On Wed, Jul 28, 2021 at 07:02:11PM +0200, bof@bof.de wrote:
Hi,
first regression report in a long while... hope I'm doing this right.
I run self-build, from time to time, mainline kernels, in production, both on HP DL hosts + inside kvm VMs on these. I usually stick to a certain long term stable line for a year or three. As such, for a bit longer already, I'm running 5.4, starting somewhere at 5.4.4x. At the moment, I started to roll out 5.4.135 on a few of the hosts, and encounter a rare "WARNING: CPU ... PID ... at kthread_is_per_cpu+0x1c/0x20" splat (see full dmesg below)
Previous kernel I was running there, never seeing that, was 5.4.114
Searching a bit, turns up this 5.4.118 commit / backport: https://lore.kernel.org/stable/162074016514831@kroah.com/ https://lore.kernel.org/lkml/20210407220628.3798191-2-valentin.schneider@arm... pointing to exactly the place shown in the dmesg trace.
Can you narrow this down to a specific commit using 'git bisect'?
thanks,
greg k-h