Linux 5.5-rc2 Kernel Released

[phoronix]

Phoronix: Linux 5.5-rc2 Kernel Released

On schedule with one week since the closure of the Linux 5.5 kernel merge window and subsequent release candidate, Linux 5.5-rc2 is out this evening for testing...

http://www.phoronix.com/scan.php?page=news_item&px=Linux-5.5-rc2-Released

[muncrief]

Darn, resume from suspend is still broken for R9-390 GPUs.

[theriddick]

Darn, resume from suspend is still broken for R9-390 GPUs.

I think I remember having this issue back in my 390 days, I believe I found a fix for it and posted it over at the whirlpool net forums but can't remember what it was.

I'll try to dig it up... can't find it, too old but I did find this bug report and a possible patch you could try applying yourself.

https://bugzilla.kernel.org/show_bug.cgi?id=204241

[Guest]

Darn, resume from suspend is still broken for R9-390 GPUs.

Wait what? Which kernel driver are you using it with - radeon or AMDGPU ? I'm using an R9 390 with AMDGPU on Linux 5.4.x and it works fine for me.

Edit: Ugh, just checked the bug report linked by theriddick and saw your comment. Damn, looks like I need to prevent updating to 5.5 until that's fixed. Or, just switch over to using the nice Intel iGPU that just works.

[R41N3R]

My AMD Ryzen 5 2500U in the HP Envy x360 is really unstable since Linux 5.4, GPU crashes and resets all the time in Plasma Wayland and in Firefox with webdender. Going back to Linux-lts in Arch Linux is not a good workaround, as then standby is creating the crashes :-(

[regis]

My AMD Ryzen 5 2500U in the HP Envy x360 is really unstable since Linux 5.4, GPU crashes and resets all the time in Plasma Wayland and in Firefox with webdender. Going back to Linux-lts in Arch Linux is not a good workaround, as then standby is creating the crashes :-(

Same here with a Matebook D 14 and the same processor, Linux 5.4 and up are completely unstable and GPU crash/reset every few minutes. Previous kernel do work but still freeze after a few hours to a few days at best.

AMD Ryzen APU support are really bad under linux...

[NateHubbard]

Same here with a Matebook D 14 and the same processor, Linux 5.4 and up are completely unstable and GPU crash/reset every few minutes. Previous kernel do work but still freeze after a few hours to a few days at best.

AMD Ryzen APU support are really bad under linux...

Maybe this is more of a laptop thing? I don't have any issues with my 2200G or 220GE.
I'm running 5.4.2 on both of them.

[muncrief]

I think I remember having this issue back in my 390 days, I believe I found a fix for it and posted it over at the whirlpool net forums but can't remember what it was.

I'll try to dig it up... can't find it, too old but I did find this bug report and a possible patch you could try applying yourself.

https://bugzilla.kernel.org/show_bug.cgi?id=204241

Thank you for the suggestion. I actually posted this problem there last night and they said it sounded like a different issue and I had to file another ticket. Everything works fine with kernel 5.4.x, it's just 5.5 that immediately reintroduced the problem. I had it when 4.2 first came out, but within a few iterations it had been fixed. And everything from then until 5.5 had worked great.

[nuetzel]

Thank you for the suggestion. I actually posted this problem there last night and they said it sounded like a different issue and I had to file another ticket. Everything works fine with kernel 5.4.x, it's just 5.5 that immediately reintroduced the problem. I had it when 4.2 first came out, but within a few iterations it had been fixed. And everything from then until 5.5 had worked great.

I've read your post and Alex's reply...
...soo, it's realy the time for a 'git bisect', even that it was fixed after 4.20+ and now broken, again.
Start with
git bisect start
git bisect bad v5.5-rc1
git bisect good v5.4

You can start with only amdgpu changes, too.

[muncrief]

I've read your post and Alex's reply...
...soo, it's realy the time for a 'git bisect', even that it was fixed after 4.20+ and now broken, again.
Start with
git bisect start
git bisect bad v5.5-rc1
git bisect good v5.4

You can start with only amdgpu changes, too.

Yes, I actually spent 3 days trying to bisect the original 4.20+ problem but was never able to do so. I remember learning that resume problems were particularly difficult to bisect, and that someone was working on some kind of new system specifically for suspend/resume problems, but thank goodness someone else was able to fix it.

The problem was that bisect would always end with something that had nothing to do with the problem, and I was actually ridiculed quite harshly for not being able to do it correctly.

I don't know why but many Linux devs seem to take a lot of joy in being down right mean, especially in the Arch community, and I really don't want to go there again.