Bug 20086 - [UXA/EXA G45] xorg freeze in hyperspace screensaver
Summary: [UXA/EXA G45] xorg freeze in hyperspace screensaver
Status: RESOLVED DUPLICATE of bug 18964
Alias: None
Product: xorg
Classification: Unclassified
Component: Driver/intel (show other bugs)
Version: unspecified
Hardware: Other All
: medium normal
Assignee: Eric Anholt
QA Contact: Xorg Project Team
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2009-02-12 09:40 UTC by martin
Modified: 2009-02-18 13:04 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments
gdb backtrace when UXA xorg freeze was in effect (5.90 KB, text/x-log)
2009-02-12 09:40 UTC, martin
no flags Details
my normal xorg_log (it's running UXA but log was not captured when bug happened) (34.68 KB, text/x-log)
2009-02-12 09:44 UTC, martin
no flags Details
my normal dmesg (it's running UXA but log was not captured when bug happened) (42.73 KB, text/x-log)
2009-02-12 09:45 UTC, martin
no flags Details
dmesg (while hung) (78.46 KB, text/plain)
2009-02-16 10:19 UTC, martin
no flags Details
xorg.log (while hung) (37.00 KB, text/plain)
2009-02-16 10:20 UTC, martin
no flags Details
xorg.log.old (while hung) (34.75 KB, application/x-trash)
2009-02-16 10:20 UTC, martin
no flags Details

Description martin 2009-02-12 09:40:19 UTC
Created attachment 22864 [details]
gdb backtrace when UXA xorg freeze was in effect

ubuntu jaunty, updated as of feb 14th, which means:

libdrm-intel1 and libdrm2 is 2.4.4-0ubuntu6
xserver-xorg-video-intel is 2:2.6.1-1ubuntu2
libgl1-mesa-dev is 7.3-1ubuntu1
Linux kingfish 2.6.28-7-generic #20-Ubuntu SMP Mon Feb 9 15:42:34 UTC 2009 x86_64 GNU/Linux

Board is a Gigabyte GA-EG45M-DS2H with G45-chipset (PCI id is 8086:2e22):
Nothing suspicious in xorg_log or dmesg.

First it happened once at random, after reboot I was able to repro it once but then it didn't happen again. The repro steps I used was (these may or may not work, probably they won't):

1. open this URL in firefox:
   http://savage2.s2games.com/download.php
2. click on the "linux 64-bit" link in the "big download column"
3. wait 2-4 seconds, press BACK
4. go to step 2 (until you click it like 5-10 times)

The stack trace has a reference to something like "animate mouse cursor" and my Ubuntu jaunty uses a small animated circle thing as it's xlib "WATCH" cursor.

I also tried to write a small pygtk program that showed this "waiting" cursor and even running that cursor animation for 1-2 minutes didn't repro the bug. 

---

I forgot to dump "cat /proc/dri/0/i915_gem_interrupt" two times to see if it's still firing interrupts.

If this happens again, what else would you like me to collect?
Comment 1 martin 2009-02-12 09:44:52 UTC
Created attachment 22865 [details]
my normal xorg_log (it's running UXA but log was not captured when bug happened)
Comment 2 martin 2009-02-12 09:45:46 UTC
Created attachment 22866 [details]
my normal dmesg (it's running UXA but log was not captured when bug happened)
Comment 3 Eric Anholt 2009-02-13 17:37:06 UTC
Looks like a pretty generic "the chip is hung".  The /proc/dri/0/i915_gem_interrupt file as you noted, /proc/dri/0/gem_objects are both useful, dmesg, Xorg.0.log from the bug actually happening, and that's about all the information we can get out of the system so far.  We're working on getting more information exported for 2.6.30 so we can debug this stuff better.
Comment 4 martin 2009-02-16 10:17:51 UTC
I tried out the 2.6.29-020629rc5-generic kernel today (I installed it on top of my normal jaunty system using this deb package http://kernel.ubuntu.com/~apw/mainline/v2.6.29-rc5/linux-image-2.6.29-020629rc5-generic_2.6.29-020629rc5_amd64.deb ). I still got the same userspace (i.e. same intel ddx and dri drivers and same mesa etc).

While I was stepping through the fullscreen preview of all the screensaver I eventually hit this bug again.

I current have the machine frozen (xorg stack is the same as I attached previously).

/proc/.../gem_objects contains:
3656 objects
314982400 object bytes
6 pinned
17076224 pin bytes
179490816 gtt bytes
234885120 gtt total

/proc/.../i915_interrupt contains:
Interrupt enable:    00000053
Interrupt identity:  00000000
Interrupt mask:      fffedfac
Pipe A stat:         00000206
Pipe B stat:         00000000
Interrupts received: 20576
Current sequence:    169153
Waiter sequence:     169324
IRQ sequence:        167120

(and it also doesnt change if I re-run this command again after 10 secs)
Comment 5 martin 2009-02-16 10:19:53 UTC
Created attachment 22987 [details]
dmesg (while hung)
Comment 6 martin 2009-02-16 10:20:18 UTC
Created attachment 22988 [details]
xorg.log (while hung)
Comment 7 martin 2009-02-16 10:20:43 UTC
Created attachment 22989 [details]
xorg.log.old (while hung)
Comment 8 martin 2009-02-16 11:29:32 UTC
Eric...

_GREAT_ news... I found a 100% consistent set of repro steps to trigger this freeze. Basically, all I need to do is to open System::Preferences::Screensaver and then press the Preview button and finally scroll down to the "Hyperspace" screensaver (in ubuntu this screensaver ships in a package called "rss-glx".

Note #1: I even tried to cold boot just to make sure it wasnt some crappy GPU state or whatever that made it consistent. Even with cold boot I was able to repro it.

Note: I also tried it with EXA and that also hangs (still on the .29rc5 vanilla kernel though).
Comment 9 Gordon Jin 2009-02-17 05:39:35 UTC
(In reply to comment #8)
> Eric...
> 
> _GREAT_ news... I found a 100% consistent set of repro steps to trigger this
> freeze. Basically, all I need to do is to open System::Preferences::Screensaver
> and then press the Preview button and finally scroll down to the "Hyperspace"
> screensaver (in ubuntu this screensaver ships in a package called "rss-glx".

This is mentioned in bug#18964.
Comment 10 martin 2009-02-17 05:46:59 UTC
He has a different xorg stack that I got and I also don't have any "[mi] EQ overflow" in my xorg.log; otherwise it looks a lot like the same bug.

If/Once there is a patch available for bug#18964 ping me and I will test it on my GA-EG45M-DS2H system.
Comment 11 Eric Anholt 2009-02-18 12:55:05 UTC
The screensaver (GL app) hang is probably different from "I was browsing things in firefox and it hung" (no GL involved).  If you want this bug to be about the screensaver hang, though, it should be closed as a dupe of the other one that was originally about the hyperspace hang.
Comment 12 martin 2009-02-18 13:04:49 UTC
Agree, it's sufficient if I test the patch for 18964 once it becomes available.
Closing as a dup of bug #18964

*** This bug has been marked as a duplicate of bug 18964 ***


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.