From: Andrey Ryabinin arbn@yandex-team.com

commit dd02d4234c9a2214a81c57a16484304a1a51872a upstream.

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 [OP: adjusted context for v5.10] Signed-off-by: Ovidiu Panait ovidiu.panait@windriver.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 941c28cf9738..8a260115a137 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;

/* @@ -115,14 +112,17 @@ static u64 cpuacct_cpuusage_read(struct cpuacct *ca, int cpu, raw_spin_lock_irq(&cpu_rq(cpu)->lock); #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 @@ -132,10 +132,14 @@ 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 /* @@ -143,9 +147,10 @@ static void cpuacct_cpuusage_write(struct cpuacct *ca, int cpu, u64 val) */ raw_spin_lock_irq(&cpu_rq(cpu)->lock); #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_unlock_irq(&cpu_rq(cpu)->lock); @@ -196,7 +201,7 @@ 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; } @@ -243,25 +248,10 @@ 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_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 - } + for (index = 0; index < CPUACCT_STAT_NSTATS; index++) + seq_printf(m, " %llu", + cpuacct_cpuusage_read(ca, cpu, index)); seq_puts(m, "\n"); } return 0; @@ -339,16 +329,11 @@ 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(); }

From: Chengming Zhou zhouchengming@bytedance.com

commit dc6e0818bc9a0336d9accf3ea35d146d72aa7a18 upstream.

Since cpuacct_charge() is called from the scheduler update_curr(), we must already have rq lock held, then the RCU read lock can be optimized away.

And do the same thing in it's wrapper cgroup_account_cputime(), but we can't use lockdep_assert_rq_held() there, which defined in kernel/sched/sched.h.

Suggested-by: Peter Zijlstra (Intel) peterz@infradead.org Signed-off-by: Chengming Zhou zhouchengming@bytedance.com Signed-off-by: Peter Zijlstra (Intel) peterz@infradead.org Link: https://lore.kernel.org/r/20220220051426.5274-2-zhouchengming@bytedance.com [OP: adjusted lockdep_assert_rq_held() -> lockdep_assert_held()] Signed-off-by: Ovidiu Panait ovidiu.panait@windriver.com --- include/linux/cgroup.h | 2 -- kernel/sched/cpuacct.c | 4 +--- 2 files changed, 1 insertion(+), 5 deletions(-)

diff --git a/include/linux/cgroup.h b/include/linux/cgroup.h index 959b370733f0..7653f5418950 100644 --- a/include/linux/cgroup.h +++ b/include/linux/cgroup.h @@ -779,11 +779,9 @@ static inline void cgroup_account_cputime(struct task_struct *task,

cpuacct_charge(task, delta_exec);

- rcu_read_lock(); cgrp = task_dfl_cgroup(task); if (cgroup_parent(cgrp)) __cgroup_account_cputime(cgrp, delta_exec); - rcu_read_unlock(); }

static inline void cgroup_account_cputime_field(struct task_struct *task, diff --git a/kernel/sched/cpuacct.c b/kernel/sched/cpuacct.c index 3c59c541dd31..8ee298321d78 100644 --- a/kernel/sched/cpuacct.c +++ b/kernel/sched/cpuacct.c @@ -331,12 +331,10 @@ void cpuacct_charge(struct task_struct *tsk, u64 cputime) unsigned int cpu = task_cpu(tsk); struct cpuacct *ca;

- rcu_read_lock(); + lockdep_assert_held(&cpu_rq(cpu)->lock);

for (ca = task_ca(tsk); ca; ca = parent_ca(ca)) *per_cpu_ptr(ca->cpuusage, cpu) += cputime; - - rcu_read_unlock(); }

/*