Project

General

Profile

Actions

Bug #2866

closed

kernel dumps core after upgrade dri to 11.0.7

Added by arcade@b1t.name over 8 years ago. Updated about 8 years ago.

Status:
Closed
Priority:
Normal
Assignee:
Category:
Driver
Target version:
-
Start date:
12/18/2015
Due date:
% Done:

0%

Estimated time:

Description

Well, previously my system generally worked with X, drm and stuff. But after upgrade when I try to start Xfce I get this:

panic: lockmgr: LK_RELEASE: no lock held
cpuid = 2
Trace beginning at frame 0xffffffe0f5e20558
panic() at panic+0x238 0xffffffff8027ba64
panic() at panic+0x238 0xffffffff8027ba64
lockmgr() at lockmgr+0x685 0xffffffff8026e132
drm_wait_vblank() at drm_wait_vblank+0x444 0xffffffff829e7c31
drm_ioctl() at drm_ioctl+0x37a 0xffffffff829e58bc
dev_dioctl() at dev_dioctl+0x76 0xffffffff8025b753
boot() called on cpu#2
Uptime: 36m18s
Physical memory: 7337 MB

Actions #1

Updated by arcade@b1t.name over 8 years ago

Well, after recent drm update with introduction of permissions for /dev/dri this looks like:

1. Everything kindda work without permissions (except for OpenGL).
2. Kernel panics when something tries to use OpenGL (e.g. with permissions).

Actions #2

Updated by dillon over 8 years ago

  • Assignee set to dillon

Looks like the DRM_WAIT_ON() macro is assuming that the DRM device is locked and goes through an unlock/relock sequence on it. But the DRM device is not locked by either of the two callers of that routine from what I can tell.

I will reproduce the bug and test and commit a fix (which will probably end up being removing the unlock/relock sequence from the DRM_WAIT_ON() macro).

-Matt

Actions #3

Updated by dillon over 8 years ago

Fix pushed, please retest with master.

-Matt

Actions #4

Updated by arcade@b1t.name about 8 years ago

Yes, it's working now. Thank you.

Actions #5

Updated by dillon about 8 years ago

  • Status changed from New to Closed

closed - should now be fixed.

Actions

Also available in: Atom PDF