Bug 81175

Summary:	[gm45 3.15] Lost HWS seqno writes g on boot or after wake from suspend
Product:	DRI	Reporter:	Jesse McClure <jmcclure>
Component:	DRM/Intel	Assignee:	Intel GFX Bugs mailing list <intel-gfx-bugs>
Status:	CLOSED FIXED	QA Contact:	Intel GFX Bugs mailing list <intel-gfx-bugs>
Severity:	normal
Priority:	medium	CC:	intel-gfx-bugs, jmcclure
Version:	unspecified
Hardware:	x86-64 (AMD64)
OS:	Linux (All)
Whiteboard:
i915 platform:		i915 features:

Description Jesse McClure 2014-07-10 14:16:18 UTC

I've been experiencing frequent kernel panics during boot or shortly after waking from suspend.  I'm running ArchLinux on a Lenovo X200 laptop.  I had traced the problem to any/all of intel-dri, mesa, or mesa-libgl as downgrading all three to version 10.1.4 prevents the issues.  The problem persists with all 10.2.x upgrades up to and including 10.2.3.

Based on logs, it seems the failed boots actually do completely boot, but nothing appears on screen after boot loader and kernel line.  If I do nothing, the kernel panics several seconds later - I am able to shutdown with the power button during this period which appears as a normal shutdown in the "failed boot" log.  I have not been able to get logs from sessions that end in a kernel panic.

The failed boots are frequent, but sporadic - roughly every other boot attempt.

I have a journalctl log from a failed boot [1], a successful boot for reference [2], and the drm/card0/error content [3].

The successful boot log includes the following which led me to post here:

[drm] stuck on render ring
[drm] GPU HANG: ecode 0:0x880e2084, in X [228], reason: Ring hung, action: reset
[drm] GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace.
[drm] Please file a _new_ bug report on bugs.freedesktop.org against DRI -> DRM/Intel
[drm] drm/i915 developers can then reassign to the right component if it's not a kernel issue.
[drm] The gpu crash dump is required to analyze gpu hangs, so please always attach it.
[drm] GPU crash dump saved to /sys/class/drm/card0/error
[drm:intel_pipe_set_base] *ERROR* pin & fence failed


While I've checked open bugs on DRM/Intel, I've not been able to assess whether this is related to any open reports.  Pardon the noise if it is.

-Jesse
jmcclure@cns.umass.edu

[1] https://raw.githubusercontent.com/TrilbyWhite/junkdrawer/master/boot_fail

[2] https://raw.githubusercontent.com/TrilbyWhite/junkdrawer/master/boot_success

[3 LARGE 1.4M] https://raw.githubusercontent.com/TrilbyWhite/junkdrawer/master/drm_error

Comment 1 Jani Nikula 2014-09-08 15:42:19 UTC

> [1] https://raw.githubusercontent.com/TrilbyWhite/junkdrawer/master/boot_fail
> 
> [2]
> https://raw.githubusercontent.com/TrilbyWhite/junkdrawer/master/boot_success
> 
> [3 LARGE 1.4M]
> https://raw.githubusercontent.com/TrilbyWhite/junkdrawer/master/drm_error

For future reference, please attach the information on the bug.

Comment 2 Chris Wilson 2015-06-13 12:24:26 UTC

Please try a recent kernel, or old kernel - it's a kernel bug. Hopefully we behave better, but if it is a regression that will also help narrow down where the fault was introduced.

Comment 3 Jesse McClure 2015-06-13 13:12:21 UTC

This was fixed - I think last october with mesa 10.3.  I forgot this bug was here as it was never assign, acknowledged, or paid any attention.

It is now resolved.

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.