Bug 99784

Created attachment 129539 [details]
Latest xOrg log file (Ubuntu 17.04 mainline kernel 4.10.rc7)

My wild guess is that problem has something to do with "Driver allocated offscreen pixmaps" warnings (as I can see 10-ish pixels from another monitor on each output device)

Version xorg is 7.7 but xserver-xorg-video-ati is 7.8.0.

Please attach the output of dmesg and

 xrandr --verbose

when the problem occurs.

Created attachment 129546 [details]
xrandr --verbose output

When xrandr command is executed HDMI output stops working

Created attachment 129547 [details]
dmesg output

Problem persists with DRM 2.49.0 / 4.10.1-041001-lowlatency, LLVM 3.9.1

Problem persists with kernel 4.11.rc4. I also made imgur post with relevant screenshots and explanations http://imgur.com/a/3kwiq

I wonder if at least some of the issues could be related to the internal panel's horizontal resolution of 1366... Does setting a lower resolution on the internal panel make any difference?

I changed resolution to 1280x768, and it disabled both internal monitor and VGA. After message box timed out and reverted resolution to 1366x768, VGA is now off and HDMI and internal are on.

Created attachment 130604 [details]
xrandr outputs for radeon and fglrx

Also there are a lot differences in xrandr outputs for fglrx and radeon, maybe timings are off too?

Problem persists with 4.11rc7 kernel. Should I try updating some other package?

Also, I noticed that problem starts before X is active (at grub/systemd/ubuntu splash logo). With fglrx only internal monitor would display image at that time - until lightdm and X start (when all three monitors would turn on). Using radeon drivers/drm however, external monitors would display mirrored systemd log/splash screen and internal monitor is off right after grub menu.

Problem persists with Ubuntu 17.10 / DRM 2.49.0 / 4.11.0-041100rc8-lowlatency, LLVM 4.0.0

Tested all resolution options for all three of the monitors, and has been able to output either on HDMI and VGA or HDMI and LVDS (internal). I have no idea what else could I test (or which code could I look at). Would setting drm.debug flag  to 0x4 in GRUB provide some more relevant data?  

Also, I have found this forum thread about the same issue, which dates from 2013: https://ubuntuforums.org/showthread.php?t=2194601

4.12.rc1 kernel (from http://kernel.ubuntu.com/~kernel-ppa/mainline/) crashes at boot during DRM loading phase, just after internal display turns off.
http://imgur.com/a/VBKv4

Created attachment 131429 [details]
Output of xrandr and dmesg

I have the same issue on Ubuntu 16.04 LTS with Mate desktop that runs on an iMac 11. The second monitor is connected through a Display Port to VGA adapter and the kernel selects it to be primary by default. 

After a fresh boot, since the second monitor is the primary one it contains the login screen, however the screen is black. So I have to unplug it to make the internal monitor (eDP) primary. If I plug again the external one, the internal monitor turns black and the external becomes active and starts to display it's content.

If I take a screenshot, I can see what is displayed on both monitors.

Created attachment 131434 [details]
Kernel oops at drm module (4.12.rc2 kernel)

Just tested 4.12.rc2 kernel, and had another kernel oops in drm related functions (at line 909 in attachment).

Created attachment 131532 [details]
"dmesg | grep drm" output (4.12.rc2, libdrm 2.4.8)

I updated libdrm and xorg from edgers ppa, and drm-tip kernel is now able to boot.
I attached latest "dmesg | grep drm" output. I went trough logs while changing drm.debug mask, but have failed to see any errors.  

Package versions are:
libdrm-radeon1/artful,now 2.4.80+git20170524.eeb51df3-0ubuntu0ricotz,
4.12.0-994-lowlatency kernel,
xorg 1:7.7+19 with xserver-xorg-video-radeon 1:7.9.0-0ubuntu1

Created attachment 136058 [details]
Workaround (Ubuntu 17.10 DRM 2.50.0 / 4.13.0-19-generic, LLVM 5.0.0)

After testing new version of kernel on Ubuntu 17.10 installation (wayland + systemd) I discovered by chance that after suspend and unsuspend (systemctl suspend), internal monitor starts displaying image and everything works (for the first time with radeon driver).

So as a workaround I used /etc/rc.local compatibility to trigger suspend after appropriate ring level is initialized during boot(sudo chmod +x /etc/rc.local is also required). 

I tested same method with 14.04 LTS with no success (so I don't know exactly what upgrade made it work).

-- GitLab Migration Automatic Message --

This bug has been migrated to freedesktop.org's GitLab instance and has been closed from further activity.

You can subscribe and participate further through the new bug through this link to our GitLab instance: https://gitlab.freedesktop.org/drm/amd/issues/772.