Re: Help me fix a regression caused by 56b9918490 (PM: sleep: Simplify suspend-to-idle control flow)

["Kenneth R. Crudup" <kenny@panix.com>]

On Fri, 22 Nov 2019, Rafael J. Wysocki wrote:

> First off, please try to change the sleep_no_lps0 ACPI module
> parameter (/sys/module/acpi/parameters/sleep_no_lps0) to 1 and see if
> that makes any difference.

I did try that; didn't seem to help. I will run a few tests to make sure,
though.

> If it helps, I don't think we can do much except for blacklisting your machine.

Urgh. Blacklisting from what/which facility?

Every now and then, when this thing goes to sleep it mostly goes down to 50mA
(measured by the USB-C charging port) idle current; I wish it were less, but
this is "modern" suspend. But every Nth suspend it will draw 180 to 240 mA for
reasons I can't deduce, and I was wondering if perhaps we're getting wedged in
an intermediate state on the way into s2idle. That's why I'm trying to make
sure I have access to every possible suspend enhancement.

At one point it was a regression in i915 that caused a high suspend current
at all times and I helped Chris Wilson and those guys track down, but that's
been OK for a while.

One thing that I do get is the dump on resume when it doesn't go into S0ix
(intel_pmc_core.warn_on_s0ix_failures=1) and there's a number of what I
assume are clocks and power rails that are still "on"; I imagine some of
that is done by the BIOS (e.g., "MPHY_CORE_GATED", "CNV_VNNAON_REQ_ACT").
Do you know what these are mnemonics for?

Thanks,

	-Kenny

-- 
Kenneth R. Crudup  Sr. SW Engineer, Scott County Consulting, Silicon Valley