Re: [PATCH 1/5] extcon: max14577: cancel wq_detcable on driver unbind

[Yaohan Li]

Hello Tristan,

We independently reproduced the max14577 delayed-work use-after-free
and tested your [PATCH 1/5] ("extcon: max14577: cancel wq_detcable
on driver unbind") on a clean Linux v7.4-rc1 tree (6628e7d4c6b2).

Test setup:
  - QEMU/KVM with KASAN enabled
  - Self-registered I2C adapter exercising the real max14577
    MFD and extcon probe/remove paths
  - wq_detcable scheduled with the driver's 17-second delay
  - Device unregistered before the delayed work would run

Results:
  - Unpatched v7.4-rc1: 80 slab-use-after-free reports,
    starting when the delayed-work timer expires
  - Your [1/5] applied to the clean tree: 0 KASAN reports,
    with the test running more than four seconds beyond
    the original expiry point
  - Control run without early unbind: 0 KASAN reports

This test applied your posted patch rather than only our own
equivalent devm_delayed_work_autocancel() change. The result
supports the fix for the tested driver-unbind sequence; the
synthetic I2C setup has not been checked against physical
hardware.

Tested-by: Yaohan Li <shunsheng1218@xxxxxxxxx>

We can provide the complete console logs, configuration, and
reproduction details privately if useful. The investigation
was AI-assisted, with source analysis and test results
reviewed by the human tester.

Best regards,
Yaohan Li