From: Sasha Levin <levinsasha928@gmail.com>
To: David Gibson <david@gibson.dropbear.id.au>
Cc: aliguori@us.ibm.com, aik@ozlabs.ru, rusty@rustcorp.com.au,
qemu-devel@nongnu.org, agraf@suse.de, pbonzini@redhat.com
Subject: Re: [Qemu-devel] [PATCH] virtio: Make memory barriers be memory barriers
Date: Thu, 01 Sep 2011 10:37:02 +0300 [thread overview]
Message-ID: <1314862622.26813.6.camel@lappy> (raw)
In-Reply-To: <1314857389-13363-1-git-send-email-david@gibson.dropbear.id.au>
On Thu, 2011-09-01 at 16:09 +1000, David Gibson wrote:
> The virtio code already has memory barrier wmb() macros in the code.
> However they are was defined as no-ops. The comment claims that real
> barriers are not necessary because the code does not run concurrent.
> However, with kvm and io-thread enabled, this is not true and this qemu
> code can indeed run concurrently with the guest kernel. This does not
> cause problems on x86 due to it's strongly ordered storage model, but it
> causes a race leading to virtio errors on POWER which has a relaxed storage
> ordering model.
>
> Specifically, the QEMU puts new element into the "used" ring and then
> updates the ring free-running counter. Without a barrier between these
> under the right circumstances, the guest linux driver can receive an
> interrupt, read the counter change but find the ring element to be handled
> still has an old value, leading to an "id %u is not a head!\n" error
> message.
>
> The problem is easy to reproduce on POWER using virtio-net with heavy
> traffic.
>
> The patch defines wmb() as __sync_synchronize(), a cross platform memory
> barrier primitive available in sufficiently recent gcc versions (gcc 4.2
> and after?). If we care about older gccs then this patch will need to
> be updated with some sort of fallback.
>
> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
> Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
> ---
> hw/virtio.c | 10 ++--------
> 1 files changed, 2 insertions(+), 8 deletions(-)
>
> diff --git a/hw/virtio.c b/hw/virtio.c
> index 13aa0fa..c9f0e75 100644
> --- a/hw/virtio.c
> +++ b/hw/virtio.c
> @@ -21,14 +21,8 @@
> * x86 pagesize again. */
> #define VIRTIO_PCI_VRING_ALIGN 4096
>
> -/* QEMU doesn't strictly need write barriers since everything runs in
> - * lock-step. We'll leave the calls to wmb() in though to make it obvious for
> - * KVM or if kqemu gets SMP support.
> - * In any case, we must prevent the compiler from reordering the code.
> - * TODO: we likely need some rmb()/mb() as well.
> - */
> -
> -#define wmb() __asm__ __volatile__("": : :"memory")
> + /* TODO: we may also need rmb()s. It hasn't bitten us yet, but.. */
> + #define wmb() __sync_synchronize()
That asm directive also implicitly provided a compiler barrier, I could
find whether __sync_synchronize() provides one as well.
Any idea if it does?
>
> typedef struct VRingDesc
> {
--
Sasha.
next prev parent reply other threads:[~2011-09-01 7:37 UTC|newest]
Thread overview: 32+ messages / expand[flat|nested] mbox.gz Atom feed top
2011-09-01 6:09 [Qemu-devel] [PATCH] virtio: Make memory barriers be memory barriers David Gibson
2011-09-01 6:47 ` Paolo Bonzini
2011-09-02 0:08 ` David Gibson
2011-09-02 6:49 ` Paolo Bonzini
2011-09-03 11:53 ` Blue Swirl
2011-09-01 7:37 ` Sasha Levin [this message]
2011-09-01 7:38 ` Sasha Levin
2011-09-01 7:56 ` Paolo Bonzini
2011-09-02 0:09 ` David Gibson
2011-09-01 15:30 ` Michael S. Tsirkin
2011-09-01 16:14 ` Paolo Bonzini
2011-09-01 16:34 ` Michael S. Tsirkin
2011-09-01 20:31 ` Paolo Bonzini
2011-09-02 15:45 ` Michael S. Tsirkin
2011-09-03 14:46 ` David Gibson
2011-09-04 9:16 ` Michael S. Tsirkin
2011-09-05 4:43 ` David Gibson
2011-09-05 9:19 ` Michael S. Tsirkin
2011-09-06 3:12 ` David Gibson
2011-09-06 6:55 ` Paolo Bonzini
2011-09-06 9:02 ` David Gibson
2011-09-06 9:28 ` Avi Kivity
2011-09-06 9:35 ` Michael S. Tsirkin
2011-09-06 9:38 ` Paolo Bonzini
2011-09-05 7:41 ` Paolo Bonzini
2011-09-05 8:06 ` Michael S. Tsirkin
2011-09-05 9:42 ` Paolo Bonzini
2011-09-03 16:19 ` Paolo Bonzini
2011-09-04 8:47 ` Michael S. Tsirkin
2011-09-02 0:11 ` David Gibson
2011-09-02 6:11 ` Paolo Bonzini
2011-09-02 15:57 ` Michael S. Tsirkin
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1314862622.26813.6.camel@lappy \
--to=levinsasha928@gmail.com \
--cc=agraf@suse.de \
--cc=aik@ozlabs.ru \
--cc=aliguori@us.ibm.com \
--cc=david@gibson.dropbear.id.au \
--cc=pbonzini@redhat.com \
--cc=qemu-devel@nongnu.org \
--cc=rusty@rustcorp.com.au \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).