From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:53726) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1Z8APx-0004nn-Dn for qemu-devel@nongnu.org; Thu, 25 Jun 2015 12:54:30 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1Z8APu-0006sJ-70 for qemu-devel@nongnu.org; Thu, 25 Jun 2015 12:54:25 -0400 Received: from mx1.redhat.com ([209.132.183.28]:45471) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1Z8APt-0006s7-Vs for qemu-devel@nongnu.org; Thu, 25 Jun 2015 12:54:22 -0400 Message-ID: <558C3239.5070300@redhat.com> Date: Thu, 25 Jun 2015 19:54:17 +0300 From: Marcel Apfelbaum MIME-Version: 1.0 References: <1434650744-6585-1-git-send-email-marcel@redhat.com> <558A6399.6090906@redhat.com> In-Reply-To: <558A6399.6090906@redhat.com> Content-Type: text/plain; charset=utf-8; format=flowed Content-Transfer-Encoding: 7bit Subject: Re: [Qemu-devel] [PATCH] net/virtio: fix multi-queue negotiation List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Jason Wang , qemu-devel@nongnu.org Cc: famz@redhat.com, n.nikolaev@virtualopensystems.com, mst@redhat.com On 06/24/2015 11:00 AM, Jason Wang wrote: > > > On 06/19/2015 02:05 AM, Marcel Apfelbaum wrote: >> Clear host multi-queue related features if the peer >> doesn't support it. >> >> Signed-off-by: Marcel Apfelbaum >> --- >> Notes: >> This fixes a guest CPU soft lock, however the virtio-net >> device will not work correctly. It seems that is >> peer's "fault", not knowing how to handle the situation. >> However, I submit this patch since it corrects the negotiation >> and saves us from a guest crash (!). > > Could you please describe how to reproduce this issue? Hi Jason, Sorry for the late reply. I was hoping that a vhost-user multi-queue "insider" will ask questions :) This happens when we have OVS as backend without multi-queue support and qemu/guest with multi-queue-support. > >> >> Any ideas from the virtio/multi-queue developers on how to debug >> this further are welcomed. >> >> hw/net/virtio-net.c | 13 ++++++++++++- >> 1 file changed, 12 insertions(+), 1 deletion(-) >> >> diff --git a/hw/net/virtio-net.c b/hw/net/virtio-net.c >> index 9281aa1..63e59e8 100644 >> --- a/hw/net/virtio-net.c >> +++ b/hw/net/virtio-net.c >> @@ -367,6 +367,11 @@ static int peer_has_ufo(VirtIONet *n) >> return n->has_ufo; >> } >> >> +static int peer_has_multiqueue(VirtIONet *n) >> +{ >> + return n->multiqueue; >> +} > > The name is confusing, this is in fact whether or not guest support > multiqueue. To check peer's ability, you need check > n->nic_conf.peers.queues instead. I just wanted to mimic similar code, I have no issue against this approach. > >> + >> static void virtio_net_set_mrg_rx_bufs(VirtIONet *n, int mergeable_rx_bufs, >> int version_1) >> { >> @@ -469,6 +474,13 @@ static uint64_t virtio_net_get_features(VirtIODevice *vdev, uint64_t features) >> virtio_clear_feature(&features, VIRTIO_NET_F_HOST_UFO); >> } >> >> + if (!peer_has_multiqueue(n)) { >> + virtio_clear_feature(&features, VIRTIO_NET_F_MQ); > > I'm not quite understand this, VIRTIO_NET_F_MQ should work if peer has > only 1 queue. Please explain, I thought VIRTIO_NET_F_MQ flag *enables* multi-queue, and in our case we pass queue=2 to vhost. > >> + virtio_clear_feature(&features, VIRTIO_NET_F_GUEST_ANNOUNCE); >> + virtio_clear_feature(&features, VIRTIO_NET_F_CTRL_VQ); >> + virtio_clear_feature(&features, VIRTIO_NET_F_CTRL_RX); >> + } >> + > > Those features don't depend on multiqueue, why clear them? Once VIRTIO_NET_F_MQ is cleared, the virtio driver from guest is complaining about those other flags too. Any explanations about how *it should* work are welcomed. Again, the scenario is: backend doesn't support multi-queue, QEMU/guest do, and queues=2 is passed on command line. Thanks, Marcel > >> if (!get_vhost_net(nc->peer)) { >> virtio_add_feature(&features, VIRTIO_F_VERSION_1); >> return features; >> @@ -1314,7 +1326,6 @@ static void virtio_net_tx_bh(void *opaque) >> static void virtio_net_set_multiqueue(VirtIONet *n, int multiqueue) >> { >> n->multiqueue = multiqueue; >> - >> virtio_net_set_queues(n); >> } >> >