Bug 368587 – sys-apps/module-init-tools-3.12-r1 with 2.6.37 - [drm] nouveau 0000:01:00.0: PFIFO_BAR_FAULT - VM: Trapped write at 00016f6c10 status 00001a26 channel 0 (0x00000040)

• Product: Gentoo Linux
• Status: RESOLVED NEEDINFO

[Description Joerg Schaible 2011-05-24 13:18:30 UTC]

After updating module-init-tools from 3.5 to 3.12-r1 my notebook switches info powersave mode when loading the modules (directly after processing uevents) and I am not able to bring it back then. I have to invoke reboot by power off the hard way. This happens only with kernel gentoo-sources-3.6.37-r4, boot process for gentoo-sources-3.6.36-r8 runs fine. Originally I suspected upower, because it was updated at the same time, but the behavior is the same for upower versions 0.9.8, 0.9.9 and 0.9.10. Only a downgrade of module-init-tools to 3.5 allows me to boot properly from the 3.6.37 kernel.

I am running my Dell notebook in the docking station.

[Comment 1 Jeroen Roovers (RETIRED) 2011-05-24 13:50:09 UTC]

1) What is it that you call "powersave mode"? Suspend mode? Or does it just
   freeze and /look/ like it's gone into suspend/sleep/standby?
2) Please post your `emerge --info' output too.
3) Attach whatever logs are kept of the boot process, like the syslog, dmesg and
   rc output.

[Comment 2 Joerg Schaible 2011-05-26 09:37:06 UTC]

OK. I reinstalled module-init-tools-3.12-r1 now and rebooted with 2.6.37. The screen went blank directly after the line "Waiting for uevents being processed ... [OK]". Normally the line "Device initiated services: net.eth0 [OK]" follows. Seems when modules are loaded.

Went blank means: Screen gets black, OSD of screen appears with the mesage "Entering power save mode", which disappears again after some time. That's it. Nothing I can do will bring the screen back.

The boot process actually continues. When I log in into the console (typing blindly) I can save dmesg output to a file, but it contains only lines like:

========== %< ===========
[drm] nouveau 0000:01:00.0: PFIFO_BAR_FAULT - VM: Trapped write at 00016f6c10 status 00001a26 channel 0 (0x00000040)
[drm] nouveau 0000:01:00.0: PFIFO_BAR_FAULT - VM: Trapped write at 00016ea428 status 00001a26 channel 0 (0x00000040)
[drm] nouveau 0000:01:00.0: PFIFO_BAR_FAULT - VM: Trapped write at 00016f531c status 00001a26 channel 0 (0x00000040)
========== %< ===========

[Comment 3 Joerg Schaible 2011-05-26 09:38:40 UTC]

Created attachment 274677 [details]
dmesg-good.txt

dmesg output with module-init-tools-3.5 in kernel 2.6.37

[Comment 4 Joerg Schaible 2011-05-26 09:45:50 UTC]

In the kernel (2.6.37) log I can find the dmesg entries directly after initializing syslog-ng:

============= %< =============
May 26 10:25:47 josix syslog-ng[1857]: syslog-ng starting up; version='3.1.4'
May 26 10:25:47 josix u 0000:01:00.0: PFIFO_BAR_FAULT - VM: Trapped write at 000169ade4 status 00001a26 channel 0 (0x00000040)
May 26 10:25:47 josix kernel: [drm] nouveau 0000:01:00.0: PFIFO_BAR_FAULT - VM: Trapped write at 00016882c8 status 00001a26 ch
annel 0 (0x00000040)
[...]
============= %< =============

With old module-init-tools:
============= %< =============
May 26 10:36:19 josix syslog-ng[1812]: syslog-ng starting up; version='3.1.4'
May 26 10:36:19 josix kernel: Initializing cgroup subsys cpu
May 26 10:36:19 josix kernel: Linux version 2.6.37-gentoo-r4 (root@josix) (gcc version 4.4.5 (Gentoo 4.4.5 p1.2, pie-0.4.5) ) #3 SMP PREEMPT Thu May 12 10:38:23 CEST 2011
============= %< =============

[Comment 5 Joerg Schaible 2011-05-26 09:46:57 UTC]

Created attachment 274679 [details]
.config-2.6.37-gentoo-r4

Kernel configuration

[Comment 6 Joerg Schaible 2011-06-01 11:02:27 UTC]

After removing DRM and nouveau drivers from kernel, I can boot without hassle with 2.6.37.

[Comment 7 Mike Pagano 2011-06-03 19:58:58 UTC]

Does this issue repeat with 2.6.39 kernels?