On Donnerstag, 3. September 2020 11:35:14 CEST Greg Kurz wrote:
> > On the long term that would be possible, however only with a protocol
> > change allowing server to send minimum, maximum and recommended msize to
> > client.
> Hmm... not sure adding a new 9P protocol version for this is the
> way to go. Not speaking of all the hustle this would cause, these
> msizes rather look like properties of the device that the guest
> can use to configure the 9P session.
>
> What about adding them to the virtio-9p device config along with
> the mount 'tag' and teach trans_virtio.c in linux to expose them
> as well in the sysfs entry of the device ?
>
> This could also be the occasion to describe virtio-9p in the virtio
> spec [1]. Something that has been sitting on my TODO list for years
> but I could never find time to consider...
>
> [1] https://docs.oasis-open.org/virtio/virtio/v1.1/virtio-v1.1.html
Sounds like a good idea!
> > As you know, right now server only has a say in maximum msize.
>
> I guess you mean minimum size ?
Well, it's ambiguous, but we mean the same thing: server may optionally lower
the msize previously suggested by client; server must not raise client's
suggested msize though.
If you see it in the context of Rversion response handling then that's a
"minimum" operation, yes.
If you see it as prose then it is "maximum msize", e.g.:
Server's max. supported msize: 20 MiB <- covered by 9P protocol
Server's min. supported msize: 4 kiB <- not officially covered by 9P
Client suggests msize 100 MiB -> server lowers that to
min(100MiB, 20MiB) = 20MiB.
Best regards,
Christian Schoenebeck
