Re: [Qemu-devel] [virtio-dev] [PATCH v3 0/7] Vhost-pci for inter-VM communication

[Stefan Hajnoczi <stefanha@redhat.com>]

[-- Attachment #1: Type: text/plain, Size: 2633 bytes --]

On Wed, Dec 13, 2017 at 10:59:10PM +0200, Michael S. Tsirkin wrote:
> >  * VHOST_USER_SET_VRING_KICK
> > 
> >    Set up vring kick doorbell (unless bit 8 is set) before sending
> >    VHOST_USER_SET_VRING_KICK to the guest.
> 
> But guest can't use it, now can it?
> 
> What guest needs is a mapping to interrupts.
...
> >  * VHOST_USER_SET_VRING_CALL
> > 
> >    Set up the vring call doorbell (unless bit 8 is set) before sending
> >    VHOST_USER_SET_VRING_CALL to the guest.
> 
> Same here. what guest needs is mapping from io to notifications,
> right?

The PCI device should contain a BAR with doorbell registers.  I don't
think a fancy mapping is necessary, instead the device spec should
define the BAR layout.

When the guest vhost-user slave receives this message it knows it can
now begin using the doorbell register.

> ---
> 
> > > I took a quick look and I doubt we can do something that is both
> > > compatible with the existing vhost-user and will make it possible to
> > > extend the protocol without qemu changes. Let's assume I pass a new
> > > message over the vhost-user channel.  How do we know it's safe to pass
> > > it to the guest?
> > >
> > > That's why we gate any protocol change on a feature bit and must parse
> > > all messages.
> > 
> > QEMU must parse all messages and cannot pass through unknown messages.
> > Think of QEMU vhost-pci as both a vhost-user slave to the other VM and
> > a vhost-user master to the guest.
> > 
> > QEMU changes are necessary when the vhost protocol is extended.
> > Device interface changes are only necessary if doorbells or shared
> > memory regions are added, any other protocol changes do not change the
> > device interface.
> > 
> > Stefan
> 
> I guess you have a different definition of a device interface than
> myself - I consider it an interface change if a feature bit changes :)

The feature bits are defined in the vhost-user protocol specification,
not in the vhost-pci PCI device specification.

For example, imagine we are adding the virtio-net MTU feature to the
protocol:

1. The VHOST_USER_PROTOCOL_F_MTU feature bit is added to the
   vhost-user protocol specification.

2. The VHOST_USER_NET_SET_MTU message is added to the vhost-user
   protocol specification.

As a result of this:

1. No PCI adapter resources (BARs, register layout, etc) change.  This
   is why I say the device interface is unchanged.  The vhost-pci
   specification does not change.

2. QEMU vhost-pci code needs to unmask VHOST_USER_PROTOCOL_F_MTU and
   pass through VHOST_USER_NET_SET_MTU.

Stefan

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 455 bytes --]