crypto: simplify entropy path after audit review

- Lift duplicated identity-gen block from main_companion.cpp +
  main_repeater.cpp into ZephyrRNG::generateFirstBootIdentity().
  Both mains shrink from ~40 lines to a 3-line helper call.
- Add LocalIdentity::fromSeed() so seed-derived keygen doesn't need
  a one-shot RNG wrapper; delete SeededRNG.
- Drop the per-byte ADC sampling loop: getBattMilliVolts() does an
  8-sample average + 10ms regulator settle internally, costing
  300-480ms of real wall-time and actively destroying the LSB jitter
  it was meant to harvest. Jitter mixer already dwarfs it.
- Centralize the printk + sys_reboot pattern as
  Utils::cryptoPanicReboot(); drop the 2000ms pre-reboot k_msleep
  (printk is synchronous, sleep just blocked the mesh thread on
  the ZephyrRNG::random() retry-failure path).
- Inline sample_cpu_jitter health check via online scalars instead
  of a 512-byte deltas[] array. Saves 1.5KB stack churn across boot
  and tracks every sample instead of only the first 128.
- extract_via_aes_ctr now uses Utils::sha256 instead of open-coding
  psa_hash_compute.
This commit is contained in:
liquidraver
2026-05-29 07:58:54 +02:00
parent b692ca72ed
commit 799d694914
8 changed files with 119 additions and 170 deletions
+63 -58
View File
@@ -27,12 +27,10 @@ void ZephyrRNG::random(uint8_t *dest, size_t sz)
if (sys_csrand_get(dest, sz) == 0) return;
k_msleep(10);
}
printk("ZephyrRNG: CSPRNG unavailable after retries — rebooting\n");
k_msleep(2000);
sys_reboot(SYS_REBOOT_COLD);
Utils::cryptoPanicReboot("CSPRNG unavailable after retries");
}
/* ===== Jitter sampling + health check =====================================
/* ===== Jitter sampling + online health check =============================
*
* Stephan Müller "CPU Time Jitter Based Non-Physical True Random Number
* Generator" — Linux kernel's jitterentropy_rng. NIST SP 800-90B has
@@ -44,55 +42,28 @@ void ZephyrRNG::random(uint8_t *dest, size_t sz)
* × 0.1 bits = 3,200 estimated bits. 256 needed for Ed25519 → 12×
* margin even pessimistically.
*
* Health check is the NIST SP 800-90B "repetition count" plus a
* variance check on the first 128 samples. Detects stuck-source
* Health check (NIST SP 800-90B style): online repetition count + a
* distinct-value check tracked across all samples in the window with
* scalar state — no per-sample buffer needed. Detects stuck-source
* catastrophic failure (e.g. cycle counter not advancing). Does not
* statistically prove entropy quality — that's what the literature is
* for. */
#define JITTER_TRACKED 128
static bool jitter_health_check(const uint32_t *deltas, size_t n)
{
if (n < 16) return false;
/* Repetition count: 32+ consecutive identical samples is failure. */
int consec = 1, max_consec = 1;
for (size_t i = 1; i < n; i++) {
if (deltas[i] == deltas[i - 1]) {
consec++;
if (consec > max_consec) max_consec = consec;
} else {
consec = 1;
}
}
if (max_consec >= 32) return false;
/* Variance: require at least 5 distinct delta values across the
* tracked window. A perfectly deterministic CPU would produce one
* value; even minimal jitter produces several. */
uint32_t distinct[8] = {0};
int n_distinct = 0;
for (size_t i = 0; i < n && n_distinct < 8; i++) {
bool found = false;
for (int j = 0; j < n_distinct; j++) {
if (distinct[j] == deltas[i]) { found = true; break; }
}
if (!found) distinct[n_distinct++] = deltas[i];
}
return n_distinct >= 5;
}
static bool sample_cpu_jitter(uint8_t *pool, size_t pool_size,
size_t pool_offset, uint32_t duration_ms)
{
uint32_t deltas[JITTER_TRACKED];
size_t tracked_idx = 0;
uint32_t accum = k_cycle_get_32();
int64_t deadline = k_uptime_get() + duration_ms;
size_t idx = pool_offset;
/* Online health stats: 32 bytes total vs. the previous 512-byte
* deltas[] array. Tracks every sample, not just the first 128. */
uint32_t prev_delta = 0;
int cur_consec = 0, max_consec = 0;
uint32_t distinct[8] = {0};
int n_distinct = 0;
int n_samples = 0;
while (k_uptime_get() < deadline) {
uint32_t t1 = k_cycle_get_32();
/* Variable-time work — number of iterations depends on the
@@ -113,10 +84,29 @@ static bool sample_cpu_jitter(uint8_t *pool, size_t pool_size,
pool[idx++ % pool_size] ^= (uint8_t)accum;
pool[idx++ % pool_size] ^= (uint8_t)(accum >> 8);
if (tracked_idx < JITTER_TRACKED) deltas[tracked_idx++] = delta;
/* Online repetition count */
if (n_samples > 0 && delta == prev_delta) {
if (++cur_consec > max_consec) max_consec = cur_consec;
} else {
cur_consec = 1;
}
prev_delta = delta;
/* Track first 8 distinct delta values */
if (n_distinct < 8) {
bool found = false;
for (int j = 0; j < n_distinct; j++) {
if (distinct[j] == delta) { found = true; break; }
}
if (!found) distinct[n_distinct++] = delta;
}
n_samples++;
}
return jitter_health_check(deltas, tracked_idx);
if (n_samples < 16) return false;
if (max_consec >= 32) return false; /* stuck source */
return n_distinct >= 5; /* minimal variance */
}
/* ===== Entropy extraction via AES-256-CTR ================================
@@ -142,18 +132,14 @@ static int extract_via_aes_ctr(const uint8_t *pool, size_t pool_len,
{
psa_status_t status;
uint8_t key[32];
size_t key_len = 0;
/* PSA is idempotent — already initialized via mbedTLS but a defensive
* call here costs nothing if it returns PSA_ERROR_ALREADY_EXISTS. */
(void)psa_crypto_init();
/* Extract: SHA-256(pool) → AES key */
status = psa_hash_compute(PSA_ALG_SHA_256, pool, pool_len,
key, sizeof(key), &key_len);
if (status != PSA_SUCCESS || key_len != sizeof(key)) {
return -1;
}
/* Extract: SHA-256(pool) → AES key. Reuses the codebase's PSA-backed
* SHA-256 wrapper instead of open-coding psa_hash_compute here. */
Utils::sha256(key, sizeof(key), pool, (int)pool_len);
/* Import key for AES-256-ECB */
psa_key_attributes_t attr = PSA_KEY_ATTRIBUTES_INIT;
@@ -246,22 +232,18 @@ void ZephyrRNG::mixIdentitySeed(uint8_t *out, size_t out_len,
* AES-ECB on a 128-bit counter. Per crypto consultant guidance —
* see extract_via_aes_ctr() for full rationale. */
if (extract_via_aes_ctr(pool, sizeof(pool), out, out_len) != 0) {
printk("ZephyrRNG: AES-CTR extraction failed — rebooting\n");
k_msleep(2000);
sys_reboot(SYS_REBOOT_COLD);
Utils::cryptoPanicReboot("AES-CTR seed extraction failed");
}
/* Output sanity check — reject all-zero / all-0xFF (catastrophic
* failure of every source). Reboot to retry. */
* failure of every source). */
bool all_zero = true, all_ff = true;
for (size_t i = 0; i < out_len; i++) {
if (out[i] != 0x00) all_zero = false;
if (out[i] != 0xFF) all_ff = false;
}
if (all_zero || all_ff) {
printk("ZephyrRNG: degenerate seed output — rebooting\n");
k_msleep(2000);
sys_reboot(SYS_REBOOT_COLD);
Utils::cryptoPanicReboot("degenerate seed output (all-zero / all-FF)");
}
/* Wipe sensitive intermediate buffers — secureZeroize survives the
@@ -271,4 +253,27 @@ void ZephyrRNG::mixIdentitySeed(uint8_t *out, size_t out_len,
Utils::secureZeroize(devid, sizeof(devid));
}
void ZephyrRNG::generateFirstBootIdentity(LocalIdentity &out_identity)
{
uint8_t seed[32];
mixIdentitySeed(seed, sizeof(seed));
out_identity.fromSeed(seed);
/* Reserved-prefix guard — MeshCore protocol treats pub_key[0] of
* 0x00/0xFF as reserved markers. With a working CSPRNG the first
* attempt almost always passes (P(reserved) = 2/256); the cap +
* panic-reboot is a stuck-source backstop. */
int attempt = 0;
while (out_identity.pub_key[0] == 0x00 || out_identity.pub_key[0] == 0xFF) {
if (++attempt > 100) {
Utils::cryptoPanicReboot("identity gen stuck on reserved prefix");
}
mixIdentitySeed(seed, sizeof(seed));
out_identity.fromSeed(seed);
}
Utils::secureZeroize(seed, sizeof(seed));
}
} /* namespace mesh */
+15 -26
View File
@@ -6,8 +6,8 @@
#pragma once
#include <mesh/RNG.h>
#include <mesh/Identity.h>
#include <stddef.h>
#include <string.h>
namespace mesh {
@@ -19,42 +19,31 @@ public:
* generation. Combines:
* 1. sys_csrand_get (early) — CSPRNG, strong on nRF/MG24
* 2. HWINFO unique device ID — per-device uniqueness
* 3. Optional caller-supplied data — e.g. ADC LSB noise samples
* 3. Optional caller-supplied data — e.g. external noise samples
* 4. CPU cycle-counter jitter — main entropy source
* (NIST SP 800-90B class)
* 5. sys_csrand_get (late) — catches mid-boot radio init
* 6. CPU cycle-counter jitter #2 — independent timing window
* Conditioned via SHA-512. NIST-style health checks on jitter
* samples; reboots on degenerate output.
* Blocks for ~250-300ms — only called at first-boot identity gen.
* Conditioned via AES-256-CTR (SHA-256(pool) → key, ECB on counter).
* NIST-style health checks on jitter samples; reboots on degenerate
* output. Blocks for ~250ms — only called at first-boot identity gen.
*
* Output is suitable as an Ed25519 seed regardless of platform
* TRNG state at boot. */
static void mixIdentitySeed(uint8_t *out, size_t out_len,
const uint8_t *extra = nullptr,
size_t extra_len = 0);
};
/* Thin wrapper that returns pre-supplied bytes from the mesh::RNG
* interface. Used to feed a mixed seed into mesh::LocalIdentity(RNG*)
* without changing the LocalIdentity API. One-shot — bytes are
* consumed sequentially across calls. */
class SeededRNG : public RNG {
public:
SeededRNG(const uint8_t *seed_data, size_t len)
: _data(seed_data), _avail(len) {}
void random(uint8_t *dest, size_t sz) override {
size_t n = (sz <= _avail) ? sz : _avail;
if (n > 0) memcpy(dest, _data, n);
_data += n;
_avail -= n;
if (n < sz) memset(dest + n, 0, sz - n);
}
private:
const uint8_t *_data;
size_t _avail;
/* End-to-end first-boot identity generation. Mixes a fresh seed,
* derives the Ed25519 keypair, and retries (up to 100×) if the
* MeshCore protocol-reserved 0x00/0xFF public-key prefix happens
* to land. Panics-and-reboots on cap exhaustion (essentially
* impossible with a working CSPRNG: P(100 reserved in a row) ≈ 10⁻²¹¹).
* Wipes the intermediate seed before return.
*
* Use this from main()'s `loadIdentity` fall-through path instead
* of open-coding the mix+derive+retry+zeroize sequence. */
static void generateFirstBootIdentity(LocalIdentity &out_identity);
};
} /* namespace mesh */
+7
View File
@@ -49,6 +49,13 @@ public:
LocalIdentity(const char *prv_hex, const char *pub_hex);
LocalIdentity(RNG *rng);
/* Derive Ed25519 keypair from a 32-byte seed. Use this when you've
* already produced a high-quality seed externally (e.g. via the
* layered ZephyrRNG::mixIdentitySeed entropy mixer) — avoids the
* one-shot-RNG-wrapper dance otherwise needed to feed bytes
* through the LocalIdentity(RNG*) constructor. */
void fromSeed(const uint8_t seed[SEED_SIZE]);
void sign(uint8_t *sig, const uint8_t *message, int msg_len) const;
void calcSharedSecret(uint8_t *secret, const Identity &other) const { calcSharedSecret(secret, other.pub_key); }
void calcSharedSecret(uint8_t *secret, const uint8_t *other_pub_key) const;
+7
View File
@@ -34,6 +34,13 @@ public:
* secrets after their last use. */
static void secureZeroize(void *buf, size_t n);
/* Log a crypto-invariant failure to printk and cold-reboot.
* Used when an entropy source, KDF, or other primitive cannot
* produce a safe result — proceeding would risk weak keys or
* bypassed authentication, so we restart rather than continue.
* Does not return. */
[[noreturn]] static void cryptoPanicReboot(const char *msg);
static void toHex(char *dest, const uint8_t *src, size_t len);
static bool fromHex(uint8_t *dest, int dest_size, const char *src_hex);
static bool isHexChar(char c);
+6
View File
@@ -54,6 +54,12 @@ LocalIdentity::LocalIdentity(RNG *rng)
uint8_t seed[SEED_SIZE];
rng->random(seed, SEED_SIZE);
ed25519_create_keypair(pub_key, prv_key, seed);
Utils::secureZeroize(seed, sizeof(seed));
}
void LocalIdentity::fromSeed(const uint8_t seed[SEED_SIZE])
{
ed25519_create_keypair(pub_key, prv_key, seed);
}
bool LocalIdentity::validatePrivateKey(const uint8_t prv[64])
+13
View File
@@ -6,6 +6,8 @@
#include <mesh/Utils.h>
#include <psa/crypto.h>
#include <string.h>
#include <zephyr/sys/printk.h>
#include <zephyr/sys/reboot.h>
#include <zephyr/logging/log.h>
LOG_MODULE_REGISTER(zephcore_utils, CONFIG_ZEPHCORE_MAIN_LOG_LEVEL);
@@ -205,6 +207,17 @@ bool Utils::constantTimeEqual(const void *a, const void *b, size_t n)
return result == 0;
}
void Utils::cryptoPanicReboot(const char *msg)
{
/* No pre-reboot k_msleep: printk is synchronous on RTT/UART so the
* line is already on the wire by the time sys_reboot fires, and the
* 2-second delay we used to do here just blocked the mesh thread on
* the rare-but-realistic ZephyrRNG::random() retry failure path. */
printk("crypto panic: %s — rebooting\n", msg ? msg : "(no detail)");
sys_reboot(SYS_REBOOT_COLD);
for (;;) { /* sys_reboot is FUNC_NORETURN, but satisfy [[noreturn]] */ }
}
void Utils::secureZeroize(void *buf, size_t n)
{
/* Volatile pointer prevents the compiler from eliminating the
+4 -42
View File
@@ -660,51 +660,13 @@ int main(void)
LOG_INF("Added default Public channel");
}
/* Load or generate identity.
*
* First-boot keygen uses ZephyrRNG::mixIdentitySeed() — a layered
* entropy mixer combining sys_csrand_get + HWINFO unique ID + ADC
* LSB noise + CPU cycle-counter jitter, conditioned via SHA-512.
* This compensates for ESP32's hardware TRNG being only seeded
* after WiFi/BT radio init (identity gen happens before bt_enable). */
/* Load or generate identity. First-boot keygen runs the layered
* entropy mixer + Ed25519 derive + reserved-prefix guard inside
* ZephyrRNG::generateFirstBootIdentity. */
mesh::LocalIdentity self_identity;
if (!data_store.loadMainIdentity(self_identity)) {
/* Sample ADC LSB noise — best-effort independent physical
* source. Boards without battery ADC return 0; jitter remains
* the primary entropy source either way. */
uint8_t adc_noise[32] = {0};
for (size_t i = 0; i < sizeof(adc_noise); i++) {
adc_noise[i] = (uint8_t)zephyr_board.getBattMilliVolts();
k_msleep(1);
}
uint8_t seed[32];
mesh::ZephyrRNG::mixIdentitySeed(seed, sizeof(seed),
adc_noise, sizeof(adc_noise));
{
mesh::SeededRNG seed_rng(seed, sizeof(seed));
self_identity = mesh::LocalIdentity(&seed_rng);
}
/* Ensure pub_key[0] is not reserved (0x00 or 0xFF in MeshCore protocol).
* With a properly mixed seed this almost never triggers; the
* cap+reboot is a safety net against pathological entropy failure. */
int attempt = 0;
while (self_identity.pub_key[0] == 0x00 || self_identity.pub_key[0] == 0xFF) {
if (++attempt > 100) {
LOG_ERR("Identity gen stuck on reserved prefix; rebooting");
k_msleep(2000);
sys_reboot(SYS_REBOOT_COLD);
}
mesh::ZephyrRNG::mixIdentitySeed(seed, sizeof(seed));
mesh::SeededRNG retry_rng(seed, sizeof(seed));
self_identity = mesh::LocalIdentity(&retry_rng);
}
mesh::ZephyrRNG::generateFirstBootIdentity(self_identity);
data_store.saveMainIdentity(self_identity);
mesh::Utils::secureZeroize(seed, sizeof(seed));
mesh::Utils::secureZeroize(adc_noise, sizeof(adc_noise));
}
companion_mesh.self_id = self_identity;
+4 -44
View File
@@ -454,55 +454,15 @@ int main(void)
lora_radio.setTxDoneCallback(lora_tx_done_callback, nullptr);
repeater_mesh.setTxQueuedCallback(tx_queued_callback, nullptr);
/* Load or generate identity BEFORE begin().
*
* First-boot keygen uses ZephyrRNG::mixIdentitySeed() a layered
* entropy mixer combining sys_csrand_get + HWINFO unique ID + ADC
* LSB noise + CPU cycle-counter jitter, conditioned via SHA-512.
* This compensates for ESP32's hardware TRNG being only seeded
* after WiFi/BT radio init (which on a repeater is on-demand for
* WiFi OTA there's no guaranteed radio activity at boot). */
/* Load or generate identity BEFORE begin(). First-boot keygen runs
* the layered entropy mixer + Ed25519 derive + reserved-prefix
* guard inside ZephyrRNG::generateFirstBootIdentity. */
mesh::LocalIdentity self_identity;
if (!data_store.loadIdentity(self_identity)) {
LOG_INF("No identity found, generating new keypair...");
/* Sample ADC LSB noise — best-effort independent physical
* source. Boards without battery ADC return 0; jitter remains
* the primary entropy source either way. */
uint8_t adc_noise[32] = {0};
for (size_t i = 0; i < sizeof(adc_noise); i++) {
adc_noise[i] = (uint8_t)zephyr_board.getBattMilliVolts();
k_msleep(1);
}
uint8_t seed[32];
mesh::ZephyrRNG::mixIdentitySeed(seed, sizeof(seed),
adc_noise, sizeof(adc_noise));
{
mesh::SeededRNG seed_rng(seed, sizeof(seed));
self_identity = mesh::LocalIdentity(&seed_rng);
}
/* Ensure pub_key[0] is not reserved (0x00 or 0xFF in MeshCore protocol).
* With a properly mixed seed this almost never triggers; the
* cap+reboot is a safety net against pathological entropy failure. */
int attempt = 0;
while (self_identity.pub_key[0] == 0x00 || self_identity.pub_key[0] == 0xFF) {
if (++attempt > 100) {
LOG_ERR("Identity gen stuck on reserved prefix; rebooting");
k_msleep(2000);
sys_reboot(SYS_REBOOT_COLD);
}
mesh::ZephyrRNG::mixIdentitySeed(seed, sizeof(seed));
mesh::SeededRNG retry_rng(seed, sizeof(seed));
self_identity = mesh::LocalIdentity(&retry_rng);
}
mesh::ZephyrRNG::generateFirstBootIdentity(self_identity);
data_store.saveIdentity(self_identity);
LOG_INF("New identity saved");
mesh::Utils::secureZeroize(seed, sizeof(seed));
mesh::Utils::secureZeroize(adc_noise, sizeof(adc_noise));
}
repeater_mesh.self_id = self_identity;