Kernel Bug Tracker – Bug 45181
Kernel 3.5 OOPS on i915 DRM in skip_spaces()
Last modified: 2012-10-15 20:46:03 UTC
Created attachment 76191 [details]
i915 Log output and subsequent OOPS
Ben Widawsky added a patch that went into the 3.5 mainline kernel that adds sysfs entries for the i915/DRM driver. Unfortunately, that patches causes the kernel to OOPS in the function skip_spaces().
I bisected to locate the patch, details of which are attached, along with the oops log itself. CC'ing the patch author too.
Hardware: Dell OptiPlex 990
CPU: Intel Core i7-2600
i915 variant: Intel Corporation Second Generation Core Processor Family
Integrated Graphics Controller (rev 09)
Created attachment 76201 [details]
Result of bisecting kernel, showing patch information
Please attach your kernel .config, too.
Created attachment 76271 [details]
The .config from kernel 3.5
OK, attaching the .config to the bug report.
I neglected to mention (in case it is not obvious) that the bug occurs at the same time on every boot, with the same stack trace. I switched the i915 from a built-in to a loadable module, which allows the kernel to boot into a minimally functional state without triggering a panic.
Pardon the delay, I've been on vacation a bit. I'll look at your .config shortly. But first to confirm: When you build i915 as a module (and then load it), does the kernel still oops or does it only oops when you build i915 compiled-in?
*** Bug 45361 has been marked as a duplicate of this bug. ***
Do you have CONFIG_PM set?
I think this is the issue which was mentioned here:
Unfortunately the original poster seems to have gone away without satisfying my request. Can you see if this patch fixes the problem If this patch works, I'll submit another patch with what I want. The patch should apply to drm-intel-fixes.
Created attachment 76901 [details]
Working patch from Ben Widawsky
No problem on the delay. :-)
Indeed, I do not have CONFIG_PM set (no suspend/hibernate) as this is a server that is effectively never powered off.
Ben, your patch works just fine! No OOPSen on module load, and no panic on boot with i915 built-in.
FWIW, I'm attaching the patch Ben mentioned above, for convenience.
Marking this as resolved; I've tested the patch as module and as built-in, with modeset=0 and modeset=1 (moot, but for completeness) and it works fine.
A patch referencing this bug report has been merged in Linux v3.6-rc2:
Author: Hunt Xu <email@example.com>
Date: Sun Jul 1 03:45:07 2012 +0000
drm/i915: make rc6 in sysfs functions conditional
A patch referencing a commit referencing this bug report has been merged in Linux v3.7-rc1:
Author: Ben Widawsky <firstname.lastname@example.org>
Date: Sun Sep 2 00:24:40 2012 -0700
drm/i915: Enable some sysfs stuff without CONFIG_PM