Bug 106427 - RX560, 5k@60Hz, Xorg: Black screen, “soft” kernel lockup
Summary: RX560, 5k@60Hz, Xorg: Black screen, “soft” kernel lockup
Status: RESOLVED NOTABUG
Alias: None
Product: DRI
Classification: Unclassified
Component: DRM/AMDgpu (show other bugs)
Version: unspecified
Hardware: x86-64 (AMD64) Linux (All)
: medium normal
Assignee: Default DRI bug account
QA Contact:
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2018-05-07 11:13 UTC by txtoxtox285
Modified: 2018-10-27 17:07 UTC (History)
1 user (show)

See Also:
i915 platform:
i915 features:


Attachments
dmesg-4.16.7 5K@60 (195.52 KB, text/plain)
2018-05-07 11:14 UTC, txtoxtox285
no flags Details
Xorg.log 5k@60 (38.21 KB, text/plain)
2018-05-07 11:14 UTC, txtoxtox285
no flags Details
dmesg-4.16.7 4K@60 (174.20 KB, text/plain)
2018-05-12 10:19 UTC, txtoxtox285
no flags Details
Xorg.log 4k@60 (30.21 KB, text/plain)
2018-05-12 10:20 UTC, txtoxtox285
no flags Details
dmesg-4.19.0 with drm.debug=0x1ff (232.21 KB, text/plain)
2018-10-27 17:06 UTC, txtoxtox285
no flags Details

Description txtoxtox285 2018-05-07 11:13:35 UTC
Kernel: 4.16.7

My monitor’s native mode is (some variant of) 5120x2880 at 60Hz. When amdgpu is loaded during startup, it picks up this mode from the monitor and switches to it just fine.

However, when I start a minimal X (X.Org X server 1.19.5, amdgpu 1.3.0) without /etc/X11/xorg.conf and just an xterm and no window manager, I can see the for a fraction of a second the xterm window in the upper left corner of the screen before the monitor turns black. It’s not turned off, the monitor’s power LED shows that it’s on.

When I try to switch to another Linux console with Alt+Tab+F6, nothing happens; the console is not switched (I can verify this because typing “echo -en '\a'” doesn’t beep).

After ca. 1:30 min the console finally *is* switched but remains black. The kernel spews out messages about timeouts, and I have to reboot the machine (from ssh or Ctrl+Alt+Del).

I can use X with 5K if I restrict the refresh rate to 30 Hz (from “cvt 5120 2880 30”).

This bug is present since kernel 4.15.10 or thereabouts, which was the first version where amdgpu would work with dc = 1.
Comment 1 txtoxtox285 2018-05-07 11:14:21 UTC
Created attachment 139402 [details]
dmesg-4.16.7 5K@60
Comment 2 txtoxtox285 2018-05-07 11:14:48 UTC
Created attachment 139403 [details]
Xorg.log 5k@60
Comment 3 txtoxtox285 2018-05-12 10:18:35 UTC
I would like to mention that “cvt 4096 2304 -r” is a mode that works under X so it may not be the 60 Hz as such but rather the increased bandwidth, which needs DisplayPort 1.3 instead of just 1.2.

(However, when X is terminated the Linux console goes black because it seems that the driver cannot switch back to 5k@60Hz.)
Comment 4 txtoxtox285 2018-05-12 10:19:52 UTC
Created attachment 139514 [details]
dmesg-4.16.7 4K@60
Comment 5 txtoxtox285 2018-05-12 10:20:15 UTC
Created attachment 139515 [details]
Xorg.log 4k@60
Comment 6 txtoxtox285 2018-06-09 11:21:10 UTC
Still present with kernel 4.17.0, xorg amdgpu 18.0.1.
Comment 7 txtoxtox285 2018-08-19 11:02:45 UTC
Kernel 4.18.1, xorg amdgpu 18.0.1: All quiet on the Western Front.
Comment 8 txtoxtox285 2018-10-27 17:03:24 UTC
I have always wondered why 5k works in the frame buffer but not Xorg. 4.19 bring some progress in this regard and made things more consistent: Now the console stays blank too.
Comment 9 txtoxtox285 2018-10-27 17:06:20 UTC
Created attachment 142235 [details]
dmesg-4.19.0 with drm.debug=0x1ff


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.