drivers: sensor: icm45686: fix stream init and early-interrupt handling - #117394
Open
bperseghetti wants to merge 3 commits into
Open
drivers: sensor: icm45686: fix stream init and early-interrupt handling#117394bperseghetti wants to merge 3 commits into
bperseghetti wants to merge 3 commits into
Conversation
icm45686_stream_init passed an uninitialized int_config to icm456xx_set_config_int, whose purpose there is to disable every INT1 source. The unwritten fields are stack garbage, so the call could instead enable arbitrary INT1 sources and assert a data-ready edge before the RTIO stream is armed. The APEX (icm45686.c) and trigger (icm45686_trigger.c) init paths already memset it to INV_IMU_DISABLE first, so do the same on the streaming path. Signed-off-by: Benjamin Perseghetti <bperseghetti@rudislabs.com>
icm45686_event_handler loaded read_cfg from data->stream.iodev_sqe->sqe.iodev->data at function entry, before the guard that checks whether iodev_sqe is NULL. A data-ready edge that arrives before a streaming submission is armed therefore dereferenced a NULL pointer. Handle the no-submission case in its own guard that ignores the spurious interrupt and returns, keep the cancelled submission path separate, and defer the read_cfg load until after both checks. Add a driver test that invokes the handler with no submission armed and confirms the interrupt is ignored. The test compiles the stream translation unit directly with stubbed bus helpers so the guard can be exercised without a full bus and device instance, and enables CONFIG_ICM45686_STREAM through Kconfig so the driver data layout matches the streaming build. Signed-off-by: James Goppert <james.goppert@gmail.com> Signed-off-by: Benjamin Perseghetti <bperseghetti@rudislabs.com>
The event handler drops an interrupt whenever a stream is already in progress or no submission is armed. During a stall these fire on every data-ready edge, and one warning per event floods the log backend and can starve the threads that would clear the stall. Count the dropped events and emit at most one summary per second so the fault stays visible without flooding. A summary that carries the suppressed count is used rather than a plain rate-limited log (LOG_WRN_RATELIMIT) so the operator sees how many edges were dropped in the interval, not just that dropping occurred. The counters and the report deadline live in per-instance driver data so a two-IMU system attributes and rate-limits each sensor independently instead of aggregating them into one shared total. An alternative that removes the busy-path re-entries outright, disabling the DRDY interrupt on entry to ICM45686_STREAM_BUSY and re-arming it from icm45686_stream_submit, was considered and set aside: it changes the pulse-mode interrupt timing for every user of this driver, so it is left for a separate change validated on hardware. Signed-off-by: Benjamin Perseghetti <bperseghetti@rudislabs.com>
zephyrbot
added
area: Sensors
zephyrbot
requested review from
MaureenHelm,
afontaine-invn,
asemjonovs,
avisconti,
gjabouley-invn,
jeppenodgaard,
nashif,
rbuisson-invn,
sriccardi-invn,
teburd,
tristan-google,
ubieda and
yperess
August 26, 2026 01:28
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Three fixes to the ICM45686 streaming path:
Zero int_config before disabling interrupts in stream init. The struct was passed uninitialized to the call whose purpose is to disable every INT1 source, so stack garbage could instead enable arbitrary sources and assert a data-ready edge before the RTIO stream was armed. The APEX and trigger init paths already memset it first.
Ignore interrupts that arrive before a streaming submission is armed. The event handler loaded the read configuration from the submission pointer before the NULL guard, so an early data-ready edge dereferenced a NULL pointer. A driver test invokes the handler with no submission armed and confirms the interrupt is ignored. The test compiles the stream translation unit directly with stubbed bus helpers so the guard can be exercised without a bus and device instance.
Rate-limit the ignore-path logs. During a stall both ignore paths fire on every data-ready edge, and one warning per event floods the log backend and can starve the threads that would clear the stall. Count the dropped events instead and emit at most one summary per second carrying the suppressed counts, kept per instance so a multi-IMU system attributes each sensor separately.
Validated on NXP MCXN947 hardware with the IMU streaming at 800 Hz.