Re: workqueue panic in 3.4 kernel

[Lei Wen <adrian.wenl@gmail.com>]

Tejun,

On Tue, Mar 12, 2013 at 1:12 PM, Tejun Heo <tj@kernel.org> wrote:
> Hello,
>
> On Tue, Mar 12, 2013 at 01:08:15PM +0800, Lei Wen wrote:
>> diff --git a/include/linux/workqueue.h b/include/linux/workqueue.h
>> index 8afab27..425d5a2 100644
>> --- a/include/linux/workqueue.h
>> +++ b/include/linux/workqueue.h
>> @@ -189,12 +189,16 @@ static inline unsigned int work_static(struct
>> work_struct *work) { return 0; }
>>   * NOTE! No point in using "atomic_long_set()": using a direct
>>   * assignment of the work data initializer allows the compiler
>>   * to generate better code.
>> + *
>> + * We take the assumption that work should not be inited if it already
>> + * hold the pending bit, or bug would be reported.
>>   */
>>  #ifdef CONFIG_LOCKDEP
>>  #define __INIT_WORK(_work, _func, _onstack)                            \
>>         do {                                                            \
>>                 static struct lock_class_key __key;                     \
>>                                                                         \
>> +               BUG_ON(work_pending(_work));                            \
>
> You're initializing random piece of memory which may contain any
> garbage and triggering BUG if some bit is set on it. No, you can't do
> that. debugobj is the right tool for debugging object lifetime issues
> and is already supported.

The debugobj is not helping on this issue, I have enabled both
CONFIG_DEBUG_OBJECTS_WORK and CONFIG_DEBUG_OBJECTS.
But find they didn't report any issue at all.

And I am not init random memory, original issue is call init multi-times
for one structure and that piece of memory already been allocated.
And __INIT_WORK shouldn't call over random memory, right?

All this patch is adding a check here.

Thanks,
Lei