Bug 89842 - Nvidia nouveau drivers hangs all gnome when opening the new legacy tray (and many other situations) in 3.16.0
Summary: Nvidia nouveau drivers hangs all gnome when opening the new legacy tray (and ...
Status: RESOLVED FIXED
Alias: None
Product: xorg
Classification: Unclassified
Component: Driver/nouveau (show other bugs)
Version: unspecified
Hardware: x86-64 (AMD64) Linux (All)
: medium critical
Assignee: Nouveau Project
QA Contact: Xorg Project Team
URL:
Whiteboard:
Keywords:
: 90201 (view as bug list)
Depends on:
Blocks:
 
Reported: 2015-03-31 14:39 UTC by Óscar García Amor
Modified: 2015-08-06 12:13 UTC (History)
8 users (show)

See Also:
i915 platform:
i915 features:


Attachments
Nouveau with Gnome 3.16.0 crash log (70.35 KB, text/plain)
2015-03-31 14:39 UTC, Óscar García Amor
no flags Details
Attaching journalctl output after gnome 3.16 freeze. Freeze happened @ ~ 23:06 (596.32 KB, text/plain)
2015-04-02 15:53 UTC, Graham
no flags Details
jounralctl events when gnome-shell freezes (113.94 KB, text/plain)
2015-05-11 11:39 UTC, Joe Verreau
no flags Details
when manipulating Gnome tray (96.65 KB, text/plain)
2015-06-28 18:19 UTC, Victor Porton
no flags Details

Description Óscar García Amor 2015-03-31 14:39:35 UTC
Created attachment 114766 [details]
Nouveau with Gnome 3.16.0 crash log

Using nouveau drivers, when plays with the new legacy tray in Gnome 3.16.0 (open it, close, and reopen again) the entire system hangs and must to restart Gnome Shell & GDM.

Card: GeForce GTS 250

Versions:
Kernel 3.19.3-1-ARCH
xf86-video-nouveau 1.0.11-3
mesa 10.5.2-1

I attach journal log.

Note: I open related bug in GNOME https://bugzilla.gnome.org/show_bug.cgi?id=747115 They advised me that open bug here too.
Comment 1 Graham 2015-04-02 15:53:26 UTC
Created attachment 114831 [details]
Attaching journalctl output after gnome 3.16 freeze. Freeze happened @ ~ 23:06
Comment 2 Graham 2015-04-02 15:55:43 UTC
The gnome team thinks I may be hitting the same issue

NVIDIA Corporation GF119 [GeForce GT 610] (rev a1)

Versions:
Kernel 3.19.3-1-ARCH
xf86-video-nouveau 1.0.11-3
mesa 10.5.2-1

Attached my journal log
Comment 3 Ilia Mirkin 2015-04-02 16:19:48 UTC
Somehow gnome-shell is able to convince nouveau to do something very dumb. I didn't even think this was possible... libdrm is supposed to de-dup these, no?

nouveau E[gnome-shell[1773]] multiple instances of buffer 215 on validation list
nouveau E[gnome-shell[1773]] validate_init
nouveau E[gnome-shell[1773]] validate: -22
Comment 4 Arjen 2015-04-03 09:35:12 UTC
I'm also getting the same errors since april 1st:

Apr 01 11:07:08 arjen-imac.office.react.nl kernel: nouveau E[gnome-shell[4997]] multiple instances of buffer 228 on validation list
Apr 01 11:36:47 arjen-imac.office.react.nl kernel: nouveau E[gnome-shell[905]] multiple instances of buffer 255 on validation list
Apr 01 13:51:37 arjen-imac.office.react.nl kernel: nouveau E[gnome-shell[2939]] multiple instances of buffer 146 on validation list
Apr 02 12:20:26 arjen-imac.office.react.nl kernel: nouveau E[gnome-shell[2939]] multiple instances of buffer 415 on validation list
Apr 02 17:00:46 arjen-imac.office.react.nl kernel: nouveau E[gnome-shell[895]] multiple instances of buffer 327 on validation list


Just before the 1st crash I upgrade mesa:

[2015-04-01 09:55] [ALPM] upgraded mesa (10.5.1-2 -> 10.5.2-1)


Versions:
Kernel 3.19.2-1-ARCH
xf86-video-nouveau 1.0.11-3
mesa 10.5.2-1
Gnome 3.14.2

So I think this is mesa related, and not related to Gnome 3.16.
Comment 5 Ilia Mirkin 2015-04-05 19:50:35 UTC
I'm guessing all you guys have libdrm-2.4.60 -- can you try downgrading to libdrm-2.4.59?
Comment 6 Rennie 2015-04-11 00:50:15 UTC
I have the same issue and can confirm that downgrading
from libdrm-2.4.60 to libdrm-2.4.59 seems to stop the issue from
happening as there are no more hangs.

Thanks,
Rennie
Comment 7 Sean Bogie 2015-04-26 16:44:29 UTC
git bisect puts the first bad commit @
commit 5ea6f1c32628887c9df0c53bc8c199eb12633fec
Author: Maarten Lankhorst <maarten.lankhorst@ubuntu.com>
Date:   Thu Feb 26 11:54:03 2015 +0100

    nouveau: make nouveau importing global buffers completely thread-safe, with tests
...


ArchLinux bug report (https://bugs.archlinux.org/task/44680) suggests an additional reproduction method "when I move my mouse over VLC's seekbar and it shows a small tooltip to show the time gnome-shell freezes".
Comment 8 Ilia Mirkin 2015-04-26 16:48:08 UTC
My favourite is "run mplayer with vdpau, then move the window". I arrived at that one by accident, but that repros it 100%. No compositors or anything like that.
Comment 9 Ilia Mirkin 2015-04-27 23:22:39 UTC
*** Bug 90201 has been marked as a duplicate of this bug. ***
Comment 10 Ben Skeggs 2015-05-06 04:52:54 UTC
I just push a commit[1] to libdrm which should fix this issue.

[1] http://cgit.freedesktop.org/mesa/drm/commit/?id=812e8fe6ce46d733c30207ee26c788c61f546294
Comment 11 Ilia Mirkin 2015-05-06 16:03:24 UTC
(In reply to Ben Skeggs from comment #10)
> I just push a commit[1] to libdrm which should fix this issue.
> 
> [1]
> http://cgit.freedesktop.org/mesa/drm/commit/
> ?id=812e8fe6ce46d733c30207ee26c788c61f546294

I can confirm that this fixes my repro case (move mplayer vdpau window around). I knew it was something relating to named bo's, so good to see that the fix also involved those.
Comment 12 Joe Verreau 2015-05-11 11:39:25 UTC
Created attachment 115697 [details]
jounralctl events when gnome-shell freezes

Gnome version is 3.14.4-2-fc21.  I had reported this event to the gnome team #749128; they referred me here.  I had initiated a download in firefox when this freeze occurred but i have experienced it in other applications
Comment 13 Joe Verreau 2015-05-11 11:54:31 UTC
In closer review of the thread above I checked on downgrading libdrm from 2.4.60.  In my installation yum tells me I need to also downgrade libdrm-devel and apparently the downgrade version is 2.4.58 rather than 2.4.59.  Is that what you recommend?
Comment 14 Matthew Miller 2015-05-11 12:53:38 UTC
(In reply to Joe Verreau from comment #13)
> In closer review of the thread above I checked on downgrading libdrm from
> 2.4.60.  In my installation yum tells me I need to also downgrade
> libdrm-devel and apparently the downgrade version is 2.4.58 rather than
> 2.4.59.  Is that what you recommend?

Joe, are you on the Fedora 22 bet? If so, an update will be going out soon. You can get it immediately from https://admin.fedoraproject.org/updates/FEDORA-2015-7930/libdrm-2.4.61-3.fc22

This is version 2.4.61, which fixes the regression. If you're not on F22, I expect updates will be coming soon. In the meantime, go ahead and downgrade to whatever works.
Comment 15 Joe Verreau 2015-05-11 13:03:15 UTC
Matthew, actually I'm on fedora 21 so I will downgrade libdrm, libdrm-devel to 2.4.58 and await the update to 2.4.61 in the normal distribution.  thanks.
Comment 16 Rich Barlow 2015-06-03 09:51:52 UTC
I too am experiencing the issue described in this ticket. It is affecting 6 machines all running Fedora 21. They generally hang around 2-3 times a day. Is there going to be an update to 2.4.61 pushed for Fedora 21 at some point?
Comment 17 Victor Porton 2015-06-28 15:07:14 UTC
I have a similar bug with

Debian Linux "testing" ("stretch")
GeForce 8400 GS Rev. 3
Linux 3.10-2-amd64
xserver-xorg-video-nouveau 1:1.0.11-1+b1
libdrm-nouveau1a 2.4.40-1~deb7u2
libdrm-nouveau2 2.4.60-3
libgl1-mesa-dri 10.5.7-1
libgl1-mesa-glx 10.5.7-1
Gnome 3.14.0-1
Comment 18 Victor Porton 2015-06-28 18:19:02 UTC
Created attachment 116767 [details]
when manipulating Gnome tray

See the error log produced by journalctl when manipulating Gnome tray.
Comment 19 Ilia Mirkin 2015-06-28 18:21:36 UTC
This is fixed by not using libdrm 2.4.60 which was a buggy release on the nouveau end. libdrm 2.4.59 or libdrm 2.4.61 should work fine.
Comment 20 Joe Verreau 2015-08-06 12:13:20 UTC
... at this point I'm inferring there will not be upgraded versions of libdrm, libdrm-devel for fc21?  I did downgrade my laptop from 2.4.60 to 2.4.58 in Jun as noted below.  I'm guessing the fix really is to go to fc22.  I ask because now my desktop is also experiencing these freeze ups tho not in the frequency that others have reported.


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.