Re: xfs vs. lockdep

[Eric Sandeen <sandeen@sandeen.net>]

Timothy Shimmin wrote:
>> Hence by the time we have the lock here in xfs_ireclaim we have 
>> guaranteed
>> that there are no other outstanding references and no new references
>> can occur. Therefore it should be safe to drop the lock before destroying
>> it.
>>
> Yeah, there really seems like something would be wrong if you can't
> unlock it before destroying it.

I thought so too; if anybody else might catch it post-unlock pre-free, they're 
in for a big surprise anyway :)

> I would have thought you'd need to guarantee that you are the only
> one with access to it before destroying it otherwise there'd be problems :)
> (Which as you say we do)

right.

> This one rings a bell. I seem to recall multiple places where we destroy
> without releasing the lock first.
> And I vaguely remember Nathan mentioning that this was causing grief
> for lockdep:)


Ok, cool.  Want a formal patch or you guys want to just free it up...

         /*
          * Free all memory associated with the inode.
          */
+       xfs_iunlock(ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
         xfs_idestroy(ip);

Thanks,

-Eric