| Bug ID | 99275 |
|---|---|
| Summary | Kernel 4.9: amdgpu regression; gui flickers; amd radeon rx 460 |
| Product | DRI |
| Version | XOrg git |
| Hardware | x86-64 (AMD64) |
| OS | Linux (All) |
| Status | NEW |
| Severity | normal |
| Priority | medium |
| Component | DRM/AMDgpu |
| Assignee | dri-devel@lists.freedesktop.org |
| Reporter | tuxuser@quantentunnel.de |
| CC | tiwai@suse.de |
Created attachment 128754 [details] hwinfo System: openSuse Leap 42.2, kernel from download.opensuse.org/repositories/Kernel:/stable/standard/ 4.9.0-4.g1af4b0f Mesa from download.opensuse.org/repositories/X11:/XOrg/openSUSE_Leap_42.2/ graphics adapter: amd radeon rx 460 cpu: i5-6402p desktop: kde5 (from 42.2-oss/update) Problem: Address-drop-down in firefox flickers. Steps to reproduce: start firefox enter address (example: www.heise.de) enter an other address address drop down starts to flicker. Perhaps problem shows up only when virtual desktop has been changed before firefox was started. (I use cube animation for virtual desktop change.) Problem also occurs with tumbleweed live system. There is no problem with kernel 4.8.x (4.8.14) I've also reported this bug at https://bug= zilla.suse.com/show_bug.cgi?id=3D1017938
Please attach your xorg log and dmesg output. Is it just the = menu that flickers or the whole screen? Did you also change any other components (me= sa, ddx, etc.)?
Created attachment 12875=
8 [details]
rpm -qa --queryformat "%{Name} - %{Version} - %{Vendor}\n" | grep
build.opensuse.org/X11
I've attached the logs and the info about changed components. They are all from download.opensuse.org/repositories/X11:/XOrg/openSUSE_Leap_42.2/ The flicker problem is not about the menu but the drop down list that opens when you enter an address to the firefox address field. A second effect: When I re-size a window using kernel 4.8.14 the background (or underlying windows) shine through. When I do so using kernel 4.9 it also starts flickering and shows artifacts from the background.
Just something I do not understand: While having a look at the Xorg.log I couldn't find anything about loading xf86-video-amdgpu. I just found lines about=20 /usr/lib64/xorg/modules/drivers/ati_drv.so and /usr/lib64/xorg/modules/drivers/radeon_drv.so Do I need a special config for loading the amdgpu driver? There is no video-driver specific config in /etc/X11/xorg.conf.d/
Any chance you can use git bisect to find the change between 4= .8 and 4.9 which introduced the problem? (In reply to Reimar Imhof from comme= nt #6) > Do I need a special config for loading the amdgp= u driver? >=20 > There is no video-driver specific config in /etc/X11/xorg.conf.d/ There should be a /usr/share/X11/xorg.conf.d/10-amdgpu.conf file which caus= es the amdgpu driver to be used automatically for GPUs controlled by the amdgpu kernel driver. If that file is missing, please report the problem to openSu= se.
Created attachment 128779 [details]
Xorg.0.log with activated Xorg amdgup driver
I found an Xorg amdgpu.conf file at the arch linux wiki:
Section "Device"
Identifier "AMD"
Driver "amdgpu"
EndSection
With this config the amdgpu Xorg driver gets loaded. But the kernel bug sti=
ll
exists.
(In reply to Michel D=C3=A4nzer from comment #7) > Any chance you can use git bisect to find the ch= ange between 4.8 and 4.9 > which introduced the problem? Sorry, I can't. > There should be a /usr/share/X11/xorg.conf.d/10-= amdgpu.conf file which > causes the amdgpu driver to be used automatically for GPUs controlled = by the > amdgpu kernel driver. If that file is missing, please report the probl= em to > openSuse. I filed a bug report at https://bugzilla.opensuse.org/show_bug.cgi?id=3D1018406
I've just tried kernel 4.9.4 from download.opensuse.org/repositories/Kernel:/stable/standard/ Same bug. Do you have any news?
tried a bisect: git bisect start # good: [c8d2bc9bc39ebea8437fd974fdbc21847bb897a3] Linux 4.8 git bisect good c8d2bc9bc39ebea8437fd974fdbc21847bb897a3 # bad: [69973b830859bc6529a7a0468ba0d80ee5117826] Linux 4.9 git bisect bad 69973b830859bc6529a7a0468ba0d80ee5117826 # bad: [a5af7e1fc69a46f29b977fd4b570e0ac414c2338] rxrpc: Fix loss of PING RESPONSE ACK production due to PING ACKs git bisect bad a5af7e1fc69a46f29b977fd4b570e0ac414c2338 # bad: [d268dbe76a53d72cc41316eb59e7968db60e77ad] Merge tag 'pinctrl-v4.9-1= ' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux-pinctrl git bisect bad d268dbe76a53d72cc41316eb59e7968db60e77ad # bad: [02bafd96f3a5d8e610b19033ffec55b92459aaae] Merge tag 'docs-4.9' of git://git.lwn.net/linux git bisect bad 02bafd96f3a5d8e610b19033ffec55b92459aaae # bad: [9929780e86854833e649b39b290b5fe921eb1701] Merge tag 'driver-core-4.9-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core git bisect bad 9929780e86854833e649b39b290b5fe921eb1701 # good: [12b7bcb43e6ea834ab2f5dc52d971e379a0ca109] Merge branch 'perf-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/= tip git bisect good 12b7bcb43e6ea834ab2f5dc52d971e379a0ca109 # bad: [5e1b834b27fb2c27cde33a0752425f11d10c0b2d] Merge branch 'timers-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip git bisect bad 5e1b834b27fb2c27cde33a0752425f11d10c0b2d # bad: [110a9e42b68719f584879c5c5c727bbae90d15f9] Merge branch 'x86-apic-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/t= ip git bisect bad 110a9e42b68719f584879c5c5c727bbae90d15f9 # bad: [a399d233078edbba7cf7902a6d080100cdf75636] sched/core: Fix incorrect utilization accounting when switching to fair class git bisect bad a399d233078edbba7cf7902a6d080100cdf75636 # good: [1a3d027c5a6847e5d349c8527f99aada47e5467a] sched/debug: Rename and = move enqueue_sleeper() git bisect good 1a3d027c5a6847e5d349c8527f99aada47e5467a # good: [53061afee43bc5041b67a45b6d793e7afdcf9ca7] Merge branch 'akpm' (pat= ches from Andrew) git bisect good 53061afee43bc5041b67a45b6d793e7afdcf9ca7 # good: [a18a579e5f84daa74f64b1f1b652b4a6a8d6f8b4] sched/debug: Hide printk= () by default git bisect good a18a579e5f84daa74f64b1f1b652b4a6a8d6f8b4 # good: [eaf9ef52241b545fe63621266bfc6fd8b06559ff] sched/wait: Avoid abort_exclusive_wait() in __wait_on_bit_lock() git bisect good eaf9ef52241b545fe63621266bfc6fd8b06559ff # good: [24fc7edb92eea05946119cc0258c891c26b3b469] sched/core: Introduce 'struct sched_domain_shared' git bisect good 24fc7edb92eea05946119cc0258c891c26b3b469 # good: [10e2f1acd0106c05229f94c70a344ce3a2c8008b] sched/core: Rewrite and improve select_idle_siblings() git bisect good 10e2f1acd0106c05229f94c70a344ce3a2c8008b # bad: [1b568f0aabf280555125bc7cefc08321ff0ebaba] sched/core: Optimize SCHED_SMT git bisect bad 1b568f0aabf280555125bc7cefc08321ff0ebaba # first bad commit: [1b568f0aabf280555125bc7cefc08321ff0ebaba] sched/core: Optimize SCHED_SMT > git bisect bad 1b568f0aabf280555125bc7cefc08321ff0ebaba is the first bad commit commit 1b568f0aabf280555125bc7cefc08321ff0ebaba Author: Peter Zijlstra <pete= rz@infradead.org> Date: Mon May 9 10:38:41 2016 +0200 sched/core: Optimize SCHED_SMT Avoid pointless SCHED_SMT code when running on !SMT hardware. Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Mike Galbraith <efault@= gmx.de> Cc: Peter Zijlstra <pete= rz@infradead.org> Cc: Thomas Gleixner <tglx&= #64;linutronix.de> Cc: linux-kernel= 4;vger.kernel.org Signed-off-by: Ingo Molnar <= mingo@kernel.org> :040000 040000 2ce55347eeb85f4e6044fac184e944f5235de1d3 454406febf88e80e4c5a8e37a403ad6c79e097d6 M kernel
tried an other bisect git bisect start # bad: [9929780e86854833e649b39b290b5fe921eb1701] Merge tag 'driver-core-4.9-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core git bisect bad 9929780e86854833e649b39b290b5fe921eb1701 # good: [12b7bcb43e6ea834ab2f5dc52d971e379a0ca109] Merge branch 'perf-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/= tip git bisect good 12b7bcb43e6ea834ab2f5dc52d971e379a0ca109 # bad: [5e1b834b27fb2c27cde33a0752425f11d10c0b2d] Merge branch 'timers-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip git bisect bad 5e1b834b27fb2c27cde33a0752425f11d10c0b2d # bad: [110a9e42b68719f584879c5c5c727bbae90d15f9] Merge branch 'x86-apic-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/t= ip git bisect bad 110a9e42b68719f584879c5c5c727bbae90d15f9 # bad: [a399d233078edbba7cf7902a6d080100cdf75636] sched/core: Fix incorrect utilization accounting when switching to fair class git bisect bad a399d233078edbba7cf7902a6d080100cdf75636 # bad: [1a3d027c5a6847e5d349c8527f99aada47e5467a] sched/debug: Rename and m= ove enqueue_sleeper() git bisect bad 1a3d027c5a6847e5d349c8527f99aada47e5467a # good: [0e6d2a67a41321b3ef650b780a279a37855de08e] sched/core: Remove unnecessary NULL-pointer check git bisect good 0e6d2a67a41321b3ef650b780a279a37855de08e # good: [62cc20bcf25617dd5ad23356ea46830da3ef7356] Merge branch 'sched/urge= nt' into sched/core, to pick up fixes git bisect good 62cc20bcf25617dd5ad23356ea46830da3ef7356 # good: [d8206bb3ffe0eaee03abfad46fd44d8b17142e88] sched/deadline: Split cpudl_set() into cpudl_set() and cpudl_clear() git bisect good d8206bb3ffe0eaee03abfad46fd44d8b17142e88 # good: [efca03ecbe29a46c2c5ae539563b6326af9dcba7] schedcore: Remove duplic= ated init_task's preempt_notifiers init git bisect good efca03ecbe29a46c2c5ae539563b6326af9dcba7 # good: [61c7aca695b6fabe85d0fc424fe8ae2f66f267dd] sched/deadline: Fix the intention to re-evalute tick dependency for offline CPU git bisect good 61c7aca695b6fabe85d0fc424fe8ae2f66f267dd # first bad commit: [1a3d027c5a6847e5d349c8527f99aada47e5467a] sched/debug: Rename and move enqueue_sleeper() > git bisect good 1a3d027c5a6847e5d349c8527f99aada47e5467a is the first bad commit commit 1a3d027c5a6847e5d349c8527f99aada47e5467a Author: Josh Poimboeuf <jpoim= boe@redhat.com> Date: Fri Jun 17 12:43:23 2016 -0500 sched/debug: Rename and move enqueue_sleeper() enqueue_sleeper() doesn't actually enqueue, it just handles some statistics and tracepoints. Rename it to update_stats_enqueue_sleeper() and call it from update_stats_enqueue(). Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Matt Fleming <ma= tt@codeblueprint.co.uk> Cc: Mel Gorman <m= gorman@techsingularity.net> Cc: Peter Zijlstra <pete= rz@infradead.org> Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com> Cc: Thomas Gleixner <tglx&= #64;linutronix.de> Link: http://lkml.kernel.org/r/fb20b715= 9dc4d028c406c0e8d5f8c439b741615b.1466184592.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <= mingo@kernel.org> :040000 040000 18aa42131bb6e286cc94eb481b7edf59c0b8d3b5 01d1206384de4d8a232427b39e9940806d9dacf3 M kernel
after I've had seen, 4.8 hat the same problem I tried to find = the first good 4.8.x: > git bisect new 107d026ae1c80ac0881f791a58cd115321d111ca is the first new commit commit 107d026ae1c80ac0881f791a58cd115321d111ca Author: Will Deacon <will.dea= con@arm.com> Date: Fri Aug 26 11:36:39 2016 +0100 arm64: debug: avoid resetting stepping state machine when TIF_SINGLESTEP commit 3a402a709500c5a3faca2111668c33d96555e35a upstream. When TIF_SINGLESTEP is set for a task, the single-step state machine is enabled and we must take care not to reset it to the active-not-pending state if it is already in the active-pending state. Unfortunately, that's exactly what user_enable_single_step does, by unconditionally setting the SS bit in the SPSR for the current task. This causes failures in the GDB testsuite, where GDB ends up missing expected step traps if the instruction being stepped generates another trap, e.g. PTRACE_EVENT_FORK from an SVC instruction. This patch fixes the problem by preserving the current state of the stepping state machine when TIF_SINGLESTEP is set on the current thread. Cc: <stable@vger.k= ernel.org> Reported-by: Yao Qi <yao.qi= 4;arm.com> Signed-off-by: Will Deacon <will.deacon@arm.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> :040000 040000 98b6a8a0d27ab29b17b6b617e52beb33d56dc604 46030e41ff5783555a6d6e895799aabbe3a0c46e M arch
I've also tried 4.9.6 and 4.10-rc5. Again error still existing.
There are two basic explanations for not getting consistent bi= section results: * In each case, at least one bad commit was accidentally marked as good. Te= st longer / more times before declaring a commit as good to avoid this. * The problem (or at least the trigger) isn't in the kernel but somewhere e= lse.
I had an other look at 4.8.
4.8 is ok.
Next try was 4.9-rc1.
That one is buggy.
Trying to bisect ended up in lots of unbootable kernels.
git bisect start
# good: [c8d2bc9bc39ebea8437fd974fdbc21847bb897a3] Linux 4.8
git bisect good c8d2bc9bc39ebea8437fd974fdbc21847bb897a3
# bad: [1001354ca34179f3db924eb66672442a173147dc] Linux 4.9-rc1
git bisect bad 1001354ca34179f3db924eb66672442a173147dc
# skip: [41844e36206be90cd4d962ea49b0abc3612a99d0] Merge tag 'staging-4.9-r=
c1'
of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging
git bisect skip 41844e36206be90cd4d962ea49b0abc3612a99d0
# skip: [7906a4da0ef845d01e76f2187c23cc71ae00fa1d] net/faraday: Make EDO{R,=
T}R
bits configurable
git bisect skip 7906a4da0ef845d01e76f2187c23cc71ae00fa1d
# skip: [86ec64bc382ad1e0e6bc66f5f705a53e299a3445] ACPICA: Tables: Fix a
regression in acpi_tb_find_table()
git bisect skip 86ec64bc382ad1e0e6bc66f5f705a53e299a3445
So that took quite some time but didn't help at all.
I can't see that this will come to an end. Hopefully it will not break my
machine.
So I'm stopping this.
Does using the new ucode here help? htt= ps://people.freedesktop.org/~agd5f/radeon_ucode/polaris/
That firmware kills kde / plasma5 and freezes the machine!
What I've done:
Copied the firmware files to /lib/firmware/amdgpu/
ran dracut to build new initrd
reboot
graphical login comes up
after logging in to kde plasma 5 the machine freezes
After a hard power of I logged into icewm.
That desktop came up. I opened an console. Ok.=20
Started dolphin. Machine has frozen immediately. Even the clock in the desk=
top
toolbar stopped working.
Before starting dolphin I had a look at dmesg. I couldn't find any new error
messages. Unfortunately I do not have these messages around any more.
I also had a look at the /var/log/Xorg.<i>.log
There was no log written from the kde session freezing the machine.
Could you please test that firmware with a kde plasma 5 desktop? Please!
(In reply to Reimar Imhof from comment #18) > That firmware kills kde / plasma5 and freezes th= e machine! >=20 > What I've done: > Copied the firmware files to /lib/firmware/amdgpu/ > ran dracut to build new initrd > reboot >=20 > graphical login comes up > after logging in to kde plasma 5 the machine freezes >=20 > After a hard power of I logged into icewm. > That desktop came up. I opened an console. Ok.=20 > Started dolphin. Machine has frozen immediately. Even the clock in the > desktop toolbar stopped working. >=20 > Before starting dolphin I had a look at dmesg. I couldn't find any new= error > messages. Unfortunately I do not have these messages around any more. > I also had a look at the /var/log/Xorg.<i>.log > There was no log written from the kde session freezing the machine. >=20 > Could you please test that firmware with a kde plasma 5 desktop? Pleas= e! Can you just try updating the smc firmware?