Bug 603139 - XServer locks screen, HW-Cursor works, Thinkpad T510
Summary: XServer locks screen, HW-Cursor works, Thinkpad T510
Status: CLOSED DUPLICATE of bug 596330
Alias: None
Product: Fedora
Classification: Fedora
Component: xorg-x11-drv-nouveau   
(Show other bugs)
Version: 13
Hardware: x86_64
OS: Linux
Target Milestone: ---
Assignee: Ben Skeggs
QA Contact: Fedora Extras Quality Assurance
Keywords: Triaged
Depends On:
TreeView+ depends on / blocked
Reported: 2010-06-11 16:10 UTC by Philipp K.
Modified: 2018-04-11 08:45 UTC (History)
5 users (show)

Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of:
Last Closed: 2010-06-23 16:00:33 UTC
Type: ---
Regression: ---
Mount Type: ---
Documentation: ---
Verified Versions:
Category: ---
oVirt Team: ---
RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: ---

Attachments (Terms of Use)
GDB backtrace of X-Server (6.03 KB, text/plain)
2010-06-11 16:10 UTC, Philipp K.
no flags Details
dmesg output (56.28 KB, text/plain)
2010-06-11 16:12 UTC, Philipp K.
no flags Details
peek output (184.00 KB, text/plain)
2010-06-11 16:13 UTC, Philipp K.
no flags Details
kernel function trace (173.83 KB, text/plain)
2010-06-12 10:29 UTC, Philipp K.
no flags Details

Description Philipp K. 2010-06-11 16:10:05 UTC
Created attachment 423324 [details]
GDB backtrace of X-Server

Description of problem:
The X-Server gets stuck in an infinite loop. It appears to be the same problem as described in https://bugs.freedesktop.org/show_bug.cgi?id=28320

Attaching gdb to the X-Server unveils (see attachment backtrace.log), that the X-Server is calling drmIoctl with the requestcode 0x40086482. The infinite loop is in drmIoctl itself.
Looking at the sourcecode of nouveau_bo_wait shows, that it should be DRM_NOUVEAU_GEM_CPU_PREP=0x42. I'm a little bit confused, that requestcode ends with 82, although the definition of DRM_NOUVEAU_GEM_CPU_PREP indicates, that it should be 42.

dmesg shows "[drm] nouveau 0000:01:00.0: PFIFO_DMA_PUSHER - Ch 2"

I also attached the output of the program peek (http://0x04.net/~mwk/pgtest/) which i have seen in some other bugreports.

Version-Release number of selected component (if applicable):
cat /sys/module/nouveau/srcversion -> 117DEC6255413EEBAF88071
yum info -> 6.20100423git13c1043.fc13

How reproducible:
Happens randomly, at least once a day. Looking a movie with VLC increases the chance.

Steps to Reproduce:
1. Watch a movie with VLC.
Actual results:
X-Server stuck in an infinite loop

Expected results:
X-Server should work properly.

Additional info:

Comment 1 Philipp K. 2010-06-11 16:12:04 UTC
Created attachment 423327 [details]
dmesg output

Comment 2 Philipp K. 2010-06-11 16:13:53 UTC
Created attachment 423328 [details]
peek output

Comment 3 Philipp K. 2010-06-12 10:29:07 UTC
Created attachment 423478 [details]
kernel function trace

The ftrace was produced with 'echo nouveau_* drm_* ttm_* > set_ftrace_filter' option.

Seems that the nouveau kernel module hangs in an infinte loop in nouveau_fence_wait.

Comment 4 Matěj Cepl 2010-06-15 14:13:44 UTC
That " [mi] EQ overflowing" message is completely meaningless. Please, read
http://marc.info/?l=fedora-devel-list&m=124101535025331&w=2 and then take a
look at the disaster which happens when people start to file this message as a
bug at bug 465884. 

The rest of your analysis is much more interesting. Passing to developers.

Comment 5 cvita 2010-06-17 21:31:42 UTC
Same problem here.

I've thought the problem was with firefox+adobe flash player (64 bits), but even without the flash plugin, the problem still exists.

I'm running a Thinkpad T510i with the mesa-dri-drivers-experimental installed. Without this package the system also hangs.

I have:
01:00.0 VGA compatible controller: nVidia Corporation GT218 [NVS 3100M] (rev a2)

Another curious thing is that when the screen locks, if I press the caps lock button, the led doesn't turn on or off. However, if I press the powerbutton (which is configured to "ask what to do"), and waiting 60 seconds, the system sometimes starts to shutdown and powerdown (screen image is not updated. I know it's shutting down due to hard disk activity led, and the power off at the end).

Brightness control stops working, and pressing Fn+F5 makes the hard drive led to blink.

My conclusion is that the system is not completely dead, only the screen image and keyboard.
I'll try next time to make a ssh connection from another PC to see if the system allows me to login.

Comment 6 Ben Skeggs 2010-06-23 16:00:33 UTC

*** This bug has been marked as a duplicate of bug 596330 ***

Note You need to log in before you can comment on or make changes to this bug.