Re: [Qemu-devel] [PATCH] hw/qxl: vaildate surface->data

[Gerd Hoffmann]

On 10/25/12 14:27, Alon Levy wrote:
> Signed-off-by: Alon Levy <address@hidden>

Looks sane at a quick glance.

But: how far we wanna take this?  Add checks to qxl for each and every
assert() guests can trigger in spice-server?  So we end up
sanity-checking everything twice long-term?

I think instead we'll need a way for spice-server to report back errors
to qxl.  So spice-server would just notify qxl and go on (or stop
processing until reset) instead of aborting.  qxl in turn will notify
the guest.

[ The alternative would be to basically move server/red_parse_qxl.c
  into the qemu codebase.  I don't think we want that because that
  would make a bunch of data structures which are spice-server internal
  today (for good reasons) a libspice-server ABI+API. ]

cheers,
  Gerd