Created attachment 72981 [details] I-G-T/kms_flip subtest: flip-vs-panning dmesg Environment: -------------------- Kernel: (drm-intel-next-queued)5b687ca1691aa6da6e16875d2ffc52a0898bd4ac Some additional commit info: Author: Egbert Eich <eich@suse.de> Date: Thu Jan 10 10:02:39 2013 -0500 drm/i915: Remove pch_rq_mask from struct drm_i915_private. Step: -------------------- 1. /GFX/Test/Intel_gpu_tools/intel-gpu-tools/tests/kms_flip --run-subtest flip-vs-panning 2. $?=10 3. STD-OUT: Using monotonic timestamps running testcase: flip-vs-panning Beginning flip-vs-panning on crtc 3, connector 7 1280x1024 60 1280 1328 1440 1688 1024 1025 1028 1066 0x5 0x48 108000 .......unexpected flip seq 7170, should be >= 7171 lspci -nn : -------------------- Description : -------------------- The kernel disabled most configurations in kernel hacking will cause this case failure. If I use the kernel with "kernel hacking", all subtests worked nicely. I have attached these two configs, maybe it is useful for you.
Created attachment 72982 [details] lspci message message of : lspci -nn
Created attachment 72984 [details] config with most kernel hacking disabled
Created attachment 72985 [details] config : kernel hacking enabled
Please retest all kms_flip failures with latest -nightly, specifically: commit 735dc0d1e29329ff34ec97f66e130cce481c9607 Merge: bac4b7c 20c60c3 Author: Dave Airlie <airlied@redhat.com> Date: Mon Jan 21 07:44:58 2013 +1000 Merge branch 'drm-kms-locking' of git://people.freedesktop.org/~danvet/drm-intel into drm-next Note that this commit is in the drm-next branch and so _not_ included in either drm-intel-fixes or drm-intel-next-queued. Hence please update the bug status with the result from -nightly only.
(In reply to comment #4) > Please retest all kms_flip failures with latest -nightly, specifically: > > commit 735dc0d1e29329ff34ec97f66e130cce481c9607 > Merge: bac4b7c 20c60c3 > Author: Dave Airlie <airlied@redhat.com> > Date: Mon Jan 21 07:44:58 2013 +1000 > > Merge branch 'drm-kms-locking' of > git://people.freedesktop.org/~danvet/drm-intel into drm-next > > Note that this commit is in the drm-next branch and so _not_ included in > either drm-intel-fixes or drm-intel-next-queued. Hence please update the bug > status with the result from -nightly only. I had a try by using latest -nightly and run "./kms_flip --run-subtest flip-vs-panning" but still fail. Environment: -------------------- Kernel: (drm-intel-nightly)b43ff00f39eb0c395b57e8f40b55a269c6b2e7ed Some additional commit info: Merge: 06ee1a3 735dc0d Author: Daniel Vetter <daniel.vetter@ffwll.ch> Date: Wed Jan 23 01:06:27 2013 +0100 Description : -------------------- stdout: unexpected flip seq 8275, should be >= 8276
Created attachment 74690 [details] Trace script I can't really think if any sensible reason for this problem. Even if the pan operation would overtake the flip (which it can with the current code) the frame count should still go forwards. Here it seems to stuck to the previous value. Well, I suppose the vblank code can be buggy (at least it's needlessly complex). Anyways, if you can reproduce with CONFIG_FUNCTION_TRACER enabled, this script may help us see what's going on. 1. start the trace script as root and redirect the output to some file 2. run kms_flip --run-subtest flip-vs-panning 3. stop the trace script 4. attach the trace log _and_ the errors from kms_flip to the bug the kms_flip errors are needed so that we know where in the log to start looking
# trace-cmd record -e drm -e i915 ./kms_flip -run-subtest flip-vs-panning Attach trace.dat
By using the key from Chris Wilson,I got the trace.dat and attached.
Created attachment 74963 [details] trace.dat
Can you please retest with the following patch applied to latest -nightly? https://patchwork.kernel.org/patch/2709081/
(In reply to comment #10) > Can you please retest with the following patch applied to latest -nightly? > > https://patchwork.kernel.org/patch/2709081/ OK, I applied it the patch on latest -nightly, and this case worked well, I find from June 10, it all passed in our nightly tests. I think this bug can be closed.
It's a drm core patch. I'll close the bug as soon the patch has landed in upstream drm branch and is merged into -nightly.
Ok, the patch was finally merged to drm-fixes and so can be tested in drm-intel-nightly: commit e91abf80a0998f326107874c88d549f94839f13c Author: Michel Dänzer <michel.daenzer@amd.com> Date: Wed Jun 12 11:58:44 2013 +0200 drm: Don't pass negative delta to ktime_sub_ns()
(In reply to comment #13) > Ok, the patch was finally merged to drm-fixes and so can be tested in > drm-intel-nightly: > > commit e91abf80a0998f326107874c88d549f94839f13c > Author: Michel Dänzer <michel.daenzer@amd.com> > Date: Wed Jun 12 11:58:44 2013 +0200 > > drm: Don't pass negative delta to ktime_sub_ns() By looking at these days nightly test results, this case worked well, so I verified here.
Closing old verified.
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.