X86_PM_TIMER: /proc/cpuinfo doesn't get updated

X86_PM_TIMER: /proc/cpuinfo doesn't get updated

[Luca]

Hi,
I have a notebook with an Athlon-M CPU. I tried linux 2.6.4 with
CONFIG_X86_PM_TIMER=y and I noticed that /proc/cpuinfo doesn't get
updated when I switch frequency (via sysfs, using powernow-k7). The is
issue seems cosmetic only, CPU frequency changes (watching
temperature/battery life).

Without X86_PM_TIMER I see the actual speed in /proc/cpuinfo. It seems
that cpu_khz isn't updated on frequency change, but I don't know how
this can affect the rest of the system...

Luca
-- 
Home: http://kronoz.cjb.net
Tentare e` il primo passo verso il fallimento.
Homer J. Simpson


-------------------------------------------------------
This SF.Net email is sponsored by: IBM Linux Tutorials
Free Linux tutorial presented by Daniel Robbins, President and CEO of
GenToo technologies. Learn everything from fundamentals to system
administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click

Re: X86_PM_TIMER: /proc/cpuinfo doesn't get updated

[Daniele Venzano]

On Tue, Mar 16, 2004 at 07:22:57PM +0100, Luca wrote:
> Hi,
> I have a notebook with an Athlon-M CPU. I tried linux 2.6.4 with
> CONFIG_X86_PM_TIMER=y and I noticed that /proc/cpuinfo doesn't get
> updated when I switch frequency (via sysfs, using powernow-k7). The is
> issue seems cosmetic only, CPU frequency changes (watching
> temperature/battery life).

I can confirm, I'm seeing the same behavior. Please note that the
bogomips count gets updated, it's only the frequency that doesn't
change.

Thanks.

-- 
-----------------------------
Daniele Venzano
Web: http://teg.homeunix.org



-------------------------------------------------------
This SF.Net email is sponsored by: IBM Linux Tutorials
Free Linux tutorial presented by Daniel Robbins, President and CEO of
GenToo technologies. Learn everything from fundamentals to system
administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click

Re: [ACPI] X86_PM_TIMER: /proc/cpuinfo doesn't get updated

[Karol Kozimor]

Thus wrote Daniele Venzano:
> > I have a notebook with an Athlon-M CPU. I tried linux 2.6.4 with
> > CONFIG_X86_PM_TIMER=y and I noticed that /proc/cpuinfo doesn't get
> > updated when I switch frequency (via sysfs, using powernow-k7). The is
> > issue seems cosmetic only, CPU frequency changes (watching
> > temperature/battery life).
> I can confirm, I'm seeing the same behavior. Please note that the
> bogomips count gets updated, it's only the frequency that doesn't
> change.

Same here with a P4-M, follow-up to John and Dmitry.
Best regards,

-- 
Karol 'sziwan' Kozimor
sziwan@hell.org.pl

Re: [ACPI] X86_PM_TIMER: /proc/cpuinfo doesn't get updated

[Dmitry Torokhov]

On Tuesday 16 March 2004 04:42 pm, Karol Kozimor wrote:
> Thus wrote Daniele Venzano:
> > > I have a notebook with an Athlon-M CPU. I tried linux 2.6.4 with
> > > CONFIG_X86_PM_TIMER=y and I noticed that /proc/cpuinfo doesn't get
> > > updated when I switch frequency (via sysfs, using powernow-k7). The is
> > > issue seems cosmetic only, CPU frequency changes (watching
> > > temperature/battery life).
> > I can confirm, I'm seeing the same behavior. Please note that the
> > bogomips count gets updated, it's only the frequency that doesn't
> > change.
> 
> Same here with a P4-M, follow-up to John and Dmitry.
> Best regards,
> 

PM timer does not install CPUFREQ handler which would scale cpu_khz to
give proper display. I might cook up something later tonight.

-- 
Dmitry

Re: [ACPI] X86_PM_TIMER: /proc/cpuinfo doesn't get updated

[john stultz]

On Tue, 2004-03-16 at 13:42, Karol Kozimor wrote:
> Thus wrote Daniele Venzano:
> > > I have a notebook with an Athlon-M CPU. I tried linux 2.6.4 with
> > > CONFIG_X86_PM_TIMER=y and I noticed that /proc/cpuinfo doesn't get
> > > updated when I switch frequency (via sysfs, using powernow-k7). The is
> > > issue seems cosmetic only, CPU frequency changes (watching
> > > temperature/battery life).
> > I can confirm, I'm seeing the same behavior. Please note that the
> > bogomips count gets updated, it's only the frequency that doesn't
> > change.
> 
> Same here with a P4-M, follow-up to John and Dmitry.

Hmm. This is untested, but I think this should do the trick.

Dominik: Is there any reason I'm not seeing why cpu_khz should only be
updated when using the TSC?

thanks
-john

===== arch/i386/kernel/timers/timer_tsc.c 1.36 vs edited =====
--- 1.36/arch/i386/kernel/timers/timer_tsc.c	Tue Feb  3 21:35:49 2004
+++ edited/arch/i386/kernel/timers/timer_tsc.c	Tue Mar 16 15:23:49 2004
@@ -360,8 +360,8 @@
 		if (variable_tsc)
 			cpu_data[freq->cpu].loops_per_jiffy = cpufreq_scale(loops_per_jiffy_ref, ref_freq, freq->new);
 #ifndef CONFIG_SMP
+		cpu_khz = cpufreq_scale(cpu_khz_ref, ref_freq, freq->new);
 		if (use_tsc) {
-			cpu_khz = cpufreq_scale(cpu_khz_ref, ref_freq, freq->new);
 			if (variable_tsc) {
 				fast_gettimeoffset_quotient = cpufreq_scale(fast_gettimeoffset_ref, freq->new, ref_freq);
 				set_cyc2ns_scale(cpu_khz/1000);

Re: X86_PM_TIMER: /proc/cpuinfo doesn't get updated

[john stultz]

On Tue, 2004-03-16 at 15:19, Dmitry Torokhov wrote:
> On Tuesday 16 March 2004 04:42 pm, Karol Kozimor wrote:
> > Thus wrote Daniele Venzano:
> > > > I have a notebook with an Athlon-M CPU. I tried linux 2.6.4 with
> > > > CONFIG_X86_PM_TIMER=y and I noticed that /proc/cpuinfo doesn't get
> > > > updated when I switch frequency (via sysfs, using powernow-k7). The is
> > > > issue seems cosmetic only, CPU frequency changes (watching
> > > > temperature/battery life).
> > > I can confirm, I'm seeing the same behavior. Please note that the
> > > bogomips count gets updated, it's only the frequency that doesn't
> > > change.
> > 
> > Same here with a P4-M, follow-up to John and Dmitry.
> > Best regards,
> > 
> 
> PM timer does not install CPUFREQ handler which would scale cpu_khz to
> give proper display. I might cook up something later tonight.

Actually, the cpufreq handler is installed by an initcall regardless of
which time-source is used. However as the handler changes a few TSC
specific variables, it exits in timer_tsc.c.

I think the fix I just mailed should do the trick. Let me know if it
doesn't.

thanks
-john





-------------------------------------------------------
This SF.Net email is sponsored by: IBM Linux Tutorials
Free Linux tutorial presented by Daniel Robbins, President and CEO of
GenToo technologies. Learn everything from fundamentals to system
administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click

Re: X86_PM_TIMER: /proc/cpuinfo doesn't get updated

[Dominik Brodowski]

[-- Attachment #1: Type: text/plain, Size: 987 bytes --]

On Tue, Mar 16, 2004 at 03:28:15PM -0800, john stultz wrote:
> On Tue, 2004-03-16 at 13:42, Karol Kozimor wrote:
> > Thus wrote Daniele Venzano:
> > > > I have a notebook with an Athlon-M CPU. I tried linux 2.6.4 with
> > > > CONFIG_X86_PM_TIMER=y and I noticed that /proc/cpuinfo doesn't get
> > > > updated when I switch frequency (via sysfs, using powernow-k7). The is
> > > > issue seems cosmetic only, CPU frequency changes (watching
> > > > temperature/battery life).
> > > I can confirm, I'm seeing the same behavior. Please note that the
> > > bogomips count gets updated, it's only the frequency that doesn't
> > > change.
> > 
> > Same here with a P4-M, follow-up to John and Dmitry.
> 
> Hmm. This is untested, but I think this should do the trick.
> 
> Dominik: Is there any reason I'm not seeing why cpu_khz should only be
> updated when using the TSC?

Is cpu_khz always correct (or, at least, nonzero) when we're reaching this 
code path?

	Dominik

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: X86_PM_TIMER: /proc/cpuinfo doesn't get updated

[Karol Kozimor]

Thus wrote john stultz:
> Hmm. This is untested, but I think this should do the trick.

Hmm... without the patch, neither cpu MHz nor bogomips are updated, with
the patch cpu MHz value seems correct (both using acpi.ko and
speedstep-ich.ko, but the bogomips is still at its initial value.
Best regards,

-- 
Karol 'sziwan' Kozimor
sziwan-DETuoxkZsSqrDJvtcaxF/A@public.gmane.org


-------------------------------------------------------
This SF.Net email is sponsored by: IBM Linux Tutorials
Free Linux tutorial presented by Daniel Robbins, President and CEO of
GenToo technologies. Learn everything from fundamentals to system
administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click

Re: [ACPI] X86_PM_TIMER: /proc/cpuinfo doesn't get updated

[Dmitry Torokhov]

On Tuesday 16 March 2004 06:32 pm, john stultz wrote:
> On Tue, 2004-03-16 at 15:19, Dmitry Torokhov wrote:
> > PM timer does not install CPUFREQ handler which would scale cpu_khz to
> > give proper display. I might cook up something later tonight.
> 
> Actually, the cpufreq handler is installed by an initcall regardless of
> which time-source is used. However as the handler changes a few TSC
> specific variables, it exits in timer_tsc.c.
>

Yes, you are of course right, I missed that fact, sorry.

-- 
Dmitry

Re: [ACPI] X86_PM_TIMER: /proc/cpuinfo doesn't get updated

[Dmitry Torokhov]

On Tuesday 16 March 2004 07:13 pm, Karol Kozimor wrote:
> Thus wrote john stultz:
> > Hmm. This is untested, but I think this should do the trick.
> 
> Hmm... without the patch, neither cpu MHz nor bogomips are updated, with
> the patch cpu MHz value seems correct (both using acpi.ko and
> speedstep-ich.ko, but the bogomips is still at its initial value.
> Best regards,
> 

Karol, do you have a P4? AFAIK P4's TSC is stable even if core frequence
changes so loop_per_juffy (== bogomips) need not be updated.

-- 
Dmitry

Re: X86_PM_TIMER: /proc/cpuinfo doesn't get updated

[john stultz]

On Tue, 2004-03-16 at 15:33, Dominik Brodowski wrote:
> On Tue, Mar 16, 2004 at 03:28:15PM -0800, john stultz wrote:
> > On Tue, 2004-03-16 at 13:42, Karol Kozimor wrote:
> > > Thus wrote Daniele Venzano:
> > > > > I have a notebook with an Athlon-M CPU. I tried linux 2.6.4 with
> > > > > CONFIG_X86_PM_TIMER=y and I noticed that /proc/cpuinfo doesn't get
> > > > > updated when I switch frequency (via sysfs, using powernow-k7). The is
> > > > > issue seems cosmetic only, CPU frequency changes (watching
> > > > > temperature/battery life).
> > > > I can confirm, I'm seeing the same behavior. Please note that the
> > > > bogomips count gets updated, it's only the frequency that doesn't
> > > > change.
> > > 
> > > Same here with a P4-M, follow-up to John and Dmitry.
> > 
> > Hmm. This is untested, but I think this should do the trick.
> > 
> > Dominik: Is there any reason I'm not seeing why cpu_khz should only be
> > updated when using the TSC?
> 
> Is cpu_khz always correct (or, at least, nonzero) when we're reaching this 
> code path?

Using the PIT time source, cpu_khz is zero, so maybe it should be
conditional on if(cpu_khz)?

thanks
-john



-------------------------------------------------------
This SF.Net email is sponsored by: IBM Linux Tutorials
Free Linux tutorial presented by Daniel Robbins, President and CEO of
GenToo technologies. Learn everything from fundamentals to system
administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click

Re: X86_PM_TIMER: /proc/cpuinfo doesn't get updated

[Dominik Brodowski]

[-- Attachment #1: Type: text/plain, Size: 1321 bytes --]

On Tue, Mar 16, 2004 at 04:46:54PM -0800, john stultz wrote:
> On Tue, 2004-03-16 at 15:33, Dominik Brodowski wrote:
> > On Tue, Mar 16, 2004 at 03:28:15PM -0800, john stultz wrote:
> > > On Tue, 2004-03-16 at 13:42, Karol Kozimor wrote:
> > > > Thus wrote Daniele Venzano:
> > > > > > I have a notebook with an Athlon-M CPU. I tried linux 2.6.4 with
> > > > > > CONFIG_X86_PM_TIMER=y and I noticed that /proc/cpuinfo doesn't get
> > > > > > updated when I switch frequency (via sysfs, using powernow-k7). The is
> > > > > > issue seems cosmetic only, CPU frequency changes (watching
> > > > > > temperature/battery life).
> > > > > I can confirm, I'm seeing the same behavior. Please note that the
> > > > > bogomips count gets updated, it's only the frequency that doesn't
> > > > > change.
> > > > 
> > > > Same here with a P4-M, follow-up to John and Dmitry.
> > > 
> > > Hmm. This is untested, but I think this should do the trick.
> > > 
> > > Dominik: Is there any reason I'm not seeing why cpu_khz should only be
> > > updated when using the TSC?
> > 
> > Is cpu_khz always correct (or, at least, nonzero) when we're reaching this 
> > code path?
> 
> Using the PIT time source, cpu_khz is zero, so maybe it should be
> conditional on if(cpu_khz)?

That will do the trick.

	Dominik

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

[PATCH] linux-2.6.5-rc1_cpukhz-fix_A0.patch

[john stultz]

On Wed, 2004-03-17 at 01:50, Dominik Brodowski wrote:
> On Tue, Mar 16, 2004 at 04:46:54PM -0800, john stultz wrote:
> > On Tue, 2004-03-16 at 15:33, Dominik Brodowski wrote:
> > > On Tue, Mar 16, 2004 at 03:28:15PM -0800, john stultz wrote:
> > > > On Tue, 2004-03-16 at 13:42, Karol Kozimor wrote:
> > > > > Thus wrote Daniele Venzano:
> > > > > > > I have a notebook with an Athlon-M CPU. I tried linux 2.6.4 with
> > > > > > > CONFIG_X86_PM_TIMER=y and I noticed that /proc/cpuinfo doesn't get
> > > > > > > updated when I switch frequency (via sysfs, using powernow-k7). The is
> > > > > > > issue seems cosmetic only, CPU frequency changes (watching
> > > > > > > temperature/battery life).
> > > > > > I can confirm, I'm seeing the same behavior. Please note that the
> > > > > > bogomips count gets updated, it's only the frequency that doesn't
> > > > > > change.
> > > > > 
> > > > > Same here with a P4-M, follow-up to John and Dmitry.
> > > > 
> > > > Hmm. This is untested, but I think this should do the trick.
> > > > 
> > > > Dominik: Is there any reason I'm not seeing why cpu_khz should only be
> > > > updated when using the TSC?
> > > 
> > > Is cpu_khz always correct (or, at least, nonzero) when we're reaching this 
> > > code path?
> > 
> > Using the PIT time source, cpu_khz is zero, so maybe it should be
> > conditional on if(cpu_khz)?
> 
> That will do the trick.

Patch below.

This small patch insures that cpu_khz is adjusted on cpufreq
notifications even when the tsc timesource is not in use. It fixes the
mostly cosmetic issue when using the ACPI PM timesource of /proc/cpuinfo
not being properly updated when cpu frequency was lowered. 

Please let me know if you have any additional feedback.

thanks
-john

diff -Nru a/arch/i386/kernel/timers/timer_tsc.c b/arch/i386/kernel/timers/timer_tsc.c
--- a/arch/i386/kernel/timers/timer_tsc.c	Thu Mar 18 12:50:08 2004
+++ b/arch/i386/kernel/timers/timer_tsc.c	Thu Mar 18 12:50:08 2004
@@ -360,8 +360,9 @@
 		if (variable_tsc)
 			cpu_data[freq->cpu].loops_per_jiffy = cpufreq_scale(loops_per_jiffy_ref, ref_freq, freq->new);
 #ifndef CONFIG_SMP
-		if (use_tsc) {
+		if(cpu_khz)
 			cpu_khz = cpufreq_scale(cpu_khz_ref, ref_freq, freq->new);
+		if (use_tsc) {
 			if (variable_tsc) {
 				fast_gettimeoffset_quotient = cpufreq_scale(fast_gettimeoffset_ref, freq->new, ref_freq);
 				set_cyc2ns_scale(cpu_khz/1000);