On Tue, Apr 18, 2023 at 07:20:29PM +0200, Andrey Ryabinin wrote: > > > On 4/18/23 08:17, Michael S. Tsirkin wrote: > > On Tue, Apr 18, 2023 at 05:13:11AM +0000, Raphael Norwitz wrote: > >> Hey Andrey - apologies for the late reply here. > >> > >> It sounds like you are dealing with a buggy guest, rather than a QEMU > >> issue. > >> > >>> On Apr 10, 2023, at 11:39 AM, Andrey Ryabinin <a...@yandex-team.com> > >>> wrote: > >>> > >>> > >>> > >>> On 4/10/23 10:35, Andrey Ryabinin wrote: > >>>> Some guests hang on boot when using the vhost-user-blk-pci device, > >>>> but boot normally when using the virtio-blk device. The problem occurs > >>>> because the guest advertises VIRTIO_F_VERSION_1 but kicks the virtqueue > >>>> before setting VIRTIO_CONFIG_S_DRIVER_OK, causing vdev->start_on_kick to > >> > >> Virtio 1.1 Section 3.1.1, says during setup “[t]he driver MUST NOT notify > >> the device before setting DRIVER_OK.” > >> > >> Therefore what you are describing is buggy guest behavior. Sounds like the > >> driver should be made to either > >> - not advertise VIRTIO_F_VERSION_1 > >> - not kick before setting VIRTIO_CONFIG_S_DRIVER_OK > >> > >> If anything, the virtio-blk virtio_blk_handle_output() function should > >> probably check start_on_kick? > > > > Question is, how easy is this guest to fix. > > > > I wouldn't count on that. > > In this case the guest is Foritgate firewall, apparently from this guys > https://www.fortinet.com/ > It seems that the kernel they use claims itself as 3.2.16 Linux kernel, > however it looks like > it's not vanilla kernel, but modified with some backports. I'm guessing that > they backported > the patches introducing VIRTIO_F_VERSTION_1, but they didn't add this patch > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=7a11370e5e6c26566904bb7f08281093a3002ff2 > > I've tried to look the sources of the kernel they use but failed to find any. > Found only some news about gpl voilation from 2005 )
All I remember is this is a can of worms, the change seems risky frankly. If you like, I'm open to an unstable property (starting with "x-") that forces start on kick. Will be handy for debugging too. -- MST
