[PATCH xserver 3/3] ramdac: Handle master and slave cursors independently

[Alex Goins]

On Fri, 10 Nov 2017, Michel Dänzer wrote:

> On 09/11/17 08:09 PM, Alex Goins wrote:
> >> On 09/11/17 02:34 AM, Alex Goins wrote:
> >>> On Wed, 8 Nov 2017, Michel Dänzer wrote:
> > 
> >> I'm asking this to avoid ending up in a similar situation as with the
> >> PRIME synchronization functionality, where it seems like the modesetting
> >> driver still can't use it with itself as the peer.
> > 
> > The modesetting driver has always been able to use PRIME Sync with itself as the
> > peer. See changes b83dede9 and 80e64dae, source and sink implementations,
> > respectively. Both went in at the same time.
> 
> I based the above on
> https://bugs.freedesktop.org/show_bug.cgi?id=103202#c8 (see the attached
> log file). Maybe I'm misunderstanding something.

Reverse PRIME does not support PRIME Sync, causing the fallback. Reverse PRIME
requires more implementation due to the additional copy to vidmem before
presentation. If the user used Nouveau as the master and i915 as the slave, both
with modesetting, it would work without the fallback. It's not a case of missing
feature parity between modesetting and the NVIDIA driver.

Implementing PRIME Sync source support for modesetting wasn't too much of an
additional challenge, but I suspect that implementing cursor composition into
the shared pixmap in a synchronized fashion will require a significant amount of
additional infrastructure in the modesetting driver. The NVIDIA driver already
had this functionality, it just got masked by change 7b634067.

Requiring cursor composition in the modesetting driver to get this fixed would
be unfortunate, as I'm simply trying to fix an upstream functional regression
that affects DisplayLink + NVIDIA users while maintaining the new functionality,
rather than implementing something completely new. All the other feedback is
doable in a short timeframe, but this would likely delay the solution
significantly.

> P.S. Did you see https://bugs.freedesktop.org/show_bug.cgi?id=101998#c37 ?

I see it, should be easy enough to fix, just the configurations I usually test
are either NVIDIA->modesetting or modesetting->modesetting.

Thanks,
Alex