Re: [PATCH v10 25/41] virtio_pci: struct virtio_pci_common_cfg add queue_notify_data

["Michael S. Tsirkin" <mst@redhat.com>]

On Mon, Jun 27, 2022 at 04:14:20PM +0800, Jason Wang wrote:
> On Mon, Jun 27, 2022 at 3:58 PM Michael S. Tsirkin <mst@redhat.com> wrote:
> >
> > On Mon, Jun 27, 2022 at 03:45:30PM +0800, Jason Wang wrote:
> > > On Mon, Jun 27, 2022 at 2:39 PM Michael S. Tsirkin <mst@redhat.com> wrote:
> > > >
> > > > On Mon, Jun 27, 2022 at 10:30:42AM +0800, Jason Wang wrote:
> > > > > On Fri, Jun 24, 2022 at 2:59 PM Michael S. Tsirkin <mst@redhat.com> wrote:
> > > > > >
> > > > > > On Fri, Jun 24, 2022 at 10:56:05AM +0800, Xuan Zhuo wrote:
> > > > > > > Add queue_notify_data in struct virtio_pci_common_cfg, which comes from
> > > > > > > here https://github.com/oasis-tcs/virtio-spec/issues/89
> > > > > > >
> > > > > > > For not breaks uABI, add a new struct virtio_pci_common_cfg_notify.
> > > > > >
> > > > > > What exactly is meant by not breaking uABI?
> > > > > > Users are supposed to be prepared for struct size to change ... no?
> > > > >
> > > > > Not sure, any doc for this?
> > > > >
> > > > > Thanks
> > > >
> > > >
> > > > Well we have this:
> > > >
> > > >         The drivers SHOULD only map part of configuration structure
> > > >         large enough for device operation.  The drivers MUST handle
> > > >         an unexpectedly large \field{length}, but MAY check that \field{length}
> > > >         is large enough for device operation.
> > >
> > > Yes, but that's the device/driver interface. What's done here is the
> > > userspace/kernel.
> > >
> > > Userspace may break if it uses e.g sizeof(struct virtio_pci_common_cfg)?
> > >
> > > Thanks
> >
> > Hmm I guess there's risk... but then how are we going to maintain this
> > going forward?  Add a new struct on any change?
> 
> This is the way we have used it for the past 5 or more years. I don't
> see why this must be handled in the vq reset feature.
> 
> >Can we at least
> > prevent this going forward somehow?
> 
> Like have some padding?
> 
> Thanks

Maybe - this is what QEMU does ...

> >
> >
> > > >
> > > >
> > > >
> > > > >
> > > > > >
> > > > > >
> > > > > > > Since I want to add queue_reset after queue_notify_data, I submitted
> > > > > > > this patch first.
> > > > > > >
> > > > > > > Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
> > > > > > > Acked-by: Jason Wang <jasowang@redhat.com>
> > > > > > > ---
> > > > > > >  include/uapi/linux/virtio_pci.h | 7 +++++++
> > > > > > >  1 file changed, 7 insertions(+)
> > > > > > >
> > > > > > > diff --git a/include/uapi/linux/virtio_pci.h b/include/uapi/linux/virtio_pci.h
> > > > > > > index 3a86f36d7e3d..22bec9bd0dfc 100644
> > > > > > > --- a/include/uapi/linux/virtio_pci.h
> > > > > > > +++ b/include/uapi/linux/virtio_pci.h
> > > > > > > @@ -166,6 +166,13 @@ struct virtio_pci_common_cfg {
> > > > > > >       __le32 queue_used_hi;           /* read-write */
> > > > > > >  };
> > > > > > >
> > > > > > > +struct virtio_pci_common_cfg_notify {
> > > > > > > +     struct virtio_pci_common_cfg cfg;
> > > > > > > +
> > > > > > > +     __le16 queue_notify_data;       /* read-write */
> > > > > > > +     __le16 padding;
> > > > > > > +};
> > > > > > > +
> > > > > > >  /* Fields in VIRTIO_PCI_CAP_PCI_CFG: */
> > > > > > >  struct virtio_pci_cfg_cap {
> > > > > > >       struct virtio_pci_cap cap;
> > > > > > > --
> > > > > > > 2.31.0
> > > > > >
> > > >
> >