With that USE flag, LXC, when init'd to something like bash, will stall, showing: read(5, 0x7ffd4fc96a20, 1024) = ? ERESTARTSYS (To be restarted if SA_RESTART is set) in strace Reproducible: Always Steps to Reproduce: 1. Install lxc with io-uring enabled 2. Import the stage3 tarball using 'https://wiki.gentoo.org/wiki/LXC#Local_template ( touch config tar -cJf metadata.tar.xz config lxc-create -t local -n gentoo-guest -- --fstree stage3.tar.xz --metadata metadata.tar.xz 3. Set 'lxc.init.cmd = /bin/bash' in the container config (/var/lib/lxc/gentoo-guest/config) 4. run "lxc-start gentoo-guest -F" 5. run a few commands until it stalls Actual Results: Bash eventually stalls, strace says read(5, 0x7ffd4fc96a20, 1024) = ? ERESTARTSYS (To be restarted if SA_RESTART is set)
Please give a gdb backtrace when stuck (gdb -p PID, then ^C, then bt). Also, do older versions work, and emerge --info please?
I just tested on my desktop and it happens on there too, System info: https://bpa.st/WWYQ Original system info: https://bpa.st/7A3A I rebuilt with debug syms: https://bpa.st/KDDA This was reproduced on my desktop, I don't think it matters much, but my desktop is using the dist kernel, unlike the original system
(In reply to zen from comment #2) > I just tested on my desktop and it happens on there too, System info: > https://bpa.st/WWYQ > > Original system info: > https://bpa.st/7A3A > > I rebuilt with debug syms: > https://bpa.st/KDDA > > This was reproduced on my desktop, I don't think it matters much, but my > desktop is using the dist kernel, unlike the original system Another BT, the first one stalled shortly after i attached, but resumed after i detached GDB. I've found opening nano reliably breaks it, so I opened that, then attached and ran BT. https://bpa.st/BT4A
Please always attach data, no pastebins, as they expire.
Created attachment 867719 [details] Emerge --info, and gdb data, in order
*sigh* not io-uring again. Would you happen to have any older kernels lying around to test with?
(In reply to Joonas Niilola from comment #6) > *sigh* not io-uring again. Would you happen to have any older kernels lying > around to test with? I do not, is there a specific version you have in mind? It takes around 10m per reboot on that first system, and I use it as a router, so would prefer to not be offline most of the time. The second system is my desktop and audio doesn't work below 6.3 (onboard audio).
So upstream did have some issues with io-uring, apparently they're supposed to be fixed in 6.8.4 which you're running. Did the issues start happening after an update to 6.8.4? can't really say what kernel version to test, apparently the issue is present in 6.1-LTS too. Fixed in 6.1.45. So if only you _can_, testing e.g. 6.1.41 and 6.1.45 would be valuable. But I guess if you connect the issues starting after 6.4, or 6.4.8, that's a hint too.