| Summary: | [ILK] random missed interrupts from BSD | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Product: | DRI | Reporter: | lu hua <huax.lu> | ||||||
| Component: | DRM/Intel | Assignee: | cprigent <christophe.prigent> | ||||||
| Status: | CLOSED DUPLICATE | QA Contact: | Intel GFX Bugs mailing list <intel-gfx-bugs> | ||||||
| Severity: | minor | ||||||||
| Priority: | medium | CC: | christophe.prigent, intel-gfx-bugs, jinxianx.guo, yi.sun | ||||||
| Version: | unspecified | ||||||||
| Hardware: | All | ||||||||
| OS: | Linux (All) | ||||||||
| Whiteboard: | |||||||||
| i915 platform: | ILK | i915 features: | GEM/Other | ||||||
| Attachments: |
|
||||||||
Sounds like a regression between -fixes and -nightly. Can you try to bisect this? I will try to bisect it. Bisect it(Run each commit 20 cycles). Bisect shows b8866ef82d4f6c5361c8edd5199fe0cd19b947e3 is the first bad commit commit b8866ef82d4f6c5361c8edd5199fe0cd19b947e3 Author: Daniel Vetter <daniel.vetter@ffwll.ch> AuthorDate: Thu Apr 24 23:54:40 2014 +0200 Commit: Daniel Vetter <daniel.vetter@ffwll.ch> CommitDate: Mon May 5 10:56:57 2014 +0200 drm/i915/tv: extract set_color_conversion intel_tv_mode_set is still too bug. Reviewed-by: Imre Deak <imre.deak@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Pretty sure that bisect is bogus - it changes a bit of kernel code which is never run on ilk. Please try to redo the bisect. Since this is tricky to reproduce please first double-check the bisected commit: - Re-run b8866ef82d4f6c5361c8edd5199fe0cd19b947e3 and check that it is indeed bad. - Re-run the parent commit, i.e. 8cb92203bf223053ab6044211cfffe7b674cf526 and make really sure it works well. (In reply to comment #5) > Since this is tricky to reproduce please first double-check the bisected > commit: > - Re-run b8866ef82d4f6c5361c8edd5199fe0cd19b947e3 and check that it is > indeed bad. > - Re-run the parent commit, i.e. 8cb92203bf223053ab6044211cfffe7b674cf526 > and make really sure it works well. The bisect result is incorrect. Run it on 8cb92203bf223053ab6044211cfffe7b674cf526, it fails in the 29th run. We found many igt cases randomly have this error <3>[ 78.814855] [drm:i915_hangcheck_elapsed] *ERROR* Hangcheck timer elapsed... bsd ring idle Run core_getclient [root@x-pk1 tests]# ./core_getclient IGT-Version: 1.7-g27d37a1 (x86_64) (Linux: 3.15.0-rc8_drm-intel-fixes_ce9557_20140612+ x86_64) [root@x-pk1 tests]# dmesg -r | egrep "<[1-3]>" |grep drm [root@x-pk1 tests]# ./core_getclient IGT-Version: 1.7-g27d37a1 (x86_64) (Linux: 3.15.0-rc8_drm-intel-fixes_ce9557_20140612+ x86_64) [root@x-pk1 tests]# dmesg -r | egrep "<[1-3]>" |grep drm <3>[ 222.829544] [drm:i915_hangcheck_elapsed] *ERROR* Hangcheck timer elapsed... bsd ring idle Candidate for testing: git://people.freedesktop.org/~ickle/linux-2.6 requests (the seqno to request rework) (In reply to comment #8) > Candidate for testing: > > git://people.freedesktop.org/~ickle/linux-2.6 requests > > (the seqno to request rework) Test on branch requests, commit 299421c8f49d90fd24d4bdacd01cb91b3606a95a. Run ./gem_ring_sync_copy 20 cycles, it doesn't have this error. Run ./core_getclient cycles, it has this error once. [ 266.671526] [drm:i915_gem_open] [ 266.671529] [drm:i915_gem_object_create_stolen] creating stolen object: size=1000 [ 266.671532] [drm:i915_pages_create_for_stolen] offset=0x82a000, size=4096 [ 266.671613] [drm:i915_gem_open] [ 266.671615] [drm:i915_gem_object_create_stolen] creating stolen object: size=1000 [ 266.671616] [drm:i915_pages_create_for_stolen] offset=0x82b000, size=4096 [ 269.806762] [drm:i915_hangcheck_elapsed] *ERROR* Hangcheck timer elapsed... bsd ring idle [ 269.806925] [drm:intel_crtc_set_config] [CRTC:8] [FB:65] #connectors=1 (x y) (0 0) [ 269.806931] [drm:intel_set_config_compute_mode_changes] computed changes for [CRTC:8], mode_changed=0, fb_changed=0 [ 269.806933] [drm:intel_modeset_stage_output_state] [CONNECTOR:14:VGA-1] to [CRTC:8] [ 269.806934] [drm:intel_modeset_stage_output_state] [CONNECTOR:30:DP-3] to [CRTC:12] [ 269.806937] [drm:intel_crtc_set_config] [CRTC:12] [FB:65] #connectors=1 (x y) (0 0) [ 269.806939] [drm:intel_set_config_compute_mode_changes] computed changes for [CRTC:12], mode_changed=0, fb_changed=0 [ 269.806940] [drm:intel_modeset_stage_output_state] [CONNECTOR:14:VGA-1] to [CRTC:8] [ 269.806941] [drm:intel_modeset_stage_output_state] [CONNECTOR:30:DP-3] to [CRTC:12] [ 269.814505] core_getclient: executing [ 269.815072] [drm:i915_gem_open] [ 269.815074] [drm:i915_gem_object_create_stolen] creating stolen object: size=1000 [ 269.815077] [drm:i915_pages_create_for_stolen] offset=0x82a000, size=4096 [ 269.815166] [drm:i915_gem_open] [ 269.815168] [drm:i915_gem_object_create_stolen] creating stolen object: size=1000 [ 269.815170] [drm:i915_pages_create_for_stolen] offset=0x82b000, size=4096 [ 269.815300] [drm:intel_crtc_set_config] [CRTC:8] [FB:65] #connectors=1 (x y) (0 0) [ 269.815304] [drm:intel_set_config_compute_mode_changes] computed changes for [CRTC:8], mode_changed=0, fb_changed=0 [ 269.815306] [drm:intel_modeset_stage_output_state] [CONNECTOR:14:VGA-1] to [CRTC:8] [ 269.815307] [drm:intel_modeset_stage_output_state] [CONNECTOR:30:DP-3] to [CRTC:12] [ 269.815309] [drm:intel_crtc_set_config] [CRTC:12] [FB:65] #connectors=1 (x y) (0 0) [ 269.815311] [drm:intel_set_config_compute_mode_changes] computed changes for [CRTC:12], mode_changed=0, fb_changed=0 [ 269.815313] [drm:intel_modeset_stage_output_state] [CONNECTOR:14:VGA-1] to [CRTC:8] [ 269.815314] [drm:intel_modeset_stage_output_state] [CONNECTOR:30:DP-3] to [CRTC:12] [ 269.823407] core_getclient: executing [ 269.823965] [drm:i915_gem_open] [ 269.823968] [drm:i915_gem_object_create_stolen] creating stolen object: size=1000 [ 269.823970] [drm:i915_pages_create_for_stolen] offset=0x82a000, size=4096 [ 269.824052] [drm:i915_gem_open] [ 269.824054] [drm:i915_gem_object_create_stolen] creating stolen object: size=1000 [ 269.824056] [drm:i915_pages_create_for_stolen] offset=0x82b000, size=4096 [ 269.824125] [drm:intel_crtc_set_config] [CRTC:8] [FB:65] #connectors=1 (x y) (0 0) [ 269.824128] [drm:intel_set_config_compute_mode_changes] computed changes for [CRTC:8], mode_changed=0, fb_changed=0 [ 269.824130] [drm:intel_modeset_stage_output_state] [CONNECTOR:14:VGA-1] to [CRTC:8] [ 269.824131] [drm:intel_modeset_stage_output_state] [CONNECTOR:30:DP-3] to [CRTC:12] [ 269.824133] [drm:intel_crtc_set_config] [CRTC:12] [FB:65] #connectors=1 (x y) (0 0) [ 269.824135] [drm:intel_set_config_compute_mode_changes] computed changes for [CRTC:12], mode_changed=0, fb_changed=0 [ 269.824136] [drm:intel_modeset_stage_output_state] [CONNECTOR:14:VGA-1] to [CRTC:8] [ 269.824137] [drm:intel_modeset_stage_output_state] [CONNECTOR:30:DP-3] to [CRTC:12] Ok, the bug is still present. Oh well, I was being hopeful. I am pretty sure it is not a regression (it looks like a random hw issue), and we have a kernel workaround inplace should this misbehaviour impact real workarounds. s/real workarounds/real workloads/ We still confuse with this bug. The result is unstable and it impacts most of igt cases. [root@x-pk5 tests]# ./core_getversion IGT-Version: 1.8-gc049c39 (x86_64) (Linux: 3.18.0-rc3_drm-intel-nightly_de6d6c_20141111+ x86_64) [root@x-pk5 tests]# dmesg -r|egrep "<[1-4]>"|grep drm <3>[ 112.807912] [drm:i915_hangcheck_elapsed [i915]] *ERROR* Hangcheck timer elapsed... bsd ring idle Created attachment 115002 [details]
dmesg(kernel 333cf6_20150403)
run ./core_getversion 20 cycles, this error still exists.
Bug scrub: Lower priority as it deals with IRL. Assigned to Christophe to check if still reproduced I launched gem_ring_sync_copy and core_getversion 30 times each, I don't see "*ERROR* Hangcheck timer elapsed... bsd ring idle" in kernel log Not reproduced with kernel 4.3-rc4 So closed as not reproduced Never fixed, just undertested. *** This bug has been marked as a duplicate of bug 94307 *** My ILK does not work at all. Assigned to Ricardo to check if the test is possible on your side. We do not have a ILK locally Closing as duplicate closed + fixed (see comment #18 on bug 94307) |
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.
Created attachment 99006 [details] dmesg System Environment: -------------------------- Platform: Ironlake Kernel: (drm-intel-nightly)2be456541ea41728002ccca2de5235f48d14326e Bug detailed description: ------------------------- It randomly causes [drm:i915_hangcheck_elapsed] *ERROR* Hangcheck timer elapsed... bsd ring idle on Ironlake with -nightly and -queued kernel.It happens 1 in 10 runs. Run 20 cycles on -fixes kernel,it works well. output: IGT-Version: 1.6-g351e7d3 (x86_64) (Linux: 3.15.0-rc3_drm-intel-nightly_2be456_20140514+ x86_64) Test requirement not met in function __real_main319, file gem_ring_sync_copy.c:342: Last errno: 0, Success Test requirement: (!(data.render.copy)) no render-copy function Subtest sync-render-blitter-write-read: SKIP Subtest sync-render-blitter-read-write: SKIP Subtest sync-render-blitter-write-write: SKIP Subtest sync-blitter-render-write-read: SKIP Subtest sync-blitter-render-read-write: SKIP Subtest sync-blitter-render-write-write: SKIP Reproduce steps: ---------------- 1. ./gem_ring_sync_copy