From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:33300) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1cH9Pw-0002R4-CX for qemu-devel@nongnu.org; Wed, 14 Dec 2016 08:16:25 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1cH9Pr-0008Gg-Ff for qemu-devel@nongnu.org; Wed, 14 Dec 2016 08:16:20 -0500 Received: from mx1.redhat.com ([209.132.183.28]:43056) by eggs.gnu.org with esmtps (TLS1.0:DHE_RSA_AES_256_CBC_SHA1:32) (Exim 4.71) (envelope-from ) id 1cH9Pr-0008GM-7j for qemu-devel@nongnu.org; Wed, 14 Dec 2016 08:16:15 -0500 Date: Wed, 14 Dec 2016 15:16:13 +0200 From: "Michael S. Tsirkin" Message-ID: <20161214151332-mutt-send-email-mst@kernel.org> References: <20161214125237.20850-1-maxime.coquelin@redhat.com> <20161214141210.22d12962.cornelia.huck@de.ibm.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20161214141210.22d12962.cornelia.huck@de.ibm.com> Subject: Re: [Qemu-devel] [PATCH] virtio-pci: Fix cross-version migration with older machines List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Cornelia Huck Cc: Maxime Coquelin , mdroth@linux.vnet.ibm.com, stefanha@redhat.com, qemu-devel@nongnu.org, marcel@redhat.com, "Dr . David Alan Gilbert" On Wed, Dec 14, 2016 at 02:12:10PM +0100, Cornelia Huck wrote: > On Wed, 14 Dec 2016 13:52:37 +0100 > Maxime Coquelin wrote: > > > This patch fixes a cross-version migration regression introduced > > by commit d1b4259f ("virtio-bus: Plug devices after features are > > negotiated"). > > > > The problem is encountered when host's vhost backend does not support > > VIRTIO_F_VERSION_1, and migration is initiated from a v2.7 or prior > > machine with virtio-pci modern capabilities enabled to a v2.8 machine. > > Wait, machine versions or qemu versions? > > > > > In this case, modern capabilities get exposed to the guest by the source, > > whereas the target will detect version 1 is not supported so will only > > expose legacy capabilities. > > > > The problem is fixed by introducing a new "x-modern-broken" property, > > which is set in v2.7 and prior compatibility modes. Doing this, v2.7 > > machine keeps its broken behaviour (enabling modern while version is > > not supported), and newer machines will behave correctly. > > > > Reported-by: Michael Roth > > Suggested-by: Stefan Hajnoczi > > Cc: Michael S. Tsirkin > > Cc: Cornelia Huck > > Cc: Marcel Apfelbaum > > Cc: Dr. David Alan Gilbert > > Signed-off-by: Maxime Coquelin > > --- > > > > I'm not sure about the property name, let me know if you have better ideas. > > I didn't tested migration yet, but I wanted to share the patch while I test it. > > I tested booting v2.8 and v2.7 machines with !VERSION_1 and get expected result: > > - v2.8: Virtio-pci probe succeed > > - v2.7: Virtio-pci probe fails > > > > Thanks, > > Maxime > > > > hw/virtio/virtio-pci.c | 4 +++- > > hw/virtio/virtio-pci.h | 1 + > > include/hw/compat.h | 4 ++++ > > 3 files changed, 8 insertions(+), 1 deletion(-) > > > > diff --git a/hw/virtio/virtio-pci.c b/hw/virtio/virtio-pci.c > > index 521ba0b..93f6b54 100644 > > --- a/hw/virtio/virtio-pci.c > > +++ b/hw/virtio/virtio-pci.c > > @@ -1580,7 +1580,8 @@ static void virtio_pci_device_plugged(DeviceState *d, Error **errp) > > * Virtio capabilities present without > > * VIRTIO_F_VERSION_1 confuses guests > > */ > > - if (!virtio_has_feature(vdev->host_features, VIRTIO_F_VERSION_1)) { > > + if (!proxy->modern_broken && > > What you are de facto doing here is ignoring the features supported by > the backend. Call this ->ignore_backend_features or so? > > > + !virtio_has_feature(vdev->host_features, VIRTIO_F_VERSION_1)) { > > virtio_pci_disable_modern(proxy); > > > > if (!legacy) { > > @@ -1852,6 +1853,7 @@ static Property virtio_pci_properties[] = { > > VIRTIO_PCI_FLAG_DISABLE_PCIE_BIT, false), > > DEFINE_PROP_BIT("page-per-vq", VirtIOPCIProxy, flags, > > VIRTIO_PCI_FLAG_PAGE_PER_VQ_BIT, false), > > + DEFINE_PROP_BOOL("x-modern-broken", VirtIOPCIProxy, modern_broken, false), > > What about "x-ignore-backend-features"? It's just the capability that ignores that backend is legacy. x-cap-ignore-backend-legacy ? > > DEFINE_PROP_END_OF_LIST(), > > }; > > > > diff --git a/hw/virtio/virtio-pci.h b/hw/virtio/virtio-pci.h > > index b2a996f..1dca223 100644 > > --- a/hw/virtio/virtio-pci.h > > +++ b/hw/virtio/virtio-pci.h > > @@ -153,6 +153,7 @@ struct VirtIOPCIProxy { > > int config_cap; > > uint32_t flags; > > bool disable_modern; > > + bool modern_broken; > > OnOffAuto disable_legacy; > > uint32_t class_code; > > uint32_t nvectors; > > diff --git a/include/hw/compat.h b/include/hw/compat.h > > index 0f06e11..fe11723 100644 > > --- a/include/hw/compat.h > > +++ b/include/hw/compat.h > > @@ -18,6 +18,10 @@ > > .driver = "intel-iommu",\ > > .property = "x-buggy-eim",\ > > .value = "true",\ > > + },{\ > > + .driver = "virtio-pci",\ > > + .property = "x-modern-broken",\ > > + .value = "on",\ > > This unfortunately has the same problem wrt -global propagation as the > other virtio-pci compat props (unexpeced overrides). There's basically > zero chance, however, that somebody will want to set this property > manually (unline disable-legacy/disable-modern), and there's a patch > targeting 2.8.1, so I think we can live with it. (Alternatively, you > would need to set this property on all possible virtio-pci devices.) YEs. > > }, > > > > #define HW_COMPAT_2_6 \