Linux Virtualization - May 2016 - [PATCH v3 7/7] [wip] virtio-gpu: add page flip support

On Mi, 2016-05-25 at 18:37 +0200, Daniel Vetter wrote:
> On Fri, Oct 2, 2015 at 1:58 PM, Gerd Hoffmann <kraxel at redhat.com>
wrote:
> > Signed-off-by: Gerd Hoffmann <kraxel at redhat.com>
> 
> So I entirely missed this, but this isn't really how to implement
> page_flip for an atomic driver. Working on some stuff and will hack up
> a likely totally broken patch, but should be enough as guideline.
> -Daniel
Hmm, no patch in my inbox yet.  Care to send it over or give some hints
what is wrong with this?

thanks,
  Gerd

On Fri, May 27, 2016 at 09:46:03AM +0200, Gerd Hoffmann
wrote:
> On Mi, 2016-05-25 at 18:37 +0200, Daniel Vetter wrote:
> > On Fri, Oct 2, 2015 at 1:58 PM, Gerd Hoffmann <kraxel at
redhat.com> wrote:
> > > Signed-off-by: Gerd Hoffmann <kraxel at redhat.com>
> > 
> > So I entirely missed this, but this isn't really how to implement
> > page_flip for an atomic driver. Working on some stuff and will hack up
> > a likely totally broken patch, but should be enough as guideline.
> > -Daniel
> 
> Hmm, no patch in my inbox yet.  Care to send it over or give some hints
> what is wrong with this?
You should use drm_atomic_helper_page_flip as .page_flip hook instead of
rolling your own, when you have a brand-new atomic driver. If that helper
doesn't work that means you have a bug in your driver somewhere.

And indeed virtio just plugs in drm_atomic_helper_commit, which thus far
doesnt' do nonblocking commits, which means the page flip stuff won't
work. I'm working on generic nonblocking atomic to fix this up. Currently
still wip and buggy though, hence no patches.

But I'll take you up on the implied offer to help out and test ;-)
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch

On Fri, May 27, 2016 at 09:50:27AM +0200, Daniel Vetter
wrote:
> On Fri, May 27, 2016 at 09:46:03AM +0200, Gerd Hoffmann wrote:
> > On Mi, 2016-05-25 at 18:37 +0200, Daniel Vetter wrote:
> > > On Fri, Oct 2, 2015 at 1:58 PM, Gerd Hoffmann <kraxel at
redhat.com> wrote:
> > > > Signed-off-by: Gerd Hoffmann <kraxel at redhat.com>
> > > 
> > > So I entirely missed this, but this isn't really how to
implement
> > > page_flip for an atomic driver. Working on some stuff and will
hack up
> > > a likely totally broken patch, but should be enough as guideline.
> > > -Daniel
> > 
> > Hmm, no patch in my inbox yet.  Care to send it over or give some
hints
> > what is wrong with this?
> 
> You should use drm_atomic_helper_page_flip as .page_flip hook instead of
> rolling your own, when you have a brand-new atomic driver. If that helper
> doesn't work that means you have a bug in your driver somewhere.
> 
> And indeed virtio just plugs in drm_atomic_helper_commit, which thus far
> doesnt' do nonblocking commits, which means the page flip stuff
won't
> work. I'm working on generic nonblocking atomic to fix this up.
Currently
> still wip and buggy though, hence no patches.
> 
> But I'll take you up on the implied offer to help out and test ;-)
Ok, patches are now on the list and also in my fdo repo at:

https://cgit.freedesktop.org/~danvet/drm/log/

git://people.freedesktop.org/~danvet/drm stuff

Would be really awesome if you could test this on virtio. Note that the
new nonblocking helpers require that your atomic backend gets the drm
event handling right. So if there's a bug in that logic then you'll see
lots of dmesg noise about waits timing out (after 10s of waiting). From a
quick inspection it should work though.

Upshot is that once you get normal modeset to work with those new helpers,
then nonblocking should Just Work, and there's no need anymore for
additional testing. Or at least shouldn't.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch