Created attachment 71069 [details] gem_cacheing fail dmesg good commit: Kernel: (drm-intel-next-queued)1a240d4de2ccf40de5796a4d1dbb3a0236051fc9 bad commit: Kernel: (drm-intel-next-queued)46cddd0246ecca425a2d0f07f3102030665afc77 git bisect log: --------------------- # bad: [46cddd0246ecca425a2d0f07f3102030665afc77] drm/i915: Remove duplicate and unused register #defines in i915_reg.h # good: [1a240d4de2ccf40de5796a4d1dbb3a0236051fc9] drm/i915: fixup sparse warnings git bisect start '46cddd0246ecca425a2d0f07f3102030665afc77' '1a240d4de2ccf40de5796a4d1dbb3a0236051fc9' # bad: [3cb49d22561ad4063856ad809b175478c1f09c8a] drm/i915: setup the hangcheck timer early git bisect bad 3cb49d22561ad4063856ad809b175478c1f09c8a # good: [29bc0d33dcd509a87cfc6eaf4ac29ce6cc2df40b] drm/i915: Split intel_ring_begin git bisect good 29bc0d33dcd509a87cfc6eaf4ac29ce6cc2df40b # bad: [7c7b4d55fb69f679c27f8ed056c6f304cd69c3b3] drm/i915: Set initial seqno value close to wrap boundary git bisect bad 7c7b4d55fb69f679c27f8ed056c6f304cd69c3b3 # good: [d2ef0e20dd0d6257526ad41b1b5bcfdc4cff4054] drm/i915: Add intel_ring_handle_seqno wrap git bisect good d2ef0e20dd0d6257526ad41b1b5bcfdc4cff4054
Created attachment 71070 [details] gem_cpu_concurrent_blit faild dmesg I bisect this failed case, and get the same bisect result. So I conbine these two cases as the same bug.
Just to check: Does reverting commit 7c7b4d55fb69f679c27f8ed056c6f304cd69c3b3 Author: Mika Kuoppala <mika.kuoppala@linux.intel.com> Date: Tue Dec 4 15:12:05 2012 +0200 drm/i915: Set initial seqno value close to wrap boundary To gain confidence in the wrap handling, make it happen quite soon after the boot. Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Fix the issues again? btw, when adding bisect results, please cite the commit that introduce the regression like above, since that's the most important thing for us. Only if the bisect turns out to not be stable do we look at the bisect log to figure out where the bisect might have gone wrong. I'll update BKMs.
Assigning to Mika ...
> btw, when adding bisect results, please cite the commit that introduce the > regression like above, since that's the most important thing for us. Only if > the bisect turns out to not be stable do we look at the bisect log to figure > out where the bisect might have gone wrong. I'll update BKMs. commit 7c7b4d55fb69f679c27f8ed056c6f304cd69c3b3 Revert this commit at the top of the -next-queued branch, after test, I'm sure it's good.
See https://patchwork.kernel.org/patch/1844941/
(In reply to comment #5) > See https://patchwork.kernel.org/patch/1844941/ I can't patch below in your given patch. which branch could I attach this patch? >>@@ -628,7 +634,7 @@ gen6_ring_sync(struct intel_ring_buffer *waiter, >> BUG_ON(!waiter->outstanding_lazy_request); >> >> /* If seqno wrap happened, omit the wait with no-ops */ >>- if (likely(waiter->outstanding_lazy_request > seqno)) { >>+ if (likely(i915_gem_next_seqno(waiter->dev) > seqno)) { >> intel_ring_emit(waiter, >> dw1 | >> signaller->semaphore_register[waiter->id]);
case gem_gtt_concurrent_blit core dumped, has same bisect commit. output: gem_gtt_concurrent_blit: gem_gtt_concurrent_blit.c:72: cmp_bo: Assertion `*vaddr++ == val' failed. Aborted (core dumped) dmesg: [ 47.508457] [drm:i915_driver_open], [ 47.508472] [drm:intel_crtc_set_config], [CRTC:3] [FB:21] #connectors=1 (x y) (0 0) [ 47.508478] [drm:intel_modeset_stage_output_state], [CONNECTOR:17:HDMI-A-2] to [CRTC:3] [ 47.508480] [drm:intel_crtc_set_config], [CRTC:5] [NOFB] [ 47.508482] [drm:intel_modeset_stage_output_state], [CONNECTOR:17:HDMI-A-2] to [CRTC:3] [ 47.508484] [drm:intel_crtc_set_config], [CRTC:7] [NOFB] [ 47.508485] [drm:intel_modeset_stage_output_state], [CONNECTOR:17:HDMI-A-2] to [CRTC:3] [ 47.508490] [drm:sandybridge_update_wm], FIFO watermarks For pipe A - plane 9, cursor: 6 [ 47.508495] [drm:sandybridge_update_wm], FIFO watermarks For pipe A - plane 9, cursor: 6 [ 47.508500] [drm:sandybridge_update_wm], FIFO watermarks For pipe A - plane 9, cursor: 6 [ 47.508505] [drm:i915_driver_open], [ 47.781840] [drm:intel_crtc_set_config], [CRTC:3] [FB:21] #connectors=1 (x y) (0 0) [ 47.781847] [drm:intel_modeset_stage_output_state], [CONNECTOR:17:HDMI-A-2] to [CRTC:3] [ 47.781849] [drm:intel_crtc_set_config], [CRTC:5] [NOFB] [ 47.781851] [drm:intel_modeset_stage_output_state], [CONNECTOR:17:HDMI-A-2] to [CRTC:3] [ 47.781853] [drm:intel_crtc_set_config], [CRTC:7] [NOFB] [ 47.781855] [drm:intel_modeset_stage_output_state], [CONNECTOR:17:HDMI-A-2] to [CRTC:3] [ 47.781859] [drm:sandybridge_update_wm], FIFO watermarks For pipe A - plane 9, cursor: 6 [ 47.781864] [drm:sandybridge_update_wm], FIFO watermarks For pipe A - plane 9, cursor: 6 [ 47.781869] [drm:sandybridge_update_wm], FIFO watermarks For pipe A - plane 9, cursor: 6
Chris patch is on top of the old tree (the broken one), latest -nightly has the offeding patch take out already, so should work again. Can you please retest?
(In reply to comment #8) > Chris patch is on top of the old tree (the broken one), latest -nightly has > the offeding patch take out already, so should work again. Can you please > retest? Yeah, I build a kernel with the top commit on -nightly branch, the case passed. commit : 3fe1b3064d39c0930796aaae81c44f9e8a5264bd
Is this RESOLVED FIXED?
(In reply to comment #10) > Is this RESOLVED FIXED? Nope, offending patch was only taken out, we still need to fix up seqno wrapping for real. So I think we can keep this one here around as a reminder.
commit f72b3435c1a75406d82d6e252bb78f009efd4bd9 Author: Mika Kuoppala <mika.kuoppala@linux.intel.com> Date: Mon Dec 10 15:41:48 2012 +0200 and subsequent patches have landed, should be ok to close now? Yangwei, can you confirm?
(In reply to comment #12) > commit f72b3435c1a75406d82d6e252bb78f009efd4bd9 > Author: Mika Kuoppala <mika.kuoppala@linux.intel.com> > Date: Mon Dec 10 15:41:48 2012 +0200 > > and subsequent patches have landed, should be ok to close now? Yangwei, can > you confirm? I have tested this commit, these two cases worked well. We can close this bug now.
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.