Date: Wed, 25 May 2011 09:25:12 -0500 (CDT)
From: Christoph Lameter <cl@linux.com>
To: Linus Torvalds <torvalds@linux-foundation.org>
cc: James Morris <jmorris@namei.org>, Pekka Enberg <penberg@kernel.org>,
        linux-kernel@vger.kernel.org
Subject: Re: SLUB regression in current Linus
In-Reply-To: <BANLkTik=H=1+oqQ9_5ZZaHpFn8_9x5V-nQ@mail.gmail.com>
Message-ID: <alpine.DEB.2.00.1105250921560.23309@router.home>
References: <alpine.LRH.2.00.1105242146240.12134@tundra.namei.org> <BANLkTik=H=1+oqQ9_5ZZaHpFn8_9x5V-nQ@mail.gmail.com>
User-Agent: Alpine 2.00 (DEB 1167 2008-08-23)
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1135
Lines: 30

On Tue, 24 May 2011, Linus Torvalds wrote:

> Look at __slab_alloc: we have:
>
>
>         page = c->page;
>         if (!page)
>                 goto new_slab;
>
>         slab_lock(page);
>         if (unlikely(!node_match(c, node)))
>                 goto another_slab;
>
> and let's assume we have two users racing on that "c->page". The
> "slab_lock()" is going to work for one of them, right?

There cannot be two users racing through this code segment since we have
interrupts disabled and c is pointing to a per cpu structure. c->page
points to a page that can only be allocated from from the current
processor (from the freelist in c->freelist) but it can be freed to from
multiple cpus (via the page->freelist). The code that you are discussing
is copying the freed objects from the page->freelist to the per cpu
freelist and it needs to lock out the slab_free path to do that.


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/