So Qubes 4.1.1 is released and some part of that update causes intermittent freezing/crashing of the entire system. This has been reported by many and there doesn’t appear to be any obvious fix.
My solution is to reinstall Qubes 4.1.0 and not update dom0, but the issue I now face is that my Whonix VMs auto restart as I can’t apply the dom0 update that prevents this.
I think I can get around this by updating only the necessary package in dom0 and hoping it doesn’t pull in any dependencies that cause the freezing to return (I assume it won’t). I would be doing something like this in dom0:
My question is, am I potentially setting myself up for something malicious to occur? Obviously I’m bypassing the salt update but I assume that as soon as the issue is fixed in Qubes upstream I can just go back to normal update and everything will be fine.
The only alternative would be to manually modify the files that need changing but I would really prefer to avoid this.
On the Whonix side, updating the Templates should be sufficient. There was no need for any security fixes in qubes-core-admin-addon-whonix ever at time of writing. (If there was, check git history for actual changes made in the repository.
I’ve got a link in there to the discussion of the crashing in case you’re interested.
As a related question, is my assumption that it is definitely possible for a crash to introduce unnoticed filesystem corruption correct? I personally don’t feel comfortable using a whonix workstation based on a template that was being modified at the time of a crash because who knows what could’ve happened. Is this a valid concern or am I being too paranoid here?
The Qubes forums and several bug reports have various users complaining about system freezes and/or crashes which aren’t recoverable. Even unman was complaining about his “rock solid” system experiencing the same issue, usually soon after launching VMs.
Could be a serious kernel issue? As Kees Cook notes, stable kernel releases have around 100 fixes per week, so maybe one of those screwed the pooch in recent times in the 5.15 kernel series. (BTW there is around 1,000 outstanding bugs on the syzkaller list these days) Or it could have been one of those recent firmware updates, as that happened around the same time people started complaining.
Possibly also the Xen grant tables issue was related, as Marek cherry-picked a patch for the recent 5.15.63-1 dom0 stable kernel update as there was some critical fault there in the logic - hopefully that improves things.
As to the original question, I’d recommend updating as usual and staying on the stable release channel and hope for the best. They’ll work it out sooner or later.
I was referring to the bit flip leading to a leak, which I agree is highly unlikely but it is slightly more likely than highly unlikely when one’s system is crashing 3-4 times a day, so I either run that risk or use an outdated dom0. Neither is ideal.