Re: [PATCH 2/2] sysfs: fix race condition around sd->s_dentry, take#2

[Tejun Heo <htejun@gmail.com>]

Tejun Heo wrote:
> Tejun Heo wrote:
>> Greg KH wrote:
>>> On Mon, May 21, 2007 at 10:31:01AM +0530, Maneesh Soni wrote:
>>>> On Thu, May 17, 2007 at 08:16:10PM +0200, Tejun Heo wrote:
>>>>> Allowing attribute and symlink dentries to be reclaimed means
>>>>> sd->s_dentry can change dynamically.  However, updates to the field
>>>>> are unsynchronized leading to race conditions.  This patch adds
>>>>> sysfs_lock and use it to synchronize updates to sd->s_dentry.
>>>>>
>>>>> Due to the locking around ->d_iput, the check in sysfs_drop_dentry()
>>>>> is complex.  sysfs_lock only protect sd->s_dentry pointer itself.  The
>>>>> validity of the dentry is protected by dcache_lock, so whether dentry
>>>>> is alive or not can only be tested while holding both locks.
>>>>>
>>>>> This is minimal backport of sysfs_drop_dentry() rewrite in devel
>>>>> branch.
>>>>>
>>>>> DONT APPLY JUST YET
>>>> Looks ok to me.. I have tested it it but unfortunately I couldn't
>>>> recreate the race without the patch also. It would be helpful if
>>>> people actually seeing the race, provide the test results.
>>>>
>>>> Greg, please merge this one once we have some test results.
>>> Can someone just resend it after those test results are in, with the
>>> proper signed-off-by, so I know it's safe to apply?
>> I'll resend with S-O-B after someone verifies it fixes the problem.
> 
> Have been trying to reproduce the problem on the latest -rc but haven't
> succeeded yet.  It seems I lost my magic recipe for these races.  Did
> anyone succeed?

Alright, overnight test finally hit the race condition.  It takes me
much longer now but the race condition is definitely still there.  I'll
verify each proposed patch and send it for 2.6.22 inclusion and if
they're accepted to -stable.

Thanks.

-- 
tejun