Re: [PATCH] fix spurious OOM kills

[Andrea Arcangeli]

On Wed, Dec 15, 2004 at 01:16:23AM +0100, Thomas Gleixner wrote:
> It solves one of the problems, but your fix is really the only complete
> fix I have in hands since this thread(s) started. + my simple changes to
> the whom to kill selection :)

That patch prevents the machine to trigger "early" "suprious" oom kills
(I had reports of suprious oom kills myself, oom killer triggered
despite lots of swapcache was freeable), so it cannot help when a true
oom happens like with your workload. In your workload the oom isn't
a suprious error.

The two patches to apply are out there (you posted a version that merges
both of them and doesn't even require to fix the caller of alloc_pages
that should be using GFP_ATOMIC instead of GFP_KERNEL). I would like to
fix those callers too from my part ;), but I understand if it's not
something for a mainline kernel (at least I'm very glad I didn't have to
find this bug in the drivers the hard way ;).
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/