Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > linux.kernel > #1426440 > unrolled thread

[PATCH v4 0/3] cpuacct: Introduce cpuacct.usage_all to show all cpu stats together(was: [PATCH v2 0/3] cpuacct: Show all detail stats in one file)

Started byZhao Lei <zhaolei@cn.fujitsu.com>
First post2016-06-20 11:50 +0200
Last post2016-06-20 11:50 +0200
Articles 3 — 1 participant

Back to article view | Back to linux.kernel


Contents

  [PATCH v4 0/3] cpuacct: Introduce cpuacct.usage_all to show all cpu stats together(was: [PATCH v2 0/3] cpuacct: Show all detail stats in one file) Zhao Lei <zhaolei@cn.fujitsu.com> - 2016-06-20 11:50 +0200
    [PATCH v4 3/3] cpuacct: Introduce cpuacct.usage_all to show all cpu stats together Zhao Lei <zhaolei@cn.fujitsu.com> - 2016-06-20 11:50 +0200
    [PATCH v4 2/3] cpuacct: Use loop to avoid copies of the similar code in cpuacct_stats_show() Zhao Lei <zhaolei@cn.fujitsu.com> - 2016-06-20 11:50 +0200

#1426440 — [PATCH v4 0/3] cpuacct: Introduce cpuacct.usage_all to show all cpu stats together(was: [PATCH v2 0/3] cpuacct: Show all detail stats in one file)

FromZhao Lei <zhaolei@cn.fujitsu.com>
Date2016-06-20 11:50 +0200
Subject[PATCH v4 0/3] cpuacct: Introduce cpuacct.usage_all to show all cpu stats together(was: [PATCH v2 0/3] cpuacct: Show all detail stats in one file)
Message-ID<rM40a-6wa-9@gated-at.bofh.it>
In current code, we can get cpuacct data from several files,
but each file have its limit.

For example:
we can get cpu usage in user and kernel mode by cpuacct.stat,
but we can't get detail data of each cpu in above file.
we can get each cpu's kernel mode usage in cpuacct.usage_percpu_sys,
but we can't get user mode data in the same time.

This patch introduce cpuacct.usage_all, to show all detailed
cpu accounting data altogether:
 # cat cpuacct.usage_all
 cpu user system
 0 3809760299 5807968992
 1 3250329855 454612211
 ..

Any aspects of statistics data can be get from this file
with a simple data-processing code.

Changelog v3->v4:
 1: Rebase on top of 4.7-rc4

Changelog v2->v3:
 1: Change title of PATCH 3/3 from:
    cpuacct: Show all detail stats in one file
    to
    cpuacct: Introduce cpuacct.usage_all to show all cpu stats together
 2: Fix spelling typos in patch's description.
 Suggested by: Ingo Molnar <mingo.kernel.org@gmail.com>

Changelog v1->v2:
 1: Rewrite subject in PATCH 1/3 and 2/3 to fix typos and
    make it more accurate.
 2: Use a variable for the cpustat array itself to make
    code clean(avoid lot of 'kcpustat->').
 3: Add missed newline between variable definitions and
    the first non-definition C statement.
 Above are suggested by: Ingo Molnar <mingo.kernel.org@gmail.com>
 4: Remove '[]' in CPUID, to make it easier to parse.
 Suggested by: Peter Zijlstra <peterz@infradead.org>

Zhao Lei (3):
  cpuacct: Merge cpuacct_usage_index and cpuacct_stat_index
  cpuacct: Use loop to avoid copies of the similar code in
    cpuacct_stats_show()
  cpuacct: Introduce cpuacct.usage_all to show all cpu stats together

 kernel/sched/cpuacct.c | 114 +++++++++++++++++++++++++++++++------------------
 1 file changed, 73 insertions(+), 41 deletions(-)

-- 
1.8.5.1

[toc] | [next] | [standalone]


#1426442 — [PATCH v4 3/3] cpuacct: Introduce cpuacct.usage_all to show all cpu stats together

FromZhao Lei <zhaolei@cn.fujitsu.com>
Date2016-06-20 11:50 +0200
Subject[PATCH v4 3/3] cpuacct: Introduce cpuacct.usage_all to show all cpu stats together
Message-ID<rM40a-6wa-27@gated-at.bofh.it>
In reply to#1426440
In current code, we can get cpuacct data from several files,
but each file have its limit.

For example:
we can get cpu usage in user and kernel mode by cpuacct.stat,
but we can't get detail data of each cpu in above file.
we can get each cpu's kernel mode usage in cpuacct.usage_percpu_sys,
but we can't get user mode data in the same time.

This patch introduce cpuacct.usage_all, to show all detailed
cpu accounting data altogether:
 # cat cpuacct.usage_all
 cpu user system
 0 3809760299 5807968992
 1 3250329855 454612211
 ..

Any aspects of statistics data can be get from this file
with a simple data-processing code.

Signed-off-by: Zhao Lei <zhaolei@cn.fujitsu.com>
---
 kernel/sched/cpuacct.c | 40 ++++++++++++++++++++++++++++++++++++++++
 1 file changed, 40 insertions(+)

diff --git a/kernel/sched/cpuacct.c b/kernel/sched/cpuacct.c
index 677cd1a..bc0b309c 100644
--- a/kernel/sched/cpuacct.c
+++ b/kernel/sched/cpuacct.c
@@ -240,6 +240,42 @@ static int cpuacct_percpu_seq_show(struct seq_file *m, void *V)
 	return __cpuacct_percpu_seq_show(m, CPUACCT_STAT_NSTATS);
 }
 
+static int cpuacct_all_seq_show(struct seq_file *m, void *V)
+{
+	struct cpuacct *ca = css_ca(seq_css(m));
+	int index;
+	int cpu;
+
+	seq_puts(m, "cpu");
+	for (index = 0; index < CPUACCT_STAT_NSTATS; index++)
+		seq_printf(m, " %s", cpuacct_stat_desc[index]);
+	seq_puts(m, "\n");
+
+	for_each_possible_cpu(cpu) {
+		struct cpuacct_usage *cpuusage = per_cpu_ptr(ca->cpuusage, cpu);
+
+		seq_printf(m, "%d", cpu);
+
+		for (index = 0; index < CPUACCT_STAT_NSTATS; index++) {
+#ifndef CONFIG_64BIT
+			/*
+			 * Take rq->lock to make 64-bit read safe on 32-bit
+			 * platforms.
+			 */
+			raw_spin_lock_irq(&cpu_rq(cpu)->lock);
+#endif
+
+			seq_printf(m, " %llu", cpuusage->usages[index]);
+
+#ifndef CONFIG_64BIT
+			raw_spin_unlock_irq(&cpu_rq(cpu)->lock);
+#endif
+		}
+		seq_puts(m, "\n");
+	}
+	return 0;
+}
+
 static int cpuacct_stats_show(struct seq_file *sf, void *v)
 {
 	struct cpuacct *ca = css_ca(seq_css(sf));
@@ -294,6 +330,10 @@ static struct cftype files[] = {
 		.seq_show = cpuacct_percpu_sys_seq_show,
 	},
 	{
+		.name = "usage_all",
+		.seq_show = cpuacct_all_seq_show,
+	},
+	{
 		.name = "stat",
 		.seq_show = cpuacct_stats_show,
 	},
-- 
1.8.5.1

[toc] | [prev] | [next] | [standalone]


#1426443 — [PATCH v4 2/3] cpuacct: Use loop to avoid copies of the similar code in cpuacct_stats_show()

FromZhao Lei <zhaolei@cn.fujitsu.com>
Date2016-06-20 11:50 +0200
Subject[PATCH v4 2/3] cpuacct: Use loop to avoid copies of the similar code in cpuacct_stats_show()
Message-ID<rM40a-6wa-29@gated-at.bofh.it>
In reply to#1426440
Currently we have copies of the similar code for each cpustat(system/user)
in cpuacct_stats_show(), this patch use loop instead.
Only a little cleanup.

Signed-off-by: Zhao Lei <zhaolei@cn.fujitsu.com>
---
 kernel/sched/cpuacct.c | 29 ++++++++++++++---------------
 1 file changed, 14 insertions(+), 15 deletions(-)

diff --git a/kernel/sched/cpuacct.c b/kernel/sched/cpuacct.c
index 74241eb..677cd1a 100644
--- a/kernel/sched/cpuacct.c
+++ b/kernel/sched/cpuacct.c
@@ -243,27 +243,26 @@ static int cpuacct_percpu_seq_show(struct seq_file *m, void *V)
 static int cpuacct_stats_show(struct seq_file *sf, void *v)
 {
 	struct cpuacct *ca = css_ca(seq_css(sf));
+	s64 val[CPUACCT_STAT_NSTATS];
 	int cpu;
-	s64 val = 0;
+	int stat;
 
+	memset(val, 0, sizeof(val));
 	for_each_possible_cpu(cpu) {
-		struct kernel_cpustat *kcpustat = per_cpu_ptr(ca->cpustat, cpu);
-		val += kcpustat->cpustat[CPUTIME_USER];
-		val += kcpustat->cpustat[CPUTIME_NICE];
-	}
-	val = cputime64_to_clock_t(val);
-	seq_printf(sf, "%s %lld\n", cpuacct_stat_desc[CPUACCT_STAT_USER], val);
+		u64 *cpustat = per_cpu_ptr(ca->cpustat, cpu)->cpustat;
 
-	val = 0;
-	for_each_possible_cpu(cpu) {
-		struct kernel_cpustat *kcpustat = per_cpu_ptr(ca->cpustat, cpu);
-		val += kcpustat->cpustat[CPUTIME_SYSTEM];
-		val += kcpustat->cpustat[CPUTIME_IRQ];
-		val += kcpustat->cpustat[CPUTIME_SOFTIRQ];
+		val[CPUACCT_STAT_USER]   += cpustat[CPUTIME_USER];
+		val[CPUACCT_STAT_USER]   += cpustat[CPUTIME_NICE];
+		val[CPUACCT_STAT_SYSTEM] += cpustat[CPUTIME_SYSTEM];
+		val[CPUACCT_STAT_SYSTEM] += cpustat[CPUTIME_IRQ];
+		val[CPUACCT_STAT_SYSTEM] += cpustat[CPUTIME_SOFTIRQ];
 	}
 
-	val = cputime64_to_clock_t(val);
-	seq_printf(sf, "%s %lld\n", cpuacct_stat_desc[CPUACCT_STAT_SYSTEM], val);
+	for (stat = 0; stat < CPUACCT_STAT_NSTATS; stat++) {
+		seq_printf(sf, "%s %lld\n",
+			   cpuacct_stat_desc[stat],
+			   cputime64_to_clock_t(val[stat]));
+	}
 
 	return 0;
 }
-- 
1.8.5.1

[toc] | [prev] | [standalone]


Back to top | Article view | linux.kernel


csiph-web