Distorted colours after suspend / resume cycle

[Jan-Marek Glogowski]

When waking up the system from a suspend-resume cycle, the display most times shows distorted colours.
As a workaround one can switch to the linux console (VT) and back to fix the display.
Alternatively triggering DPMS also fixes the problem (xset dpms force off ; sleep 0.1; xset dpms force on)

Hardware:
 01:00.0 VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI]
 Caicos [Radeon HD 6450/7450/8450 / R5 230 OEM]
 PCIID: 1002:6779

I've tested Ubuntu Xenial (16.04) with the latest updates:

mesa 11.2.0-1ubuntu2.2
xserver-xorg-video-radeon 1:7.7.0-1
linux-image-4.4.0-47-generic 4.4.0-47.68
xserver-xorg-core 2:1.18.3-1ubuntu2.3

I've tried older kernels, xorg, linux-firmware, as a colleague claimed it started for him after some of the latest Ubuntu updates, but couldn't find a working configuration.

I also checked Ubuntu Yakkety (16.10), where I couldn't reproduce with XFCE, but instantly with the Unity session, but probably XFCE was just lucky, as sometimes I also had series of working (~5) cycles. Current versions:

mesa 12.0.3-1ubuntu2
xserver-xorg-video-radeon 1:7.7.1-1
linux-image-4.8.0-27-generic 4.8.0-27.29
xserver-xorg-core 2:1.18.4-1ubuntu6.1

I'm also going to test Trusty (14.04) to check if it's really a regression. At least I have never seen this problem while running it.

ProblemType: Bug
DistroRelease: Ubuntu 16.10
Package: xorg 1:7.7+13ubuntu4
ProcVersionSignature: Ubuntu 4.8.0-27.29-generic 4.8.1
Uname: Linux 4.8.0-27-generic x86_64
.tmp.unity_support_test.0:

ApportVersion: 2.20.3-0ubuntu8
Architecture: amd64
CompizPlugins: No value set for `/apps/compiz-1/general/screen0/options/active_plugins'
CompositorRunning: compiz
CompositorUnredirectDriverBlacklist: '(nouveau|Intel).*Mesa 8.0'
CompositorUnredirectFSW: true
CurrentDesktop: Unity
Date: Tue Nov 22 12:21:24 2016
DistUpgraded: Fresh install
DistroCodename: yakkety
DistroVariant: ubuntu
ExtraDebuggingInterest: Yes, including running git bisection searches
GraphicsCard:
 Advanced Micro Devices, Inc. [AMD/ATI] Caicos [Radeon HD 6450/7450/8450 / R5 230 OEM] [1002:6779] (prog-if 00 [VGA controller])
   Subsystem: Bitland(ShenZhen) Information Technology Co., Ltd. Radeon HD 6450 [1642:3a75]
Lsusb:
 Bus 002 Device 002: ID 8087:0024 Intel Corp. Integrated Rate Matching Hub
 Bus 002 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub
 Bus 001 Device 002: ID 8087:0024 Intel Corp. Integrated Rate Matching Hub
 Bus 001 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub
MachineType: LENOVO 4524BL8
ProcKernelCmdLine: BOOT_IMAGE=/boot/vmlinuz-4.8.0-27-generic root=UUID=4dda3a4b-6e9c-49ba-a969-5d2146f02b60 ro splash quiet vt.handoff=7
SourcePackage: xorg
Symptom: display
UpgradeStatus: No upgrade log present (probably fresh install)
dmi.bios.date: 10/16/2012
dmi.bios.vendor: LENOVO
dmi.bios.version: 9HKT54AUS
dmi.board.vendor: LENOVO
dmi.chassis.type: 3
dmi.chassis.vendor: To Be Filled By O.E.M.
dmi.chassis.version: To Be Filled By O.E.M.
dmi.modalias: dmi:bvnLENOVO:bvr9HKT54AUS:bd10/16/2012:svnLENOVO:pn4524BL8:pvrThinkCentreM91p:rvnLENOVO:rn:rvr:cvnToBeFilledByO.E.M.:ct3:cvrToBeFilledByO.E.M.:
dmi.product.name: 4524BL8
dmi.product.version: ThinkCentre M91p
dmi.sys.vendor: LENOVO
version.compiz: compiz 1:0.9.13.0+16.10.20160818.2-0ubuntu2
version.ia32-libs: ia32-libs N/A
version.libdrm2: libdrm2 2.4.70-1
version.libgl1-mesa-dri: libgl1-mesa-dri 12.0.3-1ubuntu2
version.libgl1-mesa-dri-experimental: libgl1-mesa-dri-experimental N/A
version.libgl1-mesa-glx: libgl1-mesa-glx 12.0.3-1ubuntu2
version.xserver-xorg-core: xserver-xorg-core 2:1.18.4-1ubuntu6.1
version.xserver-xorg-input-evdev: xserver-xorg-input-evdev 1:2.10.2-1ubuntu1
version.xserver-xorg-video-ati: xserver-xorg-video-ati 1:7.7.1-1
version.xserver-xorg-video-intel: xserver-xorg-video-intel 2:2.99.917+git20160706-1ubuntu1
version.xserver-xorg-video-nouveau: xserver-xorg-video-nouveau 1:1.0.12-2
xserver.bootTime: Tue Nov 22 11:56:40 2016
xserver.configfile: default
xserver.errors:

xserver.logfile: /var/log/Xorg.0.log
xserver.version: 2:1.18.4-1ubuntu6.1
xserver.video_driver: radeon

[Jan-Marek Glogowski (jmglogow)]

This is just a display problem - a snapshot is correct, so it really looks like a radeon driver problem, instead of an xorg problem.

The wrong colors are always different, so every cycle produces a different result. For a photo as background it almost looks like modern art ;-)

[Jan-Marek Glogowski (jmglogow)]

I couldn't reproduce the problem with Trusty (14.04) using its original HWE stack (AKA Kernel 3.13).

[Jan-Marek Glogowski (jmglogow)]

Now I tried all the Trusty HWEs and it seems to have started with Wily.

xserver-xorg-video-radeon-lts-wily 1:7.5.0+git20150819-0ubuntu1~trusty1
linux-image-generic-lts-wily 4.2.0.42.34
xserver-xorg-core-lts-wily 2:1.17.2-1ubuntu9.1~trusty1
mesa-lts-wily 11.0.2-1ubuntu4~trusty1

I had once wrong colors with the Trusty initial stack (3.13), which I couldn't reproduce.

And I was just told that disabling DPMS / monitor power management in KDE also prevents the problem, which might correlate with the fact, that running a DPMS "force off" / "force on" cycle also restores the colors correctly.

[Jan-Marek Glogowski (jmglogow)]

While keeping the Xenial HWE stack on Ubuntu Trusty (14.04), I tried various installed HWE kernels.

It seems it really broke with the Wily kernel (4.2.0.42.34). The Vivid kernel (3.19.0.74.56) has survived multiple resume cycles, while Wily instantly created the problem.

[Jan-Marek Glogowski (jmglogow)]

So I downloaded some kernels from http://kernel.ubuntu.com/~kernel-ppa/mainline/?C=N;O=D

Working: linux-image-4.1.35-040135-generic_4.1.35-040135.201610241431_amd64.deb
Broken: linux-image-4.2.7-040207-generic_4.2.7-040207.201512091533_amd64.deb

[Jan-Marek Glogowski (jmglogow)]

Hmm - so I'm not sure what's going on here. I started bisecting v4.1 .. v4.2.8 and I couldn't find a working version in 6-7 bisects and when I checked rest of the commit set in the bisect, it couldn't find any commits to drivers/gpu?!

So I decided to test Ubuntu mainline kernels a little bit more and the current status is:
Broken: linux-image-4.1.3-040103-generic_4.1.3-040103.201507220129_amd64.deb
Working: linux-image-4.1.4-040104-generic_4.1.4-040104.201508031330_amd64.deb

So probably something was fixed and in a short time broken again during the v4.2 development cycle?!

And probably I should check with other connectors too, as I currently run the monitor via DVI => VGA connector.

[Jan-Marek Glogowski (jmglogow)]

$ git bisect log
# bad: [89e419960fb6a260f6a112821507d516117d5aa1] Linux 4.1.4
# good: [c8bde72f9af412de57f0ceae218d648640118b0b] Linux 4.1.3
git bisect start 'v4.1.4' 'v4.1.3'
# good: [e0cf83cc3de0341d8cabbb23097ad85f5ce97a11] drm/qxl: Do not leak memory if qxl_release_list_add fails
git bisect good e0cf83cc3de0341d8cabbb23097ad85f5ce97a11
# bad: [9d680e03989324f000596be4862728a7c30f22c1] selinux: don't waste ebitmap space when importing NetLabel categories
git bisect bad 9d680e03989324f000596be4862728a7c30f22c1
# bad: [510c99974fdbc18f143c41cbd461c522f5ad7164] tpm, tpm_crb: fix le64_to_cpu conversions in crb_acpi_add()
git bisect bad 510c99974fdbc18f143c41cbd461c522f5ad7164
# bad: [8b941a43ea7709111d3cbea2bdfcc678975255da] drm/radeon: only check the sink type on DP connectors
git bisect bad 8b941a43ea7709111d3cbea2bdfcc678975255da
# good: [0f2bb042f21bdb28f20efcf1ff1c507e2f8b3caa] drm/i915: Declare the swizzling unknown for L-shaped configurations
git bisect good 0f2bb042f21bdb28f20efcf1ff1c507e2f8b3caa
# good: [1f977d7e942519127aea0a08e9d08437b363cf19] drm/i915: Use two 32bit reads for select 64bit REG_READ ioctls
git bisect good 1f977d7e942519127aea0a08e9d08437b363cf19
# good: [7b49262b642511a16699cc63cf2a716739f0c43f] drm/radeon: SDMA fix hibernation (CI GPU family).
git bisect good 7b49262b642511a16699cc63cf2a716739f0c43f
# bad: [d1a4362d41e4feb52df6464f70fb64f21b894623] Revert "drm/radeon: dont switch vt on suspend"
git bisect bad d1a4362d41e4feb52df6464f70fb64f21b894623
# first bad commit: [d1a4362d41e4feb52df6464f70fb64f21b894623] Revert "drm/radeon: dont switch vt on suspend"

Just remember that good and bad is inverted, as I was looking for the commit which fixes my colour problem in v4.1.4. For me it fixes more then just the cursor.

[Jan-Marek Glogowski (jmglogow)]

So I applied this "big hammer" solution of the patch to the Trusty (14.04) Xenial HWE kernel (4.4.0-47.68~14.04.1) and this fixes the bug. I'll give it some more testing, while bisecting v4.0 to v4.1.

[Jan-Marek Glogowski (jmglogow)]

I can't bisect the problem, because for the two kernels I build, suspend was broken and the PC didn't wake up from 2nd suspend :-(
I also tried 4.9-rc6 (albeit on an Arch Linux), which shows the same bug.