On 12/8/22 03:44, Luis Chamberlain wrote:

On Wed, Dec 07, 2022 at 04:15:21PM +0100, Petr Mladek wrote:

quoted

Reviewed-by: Petr Mladek <pmladek@suse.com>

Queued onto modules-next.

Thanks both.

quoted

Of course, the ideal solution would be to avoid the multiple
loads in the first place. AFAIK, everything starts in the kernel
that sends the same udev events for each CPU...

Fixes go first, *then we can address enhancements. I have some old
fixes I can send after htis is merged. I believe folks have others.

This simpler fix should be sufficient and so the original more complex
approach can be scraped. My plan is then to only progress the additional
developed patches: a test case for the issue, minor clean-up patches, and an
ACPI change to load acpi_cpufreq/pcc_cpufreq only once.

Note that none of this work now addresses the problem that David Hildenbrand
encountered with vmap allocation issues [1]. That needs to be fixed
separately.

[1] https://lore.kernel.org/linux-modules/20221013180518.217405-1-david@redhat.com/ (local)

Cheers,
Petr