Public bug reported:

Impact:
Acer Chromebook 314 CB314-1H-C3JX (Google Bobba), Intel Celeron N4020, x86_64, 
BIOS Google_Bobba.11297.357.0 (2021-12-24). USB-C charging-related hangs 
require forced recovery. Please backport the existing upstream fix rather than 
treat this as a new upstream issue.

Confirmed evidence:
On unmodified Ubuntu 7.0.0-42-generic (linux 7.0.0-42.42), the saved kernel 
journal records UBSAN array-index-out-of-bounds at 
kernel/locking/qspinlock.h:68, index 8242 for an array of 8192 elements. Oops 
#1 occurs in native_queued_spin_lock_slowpath while cros_ec_cmd_xfer is called 
from cros_usbpd_get_event_and_notify/cros_usbpd_notify_acpi. The first Oops 
reports Not tainted. Oops #2 follows in acpi_ec_submit_query. Actual log 
excerpts are attached.

On this hardware, GOOG0003 (cros-usbpd-notify-acpi) and GOOG0004
(cros_ec_lpcs) are siblings under PNP0C09 (acpi-ec). The driver must not
interpret the ACPI EC parent's drvdata as struct cros_ec_device.

The exact Ubuntu git tags Ubuntu-7.0.0-43.43 and Ubuntu-7.0.0-44.44
still read dev_get_drvdata(dev->parent) before validating the parent's
HID. Disassembly of the installed, package-checksum-matching 7.0.0-43
module confirms this unfixed logic. Version 44 was inspected in source
only, not boot-tested.

Reporter observations and limits:
A local backport of ca75b38b1a3c, built only for 7.0.0-42, resolved the 
observed problem during subsequent normal use. This is not a controlled stress 
test. The running system now uses that out-of-tree module (srcversion 
A6A4CB64AE95650A2A66277); it is not a stock-kernel test.

The reporter experienced another hang during charging on stock
7.0.0-43.43. Its journal ends without a normal system shutdown; the
following boot reports unclean filesystem/journal recovery. No
Oops/UBSAN stack or PD event was saved for that boot. Attributing that
particular hang to this defect remains an inference, although the
affected code is present.

Fix:
Regression identified upstream: db65a06d10b3 (ACPI: EC: Convert the driver to a 
platform one).
Requested fix: ca75b38b1a3c2eabe81d231dae4ea180cf6287cd (platform/chrome: 
cros_usbpd_notify: Only use drvdata when parent is GOOG0004).
https://github.com/torvalds/linux/commit/ca75b38b1a3c2eabe81d231dae4ea180cf6287cd

Ubuntu source evidence:
https://git.launchpad.net/~ubuntu-kernel/ubuntu/+source/linux/+git/resolute/plain/drivers/platform/chrome/cros_usbpd_notify.c?h=Ubuntu-7.0.0-43.43
https://git.launchpad.net/~ubuntu-kernel/ubuntu/+source/linux/+git/resolute/plain/drivers/platform/chrome/cros_usbpd_notify.c?h=Ubuntu-7.0.0-44.44

Test plan / regression considerations:
No deterministic reproducer is claimed and no additional hang was induced for 
this report. For SRU validation, verify the parent-HID guard and NULL fallback 
on this topology, then assess normal charging and suspend/resume on affected 
hardware under a safe test plan. Also validate the GOOG0003-under-GOOG0004 
path. Incorrect probing/defer handling could regress PD notifications or 
charging-state updates. Full Ubuntu-kernel integration and that alternate 
topology have not been tested by this reporter.

** Affects: linux (Ubuntu)
     Importance: Undecided
         Status: New

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2174339

Title:
  [Resolute] cros_usbpd_notify crashes on Google Bobba; please backport
  ca75b38b1a3c

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2174339/+subscriptions


-- 
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to