FS#76072 - [boinc] OpenCL tasks error out on for all applications when using AMD GPUs

Attached to Project: Community Packages
Opened by Mattia (mattblack) - Friday, 30 September 2022, 10:01 GMT
Last edited by Toolybird (Toolybird) - Sunday, 16 April 2023, 08:20 GMT
Task Type Bug Report
Category Packages
Status Closed
Assigned To Felix Yan (felixonmars)
Architecture All
Severity Low
Priority Normal
Reported Version
Due in Version Undecided
Due Date Undecided
Percent Complete 100%
Votes 0
Private No

Details

Description:
It seems that the systemd unit file for boinc, specifically this line `ProtectSystem=strict` doesn't allow GPU task to execute.
I attched the log entry which is a bit misleading, then I tried to run boinc as another user in another folder without systemd and noticed everything was working, from that I tried different configs till I get to the root issue, ProtectSystem.
If I comment it out GPU tasks are worked correctly and finish with no issues

Additional info:
* 7.20.2-1
* [13:01:23][57640][INFO ] Application startup - thank you for supporting Einstein@Home!
[13:01:23][57640][INFO ] Starting data processing...
[13:01:23][57640][INFO ] Using OpenCL platform provided by: Advanced Micro Devices, Inc.
[13:01:23][57640][INFO ] Using OpenCL device "gfx1031" by: Advanced Micro Devices, Inc.
[13:01:23][57640][ERROR] Couldn't create OpenCL command queue (error: -6)!
[13:01:23][57640][INFO ] OpenCL shutdown complete!
[13:01:23][57640][ERROR] Demodulation failed (error: 2013)!
[13:01:23][57640][WARN ] Sorry, at the moment your system doesn't have enough free CPU/GPU memory to run this task!


Steps to reproduce:
1) install boinc with `pacman -S boinc`
2) install opencl-amd opencl-headers from AUR
3) Launch any GPU task, a `Computation Error` is returned every time
This task depends upon

Closed by  Toolybird (Toolybird)
Sunday, 16 April 2023, 08:20 GMT
Reason for closing:  Upstream
Additional comments about closing:  See the upstream ticket. The cause is understood and can be worked around with a systemd drop-in until upstream get around to fixing it.
Comment by Toolybird (Toolybird) - Monday, 03 October 2022, 05:20 GMT
That setting is a security hardening measure implemented by upstream [1]. Have you raised your issue with upstream? I found a related issue here [2]. Have you completely followed the Arch wiki [3]? In particular, added the boinc user to the "video" user group?

[1] https://github.com/BOINC/boinc/pull/2873
[2] https://github.com/BOINC/boinc/issues/3355
[3] https://wiki.archlinux.org/title/BOINC#Projects_using_GPU
Comment by Mattia (mattblack) - Monday, 03 October 2022, 09:05 GMT
Hi,

I have raised now an issue against upstream here => https://github.com/BOINC/boinc/issues/4948, I didn't find interestingly the linked issue when I was debugging.
I followed fully the wiki, this is the output of checking groups

~ » groups boinc
video boinc
Comment by Toolybird (Toolybird) - Sunday, 09 October 2022, 02:17 GMT
New post in the linked upstream issue pointing towards this not being an Arch packaging issue, but rather an upstream problem.

Loading...