[Bug 1104851] New: Hard freeze with INTEL GPU
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851 Bug ID: 1104851 Summary: Hard freeze with INTEL GPU Classification: openSUSE Product: openSUSE Distribution Version: Leap 15.0 Hardware: x86-64 OS: SUSE Other Status: NEW Severity: Critical Priority: P5 - None Component: X.Org Assignee: xorg-maintainer-bugs@forge.provo.novell.com Reporter: werner@suse.com QA Contact: xorg-maintainer-bugs@forge.provo.novell.com Found By: --- Blocker: --- Created attachment 779743 --> http://bugzilla.opensuse.org/attachment.cgi?id=779743&action=edit hwinfo on an older i915 I face a hard freeze if X server starts up ... sometimes it takes a while but then it freeze. Many times a see a black screen with a small vertical flakes on the left border. As this is the home workstation for my family[1] and me I need at least a workaround like a kernel/module parameter to avoid this. [1] My daugther use the workstation to do phase6 in english and french and my wife does here private correspondence -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851#c1
--- Comment #1 from Dr. Werner Fink
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851
Dr. Werner Fink
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851#c2
Stefan Dirsch
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851#c3
Dr. Werner Fink
Still Sandybridge. In case, instead of modesetting driver please also try xf86-video-intel driver. You could also try with
Option "AccelMethod" "UXA"
with intel driver.
Thanks a lot! It seems that in the kernels command line the option i915.modeset=1 together with Driver="intel" in the old /etc/X11/xorg.conf.d/99-local.conf do not work together. After using Driver="modesetting" and the sepcific option AccelMethod set to "UXA" it seems to be stable. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851#c4
--- Comment #4 from Stefan Dirsch
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851#c5
Neil Rickert
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851#c6
--- Comment #6 from Dr. Werner Fink
Hmm. Usually "modesetting" is less stable on older Intel platforms than the "intel" driver itself, since it it uses glamor/Mesa GL driver for 2D acceleration, and this can be more broken on older platforms than on newer ones. But ok, as long as it does work for you ...
But AccelMethod "UXA" is certainly no valid option for "modesetting" driver. Only "none" instead of "glamor" is, but this kills 2D acceleration and then you can use "fbdev" driver as well ...
Next problem: even with hibernate leads to freeze. With "intel" which kernel mode line options should I avoid and which should I set. Currently I have i915.modeset=1 i915.i915_enable_rc6 ... and some others which I can not remember and the system is currently down. (In reply to Neil Rickert from comment #5)
I have a laptop with sandybridge graphics, purchased in 2009 (I think). I have freezes.
What I have found out, is that the freeze appear to be related to brightness change. In particular, if the desktop dims the display, then it often freezes when it returns to full brightness.
Using KDE, I set it to never dim the display (but I do allow switching off the display after a period of idle). And I don't think I have had the desktop freeze since then.
I still get occasional freezes, though they are rare enough to not be a serious problem. When they occur, it is during bootup, before the desktop environment is fully up and running.
Yep I can second this ... sometimes it takes a while then it had freezed likly at xscreensaver or similar ... sometimes it had happen at start of greeter of the sddm. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851#c7
--- Comment #7 from Dr. Werner Fink
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851#c8
--- Comment #8 from Dr. Werner Fink
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851#c9
Stefan Dirsch
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851#c10
--- Comment #10 from Dr. Werner Fink
I have to switch to the old kernel of the former Leap to get a working setup ... 4.12 does not work with old sandybridge
Interesting that the three greeters running with the old kernel do not show load higher then 1% within top whereas wth 4.12 those processes have up to 8% -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851#c11
--- Comment #11 from Dr. Werner Fink
Looks like a regression in Kernel 4.12, which may meanwhile been adressed in newer kernels. I would first check kernel-of-the-day before digging into this any further.
Hmmm ... Solution 1: Following actions will be done: ignore the warning of a broken system (requires:glibc) do not install kernel-default-4.18.0-3.1.g07db4aa.i586 Solution 2: do not install kernel-default-4.18.0-3.1.g07db4aa.i586 seems that the libc on Leap 15.0 is a bit to old or simply missing a feature -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851#c12
--- Comment #12 from Dr. Werner Fink
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851#c13
--- Comment #13 from Dr. Werner Fink
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851#c14
--- Comment #14 from Dr. Werner Fink
Why the heck zypper tries to install i586 instead oc x86_64???
Hard way with rpm rpm -ihv http://download.opensuse.org/repositories/Kernel:/HEAD/standard/x86_64/kerne... --force ... and booting 4.18 ... switching to first greeter, blank, blank, freeze. The regression is not fixed in 4.18.0 -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851
Dr. Werner Fink
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851#c20
--- Comment #20 from Dr. Werner Fink
So the 4.4 kernel from leap 42.* works for you, right?
Exactly
I am afraid, there is nothing else than bisecting the kernel. As a starting point, you can use Takashi's old kernels: https://build.opensuse.org/project/subprojects/home:tiwai
I would try 4.8 and then 4.6 or 4.10 depending on the result.
https://build.opensuse.org/project/show/home:tiwai:kernel:4.8 https://build.opensuse.org/project/show/home:tiwai:kernel:4.6 https://build.opensuse.org/project/show/home:tiwai:kernel:4.10 etc.
Meanwhile I've a cheap AMD graphic card in the system and the INTEL GPU is diabled by kernel command line. And this is a system which is used by my daugther as well as by my whife. First of all I'll try the kernel of the day if this works with INTEL it is OK for a testm then I switch back to kernel of Leap 15 -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851#c21
Dr. Werner Fink
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851#c22
--- Comment #22 from Stefan Dirsch
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851#c23
--- Comment #23 from Dr. Werner Fink
Too bad. But thanks a lot for giving it a try!
No problem ;) The only nasty thing is that the BIOS of the CPU detects which line of the monitor is enabled (Display Port or HDMI1) and you have to reset the system to be able to switch and seee something. And there is no option in the CMOS/BIOS to ignore this ... only the order of using Online or PCI/PCI Express can be changed but no headless configuration is possible ;} -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851
http://bugzilla.opensuse.org/show_bug.cgi?id=1104851#c24
--- Comment #24 from Dr. Werner Fink
participants (1)
-
bugzilla_noreply@novell.com