From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([208.118.235.92]:57709) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1SALHh-0006W0-2O for qemu-devel@nongnu.org; Wed, 21 Mar 2012 09:09:07 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1SALHa-0007aJ-J8 for qemu-devel@nongnu.org; Wed, 21 Mar 2012 09:09:00 -0400 Date: Wed, 21 Mar 2012 15:08:53 +0200 From: "Michael S. Tsirkin" Message-ID: <20120321130852.GC21467@redhat.com> References: <1332133163-7890-1-git-send-email-david@gibson.dropbear.id.au> <20120319113310.GD30033@stefanha-thinkpad.localdomain> <20120320004206.GB22089@truffala.fritz.box> <20120320101947.GA19542@truffala.fritz.box> <20120321112615.GA15090@stefanha-thinkpad.localdomain> MIME-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Disposition: inline In-Reply-To: <20120321112615.GA15090@stefanha-thinkpad.localdomain> Content-Transfer-Encoding: quoted-printable Subject: Re: [Qemu-devel] [PATCH] Remove PCI class code from virtio balloon device List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Stefan Hajnoczi Cc: qemu-trivial@nongnu.org, Rusty Russell , qemu-devel@nongnu.org, anthony@codemonkey.ws On Wed, Mar 21, 2012 at 11:26:15AM +0000, Stefan Hajnoczi wrote: > On Tue, Mar 20, 2012 at 09:19:47PM +1100, David Gibson wrote: > > On Tue, Mar 20, 2012 at 09:54:20AM +0000, Stefan Hajnoczi wrote: > > > On Tue, Mar 20, 2012 at 12:42 AM, David Gibson > > > wrote: > > > > On Mon, Mar 19, 2012 at 11:33:10AM +0000, Stefan Hajnoczi wrote: > > > >> On Mon, Mar 19, 2012 at 03:59:23PM +1100, David Gibson wrote: > > > >> > Currently the virtio balloon device, when using the virtio-pci= interface > > > >> > advertises itself with PCI class code MEMORY_RAM. =A0This is w= rong; the > > > >> > balloon is vaguely related to memory, but is nothing like a PC= I memory > > > >> > device in the meaning of the class code, and this code is not = required or > > > >> > suggested by the virtio PCI specification. > > > >> > > > > >> > Worse, this patch causes problems on the pseries machine, beca= use the > > > >> > firmware, seeing this class code, advertises the device as mem= ory in the > > > >> > device tree, and then a guest kernel bug causes it to see this= "memory" > > > >> > before the real system memory, leading to a crash in early boo= t. > > > >> > > > > >> > This patch fixes the problem by removing the bogus PCI class c= ode on the > > > >> > balloon device. > > > >> > > > > >> > Cc: Michael S. Tsirkin > > > >> > Cc: Rusty Russell > > > >> > > > > >> > Signed-off-by: David Gibson > > > >> > --- > > > >> > =A0hw/virtio-pci.c | =A0 =A02 +- > > > >> > =A01 files changed, 1 insertions(+), 1 deletions(-) > > > >> > > > >> Since this is a guest-visible change we might need to be careful= about > > > >> how it's introduced. > > > >> > > > >> Do we need to keep the old class code for existing machine types= ? =A0The > > > >> new class code could be introduced only for 1.1 and later machin= e types > > > >> if we want to be extra careful about introducing guest-visible > > > >> changes. > > > > > > > > So as a general rule, I like to be very careful about user-visibl= e > > > > changes. =A0But in this case, I don't think we want to be too hes= itant. > > > > In particular, it's not just a question of the machine type, but = also > > > > of how the guest OS will deal with the PCI class code. > > > > > > > > The class code we were using was Just Plain Wrong. =A0It was not > > > > suggetsed by the virtio spec, and it makes no sense. =A0It happen= s that > > > > so far this caused problems only for a guest on a particular mach= ine > > > > type, but there's no reason it couldn't cause (different) problem= s for > > > > guests on any machine type. > > > > > > > > More to the point, it seems reasonably unlikely for existing gues= ts to > > > > rely on the broken behaviour: again, there's no reason they'd thi= nk > > > > they need to based on the spec, and the usual way of matching dri= vers > > > > to PCI devices is with the vendor/device IDs which are correct an= d not > > > > changed by this patch. > > > > > > > > So, unless we have a known example of an existing guest that woul= d be > > > > broken by this change, I think we should implement it ASAP for al= l > > > > machine types. > > >=20 > > > I agree that in practice the risk is low because working guests are > > > probably not using the class code. On the other hand I don't see a > > > downside to making this part of the 1.1 machine type, > >=20 > > Well.. there's the fact that I can't what mechanism we would use to > > make this per-machine... >=20 > Not sure I parsed this correctly, but I think you're asking how to do > it. >=20 > Looking at hw/pc_piix.c there are QEMUMachine types for each QEMU > release. Legacy machine types (e.g. pc_machine_v0_14) have a > .compat_props array that can override qdev properties. >=20 > Perhaps Michael Tsirkin or someone else can comment on how to wire up > hw/virtio-pci.c so that the class code can be overridden. >=20 > Stefan afaik we already let users over-write it for some other pci devices, look there for examples.