[WORKING] Amd RX 9070xt support

TomiWebPro said:
Code: 
root@tomiwebpro:~# lspci -nn | grep -i amd
02:00.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD/ATI] Navi 10 XL Upstream Port of PCI Express Switch [1002:1478] (rev 24)
03:00.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD/ATI] Navi 10 XL Downstream Port of PCI Express Switch [1002:1479] (rev 24)
only two pcie switches, which I think is not the vga device, how to resolve this? Many thanks!
Click to expand...
What is the output of lspci -nnks 04:00? Have you tried running update-pciids to update the PCI(e) informational database?
 
leesteken said:
What is the output of lspci -nnks 04:00? Have you tried running update-pciids to update the PCI(e) informational database?
Click to expand...
After I ran update-pciids :
Code: 
root@tomiwebpro:~# lspci -nnks 04:00
root@tomiwebpro:~# lspci -nnks 03:00
03:00.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD/ATI] Navi 10 XL Downstream Port of PCI Express Switch [1002:1479] (rev 24)
        Subsystem: Advanced Micro Devices, Inc. [AMD/ATI] Navi 10 XL Downstream Port of PCI Express Switch [1002:1479]
        Kernel driver in use: pcieport
root@tomiwebpro:~# lspci -nnks 02:00
02:00.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD/ATI] Navi 10 XL Upstream Port of PCI Express Switch [1002:1478] (rev 24)
        Subsystem: Sapphire Technology Limited Navi 10 XL Upstream Port of PCI Express Switch [1da2:1478]
        Kernel driver in use: pcieport
root@tomiwebpro:~#

There was no output for 04:00
 
leesteken said:
Did you plug in the power cables to GPU? What is the PSU rating/capacity? Can you boot any other operating system (from DVD or USB) and does it show the GPU?
Click to expand...
Yes I did plug in GPU power, PSU is a 1000W jump started only for the gpu, I'll check it for other systems tomorrow with another computer
 
Does it work properly in Windows with that computor?

I had a similar situation and the problem was that it was not recognized by BIOS.

The rx 9070 series requires a Bios update to be recognized, but my too old computor (x10srm-tf) did not have this.

It was not an OS or other problem, but a BIOS problem with the motherboard

I solved this problem by replacing the motherboard/CPU/memory.
 
uzumo said:
Does it work properly in Windows with that computor?

I had a similar situation and the problem was that it was not recognized by BIOS.

The rx 9070 series requires a Bios update to be recognized, but my too old computor (x10srm-tf) did not have this.

It was not an OS or other problem, but a BIOS problem with the motherboard

I solved this problem by replacing the motherboard/CPU/memory.
Click to expand...
That might be the case for me, but all the motherboard I have are x99 and h61 based, I hope there is still legacy support for older motherboards, I would have to check with my friend's computer 12100F then.
 
NoobSaidiop said:
@kprinssu Thank you. No longer having issues when rebooting. The VM seems to crash (internal-error status in proxmox) when I run games though.
Click to expand...
I have confirmed that when I start the game it stops with the same error.

internal-error

Intel Core Ultra 265k
Asrock Z890 Pro RS WiFi White
Hellhound Spectral White AMD Radeon RX 9070 XT 16GB GDDR6

rarely works fine.
The logs are shown attached.

Any advice would be appreciated.
 

Attachments

  • IMG_3067.jpeg
    IMG_3067.jpeg
    24.6 KB · Views: 10
  • IMG_3068.jpeg
    IMG_3068.jpeg
    493 KB · Views: 9
  • IMG_3065.jpeg
    IMG_3065.jpeg
    314.8 KB · Views: 9
kprinssu said:
Can you give this script a go? There are certain edge cases where the VM may shutdown (e.g. manually shutting the VM via a command within the VM) which causes the hook not to run. When the VM starts, the script belo, will always unbind the vfio kernel module and bind the amdgpu module.

Code: 
#!/bin/bash

phase="$2"
if [ "$phase" == "pre-start" ]; then
     echo "vmGPU: Running pre-start hook"
     # Unbind gpu from vfio-pci
     sleep 5
     echo "0000:0f:00.0" > /sys/bus/pci/drivers/vfio-pci/unbind 2>/dev/null
     sleep 2
     echo "vmGPU: unbound vfio kernel module"

     # Bind amdgpu
     echo "0000:0f:00.0" > /sys/bus/pci/drivers/amdgpu/bind 2>/dev/null
     sleep 2

     echo "vmGPU: bound amdgpu kernel module"
elif [ "$phase" == "post-stop" ]; then
    echo "vmGPU: Running post-stop hook"
    # Unbind gpu from vfio-pci
    sleep 5
    echo "0000:0f:00.0" > /sys/bus/pci/drivers/vfio-pci/unbind 2>/dev/null
    sleep 2
    echo "vmGPU: unbound vfio kernel module"

    # Bind amdgpu
    echo "0000:0f:00.0" > /sys/bus/pci/drivers/amdgpu/bind 2>/dev/null
    sleep 2
    echo "vmGPU: bound amdgpu kernel module"
fi
Click to expand...
I think I figured out my issue. Where would you add the bar resize line (echo 3 > /sys/bus/pci/devices/0000:0f:00.0/resource2_resize) in this script?
 
NoobSaidiop said:
I think I figured out my issue. Where would you add the bar resize line (echo 3 > /sys/bus/pci/devices/0000:0f:00.0/resource2_resize) in this script?
Click to expand...
In my experience this need to be done without any driver bound to the device. I (would) do it in the pre-start phase after unbinding any current driver (which might be vfio-pci but could also be the actual device driver).
 
NoobSaidiop said:
@uzumo Disabling Resizable Bar and 4G decoding in the BIOS settings fixed my issue
Click to expand...

Thank you

I may have solved this in another way.

I had foolishly gone with 6.14.

I had this problem with kernel 6.14, but when I pinned to kernel 6.11, the problem stopped earlier.

The rebar is still working with it enabled.

apt policy proxmox-kernel-6.11
apt update
apt install proxmox-kernel-6.11
proxmox-boot-tool kernel pin 6.11.11-2-pve
reboot
 

Attachments

  • IMG_3212.jpeg
    IMG_3212.jpeg
    322.1 KB · Views: 9
  • IMG_3213.jpeg
    IMG_3213.jpeg
    48.6 KB · Views: 8
uzumo said:
Thank you

I may have solved this in another way.

I had foolishly gone with 6.14.

I had this problem with kernel 6.14, but when I pinned to kernel 6.11, the problem stopped earlier.

The rebar is still working with it enabled.

apt policy proxmox-kernel-6.11
apt update
apt install proxmox-kernel-6.11
proxmox-boot-tool kernel pin 6.11.11-2-pve
reboot
Click to expand...
Thanks @uzumo , I'll give this a shot.
Can you also share the hookup script you're using for your VM?
 
NoobSaidiop said:
Thanks @uzumo , I'll give this a shot.
Can you also share the hookup script you're using for your VM?
Click to expand...
The scripts made by kprinssu are the best.

Can be used as is.

nano rx9070_reset.sh

code_language.shell: 
#!/bin/bash

phase="$2"

echo "Phase is $phase"

if [ "$phase" == "pre-start" ]; then

    # Unbind gpu from amdgpu

    echo "0000:0f:00.0" > /sys/bus/pci/drivers/amdgpu/unbind 2>/dev/null

    sleep 2

    # Resize the GPU's BAR2 memory region (useful for PCI passthrough)

    echo 3 > /sys/bus/pci/devices/0000:0f:00.0/resource2_resize

    sleep 2

elif [ "$phase" == "post-stop" ]; then

    # Unbind gpu from vfio-pci

    sleep 5

    echo "0000:0f:00.0" > /sys/bus/pci/drivers/vfio-pci/unbind 2>/dev/null

    sleep 2

    # Bind amdgpu

    echo "0000:0f:00.0" > /sys/bus/pci/drivers/amdgpu/bind 2>/dev/null

    sleep 2

fi

cd /var/lib/vz/
mkdir snippets
cd snippets
cp /root/rx9070_reset.sh ./
chmod +x rx9070_reset.sh
qm set <vmid> --hookscript local:snippets/rx9070_reset.sh

0000:0f:00.0 needs to be changed.
 
Last edited:
  • Like
Reactions: NoobSaidiop
My environment is kernel 6.11 and this procedure works, but is anyone else having success with kernel 6.14?

I would like to review it if it is just my environment, but I would like to know if there are others.

If it always happens, I want to know because eventually I won't be able to stay on kernel 6.11.

The following thread discusses a problem with nvidia also causing virtual machines to stop while running.

A

Post in thread 'Opt-in Linux 6.14 Kernel for Proxmox VE 8 available on test & no-subscription'

GPU passthrough from pre 6.12 upwards crashes my linux VMs (I have no idea what happens in windows, I don't use it) - I found the problematic kernel commit:
Problematic commit

This commit produces an instant VM crash in certain gpu utilizations (in my case a simple youtube video on brave browser) with the following report from the kernel:

Apr 11 17:28:32 pve QEMU[92966]: RAX=000018cc0e283840 RBX=0000000000000780 RCX=0000000000000780 RDX=0000000000000780
Apr 11 17:28:32 pve QEMU[92966]: RSI=000018cc0e283840 RDI=00007a71e5c0f000 RBP=00007a71ed884960 RSP=00007a71ed884960
Apr 11 17:28:32...
  • Like

https://bugzilla.kernel.org/show_bug.cgi?id=220057
 
Last edited:
uzumo said:
My environment is kernel 6.11 and this procedure works, but is anyone else having success with kernel 6.14?

I would like to review it if it is just my environment, but I would like to know if there are others.

If it always happens, I want to know because eventually I won't be able to stay on kernel 6.11.

The following thread discusses a problem with nvidia also causing virtual machines to stop while running.

A

Post in thread 'Opt-in Linux 6.14 Kernel for Proxmox VE 8 available on test & no-subscription'

GPU passthrough from pre 6.12 upwards crashes my linux VMs (I have no idea what happens in windows, I don't use it) - I found the problematic kernel commit:
Problematic commit

This commit produces an instant VM crash in certain gpu utilizations (in my case a simple youtube video on brave browser) with the following report from the kernel:

Apr 11 17:28:32 pve QEMU[92966]: RAX=000018cc0e283840 RBX=0000000000000780 RCX=0000000000000780 RDX=0000000000000780
Apr 11 17:28:32 pve QEMU[92966]: RSI=000018cc0e283840 RDI=00007a71e5c0f000 RBP=00007a71ed884960 RSP=00007a71ed884960
Apr 11 17:28:32...
  • Like

https://bugzilla.kernel.org/show_bug.cgi?id=220057
Click to expand...
Looks like it's resolved
Thanks so much!
 
You must log in or register to reply here.
Top Bottom