| Age | Commit message (Collapse) | Author |
|---|
|
Correct harness docs the code refutes and drop dated prompt patterns
|
|
ci: one Claude review per PR at a time, effort high, Opus 5
|
|
The doc half landed in this branch. The test half it asked for already
exists: NoBoardsExitRespectsFreshness and TheNoBoardsCallSiteIsWired in
test/hil/test/test_hil_report.py (375e20090) cover the fresh and
--accumulate no-boards exits through mark_report_no_boards.
|
|
The WCH lock header path sat across a code-span line break and rendered
with a space in it. The pool-check reporting list had dropped the
scan-mode disclosure the same file requires. The pr3840 handoff still
described the no-boards bullet as unfixed; it now records the doc half as
landed and keeps only the missing test.
|
|
|
|
A new push now cancels the stale in-flight review instead of stacking a
second run on the same PR. Effort max on every synchronize is the only
place the routine review needs the top setting; high is the documented
default. Drop the transcript dump that was marked temporary.
|
|
device/dfu: fix transfer buffer overwrite issue
|
|
device/cdc: avoid flushing tx buffer on connection
|
|
prompt patterns
Prompt audit of the harness files. hil/SKILL.md's banner rules predate
3963a1b70, which keeps the finished boards' rows on a pool timeout or a
worker crash and says so in the banner; the no-boards bullet is the drift
docs/superpowers/followup/pr3840-skill-md-no-boards-drift.md records. The
unit-suite paragraph counted five suites (there are seven; measured 64 s,
not 48) and the hil_lock.py line citations had already moved. The rest
replaces numeric commentary caps, a history clause, a branch pointer, an
install path and one accreted paragraph with the rule each stood in for.
|
|
hcd/dwc2: fix FIFO allocation and periodic transfer
|
|
rp2040: do not require PICO_DEFAULT_PIO_USB_DP_PIN to be set for pio_usb
|
|
|
|
HFNUM retains only 16384 host-frame positions, while valid periodic endpoint intervals can be longer. Resubmission after the counter wraps can therefore alias the elapsed time and skip the next established service phase.
Cap the host-selected interval to one HFNUM cycle using the root-port frame unit. USB permits a shorter host-provided period, and the bounded interval keeps phase calculation unambiguous for native and split endpoints.
|
|
Periodic DMA channels use their natural service-boundary halt instead of a software CHDIS request. Keep the endpoint busy after an abort so a replacement transfer cannot reuse its state or buffer while the channel remains active.
When HCINT.HALTED arrives, release the channel without reporting completion for the aborted transfer. If endpoint closure is also pending, release the endpoint from the same halt path.
|
|
agent/fix-dwc2-host-fifo-allocation
|
|
A disconnect can close endpoint zero after an enumeration control stage completes but before USBH submits the next stage or request. HCD submission then legitimately returns false; treating that result as an invariant violation asserts during rapid replug and can leave enumeration unfinished.
Complete an in-progress control request as failed when its DATA or status stage cannot be submitted. Propagate submission failures from every asynchronous enumeration continuation and finish enumeration through the normal failure cleanup path. This keeps controller teardown races out of assertions without fabricating a successful transfer.
Validated by interrupting enumeration during rapid STM32U5A5 replug tests with DWC2 DMA and slave modes.
|
|
A root-port disconnect invalidates every active transfer. Retire channel and FIFO interrupt sources plus host-channel state in the disconnect ISR using the Linux DWC2 cleanup model instead of reinitializing the core and PHY, which can sleep on STM32 HS PHYs.
Flush posted slave requests, request halts for enabled channels, clear channel interrupt and software ownership, and keep endpoint records closing until USBH processes the remove event. Reject transfer submissions to closing endpoints, preserve fast-replug notification, and re-enable the global host-channel interrupt when a new channel is initialized.
|
|
agent/fix-dwc2-host-fifo-allocation
|
|
In slave mode, channel_xfer_start() enabled an OUT channel but left every FIFO write to a later PTXFEMP interrupt. DWC2 creates the request-queue entry only when the packet's final FIFO word is written, so unrelated interrupt work could consume the selected service frame before the transfer was actually queued.
Factor FIFO writes into a capacity-checked helper and write the initial packet while the channel-enable operation is still protected from DWC2 interrupts. Keep FIFO-empty interrupts only for data that does not fit immediately. The protected section never waits for FIFO or request-queue space.
When initial periodic OUT submission is too close to the frame boundary, release the unused channel and defer the still-pending endpoint to the next SOF. Internal retries bypass this initial boundary guard.
A hardware trace showed HCCHAR enabled for frame 0x0378 while the packet's final FIFO word was delayed until frame 0x03ae. The complete five-commit fix set passed 600 seconds in every O0/O2 and slave/DMA mode.
Signed-off-by: HiFiPHile <[email protected]>
|
|
Periodic IN and DMA-backed transfers selected ODDFRM before waiting for request-queue space. A DWC2 interrupt could also run between reading HFNUM and writing HCCHAR.CHENA, allowing the selected frame to pass while the transfer still appeared active.
Wait for request-queue capacity with controller interrupts enabled, then mask only GAHBCFG.GINT while sampling HFNUM and enabling a new periodic channel. Record the periodic phase from that same HFNUM sample so a boundary after channel enable cannot shift later interval calculations. The bounded critical section contains no queue wait, callback, disable, or allocation loop. Retries that already selected their frame bypass the new selection step.
Also clear a retained HCCHAR.CHDIS before every channel enable. A halted channel can otherwise be re-enabled as CHENA|CHDIS and wait for a terminal interrupt that never arrives.
Hardware traces captured periodic IN selections at frames 0x3303 and 0x3266 but activation only after 0x330c and 0x3273, respectively.
|
|
A frame-overrun interrupt means the selected periodic service interval has already been missed. Retrying an isochronous transfer after that point cannot deliver the original packet and can leave the class waiting indefinitely for a terminal result.
Enable frame-overrun interrupts for slave periodic channels. Complete isochronous IN and OUT overruns as XFER_RESULT_FAILED in both slave and DMA modes, accounting for bytes already written on OUT. Preserve the existing retry behavior for non-isochronous DMA transfers.
This reports the missed packet honestly through the normal HCD completion path: no fabricated success and no class-level abort workaround.
|
|
DWC2 buffer/external DMA mode automatically halts a periodic channel at its next service boundary. Programming HCCHAR.CHDIS|CHENA for a non-split periodic channel is explicitly disallowed by the controller programming guide, yet channel_disable() skipped that write only for split periodic transfers.
Return without programming channel disable for every periodic DMA channel. Non-periodic DMA and slave-mode channels retain the existing explicit-disable path.
The previous path reproduced after 420 seconds in O2/DMA with a closing capture transfer left INVALID while HCCHAR retained CHENA|CHDIS and HCINT was clear.
|
|
Slave-mode channel handlers process one interrupt cause per pass, but the dispatcher acknowledged every HCINT bit before invoking them. When ChHltd arrived together with another cause, the handler consumed the other cause and the halt was lost. A subsequent disable could then leave CHENA|CHDIS asserted with HCINT and HAINT clear, so the submitted periodic transfer never completed.
When a slave channel reports ChHltd with another cause, acknowledge only the non-halt causes and leave ChHltd pending for the next channel-IRQ pass. DMA handlers retain their existing combined-cause behavior.
The uninstrumented negative capture reproduced the lost terminal state with HCCHAR=0xe044881c, HCTSIZ=0x0008001c, HCINT=0, and XFER_RESULT_INVALID.
|
|
Protect periodic deferral cancellation from the SOF interrupt. Re-enable the host interrupt before disabling an active channel because slave-mode channel disable may wait for request-queue space.
|
|
agent/fix-dwc2-host-fifo-allocation
|
|
Popping an IN transfer-completion entry from GRXSTSP asserts HCINT.XferCompl. Drain the receive FIFO first, then read the live masked global status so the newly asserted channel completion is handled without waiting for another interrupt.
|
|
|
|
Anchor resubmitted periodic transfers to the endpoint service interval and defer early submissions through SOF. This prevents callback latency from shifting the cadence or causing intervals to be skipped, while keeping pending transfers abortable.
Signed-off-by: HiFiPHile <[email protected]>
|
|
Fix DWC2 DMA data toggle mismatch in IN-transfers
|
|
bound endpoint number in tu_bind_driver_to_ep_itf
|
|
tu_bind_driver_to_ep_itf() now returns false when ep_num >= CFG_TUH_ENDPOINT_MAX, but the host caller ignores that result and continues enumeration. Configurations such as host/bare_api set the limit to 8, while valid USB devices may use endpoints 8–15. A recognized class can therefore continue and later index ep_status[epnum] or ep2drv[epnum] out of bounds. Wrap this call in TU_ASSERT(...), as the device path already does, so parsing fails immediately.
Signed-off-by: HiFiPHile <[email protected]>
|
|
|
|
ucpd: Add "attached" state callback for CC state changes
|
|
|
|
disconnected->connected or connected->disconnected
|
|
|
|
rp2350: ETM trace board pico2_etm_trace
|
|
|
|
Closes the sys-clock question. rp2350: the dynamic 156 MHz switch is
removed - 150 MHz soak-tested clean (and a runtime switch truncates ETM
capture on the trace carrier). rp2040 keeps its existing 120 MHz: soak
sweeps show 120 = 8/8, stock 125 = 0/3 (bulk-OUT collapses, device NAKs
~600:1 with zero CRC errors on the wire), and 132 = 2/10 flaky despite an
exact 12n/375k divider - no divider criterion predicts rp2040 PIO-USB
health (the RX state machine samples at raw sysclk), so only
soak-validated clocks ship.
|
|
Board files for the trace carrier (console GP12/13, LED GP10, I2C GP8/9,
PIO-USB host on GP20, all retargeted in board.cmake so the SDK defaults
cannot mux a trace pin), compile-time trace pin-conflict checks, the
measured DBGPAUSE rationale, Ozone project, and skill/docs updates.
Trace validated at the stock 150 MHz (75 MHz TRACECLK, +1 ns sampling):
zero overflow through a 15 s throughput soak; V2 probe ceiling 120 MHz.
|
|
hil: drop the sysfs blindness subsystem and derive the recovery reserve
|
|
The modules were 21% comment, much of it review-cycle argument rather than
guidance -- _kill_kids stated 'descendant by construction, no argv check needed'
twice, eight lines apart. Deleting such comments outright makes maintenance
worse: the next reader simplifies the thing the comment was defending. So the
history moves to the 2026-07-30 fleet-wedge design doc, which gains a trim
addendum recording what was removed, what was deliberately kept, and the rule
that decided each -- the CI ceiling bounds how long a run burns, and does nothing
about state that outlives it.
One comment was not merely long but WRONG: the report wipe carried 'The unlink is
DEFERRED to inside the pool try/except below', which is the opposite of what the
code does -- it sits before Manager() with its own comment explaining why. That
is the failure mode this pass is about, so it is deleted rather than reworded.
Kept everywhere: citations that refute a plausible wrong reading. That
usb_lock_device_interruptible is why the readers are killable, that usblp_mutex
is driver-global, that rawmidi honours O_NONBLOCK where usblp does not.
Two follow-ups are retired with them: pr3803-hil-blindness-reporting.md (there
is no blindness to report any more) and pr3803-usbtest-recovery-reserve.md (the
reserve is derived now). Kept: pr3803-flasher-recover.md, which PR #3832
implements, plus pr3803-pci-rebind-stranding.md and pr3803-hil-iar-rerun-spec.md,
both independent of this work.
|
|
Two layers whose cost was a contract to reason about rather than an outcome.
SYSFS_UNKNOWN was a three-valued return five files had to keep apart, and
misreading unknown as absence was silent: a healthy board reported as a firmware
regression. What it guarded is real -- `serial` is served by usb_string_attr,
which takes usb_lock_device_interruptible (v6.12.96 sysfs.c:141-143), the same
lock a wedged usbfs ioctl holds -- so the BOUND stays, on every caller by
default. usb_scan reads `serial` on every device matching the VID, and hil_lock's
controller_of does that on essentially every board, so one wedged DUT would
otherwise stall every worker, not one. What goes is the third value.
read_sysfs now returns str or None, and the question the third value existed to
answer is asked directly instead, by two predicates that say which question they
answer: sysfs_stranded() is process-wide and sticky, for hil_pool_check's footer
("could anything here be the tool losing sight of healthy hardware?"), and
path_stranded(path) is per-device, which is what usbtest needs to tell a DUT
whose `serial` is held under device_lock from one that genuinely left the bus --
that difference decides whether it performs driver-registry writes that take the
uninterruptible device_lock.
Gone: _SysfsUnknown, SYSFS_UNKNOWN, sysfs_blind, sysfs_blind_note,
note_sysfs_strand, the cross-process blindness publishing and its report banner,
usb_scan's (list, bool) return, usbtest's inconclusive abort, _blind_note's slot
in the result tuple, and bounded_open, whose last caller went in the previous
commit.
The strand memo is rewritten around the one invariant that makes it safe to
reuse: it is keyed by the path's kernfs inode, captured BEFORE the read. A
busport does not change when a board returns to the same physical port, so a
path-only blacklist outlives the wedge and hil_pool_check's own recovery flow --
reset, reflash, wait_device polling that busport -- would never look at the
board again. A re-enumeration destroys the kernfs node and makes a new one, so a
changed inode is the all-clear. Two ceilings bound different things: per path
(_PATH_STRAND_MAX) for a board that flaps while still wedged, and per process
(_STRAND_MAX) as a backstop against RLIMIT_NOFILE, counted per PATH rather than
per reader because hil_pool_check runs four poll threads over one bus.
A board the pool guard never reached is now reported as run-aborted rather than
pool-timed-out, and outranks a stale board-locked cell for the same reason the
pool-timeout cell does.
Both predicates answer conservatively where they are consulted before something
irreversible. path_stranded() covers the paths read_sysfs answered None for
WITHOUT reading -- past _STRAND_MAX it declines to start another reader, and
vouching for a path nobody looked at hands usbtest's fail-CLOSED guard a
fabricated all-clear, running remove_id/unbind against a wedged device. usbtest's
startup lookup carries the same caveat hil_test's absent arm already did, because
its stderr is relayed verbatim into the report cell.
strand_note() survives the removal for the same reason master had it: every
caller that can say "not found" needs the same sentence, and the one site left to
re-invent it got missed -- a wedged-but-enumerated printer was reported as an
enumeration failure, sending a maintainer after firmware.
The two predicates are not interchangeable, and usbtest needs both. Its per-case
verdict is per-DUT -- a peer that stranded at case 2 must not make our board
report wedged at case 29 -- but the finally block's cleanup is process-wide:
remove_id plus an unbind of EVERY interface under the driver, including that
peer's, each taking the uninterruptible device_lock. So the verdict uses
path_stranded() and the global cleanup stays gated on sysfs_stranded().
USBTEST_RECOVERY_BUDGET was an independent 250s that could not actually contain
the ladder it reserved for, which is why usbtest.py carried a _time_left() gate
re-deciding before every step -- with a bare '- 35' for costs paid downstream
that nobody could re-derive. Between them the two produced a recovery that
skipped its own steps for most real hangs.
The reserve is now derived from the bounds usbtest itself declares, per flasher
and per target: a probe reset, a reflash, and the Rescue-DP POR plus retry a
wedged RP DAP needs, plus the settles and hil_util.REAP_GRACE for each bounded
step. The Rescue-DP legs are openocd-only and gated on the RP target cfg, and
a stub reset is screened out, so the reserve tracks each board's real ladder
instead of one fleet number: 390s for the two RP boards -- whose ladder the old
250 could not contain, which is exactly why the gates skipped their steps --
190s for the other seventeen probe-reset boards, and 150s for esptool and
lm4flash, whose reset is a no-op. Changing a bound in usbtest moves the
reserve with it, and a unit test asserts it covers the ladder.
With the room actually reserved, the child runs the ladder straight through:
recovery_steps, _time_left, the three per-step gates and the parsed-but-unused
--outer-timeout are gone. What stays is what decides outcomes -- the convoy_safe
gate, reset-before-reflash, the no_op screen so a stub that resets nothing is not
claimed, and wedged_pids() as the arbiter, because a clean flash only proves the
probe wrote the MCU.
hil_util.py 616 -> 514 lines.
|
|
test_device_printer_to_cdc opened /dev/usb/lp* on the worker itself and let
hil_util.bounded_open abandon a thread when the open blocked. usblp allows one
opener -- usblp_open() returns -EBUSY while usblp->used (v6.12.96 usblp.c) --
so the abandoned thread's fd poisoned the node for every later test that worker
ran. The read half already avoided this by forking; the write half now does too,
via the same run_alongside, and a killed child takes its fd with it.
This removes the only production caller of bounded_open.
|
|
support (#3853)
Promote SEGGER RTT from an inline debugging technique to a standalone skill
backed by one stdlib-only implementation in tools/rtt.py: a CLI and importable
module for console/capture over J-Link (RTTTelnetPort) and OpenOCD (rtt server)
probes, with probe selection by serial or VID:PID, control-block address via
--elf or --addr, bidirectional console, post-mortem ring dump, and
--reset-before-attach for boot-time capture. The HIL harness reads a board's
console over RTT when its probe has no VCOM ("logger": "rtt" plus a LOGGER=rtt
variant define), covering device_info, pool-check aliveness, and CI wiring.
Validated on 22 boards across both backends; 26 unit tests run in pre-commit.
|
|
validate workflow: loop validate -> fix cycles until green
Turn the single-pass validate gate into a loop: run unit + builds +
size + PVS + claude/codex reviews in parallel; on a red verdict one
fix agent repairs the gate-failing evidence (CONFIRMED findings,
codex P0/P1, failed stages - PLAUSIBLE/quality stay report-only),
commits, and the affected stages re-run, up to maxCycles (default 5).
Hardened per review: fix-commit paths verified from git rather than
self-report, restartRequired when a fix edits the workflow itself,
per-stage evidence budgeting so the fixer prompt JSON never truncates
mid-document, dirty-tree and moving-base-ref guards, dead stage
agents retried instead of ending the loop, and only pure-docs fixes
skip a full stage re-run.
|
|
pr-review-validator: done waits for every auto-reviewer to settle
|
|
head SHA
A cycle running before the bots posted saw zero findings and reported done;
with a fast-green CI the babysit loop could exit unreviewed. done now needs
every reviewer settled for the current head: Copilot's verdict review
(commit_id), Codex's verdict comment (Reviewed-commit line), its thumbs-up
reaction on the PR body, or the named claude-review check run — with quota/
error notices and the reaction freshness-gated on push time (check-suite
creation, not committer date) and every lookup paginated. pr-babysit re-arms
with backoff on a pending reviewer instead of exiting unactionable, skipping
the pointless final-cycle wait.
|
|
pr-babysit: never edit HIL rig configs without user approval
|
|
The workflow's fix lane once skipped two host tests in test/hil/tinyusb.json
to green a check whose root cause was a failing fixture drive (reverted in
4b11d59a4). Rig rosters describe physical hardware: papering over a fixture
fault hides it from the user who has to swap the part. Now fixAndVerify
strips test/hil/*.json from every fix scope (a group left with no other
files is withheld and logged), the code-writer prompt carries the
constraint, and ok=false keeps such cycles from pushing. HIL stays red
when the fix is a hardware swap - that red is the signal.
|
