Bug 201139 - amdgpu: [drm] enabling link 1 failed: 15 (vega)
Summary: amdgpu: [drm] enabling link 1 failed: 15 (vega)
Status: NEW
Alias: None
Product: Drivers
Classification: Unclassified
Component: Video(DRI - non Intel) (show other bugs)
Hardware: All Linux
: P1 normal
Assignee: drivers_video-dri
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2018-09-15 14:29 UTC by Aleksandr Mezin
Modified: 2020-12-12 10:39 UTC (History)
9 users (show)

See Also:
Kernel Version: 4.19
Subsystem:
Regression: No
Bisected commit-id:


Attachments
Kernel log (101.04 KB, text/plain)
2018-11-01 08:38 UTC, Aleksandr Mezin
Details
journalctl --user (Xorg log is here) (889.26 KB, text/plain)
2018-11-01 08:39 UTC, Aleksandr Mezin
Details
Kernel log (104.48 KB, text/plain)
2018-11-10 02:11 UTC, Aleksandr Mezin
Details
Further info, kernel log (11.31 KB, text/plain)
2020-04-10 20:09 UTC, Christian Thäter
Details
Another instance on 5.9.11-3-MANJARO #1 SMP PREEMPT (6.60 KB, text/plain)
2020-12-03 05:24 UTC, Michal Turecki
Details

Description Aleksandr Mezin 2018-09-15 14:29:33 UTC
My setup:
- RX Vega 64
- Arch Linux
- GNOME 3.28.3 on Xorg 1.20.1
- modesetting Xorg driver
- LG 27UD69P and Dell P2415Q

I have "screen blanking" enabled in Gnome. Sometimes (very rarely) one display (Dell, connected to 2nd DisplayPort) doesn't wake up correctly. It turns on, shows only black screen, then quickly shows "no signal" message. In kernel log I see:

[39215.008773] [drm] enabling link 1 failed: 15

If I manually turn the display off and on, it starts to work.

Can't tell if it is a regression (hardly it is) because on 4.18 and earlier screen blanking just makes the driver hang: https://bugzilla.kernel.org/show_bug.cgi?id=200531

I've never seen a similar issue on Windows on the same machine.
If I'm not mistaken, I've seen the same issue with Gnome on Wayland too (but I'm not sure).
Comment 1 Michel Dänzer 2018-09-17 10:25:42 UTC
Please attach the full dmesg output.
Comment 2 Aleksandr Mezin 2018-11-01 08:38:09 UTC
Created attachment 279279 [details]
Kernel log

Today I've seen this issue again on 4.19.

Monitors were turned off by dpms, only one of them correctly woke up. Another one turned on, but was only showing a black screen. What's more interesting is when I turned off that one with the black screen, another one started showing the black screen. And when I switched second monitor on again - the black screen moved back to it.

In kernel log again:
[46082.161234] [drm] enabling link 1 failed: 15
[46118.947574] [drm] enabling link 0 failed: 15
Comment 3 Aleksandr Mezin 2018-11-01 08:39:38 UTC
Created attachment 279281 [details]
journalctl --user (Xorg log is here)
Comment 4 Aleksandr Mezin 2018-11-01 08:56:42 UTC
Also tried loggig out & in, restarting GDM, changing display configurations in Gnome. Even in another VT and in GDM, one monitor was showing only a black screen. Only reboot made it work again.
Comment 5 Harry Wentland 2018-11-01 12:57:37 UTC
It looks like link training failed. Can you check on your monitor's OSD whether DisplayPort or Auto is selected as input? I've seen this behavior on some monitors when Auto input was selected. If so try selecting the DP input and see if that helps.
Comment 6 Aleksandr Mezin 2018-11-02 23:26:00 UTC
On both monitors there are no "Auto" input option. DisplayPort is selected on both of them.

Also, I've never seen similar problem on RX 580 with the same monitors, either on Linux or Windows.

On Windows I frequently see notifications about "link failure" or something like that, but on Windows the driver/card always recovers successfully (so no other problems except the annoying notification).

Also, more details, maybe it wasn't clear:
1. After dpms suspend, both monitors woke up, but one of them was showing black screen (but no "no signal" osd or anything like that). No mouse pointer too. But I was able to move the cursor away from the working monitor. xrandr and Gnome settings were showing both monitors too.
2. I turned that "non-working" monitor off. Now the previously working monitor only shows the black screen too.
3. I turned on the second monitor. Black screen moved to it, first monitor started working completely fine.
4. I thought that it's maybe a bug in compositor, so I tried switching VTs, logging out, restarting GDM. It didn't help.
5. Logged in back to Gnome. By changing settings/turning displays on and off I somehow managed to "move" the black screen to the first monitor, and the second one started working. But exactly one monitor (either first or second one) was always showing the black screen.
6. Rebooted, both monitors started working fine
Comment 7 Aleksandr Mezin 2018-11-10 02:11:23 UTC
Created attachment 279403 [details]
Kernel log

Encountered the same issue (monitor turned on, but shows black screen only) just now, on 4.19.1.
But now the error message is different:

[ 8154.090497] [drm:wait_for_reset_trigger_to_occur.isra.12 [amdgpu]] *ERROR* TG counter is not moving!
[ 8154.090533] [drm:wait_for_reset_trigger_to_occur.isra.12 [amdgpu]] *ERROR* GSL: Timeout on reset trigger!
Comment 8 Nick Price 2020-01-01 18:46:52 UTC
Seeing the same behavior on 5.3.0 fairly regularly - is there any other info I could provide to help?
Comment 9 John Andrew McInnes 2020-02-24 00:36:56 UTC
openSUSE Tumbleweed x86_64
Kernel: 5.5.4-1-default 
Radeon RX 480
LG Ultra HD monitor
KDE Plasma

[drm] enabling link 1 failed: 15

I am seeing this also only I have just a single monitor. It will say no signal, and it turns off. Moving the mouse or pressing the keyboard doesn't turn the monitor on. I have to turn it on using the power button.
Comment 10 Christian Thäter 2020-04-10 20:09:48 UTC
Created attachment 288335 [details]
Further info, kernel log

I had some hard crashes recently (screen frozen or blank), sometimes as much as once a day, sometimes the System runs for days to weeks.

Today one screen didn't come up after I turned it on and i found a smoking gun in the attached dmesg. Maybe thats helpful.

System is AMD Ryzen 1700, Debian Buster, self built Kernel.
Comment 11 Adarion from userland 2020-06-25 17:33:01 UTC
Potentially related
https://bugzilla.kernel.org/show_bug.cgi?id=208115
Comment 12 Michal Turecki 2020-12-03 05:24:45 UTC
Created attachment 293911 [details]
Another instance on 5.9.11-3-MANJARO #1 SMP PREEMPT

Same problem happens often on my desktop PC when resuming from power saving mode with monitor switched off. Like in other cases, switching monitor off/on helps but all open windows although seemingly restored to the original size before power saving, must have been temporarily resized to 640x480 during DP resolution negotiation which results in XFCE compiz (another, unrelated bug) shrinking them to 640x480 after moving for example.

Issue occurs very often, maybe about 1 of 5 times power saving mode is on. I am happy to do some more debugging if pointed to the right direction how to do it.

Some specs:
AMD RX5700 (Navi 10)
MSI MAG321CURV monitor.
Comment 13 James Ettle 2020-12-12 10:39:22 UTC
Also seen here under kernel 5.9.12 with a Ryzen 3400G and a Samsung LU32H850UM on DisplayPort:

[drm] enabling link 1 failed: 15

This is somewhat intermittent, and it comes back after power-cycling the monitor (which is configured with manual input selection).

Not seen with Dell or Lenovo monitors on same port.

I'll try a different DisplayPort cable just to check...

Note You need to log in before you can comment on or make changes to this bug.