[SOLVED] GPU passthrough "AER: can't recover (no error_detected callback)"

[turborierer]

Hi all,

I'm struggling for 2 weeks now to the the GPU passthrough running again. It worked but all of a sudden the VM started to display the message:

guest has not initialized the display (yet)

So I did a lot of variants in the GRUB file, all started with

GRUB_CMDLINE_LINUX_DEFAULT="quiet intel_iommu=on pcie_acs_override=downstream,multifunction video=efifb:off video=vesa:off vfio-pci.ids=10de:1f06,10de:10f9,10de:1ada,10de:1adb vfio_iommu_type1.allow_unsafe_interrupts=1 kvm.ignore_msrs=1 modprobe.blacklist=radeon,nouveau,nvidia,nvidiafb,nvidia-gpu"

Which lead to the following errors on the host:

   37.872433] nvidia-gpu 0000:02:00.3: AER: can't recover (no error_detected callback)
[   37.872436] pcieport 0000:00:02.0: AER: device recovery failed
[   37.872437] pcieport 0000:00:02.0: AER: Multiple Uncorrected (Non-Fatal) error received: 0000:02:00.0
[   37.872589] vfio-pci 0000:02:00.0: PCIe Bus Error: severity=Uncorrected (Non-Fatal), type=Transaction Layer, (Requester ID)
[   37.875688] vfio-pci 0000:02:00.0:   device [10de:1f06] error status/mask=00100000/00000000
[   37.877501] vfio-pci 0000:02:00.0:    [20] UnsupReq               (First)
[   37.879289] vfio-pci 0000:02:00.0: AER:   TLP Header: 40000008 000002ff e0062000 f7f7f7f7
[   37.881334] xhci_hcd 0000:02:00.2: AER: can't recover (no error_detected callback)

Than I changed the GRUB config with the

pcie_aspm=off

command what I found somewhere in the net, and indeed the VM was booting again but now it gets stuck here:

Started Update UTMP about System Runlevel Changes

Some research pointed me to remove and reinstall gnome but after the reboot the VM stops at the exact same point.

Im at the end of my Linux knowledge and maybe someone sees the real bug.

The machine is a HP Z440 workstation with a GeForce RTX 2060 SUPER.

Thank you so much.

 

[leesteken]

Notihng changed? You did no switch from pve-kernel-5.13 to version 5.15? Maybe unplug and replug the GPU (a few times to remove corrosion on the PCIe connectors)?

 

[turborierer]

Actually I already tried a 2nd GPU with the exact same problem.
Im running pve-manager/7.2-3/c743d6c1 (running kernel: 5.15.30-2-pve).

 

[Neobin]

5.15.30-2-pve

https://forum.proxmox.com/threads/q...king-with-gpu-passthrough.117550/#post-509094

 

[turborierer]

Ok, so finally I got it running.
Updating 5.15 was not enough, I had to use the experimental 5.19.

After that I enden up with the BAR 0: can't reserve [mem 0xe0000000-0xefffffff 64bit pref] error.

But this error I had already fixed on a different machine according to this thread:https://forum.proxmox.com/threads/problem-with-gpu-passthrough.55918/page-3#post-484395

Thank you so much to point that update out @Neobin