Support saving boot error log files somewhere besides the boot drive (e.g., a USB drive)

[brenthl88]

(and for some reason, the keyboard doesn't work)

The solution you'd like

Allow automatically saving the text file on another USB storage device (besides the boot drive)

[andrewdavidwong]

This enhancement request is rather underspecified. Could you elaborate on your use case?

[brenthl88]

SVM Mode: Enabled SMT Mode: Auto IOMMU: Enabled

I was attempting to install QubesOS on my DIY computer with a x570 AORUS Ultra 1.1, 5950X and a 5700 XT (Phantom Gaming D). However, I encountered problems like dracut-pre-udev[805]: modprobe: ERROR: could not insert 'floppy'. No such device. All other items say [ OK ]

After which, I get a long line of: dracut - initqueue[1230]: Warning: dracut - initqueue timeout - starting timeout scripts Warning: Could not boot. Starting Dracut emergency shell. Warning: /dev/root does not exist

And for some reason at that point, my keyboard doesn't work (it probably gets unmounted after the dracut timeout processes)

These errors below were during my installation, but with the speed the logs popped up, I wasn't able to check them.

audit: typ=1334: prog-id=7 (and down to 0) op=UNLOAD

systemd-gpt-auto-ererator[1242]: EFI loader partition unknown, exiting

systemd-gpt-auto-ererator[1242]: (The boot loader did not set EFI variable LoaderDevicePart UUID)

While the errors I got were from a past installation of Qubes 4.0.4 and my objective was to upgrade from 4.0.4 to 4.1.0 (same hardware), using journalctl:

Fast TSC calibration failed

dom0_kernel: disabled

Setting APIC routing to Xen PV

NUMA turned off

Faking a mode

Yama becoming mindful

cpu 0 spinlock event irq 57 VPMU disabled by hypervisor (probably to mitigate Spectre attacks on a Zen 3?)

PERF NMI watchdog permanently disabled

hugetlbfs: disabling because there are no supported hugepages sizes

PM-timer failed consistency check (0xffffff) - aborting

usb: port power management may be unreliable

AMD-VI: AMD IOMMUv2 driver by Joerg Rodel

AMD-VI: AMD IOMMUv2 functionality not available on this system

amdgpu 0000:0a:00.0: amdgpu: SMU driver if not matched

KD_FONT_OP_GET failed while trying to cppy the font metadata: Invalid argument

Failed to activate with specified passphrase (Passphrase incorrect)? Too many attempts. Giving up. Failed with result exit code. Failed to start cryptography setup. Dependency failed for local encrypted volumes.

cryptsetup.target Job cryptsetup.target/start failed with result dependency

kauditd_printk_skb: 14 callbacks suppressed

audit type 1130 res=failed

Other roundabout methods I tried to install 4.1.0 (still hasn't worked)

• using Ubuntu to create a live mount of Qubes while verifying the images
• installing Ubuntu to remove any encrypted partition then installing 4.1.0

Currently using Fedora Media Writer to create a live QubesOS device (didn't work)

EDIT:

[ghost]

The main question at hand is, how can one configure the install to save logs on another storage medium? In scenarios where cpu hardlocks occur and no usb-keyboard input works, manually copying the logs isnt possible, unless I have overlooked a kernel option to overwrite the log path to another storage medium?

@brenthl88

I had very similar experiences from OP's issues in the comment, with my 5950X, instead with a B550M Aurous Elite (Rev 1.3). I'm guessing AMD Agesa updates could be part of the OP's issue at hand with hardware support.

Unfortunately my board lacks older bios versions with AMD Agesa V2 1.2.0.7 <=> 1.2.0.8 as my only options. Both HCL reports i see with a 5950X + Asus B550 board, advise their bios version of 2423 (AGESA V2 PI 1.2.0.3 Patch C), I just looked through your mobo rev and you have more chance of rolling back bios to find a more stable bios for xen.

The only way for me to get through the install without any critical errors (during/post install) is to manually change CCD Control from auto to 1 in my bios options, effectively telling my 16C CPU to only use 8 cores, disabling the last CCD.

Once installed (including the initial template setup), I can manage to set CCD Control back to auto, amending dom0_max_vcpus=8 dom0_vcpus_pin=0-7 to the default xen cmdline opts in grub, I could manage to get pass Luks, but couldn't get into dom0 as hard locks would occur, presumably with the default VM's autostart using 2 vCPUs.

Soo back to bios I went, disabled the second CCD again, back into dom0 and changed all sys VM's to 1 core. This gave me more of a shot at getting into dom0 but hard locks still occurred before logging in, so I left sys-usb as the only autostart which helped.

I haven't managed to wrap my head around xen vcpu pools yet but I think it would be the only way to utilise all 16 cores in a particular workload that doesn't cause hardlocks.