You are not logged in.

#1 2018-04-09 11:09:29

vanquish
Member
Registered: 2013-12-28
Posts: 48

AMD Raven Ridge 2200g/2400g Issues

Hello,

i've started to use a AMD 2200g on a Asus Rog Strix B350-F with RX580. And suprise surprise wink ... I have issues.

Kernel 4.15.x-x and 4.16.0-2 have same behaviour.

If I start the system with iGPU and dGPU enabled I have "only" ACPI issues and on Kernel 4.16.0-2 I have in addtion to that a flickering Xorg/KDE-Plasma Desktop.*

GhCVl1px.jpg

If i try to start the system with iGPU DISABLED and dGPU ENABLED the system will PANIC on both Kernel. System won't boot in any case (tried several bootparameters).

EW8IFFtJ.jpg

Before I will file any Bug to anyone. I want to clear if its a Hardware/Bios issue and/or a Kernel/Bug.

The flickering Desktop is obviously caused by the amdgpu driver (I guess).

Enable and Disable iGPU is no problem for Windows 10 with this system. So I think the Bios is doing it right. But I'm unsure about that. Because if the HDMI Cable is plugged in on iGPU the Bios will bring up iGPU first and enable it in Bios settings by itself. So disable means not disable for sure. If the cable is unplugged it's working as intended.

* On this specific System i would like to disable the iGPU. This is because Xorg/SDDM is quite confused about two amdgpu driven GPUs. Mostly SDDM/Xorg is starting on dGPU and iGPU is blank (if both is enabled). Sometimes I only get the Shell. On Windows I'm able to enable both GPUs but the Radeon Settings Tool is confused about both GPUs and don't want to start. Even the Installer is confused and is not able to find any GPU if both are enabled. -.-

So what options do I have: Compile Kernel 4.17, Compile Mesa 18 and LLVM? Compile Plasma 5.13? Is it worth to do this work? Are there any RR specific Kernel patches which I could use?

Thanks for any advise to track this down.
Kind regards

Last edited by vanquish (2018-04-09 11:11:22)

Offline

#2 2018-04-09 12:53:48

Buddlespit
Member
From: Chesapeake, Va.
Registered: 2014-02-07
Posts: 449

Re: AMD Raven Ridge 2200g/2400g Issues

I've had ACPI errors on my Zen 1700X from the very beginning (08/17). I've tried adding/removing chipset drivers, adjusting UEFI and/or various kernel boot options. I've gotten to the point where I just ignore the errors because my system still boots and runs just fine. As for your AMD-Vi errors, have you tried turning off AMD-Vi and IOMMU in the UEFI?


AMD Ryzen 7 1700X | Asus PRIME X370-A | GeForce GTX 980 Ti | 16G Ram | KDE Plasma | Zen Kernel
Until linuxcounter.com closed, I was registered linux user #560427

Offline

#3 2018-04-09 14:22:51

vanquish
Member
Registered: 2013-12-28
Posts: 48

Re: AMD Raven Ridge 2200g/2400g Issues

Hello Buddlespit,

thanks for your reply.

I guess the ACPI errors are not essential for running Linux on this System. I think they are just the connectors for the ASUS proprietary reading/settings stuff like LED Lights/Tool and Afterburner Tool. But I don't know for sure.

Yeah AMD-V can be disabled I know. But I'm using 2 virtual Machines were I pass one Harddrive. So it would be nice if I could use that. For the moment I'm running with iommu=soft to disable it.

On this specific system I do not understand why iommu is an issue with just iGPU disabled and enabled its not. Of course the message "Unable tow rite to IOMMU perf counter" is there too.

Is it worth to file a bug report or not? I guess it's not if I look arround for Ryzen came out one year ago. I don't understand why IOMMU is still issue nowdays. Is it a chipset bug like on Haswell Chipsets x87 where I had to fuck up passthrough on kvm with intremap=off 4 years ago? It's just frustrating. ....

Offline

#4 2018-04-09 19:47:08

Buddlespit
Member
From: Chesapeake, Va.
Registered: 2014-02-07
Posts: 449

Re: AMD Raven Ridge 2200g/2400g Issues

It could very well be a firmware issue. I had problems up until about November/December of last year. Once the agesa 1006a was released, I stoppped having firmware/ uefi issues (other than the acpi errors). Have you upgraded your firmware yet? How about your chipset driver? It took me a while to find the right chipset for my PRIME X370-A. I was using the it87-dkms-git until it stopped working for me, about the same time linux-4.15 and systemd-238 were released.


AMD Ryzen 7 1700X | Asus PRIME X370-A | GeForce GTX 980 Ti | 16G Ram | KDE Plasma | Zen Kernel
Until linuxcounter.com closed, I was registered linux user #560427

Offline

#5 2018-04-10 12:12:40

Lone_Wolf
Member
From: Netherlands, Europe
Registered: 2005-10-04
Posts: 7,135

Re: AMD Raven Ridge 2200g/2400g Issues

https://en.wikipedia.org/wiki/Input-out … ement_unit

Contrary to Intel, AMD uses IOMMU for many things on x86_64 systems.
So much infact that the iommu user virtualization benefits are similar to an unimportant extra.

The fact you switched to iommu = soft degrades performance and can result in loss of functionality.

A Firmware issue seems to be the most likely cause.


Multi-init booting with apg Openrc and systemd coexisting
Automounting : not needed, i prefer pmount
Aur helpers : makepkg + my own local repo === rarely need them

Offline

#6 2018-04-11 09:13:55

vanquish
Member
Registered: 2013-12-28
Posts: 48

Re: AMD Raven Ridge 2200g/2400g Issues

Many Thanks to you both helping me to come to a realistic assessment. So I have to blame AMD and/or the motherboard manufacturer Asus. I'll file a bug to Asus in hope they will fix this. *ingodwetrust* ;P

One further question:

The Error "AMD-Vi: unable to write to IOMMU perf counter" essential for faultless working of AMD-Vi/IOMMU?

Last edited by vanquish (2018-04-11 09:14:24)

Offline

#7 2018-04-11 11:08:37

Buddlespit
Member
From: Chesapeake, Va.
Registered: 2014-02-07
Posts: 449

Re: AMD Raven Ridge 2200g/2400g Issues

vanquish wrote:

Many Thanks to you both helping me to come to a realistic assessment. So I have to blame AMD and/or the motherboard manufacturer Asus. I'll file a bug to Asus in hope they will fix this. *ingodwetrust* ;P

One further question:

The Error "AMD-Vi: unable to write to IOMMU perf counter" essential for faultless working of AMD-Vi/IOMMU?

Since Windows doesn't have the virtualization issues that linux users have, I wouldn't surmise that it's the hardware itself that's to blame. And yes, just about any IOMMU error is going to stop hardware sharing of resources (GPU).


AMD Ryzen 7 1700X | Asus PRIME X370-A | GeForce GTX 980 Ti | 16G Ram | KDE Plasma | Zen Kernel
Until linuxcounter.com closed, I was registered linux user #560427

Offline

#8 2018-04-11 12:10:18

vanquish
Member
Registered: 2013-12-28
Posts: 48

Re: AMD Raven Ridge 2200g/2400g Issues

Sure, not the Hardware itself. For Windows 10 i cannot say that IOMMU/AMD-Vi is working flawlessly. I think it is a Bios and/or AMD Firmware issue paired with some OS specific issues. Further I think Windows is just booting regardless if there are any AMD-Vi/IOMMU errors. Linux is just saying "panic".

Here is a thread about IOMMU on Raven Ridge I've found recently:

https://github.com/RadeonOpenCompute/RO … /issues/42

I think it is not a mistake to address this issue to Asus. But I do not expect anything from them. I never got any serious support from any hardware vendor in connection with Linux issues. -.-

Last edited by vanquish (2018-04-11 12:16:22)

Offline

Board footer

Powered by FluxBB