Re: [Qemu-devel] [RFC v3] virtio-crypto specification

[Gonglei (Arei)]

Hi Stefan,

> From: Stefan Hajnoczi [mailto:address@hidden
> Sent: Thursday, April 21, 2016 11:24 PM
> Subject: Re: [RFC v3] virtio-crypto specification
> 
> On Tue, Apr 19, 2016 at 08:37:13AM +0000, Gonglei (Arei) wrote:
> > 1.6.1.1 Driver Requirements: Session operation
> > The driver MUST set the control type with
> VIRTIO_CRYPTO_CTRL_CREATE_SESSION before the request is preceded by an
> operation header when executing session creation:
> 
> Based on the text I guess you need the following to create a session:
> struct session_create {
>     struct virtio_crypto_sym_ctlhdr ctlhdr; /* OUT */
>     struct virtio_crypto_sym_session_op session_op; /* OUT */
>     struct virtio_crypto_sym_session_op_inhdr inhdr; /* IN */
> };
> 
> Please show the full C struct so the spec is clear on how to lay out the
> request.  The same applies for other operations.
> 
Yes, it's much clearer. I'll fix them in the next version. Thank you!

> > 1.6.2.1 Steps of encryption Operation
> > Both controlq and dataq virtqueue are bidirectional.
> 
> Virtqueues are always "bidirectional" so this sentence doesn't mean
> anything.  It can be dropped.

OK, my initial mean is we need to get return status etc. from each crypto 
operation
for those two queues, which is not similar with virtio-net's tx/rx virtqueue. :)

Regards,
-Gonglei