Re: [virtio] [OASIS Issue Tracker] (VIRTIO-106) Virtio SCSI

[Paolo Bonzini <pbonzini@redhat.com>]

> Hi Paolo,
> 
>         I tried to assign this to you, but you don't have a JIRA id in
> OASIS it seems.

No, I hadn't realized that you need to do it separately.  I'll reply here
and send a patch next week.

> max_channel, max_target, max_lun
>
> Are comparisons here less-than or less-than-or-equal?  Both interpretations are in the fields.

It is always less-than-or-equal.  Confusion stems from Linux SCSI code
that requires less-than-or-equal for max_channel, and less-than for max_id
aka max_target and max_lun.

> Could we standardize on returning the VIRTIO_SCSI_S_TRANSPORT_FAILURE code for
> in flight commands when the device is unplugged.

Yeah, that can work.

> The spec states “all task attributes may be mapped to SIMPLE by the device”
>  - this makes S_ORDERED, S_HEAD_OF_QUEUE or S_ACA commands impossible to issue

Yes, that's explicitly allowed by the SCSI spec.  These task attributes
are just suggestions (in the case of ACA, nobody implements it anyway).
ORDERED is the same as the old virtio-blk barrier, and you can understand
why it's a good idea to ignore it.

>  - plus some commands have implicit head of queue attribute conflicting with this statement.

The device will still provide implicit head of queue semantics for
commands that require that, even if the task attribute is SIMPLE.
S_HEAD_OF_QUEUE is for commands that do not have implicit head
of queue semantics, and the device is free to ignore that.

> VIRTIO_SCSI_S_OVERRUN - could the length referred to be clarified - the virtio buffers or the allocation_length in the CBD?

It means that the allocation length or parameter length in the CDB is greater
than (respectively) the size of the dataout/datain virtio buffers.

> VIRTIO_SCSI_T_TMF_LOGICAL_UNIT_RESET - what is the expected behavior when commands are in flight?

Commands that are in flight will either be completed, or will return
VIRTIO_SCSI_S_RESET.

> What about an incomplete TMF_ABORT?

Since LUN reset will also abort the command, the TMF_ABORT will succeed.
It is undefined whether the commands will return VIRTIO_SCSI_S_RESET or
VIRTIO_SCSI_S_ABORTED.

Paolo