Re: Kernel oops with bluetooth usb dongle

[Marcel Holtmann <marcel@holtmann.org>]

Hi Thomas,

>>> Can you please enable CONFIG_SLUB_DEBUG=y and CONFIG_SLUB_DEBUG_ON=y
>>> and give it another try?
>>>
>>> If we can not catch it that way, I'll whip up a patch which points  
>>> us
>>> to the code which added the offending timer.
>>>
>> Hi,
>>
>> Note: I switched to 2.6.25-rc2. The only new thing I see is this  
>> message:
>>
>> hci_cmd_task: hci0 command tx timeout
>>
>> This comes from net/bluetooth/hci_core.c, line 1547
>>
>> There is indeed a timeout message in the log (at the end of this
>> email). I tried to boot with slub_debug but did not get anything
>> more. slabinfo -v does not report anything either.
>>
>> Crash log:
>>
>> hci_cmd_task: hci0 command tx timeout
>> BUG: unable to handle kernel paging request at 6b6b6b6b
>
> We got some more info ---------------------------^^^^^^^^
> #define POISON_FREE      0x6b    /* for use-after-free poisoning */
>
> So the timer is in an allocated data structure, which is
> freed without having removed the timer first.
>
>> Sorry for the meager yield.
>
> Hey, we know already more :)
>
> Marcel, any idea on this one ?

I don't really have any idea. Nothing has been changed in this area  
for a couple of years. The command TX timeout is the timeout that  
indicates a missing answer to a command sent down to the Bluetooth chip.

However this involves some atomic and tasklet stuff. Did we have some  
changes that I missed and might now render this usage as broken.

Regards

Marcel