RE: 2.6.18.4: flush_workqueue calls mutex_lock in interrupt environment

["Chen, Kenneth W" <kenneth.w.chen@intel.com>]

Andrew Morton wrote on Thursday, December 14, 2006 5:20 PM
> it's hard to disagree.
> 
> Begin forwarded message:
> > On Wed, 2006-12-13 at 08:25 +0100, xb wrote:
> > > Hi all,
> > > 
> > > Running some IO stress tests on a 8*ways IA64 platform, we got:
> > >      BUG: warning at kernel/mutex.c:132/__mutex_lock_common()  message
> > > followed by:
> > >      Unable to handle kernel paging request at virtual address
> > > 0000000000200200
> > > oops corresponding to anon_vma_unlink() calling list_del() on a
> > > poisonned list.
> > > 
> > > Having a look to the stack, we see that flush_workqueue() calls
> > > mutex_lock() with softirqs disabled.
> > 
> > something is wrong here... flush_workqueue() is a sleeping function and
> > is not allowed to be called in such a context!
> 
> It seems utterly insane to have aio_complete() flush a workqueue. That
> function has to be called from a number of different environments,
> including non-sleep tolerant environments.
> 
> For instance it means that directIO on NFS will now cause the rpciod
> workqueues to call flush_workqueue(aio_wq), thus slowing down all RPC
> activity.

The bug appears to be somewhere else, somehow the ref count on ioctx is
all messed up.

In aio_complete, __put_ioctx() should not be invoked because ref count
on ioctx is supposedly more than 2, aio_complete decrement it once and
should return without invoking the free function.

The real freeing ioctx should be coming from exit_aio() or io_destroy(),
in which case both wait until no further pending AIO request via
wait_for_all_aios().

- Ken