On Tue, Jun 26, 2018 at 01:55:09PM +0200, Cornelia Huck wrote:

On Tue, 26 Jun 2018 04:46:03 +0300
"Michael S. Tsirkin" [off-list ref] wrote:

quoted

On Mon, Jun 25, 2018 at 11:55:12AM +0200, Cornelia Huck wrote:

quoted

On Fri, 22 Jun 2018 22:05:50 +0300
"Michael S. Tsirkin" [off-list ref] wrote:
  

quoted

On Fri, Jun 22, 2018 at 05:09:55PM +0200, Cornelia Huck wrote:  

quoted

On Thu, 21 Jun 2018 21:20:13 +0300
"Michael S. Tsirkin" [off-list ref] wrote:
    

quoted

On Thu, Jun 21, 2018 at 04:59:13PM +0200, Cornelia Huck wrote:    

quoted

OK, so what about the following:

- introduce a new feature bit, VIRTIO_NET_F_STANDBY_UUID that indicates
  that we have a new uuid field in the virtio-net config space
- in QEMU, add a property for virtio-net that allows to specify a uuid,
  offer VIRTIO_NET_F_STANDBY_UUID if set
- when configuring, set the property to the group UUID of the vfio-pci
  device
- in the guest, use the uuid from the virtio-net device's config space
  if applicable; else, fall back to matching by MAC as done today

That should work for all virtio transports.      

True. I'm a bit unhappy that it's virtio net specific though
since down the road I expect we'll have a very similar feature
for scsi (and maybe others).

But we do not have a way to have fields that are portable
both across devices and transports, and I think it would
be a useful addition. How would this work though? Any idea?    

Can we introduce some kind of device-independent config space area?
Pushing back the device-specific config space by a certain value if the
appropriate feature is negotiated and use that for things like the uuid?    

So config moves back and forth?
Reminds me of the msi vector mess we had with pci.  

Yes, that would be a bit unfortunate.
  

quoted

I'd rather have every transport add a new config.  

You mean via different mechanisms?  

I guess so.

Is there an alternate mechanism for pci to use? (Not so familiar with
it.)

We have a device and transport config capability.
We could add a generic config capability too.

For ccw, this needs more thought. We already introduced two commands
for reading/writing the config space (a concept that does not really
exist on s390). There's the generic read configuration data command,
but the data returned by it is not really generic enough. So we would
need one new command (or two, if we need to write as well). I'm not
sure about that yet.

quoted

quoted

quoted

quoted

But regardless of that, I'm not sure whether extending this approach to
other device types is the way to go. Tying together two different
devices is creating complicated situations at least in the hypervisor
(even if it's fairly straightforward in the guest). [I have not come
around again to look at the "how to handle visibility in QEMU"
questions due to lack of cycles, sorry about that.]

So, what's the goal of this approach? Only to allow migration with
vfio-pci, or also to plug in a faster device and use it instead of an
already attached paravirtualized device?    

These are two sides of the same coin, I think the second approach
is closer to what we are doing here.  

Thinking about it, do we need any knob to keep the vfio device
invisible if the virtio device is not present? IOW, how does the
hypervisor know that the vfio device is supposed to be paired with a
virtio device? It seems we need an explicit tie-in.  

If we are going the way of the bridge, both bridge and
virtio would have some kind of id.

So the presence of the id would indicate "this is one part of a pair"?

I guess so, yes.

quoted

When pairing using mac, I'm less sure. PAss vfio device mac to qemu
as a property?

That feels a bit odd. "This is the vfio device's mac, use this instead
of your usual mac property"? As we have not designed the QEMU interface
yet, just go with the id in any case? The guest can still match by mac.

OK

quoted

quoted

quoted

quoted

What about migration of vfio devices that are not easily replaced by a
paravirtualized device? I'm thinking of vfio-ccw, where our main (and
currently only) supported device is dasd (disks) -- which can do a lot
of specialized things that virtio-blk does not support (and should not
or even cannot support).    

But maybe virtio-scsi can?  

I don't think so. Dasds have some channel commands that don't map
easily to scsi commands.  

There's always a choice of adding these to the spec.
E.g. FC extensions were proposed, I don't remember why they
are still stuck.

FC extensions are a completely different kind of enhancements, though.
For a start, they are not unique to a certain transport.

Also, we have a whole list of special dasd issues. Weird disk layout
for eckd, low-level disk formatting, etc. (See the list of commands in
drivers/s390/block/dasd_eckd.h for an idea. There's also no public
documentation AFAICS; https://en.wikipedia.org/wiki/ECKD does not link
to anything interesting.) I don't think we want to cram stuff like this
into a completely different framework.