Re: [Qemu-devel] [PATCH] virtio-serial: PCI device for simple host <-> guest communication

[Amit Shah <amit.shah@redhat.com>]

On (Thu) Jun 11 2009 [19:07:26], Blue Swirl wrote:
> On 6/11/09, Amit Shah <amit.shah@redhat.com> wrote:
> > On (Tue) Jun 09 2009 [13:56:57], Anthony Liguori wrote:
> >  > Richard W.M. Jones wrote:
> >  >> On Tue, Jun 09, 2009 at 10:12:48PM +0530, Amit Shah wrote:
> >  >> [...]
> >  >>
> >  >> Are you going to post the virtio driver for the Linux kernel later?
> >  >>
> >  >
> >  > Yeah, I'd like to see the Linux driver before drawing conclusions.
> >
> >  Pasting here a slightly older version of the patch. I'm actively working
> >  on it; a few things might have changed but the basic idea remains the
> >  same.
> 
> >  + * along with this program; if not, write to the Free Software
> >  + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
> 
> Old FSF address, the new one is on Franklin St.

OK, will update.

> >  +struct virtio_serial_port {
> 
> >  +       unsigned int read_len;
> 
> This limits the maximum transfer length to 4G, which is way too short
> on 64 bit hosts. ;-)

This is only valid for a single packet that's transmitted via the queue
at a time and I doubt anyone's going to use bigger packets :-)

(That size is also limited by the kernel -- and the current code only
alloctes one PAGE_SIZE buffer for the transfer.)

> >  +struct virtio_serial_config {
> >  +       __u8 nr_ports;
> >  +       __u16 status;
> >  +} __attribute__((packed));
> 
> The structure should be redesigned to work without packed attribute.
> status field is unaligned.

I'll see how to untangle it.

		Amit