Accidentally attached SATA SSD to appvm and crashed dom 0, anyway to revert?

User Support
1

I was testing giving a PCI based storage device their own sys-qubes for backups, and accidentally attached the SATA device which is my Dom 0 generic SSD, to the new qube and slowly watched my system die, as all vms started halting, also got an error that said something like "can’t find dom 0 sata’ or something similar. I knew it was hopeless but I tried to detach and then delete that qube from the gui, but with no hard drive connected anymore… it didn’t respond.

I thought maybe this wouldn’t save and just hard reset. Nope, doesn’t boot anymore.

Is there any way to access shell and qvm-something to delete that sys-qube and reattach my sata drive mapping?

I have qubes I never backed up :frowning:

Damn I feel like an idiot.

2

not sure if it will work, but I’ll try attaching the drive of the crashed install to a working qubes install using this method, How to mount a Qubes partition from another OS | Qubes OS

Then will try cloning the vms, as mentioned here, Secondary storage | Qubes OS

But it doesn’t make sense to me how I’ll be able to clone the vms of the external and get them detected or in a usable state in the working qubes install. So it’s probably pointless, but I’m getting desperate. Been a month without a properly working system.

3

You can boot from whatever linux rescue media you like, decrypt & mount the Qubes OS file system as with any other Linux.

Then either edit the Qubes config file that contains the attached device (probably /var/lib/qubes/qubes.xml - better create a backup first) or chroot to that install and use qvm-block etc. to detach the drive.

Anyway I don’t see too much security benefits from a dedicated SATA PCI controller.

4

Are you sure it’s possible to use chroot?

I’m new to chroot, but mounted the messed up drive in a new qubes install’s dom 0, then followed this guide How to: Chroot into a broken system via live CD/ISO or alternate Linux system | TurnKey GNU/Linux

Then tried commands

qvm-device pci detach sys-MSATA:00:1f.2

Which didn’t seem to work.

Then tried to see if I can just delete the qube. Seems whatever command I enter in chroot I get the following:

failed to connect to qubesd service

And a bunch of other commands.

I tried going back and mounting all the folders on the drive for chroot, and redid the command to remove the template, now it says doesn’t exist, but when I do sudo pvs and lvs, still clearly there.

:frowning:

5

Well I was able to backup some stuff by getting access to the various home directories.

The error has changed from a hang to now a kernal panic.

Not sure what to do next. I’ll save this HD for troubleshooting later and will pick up another HD today and just start over.

6

Hm looks like the qvm-* commands require qubesd to be running which of course it is not in a chroot environment. You also wouldn’t want it to be running as it would attempt to attach the device to that VM etc.

Then edit /var/lib/qubes.xml manually and remove the incorrect PCI device (should be in some tag). Ideally Qubes should start just fine afterwards. Better make a backup of the config file though…

7

I just ran into this same problem. I fixed my problem thanks to this forum post! Here’s what I did:
I booted the system in TAILS.
Found the Dom0 partition as per the instructions here: How to mount a Qubes partition from another OS | Qubes OS
Found /var/lib/Qubes/qubes.xml in Dom0.
Ran in TAILS lspci and found that SATA is: id="00_1f.2
Found the line that had 00_1f.2, and that was under sys-usb.
I deleted that line.
Rebooted the system.
It worked!