| Summary: | System hangs when some video output is disabled/enabled on Skylake platform in double/triple display configuration | ||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Product: | DRI | Reporter: | Edgar Cherkasov <echerkasov> | ||||||||||||||||
| Component: | DRM/Intel | Assignee: | Intel GFX Bugs mailing list <intel-gfx-bugs> | ||||||||||||||||
| Status: | CLOSED FIXED | QA Contact: | Intel GFX Bugs mailing list <intel-gfx-bugs> | ||||||||||||||||
| Severity: | normal | ||||||||||||||||||
| Priority: | medium | CC: | intel-gfx-bugs, jaffer.intel | ||||||||||||||||
| Version: | unspecified | ||||||||||||||||||
| Hardware: | x86-64 (AMD64) | ||||||||||||||||||
| OS: | Linux (All) | ||||||||||||||||||
| Whiteboard: | |||||||||||||||||||
| i915 platform: | SKL | i915 features: | display/watermark | ||||||||||||||||
| Attachments: |
|
||||||||||||||||||
|
Description
Edgar Cherkasov
2016-04-04 16:32:08 UTC
Created attachment 123470 [details] [review] FB Connector Patch #1 Created attachment 123471 [details] [review] FB Connector Patch #2 There are a couple of odd things that I’ve seen while looking over the first dmesg log. The first is that the sink devices advertise that they only support up to 2.7Gbps transfer rate, which makes me suspect that they are older models. Could you please tell us the make and model numbers of the displays in question? The second is that there are quite a number of watermark WARNs. The kernel log lists the version as 4.6.999, so I really don’t have a clue specifically how recent this is, and I have to assume that it’s based on the kernel.org kernel. There have been some fixes in the watermarking area recently, so it would be helpful to try and reproduce this issue using the same hardware and drm-intel-nightly (assuming this is not drm-intel with the version info changed.) Finally, although the stated reason for the patches referenced in the thread is DP MST, it might be worth trying the two attached patches on top of drm-intel-nightly, assuming the problem reproduces against that kernel. The issue being fixed is exacerbated by MST, but it is a general racy problem with the frame buffer connector counting if I’m interpreting the information correctly. (In reply to Jim Bride from comment #3) Hi Jim, thanks a lot for your feedback! First of all let me inform your regarding the state of the main problem(Linux hangs when video output is disabled/enabled during boot) - it can't be reproduced with vanilla 4.6.rc5 kernel. And please find answers to your questions below: > There are a couple of odd things that I’ve seen while looking over the first > dmesg log. The first is that the sink devices advertise that they only > support up to 2.7Gbps transfer rate, which makes me suspect that they are > older models. Could you please tell us the make and model numbers of the > displays in question? This can be caused by our DP to VGA and DP to DVI converters. If needed I can repeat the test using a "pure" DP display without any converters. Description of our current h/w configuration: 1) First DDI output -> DP to DVI-I converter(Chrontel CH7318C) -> DVI display(Acer V223W); 2) Second DDI output -> DP to VGA converter(noname) -> VGA display(RoverScan JM678); 3) eDP output -> on-board eDP to LVDS converter(PTN3460) -> LVDS display(LG LB121S03). > The second is that there are quite a number of > watermark WARNs. The kernel log lists the version as 4.6.999, so I really > don’t have a clue specifically how recent this is, and I have to assume that > it’s based on the kernel.org kernel. These tests were performed with the latest at that time(first days of April) version of drm-intel-nightly from prebuilt Ubuntu repository: http://kernel.ubuntu.com/~kernel-ppa/mainline/drm-intel-nightly/ If needed, I can try to find the exact version. > There have been some fixes in the > watermarking area recently, so it would be helpful to try and reproduce this > issue using the same hardware and drm-intel-nightly (assuming this is not > drm-intel with the version info changed.) I've tried the latest drm-intel-nightly from the same Ubuntu repository: cod/tip/drm-intel-nightly/2016-05-05 bbb6a0d936b5648b31f173d47f93f389a6c6b653 and, unfortunately, all the same - system hangs during the execution of kms_plane test from intel-gpu-tools. The only difference is broken S3(but it does not relate to Intel drm). Please find logs in the attachment. > Finally, although the stated > reason for the patches referenced in the thread is DP MST, it might be worth > trying the two attached patches on top of drm-intel-nightly, assuming the > problem reproduces against that kernel. The issue being fixed is > exacerbated by MST, but it is a general racy problem with the frame buffer > connector counting if I’m interpreting the information correctly. Thanks! I will give it a try and provide a feedback in 1-2 days. Summary: As the problem can't be reproduced on vanilla 4.6.rc5 kernel, the bug can be closed, but: - I still want to test the attached patches and provide feedback; - please tell if additional experiments/information are needed(testing on pure DP display, testing different kernels, different versions of drm-intel-nightly, ...). Best regards, Edgar Cherkasov. Created attachment 123494 [details]
kms test results
On "Testing connector eDP-1 using pipe C plane 1" system hangs
Created attachment 123495 [details]
test script
Created attachment 123496 [details]
errors on display before system hangs
Intel OTC graphics driver QA performed tests using eDP+HDMI on 3 different SKL platforms… (a production one, SKL-y and SKL-H) In all three configurations, the bug was not reproducible using latest kernel and graphics stack (The platform never hung) Next QA test is to connect two displays on the platform (+eDP that makes 3 pipes), unfortunately the platforms QA has available have only 2 pipes available , eDP+HDMI. Testing will be done later ths week on a non-production platform with 3 pipes in order to test this as described in the bug. Note that the bug report is focused on Fedora 22, a distro that is a year old. (In reply to dog from comment #8) > Intel OTC graphics driver QA performed tests using eDP+HDMI on 3 different > SKL platforms… (a production one, SKL-y and SKL-H) > > In all three configurations, the bug was not reproducible using latest > kernel and graphics stack (The platform never hung) > > Next QA test is to connect two displays on the platform (+eDP that makes 3 > pipes), unfortunately the platforms QA has available have only 2 pipes > available , eDP+HDMI. Testing will be done later ths week on a > non-production platform with 3 pipes in order to test this as described in > the bug. > > Note that the bug report is focused on Fedora 22, a distro that is a year > old. Thank you for the information ! I tried to boot Fedora 22 with 4.0.4-301.fc22 kernel with enabled preliminary hw support, as result system hangs then booting. Second I boot system without preliminary hw support and run i915 manually. As result - system hangs then i915 loads. Best regards, Edgar Cherkasov. Hi Jim, I have tried latest drm-intel-nightly(from Ubuntu repository) with patches from the attachments: - system still hangs during kms_plane test(in the same place with the same error as without these patches. Refer to log here: https://bugs.freedesktop.org/show_bug.cgi?id=94822#c5 ) - but I haven't tried yes to reproduce the "main" problem with these patches(kernel hangs during boot when several displays are connected). I'm going to provide the results of this test on Tuesday. It seems that there is some mess in the discussion. Let me put it in order: - "main problem"(kernel freezes during boot) can't be reproduced with vanilla 4.6.rc5 kernel; - but kms_plane test still hangs on all available kernels/drm-intel-nighty with and without patches from https://bugs.freedesktop.org/show_bug.cgi?id=94822#c1 and https://bugs.freedesktop.org/show_bug.cgi?id=94822#c2 The second problem is not important for me, but I can conduct more tests or collect more detailed debug output if needed. Please also note that I'm using SKL-U SoC system, not Y or H variant. Best regards, Edgar Cherkasov. Created attachment 123571 [details] SKL-Y_4.6-rc6-447f243_kern.log I reproduced a similar problem on SKL-Y with fresh kernel. Steps: ------- 1. Connect HDMI and DP screens 2. Boot 3. xrandr --output eDP1 --off 4. xrandr --outout eDP1 --auto Actual result: ---------------- 4. eDP stays black. Command does not respond Info ----- Could be the same as bug 89055 Not reproduced with 4.6 rc6 from https://cgit.freedesktop.org/drm-intel/ commit 871c895c2d3192a93a701b31d2576184ab070c66 Author: Rodrigo Vivi <rodrigo.vivi@intel.com> Date: Thu May 5 16:49:48 2016 -0700 drm-intel-nightly: 2016y-05m-05d-23h-49m-21s UTC integration manifest Setup ------ Hardware Platform: CPU : Intel(R) Core(TM) m5-6Y57 CPU @ 1.10GHz (family: 6, model: 78 stepping: 3) MCP : SKL-Y D1 2+2 QDF : QJK9 CPU : SKL D0 Chipset PCH: Sunrise Point LP C0 CRB : SKY LAKE Y LPDDR3 RVP3 CRB FAB2 Software Linux OS : Ubuntu 16.10 64 bits BIOS : SKLSE2R1.R00.B104.B01.1511110114 ME FW : 11.0.0.1191 Ksc (EC FW): 1.20 Kernel: drm-intel-nightly 4.6.0-rc6 447f243 from http://cgit.freedesktop.org/drm-intel commit 447f2438e5ee3e9bb7d5cbe88fc44cdff74cc165 Author: Daniel Vetter <daniel.vetter@ffwll.ch> Date: Sun May 8 18:21:27 2016 +0200 drm-intel-nightly: 2016y-05m-08d-16h-21m-00s UTC integration manifest drm: libdrm-2.4.67-11 ea78c17 from http://cgit.freedesktop.org/mesa/drm/ xorg-server-1.18.0-254 44e1c97 from git://git.freedesktop.org/git/xorg/xserver mesa: mesa-11.1.2 7bcd827 from http://cgit.freedesktop.org/mesa/mesa/ cairo: tag 1.15.2 db8a7f1 from http://cgit.freedesktop.org/cairo libva: libva-1.7.0-1 2339d10 from http://cgit.freedesktop.org/libva/ vaapi-intel-driver: 1.7.0.pre1 ab8e4e3 from http://cgit.freedesktop.org/vaapi/intel-driver (In reply to cprigent from comment #11) > Created attachment 123571 [details] > SKL-Y_4.6-rc6-447f243_kern.log > > I reproduced a similar problem on SKL-Y with fresh kernel. > > Steps: > ------- > 1. Connect HDMI and DP screens > 2. Boot > 3. xrandr --output eDP1 --off > 4. xrandr --outout eDP1 --auto > > Actual result: > ---------------- > 4. eDP stays black. Command does not respond > > Info > ----- > Could be the same as bug 89055 > > Not reproduced with 4.6 rc6 from https://cgit.freedesktop.org/drm-intel/ > commit 871c895c2d3192a93a701b31d2576184ab070c66 > Author: Rodrigo Vivi <rodrigo.vivi@intel.com> > Date: Thu May 5 16:49:48 2016 -0700 > drm-intel-nightly: 2016y-05m-05d-23h-49m-21s UTC integration manifest > > Setup > ------ > Hardware > Platform: > CPU : Intel(R) Core(TM) m5-6Y57 CPU @ 1.10GHz (family: 6, model: 78 > stepping: 3) > MCP : SKL-Y D1 2+2 > QDF : QJK9 > CPU : SKL D0 > Chipset PCH: Sunrise Point LP C0 > CRB : SKY LAKE Y LPDDR3 RVP3 CRB FAB2 > Software > Linux OS : Ubuntu 16.10 64 bits > BIOS : SKLSE2R1.R00.B104.B01.1511110114 > ME FW : 11.0.0.1191 > Ksc (EC FW): 1.20 > Kernel: drm-intel-nightly 4.6.0-rc6 447f243 from > http://cgit.freedesktop.org/drm-intel > commit 447f2438e5ee3e9bb7d5cbe88fc44cdff74cc165 > Author: Daniel Vetter <daniel.vetter@ffwll.ch> > Date: Sun May 8 18:21:27 2016 +0200 > drm-intel-nightly: 2016y-05m-08d-16h-21m-00s UTC integration manifest > drm: libdrm-2.4.67-11 ea78c17 from http://cgit.freedesktop.org/mesa/drm/ > xorg-server-1.18.0-254 44e1c97 from > git://git.freedesktop.org/git/xorg/xserver > mesa: mesa-11.1.2 7bcd827 from http://cgit.freedesktop.org/mesa/mesa/ > cairo: tag 1.15.2 db8a7f1 from http://cgit.freedesktop.org/cairo > libva: libva-1.7.0-1 2339d10 from http://cgit.freedesktop.org/libva/ > vaapi-intel-driver: 1.7.0.pre1 ab8e4e3 from > http://cgit.freedesktop.org/vaapi/intel-driver I reproduced this on my machine with same results and errors. Best regards, Edgar Cherkasov Please try current drm-intel-nightly. At least the [ 456.516554] WARNING: CPU: 2 PID: 695 at /home/kernel/COD/linux/drivers/gpu/drm/i915/intel_pm.c:3633 skl_update_other_pipe_wm+0x16c/0x180 [i915] [ 456.516559] WARN_ON(!wm_changed) part is fixed now (bug 89055). This bug in not reproduced with last nightly kernel 4.8.0-rc7 from https://cgit.freedesktop.org/drm-intel/: Setup: ====== Hardware Platform: CPU : Intel(R) Core(TM) i7-6700 CPU @ 3.40GHz (family: 6, model: 94 stepping: 3) Software Linux OS : Ubuntu 16.10 64 bits Kernel:drm-intel-nightly: 2016y-09m-19d-13h-29m-20s UTC integration manifest' author: Daniel Vetter <daniel.vetter@ffwll.ch> authored_date: Mon Sep 19 13:29:41 2016 +0000 commit: 24e02e82fe9870ae4951d6e5bb8886097f0aa70b drm: libdrm-2.4.70-14-g0659558 from http://cgit.freedesktop.org/mesa/drm/ xorg-server-1.18.99.2 from git://git.freedesktop.org/git/xorg/xserver mesa: mesa-12.0.0 78b061 from http://cgit.freedesktop.org/mesa/mesa/ cairo: tag 1.15.2 db8a7f1 from http://cgit.freedesktop.org/cairo libva: libva-1.7.0-50-g7aa2dd9 from http://cgit.freedesktop.org/libva/ vaapi-intel-driver: 1.7.0-136-g36fbd81 from http://cgit.freedesktop.org/vaapi/intel-driver |
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.