On Sun, Jan 24, 2021 at 2:49 PM Stephen Berman [off-list ref] wrote:

On Mon, 04 Jan 2021 16:38:43 +0100 Stephen Berman [off-list ref] wrote:

quoted

On Thu, 31 Dec 2020 21:46:11 +0100 "Rafael J. Wysocki" [off-list ref] wrote:

quoted

ATM, I'm tempted to do something like the patch below (with the rationale
that it shouldn't be necessary to read the temperature right after updating
the trip points if polling is in use, because the next update through polling
will cause it to be read anyway and it will trigger trip point actions as
needed).

Stephen, can you give it a go, please?

On Sat, 02 Jan 2021 12:03:17 +0100 "Rafael J. Wysocki" [off-list ref] wrote:

quoted

There is one more way to address this, probably better: instead of checking the
temperature right away in acpi_thermal_notify(), queue that on
acpi_thermal_pm_queue
and so only if another thermal check is not pending.

This way there will be at most one temperature check coming from
acpi_thermal_notify() queued up at any time which should prevent the
build-up of work items from taking place.

So something like this:

Thanks for the patches.  I'll try them as soon as I can.

FTR, since this is the thread I started for this bug, I've confirmed in
https://lore.kernel.org/lkml/87y2gi78sg.fsf@gmx.net/T/#t (local) that the latest
patch fixes the bug.

OK, thanks!

The patch has been applied as 5.11-rc material.