Installer does not boot - nvme timeout completion polled

Hello everyone, after some reading I discovered that iso images with the newest kernels exist, so I decided to try the newest available (5.18.16-1.fc32.qubes.x86_64), since my hardware requires it.

The installer however never completes the boot sequence and when I launch it in verbose mode I get a few errors, with the last one listed here being the one that repeats indefinitely every 1 to 3 minutes or so.

AMD-Vi: AMD IOMMUv2 functionality not available on this system - This is not a bug.
A start job is running for udev Wait for Complete Device Initialization
xhci_hcd : Error while assigning device slot ID
xhci_hcd : Max number of devices this xHCI host supports is 64.
xhci_hcd : WARN: xHC save state timeout
usb usb3-port3: couldn't allocate usb_device
nvme nvme0: I/O # QID 0 timeout, completion polled

(with # being different numbers, since the last error occurs multiple times with different IDs).

My machine: 2022 G14 GA402RK, Ryzen 9 6900HS, RX 6800S.
AHCI drive mode, secure boot disabled.

Currently running Fedora 36 (5.19.7-200.fc36.x86_64) with no issues whatsoever.

Contrary to what the first error in the qubes installer shows, IOMMU is loaded:

journalctl -b0 | grep IOMMUv2
Sep 09 10:47:19 fedora kernel: AMD-Vi: AMD IOMMUv2 loaded and initialized

Try disabling hyperthreading in your firmware configuration utility. I’m trying to get an install done on a recent AMD system (5700X, recent board, yes, I know this is too bleeding edge for reliable Qubes), and I was having the same error. After some investigation revealed other people were seeing similar errors after dropping CPU cores offline, my speculation was that Qubes not using all the cores was triggering the same sort of kernel error.

I had to dig into some advanced CPU overclocking settings to find the option to disable HT entirely, but after that, the installer booted on the stock “default latest download” ISO, despite having thrown the NVMe timeout errors previously.

We’ll see how it works. It informed me that the CPU lacks all the other stuff needed, but I’m pretty sure a 5700X has an IOMMU/hardware virt/etc…

Just wanted to point you to my issues booting the installer a Ryzen 7 6800U, which appear identical to yours.

I’ve encountered the same issue.
From my research, this seems to be an issue with earlier versions of Fedora up to f35 (USB installer uses f32).
I will test with a different NVMe drive soon to see if it persists.
You could try booting from a detatched LUKs header, which managed to get me a little further.
See guide here: Playing with qubes part 2 (Note: follow video instructions vs text, as there are some inconsistencies if you choose BTRFS)

You could also try disabling IOMMU, this should get past these errors into the installer, and re-enabling once installation is complete.