Bug 27099

Summary: Compiz (+everything) hang after VT switch
Product: Mesa Reporter: North West <pi3orama>
Component: Drivers/DRI/i915Assignee: Jesse Barnes <jbarnes>
Status: RESOLVED FIXED QA Contact:
Severity: normal    
Priority: medium CC: remi
Version: unspecifiedKeywords: NEEDINFO
Hardware: Other   
OS: All   
Whiteboard:
i915 platform: i915 features:
Attachments: dmesg > /tmp/dmesg
my /etc/X11/xorg.conf
/var/log/Xorg.0.log

Description North West 2010-03-15 16:09:23 UTC
I report a MTRR bug in gentoo bugzilla, and Rémi Cardona suggests me to open a new bug here.

http://bugs.gentoo.org/show_bug.cgi?id=308459

Another problem is if set i915.modeset to 0, compiz-fusion can get only a full white screen with a movable mouse. I don't know whether the 2 problems are related.
Comment 1 Rémi Cardona 2010-03-15 23:58:46 UTC
Please attach your Xorg.0.log, your demsg and your xorg.conf.

Thanks
Comment 2 North West 2010-03-16 00:38:06 UTC
Created attachment 34098 [details]
dmesg > /tmp/dmesg
Comment 3 North West 2010-03-16 00:38:36 UTC
Created attachment 34099 [details]
my /etc/X11/xorg.conf
Comment 4 North West 2010-03-16 00:39:00 UTC
Created attachment 34100 [details]
/var/log/Xorg.0.log
Comment 5 Eric Anholt 2010-03-16 10:01:05 UTC
Disabling KMS is not supported, so the MTRR complaint is a non-issue.  For hangs, you need to be running the current release of our driver for us to help.  Also, please include in this bug report how to reproduce the hang.
Comment 6 North West 2010-03-16 18:43:24 UTC
I've list the steps of how to reproduce a hang in

http://bugs.gentoo.org/show_bug.cgi?id=308459

Now I copy the text here:

After "startx" from a console (say, console 1), everything works fine. However,
if switch back to the console (console 1, using C-M-F1) and press enter,
programs in X11 server stop (I know they stop before switching back because
rhythmbox stops to playback music after I press enter). When switch back
(C-M-F7), all applications hang, left blank windows on the screen. But mouse
can move, keyboard shortcut of C-M-F? and C-M-Backspace still work.

1. reboot system using init 3
2. execute 'startx' in console 1
3. run a graphical music playback and play a song
4. press C-M-F1 to switch back to console 1
5. press Enter, Enter, Enter ...
6. Switch back using C-M-F7

In step 5, music playback stops

In step 6, every X11 programs hang, but mouse still movable, C-M-F? and
C-M-Backspace still work
Comment 7 Rémi Cardona 2010-03-17 00:03:04 UTC
(In reply to comment #6)
> I've list the steps of how to reproduce a hang in
> 
> http://bugs.gentoo.org/show_bug.cgi?id=308459
> 
> Now I copy the text here:
> 
> After "startx" from a console (say, console 1), everything works fine. However,
> if switch back to the console (console 1, using C-M-F1) and press enter,
> programs in X11 server stop (I know they stop before switching back because
> rhythmbox stops to playback music after I press enter). When switch back
> (C-M-F7), all applications hang, left blank windows on the screen. But mouse
> can move, keyboard shortcut of C-M-F? and C-M-Backspace still work.
> 
> 1. reboot system using init 3
> 2. execute 'startx' in console 1
> 3. run a graphical music playback and play a song
> 4. press C-M-F1 to switch back to console 1
> 5. press Enter, Enter, Enter ...
> 6. Switch back using C-M-F7
> 
> In step 5, music playback stops
> 
> In step 6, every X11 programs hang, but mouse still movable, C-M-F? and
> C-M-Backspace still work
> 

You'll also need to try git versions of libdrm, xf86-video-intel and the kernel. You can use -9999 ebuilds from the x11 overlay for that by unmasking them.

As for the kernel, you should try the latest sys-kernel/git-sources from ~arch.

Thanks
Comment 8 North West 2010-03-17 00:22:30 UTC
I have tried the newest version of kernel (~x86 gentoo-source) and libdrm (~x86 2.4.18) and the hang bug still exist. I will try git-sources later.

Can you reproduce the hang bug using my steps? I can reproduce the hang bug in a NVIDIA card laptop with NVIDIA driver, so I believe the hang bug is unrelated to xf86-video-intel.
Comment 9 Chris Wilson 2010-07-24 04:40:08 UTC
As far as -intel is concerned this feels like one of the many page-flipping across VT switch that Jesse has fixed. Can you please retest with current bits?
Comment 10 North West 2010-07-24 04:54:32 UTC
Currently I start X by a script in my home directory (~/bin/startx):
---------------------------
#!/bin/bash

xserverauthfile=~/.serverauth
nohup /usr/bin/startx < /dev/null > /tmp/xorg.output 2>&1
------------------------------

Now I can switch to VT1 and do some works.

But problem still exist after more than 4 months.

Below is the current version of related packages:

x11-base/xorg-server-1.7.6
x11-libs/libdrm-2.4.20-r1
x11-drivers/xf86-video-intel-2.9.1
Comment 11 Jesse Barnes 2010-07-24 13:25:53 UTC
On Sat, 24 Jul 2010 04:54:32 -0700 (PDT)
bugzilla-daemon@freedesktop.org wrote:

> https://bugs.freedesktop.org/show_bug.cgi?id=27099
> 
> --- Comment #10 from North West <pi3orama@gmail.com> 2010-07-24 04:54:32 PDT ---
> Currently I start X by a script in my home directory (~/bin/startx):
> ---------------------------
> #!/bin/bash
> 
> xserverauthfile=~/.serverauth
> nohup /usr/bin/startx < /dev/null > /tmp/xorg.output 2>&1
> ------------------------------
> 
> Now I can switch to VT1 and do some works.
> 
> But problem still exist after more than 4 months.
> 
> Below is the current version of related packages:
> 
> x11-base/xorg-server-1.7.6
> x11-libs/libdrm-2.4.20-r1
> x11-drivers/xf86-video-intel-2.9.1
> 

Fix for this was in the X server, please try the X server from git or
from xorg edgers if you're on Ubuntu.
Comment 12 Jesse Barnes 2010-07-26 11:53:03 UTC
Pretty sure this was fixed by these two commits in xserver master, please re-open if not.

commit d75e8146c414bfd512ba5dbd4a83acb334bbe19b
Author: Keith Packard <keithp@keithp.com>
Date:   Mon Jul 12 16:01:34 2010 -0700

    Unwrap/rewrap EnterVT/LeaveVT completely, Fixes 28998

commit 28e33ae6f69f716ece5d68e63fc52557236c5f6e
Author: Jesse Barnes <jbarnes@virtuousgeek.org>
Date:   Wed Jun 30 07:59:04 2010 -0700

    OS support: fix writeable client vs IgnoreClient behavior

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.