Bug 56423

This is actually a log from SNA from a suse RPM. I wanted to give it a try before reporting this and it doesn't work the same as UXA.

Seems to work with 2.20.12-15-g40bda54.

(In reply to comment #2)
> Seems to work with 2.20.12-15-g40bda54.

(sna)

Annoying, this is the symptom that we tried to fix in 2.20.10. What's your DE setup, in particular are you using any form of compositor?

(In reply to comment #4)
> Annoying, this is the symptom that we tried to fix in 2.20.10. What's your
> DE setup, in particular are you using any form of compositor?

Yes, but suse provides 2.20.3 which is older. I'm going to tell the fellows to update the pkg. Current git works for me...

I'll close this for now presuming that it is the earlier bug. If the problem persists with an updated SuSE package, please let me know. Thanks.

Just correcting myself, 2.20.9 had the bugfix for the blank/stuck screen in UXA across resume (and DPMS).

(In reply to comment #7)
> Just correcting myself, 2.20.9 had the bugfix for the blank/stuck screen in
> UXA across resume (and DPMS).

The guys ask me to bisect that to know what patch fixes it. Don't you remember?

[I'm too lazy to bisect now, so I won't do that and leave suse 12.2 broken until I'm back from all the conferences.]

edc5b7f741a4bb8e9a58b5eb2201fa9304f94ef7, 5a45cbacb777e478d8fbda9223b0fb5c705d7249,
feef53005b57d6b4489a15f8d0db8245dcb0cfaa,
cb0b72c4fbb7b1bc941522d8679fef856f2f8936,
88cfd23379950d1fe4e682519c48ef6e3091c2f3

This does not seem to be solved completely. It's pretty rare with the commits, but it still happens occasionally.

(In reply to comment #10)
> This does not seem to be solved completely. It's pretty rare with the
> commits, but it still happens occasionally.

With SNA -- is this is a different bug?

Try using the repository
http://download.opensuse.org/repositories/X11:/XOrg/openSUSE_12.2/
for opensuse 12.2.

(In reply to comment #11)
> (In reply to comment #10)
> > This does not seem to be solved completely. It's pretty rare with the
> > commits, but it still happens occasionally.
> 
> With SNA -- is this is a different bug?

Yes. SNA did not have the same race as UXA, but there were a few bugs in the kernel as well that could result in very similar failures but were much harder to trigger than the userspace race. The most recent of which was

commit 5bb61643f6a70d48de9cfe91ad0fee0d618b6816
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Thu Sep 27 21:25:58 2012 +0100

    drm/i915: Flush the pending flips on the CRTC before modification

(In reply to comment #13)
> commit 5bb61643f6a70d48de9cfe91ad0fee0d618b6816
> Author: Chris Wilson <chris@chris-wilson.co.uk>
> Date:   Thu Sep 27 21:25:58 2012 +0100
> 
>     drm/i915: Flush the pending flips on the CRTC before modification

Sadly, this won't help me as I'm running on 3.6.3 and this is in there already.

And the issue is not that very rare. It is rather pretty frequent. Like every second or third suspend results in this.

Across suspend, there is a known issue that we don't flush all the pending work... So I have no doubt that we have some residual bugs here. :|

Anyway, the UXA case seems to work fine now. Should I create a new bug for SNA? Is this something for Daniel or for you?

I'm scratching my head on this one. I definitely also see an effect whereby only the pointer is visible after DPMS until a redraw. But looking at the debug log suggests no foul play.

I'm also chasing a bug common to UXA and SNA where gnome-shell appears to randomly freeze when starting an OpenGL client (e.g. openarena) where all that is visible is the pointer against a black screen. And another bug in gnome-shell were it freezes after xrandr --off ; xrandr --auto. All signs point towards g-s; rather I have not yet found a reason within the ddx for gnome-shell to block.

Back to this bug... Jiri are you using gnome-shell?

(In reply to comment #18)
> Back to this bug... Jiri are you using gnome-shell?

Negative. I use KDE4 with opengl compositing.

No worries, a bit of tracing later and I can see that a SwapCompletion event is going missing...

Looks like this should fix it (or at least my scenarios it does):

commit 9ab1d1f94e502e5fde87e7c171f3502f8a55f22b
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Tue Nov 20 18:42:58 2012 +0000

    sna/dri: Queue a vblank-continuation after flip-completion
    
    If a vblank request was delayed due to a pending flip, we need to make
    sure that we then queue it after that flip or else progress ceases.
    
    Reported-by: Jiri Slaby <jirislaby@gmail.com>
    Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=56423
    Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=57156
    Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

(In reply to comment #21)
> Looks like this should fix it (or at least my scenarios it does):

Yes, it looks good so far.