Two runs on one host destroyed each other's containers, producing
mid-test "No such container" failures that look like real defects. The
automated builder runs a full local CI on the same box every few minutes,
so the machine is contended almost always and this has red-ed both a hand
run and an automated gate.
Two independent causes are fixed here. Container names were hardcoded, and
docker names are global rather than scoped by compose project, so two runs
collided on the same name; every name now takes an optional suffix that
the harness sets from the run id. And the cleanup sweep matched a label
shared by every run, so one run's teardown force-removed another's
containers; resources now also carry a per-run label and the sweep can be
narrowed to it.
A third hazard turned up that was not in the original report: the sidecar
suite passes explicit compose project names, which override the run-scoped
project and put it outside the shared prefix entirely. Its project names,
network, and derived container references are now scoped too.
Both are default-off. With the suffix unset, names render exactly as they
do today and a bare compose invocation is unchanged, which is what keeps
the hosted CI and the documentation correct. A cleanup run with no run id
still reaps everything, which is what a manual "clear the box" wants.
The literal-name sweep was not sufficient: six scripts build container
names dynamically from node labels, and two suites create their own
containers outside compose. Those are handled at their construction sites.
Verified: syntax check on all modified scripts; compose validation on
every modified file with the suffix both set and unset; and a synthetic
two-run reproduction that shows the old cleanup destroying a bystander run
and the new one leaving it alone.
Known gap: subnets are still hardcoded, so two concurrent full runs will
still collide on address-pool overlap. That fix reaches into topology
configs, chaos scenarios, diagrams and production source, so it is left
for its own change rather than half-done here.
The rekey suites' Phase 4 polled until the initiator-cutover count
reached its value at phase entry plus two, while Phase 6 asserts an
absolute minimum of four. On an unloaded host the ping phases can
outrun the jittered first rekey cycle, so Phase 4 can begin with a
single cutover on the books, release at three, and leave Phase 6 to
lose a sub-second race against the remaining first-cycle cutovers.
A rekey-outbound-only run failed exactly this way: five counted
cutover lines by t=32.6s of container life, with the Phase-6 count
snapshot at t~32.4s seeing three.
Floor the Phase-4 wait target at four so a released wait guarantees
the Phase-6 assertion — the docker-logs count is monotone over a
container's lifetime. A genuine rekey stall still times out the
bounded wait and fails Phase 6 with the real count.
Also drop the second-cycle framing from the comment, the phase
banner, and the assertion description: with rekey timers resetting at
each cutover a second cycle cannot begin inside this test's window,
so the counted events are first-cycle cutovers spread across the
topology's links, emitted once or twice per completed link rekey
depending on which side's promotion path logs them.
Phase 4 of the rekey suite waited for only one more initiator cutover
(guaranteeing three) while Phase 6 asserts at least four. The fourth
cutover then had to land in the brief window between the wait returning
and the Phase 6 log snapshot, so host load could push it past the window
and fail the run even though every cutover completed correctly. Wait for
two more cutovers, the full second rekey cycle, matching the Phase 6
threshold, so the asserted count is guaranteed before it is checked.
Replace the fixed post-first-rekey and post-second-rekey settle sleeps in
the rekey integration test, and the fixed-deadline baseline wait in the
interop test, with the deterministic wait_until_connected progress-aware
polling helper already used for the initial baseline convergence.
Each converted site fails fast with structured diagnostics when the mesh
is stuck and extends its deadline only while pairwise reachability is
still climbing, removing the wall-clock settle windows that produced
intermittent connectivity failures after a rekey.
Add wait_until_connected to the shared convergence helpers: it polls a
suite's own pairwise pings (the signal it actually asserts on), returns
as soon as every pair is reachable, extends its deadline while the
reachable-pair count is still climbing, and gives up only when progress
stalls.
Use it in the rekey, static-mesh, and sidecar suites in place of the
fixed wall-clock baseline timeout and the blind sleep, which timed out
under concurrent CI load while the mesh was still converging.
Closes the eventually-consistent gap in spanning-tree state
distribution. Every existing send_tree_announce_to_all call site
gates on a local state-change event (parent switch, self-root
promotion, ancestry change, peer promotion, parent loss). Once a
partition latches — for example a parent-switch announce stranded
in the brief cross-init handshake swap window, where the announce
arrives on a session-index whose decrypt-worker entry has been
unregistered — neither side's state changes again, so neither
side ever re-broadcasts. The existing 60 s check_periodic_parent_reeval
was a re-evaluation, not a re-broadcast: it short-circuited
silently on no-change. Production-side healing depended on
incidental link churn; lab harnesses with stable docker-bridge
links had no equivalent path.
Add a final else branch that fires send_tree_announce_to_all
unconditionally on the no-change path, alongside the existing
switch and self-promote arms. Receivers coalesce by sequence
comparison (ParentDeclaration::is_fresher_than) and short-circuit
at the `if !updated` gate in handle_tree_announce; same-sequence
repeats drop silently with no cascade. The per-peer 500 ms
rate-limiter is well below this 60 s cadence and does not suppress
the heartbeat broadcast.
The fix is a general protocol-robustness improvement: it addresses
any in-flight TreeAnnounce loss class, not only the specific
cross-init swap-window drop site.
testing/static/scripts/rekey-test.sh BASELINE_CONVERGENCE_TIMEOUT
60 -> 65 so a partition healed by the periodic broadcast at T+60
lands inside the convergence window. wait_for_full_baseline
early-exits on PASS, so successful reps see no extra wall-clock.
An FSP session rekey could leave the two endpoints holding different
key sets for a brief window: if a handshake message was lost in
transit, one side rotated to the new keys while the other did not.
Traffic sealed in one key epoch then reached a peer still on the
other epoch and failed to decrypt, producing bursts of AEAD
decryption failures and dropped connectivity until a later rekey
cycle reconverged the pair. Choreographing the cutover order cannot
close this window: any fixed ordering still leaves a skew that
packet reordering widens.
Make rekey correctness independent of cutover timing by overlapping
the key epochs on the receive path. During a rekey transition the
receiver trial-decrypts each frame against every live session it
holds: current, the not-yet-promoted pending session, and the
draining previous session. The K-bit becomes a hint that orders the
trial-decrypt cascade rather than a hard gate, and a frame that
authenticates against the pending session is itself the cutover
signal. No rotation ordering and no packet reordering can then cause
a decryption failure.
The pre-rekey Noise session is held in the `previous` slot until the
peer has demonstrably moved off it. Its drain deadline is anchored
on the most recent frame the peer authenticated against that slot,
refreshed each time the trial-decrypt cascade lands there, rather
than on a fixed wall-clock timer started unilaterally at the local
cutover. A peer that never received the new keys keeps authenticating
against `previous` and the slot stays live; without this, a fixed
timer would erase the only key set that could decrypt the peer's
frames, producing a permanent silent decrypt failure on a live data
path. A peer that never catches up is handled by the existing FSP
session liveness path rather than by silent decrypt failure.
The lost-handshake liveness gap is closed separately by retransmitting
the third rekey handshake message until the peer is confirmed on the
new keys, with a bounded retry budget after which the rekey cycle is
cleanly abandoned and retried on the next timer.
Adds unit tests covering the trial-decrypt cascade (epoch selection,
promotion on pending decrypt, reordered old-epoch stragglers after
cutover, per-slot replay-window integrity), the msg3 retransmission
lifecycle, and the peer-progress-aware drain retirement.
The Phase 1, Phase 3, and Phase 5 strict asserts each fire a
single ping per directed pair. Under low-level packet loss
(e.g. 1% i.i.d. per-direction loss from a CI runner under
pressure), a single-shot round-trip fails at ~2% per pair, so a
20-pair strict assert misses with probability
1 - (0.98)^20 = ~33% per phase from ICMP noise alone, well above
the routing-state signal the asserts are meant to catch.
ping_one gains a max_attempts parameter (default 1, preserving
existing call sites). On failure it retries up to
MAX_PING_ATTEMPTS-1 additional times with PING_RETRY_DELAY
seconds between attempts. Per-pair worst case under the defaults
(4 attempts, 1 s spacing, 5 s ping6 -W timeout) is 4*5 + 3 = 23 s;
per-rep worst case scales with the failing-pair count.
Successful retries log "OK (RTT, attempt N)"; exhausted retries
log "FAIL (after N attempts)".
The retry budget is wired into all three strict asserts:
- Phase 1 final ping_all (after wait_for_full_baseline converges)
- Phase 3 ping_all (post-first-rekey)
- Phase 5 ping_all (post-second-rekey)
The wait_for_full_baseline convergence loop itself stays
single-shot. Its job is to detect when the mesh first sees a
fully clean 20-pair batch, and retries inside the loop would
conflate transient ping loss with still-converging routing
state.
No daemon code changes.
Phase 5's per-pair connectivity check ran immediately after the second
rekey cycle, with no settle for routing reconvergence. Under
GitHub-runner CPU contention, post-rekey parent-switches and
coord-cache flushes can take longer than the per-ping 5s timeout for
a small fraction of pairs (1-3 of 20 typically), even though the
rekey mechanism itself completes cleanly. Phase 6 log analysis stays
all-green on these failed runs; the failure is purely connectivity
timing.
Mirror Phase 3's existing 12-second settle pattern: reuse REKEY_SETTLE
and emit the same banner shape. Two-line change. Cost on the success
path is a fixed 12s per suite run.
The Phase 1 pre-rekey baseline in `wait_for_full_baseline`
occasionally times out on GitHub-hosted runners with one ping pair
failing to converge inside the BASELINE_CONVERGENCE_TIMEOUT window.
Phases 2–6 always pass cleanly when this happens — the rekey itself
is fine, the mesh just hasn't finished spanning-tree + bloom-filter
convergence by the time Phase 1 starts pinging.
The wait loop returns as soon as all 20 pairs converge, so the cost
on the success path is unchanged (typical local CI returns well
under the old 36s). The bump only adds headroom on the failure
path. Operators previously worked around this with
`gh run rerun --failed`; this aims to retire that workaround for
the IK/XK lines.
Distinct from the next-branch XX rekey dual-init race, which is a
real protocol bug tracked separately.
Companion to the ethernet `accept_connections: false` rekey-deadlock
fix from earlier this release: the same dual-init failure mode shows
up over UDP when peers register by hostname, and the existing
addr_to_link-only carve-out in `should_admit_msg1` doesn't cover it.
The carve-out's first predicate keys `addr_to_link` by the literal
`TransportAddr` that `initiate_connection` inserted, which is the
hostname-form when a peer config carries a hostname (e.g.,
`core-vm.tail65015.ts.net:2121`). Inbound packets always arrive with
numeric source addrs because `udp_receive_loop` builds the
`TransportAddr` from the `SocketAddr` the kernel reports via
`recvfrom`. `TransportAddr` equality is byte-exact, so the two forms
don't match and the lookup misses. With `udp.accept_connections:
false` (or `udp.outbound_only: true`, which forces it false) the
gate then rejects the rekey msg1 from an established peer. The
dual-init tie-breaker stalls because the loser side never produces
msg2; both sides retry indefinitely and the winner side keeps
logging "Dual rekey initiation: we win, dropping their msg1" at 1Hz.
The earlier ethernet fix didn't generalize to this variant because
ethernet TransportAddrs are always numeric MAC bytes — both the
config-time form and the inbound-arrival form match identically.
Add a second predicate to `should_admit_msg1`: an active peer's
`current_addr()` matching `(transport_id, remote_addr)`.
`current_addr` is updated and refreshed from inbound encrypted-frame
source addrs (`handlers/encrypted.rs`), which are always numeric
`SocketAddr`-form, so this catches the established peer regardless
of how its `addr_to_link` key was originally inserted. The fast
`addr_to_link` check stays first; the iteration over peers is
bounded by peer count and only runs when the first predicate misses.
Regression coverage in this commit:
- Unit test `test_should_admit_msg1_admits_rekey_when_addr_form_differs`
in `src/node/tests/handshake.rs`. Constructs the failing scenario
in-process: `addr_to_link` populated with hostname-form key, peer's
`current_addr` at the resolved numeric form, query with numeric form.
Without the new predicate this fails immediately.
- New integration topology `rekey-outbound-only` plus matching
docker-compose profile. Same 5-node mesh shape as `rekey-accept-off`
but `inject-config` sets `udp.outbound_only: true` on node-b and
rewrites node-b's peer-c address from the numeric docker IP to the
docker hostname (`node-c:2121`), reproducing the production
hostname-vs-numeric mismatch. The test asserts no sustained
"Dual rekey initiation: we win" log lines on any node (>10 = bug)
and the existing rekey health checks catch the connectivity loss
the loop produces.
- `testing/ci-local.sh` and `.github/workflows/ci.yml` extended to
run the new variant in the local sweep and the GitHub CI integration
matrix alongside `rekey` and `rekey-accept-off`.
Verified locally: full `bash testing/ci-local.sh` sweep passes 29/29
suites (23m 12s) with the new variant green; 1084 unit tests pass.
Single combined commit covering five interlocking pieces of test and
CI work that landed during the v0.3.0-prep cycle.
## fips-gateway robustness
- src/bin/fips-gateway.rs DNS upstream probe converted from a 3-second
hard-fail to a bounded retry loop (5 attempts × 1s timeout, 1s sleep
between attempts; ~10s worst case). Covers the cold-boot race where
the daemon's TUN is up but the DNS responder at [::1]:5354 is still
binding. Each failed attempt logs at INFO. In production the binary's
retry is the live recovery mechanism; with retry it recovers silently
instead of relying on Restart=on-failure (~5s blip + spurious ERROR
per cycle).
- packaging/debian/fips-gateway.service `ExecStartPre` now waits up to
30 seconds for the daemon's `fips0` TUN to appear before exec'ing
the gateway binary. Eliminates the cold-boot race where the gateway
exits with `fips0 interface not found` and recovers via
`Restart=on-failure`, producing a 5-second blip and a spurious error
log per restart cycle.
- testing/docker/entrypoint.sh gateway-mode waits up to 30s for the
daemon's DNS responder to bind [::1]:5354 (probes once per second
with `dig @::1 -p 5354 ... test.fips`) before exec'ing fips-gateway.
Belt-and-suspenders with the binary's own retry: in CI we want
deterministic startup ordering. On timeout, fall through so the
binary's probe reports the definitive error.
## Test infrastructure DNS bind migration to ::1
After session 359's daemon DNS-bind default flipped from `127.0.0.1`
to `::1` (the production fix for ISSUE-2026-0002), the static-test
infrastructure was carrying a stale workaround that overrode the
default back to IPv4 loopback. The fips-gateway integration test
exposed the divergence: the gateway probes its DNS upstream at
`[::1]:5354` (production default) while the daemon was binding
`127.0.0.1:5354` from the template override — IPv6-explicit sockets
do not accept v4-mapped traffic, so the upstream probe exhausted
retries and the gateway exited.
- Drop the explicit `bind_addr: "127.0.0.1"` line from every test
config that emits it: testing/static/configs/node.template.yaml,
testing/chaos/configs/node.template.yaml, the sidecar heredoc in
testing/docker/entrypoint.sh, testing/acl-allowlist/generate-configs.sh
(six per-node blocks), testing/nat/scripts/generate-configs.sh, and
the four tor templates under testing/tor/. Daemon picks up its
production `::1` default.
- Flip the dnsmasq forwarder for `.fips` in testing/docker/Dockerfile
from `127.0.0.1#5354` to `::1#5354` so dnsmasq on the shared test
image continues to reach the daemon. Template and Dockerfile must
move together since most static suites resolve `<npub>.fips` via
the test-image dnsmasq.
## rekey-accept-off integration variant + UDP unit test
- New `rekey-accept-off` topology and docker-compose profile under
testing/static/. 2-node variant where node-b runs with
`udp.accept_connections: false`. Pins the regression class that
ISSUE-2026-0004 fixed (cross-connection winner's rekey msg1 was
being filtered by the accept_connections gate, breaking rekey).
- testing/static/scripts/rekey-test.sh accepts REKEY_TOPOLOGY and
REKEY_ACCEPT_OFF_NODES env vars; its inject-config subcommand
applies the per-node `udp.accept_connections: false` edit, and
the test asserts no sustained "Dual rekey initiation" log lines.
- New UDP variant of `should_admit_msg1` admit-rekey unit test in
src/node/tests/handshake.rs.
## ci-local.sh full integration coverage
- New runner functions and dispatcher entries for `acl-allowlist`,
`nat-cone` / `nat-symmetric` / `nat-lan`, `rekey-accept-off`,
`dns-resolver`, `deb-install`. Each integrates with the existing
summary tracking via `record`.
- New `--with-tor` flag (off by default) gates `tor-socks5-outbound`
and `tor-directory-mode` runners. Tor stays opt-in because both
harnesses depend on the live Tor network and would introduce a
flake source unrelated to the FIPS code.
- New suite arrays (`ACL_SUITES`, `NAT_SUITES`, `DNS_RESOLVER_SUITES`,
`DEB_INSTALL_SUITES`, `TOR_SUITES`) drive both the default sweep
and `--list` output.
- `run_suite` extended to accept the new suite names for `--only`
invocations.
## GitHub CI matrix expansions
- `gateway` matrix entry runs testing/static/scripts/gateway-test.sh
against the existing docker-compose `gateway` profile.
- `rekey-accept-off` matrix entry exercises the new topology with
REKEY_ACCEPT_OFF_NODES=b.
- `deb-install` matrix (debian12 + ubuntu24 + ubuntu26) runs
testing/deb-install/test.sh with privileged systemd containers.
~5-7 min cold cache, ~2 min warm per distro. Self-contained: builds
its own .deb in a Debian 12 cargo-deb builder image; does not
depend on the build job's pre-built artifact.
- `dns-resolver` matrix entry runs the full 13-scenario harness
(per-distro systemd resolver-backend tests + real-fips end-to-end
scenarios) in a single job. Pins the production DNS bind path that
ISSUE-2026-0002 lived in. ~7-12 min warm, ~12-15 min cold.
Verified locally: full `bash testing/ci-local.sh` sweep passes,
including 5/5 deb-install distros and all 13 dns-resolver scenarios.
Tor-inclusive sweep (`--with-tor`) verified in a follow-up run.
- testing/static/scripts/rekey-test.sh: bump REKEY_SETTLE from 5s to
12s so post-rekey ping_all samples are taken after the 10s old-
session drain window has closed (DRAIN_WINDOW_SECS in
handlers/rekey.rs). The 5s sample window straddled the drain,
occasionally catching old-session decrypts mid-cutover and producing
spurious ping failures.
- testing/lib/log_analysis.py: discovery-succeeded match string is
"proof verified, route cached" (the wording in handlers/discovery.rs);
the regex looked for the older "caching route" wording, so the chaos
analysis summary always reported `Succeeded: 0` regardless of how
many successes the raw node logs actually showed.
- testing/scripts/build.sh: drop the second cargo invocation that
passed `--features gateway --bin fips-gateway`. PR #79 removed the
`gateway` cargo feature in favor of platform cfg gates, so this
invocation now fails with `the package 'fips' does not contain this
feature: gateway`. The first cargo build already produces
fips-gateway as a workspace binary on every supported target.
Phase 1 pre-rekey baseline was failing intermittently on CI runners
because 5s wasn't enough for multi-hop discovery (B→D requires
B→C→D). The mesh always converged by Phase 3, confirming this was
purely a timing issue.
The FSP XK rekey handshake has a race condition where the initiator
can cut over (K-bit flip) and send data encrypted with the new session
before msg3 reaches the responder. The responder has no pending session
yet, so K-bit detection fails and packets are dropped.
Defer FSP initiator cutover by 2 seconds after handshake completion
(FSP_CUTOVER_DELAY_MS) to give msg3 time to traverse the mesh. FMP
(IK, 2 messages) is unaffected since the responder completes during
msg1 processing.
Also fix MMP metric corruption after rekey cutover: the new session
starts with counter 0 but MMP state carries highest_counter and
GapTracker.expected_next from the old session, producing false
reorder counts, jitter spikes, and invalid OWD trends. Add
reset_for_rekey() methods that clear counter-dependent state while
preserving RTT estimates.
Additional fixes:
- Remove stale peers_by_index entry on abandon_rekey error path
- Replace redundant peers_by_index inserts with debug assertions
verifying the pre-registration invariant
- Tighten rekey integration test to zero tolerance (was 4 failures)
Phase 3 (post-first-rekey connectivity) can see brief disruptions while
links are mid-cutover. Add a 5s settle time after the rekey wait and
allow up to 4 transient pair failures in Phase 3. Phase 5 remains
strict, ensuring full recovery after the second rekey cycle.
Implement periodic full rekey at both protocol layers using fresh DH
key exchanges. Uses the existing K-bit flag (FLAG_KEY_EPOCH /
FSP_FLAG_K) to coordinate cutover between peers.
FMP layer (IK pattern):
- ActivePeer gains rekey state: pending/previous sessions, K-bit epoch
tracking, drain window, dampening timer
- Handshake state stored on ActivePeer with msg1 sent on existing link
- Encrypted frame handler detects K-bit flips, promotes pending
sessions, falls back to previous session during drain
- Handshake handlers distinguish rekey from new connections using
addr_to_link lookup with identity-based fallback
- Free all session indices (current, rekey, pending, previous) on
peer removal
FSP layer (XK pattern):
- SessionEntry gains parallel rekey fields with XK-specific state
for the 3-message handshake
- Route availability check before FSP rekey initiation
- Encrypted session handler adds K-bit flip detection and dual-session
decrypt fallback
- SessionSetup/Ack/Msg3 handlers extended for rekey paths
Defense-in-depth:
- Consecutive decryption failure detector (threshold=20) triggers
forced peer removal instead of waiting for link-dead timeout
- Identity-based rekey detection as fallback when addr_to_link
doesn't match (e.g., TCP ephemeral ports)
Configuration: RekeyConfig with enabled flag, after_secs (default 120),
and after_messages (default 65536) thresholds.
Logging: info for successful K-bit cutover completions, warn for
failures, debug for intermediate handshake steps, trace for routine
operations (resends, drain cleanup).
Rekey lifecycle:
1. Timer/counter fires -> initiator starts new handshake
2. Old session continues handling traffic during handshake
3. Handshake completes -> initiator cuts over, flips K-bit
4. Responder sees flipped K-bit -> promotes new session
5. Both keep old session for 10s drain window
6. After drain, old session discarded
Integration test: Docker-based multi-phase test exercising both FMP
and FSP rekey with aggressive timers (35s). Verifies connectivity
across all 20 directed pairs survives two consecutive rekey cycles.
Includes rekey topology, docker-compose profile, and CI matrix entry.
Increase ping test convergence wait from 3s to 5s for CI reliability.
Johnathan Corgan
joandJohnathan Corgan