Re: [Qemu-devel] [PATCH 3/4] net/virtio: add failover support

[Michael S. Tsirkin]

On Fri, May 31, 2019 at 07:45:13PM +0100, Dr. David Alan Gilbert wrote:
> * Michael S. Tsirkin (address@hidden) wrote:
> > On Fri, May 31, 2019 at 02:01:54PM -0300, Eduardo Habkost wrote:
> > > > Yes. It's just lots of extremely low level interfaces
> > > > and all rather pointless.
> > > > 
> > > > And down the road extensions like surprise removal support will make it
> > > > all cleaner and more transparent. Floating things up to libvirt means
> > > > all these low level details will require more and more hacks.
> > > 
> > > Why do you call it pointless?
> > 
> > We'd need APIs to manipulate device visibility to guest, hotplug
> > controller state and separately manipulate the resources allocated. This
> > is low level stuff that users really have no idea what to do about.
> > Exposing such a level of detail to management is imho pointless.
> > We are better off with a high level API, see below.
> 
> so I don't know much about vfio; but to me it strikes me that
> you wouldn't need that low level detail if we just reworked vfio
> to look more like all our other devices; something like:
> 
>   -vfiodev  host=02:00.0,id=gpu
>   -device vfio-pci,dev=gpu
> 
> The 'vfiodev' would own the resources; so to do this trick, the
> management layer would:
>    hotunplug the vfio-pci
>    migrate
> 
> if anything went wrong it would
>    hotplug the vfio-pci backin
> 
> you wouldn't have free'd up any resources because they belonged
> to the vfiodev.


IIUC that doesn't really work with passthrough
unless guests support surprise removal.


> > > If we want this to work before
> > > surprise removal is implemented, we need to provide an API that
> > > works for management software.
> > >  Don't we want to make this work
> > > without surprise removal too?
> > 
> > This patchset adds an optional, off by default support for
> > migrating guests with an assigned network device.
> > If enabled this requires guest to allow migration.
> > 
> > Of course this can be viewed as a security problem since it allows guest
> > to block migration. We can't detect a malicious guest reliably imho.
> > What we can do is report to management when guest allows migration.
> > Policy such what to do when this does not happen for a while and
> > what timeout to set would be up to management.
> > 
> > The API in question would be a high level one, something
> > along the lines of a single "guest allowed migration" event.
> 
> This is all fairly normal problems with hot unplugging - that's
> already dealt with at higher levels for normal hot unplugging.
> 
> The question here is to try to avoid duplicating that fairly
> painful process in qemu.
> 
> Dave
> > 
> > -- 
> > MST
> --
> Dr. David Alan Gilbert / address@hidden / Manchester, UK