smp server: combine all queue IDs to one map

* rfc: moving queue records to journal storage and queue blobs

* rfc: expiring messages in journal storage

* add comment

.github/workflows/build.yml

@@ -63,7 +63,7 @@ jobs:
           mv $(cabal list-bin xftp) xftp-ubuntu-${{ matrix.platform_name}}
 
       - name: Build changelog
-        if: startsWith(github.ref, 'refs/tags/v') && matrix.os == 'ubuntu-22.04'
+        if: startsWith(github.ref, 'refs/tags/v')
         id: build_changelog
         uses: mikepenz/release-changelog-builder-action@v1
         with:

CHANGELOG.md

@@ -1,3 +1,82 @@
+# 6.1.3
+
+SMP server: fix restoring notification credentials.
+
+# 6.1.2
+
+Servers: more reliable restoring of state.
+
+SMP server: reduced memory usage and faster start.
+
+Notifications: compensate for iOS notifications being droppted by Apple while device is offline (#1378):
+- Ntf server: send multiple SMP notifications in one iOS notification.
+- Agent: get multiple messages for one iOS notification.
+
+# 6.1.1
+
+SMP:
+- stop server faster (#1371)
+- add STORE error (#1372)
+
+# 6.1.0
+
+Version 6.1.0.7
+
+SMP server and client:
+- transport block encryption (#1317).
+
+Agent:
+- batch and optimize iOS notifications processing (#1308, #1311, #1313, #1316, #1330, #1331, #1333, #1337, #1346).
+- allow receiving multiple messages from single iOS notification (#1355, #1362).
+- prepare connection to accept to avoid race condition with events (#1365).
+- transport isolation mode "Session" (default) to use new SOCKS credentials when client restarts or SOCKS proxy configuration changes (#1321).
+
+Ntf server:
+- control port (#1354).
+- enable pings on ntf subscriptions, to resubscribe on reconnection (#1353).
+
+SMP server:
+- support multiple server ports (#1319).
+- support serving HTTPS and SMP transport on the same port (#1326, #1327).
+- persist iOS notifications to avoid losing them when Ntf server is offline (#1336, #1339, #1350).
+- fix lost notification subscriptions (#1347).
+- reject SKEY with different key earlier, at verification step (#1366).
+- pass server information via CLI during server initialization (#1356).
+- show version on server page (#1341).
+- explicit graceful shutdown on SIGINT (#1360).
+
+XRCP (remote access protocol):
+- use SHA3-256 in hybrid key agreement (#1302).
+- session encryption with forward secrecy (#1328).
+
+# 6.0.5
+
+SMP agent:
+- support generic SOCKS proxy (without isolate-by-auth).
+- reduce max message sizes
+
+# 6.0.4
+
+SMP server:
+- better performance/memory: fewer map updates on re-subscriptions (#1297), split and reduce STM transactions (#1294)
+- send DELD when subscribed queue is deleted (#1312)
+- add created/updated/used date to queues to manage expiration (#1306)
+
+XFTP server: truncate file creation time to 1 hour (#1310)
+
+Servers:
+- bind control port only to 127.0.0.1 for better security in case of firewall misconfiguration (#1280)
+- reduce memory used for period stats (#1298)
+
+Agent: process last notification from list (#1307)
+- report receive file error with redirected file ID, when redirect is present (#1304)
+- special error when deleted user record is not in database (#1303)
+- fix race when sending a message to the deleted connection (#1296)
+- support for multiple messages in a single notification
+
+Ntf server:
+- only use SOCKS proxy for servers without public address (#1314)
+
 # 6.0.3
 
 Agent:

README.md

@@ -149,8 +149,15 @@ On Linux, you can deploy smp and xftp server using Docker. This will download im
 You can install and setup servers automatically using our script:
 
 ```sh
-curl --proto '=https' --tlsv1.2 -sSf https://raw.githubusercontent.com/simplex-chat/simplexmq/stable/install.sh -o simplex-server-install.sh \
-&& if echo 'c90886104cd640b2ed64921dba80e90691db36788e8d6dcc13d8f33f92f0ea54 simplex-server-install.sh' | sha256sum -c; then chmod +x ./simplex-server-install.sh && ./simplex-server-install.sh; rm ./simplex-server-install.sh; else echo "SHA-256 checksum is incorrect!" && rm ./simplex-server-install.sh; fi
+curl --proto '=https' --tlsv1.2 -sSf https://raw.githubusercontent.com/simplex-chat/simplexmq/stable/install.sh -o simplex-server-install.sh &&\
+if echo '53fcdb4ceab324316e2c4cda7e84dbbb344f32550a65975a7895425e5a1be757 simplex-server-install.sh' | sha256sum -c; then
+  chmod +x ./simplex-server-install.sh
+  ./simplex-server-install.sh
+  rm ./simplex-server-install.sh
+else
+  echo "SHA-256 checksum is incorrect!"
+  rm ./simplex-server-install.sh
+fi
 ```
 
 ### Build from source

apps/smp-server/Main.hs

@@ -19,4 +19,4 @@ main = do
   setLogLevel LogDebug
   cfgPath <- getEnvPath "SMP_SERVER_CFG_PATH" defaultCfgPath
   logPath <- getEnvPath "SMP_SERVER_LOG_PATH" defaultLogPath
-  withGlobalLogging logCfg $ smpServerCLI_ Static.generateSite Static.serveStaticFiles cfgPath logPath
+  withGlobalLogging logCfg $ smpServerCLI_ Static.generateSite Static.serveStaticFiles Static.attachStaticFiles cfgPath logPath

apps/smp-server/static/index.html

@@ -221,6 +221,10 @@
               Public information
             </h2>
             <table id="public-info">
+              <tr class="text-grey-black dark:text-white text-base">
+                <td>Server version:</td>
+                <td>${version}</td>
+              </tr>
               <tr class="text-grey-black dark:text-white text-base">
                 <td>Source code:</td>
                 <td><a href="${sourceCode}" target="_blank">${sourceCode}</a></td>
@@ -291,6 +295,12 @@
                   <td>${hostingEntity} (${hostingCountry})</td>
                 </tr>
               </x-hosting>
+              <x-hostingType>
+                <tr class="text-grey-black dark:text-white text-base">
+                  <td>Hosting type:</td>
+                  <td>${hostingType}</td>
+                </tr>
+              </x-hostingType>
               <x-serverCountry>
                 <tr class="text-grey-black dark:text-white text-base">
                   <td>Server country:</td>

apps/smp-server/web/Static.hs

@@ -7,22 +7,30 @@ module Static where
 import Control.Logger.Simple
 import Control.Monad
 import Data.ByteString (ByteString)
-import qualified Data.ByteString as B
+import qualified Data.ByteString.Char8 as B
+import Data.Char (toUpper)
+import Data.IORef (readIORef)
 import Data.Maybe (fromMaybe)
 import Data.String (fromString)
 import Data.Text.Encoding (encodeUtf8)
-import Network.Wai.Application.Static as S
-import Network.Wai.Handler.Warp as W
-import qualified Network.Wai.Handler.WarpTLS as W
+import Network.Socket (getPeerName)
+import Network.Wai (Application)
+import qualified Network.Wai.Application.Static as S
+import qualified Network.Wai.Handler.Warp as W
+import qualified Network.Wai.Handler.Warp.Internal as WI
+import qualified Network.Wai.Handler.WarpTLS as WT
 import Simplex.Messaging.Encoding.String (strEncode)
+import Simplex.Messaging.Server (AttachHTTP)
 import Simplex.Messaging.Server.Information
 import Simplex.Messaging.Server.Main (EmbeddedWebParams (..), WebHttpsParams (..))
+import Simplex.Messaging.Transport (simplexMQVersion)
 import Simplex.Messaging.Transport.Client (TransportHost (..))
 import Simplex.Messaging.Util (tshow)
 import Static.Embedded as E
 import System.Directory (createDirectoryIfMissing)
 import System.FilePath
 import UnliftIO.Concurrent (forkFinally)
+import UnliftIO.Exception (bracket, finally)
 
 serveStaticFiles :: EmbeddedWebParams -> IO ()
 serveStaticFiles EmbeddedWebParams {webStaticPath, webHttpPort, webHttpsParams} = do
@@ -31,9 +39,44 @@ serveStaticFiles EmbeddedWebParams {webStaticPath, webHttpPort, webHttpsParams}
     W.runSettings (mkSettings port) (S.staticApp $ S.defaultFileServerSettings webStaticPath)
   forM_ webHttpsParams $ \WebHttpsParams {port, cert, key} -> flip forkFinally (\e -> logError $ "HTTPS server crashed: " <> tshow e) $ do
     logInfo $ "Serving static site on port " <> tshow port <> " (TLS)"
-    W.runTLS (W.tlsSettings cert key) (mkSettings port) (S.staticApp $ S.defaultFileServerSettings webStaticPath)
+    WT.runTLS (WT.tlsSettings cert key) (mkSettings port) app
   where
-    mkSettings port = setPort port defaultSettings
+    app = staticFiles webStaticPath
+    mkSettings port = W.setPort port warpSettings
+
+-- | Prepare context and prepare HTTP handler for TLS connections that already passed TLS.handshake and ALPN check.
+attachStaticFiles :: FilePath -> (AttachHTTP -> IO ()) -> IO ()
+attachStaticFiles path action =
+  -- Initialize global internal state for http server.
+  WI.withII warpSettings $ \ii -> do
+    action $ \socket cxt -> do
+      -- Initialize internal per-connection resources.
+      addr <- getPeerName socket
+      withConnection addr cxt $ \(conn, transport) ->
+        withTimeout ii conn $ \th ->
+          -- Run Warp connection handler to process HTTP requests for static files.
+          WI.serveConnection conn ii th addr transport warpSettings app
+  where
+    app = staticFiles path
+    -- from warp-tls
+    withConnection socket cxt = bracket (WT.attachConn socket cxt) (terminate . fst)
+    -- from warp
+    withTimeout ii conn =
+      bracket
+        (WI.registerKillThread (WI.timeoutManager ii) (WI.connClose conn))
+        WI.cancel
+    -- shared clean up
+    terminate conn = WI.connClose conn `finally` (readIORef (WI.connWriteBuffer conn) >>= WI.bufFree)
+
+warpSettings :: W.Settings
+warpSettings = W.setGracefulShutdownTimeout (Just 1) W.defaultSettings
+
+staticFiles :: FilePath -> Application
+staticFiles root = S.staticApp settings
+  where
+    settings = (S.defaultFileServerSettings root)
+      { S.ssListing = Nothing
+      }
 
 generateSite :: ServerInformation -> Maybe TransportHost -> FilePath -> IO ()
 generateSite si onionHost sitePath = do
@@ -78,6 +121,7 @@ serverInformation ServerInformation {config, information} onionHost = render E.i
       where
         basic =
           [ ("sourceCode", Just . encodeUtf8 $ sourceCode spi),
+            ("version", Just $ B.pack simplexMQVersion),
             ("website", encodeUtf8 <$> website spi)
           ]
         conds ServerConditions {conditions, amendments} =
@@ -109,7 +153,8 @@ serverInformation ServerInformation {config, information} onionHost = render E.i
             ("hostingCountry", encodeUtf8 <$> country)
           ]
         server =
-          [ ("serverCountry", fmap encodeUtf8 $ serverCountry =<< information)
+          [ ("serverCountry", encodeUtf8 <$> serverCountry spi),
+            ("hostingType",  (\s -> maybe s (\(c, rest) -> toUpper c `B.cons` rest) $ B.uncons s) . strEncode <$> hostingType spi)
           ]
 
 -- Copy-pasted from simplex-chat Simplex.Chat.Types.Preferences

cabal.project

@@ -4,6 +4,14 @@ packages: .
 -- packages: . ../http2
 -- packages: . ../network-transport
 
+-- package *
+   -- coverage: True
+   -- library-coverage: True
+
+-- package attoparsec
+   -- coverage: False
+   -- library-coverage: False
+
 index-state: 2023-12-12T00:00:00Z
 
 package cryptostore
@@ -28,3 +36,17 @@ source-repository-package
     type: git
     location: https://github.com/simplex-chat/sqlcipher-simple.git
     tag: a46bd361a19376c5211f1058908fc0ae6bf42446
+
+-- waiting for published warp-tls-3.4.7
+source-repository-package
+    type: git
+    location: https://github.com/yesodweb/wai.git
+    tag: ec5e017d896a78e787a5acea62b37a4e677dec2e
+    subdir: warp-tls
+
+-- backported fork due http-5.0
+source-repository-package
+    type: git
+    location: https://github.com/simplex-chat/wai.git
+    tag: 2f6e5aa5f05ba9140ac99e195ee647b4f7d926b0
+    subdir: warp

install.sh

@@ -2,9 +2,6 @@
 set -eu
 
 # Links to scripts/configs
-bin="https://github.com/simplex-chat/simplexmq/releases/latest/download"
-remote_version="$(curl --proto '=https' --tlsv1.2 -sSf -L https://api.github.com/repos/simplex-chat/simplexmq/releases/latest | grep -i "tag_name" | awk -F \" '{print $4}')"
-
 scripts="https://raw.githubusercontent.com/simplex-chat/simplexmq/stable/scripts/main"
 scripts_systemd_smp="$scripts/smp-server.service"
 scripts_systemd_xftp="$scripts/xftp-server.service"
@@ -57,7 +54,7 @@ ${GRN}1.${NC} Install latest binaries from GitHub releases:
 ${GRN}2.${NC} Create server directories:
     - smp: ${YLW}${path_conf_smp}${NC}
     - xftp: ${YLW}${path_conf_xftp}${NC}
-${GRN}3.${NC} Setup user for each server:
+${GRN}3.${NC} Setup user for server:
     - xmp: ${YLW}${user_smp}${NC}
     - xftp: ${YLW}${user_xftp}${NC}
 ${GRN}4.${NC} Create systemd services:
@@ -67,9 +64,8 @@ ${GRN}5.${NC} Install stopscript (systemd), update and uninstallation script:
     - all: ${YLW}${path_bin_update}${NC}, ${YLW}${path_bin_uninstall}${NC}, ${YLW}${path_bin_stopscript}${NC}
 
 Press:
-  - ${GRN}ENTER${NC} to continue installing both xftp and smp servers
-  - ${GRN}1${NC} to install only smp server
-  - ${GRN}2${NC} to install only xftp server
+  - ${GRN}1${NC} to install smp server
+  - ${GRN}2${NC} to install xftp server
   - ${RED}Ctrl+C${NC} to cancel installation
   
 Selection: "
@@ -83,6 +79,21 @@ Please checkout our server guides:
 To uninstall with full clean-up, simply run: ${YLW}sudo /usr/local/bin/simplex-servers-uninstall${NC}
 "
 
+set_version() {
+  ver="${VER:-latest}"
+
+  case "$ver" in 
+    latest)
+      bin="https://github.com/simplex-chat/simplexmq/releases/latest/download"
+      remote_version="$(curl --proto '=https' --tlsv1.2 -sSf -L https://api.github.com/repos/simplex-chat/simplexmq/releases/latest | grep -i "tag_name" | awk -F \" '{print $4}')"
+      ;;
+    *)
+      bin="https://github.com/simplex-chat/simplexmq/releases/download/${ver}"
+      remote_version="${ver}"
+      ;;
+  esac
+}
+
 os_test() {
  . /etc/os-release
 
@@ -155,6 +166,7 @@ checks() {
   exit 1
  fi
  
+ set_version
  os_test
 
  mkdir -p $path_conf_info  
@@ -166,13 +178,11 @@ main() {
  printf "%b\n%b" "${BLU}$logo${NC}" "$welcome"
  read ans
 
- if [ "$ans" = '1' ]; then
-  setup='smp'
- elif [ "$ans" = '2' ]; then
-  setup='xftp'
- else
-  setup='smp xftp'
- fi
+ case "$ans" in
+  1) setup='smp' ;;
+  2) setup='xftp' ;;
+  *) printf 'Installation aborted.\n' && exit 0 ;;
+ esac
 
  printf "Installing binaries..."
  

package.yaml

@@ -1,212 +0,0 @@
-name: simplexmq
-version: 6.0.3.0
-synopsis: SimpleXMQ message broker
-description: |
-  This package includes <./docs/Simplex-Messaging-Server.html server>,
-  <./docs/Simplex-Messaging-Client.html client> and
-  <./docs/Simplex-Messaging-Agent.html agent> for SMP protocols:
-  .
-  * <https://github.com/simplex-chat/simplexmq/blob/master/protocol/simplex-messaging.md SMP protocol>
-  * <https://github.com/simplex-chat/simplexmq/blob/master/protocol/agent-protocol.md SMP agent protocol>
-  .
-  See <https://github.com/simplex-chat/simplex-chat terminal chat prototype> built with SimpleXMQ broker.
-
-homepage: https://github.com/simplex-chat/simplexmq#readme
-license: AGPL-3
-author: simplex.chat
-maintainer: chat@simplex.chat
-copyright: 2020-2022 simplex.chat
-category: Chat, Network, Web, System, Cryptography
-extra-source-files:
-  - README.md
-  - CHANGELOG.md
-  - cbits/sha512.h
-  - cbits/sntrup761.h
-  - apps/smp-server/static/*.html
-  - apps/smp-server/static/media/*
-
-dependencies:
-  - aeson == 2.2.*
-  - ansi-terminal >= 0.10 && < 0.12
-  - asn1-encoding == 0.9.*
-  - asn1-types == 0.3.*
-  - async == 2.2.*
-  - attoparsec == 0.14.*
-  - base >= 4.14 && < 5
-  - base64-bytestring >= 1.0 && < 1.3
-  - case-insensitive == 1.2.*
-  - composition == 1.0.*
-  - constraints >= 0.12 && < 0.14
-  - containers == 0.6.*
-  - crypton == 0.34.*
-  - crypton-x509 == 1.7.*
-  - crypton-x509-store == 1.6.*
-  - crypton-x509-validation == 1.6.*
-  - cryptostore == 0.3.*
-  - data-default == 0.7.*
-  - direct-sqlcipher == 2.3.*
-  - directory == 1.3.*
-  - filepath == 1.4.*
-  - hashable == 1.4.*
-  - hourglass == 0.2.*
-  - http-types == 0.12.*
-  - http2 >= 4.2.2 && < 4.3
-  - ini == 0.4.1
-  - iproute == 1.7.*
-  - iso8601-time == 0.1.*
-  - memory == 0.18.*
-  - mtl >= 2.3.1 && < 3.0
-  - network >= 3.1.2.7 && < 3.2
-  - network-info >= 0.2 && < 0.3
-  - network-transport == 0.5.6
-  - network-udp >= 0.0 && < 0.1
-  - optparse-applicative >= 0.15 && < 0.17
-  - process == 1.6.*
-  - random >= 1.1 && < 1.3
-  - simple-logger == 0.1.*
-  - socks == 0.6.*
-  - sqlcipher-simple == 0.4.*
-  - stm == 2.5.*
-  - temporary == 1.3.*
-  - time == 1.12.*
-  - time-manager == 0.0.*
-  - tls >= 1.9.0 && < 1.10
-  - transformers == 0.6.*
-  - unliftio == 0.2.*
-  - unliftio-core == 0.2.*
-  - websockets == 0.12.*
-  - yaml == 0.11.*
-  - zstd == 0.1.3.*
-
-flags:
-  swift:
-    description: Enable swift JSON format
-    manual: True
-    default: False
-  use_crypton:
-    description: Use crypton etc. in cryptostore
-    manual: True
-    default: True
-
-# cpp-options:
-#   - -Dslow_servers
-
-when:
-  - condition: flag(swift)
-    cpp-options:
-      - -DswiftJSON
-  - condition: impl(ghc >= 9.6.2)
-    dependencies:
-      - bytestring == 0.11.*
-      - template-haskell == 2.20.*
-      - text >= 2.0.1 && < 2.2
-  - condition: impl(ghc < 9.6.2)
-    dependencies:
-      - bytestring == 0.10.*
-      - template-haskell == 2.16.*
-      - text >= 1.2.3.0 && < 1.3
-
-library:
-  source-dirs: src
-  c-sources:
-    - cbits/sha512.c
-    - cbits/sntrup761.c
-  include-dirs: cbits
-  extra-libraries: crypto
-
-executables:
-  smp-server:
-    source-dirs:
-    - apps/smp-server
-    - apps/smp-server/web
-    main: Main.hs
-    dependencies:
-      - file-embed
-      - simplexmq
-      - wai-app-static
-      - warp
-      - warp-tls
-    ghc-options:
-      - -threaded
-      - -rtsopts
-
-  ntf-server:
-    source-dirs: apps/ntf-server
-    main: Main.hs
-    dependencies:
-      - simplexmq
-    ghc-options:
-      - -threaded
-      - -rtsopts
-
-  xftp-server:
-    source-dirs: apps/xftp-server
-    main: Main.hs
-    dependencies:
-      - simplexmq
-    ghc-options:
-      - -threaded
-      - -rtsopts
-
-  xftp:
-    source-dirs: apps/xftp
-    main: Main.hs
-    dependencies:
-      - simplexmq
-    ghc-options:
-      - -threaded
-      - -rtsopts
-
-tests:
-  simplexmq-test:
-    source-dirs: tests
-    main: Test.hs
-    dependencies:
-      - simplexmq
-      - deepseq == 1.4.*
-      - generic-random == 1.5.*
-      - hspec == 2.11.*
-      - hspec-core == 2.11.*
-      - HUnit == 1.6.*
-      - QuickCheck == 2.14.*
-      - silently == 1.2.*
-      - main-tester == 0.2.*
-      - timeit == 2.0.*
-    ghc-options:
-      - -threaded
-      - -rtsopts
-      - -with-rtsopts=-A64M
-      - -with-rtsopts=-N1
-
-ghc-options:
-  # - -haddock
-  - -Weverything
-  - -Wno-missing-exported-signatures
-  - -Wno-missing-import-lists
-  - -Wno-missed-specialisations
-  - -Wno-all-missed-specialisations
-  - -Wno-unsafe
-  - -Wno-safe
-  - -Wno-missing-local-signatures
-  - -Wno-missing-kind-signatures
-  - -Wno-missing-deriving-strategies
-  - -Wno-monomorphism-restriction
-  - -Wno-prepositive-qualified-module
-  - -Wno-unused-packages
-  - -Wno-implicit-prelude
-  - -Wno-missing-safe-haskell-mode
-  - -Wno-missing-export-lists
-  - -Wno-partial-fields
-  - -Wcompat
-  - -Werror=incomplete-record-updates
-  - -Werror=incomplete-patterns
-  - -Werror=incomplete-uni-patterns
-  - -Werror=missing-methods
-  - -Werror=tabs
-  - -Wredundant-constraints
-  - -Wincomplete-record-updates
-  - -Wunused-type-patterns
-  - -O2
-
-default-extensions:
-  - StrictData

protocol/xftp.md

@@ -559,11 +559,13 @@ In current implementation of XFTP protocol in SimpleX Chat clients don't use FAC
 
  - perform traffic correlation attacks against senders and recipients and correlate senders and recipients within the monitored set, frustrated by the number of users on the servers.
 
- - observe how much traffic is being sent, and make guesses as to its purpose
+ - observe how much traffic is being sent, and make guesses as to its purpose.
 
-*cannot, even in case of a compromised transport protocol:*
+ - in case of a compromised transport protocol, correlate file senders and receivers.
 
- - perform traffic correlation attacks with any increase in efficiency over a non-compromised transport protocol
+*cannot, in case of a non-compromised transport protocol:*
+
+ - perform traffic correlation attacks.
 
 #### XFTP server
 

protocol/xrcp.md

@@ -67,7 +67,7 @@ The session invitation contains this data:
 - CA TLS certificate fingerprint of the controller - this is part of long term identity of the controller established during the first session, and repeated in the subsequent session announcements.
 - Session Ed25519 public key used to verify the announcement and commands - this mitigates the compromise of the long term signature key, as the controller will have to sign each command with this key first.
 - Long-term Ed25519 public key used to verify the announcement and commands - this is part of the long term controller identity.
-- Session X25519 DH key and SNTRUP761 KEM encapsulation key to agree session encryption (both for multicast announcement and for commands and responses in TLS), as described in https://datatracker.ietf.org/doc/draft-josefsson-ntruprime-hybrid/. The new keys are used for each session, and if client key is already available (from the previous session), the computed shared secret will be used to encrypt the announcement multicast packet. The out-of-band invitation is unencrypted. DH public key and KEM encapsulation key are sent unencrypted. NaCL crypto_box is used for encryption.
+- Session X25519 DH key to agree session encryption (both for multicast announcement and for commands and responses in TLS), as described in https://datatracker.ietf.org/doc/draft-josefsson-ntruprime-hybrid/. The new keys are used for each session, and if client key is already available (from the previous session), the computed shared secret will be used to encrypt the announcement multicast packet. The out-of-band invitation is unencrypted. DH public key and KEM encapsulation key are sent unencrypted. NaCL crypto_box is used for encryption.
 
 Host application decrypts (except the first session) and validates the invitation:
 - Session signature is valid.
@@ -184,7 +184,7 @@ The controller decrypts (including the first session) and validates the received
 The controller should reply with with `ctrlHello` or `ctrlError` response:
 
 ```abnf
-ctrlHello = %s"HELLO " kemCiphertext nonce encrypted(unpaddedSize ctrlHelloJSON helloPad) pad
+ctrlHello = %s"HELLO " kemCiphertext encrypted(unpaddedSize ctrlHelloJSON helloPad) pad
 ; ctrlHelloJSON is encrypted with the hybrid secret,
 ; including both previously agreed DH secret and KEM secret from kemCiphertext
 unpaddedSize = largeLength
@@ -206,6 +206,8 @@ JTD schema for the encrypted part of controller HELLO block `ctrlHelloJSON`:
 }
 ```
 
+Controller `hello` block and all subsequent protocol messages are encrypted with the chain keys derived from the hybrid key (see key exchange below) - that is why conntroller hello block does not include nonce. That provides forward secrecy within the XRCP session. Receiving this `hello` block allows host to compute the same hybrid keys and to derive the same chain keys.
+
 Once the controller replies HELLO to the valid host HELLO block, it should stop accepting new TCP connections.
 
 ### Controller/host session operation
@@ -223,10 +225,12 @@ tlsunique channel binding from TLS session MUST be included in commands (include
 The syntax for encrypted command and response body encoding:
 
 ```abnf
-commandBody = encBody sessSignature idSignature [attachment]
-responseBody = encBody [attachment] ; counter must match command
-encBody = nonce encLength32 encrypted(tlsunique counter body)
-attachment = %x01 nonce encLength32 encrypted(attachment)
+commandBody = counter encBody sessSignature idSignature [attachment]
+responseBody = counter encBody [attachment] ; counter must match command
+; counter is placed outside of encrypted body to allow correlating encryption keys
+; with the chain keys (each command and response are encrypted by different keys)
+encBody = encLength32 encrypted(tlsunique body)
+attachment = %x01 encLength32 encrypted(attachment)
 noAttachment = %x00
 tlsunique = length 1*OCTET
 counter = 8*8 OCTET ; int64
@@ -239,7 +243,7 @@ If the command or response includes attachment, its hash must be included in com
 
 Initial announcement is shared out-of-band (URI with xrcp scheme), and it is not encrypted.
 
-This announcement contains only DH keys, as KEM key is too large to include in QR code, which are used to agree encryption key for host HELLO block. The host HELLO block will contain DH key in plaintext part and KEM encapsulation (public) key in encrypted part, that will be used to determine the shared secret (using SHA256 over concatenated DH shared secret and KEM encapsulated secret) both for controller HELLO response (that contains KEM ciphertext in plaintext part) and subsequent session commands and responses.
+This announcement contains only DH keys, as KEM key is too large to include in QR code, which are used to agree encryption key for host HELLO block. The host HELLO block will contain DH key in plaintext part and KEM encapsulation (public) key in encrypted part, that will be used to determine the shared secret (using SHA3-256 over concatenated DH shared secret and KEM encapsulated secret) to derive keys for controller HELLO response (that contains KEM ciphertext in plaintext part) and subsequent session commands and responses.
 
 During the next session the announcement is sent via encrypted multicast block. The shared key for this announcement and for host HELLO block is determined using the KEM shared secret from the previous session and DH shared secret computed using the host DH key from the previous session and the new controller DH key from the announcement.
 
@@ -273,6 +277,22 @@ If controller fails to store the new host DH key after receiving HELLO block, th
 
 To decrypt a multicast announcement, the host should try to decrypt it using the keys of all known (paired) remote controllers.
 
+Once kemSecret is agreed for the session, it is used to derive two chain keys, to receive and to send messages:
+
+```
+host: sndKey, rcvKey = HKDF(kemSecret, "SimpleXSbChainInit", 64)
+controller: rcvKey, sndKey  = HKDF(kemSecret, "SimpleXSbChainInit", 64)
+```
+
+where HKDF is based on SHA512, with empty salt.
+
+Actual keys and nonces to encrypt and decrypt messages are derived from these chain keys:
+
+```
+to send: (sndKey', sk, nonce) = HKDF(sndKey, "SimpleXSbChain", 88)
+to receive: (rcvKey', sk, nonce) = HKDF(rcvKey, "SimpleXSbChain", 88)
+```
+
 ## Threat model
 
 #### A passive network adversary able to monitor the site-local traffic:

rfcs/2024-06-21-short-links.md

@@ -0,0 +1,124 @@
+# Short invitation links
+
+## Problem
+
+Long links look scary and unsafe for many users. While this is a perceived problem, rather than a real one, it hurts adoption.
+
+What is worse, long links do not fit in profile descriptions of other social networks where people might want to advertize their contact addresses.
+
+The current link size limitation is also the reason for not including PQ KEM keys into invitation links and addresses, postponing the moment when PQ-resistant encryption kicks in - if we include PQ KEM key into the link, the QR code will not be scannable.
+
+Additionally, if we store short links, they can also include chat preferences and public profile data.
+
+## Solution
+
+MITM-resistant link shortening.
+
+Instead of generating the random address that would resolve into the link - doing so would create the possibility of MITM by the server hosting this link - we can use private key as the link ID that will be passed to the accepting party, and the hash of the public key as ID for the server - the accepting party would present this key itself as ID and it will also be used for server to client encryption (see Protocol below). HKDF will be used to derive symmetric key from private key and used in secret_box together with random nonce (to allow replacing data with the same key but with a different nonce - nonce will be sent to the server too). secret_box construction is authenticated encryption, so it would protect from MITM.
+
+The proposed syntax:
+
+```abnf
+shortConnectionRequest = connectionScheme "/" connReqType "#/" smpServer "/" linkHash
+connReqType = %s"invitation" / %s"contact"
+connectionScheme = (%s"https://" clientAppServer) / %s"simplex:"
+clientAppServer = hostname [ ":" port ]
+  ; client app server, e.g. simplex.chat
+smpServer = serverIdentity "@" srvHosts [":" port] ; no smp:// prefix, no escaping
+srvHosts = <hostname> ["," srvHosts] ; RFC1123, RFC5891
+linkHash = <base64url encoded SHA256 or SHA512 hash of the original link>
+```
+
+If SMP server supports pages, its name can be used as clientAppServer, without repeating it after #, for a shorter link.
+
+Example link:
+
+```
+https://simplex.chat/contact/#0YuTwO05YJWS8rkjn9eLJDjQhFKvIYd8d4xG8X1blIU=@smp8.simplex.im/abcdefghij0123456789abcdefghij0123456789abc=
+```
+
+This link has the length of ~136 characters (256 bits), which is shorter than the full contact address (~310 characters) and much shorter than invitation links (~528 characters) even without post-quantum keys added to them.
+
+This size can be further reduced by
+- use server domain in the link.
+- do not include onion address, as the connection happens via proxy anyway, if it's untrusted server.
+- not pinning server TLS certificate - the downside here is that while the attack that compromises TLS will not be able to substitute the link (because it's hash will not match), it will be able to intercept and to block it.
+- using shorter hash, e.g. SHA128 - reducing the collision resistance.
+
+If the server is known, the client could use it's hash and onion address, otherwise it could trust the proxy to use any existing session with the same hostname or to accept the risk of interception - given that there is no risk of substitution.
+
+With the first two of these "improvements" the link could be ~122 characters:
+
+```
+https://smp8.simplex.im/contact/#0YuTwO05YJWS8rkjn9eLJDjQhFKvIYd8d4xG8X1blIU@/abcdefghij0123456789abcdefghij0123456789abc
+```
+
+If onion address is preserved the link will be ~184 characters (won't fit in Twitter 160 characters bio):
+
+```
+https://smp8.simplex.im/contact/#0YuTwO05YJWS8rkjn9eLJDjQhFKvIYd8d4xG8X1blIU@beccx4yfxxbvyhqypaavemqurytl6hozr47wfc7uuecacjqdvwpw2xid.onion/abcdefghij0123456789abcdefghij0123456789abc
+```
+
+If we implement it, the request to resolve the link would be made via proxied SMP command (to avoid the direct connection between the client and the recipient's server).
+
+Pros:
+- a bit shorter link.
+- possibility to include post-quantum keys into the full link keeping the same shortened link size.
+- possibility to include chat profile of contact or group, and preferences, for a much better connection experience, and to show this information when the link sent in the conversation (clients can resolve them automatically, without connecting - it can be resolved by the sending clients).
+- server will not have access to the link.
+
+Cons:
+- protocol complexity.
+- observers can access the link content, so for 1-time invitation we should only include permissions and not profile.
+
+Pros are a huge improvement of UX of connecting both within and from outside of the app (e.g., link can be resolved even before creating chat profile, as part of the onboarding).
+
+## Protocol
+
+To support short links, the SMP servers would provide a simple key-value store enabled by three additional commands: `WRT`, `CLR` and `READ`
+
+`WRT` command is used to store and to update values in the store. The size of the value is limited by the same size as sent messages (or, possibly, smaller - as connection information size used in confirmation messages) - the clients would use this fixed size irrespective of the content. `WRT` command will be sent with the data blob ID in the transaction entityId field, public authorization key used to authorize `WRT` and `CLR` commands (subsequent WRT commands to the existing key must use the same key), and the data blob.
+
+`CLR` command must use with the same entity ID and must be authorized by the same key.
+
+`READ` command must use the ID which hash would be equal of the ID used to create the data blob, and this ID would also be used as public authorization
+
+## Algorithm to store and to retrieve data blob.
+
+**Store data blob**
+
+- the data blob owner generates X25519 key pair: `(k, pk)`.
+- private key `pk` will be included in the short link shared with the other party (only base64url encoded key bytes, not X509 encoding).
+- `HKDF(pk)` will be used to encrypt the link data with secret_box before storing it on the server.
+- the hash of public key `sha256(k)` will be used as ID by the owner to store and to remove the data blob (`WRT` and `CLR` commands).
+
+**Retrieve data blob**
+
+- the sender uses the public key `k` derived from the private key `pk` included in the link as entity ID to retrieve data blob (the server will compute the ID used by the owner as `sha256(k)` and will be able to look it up). This provides the quality that the traffic of the parties has no shared IDs inside TLS. It also means that unlike message queue creation, the ID to retrieve the blob was never sent to the blob creator, and also is not known to the server in advance (the second part is only an observation, in itself it does not increase security, as server has access to an encrypted blob anyway).
+- note that the sender does not authorize the request to retrieve the blob, as it would not increase security unless a different key is used to authorize, and adding a key would increase link size.
+- server session keys with the sender will be `(sk, spk)`, where `sk` is public key shared with the sender during session handshake, and `spk` is the private key known only to the server.
+- this public key `k` will also be combined with server session key `spk` using `dh(k, spk)` to encrypt the response, so that there is no ciphertext in common in sent and received traffic for these blobs. Correlation ID will be used as a nonce for this encryption.
+- having received the blob, the client can now decrypt it using secret_box with `HKDF(pk)`.
+
+Using the same key as ID for the request, and also to additionally encrypt the response allows to use a single key in the link, without increasing the link size.
+
+## Threat model
+
+**Compromised SMP server**
+
+can:
+- delete link data.
+- hide link selectively from some requests.
+
+cannot:
+- undetectably replace link data.
+- access unencrypted link data, whether it was or was not accessed by the accepting party.
+- observe IP addresses of the users accessing link data.
+
+**Passive observer who observed short link**:
+
+can:
+- access original unencrypted link data
+
+cannot:
+- replace or delete the link data

rfcs/2024-09-01-smp-message-storage.md

@@ -0,0 +1,179 @@
+# SMP server message storage
+
+## Problem
+
+Currently SMP servers store all queues in server memory. As the traffic grows, so does the number of undelivered messages. What is worse, Haskell is not avoiding heap fragmentation when messages are allocated and then de-allocated - undelivered messages use ByteString and GC cannot move them around, as they use pinned memory.
+
+## Possible solutions
+
+### Solution 1: solve only GC fragmentation problem
+
+Move from ByteString to some other primitive to store messages in memory long term, e.g. ShortByteString, or manage allocation/de-allocation of stored messages manually in some other way.
+
+Pros: the simplest solution that avoids substantial re-engineering of the server.
+
+Cons:
+- not a long term solution, as memory growth still has limits.
+- may be ineffective, as it introduces additional copying of bytes.
+
+### Solution 2: move message storage to hard drive
+
+Use files or RocksDB to store messages.
+
+Pros:
+- much lower memory usage.
+- no message loss in case of abnormal server termination (important until clients have delivery redundancy).
+- this is a long term solution, and at some point it might need to be done anyway.
+
+Cons:
+- substantial re-engineering costs and risks.
+- metadata privacy. Currently we only save undelivered messages when server is restarted, with this approach all messages will be stored for some time. this argument is limited, as hosting providers of VMs can make memory snapshots too, on the other hand they are harder to analyze than files. On another hand, with this approach messages will be stored for a shorter time.
+
+#### RocksDB and other key-value stores
+
+The downside of any key-value stores is that they don't seem to have efficient primitives for sequential delivery. While sequential delivery can be modelled with linked lists, they would require 1 key insert (on send), 3 key updates (1 update to update queue data on send, 1 update of the last message to point to the next, 1 update on delivery or message expiration) and 1 key deletion (on delivery or message expiration) for each delivered message.
+
+This might result in substantial write amplification and compacting costs.
+
+In general, tree structures that are efficient for quick lookups and updates, given approximately fixed value size, are inefficient for modelling queues.
+
+#### Files
+
+The upside of files is that they are well suited for sequential delivery and don't result in the same churn, with careful design, as trees do.
+
+The downside of filesystem is that it does not scale well with the large number of files in a folder, so queues will have to be spread across multiple folders, following tree-like structure.
+
+I could not find an available library that efficiently models sequential delivery in highly concurrent environment.
+
+A possible design could be the following.
+
+##### Queue folder and files
+
+Each message queue is stored in its own folder (see below on folder locations). Folder would contain these files:
+
+- `messages.abcd.log` - the file that is used to read messages from, sequentially
+- `messages.efgh.log` - the optional file that is used to write messages, in case it is different from read file.
+- `queue.log` - append-only file where the last line represents the current queue state
+- `queue.timestamp.log` - previous states of queue.log file
+
+Each line in "queue.log" file has this syntax
+
+```abnf
+queueLogLine =
+    %s"read_file=" base64
+    %s"read_msg=" digits
+    %s"read_byte=" digits
+    %s"write_file=" base64
+    %s"write_msg=" digits
+```
+
+When queue is first requested by the server:
+
+```c
+if queue folder exists:
+    read queue state from last line of queue.log
+    if queue.log contained more than one line: // compaction
+        copy queue.log to queue.timestamp.log
+        write one line queue state to queue.log
+else:
+    create queue folder
+    create messages.abcd.log (abcd is some random string)
+    read_msg = 0
+    read_byte = 0
+    create queue.log with one line: "read_file=abcd read_msg=0 read_byte=0 write_files=abcd write_msg=0"
+open read_file in ReadMode and seek to read_byte position
+nextReadByte = read_byte
+nextReadMsg = read_msg
+open write_file in AppendMode
+```
+
+When message is added to the queue (assumes that queue state is loaded to server memory, if not the previous section will be done first):
+
+```c
+if write_msg > max_queue_messages:
+    return quota error
+else if write_msg = max_queue_messages:
+    add quota_exceeded message to write_file
+    update queue state: write_msg += 1
+    append updated queue state to queue.log
+else
+    // It is required that `max_queue_messages < max_file_messages`,
+    // so that we never need more than one additional write file.
+    if write_msg >= max_file_messages: // queue file rotation
+        create messages.efgh.log // efgh is some random string
+        update queue state: write_file=efgh write_msg=0 // read file remains the same as it was
+        append updated queue state to queue.log
+        copy queue.log to queue.timestamp.log
+        // `old` needs to be defined to limit the number and storage duration,
+        // preserving not more than N files, and not more than M days files, "and then some"
+        // (that is if the queue has high churn, we have file from M days before in any case, for any debugging).
+        delete `old` `queue.timestamp.log` files
+        write one line queue state to queue.log // compaction
+
+    add message to write_file
+    update queue state: write_msg += 1
+    append updated queue state to queue.log
+```
+
+The above algorithm assumes `max_queue_messages < than max_file_messages`, so that we never need more than one write file.
+
+When message is delivered, it is simply read from the read queue, queue state does not change yet:
+
+```c
+if nextReadMsg > read_msg:
+    deliver cached message, no need to read it again
+else
+    read message from read_file handle
+    nextReadMsg = read_msg + 1
+    nextReadByte = current position in file
+```
+
+When message delivery is acknowledged, the read queue needs to be advanced, and possibly switched to read from the current write_queue:
+
+```c
+if nextReadByte == read_byte:
+    return error // nothing was delivered
+else if nextReadByte = EOF:
+    // end of file is reached, possibly some other condition,
+    // but it should allow changing max_file_messages on server restart
+    currReadFile = read_file
+    read_file = write_file
+    read_msg = 0
+    read_byte = 0
+    append updated queue state to queue.log
+    delete currReadFile
+else
+    read_msg += 1
+    read_byte = nextReadByte
+    // `seek` should not be necessary, as the handle is already at nextReadByte position here
+    // seek to read_byte
+    append updated queue state to queue.log
+```
+
+The above algorithm delegates the problem of compaction and fragmentation management to file system, that is very optimized for such scenarios.
+
+Also, read and write files will grow to almost a constant size, so the space they used may be re-used.
+
+An important consideration is that writes to queue.log and message.log files and queue state modifications have to be sequential, without concurrency - it can be managed with the usual locks.
+
+##### Queue folders structure
+
+Most Linux systems use EXT4 filesystem where the file lookup time scales linearly to the number of files. While alternatives with logarithmic lookup time exist (XFS), they may be very complex to configure on the existing systems.
+
+So storing all queue folders in one folder won't scale.
+
+To solve this problem we could use recipient queue ID in base64url format not as a folder name, but as a folder path, splitting it to path fragments of some length. The number of fragments can be configurable and migration to a different fragment size can be supported as the number of queues on a given server grows.
+
+Currently, queue ID is 24 bytes random number, thus allowing 2^192 possible queue IDs. If we assume that a server must hold 1b queues, it means that we have ~2^162 possible addresses for each existing queue. 24 bytes in base64 is 32 characters that can be split into say 8 fragments with 4 characters each, so that queue folder path for queue with ID `abcdefghijklmnopqrstuvwxyz012345` would be:
+
+`/var/opt/simplex/messages/abcd/efgh/ijkl/mnop/qrst/uvwx/yz01/2345`
+
+The maximum theoretic number of the folders on the 1st level is 64^4, or 2^24 ~ 16m - this is probably still a large number of subfolders for EXT4. Given that addresses are random, all the possible combinations in the first folder can be used with a large number of queues.
+
+So we could use an unequal split of path, two letters each and the last being long:
+
+`/var/opt/simplex/messages/ab/cd/ef/ghijklmnopqrstuvwxyz012345`
+
+The first three levels in this case can have 4096 subfolders each, and it gives 68b possible subfolders (64^2^3), so the last level will be sparse in case of 1b queues on the server. So we could make it 4 levels with 2 letters to never think about it, accounting for a large variance of the random numbers distribution:
+
+`/var/opt/simplex/messages/ab/cd/ef/gh/ijklmnopqrstuvwxyz012345`

rfcs/2024-09-05-queue-storage.md

@@ -0,0 +1,81 @@
+# Storage considerations for SMP queues
+
+See [Short invitation links](./2024-06-21-short-links.md).
+
+## Problem
+
+1) queue records are created permanently, until the clients delete them.
+
+2) clients only delete queue records based on some user action, pending connections do not expire.
+
+While part 2 should be improved in the client, indefinite storage of queue records becomes a much bigger issue if each of them would result in a permanent storage of 4-16kb blob in server memory, without server-side expiration for short invitation links.
+
+## Possible solutions
+
+1) Add some queue timestamp, e.g. queue creation date, to expire unsecured queues after say 3 weeks.
+
+The problem with this approach is that contact addresses are also unsecured queues, and they should not be expired.
+
+We could set really large expiration time, and require that clients "update" the unsecured queues they need at least every 1-2 years, but it would not solve the problem of storing a large number of blobs in the server memory for unused/abandoned 1-time invitations.
+
+2) Do not store blobs in memory / append-only log, and instead use something like RocksDB. While it may be a correct long term solution, it may be not expedient enough at the current POC stage for this feature. Also, the lack of expiration is wrong in any case and would indefinitely grow server storage.
+
+3) Add flag allowing the server to differentiate permanent queues used as contact addresses, also using different blob sizes for them. In this case, messaging queues will be expired if not secured after 3 weeks, and contact address queues would be expired if not "updated" by the owner within 2 years.
+
+Probably all three solutions need to be used, to avoid creating a non-expiring blob storage in memory, as in case too many of such blobs are created it would not be possible to differentiate between real users and resource exhaustion attacks, and unlike with messages, they won't be expiring too.
+
+Servers already can differentiate messaging queues and contact address queues, if they want to:
+- with the old 4-message handshake, the confirmation message on a normal queue was different, and also KEY command was eventually used.
+- with the fast 2-message handshake, while the confirmation message has the same syntax, and the differences are inside encrypted envelope, the client still uses SKEY command.
+- in both cases, the usual messaging queues are secured, and contact addresses are not, so this difference is visible in the storage as well (although it is not easy to differentiate between abandoned 1-time invitations and contact addresses).
+
+Differentiating these queues can also allow different message retention times - e.g., the queues for contact addresses could have bigger size, but have lower message retention time.
+
+## Proposed solution
+
+1. Add queue updated_at date into queue records. While it adds some metadata, it seems necessary to manage retention and quality of service. It will not include exact time, only date, and the time of creation will be replaced by the time of any update - queue secured, a message is sent, or queue owner subscribes to the queue. To avoid the need to update store log on every message this information can be appended to store log on server termination. Or given that only one update per day is needed it may be ok to make these updates as they happen (temporarily making the sequence and time of these events available in storage).
+
+2. Add flag to indicate the queue usage - messaging queue or queue for contact address connection requests. This would result in different queue size and different retention policy for queue and its messages. We already have "sender can secure flag" which is, effectively, this flag - contact address queues are never secured. So this does not increase stored metadata in any way.
+
+## Possible changes to short links
+
+This is a design considerations and a concept, not a design yet.
+
+Instead of implementing a generic blob storage that can be used as an attack vector, and adds additional failure point (another server storing blob that is necessary to connect to the queue on the current server), but instead adds an extended queue information blobs, most of which could be dropped without the loss of connectivity, so that the attack can be mitigated by deleting these blobs without users losing the ability to connect, as long as the queue and minimal extended information is retained.
+
+So, to make the connection there need to be these elements:
+
+- queue server and queue ID - mandatory part, that can be included in short link
+- SMP key - mandatory part for all queues. We are considering initializing ratchets earlier for contact addresses, and include ratchet keys and pre-keys into queue data as well, but it is out of scope here.
+- Ratchet keys - mandatory part for 1-time invitation that won't fit in short link.
+- PQ key - optional part that can be stored with addresses if ratchet keys are added and with 1-time invitations.
+- App blobs - chat preferences for 1-time invitation links and profile information for contact addresses.
+
+So rather that storing one blob with a large address inside it, not associated with the queue, increasing probability of failure and reducing our ability to mitigate resource exhaustion, we could store extended blobs associated with the queues.
+
+Also, we need the address shared with the sender (party accepting the connection) to be short. We could use a similar approach that was proposed for data blobs, using a single random seed per queues to derive multiple keys and IDs from it. For example:
+
+1. The queue owner:
+  - generates Ed25529 key pair `(sk, spk)` and X25519 key pair `(dhk, dhpk)` to use with the server, same as now sent in NEW command.
+  - generates queue recipient ID (this ID can still be server-generated).
+  - generates X25519 key pair `(k, pk)` to use with the accepting party.
+  - derives from `k`:
+    - sender ID.
+    - symmetric key for authenticated encryption of blobs.
+  - `k` will be used as short link.
+2. All other data from the invitation can be included in queue creation request and be associated with the queue as 1-3 blobs with different priority:
+  - ratchet keys - it will have a small size, so only this blob cannot be removed, while other blobs can be removed in case of resource exhaustion.
+  - PQ keys - optional blob.
+  - conversation preferences and profile - can be removed depending on creation time, e.g. all new blobs can be removed.
+
+The algorithm used to derive key and ID from `k` needs to be cryptographically secure, e.g. it could be some KDF or ChaCha DRG initialized with `k` as seed, TBC.
+
+So, coupling blob storage with messaging queues has these pros/cons:
+
+Cons:
+- no additional layer of privacy - the server used for connection is visible in the link, even after the blobs are removed from the server.
+
+Pros:
+- no additional point of failure in the connection process - the same server will be used to retrieve necessary blobs as for connection.
+- queue blobs of messaging blobs will be automatically removed once the queue is secured or expired, without additional request from the recipient - reducing the storage and the time these blobs are available.
+- queue blobs for contact addresses will be structured and some of the large blobs can be removed in case of resource exhaustion attack (and recreated by the client if needed), with the only downside that PQ handshake will be postponed (which is the case now) and profile will not be available at a point of connection.

rfcs/2024-09-09-smp-blobs.md

@@ -0,0 +1,80 @@
+# Blob extensions for SMP queues
+
+Evolution of the design for short links, see [here](./2024-06-21-short-links.md) and [here](./2024-09-05-queue-storage.md).
+
+## Problems
+
+Allow storing extended information with SMP queues to improve UX and security of making connections:
+- short invitation links and contact addresses.
+- PQ encryption from the first message.
+- present user profile with chat preferences and welcome message when the public address link is scanned.
+
+## Design
+
+1. Queue creation/update date is already added to server persistence, allowing to expire queues and blobs, depending on their usage.
+2. Add "queue type" metadata to NEW command to indicate whether messaging queue is used as public address or as messaging queue (see previous docs on why it doesn't change threat model). While at the moment it would match sndSecure flag there may be future scenarios when they diverge. Initially only "invitation" and "contact" types will be supported.
+3. Prohibit sndSecure flag for "contact" queues, prohibit securing contact queues.
+4. Add "queue blobs" to NEW command:
+  - blob0: ratchetKeys up to N0 bytes - priority 0, can't be removed by the server, only in "invitation"
+  - blob1: PQ key up to N1 bytes - priority 1, can be removed by the server, only used in "invitation"
+  - blob2: Application data up to N2 bytes - priority 2, can be removed by the server.
+5. Add linkId to NEW command
+6. linkId and blobs will be removed when queue is secured.
+7. Add recipient command to remove/upsert blob2 for contact queues.
+8. Add sender command to retrieve blobs.
+
+## Protocol
+
+### Creating a queue:
+
+The queue owner:
+- generates Ed25529 key pair `(sk, spk)` and X25519 key pair `(dhk, dhpk)` to use with the server, same as now. `sk` and `dhk` will be sent in NEW command.
+- generates X25519 key pair `(k, pk)` to use with the accepting party to encrypt queue messages.
+- derives from `k` using HKDF:
+- symmetric key `bk` for authenticated encryption of blobs.
+- `linkId`, will be sent in NEW command.
+- `k` will be used as short link.
+- sends NEW command.
+
+NEW command syntax:
+
+```abnf
+create = %s"NEW " linkId queueType recipientAuthPublicKey recipientDhPublicKey
+                  basicAuth subscribe sndSecure [ "0" blob0 ] [ "1" blob1 ] [ "2" blob2 ] 
+queueType = %s"I" / %s "C" ; new parameter
+linkId = length *OCTET ; new parameter,
+                        ; can be empty in which case blobs won't be allowed
+blob0 = word16 *OCTET ; new parameter, encrypted ratchet keys,
+                      ; including nonce and auth tag
+blob1 = word16 *OCTET ; new parameter, encrypted PQ key
+blob2 = word16 *OCTET ; new parameter, encrypted application data
+```
+
+SET - command to update queue blobs (recipientId is used as entity ID):
+
+```abnf
+set = %s"SET " linkId [ "2" blob2 ] ; passing empty blob removes it
+linkId ; updated (or the same) linkId, can be empty to remove blobs
+        ; allows to change the address without removing the queue / changing blobs
+        ; (e.g., to avoid losing the messages).
+```
+
+### Sending messages to the queue
+
+GET - command to get queue blobs (linkId is used as entity ID):
+
+```abnf
+get = %s"GET"
+```
+
+Response to GET:
+
+```abnf
+blobs = %s"BLOB" senderId [ "0" blob0 ] [ "1" blob1 ] [ "2" blob2 ]
+```
+
+As blobs are retrieved using a separate linkId, once blobs are removed it will be impossible to find senderId from short link - it is a threat model improvement. Once server storage is compacted, it will be impossible to find queue related to the link even with the access to server data (unless server preserves the data).
+
+### Possible privacy improvement
+
+We could only allow unauthorized GET and authorized SET commands for long-term "contact" queues, and return BLOB in response to SKEY (or require that GET is authorized) - so that only the person who secures the queue will get access to data blobs. This way it ensures that the parties transmitting the invitation links cannot retrieve their content without the sender noticing it.

rfcs/2024-09-10-private-rendezvous.md

@@ -0,0 +1,26 @@
+# Private rendezvous protocol
+
+## Problem
+
+Our current handshake protocol is open to this attack: whoever observes the link exchange, knows on which server connection is being made, and if the traffic on this server is observed, then it can confirm communication between parties. Further, even with the [last proposal](./2024-09-09-smp-blobs.md#possible-privacy-improvement), having real-time access to the server data allows to establish the exact messaging queue that is used to send messages.
+
+## Solution
+
+We could make the initial link exchange more private by making it harder for any observer to discover which server will be used for messaging by hiding this information from the server that hosts the initial link.
+
+Preliminary, the protocol could be the following:
+
+1. Connection initiator stores 224-256 bytes of encrypted connection link on a rendezvous server (link contains server host and linkId on another messaging server, not a rendezvous one).
+
+2. Rendezvous server adds these links to buckets, up to 64 links per bucket. Bucket ID is the timestamp when the bucket was created + a sequential bucket number, in case more than one bucket is created per second.
+
+3. The server responds to the link creator with a bucket ID where this link was added. That bucket ID is its timestamp + a number prevents server "fingerprinting" clients and using say one bucket for each client. If timestamp is different or a bucket number within this timestamp is too large, the client can refuse to use it, depending on the client settings.
+
+4. The initiating party will pass to the accepting party the rendezvous server host, the hash of this bucket ID (bucket link) and the passphrase to derive the key from. The initiating party has an option to pass a link and passphrase via two channels - in which case the link will only contain the bucket ID.
+
+5. The accepting party would then request the bucket via its ID hash (the server would store hashes to be able to look up - hash is used to prevent showing time in the link) and attempt to decrypt all contained links using the provided key.
+The accepting party then will continue the connection via the decrypted link.
+
+This obviously does not protect accepting party from the initiating party, if it can choose rendezvous server it controls. It also does not protect from the malicious rendezvous server that would collaborate with link observers. I think reunion doesn’t protect from it too.
+
+But it does protect connection from whoever observes the link, particularly if this link only contains the bucket and the key is passed separately, via some other channel.

rfcs/2024-09-15-shared-port.md

@@ -0,0 +1,163 @@
+# Sharing protocol ports with HTTPS
+
+Some networks block all ports other than web ports, including port 5223 used for SMP protocol by default. Running SMP servers on a common web port 443 would allow them to work on more networks. The servers would need to provide an HTTPS page for browsers (and probes).
+
+## Problem
+
+Browsers and tools rely on system CA bundles instead of certificate pinning.
+The crypto parameters used by HTTPS are different from what the protocols use.
+Public certificate providers like LetsEncrypt can only sign specific types of keys and Ed25519 isn't one of them.
+
+This means a server should distinguish browser and protocol clients and adjust its behavior to match.
+
+## Solution
+
+`tls` package has a server hook that allows producing a different set of `TLS.Credentials` according to a client-provided "Server Name Indication" extension.
+
+Since LE certificates are only handed out to domain names, TLS client will be sending the SNI.
+However client transports are constructed over connected sockets and the SNI wouldn't be present unless explicitly requested.
+When a client sends SNI, then it's a browser and a web credentials should be used.
+Otherwise it's a protocol client to be offered the self-signed ca, cert and key.
+
+When a transport colocated with a HTTPS, its ALPN list should be extended with `h2 http/1.1`.
+The browsers will send it, and it should be checked before running transport client.
+If HTTP ALPN is detected, then the client connection is served with HTTP `Application` instead (the same "server information" page).
+
+If some client connects to server IP, doesn't send SNI and doesn't send ALPN, it will look like a pre-handshake client.
+In that case a server will send its handshake first.
+This can be mitigated by delaying its handshake and letting the probe to issue its HTTP request.
+
+## Implementation plan
+
+An unmodified client should be able to use protocols on port 443 right away.
+
+The switchover happens inside `runTransportServerState` before `runClient`:
+
+```haskell
+runServer (tcpPort, ATransport t) = do
+  -- ...
+  runTransportServerState_ ss started tcpPort serverParams tCfg $ \socket h -> do -- expose raw socket for warp-tls internals to attach
+    negotiated <- getSessionALPN
+    if allowHTTP t && isHTTP negotiated -- only attempt the switch for the TLS transport
+      then runHTTP socket (tlsContext h)-- ... collect data and produce values needed to run WAI Application
+      else runClient serverSignKey t h `runReaderT` env -- performs serverHandshake etc as usual
+```
+
+The web app and server live outside, so `runHttp` has to be provided by the `runSMPServer` caller.
+Additonally, Warp is using its `InternalInfo` object that's scoped to `withII` bracket.
+
+```haskell
+  runServer ini = do
+    -- ...
+
+    runWebServer ini ServerInformation {config, information} $ if sharedHttps then Nothing else webHttpsParams -- suppress serving https
+    if sharedHttps
+      then withRunHTTP staticFilesPath \attachStatic -> runSMPServer cfg (Just attachStatic) -- provide wrapped application runner
+      else runSMPServer cfg Nothing
+```
+
+### Upstream
+
+The implementation relies on a few modification to upstream code:
+- `warp-tls`: The library provides `httpOverTls`, but it wants to do handshake itself.
+  Since we have to do the handshake to switch on ALPN, the setup function has to be split.
+  This is a resonable change that may be upstreamed and nothing blocks us from using the recent version.
+- `warp`: Only the re-export of `serveConnection` is needed.
+  Unfortunately the most recent `warp` version can't be used right away due to dependency cascade around `http-5` and `auto-update-2`.
+  So a fork containing the backported re-export has to be used until the dependencies are refreshed.
+
+
+### TLS.ServerParams
+
+When a server has port sharing enabled, a new set of TLS params is loaded and combined with transport params:
+
+```haskell
+newEnv config = do
+  -- ...
+  tlsServerParams <- loadTLSServerParams caCertificateFile certificateFile privateKeyFile (alpn transportConfig)
+  sharedServerParams <- forM ((,) <$> sharedHttpsCredentials config <*> alpn transportConfig) $ \((chain, key), alpn) ->
+    let ca = Nothing -- It is possible to provide CA certificate, but it is typical for web server to use combined certificate chains
+    loadHTTPSServerParams tlsServerParams ca chain key alpn
+```
+
+`loadHTTPSServerParams` extends params with:
+1. `onALPNClientSuggest` hook gets `["h2", "http/1.1"]` added to the ALPN list which is now required.
+2. `onServerNameIndication` hook added, which upon detecting client SNI prepends the web credentials.
+3. `sharedCredentials = T.Credentials []` should be done to prevent transport credentials confusing browsers.
+    But that aborts key exchange somewhere in tls internals, so disabled for now.
+    As a workaround, another set of dummy credentials can be provided in the hope that any sane browser would reject them.
+    Like, RC4 ciphers, "impossible" digest combination, etc.
+
+### supportedParameters
+
+TLS certificate chains provided by LetsEncrypt use ECDSA/P256 and that requires extending `supportedParameters` with things disabled in transports:
+
+```haskell
+    browserCiphers =
+      [ TE.cipher_TLS13_AES128CCM8_SHA256
+      , TE.cipher_ECDHE_ECDSA_AES128CCM8_SHA256
+      , TE.cipher_ECDHE_ECDSA_AES256CCM8_SHA256
+      ]
+    browserGroups =
+      [ T.P256
+      ]
+    browserSigs =
+      [ (T.HashSHA256, T.SignatureECDSA),
+        (T.HashSHA384, T.SignatureECDSA)
+      ]
+```
+
+This may not be enough for other certificate providers.
+
+## Configuration
+
+> XXX: This is for the current implementation and should be updated.
+
+Web certificate chain is picked up from the WEB section:
+
+```ini
+[TRANSPORT]
+port: 443
+
+[WEB]
+https: 443
+cert: /etc/opt/simplex/web.cert
+key: /etc/opt/simplex/web.key
+
+# Alternatively, with a proper access configuration, the paths can point to the LE creds directly:
+# cert: /etc/letsencrypt/live/smp.hostname.tld/fullchain.pem
+# key: /etc/letsencrypt/live/smp.hostname.tld/privkey.pem
+```
+
+When `TRANSPORT.port` matches `WEB.https` the transport server becomes shared.
+
+Perhaps a more desirable option would be explicit configuration resulting in additional transported to run:
+
+```ini
+[TRANSPORT]
+port: 5223 ; pure protocol transport
+# control_port: 5224
+shared_port: 443 ; variant 1: register in TRANSPORT
+
+[WEB]
+https: 443
+cert: /etc/opt/simplex/web.cert
+key: /etc/opt/simplex/web.key
+# transport: on ; variant 2:
+```
+
+## Caveats
+
+Serving static files and the protocols togother may pose a problem for those who currently use dedicated web servers as they should switch to embedded http handlers.
+
+As before, using embedded HTTP server is increasing attack surface.
+
+Users who want to run everything on a single host will have to add and extra IP address and bind servers to specific IPs instead of 0.0.0.0.
+An amalgamated server binary can be provided that would contain both SMP and XFTP servers, where transport will dispatch connections by handshake ALPN.
+
+## Alternative: Use transports routable with reverse-proxies
+
+An "industrial" reverse proxy may do the ALPN routing, serving HTTP by itself and delegating `smp` and `xftp` to protocol servers.
+Same with the `websockets`.
+
+Since this in effect does TLS termination, the protocol servers will have to rely on credentials from protocol handshakes.

rfcs/2024-09-25-ios-notifications-2.md

@@ -0,0 +1,49 @@
+# iOS notifications delivery
+
+## Problem
+
+For iOS notifications to be delivered the client has to create credentials for notification subscription on SMP server using NKEY command and after that create a subscription on notification server using SNEW command. These two commands are sent in sequence, after the connections are created, and for it to happen the client needs to be online and in foreground.
+
+iOS users tend to close the app when it is not used, and iOS has very limited permissions for background activities, so these notification subscriptions are created with a substantial delay, and notifications do not work.
+
+This problem is distinct from and probably more common than other problems affecting notifications delivery described [here](./2024-07-06-ios-notifications.md).
+
+## Solution
+
+1. When the new connection is created, the client already knows if it needs to create notification subscription or not, based on the conversation setting (e.g., if the group is muted, the client will not create notification subscription as well.). We should extend NEW command to avoid the need to send additional NKEY command with an option to create notification subscription at the point where connection is created. NDEL would still be used to disable this notification, and NKEY will be used to re-enable it.
+
+2. In the same way we stopped using SDEL command (NDEL sends notification DELD to subscribed notification server) to delete notificaiton subscriptions from notification server, we should delegate creating notification subscription on notification server to SMP servers. Clients could use keys agreed with ntf server for e2e encryption and for command authorization to encrypt and sign instruction to create notification subscription that will be forwarded to notification server using protocol similar to SMP proxies. This will avoid the need for clients to separately contact notification servers that won't happen until they are online.
+
+3. Instead of making Ntf server trust DELD notifications, we could send deletion instructions signed by the client, which will only fail to send in case notification server is down (and they won't be sent later after server restart).
+
+Cons:
+- If SMP servers were to retain in the storage the information about which notification server is used for which queue, it would reduce metadata privacy. While currently it is not an issue, as all notification servers are known and operated by us, once there are other client apps, this can be used for app users fingerprinting, which would act as a deterrence from using new apps – but only if app users use servers of operators who are different from the app provider. To mitigate it, we could only store it in server memory and include notification instruction in subscription commands (SUB) and include notification subscription status in SUB responses. We don't need to mitigate the problem of server being able to store this information, as messaging servers can observe which notification servers connect to them anyway.
+- If SMP server is restarted before the subscription request is forwared to the notification server, then it will have to be forwarded again, once the client subscribes. The problem here is that if the client is offline, it will neither subscribe to the queue to send notification subscription request, nor receive notifications from this queue. Storing notification server and subscription request would mitigate that, as in this case we could send all pending requests on server start, without depending on client subscriptions.
+- "Small" agent will need to support connections to ntf servers and manage workers that retry sending pending subscription requests.
+- Until the client learns the public keys of notification server, it will not be able to decrypt notifications. It potentially can be mitigated by using the public key of the server returned when token is created, in this way different client keys (per-queue) will be combined with the same ntf server key (per-token).
+
+## Implementation details
+
+1. NEW and NKEY commands will need to be extended to include notification subscription request. As the notifier ID needs to be sent to notification server, this notifier ID will have to be client-generated and supplied as part of NEW command.
+
+now:
+
+```haskell
+NEW :: RcvPublicAuthKey -> RcvPublicDhKey -> Maybe BasicAuth -> SubscriptionMode -> SenderCanSecure -> Command Recipient
+NKEY :: NtfPublicAuthKey -> RcvNtfPublicDhKey -> Command Recipient
+```
+
+extended:
+
+```haskell
+NEW :: RcvPublicAuthKey -> RcvPublicDhKey -> Maybe BasicAuth -> SubscriptionMode -> SenderCanSecure -> Maybe NtfRequest -> Command Recipient
+
+data NtfRequest = NtfRequest NotifierId NtfPublicAuthKey RcvNtfPublicDhKey NtfServerRequest
+
+data NtfServerRequest = NtfServerRequest NtfServer EncSingedNtfCmd
+
+NKEY :: NtfPublicAuthKey -> RcvNtfPublicDhKey -> Maybe NtfServerRequest -> Command Recipient
+-- NotifierID is passed in entity ID field of the transmission
+```
+
+2. Notification server will need to support an additional command to receive "proxied" subscription commands, `SFWD`, that would include `NtfServerRequest`. This command can include both `SNEW` and `SDEL` commands.

rfcs/2024-11-25-journal-expiration.md

@@ -0,0 +1,15 @@
+# Expiring messages in journal storage
+
+## Problem
+
+The journal storage servers recently migrated to do not delete delivered or expired messages, they only update pointers to journal file lines. The messages are actually deleted when the whole journal file is deleted (when fully deleted or fully expired).
+
+The problem is that in case the queue stops receiving the new messages then writing of messages won't switch to the new journal file, and the current journal file containing delivered or expired messages would never be deleted.
+
+## Solution
+
+Remove current journal file and update queue_state.log during message expiration of "idle" queue (that is, without any new messages received or delivered within 3 hours) in case when:
+- the queue is "empty" after the expiration
+- the queue contains only quota marker(s), in which case move them to a new journal file and update the queue_state accordingly. Quota markers can be kept indefinitely to prevent writing the new messages to the dormant queues that reached capacity, so it's important to handle this case.
+
+Also remove current journal file when the queue is opened in case it is empty (as it would not be ever expired in case it remains empty), and also update queue_state.log

rfcs/2024-11-25-queue-blobs-2.md

@@ -0,0 +1,58 @@
+# Blob extensions for SMP queues 2 and queue storage
+
+This document evolves the design proposed [here](./2024-09-09-smp-blobs.md).
+
+## Problems
+
+In addition to problems in the first doc, we have these issues with in-memory queue record storage:
+- many queues are idle or rarely used, but they are loaded to memory, and currently just loading all queues uses 20gb RAM on each server, and takes 10 min to process, increasing downtimes during restarts.
+- adding blobs to memory would make this problem much worse.
+
+## Proposed solution
+
+Move queues to the same journalling approach as [used for messages](./2024-09-01-smp-message-storage.md) now, with independent file names in the same folders.
+
+Each queue change would be logged to its own file, and every time the queue is opened the whole file will be read and compacted to a single line - replacing one store log for all queues, with individual log files for each queue.
+
+Queue deletion would not be making a record in the file, instead it would be deleting the entire folder - it would reduce retention period for any metadata of deleted queues.
+
+We could additionally record deletions to the central log, for debugging, and reset it on every start. But in this case we should not remove folders at the point of deletion, but rather mark them as deleted and delete on restart. TBC
+
+It would also allow simplifying blob storage by having only one blob per queue - for example, limied to 16kb (a bit smaller to fit in block) for contact address queues and 4-8kb for invitations (to fit PQ keys and conversation preferences).
+
+We would also need to be able to lookup recipient ID via sender/notifier/link IDs.
+
+One possible solution is to use and load to memory a central index file. But it is likely to also consume a lot of memory and result in slow starts.
+
+Another solution that is probably better is to use the same folder structure and put notifier/sender/link files with the ID of the recipient queue inside the files. So to locate recipient queue the sender would have to locate folder containing the reference file pointing to the recipient queue and then to locate the actual queue data.
+
+## Implementation details
+
+Each queue folder would these files:
+
+- queue_state.log (and timestamped backups) - to store pointers to message journals (already implemented)
+- messages.randomBase64.log - message journals (already implemented)
+- queue_rec.log (and timestamped backups) - to log complete queue record every time it is changed (so only the last line needs to be read following the same logic as with queue_state.log, to prevent file corruption).
+- blob.data, blob.data.bak, blob.timestamp.data - files for data blobs (to make sure some copy of this file is readable/correct in case of write corruption) - the same two step overwrite process will be used as currently with store log compacting:
+  - on write: 1. if file exists, move it to .bak, 2. store new blob to .data, 3. move .bak to .timestamp.data
+  - on read: 1. if .bak exists, move it to .data 2. use .data
+
+Additional suggestion to reduce probability of queue_state.log and queue_rec.log file corruption is to do one of the following:
+- log end of lines in the beginning of the output, not in the end, to prevent the last line from being corrupted in case the previous line was not fully stored. The downside is that the file will not be EOL terminated, and there will be no confirmation that the output was fully made.
+- log EOL both in the beginning and at the end of output, and ignore empty lines in between - this would both confirm that the last line is fully logged and prevent corruption of the next line in case it was not.
+- check the last byte of the file and log EOL if it is not EOL. Probably cleanest approach, but with a small performance cost.
+
+If queue folder is a reference to the queue, it may have one of these files:
+- notifier.id
+- sender.id
+- link.id
+
+These files would contain a one line with the recipient ID of the queue. These files would never change, they can only be deleted when queue is deleted or when notifier/link is deleted.
+
+There is logic in code preventing using the same ID in different contexts, and the ID size is large enough to make any collisions unlikely (192 bits), so with correctly working code the queue folder would either have one of reference files, and nothing else, or the queue and message files from the beginning of this section. But even if the same ID is re-used in different context, it should not cause any problems as file names don't overlap.
+
+While we could store different types of references in different types of folders, it would have additional costs of maintaining 4 folder hierarchies. Instead we could use the fact that it is one hierarchy to prevent using the same ID in different contexts.
+
+## Protocol
+
+The only change in protocol is that there will be only one blob per queue, without markers (see the previous doc). Otherwise the protocol and proposed privacy improvement seem reasonable.

rfcs/done/2023-10-25-remote-control.md

@@ -53,7 +53,7 @@ The session invitation contains this data:
 - CA TLS certificate fingerprint of the controller - this is part of long term identity of the controller established during the first session, and repeated in the subsequent session announcements.
 - Session Ed25519 public key used to verify the announcement and commands - this mitigates the compromise of the long term signature key, as the controller will have to sign each command with this key first.
 - Long-term Ed25519 public key used to verify the announcement and commands - this is part of the long term controller identity.
-- Session X25519 DH key and sntrup761 KEM encapsulation key to agree session encryption (both for multicast announcement and for commands and responses in TLS), as described in https://datatracker.ietf.org/doc/draft-josefsson-ntruprime-hybrid/. The new keys are used for each session, and if client key is already available (from the previous session), the computed shared secret will be used to encrypt the announcement multicast packet. The out-of-band invitation is unencrypted. These DH public key and KEM encapsulation key are always sent unencrypted. NaCL Cryptobox is used for encryption.
+- Session X25519 DH key and sntrup761 KEM encapsulation key to agree session encryption (both for multicast announcement and for commands and responses in TLS), as described in https://datatracker.ietf.org/doc/draft-josefsson-ntruprime-hybrid/. The new keys are used for each session, and if client key is already available (from the previous session), the computed shared secret will be used to encrypt the announcement multicast packet. The out-of-band invitation is unencrypted. This DH public key is always sent unencrypted. NaCL Cryptobox is used for encryption.
 
 Host device decrypts (except the first session) and validates the invitation:
 - Session signature is valid.

scripts/main/simplex-servers-stopscript

@@ -1,30 +1,174 @@
 #!/usr/bin/env sh
 set -eu
 
-path_conf_var="/var/opt"
-path_conf_smp="$path_conf_var/simplex"
-path_conf_xftp="$path_conf_var/simplex-xftp"
-path_conf_storelog_smp="$path_conf_smp/smp-server-store.log"
-path_conf_storelog_xftp="$path_conf_xftp/file-server-store.log"
+# Common
+# ------
 date="$(date -u '+%Y-%m-%dT%H:%M:%S')"
 
-backup_smp() {
- if [ -e "$path_conf_storelog_smp" ]; then
-  cp "$path_conf_storelog_smp" "${path_conf_storelog_smp}.${date:-date-failed}"
- fi
+GRN='\033[0;32m'
+YLW='\033[1;33m'
+BLU='\033[1;34m'
+RED='\033[0;31m'
+NC='\033[0m'
+
+path_conf_var="/var/opt"
+
+smp_variables() {
+	path_conf_smp="$path_conf_var/simplex"
+	path_conf_smp_archive="$path_conf_smp/backups"
+	path_conf_smp_archive_storelog="$path_conf_smp_archive/queues"
+	path_conf_smp_archive_stats="$path_conf_smp_archive/stats"
+	path_conf_smp_archive_messages="$path_conf_smp_archive/messages"
+	path_conf_storelog_smp="$path_conf_smp/smp-server-store.log"
+	path_conf_storelog_smp_out="$path_conf_smp_archive_storelog/smp-server-store.log.${date:-date-failed}"
+
+	path_conf_stats_smp="$path_conf_smp/smp-server-stats.log"
+	path_conf_stats_smp_out="$path_conf_smp_archive_stats/smp-server-stats.log.${date:-date-failed}"
+
+	path_conf_messages_smp="$path_conf_smp/smp-server-messages.log"
+	path_conf_messages_smp_out="$path_conf_smp_archive_messages/smp-server-messages.log.${date:-date-failed}"
 }
 
-backup_xftp() {
- if [ -e "$path_conf_storelog_xftp" ]; then
-  cp "$path_conf_storelog_xftp" "${path_conf_storelog_xftp}.${date:-date-failed}"
- fi
+xftp_variables() {
+	path_conf_xftp="$path_conf_var/simplex-xftp"
+	path_conf_xftp_archive="$path_conf_xftp/backups"
+
+	path_conf_xftp_archive_storelog="$path_conf_xftp_archive/queues"
+	path_conf_xftp_archive_stats="$path_conf_xftp_archive/stats"
+
+	path_conf_storelog_xftp="$path_conf_xftp/file-server-store.log"
+	path_conf_storelog_xftp_out="$path_conf_xftp_archive_storelog/file-server-store.log.${date:-date-failed}"
+
+	path_conf_stats_xftp="$path_conf_xftp/file-server-stats.log"
+	path_conf_stats_xftp_out="$path_conf_xftp_archive_stats/file-server-stats.log.${date:-date-failed}"
 }
 
-if [ "$1" = 'smp-server' ]; then
- backup_smp
-elif [ "$1" = 'xftp-server' ]; then
- backup_xftp
-else
- backup_smp
- backup_xftp
-fi
+checks() {
+	result=${SERVICE_RESULT:-exit-code}
+	status=${EXIT_STATUS:-TERM}
+
+	case "$result" in
+		success)
+			case "$status" in
+				TERM)
+					printf "${RED}Refusing to backup files with failed service state${NC}\n"
+					exit 1
+					;;
+				*)
+					:
+					;;
+			esac
+			;;
+		*)
+			printf "${RED}Refusing to backup files with failed service state${NC}\n"
+			exit 1
+			;;
+	esac
+}
+
+smp_check() {
+	if [ ! -d "$path_conf_smp_archive_storelog" ]; then
+		mkdir -p "$path_conf_smp_archive_storelog"
+	fi
+	if [ ! -d "$path_conf_smp_archive_messages" ]; then
+		mkdir -p "$path_conf_smp_archive_messages"
+	fi
+	if [ ! -d "$path_conf_smp_archive_stats" ]; then
+		mkdir -p "$path_conf_smp_archive_stats"
+	fi
+}
+
+xftp_check() {
+	if [ ! -d "$path_conf_xftp_archive_storelog" ]; then
+		mkdir -p "$path_conf_xftp_archive_storelog"
+	fi
+	if [ ! -d "$path_conf_xftp_archive_stats" ]; then
+		mkdir -p "$path_conf_xftp_archive_stats"
+	fi
+}
+
+backup() {
+	file="$1"
+	out="$2"
+	file_type="$3"
+
+	if [ -e "$file" ]; then
+		if cp "$file" "$out"; then
+			printf "${YLW}${file_type}${NC} ${GRN}backup successful:${NC} ${BLU}%s${NC}\n" "${out}"
+		else
+			printf "${YLW}${file_type}${NC} ${RED}backup failed!${NC}\n"
+		fi
+	fi
+
+	unset file out file_type
+}
+
+cleanup() {
+	directory="$1"
+	
+	file_type="$2"
+	
+	files_date=$(find "$directory" -type f -exec stat --format="%y" {} + | awk '{print $1}' | sort -nr | uniq | awk 'NR==2')
+	
+	if [ -n "$files_date" ]; then
+		files=$(find "$directory" -type f -not -newermt "$files_date" -printf "%T@ %Tc %p\n" | sort -n | awk '{print $NF}')
+
+		if [ -n "$files" ]; then
+			printf '%s' "$files" | xargs rm -f
+			printf "${YLW}Old ${file_type} files${NC}${GRN} has been deleted:${NC}\n"
+			files_colored=$(printf '%s' "$files" | awk '{print "\033[1;34m"$0"\033[0m"}')
+			printf "${files_colored}\n"
+		fi
+	fi
+
+	unset directory file_type files_date files
+}
+
+smp_backup() {
+	backup "$path_conf_storelog_smp" "$path_conf_storelog_smp_out" 'Storelog'
+	backup "$path_conf_messages_smp" "$path_conf_messages_smp_out" 'Messages'
+	backup "$path_conf_stats_smp" "$path_conf_stats_smp_out" 'Stats'
+}
+
+smp_cleanup() {
+	cleanup "$path_conf_smp_archive_storelog" 'storelog'
+	cleanup "$path_conf_smp_archive_stats" 'stats'
+	cleanup "$path_conf_smp_archive_messages" 'messages'
+}
+
+xftp_backup() {
+	backup "$path_conf_storelog_xftp" "$path_conf_storelog_xftp_out" 'Storelog'
+	backup "$path_conf_stats_xftp" "$path_conf_stats_xftp_out" 'Stats'
+}
+
+xftp_cleanup() {
+	cleanup "$path_conf_xftp_archive_storelog" 'storelog'
+	cleanup "$path_conf_xftp_archive_stats" 'stats'
+}
+
+main() {
+	type="${1:-}"
+  
+	checks
+
+	case "$type" in
+		smp-server)
+			smp_variables
+			smp_check
+			smp_backup
+			smp_cleanup
+			;;
+		xftp-server)
+			xftp_variables
+			xftp_check
+			xftp_backup
+			xftp_cleanup
+			;;
+		*)
+			printf "${YLW}Unknown server type.${NC}\n"
+			exit 1
+			;;
+	esac 
+}
+
+main "$@"

scripts/main/simplex-servers-update

@@ -2,8 +2,6 @@
 set -eu
 
 # Links to scripts/configs
-bin="https://github.com/simplex-chat/simplexmq/releases/latest/download"
-
 scripts="https://raw.githubusercontent.com/simplex-chat/simplexmq/stable/scripts/main"
 scripts_systemd_smp="$scripts/smp-server.service"
 scripts_systemd_xftp="$scripts/xftp-server.service"
@@ -76,6 +74,21 @@ installed_test() {
   set -u
 }
 
+set_version() {
+  ver="${VER:-latest}"
+
+  case "$ver" in 
+    latest)
+      bin="https://github.com/simplex-chat/simplexmq/releases/latest/download"
+      remote_version="$(curl --proto '=https' --tlsv1.2 -sSf -L https://api.github.com/repos/simplex-chat/simplexmq/releases/latest | grep -i "tag_name" | awk -F \" '{print $4}')"
+      ;;
+    *)
+      bin="https://github.com/simplex-chat/simplexmq/releases/download/${ver}"
+      remote_version="${ver}"
+      ;;
+  esac
+}
+
 update_scripts() {
   curl --proto '=https' --tlsv1.2 -sSf -L "$scripts_update" -o "$path_tmp_bin_update" && chmod +x "$path_tmp_bin_update"
   curl --proto '=https' --tlsv1.2 -sSf -L "$scripts_uninstall" -o "$path_tmp_bin_uninstall" && chmod +x "$path_tmp_bin_uninstall"
@@ -137,8 +150,6 @@ update_bins() {
   eval "bin=\$bin_${1}"
   eval "path_bin=\$path_bin_${1}"
 
-  remote_version="$(curl --proto '=https' --tlsv1.2 -sSf -L https://api.github.com/repos/simplex-chat/simplexmq/releases/latest | grep -i "tag_name" | awk -F \" '{print $4}')"
-
   set_ver() {
     local_version='unset'
     sed -i -- "s/local_version_${1}=.*/local_version_${1}='${remote_version}'/" "$path_conf_info/release"
@@ -192,6 +203,7 @@ checks() {
     exit 1
   fi
 
+  set_version
   os_test
   installed_test
 

scripts/main/smp-server.service

@@ -9,7 +9,9 @@ ExecStart=/usr/local/bin/smp-server start +RTS -N -RTS
 ExecStopPost=/usr/local/bin/simplex-servers-stopscript smp-server
 LimitNOFILE=65535
 KillSignal=SIGINT
+TimeoutStartSec=infinity
 TimeoutStopSec=infinity
+AmbientCapabilities=CAP_NET_BIND_SERVICE
 
 [Install]
 WantedBy=multi-user.target

scripts/main/xftp-server.service

@@ -9,6 +9,7 @@ ExecStart=/usr/local/bin/xftp-server start +RTS -N -RTS
 ExecStopPost=/usr/local/bin/simplex-servers-stopscript xftp-server
 LimitNOFILE=65535
 KillSignal=SIGINT
+TimeoutStartSec=infinity
 TimeoutStopSec=infinity
 AmbientCapabilities=CAP_NET_BIND_SERVICE
    

simplexmq.cabal

@@ -1,11 +1,7 @@
 cabal-version: 1.12
 
--- This file has been generated from package.yaml by hpack version 0.35.0.
---
--- see: https://github.com/sol/hpack
-
 name:           simplexmq
-version:        6.0.3.0
+version:        6.2.0.4
 synopsis:       SimpleXMQ message broker
 description:    This package includes <./docs/Simplex-Messaging-Server.html server>,
                 <./docs/Simplex-Messaging-Client.html client> and
@@ -66,6 +62,11 @@ flag use_crypton
   manual: True
   default: True
 
+flag client_library
+  description: Don't build server-related code.
+  manual: True
+  default: False
+
 library
   exposed-modules:
       Simplex.FileTransfer.Agent
@@ -77,13 +78,6 @@ library
       Simplex.FileTransfer.Crypto
       Simplex.FileTransfer.Description
       Simplex.FileTransfer.Protocol
-      Simplex.FileTransfer.Server
-      Simplex.FileTransfer.Server.Control
-      Simplex.FileTransfer.Server.Env
-      Simplex.FileTransfer.Server.Main
-      Simplex.FileTransfer.Server.Stats
-      Simplex.FileTransfer.Server.Store
-      Simplex.FileTransfer.Server.StoreLog
       Simplex.FileTransfer.Transport
       Simplex.FileTransfer.Types
       Simplex.FileTransfer.Util
@@ -135,6 +129,8 @@ library
       Simplex.Messaging.Agent.Store.SQLite.Migrations.M20240417_rcv_files_approved_relays
       Simplex.Messaging.Agent.Store.SQLite.Migrations.M20240624_snd_secure
       Simplex.Messaging.Agent.Store.SQLite.Migrations.M20240702_servers_stats
+      Simplex.Messaging.Agent.Store.SQLite.Migrations.M20240930_ntf_tokens_to_delete
+      Simplex.Messaging.Agent.Store.SQLite.Migrations.M20241007_rcv_queues_last_broker_ts
       Simplex.Messaging.Agent.TRcvQueues
       Simplex.Messaging.Client
       Simplex.Messaging.Client.Agent
@@ -152,32 +148,13 @@ library
       Simplex.Messaging.Encoding.String
       Simplex.Messaging.Notifications.Client
       Simplex.Messaging.Notifications.Protocol
-      Simplex.Messaging.Notifications.Server
-      Simplex.Messaging.Notifications.Server.Env
-      Simplex.Messaging.Notifications.Server.Main
-      Simplex.Messaging.Notifications.Server.Push.APNS
-      Simplex.Messaging.Notifications.Server.Push.APNS.Internal
-      Simplex.Messaging.Notifications.Server.Stats
-      Simplex.Messaging.Notifications.Server.Store
-      Simplex.Messaging.Notifications.Server.StoreLog
       Simplex.Messaging.Notifications.Transport
       Simplex.Messaging.Notifications.Types
       Simplex.Messaging.Parsers
       Simplex.Messaging.Protocol
-      Simplex.Messaging.Server
       Simplex.Messaging.Server.CLI
-      Simplex.Messaging.Server.Control
-      Simplex.Messaging.Server.Env.STM
       Simplex.Messaging.Server.Expiration
-      Simplex.Messaging.Server.Information
-      Simplex.Messaging.Server.Main
-      Simplex.Messaging.Server.MsgStore
-      Simplex.Messaging.Server.MsgStore.STM
-      Simplex.Messaging.Server.QueueStore
       Simplex.Messaging.Server.QueueStore.QueueInfo
-      Simplex.Messaging.Server.QueueStore.STM
-      Simplex.Messaging.Server.Stats
-      Simplex.Messaging.Server.StoreLog
       Simplex.Messaging.ServiceScheme
       Simplex.Messaging.Session
       Simplex.Messaging.TMap
@@ -200,13 +177,46 @@ library
       Simplex.RemoteControl.Discovery.Multicast
       Simplex.RemoteControl.Invitation
       Simplex.RemoteControl.Types
+  if !flag(client_library)
+      exposed-modules:
+          Simplex.FileTransfer.Server
+          Simplex.FileTransfer.Server.Control
+          Simplex.FileTransfer.Server.Env
+          Simplex.FileTransfer.Server.Main
+          Simplex.FileTransfer.Server.Stats
+          Simplex.FileTransfer.Server.Store
+          Simplex.FileTransfer.Server.StoreLog
+          Simplex.Messaging.Server
+          Simplex.Messaging.Server.Control
+          Simplex.Messaging.Server.Env.STM
+          Simplex.Messaging.Server.Information
+          Simplex.Messaging.Server.Main
+          Simplex.Messaging.Server.MsgStore
+          Simplex.Messaging.Server.MsgStore.Journal
+          Simplex.Messaging.Server.MsgStore.STM
+          Simplex.Messaging.Server.MsgStore.Types
+          Simplex.Messaging.Server.NtfStore
+          Simplex.Messaging.Server.QueueStore
+          Simplex.Messaging.Server.QueueStore.STM
+          Simplex.Messaging.Server.Stats
+          Simplex.Messaging.Server.StoreLog
+          Simplex.Messaging.Server.StoreLog.Types
+          Simplex.Messaging.Notifications.Server
+          Simplex.Messaging.Notifications.Server.Control
+          Simplex.Messaging.Notifications.Server.Env
+          Simplex.Messaging.Notifications.Server.Main
+          Simplex.Messaging.Notifications.Server.Push.APNS
+          Simplex.Messaging.Notifications.Server.Push.APNS.Internal
+          Simplex.Messaging.Notifications.Server.Stats
+          Simplex.Messaging.Notifications.Server.Store
+          Simplex.Messaging.Notifications.Server.StoreLog
   other-modules:
       Paths_simplexmq
   hs-source-dirs:
       src
   default-extensions:
       StrictData
-  ghc-options: -Weverything -Wno-missing-exported-signatures -Wno-missing-import-lists -Wno-missed-specialisations -Wno-all-missed-specialisations -Wno-unsafe -Wno-safe -Wno-missing-local-signatures -Wno-missing-kind-signatures -Wno-missing-deriving-strategies -Wno-monomorphism-restriction -Wno-prepositive-qualified-module -Wno-unused-packages -Wno-implicit-prelude -Wno-missing-safe-haskell-mode -Wno-missing-export-lists -Wno-partial-fields -Wcompat -Werror=incomplete-record-updates -Werror=incomplete-patterns -Werror=incomplete-uni-patterns -Werror=missing-methods -Werror=tabs -Wredundant-constraints -Wincomplete-record-updates -Wunused-type-patterns -O2
+  ghc-options: -Weverything -Wno-missing-exported-signatures -Wno-missing-import-lists -Wno-missed-specialisations -Wno-all-missed-specialisations -Wno-unsafe -Wno-safe -Wno-missing-local-signatures -Wno-missing-kind-signatures -Wno-missing-deriving-strategies -Wno-monomorphism-restriction -Wno-prepositive-qualified-module -Wno-implicit-prelude -Wno-missing-safe-haskell-mode -Wno-missing-export-lists -Wno-partial-fields -Wcompat -Werror=incomplete-record-updates -Werror=incomplete-patterns -Werror=incomplete-uni-patterns -Werror=missing-home-modules -Werror=missing-methods -Werror=tabs -Wredundant-constraints -Wincomplete-record-updates -Wunused-type-patterns -O2
   include-dirs:
       cbits
   c-sources:
@@ -216,14 +226,12 @@ library
       crypto
   build-depends:
       aeson ==2.2.*
-    , ansi-terminal >=0.10 && <0.12
     , asn1-encoding ==0.9.*
     , asn1-types ==0.3.*
     , async ==2.2.*
     , attoparsec ==0.14.*
     , base >=4.14 && <5
     , base64-bytestring >=1.0 && <1.3
-    , case-insensitive ==1.2.*
     , composition ==1.0.*
     , constraints >=0.12 && <0.14
     , containers ==0.6.*
@@ -236,7 +244,6 @@ library
     , direct-sqlcipher ==2.3.*
     , directory ==1.3.*
     , filepath ==1.4.*
-    , hashable ==1.4.*
     , hourglass ==0.2.*
     , http-types ==0.12.*
     , http2 >=4.2.2 && <4.3
@@ -269,18 +276,22 @@ library
   default-language: Haskell2010
   if flag(swift)
     cpp-options: -DswiftJSON
+  if !flag(client_library)
+    build-depends:
+        case-insensitive ==1.2.*
+      , hashable ==1.4.*
   if impl(ghc >= 9.6.2)
     build-depends:
         bytestring ==0.11.*
-      , template-haskell ==2.20.*
       , text >=2.0.1 && <2.2
   if impl(ghc < 9.6.2)
     build-depends:
         bytestring ==0.10.*
-      , template-haskell ==2.16.*
       , text >=1.2.3.0 && <1.3
 
 executable ntf-server
+  if flag(client_library)
+      buildable: False
   main-is: Main.hs
   other-modules:
       Paths_simplexmq
@@ -288,75 +299,16 @@ executable ntf-server
       apps/ntf-server
   default-extensions:
       StrictData
-  ghc-options: -Weverything -Wno-missing-exported-signatures -Wno-missing-import-lists -Wno-missed-specialisations -Wno-all-missed-specialisations -Wno-unsafe -Wno-safe -Wno-missing-local-signatures -Wno-missing-kind-signatures -Wno-missing-deriving-strategies -Wno-monomorphism-restriction -Wno-prepositive-qualified-module -Wno-unused-packages -Wno-implicit-prelude -Wno-missing-safe-haskell-mode -Wno-missing-export-lists -Wno-partial-fields -Wcompat -Werror=incomplete-record-updates -Werror=incomplete-patterns -Werror=incomplete-uni-patterns -Werror=missing-methods -Werror=tabs -Wredundant-constraints -Wincomplete-record-updates -Wunused-type-patterns -O2 -threaded -rtsopts
+  ghc-options: -Weverything -Wno-missing-exported-signatures -Wno-missing-import-lists -Wno-missed-specialisations -Wno-all-missed-specialisations -Wno-unsafe -Wno-safe -Wno-missing-local-signatures -Wno-missing-kind-signatures -Wno-missing-deriving-strategies -Wno-monomorphism-restriction -Wno-prepositive-qualified-module -Wno-implicit-prelude -Wno-missing-safe-haskell-mode -Wno-missing-export-lists -Wno-partial-fields -Wcompat -Werror=incomplete-record-updates -Werror=incomplete-patterns -Werror=incomplete-uni-patterns -Werror=missing-methods -Werror=tabs -Wredundant-constraints -Wincomplete-record-updates -Wunused-type-patterns -O2 -threaded -rtsopts
   build-depends:
-      aeson ==2.2.*
-    , ansi-terminal >=0.10 && <0.12
-    , asn1-encoding ==0.9.*
-    , asn1-types ==0.3.*
-    , async ==2.2.*
-    , attoparsec ==0.14.*
-    , base >=4.14 && <5
-    , base64-bytestring >=1.0 && <1.3
-    , case-insensitive ==1.2.*
-    , composition ==1.0.*
-    , constraints >=0.12 && <0.14
-    , containers ==0.6.*
-    , crypton ==0.34.*
-    , crypton-x509 ==1.7.*
-    , crypton-x509-store ==1.6.*
-    , crypton-x509-validation ==1.6.*
-    , cryptostore ==0.3.*
-    , data-default ==0.7.*
-    , direct-sqlcipher ==2.3.*
-    , directory ==1.3.*
-    , filepath ==1.4.*
-    , hashable ==1.4.*
-    , hourglass ==0.2.*
-    , http-types ==0.12.*
-    , http2 >=4.2.2 && <4.3
-    , ini ==0.4.1
-    , iproute ==1.7.*
-    , iso8601-time ==0.1.*
-    , memory ==0.18.*
-    , mtl >=2.3.1 && <3.0
-    , network >=3.1.2.7 && <3.2
-    , network-info ==0.2.*
-    , network-transport ==0.5.6
-    , network-udp ==0.0.*
-    , optparse-applicative >=0.15 && <0.17
-    , process ==1.6.*
-    , random >=1.1 && <1.3
-    , simple-logger ==0.1.*
+      base
+    , simple-logger
     , simplexmq
-    , socks ==0.6.*
-    , sqlcipher-simple ==0.4.*
-    , stm ==2.5.*
-    , temporary ==1.3.*
-    , time ==1.12.*
-    , time-manager ==0.0.*
-    , tls >=1.9.0 && <1.10
-    , transformers ==0.6.*
-    , unliftio ==0.2.*
-    , unliftio-core ==0.2.*
-    , websockets ==0.12.*
-    , yaml ==0.11.*
-    , zstd ==0.1.3.*
   default-language: Haskell2010
-  if flag(swift)
-    cpp-options: -DswiftJSON
-  if impl(ghc >= 9.6.2)
-    build-depends:
-        bytestring ==0.11.*
-      , template-haskell ==2.20.*
-      , text >=2.0.1 && <2.2
-  if impl(ghc < 9.6.2)
-    build-depends:
-        bytestring ==0.10.*
-      , template-haskell ==2.16.*
-      , text >=1.2.3.0 && <1.3
 
 executable smp-server
+  if flag(client_library)
+      buildable: False
   main-is: Main.hs
   other-modules:
       Static
@@ -367,79 +319,27 @@ executable smp-server
       apps/smp-server/web
   default-extensions:
       StrictData
-  ghc-options: -Weverything -Wno-missing-exported-signatures -Wno-missing-import-lists -Wno-missed-specialisations -Wno-all-missed-specialisations -Wno-unsafe -Wno-safe -Wno-missing-local-signatures -Wno-missing-kind-signatures -Wno-missing-deriving-strategies -Wno-monomorphism-restriction -Wno-prepositive-qualified-module -Wno-unused-packages -Wno-implicit-prelude -Wno-missing-safe-haskell-mode -Wno-missing-export-lists -Wno-partial-fields -Wcompat -Werror=incomplete-record-updates -Werror=incomplete-patterns -Werror=incomplete-uni-patterns -Werror=missing-methods -Werror=tabs -Wredundant-constraints -Wincomplete-record-updates -Wunused-type-patterns -O2 -threaded -rtsopts
+  ghc-options: -Weverything -Wno-missing-exported-signatures -Wno-missing-import-lists -Wno-missed-specialisations -Wno-all-missed-specialisations -Wno-unsafe -Wno-safe -Wno-missing-local-signatures -Wno-missing-kind-signatures -Wno-missing-deriving-strategies -Wno-monomorphism-restriction -Wno-prepositive-qualified-module -Wno-implicit-prelude -Wno-missing-safe-haskell-mode -Wno-missing-export-lists -Wno-partial-fields -Wcompat -Werror=incomplete-record-updates -Werror=incomplete-patterns -Werror=incomplete-uni-patterns -Werror=missing-methods -Werror=tabs -Wredundant-constraints -Wincomplete-record-updates -Wunused-type-patterns -O2 -threaded -rtsopts
   build-depends:
-      aeson ==2.2.*
-    , ansi-terminal >=0.10 && <0.12
-    , asn1-encoding ==0.9.*
-    , asn1-types ==0.3.*
-    , async ==2.2.*
-    , attoparsec ==0.14.*
-    , base >=4.14 && <5
-    , base64-bytestring >=1.0 && <1.3
-    , case-insensitive ==1.2.*
-    , composition ==1.0.*
-    , constraints >=0.12 && <0.14
-    , containers ==0.6.*
-    , crypton ==0.34.*
-    , crypton-x509 ==1.7.*
-    , crypton-x509-store ==1.6.*
-    , crypton-x509-validation ==1.6.*
-    , cryptostore ==0.3.*
-    , data-default ==0.7.*
-    , direct-sqlcipher ==2.3.*
-    , directory ==1.3.*
+      base
+    , bytestring
+    , directory
     , file-embed
-    , filepath ==1.4.*
-    , hashable ==1.4.*
-    , hourglass ==0.2.*
-    , http-types ==0.12.*
-    , http2 >=4.2.2 && <4.3
-    , ini ==0.4.1
-    , iproute ==1.7.*
-    , iso8601-time ==0.1.*
-    , memory ==0.18.*
-    , mtl >=2.3.1 && <3.0
-    , network >=3.1.2.7 && <3.2
-    , network-info ==0.2.*
-    , network-transport ==0.5.6
-    , network-udp ==0.0.*
-    , optparse-applicative >=0.15 && <0.17
-    , process ==1.6.*
-    , random >=1.1 && <1.3
-    , simple-logger ==0.1.*
+    , filepath
+    , network
+    , simple-logger
     , simplexmq
-    , socks ==0.6.*
-    , sqlcipher-simple ==0.4.*
-    , stm ==2.5.*
-    , temporary ==1.3.*
-    , time ==1.12.*
-    , time-manager ==0.0.*
-    , tls >=1.9.0 && <1.10
-    , transformers ==0.6.*
-    , unliftio ==0.2.*
-    , unliftio-core ==0.2.*
+    , text
+    , unliftio
+    , wai
     , wai-app-static
-    , warp
-    , warp-tls
-    , websockets ==0.12.*
-    , yaml ==0.11.*
-    , zstd ==0.1.3.*
+    , warp ==3.3.30
+    , warp-tls ==3.4.7
   default-language: Haskell2010
-  if flag(swift)
-    cpp-options: -DswiftJSON
-  if impl(ghc >= 9.6.2)
-    build-depends:
-        bytestring ==0.11.*
-      , template-haskell ==2.20.*
-      , text >=2.0.1 && <2.2
-  if impl(ghc < 9.6.2)
-    build-depends:
-        bytestring ==0.10.*
-      , template-haskell ==2.16.*
-      , text >=1.2.3.0 && <1.3
 
 executable xftp
+  if flag(client_library)
+      buildable: False
   main-is: Main.hs
   other-modules:
       Paths_simplexmq
@@ -447,75 +347,15 @@ executable xftp
       apps/xftp
   default-extensions:
       StrictData
-  ghc-options: -Weverything -Wno-missing-exported-signatures -Wno-missing-import-lists -Wno-missed-specialisations -Wno-all-missed-specialisations -Wno-unsafe -Wno-safe -Wno-missing-local-signatures -Wno-missing-kind-signatures -Wno-missing-deriving-strategies -Wno-monomorphism-restriction -Wno-prepositive-qualified-module -Wno-unused-packages -Wno-implicit-prelude -Wno-missing-safe-haskell-mode -Wno-missing-export-lists -Wno-partial-fields -Wcompat -Werror=incomplete-record-updates -Werror=incomplete-patterns -Werror=incomplete-uni-patterns -Werror=missing-methods -Werror=tabs -Wredundant-constraints -Wincomplete-record-updates -Wunused-type-patterns -O2 -threaded -rtsopts
+  ghc-options: -Weverything -Wno-missing-exported-signatures -Wno-missing-import-lists -Wno-missed-specialisations -Wno-all-missed-specialisations -Wno-unsafe -Wno-safe -Wno-missing-local-signatures -Wno-missing-kind-signatures -Wno-missing-deriving-strategies -Wno-monomorphism-restriction -Wno-prepositive-qualified-module -Wno-implicit-prelude -Wno-missing-safe-haskell-mode -Wno-missing-export-lists -Wno-partial-fields -Wcompat -Werror=incomplete-record-updates -Werror=incomplete-patterns -Werror=incomplete-uni-patterns -Werror=missing-methods -Werror=tabs -Wredundant-constraints -Wincomplete-record-updates -Wunused-type-patterns -O2 -threaded -rtsopts
   build-depends:
-      aeson ==2.2.*
-    , ansi-terminal >=0.10 && <0.12
-    , asn1-encoding ==0.9.*
-    , asn1-types ==0.3.*
-    , async ==2.2.*
-    , attoparsec ==0.14.*
-    , base >=4.14 && <5
-    , base64-bytestring >=1.0 && <1.3
-    , case-insensitive ==1.2.*
-    , composition ==1.0.*
-    , constraints >=0.12 && <0.14
-    , containers ==0.6.*
-    , crypton ==0.34.*
-    , crypton-x509 ==1.7.*
-    , crypton-x509-store ==1.6.*
-    , crypton-x509-validation ==1.6.*
-    , cryptostore ==0.3.*
-    , data-default ==0.7.*
-    , direct-sqlcipher ==2.3.*
-    , directory ==1.3.*
-    , filepath ==1.4.*
-    , hashable ==1.4.*
-    , hourglass ==0.2.*
-    , http-types ==0.12.*
-    , http2 >=4.2.2 && <4.3
-    , ini ==0.4.1
-    , iproute ==1.7.*
-    , iso8601-time ==0.1.*
-    , memory ==0.18.*
-    , mtl >=2.3.1 && <3.0
-    , network >=3.1.2.7 && <3.2
-    , network-info ==0.2.*
-    , network-transport ==0.5.6
-    , network-udp ==0.0.*
-    , optparse-applicative >=0.15 && <0.17
-    , process ==1.6.*
-    , random >=1.1 && <1.3
-    , simple-logger ==0.1.*
+      base
     , simplexmq
-    , socks ==0.6.*
-    , sqlcipher-simple ==0.4.*
-    , stm ==2.5.*
-    , temporary ==1.3.*
-    , time ==1.12.*
-    , time-manager ==0.0.*
-    , tls >=1.9.0 && <1.10
-    , transformers ==0.6.*
-    , unliftio ==0.2.*
-    , unliftio-core ==0.2.*
-    , websockets ==0.12.*
-    , yaml ==0.11.*
-    , zstd ==0.1.3.*
   default-language: Haskell2010
-  if flag(swift)
-    cpp-options: -DswiftJSON
-  if impl(ghc >= 9.6.2)
-    build-depends:
-        bytestring ==0.11.*
-      , template-haskell ==2.20.*
-      , text >=2.0.1 && <2.2
-  if impl(ghc < 9.6.2)
-    build-depends:
-        bytestring ==0.10.*
-      , template-haskell ==2.16.*
-      , text >=1.2.3.0 && <1.3
 
 executable xftp-server
+  if flag(client_library)
+      buildable: False
   main-is: Main.hs
   other-modules:
       Paths_simplexmq
@@ -523,75 +363,16 @@ executable xftp-server
       apps/xftp-server
   default-extensions:
       StrictData
-  ghc-options: -Weverything -Wno-missing-exported-signatures -Wno-missing-import-lists -Wno-missed-specialisations -Wno-all-missed-specialisations -Wno-unsafe -Wno-safe -Wno-missing-local-signatures -Wno-missing-kind-signatures -Wno-missing-deriving-strategies -Wno-monomorphism-restriction -Wno-prepositive-qualified-module -Wno-unused-packages -Wno-implicit-prelude -Wno-missing-safe-haskell-mode -Wno-missing-export-lists -Wno-partial-fields -Wcompat -Werror=incomplete-record-updates -Werror=incomplete-patterns -Werror=incomplete-uni-patterns -Werror=missing-methods -Werror=tabs -Wredundant-constraints -Wincomplete-record-updates -Wunused-type-patterns -O2 -threaded -rtsopts
+  ghc-options: -Weverything -Wno-missing-exported-signatures -Wno-missing-import-lists -Wno-missed-specialisations -Wno-all-missed-specialisations -Wno-unsafe -Wno-safe -Wno-missing-local-signatures -Wno-missing-kind-signatures -Wno-missing-deriving-strategies -Wno-monomorphism-restriction -Wno-prepositive-qualified-module -Wno-implicit-prelude -Wno-missing-safe-haskell-mode -Wno-missing-export-lists -Wno-partial-fields -Wcompat -Werror=incomplete-record-updates -Werror=incomplete-patterns -Werror=incomplete-uni-patterns -Werror=missing-methods -Werror=tabs -Wredundant-constraints -Wincomplete-record-updates -Wunused-type-patterns -O2 -threaded -rtsopts
   build-depends:
-      aeson ==2.2.*
-    , ansi-terminal >=0.10 && <0.12
-    , asn1-encoding ==0.9.*
-    , asn1-types ==0.3.*
-    , async ==2.2.*
-    , attoparsec ==0.14.*
-    , base >=4.14 && <5
-    , base64-bytestring >=1.0 && <1.3
-    , case-insensitive ==1.2.*
-    , composition ==1.0.*
-    , constraints >=0.12 && <0.14
-    , containers ==0.6.*
-    , crypton ==0.34.*
-    , crypton-x509 ==1.7.*
-    , crypton-x509-store ==1.6.*
-    , crypton-x509-validation ==1.6.*
-    , cryptostore ==0.3.*
-    , data-default ==0.7.*
-    , direct-sqlcipher ==2.3.*
-    , directory ==1.3.*
-    , filepath ==1.4.*
-    , hashable ==1.4.*
-    , hourglass ==0.2.*
-    , http-types ==0.12.*
-    , http2 >=4.2.2 && <4.3
-    , ini ==0.4.1
-    , iproute ==1.7.*
-    , iso8601-time ==0.1.*
-    , memory ==0.18.*
-    , mtl >=2.3.1 && <3.0
-    , network >=3.1.2.7 && <3.2
-    , network-info ==0.2.*
-    , network-transport ==0.5.6
-    , network-udp ==0.0.*
-    , optparse-applicative >=0.15 && <0.17
-    , process ==1.6.*
-    , random >=1.1 && <1.3
-    , simple-logger ==0.1.*
+      base
+    , simple-logger
     , simplexmq
-    , socks ==0.6.*
-    , sqlcipher-simple ==0.4.*
-    , stm ==2.5.*
-    , temporary ==1.3.*
-    , time ==1.12.*
-    , time-manager ==0.0.*
-    , tls >=1.9.0 && <1.10
-    , transformers ==0.6.*
-    , unliftio ==0.2.*
-    , unliftio-core ==0.2.*
-    , websockets ==0.12.*
-    , yaml ==0.11.*
-    , zstd ==0.1.3.*
   default-language: Haskell2010
-  if flag(swift)
-    cpp-options: -DswiftJSON
-  if impl(ghc >= 9.6.2)
-    build-depends:
-        bytestring ==0.11.*
-      , template-haskell ==2.20.*
-      , text >=2.0.1 && <2.2
-  if impl(ghc < 9.6.2)
-    build-depends:
-        bytestring ==0.10.*
-      , template-haskell ==2.16.*
-      , text >=1.2.3.0 && <1.3
 
 test-suite simplexmq-test
+  if flag(client_library)
+      buildable: False
   type: exitcode-stdio-1.0
   main-is: Test.hs
   other-modules:
@@ -603,13 +384,17 @@ test-suite simplexmq-test
       AgentTests.MigrationTests
       AgentTests.NotificationTests
       AgentTests.SchemaDump
+      AgentTests.ServerChoice
       AgentTests.SQLiteTests
       CLITests
       CoreTests.BatchingTests
       CoreTests.CryptoFileTests
       CoreTests.CryptoTests
       CoreTests.EncodingTests
+      CoreTests.MsgStoreTests
       CoreTests.RetryIntervalTests
+      CoreTests.SOCKSSettings
+      CoreTests.StoreLogTests
       CoreTests.TRcvQueuesTests
       CoreTests.UtilTests
       CoreTests.VersionRangeTests
@@ -626,84 +411,61 @@ test-suite simplexmq-test
       XFTPCLI
       XFTPClient
       XFTPServerTests
+      Static
+      Static.Embedded
       Paths_simplexmq
   hs-source-dirs:
       tests
+      apps/smp-server/web
   default-extensions:
       StrictData
-  ghc-options: -Weverything -Wno-missing-exported-signatures -Wno-missing-import-lists -Wno-missed-specialisations -Wno-all-missed-specialisations -Wno-unsafe -Wno-safe -Wno-missing-local-signatures -Wno-missing-kind-signatures -Wno-missing-deriving-strategies -Wno-monomorphism-restriction -Wno-prepositive-qualified-module -Wno-unused-packages -Wno-implicit-prelude -Wno-missing-safe-haskell-mode -Wno-missing-export-lists -Wno-partial-fields -Wcompat -Werror=incomplete-record-updates -Werror=incomplete-patterns -Werror=incomplete-uni-patterns -Werror=missing-methods -Werror=tabs -Wredundant-constraints -Wincomplete-record-updates -Wunused-type-patterns -O2 -threaded -rtsopts -with-rtsopts=-A64M -with-rtsopts=-N1
+  ghc-options: -Weverything -Wno-missing-exported-signatures -Wno-missing-import-lists -Wno-missed-specialisations -Wno-all-missed-specialisations -Wno-unsafe -Wno-safe -Wno-missing-local-signatures -Wno-missing-kind-signatures -Wno-missing-deriving-strategies -Wno-monomorphism-restriction -Wno-prepositive-qualified-module -Wno-implicit-prelude -Wno-missing-safe-haskell-mode -Wno-missing-export-lists -Wno-partial-fields -Wcompat -Werror=incomplete-record-updates -Werror=incomplete-patterns -Werror=incomplete-uni-patterns -Werror=missing-methods -Werror=tabs -Wredundant-constraints -Wincomplete-record-updates -Wunused-type-patterns -O2 -threaded -rtsopts -with-rtsopts=-A64M -with-rtsopts=-N1
   build-depends:
-      HUnit ==1.6.*
-    , QuickCheck ==2.14.*
-    , aeson ==2.2.*
-    , ansi-terminal >=0.10 && <0.12
-    , asn1-encoding ==0.9.*
-    , asn1-types ==0.3.*
-    , async ==2.2.*
-    , attoparsec ==0.14.*
-    , base >=4.14 && <5
-    , base64-bytestring >=1.0 && <1.3
-    , case-insensitive ==1.2.*
-    , composition ==1.0.*
-    , constraints >=0.12 && <0.14
-    , containers ==0.6.*
-    , crypton ==0.34.*
-    , crypton-x509 ==1.7.*
-    , crypton-x509-store ==1.6.*
-    , crypton-x509-validation ==1.6.*
-    , cryptostore ==0.3.*
-    , data-default ==0.7.*
+      base
+    , aeson
+    , async
+    , base64-bytestring
+    , bytestring
+    , containers
+    , crypton
+    , crypton-x509
+    , crypton-x509-store
+    , crypton-x509-validation
     , deepseq ==1.4.*
-    , direct-sqlcipher ==2.3.*
-    , directory ==1.3.*
-    , filepath ==1.4.*
+    , directory
+    , file-embed
+    , filepath
     , generic-random ==1.5.*
-    , hashable ==1.4.*
-    , hourglass ==0.2.*
+    , hashable
     , hspec ==2.11.*
-    , hspec-core ==2.11.*
-    , http-types ==0.12.*
-    , http2 >=4.2.2 && <4.3
-    , ini ==0.4.1
-    , iproute ==1.7.*
-    , iso8601-time ==0.1.*
+    , http-client
+    , http-types
+    , http2
+    , HUnit ==1.6.*
+    , ini
+    , iso8601-time
     , main-tester ==0.2.*
-    , memory ==0.18.*
-    , mtl >=2.3.1 && <3.0
-    , network >=3.1.2.7 && <3.2
-    , network-info ==0.2.*
-    , network-transport ==0.5.6
-    , network-udp ==0.0.*
-    , optparse-applicative >=0.15 && <0.17
-    , process ==1.6.*
-    , random >=1.1 && <1.3
+    , memory
+    , mtl
+    , network
+    , process
+    , QuickCheck ==2.14.*
+    , random
     , silently ==1.2.*
-    , simple-logger ==0.1.*
+    , simple-logger
     , simplexmq
-    , socks ==0.6.*
-    , sqlcipher-simple ==0.4.*
-    , stm ==2.5.*
-    , temporary ==1.3.*
-    , time ==1.12.*
-    , time-manager ==0.0.*
+    , sqlcipher-simple
+    , stm
+    , text
+    , time
     , timeit ==2.0.*
-    , tls >=1.9.0 && <1.10
-    , transformers ==0.6.*
-    , unliftio ==0.2.*
-    , unliftio-core ==0.2.*
-    , websockets ==0.12.*
-    , yaml ==0.11.*
-    , zstd ==0.1.3.*
+    , transformers
+    , unliftio
+    , unliftio-core
+    , unordered-containers
+    , wai
+    , wai-app-static
+    , warp
+    , warp-tls
+    , yaml
   default-language: Haskell2010
-  if flag(swift)
-    cpp-options: -DswiftJSON
-  if impl(ghc >= 9.6.2)
-    build-depends:
-        bytestring ==0.11.*
-      , template-haskell ==2.20.*
-      , text >=2.0.1 && <2.2
-  if impl(ghc < 9.6.2)
-    build-depends:
-        bytestring ==0.10.*
-      , template-haskell ==2.16.*
-      , text >=1.2.3.0 && <1.3

src/Simplex/FileTransfer/Agent.hs

@@ -461,14 +461,14 @@ runXFTPSndPrepareWorker c Worker {doWork} = do
           pure srv
           where
             tryCreate = do
-              usedSrvs <- newTVarIO ([] :: [XFTPServer])
+              triedHosts <- newTVarIO S.empty
               let AgentClient {xftpServers} = c
               userSrvCount <- liftIO $ length <$> TM.lookupIO userId xftpServers
               withRetryIntervalCount (riFast ri) $ \n _ loop -> do
                 liftIO $ waitWhileSuspended c
                 liftIO $ waitForUserNetwork c
                 let triedAllSrvs = n > userSrvCount
-                createWithNextSrv usedSrvs
+                createWithNextSrv triedHosts
                   `catchAgentError` \e -> retryOnError "XFTP prepare worker" (retryLoop loop triedAllSrvs e) (throwE e) e
               where
                 -- we don't do closeXFTPServerClient here to not risk closing connection for concurrent chunk upload
@@ -477,10 +477,10 @@ runXFTPSndPrepareWorker c Worker {doWork} = do
                     when (triedAllSrvs && serverHostError e) $ notify c sndFileEntityId $ SFWARN e
                   liftIO $ assertAgentForeground c
                   loop
-            createWithNextSrv usedSrvs = do
+            createWithNextSrv triedHosts = do
               deleted <- withStore' c $ \db -> getSndFileDeleted db sndFileId
               when deleted $ throwE $ FILE NO_FILE
-              withNextSrv c userId usedSrvs [] $ \srvAuth -> do
+              withNextSrv c userId storageSrvs triedHosts [] $ \srvAuth -> do
                 replica <- agentXFTPNewChunk c ch numRecipients' srvAuth
                 pure (replica, srvAuth)
 

src/Simplex/FileTransfer/Client.hs

@@ -22,13 +22,13 @@ import Data.ByteString.Char8 (ByteString)
 import qualified Data.ByteString.Char8 as B
 import Data.Int (Int64)
 import Data.List.NonEmpty (NonEmpty (..))
+import Data.Time.Clock (UTCTime)
 import Data.Word (Word32)
 import qualified Data.X509 as X
 import qualified Data.X509.Validation as XV
 import qualified Network.HTTP.Types as N
 import qualified Network.HTTP2.Client as H
 import Simplex.FileTransfer.Protocol
-import Simplex.FileTransfer.Server.Env (supportedXFTPhandshakes)
 import Simplex.FileTransfer.Transport
 import Simplex.Messaging.Client
   ( NetworkConfig (..),
@@ -36,8 +36,8 @@ import Simplex.Messaging.Client
     TransportSession,
     chooseTransportHost,
     defaultNetworkConfig,
-    proxyUsername,
     transportClientConfig,
+    clientSocksCredentials,
     unexpectedResponse,
   )
 import qualified Simplex.Messaging.Crypto as C
@@ -52,7 +52,7 @@ import Simplex.Messaging.Protocol
     SenderId,
     pattern NoEntity,
   )
-import Simplex.Messaging.Transport (ALPN, HandshakeError (..), THandleAuth (..), THandleParams (..), TransportError (..), TransportPeer (..), supportedParameters)
+import Simplex.Messaging.Transport (ALPN, HandshakeError (..), THandleAuth (..), THandleParams (..), TransportError (..), TransportPeer (..), defaultSupportedParams)
 import Simplex.Messaging.Transport.Client (TransportClientConfig, TransportHost, alpn)
 import Simplex.Messaging.Transport.HTTP2
 import Simplex.Messaging.Transport.HTTP2.Client
@@ -98,21 +98,21 @@ defaultXFTPClientConfig =
       clientALPN = Just supportedXFTPhandshakes
     }
 
-getXFTPClient :: TransportSession FileResponse -> XFTPClientConfig -> (XFTPClient -> IO ()) -> IO (Either XFTPClientError XFTPClient)
-getXFTPClient transportSession@(_, srv, _) config@XFTPClientConfig {clientALPN, xftpNetworkConfig, serverVRange} disconnected = runExceptT $ do
-  let username = proxyUsername transportSession
+getXFTPClient :: TransportSession FileResponse -> XFTPClientConfig -> UTCTime -> (XFTPClient -> IO ()) -> IO (Either XFTPClientError XFTPClient)
+getXFTPClient transportSession@(_, srv, _) config@XFTPClientConfig {clientALPN, xftpNetworkConfig, serverVRange} proxySessTs disconnected = runExceptT $ do
+  let socksCreds = clientSocksCredentials xftpNetworkConfig proxySessTs transportSession
       ProtocolServer _ host port keyHash = srv
   useHost <- liftEither $ chooseTransportHost xftpNetworkConfig host
-  let tcConfig = (transportClientConfig xftpNetworkConfig useHost) {alpn = clientALPN}
+  let tcConfig = (transportClientConfig xftpNetworkConfig useHost False) {alpn = clientALPN}
       http2Config = xftpHTTP2Config tcConfig config
   clientVar <- newTVarIO Nothing
   let usePort = if null port then "443" else port
       clientDisconnected = readTVarIO clientVar >>= mapM_ disconnected
-  http2Client <- liftError' xftpClientError $ getVerifiedHTTP2Client (Just username) useHost usePort (Just keyHash) Nothing http2Config clientDisconnected
+  http2Client <- liftError' xftpClientError $ getVerifiedHTTP2Client socksCreds useHost usePort (Just keyHash) Nothing http2Config clientDisconnected
   let HTTP2Client {sessionId, sessionALPN} = http2Client
       v = VersionXFTP 1
       thServerVRange = versionToRange v
-      thParams0 = THandleParams {sessionId, blockSize = xftpBlockSize, thVersion = v, thServerVRange, thAuth = Nothing, implySessId = False, batch = True}
+      thParams0 = THandleParams {sessionId, blockSize = xftpBlockSize, thVersion = v, thServerVRange, thAuth = Nothing, implySessId = False, encryptBlock = Nothing, batch = True}
   logDebug $ "Client negotiated handshake protocol: " <> tshow sessionALPN
   thParams@THandleParams {thVersion} <- case sessionALPN of
     Just "xftp/1" -> xftpClientHandshakeV1 serverVRange keyHash http2Client thParams0
@@ -172,7 +172,7 @@ xftpHTTP2Config :: TransportClientConfig -> XFTPClientConfig -> HTTP2ClientConfi
 xftpHTTP2Config transportConfig XFTPClientConfig {xftpNetworkConfig = NetworkConfig {tcpConnectTimeout}} =
   defaultHTTP2ClientConfig
     { bodyHeadSize = xftpBlockSize,
-      suportedTLSParams = supportedParameters,
+      suportedTLSParams = defaultSupportedParams,
       connTimeout = tcpConnectTimeout,
       transportConfig
     }

src/Simplex/FileTransfer/Client/Agent.hs

@@ -16,6 +16,7 @@ import Data.Bifunctor (first)
 import qualified Data.ByteString.Char8 as B
 import Data.Text (Text)
 import Data.Text.Encoding (decodeUtf8)
+import Data.Time.Clock (UTCTime, getCurrentTime)
 import Simplex.FileTransfer.Client
 import Simplex.Messaging.Agent.RetryInterval
 import Simplex.Messaging.Client (NetworkConfig (..), ProtocolClientError (..), temporaryClientError)
@@ -30,6 +31,7 @@ type XFTPClientVar = TMVar (Either XFTPClientAgentError XFTPClient)
 
 data XFTPClientAgent = XFTPClientAgent
   { xftpClients :: TMap XFTPServer XFTPClientVar,
+    startedAt :: UTCTime,
     config :: XFTPClientAgentConfig
   }
 
@@ -56,19 +58,20 @@ data XFTPClientAgentError = XFTPClientAgentError XFTPServer XFTPClientError
 newXFTPAgent :: XFTPClientAgentConfig -> IO XFTPClientAgent
 newXFTPAgent config = do
   xftpClients <- TM.emptyIO
-  pure XFTPClientAgent {xftpClients, config}
+  startedAt <- getCurrentTime
+  pure XFTPClientAgent {xftpClients, startedAt, config}
 
 type ME a = ExceptT XFTPClientAgentError IO a
 
 getXFTPServerClient :: XFTPClientAgent -> XFTPServer -> ME XFTPClient
-getXFTPServerClient XFTPClientAgent {xftpClients, config} srv = do
+getXFTPServerClient XFTPClientAgent {xftpClients, startedAt, config} srv = do
   atomically getClientVar >>= either newXFTPClient waitForXFTPClient
   where
     connectClient :: ME XFTPClient
     connectClient =
       ExceptT $
         first (XFTPClientAgentError srv)
-          <$> getXFTPClient (1, srv, Nothing) (xftpConfig config) clientDisconnected
+          <$> getXFTPClient (1, srv, Nothing) (xftpConfig config) startedAt clientDisconnected
 
     clientDisconnected :: XFTPClient -> IO ()
     clientDisconnected _ = do

src/Simplex/FileTransfer/Server.hs

@@ -33,7 +33,6 @@ import qualified Data.Map.Strict as M
 import Data.Maybe (fromMaybe, isJust)
 import qualified Data.Text as T
 import Data.Time.Clock (UTCTime (..), diffTimeToPicoseconds, getCurrentTime)
-import Data.Time.Clock.System (SystemTime (..), getSystemTime)
 import Data.Time.Format.ISO8601 (iso8601Show)
 import Data.Word (Word32)
 import qualified Data.X509 as X
@@ -56,16 +55,18 @@ import Simplex.Messaging.Encoding
 import Simplex.Messaging.Encoding.String
 import Simplex.Messaging.Protocol (CorrId (..), EntityId (..), RcvPublicAuthKey, RcvPublicDhKey, RecipientId, TransmissionAuth, pattern NoEntity)
 import Simplex.Messaging.Server (dummyVerifyCmd, verifyCmdAuthorization)
+import Simplex.Messaging.Server.Control (CPClientRole (..))
 import Simplex.Messaging.Server.Expiration
+import Simplex.Messaging.Server.QueueStore (RoundedSystemTime, getRoundedSystemTime)
 import Simplex.Messaging.Server.Stats
 import Simplex.Messaging.TMap (TMap)
 import qualified Simplex.Messaging.TMap as TM
-import Simplex.Messaging.Transport (SessionId, THandleAuth (..), THandleParams (..), TransportPeer (..))
+import Simplex.Messaging.Transport (ALPN, SessionId, THandleAuth (..), THandleParams (..), TransportPeer (..), defaultSupportedParams)
 import Simplex.Messaging.Transport.Buffer (trimCR)
 import Simplex.Messaging.Transport.HTTP2
 import Simplex.Messaging.Transport.HTTP2.File (fileBlockSize)
 import Simplex.Messaging.Transport.HTTP2.Server
-import Simplex.Messaging.Transport.Server (runLocalTCPServer, tlsServerCredentials)
+import Simplex.Messaging.Transport.Server (runLocalTCPServer)
 import Simplex.Messaging.Util
 import Simplex.Messaging.Version
 import System.Exit (exitFailure)
@@ -91,36 +92,35 @@ data XFTPTransportRequest = XFTPTransportRequest
 runXFTPServer :: XFTPServerConfig -> IO ()
 runXFTPServer cfg = do
   started <- newEmptyTMVarIO
-  runXFTPServerBlocking started cfg
+  runXFTPServerBlocking started cfg $ Just supportedXFTPhandshakes
 
-runXFTPServerBlocking :: TMVar Bool -> XFTPServerConfig -> IO ()
-runXFTPServerBlocking started cfg = newXFTPServerEnv cfg >>= runReaderT (xftpServer cfg started)
+runXFTPServerBlocking :: TMVar Bool -> XFTPServerConfig -> Maybe [ALPN] -> IO ()
+runXFTPServerBlocking started cfg alpn_ = newXFTPServerEnv cfg >>= runReaderT (xftpServer cfg started alpn_)
 
 data Handshake
   = HandshakeSent C.PrivateKeyX25519
   | HandshakeAccepted (THandleParams XFTPVersion 'TServer)
 
-xftpServer :: XFTPServerConfig -> TMVar Bool -> M ()
-xftpServer cfg@XFTPServerConfig {xftpPort, transportConfig, inactiveClientExpiration, fileExpiration, xftpServerVRange} started = do
+xftpServer :: XFTPServerConfig -> TMVar Bool -> Maybe [ALPN] -> M ()
+xftpServer cfg@XFTPServerConfig {xftpPort, transportConfig, inactiveClientExpiration, fileExpiration, xftpServerVRange} started alpn_ = do
   mapM_ (expireServerFiles Nothing) fileExpiration
   restoreServerStats
   raceAny_ (runServer : expireFilesThread_ cfg <> serverStatsThread_ cfg <> controlPortThread_ cfg) `finally` stopServer
   where
     runServer :: M ()
     runServer = do
-      serverParams <- asks tlsServerParams
-      let (chain, pk) = tlsServerCredentials serverParams
+      srvCreds@(chain, pk) <- asks tlsServerCreds
       signKey <- liftIO $ case C.x509ToPrivate (pk, []) >>= C.privKey of
         Right pk' -> pure pk'
         Left e -> putStrLn ("servers has no valid key: " <> show e) >> exitFailure
       env <- ask
       sessions <- liftIO TM.emptyIO
       let cleanup sessionId = atomically $ TM.delete sessionId sessions
-      liftIO . runHTTP2Server started xftpPort defaultHTTP2BufferSize serverParams transportConfig inactiveClientExpiration cleanup $ \sessionId sessionALPN r sendResponse -> do
+      liftIO . runHTTP2Server started xftpPort defaultHTTP2BufferSize defaultSupportedParams srvCreds alpn_ transportConfig inactiveClientExpiration cleanup $ \sessionId sessionALPN r sendResponse -> do
         reqBody <- getHTTP2Body r xftpBlockSize
         let v = VersionXFTP 1
             thServerVRange = versionToRange v
-            thParams0 = THandleParams {sessionId, blockSize = xftpBlockSize, thVersion = v, thServerVRange, thAuth = Nothing, implySessId = False, batch = True}
+            thParams0 = THandleParams {sessionId, blockSize = xftpBlockSize, thVersion = v, thServerVRange, thAuth = Nothing, implySessId = False, encryptBlock = Nothing, batch = True}
             req0 = XFTPTransportRequest {thParams = thParams0, request = r, reqBody, sendResponse}
         flip runReaderT env $ case sessionALPN of
           Nothing -> processRequest req0
@@ -181,6 +181,7 @@ xftpServer cfg@XFTPServerConfig {xftpPort, transportConfig, inactiveClientExpira
     stopServer = do
       withFileLog closeStoreLog
       saveServerStats
+      logInfo "Server stopped"
 
     expireFilesThread_ :: XFTPServerConfig -> [M ()]
     expireFilesThread_ XFTPServerConfig {fileExpiration = Just fileExp} = [expireFiles fileExp]
@@ -399,7 +400,7 @@ processXFTPRequest HTTP2Body {bodyPart} = \case
       r <- runExceptT $ do
         sizes <- asks $ allowedChunkSizes . config
         unless (size file `elem` sizes) $ throwE SIZE
-        ts <- liftIO getSystemTime
+        ts <- liftIO getFileTime
         -- TODO validate body empty
         sId <- ExceptT $ addFileRetry st file 3 ts
         rcps <- mapM (ExceptT . addRecipientRetry st 3 sId) rks
@@ -412,7 +413,7 @@ processXFTPRequest HTTP2Body {bodyPart} = \case
         let rIds = L.map (\(FileRecipient rId _) -> rId) rcps
         pure $ FRSndIds sId rIds
       pure $ either FRErr id r
-    addFileRetry :: FileStore -> FileInfo -> Int -> SystemTime -> M (Either XFTPErrorType XFTPFileId)
+    addFileRetry :: FileStore -> FileInfo -> Int -> RoundedSystemTime -> M (Either XFTPErrorType XFTPFileId)
     addFileRetry st file n ts =
       retryAdd n $ \sId -> runExceptT $ do
         ExceptT $ addFile st sId file ts
@@ -531,6 +532,9 @@ deleteServerFile_ FileRec {senderId, fileInfo, filePath} = do
       liftIO $ atomicModifyIORef'_ (filesCount stats) (subtract 1)
       liftIO $ atomicModifyIORef'_ (filesSize stats) (subtract $ fromIntegral $ size fileInfo)
 
+getFileTime :: IO RoundedSystemTime
+getFileTime = getRoundedSystemTime fileTimePrecision
+
 expireServerFiles :: Maybe Int -> ExpirationConfig -> M ()
 expireServerFiles itemDelay expCfg = do
   st <- asks store

src/Simplex/FileTransfer/Server/Control.hs

@@ -8,8 +8,6 @@ import Simplex.FileTransfer.Protocol (XFTPFileId)
 import Simplex.Messaging.Encoding.String
 import Simplex.Messaging.Protocol (BasicAuth)
 
-data CPClientRole = CPRNone | CPRUser | CPRAdmin
-
 data ControlProtocol
   = CPAuth BasicAuth
   | CPStatsRTS

src/Simplex/FileTransfer/Server/Env.hs

@@ -28,8 +28,7 @@ import Simplex.FileTransfer.Transport (VersionRangeXFTP)
 import qualified Simplex.Messaging.Crypto as C
 import Simplex.Messaging.Protocol (BasicAuth, RcvPublicAuthKey)
 import Simplex.Messaging.Server.Expiration
-import Simplex.Messaging.Transport (ALPN)
-import Simplex.Messaging.Transport.Server (TransportServerConfig (..), loadFingerprint, loadTLSServerParams)
+import Simplex.Messaging.Transport.Server (ServerCredentials (..), TransportServerConfig (..), loadFingerprint, loadServerCredential)
 import Simplex.Messaging.Util (tshow)
 import System.IO (IOMode (..))
 import UnliftIO.STM
@@ -57,10 +56,7 @@ data XFTPServerConfig = XFTPServerConfig
     fileTimeout :: Int,
     -- | time after which inactive clients can be disconnected and check interval, seconds
     inactiveClientExpiration :: Maybe ExpirationConfig,
-    -- CA certificate private key is not needed for initialization
-    caCertificateFile :: FilePath,
-    privateKeyFile :: FilePath,
-    certificateFile :: FilePath,
+    xftpCredentials :: ServerCredentials,
     -- | XFTP client-server protocol version range
     xftpServerVRange :: VersionRangeXFTP,
     -- stats config - see SMP server config
@@ -75,7 +71,7 @@ data XFTPServerConfig = XFTPServerConfig
 defaultInactiveClientExpiration :: ExpirationConfig
 defaultInactiveClientExpiration =
   ExpirationConfig
-    { ttl = 43200, -- seconds, 12 hours
+    { ttl = 21600, -- seconds, 6 hours
       checkInterval = 3600 -- seconds, 1 hours
     }
 
@@ -85,7 +81,7 @@ data XFTPEnv = XFTPEnv
     storeLog :: Maybe (StoreLog 'WriteMode),
     random :: TVar ChaChaDRG,
     serverIdentity :: C.KeyHash,
-    tlsServerParams :: T.ServerParams,
+    tlsServerCreds :: T.Credential,
     serverStats :: FileServerStats
   }
 
@@ -99,11 +95,8 @@ defaultFileExpiration =
       checkInterval = 2 * 3600 -- seconds, 2 hours
     }
 
-supportedXFTPhandshakes :: [ALPN]
-supportedXFTPhandshakes = ["xftp/1"]
-
 newXFTPServerEnv :: XFTPServerConfig -> IO XFTPEnv
-newXFTPServerEnv config@XFTPServerConfig {storeLogFile, fileSizeQuota, caCertificateFile, certificateFile, privateKeyFile, transportConfig} = do
+newXFTPServerEnv config@XFTPServerConfig {storeLogFile, fileSizeQuota, xftpCredentials} = do
   random <- C.newRandom
   store <- newFileStore
   storeLog <- mapM (`readWriteFileStore` store) storeLogFile
@@ -112,10 +105,10 @@ newXFTPServerEnv config@XFTPServerConfig {storeLogFile, fileSizeQuota, caCertifi
   forM_ fileSizeQuota $ \quota -> do
     logInfo $ "Total / available storage: " <> tshow quota <> " / " <> tshow (quota - used)
     when (quota < used) $ logInfo "WARNING: storage quota is less than used storage, no files can be uploaded!"
-  tlsServerParams <- loadTLSServerParams caCertificateFile certificateFile privateKeyFile (alpn transportConfig)
-  Fingerprint fp <- loadFingerprint caCertificateFile
+  tlsServerCreds <- loadServerCredential xftpCredentials
+  Fingerprint fp <- loadFingerprint xftpCredentials
   serverStats <- newFileServerStats =<< getCurrentTime
-  pure XFTPEnv {config, store, storeLog, random, tlsServerParams, serverIdentity = C.KeyHash fp, serverStats}
+  pure XFTPEnv {config, store, storeLog, random, tlsServerCreds, serverIdentity = C.KeyHash fp, serverStats}
 
 countUsedStorage :: M.Map k FileRec -> Int64
 countUsedStorage = M.foldl' (\acc FileRec {fileInfo = FileInfo {size}} -> acc + fromIntegral size) 0

src/Simplex/FileTransfer/Server/Main.hs

@@ -19,7 +19,7 @@ import Options.Applicative
 import Simplex.FileTransfer.Chunks
 import Simplex.FileTransfer.Description (FileSize (..))
 import Simplex.FileTransfer.Server (runXFTPServer)
-import Simplex.FileTransfer.Server.Env (XFTPServerConfig (..), defFileExpirationHours, defaultFileExpiration, defaultInactiveClientExpiration, supportedXFTPhandshakes)
+import Simplex.FileTransfer.Server.Env (XFTPServerConfig (..), defFileExpirationHours, defaultFileExpiration, defaultInactiveClientExpiration)
 import Simplex.FileTransfer.Transport (supportedFileServerVRange)
 import qualified Simplex.Messaging.Crypto as C
 import Simplex.Messaging.Encoding.String
@@ -28,7 +28,7 @@ import Simplex.Messaging.Server.CLI
 import Simplex.Messaging.Server.Expiration
 import Simplex.Messaging.Transport (simplexMQVersion)
 import Simplex.Messaging.Transport.Client (TransportHost (..))
-import Simplex.Messaging.Transport.Server (TransportServerConfig (..), defaultTransportServerConfig)
+import Simplex.Messaging.Transport.Server (ServerCredentials (..), TransportServerConfig (..), defaultTransportServerConfig)
 import Simplex.Messaging.Util (safeDecodeUtf8, tshow)
 import System.Directory (createDirectoryIfMissing, doesFileExist)
 import System.FilePath (combine)
@@ -51,7 +51,9 @@ xftpServerCLI cfgPath logPath = do
         True -> readIniFile iniFile >>= either exitError runServer
         _ -> exitError $ "Error: server is not initialized (" <> iniFile <> " does not exist).\nRun `" <> executableName <> " init`."
     Delete -> do
-      confirmOrExit "WARNING: deleting the server will make all queues inaccessible, because the server identity (certificate fingerprint) will change.\nTHIS CANNOT BE UNDONE!"
+      confirmOrExit
+        "WARNING: deleting the server will make all queues inaccessible, because the server identity (certificate fingerprint) will change.\nTHIS CANNOT BE UNDONE!"
+        "Server NOT deleted"
       deleteDirIfExists cfgPath
       deleteDirIfExists logPath
       putStrLn "Deleted configuration and log files"
@@ -101,6 +103,7 @@ xftpServerCLI cfgPath logPath = do
                \\n\
                \# control_port_admin_password:\n\
                \# control_port_user_password:\n\
+               \\n\
                \[TRANSPORT]\n\
                \# host is only used to print server address on start\n"
             <> ("host: " <> T.pack host <> "\n")
@@ -173,9 +176,12 @@ xftpServerCLI cfgPath logPath = do
                     { ttl = readStrictIni "INACTIVE_CLIENTS" "ttl" ini,
                       checkInterval = readStrictIni "INACTIVE_CLIENTS" "check_interval" ini
                     },
-              caCertificateFile = c caCrtFile,
-              privateKeyFile = c serverKeyFile,
-              certificateFile = c serverCrtFile,
+              xftpCredentials =
+                ServerCredentials
+                  { caCertificateFile = Just $ c caCrtFile,
+                    privateKeyFile = c serverKeyFile,
+                    certificateFile = c serverCrtFile
+                  },
               xftpServerVRange = supportedFileServerVRange,
               logStatsInterval = logStats $> 86400, -- seconds
               logStatsStartTime = 0, -- seconds from 00:00 UTC
@@ -183,8 +189,7 @@ xftpServerCLI cfgPath logPath = do
               serverStatsBackupFile = logStats $> combine logPath "file-server-stats.log",
               transportConfig =
                 defaultTransportServerConfig
-                  { logTLSErrors = fromMaybe False $ iniOnOff "TRANSPORT" "log_tls_errors" ini,
-                    alpn = Just supportedXFTPhandshakes
+                  { logTLSErrors = fromMaybe False $ iniOnOff "TRANSPORT" "log_tls_errors" ini
                   },
               responseDelay = 0
             }

src/Simplex/FileTransfer/Server/Store.hs

@@ -17,6 +17,7 @@ module Simplex.FileTransfer.Server.Store
     expiredFilePath,
     getFile,
     ackFile,
+    fileTimePrecision,
   )
 where
 
@@ -25,12 +26,12 @@ import qualified Data.Attoparsec.ByteString.Char8 as A
 import Data.Int (Int64)
 import Data.Set (Set)
 import qualified Data.Set as S
-import Data.Time.Clock.System (SystemTime (..))
 import Simplex.FileTransfer.Protocol (FileInfo (..), SFileParty (..), XFTPFileId)
 import Simplex.FileTransfer.Transport (XFTPErrorType (..))
 import qualified Simplex.Messaging.Crypto as C
 import Simplex.Messaging.Encoding.String
 import Simplex.Messaging.Protocol (RcvPublicAuthKey, RecipientId, SenderId)
+import Simplex.Messaging.Server.QueueStore (RoundedSystemTime (..))
 import Simplex.Messaging.TMap (TMap)
 import qualified Simplex.Messaging.TMap as TM
 import Simplex.Messaging.Util (ifM, ($>>=))
@@ -46,10 +47,14 @@ data FileRec = FileRec
     fileInfo :: FileInfo,
     filePath :: TVar (Maybe FilePath),
     recipientIds :: TVar (Set RecipientId),
-    createdAt :: SystemTime
+    createdAt :: RoundedSystemTime
   }
 
+fileTimePrecision :: Int64
+fileTimePrecision = 3600 -- truncate creation time to 1 hour
+
 data FileRecipient = FileRecipient RecipientId RcvPublicAuthKey
+  deriving (Show)
 
 instance StrEncoding FileRecipient where
   strEncode (FileRecipient rId rKey) = strEncode rId <> ":" <> strEncode rKey
@@ -62,14 +67,14 @@ newFileStore = do
   usedStorage <- newTVarIO 0
   pure FileStore {files, recipients, usedStorage}
 
-addFile :: FileStore -> SenderId -> FileInfo -> SystemTime -> STM (Either XFTPErrorType ())
+addFile :: FileStore -> SenderId -> FileInfo -> RoundedSystemTime -> STM (Either XFTPErrorType ())
 addFile FileStore {files} sId fileInfo createdAt =
   ifM (TM.member sId files) (pure $ Left DUPLICATE_) $ do
     f <- newFileRec sId fileInfo createdAt
     TM.insert sId f files
     pure $ Right ()
 
-newFileRec :: SenderId -> FileInfo -> SystemTime -> STM FileRec
+newFileRec :: SenderId -> FileInfo -> RoundedSystemTime -> STM FileRec
 newFileRec senderId fileInfo createdAt = do
   recipientIds <- newTVar S.empty
   filePath <- newTVar Nothing
@@ -120,8 +125,8 @@ getFile st party fId = case party of
 expiredFilePath :: FileStore -> XFTPFileId -> Int64 -> STM (Maybe (Maybe FilePath))
 expiredFilePath FileStore {files} sId old =
   TM.lookup sId files
-    $>>= \FileRec {filePath, createdAt} ->
-      if systemSeconds createdAt < old
+    $>>= \FileRec {filePath, createdAt = RoundedSystemTime createdAt} ->
+      if createdAt + fileTimePrecision < old
         then Just <$> readTVar filePath
         else pure Nothing
 

src/Simplex/FileTransfer/Server/StoreLog.hs

@@ -28,22 +28,22 @@ import Data.List.NonEmpty (NonEmpty)
 import qualified Data.List.NonEmpty as L
 import Data.Map.Strict (Map)
 import qualified Data.Map.Strict as M
-import Data.Time.Clock.System (SystemTime)
 import Simplex.FileTransfer.Protocol (FileInfo (..))
 import Simplex.FileTransfer.Server.Store
 import Simplex.Messaging.Encoding.String
 import Simplex.Messaging.Protocol (RcvPublicAuthKey, RecipientId, SenderId)
+import Simplex.Messaging.Server.QueueStore (RoundedSystemTime)
 import Simplex.Messaging.Server.StoreLog
-import Simplex.Messaging.Util (bshow, whenM)
-import System.Directory (doesFileExist, renameFile)
+import Simplex.Messaging.Util (bshow)
 import System.IO
 
 data FileStoreLogRecord
-  = AddFile SenderId FileInfo SystemTime
+  = AddFile SenderId FileInfo RoundedSystemTime
   | PutFile SenderId FilePath
   | AddRecipients SenderId (NonEmpty FileRecipient)
   | DeleteFile SenderId
   | AckFile RecipientId
+  deriving (Show)
 
 instance StrEncoding FileStoreLogRecord where
   strEncode = \case
@@ -64,7 +64,7 @@ instance StrEncoding FileStoreLogRecord where
 logFileStoreRecord :: StoreLog 'WriteMode -> FileStoreLogRecord -> IO ()
 logFileStoreRecord = writeStoreLogRecord
 
-logAddFile :: StoreLog 'WriteMode -> SenderId -> FileInfo -> SystemTime -> IO ()
+logAddFile :: StoreLog 'WriteMode -> SenderId -> FileInfo -> RoundedSystemTime -> IO ()
 logAddFile s = logFileStoreRecord s .:. AddFile
 
 logPutFile :: StoreLog 'WriteMode -> SenderId -> FilePath -> IO ()
@@ -80,13 +80,7 @@ logAckFile :: StoreLog 'WriteMode -> RecipientId -> IO ()
 logAckFile s = logFileStoreRecord s . AckFile
 
 readWriteFileStore :: FilePath -> FileStore -> IO (StoreLog 'WriteMode)
-readWriteFileStore f st = do
-  whenM (doesFileExist f) $ do
-    readFileStore f st
-    renameFile f $ f <> ".bak"
-  s <- openWriteStoreLog f
-  writeFileStore s st
-  pure s
+readWriteFileStore = readWriteStoreLog readFileStore writeFileStore
 
 readFileStore :: FilePath -> FileStore -> IO ()
 readFileStore f st = mapM_ (addFileLogRecord . LB.toStrict) . LB.lines =<< LB.readFile f

src/Simplex/FileTransfer/Transport.hs

@@ -12,6 +12,7 @@ module Simplex.FileTransfer.Transport
   ( supportedFileServerVRange,
     authCmdsXFTPVersion,
     xftpClientHandshakeStub,
+    supportedXFTPhandshakes,
     XFTPClientHandshake (..),
     -- xftpClientHandshake,
     XFTPServerHandshake (..),
@@ -57,7 +58,7 @@ import Simplex.Messaging.Encoding
 import Simplex.Messaging.Encoding.String
 import Simplex.Messaging.Parsers
 import Simplex.Messaging.Protocol (CommandError)
-import Simplex.Messaging.Transport (SessionId, THandle (..), THandleParams (..), TransportError (..), TransportPeer (..))
+import Simplex.Messaging.Transport (ALPN, SessionId, THandle (..), THandleParams (..), TransportError (..), TransportPeer (..))
 import Simplex.Messaging.Transport.HTTP2.File
 import Simplex.Messaging.Util (bshow, tshow)
 import Simplex.Messaging.Version
@@ -101,6 +102,9 @@ supportedFileServerVRange = mkVersionRange initialXFTPVersion currentXFTPVersion
 xftpClientHandshakeStub :: c -> Maybe C.KeyPairX25519 -> C.KeyHash -> VersionRangeXFTP -> ExceptT TransportError IO (THandle XFTPVersion c 'TClient)
 xftpClientHandshakeStub _c _ks _keyHash _xftpVRange = throwE TEVersion
 
+supportedXFTPhandshakes :: [ALPN]
+supportedXFTPhandshakes = ["xftp/1"]
+
 data XFTPServerHandshake = XFTPServerHandshake
   { xftpVersionRange :: VersionRangeXFTP,
     sessionId :: SessionId,

src/Simplex/Messaging/Agent.hs

@@ -1,3 +1,4 @@
+{-# LANGUAGE BangPatterns #-}
 {-# LANGUAGE ConstraintKinds #-}
 {-# LANGUAGE DataKinds #-}
 {-# LANGUAGE DuplicateRecordFields #-}
@@ -55,14 +56,15 @@ module Simplex.Messaging.Agent
     createConnection,
     changeConnectionUser,
     prepareConnectionToJoin,
+    prepareConnectionToAccept,
     joinConnection,
     allowConnection,
     acceptContact,
     rejectContact,
     subscribeConnection,
     subscribeConnections,
-    getConnectionMessage,
-    getNotificationMessage,
+    getConnectionMessages,
+    getNotificationConns,
     resubscribeConnection,
     resubscribeConnections,
     sendMessage,
@@ -118,6 +120,8 @@ module Simplex.Messaging.Agent
     debugAgentLocks,
     getAgentSubscriptions,
     logConnection,
+    -- for tests
+    withAgentEnv,
   )
 where
 
@@ -132,7 +136,7 @@ import Data.Bifunctor (bimap, first, second)
 import Data.ByteString.Char8 (ByteString)
 import qualified Data.ByteString.Char8 as B
 import Data.Composition ((.:), (.:.), (.::), (.::.))
-import Data.Either (isRight, rights)
+import Data.Either (isRight, partitionEithers, rights)
 import Data.Foldable (foldl', toList)
 import Data.Functor (($>))
 import Data.Functor.Identity
@@ -166,22 +170,21 @@ import Simplex.Messaging.Agent.Store
 import Simplex.Messaging.Agent.Store.SQLite
 import qualified Simplex.Messaging.Agent.Store.SQLite.DB as DB
 import qualified Simplex.Messaging.Agent.Store.SQLite.Migrations as Migrations
-import Simplex.Messaging.Client (ProtocolClient (..), SMPClientError, ServerTransmission (..), ServerTransmissionBatch, temporaryClientError, unexpectedResponse)
+import Simplex.Messaging.Client (SMPClientError, ServerTransmission (..), ServerTransmissionBatch, temporaryClientError, unexpectedResponse)
 import qualified Simplex.Messaging.Crypto as C
 import Simplex.Messaging.Crypto.File (CryptoFile, CryptoFileArgs)
 import Simplex.Messaging.Crypto.Ratchet (PQEncryption, PQSupport (..), pattern PQEncOff, pattern PQEncOn, pattern PQSupportOff, pattern PQSupportOn)
 import qualified Simplex.Messaging.Crypto.Ratchet as CR
 import Simplex.Messaging.Encoding
 import Simplex.Messaging.Encoding.String
-import Simplex.Messaging.Notifications.Protocol (DeviceToken, NtfRegCode (NtfRegCode), NtfTknStatus (..), NtfTokenId)
-import Simplex.Messaging.Notifications.Server.Push.APNS (PNMessageData (..), pnMessagesP)
+import Simplex.Messaging.Notifications.Protocol (DeviceToken, NtfRegCode (NtfRegCode), NtfTknStatus (..), NtfTokenId, PNMessageData (..), pnMessagesP)
 import Simplex.Messaging.Notifications.Types
 import Simplex.Messaging.Parsers (parse)
 import Simplex.Messaging.Protocol (BrokerMsg, Cmd (..), ErrorType (AUTH), MsgBody, MsgFlags (..), NtfServer, ProtoServerWithAuth, ProtocolType (..), ProtocolTypeI (..), SMPMsgMeta, SParty (..), SProtocolType (..), SndPublicAuthKey, SubscriptionMode (..), UserProtocol, VersionSMPC, sndAuthKeySMPClientVersion)
 import qualified Simplex.Messaging.Protocol as SMP
 import Simplex.Messaging.ServiceScheme (ServiceScheme (..))
 import qualified Simplex.Messaging.TMap as TM
-import Simplex.Messaging.Transport (SMPVersion, THandleParams (sessionId))
+import Simplex.Messaging.Transport (SMPVersion)
 import Simplex.Messaging.Util
 import Simplex.Messaging.Version
 import Simplex.RemoteControl.Client
@@ -331,7 +334,7 @@ deleteConnectionsAsync c waitDelivery = withAgentEnv c . deleteConnectionsAsync'
 
 -- | Create SMP agent connection (NEW command)
 createConnection :: AgentClient -> UserId -> Bool -> SConnectionMode c -> Maybe CRClientData -> CR.InitialKeys -> SubscriptionMode -> AE (ConnId, ConnectionRequestUri c)
-createConnection c userId enableNtfs = withAgentEnv c .:: newConn c userId "" enableNtfs
+createConnection c userId enableNtfs = withAgentEnv c .:: newConn c userId enableNtfs
 {-# INLINE createConnection #-}
 
 -- | Changes the user id associated with a connection
@@ -347,10 +350,13 @@ changeConnectionUser c oldUserId connId newUserId = withAgentEnv c $ changeConne
 prepareConnectionToJoin :: AgentClient -> UserId -> Bool -> ConnectionRequestUri c -> PQSupport -> AE ConnId
 prepareConnectionToJoin c userId enableNtfs = withAgentEnv c .: newConnToJoin c userId "" enableNtfs
 
+-- | Create SMP agent connection without queue (to be joined with acceptContact passing invitation ID).
+prepareConnectionToAccept :: AgentClient -> Bool -> ConfirmationId -> PQSupport -> AE ConnId
+prepareConnectionToAccept c enableNtfs = withAgentEnv c .: newConnToAccept c "" enableNtfs
+
 -- | Join SMP agent connection (JOIN command).
-joinConnection :: AgentClient -> UserId -> Maybe ConnId -> Bool -> ConnectionRequestUri c -> ConnInfo -> PQSupport -> SubscriptionMode -> AE (ConnId, SndQueueSecured)
-joinConnection c userId Nothing enableNtfs = withAgentEnv c .:: joinConn c userId "" False enableNtfs
-joinConnection c userId (Just connId) enableNtfs = withAgentEnv c .:: joinConn c userId connId True enableNtfs
+joinConnection :: AgentClient -> UserId -> ConnId -> Bool -> ConnectionRequestUri c -> ConnInfo -> PQSupport -> SubscriptionMode -> AE SndQueueSecured
+joinConnection c userId connId enableNtfs = withAgentEnv c .:: joinConn c userId connId enableNtfs
 {-# INLINE joinConnection #-}
 
 -- | Allow connection to continue after CONF notification (LET command)
@@ -359,8 +365,8 @@ allowConnection c = withAgentEnv c .:. allowConnection' c
 {-# INLINE allowConnection #-}
 
 -- | Accept contact after REQ notification (ACPT command)
-acceptContact :: AgentClient -> Bool -> ConfirmationId -> ConnInfo -> PQSupport -> SubscriptionMode -> AE (ConnId, SndQueueSecured)
-acceptContact c enableNtfs = withAgentEnv c .:: acceptContact' c "" enableNtfs
+acceptContact :: AgentClient -> ConnId -> Bool -> ConfirmationId -> ConnInfo -> PQSupport -> SubscriptionMode -> AE SndQueueSecured
+acceptContact c connId enableNtfs = withAgentEnv c .:: acceptContact' c connId enableNtfs
 {-# INLINE acceptContact #-}
 
 -- | Reject contact (RJCT command)
@@ -378,15 +384,15 @@ subscribeConnections :: AgentClient -> [ConnId] -> AE (Map ConnId (Either AgentE
 subscribeConnections c = withAgentEnv c . subscribeConnections' c
 {-# INLINE subscribeConnections #-}
 
--- | Get connection message (GET command)
-getConnectionMessage :: AgentClient -> ConnId -> AE (Maybe SMPMsgMeta)
-getConnectionMessage c = withAgentEnv c . getConnectionMessage' c
-{-# INLINE getConnectionMessage #-}
+-- | Get messages for connections (GET commands)
+getConnectionMessages :: AgentClient -> NonEmpty ConnId -> IO (NonEmpty (Maybe SMPMsgMeta))
+getConnectionMessages c = withAgentEnv' c . getConnectionMessages' c
+{-# INLINE getConnectionMessages #-}
 
--- | Get connection message for received notification
-getNotificationMessage :: AgentClient -> C.CbNonce -> ByteString -> AE (NotificationInfo, Maybe SMPMsgMeta)
-getNotificationMessage c = withAgentEnv c .: getNotificationMessage' c
-{-# INLINE getNotificationMessage #-}
+-- | Get connections for received notification
+getNotificationConns :: AgentClient -> C.CbNonce -> ByteString -> AE (NonEmpty NotificationInfo)
+getNotificationConns c = withAgentEnv c .: getNotificationConns' c
+{-# INLINE getNotificationConns #-}
 
 resubscribeConnection :: AgentClient -> ConnId -> AE ()
 resubscribeConnection c = withAgentEnv c . resubscribeConnection' c
@@ -473,14 +479,14 @@ testProtocolServer c userId srv = withAgentEnv' c $ case protocolTypeI @p of
 
 -- | set SOCKS5 proxy on/off and optionally set TCP timeouts for fast network
 setNetworkConfig :: AgentClient -> NetworkConfig -> IO ()
-setNetworkConfig c@AgentClient {useNetworkConfig} cfg' = do
-  changed <- atomically $ do
+setNetworkConfig c@AgentClient {useNetworkConfig, proxySessTs} cfg' = do
+  (spChanged, changed) <- atomically $ do
     (_, cfg) <- readTVar useNetworkConfig
-    if cfg == cfg'
-      then pure False
-      else
-        let cfgSlow = slowNetworkConfig cfg'
-         in True <$ (cfgSlow `seq` writeTVar useNetworkConfig (cfgSlow, cfg'))
+    let changed = cfg /= cfg'
+        !cfgSlow = slowNetworkConfig cfg'
+    when changed $ writeTVar useNetworkConfig (cfgSlow, cfg')
+    pure (socksProxy cfg /= socksProxy cfg', changed)
+  when spChanged $ getCurrentTime >>= atomically . writeTVar proxySessTs
   when changed $ reconnectAllServers c
 
 setUserNetworkInfo :: AgentClient -> UserNetworkInfo -> IO ()
@@ -648,15 +654,15 @@ deleteUser' c@AgentClient {smpServersStats, xftpServersStats} userId delSMPQueue
 
 newConnAsync :: ConnectionModeI c => AgentClient -> UserId -> ACorrId -> Bool -> SConnectionMode c -> CR.InitialKeys -> SubscriptionMode -> AM ConnId
 newConnAsync c userId corrId enableNtfs cMode pqInitKeys subMode = do
-  connId <- newConnNoQueues c userId "" enableNtfs cMode (CR.connPQEncryption pqInitKeys)
+  connId <- newConnNoQueues c userId enableNtfs cMode (CR.connPQEncryption pqInitKeys)
   enqueueCommand c corrId connId Nothing $ AClientCommand $ NEW enableNtfs (ACM cMode) pqInitKeys subMode
   pure connId
 
-newConnNoQueues :: AgentClient -> UserId -> ConnId -> Bool -> SConnectionMode c -> PQSupport -> AM ConnId
-newConnNoQueues c userId connId enableNtfs cMode pqSupport = do
+newConnNoQueues :: AgentClient -> UserId -> Bool -> SConnectionMode c -> PQSupport -> AM ConnId
+newConnNoQueues c userId enableNtfs cMode pqSupport = do
   g <- asks random
   connAgentVersion <- asks $ maxVersion . smpAgentVRange . config
-  let cData = ConnData {userId, connId, connAgentVersion, enableNtfs, lastExternalSndId = 0, deleted = False, ratchetSyncState = RSOk, pqSupport}
+  let cData = ConnData {userId, connId = "", connAgentVersion, enableNtfs, lastExternalSndId = 0, deleted = False, ratchetSyncState = RSOk, pqSupport}
   withStore c $ \db -> createNewConn db g cData cMode
 
 joinConnAsync :: AgentClient -> UserId -> ACorrId -> Bool -> ConnectionRequestUri c -> ConnInfo -> PQSupport -> SubscriptionMode -> AM ConnId
@@ -743,9 +749,12 @@ switchConnectionAsync' c corrId connId =
             pure . connectionStats $ DuplexConnection cData rqs' sqs
       _ -> throwE $ CMD PROHIBITED "switchConnectionAsync: not duplex"
 
-newConn :: AgentClient -> UserId -> ConnId -> Bool -> SConnectionMode c -> Maybe CRClientData -> CR.InitialKeys -> SubscriptionMode -> AM (ConnId, ConnectionRequestUri c)
-newConn c userId connId enableNtfs cMode clientData pqInitKeys subMode =
-  getSMPServer c userId >>= newConnSrv c userId connId False enableNtfs cMode clientData pqInitKeys subMode
+newConn :: AgentClient -> UserId -> Bool -> SConnectionMode c -> Maybe CRClientData -> CR.InitialKeys -> SubscriptionMode -> AM (ConnId, ConnectionRequestUri c)
+newConn c userId enableNtfs cMode clientData pqInitKeys subMode = do
+  srv <- getSMPServer c userId
+  connId <- newConnNoQueues c userId enableNtfs cMode (CR.connPQEncryption pqInitKeys)
+  cReq <- newRcvConnSrv c userId connId enableNtfs cMode clientData pqInitKeys subMode srv
+  pure (connId, cReq)
 
 changeConnectionUser' :: AgentClient -> UserId -> ConnId -> UserId -> AM ()
 changeConnectionUser' c oldUserId connId newUserId = do
@@ -757,15 +766,7 @@ changeConnectionUser' c oldUserId connId newUserId = do
   where
     updateConn = withStore' c $ \db -> setConnUserId db oldUserId connId newUserId
 
-newConnSrv :: AgentClient -> UserId -> ConnId -> Bool -> Bool -> SConnectionMode c -> Maybe CRClientData -> CR.InitialKeys -> SubscriptionMode -> SMPServerWithAuth -> AM (ConnId, ConnectionRequestUri c)
-newConnSrv c userId connId hasNewConn enableNtfs cMode clientData pqInitKeys subMode srv = do
-  connId' <-
-    if hasNewConn
-      then pure connId
-      else newConnNoQueues c userId connId enableNtfs cMode (CR.connPQEncryption pqInitKeys)
-  newRcvConnSrv c userId connId' enableNtfs cMode clientData pqInitKeys subMode srv
-
-newRcvConnSrv :: AgentClient -> UserId -> ConnId -> Bool -> SConnectionMode c -> Maybe CRClientData -> CR.InitialKeys -> SubscriptionMode -> SMPServerWithAuth -> AM (ConnId, ConnectionRequestUri c)
+newRcvConnSrv :: AgentClient -> UserId -> ConnId -> Bool -> SConnectionMode c -> Maybe CRClientData -> CR.InitialKeys -> SubscriptionMode -> SMPServerWithAuth -> AM (ConnectionRequestUri c)
 newRcvConnSrv c userId connId enableNtfs cMode clientData pqInitKeys subMode srvWithAuth@(ProtoServerWithAuth srv _) = do
   case (cMode, pqInitKeys) of
     (SCMContact, CR.IKUsePQ) -> throwE $ CMD PROHIBITED "newRcvConnSrv"
@@ -778,15 +779,15 @@ newRcvConnSrv c userId connId enableNtfs cMode clientData pqInitKeys subMode srv
   lift . when (subMode == SMSubscribe) $ addNewQueueSubscription c rq' tSess sessId
   when enableNtfs $ do
     ns <- asks ntfSupervisor
-    atomically $ sendNtfSubCommand ns (connId, NSCCreate)
+    atomically $ sendNtfSubCommand ns (NSCCreate, [connId])
   let crData = ConnReqUriData SSSimplex smpAgentVRange [qUri] clientData
   case cMode of
-    SCMContact -> pure (connId, CRContactUri crData)
+    SCMContact -> pure $ CRContactUri crData
     SCMInvitation -> do
       g <- asks random
       (pk1, pk2, pKem, e2eRcvParams) <- liftIO $ CR.generateRcvE2EParams g (maxVersion e2eEncryptVRange) (CR.initialPQEncryption pqInitKeys)
       withStore' c $ \db -> createRatchetX3dhKeys db connId pk1 pk2 pKem
-      pure (connId, CRInvitationUri crData $ toVersionRangeT e2eRcvParams e2eEncryptVRange)
+      pure $ CRInvitationUri crData $ toVersionRangeT e2eRcvParams e2eEncryptVRange
 
 newConnToJoin :: forall c. AgentClient -> UserId -> ConnId -> Bool -> ConnectionRequestUri c -> PQSupport -> AM ConnId
 newConnToJoin c userId connId enableNtfs cReq pqSup = case cReq of
@@ -806,16 +807,26 @@ newConnToJoin c userId connId enableNtfs cReq pqSup = case cReq of
           cData = ConnData {userId, connId, connAgentVersion, enableNtfs, lastExternalSndId = 0, deleted = False, ratchetSyncState = RSOk, pqSupport}
       withStore c $ \db -> createNewConn db g cData SCMInvitation
 
-joinConn :: AgentClient -> UserId -> ConnId -> Bool -> Bool -> ConnectionRequestUri c -> ConnInfo -> PQSupport -> SubscriptionMode -> AM (ConnId, SndQueueSecured)
-joinConn c userId connId hasNewConn enableNtfs cReq cInfo pqSupport subMode = do
-  srv <- case cReq of
-    CRInvitationUri ConnReqUriData {crSmpQueues = q :| _} _ ->
-      getNextServer c userId [qServer q]
-    _ -> getSMPServer c userId
-  joinConnSrv c userId connId hasNewConn enableNtfs cReq cInfo pqSupport subMode srv
+newConnToAccept :: AgentClient -> ConnId -> Bool -> ConfirmationId -> PQSupport -> AM ConnId
+newConnToAccept c connId enableNtfs invId pqSup = do
+  Invitation {connReq, contactConnId} <- withStore c (`getInvitation` invId)
+  withStore c (`getConn` contactConnId) >>= \case
+    SomeConn _ (ContactConnection ConnData {userId} _) ->
+      newConnToJoin c userId connId enableNtfs connReq pqSup
+    _ -> throwE $ CMD PROHIBITED "newConnToAccept"
 
-startJoinInvitation :: UserId -> ConnId -> Maybe SndQueue -> Bool -> ConnectionRequestUri 'CMInvitation -> PQSupport -> AM (ConnData, NewSndQueue, C.PublicKeyX25519, CR.Ratchet 'C.X448, CR.SndE2ERatchetParams 'C.X448)
-startJoinInvitation userId connId sq_ enableNtfs cReqUri pqSup =
+joinConn :: AgentClient -> UserId -> ConnId -> Bool -> ConnectionRequestUri c -> ConnInfo -> PQSupport -> SubscriptionMode -> AM SndQueueSecured
+joinConn c userId connId enableNtfs cReq cInfo pqSupport subMode = do
+  srv <- getNextSMPServer c userId [qServer cReqQueue]
+  joinConnSrv c userId connId enableNtfs cReq cInfo pqSupport subMode srv
+  where
+    cReqQueue :: SMPQueueUri
+    cReqQueue = case cReq of
+      CRInvitationUri ConnReqUriData {crSmpQueues = q :| _} _ -> q
+      CRContactUri ConnReqUriData {crSmpQueues = q :| _} -> q
+
+startJoinInvitation :: AgentClient -> UserId -> ConnId -> Maybe SndQueue -> Bool -> ConnectionRequestUri 'CMInvitation -> PQSupport -> AM (ConnData, SndQueue, CR.SndE2ERatchetParams 'C.X448)
+startJoinInvitation c userId connId sq_ enableNtfs cReqUri pqSup =
   lift (compatibleInvitationUri cReqUri) >>= \case
     Just (qInfo, Compatible e2eRcvParams@(CR.E2ERatchetParams v _ rcDHRr kem_), Compatible connAgentVersion) -> do
       g <- asks random
@@ -828,11 +839,14 @@ startJoinInvitation userId connId sq_ enableNtfs cReqUri pqSup =
           rc = CR.initSndRatchet rcVs rcDHRr rcDHRs rcParams
       -- this case avoids re-generating queue keys and subsequent failure of SKEY that timed out
       -- e2ePubKey is always present, it's Maybe historically
-      (q, e2ePubKey) <- case sq_ of
-        Just sq@SndQueue {e2ePubKey = Just k} -> pure ((sq :: SndQueue) {dbQueueId = DBNewQueue}, k)
-        _ -> lift $ newSndQueue userId "" qInfo
+      q <- case sq_ of
+        Just sq@SndQueue {e2ePubKey = Just _k} -> pure (sq :: SndQueue) {dbQueueId = DBNewQueue}
+        _ -> lift $ fst <$> newSndQueue userId "" qInfo
       let cData = ConnData {userId, connId, connAgentVersion, enableNtfs, lastExternalSndId = 0, deleted = False, ratchetSyncState = RSOk, pqSupport}
-      pure (cData, q, e2ePubKey, rc, e2eSndParams)
+      sq' <- withStore c $ \db -> runExceptT $ do
+        liftIO $ createRatchet db connId rc
+        maybe (ExceptT $ updateNewConnSnd db connId q) pure sq_
+      pure (cData, sq', e2eSndParams)
     Nothing -> throwE $ AGENT A_VERSION
 
 connRequestPQSupport :: AgentClient -> PQSupport -> ConnectionRequestUri c -> IO (Maybe (VersionSMPA, PQSupport))
@@ -865,34 +879,25 @@ versionPQSupport_ :: VersionSMPA -> Maybe CR.VersionE2E -> PQSupport
 versionPQSupport_ agentV e2eV_ = PQSupport $ agentV >= pqdrSMPAgentVersion && maybe True (>= CR.pqRatchetE2EEncryptVersion) e2eV_
 {-# INLINE versionPQSupport_ #-}
 
-joinConnSrv :: AgentClient -> UserId -> ConnId -> Bool -> Bool -> ConnectionRequestUri c -> ConnInfo -> PQSupport -> SubscriptionMode -> SMPServerWithAuth -> AM (ConnId, SndQueueSecured)
-joinConnSrv c userId connId hasNewConn enableNtfs inv@CRInvitationUri {} cInfo pqSup subMode srv =
+joinConnSrv :: AgentClient -> UserId -> ConnId -> Bool -> ConnectionRequestUri c -> ConnInfo -> PQSupport -> SubscriptionMode -> SMPServerWithAuth -> AM SndQueueSecured
+joinConnSrv c userId connId enableNtfs inv@CRInvitationUri {} cInfo pqSup subMode srv =
   withInvLock c (strEncode inv) "joinConnSrv" $ do
-    (cData, q, _, rc, e2eSndParams) <- startJoinInvitation userId connId Nothing enableNtfs inv pqSup
-    g <- asks random
-    (connId', sq) <- withStore c $ \db -> runExceptT $ do
-      r@(connId', _) <-
-        if hasNewConn
-          then (connId,) <$> ExceptT (updateNewConnSnd db connId q)
-          else ExceptT $ createSndConn db g cData q
-      liftIO $ createRatchet db connId' rc
-      pure r
-    let cData' = (cData :: ConnData) {connId = connId'}
-    -- joinConnSrv is only used on user interaction, and its failure is permanent,
-    -- otherwise we would need to manage retries here to avoid SndQueue recreated with a different key,
-    -- similar to how joinConnAsync does that.
-    tryError (secureConfirmQueue c cData' sq srv cInfo (Just e2eSndParams) subMode) >>= \case
-      Right sqSecured -> pure (connId', sqSecured)
-      Left e -> do
-        -- possible improvement: recovery for failure on network timeout, see rfcs/2022-04-20-smp-conf-timeout-recovery.md
-        void $ withStore' c $ \db -> deleteConn db Nothing connId'
-        throwE e
-joinConnSrv c userId connId hasNewConn enableNtfs cReqUri@CRContactUri {} cInfo pqSup subMode srv =
+    SomeConn cType conn <- withStore c (`getConn` connId)
+    case conn of
+      NewConnection _ -> doJoin Nothing
+      SndConnection _ sq -> doJoin $ Just sq
+      _ -> throwE $ CMD PROHIBITED $ "joinConnSrv: bad connection " <> show cType
+  where
+    doJoin :: Maybe SndQueue -> AM SndQueueSecured
+    doJoin sq_ = do
+      (cData, sq, e2eSndParams) <- startJoinInvitation c userId connId sq_ enableNtfs inv pqSup
+      secureConfirmQueue c cData sq srv cInfo (Just e2eSndParams) subMode
+joinConnSrv c userId connId enableNtfs cReqUri@CRContactUri {} cInfo pqSup subMode srv =
   lift (compatibleContactUri cReqUri) >>= \case
     Just (qInfo, vrsn) -> do
-      (connId', cReq) <- newConnSrv c userId connId hasNewConn enableNtfs SCMInvitation Nothing (CR.IKNoPQ pqSup) subMode srv
-      void $ sendInvitation c userId connId' qInfo vrsn cReq cInfo
-      pure (connId', False)
+      cReq <- newRcvConnSrv c userId connId enableNtfs SCMInvitation Nothing (CR.IKNoPQ pqSup) subMode srv
+      void $ sendInvitation c userId connId qInfo vrsn cReq cInfo
+      pure False
     Nothing -> throwE $ AGENT A_VERSION
 
 joinConnSrvAsync :: AgentClient -> UserId -> ConnId -> Bool -> ConnectionRequestUri c -> ConnInfo -> PQSupport -> SubscriptionMode -> SMPServerWithAuth -> AM SndQueueSecured
@@ -905,11 +910,8 @@ joinConnSrvAsync c userId connId enableNtfs inv@CRInvitationUri {} cInfo pqSuppo
   where
     doJoin :: Maybe SndQueue -> AM SndQueueSecured
     doJoin sq_ = do
-      (cData, sq, _, rc, e2eSndParams) <- startJoinInvitation userId connId sq_ enableNtfs inv pqSupport
-      sq' <- withStore c $ \db -> runExceptT $ do
-        liftIO $ createRatchet db connId rc
-        maybe (ExceptT $ updateNewConnSnd db connId sq) pure sq_
-      secureConfirmQueueAsync c cData sq' srv cInfo (Just e2eSndParams) subMode
+      (cData, sq, e2eSndParams) <- startJoinInvitation c userId connId sq_ enableNtfs inv pqSupport
+      secureConfirmQueueAsync c cData sq srv cInfo (Just e2eSndParams) subMode
 joinConnSrvAsync _c _userId _connId _enableNtfs (CRContactUri _) _cInfo _subMode _pqSupport _srv = do
   throwE $ CMD PROHIBITED "joinConnSrvAsync"
 
@@ -923,7 +925,7 @@ createReplyQueue c ConnData {userId, connId, enableNtfs} SndQueue {smpClientVers
   lift . when (subMode == SMSubscribe) $ addNewQueueSubscription c rq' tSess sessId
   when enableNtfs $ do
     ns <- asks ntfSupervisor
-    atomically $ sendNtfSubCommand ns (connId, NSCCreate)
+    atomically $ sendNtfSubCommand ns (NSCCreate, [connId])
   pure qInfo
 
 -- | Approve confirmation (LET command) in Reader monad
@@ -937,13 +939,13 @@ allowConnection' c connId confId ownConnInfo = withConnLock c connId "allowConne
     _ -> throwE $ CMD PROHIBITED "allowConnection"
 
 -- | Accept contact (ACPT command) in Reader monad
-acceptContact' :: AgentClient -> ConnId -> Bool -> InvitationId -> ConnInfo -> PQSupport -> SubscriptionMode -> AM (ConnId, SndQueueSecured)
+acceptContact' :: AgentClient -> ConnId -> Bool -> InvitationId -> ConnInfo -> PQSupport -> SubscriptionMode -> AM SndQueueSecured
 acceptContact' c connId enableNtfs invId ownConnInfo pqSupport subMode = withConnLock c connId "acceptContact" $ do
   Invitation {contactConnId, connReq} <- withStore c (`getInvitation` invId)
   withStore c (`getConn` contactConnId) >>= \case
     SomeConn _ (ContactConnection ConnData {userId} _) -> do
       withStore' c $ \db -> acceptInvitation db invId ownConnInfo
-      joinConn c userId connId False enableNtfs connReq ownConnInfo pqSupport subMode `catchAgentError` \err -> do
+      joinConn c userId connId enableNtfs connReq ownConnInfo pqSupport subMode `catchAgentError` \err -> do
         withStore' c (`unacceptInvitation` invId)
         throwE err
     _ -> throwE $ CMD PROHIBITED "acceptContact"
@@ -1013,14 +1015,13 @@ subscribeConnections' c connIds = do
         order _ = 4
     sendNtfCreate :: NtfSupervisor -> Map ConnId (Either AgentErrorType ()) -> Map ConnId SomeConn -> AM' ()
     sendNtfCreate ns rcvRs cs = do
-      -- TODO this needs to be batched end to end.
-      -- Currently, the only change is to ignore failed subscriptions.
       let oks = M.keysSet $ M.filter (either temporaryAgentError $ const True) rcvRs
-      forM_ (M.restrictKeys cs oks) $ \case
-        SomeConn _ conn -> do
-          let cmd = if enableNtfs $ toConnData conn then NSCCreate else NSCDelete
-              ConnData {connId} = toConnData conn
-          atomically $ writeTBQueue (ntfSubQ ns) (connId, cmd)
+          cs' = M.restrictKeys cs oks
+          (csCreate, csDelete) = M.partition (\(SomeConn _ conn) -> enableNtfs $ toConnData conn) cs'
+      sendNtfCmd NSCCreate csCreate
+      sendNtfCmd NSCSmpDelete csDelete
+      where
+        sendNtfCmd cmd cs' = forM_ (L.nonEmpty $ M.keys cs') $ \cids -> atomically $ writeTBQueue (ntfSubQ ns) (cmd, cids)
     resumeDelivery :: Map ConnId SomeConn -> AM ()
     resumeDelivery conns = do
       conns' <- M.restrictKeys conns . S.fromList <$> withStore' c getConnectionsForDelivery
@@ -1049,29 +1050,57 @@ resubscribeConnections' c connIds = do
   -- union is left-biased, so results returned by subscribeConnections' take precedence
   (`M.union` r) <$> subscribeConnections' c connIds'
 
-getConnectionMessage' :: AgentClient -> ConnId -> AM (Maybe SMPMsgMeta)
-getConnectionMessage' c connId = do
-  whenM (atomically $ hasActiveSubscription c connId) . throwE $ CMD PROHIBITED "getConnectionMessage: subscribed"
-  SomeConn _ conn <- withStore c (`getConn` connId)
-  case conn of
-    DuplexConnection _ (rq :| _) _ -> getQueueMessage c rq
-    RcvConnection _ rq -> getQueueMessage c rq
-    ContactConnection _ rq -> getQueueMessage c rq
-    SndConnection _ _ -> throwE $ CONN SIMPLEX
-    NewConnection _ -> throwE $ CMD PROHIBITED "getConnectionMessage: NewConnection"
+getConnectionMessages' :: AgentClient -> NonEmpty ConnId -> AM' (NonEmpty (Maybe SMPMsgMeta))
+getConnectionMessages' c = mapM getMsg
+  where
+    getMsg :: ConnId -> AM' (Maybe SMPMsgMeta)
+    getMsg connId =
+      getConnectionMessage connId `catchAgentError'` \e -> do
+        logError $ "Error loading message: " <> tshow e
+        pure Nothing
+    getConnectionMessage :: ConnId -> AM (Maybe SMPMsgMeta)
+    getConnectionMessage connId = do
+      whenM (atomically $ hasActiveSubscription c connId) . throwE $ CMD PROHIBITED "getConnectionMessage: subscribed"
+      SomeConn _ conn <- withStore c (`getConn` connId)
+      case conn of
+        DuplexConnection _ (rq :| _) _ -> getQueueMessage c rq
+        RcvConnection _ rq -> getQueueMessage c rq
+        ContactConnection _ rq -> getQueueMessage c rq
+        SndConnection _ _ -> throwE $ CONN SIMPLEX
+        NewConnection _ -> throwE $ CMD PROHIBITED "getConnectionMessage: NewConnection"
 
-getNotificationMessage' :: AgentClient -> C.CbNonce -> ByteString -> AM (NotificationInfo, Maybe SMPMsgMeta)
-getNotificationMessage' c nonce encNtfInfo = do
+getNotificationConns' :: AgentClient -> C.CbNonce -> ByteString -> AM (NonEmpty NotificationInfo)
+getNotificationConns' c nonce encNtfInfo =
   withStore' c getActiveNtfToken >>= \case
     Just NtfToken {ntfDhSecret = Just dhSecret} -> do
-      ntfData <- agentCbDecrypt dhSecret nonce encNtfInfo
+      ntfData <- liftEither $ agentCbDecrypt dhSecret nonce encNtfInfo
       pnMsgs <- liftEither (parse pnMessagesP (INTERNAL "error parsing PNMessageData") ntfData)
-      let PNMessageData {smpQueue, ntfTs, nmsgNonce, encNMsgMeta} = L.last pnMsgs
-      (ntfConnId, rcvNtfDhSecret) <- withStore c (`getNtfRcvQueue` smpQueue)
-      ntfMsgMeta <- (eitherToMaybe . smpDecode <$> agentCbDecrypt rcvNtfDhSecret nmsgNonce encNMsgMeta) `catchAgentError` \_ -> pure Nothing
-      msgMeta <- getConnectionMessage' c ntfConnId
-      pure (NotificationInfo {ntfConnId, ntfTs, ntfMsgMeta}, msgMeta)
-    _ -> throwE $ CMD PROHIBITED "getNotificationMessage"
+      let (initNtfs, lastNtf) = (L.init pnMsgs, L.last pnMsgs)
+      rs <-
+        lift $ withStoreBatch c $ \db ->
+          let initNtfInfos = map (getInitNtfInfo db) initNtfs
+              lastNtfInfo = Just . fst <$$> getNtfInfo db lastNtf
+           in initNtfInfos <> [lastNtfInfo]
+      let (errs, ntfInfos_) = partitionEithers rs
+      logError $ "Error(s) loading notifications: " <> tshow errs
+      case L.nonEmpty $ catMaybes ntfInfos_ of
+        Just r -> pure r
+        Nothing -> throwE $ INTERNAL "getNotificationConns: couldn't get conn info"
+    _ -> throwE $ CMD PROHIBITED "getNotificationConns"
+  where
+    getNtfInfo :: DB.Connection -> PNMessageData -> IO (Either AgentErrorType (NotificationInfo, Maybe UTCTime))
+    getNtfInfo db PNMessageData {smpQueue, ntfTs, nmsgNonce, encNMsgMeta} = runExceptT $ do
+      (ntfConnId, rcvNtfDhSecret, lastBrokerTs_) <- liftError' storeError $ getNtfRcvQueue db smpQueue
+      let ntfMsgMeta = eitherToMaybe $ smpDecode =<< first show (C.cbDecrypt rcvNtfDhSecret nmsgNonce encNMsgMeta)
+          ntfInfo = NotificationInfo {ntfConnId, ntfTs, ntfMsgMeta}
+      pure (ntfInfo, lastBrokerTs_)
+    getInitNtfInfo :: DB.Connection -> PNMessageData -> IO (Either AgentErrorType (Maybe NotificationInfo))
+    getInitNtfInfo db msgData = runExceptT $ do
+      (nftInfo, lastBrokerTs_) <- ExceptT $ getNtfInfo db msgData
+      pure $ case (ntfMsgMeta nftInfo, lastBrokerTs_) of
+        (Just SMP.NMsgMeta {msgTs}, Just lastBrokerTs)
+          | systemToUTCTime msgTs > lastBrokerTs -> Just nftInfo
+        _ -> Nothing
 
 -- | Send message to the connection (SEND command) in Reader monad
 sendMessage' :: AgentClient -> ConnId -> PQEncryption -> MsgFlags -> MsgBody -> AM (AgentMsgId, PQEncryption)
@@ -1169,14 +1198,13 @@ runCommandProcessing c@AgentClient {subQ} connId server_ Worker {doWork} = do
     processCmd ri PendingCommand {cmdId, corrId, userId, command} pendingCmds = case command of
       AClientCommand cmd -> case cmd of
         NEW enableNtfs (ACM cMode) pqEnc subMode -> noServer $ do
-          usedSrvs <- newTVarIO ([] :: [SMPServer])
-          tryCommand . withNextSrv c userId usedSrvs [] $ \srv -> do
-            (_, cReq) <- newRcvConnSrv c userId connId enableNtfs cMode Nothing pqEnc subMode srv
+          triedHosts <- newTVarIO S.empty
+          tryCommand . withNextSrv c userId storageSrvs triedHosts [] $ \srv -> do
+            cReq <- newRcvConnSrv c userId connId enableNtfs cMode Nothing pqEnc subMode srv
             notify $ INV (ACR cMode cReq)
         JOIN enableNtfs (ACR _ cReq@(CRInvitationUri ConnReqUriData {crSmpQueues = q :| _} _)) pqEnc subMode connInfo -> noServer $ do
-          let initUsed = [qServer q]
-          usedSrvs <- newTVarIO initUsed
-          tryCommand . withNextSrv c userId usedSrvs initUsed $ \srv -> do
+          triedHosts <- newTVarIO S.empty
+          tryCommand . withNextSrv c userId storageSrvs triedHosts [qServer q] $ \srv -> do
             sqSecured <- joinConnSrvAsync c userId connId enableNtfs cReq connInfo pqEnc subMode srv
             notify $ JOINED sqSecured
         LET confId ownCInfo -> withServer' . tryCommand $ allowConnection' c connId confId ownCInfo >> notify OK
@@ -1259,7 +1287,7 @@ runCommandProcessing c@AgentClient {subQ} connId server_ Worker {doWork} = do
                     withStore' c $ \db -> deleteConnRcvQueue db rq'
                     when (enableNtfs cData) $ do
                       ns <- asks ntfSupervisor
-                      atomically $ sendNtfSubCommand ns (connId, NSCCreate)
+                      atomically $ sendNtfSubCommand ns (NSCCreate, [connId])
                     let conn' = DuplexConnection cData (rq'' :| rqs') sqs
                     notify $ SWITCH QDRcv SPCompleted $ connectionStats conn'
               _ -> internalErr "ICQDelete: cannot delete the only queue in connection"
@@ -1624,8 +1652,8 @@ switchDuplexConnection c (DuplexConnection cData@ConnData {connId, userId} rqs s
   checkRQSwchStatus rq RSSwitchStarted
   clientVRange <- asks $ smpClientVRange . config
   -- try to get the server that is different from all queues, or at least from the primary rcv queue
-  srvAuth@(ProtoServerWithAuth srv _) <- getNextServer c userId $ map qServer (L.toList rqs) <> map qServer (L.toList sqs)
-  srv' <- if srv == server then getNextServer c userId [server] else pure srvAuth
+  srvAuth@(ProtoServerWithAuth srv _) <- getNextSMPServer c userId $ map qServer (L.toList rqs) <> map qServer (L.toList sqs)
+  srv' <- if srv == server then getNextSMPServer c userId [server] else pure srvAuth
   (q, qUri, tSess, sessId) <- newRcvQueue c userId connId srv' clientVRange SMSubscribe False
   let rq' = (q :: NewRcvQueue) {primary = True, dbReplaceQueueId = Just dbQueueId}
   rq'' <- withStore c $ \db -> addConnRcvQueue db connId rq'
@@ -1681,14 +1709,21 @@ synchronizeRatchet' c connId pqSupport' force = withConnLock c connId "synchroni
     _ -> throwE $ CMD PROHIBITED "synchronizeRatchet: not duplex"
 
 ackQueueMessage :: AgentClient -> RcvQueue -> SMP.MsgId -> AM ()
-ackQueueMessage c rq@RcvQueue {userId, server} srvMsgId = do
+ackQueueMessage c rq@RcvQueue {userId, connId, server} srvMsgId = do
   atomically $ incSMPServerStat c userId server ackAttempts
   tryAgentError (sendAck c rq srvMsgId) >>= \case
-    Right _ -> atomically $ incSMPServerStat c userId server ackMsgs
-    Left (SMP _ SMP.NO_MSG) -> atomically $ incSMPServerStat c userId server ackNoMsgErrs
+    Right _ -> sendMsgNtf ackMsgs
+    Left (SMP _ SMP.NO_MSG) -> sendMsgNtf ackNoMsgErrs
     Left e -> do
       unless (temporaryOrHostError e) $ atomically $ incSMPServerStat c userId server ackOtherErrs
       throwE e
+  where
+    sendMsgNtf stat = do
+      atomically $ incSMPServerStat c userId server stat
+      whenM (liftIO $ hasGetLock c rq) $ do
+        atomically $ releaseGetLock c rq
+        brokerTs_ <- eitherToMaybe <$> tryAgentError (withStore c $ \db -> getRcvMsgBrokerTs db connId srvMsgId)
+        atomically $ writeTBQueue (subQ c) ("", connId, AEvt SAEConn $ MSGNTF srvMsgId brokerTs_)
 
 -- | Suspend SMP agent connection (OFF command) in Reader monad
 suspendConnection' :: AgentClient -> ConnId -> AM ()
@@ -1716,12 +1751,6 @@ connRcvQueues = \case
   SndConnection _ _ -> []
   NewConnection _ -> []
 
-disableConn :: AgentClient -> ConnId -> AM' ()
-disableConn c connId = do
-  atomically $ removeSubscription c connId
-  ns <- asks ntfSupervisor
-  atomically $ writeTBQueue (ntfSubQ ns) (connId, NSCDelete)
-
 -- Unlike deleteConnectionsAsync, this function does not mark connections as deleted in case of deletion failure.
 deleteConnections' :: AgentClient -> [ConnId] -> AM (Map ConnId (Either AgentErrorType ()))
 deleteConnections' = deleteConnections_ getConns False False
@@ -1748,7 +1777,7 @@ prepareDeleteConnections_ getConnections c waitDelivery connIds = do
       (delRs, rcvQs) = M.mapEither rcvQueues cs
       rqs = concat $ M.elems rcvQs
       connIds' = M.keys rcvQs
-  lift . forM_ connIds' $ disableConn c
+  lift $ forM_ (L.nonEmpty connIds') unsubConnIds
   -- ! delRs is not used to notify about the result in any of the calling functions,
   -- ! it is only used to check results count in deleteConnections_;
   -- ! if it was used to notify about the result, it might be necessary to differentiate
@@ -1762,6 +1791,12 @@ prepareDeleteConnections_ getConnections c waitDelivery connIds = do
     rcvQueues (SomeConn _ conn) = case connRcvQueues conn of
       [] -> Left $ Right ()
       rqs -> Right rqs
+    unsubConnIds :: NonEmpty ConnId -> AM' ()
+    unsubConnIds connIds' = do
+      forM_ connIds' $ \connId ->
+        atomically $ removeSubscription c connId
+      ns <- asks ntfSupervisor
+      atomically $ writeTBQueue (ntfSubQ ns) (NSCDeleteSub, connIds')
     notify = atomically . writeTBQueue (subQ c)
 
 deleteConnQueues :: AgentClient -> Bool -> Bool -> [RcvQueue] -> AM' (Map ConnId (Either AgentErrorType ()))
@@ -1902,16 +1937,12 @@ registerNtfToken' c suppliedDeviceToken suppliedNtfMode =
                 cron <- asks $ ntfCron . config
                 agentNtfEnableCron c tknId tkn cron
                 when (suppliedNtfMode == NMInstant) $ initializeNtfSubs c
-                when (suppliedNtfMode == NMPeriodic && savedNtfMode == NMInstant) $ deleteNtfSubs c NSCDelete
+                when (suppliedNtfMode == NMPeriodic && savedNtfMode == NMInstant) $ deleteNtfSubs c NSCSmpDelete
               -- possible improvement: get updated token status from the server, or maybe TCRON could return the current status
               pure ntfTknStatus
           | otherwise -> replaceToken tknId
-        (Just tknId, Just NTADelete) -> do
-          agentNtfDeleteToken c tknId tkn
-          withStore' c (`removeNtfToken` tkn)
-          ns <- asks ntfSupervisor
-          atomically $ nsRemoveNtfToken ns
-          pure NTExpired
+        -- deprecated
+        (Just _tknId, Just NTADelete) -> deleteToken c tkn $> NTExpired
         _ -> pure ntfTknStatus
       withStore' c $ \db -> updateNtfMode db tkn suppliedNtfMode
       pure status
@@ -1984,8 +2015,8 @@ deleteNtfToken' :: AgentClient -> DeviceToken -> AM ()
 deleteNtfToken' c deviceToken =
   withStore' c getSavedNtfToken >>= \case
     Just tkn@NtfToken {deviceToken = savedDeviceToken} -> do
-      when (deviceToken /= savedDeviceToken) . throwE $ CMD PROHIBITED "deleteNtfToken: different token"
-      deleteToken_ c tkn
+      when (deviceToken /= savedDeviceToken) $ logWarn "deleteNtfToken: different token"
+      deleteToken c tkn
       deleteNtfSubs c NSCSmpDelete
     _ -> throwE $ CMD PROHIBITED "deleteNtfToken: no token"
 
@@ -2017,21 +2048,8 @@ toggleConnectionNtfs' c connId enable = do
       | otherwise = do
           withStore' c $ \db -> setConnectionNtfs db connId enable
           ns <- asks ntfSupervisor
-          let cmd = if enable then NSCCreate else NSCDelete
-          atomically $ sendNtfSubCommand ns (connId, cmd)
-
-deleteToken_ :: AgentClient -> NtfToken -> AM ()
-deleteToken_ c tkn@NtfToken {ntfTokenId, ntfTknStatus} = do
-  ns <- asks ntfSupervisor
-  forM_ ntfTokenId $ \tknId -> do
-    let ntfTknAction = Just NTADelete
-    withStore' c $ \db -> updateNtfToken db tkn ntfTknStatus ntfTknAction
-    atomically $ nsUpdateToken ns tkn {ntfTknStatus, ntfTknAction}
-    agentNtfDeleteToken c tknId tkn `catchAgentError` \case
-      NTF _ AUTH -> pure ()
-      e -> throwE e
-  withStore' c $ \db -> removeNtfToken db tkn
-  atomically $ nsRemoveNtfToken ns
+          let cmd = if enable then NSCCreate else NSCSmpDelete
+          atomically $ sendNtfSubCommand ns (cmd, [connId])
 
 withToken :: AgentClient -> NtfToken -> Maybe (NtfTknStatus, NtfTknAction) -> (NtfTknStatus, Maybe NtfTknAction) -> AM a -> AM NtfTknStatus
 withToken c tkn@NtfToken {deviceToken, ntfMode} from_ (toStatus, toAction_) f = do
@@ -2065,13 +2083,24 @@ deleteNtfSubs c deleteCmd = do
 sendNtfConnCommands :: AgentClient -> NtfSupervisorCommand -> AM ()
 sendNtfConnCommands c cmd = do
   ns <- asks ntfSupervisor
-  connIds <- liftIO $ getSubscriptions c
-  forM_ connIds $ \connId -> do
-    withStore' c (`getConnData` connId) >>= \case
-      Just (ConnData {enableNtfs}, _) ->
-        when enableNtfs . atomically $ writeTBQueue (ntfSubQ ns) (connId, cmd)
-      _ ->
-        atomically $ writeTBQueue (subQ c) ("", connId, AEvt SAEConn $ ERR $ INTERNAL "no connection data")
+  connIds <- liftIO $ S.toList <$> getSubscriptions c
+  rs <- lift $ withStoreBatch' c (\db -> map (getConnData db) connIds)
+  let (connIds', cErrs) = enabledNtfConns (zip connIds rs)
+  forM_ (L.nonEmpty connIds') $ \connIds'' ->
+    atomically $ writeTBQueue (ntfSubQ ns) (cmd, connIds'')
+  unless (null cErrs) $ atomically $ writeTBQueue (subQ c) ("", "", AEvt SAENone $ ERRS cErrs)
+  where
+    enabledNtfConns :: [(ConnId, Either AgentErrorType (Maybe (ConnData, ConnectionMode)))] -> ([ConnId], [(ConnId, AgentErrorType)])
+    enabledNtfConns = foldr addEnabledConn ([], [])
+      where
+        addEnabledConn ::
+          (ConnId, Either AgentErrorType (Maybe (ConnData, ConnectionMode))) ->
+          ([ConnId], [(ConnId, AgentErrorType)]) ->
+          ([ConnId], [(ConnId, AgentErrorType)])
+        addEnabledConn cData_ (cIds, errs) = case cData_ of
+          (_, Right (Just (ConnData {connId, enableNtfs}, _))) -> if enableNtfs then (connId : cIds, errs) else (cIds, errs)
+          (connId, Right Nothing) -> (cIds, (connId, INTERNAL "no connection data") : errs)
+          (connId, Left e) -> (cIds, (connId, e) : errs)
 
 setNtfServers :: AgentClient -> [NtfServer] -> IO ()
 setNtfServers c = atomically . writeTVar (ntfServers c)
@@ -2132,9 +2161,13 @@ debugAgentLocks AgentClient {connLocks = cs, invLocks = is, deleteLock = d} = do
     getLocks ls = atomically $ M.mapKeys (B.unpack . strEncode) . M.mapMaybe id <$> (mapM tryReadTMVar =<< readTVar ls)
 
 getSMPServer :: AgentClient -> UserId -> AM SMPServerWithAuth
-getSMPServer c userId = withUserServers c userId pickServer
+getSMPServer c userId = getNextSMPServer c userId []
 {-# INLINE getSMPServer #-}
 
+getNextSMPServer :: AgentClient -> UserId -> [SMPServer] -> AM SMPServerWithAuth
+getNextSMPServer c userId = getNextServer c userId storageSrvs
+{-# INLINE getNextSMPServer #-}
+
 subscriber :: AgentClient -> AM' ()
 subscriber c@AgentClient {msgQ} = forever $ do
   t <- atomically $ readTBQueue msgQ
@@ -2152,6 +2185,7 @@ cleanupManager c@AgentClient {subQ} = do
     run ERR $ withStore' c (`deleteRcvMsgHashesExpired` ttl)
     run ERR $ withStore' c (`deleteSndMsgsExpired` ttl)
     run ERR $ withStore' c (`deleteRatchetKeyHashesExpired` ttl)
+    run ERR $ withStore' c (`deleteExpiredNtfTokensToDelete` ttl)
     run RFERR deleteRcvFilesExpired
     run RFERR deleteRcvFilesDeleted
     run RFERR deleteRcvFilesTmpPaths
@@ -2290,12 +2324,9 @@ processSMPTransmissions c@AgentClient {subQ} (tSess@(userId, srv, _), _v, sessId
             atomically $ incSMPServerStat c userId srv recvMsgs
             void . handleNotifyAck $ do
               msg' <- decryptSMPMessage rq msg
-              ack' <- handleNotifyAck $ case msg' of
+              handleNotifyAck $ case msg' of
                 SMP.ClientRcvMsgBody {msgTs = srvTs, msgFlags, msgBody} -> processClientMsg srvTs msgFlags msgBody
                 SMP.ClientRcvMsgQuota {} -> queueDrained >> ack
-              whenM (liftIO $ hasGetLock c rq) $
-                notify (MSGNTF $ SMP.rcvMessageMeta srvMsgId msg')
-              pure ack'
             where
               queueDrained = case conn of
                 DuplexConnection _ _ sqs -> void $ enqueueMessages c cData sqs SMP.noMsgFlags $ A_QCONT (sndAddress rq)
@@ -2450,17 +2481,14 @@ processSMPTransmissions c@AgentClient {subQ} (tSess@(userId, srv, _), _v, sessId
               handleNotifyAck :: AM ACKd -> AM ACKd
               handleNotifyAck m = m `catchAgentError` \e -> notify (ERR e) >> ack
           SMP.END ->
-            atomically (TM.lookup tSess (smpClients c) $>>= (tryReadTMVar . sessionVar) >>= processEND)
+            atomically (ifM (activeClientSession c tSess sessId) (removeSubscription c connId $> True) (pure False))
               >>= notifyEnd
             where
-              processEND = \case
-                Just (Right clnt)
-                  | sessId == sessionId (thParams $ connectedClient clnt) ->
-                      removeSubscription c connId $> True
-                _ -> pure False
               notifyEnd removed
                 | removed = notify END >> logServer "<--" c srv rId "END"
                 | otherwise = logServer "<--" c srv rId "END from disconnected client - ignored"
+          -- Possibly, we need to add some flag to connection that it was deleted
+          SMP.DELD -> atomically (removeSubscription c connId) >> notify DELD
           SMP.ERR e -> notify $ ERR $ SMP (B.unpack $ strEncode srv) e
           r -> unexpected r
         where
@@ -2485,7 +2513,7 @@ processSMPTransmissions c@AgentClient {subQ} (tSess@(userId, srv, _), _v, sessId
 
           decryptClientMessage :: C.DhSecretX25519 -> SMP.ClientMsgEnvelope -> AM (SMP.PrivHeader, AgentMsgEnvelope)
           decryptClientMessage e2eDh SMP.ClientMsgEnvelope {cmNonce, cmEncBody} = do
-            clientMsg <- agentCbDecrypt e2eDh cmNonce cmEncBody
+            clientMsg <- liftEither $ agentCbDecrypt e2eDh cmNonce cmEncBody
             SMP.ClientMessage privHeader clientBody <- parseMessage clientMsg
             agentEnvelope <- parseMessage clientBody
             -- Version check is removed here, because when connecting via v1 contact address the agent still sends v2 message,

src/Simplex/Messaging/Agent/Client.hs

@@ -30,6 +30,7 @@ module Simplex.Messaging.Agent.Client
     withConnLocks,
     withInvLock,
     withLockMap,
+    getMapLock,
     ipAddressProtected,
     closeAgentClient,
     closeProtocolServerClients,
@@ -56,8 +57,10 @@ module Simplex.Messaging.Agent.Client
     secureQueue,
     secureSndQueue,
     enableQueueNotifications,
+    EnableQueueNtfReq (..),
     enableQueuesNtfs,
     disableQueueNotifications,
+    DisableQueueNtfReq,
     disableQueuesNtfs,
     sendAgentMessage,
     getQueueInfo,
@@ -68,7 +71,9 @@ module Simplex.Messaging.Agent.Client
     agentNtfDeleteToken,
     agentNtfEnableCron,
     agentNtfCreateSubscription,
+    agentNtfCreateSubscriptions,
     agentNtfCheckSubscription,
+    agentNtfCheckSubscriptions,
     agentNtfDeleteSubscription,
     agentXFTPDownloadChunk,
     agentXFTPNewChunk,
@@ -88,6 +93,7 @@ module Simplex.Messaging.Agent.Client
     hasActiveSubscription,
     hasPendingSubscription,
     hasGetLock,
+    releaseGetLock,
     activeClientSession,
     agentClientStore,
     agentDRG,
@@ -115,6 +121,7 @@ module Simplex.Messaging.Agent.Client
     hasWorkToDo,
     hasWorkToDo',
     withWork,
+    withWorkItems,
     agentOperations,
     agentOperationBracket,
     waitUntilActive,
@@ -142,7 +149,6 @@ module Simplex.Messaging.Agent.Client
     userServers,
     pickServer,
     getNextServer,
-    withUserServers,
     withNextSrv,
     incSMPServerStat,
     incSMPServerStat',
@@ -150,6 +156,7 @@ module Simplex.Messaging.Agent.Client
     incXFTPServerStat',
     incXFTPServerSizeStat,
     incNtfServerStat,
+    incNtfServerStat',
     AgentWorkersDetails (..),
     getAgentWorkersDetails,
     AgentWorkersSummary (..),
@@ -185,12 +192,12 @@ import qualified Data.ByteString.Char8 as B
 import Data.Either (isRight, partitionEithers)
 import Data.Functor (($>))
 import Data.Int (Int64)
-import Data.List (deleteFirstsBy, foldl', partition, (\\))
+import Data.List (find, foldl', partition)
 import Data.List.NonEmpty (NonEmpty (..), (<|))
 import qualified Data.List.NonEmpty as L
 import Data.Map.Strict (Map)
 import qualified Data.Map.Strict as M
-import Data.Maybe (catMaybes, fromMaybe, isJust, isNothing, listToMaybe)
+import Data.Maybe (catMaybes, fromMaybe, isJust, isNothing, listToMaybe, mapMaybe)
 import Data.Set (Set)
 import qualified Data.Set as S
 import Data.Text (Text)
@@ -257,7 +264,6 @@ import Simplex.Messaging.Protocol
     XFTPServer,
     XFTPServerWithAuth,
     pattern NoEntity,
-    sameSrvAddr',
   )
 import qualified Simplex.Messaging.Protocol as SMP
 import Simplex.Messaging.Server.QueueStore.QueueInfo
@@ -333,6 +339,7 @@ data AgentClient = AgentClient
     smpSubWorkers :: TMap SMPTransportSession (SessionVar (Async ())),
     clientId :: Int,
     agentEnv :: Env,
+    proxySessTs :: TVar UTCTime,
     smpServersStats :: TMap (UserId, SMPServer) AgentSMPServerStats,
     xftpServersStats :: TMap (UserId, XFTPServer) AgentXFTPServerStats,
     ntfServersStats :: TMap (UserId, NtfServer) AgentNtfServerStats,
@@ -461,6 +468,7 @@ newAgentClient :: Int -> InitialAgentServers -> UTCTime -> Env -> IO AgentClient
 newAgentClient clientId InitialAgentServers {smp, ntf, xftp, netCfg} currentTs agentEnv = do
   let cfg = config agentEnv
       qSize = tbqSize cfg
+  proxySessTs <- newTVarIO =<< getCurrentTime
   acThread <- newTVarIO Nothing
   active <- newTVarIO True
   subQ <- newTBQueueIO qSize
@@ -491,7 +499,7 @@ newAgentClient clientId InitialAgentServers {smp, ntf, xftp, netCfg} currentTs a
   getMsgLocks <- TM.emptyIO
   connLocks <- TM.emptyIO
   invLocks <- TM.emptyIO
-  deleteLock <- atomically createLock
+  deleteLock <- createLockIO
   smpSubWorkers <- TM.emptyIO
   smpServersStats <- TM.emptyIO
   xftpServersStats <- TM.emptyIO
@@ -533,6 +541,7 @@ newAgentClient clientId InitialAgentServers {smp, ntf, xftp, netCfg} currentTs a
         smpSubWorkers,
         clientId,
         agentEnv,
+        proxySessTs,
         smpServersStats,
         xftpServersStats,
         ntfServersStats,
@@ -608,7 +617,7 @@ getSMPServerClient c@AgentClient {active, smpClients, workerSeq} tSess = do
 getSMPProxyClient :: AgentClient -> Maybe SMPServerWithAuth -> SMPTransportSession -> AM (SMPConnectedClient, Either AgentErrorType ProxiedRelay)
 getSMPProxyClient c@AgentClient {active, smpClients, smpProxiedRelays, workerSeq} proxySrv_ destSess@(userId, destSrv, qId) = do
   unlessM (readTVarIO active) $ throwE INACTIVE
-  proxySrv <- maybe (getNextServer c userId [destSrv]) pure proxySrv_
+  proxySrv <- maybe (getNextServer c userId proxySrvs [destSrv]) pure proxySrv_
   ts <- liftIO getCurrentTime
   atomically (getClientVar proxySrv ts) >>= \(tSess, auth, v) ->
     either (newProxyClient tSess auth ts) (waitForProxyClient tSess auth) v
@@ -657,7 +666,7 @@ getSMPProxyClient c@AgentClient {active, smpClients, smpProxiedRelays, workerSeq
         Nothing -> Left $ BROKER (B.unpack $ strEncode srv) TIMEOUT
 
 smpConnectClient :: AgentClient -> SMPTransportSession -> TMap SMPServer ProxiedRelayVar -> SMPClientVar -> AM SMPConnectedClient
-smpConnectClient c@AgentClient {smpClients, msgQ} tSess@(_, srv, _) prs v =
+smpConnectClient c@AgentClient {smpClients, msgQ, proxySessTs} tSess@(_, srv, _) prs v =
   newProtocolClient c tSess smpClients connectClient v
     `catchAgentError` \e -> lift (resubscribeSMPSession c tSess) >> throwE e
   where
@@ -667,7 +676,8 @@ smpConnectClient c@AgentClient {smpClients, msgQ} tSess@(_, srv, _) prs v =
       g <- asks random
       env <- ask
       liftError (protocolClientError SMP $ B.unpack $ strEncode srv) $ do
-        smp <- ExceptT $ getProtocolClient g tSess cfg (Just msgQ) $ smpClientDisconnected c tSess env v' prs
+        ts <- readTVarIO proxySessTs
+        smp <- ExceptT $ getProtocolClient g tSess cfg (Just msgQ) ts $ smpClientDisconnected c tSess env v' prs
         pure SMPConnectedClient {connectedClient = smp, proxiedRelays = prs}
 
 smpClientDisconnected :: AgentClient -> SMPTransportSession -> Env -> SMPClientVar -> TMap SMPServer ProxiedRelayVar -> SMPClient -> IO ()
@@ -756,7 +766,7 @@ reconnectSMPClient c tSess@(_, srv, _) qs = handleNotify $ do
     notifySub connId cmd = atomically $ writeTBQueue (subQ c) ("", connId, AEvt (sAEntity @e) cmd)
 
 getNtfServerClient :: AgentClient -> NtfTransportSession -> AM NtfClient
-getNtfServerClient c@AgentClient {active, ntfClients, workerSeq} tSess@(_, srv, _) = do
+getNtfServerClient c@AgentClient {active, ntfClients, workerSeq, proxySessTs} tSess@(_, srv, _) = do
   unlessM (readTVarIO active) $ throwE INACTIVE
   ts <- liftIO getCurrentTime
   atomically (getSessVar workerSeq tSess ntfClients ts)
@@ -768,8 +778,9 @@ getNtfServerClient c@AgentClient {active, ntfClients, workerSeq} tSess@(_, srv,
     connectClient v = do
       cfg <- lift $ getClientConfig c ntfCfg
       g <- asks random
+      ts <- readTVarIO proxySessTs
       liftError' (protocolClientError NTF $ B.unpack $ strEncode srv) $
-        getProtocolClient g tSess cfg Nothing $
+        getProtocolClient g tSess cfg Nothing ts $
           clientDisconnected v
 
     clientDisconnected :: NtfClientVar -> NtfClient -> IO ()
@@ -779,7 +790,7 @@ getNtfServerClient c@AgentClient {active, ntfClients, workerSeq} tSess@(_, srv,
       logInfo . decodeUtf8 $ "Agent disconnected from " <> showServer srv
 
 getXFTPServerClient :: AgentClient -> XFTPTransportSession -> AM XFTPClient
-getXFTPServerClient c@AgentClient {active, xftpClients, workerSeq} tSess@(_, srv, _) = do
+getXFTPServerClient c@AgentClient {active, xftpClients, workerSeq, proxySessTs} tSess@(_, srv, _) = do
   unlessM (readTVarIO active) $ throwE INACTIVE
   ts <- liftIO getCurrentTime
   atomically (getSessVar workerSeq tSess xftpClients ts)
@@ -791,8 +802,9 @@ getXFTPServerClient c@AgentClient {active, xftpClients, workerSeq} tSess@(_, srv
     connectClient v = do
       cfg <- asks $ xftpCfg . config
       xftpNetworkConfig <- getNetworkConfig c
+      ts <- readTVarIO proxySessTs
       liftError' (protocolClientError XFTP $ B.unpack $ strEncode srv) $
-        X.getXFTPClient tSess cfg {xftpNetworkConfig} $
+        X.getXFTPClient tSess cfg {xftpNetworkConfig} ts $
           clientDisconnected v
 
     clientDisconnected :: XFTPClientVar -> XFTPClient -> IO ()
@@ -1060,7 +1072,7 @@ sendOrProxySMPCommand ::
   (SMPClient -> ProxiedRelay -> ExceptT SMPClientError IO (Either ProxyClientError ())) ->
   (SMPClient -> ExceptT SMPClientError IO ()) ->
   AM (Maybe SMPServer)
-sendOrProxySMPCommand c userId destSrv connId cmdStr senderId sendCmdViaProxy sendCmdDirectly = do
+sendOrProxySMPCommand c userId destSrv@ProtocolServer {host = destHosts} connId cmdStr senderId sendCmdViaProxy sendCmdDirectly = do
   tSess <- mkTransportSession c userId destSrv connId
   ifM shouldUseProxy (sendViaProxy Nothing tSess) (sendDirectly tSess $> Nothing)
   where
@@ -1079,7 +1091,7 @@ sendOrProxySMPCommand c userId destSrv connId cmdStr senderId sendCmdViaProxy se
         SPFAllow -> True
         SPFAllowProtected -> ipAddressProtected cfg destSrv
         SPFProhibit -> False
-    unknownServer = liftIO $ maybe True (notElem destSrv . knownSrvs) <$> TM.lookupIO userId (smpServers c)
+    unknownServer = liftIO $ maybe True (\srvs -> all (`S.notMember` knownHosts srvs) destHosts) <$> TM.lookupIO userId (smpServers c)
     sendViaProxy :: Maybe SMPServerWithAuth -> SMPTransportSession -> AM (Maybe SMPServer)
     sendViaProxy proxySrv_ destSess@(_, _, connId_) = do
       r <- tryAgentError . withProxySession c proxySrv_ destSess senderId ("PFWD " <> cmdStr) $ \(SMPConnectedClient smp _, proxySess@ProxiedRelay {prBasicAuth}) -> do
@@ -1199,7 +1211,8 @@ runSMPServerTest c userId (ProtoServerWithAuth srv auth) = do
   g <- asks random
   liftIO $ do
     let tSess = (userId, srv, Nothing)
-    getProtocolClient g tSess cfg Nothing (\_ -> pure ()) >>= \case
+    ts <- readTVarIO $ proxySessTs c
+    getProtocolClient g tSess cfg Nothing ts (\_ -> pure ()) >>= \case
       Right smp -> do
         rKeys@(_, rpKey) <- atomically $ C.generateAuthKeyPair ra g
         (sKey, spKey) <- atomically $ C.generateAuthKeyPair sa g
@@ -1229,7 +1242,8 @@ runXFTPServerTest c userId (ProtoServerWithAuth srv auth) = do
   rcvPath <- getTempFilePath workDir
   liftIO $ do
     let tSess = (userId, srv, Nothing)
-    X.getXFTPClient tSess cfg {xftpNetworkConfig} (\_ -> pure ()) >>= \case
+    ts <- readTVarIO $ proxySessTs c
+    X.getXFTPClient tSess cfg {xftpNetworkConfig} ts (\_ -> pure ()) >>= \case
       Right xftp -> withTestChunk filePath $ do
         (sndKey, spKey) <- atomically $ C.generateAuthKeyPair C.SEd25519 g
         (rcvKey, rpKey) <- atomically $ C.generateAuthKeyPair C.SEd25519 g
@@ -1273,7 +1287,8 @@ runNTFServerTest c userId (ProtoServerWithAuth srv _) = do
   g <- asks random
   liftIO $ do
     let tSess = (userId, srv, Nothing)
-    getProtocolClient g tSess cfg Nothing (\_ -> pure ()) >>= \case
+    ts <- readTVarIO $ proxySessTs c
+    getProtocolClient g tSess cfg Nothing ts (\_ -> pure ()) >>= \case
       Right ntf -> do
         (nKey, npKey) <- atomically $ C.generateAuthKeyPair a g
         (dhKey, _) <- atomically $ C.generateKeyPair g
@@ -1371,6 +1386,7 @@ temporaryAgentError = \case
   PROXY _ _ (ProxyProtocolError (SMP.PROXY (SMP.BROKER e))) -> tempBrokerError e
   PROXY _ _ (ProxyProtocolError (SMP.PROXY SMP.NO_SESSION)) -> True
   INACTIVE -> True
+  CRITICAL True _ -> True -- critical errors that do not show restart button are likely to be permanent
   _ -> False
   where
     tempBrokerError = \case
@@ -1410,7 +1426,7 @@ subscribeQueues c qs = do
     checkQueue rq = do
       prohibited <- liftIO $ hasGetLock c rq
       pure $ if prohibited then Left (rq, Left $ CMD PROHIBITED "subscribeQueues") else Right rq
-    subscribeQueues_ :: Env -> TVar (Maybe SessionId) -> SMPClient -> NonEmpty RcvQueue -> IO (BatchResponses SMPClientError ())
+    subscribeQueues_ :: Env -> TVar (Maybe SessionId) -> SMPClient -> NonEmpty RcvQueue -> IO (BatchResponses RcvQueue SMPClientError ())
     subscribeQueues_ env session smp qs' = do
       let (userId, srv, _) = transportSession' smp
       atomically $ incSMPServerStat' c userId srv connSubAttempts $ length qs'
@@ -1441,31 +1457,33 @@ activeClientSession c tSess sessId = sameSess <$> tryReadSessVar tSess (smpClien
       Just (Right (SMPConnectedClient smp _)) -> sessId == sessionId (thParams smp)
       _ -> False
 
-type BatchResponses e r = NonEmpty (RcvQueue, Either e r)
+type BatchResponses q e r = NonEmpty (q, Either e r)
 
-sendTSessionBatches :: forall q r. ByteString -> (q -> RcvQueue) -> (SMPClient -> NonEmpty q -> IO (BatchResponses SMPClientError r)) -> AgentClient -> [q] -> AM' [(RcvQueue, Either AgentErrorType r)]
+-- Please note: this function does not preserve order of results to be the same as the order of arguments,
+-- it includes arguments in the results instead.
+sendTSessionBatches :: forall q r. ByteString -> (q -> RcvQueue) -> (SMPClient -> NonEmpty q -> IO (BatchResponses q SMPClientError r)) -> AgentClient -> [q] -> AM' [(q, Either AgentErrorType r)]
 sendTSessionBatches statCmd toRQ action c qs =
   concatMap L.toList <$> (mapConcurrently sendClientBatch =<< batchQueues)
   where
     batchQueues :: AM' [(SMPTransportSession, NonEmpty q)]
     batchQueues = do
       mode <- getSessionMode c
-      pure . M.assocs $ foldl' (batch mode) M.empty qs
+      pure . M.assocs $ foldr (batch mode) M.empty qs
       where
-        batch mode m q =
+        batch mode q m =
           let tSess = mkSMPTSession (toRQ q) mode
            in M.alter (Just . maybe [q] (q <|)) tSess m
-    sendClientBatch :: (SMPTransportSession, NonEmpty q) -> AM' (BatchResponses AgentErrorType r)
+    sendClientBatch :: (SMPTransportSession, NonEmpty q) -> AM' (BatchResponses q AgentErrorType r)
     sendClientBatch (tSess@(_, srv, _), qs') =
       tryAgentError' (getSMPServerClient c tSess) >>= \case
-        Left e -> pure $ L.map ((,Left e) . toRQ) qs'
+        Left e -> pure $ L.map (,Left e) qs'
         Right (SMPConnectedClient smp _) -> liftIO $ do
           logServer' "-->" c srv (bshow (length qs') <> " queues") statCmd
           L.map agentError <$> action smp qs'
           where
             agentError = second . first $ protocolClientError SMP $ clientServer smp
 
-sendBatch :: (SMPClient -> NonEmpty (SMP.RcvPrivateAuthKey, SMP.RecipientId) -> IO (NonEmpty (Either SMPClientError ()))) -> SMPClient -> NonEmpty RcvQueue -> IO (BatchResponses SMPClientError ())
+sendBatch :: (SMPClient -> NonEmpty (SMP.RcvPrivateAuthKey, SMP.RecipientId) -> IO (NonEmpty (Either SMPClientError ()))) -> SMPClient -> NonEmpty RcvQueue -> IO (BatchResponses RcvQueue SMPClientError ())
 sendBatch smpCmdFunc smp qs = L.zip qs <$> smpCmdFunc smp (L.map queueCreds qs)
   where
     queueCreds RcvQueue {rcvPrivateKey, rcvId} = (rcvPrivateKey, rcvId)
@@ -1572,7 +1590,7 @@ getQueueMessage c rq@RcvQueue {server, rcvId, rcvPrivateKey} = do
 
 decryptSMPMessage :: RcvQueue -> SMP.RcvMessage -> AM SMP.ClientRcvMsgBody
 decryptSMPMessage rq SMP.RcvMessage {msgId, msgBody = SMP.EncRcvMsgBody body} =
-  liftEither . parse SMP.clientRcvMsgBodyP (AGENT A_MESSAGE) =<< decrypt body
+  liftEither $ parse SMP.clientRcvMsgBodyP (AGENT A_MESSAGE) =<< decrypt body
   where
     decrypt = agentCbDecrypt (rcvDhSecret rq) (C.cbNonce msgId)
 
@@ -1594,28 +1612,44 @@ enableQueueNotifications c rq@RcvQueue {rcvId, rcvPrivateKey} notifierKey rcvNtf
   withSMPClient c rq "NKEY <nkey>" $ \smp ->
     enableSMPQueueNotifications smp rcvPrivateKey rcvId notifierKey rcvNtfPublicDhKey
 
-enableQueuesNtfs :: AgentClient -> [(RcvQueue, SMP.NtfPublicAuthKey, SMP.RcvNtfPublicDhKey)] -> AM' [(RcvQueue, Either AgentErrorType (SMP.NotifierId, SMP.RcvNtfPublicDhKey))]
-enableQueuesNtfs = sendTSessionBatches "NKEY" fst3 enableQueues_
+data EnableQueueNtfReq = EnableQueueNtfReq
+  { eqnrNtfSub :: NtfSubscription,
+    eqnrRq :: RcvQueue,
+    eqnrAuthKeyPair :: C.AAuthKeyPair,
+    eqnrRcvKeyPair :: C.KeyPairX25519
+  }
+
+enableQueuesNtfs :: AgentClient -> [EnableQueueNtfReq] -> AM' [(EnableQueueNtfReq, Either AgentErrorType (SMP.NotifierId, SMP.RcvNtfPublicDhKey))]
+enableQueuesNtfs = sendTSessionBatches "NKEY" eqnrRq enableQueues_
   where
-    fst3 (x, _, _) = x
-    enableQueues_ :: SMPClient -> NonEmpty (RcvQueue, SMP.NtfPublicAuthKey, SMP.RcvNtfPublicDhKey) -> IO (NonEmpty (RcvQueue, Either (ProtocolClientError ErrorType) (SMP.NotifierId, RcvNtfPublicDhKey)))
-    enableQueues_ smp qs' = L.zipWith ((,) . fst3) qs' <$> enableSMPQueuesNtfs smp (L.map queueCreds qs')
-    queueCreds :: (RcvQueue, SMP.NtfPublicAuthKey, SMP.RcvNtfPublicDhKey) -> (SMP.RcvPrivateAuthKey, SMP.RecipientId, SMP.NtfPublicAuthKey, SMP.RcvNtfPublicDhKey)
-    queueCreds (RcvQueue {rcvPrivateKey, rcvId}, notifierKey, rcvNtfPublicDhKey) = (rcvPrivateKey, rcvId, notifierKey, rcvNtfPublicDhKey)
+    enableQueues_ :: SMPClient -> NonEmpty EnableQueueNtfReq -> IO (NonEmpty (EnableQueueNtfReq, Either (ProtocolClientError ErrorType) (SMP.NotifierId, RcvNtfPublicDhKey)))
+    enableQueues_ smp qs' = L.zip qs' <$> enableSMPQueuesNtfs smp (L.map queueCreds qs')
+    queueCreds :: EnableQueueNtfReq -> (SMP.RcvPrivateAuthKey, SMP.RecipientId, SMP.NtfPublicAuthKey, SMP.RcvNtfPublicDhKey)
+    queueCreds EnableQueueNtfReq {eqnrRq, eqnrAuthKeyPair, eqnrRcvKeyPair} =
+      let RcvQueue {rcvPrivateKey, rcvId} = eqnrRq
+          (ntfPublicKey, _) = eqnrAuthKeyPair
+          (rcvNtfPubDhKey, _) = eqnrRcvKeyPair
+       in (rcvPrivateKey, rcvId, ntfPublicKey, rcvNtfPubDhKey)
 
 disableQueueNotifications :: AgentClient -> RcvQueue -> AM ()
 disableQueueNotifications c rq@RcvQueue {rcvId, rcvPrivateKey} =
   withSMPClient c rq "NDEL" $ \smp ->
     disableSMPQueueNotifications smp rcvPrivateKey rcvId
 
-disableQueuesNtfs :: AgentClient -> [RcvQueue] -> AM' [(RcvQueue, Either AgentErrorType ())]
-disableQueuesNtfs = sendTSessionBatches "NDEL" id $ sendBatch disableSMPQueuesNtfs
+type DisableQueueNtfReq = (NtfSubscription, RcvQueue)
+
+disableQueuesNtfs :: AgentClient -> [DisableQueueNtfReq] -> AM' [(DisableQueueNtfReq, Either AgentErrorType ())]
+disableQueuesNtfs = sendTSessionBatches "NDEL" snd disableQueues_
+  where
+    disableQueues_ :: SMPClient -> NonEmpty DisableQueueNtfReq -> IO (NonEmpty (DisableQueueNtfReq, Either (ProtocolClientError ErrorType) ()))
+    disableQueues_ smp qs' = L.zip qs' <$> disableSMPQueuesNtfs smp (L.map queueCreds qs')
+    queueCreds :: DisableQueueNtfReq -> (SMP.RcvPrivateAuthKey, SMP.RecipientId)
+    queueCreds (_, RcvQueue {rcvPrivateKey, rcvId}) = (rcvPrivateKey, rcvId)
 
 sendAck :: AgentClient -> RcvQueue -> MsgId -> AM ()
-sendAck c rq@RcvQueue {rcvId, rcvPrivateKey} msgId = do
+sendAck c rq@RcvQueue {rcvId, rcvPrivateKey} msgId =
   withSMPClient c rq ("ACK:" <> logSecret' msgId) $ \smp ->
-    ackSMPMessage smp rcvPrivateKey rcvId msgId
-  atomically $ releaseGetLock c rq
+    ackSMPMessage smp rcvPrivateKey rcvId msgId      
 
 hasGetLock :: AgentClient -> RcvQueue -> IO Bool
 hasGetLock c RcvQueue {server, rcvId} =
@@ -1687,8 +1721,8 @@ agentNtfReplaceToken :: AgentClient -> NtfTokenId -> NtfToken -> DeviceToken ->
 agentNtfReplaceToken c tknId NtfToken {ntfServer, ntfPrivKey} token =
   withNtfClient c ntfServer tknId "TRPL" $ \ntf -> ntfReplaceToken ntf ntfPrivKey tknId token
 
-agentNtfDeleteToken :: AgentClient -> NtfTokenId -> NtfToken -> AM ()
-agentNtfDeleteToken c tknId NtfToken {ntfServer, ntfPrivKey} =
+agentNtfDeleteToken :: AgentClient -> NtfServer -> C.APrivateAuthKey -> NtfTokenId -> AM ()
+agentNtfDeleteToken c ntfServer ntfPrivKey tknId =
   withNtfClient c ntfServer tknId "TDEL" $ \ntf -> ntfDeleteToken ntf ntfPrivKey tknId
 
 agentNtfEnableCron :: AgentClient -> NtfTokenId -> NtfToken -> Word16 -> AM ()
@@ -1699,10 +1733,34 @@ agentNtfCreateSubscription :: AgentClient -> NtfTokenId -> NtfToken -> SMPQueueN
 agentNtfCreateSubscription c tknId NtfToken {ntfServer, ntfPrivKey} smpQueue nKey =
   withNtfClient c ntfServer tknId "SNEW" $ \ntf -> ntfCreateSubscription ntf ntfPrivKey (NewNtfSub tknId smpQueue nKey)
 
-agentNtfCheckSubscription :: AgentClient -> NtfSubscriptionId -> NtfToken -> AM NtfSubStatus
-agentNtfCheckSubscription c subId NtfToken {ntfServer, ntfPrivKey} =
+agentNtfCreateSubscriptions :: AgentClient -> NtfToken -> NonEmpty (NewNtfEntity 'Subscription) -> AM' (NonEmpty (Either AgentErrorType NtfSubscriptionId))
+agentNtfCreateSubscriptions = withNtfBatch "SNEW" ntfCreateSubscriptions
+
+agentNtfCheckSubscription :: AgentClient -> NtfToken -> NtfSubscriptionId -> AM NtfSubStatus
+agentNtfCheckSubscription c NtfToken {ntfServer, ntfPrivKey} subId =
   withNtfClient c ntfServer subId "SCHK" $ \ntf -> ntfCheckSubscription ntf ntfPrivKey subId
 
+agentNtfCheckSubscriptions :: AgentClient -> NtfToken -> NonEmpty NtfSubscriptionId -> AM' (NonEmpty (Either AgentErrorType NtfSubStatus))
+agentNtfCheckSubscriptions = withNtfBatch "SCHK" ntfCheckSubscriptions
+
+-- This batch sends all commands to one ntf server (client can only use one server at a time)
+withNtfBatch ::
+  ByteString ->
+  (NtfClient -> C.APrivateAuthKey -> NonEmpty a -> IO (NonEmpty (Either NtfClientError r))) ->
+  AgentClient ->
+  NtfToken ->
+  NonEmpty a ->
+  AM' (NonEmpty (Either AgentErrorType r))
+withNtfBatch cmdStr action c NtfToken {ntfServer, ntfPrivKey} subs = do
+  let tSess = (0, ntfServer, Nothing)
+  tryAgentError' (getNtfServerClient c tSess) >>= \case
+    Left e -> pure $ L.map (\_ -> Left e) subs
+    Right ntf -> liftIO $ do
+      logServer' "-->" c ntfServer (bshow (length subs) <> " subscriptions") cmdStr
+      L.map agentError <$> action ntf ntfPrivKey subs
+      where
+        agentError = first $ protocolClientError NTF $ clientServer ntf
+
 agentNtfDeleteSubscription :: AgentClient -> NtfSubscriptionId -> NtfToken -> AM ()
 agentNtfDeleteSubscription c subId NtfToken {ntfServer, ntfPrivKey} =
   withNtfClient c ntfServer subId "SDEL" $ \ntf -> ntfDeleteSubscription ntf ntfPrivKey subId
@@ -1772,9 +1830,9 @@ agentCbEncryptOnce clientVersion dhRcvPubKey msg = do
 
 -- | NaCl crypto-box decrypt - both for messages received from the server
 -- and per-queue E2E encrypted messages from the sender that were inside.
-agentCbDecrypt :: C.DhSecretX25519 -> C.CbNonce -> ByteString -> AM ByteString
+agentCbDecrypt :: C.DhSecretX25519 -> C.CbNonce -> ByteString -> Either AgentErrorType ByteString
 agentCbDecrypt dhSecret nonce msg =
-  liftEither . first cryptoError $
+  first cryptoError $
     C.cbDecrypt dhSecret nonce msg
 
 cryptoError :: C.CryptoError -> AgentErrorType
@@ -1800,12 +1858,39 @@ withWork c doWork getWork action =
   withStore' c getWork >>= \case
     Right (Just r) -> action r
     Right Nothing -> noWork
+    -- worker is stopped here (noWork) because the next iteration is likely to produce the same result
     Left e@SEWorkItemError {} -> noWork >> notifyErr (CRITICAL False) e
     Left e -> notifyErr INTERNAL e
   where
     noWork = liftIO $ noWorkToDo doWork
     notifyErr err e = atomically $ writeTBQueue (subQ c) ("", "", AEvt SAEConn $ ERR $ err $ show e)
 
+withWorkItems :: AgentClient -> TMVar () -> (DB.Connection -> IO (Either StoreError [Either StoreError a])) -> (NonEmpty a -> AM ()) -> AM ()
+withWorkItems c doWork getWork action = do
+  withStore' c getWork >>= \case
+    Right [] -> noWork
+    Right rs -> do
+      let (errs, items) = partitionEithers rs
+      case L.nonEmpty items of
+        Just items' -> action items'
+        Nothing -> do
+          let criticalErr = find workItemError errs
+          forM_ criticalErr $ \err -> do
+            notifyErr (CRITICAL False) err
+            when (all workItemError errs) noWork
+      unless (null errs) $
+        atomically $
+          writeTBQueue (subQ c) ("", "", AEvt SAENone $ ERRS $ map (\e -> ("", INTERNAL $ show e)) errs)
+    Left e
+      | workItemError e -> noWork >> notifyErr (CRITICAL False) e
+      | otherwise -> notifyErr INTERNAL e
+  where
+    workItemError = \case
+      SEWorkItemError {} -> True
+      _ -> False
+    noWork = liftIO $ noWorkToDo doWork
+    notifyErr err e = atomically $ writeTBQueue (subQ c) ("", "", AEvt SAEConn $ ERR $ err $ show e)
+
 noWorkToDo :: TMVar () -> IO ()
 noWorkToDo = void . atomically . tryTakeTMVar
 {-# INLINE noWorkToDo #-}
@@ -1949,33 +2034,82 @@ userServers c = case protocolTypeI @p of
   SPXFTP -> xftpServers c
 {-# INLINE userServers #-}
 
-pickServer :: forall p. NonEmpty (ProtoServerWithAuth p) -> AM (ProtoServerWithAuth p)
+pickServer :: NonEmpty (Maybe OperatorId, ProtoServerWithAuth p) -> AM (ProtoServerWithAuth p)
 pickServer = \case
-  srv :| [] -> pure srv
+  (_, srv) :| [] -> pure srv
   servers -> do
     gen <- asks randomServer
-    atomically $ (servers L.!!) <$> stateTVar gen (randomR (0, L.length servers - 1))
+    atomically $ snd . (servers L.!!) <$> stateTVar gen (randomR (0, L.length servers - 1))
 
-getNextServer :: forall p. (ProtocolTypeI p, UserProtocol p) => AgentClient -> UserId -> [ProtocolServer p] -> AM (ProtoServerWithAuth p)
-getNextServer c userId usedSrvs = withUserServers c userId $ \srvs ->
-  case L.nonEmpty $ deleteFirstsBy sameSrvAddr' (L.toList srvs) (map noAuthSrv usedSrvs) of
-    Just srvs' -> pickServer srvs'
-    _ -> pickServer srvs
+getNextServer ::
+  (ProtocolTypeI p, UserProtocol p) =>
+  AgentClient ->
+  UserId -> 
+  (UserServers p -> NonEmpty (Maybe OperatorId, ProtoServerWithAuth p)) ->
+  [ProtocolServer p] ->
+  AM (ProtoServerWithAuth p)
+getNextServer c userId srvsSel usedSrvs = do
+  srvs <- getUserServers_ c userId srvsSel
+  snd <$> getNextServer_ srvs (usedOperatorsHosts srvs usedSrvs)
 
-withUserServers :: forall p a. (ProtocolTypeI p, UserProtocol p) => AgentClient -> UserId -> (NonEmpty (ProtoServerWithAuth p) -> AM a) -> AM a
-withUserServers c userId action =
+usedOperatorsHosts :: NonEmpty (Maybe OperatorId, ProtoServerWithAuth p) -> [ProtocolServer p] -> (Set (Maybe OperatorId), Set TransportHost)
+usedOperatorsHosts srvs usedSrvs = (usedOperators, usedHosts)
+  where
+    usedHosts = S.unions $ map serverHosts usedSrvs
+    usedOperators = S.fromList $ mapMaybe usedOp $ L.toList srvs
+    usedOp (op, srv) = if hasUsedHost srv then Just op else Nothing
+    hasUsedHost (ProtoServerWithAuth srv _) = any (`S.member` usedHosts) $ serverHosts srv
+
+getNextServer_ ::
+  (ProtocolTypeI p, UserProtocol p) =>
+  NonEmpty (Maybe OperatorId, ProtoServerWithAuth p) ->
+  (Set (Maybe OperatorId), Set TransportHost) ->
+  AM (NonEmpty (Maybe OperatorId, ProtoServerWithAuth p), ProtoServerWithAuth p)
+getNextServer_ servers (usedOperators, usedHosts) = do
+  -- choose from servers of unused operators, when possible
+  let otherOpsSrvs = filterOrAll ((`S.notMember` usedOperators) . fst) servers
+      -- choose from servers with unused hosts when possible
+      unusedSrvs = filterOrAll (isUnusedServer usedHosts) otherOpsSrvs
+  (otherOpsSrvs,) <$> pickServer unusedSrvs
+  where
+    filterOrAll p srvs = fromMaybe srvs $ L.nonEmpty $ L.filter p srvs
+
+isUnusedServer :: Set TransportHost -> (Maybe OperatorId, ProtoServerWithAuth p) -> Bool
+isUnusedServer usedHosts (_, ProtoServerWithAuth ProtocolServer {host} _) = all (`S.notMember` usedHosts) host
+
+getUserServers_ ::
+  (ProtocolTypeI p, UserProtocol p) =>
+  AgentClient ->
+  UserId ->
+  (UserServers p -> NonEmpty (Maybe OperatorId, ProtoServerWithAuth p)) ->
+  AM (NonEmpty (Maybe OperatorId, ProtoServerWithAuth p))
+getUserServers_ c userId srvsSel =
   liftIO (TM.lookupIO userId $ userServers c) >>= \case
-    Just srvs -> action $ enabledSrvs srvs
+    Just srvs -> pure $ srvsSel srvs
     _ -> throwE $ INTERNAL "unknown userId - no user servers"
 
-withNextSrv :: forall p a. (ProtocolTypeI p, UserProtocol p) => AgentClient -> UserId -> TVar [ProtocolServer p] -> [ProtocolServer p] -> (ProtoServerWithAuth p -> AM a) -> AM a
-withNextSrv c userId usedSrvs initUsed action = do
-  used <- readTVarIO usedSrvs
-  srvAuth@(ProtoServerWithAuth srv _) <- getNextServer c userId used
-  srvs_ <- liftIO $ TM.lookupIO userId $ userServers c
-  let unused = maybe [] ((\\ used) . map protoServer . L.toList . enabledSrvs) srvs_
-      used' = if null unused then initUsed else srv : used
-  atomically $ writeTVar usedSrvs $! used'
+-- This function checks used servers and operators every time to allow
+-- changing configuration while retry look is executing.
+-- This function is not thread safe.
+withNextSrv ::
+  (ProtocolTypeI p, UserProtocol p) =>
+  AgentClient ->
+  UserId ->
+  (UserServers p -> NonEmpty (Maybe OperatorId, ProtoServerWithAuth p)) ->
+  TVar (Set TransportHost) ->
+  [ProtocolServer p] -> 
+  (ProtoServerWithAuth p -> AM a) ->
+  AM a
+withNextSrv c userId srvsSel triedHosts usedSrvs action = do
+  srvs <- getUserServers_ c userId srvsSel
+  let (usedOperators, usedHosts) = usedOperatorsHosts srvs usedSrvs
+  tried <- readTVarIO triedHosts
+  let triedOrUsed = S.union tried usedHosts
+  (otherOpsSrvs, srvAuth@(ProtoServerWithAuth srv _)) <- getNextServer_ srvs (usedOperators, triedOrUsed)
+  let newHosts = serverHosts srv
+      unusedSrvs = L.filter (isUnusedServer $ S.union triedOrUsed newHosts) otherOpsSrvs
+      !tried' = if null unusedSrvs then S.empty else S.union tried newHosts
+  atomically $ writeTVar triedHosts tried'
   action srvAuth
 
 incSMPServerStat :: AgentClient -> UserId -> SMPServer -> (AgentSMPServerStats -> TVar Int) -> STM ()
@@ -2001,9 +2135,13 @@ incXFTPServerStat_ = incServerStat (\AgentClient {xftpServersStats = s} -> s) ne
 {-# INLINE incXFTPServerStat_ #-}
 
 incNtfServerStat :: AgentClient -> UserId -> NtfServer -> (AgentNtfServerStats -> TVar Int) -> STM ()
-incNtfServerStat c userId srv sel = incServerStat (\AgentClient {ntfServersStats = s} -> s) newAgentNtfServerStats c userId srv sel 1
+incNtfServerStat c userId srv sel = incNtfServerStat' c userId srv sel 1
 {-# INLINE incNtfServerStat #-}
 
+incNtfServerStat' :: AgentClient -> UserId -> NtfServer -> (AgentNtfServerStats -> TVar Int) -> Int -> STM ()
+incNtfServerStat' = incServerStat (\AgentClient {ntfServersStats = s} -> s) newAgentNtfServerStats
+{-# INLINE incNtfServerStat' #-}
+
 incServerStat :: Num n => (AgentClient -> TMap (UserId, ProtocolServer p) s) -> STM s -> AgentClient -> UserId -> ProtocolServer p -> (s -> TVar n) -> n -> STM ()
 incServerStat statsSel mkNewStats c userId srv sel n = do
   TM.lookup (userId, srv) (statsSel c) >>= \case

src/Simplex/Messaging/Agent/Env/SQLite.hs

@@ -17,11 +17,14 @@ module Simplex.Messaging.Agent.Env.SQLite
     AgentConfig (..),
     InitialAgentServers (..),
     ServerCfg (..),
+    ServerRoles (..),
+    OperatorId,
     UserServers (..),
     NetworkConfig (..),
     presetServerCfg,
-    enabledServerCfg,
+    allRoles,
     mkUserServers,
+    serverHosts,
     defaultAgentConfig,
     defaultReconnectInterval,
     tryAgentError,
@@ -42,6 +45,7 @@ module Simplex.Messaging.Agent.Env.SQLite
 where
 
 import Control.Concurrent (ThreadId)
+import Control.Exception (BlockedIndefinitelyOnSTM (..), SomeException, fromException)
 import Control.Monad.Except
 import Control.Monad.IO.Unlift
 import Control.Monad.Reader
@@ -54,6 +58,8 @@ import Data.List.NonEmpty (NonEmpty)
 import qualified Data.List.NonEmpty as L
 import Data.Map.Strict (Map)
 import Data.Maybe (fromMaybe)
+import Data.Set (Set)
+import qualified Data.Set as S
 import Data.Time.Clock (NominalDiffTime, nominalDay)
 import Data.Time.Clock.System (SystemTime (..))
 import Data.Word (Word16)
@@ -71,15 +77,14 @@ import Simplex.Messaging.Notifications.Client (defaultNTFClientConfig)
 import Simplex.Messaging.Notifications.Transport (NTFVersion)
 import Simplex.Messaging.Notifications.Types
 import Simplex.Messaging.Parsers (defaultJSON)
-import Simplex.Messaging.Protocol (NtfServer, ProtoServerWithAuth, ProtocolServer, ProtocolType (..), ProtocolTypeI, VersionRangeSMPC, XFTPServer, supportedSMPClientVRange)
+import Simplex.Messaging.Protocol (NtfServer, ProtoServerWithAuth (..), ProtocolServer (..), ProtocolType (..), ProtocolTypeI, VersionRangeSMPC, XFTPServer, supportedSMPClientVRange)
 import Simplex.Messaging.TMap (TMap)
 import qualified Simplex.Messaging.TMap as TM
-import Simplex.Messaging.Transport (SMPVersion, TLS, Transport (..))
-import Simplex.Messaging.Transport.Client (defaultSMPPort)
+import Simplex.Messaging.Transport (SMPVersion)
+import Simplex.Messaging.Transport.Client (TransportHost)
 import Simplex.Messaging.Util (allFinally, catchAllErrors, catchAllErrors', tryAllErrors, tryAllErrors')
 import System.Mem.Weak (Weak)
 import System.Random (StdGen, newStdGen)
-import UnliftIO (SomeException)
 import UnliftIO.STM
 
 type AM' a = ReaderT Env IO a
@@ -95,29 +100,42 @@ data InitialAgentServers = InitialAgentServers
 
 data ServerCfg p = ServerCfg
   { server :: ProtoServerWithAuth p,
-    preset :: Bool,
-    tested :: Maybe Bool,
-    enabled :: Bool
+    operator :: Maybe OperatorId,
+    enabled :: Bool,
+    roles :: ServerRoles
   }
   deriving (Show)
 
-enabledServerCfg :: ProtoServerWithAuth p -> ServerCfg p
-enabledServerCfg server = ServerCfg {server, preset = False, tested = Nothing, enabled = True}
+data ServerRoles = ServerRoles
+  { storage :: Bool,
+    proxy :: Bool
+  }
+  deriving (Show)
 
-presetServerCfg :: Bool -> ProtoServerWithAuth p -> ServerCfg p
-presetServerCfg enabled server = ServerCfg {server, preset = True, tested = Nothing, enabled}
+allRoles :: ServerRoles
+allRoles = ServerRoles True True
+
+presetServerCfg :: Bool -> ServerRoles -> Maybe OperatorId -> ProtoServerWithAuth p -> ServerCfg p
+presetServerCfg enabled roles operator server =
+  ServerCfg {server, operator, enabled, roles}
 
 data UserServers p = UserServers
-  { enabledSrvs :: NonEmpty (ProtoServerWithAuth p),
-    knownSrvs :: NonEmpty (ProtocolServer p)
+  { storageSrvs :: NonEmpty (Maybe OperatorId, ProtoServerWithAuth p),
+    proxySrvs :: NonEmpty (Maybe OperatorId, ProtoServerWithAuth p),
+    knownHosts :: Set TransportHost
   }
 
+type OperatorId = Int64
+
 -- This function sets all servers as enabled in case all passed servers are disabled.
 mkUserServers :: NonEmpty (ServerCfg p) -> UserServers p
-mkUserServers srvs = UserServers {enabledSrvs, knownSrvs}
+mkUserServers srvs = UserServers {storageSrvs = filterSrvs storage, proxySrvs = filterSrvs proxy, knownHosts}
   where
-    enabledSrvs = L.map (\ServerCfg {server} -> server) $ fromMaybe srvs $ L.nonEmpty $ L.filter (\ServerCfg {enabled} -> enabled) srvs
-    knownSrvs = L.map (\ServerCfg {server = ProtoServerWithAuth srv _} -> srv) srvs
+    filterSrvs role = L.map (\ServerCfg {operator, server} -> (operator, server)) $ fromMaybe srvs $ L.nonEmpty $ L.filter (\ServerCfg {enabled, roles} -> enabled && role roles) srvs
+    knownHosts = S.unions $ L.map (\ServerCfg {server = ProtoServerWithAuth srv _} -> serverHosts srv) srvs
+
+serverHosts :: ProtocolServer p -> Set TransportHost
+serverHosts ProtocolServer {host} = S.fromList $ L.toList host
 
 data AgentConfig = AgentConfig
   { tcpPort :: Maybe ServiceName,
@@ -150,6 +168,8 @@ data AgentConfig = AgentConfig
     xftpMaxRecipientsPerRequest :: Int,
     deleteErrorCount :: Int,
     ntfCron :: Word16,
+    ntfBatchSize :: Int,
+    ntfSubFirstCheckInterval :: NominalDiffTime,
     ntfSubCheckInterval :: NominalDiffTime,
     caCertificateFile :: FilePath,
     privateKeyFile :: FilePath,
@@ -193,9 +213,9 @@ defaultAgentConfig =
       rcvAuthAlg = C.AuthAlg C.SEd25519, -- this will stay as Ed25519
       sndAuthAlg = C.AuthAlg C.SEd25519, -- TODO replace with X25519 when switching to v7
       connIdBytes = 12,
-      tbqSize = 64,
-      smpCfg = defaultSMPClientConfig {defaultTransport = (show defaultSMPPort, transport @TLS)},
-      ntfCfg = defaultNTFClientConfig {defaultTransport = ("443", transport @TLS)},
+      tbqSize = 128,
+      smpCfg = defaultSMPClientConfig,
+      ntfCfg = defaultNTFClientConfig,
       xftpCfg = defaultXFTPClientConfig,
       reconnectInterval = defaultReconnectInterval,
       messageRetryInterval = defaultMessageRetryInterval,
@@ -219,7 +239,9 @@ defaultAgentConfig =
       xftpMaxRecipientsPerRequest = 200,
       deleteErrorCount = 10,
       ntfCron = 20, -- minutes
-      ntfSubCheckInterval = nominalDay,
+      ntfBatchSize = 150,
+      ntfSubFirstCheckInterval = nominalDay,
+      ntfSubCheckInterval = 3 * nominalDay,
       -- CA certificate private key is not needed for initialization
       -- ! we do not generate these
       caCertificateFile = "/etc/opt/simplex-agent/ca.crt",
@@ -254,12 +276,13 @@ createAgentStore dbFilePath dbKey keepKey = createSQLiteStore dbFilePath dbKey k
 
 data NtfSupervisor = NtfSupervisor
   { ntfTkn :: TVar (Maybe NtfToken),
-    ntfSubQ :: TBQueue (ConnId, NtfSupervisorCommand),
+    ntfSubQ :: TBQueue (NtfSupervisorCommand, NonEmpty ConnId),
     ntfWorkers :: TMap NtfServer Worker,
-    ntfSMPWorkers :: TMap SMPServer Worker
+    ntfSMPWorkers :: TMap SMPServer Worker,
+    ntfTknDelWorkers :: TMap NtfServer Worker
   }
 
-data NtfSupervisorCommand = NSCCreate | NSCDelete | NSCSmpDelete | NSCNtfWorker NtfServer | NSCNtfSMPWorker SMPServer
+data NtfSupervisorCommand = NSCCreate | NSCSmpDelete | NSCDeleteSub
   deriving (Show)
 
 newNtfSubSupervisor :: Natural -> IO NtfSupervisor
@@ -268,7 +291,8 @@ newNtfSubSupervisor qSize = do
   ntfSubQ <- newTBQueueIO qSize
   ntfWorkers <- TM.emptyIO
   ntfSMPWorkers <- TM.emptyIO
-  pure NtfSupervisor {ntfTkn, ntfSubQ, ntfWorkers, ntfSMPWorkers}
+  ntfTknDelWorkers <- TM.emptyIO
+  pure NtfSupervisor {ntfTkn, ntfSubQ, ntfWorkers, ntfSMPWorkers, ntfTknDelWorkers}
 
 data XFTPAgent = XFTPAgent
   { -- if set, XFTP file paths will be considered as relative to this directory
@@ -308,7 +332,9 @@ agentFinally = allFinally mkInternal
 {-# INLINE agentFinally #-}
 
 mkInternal :: SomeException -> AgentErrorType
-mkInternal = INTERNAL . show
+mkInternal e = case fromException e of
+  Just BlockedIndefinitelyOnSTM -> CRITICAL True "Thread blocked indefinitely in STM transaction"
+  _ -> INTERNAL $ show e
 {-# INLINE mkInternal #-}
 
 data Worker = Worker
@@ -330,6 +356,8 @@ updateRestartCount t (RestartCount minute count) = do
 
 $(pure [])
 
+$(JQ.deriveJSON defaultJSON ''ServerRoles)
+
 instance ProtocolTypeI p => ToJSON (ServerCfg p) where
   toEncoding = $(JQ.mkToEncoding defaultJSON ''ServerCfg)
   toJSON = $(JQ.mkToJSON defaultJSON ''ServerCfg)

src/Simplex/Messaging/Agent/Lock.hs

@@ -1,6 +1,7 @@
 module Simplex.Messaging.Agent.Lock
   ( Lock,
     createLock,
+    createLockIO,
     withLock,
     withLock',
     withGetLock,
@@ -24,6 +25,10 @@ createLock :: STM Lock
 createLock = newEmptyTMVar
 {-# INLINE createLock #-}
 
+createLockIO :: IO Lock
+createLockIO = newEmptyTMVarIO
+{-# INLINE createLockIO #-}
+
 withLock :: MonadUnliftIO m => Lock -> String -> ExceptT e m a -> ExceptT e m a
 withLock lock name = ExceptT . withLock' lock name . runExceptT
 {-# INLINE withLock #-}

src/Simplex/Messaging/Agent/NtfSubSupervisor.hs

@@ -1,7 +1,9 @@
+{-# LANGUAGE DataKinds #-}
 {-# LANGUAGE DuplicateRecordFields #-}
 {-# LANGUAGE FlexibleContexts #-}
 {-# LANGUAGE LambdaCase #-}
 {-# LANGUAGE NamedFieldPuns #-}
+{-# LANGUAGE OverloadedLists #-}
 {-# LANGUAGE OverloadedStrings #-}
 {-# LANGUAGE ScopedTypeVariables #-}
 {-# LANGUAGE TupleSections #-}
@@ -13,6 +15,7 @@ module Simplex.Messaging.Agent.NtfSubSupervisor
     nsRemoveNtfToken,
     sendNtfSubCommand,
     instantNotifications,
+    deleteToken,
     closeNtfSupervisor,
     getNtfServer,
   )
@@ -22,23 +25,32 @@ import Control.Logger.Simple (logError, logInfo)
 import Control.Monad
 import Control.Monad.Reader
 import Control.Monad.Trans.Except
+import Crypto.Random (ChaChaDRG)
 import Data.Bifunctor (first)
+import Data.Either (fromRight, partitionEithers)
+import Data.Functor (($>))
+import Data.List (foldl')
+import Data.List.NonEmpty (NonEmpty (..))
+import qualified Data.List.NonEmpty as L
 import qualified Data.Map.Strict as M
-import Data.Text (Text)
+import Data.Maybe (catMaybes)
+import qualified Data.Set as S
 import Data.Time (UTCTime, addUTCTime, getCurrentTime)
 import Data.Time.Clock (diffUTCTime)
 import Simplex.Messaging.Agent.Client
 import Simplex.Messaging.Agent.Env.SQLite
-import Simplex.Messaging.Agent.Protocol (AEvent (..), AEvt (..), AgentErrorType (..), BrokerErrorType (..), ConnId, NotificationsMode (..), SAEntity (..))
+import Simplex.Messaging.Agent.Protocol
 import Simplex.Messaging.Agent.RetryInterval
 import Simplex.Messaging.Agent.Stats
 import Simplex.Messaging.Agent.Store
 import Simplex.Messaging.Agent.Store.SQLite
+import qualified Simplex.Messaging.Agent.Store.SQLite.DB as DB
 import qualified Simplex.Messaging.Crypto as C
-import Simplex.Messaging.Notifications.Protocol (NtfSubStatus (..), NtfTknStatus (..), SMPQueueNtf (..))
+import Simplex.Messaging.Notifications.Protocol
 import Simplex.Messaging.Notifications.Types
-import Simplex.Messaging.Protocol (NtfServer, SMPServer, sameSrvAddr)
-import Simplex.Messaging.Util (diffToMicroseconds, threadDelay', tshow, unlessM)
+import Simplex.Messaging.Protocol (NtfServer, sameSrvAddr)
+import qualified Simplex.Messaging.Protocol as SMP
+import Simplex.Messaging.Util (diffToMicroseconds, threadDelay', tshow)
 import System.Random (randomR)
 import UnliftIO
 import UnliftIO.Concurrent (forkIO)
@@ -47,91 +59,134 @@ import qualified UnliftIO.Exception as E
 runNtfSupervisor :: AgentClient -> AM' ()
 runNtfSupervisor c = do
   ns <- asks ntfSupervisor
+  runExceptT startTknDelete >>= \case
+    Left e -> notifyErr e
+    Right _ -> pure ()
   forever $ do
-    cmd@(connId, _) <- atomically . readTBQueue $ ntfSubQ ns
-    handleErr connId . agentOperationBracket c AONtfNetwork waitUntilActive $
-      runExceptT (processNtfSub c cmd) >>= \case
-        Left e -> notifyErr connId e
+    cmd <- atomically . readTBQueue $ ntfSubQ ns
+    handleErr . agentOperationBracket c AONtfNetwork waitUntilActive $
+      runExceptT (processNtfCmd c cmd) >>= \case
+        Left e -> notifyErr e
         Right _ -> return ()
   where
-    handleErr :: ConnId -> AM' () -> AM' ()
-    handleErr connId = E.handle $ \(e :: E.SomeException) -> do
+    startTknDelete :: AM ()
+    startTknDelete = do
+      pendingDelServers <- withStore' c getPendingDelTknServers
+      lift . forM_ pendingDelServers $ getNtfTknDelWorker True c
+    handleErr :: AM' () -> AM' ()
+    handleErr = E.handle $ \(e :: E.SomeException) -> do
       logError $ "runNtfSupervisor error " <> tshow e
-      notifyErr connId e
-    notifyErr connId e = notifyInternalError c connId $ "runNtfSupervisor error " <> show e
+      notifyErr e
+    notifyErr e = notifyInternalError' c $ "runNtfSupervisor error " <> show e
 
-processNtfSub :: AgentClient -> (ConnId, NtfSupervisorCommand) -> AM ()
-processNtfSub c (connId, cmd) = do
-  logInfo $ "processNtfSub - connId = " <> tshow connId <> " - cmd = " <> tshow cmd
+partitionErrs :: (a -> ConnId) -> [a] -> [Either AgentErrorType b] -> ([(ConnId, AgentErrorType)], [b])
+partitionErrs f xs = partitionEithers . zipWith (\x -> first (f x,)) xs
+{-# INLINE partitionErrs #-}
+
+ntfSubConnId :: NtfSubscription -> ConnId
+ntfSubConnId NtfSubscription {connId} = connId
+
+processNtfCmd :: AgentClient -> (NtfSupervisorCommand, NonEmpty ConnId) -> AM ()
+processNtfCmd c (cmd, connIds) = do
+  logInfo $ "processNtfCmd - cmd = " <> tshow cmd
+  let connIds' = L.toList connIds
   case cmd of
     NSCCreate -> do
-      (a, RcvQueue {userId, server = smpServer, clientNtfCreds}) <- withStore c $ \db -> runExceptT $ do
-        a <- liftIO $ getNtfSubscription db connId
-        q <- ExceptT $ getPrimaryRcvQueue db connId
-        pure (a, q)
-      logInfo $ "processNtfSub, NSCCreate - a = " <> tshow a
-      case a of
-        Nothing -> do
+      (cErrs, rqSubActions) <- lift $ partitionErrs id connIds' <$> withStoreBatch c (\db -> map (getQueueSub db) connIds')
+      notifyErrs c cErrs
+      logInfo $ "processNtfCmd, NSCCreate - length rqSubs = " <> tshow (length rqSubActions)
+      let (ns, rs, css, cns) = partitionQueueSubActions rqSubActions
+      createNewSubs ns
+      resetSubs rs
+      lift $ do
+        mapM_ (getNtfSMPWorker True c) (S.fromList css)
+        mapM_ (getNtfNTFWorker True c) (S.fromList cns)
+      where
+        getQueueSub ::
+          DB.Connection ->
+          ConnId ->
+          IO (Either AgentErrorType (RcvQueue, Maybe NtfSupervisorSub))
+        getQueueSub db connId = fmap (first storeError) $ runExceptT $ do
+          rq <- ExceptT $ getPrimaryRcvQueue db connId
+          sub <- liftIO $ getNtfSubscription db connId
+          pure (rq, sub)
+        createNewSubs :: [RcvQueue] -> AM ()
+        createNewSubs rqs = do
           withTokenServer $ \ntfServer -> do
-            case clientNtfCreds of
-              Just ClientNtfCreds {notifierId} -> do
-                let newSub = newNtfSubscription userId connId smpServer (Just notifierId) ntfServer NASKey
-                withStore c $ \db -> createNtfSubscription db newSub $ NSANtf NSACreate
-                lift . void $ getNtfNTFWorker True c ntfServer
-              Nothing -> do
-                let newSub = newNtfSubscription userId connId smpServer Nothing ntfServer NASNew
-                withStore c $ \db -> createNtfSubscription db newSub $ NSASMP NSASmpKey
-                lift . void $ getNtfSMPWorker True c smpServer
-        (Just (sub@NtfSubscription {ntfSubStatus, ntfServer = subNtfServer, smpServer = smpServer', ntfQueueId}, action_)) -> do
-          case (clientNtfCreds, ntfQueueId) of
-            (Just ClientNtfCreds {notifierId}, Just ntfQueueId')
-              | sameSrvAddr smpServer smpServer' && notifierId == ntfQueueId' -> create
-              | otherwise -> rotate
-            (Nothing, Nothing) -> create
-            _ -> rotate
+            let newSubs = map (rqToNewSub ntfServer) rqs
+            (cErrs, _) <- lift $ partitionErrs ntfSubConnId newSubs <$> withStoreBatch c (\db -> map (storeNewSub db) newSubs)
+            notifyErrs c cErrs
+            kickSMPWorkers rqs
           where
-            create :: AM ()
-            create = case action_ of
-              -- action was set to NULL after worker internal error
-              Nothing -> resetSubscription
-              Just (action, _)
-                -- subscription was marked for deletion / is being deleted
-                | isDeleteNtfSubAction action -> do
-                    if ntfSubStatus == NASNew || ntfSubStatus == NASOff || ntfSubStatus == NASDeleted
-                      then resetSubscription
-                      else withTokenServer $ \ntfServer -> do
-                        withStore' c $ \db -> supervisorUpdateNtfSub db sub {ntfServer} (NSANtf NSACreate)
-                        lift . void $ getNtfNTFWorker True c ntfServer
-                | otherwise -> case action of
-                    NSANtf _ -> lift . void $ getNtfNTFWorker True c subNtfServer
-                    NSASMP _ -> lift . void $ getNtfSMPWorker True c smpServer
-            rotate :: AM ()
-            rotate = do
-              withStore' c $ \db -> supervisorUpdateNtfSub db sub (NSANtf NSARotate)
-              lift . void $ getNtfNTFWorker True c subNtfServer
-            resetSubscription :: AM ()
-            resetSubscription =
-              withTokenServer $ \ntfServer -> do
-                let sub' = sub {ntfQueueId = Nothing, ntfServer, ntfSubId = Nothing, ntfSubStatus = NASNew}
-                withStore' c $ \db -> supervisorUpdateNtfSub db sub' (NSASMP NSASmpKey)
-                lift . void $ getNtfSMPWorker True c smpServer
-    NSCDelete -> do
-      sub_ <- withStore' c $ \db -> do
-        supervisorUpdateNtfAction db connId (NSANtf NSADelete)
-        getNtfSubscription db connId
-      logInfo $ "processNtfSub, NSCDelete - sub_ = " <> tshow sub_
-      case sub_ of
-        (Just (NtfSubscription {ntfServer}, _)) -> lift . void $ getNtfNTFWorker True c ntfServer
-        _ -> pure () -- err "NSCDelete - no subscription"
+            rqToNewSub :: NtfServer -> RcvQueue -> NtfSubscription
+            rqToNewSub ntfServer RcvQueue {userId, connId, server} = newNtfSubscription userId connId server Nothing ntfServer NASNew
+            storeNewSub :: DB.Connection -> NtfSubscription -> IO (Either AgentErrorType ())
+            storeNewSub db sub = first storeError <$> createNtfSubscription db sub (NSASMP NSASmpKey)
+        resetSubs :: [(RcvQueue, NtfSubscription)] -> AM ()
+        resetSubs rqSubs = do
+          withTokenServer $ \ntfServer -> do
+            let subsToReset = map (toResetSub ntfServer) rqSubs
+            (cErrs, _) <- lift $ partitionErrs ntfSubConnId subsToReset <$> withStoreBatch' c (\db -> map (storeResetSub db) subsToReset)
+            notifyErrs c cErrs
+            let rqs = map fst rqSubs
+            kickSMPWorkers rqs
+          where
+            toResetSub :: NtfServer -> (RcvQueue, NtfSubscription) -> NtfSubscription
+            toResetSub ntfServer (rq, sub) =
+              let RcvQueue {server = smpServer} = rq
+               in sub {smpServer, ntfQueueId = Nothing, ntfServer, ntfSubId = Nothing, ntfSubStatus = NASNew}
+            storeResetSub :: DB.Connection -> NtfSubscription -> IO ()
+            storeResetSub db sub = supervisorUpdateNtfSub db sub (NSASMP NSASmpKey)
+        partitionQueueSubActions ::
+          [(RcvQueue, Maybe NtfSupervisorSub)] ->
+          ( [RcvQueue], -- new subs
+            [(RcvQueue, NtfSubscription)], -- reset subs
+            [SMPServer], -- continue work (SMP)
+            [NtfServer] -- continue work (Ntf)
+          )
+        partitionQueueSubActions = foldr decideSubWork ([], [], [], [])
+          where
+            -- sub = Nothing, needs to be created
+            decideSubWork (rq, Nothing) (ns, rs, css, cns) = (rq : ns, rs, css, cns)
+            decideSubWork (rq, Just (sub, subAction_)) (ns, rs, css, cns) =
+              case (clientNtfCreds rq, ntfQueueId sub) of
+                -- notifier ID created on SMP server (on ntf server subscription can be registered or not yet),
+                -- need to clarify action
+                (Just ClientNtfCreds {notifierId}, Just ntfQueueId')
+                  | sameSrvAddr (qServer rq) subSMPServer && notifierId == ntfQueueId' -> contOrReset
+                  | otherwise -> reset
+                (Nothing, Nothing) -> contOrReset
+                _ -> reset
+              where
+                NtfSubscription {ntfServer = subNtfServer, smpServer = subSMPServer} = sub
+                contOrReset = case subAction_ of
+                  -- action was set to NULL after worker internal error
+                  Nothing -> reset
+                  Just (action, _)
+                    -- subscription was marked for deletion / is being deleted
+                    | isDeleteNtfSubAction action -> reset
+                    -- continue work on subscription (e.g. supervisor was repeatedly tasked with creating a subscription)
+                    | otherwise -> case action of
+                        NSASMP _ -> (ns, rs, qServer rq : css, cns)
+                        NSANtf _ -> (ns, rs, css, subNtfServer : cns)
+                reset = (ns, (rq, sub) : rs, css, cns)
     NSCSmpDelete -> do
-      withStore' c (`getPrimaryRcvQueue` connId) >>= \case
-        Right rq@RcvQueue {server = smpServer} -> do
-          logInfo $ "processNtfSub, NSCSmpDelete - rq = " <> tshow rq
-          withStore' c $ \db -> supervisorUpdateNtfAction db connId (NSASMP NSASmpDelete)
-          lift . void $ getNtfSMPWorker True c smpServer
-        _ -> notifyInternalError c connId "NSCSmpDelete - no rcv queue"
-    NSCNtfWorker ntfServer -> lift . void $ getNtfNTFWorker True c ntfServer
-    NSCNtfSMPWorker smpServer -> lift . void $ getNtfSMPWorker True c smpServer
+      (cErrs, rqs) <- lift $ partitionErrs id connIds' <$> withStoreBatch c (\db -> map (getQueue db) connIds')
+      logInfo $ "processNtfCmd, NSCSmpDelete - length rqs = " <> tshow (length rqs)
+      (cErrs', _) <- lift $ partitionErrs qConnId rqs <$> withStoreBatch' c (\db -> map (updateAction db) rqs)
+      notifyErrs c (cErrs <> cErrs')
+      kickSMPWorkers rqs
+      where
+        getQueue :: DB.Connection -> ConnId -> IO (Either AgentErrorType RcvQueue)
+        getQueue db connId = first storeError <$> getPrimaryRcvQueue db connId
+        updateAction :: DB.Connection -> RcvQueue -> IO ()
+        updateAction db rq = supervisorUpdateNtfAction db (qConnId rq) (NSASMP NSASmpDelete)
+    NSCDeleteSub -> void $ lift $ withStoreBatch' c $ \db -> map (deleteNtfSubscription' db) connIds'
+  where
+    kickSMPWorkers :: [RcvQueue] -> AM ()
+    kickSMPWorkers rqs = do
+      let smpServers = S.fromList $ map qServer rqs
+      lift $ mapM_ (getNtfSMPWorker True c) smpServers
 
 getNtfNTFWorker :: Bool -> AgentClient -> NtfServer -> AM' Worker
 getNtfNTFWorker hasWork c server = do
@@ -143,6 +198,11 @@ getNtfSMPWorker hasWork c server = do
   ws <- asks $ ntfSMPWorkers . ntfSupervisor
   getAgentWorker "ntf_smp" hasWork c server ws $ runNtfSMPWorker c server
 
+getNtfTknDelWorker :: Bool -> AgentClient -> NtfServer -> AM' Worker
+getNtfTknDelWorker hasWork c server = do
+  ws <- asks $ ntfTknDelWorkers . ntfSupervisor
+  getAgentWorker "ntf_tkn_del" hasWork c server ws $ runNtfTknDelWorker c server
+
 withTokenServer :: (NtfServer -> AM ()) -> AM ()
 withTokenServer action = lift getNtfToken >>= mapM_ (\NtfToken {ntfServer} -> action ntfServer)
 
@@ -153,153 +213,288 @@ runNtfWorker c srv Worker {doWork} =
     ExceptT $ agentOperationBracket c AONtfNetwork throwWhenInactive $ runExceptT runNtfOperation
   where
     runNtfOperation :: AM ()
-    runNtfOperation =
-      withWork c doWork (`getNextNtfSubNTFAction` srv) $
-        \nextSub@(NtfSubscription {connId}, _, _) -> do
-          logInfo $ "runNtfWorker, nextSub " <> tshow nextSub
-          ri <- asks $ reconnectInterval . config
-          withRetryInterval ri $ \_ loop -> do
-            liftIO $ waitWhileSuspended c
-            liftIO $ waitForUserNetwork c
-            processSub nextSub
-              `catchAgentError` retryOnError c "NtfWorker" loop (workerInternalError c connId . show)
-    processSub :: (NtfSubscription, NtfSubNTFAction, NtfActionTs) -> AM ()
-    processSub (sub@NtfSubscription {userId, connId, smpServer, ntfSubId}, action, actionTs) = do
-      ts <- liftIO getCurrentTime
-      unlessM (lift $ rescheduleAction doWork ts actionTs) $
-        case action of
-          NSACreate ->
-            lift getNtfToken >>= \case
-              Just tkn@NtfToken {ntfServer, ntfTokenId = Just tknId, ntfTknStatus = NTActive, ntfMode = NMInstant} -> do
-                RcvQueue {clientNtfCreds} <- withStore c (`getPrimaryRcvQueue` connId)
-                case clientNtfCreds of
-                  Just ClientNtfCreds {ntfPrivateKey, notifierId} -> do
-                    atomically $ incNtfServerStat c userId ntfServer ntfCreateAttempts
-                    nSubId <- agentNtfCreateSubscription c tknId tkn (SMPQueueNtf smpServer notifierId) ntfPrivateKey
-                    atomically $ incNtfServerStat c userId ntfServer ntfCreated
-                    -- possible improvement: smaller retry until Active, less frequently (daily?) once Active
-                    let actionTs' = addUTCTime 30 ts
-                    withStore' c $ \db ->
-                      updateNtfSubscription db sub {ntfSubId = Just nSubId, ntfSubStatus = NASCreated NSNew} (NSANtf NSACheck) actionTs'
-                  _ -> workerInternalError c connId "NSACreate - no notifier queue credentials"
-              _ -> workerInternalError c connId "NSACreate - no active token"
-          NSACheck ->
-            lift getNtfToken >>= \case
-              Just tkn@NtfToken {ntfServer} ->
-                case ntfSubId of
-                  Just nSubId -> do
-                    atomically $ incNtfServerStat c userId ntfServer ntfCheckAttempts
-                    agentNtfCheckSubscription c nSubId tkn >>= \case
-                      NSAuth -> do
-                        withStore' c $ \db ->
-                          updateNtfSubscription db sub {ntfServer, ntfQueueId = Nothing, ntfSubId = Nothing, ntfSubStatus = NASNew} (NSASMP NSASmpKey) ts
-                        ns <- asks ntfSupervisor
-                        atomically $ writeTBQueue (ntfSubQ ns) (connId, NSCNtfSMPWorker smpServer)
-                      status -> updateSubNextCheck ts status
-                    atomically $ incNtfServerStat c userId ntfServer ntfChecked
-                  Nothing -> workerInternalError c connId "NSACheck - no subscription ID"
-              _ -> workerInternalError c connId "NSACheck - no active token"
-          NSADelete ->
-            deleteNtfSub $ do
-              let sub' = sub {ntfSubId = Nothing, ntfSubStatus = NASOff}
-              withStore' c $ \db -> updateNtfSubscription db sub' (NSASMP NSASmpDelete) ts
-              ns <- asks ntfSupervisor
-              atomically $ writeTBQueue (ntfSubQ ns) (connId, NSCNtfSMPWorker smpServer)
-          NSARotate ->
-            deleteNtfSub $ do
-              withStore' c $ \db -> deleteNtfSubscription db connId
-              ns <- asks ntfSupervisor
-              atomically $ writeTBQueue (ntfSubQ ns) (connId, NSCCreate)
+    runNtfOperation = do
+      ntfBatchSize <- asks $ ntfBatchSize . config
+      withWorkItems c doWork (\db -> getNextNtfSubNTFActions db srv ntfBatchSize) $ \nextSubs -> do
+        logInfo $ "runNtfWorker - length nextSubs = " <> tshow (length nextSubs)
+        currTs <- liftIO getCurrentTime
+        let (creates, checks, deletes, rotates) = splitActions currTs nextSubs
+        if null creates && null checks && null deletes && null rotates
+          then
+            let (_, _, firstActionTs) = L.head nextSubs
+             in lift $ rescheduleWork doWork currTs firstActionTs
+          else do
+            retrySubActions c creates createSubs
+            retrySubActions c checks checkSubs
+            retrySubActions c deletes deleteSubs
+            retrySubActions c rotates rotateSubs
+    splitActions :: UTCTime -> NonEmpty (NtfSubNTFAction, NtfSubscription, NtfActionTs) -> ([NtfSubscription], [NtfSubscription], [NtfSubscription], [NtfSubscription])
+    splitActions currTs = foldr addAction ([], [], [], [])
       where
-        deleteNtfSub continue = case ntfSubId of
-          Just nSubId ->
-            lift getNtfToken >>= \case
-              Just tkn@NtfToken {ntfServer} -> do
-                atomically $ incNtfServerStat c userId ntfServer ntfDelAttempts
-                tryAgentError (agentNtfDeleteSubscription c nSubId tkn) >>= \case
-                  Left e | temporaryOrHostError e -> throwE e
-                  _ -> continue
+        addAction (cmd, sub, ts) acc@(creates, checks, deletes, rotates) = case cmd of
+          NSACreate -> (sub : creates, checks, deletes, rotates)
+          NSACheck
+            | ts <= currTs -> (creates, sub : checks, deletes, rotates)
+            | otherwise -> acc
+          NSADelete -> (creates, checks, sub : deletes, rotates)
+          NSARotate -> (creates, checks, deletes, sub : rotates)
+    createSubs :: [NtfSubscription] -> AM' [NtfSubscription]
+    createSubs ntfSubs =
+      getNtfToken >>= \case
+        Just tkn@NtfToken {ntfServer, ntfTokenId = Just tknId, ntfTknStatus = NTActive, ntfMode = NMInstant} -> do
+          subsRqs_ <- zip ntfSubs <$> withStoreBatch c (\db -> map (getQueue db) ntfSubs)
+          let (errs1, subs_, newSubs_) = splitSubs tknId subsRqs_
+          incStatByUserId ntfServer ntfCreateAttempts subs_
+          case (L.nonEmpty subs_, L.nonEmpty newSubs_) of
+            (Just subs, Just newSubs) -> do
+              rs <- L.zip subs <$> agentNtfCreateSubscriptions c tkn newSubs
+              let (ntfSubs', errs2, nSubIds) = splitResults $ L.toList rs
+                  subs' = map fst nSubIds
+                  errs2' = map (first ntfSubConnId) errs2
+              incStatByUserId ntfServer ntfCreated subs'
+              ts <- liftIO getCurrentTime
+              int <- asks $ ntfSubFirstCheckInterval . config
+              let checkTs = addUTCTime int ts
+              (errs3, _) <- partitionErrs ntfSubConnId subs' <$> withStoreBatch' c (\db -> map (updateSubNSACheck db checkTs) nSubIds)
+              workerErrors c $ errs1 <> errs2' <> errs3
+              pure ntfSubs'
+            _ -> workerErrors c errs1 $> []
+        _ -> do
+          let errs = map (\sub -> (ntfSubConnId sub, INTERNAL "NSACreate - no active token")) ntfSubs
+          workerErrors c errs
+          pure []
+      where
+        getQueue :: DB.Connection -> NtfSubscription -> IO (Either AgentErrorType RcvQueue)
+        getQueue db NtfSubscription {connId} = first storeError <$> getPrimaryRcvQueue db connId
+        splitSubs :: NtfTokenId -> [(NtfSubscription, Either AgentErrorType RcvQueue)] -> ([(ConnId, AgentErrorType)], [NtfSubscription], [NewNtfEntity 'Subscription])
+        splitSubs tknId = foldr splitSub ([], [], [])
+          where
+            splitSub (sub, rq) (errs, subs, newSubs) = case rq of
+              Right RcvQueue {clientNtfCreds = Just creds} -> (errs, sub : subs, toNewSub sub creds : newSubs)
+              Right _ -> ((ntfSubConnId sub, INTERNAL "NSACreate - no notifier queue credentials") : errs, subs, newSubs)
+              Left e -> ((ntfSubConnId sub, e) : errs, subs, newSubs)
+            toNewSub NtfSubscription {smpServer} ClientNtfCreds {ntfPrivateKey, notifierId} =
+              NewNtfSub tknId (SMPQueueNtf smpServer notifierId) ntfPrivateKey
+        updateSubNSACheck :: DB.Connection -> UTCTime -> (NtfSubscription, NtfSubscriptionId) -> IO ()
+        updateSubNSACheck db checkTs (sub, nSubId) = updateNtfSubscription db sub {ntfSubId = Just nSubId, ntfSubStatus = NASCreated NSNew} (NSANtf NSACheck) checkTs
+    checkSubs :: [NtfSubscription] -> AM' [NtfSubscription]
+    checkSubs ntfSubs =
+      getNtfToken >>= \case
+        Just tkn@NtfToken {ntfServer, ntfTknStatus = NTActive, ntfMode = NMInstant} -> do
+          let (errs1, subs_, subIds_) = splitSubs ntfSubs
+          incStatByUserId ntfServer ntfCheckAttempts subs_
+          case (L.nonEmpty subs_, L.nonEmpty subIds_) of
+            (Just subs, Just subIds) -> do
+              rs <- L.zip subs <$> agentNtfCheckSubscriptions c tkn subIds
+              let (ntfSubs', errs2, nSubStatuses) = splitResults $ L.toList rs
+                  subs' = map fst nSubStatuses
+                  (errs2', authSubs) = partitionEithers $ map (\case (sub, NTF _ SMP.AUTH) -> Right sub; e -> Left $ first ntfSubConnId e) errs2
+              incStatByUserId ntfServer ntfChecked subs'
+              ts <- liftIO getCurrentTime
+              int <- asks $ ntfSubCheckInterval . config
+              let nextCheckTs = addUTCTime int ts
+              (errs3, srvs) <- partitionErrs ntfSubConnId subs' <$> withStoreBatch' c (\db -> map (updateSub db ntfServer ts nextCheckTs) nSubStatuses)
+              (errs4, srvs') <- partitionErrs ntfSubConnId authSubs <$> withStoreBatch' c (\db -> map (recreateNtfSub db ntfServer ts) authSubs)
+              mapM_ (getNtfSMPWorker True c) $ S.fromList (catMaybes srvs <> srvs')
+              workerErrors c $ errs1 <> errs2' <> errs3 <> errs4
+              pure ntfSubs'
+            _ -> workerErrors c errs1 $> []
+        _ -> do
+          let errs = map (\sub -> (ntfSubConnId sub, INTERNAL "NSACheck - no active token")) ntfSubs
+          workerErrors c errs
+          pure []
+      where
+        splitSubs :: [NtfSubscription] -> ([(ConnId, AgentErrorType)], [NtfSubscription], [NtfSubscriptionId])
+        splitSubs = foldr splitSub ([], [], [])
+          where
+            splitSub sub (errs, subs, subIds) = case sub of
+              NtfSubscription {ntfSubId = Just subId} -> (errs, sub : subs, subId : subIds)
+              _ -> ((ntfSubConnId sub, INTERNAL "NSACheck - no subscription ID") : errs, subs, subIds)
+        updateSub :: DB.Connection -> NtfServer -> UTCTime -> UTCTime -> (NtfSubscription, NtfSubStatus) -> IO (Maybe SMPServer)
+        updateSub db ntfServer ts nextCheckTs (sub, status)
+          | ntfShouldSubscribe status =
+              let sub' = sub {ntfSubStatus = NASCreated status}
+               in Nothing <$ updateNtfSubscription db sub' (NSANtf NSACheck) nextCheckTs
+          -- ntf server stopped subscribing to this queue
+          | otherwise = Just <$> recreateNtfSub db ntfServer ts sub
+        recreateNtfSub :: DB.Connection -> NtfServer -> UTCTime -> NtfSubscription -> IO SMPServer
+        recreateNtfSub db ntfServer ts sub@NtfSubscription {smpServer} =
+          let sub' = sub {ntfServer, ntfQueueId = Nothing, ntfSubId = Nothing, ntfSubStatus = NASNew}
+           in smpServer <$ updateNtfSubscription db sub' (NSASMP NSASmpKey) ts
+    incStatByUserId :: NtfServer -> (AgentNtfServerStats -> TVar Int) -> [NtfSubscription] -> AM' ()
+    incStatByUserId ntfServer sel ss =
+      forM_ (M.assocs userIdsCounts) $ \(userId, count) ->
+        atomically $ incNtfServerStat' c userId ntfServer sel count
+      where
+        userIdsCounts = foldl' (\acc NtfSubscription {userId} -> M.insertWith (+) userId 1 acc) M.empty ss
+    -- NSADelete and NSARotate are deprecated, but their processing is kept for legacy db records;
+    -- These actions are not batched
+    deleteSubs :: [NtfSubscription] -> AM' [NtfSubscription]
+    deleteSubs ntfSubs = do
+      retrySubs_ <- mapM (runCatching deleteSub) ntfSubs
+      pure $ catMaybes retrySubs_
+      where
+        deleteSub :: NtfSubscription -> AM (Maybe NtfSubscription)
+        deleteSub sub@NtfSubscription {smpServer} =
+          deleteNtfSub sub $ do
+            let sub' = sub {ntfSubId = Nothing, ntfSubStatus = NASOff}
+            ts <- liftIO getCurrentTime
+            withStore' c $ \db -> updateNtfSubscription db sub' (NSASMP NSASmpDelete) ts
+            lift . void $ getNtfSMPWorker True c smpServer
+    rotateSubs :: [NtfSubscription] -> AM' [NtfSubscription]
+    rotateSubs ntfSubs = do
+      retrySubs_ <- mapM (runCatching rotateSub) ntfSubs
+      pure $ catMaybes retrySubs_
+      where
+        rotateSub :: NtfSubscription -> AM (Maybe NtfSubscription)
+        rotateSub sub@NtfSubscription {connId} =
+          deleteNtfSub sub $ do
+            withStore' c $ \db -> deleteNtfSubscription db connId
+            ns <- asks ntfSupervisor
+            atomically $ writeTBQueue (ntfSubQ ns) (NSCCreate, [connId])
+    runCatching :: (NtfSubscription -> AM (Maybe NtfSubscription)) -> NtfSubscription -> AM' (Maybe NtfSubscription)
+    runCatching action sub@NtfSubscription {connId} =
+      fromRight Nothing
+        <$> runExceptT (action sub `catchAgentError` \e -> workerInternalError c connId (show e) $> Nothing)
+    -- deleteNtfSub is only used in NSADelete and NSARotate, so also deprecated
+    deleteNtfSub :: NtfSubscription -> AM () -> AM (Maybe NtfSubscription)
+    deleteNtfSub sub@NtfSubscription {userId, ntfSubId} continue = case ntfSubId of
+      Just nSubId ->
+        lift getNtfToken >>= \case
+          Just tkn@NtfToken {ntfServer} -> do
+            atomically $ incNtfServerStat c userId ntfServer ntfDelAttempts
+            tryAgentError (agentNtfDeleteSubscription c nSubId tkn) >>= \case
+              Right _ -> do
                 atomically $ incNtfServerStat c userId ntfServer ntfDeleted
-              Nothing -> continue
-          _ -> continue
-        updateSubNextCheck ts toStatus = do
-          checkInterval <- asks $ ntfSubCheckInterval . config
-          let nextCheckTs = addUTCTime checkInterval ts
-          updateSub (NASCreated toStatus) (NSANtf NSACheck) nextCheckTs
-        updateSub toStatus toAction actionTs' =
-          withStore' c $ \db ->
-            updateNtfSubscription db sub {ntfSubStatus = toStatus} toAction actionTs'
+                continue'
+              Left e
+                | temporaryOrHostError e -> pure $ Just sub -- don't continue, retry
+                | otherwise -> continue'
+          Nothing -> continue'
+      _ -> continue'
+      where
+        continue' = continue $> Nothing -- continue without retry
 
 runNtfSMPWorker :: AgentClient -> SMPServer -> Worker -> AM ()
-runNtfSMPWorker c srv Worker {doWork} = do
-  env <- ask
-  forever $ do
-    waitForWork doWork
-    ExceptT . liftIO . agentOperationBracket c AONtfNetwork throwWhenInactive $
-      runReaderT (runExceptT runNtfSMPOperation) env
+runNtfSMPWorker c srv Worker {doWork} = forever $ do
+  waitForWork doWork
+  ExceptT $ agentOperationBracket c AONtfNetwork throwWhenInactive $ runExceptT runNtfSMPOperation
   where
-    runNtfSMPOperation =
-      withWork c doWork (`getNextNtfSubSMPAction` srv) $
-        \nextSub@(NtfSubscription {connId}, _, _) -> do
-          logInfo $ "runNtfSMPWorker, nextSub " <> tshow nextSub
-          ri <- asks $ reconnectInterval . config
-          withRetryInterval ri $ \_ loop -> do
-            liftIO $ waitWhileSuspended c
-            liftIO $ waitForUserNetwork c
-            processSub nextSub
-              `catchAgentError` retryOnError c "NtfSMPWorker" loop (workerInternalError c connId . show)
-    processSub :: (NtfSubscription, NtfSubSMPAction, NtfActionTs) -> AM ()
-    processSub (sub@NtfSubscription {connId, ntfServer}, smpAction, actionTs) = do
-      ts <- liftIO getCurrentTime
-      unlessM (lift $ rescheduleAction doWork ts actionTs) $
-        case smpAction of
-          NSASmpKey ->
-            lift getNtfToken >>= \case
-              Just NtfToken {ntfTknStatus = NTActive, ntfMode = NMInstant} -> do
-                rq <- withStore c (`getPrimaryRcvQueue` connId)
-                C.AuthAlg a <- asks (rcvAuthAlg . config)
-                g <- asks random
-                (ntfPublicKey, ntfPrivateKey) <- atomically $ C.generateAuthKeyPair a g
-                (rcvNtfPubDhKey, rcvNtfPrivDhKey) <- atomically $ C.generateKeyPair g
-                (notifierId, rcvNtfSrvPubDhKey) <- enableQueueNotifications c rq ntfPublicKey rcvNtfPubDhKey
-                let rcvNtfDhSecret = C.dh' rcvNtfSrvPubDhKey rcvNtfPrivDhKey
-                withStore' c $ \db -> do
-                  setRcvQueueNtfCreds db connId $ Just ClientNtfCreds {ntfPublicKey, ntfPrivateKey, notifierId, rcvNtfDhSecret}
-                  updateNtfSubscription db sub {ntfQueueId = Just notifierId, ntfSubStatus = NASKey} (NSANtf NSACreate) ts
-                ns <- asks ntfSupervisor
-                atomically $ sendNtfSubCommand ns (connId, NSCNtfWorker ntfServer)
-              _ -> workerInternalError c connId "NSASmpKey - no active token"
-          NSASmpDelete -> do
-            -- TODO should we remove it after successful removal from the server?
-            rq_ <- withStore' c $ \db -> do
-              setRcvQueueNtfCreds db connId Nothing
-              getPrimaryRcvQueue db connId
-            mapM_ (disableQueueNotifications c) rq_
-            withStore' c $ \db -> deleteNtfSubscription db connId
+    runNtfSMPOperation :: AM ()
+    runNtfSMPOperation = do
+      ntfBatchSize <- asks $ ntfBatchSize . config
+      withWorkItems c doWork (\db -> getNextNtfSubSMPActions db srv ntfBatchSize) $ \nextSubs -> do
+        logInfo $ "runNtfSMPWorker - length nextSubs = " <> tshow (length nextSubs)
+        let (creates, deletes) = splitActions nextSubs
+        retrySubActions c creates createNotifierKeys
+        retrySubActions c deletes deleteNotifierKeys
+    splitActions :: NonEmpty (NtfSubSMPAction, NtfSubscription) -> ([NtfSubscription], [NtfSubscription])
+    splitActions = foldr addAction ([], [])
+      where
+        addAction (cmd, sub) (creates, deletes) = case cmd of
+          NSASmpKey -> (sub : creates, deletes)
+          NSASmpDelete -> (creates, sub : deletes)
+    createNotifierKeys :: [NtfSubscription] -> AM' [NtfSubscription]
+    createNotifierKeys ntfSubs =
+      getNtfToken >>= \case
+        Just NtfToken {ntfTknStatus = NTActive, ntfMode = NMInstant} -> do
+          (errs1, subRqKeys) <- prepareQueueSmpKey ntfSubs
+          rs <- enableQueuesNtfs c subRqKeys
+          let (subRqKeys', errs2, successes) = splitResults rs
+              ntfSubs' = map eqnrNtfSub subRqKeys'
+              errs2' = map (first (qConnId . eqnrRq)) errs2
+          ts <- liftIO getCurrentTime
+          (errs3, srvs) <- partitionErrs (qConnId . eqnrRq . fst) successes <$> withStoreBatch' c (\db -> map (storeNtfSubCreds db ts) successes)
+          mapM_ (getNtfNTFWorker True c) $ S.fromList srvs
+          workerErrors c $ errs1 <> errs2' <> errs3
+          pure ntfSubs'
+        _ -> do
+          let errs = map (\sub -> (ntfSubConnId sub, INTERNAL "NSASmpKey - no active token")) ntfSubs
+          workerErrors c errs
+          pure []
+      where
+        prepareQueueSmpKey :: [NtfSubscription] -> AM' ([(ConnId, AgentErrorType)], [EnableQueueNtfReq])
+        prepareQueueSmpKey subs = do
+          alg <- asks (rcvAuthAlg . config)
+          g <- asks random
+          partitionErrs ntfSubConnId subs <$> withStoreBatch c (\db -> map (getQueue db alg g) subs)
+          where
+            getQueue :: DB.Connection -> C.AuthAlg -> TVar ChaChaDRG -> NtfSubscription -> IO (Either AgentErrorType EnableQueueNtfReq)
+            getQueue db (C.AuthAlg a) g sub = fmap (first storeError) $ runExceptT $ do
+              rq <- ExceptT $ getPrimaryRcvQueue db (ntfSubConnId sub)
+              authKeyPair <- atomically $ C.generateAuthKeyPair a g
+              rcvNtfKeyPair <- atomically $ C.generateKeyPair g
+              pure (EnableQueueNtfReq sub rq authKeyPair rcvNtfKeyPair)
+        storeNtfSubCreds :: DB.Connection -> UTCTime -> (EnableQueueNtfReq, (SMP.NotifierId, SMP.RcvNtfPublicDhKey)) -> IO NtfServer
+        storeNtfSubCreds db ts (EnableQueueNtfReq {eqnrNtfSub, eqnrAuthKeyPair = (ntfPublicKey, ntfPrivateKey), eqnrRcvKeyPair = (_, pk)}, (notifierId, srvPubDhKey)) = do
+          let NtfSubscription {ntfServer} = eqnrNtfSub
+              rcvNtfDhSecret = C.dh' srvPubDhKey pk
+          setRcvQueueNtfCreds db (ntfSubConnId eqnrNtfSub) $ Just ClientNtfCreds {ntfPublicKey, ntfPrivateKey, notifierId, rcvNtfDhSecret}
+          updateNtfSubscription db eqnrNtfSub {ntfQueueId = Just notifierId, ntfSubStatus = NASKey} (NSANtf NSACreate) ts
+          pure ntfServer
+    deleteNotifierKeys :: [NtfSubscription] -> AM' [NtfSubscription]
+    deleteNotifierKeys ntfSubs = do
+      (errs1, subRqs) <- partitionErrs ntfSubConnId ntfSubs <$> withStoreBatch c (\db -> map (resetCredsGetQueue db) ntfSubs)
+      rs <- disableQueuesNtfs c subRqs
+      let (subRqs', errs2, successes) = splitResults rs
+          ntfSubs' = map fst subRqs'
+          errs2' = map (first (qConnId . snd)) errs2
+          disabledRqs = map (snd . fst) successes
+      (errs3, _) <- partitionErrs qConnId disabledRqs <$> withStoreBatch' c (\db -> map (deleteSub db) disabledRqs)
+      workerErrors c $ errs1 <> errs2' <> errs3
+      pure ntfSubs'
+      where
+        resetCredsGetQueue :: DB.Connection -> NtfSubscription -> IO (Either AgentErrorType DisableQueueNtfReq)
+        resetCredsGetQueue db sub@NtfSubscription {connId} = fmap (first storeError) $ runExceptT $ do
+          liftIO $ setRcvQueueNtfCreds db connId Nothing
+          rq <- ExceptT $ getPrimaryRcvQueue db connId
+          pure (sub, rq)
+        deleteSub :: DB.Connection -> RcvQueue -> IO ()
+        deleteSub db rq = deleteNtfSubscription db (qConnId rq)
 
-rescheduleAction :: TMVar () -> UTCTime -> UTCTime -> AM' Bool
-rescheduleAction doWork ts actionTs
-  | actionTs <= ts = pure False
-  | otherwise = do
-      void . atomically $ tryTakeTMVar doWork
-      void . forkIO $ do
-        liftIO $ threadDelay' $ diffToMicroseconds $ diffUTCTime actionTs ts
-        atomically $ hasWorkToDo' doWork
-      pure True
+retrySubActions :: AgentClient -> [NtfSubscription] -> ([NtfSubscription] -> AM' [NtfSubscription]) -> AM ()
+retrySubActions _ [] _ = pure ()
+retrySubActions c subs action = do
+  v <- newTVarIO subs
+  ri <- asks $ reconnectInterval . config
+  withRetryInterval ri $ \_ loop -> do
+    liftIO $ waitWhileSuspended c
+    liftIO $ waitForUserNetwork c
+    subs' <- readTVarIO v
+    retrySubs <- lift $ action subs'
+    unless (null retrySubs) $ do
+      atomically $ writeTVar v retrySubs
+      retryNetworkLoop c loop
 
-retryOnError :: AgentClient -> Text -> AM () -> (AgentErrorType -> AM ()) -> AgentErrorType -> AM ()
-retryOnError c name loop done e = do
-  logError $ name <> " error: " <> tshow e
-  case e of
-    BROKER _ NETWORK -> retryLoop
-    BROKER _ TIMEOUT -> retryLoop
-    _ -> done e
+--                                                (temporary errs, other errs, successes)
+splitResults :: [(a, Either AgentErrorType r)] -> ([a], [(a, AgentErrorType)], [(a, r)])
+splitResults = foldr addRes ([], [], [])
   where
-    retryLoop = do
-      atomically $ endAgentOperation c AONtfNetwork
-      liftIO $ throwWhenInactive c
-      atomically $ beginAgentOperation c AONtfNetwork
-      loop
+    addRes (a, r_) (as, errs, rs) = case r_ of
+      Right r -> (as, errs, (a, r) : rs)
+      Left e
+        | temporaryOrHostError e -> (a : as, errs, rs)
+        | otherwise -> (as, (a, e) : errs, rs)
+
+rescheduleWork :: TMVar () -> UTCTime -> UTCTime -> AM' ()
+rescheduleWork doWork ts actionTs = do
+  void . atomically $ tryTakeTMVar doWork
+  void . forkIO $ do
+    liftIO $ threadDelay' $ diffToMicroseconds $ diffUTCTime actionTs ts
+    atomically $ hasWorkToDo' doWork
+
+retryNetworkLoop :: AgentClient -> AM () -> AM ()
+retryNetworkLoop c loop = do
+  atomically $ endAgentOperation c AONtfNetwork
+  liftIO $ throwWhenInactive c
+  atomically $ beginAgentOperation c AONtfNetwork
+  loop
+
+workerErrors :: AgentClient -> [(ConnId, AgentErrorType)] -> AM' ()
+workerErrors c connErrs =
+  unless (null connErrs) $ do
+    void $ withStoreBatch' c (\db -> map (setNullNtfSubscriptionAction db . fst) connErrs)
+    notifyErrs c connErrs
 
 workerInternalError :: AgentClient -> ConnId -> String -> AM ()
 workerInternalError c connId internalErrStr = do
@@ -311,6 +506,14 @@ notifyInternalError :: MonadIO m => AgentClient -> ConnId -> String -> m ()
 notifyInternalError AgentClient {subQ} connId internalErrStr = atomically $ writeTBQueue subQ ("", connId, AEvt SAEConn $ ERR $ INTERNAL internalErrStr)
 {-# INLINE notifyInternalError #-}
 
+notifyInternalError' :: MonadIO m => AgentClient -> String -> m ()
+notifyInternalError' AgentClient {subQ} internalErrStr = atomically $ writeTBQueue subQ ("", "", AEvt SAEConn $ ERR $ INTERNAL internalErrStr)
+{-# INLINE notifyInternalError' #-}
+
+notifyErrs :: MonadIO m => AgentClient -> [(ConnId, AgentErrorType)] -> m ()
+notifyErrs AgentClient {subQ} connErrs = unless (null connErrs) $ atomically $ writeTBQueue subQ ("", "", AEvt SAENone $ ERRS connErrs)
+{-# INLINE notifyErrs #-}
+
 getNtfToken :: AM' (Maybe NtfToken)
 getNtfToken = do
   tkn <- asks $ ntfTkn . ntfSupervisor
@@ -322,7 +525,7 @@ nsUpdateToken ns tkn = writeTVar (ntfTkn ns) $ Just tkn
 nsRemoveNtfToken :: NtfSupervisor -> STM ()
 nsRemoveNtfToken ns = writeTVar (ntfTkn ns) Nothing
 
-sendNtfSubCommand :: NtfSupervisor -> (ConnId, NtfSupervisorCommand) -> STM ()
+sendNtfSubCommand :: NtfSupervisor -> (NtfSupervisorCommand, NonEmpty ConnId) -> STM ()
 sendNtfSubCommand ns cmd = do
   tkn <- readTVar (ntfTkn ns)
   when (instantNotifications tkn) $ writeTBQueue (ntfSubQ ns) cmd
@@ -332,10 +535,51 @@ instantNotifications = \case
   Just NtfToken {ntfTknStatus = NTActive, ntfMode = NMInstant} -> True
   _ -> False
 
+deleteToken :: AgentClient -> NtfToken -> AM ()
+deleteToken c tkn@NtfToken {ntfServer, ntfTokenId, ntfPrivKey} = do
+  setToDelete <- withStore' c $ \db -> do
+    removeNtfToken db tkn
+    case ntfTokenId of
+      Just tknId -> addNtfTokenToDelete db ntfServer ntfPrivKey tknId $> True
+      Nothing -> pure False
+  ns <- asks ntfSupervisor
+  atomically $ nsRemoveNtfToken ns
+  when setToDelete $ void $ lift $ getNtfTknDelWorker True c ntfServer
+
+runNtfTknDelWorker :: AgentClient -> NtfServer -> Worker -> AM ()
+runNtfTknDelWorker c srv Worker {doWork} =
+  forever $ do
+    waitForWork doWork
+    ExceptT $ agentOperationBracket c AONtfNetwork throwWhenInactive $ runExceptT runNtfOperation
+  where
+    runNtfOperation :: AM ()
+    runNtfOperation =
+      withWork c doWork (`getNextNtfTokenToDelete` srv) $
+        \nextTknToDelete -> do
+          logInfo $ "runNtfTknDelWorker, nextTknToDelete " <> tshow nextTknToDelete
+          ri <- asks $ reconnectInterval . config
+          withRetryInterval ri $ \_ loop -> do
+            liftIO $ waitWhileSuspended c
+            liftIO $ waitForUserNetwork c
+            processTknToDelete nextTknToDelete `catchAgentError` retryTmpError loop nextTknToDelete
+    retryTmpError :: AM () -> NtfTokenToDelete -> AgentErrorType -> AM ()
+    retryTmpError loop (tknDbId, _, _) e = do
+      logError $ "ntf tkn del error: " <> tshow e
+      if temporaryOrHostError e
+        then retryNetworkLoop c loop
+        else do
+          withStore' c $ \db -> deleteNtfTokenToDelete db tknDbId
+          notifyInternalError' c (show e)
+    processTknToDelete :: NtfTokenToDelete -> AM ()
+    processTknToDelete (tknDbId, ntfPrivKey, tknId) = do
+      agentNtfDeleteToken c srv ntfPrivKey tknId
+      withStore' c $ \db -> deleteNtfTokenToDelete db tknDbId
+
 closeNtfSupervisor :: NtfSupervisor -> IO ()
 closeNtfSupervisor ns = do
   stopWorkers $ ntfWorkers ns
   stopWorkers $ ntfSMPWorkers ns
+  stopWorkers $ ntfTknDelWorkers ns
   where
     stopWorkers workers = atomically (swapTVar workers M.empty) >>= mapM_ (liftIO . cancelWorker)
 

src/Simplex/Messaging/Agent/Protocol.hs

@@ -198,7 +198,6 @@ import Simplex.Messaging.Protocol
     NMsgMeta,
     ProtocolServer (..),
     SMPClientVersion,
-    SMPMsgMeta,
     SMPServer,
     SMPServerWithAuth,
     SndPublicAuthKey,
@@ -277,14 +276,14 @@ supportedSMPAgentVRange = mkVersionRange minSupportedSMPAgentVersion currentSMPA
 e2eEncConnInfoLength :: VersionSMPA -> PQSupport -> Int
 e2eEncConnInfoLength v = \case
   -- reduced by 3726 (roughly the increase of message ratchet header size + key and ciphertext in reply link)
-  PQSupportOn | v >= pqdrSMPAgentVersion -> 11122
-  _ -> 14848
+  PQSupportOn | v >= pqdrSMPAgentVersion -> 11106
+  _ -> 14832
 
 e2eEncAgentMsgLength :: VersionSMPA -> PQSupport -> Int
 e2eEncAgentMsgLength v = \case
   -- reduced by 2222 (the increase of message ratchet header size)
-  PQSupportOn | v >= pqdrSMPAgentVersion -> 13634
-  _ -> 15856
+  PQSupportOn | v >= pqdrSMPAgentVersion -> 13618
+  _ -> 15840
 
 -- | SMP agent event
 type ATransmission = (ACorrId, AEntityId, AEvt)
@@ -344,6 +343,7 @@ data AEvent (e :: AEntity) where
   INFO :: PQSupport -> ConnInfo -> AEvent AEConn
   CON :: PQEncryption -> AEvent AEConn -- notification that connection is established
   END :: AEvent AEConn
+  DELD :: AEvent AEConn
   CONNECT :: AProtocolType -> TransportHost -> AEvent AENone
   DISCONNECT :: AProtocolType -> TransportHost -> AEvent AENone
   DOWN :: SMPServer -> [ConnId] -> AEvent AENone
@@ -355,7 +355,7 @@ data AEvent (e :: AEntity) where
   MERR :: AgentMsgId -> AgentErrorType -> AEvent AEConn
   MERRS :: NonEmpty AgentMsgId -> AgentErrorType -> AEvent AEConn
   MSG :: MsgMeta -> MsgFlags -> MsgBody -> AEvent AEConn
-  MSGNTF :: SMPMsgMeta -> AEvent AEConn
+  MSGNTF :: MsgId -> Maybe UTCTime -> AEvent AEConn
   RCVD :: MsgMeta -> NonEmpty MsgReceipt -> AEvent AEConn
   QCONT :: AEvent AEConn
   DEL_RCVQ :: SMPServer -> SMP.RecipientId -> Maybe AgentErrorType -> AEvent AEConn
@@ -365,6 +365,7 @@ data AEvent (e :: AEntity) where
   OK :: AEvent AEConn
   JOINED :: SndQueueSecured -> AEvent AEConn
   ERR :: AgentErrorType -> AEvent AEConn
+  ERRS :: [(ConnId, AgentErrorType)] -> AEvent AENone
   SUSPENDED :: AEvent AENone
   RFPROG :: Int64 -> Int64 -> AEvent AERcvFile
   RFDONE :: FilePath -> AEvent AERcvFile
@@ -413,6 +414,7 @@ data AEventTag (e :: AEntity) where
   INFO_ :: AEventTag AEConn
   CON_ :: AEventTag AEConn
   END_ :: AEventTag AEConn
+  DELD_ :: AEventTag AEConn
   CONNECT_ :: AEventTag AENone
   DISCONNECT_ :: AEventTag AENone
   DOWN_ :: AEventTag AENone
@@ -434,6 +436,7 @@ data AEventTag (e :: AEntity) where
   OK_ :: AEventTag AEConn
   JOINED_ :: AEventTag AEConn
   ERR_ :: AEventTag AEConn
+  ERRS_ :: AEventTag AENone
   SUSPENDED_ :: AEventTag AENone
   -- XFTP commands and responses
   RFDONE_ :: AEventTag AERcvFile
@@ -466,6 +469,7 @@ aEventTag = \case
   INFO {} -> INFO_
   CON _ -> CON_
   END -> END_
+  DELD -> DELD_
   CONNECT {} -> CONNECT_
   DISCONNECT {} -> DISCONNECT_
   DOWN {} -> DOWN_
@@ -487,6 +491,7 @@ aEventTag = \case
   OK -> OK_
   JOINED _ -> JOINED_
   ERR _ -> ERR_
+  ERRS _ -> ERRS_
   SUSPENDED -> SUSPENDED_
   RFPROG {} -> RFPROG_
   RFDONE {} -> RFDONE_

src/Simplex/Messaging/Agent/Store/SQLite.hs

@@ -122,6 +122,7 @@ module Simplex.Messaging.Agent.Store.SQLite
     getRcvMsg,
     getLastMsg,
     checkRcvMsgHashExists,
+    getRcvMsgBrokerTs,
     deleteMsg,
     deleteDeliveredSndMsg,
     deleteSndMsgDelivery,
@@ -150,7 +151,15 @@ module Simplex.Messaging.Agent.Store.SQLite
     updateNtfMode,
     updateNtfToken,
     removeNtfToken,
+    addNtfTokenToDelete,
+    deleteExpiredNtfTokensToDelete,
+    NtfTokenToDelete,
+    getNextNtfTokenToDelete,
+    markNtfTokenToDeleteFailed_, -- exported for tests
+    getPendingDelTknServers,
+    deleteNtfTokenToDelete,
     -- Notification subscription persistence
+    NtfSupervisorSub,
     getNtfSubscription,
     createNtfSubscription,
     supervisorUpdateNtfSub,
@@ -158,9 +167,10 @@ module Simplex.Messaging.Agent.Store.SQLite
     updateNtfSubscription,
     setNullNtfSubscriptionAction,
     deleteNtfSubscription,
-    getNextNtfSubNTFAction,
+    deleteNtfSubscription',
+    getNextNtfSubNTFActions,
     markNtfSubActionNtfFailed_, -- exported for tests
-    getNextNtfSubSMPAction,
+    getNextNtfSubSMPActions,
     markNtfSubActionSMPFailed_, -- exported for tests
     getActiveNtfToken,
     getNtfRcvQueue,
@@ -966,10 +976,11 @@ updateRcvIds db connId = do
   pure (internalId, internalRcvId, lastExternalSndId, lastRcvHash)
 
 createRcvMsg :: DB.Connection -> ConnId -> RcvQueue -> RcvMsgData -> IO ()
-createRcvMsg db connId rq rcvMsgData@RcvMsgData {msgMeta = MsgMeta {sndMsgId}, internalRcvId, internalHash} = do
+createRcvMsg db connId rq@RcvQueue {dbQueueId} rcvMsgData@RcvMsgData {msgMeta = MsgMeta {sndMsgId, broker = (_, brokerTs)}, internalRcvId, internalHash} = do
   insertRcvMsgBase_ db connId rcvMsgData
   insertRcvMsgDetails_ db connId rq rcvMsgData
   updateRcvMsgHash db connId sndMsgId internalRcvId internalHash
+  DB.execute db "UPDATE rcv_queues SET last_broker_ts = ? WHERE conn_id = ? AND rcv_queue_id = ?" (brokerTs, connId, dbQueueId)
 
 updateSndIds :: DB.Connection -> ConnId -> IO (Either StoreError (InternalId, InternalSndId, PrevSndMsgHash))
 updateSndIds db connId = runExceptT $ do
@@ -1061,17 +1072,26 @@ getPendingQueueMsg db connId SndQueue {dbQueueId} =
 
 getWorkItem :: Show i => ByteString -> IO (Maybe i) -> (i -> IO (Either StoreError a)) -> (i -> IO ()) -> IO (Either StoreError (Maybe a))
 getWorkItem itemName getId getItem markFailed =
-  runExceptT $ handleErr "getId" getId >>= mapM tryGetItem
+  runExceptT $ handleWrkErr itemName "getId" getId >>= mapM (tryGetItem itemName getItem markFailed)
+
+getWorkItems :: Show i => ByteString -> IO [i] -> (i -> IO (Either StoreError a)) -> (i -> IO ()) -> IO (Either StoreError [Either StoreError a])
+getWorkItems itemName getIds getItem markFailed =
+  runExceptT $ handleWrkErr itemName "getIds" getIds >>= mapM (tryE . tryGetItem itemName getItem markFailed)
+
+tryGetItem :: Show i => ByteString -> (i -> IO (Either StoreError a)) -> (i -> IO ()) -> i -> ExceptT StoreError IO a
+tryGetItem itemName getItem markFailed itemId = ExceptT (getItem itemId) `catchStoreError` \e -> mark >> throwE e
   where
-    tryGetItem itemId = ExceptT (getItem itemId) `catchStoreErrors` \e -> mark itemId >> throwE e
-    mark itemId = handleErr ("markFailed ID " <> bshow itemId) $ markFailed itemId
-    catchStoreErrors = catchAllErrors (SEInternal . bshow)
-    -- Errors caught by this function will suspend worker as if there is no more work,
-    handleErr :: ByteString -> IO a -> ExceptT StoreError IO a
-    handleErr opName action = ExceptT $ first mkError <$> E.try action
-      where
-        mkError :: E.SomeException -> StoreError
-        mkError e = SEWorkItemError $ itemName <> " " <> opName <> " error: " <> bshow e
+    mark = handleWrkErr itemName ("markFailed ID " <> bshow itemId) $ markFailed itemId
+
+catchStoreError :: ExceptT StoreError IO a -> (StoreError -> ExceptT StoreError IO a) -> ExceptT StoreError IO a
+catchStoreError = catchAllErrors (SEInternal . bshow)
+
+-- Errors caught by this function will suspend worker as if there is no more work,
+handleWrkErr :: ByteString -> ByteString -> IO a -> ExceptT StoreError IO a
+handleWrkErr itemName opName action = ExceptT $ first mkError <$> E.try action
+  where
+    mkError :: E.SomeException -> StoreError
+    mkError e = SEWorkItemError $ itemName <> " " <> opName <> " error: " <> bshow e
 
 updatePendingMsgRIState :: DB.Connection -> ConnId -> InternalId -> RI2State -> IO ()
 updatePendingMsgRIState db connId msgId RI2State {slowInterval, fastInterval} =
@@ -1167,6 +1187,11 @@ checkRcvMsgHashExists db connId hash = do
           (connId, hash)
       )
 
+getRcvMsgBrokerTs :: DB.Connection -> ConnId -> SMP.MsgId -> IO (Either StoreError BrokerTs)
+getRcvMsgBrokerTs db connId msgId =
+  firstRow fromOnly SEMsgNotFound $
+    DB.query db "SELECT broker_ts FROM rcv_messages WHERE conn_id = ? AND broker_id = ?" (connId, msgId)
+
 deleteMsg :: DB.Connection -> ConnId -> InternalId -> IO ()
 deleteMsg db connId msgId =
   DB.execute db "DELETE FROM messages WHERE conn_id = ? AND internal_id = ?;" (connId, msgId)
@@ -1475,7 +1500,73 @@ removeNtfToken db NtfToken {deviceToken = DeviceToken provider token, ntfServer
     |]
     (provider, token, host, port)
 
-getNtfSubscription :: DB.Connection -> ConnId -> IO (Maybe (NtfSubscription, Maybe (NtfSubAction, NtfActionTs)))
+addNtfTokenToDelete :: DB.Connection -> NtfServer -> C.APrivateAuthKey -> NtfTokenId -> IO ()
+addNtfTokenToDelete db ProtocolServer {host, port, keyHash} ntfPrivKey tknId =
+  DB.execute db "INSERT INTO ntf_tokens_to_delete (ntf_host, ntf_port, ntf_key_hash, tkn_id, tkn_priv_key) VALUES (?,?,?,?,?)" (host, port, keyHash, tknId, ntfPrivKey)
+
+deleteExpiredNtfTokensToDelete :: DB.Connection -> NominalDiffTime -> IO ()
+deleteExpiredNtfTokensToDelete db ttl = do
+  cutoffTs <- addUTCTime (-ttl) <$> getCurrentTime
+  DB.execute db "DELETE FROM ntf_tokens_to_delete WHERE created_at < ?" (Only cutoffTs)
+
+type NtfTokenToDelete = (Int64, C.APrivateAuthKey, NtfTokenId)
+
+getNextNtfTokenToDelete :: DB.Connection -> NtfServer -> IO (Either StoreError (Maybe NtfTokenToDelete))
+getNextNtfTokenToDelete db (NtfServer ntfHost ntfPort _) =
+  getWorkItem "ntf tkn del" getNtfTknDbId getNtfTknToDelete (markNtfTokenToDeleteFailed_ db)
+  where
+    getNtfTknDbId :: IO (Maybe Int64)
+    getNtfTknDbId =
+      maybeFirstRow fromOnly $
+        DB.query
+          db
+          [sql|
+            SELECT ntf_token_to_delete_id
+            FROM ntf_tokens_to_delete
+            WHERE ntf_host = ? AND ntf_port = ?
+              AND del_failed = 0
+            ORDER BY created_at ASC
+            LIMIT 1
+          |]
+          (ntfHost, ntfPort)
+    getNtfTknToDelete :: Int64 -> IO (Either StoreError NtfTokenToDelete)
+    getNtfTknToDelete tknDbId =
+      firstRow ntfTokenToDelete err $
+        DB.query
+          db
+          [sql|
+            SELECT tkn_priv_key, tkn_id
+            FROM ntf_tokens_to_delete
+            WHERE ntf_token_to_delete_id = ?
+          |]
+          (Only tknDbId)
+      where
+        err = SEInternal $ "ntf token to delete " <> bshow tknDbId <> " returned []"
+        ntfTokenToDelete (tknPrivKey, tknId) = (tknDbId, tknPrivKey, tknId)
+
+markNtfTokenToDeleteFailed_ :: DB.Connection -> Int64 -> IO ()
+markNtfTokenToDeleteFailed_ db tknDbId =
+  DB.execute db "UPDATE ntf_tokens_to_delete SET del_failed = 1 where ntf_token_to_delete_id = ?" (Only tknDbId)
+
+getPendingDelTknServers :: DB.Connection -> IO [NtfServer]
+getPendingDelTknServers db =
+  map toNtfServer
+    <$> DB.query_
+      db
+      [sql|
+        SELECT DISTINCT ntf_host, ntf_port, ntf_key_hash
+        FROM ntf_tokens_to_delete
+      |]
+  where
+    toNtfServer (host, port, keyHash) = NtfServer host port keyHash
+
+deleteNtfTokenToDelete :: DB.Connection -> Int64 -> IO ()
+deleteNtfTokenToDelete db tknDbId =
+  DB.execute db "DELETE FROM ntf_tokens_to_delete WHERE ntf_token_to_delete_id = ?" (Only tknDbId)
+
+type NtfSupervisorSub = (NtfSubscription, Maybe (NtfSubAction, NtfActionTs))
+
+getNtfSubscription :: DB.Connection -> ConnId -> IO (Maybe NtfSupervisorSub)
 getNtfSubscription db connId =
   maybeFirstRow ntfSubscription $
     DB.query
@@ -1491,7 +1582,7 @@ getNtfSubscription db connId =
       |]
       (Only connId)
   where
-    ntfSubscription ((userId, smpHost, smpPort, smpKeyHash, ntfHost, ntfPort, ntfKeyHash ) :. (ntfQueueId, ntfSubId, ntfSubStatus, ntfAction_, smpAction_, actionTs_)) =
+    ntfSubscription ((userId, smpHost, smpPort, smpKeyHash, ntfHost, ntfPort, ntfKeyHash) :. (ntfQueueId, ntfSubId, ntfSubStatus, ntfAction_, smpAction_, actionTs_)) =
       let smpServer = SMPServer smpHost smpPort smpKeyHash
           ntfServer = NtfServer ntfHost ntfPort ntfKeyHash
           action = case (ntfAction_, smpAction_, actionTs_) of
@@ -1521,16 +1612,19 @@ createNtfSubscription db ntfSubscription action = runExceptT $ do
     (ntfSubAction, ntfSubSMPAction) = ntfSubAndSMPAction action
 
 supervisorUpdateNtfSub :: DB.Connection -> NtfSubscription -> NtfSubAction -> IO ()
-supervisorUpdateNtfSub db NtfSubscription {connId, ntfQueueId, ntfServer = (NtfServer ntfHost ntfPort _), ntfSubId, ntfSubStatus} action = do
+supervisorUpdateNtfSub db NtfSubscription {connId, smpServer = (SMPServer smpHost smpPort _), ntfQueueId, ntfServer = (NtfServer ntfHost ntfPort _), ntfSubId, ntfSubStatus} action = do
   ts <- getCurrentTime
   DB.execute
     db
     [sql|
       UPDATE ntf_subscriptions
-      SET smp_ntf_id = ?, ntf_host = ?, ntf_port = ?, ntf_sub_id = ?, ntf_sub_status = ?, ntf_sub_action = ?, ntf_sub_smp_action = ?, ntf_sub_action_ts = ?, updated_by_supervisor = ?, updated_at = ?
+      SET smp_host = ?, smp_port = ?, smp_ntf_id = ?, ntf_host = ?, ntf_port = ?, ntf_sub_id = ?,
+          ntf_sub_status = ?, ntf_sub_action = ?, ntf_sub_smp_action = ?, ntf_sub_action_ts = ?, updated_by_supervisor = ?, updated_at = ?
       WHERE conn_id = ?
     |]
-    ((ntfQueueId, ntfHost, ntfPort, ntfSubId) :. (ntfSubStatus, ntfSubAction, ntfSubSMPAction, ts, True, ts, connId))
+    ( (smpHost, smpPort, ntfQueueId, ntfHost, ntfPort, ntfSubId)
+        :. (ntfSubStatus, ntfSubAction, ntfSubSMPAction, ts, True, ts, connId)
+    )
   where
     (ntfSubAction, ntfSubSMPAction) = ntfSubAndSMPAction action
 
@@ -1605,16 +1699,20 @@ deleteNtfSubscription db connId = do
             WHERE conn_id = ?
           |]
           (Nothing :: Maybe SMP.NotifierId, Nothing :: Maybe NtfSubscriptionId, NASDeleted, False, updatedAt, connId)
-      else DB.execute db "DELETE FROM ntf_subscriptions WHERE conn_id = ?" (Only connId)
+      else deleteNtfSubscription' db connId
 
-getNextNtfSubNTFAction :: DB.Connection -> NtfServer -> IO (Either StoreError (Maybe (NtfSubscription, NtfSubNTFAction, NtfActionTs)))
-getNextNtfSubNTFAction db ntfServer@(NtfServer ntfHost ntfPort _) =
-  getWorkItem "ntf NTF" getNtfConnId getNtfSubAction (markNtfSubActionNtfFailed_ db)
+deleteNtfSubscription' :: DB.Connection -> ConnId -> IO ()
+deleteNtfSubscription' db connId = do
+  DB.execute db "DELETE FROM ntf_subscriptions WHERE conn_id = ?" (Only connId)
+
+getNextNtfSubNTFActions :: DB.Connection -> NtfServer -> Int -> IO (Either StoreError [Either StoreError (NtfSubNTFAction, NtfSubscription, NtfActionTs)])
+getNextNtfSubNTFActions db ntfServer@(NtfServer ntfHost ntfPort _) ntfBatchSize =
+  getWorkItems "ntf NTF" getNtfConnIds getNtfSubAction (markNtfSubActionNtfFailed_ db)
   where
-    getNtfConnId :: IO (Maybe ConnId)
-    getNtfConnId =
-      maybeFirstRow fromOnly $
-        DB.query
+    getNtfConnIds :: IO [ConnId]
+    getNtfConnIds =
+      map fromOnly
+        <$> DB.query
           db
           [sql|
             SELECT conn_id
@@ -1622,10 +1720,10 @@ getNextNtfSubNTFAction db ntfServer@(NtfServer ntfHost ntfPort _) =
             WHERE ntf_host = ? AND ntf_port = ? AND ntf_sub_action IS NOT NULL
               AND (ntf_failed = 0 OR updated_by_supervisor = 1)
             ORDER BY ntf_sub_action_ts ASC
-            LIMIT 1
+            LIMIT ?
           |]
-          (ntfHost, ntfPort)
-    getNtfSubAction :: ConnId -> IO (Either StoreError (NtfSubscription, NtfSubNTFAction, NtfActionTs))
+          (ntfHost, ntfPort, ntfBatchSize)
+    getNtfSubAction :: ConnId -> IO (Either StoreError (NtfSubNTFAction, NtfSubscription, NtfActionTs))
     getNtfSubAction connId = do
       markUpdatedByWorker db connId
       firstRow ntfSubAction err $
@@ -1645,20 +1743,20 @@ getNextNtfSubNTFAction db ntfServer@(NtfServer ntfHost ntfPort _) =
         ntfSubAction (userId, smpHost, smpPort, smpKeyHash, ntfQueueId, ntfSubId, ntfSubStatus, actionTs, action) =
           let smpServer = SMPServer smpHost smpPort smpKeyHash
               ntfSubscription = NtfSubscription {userId, connId, smpServer, ntfQueueId, ntfServer, ntfSubId, ntfSubStatus}
-           in (ntfSubscription, action, actionTs)
+           in (action, ntfSubscription, actionTs)
 
 markNtfSubActionNtfFailed_ :: DB.Connection -> ConnId -> IO ()
 markNtfSubActionNtfFailed_ db connId =
   DB.execute db "UPDATE ntf_subscriptions SET ntf_failed = 1 where conn_id = ?" (Only connId)
 
-getNextNtfSubSMPAction :: DB.Connection -> SMPServer -> IO (Either StoreError (Maybe (NtfSubscription, NtfSubSMPAction, NtfActionTs)))
-getNextNtfSubSMPAction db smpServer@(SMPServer smpHost smpPort _) =
-  getWorkItem "ntf SMP" getNtfConnId getNtfSubAction (markNtfSubActionSMPFailed_ db)
+getNextNtfSubSMPActions :: DB.Connection -> SMPServer -> Int -> IO (Either StoreError [Either StoreError (NtfSubSMPAction, NtfSubscription)])
+getNextNtfSubSMPActions db smpServer@(SMPServer smpHost smpPort _) ntfBatchSize =
+  getWorkItems "ntf SMP" getNtfConnIds getNtfSubAction (markNtfSubActionSMPFailed_ db)
   where
-    getNtfConnId :: IO (Maybe ConnId)
-    getNtfConnId =
-      maybeFirstRow fromOnly $
-        DB.query
+    getNtfConnIds :: IO [ConnId]
+    getNtfConnIds =
+      map fromOnly
+        <$> DB.query
           db
           [sql|
             SELECT conn_id
@@ -1666,10 +1764,10 @@ getNextNtfSubSMPAction db smpServer@(SMPServer smpHost smpPort _) =
             WHERE smp_host = ? AND smp_port = ? AND ntf_sub_smp_action IS NOT NULL AND ntf_sub_action_ts IS NOT NULL
               AND (smp_failed = 0 OR updated_by_supervisor = 1)
             ORDER BY ntf_sub_action_ts ASC
-            LIMIT 1
+            LIMIT ?
           |]
-          (smpHost, smpPort)
-    getNtfSubAction :: ConnId -> IO (Either StoreError (NtfSubscription, NtfSubSMPAction, NtfActionTs))
+          (smpHost, smpPort, ntfBatchSize)
+    getNtfSubAction :: ConnId -> IO (Either StoreError (NtfSubSMPAction, NtfSubscription))
     getNtfSubAction connId = do
       markUpdatedByWorker db connId
       firstRow ntfSubAction err $
@@ -1677,7 +1775,7 @@ getNextNtfSubSMPAction db smpServer@(SMPServer smpHost smpPort _) =
           db
           [sql|
             SELECT c.user_id, s.ntf_host, s.ntf_port, s.ntf_key_hash,
-              ns.smp_ntf_id, ns.ntf_sub_id, ns.ntf_sub_status, ns.ntf_sub_action_ts, ns.ntf_sub_smp_action
+              ns.smp_ntf_id, ns.ntf_sub_id, ns.ntf_sub_status, ns.ntf_sub_smp_action
             FROM ntf_subscriptions ns
             JOIN connections c USING (conn_id)
             JOIN ntf_servers s USING (ntf_host, ntf_port)
@@ -1686,10 +1784,10 @@ getNextNtfSubSMPAction db smpServer@(SMPServer smpHost smpPort _) =
           (Only connId)
       where
         err = SEInternal $ "ntf subscription " <> bshow connId <> " returned []"
-        ntfSubAction (userId, ntfHost, ntfPort, ntfKeyHash, ntfQueueId, ntfSubId, ntfSubStatus, actionTs, action) =
+        ntfSubAction (userId, ntfHost, ntfPort, ntfKeyHash, ntfQueueId, ntfSubId, ntfSubStatus, action) =
           let ntfServer = NtfServer ntfHost ntfPort ntfKeyHash
               ntfSubscription = NtfSubscription {userId, connId, smpServer, ntfQueueId, ntfServer, ntfSubId, ntfSubStatus}
-           in (ntfSubscription, action, actionTs)
+           in (action, ntfSubscription)
 
 markNtfSubActionSMPFailed_ :: DB.Connection -> ConnId -> IO ()
 markNtfSubActionSMPFailed_ db connId =
@@ -1720,19 +1818,19 @@ getActiveNtfToken db =
           ntfMode = fromMaybe NMPeriodic ntfMode_
        in NtfToken {deviceToken = DeviceToken provider dt, ntfServer, ntfTokenId, ntfPubKey, ntfPrivKey, ntfDhKeys, ntfDhSecret, ntfTknStatus, ntfTknAction, ntfMode}
 
-getNtfRcvQueue :: DB.Connection -> SMPQueueNtf -> IO (Either StoreError (ConnId, RcvNtfDhSecret))
+getNtfRcvQueue :: DB.Connection -> SMPQueueNtf -> IO (Either StoreError (ConnId, RcvNtfDhSecret, Maybe UTCTime))
 getNtfRcvQueue db SMPQueueNtf {smpServer = (SMPServer host port _), notifierId} =
   firstRow' res SEConnNotFound $
     DB.query
       db
       [sql|
-        SELECT conn_id, rcv_ntf_dh_secret
+        SELECT conn_id, rcv_ntf_dh_secret, last_broker_ts
         FROM rcv_queues
         WHERE host = ? AND port = ? AND ntf_id = ? AND deleted = 0
       |]
       (host, port, notifierId)
   where
-    res (connId, Just rcvNtfDhSecret) = Right (connId, rcvNtfDhSecret)
+    res (connId, Just rcvNtfDhSecret, lastBrokerTs_) = Right (connId, rcvNtfDhSecret, lastBrokerTs_)
     res _ = Left SEConnNotFound
 
 setConnectionNtfs :: DB.Connection -> ConnId -> Bool -> IO ()
@@ -2004,7 +2102,7 @@ setConnDeleted db waitDelivery connId
       DB.execute db "UPDATE connections SET deleted = ? WHERE conn_id = ?" (True, connId)
 
 setConnUserId :: DB.Connection -> UserId -> ConnId -> UserId -> IO ()
-setConnUserId db oldUserId connId newUserId = 
+setConnUserId db oldUserId connId newUserId =
   DB.execute db "UPDATE connections SET user_id = ? WHERE conn_id = ? and user_id = ?" (newUserId, connId, oldUserId)
 
 setConnAgentVersion :: DB.Connection -> ConnId -> VersionSMPA -> IO ()

src/Simplex/Messaging/Agent/Store/SQLite/Migrations.hs

@@ -74,6 +74,8 @@ import Simplex.Messaging.Agent.Store.SQLite.Migrations.M20240225_ratchet_kem
 import Simplex.Messaging.Agent.Store.SQLite.Migrations.M20240417_rcv_files_approved_relays
 import Simplex.Messaging.Agent.Store.SQLite.Migrations.M20240624_snd_secure
 import Simplex.Messaging.Agent.Store.SQLite.Migrations.M20240702_servers_stats
+import Simplex.Messaging.Agent.Store.SQLite.Migrations.M20240930_ntf_tokens_to_delete
+import Simplex.Messaging.Agent.Store.SQLite.Migrations.M20241007_rcv_queues_last_broker_ts
 import Simplex.Messaging.Encoding.String
 import Simplex.Messaging.Parsers (dropPrefix, sumTypeJSON)
 import Simplex.Messaging.Transport.Client (TransportHost)
@@ -116,7 +118,9 @@ schemaMigrations =
     ("m20240225_ratchet_kem", m20240225_ratchet_kem, Just down_m20240225_ratchet_kem),
     ("m20240417_rcv_files_approved_relays", m20240417_rcv_files_approved_relays, Just down_m20240417_rcv_files_approved_relays),
     ("m20240624_snd_secure", m20240624_snd_secure, Just down_m20240624_snd_secure),
-    ("m20240702_servers_stats", m20240702_servers_stats, Just down_m20240702_servers_stats)
+    ("m20240702_servers_stats", m20240702_servers_stats, Just down_m20240702_servers_stats),
+    ("m20240930_ntf_tokens_to_delete", m20240930_ntf_tokens_to_delete, Just down_m20240930_ntf_tokens_to_delete),
+    ("m20241007_rcv_queues_last_broker_ts", m20241007_rcv_queues_last_broker_ts, Just down_m20241007_rcv_queues_last_broker_ts)
   ]
 
 -- | The list of migrations in ascending order by date

src/Simplex/Messaging/Agent/Store/SQLite/Migrations/M20240930_ntf_tokens_to_delete.hs

@@ -0,0 +1,27 @@
+{-# LANGUAGE QuasiQuotes #-}
+
+module Simplex.Messaging.Agent.Store.SQLite.Migrations.M20240930_ntf_tokens_to_delete where
+
+import Database.SQLite.Simple (Query)
+import Database.SQLite.Simple.QQ (sql)
+
+m20240930_ntf_tokens_to_delete :: Query
+m20240930_ntf_tokens_to_delete =
+    [sql|
+CREATE TABLE ntf_tokens_to_delete (
+  ntf_token_to_delete_id INTEGER PRIMARY KEY,
+  ntf_host TEXT NOT NULL,
+  ntf_port TEXT NOT NULL,
+  ntf_key_hash BLOB NOT NULL,
+  tkn_id BLOB NOT NULL, -- token ID assigned by notifications server
+  tkn_priv_key BLOB NOT NULL, -- client's private key to sign token commands,
+  del_failed INTEGER DEFAULT 0,
+  created_at TEXT NOT NULL DEFAULT (datetime('now'))
+);
+|]
+
+down_m20240930_ntf_tokens_to_delete :: Query
+down_m20240930_ntf_tokens_to_delete =
+    [sql|
+DROP TABLE ntf_tokens_to_delete;
+|]

src/Simplex/Messaging/Agent/Store/SQLite/Migrations/M20241007_rcv_queues_last_broker_ts.hs

@@ -0,0 +1,18 @@
+{-# LANGUAGE QuasiQuotes #-}
+
+module Simplex.Messaging.Agent.Store.SQLite.Migrations.M20241007_rcv_queues_last_broker_ts where
+
+import Database.SQLite.Simple (Query)
+import Database.SQLite.Simple.QQ (sql)
+
+m20241007_rcv_queues_last_broker_ts :: Query
+m20241007_rcv_queues_last_broker_ts =
+    [sql|
+ALTER TABLE rcv_queues ADD COLUMN last_broker_ts TEXT;
+|]
+
+down_m20241007_rcv_queues_last_broker_ts :: Query
+down_m20241007_rcv_queues_last_broker_ts =
+    [sql|
+ALTER TABLE rcv_queues DROP COLUMN last_broker_ts;
+|]

src/Simplex/Messaging/Agent/Store/SQLite/Migrations/agent_lint.sql

@@ -0,0 +1 @@
+CREATE INDEX 'ratchets_conn_id' ON 'ratchets'('conn_id'); --> connections(conn_id)

src/Simplex/Messaging/Agent/Store/SQLite/Migrations/agent_schema.sql

@@ -56,6 +56,7 @@ CREATE TABLE rcv_queues(
   switch_status TEXT,
   deleted INTEGER NOT NULL DEFAULT 0,
   snd_secure INTEGER NOT NULL DEFAULT 0,
+  last_broker_ts TEXT,
   PRIMARY KEY(host, port, rcv_id),
   FOREIGN KEY(host, port) REFERENCES servers
   ON DELETE RESTRICT ON UPDATE CASCADE,
@@ -403,6 +404,16 @@ CREATE TABLE servers_stats(
   created_at TEXT NOT NULL DEFAULT(datetime('now')),
   updated_at TEXT NOT NULL DEFAULT(datetime('now'))
 );
+CREATE TABLE ntf_tokens_to_delete(
+  ntf_token_to_delete_id INTEGER PRIMARY KEY,
+  ntf_host TEXT NOT NULL,
+  ntf_port TEXT NOT NULL,
+  ntf_key_hash BLOB NOT NULL,
+  tkn_id BLOB NOT NULL, -- token ID assigned by notifications server
+  tkn_priv_key BLOB NOT NULL, -- client's private key to sign token commands,
+del_failed INTEGER DEFAULT 0,
+created_at TEXT NOT NULL DEFAULT(datetime('now'))
+);
 CREATE UNIQUE INDEX idx_rcv_queues_ntf ON rcv_queues(host, port, ntf_id);
 CREATE UNIQUE INDEX idx_rcv_queue_id ON rcv_queues(conn_id, rcv_queue_id);
 CREATE UNIQUE INDEX idx_snd_queue_id ON snd_queues(conn_id, snd_queue_id);

src/Simplex/Messaging/Client.hs

@@ -63,11 +63,13 @@ module Simplex.Messaging.Client
     forwardSMPTransmission,
     getSMPQueueInfo,
     sendProtocolCommand,
+    sendProtocolCommands,
 
     -- * Supporting types and client configuration
     ProtocolClientError (..),
     SMPClientError,
     ProxyClientError (..),
+    Response (..),
     unexpectedResponse,
     ProtocolClientConfig (..),
     NetworkConfig (..),
@@ -80,10 +82,11 @@ module Simplex.Messaging.Client
     defaultSMPClientConfig,
     defaultNetworkConfig,
     transportClientConfig,
+    clientSocksCredentials,
     chooseTransportHost,
-    proxyUsername,
     temporaryClientError,
     smpProxyError,
+    textToHostMode,
     ServerTransmissionBatch,
     ServerTransmission (..),
     ClientCommand,
@@ -116,16 +119,20 @@ import qualified Data.Aeson.TH as J
 import qualified Data.Attoparsec.ByteString.Char8 as A
 import Data.ByteString.Char8 (ByteString)
 import qualified Data.ByteString.Char8 as B
+import qualified Data.ByteString.Base64 as B64
 import Data.Functor (($>))
 import Data.Int (Int64)
 import Data.List (find)
 import Data.List.NonEmpty (NonEmpty (..))
 import qualified Data.List.NonEmpty as L
 import Data.Maybe (catMaybes, fromMaybe)
+import Data.Text (Text)
+import qualified Data.Text as T
 import Data.Time.Clock (UTCTime (..), diffUTCTime, getCurrentTime)
 import qualified Data.X509 as X
 import qualified Data.X509.Validation as XV
 import Network.Socket (ServiceName)
+import Network.Socks5 (SocksCredentials (..))
 import Numeric.Natural
 import qualified Simplex.Messaging.Crypto as C
 import Simplex.Messaging.Encoding
@@ -136,7 +143,7 @@ import Simplex.Messaging.Server.QueueStore.QueueInfo
 import Simplex.Messaging.TMap (TMap)
 import qualified Simplex.Messaging.TMap as TM
 import Simplex.Messaging.Transport
-import Simplex.Messaging.Transport.Client (SocksProxy, TransportClientConfig (..), TransportHost (..), defaultTcpConnectTimeout, runTransportClient)
+import Simplex.Messaging.Transport.Client (SocksAuth (..), SocksProxyWithAuth (..), TransportClientConfig (..), TransportHost (..), defaultSMPPort, defaultTcpConnectTimeout, runTransportClient)
 import Simplex.Messaging.Transport.KeepAlive
 import Simplex.Messaging.Transport.WebSockets (WS)
 import Simplex.Messaging.Util (bshow, diffToMicroseconds, ifM, liftEitherWith, raceAny_, threadDelay', tshow, whenM)
@@ -192,6 +199,7 @@ smpClientStub g sessionId thVersion thAuth = do
               thAuth,
               blockSize = smpBlockSize,
               implySessId = thVersion >= authCmdsSMPVersion,
+              encryptBlock = Nothing,
               batch = True
             },
         sessionTs = ts,
@@ -236,6 +244,12 @@ data HostMode
     HMPublic
   deriving (Eq, Show)
 
+textToHostMode :: Text -> Either String HostMode
+textToHostMode = \case
+  "public" -> Right HMPublic
+  "onion" -> Right HMOnionViaSocks
+  s -> Left $ T.unpack $ "Invalid host_mode: " <> s
+
 data SocksMode
   = -- | always use SOCKS proxy when enabled
     SMAlways
@@ -257,7 +271,7 @@ instance StrEncoding SocksMode where
 -- | network configuration for the client
 data NetworkConfig = NetworkConfig
   { -- | use SOCKS5 proxy
-    socksProxy :: Maybe SocksProxy,
+    socksProxy :: Maybe SocksProxyWithAuth,
     -- | when to use SOCKS proxy
     socksMode :: SocksMode,
     -- | determines critera which host is chosen from the list
@@ -270,6 +284,8 @@ data NetworkConfig = NetworkConfig
     smpProxyMode :: SMPProxyMode,
     -- | Fallback to direct connection when destination SMP relay does not support SMP proxy protocol extensions
     smpProxyFallback :: SMPProxyFallback,
+    -- | use web port 443 for SMP protocol
+    smpWebPort :: Bool,
     -- | timeout for the initial client TCP/TLS connection (microseconds)
     tcpConnectTimeout :: Int,
     -- | timeout of protocol commands (microseconds)
@@ -288,7 +304,7 @@ data NetworkConfig = NetworkConfig
   }
   deriving (Eq, Show)
 
-data TransportSessionMode = TSMUser | TSMEntity
+data TransportSessionMode = TSMUser | TSMSession | TSMServer | TSMEntity
   deriving (Eq, Show)
 
 -- SMP proxy mode for sending messages
@@ -338,9 +354,10 @@ defaultNetworkConfig =
       socksMode = SMAlways,
       hostMode = HMOnionViaSocks,
       requiredHostMode = False,
-      sessionMode = TSMUser,
+      sessionMode = TSMSession,
       smpProxyMode = SPMNever,
       smpProxyFallback = SPFAllow,
+      smpWebPort = False,
       tcpConnectTimeout = defaultTcpConnectTimeout,
       tcpTimeout = 15_000_000,
       tcpTimeoutPerKb = 5_000,
@@ -351,15 +368,31 @@ defaultNetworkConfig =
       logTLSErrors = False
     }
 
-transportClientConfig :: NetworkConfig -> TransportHost -> TransportClientConfig
-transportClientConfig NetworkConfig {socksProxy, socksMode, tcpConnectTimeout, tcpKeepAlive, logTLSErrors} host =
-  TransportClientConfig {socksProxy = useSocksProxy socksMode, tcpConnectTimeout, tcpKeepAlive, logTLSErrors, clientCredentials = Nothing, alpn = Nothing}
+transportClientConfig :: NetworkConfig -> TransportHost -> Bool -> TransportClientConfig
+transportClientConfig NetworkConfig {socksProxy, socksMode, tcpConnectTimeout, tcpKeepAlive, logTLSErrors} host useSNI =
+  TransportClientConfig {socksProxy = useSocksProxy socksMode, tcpConnectTimeout, tcpKeepAlive, logTLSErrors, clientCredentials = Nothing, alpn = Nothing, useSNI}
   where
-    useSocksProxy SMAlways = socksProxy
+    socksProxy' = (\(SocksProxyWithAuth _ proxy) -> proxy) <$> socksProxy
+    useSocksProxy SMAlways = socksProxy'
     useSocksProxy SMOnion = case host of
-      THOnionHost _ -> socksProxy
+      THOnionHost _ -> socksProxy'
       _ -> Nothing
-{-# INLINE transportClientConfig #-}
+
+clientSocksCredentials :: ProtocolTypeI (ProtoType msg) => NetworkConfig -> UTCTime -> TransportSession msg -> Maybe SocksCredentials
+clientSocksCredentials NetworkConfig {socksProxy, sessionMode} proxySessTs (userId, srv, entityId_) = case socksProxy of
+  Just (SocksProxyWithAuth auth _) -> case auth of
+    SocksAuthUsername {username, password} -> Just $ SocksCredentials username password
+    SocksAuthNull -> Nothing
+    SocksIsolateByAuth -> Just $ SocksCredentials sessionUsername ""
+  Nothing -> Nothing
+  where
+    sessionUsername =
+      B64.encode $ C.sha256Hash $
+        bshow userId <> case sessionMode of
+          TSMUser -> ""
+          TSMSession -> ":" <> bshow proxySessTs
+          TSMServer -> ":" <> bshow proxySessTs <> "@" <> strEncode srv
+          TSMEntity -> ":" <> bshow proxySessTs <> "@" <> strEncode srv <> maybe "" ("/" <>) entityId_
 
 -- | protocol client configuration.
 data ProtocolClientConfig v = ProtocolClientConfig
@@ -373,24 +406,31 @@ data ProtocolClientConfig v = ProtocolClientConfig
     -- | client-server protocol version range
     serverVRange :: VersionRange v,
     -- | agree shared session secret (used in SMP proxy for additional encryption layer)
-    agreeSecret :: Bool
+    agreeSecret :: Bool,
+    -- | send SNI to server, False for SMP
+    useSNI :: Bool
   }
 
 -- | Default protocol client configuration.
-defaultClientConfig :: Maybe [ALPN] -> VersionRange v -> ProtocolClientConfig v
-defaultClientConfig clientALPN serverVRange =
+defaultClientConfig :: Maybe [ALPN] -> Bool -> VersionRange v -> ProtocolClientConfig v
+defaultClientConfig clientALPN useSNI serverVRange =
   ProtocolClientConfig
     { qSize = 64,
       defaultTransport = ("443", transport @TLS),
       networkConfig = defaultNetworkConfig,
       clientALPN,
       serverVRange,
-      agreeSecret = False
+      agreeSecret = False,
+      useSNI
     }
 {-# INLINE defaultClientConfig #-}
 
 defaultSMPClientConfig :: ProtocolClientConfig SMPVersion
-defaultSMPClientConfig = defaultClientConfig (Just supportedSMPHandshakes) supportedClientSMPRelayVRange
+defaultSMPClientConfig =
+  (defaultClientConfig (Just supportedSMPHandshakes) False supportedClientSMPRelayVRange)
+    { defaultTransport = (show defaultSMPPort, transport @TLS),
+      agreeSecret = True
+    }
 {-# INLINE defaultSMPClientConfig #-}
 
 data Request err msg = Request
@@ -449,15 +489,15 @@ type TransportSession msg = (UserId, ProtoServer msg, Maybe ByteString)
 --
 -- A single queue can be used for multiple 'SMPClient' instances,
 -- as 'SMPServerTransmission' includes server information.
-getProtocolClient :: forall v err msg. Protocol v err msg => TVar ChaChaDRG -> TransportSession msg -> ProtocolClientConfig v -> Maybe (TBQueue (ServerTransmissionBatch v err msg)) -> (ProtocolClient v err msg -> IO ()) -> IO (Either (ProtocolClientError err) (ProtocolClient v err msg))
-getProtocolClient g transportSession@(_, srv, _) cfg@ProtocolClientConfig {qSize, networkConfig, clientALPN, serverVRange, agreeSecret} msgQ disconnected = do
+getProtocolClient :: forall v err msg. Protocol v err msg => TVar ChaChaDRG -> TransportSession msg -> ProtocolClientConfig v -> Maybe (TBQueue (ServerTransmissionBatch v err msg)) -> UTCTime -> (ProtocolClient v err msg -> IO ()) -> IO (Either (ProtocolClientError err) (ProtocolClient v err msg))
+getProtocolClient g transportSession@(_, srv, _) cfg@ProtocolClientConfig {qSize, networkConfig, clientALPN, serverVRange, agreeSecret, useSNI} msgQ proxySessTs disconnected = do
   case chooseTransportHost networkConfig (host srv) of
     Right useHost ->
       (getCurrentTime >>= mkProtocolClient useHost >>= runClient useTransport useHost)
         `catch` \(e :: IOException) -> pure . Left $ PCEIOError e
     Left e -> pure $ Left e
   where
-    NetworkConfig {tcpConnectTimeout, tcpTimeout, smpPingInterval} = networkConfig
+    NetworkConfig {smpWebPort, tcpConnectTimeout, tcpTimeout, smpPingInterval} = networkConfig
     mkProtocolClient :: TransportHost -> UTCTime -> IO (PClient v err msg)
     mkProtocolClient transportHost ts = do
       connected <- newTVarIO False
@@ -488,10 +528,10 @@ getProtocolClient g transportSession@(_, srv, _) cfg@ProtocolClientConfig {qSize
     runClient :: (ServiceName, ATransport) -> TransportHost -> PClient v err msg -> IO (Either (ProtocolClientError err) (ProtocolClient v err msg))
     runClient (port', ATransport t) useHost c = do
       cVar <- newEmptyTMVarIO
-      let tcConfig = (transportClientConfig networkConfig useHost) {alpn = clientALPN}
-          username = proxyUsername transportSession
+      let tcConfig = (transportClientConfig networkConfig useHost useSNI) {alpn = clientALPN}
+          socksCreds = clientSocksCredentials networkConfig proxySessTs transportSession
       tId <-
-        runTransportClient tcConfig (Just username) useHost port' (Just $ keyHash srv) (client t c cVar)
+        runTransportClient tcConfig socksCreds useHost port' (Just $ keyHash srv) (client t c cVar)
           `forkFinally` \_ -> void (atomically . tryPutTMVar cVar $ Left PCENetworkError)
       c_ <- tcpConnectTimeout `timeout` atomically (takeTMVar cVar)
       case c_ of
@@ -501,7 +541,9 @@ getProtocolClient g transportSession@(_, srv, _) cfg@ProtocolClientConfig {qSize
 
     useTransport :: (ServiceName, ATransport)
     useTransport = case port srv of
-      "" -> defaultTransport cfg
+      "" -> case protocolTypeI @(ProtoType msg) of
+        SPSMP | smpWebPort -> ("443", transport @TLS)
+        _ -> defaultTransport cfg
       "80" -> ("80", transport @WS)
       p -> (p, transport @TLS)
 
@@ -597,10 +639,6 @@ getProtocolClient g transportSession@(_, srv, _) cfg@ProtocolClientConfig {qSize
 unexpectedResponse :: Show r => r -> ProtocolClientError err
 unexpectedResponse = PCEUnexpectedResponse . B.pack . take 32 . show
 
-proxyUsername :: TransportSession msg -> ByteString
-proxyUsername (userId, _, entityId_) = C.sha256Hash $ bshow userId <> maybe "" (":" <>) entityId_
-{-# INLINE proxyUsername #-}
-
 -- | Disconnects client from the server and terminates client threads.
 closeProtocolClient :: ProtocolClient v err msg -> IO ()
 closeProtocolClient = mapM_ (deRefWeak >=> mapM_ killThread) . action
@@ -679,8 +717,8 @@ createSMPQueue c (rKey, rpKey) dhKey auth subMode sndSecure =
 --
 -- https://github.com/simplex-chat/simplexmq/blob/master/protocol/simplex-messaging.md#subscribe-to-queue
 subscribeSMPQueue :: SMPClient -> RcvPrivateAuthKey -> RecipientId -> ExceptT SMPClientError IO ()
-subscribeSMPQueue c@ProtocolClient {client_ = PClient {sendPings}} rpKey rId = do
-  liftIO . atomically $ writeTVar sendPings True
+subscribeSMPQueue c rpKey rId = do
+  liftIO $ enablePings c
   sendSMPCommand c (Just rpKey) rId SUB >>= \case
     OK -> pure ()
     cmd@MSG {} -> liftIO $ writeSMPMessage c rId cmd
@@ -688,8 +726,8 @@ subscribeSMPQueue c@ProtocolClient {client_ = PClient {sendPings}} rpKey rId = d
 
 -- | Subscribe to multiple SMP queues batching commands if supported.
 subscribeSMPQueues :: SMPClient -> NonEmpty (RcvPrivateAuthKey, RecipientId) -> IO (NonEmpty (Either SMPClientError ()))
-subscribeSMPQueues c@ProtocolClient {client_ = PClient {sendPings}} qs = do
-  atomically $ writeTVar sendPings True
+subscribeSMPQueues c qs = do
+  liftIO $ enablePings c
   sendProtocolCommands c cs >>= mapM (processSUBResponse c)
   where
     cs = L.map (\(rpKey, rId) -> (Just rpKey, rId, Cmd SRecipient SUB)) qs
@@ -728,14 +766,22 @@ getSMPMessage c rpKey rId =
 --
 -- https://github.com/simplex-chat/simplexmq/blob/master/protocol/simplex-messaging.md#subscribe-to-queue-notifications
 subscribeSMPQueueNotifications :: SMPClient -> NtfPrivateAuthKey -> NotifierId -> ExceptT SMPClientError IO ()
-subscribeSMPQueueNotifications = okSMPCommand NSUB
+subscribeSMPQueueNotifications c npKey nId = do
+  liftIO $ enablePings c
+  okSMPCommand NSUB c npKey nId
 {-# INLINE subscribeSMPQueueNotifications #-}
 
 -- | Subscribe to multiple SMP queues notifications batching commands if supported.
 subscribeSMPQueuesNtfs :: SMPClient -> NonEmpty (NtfPrivateAuthKey, NotifierId) -> IO (NonEmpty (Either SMPClientError ()))
-subscribeSMPQueuesNtfs = okSMPCommands NSUB
+subscribeSMPQueuesNtfs c qs = do
+  liftIO $ enablePings c
+  okSMPCommands NSUB c qs
 {-# INLINE subscribeSMPQueuesNtfs #-}
 
+enablePings :: SMPClient -> IO ()
+enablePings ProtocolClient {client_ = PClient {sendPings}} = atomically $ writeTVar sendPings True
+{-# INLINE enablePings #-}
+
 -- | Secure the SMP queue by adding a sender public key.
 --
 -- https://github.com/simplex-chat/simplexmq/blob/master/protocol/simplex-messaging.md#secure-queue-command

src/Simplex/Messaging/Client/Agent.hs

@@ -76,7 +76,7 @@ data SMPClientAgentConfig = SMPClientAgentConfig
 defaultSMPClientAgentConfig :: SMPClientAgentConfig
 defaultSMPClientAgentConfig =
   SMPClientAgentConfig
-    { smpCfg = defaultSMPClientConfig {defaultTransport = ("5223", transport @TLS)},
+    { smpCfg = defaultSMPClientConfig,
       reconnectInterval =
         RetryInterval
           { initialInterval = second,
@@ -84,8 +84,8 @@ defaultSMPClientAgentConfig =
             maxInterval = 10 * second
           },
       persistErrorInterval = 30, -- seconds
-      msgQSize = 1024,
-      agentQSize = 1024,
+      msgQSize = 2048,
+      agentQSize = 2048,
       agentSubsBatchSize = 1360,
       ownServerDomains = []
     }
@@ -95,6 +95,7 @@ defaultSMPClientAgentConfig =
 data SMPClientAgent = SMPClientAgent
   { agentCfg :: SMPClientAgentConfig,
     active :: TVar Bool,
+    startedAt :: UTCTime,
     msgQ :: TBQueue (ServerTransmissionBatch SMPVersion ErrorType BrokerMsg),
     agentQ :: TBQueue SMPClientAgentEvent,
     randomDrg :: TVar ChaChaDRG,
@@ -111,6 +112,7 @@ type OwnServer = Bool
 newSMPClientAgent :: SMPClientAgentConfig -> TVar ChaChaDRG -> IO SMPClientAgent
 newSMPClientAgent agentCfg@SMPClientAgentConfig {msgQSize, agentQSize} randomDrg = do
   active <- newTVarIO True
+  startedAt <- getCurrentTime
   msgQ <- newTBQueueIO msgQSize
   agentQ <- newTBQueueIO agentQSize
   smpClients <- TM.emptyIO
@@ -123,6 +125,7 @@ newSMPClientAgent agentCfg@SMPClientAgentConfig {msgQSize, agentQSize} randomDrg
     SMPClientAgent
       { agentCfg,
         active,
+        startedAt,
         msgQ,
         agentQ,
         randomDrg,
@@ -194,8 +197,8 @@ isOwnServer SMPClientAgent {agentCfg} ProtocolServer {host} =
 
 -- | Run an SMP client for SMPClientVar
 connectClient :: SMPClientAgent -> SMPServer -> SMPClientVar -> IO (Either SMPClientError SMPClient)
-connectClient ca@SMPClientAgent {agentCfg, smpClients, smpSessions, msgQ, randomDrg} srv v =
-  getProtocolClient randomDrg (1, srv, Nothing) (smpCfg agentCfg) (Just msgQ) clientDisconnected
+connectClient ca@SMPClientAgent {agentCfg, smpClients, smpSessions, msgQ, randomDrg, startedAt} srv v =
+  getProtocolClient randomDrg (1, srv, Nothing) (smpCfg agentCfg) (Just msgQ) startedAt clientDisconnected
   where
     clientDisconnected :: SMPClient -> IO ()
     clientDisconnected smp = do

src/Simplex/Messaging/Crypto.hs

@@ -152,6 +152,12 @@ module Simplex.Messaging.Crypto
     unsafeSbKey,
     randomSbKey,
 
+    -- * secret_box chains
+    SbChainKey,
+    SbKeyNonce,
+    sbcInit,
+    sbcHkdf,
+
     -- * pseudo-random bytes
     randomBytes,
 
@@ -198,6 +204,7 @@ import qualified Crypto.Cipher.Types as AES
 import qualified Crypto.Cipher.XSalsa as XSalsa
 import qualified Crypto.Error as CE
 import Crypto.Hash (Digest, SHA256 (..), SHA512 (..), hash, hashDigestSize)
+import qualified Crypto.KDF.HKDF as H
 import qualified Crypto.MAC.Poly1305 as Poly1305
 import qualified Crypto.PubKey.Curve25519 as X25519
 import qualified Crypto.PubKey.Curve448 as X448
@@ -1334,6 +1341,26 @@ unsafeSbKey s = either error id $ sbKey s
 randomSbKey :: TVar ChaChaDRG -> STM SbKey
 randomSbKey gVar = SecretBoxKey <$> randomBytes 32 gVar
 
+newtype SbChainKey = SecretBoxChainKey {unSbChainKey :: ByteString}
+  deriving (Eq, Show)
+
+sbcInit :: ByteArrayAccess secret => ByteString -> secret -> (SbChainKey, SbChainKey)
+sbcInit salt secret = (SecretBoxChainKey ck1, SecretBoxChainKey ck2)
+  where
+    prk = H.extract salt secret :: H.PRK SHA512
+    out = H.expand prk ("SimpleXSbChainInit" :: ByteString) 64
+    (ck1, ck2) = B.splitAt 32 out
+
+type SbKeyNonce = (SbKey, CbNonce)
+
+sbcHkdf :: SbChainKey -> (SbKeyNonce, SbChainKey)
+sbcHkdf (SecretBoxChainKey ck) = ((SecretBoxKey sk, CryptoBoxNonce nonce), SecretBoxChainKey ck')
+  where
+    prk = H.extract B.empty ck :: H.PRK SHA512
+    out = H.expand prk ("SimpleXSbChain" :: ByteString) 88 -- = 32 (new chain key) + 32 (secret_box key) + 24 (nonce)
+    (ck', rest) = B.splitAt 32 out
+    (sk, nonce) = B.splitAt 32 rest
+
 xSalsa20 :: ByteArrayAccess key => key -> ByteString -> ByteString -> (ByteString, ByteString)
 xSalsa20 secret nonce msg = (rs, msg')
   where

src/Simplex/Messaging/Crypto/Lazy.hs

@@ -17,6 +17,8 @@ module Simplex.Messaging.Crypto.Lazy
     kcbEncryptTailTag,
     sbDecryptTailTag,
     kcbDecryptTailTag,
+    sbEncryptTailTagNoPad,
+    sbDecryptTailTagNoPad,
     fastReplicate,
     secretBox,
     secretBoxTailTag,
@@ -49,7 +51,7 @@ import Data.Composition ((.:.))
 import Data.Int (Int64)
 import Data.List.NonEmpty (NonEmpty (..))
 import Foreign (sizeOf)
-import Simplex.Messaging.Crypto (CbNonce, CryptoError (..), DhSecret (..), DhSecretX25519, SbKey, pattern CbNonce, pattern SbKey)
+import Simplex.Messaging.Crypto (CbNonce, CryptoError (..), DhSecret (..), DhSecretX25519, SbKey, SbKeyNonce, pattern CbNonce, pattern SbKey)
 import Simplex.Messaging.Crypto.SNTRUP761 (KEMHybridSecret (..))
 import Simplex.Messaging.Encoding
 
@@ -142,6 +144,10 @@ sbEncryptTailTag_ :: ByteArrayAccess key => key -> CbNonce -> LazyByteString ->
 sbEncryptTailTag_ key (CbNonce nonce) msg len paddedLen =
   LB.fromChunks <$> (secretBoxTailTag sbEncryptChunk key nonce =<< pad msg len paddedLen)
 
+sbEncryptTailTagNoPad :: SbKeyNonce -> LazyByteString -> Either CryptoError LazyByteString
+sbEncryptTailTagNoPad (SbKey key, CbNonce nonce) msg =
+  LB.fromChunks <$> secretBoxTailTag sbEncryptChunk key nonce msg
+
 -- | NaCl @secret_box@ decrypt with a symmetric 256-bit key and 192-bit nonce with appended auth tag (more efficient with large files).
 -- paddedLen should NOT include the tag length, it should be the same number that is passed to sbEncrypt / sbEncryptTailTag.
 sbDecryptTailTag :: SbKey -> CbNonce -> Int64 -> LazyByteString -> Either CryptoError (Bool, LazyByteString)
@@ -165,6 +171,15 @@ sbDecryptTailTag_ key (CbNonce nonce) paddedLen packet =
   where
     (c, tag') = LB.splitAt paddedLen packet
 
+sbDecryptTailTagNoPad :: SbKeyNonce -> Int64 -> LazyByteString -> Either CryptoError (Bool, LazyByteString)
+sbDecryptTailTagNoPad (SbKey key, CbNonce nonce) paddedLen packet =
+  result <$> secretBox sbDecryptChunk key nonce c
+  where
+    result (tag :| cs) =
+      let valid = LB.length tag' == 16 && BA.constEq (LB.toStrict tag') tag
+       in (valid, LB.fromChunks cs)
+    (c, tag') = LB.splitAt paddedLen packet
+
 secretBoxTailTag :: ByteArrayAccess key => (SbState -> ByteString -> (ByteString, SbState)) -> key -> ByteString -> LazyByteString -> Either CryptoError [ByteString]
 secretBoxTailTag sbProcess secret nonce msg = run <$> sbInit_ secret nonce
   where

src/Simplex/Messaging/Notifications/Client.hs

@@ -2,24 +2,31 @@
 {-# LANGUAGE LambdaCase #-}
 {-# LANGUAGE OverloadedStrings #-}
 {-# LANGUAGE PatternSynonyms #-}
+{-# LANGUAGE TypeApplications #-}
 
 module Simplex.Messaging.Notifications.Client where
 
 import Control.Monad.Except
 import Control.Monad.Trans.Except
+import Data.List.NonEmpty (NonEmpty (..))
+import qualified Data.List.NonEmpty as L
 import Data.Word (Word16)
 import Simplex.Messaging.Client
 import qualified Simplex.Messaging.Crypto as C
 import Simplex.Messaging.Notifications.Protocol
 import Simplex.Messaging.Notifications.Transport (NTFVersion, supportedClientNTFVRange, supportedNTFHandshakes)
 import Simplex.Messaging.Protocol (ErrorType, pattern NoEntity)
+import Simplex.Messaging.Transport (TLS, Transport (..))
 
 type NtfClient = ProtocolClient NTFVersion ErrorType NtfResponse
 
 type NtfClientError = ProtocolClientError ErrorType
 
 defaultNTFClientConfig :: ProtocolClientConfig NTFVersion
-defaultNTFClientConfig = defaultClientConfig (Just supportedNTFHandshakes) supportedClientNTFVRange
+defaultNTFClientConfig =
+  (defaultClientConfig (Just supportedNTFHandshakes) False supportedClientNTFVRange)
+    {defaultTransport = ("443", transport @TLS)}
+{-# INLINE defaultNTFClientConfig #-}
 
 ntfRegisterToken :: NtfClient -> C.APrivateAuthKey -> NewNtfEntity 'Token -> ExceptT NtfClientError IO (NtfTokenId, C.PublicKeyX25519)
 ntfRegisterToken c pKey newTkn =
@@ -51,12 +58,30 @@ ntfCreateSubscription c pKey newSub =
     NRSubId subId -> pure subId
     r -> throwE $ unexpectedResponse r
 
+ntfCreateSubscriptions :: NtfClient -> C.APrivateAuthKey -> NonEmpty (NewNtfEntity 'Subscription) -> IO (NonEmpty (Either NtfClientError NtfSubscriptionId))
+ntfCreateSubscriptions c pKey newSubs = L.map process <$> sendProtocolCommands c cs
+  where
+    cs = L.map (\newSub -> (Just pKey, NoEntity, NtfCmd SSubscription $ SNEW newSub)) newSubs
+    process (Response _ r) = case r of
+      Right (NRSubId subId) -> Right subId
+      Right r' -> Left $ unexpectedResponse r'
+      Left e -> Left e
+
 ntfCheckSubscription :: NtfClient -> C.APrivateAuthKey -> NtfSubscriptionId -> ExceptT NtfClientError IO NtfSubStatus
 ntfCheckSubscription c pKey subId =
   sendNtfCommand c (Just pKey) subId SCHK >>= \case
     NRSub stat -> pure stat
     r -> throwE $ unexpectedResponse r
 
+ntfCheckSubscriptions :: NtfClient -> C.APrivateAuthKey -> NonEmpty NtfSubscriptionId -> IO (NonEmpty (Either NtfClientError NtfSubStatus))
+ntfCheckSubscriptions c pKey subIds = L.map process <$> sendProtocolCommands c cs
+  where
+    cs = L.map (\subId -> (Just pKey, subId, NtfCmd SSubscription SCHK)) subIds
+    process (Response _ r) = case r of
+      Right (NRSub stat) -> Right stat
+      Right r' -> Left $ unexpectedResponse r'
+      Left e -> Left e
+
 ntfDeleteSubscription :: NtfClient -> C.APrivateAuthKey -> NtfSubscriptionId -> ExceptT NtfClientError IO ()
 ntfDeleteSubscription = okNtfCommand SDEL
 

src/Simplex/Messaging/Notifications/Protocol.hs

@@ -20,8 +20,11 @@ import Data.ByteString.Char8 (ByteString)
 import qualified Data.ByteString.Char8 as B
 import Data.Functor (($>))
 import Data.Kind
+import Data.List.NonEmpty (NonEmpty (..))
+import qualified Data.List.NonEmpty as L
 import Data.Maybe (isNothing)
 import Data.Text.Encoding (decodeLatin1, encodeUtf8)
+import Data.Time.Clock.System
 import Data.Type.Equality
 import Data.Word (Word16)
 import Database.SQLite.Simple.FromField (FromField (..))
@@ -426,6 +429,30 @@ instance FromJSON DeviceToken where
     t <- encodeUtf8 <$> o .: "token"
     pure $ DeviceToken pp t
 
+-- List of PNMessageData uses semicolon-separated encoding instead of strEncode,
+-- because strEncode of NonEmpty list uses comma for separator,
+-- and encoding of PNMessageData's smpQueue has comma in list of hosts
+encodePNMessages :: NonEmpty PNMessageData -> ByteString
+encodePNMessages = B.intercalate ";" . map strEncode . L.toList
+
+pnMessagesP :: A.Parser (NonEmpty PNMessageData)
+pnMessagesP = L.fromList <$> strP `A.sepBy1` A.char ';'
+
+data PNMessageData = PNMessageData
+  { smpQueue :: SMPQueueNtf,
+    ntfTs :: SystemTime,
+    nmsgNonce :: C.CbNonce,
+    encNMsgMeta :: EncNMsgMeta
+  }
+  deriving (Show)
+
+instance StrEncoding PNMessageData where
+  strEncode PNMessageData {smpQueue, ntfTs, nmsgNonce, encNMsgMeta} =
+    strEncode (smpQueue, ntfTs, nmsgNonce, encNMsgMeta)
+  strP = do
+    (smpQueue, ntfTs, nmsgNonce, encNMsgMeta) <- strP
+    pure PNMessageData {smpQueue, ntfTs, nmsgNonce, encNMsgMeta}
+
 type NtfEntityId = EntityId
 
 type NtfSubscriptionId = NtfEntityId
@@ -443,6 +470,8 @@ data NtfSubStatus
     NSInactive
   | -- | END received
     NSEnd
+  | -- | DELD received (connection was deleted)
+    NSDeleted
   | -- | SMP AUTH error
     NSAuth
   | -- | SMP error other than AUTH
@@ -456,6 +485,7 @@ ntfShouldSubscribe = \case
   NSActive -> True
   NSInactive -> True
   NSEnd -> False
+  NSDeleted -> False
   NSAuth -> False
   NSErr _ -> False
 
@@ -466,6 +496,7 @@ instance Encoding NtfSubStatus where
     NSActive -> "ACTIVE"
     NSInactive -> "INACTIVE"
     NSEnd -> "END"
+    NSDeleted -> "DELETED"
     NSAuth -> "AUTH"
     NSErr err -> "ERR " <> err
   smpP =
@@ -475,6 +506,7 @@ instance Encoding NtfSubStatus where
       "ACTIVE" -> pure NSActive
       "INACTIVE" -> pure NSInactive
       "END" -> pure NSEnd
+      "DELETED" -> pure NSDeleted
       "AUTH" -> pure NSAuth
       "ERR" -> NSErr <$> (A.space *> A.takeByteString)
       _ -> fail "bad NtfSubStatus"

src/Simplex/Messaging/Notifications/Server.hs

@@ -1,3 +1,5 @@
+{-# LANGUAGE CPP #-}
+{-# LANGUAGE BangPatterns #-}
 {-# LANGUAGE DataKinds #-}
 {-# LANGUAGE DuplicateRecordFields #-}
 {-# LANGUAGE FlexibleContexts #-}
@@ -17,12 +19,17 @@ import Control.Logger.Simple
 import Control.Monad
 import Control.Monad.Except
 import Control.Monad.Reader
+import Data.Bifunctor (first)
+import qualified Data.ByteString.Builder as BLD
 import Data.ByteString.Char8 (ByteString)
 import qualified Data.ByteString.Char8 as B
+import qualified Data.ByteString.Lazy.Char8 as LB
+import Data.Either (partitionEithers)
 import Data.Functor (($>))
 import Data.IORef
 import Data.Int (Int64)
-import Data.List (intercalate, sort)
+import qualified Data.IntSet as IS
+import Data.List (intercalate, partition, sort)
 import Data.List.NonEmpty (NonEmpty (..))
 import qualified Data.List.NonEmpty as L
 import qualified Data.Map.Strict as M
@@ -33,34 +40,42 @@ import Data.Time.Clock (UTCTime (..), diffTimeToPicoseconds, getCurrentTime)
 import Data.Time.Clock.System (getSystemTime)
 import Data.Time.Format.ISO8601 (iso8601Show)
 import GHC.IORef (atomicSwapIORef)
-import Network.Socket (ServiceName)
+import GHC.Stats (getRTSStats)
+import Network.Socket (ServiceName, Socket, socketToHandle)
 import Simplex.Messaging.Client (ProtocolClientError (..), SMPClientError, ServerTransmission (..))
 import Simplex.Messaging.Client.Agent
 import qualified Simplex.Messaging.Crypto as C
 import Simplex.Messaging.Encoding.String
 import Simplex.Messaging.Notifications.Protocol
+import Simplex.Messaging.Notifications.Server.Control
 import Simplex.Messaging.Notifications.Server.Env
-import Simplex.Messaging.Notifications.Server.Push.APNS (PNMessageData (..), PushNotification (..), PushProviderError (..))
+import Simplex.Messaging.Notifications.Server.Push.APNS (PushNotification (..), PushProviderError (..))
 import Simplex.Messaging.Notifications.Server.Stats
 import Simplex.Messaging.Notifications.Server.Store
 import Simplex.Messaging.Notifications.Server.StoreLog
 import Simplex.Messaging.Notifications.Transport
-import Simplex.Messaging.Protocol (EntityId (..), ErrorType (..), ProtocolServer (host), SMPServer, SignedTransmission, Transmission, pattern NoEntity, encodeTransmission, tGet, tPut)
+import Simplex.Messaging.Protocol (EntityId (..), ErrorType (..), ProtocolServer (host), SMPServer, SignedTransmission, Transmission, pattern NoEntity, pattern SMPServer, encodeTransmission, tGet, tPut)
 import qualified Simplex.Messaging.Protocol as SMP
 import Simplex.Messaging.Server
-import Simplex.Messaging.Server.Stats
+import Simplex.Messaging.Server.Control (CPClientRole (..))
+import Simplex.Messaging.Server.Stats (PeriodStats (..), PeriodStatCounts (..), periodStatCounts, updatePeriodStats)
+import Simplex.Messaging.TMap (TMap)
 import qualified Simplex.Messaging.TMap as TM
-import Simplex.Messaging.Transport (ATransport (..), THandle (..), THandleAuth (..), THandleParams (..), TProxy, Transport (..), TransportPeer (..))
-import Simplex.Messaging.Transport.Server (runTransportServer, tlsServerCredentials)
+import Simplex.Messaging.Transport (ATransport (..), THandle (..), THandleAuth (..), THandleParams (..), TProxy, Transport (..), TransportPeer (..), defaultSupportedParams)
+import Simplex.Messaging.Transport.Buffer (trimCR)
+import Simplex.Messaging.Transport.Server (AddHTTP, runTransportServer, runLocalTCPServer)
 import Simplex.Messaging.Util
 import System.Exit (exitFailure)
-import System.IO (BufferMode (..), hPutStrLn, hSetBuffering)
+import System.IO (BufferMode (..), hClose, hPrint, hPutStrLn, hSetBuffering, hSetNewlineMode, universalNewlineMode)
 import System.Mem.Weak (deRefWeak)
-import UnliftIO (IOMode (..), async, uninterruptibleCancel, withFile)
+import UnliftIO (IOMode (..), UnliftIO, askUnliftIO, async, uninterruptibleCancel, unliftIO, withFile)
 import UnliftIO.Concurrent (forkIO, killThread, mkWeakThreadId)
 import UnliftIO.Directory (doesFileExist, renameFile)
 import UnliftIO.Exception
 import UnliftIO.STM
+#if MIN_VERSION_base(4,18,0)
+import GHC.Conc (listThreads)
+#endif
 
 runNtfServer :: NtfServerConfig -> IO ()
 runNtfServer cfg = do
@@ -74,18 +89,19 @@ type M a = ReaderT NtfEnv IO a
 
 ntfServer :: NtfServerConfig -> TMVar Bool -> M ()
 ntfServer cfg@NtfServerConfig {transports, transportConfig = tCfg} started = do
+  restoreServerLastNtfs
   restoreServerStats
   s <- asks subscriber
   ps <- asks pushServer
   resubscribe s
-  raceAny_ (ntfSubscriber s : ntfPush ps : map runServer transports <> serverStatsThread_ cfg) `finally` stopServer
+  raceAny_ (ntfSubscriber s : ntfPush ps : map runServer transports <> serverStatsThread_ cfg <> controlPortThread_ cfg) `finally` stopServer
   where
-    runServer :: (ServiceName, ATransport) -> M ()
-    runServer (tcpPort, ATransport t) = do
-      serverParams <- asks tlsServerParams
-      serverSignKey <- either fail pure . fromTLSCredentials $ tlsServerCredentials serverParams
+    runServer :: (ServiceName, ATransport, AddHTTP) -> M ()
+    runServer (tcpPort, ATransport t, _addHTTP) = do
+      srvCreds <- asks tlsServerCreds
+      serverSignKey <- either fail pure $ fromTLSCredentials srvCreds
       env <- ask
-      liftIO $ runTransportServer started tcpPort serverParams tCfg $ \h -> runClient serverSignKey t h `runReaderT` env
+      liftIO $ runTransportServer started tcpPort defaultSupportedParams srvCreds (Just supportedNTFHandshakes) tCfg $ \h -> runClient serverSignKey t h `runReaderT` env
     fromTLSCredentials (_, pk) = C.x509ToPrivate (pk, []) >>= C.privKey
 
     runClient :: Transport c => C.APrivateSignKey -> TProxy c -> c -> M ()
@@ -99,11 +115,15 @@ ntfServer cfg@NtfServerConfig {transports, transportConfig = tCfg} started = do
 
     stopServer :: M ()
     stopServer = do
-      withNtfLog closeStoreLog
-      saveServerStats
+      logInfo "Saving server state..."
+      saveServer
       NtfSubscriber {smpSubscribers, smpAgent} <- asks subscriber
       liftIO $ readTVarIO smpSubscribers >>= mapM_ (\SMPSubscriber {subThreadId} -> readTVarIO subThreadId >>= mapM_ (deRefWeak >=> mapM_ killThread))
       liftIO $ closeSMPClientAgent smpAgent
+      logInfo "Server stopped"
+
+    saveServer :: M ()
+    saveServer = withNtfLog closeStoreLog >> saveServerLastNtfs >> saveServerStats
 
     serverStatsThread_ :: NtfServerConfig -> [M ()]
     serverStatsThread_ NtfServerConfig {logStatsInterval = Just interval, logStatsStartTime, serverStatsLogFile} =
@@ -151,6 +171,131 @@ ntfServer cfg@NtfServerConfig {transports, transportConfig = tCfg} started = do
               ]
         liftIO $ threadDelay' interval
 
+    controlPortThread_ :: NtfServerConfig -> [M ()]
+    controlPortThread_ NtfServerConfig {controlPort = Just port} = [runCPServer port]
+    controlPortThread_ _ = []
+
+    runCPServer :: ServiceName -> M ()
+    runCPServer port = do
+      cpStarted <- newEmptyTMVarIO
+      u <- askUnliftIO
+      liftIO $ do
+        labelMyThread "control port server"
+        runLocalTCPServer cpStarted port $ runCPClient u
+      where
+        runCPClient :: UnliftIO (ReaderT NtfEnv IO) -> Socket -> IO ()
+        runCPClient u sock = do
+          labelMyThread "control port client"
+          h <- socketToHandle sock ReadWriteMode
+          hSetBuffering h LineBuffering
+          hSetNewlineMode h universalNewlineMode
+          hPutStrLn h "Ntf server control port\n'help' for supported commands"
+          role <- newTVarIO CPRNone
+          cpLoop h role
+          where
+            cpLoop h role = do
+              s <- trimCR <$> B.hGetLine h
+              case strDecode s of
+                Right CPQuit -> hClose h
+                Right cmd -> logCmd s cmd >> processCP h role cmd >> cpLoop h role
+                Left err -> hPutStrLn h ("error: " <> err) >> cpLoop h role
+            logCmd s cmd = when shouldLog $ logWarn $ "ControlPort: " <> tshow s
+              where
+                shouldLog = case cmd of
+                  CPAuth _ -> False
+                  CPHelp -> False
+                  CPQuit -> False
+                  CPSkip -> False
+                  _ -> True
+            processCP h role = \case
+              CPAuth auth -> atomically $ writeTVar role $! newRole cfg
+                where
+                  newRole NtfServerConfig {controlPortUserAuth = user, controlPortAdminAuth = admin}
+                    | Just auth == admin = CPRAdmin
+                    | Just auth == user = CPRUser
+                    | otherwise = CPRNone
+              CPStats -> withUserRole $ do
+                ss <- unliftIO u $ asks serverStats
+                let getStat :: (NtfServerStats -> IORef a) -> IO a
+                    getStat var = readIORef (var ss)
+                    putStat :: Show a => String -> (NtfServerStats -> IORef a) -> IO ()
+                    putStat label var = getStat var >>= \v -> hPutStrLn h $ label <> ": " <> show v
+                putStat "fromTime" fromTime
+                putStat "tknCreated" tknCreated
+                putStat "tknVerified" tknVerified
+                putStat "tknDeleted" tknDeleted
+                putStat "subCreated" subCreated
+                putStat "subDeleted" subDeleted
+                putStat "ntfReceived" ntfReceived
+                getStat (day . activeTokens) >>= \v -> hPutStrLn h $ "daily active tokens: " <> show (IS.size v)
+                getStat (day . activeSubs) >>= \v -> hPutStrLn h $ "daily active subscriptions: " <> show (IS.size v)
+              CPStatsRTS -> tryAny getRTSStats >>= either (hPrint h) (hPrint h)
+              CPServerInfo -> readTVarIO role >>= \case
+                CPRNone -> do
+                  logError "Unauthorized control port command"
+                  hPutStrLn h "AUTH"
+                r -> do
+#if MIN_VERSION_base(4,18,0)
+                  threads <- liftIO listThreads
+                  hPutStrLn h $ "Threads: " <> show (length threads)
+#else
+                  hPutStrLn h "Threads: not available on GHC 8.10"
+#endif
+                  NtfSubscriber {smpSubscribers, smpAgent = a} <- unliftIO u $ asks subscriber
+                  putSMPWorkers a "SMP subcscribers" smpSubscribers
+                  let SMPClientAgent {smpClients, smpSessions, srvSubs, pendingSrvSubs, smpSubWorkers} = a
+                  putSMPWorkers a "SMP clients" smpClients
+                  putSMPWorkers a "SMP subscription workers" smpSubWorkers
+                  sessions <- readTVarIO smpSessions
+                  hPutStrLn h $ "SMP sessions count: " <> show (M.size sessions)
+                  putSMPSubs a "SMP subscriptions" srvSubs
+                  putSMPSubs a "Pending SMP subscriptions" pendingSrvSubs
+                  where
+                    putSMPSubs :: SMPClientAgent -> String -> TMap SMPServer (TMap SMPSub a) -> IO ()
+                    putSMPSubs a name v = do
+                      subs <- readTVarIO v
+                      (totalCnt, ownCount, otherCnt, servers, ownByServer) <- foldM countSubs (0, 0, 0, [], M.empty) $ M.assocs subs
+                      showServers a name servers
+                      hPutStrLn h $ name <> " total: " <> show totalCnt
+                      hPutStrLn h $ name <> " on own servers: " <> show ownCount
+                      when (r == CPRAdmin && not (null ownByServer)) $
+                        forM_ (M.assocs ownByServer) $ \(SMPServer (host :| _) _ _, cnt) ->
+                          hPutStrLn h $ name <> " on " <> B.unpack (strEncode host) <> ": " <> show cnt
+                      hPutStrLn h $ name <> " on other servers: " <> show otherCnt
+                      where
+                        countSubs :: (Int, Int, Int, [SMPServer], M.Map SMPServer Int) -> (SMPServer, TMap SMPSub a) -> IO (Int, Int, Int, [SMPServer], M.Map SMPServer Int)
+                        countSubs (!totalCnt, !ownCount, !otherCnt, !servers, !ownByServer) (srv, srvSubs) = do
+                          cnt <- M.size <$> readTVarIO srvSubs
+                          let totalCnt' = totalCnt + cnt
+                              ownServer = isOwnServer a srv
+                              (ownCount', otherCnt')
+                                | ownServer = (ownCount + cnt, otherCnt)
+                                | otherwise = (ownCount, otherCnt + cnt)
+                              servers' = if cnt > 0 then srv : servers else servers
+                              ownByServer'
+                                | r == CPRAdmin && ownServer && cnt > 0 = M.alter (Just . maybe cnt (+ cnt)) srv ownByServer
+                                | otherwise = ownByServer
+                          pure (totalCnt', ownCount', otherCnt', servers', ownByServer')
+                    putSMPWorkers :: SMPClientAgent -> String -> TMap SMPServer a -> IO ()
+                    putSMPWorkers a name v = readTVarIO v >>= showServers a name . M.keys
+                    showServers :: SMPClientAgent -> String -> [SMPServer] -> IO ()
+                    showServers a name srvs = do
+                      let (ownSrvs, otherSrvs) = partition (isOwnServer a) srvs
+                      hPutStrLn h $ name <> " own servers count: " <> show (length ownSrvs)
+                      when (r == CPRAdmin) $ hPutStrLn h $ name <> " own servers: " <> intercalate "," (sort $ map (\(SMPServer (host :| _) _ _) -> B.unpack $ strEncode host) ownSrvs)
+                      hPutStrLn h $ name <> " other servers count: " <> show (length otherSrvs)
+              CPHelp -> hPutStrLn h "commands: stats, stats-rts, server-info, help, quit"
+              CPQuit -> pure ()
+              CPSkip -> pure ()
+              where
+                withUserRole action =
+                  readTVarIO role >>= \case
+                    CPRAdmin -> action
+                    CPRUser -> action
+                    _ -> do
+                      logError "Unauthorized control port command"
+                      hPutStrLn h "AUTH"
+
 resubscribe :: NtfSubscriber -> M ()
 resubscribe NtfSubscriber {newSubQ} = do
   logInfo "Preparing SMP resubscriptions..."
@@ -219,13 +364,16 @@ ntfSubscriber NtfSubscriber {smpSubscribers, newSubQ, smpAgent = ca@SMPClientAge
               NtfPushServer {pushQ} <- asks pushServer
               stats <- asks serverStats
               liftIO $ updatePeriodStats (activeSubs stats) ntfId
-              atomically $
-                findNtfSubscriptionToken st smpQueue
-                  >>= mapM_ (\tkn -> writeTBQueue pushQ (tkn, PNMessage (PNMessageData {smpQueue, ntfTs, nmsgNonce, encNMsgMeta} :| [])))
+              tkn_ <- atomically (findNtfSubscriptionToken st smpQueue)
+              forM_ tkn_ $ \tkn -> do
+                let newNtf = PNMessageData {smpQueue, ntfTs, nmsgNonce, encNMsgMeta}
+                lastNtfs <- liftIO $ addTokenLastNtf st (ntfTknId tkn) newNtf
+                atomically (writeTBQueue pushQ (tkn, PNMessage lastNtfs))
               incNtfStat ntfReceived
             Right SMP.END ->
               whenM (atomically $ activeClientSession' ca sessionId srv) $
                 updateSubStatus smpQueue NSEnd
+            Right SMP.DELD -> updateSubStatus smpQueue NSDeleted
             Right (SMP.ERR e) -> logError $ "SMP server error: " <> tshow e
             Right _ -> logError "SMP server unexpected response"
             Left e -> logError $ "SMP client error: " <> tshow e
@@ -353,28 +501,29 @@ clientDisconnected NtfServerClient {connected} = atomically $ writeTVar connecte
 
 receive :: Transport c => THandleNTF c 'TServer -> NtfServerClient -> M ()
 receive th@THandle {params = THandleParams {thAuth}} NtfServerClient {rcvQ, sndQ, rcvActiveAt} = forever $ do
-  ts <- liftIO $ tGet th
-  forM_ ts $ \t@(_, _, (corrId, entId, cmdOrError)) -> do
-    atomically . writeTVar rcvActiveAt =<< liftIO getSystemTime
-    logDebug "received transmission"
-    case cmdOrError of
-      Left e -> write sndQ (corrId, entId, NRErr e)
-      Right cmd ->
-        verifyNtfTransmission ((,C.cbNonce (SMP.bs corrId)) <$> thAuth) t cmd >>= \case
-          VRVerified req -> write rcvQ req
-          VRFailed -> write sndQ (corrId, entId, NRErr AUTH)
+  ts <- L.toList <$> liftIO (tGet th)
+  atomically . (writeTVar rcvActiveAt $!) =<< liftIO getSystemTime
+  (errs, cmds) <- partitionEithers <$> mapM cmdAction ts
+  write sndQ errs
+  write rcvQ cmds
   where
-    write q t = atomically $ writeTBQueue q t
+    cmdAction t@(_, _, (corrId, entId, cmdOrError)) =
+      case cmdOrError of
+        Left e -> pure $ Left (corrId, entId, NRErr e)
+        Right cmd ->
+          verified <$> verifyNtfTransmission ((,C.cbNonce (SMP.bs corrId)) <$> thAuth) t cmd
+          where
+            verified = \case
+              VRVerified req -> Right req
+              VRFailed -> Left (corrId, entId, NRErr AUTH)
+    write q = mapM_ (atomically . writeTBQueue q) . L.nonEmpty
 
 send :: Transport c => THandleNTF c 'TServer -> NtfServerClient -> IO ()
 send h@THandle {params} NtfServerClient {sndQ, sndActiveAt} = forever $ do
-  t <- atomically $ readTBQueue sndQ
-  void . liftIO $ tPut h [Right (Nothing, encodeTransmission params t)]
+  ts <- atomically $ readTBQueue sndQ
+  void . liftIO $ tPut h $ L.map (\t -> Right (Nothing, encodeTransmission params t)) ts
   atomically . (writeTVar sndActiveAt $!) =<< liftIO getSystemTime
 
--- instance Show a => Show (TVar a) where
---   show x = unsafePerformIO $ show <$> readTVarIO x
-
 data VerificationResult = VRVerified NtfRequest | VRFailed
 
 verifyNtfTransmission :: Maybe (THandleAuth 'TServer, C.CbNonce) -> SignedTransmission ErrorType NtfCmd -> NtfCmd -> M VerificationResult
@@ -432,7 +581,7 @@ client :: NtfServerClient -> NtfSubscriber -> NtfPushServer -> M ()
 client NtfServerClient {rcvQ, sndQ} NtfSubscriber {newSubQ, smpAgent = ca} NtfPushServer {pushQ, intervalNotifiers} =
   forever $
     atomically (readTBQueue rcvQ)
-      >>= processCommand
+      >>= mapM processCommand
       >>= atomically . writeTBQueue sndQ
   where
     processCommand :: NtfRequest -> M (Transmission NtfResponse)
@@ -587,6 +736,44 @@ incNtfStat statSel = do
   stats <- asks serverStats
   liftIO $ atomicModifyIORef'_ (statSel stats) (+ 1)
 
+saveServerLastNtfs :: M ()
+saveServerLastNtfs = asks (storeLastNtfsFile . config) >>= mapM_ saveLastNtfs
+  where
+    saveLastNtfs f = do
+      logInfo $ "saving last notifications to file " <> T.pack f
+      NtfStore {tokenLastNtfs} <- asks store
+      liftIO . withFile f WriteMode $ \h ->
+        readTVarIO tokenLastNtfs >>= mapM_ (saveTokenLastNtfs h) . M.assocs
+      logInfo "notifications saved"
+      where
+        -- reverse on save, to save notifications in order, will become reversed again when restoring.
+        saveTokenLastNtfs h (tknId, v) = BLD.hPutBuilder h . encodeLastNtfs tknId . L.reverse =<< readTVarIO v
+        encodeLastNtfs tknId = mconcat . L.toList . L.map (\ntf -> BLD.byteString (strEncode $ TNMRv1 tknId ntf) <> BLD.char8 '\n')
+
+restoreServerLastNtfs :: M ()
+restoreServerLastNtfs =
+  asks (storeLastNtfsFile . config) >>= mapM_ restoreLastNtfs
+  where
+    restoreLastNtfs f =
+      whenM (doesFileExist f) $ do
+        logInfo $ "restoring last notifications from file " <> T.pack f
+        st <- asks store
+        runExceptT (liftIO (LB.readFile f) >>= mapM (restoreNtf st) . LB.lines) >>= \case
+          Left e -> do
+            logError . T.pack $ "error restoring last notifications: " <> e
+            liftIO exitFailure
+          Right _ -> do
+            renameFile f $ f <> ".bak"
+            logInfo "last notifications restored"
+      where
+        restoreNtf st s' = do
+          TNMRv1 tknId ntf <- liftEither . first (ntfErr "parsing") $ strDecode s
+          liftIO $ storeTokenLastNtf st tknId ntf
+          where
+            s = LB.toStrict s'
+            ntfErr :: Show e => String -> e -> String
+            ntfErr op e = op <> " error (" <> show e <> "): " <> B.unpack (B.take 100 s)
+
 saveServerStats :: M ()
 saveServerStats =
   asks (serverStatsBackupFile . config)

src/Simplex/Messaging/Notifications/Server/Control.hs

@@ -0,0 +1,37 @@
+{-# LANGUAGE LambdaCase #-}
+{-# LANGUAGE OverloadedStrings #-}
+
+module Simplex.Messaging.Notifications.Server.Control where
+
+import qualified Data.Attoparsec.ByteString.Char8 as A
+import Simplex.Messaging.Encoding.String
+import Simplex.Messaging.Protocol (BasicAuth)
+
+data ControlProtocol
+  = CPAuth BasicAuth
+  | CPStats
+  | CPStatsRTS
+  | CPServerInfo
+  | CPHelp
+  | CPQuit
+  | CPSkip
+
+instance StrEncoding ControlProtocol where
+  strEncode = \case
+    CPAuth tok -> "auth " <> strEncode tok
+    CPStats -> "stats"
+    CPStatsRTS -> "stats-rts"
+    CPServerInfo -> "server-info"
+    CPHelp -> "help"
+    CPQuit -> "quit"
+    CPSkip -> ""
+  strP =
+    A.takeTill (== ' ') >>= \case
+      "auth" -> CPAuth <$> _strP
+      "stats" -> pure CPStats
+      "stats-rts" -> pure CPStatsRTS
+      "server-info" -> pure CPServerInfo
+      "help" -> pure CPHelp
+      "quit" -> pure CPQuit
+      "" -> pure CPSkip
+      _ -> fail "bad ControlProtocol command"

src/Simplex/Messaging/Notifications/Server/Env.hs

@@ -28,18 +28,21 @@ import Simplex.Messaging.Notifications.Server.Stats
 import Simplex.Messaging.Notifications.Server.Store
 import Simplex.Messaging.Notifications.Server.StoreLog
 import Simplex.Messaging.Notifications.Transport (NTFVersion, VersionRangeNTF)
-import Simplex.Messaging.Protocol (CorrId, SMPServer, Transmission)
+import Simplex.Messaging.Protocol (BasicAuth, CorrId, SMPServer, Transmission)
 import Simplex.Messaging.Server.Expiration
 import Simplex.Messaging.TMap (TMap)
 import qualified Simplex.Messaging.TMap as TM
 import Simplex.Messaging.Transport (ATransport, THandleParams, TransportPeer (..))
-import Simplex.Messaging.Transport.Server (TransportServerConfig, alpn, loadFingerprint, loadTLSServerParams)
+import Simplex.Messaging.Transport.Server (AddHTTP, ServerCredentials, TransportServerConfig, loadFingerprint, loadServerCredential)
 import System.IO (IOMode (..))
 import System.Mem.Weak (Weak)
 import UnliftIO.STM
 
 data NtfServerConfig = NtfServerConfig
-  { transports :: [(ServiceName, ATransport)],
+  { transports :: [(ServiceName, ATransport, AddHTTP)],
+    controlPort :: Maybe ServiceName,
+    controlPortUserAuth :: Maybe BasicAuth,
+    controlPortAdminAuth :: Maybe BasicAuth,
     subIdBytes :: Int,
     regCodeBytes :: Int,
     clientQSize :: Natural,
@@ -50,10 +53,8 @@ data NtfServerConfig = NtfServerConfig
     subsBatchSize :: Int,
     inactiveClientExpiration :: Maybe ExpirationConfig,
     storeLogFile :: Maybe FilePath,
-    -- CA certificate private key is not needed for initialization
-    caCertificateFile :: FilePath,
-    privateKeyFile :: FilePath,
-    certificateFile :: FilePath,
+    storeLastNtfsFile :: Maybe FilePath,
+    ntfCredentials :: ServerCredentials,
     -- stats config - see SMP server config
     logStatsInterval :: Maybe Int64,
     logStatsStartTime :: Int64,
@@ -77,13 +78,13 @@ data NtfEnv = NtfEnv
     store :: NtfStore,
     storeLog :: Maybe (StoreLog 'WriteMode),
     random :: TVar ChaChaDRG,
-    tlsServerParams :: T.ServerParams,
+    tlsServerCreds :: T.Credential,
     serverIdentity :: C.KeyHash,
     serverStats :: NtfServerStats
   }
 
 newNtfServerEnv :: NtfServerConfig -> IO NtfEnv
-newNtfServerEnv config@NtfServerConfig {subQSize, pushQSize, smpAgentCfg, apnsConfig, storeLogFile, caCertificateFile, certificateFile, privateKeyFile, transportConfig} = do
+newNtfServerEnv config@NtfServerConfig {subQSize, pushQSize, smpAgentCfg, apnsConfig, storeLogFile, ntfCredentials} = do
   random <- C.newRandom
   store <- newNtfStore
   logInfo "restoring subscriptions..."
@@ -91,10 +92,10 @@ newNtfServerEnv config@NtfServerConfig {subQSize, pushQSize, smpAgentCfg, apnsCo
   logInfo "restored subscriptions"
   subscriber <- newNtfSubscriber subQSize smpAgentCfg random
   pushServer <- newNtfPushServer pushQSize apnsConfig
-  tlsServerParams <- loadTLSServerParams caCertificateFile certificateFile privateKeyFile (alpn transportConfig)
-  Fingerprint fp <- loadFingerprint caCertificateFile
+  tlsServerCreds <- loadServerCredential ntfCredentials
+  Fingerprint fp <- loadFingerprint ntfCredentials
   serverStats <- newNtfServerStats =<< getCurrentTime
-  pure NtfEnv {config, subscriber, pushServer, store, storeLog, random, tlsServerParams, serverIdentity = C.KeyHash fp, serverStats}
+  pure NtfEnv {config, subscriber, pushServer, store, storeLog, random, tlsServerCreds, serverIdentity = C.KeyHash fp, serverStats}
 
 data NtfSubscriber = NtfSubscriber
   { smpSubscribers :: TMap SMPServer SMPSubscriber,
@@ -158,8 +159,8 @@ data NtfRequest
   | NtfReqPing CorrId NtfEntityId
 
 data NtfServerClient = NtfServerClient
-  { rcvQ :: TBQueue NtfRequest,
-    sndQ :: TBQueue (Transmission NtfResponse),
+  { rcvQ :: TBQueue (NonEmpty NtfRequest),
+    sndQ :: TBQueue (NonEmpty (Transmission NtfResponse)),
     ntfThParams :: THandleParams NTFVersion 'TServer,
     connected :: TVar Bool,
     rcvActiveAt :: TVar SystemTime,

src/Simplex/Messaging/Notifications/Server/Main.hs

@@ -5,6 +5,7 @@
 {-# LANGUAGE OverloadedLists #-}
 {-# LANGUAGE OverloadedStrings #-}
 {-# LANGUAGE PatternSynonyms #-}
+{-# OPTIONS_GHC -fno-warn-ambiguous-fields #-}
 
 module Simplex.Messaging.Notifications.Server.Main where
 
@@ -13,22 +14,23 @@ import Data.Functor (($>))
 import Data.Ini (lookupValue, readIniFile)
 import Data.Maybe (fromMaybe)
 import qualified Data.Text as T
+import Data.Text.Encoding (encodeUtf8)
 import qualified Data.Text.IO as T
 import Network.Socket (HostName)
 import Options.Applicative
-import Simplex.Messaging.Client (NetworkConfig (..), ProtocolClientConfig (..), SocksMode (..), defaultNetworkConfig)
+import Simplex.Messaging.Client (HostMode (..), NetworkConfig (..), ProtocolClientConfig (..), SocksMode (..), defaultNetworkConfig, textToHostMode)
 import Simplex.Messaging.Client.Agent (SMPClientAgentConfig (..), defaultSMPClientAgentConfig)
 import qualified Simplex.Messaging.Crypto as C
 import Simplex.Messaging.Notifications.Server (runNtfServer)
 import Simplex.Messaging.Notifications.Server.Env (NtfServerConfig (..), defaultInactiveClientExpiration)
 import Simplex.Messaging.Notifications.Server.Push.APNS (defaultAPNSPushClientConfig)
-import Simplex.Messaging.Notifications.Transport (supportedNTFHandshakes, supportedServerNTFVRange)
+import Simplex.Messaging.Notifications.Transport (supportedServerNTFVRange)
 import Simplex.Messaging.Protocol (ProtoServerWithAuth (..), pattern NtfServer)
 import Simplex.Messaging.Server.CLI
 import Simplex.Messaging.Server.Expiration
 import Simplex.Messaging.Transport (simplexMQVersion)
 import Simplex.Messaging.Transport.Client (TransportHost (..))
-import Simplex.Messaging.Transport.Server (TransportServerConfig (..), defaultTransportServerConfig)
+import Simplex.Messaging.Transport.Server (ServerCredentials (..), TransportServerConfig (..), defaultTransportServerConfig)
 import Simplex.Messaging.Util (tshow)
 import System.Directory (createDirectoryIfMissing, doesFileExist)
 import System.FilePath (combine)
@@ -51,7 +53,9 @@ ntfServerCLI cfgPath logPath =
         True -> readIniFile iniFile >>= either exitError runServer
         _ -> exitError $ "Error: server is not initialized (" <> iniFile <> " does not exist).\nRun `" <> executableName <> " init`."
     Delete -> do
-      confirmOrExit "WARNING: deleting the server will make all queues inaccessible, because the server identity (certificate fingerprint) will change.\nTHIS CANNOT BE UNDONE!"
+      confirmOrExit
+        "WARNING: deleting the server will make all queues inaccessible, because the server identity (certificate fingerprint) will change.\nTHIS CANNOT BE UNDONE!"
+        "Server NOT deleted"
       deleteDirIfExists cfgPath
       deleteDirIfExists logPath
       putStrLn "Deleted configuration and log files"
@@ -83,21 +87,38 @@ ntfServerCLI cfgPath logPath =
           \# and restoring it when the server is started.\n\
           \# Log is compacted on start (deleted objects are removed).\n"
             <> ("enable: " <> onOff enableStoreLog <> "\n\n")
+            <> "# Last notifications are optionally saved and restored when the server restarts,\n\
+               \# they are preserved in the .bak file until the next restart.\n"
+            <> ("restore_last_notifications: " <> onOff enableStoreLog <> "\n\n")
             <> "log_stats: off\n\n\
+               \[AUTH]\n\
+               \# control_port_admin_password:\n\
+               \# control_port_user_password:\n\
+               \\n\
                \[TRANSPORT]\n\
-               \# host is only used to print server address on start\n"
+               \# Host is only used to print server address on start.\n\
+               \# You can specify multiple server ports.\n"
             <> ("host: " <> T.pack host <> "\n")
             <> ("port: " <> T.pack defaultServerPort <> "\n")
-            <> "log_tls_errors: off\n"
-            <> "websockets: off\n\n\
+            <> "log_tls_errors: off\n\n\
+               \# Use `websockets: 443` to run websockets server in addition to plain TLS.\n\
+               \websockets: off\n\n\
+               \# control_port: 5227\n\
+               \\n\
                \[SUBSCRIBER]\n\
                \# Network configuration for notification server client.\n\
+               \# `host_mode` can be 'public' (default) or 'onion'.\n\
+               \# It defines prefferred hostname for destination servers with multiple hostnames.\n\
+               \# host_mode: public\n\
+               \# required_host_mode: off\n\n\
                \# SOCKS proxy port for subscribing to SMP servers.\n\
                \# You may need a separate instance of SOCKS proxy for incoming single-hop requests.\n\
                \# socks_proxy: localhost:9050\n\n\
                \# `socks_mode` can be 'onion' for SOCKS proxy to be used for .onion destination hosts only (default)\n\
                \# or 'always' to be used for all destination hosts (can be used if it is an .onion server).\n\
                \# socks_mode: onion\n\n\
+               \# The domain suffixes of the relays you operate (space-separated) to count as separate proxy statistics.\n\
+               \# own_server_domains: \n\n\
                \[INACTIVE_CLIENTS]\n\
                \# TTL and interval to check inactive clients\n\
                \disconnect: off\n"
@@ -118,9 +139,17 @@ ntfServerCLI cfgPath logPath =
         enableStoreLog = settingIsOn "STORE_LOG" "enable" ini
         logStats = settingIsOn "STORE_LOG" "log_stats" ini
         c = combine cfgPath . ($ defaultX509Config)
+        restoreLastNtfsFile path = case iniOnOff "STORE_LOG" "restore_last_notifications" ini of
+          Just True -> Just path
+          Just False -> Nothing
+          -- if the setting is not set, it is enabled when store log is enabled
+          _ -> enableStoreLog $> path
         serverConfig =
           NtfServerConfig
             { transports = iniTransports ini,
+              controlPort = either (const Nothing) (Just . T.unpack) $ lookupValue "TRANSPORT" "control_port" ini,
+              controlPortAdminAuth = either error id <$!> strDecodeIni "AUTH" "control_port_admin_password" ini,
+              controlPortUserAuth = either error id <$!> strDecodeIni "AUTH" "control_port_user_password" ini,
               subIdBytes = 24,
               regCodeBytes = 32,
               clientQSize = 64,
@@ -134,9 +163,12 @@ ntfServerCLI cfgPath logPath =
                             defaultNetworkConfig
                               { socksProxy = either error id <$!> strDecodeIni "SUBSCRIBER" "socks_proxy" ini,
                                 socksMode = maybe SMOnion (either error id) $! strDecodeIni "SUBSCRIBER" "socks_mode" ini,
-                                smpPingInterval = 60_000_000 -- 1 minutes
+                                hostMode = either (const HMPublic) (either error id . textToHostMode) $ lookupValue "SUBSCRIBER" "host_mode" ini,
+                                requiredHostMode = fromMaybe False $ iniOnOff "SUBSCRIBER" "required_host_mode" ini,
+                                smpPingInterval = 60_000_000 -- 1 minute
                               }
                         },
+                    ownServerDomains = either (const []) (map encodeUtf8 . T.words) $ lookupValue "SUBSCRIBER" "own_server_domains" ini,
                     persistErrorInterval = 0 -- seconds
                   },
               apnsConfig = defaultAPNSPushClientConfig,
@@ -148,9 +180,13 @@ ntfServerCLI cfgPath logPath =
                       checkInterval = readStrictIni "INACTIVE_CLIENTS" "check_interval" ini
                     },
               storeLogFile = enableStoreLog $> storeLogFilePath,
-              caCertificateFile = c caCrtFile,
-              privateKeyFile = c serverKeyFile,
-              certificateFile = c serverCrtFile,
+              storeLastNtfsFile = restoreLastNtfsFile $ combine logPath "ntf-server-last-notifications.log",
+              ntfCredentials =
+                ServerCredentials
+                  { caCertificateFile = Just $ c caCrtFile,
+                    privateKeyFile = c serverKeyFile,
+                    certificateFile = c serverCrtFile
+                  },
               logStatsInterval = logStats $> 86400, -- seconds
               logStatsStartTime = 0, -- seconds from 00:00 UTC
               serverStatsLogFile = combine logPath "ntf-server-stats.daily.log",
@@ -158,8 +194,7 @@ ntfServerCLI cfgPath logPath =
               ntfServerVRange = supportedServerNTFVRange,
               transportConfig =
                 defaultTransportServerConfig
-                  { logTLSErrors = fromMaybe False $ iniOnOff "TRANSPORT" "log_tls_errors" ini,
-                    alpn = Just supportedNTFHandshakes
+                  { logTLSErrors = fromMaybe False $ iniOnOff "TRANSPORT" "log_tls_errors" ini
                   }
             }
 

src/Simplex/Messaging/Notifications/Server/Push/APNS.hs

@@ -28,16 +28,13 @@ import Data.Aeson (ToJSON, (.=))
 import qualified Data.Aeson as J
 import qualified Data.Aeson.Encoding as JE
 import qualified Data.Aeson.TH as JQ
-import qualified Data.Attoparsec.ByteString.Char8 as A
 import Data.Bifunctor (first)
 import qualified Data.ByteString.Base64.URL as U
 import Data.ByteString.Builder (lazyByteString)
 import Data.ByteString.Char8 (ByteString)
-import qualified Data.ByteString.Char8 as B
 import qualified Data.ByteString.Lazy.Char8 as LB
 import Data.Int (Int64)
 import Data.List.NonEmpty (NonEmpty (..))
-import qualified Data.List.NonEmpty as L
 import Data.Map.Strict (Map)
 import Data.Maybe (isNothing)
 import Data.Text (Text)
@@ -51,12 +48,10 @@ import Network.HTTP2.Client (Request)
 import qualified Network.HTTP2.Client as H
 import Network.Socket (HostName, ServiceName)
 import qualified Simplex.Messaging.Crypto as C
-import Simplex.Messaging.Encoding.String
 import Simplex.Messaging.Notifications.Protocol
 import Simplex.Messaging.Notifications.Server.Push.APNS.Internal
 import Simplex.Messaging.Notifications.Server.Store (NtfTknData (..))
 import Simplex.Messaging.Parsers (defaultJSON)
-import Simplex.Messaging.Protocol (EncNMsgMeta)
 import Simplex.Messaging.Transport.HTTP2 (HTTP2Body (..))
 import Simplex.Messaging.Transport.HTTP2.Client
 import Simplex.Messaging.Util (safeDecodeUtf8, tshow)
@@ -112,30 +107,6 @@ data PushNotification
     PNCheckMessages
   deriving (Show)
 
--- List of PNMessageData uses semicolon-separated encoding instead of strEncode,
--- because strEncode of NonEmpty list uses comma for separator,
--- and encoding of PNMessageData's smpQueue has comma in list of hosts
-encodePNMessages :: NonEmpty PNMessageData -> ByteString
-encodePNMessages = B.intercalate ";" . map strEncode . L.toList
-
-pnMessagesP :: A.Parser (NonEmpty PNMessageData)
-pnMessagesP = L.fromList <$> strP `A.sepBy1` A.char ';'
-
-data PNMessageData = PNMessageData
-  { smpQueue :: SMPQueueNtf,
-    ntfTs :: SystemTime,
-    nmsgNonce :: C.CbNonce,
-    encNMsgMeta :: EncNMsgMeta
-  }
-  deriving (Show)
-
-instance StrEncoding PNMessageData where
-  strEncode PNMessageData {smpQueue, ntfTs, nmsgNonce, encNMsgMeta} =
-    strEncode (smpQueue, ntfTs, nmsgNonce, encNMsgMeta)
-  strP = do
-    (smpQueue, ntfTs, nmsgNonce, encNMsgMeta) <- strP
-    pure PNMessageData {smpQueue, ntfTs, nmsgNonce, encNMsgMeta}
-
 data APNSNotification = APNSNotification {aps :: APNSNotificationBody, notificationData :: Maybe J.Value}
   deriving (Show)
 
@@ -221,7 +192,7 @@ defaultAPNSPushClientConfig =
       authKeyFileEnv = "APNS_KEY_FILE", -- the environment variables APNS_KEY_FILE and APNS_KEY_ID must be set, or the server would fail to start
       authKeyAlg = "ES256",
       authKeyIdEnv = "APNS_KEY_ID",
-      paddedNtfLength = 512,
+      paddedNtfLength = 3072,
       appName = "chat.simplex.app",
       appTeamId = "5NN7GUYB6T",
       apnsPort = "443",

src/Simplex/Messaging/Notifications/Server/Store.hs

@@ -1,9 +1,12 @@
+{-# LANGUAGE BangPatterns #-}
 {-# LANGUAGE DataKinds #-}
 {-# LANGUAGE DuplicateRecordFields #-}
 {-# LANGUAGE GADTs #-}
 {-# LANGUAGE KindSignatures #-}
 {-# LANGUAGE LambdaCase #-}
 {-# LANGUAGE NamedFieldPuns #-}
+{-# LANGUAGE OverloadedLists #-}
+{-# LANGUAGE OverloadedStrings #-}
 {-# LANGUAGE ScopedTypeVariables #-}
 
 module Simplex.Messaging.Notifications.Server.Store where
@@ -12,12 +15,14 @@ import Control.Concurrent.STM
 import Control.Monad
 import Data.ByteString.Char8 (ByteString)
 import Data.Functor (($>))
+import Data.List.NonEmpty (NonEmpty (..), (<|))
 import qualified Data.Map.Strict as M
 import Data.Maybe (catMaybes)
 import Data.Set (Set)
 import qualified Data.Set as S
 import Data.Word (Word16)
 import qualified Simplex.Messaging.Crypto as C
+import Simplex.Messaging.Encoding.String
 import Simplex.Messaging.Notifications.Protocol
 import Simplex.Messaging.Protocol (NtfPrivateAuthKey, NtfPublicAuthKey, SMPServer)
 import Simplex.Messaging.TMap (TMap)
@@ -30,7 +35,8 @@ data NtfStore = NtfStore
     tokenRegistrations :: TMap DeviceToken (TMap ByteString NtfTokenId),
     subscriptions :: TMap NtfSubscriptionId NtfSubData,
     tokenSubscriptions :: TMap NtfTokenId (TVar (Set NtfSubscriptionId)),
-    subscriptionLookup :: TMap SMPQueueNtf NtfSubscriptionId
+    subscriptionLookup :: TMap SMPQueueNtf NtfSubscriptionId,
+    tokenLastNtfs :: TMap NtfTokenId (TVar (NonEmpty PNMessageData))
   }
 
 newNtfStore :: IO NtfStore
@@ -40,7 +46,8 @@ newNtfStore = do
   subscriptions <- TM.emptyIO
   tokenSubscriptions <- TM.emptyIO
   subscriptionLookup <- TM.emptyIO
-  pure NtfStore {tokens, tokenRegistrations, subscriptions, tokenSubscriptions, subscriptionLookup}
+  tokenLastNtfs <- TM.emptyIO
+  pure NtfStore {tokens, tokenRegistrations, subscriptions, tokenSubscriptions, subscriptionLookup, tokenLastNtfs}
 
 data NtfTknData = NtfTknData
   { ntfTknId :: NtfTokenId,
@@ -111,6 +118,7 @@ removeInactiveTokenRegistrations st NtfTknData {ntfTknId = tId, token} =
       forM_ tIds $ \(regKey, tId') -> do
         TM.delete regKey tknRegs
         TM.delete tId' $ tokens st
+        TM.delete tId' $ tokenLastNtfs st
         void $ deleteTokenSubs st tId'
       pure $ map snd tIds
 
@@ -130,6 +138,7 @@ deleteNtfToken st tknId = do
       TM.lookup token regs $>>= \tIds ->
         TM.delete (regKey tknVerifyKey) tIds
           >> whenM (TM.null tIds) (TM.delete token regs) $> Just ()
+  TM.delete tknId $ tokenLastNtfs st
   deleteTokenSubs st tknId
   where
     regs = tokenRegistrations st
@@ -196,3 +205,42 @@ deleteNtfSubscription st subId = do
           ts_ <- TM.lookup tokenId (tokenSubscriptions st)
           forM_ ts_ $ \ts -> modifyTVar' ts $ S.delete subId
       )
+
+addTokenLastNtf :: NtfStore -> NtfTokenId -> PNMessageData -> IO (NonEmpty PNMessageData)
+addTokenLastNtf st tknId newNtf =
+  TM.lookupIO tknId (tokenLastNtfs st) >>= maybe (atomically maybeNewTokenLastNtfs) (atomically . addNtf)
+  where
+    maybeNewTokenLastNtfs =
+      TM.lookup tknId (tokenLastNtfs st) >>= maybe newTokenLastNtfs addNtf
+    newTokenLastNtfs = do
+      v <- newTVar [newNtf]
+      TM.insert tknId v $ tokenLastNtfs st
+      pure [newNtf]
+    addNtf v =
+      stateTVar v $ \ntfs -> let !ntfs' = rebuildList ntfs in (ntfs', ntfs')
+      where
+        rebuildList :: NonEmpty PNMessageData -> NonEmpty PNMessageData
+        rebuildList = foldr keepPrevNtf [newNtf]
+          where
+            PNMessageData {smpQueue = newNtfQ} = newNtf
+            keepPrevNtf ntf@PNMessageData {smpQueue} ntfs
+              | smpQueue /= newNtfQ && length ntfs < maxNtfs = ntf <| ntfs
+              | otherwise = ntfs
+        maxNtfs = 6
+
+-- This function is expected to be called after store log is read,
+-- as it checks for token existence when adding last notification.
+storeTokenLastNtf :: NtfStore -> NtfTokenId -> PNMessageData -> IO ()
+storeTokenLastNtf (NtfStore {tokens, tokenLastNtfs}) tknId ntf = do
+  TM.lookupIO tknId tokenLastNtfs >>= atomically . maybe newTokenLastNtfs (`modifyTVar'` (ntf <|))
+  where
+    newTokenLastNtfs = TM.lookup tknId tokenLastNtfs >>= maybe insertForExistingToken (`modifyTVar'` (ntf <|))
+    insertForExistingToken =
+      whenM (TM.member tknId tokens) $
+        TM.insertM tknId (newTVar [ntf]) tokenLastNtfs
+
+data TokenNtfMessageRecord = TNMRv1 NtfTokenId PNMessageData
+
+instance StrEncoding TokenNtfMessageRecord where
+  strEncode (TNMRv1 tknId ntf) = strEncode (Str "v1", tknId, ntf)
+  strP = "v1 " *> (TNMRv1 <$> strP_ <*> strP)

src/Simplex/Messaging/Notifications/Server/StoreLog.hs

@@ -37,8 +37,7 @@ import Simplex.Messaging.Notifications.Protocol
 import Simplex.Messaging.Notifications.Server.Store
 import Simplex.Messaging.Protocol (NtfPrivateAuthKey)
 import Simplex.Messaging.Server.StoreLog
-import Simplex.Messaging.Util (safeDecodeUtf8, whenM)
-import System.Directory (doesFileExist, renameFile)
+import Simplex.Messaging.Util (safeDecodeUtf8)
 import System.IO
 
 data NtfStoreLogRecord
@@ -50,6 +49,7 @@ data NtfStoreLogRecord
   | CreateSubscription NtfSubRec
   | SubscriptionStatus NtfSubscriptionId NtfSubStatus
   | DeleteSubscription NtfSubscriptionId
+  deriving (Show)
 
 data NtfTknRec = NtfTknRec
   { ntfTknId :: NtfTokenId,
@@ -61,6 +61,7 @@ data NtfTknRec = NtfTknRec
     tknRegCode :: NtfRegCode,
     tknCronInterval :: Word16
   }
+  deriving (Show)
 
 mkTknData :: NtfTknRec -> STM NtfTknData
 mkTknData NtfTknRec {ntfTknId, token, tknStatus = status, tknVerifyKey, tknDhKeys, tknDhSecret, tknRegCode, tknCronInterval = cronInt} = do
@@ -81,6 +82,7 @@ data NtfSubRec = NtfSubRec
     tokenId :: NtfTokenId,
     subStatus :: NtfSubStatus
   }
+  deriving (Show)
 
 mkSubData :: NtfSubRec -> STM NtfSubData
 mkSubData NtfSubRec {ntfSubId, smpQueue, notifierKey, tokenId, subStatus = status} = do
@@ -183,13 +185,7 @@ logDeleteSubscription :: StoreLog 'WriteMode -> NtfSubscriptionId -> IO ()
 logDeleteSubscription s subId = logNtfStoreRecord s $ DeleteSubscription subId
 
 readWriteNtfStore :: FilePath -> NtfStore -> IO (StoreLog 'WriteMode)
-readWriteNtfStore f st = do
-  whenM (doesFileExist f) $ do
-    readNtfStore f st
-    renameFile f $ f <> ".bak"
-  s <- openWriteStoreLog f
-  writeNtfStore s st
-  pure s
+readWriteNtfStore = readWriteStoreLog readNtfStore writeNtfStore
 
 readNtfStore :: FilePath -> NtfStore -> IO ()
 readNtfStore f st = mapM_ (addNtfLogRecord . LB.toStrict) . LB.lines =<< LB.readFile f

src/Simplex/Messaging/Notifications/Transport.hs

@@ -169,5 +169,6 @@ ntfTHandle c = THandle {connection = c, params}
           thServerVRange = versionToRange v,
           thAuth = Nothing,
           implySessId = False,
+          encryptBlock = Nothing,
           batch = False
         }

src/Simplex/Messaging/Notifications/Types.hs

@@ -99,8 +99,8 @@ type NtfActionTs = UTCTime
 data NtfSubNTFAction
   = NSACreate
   | NSACheck
-  | NSADelete
-  | NSARotate
+  | NSADelete -- deprecated
+  | NSARotate -- deprecated
   deriving (Show)
 
 instance Encoding NtfSubNTFAction where

src/Simplex/Messaging/Protocol.hs

@@ -258,13 +258,14 @@ supportedSMPClientVRange = mkVersionRange initialSMPClientVersion currentSMPClie
 -- TODO v6.0 remove dependency on version
 maxMessageLength :: VersionSMP -> Int
 maxMessageLength v
+  | v >= encryptedBlockSMPVersion = 16048 -- max 16051
   | v >= sendingProxySMPVersion = 16064 -- max 16067
-  | otherwise = 16088 -- 16064 - always use this size to determine allowed ranges
+  | otherwise = 16088 -- 16048 - always use this size to determine allowed ranges
 
 paddedProxiedTLength :: Int
-paddedProxiedTLength = 16242 -- 16241 .. 16243
+paddedProxiedTLength = 16226 -- 16225 .. 16227
 
--- TODO v6.0 change to 16064
+-- TODO v7.0 change to 16048
 type MaxMessageLen = 16088
 
 -- 16 extra bytes: 8 for timestamp and 8 for flags (7 flags and the space, only 1 flag is currently used)
@@ -272,10 +273,10 @@ type MaxRcvMessageLen = MaxMessageLen + 16 -- 16104, the padded size is 16106
 
 -- it is shorter to allow per-queue e2e encryption DH key in the "public" header
 e2eEncConfirmationLength :: Int
-e2eEncConfirmationLength = 15920 -- 15881 .. 15976
+e2eEncConfirmationLength = 15904 -- 15865 .. 15960
 
 e2eEncMessageLength :: Int
-e2eEncMessageLength = 16016 -- 16004 .. 16021
+e2eEncMessageLength = 16000 -- 15988 .. 16005
 
 -- | SMP protocol clients
 data Party = Recipient | Sender | Notifier | ProxiedClient
@@ -398,8 +399,6 @@ data Command (p :: Party) where
   NKEY :: NtfPublicAuthKey -> RcvNtfPublicDhKey -> Command Recipient
   NDEL :: Command Recipient
   GET :: Command Recipient
-  -- ACK v1 has to be supported for encoding/decoding
-  -- ACK :: Command Recipient
   ACK :: MsgId -> Command Recipient
   OFF :: Command Recipient
   DEL :: Command Recipient
@@ -484,6 +483,7 @@ data BrokerMsg where
   RRES :: EncFwdResponse -> BrokerMsg -- relay to proxy
   PRES :: EncResponse -> BrokerMsg -- proxy to client
   END :: BrokerMsg
+  DELD :: BrokerMsg
   INFO :: QueueInfo -> BrokerMsg
   OK :: BrokerMsg
   ERR :: ErrorType -> BrokerMsg
@@ -538,11 +538,13 @@ messageId :: Message -> MsgId
 messageId = \case
   Message {msgId} -> msgId
   MessageQuota {msgId} -> msgId
+{-# INLINE messageId #-}
 
 messageTs :: Message -> SystemTime
 messageTs = \case
   Message {msgTs} -> msgTs
   MessageQuota {msgTs} -> msgTs
+{-# INLINE messageTs #-}
 
 newtype EncRcvMsgBody = EncRcvMsgBody ByteString
   deriving (Eq, Show)
@@ -705,6 +707,7 @@ data BrokerMsgTag
   | RRES_
   | PRES_
   | END_
+  | DELD_
   | INFO_
   | OK_
   | ERR_
@@ -778,6 +781,7 @@ instance Encoding BrokerMsgTag where
     RRES_ -> "RRES"
     PRES_ -> "PRES"
     END_ -> "END"
+    DELD_ -> "DELD"
     INFO_ -> "INFO"
     OK_ -> "OK"
     ERR_ -> "ERR"
@@ -794,6 +798,7 @@ instance ProtocolMsgTag BrokerMsgTag where
     "RRES" -> Just RRES_
     "PRES" -> Just PRES_
     "END" -> Just END_
+    "DELD" -> Just DELD_
     "INFO" -> Just INFO_
     "OK" -> Just OK_
     "ERR" -> Just ERR_
@@ -1193,6 +1198,8 @@ data ErrorType
     CRYPTO
   | -- | SMP queue capacity is exceeded on the server
     QUOTA
+  | -- | SMP server storage error
+    STORE {storeErr :: String}
   | -- | ACK command is sent without message to be acknowledged
     NO_MSG
   | -- | sent message is too large (> maxMessageLength = 16088 bytes)
@@ -1274,7 +1281,7 @@ transmissionP THandleParams {sessionId, implySessId} = do
       command <- A.takeByteString
       pure RawTransmission {authenticator, authorized = authorized', sessId, corrId, entityId, command}
 
-class (ProtocolEncoding v err msg, ProtocolEncoding v err (ProtoCommand msg), Show err, Show msg) => Protocol v err msg | msg -> v, msg -> err where
+class (ProtocolTypeI (ProtoType msg), ProtocolEncoding v err msg, ProtocolEncoding v err (ProtoCommand msg), Show err, Show msg) => Protocol v err msg | msg -> v, msg -> err where
   type ProtoCommand msg = cmd | cmd -> msg
   type ProtoType msg = (sch :: ProtocolType) | sch -> msg
   protocolClientHandshake :: forall c. Transport c => c -> Maybe C.KeyPairX25519 -> C.KeyHash -> VersionRange v -> ExceptT TransportError IO (THandle v c 'TClient)
@@ -1423,6 +1430,9 @@ instance ProtocolEncoding SMPVersion ErrorType BrokerMsg where
     RRES (EncFwdResponse encBlock) -> e (RRES_, ' ', Tail encBlock)
     PRES (EncResponse encBlock) -> e (PRES_, ' ', Tail encBlock)
     END -> e END_
+    DELD
+      | v >= deletedEventSMPVersion -> e DELD_
+      | otherwise -> e END_
     INFO info -> e (INFO_, ' ', info)
     OK -> e OK_
     ERR err -> e (ERR_, ' ', err)
@@ -1448,6 +1458,7 @@ instance ProtocolEncoding SMPVersion ErrorType BrokerMsg where
     RRES_ -> RRES <$> (EncFwdResponse . unTail <$> _smpP)
     PRES_ -> PRES <$> (EncResponse . unTail <$> _smpP)
     END_ -> pure END
+    DELD_ -> pure DELD
     INFO_ -> INFO <$> _smpP
     OK_ -> pure OK
     ERR_ -> ERR <$> _smpP
@@ -1504,6 +1515,7 @@ instance Encoding ErrorType where
     AUTH -> "AUTH"
     CRYPTO -> "CRYPTO"
     QUOTA -> "QUOTA"
+    STORE err -> "STORE " <> smpEncode err
     EXPIRED -> "EXPIRED"
     NO_MSG -> "NO_MSG"
     LARGE_MSG -> "LARGE_MSG"
@@ -1519,6 +1531,7 @@ instance Encoding ErrorType where
       "AUTH" -> pure AUTH
       "CRYPTO" -> pure CRYPTO
       "QUOTA" -> pure QUOTA
+      "STORE" -> STORE <$> _smpP
       "EXPIRED" -> pure EXPIRED
       "NO_MSG" -> pure NO_MSG
       "LARGE_MSG" -> pure LARGE_MSG
@@ -1655,8 +1668,8 @@ batchTransmissions' batch bSize ts
 batchTransmissions_ :: Int -> NonEmpty (Either TransportError ByteString, r) -> [TransportBatch r]
 batchTransmissions_ bSize = addBatch . foldr addTransmission ([], 0, 0, [], [])
   where
-    -- 3 = 2 bytes reserved for pad size + 1 for transmission count
-    bSize' = bSize - 3
+    -- 19 = 2 bytes reserved for pad size + 1 for transmission count + 16 auth tag from block encryption
+    bSize' = bSize - 19
     addTransmission :: (Either TransportError ByteString, r) -> ([TransportBatch r], Int, Int, [ByteString], [r]) -> ([TransportBatch r], Int, Int, [ByteString], [r])
     addTransmission (t_, r) acc@(bs, !len, !n, ss, rs) = case t_ of
       Left e -> (TBError e r : addBatch acc, 0, 0, [], [])

src/Simplex/Messaging/Server/CLI.hs

@@ -6,6 +6,8 @@
 {-# LANGUAGE LambdaCase #-}
 {-# LANGUAGE NamedFieldPuns #-}
 {-# LANGUAGE OverloadedStrings #-}
+{-# LANGUAGE ScopedTypeVariables #-}
+{-# LANGUAGE TupleSections #-}
 {-# LANGUAGE TypeApplications #-}
 
 module Simplex.Messaging.Server.CLI where
@@ -16,6 +18,7 @@ import Data.ByteString.Char8 (ByteString)
 import qualified Data.ByteString.Char8 as B
 import Data.Either (fromRight)
 import Data.Ini (Ini, lookupValue)
+import Data.List ((\\))
 import Data.Text (Text)
 import qualified Data.Text as T
 import Data.Text.Encoding (encodeUtf8)
@@ -27,7 +30,7 @@ import Options.Applicative
 import Simplex.Messaging.Encoding.String
 import Simplex.Messaging.Protocol (ProtoServerWithAuth (..), ProtocolServer (..), ProtocolTypeI)
 import Simplex.Messaging.Transport (ATransport (..), TLS, Transport (..))
-import Simplex.Messaging.Transport.Server (loadFingerprint)
+import Simplex.Messaging.Transport.Server (AddHTTP, loadFileFingerprint)
 import Simplex.Messaging.Transport.WebSockets (WS)
 import Simplex.Messaging.Util (eitherToMaybe, whenM)
 import System.Directory (doesDirectoryExist, listDirectory, removeDirectoryRecursive, removePathForcibly)
@@ -41,11 +44,11 @@ import Text.Read (readMaybe)
 exitError :: String -> IO a
 exitError msg = putStrLn msg >> exitFailure
 
-confirmOrExit :: String -> IO ()
-confirmOrExit s =
+confirmOrExit :: String -> String -> IO ()
+confirmOrExit s no =
   withPrompt (s <> "\nContinue (Y/n): ") $ do
     ok <- getLine
-    when (ok /= "Y") $ putStrLn "Server NOT deleted" >> exitFailure
+    when (ok /= "Y") $ putStrLn no >> exitFailure
 
 data SignAlgorithm = ED448 | ED25519
   deriving (Read, Show)
@@ -137,7 +140,7 @@ createServerX509_ createCA cfgPath x509cfg = do
         )
 
     saveFingerprint = do
-      Fingerprint fp <- loadFingerprint $ c caCrtFile
+      Fingerprint fp <- loadFileFingerprint $ c caCrtFile
       withFile (c fingerprintFile) WriteMode (`B.hPutStrLn` strEncode fp)
       pure fp
 
@@ -266,26 +269,37 @@ settingIsOn section name ini
 checkSavedFingerprint :: FilePath -> X509Config -> IO ByteString
 checkSavedFingerprint cfgPath x509cfg = do
   savedFingerprint <- withFile (c fingerprintFile) ReadMode hGetLine
-  Fingerprint fp <- loadFingerprint (c caCrtFile)
+  Fingerprint fp <- loadFileFingerprint (c caCrtFile)
   when (B.pack savedFingerprint /= strEncode fp) $
     exitError "Stored fingerprint is invalid."
   pure fp
   where
     c = combine cfgPath . ($ x509cfg)
 
-iniTransports :: Ini -> [(String, ATransport)]
+iniTransports :: Ini -> [(ServiceName, ATransport, AddHTTP)]
 iniTransports ini =
-  let port = T.unpack $ strictIni "TRANSPORT" "port" ini
-      enableWebsockets = (== "on") $ strictIni "TRANSPORT" "websockets" ini
-   in (port, transport @TLS) : [("80", transport @WS) | enableWebsockets]
+  let smpPorts = ports $ strictIni "TRANSPORT" "port" ini
+      ws = strictIni "TRANSPORT" "websockets" ini
+      wsPorts
+        | ws == "off" = []
+        | ws == "on" = ["80"]
+        | otherwise = ports ws \\ smpPorts
+   in ts (transport @TLS) smpPorts <> ts (transport @WS) wsPorts
+  where
+    ts :: ATransport -> [ServiceName] -> [(ServiceName, ATransport, AddHTTP)]
+    ts t = map (\port -> (port, t, webPort == Just port))
+    webPort = T.unpack <$> eitherToMaybe (lookupValue "WEB" "https" ini)
+    ports = map T.unpack . T.splitOn ","
 
-printServerConfig :: [(ServiceName, ATransport)] -> Maybe FilePath -> IO ()
+printServerConfig :: [(ServiceName, ATransport, AddHTTP)] -> Maybe FilePath -> IO ()
 printServerConfig transports logFile = do
   putStrLn $ case logFile of
     Just f -> "Store log: " <> f
     _ -> "Store log disabled."
-  forM_ transports $ \(p, ATransport t) ->
-    putStrLn $ "Listening on port " <> p <> " (" <> transportName t <> ")..."
+  forM_ transports $ \(p, ATransport t, addHTTP) -> do
+    let descr = p <> " (" <> transportName t <> ")..."
+    putStrLn $ "Serving SMP protocol on port " <> descr
+    when addHTTP $ putStrLn $ "Serving static site on port " <> descr
 
 deleteDirIfExists :: FilePath -> IO ()
 deleteDirIfExists path = whenM (doesDirectoryExist path) $ removeDirectoryRecursive path

src/Simplex/Messaging/Server/Env/STM.hs

@@ -1,27 +1,33 @@
 {-# LANGUAGE DataKinds #-}
 {-# LANGUAGE DuplicateRecordFields #-}
+{-# LANGUAGE FlexibleContexts #-}
+{-# LANGUAGE GADTs #-}
+{-# LANGUAGE KindSignatures #-}
+{-# LANGUAGE LambdaCase #-}
 {-# LANGUAGE NamedFieldPuns #-}
 {-# LANGUAGE OverloadedStrings #-}
 {-# LANGUAGE ScopedTypeVariables #-}
 {-# LANGUAGE StrictData #-}
+{-# LANGUAGE TypeFamilies #-}
 
 module Simplex.Messaging.Server.Env.STM where
 
 import Control.Concurrent (ThreadId)
 import Control.Logger.Simple
 import Control.Monad
+import qualified Crypto.PubKey.RSA as RSA
 import Crypto.Random
 import Data.ByteString.Char8 (ByteString)
 import Data.Int (Int64)
 import Data.IntMap.Strict (IntMap)
 import qualified Data.IntMap.Strict as IM
+import Data.List (intercalate)
 import Data.List.NonEmpty (NonEmpty)
-import Data.Map.Strict (Map)
-import qualified Data.Map.Strict as M
 import Data.Maybe (isJust, isNothing)
 import qualified Data.Text as T
 import Data.Time.Clock (getCurrentTime)
 import Data.Time.Clock.System (SystemTime)
+import qualified Data.X509 as X
 import Data.X509.Validation (Fingerprint (..))
 import Network.Socket (ServiceName)
 import qualified Network.TLS as T
@@ -33,28 +39,37 @@ import qualified Simplex.Messaging.Crypto as C
 import Simplex.Messaging.Protocol
 import Simplex.Messaging.Server.Expiration
 import Simplex.Messaging.Server.Information
+import Simplex.Messaging.Server.MsgStore.Journal
 import Simplex.Messaging.Server.MsgStore.STM
-import Simplex.Messaging.Server.QueueStore (NtfCreds (..), QueueRec (..))
+import Simplex.Messaging.Server.MsgStore.Types
+import Simplex.Messaging.Server.NtfStore
+import Simplex.Messaging.Server.QueueStore
 import Simplex.Messaging.Server.QueueStore.STM
 import Simplex.Messaging.Server.Stats
 import Simplex.Messaging.Server.StoreLog
 import Simplex.Messaging.TMap (TMap)
 import qualified Simplex.Messaging.TMap as TM
 import Simplex.Messaging.Transport (ATransport, VersionRangeSMP, VersionSMP)
-import Simplex.Messaging.Transport.Server (SocketState, TransportServerConfig, alpn, loadFingerprint, loadTLSServerParams, newSocketState)
+import Simplex.Messaging.Transport.Server
+import System.Directory (doesFileExist)
+import System.Exit (exitFailure)
 import System.IO (IOMode (..))
 import System.Mem.Weak (Weak)
 import UnliftIO.STM
 
 data ServerConfig = ServerConfig
-  { transports :: [(ServiceName, ATransport)],
+  { transports :: [(ServiceName, ATransport, AddHTTP)],
     smpHandshakeTimeout :: Int,
     tbqSize :: Natural,
+    msgStoreType :: AMSType,
     msgQueueQuota :: Int,
+    maxJournalMsgCount :: Int,
+    maxJournalStateLines :: Int,
     queueIdBytes :: Int,
     msgIdBytes :: Int,
     storeLogFile :: Maybe FilePath,
     storeMsgsFile :: Maybe FilePath,
+    storeNtfsFile :: Maybe FilePath,
     -- | set to False to prohibit creating new queues
     allowNewQueues :: Bool,
     -- | simple password that the clients need to pass in handshake to be able to create new queues
@@ -64,6 +79,11 @@ data ServerConfig = ServerConfig
     controlPortAdminAuth :: Maybe BasicAuth,
     -- | time after which the messages can be removed from the queues and check interval, seconds
     messageExpiration :: Maybe ExpirationConfig,
+    expireMessagesOnStart :: Bool,
+    -- | interval of inactivity after which journal queue is closed
+    idleQueueInterval :: Int64,
+    -- | notification expiration interval (seconds)
+    notificationExpiration :: ExpirationConfig,
     -- | time after which the socket with inactive client can be disconnected (without any messages or commands, incl. PING),
     -- and check interval, seconds
     inactiveClientExpiration :: Maybe ExpirationConfig,
@@ -76,12 +96,12 @@ data ServerConfig = ServerConfig
     serverStatsLogFile :: FilePath,
     -- | file to save and restore stats
     serverStatsBackupFile :: Maybe FilePath,
+    -- | notification delivery interval
+    ntfDeliveryInterval :: Int,
     -- | interval between sending pending END events to unsubscribed clients, seconds
     pendingENDInterval :: Int,
-    -- | CA certificate private key is not needed for initialization
-    caCertificateFile :: FilePath,
-    privateKeyFile :: FilePath,
-    certificateFile :: FilePath,
+    smpCredentials :: ServerCredentials,
+    httpCredentials :: Maybe ServerCredentials,
     -- | SMP client-server protocol version range
     smpServerVRange :: VersionRangeSMP,
     -- | TCP transport config
@@ -103,7 +123,20 @@ defaultMessageExpiration :: ExpirationConfig
 defaultMessageExpiration =
   ExpirationConfig
     { ttl = defMsgExpirationDays * 86400, -- seconds
-      checkInterval = 43200 -- seconds, 12 hours
+      checkInterval = 14400 -- seconds, 4 hours
+    }
+
+defaultIdleQueueInterval :: Int64
+defaultIdleQueueInterval = 28800 -- seconds, 8 hours
+
+defNtfExpirationHours :: Int64
+defNtfExpirationHours = 24
+
+defaultNtfExpiration :: ExpirationConfig
+defaultNtfExpiration =
+  ExpirationConfig
+    { ttl = defNtfExpirationHours * 3600, -- seconds
+      checkInterval = 3600 -- seconds, 1 hour
     }
 
 defaultInactiveClientExpiration :: ExpirationConfig
@@ -116,32 +149,64 @@ defaultInactiveClientExpiration =
 defaultProxyClientConcurrency :: Int
 defaultProxyClientConcurrency = 32
 
+journalMsgStoreDepth :: Int
+journalMsgStoreDepth = 5
+
+defaultMaxJournalStateLines :: Int
+defaultMaxJournalStateLines = 16
+
+defaultMaxJournalMsgCount :: Int
+defaultMaxJournalMsgCount = 256
+
+defaultMsgQueueQuota :: Int
+defaultMsgQueueQuota = 128
+
+defaultStateTailSize :: Int
+defaultStateTailSize = 512
+
 data Env = Env
   { config :: ServerConfig,
+    serverActive :: TVar Bool,
     serverInfo :: ServerInformation,
     server :: Server,
     serverIdentity :: KeyHash,
-    queueStore :: QueueStore,
-    msgStore :: STMMsgStore,
+    msgStore :: AMsgStore,
+    ntfStore :: NtfStore,
     random :: TVar ChaChaDRG,
-    storeLog :: Maybe (StoreLog 'WriteMode),
-    tlsServerParams :: T.ServerParams,
+    tlsServerCreds :: T.Credential,
+    httpServerCreds :: Maybe T.Credential,
     serverStats :: ServerStats,
-    sockets :: SocketState,
+    sockets :: TVar [(ServiceName, SocketState)],
     clientSeq :: TVar ClientId,
-    clients :: TVar (IntMap (Maybe Client)),
+    clients :: TVar (IntMap (Maybe AClient)),
     proxyAgent :: ProxyAgent -- senders served on this proxy
   }
 
+type family MsgStore s where
+  MsgStore 'MSMemory = STMMsgStore
+  MsgStore 'MSJournal = JournalMsgStore
+
+data AMsgStore = forall s. (STMQueueStore (MsgStore s), MsgStoreClass (MsgStore s)) => AMS (SMSType s) (MsgStore s)
+
+data AStoreQueue = forall s. MsgStoreClass (MsgStore s) => ASQ (SMSType s) (StoreQueue (MsgStore s))
+
+data AMsgStoreCfg = forall s. MsgStoreClass (MsgStore s) => AMSC (SMSType s) (MsgStoreConfig (MsgStore s))
+
+msgPersistence :: AMsgStoreCfg -> Bool
+msgPersistence (AMSC SMSMemory (STMStoreConfig {storePath})) = isJust storePath
+msgPersistence (AMSC SMSJournal _) = True
+
 type Subscribed = Bool
 
 data Server = Server
   { subscribedQ :: TQueue (RecipientId, ClientId, Subscribed),
-    subscribers :: TMap RecipientId (TVar Client),
+    subscribers :: TMap RecipientId (TVar AClient),
     ntfSubscribedQ :: TQueue (NotifierId, ClientId, Subscribed),
-    notifiers :: TMap NotifierId (TVar Client),
-    pendingENDs :: TVar (IntMap (NonEmpty RecipientId)),
-    pendingNtfENDs :: TVar (IntMap (NonEmpty NotifierId)),
+    notifiers :: TMap NotifierId (TVar AClient),
+    subClients :: TVar (IntMap AClient), -- clients with SMP subscriptions
+    ntfSubClients :: TVar (IntMap AClient), -- clients with Ntf subscriptions
+    pendingSubEvents :: TVar (IntMap (NonEmpty (RecipientId, Subscribed))),
+    pendingNtfSubEvents :: TVar (IntMap (NonEmpty (NotifierId, Subscribed))),
     savingLock :: Lock
   }
 
@@ -151,11 +216,16 @@ newtype ProxyAgent = ProxyAgent
 
 type ClientId = Int
 
-data Client = Client
+data AClient = forall s. MsgStoreClass (MsgStore s) => AClient (SMSType s) (Client (MsgStore s))
+
+clientId' :: AClient -> ClientId
+clientId' (AClient _ Client {clientId}) = clientId
+
+data Client s = Client
   { clientId :: ClientId,
     subscriptions :: TMap RecipientId Sub,
     ntfSubscriptions :: TMap NotifierId (),
-    rcvQ :: TBQueue (NonEmpty (Maybe QueueRec, Transmission Cmd)),
+    rcvQ :: TBQueue (NonEmpty (Maybe (StoreQueue s, QueueRec), Transmission Cmd)),
     sndQ :: TBQueue (NonEmpty (Transmission BrokerMsg)),
     msgQ :: TBQueue (NonEmpty (Transmission BrokerMsg)),
     procThreads :: TVar Int,
@@ -184,13 +254,15 @@ newServer = do
   subscribers <- TM.emptyIO
   ntfSubscribedQ <- newTQueueIO
   notifiers <- TM.emptyIO
-  pendingENDs <- newTVarIO IM.empty
-  pendingNtfENDs <- newTVarIO IM.empty
-  savingLock <- atomically createLock
-  return Server {subscribedQ, subscribers, ntfSubscribedQ, notifiers, pendingENDs, pendingNtfENDs, savingLock}
+  subClients <- newTVarIO IM.empty
+  ntfSubClients <- newTVarIO IM.empty
+  pendingSubEvents <- newTVarIO IM.empty
+  pendingNtfSubEvents <- newTVarIO IM.empty
+  savingLock <- createLockIO
+  return Server {subscribedQ, subscribers, ntfSubscribedQ, notifiers, subClients, ntfSubClients, pendingSubEvents, pendingNtfSubEvents, savingLock}
 
-newClient :: ClientId -> Natural -> VersionSMP -> ByteString -> SystemTime -> IO Client
-newClient clientId qSize thVersion sessionId createdAt = do
+newClient :: SMSType s -> ClientId -> Natural -> VersionSMP -> ByteString -> SystemTime -> IO (Client (MsgStore s))
+newClient _msType clientId qSize thVersion sessionId createdAt = do
   subscriptions <- TM.emptyIO
   ntfSubscriptions <- TM.emptyIO
   rcvQ <- newTBQueueIO qSize
@@ -216,38 +288,54 @@ newProhibitedSub = do
   return Sub {subThread = ProhibitSub, delivered}
 
 newEnv :: ServerConfig -> IO Env
-newEnv config@ServerConfig {caCertificateFile, certificateFile, privateKeyFile, storeLogFile, smpAgentCfg, transportConfig, information, messageExpiration} = do
+newEnv config@ServerConfig {smpCredentials, httpCredentials, storeLogFile, msgStoreType, storeMsgsFile, smpAgentCfg, information, messageExpiration, idleQueueInterval, msgQueueQuota, maxJournalMsgCount, maxJournalStateLines} = do
+  serverActive <- newTVarIO True
   server <- newServer
-  queueStore <- newQueueStore
-  msgStore <- newMsgStore
+  msgStore@(AMS _ store) <- case msgStoreType of
+    AMSType SMSMemory -> AMS SMSMemory <$> newMsgStore STMStoreConfig {storePath = storeMsgsFile, quota = msgQueueQuota}
+    AMSType SMSJournal -> case storeMsgsFile of
+      Just storePath -> 
+        let cfg = JournalStoreConfig {storePath, quota = msgQueueQuota, pathParts = journalMsgStoreDepth, maxMsgCount = maxJournalMsgCount, maxStateLines = maxJournalStateLines, stateTailSize = defaultStateTailSize, idleInterval = idleQueueInterval}
+         in AMS SMSJournal <$> newMsgStore cfg
+      Nothing -> putStrLn "Error: journal msg store require path in [STORE_LOG], restore_messages" >> exitFailure
+  ntfStore <- NtfStore <$> TM.emptyIO
   random <- C.newRandom
-  storeLog <-
-    forM storeLogFile $ \f -> do
-      logInfo $ "restoring queues from file " <> T.pack f
-      restoreQueues queueStore f
-  tlsServerParams <- loadTLSServerParams caCertificateFile certificateFile privateKeyFile (alpn transportConfig)
-  Fingerprint fp <- loadFingerprint caCertificateFile
+  forM_ storeLogFile $ \f -> do
+    logInfo $ "restoring queues from file " <> T.pack f
+    sl <- readWriteQueueStore f store
+    setStoreLog store sl
+  tlsServerCreds <- getCredentials "SMP" smpCredentials
+  httpServerCreds <- mapM (getCredentials "HTTPS") httpCredentials
+  mapM_ checkHTTPSCredentials httpServerCreds
+  Fingerprint fp <- loadFingerprint smpCredentials
   let serverIdentity = KeyHash fp
   serverStats <- newServerStats =<< getCurrentTime
-  sockets <- newSocketState
+  sockets <- newTVarIO []
   clientSeq <- newTVarIO 0
   clients <- newTVarIO mempty
   proxyAgent <- newSMPProxyAgent smpAgentCfg random
-  pure Env {config, serverInfo, server, serverIdentity, queueStore, msgStore, random, storeLog, tlsServerParams, serverStats, sockets, clientSeq, clients, proxyAgent}
+  pure Env {serverActive, config, serverInfo, server, serverIdentity, msgStore, ntfStore, random, tlsServerCreds, httpServerCreds, serverStats, sockets, clientSeq, clients, proxyAgent}
   where
-    restoreQueues :: QueueStore -> FilePath -> IO (StoreLog 'WriteMode)
-    restoreQueues QueueStore {queues, senders, notifiers} f = do
-      (qs, s) <- readWriteStoreLog f
-      atomically . writeTVar queues =<< mapM newTVarIO qs
-      atomically $ writeTVar senders $! M.foldr' addSender M.empty qs
-      atomically $ writeTVar notifiers $! M.foldr' addNotifier M.empty qs
-      pure s
-    addSender :: QueueRec -> Map SenderId RecipientId -> Map SenderId RecipientId
-    addSender q = M.insert (senderId q) (recipientId q)
-    addNotifier :: QueueRec -> Map NotifierId RecipientId -> Map NotifierId RecipientId
-    addNotifier q = case notifier q of
-      Nothing -> id
-      Just NtfCreds {notifierId} -> M.insert notifierId (recipientId q)
+    getCredentials protocol creds = do
+      files <- missingCreds
+      unless (null files) $ do
+        putStrLn $ "Error: no " <> protocol <> " credentials: " <> intercalate ", " files
+        when (protocol == "HTTPS") $ putStrLn letsEncrypt
+        exitFailure
+      loadServerCredential creds
+      where
+        missingfile f = (\y -> [f | not y]) <$> doesFileExist f
+        missingCreds = do
+          let files = maybe id (:) (caCertificateFile creds) [certificateFile creds, privateKeyFile creds]
+           in concat <$> mapM missingfile files
+    checkHTTPSCredentials (X.CertificateChain cc, _k) =
+      -- LetsEncrypt provides ECDSA with insecure curve p256 (https://safecurves.cr.yp.to)
+      case map (X.signedObject . X.getSigned) cc of
+        X.Certificate {X.certPubKey = X.PubKeyRSA rsa} : _ca | RSA.public_size rsa >= 512 -> pure ()
+        _ -> do
+          putStrLn $ "Error: unsupported HTTPS credentials, required 4096-bit RSA\n" <> letsEncrypt
+          exitFailure
+    letsEncrypt = "Use Let's Encrypt to generate: certbot certonly --standalone -d yourdomainname --key-type rsa --rsa-key-size 4096"
     serverInfo =
       ServerInformation
         { information,
@@ -263,10 +351,13 @@ newEnv config@ServerConfig {caCertificateFile, certificateFile, privateKeyFile,
       where
         persistence
           | isNothing storeLogFile = SPMMemoryOnly
-          | isJust (storeMsgsFile config) = SPMMessages
+          | isJust storeMsgsFile = SPMMessages
           | otherwise = SPMQueues
 
 newSMPProxyAgent :: SMPClientAgentConfig -> TVar ChaChaDRG -> IO ProxyAgent
 newSMPProxyAgent smpAgentCfg random = do
   smpAgent <- newSMPClientAgent smpAgentCfg random
   pure ProxyAgent {smpAgent}
+
+readWriteQueueStore :: STMQueueStore s => FilePath -> s -> IO (StoreLog 'WriteMode)
+readWriteQueueStore = readWriteStoreLog readQueueStore writeQueueStore

src/Simplex/Messaging/Server/Information.hs

@@ -1,14 +1,21 @@
 {-# LANGUAGE DataKinds #-}
 {-# LANGUAGE DuplicateRecordFields #-}
+{-# LANGUAGE LambdaCase #-}
+{-# LANGUAGE NamedFieldPuns #-}
+{-# LANGUAGE OverloadedStrings #-}
 {-# LANGUAGE StrictData #-}
 {-# LANGUAGE TemplateHaskell #-}
 
 module Simplex.Messaging.Server.Information where
 
+import Data.Aeson (FromJSON (..), ToJSON (..))
 import qualified Data.Aeson.TH as J
+import qualified Data.Attoparsec.ByteString.Char8 as A
 import Data.Int (Int64)
+import Data.Maybe (isJust)
 import Data.Text (Text)
 import Simplex.Messaging.Agent.Protocol (ConnectionMode (..), ConnectionRequestUri)
+import Simplex.Messaging.Encoding.String
 import Simplex.Messaging.Parsers (defaultJSON, dropPrefix, enumJSON)
 
 data ServerInformation = ServerInformation
@@ -36,16 +43,59 @@ data ServerPublicInfo = ServerPublicInfo
     adminContacts :: Maybe ServerContactAddress,
     complaintsContacts :: Maybe ServerContactAddress,
     hosting :: Maybe Entity,
+    hostingType :: Maybe HostingType,
     serverCountry :: Maybe Text
   }
   deriving (Show)
 
+emptyServerInfo :: Text -> ServerPublicInfo
+emptyServerInfo sourceCode =
+  ServerPublicInfo
+    { sourceCode,
+      usageConditions = Nothing,
+      operator = Nothing,
+      website = Nothing,
+      adminContacts = Nothing,
+      complaintsContacts = Nothing,
+      hosting = Nothing,
+      hostingType = Nothing,
+      serverCountry = Nothing
+    }
+
+hasServerInfo :: ServerPublicInfo -> Bool
+hasServerInfo ServerPublicInfo {usageConditions, operator, website, adminContacts, complaintsContacts, hosting, hostingType, serverCountry} =
+  isJust usageConditions || isJust operator || isJust website || isJust adminContacts || isJust complaintsContacts || isJust hosting || isJust hostingType || isJust serverCountry
+
 data ServerPersistenceMode = SPMMemoryOnly | SPMQueues | SPMMessages
   deriving (Show)
 
 data ServerConditions = ServerConditions {conditions :: Text, amendments :: Maybe Text}
   deriving (Show)
 
+data HostingType = HTVirtual | HTDedicated | HTColocation | HTOwned
+  deriving (Show)
+
+instance StrEncoding HostingType where
+  strEncode = \case
+    HTVirtual -> "virtual"
+    HTDedicated -> "dedicated"
+    HTColocation -> "colocation"
+    HTOwned -> "owned"
+  strP =
+    A.takeTill (== ' ') >>= \case
+      "virtual" -> pure HTVirtual
+      "dedicated" -> pure HTDedicated
+      "colocation" -> pure HTColocation
+      "owned" -> pure HTOwned
+      _ -> fail "bad HostingType"
+
+instance FromJSON HostingType where
+  parseJSON = strParseJSON "HostingType"
+
+instance ToJSON HostingType where
+  toJSON = strToJSON
+  toEncoding = strToJEncoding
+
 data Entity = Entity {name :: Text, country :: Maybe Text}
   deriving (Show)
 

src/Simplex/Messaging/Server/Main.hs

@@ -1,5 +1,6 @@
 {-# LANGUAGE ApplicativeDo #-}
 {-# LANGUAGE DuplicateRecordFields #-}
+{-# LANGUAGE GADTs #-}
 {-# LANGUAGE LambdaCase #-}
 {-# LANGUAGE NamedFieldPuns #-}
 {-# LANGUAGE OverloadedLists #-}
@@ -8,15 +9,18 @@
 {-# LANGUAGE ScopedTypeVariables #-}
 {-# LANGUAGE StrictData #-}
 {-# LANGUAGE TypeApplications #-}
+{-# OPTIONS_GHC -fno-warn-ambiguous-fields #-}
 
 module Simplex.Messaging.Server.Main where
 
 import Control.Concurrent.STM
+import Control.Exception (finally)
 import Control.Logger.Simple
-import Control.Monad (void, when, (<$!>))
+import Control.Monad
 import Data.ByteString.Char8 (ByteString)
 import qualified Data.ByteString.Char8 as B
 import Data.Char (isAlpha, isAscii, toUpper)
+import Data.Either (fromRight)
 import Data.Functor (($>))
 import Data.Ini (Ini, lookupValue, readIniFile)
 import Data.List (find, isPrefixOf)
@@ -24,68 +28,149 @@ import qualified Data.List.NonEmpty as L
 import Data.Maybe (fromMaybe, isJust, isNothing)
 import Data.Text (Text)
 import qualified Data.Text as T
-import Data.Text.Encoding (encodeUtf8)
+import Data.Text.Encoding (decodeLatin1, encodeUtf8)
 import qualified Data.Text.IO as T
 import Network.Socket (HostName)
 import Options.Applicative
 import Simplex.Messaging.Agent.Protocol (connReqUriP')
-import Simplex.Messaging.Client (HostMode (..), NetworkConfig (..), ProtocolClientConfig (..), SocksMode (..), defaultNetworkConfig)
+import Simplex.Messaging.Client (HostMode (..), NetworkConfig (..), ProtocolClientConfig (..), SocksMode (..), defaultNetworkConfig, textToHostMode)
 import Simplex.Messaging.Client.Agent (SMPClientAgentConfig (..), defaultSMPClientAgentConfig)
 import qualified Simplex.Messaging.Crypto as C
 import Simplex.Messaging.Encoding.String
 import Simplex.Messaging.Parsers (parseAll)
 import Simplex.Messaging.Protocol (BasicAuth (..), ProtoServerWithAuth (ProtoServerWithAuth), pattern SMPServer)
-import Simplex.Messaging.Server (runSMPServer)
+import Simplex.Messaging.Server (AttachHTTP, exportMessages, importMessages, printMessageStats, runSMPServer)
 import Simplex.Messaging.Server.CLI
-import Simplex.Messaging.Server.Env.STM (ServerConfig (..), defMsgExpirationDays, defaultInactiveClientExpiration, defaultMessageExpiration, defaultProxyClientConcurrency)
+import Simplex.Messaging.Server.Env.STM
 import Simplex.Messaging.Server.Expiration
 import Simplex.Messaging.Server.Information
-import Simplex.Messaging.Transport (batchCmdsSMPVersion, sendingProxySMPVersion, simplexMQVersion, supportedSMPHandshakes, supportedServerSMPRelayVRange)
-import Simplex.Messaging.Transport.Client (TransportHost (..))
-import Simplex.Messaging.Transport.Server (TransportServerConfig (..), defaultTransportServerConfig)
-import Simplex.Messaging.Util (eitherToMaybe, safeDecodeUtf8, tshow)
+import Simplex.Messaging.Server.MsgStore.Journal (JournalStoreConfig (..))
+import Simplex.Messaging.Server.MsgStore.Types (AMSType (..), SMSType (..), newMsgStore)
+import Simplex.Messaging.Server.QueueStore.STM (readQueueStore)
+import Simplex.Messaging.Transport (batchCmdsSMPVersion, sendingProxySMPVersion, simplexMQVersion, supportedServerSMPRelayVRange)
+import Simplex.Messaging.Transport.Client (SocksProxy, TransportHost (..), defaultSocksProxy)
+import Simplex.Messaging.Transport.Server (ServerCredentials (..), TransportServerConfig (..), defaultTransportServerConfig)
+import Simplex.Messaging.Util (eitherToMaybe, ifM, safeDecodeUtf8, tshow)
 import Simplex.Messaging.Version (mkVersionRange)
-import System.Directory (createDirectoryIfMissing, doesFileExist)
+import System.Directory (createDirectoryIfMissing, doesDirectoryExist, doesFileExist)
+import System.Exit (exitFailure)
 import System.FilePath (combine)
 import System.IO (BufferMode (..), hSetBuffering, stderr, stdout)
 import Text.Read (readMaybe)
 
 smpServerCLI :: FilePath -> FilePath -> IO ()
-smpServerCLI = smpServerCLI_ (\_ _ _ -> pure ()) (\_ -> pure ())
+smpServerCLI = smpServerCLI_ (\_ _ _ -> pure ()) (\_ -> pure ()) (\_ -> error "attachStaticFiles not available")
 
-smpServerCLI_ :: (ServerInformation -> Maybe TransportHost -> FilePath -> IO ()) -> (EmbeddedWebParams -> IO ()) -> FilePath -> FilePath -> IO ()
-smpServerCLI_ generateSite serveStaticFiles cfgPath logPath =
+smpServerCLI_ ::
+  (ServerInformation -> Maybe TransportHost -> FilePath -> IO ()) ->
+  (EmbeddedWebParams -> IO ()) ->
+  (FilePath -> (AttachHTTP -> IO ()) -> IO ()) ->
+  FilePath ->
+  FilePath ->
+  IO ()
+smpServerCLI_ generateSite serveStaticFiles attachStaticFiles cfgPath logPath =
   getCliCommand' (cliCommandP cfgPath logPath iniFile) serverVersion >>= \case
     Init opts ->
       doesFileExist iniFile >>= \case
         True -> exitError $ "Error: server is already initialized (" <> iniFile <> " exists).\nRun `" <> executableName <> " start`."
         _ -> initializeServer opts
-    OnlineCert certOpts ->
-      doesFileExist iniFile >>= \case
-        True -> genOnline cfgPath certOpts
-        _ -> exitError $ "Error: server is not initialized (" <> iniFile <> " does not exist).\nRun `" <> executableName <> " init`."
-    Start ->
-      doesFileExist iniFile >>= \case
-        True -> readIniFile iniFile >>= either exitError runServer
-        _ -> exitError $ "Error: server is not initialized (" <> iniFile <> " does not exist).\nRun `" <> executableName <> " init`."
+    OnlineCert certOpts -> withIniFile $ \_ -> genOnline cfgPath certOpts
+    Start -> withIniFile runServer
     Delete -> do
-      confirmOrExit "WARNING: deleting the server will make all queues inaccessible, because the server identity (certificate fingerprint) will change.\nTHIS CANNOT BE UNDONE!"
+      confirmOrExit
+        "WARNING: deleting the server will make all queues inaccessible, because the server identity (certificate fingerprint) will change.\nTHIS CANNOT BE UNDONE!"
+        "Server NOT deleted"
       deleteDirIfExists cfgPath
       deleteDirIfExists logPath
       putStrLn "Deleted configuration and log files"
+    Journal cmd -> withIniFile $ \ini -> do
+      msgsDirExists <- doesDirectoryExist storeMsgsJournalDir
+      msgsFileExists <- doesFileExist storeMsgsFilePath
+      let enableStoreLog = settingIsOn "STORE_LOG" "enable" ini
+      storeLogFile <- case enableStoreLog $> storeLogFilePath of
+        Just storeLogFile -> do
+          ifM
+            (doesFileExist storeLogFile)
+            (pure storeLogFile)
+            (putStrLn ("Store log file " <> storeLogFile <> " not found") >> exitFailure)
+        Nothing -> putStrLn "Store log disabled, see `[STORE_LOG] enable`" >> exitFailure
+      case cmd of
+        JCImport
+          | msgsFileExists && msgsDirExists -> exitConfigureMsgStorage
+          | msgsDirExists -> do
+              putStrLn $ storeMsgsJournalDir <> " directory already exists."
+              exitFailure
+          | not msgsFileExists -> do
+              putStrLn $ storeMsgsFilePath <> " file does not exists."
+              exitFailure
+          | otherwise -> do
+              confirmOrExit
+                ("WARNING: message log file " <> storeMsgsFilePath <> " will be imported to journal directory " <> storeMsgsJournalDir)
+                "Messages not imported"
+              ms <- newJournalMsgStore
+              readQueueStore storeLogFile ms
+              msgStats <- importMessages True ms storeMsgsFilePath Nothing -- no expiration
+              putStrLn "Import completed"
+              printMessageStats "Messages" msgStats
+              putStrLn $ case readMsgStoreType ini of
+                Right (AMSType SMSMemory) -> "store_messages set to `memory`, update it to `journal` in INI file"
+                Right (AMSType SMSJournal) -> "store_messages set to `journal`"
+                Left e -> e <> ", update it to `journal` in INI file"
+        JCExport
+          | msgsFileExists && msgsDirExists -> exitConfigureMsgStorage
+          | msgsFileExists -> do
+              putStrLn $ storeMsgsFilePath <> " file already exists."
+              exitFailure
+          | otherwise -> do
+              confirmOrExit
+                ("WARNING: journal directory " <> storeMsgsJournalDir <> " will be exported to message log file " <> storeMsgsFilePath)
+                "Journal not exported"
+              ms <- newJournalMsgStore
+              readQueueStore storeLogFile ms
+              exportMessages True ms storeMsgsFilePath False
+              putStrLn "Export completed"
+              putStrLn $ case readMsgStoreType ini of
+                Right (AMSType SMSMemory) -> "store_messages set to `memory`"
+                Right (AMSType SMSJournal) -> "store_messages set to `journal`, update it to `memory` in INI file"
+                Left e -> e <> ", update it to `memory` in INI file"
+        JCDelete
+          | not msgsDirExists -> do
+              putStrLn $ storeMsgsJournalDir <> " directory does not exists."
+              exitFailure
+          | otherwise -> do
+              confirmOrExit
+                ("WARNING: journal directory " <> storeMsgsJournalDir <> " will be permanently deleted.\nTHIS CANNOT BE UNDONE!")
+                "Messages NOT deleted"
+              deleteDirIfExists storeMsgsJournalDir
+              putStrLn $ "Deleted all messages in journal " <> storeMsgsJournalDir
   where
+    withIniFile a =
+      doesFileExist iniFile >>= \case
+        True -> readIniFile iniFile >>= either exitError a
+        _ -> exitError $ "Error: server is not initialized (" <> iniFile <> " does not exist).\nRun `" <> executableName <> " init`."
+    newJournalMsgStore = newMsgStore JournalStoreConfig {storePath = storeMsgsJournalDir, pathParts = journalMsgStoreDepth, quota = defaultMsgQueueQuota, maxMsgCount = defaultMaxJournalMsgCount, maxStateLines = defaultMaxJournalStateLines, stateTailSize = defaultStateTailSize, idleInterval = checkInterval defaultMessageExpiration}
     iniFile = combine cfgPath "smp-server.ini"
     serverVersion = "SMP server v" <> simplexMQVersion
-    defaultServerPort = "5223"
+    defaultServerPorts = "5223,443"
     executableName = "smp-server"
     storeLogFilePath = combine logPath "smp-server-store.log"
-    httpsCertFile = combine cfgPath "web.cert"
+    storeMsgsFilePath = combine logPath "smp-server-messages.log"
+    storeMsgsJournalDir = combine logPath "messages"
+    storeNtfsFilePath = combine logPath "smp-server-ntfs.log"
+    readMsgStoreType :: Ini -> Either String AMSType
+    readMsgStoreType = textToMsgStoreType . fromRight "memory" . lookupValue "STORE_LOG" "store_messages"
+    textToMsgStoreType = \case
+      "memory" -> Right $ AMSType SMSMemory
+      "journal" -> Right $ AMSType SMSJournal
+      s -> Left $ "invalid store_messages: " <> T.unpack s
+    httpsCertFile = combine cfgPath "web.crt"
     httpsKeyFile = combine cfgPath "web.key"
     defaultStaticPath = combine logPath "www"
     initializeServer opts@InitOptions {ip, fqdn, sourceCode = src', webStaticPath = sp', disableWeb = noWeb', scripted}
       | scripted = initialize opts
       | otherwise = do
           putStrLn "Use `smp-server init -h` for available options."
+          checkInitOptions opts
           void $ withPrompt "SMP server will be initialized (press Enter)" getLine
           enableStoreLog <- onOffPrompt "Enable store log to restore queues and messages on server restart" True
           logStats <- onOffPrompt "Enable logging daily statistics" False
@@ -95,16 +180,15 @@ smpServerCLI_ generateSite serveStaticFiles cfgPath logPath =
           host' <- withPrompt ("Enter server FQDN or IP address for certificate (" <> host <> "): ") getLine
           sourceCode' <- withPrompt ("Enter server source code URI (" <> maybe simplexmqSource T.unpack src' <> "): ") getServerSourceCode
           staticPath' <- withPrompt ("Enter path to store generated static site with server information (" <> fromMaybe defaultStaticPath sp' <> "): ") getLine
-          enableWeb <- onOffPrompt "Enable built-in web server for static site" (not noWeb')
           initialize
             opts
               { enableStoreLog,
                 logStats,
                 fqdn = if null host' then fqdn else Just host',
                 password,
-                sourceCode = (T.pack <$> sourceCode') <|> src',
+                sourceCode = (T.pack <$> sourceCode') <|> src' <|> Just (T.pack simplexmqSource),
                 webStaticPath = if null staticPath' then sp' else Just staticPath',
-                disableWeb = not enableWeb
+                disableWeb = noWeb'
               }
       where
         serverPassword =
@@ -116,7 +200,18 @@ smpServerCLI_ generateSite serveStaticFiles cfgPath logPath =
               case strDecode $ encodeUtf8 $ T.pack s of
                 Right auth -> pure . Just $ ServerPassword auth
                 _ -> putStrLn "Invalid password. Only latin letters, digits and symbols other than '@' and ':' are allowed" >> serverPassword
-        initialize InitOptions {enableStoreLog, logStats, signAlgorithm, password, sourceCode, webStaticPath, disableWeb} = do
+        checkInitOptions InitOptions {sourceCode, serverInfo, operatorCountry, hostingCountry} = do
+          let err_
+                | isNothing sourceCode && hasServerInfo serverInfo =
+                    Just "Error: passing any server information requires passing --source-code"
+                | isNothing (operator serverInfo) && isJust operatorCountry =
+                    Just "Error: passing --operator-country requires passing --operator"
+                | isNothing (hosting serverInfo) && isJust hostingCountry =
+                    Just "Error: passing --hosting-country requires passing --hosting"
+                | otherwise = Nothing
+          forM_ err_ $ \err -> putStrLn err >> exitFailure
+        initialize opts'@InitOptions {enableStoreLog, logStats, signAlgorithm, password, controlPort, socksProxy, ownDomains, sourceCode, webStaticPath, disableWeb} = do
+          checkInitOptions opts'
           clearDirIfExists cfgPath
           clearDirIfExists logPath
           createDirectoryIfMissing True cfgPath
@@ -124,9 +219,10 @@ smpServerCLI_ generateSite serveStaticFiles cfgPath logPath =
           let x509cfg = defaultX509Config {commonName = fromMaybe ip fqdn, signAlgorithm}
           fp <- createServerX509 cfgPath x509cfg
           basicAuth <- mapM createServerPassword password
+          controlPortPwds <- forM controlPort $ \_ -> let pwd = decodeLatin1 <$> randomBase64 18 in (,) <$> pwd <*> pwd
           let host = fromMaybe (if ip == "127.0.0.1" then "<hostnames>" else ip) fqdn
               srv = ProtoServerWithAuth (SMPServer [THDomainName host] "" (C.KeyHash fp)) basicAuth
-          T.writeFile iniFile $ iniFileContent host basicAuth $ Just "https://github.com/simplex-chat/simplexmq"
+          T.writeFile iniFile $ iniFileContent host basicAuth controlPortPwds
           putStrLn $ "Server initialized, please provide additional server information in " <> iniFile <> "."
           putStrLn $ "Run `" <> executableName <> " start` to start server."
           warnCAPrivateKeyFile cfgPath x509cfg
@@ -135,9 +231,10 @@ smpServerCLI_ generateSite serveStaticFiles cfgPath logPath =
           where
             createServerPassword = \case
               ServerPassword s -> pure s
-              SPRandom -> BasicAuth . strEncode <$> (atomically . C.randomBytes 32 =<< C.newRandom)
-            iniFileContent host basicAuth sourceCode' =
-              informationIniContent sourceCode'
+              SPRandom -> BasicAuth <$> randomBase64 32
+            randomBase64 n = strEncode <$> (atomically . C.randomBytes n =<< C.newRandom)
+            iniFileContent host basicAuth controlPortPwds =
+              informationIniContent opts'
                 <> "[STORE_LOG]\n\
                    \# The server uses STM memory for persistence,\n\
                    \# that will be lost on restart (e.g., as with redis).\n\
@@ -145,10 +242,15 @@ smpServerCLI_ generateSite serveStaticFiles cfgPath logPath =
                    \# and restoring it when the server is started.\n\
                    \# Log is compacted on start (deleted objects are removed).\n"
                 <> ("enable: " <> onOff enableStoreLog <> "\n\n")
-                <> "# Undelivered messages are optionally saved and restored when the server restarts,\n\
-                   \# they are preserved in the .bak file until the next restart.\n"
-                <> ("restore_messages: " <> onOff enableStoreLog <> "\n")
-                <> ("expire_messages_days: " <> tshow defMsgExpirationDays <> "\n\n")
+                <> "# Message storage mode: `memory` or `journal`.\n\
+                   \store_messages: memory\n\n\
+                   \# When store_messages is `memory`, undelivered messages are optionally saved and restored\n\
+                   \# when the server restarts, they are preserved in the .bak file until the next restart.\n"
+                <> ("restore_messages: " <> onOff enableStoreLog <> "\n\n")
+                <> "# Messages and notifications expiration periods.\n"
+                <> ("expire_messages_days: " <> tshow defMsgExpirationDays <> "\n")
+                <> "expire_messages_on_start: on\n"
+                <> ("expire_ntfs_hours: " <> tshow defNtfExpirationHours <> "\n\n")
                 <> "# Log daily server statistics to CSV file\n"
                 <> ("log_stats: " <> onOff logStats <> "\n\n")
                 <> "[AUTH]\n\
@@ -160,66 +262,77 @@ smpServerCLI_ generateSite serveStaticFiles cfgPath logPath =
                    \# smp://fingerprint:password@host1,host2\n\
                    \# The password will not be shared with the connecting contacts, you must share it only\n\
                    \# with the users who you want to allow creating messaging queues on your server.\n"
-                <> ( case basicAuth of
-                      Just auth -> "create_password: " <> safeDecodeUtf8 (strEncode auth)
-                      _ -> "# create_password: password to create new queues (any printable ASCII characters without whitespace, '@', ':' and '/')"
+                <> ( let noPassword = "password to create new queues and forward messages (any printable ASCII characters without whitespace, '@', ':' and '/')"
+                      in optDisabled basicAuth <> "create_password: " <> maybe noPassword (safeDecodeUtf8 . strEncode) basicAuth
                    )
-                <> "\n\n\
-                   \# control_port_admin_password:\n\
-                   \# control_port_user_password:\n\n\
+                <> "\n\n"
+                <> (optDisabled controlPortPwds <> "control_port_admin_password: " <> maybe "" fst controlPortPwds <> "\n")
+                <> (optDisabled controlPortPwds <> "control_port_user_password: " <> maybe "" snd controlPortPwds <> "\n")
+                <> "\n\
                    \[TRANSPORT]\n\
-                   \# host is only used to print server address on start\n"
+                   \# Host is only used to print server address on start.\n\
+                   \# You can specify multiple server ports.\n"
                 <> ("host: " <> T.pack host <> "\n")
-                <> ("port: " <> T.pack defaultServerPort <> "\n")
-                <> "log_tls_errors: off\n\
-                   \websockets: off\n\
-                   \# control_port: 5224\n\n\
+                <> ("port: " <> T.pack defaultServerPorts <> "\n")
+                <> "log_tls_errors: off\n\n\
+                   \# Use `websockets: 443` to run websockets server in addition to plain TLS.\n\
+                   \# This option is deprecated and should be used for testing only.\n\
+                   \# , port 443 should be specified in port above\n\
+                   \websockets: off\n"
+                <> (optDisabled controlPort <> "control_port: " <> tshow (fromMaybe defaultControlPort controlPort))
+                <> "\n\n\
                    \[PROXY]\n\
                    \# Network configuration for SMP proxy client.\n\
                    \# `host_mode` can be 'public' (default) or 'onion'.\n\
                    \# It defines prefferred hostname for destination servers with multiple hostnames.\n\
                    \# host_mode: public\n\
                    \# required_host_mode: off\n\n\
-                   \# The domain suffixes of the relays you operate (space-separated) to count as separate proxy statistics.\n\
-                   \# own_server_domains: \n\n\
+                   \# The domain suffixes of the relays you operate (space-separated) to count as separate proxy statistics.\n"
+                <> (optDisabled ownDomains <> "own_server_domains: " <> maybe "" (safeDecodeUtf8 . strEncode) ownDomains)
+                <> "\n\n\
                    \# SOCKS proxy port for forwarding messages to destination servers.\n\
-                   \# You may need a separate instance of SOCKS proxy for incoming single-hop requests.\n\
-                   \# socks_proxy: localhost:9050\n\n\
+                   \# You may need a separate instance of SOCKS proxy for incoming single-hop requests.\n"
+                <> (optDisabled socksProxy <> "socks_proxy: " <> maybe "localhost:9050" (safeDecodeUtf8 . strEncode) socksProxy)
+                <> "\n\n\
                    \# `socks_mode` can be 'onion' for SOCKS proxy to be used for .onion destination hosts only (default)\n\
                    \# or 'always' to be used for all destination hosts (can be used if it is an .onion server).\n\
                    \# socks_mode: onion\n\n\
                    \# Limit number of threads a client can spawn to process proxy commands in parrallel.\n"
-                <> ("# client_concurrency: " <> tshow defaultProxyClientConcurrency <> "\n\n")
-                <> "[INACTIVE_CLIENTS]\n\
+                <> ("# client_concurrency: " <> tshow defaultProxyClientConcurrency)
+                <> "\n\n\
+                   \[INACTIVE_CLIENTS]\n\
                    \# TTL and interval to check inactive clients\n\
-                   \disconnect: off\n"
-                <> ("# ttl: " <> tshow (ttl defaultInactiveClientExpiration) <> "\n")
-                <> ("# check_interval: " <> tshow (checkInterval defaultInactiveClientExpiration) <> "\n")
+                   \disconnect: on\n"
+                <> ("ttl: " <> tshow (ttl defaultInactiveClientExpiration) <> "\n")
+                <> ("check_interval: " <> tshow (checkInterval defaultInactiveClientExpiration))
                 <> "\n\n\
                    \[WEB]\n\
                    \# Set path to generate static mini-site for server information and qr codes/links\n"
                 <> ("static_path: " <> T.pack (fromMaybe defaultStaticPath webStaticPath) <> "\n\n")
                 <> "# Run an embedded server on this port\n\
                    \# Onion sites can use any port and register it in the hidden service config.\n\
-                   \# Running on a port 80 may require setting process capabilities.\n"
-                <> ((if disableWeb then "# " else "") <> "http: 8000\n\n")
-                <> "# You can run an embedded TLS web server too if you provide port and cert and key files.\n\
-                   \# Not required for running relay on onion address.\n\
-                   \# https: 443\n"
-                <> ("# cert: " <> T.pack httpsCertFile <> "\n")
-                <> ("# key: " <> T.pack httpsKeyFile <> "\n")
+                   \# Running on a port 80 may require setting process capabilities.\n\
+                   \# http: 8000\n\n\
+                   \# You can run an embedded TLS web server too if you provide port and cert and key files.\n\
+                   \# Not required for running relay on onion address.\n"
+                <> (webDisabled <> "https: 443\n")
+                <> (webDisabled <> "cert: " <> T.pack httpsCertFile <> "\n")
+                <> (webDisabled <> "key: " <> T.pack httpsKeyFile <> "\n")
+              where
+                webDisabled = if disableWeb then "# " else ""
     runServer ini = do
       hSetBuffering stdout LineBuffering
       hSetBuffering stderr LineBuffering
       fp <- checkSavedFingerprint cfgPath defaultX509Config
       let host = either (const "<hostnames>") T.unpack $ lookupValue "TRANSPORT" "host" ini
           port = T.unpack $ strictIni "TRANSPORT" "port" ini
-          cfg@ServerConfig {information, transports, storeLogFile, newQueueBasicAuth, messageExpiration, inactiveClientExpiration} = serverConfig
+          cfg@ServerConfig {information, storeLogFile, msgStoreType, newQueueBasicAuth, messageExpiration, inactiveClientExpiration} = serverConfig
           sourceCode' = (\ServerPublicInfo {sourceCode} -> sourceCode) <$> information
           srv = ProtoServerWithAuth (SMPServer [THDomainName host] (if port == "5223" then "" else port) (C.KeyHash fp)) newQueueBasicAuth
       printServiceInfo serverVersion srv
       printSourceCode sourceCode'
       printServerConfig transports storeLogFile
+      checkMsgStoreMode msgStoreType
       putStrLn $ case messageExpiration of
         Just ExpirationConfig {ttl} -> "expiring messages after " <> showTTL ttl
         _ -> "not expiring messages"
@@ -244,31 +357,54 @@ smpServerCLI_ generateSite serveStaticFiles cfgPath logPath =
                 newQueuesAllowed = allowNewQueues cfg,
                 basicAuthEnabled = isJust newQueueBasicAuth
               }
-      runWebServer ini ServerInformation {config, information}
-      runSMPServer cfg
+      case webStaticPath' of
+        Just path | sharedHTTP -> do
+          runWebServer path Nothing ServerInformation {config, information}
+          attachStaticFiles path $ \attachHTTP -> do
+            logDebug "Allocated web server resources"
+            runSMPServer cfg (Just attachHTTP) `finally` logDebug "Releasing web server resources..."
+        Just path -> do
+          runWebServer path webHttpsParams' ServerInformation {config, information}
+          runSMPServer cfg Nothing
+        Nothing -> do
+          logWarn "No server static path set"
+          runSMPServer cfg Nothing
+      logDebug "Bye"
       where
         enableStoreLog = settingIsOn "STORE_LOG" "enable" ini
         logStats = settingIsOn "STORE_LOG" "log_stats" ini
         c = combine cfgPath . ($ defaultX509Config)
+        restoreMessagesFile path = case iniOnOff "STORE_LOG" "restore_messages" ini of
+          Just True -> Just path
+          Just False -> Nothing
+          -- if the setting is not set, it is enabled when store log is enabled
+          _ -> enableStoreLog $> path
+        transports = iniTransports ini
+        sharedHTTP = any (\(_, _, addHTTP) -> addHTTP) transports
+        iniMsgStoreType = either error id $! readMsgStoreType ini
         serverConfig =
           ServerConfig
-            { transports = iniTransports ini,
+            { transports,
               smpHandshakeTimeout = 120000000,
-              tbqSize = 64,
-              msgQueueQuota = 128,
+              tbqSize = 128,
+              msgStoreType = iniMsgStoreType,
+              msgQueueQuota = defaultMsgQueueQuota,
+              maxJournalMsgCount = defaultMaxJournalMsgCount,
+              maxJournalStateLines = defaultMaxJournalStateLines,
               queueIdBytes = 24,
               msgIdBytes = 24, -- must be at least 24 bytes, it is used as 192-bit nonce for XSalsa20
-              caCertificateFile = c caCrtFile,
-              privateKeyFile = c serverKeyFile,
-              certificateFile = c serverCrtFile,
+              smpCredentials =
+                ServerCredentials
+                  { caCertificateFile = Just $ c caCrtFile,
+                    privateKeyFile = c serverKeyFile,
+                    certificateFile = c serverCrtFile
+                  },
+              httpCredentials = (\WebHttpsParams {key, cert} -> ServerCredentials {caCertificateFile = Nothing, privateKeyFile = key, certificateFile = cert}) <$> webHttpsParams',
               storeLogFile = enableStoreLog $> storeLogFilePath,
-              storeMsgsFile =
-                let messagesPath = combine logPath "smp-server-messages.log"
-                 in case iniOnOff "STORE_LOG" "restore_messages" ini of
-                      Just True -> Just messagesPath
-                      Just False -> Nothing
-                      -- if the setting is not set, it is enabled when store log is enabled
-                      _ -> enableStoreLog $> messagesPath,
+              storeMsgsFile = case iniMsgStoreType of
+                AMSType SMSMemory -> restoreMessagesFile storeMsgsFilePath
+                AMSType SMSJournal -> Just storeMsgsJournalDir,
+              storeNtfsFile = restoreMessagesFile storeNtfsFilePath,
               -- allow creating new queues by default
               allowNewQueues = fromMaybe True $ iniOnOff "AUTH" "new_queues" ini,
               newQueueBasicAuth = either error id <$!> strDecodeIni "AUTH" "create_password" ini,
@@ -279,6 +415,12 @@ smpServerCLI_ generateSite serveStaticFiles cfgPath logPath =
                   defaultMessageExpiration
                     { ttl = 86400 * readIniDefault defMsgExpirationDays "STORE_LOG" "expire_messages_days" ini
                     },
+              expireMessagesOnStart = fromMaybe True $ iniOnOff "STORE_LOG" "expire_messages_on_start" ini,
+              idleQueueInterval = defaultIdleQueueInterval,
+              notificationExpiration =
+                defaultNtfExpiration
+                  { ttl = 3600 * readIniDefault defNtfExpirationHours "STORE_LOG" "expire_ntfs_hours" ini
+                  },
               inactiveClientExpiration =
                 settingIsOn "INACTIVE_CLIENTS" "disconnect" ini
                   $> ExpirationConfig
@@ -290,11 +432,11 @@ smpServerCLI_ generateSite serveStaticFiles cfgPath logPath =
               serverStatsLogFile = combine logPath "smp-server-stats.daily.log",
               serverStatsBackupFile = logStats $> combine logPath "smp-server-stats.log",
               pendingENDInterval = 15000000, -- 15 seconds
+              ntfDeliveryInterval = 3000000, -- 3 seconds
               smpServerVRange = supportedServerSMPRelayVRange,
               transportConfig =
                 defaultTransportServerConfig
-                  { logTLSErrors = fromMaybe False $ iniOnOff "TRANSPORT" "log_tls_errors" ini,
-                    alpn = Just supportedSMPHandshakes
+                  { logTLSErrors = fromMaybe False $ iniOnOff "TRANSPORT" "log_tls_errors" ini
                   },
               controlPort = eitherToMaybe $ T.unpack <$> lookupValue "TRANSPORT" "control_port" ini,
               smpAgentCfg =
@@ -307,7 +449,7 @@ smpServerCLI_ generateSite serveStaticFiles cfgPath logPath =
                             defaultNetworkConfig
                               { socksProxy = either error id <$!> strDecodeIni "PROXY" "socks_proxy" ini,
                                 socksMode = maybe SMOnion (either error id) $! strDecodeIni "PROXY" "socks_mode" ini,
-                                hostMode = either (const HMPublic) textToHostMode $ lookupValue "PROXY" "host_mode" ini,
+                                hostMode = either (const HMPublic) (either error id . textToHostMode) $ lookupValue "PROXY" "host_mode" ini,
                                 requiredHostMode = fromMaybe False $ iniOnOff "PROXY" "required_host_mode" ini
                               }
                         },
@@ -318,33 +460,50 @@ smpServerCLI_ generateSite serveStaticFiles cfgPath logPath =
               serverClientConcurrency = readIniDefault defaultProxyClientConcurrency "PROXY" "client_concurrency" ini,
               information = serverPublicInfo ini
             }
-        textToHostMode :: Text -> HostMode
-        textToHostMode = \case
-          "public" -> HMPublic
-          "onion" -> HMOnionViaSocks
-          s -> error . T.unpack $ "Invalid host_mode: " <> s
         textToOwnServers :: Text -> [ByteString]
         textToOwnServers = map encodeUtf8 . T.words
-
-    runWebServer ini si =
-      case eitherToMaybe $ T.unpack <$> lookupValue "WEB" "static_path" ini of
-        Nothing -> logWarn "No server static path set"
-        Just webStaticPath -> do
+        runWebServer webStaticPath webHttpsParams si = do
           let onionHost =
                 either (const Nothing) (find isOnion) $
                   strDecode @(L.NonEmpty TransportHost) . encodeUtf8 =<< lookupValue "TRANSPORT" "host" ini
               webHttpPort = eitherToMaybe $ read . T.unpack <$> lookupValue "WEB" "http" ini
-              webHttpsParams =
-                eitherToMaybe $ do
-                  port <- read . T.unpack <$> lookupValue "WEB" "https" ini
-                  cert <- T.unpack <$> lookupValue "WEB" "cert" ini
-                  key <- T.unpack <$> lookupValue "WEB" "key" ini
-                  pure WebHttpsParams {port, cert, key}
           generateSite si onionHost webStaticPath
           when (isJust webHttpPort || isJust webHttpsParams) $
             serveStaticFiles EmbeddedWebParams {webStaticPath, webHttpPort, webHttpsParams}
           where
             isOnion = \case THOnionHost _ -> True; _ -> False
+        webHttpsParams' =
+          eitherToMaybe $ do
+            port <- read . T.unpack <$> lookupValue "WEB" "https" ini
+            cert <- T.unpack <$> lookupValue "WEB" "cert" ini
+            key <- T.unpack <$> lookupValue "WEB" "key" ini
+            pure WebHttpsParams {port, cert, key}
+        webStaticPath' = eitherToMaybe $ T.unpack <$> lookupValue "WEB" "static_path" ini
+
+    checkMsgStoreMode :: AMSType -> IO ()
+    checkMsgStoreMode mode = do
+      msgsDirExists <- doesDirectoryExist storeMsgsJournalDir
+      msgsFileExists <- doesFileExist storeMsgsFilePath
+      case mode of
+        _ | msgsFileExists && msgsDirExists -> exitConfigureMsgStorage
+        AMSType SMSJournal
+          | msgsFileExists -> do
+              putStrLn $ "Error: store_messages is `journal` with " <> storeMsgsFilePath <> " file present."
+              putStrLn "Set store_messages to `memory` or use `smp-server journal export` to migrate."
+              exitFailure
+          | not msgsDirExists ->
+              putStrLn $ "store_messages is `journal`, " <> storeMsgsJournalDir <> " directory will be created."
+        AMSType SMSMemory
+          | msgsDirExists -> do
+              putStrLn $ "Error: store_messages is `memory` with " <> storeMsgsJournalDir <> " directory present."
+              putStrLn "Set store_messages to `journal` or use `smp-server journal import` to migrate."
+              exitFailure
+        _ -> pure ()
+
+    exitConfigureMsgStorage = do
+      putStrLn $ "Error: both " <> storeMsgsFilePath <> " file and " <> storeMsgsJournalDir <> " directory are present."
+      putStrLn "Configure memory storage."
+      exitFailure
 
 data EmbeddedWebParams = EmbeddedWebParams
   { webStaticPath :: FilePath,
@@ -368,28 +527,31 @@ getServerSourceCode =
 simplexmqSource :: String
 simplexmqSource = "https://github.com/simplex-chat/simplexmq"
 
-informationIniContent :: Maybe Text -> Text
-informationIniContent sourceCode_ =
+defaultControlPort :: Int
+defaultControlPort = 5224
+
+informationIniContent :: InitOptions -> Text
+informationIniContent InitOptions {sourceCode, serverInfo} =
   "[INFORMATION]\n\
   \# AGPLv3 license requires that you make any source code modifications\n\
   \# available to the end users of the server.\n\
   \# LICENSE: https://github.com/simplex-chat/simplexmq/blob/stable/LICENSE\n\
   \# Include correct source code URI in case the server source code is modified in any way.\n\
   \# If any other information fields are present, source code property also MUST be present.\n\n"
-    <> (maybe "# source_code: URI" ("source_code: " <>) sourceCode_ <> "\n\n")
-    <> "# Declaring all below information is optional, any of these fields can be omitted.\n\
+    <> (optDisabled sourceCode <> "source_code: " <> fromMaybe "URI" sourceCode)
+    <> "\n\n\
+       \# Declaring all below information is optional, any of these fields can be omitted.\n\
        \\n\
        \# Server usage conditions and amendments.\n\
        \# It is recommended to use standard conditions with any amendments in a separate document.\n\
        \# usage_conditions: https://github.com/simplex-chat/simplex-chat/blob/stable/PRIVACY.md\n\
        \# condition_amendments: link\n\
        \\n\
-       \# Server location and operator.\n\
-       \# server_country: ISO-3166 2-letter code\n\
-       \# operator: entity (organization or person name)\n\
-       \# operator_country: ISO-3166 2-letter code\n\
-       \# website:\n\
-       \\n\
+       \# Server location and operator.\n"
+    <> countryStr "server" serverCountry
+    <> enitiyStrs "operator" operator
+    <> (optDisabled website <> "website: " <> fromMaybe "" website)
+    <> "\n\n\
        \# Administrative contacts.\n\
        \# admin_simplex: SimpleX address\n\
        \# admin_email:\n\
@@ -402,9 +564,21 @@ informationIniContent sourceCode_ =
        \# complaints_pgp:\n\
        \# complaints_pgp_fingerprint:\n\
        \\n\
-       \# Hosting provider.\n\
-       \# hosting: entity (organization or person name)\n\
-       \# hosting_country: ISO-3166 2-letter code\n\n"
+       \# Hosting provider.\n"
+    <> enitiyStrs "hosting" hosting
+    <> "\n\
+       \# Hosting type can be `virtual`, `dedicated`, `colocation`, `owned`\n"
+    <> ("hosting_type: " <> maybe "virtual" (decodeLatin1 . strEncode) hostingType <> "\n\n")
+  where
+    ServerPublicInfo {operator, website, hosting, hostingType, serverCountry} = serverInfo
+    countryStr optName country = optDisabled country <> optName <> "_country: " <> fromMaybe "ISO-3166 2-letter code" country <> "\n"
+    enitiyStrs optName entity =
+      optDisabled entity
+        <> optName
+        <> ": "
+        <> maybe "entity (organization or person name)" name entity
+        <> "\n"
+        <> countryStr optName (country =<< entity)
 
 serverPublicInfo :: Ini -> Maybe ServerPublicInfo
 serverPublicInfo ini = serverInfo <$!> infoValue "source_code"
@@ -420,15 +594,14 @@ serverPublicInfo ini = serverInfo <$!> infoValue "source_code"
           website = infoValue "website",
           adminContacts = iniContacts "admin_simplex" "admin_email" "admin_pgp" "admin_pgp_fingerprint",
           complaintsContacts = iniContacts "complaints_simplex" "complaints_email" "complaints_pgp" "complaints_pgp_fingerprint",
-          hosting = iniEntity "hosting" "hosting_country"
+          hosting = iniEntity "hosting" "hosting_country",
+          hostingType = either error id <$!> strDecodeIni "INFORMATION" "hosting_type" ini
         }
     infoValue name = eitherToMaybe $ lookupValue "INFORMATION" name ini
     iniEntity nameField countryField =
       (\name -> Entity {name, country = countryValue countryField})
         <$!> infoValue nameField
-    countryValue field =
-      (\cs -> if T.length cs == 2 && T.all (\c -> isAscii c && isAlpha c) cs then T.map toUpper cs else error $ "Use ISO3166 2-letter code for " <> T.unpack field)
-        <$!> infoValue field
+    countryValue field = (either error id . validCountryValue (T.unpack field) . T.unpack) <$!> infoValue field
     iniContacts simplexField emailField pgpKeyUriField pgpKeyFingerprintField =
       let simplex = either error id . parseAll (connReqUriP' Nothing) . encodeUtf8 <$!> eitherToMaybe (lookupValue "INFORMATION" simplexField ini)
           email = infoValue emailField
@@ -439,6 +612,14 @@ serverPublicInfo ini = serverInfo <$!> infoValue "source_code"
             (Nothing, Nothing, _, Nothing) -> Nothing
             (_, _, pkURI, pkFingerprint) -> Just ServerContactAddress {simplex, email, pgp = PGPKey <$> pkURI <*> pkFingerprint}
 
+optDisabled :: Maybe a -> Text
+optDisabled p = if isNothing p then "# " else ""
+
+validCountryValue :: String -> String -> Either String Text
+validCountryValue field s
+  | length s == 2 && all (\c -> isAscii c && isAlpha c) s = Right $ T.pack $ map toUpper s
+  | otherwise = Left $ "Use ISO3166 2-letter code for " <> field
+
 printSourceCode :: Maybe Text -> IO ()
 printSourceCode = \case
   Just sourceCode -> T.putStrLn $ "Server source code: " <> sourceCode
@@ -451,6 +632,9 @@ data CliCommand
   | OnlineCert CertOptions
   | Start
   | Delete
+  | Journal JournalCmd
+
+data JournalCmd = JCImport | JCExport | JCDelete
 
 data InitOptions = InitOptions
   { enableStoreLog :: Bool,
@@ -459,7 +643,13 @@ data InitOptions = InitOptions
     ip :: HostName,
     fqdn :: Maybe HostName,
     password :: Maybe ServerPassword,
+    controlPort :: Maybe Int,
+    socksProxy :: Maybe SocksProxy,
+    ownDomains :: Maybe (L.NonEmpty TransportHost),
     sourceCode :: Maybe Text,
+    serverInfo :: ServerPublicInfo,
+    operatorCountry :: Maybe Text,
+    hostingCountry :: Maybe Text,
     webStaticPath :: Maybe FilePath,
     disableWeb :: Bool,
     scripted :: Bool
@@ -476,6 +666,7 @@ cliCommandP cfgPath logPath iniFile =
         <> command "cert" (info (OnlineCert <$> certOptionsP) (progDesc $ "Generate new online TLS server credentials (configuration: " <> iniFile <> ")"))
         <> command "start" (info (pure Start) (progDesc $ "Start server (configuration: " <> iniFile <> ")"))
         <> command "delete" (info (pure Delete) (progDesc "Delete configuration and log files"))
+        <> command "journal" (info (Journal <$> journalCmdP) (progDesc "Import/export messages to/from journal storage"))
     )
   where
     initP :: Parser InitOptions
@@ -528,11 +719,45 @@ cliCommandP cfgPath logPath iniFile =
                   <> help "Set password to create new messaging queues"
                   <> value SPRandom
               )
+      controlPort <-
+        flag' (Just defaultControlPort) (long "control-port" <> help ("Enable control port on " <> show defaultControlPort))
+          <|> option strParse (long "control-port" <> help "Enable control port" <> metavar "PORT" <> value Nothing)
+      socksProxy <-
+        flag' (Just defaultSocksProxy) (long "socks-proxy" <> help "Outgoing SOCKS proxy on port 9050")
+          <|> option
+            strParse
+            ( long "socks-proxy"
+                <> metavar "PROXY"
+                <> help "Outgoing SOCKS proxy to forward messages to onion-only servers"
+                <> value Nothing
+            )
+      ownDomains :: Maybe (L.NonEmpty TransportHost) <-
+        option
+          strParse
+          ( long "own-domains"
+              <> metavar "DOMAINS"
+              <> help "Own server domain names (comma-separated)"
+              <> value Nothing
+          )
       sourceCode <-
+        flag' (Just simplexmqSource) (long "source-code" <> help ("Server source code (default: " <> simplexmqSource <> ")"))
+          <|> (optional . strOption) (long "source-code" <> metavar "URI" <> help "Server source code")
+      operator_ <- entityP "operator" "OPERATOR" "Server operator"
+      hosting_ <- entityP "hosting" "HOSTING" "Hosting provider"
+      hostingType <-
+        option
+          strParse
+          ( long "hosting-type"
+              <> metavar "HOSTING_TYPE"
+              <> help "Hosting type: virtual, dedicated, colocation, owned"
+              <> value Nothing
+          )
+      serverCountry <- countryP "server" "SERVER" "Server datacenter"
+      website <-
         (optional . strOption)
-          ( long "source-code"
-              <> help "Server source code will be communicated to the users"
-              <> metavar "URI"
+          ( long "operator-website"
+              <> help "Operator public website"
+              <> metavar "WEBSITE"
           )
       webStaticPath <-
         (optional . strOption)
@@ -559,10 +784,53 @@ cliCommandP cfgPath logPath iniFile =
             ip,
             fqdn,
             password,
-            sourceCode,
+            controlPort,
+            socksProxy,
+            ownDomains,
+            sourceCode = T.pack <$> sourceCode,
+            serverInfo =
+              ServerPublicInfo
+                { sourceCode = T.pack simplexmqSource,
+                  usageConditions = Nothing,
+                  operator = fst operator_,
+                  website,
+                  adminContacts = Nothing,
+                  complaintsContacts = Nothing,
+                  hosting = fst hosting_,
+                  hostingType,
+                  serverCountry
+                },
+            operatorCountry = snd operator_,
+            hostingCountry = snd hosting_,
             webStaticPath,
             disableWeb,
             scripted
           }
+    journalCmdP =
+      hsubparser
+        ( command "import" (info (pure JCImport) (progDesc "Import message log file into a new journal storage"))
+            <> command "export" (info (pure JCExport) (progDesc "Export journal storage to message log file"))
+            <> command "delete" (info (pure JCDelete) (progDesc "Delete journal storage"))
+        )
+
     parseBasicAuth :: ReadM ServerPassword
     parseBasicAuth = eitherReader $ fmap ServerPassword . strDecode . B.pack
+    entityP :: String -> String -> String -> Parser (Maybe Entity, Maybe Text)
+    entityP opt' metavar' help' = do
+      name_ <-
+        (optional . strOption)
+          ( long opt'
+              <> metavar (metavar' <> "_NAME")
+              <> help (help' <> " name")
+          )
+      country <- countryP opt' metavar' help'
+      pure ((\name -> Entity {name, country}) <$> name_, country)
+    countryP :: String -> String -> String -> Parser (Maybe Text)
+    countryP opt' metavar' help' =
+      (optional . option (eitherReader $ validCountryValue opt'))
+        ( long (opt' <> "-country")
+            <> metavar (metavar' <> "_COUNTRY")
+            <> help (help' <> " country")
+        )
+    strParse :: StrEncoding a => ReadM a
+    strParse = eitherReader $ parseAll strP . encodeUtf8 . T.pack

src/Simplex/Messaging/Server/MsgStore.hs

@@ -1,4 +1,3 @@
-{-# LANGUAGE LambdaCase #-}
 {-# LANGUAGE OverloadedStrings #-}
 
 module Simplex.Messaging.Server.MsgStore where

src/Simplex/Messaging/Server/MsgStore/Journal.hs

@@ -0,0 +1,800 @@
+{-# LANGUAGE BangPatterns #-}
+{-# LANGUAGE DataKinds #-}
+{-# LANGUAGE DerivingStrategies #-}
+{-# LANGUAGE DuplicateRecordFields #-}
+{-# LANGUAGE GADTs #-}
+{-# LANGUAGE GeneralizedNewtypeDeriving #-}
+{-# LANGUAGE InstanceSigs #-}
+{-# LANGUAGE LambdaCase #-}
+{-# LANGUAGE MultiWayIf #-}
+{-# LANGUAGE NamedFieldPuns #-}
+{-# LANGUAGE OverloadedStrings #-}
+{-# LANGUAGE ScopedTypeVariables #-}
+{-# LANGUAGE StandaloneDeriving #-}
+{-# LANGUAGE TupleSections #-}
+{-# LANGUAGE TypeFamilies #-}
+
+module Simplex.Messaging.Server.MsgStore.Journal
+  ( JournalMsgStore (queues, random),
+    JournalQueue,
+    JournalMsgQueue (queue, state),
+    JMQueue (queueDirectory, statePath),
+    JournalStoreConfig (..),
+    closeMsgQueue,
+    closeMsgQueueHandles,
+    -- below are exported for tests
+    MsgQueueState (..),
+    JournalState (..),
+    SJournalType (..),
+    msgQueueDirectory,
+    msgQueueStatePath,
+    readWriteQueueState,
+    newMsgQueueState,
+    newJournalId,
+    appendState,
+    queueLogFileName,
+    journalFilePath,
+    logFileExt,
+  )
+where
+
+import Control.Concurrent.STM
+import qualified Control.Exception as E
+import Control.Logger.Simple
+import Control.Monad
+import Control.Monad.Trans.Except
+import qualified Data.Attoparsec.ByteString.Char8 as A
+import Data.ByteString.Char8 (ByteString)
+import qualified Data.ByteString.Char8 as B
+import Data.Functor (($>))
+import Data.Int (Int64)
+import Data.List (intercalate)
+import Data.Maybe (catMaybes, fromMaybe, isNothing)
+import qualified Data.Text as T
+import Data.Time.Clock (getCurrentTime)
+import Data.Time.Clock.System (SystemTime (..), getSystemTime)
+import Data.Time.Format.ISO8601 (iso8601Show)
+import GHC.IO (catchAny)
+import Simplex.Messaging.Agent.Client (getMapLock, withLockMap)
+import Simplex.Messaging.Agent.Lock
+import Simplex.Messaging.Encoding.String
+import Simplex.Messaging.Protocol
+import Simplex.Messaging.Server.MsgStore.Types
+import Simplex.Messaging.Server.QueueStore
+import Simplex.Messaging.Server.QueueStore.STM
+import Simplex.Messaging.Server.StoreLog
+import Simplex.Messaging.TMap (TMap)
+import qualified Simplex.Messaging.TMap as TM
+import Simplex.Messaging.Util (ifM, tshow, ($>>=), (<$$>))
+import System.Directory
+import System.Exit
+import System.FilePath ((</>))
+import System.IO (BufferMode (..), Handle, IOMode (..), SeekMode (..), stdout)
+import qualified System.IO as IO
+import System.Random (StdGen, genByteString, newStdGen)
+
+data JournalMsgStore = JournalMsgStore
+  { config :: JournalStoreConfig,
+    random :: TVar StdGen,
+    queueLocks :: TMap RecipientId Lock,
+    queues :: TMap RecipientId (QueueReference JournalQueue),
+    queueCount :: TVar Int,
+    notifierCount :: TVar Int,
+    -- senders :: TMap SenderId RecipientId,
+    -- notifiers :: TMap NotifierId RecipientId,
+    storeLog :: TVar (Maybe (StoreLog 'WriteMode))
+  }
+
+data JournalStoreConfig = JournalStoreConfig
+  { storePath :: FilePath,
+    pathParts :: Int,
+    quota :: Int,
+    -- Max number of messages per journal file - ignored in STM store.
+    -- When this limit is reached, the file will be changed.
+    -- This number should be set bigger than queue quota.
+    maxMsgCount :: Int,
+    maxStateLines :: Int,
+    stateTailSize :: Int,
+    -- time in seconds after which the queue will be closed after message expiration
+    idleInterval :: Int64
+  }
+
+data JournalQueue = JournalQueue
+  { queueLock :: Lock,
+    -- To avoid race conditions and errors when restoring queues,
+    -- Nothing is written to TVar when queue is deleted.
+    queueRec :: TVar (Maybe QueueRec),
+    msgQueue_ :: TVar (Maybe JournalMsgQueue),
+    -- system time in seconds since epoch
+    activeAt :: TVar Int64,
+    -- Just True - empty, Just False - non-empty, Nothing - unknown
+    isEmpty :: TVar (Maybe Bool)
+  }
+
+data JMQueue = JMQueue
+  { queueDirectory :: FilePath,
+    statePath :: FilePath
+  }
+
+data JournalMsgQueue = JournalMsgQueue
+  { queue :: JMQueue,
+    state :: TVar MsgQueueState,
+    -- tipMsg contains last message and length incl. newline
+    -- Nothing - unknown, Just Nothing - empty queue.
+    -- It  prevents reading each message twice,
+    -- and reading it after it was just written.
+    tipMsg :: TVar (Maybe (Maybe (Message, Int64))),
+    handles :: TVar (Maybe MsgQueueHandles)
+  }
+
+data MsgQueueState = MsgQueueState
+  { readState :: JournalState 'JTRead,
+    writeState :: JournalState 'JTWrite,
+    canWrite :: Bool,
+    size :: Int
+  }
+  deriving (Show)
+
+data MsgQueueHandles = MsgQueueHandles
+  { stateHandle :: Handle, -- handle to queue state log file, rotates and removes old backups when server is restarted
+    readHandle :: Handle,
+    writeHandle :: Maybe Handle -- optional, used when write file is different from read file
+  }
+
+data JournalState t = JournalState
+  { journalType :: SJournalType t,
+    journalId :: ByteString,
+    msgPos :: Int,
+    msgCount :: Int,
+    bytePos :: Int64,
+    byteCount :: Int64
+  }
+  deriving (Show)
+
+data JournalType = JTRead | JTWrite
+
+data SJournalType (t :: JournalType) where
+  SJTRead :: SJournalType 'JTRead
+  SJTWrite :: SJournalType 'JTWrite
+
+class JournalTypeI t where sJournalType :: SJournalType t
+
+instance JournalTypeI 'JTRead where sJournalType = SJTRead
+
+instance JournalTypeI 'JTWrite where sJournalType = SJTWrite
+
+deriving instance Show (SJournalType t)
+
+newMsgQueueState :: ByteString -> MsgQueueState
+newMsgQueueState journalId =
+  MsgQueueState
+    { writeState = newJournalState journalId,
+      readState = newJournalState journalId,
+      canWrite = True,
+      size = 0
+    }
+
+newJournalState :: JournalTypeI t => ByteString -> JournalState t
+newJournalState journalId = JournalState sJournalType journalId 0 0 0 0
+
+journalFilePath :: FilePath -> ByteString -> FilePath
+journalFilePath dir journalId = dir </> (msgLogFileName <> "." <> B.unpack journalId <> logFileExt)
+
+instance StrEncoding MsgQueueState where
+  strEncode MsgQueueState {writeState, readState, canWrite, size} =
+    B.unwords
+      [ "write=" <> strEncode writeState,
+        "read=" <> strEncode readState,
+        "canWrite=" <> strEncode canWrite,
+        "size=" <> strEncode size
+      ]
+  strP = do
+    writeState <- "write=" *> strP
+    readState <- " read=" *> strP
+    canWrite <- " canWrite=" *> strP
+    size <- " size=" *> strP
+    pure MsgQueueState {writeState, readState, canWrite, size}
+
+instance JournalTypeI t => StrEncoding (JournalState t) where
+  strEncode JournalState {journalId, msgPos, msgCount, bytePos, byteCount} =
+    B.intercalate "," [journalId, e msgPos, e msgCount, e bytePos, e byteCount]
+    where
+      e :: StrEncoding a => a -> ByteString
+      e = strEncode
+  strP = do
+    journalId <- A.takeTill (== ',')
+    JournalState sJournalType journalId <$> i <*> i <*> i <*> i
+    where
+      i :: Integral a => A.Parser a
+      i = A.char ',' *> A.decimal
+
+queueLogFileName :: String
+queueLogFileName = "queue_state"
+
+msgLogFileName :: String
+msgLogFileName = "messages"
+
+logFileExt :: String
+logFileExt = ".log"
+
+newtype StoreIO a = StoreIO {unStoreIO :: IO a}
+  deriving newtype (Functor, Applicative, Monad)
+
+instance STMQueueStore JournalMsgStore where
+  queues' = queues
+  queueCount' = queueCount
+  notifierCount' = notifierCount
+
+  -- senders' = senders
+  -- notifiers' = notifiers
+  storeLog' = storeLog
+  mkQueue st qr = do
+    lock <- getMapLock (queueLocks st) $ recipientId qr
+    q <- newTVar $ Just qr
+    mq <- newTVar Nothing
+    activeAt <- newTVar 0
+    isEmpty <- newTVar Nothing
+    pure $ JournalQueue lock q mq activeAt isEmpty
+  msgQueue_' = msgQueue_
+
+instance MsgStoreClass JournalMsgStore where
+  type StoreMonad JournalMsgStore = StoreIO
+  type StoreQueue JournalMsgStore = JournalQueue
+  type MsgQueue JournalMsgStore = JournalMsgQueue
+  type MsgStoreConfig JournalMsgStore = JournalStoreConfig
+
+  newMsgStore :: JournalStoreConfig -> IO JournalMsgStore
+  newMsgStore config = do
+    random <- newTVarIO =<< newStdGen
+    queueLocks <- TM.emptyIO
+    queues <- TM.emptyIO
+    queueCount <- newTVarIO 0
+    notifierCount <- newTVarIO 0
+    -- senders <- TM.emptyIO
+    -- notifiers <- TM.emptyIO
+    storeLog <- newTVarIO Nothing
+    pure JournalMsgStore {config, random, queueLocks, queues, queueCount, notifierCount, storeLog}
+
+  setStoreLog :: JournalMsgStore -> StoreLog 'WriteMode -> IO ()
+  setStoreLog st sl = atomically $ writeTVar (storeLog st) (Just sl)
+
+  closeMsgStore st = do
+    readTVarIO (storeLog st) >>= mapM_ closeStoreLog
+    readTVarIO (queues st) >>= mapM_ (\case QRRecipient q -> closeMsgQueue q; _ -> pure ())
+
+  activeMsgQueues = queues
+  {-# INLINE activeMsgQueues #-}
+
+  -- This function is a "foldr" that opens and closes all queues, processes them as defined by action and accumulates the result.
+  -- It is used to export storage to a single file and also to expire messages and validate all queues when server is started.
+  -- TODO this function requires case-sensitive file system, because it uses queue directory as recipient ID.
+  -- It can be made to support case-insensite FS by supporting more than one queue per directory, by getting recipient ID from state file name.
+  withAllMsgQueues :: forall a. Monoid a => Bool -> JournalMsgStore -> (RecipientId -> JournalQueue -> IO a) -> IO a
+  withAllMsgQueues tty ms@JournalMsgStore {config} action = ifM (doesDirectoryExist storePath) processStore (pure mempty)
+    where
+      processStore = do
+        (!count, !res) <- foldQueues 0 processQueue (0, mempty) ("", storePath)
+        putStrLn $ progress count
+        pure res
+      JournalStoreConfig {storePath, pathParts} = config
+      processQueue :: (Int, a) -> (String, FilePath) -> IO (Int, a)
+      processQueue (!i, !r) (queueId, dir) = do
+        when (tty && i `mod` 100 == 0) $ putStr (progress i <> "\r") >> IO.hFlush stdout
+        r' <- case strDecode $ B.pack queueId of
+          Right rId ->
+            getQueue ms SRecipient rId >>= \case
+              Right q -> unStoreIO (getMsgQueue ms rId q) *> action rId q <* closeMsgQueue q
+              Left AUTH -> do
+                logWarn $ "STORE: processQueue, queue " <> T.pack queueId <> " was removed, removing " <> T.pack dir
+                removeQueueDirectory_ dir
+                pure mempty
+              Left e -> do
+                logError $ "STORE: processQueue, error getting queue " <> T.pack queueId <> ", " <> tshow e
+                exitFailure
+          Left e -> do
+            logError $ "STORE: processQueue, message queue directory " <> T.pack dir <> " is invalid, " <> tshow e
+            exitFailure
+        pure (i + 1, r <> r')
+      progress i = "Processed: " <> show i <> " queues"
+      foldQueues depth f acc (queueId, path) = do
+        let f' = if depth == pathParts - 1 then f else foldQueues (depth + 1) f
+        listDirs >>= foldM f' acc
+        where
+          listDirs = fmap catMaybes . mapM queuePath =<< listDirectory path
+          queuePath dir = do
+            let !path' = path </> dir
+                !queueId' = queueId <> dir
+            ifM
+              (doesDirectoryExist path')
+              (pure $ Just (queueId', path'))
+              (Nothing <$ putStrLn ("Error: path " <> path' <> " is not a directory, skipping"))
+
+  logQueueStates :: JournalMsgStore -> IO ()
+  logQueueStates ms = withActiveMsgQueues ms $ \_ -> unStoreIO . logQueueState
+
+  logQueueState :: JournalQueue -> StoreIO ()
+  logQueueState q =
+    StoreIO . void $
+      readTVarIO (msgQueue_ q)
+        $>>= \mq ->
+          readTVarIO (handles mq)
+            $>>= (\hs -> (readTVarIO (state mq) >>= appendState (stateHandle hs)) $> Just ())
+
+  queueRec' = queueRec
+  {-# INLINE queueRec' #-}
+
+  getMsgQueue :: JournalMsgStore -> RecipientId -> JournalQueue -> StoreIO JournalMsgQueue
+  getMsgQueue ms@JournalMsgStore {random} rId JournalQueue {msgQueue_} =
+    StoreIO $ readTVarIO msgQueue_ >>= maybe newQ pure
+    where
+      newQ = do
+        let dir = msgQueueDirectory ms rId
+            statePath = msgQueueStatePath dir $ B.unpack (strEncode rId)
+            queue = JMQueue {queueDirectory = dir, statePath}
+        q <- ifM (doesDirectoryExist dir) (openMsgQueue ms queue) (createQ queue)
+        atomically $ writeTVar msgQueue_ $ Just q
+        pure q
+        where
+          createQ :: JMQueue -> IO JournalMsgQueue
+          createQ queue = do
+            -- folder and files are not created here,
+            -- to avoid file IO for queues without messages during subscription
+            journalId <- newJournalId random
+            mkJournalQueue queue (newMsgQueueState journalId) Nothing
+
+  getPeekMsgQueue :: JournalMsgStore -> RecipientId -> JournalQueue -> StoreIO (Maybe (JournalMsgQueue, Message))
+  getPeekMsgQueue ms rId q@JournalQueue {isEmpty} =
+    StoreIO (readTVarIO isEmpty) >>= \case
+      Just True -> pure Nothing
+      Just False -> peek
+      Nothing -> do
+        -- We only close the queue if we just learnt it's empty.
+        -- This is needed to reduce file descriptors and memory usage
+        -- after the server just started and many clients subscribe.
+        -- In case the queue became non-empty on write and then again empty on read
+        -- we won't be closing it, to avoid frequent open/close on active queues.
+        r <- peek
+        when (isNothing r) $ StoreIO $ closeMsgQueue q
+        pure r
+    where
+      peek = do
+        mq <- getMsgQueue ms rId q
+        (mq,) <$$> tryPeekMsg_ q mq
+
+  -- only runs action if queue is not empty
+  withIdleMsgQueue :: Int64 -> JournalMsgStore -> RecipientId -> JournalQueue -> (JournalMsgQueue -> StoreIO a) -> StoreIO (Maybe a, Int)
+  withIdleMsgQueue now ms@JournalMsgStore {config} rId q action =
+    StoreIO $
+      readTVarIO (msgQueue_ q) >>= \case
+        Nothing ->
+          E.bracket
+            (unStoreIO $ getPeekMsgQueue ms rId q)
+            (mapM_ $ \_ -> closeMsgQueue q)
+            (maybe (pure (Nothing, 0)) (unStoreIO . run))
+          where
+            run (mq, _) = do
+              r <- action mq
+              sz <- getQueueSize_ mq
+              pure (Just r, sz)
+        Just mq -> do
+          ts <- readTVarIO $ activeAt q
+          r <-
+            if now - ts >= idleInterval config
+              then Just <$> unStoreIO (action mq) `E.finally` closeMsgQueue q
+              else pure Nothing
+          sz <- unStoreIO $ getQueueSize_ mq
+          pure (r, sz)
+
+  deleteQueue :: JournalMsgStore -> RecipientId -> JournalQueue -> IO (Either ErrorType QueueRec)
+  deleteQueue ms rId q =
+    fst <$$> deleteQueue_ ms rId q
+
+  deleteQueueSize :: JournalMsgStore -> RecipientId -> JournalQueue -> IO (Either ErrorType (QueueRec, Int))
+  deleteQueueSize ms rId q =
+    deleteQueue_ ms rId q >>= mapM (traverse getSize)
+    where
+      -- traverse operates on the second tuple element
+
+      getSize = maybe (pure (-1)) (fmap size . readTVarIO . state)
+
+  getQueueMessages_ :: Bool -> JournalMsgQueue -> StoreIO [Message]
+  getQueueMessages_ drainMsgs q = StoreIO (run [])
+    where
+      run msgs = readTVarIO (handles q) >>= maybe (pure []) (getMsg msgs)
+      getMsg msgs hs = chooseReadJournal q drainMsgs hs >>= maybe (pure msgs) readMsg
+        where
+          readMsg (rs, h) = do
+            (msg, len) <- hGetMsgAt h $ bytePos rs
+            updateReadPos q drainMsgs len hs
+            (msg :) <$> run msgs
+
+  writeMsg :: JournalMsgStore -> RecipientId -> JournalQueue -> Bool -> Message -> ExceptT ErrorType IO (Maybe (Message, Bool))
+  writeMsg ms rId q' logState msg = isolateQueue rId q' "writeMsg" $ do
+    q <- getMsgQueue ms rId q'
+    StoreIO $ (`E.finally` updateActiveAt q') $ do
+      st@MsgQueueState {canWrite, size} <- readTVarIO (state q)
+      let empty = size == 0
+      if canWrite || empty
+        then do
+          atomically $ writeTVar (isEmpty q') (Just False)
+          let canWrt' = quota > size
+          if canWrt'
+            then writeToJournal q st canWrt' msg $> Just (msg, empty)
+            else writeToJournal q st canWrt' msgQuota $> Nothing
+        else pure Nothing
+    where
+      JournalStoreConfig {quota, maxMsgCount} = config ms
+      msgQuota = MessageQuota {msgId = messageId msg, msgTs = messageTs msg}
+      writeToJournal q st@MsgQueueState {writeState, readState = rs, size} canWrt' !msg' = do
+        let msgStr = strEncode msg' `B.snoc` '\n'
+            msgLen = fromIntegral $ B.length msgStr
+        hs <- maybe createQueueDir pure =<< readTVarIO handles
+        (ws, wh) <- case writeHandle hs of
+          Nothing | msgCount writeState >= maxMsgCount -> switchWriteJournal hs
+          wh_ -> pure (writeState, fromMaybe (readHandle hs) wh_)
+        let msgPos' = msgPos ws + 1
+            bytePos' = bytePos ws + msgLen
+            ws' = ws {msgPos = msgPos', msgCount = msgPos', bytePos = bytePos', byteCount = bytePos'}
+            rs' = if journalId ws == journalId rs then rs {msgCount = msgPos', byteCount = bytePos'} else rs
+            !st' = st {writeState = ws', readState = rs', canWrite = canWrt', size = size + 1}
+        hAppend wh (bytePos ws) msgStr
+        updateQueueState q logState hs st' $
+          when (size == 0) $
+            writeTVar (tipMsg q) $
+              Just (Just (msg, msgLen))
+        where
+          JournalMsgQueue {queue = JMQueue {queueDirectory, statePath}, handles} = q
+          createQueueDir = do
+            createDirectoryIfMissing True queueDirectory
+            sh <- openFile statePath AppendMode
+            B.hPutStr sh ""
+            rh <- createNewJournal queueDirectory $ journalId rs
+            let hs = MsgQueueHandles {stateHandle = sh, readHandle = rh, writeHandle = Nothing}
+            atomically $ writeTVar handles $ Just hs
+            pure hs
+          switchWriteJournal hs = do
+            journalId <- newJournalId $ random ms
+            wh <- createNewJournal queueDirectory journalId
+            atomically $ writeTVar handles $ Just $ hs {writeHandle = Just wh}
+            pure (newJournalState journalId, wh)
+
+  -- can ONLY be used while restoring messages, not while server running
+  setOverQuota_ :: JournalQueue -> IO ()
+  setOverQuota_ q =
+    readTVarIO (msgQueue_ q)
+      >>= mapM_ (\JournalMsgQueue {state} -> atomically $ modifyTVar' state $ \st -> st {canWrite = False})
+
+  getQueueSize_ :: JournalMsgQueue -> StoreIO Int
+  getQueueSize_ JournalMsgQueue {state} = StoreIO $ size <$> readTVarIO state
+
+  tryPeekMsg_ :: JournalQueue -> JournalMsgQueue -> StoreIO (Maybe Message)
+  tryPeekMsg_ q mq@JournalMsgQueue {tipMsg, handles} =
+    StoreIO $ (readTVarIO handles $>>= chooseReadJournal mq True $>>= peekMsg) >>= setEmpty
+    where
+      peekMsg (rs, h) = readTVarIO tipMsg >>= maybe readMsg (pure . fmap fst)
+        where
+          readMsg = do
+            ml@(msg, _) <- hGetMsgAt h $ bytePos rs
+            atomically $ writeTVar tipMsg $ Just (Just ml)
+            pure $ Just msg
+      setEmpty msg = do
+        atomically $ writeTVar (isEmpty q) (Just $ isNothing msg)
+        pure msg
+
+  tryDeleteMsg_ :: JournalQueue -> JournalMsgQueue -> Bool -> StoreIO ()
+  tryDeleteMsg_ q mq@JournalMsgQueue {tipMsg, handles} logState =
+    StoreIO $
+      (`E.finally` when logState (updateActiveAt q)) $
+        void $
+          readTVarIO tipMsg -- if there is no cached tipMsg, do nothing
+            $>>= (pure . fmap snd)
+            $>>= \len ->
+              readTVarIO handles
+                $>>= \hs -> updateReadPos mq logState len hs $> Just ()
+
+  isolateQueue :: RecipientId -> JournalQueue -> String -> StoreIO a -> ExceptT ErrorType IO a
+  isolateQueue rId JournalQueue {queueLock} op =
+    tryStore' op rId . withLock' queueLock op . unStoreIO
+
+updateActiveAt :: JournalQueue -> IO ()
+updateActiveAt q = atomically . writeTVar (activeAt q) . systemSeconds =<< getSystemTime
+
+tryStore' :: String -> RecipientId -> IO a -> ExceptT ErrorType IO a
+tryStore' op rId = tryStore op rId . fmap Right
+
+tryStore :: forall a. String -> RecipientId -> IO (Either ErrorType a) -> ExceptT ErrorType IO a
+tryStore op rId a = ExceptT $ E.mask_ $ E.try a >>= either storeErr pure
+  where
+    storeErr :: E.SomeException -> IO (Either ErrorType a)
+    storeErr e =
+      let e' = intercalate ", " [op, B.unpack $ strEncode rId, show e]
+       in logError ("STORE: " <> T.pack e') $> Left (STORE e')
+
+isolateQueueId :: String -> JournalMsgStore -> RecipientId -> IO (Either ErrorType a) -> ExceptT ErrorType IO a
+isolateQueueId op ms rId = tryStore op rId . withLockMap (queueLocks ms) rId op
+
+openMsgQueue :: JournalMsgStore -> JMQueue -> IO JournalMsgQueue
+openMsgQueue ms q@JMQueue {queueDirectory = dir, statePath} = do
+  (st, sh) <- readWriteQueueState ms statePath
+  (st', rh, wh_) <- closeOnException sh $ openJournals ms dir st sh
+  let hs = MsgQueueHandles {stateHandle = sh, readHandle = rh, writeHandle = wh_}
+  mkJournalQueue q st' (Just hs)
+
+mkJournalQueue :: JMQueue -> MsgQueueState -> Maybe MsgQueueHandles -> IO JournalMsgQueue
+mkJournalQueue queue st hs_ = do
+  state <- newTVarIO st
+  tipMsg <- newTVarIO Nothing
+  handles <- newTVarIO hs_
+  -- using the same queue lock which is currently locked,
+  -- to avoid map lookup on queue operations
+  pure JournalMsgQueue {queue, state, tipMsg, handles}
+
+chooseReadJournal :: JournalMsgQueue -> Bool -> MsgQueueHandles -> IO (Maybe (JournalState 'JTRead, Handle))
+chooseReadJournal q log' hs = do
+  st@MsgQueueState {writeState = ws, readState = rs} <- readTVarIO (state q)
+  case writeHandle hs of
+    Just wh | msgPos rs >= msgCount rs && journalId rs /= journalId ws -> do
+      -- switching to write journal
+      atomically $ writeTVar (handles q) $ Just hs {readHandle = wh, writeHandle = Nothing}
+      hClose $ readHandle hs
+      when log' $ removeJournal (queueDirectory $ queue q) rs
+      let !rs' = (newJournalState $ journalId ws) {msgCount = msgCount ws, byteCount = byteCount ws}
+          !st' = st {readState = rs'}
+      updateQueueState q log' hs st' $ pure ()
+      pure $ Just (rs', wh)
+    _ | msgPos rs >= msgCount rs && journalId rs == journalId ws -> pure Nothing
+    _ -> pure $ Just (rs, readHandle hs)
+
+updateQueueState :: JournalMsgQueue -> Bool -> MsgQueueHandles -> MsgQueueState -> STM () -> IO ()
+updateQueueState q log' hs st a = do
+  unless (validQueueState st) $ E.throwIO $ userError $ "updateQueueState invalid state: " <> show st
+  when log' $ appendState (stateHandle hs) st
+  atomically $ writeTVar (state q) st >> a
+
+appendState :: Handle -> MsgQueueState -> IO ()
+appendState h st = E.uninterruptibleMask_ $ B.hPutStr h $ strEncode st `B.snoc` '\n'
+
+updateReadPos :: JournalMsgQueue -> Bool -> Int64 -> MsgQueueHandles -> IO ()
+updateReadPos q log' len hs = do
+  st@MsgQueueState {readState = rs, size} <- readTVarIO (state q)
+  let JournalState {msgPos, bytePos} = rs
+  let msgPos' = msgPos + 1
+      rs' = rs {msgPos = msgPos', bytePos = bytePos + len}
+      st' = st {readState = rs', size = size - 1}
+  updateQueueState q log' hs st' $ writeTVar (tipMsg q) Nothing
+
+msgQueueDirectory :: JournalMsgStore -> RecipientId -> FilePath
+msgQueueDirectory JournalMsgStore {config = JournalStoreConfig {storePath, pathParts}} rId =
+  storePath </> B.unpack (B.intercalate "/" $ splitSegments pathParts $ strEncode rId)
+  where
+    splitSegments _ "" = []
+    splitSegments 1 s = [s]
+    splitSegments n s =
+      let (seg, s') = B.splitAt 2 s
+       in seg : splitSegments (n - 1) s'
+
+msgQueueStatePath :: FilePath -> String -> FilePath
+msgQueueStatePath dir queueId = dir </> (queueLogFileName <> "." <> queueId <> logFileExt)
+
+createNewJournal :: FilePath -> ByteString -> IO Handle
+createNewJournal dir journalId = do
+  let path = journalFilePath dir journalId -- TODO retry if file exists
+  h <- openFile path ReadWriteMode
+  B.hPutStr h ""
+  pure h
+
+newJournalId :: TVar StdGen -> IO ByteString
+newJournalId g = strEncode <$> atomically (stateTVar g $ genByteString 12)
+
+openJournals :: JournalMsgStore -> FilePath -> MsgQueueState -> Handle -> IO (MsgQueueState, Handle, Maybe Handle)
+openJournals ms dir st@MsgQueueState {readState = rs, writeState = ws} sh = do
+  let rjId = journalId rs
+      wjId = journalId ws
+  openJournal rs >>= \case
+    Left path
+      | rjId == wjId -> do
+          logError $ "STORE: openJournals, no read/write file - creating new file, " <> T.pack path
+          newReadJournal
+      | otherwise -> do
+          let rs' = (newJournalState wjId) {msgCount = msgCount ws, byteCount = byteCount ws}
+              st' = st {readState = rs', size = msgCount ws}
+          openJournal rs' >>= \case
+            Left path' -> do
+              logError $ "STORE: openJournals, no read and write files - creating new file, read: " <> T.pack path <> ", write: " <> T.pack path'
+              newReadJournal
+            Right rh -> do
+              logError $ "STORE: openJournals, no read file - switched to write file, " <> T.pack path
+              closeOnException rh $ fixFileSize rh $ bytePos ws
+              pure (st', rh, Nothing)
+    Right rh
+      | rjId == wjId -> do
+          closeOnException rh $ fixFileSize rh $ bytePos ws
+          pure (st, rh, Nothing)
+      | otherwise -> closeOnException rh $ do
+          fixFileSize rh $ byteCount rs
+          openJournal ws >>= \case
+            Left path -> do
+              let msgs = msgCount rs
+                  bytes = byteCount rs
+                  size' = msgs - msgPos rs
+                  ws' = (newJournalState rjId) {msgPos = msgs, msgCount = msgs, bytePos = bytes, byteCount = bytes}
+                  st' = st {writeState = ws', size = size'} -- we don't amend canWrite to trigger QCONT
+              logError $ "STORE: openJournals, no write file, " <> T.pack path
+              pure (st', rh, Nothing)
+            Right wh -> do
+              closeOnException wh $ fixFileSize wh $ bytePos ws
+              pure (st, rh, Just wh)
+  where
+    newReadJournal = do
+      rjId' <- newJournalId $ random ms
+      rh <- createNewJournal dir rjId'
+      let st' = newMsgQueueState rjId'
+      closeOnException rh $ appendState sh st'
+      pure (st', rh, Nothing)
+    openJournal :: JournalState t -> IO (Either FilePath Handle)
+    openJournal JournalState {journalId} =
+      let path = journalFilePath dir journalId
+       in ifM (doesFileExist path) (Right <$> openFile path ReadWriteMode) (pure $ Left path)
+
+-- do that for all append operations
+
+fixFileSize :: Handle -> Int64 -> IO ()
+fixFileSize h pos = do
+  let pos' = fromIntegral pos
+  size <- IO.hFileSize h
+  if
+    | size > pos' -> do
+        name <- IO.hShow h
+        logWarn $ "STORE: fixFileSize, size " <> tshow size <> " > pos " <> tshow pos <> " - truncating, " <> T.pack name
+        IO.hSetFileSize h pos'
+    | size < pos' -> do
+        -- From code logic this can't happen.
+        name <- IO.hShow h
+        E.throwIO $ userError $ "fixFileSize size " <> show size <> " < pos " <> show pos <> " - aborting: " <> name
+    | otherwise -> pure ()
+
+removeJournal :: FilePath -> JournalState t -> IO ()
+removeJournal dir JournalState {journalId} = do
+  let path = journalFilePath dir journalId
+  removeFile path `catchAny` (\e -> logError $ "STORE: removeJournal, " <> T.pack path <> ", " <> tshow e)
+
+-- This function is supposed to be resilient to crashes while updating state files,
+-- and also resilient to crashes during its execution.
+readWriteQueueState :: JournalMsgStore -> FilePath -> IO (MsgQueueState, Handle)
+readWriteQueueState JournalMsgStore {random, config} statePath =
+  ifM
+    (doesFileExist tempBackup)
+    (renameFile tempBackup statePath >> readQueueState)
+    (ifM (doesFileExist statePath) readQueueState writeNewQueueState)
+  where
+    tempBackup = statePath <> ".bak"
+    readQueueState = do
+      ls <- B.lines <$> readFileTail
+      case ls of
+        [] -> writeNewQueueState
+        _ -> do
+          r@(st, _) <- useLastLine (length ls) True ls
+          unless (validQueueState st) $ E.throwIO $ userError $ "readWriteQueueState inconsistent state: " <> show st
+          pure r
+    writeNewQueueState = do
+      logWarn $ "STORE: readWriteQueueState, empty queue state - initialized, " <> T.pack statePath
+      st <- newMsgQueueState <$> newJournalId random
+      writeQueueState st
+    useLastLine len isLastLine ls = case strDecode $ last ls of
+      Right st
+        | len > maxStateLines config || not isLastLine ->
+            backupWriteQueueState st
+        | otherwise -> do
+            -- when state file has fewer than maxStateLines, we don't compact it
+            sh <- openFile statePath AppendMode
+            pure (st, sh)
+      Left e -- if the last line failed to parse
+        | isLastLine -> case init ls of -- or use the previous line
+            [] -> do
+              logWarn $ "STORE: readWriteQueueState, invalid 1-line queue state - initialized, " <> T.pack statePath
+              st <- newMsgQueueState <$> newJournalId random
+              backupWriteQueueState st
+            ls' -> do
+              logWarn $ "STORE: readWriteQueueState, invalid last line in queue state - using the previous line, " <> T.pack statePath
+              useLastLine len False ls'
+        | otherwise -> E.throwIO $ userError $ "readWriteQueueState invalid state " <> statePath <> ": " <> show e
+    backupWriteQueueState st = do
+      -- State backup is made in two steps to mitigate the crash during the backup.
+      -- Temporary backup file will be used when it is present.
+      renameFile statePath tempBackup -- 1) temp backup
+      r <- writeQueueState st -- 2) save state
+      ts <- getCurrentTime
+      renameFile tempBackup (statePath <> "." <> iso8601Show ts <> ".bak") -- 3) timed backup
+      pure r
+    writeQueueState st = do
+      sh <- openFile statePath AppendMode
+      closeOnException sh $ appendState sh st
+      pure (st, sh)
+    readFileTail =
+      IO.withFile statePath ReadMode $ \h -> do
+        size <- IO.hFileSize h
+        let sz = stateTailSize config
+            sz' = fromIntegral sz
+        if size > sz'
+          then IO.hSeek h AbsoluteSeek (size - sz') >> B.hGet h sz
+          else B.hGet h (fromIntegral size)
+
+validQueueState :: MsgQueueState -> Bool
+validQueueState MsgQueueState {readState = rs, writeState = ws, size}
+  | journalId rs == journalId ws =
+      alwaysValid
+        && msgPos rs <= msgPos ws
+        && msgCount rs == msgCount ws
+        && bytePos rs <= bytePos ws
+        && byteCount rs == byteCount ws
+        && size == msgCount rs - msgPos rs
+  | otherwise =
+      alwaysValid
+        && size == msgCount ws + msgCount rs - msgPos rs
+  where
+    alwaysValid =
+      msgPos rs <= msgCount rs
+        && bytePos rs <= byteCount rs
+        && msgPos ws == msgCount ws
+        && bytePos ws == byteCount ws
+
+deleteQueue_ :: JournalMsgStore -> RecipientId -> JournalQueue -> IO (Either ErrorType (QueueRec, Maybe JournalMsgQueue))
+deleteQueue_ ms rId q =
+  runExceptT $
+    isolateQueueId "deleteQueue_" ms rId $
+      deleteQueue' ms rId q >>= mapM remove
+  where
+    remove r@(_, mq_) = do
+      mapM_ closeMsgQueueHandles mq_
+      removeQueueDirectory ms rId
+      pure r
+
+closeMsgQueue :: JournalQueue -> IO ()
+closeMsgQueue JournalQueue {msgQueue_} = atomically (swapTVar msgQueue_ Nothing) >>= mapM_ closeMsgQueueHandles
+
+closeMsgQueueHandles :: JournalMsgQueue -> IO ()
+closeMsgQueueHandles q = readTVarIO (handles q) >>= mapM_ closeHandles
+  where
+    closeHandles (MsgQueueHandles sh rh wh_) = do
+      hClose sh
+      hClose rh
+      mapM_ hClose wh_
+
+removeQueueDirectory :: JournalMsgStore -> RecipientId -> IO ()
+removeQueueDirectory st = removeQueueDirectory_ . msgQueueDirectory st
+
+removeQueueDirectory_ :: FilePath -> IO ()
+removeQueueDirectory_ dir =
+  removePathForcibly dir `catchAny` \e ->
+    logError $ "STORE: removeQueueDirectory, " <> T.pack dir <> ", " <> tshow e
+
+hAppend :: Handle -> Int64 -> ByteString -> IO ()
+hAppend h pos s = do
+  fixFileSize h pos
+  IO.hSeek h SeekFromEnd 0
+  B.hPutStr h s
+
+hGetMsgAt :: Handle -> Int64 -> IO (Message, Int64)
+hGetMsgAt h pos = do
+  IO.hSeek h AbsoluteSeek $ fromIntegral pos
+  s <- B.hGetLine h
+  case strDecode s of
+    Right !msg ->
+      let !len = fromIntegral (B.length s) + 1
+       in pure (msg, len)
+    Left e -> E.throwIO $ userError $ "hGetMsgAt invalid message: " <> e
+
+openFile :: FilePath -> IOMode -> IO Handle
+openFile f mode = do
+  h <- IO.openFile f mode
+  IO.hSetBuffering h LineBuffering
+  pure h
+
+hClose :: Handle -> IO ()
+hClose h =
+  IO.hClose h `catchAny` \e -> do
+    name <- IO.hShow h
+    logError $ "STORE: hClose, " <> T.pack name <> ", " <> tshow e
+
+closeOnException :: Handle -> IO a -> IO a
+closeOnException h a = a `E.onException` hClose h

src/Simplex/Messaging/Server/MsgStore/STM.hs

@@ -1,129 +1,184 @@
-{-# LANGUAGE BangPatterns #-}
 {-# LANGUAGE ConstraintKinds #-}
+{-# LANGUAGE DataKinds #-}
+{-# LANGUAGE DuplicateRecordFields #-}
 {-# LANGUAGE FlexibleContexts #-}
 {-# LANGUAGE FlexibleInstances #-}
+{-# LANGUAGE InstanceSigs #-}
 {-# LANGUAGE LambdaCase #-}
 {-# LANGUAGE MultiParamTypeClasses #-}
 {-# LANGUAGE NamedFieldPuns #-}
 {-# LANGUAGE TupleSections #-}
+{-# LANGUAGE TypeFamilies #-}
 
 module Simplex.Messaging.Server.MsgStore.STM
-  ( STMMsgStore,
-    MsgQueue (msgQueue),
-    newMsgStore,
-    getMsgQueue,
-    delMsgQueue,
-    delMsgQueueSize,
-    writeMsg,
-    tryPeekMsg,
-    tryPeekMsgIO,
-    tryDelMsg,
-    tryDelPeekMsg,
-    deleteExpiredMsgs,
-    getQueueSize,
+  ( STMMsgStore (..),
+    STMStoreConfig (..),
   )
 where
 
-import qualified Data.ByteString.Char8 as B
+import Control.Concurrent.STM
+import Control.Monad.IO.Class
+import Control.Monad.Trans.Except
 import Data.Functor (($>))
 import Data.Int (Int64)
-import Data.Time.Clock.System (SystemTime (systemSeconds))
-import Simplex.Messaging.Protocol (Message (..), MsgId, RecipientId)
+import Simplex.Messaging.Protocol
+import Simplex.Messaging.Server.MsgStore.Types
+import Simplex.Messaging.Server.QueueStore
+import Simplex.Messaging.Server.QueueStore.STM
+import Simplex.Messaging.Server.StoreLog
 import Simplex.Messaging.TMap (TMap)
 import qualified Simplex.Messaging.TMap as TM
-import UnliftIO.STM
+import Simplex.Messaging.Util (($>>=), (<$$>))
+import System.IO (IOMode (..))
 
-data MsgQueue = MsgQueue
+data STMMsgStore = STMMsgStore
+  { storeConfig :: STMStoreConfig,
+    queues :: TMap RecipientId (QueueReference STMQueue),
+    queueCount :: TVar Int,
+    notifierCount :: TVar Int,
+    -- senders :: TMap SenderId RecipientId,
+    -- notifiers :: TMap NotifierId RecipientId,
+    storeLog :: TVar (Maybe (StoreLog 'WriteMode))
+  }
+
+data STMQueue = STMQueue
+  { -- To avoid race conditions and errors when restoring queues,
+    -- Nothing is written to TVar when queue is deleted.
+    queueRec :: TVar (Maybe QueueRec),
+    msgQueue_ :: TVar (Maybe STMMsgQueue)
+  }
+
+data STMMsgQueue = STMMsgQueue
   { msgQueue :: TQueue Message,
-    quota :: Int,
     canWrite :: TVar Bool,
     size :: TVar Int
   }
 
-type STMMsgStore = TMap RecipientId MsgQueue
+data STMStoreConfig = STMStoreConfig
+  { storePath :: Maybe FilePath,
+    quota :: Int
+  }
 
-newMsgStore :: IO STMMsgStore
-newMsgStore = TM.emptyIO
+instance STMQueueStore STMMsgStore where
+  queues' = queues
+  queueCount' = queueCount
+  notifierCount' = notifierCount
 
--- The reason for double lookup is that majority of messaging queues exist,
--- because multiple messages are sent to the same queue,
--- so the first lookup without STM transaction will return the queue faster.
--- In case the queue does not exist, it needs to be looked-up again inside transaction.
-getMsgQueue :: STMMsgStore -> RecipientId -> Int -> IO MsgQueue
-getMsgQueue st rId quota = TM.lookupIO rId st >>= maybe (atomically maybeNewQ) pure
-  where
-    maybeNewQ = TM.lookup rId st >>= maybe newQ pure
-    newQ = do
-      msgQueue <- newTQueue
-      canWrite <- newTVar True
-      size <- newTVar 0
-      let q = MsgQueue {msgQueue, quota, canWrite, size}
-      TM.insert rId q st
-      pure q
+  -- senders' = senders
+  -- notifiers' = notifiers
+  storeLog' = storeLog
+  mkQueue _ qr = STMQueue <$> newTVar (Just qr) <*> newTVar Nothing
+  msgQueue_' = msgQueue_
 
-delMsgQueue :: STMMsgStore -> RecipientId -> IO ()
-delMsgQueue st rId = atomically $ TM.delete rId st
+instance MsgStoreClass STMMsgStore where
+  type StoreMonad STMMsgStore = STM
+  type StoreQueue STMMsgStore = STMQueue
+  type MsgQueue STMMsgStore = STMMsgQueue
+  type MsgStoreConfig STMMsgStore = STMStoreConfig
 
-delMsgQueueSize :: STMMsgStore -> RecipientId -> IO Int
-delMsgQueueSize st rId = atomically (TM.lookupDelete rId st) >>= maybe (pure 0) (\MsgQueue {size} -> readTVarIO size)
+  newMsgStore :: STMStoreConfig -> IO STMMsgStore
+  newMsgStore storeConfig = do
+    queues <- TM.emptyIO
+    queueCount <- newTVarIO 0
+    notifierCount <- newTVarIO 0
+    -- senders <- TM.emptyIO
+    -- notifiers <- TM.emptyIO
+    storeLog <- newTVarIO Nothing
+    pure STMMsgStore {storeConfig, queues, queueCount, notifierCount, storeLog}
 
-writeMsg :: MsgQueue -> Message -> IO (Maybe (Message, Bool))
-writeMsg MsgQueue {msgQueue = q, quota, canWrite, size} !msg = atomically $ do
-  canWrt <- readTVar canWrite
-  empty <- isEmptyTQueue q
-  if canWrt || empty
-    then do
-      canWrt' <- (quota >) <$> readTVar size
-      writeTVar canWrite $! canWrt'
-      modifyTVar' size (+ 1)
-      if canWrt'
-        then writeTQueue q msg $> Just (msg, empty)
-        else (writeTQueue q $! msgQuota) $> Nothing
-    else pure Nothing
-  where
-    msgQuota = MessageQuota {msgId = msgId msg, msgTs = msgTs msg}
+  setStoreLog :: STMMsgStore -> StoreLog 'WriteMode -> IO ()
+  setStoreLog st sl = atomically $ writeTVar (storeLog st) (Just sl)
 
-tryPeekMsgIO :: MsgQueue -> IO (Maybe Message)
-tryPeekMsgIO = atomically . tryPeekTQueue . msgQueue
-{-# INLINE tryPeekMsgIO #-}
+  closeMsgStore st = readTVarIO (storeLog st) >>= mapM_ closeStoreLog
 
--- TODO remove once deliverToSub is split
-tryPeekMsg :: MsgQueue -> STM (Maybe Message)
-tryPeekMsg = tryPeekTQueue . msgQueue
-{-# INLINE tryPeekMsg #-}
+  activeMsgQueues = queues
+  {-# INLINE activeMsgQueues #-}
 
-tryDelMsg :: MsgQueue -> MsgId -> IO (Maybe Message)
-tryDelMsg mq msgId' = atomically $
-  tryPeekMsg mq >>= \case
-    msg_@(Just msg)
-      | msgId msg == msgId' || B.null msgId' -> tryDeleteMsg_ mq >> pure msg_
-      | otherwise -> pure Nothing
-    _ -> pure Nothing
+  withAllMsgQueues _ = withActiveMsgQueues
+  {-# INLINE withAllMsgQueues #-}
 
--- atomic delete (== read) last and peek next message if available
-tryDelPeekMsg :: MsgQueue -> MsgId -> IO (Maybe Message, Maybe Message)
-tryDelPeekMsg mq msgId' = atomically $
-  tryPeekMsg mq >>= \case
-    msg_@(Just msg)
-      | msgId msg == msgId' || B.null msgId' -> (msg_,) <$> (tryDeleteMsg_ mq >> tryPeekMsg mq)
-      | otherwise -> pure (Nothing, msg_)
-    _ -> pure (Nothing, Nothing)
+  logQueueStates _ = pure ()
 
-deleteExpiredMsgs :: MsgQueue -> Int64 -> IO Int
-deleteExpiredMsgs mq old = atomically $ loop 0
-  where
-    loop dc =
-      tryPeekMsg mq >>= \case
-        Just Message {msgTs}
-          | systemSeconds msgTs < old ->
-              tryDeleteMsg_ mq >> loop (dc + 1)
-        _ -> pure dc
+  logQueueState _ = pure ()
 
-tryDeleteMsg_ :: MsgQueue -> STM ()
-tryDeleteMsg_ MsgQueue {msgQueue = q, size} =
-  tryReadTQueue q >>= \case
-    Just _ -> modifyTVar' size (subtract 1)
-    _ -> pure ()
+  queueRec' = queueRec
+  {-# INLINE queueRec' #-}
 
-getQueueSize :: MsgQueue -> IO Int
-getQueueSize MsgQueue {size} = readTVarIO size
+  getMsgQueue :: STMMsgStore -> RecipientId -> STMQueue -> STM STMMsgQueue
+  getMsgQueue _ _ STMQueue {msgQueue_} = readTVar msgQueue_ >>= maybe newQ pure
+    where
+      newQ = do
+        msgQueue <- newTQueue
+        canWrite <- newTVar True
+        size <- newTVar 0
+        let q = STMMsgQueue {msgQueue, canWrite, size}
+        writeTVar msgQueue_ (Just q)
+        pure q
+
+  getPeekMsgQueue :: STMMsgStore -> RecipientId -> STMQueue -> STM (Maybe (STMMsgQueue, Message))
+  getPeekMsgQueue _ _ q@STMQueue {msgQueue_} = readTVar msgQueue_ $>>= \mq -> (mq,) <$$> tryPeekMsg_ q mq
+
+  -- does not create queue if it does not exist, does not delete it if it does (can't just close in-memory queue)
+  withIdleMsgQueue :: Int64 -> STMMsgStore -> RecipientId -> STMQueue -> (STMMsgQueue -> STM a) -> STM (Maybe a, Int)
+  withIdleMsgQueue _ _ _ STMQueue {msgQueue_} action =
+    readTVar msgQueue_ >>= \case
+      Just q -> do
+        r <- action q
+        sz <- getQueueSize_ q
+        pure (Just r, sz)
+      Nothing -> pure (Nothing, 0)
+
+  deleteQueue :: STMMsgStore -> RecipientId -> STMQueue -> IO (Either ErrorType QueueRec)
+  deleteQueue ms rId q = fst <$$> deleteQueue' ms rId q
+
+  deleteQueueSize :: STMMsgStore -> RecipientId -> STMQueue -> IO (Either ErrorType (QueueRec, Int))
+  deleteQueueSize ms rId q = deleteQueue' ms rId q >>= mapM (traverse getSize)
+    where
+      -- traverse operates on the second tuple element
+
+      getSize = maybe (pure 0) (\STMMsgQueue {size} -> readTVarIO size)
+
+  getQueueMessages_ :: Bool -> STMMsgQueue -> STM [Message]
+  getQueueMessages_ drainMsgs = (if drainMsgs then flushTQueue else snapshotTQueue) . msgQueue
+    where
+      snapshotTQueue q = do
+        msgs <- flushTQueue q
+        mapM_ (writeTQueue q) msgs
+        pure msgs
+
+  writeMsg :: STMMsgStore -> RecipientId -> STMQueue -> Bool -> Message -> ExceptT ErrorType IO (Maybe (Message, Bool))
+  writeMsg ms rId q' _logState msg = liftIO $ atomically $ do
+    STMMsgQueue {msgQueue = q, canWrite, size} <- getMsgQueue ms rId q'
+    canWrt <- readTVar canWrite
+    empty <- isEmptyTQueue q
+    if canWrt || empty
+      then do
+        canWrt' <- (quota >) <$> readTVar size
+        writeTVar canWrite $! canWrt'
+        modifyTVar' size (+ 1)
+        if canWrt'
+          then (writeTQueue q $! msg) $> Just (msg, empty)
+          else (writeTQueue q $! msgQuota) $> Nothing
+      else pure Nothing
+    where
+      STMMsgStore {storeConfig = STMStoreConfig {quota}} = ms
+      msgQuota = MessageQuota {msgId = messageId msg, msgTs = messageTs msg}
+
+  setOverQuota_ :: STMQueue -> IO ()
+  setOverQuota_ q = readTVarIO (msgQueue_ q) >>= mapM_ (\mq -> atomically $ writeTVar (canWrite mq) False)
+
+  getQueueSize_ :: STMMsgQueue -> STM Int
+  getQueueSize_ STMMsgQueue {size} = readTVar size
+
+  tryPeekMsg_ :: STMQueue -> STMMsgQueue -> STM (Maybe Message)
+  tryPeekMsg_ _ = tryPeekTQueue . msgQueue
+  {-# INLINE tryPeekMsg_ #-}
+
+  tryDeleteMsg_ :: STMQueue -> STMMsgQueue -> Bool -> STM ()
+  tryDeleteMsg_ _ STMMsgQueue {msgQueue = q, size} _logState =
+    tryReadTQueue q >>= \case
+      Just _ -> modifyTVar' size (subtract 1)
+      _ -> pure ()
+
+  isolateQueue :: RecipientId -> STMQueue -> String -> STM a -> ExceptT ErrorType IO a
+  isolateQueue _ _ _ = liftIO . atomically

src/Simplex/Messaging/Server/MsgStore/Types.hs

@@ -0,0 +1,145 @@
+{-# LANGUAGE BangPatterns #-}
+{-# LANGUAGE DataKinds #-}
+{-# LANGUAGE DuplicateRecordFields #-}
+{-# LANGUAGE FlexibleContexts #-}
+{-# LANGUAGE GADTs #-}
+{-# LANGUAGE LambdaCase #-}
+{-# LANGUAGE MultiWayIf #-}
+{-# LANGUAGE NamedFieldPuns #-}
+{-# LANGUAGE TupleSections #-}
+{-# LANGUAGE TypeFamilyDependencies #-}
+{-# OPTIONS_GHC -Wno-unrecognised-pragmas #-}
+
+{-# HLINT ignore "Redundant multi-way if" #-}
+
+module Simplex.Messaging.Server.MsgStore.Types where
+
+import Control.Concurrent.STM
+import Control.Monad (foldM)
+import Control.Monad.Trans.Except
+import Data.Functor (($>))
+import Data.Int (Int64)
+import Data.Kind
+import qualified Data.Map.Strict as M
+import Data.Time.Clock.System (SystemTime (systemSeconds))
+import Simplex.Messaging.Protocol
+import Simplex.Messaging.Server.QueueStore
+import Simplex.Messaging.Server.StoreLog.Types
+import Simplex.Messaging.TMap (TMap)
+import Simplex.Messaging.Util ((<$$>))
+import System.IO (IOMode (..))
+
+class MsgStoreClass s => STMQueueStore s where
+  queues' :: s -> TMap QueueId (QueueReference (StoreQueue s))
+  queueCount' :: s -> TVar Int
+  notifierCount' :: s -> TVar Int
+
+  -- senders' :: s -> TMap SenderId RecipientId
+  -- notifiers' :: s -> TMap NotifierId RecipientId
+  storeLog' :: s -> TVar (Maybe (StoreLog 'WriteMode))
+  mkQueue :: s -> QueueRec -> STM (StoreQueue s)
+  msgQueue_' :: StoreQueue s -> TVar (Maybe (MsgQueue s))
+
+class Monad (StoreMonad s) => MsgStoreClass s where
+  type StoreMonad s = (m :: Type -> Type) | m -> s
+  type MsgStoreConfig s = c | c -> s
+  type StoreQueue s = q | q -> s
+  type MsgQueue s = q | q -> s
+  newMsgStore :: MsgStoreConfig s -> IO s
+  setStoreLog :: s -> StoreLog 'WriteMode -> IO ()
+  closeMsgStore :: s -> IO ()
+  activeMsgQueues :: s -> TMap RecipientId (QueueReference (StoreQueue s))
+  withAllMsgQueues :: Monoid a => Bool -> s -> (RecipientId -> StoreQueue s -> IO a) -> IO a
+  logQueueStates :: s -> IO ()
+  logQueueState :: StoreQueue s -> StoreMonad s ()
+  queueRec' :: StoreQueue s -> TVar (Maybe QueueRec)
+  getPeekMsgQueue :: s -> RecipientId -> StoreQueue s -> StoreMonad s (Maybe (MsgQueue s, Message))
+  getMsgQueue :: s -> RecipientId -> StoreQueue s -> StoreMonad s (MsgQueue s)
+
+  -- the journal queue will be closed after action if it was initially closed or idle longer than interval in config
+  withIdleMsgQueue :: Int64 -> s -> RecipientId -> StoreQueue s -> (MsgQueue s -> StoreMonad s a) -> StoreMonad s (Maybe a, Int)
+  deleteQueue :: s -> RecipientId -> StoreQueue s -> IO (Either ErrorType QueueRec)
+  deleteQueueSize :: s -> RecipientId -> StoreQueue s -> IO (Either ErrorType (QueueRec, Int))
+  getQueueMessages_ :: Bool -> MsgQueue s -> StoreMonad s [Message]
+  writeMsg :: s -> RecipientId -> StoreQueue s -> Bool -> Message -> ExceptT ErrorType IO (Maybe (Message, Bool))
+  setOverQuota_ :: StoreQueue s -> IO () -- can ONLY be used while restoring messages, not while server running
+  getQueueSize_ :: MsgQueue s -> StoreMonad s Int
+  tryPeekMsg_ :: StoreQueue s -> MsgQueue s -> StoreMonad s (Maybe Message)
+  tryDeleteMsg_ :: StoreQueue s -> MsgQueue s -> Bool -> StoreMonad s ()
+  isolateQueue :: RecipientId -> StoreQueue s -> String -> StoreMonad s a -> ExceptT ErrorType IO a
+
+data MSType = MSMemory | MSJournal
+
+data SMSType :: MSType -> Type where
+  SMSMemory :: SMSType 'MSMemory
+  SMSJournal :: SMSType 'MSJournal
+
+data AMSType = forall s. AMSType (SMSType s)
+
+withActiveMsgQueues :: (MsgStoreClass s, Monoid a) => s -> (RecipientId -> StoreQueue s -> IO a) -> IO a
+withActiveMsgQueues st f = readTVarIO (activeMsgQueues st) >>= foldM run mempty . M.assocs
+  where
+    run !acc (k, QRRecipient v) = do
+      r <- f k v
+      pure $! acc <> r
+    run acc _ = pure acc
+
+getQueueMessages :: MsgStoreClass s => Bool -> s -> RecipientId -> StoreQueue s -> ExceptT ErrorType IO [Message]
+getQueueMessages drainMsgs st rId q = withPeekMsgQueue st rId q "getQueueSize" $ maybe (pure []) (getQueueMessages_ drainMsgs . fst)
+{-# INLINE getQueueMessages #-}
+
+getQueueSize :: MsgStoreClass s => s -> RecipientId -> StoreQueue s -> ExceptT ErrorType IO Int
+getQueueSize st rId q = withPeekMsgQueue st rId q "getQueueSize" $ maybe (pure 0) (getQueueSize_ . fst)
+{-# INLINE getQueueSize #-}
+
+tryPeekMsg :: MsgStoreClass s => s -> RecipientId -> StoreQueue s -> ExceptT ErrorType IO (Maybe Message)
+tryPeekMsg st rId q = snd <$$> withPeekMsgQueue st rId q "tryPeekMsg" pure
+{-# INLINE tryPeekMsg #-}
+
+tryDelMsg :: MsgStoreClass s => s -> RecipientId -> StoreQueue s -> MsgId -> ExceptT ErrorType IO (Maybe Message)
+tryDelMsg st rId q msgId' =
+  withPeekMsgQueue st rId q "tryDelMsg" $
+    maybe (pure Nothing) $ \(mq, msg) ->
+      if
+        | messageId msg == msgId' ->
+            tryDeleteMsg_ q mq True $> Just msg
+        | otherwise -> pure Nothing
+
+-- atomic delete (== read) last and peek next message if available
+tryDelPeekMsg :: MsgStoreClass s => s -> RecipientId -> StoreQueue s -> MsgId -> ExceptT ErrorType IO (Maybe Message, Maybe Message)
+tryDelPeekMsg st rId q msgId' =
+  withPeekMsgQueue st rId q "tryDelPeekMsg" $
+    maybe (pure (Nothing, Nothing)) $ \(mq, msg) ->
+      if
+        | messageId msg == msgId' -> (Just msg,) <$> (tryDeleteMsg_ q mq True >> tryPeekMsg_ q mq)
+        | otherwise -> pure (Nothing, Just msg)
+
+-- The action is called with Nothing when it is known that the queue is empty
+withPeekMsgQueue :: MsgStoreClass s => s -> RecipientId -> StoreQueue s -> String -> (Maybe (MsgQueue s, Message) -> StoreMonad s a) -> ExceptT ErrorType IO a
+withPeekMsgQueue st rId q op a = isolateQueue rId q op $ getPeekMsgQueue st rId q >>= a
+{-# INLINE withPeekMsgQueue #-}
+
+deleteExpiredMsgs :: MsgStoreClass s => s -> RecipientId -> StoreQueue s -> Int64 -> ExceptT ErrorType IO Int
+deleteExpiredMsgs st rId q old =
+  isolateQueue rId q "deleteExpiredMsgs" $
+    getMsgQueue st rId q >>= deleteExpireMsgs_ old q
+
+-- closed and idle queues will be closed after expiration
+-- returns (expired count, queue size after expiration)
+idleDeleteExpiredMsgs :: MsgStoreClass s => Int64 -> s -> RecipientId -> StoreQueue s -> Int64 -> ExceptT ErrorType IO (Maybe Int, Int)
+idleDeleteExpiredMsgs now st rId q old =
+  isolateQueue rId q "idleDeleteExpiredMsgs" $
+    withIdleMsgQueue now st rId q (deleteExpireMsgs_ old q)
+
+deleteExpireMsgs_ :: MsgStoreClass s => Int64 -> StoreQueue s -> MsgQueue s -> StoreMonad s Int
+deleteExpireMsgs_ old q mq = do
+  n <- loop 0
+  logQueueState q
+  pure n
+  where
+    loop dc =
+      tryPeekMsg_ q mq >>= \case
+        Just Message {msgTs}
+          | systemSeconds msgTs < old ->
+              tryDeleteMsg_ q mq False >> loop (dc + 1)
+        _ -> pure dc

src/Simplex/Messaging/Server/NtfStore.hs

@@ -0,0 +1,67 @@
+{-# LANGUAGE BangPatterns #-}
+{-# LANGUAGE LambdaCase #-}
+{-# LANGUAGE NamedFieldPuns #-}
+{-# LANGUAGE OverloadedStrings #-}
+
+module Simplex.Messaging.Server.NtfStore where
+
+import Control.Concurrent.STM
+import Control.Monad (foldM)
+import Data.Int (Int64)
+import qualified Data.Map.Strict as M
+import Data.Time.Clock.System (SystemTime (..))
+import qualified Simplex.Messaging.Crypto as C
+import Simplex.Messaging.Encoding.String
+import Simplex.Messaging.Protocol (EncNMsgMeta, MsgId, NotifierId)
+import Simplex.Messaging.TMap (TMap)
+import qualified Simplex.Messaging.TMap as TM
+
+newtype NtfStore = NtfStore (TMap NotifierId (TVar [MsgNtf]))
+
+data MsgNtf = MsgNtf
+  { ntfMsgId :: MsgId,
+    ntfTs :: SystemTime,
+    ntfNonce :: C.CbNonce,
+    ntfEncMeta :: EncNMsgMeta
+  }
+
+storeNtf :: NtfStore -> NotifierId -> MsgNtf -> IO ()
+storeNtf (NtfStore ns) nId ntf = do
+  TM.lookupIO nId ns >>= atomically . maybe newNtfs (`modifyTVar'` (ntf :))
+  -- TODO coalesce messages here once the client is updated to process multiple messages
+  -- for single notification.
+  -- when (isJust prevNtf) $ incStat $ msgNtfReplaced stats
+  where
+    newNtfs = TM.lookup nId ns >>= maybe (TM.insertM nId (newTVar [ntf]) ns) (`modifyTVar'` (ntf :))
+
+deleteNtfs :: NtfStore -> NotifierId -> IO Int
+deleteNtfs (NtfStore ns) nId = atomically (TM.lookupDelete nId ns) >>= maybe (pure 0) (fmap length . readTVarIO)
+
+deleteExpiredNtfs :: NtfStore -> Int64 -> IO Int
+deleteExpiredNtfs (NtfStore ns) old =
+  foldM (\expired -> fmap (expired +) . expireQueue) 0 . M.keys =<< readTVarIO ns
+  where
+    expireQueue nId = TM.lookupIO nId ns >>= maybe (pure 0) expire
+    expire v = readTVarIO v >>= \case
+      [] -> pure 0
+      _ ->
+        atomically $ readTVar v >>= \case
+          [] -> pure 0
+          -- check the last message first, it is the earliest
+          ntfs | systemSeconds (ntfTs $ last $ ntfs) < old -> do
+            let !ntfs' = filter (\MsgNtf {ntfTs = ts} -> systemSeconds ts >= old) ntfs
+            writeTVar v ntfs'
+            pure $! length ntfs - length ntfs'
+          _ -> pure 0
+
+data NtfLogRecord = NLRv1 NotifierId MsgNtf
+
+instance StrEncoding MsgNtf where
+  strEncode MsgNtf {ntfMsgId, ntfTs, ntfNonce, ntfEncMeta} = strEncode (ntfMsgId, ntfTs, ntfNonce, ntfEncMeta)
+  strP = do
+    (ntfMsgId, ntfTs, ntfNonce, ntfEncMeta) <- strP
+    pure MsgNtf {ntfMsgId, ntfTs, ntfNonce, ntfEncMeta}
+
+instance StrEncoding NtfLogRecord where
+  strEncode (NLRv1 nId ntf) = strEncode (Str "v1", nId, ntf)
+  strP = "v1 " *> (NLRv1 <$> strP_ <*> strP)

src/Simplex/Messaging/Server/QueueStore.hs

@@ -11,6 +11,11 @@ import Data.Time.Clock.System (SystemTime (..), getSystemTime)
 import Simplex.Messaging.Encoding.String
 import Simplex.Messaging.Protocol
 
+data QueueReference q
+  = QRRecipient q
+  | QRSender q
+  | QRNotifier q
+
 data QueueRec = QueueRec
   { recipientId :: !RecipientId,
     recipientKey :: !RcvPublicAuthKey,

src/Simplex/Messaging/Server/QueueStore/STM.hs

@@ -1,117 +1,215 @@
 {-# LANGUAGE BangPatterns #-}
 {-# LANGUAGE DataKinds #-}
+{-# LANGUAGE DuplicateRecordFields #-}
 {-# LANGUAGE FlexibleInstances #-}
 {-# LANGUAGE GADTs #-}
 {-# LANGUAGE KindSignatures #-}
 {-# LANGUAGE LambdaCase #-}
 {-# LANGUAGE MultiParamTypeClasses #-}
 {-# LANGUAGE NamedFieldPuns #-}
+{-# LANGUAGE OverloadedStrings #-}
 {-# LANGUAGE RankNTypes #-}
 {-# LANGUAGE ScopedTypeVariables #-}
+{-# LANGUAGE TupleSections #-}
 {-# LANGUAGE UndecidableInstances #-}
 
 module Simplex.Messaging.Server.QueueStore.STM
-  ( QueueStore (..),
-    newQueueStore,
-    addQueue,
+  ( addQueue,
     getQueue,
+    getQueueRec,
     secureQueue,
     addQueueNotifier,
     deleteQueueNotifier,
     suspendQueue,
     updateQueueTime,
-    deleteQueue,
+    deleteQueue',
+    readQueueStore,
+    withLog',
   )
 where
 
+import qualified Control.Exception as E
+import Control.Logger.Simple
 import Control.Monad
+import Control.Monad.IO.Class
+import Control.Monad.Trans.Except
+import Data.Bitraversable (bimapM)
+import qualified Data.ByteString.Char8 as B
+import qualified Data.ByteString.Lazy.Char8 as LB
 import Data.Functor (($>))
+import qualified Data.Text as T
+import Data.Text.Encoding (decodeLatin1)
+import Simplex.Messaging.Encoding.String
 import Simplex.Messaging.Protocol
+import Simplex.Messaging.Server.MsgStore.Types
 import Simplex.Messaging.Server.QueueStore
-import Simplex.Messaging.TMap (TMap)
+import Simplex.Messaging.Server.StoreLog
 import qualified Simplex.Messaging.TMap as TM
-import Simplex.Messaging.Util (ifM, ($>>=))
+import Simplex.Messaging.Util (ifM, tshow, ($>>=), (<$$))
+import System.IO
 import UnliftIO.STM
 
-data QueueStore = QueueStore
-  { queues :: TMap RecipientId (TVar QueueRec),
-    senders :: TMap SenderId RecipientId,
-    notifiers :: TMap NotifierId RecipientId
-  }
-
-newQueueStore :: IO QueueStore
-newQueueStore = do
-  queues <- TM.emptyIO
-  senders <- TM.emptyIO
-  notifiers <- TM.emptyIO
-  pure QueueStore {queues, senders, notifiers}
-
-addQueue :: QueueStore -> QueueRec -> IO (Either ErrorType ())
-addQueue QueueStore {queues, senders} q@QueueRec {recipientId = rId, senderId = sId} = atomically $ do
-  ifM hasId (pure $ Left DUPLICATE_) $ do
-    qVar <- newTVar q
-    TM.insert rId qVar queues
-    TM.insert sId rId senders
-    pure $ Right ()
+addQueue :: STMQueueStore s => s -> QueueRec -> IO (Either ErrorType (StoreQueue s))
+addQueue st qr@QueueRec {recipientId = rId, senderId = sId, notifier} =
+  atomically add
+    $>>= \q -> q <$$ withLog "addQueue" st (`logCreateQueue` qr)
   where
-    hasId = (||) <$> TM.member rId queues <*> TM.member sId senders
+    add = ifM duplicateIds (pure $ Left DUPLICATE_) $ do
+      q <- mkQueue st qr
+      TM.insert rId (QRRecipient q) qs
+      TM.insert sId (QRSender q) qs
+      modifyTVar' (queueCount' st) (+ 1)
+      forM_ notifier $ \NtfCreds {notifierId} -> do
+        TM.insert notifierId (QRNotifier q) qs
+        modifyTVar' (notifierCount' st) (+ 1)
+      pure $ Right q
+    duplicateIds
+      | rId == sId || sameNtf rId || sameNtf sId = pure False
+      | otherwise = or <$> sequence [TM.member rId qs, TM.member sId qs, hasNotifier]
+    sameNtf qId = maybe False ((qId ==) . notifierId) notifier
+    hasNotifier = maybe (pure False) (\NtfCreds {notifierId} -> TM.member notifierId qs) notifier
+    qs = queues' st
 
-getQueue :: DirectParty p => QueueStore -> SParty p -> QueueId -> IO (Either ErrorType QueueRec)
-getQueue QueueStore {queues, senders, notifiers} party qId =
-  toResult <$> (mapM readTVarIO =<< getVar)
+getQueue :: (STMQueueStore s, DirectParty p) => s -> SParty p -> QueueId -> IO (Either ErrorType (StoreQueue s))
+getQueue st party qId = fmap (maybe (Left AUTH) Right) $ do
+  q_ <- TM.lookupIO qId qs
+  pure $ case (q_, party) of
+    (Just (QRRecipient q), SRecipient) -> Just q
+    (Just (QRSender q), SSender) -> Just q --  -> TM.lookupIO sId qs $>>= \case QRRecipient q -> pure $ Just q; _ -> pure Nothing
+    (Just (QRNotifier q), SNotifier) -> Just q --  -> TM.lookupIO nId qs $>>= \case QRRecipient q -> pure $ Just q; _ -> pure Nothing
+    _ -> Nothing
   where
-    getVar = case party of
-      SRecipient -> TM.lookupIO qId queues
-      SSender -> TM.lookupIO qId senders $>>= (`TM.lookupIO` queues)
-      SNotifier -> TM.lookupIO qId notifiers $>>= (`TM.lookupIO` queues)
+    qs = queues' st
 
-secureQueue :: QueueStore -> RecipientId -> SndPublicAuthKey -> IO (Either ErrorType QueueRec)
-secureQueue QueueStore {queues} rId sKey = toResult <$> do
-  TM.lookupIO rId queues $>>= \qVar -> atomically $ 
-    readTVar qVar >>= \q -> case senderKey q of
-      Just k -> pure $ if sKey == k then Just q else Nothing
-      _ ->
-        let !q' = q {senderKey = Just sKey}
-         in writeTVar qVar q' $> Just q'
+getQueueRec :: (STMQueueStore s, DirectParty p) => s -> SParty p -> QueueId -> IO (Either ErrorType (StoreQueue s, QueueRec))
+getQueueRec st party qId =
+  getQueue st party qId
+    $>>= (\q -> maybe (Left AUTH) (Right . (q,)) <$> readTVarIO (queueRec' q))
 
-addQueueNotifier :: QueueStore -> RecipientId -> NtfCreds -> IO (Either ErrorType QueueRec)
-addQueueNotifier QueueStore {queues, notifiers} rId ntfCreds@NtfCreds {notifierId = nId} = do
-  ifM (TM.memberIO nId notifiers) (pure $ Left DUPLICATE_) $
-    withQueue rId queues $ \qVar -> do
-      q <- readTVar qVar
-      forM_ (notifier q) $ (`TM.delete` notifiers) . notifierId
+secureQueue :: STMQueueStore s => s -> StoreQueue s -> SndPublicAuthKey -> IO (Either ErrorType ())
+secureQueue st sq sKey =
+  atomically (readQueueRec qr $>>= secure)
+    $>>= \rId -> withLog "secureQueue" st $ \s -> logSecureQueue s rId sKey
+  where
+    qr = queueRec' sq
+    secure q@QueueRec {recipientId = rId} = case senderKey q of
+      Just k -> pure $ if sKey == k then Right rId else Left AUTH
+      Nothing -> do
+        writeTVar qr $ Just q {senderKey = Just sKey}
+        pure $ Right rId
+
+addQueueNotifier :: STMQueueStore s => s -> StoreQueue s -> NtfCreds -> IO (Either ErrorType (Maybe NotifierId))
+addQueueNotifier st sq ntfCreds@NtfCreds {notifierId = nId} =
+  atomically (readQueueRec qr $>>= add)
+    $>>= \(rId, nId_) -> nId_ <$$ withLog "addQueueNotifier" st (\s -> logAddNotifier s rId ntfCreds)
+  where
+    qr = queueRec' sq
+    qs = queues' st
+    add q@QueueRec {recipientId = rId} = ifM (TM.member nId qs) (pure $ Left DUPLICATE_) $ do
+      nId_ <- forM (notifier q) $ \NtfCreds {notifierId} -> TM.delete notifierId qs $> notifierId
       let !q' = q {notifier = Just ntfCreds}
-      writeTVar qVar q'
-      TM.insert nId rId notifiers
-      pure q'
+      writeTVar qr $ Just q'
+      TM.insert nId (QRNotifier sq) qs
+      modifyTVar' (notifierCount' st) (+ 1)
+      pure $ Right (rId, nId_)
 
-deleteQueueNotifier :: QueueStore -> RecipientId -> IO (Either ErrorType ())
-deleteQueueNotifier QueueStore {queues, notifiers} rId =
-  withQueue rId queues $ \qVar -> do
-    q <- readTVar qVar
-    forM_ (notifier q) $ \NtfCreds {notifierId} -> TM.delete notifierId notifiers
-    writeTVar qVar $! q {notifier = Nothing}
+deleteQueueNotifier :: STMQueueStore s => s -> StoreQueue s -> IO (Either ErrorType (Maybe NotifierId))
+deleteQueueNotifier st sq =
+  atomically (readQueueRec qr >>= mapM delete)
+    $>>= \(rId, nId_) -> nId_ <$$ withLog "deleteQueueNotifier" st (`logDeleteNotifier` rId)
+  where
+    qr = queueRec' sq
+    delete q = fmap (recipientId q,) $ forM (notifier q) $ \NtfCreds {notifierId} -> do
+      TM.delete notifierId $ queues' st
+      modifyTVar' (notifierCount' st) (subtract 1)
+      writeTVar qr $ Just q {notifier = Nothing}
+      pure notifierId
 
-suspendQueue :: QueueStore -> RecipientId -> IO (Either ErrorType ())
-suspendQueue QueueStore {queues} rId =
-  withQueue rId queues (`modifyTVar'` \q -> q {status = QueueOff})
+suspendQueue :: STMQueueStore s => s -> StoreQueue s -> IO (Either ErrorType ())
+suspendQueue st sq =
+  atomically (readQueueRec qr >>= mapM suspend)
+    $>>= \rId -> withLog "suspendQueue" st (`logSuspendQueue` rId)
+  where
+    qr = queueRec' sq
+    suspend q = do
+      writeTVar qr $ Just q {status = QueueOff}
+      pure $ recipientId q
 
-updateQueueTime :: QueueStore -> RecipientId -> RoundedSystemTime -> IO ()
-updateQueueTime QueueStore {queues} rId t =
-  void $ withQueue rId queues (`modifyTVar'` \q -> q {updatedAt = Just t})
+updateQueueTime :: STMQueueStore s => s -> StoreQueue s -> RoundedSystemTime -> IO (Either ErrorType QueueRec)
+updateQueueTime st sq t = atomically (readQueueRec qr >>= mapM update) $>>= log'
+  where
+    qr = queueRec' sq
+    update q@QueueRec {updatedAt}
+      | updatedAt == Just t = pure (q, False)
+      | otherwise =
+          let !q' = q {updatedAt = Just t}
+           in writeTVar qr (Just q') $> (q', True)
+    log' (q, changed)
+      | changed = q <$$ withLog "updateQueueTime" st (\sl -> logUpdateQueueTime sl (recipientId q) t)
+      | otherwise = pure $ Right q
 
-deleteQueue :: QueueStore -> RecipientId -> IO (Either ErrorType QueueRec)
-deleteQueue QueueStore {queues, senders, notifiers} rId = atomically $ do
-  TM.lookupDelete rId queues >>= \case
-    Just qVar ->
-      readTVar qVar >>= \q -> do
-        TM.delete (senderId q) senders
-        forM_ (notifier q) $ \NtfCreds {notifierId} -> TM.delete notifierId notifiers
-        pure $ Right q
-    _ -> pure $ Left AUTH
+deleteQueue' :: STMQueueStore s => s -> RecipientId -> StoreQueue s -> IO (Either ErrorType (QueueRec, Maybe (MsgQueue s)))
+deleteQueue' st rId sq =
+  atomically (readQueueRec qr >>= mapM delete)
+    $>>= \q ->
+      withLog "deleteQueue" st (`logDeleteQueue` rId)
+        >>= bimapM pure (\_ -> (q,) <$> atomically (swapTVar (msgQueue_' sq) Nothing))
+  where
+    qr = queueRec' sq
+    qs = queues' st
+    delete q = do
+      writeTVar qr Nothing
+      TM.delete (senderId q) qs
+      modifyTVar' (queueCount' st) (subtract 1)
+      forM_ (notifier q) $ \NtfCreds {notifierId} -> do
+        TM.delete notifierId qs
+        modifyTVar' (notifierCount' st) (subtract 1)
+      pure q
 
-toResult :: Maybe a -> Either ErrorType a
-toResult = maybe (Left AUTH) Right
+readQueueRec :: TVar (Maybe QueueRec) -> STM (Either ErrorType QueueRec)
+readQueueRec qr = maybe (Left AUTH) Right <$> readTVar qr
+{-# INLINE readQueueRec #-}
 
-withQueue :: RecipientId -> TMap RecipientId (TVar QueueRec) -> (TVar QueueRec -> STM a) -> IO (Either ErrorType a)
-withQueue rId queues f = toResult <$> TM.lookupIO rId queues >>= atomically . mapM f
+withLog' :: String -> TVar (Maybe (StoreLog 'WriteMode)) -> (StoreLog 'WriteMode -> IO ()) -> IO (Either ErrorType ())
+withLog' name sl action =
+  readTVarIO sl
+    >>= maybe (pure $ Right ()) (E.try . action >=> bimapM logErr pure)
+  where
+    logErr :: E.SomeException -> IO ErrorType
+    logErr e = logError ("STORE: " <> T.pack err) $> STORE err
+      where
+        err = name <> ", withLog, " <> show e
+
+withLog :: STMQueueStore s => String -> s -> (StoreLog 'WriteMode -> IO ()) -> IO (Either ErrorType ())
+withLog name = withLog' name . storeLog'
+
+readQueueStore :: forall s. STMQueueStore s => FilePath -> s -> IO ()
+readQueueStore f st = withFile f ReadMode $ LB.hGetContents >=> mapM_ processLine . LB.lines
+  where
+    processLine :: LB.ByteString -> IO ()
+    processLine s' = either printError procLogRecord (strDecode s)
+      where
+        s = LB.toStrict s'
+        procLogRecord :: StoreLogRecord -> IO ()
+        procLogRecord = \case
+          CreateQueue q -> addQueue st q >>= qError (recipientId q) "CreateQueue"
+          SecureQueue qId sKey -> withQueue qId "SecureQueue" $ \q -> secureQueue st q sKey
+          AddNotifier qId ntfCreds -> withQueue qId "AddNotifier" $ \q -> addQueueNotifier st q ntfCreds
+          SuspendQueue qId -> withQueue qId "SuspendQueue" $ suspendQueue st
+          DeleteQueue qId -> withQueue qId "DeleteQueue" $ deleteQueue st qId
+          DeleteNotifier qId -> withQueue qId "DeleteNotifier" $ deleteQueueNotifier st
+          UpdateTime qId t -> withQueue qId "UpdateTime" $ \q -> updateQueueTime st q t
+        printError :: String -> IO ()
+        printError e = B.putStrLn $ "Error parsing log: " <> B.pack e <> " - " <> s
+        withQueue :: forall a. RecipientId -> T.Text -> (StoreQueue s -> IO (Either ErrorType a)) -> IO ()
+        withQueue qId op a = runExceptT go >>= qError qId op
+          where
+            go = do
+              q <- ExceptT $ getQueue st SRecipient qId
+              liftIO (readTVarIO $ queueRec' q) >>= \case
+                Nothing -> logWarn $ logPfx qId op <> "already deleted"
+                Just _ -> void $ ExceptT $ a q
+        qError qId op = \case
+          Left e -> logError $ logPfx qId op <> tshow e
+          Right _ -> pure ()
+        logPfx qId op = "STORE: " <> op <> ", stored queue " <> decodeLatin1 (strEncode qId) <> ", "

src/Simplex/Messaging/Server/Stats.hs

@@ -67,15 +67,18 @@ data ServerStats = ServerStats
     msgRecvNtf :: IORef Int, -- received messages with NTF flag
     activeQueuesNtf :: PeriodStats,
     msgNtfs :: IORef Int, -- messages notications delivered to NTF server (<= msgSentNtf)
+    msgNtfsB :: IORef Int, -- messages notication batches delivered to NTF server
     msgNtfNoSub :: IORef Int, -- no subscriber to notifications (e.g., NTF server not connected)
     msgNtfLost :: IORef Int, -- notification is lost because NTF delivery queue is full
+    msgNtfExpired :: IORef Int, -- expired
     pRelays :: ProxyStats,
     pRelaysOwn :: ProxyStats,
     pMsgFwds :: ProxyStats,
     pMsgFwdsOwn :: ProxyStats,
     pMsgFwdsRecv :: IORef Int,
     qCount :: IORef Int,
-    msgCount :: IORef Int
+    msgCount :: IORef Int,
+    ntfCount :: IORef Int
   }
 
 data ServerStatsData = ServerStatsData
@@ -117,15 +120,18 @@ data ServerStatsData = ServerStatsData
     _msgRecvNtf :: Int,
     _activeQueuesNtf :: PeriodStatsData,
     _msgNtfs :: Int,
+    _msgNtfsB :: Int,
     _msgNtfNoSub :: Int,
     _msgNtfLost :: Int,
+    _msgNtfExpired :: Int,
     _pRelays :: ProxyStatsData,
     _pRelaysOwn :: ProxyStatsData,
     _pMsgFwds :: ProxyStatsData,
     _pMsgFwdsOwn :: ProxyStatsData,
     _pMsgFwdsRecv :: Int,
     _qCount :: Int,
-    _msgCount :: Int
+    _msgCount :: Int,
+    _ntfCount :: Int
   }
   deriving (Show)
 
@@ -169,8 +175,10 @@ newServerStats ts = do
   msgRecvNtf <- newIORef 0
   activeQueuesNtf <- newPeriodStats
   msgNtfs <- newIORef 0
+  msgNtfsB <- newIORef 0
   msgNtfNoSub <- newIORef 0
   msgNtfLost <- newIORef 0
+  msgNtfExpired <- newIORef 0
   pRelays <- newProxyStats
   pRelaysOwn <- newProxyStats
   pMsgFwds <- newProxyStats
@@ -178,6 +186,7 @@ newServerStats ts = do
   pMsgFwdsRecv <- newIORef 0
   qCount <- newIORef 0
   msgCount <- newIORef 0
+  ntfCount <- newIORef 0
   pure
     ServerStats
       { fromTime,
@@ -218,15 +227,18 @@ newServerStats ts = do
         msgRecvNtf,
         activeQueuesNtf,
         msgNtfs,
+        msgNtfsB,
         msgNtfNoSub,
         msgNtfLost,
+        msgNtfExpired,
         pRelays,
         pRelaysOwn,
         pMsgFwds,
         pMsgFwdsOwn,
         pMsgFwdsRecv,
         qCount,
-        msgCount
+        msgCount,
+        ntfCount
       }
 
 getServerStatsData :: ServerStats -> IO ServerStatsData
@@ -269,8 +281,10 @@ getServerStatsData s = do
   _msgRecvNtf <- readIORef $ msgRecvNtf s
   _activeQueuesNtf <- getPeriodStatsData $ activeQueuesNtf s
   _msgNtfs <- readIORef $ msgNtfs s
+  _msgNtfsB <- readIORef $ msgNtfsB s
   _msgNtfNoSub <- readIORef $ msgNtfNoSub s
   _msgNtfLost <- readIORef $ msgNtfLost s
+  _msgNtfExpired <- readIORef $ msgNtfExpired s
   _pRelays <- getProxyStatsData $ pRelays s
   _pRelaysOwn <- getProxyStatsData $ pRelaysOwn s
   _pMsgFwds <- getProxyStatsData $ pMsgFwds s
@@ -278,6 +292,7 @@ getServerStatsData s = do
   _pMsgFwdsRecv <- readIORef $ pMsgFwdsRecv s
   _qCount <- readIORef $ qCount s
   _msgCount <- readIORef $ msgCount s
+  _ntfCount <- readIORef $ ntfCount s
   pure
     ServerStatsData
       { _fromTime,
@@ -318,15 +333,18 @@ getServerStatsData s = do
         _msgRecvNtf,
         _activeQueuesNtf,
         _msgNtfs,
+        _msgNtfsB,
         _msgNtfNoSub,
         _msgNtfLost,
+        _msgNtfExpired,
         _pRelays,
         _pRelaysOwn,
         _pMsgFwds,
         _pMsgFwdsOwn,
         _pMsgFwdsRecv,
         _qCount,
-        _msgCount
+        _msgCount,
+        _ntfCount
       }
 
 -- this function is not thread safe, it is used on server start only
@@ -370,8 +388,10 @@ setServerStats s d = do
   writeIORef (msgRecvNtf s) $! _msgRecvNtf d
   setPeriodStats (activeQueuesNtf s) (_activeQueuesNtf d)
   writeIORef (msgNtfs s) $! _msgNtfs d
+  writeIORef (msgNtfsB s) $! _msgNtfsB d
   writeIORef (msgNtfNoSub s) $! _msgNtfNoSub d
   writeIORef (msgNtfLost s) $! _msgNtfLost d
+  writeIORef (msgNtfExpired s) $! _msgNtfExpired d
   setProxyStats (pRelays s) $! _pRelays d
   setProxyStats (pRelaysOwn s) $! _pRelaysOwn d
   setProxyStats (pMsgFwds s) $! _pMsgFwds d
@@ -379,6 +399,7 @@ setServerStats s d = do
   writeIORef (pMsgFwdsRecv s) $! _pMsgFwdsRecv d
   writeIORef (qCount s) $! _qCount d
   writeIORef (msgCount s) $! _msgCount d
+  writeIORef (ntfCount s) $! _ntfCount d
 
 instance StrEncoding ServerStatsData where
   strEncode d =
@@ -420,8 +441,10 @@ instance StrEncoding ServerStatsData where
         "msgSentNtf=" <> strEncode (_msgSentNtf d),
         "msgRecvNtf=" <> strEncode (_msgRecvNtf d),
         "msgNtfs=" <> strEncode (_msgNtfs d),
+        "msgNtfsB=" <> strEncode (_msgNtfsB d),
         "msgNtfNoSub=" <> strEncode (_msgNtfNoSub d),
         "msgNtfLost=" <> strEncode (_msgNtfLost d),
+        "msgNtfExpired=" <> strEncode (_msgNtfExpired d),
         "activeQueues:",
         strEncode (_activeQueues d),
         "activeQueuesNtf:",
@@ -475,8 +498,10 @@ instance StrEncoding ServerStatsData where
     _msgSentNtf <- opt "msgSentNtf="
     _msgRecvNtf <- opt "msgRecvNtf="
     _msgNtfs <- opt "msgNtfs="
+    _msgNtfsB <- opt "msgNtfsB="
     _msgNtfNoSub <- opt "msgNtfNoSub="
     _msgNtfLost <- opt "msgNtfLost="
+    _msgNtfExpired <- opt "msgNtfExpired="
     _activeQueues <-
       optional ("activeQueues:" <* A.endOfLine) >>= \case
         Just _ -> strP <* optional A.endOfLine
@@ -536,8 +561,10 @@ instance StrEncoding ServerStatsData where
           _msgSentNtf,
           _msgRecvNtf,
           _msgNtfs,
+          _msgNtfsB,
           _msgNtfNoSub,
           _msgNtfLost,
+          _msgNtfExpired,
           _activeQueues,
           _activeQueuesNtf,
           _pRelays,
@@ -546,7 +573,8 @@ instance StrEncoding ServerStatsData where
           _pMsgFwdsOwn,
           _pMsgFwdsRecv,
           _qCount,
-          _msgCount = 0
+          _msgCount = 0,
+          _ntfCount = 0
         }
     where
       opt s = A.string s *> strP <* A.endOfLine <|> pure 0

src/Simplex/Messaging/Server/StoreLog.hs

@@ -1,3 +1,4 @@
+{-# LANGUAGE BangPatterns #-}
 {-# LANGUAGE DataKinds #-}
 {-# LANGUAGE DuplicateRecordFields #-}
 {-# LANGUAGE GADTs #-}
@@ -5,10 +6,12 @@
 {-# LANGUAGE LambdaCase #-}
 {-# LANGUAGE NamedFieldPuns #-}
 {-# LANGUAGE OverloadedStrings #-}
+{-# LANGUAGE ScopedTypeVariables #-}
 {-# OPTIONS_GHC -fno-warn-orphans #-}
 
 module Simplex.Messaging.Server.StoreLog
   ( StoreLog, -- constructors are not exported
+    StoreLogRecord (..), -- used in tests
     openWriteStoreLog,
     openReadStoreLog,
     storeLogFilePath,
@@ -22,31 +25,32 @@ module Simplex.Messaging.Server.StoreLog
     logDeleteNotifier,
     logUpdateQueueTime,
     readWriteStoreLog,
+    writeQueueStore,
   )
 where
 
 import Control.Applicative (optional, (<|>))
-import Control.Monad (foldM, unless, when)
+import Control.Concurrent.STM
+import qualified Control.Exception as E
+import Control.Logger.Simple
+import Control.Monad
 import qualified Data.Attoparsec.ByteString.Char8 as A
 import qualified Data.ByteString.Char8 as B
-import qualified Data.ByteString.Lazy.Char8 as LB
-import Data.Functor (($>))
-import Data.Map.Strict (Map)
 import qualified Data.Map.Strict as M
+import qualified Data.Text as T
+import Data.Time.Clock (getCurrentTime)
+import Data.Time.Format.ISO8601 (iso8601Show)
+import GHC.IO (catchAny)
 import Simplex.Messaging.Encoding.String
 import Simplex.Messaging.Protocol
+import Simplex.Messaging.Server.MsgStore.Types
 import Simplex.Messaging.Server.QueueStore
-import Simplex.Messaging.Transport.Buffer (trimCR)
-import Simplex.Messaging.Util (ifM)
+import Simplex.Messaging.Server.StoreLog.Types
+import qualified Simplex.Messaging.TMap as TM
+import Simplex.Messaging.Util (ifM, tshow, unlessM, whenM)
 import System.Directory (doesFileExist, renameFile)
 import System.IO
 
--- | opaque container for file handle with a type-safe IOMode
--- constructors are not exported, openWriteStoreLog and openReadStoreLog should be used instead
-data StoreLog (a :: IOMode) where
-  ReadStoreLog :: FilePath -> Handle -> StoreLog 'ReadMode
-  WriteStoreLog :: FilePath -> Handle -> StoreLog 'WriteMode
-
 data StoreLogRecord
   = CreateQueue QueueRec
   | SecureQueue QueueId SndPublicAuthKey
@@ -55,6 +59,7 @@ data StoreLogRecord
   | DeleteQueue QueueId
   | DeleteNotifier QueueId
   | UpdateTime QueueId RoundedSystemTime
+  deriving (Show)
 
 data SLRTag
   = CreateQueue_
@@ -74,10 +79,11 @@ instance StrEncoding QueueRec where
         "sid=" <> strEncode senderId,
         "sk=" <> strEncode senderKey
       ]
-      <> if sndSecure then " sndSecure=" <> strEncode sndSecure else ""
+      <> sndSecureStr
       <> maybe "" notifierStr notifier
       <> maybe "" updatedAtStr updatedAt
     where
+      sndSecureStr = if sndSecure then " sndSecure=" <> strEncode sndSecure else ""
       notifierStr ntfCreds = " notifier=" <> strEncode ntfCreds
       updatedAtStr t = " updated_at=" <> strEncode t
 
@@ -121,7 +127,7 @@ instance StrEncoding StoreLogRecord where
     SuspendQueue rId -> strEncode (SuspendQueue_, rId)
     DeleteQueue rId -> strEncode (DeleteQueue_, rId)
     DeleteNotifier rId -> strEncode (DeleteNotifier_, rId)
-    UpdateTime rId t ->  strEncode (UpdateTime_, rId, t)
+    UpdateTime rId t -> strEncode (UpdateTime_, rId, t)
 
   strP =
     strP_ >>= \case
@@ -141,7 +147,7 @@ openWriteStoreLog f = do
 
 openReadStoreLog :: FilePath -> IO (StoreLog 'ReadMode)
 openReadStoreLog f = do
-  doesFileExist f >>= (`unless` writeFile f "")
+  unlessM (doesFileExist f) (writeFile f "")
   ReadStoreLog f <$> openFile f ReadMode
 
 storeLogFilePath :: StoreLog a -> FilePath
@@ -151,11 +157,13 @@ storeLogFilePath = \case
 
 closeStoreLog :: StoreLog a -> IO ()
 closeStoreLog = \case
-  WriteStoreLog _ h -> hClose h
-  ReadStoreLog _ h -> hClose h
+  WriteStoreLog _ h -> close_ h
+  ReadStoreLog _ h -> close_ h
+  where
+    close_ h = hClose h `catchAny` \e -> logError ("STORE: closeStoreLog, error closing, " <> tshow e)
 
 writeStoreLogRecord :: StrEncoding r => StoreLog 'WriteMode -> r -> IO ()
-writeStoreLogRecord (WriteStoreLog _ h) r = do
+writeStoreLogRecord (WriteStoreLog _ h) r = E.uninterruptibleMask_ $ do
   B.hPut h $ strEncode r `B.snoc` '\n' -- hPutStrLn makes write non-atomic for length > 1024
   hFlush h
 
@@ -180,37 +188,47 @@ logDeleteNotifier s = writeStoreLogRecord s . DeleteNotifier
 logUpdateQueueTime :: StoreLog 'WriteMode -> QueueId -> RoundedSystemTime -> IO ()
 logUpdateQueueTime s qId t = writeStoreLogRecord s $ UpdateTime qId t
 
-readWriteStoreLog :: FilePath -> IO (Map RecipientId QueueRec, StoreLog 'WriteMode)
-readWriteStoreLog f = do
-  qs <- ifM (doesFileExist f) readQS (pure M.empty)
-  s <- openWriteStoreLog f
-  writeQueues s qs
-  pure (qs, s)
+readWriteStoreLog :: (FilePath -> s -> IO ()) -> (StoreLog 'WriteMode -> s -> IO ()) -> FilePath -> s -> IO (StoreLog 'WriteMode)
+readWriteStoreLog readStore writeStore f st =
+  ifM
+    (doesFileExist tempBackup)
+    (useTempBackup >> readWriteLog)
+    (ifM (doesFileExist f) readWriteLog (writeLog "creating store log..."))
   where
-    readQS = readQueues f <* renameFile f (f <> ".bak")
+    f' = T.pack f
+    tempBackup = f <> ".start"
+    useTempBackup = do
+      -- preserve current file, use temp backup
+      logWarn $ "Server terminated abnormally on last start, restoring state from " <> T.pack tempBackup
+      whenM (doesFileExist f) $ do
+        renameFile f (f <> ".bak")
+        logInfo $ "preserved incomplete state " <> f' <> " as " <> (f' <> ".bak")
+      renameFile tempBackup f
+    readWriteLog = do
+      -- log backup is made in two steps to mitigate the crash during the compacting.
+      -- Temporary backup file .start will be used when it is present.
+      readStore f st
+      renameFile f tempBackup -- 1) make temp backup
+      s <- writeLog "compacting store log (do not terminate)..." -- 2) save state
+      renameBackup -- 3) timed backup
+      pure s
+    writeLog msg = do
+      s <- openWriteStoreLog f
+      logInfo msg
+      writeStore s st
+      pure s
+    renameBackup = do
+      ts <- getCurrentTime
+      let timedBackup = f <> "." <> iso8601Show ts
+      renameFile tempBackup timedBackup
+      logInfo $ "original state preserved as " <> T.pack timedBackup
 
-writeQueues :: StoreLog 'WriteMode -> Map RecipientId QueueRec -> IO ()
-writeQueues s = mapM_ $ \q -> when (active q) $ logCreateQueue s q
+writeQueueStore :: STMQueueStore s => StoreLog 'WriteMode -> s -> IO ()
+writeQueueStore s st = readTVarIO (activeMsgQueues st) >>= mapM_ writeQueue . M.assocs
   where
+    writeQueue (rId, QRRecipient q) =
+      readTVarIO (queueRec' q) >>= \case
+        Just q' -> when (active q') $ logCreateQueue s q' -- TODO we should log suspended queues when we use them
+        Nothing -> atomically $ TM.delete rId $ queues' st
+    writeQueue _ = pure ()
     active QueueRec {status} = status == QueueActive
-
-readQueues :: FilePath -> IO (Map RecipientId QueueRec)
-readQueues f = foldM processLine M.empty . LB.lines =<< LB.readFile f
-  where
-    processLine :: Map RecipientId QueueRec -> LB.ByteString -> IO (Map RecipientId QueueRec)
-    processLine m s' = case strDecode $ trimCR s of
-      Right r -> pure $ procLogRecord r
-      Left e -> printError e $> m
-      where
-        s = LB.toStrict s'
-        procLogRecord :: StoreLogRecord -> Map RecipientId QueueRec
-        procLogRecord = \case
-          CreateQueue q -> M.insert (recipientId q) q m
-          SecureQueue qId sKey -> M.adjust (\q -> q {senderKey = Just sKey}) qId m
-          AddNotifier qId ntfCreds -> M.adjust (\q -> q {notifier = Just ntfCreds}) qId m
-          SuspendQueue qId -> M.adjust (\q -> q {status = QueueOff}) qId m
-          DeleteQueue qId -> M.delete qId m
-          DeleteNotifier qId -> M.adjust (\q -> q {notifier = Nothing}) qId m
-          UpdateTime qId t -> M.adjust (\q -> q {updatedAt = Just t}) qId m
-        printError :: String -> IO ()
-        printError e = B.putStrLn $ "Error parsing log: " <> B.pack e <> " - " <> s

src/Simplex/Messaging/Server/StoreLog/Types.hs

@@ -0,0 +1,13 @@
+{-# LANGUAGE DataKinds #-}
+{-# LANGUAGE GADTs #-}
+{-# LANGUAGE KindSignatures #-}
+
+module Simplex.Messaging.Server.StoreLog.Types where
+
+import System.IO (Handle, IOMode (..))
+
+-- | opaque container for file handle with a type-safe IOMode
+-- constructors are not exported, openWriteStoreLog and openReadStoreLog should be used instead
+data StoreLog (a :: IOMode) where
+  ReadStoreLog :: FilePath -> Handle -> StoreLog 'ReadMode
+  WriteStoreLog :: FilePath -> Handle -> StoreLog 'WriteMode

src/Simplex/Messaging/TMap.hs

@@ -9,6 +9,7 @@ module Simplex.Messaging.TMap
     member,
     memberIO,
     insert,
+    insertM,
     delete,
     lookupInsert,
     lookupDelete,
@@ -62,6 +63,10 @@ insert :: Ord k => k -> a -> TMap k a -> STM ()
 insert k v m = modifyTVar' m $ M.insert k v
 {-# INLINE insert #-}
 
+insertM :: Ord k => k -> STM a -> TMap k a -> STM ()
+insertM k f m = modifyTVar' m . M.insert k =<< f
+{-# INLINE insertM #-}
+
 delete :: Ord k => k -> TMap k a -> STM ()
 delete k m = modifyTVar' m $ M.delete k
 {-# INLINE delete #-}

src/Simplex/Messaging/Transport.hs

@@ -47,6 +47,8 @@ module Simplex.Messaging.Transport
     authCmdsSMPVersion,
     sendingProxySMPVersion,
     sndAuthKeySMPVersion,
+    deletedEventSMPVersion,
+    encryptedBlockSMPVersion,
     simplexMQVersion,
     smpBlockSize,
     TransportConfig (..),
@@ -64,13 +66,15 @@ module Simplex.Messaging.Transport
     ALPN,
     connectTLS,
     closeTLS,
-    supportedParameters,
+    defaultSupportedParams,
+    defaultSupportedParamsHTTPS,
     withTlsUnique,
 
     -- * SMP transport
     THandle (..),
     THandleParams (..),
     THandleAuth (..),
+    TSbChainKeys (..),
     TransportError (..),
     HandshakeError (..),
     smpServerHandshake,
@@ -84,8 +88,10 @@ module Simplex.Messaging.Transport
 where
 
 import Control.Applicative (optional)
+import Control.Concurrent.STM
 import Control.Monad (forM, (<$!>))
 import Control.Monad.Except
+import Control.Monad.IO.Class
 import Control.Monad.Trans.Except (throwE)
 import qualified Data.Aeson.TH as J
 import Data.Attoparsec.ByteString.Char8 (Parser)
@@ -97,6 +103,8 @@ import qualified Data.ByteString.Char8 as B
 import qualified Data.ByteString.Lazy.Char8 as LB
 import Data.Default (def)
 import Data.Functor (($>))
+import Data.Tuple (swap)
+import Data.Typeable (Typeable)
 import Data.Version (showVersion)
 import Data.Word (Word16)
 import qualified Data.X509 as X
@@ -130,6 +138,10 @@ smpBlockSize = 16384
 -- 5 - basic auth for SMP servers (11/12/2022)
 -- 6 - allow creating queues without subscribing (9/10/2023)
 -- 7 - support authenticated encryption to verify senders' commands, imply but do NOT send session ID in signed part (4/30/2024)
+-- 8 - SMP proxy for sender commands
+-- 9 - faster handshake: SKEY command for sender to secure queue
+-- 10 - DELD event to subscriber when queue is deleted via another connnection
+-- 11 - additional encryption of transport blocks with forward secrecy (9/14/2024)
 
 data SMPVersion
 
@@ -160,14 +172,20 @@ sendingProxySMPVersion = VersionSMP 8
 sndAuthKeySMPVersion :: VersionSMP
 sndAuthKeySMPVersion = VersionSMP 9
 
+deletedEventSMPVersion :: VersionSMP
+deletedEventSMPVersion = VersionSMP 10
+
+encryptedBlockSMPVersion :: VersionSMP
+encryptedBlockSMPVersion = VersionSMP 11
+
 currentClientSMPRelayVersion :: VersionSMP
-currentClientSMPRelayVersion = VersionSMP 9
+currentClientSMPRelayVersion = VersionSMP 11
 
 legacyServerSMPRelayVersion :: VersionSMP
 legacyServerSMPRelayVersion = VersionSMP 6
 
 currentServerSMPRelayVersion :: VersionSMP
-currentServerSMPRelayVersion = VersionSMP 9
+currentServerSMPRelayVersion = VersionSMP 11
 
 -- Max SMP protocol version to be used in e2e encrypted
 -- connection between client and server, as defined by SMP proxy.
@@ -205,7 +223,7 @@ data TransportConfig = TransportConfig
     transportTimeout :: Maybe Int
   }
 
-class Transport c where
+class Typeable c => Transport c where
   transport :: ATransport
   transport = ATransport (TProxy @c)
 
@@ -303,8 +321,8 @@ closeTLS ctx =
     `E.finally` T.contextClose ctx
     `catchAll_` pure ()
 
-supportedParameters :: T.Supported
-supportedParameters =
+defaultSupportedParams :: T.Supported
+defaultSupportedParams =
   def
     { T.supportedVersions = [T.TLS13, T.TLS12],
       T.supportedCiphers =
@@ -312,8 +330,29 @@ supportedParameters =
           TE.cipher_ECDHE_ECDSA_CHACHA20POLY1305_SHA256 -- for TLS12
         ],
       T.supportedHashSignatures = [(T.HashIntrinsic, T.SignatureEd448), (T.HashIntrinsic, T.SignatureEd25519)],
-      T.supportedSecureRenegotiation = False,
-      T.supportedGroups = [T.X448, T.X25519]
+      T.supportedGroups = [T.X448, T.X25519],
+      T.supportedSecureRenegotiation = False
+    }
+
+-- | A selection of extra parameters to accomodate browser chains
+defaultSupportedParamsHTTPS :: T.Supported
+defaultSupportedParamsHTTPS =
+  defaultSupportedParams
+    { T.supportedCiphers = TE.ciphersuite_strong,
+      T.supportedGroups = [T.X25519, T.X448, T.FFDHE4096, T.FFDHE6144, T.FFDHE8192, T.P521],
+      T.supportedHashSignatures =
+        [ (T.HashIntrinsic, T.SignatureEd448),
+          (T.HashIntrinsic, T.SignatureEd25519),
+          (T.HashSHA256, T.SignatureECDSA),
+          (T.HashSHA384, T.SignatureECDSA),
+          (T.HashSHA512, T.SignatureECDSA),
+          (T.HashIntrinsic, T.SignatureRSApssRSAeSHA512),
+          (T.HashIntrinsic, T.SignatureRSApssRSAeSHA384),
+          (T.HashIntrinsic, T.SignatureRSApssRSAeSHA256),
+          (T.HashSHA512, T.SignatureRSA),
+          (T.HashSHA384, T.SignatureRSA),
+          (T.HashSHA256, T.SignatureRSA)
+        ]
     }
 
 instance Transport TLS where
@@ -368,6 +407,8 @@ data THandleParams v p = THandleParams
     -- | do NOT send session ID in transmission, but include it into signed message
     -- based on protocol version
     implySessId :: Bool,
+    -- -- | additional block encryption
+    encryptBlock :: Maybe TSbChainKeys,
     -- | send multiple transmissions in a single block
     -- based on protocol version
     batch :: Bool
@@ -386,6 +427,11 @@ data THandleAuth (p :: TransportPeer) where
     } ->
     THandleAuth 'TServer
 
+data TSbChainKeys = TSbChainKeys
+  { sndKey :: TVar C.SbChainKey,
+    rcvKey :: TVar C.SbChainKey
+  }
+
 -- | TLS-unique channel binding
 type SessionId = ByteString
 
@@ -488,16 +534,26 @@ instance Encoding TransportError where
 
 -- | Pad and send block to SMP transport.
 tPutBlock :: Transport c => THandle v c p -> ByteString -> IO (Either TransportError ())
-tPutBlock THandle {connection = c, params = THandleParams {blockSize}} block =
-  bimapM (const $ pure TELargeMsg) (cPut c) $
-    C.pad block blockSize
+tPutBlock THandle {connection = c, params = THandleParams {blockSize, encryptBlock}} block = do
+  block_ <- case encryptBlock of
+    Just TSbChainKeys {sndKey} -> do
+      (sk, nonce) <- atomically $ stateTVar sndKey C.sbcHkdf
+      pure $ C.sbEncrypt sk nonce block (blockSize - 16)
+    Nothing -> pure $ C.pad block blockSize
+  bimapM (const $ pure TELargeMsg) (cPut c) block_
 
 -- | Receive block from SMP transport.
 tGetBlock :: Transport c => THandle v c p -> IO (Either TransportError ByteString)
-tGetBlock THandle {connection = c, params = THandleParams {blockSize}} = do
+tGetBlock THandle {connection = c, params = THandleParams {blockSize, encryptBlock}} = do
   msg <- cGet c blockSize
   if B.length msg == blockSize
-    then pure . first (const TELargeMsg) $ C.unPad msg
+    then
+      first (const TELargeMsg) <$>
+        case encryptBlock of
+          Just TSbChainKeys {rcvKey} -> do
+            (sk, nonce) <- atomically $ stateTVar rcvKey C.sbcHkdf
+            pure $ C.sbDecrypt sk nonce msg
+          Nothing -> pure $ C.unPad msg
     else ioe_EOF
 
 -- | Server SMP transport handshake.
@@ -516,7 +572,7 @@ smpServerHandshake serverSignKey c (k, pk) kh smpVRange = do
           throwE $ TEHandshake IDENTITY
       | otherwise ->
           case compatibleVRange' smpVersionRange v of
-            Just (Compatible vr) -> pure $ smpTHandleServer th v vr pk k'
+            Just (Compatible vr) -> liftIO $ smpTHandleServer th v vr pk k'
             Nothing -> throwE TEVersion
 
 -- | Client SMP transport handshake.
@@ -540,23 +596,36 @@ smpClientHandshake c ks_ keyHash@(C.KeyHash kh) smpVRange = do
             (,certKey) <$> (C.x509ToPublic (pubKey, []) >>= C.pubKey)
         let v = maxVersion vr
         sendHandshake th $ ClientHandshake {smpVersion = v, keyHash, authPubKey = fst <$> ks_}
-        pure $ smpTHandleClient th v vr (snd <$> ks_) ck_
+        liftIO $ smpTHandleClient th v vr (snd <$> ks_) ck_
       Nothing -> throwE TEVersion
 
-smpTHandleServer :: forall c. THandleSMP c 'TServer -> VersionSMP -> VersionRangeSMP -> C.PrivateKeyX25519 -> Maybe C.PublicKeyX25519 -> THandleSMP c 'TServer
-smpTHandleServer th v vr pk k_ =
-  let thAuth = THAuthServer {serverPrivKey = pk, sessSecret' = (`C.dh'` pk) <$!> k_}
-   in smpTHandle_ th v vr (Just thAuth)
+smpTHandleServer :: forall c. THandleSMP c 'TServer -> VersionSMP -> VersionRangeSMP -> C.PrivateKeyX25519 -> Maybe C.PublicKeyX25519 -> IO (THandleSMP c 'TServer)
+smpTHandleServer th v vr pk k_ = do
+  let thAuth = Just THAuthServer {serverPrivKey = pk, sessSecret' = (`C.dh'` pk) <$!> k_}
+  be <- blockEncryption th v thAuth
+  pure $ smpTHandle_ th v vr thAuth $ uncurry TSbChainKeys <$> be
 
-smpTHandleClient :: forall c. THandleSMP c 'TClient -> VersionSMP -> VersionRangeSMP -> Maybe C.PrivateKeyX25519 -> Maybe (C.PublicKeyX25519, (X.CertificateChain, X.SignedExact X.PubKey)) -> THandleSMP c 'TClient
-smpTHandleClient th v vr pk_ ck_ =
+smpTHandleClient :: forall c. THandleSMP c 'TClient -> VersionSMP -> VersionRangeSMP -> Maybe C.PrivateKeyX25519 -> Maybe (C.PublicKeyX25519, (X.CertificateChain, X.SignedExact X.PubKey)) -> IO (THandleSMP c 'TClient)
+smpTHandleClient th v vr pk_ ck_ = do
   let thAuth = (\(k, ck) -> THAuthClient {serverPeerPubKey = k, serverCertKey = forceCertChain ck, sessSecret = C.dh' k <$!> pk_}) <$!> ck_
-   in smpTHandle_ th v vr thAuth
+  be <- blockEncryption th v thAuth
+  -- swap is needed to use client's sndKey as server's rcvKey and vice versa
+  pure $ smpTHandle_ th v vr thAuth $ uncurry TSbChainKeys . swap <$> be
 
-smpTHandle_ :: forall c p. THandleSMP c p -> VersionSMP -> VersionRangeSMP -> Maybe (THandleAuth p) -> THandleSMP c p
-smpTHandle_ th@THandle {params} v vr thAuth =
+blockEncryption :: THandleSMP c p -> VersionSMP -> Maybe (THandleAuth p) -> IO (Maybe (TVar C.SbChainKey, TVar C.SbChainKey))
+blockEncryption THandle {params = THandleParams {sessionId}} v = \case
+  Just thAuth | v >= encryptedBlockSMPVersion -> case thAuth of
+    THAuthClient {sessSecret} -> be sessSecret
+    THAuthServer {sessSecret'} -> be sessSecret'
+  _ -> pure Nothing
+  where
+    be :: Maybe C.DhSecretX25519 -> IO (Maybe (TVar C.SbChainKey, TVar C.SbChainKey))
+    be = mapM $ \(C.DhSecretX25519 secret) -> bimapM newTVarIO newTVarIO $ C.sbcInit sessionId secret
+
+smpTHandle_ :: forall c p. THandleSMP c p -> VersionSMP -> VersionRangeSMP -> Maybe (THandleAuth p) -> Maybe TSbChainKeys -> THandleSMP c p
+smpTHandle_ th@THandle {params} v vr thAuth encryptBlock =
   -- TODO drop SMP v6: make thAuth non-optional
-  let params' = params {thVersion = v, thServerVRange = vr, thAuth, implySessId = v >= authCmdsSMPVersion}
+  let params' = params {thVersion = v, thServerVRange = vr, thAuth, implySessId = v >= authCmdsSMPVersion, encryptBlock}
    in (th :: THandleSMP c p) {params = params'}
 
 {-# INLINE forceCertChain #-}
@@ -588,6 +657,7 @@ smpTHandle c = THandle {connection = c, params}
           thVersion = v,
           thAuth = Nothing,
           implySessId = False,
+          encryptBlock = Nothing,
           batch = True
         }
 

src/Simplex/Messaging/Transport/Client.hs

@@ -13,9 +13,13 @@ module Simplex.Messaging.Transport.Client
     defaultSMPPort,
     defaultTcpConnectTimeout,
     defaultTransportClientConfig,
+    defaultSocksProxyWithAuth,
     defaultSocksProxy,
+    defaultSocksHost,
     TransportClientConfig (..),
-    SocksProxy,
+    SocksProxy (..),
+    SocksProxyWithAuth (..),
+    SocksAuth (..),
     TransportHost (..),
     TransportHosts (..),
     TransportHosts_ (..),
@@ -23,7 +27,7 @@ module Simplex.Messaging.Transport.Client
   )
 where
 
-import Control.Applicative (optional)
+import Control.Applicative (optional, (<|>))
 import Control.Logger.Simple (logError)
 import Control.Monad (when)
 import Data.Aeson (FromJSON (..), ToJSON (..))
@@ -79,7 +83,7 @@ instance StrEncoding TransportHost where
   strP =
     A.choice
       [ THIPv4 <$> ((,,,) <$> ipNum <*> ipNum <*> ipNum <*> A.decimal),
-        maybe (Left "bad IPv6") (Right . THIPv6 . fromIPv6w) . readMaybe . B.unpack <$?> A.takeWhile1 (\c -> isHexDigit c || c == ':'),
+        maybe (Left "bad IPv6") (Right . THIPv6 . fromIPv6w) . readMaybe . B.unpack <$?> ipv6StrP,
         THOnionHost <$> ((<>) <$> A.takeWhile (\c -> isAsciiLower c || isDigit c) <*> A.string ".onion"),
         THDomainName . B.unpack <$> (notOnion <$?> A.takeWhile1 (A.notInClass ":#,;/ \n\r\t"))
       ]
@@ -87,6 +91,9 @@ instance StrEncoding TransportHost where
       ipNum = validIP <$?> (A.decimal <* A.char '.')
       validIP :: Int -> Either String Word8
       validIP n = if 0 <= n && n <= 255 then Right $ fromIntegral n else Left "invalid IP address"
+      ipv6StrP =
+        A.char '[' *> A.takeWhile1 (/= ']') <* A.char ']'
+          <|> A.takeWhile1 (\c -> isHexDigit c || c == ':')
       notOnion s = if ".onion" `B.isSuffixOf` s then Left "invalid onion host" else Right s
 
 instance ToJSON TransportHost where
@@ -118,7 +125,8 @@ data TransportClientConfig = TransportClientConfig
     tcpKeepAlive :: Maybe KeepAliveOpts,
     logTLSErrors :: Bool,
     clientCredentials :: Maybe (X.CertificateChain, T.PrivKey),
-    alpn :: Maybe [ALPN]
+    alpn :: Maybe [ALPN],
+    useSNI :: Bool
   }
   deriving (Eq, Show)
 
@@ -127,23 +135,23 @@ defaultTcpConnectTimeout :: Int
 defaultTcpConnectTimeout = 25_000_000
 
 defaultTransportClientConfig :: TransportClientConfig
-defaultTransportClientConfig = TransportClientConfig Nothing defaultTcpConnectTimeout (Just defaultKeepAliveOpts) True Nothing Nothing
+defaultTransportClientConfig = TransportClientConfig Nothing defaultTcpConnectTimeout (Just defaultKeepAliveOpts) True Nothing Nothing True
 
 clientTransportConfig :: TransportClientConfig -> TransportConfig
 clientTransportConfig TransportClientConfig {logTLSErrors} =
   TransportConfig {logTLSErrors, transportTimeout = Nothing}
 
 -- | Connect to passed TCP host:port and pass handle to the client.
-runTransportClient :: Transport c => TransportClientConfig -> Maybe ByteString -> TransportHost -> ServiceName -> Maybe C.KeyHash -> (c -> IO a) -> IO a
-runTransportClient = runTLSTransportClient supportedParameters Nothing
+runTransportClient :: Transport c => TransportClientConfig -> Maybe SocksCredentials -> TransportHost -> ServiceName -> Maybe C.KeyHash -> (c -> IO a) -> IO a
+runTransportClient = runTLSTransportClient defaultSupportedParams Nothing
 
-runTLSTransportClient :: Transport c => T.Supported -> Maybe XS.CertificateStore -> TransportClientConfig -> Maybe ByteString -> TransportHost -> ServiceName -> Maybe C.KeyHash -> (c -> IO a) -> IO a
-runTLSTransportClient tlsParams caStore_ cfg@TransportClientConfig {socksProxy, tcpKeepAlive, clientCredentials, alpn} proxyUsername host port keyHash client = do
+runTLSTransportClient :: Transport c => T.Supported -> Maybe XS.CertificateStore -> TransportClientConfig -> Maybe SocksCredentials -> TransportHost -> ServiceName -> Maybe C.KeyHash -> (c -> IO a) -> IO a
+runTLSTransportClient tlsParams caStore_ cfg@TransportClientConfig {socksProxy, tcpKeepAlive, clientCredentials, alpn, useSNI} socksCreds host port keyHash client = do
   serverCert <- newEmptyTMVarIO
   let hostName = B.unpack $ strEncode host
-      clientParams = mkTLSClientParams tlsParams caStore_ hostName port keyHash clientCredentials alpn serverCert
+      clientParams = mkTLSClientParams tlsParams caStore_ hostName port keyHash clientCredentials alpn useSNI serverCert
       connectTCP = case socksProxy of
-        Just proxy -> connectSocksClient proxy proxyUsername (hostAddr host)
+        Just proxy -> connectSocksClient proxy socksCreds (hostAddr host)
         _ -> connectTCPClient hostName
   c <- do
     sock <- connectTCP port
@@ -191,45 +199,76 @@ connectTCPClient host port = withSocketsDo $ resolve >>= tryOpen err
 defaultSMPPort :: PortNumber
 defaultSMPPort = 5223
 
-connectSocksClient :: SocksProxy -> Maybe ByteString -> SocksHostAddress -> ServiceName -> IO Socket
-connectSocksClient (SocksProxy addr) proxyUsername hostAddr _port = do
+connectSocksClient :: SocksProxy -> Maybe SocksCredentials -> SocksHostAddress -> ServiceName -> IO Socket
+connectSocksClient (SocksProxy addr) socksCreds hostAddr _port = do
   let port = if null _port then defaultSMPPort else fromMaybe defaultSMPPort $ readMaybe _port
-  fst <$> case proxyUsername of
-    Just username -> socksConnectAuth (defaultSocksConf addr) (SocksAddress hostAddr port) (SocksCredentials username "")
+  fst <$> case socksCreds of
+    Just creds -> socksConnectAuth (defaultSocksConf addr) (SocksAddress hostAddr port) creds
     _ -> socksConnect (defaultSocksConf addr) (SocksAddress hostAddr port)
 
-defaultSocksHost :: HostAddress
-defaultSocksHost = tupleToHostAddress (127, 0, 0, 1)
+defaultSocksHost :: (Word8, Word8, Word8, Word8)
+defaultSocksHost = (127, 0, 0, 1)
+
+defaultSocksProxyWithAuth :: SocksProxyWithAuth
+defaultSocksProxyWithAuth = SocksProxyWithAuth SocksIsolateByAuth defaultSocksProxy
 
 defaultSocksProxy :: SocksProxy
-defaultSocksProxy = SocksProxy $ SockAddrInet 9050 defaultSocksHost
+defaultSocksProxy = SocksProxy $ SockAddrInet 9050 $ tupleToHostAddress defaultSocksHost
 
 newtype SocksProxy = SocksProxy SockAddr
   deriving (Eq)
 
+data SocksProxyWithAuth = SocksProxyWithAuth SocksAuth SocksProxy
+  deriving (Eq, Show)
+
+data SocksAuth
+  = SocksAuthUsername {username :: ByteString, password :: ByteString}
+  | SocksAuthNull
+  | SocksIsolateByAuth -- this is default
+  deriving (Eq, Show)
+
 instance Show SocksProxy where show (SocksProxy addr) = show addr
 
 instance StrEncoding SocksProxy where
   strEncode = B.pack . show
   strP = do
-    host <- maybe defaultSocksHost tupleToHostAddress <$> optional ipv4P
+    host <- fromMaybe (THIPv4 defaultSocksHost) <$> optional strP
     port <- fromMaybe 9050 <$> optional (A.char ':' *> (fromInteger <$> A.decimal))
-    pure . SocksProxy $ SockAddrInet port host
+    SocksProxy <$> socksAddr port host
     where
-      ipv4P = (,,,) <$> ipNum <*> ipNum <*> ipNum <*> A.decimal
-      ipNum = A.decimal <* A.char '.'
+      socksAddr port = \case
+        THIPv4 addr -> pure $ SockAddrInet port $ tupleToHostAddress addr
+        THIPv6 addr -> pure $ SockAddrInet6 port 0 addr 0
+        _ -> fail "SOCKS5 host should be IPv4 or IPv6 address"
 
-instance ToJSON SocksProxy where
+instance StrEncoding SocksProxyWithAuth where
+  strEncode (SocksProxyWithAuth auth proxy) = strEncode auth <> strEncode proxy
+  strP = SocksProxyWithAuth <$> strP <*> strP
+
+instance ToJSON SocksProxyWithAuth where
   toJSON = strToJSON
   toEncoding = strToJEncoding
 
-instance FromJSON SocksProxy where
-  parseJSON = strParseJSON "SocksProxy"
+instance FromJSON SocksProxyWithAuth where
+  parseJSON = strParseJSON "SocksProxyWithAuth"
 
-mkTLSClientParams :: T.Supported -> Maybe XS.CertificateStore -> HostName -> ServiceName -> Maybe C.KeyHash -> Maybe (X.CertificateChain, T.PrivKey) -> Maybe [ALPN] -> TMVar X.CertificateChain -> T.ClientParams
-mkTLSClientParams supported caStore_ host port cafp_ clientCreds_ alpn_ serverCerts =
+instance StrEncoding SocksAuth where
+  strEncode = \case
+    SocksAuthUsername {username, password} -> username <> ":" <> password <> "@"
+    SocksAuthNull -> "@"
+    SocksIsolateByAuth -> ""
+  strP = usernameP <|> (SocksAuthNull <$ A.char '@') <|> pure SocksIsolateByAuth
+    where
+      usernameP = do
+        username <- A.takeTill (== ':') <* A.char ':'
+        password <- A.takeTill (== '@') <* A.char '@'
+        pure SocksAuthUsername {username, password}
+
+mkTLSClientParams :: T.Supported -> Maybe XS.CertificateStore -> HostName -> ServiceName -> Maybe C.KeyHash -> Maybe (X.CertificateChain, T.PrivKey) -> Maybe [ALPN] -> Bool -> TMVar X.CertificateChain -> T.ClientParams
+mkTLSClientParams supported caStore_ host port cafp_ clientCreds_ alpn_ sni serverCerts =
   (T.defaultParamsClient host p)
-    { T.clientShared = def {T.sharedCAStore = fromMaybe (T.sharedCAStore def) caStore_},
+    { T.clientUseServerNameIndication = sni,
+      T.clientShared = def {T.sharedCAStore = fromMaybe (T.sharedCAStore def) caStore_},
       T.clientHooks =
         def
           { T.onServerCertificate = onServerCert,

src/Simplex/Messaging/Transport/Credentials.hs

@@ -32,8 +32,8 @@ import qualified Time.System as Hourglass
 --   leaf <- genCredentials (Just ca) (0, 1) "Entity" -- session-signing cert
 --   pure $ tlsCredentials (leaf :| [ca])
 -- @
-tlsCredentials :: NonEmpty Credentials -> (C.KeyHash, TLS.Credentials)
-tlsCredentials credentials = (C.KeyHash rootFP, TLS.Credentials [(X509.CertificateChain certs, privateToTls $ snd leafKey)])
+tlsCredentials :: NonEmpty Credentials -> (C.KeyHash, TLS.Credential)
+tlsCredentials credentials = (C.KeyHash rootFP, (X509.CertificateChain certs, privateToTls $ snd leafKey))
   where
     Fingerprint rootFP = getFingerprint root X509.HashSHA256
     leafKey = fst $ L.head credentials

src/Simplex/Messaging/Transport/HTTP2/Client.hs

@@ -11,7 +11,6 @@ import Control.Concurrent.Async
 import Control.Exception (IOException, try)
 import qualified Control.Exception as E
 import Control.Monad
-import Data.ByteString.Char8 (ByteString)
 import Data.Functor (($>))
 import Data.Time (UTCTime, getCurrentTime)
 import qualified Data.X509 as X
@@ -20,6 +19,7 @@ import Network.HPACK (BufferSize)
 import Network.HTTP2.Client (ClientConfig (..), Request, Response)
 import qualified Network.HTTP2.Client as H
 import Network.Socket (HostName, ServiceName)
+import Network.Socks5 (SocksCredentials)
 import qualified Network.TLS as T
 import Numeric.Natural (Natural)
 import qualified Simplex.Messaging.Crypto as C
@@ -78,7 +78,8 @@ defaultHTTP2ClientConfig =
             tcpKeepAlive = Nothing,
             logTLSErrors = True,
             clientCredentials = Nothing,
-            alpn = Nothing
+            alpn = Nothing,
+            useSNI = False
           },
       bufferSize = defaultHTTP2BufferSize,
       bodyHeadSize = 16384,
@@ -91,10 +92,10 @@ data HTTP2ClientError = HCResponseTimeout | HCNetworkError | HCIOError IOExcepti
 getHTTP2Client :: HostName -> ServiceName -> Maybe XS.CertificateStore -> HTTP2ClientConfig -> IO () -> IO (Either HTTP2ClientError HTTP2Client)
 getHTTP2Client host port = getVerifiedHTTP2Client Nothing (THDomainName host) port Nothing
 
-getVerifiedHTTP2Client :: Maybe ByteString -> TransportHost -> ServiceName -> Maybe C.KeyHash -> Maybe XS.CertificateStore -> HTTP2ClientConfig -> IO () -> IO (Either HTTP2ClientError HTTP2Client)
-getVerifiedHTTP2Client proxyUsername host port keyHash caStore config disconnected = getVerifiedHTTP2ClientWith config host port disconnected setup
+getVerifiedHTTP2Client :: Maybe SocksCredentials -> TransportHost -> ServiceName -> Maybe C.KeyHash -> Maybe XS.CertificateStore -> HTTP2ClientConfig -> IO () -> IO (Either HTTP2ClientError HTTP2Client)
+getVerifiedHTTP2Client socksCreds host port keyHash caStore config disconnected = getVerifiedHTTP2ClientWith config host port disconnected setup
   where
-    setup = runHTTP2Client (suportedTLSParams config) caStore (transportConfig config) (bufferSize config) proxyUsername host port keyHash
+    setup = runHTTP2Client (suportedTLSParams config) caStore (transportConfig config) (bufferSize config) socksCreds host port keyHash
 
 attachHTTP2Client :: HTTP2ClientConfig -> TransportHost -> ServiceName -> IO () -> Int -> TLS -> IO (Either HTTP2ClientError HTTP2Client)
 attachHTTP2Client config host port disconnected bufferSize tls = getVerifiedHTTP2ClientWith config host port disconnected setup
@@ -178,11 +179,11 @@ sendRequestDirect HTTP2Client {client_ = HClient {config, disconnected}, sendReq
 http2RequestTimeout :: HTTP2ClientConfig -> Maybe Int -> Int
 http2RequestTimeout HTTP2ClientConfig {connTimeout} = maybe connTimeout (connTimeout +)
 
-runHTTP2Client :: forall a. T.Supported -> Maybe XS.CertificateStore -> TransportClientConfig -> BufferSize -> Maybe ByteString -> TransportHost -> ServiceName -> Maybe C.KeyHash -> (TLS -> H.Client a) -> IO a
-runHTTP2Client tlsParams caStore tcConfig bufferSize proxyUsername host port keyHash = runHTTP2ClientWith bufferSize host setup
+runHTTP2Client :: forall a. T.Supported -> Maybe XS.CertificateStore -> TransportClientConfig -> BufferSize -> Maybe SocksCredentials -> TransportHost -> ServiceName -> Maybe C.KeyHash -> (TLS -> H.Client a) -> IO a
+runHTTP2Client tlsParams caStore tcConfig bufferSize socksCreds host port keyHash = runHTTP2ClientWith bufferSize host setup
   where
     setup :: (TLS -> IO a) -> IO a
-    setup = runTLSTransportClient tlsParams caStore tcConfig proxyUsername host port keyHash
+    setup = runTLSTransportClient tlsParams caStore tcConfig socksCreds host port keyHash
 
 runHTTP2ClientWith :: forall a. BufferSize -> TransportHost -> ((TLS -> IO a) -> IO a) -> (TLS -> H.Client a) -> IO a
 runHTTP2ClientWith bufferSize host setup client = setup $ \tls -> withHTTP2 bufferSize (run tls) (pure ()) tls

src/Simplex/Messaging/Transport/HTTP2/Server.hs

@@ -15,7 +15,7 @@ import Numeric.Natural (Natural)
 import Simplex.Messaging.Server.Expiration
 import Simplex.Messaging.Transport (ALPN, SessionId, TLS, closeConnection, tlsALPN, tlsUniq)
 import Simplex.Messaging.Transport.HTTP2
-import Simplex.Messaging.Transport.Server (TransportServerConfig (..), loadSupportedTLSServerParams, runTransportServer)
+import Simplex.Messaging.Transport.Server (ServerCredentials, TransportServerConfig (..), loadServerCredential, runTransportServer)
 import Simplex.Messaging.Util (threadDelay')
 import UnliftIO (finally)
 import UnliftIO.Concurrent (forkIO, killThread)
@@ -28,9 +28,7 @@ data HTTP2ServerConfig = HTTP2ServerConfig
     bufferSize :: BufferSize,
     bodyHeadSize :: Int,
     serverSupported :: T.Supported,
-    caCertificateFile :: FilePath,
-    privateKeyFile :: FilePath,
-    certificateFile :: FilePath,
+    https2Credentials :: ServerCredentials,
     transportConfig :: TransportServerConfig
   }
   deriving (Show)
@@ -49,13 +47,13 @@ data HTTP2Server = HTTP2Server
   }
 
 -- This server is for testing only, it processes all requests in a single queue.
-getHTTP2Server :: HTTP2ServerConfig -> IO HTTP2Server
-getHTTP2Server HTTP2ServerConfig {qSize, http2Port, bufferSize, bodyHeadSize, serverSupported, caCertificateFile, certificateFile, privateKeyFile, transportConfig} = do
-  tlsServerParams <- loadSupportedTLSServerParams serverSupported caCertificateFile certificateFile privateKeyFile (alpn transportConfig)
+getHTTP2Server :: HTTP2ServerConfig -> Maybe [ALPN] -> IO HTTP2Server
+getHTTP2Server HTTP2ServerConfig {qSize, http2Port, bufferSize, bodyHeadSize, serverSupported, https2Credentials, transportConfig} alpn_ = do
+  srvCreds <- loadServerCredential https2Credentials
   started <- newEmptyTMVarIO
   reqQ <- newTBQueueIO qSize
   action <- async $
-    runHTTP2Server started http2Port bufferSize tlsServerParams transportConfig Nothing (const $ pure ()) $ \sessionId sessionALPN r sendResponse -> do
+    runHTTP2Server started http2Port bufferSize serverSupported srvCreds alpn_ transportConfig Nothing (const $ pure ()) $ \sessionId sessionALPN r sendResponse -> do
       reqBody <- getHTTP2Body r bodyHeadSize
       atomically $ writeTBQueue reqQ HTTP2Request {sessionId, sessionALPN, request = r, reqBody, sendResponse}
   void . atomically $ takeTMVar started
@@ -64,10 +62,10 @@ getHTTP2Server HTTP2ServerConfig {qSize, http2Port, bufferSize, bodyHeadSize, se
 closeHTTP2Server :: HTTP2Server -> IO ()
 closeHTTP2Server = uninterruptibleCancel . action
 
-runHTTP2Server :: TMVar Bool -> ServiceName -> BufferSize -> T.ServerParams -> TransportServerConfig -> Maybe ExpirationConfig -> (SessionId -> IO ()) -> HTTP2ServerFunc -> IO ()
-runHTTP2Server started port bufferSize serverParams transportConfig expCfg_ clientFinished = runHTTP2ServerWith_ expCfg_ clientFinished bufferSize setup
+runHTTP2Server :: TMVar Bool -> ServiceName -> BufferSize -> T.Supported -> T.Credential -> Maybe [ALPN] -> TransportServerConfig -> Maybe ExpirationConfig -> (SessionId -> IO ()) -> HTTP2ServerFunc -> IO ()
+runHTTP2Server started port bufferSize srvSupported srvCreds alpn_ transportConfig expCfg_ clientFinished = runHTTP2ServerWith_ expCfg_ clientFinished bufferSize setup
   where
-    setup = runTransportServer started port serverParams transportConfig
+    setup = runTransportServer started port srvSupported srvCreds alpn_ transportConfig
 
 runHTTP2ServerWith :: BufferSize -> ((TLS -> IO ()) -> a) -> HTTP2ServerFunc -> a
 runHTTP2ServerWith = runHTTP2ServerWith_ Nothing (\_sessId -> pure ())

src/Simplex/Messaging/Transport/Server.hs

@@ -6,8 +6,11 @@
 
 module Simplex.Messaging.Transport.Server
   ( TransportServerConfig (..),
+    ServerCredentials (..),
+    AddHTTP,
     defaultTransportServerConfig,
     runTransportServerState,
+    runTransportServerState_,
     SocketState,
     newSocketState,
     runTransportServer,
@@ -15,11 +18,12 @@ module Simplex.Messaging.Transport.Server
     runLocalTCPServer,
     runTCPServerSocket,
     startTCPServer,
-    loadSupportedTLSServerParams,
-    loadTLSServerParams,
+    loadServerCredential,
+    supportedTLSServerParams,
+    supportedTLSServerParams_,
     loadFingerprint,
+    loadFileFingerprint,
     smpServerHandshake,
-    tlsServerCredentials,
   )
 where
 
@@ -31,7 +35,7 @@ import Data.Default (def)
 import Data.IntMap.Strict (IntMap)
 import qualified Data.IntMap.Strict as IM
 import Data.List (find)
-import Data.Maybe (fromJust, fromMaybe)
+import Data.Maybe (fromJust, fromMaybe, maybeToList)
 import qualified Data.X509 as X
 import Data.X509.Validation (Fingerprint (..))
 import qualified Data.X509.Validation as XV
@@ -52,18 +56,25 @@ import UnliftIO.STM
 data TransportServerConfig = TransportServerConfig
   { logTLSErrors :: Bool,
     tlsSetupTimeout :: Int,
-    transportTimeout :: Int,
-    alpn :: Maybe [ALPN]
+    transportTimeout :: Int
   }
   deriving (Eq, Show)
 
+data ServerCredentials = ServerCredentials
+  { caCertificateFile :: Maybe FilePath, -- CA certificate private key is not needed for initialization
+    privateKeyFile :: FilePath,
+    certificateFile :: FilePath
+  }
+  deriving (Show)
+
+type AddHTTP = Bool
+
 defaultTransportServerConfig :: TransportServerConfig
 defaultTransportServerConfig =
   TransportServerConfig
     { logTLSErrors = True,
       tlsSetupTimeout = 60000000,
-      transportTimeout = 40000000,
-      alpn = Nothing
+      transportTimeout = 40000000
     }
 
 serverTransportConfig :: TransportServerConfig -> TransportConfig
@@ -74,37 +85,41 @@ serverTransportConfig TransportServerConfig {logTLSErrors} =
 -- | Run transport server (plain TCP or WebSockets) on passed TCP port and signal when server started and stopped via passed TMVar.
 --
 -- All accepted connections are passed to the passed function.
-runTransportServer :: forall c. Transport c => TMVar Bool -> ServiceName -> T.ServerParams -> TransportServerConfig -> (c -> IO ()) -> IO ()
-runTransportServer started port params cfg server = do
+runTransportServer :: forall c. Transport c => TMVar Bool -> ServiceName -> T.Supported -> T.Credential -> Maybe [ALPN] -> TransportServerConfig -> (c -> IO ()) -> IO ()
+runTransportServer started port srvSupported srvCreds alpn_ cfg server = do
   ss <- newSocketState
-  runTransportServerState ss started port params cfg server
+  runTransportServerState ss started port srvSupported srvCreds alpn_ cfg server
 
-runTransportServerState :: forall c . Transport c => SocketState -> TMVar Bool -> ServiceName -> T.ServerParams -> TransportServerConfig -> (c -> IO ()) -> IO ()
-runTransportServerState ss started port = runTransportServerSocketState ss started (startTCPServer started Nothing port) (transportName (TProxy :: TProxy c))
+runTransportServerState :: forall c . Transport c => SocketState -> TMVar Bool -> ServiceName -> T.Supported -> T.Credential -> Maybe [ALPN] -> TransportServerConfig -> (c -> IO ()) -> IO ()
+runTransportServerState ss started port srvSupported srvCreds alpn_ cfg server = runTransportServerState_ ss started port srvSupported (const srvCreds) alpn_ cfg (const server)
+
+runTransportServerState_ :: forall c . Transport c => SocketState -> TMVar Bool -> ServiceName -> T.Supported -> (Maybe HostName -> T.Credential) -> Maybe [ALPN] -> TransportServerConfig -> (Socket -> c -> IO ()) -> IO ()
+runTransportServerState_ ss started port = runTransportServerSocketState ss started (startTCPServer started Nothing port) (transportName (TProxy :: TProxy c))
 
 -- | Run a transport server with provided connection setup and handler.
-runTransportServerSocket :: Transport a => TMVar Bool -> IO Socket -> String -> T.ServerParams -> TransportServerConfig -> (a -> IO ()) -> IO ()
-runTransportServerSocket started getSocket threadLabel serverParams cfg server = do
+runTransportServerSocket :: Transport a => TMVar Bool -> IO Socket -> String -> T.Credential -> T.ServerParams -> TransportServerConfig -> (a -> IO ()) -> IO ()
+runTransportServerSocket started getSocket threadLabel srvCreds srvParams cfg server = do
   ss <- newSocketState
-  runTransportServerSocketState ss started getSocket threadLabel serverParams cfg server
+  runTransportServerSocketState_ ss started getSocket threadLabel (const srvCreds) srvParams cfg (const server)
+
+runTransportServerSocketState :: Transport a => SocketState -> TMVar Bool -> IO Socket -> String -> T.Supported -> (Maybe HostName -> T.Credential) -> Maybe [ALPN] -> TransportServerConfig -> (Socket -> a -> IO ()) -> IO ()
+runTransportServerSocketState ss started getSocket threadLabel srvSupported srvCreds alpn_ =
+  runTransportServerSocketState_ ss started getSocket threadLabel srvCreds srvParams
+  where
+    srvParams = supportedTLSServerParams_ srvSupported srvCreds alpn_
 
 -- | Run a transport server with provided connection setup and handler.
-runTransportServerSocketState :: Transport a => SocketState -> TMVar Bool -> IO Socket -> String -> T.ServerParams -> TransportServerConfig -> (a -> IO ()) -> IO ()
-runTransportServerSocketState ss started getSocket threadLabel serverParams cfg server = do
+runTransportServerSocketState_ :: Transport a => SocketState -> TMVar Bool -> IO Socket -> String -> (Maybe HostName -> (X.CertificateChain, X.PrivKey)) -> T.ServerParams -> TransportServerConfig -> (Socket -> a -> IO ()) -> IO ()
+runTransportServerSocketState_ ss started getSocket threadLabel srvCreds srvParams cfg server = do
   labelMyThread $ "transport server for " <> threadLabel
   runTCPServerSocket ss started getSocket $ \conn ->
-    E.bracket (setup conn >>= maybe (fail "tls setup timeout") pure) closeConnection server
+    E.bracket (setup conn >>= maybe (fail "tls setup timeout") pure) closeConnection (server conn)
   where
     tCfg = serverTransportConfig cfg
     setup conn = timeout (tlsSetupTimeout cfg) $ do
       labelMyThread $ threadLabel <> "/setup"
-      tls <- connectTLS Nothing tCfg serverParams conn
-      getServerConnection tCfg (fst $ tlsServerCredentials serverParams) tls
-
-tlsServerCredentials :: T.ServerParams -> (X.CertificateChain, X.PrivKey)
-tlsServerCredentials serverParams = case T.sharedCredentials $ T.serverShared serverParams of
-  T.Credentials [creds] -> creds
-  _ -> error "server has more than one key"
+      tls <- connectTLS Nothing tCfg srvParams conn
+      getServerConnection tCfg (fst $ srvCreds Nothing) tls
 
 -- | Run TCP server without TLS
 runLocalTCPServer :: TMVar Bool -> ServiceName -> (Socket -> IO ()) -> IO ()
@@ -117,7 +132,7 @@ runTCPServerSocket :: SocketState -> TMVar Bool -> IO Socket -> (Socket -> IO ()
 runTCPServerSocket (accepted, gracefullyClosed, clients) started getSocket server =
   E.bracket getSocket (closeServer started clients) $ \sock ->
     forever . E.bracketOnError (safeAccept sock) (close . fst) $ \(conn, _peer) -> do
-      cId <- atomically $ stateTVar accepted $ \cId -> let cId' = cId + 1 in cId `seq` (cId', cId')
+      cId <- atomically $ stateTVar accepted $ \cId -> let cId' = cId + 1 in cId' `seq` (cId', cId')
       let closeConn _ = do
             atomically $ modifyTVar' clients $ IM.delete cId
             gracefulClose conn 5000 `catchAll_` pure () -- catchAll_ is needed here in case the connection was closed earlier
@@ -153,8 +168,8 @@ newSocketState = (,,) <$> newTVarIO 0 <*> newTVarIO 0 <*> newTVarIO mempty
 
 closeServer :: TMVar Bool -> TVar (IntMap (Weak ThreadId)) -> Socket -> IO ()
 closeServer started clients sock = do
-  readTVarIO clients >>= mapM_ (deRefWeak >=> mapM_ killThread)
   close sock
+  readTVarIO clients >>= mapM_ (deRefWeak >=> mapM_ killThread)
   void . atomically $ tryPutTMVar started False
 
 startTCPServer :: TMVar Bool -> Maybe HostName -> ServiceName -> IO Socket
@@ -176,30 +191,33 @@ startTCPServer started host port = withSocketsDo $ resolve >>= open >>= setStart
       pure sock
     setStarted sock = atomically (tryPutTMVar started True) >> pure sock
 
-loadTLSServerParams :: FilePath -> FilePath -> FilePath -> Maybe [ALPN] -> IO T.ServerParams
-loadTLSServerParams = loadSupportedTLSServerParams supportedParameters
+loadServerCredential :: ServerCredentials -> IO T.Credential
+loadServerCredential ServerCredentials {caCertificateFile, certificateFile, privateKeyFile} =
+  T.credentialLoadX509Chain certificateFile (maybeToList caCertificateFile) privateKeyFile >>= \case
+    Right credential -> pure credential
+    Left _ -> putStrLn "invalid credential" >> exitFailure
 
-loadSupportedTLSServerParams :: T.Supported -> FilePath -> FilePath -> FilePath -> Maybe [ALPN] -> IO T.ServerParams
-loadSupportedTLSServerParams serverSupported caCertificateFile certificateFile privateKeyFile alpn_ = do
-  tlsServerParams <- fromCredential <$> loadServerCredential
-  pure tlsServerParams {T.serverHooks = maybe def alpnHooks alpn_}
-  where
-    loadServerCredential :: IO T.Credential
-    loadServerCredential =
-      T.credentialLoadX509Chain certificateFile [caCertificateFile] privateKeyFile >>= \case
-        Right credential -> pure credential
-        Left _ -> putStrLn "invalid credential" >> exitFailure
-    fromCredential :: T.Credential -> T.ServerParams
-    fromCredential credential =
-      def
-        { T.serverWantClientCert = False,
-          T.serverShared = def {T.sharedCredentials = T.Credentials [credential]},
-          T.serverHooks = def,
-          T.serverSupported = serverSupported
-        }
-    alpnHooks supported = def {T.onALPNClientSuggest = Just $ pure . fromMaybe "" . find (`elem` supported)}
+supportedTLSServerParams :: T.Credential -> Maybe [ALPN] -> T.ServerParams
+supportedTLSServerParams = supportedTLSServerParams_ defaultSupportedParams . const
 
-loadFingerprint :: FilePath -> IO Fingerprint
-loadFingerprint certificateFile = do
+supportedTLSServerParams_ :: T.Supported -> (Maybe HostName -> T.Credential) -> Maybe [ALPN] -> T.ServerParams
+supportedTLSServerParams_ serverSupported creds alpn_ =
+  def
+    { T.serverWantClientCert = False,
+      T.serverHooks =
+        def
+          { T.onServerNameIndication = \host_ -> pure $ T.Credentials [creds host_],
+            T.onALPNClientSuggest = (\alpn -> pure . fromMaybe "" . find (`elem` alpn)) <$> alpn_
+          },
+      T.serverSupported = serverSupported
+    }
+
+loadFingerprint :: ServerCredentials -> IO Fingerprint
+loadFingerprint ServerCredentials {caCertificateFile} = case caCertificateFile of
+  Just certificateFile -> loadFileFingerprint certificateFile
+  Nothing -> error "CA file must be used in protocol credentials"
+
+loadFileFingerprint :: FilePath -> IO Fingerprint
+loadFileFingerprint certificateFile = do
   (cert : _) <- SX.readSignedObject certificateFile
   pure $ XV.getFingerprint (cert :: X.SignedExact X.Certificate) X.HashSHA256

src/Simplex/Messaging/Util.hs

@@ -14,11 +14,13 @@ import Data.Bifunctor (first)
 import Data.ByteString.Char8 (ByteString)
 import qualified Data.ByteString.Char8 as B
 import qualified Data.ByteString.Lazy.Char8 as LB
-import Data.Int (Int64)
 import Data.IORef
+import Data.Int (Int64)
 import Data.List (groupBy, sortOn)
 import Data.List.NonEmpty (NonEmpty)
 import qualified Data.List.NonEmpty as L
+import Data.Map.Strict (Map)
+import qualified Data.Map.Strict as M
 import Data.Text (Text)
 import qualified Data.Text as T
 import Data.Text.Encoding (decodeUtf8With, encodeUtf8)
@@ -33,12 +35,16 @@ raceAny_ = r []
     r as (m : ms) = withAsync m $ \a -> r (a : as) ms
     r as [] = void $ waitAnyCancel as
 
-infixl 4 <$$>, <$?>
+infixl 4 <$$>, <$$, <$?>
 
 (<$$>) :: (Functor f, Functor g) => (a -> b) -> f (g a) -> f (g b)
 (<$$>) = fmap . fmap
 {-# INLINE (<$$>) #-}
 
+(<$$) :: (Functor f, Functor g) => b -> f (g a) -> f (g b)
+(<$$) = fmap . fmap . const
+{-# INLINE (<$$) #-}
+
 (<$?>) :: MonadFail m => (a -> Either String b) -> m a -> m b
 f <$?> m = either fail pure . f =<< m
 {-# INLINE (<$?>) #-}
@@ -185,3 +191,7 @@ encodeJSON = safeDecodeUtf8 . LB.toStrict . J.encode
 
 decodeJSON :: FromJSON a => Text -> Maybe a
 decodeJSON = J.decode . LB.fromStrict . encodeUtf8
+
+traverseWithKey_ :: Monad m => (k -> v -> m ()) -> Map k v -> m ()
+traverseWithKey_ f = M.foldrWithKey (\k v -> (f k v >>)) (pure ())
+{-# INLINE traverseWithKey_ #-}

src/Simplex/RemoteControl/Client.hs

@@ -23,6 +23,9 @@ module Simplex.RemoteControl.Client
     rcEncryptBody,
     rcDecryptBody,
     xrcpBlockSize,
+    -- for tests only
+    sendRCPacket,
+    receiveRCPacket,
   ) where
 
 import Control.Applicative ((<|>))
@@ -33,6 +36,7 @@ import Control.Monad.IO.Class
 import Control.Monad.Trans.Except
 import Crypto.Random (ChaChaDRG)
 import qualified Data.Aeson as J
+import Data.Bitraversable (bimapM)
 import Data.ByteString (ByteString)
 import qualified Data.ByteString as B
 import qualified Data.ByteString.Lazy as LB
@@ -43,6 +47,7 @@ import qualified Data.List.NonEmpty as L
 import Data.Maybe (isNothing)
 import qualified Data.Text as T
 import Data.Time.Clock.System (getSystemTime)
+import Data.Tuple (swap)
 import Data.Word (Word16)
 import qualified Data.X509 as X509
 import Data.X509.Validation (Fingerprint (..), getFingerprint)
@@ -57,7 +62,7 @@ import Simplex.Messaging.Crypto.SNTRUP761
 import Simplex.Messaging.Crypto.SNTRUP761.Bindings
 import Simplex.Messaging.Encoding
 import Simplex.Messaging.Encoding.String (StrEncoding (..))
-import Simplex.Messaging.Transport (TLS (..), cGet, cPut)
+import Simplex.Messaging.Transport (TSbChainKeys (..), TLS (..), cGet, cPut)
 import Simplex.Messaging.Transport.Buffer (peekBuffered)
 import Simplex.Messaging.Transport.Client (TransportClientConfig (..), TransportHost (..), defaultTransportClientConfig, runTransportClient)
 import Simplex.Messaging.Transport.Credentials (genCredentials, tlsCredentials)
@@ -185,7 +190,7 @@ connectRCHost drg pairing@RCHostPairing {caKey, caCert, idPrivKey, knownHost} ct
               }
       pure $ signInvitation (snd sessKeys) idPrivKey inv
 
-genTLSCredentials :: TVar ChaChaDRG -> C.APrivateSignKey -> C.SignedCertificate -> IO TLS.Credentials
+genTLSCredentials :: TVar ChaChaDRG -> C.APrivateSignKey -> C.SignedCertificate -> IO TLS.Credential
 genTLSCredentials drg caKey caCert = do
   let caCreds = (C.signatureKeyPair caKey, caCert)
   leaf <- genCredentials drg (Just caCreds) (0, 24 * 999999) "localhost" -- session-signing cert
@@ -214,15 +219,16 @@ prepareHostSession
     hostHello@RCHostHello {v, ca, kem = kemPubKey} <- liftEitherWith RCESyntax $ J.eitherDecodeStrict helloBody
     unless (ca == tlsHostFingerprint) $ throwE RCEIdentity
     (kemCiphertext, kemSharedKey) <- liftIO $ sntrup761Enc drg kemPubKey
-    let hybridKey = kemHybridSecret dhPubKey dhPrivKey kemSharedKey
+    let KEMHybridSecret hybridKey = kemHybridSecret dhPubKey dhPrivKey kemSharedKey
     unless (isCompatible v supportedRCPVRange) $ throwE RCEVersion
-    let keys = HostSessKeys {hybridKey, idPrivKey, sessPrivKey}
+    (sndKey, rcvKey) <- bimapM newTVarIO newTVarIO $ C.sbcInit "" hybridKey
+    let keys = HostSessKeys {chainKeys = TSbChainKeys {sndKey, rcvKey}, idPrivKey, sessPrivKey}
     knownHost' <- updateKnownHost ca dhPubKey
     let ctrlHello = RCCtrlHello {}
     -- TODO send error response if something fails
-    nonce' <- liftIO . atomically $ C.randomCbNonce drg
-    encBody' <- liftEitherWith (const RCEBlockSize) $ kcbEncrypt hybridKey nonce' (LB.toStrict $ J.encode ctrlHello) helloBlockSize
-    let ctrlEncHello = RCCtrlEncHello {kem = kemCiphertext, nonce = nonce', encBody = encBody'}
+    (sk, nonce') <- atomically $ stateTVar sndKey C.sbcHkdf
+    encBody' <- liftEitherWith (const RCEBlockSize) $ C.sbEncrypt sk nonce' (LB.toStrict $ J.encode ctrlHello) helloBlockSize
+    let ctrlEncHello = RCCtrlEncHello {kem = kemCiphertext, encBody = encBody'}
     pure (ctrlEncHello, keys, hostHello, pairing {knownHost = Just knownHost'})
     where
       updateKnownHost :: C.KeyHash -> C.PublicKeyX25519 -> ExceptT RCErrorType IO KnownHostPairing
@@ -276,10 +282,7 @@ connectRCCtrl_ drg pairing'@RCCtrlPairing {caKey, caCert} inv@RCInvitation {ca,
       pure RCCClient_ {confirmSession, endSession}
     runClient :: RCCClient_ -> RCStepTMVar (SessionCode, TLS, RCStepTMVar (RCCtrlSession, RCCtrlPairing)) -> ExceptT RCErrorType IO ()
     runClient RCCClient_ {confirmSession, endSession} r = do
-      clientCredentials <-
-        liftIO (genTLSCredentials drg caKey caCert) >>= \case
-          TLS.Credentials (creds : _) -> pure $ Just creds
-          _ -> throwE $ RCEInternal "genTLSCredentials must generate credentials"
+      clientCredentials <- liftIO $ Just <$> genTLSCredentials drg caKey caCert
       let clientConfig = defaultTransportClientConfig {clientCredentials}
       ExceptT . runTransportClient clientConfig Nothing host (show port) (Just ca) $ \tls@TLS {tlsBuffer, tlsContext} -> runExceptT $ do
         -- pump socket to detect connection problems
@@ -349,13 +352,16 @@ prepareCtrlSession
   RCInvitation {skey, dh = dhPubKey}
   sharedKey
   kemPrivKey = \case
-    RCCtrlEncHello {kem = kemCiphertext, nonce, encBody} -> do
+    RCCtrlEncHello {kem = kemCiphertext, encBody} -> do
       kemSharedKey <- liftIO $ sntrup761Dec kemCiphertext kemPrivKey
-      let hybridKey = kemHybridSecret dhPubKey dhPrivKey kemSharedKey
-      helloBody <- liftEitherWith (const RCEDecrypt) $ kcbDecrypt hybridKey nonce encBody
+      let KEMHybridSecret hybridKey = kemHybridSecret dhPubKey dhPrivKey kemSharedKey
+      -- keys are swapped in controller
+      (sndKey, rcvKey) <- swap <$> bimapM newTVarIO newTVarIO (C.sbcInit "" hybridKey)
+      (sk, nonce) <- atomically $ stateTVar rcvKey C.sbcHkdf
+      helloBody <- liftEitherWith (const RCEDecrypt) $ C.sbDecrypt sk nonce encBody
       logDebug "Decrypted ctrl HELLO"
       RCCtrlHello {} <- liftEitherWith RCESyntax $ J.eitherDecodeStrict helloBody
-      pure CtrlSessKeys {hybridKey, idPubKey, sessPubKey = skey}
+      pure CtrlSessKeys {chainKeys = TSbChainKeys {sndKey, rcvKey}, idPubKey, sessPubKey = skey}
     RCCtrlEncError {nonce, encMessage} -> do
       message <- liftEitherWith (const RCEDecrypt) $ C.cbDecrypt sharedKey nonce encMessage
       throwE $ RCECtrlError $ T.unpack $ safeDecodeUtf8 message
@@ -426,17 +432,14 @@ cancelCtrlClient RCCtrlClient {action, client_ = RCCClient_ {endSession}} = do
 
 -- * Session encryption
 
-rcEncryptBody :: TVar ChaChaDRG -> KEMHybridSecret -> LazyByteString -> ExceptT RCErrorType IO (C.CbNonce, LazyByteString)
-rcEncryptBody drg hybridKey s = do
-  nonce <- atomically $ C.randomCbNonce drg
-  let len = LB.length s
-  ct <- liftEitherWith (const RCEEncrypt) $ LC.kcbEncryptTailTag hybridKey nonce s len (len + 8)
-  pure (nonce, ct)
+rcEncryptBody :: C.SbKeyNonce -> LazyByteString -> ExceptT RCErrorType IO LazyByteString
+rcEncryptBody keyNonce s = do
+  liftEitherWith (const RCEEncrypt) $ LC.sbEncryptTailTagNoPad keyNonce s
 
-rcDecryptBody :: KEMHybridSecret -> C.CbNonce -> LazyByteString -> ExceptT RCErrorType IO LazyByteString
-rcDecryptBody hybridKey nonce ct = do
+rcDecryptBody :: C.SbKeyNonce -> LazyByteString -> ExceptT RCErrorType IO LazyByteString
+rcDecryptBody keyNonce ct = do
   let len = LB.length ct - 16
   when (len < 0) $ throwE RCEDecrypt
-  (ok, s) <- liftEitherWith (const RCEDecrypt) $ LC.kcbDecryptTailTag hybridKey nonce len ct
+  (ok, s) <- liftEitherWith (const RCEDecrypt) $ LC.sbDecryptTailTagNoPad keyNonce len ct
   unless ok $ throwE RCEDecrypt
   pure s

src/Simplex/RemoteControl/Discovery.hs

@@ -23,7 +23,7 @@ import Network.Info (IPv4 (..), NetworkInterface (..), getNetworkInterfaces)
 import qualified Network.Socket as N
 import qualified Network.TLS as TLS
 import qualified Network.UDP as UDP
-import Simplex.Messaging.Transport (supportedParameters)
+import Simplex.Messaging.Transport (defaultSupportedParams)
 import qualified Simplex.Messaging.Transport as Transport
 import Simplex.Messaging.Transport.Client (TransportHost (..))
 import Simplex.Messaging.Transport.Server (defaultTransportServerConfig, runTransportServerSocket, startTCPServer)
@@ -68,7 +68,7 @@ preferAddress RCCtrlAddress {address, interface} addrs =
     matchAddr RCCtrlAddress {address = a} = a == address
     matchIface RCCtrlAddress {interface = i} = i == interface
 
-startTLSServer :: Maybe Word16 -> TMVar (Maybe N.PortNumber) -> TLS.Credentials -> TLS.ServerHooks -> (Transport.TLS -> IO ()) -> IO (Async ())
+startTLSServer :: Maybe Word16 -> TMVar (Maybe N.PortNumber) -> TLS.Credential -> TLS.ServerHooks -> (Transport.TLS -> IO ()) -> IO (Async ())
 startTLSServer port_ startedOnPort credentials hooks server = async . liftIO $ do
   started <- newEmptyTMVarIO
   bracketOnError (startTCPServer started Nothing $ maybe "0" show port_) (\_e -> setPort Nothing) $ \socket ->
@@ -81,14 +81,14 @@ startTLSServer port_ startedOnPort credentials hooks server = async . liftIO $ d
       port <- N.socketPort socket
       logInfo $ "System-assigned port: " <> tshow port
       setPort $ Just port
-      runTransportServerSocket started (pure socket) "RCP TLS" serverParams defaultTransportServerConfig server
+      runTransportServerSocket started (pure socket) "RCP TLS" credentials serverParams defaultTransportServerConfig server
     setPort = void . atomically . tryPutTMVar startedOnPort
     serverParams =
       def
         { TLS.serverWantClientCert = True,
-          TLS.serverShared = def {TLS.sharedCredentials = credentials},
+          TLS.serverShared = def {TLS.sharedCredentials = TLS.Credentials [credentials]},
           TLS.serverHooks = hooks,
-          TLS.serverSupported = supportedParameters
+          TLS.serverSupported = defaultSupportedParams
         }
 
 withSender :: (UDP.UDPSocket -> IO a) -> IO a

src/Simplex/RemoteControl/Types.hs

@@ -20,12 +20,11 @@ import qualified Data.Text as T
 import Data.Text.Encoding (encodeUtf8)
 import Data.Word (Word16)
 import qualified Simplex.Messaging.Crypto as C
-import Simplex.Messaging.Crypto.SNTRUP761
 import Simplex.Messaging.Crypto.SNTRUP761.Bindings
 import Simplex.Messaging.Encoding
 import Simplex.Messaging.Encoding.String
 import Simplex.Messaging.Parsers (defaultJSON, dropPrefix, sumTypeJSON)
-import Simplex.Messaging.Transport (TLS)
+import Simplex.Messaging.Transport (TLS, TSbChainKeys)
 import Simplex.Messaging.Transport.Client (TransportHost)
 import Simplex.Messaging.Util (safeDecodeUtf8)
 import Simplex.Messaging.Version (VersionRange, VersionScope, mkVersionRange)
@@ -180,7 +179,7 @@ data RCHostSession = RCHostSession
   }
 
 data HostSessKeys = HostSessKeys
-  { hybridKey :: KEMHybridSecret,
+  { chainKeys :: TSbChainKeys,
     idPrivKey :: C.PrivateKeyEd25519,
     sessPrivKey :: C.PrivateKeyEd25519
   }
@@ -193,7 +192,7 @@ data RCCtrlSession = RCCtrlSession
   }
 
 data CtrlSessKeys = CtrlSessKeys
-  { hybridKey :: KEMHybridSecret,
+  { chainKeys :: TSbChainKeys,
     idPubKey :: C.PublicKeyEd25519,
     sessPubKey :: C.PublicKeyEd25519
   }
@@ -213,19 +212,19 @@ instance Encoding RCHostEncHello where
     pure RCHostEncHello {dhPubKey, nonce, encBody}
 
 data RCCtrlEncHello
-  = RCCtrlEncHello {kem :: KEMCiphertext, nonce :: C.CbNonce, encBody :: ByteString}
+  = RCCtrlEncHello {kem :: KEMCiphertext, encBody :: ByteString}
   | RCCtrlEncError {nonce :: C.CbNonce, encMessage :: ByteString}
   deriving (Show)
 
 instance Encoding RCCtrlEncHello where
   smpEncode = \case
-    RCCtrlEncHello {kem, nonce, encBody} -> "HELLO " <> smpEncode (kem, nonce, Tail encBody)
+    RCCtrlEncHello {kem, encBody} -> "HELLO " <> smpEncode (kem, Tail encBody)
     RCCtrlEncError {nonce, encMessage} -> "ERROR " <> smpEncode (nonce, Tail encMessage)
   smpP =
     A.takeTill (== ' ') >>= \case
       "HELLO" -> do
-        (kem, nonce, Tail encBody) <- _smpP
-        pure RCCtrlEncHello {kem, nonce, encBody}
+        (kem, Tail encBody) <- _smpP
+        pure RCCtrlEncHello {kem, encBody}
       "ERROR" -> do
         (nonce, Tail encMessage) <- _smpP
         pure RCCtrlEncError {nonce, encMessage}

tests/AgentTests.hs

@@ -16,6 +16,7 @@ import AgentTests.FunctionalAPITests (functionalAPITests)
 import AgentTests.MigrationTests (migrationTests)
 import AgentTests.NotificationTests (notificationTests)
 import AgentTests.SQLiteTests (storeTests)
+import AgentTests.ServerChoice (serverChoiceTests)
 import Simplex.Messaging.Transport (ATransport (..))
 import Test.Hspec
 
@@ -26,4 +27,5 @@ agentTests (ATransport t) = do
   describe "Functional API" $ functionalAPITests (ATransport t)
   describe "Notification tests" $ notificationTests (ATransport t)
   describe "SQLite store" storeTests
+  describe "Chosen servers" serverChoiceTests
   describe "Migration tests" migrationTests

tests/AgentTests/FunctionalAPITests.hs

@@ -12,6 +12,7 @@
 {-# LANGUAGE PatternSynonyms #-}
 {-# LANGUAGE RankNTypes #-}
 {-# LANGUAGE ScopedTypeVariables #-}
+{-# LANGUAGE TupleSections #-}
 {-# LANGUAGE TypeApplications #-}
 {-# OPTIONS_GHC -Wno-orphans #-}
 
@@ -76,7 +77,7 @@ import Data.Word (Word16)
 import qualified Database.SQLite.Simple as SQL
 import GHC.Stack (withFrozenCallStack)
 import SMPAgentClient
-import SMPClient (cfg, prevRange, prevVersion, testPort, testPort2, testStoreLogFile2, withSmpServer, withSmpServerConfigOn, withSmpServerOn, withSmpServerProxy, withSmpServerStoreLogOn, withSmpServerStoreMsgLogOn)
+import SMPClient (cfg, prevRange, prevVersion, testPort, testPort2, testStoreLogFile2, testStoreMsgsDir2, withSmpServer, withSmpServerConfigOn, withSmpServerProxy, withSmpServerStoreLogOn, withSmpServerStoreMsgLogOn)
 import Simplex.Messaging.Agent hiding (createConnection, joinConnection, sendMessage)
 import qualified Simplex.Messaging.Agent as A
 import Simplex.Messaging.Agent.Client (ProtocolTestFailure (..), ProtocolTestStep (..), ServerQueueInfo (..), UserNetworkInfo (..), UserNetworkType (..), waitForUserNetwork)
@@ -85,7 +86,7 @@ import Simplex.Messaging.Agent.Protocol hiding (CON, CONF, INFO, REQ, SENT)
 import qualified Simplex.Messaging.Agent.Protocol as A
 import Simplex.Messaging.Agent.Store.SQLite (MigrationConfirmation (..), SQLiteStore (dbNew))
 import Simplex.Messaging.Agent.Store.SQLite.Common (withTransaction')
-import Simplex.Messaging.Client (NetworkConfig (..), ProtocolClientConfig (..), SMPProxyFallback (..), SMPProxyMode (..), TransportSessionMode (TSMEntity, TSMUser), defaultClientConfig)
+import Simplex.Messaging.Client (NetworkConfig (..), ProtocolClientConfig (..), SMPProxyFallback (..), SMPProxyMode (..), TransportSessionMode (..), defaultClientConfig)
 import qualified Simplex.Messaging.Crypto as C
 import Simplex.Messaging.Crypto.Ratchet (InitialKeys (..), PQEncryption (..), PQSupport (..), pattern IKPQOff, pattern IKPQOn, pattern PQEncOff, pattern PQEncOn, pattern PQSupportOff, pattern PQSupportOn)
 import qualified Simplex.Messaging.Crypto.Ratchet as CR
@@ -95,6 +96,7 @@ import Simplex.Messaging.Protocol (BasicAuth, ErrorType (..), MsgBody, ProtocolS
 import qualified Simplex.Messaging.Protocol as SMP
 import Simplex.Messaging.Server.Env.STM (ServerConfig (..))
 import Simplex.Messaging.Server.Expiration
+import Simplex.Messaging.Server.MsgStore.Types (AMSType (..), SMSType (..))
 import Simplex.Messaging.Server.QueueStore.QueueInfo
 import Simplex.Messaging.Transport (ATransport (..), SMPVersion, VersionSMP, authCmdsSMPVersion, basicAuthSMPVersion, batchCmdsSMPVersion, currentServerSMPRelayVersion, sndAuthKeySMPVersion, supportedSMPHandshakes)
 import Simplex.Messaging.Util (bshow, diffToMicroseconds)
@@ -246,7 +248,9 @@ createConnection :: AgentClient -> UserId -> Bool -> SConnectionMode c -> Maybe
 createConnection c userId enableNtfs cMode clientData = A.createConnection c userId enableNtfs cMode clientData (IKNoPQ PQSupportOn)
 
 joinConnection :: AgentClient -> UserId -> Bool -> ConnectionRequestUri c -> ConnInfo -> SubscriptionMode -> AE (ConnId, SndQueueSecured)
-joinConnection c userId enableNtfs cReq connInfo = A.joinConnection c userId Nothing enableNtfs cReq connInfo PQSupportOn
+joinConnection c userId enableNtfs cReq connInfo subMode = do
+  connId <- A.prepareConnectionToJoin c userId enableNtfs cReq PQSupportOn
+  (connId,) <$> A.joinConnection c userId connId enableNtfs cReq connInfo PQSupportOn subMode
 
 sendMessage :: AgentClient -> ConnId -> SMP.MsgFlags -> MsgBody -> AE AgentMsgId
 sendMessage c connId msgFlags msgBody = do
@@ -527,7 +531,7 @@ testRatchetMatrix2 t runTest = do
 testServerMatrix2 :: HasCallStack => ATransport -> (InitialAgentServers -> IO ()) -> Spec
 testServerMatrix2 t runTest = do
   it "1 server" $ withSmpServer t $ runTest initAgentServers
-  it "2 servers" $ withSmpServer t . withSmpServerOn t testPort2 $ runTest initAgentServers2
+  it "2 servers" $ withSmpServer t $ withSmpServerConfigOn t cfg {storeLogFile = Just testStoreLogFile2, storeMsgsFile = Just testStoreMsgsDir2} testPort2 $ \_ -> runTest initAgentServers2
 
 testPQMatrix2 :: HasCallStack => ATransport -> (HasCallStack => (AgentClient, InitialKeys) -> (AgentClient, PQSupport) -> AgentMsgId -> IO ()) -> Spec
 testPQMatrix2 = pqMatrix2_ True
@@ -605,7 +609,8 @@ runAgentClientTestPQ :: HasCallStack => SndQueueSecured -> Bool -> (AgentClient,
 runAgentClientTestPQ sqSecured viaProxy (alice, aPQ) (bob, bPQ) baseId =
   runRight_ $ do
     (bobId, qInfo) <- A.createConnection alice 1 True SCMInvitation Nothing aPQ SMSubscribe
-    (aliceId, sqSecured') <- A.joinConnection bob 1 Nothing True qInfo "bob's connInfo" bPQ SMSubscribe
+    aliceId <- A.prepareConnectionToJoin bob 1 True qInfo bPQ
+    sqSecured' <- A.joinConnection bob 1 aliceId True qInfo "bob's connInfo" bPQ SMSubscribe
     liftIO $ sqSecured' `shouldBe` sqSecured
     ("", _, A.CONF confId pqSup' _ "bob's connInfo") <- get alice
     liftIO $ pqSup' `shouldBe` CR.connPQEncryption aPQ
@@ -807,13 +812,12 @@ runAgentClientContactTestPQ sqSecured viaProxy reqPQSupport (alice, aPQ) (bob, b
   runRight_ $ do
     (_, qInfo) <- A.createConnection alice 1 True SCMContact Nothing aPQ SMSubscribe
     aliceId <- A.prepareConnectionToJoin bob 1 True qInfo bPQ
-    (aliceId', sqSecuredJoin) <- A.joinConnection bob 1 (Just aliceId) True qInfo "bob's connInfo" bPQ SMSubscribe
-    liftIO $ do
-      aliceId' `shouldBe` aliceId
-      sqSecuredJoin `shouldBe` False -- joining via contact address connection
+    sqSecuredJoin <- A.joinConnection bob 1 aliceId True qInfo "bob's connInfo" bPQ SMSubscribe
+    liftIO $ sqSecuredJoin `shouldBe` False -- joining via contact address connection
     ("", _, A.REQ invId pqSup' _ "bob's connInfo") <- get alice
     liftIO $ pqSup' `shouldBe` reqPQSupport
-    (bobId, sqSecured') <- acceptContact alice True invId "alice's connInfo" (CR.connPQEncryption aPQ) SMSubscribe
+    bobId <- A.prepareConnectionToAccept alice True invId (CR.connPQEncryption aPQ)
+    sqSecured' <- acceptContact alice bobId True invId "alice's connInfo" (CR.connPQEncryption aPQ) SMSubscribe
     liftIO $ sqSecured' `shouldBe` sqSecured
     ("", _, A.CONF confId pqSup'' _ "alice's connInfo") <- get bob
     liftIO $ pqSup'' `shouldBe` bPQ
@@ -859,13 +863,12 @@ runAgentClientContactTestPQ3 viaProxy (alice, aPQ) (bob, bPQ) (tom, tPQ) baseId
     msgId = subtract baseId . fst
     connectViaContact b pq qInfo = do
       aId <- A.prepareConnectionToJoin b 1 True qInfo pq
-      (aId', sqSecuredJoin) <- A.joinConnection b 1 (Just aId) True qInfo "bob's connInfo" pq SMSubscribe
-      liftIO $ do
-        aId' `shouldBe` aId
-        sqSecuredJoin `shouldBe` False -- joining via contact address connection
+      sqSecuredJoin <- A.joinConnection b 1 aId True qInfo "bob's connInfo" pq SMSubscribe
+      liftIO $ sqSecuredJoin `shouldBe` False -- joining via contact address connection
       ("", _, A.REQ invId pqSup' _ "bob's connInfo") <- get alice
       liftIO $ pqSup' `shouldBe` PQSupportOn
-      (bId, sqSecuredAccept) <- acceptContact alice True invId "alice's connInfo" (CR.connPQEncryption aPQ) SMSubscribe
+      bId <- A.prepareConnectionToAccept alice True invId (CR.connPQEncryption aPQ)
+      sqSecuredAccept <- acceptContact alice bId True invId "alice's connInfo" (CR.connPQEncryption aPQ) SMSubscribe
       liftIO $ sqSecuredAccept `shouldBe` False -- agent cfg is v8
       ("", _, A.CONF confId pqSup'' _ "alice's connInfo") <- get b
       liftIO $ pqSup'' `shouldBe` pq
@@ -906,10 +909,8 @@ testRejectContactRequest =
   withAgentClients2 $ \alice bob -> runRight_ $ do
     (addrConnId, qInfo) <- A.createConnection alice 1 True SCMContact Nothing IKPQOn SMSubscribe
     aliceId <- A.prepareConnectionToJoin bob 1 True qInfo PQSupportOn
-    (aliceId', sqSecured) <- A.joinConnection bob 1 (Just aliceId) True qInfo "bob's connInfo" PQSupportOn SMSubscribe
-    liftIO $ do
-      aliceId' `shouldBe` aliceId
-      sqSecured `shouldBe` False -- joining via contact address connection
+    sqSecured <- A.joinConnection bob 1 aliceId True qInfo "bob's connInfo" PQSupportOn SMSubscribe
+    liftIO $ sqSecured `shouldBe` False -- joining via contact address connection
     ("", _, A.REQ invId PQSupportOn _ "bob's connInfo") <- get alice
     liftIO $ runExceptT (rejectContact alice "abcd" invId) `shouldReturn` Left (CONN NOT_FOUND)
     rejectContact alice addrConnId invId
@@ -922,10 +923,8 @@ testUpdateConnectionUserId =
     newUserId <- createUser alice [noAuthSrvCfg testSMPServer] [noAuthSrvCfg testXFTPServer]
     _ <- changeConnectionUser alice 1 connId newUserId
     aliceId <- A.prepareConnectionToJoin bob 1 True qInfo PQSupportOn
-    (aliceId', sqSecured') <- A.joinConnection bob 1 (Just aliceId) True qInfo "bob's connInfo" PQSupportOn SMSubscribe
-    liftIO $ do
-      aliceId' `shouldBe` aliceId
-      sqSecured' `shouldBe` True
+    sqSecured' <- A.joinConnection bob 1 aliceId True qInfo "bob's connInfo" PQSupportOn SMSubscribe
+    liftIO $ sqSecured' `shouldBe` True
     ("", _, A.CONF confId pqSup' _ "bob's connInfo") <- get alice
     liftIO $ pqSup' `shouldBe` PQSupportOn
     allowConnection alice connId confId "alice's connInfo"
@@ -1043,7 +1042,7 @@ testAllowConnectionClientRestart t = do
   bob <- getSMPAgentClient' 2 agentCfg initAgentServersSrv2 testDB2
   withSmpServerStoreLogOn t testPort $ \_ -> do
     (aliceId, bobId, confId) <-
-      withSmpServerConfigOn t cfg {storeLogFile = Just testStoreLogFile2} testPort2 $ \_ -> do
+      withSmpServerConfigOn t cfg {storeLogFile = Just testStoreLogFile2, storeMsgsFile = Just testStoreMsgsDir2} testPort2 $ \_ -> do
         runRight $ do
           (bobId, qInfo) <- createConnection alice 1 True SCMInvitation Nothing SMSubscribe
           (aliceId, sqSecured) <- joinConnection bob 1 True qInfo "bob's connInfo" SMSubscribe
@@ -1065,8 +1064,7 @@ testAllowConnectionClientRestart t = do
     alice2 <- getSMPAgentClient' 3 agentCfg initAgentServers testDB
     runRight_ $ subscribeConnection alice2 bobId
     threadDelay 500000
-
-    withSmpServerConfigOn t cfg {storeLogFile = Just testStoreLogFile2} testPort2 $ \_ -> do
+    withSmpServerConfigOn t cfg {storeLogFile = Just testStoreLogFile2, storeMsgsFile = Just testStoreMsgsDir2} testPort2 $ \_ -> do
       runRight $ do
         ("", "", UP _ _) <- nGet bob
         get alice2 ##> ("", bobId, CON)
@@ -1284,7 +1282,7 @@ testSkippedMessages :: HasCallStack => ATransport -> IO ()
 testSkippedMessages t = do
   alice <- getSMPAgentClient' 1 agentCfg initAgentServers testDB
   bob <- getSMPAgentClient' 2 agentCfg initAgentServers testDB2
-  (aliceId, bobId) <- withSmpServerStoreLogOn t testPort $ \_ -> do
+  (aliceId, bobId) <- withSmpServerConfigOn t cfg' testPort $ \_ -> do
     (aliceId, bobId) <- runRight $ makeConnection alice bob
     runRight_ $ do
       2 <- sendMessage alice bobId SMP.noMsgFlags "hello"
@@ -1312,7 +1310,7 @@ testSkippedMessages t = do
   alice2 <- getSMPAgentClient' 3 agentCfg initAgentServers testDB
   bob2 <- getSMPAgentClient' 4 agentCfg initAgentServers testDB2
 
-  withSmpServerStoreLogOn t testPort $ \_ -> do
+  withSmpServerConfigOn t cfg' testPort $ \_ -> do
     runRight_ $ do
       subscribeConnection bob2 aliceId
       subscribeConnection alice2 bobId
@@ -1328,6 +1326,8 @@ testSkippedMessages t = do
       ackMessage bob2 aliceId 4 Nothing
   disposeAgentClient alice2
   disposeAgentClient bob2
+  where
+    cfg' = cfg {msgStoreType = AMSType SMSMemory, storeMsgsFile = Nothing}
 
 testDeliveryAfterSubscriptionError :: HasCallStack => ATransport -> IO ()
 testDeliveryAfterSubscriptionError t = do
@@ -1431,7 +1431,7 @@ withUP a bId p =
       ]
 
 testExpireMessageQuota :: HasCallStack => ATransport -> IO ()
-testExpireMessageQuota t = withSmpServerConfigOn t cfg {msgQueueQuota = 1} testPort $ \_ -> do
+testExpireMessageQuota t = withSmpServerConfigOn t cfg {msgQueueQuota = 1, maxJournalMsgCount = 2} testPort $ \_ -> do
   a <- getSMPAgentClient' 1 agentCfg {quotaExceededTimeout = 1, messageRetryInterval = fastMessageRetryInterval} initAgentServers testDB
   b <- getSMPAgentClient' 2 agentCfg initAgentServers testDB2
   (aId, bId) <- runRight $ do
@@ -1457,7 +1457,7 @@ testExpireMessageQuota t = withSmpServerConfigOn t cfg {msgQueueQuota = 1} testP
   disposeAgentClient a
 
 testExpireManyMessagesQuota :: ATransport -> IO ()
-testExpireManyMessagesQuota t = withSmpServerConfigOn t cfg {msgQueueQuota = 1} testPort $ \_ -> do
+testExpireManyMessagesQuota t = withSmpServerConfigOn t cfg {msgQueueQuota = 1, maxJournalMsgCount = 2} testPort $ \_ -> do
   a <- getSMPAgentClient' 1 agentCfg {quotaExceededTimeout = 2, messageRetryInterval = fastMessageRetryInterval} initAgentServers testDB
   b <- getSMPAgentClient' 2 agentCfg initAgentServers testDB2
   (aId, bId) <- runRight $ do
@@ -1681,34 +1681,45 @@ testOnlyCreatePullSlowHandshake = withAgentClientsCfg2 agentProxyCfgV8 agentProx
   (aliceId, sqSecured) <- joinConnection bob 1 True qInfo "bob's connInfo" SMOnlyCreate
   liftIO $ sqSecured `shouldBe` False
   Just ("", _, CONF confId _ "bob's connInfo") <- getMsg alice bobId $ timeout 5_000000 $ get alice
+  getMSGNTF alice bobId
   allowConnection alice bobId confId "alice's connInfo"
   liftIO $ threadDelay 1_000000
   getMsg bob aliceId $
     get bob ##> ("", aliceId, INFO "alice's connInfo")
+  getMSGNTF bob aliceId
   liftIO $ threadDelay 1_000000
   getMsg alice bobId $ pure ()
-  get alice ##> ("", bobId, CON)
+  inAnyOrder
+    (get alice)
+    [ \case ("", c, CON) -> c == bobId; _ -> False,
+      \case ("", c, MSGNTF {}) -> c == bobId; _ -> False
+    ]
   getMsg bob aliceId $
     get bob ##> ("", aliceId, CON)
+  getMSGNTF bob aliceId
   -- exchange messages
   4 <- sendMessage alice bobId SMP.noMsgFlags "hello"
   get alice ##> ("", bobId, SENT 4)
   getMsg bob aliceId $
     get bob =##> \case ("", c, Msg "hello") -> c == aliceId; _ -> False
   ackMessage bob aliceId 4 Nothing
+  getMSGNTF bob aliceId
   5 <- sendMessage bob aliceId SMP.noMsgFlags "hello too"
   get bob ##> ("", aliceId, SENT 5)
   getMsg alice bobId $
     get alice =##> \case ("", c, Msg "hello too") -> c == bobId; _ -> False
   ackMessage alice bobId 5 Nothing
+  getMSGNTF alice bobId
 
 getMsg :: AgentClient -> ConnId -> ExceptT AgentErrorType IO a -> ExceptT AgentErrorType IO a
 getMsg c cId action = do
   liftIO $ noMessages c "nothing should be delivered before GET"
-  Just _ <- getConnectionMessage c cId
-  r <- action
-  get c =##> \case ("", cId', MSGNTF _) -> cId == cId'; _ -> False
-  pure r
+  [Just _] <- lift $ getConnectionMessages c [cId]
+  action
+
+getMSGNTF :: AgentClient -> ConnId -> ExceptT AgentErrorType IO ()
+getMSGNTF c cId =
+  get c =##> \case ("", c', MSGNTF {}) -> c' == cId; _ -> False
 
 testOnlyCreatePull :: IO ()
 testOnlyCreatePull = withAgentClients2 $ \alice bob -> runRight_ $ do
@@ -1716,11 +1727,13 @@ testOnlyCreatePull = withAgentClients2 $ \alice bob -> runRight_ $ do
   (aliceId, sqSecured) <- joinConnection bob 1 True qInfo "bob's connInfo" SMOnlyCreate
   liftIO $ sqSecured `shouldBe` True
   Just ("", _, CONF confId _ "bob's connInfo") <- getMsg alice bobId $ timeout 5_000000 $ get alice
+  getMSGNTF alice bobId
   allowConnection alice bobId confId "alice's connInfo"
   liftIO $ threadDelay 1_000000
   getMsg bob aliceId $ do
     get bob ##> ("", aliceId, INFO "alice's connInfo")
     get bob ##> ("", aliceId, CON)
+  getMSGNTF bob aliceId
   liftIO $ threadDelay 1_000000
   get alice ##> ("", bobId, CON) -- sent to initiating party after sending confirmation
   -- exchange messages
@@ -1729,11 +1742,13 @@ testOnlyCreatePull = withAgentClients2 $ \alice bob -> runRight_ $ do
   getMsg bob aliceId $
     get bob =##> \case ("", c, Msg "hello") -> c == aliceId; _ -> False
   ackMessage bob aliceId 2 Nothing
+  getMSGNTF bob aliceId
   3 <- sendMessage bob aliceId SMP.noMsgFlags "hello too"
   get bob ##> ("", aliceId, SENT 3)
   getMsg alice bobId $
     get alice =##> \case ("", c, Msg "hello too") -> c == bobId; _ -> False
   ackMessage alice bobId 3 Nothing
+  getMSGNTF alice bobId
 
 makeConnection :: AgentClient -> AgentClient -> ExceptT AgentErrorType IO (ConnId, ConnId)
 makeConnection = makeConnection_ PQSupportOn True
@@ -1748,10 +1763,8 @@ makeConnectionForUsers_ :: HasCallStack => PQSupport -> SndQueueSecured -> Agent
 makeConnectionForUsers_ pqSupport sqSecured alice aliceUserId bob bobUserId = do
   (bobId, qInfo) <- A.createConnection alice aliceUserId True SCMInvitation Nothing (CR.IKNoPQ pqSupport) SMSubscribe
   aliceId <- A.prepareConnectionToJoin bob bobUserId True qInfo pqSupport
-  (aliceId', sqSecured') <- A.joinConnection bob bobUserId (Just aliceId) True qInfo "bob's connInfo" pqSupport SMSubscribe
-  liftIO $ do
-    aliceId' `shouldBe` aliceId
-    sqSecured' `shouldBe` sqSecured
+  sqSecured' <- A.joinConnection bob bobUserId aliceId True qInfo "bob's connInfo" pqSupport SMSubscribe
+  liftIO $ sqSecured' `shouldBe` sqSecured
   ("", _, A.CONF confId pqSup' _ "bob's connInfo") <- get alice
   liftIO $ pqSup' `shouldBe` pqSupport
   allowConnection alice bobId confId "alice's connInfo"
@@ -1933,7 +1946,7 @@ testBatchedSubscriptions nCreate nDel t =
     runServers :: ExceptT AgentErrorType IO a -> IO a
     runServers a = do
       withSmpServerStoreLogOn t testPort $ \t1 -> do
-        res <- withSmpServerConfigOn t cfg {storeLogFile = Just testStoreLogFile2} testPort2 $ \t2 ->
+        res <- withSmpServerConfigOn t cfg {storeLogFile = Just testStoreLogFile2, storeMsgsFile = Just testStoreMsgsDir2} testPort2 $ \t2 ->
           runRight a `finally` killThread t2
         killThread t1
         pure res
@@ -1995,7 +2008,7 @@ pattern SentB :: AgentMsgId -> Either AgentErrorType (AgentMsgId, PQEncryption)
 pattern SentB msgId <- Right (msgId, PQEncOn)
 
 receiveMsg :: AgentClient -> ConnId -> AgentMsgId -> MsgBody -> ExceptT AgentErrorType IO ()
-receiveMsg c cId msgId msg  = do
+receiveMsg c cId msgId msg = do
   get c =##> \case ("", cId', Msg' mId' PQEncOn msg') -> cId' == cId && mId' == msgId && msg' == msg; _ -> False
   ackMessage c cId msgId Nothing
 
@@ -2345,7 +2358,7 @@ testJoinConnectionAsyncReplyErrorV8 t = do
         ConnectionStats {rcvQueuesInfo = [], sndQueuesInfo = [SndQueueInfo {}]} <- getConnectionServers b aId
         pure (aId, bId)
       nGet a =##> \case ("", "", DOWN _ [c]) -> c == bId; _ -> False
-      withSmpServerOn t testPort2 $ do
+      withSmpServerConfigOn t cfg {storeLogFile = Just testStoreLogFile2, storeMsgsFile = Just testStoreMsgsDir2} testPort2 $ \_ -> do
         get b =##> \case ("2", c, JOINED sqSecured) -> c == aId && not sqSecured; _ -> False
         confId <- withSmpServerStoreLogOn t testPort $ \_ -> do
           pGet a >>= \case
@@ -2384,7 +2397,7 @@ testJoinConnectionAsyncReplyError t = do
         ConnectionStats {rcvQueuesInfo = [], sndQueuesInfo = [SndQueueInfo {}]} <- getConnectionServers b aId
         pure (aId, bId)
       nGet a =##> \case ("", "", DOWN _ [c]) -> c == bId; _ -> False
-      withSmpServerOn t testPort2 $ do
+      withSmpServerConfigOn t cfg {storeLogFile = Just testStoreLogFile2, storeMsgsFile = Just testStoreMsgsDir2} testPort2 $ \_ -> do
         confId <- withSmpServerStoreLogOn t testPort $ \_ -> do
           -- both servers need to be online for connection to progress because of SKEY
           get b =##> \case ("2", c, JOINED sqSecured) -> c == aId && sqSecured; _ -> False
@@ -2869,7 +2882,7 @@ testCreateQueueAuth srvVersion clnt1 clnt2 sqSecured baseId = do
     getClient clientId (clntAuth, clntVersion) db =
       let servers = initAgentServers {smp = userServers' [ProtoServerWithAuth testSMPServer clntAuth]}
           alpn_ = if clntVersion >= authCmdsSMPVersion then Just supportedSMPHandshakes else Nothing
-          smpCfg = defaultClientConfig alpn_ $ V.mkVersionRange (prevVersion basicAuthSMPVersion) clntVersion
+          smpCfg = defaultClientConfig alpn_ False $ V.mkVersionRange (prevVersion basicAuthSMPVersion) clntVersion
           sndAuthAlg = if srvVersion >= authCmdsSMPVersion && clntVersion >= authCmdsSMPVersion then C.AuthAlg C.SX25519 else C.AuthAlg C.SEd25519
        in getSMPAgentClient' clientId agentCfg {smpCfg, sndAuthAlg} servers db
 
@@ -2963,7 +2976,7 @@ testDeliveryReceiptsVersion t = do
 
 testDeliveryReceiptsConcurrent :: HasCallStack => ATransport -> IO ()
 testDeliveryReceiptsConcurrent t =
-  withSmpServerConfigOn t cfg {msgQueueQuota = 256} testPort $ \_ -> do
+  withSmpServerConfigOn t cfg {msgQueueQuota = 256, maxJournalMsgCount = 512} testPort $ \_ -> do
     withAgentClients2 $ \a b -> do
       (aId, bId) <- runRight $ makeConnection a b
       t1 <- liftIO getCurrentTime
@@ -3014,7 +3027,7 @@ testDeliveryReceiptsConcurrent t =
 testTwoUsers :: HasCallStack => IO ()
 testTwoUsers = withAgentClients2 $ \a b -> do
   let nc = netCfg initAgentServers
-  sessionMode nc `shouldBe` TSMUser
+  sessionMode nc `shouldBe` TSMSession
   runRight_ $ do
     (aId1, bId1) <- makeConnectionForUsers a 1 b 1
     exchangeGreetings a bId1 b aId1
@@ -3063,8 +3076,9 @@ testTwoUsers = withAgentClients2 $ \a b -> do
     ("", "", DOWN _ _) <- nGet a
     ("", "", DOWN _ _) <- nGet a
     ("", "", DOWN _ _) <- nGet a
-    ("", "", DOWN _ _) <- nGet a
-    ("", "", UP _ _) <- nGet a
+    -- to avoice race condition
+    nGet a =##> \case ("", "", DOWN _ _) -> True; ("", "", UP _ _) -> True; _ -> False
+    nGet a =##> \case ("", "", UP _ _) -> True; ("", "", DOWN _ _) -> True; _ -> False
     ("", "", UP _ _) <- nGet a
     ("", "", UP _ _) <- nGet a
     ("", "", UP _ _) <- nGet a

tests/AgentTests/NotificationTests.hs

@@ -3,6 +3,7 @@
 {-# LANGUAGE GADTs #-}
 {-# LANGUAGE LambdaCase #-}
 {-# LANGUAGE NamedFieldPuns #-}
+{-# LANGUAGE OverloadedLists #-}
 {-# LANGUAGE OverloadedStrings #-}
 {-# LANGUAGE PatternSynonyms #-}
 {-# LANGUAGE QuasiQuotes #-}
@@ -49,12 +50,13 @@ import Data.Bifunctor (bimap, first)
 import qualified Data.ByteString.Base64.URL as U
 import Data.ByteString.Char8 (ByteString)
 import qualified Data.ByteString.Char8 as B
+import Data.List.NonEmpty (NonEmpty (..))
 import qualified Data.List.NonEmpty as L
 import Data.Text.Encoding (encodeUtf8)
 import Database.SQLite.Simple.QQ (sql)
 import NtfClient
 import SMPAgentClient (agentCfg, initAgentServers, initAgentServers2, testDB, testDB2, testNtfServer, testNtfServer2)
-import SMPClient (cfg, cfgVPrev, testPort, testPort2, testStoreLogFile2, withSmpServer, withSmpServerConfigOn, withSmpServerStoreLogOn)
+import SMPClient (cfg, cfgVPrev, testPort, testPort2, testStoreLogFile2, testStoreMsgsDir2, withSmpServer, withSmpServerConfigOn, withSmpServerStoreLogOn, withSmpServerStoreMsgLogOn)
 import Simplex.Messaging.Agent hiding (createConnection, joinConnection, sendMessage)
 import Simplex.Messaging.Agent.Client (ProtocolTestFailure (..), ProtocolTestStep (..), withStore')
 import Simplex.Messaging.Agent.Env.SQLite (AgentConfig, Env (..), InitialAgentServers)
@@ -134,9 +136,8 @@ notificationTests t = do
           withNtfServer t $ testChangeToken apns
   describe "Notifications server store log" $
     it "should save and restore tokens and subscriptions" $
-      withSmpServer t $
-        withAPNSMockServer $ \apns ->
-          testNotificationsStoreLog t apns
+      withAPNSMockServer $ \apns ->
+        testNotificationsStoreLog t apns
   describe "Notifications after SMP server restart" $
     it "should resume subscriptions after SMP server is restarted" $
       withAPNSMockServer $ \apns ->
@@ -166,7 +167,7 @@ testNtfMatrix t runTest = do
     it "curr servers; curr clients" $ runNtfTestCfg t 1 cfg ntfServerCfg agentCfg agentCfg runTest
     it "curr servers; prev clients" $ runNtfTestCfg t 3 cfg ntfServerCfg agentCfgVPrevPQ agentCfgVPrevPQ runTest
     it "prev servers; prev clients" $ runNtfTestCfg t 3 cfgVPrev ntfServerCfgVPrev agentCfgVPrevPQ agentCfgVPrevPQ runTest
-    it "prev servers; curr clients" $ runNtfTestCfg t 3 cfgVPrev ntfServerCfgVPrev agentCfg agentCfg runTest
+    it "prev servers; curr clients" $ runNtfTestCfg t 1 cfgVPrev ntfServerCfgVPrev agentCfg agentCfg runTest
     -- servers can be upgraded in any order
     it "servers: curr SMP, prev NTF; prev clients" $ runNtfTestCfg t 3 cfg ntfServerCfgVPrev agentCfgVPrevPQ agentCfgVPrevPQ runTest
     it "servers: prev SMP, curr NTF; prev clients" $ runNtfTestCfg t 3 cfgVPrev ntfServerCfg agentCfgVPrevPQ agentCfgVPrevPQ runTest
@@ -178,20 +179,19 @@ runNtfTestCfg :: HasCallStack => ATransport -> AgentMsgId -> ServerConfig -> Ntf
 runNtfTestCfg t baseId smpCfg ntfCfg aCfg bCfg runTest = do
   withSmpServerConfigOn t smpCfg testPort $ \_ ->
     withAPNSMockServer $ \apns ->
-      withNtfServerCfg ntfCfg {transports = [(ntfTestPort, t)]} $ \_ ->
+      withNtfServerCfg ntfCfg {transports = [(ntfTestPort, t, False)]} $ \_ ->
         withAgentClientsCfg2 aCfg bCfg $ runTest apns baseId
   threadDelay 100000
 
 testNotificationToken :: APNSMockServer -> IO ()
-testNotificationToken APNSMockServer {apnsQ} = do
+testNotificationToken apns = do
   withAgent 1 agentCfg initAgentServers testDB $ \a -> runRight_ $ do
     let tkn = DeviceToken PPApnsTest "abcd"
     NTRegistered <- registerNtfToken a tkn NMPeriodic
-    APNSMockRequest {notification = APNSNotification {aps = APNSBackground _, notificationData = Just ntfData}, sendApnsResponse} <-
-      atomically $ readTBQueue apnsQ
+    APNSMockRequest {notification = APNSNotification {aps = APNSBackground _, notificationData = Just ntfData}} <-
+      getMockNotification apns tkn
     verification <- ntfData .-> "verification"
     nonce <- C.cbNonce <$> ntfData .-> "nonce"
-    liftIO $ sendApnsResponse APNSRespOk
     verifyNtfToken a tkn nonce verification
     NTActive <- checkNtfToken a tkn
     deleteNtfToken a tkn
@@ -208,48 +208,44 @@ v .-> key = do
 -- logCfg = LogConfig {lc_file = Nothing, lc_stderr = True}
 
 testNtfTokenRepeatRegistration :: APNSMockServer -> IO ()
-testNtfTokenRepeatRegistration APNSMockServer {apnsQ} = do
+testNtfTokenRepeatRegistration apns = do
   -- setLogLevel LogError -- LogDebug
   -- withGlobalLogging logCfg $ do
   withAgent 1 agentCfg initAgentServers testDB $ \a -> runRight_ $ do
     let tkn = DeviceToken PPApnsTest "abcd"
     NTRegistered <- registerNtfToken a tkn NMPeriodic
-    APNSMockRequest {notification = APNSNotification {aps = APNSBackground _, notificationData = Just ntfData}, sendApnsResponse} <-
-      atomically $ readTBQueue apnsQ
+    APNSMockRequest {notification = APNSNotification {aps = APNSBackground _, notificationData = Just ntfData}} <-
+      getMockNotification apns tkn
     verification <- ntfData .-> "verification"
     nonce <- C.cbNonce <$> ntfData .-> "nonce"
-    liftIO $ sendApnsResponse APNSRespOk
     NTRegistered <- registerNtfToken a tkn NMPeriodic
-    APNSMockRequest {notification = APNSNotification {aps = APNSBackground _, notificationData = Just ntfData'}, sendApnsResponse = sendApnsResponse'} <-
-      atomically $ readTBQueue apnsQ
+    APNSMockRequest {notification = APNSNotification {aps = APNSBackground _, notificationData = Just ntfData'}} <-
+      getMockNotification apns tkn
     _ <- ntfData' .-> "verification"
     _ <- C.cbNonce <$> ntfData' .-> "nonce"
-    liftIO $ sendApnsResponse' APNSRespOk
     -- can still use the first verification code, it is the same after decryption
     verifyNtfToken a tkn nonce verification
     NTActive <- checkNtfToken a tkn
     pure ()
 
 testNtfTokenSecondRegistration :: APNSMockServer -> IO ()
-testNtfTokenSecondRegistration APNSMockServer {apnsQ} =
+testNtfTokenSecondRegistration apns =
   -- setLogLevel LogError -- LogDebug
   -- withGlobalLogging logCfg $ do
   withAgentClients2 $ \a a' -> runRight_ $ do
     let tkn = DeviceToken PPApnsTest "abcd"
     NTRegistered <- registerNtfToken a tkn NMPeriodic
-    APNSMockRequest {notification = APNSNotification {aps = APNSBackground _, notificationData = Just ntfData}, sendApnsResponse} <-
-      atomically $ readTBQueue apnsQ
+    APNSMockRequest {notification = APNSNotification {aps = APNSBackground _, notificationData = Just ntfData}} <-
+      getMockNotification apns tkn
     verification <- ntfData .-> "verification"
     nonce <- C.cbNonce <$> ntfData .-> "nonce"
-    liftIO $ sendApnsResponse APNSRespOk
     verifyNtfToken a tkn nonce verification
 
     NTRegistered <- registerNtfToken a' tkn NMPeriodic
-    APNSMockRequest {notification = APNSNotification {aps = APNSBackground _, notificationData = Just ntfData'}, sendApnsResponse = sendApnsResponse'} <-
-      atomically $ readTBQueue apnsQ
+    APNSMockRequest {notification = APNSNotification {aps = APNSBackground _, notificationData = Just ntfData'}} <-
+      getMockNotification apns tkn
     verification' <- ntfData' .-> "verification"
     nonce' <- C.cbNonce <$> ntfData' .-> "nonce"
-    liftIO $ sendApnsResponse' APNSRespOk
 
     -- at this point the first token is still active
     NTActive <- checkNtfToken a tkn
@@ -265,14 +261,13 @@ testNtfTokenSecondRegistration APNSMockServer {apnsQ} =
     pure ()
 
 testNtfTokenServerRestart :: ATransport -> APNSMockServer -> IO ()
-testNtfTokenServerRestart t APNSMockServer {apnsQ} = do
+testNtfTokenServerRestart t apns = do
   let tkn = DeviceToken PPApnsTest "abcd"
   ntfData <- withAgent 1 agentCfg initAgentServers testDB $ \a ->
     withNtfServerStoreLog t $ \_ -> runRight $ do
       NTRegistered <- registerNtfToken a tkn NMPeriodic
-      APNSMockRequest {notification = APNSNotification {aps = APNSBackground _, notificationData = Just ntfData}, sendApnsResponse} <-
-        atomically $ readTBQueue apnsQ
-      liftIO $ sendApnsResponse APNSRespOk
+      APNSMockRequest {notification = APNSNotification {aps = APNSBackground _, notificationData = Just ntfData}} <-
+        getMockNotification apns tkn
       pure ntfData
   -- the new agent is created as otherwise when running the tests in CI the old agent was keeping the connection to the server
   threadDelay 1000000
@@ -287,14 +282,13 @@ testNtfTokenServerRestart t APNSMockServer {apnsQ} = do
       pure ()
 
 testNtfTokenServerRestartReverify :: ATransport -> APNSMockServer -> IO ()
-testNtfTokenServerRestartReverify t APNSMockServer {apnsQ} = do
+testNtfTokenServerRestartReverify t apns = do
   let tkn = DeviceToken PPApnsTest "abcd"
   withAgent 1 agentCfg initAgentServers testDB $ \a -> do
     ntfData <- withNtfServerStoreLog t $ \_ -> runRight $ do
       NTRegistered <- registerNtfToken a tkn NMPeriodic
-      APNSMockRequest {notification = APNSNotification {aps = APNSBackground _, notificationData = Just ntfData}, sendApnsResponse} <-
-        atomically $ readTBQueue apnsQ
-      liftIO $ sendApnsResponse APNSRespOk
+      APNSMockRequest {notification = APNSNotification {aps = APNSBackground _, notificationData = Just ntfData}} <-
+        getMockNotification apns tkn
       pure ntfData
     runRight_ $ do
       verification <- ntfData .-> "verification"
@@ -311,14 +305,13 @@ testNtfTokenServerRestartReverify t APNSMockServer {apnsQ} = do
       pure ()
 
 testNtfTokenServerRestartReverifyTimeout :: ATransport -> APNSMockServer -> IO ()
-testNtfTokenServerRestartReverifyTimeout t APNSMockServer {apnsQ} = do
+testNtfTokenServerRestartReverifyTimeout t apns = do
   let tkn = DeviceToken PPApnsTest "abcd"
   withAgent 1 agentCfg initAgentServers testDB $ \a@AgentClient {agentEnv = Env {store}} -> do
     (nonce, verification) <- withNtfServerStoreLog t $ \_ -> runRight $ do
       NTRegistered <- registerNtfToken a tkn NMPeriodic
-      APNSMockRequest {notification = APNSNotification {aps = APNSBackground _, notificationData = Just ntfData}, sendApnsResponse} <-
-        atomically $ readTBQueue apnsQ
-      liftIO $ sendApnsResponse APNSRespOk
+      APNSMockRequest {notification = APNSNotification {aps = APNSBackground _, notificationData = Just ntfData}} <-
+        getMockNotification apns tkn
       verification <- ntfData .-> "verification"
       nonce <- C.cbNonce <$> ntfData .-> "nonce"
       verifyNtfToken a tkn nonce verification
@@ -347,14 +340,14 @@ testNtfTokenServerRestartReverifyTimeout t APNSMockServer {apnsQ} = do
       pure ()
 
 testNtfTokenServerRestartReregister :: ATransport -> APNSMockServer -> IO ()
-testNtfTokenServerRestartReregister t APNSMockServer {apnsQ} = do
+testNtfTokenServerRestartReregister t apns = do
   let tkn = DeviceToken PPApnsTest "abcd"
   withAgent 1 agentCfg initAgentServers testDB $ \a ->
     withNtfServerStoreLog t $ \_ -> runRight $ do
       NTRegistered <- registerNtfToken a tkn NMPeriodic
-      APNSMockRequest {notification = APNSNotification {aps = APNSBackground _, notificationData = Just _}, sendApnsResponse} <-
-        atomically $ readTBQueue apnsQ
-      liftIO $ sendApnsResponse APNSRespOk
+      APNSMockRequest {notification = APNSNotification {aps = APNSBackground _, notificationData = Just _}} <-
+        getMockNotification apns tkn
+      pure ()
   -- the new agent is created as otherwise when running the tests in CI the old agent was keeping the connection to the server
   threadDelay 1000000
   withAgent 2 agentCfg initAgentServers testDB $ \a' ->
@@ -362,9 +355,8 @@ testNtfTokenServerRestartReregister t APNSMockServer {apnsQ} = do
     -- so that repeat registration happens when client is restarted.
     withNtfServerStoreLog t $ \_ -> runRight_ $ do
       NTRegistered <- registerNtfToken a' tkn NMPeriodic
-      APNSMockRequest {notification = APNSNotification {aps = APNSBackground _, notificationData = Just ntfData}, sendApnsResponse} <-
-        atomically $ readTBQueue apnsQ
-      liftIO $ sendApnsResponse APNSRespOk
+      APNSMockRequest {notification = APNSNotification {aps = APNSBackground _, notificationData = Just ntfData}} <-
+        getMockNotification apns tkn
       verification <- ntfData .-> "verification"
       nonce <- C.cbNonce <$> ntfData .-> "nonce"
       verifyNtfToken a' tkn nonce verification
@@ -372,14 +364,14 @@ testNtfTokenServerRestartReregister t APNSMockServer {apnsQ} = do
       pure ()
 
 testNtfTokenServerRestartReregisterTimeout :: ATransport -> APNSMockServer -> IO ()
-testNtfTokenServerRestartReregisterTimeout t APNSMockServer {apnsQ} = do
+testNtfTokenServerRestartReregisterTimeout t apns = do
   let tkn = DeviceToken PPApnsTest "abcd"
   withAgent 1 agentCfg initAgentServers testDB $ \a@AgentClient {agentEnv = Env {store}} -> do
     withNtfServerStoreLog t $ \_ -> runRight $ do
       NTRegistered <- registerNtfToken a tkn NMPeriodic
-      APNSMockRequest {notification = APNSNotification {aps = APNSBackground _, notificationData = Just _}, sendApnsResponse} <-
-        atomically $ readTBQueue apnsQ
-      liftIO $ sendApnsResponse APNSRespOk
+      APNSMockRequest {notification = APNSNotification {aps = APNSBackground _, notificationData = Just _}} <-
+        getMockNotification apns tkn
+      pure ()
     -- this emulates the situation when server registered token but the client did not receive the response
     withTransaction store $ \db ->
       DB.execute
@@ -398,9 +390,8 @@ testNtfTokenServerRestartReregisterTimeout t APNSMockServer {apnsQ} = do
     -- so that repeat registration happens when client is restarted.
     withNtfServerStoreLog t $ \_ -> runRight_ $ do
       NTRegistered <- registerNtfToken a' tkn NMPeriodic
-      APNSMockRequest {notification = APNSNotification {aps = APNSBackground _, notificationData = Just ntfData}, sendApnsResponse} <-
-        atomically $ readTBQueue apnsQ
-      liftIO $ sendApnsResponse APNSRespOk
+      APNSMockRequest {notification = APNSNotification {aps = APNSBackground _, notificationData = Just ntfData}} <-
+        getMockNotification apns tkn
       verification <- ntfData .-> "verification"
       nonce <- C.cbNonce <$> ntfData .-> "nonce"
       verifyNtfToken a' tkn nonce verification
@@ -414,18 +405,17 @@ getTestNtfTokenPort a =
     Nothing -> error "no active NtfToken"
 
 testNtfTokenMultipleServers :: ATransport -> APNSMockServer -> IO ()
-testNtfTokenMultipleServers t APNSMockServer {apnsQ} = do
+testNtfTokenMultipleServers t apns = do
   let tkn = DeviceToken PPApnsTest "abcd"
   withAgent 1 agentCfg initAgentServers2 testDB $ \a ->
     withNtfServerThreadOn t ntfTestPort $ \ntf ->
       withNtfServerThreadOn t ntfTestPort2 $ \ntf2 -> runRight_ $ do
         -- register a new token, the agent picks a server and stores its choice
         NTRegistered <- registerNtfToken a tkn NMPeriodic
-        APNSMockRequest {notification = APNSNotification {aps = APNSBackground _, notificationData = Just ntfData}, sendApnsResponse} <-
-          atomically $ readTBQueue apnsQ
+        APNSMockRequest {notification = APNSNotification {aps = APNSBackground _, notificationData = Just ntfData}} <-
+          getMockNotification apns tkn
         verification <- ntfData .-> "verification"
         nonce <- C.cbNonce <$> ntfData .-> "nonce"
-        liftIO $ sendApnsResponse APNSRespOk
         verifyNtfToken a tkn nonce verification
         NTActive <- checkNtfToken a tkn
         -- shut down the "other" server
@@ -439,10 +429,10 @@ testNtfTokenMultipleServers t APNSMockServer {apnsQ} = do
         pure ()
 
 testNtfTokenChangeServers :: ATransport -> APNSMockServer -> IO ()
-testNtfTokenChangeServers t APNSMockServer {apnsQ} =
+testNtfTokenChangeServers t apns =
   withNtfServerThreadOn t ntfTestPort $ \ntf -> do
     tkn1 <- withAgent 1 agentCfg initAgentServers testDB $ \a -> runRight $ do
-      tkn <- registerTestToken a "abcd" NMInstant apnsQ
+      tkn <- registerTestToken a "abcd" NMInstant apns
       NTActive <- checkNtfToken a tkn
       liftIO $ setNtfServers a [testNtfServer2]
       NTActive <- checkNtfToken a tkn -- still works on old server
@@ -457,15 +447,15 @@ testNtfTokenChangeServers t APNSMockServer {apnsQ} =
         liftIO $ setNtfServers a [testNtfServer2] -- just change configured server list
         getTestNtfTokenPort a >>= \port -> liftIO $ port `shouldBe` ntfTestPort -- not yet changed
         -- trigger token replace
-        tkn2 <- registerTestToken a "xyzw" NMInstant apnsQ
+        tkn2 <- registerTestToken a "xyzw" NMInstant apns
         getTestNtfTokenPort a >>= \port -> liftIO $ port `shouldBe` ntfTestPort -- not yet changed
         deleteNtfToken a tkn2 -- force server switch
-        Left BROKER {brokerErr = NETWORK} <- tryError $ registerTestToken a "qwer" NMInstant apnsQ -- ok, it's down for now
+        Left BROKER {brokerErr = NETWORK} <- tryError $ registerTestToken a "qwer" NMInstant apns -- ok, it's down for now
         getTestNtfTokenPort a >>= \port2 -> liftIO $ port2 `shouldBe` ntfTestPort2 -- but the token got updated
       killThread ntf
       withNtfServerOn t ntfTestPort2 $ runRight_ $ do
         liftIO $ threadDelay 1000000 -- for notification server to reconnect
-        tkn <- registerTestToken a "qwer" NMInstant apnsQ
+        tkn <- registerTestToken a "qwer" NMInstant apns
         checkNtfToken a tkn >>= \r -> liftIO $ r `shouldBe` NTActive
 
 testRunNTFServerTests :: ATransport -> NtfServer -> IO (Maybe ProtocolTestFailure)
@@ -475,7 +465,7 @@ testRunNTFServerTests t srv =
       testProtocolServer a 1 $ ProtoServerWithAuth srv Nothing
 
 testNotificationSubscriptionExistingConnection :: APNSMockServer -> AgentMsgId -> AgentClient -> AgentClient -> IO ()
-testNotificationSubscriptionExistingConnection APNSMockServer {apnsQ} baseId alice@AgentClient {agentEnv = Env {config = aliceCfg, store}} bob = do
+testNotificationSubscriptionExistingConnection apns baseId alice@AgentClient {agentEnv = Env {config = aliceCfg, store}} bob = do
   (bobId, aliceId, nonce, message) <- runRight $ do
     -- establish connection
     (bobId, qInfo) <- createConnection alice 1 True SCMInvitation Nothing SMSubscribe
@@ -488,11 +478,10 @@ testNotificationSubscriptionExistingConnection APNSMockServer {apnsQ} baseId ali
     -- register notification token
     let tkn = DeviceToken PPApnsTest "abcd"
     NTRegistered <- registerNtfToken alice tkn NMInstant
-    APNSMockRequest {notification = APNSNotification {aps = APNSBackground _, notificationData = Just ntfData}, sendApnsResponse} <-
-      atomically $ readTBQueue apnsQ
+    APNSMockRequest {notification = APNSNotification {aps = APNSBackground _, notificationData = Just ntfData}} <-
+      getMockNotification apns tkn
     verification <- ntfData .-> "verification"
     vNonce <- C.cbNonce <$> ntfData .-> "nonce"
-    liftIO $ sendApnsResponse APNSRespOk
     verifyNtfToken alice tkn vNonce verification
     NTActive <- checkNtfToken alice tkn
     -- send message
@@ -500,11 +489,14 @@ testNotificationSubscriptionExistingConnection APNSMockServer {apnsQ} baseId ali
     1 <- msgId <$> sendMessage bob aliceId (SMP.MsgFlags True) "hello"
     get bob ##> ("", aliceId, SENT $ baseId + 1)
     -- notification
-    (nonce, message) <- messageNotification apnsQ
+    (nonce, message) <- messageNotification apns tkn
     pure (bobId, aliceId, nonce, message)
 
-  -- alice client already has subscription for the connection
-  Left (CMD PROHIBITED _) <- runExceptT $ getNotificationMessage alice nonce message
+  Right [NotificationInfo {ntfConnId = cId}] <- runExceptT $ getNotificationConns alice nonce message
+  cId `shouldBe` bobId
+  -- alice client already has subscription for the connection,
+  -- so get fails with CMD PROHIBITED (transformed into Nothing in catch)
+  [Nothing] <- getConnectionMessages alice [cId]
 
   threadDelay 500000
   suspendAgent alice 0
@@ -513,8 +505,8 @@ testNotificationSubscriptionExistingConnection APNSMockServer {apnsQ} baseId ali
   putStrLn "before opening the database from another agent"
 
   -- aliceNtf client doesn't have subscription and is allowed to get notification message
-  withAgent 3 aliceCfg initAgentServers testDB $ \aliceNtf -> runRight_ $ do
-    (_, Just SMPMsgMeta {msgFlags = MsgFlags True}) <- getNotificationMessage aliceNtf nonce message
+  withAgent 3 aliceCfg initAgentServers testDB $ \aliceNtf -> do
+    (Just SMPMsgMeta {msgFlags = MsgFlags True}) :| _ <- getConnectionMessages aliceNtf [cId]
     pure ()
 
   threadDelay 1000000
@@ -533,65 +525,64 @@ testNotificationSubscriptionExistingConnection APNSMockServer {apnsQ} baseId ali
     2 <- msgId <$> sendMessage bob aliceId (SMP.MsgFlags True) "hello again"
     get bob ##> ("", aliceId, SENT $ baseId + 2)
     -- no notifications should follow
-    noNotification apnsQ
+    noNotification alice apns
   where
     msgId = subtract baseId
 
 testNotificationSubscriptionNewConnection :: HasCallStack => APNSMockServer -> AgentMsgId -> AgentClient -> AgentClient -> IO ()
-testNotificationSubscriptionNewConnection APNSMockServer {apnsQ} baseId alice bob =
+testNotificationSubscriptionNewConnection apns baseId alice bob =
   runRight_ $ do
     -- alice registers notification token
-    DeviceToken {} <- registerTestToken alice "abcd" NMInstant apnsQ
+    DeviceToken {} <- registerTestToken alice "abcd" NMInstant apns
     -- bob registers notification token
-    DeviceToken {} <- registerTestToken bob "bcde" NMInstant apnsQ
+    DeviceToken {} <- registerTestToken bob "bcde" NMInstant apns
     -- establish connection
     liftIO $ threadDelay 50000
     (bobId, qInfo) <- createConnection alice 1 True SCMInvitation Nothing SMSubscribe
     liftIO $ threadDelay 1000000
     (aliceId, _sqSecured) <- joinConnection bob 1 True qInfo "bob's connInfo" SMSubscribe
     liftIO $ threadDelay 750000
-    void $ messageNotificationData alice apnsQ
+    void $ messageNotificationData alice apns
     ("", _, CONF confId _ "bob's connInfo") <- get alice
     liftIO $ threadDelay 500000
     allowConnection alice bobId confId "alice's connInfo"
-    void $ messageNotificationData bob apnsQ
+    void $ messageNotificationData bob apns
     get bob ##> ("", aliceId, INFO "alice's connInfo")
-    when (baseId == 3) $ void $ messageNotificationData alice apnsQ
+    when (baseId == 3) $ void $ messageNotificationData alice apns
     get alice ##> ("", bobId, CON)
-    when (baseId == 3) $ void $ messageNotificationData bob apnsQ
+    when (baseId == 3) $ void $ messageNotificationData bob apns
     get bob ##> ("", aliceId, CON)
     -- bob sends message
     1 <- msgId <$> sendMessage bob aliceId (SMP.MsgFlags True) "hello"
     get bob ##> ("", aliceId, SENT $ baseId + 1)
-    void $ messageNotificationData alice apnsQ
+    void $ messageNotificationData alice apns
     get alice =##> \case ("", c, Msg "hello") -> c == bobId; _ -> False
     ackMessage alice bobId (baseId + 1) Nothing
     -- alice sends message
     2 <- msgId <$> sendMessage alice bobId (SMP.MsgFlags True) "hey there"
     get alice ##> ("", bobId, SENT $ baseId + 2)
-    void $ messageNotificationData bob apnsQ
+    void $ messageNotificationData bob apns
     get bob =##> \case ("", c, Msg "hey there") -> c == aliceId; _ -> False
     ackMessage bob aliceId (baseId + 2) Nothing
     -- no unexpected notifications should follow
-    noNotification apnsQ
+    noNotifications apns
   where
     msgId = subtract baseId
 
-registerTestToken :: AgentClient -> ByteString -> NotificationsMode -> TBQueue APNSMockRequest -> ExceptT AgentErrorType IO DeviceToken
-registerTestToken a token mode apnsQ = do
+registerTestToken :: AgentClient -> ByteString -> NotificationsMode -> APNSMockServer -> ExceptT AgentErrorType IO DeviceToken
+registerTestToken a token mode apns = do
   let tkn = DeviceToken PPApnsTest token
   NTRegistered <- registerNtfToken a tkn mode
-  Just APNSMockRequest {notification = APNSNotification {aps = APNSBackground _, notificationData = Just ntfData'}, sendApnsResponse = sendApnsResponse'} <-
-    timeout 1000000 . atomically $ readTBQueue apnsQ
+  Just APNSMockRequest {notification = APNSNotification {aps = APNSBackground _, notificationData = Just ntfData'}} <-
+    timeout 1000000 $ getMockNotification apns tkn
   verification' <- ntfData' .-> "verification"
   nonce' <- C.cbNonce <$> ntfData' .-> "nonce"
-  liftIO $ sendApnsResponse' APNSRespOk
   verifyNtfToken a tkn nonce' verification'
   NTActive <- checkNtfToken a tkn
   pure tkn
 
-testChangeNotificationsMode :: APNSMockServer -> IO ()
-testChangeNotificationsMode APNSMockServer {apnsQ} =
+testChangeNotificationsMode :: HasCallStack => APNSMockServer -> IO ()
+testChangeNotificationsMode apns =
   withAgentClients2 $ \alice bob -> runRight_ $ do
     -- establish connection
     (bobId, qInfo) <- createConnection alice 1 True SCMInvitation Nothing SMSubscribe
@@ -603,12 +594,12 @@ testChangeNotificationsMode APNSMockServer {apnsQ} =
     get alice ##> ("", bobId, CON)
     get bob ##> ("", aliceId, CON)
     -- register notification token, set mode to NMInstant
-    tkn <- registerTestToken alice "abcd" NMInstant apnsQ
+    tkn <- registerTestToken alice "abcd" NMInstant apns
     -- send message, receive notification
     liftIO $ threadDelay 500000
     1 <- msgId <$> sendMessage bob aliceId (SMP.MsgFlags True) "hello"
     get bob ##> ("", aliceId, SENT $ baseId + 1)
-    void $ messageNotificationData alice apnsQ
+    void $ messageNotificationData alice apns
     get alice =##> \case ("", c, Msg "hello") -> c == bobId; _ -> False
     ackMessage alice bobId (baseId + 1) Nothing
     -- set mode to NMPeriodic
@@ -617,7 +608,7 @@ testChangeNotificationsMode APNSMockServer {apnsQ} =
     liftIO $ threadDelay 750000
     2 <- msgId <$> sendMessage bob aliceId (SMP.MsgFlags True) "hello again"
     get bob ##> ("", aliceId, SENT $ baseId + 2)
-    noNotification apnsQ
+    noNotification alice apns
     get alice =##> \case ("", c, Msg "hello again") -> c == bobId; _ -> False
     ackMessage alice bobId (baseId + 2) Nothing
     -- set mode to NMInstant
@@ -626,7 +617,7 @@ testChangeNotificationsMode APNSMockServer {apnsQ} =
     liftIO $ threadDelay 500000
     3 <- msgId <$> sendMessage bob aliceId (SMP.MsgFlags True) "hello there"
     get bob ##> ("", aliceId, SENT $ baseId + 3)
-    void $ messageNotificationData alice apnsQ
+    void $ messageNotificationData alice apns
     get alice =##> \case ("", c, Msg "hello there") -> c == bobId; _ -> False
     ackMessage alice bobId (baseId + 3) Nothing
     -- turn off notifications
@@ -635,26 +626,26 @@ testChangeNotificationsMode APNSMockServer {apnsQ} =
     liftIO $ threadDelay 500000
     4 <- msgId <$> sendMessage bob aliceId (SMP.MsgFlags True) "why hello there"
     get bob ##> ("", aliceId, SENT $ baseId + 4)
-    noNotification apnsQ
+    noNotifications apns
     get alice =##> \case ("", c, Msg "why hello there") -> c == bobId; _ -> False
     ackMessage alice bobId (baseId + 4) Nothing
     -- turn on notifications, set mode to NMInstant
-    void $ registerTestToken alice "abcd" NMInstant apnsQ
+    void $ registerTestToken alice "abcd" NMInstant apns
     -- send message, receive notification
     liftIO $ threadDelay 500000
     5 <- msgId <$> sendMessage bob aliceId (SMP.MsgFlags True) "hey"
     get bob ##> ("", aliceId, SENT $ baseId + 5)
-    void $ messageNotificationData alice apnsQ
+    void $ messageNotificationData alice apns
     get alice =##> \case ("", c, Msg "hey") -> c == bobId; _ -> False
     ackMessage alice bobId (baseId + 5) Nothing
     -- no notifications should follow
-    noNotification apnsQ
+    noNotification alice apns
   where
     baseId = 1
     msgId = subtract baseId
 
 testChangeToken :: APNSMockServer -> IO ()
-testChangeToken APNSMockServer {apnsQ} = withAgent 1 agentCfg initAgentServers testDB2 $ \bob -> do
+testChangeToken apns = withAgent 1 agentCfg initAgentServers testDB2 $ \bob -> do
   (aliceId, bobId) <- withAgent 2 agentCfg initAgentServers testDB $ \alice -> runRight $ do
     -- establish connection
     (bobId, qInfo) <- createConnection alice 1 True SCMInvitation Nothing SMSubscribe
@@ -666,12 +657,12 @@ testChangeToken APNSMockServer {apnsQ} = withAgent 1 agentCfg initAgentServers t
     get alice ##> ("", bobId, CON)
     get bob ##> ("", aliceId, CON)
     -- register notification token, set mode to NMInstant
-    void $ registerTestToken alice "abcd" NMInstant apnsQ
+    void $ registerTestToken alice "abcd" NMInstant apns
     -- send message, receive notification
     liftIO $ threadDelay 500000
     1 <- msgId <$> sendMessage bob aliceId (SMP.MsgFlags True) "hello"
     get bob ##> ("", aliceId, SENT $ baseId + 1)
-    void $ messageNotificationData alice apnsQ
+    void $ messageNotificationData alice apns
     get alice =##> \case ("", c, Msg "hello") -> c == bobId; _ -> False
     ackMessage alice bobId (baseId + 1) Nothing
     pure (aliceId, bobId)
@@ -679,53 +670,66 @@ testChangeToken APNSMockServer {apnsQ} = withAgent 1 agentCfg initAgentServers t
   withAgent 3 agentCfg initAgentServers testDB $ \alice1 -> runRight_ $ do
     subscribeConnection alice1 bobId
     -- change notification token
-    void $ registerTestToken alice1 "bcde" NMInstant apnsQ
+    void $ registerTestToken alice1 "bcde" NMInstant apns
     -- send message, receive notification
     liftIO $ threadDelay 500000
     2 <- msgId <$> sendMessage bob aliceId (SMP.MsgFlags True) "hello there"
     get bob ##> ("", aliceId, SENT $ baseId + 2)
-    void $ messageNotificationData alice1 apnsQ
+    void $ messageNotificationData alice1 apns
     get alice1 =##> \case ("", c, Msg "hello there") -> c == bobId; _ -> False
     ackMessage alice1 bobId (baseId + 2) Nothing
     -- no notifications should follow
-    noNotification apnsQ
+    noNotification alice1 apns
   where
     baseId = 1
     msgId = subtract baseId
 
 testNotificationsStoreLog :: ATransport -> APNSMockServer -> IO ()
-testNotificationsStoreLog t APNSMockServer {apnsQ} = withAgentClients2 $ \alice bob -> do
-  (aliceId, bobId) <- withNtfServerStoreLog t $ \threadId -> runRight $ do
-    (aliceId, bobId) <- makeConnection alice bob
-    _ <- registerTestToken alice "abcd" NMInstant apnsQ
-    liftIO $ threadDelay 250000
-    2 <- sendMessage bob aliceId (SMP.MsgFlags True) "hello"
-    get bob ##> ("", aliceId, SENT 2)
-    void $ messageNotificationData alice apnsQ
-    get alice =##> \case ("", c, Msg "hello") -> c == bobId; _ -> False
-    ackMessage alice bobId 2 Nothing
-    liftIO $ killThread threadId
-    pure (aliceId, bobId)
+testNotificationsStoreLog t apns = withAgentClients2 $ \alice bob -> do
+  withSmpServerStoreMsgLogOn t testPort $ \_ -> do
+    (aliceId, bobId) <- withNtfServerStoreLog t $ \threadId -> runRight $ do
+      (aliceId, bobId) <- makeConnection alice bob
+      _ <- registerTestToken alice "abcd" NMInstant apns
+      liftIO $ threadDelay 250000
+      2 <- sendMessage bob aliceId (SMP.MsgFlags True) "hello"
+      get bob ##> ("", aliceId, SENT 2)
+      void $ messageNotificationData alice apns
+      get alice =##> \case ("", c, Msg "hello") -> c == bobId; _ -> False
+      ackMessage alice bobId 2 Nothing
+      liftIO $ killThread threadId
+      pure (aliceId, bobId)
 
-  liftIO $ threadDelay 250000
-
-  withNtfServerStoreLog t $ \threadId -> runRight_ $ do
     liftIO $ threadDelay 250000
-    3 <- sendMessage bob aliceId (SMP.MsgFlags True) "hello again"
-    get bob ##> ("", aliceId, SENT 3)
-    void $ messageNotificationData alice apnsQ
-    get alice =##> \case ("", c, Msg "hello again") -> c == bobId; _ -> False
-    liftIO $ killThread threadId
+
+    withNtfServerStoreLog t $ \threadId -> runRight_ $ do
+      liftIO $ threadDelay 250000
+      3 <- sendMessage bob aliceId (SMP.MsgFlags True) "hello again"
+      get bob ##> ("", aliceId, SENT 3)
+      void $ messageNotificationData alice apns
+      get alice =##> \case ("", c, Msg "hello again") -> c == bobId; _ -> False
+      ackMessage alice bobId 3 Nothing
+      liftIO $ killThread threadId
+
+    runRight_ $ do
+      4 <- sendMessage bob aliceId (SMP.MsgFlags True) "message 4"
+      get bob ##> ("", aliceId, SENT 4)
+      get alice =##> \case ("", c, Msg "message 4") -> c == bobId; _ -> False
+      ackMessage alice bobId 4 Nothing
+      noNotifications apns
+
+  withSmpServerStoreMsgLogOn t testPort $ \_ ->
+    withNtfServerStoreLog t $ \_ -> runRight_ $ do
+      void $ messageNotificationData alice apns
 
 testNotificationsSMPRestart :: ATransport -> APNSMockServer -> IO ()
-testNotificationsSMPRestart t APNSMockServer {apnsQ} = withAgentClients2 $ \alice bob -> do
+testNotificationsSMPRestart t apns = withAgentClients2 $ \alice bob -> do
   (aliceId, bobId) <- withSmpServerStoreLogOn t testPort $ \threadId -> runRight $ do
     (aliceId, bobId) <- makeConnection alice bob
-    _ <- registerTestToken alice "abcd" NMInstant apnsQ
+    _ <- registerTestToken alice "abcd" NMInstant apns
     liftIO $ threadDelay 250000
     2 <- sendMessage bob aliceId (SMP.MsgFlags True) "hello"
     get bob ##> ("", aliceId, SENT 2)
-    void $ messageNotificationData alice apnsQ
+    void $ messageNotificationData alice apns
     get alice =##> \case ("", c, Msg "hello") -> c == bobId; _ -> False
     ackMessage alice bobId 2 Nothing
     liftIO $ killThread threadId
@@ -741,22 +745,22 @@ testNotificationsSMPRestart t APNSMockServer {apnsQ} = withAgentClients2 $ \alic
     liftIO $ threadDelay 1000000
     3 <- sendMessage bob aliceId (SMP.MsgFlags True) "hello again"
     get bob ##> ("", aliceId, SENT 3)
-    _ <- messageNotificationData alice apnsQ
+    _ <- messageNotificationData alice apns
     get alice =##> \case ("", c, Msg "hello again") -> c == bobId; _ -> False
     liftIO $ killThread threadId
 
 testNotificationsSMPRestartBatch :: Int -> ATransport -> APNSMockServer -> IO ()
-testNotificationsSMPRestartBatch n t APNSMockServer {apnsQ} =
+testNotificationsSMPRestartBatch n t apns =
   withAgentClientsCfgServers2 agentCfg agentCfg initAgentServers2 $ \a b -> do
     threadDelay 1000000
     conns <- runServers $ do
       conns <- replicateM (n :: Int) $ makeConnection a b
-      _ <- registerTestToken a "abcd" NMInstant apnsQ
+      _ <- registerTestToken a "abcd" NMInstant apns
       liftIO $ threadDelay 5000000
       forM_ conns $ \(aliceId, bobId) -> do
         msgId <- sendMessage b aliceId (SMP.MsgFlags True) "hello"
         get b ##> ("", aliceId, SENT msgId)
-        void $ messageNotificationData a apnsQ
+        void $ messageNotificationData a apns
         get a =##> \case ("", c, Msg "hello") -> c == bobId; _ -> False
         ackMessage a bobId msgId Nothing
       pure conns
@@ -780,28 +784,28 @@ testNotificationsSMPRestartBatch n t APNSMockServer {apnsQ} =
       forM_ conns $ \(aliceId, bobId) -> do
         msgId <- sendMessage b aliceId (SMP.MsgFlags True) "hello again"
         get b ##> ("", aliceId, SENT msgId)
-        _ <- messageNotificationData a apnsQ
+        _ <- messageNotificationData a apns
         get a =##> \case ("", c, Msg "hello again") -> c == bobId; _ -> False
   where
     runServers :: ExceptT AgentErrorType IO a -> IO a
     runServers a = do
       withSmpServerStoreLogOn t testPort $ \t1 -> do
-        res <- withSmpServerConfigOn t (cfg :: ServerConfig) {storeLogFile = Just testStoreLogFile2} testPort2 $ \t2 ->
+        res <- withSmpServerConfigOn t (cfg :: ServerConfig) {storeLogFile = Just testStoreLogFile2, storeMsgsFile = Just testStoreMsgsDir2} testPort2 $ \t2 ->
           runRight a `finally` killThread t2
         killThread t1
         pure res
 
 testSwitchNotifications :: InitialAgentServers -> APNSMockServer -> IO ()
-testSwitchNotifications servers APNSMockServer {apnsQ} =
+testSwitchNotifications servers apns =
   withAgentClientsCfgServers2 agentCfg agentCfg servers $ \a b -> runRight_ $ do
     (aId, bId) <- makeConnection a b
     exchangeGreetings a bId b aId
-    _ <- registerTestToken a "abcd" NMInstant apnsQ
+    _ <- registerTestToken a "abcd" NMInstant apns
     liftIO $ threadDelay 250000
     let testMessage msg = do
           msgId <- sendMessage b aId (SMP.MsgFlags True) msg
           get b ##> ("", aId, SENT msgId)
-          void $ messageNotificationData a apnsQ
+          void $ messageNotificationData a apns
           get a =##> \case ("", c, Msg msg') -> c == bId && msg == msg'; _ -> False
           ackMessage a bId msgId Nothing
     testMessage "hello"
@@ -811,74 +815,80 @@ testSwitchNotifications servers APNSMockServer {apnsQ} =
     testMessage "hello again"
 
 testNotificationsOldToken :: APNSMockServer -> IO ()
-testNotificationsOldToken APNSMockServer {apnsQ} =
+testNotificationsOldToken apns =
   withAgentClients3 $ \a b c -> runRight_ $ do
     (abId, baId) <- makeConnection a b
-    let testMessageAB = testMessage_ apnsQ a abId b baId
-    _ <- registerTestToken a "abcd" NMInstant apnsQ
+    let testMessageAB = testMessage_ apns a abId b baId
+    _ <- registerTestToken a "abcd" NMInstant apns
     liftIO $ threadDelay 250000
     testMessageAB "hello"
     -- change server
     liftIO $ setNtfServers a [testNtfServer2] -- server 2 isn't running now, don't use
     -- replacing token keeps server
-    _ <- registerTestToken a "xyzw" NMInstant apnsQ
+    _ <- registerTestToken a "xyzw" NMInstant apns
     getTestNtfTokenPort a >>= \port -> liftIO $ port `shouldBe` ntfTestPort
     testMessageAB "still there"
     -- new connections keep server
     (acId, caId) <- makeConnection a c
-    let testMessageAC = testMessage_ apnsQ a acId c caId
+    let testMessageAC = testMessage_ apns a acId c caId
     testMessageAC "greetings"
 
 testNotificationsNewToken :: APNSMockServer -> ThreadId -> IO ()
-testNotificationsNewToken APNSMockServer {apnsQ} oldNtf =
+testNotificationsNewToken apns oldNtf =
   withAgentClients3 $ \a b c -> runRight_ $ do
     (abId, baId) <- makeConnection a b
-    let testMessageAB = testMessage_ apnsQ a abId b baId
-    tkn <- registerTestToken a "abcd" NMInstant apnsQ
+    let testMessageAB = testMessage_ apns a abId b baId
+    tkn <- registerTestToken a "abcd" NMInstant apns
     getTestNtfTokenPort a >>= \port -> liftIO $ port `shouldBe` ntfTestPort
     liftIO $ threadDelay 250000
     testMessageAB "hello"
     -- switch
     liftIO $ setNtfServers a [testNtfServer2]
     deleteNtfToken a tkn
-    _ <- registerTestToken a "abcd" NMInstant apnsQ
+    _ <- registerTestToken a "abcd" NMInstant apns
     getTestNtfTokenPort a >>= \port -> liftIO $ port `shouldBe` ntfTestPort2
     liftIO $ threadDelay 250000
     liftIO $ killThread oldNtf
     -- -- back to work
     testMessageAB "hello again"
     (acId, caId) <- makeConnection a c
-    let testMessageAC = testMessage_ apnsQ a acId c caId
+    let testMessageAC = testMessage_ apns a acId c caId
     testMessageAC "greetings"
 
-testMessage_ :: HasCallStack => TBQueue APNSMockRequest -> AgentClient -> ConnId -> AgentClient -> ConnId -> SMP.MsgBody -> ExceptT AgentErrorType IO ()
-testMessage_ apnsQ a aId b bId msg = do
+testMessage_ :: HasCallStack => APNSMockServer -> AgentClient -> ConnId -> AgentClient -> ConnId -> SMP.MsgBody -> ExceptT AgentErrorType IO ()
+testMessage_ apns a aId b bId msg = do
   msgId <- sendMessage b aId (SMP.MsgFlags True) msg
   get b ##> ("", aId, SENT msgId)
-  void $ messageNotificationData a apnsQ
+  void $ messageNotificationData a apns
   get a =##> \case ("", c, Msg msg') -> c == bId && msg == msg'; _ -> False
   ackMessage a bId msgId Nothing
 
-messageNotification :: HasCallStack => TBQueue APNSMockRequest -> ExceptT AgentErrorType IO (C.CbNonce, ByteString)
-messageNotification apnsQ = do
-  1000000 `timeout` atomically (readTBQueue apnsQ) >>= \case
+messageNotification :: HasCallStack => APNSMockServer -> DeviceToken -> ExceptT AgentErrorType IO (C.CbNonce, ByteString)
+messageNotification apns tkn = do
+  500000 `timeout` getMockNotification apns tkn >>= \case
     Nothing -> error "no notification"
-    Just APNSMockRequest {notification = APNSNotification {aps = APNSMutableContent {}, notificationData = Just ntfData}, sendApnsResponse} -> do
+    Just APNSMockRequest {notification = APNSNotification {aps = APNSMutableContent {}, notificationData = Just ntfData}} -> do
       nonce <- C.cbNonce <$> ntfData .-> "nonce"
       message <- ntfData .-> "message"
-      liftIO $ sendApnsResponse APNSRespOk
       pure (nonce, message)
     _ -> error "bad notification"
 
-messageNotificationData :: HasCallStack => AgentClient -> TBQueue APNSMockRequest -> ExceptT AgentErrorType IO PNMessageData
-messageNotificationData c apnsQ = do
-  (nonce, message) <- messageNotification apnsQ
-  NtfToken {ntfDhSecret = Just dhSecret} <- getNtfTokenData c
+messageNotificationData :: HasCallStack => AgentClient -> APNSMockServer -> ExceptT AgentErrorType IO PNMessageData
+messageNotificationData c apns = do
+  NtfToken {deviceToken, ntfDhSecret = Just dhSecret} <- getNtfTokenData c
+  (nonce, message) <- messageNotification apns deviceToken
   Right pnMsgs <- liftEither . first INTERNAL $ Right . parseAll pnMessagesP =<< first show (C.cbDecrypt dhSecret nonce message)
   pure $ L.last pnMsgs
 
-noNotification :: TBQueue APNSMockRequest -> ExceptT AgentErrorType IO ()
-noNotification apnsQ = do
-  500000 `timeout` atomically (readTBQueue apnsQ) >>= \case
+noNotification :: AgentClient -> APNSMockServer -> ExceptT AgentErrorType IO ()
+noNotification c apns = do
+  NtfToken {deviceToken} <- getNtfTokenData c
+  500000 `timeout` getMockNotification apns deviceToken >>= \case
+    Nothing -> pure ()
+    _ -> error "unexpected notification"
+
+noNotifications :: APNSMockServer -> ExceptT AgentErrorType IO ()
+noNotifications apns = do
+  500000 `timeout` getAnyMockNotification apns >>= \case
     Nothing -> pure ()
     _ -> error "unexpected notification"

tests/AgentTests/SQLiteTests.hs

@@ -13,7 +13,7 @@
 {-# OPTIONS_GHC -Wno-orphans #-}
 {-# OPTIONS_GHC -fno-warn-ambiguous-fields #-}
 
-module AgentTests.SQLiteTests (storeTests) where
+module AgentTests.SQLiteTests where
 
 import AgentTests.EqInstances ()
 import Control.Concurrent.Async (concurrently_)
@@ -141,6 +141,7 @@ storeTests = do
         it "should getNextDeletedSndChunkReplica" testGetNextDeletedSndChunkReplica
         it "should markNtfSubActionNtfFailed_" testMarkNtfSubActionNtfFailed
         it "should markNtfSubActionSMPFailed_" testMarkNtfSubActionSMPFailed
+        it "should markNtfTokenToDeleteFailed_" testMarkNtfTokenToDeleteFailed
   describe "open/close store" $ do
     it "should close and re-open" testCloseReopenStore
     it "should close and re-open encrypted store" testCloseReopenEncryptedStore
@@ -838,3 +839,8 @@ testMarkNtfSubActionSMPFailed :: SQLiteStore -> Expectation
 testMarkNtfSubActionSMPFailed st = do
   withTransaction st $ \db -> do
     markNtfSubActionSMPFailed_ db "abc"
+
+testMarkNtfTokenToDeleteFailed :: SQLiteStore -> Expectation
+testMarkNtfTokenToDeleteFailed st = do
+  withTransaction st $ \db -> do
+    markNtfTokenToDeleteFailed_ db 1

tests/AgentTests/SchemaDump.hs

@@ -4,6 +4,7 @@
 module AgentTests.SchemaDump where
 
 import Control.DeepSeq
+import Control.Exception (bracket_)
 import Control.Monad (unless, void)
 import Data.List (dropWhileEnd)
 import Data.Maybe (fromJust, isJust)
@@ -14,7 +15,7 @@ import Simplex.Messaging.Agent.Store.SQLite.Common (withTransaction')
 import Simplex.Messaging.Agent.Store.SQLite.Migrations (Migration (..), MigrationsToRun (..), toDownMigration)
 import qualified Simplex.Messaging.Agent.Store.SQLite.Migrations as Migrations
 import Simplex.Messaging.Util (ifM)
-import System.Directory (doesFileExist, removeFile)
+import System.Directory (createDirectoryIfMissing, doesFileExist, removeDirectoryRecursive, removeFile)
 import System.Process (readCreateProcess, shell)
 import Test.Hspec
 
@@ -24,12 +25,22 @@ testDB = "tests/tmp/test_agent_schema.db"
 appSchema :: FilePath
 appSchema = "src/Simplex/Messaging/Agent/Store/SQLite/Migrations/agent_schema.sql"
 
+-- Some indexes found by `.lint fkey-indexes` are not added to schema, explanation:
+--
+-- - CREATE INDEX 'ratchets_conn_id' ON 'ratchets'('conn_id'); --> connections(conn_id)
+--   Primary key is used instead. See for example:
+--   EXPLAIN QUERY PLAN DELETE FROM connections;
+--   (uses conn_id for ratchets table)
+appLint :: FilePath
+appLint = "src/Simplex/Messaging/Agent/Store/SQLite/Migrations/agent_lint.sql"
+
 testSchema :: FilePath
 testSchema = "tests/tmp/test_agent_schema.sql"
 
 schemaDumpTest :: Spec
 schemaDumpTest = do
   it "verify and overwrite schema dump" testVerifySchemaDump
+  it "verify .lint fkey-indexes" testVerifyLintFKeyIndexes
   it "verify schema down migrations" testSchemaMigrations
   it "should NOT create user record for new database" testUsersMigrationNew
   it "should create user record for old database" testUsersMigrationOld
@@ -42,6 +53,20 @@ testVerifySchemaDump = do
   getSchema testDB appSchema `shouldReturn` savedSchema
   removeFile testDB
 
+testVerifyLintFKeyIndexes :: IO ()
+testVerifyLintFKeyIndexes = do
+  savedLint <- ifM (doesFileExist appLint) (readFile appLint) (pure "")
+  savedLint `deepseq` pure ()
+  void $ createSQLiteStore testDB "" False Migrations.app MCConsole
+  getLintFKeyIndexes testDB "tests/tmp/agent_lint.sql" `shouldReturn` savedLint
+  removeFile testDB
+
+withTmpFiles :: IO () -> IO ()
+withTmpFiles =
+  bracket_
+    (createDirectoryIfMissing False "tests/tmp")
+    (removeDirectoryRecursive "tests/tmp")
+
 testSchemaMigrations :: IO ()
 testSchemaMigrations = do
   let noDownMigrations = dropWhileEnd (\Migration {down} -> isJust down) Migrations.app
@@ -92,7 +117,13 @@ skipComparisonForDownMigrations =
   ]
 
 getSchema :: FilePath -> FilePath -> IO String
-getSchema dpPath schemaPath = do
-  void $ readCreateProcess (shell $ "sqlite3 " <> dpPath <> " '.schema --indent' > " <> schemaPath) ""
+getSchema dbPath schemaPath = do
+  void $ readCreateProcess (shell $ "sqlite3 " <> dbPath <> " '.schema --indent' > " <> schemaPath) ""
   sch <- readFile schemaPath
   sch `deepseq` pure sch
+
+getLintFKeyIndexes :: FilePath -> FilePath -> IO String
+getLintFKeyIndexes dbPath lintPath = do
+  void $ readCreateProcess (shell $ "sqlite3 " <> dbPath <> " '.lint fkey-indexes' > " <> lintPath) ""
+  lint <- readFile lintPath
+  lint `deepseq` pure lint

tests/AgentTests/ServerChoice.hs

@@ -0,0 +1,85 @@
+{-# LANGUAGE DataKinds #-}
+{-# LANGUAGE OverloadedLists #-}
+{-# LANGUAGE OverloadedStrings #-}
+
+module AgentTests.ServerChoice where
+
+import AgentTests.FunctionalAPITests
+import Control.Monad.IO.Class
+import Data.List.NonEmpty (NonEmpty (..))
+import qualified Data.Map.Strict as M
+import SMPAgentClient
+import Simplex.Messaging.Agent (withAgentEnv)
+import Simplex.Messaging.Agent.Client hiding (userServers)
+import Simplex.Messaging.Agent.Env.SQLite
+import Simplex.Messaging.Client (defaultNetworkConfig)
+import Simplex.Messaging.Protocol
+import Test.Hspec
+import Test.QuickCheck
+import XFTPClient (testXFTPServer)
+
+serverChoiceTests :: Spec
+serverChoiceTests = do
+  describe "Server operators" $ do
+    it "should choose server of different operator" $ ioProperty $ testChooseDifferentOperator
+
+operatorSimpleX :: Maybe OperatorId
+operatorSimpleX = Just 1
+
+operator2 :: Maybe OperatorId
+operator2 = Just 2
+
+testOp1Srv1 :: ProtoServerWithAuth 'PSMP
+testOp1Srv1 = "smp://LcJU@test1.simplex.im"
+
+testOp1Srv2 :: ProtoServerWithAuth 'PSMP
+testOp1Srv2 = "smp://LcJU@test2.simplex.im"
+
+testOp2Srv1 :: ProtoServerWithAuth 'PSMP
+testOp2Srv1 = "smp://LcJU@srv1.example.com"
+
+testOp2Srv2 :: ProtoServerWithAuth 'PSMP
+testOp2Srv2 = "smp://LcJU@srv2.example.com"
+
+testSMPServers :: NonEmpty (ServerCfg 'PSMP)
+testSMPServers =
+  [ presetServerCfg True allRoles operatorSimpleX testOp1Srv1,
+    presetServerCfg True allRoles operatorSimpleX testOp1Srv2,
+    presetServerCfg True proxyOnly operator2 testOp2Srv1,
+    presetServerCfg True proxyOnly operator2 testOp2Srv2
+  ]
+
+storageOnly :: ServerRoles
+storageOnly = ServerRoles {storage = True, proxy = False}
+
+proxyOnly :: ServerRoles
+proxyOnly = ServerRoles {storage = False, proxy = True}
+
+initServers :: InitialAgentServers
+initServers =
+  InitialAgentServers
+    { smp = M.fromList [(1, testSMPServers)],
+      ntf = [testNtfServer],
+      xftp = userServers [testXFTPServer],
+      netCfg = defaultNetworkConfig
+    }
+
+testChooseDifferentOperator :: IO ()
+testChooseDifferentOperator = do
+  c <- getSMPAgentClient' 1 agentCfg initServers testDB
+  runRight_ $ do
+    -- chooses the only operator with storage role
+    srv1 <- withAgentEnv c $ getNextServer c 1 storageSrvs []
+    liftIO $ srv1 == testOp1Srv1 || srv1 == testOp1Srv2 `shouldBe` True
+    -- chooses another server for storage
+    srv2 <- withAgentEnv c $ getNextServer c 1 storageSrvs [protoServer testOp1Srv1]
+    liftIO $ srv2 `shouldBe` testOp1Srv2
+    -- chooses another operator for proxy
+    srv3 <- withAgentEnv c $ getNextServer c 1 proxySrvs [protoServer srv1]
+    liftIO $ srv3 == testOp2Srv1 || srv3 == testOp2Srv2 `shouldBe` True
+    -- chooses another operator for proxy
+    srv3' <- withAgentEnv c $ getNextServer c 1 proxySrvs [protoServer testOp1Srv1, protoServer testOp1Srv2]
+    liftIO $ srv3' == testOp2Srv1 || srv3' == testOp2Srv2 `shouldBe` True
+    -- chooses any other server
+    srv4 <- withAgentEnv c $ getNextServer c 1 proxySrvs [protoServer testOp1Srv1, protoServer testOp2Srv1]
+    liftIO $ srv4 == testOp1Srv2 || srv4 == testOp2Srv2 `shouldBe` True

tests/CLITests.hs

@@ -3,15 +3,31 @@
 
 module CLITests where
 
-import Data.Ini (lookupValue, readIniFile)
+import AgentTests.FunctionalAPITests (runRight_)
+import Control.Logger.Simple
+import Control.Monad
+import qualified Crypto.PubKey.RSA as RSA
+import qualified Data.ByteString.Lazy as BL
+import qualified Data.HashMap.Strict as HM
+import Data.Ini (Ini (..), lookupValue, readIniFile, writeIniFile)
 import Data.List (isPrefixOf)
+import qualified Data.Text as T
 import qualified Data.X509 as X
 import qualified Data.X509.File as XF
+import Data.X509.Validation (Fingerprint (..))
+import qualified Network.HTTP.Client as H1
+import qualified Network.HTTP2.Client as H2
 import Simplex.FileTransfer.Server.Main (xftpServerCLI)
+import qualified Simplex.Messaging.Crypto as C
 import Simplex.Messaging.Notifications.Server.Main
-import Simplex.Messaging.Server.Main
-import Simplex.Messaging.Transport (simplexMQVersion)
+import Simplex.Messaging.Server.Main (smpServerCLI, smpServerCLI_)
+import Simplex.Messaging.Transport (TLS (..), defaultSupportedParams, defaultSupportedParamsHTTPS, simplexMQVersion, supportedClientSMPRelayVRange)
+import Simplex.Messaging.Transport.Client (TransportClientConfig (..), defaultTransportClientConfig, runTLSTransportClient, smpClientHandshake)
+import Simplex.Messaging.Transport.HTTP2 (HTTP2Body (..))
+import qualified Simplex.Messaging.Transport.HTTP2.Client as HC
+import Simplex.Messaging.Transport.Server (loadFileFingerprint)
 import Simplex.Messaging.Util (catchAll_)
+import qualified Static
 import System.Directory (doesFileExist)
 import System.Environment (withArgs)
 import System.FilePath ((</>))
@@ -19,6 +35,10 @@ import System.IO.Silently (capture_)
 import System.Timeout (timeout)
 import Test.Hspec
 import Test.Main (withStdin)
+import UnliftIO (catchAny)
+import UnliftIO.Async (async, cancel)
+import UnliftIO.Concurrent (threadDelay)
+import UnliftIO.Exception (bracket)
 
 cfgPath :: FilePath
 cfgPath = "tests/tmp/cli/etc/opt/simplex"
@@ -26,6 +46,9 @@ cfgPath = "tests/tmp/cli/etc/opt/simplex"
 logPath :: FilePath
 logPath = "tests/tmp/cli/etc/var/simplex"
 
+webPath :: FilePath
+webPath = "tests/tmp/cli/var/www"
+
 ntfCfgPath :: FilePath
 ntfCfgPath = "tests/tmp/cli/etc/opt/simplex-notifications"
 
@@ -46,6 +69,7 @@ cliTests = do
       it "with store log, random password (default)" $ smpServerTest True True
       it "no store log, no password" $ smpServerTest False False
       it "with store log, no password" $ smpServerTest True False
+      it "static files" smpServerTestStatic
   describe "Ntf server CLI" $ do
     it "should initialize, start and delete the server (no store log)" $ ntfServerTest False
     it "should initialize, start and delete the server (with store log)" $ ntfServerTest True
@@ -61,17 +85,17 @@ smpServerTest storeLog basicAuth = do
   Right ini <- readIniFile $ cfgPath <> "/smp-server.ini"
   lookupValue "STORE_LOG" "enable" ini `shouldBe` Right (if storeLog then "on" else "off")
   lookupValue "STORE_LOG" "log_stats" ini `shouldBe` Right "off"
-  lookupValue "TRANSPORT" "port" ini `shouldBe` Right "5223"
+  lookupValue "TRANSPORT" "port" ini `shouldBe` Right "5223,443"
   lookupValue "TRANSPORT" "websockets" ini `shouldBe` Right "off"
   lookupValue "AUTH" "new_queues" ini `shouldBe` Right "on"
-  lookupValue "INACTIVE_CLIENTS" "disconnect" ini `shouldBe` Right "off"
+  lookupValue "INACTIVE_CLIENTS" "disconnect" ini `shouldBe` Right "on"
   doesFileExist (cfgPath <> "/ca.key") `shouldReturn` True
   -- start
   r <- lines <$> capture_ (withArgs ["start"] $ (100000 `timeout` smpServerCLI cfgPath logPath) `catchAll_` pure (Just ()))
   r `shouldContain` ["SMP server v" <> simplexMQVersion]
   r `shouldContain` (if storeLog then ["Store log: " <> logPath <> "/smp-server-store.log"] else ["Store log disabled."])
-  r `shouldContain` ["Listening on port 5223 (TLS)..."]
-  r `shouldContain` ["not expiring inactive clients"]
+  r `shouldContain` ["Serving SMP protocol on port 5223 (TLS)...", "Serving SMP protocol on port 443 (TLS)...", "Serving static site on port 443 (TLS)..."]
+  r `shouldContain` ["expiring clients inactive for 21600 seconds every 3600 seconds"]
   r `shouldContain` (if basicAuth then ["creating new queues requires password"] else ["creating new queues allowed"])
   -- cert
   let certPath = cfgPath </> "server.crt"
@@ -94,6 +118,70 @@ smpServerTest storeLog basicAuth = do
     >>= (`shouldSatisfy` ("WARNING: deleting the server will make all queues inaccessible" `isPrefixOf`))
   doesFileExist (cfgPath <> "/ca.key") `shouldReturn` False
 
+smpServerTestStatic :: HasCallStack => IO ()
+smpServerTestStatic = do
+  let iniFile = cfgPath <> "/smp-server.ini"
+  capture_ (withArgs ["init", "-y", "--no-password", "--web-path", webPath] $ smpServerCLI cfgPath logPath)
+    >>= (`shouldSatisfy` (("Server initialized, please provide additional server information in " <> iniFile) `isPrefixOf`))
+  doesFileExist (cfgPath <> "/ca.key") `shouldReturn` True
+  Right ini <- readIniFile iniFile
+  lookupValue "WEB" "static_path" ini `shouldBe` Right (T.pack webPath)
+  let transport = [("host", "localhost"), ("port", "5223"), ("log_tls_errors", "off"), ("websockets", "off")]
+      web = [("http", "8000"), ("https", "5223"), ("cert", "tests/fixtures/web.crt"), ("key", "tests/fixtures/web.key"), ("static_path", T.pack webPath)]
+      ini' = ini {iniSections = HM.insert "TRANSPORT" transport $ HM.insert "WEB" web (iniSections ini)}
+  writeIniFile iniFile ini'
+
+  Right ini_ <- readIniFile iniFile
+  lookupValue "WEB" "https" ini_ `shouldBe` Right "5223"
+
+  let smpServerCLI' = smpServerCLI_ Static.generateSite Static.serveStaticFiles Static.attachStaticFiles
+  let server = capture_ (withArgs ["start"] $ smpServerCLI' cfgPath logPath `catchAny` print)
+  bracket (async server) cancel $ \_t -> do
+    threadDelay 1000000
+    html <- BL.readFile $ webPath <> "/index.html"
+
+    -- "external" CA signing HTTP credentials
+    Fingerprint fpHTTP <- loadFileFingerprint "tests/fixtures/ca.crt"
+    let caHTTP = C.KeyHash fpHTTP
+    manager <- H1.newManager H1.defaultManagerSettings
+    H1.responseBody <$> H1.httpLbs "http://127.0.0.1:8000" manager `shouldReturn` html
+    logDebug "Plain HTTP works"
+
+    threadDelay 2000000
+
+    let cfgHttp = defaultTransportClientConfig {alpn = Just ["h2"], useSNI = True}
+    runTLSTransportClient defaultSupportedParamsHTTPS Nothing cfgHttp Nothing "localhost" "5223" (Just caHTTP) $ \tls -> do
+      tlsALPN tls `shouldBe` Just "h2"
+      case getCerts tls of
+        X.Certificate {X.certPubKey = X.PubKeyRSA rsa} : _ca -> RSA.public_size rsa `shouldBe` 512
+        leaf : _ -> error $ "Unexpected leaf cert: " <> show leaf
+        [] -> error "Empty chain"
+
+      let h2cfg = HC.defaultHTTP2ClientConfig {HC.bodyHeadSize = 1024 * 1024}
+      h2 <- either (error . show) pure =<< HC.attachHTTP2Client h2cfg "localhost" "5223" mempty 65536 tls
+      let req = H2.requestNoBody "GET" "/" []
+      HC.HTTP2Response {HC.respBody = HTTP2Body {bodyHead = shsBody}} <- either (error . show) pure =<< HC.sendRequest h2 req (Just 1000000)
+      BL.fromStrict shsBody `shouldBe` html
+    logDebug "Combined HTTPS works"
+
+    -- "local" CA signing SMP credentials
+    Fingerprint fpSMP <- loadFileFingerprint (cfgPath <> "/ca.crt")
+    let caSMP = C.KeyHash fpSMP
+    let cfgSmp = defaultTransportClientConfig {alpn = Just ["smp/1"], useSNI = False}
+    runTLSTransportClient defaultSupportedParams Nothing cfgSmp Nothing "localhost" "5223" (Just caSMP) $ \tls -> do
+      tlsALPN tls `shouldBe` Just "smp/1"
+      case getCerts tls of
+        X.Certificate {X.certPubKey = X.PubKeyEd25519 _k} : _ca -> print _ca -- pure ()
+        leaf : _ -> error $ "Unexpected leaf cert: " <> show leaf
+        [] -> error "Empty chain"
+      runRight_ . void $ smpClientHandshake tls Nothing caSMP supportedClientSMPRelayVRange
+    logDebug "Combined SMP works"
+  where
+    getCerts :: TLS -> [X.Certificate]
+    getCerts tls =
+      let X.CertificateChain cc = tlsServerCerts tls
+       in map (X.signedObject . X.getSigned) cc
+
 ntfServerTest :: Bool -> IO ()
 ntfServerTest storeLog = do
   capture_ (withArgs (["init"] <> ["-l" | storeLog]) $ ntfServerCLI ntfCfgPath ntfLogPath)
@@ -107,7 +195,7 @@ ntfServerTest storeLog = do
   r <- lines <$> capture_ (withArgs ["start"] $ (100000 `timeout` ntfServerCLI ntfCfgPath ntfLogPath) `catchAll_` pure (Just ()))
   r `shouldContain` ["SMP notifications server v" <> simplexMQVersion]
   r `shouldContain` (if storeLog then ["Store log: " <> ntfLogPath <> "/ntf-server-store.log"] else ["Store log disabled."])
-  r `shouldContain` ["Listening on port 443 (TLS)..."]
+  r `shouldContain` ["Serving SMP protocol on port 443 (TLS)..."]
   capture_ (withStdin "Y" . withArgs ["delete"] $ ntfServerCLI ntfCfgPath ntfLogPath)
     >>= (`shouldSatisfy` ("WARNING: deleting the server will make all queues inaccessible" `isPrefixOf`))
   doesFileExist (cfgPath <> "/ca.key") `shouldReturn` False

tests/CoreTests/BatchingTests.hs

@@ -1,4 +1,5 @@
 {-# LANGUAGE DataKinds #-}
+{-# LANGUAGE DuplicateRecordFields #-}
 {-# LANGUAGE GADTs #-}
 {-# LANGUAGE LambdaCase #-}
 {-# LANGUAGE NamedFieldPuns #-}
@@ -14,11 +15,13 @@ import Crypto.Random (ChaChaDRG)
 import qualified Data.ByteString as B
 import Data.ByteString.Char8 (ByteString)
 import qualified Data.List.NonEmpty as L
+import Data.Time.Clock.System (SystemTime, getSystemTime)
 import qualified Data.X509 as X
 import qualified Data.X509.CertificateStore as XS
 import qualified Data.X509.File as XF
 import Simplex.Messaging.Client
 import qualified Simplex.Messaging.Crypto as C
+import Simplex.Messaging.Encoding
 import Simplex.Messaging.Protocol
 import Simplex.Messaging.Transport
 import Test.Hspec
@@ -26,42 +29,43 @@ import Test.Hspec
 batchingTests :: Spec
 batchingTests = do
   describe "batchTransmissions" $ do
-    describe "SMP v6 (current)" $ do
-      it "should batch with 107 subscriptions per batch" testBatchSubscriptions
+    describe "SMP v6 (previous)" $ do
+      it "should batch with 106 subscriptions per batch" testBatchSubscriptionsV6
+      it "should break on message that does not fit" testBatchWithMessageV6
+      it "should break on large message" testBatchWithLargeMessageV6
+    describe "SMP current" $ do
+      it "should batch with 136 subscriptions per batch" testBatchSubscriptions
       it "should break on message that does not fit" testBatchWithMessage
       it "should break on large message" testBatchWithLargeMessage
-    describe "v7 (next)" $ do
-      it "should batch with 136 subscriptions per batch" testBatchSubscriptionsV7
-      it "should break on message that does not fit" testBatchWithMessageV7
-      it "should break on large message" testBatchWithLargeMessageV7
   describe "batchTransmissions'" $ do
-    describe "SMP v6 (current)" $ do
-      it "should batch with 107 subscriptions per batch" testClientBatchSubscriptions
+    describe "SMP v6 (previous)" $ do
+      it "should batch with 106 subscriptions per batch" testClientBatchSubscriptionsV6
+      it "should break on message that does not fit" testClientBatchWithMessageV6
+      it "should break on large message" testClientBatchWithLargeMessageV6
+    describe "SMP current" $ do
+      it "should batch with 136 subscriptions per batch" testClientBatchSubscriptions
+      it "should batch with 255 ENDs per batch" testClientBatchENDs
+      it "should batch with 80 NMSGs per batch" testClientBatchNMSGs
       it "should break on message that does not fit" testClientBatchWithMessage
       it "should break on large message" testClientBatchWithLargeMessage
-    describe "v7 (next)" $ do
-      it "should batch with 136 subscriptions per batch" testClientBatchSubscriptionsV7
-      it "should batch with N ENDs per batch" testClientBatchENDs
-      it "should break on message that does not fit" testClientBatchWithMessageV7
-      it "should break on large message" testClientBatchWithLargeMessageV7
 
-testBatchSubscriptions :: IO ()
-testBatchSubscriptions = do
+testBatchSubscriptionsV6 :: IO ()
+testBatchSubscriptionsV6 = do
   sessId <- atomically . C.randomBytes 32 =<< C.newRandom
-  subs <- replicateM 250 $ randomSUB sessId
+  subs <- replicateM 250 $ randomSUBv6 sessId
   let batches1 = batchTransmissions False smpBlockSize $ L.fromList subs
   all lenOk1 batches1 `shouldBe` True
   length batches1 `shouldBe` 250
   let batches = batchTransmissions True smpBlockSize $ L.fromList subs
   length batches `shouldBe` 3
   [TBTransmissions s1 n1 _, TBTransmissions s2 n2 _, TBTransmissions s3 n3 _] <- pure batches
-  (n1, n2, n3) `shouldBe` (36, 107, 107)
+  (n1, n2, n3) `shouldBe` (38, 106, 106)
   all lenOk [s1, s2, s3] `shouldBe` True
 
-testBatchSubscriptionsV7 :: IO ()
-testBatchSubscriptionsV7 = do
+testBatchSubscriptions :: IO ()
+testBatchSubscriptions = do
   sessId <- atomically . C.randomBytes 32 =<< C.newRandom
-  subs <- replicateM 300 $ randomSUBv7 sessId
+  subs <- replicateM 300 $ randomSUB sessId
   let batches1 = batchTransmissions False smpBlockSize $ L.fromList subs
   all lenOk1 batches1 `shouldBe` True
   length batches1 `shouldBe` 300
@@ -71,6 +75,22 @@ testBatchSubscriptionsV7 = do
   (n1, n2, n3) `shouldBe` (28, 136, 136)
   all lenOk [s1, s2, s3] `shouldBe` True
 
+testBatchWithMessageV6 :: IO ()
+testBatchWithMessageV6 = do
+  sessId <- atomically . C.randomBytes 32 =<< C.newRandom
+  subs1 <- replicateM 60 $ randomSUBv6 sessId
+  send <- randomSENDv6 sessId 8000
+  subs2 <- replicateM 40 $ randomSUBv6 sessId
+  let cmds = subs1 <> [send] <> subs2
+      batches1 = batchTransmissions False smpBlockSize $ L.fromList cmds
+  all lenOk1 batches1 `shouldBe` True
+  length batches1 `shouldBe` 101
+  let batches = batchTransmissions True smpBlockSize $ L.fromList cmds
+  length batches `shouldBe` 2
+  [TBTransmissions s1 n1 _, TBTransmissions s2 n2 _] <- pure batches
+  (n1, n2) `shouldBe` (47, 54)
+  all lenOk [s1, s2] `shouldBe` True
+
 testBatchWithMessage :: IO ()
 testBatchWithMessage = do
   sessId <- atomically . C.randomBytes 32 =<< C.newRandom
@@ -84,31 +104,15 @@ testBatchWithMessage = do
   let batches = batchTransmissions True smpBlockSize $ L.fromList cmds
   length batches `shouldBe` 2
   [TBTransmissions s1 n1 _, TBTransmissions s2 n2 _] <- pure batches
-  (n1, n2) `shouldBe` (47, 54)
-  all lenOk [s1, s2] `shouldBe` True
-
-testBatchWithMessageV7 :: IO ()
-testBatchWithMessageV7 = do
-  sessId <- atomically . C.randomBytes 32 =<< C.newRandom
-  subs1 <- replicateM 60 $ randomSUBv7 sessId
-  send <- randomSENDv7 sessId 8000
-  subs2 <- replicateM 40 $ randomSUBv7 sessId
-  let cmds = subs1 <> [send] <> subs2
-      batches1 = batchTransmissions False smpBlockSize $ L.fromList cmds
-  all lenOk1 batches1 `shouldBe` True
-  length batches1 `shouldBe` 101
-  let batches = batchTransmissions True smpBlockSize $ L.fromList cmds
-  length batches `shouldBe` 2
-  [TBTransmissions s1 n1 _, TBTransmissions s2 n2 _] <- pure batches
   (n1, n2) `shouldBe` (32, 69)
   all lenOk [s1, s2] `shouldBe` True
 
-testBatchWithLargeMessage :: IO ()
-testBatchWithLargeMessage = do
+testBatchWithLargeMessageV6 :: IO ()
+testBatchWithLargeMessageV6 = do
   sessId <- atomically . C.randomBytes 32 =<< C.newRandom
-  subs1 <- replicateM 50 $ randomSUB sessId
-  send <- randomSEND sessId 17000
-  subs2 <- replicateM 150 $ randomSUB sessId
+  subs1 <- replicateM 50 $ randomSUBv6 sessId
+  send <- randomSENDv6 sessId 17000
+  subs2 <- replicateM 150 $ randomSUBv6 sessId
   let cmds = subs1 <> [send] <> subs2
       batches1 = batchTransmissions False smpBlockSize $ L.fromList cmds
   all lenOk1 batches1 `shouldBe` False
@@ -119,15 +123,15 @@ testBatchWithLargeMessage = do
   let batches = batchTransmissions True smpBlockSize $ L.fromList cmds
   length batches `shouldBe` 4
   [TBTransmissions s1 n1 _, TBError TELargeMsg _, TBTransmissions s2 n2 _, TBTransmissions s3 n3 _] <- pure batches
-  (n1, n2, n3) `shouldBe` (50, 43, 107)
+  (n1, n2, n3) `shouldBe` (50, 44, 106)
   all lenOk [s1, s2, s3] `shouldBe` True
 
-testBatchWithLargeMessageV7 :: IO ()
-testBatchWithLargeMessageV7 = do
+testBatchWithLargeMessage :: IO ()
+testBatchWithLargeMessage = do
   sessId <- atomically . C.randomBytes 32 =<< C.newRandom
-  subs1 <- replicateM 60 $ randomSUBv7 sessId
-  send <- randomSENDv7 sessId 17000
-  subs2 <- replicateM 150 $ randomSUBv7 sessId
+  subs1 <- replicateM 60 $ randomSUB sessId
+  send <- randomSEND sessId 17000
+  subs2 <- replicateM 150 $ randomSUB sessId
   let cmds = subs1 <> [send] <> subs2
       batches1 = batchTransmissions False smpBlockSize $ L.fromList cmds
   all lenOk1 batches1 `shouldBe` False
@@ -141,23 +145,23 @@ testBatchWithLargeMessageV7 = do
   (n1, n2, n3) `shouldBe` (60, 14, 136)
   all lenOk [s1, s2, s3] `shouldBe` True
 
-testClientBatchSubscriptions :: IO ()
-testClientBatchSubscriptions = do
-  client <- testClientStub
-  subs <- replicateM 250 $ randomSUBCmd client
+testClientBatchSubscriptionsV6 :: IO ()
+testClientBatchSubscriptionsV6 = do
+  client <- testClientStubV6
+  subs <- replicateM 250 $ randomSUBCmdV6 client
   let batches1 = batchTransmissions' False smpBlockSize $ L.fromList subs
   all lenOk1 batches1 `shouldBe` True
   let batches = batchTransmissions' True smpBlockSize $ L.fromList subs
   length batches `shouldBe` 3
   [TBTransmissions s1 n1 rs1, TBTransmissions s2 n2 rs2, TBTransmissions s3 n3 rs3] <- pure batches
-  (n1, n2, n3) `shouldBe` (36, 107, 107)
-  (length rs1, length rs2, length rs3) `shouldBe` (36, 107, 107)
+  (n1, n2, n3) `shouldBe` (38, 106, 106)
+  (length rs1, length rs2, length rs3) `shouldBe` (38, 106, 106)
   all lenOk [s1, s2, s3] `shouldBe` True
 
-testClientBatchSubscriptionsV7 :: IO ()
-testClientBatchSubscriptionsV7 = do
-  client <- clientStubV7
-  subs <- replicateM 300 $ randomSUBCmdV7 client
+testClientBatchSubscriptions :: IO ()
+testClientBatchSubscriptions = do
+  client <- testClientStub
+  subs <- replicateM 300 $ randomSUBCmd client
   let batches1 = batchTransmissions' False smpBlockSize $ L.fromList subs
   all lenOk1 batches1 `shouldBe` True
   let batches = batchTransmissions' True smpBlockSize $ L.fromList subs
@@ -169,7 +173,7 @@ testClientBatchSubscriptionsV7 = do
 
 testClientBatchENDs :: IO ()
 testClientBatchENDs = do
-  client <- clientStubV7
+  client <- testClientStub
   ends <- replicateM 300 randomENDCmd
   let ends' = map (\t -> Right (Nothing, encodeTransmission (thParams client) t)) ends
       batches1 = batchTransmissions False smpBlockSize $ L.fromList ends'
@@ -181,6 +185,38 @@ testClientBatchENDs = do
   (length rs1, length rs2) `shouldBe` (45, 255)
   all lenOk [s1, s2] `shouldBe` True
 
+testClientBatchNMSGs :: IO ()
+testClientBatchNMSGs = do
+  client <- testClientStub
+  ts <- getSystemTime
+  ntfs <- replicateM 200 $ randomNMSGCmd ts
+  let ntfs' = map (\t -> Right (Nothing, encodeTransmission (thParams client) t)) ntfs
+      batches1 = batchTransmissions False smpBlockSize $ L.fromList ntfs'
+  all lenOk1 batches1 `shouldBe` True
+  let batches = batchTransmissions True smpBlockSize $ L.fromList ntfs'
+  length batches `shouldBe` 3
+  [TBTransmissions s1 n1 rs1, TBTransmissions s2 n2 rs2, TBTransmissions s3 n3 rs3] <- pure batches
+  (n1, n2, n3) `shouldBe` (40, 80, 80)
+  (length rs1, length rs2, length rs3) `shouldBe` (40, 80, 80)
+  all lenOk [s1, s2, s3] `shouldBe` True
+
+testClientBatchWithMessageV6 :: IO ()
+testClientBatchWithMessageV6 = do
+  client <- testClientStubV6
+  subs1 <- replicateM 60 $ randomSUBCmdV6 client
+  send <- randomSENDCmdV6 client 8000
+  subs2 <- replicateM 40 $ randomSUBCmdV6 client
+  let cmds = subs1 <> [send] <> subs2
+      batches1 = batchTransmissions' False smpBlockSize $ L.fromList cmds
+  all lenOk1 batches1 `shouldBe` True
+  length batches1 `shouldBe` 101
+  let batches = batchTransmissions' True smpBlockSize $ L.fromList cmds
+  length batches `shouldBe` 2
+  [TBTransmissions s1 n1 rs1, TBTransmissions s2 n2 rs2] <- pure batches
+  (n1, n2) `shouldBe` (47, 54)
+  (length rs1, length rs2) `shouldBe` (47, 54)
+  all lenOk [s1, s2] `shouldBe` True
+
 testClientBatchWithMessage :: IO ()
 testClientBatchWithMessage = do
   client <- testClientStub
@@ -194,33 +230,16 @@ testClientBatchWithMessage = do
   let batches = batchTransmissions' True smpBlockSize $ L.fromList cmds
   length batches `shouldBe` 2
   [TBTransmissions s1 n1 rs1, TBTransmissions s2 n2 rs2] <- pure batches
-  (n1, n2) `shouldBe` (47, 54)
-  (length rs1, length rs2) `shouldBe` (47, 54)
-  all lenOk [s1, s2] `shouldBe` True
-
-testClientBatchWithMessageV7 :: IO ()
-testClientBatchWithMessageV7 = do
-  client <- clientStubV7
-  subs1 <- replicateM 60 $ randomSUBCmdV7 client
-  send <- randomSENDCmdV7 client 8000
-  subs2 <- replicateM 40 $ randomSUBCmdV7 client
-  let cmds = subs1 <> [send] <> subs2
-      batches1 = batchTransmissions' False smpBlockSize $ L.fromList cmds
-  all lenOk1 batches1 `shouldBe` True
-  length batches1 `shouldBe` 101
-  let batches = batchTransmissions' True smpBlockSize $ L.fromList cmds
-  length batches `shouldBe` 2
-  [TBTransmissions s1 n1 rs1, TBTransmissions s2 n2 rs2] <- pure batches
   (n1, n2) `shouldBe` (32, 69)
   (length rs1, length rs2) `shouldBe` (32, 69)
   all lenOk [s1, s2] `shouldBe` True
 
-testClientBatchWithLargeMessage :: IO ()
-testClientBatchWithLargeMessage = do
-  client <- testClientStub
-  subs1 <- replicateM 50 $ randomSUBCmd client
-  send <- randomSENDCmd client 17000
-  subs2 <- replicateM 150 $ randomSUBCmd client
+testClientBatchWithLargeMessageV6 :: IO ()
+testClientBatchWithLargeMessageV6 = do
+  client <- testClientStubV6
+  subs1 <- replicateM 50 $ randomSUBCmdV6 client
+  send <- randomSENDCmdV6 client 17000
+  subs2 <- replicateM 150 $ randomSUBCmdV6 client
   let cmds = subs1 <> [send] <> subs2
       batches1 = batchTransmissions' False smpBlockSize $ L.fromList cmds
   all lenOk1 batches1 `shouldBe` False
@@ -232,24 +251,24 @@ testClientBatchWithLargeMessage = do
   let batches = batchTransmissions' True smpBlockSize $ L.fromList cmds
   length batches `shouldBe` 4
   [TBTransmissions s1 n1 rs1, TBError TELargeMsg _, TBTransmissions s2 n2 rs2, TBTransmissions s3 n3 rs3] <- pure batches
-  (n1, n2, n3) `shouldBe` (50, 43, 107)
-  (length rs1, length rs2, length rs3) `shouldBe` (50, 43, 107)
+  (n1, n2, n3) `shouldBe` (50, 44, 106)
+  (length rs1, length rs2, length rs3) `shouldBe` (50, 44, 106)
   all lenOk [s1, s2, s3] `shouldBe` True
   --
   let cmds' = [send] <> subs1 <> subs2
   let batches' = batchTransmissions' True smpBlockSize $ L.fromList cmds'
   length batches' `shouldBe` 3
   [TBError TELargeMsg _, TBTransmissions s1' n1' rs1', TBTransmissions s2' n2' rs2'] <- pure batches'
-  (n1', n2') `shouldBe` (93, 107)
-  (length rs1', length rs2') `shouldBe` (93, 107)
+  (n1', n2') `shouldBe` (94, 106)
+  (length rs1', length rs2') `shouldBe` (94, 106)
   all lenOk [s1', s2'] `shouldBe` True
 
-testClientBatchWithLargeMessageV7 :: IO ()
-testClientBatchWithLargeMessageV7 = do
-  client <- clientStubV7
-  subs1 <- replicateM 60 $ randomSUBCmdV7 client
-  send <- randomSENDCmdV7 client 17000
-  subs2 <- replicateM 150 $ randomSUBCmdV7 client
+testClientBatchWithLargeMessage :: IO ()
+testClientBatchWithLargeMessage = do
+  client <- testClientStub
+  subs1 <- replicateM 60 $ randomSUBCmd client
+  send <- randomSENDCmd client 17000
+  subs2 <- replicateM 150 $ randomSUBCmd client
   let cmds = subs1 <> [send] <> subs2
       batches1 = batchTransmissions' False smpBlockSize $ L.fromList cmds
   all lenOk1 batches1 `shouldBe` False
@@ -273,25 +292,25 @@ testClientBatchWithLargeMessageV7 = do
   (length rs1', length rs2') `shouldBe` (74, 136)
   all lenOk [s1', s2'] `shouldBe` True
 
-testClientStub :: IO (ProtocolClient SMPVersion ErrorType BrokerMsg)
-testClientStub = do
+testClientStubV6 :: IO (ProtocolClient SMPVersion ErrorType BrokerMsg)
+testClientStubV6 = do
   g <- C.newRandom
   sessId <- atomically $ C.randomBytes 32 g
   smpClientStub g sessId subModeSMPVersion Nothing
 
-clientStubV7 :: IO (ProtocolClient SMPVersion ErrorType BrokerMsg)
-clientStubV7 = do
+testClientStub :: IO (ProtocolClient SMPVersion ErrorType BrokerMsg)
+testClientStub = do
   g <- C.newRandom
   sessId <- atomically $ C.randomBytes 32 g
   (rKey, _) <- atomically $ C.generateAuthKeyPair C.SX25519 g
-  thAuth_ <- testTHandleAuth authCmdsSMPVersion g rKey
-  smpClientStub g sessId authCmdsSMPVersion thAuth_
+  thAuth_ <- testTHandleAuth currentClientSMPRelayVersion g rKey
+  smpClientStub g sessId currentClientSMPRelayVersion thAuth_
+
+randomSUBv6 :: ByteString -> IO (Either TransportError (Maybe TransmissionAuth, ByteString))
+randomSUBv6 = randomSUB_ C.SEd25519 subModeSMPVersion
 
 randomSUB :: ByteString -> IO (Either TransportError (Maybe TransmissionAuth, ByteString))
-randomSUB = randomSUB_ C.SEd25519 subModeSMPVersion
-
-randomSUBv7 :: ByteString -> IO (Either TransportError (Maybe TransmissionAuth, ByteString))
-randomSUBv7 = randomSUB_ C.SEd25519 authCmdsSMPVersion
+randomSUB = randomSUB_ C.SEd25519 currentClientSMPRelayVersion
 
 randomSUB_ :: (C.AlgorithmI a, C.AuthAlgorithm a) => C.SAlgorithm a -> VersionSMP -> ByteString -> IO (Either TransportError (Maybe TransmissionAuth, ByteString))
 randomSUB_ a v sessId = do
@@ -304,11 +323,11 @@ randomSUB_ a v sessId = do
       TransmissionForAuth {tForAuth, tToSend} = encodeTransmissionForAuth thParams (CorrId corrId, EntityId rId, Cmd SRecipient SUB)
   pure $ (,tToSend) <$> authTransmission thAuth_ (Just rpKey) nonce tForAuth
 
-randomSUBCmd :: ProtocolClient SMPVersion ErrorType BrokerMsg -> IO (PCTransmission ErrorType BrokerMsg)
-randomSUBCmd = randomSUBCmd_ C.SEd25519
+randomSUBCmdV6 :: ProtocolClient SMPVersion ErrorType BrokerMsg -> IO (PCTransmission ErrorType BrokerMsg)
+randomSUBCmdV6 = randomSUBCmd_ C.SEd25519
 
-randomSUBCmdV7 :: ProtocolClient SMPVersion ErrorType BrokerMsg -> IO (PCTransmission ErrorType BrokerMsg)
-randomSUBCmdV7 = randomSUBCmd_ C.SEd25519 -- same as v6
+randomSUBCmd :: ProtocolClient SMPVersion ErrorType BrokerMsg -> IO (PCTransmission ErrorType BrokerMsg)
+randomSUBCmd = randomSUBCmd_ C.SEd25519 -- same as v6
 
 randomSUBCmd_ :: (C.AlgorithmI a, C.AuthAlgorithm a) => C.SAlgorithm a -> ProtocolClient SMPVersion ErrorType BrokerMsg -> IO (PCTransmission ErrorType BrokerMsg)
 randomSUBCmd_ a c = do
@@ -323,11 +342,22 @@ randomENDCmd = do
   rId <- atomically $ C.randomBytes 24 g
   pure (CorrId "", EntityId rId, END)
 
-randomSEND :: ByteString -> Int -> IO (Either TransportError (Maybe TransmissionAuth, ByteString))
-randomSEND = randomSEND_ C.SEd25519 subModeSMPVersion
+randomNMSGCmd :: SystemTime -> IO (Transmission BrokerMsg)
+randomNMSGCmd ts = do
+  g <- C.newRandom
+  nId <- atomically $ C.randomBytes 24 g
+  msgId <- atomically $ C.randomBytes 24 g
+  (k, pk) <- atomically $ C.generateKeyPair g
+  nonce <- atomically $ C.randomCbNonce g
+  let msgMeta = NMsgMeta {msgId, msgTs = ts}
+  Right encNMsgMeta <- pure $ C.cbEncrypt (C.dh' k pk) nonce (smpEncode msgMeta) 128
+  pure (CorrId "", EntityId nId, NMSG nonce encNMsgMeta)
 
-randomSENDv7 :: ByteString -> Int -> IO (Either TransportError (Maybe TransmissionAuth, ByteString))
-randomSENDv7 = randomSEND_ C.SX25519 authCmdsSMPVersion
+randomSENDv6 :: ByteString -> Int -> IO (Either TransportError (Maybe TransmissionAuth, ByteString))
+randomSENDv6 = randomSEND_ C.SEd25519 subModeSMPVersion
+
+randomSEND :: ByteString -> Int -> IO (Either TransportError (Maybe TransmissionAuth, ByteString))
+randomSEND = randomSEND_ C.SX25519 currentClientSMPRelayVersion
 
 randomSEND_ :: (C.AlgorithmI a, C.AuthAlgorithm a) => C.SAlgorithm a -> VersionSMP -> ByteString -> Int -> IO (Either TransportError (Maybe TransmissionAuth, ByteString))
 randomSEND_ a v sessId len = do
@@ -350,6 +380,7 @@ testTHandleParams v sessionId =
       thServerVRange = supportedServerSMPRelayVRange,
       thAuth = Nothing,
       implySessId = v >= authCmdsSMPVersion,
+      encryptBlock = Nothing,
       batch = True
     }
 
@@ -365,11 +396,11 @@ testTHandleAuth v g (C.APublicAuthKey a serverPeerPubKey) = case a of
     pure $ Just THAuthClient {serverPeerPubKey, serverCertKey, sessSecret = Nothing}
   _ -> pure Nothing
 
-randomSENDCmd :: ProtocolClient SMPVersion ErrorType BrokerMsg -> Int -> IO (PCTransmission ErrorType BrokerMsg)
-randomSENDCmd = randomSENDCmd_ C.SEd25519
+randomSENDCmdV6 :: ProtocolClient SMPVersion ErrorType BrokerMsg -> Int -> IO (PCTransmission ErrorType BrokerMsg)
+randomSENDCmdV6 = randomSENDCmd_ C.SEd25519
 
-randomSENDCmdV7 :: ProtocolClient SMPVersion ErrorType BrokerMsg -> Int -> IO (PCTransmission ErrorType BrokerMsg)
-randomSENDCmdV7 = randomSENDCmd_ C.SX25519
+randomSENDCmd :: ProtocolClient SMPVersion ErrorType BrokerMsg -> Int -> IO (PCTransmission ErrorType BrokerMsg)
+randomSENDCmd = randomSENDCmd_ C.SX25519
 
 randomSENDCmd_ :: (C.AlgorithmI a, C.AuthAlgorithm a) => C.SAlgorithm a -> ProtocolClient SMPVersion ErrorType BrokerMsg -> Int -> IO (PCTransmission ErrorType BrokerMsg)
 randomSENDCmd_ a c len = do

tests/CoreTests/MsgStoreTests.hs

@@ -0,0 +1,414 @@
+{-# LANGUAGE DataKinds #-}
+{-# LANGUAGE DuplicateRecordFields #-}
+{-# LANGUAGE FlexibleContexts #-}
+{-# LANGUAGE GADTs #-}
+{-# LANGUAGE NamedFieldPuns #-}
+{-# LANGUAGE OverloadedStrings #-}
+{-# LANGUAGE PatternSynonyms #-}
+{-# LANGUAGE RankNTypes #-}
+{-# LANGUAGE ScopedTypeVariables #-}
+{-# LANGUAGE StandaloneDeriving #-}
+{-# LANGUAGE TypeApplications #-}
+{-# OPTIONS_GHC -fno-warn-ambiguous-fields #-}
+{-# OPTIONS_GHC -Wno-orphans #-}
+
+module CoreTests.MsgStoreTests where
+
+import AgentTests.FunctionalAPITests (runRight, runRight_)
+import Control.Concurrent.STM
+import Control.Exception (bracket)
+import Control.Monad
+import Control.Monad.IO.Class
+import Control.Monad.Trans.Except
+import Crypto.Random (ChaChaDRG)
+import Data.ByteString.Char8 (ByteString)
+import qualified Data.ByteString.Char8 as B
+import qualified Data.ByteString.Base64.URL as B64
+import Data.Maybe (fromJust)
+import Data.Time.Clock.System (getSystemTime)
+import Simplex.Messaging.Crypto (pattern MaxLenBS)
+import qualified Simplex.Messaging.Crypto as C
+import Simplex.Messaging.Protocol (EntityId (..), Message (..), RecipientId, SParty (..), noMsgFlags)
+import Simplex.Messaging.Server (MessageStats (..), exportMessages, importMessages, printMessageStats)
+import Simplex.Messaging.Server.Env.STM (journalMsgStoreDepth, readWriteQueueStore)
+import Simplex.Messaging.Server.MsgStore.Journal
+import Simplex.Messaging.Server.MsgStore.STM
+import Simplex.Messaging.Server.MsgStore.Types
+import Simplex.Messaging.Server.QueueStore
+import Simplex.Messaging.Server.QueueStore.STM
+import Simplex.Messaging.Server.StoreLog (closeStoreLog, logCreateQueue)
+import SMPClient (testStoreLogFile, testStoreMsgsDir, testStoreMsgsDir2, testStoreMsgsFile, testStoreMsgsFile2)
+import System.Directory (copyFile, createDirectoryIfMissing, listDirectory, removeFile, renameFile)
+import System.FilePath ((</>))
+import System.IO (IOMode (..), hClose, withFile)
+import Test.Hspec
+
+msgStoreTests :: Spec
+msgStoreTests = do
+  around (withMsgStore testSMTStoreConfig) $ describe "STM message store" someMsgStoreTests
+  around (withMsgStore testJournalStoreCfg) $ describe "Journal message store" $ do
+    someMsgStoreTests
+    it "should export and import journal store" testExportImportStore
+    describe "queue state" $ do
+      it "should restore queue state from the last line" testQueueState
+      it "should recover when message is written and state is not" testMessageState
+    describe "missing files" $ do
+      it "should create read file when missing" testReadFileMissing
+      it "should switch to write file when read file missing" testReadFileMissingSwitch
+      it "should create write file when missing" testWriteFileMissing
+      it "should create read file when read and write files are missing" testReadAndWriteFilesMissing
+  where
+    someMsgStoreTests :: STMQueueStore s => SpecWith s
+    someMsgStoreTests = do
+      it "should get queue and store/read messages" testGetQueue
+      it "should not fail on EOF when changing read journal" testChangeReadJournal
+
+withMsgStore :: STMQueueStore s => MsgStoreConfig s -> (s -> IO ()) -> IO ()
+withMsgStore cfg = bracket (newMsgStore cfg) closeMsgStore
+
+testSMTStoreConfig :: STMStoreConfig
+testSMTStoreConfig = STMStoreConfig {storePath = Nothing, quota = 3}
+
+testJournalStoreCfg :: JournalStoreConfig
+testJournalStoreCfg =
+  JournalStoreConfig
+    { storePath = testStoreMsgsDir,
+      pathParts = journalMsgStoreDepth,
+      quota = 3,
+      maxMsgCount = 4,
+      maxStateLines = 2,
+      stateTailSize = 256,
+      idleInterval = 21600
+    }
+
+mkMessage :: MonadIO m => ByteString -> m Message
+mkMessage body = liftIO $ do
+  g <- C.newRandom
+  msgTs <- getSystemTime
+  msgId <- atomically $ C.randomBytes 24 g
+  pure Message {msgId, msgTs, msgFlags = noMsgFlags, msgBody = C.unsafeMaxLenBS body}
+
+pattern Msg :: ByteString -> Maybe Message
+pattern Msg s <- Just Message {msgBody = MaxLenBS s}
+
+deriving instance Eq MsgQueueState
+
+deriving instance Eq (JournalState t)
+
+deriving instance Eq (SJournalType t)
+
+testNewQueueRec :: TVar ChaChaDRG -> Bool -> IO (RecipientId, QueueRec)
+testNewQueueRec g sndSecure = do
+  rId <- atomically $ EntityId <$> C.randomBytes 24 g
+  senderId <- atomically $ EntityId <$> C.randomBytes 24 g
+  (recipientKey, _) <- atomically $ C.generateAuthKeyPair C.SX25519 g
+  (k, pk) <- atomically $ C.generateKeyPair @'C.X25519 g
+  let qr =
+        QueueRec
+          { recipientId = rId,
+            recipientKey,
+            rcvDhSecret = C.dh' k pk,
+            senderId,
+            senderKey = Nothing,
+            sndSecure,
+            notifier = Nothing,
+            status = QueueActive,
+            updatedAt = Nothing
+          }
+  pure (rId, qr)
+
+testGetQueue :: STMQueueStore s => s -> IO ()
+testGetQueue ms = do
+  g <- C.newRandom
+  (rId, qr) <- testNewQueueRec g True
+  runRight_ $ do
+    q <- ExceptT $ addQueue ms qr
+    let write s = writeMsg ms rId q True =<< mkMessage s
+    Just (Message {msgId = mId1}, True) <- write "message 1"
+    Just (Message {msgId = mId2}, False) <- write "message 2"
+    Just (Message {msgId = mId3}, False) <- write "message 3"
+    Msg "message 1" <- tryPeekMsg ms rId q
+    Msg "message 1" <- tryPeekMsg ms rId q
+    Nothing <- tryDelMsg ms rId q mId2
+    Msg "message 1" <- tryDelMsg ms rId q mId1
+    Nothing <- tryDelMsg ms rId q mId1
+    Msg "message 2" <- tryPeekMsg ms rId q
+    Nothing <- tryDelMsg ms rId q mId1
+    (Nothing, Msg "message 2") <- tryDelPeekMsg ms rId q mId1
+    (Msg "message 2", Msg "message 3") <- tryDelPeekMsg ms rId q mId2
+    (Nothing, Msg "message 3") <- tryDelPeekMsg ms rId q mId2
+    Msg "message 3" <- tryPeekMsg ms rId q
+    (Msg "message 3", Nothing) <- tryDelPeekMsg ms rId q mId3
+    Nothing <- tryDelMsg ms rId q mId2
+    Nothing <- tryDelMsg ms rId q mId3
+    Nothing <- tryPeekMsg ms rId q
+    Just (Message {msgId = mId4}, True) <- write "message 4"
+    Msg "message 4" <- tryPeekMsg ms rId q
+    Just (Message {msgId = mId5}, False) <- write "message 5"
+    (Nothing, Msg "message 4") <- tryDelPeekMsg ms rId q mId3
+    (Msg "message 4", Msg "message 5") <- tryDelPeekMsg ms rId q mId4
+    Just (Message {msgId = mId6}, False) <- write "message 6"
+    Just (Message {msgId = mId7}, False) <- write "message 7"
+    Nothing <- write "message 8"
+    Msg "message 5" <- tryPeekMsg ms rId q
+    (Nothing, Msg "message 5") <- tryDelPeekMsg ms rId q mId4
+    (Msg "message 5", Msg "message 6") <- tryDelPeekMsg ms rId q mId5
+    (Msg "message 6", Msg "message 7") <- tryDelPeekMsg ms rId q mId6
+    (Msg "message 7", Just MessageQuota {msgId = mId8}) <- tryDelPeekMsg ms rId q mId7
+    (Just MessageQuota {}, Nothing) <- tryDelPeekMsg ms rId q mId8
+    (Nothing, Nothing) <- tryDelPeekMsg ms rId q mId8
+    void $ ExceptT $ deleteQueue ms rId q
+
+testChangeReadJournal :: STMQueueStore s => s -> IO ()
+testChangeReadJournal ms = do
+  g <- C.newRandom
+  (rId, qr) <- testNewQueueRec g True
+  runRight_ $ do
+    q <- ExceptT $ addQueue ms qr
+    let write s = writeMsg ms rId q True =<< mkMessage s
+    Just (Message {msgId = mId1}, True) <- write "message 1"
+    (Msg "message 1", Nothing) <- tryDelPeekMsg ms rId q mId1
+    Just (Message {msgId = mId2}, True) <- write "message 2"
+    (Msg "message 2", Nothing) <- tryDelPeekMsg ms rId q mId2
+    Just (Message {msgId = mId3}, True) <- write "message 3"
+    (Msg "message 3", Nothing) <- tryDelPeekMsg ms rId q mId3
+    Just (Message {msgId = mId4}, True) <- write "message 4"
+    (Msg "message 4", Nothing) <- tryDelPeekMsg ms rId q mId4
+    Just (Message {msgId = mId5}, True) <- write "message 5"
+    (Msg "message 5", Nothing) <- tryDelPeekMsg ms rId q mId5
+    void $ ExceptT $ deleteQueue ms rId q
+
+testExportImportStore :: JournalMsgStore -> IO ()
+testExportImportStore ms = do
+  g <- C.newRandom
+  (rId1, qr1) <- testNewQueueRec g True
+  (rId2, qr2) <- testNewQueueRec g True
+  sl <- readWriteQueueStore testStoreLogFile ms
+  runRight_ $ do
+    let write rId q s = writeMsg ms rId q True =<< mkMessage s
+    q1 <- ExceptT $ addQueue ms qr1
+    liftIO $ logCreateQueue sl qr1
+    Just (Message {}, True) <- write rId1 q1 "message 1"
+    Just (Message {}, False) <- write rId1 q1 "message 2"
+    q2 <- ExceptT $ addQueue ms qr2
+    liftIO $ logCreateQueue sl qr2
+    Just (Message {msgId = mId3}, True) <- write rId2 q2 "message 3"
+    Just (Message {msgId = mId4}, False) <- write rId2 q2 "message 4"
+    (Msg "message 3", Msg "message 4") <- tryDelPeekMsg ms rId2 q2 mId3
+    (Msg "message 4", Nothing) <- tryDelPeekMsg ms rId2 q2 mId4
+    Just (Message {}, True) <- write rId2 q2 "message 5"
+    Just (Message {}, False) <- write rId2 q2 "message 6"
+    Just (Message {}, False) <- write rId2 q2 "message 7"
+    Nothing <- write rId2 q2 "message 8"
+    pure ()
+  length <$> listDirectory (msgQueueDirectory ms rId1) `shouldReturn` 2
+  length <$> listDirectory (msgQueueDirectory ms rId2) `shouldReturn` 3
+  exportMessages False ms testStoreMsgsFile False
+  renameFile testStoreMsgsFile (testStoreMsgsFile <> ".copy")
+  closeMsgStore ms
+  closeStoreLog sl
+  exportMessages False ms testStoreMsgsFile False
+  (B.readFile testStoreMsgsFile `shouldReturn`) =<< B.readFile (testStoreMsgsFile <> ".copy")
+  let cfg = (testJournalStoreCfg :: JournalStoreConfig) {storePath = testStoreMsgsDir2}
+  ms' <- newMsgStore cfg
+  readWriteQueueStore testStoreLogFile ms' >>= closeStoreLog
+  stats@MessageStats {storedMsgsCount = 5, expiredMsgsCount = 0, storedQueues = 2} <-
+    importMessages False ms' testStoreMsgsFile Nothing
+  printMessageStats "Messages" stats
+  length <$> listDirectory (msgQueueDirectory ms rId1) `shouldReturn` 2
+  length <$> listDirectory (msgQueueDirectory ms rId2) `shouldReturn` 4 -- state file is backed up, 2 message files
+  exportMessages False ms' testStoreMsgsFile2 False
+  (B.readFile testStoreMsgsFile2 `shouldReturn`) =<< B.readFile (testStoreMsgsFile <> ".bak")
+  stmStore <- newMsgStore testSMTStoreConfig
+  readWriteQueueStore testStoreLogFile stmStore >>= closeStoreLog
+  MessageStats {storedMsgsCount = 5, expiredMsgsCount = 0, storedQueues = 2} <-
+    importMessages False stmStore testStoreMsgsFile2 Nothing
+  exportMessages False stmStore testStoreMsgsFile False
+  (B.sort <$> B.readFile testStoreMsgsFile `shouldReturn`) =<< (B.sort <$> B.readFile (testStoreMsgsFile2 <> ".bak"))
+
+testQueueState :: JournalMsgStore -> IO ()
+testQueueState ms = do
+  g <- C.newRandom
+  rId <- EntityId <$> atomically (C.randomBytes 24 g)
+  let dir = msgQueueDirectory ms rId
+      statePath = msgQueueStatePath dir $ B.unpack (B64.encode $ unEntityId rId)
+  createDirectoryIfMissing True dir
+  state <- newMsgQueueState <$> newJournalId (random ms)
+  withFile statePath WriteMode (`appendState` state)
+  length . lines <$> readFile statePath `shouldReturn` 1
+  readQueueState statePath `shouldReturn` state
+  length <$> listDirectory dir `shouldReturn` 1 -- no backup
+
+  let state1 =
+        state
+          { size = 1,
+            readState = (readState state) {msgCount = 1, byteCount = 100},
+            writeState = (writeState state) {msgPos = 1, msgCount = 1, bytePos = 100, byteCount = 100}
+          }
+  withFile statePath AppendMode (`appendState` state1)
+  length . lines <$> readFile statePath `shouldReturn` 2
+  readQueueState statePath `shouldReturn` state1
+  length <$> listDirectory dir `shouldReturn` 1 -- no backup
+
+  let state2 =
+        state
+          { size = 2,
+            readState = (readState state) {msgCount = 2, byteCount = 200},
+            writeState = (writeState state) {msgPos = 2, msgCount = 2, bytePos = 200, byteCount = 200}
+          }
+  withFile statePath AppendMode (`appendState` state2)
+  length . lines <$> readFile statePath `shouldReturn` 3
+  copyFile statePath (statePath <> ".2")
+  readQueueState statePath `shouldReturn` state2
+  length <$> listDirectory dir `shouldReturn` 3 -- new state, copy + backup
+  length . lines <$> readFile statePath `shouldReturn` 1
+
+  -- corrupt the only line
+  corruptFile statePath
+  newState <- readQueueState statePath
+  newState `shouldBe` newMsgQueueState (journalId $ writeState newState)
+
+  -- corrupt the last line
+  renameFile (statePath <> ".2") statePath
+  removeOtherFiles dir statePath
+  length . lines <$> readFile statePath `shouldReturn` 3
+  corruptFile statePath
+  readQueueState statePath `shouldReturn` state1
+  length <$> listDirectory dir `shouldReturn` 2
+  length . lines <$> readFile statePath `shouldReturn` 1
+  where
+    readQueueState statePath = do
+      (state, h) <- readWriteQueueState ms statePath
+      hClose h
+      pure state
+    corruptFile f = do
+      s <- readFile f
+      removeFile f
+      writeFile f $ take (length s - 4) s
+    removeOtherFiles dir keep = do
+      names <- listDirectory dir
+      forM_ names $ \name ->
+        let f = dir </> name
+         in unless (f == keep) $ removeFile f
+
+testMessageState :: JournalMsgStore -> IO ()
+testMessageState ms = do
+  g <- C.newRandom
+  (rId, qr) <- testNewQueueRec g True
+  let dir = msgQueueDirectory ms rId
+      statePath = msgQueueStatePath dir $ B.unpack (B64.encode $ unEntityId rId)
+      write q s = writeMsg ms rId q True =<< mkMessage s
+
+  mId1 <- runRight $ do
+    q <- ExceptT $ addQueue ms qr
+    Just (Message {msgId = mId1}, True) <- write q "message 1"
+    Just (Message {}, False) <- write q "message 2"
+    liftIO $ closeMsgQueue q
+    pure mId1
+
+  ls <- B.lines <$> B.readFile statePath
+  B.writeFile statePath $ B.unlines $ take (length ls - 1) ls
+
+  runRight_ $ do
+    q <- ExceptT $ getQueue ms SRecipient rId
+    Just (Message {msgId = mId3}, False) <- write q "message 3"
+    (Msg "message 1", Msg "message 3") <- tryDelPeekMsg ms rId q mId1
+    (Msg "message 3", Nothing) <- tryDelPeekMsg ms rId q mId3
+    liftIO $ closeMsgQueue q
+
+testReadFileMissing :: JournalMsgStore -> IO ()
+testReadFileMissing ms = do
+  g <- C.newRandom
+  (rId, qr) <- testNewQueueRec g True
+  let write q s = writeMsg ms rId q True =<< mkMessage s
+  q <- runRight $ do
+    q <- ExceptT $ addQueue ms qr
+    Just (Message {}, True) <- write q "message 1"
+    Msg "message 1" <- tryPeekMsg ms rId q
+    pure q
+
+  mq <- fromJust <$> readTVarIO (msgQueue_' q)
+  MsgQueueState {readState = rs} <- readTVarIO $ state mq
+  closeMsgStore ms
+  let path = journalFilePath (queueDirectory $ queue mq) $ journalId rs
+  removeFile path
+
+  runRight_ $ do
+    q' <- ExceptT $ getQueue ms SRecipient rId
+    Nothing <- tryPeekMsg ms rId q'
+    Just (Message {}, True) <- write q' "message 2"
+    Msg "message 2" <- tryPeekMsg ms rId q'
+    pure ()
+
+testReadFileMissingSwitch :: JournalMsgStore -> IO ()
+testReadFileMissingSwitch ms = do
+  g <- C.newRandom
+  (rId, qr) <- testNewQueueRec g True
+  q <- writeMessages ms rId qr
+
+  mq <- fromJust <$> readTVarIO (msgQueue_' q)
+  MsgQueueState {readState = rs} <- readTVarIO $ state mq
+  closeMsgStore ms
+  let path = journalFilePath (queueDirectory $ queue mq) $ journalId rs
+  removeFile path
+
+  runRight_ $ do
+    q' <- ExceptT $ getQueue ms SRecipient rId
+    Just (Message {}, False) <- writeMsg ms rId q' True =<< mkMessage "message 6"
+    Msg "message 5" <- tryPeekMsg ms rId q'
+    pure ()
+
+testWriteFileMissing :: JournalMsgStore -> IO ()
+testWriteFileMissing ms = do
+  g <- C.newRandom
+  (rId, qr) <- testNewQueueRec g True
+  q <- writeMessages ms rId qr
+
+  mq <- fromJust <$> readTVarIO (msgQueue_' q)
+  MsgQueueState {writeState = ws} <- readTVarIO $ state mq
+  closeMsgStore ms
+  let path = journalFilePath (queueDirectory $ queue mq) $ journalId ws
+  print path
+  removeFile path
+
+  runRight_ $ do
+    q' <- ExceptT $ getQueue ms SRecipient rId
+    Just Message {msgId = mId3} <- tryPeekMsg ms rId q'
+    (Msg "message 3", Msg "message 4") <- tryDelPeekMsg ms rId q' mId3
+    Just Message {msgId = mId4} <- tryPeekMsg ms rId q'
+    (Msg "message 4", Nothing) <- tryDelPeekMsg ms rId q' mId4
+    Just (Message {}, True) <- writeMsg ms rId q' True =<< mkMessage "message 6"
+    Msg "message 6" <- tryPeekMsg ms rId q'
+    pure ()
+
+testReadAndWriteFilesMissing :: JournalMsgStore -> IO ()
+testReadAndWriteFilesMissing ms = do
+  g <- C.newRandom
+  (rId, qr) <- testNewQueueRec g True
+  q <- writeMessages ms rId qr
+
+  mq <- fromJust <$> readTVarIO (msgQueue_' q)
+  MsgQueueState {readState = rs, writeState = ws} <- readTVarIO $ state mq
+  closeMsgStore ms
+  removeFile $ journalFilePath (queueDirectory $ queue mq) $ journalId rs
+  removeFile $ journalFilePath (queueDirectory $ queue mq) $ journalId ws
+
+  runRight_ $ do
+    q' <- ExceptT $ getQueue ms SRecipient rId
+    Nothing <- tryPeekMsg ms rId q'
+    Just (Message {}, True) <- writeMsg ms rId q' True =<< mkMessage "message 6"
+    Msg "message 6" <- tryPeekMsg ms rId q'
+    pure ()
+
+writeMessages :: JournalMsgStore -> RecipientId -> QueueRec -> IO JournalQueue
+writeMessages ms rId qr = runRight $ do
+  q <- ExceptT $ addQueue ms qr
+  let write s = writeMsg ms rId q True =<< mkMessage s
+  Just (Message {msgId = mId1}, True) <- write "message 1"
+  Just (Message {msgId = mId2}, False) <- write "message 2"
+  Just (Message {}, False) <- write "message 3"
+  (Msg "message 1", Msg "message 2") <- tryDelPeekMsg ms rId q mId1
+  (Msg "message 2", Msg "message 3") <- tryDelPeekMsg ms rId q mId2
+  Just (Message {}, False) <- write "message 4"
+  Just (Message {}, False) <- write "message 5"
+  pure q

tests/CoreTests/RetryIntervalTests.hs

@@ -66,7 +66,7 @@ testRetryIntervalSameMode =
 
 testRetryIntervalSwitchMode :: Spec
 testRetryIntervalSwitchMode =
-  it "should increase elapased time and interval when the mode stays the same" $ do
+  it "should increase elapased time and interval when the mode switches" $ do
     lock <- newEmptyTMVarIO
     intervals <- newTVarIO []
     reportedIntervals <- newTVarIO []