Re: [virtio-dev] Re: [virtio-comment] [RFC PATCH v3] virtio-can: Device specification.

[Cornelia Huck <cohuck@redhat.com>]

On Wed, Feb 21 2024, Marc Kleine-Budde <mkl@pengutronix.de> wrote:

> On 21.02.2024 14:16:54, Matias Ezequiel Vara Larsen wrote:
>> On Wed, Feb 21, 2024 at 01:49:31PM +0100, Marc Kleine-Budde wrote:
>> > On 21.02.2024 11:37:58, Matias Ezequiel Vara Larsen wrote:
>> > > > > +The length of the \field{sdu} is determined by the \field{length}.
>> > > > > +
>> > > > > +The type of a CAN message identifier is determined by \field{flags}. The
>> > > > > +3 most significant bits of \field{can_id} do not bear the information
>> > > > > +about the type of the CAN message identifier and are 0.
>> > > > > +
>> > > > > +The device MUST reject any CAN frame type for which support has not been
>> > > > > +negotiated with VIRTIO_CAN_RESULT_NOT_OK in \field{result} and MUST NOT
>> > > > > +schedule the message for transmission. A CAN frame with an undefined bit
>> > > > > +set in \field{flags} is treated like a CAN frame for which support has
>> > > > > +not been negotiated.
>> > > > > +
>> > > > > +The device MUST reject any CAN frame for which \field{can_id} or
>> > > > > +\field{sdu} length are out of range or the CAN controller is in an
>> > > > > +invalid state with VIRTIO_CAN_RESULT_NOT_OK in \field{result} and MUST
>> > > > > +NOT schedule the message for transmission.
>> > > > > +
>> > > I am not very familiar with CAN but how does the device figure out that
>> > > the can_id is out of range?
>> > 
>> > In classical CAN we have the standard CAN frames, which have an 11 bit
>> > ID, and there are extended CAN frames, which have 29 bits ID. Extended
>> > frames are signaled with VIRTIO_CAN_FLAGS_EXTENDED set.
>> > 
>> > So if a standard frame uses more than 11 Bits of CAN-ID, it's considered
>> > out of range.
>
> Another option would be an extended frame (VIRTIO_CAN_FLAGS_EXTENDED
> set) and using more than 29 bits.
>
>> Thanks Marc for the explanation. Do you think that it would be
>> worthwhile to add that to the spec at some point?
>
> Yes that makes sense as it clarifies what's meant by out of range for
> CAN-IDs, for the valid length a reference to
> \item[VIRTIO_CAN_F_CAN_CLASSIC (0)] and \item[VIRTIO_CAN_F_CAN_FD (1)]
> might be added.

[virtio mailing lists are supposedly down for migration right now, I
hope there's some kind of backfill happening later...]

If the question comes up, it does make sense to add a
clarification... as the virtio-can spec is already voted upon and
merged, we'd need a patch on top. Not sure if it would qualify as an
editorial update or a vote would be needed, best to see it first :)