Update notifications

[M-Bab]

This isn't a real issue. It is just there to inform interested people about pushed updates. I will add an entry on each pushed major update.

Please do not comment to this issue.

[M-Bab]

Updated to kernel 4.2 with updated firmware binaries.

[M-Bab]

Updated to kernel 5.4.8.

[JoneKone]

Oh man, after i bought Vega56 for me and my kid, this kernel has been the only thing that kept me sane, the computers just stopped crashing randomly thank you. (Black screen of death)

[M-Bab]

Updated to kernel 5.6.12 (debian+ubuntu variants). Updated firmware binaries package.

@JoneKone is the "oh man" an expression of concern this kernel died? It didn't - I just did not have a lot time recently. Another factor is that may graphics card is so old in the meantime it does not ultimately need this kernel. This was different when I started to build it (no HDMI sound and stuff).

[JoneKone]

@M-Bab It was a sigh of relief =) Vega 56 isnt' that old, and a guy on Linux gaming group telegram, is having so many issues with rx 5700xt... Similar issues that i'm having. So thankyou :)

[M-Bab]

Updated kernel to 5.6.19, updated firmware binaries. Kernel already includes code for upcoming sienna but the firmware binaries are not there yet.

[M-Bab]

Updated to kernel 5.9.0 -rc5. Happily recognized AMD rebased their kernel from 5.6 to a newer one. Also nice: Code for future graphic cards (sienna cichlid, navy flounder) already included. I will look out for the matching firmware binaries.

[M-Bab]

For the first time I also published it as release with assets: https://github.com/M-Bab/linux-kernel-amdgpu/releases/tag/linux-5.9.0-rc5-20.09.19.amdgpu . But not sure if this is more helpful.

[JoneKone]

https://www.phoronix.com/scan.php?page=news_item&px=Linux-5.9-rc6-Released Released With Soft Scrollback Removed, Performance Regression Fixed.

Hmm

[stuaxo]

I recently got an HP ds000xxx with a Ryzen 3700u and it's really broken as in ACPI they set "reduced hardware" to 1, so it doesn't boot by default [1].

Any idea if it will work any better with 5.9 ?

(I believe there was a bug where Linux didn't handle the "reduced hardware" setting very well)

[1 ]https://askubuntu.com/questions/1162746/why-wont-my-hp-envy-x360-laptop-with-ryzen-7-3700u-boot-ubuntu-live-cd

[M-Bab]

Updated to kernel 5.9.0 -rc6/-rc7. The last one wasn't a very good one unfortunately. On the one side the performance regression mentioned above and - at least for me - hibernation did not work. Both should be fixed now.

[lanigb]

Hello, thanks for your work. I installed the latest kernel you compiled (5.9.0-rc8-20.10.05.amdgpu.ubuntu) but was not able to get out of hibernation.

Oct 24 09:12:08 logodenn kernel: [  206.513174] amdgpu 0000:38:00.0: [drm:amdgpu_ring_test_helper [amdgpu]] *ERROR* ring gfx test failed (-110)
Oct 24 09:12:08 logodenn kernel: [  206.513210] [drm:amdgpu_device_ip_resume_phase2 [amdgpu]] *ERROR* resume of IP block <gfx_v9_0> failed -110
Oct 24 09:12:08 logodenn kernel: [  206.513212] amdgpu 0000:38:00.0: amdgpu: amdgpu_device_ip_resume failed (-110).
Oct 24 09:12:08 logodenn kernel: [  206.513216] PM: dpm_run_callback(): pci_pm_restore+0x0/0x90 returns -110
Oct 24 09:12:08 logodenn kernel: [  206.513223] PM: Device 0000:38:00.0 failed to restore async: error -110

I didn't have the amdgpu.dc=1 set in the boot parameters then. Do you understand the reason and would you know a fix ?

Moreover, syslog is full of messages like evbug: Event. Dev: input2, Type: 1, Code: 103, Value: 1

Do this messages relate to a real problem or may I safely blacklist the module?

TIA

[M-Bab]

Updated to kernel 5.9.1.

[M-Bab]

@lanigb Please first try the fresh 5.9.1 - there were some issues with hibernation in the past week. If it is not fixed please open with error log as a new issue.

[lanigb]

Hi, Came back home and tried the new kernel. It is actually worst. What log should I post ? syslog ? I don't have an error log.

[JoneKone]

Any news on getting 5.10.1 working?

[M-Bab]

@JoneKone Unfortunately amd-staging-drm-next has still not been rebased. I will precisely start to build kernels again once the first lines of this file: https://cgit.freedesktop.org/~agd5f/linux/tree/Makefile?h=amd-staging-drm-next show it has been lifted on any newer kernel than 5.9.

For know the deviation to the official sources of 5.9 are too big so it would be really effortful to get it merged and building (not even to mention working!) again. This also increases the risks of crashes and so on - so I would like to avoid that. So far the AMD guys were pretty fast jumping on the most recent kernel (except they stuck really long with 4.8 - but thats a while ago :smile: ).

[stuaxo]

Can us AMD users crowdsource an AMD focused kernel dev ? I don't see AMD doing enough to fix things that have been broken for years in Ryzen.

[M-Bab]

No idea. Usually the AMD open source team was pretty strong and active. No idea what is happening now. So far their usual repository feels somehow abandoned since almost 2 months.

I hope they will pick it up again.

[stuaxo]

Just airing a bit of frustration, certainly things are a lot better than they were.. a lot of my issues seem to be down to HPs broken bios. I have a laptop with a Ryzen 2500u that is pretty OK (but still with the odd freeze here and there). One with a 3700u that is unused rn because things were such a mess (very very broken bios), that I'm leaving it for a while to see if some of the fixes for the broken bios end up upstream.

[JoneKone]

Nice :) thx

[M-Bab]

Finally after quite a while I simply asked the AMD developers and they actually moved their repo (to https://gitlab.freedesktop.org/agd5f/linux/-/tree/amd-staging-drm-next ). The other repo was indeed abandoned.

So here we go. Kernel 5.11 based amd-staging-drm-next. The firmware files are also really outdated - maybe on the weekend I have time to collect all the neat little blobs together. As always: Use at your own risk.

[stuaxo]

Oh wow, fantastic news :)

[M-Bab]

Updated to kernel 5.11.8. Updated firmware binaries.

[M-Bab]

Updated to kernel 5.11.15. Updated firmware binaries.

Completely started from scratch because the previous kernel had graphic artifacts. Testing this one so far it looks fixed. Also updated the firmware binaries with ROCM blobs - so there are less warnings when this kernel is installed. And maybe someone actually wants to use this for open compute stuff.

[M-Bab]

Updated to kernel 5.11.18.

[stuaxo]

Alright, it's not out yet, but looks like 5.15 will finally do temperature reporting for Zen 2 / 3 APUs via k10temp with much qudos to Mario Limonciello.

https://www.phoronix.com/scan.php?page=news_item&px=Linux-5.15-HWMON

[M-Bab]

Its been a while ... Updated to kernel 5.13.18

@stuaxo I assume, although the kernel provides the newest AMD GPU features, such CPU features will arrive as slowly as for all other kernels ...

[M-Bab]

Updated to kernel 5.13.19

5.13 is end-of-life but amd-staging-drm-next is still on this version.

[JoneKone]

seems to be updated to 5.16

[JoneKone]

I'm little bit afraid to ask this.. Are you still alive?

[M-Bab]

I am :smile:

I did try to build 5.16 - but it did not build very well. Lots of merge conflicts - resolved them wrong it seems. Will try another round when I am back from vacation.

And also bad: Ubuntu abandoned their ubuntu-unstable repo where I could get all the Ubuntu modifications for each kernel versions. Now they only support the kernels they actually have in a distro (kind of makes sense, but bad for me). So probably the 5.16 kernel will not be available in Ubuntu flavor ...

[JoneKone]

:I uh oh, i have ubuntu :) have a nice vacation..

[stuaxo]

What a pain - onwards and upwards, hope the vacation is good.

[Mr-Precise]

I was able to build the 5.16 kernel, but it causes the graphics card to freeze for unknown reasons (HD 7750)

[M-Bab]

Update

My latest tries to build a mixed up kernel from the AMD code (and official linux kernel code and Ubuntu sauce) did not work well. There are massive merge conflicts and build errors which I dont have the time to figure out. But the reason for that is also kind of good news: The development in the official kernels is almost on-par with the latest AMD kernel developments.

I started this kernel because back in the days users didn't even have HDMI sound via AMD graphics card without such a special kernel. Things have improved a lot till then. The amdgpu dc code is merged for several years now and the software development there is so fast you usually get release day support for their new GPUs.

In short: This kernel is not really necessary anymore. I am not using it anymore either. I will shut these two repos down. Thanks to all users!

[themoonisacheese]

Nice job while you were there to help.

I'm a bit surprised that you say support is now release day because I had to use this no earlier than September on a 5600G. That said, installing Ubuntu in place of Debian fixed it as well. Godspeed

[FlorianMittag]

Thanks for the work you put into this. Without your kernel my work notebook made all sorts of problems.

Godspeed and please know that some people are really grateful for your time and effort you put into this.

[JoneKone]

Thank you :)

[stuaxo]

Thanks for all the work over the years keeping two successive laptops of mine working, it has been much appreciated.

[JoneKone]

Well umm so, there is this weird project going on. https://forum.level1techs.com/t/mi25-stable-diffusions-100-hidden-beast/194172 aand couldn't help but notice that weirdly specific kernel versions seem to match. to your "old" project.

[stuaxo]

I know people used this repo for various different issues - one issue I had was my laptop not coming back from suspend. amdgpu / kernel devs Mario Limonciello, Felix Kuehling and Aaron Liu helped me submit enough info that it will be mostly fixed in 6.4.

More info -

The issue: On an HP x360 with Ryzen 3700u: On returning from suspend the screen would be glitched or black, and I'd have to hard reset.

We found that fixes behaved differently depending on where the laptop gets power from (YMMV), the fix doesn't help if the laptop is powered by AC, but via USBC from a monitor or battery it now wakes up almost every time.

If you are having similar issues it may be worth building this kernel and testing.

Issue https://gitlab.freedesktop.org/drm/amd/-/issues/2449

Quick thread https://twitter.com/stu_axon/status/1648660362571464704

[M-Bab]

Pretty exciting there is still occasional news to this. And I am not even on Ubuntu anymore which means it would be even harder for me to build this kernel :sweat_smile: