Bug 88652

gem_evict_everything/minor-hang also has this issue.

*** Bug 88655 has been marked as a duplicate of this bug. ***

It also happens on BDW.

./gem_evict_everything --run-subtest swapping-hang also causes system hang on SNB.

(In reply to lu hua from comment #4)
> ./gem_evict_everything --run-subtest swapping-hang also causes system hang
> on SNB.

I doubt it is the same bug. Please file it separately and we can dup if it does match.

(In reply to Chris Wilson from comment #5)
> (In reply to lu hua from comment #4)
> > ./gem_evict_everything --run-subtest swapping-hang also causes system hang
> > on SNB.
> 
> I doubt it is the same bug. Please file it separately and we can dup if it
> does match.

OK, report bug 88821 to track it on SNB, Thanks.

Report bug 89000 to track BDW.
Test on BSW with i915.enable_ppgtt=0, it works well.

Bisect shows: 6d3d8274bc45de4babb62d64562d92af984dd238 is the first bad commit.
commit 6d3d8274bc45de4babb62d64562d92af984dd238
Author:     Nick Hoath <nicholas.hoath@intel.com>
AuthorDate: Thu Jan 15 13:10:39 2015 +0000
Commit:     Daniel Vetter <daniel.vetter@ffwll.ch>
CommitDate: Tue Jan 27 09:50:53 2015 +0100

    drm/i915: Subsume intel_ctx_submit_request in to drm_i915_gem_request

    Move all remaining elements that were unique to execlists queue items
    in to the associated request.

    Issue: VIZ-4274

    v2: Rebase. Fixed issue of overzealous freeing of request.
    v3: Removed re-addition of cleanup work queue (found by Daniel Vetter)
    v4: Rebase.
    v5: Actual removal of intel_ctx_submit_request. Update both tail and postfix
    pointer in __i915_add_request (found by Thomas Daniel)
    v6: Removed unrelated changes

    Signed-off-by: Nick Hoath <nicholas.hoath@intel.com>
    Reviewed-by: Thomas Daniel <thomas.daniel@intel.com>
    [danvet: Reformat comment with strange linebreaks.]
    Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

Please retest with current drm-intel-nightly that has

commit f82107950e9bda3779610e37bdfdccae6fc16f87
Author: Nick Hoath <nicholas.hoath@intel.com>
Date:   Thu Jan 29 16:55:07 2015 +0000

    drm/i915: Fix a use-after-free in intel_execlists_retire_requests

*** Bug 88790 has been marked as a duplicate of this bug. ***

*** Bug 88845 has been marked as a duplicate of this bug. ***

*** Bug 88688 has been marked as a duplicate of this bug. ***

*** Bug 88840 has been marked as a duplicate of this bug. ***

*** Bug 89000 has been marked as a duplicate of this bug. ***

*** Bug 89005 has been marked as a duplicate of this bug. ***

*** Bug 88817 has been marked as a duplicate of this bug. ***

All the dupes have the same bisected bad commit.

*** Bug 88987 has been marked as a duplicate of this bug. ***

Note, when you test and verify this bug, please have a look at what the steps to reproduce were in the duplicate bugs, and see if they are truly fixed too. Thanks.

Test on the latest drm-intel-nightly kernel, this issue still exists.
Test commit ad95125eaef18eebb9f47261ce3c99957f5953de
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Mon Feb 9 21:31:04 2015 +0100

    drm-intel-nightly: 2015y-02m-09d-20h-26m-16s UTC integration manifest

Test on the latest -nightly kernel, gem_concurrent_blit/cpu-bcs-early-read-forked-hang-blt also causes system hang, add i915.enable_ppgtt=0, it doesn't have hang issue.

Test some gem_concurrent_blit*hang* cases on BDW and BSW, they all cause system hang and have the same bisect commit.

I've submitted a fix for this issue:
https://patchwork.kernel.org/patch/5819071/

Nick, for tracking purposes we keep the bugs open until we've merged the patches upstream.

lu hua, please test Nick's patch.

(In reply to Nick Hoath from comment #23)
> I've submitted a fix for this issue:
> https://patchwork.kernel.org/patch/5819071/

Apply this patch on the latest -nightly kernel. Test ./gem_evict_everything --run-subtest major-hang on BDW and BSW, it works well.
I will test the duplicate bugs later.

It also impacts SKL.

Nick's latest patch is at 
http://patchwork.freedesktop.org/patch/42508

Please test this, also against the tests in the duplicates.

(In reply to Jani Nikula from comment #27)
> Nick's latest patch is at 

Another update, http://patchwork.freedesktop.org/patch/42729

> Please test this, also against the tests in the duplicates.

Fixed by

commit b3a38998f042b862f5ba4d7f2268f3a8dfb4883a
Author: Nick Hoath <nicholas.hoath@intel.com>
Date:   Thu Feb 19 16:30:47 2015 +0000

    drm/i915: Fix a use after free, and unbalanced refcounting

in drm-intel-fixes.

(In reply to Jani Nikula from comment #29)
> Fixed by

commit b3a38998f042b862f5ba4d7f2268f3a8dfb4883a
Author: Nick Hoath
> <nicholas.hoath@intel.com>
Date:   Thu Feb 19 16:30:47 2015 +0000

   
> drm/i915: Fix a use after free, and unbalanced refcounting

in
> drm-intel-fixes.

I had test with kernel commit b3a38998f042b862f5ba4d7f2268f3a8dfb4883a

./gem_evict_everything --run-subtest swapping-hang will cause system hang still.
./gem_evict_everything --run-subtest major-hang will report an error, claimed that it need 6G free mem for this case. Return number is 77.

Hi,
   What's the failure rate you're getting for 
./gem_evict_everything --run-subtest swapping-hang ?

   I've just run it on BDW with the latest nightly, and I'm getting a 1/10 failure due to paging request BUG, with nothing to indicate 6d3d8274bc45de4babb62d64562d92af984dd238 is the cause.

   I can't run ./gem_evict_everything --run-subtest major-hang as there isn't enough memory on my system.

(In reply to Nick Hoath from comment #31)
> Hi,
>    What's the failure rate you're getting for 
> ./gem_evict_everything --run-subtest swapping-hang ?
> 
>    I've just run it on BDW with the latest nightly, and I'm getting a 1/10
> failure due to paging request BUG, with nothing to indicate
> 6d3d8274bc45de4babb62d64562d92af984dd238 is the cause.
> 
>    I can't run ./gem_evict_everything --run-subtest major-hang as there
> isn't enough memory on my system.


./gem_evict_everything --run-subtest swapping-hang 

I tried it with nightly commit 855932144a48a66081a62288bea6f2bbbf48e2e7(2015-02-28) on BDW and 0b2a1076c5cb4f383d6a8c940ffab1e27f241097(2015-02-25) on BSW, the reproducible is 100%.

./gem_evict_everything --run-subtest major-hang 
This case need 6G free mem, I don't have a machine with so much mem available. However, this case should be able to run before(refer to comment 25), and the machine I use is the same as Lu Hua. What's the reason for this mem requirement increasememn?

Hi,
   I'm investigating the lockup I see, but please can I have your kernel console output when the hang occurs from:
./gem_evict_everything --run-subtest swapping-hang
   to see if it's a difference problem.

FWIW the hang I am investigating still occurs with without 6d3d8274bc45de4babb62d64562d92af984dd238, so it will need a new bug if the kernel console output matches.

Created attachment 114019 [details]
console output---call trace from running case to hang

(In reply to wendy.wang from comment #35)
> Created attachment 114019 [details]
> console output---call trace from running case to hang

This log was based 0b2a1076c5cb4f383d6a8c940ffab1e27f241097(2015-02-25) drm-intel-nightly kernel testing result.

This latest trace is the same problem I'm investigating.
It pre-dates 6d3d8274bc45de4babb62d64562d92af984dd238.
The original bug introduced in 6d3d8274bc45de4babb62d64562d92af984dd238 is fixed and the fix upstreamed, and as such I am closing this bug as fixed. I have created bug 89441 to track the newly reported (pre-existing) issue.

Moving old bug from Verified to Closed.