Re: upcoming kerneloops.org item: get_page_from_freelist

[Andrew Morton <akpm@linux-foundation.org>]

On Wed, 24 Jun 2009 12:46:02 -0700 (PDT)
Linus Torvalds <torvalds@linux-foundation.org> wrote:

> 
> 
> On Wed, 24 Jun 2009, Andrew Morton wrote:
> 
> > On Wed, 24 Jun 2009 12:16:20 -0700 (PDT)
> > Linus Torvalds <torvalds@linux-foundation.org> wrote:
> > 
> > > Lookie here. This is 2.6.0:mm/page_alloc.c:
> > > 
> > >         do_retry = 0;
> > >         if (!(gfp_mask & __GFP_NORETRY)) {
> > >                 if ((order <= 3) || (gfp_mask & __GFP_REPEAT))
> > >                         do_retry = 1;
> > >                 if (gfp_mask & __GFP_NOFAIL)
> > >                         do_retry = 1;
> > >         }
> > >         if (do_retry) {
> > >                 blk_congestion_wait(WRITE, HZ/50);
> > >                 goto rebalance;
> > >         }
> > 
> > rebalance:
> > 	if ((p->flags & (PF_MEMALLOC | PF_MEMDIE)) && !in_interrupt()) {
> > 		/* go through the zonelist yet again, ignoring mins */
> > 		for (i = 0; zones[i] != NULL; i++) {
> > 			struct zone *z = zones[i];
> > 
> > 			page = buffered_rmqueue(z, order, cold);
> > 			if (page)
> > 				goto got_pg;
> > 		}
> > 		goto nopage;
> > 	}
> 
> Your point?

That allocation attempts of any order can fail.

> That's the recursive allocation or oom case. Not the normal case at all.
> 
> The _normal_ case is to do the whole "try_to_free_pages()" case and try 
> and try again. Forever.

If the caller gets oom-killed, the allocation attempt fails.  Callers need
to handle that.

> IOW, we have traditionally never failed small kernel allocations. It makes 
> perfect sense that people _depend_ on that.
> 
> Now, we have since relaxed that (a lot). And in answer to that, people 
> have added more __GFP_NOFAIL flags, I bet. It's all very natural. Claiming 
> that this is some "new error" and that we should warn about NOFAIL 
> allocations with big orders is just silly and simply not true.
> 

There are situations in which the allocation attempt simply will not
succeed, so a __GFP_NOFAIL attempt will lock up.  Hence callers should
stop using __GFP_NOFAIL and should handle the allocation error like
99.9999% of the rest of the kernel does.

The chances of the allocation attempt failing increase with
higher-order allocations, hence the combination of __GFP_NOFAIL with
order>0 should be avoided more strenuously than __GFP_NOFAIL &&
order==0.


<Note that the TIF_MEMDIE handling has changed post-2.6.30.  I still
need to get my head around the end result of what we did there.  Did we
break the __alloc_pages-fails-if-TFI_MEMDIE logic?>