Re: [PATCH 1/2] sunrpc: add missing spin_unlock

[Peter Zijlstra <a.p.zijlstra@chello.nl>]

On Sun, 2006-11-05 at 11:45 -0800, Andrew Morton wrote:
> On Sun, 05 Nov 2006 17:35:16 +0100
> Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:
> 
> > On Sun, 2006-10-29 at 22:37 +0900, Akinobu Mita wrote:
> > > auth_domain_put() forgot to unlock acquired spinlock.
> > > 
> > > Cc: Olaf Kirch <okir@monad.swb.de>
> > > Cc: Andy Adamson <andros@citi.umich.edu>
> > > Cc: J. Bruce Fields <bfields@citi.umich.edu>
> > > Cc: Trond Myklebust <Trond.Myklebust@netapp.com>
> > > Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
> > 
> > Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
> > 
> > I just found this too while trying to get .19-rc4-git up and running on
> > a machine here - took me a few hours.
> > 
> > It made my kernel decidedly unhappy :-(
> > 
> > Andrew, could you push this and:
> >   http://lkml.org/lkml/2006/11/3/109
> > into .19 still? - those patches are needed to make todays git happy on
> > my machine.
> 
> OK.

Thanks!

> I wonder if this will fix http://bugzilla.kernel.org/show_bug.cgi?id=7457

The scheduling while atomic part looks familiar, the rest not so.

Worth giving it a shot though...

On my machine it was the keventd workqueue that got messed up.
I have some patches that:
 - add debug_show_held_locks(current) to might_sleep() and schedule()
 - check in_atomic() and lockdep_depth after each workqueue function
   and print the last function executed
 - name some 'old_style_spin_init' locks

I'll post those patches after a cleanup...