diff options
Diffstat (limited to 'kernel/events')
| -rw-r--r-- | kernel/events/callchain.c | 3 | ||||
| -rw-r--r-- | kernel/events/core.c | 1789 | ||||
| -rw-r--r-- | kernel/events/hw_breakpoint.c | 193 | ||||
| -rw-r--r-- | kernel/events/internal.h | 41 | ||||
| -rw-r--r-- | kernel/events/ring_buffer.c | 142 | ||||
| -rw-r--r-- | kernel/events/uprobes.c | 1058 |
6 files changed, 2140 insertions, 1086 deletions
diff --git a/kernel/events/callchain.c b/kernel/events/callchain.c index c77206184b8..97b67df8fbf 100644 --- a/kernel/events/callchain.c +++ b/kernel/events/callchain.c @@ -116,6 +116,9 @@ int get_callchain_buffers(void) err = alloc_callchain_buffers(); exit: + if (err) + atomic_dec(&nr_callchain_events); + mutex_unlock(&callchain_mutex); return err; diff --git a/kernel/events/core.c b/kernel/events/core.c index 301079d06f2..6b17ac1b0c2 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -18,6 +18,7 @@ #include <linux/poll.h> #include <linux/slab.h> #include <linux/hash.h> +#include <linux/tick.h> #include <linux/sysfs.h> #include <linux/dcache.h> #include <linux/percpu.h> @@ -37,6 +38,9 @@ #include <linux/ftrace_event.h> #include <linux/hw_breakpoint.h> #include <linux/mm_types.h> +#include <linux/cgroup.h> +#include <linux/module.h> +#include <linux/mman.h> #include "internal.h" @@ -117,7 +121,8 @@ static int cpu_function_call(int cpu, int (*func) (void *info), void *info) #define PERF_FLAG_ALL (PERF_FLAG_FD_NO_GROUP |\ PERF_FLAG_FD_OUTPUT |\ - PERF_FLAG_PID_CGROUP) + PERF_FLAG_PID_CGROUP |\ + PERF_FLAG_FD_CLOEXEC) /* * branch priv levels that need permission checks @@ -143,6 +148,7 @@ static DEFINE_PER_CPU(atomic_t, perf_branch_stack_events); static atomic_t nr_mmap_events __read_mostly; static atomic_t nr_comm_events __read_mostly; static atomic_t nr_task_events __read_mostly; +static atomic_t nr_freq_events __read_mostly; static LIST_HEAD(pmus); static DEFINE_MUTEX(pmus_lock); @@ -163,25 +169,130 @@ int sysctl_perf_event_mlock __read_mostly = 512 + (PAGE_SIZE / 1024); /* 'free' /* * max perf event sample rate */ -#define DEFAULT_MAX_SAMPLE_RATE 100000 -int sysctl_perf_event_sample_rate __read_mostly = DEFAULT_MAX_SAMPLE_RATE; -static int max_samples_per_tick __read_mostly = - DIV_ROUND_UP(DEFAULT_MAX_SAMPLE_RATE, HZ); +#define DEFAULT_MAX_SAMPLE_RATE 100000 +#define DEFAULT_SAMPLE_PERIOD_NS (NSEC_PER_SEC / DEFAULT_MAX_SAMPLE_RATE) +#define DEFAULT_CPU_TIME_MAX_PERCENT 25 + +int sysctl_perf_event_sample_rate __read_mostly = DEFAULT_MAX_SAMPLE_RATE; + +static int max_samples_per_tick __read_mostly = DIV_ROUND_UP(DEFAULT_MAX_SAMPLE_RATE, HZ); +static int perf_sample_period_ns __read_mostly = DEFAULT_SAMPLE_PERIOD_NS; + +static int perf_sample_allowed_ns __read_mostly = + DEFAULT_SAMPLE_PERIOD_NS * DEFAULT_CPU_TIME_MAX_PERCENT / 100; + +void update_perf_cpu_limits(void) +{ + u64 tmp = perf_sample_period_ns; + + tmp *= sysctl_perf_cpu_time_max_percent; + do_div(tmp, 100); + ACCESS_ONCE(perf_sample_allowed_ns) = tmp; +} + +static int perf_rotate_context(struct perf_cpu_context *cpuctx); int perf_proc_update_handler(struct ctl_table *table, int write, void __user *buffer, size_t *lenp, loff_t *ppos) { - int ret = proc_dointvec(table, write, buffer, lenp, ppos); + int ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos); if (ret || !write) return ret; max_samples_per_tick = DIV_ROUND_UP(sysctl_perf_event_sample_rate, HZ); + perf_sample_period_ns = NSEC_PER_SEC / sysctl_perf_event_sample_rate; + update_perf_cpu_limits(); return 0; } +int sysctl_perf_cpu_time_max_percent __read_mostly = DEFAULT_CPU_TIME_MAX_PERCENT; + +int perf_cpu_time_max_percent_handler(struct ctl_table *table, int write, + void __user *buffer, size_t *lenp, + loff_t *ppos) +{ + int ret = proc_dointvec(table, write, buffer, lenp, ppos); + + if (ret || !write) + return ret; + + update_perf_cpu_limits(); + + return 0; +} + +/* + * perf samples are done in some very critical code paths (NMIs). + * If they take too much CPU time, the system can lock up and not + * get any real work done. This will drop the sample rate when + * we detect that events are taking too long. + */ +#define NR_ACCUMULATED_SAMPLES 128 +static DEFINE_PER_CPU(u64, running_sample_length); + +static void perf_duration_warn(struct irq_work *w) +{ + u64 allowed_ns = ACCESS_ONCE(perf_sample_allowed_ns); + u64 avg_local_sample_len; + u64 local_samples_len; + + local_samples_len = __get_cpu_var(running_sample_length); + avg_local_sample_len = local_samples_len/NR_ACCUMULATED_SAMPLES; + + printk_ratelimited(KERN_WARNING + "perf interrupt took too long (%lld > %lld), lowering " + "kernel.perf_event_max_sample_rate to %d\n", + avg_local_sample_len, allowed_ns >> 1, + sysctl_perf_event_sample_rate); +} + +static DEFINE_IRQ_WORK(perf_duration_work, perf_duration_warn); + +void perf_sample_event_took(u64 sample_len_ns) +{ + u64 allowed_ns = ACCESS_ONCE(perf_sample_allowed_ns); + u64 avg_local_sample_len; + u64 local_samples_len; + + if (allowed_ns == 0) + return; + + /* decay the counter by 1 average sample */ + local_samples_len = __get_cpu_var(running_sample_length); + local_samples_len -= local_samples_len/NR_ACCUMULATED_SAMPLES; + local_samples_len += sample_len_ns; + __get_cpu_var(running_sample_length) = local_samples_len; + + /* + * note: this will be biased artifically low until we have + * seen NR_ACCUMULATED_SAMPLES. Doing it this way keeps us + * from having to maintain a count. + */ + avg_local_sample_len = local_samples_len/NR_ACCUMULATED_SAMPLES; + + if (avg_local_sample_len <= allowed_ns) + return; + + if (max_samples_per_tick <= 1) + return; + + max_samples_per_tick = DIV_ROUND_UP(max_samples_per_tick, 2); + sysctl_perf_event_sample_rate = max_samples_per_tick * HZ; + perf_sample_period_ns = NSEC_PER_SEC / sysctl_perf_event_sample_rate; + + update_perf_cpu_limits(); + + if (!irq_work_queue(&perf_duration_work)) { + early_printk("perf interrupt took too long (%lld > %lld), lowering " + "kernel.perf_event_max_sample_rate to %d\n", + avg_local_sample_len, allowed_ns >> 1, + sysctl_perf_event_sample_rate); + } +} + static atomic64_t perf_event_id; static void cpu_ctx_sched_out(struct perf_cpu_context *cpuctx, @@ -194,9 +305,6 @@ static void cpu_ctx_sched_in(struct perf_cpu_context *cpuctx, static void update_context_time(struct perf_event_context *ctx); static u64 perf_event_time(struct perf_event *event); -static void ring_buffer_attach(struct perf_event *event, - struct ring_buffer *rb); - void __weak perf_event_print_debug(void) { } extern __weak const char *perf_pmu_name(void) @@ -234,6 +342,20 @@ static void perf_ctx_unlock(struct perf_cpu_context *cpuctx, #ifdef CONFIG_CGROUP_PERF /* + * perf_cgroup_info keeps track of time_enabled for a cgroup. + * This is a per-cpu dynamically allocated data structure. + */ +struct perf_cgroup_info { + u64 time; + u64 timestamp; +}; + +struct perf_cgroup { + struct cgroup_subsys_state css; + struct perf_cgroup_info __percpu *info; +}; + +/* * Must ensure cgroup is pinned (css_get) before calling * this function. In other words, we cannot call this function * if there is no cgroup event for the current CPU context. @@ -241,8 +363,8 @@ static void perf_ctx_unlock(struct perf_cpu_context *cpuctx, static inline struct perf_cgroup * perf_cgroup_from_task(struct task_struct *task) { - return container_of(task_subsys_state(task, perf_subsys_id), - struct perf_cgroup, css); + return container_of(task_css(task, perf_event_cgrp_id), + struct perf_cgroup, css); } static inline bool @@ -251,12 +373,22 @@ perf_cgroup_match(struct perf_event *event) struct perf_event_context *ctx = event->ctx; struct perf_cpu_context *cpuctx = __get_cpu_context(ctx); - return !event->cgrp || event->cgrp == cpuctx->cgrp; -} + /* @event doesn't care about cgroup */ + if (!event->cgrp) + return true; -static inline bool perf_tryget_cgroup(struct perf_event *event) -{ - return css_tryget(&event->cgrp->css); + /* wants specific cgroup scope but @cpuctx isn't associated with any */ + if (!cpuctx->cgrp) + return false; + + /* + * Cgroup scoping is recursive. An event enabled for a cgroup is + * also enabled for all its descendant cgroups. If @cpuctx's + * cgroup is a descendant of @event's (the test covers identity + * case), it's a match. + */ + return cgroup_is_descendant(cpuctx->cgrp->css.cgroup, + event->cgrp->css.cgroup); } static inline void perf_put_cgroup(struct perf_event *event) @@ -477,7 +609,8 @@ static inline int perf_cgroup_connect(int fd, struct perf_event *event, if (!f.file) return -EBADF; - css = cgroup_css_from_dir(f.file, perf_subsys_id); + css = css_tryget_online_from_dir(f.file->f_dentry, + &perf_event_cgrp_subsys); if (IS_ERR(css)) { ret = PTR_ERR(css); goto out; @@ -486,13 +619,6 @@ static inline int perf_cgroup_connect(int fd, struct perf_event *event, cgrp = container_of(css, struct perf_cgroup, css); event->cgrp = cgrp; - /* must be done before we fput() the file */ - if (!perf_tryget_cgroup(event)) { - event->cgrp = NULL; - ret = -ENOENT; - goto out; - } - /* * all events in a group must monitor * the same cgroup because a task belongs @@ -627,6 +753,106 @@ perf_cgroup_mark_enabled(struct perf_event *event, } #endif +/* + * set default to be dependent on timer tick just + * like original code + */ +#define PERF_CPU_HRTIMER (1000 / HZ) +/* + * function must be called with interrupts disbled + */ +static enum hrtimer_restart perf_cpu_hrtimer_handler(struct hrtimer *hr) +{ + struct perf_cpu_context *cpuctx; + enum hrtimer_restart ret = HRTIMER_NORESTART; + int rotations = 0; + + WARN_ON(!irqs_disabled()); + + cpuctx = container_of(hr, struct perf_cpu_context, hrtimer); + + rotations = perf_rotate_context(cpuctx); + + /* + * arm timer if needed + */ + if (rotations) { + hrtimer_forward_now(hr, cpuctx->hrtimer_interval); + ret = HRTIMER_RESTART; + } + + return ret; +} + +/* CPU is going down */ +void perf_cpu_hrtimer_cancel(int cpu) +{ + struct perf_cpu_context *cpuctx; + struct pmu *pmu; + unsigned long flags; + + if (WARN_ON(cpu != smp_processor_id())) + return; + + local_irq_save(flags); + + rcu_read_lock(); + + list_for_each_entry_rcu(pmu, &pmus, entry) { + cpuctx = this_cpu_ptr(pmu->pmu_cpu_context); + + if (pmu->task_ctx_nr == perf_sw_context) + continue; + + hrtimer_cancel(&cpuctx->hrtimer); + } + + rcu_read_unlock(); + + local_irq_restore(flags); +} + +static void __perf_cpu_hrtimer_init(struct perf_cpu_context *cpuctx, int cpu) +{ + struct hrtimer *hr = &cpuctx->hrtimer; + struct pmu *pmu = cpuctx->ctx.pmu; + int timer; + + /* no multiplexing needed for SW PMU */ + if (pmu->task_ctx_nr == perf_sw_context) + return; + + /* + * check default is sane, if not set then force to + * default interval (1/tick) + */ + timer = pmu->hrtimer_interval_ms; + if (timer < 1) + timer = pmu->hrtimer_interval_ms = PERF_CPU_HRTIMER; + + cpuctx->hrtimer_interval = ns_to_ktime(NSEC_PER_MSEC * timer); + + hrtimer_init(hr, CLOCK_MONOTONIC, HRTIMER_MODE_REL_PINNED); + hr->function = perf_cpu_hrtimer_handler; +} + +static void perf_cpu_hrtimer_restart(struct perf_cpu_context *cpuctx) +{ + struct hrtimer *hr = &cpuctx->hrtimer; + struct pmu *pmu = cpuctx->ctx.pmu; + + /* not for SW PMU */ + if (pmu->task_ctx_nr == perf_sw_context) + return; + + if (hrtimer_active(hr)) + return; + + if (!hrtimer_callback_running(hr)) + __hrtimer_start_range_ns(hr, cpuctx->hrtimer_interval, + 0, HRTIMER_MODE_REL_PINNED, 0); +} + void perf_pmu_disable(struct pmu *pmu) { int *count = this_cpu_ptr(pmu->pmu_disable_count); @@ -681,6 +907,7 @@ static void unclone_ctx(struct perf_event_context *ctx) put_ctx(ctx->parent_ctx); ctx->parent_ctx = NULL; } + ctx->generation++; } static u32 perf_event_pid(struct perf_event *event, struct task_struct *p) @@ -729,8 +956,18 @@ perf_lock_task_context(struct task_struct *task, int ctxn, unsigned long *flags) { struct perf_event_context *ctx; - rcu_read_lock(); retry: + /* + * One of the few rules of preemptible RCU is that one cannot do + * rcu_read_unlock() while holding a scheduler (or nested) lock when + * part of the read side critical section was preemptible -- see + * rcu_read_unlock_special(). + * + * Since ctx->lock nests under rq->lock we must ensure the entire read + * side critical section is non-preemptible. + */ + preempt_disable(); + rcu_read_lock(); ctx = rcu_dereference(task->perf_event_ctxp[ctxn]); if (ctx) { /* @@ -746,6 +983,8 @@ retry: raw_spin_lock_irqsave(&ctx->lock, *flags); if (ctx != rcu_dereference(task->perf_event_ctxp[ctxn])) { raw_spin_unlock_irqrestore(&ctx->lock, *flags); + rcu_read_unlock(); + preempt_enable(); goto retry; } @@ -755,6 +994,7 @@ retry: } } rcu_read_unlock(); + preempt_enable(); return ctx; } @@ -905,6 +1145,17 @@ list_add_event(struct perf_event *event, struct perf_event_context *ctx) ctx->nr_events++; if (event->attr.inherit_stat) ctx->nr_stat++; + + ctx->generation++; +} + +/* + * Initialize event state based on the perf_event_attr::disabled. + */ +static inline void perf_event__state_init(struct perf_event *event) +{ + event->state = event->attr.disabled ? PERF_EVENT_STATE_OFF : + PERF_EVENT_STATE_INACTIVE; } /* @@ -952,9 +1203,18 @@ static void perf_event__header_size(struct perf_event *event) if (sample_type & PERF_SAMPLE_PERIOD) size += sizeof(data->period); + if (sample_type & PERF_SAMPLE_WEIGHT) + size += sizeof(data->weight); + if (sample_type & PERF_SAMPLE_READ) size += event->read_size; + if (sample_type & PERF_SAMPLE_DATA_SRC) + size += sizeof(data->data_src.val); + + if (sample_type & PERF_SAMPLE_TRANSACTION) + size += sizeof(data->txn); + event->header_size = size; } @@ -970,6 +1230,9 @@ static void perf_event__id_header_size(struct perf_event *event) if (sample_type & PERF_SAMPLE_TIME) size += sizeof(data->time); + if (sample_type & PERF_SAMPLE_IDENTIFIER) + size += sizeof(data->id); + if (sample_type & PERF_SAMPLE_ID) size += sizeof(data->id); @@ -1061,6 +1324,8 @@ list_del_event(struct perf_event *event, struct perf_event_context *ctx) */ if (event->state > PERF_EVENT_STATE_OFF) event->state = PERF_EVENT_STATE_OFF; + + ctx->generation++; } static void perf_group_detach(struct perf_event *event) @@ -1139,6 +1404,8 @@ event_sched_out(struct perf_event *event, if (event->state != PERF_EVENT_STATE_ACTIVE) return; + perf_pmu_disable(event->pmu); + event->state = PERF_EVENT_STATE_INACTIVE; if (event->pending_disable) { event->pending_disable = 0; @@ -1155,6 +1422,8 @@ event_sched_out(struct perf_event *event, ctx->nr_freq--; if (event->attr.exclusive || !cpuctx->active_oncpu) cpuctx->exclusive = 0; + + perf_pmu_enable(event->pmu); } static void @@ -1177,6 +1446,11 @@ group_sched_out(struct perf_event *group_event, cpuctx->exclusive = 0; } +struct remove_event { + struct perf_event *event; + bool detach_group; +}; + /* * Cross CPU call to remove a performance event * @@ -1185,12 +1459,15 @@ group_sched_out(struct perf_event *group_event, */ static int __perf_remove_from_context(void *info) { - struct perf_event *event = info; + struct remove_event *re = info; + struct perf_event *event = re->event; struct perf_event_context *ctx = event->ctx; struct perf_cpu_context *cpuctx = __get_cpu_context(ctx); raw_spin_lock(&ctx->lock); event_sched_out(event, cpuctx, ctx); + if (re->detach_group) + perf_group_detach(event); list_del_event(event, ctx); if (!ctx->nr_events && cpuctx->task_ctx == ctx) { ctx->is_active = 0; @@ -1215,10 +1492,14 @@ static int __perf_remove_from_context(void *info) * When called from perf_event_exit_task, it's OK because the * context has been detached from its task. */ -static void perf_remove_from_context(struct perf_event *event) +static void perf_remove_from_context(struct perf_event *event, bool detach_group) { struct perf_event_context *ctx = event->ctx; struct task_struct *task = ctx->task; + struct remove_event re = { + .event = event, + .detach_group = detach_group, + }; lockdep_assert_held(&ctx->mutex); @@ -1227,12 +1508,12 @@ static void perf_remove_from_context(struct perf_event *event) * Per cpu events are removed via an smp call and * the removal is always successful. */ - cpu_function_call(event->cpu, __perf_remove_from_context, event); + cpu_function_call(event->cpu, __perf_remove_from_context, &re); return; } retry: - if (!task_function_call(task, __perf_remove_from_context, event)) + if (!task_function_call(task, __perf_remove_from_context, &re)) return; raw_spin_lock_irq(&ctx->lock); @@ -1249,6 +1530,8 @@ retry: * Since the task isn't running, its safe to remove the event, us * holding the ctx->lock ensures the task won't get scheduled in. */ + if (detach_group) + perf_group_detach(event); list_del_event(event, ctx); raw_spin_unlock_irq(&ctx->lock); } @@ -1395,6 +1678,9 @@ event_sched_in(struct perf_event *event, struct perf_event_context *ctx) { u64 tstamp = perf_event_time(event); + int ret = 0; + + lockdep_assert_held(&ctx->lock); if (event->state <= PERF_EVENT_STATE_OFF) return 0; @@ -1417,10 +1703,13 @@ event_sched_in(struct perf_event *event, */ smp_wmb(); + perf_pmu_disable(event->pmu); + if (event->pmu->add(event, PERF_EF_START)) { event->state = PERF_EVENT_STATE_INACTIVE; event->oncpu = -1; - return -EAGAIN; + ret = -EAGAIN; + goto out; } event->tstamp_running += tstamp - event->tstamp_stopped; @@ -1436,7 +1725,10 @@ event_sched_in(struct perf_event *event, if (event->attr.exclusive) cpuctx->exclusive = 1; - return 0; +out: + perf_pmu_enable(event->pmu); + + return ret; } static int @@ -1445,7 +1737,7 @@ group_sched_in(struct perf_event *group_event, struct perf_event_context *ctx) { struct perf_event *event, *partial_group = NULL; - struct pmu *pmu = group_event->pmu; + struct pmu *pmu = ctx->pmu; u64 now = ctx->time; bool simulate = false; @@ -1456,6 +1748,7 @@ group_sched_in(struct perf_event *group_event, if (event_sched_in(group_event, cpuctx, ctx)) { pmu->cancel_txn(pmu); + perf_cpu_hrtimer_restart(cpuctx); return -EAGAIN; } @@ -1502,6 +1795,8 @@ group_error: pmu->cancel_txn(pmu); + perf_cpu_hrtimer_restart(cpuctx); + return -EAGAIN; } @@ -1714,7 +2009,16 @@ static int __perf_event_enable(void *info) struct perf_cpu_context *cpuctx = __get_cpu_context(ctx); int err; - if (WARN_ON_ONCE(!ctx->is_active)) + /* + * There's a time window between 'ctx->is_active' check + * in perf_event_enable function and this place having: + * - IRQs on + * - ctx->lock unlocked + * + * where the task could be killed and 'ctx' deactivated + * by perf_event_exit_task. + */ + if (!ctx->is_active) return -EINVAL; raw_spin_lock(&ctx->lock); @@ -1757,8 +2061,10 @@ static int __perf_event_enable(void *info) * If this event can't go on and it's part of a * group, then the whole group has to come off. */ - if (leader != event) + if (leader != event) { group_sched_out(leader, cpuctx, ctx); + perf_cpu_hrtimer_restart(cpuctx); + } if (leader->attr.pinned) { update_group_times(leader); leader->state = PERF_EVENT_STATE_ERROR; @@ -1883,22 +2189,38 @@ static void ctx_sched_out(struct perf_event_context *ctx, } /* - * Test whether two contexts are equivalent, i.e. whether they - * have both been cloned from the same version of the same context - * and they both have the same number of enabled events. - * If the number of enabled events is the same, then the set - * of enabled events should be the same, because these are both - * inherited contexts, therefore we can't access individual events - * in them directly with an fd; we can only enable/disable all - * events via prctl, or enable/disable all events in a family - * via ioctl, which will have the same effect on both contexts. + * Test whether two contexts are equivalent, i.e. whether they have both been + * cloned from the same version of the same context. + * + * Equivalence is measured using a generation number in the context that is + * incremented on each modification to it; see unclone_ctx(), list_add_event() + * and list_del_event(). */ static int context_equiv(struct perf_event_context *ctx1, struct perf_event_context *ctx2) { - return ctx1->parent_ctx && ctx1->parent_ctx == ctx2->parent_ctx - && ctx1->parent_gen == ctx2->parent_gen - && !ctx1->pin_count && !ctx2->pin_count; + /* Pinning disables the swap optimization */ + if (ctx1->pin_count || ctx2->pin_count) + return 0; + + /* If ctx1 is the parent of ctx2 */ + if (ctx1 == ctx2->parent_ctx && ctx1->generation == ctx2->parent_gen) + return 1; + + /* If ctx2 is the parent of ctx1 */ + if (ctx1->parent_ctx == ctx2 && ctx1->parent_gen == ctx2->generation) + return 1; + + /* + * If ctx1 and ctx2 have the same parent; we flatten the parent + * hierarchy, see perf_event_init_context(). + */ + if (ctx1->parent_ctx && ctx1->parent_ctx == ctx2->parent_ctx && + ctx1->parent_gen == ctx2->parent_gen) + return 1; + + /* Unmatched */ + return 0; } static void __perf_event_sync_stat(struct perf_event *event, @@ -1947,9 +2269,6 @@ static void __perf_event_sync_stat(struct perf_event *event, perf_event_update_userpage(next_event); } -#define list_next_entry(pos, member) \ - list_entry(pos->member.next, typeof(*pos), member) - static void perf_event_sync_stat(struct perf_event_context *ctx, struct perf_event_context *next_ctx) { @@ -1981,7 +2300,7 @@ static void perf_event_context_sched_out(struct task_struct *task, int ctxn, { struct perf_event_context *ctx = task->perf_event_ctxp[ctxn]; struct perf_event_context *next_ctx; - struct perf_event_context *parent; + struct perf_event_context *parent, *next_parent; struct perf_cpu_context *cpuctx; int do_switch = 1; @@ -1993,10 +2312,18 @@ static void perf_event_context_sched_out(struct task_struct *task, int ctxn, return; rcu_read_lock(); - parent = rcu_dereference(ctx->parent_ctx); next_ctx = next->perf_event_ctxp[ctxn]; - if (parent && next_ctx && - rcu_dereference(next_ctx->parent_ctx) == parent) { + if (!next_ctx) + goto unlock; + + parent = rcu_dereference(ctx->parent_ctx); + next_parent = rcu_dereference(next_ctx->parent_ctx); + + /* If neither context have a parent context; they cannot be clones. */ + if (!parent || !next_parent) + goto unlock; + + if (next_parent == ctx || next_ctx == parent || next_parent == parent) { /* * Looks like the two contexts are clones, so we might be * able to optimize the context switch. We lock both @@ -2024,6 +2351,7 @@ static void perf_event_context_sched_out(struct task_struct *task, int ctxn, raw_spin_unlock(&next_ctx->lock); raw_spin_unlock(&ctx->lock); } +unlock: rcu_read_unlock(); if (do_switch) { @@ -2258,8 +2586,6 @@ static void perf_branch_stack_sched_in(struct task_struct *prev, if (cpuctx->ctx.nr_branch_stack > 0 && pmu->flush_branch_stack) { - pmu = cpuctx->ctx.pmu; - perf_ctx_lock(cpuctx, cpuctx->task_ctx); perf_pmu_disable(pmu); @@ -2450,16 +2776,18 @@ static void perf_adjust_freq_unthr_context(struct perf_event_context *ctx, if (!event_filter_match(event)) continue; + perf_pmu_disable(event->pmu); + hwc = &event->hw; - if (needs_unthr && hwc->interrupts == MAX_INTERRUPTS) { + if (hwc->interrupts == MAX_INTERRUPTS) { hwc->interrupts = 0; perf_log_throttle(event, 1); event->pmu->start(event, 0); } if (!event->attr.freq || !event->attr.sample_freq) - continue; + goto next; /* * stop the event and update event->count @@ -2481,6 +2809,8 @@ static void perf_adjust_freq_unthr_context(struct perf_event_context *ctx, perf_adjust_period(event, period, delta, false); event->pmu->start(event, delta > 0 ? PERF_EF_RELOAD : 0); + next: + perf_pmu_enable(event->pmu); } perf_pmu_enable(ctx->pmu); @@ -2505,7 +2835,7 @@ static void rotate_ctx(struct perf_event_context *ctx) * because they're strictly cpu affine and rotate_start is called with IRQs * disabled, while rotate_context is called from IRQ context. */ -static void perf_rotate_context(struct perf_cpu_context *cpuctx) +static int perf_rotate_context(struct perf_cpu_context *cpuctx) { struct perf_event_context *ctx = NULL; int rotate = 0, remove = 1; @@ -2544,8 +2874,21 @@ static void perf_rotate_context(struct perf_cpu_context *cpuctx) done: if (remove) list_del_init(&cpuctx->rotation_list); + + return rotate; } +#ifdef CONFIG_NO_HZ_FULL +bool perf_event_can_stop_tick(void) +{ + if (atomic_read(&nr_freq_events) || + __this_cpu_read(perf_throttled_count)) + return false; + else + return true; +} +#endif + void perf_event_task_tick(void) { struct list_head *head = &__get_cpu_var(rotation_list); @@ -2565,10 +2908,6 @@ void perf_event_task_tick(void) ctx = cpuctx->task_ctx; if (ctx) perf_adjust_freq_unthr_context(ctx, throttled); - - if (cpuctx->jiffies_interval == 1 || - !(jiffies % cpuctx->jiffies_interval)) - perf_rotate_context(cpuctx); } } @@ -2636,6 +2975,22 @@ out: local_irq_restore(flags); } +void perf_event_exec(void) +{ + struct perf_event_context *ctx; + int ctxn; + + rcu_read_lock(); + for_each_task_context_nr(ctxn) { + ctx = current->perf_event_ctxp[ctxn]; + if (!ctx) + continue; + + perf_event_enable_on_exec(ctx); + } + rcu_read_unlock(); +} + /* * Cross CPU call to read the hardware event */ @@ -2858,43 +3213,51 @@ static void free_event_rcu(struct rcu_head *head) } static void ring_buffer_put(struct ring_buffer *rb); +static void ring_buffer_attach(struct perf_event *event, + struct ring_buffer *rb); -static void free_event(struct perf_event *event) +static void unaccount_event_cpu(struct perf_event *event, int cpu) { - irq_work_sync(&event->pending); - - if (!event->parent) { - if (event->attach_state & PERF_ATTACH_TASK) - static_key_slow_dec_deferred(&perf_sched_events); - if (event->attr.mmap || event->attr.mmap_data) - atomic_dec(&nr_mmap_events); - if (event->attr.comm) - atomic_dec(&nr_comm_events); - if (event->attr.task) - atomic_dec(&nr_task_events); - if (event->attr.sample_type & PERF_SAMPLE_CALLCHAIN) - put_callchain_buffers(); - if (is_cgroup_event(event)) { - atomic_dec(&per_cpu(perf_cgroup_events, event->cpu)); - static_key_slow_dec_deferred(&perf_sched_events); - } + if (event->parent) + return; - if (has_branch_stack(event)) { - static_key_slow_dec_deferred(&perf_sched_events); - /* is system-wide event */ - if (!(event->attach_state & PERF_ATTACH_TASK)) - atomic_dec(&per_cpu(perf_branch_stack_events, - event->cpu)); - } + if (has_branch_stack(event)) { + if (!(event->attach_state & PERF_ATTACH_TASK)) + atomic_dec(&per_cpu(perf_branch_stack_events, cpu)); } + if (is_cgroup_event(event)) + atomic_dec(&per_cpu(perf_cgroup_events, cpu)); +} - if (event->rb) { - ring_buffer_put(event->rb); - event->rb = NULL; - } +static void unaccount_event(struct perf_event *event) +{ + if (event->parent) + return; + if (event->attach_state & PERF_ATTACH_TASK) + static_key_slow_dec_deferred(&perf_sched_events); + if (event->attr.mmap || event->attr.mmap_data) + atomic_dec(&nr_mmap_events); + if (event->attr.comm) + atomic_dec(&nr_comm_events); + if (event->attr.task) + atomic_dec(&nr_task_events); + if (event->attr.freq) + atomic_dec(&nr_freq_events); if (is_cgroup_event(event)) - perf_detach_cgroup(event); + static_key_slow_dec_deferred(&perf_sched_events); + if (has_branch_stack(event)) + static_key_slow_dec_deferred(&perf_sched_events); + + unaccount_event_cpu(event, event->cpu); +} + +static void __free_event(struct perf_event *event) +{ + if (!event->parent) { + if (event->attr.sample_type & PERF_SAMPLE_CALLCHAIN) + put_callchain_buffers(); + } if (event->destroy) event->destroy(event); @@ -2902,44 +3265,58 @@ static void free_event(struct perf_event *event) if (event->ctx) put_ctx(event->ctx); + if (event->pmu) + module_put(event->pmu->module); + call_rcu(&event->rcu_head, free_event_rcu); } -int perf_event_release_kernel(struct perf_event *event) +static void _free_event(struct perf_event *event) { - struct perf_event_context *ctx = event->ctx; + irq_work_sync(&event->pending); - WARN_ON_ONCE(ctx->parent_ctx); - /* - * There are two ways this annotation is useful: - * - * 1) there is a lock recursion from perf_event_exit_task - * see the comment there. - * - * 2) there is a lock-inversion with mmap_sem through - * perf_event_read_group(), which takes faults while - * holding ctx->mutex, however this is called after - * the last filedesc died, so there is no possibility - * to trigger the AB-BA case. - */ - mutex_lock_nested(&ctx->mutex, SINGLE_DEPTH_NESTING); - raw_spin_lock_irq(&ctx->lock); - perf_group_detach(event); - raw_spin_unlock_irq(&ctx->lock); - perf_remove_from_context(event); - mutex_unlock(&ctx->mutex); + unaccount_event(event); - free_event(event); + if (event->rb) { + /* + * Can happen when we close an event with re-directed output. + * + * Since we have a 0 refcount, perf_mmap_close() will skip + * over us; possibly making our ring_buffer_put() the last. + */ + mutex_lock(&event->mmap_mutex); + ring_buffer_attach(event, NULL); + mutex_unlock(&event->mmap_mutex); + } - return 0; + if (is_cgroup_event(event)) + perf_detach_cgroup(event); + + __free_event(event); +} + +/* + * Used to free events which have a known refcount of 1, such as in error paths + * where the event isn't exposed yet and inherited events. + */ +static void free_event(struct perf_event *event) +{ + if (WARN(atomic_long_cmpxchg(&event->refcount, 1, 0) != 1, + "unexpected event refcount: %ld; ptr=%p\n", + atomic_long_read(&event->refcount), event)) { + /* leak to avoid use-after-free */ + return; + } + + _free_event(event); } -EXPORT_SYMBOL_GPL(perf_event_release_kernel); /* * Called when the last reference to the file is gone. */ static void put_event(struct perf_event *event) { + struct perf_event_context *ctx = event->ctx; struct task_struct *owner; if (!atomic_long_dec_and_test(&event->refcount)) @@ -2978,8 +3355,32 @@ static void put_event(struct perf_event *event) put_task_struct(owner); } - perf_event_release_kernel(event); + WARN_ON_ONCE(ctx->parent_ctx); + /* + * There are two ways this annotation is useful: + * + * 1) there is a lock recursion from perf_event_exit_task + * see the comment there. + * + * 2) there is a lock-inversion with mmap_sem through + * perf_event_read_group(), which takes faults while + * holding ctx->mutex, however this is called after + * the last filedesc died, so there is no possibility + * to trigger the AB-BA case. + */ + mutex_lock_nested(&ctx->mutex, SINGLE_DEPTH_NESTING); + perf_remove_from_context(event, true); + mutex_unlock(&ctx->mutex); + + _free_event(event); +} + +int perf_event_release_kernel(struct perf_event *event) +{ + put_event(event); + return 0; } +EXPORT_SYMBOL_GPL(perf_event_release_kernel); static int perf_release(struct inode *inode, struct file *file) { @@ -3128,30 +3529,13 @@ static unsigned int perf_poll(struct file *file, poll_table *wait) unsigned int events = POLL_HUP; /* - * Race between perf_event_set_output() and perf_poll(): perf_poll() - * grabs the rb reference but perf_event_set_output() overrides it. - * Here is the timeline for two threads T1, T2: - * t0: T1, rb = rcu_dereference(event->rb) - * t1: T2, old_rb = event->rb - * t2: T2, event->rb = new rb - * t3: T2, ring_buffer_detach(old_rb) - * t4: T1, ring_buffer_attach(rb1) - * t5: T1, poll_wait(event->waitq) - * - * To avoid this problem, we grab mmap_mutex in perf_poll() - * thereby ensuring that the assignment of the new ring buffer - * and the detachment of the old buffer appear atomic to perf_poll() + * Pin the event->rb by taking event->mmap_mutex; otherwise + * perf_event_set_output() can swizzle our rb and make us miss wakeups. */ mutex_lock(&event->mmap_mutex); - - rcu_read_lock(); - rb = rcu_dereference(event->rb); - if (rb) { - ring_buffer_attach(event, rb); + rb = event->rb; + if (rb) events = atomic_xchg(&rb->poll, 0); - } - rcu_read_unlock(); - mutex_unlock(&event->mmap_mutex); poll_wait(file, &event->waitq, wait); @@ -3204,7 +3588,7 @@ static void perf_event_for_each(struct perf_event *event, static int perf_event_period(struct perf_event *event, u64 __user *arg) { struct perf_event_context *ctx = event->ctx; - int ret = 0; + int ret = 0, active; u64 value; if (!is_sampling_event(event)) @@ -3228,6 +3612,20 @@ static int perf_event_period(struct perf_event *event, u64 __user *arg) event->attr.sample_period = value; event->hw.sample_period = value; } + + active = (event->state == PERF_EVENT_STATE_ACTIVE); + if (active) { + perf_pmu_disable(ctx->pmu); + event->pmu->stop(event, PERF_EF_UPDATE); + } + + local64_set(&event->hw.period_left, 0); + + if (active) { + event->pmu->start(event, PERF_EF_RELOAD); + perf_pmu_enable(ctx->pmu); + } + unlock: raw_spin_unlock_irq(&ctx->lock); @@ -3277,6 +3675,15 @@ static long perf_ioctl(struct file *file, unsigned int cmd, unsigned long arg) case PERF_EVENT_IOC_PERIOD: return perf_event_period(event, (u64 __user *)arg); + case PERF_EVENT_IOC_ID: + { + u64 id = primary_event_id(event); + + if (copy_to_user((void __user *)arg, &id, sizeof(id))) + return -EFAULT; + return 0; + } + case PERF_EVENT_IOC_SET_OUTPUT: { int ret; @@ -3358,6 +3765,26 @@ static void calc_timer_values(struct perf_event *event, *running = ctx_time - event->tstamp_running; } +static void perf_event_init_userpage(struct perf_event *event) +{ + struct perf_event_mmap_page *userpg; + struct ring_buffer *rb; + + rcu_read_lock(); + rb = rcu_dereference(event->rb); + if (!rb) + goto unlock; + + userpg = rb->user_page; + + /* Allow new userspace to detect that bit 0 is deprecated */ + userpg->cap_bit0_is_deprecated = 1; + userpg->size = offsetof(struct perf_event_mmap_page, __reserved); + +unlock: + rcu_read_unlock(); +} + void __weak arch_perf_update_userpage(struct perf_event_mmap_page *userpg, u64 now) { } @@ -3374,6 +3801,10 @@ void perf_event_update_userpage(struct perf_event *event) u64 enabled, running, now; rcu_read_lock(); + rb = rcu_dereference(event->rb); + if (!rb) + goto unlock; + /* * compute total_time_enabled, total_time_running * based on snapshot values taken when the event @@ -3384,12 +3815,8 @@ void perf_event_update_userpage(struct perf_event *event) * NMI context */ calc_timer_values(event, &now, &enabled, &running); - rb = rcu_dereference(event->rb); - if (!rb) - goto unlock; userpg = rb->user_page; - /* * Disable preemption so as to not let the corresponding user-space * spin too long if we get preempted. @@ -3455,32 +3882,47 @@ unlock: static void ring_buffer_attach(struct perf_event *event, struct ring_buffer *rb) { + struct ring_buffer *old_rb = NULL; unsigned long flags; - if (!list_empty(&event->rb_entry)) - return; + if (event->rb) { + /* + * Should be impossible, we set this when removing + * event->rb_entry and wait/clear when adding event->rb_entry. + */ + WARN_ON_ONCE(event->rcu_pending); - spin_lock_irqsave(&rb->event_lock, flags); - if (!list_empty(&event->rb_entry)) - goto unlock; + old_rb = event->rb; + event->rcu_batches = get_state_synchronize_rcu(); + event->rcu_pending = 1; - list_add(&event->rb_entry, &rb->event_list); -unlock: - spin_unlock_irqrestore(&rb->event_lock, flags); -} + spin_lock_irqsave(&old_rb->event_lock, flags); + list_del_rcu(&event->rb_entry); + spin_unlock_irqrestore(&old_rb->event_lock, flags); + } -static void ring_buffer_detach(struct perf_event *event, - struct ring_buffer *rb) -{ - unsigned long flags; + if (event->rcu_pending && rb) { + cond_synchronize_rcu(event->rcu_batches); + event->rcu_pending = 0; + } - if (list_empty(&event->rb_entry)) - return; + if (rb) { + spin_lock_irqsave(&rb->event_lock, flags); + list_add_rcu(&event->rb_entry, &rb->event_list); + spin_unlock_irqrestore(&rb->event_lock, flags); + } + + rcu_assign_pointer(event->rb, rb); - spin_lock_irqsave(&rb->event_lock, flags); - list_del_init(&event->rb_entry); - wake_up_all(&event->waitq); - spin_unlock_irqrestore(&rb->event_lock, flags); + if (old_rb) { + ring_buffer_put(old_rb); + /* + * Since we detached before setting the new rb, so that we + * could attach the new rb, we could have missed a wakeup. + * Provide it now. + */ + wake_up_all(&event->waitq); + } } static void ring_buffer_wakeup(struct perf_event *event) @@ -3489,13 +3931,10 @@ static void ring_buffer_wakeup(struct perf_event *event) rcu_read_lock(); rb = rcu_dereference(event->rb); - if (!rb) - goto unlock; - - list_for_each_entry_rcu(event, &rb->event_list, rb_entry) - wake_up_all(&event->waitq); - -unlock: + if (rb) { + list_for_each_entry_rcu(event, &rb->event_list, rb_entry) + wake_up_all(&event->waitq); + } rcu_read_unlock(); } @@ -3524,18 +3963,10 @@ static struct ring_buffer *ring_buffer_get(struct perf_event *event) static void ring_buffer_put(struct ring_buffer *rb) { - struct perf_event *event, *n; - unsigned long flags; - if (!atomic_dec_and_test(&rb->refcount)) return; - spin_lock_irqsave(&rb->event_lock, flags); - list_for_each_entry_safe(event, n, &rb->event_list, rb_entry) { - list_del_init(&event->rb_entry); - wake_up_all(&event->waitq); - } - spin_unlock_irqrestore(&rb->event_lock, flags); + WARN_ON_ONCE(!list_empty(&rb->event_list)); call_rcu(&rb->rcu_head, rb_free_rcu); } @@ -3545,26 +3976,95 @@ static void perf_mmap_open(struct vm_area_struct *vma) struct perf_event *event = vma->vm_file->private_data; atomic_inc(&event->mmap_count); + atomic_inc(&event->rb->mmap_count); } +/* + * A buffer can be mmap()ed multiple times; either directly through the same + * event, or through other events by use of perf_event_set_output(). + * + * In order to undo the VM accounting done by perf_mmap() we need to destroy + * the buffer here, where we still have a VM context. This means we need + * to detach all events redirecting to us. + */ static void perf_mmap_close(struct vm_area_struct *vma) { struct perf_event *event = vma->vm_file->private_data; - if (atomic_dec_and_mutex_lock(&event->mmap_count, &event->mmap_mutex)) { - unsigned long size = perf_data_size(event->rb); - struct user_struct *user = event->mmap_user; - struct ring_buffer *rb = event->rb; + struct ring_buffer *rb = ring_buffer_get(event); + struct user_struct *mmap_user = rb->mmap_user; + int mmap_locked = rb->mmap_locked; + unsigned long size = perf_data_size(rb); + + atomic_dec(&rb->mmap_count); + + if (!atomic_dec_and_mutex_lock(&event->mmap_count, &event->mmap_mutex)) + goto out_put; + + ring_buffer_attach(event, NULL); + mutex_unlock(&event->mmap_mutex); + + /* If there's still other mmap()s of this buffer, we're done. */ + if (atomic_read(&rb->mmap_count)) + goto out_put; + + /* + * No other mmap()s, detach from all other events that might redirect + * into the now unreachable buffer. Somewhat complicated by the + * fact that rb::event_lock otherwise nests inside mmap_mutex. + */ +again: + rcu_read_lock(); + list_for_each_entry_rcu(event, &rb->event_list, rb_entry) { + if (!atomic_long_inc_not_zero(&event->refcount)) { + /* + * This event is en-route to free_event() which will + * detach it and remove it from the list. + */ + continue; + } + rcu_read_unlock(); + + mutex_lock(&event->mmap_mutex); + /* + * Check we didn't race with perf_event_set_output() which can + * swizzle the rb from under us while we were waiting to + * acquire mmap_mutex. + * + * If we find a different rb; ignore this event, a next + * iteration will no longer find it on the list. We have to + * still restart the iteration to make sure we're not now + * iterating the wrong list. + */ + if (event->rb == rb) + ring_buffer_attach(event, NULL); - atomic_long_sub((size >> PAGE_SHIFT) + 1, &user->locked_vm); - vma->vm_mm->pinned_vm -= event->mmap_locked; - rcu_assign_pointer(event->rb, NULL); - ring_buffer_detach(event, rb); mutex_unlock(&event->mmap_mutex); + put_event(event); - ring_buffer_put(rb); - free_uid(user); + /* + * Restart the iteration; either we're on the wrong list or + * destroyed its integrity by doing a deletion. + */ + goto again; } + rcu_read_unlock(); + + /* + * It could be there's still a few 0-ref events on the list; they'll + * get cleaned up by free_event() -- they'll also still have their + * ref on the rb and will free it whenever they are done with it. + * + * Aside from that, this buffer is 'fully' detached and unmapped, + * undo the VM accounting. + */ + + atomic_long_sub((size >> PAGE_SHIFT) + 1, &mmap_user->locked_vm); + vma->vm_mm->pinned_vm -= mmap_locked; + free_uid(mmap_user); + +out_put: + ring_buffer_put(rb); /* could be last */ } static const struct vm_operations_struct perf_mmap_vmops = { @@ -3614,12 +4114,24 @@ static int perf_mmap(struct file *file, struct vm_area_struct *vma) return -EINVAL; WARN_ON_ONCE(event->ctx->parent_ctx); +again: mutex_lock(&event->mmap_mutex); if (event->rb) { - if (event->rb->nr_pages == nr_pages) - atomic_inc(&event->rb->refcount); - else + if (event->rb->nr_pages != nr_pages) { ret = -EINVAL; + goto unlock; + } + + if (!atomic_inc_not_zero(&event->rb->mmap_count)) { + /* + * Raced against perf_mmap_close() through + * perf_event_set_output(). Try again, hope for better + * luck. + */ + mutex_unlock(&event->mmap_mutex); + goto again; + } + goto unlock; } @@ -3660,13 +4172,17 @@ static int perf_mmap(struct file *file, struct vm_area_struct *vma) ret = -ENOMEM; goto unlock; } - rcu_assign_pointer(event->rb, rb); + + atomic_set(&rb->mmap_count, 1); + rb->mmap_locked = extra; + rb->mmap_user = get_current_user(); atomic_long_add(user_extra, &user->locked_vm); - event->mmap_locked = extra; - event->mmap_user = get_current_user(); - vma->vm_mm->pinned_vm += event->mmap_locked; + vma->vm_mm->pinned_vm += extra; + + ring_buffer_attach(event, rb); + perf_event_init_userpage(event); perf_event_update_userpage(event); unlock: @@ -3674,7 +4190,11 @@ unlock: atomic_inc(&event->mmap_count); mutex_unlock(&event->mmap_mutex); - vma->vm_flags |= VM_DONTEXPAND | VM_DONTDUMP; + /* + * Since pinned accounting is per vm we cannot allow fork() to copy our + * vma. + */ + vma->vm_flags |= VM_DONTCOPY | VM_DONTEXPAND | VM_DONTDUMP; vma->vm_ops = &perf_mmap_vmops; return ret; @@ -3682,7 +4202,7 @@ unlock: static int perf_fasync(int fd, struct file *filp, int on) { - struct inode *inode = filp->f_path.dentry->d_inode; + struct inode *inode = file_inode(filp); struct perf_event *event = filp->private_data; int retval; @@ -3905,7 +4425,7 @@ static void __perf_event_header__init_id(struct perf_event_header *header, if (sample_type & PERF_SAMPLE_TIME) data->time = perf_clock(); - if (sample_type & PERF_SAMPLE_ID) + if (sample_type & (PERF_SAMPLE_ID | PERF_SAMPLE_IDENTIFIER)) data->id = primary_event_id(event); if (sample_type & PERF_SAMPLE_STREAM_ID) @@ -3944,6 +4464,9 @@ static void __perf_event__output_id_sample(struct perf_output_handle *handle, if (sample_type & PERF_SAMPLE_CPU) perf_output_put(handle, data->cpu_entry); + + if (sample_type & PERF_SAMPLE_IDENTIFIER) + perf_output_put(handle, data->id); } void perf_event__output_id_sample(struct perf_event *event, @@ -4009,7 +4532,8 @@ static void perf_output_read_group(struct perf_output_handle *handle, list_for_each_entry(sub, &leader->sibling_list, group_entry) { n = 0; - if (sub != event) + if ((sub != event) && + (sub->state == PERF_EVENT_STATE_ACTIVE)) sub->pmu->read(sub); values[n++] = perf_event_count(sub); @@ -4056,6 +4580,9 @@ void perf_output_sample(struct perf_output_handle *handle, perf_output_put(handle, *header); + if (sample_type & PERF_SAMPLE_IDENTIFIER) + perf_output_put(handle, data->id); + if (sample_type & PERF_SAMPLE_IP) perf_output_put(handle, data->ip); @@ -4116,20 +4643,6 @@ void perf_output_sample(struct perf_output_handle *handle, } } - if (!event->attr.watermark) { - int wakeup_events = event->attr.wakeup_events; - - if (wakeup_events) { - struct ring_buffer *rb = handle->rb; - int events = local_inc_return(&rb->events); - - if (events >= wakeup_events) { - local_sub(wakeup_events, &rb->events); - local_inc(&rb->wakeup); - } - } - } - if (sample_type & PERF_SAMPLE_BRANCH_STACK) { if (data->br_stack) { size_t size; @@ -4165,10 +4678,34 @@ void perf_output_sample(struct perf_output_handle *handle, } } - if (sample_type & PERF_SAMPLE_STACK_USER) + if (sample_type & PERF_SAMPLE_STACK_USER) { perf_output_sample_ustack(handle, data->stack_user_size, data->regs_user.regs); + } + + if (sample_type & PERF_SAMPLE_WEIGHT) + perf_output_put(handle, data->weight); + + if (sample_type & PERF_SAMPLE_DATA_SRC) + perf_output_put(handle, data->data_src.val); + + if (sample_type & PERF_SAMPLE_TRANSACTION) + perf_output_put(handle, data->txn); + + if (!event->attr.watermark) { + int wakeup_events = event->attr.wakeup_events; + + if (wakeup_events) { + struct ring_buffer *rb = handle->rb; + int events = local_inc_return(&rb->events); + + if (events >= wakeup_events) { + local_sub(wakeup_events, &rb->events); + local_inc(&rb->wakeup); + } + } + } } void perf_prepare_sample(struct perf_event_header *header, @@ -4328,10 +4865,63 @@ perf_event_read_event(struct perf_event *event, perf_output_end(&handle); } +typedef void (perf_event_aux_output_cb)(struct perf_event *event, void *data); + +static void +perf_event_aux_ctx(struct perf_event_context *ctx, + perf_event_aux_output_cb output, + void *data) +{ + struct perf_event *event; + + list_for_each_entry_rcu(event, &ctx->event_list, event_entry) { + if (event->state < PERF_EVENT_STATE_INACTIVE) + continue; + if (!event_filter_match(event)) + continue; + output(event, data); + } +} + +static void +perf_event_aux(perf_event_aux_output_cb output, void *data, + struct perf_event_context *task_ctx) +{ + struct perf_cpu_context *cpuctx; + struct perf_event_context *ctx; + struct pmu *pmu; + int ctxn; + + rcu_read_lock(); + list_for_each_entry_rcu(pmu, &pmus, entry) { + cpuctx = get_cpu_ptr(pmu->pmu_cpu_context); + if (cpuctx->unique_pmu != pmu) + goto next; + perf_event_aux_ctx(&cpuctx->ctx, output, data); + if (task_ctx) + goto next; + ctxn = pmu->task_ctx_nr; + if (ctxn < 0) + goto next; + ctx = rcu_dereference(current->perf_event_ctxp[ctxn]); + if (ctx) + perf_event_aux_ctx(ctx, output, data); +next: + put_cpu_ptr(pmu->pmu_cpu_context); + } + + if (task_ctx) { + preempt_disable(); + perf_event_aux_ctx(task_ctx, output, data); + preempt_enable(); + } + rcu_read_unlock(); +} + /* * task tracking -- fork/exit * - * enabled by: attr.comm | attr.mmap | attr.mmap_data | attr.task + * enabled by: attr.comm | attr.mmap | attr.mmap2 | attr.mmap_data | attr.task */ struct perf_task_event { @@ -4349,14 +4939,25 @@ struct perf_task_event { } event_id; }; +static int perf_event_task_match(struct perf_event *event) +{ + return event->attr.comm || event->attr.mmap || + event->attr.mmap2 || event->attr.mmap_data || + event->attr.task; +} + static void perf_event_task_output(struct perf_event *event, - struct perf_task_event *task_event) + void *data) { + struct perf_task_event *task_event = data; struct perf_output_handle handle; struct perf_sample_data sample; struct task_struct *task = task_event->task; int ret, size = task_event->event_id.header.size; + if (!perf_event_task_match(event)) + return; + perf_event_header__init_id(&task_event->event_id.header, &sample, event); ret = perf_output_begin(&handle, event, @@ -4379,61 +4980,6 @@ out: task_event->event_id.header.size = size; } -static int perf_event_task_match(struct perf_event *event) -{ - if (event->state < PERF_EVENT_STATE_INACTIVE) - return 0; - - if (!event_filter_match(event)) - return 0; - - if (event->attr.comm || event->attr.mmap || - event->attr.mmap_data || event->attr.task) - return 1; - - return 0; -} - -static void perf_event_task_ctx(struct perf_event_context *ctx, - struct perf_task_event *task_event) -{ - struct perf_event *event; - - list_for_each_entry_rcu(event, &ctx->event_list, event_entry) { - if (perf_event_task_match(event)) - perf_event_task_output(event, task_event); - } -} - -static void perf_event_task_event(struct perf_task_event *task_event) -{ - struct perf_cpu_context *cpuctx; - struct perf_event_context *ctx; - struct pmu *pmu; - int ctxn; - - rcu_read_lock(); - list_for_each_entry_rcu(pmu, &pmus, entry) { - cpuctx = get_cpu_ptr(pmu->pmu_cpu_context); - if (cpuctx->unique_pmu != pmu) - goto next; - perf_event_task_ctx(&cpuctx->ctx, task_event); - - ctx = task_event->task_ctx; - if (!ctx) { - ctxn = pmu->task_ctx_nr; - if (ctxn < 0) - goto next; - ctx = rcu_dereference(current->perf_event_ctxp[ctxn]); - } - if (ctx) - perf_event_task_ctx(ctx, task_event); -next: - put_cpu_ptr(pmu->pmu_cpu_context); - } - rcu_read_unlock(); -} - static void perf_event_task(struct task_struct *task, struct perf_event_context *task_ctx, int new) @@ -4462,7 +5008,9 @@ static void perf_event_task(struct task_struct *task, }, }; - perf_event_task_event(&task_event); + perf_event_aux(perf_event_task_output, + &task_event, + task_ctx); } void perf_event_fork(struct task_struct *task) @@ -4487,14 +5035,23 @@ struct perf_comm_event { } event_id; }; +static int perf_event_comm_match(struct perf_event *event) +{ + return event->attr.comm; +} + static void perf_event_comm_output(struct perf_event *event, - struct perf_comm_event *comm_event) + void *data) { + struct perf_comm_event *comm_event = data; struct perf_output_handle handle; struct perf_sample_data sample; int size = comm_event->event_id.header.size; int ret; + if (!perf_event_comm_match(event)) + return; + perf_event_header__init_id(&comm_event->event_id.header, &sample, event); ret = perf_output_begin(&handle, event, comm_event->event_id.header.size); @@ -4516,39 +5073,10 @@ out: comm_event->event_id.header.size = size; } -static int perf_event_comm_match(struct perf_event *event) -{ - if (event->state < PERF_EVENT_STATE_INACTIVE) - return 0; - - if (!event_filter_match(event)) - return 0; - - if (event->attr.comm) - return 1; - - return 0; -} - -static void perf_event_comm_ctx(struct perf_event_context *ctx, - struct perf_comm_event *comm_event) -{ - struct perf_event *event; - - list_for_each_entry_rcu(event, &ctx->event_list, event_entry) { - if (perf_event_comm_match(event)) - perf_event_comm_output(event, comm_event); - } -} - static void perf_event_comm_event(struct perf_comm_event *comm_event) { - struct perf_cpu_context *cpuctx; - struct perf_event_context *ctx; char comm[TASK_COMM_LEN]; unsigned int size; - struct pmu *pmu; - int ctxn; memset(comm, 0, sizeof(comm)); strlcpy(comm, comm_event->task->comm, sizeof(comm)); @@ -4558,39 +5086,15 @@ static void perf_event_comm_event(struct perf_comm_event *comm_event) comm_event->comm_size = size; comm_event->event_id.header.size = sizeof(comm_event->event_id) + size; - rcu_read_lock(); - list_for_each_entry_rcu(pmu, &pmus, entry) { - cpuctx = get_cpu_ptr(pmu->pmu_cpu_context); - if (cpuctx->unique_pmu != pmu) - goto next; - perf_event_comm_ctx(&cpuctx->ctx, comm_event); - - ctxn = pmu->task_ctx_nr; - if (ctxn < 0) - goto next; - ctx = rcu_dereference(current->perf_event_ctxp[ctxn]); - if (ctx) - perf_event_comm_ctx(ctx, comm_event); -next: - put_cpu_ptr(pmu->pmu_cpu_context); - } - rcu_read_unlock(); + perf_event_aux(perf_event_comm_output, + comm_event, + NULL); } -void perf_event_comm(struct task_struct *task) +void perf_event_comm(struct task_struct *task, bool exec) { struct perf_comm_event comm_event; - struct perf_event_context *ctx; - int ctxn; - - for_each_task_context_nr(ctxn) { - ctx = task->perf_event_ctxp[ctxn]; - if (!ctx) - continue; - - perf_event_enable_on_exec(ctx); - } if (!atomic_read(&nr_comm_events)) return; @@ -4602,7 +5106,7 @@ void perf_event_comm(struct task_struct *task) .event_id = { .header = { .type = PERF_RECORD_COMM, - .misc = 0, + .misc = exec ? PERF_RECORD_MISC_COMM_EXEC : 0, /* .size */ }, /* .pid */ @@ -4622,6 +5126,10 @@ struct perf_mmap_event { const char *file_name; int file_size; + int maj, min; + u64 ino; + u64 ino_generation; + u32 prot, flags; struct { struct perf_event_header header; @@ -4634,14 +5142,39 @@ struct perf_mmap_event { } event_id; }; +static int perf_event_mmap_match(struct perf_event *event, + void *data) +{ + struct perf_mmap_event *mmap_event = data; + struct vm_area_struct *vma = mmap_event->vma; + int executable = vma->vm_flags & VM_EXEC; + + return (!executable && event->attr.mmap_data) || + (executable && (event->attr.mmap || event->attr.mmap2)); +} + static void perf_event_mmap_output(struct perf_event *event, - struct perf_mmap_event *mmap_event) + void *data) { + struct perf_mmap_event *mmap_event = data; struct perf_output_handle handle; struct perf_sample_data sample; int size = mmap_event->event_id.header.size; int ret; + if (!perf_event_mmap_match(event, data)) + return; + + if (event->attr.mmap2) { + mmap_event->event_id.header.type = PERF_RECORD_MMAP2; + mmap_event->event_id.header.size += sizeof(mmap_event->maj); + mmap_event->event_id.header.size += sizeof(mmap_event->min); + mmap_event->event_id.header.size += sizeof(mmap_event->ino); + mmap_event->event_id.header.size += sizeof(mmap_event->ino_generation); + mmap_event->event_id.header.size += sizeof(mmap_event->prot); + mmap_event->event_id.header.size += sizeof(mmap_event->flags); + } + perf_event_header__init_id(&mmap_event->event_id.header, &sample, event); ret = perf_output_begin(&handle, event, mmap_event->event_id.header.size); @@ -4652,6 +5185,16 @@ static void perf_event_mmap_output(struct perf_event *event, mmap_event->event_id.tid = perf_event_tid(event, current); perf_output_put(&handle, mmap_event->event_id); + + if (event->attr.mmap2) { + perf_output_put(&handle, mmap_event->maj); + perf_output_put(&handle, mmap_event->min); + perf_output_put(&handle, mmap_event->ino); + perf_output_put(&handle, mmap_event->ino_generation); + perf_output_put(&handle, mmap_event->prot); + perf_output_put(&handle, mmap_event->flags); + } + __output_copy(&handle, mmap_event->file_name, mmap_event->file_size); @@ -4662,119 +5205,116 @@ out: mmap_event->event_id.header.size = size; } -static int perf_event_mmap_match(struct perf_event *event, - struct perf_mmap_event *mmap_event, - int executable) -{ - if (event->state < PERF_EVENT_STATE_INACTIVE) - return 0; - - if (!event_filter_match(event)) - return 0; - - if ((!executable && event->attr.mmap_data) || - (executable && event->attr.mmap)) - return 1; - - return 0; -} - -static void perf_event_mmap_ctx(struct perf_event_context *ctx, - struct perf_mmap_event *mmap_event, - int executable) -{ - struct perf_event *event; - - list_for_each_entry_rcu(event, &ctx->event_list, event_entry) { - if (perf_event_mmap_match(event, mmap_event, executable)) - perf_event_mmap_output(event, mmap_event); - } -} - static void perf_event_mmap_event(struct perf_mmap_event *mmap_event) { - struct perf_cpu_context *cpuctx; - struct perf_event_context *ctx; struct vm_area_struct *vma = mmap_event->vma; struct file *file = vma->vm_file; + int maj = 0, min = 0; + u64 ino = 0, gen = 0; + u32 prot = 0, flags = 0; unsigned int size; char tmp[16]; char *buf = NULL; - const char *name; - struct pmu *pmu; - int ctxn; - - memset(tmp, 0, sizeof(tmp)); + char *name; if (file) { + struct inode *inode; + dev_t dev; + + buf = kmalloc(PATH_MAX, GFP_KERNEL); + if (!buf) { + name = "//enomem"; + goto cpy_name; + } /* - * d_path works from the end of the rb backwards, so we + * d_path() works from the end of the rb backwards, so we * need to add enough zero bytes after the string to handle * the 64bit alignment we do later. */ - buf = kzalloc(PATH_MAX + sizeof(u64), GFP_KERNEL); - if (!buf) { - name = strncpy(tmp, "//enomem", sizeof(tmp)); - goto got_name; - } - name = d_path(&file->f_path, buf, PATH_MAX); + name = d_path(&file->f_path, buf, PATH_MAX - sizeof(u64)); if (IS_ERR(name)) { - name = strncpy(tmp, "//toolong", sizeof(tmp)); - goto got_name; + name = "//toolong"; + goto cpy_name; } + inode = file_inode(vma->vm_file); + dev = inode->i_sb->s_dev; + ino = inode->i_ino; + gen = inode->i_generation; + maj = MAJOR(dev); + min = MINOR(dev); + + if (vma->vm_flags & VM_READ) + prot |= PROT_READ; + if (vma->vm_flags & VM_WRITE) + prot |= PROT_WRITE; + if (vma->vm_flags & VM_EXEC) + prot |= PROT_EXEC; + + if (vma->vm_flags & VM_MAYSHARE) + flags = MAP_SHARED; + else + flags = MAP_PRIVATE; + + if (vma->vm_flags & VM_DENYWRITE) + flags |= MAP_DENYWRITE; + if (vma->vm_flags & VM_MAYEXEC) + flags |= MAP_EXECUTABLE; + if (vma->vm_flags & VM_LOCKED) + flags |= MAP_LOCKED; + if (vma->vm_flags & VM_HUGETLB) + flags |= MAP_HUGETLB; + + goto got_name; } else { - if (arch_vma_name(mmap_event->vma)) { - name = strncpy(tmp, arch_vma_name(mmap_event->vma), - sizeof(tmp)); - goto got_name; - } + name = (char *)arch_vma_name(vma); + if (name) + goto cpy_name; - if (!vma->vm_mm) { - name = strncpy(tmp, "[vdso]", sizeof(tmp)); - goto got_name; - } else if (vma->vm_start <= vma->vm_mm->start_brk && + if (vma->vm_start <= vma->vm_mm->start_brk && vma->vm_end >= vma->vm_mm->brk) { - name = strncpy(tmp, "[heap]", sizeof(tmp)); - goto got_name; - } else if (vma->vm_start <= vma->vm_mm->start_stack && + name = "[heap]"; + goto cpy_name; + } + if (vma->vm_start <= vma->vm_mm->start_stack && vma->vm_end >= vma->vm_mm->start_stack) { - name = strncpy(tmp, "[stack]", sizeof(tmp)); - goto got_name; + name = "[stack]"; + goto cpy_name; } - name = strncpy(tmp, "//anon", sizeof(tmp)); - goto got_name; + name = "//anon"; + goto cpy_name; } +cpy_name: + strlcpy(tmp, name, sizeof(tmp)); + name = tmp; got_name: - size = ALIGN(strlen(name)+1, sizeof(u64)); + /* + * 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(name)+1; + while (!IS_ALIGNED(size, sizeof(u64))) + name[size++] = '\0'; mmap_event->file_name = name; mmap_event->file_size = size; + mmap_event->maj = maj; + mmap_event->min = min; + mmap_event->ino = ino; + mmap_event->ino_generation = gen; + mmap_event->prot = prot; + mmap_event->flags = flags; - mmap_event->event_id.header.size = sizeof(mmap_event->event_id) + size; + if (!(vma->vm_flags & VM_EXEC)) + mmap_event->event_id.header.misc |= PERF_RECORD_MISC_MMAP_DATA; - rcu_read_lock(); - list_for_each_entry_rcu(pmu, &pmus, entry) { - cpuctx = get_cpu_ptr(pmu->pmu_cpu_context); - if (cpuctx->unique_pmu != pmu) - goto next; - perf_event_mmap_ctx(&cpuctx->ctx, mmap_event, - vma->vm_flags & VM_EXEC); - - ctxn = pmu->task_ctx_nr; - if (ctxn < 0) - goto next; + mmap_event->event_id.header.size = sizeof(mmap_event->event_id) + size; - ctx = rcu_dereference(current->perf_event_ctxp[ctxn]); - if (ctx) { - perf_event_mmap_ctx(ctx, mmap_event, - vma->vm_flags & VM_EXEC); - } -next: - put_cpu_ptr(pmu->pmu_cpu_context); - } - rcu_read_unlock(); + perf_event_aux(perf_event_mmap_output, + mmap_event, + NULL); kfree(buf); } @@ -4802,6 +5342,12 @@ void perf_event_mmap(struct vm_area_struct *vma) .len = vma->vm_end - vma->vm_start, .pgoff = (u64)vma->vm_pgoff << PAGE_SHIFT, }, + /* .maj (attr_mmap2 only) */ + /* .min (attr_mmap2 only) */ + /* .ino (attr_mmap2 only) */ + /* .ino_generation (attr_mmap2 only) */ + /* .prot (attr_mmap2 only) */ + /* .flags (attr_mmap2 only) */ }; perf_event_mmap_event(&mmap_event); @@ -4879,6 +5425,7 @@ static int __perf_event_overflow(struct perf_event *event, __this_cpu_inc(perf_throttled_count); hwc->interrupts = MAX_INTERRUPTS; perf_log_throttle(event, 0); + tick_nohz_full_kick(); ret = 1; } } @@ -4937,6 +5484,9 @@ struct swevent_htable { /* Recursion avoidance in each contexts */ int recursion[PERF_NR_CONTEXTS]; + + /* Keeps track of cpu being initialized/exited */ + bool online; }; static DEFINE_PER_CPU(struct swevent_htable, swevent_htable); @@ -4948,7 +5498,7 @@ static DEFINE_PER_CPU(struct swevent_htable, swevent_htable); * sign as trigger. */ -static u64 perf_swevent_set_period(struct perf_event *event) +u64 perf_swevent_set_period(struct perf_event *event) { struct hw_perf_event *hwc = &event->hw; u64 period = hwc->last_period; @@ -5117,7 +5667,6 @@ static void do_perf_sw_event(enum perf_type_id type, u32 event_id, { struct swevent_htable *swhash = &__get_cpu_var(swevent_htable); struct perf_event *event; - struct hlist_node *node; struct hlist_head *head; rcu_read_lock(); @@ -5125,7 +5674,7 @@ static void do_perf_sw_event(enum perf_type_id type, u32 event_id, if (!head) goto end; - hlist_for_each_entry_rcu(event, node, head, hlist_entry) { + hlist_for_each_entry_rcu(event, head, hlist_entry) { if (perf_swevent_match(event, type, event_id, data, regs)) perf_swevent_event(event, nr, data, regs); } @@ -5184,8 +5733,14 @@ static int perf_swevent_add(struct perf_event *event, int flags) hwc->state = !(flags & PERF_EF_START); head = find_swevent_head(swhash, event); - if (WARN_ON_ONCE(!head)) + if (!head) { + /* + * We can race with cpu hotplug code. Do not + * WARN if the cpu just got unplugged. + */ + WARN_ON_ONCE(swhash->online); return -EINVAL; + } hlist_add_head_rcu(&event->hlist_entry, head); @@ -5242,11 +5797,6 @@ static void swevent_hlist_put(struct perf_event *event) { int cpu; - if (event->cpu != -1) { - swevent_hlist_put_cpu(event, event->cpu); - return; - } - for_each_possible_cpu(cpu) swevent_hlist_put_cpu(event, cpu); } @@ -5280,9 +5830,6 @@ static int swevent_hlist_get(struct perf_event *event) int err; int cpu, failed_cpu; - if (event->cpu != -1) - return swevent_hlist_get_cpu(event, event->cpu); - get_online_cpus(); for_each_possible_cpu(cpu) { err = swevent_hlist_get_cpu(event, cpu); @@ -5319,7 +5866,7 @@ static void sw_perf_event_destroy(struct perf_event *event) static int perf_swevent_init(struct perf_event *event) { - int event_id = event->attr.config; + u64 event_id = event->attr.config; if (event->attr.type != PERF_TYPE_SOFTWARE) return -ENOENT; @@ -5410,7 +5957,6 @@ void perf_tp_event(u64 addr, u64 count, void *record, int entry_size, { struct perf_sample_data data; struct perf_event *event; - struct hlist_node *node; struct perf_raw_record raw = { .size = entry_size, @@ -5420,7 +5966,7 @@ void perf_tp_event(u64 addr, u64 count, void *record, int entry_size, perf_sample_data_init(&data, addr, 0); data.raw = &raw; - hlist_for_each_entry_rcu(event, node, head, hlist_entry) { + hlist_for_each_entry_rcu(event, head, hlist_entry) { if (perf_tp_event_match(event, &data, regs)) perf_swevent_event(event, count, &data, regs); } @@ -5640,6 +6186,7 @@ static void perf_swevent_init_hrtimer(struct perf_event *event) event->attr.sample_period = NSEC_PER_SEC / freq; hwc->sample_period = event->attr.sample_period; local64_set(&hwc->period_left, hwc->sample_period); + hwc->last_period = hwc->sample_period; event->attr.freq = 0; } } @@ -5834,7 +6381,7 @@ static int perf_event_idx_default(struct perf_event *event) * Ensures all contexts with the same task_ctx_nr have the same * pmu_cpu_context too. */ -static void *find_pmu_context(int ctxn) +static struct perf_cpu_context __percpu *find_pmu_context(int ctxn) { struct pmu *pmu; @@ -5891,16 +6438,64 @@ type_show(struct device *dev, struct device_attribute *attr, char *page) return snprintf(page, PAGE_SIZE-1, "%d\n", pmu->type); } +static DEVICE_ATTR_RO(type); + +static ssize_t +perf_event_mux_interval_ms_show(struct device *dev, + struct device_attribute *attr, + char *page) +{ + struct pmu *pmu = dev_get_drvdata(dev); + + return snprintf(page, PAGE_SIZE-1, "%d\n", pmu->hrtimer_interval_ms); +} + +static ssize_t +perf_event_mux_interval_ms_store(struct device *dev, + struct device_attribute *attr, + const char *buf, size_t count) +{ + struct pmu *pmu = dev_get_drvdata(dev); + int timer, cpu, ret; + + ret = kstrtoint(buf, 0, &timer); + if (ret) + return ret; + + if (timer < 1) + return -EINVAL; + + /* same value, noting to do */ + if (timer == pmu->hrtimer_interval_ms) + return count; + + pmu->hrtimer_interval_ms = timer; + + /* update all cpuctx for this PMU */ + for_each_possible_cpu(cpu) { + struct perf_cpu_context *cpuctx; + cpuctx = per_cpu_ptr(pmu->pmu_cpu_context, cpu); + cpuctx->hrtimer_interval = ns_to_ktime(NSEC_PER_MSEC * timer); + + if (hrtimer_active(&cpuctx->hrtimer)) + hrtimer_forward_now(&cpuctx->hrtimer, cpuctx->hrtimer_interval); + } + + return count; +} +static DEVICE_ATTR_RW(perf_event_mux_interval_ms); -static struct device_attribute pmu_dev_attrs[] = { - __ATTR_RO(type), - __ATTR_NULL, +static struct attribute *pmu_dev_attrs[] = { + &dev_attr_type.attr, + &dev_attr_perf_event_mux_interval_ms.attr, + NULL, }; +ATTRIBUTE_GROUPS(pmu_dev); static int pmu_bus_running; static struct bus_type pmu_bus = { .name = "event_source", - .dev_attrs = pmu_dev_attrs, + .dev_groups = pmu_dev_groups, }; static void pmu_dev_release(struct device *dev) @@ -5940,7 +6535,7 @@ free_dev: static struct lock_class_key cpuctx_mutex; static struct lock_class_key cpuctx_lock; -int perf_pmu_register(struct pmu *pmu, char *name, int type) +int perf_pmu_register(struct pmu *pmu, const char *name, int type) { int cpu, ret; @@ -5956,13 +6551,9 @@ int perf_pmu_register(struct pmu *pmu, char *name, int type) pmu->name = name; if (type < 0) { - int err = idr_pre_get(&pmu_idr, GFP_KERNEL); - if (!err) - goto free_pdc; - - err = idr_get_new_above(&pmu_idr, pmu, PERF_TYPE_MAX, &type); - if (err) { - ret = err; + type = idr_alloc(&pmu_idr, pmu, PERF_TYPE_MAX, 0, GFP_KERNEL); + if (type < 0) { + ret = type; goto free_pdc; } } @@ -5979,6 +6570,7 @@ skip_type: if (pmu->pmu_cpu_context) goto got_cpu_context; + ret = -ENOMEM; pmu->pmu_cpu_context = alloc_percpu(struct perf_cpu_context); if (!pmu->pmu_cpu_context) goto free_dev; @@ -5992,7 +6584,9 @@ skip_type: lockdep_set_class(&cpuctx->ctx.lock, &cpuctx_lock); cpuctx->ctx.type = cpu_context; cpuctx->ctx.pmu = pmu; - cpuctx->jiffies_interval = 1; + + __perf_cpu_hrtimer_init(cpuctx, cpu); + INIT_LIST_HEAD(&cpuctx->rotation_list); cpuctx->unique_pmu = pmu; } @@ -6042,6 +6636,7 @@ free_pdc: free_percpu(pmu->pmu_disable_count); goto unlock; } +EXPORT_SYMBOL_GPL(perf_pmu_register); void perf_pmu_unregister(struct pmu *pmu) { @@ -6063,6 +6658,7 @@ void perf_pmu_unregister(struct pmu *pmu) put_device(pmu->dev); free_pmu_context(pmu); } +EXPORT_SYMBOL_GPL(perf_pmu_unregister); struct pmu *perf_init_event(struct perf_event *event) { @@ -6076,6 +6672,10 @@ struct pmu *perf_init_event(struct perf_event *event) pmu = idr_find(&pmu_idr, event->attr.type); rcu_read_unlock(); if (pmu) { + if (!try_module_get(pmu->module)) { + pmu = ERR_PTR(-ENODEV); + goto unlock; + } event->pmu = pmu; ret = pmu->event_init(event); if (ret) @@ -6084,6 +6684,10 @@ struct pmu *perf_init_event(struct perf_event *event) } list_for_each_entry_rcu(pmu, &pmus, entry) { + if (!try_module_get(pmu->module)) { + pmu = ERR_PTR(-ENODEV); + goto unlock; + } event->pmu = pmu; ret = pmu->event_init(event); if (!ret) @@ -6101,6 +6705,44 @@ unlock: return pmu; } +static void account_event_cpu(struct perf_event *event, int cpu) +{ + if (event->parent) + return; + + if (has_branch_stack(event)) { + if (!(event->attach_state & PERF_ATTACH_TASK)) + atomic_inc(&per_cpu(perf_branch_stack_events, cpu)); + } + if (is_cgroup_event(event)) + atomic_inc(&per_cpu(perf_cgroup_events, cpu)); +} + +static void account_event(struct perf_event *event) +{ + if (event->parent) + return; + + if (event->attach_state & PERF_ATTACH_TASK) + static_key_slow_inc(&perf_sched_events.key); + if (event->attr.mmap || event->attr.mmap_data) + atomic_inc(&nr_mmap_events); + if (event->attr.comm) + atomic_inc(&nr_comm_events); + if (event->attr.task) + atomic_inc(&nr_task_events); + if (event->attr.freq) { + if (atomic_inc_return(&nr_freq_events) == 1) + tick_nohz_full_kick_all(); + } + if (has_branch_stack(event)) + static_key_slow_inc(&perf_sched_events.key); + if (is_cgroup_event(event)) + static_key_slow_inc(&perf_sched_events.key); + + account_event_cpu(event, event->cpu); +} + /* * Allocate and initialize a event structure */ @@ -6115,7 +6757,7 @@ perf_event_alloc(struct perf_event_attr *attr, int cpu, struct pmu *pmu; struct perf_event *event; struct hw_perf_event *hwc; - long err; + long err = -EINVAL; if ((unsigned)cpu >= nr_cpu_ids) { if (!task || cpu != -1) @@ -6140,6 +6782,9 @@ perf_event_alloc(struct perf_event_attr *attr, int cpu, INIT_LIST_HEAD(&event->event_entry); INIT_LIST_HEAD(&event->sibling_list); INIT_LIST_HEAD(&event->rb_entry); + INIT_LIST_HEAD(&event->active_entry); + INIT_HLIST_NODE(&event->hlist_entry); + init_waitqueue_head(&event->waitq); init_irq_work(&event->pending, perf_pending_event); @@ -6162,11 +6807,14 @@ perf_event_alloc(struct perf_event_attr *attr, int cpu, if (task) { event->attach_state = PERF_ATTACH_TASK; + + if (attr->type == PERF_TYPE_TRACEPOINT) + event->hw.tp_target = task; #ifdef CONFIG_HAVE_HW_BREAKPOINT /* * hw_breakpoint is a bit difficult here.. */ - if (attr->type == PERF_TYPE_BREAKPOINT) + else if (attr->type == PERF_TYPE_BREAKPOINT) event->hw.bp_target = task; #endif } @@ -6179,8 +6827,7 @@ perf_event_alloc(struct perf_event_attr *attr, int cpu, event->overflow_handler = overflow_handler; event->overflow_handler_context = context; - if (attr->disabled) - event->state = PERF_EVENT_STATE_OFF; + perf_event__state_init(event); pmu = NULL; @@ -6196,49 +6843,36 @@ perf_event_alloc(struct perf_event_attr *attr, int cpu, * we currently do not support PERF_FORMAT_GROUP on inherited events */ if (attr->inherit && (attr->read_format & PERF_FORMAT_GROUP)) - goto done; + goto err_ns; pmu = perf_init_event(event); - -done: - err = 0; if (!pmu) - err = -EINVAL; - else if (IS_ERR(pmu)) + goto err_ns; + else if (IS_ERR(pmu)) { err = PTR_ERR(pmu); - - if (err) { - if (event->ns) - put_pid_ns(event->ns); - kfree(event); - return ERR_PTR(err); + goto err_ns; } if (!event->parent) { - if (event->attach_state & PERF_ATTACH_TASK) - static_key_slow_inc(&perf_sched_events.key); - if (event->attr.mmap || event->attr.mmap_data) - atomic_inc(&nr_mmap_events); - if (event->attr.comm) - atomic_inc(&nr_comm_events); - if (event->attr.task) - atomic_inc(&nr_task_events); if (event->attr.sample_type & PERF_SAMPLE_CALLCHAIN) { err = get_callchain_buffers(); - if (err) { - free_event(event); - return ERR_PTR(err); - } - } - if (has_branch_stack(event)) { - static_key_slow_inc(&perf_sched_events.key); - if (!(event->attach_state & PERF_ATTACH_TASK)) - atomic_inc(&per_cpu(perf_branch_stack_events, - event->cpu)); + if (err) + goto err_pmu; } } return event; + +err_pmu: + if (event->destroy) + event->destroy(event); + module_put(pmu->module); +err_ns: + if (event->ns) + put_pid_ns(event->ns); + kfree(event); + + return ERR_PTR(err); } static int perf_copy_attr(struct perf_event_attr __user *uattr, @@ -6316,11 +6950,6 @@ static int perf_copy_attr(struct perf_event_attr __user *uattr, if (!(mask & ~PERF_SAMPLE_BRANCH_PLM_ALL)) return -EINVAL; - /* kernel level capture: check permissions */ - if ((mask & PERF_SAMPLE_BRANCH_PERM_PLM) - && perf_paranoid_kernel() && !capable(CAP_SYS_ADMIN)) - return -EACCES; - /* propagate priv level, when not set for branch */ if (!(mask & PERF_SAMPLE_BRANCH_PLM_ALL)) { @@ -6338,6 +6967,10 @@ static int perf_copy_attr(struct perf_event_attr __user *uattr, */ attr->branch_sample_type = mask; } + /* privileged levels capture (kernel, hv): check permissions */ + if ((mask & PERF_SAMPLE_BRANCH_PERM_PLM) + && perf_paranoid_kernel() && !capable(CAP_SYS_ADMIN)) + return -EACCES; } if (attr->sample_type & PERF_SAMPLE_REGS_USER) { @@ -6373,7 +7006,7 @@ err_size: static int perf_event_set_output(struct perf_event *event, struct perf_event *output_event) { - struct ring_buffer *rb = NULL, *old_rb = NULL; + struct ring_buffer *rb = NULL; int ret = -EINVAL; if (!output_event) @@ -6408,16 +7041,12 @@ set: goto unlock; } - old_rb = event->rb; - rcu_assign_pointer(event->rb, rb); - if (old_rb) - ring_buffer_detach(event, old_rb); + ring_buffer_attach(event, rb); + ret = 0; unlock: mutex_unlock(&event->mmap_mutex); - if (old_rb) - ring_buffer_put(old_rb); out: return ret; } @@ -6445,6 +7074,7 @@ SYSCALL_DEFINE5(perf_event_open, int event_fd; int move_group = 0; int err; + int f_flags = O_RDWR; /* for future expandability... */ if (flags & ~PERF_FLAG_ALL) @@ -6462,6 +7092,9 @@ SYSCALL_DEFINE5(perf_event_open, if (attr.freq) { if (attr.sample_freq > sysctl_perf_event_sample_rate) return -EINVAL; + } else { + if (attr.sample_period & (1ULL << 63)) + return -EINVAL; } /* @@ -6473,7 +7106,10 @@ SYSCALL_DEFINE5(perf_event_open, if ((flags & PERF_FLAG_PID_CGROUP) && (pid == -1 || cpu == -1)) return -EINVAL; - event_fd = get_unused_fd(); + if (flags & PERF_FLAG_FD_CLOEXEC) + f_flags |= O_CLOEXEC; + + event_fd = get_unused_fd_flags(f_flags); if (event_fd < 0) return event_fd; @@ -6496,28 +7132,38 @@ SYSCALL_DEFINE5(perf_event_open, } } + if (task && group_leader && + group_leader->attr.inherit != attr.inherit) { + err = -EINVAL; + goto err_task; + } + get_online_cpus(); event = perf_event_alloc(&attr, cpu, task, group_leader, NULL, NULL, NULL); if (IS_ERR(event)) { err = PTR_ERR(event); - goto err_task; + goto err_cpus; } if (flags & PERF_FLAG_PID_CGROUP) { err = perf_cgroup_connect(pid, event, &attr, group_leader); - if (err) + if (err) { + __free_event(event); + goto err_cpus; + } + } + + if (is_sampling_event(event)) { + if (event->pmu->capabilities & PERF_PMU_CAP_NO_INTERRUPT) { + err = -ENOTSUPP; goto err_alloc; - /* - * one more event: - * - that has cgroup constraint on event->cpu - * - that may need work on context switch - */ - atomic_inc(&per_cpu(perf_cgroup_events, event->cpu)); - static_key_slow_inc(&perf_sched_events.key); + } } + account_event(event); + /* * Special case software events and allow them to be part of * any hardware group. @@ -6598,7 +7244,8 @@ SYSCALL_DEFINE5(perf_event_open, goto err_context; } - event_file = anon_inode_getfile("[perf_event]", &perf_fops, event, O_RDWR); + event_file = anon_inode_getfile("[perf_event]", &perf_fops, event, + f_flags); if (IS_ERR(event_file)) { err = PTR_ERR(event_file); goto err_context; @@ -6608,10 +7255,18 @@ SYSCALL_DEFINE5(perf_event_open, struct perf_event_context *gctx = group_leader->ctx; mutex_lock(&gctx->mutex); - perf_remove_from_context(group_leader); + perf_remove_from_context(group_leader, false); + + /* + * Removing from the context ends up with disabled + * event. What we want here is event in the initial + * startup state, ready to be add into new context. + */ + perf_event__state_init(group_leader); list_for_each_entry(sibling, &group_leader->sibling_list, group_entry) { - perf_remove_from_context(sibling); + perf_remove_from_context(sibling, false); + perf_event__state_init(sibling); put_ctx(gctx); } mutex_unlock(&gctx->mutex); @@ -6633,7 +7288,6 @@ SYSCALL_DEFINE5(perf_event_open, } perf_install_in_context(ctx, event, event->cpu); - ++ctx->generation; perf_unpin_context(ctx); mutex_unlock(&ctx->mutex); @@ -6666,8 +7320,9 @@ err_context: put_ctx(ctx); err_alloc: free_event(event); -err_task: +err_cpus: put_online_cpus(); +err_task: if (task) put_task_struct(task); err_group_fd: @@ -6705,6 +7360,8 @@ perf_event_create_kernel_counter(struct perf_event_attr *attr, int cpu, goto err; } + account_event(event); + ctx = find_get_context(event->pmu, task, cpu); if (IS_ERR(ctx)) { err = PTR_ERR(ctx); @@ -6714,7 +7371,6 @@ perf_event_create_kernel_counter(struct perf_event_attr *attr, int cpu, WARN_ON_ONCE(ctx->parent_ctx); mutex_lock(&ctx->mutex); perf_install_in_context(ctx, event, cpu); - ++ctx->generation; perf_unpin_context(ctx); mutex_unlock(&ctx->mutex); @@ -6740,19 +7396,21 @@ void perf_pmu_migrate_context(struct pmu *pmu, int src_cpu, int dst_cpu) mutex_lock(&src_ctx->mutex); list_for_each_entry_safe(event, tmp, &src_ctx->event_list, event_entry) { - perf_remove_from_context(event); + perf_remove_from_context(event, false); + unaccount_event_cpu(event, src_cpu); put_ctx(src_ctx); - list_add(&event->event_entry, &events); + list_add(&event->migrate_entry, &events); } mutex_unlock(&src_ctx->mutex); synchronize_rcu(); mutex_lock(&dst_ctx->mutex); - list_for_each_entry_safe(event, tmp, &events, event_entry) { - list_del(&event->event_entry); + list_for_each_entry_safe(event, tmp, &events, migrate_entry) { + list_del(&event->migrate_entry); if (event->state >= PERF_EVENT_STATE_OFF) event->state = PERF_EVENT_STATE_INACTIVE; + account_event_cpu(event, dst_cpu); perf_install_in_context(dst_ctx, event, dst_cpu); get_ctx(dst_ctx); } @@ -6800,13 +7458,19 @@ __perf_event_exit_task(struct perf_event *child_event, struct perf_event_context *child_ctx, struct task_struct *child) { - if (child_event->parent) { - raw_spin_lock_irq(&child_ctx->lock); - perf_group_detach(child_event); - raw_spin_unlock_irq(&child_ctx->lock); - } - - perf_remove_from_context(child_event); + /* + * Do not destroy the 'original' grouping; because of the context + * switch optimization the original events could've ended up in a + * random child task. + * + * If we were to destroy the original group, all group related + * operations would cease to function properly after this random + * child dies. + * + * Do destroy all inherited groups, we don't care about those + * and being thorough is better. + */ + perf_remove_from_context(child_event, !!child_event->parent); /* * It can happen that the parent exits first, and has events @@ -6821,8 +7485,8 @@ __perf_event_exit_task(struct perf_event *child_event, static void perf_event_exit_task_context(struct task_struct *child, int ctxn) { - struct perf_event *child_event, *tmp; - struct perf_event_context *child_ctx; + struct perf_event *child_event, *next; + struct perf_event_context *child_ctx, *parent_ctx; unsigned long flags; if (likely(!child->perf_event_ctxp[ctxn])) { @@ -6847,6 +7511,15 @@ static void perf_event_exit_task_context(struct task_struct *child, int ctxn) raw_spin_lock(&child_ctx->lock); task_ctx_sched_out(child_ctx); child->perf_event_ctxp[ctxn] = NULL; + + /* + * In order to avoid freeing: child_ctx->parent_ctx->task + * under perf_event_context::lock, grab another reference. + */ + parent_ctx = child_ctx->parent_ctx; + if (parent_ctx) + get_ctx(parent_ctx); + /* * If this context is a clone; unclone it so it can't get * swapped to another process while we're removing all @@ -6857,6 +7530,13 @@ static void perf_event_exit_task_context(struct task_struct *child, int ctxn) raw_spin_unlock_irqrestore(&child_ctx->lock, flags); /* + * Now that we no longer hold perf_event_context::lock, drop + * our extra child_ctx->parent_ctx reference. + */ + if (parent_ctx) + put_ctx(parent_ctx); + + /* * Report the task dead after unscheduling the events so that we * won't get any samples after PERF_RECORD_EXIT. We can however still * get a few PERF_RECORD_READ events. @@ -6875,24 +7555,9 @@ static void perf_event_exit_task_context(struct task_struct *child, int ctxn) */ mutex_lock(&child_ctx->mutex); -again: - list_for_each_entry_safe(child_event, tmp, &child_ctx->pinned_groups, - group_entry) - __perf_event_exit_task(child_event, child_ctx, child); - - list_for_each_entry_safe(child_event, tmp, &child_ctx->flexible_groups, - group_entry) + list_for_each_entry_safe(child_event, next, &child_ctx->event_list, event_entry) __perf_event_exit_task(child_event, child_ctx, child); - /* - * If the last event was a group event, it will have appended all - * its siblings to the list, but we obtained 'tmp' before that which - * will still point to the list head terminating the iteration. - */ - if (!list_empty(&child_ctx->pinned_groups) || - !list_empty(&child_ctx->flexible_groups)) - goto again; - mutex_unlock(&child_ctx->mutex); put_ctx(child_ctx); @@ -7120,7 +7785,7 @@ inherit_task_group(struct perf_event *event, struct task_struct *parent, * child. */ - child_ctx = alloc_perf_context(event->pmu, child); + child_ctx = alloc_perf_context(parent_ctx->pmu, child); if (!child_ctx) return -ENOMEM; @@ -7157,6 +7822,8 @@ int perf_event_init_context(struct task_struct *child, int ctxn) * swapped under us. */ parent_ctx = perf_pin_task_context(parent, ctxn); + if (!parent_ctx) + return 0; /* * No need to check if parent_ctx != NULL here; since we saw @@ -7263,11 +7930,12 @@ static void __init perf_event_init_all_cpus(void) } } -static void __cpuinit perf_event_init_cpu(int cpu) +static void perf_event_init_cpu(int cpu) { struct swevent_htable *swhash = &per_cpu(swevent_htable, cpu); mutex_lock(&swhash->hlist_mutex); + swhash->online = true; if (swhash->hlist_refcount > 0) { struct swevent_hlist *hlist; @@ -7290,15 +7958,15 @@ static void perf_pmu_rotate_stop(struct pmu *pmu) static void __perf_event_exit_context(void *__info) { + struct remove_event re = { .detach_group = false }; struct perf_event_context *ctx = __info; - struct perf_event *event, *tmp; perf_pmu_rotate_stop(ctx->pmu); - list_for_each_entry_safe(event, tmp, &ctx->pinned_groups, group_entry) - __perf_remove_from_context(event); - list_for_each_entry_safe(event, tmp, &ctx->flexible_groups, group_entry) - __perf_remove_from_context(event); + rcu_read_lock(); + list_for_each_entry_rcu(re.event, &ctx->event_list, event_entry) + __perf_remove_from_context(&re); + rcu_read_unlock(); } static void perf_event_exit_cpu_context(int cpu) @@ -7322,11 +7990,12 @@ static void perf_event_exit_cpu(int cpu) { struct swevent_htable *swhash = &per_cpu(swevent_htable, cpu); + perf_event_exit_cpu_context(cpu); + mutex_lock(&swhash->hlist_mutex); + swhash->online = false; swevent_hlist_release(swhash); mutex_unlock(&swhash->hlist_mutex); - - perf_event_exit_cpu_context(cpu); } #else static inline void perf_event_exit_cpu(int cpu) { } @@ -7352,7 +8021,7 @@ static struct notifier_block perf_reboot_notifier = { .priority = INT_MIN, }; -static int __cpuinit +static int perf_cpu_notify(struct notifier_block *self, unsigned long action, void *hcpu) { unsigned int cpu = (long)hcpu; @@ -7368,7 +8037,6 @@ perf_cpu_notify(struct notifier_block *self, unsigned long action, void *hcpu) case CPU_DOWN_PREPARE: perf_event_exit_cpu(cpu); break; - default: break; } @@ -7434,7 +8102,8 @@ unlock: device_initcall(perf_event_sysfs_init); #ifdef CONFIG_CGROUP_PERF -static struct cgroup_subsys_state *perf_cgroup_css_alloc(struct cgroup *cont) +static struct cgroup_subsys_state * +perf_cgroup_css_alloc(struct cgroup_subsys_state *parent_css) { struct perf_cgroup *jc; @@ -7451,11 +8120,10 @@ static struct cgroup_subsys_state *perf_cgroup_css_alloc(struct cgroup *cont) return &jc->css; } -static void perf_cgroup_css_free(struct cgroup *cont) +static void perf_cgroup_css_free(struct cgroup_subsys_state *css) { - struct perf_cgroup *jc; - jc = container_of(cgroup_subsys_state(cont, perf_subsys_id), - struct perf_cgroup, css); + struct perf_cgroup *jc = container_of(css, struct perf_cgroup, css); + free_percpu(jc->info); kfree(jc); } @@ -7467,15 +8135,17 @@ static int __perf_cgroup_move(void *info) return 0; } -static void perf_cgroup_attach(struct cgroup *cgrp, struct cgroup_taskset *tset) +static void perf_cgroup_attach(struct cgroup_subsys_state *css, + struct cgroup_taskset *tset) { struct task_struct *task; - cgroup_taskset_for_each(task, cgrp, tset) + cgroup_taskset_for_each(task, tset) task_function_call(task, __perf_cgroup_move, task); } -static void perf_cgroup_exit(struct cgroup *cgrp, struct cgroup *old_cgrp, +static void perf_cgroup_exit(struct cgroup_subsys_state *css, + struct cgroup_subsys_state *old_css, struct task_struct *task) { /* @@ -7489,19 +8159,10 @@ static void perf_cgroup_exit(struct cgroup *cgrp, struct cgroup *old_cgrp, task_function_call(task, __perf_cgroup_move, task); } -struct cgroup_subsys perf_subsys = { - .name = "perf_event", - .subsys_id = perf_subsys_id, +struct cgroup_subsys perf_event_cgrp_subsys = { .css_alloc = perf_cgroup_css_alloc, .css_free = perf_cgroup_css_free, .exit = perf_cgroup_exit, .attach = perf_cgroup_attach, - - /* - * perf_event cgroup doesn't handle nesting correctly. - * ctx->nr_cgroups adjustments should be propagated through the - * cgroup hierarchy. Fix it and remove the following. - */ - .broken_hierarchy = true, }; #endif /* CONFIG_CGROUP_PERF */ diff --git a/kernel/events/hw_breakpoint.c b/kernel/events/hw_breakpoint.c index fe8a916507e..1559fb0b929 100644 --- a/kernel/events/hw_breakpoint.c +++ b/kernel/events/hw_breakpoint.c @@ -46,23 +46,26 @@ #include <linux/smp.h> #include <linux/hw_breakpoint.h> - - /* * Constraints data */ +struct bp_cpuinfo { + /* Number of pinned cpu breakpoints in a cpu */ + unsigned int cpu_pinned; + /* tsk_pinned[n] is the number of tasks having n+1 breakpoints */ + unsigned int *tsk_pinned; + /* Number of non-pinned cpu/task breakpoints in a cpu */ + unsigned int flexible; /* XXX: placeholder, see fetch_this_slot() */ +}; -/* Number of pinned cpu breakpoints in a cpu */ -static DEFINE_PER_CPU(unsigned int, nr_cpu_bp_pinned[TYPE_MAX]); - -/* Number of pinned task breakpoints in a cpu */ -static DEFINE_PER_CPU(unsigned int *, nr_task_bp_pinned[TYPE_MAX]); - -/* Number of non-pinned cpu/task breakpoints in a cpu */ -static DEFINE_PER_CPU(unsigned int, nr_bp_flexible[TYPE_MAX]); - +static DEFINE_PER_CPU(struct bp_cpuinfo, bp_cpuinfo[TYPE_MAX]); static int nr_slots[TYPE_MAX]; +static struct bp_cpuinfo *get_bp_info(int cpu, enum bp_type_idx type) +{ + return per_cpu_ptr(bp_cpuinfo + type, cpu); +} + /* Keep track of the breakpoints attached to tasks */ static LIST_HEAD(bp_task_head); @@ -96,8 +99,8 @@ static inline enum bp_type_idx find_slot_idx(struct perf_event *bp) */ static unsigned int max_task_bp_pinned(int cpu, enum bp_type_idx type) { + unsigned int *tsk_pinned = get_bp_info(cpu, type)->tsk_pinned; int i; - unsigned int *tsk_pinned = per_cpu(nr_task_bp_pinned[type], cpu); for (i = nr_slots[type] - 1; i >= 0; i--) { if (tsk_pinned[i] > 0) @@ -120,13 +123,20 @@ static int task_bp_pinned(int cpu, struct perf_event *bp, enum bp_type_idx type) list_for_each_entry(iter, &bp_task_head, hw.bp_list) { if (iter->hw.bp_target == tsk && find_slot_idx(iter) == type && - cpu == iter->cpu) + (iter->cpu < 0 || cpu == iter->cpu)) count += hw_breakpoint_weight(iter); } return count; } +static const struct cpumask *cpumask_of_bp(struct perf_event *bp) +{ + if (bp->cpu >= 0) + return cpumask_of(bp->cpu); + return cpu_possible_mask; +} + /* * Report the number of pinned/un-pinned breakpoints we have in * a given cpu (cpu > -1) or in all of them (cpu = -1). @@ -135,25 +145,15 @@ static void fetch_bp_busy_slots(struct bp_busy_slots *slots, struct perf_event *bp, enum bp_type_idx type) { - int cpu = bp->cpu; - struct task_struct *tsk = bp->hw.bp_target; - - if (cpu >= 0) { - slots->pinned = per_cpu(nr_cpu_bp_pinned[type], cpu); - if (!tsk) - slots->pinned += max_task_bp_pinned(cpu, type); - else - slots->pinned += task_bp_pinned(cpu, bp, type); - slots->flexible = per_cpu(nr_bp_flexible[type], cpu); - - return; - } + const struct cpumask *cpumask = cpumask_of_bp(bp); + int cpu; - for_each_online_cpu(cpu) { - unsigned int nr; + for_each_cpu(cpu, cpumask) { + struct bp_cpuinfo *info = get_bp_info(cpu, type); + int nr; - nr = per_cpu(nr_cpu_bp_pinned[type], cpu); - if (!tsk) + nr = info->cpu_pinned; + if (!bp->hw.bp_target) nr += max_task_bp_pinned(cpu, type); else nr += task_bp_pinned(cpu, bp, type); @@ -161,8 +161,7 @@ fetch_bp_busy_slots(struct bp_busy_slots *slots, struct perf_event *bp, if (nr > slots->pinned) slots->pinned = nr; - nr = per_cpu(nr_bp_flexible[type], cpu); - + nr = info->flexible; if (nr > slots->flexible) slots->flexible = nr; } @@ -182,29 +181,19 @@ fetch_this_slot(struct bp_busy_slots *slots, int weight) /* * Add a pinned breakpoint for the given task in our constraint table */ -static void toggle_bp_task_slot(struct perf_event *bp, int cpu, bool enable, +static void toggle_bp_task_slot(struct perf_event *bp, int cpu, enum bp_type_idx type, int weight) { - unsigned int *tsk_pinned; - int old_count = 0; - int old_idx = 0; - int idx = 0; - - old_count = task_bp_pinned(cpu, bp, type); - old_idx = old_count - 1; - idx = old_idx + weight; - - /* tsk_pinned[n] is the number of tasks having n breakpoints */ - tsk_pinned = per_cpu(nr_task_bp_pinned[type], cpu); - if (enable) { - tsk_pinned[idx]++; - if (old_count > 0) - tsk_pinned[old_idx]--; - } else { - tsk_pinned[idx]--; - if (old_count > 0) - tsk_pinned[old_idx]++; - } + unsigned int *tsk_pinned = get_bp_info(cpu, type)->tsk_pinned; + int old_idx, new_idx; + + old_idx = task_bp_pinned(cpu, bp, type) - 1; + new_idx = old_idx + weight; + + if (old_idx >= 0) + tsk_pinned[old_idx]--; + if (new_idx >= 0) + tsk_pinned[new_idx]++; } /* @@ -214,33 +203,26 @@ static void toggle_bp_slot(struct perf_event *bp, bool enable, enum bp_type_idx type, int weight) { - int cpu = bp->cpu; - struct task_struct *tsk = bp->hw.bp_target; + const struct cpumask *cpumask = cpumask_of_bp(bp); + int cpu; - /* Pinned counter cpu profiling */ - if (!tsk) { + if (!enable) + weight = -weight; - if (enable) - per_cpu(nr_cpu_bp_pinned[type], bp->cpu) += weight; - else - per_cpu(nr_cpu_bp_pinned[type], bp->cpu) -= weight; + /* Pinned counter cpu profiling */ + if (!bp->hw.bp_target) { + get_bp_info(bp->cpu, type)->cpu_pinned += weight; return; } /* Pinned counter task profiling */ - - if (!enable) - list_del(&bp->hw.bp_list); - - if (cpu >= 0) { - toggle_bp_task_slot(bp, cpu, enable, type, weight); - } else { - for_each_online_cpu(cpu) - toggle_bp_task_slot(bp, cpu, enable, type, weight); - } + for_each_cpu(cpu, cpumask) + toggle_bp_task_slot(bp, cpu, type, weight); if (enable) list_add_tail(&bp->hw.bp_list, &bp_task_head); + else + list_del(&bp->hw.bp_list); } /* @@ -261,8 +243,8 @@ __weak void arch_unregister_hw_breakpoint(struct perf_event *bp) * * - If attached to a single cpu, check: * - * (per_cpu(nr_bp_flexible, cpu) || (per_cpu(nr_cpu_bp_pinned, cpu) - * + max(per_cpu(nr_task_bp_pinned, cpu)))) < HBP_NUM + * (per_cpu(info->flexible, cpu) || (per_cpu(info->cpu_pinned, cpu) + * + max(per_cpu(info->tsk_pinned, cpu)))) < HBP_NUM * * -> If there are already non-pinned counters in this cpu, it means * there is already a free slot for them. @@ -272,8 +254,8 @@ __weak void arch_unregister_hw_breakpoint(struct perf_event *bp) * * - If attached to every cpus, check: * - * (per_cpu(nr_bp_flexible, *) || (max(per_cpu(nr_cpu_bp_pinned, *)) - * + max(per_cpu(nr_task_bp_pinned, *)))) < HBP_NUM + * (per_cpu(info->flexible, *) || (max(per_cpu(info->cpu_pinned, *)) + * + max(per_cpu(info->tsk_pinned, *)))) < HBP_NUM * * -> This is roughly the same, except we check the number of per cpu * bp for every cpu and we keep the max one. Same for the per tasks @@ -284,16 +266,16 @@ __weak void arch_unregister_hw_breakpoint(struct perf_event *bp) * * - If attached to a single cpu, check: * - * ((per_cpu(nr_bp_flexible, cpu) > 1) + per_cpu(nr_cpu_bp_pinned, cpu) - * + max(per_cpu(nr_task_bp_pinned, cpu))) < HBP_NUM + * ((per_cpu(info->flexible, cpu) > 1) + per_cpu(info->cpu_pinned, cpu) + * + max(per_cpu(info->tsk_pinned, cpu))) < HBP_NUM * - * -> Same checks as before. But now the nr_bp_flexible, if any, must keep + * -> Same checks as before. But now the info->flexible, if any, must keep * one register at least (or they will never be fed). * * - If attached to every cpus, check: * - * ((per_cpu(nr_bp_flexible, *) > 1) + max(per_cpu(nr_cpu_bp_pinned, *)) - * + max(per_cpu(nr_task_bp_pinned, *))) < HBP_NUM + * ((per_cpu(info->flexible, *) > 1) + max(per_cpu(info->cpu_pinned, *)) + * + max(per_cpu(info->tsk_pinned, *))) < HBP_NUM */ static int __reserve_bp_slot(struct perf_event *bp) { @@ -518,8 +500,8 @@ register_wide_hw_breakpoint(struct perf_event_attr *attr, perf_overflow_handler_t triggered, void *context) { - struct perf_event * __percpu *cpu_events, **pevent, *bp; - long err; + struct perf_event * __percpu *cpu_events, *bp; + long err = 0; int cpu; cpu_events = alloc_percpu(typeof(*cpu_events)); @@ -528,31 +510,21 @@ register_wide_hw_breakpoint(struct perf_event_attr *attr, get_online_cpus(); for_each_online_cpu(cpu) { - pevent = per_cpu_ptr(cpu_events, cpu); bp = perf_event_create_kernel_counter(attr, cpu, NULL, triggered, context); - - *pevent = bp; - if (IS_ERR(bp)) { err = PTR_ERR(bp); - goto fail; + break; } - } - put_online_cpus(); - return cpu_events; - -fail: - for_each_online_cpu(cpu) { - pevent = per_cpu_ptr(cpu_events, cpu); - if (IS_ERR(*pevent)) - break; - unregister_hw_breakpoint(*pevent); + per_cpu(*cpu_events, cpu) = bp; } put_online_cpus(); - free_percpu(cpu_events); + if (likely(!err)) + return cpu_events; + + unregister_wide_hw_breakpoint(cpu_events); return (void __percpu __force *)ERR_PTR(err); } EXPORT_SYMBOL_GPL(register_wide_hw_breakpoint); @@ -564,12 +536,10 @@ EXPORT_SYMBOL_GPL(register_wide_hw_breakpoint); void unregister_wide_hw_breakpoint(struct perf_event * __percpu *cpu_events) { int cpu; - struct perf_event **pevent; - for_each_possible_cpu(cpu) { - pevent = per_cpu_ptr(cpu_events, cpu); - unregister_hw_breakpoint(*pevent); - } + for_each_possible_cpu(cpu) + unregister_hw_breakpoint(per_cpu(*cpu_events, cpu)); + free_percpu(cpu_events); } EXPORT_SYMBOL_GPL(unregister_wide_hw_breakpoint); @@ -612,6 +582,11 @@ static int hw_breakpoint_add(struct perf_event *bp, int flags) if (!(flags & PERF_EF_START)) bp->hw.state = PERF_HES_STOPPED; + if (is_sampling_event(bp)) { + bp->hw.last_period = bp->hw.sample_period; + perf_swevent_set_period(bp); + } + return arch_install_hw_breakpoint(bp); } @@ -650,7 +625,6 @@ static struct pmu perf_breakpoint = { int __init init_hw_breakpoint(void) { - unsigned int **task_bp_pinned; int cpu, err_cpu; int i; @@ -659,10 +633,11 @@ int __init init_hw_breakpoint(void) for_each_possible_cpu(cpu) { for (i = 0; i < TYPE_MAX; i++) { - task_bp_pinned = &per_cpu(nr_task_bp_pinned[i], cpu); - *task_bp_pinned = kzalloc(sizeof(int) * nr_slots[i], - GFP_KERNEL); - if (!*task_bp_pinned) + struct bp_cpuinfo *info = get_bp_info(cpu, i); + + info->tsk_pinned = kcalloc(nr_slots[i], sizeof(int), + GFP_KERNEL); + if (!info->tsk_pinned) goto err_alloc; } } @@ -676,7 +651,7 @@ int __init init_hw_breakpoint(void) err_alloc: for_each_possible_cpu(err_cpu) { for (i = 0; i < TYPE_MAX; i++) - kfree(per_cpu(nr_task_bp_pinned[i], cpu)); + kfree(get_bp_info(err_cpu, i)->tsk_pinned); if (err_cpu == cpu) break; } diff --git a/kernel/events/internal.h b/kernel/events/internal.h index d56a64c99a8..569b218782a 100644 --- a/kernel/events/internal.h +++ b/kernel/events/internal.h @@ -16,7 +16,7 @@ struct ring_buffer { int page_order; /* allocation order */ #endif int nr_pages; /* nr of data pages */ - int writable; /* are we writable */ + int overwrite; /* can overwrite itself */ atomic_t poll; /* POLL_ for wakeups */ @@ -31,6 +31,10 @@ struct ring_buffer { spinlock_t event_lock; struct list_head event_list; + atomic_t mmap_count; + unsigned long mmap_locked; + struct user_struct *mmap_user; + struct perf_event_mmap_page *user_page; void *data_pages[0]; }; @@ -78,16 +82,16 @@ static inline unsigned long perf_data_size(struct ring_buffer *rb) } #define DEFINE_OUTPUT_COPY(func_name, memcpy_func) \ -static inline unsigned int \ +static inline unsigned long \ func_name(struct perf_output_handle *handle, \ - const void *buf, unsigned int len) \ + const void *buf, unsigned long len) \ { \ unsigned long size, written; \ \ do { \ - size = min_t(unsigned long, handle->size, len); \ - \ + size = min(handle->size, len); \ written = memcpy_func(handle->addr, buf, size); \ + written = size - written; \ \ len -= written; \ handle->addr += written; \ @@ -106,20 +110,37 @@ func_name(struct perf_output_handle *handle, \ return len; \ } -static inline int memcpy_common(void *dst, const void *src, size_t n) +static inline unsigned long +memcpy_common(void *dst, const void *src, unsigned long n) { memcpy(dst, src, n); - return n; + return 0; } DEFINE_OUTPUT_COPY(__output_copy, memcpy_common) -#define MEMCPY_SKIP(dst, src, n) (n) +static inline unsigned long +memcpy_skip(void *dst, const void *src, unsigned long n) +{ + return 0; +} -DEFINE_OUTPUT_COPY(__output_skip, MEMCPY_SKIP) +DEFINE_OUTPUT_COPY(__output_skip, memcpy_skip) #ifndef arch_perf_out_copy_user -#define arch_perf_out_copy_user __copy_from_user_inatomic +#define arch_perf_out_copy_user arch_perf_out_copy_user + +static inline unsigned long +arch_perf_out_copy_user(void *dst, const void *src, unsigned long n) +{ + unsigned long ret; + + pagefault_disable(); + ret = __copy_from_user_inatomic(dst, src, n); + pagefault_enable(); + + return ret; +} #endif DEFINE_OUTPUT_COPY(__output_copy_user, arch_perf_out_copy_user) diff --git a/kernel/events/ring_buffer.c b/kernel/events/ring_buffer.c index 23cb34ff397..146a5792b1d 100644 --- a/kernel/events/ring_buffer.c +++ b/kernel/events/ring_buffer.c @@ -12,28 +12,10 @@ #include <linux/perf_event.h> #include <linux/vmalloc.h> #include <linux/slab.h> +#include <linux/circ_buf.h> #include "internal.h" -static bool perf_output_space(struct ring_buffer *rb, unsigned long tail, - unsigned long offset, unsigned long head) -{ - unsigned long mask; - - if (!rb->writable) - return true; - - mask = perf_data_size(rb) - 1; - - offset = (offset - tail) & mask; - head = (head - tail) & mask; - - if ((int)(head - offset) < 0) - return false; - - return true; -} - static void perf_output_wakeup(struct perf_output_handle *handle) { atomic_set(&handle->rb->poll, POLL_IN); @@ -75,15 +57,37 @@ again: goto out; /* - * Publish the known good head. Rely on the full barrier implied - * by atomic_dec_and_test() order the rb->head read and this - * write. + * Since the mmap() consumer (userspace) can run on a different CPU: + * + * kernel user + * + * if (LOAD ->data_tail) { LOAD ->data_head + * (A) smp_rmb() (C) + * STORE $data LOAD $data + * smp_wmb() (B) smp_mb() (D) + * STORE ->data_head STORE ->data_tail + * } + * + * Where A pairs with D, and B pairs with C. + * + * In our case (A) is a control dependency that separates the load of + * the ->data_tail and the stores of $data. In case ->data_tail + * indicates there is no room in the buffer to store $data we do not. + * + * D needs to be a full barrier since it separates the data READ + * from the tail WRITE. + * + * For B a WMB is sufficient since it separates two WRITEs, and for C + * an RMB is sufficient since it separates two READs. + * + * See perf_output_begin(). */ + smp_wmb(); /* B, matches C */ rb->user_page->data_head = head; /* - * Now check if we missed an update, rely on the (compiler) - * barrier in atomic_dec_and_test() to re-read rb->head. + * Now check if we missed an update -- rely on previous implied + * compiler barriers to force a re-read. */ if (unlikely(head != local_read(&rb->head))) { local_inc(&rb->nest); @@ -102,8 +106,7 @@ int perf_output_begin(struct perf_output_handle *handle, { struct ring_buffer *rb; unsigned long tail, offset, head; - int have_lost; - struct perf_sample_data sample_data; + int have_lost, page_shift; struct { struct perf_event_header header; u64 id; @@ -118,55 +121,72 @@ int perf_output_begin(struct perf_output_handle *handle, event = event->parent; rb = rcu_dereference(event->rb); - if (!rb) + if (unlikely(!rb)) goto out; - handle->rb = rb; - handle->event = event; - - if (!rb->nr_pages) + if (unlikely(!rb->nr_pages)) goto out; + handle->rb = rb; + handle->event = event; + have_lost = local_read(&rb->lost); - if (have_lost) { - lost_event.header.size = sizeof(lost_event); - perf_event_header__init_id(&lost_event.header, &sample_data, - event); - size += lost_event.header.size; + if (unlikely(have_lost)) { + size += sizeof(lost_event); + if (event->attr.sample_id_all) + size += event->id_header_size; } perf_output_get_handle(handle); do { - /* - * Userspace could choose to issue a mb() before updating the - * tail pointer. So that all reads will be completed before the - * write is issued. - */ tail = ACCESS_ONCE(rb->user_page->data_tail); - smp_rmb(); offset = head = local_read(&rb->head); - head += size; - if (unlikely(!perf_output_space(rb, tail, offset, head))) + if (!rb->overwrite && + unlikely(CIRC_SPACE(head, tail, perf_data_size(rb)) < size)) goto fail; + + /* + * The above forms a control dependency barrier separating the + * @tail load above from the data stores below. Since the @tail + * load is required to compute the branch to fail below. + * + * A, matches D; the full memory barrier userspace SHOULD issue + * after reading the data and before storing the new tail + * position. + * + * See perf_output_put_handle(). + */ + + head += size; } while (local_cmpxchg(&rb->head, offset, head) != offset); - if (head - local_read(&rb->wakeup) > rb->watermark) + /* + * We rely on the implied barrier() by local_cmpxchg() to ensure + * none of the data stores below can be lifted up by the compiler. + */ + + if (unlikely(head - local_read(&rb->wakeup) > rb->watermark)) local_add(rb->watermark, &rb->wakeup); - handle->page = offset >> (PAGE_SHIFT + page_order(rb)); - handle->page &= rb->nr_pages - 1; - handle->size = offset & ((PAGE_SIZE << page_order(rb)) - 1); - handle->addr = rb->data_pages[handle->page]; - handle->addr += handle->size; - handle->size = (PAGE_SIZE << page_order(rb)) - handle->size; + page_shift = PAGE_SHIFT + page_order(rb); + + handle->page = (offset >> page_shift) & (rb->nr_pages - 1); + offset &= (1UL << page_shift) - 1; + handle->addr = rb->data_pages[handle->page] + offset; + handle->size = (1UL << page_shift) - offset; + + if (unlikely(have_lost)) { + struct perf_sample_data sample_data; - if (have_lost) { + lost_event.header.size = sizeof(lost_event); lost_event.header.type = PERF_RECORD_LOST; lost_event.header.misc = 0; lost_event.id = event->id; lost_event.lost = local_xchg(&rb->lost, 0); + perf_event_header__init_id(&lost_event.header, + &sample_data, event); perf_output_put(handle, lost_event); perf_event__output_id_sample(event, handle, &sample_data); } @@ -212,7 +232,9 @@ ring_buffer_init(struct ring_buffer *rb, long watermark, int flags) rb->watermark = max_size / 2; if (flags & RING_BUFFER_WRITABLE) - rb->writable = 1; + rb->overwrite = 0; + else + rb->overwrite = 1; atomic_set(&rb->refcount, 1); @@ -312,11 +334,16 @@ void rb_free(struct ring_buffer *rb) } #else +static int data_page_nr(struct ring_buffer *rb) +{ + return rb->nr_pages << page_order(rb); +} struct page * perf_mmap_to_page(struct ring_buffer *rb, unsigned long pgoff) { - if (pgoff > (1UL << page_order(rb))) + /* The '>' counts in the user page. */ + if (pgoff > data_page_nr(rb)) return NULL; return vmalloc_to_page((void *)rb->user_page + pgoff * PAGE_SIZE); @@ -336,10 +363,11 @@ static void rb_free_work(struct work_struct *work) int i, nr; rb = container_of(work, struct ring_buffer, work); - nr = 1 << page_order(rb); + nr = data_page_nr(rb); base = rb->user_page; - for (i = 0; i < nr + 1; i++) + /* The '<=' counts in the user page. */ + for (i = 0; i <= nr; i++) perf_mmap_unmark_page(base + (i * PAGE_SIZE)); vfree(base); @@ -373,7 +401,7 @@ struct ring_buffer *rb_alloc(int nr_pages, long watermark, int cpu, int flags) rb->user_page = all_buf; rb->data_pages[0] = all_buf + PAGE_SIZE; rb->page_order = ilog2(nr_pages); - rb->nr_pages = 1; + rb->nr_pages = !!nr_pages; ring_buffer_init(rb, watermark, flags); diff --git a/kernel/events/uprobes.c b/kernel/events/uprobes.c index dea7acfbb07..6f3254e8c13 100644 --- a/kernel/events/uprobes.c +++ b/kernel/events/uprobes.c @@ -27,6 +27,7 @@ #include <linux/pagemap.h> /* read_mapping_page */ #include <linux/slab.h> #include <linux/sched.h> +#include <linux/export.h> #include <linux/rmap.h> /* anon_vma_prepare */ #include <linux/mmu_notifier.h> /* set_pte_at_notify */ #include <linux/swap.h> /* try_to_free_swap */ @@ -34,6 +35,8 @@ #include <linux/kdebug.h> /* notifier mechanism */ #include "../../mm/internal.h" /* munlock_vma_page */ #include <linux/percpu-rwsem.h> +#include <linux/task_work.h> +#include <linux/shmem_fs.h> #include <linux/uprobes.h> @@ -41,66 +44,80 @@ #define MAX_UPROBE_XOL_SLOTS UINSNS_PER_PAGE static struct rb_root uprobes_tree = RB_ROOT; - -static DEFINE_SPINLOCK(uprobes_treelock); /* serialize rbtree access */ - -#define UPROBES_HASH_SZ 13 - /* - * We need separate register/unregister and mmap/munmap lock hashes because - * of mmap_sem nesting. - * - * uprobe_register() needs to install probes on (potentially) all processes - * and thus needs to acquire multiple mmap_sems (consequtively, not - * concurrently), whereas uprobe_mmap() is called while holding mmap_sem - * for the particular process doing the mmap. - * - * uprobe_register()->register_for_each_vma() needs to drop/acquire mmap_sem - * because of lock order against i_mmap_mutex. This means there's a hole in - * the register vma iteration where a mmap() can happen. - * - * Thus uprobe_register() can race with uprobe_mmap() and we can try and - * install a probe where one is already installed. + * allows us to skip the uprobe_mmap if there are no uprobe events active + * at this time. Probably a fine grained per inode count is better? */ +#define no_uprobe_events() RB_EMPTY_ROOT(&uprobes_tree) -/* serialize (un)register */ -static struct mutex uprobes_mutex[UPROBES_HASH_SZ]; - -#define uprobes_hash(v) (&uprobes_mutex[((unsigned long)(v)) % UPROBES_HASH_SZ]) +static DEFINE_SPINLOCK(uprobes_treelock); /* serialize rbtree access */ +#define UPROBES_HASH_SZ 13 /* serialize uprobe->pending_list */ static struct mutex uprobes_mmap_mutex[UPROBES_HASH_SZ]; #define uprobes_mmap_hash(v) (&uprobes_mmap_mutex[((unsigned long)(v)) % UPROBES_HASH_SZ]) static struct percpu_rw_semaphore dup_mmap_sem; -/* - * uprobe_events allows us to skip the uprobe_mmap if there are no uprobe - * events active at this time. Probably a fine grained per inode count is - * better? - */ -static atomic_t uprobe_events = ATOMIC_INIT(0); - /* Have a copy of original instruction */ #define UPROBE_COPY_INSN 0 -/* Dont run handlers when first register/ last unregister in progress*/ -#define UPROBE_RUN_HANDLER 1 -/* Can skip singlestep */ -#define UPROBE_SKIP_SSTEP 2 struct uprobe { struct rb_node rb_node; /* node in the rb tree */ atomic_t ref; + struct rw_semaphore register_rwsem; struct rw_semaphore consumer_rwsem; - struct mutex copy_mutex; /* TODO: kill me and UPROBE_COPY_INSN */ struct list_head pending_list; struct uprobe_consumer *consumers; struct inode *inode; /* Also hold a ref to inode */ loff_t offset; unsigned long flags; + + /* + * The generic code assumes that it has two members of unknown type + * owned by the arch-specific code: + * + * insn - copy_insn() saves the original instruction here for + * arch_uprobe_analyze_insn(). + * + * ixol - potentially modified instruction to execute out of + * line, copied to xol_area by xol_get_insn_slot(). + */ struct arch_uprobe arch; }; +struct return_instance { + struct uprobe *uprobe; + unsigned long func; + unsigned long orig_ret_vaddr; /* original return address */ + bool chained; /* true, if instance is nested */ + + struct return_instance *next; /* keep as stack */ +}; + +/* + * Execute out of line area: anonymous executable mapping installed + * by the probed task to execute the copy of the original instruction + * mangled by set_swbp(). + * + * On a breakpoint hit, thread contests for a slot. It frees the + * slot after singlestep. Currently a fixed number of slots are + * allocated. + */ +struct xol_area { + wait_queue_head_t wq; /* if all slots are busy */ + atomic_t slot_count; /* number of in-use slots */ + unsigned long *bitmap; /* 0 = free slot */ + struct page *page; + + /* + * We keep the vma's vm_start rather than a pointer to the vma + * itself. The probed process or a naughty kernel module could make + * the vma go away, and we must handle that reasonably gracefully. + */ + unsigned long vaddr; /* Page(s) of instruction slots */ +}; + /* * valid_vma: Verify if the specified vma is an executable vma * Relax restrictions while unregistering: vm_flags might have @@ -111,7 +128,7 @@ struct uprobe { */ static bool valid_vma(struct vm_area_struct *vma, bool is_register) { - vm_flags_t flags = VM_HUGETLB | VM_MAYEXEC | VM_SHARED; + vm_flags_t flags = VM_HUGETLB | VM_MAYEXEC | VM_MAYSHARE; if (is_register) flags |= VM_WRITE; @@ -199,10 +216,31 @@ bool __weak is_swbp_insn(uprobe_opcode_t *insn) return *insn == UPROBE_SWBP_INSN; } -static void copy_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t *opcode) +/** + * is_trap_insn - check if instruction is breakpoint instruction. + * @insn: instruction to be checked. + * Default implementation of is_trap_insn + * Returns true if @insn is a breakpoint instruction. + * + * This function is needed for the case where an architecture has multiple + * trap instructions (like powerpc). + */ +bool __weak is_trap_insn(uprobe_opcode_t *insn) +{ + return is_swbp_insn(insn); +} + +static void copy_from_page(struct page *page, unsigned long vaddr, void *dst, int len) +{ + void *kaddr = kmap_atomic(page); + memcpy(dst, kaddr + (vaddr & ~PAGE_MASK), len); + kunmap_atomic(kaddr); +} + +static void copy_to_page(struct page *page, unsigned long vaddr, const void *src, int len) { void *kaddr = kmap_atomic(page); - memcpy(opcode, kaddr + (vaddr & ~PAGE_MASK), UPROBE_SWBP_INSN_SIZE); + memcpy(kaddr + (vaddr & ~PAGE_MASK), src, len); kunmap_atomic(kaddr); } @@ -211,7 +249,16 @@ static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t uprobe_opcode_t old_opcode; bool is_swbp; - copy_opcode(page, vaddr, &old_opcode); + /* + * Note: We only check if the old_opcode is UPROBE_SWBP_INSN here. + * We do not check if it is any other 'trap variant' which could + * be conditional trap instruction such as the one powerpc supports. + * + * The logic is that we do not care if the underlying instruction + * is a trap variant; uprobes always wins over any other (gdb) + * breakpoint. + */ + copy_from_page(page, vaddr, &old_opcode, UPROBE_SWBP_INSN_SIZE); is_swbp = is_swbp_insn(&old_opcode); if (is_swbp_insn(new_opcode)) { @@ -230,28 +277,22 @@ static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t * Expect the breakpoint instruction to be the smallest size instruction for * the architecture. If an arch has variable length instruction and the * breakpoint instruction is not of the smallest length instruction - * supported by that architecture then we need to modify is_swbp_at_addr and - * write_opcode accordingly. This would never be a problem for archs that - * have fixed length instructions. - */ - -/* - * write_opcode - write the opcode at a given virtual address. + * supported by that architecture then we need to modify is_trap_at_addr and + * uprobe_write_opcode accordingly. This would never be a problem for archs + * that have fixed length instructions. + * + * uprobe_write_opcode - write the opcode at a given virtual address. * @mm: the probed process address space. * @vaddr: the virtual address to store the opcode. * @opcode: opcode to be written at @vaddr. * - * Called with mm->mmap_sem held (for read and with a reference to - * mm). - * - * For mm @mm, write the opcode at @vaddr. + * Called with mm->mmap_sem held for write. * Return 0 (success) or a negative errno. */ -static int write_opcode(struct mm_struct *mm, unsigned long vaddr, +int uprobe_write_opcode(struct mm_struct *mm, unsigned long vaddr, uprobe_opcode_t opcode) { struct page *old_page, *new_page; - void *vaddr_old, *vaddr_new; struct vm_area_struct *vma; int ret; @@ -265,28 +306,25 @@ retry: if (ret <= 0) goto put_old; + ret = anon_vma_prepare(vma); + if (ret) + goto put_old; + ret = -ENOMEM; new_page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma, vaddr); if (!new_page) goto put_old; - __SetPageUptodate(new_page); - - /* copy the page now that we've got it stable */ - vaddr_old = kmap_atomic(old_page); - vaddr_new = kmap_atomic(new_page); - - memcpy(vaddr_new, vaddr_old, PAGE_SIZE); - memcpy(vaddr_new + (vaddr & ~PAGE_MASK), &opcode, UPROBE_SWBP_INSN_SIZE); - - kunmap_atomic(vaddr_new); - kunmap_atomic(vaddr_old); - - ret = anon_vma_prepare(vma); - if (ret) + if (mem_cgroup_charge_anon(new_page, mm, GFP_KERNEL)) goto put_new; + __SetPageUptodate(new_page); + copy_highpage(new_page, old_page); + copy_to_page(new_page, vaddr, &opcode, UPROBE_SWBP_INSN_SIZE); + ret = __replace_page(vma, vaddr, old_page, new_page); + if (ret) + mem_cgroup_uncharge_page(new_page); put_new: page_cache_release(new_page); @@ -309,7 +347,7 @@ put_old: */ int __weak set_swbp(struct arch_uprobe *auprobe, struct mm_struct *mm, unsigned long vaddr) { - return write_opcode(mm, vaddr, UPROBE_SWBP_INSN); + return uprobe_write_opcode(mm, vaddr, UPROBE_SWBP_INSN); } /** @@ -324,7 +362,7 @@ int __weak set_swbp(struct arch_uprobe *auprobe, struct mm_struct *mm, unsigned int __weak set_orig_insn(struct arch_uprobe *auprobe, struct mm_struct *mm, unsigned long vaddr) { - return write_opcode(mm, vaddr, *(uprobe_opcode_t *)auprobe->insn); + return uprobe_write_opcode(mm, vaddr, *(uprobe_opcode_t *)&auprobe->insn); } static int match_uprobe(struct uprobe *l, struct uprobe *r) @@ -430,9 +468,6 @@ static struct uprobe *insert_uprobe(struct uprobe *uprobe) u = __insert_uprobe(uprobe); spin_unlock(&uprobes_treelock); - /* For now assume that the instruction need not be single-stepped */ - __set_bit(UPROBE_SKIP_SSTEP, &uprobe->flags); - return u; } @@ -452,49 +487,27 @@ static struct uprobe *alloc_uprobe(struct inode *inode, loff_t offset) uprobe->inode = igrab(inode); uprobe->offset = offset; + init_rwsem(&uprobe->register_rwsem); init_rwsem(&uprobe->consumer_rwsem); - mutex_init(&uprobe->copy_mutex); /* add to uprobes_tree, sorted on inode:offset */ cur_uprobe = insert_uprobe(uprobe); - /* a uprobe exists for this inode:offset combination */ if (cur_uprobe) { kfree(uprobe); uprobe = cur_uprobe; iput(inode); - } else { - atomic_inc(&uprobe_events); } return uprobe; } -static void handler_chain(struct uprobe *uprobe, struct pt_regs *regs) -{ - struct uprobe_consumer *uc; - - if (!test_bit(UPROBE_RUN_HANDLER, &uprobe->flags)) - return; - - down_read(&uprobe->consumer_rwsem); - for (uc = uprobe->consumers; uc; uc = uc->next) { - if (!uc->filter || uc->filter(uc, current)) - uc->handler(uc, regs); - } - up_read(&uprobe->consumer_rwsem); -} - -/* Returns the previous consumer */ -static struct uprobe_consumer * -consumer_add(struct uprobe *uprobe, struct uprobe_consumer *uc) +static void consumer_add(struct uprobe *uprobe, struct uprobe_consumer *uc) { down_write(&uprobe->consumer_rwsem); uc->next = uprobe->consumers; uprobe->consumers = uc; up_write(&uprobe->consumer_rwsem); - - return uc->next; } /* @@ -520,35 +533,23 @@ static bool consumer_del(struct uprobe *uprobe, struct uprobe_consumer *uc) return ret; } -static int -__copy_insn(struct address_space *mapping, struct file *filp, char *insn, - unsigned long nbytes, loff_t offset) +static int __copy_insn(struct address_space *mapping, struct file *filp, + void *insn, int nbytes, loff_t offset) { struct page *page; - void *vaddr; - unsigned long off; - pgoff_t idx; - - if (!filp) - return -EINVAL; - - if (!mapping->a_ops->readpage) - return -EIO; - - idx = offset >> PAGE_CACHE_SHIFT; - off = offset & ~PAGE_MASK; - /* - * Ensure that the page that has the original instruction is - * populated and in page-cache. + * Ensure that the page that has the original instruction is populated + * and in page-cache. If ->readpage == NULL it must be shmem_mapping(), + * see uprobe_register(). */ - page = read_mapping_page(mapping, idx, filp); + if (mapping->a_ops->readpage) + page = read_mapping_page(mapping, offset >> PAGE_CACHE_SHIFT, filp); + else + page = shmem_read_mapping_page(mapping, offset >> PAGE_CACHE_SHIFT); if (IS_ERR(page)) return PTR_ERR(page); - vaddr = kmap_atomic(page); - memcpy(insn, vaddr + off, nbytes); - kunmap_atomic(vaddr); + copy_from_page(page, offset, insn, nbytes); page_cache_release(page); return 0; @@ -556,28 +557,28 @@ __copy_insn(struct address_space *mapping, struct file *filp, char *insn, static int copy_insn(struct uprobe *uprobe, struct file *filp) { - struct address_space *mapping; - unsigned long nbytes; - int bytes; - - nbytes = PAGE_SIZE - (uprobe->offset & ~PAGE_MASK); - mapping = uprobe->inode->i_mapping; + struct address_space *mapping = uprobe->inode->i_mapping; + loff_t offs = uprobe->offset; + void *insn = &uprobe->arch.insn; + int size = sizeof(uprobe->arch.insn); + int len, err = -EIO; - /* Instruction at end of binary; copy only available bytes */ - if (uprobe->offset + MAX_UINSN_BYTES > uprobe->inode->i_size) - bytes = uprobe->inode->i_size - uprobe->offset; - else - bytes = MAX_UINSN_BYTES; + /* Copy only available bytes, -EIO if nothing was read */ + do { + if (offs >= i_size_read(uprobe->inode)) + break; - /* Instruction at the page-boundary; copy bytes in second page */ - if (nbytes < bytes) { - int err = __copy_insn(mapping, filp, uprobe->arch.insn + nbytes, - bytes - nbytes, uprobe->offset + nbytes); + len = min_t(int, size, PAGE_SIZE - (offs & ~PAGE_MASK)); + err = __copy_insn(mapping, filp, insn, len, offs); if (err) - return err; - bytes = nbytes; - } - return __copy_insn(mapping, filp, uprobe->arch.insn, bytes, uprobe->offset); + break; + + insn += len; + offs += len; + size -= len; + } while (size); + + return err; } static int prepare_uprobe(struct uprobe *uprobe, struct file *file, @@ -588,7 +589,8 @@ static int prepare_uprobe(struct uprobe *uprobe, struct file *file, if (test_bit(UPROBE_COPY_INSN, &uprobe->flags)) return ret; - mutex_lock(&uprobe->copy_mutex); + /* TODO: move this into _register, until then we abuse this sem. */ + down_write(&uprobe->consumer_rwsem); if (test_bit(UPROBE_COPY_INSN, &uprobe->flags)) goto out; @@ -597,14 +599,14 @@ static int prepare_uprobe(struct uprobe *uprobe, struct file *file, goto out; ret = -ENOTSUPP; - if (is_swbp_insn((uprobe_opcode_t *)uprobe->arch.insn)) + if (is_trap_insn((uprobe_opcode_t *)&uprobe->arch.insn)) goto out; ret = arch_uprobe_analyze_insn(&uprobe->arch, mm, vaddr); if (ret) goto out; - /* write_opcode() assumes we don't cross page boundary */ + /* uprobe_write_opcode() assumes we don't cross page boundary */ BUG_ON((uprobe->offset & ~PAGE_MASK) + UPROBE_SWBP_INSN_SIZE > PAGE_SIZE); @@ -612,7 +614,30 @@ static int prepare_uprobe(struct uprobe *uprobe, struct file *file, set_bit(UPROBE_COPY_INSN, &uprobe->flags); out: - mutex_unlock(&uprobe->copy_mutex); + up_write(&uprobe->consumer_rwsem); + + return ret; +} + +static inline bool consumer_filter(struct uprobe_consumer *uc, + enum uprobe_filter_ctx ctx, struct mm_struct *mm) +{ + return !uc->filter || uc->filter(uc, ctx, mm); +} + +static bool filter_chain(struct uprobe *uprobe, + enum uprobe_filter_ctx ctx, struct mm_struct *mm) +{ + struct uprobe_consumer *uc; + bool ret = false; + + down_read(&uprobe->consumer_rwsem); + for (uc = uprobe->consumers; uc; uc = uc->next) { + ret = consumer_filter(uc, ctx, mm); + if (ret) + break; + } + up_read(&uprobe->consumer_rwsem); return ret; } @@ -624,16 +649,6 @@ install_breakpoint(struct uprobe *uprobe, struct mm_struct *mm, bool first_uprobe; int ret; - /* - * If probe is being deleted, unregister thread could be done with - * the vma-rmap-walk through. Adding a probe now can be fatal since - * nobody will be able to cleanup. Also we could be from fork or - * mremap path, where the probe might have already been inserted. - * Hence behave as if probe already existed. - */ - if (!uprobe->consumers) - return 0; - ret = prepare_uprobe(uprobe, vma->vm_file, mm, vaddr); if (ret) return ret; @@ -658,14 +673,14 @@ install_breakpoint(struct uprobe *uprobe, struct mm_struct *mm, static int remove_breakpoint(struct uprobe *uprobe, struct mm_struct *mm, unsigned long vaddr) { - /* can happen if uprobe_register() fails */ - if (!test_bit(MMF_HAS_UPROBES, &mm->flags)) - return 0; - set_bit(MMF_RECALC_UPROBES, &mm->flags); return set_orig_insn(&uprobe->arch, mm, vaddr); } +static inline bool uprobe_is_active(struct uprobe *uprobe) +{ + return !RB_EMPTY_NODE(&uprobe->rb_node); +} /* * There could be threads that have already hit the breakpoint. They * will recheck the current insn and restart if find_uprobe() fails. @@ -673,12 +688,15 @@ remove_breakpoint(struct uprobe *uprobe, struct mm_struct *mm, unsigned long vad */ static void delete_uprobe(struct uprobe *uprobe) { + if (WARN_ON(!uprobe_is_active(uprobe))) + return; + spin_lock(&uprobes_treelock); rb_erase(&uprobe->rb_node, &uprobes_tree); spin_unlock(&uprobes_treelock); + RB_CLEAR_NODE(&uprobe->rb_node); /* for uprobe_is_active() */ iput(uprobe->inode); put_uprobe(uprobe); - atomic_dec(&uprobe_events); } struct map_info { @@ -764,8 +782,10 @@ build_map_info(struct address_space *mapping, loff_t offset, bool is_register) return curr; } -static int register_for_each_vma(struct uprobe *uprobe, bool is_register) +static int +register_for_each_vma(struct uprobe *uprobe, struct uprobe_consumer *new) { + bool is_register = !!new; struct map_info *info; int err = 0; @@ -787,17 +807,23 @@ static int register_for_each_vma(struct uprobe *uprobe, bool is_register) down_write(&mm->mmap_sem); vma = find_vma(mm, info->vaddr); if (!vma || !valid_vma(vma, is_register) || - vma->vm_file->f_mapping->host != uprobe->inode) + file_inode(vma->vm_file) != uprobe->inode) goto unlock; if (vma->vm_start > info->vaddr || vaddr_to_offset(vma, info->vaddr) != uprobe->offset) goto unlock; - if (is_register) - err = install_breakpoint(uprobe, mm, vma, info->vaddr); - else - err |= remove_breakpoint(uprobe, mm, info->vaddr); + if (is_register) { + /* consult only the "caller", new consumer. */ + if (consumer_filter(new, + UPROBE_FILTER_REGISTER, mm)) + err = install_breakpoint(uprobe, mm, vma, info->vaddr); + } else if (test_bit(MMF_HAS_UPROBES, &mm->flags)) { + if (!filter_chain(uprobe, + UPROBE_FILTER_UNREGISTER, mm)) + err |= remove_breakpoint(uprobe, mm, info->vaddr); + } unlock: up_write(&mm->mmap_sem); @@ -810,17 +836,23 @@ static int register_for_each_vma(struct uprobe *uprobe, bool is_register) return err; } -static int __uprobe_register(struct uprobe *uprobe) +static int __uprobe_register(struct uprobe *uprobe, struct uprobe_consumer *uc) { - return register_for_each_vma(uprobe, true); + consumer_add(uprobe, uc); + return register_for_each_vma(uprobe, uc); } -static void __uprobe_unregister(struct uprobe *uprobe) +static void __uprobe_unregister(struct uprobe *uprobe, struct uprobe_consumer *uc) { - if (!register_for_each_vma(uprobe, false)) - delete_uprobe(uprobe); + int err; + if (WARN_ON(!consumer_del(uprobe, uc))) + return; + + err = register_for_each_vma(uprobe, NULL); /* TODO : cant unregister? schedule a worker thread */ + if (!uprobe->consumers && !err) + delete_uprobe(uprobe); } /* @@ -845,31 +877,66 @@ int uprobe_register(struct inode *inode, loff_t offset, struct uprobe_consumer * struct uprobe *uprobe; int ret; - if (!inode || !uc || uc->next) + /* Uprobe must have at least one set consumer */ + if (!uc->handler && !uc->ret_handler) return -EINVAL; + /* copy_insn() uses read_mapping_page() or shmem_read_mapping_page() */ + if (!inode->i_mapping->a_ops->readpage && !shmem_mapping(inode->i_mapping)) + return -EIO; + /* Racy, just to catch the obvious mistakes */ if (offset > i_size_read(inode)) return -EINVAL; - ret = 0; - mutex_lock(uprobes_hash(inode)); + retry: uprobe = alloc_uprobe(inode, offset); - - if (!uprobe) { - ret = -ENOMEM; - } else if (!consumer_add(uprobe, uc)) { - ret = __uprobe_register(uprobe); - if (ret) { - uprobe->consumers = NULL; - __uprobe_unregister(uprobe); - } else { - set_bit(UPROBE_RUN_HANDLER, &uprobe->flags); - } + if (!uprobe) + return -ENOMEM; + /* + * We can race with uprobe_unregister()->delete_uprobe(). + * Check uprobe_is_active() and retry if it is false. + */ + down_write(&uprobe->register_rwsem); + ret = -EAGAIN; + if (likely(uprobe_is_active(uprobe))) { + ret = __uprobe_register(uprobe, uc); + if (ret) + __uprobe_unregister(uprobe, uc); } + up_write(&uprobe->register_rwsem); + put_uprobe(uprobe); - mutex_unlock(uprobes_hash(inode)); - if (uprobe) - put_uprobe(uprobe); + if (unlikely(ret == -EAGAIN)) + goto retry; + return ret; +} +EXPORT_SYMBOL_GPL(uprobe_register); + +/* + * uprobe_apply - unregister a already registered probe. + * @inode: the file in which the probe has to be removed. + * @offset: offset from the start of the file. + * @uc: consumer which wants to add more or remove some breakpoints + * @add: add or remove the breakpoints + */ +int uprobe_apply(struct inode *inode, loff_t offset, + struct uprobe_consumer *uc, bool add) +{ + struct uprobe *uprobe; + struct uprobe_consumer *con; + int ret = -ENOENT; + + uprobe = find_uprobe(inode, offset); + if (WARN_ON(!uprobe)) + return ret; + + down_write(&uprobe->register_rwsem); + for (con = uprobe->consumers; con && con != uc ; con = con->next) + ; + if (con) + ret = register_for_each_vma(uprobe, add ? uc : NULL); + up_write(&uprobe->register_rwsem); + put_uprobe(uprobe); return ret; } @@ -884,25 +951,42 @@ void uprobe_unregister(struct inode *inode, loff_t offset, struct uprobe_consume { struct uprobe *uprobe; - if (!inode || !uc) - return; - uprobe = find_uprobe(inode, offset); - if (!uprobe) + if (WARN_ON(!uprobe)) return; - mutex_lock(uprobes_hash(inode)); + down_write(&uprobe->register_rwsem); + __uprobe_unregister(uprobe, uc); + up_write(&uprobe->register_rwsem); + put_uprobe(uprobe); +} +EXPORT_SYMBOL_GPL(uprobe_unregister); - if (consumer_del(uprobe, uc)) { - if (!uprobe->consumers) { - __uprobe_unregister(uprobe); - clear_bit(UPROBE_RUN_HANDLER, &uprobe->flags); - } +static int unapply_uprobe(struct uprobe *uprobe, struct mm_struct *mm) +{ + struct vm_area_struct *vma; + int err = 0; + + down_read(&mm->mmap_sem); + for (vma = mm->mmap; vma; vma = vma->vm_next) { + unsigned long vaddr; + loff_t offset; + + if (!valid_vma(vma, false) || + file_inode(vma->vm_file) != uprobe->inode) + continue; + + offset = (loff_t)vma->vm_pgoff << PAGE_SHIFT; + if (uprobe->offset < offset || + uprobe->offset >= offset + vma->vm_end - vma->vm_start) + continue; + + vaddr = offset_to_vaddr(vma, uprobe->offset); + err |= remove_breakpoint(uprobe, mm, vaddr); } + up_read(&mm->mmap_sem); - mutex_unlock(uprobes_hash(inode)); - if (uprobe) - put_uprobe(uprobe); + return err; } static struct rb_node * @@ -979,18 +1063,23 @@ int uprobe_mmap(struct vm_area_struct *vma) struct uprobe *uprobe, *u; struct inode *inode; - if (!atomic_read(&uprobe_events) || !valid_vma(vma, true)) + if (no_uprobe_events() || !valid_vma(vma, true)) return 0; - inode = vma->vm_file->f_mapping->host; + inode = file_inode(vma->vm_file); if (!inode) return 0; mutex_lock(uprobes_mmap_hash(inode)); build_probe_list(inode, vma, vma->vm_start, vma->vm_end, &tmp_list); - + /* + * We can race with uprobe_unregister(), this uprobe can be already + * removed. But in this case filter_chain() must return false, all + * consumers have gone away. + */ list_for_each_entry_safe(uprobe, u, &tmp_list, pending_list) { - if (!fatal_signal_pending(current)) { + if (!fatal_signal_pending(current) && + filter_chain(uprobe, UPROBE_FILTER_MMAP, vma->vm_mm)) { unsigned long vaddr = offset_to_vaddr(vma, uprobe->offset); install_breakpoint(uprobe, vma->vm_mm, vma, vaddr); } @@ -1008,7 +1097,7 @@ vma_has_uprobes(struct vm_area_struct *vma, unsigned long start, unsigned long e struct inode *inode; struct rb_node *n; - inode = vma->vm_file->f_mapping->host; + inode = file_inode(vma->vm_file); min = vaddr_to_offset(vma, start); max = min + (end - start) - 1; @@ -1025,7 +1114,7 @@ vma_has_uprobes(struct vm_area_struct *vma, unsigned long start, unsigned long e */ void uprobe_munmap(struct vm_area_struct *vma, unsigned long start, unsigned long end) { - if (!atomic_read(&uprobe_events) || !valid_vma(vma, false)) + if (no_uprobe_events() || !valid_vma(vma, false)) return; if (!atomic_read(&vma->vm_mm->mm_users)) /* called by mmput() ? */ @@ -1040,29 +1129,22 @@ void uprobe_munmap(struct vm_area_struct *vma, unsigned long start, unsigned lon } /* Slot allocation for XOL */ -static int xol_add_vma(struct xol_area *area) +static int xol_add_vma(struct mm_struct *mm, struct xol_area *area) { - struct mm_struct *mm; - int ret; - - area->page = alloc_page(GFP_HIGHUSER); - if (!area->page) - return -ENOMEM; - - ret = -EALREADY; - mm = current->mm; + int ret = -EALREADY; down_write(&mm->mmap_sem); if (mm->uprobes_state.xol_area) goto fail; - ret = -ENOMEM; - - /* Try to map as high as possible, this is only a hint. */ - area->vaddr = get_unmapped_area(NULL, TASK_SIZE - PAGE_SIZE, PAGE_SIZE, 0, 0); - if (area->vaddr & ~PAGE_MASK) { - ret = area->vaddr; - goto fail; + if (!area->vaddr) { + /* Try to map as high as possible, this is only a hint. */ + area->vaddr = get_unmapped_area(NULL, TASK_SIZE - PAGE_SIZE, + PAGE_SIZE, 0, 0); + if (area->vaddr & ~PAGE_MASK) { + ret = area->vaddr; + goto fail; + } } ret = install_special_mapping(mm, area->vaddr, PAGE_SIZE, @@ -1072,55 +1154,66 @@ static int xol_add_vma(struct xol_area *area) smp_wmb(); /* pairs with get_xol_area() */ mm->uprobes_state.xol_area = area; - ret = 0; - -fail: + fail: up_write(&mm->mmap_sem); - if (ret) - __free_page(area->page); return ret; } -static struct xol_area *get_xol_area(struct mm_struct *mm) -{ - struct xol_area *area; - - area = mm->uprobes_state.xol_area; - smp_read_barrier_depends(); /* pairs with wmb in xol_add_vma() */ - - return area; -} - -/* - * xol_alloc_area - Allocate process's xol_area. - * This area will be used for storing instructions for execution out of - * line. - * - * Returns the allocated area or NULL. - */ -static struct xol_area *xol_alloc_area(void) +static struct xol_area *__create_xol_area(unsigned long vaddr) { + struct mm_struct *mm = current->mm; + uprobe_opcode_t insn = UPROBE_SWBP_INSN; struct xol_area *area; - area = kzalloc(sizeof(*area), GFP_KERNEL); + area = kmalloc(sizeof(*area), GFP_KERNEL); if (unlikely(!area)) - return NULL; + goto out; area->bitmap = kzalloc(BITS_TO_LONGS(UINSNS_PER_PAGE) * sizeof(long), GFP_KERNEL); - if (!area->bitmap) - goto fail; + goto free_area; + + area->page = alloc_page(GFP_HIGHUSER); + if (!area->page) + goto free_bitmap; + area->vaddr = vaddr; init_waitqueue_head(&area->wq); - if (!xol_add_vma(area)) + /* Reserve the 1st slot for get_trampoline_vaddr() */ + set_bit(0, area->bitmap); + atomic_set(&area->slot_count, 1); + copy_to_page(area->page, 0, &insn, UPROBE_SWBP_INSN_SIZE); + + if (!xol_add_vma(mm, area)) return area; -fail: + __free_page(area->page); + free_bitmap: kfree(area->bitmap); + free_area: kfree(area); + out: + return NULL; +} + +/* + * get_xol_area - Allocate process's xol_area if necessary. + * This area will be used for storing instructions for execution out of line. + * + * Returns the allocated area or NULL. + */ +static struct xol_area *get_xol_area(void) +{ + struct mm_struct *mm = current->mm; + struct xol_area *area; + + if (!mm->uprobes_state.xol_area) + __create_xol_area(0); - return get_xol_area(current->mm); + area = mm->uprobes_state.xol_area; + smp_read_barrier_depends(); /* pairs with wmb in xol_add_vma() */ + return area; } /* @@ -1186,43 +1279,26 @@ static unsigned long xol_take_insn_slot(struct xol_area *area) } /* - * xol_get_insn_slot - If was not allocated a slot, then - * allocate a slot. + * xol_get_insn_slot - allocate a slot for xol. * Returns the allocated slot address or 0. */ -static unsigned long xol_get_insn_slot(struct uprobe *uprobe, unsigned long slot_addr) +static unsigned long xol_get_insn_slot(struct uprobe *uprobe) { struct xol_area *area; - unsigned long offset; - void *vaddr; + unsigned long xol_vaddr; - area = get_xol_area(current->mm); - if (!area) { - area = xol_alloc_area(); - if (!area) - return 0; - } - current->utask->xol_vaddr = xol_take_insn_slot(area); + area = get_xol_area(); + if (!area) + return 0; - /* - * Initialize the slot if xol_vaddr points to valid - * instruction slot. - */ - if (unlikely(!current->utask->xol_vaddr)) + xol_vaddr = xol_take_insn_slot(area); + if (unlikely(!xol_vaddr)) return 0; - current->utask->vaddr = slot_addr; - offset = current->utask->xol_vaddr & ~PAGE_MASK; - vaddr = kmap_atomic(area->page); - memcpy(vaddr + offset, uprobe->arch.insn, MAX_UINSN_BYTES); - kunmap_atomic(vaddr); - /* - * We probably need flush_icache_user_range() but it needs vma. - * This should work on supported architectures too. - */ - flush_dcache_page(area->page); + arch_uprobe_copy_ixol(area->page, xol_vaddr, + &uprobe->arch.ixol, sizeof(uprobe->arch.ixol)); - return current->utask->xol_vaddr; + return xol_vaddr; } /* @@ -1240,8 +1316,7 @@ static void xol_free_insn_slot(struct task_struct *tsk) return; slot_addr = tsk->utask->xol_vaddr; - - if (unlikely(!slot_addr || IS_ERR_VALUE(slot_addr))) + if (unlikely(!slot_addr)) return; area = tsk->mm->uprobes_state.xol_area; @@ -1264,6 +1339,21 @@ static void xol_free_insn_slot(struct task_struct *tsk) } } +void __weak arch_uprobe_copy_ixol(struct page *page, unsigned long vaddr, + void *src, unsigned long len) +{ + /* Initialize the slot */ + copy_to_page(page, vaddr, src, len); + + /* + * We probably need flush_icache_user_range() but it needs vma. + * This should work on most of architectures by default. If + * architecture needs to do something different it can define + * its own version of the function. + */ + flush_dcache_page(page); +} + /** * uprobe_get_swbp_addr - compute address of swbp given post-swbp regs * @regs: Reflects the saved state of the task after it has hit a breakpoint @@ -1275,6 +1365,16 @@ unsigned long __weak uprobe_get_swbp_addr(struct pt_regs *regs) return instruction_pointer(regs) - UPROBE_SWBP_INSN_SIZE; } +unsigned long uprobe_get_trap_addr(struct pt_regs *regs) +{ + struct uprobe_task *utask = current->utask; + + if (unlikely(utask && utask->active_uprobe)) + return utask->vaddr; + + return instruction_pointer(regs); +} + /* * Called with no locks held. * Called in context of a exiting or a exec-ing thread. @@ -1282,6 +1382,7 @@ unsigned long __weak uprobe_get_swbp_addr(struct pt_regs *regs) void uprobe_free_utask(struct task_struct *t) { struct uprobe_task *utask = t->utask; + struct return_instance *ri, *tmp; if (!utask) return; @@ -1289,47 +1390,226 @@ void uprobe_free_utask(struct task_struct *t) if (utask->active_uprobe) put_uprobe(utask->active_uprobe); + ri = utask->return_instances; + while (ri) { + tmp = ri; + ri = ri->next; + + put_uprobe(tmp->uprobe); + kfree(tmp); + } + xol_free_insn_slot(t); kfree(utask); t->utask = NULL; } /* + * Allocate a uprobe_task object for the task if if necessary. + * Called when the thread hits a breakpoint. + * + * Returns: + * - pointer to new uprobe_task on success + * - NULL otherwise + */ +static struct uprobe_task *get_utask(void) +{ + if (!current->utask) + current->utask = kzalloc(sizeof(struct uprobe_task), GFP_KERNEL); + return current->utask; +} + +static int dup_utask(struct task_struct *t, struct uprobe_task *o_utask) +{ + struct uprobe_task *n_utask; + struct return_instance **p, *o, *n; + + n_utask = kzalloc(sizeof(struct uprobe_task), GFP_KERNEL); + if (!n_utask) + return -ENOMEM; + t->utask = n_utask; + + p = &n_utask->return_instances; + for (o = o_utask->return_instances; o; o = o->next) { + n = kmalloc(sizeof(struct return_instance), GFP_KERNEL); + if (!n) + return -ENOMEM; + + *n = *o; + atomic_inc(&n->uprobe->ref); + n->next = NULL; + + *p = n; + p = &n->next; + n_utask->depth++; + } + + return 0; +} + +static void uprobe_warn(struct task_struct *t, const char *msg) +{ + pr_warn("uprobe: %s:%d failed to %s\n", + current->comm, current->pid, msg); +} + +static void dup_xol_work(struct callback_head *work) +{ + if (current->flags & PF_EXITING) + return; + + if (!__create_xol_area(current->utask->dup_xol_addr)) + uprobe_warn(current, "dup xol area"); +} + +/* * Called in context of a new clone/fork from copy_process. */ -void uprobe_copy_process(struct task_struct *t) +void uprobe_copy_process(struct task_struct *t, unsigned long flags) { + struct uprobe_task *utask = current->utask; + struct mm_struct *mm = current->mm; + struct xol_area *area; + t->utask = NULL; + + if (!utask || !utask->return_instances) + return; + + if (mm == t->mm && !(flags & CLONE_VFORK)) + return; + + if (dup_utask(t, utask)) + return uprobe_warn(t, "dup ret instances"); + + /* The task can fork() after dup_xol_work() fails */ + area = mm->uprobes_state.xol_area; + if (!area) + return uprobe_warn(t, "dup xol area"); + + if (mm == t->mm) + return; + + t->utask->dup_xol_addr = area->vaddr; + init_task_work(&t->utask->dup_xol_work, dup_xol_work); + task_work_add(t, &t->utask->dup_xol_work, true); } /* - * Allocate a uprobe_task object for the task. - * Called when the thread hits a breakpoint for the first time. + * Current area->vaddr notion assume the trampoline address is always + * equal area->vaddr. * - * Returns: - * - pointer to new uprobe_task on success - * - NULL otherwise + * Returns -1 in case the xol_area is not allocated. */ -static struct uprobe_task *add_utask(void) +static unsigned long get_trampoline_vaddr(void) +{ + struct xol_area *area; + unsigned long trampoline_vaddr = -1; + + area = current->mm->uprobes_state.xol_area; + smp_read_barrier_depends(); + if (area) + trampoline_vaddr = area->vaddr; + + return trampoline_vaddr; +} + +static void prepare_uretprobe(struct uprobe *uprobe, struct pt_regs *regs) { + struct return_instance *ri; struct uprobe_task *utask; + unsigned long orig_ret_vaddr, trampoline_vaddr; + bool chained = false; - utask = kzalloc(sizeof *utask, GFP_KERNEL); - if (unlikely(!utask)) - return NULL; + if (!get_xol_area()) + return; + + utask = get_utask(); + if (!utask) + return; + + if (utask->depth >= MAX_URETPROBE_DEPTH) { + printk_ratelimited(KERN_INFO "uprobe: omit uretprobe due to" + " nestedness limit pid/tgid=%d/%d\n", + current->pid, current->tgid); + return; + } + + ri = kzalloc(sizeof(struct return_instance), GFP_KERNEL); + if (!ri) + goto fail; + + trampoline_vaddr = get_trampoline_vaddr(); + orig_ret_vaddr = arch_uretprobe_hijack_return_addr(trampoline_vaddr, regs); + if (orig_ret_vaddr == -1) + goto fail; + + /* + * We don't want to keep trampoline address in stack, rather keep the + * original return address of first caller thru all the consequent + * instances. This also makes breakpoint unwrapping easier. + */ + if (orig_ret_vaddr == trampoline_vaddr) { + if (!utask->return_instances) { + /* + * This situation is not possible. Likely we have an + * attack from user-space. + */ + pr_warn("uprobe: unable to set uretprobe pid/tgid=%d/%d\n", + current->pid, current->tgid); + goto fail; + } + + chained = true; + orig_ret_vaddr = utask->return_instances->orig_ret_vaddr; + } - current->utask = utask; - return utask; + atomic_inc(&uprobe->ref); + ri->uprobe = uprobe; + ri->func = instruction_pointer(regs); + ri->orig_ret_vaddr = orig_ret_vaddr; + ri->chained = chained; + + utask->depth++; + + /* add instance to the stack */ + ri->next = utask->return_instances; + utask->return_instances = ri; + + return; + + fail: + kfree(ri); } /* Prepare to single-step probed instruction out of line. */ static int -pre_ssout(struct uprobe *uprobe, struct pt_regs *regs, unsigned long vaddr) +pre_ssout(struct uprobe *uprobe, struct pt_regs *regs, unsigned long bp_vaddr) { - if (xol_get_insn_slot(uprobe, vaddr) && !arch_uprobe_pre_xol(&uprobe->arch, regs)) - return 0; + struct uprobe_task *utask; + unsigned long xol_vaddr; + int err; - return -EFAULT; + utask = get_utask(); + if (!utask) + return -ENOMEM; + + xol_vaddr = xol_get_insn_slot(uprobe); + if (!xol_vaddr) + return -ENOMEM; + + utask->xol_vaddr = xol_vaddr; + utask->vaddr = bp_vaddr; + + err = arch_uprobe_pre_xol(&uprobe->arch, regs); + if (unlikely(err)) { + xol_free_insn_slot(current); + return err; + } + + utask->active_uprobe = uprobe; + utask->state = UTASK_SSTEP; + return 0; } /* @@ -1366,20 +1646,6 @@ bool uprobe_deny_signal(void) return true; } -/* - * Avoid singlestepping the original instruction if the original instruction - * is a NOP or can be emulated. - */ -static bool can_skip_sstep(struct uprobe *uprobe, struct pt_regs *regs) -{ - if (test_bit(UPROBE_SKIP_SSTEP, &uprobe->flags)) { - if (arch_uprobe_skip_sstep(&uprobe->arch, regs)) - return true; - clear_bit(UPROBE_SKIP_SSTEP, &uprobe->flags); - } - return false; -} - static void mmf_recalc_uprobes(struct mm_struct *mm) { struct vm_area_struct *vma; @@ -1391,6 +1657,7 @@ static void mmf_recalc_uprobes(struct mm_struct *mm) * This is not strictly accurate, we can race with * uprobe_unregister() and see the already removed * uprobe if delete_uprobe() was not yet called. + * Or this uprobe can be filtered out. */ if (vma_has_uprobes(vma, vma->vm_start, vma->vm_end)) return; @@ -1399,7 +1666,7 @@ static void mmf_recalc_uprobes(struct mm_struct *mm) clear_bit(MMF_HAS_UPROBES, &mm->flags); } -static int is_swbp_at_addr(struct mm_struct *mm, unsigned long vaddr) +static int is_trap_at_addr(struct mm_struct *mm, unsigned long vaddr) { struct page *page; uprobe_opcode_t opcode; @@ -1417,10 +1684,11 @@ static int is_swbp_at_addr(struct mm_struct *mm, unsigned long vaddr) if (result < 0) return result; - copy_opcode(page, vaddr, &opcode); + copy_from_page(page, vaddr, &opcode, UPROBE_SWBP_INSN_SIZE); put_page(page); out: - return is_swbp_insn(&opcode); + /* This needs to return true for any variant of the trap insn */ + return is_trap_insn(&opcode); } static struct uprobe *find_active_uprobe(unsigned long bp_vaddr, int *is_swbp) @@ -1433,14 +1701,14 @@ static struct uprobe *find_active_uprobe(unsigned long bp_vaddr, int *is_swbp) vma = find_vma(mm, bp_vaddr); if (vma && vma->vm_start <= bp_vaddr) { if (valid_vma(vma, false)) { - struct inode *inode = vma->vm_file->f_mapping->host; + struct inode *inode = file_inode(vma->vm_file); loff_t offset = vaddr_to_offset(vma, bp_vaddr); uprobe = find_uprobe(inode, offset); } if (!uprobe) - *is_swbp = is_swbp_at_addr(mm, bp_vaddr); + *is_swbp = is_trap_at_addr(mm, bp_vaddr); } else { *is_swbp = -EFAULT; } @@ -1452,20 +1720,119 @@ static struct uprobe *find_active_uprobe(unsigned long bp_vaddr, int *is_swbp) return uprobe; } +static void handler_chain(struct uprobe *uprobe, struct pt_regs *regs) +{ + struct uprobe_consumer *uc; + int remove = UPROBE_HANDLER_REMOVE; + bool need_prep = false; /* prepare return uprobe, when needed */ + + down_read(&uprobe->register_rwsem); + for (uc = uprobe->consumers; uc; uc = uc->next) { + int rc = 0; + + if (uc->handler) { + rc = uc->handler(uc, regs); + WARN(rc & ~UPROBE_HANDLER_MASK, + "bad rc=0x%x from %pf()\n", rc, uc->handler); + } + + if (uc->ret_handler) + need_prep = true; + + remove &= rc; + } + + if (need_prep && !remove) + prepare_uretprobe(uprobe, regs); /* put bp at return */ + + if (remove && uprobe->consumers) { + WARN_ON(!uprobe_is_active(uprobe)); + unapply_uprobe(uprobe, current->mm); + } + up_read(&uprobe->register_rwsem); +} + +static void +handle_uretprobe_chain(struct return_instance *ri, struct pt_regs *regs) +{ + struct uprobe *uprobe = ri->uprobe; + struct uprobe_consumer *uc; + + down_read(&uprobe->register_rwsem); + for (uc = uprobe->consumers; uc; uc = uc->next) { + if (uc->ret_handler) + uc->ret_handler(uc, ri->func, regs); + } + up_read(&uprobe->register_rwsem); +} + +static bool handle_trampoline(struct pt_regs *regs) +{ + struct uprobe_task *utask; + struct return_instance *ri, *tmp; + bool chained; + + utask = current->utask; + if (!utask) + return false; + + ri = utask->return_instances; + if (!ri) + return false; + + /* + * TODO: we should throw out return_instance's invalidated by + * longjmp(), currently we assume that the probed function always + * returns. + */ + instruction_pointer_set(regs, ri->orig_ret_vaddr); + + for (;;) { + handle_uretprobe_chain(ri, regs); + + chained = ri->chained; + put_uprobe(ri->uprobe); + + tmp = ri; + ri = ri->next; + kfree(tmp); + utask->depth--; + + if (!chained) + break; + BUG_ON(!ri); + } + + utask->return_instances = ri; + + return true; +} + +bool __weak arch_uprobe_ignore(struct arch_uprobe *aup, struct pt_regs *regs) +{ + return false; +} + /* * Run handler and ask thread to singlestep. * Ensure all non-fatal signals cannot interrupt thread while it singlesteps. */ static void handle_swbp(struct pt_regs *regs) { - struct uprobe_task *utask; struct uprobe *uprobe; unsigned long bp_vaddr; int uninitialized_var(is_swbp); bp_vaddr = uprobe_get_swbp_addr(regs); - uprobe = find_active_uprobe(bp_vaddr, &is_swbp); + if (bp_vaddr == get_trampoline_vaddr()) { + if (handle_trampoline(regs)) + return; + pr_warn("uprobe: unable to handle uretprobe pid/tgid=%d/%d\n", + current->pid, current->tgid); + } + + uprobe = find_active_uprobe(bp_vaddr, &is_swbp); if (!uprobe) { if (is_swbp > 0) { /* No matching uprobe; signal SIGTRAP. */ @@ -1483,6 +1850,10 @@ static void handle_swbp(struct pt_regs *regs) } return; } + + /* change it in advance for ->handler() and restart */ + instruction_pointer_set(regs, bp_vaddr); + /* * TODO: move copy_insn/etc into _register and remove this hack. * After we hit the bp, _unregister + _register can install the @@ -1490,32 +1861,24 @@ static void handle_swbp(struct pt_regs *regs) */ smp_rmb(); /* pairs with wmb() in install_breakpoint() */ if (unlikely(!test_bit(UPROBE_COPY_INSN, &uprobe->flags))) - goto restart; + goto out; - utask = current->utask; - if (!utask) { - utask = add_utask(); - /* Cannot allocate; re-execute the instruction. */ - if (!utask) - goto restart; - } + /* Tracing handlers use ->utask to communicate with fetch methods */ + if (!get_utask()) + goto out; + + if (arch_uprobe_ignore(&uprobe->arch, regs)) + goto out; handler_chain(uprobe, regs); - if (can_skip_sstep(uprobe, regs)) + + if (arch_uprobe_skip_sstep(&uprobe->arch, regs)) goto out; - if (!pre_ssout(uprobe, regs, bp_vaddr)) { - utask->active_uprobe = uprobe; - utask->state = UTASK_SSTEP; + if (!pre_ssout(uprobe, regs, bp_vaddr)) return; - } -restart: - /* - * cannot singlestep; cannot skip instruction; - * re-execute the instruction. - */ - instruction_pointer_set(regs, bp_vaddr); + /* arch_uprobe_skip_sstep() succeeded, or restart if can't singlestep */ out: put_uprobe(uprobe); } @@ -1527,10 +1890,11 @@ out: static void handle_singlestep(struct uprobe_task *utask, struct pt_regs *regs) { struct uprobe *uprobe; + int err = 0; uprobe = utask->active_uprobe; if (utask->state == UTASK_SSTEP_ACK) - arch_uprobe_post_xol(&uprobe->arch, regs); + err = arch_uprobe_post_xol(&uprobe->arch, regs); else if (utask->state == UTASK_SSTEP_TRAPPED) arch_uprobe_abort_xol(&uprobe->arch, regs); else @@ -1544,6 +1908,11 @@ static void handle_singlestep(struct uprobe_task *utask, struct pt_regs *regs) spin_lock_irq(¤t->sighand->siglock); recalc_sigpending(); /* see uprobe_deny_signal() */ spin_unlock_irq(¤t->sighand->siglock); + + if (unlikely(err)) { + uprobe_warn(current, "execute the probed insn, sending SIGILL."); + force_sig_info(SIGILL, SEND_SIG_FORCED, current); + } } /* @@ -1576,7 +1945,11 @@ void uprobe_notify_resume(struct pt_regs *regs) */ int uprobe_pre_sstep_notifier(struct pt_regs *regs) { - if (!current->mm || !test_bit(MMF_HAS_UPROBES, ¤t->mm->flags)) + if (!current->mm) + return 0; + + if (!test_bit(MMF_HAS_UPROBES, ¤t->mm->flags) && + (!current->utask || !current->utask->return_instances)) return 0; set_thread_flag(TIF_UPROBE); @@ -1609,19 +1982,12 @@ static int __init init_uprobes(void) { int i; - for (i = 0; i < UPROBES_HASH_SZ; i++) { - mutex_init(&uprobes_mutex[i]); + for (i = 0; i < UPROBES_HASH_SZ; i++) mutex_init(&uprobes_mmap_mutex[i]); - } if (percpu_init_rwsem(&dup_mmap_sem)) return -ENOMEM; return register_die_notifier(&uprobe_exception_nb); } -module_init(init_uprobes); - -static void __exit exit_uprobes(void) -{ -} -module_exit(exit_uprobes); +__initcall(init_uprobes); |