[Devel] Re: [PATCH] remove BUG() in possible but rare condition

[Michal Hocko]

On Wed 11-04-12 14:12:44, Andrew Morton wrote:
> On Wed, 11 Apr 2012 17:51:57 -0300
> Glauber Costa <glommer at parallels.com> wrote:
> 
> > On 04/11/2012 05:26 PM, Andrew Morton wrote:
> > >>
> > >> >    failed:
> > >> >  -	BUG();
> > >> >    	unlock_page(page);
> > >> >    	page_cache_release(page);
> > >> >    	return NULL;
> > > Cute.
> > >
> > > AFAICT what happened was that in my April 2002 rewrite of this code I
> > > put a non-fatal buffer_error() warning in that case to tell us that
> > > something bad happened.
> > >
> > > Years later we removed the temporary buffer_error() and mistakenly
> > > replaced that warning with a BUG().  Only it*can*  happen.
> > >
> > > We can remove the BUG() and fix up callers, or we can pass retry=1 into
> > > alloc_page_buffers(), so grow_dev_page() "cannot fail".  Immortal
> > > functions are a silly fiction, so we should remove the BUG() and fix up
> > > callers.
> > >
> > Any particular caller you are concerned with ?
> 
> Didn't someone see a buggy caller in btrfs?

No I missed that __getblk (__getblk_slow) returns NULL only if
grow_buffers < 0 while it returns 0 for the allocation failure.

Sorry for confusion.
-- 
Michal Hocko
SUSE Labs
SUSE LINUX s.r.o.
Lihovarska 1060/12
190 00 Praha 9    
Czech Republic