I have had lots of troubles with GNOME Shell halting the cpu requiring power off. I could provoke the same situation using the Weston reference implementation Computer is MacBook Pro 2015 13" Have external display to the right with display bottoms aligned Launch a terminal window on the external display put it near lower left corner around the lower right corner of the internal display move the mouse cursor back and forth from the internal display to into the terminal window on the external display BUG within about 30 s, cpu halts, computer has to be powered off static video from the halt moment is displayed Halting the cpu is not helpful at all This problem did not exist in 1.08/15.10 The cause seems to be drawing the mouse cursor on multiple displays without an external display no problems. GNOME Shell does not crash until - a window exists https://bugzilla.gnome.org/show_bug.cgi?id=765285 There is some special handling of the mouse cursor on external displays within and inch of a cross-display boundary if a window edge is placed here, say a maximized window on the external display, displaying a different mouse cursor shape, the mouse shape change close to a cross-display boundary causes the halt to occur almost immediately dpkg -s weston | egrep ^V Version: 1.9.0-3 dpkg -s libwayland-server0 | egrep ^V Version: 1.10.0-1~xenial0 lsb_release -sd Ubuntu 16.04 LTS uname -r 4.4.0-21-generic
Sounds like a driver bug. It has nothing to do with Wayland, and I believe also gnome-shell and weston are just triggers, not at blame here. You say the problem did not exist on Weston 1.8.0? Does it occur with 1.8.0 now that you have the rest of the system upgraded? If yes, it is not a Weston regression but most likely with the drivers. If simply downgrading Weston alone does make the problem go away, then please try weston 1.10. If 1.10 is also problematic, it would be good to bisect which change in weston caused it.
This bugs hangs the CPU (as in not only the GPU) The system has to be powered off with a 4-second power button press It is clearly a kernel problem, does that mean it has to be drm? I was reading over here: https://01.org/linuxgraphics/documentation/how-report-bugs - So I should use drm-intel-nightly I can capture dmesg until halt on another networked machine I am not sure if cat /sys/class/drm/card0/error will be possible since the cpu halts? (kernel options: drm.debug=0x1e log_buf_len=1M) However, this is a MacBook Pro 2015 13", "everybody" has one. And it's the latest 16.04 "everybody" has that, too cpu/gpu: i5-5287U CPU @ 2.90GHz Two displays via two DisplayPort cables - it crashes with a single external display, too - it does not crash without external displays Running Weston from 16.04, it's 100% reproducible uname --kernel-release --machine 4.4.0-21-generic x86_64 I am running KDE/X now, b/c that does not crash.
(In reply to Harry from comment #0) > There is some special handling of the mouse cursor on external displays > within and inch of a cross-display boundary > if a window edge is placed here, say a maximized window on the external > display, displaying a different mouse cursor shape, the mouse shape change > close to a cross-display boundary causes the halt to occur almost immediately Sounds familiar, and I think we may have fixed this already. Can you try v4.6-rc5 or drm-intel-nightly branch of http://cgit.freedesktop.org/drm-intel?
With the following configuration on the processor i5-5200U 2.2 GHZ i can see the issue reported on this bug, i5-5200U 2.2 GHZ |=== Software information ===| ++ Kernel version : 4.6.0-rc7-drm-intel-nightly-ww20-commit-bcc6a84+ ++ Linux distribution : Ubuntu 15.10 ++ Architecture : 64-bit ++ Mesa version : 11.1.2 (git-7bcd827 ++ xf86-video-intel version : 2.99.917 ++ Xorg-Xserver version : 1.18.99.1 ++ DRM version : 2.4.67 ++ VAAPI version : Intel i965 driver for Intel(R) Broadwell - 1.7.1.pre1 (1.7.0-8-g2c1bec0) ++ Cairo version : 1.15.2 ++ Intel GPU Tools version : Tag [intel-gpu-tools-1.14-212-g1e9a3ac] / Commit [1e9a3ac] ++ Kernel driver in use : i915 ++ Hardware acceleration : Enabled ++ Bios revision : 1.69 ++ KSC revision : 1.69 |=== Hardware information ===| ++ Platform : BDW ++ Motherboard model : 80E5 ++ Motherboard type : Lenovo G50-80 Notebook ++ Motherboard manufacturer : LENOVO ++ CPU family : Core i5 ++ CPU information : Intel(R) Core(TM) i5-5200U CPU @ 2.20GHz ++ GPU Card : Intel Corporation Broadwell-U Integrated Graphics (rev 09) (prog-if 00 [VGA controller]) ++ Memory ram : 6 GB ++ Maximum memory ram allowed : 16 GB ++ Display resolution : 1366x768 ++ CPU's number : 4 Kernel information =============================================== commit bcc6a843e7e4a3f4794b90dbefb00174171365bd Author: Chris Wilson <chris@chris-wilson.co.uk> Date: Mon May 9 13:48:17 2016 +0100 drm-intel-nightly: 2016y-05m-09d-12h-47m-46s UTC integration manifest Kernel version : 4.6.0-rc7 Architecture : source amd64 all Homepage : http://www.kernel.org/
Is there work on this bug? It's making my life kind of difficult. I got out of nVidia unsupported hell by upgrading to an Intel 3x4k gpu. I get the halt approximately weekly on KDE/X which is the most stable configuration I have found. When I was on GNOME Shell 3.18 or 3.20 I had an uptime of approximately 2 hours And the 16.04 Weston has the problem 100% reproducible as described in this report. Although halts are horrible, since it is reproducible it should be "easy" to fix I did not see this prior to upgrading to 16.04, which likely means that 15.10 with 4.2.0-35.40 did not have it. I am currently on 4.4.0-21. I believe it has to do with the mouse cursor on multiple displays and the phantom redraws of the displays below the mouse cursor that occur when the mouse cursor moves over a display boundary. I just lost data again from a halt where fsck removed data that I considered safely written to disk. Like most people, I hate losing data.
Not even with a legacy KDE/X can one avoid this problem. KDE halts occasionally when using the "Present Windows" desktop effects function that displays all windows side-by-side. Halting the cpu leads to data loss, which means this bug is Critical. And fully reproducible.
4.4.0-22-generic
Updating priority accordingly: System halt
Despite everything I don’t this bug still halts my machine on occasion - no GNOME - no Wayland - the most retarded KDE available Anyone with multiple displays on Broadwell and possibly other chips is hosed. Now it ran for 10 days.
I did some tests and concluded that setting kernel command line parameter i915.enable_rc6=0 made Wayland not crash anymore - it appears acceleration is then off in Wayland - GNOME 3.18/Wayland crashed within minutes the setting can be checked: cat /sys/module/i915/parameters/enable_rc6 0 documentation: modinfo i915 | less rC-state is from ACPI, c0 is power on, c6 is deep power down, possibly off. had no effect: intel_idle.max_cstate=1 intel_pstate=disable was already set: for g in /sys/devices/system/cpu/cpu[0-9]/cpufreq/scaling_governor; do echo powersave > $g; echo cpu$i: $(cat $g); ((i++)); done
(In reply to Harry from comment #10) > I did some tests and concluded that setting kernel command line parameter > i915.enable_rc6=0 > made Wayland not crash anymore > - it appears acceleration is then off in Wayland > - GNOME 3.18/Wayland crashed within minutes > > the setting can be checked: > cat /sys/module/i915/parameters/enable_rc6 > 0 > > documentation: > modinfo i915 | less > > rC-state is from ACPI, c0 is power on, c6 is deep power down, possibly off. > > had no effect: > intel_idle.max_cstate=1 > intel_pstate=disable > > was already set: > for g in /sys/devices/system/cpu/cpu[0-9]/cpufreq/scaling_governor; do > echo powersave > $g; echo cpu$i: $(cat $g); ((i++)); done setted kernel command line parameter i915.enable_rc6=0 and These issue is not reproduced and hardware acceleration is active with the following configuration === Software information === Kernel version 4.8.0-rc4drm-intel-nighly-ww36-commit-f91144a+ Linux distribution Ubuntu 16.04.1 LTS Architecture 64-bit Mesa version 11.2.2 (git-3a9f628 xf86-video-intel version 2.99.917 DRM version 2.4.70 Cairo version 1.15.2 Bios revision 1.69 KSC revision 1.69 === Hardware information === Platform BDW Motherboard type Lenovo G50-80 Notebook Motherboard manufacturer LENOVO CPU family Core i5 CPU information Intel(R) Core(TM) i5-5200U CPU @ 2.20GHz GPU Card Intel Corporation Broadwell-U Integrated Graphics (rev 09) (prog-if 00 [VGA controller]) Memory ram 6 GB Maximum memory ram allowed 16 GB Hard drive capacity 74GiB (80GB)
mailto:ismail@inbox247.com - can you let us know if you are still having problems with the following environment (used also by Maria): Kernel version 4.8.0-rc4drm-intel-nighly-ww36-commit-f91144a+ Linux distribution Ubuntu 16.04.1 LTS
This problem is fixed by kernel command line parameter i915.enable_rc6=0 Some apps in GNOME/Wayland 3.20 hakts in the same way when moved between displays. https://bugs.freedesktop.org/show_bug.cgi?id=98985
Reporter, Harry, can this be resolved or what?
yes. The same halt occurs in https://bugs.freedesktop.org/show_bug.cgi?id=98985
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.