v4.8-rc1: thinkpad x60: running at low frequency even during kernel build

Commit Message

Pandruvada, Srinivas Nov. 5, 2016, 1:55 p.m. UTC

On Sat, 2016-11-05 at 14:37 +0100, Pavel Machek wrote:
> Hi!

> 

> > 

> > [...]

> > 

> > > 

> > > So we seem to have thermal or ACPI regression in v4.9-rc3.

> > > 

> > It is possible. Can you add either add printk

> > in acpi_processor_ppc_has_changed() or use ftrace and see do you

> > get to

> > these functions

> > 

> > acpi_processor_ppc_init()

> > acpi_processor_ppc_has_changed()

> > acpi_processor_ppc_notifier()

> > 

> > ?

> 

> Yes, these seem to be called. Here's the log:


Try this

1. Either enable dyndebug or add 

#define DEBUG 1 
at the top of acpi-cpufreq.c


2.

                                     pr->performance->states[ppc].
                                     core_frequency * 1000);

Thanks,
Srinivas

Comments

Pavel Machek Nov. 5, 2016, 8:31 p.m. UTC | #1

Hi!

BTW.. This machine has nasty habit of hanging during kernel boot when
it is "hot".. which makes reboots unplesant here. Ideas would be
welcome how to debug that.

Best regards,
									Pavel

Pavel Machek Nov. 9, 2016, 11:34 a.m. UTC | #2

Hi!

Thanks to Srinivas, bug tracking moved to bugzilla at
https://bugzilla.kernel.org/show_bug.cgi?id=187311 , it is regression
from v4.8-final.

Easiest way to observe it is that cpufreq/bios_limit does not change
in v4.9, where it goes lower with high temperature on v4.8.

Best regards,

								Pavel

Pavel Machek Nov. 14, 2016, 7:03 p.m. UTC | #3

Hi!

Bisection was not fun, but I've got the result:

# first bad commit: [6ea8c546f3655a81f82672f24b66dad6095bdd07] ACPICA:
FADT support cleanup

I've reverted the patch on top of 4.9-rc4, and thermal management now works.

More details are in https://bugzilla.kernel.org/show_bug.cgi?id=187311

As this breaks thermal management on both thinkpad X60 and T40p, can
I ask for a revert?

If you have other ideas, I can test them, but as this is just a
cleanup, it can wait for 4.10...

Thanks,
									Pavel

Rafael J. Wysocki Nov. 14, 2016, 7:54 p.m. UTC | #4

On Mon, Nov 14, 2016 at 8:03 PM, Pavel Machek <[e-mail]> wrote:
> Hi!
>
> Bisection was not fun, but I've got the result:
>
> # first bad commit: [6ea8c546f3655a81f82672f24b66dad6095bdd07] ACPICA:
> FADT support cleanup

Bob?

> I've reverted the patch on top of 4.9-rc4, and thermal management now works.
>
> More details are in https://bugzilla.kernel.org/show_bug.cgi?id=187311
>
> As this breaks thermal management on both thinkpad X60 and T40p, can
> I ask for a revert?
>
> If you have other ideas, I can test them, but as this is just a
> cleanup, it can wait for 4.10...

Right.

Besides, we're in the -rc6 time frame now, so I'll queue up a revert.

Thanks,
Rafael
--
To unsubscribe from this list: send the line "unsubscribe platform-driver-x86" in
the body of a message to [e-mail]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch

diff --git a/drivers/acpi/processor_perflib.c
b/drivers/acpi/processor_perflib.c
index bb01dea..6074995 100644
--- a/drivers/acpi/processor_perflib.c
+++ b/drivers/acpi/processor_perflib.c
@@ -94,9 +94,14 @@  static int acpi_processor_ppc_notifier(struct
notifier_block *nb,
 
        ppc = (unsigned int)pr->performance_platform_limit;
 
+       printk(KERN_ERR "ppc = %d\n", ppc);
+
        if (ppc >= pr->performance->state_count)
                goto out;
 
+       printk(KERN_ERR "ppc = %d freq-limit %d\n", ppc, pr-
>performance->states[ppc].
+                                     core_frequency * 1000);
+
        cpufreq_verify_within_limits(policy, 0,