linux-pm.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* v4.8 cpufreq regression?
@ 2016-10-07 22:59 Aaro Koskinen
  2016-10-10 21:38 ` [PATCH] cpufreq: skip invalid entries when searching the frequency Aaro Koskinen
  0 siblings, 1 reply; 3+ messages in thread
From: Aaro Koskinen @ 2016-10-07 22:59 UTC (permalink / raw)
  To: Viresh Kumar, Rafael J. Wysocki, linux-pm, linux-kernel

Hi,

I'm upgrading from v4.7 to v4.8 on MIPS/loongson2_cpufreq with
cpufreq/ondemand and I noticed that /proc/cpuinfo is always reporting
(regardless of system load):

BogoMIPS                : 697150.20

	and sysfs:

/sys/devices/system/cpu/cpufreq/policy0/affected_cpus:0
/sys/devices/system/cpu/cpufreq/policy0/cpuinfo_cur_freq:797800
/sys/devices/system/cpu/cpufreq/policy0/cpuinfo_max_freq:797800
/sys/devices/system/cpu/cpufreq/policy0/cpuinfo_min_freq:199450
/sys/devices/system/cpu/cpufreq/policy0/cpuinfo_transition_latency:0
/sys/devices/system/cpu/cpufreq/policy0/related_cpus:0
/sys/devices/system/cpu/cpufreq/policy0/scaling_available_frequencies:199450 299175 398900 498625 598350 698075 797800 
/sys/devices/system/cpu/cpufreq/policy0/scaling_available_governors:ondemand performance 
/sys/devices/system/cpu/cpufreq/policy0/scaling_cur_freq:797800
/sys/devices/system/cpu/cpufreq/policy0/scaling_driver:loongson2
/sys/devices/system/cpu/cpufreq/policy0/scaling_governor:ondemand
/sys/devices/system/cpu/cpufreq/policy0/scaling_max_freq:797800
/sys/devices/system/cpu/cpufreq/policy0/scaling_min_freq:199450
/sys/devices/system/cpu/cpufreq/policy0/scaling_setspeed:<unsupported>

although previously they reported (when idle):

BogoMIPS                : 132.09

/sys/devices/system/cpu/cpufreq/policy0/affected_cpus:0
/sys/devices/system/cpu/cpufreq/policy0/cpuinfo_cur_freq:199445
/sys/devices/system/cpu/cpufreq/policy0/cpuinfo_max_freq:797780
/sys/devices/system/cpu/cpufreq/policy0/cpuinfo_min_freq:199445
/sys/devices/system/cpu/cpufreq/policy0/cpuinfo_transition_latency:0
/sys/devices/system/cpu/cpufreq/policy0/related_cpus:0
/sys/devices/system/cpu/cpufreq/policy0/scaling_available_frequencies:199445 299167 398890 498612 598335 698057 797780 
/sys/devices/system/cpu/cpufreq/policy0/scaling_available_governors:ondemand performance 
/sys/devices/system/cpu/cpufreq/policy0/scaling_cur_freq:199445
/sys/devices/system/cpu/cpufreq/policy0/scaling_driver:loongson2
/sys/devices/system/cpu/cpufreq/policy0/scaling_governor:ondemand
/sys/devices/system/cpu/cpufreq/policy0/scaling_max_freq:797780
/sys/devices/system/cpu/cpufreq/policy0/scaling_min_freq:199445
/sys/devices/system/cpu/cpufreq/policy0/scaling_setspeed:<unsupported>

and under full load:

BogoMIPS                : 528.38

/sys/devices/system/cpu/cpufreq/policy0/affected_cpus:0
/sys/devices/system/cpu/cpufreq/policy0/cpuinfo_cur_freq:797780
/sys/devices/system/cpu/cpufreq/policy0/cpuinfo_max_freq:797780
/sys/devices/system/cpu/cpufreq/policy0/cpuinfo_min_freq:199445
/sys/devices/system/cpu/cpufreq/policy0/cpuinfo_transition_latency:0
/sys/devices/system/cpu/cpufreq/policy0/related_cpus:0
/sys/devices/system/cpu/cpufreq/policy0/scaling_available_frequencies:199445 299167 398890 498612 598335 698057 797780 
/sys/devices/system/cpu/cpufreq/policy0/scaling_available_governors:ondemand performance 
/sys/devices/system/cpu/cpufreq/policy0/scaling_cur_freq:797780
/sys/devices/system/cpu/cpufreq/policy0/scaling_driver:loongson2
/sys/devices/system/cpu/cpufreq/policy0/scaling_governor:ondemand
/sys/devices/system/cpu/cpufreq/policy0/scaling_max_freq:797780
/sys/devices/system/cpu/cpufreq/policy0/scaling_min_freq:199445
/sys/devices/system/cpu/cpufreq/policy0/scaling_setspeed:<unsupported>

The offeding commit (bisected) seems to be:

commit da0c6dc00c69d0adaae99c3b3d2ea0c869a3fb35
Author: Viresh Kumar <viresh.kumar@linaro.org>
Date:   Mon Jun 27 16:04:07 2016 +0530

    cpufreq: Handle sorted frequency tables more efficiently

Any ideas?

A.

^ permalink raw reply	[flat|nested] 3+ messages in thread

* [PATCH] cpufreq: skip invalid entries when searching the frequency
  2016-10-07 22:59 v4.8 cpufreq regression? Aaro Koskinen
@ 2016-10-10 21:38 ` Aaro Koskinen
  2016-10-12  3:16   ` Viresh Kumar
  0 siblings, 1 reply; 3+ messages in thread
From: Aaro Koskinen @ 2016-10-10 21:38 UTC (permalink / raw)
  To: Viresh Kumar, Rafael J. Wysocki, linux-pm, linux-kernel; +Cc: Aaro Koskinen

Skip invalid entries when searching the frequency. This fixes cpufreq
at least on loongson2 MIPS board.

Fixes: da0c6dc00c69 ("cpufreq: Handle sorted frequency tables more efficiently")
Signed-off-by: Aaro Koskinen <aaro.koskinen@iki.fi>
---
 include/linux/cpufreq.h | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)

diff --git a/include/linux/cpufreq.h b/include/linux/cpufreq.h
index 631ba33b..0c1b188 100644
--- a/include/linux/cpufreq.h
+++ b/include/linux/cpufreq.h
@@ -645,6 +645,9 @@ static inline int cpufreq_table_find_index_al(struct cpufreq_policy *policy,
 	for (i = 0; table[i].frequency != CPUFREQ_TABLE_END; i++) {
 		freq = table[i].frequency;
 
+		if (freq == CPUFREQ_ENTRY_INVALID)
+			continue;
+
 		if (freq >= target_freq)
 			return i;
 
@@ -665,6 +668,9 @@ static inline int cpufreq_table_find_index_dl(struct cpufreq_policy *policy,
 	for (i = 0; table[i].frequency != CPUFREQ_TABLE_END; i++) {
 		freq = table[i].frequency;
 
+		if (freq == CPUFREQ_ENTRY_INVALID)
+			continue;
+
 		if (freq == target_freq)
 			return i;
 
@@ -706,6 +712,9 @@ static inline int cpufreq_table_find_index_ah(struct cpufreq_policy *policy,
 	for (i = 0; table[i].frequency != CPUFREQ_TABLE_END; i++) {
 		freq = table[i].frequency;
 
+		if (freq == CPUFREQ_ENTRY_INVALID)
+			continue;
+
 		if (freq == target_freq)
 			return i;
 
@@ -735,6 +744,9 @@ static inline int cpufreq_table_find_index_dh(struct cpufreq_policy *policy,
 	for (i = 0; table[i].frequency != CPUFREQ_TABLE_END; i++) {
 		freq = table[i].frequency;
 
+		if (freq == CPUFREQ_ENTRY_INVALID)
+			continue;
+
 		if (freq <= target_freq)
 			return i;
 
@@ -767,6 +779,9 @@ static inline int cpufreq_table_find_index_ac(struct cpufreq_policy *policy,
 	for (i = 0; table[i].frequency != CPUFREQ_TABLE_END; i++) {
 		freq = table[i].frequency;
 
+		if (freq == CPUFREQ_ENTRY_INVALID)
+			continue;
+
 		if (freq == target_freq)
 			return i;
 
@@ -800,6 +815,9 @@ static inline int cpufreq_table_find_index_dc(struct cpufreq_policy *policy,
 	for (i = 0; table[i].frequency != CPUFREQ_TABLE_END; i++) {
 		freq = table[i].frequency;
 
+		if (freq == CPUFREQ_ENTRY_INVALID)
+			continue;
+
 		if (freq == target_freq)
 			return i;
 
-- 
2.9.2


^ permalink raw reply related	[flat|nested] 3+ messages in thread

* Re: [PATCH] cpufreq: skip invalid entries when searching the frequency
  2016-10-10 21:38 ` [PATCH] cpufreq: skip invalid entries when searching the frequency Aaro Koskinen
@ 2016-10-12  3:16   ` Viresh Kumar
  0 siblings, 0 replies; 3+ messages in thread
From: Viresh Kumar @ 2016-10-12  3:16 UTC (permalink / raw)
  To: Aaro Koskinen; +Cc: Rafael J. Wysocki, linux-pm, linux-kernel

On 11-10-16, 00:38, Aaro Koskinen wrote:
> Skip invalid entries when searching the frequency. This fixes cpufreq
> at least on loongson2 MIPS board.
> 
> Fixes: da0c6dc00c69 ("cpufreq: Handle sorted frequency tables more efficiently")
> Signed-off-by: Aaro Koskinen <aaro.koskinen@iki.fi>
> ---
>  include/linux/cpufreq.h | 18 ++++++++++++++++++
>  1 file changed, 18 insertions(+)

Hi Aaro,

I have sent a new version of this patch, but screwed up the
--in-reply-to somehow.. Can you please give that a try?

-- 
viresh

^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2016-10-12  3:16 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2016-10-07 22:59 v4.8 cpufreq regression? Aaro Koskinen
2016-10-10 21:38 ` [PATCH] cpufreq: skip invalid entries when searching the frequency Aaro Koskinen
2016-10-12  3:16   ` Viresh Kumar

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).