[PATCH v3 1/4] cputime, cpuacct: Include guest time in user time in cpuacct.stat

[PATCH v3 1/4] cputime, cpuacct: Include guest time in user time in cpuacct.stat

[Andrey Ryabinin]

cpuacct.stat in no-root cgroups shows user time without guest time
included int it. This doesn't match with user time shown in root
cpuacct.stat and /proc/<pid>/stat. This also affects cgroup2's cpu.stat
in the same way.

Make account_guest_time() to add user time to cgroup's cpustat to
fix this.

Fixes: ef12fefabf94 ("cpuacct: add per-cgroup utime/stime statistics")
Signed-off-by: Andrey Ryabinin <arbn@yandex-team.com>
Reviewed-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Acked-by: Tejun Heo <tj@kernel.org>
Cc: <stable@vger.kernel.org>
---
 kernel/sched/cputime.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/kernel/sched/cputime.c b/kernel/sched/cputime.c
index 872e481d5098..042a6dbce8f3 100644
--- a/kernel/sched/cputime.c
+++ b/kernel/sched/cputime.c
@@ -148,10 +148,10 @@ void account_guest_time(struct task_struct *p, u64 cputime)
 
 	/* Add guest time to cpustat. */
 	if (task_nice(p) > 0) {
-		cpustat[CPUTIME_NICE] += cputime;
+		task_group_account_field(p, CPUTIME_NICE, cputime);
 		cpustat[CPUTIME_GUEST_NICE] += cputime;
 	} else {
-		cpustat[CPUTIME_USER] += cputime;
+		task_group_account_field(p, CPUTIME_USER, cputime);
 		cpustat[CPUTIME_GUEST] += cputime;
 	}
 }
-- 
2.32.0

[PATCH v3 3/4] sched/cpuacct: fix user/system in shown cpuacct.usage*

[Andrey Ryabinin]

cpuacct has 2 different ways of accounting and showing user
and system times.

The first one uses cpuacct_account_field() to account times
and cpuacct.stat file to expose them. And this one seems to work ok.

The second one is uses cpuacct_charge() function for accounting and
set of cpuacct.usage* files to show times. Despite some attempts to
fix it in the past it still doesn't work. Sometimes while running KVM
guest the cpuacct_charge() accounts most of the guest time as
system time. This doesn't match with user&system times shown in
cpuacct.stat or proc/<pid>/stat.

Demonstration:
 # git clone https://github.com/aryabinin/kvmsample
 # make
 # mkdir /sys/fs/cgroup/cpuacct/test
 # echo $$ > /sys/fs/cgroup/cpuacct/test/tasks
 # ./kvmsample &
 # for i in {1..5}; do cat /sys/fs/cgroup/cpuacct/test/cpuacct.usage_sys; sleep 1; done
 1976535645
 2979839428
 3979832704
 4983603153
 5983604157

Use cpustats accounted in cpuacct_account_field() as the source
of user/sys times for cpuacct.usage* files. Make cpuacct_charge()
to account only summary execution time.

Fixes: d740037fac70 ("sched/cpuacct: Split usage accounting into user_usage and sys_usage")
Signed-off-by: Andrey Ryabinin <arbn@yandex-team.com>
Cc: <stable@vger.kernel.org>
Reviewed-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Acked-by: Tejun Heo <tj@kernel.org>
---
 kernel/sched/cpuacct.c | 79 +++++++++++++++++-------------------------
 1 file changed, 32 insertions(+), 47 deletions(-)

diff --git a/kernel/sched/cpuacct.c b/kernel/sched/cpuacct.c
index f347cf9e4634..9de7dd51beb0 100644
--- a/kernel/sched/cpuacct.c
+++ b/kernel/sched/cpuacct.c
@@ -21,15 +21,11 @@ static const char * const cpuacct_stat_desc[] = {
 	[CPUACCT_STAT_SYSTEM] = "system",
 };
 
-struct cpuacct_usage {
-	u64	usages[CPUACCT_STAT_NSTATS];
-};
-
 /* track CPU usage of a group of tasks and its child groups */
 struct cpuacct {
 	struct cgroup_subsys_state	css;
 	/* cpuusage holds pointer to a u64-type object on every CPU */
-	struct cpuacct_usage __percpu	*cpuusage;
+	u64 __percpu	*cpuusage;
 	struct kernel_cpustat __percpu	*cpustat;
 };
 
@@ -49,7 +45,7 @@ static inline struct cpuacct *parent_ca(struct cpuacct *ca)
 	return css_ca(ca->css.parent);
 }
 
-static DEFINE_PER_CPU(struct cpuacct_usage, root_cpuacct_cpuusage);
+static DEFINE_PER_CPU(u64, root_cpuacct_cpuusage);
 static struct cpuacct root_cpuacct = {
 	.cpustat	= &kernel_cpustat,
 	.cpuusage	= &root_cpuacct_cpuusage,
@@ -68,7 +64,7 @@ cpuacct_css_alloc(struct cgroup_subsys_state *parent_css)
 	if (!ca)
 		goto out;
 
-	ca->cpuusage = alloc_percpu(struct cpuacct_usage);
+	ca->cpuusage = alloc_percpu(u64);
 	if (!ca->cpuusage)
 		goto out_free_ca;
 
@@ -99,7 +95,8 @@ static void cpuacct_css_free(struct cgroup_subsys_state *css)
 static u64 cpuacct_cpuusage_read(struct cpuacct *ca, int cpu,
 				 enum cpuacct_stat_index index)
 {
-	struct cpuacct_usage *cpuusage = per_cpu_ptr(ca->cpuusage, cpu);
+	u64 *cpuusage = per_cpu_ptr(ca->cpuusage, cpu);
+	u64 *cpustat = per_cpu_ptr(ca->cpustat, cpu)->cpustat;
 	u64 data;
 
 	/*
@@ -116,14 +113,17 @@ static u64 cpuacct_cpuusage_read(struct cpuacct *ca, int cpu,
 	raw_spin_rq_lock_irq(cpu_rq(cpu));
 #endif
 
-	if (index == CPUACCT_STAT_NSTATS) {
-		int i = 0;
-
-		data = 0;
-		for (i = 0; i < CPUACCT_STAT_NSTATS; i++)
-			data += cpuusage->usages[i];
-	} else {
-		data = cpuusage->usages[index];
+	switch (index) {
+	case CPUACCT_STAT_USER:
+		data = cpustat[CPUTIME_USER] + cpustat[CPUTIME_NICE];
+		break;
+	case CPUACCT_STAT_SYSTEM:
+		data = cpustat[CPUTIME_SYSTEM] + cpustat[CPUTIME_IRQ] +
+			cpustat[CPUTIME_SOFTIRQ];
+		break;
+	case CPUACCT_STAT_NSTATS:
+		data = *cpuusage;
+		break;
 	}
 
 #ifndef CONFIG_64BIT
@@ -133,10 +133,14 @@ static u64 cpuacct_cpuusage_read(struct cpuacct *ca, int cpu,
 	return data;
 }
 
-static void cpuacct_cpuusage_write(struct cpuacct *ca, int cpu, u64 val)
+static void cpuacct_cpuusage_write(struct cpuacct *ca, int cpu)
 {
-	struct cpuacct_usage *cpuusage = per_cpu_ptr(ca->cpuusage, cpu);
-	int i;
+	u64 *cpuusage = per_cpu_ptr(ca->cpuusage, cpu);
+	u64 *cpustat = per_cpu_ptr(ca->cpustat, cpu)->cpustat;
+
+	/* Don't allow to reset global kernel_cpustat */
+	if (ca == &root_cpuacct)
+		return;
 
 #ifndef CONFIG_64BIT
 	/*
@@ -144,9 +148,10 @@ static void cpuacct_cpuusage_write(struct cpuacct *ca, int cpu, u64 val)
 	 */
 	raw_spin_rq_lock_irq(cpu_rq(cpu));
 #endif
-
-	for (i = 0; i < CPUACCT_STAT_NSTATS; i++)
-		cpuusage->usages[i] = val;
+	*cpuusage = 0;
+	cpustat[CPUTIME_USER] = cpustat[CPUTIME_NICE] = 0;
+	cpustat[CPUTIME_SYSTEM] = cpustat[CPUTIME_IRQ] = 0;
+	cpustat[CPUTIME_SOFTIRQ] = 0;
 
 #ifndef CONFIG_64BIT
 	raw_spin_rq_unlock_irq(cpu_rq(cpu));
@@ -197,7 +202,7 @@ static int cpuusage_write(struct cgroup_subsys_state *css, struct cftype *cft,
 		return -EINVAL;
 
 	for_each_possible_cpu(cpu)
-		cpuacct_cpuusage_write(ca, cpu, 0);
+		cpuacct_cpuusage_write(ca, cpu);
 
 	return 0;
 }
@@ -244,25 +249,10 @@ static int cpuacct_all_seq_show(struct seq_file *m, void *V)
 	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_rq_lock_irq(cpu_rq(cpu));
-#endif
-
-			seq_printf(m, " %llu", cpuusage->usages[index]);
-
-#ifndef CONFIG_64BIT
-			raw_spin_rq_unlock_irq(cpu_rq(cpu));
-#endif
-		}
+		for (index = 0; index < CPUACCT_STAT_NSTATS; index++)
+			seq_printf(m, " %llu",
+				   cpuacct_cpuusage_read(ca, cpu, index));
 		seq_puts(m, "\n");
 	}
 	return 0;
@@ -340,16 +330,11 @@ static struct cftype files[] = {
 void cpuacct_charge(struct task_struct *tsk, u64 cputime)
 {
 	struct cpuacct *ca;
-	int index = CPUACCT_STAT_SYSTEM;
-	struct pt_regs *regs = get_irq_regs() ? : task_pt_regs(tsk);
-
-	if (regs && user_mode(regs))
-		index = CPUACCT_STAT_USER;
 
 	rcu_read_lock();
 
 	for (ca = task_ca(tsk); ca; ca = parent_ca(ca))
-		__this_cpu_add(ca->cpuusage->usages[index], cputime);
+		__this_cpu_add(*ca->cpuusage, cputime);
 
 	rcu_read_unlock();
 }
-- 
2.32.0

[tip: sched/core] sched/cpuacct: Fix user/system in shown cpuacct.usage*

[tip-bot2 for Andrey Ryabinin]

The following commit has been merged into the sched/core branch of tip:

Commit-ID:     dd02d4234c9a2214a81c57a16484304a1a51872a
Gitweb:        https://git.kernel.org/tip/dd02d4234c9a2214a81c57a16484304a1a51872a
Author:        Andrey Ryabinin <arbn@yandex-team.com>
AuthorDate:    Mon, 15 Nov 2021 19:46:06 +03:00
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Tue, 23 Nov 2021 09:55:22 +01:00

sched/cpuacct: Fix user/system in shown cpuacct.usage*

cpuacct has 2 different ways of accounting and showing user
and system times.

The first one uses cpuacct_account_field() to account times
and cpuacct.stat file to expose them. And this one seems to work ok.

The second one is uses cpuacct_charge() function for accounting and
set of cpuacct.usage* files to show times. Despite some attempts to
fix it in the past it still doesn't work. Sometimes while running KVM
guest the cpuacct_charge() accounts most of the guest time as
system time. This doesn't match with user&system times shown in
cpuacct.stat or proc/<pid>/stat.

Demonstration:
 # git clone https://github.com/aryabinin/kvmsample
 # make
 # mkdir /sys/fs/cgroup/cpuacct/test
 # echo $$ > /sys/fs/cgroup/cpuacct/test/tasks
 # ./kvmsample &
 # for i in {1..5}; do cat /sys/fs/cgroup/cpuacct/test/cpuacct.usage_sys; sleep 1; done
 1976535645
 2979839428
 3979832704
 4983603153
 5983604157

Use cpustats accounted in cpuacct_account_field() as the source
of user/sys times for cpuacct.usage* files. Make cpuacct_charge()
to account only summary execution time.

Fixes: d740037fac70 ("sched/cpuacct: Split usage accounting into user_usage and sys_usage")
Signed-off-by: Andrey Ryabinin <arbn@yandex-team.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Acked-by: Tejun Heo <tj@kernel.org>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20211115164607.23784-3-arbn@yandex-team.com
---
 kernel/sched/cpuacct.c | 79 ++++++++++++++++-------------------------
 1 file changed, 32 insertions(+), 47 deletions(-)

diff --git a/kernel/sched/cpuacct.c b/kernel/sched/cpuacct.c
index f347cf9..9de7dd5 100644
--- a/kernel/sched/cpuacct.c
+++ b/kernel/sched/cpuacct.c
@@ -21,15 +21,11 @@ static const char * const cpuacct_stat_desc[] = {
 	[CPUACCT_STAT_SYSTEM] = "system",
 };
 
-struct cpuacct_usage {
-	u64	usages[CPUACCT_STAT_NSTATS];
-};
-
 /* track CPU usage of a group of tasks and its child groups */
 struct cpuacct {
 	struct cgroup_subsys_state	css;
 	/* cpuusage holds pointer to a u64-type object on every CPU */
-	struct cpuacct_usage __percpu	*cpuusage;
+	u64 __percpu	*cpuusage;
 	struct kernel_cpustat __percpu	*cpustat;
 };
 
@@ -49,7 +45,7 @@ static inline struct cpuacct *parent_ca(struct cpuacct *ca)
 	return css_ca(ca->css.parent);
 }
 
-static DEFINE_PER_CPU(struct cpuacct_usage, root_cpuacct_cpuusage);
+static DEFINE_PER_CPU(u64, root_cpuacct_cpuusage);
 static struct cpuacct root_cpuacct = {
 	.cpustat	= &kernel_cpustat,
 	.cpuusage	= &root_cpuacct_cpuusage,
@@ -68,7 +64,7 @@ cpuacct_css_alloc(struct cgroup_subsys_state *parent_css)
 	if (!ca)
 		goto out;
 
-	ca->cpuusage = alloc_percpu(struct cpuacct_usage);
+	ca->cpuusage = alloc_percpu(u64);
 	if (!ca->cpuusage)
 		goto out_free_ca;
 
@@ -99,7 +95,8 @@ static void cpuacct_css_free(struct cgroup_subsys_state *css)
 static u64 cpuacct_cpuusage_read(struct cpuacct *ca, int cpu,
 				 enum cpuacct_stat_index index)
 {
-	struct cpuacct_usage *cpuusage = per_cpu_ptr(ca->cpuusage, cpu);
+	u64 *cpuusage = per_cpu_ptr(ca->cpuusage, cpu);
+	u64 *cpustat = per_cpu_ptr(ca->cpustat, cpu)->cpustat;
 	u64 data;
 
 	/*
@@ -116,14 +113,17 @@ static u64 cpuacct_cpuusage_read(struct cpuacct *ca, int cpu,
 	raw_spin_rq_lock_irq(cpu_rq(cpu));
 #endif
 
-	if (index == CPUACCT_STAT_NSTATS) {
-		int i = 0;
-
-		data = 0;
-		for (i = 0; i < CPUACCT_STAT_NSTATS; i++)
-			data += cpuusage->usages[i];
-	} else {
-		data = cpuusage->usages[index];
+	switch (index) {
+	case CPUACCT_STAT_USER:
+		data = cpustat[CPUTIME_USER] + cpustat[CPUTIME_NICE];
+		break;
+	case CPUACCT_STAT_SYSTEM:
+		data = cpustat[CPUTIME_SYSTEM] + cpustat[CPUTIME_IRQ] +
+			cpustat[CPUTIME_SOFTIRQ];
+		break;
+	case CPUACCT_STAT_NSTATS:
+		data = *cpuusage;
+		break;
 	}
 
 #ifndef CONFIG_64BIT
@@ -133,10 +133,14 @@ static u64 cpuacct_cpuusage_read(struct cpuacct *ca, int cpu,
 	return data;
 }
 
-static void cpuacct_cpuusage_write(struct cpuacct *ca, int cpu, u64 val)
+static void cpuacct_cpuusage_write(struct cpuacct *ca, int cpu)
 {
-	struct cpuacct_usage *cpuusage = per_cpu_ptr(ca->cpuusage, cpu);
-	int i;
+	u64 *cpuusage = per_cpu_ptr(ca->cpuusage, cpu);
+	u64 *cpustat = per_cpu_ptr(ca->cpustat, cpu)->cpustat;
+
+	/* Don't allow to reset global kernel_cpustat */
+	if (ca == &root_cpuacct)
+		return;
 
 #ifndef CONFIG_64BIT
 	/*
@@ -144,9 +148,10 @@ static void cpuacct_cpuusage_write(struct cpuacct *ca, int cpu, u64 val)
 	 */
 	raw_spin_rq_lock_irq(cpu_rq(cpu));
 #endif
-
-	for (i = 0; i < CPUACCT_STAT_NSTATS; i++)
-		cpuusage->usages[i] = val;
+	*cpuusage = 0;
+	cpustat[CPUTIME_USER] = cpustat[CPUTIME_NICE] = 0;
+	cpustat[CPUTIME_SYSTEM] = cpustat[CPUTIME_IRQ] = 0;
+	cpustat[CPUTIME_SOFTIRQ] = 0;
 
 #ifndef CONFIG_64BIT
 	raw_spin_rq_unlock_irq(cpu_rq(cpu));
@@ -197,7 +202,7 @@ static int cpuusage_write(struct cgroup_subsys_state *css, struct cftype *cft,
 		return -EINVAL;
 
 	for_each_possible_cpu(cpu)
-		cpuacct_cpuusage_write(ca, cpu, 0);
+		cpuacct_cpuusage_write(ca, cpu);
 
 	return 0;
 }
@@ -244,25 +249,10 @@ static int cpuacct_all_seq_show(struct seq_file *m, void *V)
 	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_rq_lock_irq(cpu_rq(cpu));
-#endif
-
-			seq_printf(m, " %llu", cpuusage->usages[index]);
-
-#ifndef CONFIG_64BIT
-			raw_spin_rq_unlock_irq(cpu_rq(cpu));
-#endif
-		}
+		for (index = 0; index < CPUACCT_STAT_NSTATS; index++)
+			seq_printf(m, " %llu",
+				   cpuacct_cpuusage_read(ca, cpu, index));
 		seq_puts(m, "\n");
 	}
 	return 0;
@@ -340,16 +330,11 @@ static struct cftype files[] = {
 void cpuacct_charge(struct task_struct *tsk, u64 cputime)
 {
 	struct cpuacct *ca;
-	int index = CPUACCT_STAT_SYSTEM;
-	struct pt_regs *regs = get_irq_regs() ? : task_pt_regs(tsk);
-
-	if (regs && user_mode(regs))
-		index = CPUACCT_STAT_USER;
 
 	rcu_read_lock();
 
 	for (ca = task_ca(tsk); ca; ca = parent_ca(ca))
-		__this_cpu_add(ca->cpuusage->usages[index], cputime);
+		__this_cpu_add(*ca->cpuusage, cputime);
 
 	rcu_read_unlock();
 }

Re: [PATCH v3 1/4] cputime, cpuacct: Include guest time in user time in cpuacct.stat

[Tejun Heo]

Hello,

Ingo, Peter, this looks fine to be sans a couple typos in the descriptions.
Please let me know if you wanna take them through sched tree. Otherwise,
I'll route them through cgroup tree in a few days.

Thanks.

-- 
tejun

Re: [PATCH v3 1/4] cputime, cpuacct: Include guest time in user time in cpuacct.stat

[Peter Zijlstra]

On Mon, Nov 15, 2021 at 11:06:16AM -1000, Tejun Heo wrote:
> Hello,
> 
> Ingo, Peter, this looks fine to be sans a couple typos in the descriptions.
> Please let me know if you wanna take them through sched tree. Otherwise,
> I'll route them through cgroup tree in a few days.

This is all cgroup only changes, right? That's not immediately obvious
from the changelogs.

But yeah, I can take them I suppose. Lemme go queue them.

[tip: sched/core] cputime, cpuacct: Include guest time in user time in cpuacct.stat

[tip-bot2 for Andrey Ryabinin]

The following commit has been merged into the sched/core branch of tip:

Commit-ID:     9731698ecb9c851f353ce2496292ff9fcea39dff
Gitweb:        https://git.kernel.org/tip/9731698ecb9c851f353ce2496292ff9fcea39dff
Author:        Andrey Ryabinin <arbn@yandex-team.com>
AuthorDate:    Mon, 15 Nov 2021 19:46:04 +03:00
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Tue, 23 Nov 2021 09:55:22 +01:00

cputime, cpuacct: Include guest time in user time in cpuacct.stat

cpuacct.stat in no-root cgroups shows user time without guest time
included int it. This doesn't match with user time shown in root
cpuacct.stat and /proc/<pid>/stat. This also affects cgroup2's cpu.stat
in the same way.

Make account_guest_time() to add user time to cgroup's cpustat to
fix this.

Fixes: ef12fefabf94 ("cpuacct: add per-cgroup utime/stime statistics")
Signed-off-by: Andrey Ryabinin <arbn@yandex-team.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Acked-by: Tejun Heo <tj@kernel.org>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20211115164607.23784-1-arbn@yandex-team.com
---
 kernel/sched/cputime.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/kernel/sched/cputime.c b/kernel/sched/cputime.c
index 872e481..042a6db 100644
--- a/kernel/sched/cputime.c
+++ b/kernel/sched/cputime.c
@@ -148,10 +148,10 @@ void account_guest_time(struct task_struct *p, u64 cputime)
 
 	/* Add guest time to cpustat. */
 	if (task_nice(p) > 0) {
-		cpustat[CPUTIME_NICE] += cputime;
+		task_group_account_field(p, CPUTIME_NICE, cputime);
 		cpustat[CPUTIME_GUEST_NICE] += cputime;
 	} else {
-		cpustat[CPUTIME_USER] += cputime;
+		task_group_account_field(p, CPUTIME_USER, cputime);
 		cpustat[CPUTIME_GUEST] += cputime;
 	}
 }