diff options
Diffstat (limited to 'kernel/events/core.c')
-rw-r--r-- | kernel/events/core.c | 592 |
1 files changed, 454 insertions, 138 deletions
diff --git a/kernel/events/core.c b/kernel/events/core.c index e453589da97c..bc9b98a9af9a 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -28,6 +28,7 @@ #include <linux/export.h> #include <linux/vmalloc.h> #include <linux/hardirq.h> +#include <linux/hugetlb.h> #include <linux/rculist.h> #include <linux/uaccess.h> #include <linux/syscalls.h> @@ -49,6 +50,7 @@ #include <linux/sched/mm.h> #include <linux/proc_ns.h> #include <linux/mount.h> +#include <linux/min_heap.h> #include "internal.h" @@ -386,6 +388,7 @@ static atomic_t nr_freq_events __read_mostly; static atomic_t nr_switch_events __read_mostly; static atomic_t nr_ksymbol_events __read_mostly; static atomic_t nr_bpf_events __read_mostly; +static atomic_t nr_cgroup_events __read_mostly; static LIST_HEAD(pmus); static DEFINE_MUTEX(pmus_lock); @@ -891,6 +894,47 @@ static inline void perf_cgroup_sched_in(struct task_struct *prev, rcu_read_unlock(); } +static int perf_cgroup_ensure_storage(struct perf_event *event, + struct cgroup_subsys_state *css) +{ + struct perf_cpu_context *cpuctx; + struct perf_event **storage; + int cpu, heap_size, ret = 0; + + /* + * Allow storage to have sufficent space for an iterator for each + * possibly nested cgroup plus an iterator for events with no cgroup. + */ + for (heap_size = 1; css; css = css->parent) + heap_size++; + + for_each_possible_cpu(cpu) { + cpuctx = per_cpu_ptr(event->pmu->pmu_cpu_context, cpu); + if (heap_size <= cpuctx->heap_size) + continue; + + storage = kmalloc_node(heap_size * sizeof(struct perf_event *), + GFP_KERNEL, cpu_to_node(cpu)); + if (!storage) { + ret = -ENOMEM; + break; + } + + raw_spin_lock_irq(&cpuctx->ctx.lock); + if (cpuctx->heap_size < heap_size) { + swap(cpuctx->heap, storage); + if (storage == cpuctx->heap_default) + storage = NULL; + cpuctx->heap_size = heap_size; + } + raw_spin_unlock_irq(&cpuctx->ctx.lock); + + kfree(storage); + } + + return ret; +} + static inline int perf_cgroup_connect(int fd, struct perf_event *event, struct perf_event_attr *attr, struct perf_event *group_leader) @@ -910,6 +954,10 @@ static inline int perf_cgroup_connect(int fd, struct perf_event *event, goto out; } + ret = perf_cgroup_ensure_storage(event, css); + if (ret) + goto out; + cgrp = container_of(css, struct perf_cgroup, css); event->cgrp = cgrp; @@ -935,16 +983,10 @@ perf_cgroup_set_shadow_time(struct perf_event *event, u64 now) event->shadow_ctx_time = now - t->timestamp; } -/* - * Update cpuctx->cgrp so that it is set when first cgroup event is added and - * cleared when last cgroup event is removed. - */ static inline void -list_update_cgroup_event(struct perf_event *event, - struct perf_event_context *ctx, bool add) +perf_cgroup_event_enable(struct perf_event *event, struct perf_event_context *ctx) { struct perf_cpu_context *cpuctx; - struct list_head *cpuctx_entry; if (!is_cgroup_event(event)) return; @@ -961,28 +1003,41 @@ list_update_cgroup_event(struct perf_event *event, * because if the first would mismatch, the second would not try again * and we would leave cpuctx->cgrp unset. */ - if (add && !cpuctx->cgrp) { + if (ctx->is_active && !cpuctx->cgrp) { struct perf_cgroup *cgrp = perf_cgroup_from_task(current, ctx); if (cgroup_is_descendant(cgrp->css.cgroup, event->cgrp->css.cgroup)) cpuctx->cgrp = cgrp; } - if (add && ctx->nr_cgroups++) + if (ctx->nr_cgroups++) + return; + + list_add(&cpuctx->cgrp_cpuctx_entry, + per_cpu_ptr(&cgrp_cpuctx_list, event->cpu)); +} + +static inline void +perf_cgroup_event_disable(struct perf_event *event, struct perf_event_context *ctx) +{ + struct perf_cpu_context *cpuctx; + + if (!is_cgroup_event(event)) return; - else if (!add && --ctx->nr_cgroups) + + /* + * Because cgroup events are always per-cpu events, + * @ctx == &cpuctx->ctx. + */ + cpuctx = container_of(ctx, struct perf_cpu_context, ctx); + + if (--ctx->nr_cgroups) return; - /* no cgroup running */ - if (!add) + if (ctx->is_active && cpuctx->cgrp) cpuctx->cgrp = NULL; - cpuctx_entry = &cpuctx->cgrp_cpuctx_entry; - if (add) - list_add(cpuctx_entry, - per_cpu_ptr(&cgrp_cpuctx_list, event->cpu)); - else - list_del(cpuctx_entry); + list_del(&cpuctx->cgrp_cpuctx_entry); } #else /* !CONFIG_CGROUP_PERF */ @@ -1048,11 +1103,14 @@ static inline u64 perf_cgroup_event_time(struct perf_event *event) } static inline void -list_update_cgroup_event(struct perf_event *event, - struct perf_event_context *ctx, bool add) +perf_cgroup_event_enable(struct perf_event *event, struct perf_event_context *ctx) { } +static inline void +perf_cgroup_event_disable(struct perf_event *event, struct perf_event_context *ctx) +{ +} #endif /* @@ -1249,7 +1307,7 @@ static void put_ctx(struct perf_event_context *ctx) * function. * * Lock order: - * cred_guard_mutex + * exec_update_mutex * task_struct::perf_event_mutex * perf_event_context::mutex * perf_event::child_mutex; @@ -1531,6 +1589,30 @@ perf_event_groups_less(struct perf_event *left, struct perf_event *right) if (left->cpu > right->cpu) return false; +#ifdef CONFIG_CGROUP_PERF + if (left->cgrp != right->cgrp) { + if (!left->cgrp || !left->cgrp->css.cgroup) { + /* + * Left has no cgroup but right does, no cgroups come + * first. + */ + return true; + } + if (!right->cgrp || !right->cgrp->css.cgroup) { + /* + * Right has no cgroup but left does, no cgroups come + * first. + */ + return false; + } + /* Two dissimilar cgroups, order by id. */ + if (left->cgrp->css.cgroup->kn->id < right->cgrp->css.cgroup->kn->id) + return true; + + return false; + } +#endif + if (left->group_index < right->group_index) return true; if (left->group_index > right->group_index) @@ -1610,25 +1692,48 @@ del_event_from_groups(struct perf_event *event, struct perf_event_context *ctx) } /* - * Get the leftmost event in the @cpu subtree. + * Get the leftmost event in the cpu/cgroup subtree. */ static struct perf_event * -perf_event_groups_first(struct perf_event_groups *groups, int cpu) +perf_event_groups_first(struct perf_event_groups *groups, int cpu, + struct cgroup *cgrp) { struct perf_event *node_event = NULL, *match = NULL; struct rb_node *node = groups->tree.rb_node; +#ifdef CONFIG_CGROUP_PERF + u64 node_cgrp_id, cgrp_id = 0; + + if (cgrp) + cgrp_id = cgrp->kn->id; +#endif while (node) { node_event = container_of(node, struct perf_event, group_node); if (cpu < node_event->cpu) { node = node->rb_left; - } else if (cpu > node_event->cpu) { + continue; + } + if (cpu > node_event->cpu) { node = node->rb_right; - } else { - match = node_event; + 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->kn->id; + + if (cgrp_id < node_cgrp_id) { node = node->rb_left; + continue; } + if (cgrp_id > node_cgrp_id) { + node = node->rb_right; + continue; + } +#endif + match = node_event; + node = node->rb_left; } return match; @@ -1641,12 +1746,26 @@ static struct perf_event * perf_event_groups_next(struct perf_event *event) { struct perf_event *next; +#ifdef CONFIG_CGROUP_PERF + u64 curr_cgrp_id = 0; + u64 next_cgrp_id = 0; +#endif next = rb_entry_safe(rb_next(&event->group_node), typeof(*event), group_node); - if (next && next->cpu == event->cpu) - return next; + if (next == NULL || next->cpu != event->cpu) + return NULL; - return NULL; +#ifdef CONFIG_CGROUP_PERF + if (event->cgrp && event->cgrp->css.cgroup) + curr_cgrp_id = event->cgrp->css.cgroup->kn->id; + + if (next->cgrp && next->cgrp->css.cgroup) + next_cgrp_id = next->cgrp->css.cgroup->kn->id; + + if (curr_cgrp_id != next_cgrp_id) + return NULL; +#endif + return next; } /* @@ -1682,13 +1801,14 @@ list_add_event(struct perf_event *event, struct perf_event_context *ctx) add_event_to_groups(event, ctx); } - list_update_cgroup_event(event, ctx, true); - list_add_rcu(&event->event_entry, &ctx->event_list); ctx->nr_events++; if (event->attr.inherit_stat) ctx->nr_stat++; + if (event->state > PERF_EVENT_STATE_OFF) + perf_cgroup_event_enable(event, ctx); + ctx->generation++; } @@ -1754,6 +1874,9 @@ static void __perf_event_header_size(struct perf_event *event, u64 sample_type) if (sample_type & PERF_SAMPLE_PHYS_ADDR) size += sizeof(data->phys_addr); + if (sample_type & PERF_SAMPLE_CGROUP) + size += sizeof(data->cgroup); + event->header_size = size; } @@ -1864,8 +1987,6 @@ list_del_event(struct perf_event *event, struct perf_event_context *ctx) event->attach_state &= ~PERF_ATTACH_CONTEXT; - list_update_cgroup_event(event, ctx, false); - ctx->nr_events--; if (event->attr.inherit_stat) ctx->nr_stat--; @@ -1882,8 +2003,10 @@ list_del_event(struct perf_event *event, struct perf_event_context *ctx) * of error state is by explicit re-enabling * of the event */ - if (event->state > PERF_EVENT_STATE_OFF) + if (event->state > PERF_EVENT_STATE_OFF) { + perf_cgroup_event_disable(event, ctx); perf_event_set_state(event, PERF_EVENT_STATE_OFF); + } ctx->generation++; } @@ -1986,6 +2109,12 @@ static int perf_get_aux_event(struct perf_event *event, return 1; } +static inline struct list_head *get_event_list(struct perf_event *event) +{ + struct perf_event_context *ctx = event->ctx; + return event->attr.pinned ? &ctx->pinned_active : &ctx->flexible_active; +} + static void perf_group_detach(struct perf_event *event) { struct perf_event *sibling, *tmp; @@ -2028,12 +2157,8 @@ static void perf_group_detach(struct perf_event *event) if (!RB_EMPTY_NODE(&event->group_node)) { add_event_to_groups(sibling, event->ctx); - if (sibling->state == PERF_EVENT_STATE_ACTIVE) { - struct list_head *list = sibling->attr.pinned ? - &ctx->pinned_active : &ctx->flexible_active; - - list_add_tail(&sibling->active_list, list); - } + if (sibling->state == PERF_EVENT_STATE_ACTIVE) + list_add_tail(&sibling->active_list, get_event_list(sibling)); } WARN_ON_ONCE(sibling->ctx != event->ctx); @@ -2112,6 +2237,7 @@ event_sched_out(struct perf_event *event, if (READ_ONCE(event->pending_disable) >= 0) { WRITE_ONCE(event->pending_disable, -1); + perf_cgroup_event_disable(event, ctx); state = PERF_EVENT_STATE_OFF; } perf_event_set_state(event, state); @@ -2182,6 +2308,7 @@ __perf_remove_from_context(struct perf_event *event, if (!ctx->nr_events && ctx->is_active) { ctx->is_active = 0; + ctx->rotate_necessary = 0; if (ctx->task) { WARN_ON_ONCE(cpuctx->task_ctx != ctx); cpuctx->task_ctx = NULL; @@ -2248,6 +2375,7 @@ static void __perf_event_disable(struct perf_event *event, event_sched_out(event, cpuctx, ctx); perf_event_set_state(event, PERF_EVENT_STATE_OFF); + perf_cgroup_event_disable(event, ctx); } /* @@ -2350,6 +2478,8 @@ event_sched_in(struct perf_event *event, { int ret = 0; + WARN_ON_ONCE(event->ctx != ctx); + lockdep_assert_held(&ctx->lock); if (event->state <= PERF_EVENT_STATE_OFF) @@ -2629,7 +2759,7 @@ static int __perf_install_in_context(void *info) } #ifdef CONFIG_CGROUP_PERF - if (is_cgroup_event(event)) { + if (event->state > PERF_EVENT_STATE_OFF && is_cgroup_event(event)) { /* * If the current cgroup doesn't match the event's * cgroup, we should not try to schedule it. @@ -2789,6 +2919,7 @@ static void __perf_event_enable(struct perf_event *event, ctx_sched_out(ctx, cpuctx, EVENT_TIME); perf_event_set_state(event, PERF_EVENT_STATE_INACTIVE); + perf_cgroup_event_enable(event, ctx); if (!ctx->is_active) return; @@ -3077,12 +3208,6 @@ static void ctx_sched_out(struct perf_event_context *ctx, if (!ctx->nr_active || !(is_active & EVENT_ALL)) return; - /* - * If we had been multiplexing, no rotations are necessary, now no events - * are active. - */ - ctx->rotate_necessary = 0; - perf_pmu_disable(ctx->pmu); if (is_active & EVENT_PINNED) { list_for_each_entry_safe(event, tmp, &ctx->pinned_active, active_list) @@ -3092,6 +3217,13 @@ static void ctx_sched_out(struct perf_event_context *ctx, if (is_active & EVENT_FLEXIBLE) { list_for_each_entry_safe(event, tmp, &ctx->flexible_active, active_list) group_sched_out(event, cpuctx, ctx); + + /* + * Since we cleared EVENT_FLEXIBLE, also clear + * rotate_necessary, is will be reset by + * ctx_flexible_sched_in() when needed. + */ + ctx->rotate_necessary = 0; } perf_pmu_enable(ctx->pmu); } @@ -3388,71 +3520,104 @@ static void cpu_ctx_sched_out(struct perf_cpu_context *cpuctx, ctx_sched_out(&cpuctx->ctx, cpuctx, event_type); } -static int visit_groups_merge(struct perf_event_groups *groups, int cpu, - int (*func)(struct perf_event *, void *), void *data) +static bool perf_less_group_idx(const void *l, const void *r) { - struct perf_event **evt, *evt1, *evt2; - int ret; - - evt1 = perf_event_groups_first(groups, -1); - evt2 = perf_event_groups_first(groups, cpu); + const struct perf_event *le = *(const struct perf_event **)l; + const struct perf_event *re = *(const struct perf_event **)r; - while (evt1 || evt2) { - if (evt1 && evt2) { - if (evt1->group_index < evt2->group_index) - evt = &evt1; - else - evt = &evt2; - } else if (evt1) { - evt = &evt1; - } else { - evt = &evt2; - } - - ret = func(*evt, data); - if (ret) - return ret; + return le->group_index < re->group_index; +} - *evt = perf_event_groups_next(*evt); - } +static void swap_ptr(void *l, void *r) +{ + void **lp = l, **rp = r; - return 0; + swap(*lp, *rp); } -struct sched_in_data { - struct perf_event_context *ctx; - struct perf_cpu_context *cpuctx; - int can_add_hw; +static const struct min_heap_callbacks perf_min_heap = { + .elem_size = sizeof(struct perf_event *), + .less = perf_less_group_idx, + .swp = swap_ptr, }; -static int pinned_sched_in(struct perf_event *event, void *data) +static void __heap_add(struct min_heap *heap, struct perf_event *event) { - struct sched_in_data *sid = data; + struct perf_event **itrs = heap->data; - if (event->state <= PERF_EVENT_STATE_OFF) - return 0; + if (event) { + itrs[heap->nr] = event; + heap->nr++; + } +} - if (!event_filter_match(event)) - return 0; +static noinline int visit_groups_merge(struct perf_cpu_context *cpuctx, + struct perf_event_groups *groups, int cpu, + int (*func)(struct perf_event *, void *), + void *data) +{ +#ifdef CONFIG_CGROUP_PERF + struct cgroup_subsys_state *css = NULL; +#endif + /* Space for per CPU and/or any CPU event iterators. */ + struct perf_event *itrs[2]; + struct min_heap event_heap; + struct perf_event **evt; + int ret; + + if (cpuctx) { + event_heap = (struct min_heap){ + .data = cpuctx->heap, + .nr = 0, + .size = cpuctx->heap_size, + }; - if (group_can_go_on(event, sid->cpuctx, sid->can_add_hw)) { - if (!group_sched_in(event, sid->cpuctx, sid->ctx)) - list_add_tail(&event->active_list, &sid->ctx->pinned_active); + lockdep_assert_held(&cpuctx->ctx.lock); + +#ifdef CONFIG_CGROUP_PERF + if (cpuctx->cgrp) + css = &cpuctx->cgrp->css; +#endif + } else { + event_heap = (struct min_heap){ + .data = itrs, + .nr = 0, + .size = ARRAY_SIZE(itrs), + }; + /* Events not within a CPU context may be on any CPU. */ + __heap_add(&event_heap, perf_event_groups_first(groups, -1, NULL)); } + evt = event_heap.data; - /* - * If this pinned group hasn't been scheduled, - * put it in error state. - */ - if (event->state == PERF_EVENT_STATE_INACTIVE) - perf_event_set_state(event, PERF_EVENT_STATE_ERROR); + __heap_add(&event_heap, perf_event_groups_first(groups, cpu, NULL)); + +#ifdef CONFIG_CGROUP_PERF + for (; css; css = css->parent) + __heap_add(&event_heap, perf_event_groups_first(groups, cpu, css->cgroup)); +#endif + + min_heapify_all(&event_heap, &perf_min_heap); + + while (event_heap.nr) { + ret = func(*evt, data); + if (ret) + return ret; + + *evt = perf_event_groups_next(*evt); + if (*evt) + min_heapify(&event_heap, 0, &perf_min_heap); + else + min_heap_pop(&event_heap, &perf_min_heap); + } return 0; } -static int flexible_sched_in(struct perf_event *event, void *data) +static int merge_sched_in(struct perf_event *event, void *data) { - struct sched_in_data *sid = data; + struct perf_event_context *ctx = event->ctx; + struct perf_cpu_context *cpuctx = __get_cpu_context(ctx); + int *can_add_hw = data; if (event->state <= PERF_EVENT_STATE_OFF) return 0; @@ -3460,14 +3625,19 @@ static int flexible_sched_in(struct perf_event *event, void *data) if (!event_filter_match(event)) return 0; - if (group_can_go_on(event, sid->cpuctx, sid->can_add_hw)) { - int ret = group_sched_in(event, sid->cpuctx, sid->ctx); - if (ret) { - sid->can_add_hw = 0; - sid->ctx->rotate_necessary = 1; - return 0; + if (group_can_go_on(event, cpuctx, *can_add_hw)) { + if (!group_sched_in(event, cpuctx, ctx)) + list_add_tail(&event->active_list, get_event_list(event)); + } + + if (event->state == PERF_EVENT_STATE_INACTIVE) { + if (event->attr.pinned) { + perf_cgroup_event_disable(event, ctx); + perf_event_set_state(event, PERF_EVENT_STATE_ERROR); } - list_add_tail(&event->active_list, &sid->ctx->flexible_active); + + *can_add_hw = 0; + ctx->rotate_necessary = 1; } return 0; @@ -3477,30 +3647,28 @@ static void ctx_pinned_sched_in(struct perf_event_context *ctx, struct perf_cpu_context *cpuctx) { - struct sched_in_data sid = { - .ctx = ctx, - .cpuctx = cpuctx, - .can_add_hw = 1, - }; + int can_add_hw = 1; + + if (ctx != &cpuctx->ctx) + cpuctx = NULL; - visit_groups_merge(&ctx->pinned_groups, + visit_groups_merge(cpuctx, &ctx->pinned_groups, smp_processor_id(), - pinned_sched_in, &sid); + merge_sched_in, &can_add_hw); } static void ctx_flexible_sched_in(struct perf_event_context *ctx, struct perf_cpu_context *cpuctx) { - struct sched_in_data sid = { - .ctx = ctx, - .cpuctx = cpuctx, - .can_add_hw = 1, - }; + int can_add_hw = 1; + + if (ctx != &cpuctx->ctx) + cpuctx = NULL; - visit_groups_merge(&ctx->flexible_groups, + visit_groups_merge(cpuctx, &ctx->flexible_groups, smp_processor_id(), - flexible_sched_in, &sid); + merge_sched_in, &can_add_hw); } static void @@ -3841,6 +4009,12 @@ ctx_event_to_rotate(struct perf_event_context *ctx) typeof(*event), group_node); } + /* + * Unconditionally clear rotate_necessary; if ctx_flexible_sched_in() + * finds there are unschedulable events, it will set it again. + */ + ctx->rotate_necessary = 0; + return event; } @@ -4456,6 +4630,8 @@ static void unaccount_event(struct perf_event *event) atomic_dec(&nr_comm_events); if (event->attr.namespaces) atomic_dec(&nr_namespaces_events); + if (event->attr.cgroup) + atomic_dec(&nr_cgroup_events); if (event->attr.task) atomic_dec(&nr_task_events); if (event->attr.freq) @@ -6555,6 +6731,11 @@ static void perf_output_read(struct perf_output_handle *handle, perf_output_read_one(handle, event, enabled, running); } +static inline bool perf_sample_save_hw_index(struct perf_event *event) +{ + return event->attr.branch_sample_type & PERF_SAMPLE_BRANCH_HW_INDEX; +} + void perf_output_sample(struct perf_output_handle *handle, struct perf_event_header *header, struct perf_sample_data *data, @@ -6643,6 +6824,8 @@ void perf_output_sample(struct perf_output_handle *handle, * sizeof(struct perf_branch_entry); perf_output_put(handle, data->br_stack->nr); + if (perf_sample_save_hw_index(event)) + perf_output_put(handle, data->br_stack->hw_idx); perf_output_copy(handle, data->br_stack->entries, size); } else { /* @@ -6705,6 +6888,9 @@ void perf_output_sample(struct perf_output_handle *handle, if (sample_type & PERF_SAMPLE_PHYS_ADDR) perf_output_put(handle, data->phys_addr); + if (sample_type & PERF_SAMPLE_CGROUP) + perf_output_put(handle, data->cgroup); + if (sample_type & PERF_SAMPLE_AUX) { perf_output_put(handle, data->aux_size); @@ -6748,9 +6934,12 @@ static u64 perf_virt_to_phys(u64 virt) * Try IRQ-safe __get_user_pages_fast first. * If failed, leave phys_addr as 0. */ - if ((current->mm != NULL) && - (__get_user_pages_fast(virt, 1, 0, &p) == 1)) - phys_addr = page_to_phys(p) + virt % PAGE_SIZE; + if (current->mm != NULL) { + pagefault_disable(); + if (__get_user_pages_fast(virt, 1, 0, &p) == 1) + phys_addr = page_to_phys(p) + virt % PAGE_SIZE; + pagefault_enable(); + } if (p) put_page(p); @@ -6836,6 +7025,9 @@ void perf_prepare_sample(struct perf_event_header *header, if (sample_type & PERF_SAMPLE_BRANCH_STACK) { int size = sizeof(u64); /* nr */ if (data->br_stack) { + if (perf_sample_save_hw_index(event)) + size += sizeof(u64); + size += data->br_stack->nr * sizeof(struct perf_branch_entry); } @@ -6901,6 +7093,16 @@ void perf_prepare_sample(struct perf_event_header *header, if (sample_type & PERF_SAMPLE_PHYS_ADDR) data->phys_addr = perf_virt_to_phys(data->addr); +#ifdef CONFIG_CGROUP_PERF + if (sample_type & PERF_SAMPLE_CGROUP) { + struct cgroup *cgrp; + + /* protected by RCU */ + cgrp = task_css_check(current, perf_event_cgrp_id, 1)->cgroup; + data->cgroup = cgroup_id(cgrp); + } +#endif + if (sample_type & PERF_SAMPLE_AUX) { u64 size; @@ -7574,6 +7776,105 @@ void perf_event_namespaces(struct task_struct *task) } /* + * cgroup tracking + */ +#ifdef CONFIG_CGROUP_PERF + +struct perf_cgroup_event { + char *path; + int path_size; + struct { + struct perf_event_header header; + u64 id; + char path[]; + } event_id; +}; + +static int perf_event_cgroup_match(struct perf_event *event) +{ + return event->attr.cgroup; +} + +static void perf_event_cgroup_output(struct perf_event *event, void *data) +{ + struct perf_cgroup_event *cgroup_event = data; + struct perf_output_handle handle; + struct perf_sample_data sample; + u16 header_size = cgroup_event->event_id.header.size; + int ret; + + if (!perf_event_cgroup_match(event)) + return; + + perf_event_header__init_id(&cgroup_event->event_id.header, + &sample, event); + ret = perf_output_begin(&handle, event, + cgroup_event->event_id.header.size); + if (ret) + goto out; + + perf_output_put(&handle, cgroup_event->event_id); + __output_copy(&handle, cgroup_event->path, cgroup_event->path_size); + + perf_event__output_id_sample(event, &handle, &sample); + + perf_output_end(&handle); +out: + cgroup_event->event_id.header.size = header_size; +} + +static void perf_event_cgroup(struct cgroup *cgrp) +{ + struct perf_cgroup_event cgroup_event; + char path_enomem[16] = "//enomem"; + char *pathname; + size_t size; + + if (!atomic_read(&nr_cgroup_events)) + return; + + cgroup_event = (struct perf_cgroup_event){ + .event_id = { + .header = { + .type = PERF_RECORD_CGROUP, + .misc = 0, + .size = sizeof(cgroup_event.event_id), + }, + .id = cgroup_id(cgrp), + }, + }; + + pathname = kmalloc(PATH_MAX, GFP_KERNEL); + if (pathname == NULL) { + cgroup_event.path = path_enomem; + } else { + /* just to be sure to have enough space for alignment */ + cgroup_path(cgrp, pathname, PATH_MAX - sizeof(u64)); + cgroup_event.path = pathname; + } + + /* + * Since our buffer works in 8 byte units we need to align our string + * size to a multiple of 8. However, we must guarantee the tail end is + * zero'd out to avoid leaking random bits to userspace. + */ + size = strlen(cgroup_event.path) + 1; + while (!IS_ALIGNED(size, sizeof(u64))) + cgroup_event.path[size++] = '\0'; + + cgroup_event.event_id.header.size += size; + cgroup_event.path_size = size; + + perf_iterate_sb(perf_event_cgroup_output, + &cgroup_event, + NULL); + + kfree(pathname); +} + +#endif + +/* * mmap tracking */ @@ -7693,7 +7994,7 @@ static void perf_event_mmap_event(struct perf_mmap_event *mmap_event) flags |= MAP_EXECUTABLE; if (vma->vm_flags & VM_LOCKED) flags |= MAP_LOCKED; - if (vma->vm_flags & VM_HUGETLB) + if (is_vm_hugetlb_page(vma)) flags |= MAP_HUGETLB; if (file) { @@ -8255,23 +8556,22 @@ static void perf_event_bpf_emit_ksymbols(struct bpf_prog *prog, enum perf_bpf_event_type type) { bool unregister = type == PERF_BPF_EVENT_PROG_UNLOAD; - char sym[KSYM_NAME_LEN]; int i; if (prog->aux->func_cnt == 0) { - bpf_get_prog_name(prog, sym); perf_event_ksymbol(PERF_RECORD_KSYMBOL_TYPE_BPF, (u64)(unsigned long)prog->bpf_func, - prog->jited_len, unregister, sym); + prog->jited_len, unregister, + prog->aux->ksym.name); } else { for (i = 0; i < prog->aux->func_cnt; i++) { struct bpf_prog *subprog = prog->aux->func[i]; - bpf_get_prog_name(subprog, sym); perf_event_ksymbol( PERF_RECORD_KSYMBOL_TYPE_BPF, (u64)(unsigned long)subprog->bpf_func, - subprog->jited_len, unregister, sym); + subprog->jited_len, unregister, + prog->aux->ksym.name); } } } @@ -9206,7 +9506,6 @@ static void bpf_overflow_handler(struct perf_event *event, int ret = 0; ctx.regs = perf_arch_bpf_user_pt_regs(regs); - preempt_disable(); if (unlikely(__this_cpu_inc_return(bpf_prog_active) != 1)) goto out; rcu_read_lock(); @@ -9214,7 +9513,6 @@ static void bpf_overflow_handler(struct perf_event *event, rcu_read_unlock(); out: __this_cpu_dec(bpf_prog_active); - preempt_enable(); if (!ret) return; @@ -10349,6 +10647,9 @@ skip_type: cpuctx->online = cpumask_test_cpu(cpu, perf_online_mask); __perf_mux_hrtimer_init(cpuctx, cpu); + + cpuctx->heap_size = ARRAY_SIZE(cpuctx->heap_default); + cpuctx->heap = cpuctx->heap_default; } got_cpu_context: @@ -10616,6 +10917,8 @@ static void account_event(struct perf_event *event) atomic_inc(&nr_comm_events); if (event->attr.namespaces) atomic_inc(&nr_namespaces_events); + if (event->attr.cgroup) + atomic_inc(&nr_cgroup_events); if (event->attr.task) atomic_inc(&nr_task_events); if (event->attr.freq) @@ -10794,12 +11097,6 @@ perf_event_alloc(struct perf_event_attr *attr, int cpu, if (!has_branch_stack(event)) event->attr.branch_sample_type = 0; - if (cgroup_fd != -1) { - err = perf_cgroup_connect(cgroup_fd, event, attr, group_leader); - if (err) - goto err_ns; - } - pmu = perf_init_event(event); if (IS_ERR(pmu)) { err = PTR_ERR(pmu); @@ -10821,6 +11118,12 @@ perf_event_alloc(struct perf_event_attr *attr, int cpu, goto err_pmu; } + if (cgroup_fd != -1) { + err = perf_cgroup_connect(cgroup_fd, event, attr, group_leader); + if (err) + goto err_pmu; + } + err = exclusive_event_init(event); if (err) goto err_pmu; @@ -10881,12 +11184,12 @@ err_per_task: exclusive_event_destroy(event); err_pmu: + if (is_cgroup_event(event)) + perf_detach_cgroup(event); if (event->destroy) event->destroy(event); module_put(pmu->module); err_ns: - if (is_cgroup_event(event)) - perf_detach_cgroup(event); if (event->ns) put_pid_ns(event->ns); if (event->hw.target) @@ -10995,6 +11298,12 @@ static int perf_copy_attr(struct perf_event_attr __user *uattr, if (attr->sample_type & PERF_SAMPLE_REGS_INTR) ret = perf_reg_validate(attr->sample_regs_intr); + +#ifndef CONFIG_CGROUP_PERF + if (attr->sample_type & PERF_SAMPLE_CGROUP) + return -EINVAL; +#endif + out: return ret; @@ -11263,14 +11572,14 @@ SYSCALL_DEFINE5(perf_event_open, } if (task) { - err = mutex_lock_interruptible(&task->signal->cred_guard_mutex); + err = mutex_lock_interruptible(&task->signal->exec_update_mutex); if (err) goto err_task; /* * Reuse ptrace permission checks for now. * - * We must hold cred_guard_mutex across this and any potential + * We must hold exec_update_mutex across this and any potential * perf_install_in_context() call for this new event to * serialize against exec() altering our credentials (and the * perf_event_exit_task() that could imply). @@ -11559,7 +11868,7 @@ SYSCALL_DEFINE5(perf_event_open, mutex_unlock(&ctx->mutex); if (task) { - mutex_unlock(&task->signal->cred_guard_mutex); + mutex_unlock(&task->signal->exec_update_mutex); put_task_struct(task); } @@ -11595,7 +11904,7 @@ err_alloc: free_event(event); err_cred: if (task) - mutex_unlock(&task->signal->cred_guard_mutex); + mutex_unlock(&task->signal->exec_update_mutex); err_task: if (task) put_task_struct(task); @@ -11900,7 +12209,7 @@ static void perf_event_exit_task_context(struct task_struct *child, int ctxn) /* * When a child task exits, feed back event values to parent events. * - * Can be called with cred_guard_mutex held when called from + * Can be called with exec_update_mutex held when called from * install_exec_creds(). */ void perf_event_exit_task(struct task_struct *child) @@ -12592,6 +12901,12 @@ static void perf_cgroup_css_free(struct cgroup_subsys_state *css) kfree(jc); } +static int perf_cgroup_css_online(struct cgroup_subsys_state *css) +{ + perf_event_cgroup(css->cgroup); + return 0; +} + static int __perf_cgroup_move(void *info) { struct task_struct *task = info; @@ -12613,6 +12928,7 @@ static void perf_cgroup_attach(struct cgroup_taskset *tset) struct cgroup_subsys perf_event_cgrp_subsys = { .css_alloc = perf_cgroup_css_alloc, .css_free = perf_cgroup_css_free, + .css_online = perf_cgroup_css_online, .attach = perf_cgroup_attach, /* * Implicitly enable on dfl hierarchy so that perf events can |