Bug 213633 - Displays go blank and system semi-crashes (several hardware-related tasks block) on activation of output to certain HDMI monitors on intel UHD 605 graphics card; depending on kernel configuratuion.
Summary: Displays go blank and system semi-crashes (several hardware-related tasks blo...
Status: RESOLVED OBSOLETE
Alias: None
Product: Drivers
Classification: Unclassified
Component: Video(Other) (show other bugs)
Hardware: Intel Linux
: P1 normal
Assignee: drivers_video-other
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2021-07-02 12:52 UTC by Drei Eck
Modified: 2022-05-10 12:03 UTC (History)
1 user (show)

See Also:
Kernel Version: 5.13, 5.11.11, 5.10.27
Subsystem:
Regression: No
Bisected commit-id:


Attachments
Archive containing the additional information files mentioned in the report. (244.35 KB, application/x-xz)
2021-07-02 12:52 UTC, Drei Eck
Details
The kernel configuration used based on my configuration (258.61 KB, text/plain)
2021-07-04 09:41 UTC, Drei Eck
Details
The output of dmesg just before activating the plugged-in monitor via xrandr (85.46 KB, text/plain)
2021-07-04 09:42 UTC, Drei Eck
Details
The output of dmesg after activating the plugged-in monitor via xrandr and then after issuing chvt 1 and abeep (alsa beep) and after waiting some time (to capture the "blocked tasks" messages) (105.09 KB, text/plain)
2021-07-04 09:42 UTC, Drei Eck
Details
From dmesg_after-my.txt, just extracted the part with the "blocked tasks" messages (18.99 KB, text/plain)
2021-07-04 09:43 UTC, Drei Eck
Details
The output of hwinfo (1.04 MB, text/plain)
2021-07-04 09:43 UTC, Drei Eck
Details
The output of dmidecode (18.16 KB, text/plain)
2021-07-04 09:44 UTC, Drei Eck
Details
A brief description on how the kernel was configured (69 bytes, text/plain)
2021-07-04 09:44 UTC, Drei Eck
Details
The output of uname -a (91 bytes, text/plain)
2021-07-04 09:44 UTC, Drei Eck
Details
A description of symptoms after triggering the problem (981 bytes, text/plain)
2021-07-04 09:45 UTC, Drei Eck
Details
The make defconfig-based configuration (125.68 KB, text/plain)
2021-07-04 09:45 UTC, Drei Eck
Details
The output of dmesg just before activating the plugged-in monitor via xrandr (66.13 KB, text/plain)
2021-07-04 09:46 UTC, Drei Eck
Details
The output of dmesg after activating the plugged-in monitor via xrandr (66.77 KB, text/plain)
2021-07-04 09:46 UTC, Drei Eck
Details
The output of hwinfo (931.96 KB, text/plain)
2021-07-04 09:46 UTC, Drei Eck
Details
The output of dmidecode (18.16 KB, text/plain)
2021-07-04 09:47 UTC, Drei Eck
Details
A description of how the kernel was configured (686 bytes, text/plain)
2021-07-04 09:47 UTC, Drei Eck
Details
The output of uname -a (83 bytes, text/plain)
2021-07-04 09:48 UTC, Drei Eck
Details
A description of symptoms after activating the external monitor via xrandr (263 bytes, text/plain)
2021-07-04 09:48 UTC, Drei Eck
Details
The URL I used to download the kernel from (62 bytes, text/plain)
2021-07-04 09:48 UTC, Drei Eck
Details
For reference: The kernel configuration of the also problem-triggering precompiled Artix distribution kernel version 5.11.8-artix1-1 (229.58 KB, text/plain)
2021-07-04 09:50 UTC, Drei Eck
Details
For reference: The output of dmesg for the distribution kernel 5.11.8-artix1-1 before activating the external HDMI monitor with xrandr. (73.02 KB, text/plain)
2021-07-04 09:53 UTC, Drei Eck
Details
attachment-29200-0.sig (833 bytes, application/pgp-signature)
2021-11-10 19:30 UTC, Drei Eck
Details

Description Drei Eck 2021-07-02 12:52:34 UTC
Created attachment 297689 [details]
Archive containing the additional information files mentioned in the report.

# Displays go blank and system semi-crashes (sveral hardware-related tasks block) on activation of output to certain HDMI monitors on intel UHD 605 graphics card; depending on kernel configuratuion.

## On which kernels was this tested?

* This was tested with _vanilla_ kernels
  - 5.4.109,
  - 5.10.27,
  - 5.11.11,
  - 5.13.
  - Additionally with distribution kernels ([Artix](mirror1.artixlinux.org/repos/system/os/x86_64/), version 5.11.8-artix1-1 and [third party for Arch](https://repo.m2x.dev/current/kernel-lts/x86_64/), version 5.4.107-1-lts54) and the kernel I usually use (vanilla patched with the [`-pf` patchset](https://gitlab.com/post-factum/pf-kernel/-/wikis/README)), where I noticed the problem first.

* On vanilla kernels, this was tested with two kinds of kernel configuration:
  1. The configuration I normally use (adapted to vanilla kernel by running a `make oldconfig` after copying in my `.config`),
  2. a minimal configuration generated by `make defconfig`, with a few changes to make the kernel boot my system, display graphics correctly, and enabling `/proc/config.gz` (details see file `how_configured-def.txt` in the attachment).


## For which kernels did the problem occur?

* This problem did occur for the following kernels and configurations:
  - vanilla kernels 5.10.27, 5.11.11, 5.13 with the configuration based on my usual configuration,
  - precompiled distribution kernel 5.11.8-artix1-1,
  - the kernel I usually use self-compiled.

* This problem did _not_ occur for the following kernels and configurations:
  - vanilla kernels 5.10.27, 5.11.11, 5.13 with the `make defconfig` based minimal configuration,
  - vanilla kernel 5.4.109 with any of the two configurations,
  - Precompiled distribution kernel 5.4.107-1-lts54.

* That the problem did not occur for kernels 5.4.* can also be just because they don't have a feature yet which triggers the problem.

* I have not bisected the problem down to the kernel configuration option(s) which trigger this.
  - What I have tried is to disable all the HDMI-related `CEC` options, which did _not_ help.


## What is my system?

A Laptop, GPD MicroPC, with intel UHD 605 graphics and Intel Celeron N4100 processor. It has a HDMI and a DisplayPort (the latter via USB-C) graphics output. I tested only for HDMI, USB-C I could not test due to lack of adapters. I do not have activated any automatic monitor configuration; I have to run `xrandr` (or a graphical tool) manually.


## What is the problem?

When I connect certain "modern" HDMI monitors to the running system with `Xorg` and then activate the monitor with `xrandr`, the internal display goes black (though backlight stays on). The external monitor receives singnal with the resolution set by `xrandr` (the monitor tells so), but it stays also black. (On just connection everything is still fine, activation does trigger the problem.)

I can connect to the system via ssh.

On the system, hardware-related tasks block (even causing `dmesg`-messages like `INFO: task Xorg:3138 blocked for more than 122 seconds.`), so I cannot anymore
* change monitor configuration,
* play sound,
* shut down the system,
* change from `Xorg` to a text mode virtual console via `chvt 1`
* query or set the brightness of the laptop display,
* start `alsamixer`
(maybe more, not tested.)

I still can log in via ssh, open, edit and save files, and sync and force-shutdown via `/proc/sysrq-trigger`.


## For which monitor connections and monitors does this problem occur?

* The problem _does_ occur when I directly connect from the laptop's HDMI output to the problematic monitor's HDMI input.
* The problem does _not_ occur when I connect that same monitor to it's DVI port via a HDMI to DVI adapter.
* It _does also_ occur when I connect to the monitor's HDMI port with an intermediate HDMI to DVI and then DVI to HDMI adapter.
* It also does not occur on every HDMI monitor:
  - I have observed it with a Samsung SyncMaster P2470HD, which also is a TV.
  - I have observed it with another Telefunken monitor which also is a TV, I cannot remember, it was in a hotel room.
  - I have _not_ observed it with a Dasung Paperlike HD-FT.


## Additional Information:

* It does not happen when connection is already present at boot; but dis- and reconnect make it happen.
* It does not happen under windows 10.

I attach the following files, all related to the vanilla 5.13 kernel (symptoms where similar or the same for the 5.11.11 and 5.10.27 kernels). Those files are inside the attached archive `hdmi_semi-crash_attachments.tar.xz` (It seems I can only attach one file here, so I did fall back to an archive):

* For the problematic kernel configuration:
  - `.config-my`: The kernel configuration used based on my configuration,
  - `dmesg_before-my.txt`: The output of `dmesg` just before activating the plugged-in monitor via `xrandr`,
  - `dmesg_after-my.txt`: The output of `dmesg` after activating the plugged-in monitor via `xrandr` and then after issuing `chvt 1` and `abeep` (alsa beep) and after waiting some time (to capture the "blocked tasks" messages),
  - `hung_tasks.dmesg-my.txt`: From `dmesg_after-my.txt`, just extracted the part with the "blocked tasks" messages,
  - `hwinfo-my.txt`: The output of `hwinfo`,
  - `dmidecode-my.txt`: The output of `dmidecode`,
  - `how_configured-my.txt`: A brief description on how the kernel was configured,
  - `uname-my`: The output of `uname -a`,
  - `symptoms-my.txt`: A description of symptoms after triggering the problem.
* For the `make defconfig` based kernel configuration:
  - `.config-def`: The `make defconfig`-based configuration,
  - `dmesg_before-def.txt`: The output of `dmesg` just before activating the plugged-in monitor via `xrandr`,
  - `dmesg_after-def.txt`: The output of `dmesg` after activating the plugged-in monitor via `xrandr` and then after issuing `chvt 1` and `abeep` (alsa beep) and after waiting some time (to capture the "blocked tasks" messages),
  - `hwinfo-def.txt`: The output of `hwinfo`,
  - `dmidecode-def.txt`: The output of `dmidecode`,
  - `how_configured-def.txt`: A description of how the kernel was configured,
  - `uname-def`: The output of `uname -a`,
  - `symptoms-def.txt`: A description of symptoms after activating the external monitor via `xrandr`.
* `used-kernel-source.url`: The URL I used to download the kernel from
* For reference: `.config-artix`: The kernel configuration of the also problem-triggering precompiled Artix distribution kernel version 5.11.8-artix1-1.
Comment 1 Drei Eck 2021-07-04 09:41:52 UTC
Created attachment 297709 [details]
The kernel configuration used based on my configuration
Comment 2 Drei Eck 2021-07-04 09:42:23 UTC
Created attachment 297711 [details]
The output of dmesg just before activating the plugged-in monitor via xrandr
Comment 3 Drei Eck 2021-07-04 09:42:53 UTC
Created attachment 297713 [details]
The output of dmesg after activating the plugged-in monitor via xrandr and then after issuing chvt 1 and abeep (alsa beep) and after waiting some time (to capture the "blocked tasks" messages)
Comment 4 Drei Eck 2021-07-04 09:43:21 UTC
Created attachment 297715 [details]
From dmesg_after-my.txt, just extracted the part with the "blocked tasks" messages
Comment 5 Drei Eck 2021-07-04 09:43:47 UTC
Created attachment 297717 [details]
The output of hwinfo
Comment 6 Drei Eck 2021-07-04 09:44:13 UTC
Created attachment 297719 [details]
The output of dmidecode
Comment 7 Drei Eck 2021-07-04 09:44:37 UTC
Created attachment 297721 [details]
A brief description on how the kernel was configured
Comment 8 Drei Eck 2021-07-04 09:44:57 UTC
Created attachment 297723 [details]
The output of uname -a
Comment 9 Drei Eck 2021-07-04 09:45:16 UTC
Created attachment 297725 [details]
A description of symptoms after triggering the problem
Comment 10 Drei Eck 2021-07-04 09:45:38 UTC
Created attachment 297727 [details]
The make defconfig-based configuration
Comment 11 Drei Eck 2021-07-04 09:46:00 UTC
Created attachment 297729 [details]
The output of dmesg just before activating the plugged-in monitor via xrandr
Comment 12 Drei Eck 2021-07-04 09:46:22 UTC
Created attachment 297731 [details]
The output of dmesg after activating the plugged-in monitor via xrandr
Comment 13 Drei Eck 2021-07-04 09:46:41 UTC
Created attachment 297733 [details]
The output of hwinfo
Comment 14 Drei Eck 2021-07-04 09:47:01 UTC
Created attachment 297735 [details]
The output of dmidecode
Comment 15 Drei Eck 2021-07-04 09:47:22 UTC
Created attachment 297737 [details]
A description of how the kernel was configured
Comment 16 Drei Eck 2021-07-04 09:48:07 UTC
Created attachment 297739 [details]
The output of uname -a
Comment 17 Drei Eck 2021-07-04 09:48:28 UTC
Created attachment 297741 [details]
A description of symptoms after activating the external monitor via xrandr
Comment 18 Drei Eck 2021-07-04 09:48:53 UTC
Created attachment 297743 [details]
The URL I used to download the kernel from
Comment 19 Drei Eck 2021-07-04 09:49:59 UTC
I have noticed that further files can be attached after submitting the report. So I now have attached all the files on by one as well (the archive I still leave here, since it contains the file sorted into directories by kernel and by configuration).
Comment 20 Drei Eck 2021-07-04 09:50:34 UTC
Created attachment 297745 [details]
For reference: The kernel configuration of the also problem-triggering precompiled Artix distribution kernel version 5.11.8-artix1-1
Comment 21 Drei Eck 2021-07-04 09:53:43 UTC
Created attachment 297747 [details]
For reference: The output of dmesg for the distribution kernel 5.11.8-artix1-1 before activating the external HDMI monitor with xrandr.

.. and for reference/interest, I also attach `dmesg_before-artix.log`, which shows the output of `dmesg` for the distribution kernel 5.11.8-artix1-1 before activating the external HDMI monitor with `xrandr`.
Comment 22 Jani Nikula 2021-11-09 08:00:21 UTC
Please file drm/i915 bugs at freedesktop.org gitlab.

https://gitlab.freedesktop.org/drm/intel/-/wikis/How-to-file-i915-bugs
Comment 23 Drei Eck 2021-11-10 19:30:40 UTC
Created attachment 299513 [details]
attachment-29200-0.sig

> Please file drm/i915 bugs at freedesktop.org gitlab.
> 
> https://gitlab.freedesktop.org/drm/intel/-/wikis/How-to-file-i915-bugs

Thank you for noting that!
Comment 24 Drei Eck 2022-05-10 12:03:09 UTC
The problem seems to have disappeared over time; I re-tested one of the problematic monitors with kernel [5.17-pf3](https://gitlab.com/post-factum/pf-kernel/-/wikis/README) ([which is based on 5.17.4](https://gitlab.com/post-factum/pf-kernel/-/tags)), and don't see the problem anymore.

So I changed the "Resolved"-status to "Obsolete".

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