Re: [PATCH] mm: exempt pcp alloc from watermarks

[Nick Piggin <nickpiggin@yahoo.com.au>]

Peter Zijlstra wrote:
> On Wed, 2006-09-20 at 00:35 +1000, Nick Piggin wrote:
> 

>>Thanks for the patch! I have a slight preference for the following
>>version, which speculatively tests pcp->count without disabling
>>interrupts (the chance of being preempted or scheduled in this
>>window is basically the same as the chance of being preempted after
>>checking watermarks). What do you think?
> 
> 
> The race here allows to wrongly bypass the watermark check. My version
> raced the other way about, where you could find a non empty pcp where an
> empty one was otherwise expected.

I really doubt it matters. You could be preempted after that check
anyway, and by the time you return the previous watermark check is
meaningless.

If we really want to be strict about watermark checks, it has to be
done with the zone lock held, no other option. I doubt anybody
bothered, because the watermarks (even PF_MEMALLOC pool) are all
heuristics anyway and it is a better idea to keep fastpath code fast.

> OTOH it is much shorter, I'll see if I can shorten mine and keep the
> race safe (and perhaps do what Christoph suggests).

-- 
SUSE Labs, Novell Inc.
Send instant messages to your online friends http://au.messenger.yahoo.com 

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>