Add a per-CPU cache of the pinned and flexible perf_event_groups_first value for a cgroup avoiding an O(log(#perf events)) searches during sched_in.
This patch is derived from an original patch by Kan Liang <kan.li...@linux.intel.com>: https://lkml.org/lkml/2019/5/15/1594 Signed-off-by: Ian Rogers <irog...@google.com> --- include/linux/perf_event.h | 6 ++ kernel/events/core.c | 142 ++++++++++++++++++++++++++----------- 2 files changed, 106 insertions(+), 42 deletions(-) diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index 43f90cfa2c39..2d411786ab87 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -845,6 +845,12 @@ struct perf_cgroup_info { struct perf_cgroup { struct cgroup_subsys_state css; struct perf_cgroup_info __percpu *info; + /* A cache of the first event with the perf_cpu_context's + * perf_event_context for the first event in pinned_groups or + * flexible_groups. Avoids an rbtree search during sched_in. + */ + struct perf_event * __percpu *pinned_event; + struct perf_event * __percpu *flexible_event; }; /* diff --git a/kernel/events/core.c b/kernel/events/core.c index 2a2188908bed..c8b9c8611533 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -1592,6 +1592,25 @@ perf_event_groups_insert(struct perf_event_groups *groups, rb_link_node(&event->group_node, parent, node); rb_insert_color(&event->group_node, &groups->tree); +#ifdef CONFIG_CGROUP_PERF + if (is_cgroup_event(event)) { + struct perf_event **cgrp_event; + + if (event->attr.pinned) + cgrp_event = per_cpu_ptr(event->cgrp->pinned_event, + event->cpu); + else + cgrp_event = per_cpu_ptr(event->cgrp->flexible_event, + event->cpu); + /* + * Cgroup events for the same cgroup on the same CPU will + * always be inserted at the right because of bigger + * @groups->index. Only need to set *cgrp_event when it's NULL. + */ + if (!*cgrp_event) + *cgrp_event = event; + } +#endif } /* @@ -1606,6 +1625,9 @@ add_event_to_groups(struct perf_event *event, struct perf_event_context *ctx) perf_event_groups_insert(groups, event); } +static struct perf_event * +perf_event_groups_next(struct perf_event *event); + /* * Delete a group from a tree. */ @@ -1616,6 +1638,22 @@ perf_event_groups_delete(struct perf_event_groups *groups, WARN_ON_ONCE(RB_EMPTY_NODE(&event->group_node) || RB_EMPTY_ROOT(&groups->tree)); +#ifdef CONFIG_CGROUP_PERF + if (is_cgroup_event(event)) { + struct perf_event **cgrp_event; + + if (event->attr.pinned) + cgrp_event = per_cpu_ptr(event->cgrp->pinned_event, + event->cpu); + else + cgrp_event = per_cpu_ptr(event->cgrp->flexible_event, + event->cpu); + + if (*cgrp_event == event) + *cgrp_event = perf_event_groups_next(event); + } +#endif + rb_erase(&event->group_node, &groups->tree); init_event_group(event); } @@ -1633,20 +1671,14 @@ del_event_from_groups(struct perf_event *event, struct perf_event_context *ctx) } /* - * Get the leftmost event in the cpu/cgroup subtree. + * Get the leftmost event in the cpu subtree without a cgroup (ie task or + * system-wide). */ static struct perf_event * -perf_event_groups_first(struct perf_event_groups *groups, int cpu, - struct cgroup *cgrp) +perf_event_groups_first_no_cgroup(struct perf_event_groups *groups, int cpu) { struct perf_event *node_event = NULL, *match = NULL; struct rb_node *node = groups->tree.rb_node; -#ifdef CONFIG_CGROUP_PERF - int node_cgrp_id, cgrp_id = 0; - - if (cgrp) - cgrp_id = cgrp->id; -#endif while (node) { node_event = container_of(node, struct perf_event, group_node); @@ -1660,18 +1692,10 @@ perf_event_groups_first(struct perf_event_groups *groups, int cpu, continue; } #ifdef CONFIG_CGROUP_PERF - node_cgrp_id = 0; - if (node_event->cgrp && node_event->cgrp->css.cgroup) - node_cgrp_id = node_event->cgrp->css.cgroup->id; - - if (cgrp_id < node_cgrp_id) { + if (node_event->cgrp) { node = node->rb_left; continue; } - if (cgrp_id > node_cgrp_id) { - node = node->rb_right; - continue; - } #endif match = node_event; node = node->rb_left; @@ -3433,6 +3457,13 @@ static void min_heap_pop_push(struct perf_event_heap *heap, } } +static int pinned_sched_in(struct perf_event_context *ctx, + struct perf_cpu_context *cpuctx, + struct perf_event *event); +static int flexible_sched_in(struct perf_event_context *ctx, + struct perf_cpu_context *cpuctx, + struct perf_event *event, + int *can_add_hw); /* * Without cgroups, with a task context, there may be per-CPU and any @@ -3442,11 +3473,7 @@ static void min_heap_pop_push(struct perf_event_heap *heap, static int visit_groups_merge(struct perf_event_context *ctx, struct perf_cpu_context *cpuctx, - struct perf_event_groups *groups, - int (*func)(struct perf_event_context *, - struct perf_cpu_context *, - struct perf_event *, - int *), + bool is_pinned, int *data) { /* @@ -3472,7 +3499,11 @@ static int visit_groups_merge(struct perf_event_context *ctx, #endif int ret, cpu = smp_processor_id(); - heap.storage[0] = perf_event_groups_first(groups, cpu, NULL); + struct perf_event_groups *groups = is_pinned + ? &ctx->pinned_groups + : &ctx->flexible_groups; + + heap.storage[0] = perf_event_groups_first_no_cgroup(groups, cpu); if (heap.storage[0]) heap.num_elements++; @@ -3482,7 +3513,7 @@ static int visit_groups_merge(struct perf_event_context *ctx, * events. */ heap.storage[heap.num_elements] = - perf_event_groups_first(groups, -1, NULL); + perf_event_groups_first_no_cgroup(groups, -1); if (heap.storage[heap.num_elements]) heap.num_elements++; } else { @@ -3492,14 +3523,22 @@ static int visit_groups_merge(struct perf_event_context *ctx, * For itrs[1..MAX] add an iterator for each cgroup parent that * has events. */ - if (cpuctx->cgrp) { + struct perf_cgroup *cgrp = cpuctx->cgrp; + + if (cgrp) { struct cgroup_subsys_state *css; - for (css = &cpuctx->cgrp->css; css; css = css->parent) { - heap.storage[heap.num_elements] = - perf_event_groups_first(groups, - cpu, - css->cgroup); + for (css = &cgrp->css; css; css = css->parent) { + /* root cgroup doesn't have events */ + if (css->id == 1) + break; + + cgrp = container_of(css, struct perf_cgroup, + css); + heap.storage[heap.num_elements] = is_pinned + ? *per_cpu_ptr(cgrp->pinned_event, cpu) + : *per_cpu_ptr(cgrp->flexible_event, + cpu); if (heap.storage[heap.num_elements]) { heap.num_elements++; if (heap.num_elements == @@ -3516,7 +3555,12 @@ static int visit_groups_merge(struct perf_event_context *ctx, min_heapify_all(&heap); while (heap.num_elements > 0) { - ret = func(ctx, cpuctx, heap.storage[0], data); + if (is_pinned) + ret = pinned_sched_in(ctx, cpuctx, heap.storage[0]); + else + ret = flexible_sched_in(ctx, cpuctx, heap.storage[0], + data); + if (ret) return ret; @@ -3529,8 +3573,7 @@ static int visit_groups_merge(struct perf_event_context *ctx, static int pinned_sched_in(struct perf_event_context *ctx, struct perf_cpu_context *cpuctx, - struct perf_event *event, - int *unused) + struct perf_event *event) { if (event->state <= PERF_EVENT_STATE_OFF) return 0; @@ -3583,8 +3626,7 @@ ctx_pinned_sched_in(struct perf_event_context *ctx, { visit_groups_merge(ctx, cpuctx, - &ctx->pinned_groups, - pinned_sched_in, + /*is_pinned=*/true, NULL); } @@ -3596,8 +3638,7 @@ ctx_flexible_sched_in(struct perf_event_context *ctx, visit_groups_merge(ctx, cpuctx, - &ctx->flexible_groups, - flexible_sched_in, + /*is_pinned=*/false, &can_add_hw); } @@ -12417,18 +12458,35 @@ perf_cgroup_css_alloc(struct cgroup_subsys_state *parent_css) return ERR_PTR(-ENOMEM); jc->info = alloc_percpu(struct perf_cgroup_info); - if (!jc->info) { - kfree(jc); - return ERR_PTR(-ENOMEM); - } + if (!jc->info) + goto free_jc; + + jc->pinned_event = alloc_percpu(struct perf_event *); + if (!jc->pinned_event) + goto free_jc_info; + + jc->flexible_event = alloc_percpu(struct perf_event *); + if (!jc->flexible_event) + goto free_jc_pinned; return &jc->css; + +free_jc_pinned: + free_percpu(jc->pinned_event); +free_jc_info: + free_percpu(jc->info); +free_jc: + kfree(jc); + + return ERR_PTR(-ENOMEM); } static void perf_cgroup_css_free(struct cgroup_subsys_state *css) { struct perf_cgroup *jc = container_of(css, struct perf_cgroup, css); + free_percpu(jc->pinned_event); + free_percpu(jc->flexible_event); free_percpu(jc->info); kfree(jc); } -- 2.22.0.709.g102302147b-goog
