[PATCH] x86/cpu: disable frequency requests via aperfmperf IPI for nohz_full CPUs

[PATCH] x86/cpu: disable frequency requests via aperfmperf IPI for nohz_full CPUs

[Konstantin Khlebnikov]

Since commit 7d5905dc14a8 ("x86 / CPU: Always show current CPU frequency
in /proc/cpuinfo") open and read of /proc/cpuinfo sends IPI to all CPUs.
Many applications read /proc/cpuinfo at the start for trivial reasons like
counting cores or detecting cpu features. While sensitive workloads like
DPDK network polling don't like any interrupts.

This patch integrates this feature with cpu isolation and does not send
IPI to cpus without housekeeping flag HK_FLAG_MISC (set by nohz_full).

Code that requests cpu frequency like show_cpuinfo() falls back to last
frequency set by cpufreq driver if direct method returns 0.

Signed-off-by: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Link: http://lkml.kernel.org/r/155618921176.2006.3031084313219003524.stgit@buzz
---
 arch/x86/kernel/cpu/aperfmperf.c |   12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)

diff --git a/arch/x86/kernel/cpu/aperfmperf.c b/arch/x86/kernel/cpu/aperfmperf.c
index 64d5aec24203..0be5ef97fd8f 100644
--- a/arch/x86/kernel/cpu/aperfmperf.c
+++ b/arch/x86/kernel/cpu/aperfmperf.c
@@ -14,6 +14,7 @@
 #include <linux/percpu.h>
 #include <linux/cpufreq.h>
 #include <linux/smp.h>
+#include <linux/sched/isolation.h>
 
 #include "cpu.h"
 
@@ -86,6 +87,9 @@ unsigned int aperfmperf_get_khz(int cpu)
 	if (!boot_cpu_has(X86_FEATURE_APERFMPERF))
 		return 0;
 
+	if (!housekeeping_cpu(cpu, HK_FLAG_MISC))
+		return 0;
+
 	aperfmperf_snapshot_cpu(cpu, ktime_get(), true);
 	return per_cpu(samples.khz, cpu);
 }
@@ -102,9 +106,12 @@ void arch_freq_prepare_all(void)
 	if (!boot_cpu_has(X86_FEATURE_APERFMPERF))
 		return;
 
-	for_each_online_cpu(cpu)
+	for_each_online_cpu(cpu) {
+		if (!housekeeping_cpu(cpu, HK_FLAG_MISC))
+			continue;
 		if (!aperfmperf_snapshot_cpu(cpu, now, false))
 			wait = true;
+	}
 
 	if (wait)
 		msleep(APERFMPERF_REFRESH_DELAY_MS);
@@ -118,6 +125,9 @@ unsigned int arch_freq_get_on_cpu(int cpu)
 	if (!boot_cpu_has(X86_FEATURE_APERFMPERF))
 		return 0;
 
+	if (!housekeeping_cpu(cpu, HK_FLAG_MISC))
+		return 0;
+
 	if (aperfmperf_snapshot_cpu(cpu, ktime_get(), true))
 		return per_cpu(samples.khz, cpu);
 

[tip:x86/cpu] x86/cpu: Disable frequency requests via aperfmperf IPI for nohz_full CPUs

[tip-bot for Konstantin Khlebnikov]

Commit-ID:  cc9e303c91f5c25c49a4312552841f4c23fa2b69
Gitweb:     https://git.kernel.org/tip/cc9e303c91f5c25c49a4312552841f4c23fa2b69
Author:     Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
AuthorDate: Wed, 15 May 2019 09:59:00 +0300
Committer:  Thomas Gleixner <tglx@linutronix.de>
CommitDate: Sat, 22 Jun 2019 17:23:48 +0200

x86/cpu: Disable frequency requests via aperfmperf IPI for nohz_full CPUs

Since commit 7d5905dc14a8 ("x86 / CPU: Always show current CPU frequency
in /proc/cpuinfo") open and read of /proc/cpuinfo sends IPI to all CPUs.
Many applications read /proc/cpuinfo at the start for trivial reasons like
counting cores or detecting cpu features. While sensitive workloads like
DPDK network polling don't like any interrupts.

Integrates this feature with cpu isolation and do not send IPIs to CPUs
without housekeeping flag HK_FLAG_MISC (set by nohz_full).

Code that requests cpu frequency like show_cpuinfo() falls back to the last
frequency set by the cpufreq driver if this method returns 0.

Signed-off-by: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Len Brown <len.brown@intel.com>
Cc: Frederic Weisbecker <frederic@kernel.org>
Cc: "Rafael J. Wysocki" <rafael.j.wysocki@intel.com>
Cc: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
Link: https://lkml.kernel.org/r/155790354043.1104.15333317408370209.stgit@buzz

---
 arch/x86/kernel/cpu/aperfmperf.c | 12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)

diff --git a/arch/x86/kernel/cpu/aperfmperf.c b/arch/x86/kernel/cpu/aperfmperf.c
index e71a6ff8a67e..e2f319dc992d 100644
--- a/arch/x86/kernel/cpu/aperfmperf.c
+++ b/arch/x86/kernel/cpu/aperfmperf.c
@@ -13,6 +13,7 @@
 #include <linux/percpu.h>
 #include <linux/cpufreq.h>
 #include <linux/smp.h>
+#include <linux/sched/isolation.h>
 
 #include "cpu.h"
 
@@ -85,6 +86,9 @@ unsigned int aperfmperf_get_khz(int cpu)
 	if (!boot_cpu_has(X86_FEATURE_APERFMPERF))
 		return 0;
 
+	if (!housekeeping_cpu(cpu, HK_FLAG_MISC))
+		return 0;
+
 	aperfmperf_snapshot_cpu(cpu, ktime_get(), true);
 	return per_cpu(samples.khz, cpu);
 }
@@ -101,9 +105,12 @@ void arch_freq_prepare_all(void)
 	if (!boot_cpu_has(X86_FEATURE_APERFMPERF))
 		return;
 
-	for_each_online_cpu(cpu)
+	for_each_online_cpu(cpu) {
+		if (!housekeeping_cpu(cpu, HK_FLAG_MISC))
+			continue;
 		if (!aperfmperf_snapshot_cpu(cpu, now, false))
 			wait = true;
+	}
 
 	if (wait)
 		msleep(APERFMPERF_REFRESH_DELAY_MS);
@@ -117,6 +124,9 @@ unsigned int arch_freq_get_on_cpu(int cpu)
 	if (!boot_cpu_has(X86_FEATURE_APERFMPERF))
 		return 0;
 
+	if (!housekeeping_cpu(cpu, HK_FLAG_MISC))
+		return 0;
+
 	if (aperfmperf_snapshot_cpu(cpu, ktime_get(), true))
 		return per_cpu(samples.khz, cpu);