Ryan Wilson
Ryan Wilson
Did some debugging and found that it seems to be a regression in the `full` type in `memory.pressure` which is what systemd-oomd uses for oom kills What we see on...
Per Johannes: ``` Thanks for checking, this definitely looks like the issue I mentioned. The state warning is issued only once per boot when a corruption is first detected to...
@bluca AFAICT the regression-causing commit - 152e11f6df29 ("sched/fair: Implement delayed dequeue") - isn't on 6.11 (e.g. I don't see SCHED_FEAT(DELAY_DEQUEUE, true) in sched/features.h). However, I can follow up to make...
@yuwata looks like the kernel fix is in 6.12-rc4: https://github.com/torvalds/linux/commit/c6508124193d42bbc3224571eb75bfa4c1821fbb Worth testing on Fedora rawhide again, not sure if Rawhide is pulling latest RC build for 6.12 or using static...
@poettering I noticed we have a ProtectControlGroups setting...it might be a bit confusing for the user to have 2 settings - ProtectControlGroups and PrivateControlGroups - control how /sys/fs/cgroup is mounted....
@poettering yea `strict` sounds good to me
Tommy Unger is working on adding get-counters functionality as part of his ramp-up!