Bug 2266 - Xorg freezes when exiting session to a login manager.
Summary: Xorg freezes when exiting session to a login manager.
Status: RESOLVED FIXED
Alias: None
Product: DRI
Classification: Unclassified
Component: DRM/other (show other bugs)
Version: XOrg git
Hardware: x86 (IA32) Linux (All)
: high major
Assignee: Default DRI bug account
QA Contact:
URL: http://forums.gentoo.org/viewtopic.ph...
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2005-01-11 15:52 UTC by Antoni Grzymała
Modified: 2005-09-27 22:16 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments
My xorg.conf (2.08 KB, text/plain)
2005-01-11 15:54 UTC, Antoni Grzymała
no flags Details

Description Antoni Grzymała 2005-01-11 15:52:22 UTC
I've been having this problem since I moved from xfree to xorg:

I have a Gentoo system running with an 8MB ATI Rage128 (Mobility M3 AGPx2) on a
Dell Inspiron 4000 (Intel 440BX mainboard). My current xorg version is 6.8.0-r3,
but the problem has existed with all previously had ones. Same thing with the
kernel, currently it is at 2.6.10-gentoo-r2. I'm running the r128 driver
(obviously).

When I have DRI *enabled* and I exit from an X session to a login manager I get
a total system freeze. I tried both in-kernel DRI (2.6.{7,8,9,10}) and the
gentoo portage x11-drm package, all versions working with kernel 2.6, namely:
20040827, 20050104.

I don't get the freeze when I'm not starting X with a login manager. When I have
accelerated X started up with startx as a normal user, I can exit the session
without a problem.

This happens on any combination of WM / Login Manager (tried xdm+fluxbox,
xdm+gnome, xdm+fvwm, gdm+gnome, gdm+fvwm, gdm+fluxbox, gdm+kde, kdm+kde).

On my config to enable or disable DRI I only have to change my default bit depth
(on 16 bit I get DRI, on 24 I don't). This may or may not have anything to do
with the freezes. The rest of xorg.conf stays intact.

The problem did not appear on XFree (with identical XF86Config).

Searched the web and groups to no avail. Any help will be much appreciated.

I thought about filing a bug report, but I'm not even really sure, which exactly
is the failing component.

Excuse my convoluted sentences - English is not my native language :).
Comment 1 Antoni Grzymała 2005-01-11 15:54:05 UTC
Created attachment 1667 [details]
My xorg.conf

I'm attaching my xorg.conf for completness. There's nothing fancy in it.
Comment 2 Antoni Grzymała 2005-01-11 16:31:23 UTC
A suggestion that this bug may be related to using the text console framebuffer
lead me to removing all framebuffer code from the kernel leaving me with a plain
VGA console. Doesn't seem to solve the problem, unfortunately.
Comment 3 Dave Airlie 2005-01-11 16:52:21 UTC
don't suppose you can plug a serial console in or something, it would be nice to
know does the machine actually die or is it just X getting stuck and messing up
the graphics hardware...
Comment 4 Antoni Grzymała 2005-01-11 17:16:48 UTC
No possibility of connecting a serial console at the moment, however the machine
seems to die completely, judging by the way some pluggable hardware acts. My
wireless PCMCIA adapter doesn't get it's lights back on after unplugging and
plugging in again. The built in ethernet goes dead as well.
Comment 5 Michel Dänzer 2005-01-12 07:25:03 UTC
Sounds like a dupe of bug 1886.
Comment 6 Shawn Starr 2005-09-28 08:59:28 UTC
Can you try a newer build of Xorg CVS with dri-CVS head and report your findings?
Comment 7 Antoni Grzymała 2005-09-28 15:16:47 UTC
This has been fixed in 6.8.2. This patch was required:

--- xc/programs/Xserver/hw/xfree86/drivers/ati/r128_dri.c.orig  2004-11-15
13:26:09.725996808 +0100
+++ xc/programs/Xserver/hw/xfree86/drivers/ati/r128_dri.c 2004-11-15
13:26:21.997131312 +0100
@@ -1297,6 +1297,7 @@
     if (info->irq) {
  drmCtlUninstHandler(info->drmFD);
  info->irq = 0;
+ info->gen_int_cntl = 0;
     }  
        
        /* De-allocate vertex buffers */

Thanks.


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.