Thread (6 messages) 6 messages, 3 authors, 2021-08-31

Re: stable 5.4.135 REGRESSION / once-a-day WARNING: at kthread_is_per_cpu+0x1c/0x20

From: Patrick Schaaf <hidden>
Date: 2021-08-31 04:47:06
Also in: regressions

Looking into this again.

Unfortunately, couldn't see how I would do bisection on the issue, as
it appears with that 5.4.118 commit implicated by the call stack,and
with tha tremoved,is obviously gone(that I tested, 5.4.135 with
b56ad4febe67b8c0647c0a3e427e935a76dedb59 reverted runs smoothly for
me, while the original 5.4.135 with that 5.4.118 time commit in, now
on a dozen machines, throws the WARNING.

I got email on the side of someone (Igor, on Cc) who sees the same
with DELL servers, a newer 5.10 kernel, for him running IPVS + he sees
actual operational impact there.

I just had a look at Linus mainlinetree, and see there is this
followup / further fix from Peter Zijlstra,
https://github.com/torvalds/linux/commit/3a7956e25e1d7b3c148569e78895e1f3178122a9
; now I'm much too incompetent to try and backport that, as it looks
more involved, but I imagine such a backport would be needed to fix
the WARNING (or IPVS breakage of Igor) we see.

best regards
  Patrick
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help