During S2R, secondary cores are hotplugged and as part of this, cpufreq offline callback is unregistering cpufreq cooling device which in turn removing kobject using device_del() and notifying userspace via uevent broadcast. This causes suspend abort for S2R use cases. Callstack is provided in attachment. Issue is there quite some time after this cpufreq patch [1] and it can happen to any chipset. I see someone already posted RFC patch to address this issue in [2] sometimes back. [1]. https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=5c238a8b599f1ae25eaeb08ad0e9e13e2b9eb023 [2]. https://patchwork.kernel.org/project/linux-pm/patch/20220710164026.541466-1-r.stratiienko@gmail.com/
Created attachment 305904 [details] Callstack
Seems like this issues is reported in other mailing lists: https://issuetracker.google.com/issues/194314791
Created attachment 305905 [details] wakeup source fix for the thermal cooling device A proposal to fix the issue. Not tested because I was not able to reproduce the issue on my platform yet.
(In reply to Daniel Lezcano from comment #3) > Created attachment 305905 [details] > wakeup source fix for the thermal cooling device > > A proposal to fix the issue. Not tested because I was not able to reproduce > the issue on my platform yet. Apparently this fix does not work.