Return-Path: <linux-kernel-owner@vger.kernel.org>
Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand
	id S935974Ab0GSPVc (ORCPT <rfc822;w@1wt.eu>);
	Mon, 19 Jul 2010 11:21:32 -0400
Received: from mx1.redhat.com ([209.132.183.28]:55120 "EHLO mx1.redhat.com"
	rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP
	id S935696Ab0GSPVb (ORCPT <rfc822;linux-kernel@vger.kernel.org>);
	Mon, 19 Jul 2010 11:21:31 -0400
Date: Mon, 19 Jul 2010 18:16:10 +0300
From: "Michael S. Tsirkin" <mst@redhat.com>
To: Chris Mason <chris.mason@oracle.com>, linux-kernel@vger.kernel.org,
        Rusty Russell <rusty@rustcorp.com.au>
Subject: Re: 2.6.35 Regression/oops from virtio: return ENOMEM on out of
 memory patch
Message-ID: <20100719151609.GA4267@redhat.com>
References: <20100719150216.GC8623@think>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20100719150216.GC8623@think>
User-Agent: Mutt/1.5.20 (2009-12-10)
Sender: linux-kernel-owner@vger.kernel.org
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org
Content-Length: 2969
Lines: 83

On Mon, Jul 19, 2010 at 11:02:16AM -0400, Chris Mason wrote:
> Hi everyone,
> 
> http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=686d363786a53ed28ee875b84ef24e6d5126ef6f
> 
> I've been having problems with my long running stress runs and tracked
> it down to the above commit.  Under load I get a couple of GFP_ATOMIC
> allocation failures from virtio per day (not really surprising), and in
> the past it would carry on happily.
> 
> Now I get the atomic allocation failure followed by this:
> 
> BUG: unable to handle kernel paging request at ffff88087c37e458
> IP: [<ffffffff812e3752>] virtqueue_add_buf_gfp+0x305/0x353
> 
> (Full oops below).
> 
> Looking at virtqueue_add_buf_gfp, it does:
> 
>         /* If the host supports indirect descriptor tables, and we have multiple 
>          * buffers, then go indirect. FIXME: tune this threshold */
>         if (vq->indirect && (out + in) > 1 && vq->num_free) {
>                 head = vring_add_indirect(vq, sg, out, in, gfp);
>                 if (head != vq->vring.num)
>                         goto add_head;
>         }
> [ ... ]
> 
> add_head:
>         /* Set token. */
>         vq->data[head] = data;
> 
> Since vring_add_indirect is returning -ENOMEM, head is -ENOMEM and things
> go bad pretty quickly.  Full oops below, afraid I don't know the virtio
> code well enough to provide the clean and obvious fix (outside of
> reverting) at this late rc.

Good catch! Can you verify this fix please?

virtio: fix oops on OOM

virtio ring was changed to return an error code on OOM,
but one caller was missed and still checks for vq->vring.num.
The fix is just to check for <0 error code.

Long term it might make sense to change goto add_head to
just return an error on oom instead, but let's apply
a minimal fix for 2.6.35.

Reported-by: Chris Mason <chris.mason@oracle.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

---

diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
index dd35b34..bffec32 100644
--- a/drivers/virtio/virtio_ring.c
+++ b/drivers/virtio/virtio_ring.c
@@ -164,7 +164,8 @@ int virtqueue_add_buf_gfp(struct virtqueue *_vq,
 			  gfp_t gfp)
 {
 	struct vring_virtqueue *vq = to_vvq(_vq);
-	unsigned int i, avail, head, uninitialized_var(prev);
+	unsigned int i, avail, uninitialized_var(prev);
+	int head;
 
 	START_USE(vq);
 
@@ -174,8 +175,8 @@ int virtqueue_add_buf_gfp(struct virtqueue *_vq,
 	 * buffers, then go indirect. FIXME: tune this threshold */
 	if (vq->indirect && (out + in) > 1 && vq->num_free) {
 		head = vring_add_indirect(vq, sg, out, in, gfp);
-		if (head != vq->vring.num)
+		if (likely(head >= 0))
 			goto add_head;
 	}
 
 	BUG_ON(out + in > vq->vring.num);
--
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/