On Tue, Apr 12, 2022 at 10:09:30PM +0000, Limonciello, Mario wrote:
[Public]
Hi,
There are a variety of x86 systems that advertise LPI support and as part of negotiation with firmware they don't end up using C-states in certain circumstances. This leads to higher runtime power consumption and also failure to enter s2idle.
In mainline there have been changes to block that behavior. Can you please backport these two commits from mainline? commit 01f6c7338ce267959975da65d86ba34f44d54220 ("cpuidle: PSCI: Move the `has_lpi` check to the beginning of the function") commit eb087f305919ee8169ad65665610313e74260463 ("ACPI: processor idle: Check for architectural support for LPI")
This should go to 5.15.y and later stable kernels.
Queued for 5.4 and newer kernels. If you want this backported to older ones, please provide a working set of patches.
thanks,
greg k-h