Bug 1236337

Summary: regression: vertical pink line on HDMI port with KMS / wrong resolution
Product: [Fedora] Fedora Reporter: ivan <ivan.mitev>
Component: xorg-x11-drv-atiAssignee: Hans de Goede <hdegoede>
Status: CLOSED ERRATA QA Contact: Fedora Extras Quality Assurance <extras-qa>
Severity: medium Docs Contact:
Priority: unspecified    
Version: 22CC: hdegoede, krofnica996, xgl-maint
Target Milestone: ---   
Target Release: ---   
Hardware: x86_64   
OS: Linux   
Whiteboard:
Fixed In Version: Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2015-08-19 13:54:52 UTC Type: Bug
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Attachments:
Description Flags
projector EDID
none
[PATCH] drm/radeon: rework audio detect (v4) none

Description ivan 2015-06-27 20:26:04 UTC
Description of problem:

After updating/booting to kernel-4.0.5-300.fc22.x86_64 I now see a vertical pink line on the left side on a projector plugged in the HDMI port of an AMD APU based HTPC. It happens as soon as Xorg starts (but note that there's no pink line in the console after modesetting prior to Xorg start).

Reverting to kernel-4.0.4-303.fc22.x86_64 "fixes" the problem. Setting radeon.nomodeset=1 too, but that's not an acceptable workaround.

Version-Release number of selected component (if applicable):
xorg-x11-drv-ati-7.5.0-3.fc22.x86_64

Additional info:

cpu:
Vendor ID:             AuthenticAMD
CPU family:            21
Model:                 16
Model name:            AMD A8-5500 APU with Radeon(tm) HD Graphics
Stepping:              1

the projector (epson EH-TW4500) shows that the resolution is 1922x1080, so it's off by 2 pixels. However there's nothing in dmesg, Xorg.0.log, decoded EDID or /sys/class/drm/* that show anything different than 1920x1080.

There are similar bug reports (eg. bug 597366) or https://www.mail-archive.com/xorg-driver-ati@lists.x.org/msg12869.html, but things used to work perfectly before the kernel update so they don't seem relevant.

Comment 1 ivan 2015-06-29 18:52:21 UTC
BTW, same problem with today's kernel update - kernel-4.0.6-300.fc22.x86_64.

Comment 2 ivan 2015-07-21 08:49:30 UTC
same problem with 4.0.8-300.fc22.x86_64 :(

I'd be happy to help with debugging - only that I have no clue where to begin from.

Comment 3 Hans de Goede 2015-07-21 19:40:21 UTC
Hi,

The problem seems to be that the radeon kernel driver thinks that your beamer is hdmi-audio capable, while it is not, and then the beamer sees the audio data in the hdmi frames as video data.

Can you try adding: "radeon.audio=0" to your kernel cmdline and see if that helps? That should work around this, once that is confirmed we can see if we can make the auto-detect code work with your beamer.

Regards,

Hans

Comment 4 ivan 2015-07-21 21:02:45 UTC
Hi,

You were right, radeon.audio=0 fixed it. Cool..

Let me know when you want me to do additional tests/debug with the auto-detection code. (Or feel free to close the bug).

Thanks !
ivan

Comment 5 Hans de Goede 2015-07-23 08:51:42 UTC
Hi,

(In reply to ivan from comment #4)
> Hi,
> 
> You were right, radeon.audio=0 fixed it. Cool..

Good.

> Let me know when you want me to do additional tests/debug with the
> auto-detection code. (Or feel free to close the bug).

Can you please do:

cd /sys/class/drm
ls

You should then see something like this:

card0       card0-DP-2      card0-HDMI-A-2  controlD64  version
card0-DP-1  card0-HDMI-A-1  card0-VGA-1     renderD128

Then for each of the cord0-$connector dirs do:

cat card0-$connector/modes

Until you've found the connector which corresponds to your projector.

Then do:

cat card0-$connector/edid > ~/projector-edid.bin

And then attach projector-edid.bin here. Then I'll see if I can figure out what is going wrong in the auto-detect code for the audio setting.

Regards,

Hans

Comment 6 ivan 2015-07-23 12:01:53 UTC
Hi,

I forgot to mention that the projector is not always plugged: the PC is more a server that is used as HTPC from times to times than a real HTPC; it has a DVI monitor always plugged.
The PC's HDMI output goes into an AV amplifier and the projector is then plugged in the amplifier. The amplifier is not a "smart" one, ie. it only acts as an HDMI switch with no decoding/encoding capabilities, so it shouldn't mangle the EDID (I don't have longer cables so I couldn't try to physically plug the beamer directly into the PC's HDMI port to verify that this assumption is right though).

The projector was on card0-HDMI-A-1. By the way the output of card0-HDMI-A-1/modes didn't list 1920x1080 while the projector supports it (see output below). edid-decode properly lists it though.

Thanks for the help !
Ivan


$ for i in card0*; do echo $i; cat $i/modes; echo; done                                                                                           
card0-DVI-D-1
1280x1024
1280x1024
1280x960
1152x864
1024x768
1024x768
1024x768
832x624
800x600
800x600
800x600
800x600
640x480
640x480
640x480
640x480
720x400

card0-HDMI-A-1
1280x720
1280x720
1280x720
720x576
720x576i
720x480
720x480
720x480i
720x480i
640x480
640x480

card0-VGA-1

Comment 7 ivan 2015-07-23 12:02:37 UTC
Created attachment 1055296 [details]
projector EDID

Comment 8 Mislav 2015-07-24 09:32:10 UTC
I am also affected by this bug, but the pink line is on monitor.

Comment 9 Hans de Goede 2015-07-24 13:29:26 UTC
Hi,

I've just started a scratch build of the fedora 4.1.3 kernel with a patch from upstream which should fix this:
http://koji.fedoraproject.org/koji/taskinfo?taskID=10462458

Please download the following rpms:
kernel-4.1.3-...rpm
kernel-core-4.1.3-...rpm
kernel-modules-4.1.3-...rpm

For your arch from there  (note this is still building atm).

Save them all to the same directory (which should not contain other rpms) and do:

sudo rpm -ivh kernel-*.rpm

From within that directory.

Then boot into the new kernel (without any special kernel commandline options) and check if this fixes things.

Thanks & Regards,

Hans

Comment 10 Hans de Goede 2015-07-24 13:31:28 UTC
Created attachment 1055738 [details]
[PATCH] drm/radeon: rework audio detect (v4)

Patch from: https://bugzilla.kernel.org/show_bug.cgi?id=93701 backported to 4.1 kernel.

Comment 11 ivan 2015-07-24 16:14:55 UTC
Hi,

The test kernel fixes the problem - no more pink/purple vertical lines.

Thanks!
Ivan

Comment 12 Hans de Goede 2015-07-24 16:18:56 UTC
Hi,

(In reply to ivan from comment #11)
> Hi,
> 
> The test kernel fixes the problem - no more pink/purple vertical lines.
> 
> Thanks!
> Ivan

That is good to hear, I'm going to wait for upstream to declare the fix stable before adding it to any officia Fedora kernels though, so for now please keep using the radeon.audio=0 workaround.

Regards,

Hans

Comment 13 Hans de Goede 2015-08-19 13:54:52 UTC
The fix for this is in 4.1.6, which is in updates testing now, closing.