Created attachment 117222 [details] contains dmesg and /sys/class/drm/card0/error files. I am running fedora 22 with kernel 4.0.8 on a x86_64 system. More packages: gnome-shell-3.16.3-1.fc22.x86_64 firefox-39.0-8.fc22.x86_64 gstreamer1-1.4.5-1.fc22.x86_64 libdrm-2.4.61-3.fc22.x86_64 Hardware: iGPU of an Intel Core i5 6x0 series Since today (since updating to kernel 4.0.8, this issue was not present in kernel 4.0.7) I got this issue. Steps to reproduce: 1. log in to gnome-shell session on X 2. open firefox (works with a fresh profile) 3. load youtube.com (I tested this several times, always worked) What happens on the UI: Gnome-shell crashes (window decorations are gone) firefox repaints its application window wrong, displaying content at wrong positions gnome-session tells me that it crashed and can't recover. I am forced to log out. In syslog (systemd journal) I see this: firefox.desktop[3110]: libva info: VA-API version 0.37.0 firefox.desktop[3110]: libva info: va_getDriverName() returns 0 firefox.desktop[3110]: libva info: Trying to open /usr/lib64/dri/i965_drv_video.so firefox.desktop[3110]: libva info: Found init function __vaDriverInit_0_37 firefox.desktop[3110]: libva info: va_openDriver() returns 0 firefox.desktop[3110]: (firefox:3467): GStreamer-CRITICAL **: gst_mini_object_lock: assertion 'object != NULL' failed firefox.desktop[3110]: (firefox:3467): GStreamer-CRITICAL **: gst_memory_unmap: assertion 'mem != NULL' failed firefox.desktop[3110]: (firefox:3467): GStreamer-CRITICAL **: gst_memory_get_sizes: assertion 'mem != NULL' failed firefox.desktop[3110]: (firefox:3467): GStreamer-CRITICAL **: gst_mini_object_lock: assertion 'object != NULL' failed firefox.desktop[3110]: (firefox:3467): GStreamer-CRITICAL **: gst_memory_unmap: assertion 'mem != NULL' failed firefox.desktop[3110]: (firefox:3467): GStreamer-CRITICAL **: gst_memory_get_sizes: assertion 'mem != NULL' failed firefox.desktop[3110]: (firefox:3467): GStreamer-CRITICAL **: gst_mini_object_unlock: assertion 'object != NULL' failed firefox.desktop[3110]: (firefox:3467): GStreamer-CRITICAL **: gst_mini_object_unref: assertion 'mini_object != NULL' failed firefox.desktop[3110]: (firefox:3467): GStreamer-CRITICAL **: gst_mini_object_lock: assertion 'object != NULL' failed firefox.desktop[3110]: (firefox:3467): GStreamer-CRITICAL **: gst_memory_unmap: assertion 'mem != NULL' failed firefox.desktop[3110]: (firefox:3467): GStreamer-CRITICAL **: gst_memory_get_sizes: assertion 'mem != NULL' failed firefox.desktop[3110]: (firefox:3467): GStreamer-CRITICAL **: gst_mini_object_unlock: assertion 'object != NULL' failed firefox.desktop[3110]: (firefox:3467): GStreamer-CRITICAL **: gst_mini_object_unref: assertion 'mini_object != NULL' failed firefox.desktop[3110]: (firefox:3467): GStreamer-CRITICAL **: gst_mini_object_lock: assertion 'object != NULL' failed firefox.desktop[3110]: (firefox:3467): GStreamer-CRITICAL **: gst_memory_unmap: assertion 'mem != NULL' failed firefox.desktop[3110]: (firefox:3467): GStreamer-CRITICAL **: gst_memory_get_sizes: assertion 'mem != NULL' failed firefox.desktop[3110]: (firefox:3467): GStreamer-CRITICAL **: gst_mini_object_unlock: assertion 'object != NULL' failed firefox.desktop[3110]: (firefox:3467): GStreamer-CRITICAL **: gst_mini_object_unref: assertion 'mini_object != NULL' failed firefox.desktop[3110]: (firefox:3467): GStreamer-CRITICAL **: gst_mini_object_lock: assertion 'object != NULL' failed with the last 5 lines repeating some hundred or thousand times until I kill firefox (with sigterm or SAK). Inbetween I see this: kernel: [drm] stuck on render ring kernel: [drm] GPU HANG: ecode 5:0:0x9ffdffff, in gnome-shell [1926], reason: Ring hung, action: reset kernel: [drm] GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace. kernel: [drm] Please file a _new_ bug report on bugs.freedesktop.org against DRI -> DRM/Intel kernel: [drm] drm/i915 developers can then reassign to the right component if it's not a kernel issue. kernel: [drm] The gpu crash dump is required to analyze gpu hangs, so please always attach it. kernel: [drm] GPU crash dump saved to /sys/class/drm/card0/error and kernel: drm/i915: Resetting chip after gpu hang
I would not have expected that to be a kernel bug. But if you switch back to 4.0.7 keeping everything the same and then you cannot reproduce, it will be pretty quick to bisect.
You are right. I can reproduce it with older kernel versions too, so it is no recently introduced kernel bug. I thought so since I haven't seen it before about the time I updated the kernel for the last time.
I am forced to remove gstreamer1-vaapi because of this problem. I tried Linux 3.18.7, 3.19.3, 4.0.8, 4.1.6 and none of them works, so I think it is not related to kernel version. I also tried libva 1.5.1, 1.4.1, 1.3.1 and all corresponding libva-intel-driver versions. All of them cause GPU hang. libdrm-2.4.61-3.fc22.x86_64 mesa-10.6.3-1.20150729.fc22.x86_64 xorg-x11-server-Xorg-1.17.2-2.fc22.x86_64 xorg-x11-drv-intel-2.99.917-12.20150615.fc22.x86_64 libva info: VA-API version 0.37.0 libva info: va_getDriverName() returns 0 libva info: Trying to open /usr/lib64/dri/i965_drv_video.so libva info: Found init function __vaDriverInit_0_37 libva info: va_openDriver() returns 0 vainfo: VA-API version: 0.37 (libva 1.5.1) vainfo: Driver version: Intel i965 driver for Intel(R) Ironlake Desktop - 1.5.1 vainfo: Supported profile and entrypoints VAProfileMPEG2Simple : VAEntrypointVLD VAProfileMPEG2Main : VAEntrypointVLD VAProfileH264ConstrainedBaseline: VAEntrypointVLD VAProfileH264Main : VAEntrypointVLD VAProfileH264High : VAEntrypointVLD VAProfileNone : VAEntrypointVideoProc
Seems like this is a libva bug rather than one in Mesa?
I think yes, this only happens when programs use libva. vlc can also cause the crash.
*** This bug has been marked as a duplicate of bug 91254 ***
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.