Bug 14784

Created attachment 14772 [details]
Working startx log

Straight capture of output from startx on a successful start (a failed start produces no file)

Created attachment 14773 [details]
Working Xorg log

From a working Xorg log, again no file is built on freezing startx.

Running:

Gentoo-2.6.24 (32bit)
Xorg-Server 1.4.0.90-r3
Intel Version 2.2.1
Mesa 7.0.2
DRM 1.1.0 (20060810) (from dmesg)

Hardware:
Intel e8400
4GB RAM
Onboard Intel G35 Video


-Chad

In trying to run ABET tests (http://www.intellinuxgraphics.org/testing.html) the startx (If I don't run startx first) locks up the system.  I'm happy to enable debug and post the output, I'm just not sure which debug should be turned on :)

-Chad

(In reply to comment #4)
> In trying to run ABET tests (http://www.intellinuxgraphics.org/testing.html)
> the startx (If I don't run startx first) locks up the system.  I'm happy to
> enable debug and post the output, I'm just not sure which debug should be
> turned on :)
> 
> -Chad
> 

Heh heh, that's supposed to be ABAT. ;)
And just to clarify, that is with i915 loaded.  Further investigation appears to show that it only fails on subsequent startx launches (the first one is usually successful).

This sounds like Bug #14591.

(In reply to comment #6)
> This sounds like Bug #14591.
> 

Indeed it somewhat does.  However, I'm now having a somewhat more consistent way of starting X that makes it even stranger.  If I launch:
startx &> file.log

Then I can get it start (once).  Otherwise, so far, any standard "startx" hangs the system, and subsequent starts with or without the redirect hangs the system. 

-Chad

(In reply to comment #7)
> (In reply to comment #6)
> > This sounds like Bug #14591.
> > 
> 
> Indeed it somewhat does.  However, I'm now having a somewhat more consistent
> way of starting X that makes it even stranger.  If I launch:
> startx &> file.log
> 
> Then I can get it start (once).  Otherwise, so far, any standard "startx" hangs
> the system, and subsequent starts with or without the redirect hangs the
> system. 
> 
> -Chad
> 

Scratch that.  It isn't consistent.  I was just able to launch X with "startx" without any redirects.  It was the first attempt after the initial boot, that seems to be the only consistency; once only, if at all.

-Chad

Another update, this doesn't seem to happen (ever, regardless of how many times I kill and restart X) when I roll back to 2.2.0.90 (using the Gentoo ebuild).

-Chad

I bet this commit fixes the issue.

commit 02a3b80ee8341968dbce5b7cb871b187db5ff805
Author: Zhenyu Wang <zhenyu.z.wang@intel.com>
Date:   Fri Feb 29 11:13:34 2008 +0800

    Bug 14591: comment out MI_DISPLAY_POWER_DOWN as for 965GM only

Update to current git head.

>> When I launch startx from the command line (as root) if the i915 kernel drm
>> module isn't loaded the machine "hard" locks 

Is "without i915 module loaded" required for reproducing the lock?

Can you try the latest git code? The patch in comment#10 is committed after 2.2.1.

(In reply to comment #5)
> Heh heh, that's supposed to be ABAT. ;)

btw, I just updated ABAT, which is supposed to recover to text console if you run it before startx.

(In reply to comment #11)
> >> When I launch startx from the command line (as root) if the i915 kernel drm
> >> module isn't loaded the machine "hard" locks 
> 
> Is "without i915 module loaded" required for reproducing the lock?
> 
> Can you try the latest git code? The patch in comment#10 is committed after
> 2.2.1.
> 

Will do, thanks for the info.  I'll test tonight and report back.

-Chad

(In reply to comment #10)
> I bet this commit fixes the issue.
> 
> commit 02a3b80ee8341968dbce5b7cb871b187db5ff805
> Author: Zhenyu Wang <zhenyu.z.wang@intel.com>
> Date:   Fri Feb 29 11:13:34 2008 +0800
> 
>     Bug 14591: comment out MI_DISPLAY_POWER_DOWN as for 965GM only
> 
> Update to current git head.
> 

Indeed, that seems to have fixed it!  Much appreciated.

(In reply to comment #11)
> >> When I launch startx from the command line (as root) if the i915 kernel drm
> >> module isn't loaded the machine "hard" locks 
> 
> Is "without i915 module loaded" required for reproducing the lock?
> 
> Can you try the latest git code? The patch in comment#10 is committed after
> 2.2.1.
> 

No it wasn't, it just appeared as the first symptom while doing several tests.  It would like with or without i915 loaded.

*** This bug has been marked as a duplicate of bug 14591 ***