Bug 67064 - [sna] various corruption on screen
Summary: [sna] various corruption on screen
Status: VERIFIED FIXED
Alias: None
Product: xorg
Classification: Unclassified
Component: Driver/intel (show other bugs)
Version: git
Hardware: Other All
: medium normal
Assignee: Chris Wilson
QA Contact: Intel GFX Bugs mailing list
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2013-07-19 06:43 UTC by Jiri Slaby
Modified: 2013-07-21 14:19 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments
example of corruption (5.46 KB, image/png)
2013-07-19 06:43 UTC, Jiri Slaby
no flags Details
xorg.log (38.94 KB, text/plain)
2013-07-19 06:44 UTC, Jiri Slaby
no flags Details
another example (3.52 KB, image/png)
2013-07-19 06:47 UTC, Jiri Slaby
no flags Details
another artifact (3.41 KB, image/png)
2013-07-19 19:26 UTC, Jiri Slaby
no flags Details

Description Jiri Slaby 2013-07-19 06:43:38 UTC
Created attachment 82656 [details]
example of corruption

After the last update to 2.21.12-12-g6c8b15d I started seeing random image corruption on screen like vertical or horizontal lines. See the picture for illustration, there are two of the corruptions.
Comment 1 Jiri Slaby 2013-07-19 06:44:02 UTC
Created attachment 82657 [details]
xorg.log
Comment 2 Jiri Slaby 2013-07-19 06:47:19 UTC
Created attachment 82658 [details]
another example
Comment 3 Chris Wilson 2013-07-19 08:24:39 UTC
Do you know which commit you used last? Looks like a dup of bug 67015 or bug 67045, both of which I am unsure about yet.
Comment 4 Chris Wilson 2013-07-19 08:30:39 UTC
s/bug 67045/bug 67055/
Comment 5 Jiri Slaby 2013-07-19 08:32:55 UTC
Git reflog says:

6c8b15d HEAD@{0}: rebase finished: returning to refs/heads/master
6c8b15d HEAD@{1}: checkout: moving from master to 6c8b15d321044d4a81cb187cc5e1ac094eb82367^0
34c9b75 HEAD@{2}: rebase finished: returning to refs/heads/master
34c9b75 HEAD@{3}: checkout: moving from master to 34c9b759fbab8d548108e954d55de38c6f5bec31^0
884dd89 HEAD@{4}: rebase finished: returning to refs/heads/master
884dd89 HEAD@{5}: checkout: moving from master to 884dd896e4e959e5a284a6f70da9e5cde1bbb7e9^0
dbb585b HEAD@{6}: rebase finished: returning to refs/heads/master

34c9b759fbab8d548108e954d55de38c6f5bec31 already exposed that bug I think. I'm not sure about 884dd896e4e959e5a284a6f70da9e5cde1bbb7e9... It's not so easily reproducible to test :/.
Comment 6 Jiri Slaby 2013-07-19 08:36:01 UTC
I haven't noticed anything that is described in the two bugs above.
Comment 7 Jiri Slaby 2013-07-19 08:39:53 UTC
In the meantime I also updated the kernel from 3.9 to 3.10. So it might be relevant too.
Comment 8 Chris Wilson 2013-07-19 09:22:35 UTC
Essentially, you are doing your daily tasks under KDE until corruption shows up and spoils the party? Is the corruption gradual, or spontaneous? Is it transient?
Comment 9 Jiri Slaby 2013-07-19 09:34:14 UTC
(In reply to comment #8)
> Essentially, you are doing your daily tasks under KDE until corruption shows
> up and spoils the party?

Yes.

> Is the corruption gradual, or spontaneous? Is it transient?

What can be seen on the first image is spontaneous, I do not see it anywhere now. But since I saw what is on picture 2, I see now those thick lines on most my window decorations.
Comment 10 Daniel Vetter 2013-07-19 09:39:50 UTC
Could be the fence bug, at least if you've throw in a resume once a while. See

https://patchwork.kernel.org/patch/2828671/
Comment 11 Chris Wilson 2013-07-19 13:33:51 UTC
Also 3.10.1 doesn't have the read-read/read-write fixes yet, which could be related.
Comment 12 Jiri Slaby 2013-07-19 19:26:02 UTC
Created attachment 82704 [details]
another artifact

(In reply to comment #10)
> Could be the fence bug, at least if you've throw in a resume once a while.
> See
> 
> https://patchwork.kernel.org/patch/2828671/

I built a 3.10.1 kernel with that patch, but this one on its own does not help entirely. I still see the artifact attached. It should be the psi's star, the orange star at http://psi-im.org/. Nothing more than that happened yet.

(In reply to comment #11)
> Also 3.10.1 doesn't have the read-read/read-write fixes yet, which could be
> related.

Do you have SHAs at hand so that I can test?
Comment 13 Chris Wilson 2013-07-19 19:38:35 UTC
commit daa13e1ca587bc773c1aae415ed1af6554117bd4
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Fri Jun 28 16:54:08 2013 +0100

    drm/i915: Only clear write-domains after a successful wait-seqno


commit 02978ff57a5bdfbf703d2bc5a4d933a53ede3144
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Tue Jul 9 09:22:39 2013 +0100

    drm/i915: Fix write-read race with multiple rings
Comment 14 Jiri Slaby 2013-07-20 11:15:04 UTC
(In reply to comment #13)
> commit daa13e1ca587bc773c1aae415ed1af6554117bd4

Those 3 and 2.21.12-29-g65a0943 and I am still seeing what is depicted in attachment 82704 [details].

I haven't seen the previous corruptions yet.

Maybe this is a different bug? For me it looks like it appeared with one of the updates in the past few days...
Comment 15 Chris Wilson 2013-07-20 21:05:12 UTC
I think I have that systray corruption resolved:

commit 2c78403cb56841f342692b97bb53e995c3ff6b9f
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Sat Jul 20 21:55:16 2013 +0100

    sna: Remember to apply drawable offsets for composite memcpy
    
    Yet another missing chunk from
    commit 6921abd81017c9ed7f3b2413784068fbc609a0ea
    Author: Chris Wilson <chris@chris-wilson.co.uk>
    Date:   Thu Jul 18 16:21:27 2013 +0100
    
        sna: Add a fast path for the most common fallback for CPU-CPU blits
    
    I had the composite offsets, but not the normal window offsets!
    
    Reported-by: F.Brown <francisbrwn9@gmail.com>
    Reported-by: Jiri Slaby <jirislaby@gmail.com>
    Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=67124
    Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=67064
    Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Comment 16 Jiri Slaby 2013-07-21 14:16:23 UTC
Yeah, it works...


Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.