List corruption and crash with kernel 3.1

Quoted 1 line · Andrew Morton

erk, help, who do I blame for this?

[...]

Quoted 3 lines

WARNING: at lib/list_debug.c:47 __list_del_entry+0x8d/0x98()
Hardware name: Latitude E4200

list_del corruption, ffff880075e147b0->next is LIST_POISON1 (dead000000100100)

[...]

Quoted 8 lines

 [<ffffffff81243861>] __list_del_entry+0x8d/0x98
 [<ffffffff8124387a>] list_del+0xe/0x2d
 [<ffffffff813a7e55>] led_trigger_unregister+0x29/0x9c
 [<ffffffff813a7ee1>] led_trigger_unregister_simple+0x19/0x26
 [<ffffffff813828e2>] power_supply_remove_triggers+0x21/0x8f
 [<ffffffff81381d42>] power_supply_unregister+0x1f/0x2c
 [<ffffffff812a7d1f>] sysfs_remove_battery+0x3c/0x54
 [<ffffffff812a8c4d>] acpi_battery_notify+0x46/0xaa

As far as I know (and can see from the changelogs), the LED trigger registration code hasn't changed in a long time. led_trigger_unregister_simple() and led_trigger_unregister() are relatively simple functions and looking at the latter, the list_del causing this is fairly clear.

What puzzles me is that led_trigger_unregister_simple() wouldn't call led_trigger_unregister() twice for the same trigger as it frees the memory.

However, if that function were called multiple times in parallel with the same trigger, I can imagine it racing.

My thought is therefore is something in the power_supply* code calling this function multiple times and a race?

Resend, add proper Cc's.

Quoted 5 lines · Richard Purdie

Quoted 1 line · Andrew Morton

erk, help, who do I blame for this?

[...]

Quoted 1 line

Quoted 1 line

WARNING: at lib/list_debug.c:47 __list_del_entry+0x8d/0x98()

I wonder if that patch helps:

https://lkml.org/lkml/2011/7/12/242

The traces are pretty similar.