Random freezes on 5.4.* kernels

[kvaps]

Initially please make sure the problem hasn't been already addressed:

• Are you using the latest driver?
  (https://www.displaylink.com/downloads/ubuntu)
• Are you using the latest EVDI version?
  (https://github.com/DisplayLink/evdi/releases)
• If you are using a DisplayLink device, have you checked 'troubleshooting'
  on DisplayLink's website?
  (https://support.displaylink.com/knowledgebase/topics/103927-troubleshooting-ubuntu)
• Is this issue related to evdi/kernel?
  (if it is rather connected to DisplayLinkManager please take a look at support
  https://support.displaylink.com or forum https://www.displaylink.org/forum/)

Some information for baseline would help a lot:

• Linux distribution and its version: Arch Linux (upgrade 2020.01.08)
• Linux kernel version: linux-5.4.2.arch1-1 and linux-5.4.8.arch1-1
• Xorg version (if used) 1.20.6
• Desktop environment in use: Xorg + awesome wm

Hi, after upgrading my distro evdi started freezing quite often, it is usually happens every 1-2 minute with duration about 10-15 seconds.

I've tested linux-5.4.2.arch1-1 and linux-5.4.8.arch1-1 kernels, both have the same problem, after fallback to 5.3.13-arch1-1 evdi working as it should.

[asterlab]

got the same issue with kernel 5.4.0-2-amd64

[rghv]

Can validate; witnessed the same on 5.4.12 (taken via ukuu) on Linux Mint 19.3 Tricia.

Noticed issues similar to #180 and #160 as well.

[abucodonosor]

Are you guys using Intel GPU(s)?

[kvaps]

Are you guys using Intel GPU(s)?

My system:

• OS: Arch Linux x86_64
• Host: Lenovo ideapad 720S-13ARR
• Kernel: 5.3.10-arch1-1
• CPU: AMD Ryzen 7 2700U
• GPU: AMD Radeon RX Vega 10

[abucodonosor]

@kvaps

can you attach a dmesg from a broken kernel?

You can boot such a kernel, let it break then reboot to a good one and use journalctl
to show the dmesg. Maybe you already have such a dmesg, just use journalctl --list-boots to see your entries then journalctl -b the-entire -o cat >dmesg.txt

[abucodonosor]

Also for people with Intel HW and this issue please see #173

[kvaps]

@abucodonosor I'll do this, short note: kernel is not hanging fully, its got unstuck after certain amount of time

[abucodonosor]

@kvaps

even better, for sure there are some traces in the dmesg then.

[kvaps]

Hi @abucodonosor, after every stuckage I see the following messages:

[  133.132542] [drm:amdgpu_job_timedout [amdgpu]] *ERROR* ring gfx timeout, but soft recovered
[  161.293215] [drm:amdgpu_job_timedout [amdgpu]] *ERROR* ring gfx timeout, but soft recovered
[  190.946188] [drm:amdgpu_job_timedout [amdgpu]] *ERROR* ring gfx timeout, but soft recovered
[  200.973338] [drm:amdgpu_job_timedout [amdgpu]] *ERROR* ring gfx timeout, but soft recovered
[  214.199950] [drm:amdgpu_job_timedout [amdgpu]] *ERROR* ring gfx timeout, but soft recovered

Looks like related to https://bbs.archlinux.org/viewtopic.php?id=251159

[kvaps]

Problem solved for me by adding iommu=pt to my kernel cmdline

[abucodonosor]

@kvaps

that is a kernel bug itself it seems. Adding iommu=pt is a workaround, maybe it is fixed in 5.5.

[kvaps]

OK, let's see.

[kvaps]

@abucodonosor, you're right it seems no more freezes on 5.5 even without iommu=pt

[abucodonosor]

@kvaps

Good. They fixed a similar bug on Intel GPUs in 5.5 also.