HW overlay + animation causes "lost UI shared context" |
|||||||
Issue descriptionToT as of 7/26 ----------------- Repro step 1) Start ARC++ application in windowed mode 2) alt + ] Expected: the window bounces Actual: Screen turns black, then comes back. A few notes: * Works fine with night light enabled so I guess this has something to do with HW overlay. tfiga@ is this related to the fix you're working on? in /var/log/ui/ui.LATEST drv: DRM_IOCTL_I915_GEM_CREATE failed (size=0) Unexpected crash report id length System crash_reporter failed to process crash report. Report Id: device-enumerator: scan all dirs device-enumerator: scanning /sys/bus device-enumerator: scanning /sys/class chrome log: [7327:7434:0727/070058.830004:VERBOSE1:drm_device_handle.cc(83)] Succeeded authenticating /dev/dri/card0 in 0 ms with 1 attempt(s) [7327:7327:0727/070058.837467:VERBOSE1:drm_display_host_manager.cc(247)] Got display event ADD for /dev/dri/card0 [7327:7327:0727/070058.839931:VERBOSE1:drm_display_host_manager.cc(247)] Got display event ADD for /dev/dri/card1 [7327:11462:0727/070058.840076:VERBOSE1:drm_device_handle.cc(83)] Succeeded authenticating /dev/dri/card1 in 0 ms with 1 attempt(s) [7327:7327:0727/070058.840534:ERROR:gpu_process_transport_factory.cc(930)] Lost UI shared context. and here is the warning in dmesg: [30635.626707] CPU: 3 PID: 31572 Comm: DrmThread Tainted: G W 4.4.70-11150-ga6edf7ec6181 #1 [30635.626713] Hardware name: Google Eve/Eve, BIOS Google_Eve.9584.18.0 06/26/2017 [30635.626719] 0000000000000286 000000003c230237 ffff880403b1fb68 ffffffffbce93a1d [30635.626734] ffff880403b1fbb0 0000000000000009 ffff880403b1fba0 ffffffffbcc6981c [30635.626748] ffffffffbd026f27 ffff88046b473000 0000000000000001 ffff88046cb5c060 [30635.626764] Call Trace: [30635.626783] [<ffffffffbce93a1d>] dump_stack+0x4d/0x63 [30635.626798] [<ffffffffbcc6981c>] warn_slowpath_common+0x9f/0xb8 [30635.626809] [<ffffffffbd026f27>] ? intel_dp_detect+0x24a/0x6f1 [30635.626820] [<ffffffffbcc6988c>] warn_slowpath_fmt+0x57/0x73 [30635.626832] [<ffffffffbd026f27>] intel_dp_detect+0x24a/0x6f1 [30635.626846] [<ffffffffbcf6acc2>] drm_helper_probe_single_connector_modes_merge_bits+0xcd/0x3b6 [30635.626860] [<ffffffffbd30c78b>] ? mutex_unlock+0x22/0x34 [30635.626872] [<ffffffffbcf6afbe>] drm_helper_probe_single_connector_modes+0x13/0x15 [30635.626884] [<ffffffffbcf8e6b7>] drm_mode_getconnector+0xeb/0x35b [30635.626897] [<ffffffffbcc8b697>] ? __might_sleep+0x9f/0xa6 [30635.626910] [<ffffffffbcf79c10>] drm_ioctl+0x2a8/0x3dd [30635.626922] [<ffffffffbcf8e5cc>] ? drm_mode_connector_property_set_ioctl+0x55/0x55 [30635.626936] [<ffffffffbce431d8>] ? selinux_file_ioctl+0x141/0x185 [30635.626948] [<ffffffffbcd749cd>] do_vfs_ioctl+0x42e/0x501 [30635.626959] [<ffffffffbce3ca88>] ? security_file_ioctl+0x35/0x49 [30635.626970] [<ffffffffbcd74af7>] SyS_ioctl+0x57/0x79 [30635.626981] [<ffffffffbd30f2e1>] entry_SYSCALL_64_fastpath+0x1c/0x74 [30635.626991] ---[ end trace 340be61aff333571 ]---
,
Jul 27 2017
oops, wrong person/ldap. -> dcastagna@
,
Jul 27 2017
,
Jul 27 2017
I guess the question "is this related to the fix you're working on?" was directed to me. This issue doesn't seem related to http://crbug.com/746068 nor to http://crbug.com/743277 but it's a new bug. Thanks for reporting it. https://chromium-review.googlesource.com/c/590275/ to fix this issue.
,
Jul 28 2017
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/d457f1097aeef9de0f7559dddad1f6cec6b985c2 commit d457f1097aeef9de0f7559dddad1f6cec6b985c2 Author: Daniele Castagna <dcastagna@chromium.org> Date: Fri Jul 28 17:14:58 2017 ozone: Mark OverlayCheck_Params as non candidate if it can't be handled. DrmOverlayManager::CheckOverlaySupport marks candidates that can't be used as overlays with overlay_handled = false. In particular, if a candidate has a subpixel positioning, CheckOverlaySupport will add an empty OverlaySurfaceCandidate and mark it as overlay_handled = false before sending it to the GPU process for validation. Prior to this patch the GPU process would try to allocate a 0 sized buffer and fail in minigbm. Causing the GPU process to crash. With this patch, we'll mark OverlayCheck_Params as non overlay candidate so that the GPU process won't even try to validate the surface. Bug: 749518 Change-Id: Ib2e8e14e2fdc6506fab7ddd441488a87eec592f4 Reviewed-on: https://chromium-review.googlesource.com/590275 Reviewed-by: Robert Kroeger <rjkroege@chromium.org> Reviewed-by: Tom Sepez <tsepez@chromium.org> Commit-Queue: Daniele Castagna <dcastagna@chromium.org> Cr-Commit-Position: refs/heads/master@{#490438} [modify] https://crrev.com/d457f1097aeef9de0f7559dddad1f6cec6b985c2/ui/ozone/common/gpu/ozone_gpu_message_params.cc
,
Jul 31 2017
,
Jul 31 2017
[Auto-generated comment by a script] We noticed that this issue is targeted for M-61; it appears the fix may have landed after branch point, meaning a merge might be required. Please confirm if a merge is required here - if so add Merge-Request-61 label, otherwise remove Merge-TBD label. Thanks.
,
Jul 31 2017
AFAIU this regression happened with crrev.com/3817977756471f18782ea88f867fe22f24a7669e that didn't make it to 61. A merge to 61 is not needed.
,
Aug 2 2017
jlklein@ reported that he had similar problem on R61-9765.13.0 .
,
Aug 2 2017
Thanks Oshima. I can confirm that I'm seeing this every so often on R61.
,
Aug 3 2017
Every so often? The bug that oshima reported was reproduceable 100% of the time with his repro steps. Is it not the case for you in R61? jlklein@: Do you also get a drv: DRM_IOCTL_I915_GEM_CREATE failed (size=0) in /var/log/ui/ui.LATEST ?
,
Aug 3 2017
To be honest, I've flashed my device a ton of times today to a couple different R61 builds. In my current build, I'm not able to repro with those steps and I haven't seen the crash in a few hours. I'll keep an eye out, though.
,
Aug 3 2017
Please, in case it happens again, would you mind saving /var/log/ui/ui.LATEST and maybe the output of dmesg? I suspect it might be another bug, since this particular issue that Oshima described was happening because of a CL that landed after 61 branch. Feel free to open another bug if you notice the same issue again. Thank you!
,
Aug 3 2017
Will do. Thanks! |
|||||||
►
Sign in to add a comment |
|||||||
Comment 1 by tfiga@chromium.org
, Jul 27 2017Owner: osh...@chromium.org