[PATCH v2 10/10] sched/psi: cache parent psi_group to speed up groups iterate
From: Chengming Zhou <hidden>
Date: 2022-08-08 11:06:27
Also in:
cgroups, lkml
Subsystem:
pressure stall information (psi), scheduler, the rest · Maintainers:
Johannes Weiner, Suren Baghdasaryan, Ingo Molnar, Peter Zijlstra, Juri Lelli, Vincent Guittot, Linus Torvalds
We use iterate_groups() to iterate each level psi_group to update PSI stats, which is a very hot path. In current code, iterate_groups() have to use multiple branches and cgroup_parent() to get parent psi_group for each level, which is not very efficient. This patch cache parent psi_group in struct psi_group, only need to get psi_group of task itself first, then just use group->parent to iterate. Signed-off-by: Chengming Zhou <redacted> --- include/linux/psi_types.h | 1 + kernel/sched/psi.c | 51 ++++++++++++++++++++------------------- 2 files changed, 27 insertions(+), 25 deletions(-)
diff --git a/include/linux/psi_types.h b/include/linux/psi_types.h
index fced39e255aa..7459a47fcb1f 100644
--- a/include/linux/psi_types.h
+++ b/include/linux/psi_types.h@@ -148,6 +148,7 @@ struct psi_trigger { struct psi_group { bool enabled; + struct psi_group *parent; /* Protects data used by the aggregator */ struct mutex avgs_lock;
diff --git a/kernel/sched/psi.c b/kernel/sched/psi.c
index 9df1686ee02d..d3c1c49b9bcf 100644
--- a/kernel/sched/psi.c
+++ b/kernel/sched/psi.c@@ -786,30 +786,22 @@ static void psi_group_change(struct psi_group *group, int cpu, schedule_delayed_work(&group->avgs_work, PSI_FREQ); } -static struct psi_group *iterate_groups(struct task_struct *task, void **iter) +static inline struct psi_group *task_psi_group(struct task_struct *task) { - if (*iter == &psi_system) - return NULL; - #ifdef CONFIG_CGROUPS if (static_branch_likely(&psi_cgroups_enabled)) { - struct cgroup *cgroup = NULL; - - if (!*iter) - cgroup = task->cgroups->dfl_cgrp; - else - cgroup = cgroup_parent(*iter); + struct cgroup *cgroup = task_dfl_cgroup(task); - if (cgroup && cgroup_parent(cgroup)) { - *iter = cgroup; + if (cgroup && cgroup_parent(cgroup)) return cgroup_psi(cgroup); - } } #endif - *iter = &psi_system; return &psi_system; } +#define for_each_psi_group(group) \ + for (; group; group = group->parent) + static void psi_flags_change(struct task_struct *task, int clear, int set) { if (((task->psi_flags & set) ||
@@ -827,12 +819,11 @@ static void psi_flags_change(struct task_struct *task, int clear, int set) void psi_change_groups(struct task_struct *task, int clear, int set) { + struct psi_group *group = task_psi_group(task); int cpu = task_cpu(task); - struct psi_group *group; - void *iter = NULL; u64 now = cpu_clock(cpu); - while ((group = iterate_groups(task, &iter))) + for_each_psi_group(group) psi_group_change(group, cpu, clear, set, now, true); }
@@ -850,7 +841,6 @@ void psi_task_switch(struct task_struct *prev, struct task_struct *next, { struct psi_group *group, *common = NULL; int cpu = task_cpu(prev); - void *iter; u64 now = cpu_clock(cpu); if (next->pid) {
@@ -861,8 +851,8 @@ void psi_task_switch(struct task_struct *prev, struct task_struct *next, * we reach the first common ancestor. Iterate @next's * ancestors only until we encounter @prev's ONCPU. */ - iter = NULL; - while ((group = iterate_groups(next, &iter))) { + group = task_psi_group(next); + for_each_psi_group(group) { if (per_cpu_ptr(group->pcpu, cpu)->state_mask & PSI_ONCPU) { common = group;
@@ -903,9 +893,12 @@ void psi_task_switch(struct task_struct *prev, struct task_struct *next, psi_flags_change(prev, clear, set); - iter = NULL; - while ((group = iterate_groups(prev, &iter)) && group != common) + group = task_psi_group(prev); + for_each_psi_group(group) { + if (group == common) + break; psi_group_change(group, cpu, clear, set, now, wake_clock); + } /* * TSK_ONCPU is handled up to the common ancestor. If we're tasked
@@ -913,7 +906,7 @@ void psi_task_switch(struct task_struct *prev, struct task_struct *next, */ if (sleep || unlikely(prev->in_memstall != next->in_memstall)) { clear &= ~TSK_ONCPU; - for (; group; group = iterate_groups(prev, &iter)) + for_each_psi_group(group) psi_group_change(group, cpu, clear, set, now, wake_clock); } }
@@ -922,7 +915,6 @@ void psi_task_switch(struct task_struct *prev, struct task_struct *next, void psi_account_irqtime(struct task_struct *task, u32 delta) { int cpu = task_cpu(task); - void *iter = NULL; struct psi_group *group; struct psi_group_cpu *groupc; u64 now;
@@ -932,7 +924,8 @@ void psi_account_irqtime(struct task_struct *task, u32 delta) now = cpu_clock(cpu); - while ((group = iterate_groups(task, &iter))) { + group = task_psi_group(task); + for_each_psi_group(group) { groupc = per_cpu_ptr(group->pcpu, cpu); write_seqcount_begin(&groupc->seq);
@@ -1010,6 +1003,8 @@ void psi_memstall_leave(unsigned long *flags) #ifdef CONFIG_CGROUPS int psi_cgroup_alloc(struct cgroup *cgroup) { + struct cgroup *parent; + if (!static_branch_likely(&psi_cgroups_enabled)) return 0;
@@ -1017,6 +1012,12 @@ int psi_cgroup_alloc(struct cgroup *cgroup) if (!cgroup->psi.pcpu) return -ENOMEM; group_init(&cgroup->psi); + + parent = cgroup_parent(cgroup); + if (parent && cgroup_parent(parent)) + cgroup->psi.parent = cgroup_psi(parent); + else + cgroup->psi.parent = &psi_system; return 0; }
--
2.36.1