Re: [RFC] Badness in __mutex_unlock_slowpath with XFS stress tests

[Adrian Bunk <bunk@xxxxxxxxx>]

On Fri, Mar 10, 2006 at 11:50:20AM +1100, Nathan Scott wrote:
> On Fri, Mar 10, 2006 at 10:14:22AM +1100, Nathan Scott wrote:
> > On Thu, Mar 09, 2006 at 10:42:19PM +0000, Christoph Hellwig wrote:
> > > On Fri, Mar 10, 2006 at 09:30:42AM +1100, Nathan Scott wrote:
> > > > Not for reads AFAICT - __generic_file_aio_read + own-locking
> > > > should always have released i_mutex at the end of the direct
> > > > read - are you thinking of writes or have I missed something?
> > > 
> > > if an error occurs before a_ops->direct_IO is called 
> > > __generic_file_aio_read
> > > will return with i_mutex still locked.  Note that checking for negative
> > > return values is not enough as __blockdev_direct_IO can return errors
> > > aswell.
> > 
> > *groan* - right you are.  Another option may be to have the
> > generic dio+own-locking case reacquire i_mutex if it drops
> > it, before returning... thoughts?  Seems alot less invasive
> > than the filemap.c code dup'ing thing.
> 
> Something like this (works OK for me)...

Is this 2.6.16 material?

> cheers.
> Nathan
>...

cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed