Re: [Qemu-devel] [RFC] qemu VGA endian swap low level drawing changes

[Gerd Hoffmann <kraxel@redhat.com>]

On Di, 2014-07-01 at 10:26 +0200, Alexander Graf wrote:
> On 01.07.14 10:20, Gerd Hoffmann wrote:
> > On Mo, 2014-06-30 at 22:32 +1000, Benjamin Herrenschmidt wrote:
> >> On Mon, 2014-06-30 at 13:14 +0200, Gerd Hoffmann wrote:
> >>>    Hi,
> >>>
> >>>>  From what I can tell, we only ever call the cursor drawing callback on
> >>>> non-shared surfaces. Should I deduce that the HW cursor emulation simply
> >>>> doesn't work when using shared surfaces ? Or is there another path I
> >>>> have missed to handle it ?
> >>> Hmm.  Looks like hw-cursor-on-shared-surface broken indeed.  Need to dig
> >>> out a guest which actually uses it & go figure when testing your patch
> >>> series ...
> >> I don't think I broke it much more than it already was but then I
> >> couldn't find a guest using it. I've tried the plain cirrus DDX in X and
> >> it didn't have any problem... maybe windows ?
> > Nope.  windows xp doesn't use it.  Anything newer doesn't ship with
> > cirrus drivers any more (and uses vesa bios support).
> >
> > Looking at the code the cirrus hardware cursor supports two colors only
> > (and some funky xor mode).  Guess it simply doesn't cut it as you can't
> > have your cursors drop shadows with that, so guests are ignoring it.
> 
> Windows NT 4 might use it. I remember that I had issues running NT4 with 
> Cirrus emulation a while back.

That could be it indeed.

> >> Right. A quick fix would be to add a flag to force always using a shadow
> >> surface and set it in cirrus ... I'm not sure anybody will notice the
> >> performance difference.
> > I suspect we can rip out hw cursor emulation and nobody will notice the
> > difference either ...
> 
> Very likely ;). Though I think we're better off keeping it around to 
> make sure we're still compatible with ancient guests (Windows 3.1 might 
> use it too). Making it slow however shouldn't make any difference at all.

Especially as we can make the force-shadow mode depend on the hw cursor
enable bit, so we don't have any difference for the common case.

cheers,
  Gerd