mirror of
https://github.com/meshtastic/firmware.git
synced 2026-09-16 08:30:04 -04:00
https-heap-headroom
15
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
 Thomas Göttgens
|
a89f1920e1 |
fix(traffic): don't re-stamp dropped duplicate positions, which slid the dedup window indefinitely (#11620)
* fix(traffic): don't re-stamp dropped duplicate positions, which slid the dedup window indefinitely * test(traffic): trim the regression test comment and derive its counts |
||
|
Ben MeadorsandClaude Opus 5
|
98c88d7e19 |
fix(position): halve the stationary/fixed-position broadcast floor to 6h (#11606)
The 12h floor introduced with traffic management was too aggressive: a fixed_position or stationary node goes quiet for half a day after its boot broadcast, so anything that missed that one packet - a node that joined later, or one that restarted - shows it with no position until the next refresh. Drop the floor to 6h, and drop the traffic-management identical-position dedup window from 11h to 5h with it. The two are a pair: the dedup window was deliberately sized just under the broadcast floor so a stationary node's periodic refresh clears its neighbours' window instead of being dropped as a duplicate. Leaving it at 11h would have made the extra broadcast pure airtime - aired, then discarded by every receiver - so the mesh would still have seen a 12h refresh. Role caps are unchanged and still bind: tracker 1h, lost-and-found 15m. Both remain shorter than the new 5h default, so those exceptions apply exactly as before. Co-authored-by: Claude Opus 5 <noreply@anthropic.com> |
||
|
Tom
|
bca7c0b480 |
Tom fiddles with the test suite - again (#11517)
* test: make every suite run its own binary, and fail the run when it does not PlatformIO links every native test program to the one $BUILD_DIR/$PROGNAME path and attributes Unity output by text alone, never checking that the source file a case came from belongs to the suite it thinks it ran. Both harnesses had been split into a build pass (--without-testing) and a run pass (--without-building), and for a non-embedded platform the run pass never relinks - so all 57 suites executed whichever suite was linked last, each reporting PASSED under its own name. Introduced for CI in |
||
|
Tom
|
f5314148c2 |
Serialise AirTime behind a lock, and stop handing out its buckets (#11362)
* Copy airtime reports into a caller buffer instead of exposing the array
airtimeReport() returned a pointer into the rotating bucket arrays, so the
caller held a handle to state that logAirtime() and every accessor mutate
underneath it. Copy into a caller-supplied buffer instead, and report failure
for a null buffer, a count past the log depth, or an unknown report type.
ContentHandler owns its buffer and hoists getPeriodsToLog() out of the three
calls that repeated it.
* Cover the AirTime report API and log-dispatch contract
Half of AirTime's surface had no tests: which store each report type feeds,
what airtimeReport() does when misused, how the first sync seeds itself, and
whether calling several entry points in one interval compounds the rotation.
Eighteen tests, asserted through the public API rather than the public bucket
arrays - those arrays are meant to become private, and a test that reads them
would have to be rewritten rather than pinning a contract.
Two of them state a convention that was never written down: the report arrays
are shift-ordered with slot 0 newest, and slot 0 covers only the time since the
last rotation. channelUtilization and utilizationTX use the opposite convention
- a modular ring indexed by uptime phase - and reading one as if it were the
other is a defect that has already happened once.
* Characterise AirTime window decay, TX gates, and sleep behaviour
Thirty-three tests in three kinds. Invariants must hold forever; boundaries pin
off-by-ones a refactor would move; five characterisations encode today's wrong
numbers, each tagged with the phase that will flip it.
Readings are asserted against an event-log oracle - airtime physically on air
inside (now - window, now], computed from a list of completed packets - rather
than against hand-worked constants, so a test states "this matches the
definition" instead of "this looked right when I wrote it".
The characterisations, all measured rather than assumed:
- the window covers (N-1)p + phase but divides by Np, so a steady 10% load
reads 8.33% right after a bucket boundary -> phase 5
- the same load sweeps across bucket phase instead of holding -> phase 5
- the hour window carries the same defect, 10x smaller -> phase 5
- a packet longer than its bucket is credited whole to the bucket
it completed in, so a saturated LONG_SLOW channel reads >100% -> phase 4b
- getSilentMinutes() reads a modular ring as if the index were an
age, so identical airtime gives different answers by phase -> phase 6
Two tests needed correcting during the write, both my expectations rather than
the code: a six-bucket ring sheds whole buckets, so a 30s gap drops three of
five survivors and not "half"; and the oracle sees 59 completions in a 60s
window, not 60, because the one on the lower edge is outside it.
Not written: the planned RX_LOG/RX_ALL_LOG disjointness test. That is a
property of the two radio drivers, which choose one or the other per packet -
it is not observable from AirTime, which records what it is told. The
AirTime-side half is already covered by the routing tests.
* Drop write-only and undefined AirTime members
None of this was reachable:
air_period_tx / air_period_rx file-scope mirrors of airtimes.periodTX/RX,
accumulated, rotated and memset in lockstep
with them but never read out or serialised.
Orphaned when #2552 re-pointed the writes at
bare globals instead of deleting them.
lastUtilPeriod, lastUtilPeriodTX written on every sync, read nowhere
airtimes.lastPeriodIndex written on every rotation, read nowhere
currentPeriodIndex() computes (secs / 3600) % 8 - a modular-ring
index for the one array that is shift-ordered
rather than a ring. Its only two uses were the
dead field above and a log line. It is the
fossil of the same confusion that makes
getSilentMinutes() wrong.
UtilizationPercentTX() declared, never defined
free logAirtime()/airtimeReport() declared, never defined; the latter still
carried the array-returning signature the
previous commit removed, so it actively misled
Also fixes the rotation log line, which read currentPeriodIndex() from inside
the loop although the index is advanced before it - on a multi-hour wake it
printed the same final value once per rotation. It now reports which of the
crossed hours is being rotated.
airtimeRotatePeriod() is kept: it has no caller in the tree either, but unlike
the above it is a defined public method, so out-of-tree callers are plausible.
Measured, not estimated: sizeof(AirTime) 464 -> 456 B, plus 64 B of globals, so
-72 B of static RAM. Padding accounts for the difference from the 66 B the plan
predicted by counting declared bytes.
The whole point of writing the tests first: the suite is green here with zero
test changes.
* Document what the AirTime figures measure and how they are stored
Comments only, but four of the things they replace were false.
The header's example analytics claimed RX_ALL_LOG was "all received lora
packets" and offered "RX_ALL_LOG - RX_LOG = other lora radios". Both radio
drivers pick exactly one of the two per packet, so they are disjoint: RX_ALL_LOG
is airtime we could not parse, the subtraction can go negative, and the total is
TX + RX + RX_ALL. Replaced with the actual contract - four inputs, eight
outputs, the window each spans, and the fact that the three thresholds are
hard-coded members rather than the settings they look like.
Names the two storage conventions on their declarations, because mixing them up
is what makes getSilentMinutes() wrong: channelUtilization and utilizationTX are
modular rings indexed by uptime phase, where the oldest bucket is (current + 1)
% N; airtimes.period* is shift-ordered with slot 0 newest, where the index IS an
age and slot 0 is a partial hour.
Defines the measurement as wall time rather than awake time, and says why: a
sleeping node still hears traffic, and per-node redefinition would make two
broadcast readings incomparable. Records that the 60s figure is published to the
mesh at >= 1h cadence, so what other nodes see is a snapshot - at LONG_FAST and
1% occupancy it reads exactly 0 in about 44% of reports - and that the contention
window it feeds moves in 20-percentage-point steps, so small errors never reach
the backoff.
Finally, states that rotation happens on access rather than on the scheduler
tick, names the test that enforces it, and leaves a TODO pointing at the plan
phases that fix the characterised accuracy defects.
* Serialise AirTime behind a lock proven by a private token
Two mechanisms solving different halves. A lock-free inner core (Windows) holds
all state and all logic; it has no lock and no way to reach one, so nesting is
impossible by construction. A private Held token takes the lock in its own
constructor and is the only thing that can be passed where a core method demands
one, so the lock cannot be forgotten either.
The rule is now uniform with no exceptions to remember: every public method
takes the lock once and delegates. In particular isTxAllowed*() lock like
everything else - before the split they could not, because they called the
public accessors and the lock is not recursive. That asymmetry was the foot-gun
the previous design documented in prose and hoped nobody would trip.
getPeriodsToLog()/getSecondsPerPeriod() still take no lock; they return
compile-time constants and touch no state.
channelUtilization[] and utilizationTX[] were public, so the lock was bypassable
at compile time. They move into the private core. Four test sites reached in;
all four now use logAirtime() plus the virtual clock, and no new test seam was
needed. Nothing in src/ was affected.
The re-entry assert is guarded on PIO_UNIT_TESTING, so it exists in test builds
only. The design sketched #ifdef DEBUG, but nothing in this tree defines DEBUG
or NDEBUG, so either spelling ships the assert to every board - and
nrf52_promicro_diy_tcxo has ~128 bytes of headroom under its 0xEA000 warm-store
cap, which the assert's strings and abort path overrun. It would have worked on
hardware, since the check runs in Held's owner initialiser and so precedes the
blocking take; the objection is that abort()ing a live mesh node is a poor trade
for a bug never seen in the field. Native tests are where it earns its keep
anyway: Portduino compiles Lock::lock() to an empty body, so a nested take there
succeeds silently and nothing else would notice.
Also comments out ScopedBusyAirTime in test_traffic_management. It is inert
twice over: the module holds no reference to airTime at all since hop exhaustion
was shelved, and the fixture never worked anyway - writing the buckets on a
fresh AirTime is undone by the first accessor call, which takes the firstTime
branch and memsets them. It reported 0%, not the 100% it claimed. Left in place,
commented, with both reasons recorded.
Cost on the tightest board in the tree, nrf52_promicro_diy_tcxo: the six phases
together add 96 bytes of flash, leaving it 32 bytes clear of the warm-store
guard. RAM is 72 bytes lower from the dead-state removal. Suite green at 47/47,
with test_airtime unedited apart from the added nesting test.
* Count rotations with the loop variable, not a separate tally
LOG_DEBUG compiles to nothing under DEBUG_MUTE, so the counter's only read
disappeared with it and the tally became write-only. It does not warn today -
this build has -Wunused-but-set-variable on, and it fires for other locals, but
not for one that is only initialised and never read - so it was latent rather
than broken: a stricter flag or -Werror would have failed muted builds only.
Using the loop variable removes the class of problem, since the loop condition
reads it, and drops the elapsedAirtimePeriods-- mutation as a side benefit.
Same iteration count, same output.
Found by compiling nrf52_promicro_diy_tcxo with -D DEBUG_MUTE, which is worth
recording for its own sake: muting logs takes that image from 802 784 to
673 416 bytes, 98.5% to 82.6% of flash. Logging is 16% of the largest nrf52
image, and its 32 bytes of warm-store headroom are a logging-verbosity question
rather than a code-size one.
* Tighten the comments added by this branch
Comment-only: with comments stripped, all five files are byte-identical to the
previous commit.
Removed the references to the planning notes. Those documents are working
material and will go stale; the code should not depend on them. The five
CHARACTERISATION tags now describe the defect they pin and stop there, and the
accuracy TODO names the four defects and points at the tests instead of a plan
file.
Also removed, as noise rather than information:
- comparisons against pre-#11291 behaviour, which nobody reading this needs
- a comment describing the lock restructure as future work, written before it
landed
- speculation ("plausible", "worth pinning so a future...")
- an aside arguing with an arithmetic slip made while writing the test
Kept the mechanical facts that are slow to re-derive: the two storage orderings
and which array uses which, RX_LOG/RX_ALL_LOG disjointness, the locking rule and
the addSpanned() constraint that protects it, why the re-entry assert is
test-only, and the concrete numbers - (N-1)p + phase, 14 164 ms, the 20 pp
contention-window steps.
Net 16 comment lines out of src/, 33 out of test/.
* Gate the AirTime re-entry check on the host, not on testing
PIO_UNIT_TESTING is injected by PlatformIO purely on BUILD_TYPE, with no
platform check, so it is defined on an on-target `pio test` run too. The
check arms before the lock is taken - a nested take blocks forever, so a
later check would never run - which under preemption false-positives on
legitimate contention and races on its own write.
Derive AIRTIME_REENTRY_CHECK once from PIO_UNIT_TESTING && !HAS_FREE_RTOS
and use it at all three sites. Had the three conditions ever diverged, an
on-target test build would fail to compile on a member the header no
longer declares.
* Log AirTime outside the lock it serialises
DEBUG_PORT.log() blocks on a UART write, and `lock` is a plain binary
semaphore with no priority inheritance, so holding it across a log call
lets the main thread stall the radio thread in getTxDelayMsec().
Move logAirtime()'s LOG_DEBUG into the shell, after the Held scope
closes; the shell already has both arguments, so nothing has to be
passed back out of the core. isTxAllowed{ChannelUtil,AirUtil} read into
a local under the lock and warn after it. The log bodies are braced
because LOG_DEBUG compiles away under DEBUG_MUTE and a bare `if (x) ;`
trips -Wempty-body.
Fold the two doubled index calls into `+=` while touching the lines.
* Give each airtime report its own buffer
handleReport() reused one array across the three airtimeReport() calls
and ignored the bool. A failed report would have left the previous
type's data in place and emitted it under the next type's key. Build
each through a lambda whose buffer is zeroed per call, so a failure
emits zeros.
Unreachable today - the count is always PERIODS_TO_LOG and the type is
always valid - but the old shape only read as correct by accident.
* Drop a stray semicolon from the inert-guard comment
* Address external review: name the race, tighten the claims and the tests
The header sold the lock as mechanism without naming a second thread, which
invites the reasonable objection that this is a cooperative OSThread codebase.
There is a real race and it is nRF52-only: NRF52Bluetooth registers its ToRadio
write callback with defer == false, so a phone's packet runs handleToRadio ->
sendToMesh -> Router::send on the Bluefruit BLE task, reading
utilizationTXPercent() and getSilentMinutes() while loopTask may be inside
logAirtime(). ESP32 hands BLE work to the main task and does not have it.
Three claims in the header were wrong or overstated:
- "nesting is impossible by construction" - Windows is a nested class with an
enclosing class's access rights, and `extern AirTime *airTime` is in the
same header, so airTime->anyPublicMethod() from inside it is well-formed
and would hang. Nothing does it; the assert is the backstop. Say that
instead, because the comment below instructs contributors to add helpers
to Windows on the strength of the guarantee.
- "every public method takes the lock exactly once" - two constant accessors
take none and isTxAllowedAirUtil() takes it zero or one times. State the
exceptions where the invariant is stated, not only at the definitions.
- "both radio drivers pick exactly one per packet" - five drop paths log
neither. At most one. Recorded against plan4 rather than fixed here: it
changes a telemetry value.
getPeriodsToLog()/getSecondsPerPeriod() become static constexpr, which removes
them from the locking claim structurally and lets ContentHandler size its
buffer and its count from one constant.
Tests:
- C14's saturated AirTime is installed by a helper and restored in tearDown.
Unity's TEST_ABORT() is longjmp and does not run destructors of automatic
objects, so the scoped guard it replaces would leave airTime dangling into
an abandoned frame on any assertion failure - and the same commit that
added it removed the tearDown reset that did cover that.
- test_getSilentMinutes_counts_minutes_until_enough_ages_out asserted only
`mins <= 60`, which neither return path can violate. The answer is 59.
- test_backwards_uptime_degrades_safely stepped 600s -> 60s, which leaves
elapsedAirtimePeriods at 0, so it never reached the hourly-report branch
its own comment describes. Step by the wrap instead and assert the exact
figures.
- test_airtime leaked EU_868 out of the duty-cycle case into every later one,
and the reentry test's isTxAllowedAirUtil() coverage depended on it.
Restore the region in tearDown and set it explicitly where it is wanted.
- Rename that test to what it can actually check: no single method takes the
lock twice. The calls are sequential, so it cannot catch two methods
nesting.
* trunk: suppress trufflehog/Lob false positives in test_airtime
* Address CodeRabbit review: the rotate trace, the cap warn, the backoff
Four findings from the CodeRabbit pass. Two were introduced by this branch,
one is a real inconsistency it inherited, one is a naming slip.
The rotate trace was the one that mattered. "Log AirTime outside the lock it
serialises" moved the per-packet lines and the two TX-gate warnings out to the
shell, but missed LOG_DEBUG("Rotate airtimes, crossed hour %u") because it does
not sit in the shell at all: it is inside Windows::syncNow(), the lock-free
core, which by construction only ever runs under Held. Nothing at that line
looks like a lock, which is why it survived.
The exposure is smaller than the review suggests - runOnce() syncs at 1 Hz, so
in steady state this is one line an hour, and the PERIODS_TO_LOG - 1 burst
needs an hour of light sleep with no intervening sync - but a UART write under
a plain binary semaphore with no priority inheritance is exactly what the
comment above logAirtime() says this code does not do. syncNow() now
accumulates crossings in rotationsPendingLog and runOnce() drains it inside the
Held scope, then logs after release. Any caller can cross an hour; only that
thread reports it, so a crossing raised elsewhere is traced at most one tick
late. The `if (rotations > 0)` guard keeps the drained value read under
DEBUG_MUTE, where LOG_DEBUG expands to nothing - the write-only tally that
"Count rotations with the loop variable" removed.
addFromContact()'s favorite fallback stamped silently when the protected cap
refused it. The stamp is new on this branch; the two sibling refusals (ignore,
verify) both emit PROTECTED_CAP_WARN_FMT, so the operator lost the only signal
that the cap was hit on the one path that has a fallback.
lfs_assert() mixed clocks: Throttle read Time::getMillis(), the remainder was
computed from a second, bare millis(). The review's stated failure mode - a
native test overriding the clock - cannot happen, since the hook is behind
PIO_UNIT_TESTING and this file is nRF52-only. The real defect is the second
read: a tick landing on the 20-minute boundary between the check and the
subtraction underflows the remainder into delay(~50 days), on a device that has
just found its flash corrupt. One read, clamped, and preFSBegin() stores from
the same clock.
The eviction test is renamed to
test_eviction_prefersCurrentBootStampOverPost2038Epoch. The finding is right
that it was snake_case, but the suggested testEvictionPrefers... does not match
this file either, which is test_<area>_<camelCase> throughout.
Not taken, both pre-existing and out of scope for a rollover branch:
- t5s3_epaper's touchResumeAtMs/suppressFromMs read an active suppression as
inactive if the wake lands in the 1 ms where millis() is 0. Consequence is
one skipped 150 ms touch-settle window per 49.7-day wrap.
- NRF52Bluetooth::onPairingPasskey() busy-waits 30 s in a BLE callback. Worth
saying plainly that this branch makes it more visible: the old
`millis() < start_time + 30000` overflowed at the wrap and cut the wait
short, so the correct Throttle form is what lets it run the full 30 s.
Reworking it into an OSThread is its own change.
Native suite GREEN, 48/48, 672 cases.
|
||
|
TomandClaude Opus 4.8
|
e26b6bfc5a |
TMM & Warmstore key source naming (#11119)
* one becomes two * warmstore clarify * Address PR review: key-provenance terminology consistency - Log line now says "not key-proven" (gate is XEdDSA OR manual, not just signer) - Rename markKeySignerProvenForTest -> markKeyXeddsaSignedForTest (sets only the XEdDSA bit) - Docs + test comments: "signer bit" -> "XEdDSA-signed bit" clod helped too * Rename signer-proven -> key-proven for broadened provenance predicate Address PR #11119 review: the copyPublicKey()/copyUser() out-parameter and the cache-path replay gate now report entry->keyProven() (XEdDSA-signed OR manually verified), so the "signerProven" name and "signer-proven" comments were misleading. Rename the public out-param to keyProven, the local cachedKeySignerProven to cachedKeyProven, and update coupled callers, log strings, docs headings, and comments to say "key-proven". Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * nitpicks --------- Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com> |
||
|
Tom
|
2024bb8384 |
Arrival time fix perhaps (#11274)
* Add explicit presence for MeshPacket.rx_time (arrival time) rx_time is now proto3 optional with a has_rx_time presence bit, matching the rx_rssi treatment. A node with no GPS and no phone connected yet has no time source at all, so a bare 0 was indistinguishable from a genuine 1970-01-01 reading; downstream consumers (replay packets, JSON serialization) now check has_rx_time instead of the value. * Dedupe rx_time stamping into a shared helper; trim a debug log string Extract the repeated haveTime/rx_time/has_rx_time stamp logic (5 call sites across Router.cpp, MeshBeaconModule.cpp, MeshService.cpp) into Router::computeRxTimeStamp()/stampRxTime(). Also shorten the new RTC.cpp LOG_DEBUG string. Saves 48 bytes of flash on rak4631 (measured), no behavior change. * Fix has_rx_rssi presence carried unconditionally through StoreForward replay preparePayload() set has_rx_rssi = true unconditionally on replay, regardless of whether the packet's rx_rssi at store time was a genuine measurement (e.g. MQTT-relayed packets carry no real RSSI). Store the presence bit alongside rx_rssi in PacketHistoryStruct and restore it on replay instead. Flagged by Copilot on #11271 (same root cause the has_rx_time explicit presence work fixes) but never addressed before that PR merged. * Trim comment blocks to the repo's 1-2 line guideline .github/copilot-instructions.md:338 caps code comments at 1-2 lines; several blocks added across the rx_time explicit-presence work ran well past that. Also consolidates Time.cpp's file-level doc comment into Time.h, where the rest of the Time:: API contract already lives. No behavior change. * Add rx_time explicit-presence test coverage - test_meshpacket_serializer: has_rx_time=false fixture plus tests asserting JsonSerialize/JsonSerializeEncrypted emit 0 rather than leaking the millis() placeholder, alongside the has_rx_time=true baseline. - test_stream_api: two tests driving a real PhoneAPI handshake (want_config_id through STATE_SEND_PACKETS) that simulate a phone time-giving transaction arriving before vs. after a queued packet is drained - covering both the reconciled and the ships-with-placeholder-absent paths of MeshService::reconcilePendingRxTimes(). * Fix three correctness issues flagged in review - Time.h: drop the reserved-identifier include guard (_MT_TIME_H); pragma once already covers it, matching convention elsewhere (e.g. RTC.h). - Time.cpp: rebase getMillis64()'s wrap accumulator when the test seam swaps clock sources, so a real<->injected clock jump isn't miscounted as a genuine 32-bit wrap. - NodeInfoModule: the 12h reply-suppression window is a local dedup duration, not a wall-clock reading - switch it to Time::getMillis64() so RTC-quality jumps and replayed packets' stale rx_time can't perturb it. - StoreForwardModule: has_rx_time was derived from *current* RTC quality at replay time rather than stored at capture time, so a history entry saved while time-blind could be misreported as a valid epoch once the clock later improved. Persist the presence bit in PacketHistoryStruct instead. * tryfix CI * post review fixes * more test fixes |
||
|
|
1872e5b306 |
TMM extend to ephemeral 3rd tier store (#11050)
* Harden TMM NodeInfo direct-response: staleness, key hygiene, throttle The NodeInfo direct-response path answered queries on behalf of other nodes from an unauthenticated, never-expiring cache while suppressing the genuine request (STOP). That let a long-gone or forged identity be served as a fresh-looking, authoritative reply indefinitely. Three fixes: - Staleness gate: refuse to spoof a reply for a node not actually heard within ~6h (PSRAM cache via lastObservedMs; NodeDB fallback via sinceLastSeen, which tolerates last_heard==0 when the clock is unset). A stale entry now falls through so the real request propagates instead of being answered for a dead node. - Key hygiene: mirror NodeDB::updateUser() PKI checks when caching overheard NodeInfo - reject a packet advertising our own public key, and pin keys (drop a NodeInfo whose key mismatches an already-known key from NodeDB or from our own cache). Prevents cache poisoning / key substitution via the spoofed-reply path. - Response throttle: at most one spoofed reply per target node per 30s. Direct responses bypass the per-sender rate limiter (they STOP the request first) and the reply target is attacker-controlled, so this bounds airtime exhaustion and reflected floods. Recorded in a new NodeInfoPayloadEntry.lastResponseMs (PSRAM; self-expiring timestamp compare, no sweep needed). Tests: native NodeDB-fallback staleness (stale drops, fresh serves) plus PSRAM-guarded staleness, throttle, and key-mismatch cases. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01G7QggdwMM1CJkguuCmMfZs * Add TODO(T1-T9) markers for code-review findings on NodeInfo hardening Comment-only. Tags each site flagged in the review of the direct-response throttle/staleness/key-hygiene change so the follow-ups are visible in-context: T1 throttle black-holes distinct requestors; T2 per-target key doesn't bound aggregate TX; T3 non-PSRAM fallback unthrottled; T4 millis-wrap defeats the staleness gate; T5 redundant second O(n) scan for the throttle stamp; T6 duplicated 32-byte key compares; T7 sinceLastSeen() called twice; T8 two 6h constants can desync; T9 clockMs()==0 collides with the 0 sentinel. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01G7QggdwMM1CJkguuCmMfZs * Address code-review cleanups T6, T7, T8 in NodeInfo direct-response T8: derive kNodeInfoMaxServeAgeSecs from kNodeInfoMaxServeAgeMs so the PSRAM and NodeDB-fallback staleness windows cannot desync. T6: collapse the three duplicated 32-byte key compares in cacheNodeInfoPacket (owner-impersonation + NodeDB pin + cache TOFU pin) into one pubKeysEqual() helper, so the compare lives in one place. T7: compute sinceLastSeen(node) once on the fallback staleness path so the tested age and the logged age cannot diverge (it calls getTime() each time). Remaining review items still marked in-code: T1/T2/T3 (throttle design), T4/T9 (millis-wrap + 0-sentinel), T5 (redundant throttle-stamp scan). Native test_traffic_management suite: 48/48 passing. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01G7QggdwMM1CJkguuCmMfZs * Throttle the NodeDB fallback direct-response path (T3) The per-target NodeInfo response throttle lived only in the PSRAM NodeInfoPayloadEntry, so non-PSRAM boards - which answer from the NodeDB fallback path - emitted spoofed direct replies with no rate limit at all, leaving the airtime/reflection surface the throttle exists to close. Add a single module-global fallback stamp (nodeInfoFallbackLastResponseMs, 4 bytes, guarded by cacheLock) that throttles the fallback path to one spoofed reply per window across all targets. Coarser than the PSRAM per-target throttle, but the fallback path has no per-node slot to stamp, and a global cap still bounds spoofed transmissions - which is the point. Add a native test exercising the fallback throttle window. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01G7QggdwMM1CJkguuCmMfZs * Make NodeInfo staleness/throttle wrap- and sentinel-safe (T4, T9) T9: the millis-based fields that use 0 as a "never" sentinel (lastObservedMs, lastResponseMs, nodeInfoFallbackLastResponseMs) could be stamped with a literal 0 during the one-millisecond clockMs()==0 instant at each ~49.7-day wrap, momentarily colliding with the sentinel and disabling the staleness/throttle gate for that entry. Route all such stamps through a new nowStampMs() that maps 0 -> 1; the 1 ms skew is irrelevant to every window these fields feed. T4: the staleness gate's modular age compare is only unambiguous while true age < 2^32 ms (~49.7 days); an entry that lingered unrefreshed that long could wrap to a small age and read as fresh, defeating the gate. Add a NodeInfo eviction pass to the maintenance sweep that drops entries past the serve window, so an entry is removed long before its age can approach the wrap boundary. This also frees slots holding stale identities. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01G7QggdwMM1CJkguuCmMfZs * Stamp PSRAM throttle entry by captured index, not a rescan (T5) The post-send throttle stamp did a second full O(n) scan of the PSRAM NodeInfo array under a fresh lock, even though findNodeInfoEntry() had already located the slot at the top of shouldRespondToNodeInfo(). Capture the slot index during that initial lookup and address the entry directly on the stamp path. Because the cache lock is released between the two accesses, the slot could have been evicted or reused, so re-validate node == p->to under the lock before stamping. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01G7QggdwMM1CJkguuCmMfZs * Document accepted per-target throttle tradeoffs (T1, T2) Convert the T1/T2 TODO markers into a NOTE: the per-target PSRAM throttle keying and its lack of an aggregate bound are deliberate design choices, not pending work. A distinct requestor being throttled for one window is harmless on a redundant mesh, and keeping the PSRAM path per-target preserves throughput for legitimate multi-target responders (the fallback path already bounds aggregate via its global stamp). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01G7QggdwMM1CJkguuCmMfZs * Track signed-provenance of cached NodeInfo public keys Add keySignerProven to NodeInfoPayloadEntry, set when an observed NODEINFO frame's XEdDSA signature was verified (mp.xeddsa_signed). This distinguishes a trust-on-first-use key from one proven to belong to a signer. The flag is monotonic - once proven it stays proven, and the existing key-pin checks forbid the underlying key from changing - so a later unsigned frame cannot downgrade it. A signature can only be verified against a key we already held, so a first-contact key is always TOFU until a later signed frame upgrades it. Groundwork for using the TMM cache as a last-resort public-key source, where this flag serves as a trust tier. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01G7QggdwMM1CJkguuCmMfZs * Draw public keys from the TMM NodeInfo cache as a last resort Add TrafficManagementModule::copyPublicKey() and consult it from NodeDB::copyPublicKey() after the hot (NodeInfoLite) and warm (WarmNodeStore) tiers miss. This extends the pool of peers the node can encrypt to: a key the NodeInfo direct-response cache overheard for a node that has since aged out of both NodeDB tiers can still be used. The getter reports whether the key is signer-proven or trust-on-first-use. NodeDB serves TOFU keys here too - the same first-contact trust NodeDB already applies in updateUser() - so the pool actually expands to new long-tail nodes rather than only re-confirming known keys. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01G7QggdwMM1CJkguuCmMfZs * Retain keyed NodeInfo entries as a pubkey pool; trust-tier eviction Now that NodeDB::copyPublicKey() draws keys from the NodeInfo cache, the flat 6 h serve-window eviction would discard useful keys the moment a node stops being servable. Split retention: an entry carrying a 32-byte public key is kept for kNodeInfoKeyRetentionMs (7 d), while a keyless entry still expires at the serve window. Both windows stay well under the ~49.7-day millis wrap, preserving the T4 wrap-safety guarantee. Make LRU victim selection trust-tiered to match: a keyless slot is sacrificed before any keyed one, and a trust-on-first-use key before a signer-proven key; within a tier the oldest loses. Mirrors WarmNodeStore's keyed-first admission so the most valuable keys are the stickiest under memory pressure. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01G7QggdwMM1CJkguuCmMfZs * Test signed-provenance flag and copyPublicKey key-pool source PSRAM-gated tests (run on the ESP32 build, alongside the existing NodeInfo PSRAM tests): - copyPublicKey serves a TOFU key learned from an unsigned NodeInfo and reports signerProven=false - a later signature-verified NodeInfo upgrades provenance to signer-proven while the pinned key bytes stay unchanged - copyPublicKey reports a miss for an uncached node Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01G7QggdwMM1CJkguuCmMfZs * Inherit signer provenance from NodeDB when caching a re-found node When the TrafficManagement NodeInfo cache re-caches a node, mark its key signer-proven if NodeDB already knows the node as a verified signer for that same key - even if this particular (unicast/unsigned) frame carried no signature. Previously the flag only upgraded on a frame we verified ourselves, so a node we had already proven elsewhere looked TOFU here. Add NodeDB::isVerifiedSignerForKey(), which checks both tiers - the hot store's signed bitfield and the warm tier's cached signer bit - and requires the key to match so a rotated/mismatched key never inherits a stale verdict. Add WarmNodeStore::isVerifiedSigner() to expose the warm signer bit (the rebase onto develop added the bit itself; this surfaces it for lookups). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01G7QggdwMM1CJkguuCmMfZs * Lock NodeInfoPayloadEntry packing with a size assert keySignerProven cost zero bytes: sourceChannel, the two bools, and decodedBitfield are four 1-byte fields that fill a single 4-byte tail word (struct alignment is 4), so the flag consumed former padding rather than growing the 2000-entry PSRAM array. Add a static_assert pinning the entry to sizeof(meshtastic_User) + 20 so a future 5th trailing byte - which would open a fresh word (~8 KB PSRAM across the array) - fails the build instead of silently costing memory, prompting new flags to be packed into existing bytes. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01G7QggdwMM1CJkguuCmMfZs * Pack NodeInfo cache booleans into 1-bit fields Fold hasDecodedBitfield and keySignerProven into adjacent uint8_t:1 bitfields so they share a single byte, reserving 6 spare bits for future flags without growing the 2000-entry PSRAM array. sizeof is unchanged (still one packed tail word, sizeof(meshtastic_User) + 20); access is by name exactly as before, so no call sites change. Reorders decodedBitfield ahead of the flags so the two 1-bit members stay adjacent and the compiler packs them together. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01G7QggdwMM1CJkguuCmMfZs * Gate NodeInfo replay on signer-proven provenance (compile-time, default on) Add TMM_NODEINFO_REPLAY_REQUIRE_SIGNED (default 1): the direct-response path now only spoofs a reply for a node whose key is signer-proven, in addition to the staleness gate - so replay is based on flagging AND staleness. Replay is a courtesy feature, and vouching for an unverified (trust-on-first-use) identity to other nodes is the risk this closes, so signed-only is the safer default. Define the macro to 0 at build time to also serve fresh TOFU-only nodes. Both paths are gated: the PSRAM path checks the cached keySignerProven flag, the NodeDB fallback checks nodeInfoLiteHasXeddsaSigned(node). The effective gate (TMM_NODEINFO_REPLAY_SIGNED_GATE) is bypassed when PKI is excluded from the build, since nothing can be signed there and the courtesy feature would otherwise be disabled outright. Tests: existing reply-expecting tests establish signer-proven state (a new markKeySignerProvenForTest hook for the PSRAM cache, the NodeDB signed bit for the fallback path); add fallback and PSRAM tests asserting an unsigned/TOFU node is withheld under the default gate. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01G7QggdwMM1CJkguuCmMfZs * Rehydrate re-admitted node names from the TMM NodeInfo cache The warm tier keeps an evicted node's key but not its name (the 40-byte record has no room), so a re-admitted long-tail node is nameless until its next NodeInfo broadcast. The TrafficManagement NodeInfo cache is much larger (2000 PSRAM entries) and commonly still holds the full User, so use it as a name reservoir the warm tier structurally cannot be. On re-admission (getOrCreateMeshNode), after the warm-tier restore, copy the cached identity from the TMM cache via a new copyUser() getter - but only when its cached key matches the key just restored from warm, so a name never attaches to a different identity than the one we encrypt to (key-matched trust). Guarded by HAS_TRAFFIC_MANAGEMENT and a null check; a no-op without the PSRAM cache or when no key is present. CopyUserToNodeInfoLite sets only user-related bits, so the warm-restored signer bit is preserved. Limitations (by design): PSRAM-only, and the TMM cache is RAM-only, so this helps within an uptime session, not across reboot. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01G7QggdwMM1CJkguuCmMfZs * Drop non-portable NodeInfoPayloadEntry size assert (fixes pico build) The static_assert pinned sizeof(NodeInfoPayloadEntry) to sizeof(meshtastic_User) + 20, but nanopb packs the generated User struct differently per platform: on pico its size is not a multiple of 4, so alignment padding before the uint32 timestamps makes the overhead 22, not 20, and the assert failed the build (native happened to be +20 and passed, hiding it). The bitfield packing it was guarding still stands; replace the fixed-count assert with a comment, since no portable byte count exists. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01G7QggdwMM1CJkguuCmMfZs * Run the NodeInfo cache paths in native tests (TMM_HAS_NODEINFO_CACHE) The NodeInfo direct-response cache and everything layered on it - key pinning, signer provenance, staleness, throttle - was compiled only for ARCH_ESP32 + BOARD_HAS_PSRAM, so its entire test set was skipped by the native CI suite and ran nowhere but real hardware. Fold the scattered platform guards into one TMM_HAS_NODEINFO_CACHE macro that also enables the cache (plain heap, delete[] path already existed) for ARCH_PORTDUINO unit-test builds. Production portduino and embedded test builds are unchanged. Tests that exercise the NodeDB fallback path drop the cache explicitly via a new dropNodeInfoCacheForTest() hook, so both response paths are covered in one binary. Native suite: 60/60 (was 50 with the 10 cache tests compiled out). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01KWHDDxY4U2XLYW2WJJ2oLT (cherry picked from commit b2dafbbbd6501b42fc27175e599fb38be292b0e3) * Pin NodeInfo cache keys against the warm tier, not just the hot store cacheNodeInfoPacket() mirrored NodeDB::updateUser()'s key pin but checked only getMeshNode() - the hot store. updateUser's own pin effectively covers the warm tier too (getOrCreateMeshNode rehydrates the warm key before the check), so the mirror had a gap: for a node evicted to the warm tier whose cache slot was also gone, an attacker's NodeInfo with a bogus key passed the pin, and the cache's TOFU pin then locked the genuine node's frames out until the poisoned entry aged away. Split the authoritative lookup out of NodeDB::copyPublicKey() as copyPublicKeyAuthoritative() (hot store, then warm tier - never the opportunistic TMM tier, which would compare the cache against itself) and pin against that. Test: warm-tier-only key rejects a mismatching NodeInfo and accepts the matching one. Native suite: 61/61. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01KWHDDxY4U2XLYW2WJJ2oLT (cherry picked from commit 28e880875ef1f8bf16bdf301a4e6036537a84b4b) * Convert NodeInfo cache clocks to uint8 tick counters Replace the entry's uint32 millis stamps (lastObservedMs, lastResponseMs) with three uint8 free-running tick clocks - the same modular scheme the UnifiedCache counters already use: obsTick 3 min/tick (12.8 h period) - replay staleness gate respTick 5 s/tick (21.3 min) - per-target response throttle retTick 1 h/tick (10.67 d) - retention TTL + LRU age Validity is an explicit flag bit (hasObserved / hasResponded), not a 0 sentinel, and the maintenance sweep saturates a stamp (clears its flag) once its window passes, so no stamp can age toward its ~256-tick aliasing horizon. That retires the wrap/sentinel special cases (T4, T9) - nowStampMs() survives only for the fallback path's module-global millis stamp. obsTick is separated from retention by design: only a genuinely heard NODEINFO frame stamps it, so later membership-based retention refresh can never make a silent node look servable. Also drops lastObservedRxTime, which was only echoed in a debug log. Entry shrinks 136 -> 128 B; the 2000-entry array 272 kB -> 256 kB, 8 kB below the original develop footprint while keeping the throttle and retention features. Tick granularity costs at most one tick per window (+-5 s on 30 s, +-3 min on 6 h, +-1 h on 7 d). Native suite: 61/61. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01KWHDDxY4U2XLYW2WJJ2oLT (cherry picked from commit e48ed9cfacfed0131b77ed2f1992e8720720d387) * Keep the NodeInfo cache a superset of NodeDB (seed + keep-alive) The cache learned identities only from overheard NODEINFO frames, so its membership was opportunistic: a NodeDB-tier node whose frame was never heard this boot had no entry (no key pin to protect it, no name to rehydrate), and the retention TTL evicted entries for nodes that still lived in the hot store or warm tier. Add an anti-entropy pass to the maintenance sweep (reconcileNodeInfoMembershipLocked): every hot-store / warm-tier node gets an entry - full identity from the hot store (hasFullUser), key-only from the warm tier - with signer provenance inherited key-matched, and NodeDB's key adopted wholesale on conflict. Member entries (isMember) are re-stamped each sweep, so they never age toward the retention TTL; when a node leaves both tiers the keep-alive stops and its entry ages out from its final member re-stamp. Membership also outranks key trust in LRU victim selection, and the reconcile pass skips seeding rather than churn one member out for another when hot+warm exceeds the cache. Two properties are load-bearing: - seeding/keep-alive never touch obsTick/hasObserved, so a seeded or retained entry is never served as a spoofed reply - only genuinely heard frames make a node servable; - copyUser() now requires hasFullUser, so a key-only warm seed can never stamp HAS_USER onto a nameless node via name-rehydration. Native suite: 64/64. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01KWHDDxY4U2XLYW2WJJ2oLT (cherry picked from commit 4f1b30d838b2f7b11173373b8f1ba97cf91bef6b) * Write-through NodeDB identity commits into the NodeInfo cache updateUser() is the single chokepoint through which every remote identity/key write enters NodeDB (key-verification keys stay in CryptoEngine's pending buffer until a NodeInfo carries them here), so one hook at its tail lets the NodeInfo cache reflect a commit immediately instead of waiting up to a minute for the reconcile sweep - which stays in place as the anti-entropy backstop. onNodeIdentityCommitted() upserts the full User: NodeDB's key is authoritative (a conflicting cached key is stale residue - replaced, provenance dropped), a keyless commit keeps an already-cached TOFU key, and signer provenance transfers only for the committed key. The observation stamp is never touched: knowledge is not observation, so a hook write can never make a never-heard node servable - covered by the new test alongside immediate copyUser/copyPublicKey visibility. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01KWHDDxY4U2XLYW2WJJ2oLT (cherry picked from commit e160b899728be8d951d4e7bdf5106c777832b917) * Purge TrafficManagement caches on explicit node removal NodeDB::removeNodeByNum() already forgets the node everywhere NodeDB owns - hot store, satellite stores, warm tier - but the TrafficManagement caches kept the deleted identity: the NodeInfo entry went on feeding the key pool (NodeDB::copyPublicKey) and name rehydration, and the unified slot kept its role/next-hop/dedup state, resurrecting the node on next contact. Add purgeNode(): clears both the unified cache slot and the NodeInfo cache entry, called from removeNodeByNum() alongside the warm-tier removal. Removal means full removal; passive eviction never calls this, and the reconcile sweep will not re-seed a node absent from both NodeDB tiers. Test: an observed identity plus a next-hop hint both vanish after removeNodeByNum(). Left for CI to execute per request. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01KWHDDxY4U2XLYW2WJJ2oLT (cherry picked from commit b5564c173337de99e15ef998a5c698550238fd3c) * Document the tmm-fix-2 / tmm-fix-superset reconciliation decisions Two parallel implementations of the same superset design are being combined on this branch. This decision matrix records every divergence and which side each reconciled feature takes, ahead of the code commits that apply them. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01KWHDDxY4U2XLYW2WJJ2oLT * Reconcile retention: no timed eviction, obsTick LRU, hourly seeding Adopt tmm-fix-superset's retention model (reconciliation decisions #3, #4, #9 in .notes/tmm-super-superset-reconciliation.md): - Entries are never evicted on a timer. The 7-day retention TTL and its third tick clock (retTick) are gone; wrap-safety rests entirely on the sweep's presence-bit saturation, and a quiet entry keeps its value (pubkey pool, name rehydration). Slots are reclaimed only by trust/membership-tiered LRU on insert - age scored by obsTick, with never-observed entries counting oldest - or by an explicit purge. - Membership refresh (entry -> NodeDB contains checks) runs every sweep; the heavy seeding pass runs at boot and then hourly, with the write-through hooks carrying immediacy in between. - Tick constants and helpers move into the header beside the existing UnifiedCache tick idiom, with static_asserts tying the tick windows to the fallback path's seconds/ms forms. Kept from tmm-fix-2 (decision #4): the warm tier is still seeded (key-only records via WarmNodeStore::entryAt) so the invariant remains cache superset-of hot AND warm, and the spareMembers guard still stops seeding from churning one member out for another when hot+warm exceeds the cache. Entry shrinks to 128 B (2000 entries = 256 kB). The TTL-based retention test is superseded by a no-timed-eviction test: a quiet keyed entry survives 9 days of sweeps while the serve gate saturates as before. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01KWHDDxY4U2XLYW2WJJ2oLT * Reconcile hooks: key-commit write-through, purgeAll, key-matched signer Adopt tmm-fix-superset's hook coverage (reconciliation decisions #5-#8): - onNodeKeyCommitted(): write-through for the two key-write sites that bypass NodeDB::updateUser - the Router's admin-key learn (TOFU-grade) and KeyVerificationModule's manual-verification commit (proven=true, the strongest provenance the cache can carry). Both were coverage gaps in the tmm-fix-2 write-through design. - updateUser's hook call now transfers signer status key-matched (isVerifiedSignerForKey) instead of the node's bare signed flag, and runs on acceptance rather than only on change. - purgeAll(): factoryReset() and resetNodes() clear both TMM tables - removal-is-full-removal applies to bulk resets too, without relying on the usual post-reset reboot. - purgeNode() reuses the existing finders instead of open-coded scans and logs the (user-initiated) purge. - peekNodeInfoFlagsForTest(): flag introspection so saturation and membership tests can assert sweep effects directly. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01KWHDDxY4U2XLYW2WJJ2oLT * Merge the two branches' test suites for the reconciled design Port tmm-fix-superset's unique tests, adapted to run natively under TMM_HAS_NODEINFO_CACHE (reconciliation decision #11): - keyHook_upsertsAndGovernsProvenance: TOFU learn -> manual-verification upgrade -> NodeDB-senior rotation resets provenance. - tickSaturation_sweepClearsObserved: the sweep saturates hasObserved past the serve window, the entry persists (no TTL), and a full 256-tick clock wrap cannot alias a saturated stamp back to fresh. - sweepMembershipMarking: reconciliation seeds a hot identity as member+fullUser+unobserved; the next sweep clears membership once the node leaves NodeDB, while the entry itself persists. tmm-fix-2's tests (warm pin, hot/warm seeding, updateUser write-through, removal purge - now also asserting the entry is gone via the new peek hook - and the no-timed-eviction rewrite) were already on this branch. Suite now counts 70 test functions. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01KWHDDxY4U2XLYW2WJJ2oLT * Adapt cherry-picked and seeding tests to the reconciled semantics Two tests needed updating for interactions the validation run surfaced: - The #11035 test (ignoresUnsignedSignerIdentity) was written for a world without the native NodeInfo cache: it needs the NodeDB fallback path (dropNodeInfoCacheForTest) and the signed replay gate satisfied for its target, like the other fallback tests on this branch. - The hot-seed test's "observed frame makes it servable" step now sends a signature-verified frame: its node is a known signer, and per #11035 an unsigned frame from a signer must not (and does not) drive cache writes - the gate working as designed. Full native suite: 70/70 under ASan. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01KWHDDxY4U2XLYW2WJJ2oLT * trunk fml * Doc tidyup * TrafficManagement: key NodeInfo-cache maintenance to its own guard runOnce() nested the entire NodeInfo maintenance block (tick-stamp saturation, membership refresh, boot/hourly reconcile) inside #if TRAFFIC_MANAGEMENT_CACHE_SIZE > 0, so a variant overriding the unified cache to 0 on a PSRAM board would compile the NodeInfo cache without its maintenance: hasObserved would never saturate, obsTick would alias past its 12.8 h wrap, and the 6 h staleness gate - whose wrap-safety explicitly depends on the sweep - would serve spoofed replies for long-gone nodes. Extract maintainNodeInfoCacheLocked(), guarded by TMM_HAS_NODEINFO_CACHE alone, and give runOnce() the same independent-guard structure purgeAll() already has. Also close the runtime cousin of the same mismatch: the write-through hooks (onNodeIdentityCommitted / onNodeKeyCommitted) now no-op while moduleConfig.has_traffic_management is off. Previously they kept filling the cache from NodeDB commits while runOnce() returned INT32_MAX, accumulating entries that were never swept or reconciled. Purges and reads stay ungated: removal must always work, and reads just miss an empty cache. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * TrafficManagement: forward throttled NodeInfo requests instead of consuming them The direct-response throttle returned true, which made handleReceived() STOP the request: within the 30 s window a repeat request was neither answered nor forwarded toward the genuine target, and the call site still logged "respond" and bumped nodeinfo_cache_hits for a reply that was never sent. A requester whose first reply was lost on a noisy link got silence for the whole window. Return false instead: the request flows through normal relay handling, so the genuine node (or another cache-holder) can answer, while our own spoofed TX stays bounded. Repeats of the same packet id are already absorbed by the router's duplicate detection, and the stats counter now only counts replies actually sent. Also log purgeNode() only when a slot was actually cleared - it runs for every NodeDB removal, including nodes the caches never tracked. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * TrafficManagement: consolidate NodeInfo-cache #if sprawl into one region Fourteen scattered #if TMM_HAS_NODEINFO_CACHE guards (one per function, each with its own #else stub of (void) casts) collapse into a single guarded region holding every NodeInfo-cache-only function, terminated by one #else block of no-op stub definitions. Because the stubs now exist on every build, the call sites in runOnce(), purgeNode() and purgeAll() drop their guards too; inner guards remain only for orthogonal features (PKI, warm tier) and for real conditional work (constructor allocation, PSRAM paths in shouldRespondToNodeInfo). No behavior change. Compile-checked on both sides of the macro: the native app build (stubs) and the native unit-test build (region). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Router: resolve sender key only for PKI-decrypt candidates perhapsDecode() resolved the sender's public key unconditionally for every encrypted packet, before even checking whether the packet could be PKI-decrypted (channel 0, unicast to us). Since copyPublicKey() grew its TrafficManagement fallback tier, a full hot+warm miss - the common case for channel traffic from senders outside both NodeDB tiers - additionally walked the 2000-entry NodeInfo cache under its lock, per packet, for a key that was then discarded. Move the resolution inside the PKI-candidate branch; remotePublic and haveRemoteKey had no consumers outside it. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * TrafficManagement: refresh NodeInfo membership hourly, not per-sweep The 60 s sweep re-derived isMember with a per-entry NodeDB lookup: O(entries x members) - about 700k node-number comparisons per minute on a full PSRAM cache (2000 entries x 250 hot + 100 warm), with strided PSRAM reads, all while holding cacheLock against the packet path. Move the refresh into the hourly reconcile pass, which is already O(members x entries): after the seeding loops (so the upsert pass still sees last pass's bits for spareMembers protection), clear every isMember bit and re-mark from both NodeDB tiers - including keyless warm records, which seed nothing but are still members. Accepted tradeoff, now documented on the field: membership lags a passive NodeDB eviction by up to an hour (the entry just stays LRU-sticky slightly longer). Additions stay immediate via the write-through hooks, explicit removals via purgeNode(). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * NodeDB: centralize bare-key commits in commitRemoteKey() The two key-write sites that bypass updateUser() (admin-channel learn in Router::perhapsDecode, manual verification in KeyVerificationModule) each carried their own #if HAS_TRAFFIC_MANAGEMENT write-through boilerplate - the pattern this PR itself had to retrofit twice, and which any future direct key write would silently miss, leaving the TrafficManagement cache divergent until the next hourly reconcile. Add NodeDB::commitRemoteKey(n, key32, KeyCommitTrust): writes the key to the hot store and routes the TrafficManagement write-through in one place, with provenance explicit at the call site (AdminChannelProven maps to TOFU-grade, ManuallyVerified to proven). The bypass sites keep their reason for existing - bare-key commits with provenance that updateUser's User-payload/TOFU-pin path cannot express - but no longer know about TrafficManagement at all; both stale includes are dropped (Router's was already unused on develop). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * docs: tick wrap-safety analysis, updated cadences, CodeRabbit doc fixes node_info_stores.md gains a "Tick clocks and wrap safety" section recording which mechanism keeps each modular tick clock honest: the unified-cache clocks pair the 60 s sweep with read-time window resets, the NodeInfo obs/resp clocks are sweep-only (hence the compile-time invariant that maintainNodeInfoCacheLocked() is guarded by TMM_HAS_NODEINFO_CACHE alone), and the warm tier is immune by design - absolute unix-seconds, no wrap until 2106. Also brought current with this series: membership refresh moved to the hourly reconcile, throttled direct-response requests forward instead of being consumed, the commitRemoteKey() bare-key funnel, and the module-disabled gate on the write-through hooks. CodeRabbit doc review (PR #11050): present the NodeInfo payload cache as the third *identity* tier with the unified cache beside the chain, and describe NodeInfoLite's flattened fields / satellite copy-out accessors instead of the removed nested members. Two stale docs/tmm_node_stores.md references in the header now point at the real file. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * TrafficManagement: adapt tests to forwarded throttles and hourly membership Throttle tests (PSRAM and NodeDB-fallback paths): a request inside the 30 s window now CONTINUEs into normal relay handling instead of being consumed - assert no spoofed TX, no NAK suppression, and (PSRAM path) that nodeinfo_cache_hits counts only replies actually sent. Membership test (renamed reconcileMembershipMarking): the per-minute sweep no longer refreshes isMember, so the test now pins down both halves of the new contract - the very next sweep after a passive NodeDB eviction still shows the stale member bit (the documented up-to-an-hour lag), and a reconcile interval's worth of sweeps clears it. Disabled-module test additionally proves the write-through hooks share the has_traffic_management gate: a key commit while disabled must not land in the NodeInfo cache. Not covered here: a build permutation with TRAFFIC_MANAGEMENT_CACHE_SIZE overridden to 0 (the configuration the maintenance-guard fix protects) would need its own PlatformIO env plus guards on every unified-cache test - left as a follow-up. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * expand test coverage * nitpicks and docs update * more comment fixes * nitpicks * test: make TMM fixture cleanup abort-safe in tearDown() Several tests reset per-test global state (trafficManagementModule pointing at a stack `module`, owner.public_key, the RTC fake clock) only after their assertions. A TEST_ASSERT_* failure longjmps out and skips that trailing cleanup, leaving the state to dangle into later cases - concretely, the next setUp()'s resetNodes() would call trafficManagementModule->purgeAll() on a destroyed object. Move the resets into tearDown(), which runs unconditionally between tests, and drop the now-redundant per-test cleanups. Addresses a CodeRabbit review comment on PR #11050. clod helped too * docs tidy * Throttle NodeInfo direct responses A direct response is addressed to the requesting packet's from field, which is unauthenticated, and is sent by every neighbour that holds the target in cache. One request therefore makes several nodes transmit at an address the requester chose, and nothing limited how often that could be repeated. Replies are now spaced per requester, which bounds how much any single node can be made to receive, plus a global floor on how much airtime the feature can consume. The check sits where a reply is about to be sent, so requests declined for other reasons do not consume the budget. * test: reconcile TMM direct-response tests with #11104 throttle Makes the suite green for now. #11104's per-requester + global-floor throttle (60 s) is layered over the branch's per-target throttle, so the three existing "served again" checks (psram, fallback, sweep) now use a fresh requester to avoid the per-requester window masking the mechanism each one actually exercises. Adds a dedicated test for the new per-requester + 1 s global-floor behaviour (the reflector-flood gap the per-target throttle leaves open, and the bound that survives on non-PSRAM builds). Deferred: the branch's now-redundant fallback global stamp (nodeInfoFallbackLastResponseMs) is left in place as a follow-up cleanup. Clod wants credit. * TMM: unify direct-response throttles into per-sender + per-target RAM tables Replace the two path-specific NodeInfo-spoof throttles (per-entry respTick on the PSRAM cache path, single global stamp on the fallback path) with three symmetric bounds that behave identically with and without PSRAM: - per requester (60 s): how much any one node can be made to receive; - per target (60 s): how often we vouch for the same identity; - global floor (1 s): total airtime, the backstop once an attacker cycles requester/target past the 8-slot tables. Both axes are fixed 8-slot LRU tables in internal RAM (not the PSRAM NodeInfo cache), compared by wrap-safe uint32 ms subtraction, so there is no tick clock and no sweep to maintain. directResponseAllowed() resolves both slots before stamping either, so a reply one axis throttles never consumes the other's budget, and records the send itself. Retires: nodeInfoFallbackLastResponseMs, kNodeInfoResponseThrottleMs, nowStampMs, the respTick byte + hasResponded bit on every cache entry, currentRespTick/kNodeInfoRespTickMs/kNodeInfoThrottleTicks, and the sweep's respTick clear. Renumbers peekNodeInfoFlagsForTest (drops the responded bit). Tests: the psram/fallback throttle tests now assert the per-target axis at 60 s, isolated via a different requester, proving it holds with and without PSRAM; perRequesterAndGlobalFloor isolates the other two axes; the obsolete respTick wrap-safety test is removed. 83/83 native cases pass. clod helped too * whats up, doc? * trimming the comments * test: bump native-suite-count to 38 after upstream rebase Upstream develop carries 38 test_* suite directories but its native-suite-count file still reads 37 (a suite was added without bumping the count). Rebasing onto develop inherits that stale file, so the runner flags AMBER. Correct the registered total to 38. clod helped too --------- Co-authored-by: Claude <noreply@anthropic.com> Co-authored-by: Thomas Göttgens <tgoettgens@gmail.com> |
||
|
Thomas Göttgens
|
ed3afdbc61 | TrafficManagement: don't learn identity from a signer's unsigned NodeInfo (#11035) | ||
|
TomandClaude Fable 5
|
d846780a9b |
More fuzz tests and small fixes for the findings (#10864)
* first pass tests * more tests * Fix two crafted-admin-packet crashes found by the E5 fuzzer Both are reachable from an authorized admin (local from==0, admin channel, or PKC) - remote DoS: 1. SIGFPE in LoRa config validation. A set_config LoRaConfig with use_preset=false and bandwidth=0 makes freqSlotWidth 0, so numFreqSlots is 0 and `hash(name) % numFreqSlots` (RadioInterface.cpp) divides by zero. Guard the modulo; the existing channel_num check then rejects/ clamps the config. 2. Stack overflow in Channels::getKey. A SECONDARY channel at the primary slot with an empty PSK recursed into getKey(primaryIndex) forever. Skip the primary-key borrow when chIndex == primaryIndex. Re-enable the E5 admin fuzzer to hit both triggers again (use_preset both ways incl. bandwidth 0, plus the set_channel tag) as regression guards. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Correct fuzz-test invariants after the crash fixes - E5 admin fuzz: node eviction under a filling NodeDB is legitimate, so assert only the bounded-count invariant, not that a specific seed node survives 6000 mutating ops. - TMM blitz: scope off the nodeinfo direct-response send path (it needs a fully-wired MeshService/phone queue the fixture doesn't provide; the deterministic directResponse tests cover it). The crafted-nodenum rate/unknown/position cache stress is unchanged. clod helped too * realistic tests * test: dedup fuzz RNG into shared test/support/DeterministicRng.h The four in-tree fuzz suites (test_fuzz_decode, test_fuzz_packets, test_hop_scaling, test_traffic_management) each carried a byte-identical copy of the seeded 64-bit LCG (rngSeed/rngNext/rngByte/rngRange). Hoist it into one shared header so there is a single generator to reason about and no risk of the copies drifting. static inline keeps per-suite state per translation unit and avoids -Wunused-function for suites that don't use every helper. Also corrects a stale comment in test_traffic_management (the blitz's nodeinfo direct-response path is intentionally left off). No behavioral change: same constants, same per-suite seeds. clod helped too * test: fuzz uncovered ProtobufModule handlers and the MQTT downlink ingress Extend the in-tree fuzz coverage to packet sources that previously had none: - test_fuzz_packets E8/E9/E10: drive PositionModule, DeviceTelemetryModule and NeighborInfoModule at handleReceivedProtobuf directly (via using-shims, bypassing the ProtobufModule reply/send path so no router is needed). The fixture already stands up nodeDB/service/channels, and nodeStatus/powerStatus are auto-initialized in main.cpp, so no new globals are required. Adds a shared fuzzRxHeader() helper for crafting adversarial RX packet headers. - test_fuzz_decode: add meshtastic_KeyVerification to the decode table. The KeyVerification and StoreForward handler paths are documented as decode-level only, with the concrete reason each is intrinsic (private-state gating / PSRAM + self-pointer wiring), not a fixture gap. - test_mqtt: test_receiveFuzzServiceEnvelope blitzes the non-RF broker-push ingress (onReceiveProto) two ways - raw garbage bytes that must fail envelope decode cleanly, and a well-formed ServiceEnvelope wrapping a crafted inner MeshPacket over crafted channel_id/gateway_id - exercising the channel match, isFromUs, XEdDSA receive policy and perhapsDecode chain. Adds a deliverRaw() passthrough to MQTTUnitTest. All under the coverage env (ASan/LSan). No firmware/src changes. Full sweep GREEN 27/27, 544 cases. clod helped too * Harden LoRa/channel config against crafted admin messages; consolidate test helpers Production (review findings on the hot-fuzz crash fixes): - Clamp bandwidth at the source (clampBandwidthKHz) in checkOrClampConfigLora and applyModemConfig so numFreqSlots can never be 0 for any consumer; a bandwidth-0 set_config previously passed validation and re-armed the SIGFPE on the next applyModemConfig. - Guard applyModemConfig's hash % numFreqSlots (the validator's sibling modulo was fixed earlier but this one was still unguarded). - Enforce the primary-channel invariant in Channels::onConfigChanged: a config demoting every slot now re-promotes the stale SECONDARY slot (keeping its key) or restores the default channel if the slot is DISABLED, instead of leaving every getPrimaryIndex() reader on a non-primary slot. The getKey recursion guard stays as defense-in-depth. Tests: - New test/support/MockMeshService.h and AdminModuleTestShim.h replace four byte-identical mocks and three divergent admin shims (test_mqtt's capturing mock is genuinely different and stays). - DeterministicRng.h: add rngFill() (replaces 14 hand-rolled fill loops) and rngEdgeNodeNum() (unifies the three NodeNum boundary pools). - Extract fuzzChannelSettings() shared by the set_channel case and fuzzBeacon. - fuzzBeacon: the un-terminated branch now fills the whole buffer with non-NUL bytes so the strnlen bound is actually stressed (~50% of iterations, not ~4%). - E6 beacon fuzz: replace the TEST_ASSERT_TRUE(true) tautology with real invariants (handler never consumes; offers land in lastReceivedOffer keyed to the sender). - Trim the seven over-long comment blocks flagged against the 1-2 line rule; the FINDINGS trailer moves to this commit message (see production notes). Full native suite GREEN 27/27 under the coverage (ASan/LSan) env. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Clamp UTF-8 char length in the emote walkers; add test_fuzz_emotes A TEXT_MESSAGE payload is opaque protobuf bytes, so PB_VALIDATE_UTF8 never screens it - invalid UTF-8 and truncated multi-byte lead bytes reach the emote/width render path verbatim. EmoteRenderer's walkers advanced by utf8CharLen(lead) without clamping to the bytes actually remaining, so a truncated lead (e.g. a lone 0xF0, which claims 4 bytes) near the end of the buffer made getUtf8ChunkWidth's memcpy read past the string. ASan confirms a heap-buffer-overflow READ from measureStringWithEmotes. Add utf8CharLenClamped() and use it at every walk site (width measure, truncation cut-loop, and the draw-path text-run/chunk builders); the one already-guarded site (matchAtIgnoringModifiers) is unchanged. New test/test_fuzz_emotes drives measureStringWithEmotes and truncateToWidth over adversarial byte strings (biased to embed/end in truncated multi-byte leads) in exact-sized heap buffers so any over-read is a hard ASan fault. Its headless display uses a synthetic font (firstChar 0, fontData centered in a large buffer) so the stock OLEDDisplay::getStringWidth - which indexes the font jump table with a signed char and over-reads for any byte >= 0x80 - does not mask the finding. native-suite-count bumped 27 -> 28. Full native suite GREEN 28/28 under the coverage (ASan/LSan) env. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Keep emote width measurement in-bounds for non-ASCII bytes OLEDDisplay::getStringWidth (the utf8=false path EmoteRenderer uses on default builds) indexes the font jump table by (c - firstChar) with a signed char and no bounds check, so any byte outside printable ASCII - high bytes from UTF-8 text, but also a stray control byte like 0x0A - reads outside the font array. On-device this reads adjacent flash and returns a garbage width; under ASan the test_fuzz_emotes fuzzer flags it as a global-buffer-overflow, and it made the non-ASCII width measurement meaningless either way. The OLED driver is a pinned upstream dependency, so guard it firmware-side in EmoteRenderer's getStringWidth helper: measure a sanitized copy where any byte outside [0x20, 0x7E] counts as a '?' placeholder. Printable ASCII is unchanged and the UA/RU lookup path is untouched. test_fuzz_emotes now drives a real ArialMT font instead of the synthetic in-bounds font it needed before this fix, so the suite exercises the true production width path (utf8CharLen clamp + this sanitizer) end to end. The same fuzzer tripped the global-buffer-overflow before this change. Full native suite GREEN 28/28 under the coverage (ASan/LSan) env. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com> |
||
|
Tom
|
3becaf2d95 | emdashes begone (#10847) | ||
|
TomandClaude Sonnet 4.6
|
c51c01607d |
Traffic Management module: dedup, rate limiting, role-aware policing (#10706)
Adds the Traffic Management module (TMM) plus the NodeDB/warm-store and next-hop foundations it builds on: - Unified per-node cache (flat array, 8-bit relative ticks) shared by all features; role-aware throttles for tracker / lost-and-found. - Position deduplication: drop unchanged position rebroadcasts within a configurable interval; precision driven off the channel ceiling (clamped to the public-key max on well-known channels). Enabled by default at 11h. - Per-node rate limiting and unknown-packet filtering (config-driven; a non-zero companion field enables each feature -- no bool toggles). - NodeInfo direct response from cache with role-based hop clamps. - Persistent next-hop overflow store: confirmed hops have no TTL, are seeded from NodeInfoLite at boot, and survive hot-store eviction. - Three-tier sender-role resolution (hot NodeInfoLite -> warm store -> TMM cache). Role is cached write-time (seeded on first track, refreshed from NodeInfo), pins its cache entry like a next-hop hint, and is evicted last. - Warm store caches device role + protected category across reboot/eviction. - PositionModule stationary floor for tracker / lost-and-found. - PSRAM gating for warm/satellite/TMM cache sizes; STM32WL excluded. Protobufs: TrafficManagementConfig trimmed to the five uint32 fields actually used; submodule repointed to protobufs develop. Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com> |
||
|
|
22072c5f4b |
Pr1.5 tmm nexthop (#10745)
* TrafficManagement: flat unified cache + persistent next-hop overflow store Reworks the TrafficManagementModule cache layer (policing behaviour unchanged from upstream) and adds a routing-hint overflow store: - Flatten the ring: replace the cuckoo-hashed unified cache and the bucketed PSRAM NodeInfo index with plain flat arrays + linear scan (same idiom as WarmNodeStore). At LoRa packet rates an O(n) scan of the cache is negligible, and it removes a large amount of hashing/displacement complexity. The cache entry is 11 B; timestamps use a uniform +1 presence-offset so a 0 byte always means "empty" across every sub-store. Adds rebaseEpoch() so cached state survives the ~19 h relative-timestamp horizon instead of being flushed. - Next-hop overflow cache: setNextHop/getNextHopHint store a confirmed last-byte relay for a destination, written only from NextHopRouter's ACK-confirmed decision (and mirrored from TraceRoute). NextHopRouter::getNextHop falls back to this cache when the hot NodeDB has no hint, so DMs/relays to long-tail nodes keep routing after the node ages out of NodeInfoLite. - Persistence: preloadNextHopsFromNodeDB warm-starts the cache from persisted NodeInfoLite hints on first maintenance pass; next_hop entries are kept alive across the maintenance sweep (no TTL) and never clobbered by a stale preload. All packet-policing logic (rate limit, position dedup, unknown-packet drop, NodeInfo direct response, hop exhaustion) is the existing upstream behaviour, untouched. HAS_TRAFFIC_MANAGEMENT defaults on so the module is compiled in. (see note). Tests: upstream policing suite now actually runs (adds the MeshTypes.h include that gates HAS_TRAFFIC_MANAGEMENT) plus 4 next-hop tests. Role-aware throttles, politeness, precision clamp, port-interval and mesh-radius gating — and the rate-limit >255 saturation fix — are deferred to the advanced-TMM branch. Note: default dedup movement grid moves to ~91m, which also means 1.5km required to end up with the same signature position - coarser and therefore further than before. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * TrafficManagement: fix cppcheck constVariablePointer warning `node` in preloadNextHopsFromNodeDB() is never written through — mark it const to satisfy cppcheck's constVariablePointer check in CI. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * Add multi-hop NextHop recovery tests and unit tests for routing reliability - Introduced a new test suite for multi-hop NextHop directed-message delivery and relay recovery in `test_nexthop_multihop_recovery.py`. This includes tests for end-to-end delivery and recovery after relay drop. - Implemented unit tests in `test_main.cpp` for NextHop routing reliability mitigations, covering: - M1: Ambiguity-aware last-byte resolution. - M2: NextHopRouter's strict-neighbor gate and hop limit checks. - M3: Route-health freshness and failure decay. - Enhanced mock classes to facilitate controlled testing of node behaviors and routing logic. * grafting fixed * Address Copilot review for PR #10735 (NextHop improvements) - docs/nexthop-routing-reliability.md: update status from "no code changes yet" to reflect that mitigations and tests are implemented RAM pressure and MIGRATION_VERBOSE concerns addressed upstream in PR2.5 (per-platform TRAFFIC_MANAGEMENT_CACHE_SIZE) and PR2 (verbose default=0) respectively; (0,0) sentinel fixed in PR2.5. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * CI: fix cppcheck constVariablePointer and test include path - NextHopRouter.cpp: qualify two RouteHealth *h locals as const — only read for stale-route checks, never mutated through the pointer - Router.cpp: qualify meshtastic_NodeInfoLite *node as const in shouldDecrementHopLimit — only read for favorite/role predicate - test_position_module/test_main.cpp: change bare PositionModule.h to modules/PositionModule.h — build_flags sets -Isrc, not -Isrc/modules, so the bare form fails to resolve in the native PlatformIO test env Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * WarmStore: cache device role + protected category in last_heard low bits Steal the low 6 bits of WarmNodeEntry.last_heard to carry an evicted node's device role (4 bits) and a protected category (2 bits) for the hop-trim path, at zero record-size cost (entry stays 40 B; no RAM/flash growth). The high bits remain a real unix-seconds timestamp, quantised to 64 s — ample for warm LRU ordering of long-tail nodes. - absorb() packs role/protectedCat; place()/ring replay store the raw word so metadata round-trips through flash. LRU compares masked time (warmTimeOf). - take() rehydration masks the metadata bits and restores the cached role so a re-admitted node isn't stuck at CLIENT until its next NodeInfo. - NodeDB classifies the category (favorite/ignored/verified -> Flag; tracker/sensor/tak_tracker -> Role) at each eviction site. - WarmNodeStore::lookupMeta() exposes role/category to consumers. - Bump WARM_RING_MAGIC (WRNG->WRN2): old rings read as erased and rebuild; warm data is a non-critical evictee cache, so discard-on-upgrade is safe. Tests: test_warm_store 11/11 (new meta round-trip + quantisation-aware ordering); NodeDB compiles (test_nodedb_blocked 4/4). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * WarmStore: migrate v1 rings/files by discarding last_heard, not the data Previously the WRNG->WRN2 magic bump treated old rings as erased, discarding all warm entries — including the PKI public keys that let evicted nodes keep decrypting DMs. Instead, read v1 (WRNG / WRM1) records and keep each node's identity + public key, discarding only last_heard (its low bits would otherwise be misread as the new role/protected metadata). Records re-rank and re-learn their role on next contact. - Ring backend (nRF52840): ringReadHeader accepts both magics and reports v1 via an out-param; replay zeroes last_heard for v1 records. If the active head page is v1, force a rotation so new v2 records never land in a v1-headered page (which would discard their freshly-set role on the next load). Legacy pages convert to v2 as the ring rotates. - File backend (warm.dat): bump WARM_STORE_MAGIC WRM1->WRM2; accept WRM1, verify CRC against the stored bytes, then discard last_heard and mark dirty so the next save rewrites as v2. Tests: test_warm_store 12/12 (adds test_ws_v1_migration_discardsLastHeard: key survives, role/protected reset). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * WarmStore: guard role bit-width + test eviction carries role/protected - static_assert that the device role enum still fits the 4-bit warm metadata field (WARM_ROLE_MASK); fails the build loudly if a new role is added past 15 rather than silently truncating role on eviction. (Max role today = 12.) - Add test_migration_carriesRoleAndProtectedIntoWarm: a demoted TRACKER lands in the warm tier with its key, role=TRACKER and protected category=Role; a demoted CLIENT carries role=CLIENT/None. Exercises the NodeDB eviction path + warmProtectedCategory classification (the warm-store unit tests only cover absorb() directly). Tests: test_nodedb_blocked 5/5. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix copilot comments * fix(test): restore #if HAS_TRAFFIC_MANAGEMENT guard in TMM test The rebase onto PR1.5 lost the top-level HAS_TRAFFIC_MANAGEMENT guard that PR1.5 introduced, leaving the #else/#endif tail orphaned and causing compile errors on non-TMM builds. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Co-authored-by: Ben Meadors <benmmeadors@gmail.com> |
||
|
Ben Meadors
|
93f87c57b9 | MacOS fixes | ||
|
|
94bb21ecc7 |
2.8: NodeDB shrink, decoupling, and restructuring (#10413)
* 2.8: NodeDB refactor to decouple satellite entries and decrease size * Regen * Refactor node mute handling to use dedicated functions for clarity and consistency * Develop ref * Fix NodeDB review follow-ups Agent-Logs-Url: https://github.com/meshtastic/firmware/sessions/6b1d6cf6-ed6b-43b6-95cb-8e141757664e Co-authored-by: thebentern <9000580+thebentern@users.noreply.github.com> * Address review validation nits Agent-Logs-Url: https://github.com/meshtastic/firmware/sessions/6b1d6cf6-ed6b-43b6-95cb-8e141757664e Co-authored-by: thebentern <9000580+thebentern@users.noreply.github.com> * Trunk * Potential fix for pull request finding Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com> * Extract legacy NodeDatabase migration * Fix remaining NodeDB review issues Agent-Logs-Url: https://github.com/meshtastic/firmware/sessions/c76b9a5a-7244-4fbc-9ef0-98091d8caaea Co-authored-by: thebentern <9000580+thebentern@users.noreply.github.com> * Fixes * Trunk * Fix latest review compile follow-ups Agent-Logs-Url: https://github.com/meshtastic/firmware/sessions/5198da01-ec4c-4c16-8a09-68b8e6d5d410 Co-authored-by: thebentern <9000580+thebentern@users.noreply.github.com> * Fix cppcheck style warnings Agent-Logs-Url: https://github.com/meshtastic/firmware/sessions/e60287ba-4ece-46e0-83d8-a6d89664c0bb Co-authored-by: thebentern <9000580+thebentern@users.noreply.github.com> * Change pointer type for mesh node in set_favorite function * Change pointer types for mesh node references to const in multiple applets * Add NodeDB layout v25 documentation and migration guidelines * Remove tests for uninitialized PacketHistory state due to undefined behavior * Fix code block formatting in copilot instructions --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com> |
||
|
Clive BlackledgeandCopilot
|
016e68ec53 |
Traffic Management Module for packet forwarding logic (#9358)
* Add ESP32 Power Management lessons learned document Documents our experimentation with ESP-IDF DFS and why it doesn't work well for Meshtastic (RTOS locks, BLE locks, USB issues). Proposes simpler alternative: manual setCpuFrequencyMhz() control with explicit triggers for when to go fast vs slow. * Addition of traffic management module * Fixing compile issues, but may still need to update protobufs. * Fixing log2Floor in cuckoo hash function * Adding support for traffic management in PhoneAPI. * Making router_preserve_hops work without checking if the previous hop was a router. Also works for CLIENT_BASE. * Adding station-g2 and portduino varients to be able to use this module. * Spoofing from address for nodeinfo cache * Changing name and behavior for zero_hop_telemetry / zero_hop_position * Name change for exhausting telemetry packets and setting hop_limit to 1 so it will be 0 when sent. * Updated hop logic, including exhaustRequested flag to bypass some checks later in the code. * Reducing memory on nrf52 nodes further to 12 bytes per entry, 12KB total using 8 bit hashes with 0.4% collision. Probably ok. Adding portduino to the platforms that don't need to worry about memory as much. * Fixing hopsAway for nodeinfo responses. * traffic_management.nodeinfo_direct_response_min_hops -> traffic_management.nodeinfo_direct_response_max_hops * Removing dry run mode * Updates to UnifiedCacheEntry to use a common cache, created defaults for some values, reduced a couple bytes per entry by using a resolution-scale time selection based on configuration value. * Enhance traffic management logging and configuration. Updated log messages in NextHopRouter and Router to include more context. Adjusted traffic management configuration checks in AdminModule and improved cache handling in TrafficManagementModule. Ensured consistent enabling of traffic management across various variants. * Implement destructor for TrafficManagementModule and improve cache allocation handling. The destructor ensures proper deallocation of cache memory based on its allocation source (PSRAM or heap). Additionally, updated cache allocation logic to log warnings only when PSRAM allocation fails. * Update TrafficManagementModule with enhanced comments for clarity and improve cache handling logic. Update protobuf submodule to latest commit. * Creating consistent log messages * Remove docs/ESP32_Power_Management.md from traffic_module * Add unit tests for Traffic Management Module functionality * Fixing compile issues, but may still need to update protobufs. * Adding support for traffic management in PhoneAPI. * Making router_preserve_hops work without checking if the previous hop was a router. Also works for CLIENT_BASE. * Enhance traffic management logging and configuration. Updated log messages in NextHopRouter and Router to include more context. Adjusted traffic management configuration checks in AdminModule and improved cache handling in TrafficManagementModule. Ensured consistent enabling of traffic management across various variants. * Implement destructor for TrafficManagementModule and improve cache allocation handling. The destructor ensures proper deallocation of cache memory based on its allocation source (PSRAM or heap). Additionally, updated cache allocation logic to log warnings only when PSRAM allocation fails. * Update TrafficManagementModule with enhanced comments for clarity and improve cache handling logic. Update protobuf submodule to latest commit. * Add mock classes and unit tests for Traffic Management Module functionality. * Refactor setup and loop functions in test_main.cpp to include extern "C" linkage * Update comment to include reduced memory requirements Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com> * Re-arranging comments for programmers with the attention span of less than 5 lines of code. Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com> * Update comments in TrafficManagementModule to reflect changes in timestamp epoch handling and memory optimization details. * bug: Use node-wide config_ok_to_mqtt setting for cached NodeInfo replies. * Better way to handle clearing the ok_to_mqtt bit * Add bucketing to cuckoo hashing, allowing for 95% occupied rate before major eviction problems. * Extend nodeinfo cache for psram devices. * Refactor traffic management to make hop exhaustion packet-scoped. Nice catch. * Implement better position precision sanitization in TrafficManagementModule. * Added logic in TrafficManagementModule to invalidate stale traffic state. Also, added some tests to avoid future me from creating a regression here. * Fixing tests for native * Enhance TrafficManagementModule to improve NodeInfo response handling and position deduplication logic. Added tests to ensure local packets bypass transit filters and that NodeInfo requests correctly update the requester information in the cache. Updated deduplication checks to prevent dropping valid position packets under certain conditions. --------- Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com> |