Re: [PATCH 1/2] slub: fix leakage

[Hugh Dickins <hugh@veritas.com>]

On Tue, 6 Nov 2007, Jeff Chua wrote:
> On Nov 6, 2007 2:45 AM, Christoph Lameter <clameter@sgi.com> wrote:
> 
> > 1. We can avoid going back to the page allocator for awhile since we will
> > find the almost free slab if the current slab is exhausted.
> 
> Does this impact SLAB as well?

No, it's in code peculiar to SLUB.

Of course, there might be a similar bug somewhere in SLAB.
More likely to be something else though.

> I'm getting out of memory with kernel
> 2.6.21, 2.6.22 and 2.6.23, so I'm stuck with 2.6.20-15 for systems
> running Oracle. It only happens with lots of activities.

It would help everybody if you could get more info on this.
Give 2.6.24-rc2 a try when it appears, if you can.

Do you think it's the SLAB which is growing?  If /proc/meminfo's
Slab count goes up and up, then use /proc/slabinfo to see what
it is that's leaking.  Then there's CONFIG_DEBUG_SLAB_LEAK.

With 2.6.22 or 2.6.23 you could CONFIG_SLUB instead, apply
the leak fix from this thread, and see if SLUB behaves the same
(using Documentation/vm/slabinfo.c instead of /proc/slabinfo); but
chances are the problem's above it, and it will behave the same.

Best start a new thread when you can report back.

Thanks,
Hugh