On Mon, Nov 21, 2022 at 08:50:11PM +0100, David Hildenbrand wrote:

On 21.11.22 20:03, Luis Chamberlain wrote:

quoted

On Mon, Nov 21, 2022 at 05:00:49PM +0100, Petr Pavlu wrote:

quoted

On 11/16/22 17:03, Prarit Bhargava wrote:

quoted

On 11/15/22 14:29, Luis Chamberlain wrote:

quoted

On Mon, Nov 14, 2022 at 04:45:05PM +0100, David Hildenbrand wrote:

quoted

Note that I don't think the issue I raised is due to 6e6de3dee51a.
I don't have the machine at hand right now. But, again, I doubt this will
fix it.

There are *more* modules processed after that commit. That's all. So
testing would be appreciated.

Can anyone tell us if

https://lore.kernel.org/linux-pm/20221102195957.82871-1-stuart.w.hayes@gmail.com/ (local)

resolves the module loading delay problem?

This patch unfortunately makes no difference on my test system. In my case,
the kernel has already intel_pstate loaded when udev starts inserting a burst
of acpi_cpufreq modules. It then causes the init function acpi_cpufreq_init()
to immediately return once the check cpufreq_get_current_driver() fails. The
code modified by the patch is not reached at all.

To be clear I don't care about the patch mentioned in the above URL, I care
about this:

https://lkml.kernel.org/r/d0bc50e3-0e42-311b-20ed-7538bb918c5b@suse.com

David was this the on you tested too?

Yes, that's the one I tried without luck.

OK thanks. I'm just trying to sort out the regression first before we go
out and fix another issue. We will chase that issue down though and I
hav some other ideas to help with this further too.

  Luis