2008-06-21 23:54:47

by Linus Torvalds

[permalink] [raw]
Subject: Re: 2.6.26-rc: nfsd hangs for a few sec

On Sat, 21 Jun 2008, Christoph Lameter wrote:
> Not a direct explanation for the problem but the memory wastage could
> certainly can heretofore undiscovered locking dependencies to be
> exposed.

Well, not for these traces, no. The trace contains __slab_alloc() in the
call chain, which definitely fingers SLUB, not slab, despite the name
(slab calls its allocation routines "cache_alloc", while slub calls them
"slab_alloc" ;)

So the patch looks fine, and I applied it, but as Mel already mentioned,
it looks like it won't be making any difference for Alexander.