Merge branch 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...

[pandora-kernel.git] / kernel / perf_event.c

/*
 * Performance events core code:
 *
 *  Copyright (C) 2008 Thomas Gleixner <tglx@linutronix.de>
 *  Copyright (C) 2008-2009 Red Hat, Inc., Ingo Molnar
 *  Copyright (C) 2008-2009 Red Hat, Inc., Peter Zijlstra <pzijlstr@redhat.com>
 *  Copyright  ©  2009 Paul Mackerras, IBM Corp. <paulus@au1.ibm.com>
 *
 * For licensing details see kernel-base/COPYING
 */

#include <linux/fs.h>
#include <linux/mm.h>
#include <linux/cpu.h>
#include <linux/smp.h>
#include <linux/idr.h>
#include <linux/file.h>
#include <linux/poll.h>
#include <linux/slab.h>
#include <linux/hash.h>
#include <linux/sysfs.h>
#include <linux/dcache.h>
#include <linux/percpu.h>
#include <linux/ptrace.h>
#include <linux/reboot.h>
#include <linux/vmstat.h>
#include <linux/device.h>
#include <linux/vmalloc.h>
#include <linux/hardirq.h>
#include <linux/rculist.h>
#include <linux/uaccess.h>
#include <linux/syscalls.h>
#include <linux/anon_inodes.h>
#include <linux/kernel_stat.h>
#include <linux/perf_event.h>
#include <linux/ftrace_event.h>
#include <linux/hw_breakpoint.h>

#include <asm/irq_regs.h>

enum event_type_t {
        EVENT_FLEXIBLE = 0x1,
        EVENT_PINNED = 0x2,
        EVENT_ALL = EVENT_FLEXIBLE | EVENT_PINNED,
};

atomic_t perf_task_events __read_mostly;
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 LIST_HEAD(pmus);
static DEFINE_MUTEX(pmus_lock);
static struct srcu_struct pmus_srcu;

/*
 * perf event paranoia level:
 *  -1 - not paranoid at all
 *   0 - disallow raw tracepoint access for unpriv
 *   1 - disallow cpu events for unpriv
 *   2 - disallow kernel profiling for unpriv
 */
int sysctl_perf_event_paranoid __read_mostly = 1;

int sysctl_perf_event_mlock __read_mostly = 512; /* 'free' kb per user */

/*
 * max perf event sample rate
 */
int sysctl_perf_event_sample_rate __read_mostly = 100000;

static atomic64_t perf_event_id;

static void cpu_ctx_sched_out(struct perf_cpu_context *cpuctx,
                              enum event_type_t event_type);

static void cpu_ctx_sched_in(struct perf_cpu_context *cpuctx,
                             enum event_type_t event_type);

void __weak perf_event_print_debug(void)        { }

extern __weak const char *perf_pmu_name(void)
{
        return "pmu";
}

static inline u64 perf_clock(void)
{
        return local_clock();
}

void perf_pmu_disable(struct pmu *pmu)
{
        int *count = this_cpu_ptr(pmu->pmu_disable_count);
        if (!(*count)++)
                pmu->pmu_disable(pmu);
}

void perf_pmu_enable(struct pmu *pmu)
{
        int *count = this_cpu_ptr(pmu->pmu_disable_count);
        if (!--(*count))
                pmu->pmu_enable(pmu);
}

static DEFINE_PER_CPU(struct list_head, rotation_list);

/*
 * perf_pmu_rotate_start() and perf_rotate_context() are fully serialized
 * 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_pmu_rotate_start(struct pmu *pmu)
{
        struct perf_cpu_context *cpuctx = this_cpu_ptr(pmu->pmu_cpu_context);
        struct list_head *head = &__get_cpu_var(rotation_list);

        WARN_ON(!irqs_disabled());

        if (list_empty(&cpuctx->rotation_list))
                list_add(&cpuctx->rotation_list, head);
}

static void get_ctx(struct perf_event_context *ctx)
{
        WARN_ON(!atomic_inc_not_zero(&ctx->refcount));
}

static void free_ctx(struct rcu_head *head)
{
        struct perf_event_context *ctx;

        ctx = container_of(head, struct perf_event_context, rcu_head);
        kfree(ctx);
}

static void put_ctx(struct perf_event_context *ctx)
{
        if (atomic_dec_and_test(&ctx->refcount)) {
                if (ctx->parent_ctx)
                        put_ctx(ctx->parent_ctx);
                if (ctx->task)
                        put_task_struct(ctx->task);
                call_rcu(&ctx->rcu_head, free_ctx);
        }
}

static void unclone_ctx(struct perf_event_context *ctx)
{
        if (ctx->parent_ctx) {
                put_ctx(ctx->parent_ctx);
                ctx->parent_ctx = NULL;
        }
}

static u32 perf_event_pid(struct perf_event *event, struct task_struct *p)
{
        /*
         * only top level events have the pid namespace they were created in
         */
        if (event->parent)
                event = event->parent;

        return task_tgid_nr_ns(p, event->ns);
}

static u32 perf_event_tid(struct perf_event *event, struct task_struct *p)
{
        /*
         * only top level events have the pid namespace they were created in
         */
        if (event->parent)
                event = event->parent;

        return task_pid_nr_ns(p, event->ns);
}

/*
 * If we inherit events we want to return the parent event id
 * to userspace.
 */
static u64 primary_event_id(struct perf_event *event)
{
        u64 id = event->id;

        if (event->parent)
                id = event->parent->id;

        return id;
}

/*
 * Get the perf_event_context for a task and lock it.
 * This has to cope with with the fact that until it is locked,
 * the context could get moved to another task.
 */
static struct perf_event_context *
perf_lock_task_context(struct task_struct *task, int ctxn, unsigned long *flags)
{
        struct perf_event_context *ctx;

        rcu_read_lock();
retry:
        ctx = rcu_dereference(task->perf_event_ctxp[ctxn]);
        if (ctx) {
                /*
                 * If this context is a clone of another, it might
                 * get swapped for another underneath us by
                 * perf_event_task_sched_out, though the
                 * rcu_read_lock() protects us from any context
                 * getting freed.  Lock the context and check if it
                 * got swapped before we could get the lock, and retry
                 * if so.  If we locked the right context, then it
                 * can't get swapped on us any more.
                 */
                raw_spin_lock_irqsave(&ctx->lock, *flags);
                if (ctx != rcu_dereference(task->perf_event_ctxp[ctxn])) {
                        raw_spin_unlock_irqrestore(&ctx->lock, *flags);
                        goto retry;
                }

                if (!atomic_inc_not_zero(&ctx->refcount)) {
                        raw_spin_unlock_irqrestore(&ctx->lock, *flags);
                        ctx = NULL;
                }
        }
        rcu_read_unlock();
        return ctx;
}

/*
 * Get the context for a task and increment its pin_count so it
 * can't get swapped to another task.  This also increments its
 * reference count so that the context can't get freed.
 */
static struct perf_event_context *
perf_pin_task_context(struct task_struct *task, int ctxn)
{
        struct perf_event_context *ctx;
        unsigned long flags;

        ctx = perf_lock_task_context(task, ctxn, &flags);
        if (ctx) {
                ++ctx->pin_count;
                raw_spin_unlock_irqrestore(&ctx->lock, flags);
        }
        return ctx;
}

static void perf_unpin_context(struct perf_event_context *ctx)
{
        unsigned long flags;

        raw_spin_lock_irqsave(&ctx->lock, flags);
        --ctx->pin_count;
        raw_spin_unlock_irqrestore(&ctx->lock, flags);
        put_ctx(ctx);
}

/*
 * Update the record of the current time in a context.
 */
static void update_context_time(struct perf_event_context *ctx)
{
        u64 now = perf_clock();

        ctx->time += now - ctx->timestamp;
        ctx->timestamp = now;
}

static u64 perf_event_time(struct perf_event *event)
{
        struct perf_event_context *ctx = event->ctx;
        return ctx ? ctx->time : 0;
}

/*
 * Update the total_time_enabled and total_time_running fields for a event.
 */
static void update_event_times(struct perf_event *event)
{
        struct perf_event_context *ctx = event->ctx;
        u64 run_end;

        if (event->state < PERF_EVENT_STATE_INACTIVE ||
            event->group_leader->state < PERF_EVENT_STATE_INACTIVE)
                return;

        if (ctx->is_active)
                run_end = perf_event_time(event);
        else
                run_end = event->tstamp_stopped;

        event->total_time_enabled = run_end - event->tstamp_enabled;

        if (event->state == PERF_EVENT_STATE_INACTIVE)
                run_end = event->tstamp_stopped;
        else
                run_end = perf_event_time(event);

        event->total_time_running = run_end - event->tstamp_running;
}

/*
 * Update total_time_enabled and total_time_running for all events in a group.
 */
static void update_group_times(struct perf_event *leader)
{
        struct perf_event *event;

        update_event_times(leader);
        list_for_each_entry(event, &leader->sibling_list, group_entry)
                update_event_times(event);
}

static struct list_head *
ctx_group_list(struct perf_event *event, struct perf_event_context *ctx)
{
        if (event->attr.pinned)
                return &ctx->pinned_groups;
        else
                return &ctx->flexible_groups;
}

/*
 * Add a event from the lists for its context.
 * Must be called with ctx->mutex and ctx->lock held.
 */
static void
list_add_event(struct perf_event *event, struct perf_event_context *ctx)
{
        WARN_ON_ONCE(event->attach_state & PERF_ATTACH_CONTEXT);
        event->attach_state |= PERF_ATTACH_CONTEXT;

        /*
         * If we're a stand alone event or group leader, we go to the context
         * list, group events are kept attached to the group so that
         * perf_group_detach can, at all times, locate all siblings.
         */
        if (event->group_leader == event) {
                struct list_head *list;

                if (is_software_event(event))
                        event->group_flags |= PERF_GROUP_SOFTWARE;

                list = ctx_group_list(event, ctx);
                list_add_tail(&event->group_entry, list);
        }

        list_add_rcu(&event->event_entry, &ctx->event_list);
        if (!ctx->nr_events)
                perf_pmu_rotate_start(ctx->pmu);
        ctx->nr_events++;
        if (event->attr.inherit_stat)
                ctx->nr_stat++;
}

/*
 * Called at perf_event creation and when events are attached/detached from a
 * group.
 */
static void perf_event__read_size(struct perf_event *event)
{
        int entry = sizeof(u64); /* value */
        int size = 0;
        int nr = 1;

        if (event->attr.read_format & PERF_FORMAT_TOTAL_TIME_ENABLED)
                size += sizeof(u64);

        if (event->attr.read_format & PERF_FORMAT_TOTAL_TIME_RUNNING)
                size += sizeof(u64);

        if (event->attr.read_format & PERF_FORMAT_ID)
                entry += sizeof(u64);

        if (event->attr.read_format & PERF_FORMAT_GROUP) {
                nr += event->group_leader->nr_siblings;
                size += sizeof(u64);
        }

        size += entry * nr;
        event->read_size = size;
}

static void perf_event__header_size(struct perf_event *event)
{
        struct perf_sample_data *data;
        u64 sample_type = event->attr.sample_type;
        u16 size = 0;

        perf_event__read_size(event);

        if (sample_type & PERF_SAMPLE_IP)
                size += sizeof(data->ip);

        if (sample_type & PERF_SAMPLE_ADDR)
                size += sizeof(data->addr);

        if (sample_type & PERF_SAMPLE_PERIOD)
                size += sizeof(data->period);

        if (sample_type & PERF_SAMPLE_READ)
                size += event->read_size;

        event->header_size = size;
}

static void perf_event__id_header_size(struct perf_event *event)
{
        struct perf_sample_data *data;
        u64 sample_type = event->attr.sample_type;
        u16 size = 0;

        if (sample_type & PERF_SAMPLE_TID)
                size += sizeof(data->tid_entry);

        if (sample_type & PERF_SAMPLE_TIME)
                size += sizeof(data->time);

        if (sample_type & PERF_SAMPLE_ID)
                size += sizeof(data->id);

        if (sample_type & PERF_SAMPLE_STREAM_ID)
                size += sizeof(data->stream_id);

        if (sample_type & PERF_SAMPLE_CPU)
                size += sizeof(data->cpu_entry);

        event->id_header_size = size;
}

static void perf_group_attach(struct perf_event *event)
{
        struct perf_event *group_leader = event->group_leader, *pos;

        /*
         * We can have double attach due to group movement in perf_event_open.
         */
        if (event->attach_state & PERF_ATTACH_GROUP)
                return;

        event->attach_state |= PERF_ATTACH_GROUP;

        if (group_leader == event)
                return;

        if (group_leader->group_flags & PERF_GROUP_SOFTWARE &&
                        !is_software_event(event))
                group_leader->group_flags &= ~PERF_GROUP_SOFTWARE;

        list_add_tail(&event->group_entry, &group_leader->sibling_list);
        group_leader->nr_siblings++;

        perf_event__header_size(group_leader);

        list_for_each_entry(pos, &group_leader->sibling_list, group_entry)
                perf_event__header_size(pos);
}

/*
 * Remove a event from the lists for its context.
 * Must be called with ctx->mutex and ctx->lock held.
 */
static void
list_del_event(struct perf_event *event, struct perf_event_context *ctx)
{
        /*
         * We can have double detach due to exit/hot-unplug + close.
         */
        if (!(event->attach_state & PERF_ATTACH_CONTEXT))
                return;

        event->attach_state &= ~PERF_ATTACH_CONTEXT;

        ctx->nr_events--;
        if (event->attr.inherit_stat)
                ctx->nr_stat--;

        list_del_rcu(&event->event_entry);

        if (event->group_leader == event)
                list_del_init(&event->group_entry);

        update_group_times(event);

        /*
         * If event was in error state, then keep it
         * that way, otherwise bogus counts will be
         * returned on read(). The only way to get out
         * of error state is by explicit re-enabling
         * of the event
         */
        if (event->state > PERF_EVENT_STATE_OFF)
                event->state = PERF_EVENT_STATE_OFF;
}

static void perf_group_detach(struct perf_event *event)
{
        struct perf_event *sibling, *tmp;
        struct list_head *list = NULL;

        /*
         * We can have double detach due to exit/hot-unplug + close.
         */
        if (!(event->attach_state & PERF_ATTACH_GROUP))
                return;

        event->attach_state &= ~PERF_ATTACH_GROUP;

        /*
         * If this is a sibling, remove it from its group.
         */
        if (event->group_leader != event) {
                list_del_init(&event->group_entry);
                event->group_leader->nr_siblings--;
                goto out;
        }

        if (!list_empty(&event->group_entry))
                list = &event->group_entry;

        /*
         * If this was a group event with sibling events then
         * upgrade the siblings to singleton events by adding them
         * to whatever list we are on.
         */
        list_for_each_entry_safe(sibling, tmp, &event->sibling_list, group_entry) {
                if (list)
                        list_move_tail(&sibling->group_entry, list);
                sibling->group_leader = sibling;

                /* Inherit group flags from the previous leader */
                sibling->group_flags = event->group_flags;
        }

out:
        perf_event__header_size(event->group_leader);

        list_for_each_entry(tmp, &event->group_leader->sibling_list, group_entry)
                perf_event__header_size(tmp);
}

static inline int
event_filter_match(struct perf_event *event)
{
        return event->cpu == -1 || event->cpu == smp_processor_id();
}

static void
event_sched_out(struct perf_event *event,
                  struct perf_cpu_context *cpuctx,
                  struct perf_event_context *ctx)
{
        u64 tstamp = perf_event_time(event);
        u64 delta;
        /*
         * An event which could not be activated because of
         * filter mismatch still needs to have its timings
         * maintained, otherwise bogus information is return
         * via read() for time_enabled, time_running:
         */
        if (event->state == PERF_EVENT_STATE_INACTIVE
            && !event_filter_match(event)) {
                delta = ctx->time - event->tstamp_stopped;
                event->tstamp_running += delta;
                event->tstamp_stopped = tstamp;
        }

        if (event->state != PERF_EVENT_STATE_ACTIVE)
                return;

        event->state = PERF_EVENT_STATE_INACTIVE;
        if (event->pending_disable) {
                event->pending_disable = 0;
                event->state = PERF_EVENT_STATE_OFF;
        }
        event->tstamp_stopped = tstamp;
        event->pmu->del(event, 0);
        event->oncpu = -1;

        if (!is_software_event(event))
                cpuctx->active_oncpu--;
        ctx->nr_active--;
        if (event->attr.exclusive || !cpuctx->active_oncpu)
                cpuctx->exclusive = 0;
}

static void
group_sched_out(struct perf_event *group_event,
                struct perf_cpu_context *cpuctx,
                struct perf_event_context *ctx)
{
        struct perf_event *event;
        int state = group_event->state;

        event_sched_out(group_event, cpuctx, ctx);

        /*
         * Schedule out siblings (if any):
         */
        list_for_each_entry(event, &group_event->sibling_list, group_entry)
                event_sched_out(event, cpuctx, ctx);

        if (state == PERF_EVENT_STATE_ACTIVE && group_event->attr.exclusive)
                cpuctx->exclusive = 0;
}

static inline struct perf_cpu_context *
__get_cpu_context(struct perf_event_context *ctx)
{
        return this_cpu_ptr(ctx->pmu->pmu_cpu_context);
}

/*
 * Cross CPU call to remove a performance event
 *
 * We disable the event on the hardware level first. After that we
 * remove it from the context list.
 */
static void __perf_event_remove_from_context(void *info)
{
        struct perf_event *event = info;
        struct perf_event_context *ctx = event->ctx;
        struct perf_cpu_context *cpuctx = __get_cpu_context(ctx);

        /*
         * If this is a task context, we need to check whether it is
         * the current task context of this cpu. If not it has been
         * scheduled out before the smp call arrived.
         */
        if (ctx->task && cpuctx->task_ctx != ctx)
                return;

        raw_spin_lock(&ctx->lock);

        event_sched_out(event, cpuctx, ctx);

        list_del_event(event, ctx);

        raw_spin_unlock(&ctx->lock);
}


/*
 * Remove the event from a task's (or a CPU's) list of events.
 *
 * Must be called with ctx->mutex held.
 *
 * CPU events are removed with a smp call. For task events we only
 * call when the task is on a CPU.
 *
 * If event->ctx is a cloned context, callers must make sure that
 * every task struct that event->ctx->task could possibly point to
 * remains valid.  This is OK when called from perf_release since
 * that only calls us on the top-level context, which can't be a clone.
 * When called from perf_event_exit_task, it's OK because the
 * context has been detached from its task.
 */
static void perf_event_remove_from_context(struct perf_event *event)
{
        struct perf_event_context *ctx = event->ctx;
        struct task_struct *task = ctx->task;

        if (!task) {
                /*
                 * Per cpu events are removed via an smp call and
                 * the removal is always successful.
                 */
                smp_call_function_single(event->cpu,
                                         __perf_event_remove_from_context,
                                         event, 1);
                return;
        }

retry:
        task_oncpu_function_call(task, __perf_event_remove_from_context,
                                 event);

        raw_spin_lock_irq(&ctx->lock);
        /*
         * If the context is active we need to retry the smp call.
         */
        if (ctx->nr_active && !list_empty(&event->group_entry)) {
                raw_spin_unlock_irq(&ctx->lock);
                goto retry;
        }

        /*
         * The lock prevents that this context is scheduled in so we
         * can remove the event safely, if the call above did not
         * succeed.
         */
        if (!list_empty(&event->group_entry))
                list_del_event(event, ctx);
        raw_spin_unlock_irq(&ctx->lock);
}

/*
 * Cross CPU call to disable a performance event
 */
static void __perf_event_disable(void *info)
{
        struct perf_event *event = info;
        struct perf_event_context *ctx = event->ctx;
        struct perf_cpu_context *cpuctx = __get_cpu_context(ctx);

        /*
         * If this is a per-task event, need to check whether this
         * event's task is the current task on this cpu.
         */
        if (ctx->task && cpuctx->task_ctx != ctx)
                return;

        raw_spin_lock(&ctx->lock);

        /*
         * If the event is on, turn it off.
         * If it is in error state, leave it in error state.
         */
        if (event->state >= PERF_EVENT_STATE_INACTIVE) {
                update_context_time(ctx);
                update_group_times(event);
                if (event == event->group_leader)
                        group_sched_out(event, cpuctx, ctx);
                else
                        event_sched_out(event, cpuctx, ctx);
                event->state = PERF_EVENT_STATE_OFF;
        }

        raw_spin_unlock(&ctx->lock);
}

/*
 * Disable a event.
 *
 * If event->ctx is a cloned context, callers must make sure that
 * every task struct that event->ctx->task could possibly point to
 * remains valid.  This condition is satisifed when called through
 * perf_event_for_each_child or perf_event_for_each because they
 * hold the top-level event's child_mutex, so any descendant that
 * goes to exit will block in sync_child_event.
 * When called from perf_pending_event it's OK because event->ctx
 * is the current context on this CPU and preemption is disabled,
 * hence we can't get into perf_event_task_sched_out for this context.
 */
void perf_event_disable(struct perf_event *event)
{
        struct perf_event_context *ctx = event->ctx;
        struct task_struct *task = ctx->task;

        if (!task) {
                /*
                 * Disable the event on the cpu that it's on
                 */
                smp_call_function_single(event->cpu, __perf_event_disable,
                                         event, 1);
                return;
        }

retry:
        task_oncpu_function_call(task, __perf_event_disable, event);

        raw_spin_lock_irq(&ctx->lock);
        /*
         * If the event is still active, we need to retry the cross-call.
         */
        if (event->state == PERF_EVENT_STATE_ACTIVE) {
                raw_spin_unlock_irq(&ctx->lock);
                goto retry;
        }

        /*
         * Since we have the lock this context can't be scheduled
         * in, so we can change the state safely.
         */
        if (event->state == PERF_EVENT_STATE_INACTIVE) {
                update_group_times(event);
                event->state = PERF_EVENT_STATE_OFF;
        }

        raw_spin_unlock_irq(&ctx->lock);
}

static int
event_sched_in(struct perf_event *event,
                 struct perf_cpu_context *cpuctx,
                 struct perf_event_context *ctx)
{
        u64 tstamp = perf_event_time(event);

        if (event->state <= PERF_EVENT_STATE_OFF)
                return 0;

        event->state = PERF_EVENT_STATE_ACTIVE;
        event->oncpu = smp_processor_id();
        /*
         * The new state must be visible before we turn it on in the hardware:
         */
        smp_wmb();

        if (event->pmu->add(event, PERF_EF_START)) {
                event->state = PERF_EVENT_STATE_INACTIVE;
                event->oncpu = -1;
                return -EAGAIN;
        }

        event->tstamp_running += tstamp - event->tstamp_stopped;

        event->shadow_ctx_time = tstamp - ctx->timestamp;

        if (!is_software_event(event))
                cpuctx->active_oncpu++;
        ctx->nr_active++;

        if (event->attr.exclusive)
                cpuctx->exclusive = 1;

        return 0;
}

static int
group_sched_in(struct perf_event *group_event,
               struct perf_cpu_context *cpuctx,
               struct perf_event_context *ctx)
{
        struct perf_event *event, *partial_group = NULL;
        struct pmu *pmu = group_event->pmu;
        u64 now = ctx->time;
        bool simulate = false;

        if (group_event->state == PERF_EVENT_STATE_OFF)
                return 0;

        pmu->start_txn(pmu);

        if (event_sched_in(group_event, cpuctx, ctx)) {
                pmu->cancel_txn(pmu);
                return -EAGAIN;
        }

        /*
         * Schedule in siblings as one group (if any):
         */
        list_for_each_entry(event, &group_event->sibling_list, group_entry) {
                if (event_sched_in(event, cpuctx, ctx)) {
                        partial_group = event;
                        goto group_error;
                }
        }

        if (!pmu->commit_txn(pmu))
                return 0;

group_error:
        /*
         * Groups can be scheduled in as one unit only, so undo any
         * partial group before returning:
         * The events up to the failed event are scheduled out normally,
         * tstamp_stopped will be updated.
         *
         * The failed events and the remaining siblings need to have
         * their timings updated as if they had gone thru event_sched_in()
         * and event_sched_out(). This is required to get consistent timings
         * across the group. This also takes care of the case where the group
         * could never be scheduled by ensuring tstamp_stopped is set to mark
         * the time the event was actually stopped, such that time delta
         * calculation in update_event_times() is correct.
         */
        list_for_each_entry(event, &group_event->sibling_list, group_entry) {
                if (event == partial_group)
                        simulate = true;

                if (simulate) {
                        event->tstamp_running += now - event->tstamp_stopped;
                        event->tstamp_stopped = now;
                } else {
                        event_sched_out(event, cpuctx, ctx);
                }
        }
        event_sched_out(group_event, cpuctx, ctx);

        pmu->cancel_txn(pmu);

        return -EAGAIN;
}

/*
 * Work out whether we can put this event group on the CPU now.
 */
static int group_can_go_on(struct perf_event *event,
                           struct perf_cpu_context *cpuctx,
                           int can_add_hw)
{
        /*
         * Groups consisting entirely of software events can always go on.
         */
        if (event->group_flags & PERF_GROUP_SOFTWARE)
                return 1;
        /*
         * If an exclusive group is already on, no other hardware
         * events can go on.
         */
        if (cpuctx->exclusive)
                return 0;
        /*
         * If this group is exclusive and there are already
         * events on the CPU, it can't go on.
         */
        if (event->attr.exclusive && cpuctx->active_oncpu)
                return 0;
        /*
         * Otherwise, try to add it if all previous groups were able
         * to go on.
         */
        return can_add_hw;
}

static void add_event_to_ctx(struct perf_event *event,
                               struct perf_event_context *ctx)
{
        u64 tstamp = perf_event_time(event);

        list_add_event(event, ctx);
        perf_group_attach(event);
        event->tstamp_enabled = tstamp;
        event->tstamp_running = tstamp;
        event->tstamp_stopped = tstamp;
}

/*
 * Cross CPU call to install and enable a performance event
 *
 * Must be called with ctx->mutex held
 */
static void __perf_install_in_context(void *info)
{
        struct perf_event *event = info;
        struct perf_event_context *ctx = event->ctx;
        struct perf_event *leader = event->group_leader;
        struct perf_cpu_context *cpuctx = __get_cpu_context(ctx);
        int err;

        /*
         * If this is a task context, we need to check whether it is
         * the current task context of this cpu. If not it has been
         * scheduled out before the smp call arrived.
         * Or possibly this is the right context but it isn't
         * on this cpu because it had no events.
         */
        if (ctx->task && cpuctx->task_ctx != ctx) {
                if (cpuctx->task_ctx || ctx->task != current)
                        return;
                cpuctx->task_ctx = ctx;
        }

        raw_spin_lock(&ctx->lock);
        ctx->is_active = 1;
        update_context_time(ctx);

        add_event_to_ctx(event, ctx);

        if (!event_filter_match(event))
                goto unlock;

        /*
         * Don't put the event on if it is disabled or if
         * it is in a group and the group isn't on.
         */
        if (event->state != PERF_EVENT_STATE_INACTIVE ||
            (leader != event && leader->state != PERF_EVENT_STATE_ACTIVE))
                goto unlock;

        /*
         * An exclusive event can't go on if there are already active
         * hardware events, and no hardware event can go on if there
         * is already an exclusive event on.
         */
        if (!group_can_go_on(event, cpuctx, 1))
                err = -EEXIST;
        else
                err = event_sched_in(event, cpuctx, ctx);

        if (err) {
                /*
                 * This event couldn't go on.  If it is in a group
                 * then we have to pull the whole group off.
                 * If the event group is pinned then put it in error state.
                 */
                if (leader != event)
                        group_sched_out(leader, cpuctx, ctx);
                if (leader->attr.pinned) {
                        update_group_times(leader);
                        leader->state = PERF_EVENT_STATE_ERROR;
                }
        }

unlock:
        raw_spin_unlock(&ctx->lock);
}

/*
 * Attach a performance event to a context
 *
 * First we add the event to the list with the hardware enable bit
 * in event->hw_config cleared.
 *
 * If the event is attached to a task which is on a CPU we use a smp
 * call to enable it in the task context. The task might have been
 * scheduled away, but we check this in the smp call again.
 *
 * Must be called with ctx->mutex held.
 */
static void
perf_install_in_context(struct perf_event_context *ctx,
                        struct perf_event *event,
                        int cpu)
{
        struct task_struct *task = ctx->task;

        event->ctx = ctx;

        if (!task) {
                /*
                 * Per cpu events are installed via an smp call and
                 * the install is always successful.
                 */
                smp_call_function_single(cpu, __perf_install_in_context,
                                         event, 1);
                return;
        }

retry:
        task_oncpu_function_call(task, __perf_install_in_context,
                                 event);

        raw_spin_lock_irq(&ctx->lock);
        /*
         * we need to retry the smp call.
         */
        if (ctx->is_active && list_empty(&event->group_entry)) {
                raw_spin_unlock_irq(&ctx->lock);
                goto retry;
        }

        /*
         * The lock prevents that this context is scheduled in so we
         * can add the event safely, if it the call above did not
         * succeed.
         */
        if (list_empty(&event->group_entry))
                add_event_to_ctx(event, ctx);
        raw_spin_unlock_irq(&ctx->lock);
}

/*
 * Put a event into inactive state and update time fields.
 * Enabling the leader of a group effectively enables all
 * the group members that aren't explicitly disabled, so we
 * have to update their ->tstamp_enabled also.
 * Note: this works for group members as well as group leaders
 * since the non-leader members' sibling_lists will be empty.
 */
static void __perf_event_mark_enabled(struct perf_event *event,
                                        struct perf_event_context *ctx)
{
        struct perf_event *sub;
        u64 tstamp = perf_event_time(event);

        event->state = PERF_EVENT_STATE_INACTIVE;
        event->tstamp_enabled = tstamp - event->total_time_enabled;
        list_for_each_entry(sub, &event->sibling_list, group_entry) {
                if (sub->state >= PERF_EVENT_STATE_INACTIVE)
                        sub->tstamp_enabled = tstamp - sub->total_time_enabled;
        }
}

/*
 * Cross CPU call to enable a performance event
 */
static void __perf_event_enable(void *info)
{
        struct perf_event *event = info;
        struct perf_event_context *ctx = event->ctx;
        struct perf_event *leader = event->group_leader;
        struct perf_cpu_context *cpuctx = __get_cpu_context(ctx);
        int err;

        /*
         * If this is a per-task event, need to check whether this
         * event's task is the current task on this cpu.
         */
        if (ctx->task && cpuctx->task_ctx != ctx) {
                if (cpuctx->task_ctx || ctx->task != current)
                        return;
                cpuctx->task_ctx = ctx;
        }

        raw_spin_lock(&ctx->lock);
        ctx->is_active = 1;
        update_context_time(ctx);

        if (event->state >= PERF_EVENT_STATE_INACTIVE)
                goto unlock;
        __perf_event_mark_enabled(event, ctx);

        if (!event_filter_match(event))
                goto unlock;

        /*
         * If the event is in a group and isn't the group leader,
         * then don't put it on unless the group is on.
         */
        if (leader != event && leader->state != PERF_EVENT_STATE_ACTIVE)
                goto unlock;

        if (!group_can_go_on(event, cpuctx, 1)) {
                err = -EEXIST;
        } else {
                if (event == leader)
                        err = group_sched_in(event, cpuctx, ctx);
                else
                        err = event_sched_in(event, cpuctx, ctx);
        }

        if (err) {
                /*
                 * 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)
                        group_sched_out(leader, cpuctx, ctx);
                if (leader->attr.pinned) {
                        update_group_times(leader);
                        leader->state = PERF_EVENT_STATE_ERROR;
                }
        }

unlock:
        raw_spin_unlock(&ctx->lock);
}

/*
 * Enable a event.
 *
 * If event->ctx is a cloned context, callers must make sure that
 * every task struct that event->ctx->task could possibly point to
 * remains valid.  This condition is satisfied when called through
 * perf_event_for_each_child or perf_event_for_each as described
 * for perf_event_disable.
 */
void perf_event_enable(struct perf_event *event)
{
        struct perf_event_context *ctx = event->ctx;
        struct task_struct *task = ctx->task;

        if (!task) {
                /*
                 * Enable the event on the cpu that it's on
                 */
                smp_call_function_single(event->cpu, __perf_event_enable,
                                         event, 1);
                return;
        }

        raw_spin_lock_irq(&ctx->lock);
        if (event->state >= PERF_EVENT_STATE_INACTIVE)
                goto out;

        /*
         * If the event is in error state, clear that first.
         * That way, if we see the event in error state below, we
         * know that it has gone back into error state, as distinct
         * from the task having been scheduled away before the
         * cross-call arrived.
         */
        if (event->state == PERF_EVENT_STATE_ERROR)
                event->state = PERF_EVENT_STATE_OFF;

retry:
        raw_spin_unlock_irq(&ctx->lock);
        task_oncpu_function_call(task, __perf_event_enable, event);

        raw_spin_lock_irq(&ctx->lock);

        /*
         * If the context is active and the event is still off,
         * we need to retry the cross-call.
         */
        if (ctx->is_active && event->state == PERF_EVENT_STATE_OFF)
                goto retry;

        /*
         * Since we have the lock this context can't be scheduled
         * in, so we can change the state safely.
         */
        if (event->state == PERF_EVENT_STATE_OFF)
                __perf_event_mark_enabled(event, ctx);

out:
        raw_spin_unlock_irq(&ctx->lock);
}

static int perf_event_refresh(struct perf_event *event, int refresh)
{
        /*
         * not supported on inherited events
         */
        if (event->attr.inherit || !is_sampling_event(event))
                return -EINVAL;

        atomic_add(refresh, &event->event_limit);
        perf_event_enable(event);

        return 0;
}

static void ctx_sched_out(struct perf_event_context *ctx,
                          struct perf_cpu_context *cpuctx,
                          enum event_type_t event_type)
{
        struct perf_event *event;

        raw_spin_lock(&ctx->lock);
        perf_pmu_disable(ctx->pmu);
        ctx->is_active = 0;
        if (likely(!ctx->nr_events))
                goto out;
        update_context_time(ctx);

        if (!ctx->nr_active)
                goto out;

        if (event_type & EVENT_PINNED) {
                list_for_each_entry(event, &ctx->pinned_groups, group_entry)
                        group_sched_out(event, cpuctx, ctx);
        }

        if (event_type & EVENT_FLEXIBLE) {
                list_for_each_entry(event, &ctx->flexible_groups, group_entry)
                        group_sched_out(event, cpuctx, ctx);
        }
out:
        perf_pmu_enable(ctx->pmu);
        raw_spin_unlock(&ctx->lock);
}

/*
 * 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.
 */
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;
}

static void __perf_event_sync_stat(struct perf_event *event,
                                     struct perf_event *next_event)
{
        u64 value;

        if (!event->attr.inherit_stat)
                return;

        /*
         * Update the event value, we cannot use perf_event_read()
         * because we're in the middle of a context switch and have IRQs
         * disabled, which upsets smp_call_function_single(), however
         * we know the event must be on the current CPU, therefore we
         * don't need to use it.
         */
        switch (event->state) {
        case PERF_EVENT_STATE_ACTIVE:
                event->pmu->read(event);
                /* fall-through */

        case PERF_EVENT_STATE_INACTIVE:
                update_event_times(event);
                break;

        default:
                break;
        }

        /*
         * In order to keep per-task stats reliable we need to flip the event
         * values when we flip the contexts.
         */
        value = local64_read(&next_event->count);
        value = local64_xchg(&event->count, value);
        local64_set(&next_event->count, value);

        swap(event->total_time_enabled, next_event->total_time_enabled);
        swap(event->total_time_running, next_event->total_time_running);

        /*
         * Since we swizzled the values, update the user visible data too.
         */
        perf_event_update_userpage(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)
{
        struct perf_event *event, *next_event;

        if (!ctx->nr_stat)
                return;

        update_context_time(ctx);

        event = list_first_entry(&ctx->event_list,
                                   struct perf_event, event_entry);

        next_event = list_first_entry(&next_ctx->event_list,
                                        struct perf_event, event_entry);

        while (&event->event_entry != &ctx->event_list &&
               &next_event->event_entry != &next_ctx->event_list) {

                __perf_event_sync_stat(event, next_event);

                event = list_next_entry(event, event_entry);
                next_event = list_next_entry(next_event, event_entry);
        }
}

void perf_event_context_sched_out(struct task_struct *task, int ctxn,
                                  struct task_struct *next)
{
        struct perf_event_context *ctx = task->perf_event_ctxp[ctxn];
        struct perf_event_context *next_ctx;
        struct perf_event_context *parent;
        struct perf_cpu_context *cpuctx;
        int do_switch = 1;

        if (likely(!ctx))
                return;

        cpuctx = __get_cpu_context(ctx);
        if (!cpuctx->task_ctx)
                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) {
                /*
                 * Looks like the two contexts are clones, so we might be
                 * able to optimize the context switch.  We lock both
                 * contexts and check that they are clones under the
                 * lock (including re-checking that neither has been
                 * uncloned in the meantime).  It doesn't matter which
                 * order we take the locks because no other cpu could
                 * be trying to lock both of these tasks.
                 */
                raw_spin_lock(&ctx->lock);
                raw_spin_lock_nested(&next_ctx->lock, SINGLE_DEPTH_NESTING);
                if (context_equiv(ctx, next_ctx)) {
                        /*
                         * XXX do we need a memory barrier of sorts
                         * wrt to rcu_dereference() of perf_event_ctxp
                         */
                        task->perf_event_ctxp[ctxn] = next_ctx;
                        next->perf_event_ctxp[ctxn] = ctx;
                        ctx->task = next;
                        next_ctx->task = task;
                        do_switch = 0;

                        perf_event_sync_stat(ctx, next_ctx);
                }
                raw_spin_unlock(&next_ctx->lock);
                raw_spin_unlock(&ctx->lock);
        }
        rcu_read_unlock();

        if (do_switch) {
                ctx_sched_out(ctx, cpuctx, EVENT_ALL);
                cpuctx->task_ctx = NULL;
        }
}

#define for_each_task_context_nr(ctxn)                                  \
        for ((ctxn) = 0; (ctxn) < perf_nr_task_contexts; (ctxn)++)

/*
 * Called from scheduler to remove the events of the current task,
 * with interrupts disabled.
 *
 * We stop each event and update the event value in event->count.
 *
 * This does not protect us against NMI, but disable()
 * sets the disabled bit in the control field of event _before_
 * accessing the event control register. If a NMI hits, then it will
 * not restart the event.
 */
void __perf_event_task_sched_out(struct task_struct *task,
                                 struct task_struct *next)
{
        int ctxn;

        for_each_task_context_nr(ctxn)
                perf_event_context_sched_out(task, ctxn, next);
}

static void task_ctx_sched_out(struct perf_event_context *ctx,
                               enum event_type_t event_type)
{
        struct perf_cpu_context *cpuctx = __get_cpu_context(ctx);

        if (!cpuctx->task_ctx)
                return;

        if (WARN_ON_ONCE(ctx != cpuctx->task_ctx))
                return;

        ctx_sched_out(ctx, cpuctx, event_type);
        cpuctx->task_ctx = NULL;
}

/*
 * Called with IRQs disabled
 */
static void cpu_ctx_sched_out(struct perf_cpu_context *cpuctx,
                              enum event_type_t event_type)
{
        ctx_sched_out(&cpuctx->ctx, cpuctx, event_type);
}

static void
ctx_pinned_sched_in(struct perf_event_context *ctx,
                    struct perf_cpu_context *cpuctx)
{
        struct perf_event *event;

        list_for_each_entry(event, &ctx->pinned_groups, group_entry) {
                if (event->state <= PERF_EVENT_STATE_OFF)
                        continue;
                if (!event_filter_match(event))
                        continue;

                if (group_can_go_on(event, cpuctx, 1))
                        group_sched_in(event, cpuctx, ctx);

                /*
                 * If this pinned group hasn't been scheduled,
                 * put it in error state.
                 */
                if (event->state == PERF_EVENT_STATE_INACTIVE) {
                        update_group_times(event);
                        event->state = PERF_EVENT_STATE_ERROR;
                }
        }
}

static void
ctx_flexible_sched_in(struct perf_event_context *ctx,
                      struct perf_cpu_context *cpuctx)
{
        struct perf_event *event;
        int can_add_hw = 1;

        list_for_each_entry(event, &ctx->flexible_groups, group_entry) {
                /* Ignore events in OFF or ERROR state */
                if (event->state <= PERF_EVENT_STATE_OFF)
                        continue;
                /*
                 * Listen to the 'cpu' scheduling filter constraint
                 * of events:
                 */
                if (!event_filter_match(event))
                        continue;

                if (group_can_go_on(event, cpuctx, can_add_hw)) {
                        if (group_sched_in(event, cpuctx, ctx))
                                can_add_hw = 0;
                }
        }
}

static void
ctx_sched_in(struct perf_event_context *ctx,
             struct perf_cpu_context *cpuctx,
             enum event_type_t event_type)
{
        raw_spin_lock(&ctx->lock);
        ctx->is_active = 1;
        if (likely(!ctx->nr_events))
                goto out;

        ctx->timestamp = perf_clock();

        /*
         * First go through the list and put on any pinned groups
         * in order to give them the best chance of going on.
         */
        if (event_type & EVENT_PINNED)
                ctx_pinned_sched_in(ctx, cpuctx);

        /* Then walk through the lower prio flexible groups */
        if (event_type & EVENT_FLEXIBLE)
                ctx_flexible_sched_in(ctx, cpuctx);

out:
        raw_spin_unlock(&ctx->lock);
}

static void cpu_ctx_sched_in(struct perf_cpu_context *cpuctx,
                             enum event_type_t event_type)
{
        struct perf_event_context *ctx = &cpuctx->ctx;

        ctx_sched_in(ctx, cpuctx, event_type);
}

static void task_ctx_sched_in(struct perf_event_context *ctx,
                              enum event_type_t event_type)
{
        struct perf_cpu_context *cpuctx;

        cpuctx = __get_cpu_context(ctx);
        if (cpuctx->task_ctx == ctx)
                return;

        ctx_sched_in(ctx, cpuctx, event_type);
        cpuctx->task_ctx = ctx;
}

void perf_event_context_sched_in(struct perf_event_context *ctx)
{
        struct perf_cpu_context *cpuctx;

        cpuctx = __get_cpu_context(ctx);
        if (cpuctx->task_ctx == ctx)
                return;

        perf_pmu_disable(ctx->pmu);
        /*
         * We want to keep the following priority order:
         * cpu pinned (that don't need to move), task pinned,
         * cpu flexible, task flexible.
         */
        cpu_ctx_sched_out(cpuctx, EVENT_FLEXIBLE);

        ctx_sched_in(ctx, cpuctx, EVENT_PINNED);
        cpu_ctx_sched_in(cpuctx, EVENT_FLEXIBLE);
        ctx_sched_in(ctx, cpuctx, EVENT_FLEXIBLE);

        cpuctx->task_ctx = ctx;

        /*
         * Since these rotations are per-cpu, we need to ensure the
         * cpu-context we got scheduled on is actually rotating.
         */
        perf_pmu_rotate_start(ctx->pmu);
        perf_pmu_enable(ctx->pmu);
}

/*
 * Called from scheduler to add the events of the current task
 * with interrupts disabled.
 *
 * We restore the event value and then enable it.
 *
 * This does not protect us against NMI, but enable()
 * sets the enabled bit in the control field of event _before_
 * accessing the event control register. If a NMI hits, then it will
 * keep the event running.
 */
void __perf_event_task_sched_in(struct task_struct *task)
{
        struct perf_event_context *ctx;
        int ctxn;

        for_each_task_context_nr(ctxn) {
                ctx = task->perf_event_ctxp[ctxn];
                if (likely(!ctx))
                        continue;

                perf_event_context_sched_in(ctx);
        }
}

#define MAX_INTERRUPTS (~0ULL)

static void perf_log_throttle(struct perf_event *event, int enable);

static u64 perf_calculate_period(struct perf_event *event, u64 nsec, u64 count)
{
        u64 frequency = event->attr.sample_freq;
        u64 sec = NSEC_PER_SEC;
        u64 divisor, dividend;

        int count_fls, nsec_fls, frequency_fls, sec_fls;

        count_fls = fls64(count);
        nsec_fls = fls64(nsec);
        frequency_fls = fls64(frequency);
        sec_fls = 30;

        /*
         * We got @count in @nsec, with a target of sample_freq HZ
         * the target period becomes:
         *
         *             @count * 10^9
         * period = -------------------
         *          @nsec * sample_freq
         *
         */

        /*
         * Reduce accuracy by one bit such that @a and @b converge
         * to a similar magnitude.
         */
#define REDUCE_FLS(a, b)                \
do {                                    \
        if (a##_fls > b##_fls) {        \
                a >>= 1;                \
                a##_fls--;              \
        } else {                        \
                b >>= 1;                \
                b##_fls--;              \
        }                               \
} while (0)

        /*
         * Reduce accuracy until either term fits in a u64, then proceed with
         * the other, so that finally we can do a u64/u64 division.
         */
        while (count_fls + sec_fls > 64 && nsec_fls + frequency_fls > 64) {
                REDUCE_FLS(nsec, frequency);
                REDUCE_FLS(sec, count);
        }

        if (count_fls + sec_fls > 64) {
                divisor = nsec * frequency;

                while (count_fls + sec_fls > 64) {
                        REDUCE_FLS(count, sec);
                        divisor >>= 1;
                }

                dividend = count * sec;
        } else {
                dividend = count * sec;

                while (nsec_fls + frequency_fls > 64) {
                        REDUCE_FLS(nsec, frequency);
                        dividend >>= 1;
                }

                divisor = nsec * frequency;
        }

        if (!divisor)
                return dividend;

        return div64_u64(dividend, divisor);
}

static void perf_adjust_period(struct perf_event *event, u64 nsec, u64 count)
{
        struct hw_perf_event *hwc = &event->hw;
        s64 period, sample_period;
        s64 delta;

        period = perf_calculate_period(event, nsec, count);

        delta = (s64)(period - hwc->sample_period);
        delta = (delta + 7) / 8; /* low pass filter */

        sample_period = hwc->sample_period + delta;

        if (!sample_period)
                sample_period = 1;

        hwc->sample_period = sample_period;

        if (local64_read(&hwc->period_left) > 8*sample_period) {
                event->pmu->stop(event, PERF_EF_UPDATE);
                local64_set(&hwc->period_left, 0);
                event->pmu->start(event, PERF_EF_RELOAD);
        }
}

static void perf_ctx_adjust_freq(struct perf_event_context *ctx, u64 period)
{
        struct perf_event *event;
        struct hw_perf_event *hwc;
        u64 interrupts, now;
        s64 delta;

        raw_spin_lock(&ctx->lock);
        list_for_each_entry_rcu(event, &ctx->event_list, event_entry) {
                if (event->state != PERF_EVENT_STATE_ACTIVE)
                        continue;

                if (!event_filter_match(event))
                        continue;

                hwc = &event->hw;

                interrupts = hwc->interrupts;
                hwc->interrupts = 0;

                /*
                 * unthrottle events on the tick
                 */
                if (interrupts == MAX_INTERRUPTS) {
                        perf_log_throttle(event, 1);
                        event->pmu->start(event, 0);
                }

                if (!event->attr.freq || !event->attr.sample_freq)
                        continue;

                event->pmu->read(event);
                now = local64_read(&event->count);
                delta = now - hwc->freq_count_stamp;
                hwc->freq_count_stamp = now;

                if (delta > 0)
                        perf_adjust_period(event, period, delta);
        }
        raw_spin_unlock(&ctx->lock);
}

/*
 * Round-robin a context's events:
 */
static void rotate_ctx(struct perf_event_context *ctx)
{
        raw_spin_lock(&ctx->lock);

        /*
         * Rotate the first entry last of non-pinned groups. Rotation might be
         * disabled by the inheritance code.
         */
        if (!ctx->rotate_disable)
                list_rotate_left(&ctx->flexible_groups);

        raw_spin_unlock(&ctx->lock);
}

/*
 * perf_pmu_rotate_start() and perf_rotate_context() are fully serialized
 * 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)
{
        u64 interval = (u64)cpuctx->jiffies_interval * TICK_NSEC;
        struct perf_event_context *ctx = NULL;
        int rotate = 0, remove = 1;

        if (cpuctx->ctx.nr_events) {
                remove = 0;
                if (cpuctx->ctx.nr_events != cpuctx->ctx.nr_active)
                        rotate = 1;
        }

        ctx = cpuctx->task_ctx;
        if (ctx && ctx->nr_events) {
                remove = 0;
                if (ctx->nr_events != ctx->nr_active)
                        rotate = 1;
        }

        perf_pmu_disable(cpuctx->ctx.pmu);
        perf_ctx_adjust_freq(&cpuctx->ctx, interval);
        if (ctx)
                perf_ctx_adjust_freq(ctx, interval);

        if (!rotate)
                goto done;

        cpu_ctx_sched_out(cpuctx, EVENT_FLEXIBLE);
        if (ctx)
                task_ctx_sched_out(ctx, EVENT_FLEXIBLE);

        rotate_ctx(&cpuctx->ctx);
        if (ctx)
                rotate_ctx(ctx);

        cpu_ctx_sched_in(cpuctx, EVENT_FLEXIBLE);
        if (ctx)
                task_ctx_sched_in(ctx, EVENT_FLEXIBLE);

done:
        if (remove)
                list_del_init(&cpuctx->rotation_list);

        perf_pmu_enable(cpuctx->ctx.pmu);
}

void perf_event_task_tick(void)
{
        struct list_head *head = &__get_cpu_var(rotation_list);
        struct perf_cpu_context *cpuctx, *tmp;

        WARN_ON(!irqs_disabled());

        list_for_each_entry_safe(cpuctx, tmp, head, rotation_list) {
                if (cpuctx->jiffies_interval == 1 ||
                                !(jiffies % cpuctx->jiffies_interval))
                        perf_rotate_context(cpuctx);
        }
}

static int event_enable_on_exec(struct perf_event *event,
                                struct perf_event_context *ctx)
{
        if (!event->attr.enable_on_exec)
                return 0;

        event->attr.enable_on_exec = 0;
        if (event->state >= PERF_EVENT_STATE_INACTIVE)
                return 0;

        __perf_event_mark_enabled(event, ctx);

        return 1;
}

/*
 * Enable all of a task's events that have been marked enable-on-exec.
 * This expects task == current.
 */
static void perf_event_enable_on_exec(struct perf_event_context *ctx)
{
        struct perf_event *event;
        unsigned long flags;
        int enabled = 0;
        int ret;

        local_irq_save(flags);
        if (!ctx || !ctx->nr_events)
                goto out;

        task_ctx_sched_out(ctx, EVENT_ALL);

        raw_spin_lock(&ctx->lock);

        list_for_each_entry(event, &ctx->pinned_groups, group_entry) {
                ret = event_enable_on_exec(event, ctx);
                if (ret)
                        enabled = 1;
        }

        list_for_each_entry(event, &ctx->flexible_groups, group_entry) {
                ret = event_enable_on_exec(event, ctx);
                if (ret)
                        enabled = 1;
        }

        /*
         * Unclone this context if we enabled any event.
         */
        if (enabled)
                unclone_ctx(ctx);

        raw_spin_unlock(&ctx->lock);

        perf_event_context_sched_in(ctx);
out:
        local_irq_restore(flags);
}

/*
 * Cross CPU call to read the hardware event
 */
static void __perf_event_read(void *info)
{
        struct perf_event *event = info;
        struct perf_event_context *ctx = event->ctx;
        struct perf_cpu_context *cpuctx = __get_cpu_context(ctx);

        /*
         * If this is a task context, we need to check whether it is
         * the current task context of this cpu.  If not it has been
         * scheduled out before the smp call arrived.  In that case
         * event->count would have been updated to a recent sample
         * when the event was scheduled out.
         */
        if (ctx->task && cpuctx->task_ctx != ctx)
                return;

        raw_spin_lock(&ctx->lock);
        update_context_time(ctx);
        update_event_times(event);
        raw_spin_unlock(&ctx->lock);

        event->pmu->read(event);
}

static inline u64 perf_event_count(struct perf_event *event)
{
        return local64_read(&event->count) + atomic64_read(&event->child_count);
}

static u64 perf_event_read(struct perf_event *event)
{
        /*
         * If event is enabled and currently active on a CPU, update the
         * value in the event structure:
         */
        if (event->state == PERF_EVENT_STATE_ACTIVE) {
                smp_call_function_single(event->oncpu,
                                         __perf_event_read, event, 1);
        } else if (event->state == PERF_EVENT_STATE_INACTIVE) {
                struct perf_event_context *ctx = event->ctx;
                unsigned long flags;

                raw_spin_lock_irqsave(&ctx->lock, flags);
                /*
                 * may read while context is not active
                 * (e.g., thread is blocked), in that case
                 * we cannot update context time
                 */
                if (ctx->is_active)
                        update_context_time(ctx);
                update_event_times(event);
                raw_spin_unlock_irqrestore(&ctx->lock, flags);
        }

        return perf_event_count(event);
}

/*
 * Callchain support
 */

struct callchain_cpus_entries {
        struct rcu_head                 rcu_head;
        struct perf_callchain_entry     *cpu_entries[0];
};

static DEFINE_PER_CPU(int, callchain_recursion[PERF_NR_CONTEXTS]);
static atomic_t nr_callchain_events;
static DEFINE_MUTEX(callchain_mutex);
struct callchain_cpus_entries *callchain_cpus_entries;


__weak void perf_callchain_kernel(struct perf_callchain_entry *entry,
                                  struct pt_regs *regs)
{
}

__weak void perf_callchain_user(struct perf_callchain_entry *entry,
                                struct pt_regs *regs)
{
}

static void release_callchain_buffers_rcu(struct rcu_head *head)
{
        struct callchain_cpus_entries *entries;
        int cpu;

        entries = container_of(head, struct callchain_cpus_entries, rcu_head);

        for_each_possible_cpu(cpu)
                kfree(entries->cpu_entries[cpu]);

        kfree(entries);
}

static void release_callchain_buffers(void)
{
        struct callchain_cpus_entries *entries;

        entries = callchain_cpus_entries;
        rcu_assign_pointer(callchain_cpus_entries, NULL);
        call_rcu(&entries->rcu_head, release_callchain_buffers_rcu);
}

static int alloc_callchain_buffers(void)
{
        int cpu;
        int size;
        struct callchain_cpus_entries *entries;

        /*
         * We can't use the percpu allocation API for data that can be
         * accessed from NMI. Use a temporary manual per cpu allocation
         * until that gets sorted out.
         */
        size = sizeof(*entries) + sizeof(struct perf_callchain_entry *) *
                num_possible_cpus();

        entries = kzalloc(size, GFP_KERNEL);
        if (!entries)
                return -ENOMEM;

        size = sizeof(struct perf_callchain_entry) * PERF_NR_CONTEXTS;

        for_each_possible_cpu(cpu) {
                entries->cpu_entries[cpu] = kmalloc_node(size, GFP_KERNEL,
                                                         cpu_to_node(cpu));
                if (!entries->cpu_entries[cpu])
                        goto fail;
        }

        rcu_assign_pointer(callchain_cpus_entries, entries);

        return 0;

fail:
        for_each_possible_cpu(cpu)
                kfree(entries->cpu_entries[cpu]);
        kfree(entries);

        return -ENOMEM;
}

static int get_callchain_buffers(void)
{
        int err = 0;
        int count;

        mutex_lock(&callchain_mutex);

        count = atomic_inc_return(&nr_callchain_events);
        if (WARN_ON_ONCE(count < 1)) {
                err = -EINVAL;
                goto exit;
        }

        if (count > 1) {
                /* If the allocation failed, give up */
                if (!callchain_cpus_entries)
                        err = -ENOMEM;
                goto exit;
        }

        err = alloc_callchain_buffers();
        if (err)
                release_callchain_buffers();
exit:
        mutex_unlock(&callchain_mutex);

        return err;
}

static void put_callchain_buffers(void)
{
        if (atomic_dec_and_mutex_lock(&nr_callchain_events, &callchain_mutex)) {
                release_callchain_buffers();
                mutex_unlock(&callchain_mutex);
        }
}

static int get_recursion_context(int *recursion)
{
        int rctx;

        if (in_nmi())
                rctx = 3;
        else if (in_irq())
                rctx = 2;
        else if (in_softirq())
                rctx = 1;
        else
                rctx = 0;

        if (recursion[rctx])
                return -1;

        recursion[rctx]++;
        barrier();

        return rctx;
}

static inline void put_recursion_context(int *recursion, int rctx)
{
        barrier();
        recursion[rctx]--;
}

static struct perf_callchain_entry *get_callchain_entry(int *rctx)
{
        int cpu;
        struct callchain_cpus_entries *entries;

        *rctx = get_recursion_context(__get_cpu_var(callchain_recursion));
        if (*rctx == -1)
                return NULL;

        entries = rcu_dereference(callchain_cpus_entries);
        if (!entries)
                return NULL;

        cpu = smp_processor_id();

        return &entries->cpu_entries[cpu][*rctx];
}

static void
put_callchain_entry(int rctx)
{
        put_recursion_context(__get_cpu_var(callchain_recursion), rctx);
}

static struct perf_callchain_entry *perf_callchain(struct pt_regs *regs)
{
        int rctx;
        struct perf_callchain_entry *entry;


        entry = get_callchain_entry(&rctx);
        if (rctx == -1)
                return NULL;

        if (!entry)
                goto exit_put;

        entry->nr = 0;

        if (!user_mode(regs)) {
                perf_callchain_store(entry, PERF_CONTEXT_KERNEL);
                perf_callchain_kernel(entry, regs);
                if (current->mm)
                        regs = task_pt_regs(current);
                else
                        regs = NULL;
        }

        if (regs) {
                perf_callchain_store(entry, PERF_CONTEXT_USER);
                perf_callchain_user(entry, regs);
        }

exit_put:
        put_callchain_entry(rctx);

        return entry;
}

/*
 * Initialize the perf_event context in a task_struct:
 */
static void __perf_event_init_context(struct perf_event_context *ctx)
{
        raw_spin_lock_init(&ctx->lock);
        mutex_init(&ctx->mutex);
        INIT_LIST_HEAD(&ctx->pinned_groups);
        INIT_LIST_HEAD(&ctx->flexible_groups);
        INIT_LIST_HEAD(&ctx->event_list);
        atomic_set(&ctx->refcount, 1);
}

static struct perf_event_context *
alloc_perf_context(struct pmu *pmu, struct task_struct *task)
{
        struct perf_event_context *ctx;

        ctx = kzalloc(sizeof(struct perf_event_context), GFP_KERNEL);
        if (!ctx)
                return NULL;

        __perf_event_init_context(ctx);
        if (task) {
                ctx->task = task;
                get_task_struct(task);
        }
        ctx->pmu = pmu;

        return ctx;
}

static struct task_struct *
find_lively_task_by_vpid(pid_t vpid)
{
        struct task_struct *task;
        int err;

        rcu_read_lock();
        if (!vpid)
                task = current;
        else
                task = find_task_by_vpid(vpid);
        if (task)
                get_task_struct(task);
        rcu_read_unlock();

        if (!task)
                return ERR_PTR(-ESRCH);

        /*
         * Can't attach events to a dying task.
         */
        err = -ESRCH;
        if (task->flags & PF_EXITING)
                goto errout;

        /* Reuse ptrace permission checks for now. */
        err = -EACCES;
        if (!ptrace_may_access(task, PTRACE_MODE_READ))
                goto errout;

        return task;
errout:
        put_task_struct(task);
        return ERR_PTR(err);

}

static struct perf_event_context *
find_get_context(struct pmu *pmu, struct task_struct *task, int cpu)
{
        struct perf_event_context *ctx;
        struct perf_cpu_context *cpuctx;
        unsigned long flags;
        int ctxn, err;

        if (!task && cpu != -1) {
                /* Must be root to operate on a CPU event: */
                if (perf_paranoid_cpu() && !capable(CAP_SYS_ADMIN))
                        return ERR_PTR(-EACCES);

                if (cpu < 0 || cpu >= nr_cpumask_bits)
                        return ERR_PTR(-EINVAL);

                /*
                 * We could be clever and allow to attach a event to an
                 * offline CPU and activate it when the CPU comes up, but
                 * that's for later.
                 */
                if (!cpu_online(cpu))
                        return ERR_PTR(-ENODEV);

                cpuctx = per_cpu_ptr(pmu->pmu_cpu_context, cpu);
                ctx = &cpuctx->ctx;
                get_ctx(ctx);

                return ctx;
        }

        err = -EINVAL;
        ctxn = pmu->task_ctx_nr;
        if (ctxn < 0)
                goto errout;

retry:
        ctx = perf_lock_task_context(task, ctxn, &flags);
        if (ctx) {
                unclone_ctx(ctx);
                raw_spin_unlock_irqrestore(&ctx->lock, flags);
        }

        if (!ctx) {
                ctx = alloc_perf_context(pmu, task);
                err = -ENOMEM;
                if (!ctx)
                        goto errout;

                get_ctx(ctx);

                if (cmpxchg(&task->perf_event_ctxp[ctxn], NULL, ctx)) {
                        /*
                         * We raced with some other task; use
                         * the context they set.
                         */
                        put_task_struct(task);
                        kfree(ctx);
                        goto retry;
                }
        }

        return ctx;

errout:
        return ERR_PTR(err);
}

static void perf_event_free_filter(struct perf_event *event);

static void free_event_rcu(struct rcu_head *head)
{
        struct perf_event *event;

        event = container_of(head, struct perf_event, rcu_head);
        if (event->ns)
                put_pid_ns(event->ns);
        perf_event_free_filter(event);
        kfree(event);
}

static void perf_buffer_put(struct perf_buffer *buffer);

static void free_event(struct perf_event *event)
{
        irq_work_sync(&event->pending);

        if (!event->parent) {
                if (event->attach_state & PERF_ATTACH_TASK)
                        jump_label_dec(&perf_task_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 (event->buffer) {
                perf_buffer_put(event->buffer);
                event->buffer = NULL;
        }

        if (event->destroy)
                event->destroy(event);

        if (event->ctx)
                put_ctx(event->ctx);

        call_rcu(&event->rcu_head, free_event_rcu);
}

int perf_event_release_kernel(struct perf_event *event)
{
        struct perf_event_context *ctx = event->ctx;

        /*
         * Remove from the PMU, can't get re-enabled since we got
         * here because the last ref went.
         */
        perf_event_disable(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);
        raw_spin_lock_irq(&ctx->lock);
        perf_group_detach(event);
        list_del_event(event, ctx);
        raw_spin_unlock_irq(&ctx->lock);
        mutex_unlock(&ctx->mutex);

        free_event(event);

        return 0;
}
EXPORT_SYMBOL_GPL(perf_event_release_kernel);

/*
 * Called when the last reference to the file is gone.
 */
static int perf_release(struct inode *inode, struct file *file)
{
        struct perf_event *event = file->private_data;
        struct task_struct *owner;

        file->private_data = NULL;

        rcu_read_lock();
        owner = ACCESS_ONCE(event->owner);
        /*
         * Matches the smp_wmb() in perf_event_exit_task(). If we observe
         * !owner it means the list deletion is complete and we can indeed
         * free this event, otherwise we need to serialize on
         * owner->perf_event_mutex.
         */
        smp_read_barrier_depends();
        if (owner) {
                /*
                 * Since delayed_put_task_struct() also drops the last
                 * task reference we can safely take a new reference
                 * while holding the rcu_read_lock().
                 */
                get_task_struct(owner);
        }
        rcu_read_unlock();

        if (owner) {
                mutex_lock(&owner->perf_event_mutex);
                /*
                 * We have to re-check the event->owner field, if it is cleared
                 * we raced with perf_event_exit_task(), acquiring the mutex
                 * ensured they're done, and we can proceed with freeing the
                 * event.
                 */
                if (event->owner)
                        list_del_init(&event->owner_entry);
                mutex_unlock(&owner->perf_event_mutex);
                put_task_struct(owner);
        }

        return perf_event_release_kernel(event);
}

u64 perf_event_read_value(struct perf_event *event, u64 *enabled, u64 *running)
{
        struct perf_event *child;
        u64 total = 0;

        *enabled = 0;
        *running = 0;

        mutex_lock(&event->child_mutex);
        total += perf_event_read(event);
        *enabled += event->total_time_enabled +
                        atomic64_read(&event->child_total_time_enabled);
        *running += event->total_time_running +
                        atomic64_read(&event->child_total_time_running);

        list_for_each_entry(child, &event->child_list, child_list) {
                total += perf_event_read(child);
                *enabled += child->total_time_enabled;
                *running += child->total_time_running;
        }
        mutex_unlock(&event->child_mutex);

        return total;
}
EXPORT_SYMBOL_GPL(perf_event_read_value);

static int perf_event_read_group(struct perf_event *event,
                                   u64 read_format, char __user *buf)
{
        struct perf_event *leader = event->group_leader, *sub;
        int n = 0, size = 0, ret = -EFAULT;
        struct perf_event_context *ctx = leader->ctx;
        u64 values[5];
        u64 count, enabled, running;

        mutex_lock(&ctx->mutex);
        count = perf_event_read_value(leader, &enabled, &running);

        values[n++] = 1 + leader->nr_siblings;
        if (read_format & PERF_FORMAT_TOTAL_TIME_ENABLED)
                values[n++] = enabled;
        if (read_format & PERF_FORMAT_TOTAL_TIME_RUNNING)
                values[n++] = running;
        values[n++] = count;
        if (read_format & PERF_FORMAT_ID)
                values[n++] = primary_event_id(leader);

        size = n * sizeof(u64);

        if (copy_to_user(buf, values, size))
                goto unlock;

        ret = size;

        list_for_each_entry(sub, &leader->sibling_list, group_entry) {
                n = 0;

                values[n++] = perf_event_read_value(sub, &enabled, &running);
                if (read_format & PERF_FORMAT_ID)
                        values[n++] = primary_event_id(sub);

                size = n * sizeof(u64);

                if (copy_to_user(buf + ret, values, size)) {
                        ret = -EFAULT;
                        goto unlock;
                }

                ret += size;
        }
unlock:
        mutex_unlock(&ctx->mutex);

        return ret;
}

static int perf_event_read_one(struct perf_event *event,
                                 u64 read_format, char __user *buf)
{
        u64 enabled, running;
        u64 values[4];
        int n = 0;

        values[n++] = perf_event_read_value(event, &enabled, &running);
        if (read_format & PERF_FORMAT_TOTAL_TIME_ENABLED)
                values[n++] = enabled;
        if (read_format & PERF_FORMAT_TOTAL_TIME_RUNNING)
                values[n++] = running;
        if (read_format & PERF_FORMAT_ID)
                values[n++] = primary_event_id(event);

        if (copy_to_user(buf, values, n * sizeof(u64)))
                return -EFAULT;

        return n * sizeof(u64);
}

/*
 * Read the performance event - simple non blocking version for now
 */
static ssize_t
perf_read_hw(struct perf_event *event, char __user *buf, size_t count)
{
        u64 read_format = event->attr.read_format;
        int ret;

        /*
         * Return end-of-file for a read on a event that is in
         * error state (i.e. because it was pinned but it couldn't be
         * scheduled on to the CPU at some point).
         */
        if (event->state == PERF_EVENT_STATE_ERROR)
                return 0;

        if (count < event->read_size)
                return -ENOSPC;

        WARN_ON_ONCE(event->ctx->parent_ctx);
        if (read_format & PERF_FORMAT_GROUP)
                ret = perf_event_read_group(event, read_format, buf);
        else
                ret = perf_event_read_one(event, read_format, buf);

        return ret;
}

static ssize_t
perf_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
{
        struct perf_event *event = file->private_data;

        return perf_read_hw(event, buf, count);
}

static unsigned int perf_poll(struct file *file, poll_table *wait)
{
        struct perf_event *event = file->private_data;
        struct perf_buffer *buffer;
        unsigned int events = POLL_HUP;

        rcu_read_lock();
        buffer = rcu_dereference(event->buffer);
        if (buffer)
                events = atomic_xchg(&buffer->poll, 0);
        rcu_read_unlock();

        poll_wait(file, &event->waitq, wait);

        return events;
}

static void perf_event_reset(struct perf_event *event)
{
        (void)perf_event_read(event);
        local64_set(&event->count, 0);
        perf_event_update_userpage(event);
}

/*
 * Holding the top-level event's child_mutex means that any
 * descendant process that has inherited this event will block
 * in sync_child_event if it goes to exit, thus satisfying the
 * task existence requirements of perf_event_enable/disable.
 */
static void perf_event_for_each_child(struct perf_event *event,
                                        void (*func)(struct perf_event *))
{
        struct perf_event *child;

        WARN_ON_ONCE(event->ctx->parent_ctx);
        mutex_lock(&event->child_mutex);
        func(event);
        list_for_each_entry(child, &event->child_list, child_list)
                func(child);
        mutex_unlock(&event->child_mutex);
}

static void perf_event_for_each(struct perf_event *event,
                                  void (*func)(struct perf_event *))
{
        struct perf_event_context *ctx = event->ctx;
        struct perf_event *sibling;

        WARN_ON_ONCE(ctx->parent_ctx);
        mutex_lock(&ctx->mutex);
        event = event->group_leader;

        perf_event_for_each_child(event, func);
        func(event);
        list_for_each_entry(sibling, &event->sibling_list, group_entry)
                perf_event_for_each_child(event, func);
        mutex_unlock(&ctx->mutex);
}

static int perf_event_period(struct perf_event *event, u64 __user *arg)
{
        struct perf_event_context *ctx = event->ctx;
        int ret = 0;
        u64 value;

        if (!is_sampling_event(event))
                return -EINVAL;

        if (copy_from_user(&value, arg, sizeof(value)))
                return -EFAULT;

        if (!value)
                return -EINVAL;

        raw_spin_lock_irq(&ctx->lock);
        if (event->attr.freq) {
                if (value > sysctl_perf_event_sample_rate) {
                        ret = -EINVAL;
                        goto unlock;
                }

                event->attr.sample_freq = value;
        } else {
                event->attr.sample_period = value;
                event->hw.sample_period = value;
        }
unlock:
        raw_spin_unlock_irq(&ctx->lock);

        return ret;
}

static const struct file_operations perf_fops;

static struct perf_event *perf_fget_light(int fd, int *fput_needed)
{
        struct file *file;

        file = fget_light(fd, fput_needed);
        if (!file)
                return ERR_PTR(-EBADF);

        if (file->f_op != &perf_fops) {
                fput_light(file, *fput_needed);
                *fput_needed = 0;
                return ERR_PTR(-EBADF);
        }

        return file->private_data;
}

static int perf_event_set_output(struct perf_event *event,
                                 struct perf_event *output_event);
static int perf_event_set_filter(struct perf_event *event, void __user *arg);

static long perf_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
{
        struct perf_event *event = file->private_data;
        void (*func)(struct perf_event *);
        u32 flags = arg;

        switch (cmd) {
        case PERF_EVENT_IOC_ENABLE:
                func = perf_event_enable;
                break;
        case PERF_EVENT_IOC_DISABLE:
                func = perf_event_disable;
                break;
        case PERF_EVENT_IOC_RESET:
                func = perf_event_reset;
                break;

        case PERF_EVENT_IOC_REFRESH:
                return perf_event_refresh(event, arg);

        case PERF_EVENT_IOC_PERIOD:
                return perf_event_period(event, (u64 __user *)arg);

        case PERF_EVENT_IOC_SET_OUTPUT:
        {
                struct perf_event *output_event = NULL;
                int fput_needed = 0;
                int ret;

                if (arg != -1) {
                        output_event = perf_fget_light(arg, &fput_needed);
                        if (IS_ERR(output_event))
                                return PTR_ERR(output_event);
                }

                ret = perf_event_set_output(event, output_event);
                if (output_event)
                        fput_light(output_event->filp, fput_needed);

                return ret;
        }

        case PERF_EVENT_IOC_SET_FILTER:
                return perf_event_set_filter(event, (void __user *)arg);

        default:
                return -ENOTTY;
        }

        if (flags & PERF_IOC_FLAG_GROUP)
                perf_event_for_each(event, func);
        else
                perf_event_for_each_child(event, func);

        return 0;
}

int perf_event_task_enable(void)
{
        struct perf_event *event;

        mutex_lock(&current->perf_event_mutex);
        list_for_each_entry(event, &current->perf_event_list, owner_entry)
                perf_event_for_each_child(event, perf_event_enable);
        mutex_unlock(&current->perf_event_mutex);

        return 0;
}

int perf_event_task_disable(void)
{
        struct perf_event *event;

        mutex_lock(&current->perf_event_mutex);
        list_for_each_entry(event, &current->perf_event_list, owner_entry)
                perf_event_for_each_child(event, perf_event_disable);
        mutex_unlock(&current->perf_event_mutex);

        return 0;
}

#ifndef PERF_EVENT_INDEX_OFFSET
# define PERF_EVENT_INDEX_OFFSET 0
#endif

static int perf_event_index(struct perf_event *event)
{
        if (event->hw.state & PERF_HES_STOPPED)
                return 0;

        if (event->state != PERF_EVENT_STATE_ACTIVE)
                return 0;

        return event->hw.idx + 1 - PERF_EVENT_INDEX_OFFSET;
}

/*
 * Callers need to ensure there can be no nesting of this function, otherwise
 * the seqlock logic goes bad. We can not serialize this because the arch
 * code calls this from NMI context.
 */
void perf_event_update_userpage(struct perf_event *event)
{
        struct perf_event_mmap_page *userpg;
        struct perf_buffer *buffer;

        rcu_read_lock();
        buffer = rcu_dereference(event->buffer);
        if (!buffer)
                goto unlock;

        userpg = buffer->user_page;

        /*
         * Disable preemption so as to not let the corresponding user-space
         * spin too long if we get preempted.
         */
        preempt_disable();
        ++userpg->lock;
        barrier();
        userpg->index = perf_event_index(event);
        userpg->offset = perf_event_count(event);
        if (event->state == PERF_EVENT_STATE_ACTIVE)
                userpg->offset -= local64_read(&event->hw.prev_count);

        userpg->time_enabled = event->total_time_enabled +
                        atomic64_read(&event->child_total_time_enabled);

        userpg->time_running = event->total_time_running +
                        atomic64_read(&event->child_total_time_running);

        barrier();
        ++userpg->lock;
        preempt_enable();
unlock:
        rcu_read_unlock();
}

static unsigned long perf_data_size(struct perf_buffer *buffer);

static void
perf_buffer_init(struct perf_buffer *buffer, long watermark, int flags)
{
        long max_size = perf_data_size(buffer);

        if (watermark)
                buffer->watermark = min(max_size, watermark);

        if (!buffer->watermark)
                buffer->watermark = max_size / 2;

        if (flags & PERF_BUFFER_WRITABLE)
                buffer->writable = 1;

        atomic_set(&buffer->refcount, 1);
}

#ifndef CONFIG_PERF_USE_VMALLOC

/*
 * Back perf_mmap() with regular GFP_KERNEL-0 pages.
 */

static struct page *
perf_mmap_to_page(struct perf_buffer *buffer, unsigned long pgoff)
{
        if (pgoff > buffer->nr_pages)
                return NULL;

        if (pgoff == 0)
                return virt_to_page(buffer->user_page);

        return virt_to_page(buffer->data_pages[pgoff - 1]);
}

static void *perf_mmap_alloc_page(int cpu)
{
        struct page *page;
        int node;

        node = (cpu == -1) ? cpu : cpu_to_node(cpu);
        page = alloc_pages_node(node, GFP_KERNEL | __GFP_ZERO, 0);
        if (!page)
                return NULL;

        return page_address(page);
}

static struct perf_buffer *
perf_buffer_alloc(int nr_pages, long watermark, int cpu, int flags)
{
        struct perf_buffer *buffer;
        unsigned long size;
        int i;

        size = sizeof(struct perf_buffer);
        size += nr_pages * sizeof(void *);

        buffer = kzalloc(size, GFP_KERNEL);
        if (!buffer)
                goto fail;

        buffer->user_page = perf_mmap_alloc_page(cpu);
        if (!buffer->user_page)
                goto fail_user_page;

        for (i = 0; i < nr_pages; i++) {
                buffer->data_pages[i] = perf_mmap_alloc_page(cpu);
                if (!buffer->data_pages[i])
                        goto fail_data_pages;
        }

        buffer->nr_pages = nr_pages;

        perf_buffer_init(buffer, watermark, flags);

        return buffer;

fail_data_pages:
        for (i--; i >= 0; i--)
                free_page((unsigned long)buffer->data_pages[i]);

        free_page((unsigned long)buffer->user_page);

fail_user_page:
        kfree(buffer);

fail:
        return NULL;
}

static void perf_mmap_free_page(unsigned long addr)
{
        struct page *page = virt_to_page((void *)addr);

        page->mapping = NULL;
        __free_page(page);
}

static void perf_buffer_free(struct perf_buffer *buffer)
{
        int i;

        perf_mmap_free_page((unsigned long)buffer->user_page);
        for (i = 0; i < buffer->nr_pages; i++)
                perf_mmap_free_page((unsigned long)buffer->data_pages[i]);
        kfree(buffer);
}

static inline int page_order(struct perf_buffer *buffer)
{
        return 0;
}

#else

/*
 * Back perf_mmap() with vmalloc memory.
 *
 * Required for architectures that have d-cache aliasing issues.
 */

static inline int page_order(struct perf_buffer *buffer)
{
        return buffer->page_order;
}

static struct page *
perf_mmap_to_page(struct perf_buffer *buffer, unsigned long pgoff)
{
        if (pgoff > (1UL << page_order(buffer)))
                return NULL;

        return vmalloc_to_page((void *)buffer->user_page + pgoff * PAGE_SIZE);
}

static void perf_mmap_unmark_page(void *addr)
{
        struct page *page = vmalloc_to_page(addr);

        page->mapping = NULL;
}

static void perf_buffer_free_work(struct work_struct *work)
{
        struct perf_buffer *buffer;
        void *base;
        int i, nr;

        buffer = container_of(work, struct perf_buffer, work);
        nr = 1 << page_order(buffer);

        base = buffer->user_page;
        for (i = 0; i < nr + 1; i++)
                perf_mmap_unmark_page(base + (i * PAGE_SIZE));

        vfree(base);
        kfree(buffer);
}

static void perf_buffer_free(struct perf_buffer *buffer)
{
        schedule_work(&buffer->work);
}

static struct perf_buffer *
perf_buffer_alloc(int nr_pages, long watermark, int cpu, int flags)
{
        struct perf_buffer *buffer;
        unsigned long size;
        void *all_buf;

        size = sizeof(struct perf_buffer);
        size += sizeof(void *);

        buffer = kzalloc(size, GFP_KERNEL);
        if (!buffer)
                goto fail;

        INIT_WORK(&buffer->work, perf_buffer_free_work);

        all_buf = vmalloc_user((nr_pages + 1) * PAGE_SIZE);
        if (!all_buf)
                goto fail_all_buf;

        buffer->user_page = all_buf;
        buffer->data_pages[0] = all_buf + PAGE_SIZE;
        buffer->page_order = ilog2(nr_pages);
        buffer->nr_pages = 1;

        perf_buffer_init(buffer, watermark, flags);

        return buffer;

fail_all_buf:
        kfree(buffer);

fail:
        return NULL;
}

#endif

static unsigned long perf_data_size(struct perf_buffer *buffer)
{
        return buffer->nr_pages << (PAGE_SHIFT + page_order(buffer));
}

static int perf_mmap_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
{
        struct perf_event *event = vma->vm_file->private_data;
        struct perf_buffer *buffer;
        int ret = VM_FAULT_SIGBUS;

        if (vmf->flags & FAULT_FLAG_MKWRITE) {
                if (vmf->pgoff == 0)
                        ret = 0;
                return ret;
        }

        rcu_read_lock();
        buffer = rcu_dereference(event->buffer);
        if (!buffer)
                goto unlock;

        if (vmf->pgoff && (vmf->flags & FAULT_FLAG_WRITE))
                goto unlock;

        vmf->page = perf_mmap_to_page(buffer, vmf->pgoff);
        if (!vmf->page)
                goto unlock;

        get_page(vmf->page);
        vmf->page->mapping = vma->vm_file->f_mapping;
        vmf->page->index   = vmf->pgoff;

        ret = 0;
unlock:
        rcu_read_unlock();

        return ret;
}

static void perf_buffer_free_rcu(struct rcu_head *rcu_head)
{
        struct perf_buffer *buffer;

        buffer = container_of(rcu_head, struct perf_buffer, rcu_head);
        perf_buffer_free(buffer);
}

static struct perf_buffer *perf_buffer_get(struct perf_event *event)
{
        struct perf_buffer *buffer;

        rcu_read_lock();
        buffer = rcu_dereference(event->buffer);
        if (buffer) {
                if (!atomic_inc_not_zero(&buffer->refcount))
                        buffer = NULL;
        }
        rcu_read_unlock();

        return buffer;
}

static void perf_buffer_put(struct perf_buffer *buffer)
{
        if (!atomic_dec_and_test(&buffer->refcount))
                return;

        call_rcu(&buffer->rcu_head, perf_buffer_free_rcu);
}

static void perf_mmap_open(struct vm_area_struct *vma)
{
        struct perf_event *event = vma->vm_file->private_data;

        atomic_inc(&event->mmap_count);
}

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->buffer);
                struct user_struct *user = event->mmap_user;
                struct perf_buffer *buffer = event->buffer;

                atomic_long_sub((size >> PAGE_SHIFT) + 1, &user->locked_vm);
                vma->vm_mm->locked_vm -= event->mmap_locked;
                rcu_assign_pointer(event->buffer, NULL);
                mutex_unlock(&event->mmap_mutex);

                perf_buffer_put(buffer);
                free_uid(user);
        }
}

static const struct vm_operations_struct perf_mmap_vmops = {
        .open           = perf_mmap_open,
        .close          = perf_mmap_close,
        .fault          = perf_mmap_fault,
        .page_mkwrite   = perf_mmap_fault,
};

static int perf_mmap(struct file *file, struct vm_area_struct *vma)
{
        struct perf_event *event = file->private_data;
        unsigned long user_locked, user_lock_limit;
        struct user_struct *user = current_user();
        unsigned long locked, lock_limit;
        struct perf_buffer *buffer;
        unsigned long vma_size;
        unsigned long nr_pages;
        long user_extra, extra;
        int ret = 0, flags = 0;

        /*
         * Don't allow mmap() of inherited per-task counters. This would
         * create a performance issue due to all children writing to the
         * same buffer.
         */
        if (event->cpu == -1 && event->attr.inherit)
                return -EINVAL;

        if (!(vma->vm_flags & VM_SHARED))
                return -EINVAL;

        vma_size = vma->vm_end - vma->vm_start;
        nr_pages = (vma_size / PAGE_SIZE) - 1;

        /*
         * If we have buffer pages ensure they're a power-of-two number, so we
         * can do bitmasks instead of modulo.
         */
        if (nr_pages != 0 && !is_power_of_2(nr_pages))
                return -EINVAL;

        if (vma_size != PAGE_SIZE * (1 + nr_pages))
                return -EINVAL;

        if (vma->vm_pgoff != 0)
                return -EINVAL;

        WARN_ON_ONCE(event->ctx->parent_ctx);
        mutex_lock(&event->mmap_mutex);
        if (event->buffer) {
                if (event->buffer->nr_pages == nr_pages)
                        atomic_inc(&event->buffer->refcount);
                else
                        ret = -EINVAL;
                goto unlock;
        }

        user_extra = nr_pages + 1;
        user_lock_limit = sysctl_perf_event_mlock >> (PAGE_SHIFT - 10);

        /*
         * Increase the limit linearly with more CPUs:
         */
        user_lock_limit *= num_online_cpus();

        user_locked = atomic_long_read(&user->locked_vm) + user_extra;

        extra = 0;
        if (user_locked > user_lock_limit)
                extra = user_locked - user_lock_limit;

        lock_limit = rlimit(RLIMIT_MEMLOCK);
        lock_limit >>= PAGE_SHIFT;
        locked = vma->vm_mm->locked_vm + extra;

        if ((locked > lock_limit) && perf_paranoid_tracepoint_raw() &&
                !capable(CAP_IPC_LOCK)) {
                ret = -EPERM;
                goto unlock;
        }

        WARN_ON(event->buffer);

        if (vma->vm_flags & VM_WRITE)
                flags |= PERF_BUFFER_WRITABLE;

        buffer = perf_buffer_alloc(nr_pages, event->attr.wakeup_watermark,
                                   event->cpu, flags);
        if (!buffer) {
                ret = -ENOMEM;
                goto unlock;
        }
        rcu_assign_pointer(event->buffer, buffer);

        atomic_long_add(user_extra, &user->locked_vm);
        event->mmap_locked = extra;
        event->mmap_user = get_current_user();
        vma->vm_mm->locked_vm += event->mmap_locked;

unlock:
        if (!ret)
                atomic_inc(&event->mmap_count);
        mutex_unlock(&event->mmap_mutex);

        vma->vm_flags |= VM_RESERVED;
        vma->vm_ops = &perf_mmap_vmops;

        return ret;
}

static int perf_fasync(int fd, struct file *filp, int on)
{
        struct inode *inode = filp->f_path.dentry->d_inode;
        struct perf_event *event = filp->private_data;
        int retval;

        mutex_lock(&inode->i_mutex);
        retval = fasync_helper(fd, filp, on, &event->fasync);
        mutex_unlock(&inode->i_mutex);

        if (retval < 0)
                return retval;

        return 0;
}

static const struct file_operations perf_fops = {
        .llseek                 = no_llseek,
        .release                = perf_release,
        .read                   = perf_read,
        .poll                   = perf_poll,
        .unlocked_ioctl         = perf_ioctl,
        .compat_ioctl           = perf_ioctl,
        .mmap                   = perf_mmap,
        .fasync                 = perf_fasync,
};

/*
 * Perf event wakeup
 *
 * If there's data, ensure we set the poll() state and publish everything
 * to user-space before waking everybody up.
 */

void perf_event_wakeup(struct perf_event *event)
{
        wake_up_all(&event->waitq);

        if (event->pending_kill) {
                kill_fasync(&event->fasync, SIGIO, event->pending_kill);
                event->pending_kill = 0;
        }
}

static void perf_pending_event(struct irq_work *entry)
{
        struct perf_event *event = container_of(entry,
                        struct perf_event, pending);

        if (event->pending_disable) {
                event->pending_disable = 0;
                __perf_event_disable(event);
        }

        if (event->pending_wakeup) {
                event->pending_wakeup = 0;
                perf_event_wakeup(event);
        }
}

/*
 * We assume there is only KVM supporting the callbacks.
 * Later on, we might change it to a list if there is
 * another virtualization implementation supporting the callbacks.
 */
struct perf_guest_info_callbacks *perf_guest_cbs;

int perf_register_guest_info_callbacks(struct perf_guest_info_callbacks *cbs)
{
        perf_guest_cbs = cbs;
        return 0;
}
EXPORT_SYMBOL_GPL(perf_register_guest_info_callbacks);

int perf_unregister_guest_info_callbacks(struct perf_guest_info_callbacks *cbs)
{
        perf_guest_cbs = NULL;
        return 0;
}
EXPORT_SYMBOL_GPL(perf_unregister_guest_info_callbacks);

/*
 * Output
 */
static bool perf_output_space(struct perf_buffer *buffer, unsigned long tail,
                              unsigned long offset, unsigned long head)
{
        unsigned long mask;

        if (!buffer->writable)
                return true;

        mask = perf_data_size(buffer) - 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->buffer->poll, POLL_IN);

        if (handle->nmi) {
                handle->event->pending_wakeup = 1;
                irq_work_queue(&handle->event->pending);
        } else
                perf_event_wakeup(handle->event);
}

/*
 * We need to ensure a later event_id doesn't publish a head when a former
 * event isn't done writing. However since we need to deal with NMIs we
 * cannot fully serialize things.
 *
 * We only publish the head (and generate a wakeup) when the outer-most
 * event completes.
 */
static void perf_output_get_handle(struct perf_output_handle *handle)
{
        struct perf_buffer *buffer = handle->buffer;

        preempt_disable();
        local_inc(&buffer->nest);
        handle->wakeup = local_read(&buffer->wakeup);
}

static void perf_output_put_handle(struct perf_output_handle *handle)
{
        struct perf_buffer *buffer = handle->buffer;
        unsigned long head;

again:
        head = local_read(&buffer->head);

        /*
         * IRQ/NMI can happen here, which means we can miss a head update.
         */

        if (!local_dec_and_test(&buffer->nest))
                goto out;

        /*
         * Publish the known good head. Rely on the full barrier implied
         * by atomic_dec_and_test() order the buffer->head read and this
         * write.
         */
        buffer->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 buffer->head.
         */
        if (unlikely(head != local_read(&buffer->head))) {
                local_inc(&buffer->nest);
                goto again;
        }

        if (handle->wakeup != local_read(&buffer->wakeup))
                perf_output_wakeup(handle);

out:
        preempt_enable();
}

__always_inline void perf_output_copy(struct perf_output_handle *handle,
                      const void *buf, unsigned int len)
{
        do {
                unsigned long size = min_t(unsigned long, handle->size, len);

                memcpy(handle->addr, buf, size);

                len -= size;
                handle->addr += size;
                buf += size;
                handle->size -= size;
                if (!handle->size) {
                        struct perf_buffer *buffer = handle->buffer;

                        handle->page++;
                        handle->page &= buffer->nr_pages - 1;
                        handle->addr = buffer->data_pages[handle->page];
                        handle->size = PAGE_SIZE << page_order(buffer);
                }
        } while (len);
}

static void __perf_event_header__init_id(struct perf_event_header *header,
                                         struct perf_sample_data *data,