Bug 94759

Summary: [BAT BSW] lockdep splat in drm_gem_mmap
Product: DRI Reporter: Joonas Lahtinen <joonas.lahtinen>
Component: DRM/IntelAssignee: Chris Wilson <chris>
Status: CLOSED DUPLICATE QA Contact: Intel GFX Bugs mailing list <intel-gfx-bugs>
Severity: normal    
Priority: medium CC: intel-gfx-bugs
Version: unspecified   
Hardware: x86-64 (AMD64)   
OS: Linux (All)   
Whiteboard:
i915 platform: i915 features:
Attachments:
Description Flags
dmesg from I-G-T run none

Description Joonas Lahtinen 2016-03-30 11:07:55 UTC 
Created attachment 122637 [details]
dmesg from I-G-T run

At: /archive/results/CI_IGT_test/Patchwork_1712/bsw-nuc-2/html/bsw-nuc-2@Patchwork_1712@1/igt@gem_exec_basic@gtt-blt.html

Log beginning (full log attached):
[  149.921578] ======================================================
[  149.921640] [ INFO: possible circular locking dependency detected ]
[  149.921703] 4.5.0-gfxbench+ #1 Not tainted
[  149.921745] -------------------------------------------------------
[  149.921807] gem_exec_basic/5896 is trying to acquire lock:
[  149.921861]  (&dev->struct_mutex){+.+.+.}, at: [<ffffffff81514ea1>] drm_gem_mmap+0x1a1/0x270
[  149.921969] 
but task is already holding lock:
[  149.922027]  (&mm->mmap_sem){++++++}, at: [<ffffffff8117d8c4>] vm_mmap_pgoff+0x44/0xa0
[  149.922125] 
which lock already depends on the new lock.
Comment 1 Chris Wilson 2016-03-30 11:31:19 UTC 
That one should be the kernfs one, which goes back to the question why is this only sporadic in CI?
Comment 2 Joonas Lahtinen 2016-03-31 14:16:46 UTC 
Also in gem_mmap_gtt, same chain:
[  145.549678] ======================================================
[  145.549739] [ INFO: possible circular locking dependency detected ]
[  145.549803] 4.6.0-rc1-gfxbench+ #1 Tainted: G     U         
[  145.549858] -------------------------------------------------------
[  145.549919] gem_mmap_gtt/5996 is trying to acquire lock:
[  145.549971]  (&dev->struct_mutex){+.+.+.}, at: [<ffffffff8151c781>] drm_gem_mmap+0x1a1/0x270
[  145.550077] 
but task is already holding lock:
[  145.550134]  (&mm->mmap_sem){++++++}, at: [<ffffffff81183204>] vm_mmap_pgoff+0x44/0xa0
[  145.550230] 
which lock already depends on the new lock.
Comment 3 Joonas Lahtinen 2016-04-01 08:57:26 UTC 
Patch was merged to our local CI topic branch, seems to have been effective for past two runs (which is still quite low confidence level):

commit 6954af8b55f3b00b08f7759f479c41388fbe364f
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Thu Mar 31 11:45:06 2016 +0100

    kernfs: Move faulting copy_user operations outside of the mutex

Greg K-H will merge the patch upstream for 4.7-rc1.
Comment 4 Chris Wilson 2016-04-01 09:07:18 UTC 
Might as well keep the records straight and close the one mentioned in the patch.

*** This bug has been marked as a duplicate of bug 94350 ***
Comment 5 Jari Tahvanainen 2016-10-07 09:07:05 UTC 
Closing as duplicate of closed+fixed.

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.