Using unsquashfs to install live environments causes the environment to freeze using Calamares

[plougher]

It has been reported by Neal Gompa ([email protected])

My only issue with it is that using unsquashfs to install
live environments causes the environment to freeze, which is not
great. This doesn't affect Anaconda, but it does affect Calamares
(which Fedora itself does not use).

Unsquashfs uses getrlimit() (equivalent to Bash ulimit -n) to get the maximum number of files that can be opened. It may be the case that opening that many files causes read operations to be throttled, causing the system to hang. Obviously a solution here is to set the maximum number of files to a smaller value before calling Unsquashfs. But Unsquashfs could probably open less files automatically without affecting performance.

This issue needs to be reproduced and investigated.

[Conan-Kudo]

I will see if I can produce a test-case ISO for you to see this with. I had adjusted Calamares in Fedora to work around it, but fixing this would be better.

[plougher]

I will see if I can produce a test-case ISO for you to see this with. I had adjusted Calamares in Fedora to work around it, but fixing this would be better.

A test-case ISO would be a great help :-)

[yump]

I think the kernel has a general problem with unrelated tasks getting starved of CPU when there is heavy load in the I/O path. Last month I was btrfs receive-ing onto a disk with compress-force=zstd:12, and it lagged the GUI worse than any amount of SCHED_OTHER userspace CPU load.

Updating flatpaks often causes a few brief hiccups too, and of course there's the infamous USB stick stall problem.

IIRC Fedora kernels are CONFIG_PREEMPT_DYNAMIC now. Does the problem go away if you boot with preempt=full?