1 files changed, 311 insertions, 205 deletions

diff --git a/kernel/perf_counter.c b/kernel/perf_counter.c
index a641eb753b8c..b0b20a07f394 100644
--- a/kernel/perf_counter.c
+++ b/kernel/perf_counter.c
@@ -42,6 +42,7 @@ static int perf_overcommit __read_mostly = 1;
 static atomic_t nr_counters __read_mostly;
 static atomic_t nr_mmap_counters __read_mostly;
 static atomic_t nr_comm_counters __read_mostly;
+static atomic_t nr_task_counters __read_mostly;
 
 /*
  * perf counter paranoia level:
@@ -146,6 +147,28 @@ static void put_ctx(struct perf_counter_context *ctx)
         }
 }
 
+static void unclone_ctx(struct perf_counter_context *ctx)
+{
+        if (ctx->parent_ctx) {
+                put_ctx(ctx->parent_ctx);
+                ctx->parent_ctx = NULL;
+        }
+}
+
+/*
+ * If we inherit counters we want to return the parent counter id
+ * to userspace.
+ */
+static u64 primary_counter_id(struct perf_counter *counter)
+{
+        u64 id = counter->id;
+
+        if (counter->parent)
+                id = counter->parent->id;
+
+        return id;
+}
+
 /*
  * Get the perf_counter_context for a task and lock it.
  * This has to cope with with the fact that until it is locked,
@@ -1081,7 +1104,7 @@ static void perf_counter_sync_stat(struct perf_counter_context *ctx,
                 __perf_counter_sync_stat(counter, next_counter);
 
                 counter = list_next_entry(counter, event_entry);
-                next_counter = list_next_entry(counter, event_entry);
+                next_counter = list_next_entry(next_counter, event_entry);
         }
 }
 
@@ -1288,7 +1311,6 @@ static void perf_counter_cpu_sched_in(struct perf_cpu_context *cpuctx, int cpu)
 #define MAX_INTERRUPTS (~0ULL)
 
 static void perf_log_throttle(struct perf_counter *counter, int enable);
-static void perf_log_period(struct perf_counter *counter, u64 period);
 
 static void perf_adjust_period(struct perf_counter *counter, u64 events)
 {
@@ -1307,8 +1329,6 @@ static void perf_adjust_period(struct perf_counter *counter, u64 events)
         if (!sample_period)
                 sample_period = 1;
 
-        perf_log_period(counter, sample_period);
-
         hwc->sample_period = sample_period;
 }
 
@@ -1463,10 +1483,8 @@ static void perf_counter_enable_on_exec(struct task_struct *task)
         /*
          * Unclone this context if we enabled any counter.
          */
-        if (enabled && ctx->parent_ctx) {
-                put_ctx(ctx->parent_ctx);
-                ctx->parent_ctx = NULL;
-        }
+        if (enabled)
+                unclone_ctx(ctx);
 
         spin_unlock(&ctx->lock);
 
@@ -1526,7 +1544,6 @@ __perf_counter_init_context(struct perf_counter_context *ctx,
 
 static struct perf_counter_context *find_get_context(pid_t pid, int cpu)
 {
-        struct perf_counter_context *parent_ctx;
         struct perf_counter_context *ctx;
         struct perf_cpu_context *cpuctx;
         struct task_struct *task;
@@ -1586,11 +1603,7 @@ static struct perf_counter_context *find_get_context(pid_t pid, int cpu)
  retry:
         ctx = perf_lock_task_context(task, &flags);
         if (ctx) {
-                parent_ctx = ctx->parent_ctx;
-                if (parent_ctx) {
-                        put_ctx(parent_ctx);
-                        ctx->parent_ctx = NULL;         /* no longer a clone */
-                }
+                unclone_ctx(ctx);
                 spin_unlock_irqrestore(&ctx->lock, flags);
         }
 
@@ -1642,6 +1655,8 @@ static void free_counter(struct perf_counter *counter)
                         atomic_dec(&nr_mmap_counters);
                 if (counter->attr.comm)
                         atomic_dec(&nr_comm_counters);
+                if (counter->attr.task)
+                        atomic_dec(&nr_task_counters);
         }
 
         if (counter->destroy)
@@ -1676,6 +1691,18 @@ static int perf_release(struct inode *inode, struct file *file)
         return 0;
 }
 
+static u64 perf_counter_read_tree(struct perf_counter *counter)
+{
+        struct perf_counter *child;
+        u64 total = 0;
+
+        total += perf_counter_read(counter);
+        list_for_each_entry(child, &counter->child_list, child_list)
+                total += perf_counter_read(child);
+
+        return total;
+}
+
 /*
  * Read the performance counter - simple non blocking version for now
  */
@@ -1695,7 +1722,7 @@ perf_read_hw(struct perf_counter *counter, char __user *buf, size_t count)
 
         WARN_ON_ONCE(counter->ctx->parent_ctx);
         mutex_lock(&counter->child_mutex);
-        values[0] = perf_counter_read(counter);
+        values[0] = perf_counter_read_tree(counter);
         n = 1;
         if (counter->attr.read_format & PERF_FORMAT_TOTAL_TIME_ENABLED)
                 values[n++] = counter->total_time_enabled +
@@ -1704,7 +1731,7 @@ perf_read_hw(struct perf_counter *counter, char __user *buf, size_t count)
                 values[n++] = counter->total_time_running +
                         atomic64_read(&counter->child_total_time_running);
         if (counter->attr.read_format & PERF_FORMAT_ID)
-                values[n++] = counter->id;
+                values[n++] = primary_counter_id(counter);
         mutex_unlock(&counter->child_mutex);
 
         if (count < n * sizeof(u64))
@@ -1811,8 +1838,6 @@ static int perf_counter_period(struct perf_counter *counter, u64 __user *arg)
 
                 counter->attr.sample_freq = value;
         } else {
-                perf_log_period(counter, value);
-
                 counter->attr.sample_period = value;
                 counter->hw.sample_period = value;
         }
@@ -2661,10 +2686,14 @@ static void perf_counter_output(struct perf_counter *counter, int nmi,
         if (sample_type & PERF_SAMPLE_ID)
                 header.size += sizeof(u64);
 
+        if (sample_type & PERF_SAMPLE_STREAM_ID)
+                header.size += sizeof(u64);
+
         if (sample_type & PERF_SAMPLE_CPU) {
                 header.size += sizeof(cpu_entry);
 
                 cpu_entry.cpu = raw_smp_processor_id();
+                cpu_entry.reserved = 0;
         }
 
         if (sample_type & PERF_SAMPLE_PERIOD)
@@ -2685,6 +2714,18 @@ static void perf_counter_output(struct perf_counter *counter, int nmi,
                         header.size += sizeof(u64);
         }
 
+        if (sample_type & PERF_SAMPLE_RAW) {
+                int size = sizeof(u32);
+
+                if (data->raw)
+                        size += data->raw->size;
+                else
+                        size += sizeof(u32);
+
+                WARN_ON_ONCE(size & (sizeof(u64)-1));
+                header.size += size;
+        }
+
         ret = perf_output_begin(&handle, counter, header.size, nmi, 1);
         if (ret)
                 return;
@@ -2703,7 +2744,13 @@ static void perf_counter_output(struct perf_counter *counter, int nmi,
         if (sample_type & PERF_SAMPLE_ADDR)
                 perf_output_put(&handle, data->addr);
 
-        if (sample_type & PERF_SAMPLE_ID)
+        if (sample_type & PERF_SAMPLE_ID) {
+                u64 id = primary_counter_id(counter);
+
+                perf_output_put(&handle, id);
+        }
+
+        if (sample_type & PERF_SAMPLE_STREAM_ID)
                 perf_output_put(&handle, counter->id);
 
         if (sample_type & PERF_SAMPLE_CPU)
@@ -2726,7 +2773,7 @@ static void perf_counter_output(struct perf_counter *counter, int nmi,
                         if (sub != counter)
                                 sub->pmu->read(sub);
 
-                        group_entry.id = sub->id;
+                        group_entry.id = primary_counter_id(sub);
                         group_entry.counter = atomic64_read(&sub->count);
 
                         perf_output_put(&handle, group_entry);
@@ -2742,6 +2789,22 @@ static void perf_counter_output(struct perf_counter *counter, int nmi,
                 }
         }
 
+        if (sample_type & PERF_SAMPLE_RAW) {
+                if (data->raw) {
+                        perf_output_put(&handle, data->raw->size);
+                        perf_output_copy(&handle, data->raw->data, data->raw->size);
+                } else {
+                        struct {
+                                u32     size;
+                                u32     data;
+                        } raw = {
+                                .size = sizeof(u32),
+                                .data = 0,
+                        };
+                        perf_output_put(&handle, raw);
+                }
+        }
+
         perf_output_end(&handle);
 }
 
@@ -2786,15 +2849,8 @@ perf_counter_read_event(struct perf_counter *counter,
         }
 
         if (counter->attr.read_format & PERF_FORMAT_ID) {
-                u64 id;
-
                 event.header.size += sizeof(u64);
-                if (counter->parent)
-                        id = counter->parent->id;
-                else
-                        id = counter->id;
-
-                event.format[i++] = id;
+                event.format[i++] = primary_counter_id(counter);
         }
 
         ret = perf_output_begin(&handle, counter, event.header.size, 0, 0);
@@ -2806,48 +2862,56 @@ perf_counter_read_event(struct perf_counter *counter,
 }
 
 /*
- * fork tracking
+ * task tracking -- fork/exit
+ *
+ * enabled by: attr.comm | attr.mmap | attr.task
  */
 
-struct perf_fork_event {
-        struct task_struct      *task;
+struct perf_task_event {
+        struct task_struct              *task;
+        struct perf_counter_context     *task_ctx;
 
         struct {
                 struct perf_event_header        header;
 
                 u32                             pid;
                 u32                             ppid;
+                u32                             tid;
+                u32                             ptid;
         } event;
 };
 
-static void perf_counter_fork_output(struct perf_counter *counter,
-                                     struct perf_fork_event *fork_event)
+static void perf_counter_task_output(struct perf_counter *counter,
+                                     struct perf_task_event *task_event)
 {
         struct perf_output_handle handle;
-        int size = fork_event->event.header.size;
-        struct task_struct *task = fork_event->task;
+        int size = task_event->event.header.size;
+        struct task_struct *task = task_event->task;
         int ret = perf_output_begin(&handle, counter, size, 0, 0);
 
         if (ret)
                 return;
 
-        fork_event->event.pid = perf_counter_pid(counter, task);
-        fork_event->event.ppid = perf_counter_pid(counter, task->real_parent);
+        task_event->event.pid = perf_counter_pid(counter, task);
+        task_event->event.ppid = perf_counter_pid(counter, task->real_parent);
 
-        perf_output_put(&handle, fork_event->event);
+        task_event->event.tid = perf_counter_tid(counter, task);
+        task_event->event.ptid = perf_counter_tid(counter, task->real_parent);
+
+        perf_output_put(&handle, task_event->event);
         perf_output_end(&handle);
 }
 
-static int perf_counter_fork_match(struct perf_counter *counter)
+static int perf_counter_task_match(struct perf_counter *counter)
 {
-        if (counter->attr.comm || counter->attr.mmap)
+        if (counter->attr.comm || counter->attr.mmap || counter->attr.task)
                 return 1;
 
         return 0;
 }
 
-static void perf_counter_fork_ctx(struct perf_counter_context *ctx,
-                                  struct perf_fork_event *fork_event)
+static void perf_counter_task_ctx(struct perf_counter_context *ctx,
+                                  struct perf_task_event *task_event)
 {
         struct perf_counter *counter;
 
@@ -2856,51 +2920,62 @@ static void perf_counter_fork_ctx(struct perf_counter_context *ctx,
 
         rcu_read_lock();
         list_for_each_entry_rcu(counter, &ctx->event_list, event_entry) {
-                if (perf_counter_fork_match(counter))
-                        perf_counter_fork_output(counter, fork_event);
+                if (perf_counter_task_match(counter))
+                        perf_counter_task_output(counter, task_event);
         }
         rcu_read_unlock();
 }
 
-static void perf_counter_fork_event(struct perf_fork_event *fork_event)
+static void perf_counter_task_event(struct perf_task_event *task_event)
 {
         struct perf_cpu_context *cpuctx;
-        struct perf_counter_context *ctx;
+        struct perf_counter_context *ctx = task_event->task_ctx;
 
         cpuctx = &get_cpu_var(perf_cpu_context);
-        perf_counter_fork_ctx(&cpuctx->ctx, fork_event);
+        perf_counter_task_ctx(&cpuctx->ctx, task_event);
         put_cpu_var(perf_cpu_context);
 
         rcu_read_lock();
-        /*
-         * doesn't really matter which of the child contexts the
-         * events ends up in.
-         */
-        ctx = rcu_dereference(current->perf_counter_ctxp);
+        if (!ctx)
+                ctx = rcu_dereference(task_event->task->perf_counter_ctxp);
         if (ctx)
-                perf_counter_fork_ctx(ctx, fork_event);
+                perf_counter_task_ctx(ctx, task_event);
         rcu_read_unlock();
 }
 
-void perf_counter_fork(struct task_struct *task)
+static void perf_counter_task(struct task_struct *task,
+                              struct perf_counter_context *task_ctx,
+                              int new)
 {
-        struct perf_fork_event fork_event;
+        struct perf_task_event task_event;
 
         if (!atomic_read(&nr_comm_counters) &&
-            !atomic_read(&nr_mmap_counters))
+            !atomic_read(&nr_mmap_counters) &&
+            !atomic_read(&nr_task_counters))
                 return;
 
-        fork_event = (struct perf_fork_event){
-                .task   = task,
-                .event  = {
+        task_event = (struct perf_task_event){
+                .task     = task,
+                .task_ctx = task_ctx,
+                .event    = {
                         .header = {
-                                .type = PERF_EVENT_FORK,
-                                .size = sizeof(fork_event.event),
+                                .type = new ? PERF_EVENT_FORK : PERF_EVENT_EXIT,
+                                .misc = 0,
+                                .size = sizeof(task_event.event),
                         },
+                        /* .pid  */
+                        /* .ppid */
+                        /* .tid  */
+                        /* .ptid */
                 },
         };
 
-        perf_counter_fork_event(&fork_event);
+        perf_counter_task_event(&task_event);
+}
+
+void perf_counter_fork(struct task_struct *task)
+{
+        perf_counter_task(task, NULL, 1);
 }
 
 /*
@@ -2968,8 +3043,10 @@ static void perf_counter_comm_event(struct perf_comm_event *comm_event)
         struct perf_cpu_context *cpuctx;
         struct perf_counter_context *ctx;
         unsigned int size;
-        char *comm = comm_event->task->comm;
+        char comm[TASK_COMM_LEN];
 
+        memset(comm, 0, sizeof(comm));
+        strncpy(comm, comm_event->task->comm, sizeof(comm));
         size = ALIGN(strlen(comm)+1, sizeof(u64));
 
         comm_event->comm = comm;
@@ -3004,8 +3081,16 @@ void perf_counter_comm(struct task_struct *task)
 
         comm_event = (struct perf_comm_event){
                 .task   = task,
+                /* .comm      */
+                /* .comm_size */
                 .event  = {
-                        .header = { .type = PERF_EVENT_COMM, },
+                        .header = {
+                                .type = PERF_EVENT_COMM,
+                                .misc = 0,
+                                /* .size */
+                        },
+                        /* .pid */
+                        /* .tid */
                 },
         };
 
@@ -3088,8 +3173,15 @@ static void perf_counter_mmap_event(struct perf_mmap_event *mmap_event)
         char *buf = NULL;
         const char *name;
 
+        memset(tmp, 0, sizeof(tmp));
+
         if (file) {
-                buf = kzalloc(PATH_MAX, GFP_KERNEL);
+                /*
+                 * d_path works from the end of the buffer backwards, so we
+                 * need to add enough zero bytes after the string to handle
+                 * the 64bit alignment we do later.
+                 */
+                buf = kzalloc(PATH_MAX + sizeof(u64), GFP_KERNEL);
                 if (!buf) {
                         name = strncpy(tmp, "//enomem", sizeof(tmp));
                         goto got_name;
@@ -3100,9 +3192,11 @@ static void perf_counter_mmap_event(struct perf_mmap_event *mmap_event)
                         goto got_name;
                 }
         } else {
-                name = arch_vma_name(mmap_event->vma);
-                if (name)
+                if (arch_vma_name(mmap_event->vma)) {
+                        name = strncpy(tmp, arch_vma_name(mmap_event->vma),
+                                       sizeof(tmp));
                         goto got_name;
+                }
 
                 if (!vma->vm_mm) {
                         name = strncpy(tmp, "[vdso]", sizeof(tmp));
@@ -3147,8 +3241,16 @@ void __perf_counter_mmap(struct vm_area_struct *vma)
 
         mmap_event = (struct perf_mmap_event){
                 .vma    = vma,
+                /* .file_name */
+                /* .file_size */
                 .event  = {
-                        .header = { .type = PERF_EVENT_MMAP, },
+                        .header = {
+                                .type = PERF_EVENT_MMAP,
+                                .misc = 0,
+                                /* .size */
+                        },
+                        /* .pid */
+                        /* .tid */
                         .start  = vma->vm_start,
                         .len    = vma->vm_end - vma->vm_start,
                         .pgoff  = vma->vm_pgoff,
@@ -3159,49 +3261,6 @@ void __perf_counter_mmap(struct vm_area_struct *vma)
 }
 
 /*
- * Log sample_period changes so that analyzing tools can re-normalize the
- * event flow.
- */
-
-struct freq_event {
-        struct perf_event_header        header;
-        u64                             time;
-        u64                             id;
-        u64                             period;
-};
-
-static void perf_log_period(struct perf_counter *counter, u64 period)
-{
-        struct perf_output_handle handle;
-        struct freq_event event;
-        int ret;
-
-        if (counter->hw.sample_period == period)
-                return;
-
-        if (counter->attr.sample_type & PERF_SAMPLE_PERIOD)
-                return;
-
-        event = (struct freq_event) {
-                .header = {
-                        .type = PERF_EVENT_PERIOD,
-                        .misc = 0,
-                        .size = sizeof(event),
-                },
-                .time = sched_clock(),
-                .id = counter->id,
-                .period = period,
-        };
-
-        ret = perf_output_begin(&handle, counter, sizeof(event), 1, 0);
-        if (ret)
-                return;
-
-        perf_output_put(&handle, event);
-        perf_output_end(&handle);
-}
-
-/*
  * IRQ throttle logging
  */
 
@@ -3214,16 +3273,21 @@ static void perf_log_throttle(struct perf_counter *counter, int enable)
                 struct perf_event_header        header;
                 u64                             time;
                 u64                             id;
+                u64                             stream_id;
         } throttle_event = {
                 .header = {
-                        .type = PERF_EVENT_THROTTLE + 1,
+                        .type = PERF_EVENT_THROTTLE,
                         .misc = 0,
                         .size = sizeof(throttle_event),
                 },
-                .time   = sched_clock(),
-                .id     = counter->id,
+                .time           = sched_clock(),
+                .id             = primary_counter_id(counter),
+                .stream_id      = counter->id,
         };
 
+        if (enable)
+                throttle_event.header.type = PERF_EVENT_UNTHROTTLE;
+
         ret = perf_output_begin(&handle, counter, sizeof(throttle_event), 1, 0);
         if (ret)
                 return;
@@ -3300,87 +3364,81 @@ int perf_counter_overflow(struct perf_counter *counter, int nmi,
  * Generic software counter infrastructure
  */
 
-static void perf_swcounter_update(struct perf_counter *counter)
+/*
+ * We directly increment counter->count and keep a second value in
+ * counter->hw.period_left to count intervals. This period counter
+ * is kept in the range [-sample_period, 0] so that we can use the
+ * sign as trigger.
+ */
+
+static u64 perf_swcounter_set_period(struct perf_counter *counter)
 {
         struct hw_perf_counter *hwc = &counter->hw;
-        u64 prev, now;
-        s64 delta;
+        u64 period = hwc->last_period;
+        u64 nr, offset;
+        s64 old, val;
+
+        hwc->last_period = hwc->sample_period;
 
 again:
-        prev = atomic64_read(&hwc->prev_count);
-        now = atomic64_read(&hwc->count);
-        if (atomic64_cmpxchg(&hwc->prev_count, prev, now) != prev)
-                goto again;
+        old = val = atomic64_read(&hwc->period_left);
+        if (val < 0)
+                return 0;
 
-        delta = now - prev;
+        nr = div64_u64(period + val, period);
+        offset = nr * period;
+        val -= offset;
+        if (atomic64_cmpxchg(&hwc->period_left, old, val) != old)
+                goto again;
 
-        atomic64_add(delta, &counter->count);
-        atomic64_sub(delta, &hwc->period_left);
+        return nr;
 }
 
-static void perf_swcounter_set_period(struct perf_counter *counter)
+static void perf_swcounter_overflow(struct perf_counter *counter,
+                                    int nmi, struct perf_sample_data *data)
 {
         struct hw_perf_counter *hwc = &counter->hw;
-        s64 left = atomic64_read(&hwc->period_left);
-        s64 period = hwc->sample_period;
+        u64 overflow;
 
-        if (unlikely(left <= -period)) {
-                left = period;
-                atomic64_set(&hwc->period_left, left);
-                hwc->last_period = period;
-        }
+        data->period = counter->hw.last_period;
+        overflow = perf_swcounter_set_period(counter);
 
-        if (unlikely(left <= 0)) {
-                left += period;
-                atomic64_add(period, &hwc->period_left);
-                hwc->last_period = period;
-        }
+        if (hwc->interrupts == MAX_INTERRUPTS)
+                return;
 
-        atomic64_set(&hwc->prev_count, -left);
-        atomic64_set(&hwc->count, -left);
+        for (; overflow; overflow--) {
+                if (perf_counter_overflow(counter, nmi, data)) {
+                        /*
+                         * We inhibit the overflow from happening when
+                         * hwc->interrupts == MAX_INTERRUPTS.
+                         */
+                        break;
+                }
+        }
 }
 
-static enum hrtimer_restart perf_swcounter_hrtimer(struct hrtimer *hrtimer)
+static void perf_swcounter_unthrottle(struct perf_counter *counter)
 {
-        enum hrtimer_restart ret = HRTIMER_RESTART;
-        struct perf_sample_data data;
-        struct perf_counter *counter;
-        u64 period;
-
-        counter = container_of(hrtimer, struct perf_counter, hw.hrtimer);
-        counter->pmu->read(counter);
-
-        data.addr = 0;
-        data.regs = get_irq_regs();
         /*
-         * In case we exclude kernel IPs or are somehow not in interrupt
-         * context, provide the next best thing, the user IP.
+         * Nothing to do, we already reset hwc->interrupts.
          */
-        if ((counter->attr.exclude_kernel || !data.regs) &&
-                        !counter->attr.exclude_user)
-                data.regs = task_pt_regs(current);
+}
 
-        if (data.regs) {
-                if (perf_counter_overflow(counter, 0, &data))
-                        ret = HRTIMER_NORESTART;
-        }
+static void perf_swcounter_add(struct perf_counter *counter, u64 nr,
+                               int nmi, struct perf_sample_data *data)
+{
+        struct hw_perf_counter *hwc = &counter->hw;
 
-        period = max_t(u64, 10000, counter->hw.sample_period);
-        hrtimer_forward_now(hrtimer, ns_to_ktime(period));
+        atomic64_add(nr, &counter->count);
 
-        return ret;
-}
+        if (!hwc->sample_period)
+                return;
 
-static void perf_swcounter_overflow(struct perf_counter *counter,
-                                    int nmi, struct perf_sample_data *data)
-{
-        data->period = counter->hw.last_period;
+        if (!data->regs)
+                return;
 
-        perf_swcounter_update(counter);
-        perf_swcounter_set_period(counter);
-        if (perf_counter_overflow(counter, nmi, data))
-                /* soft-disable the counter */
-                ;
+        if (!atomic64_add_negative(nr, &hwc->period_left))
+                perf_swcounter_overflow(counter, nmi, data);
 }
 
 static int perf_swcounter_is_counting(struct perf_counter *counter)
@@ -3444,15 +3502,6 @@ static int perf_swcounter_match(struct perf_counter *counter,
         return 1;
 }
 
-static void perf_swcounter_add(struct perf_counter *counter, u64 nr,
-                               int nmi, struct perf_sample_data *data)
-{
-        int neg = atomic64_add_negative(nr, &counter->hw.count);
-
-        if (counter->hw.sample_period && !neg && data->regs)
-                perf_swcounter_overflow(counter, nmi, data);
-}
-
 static void perf_swcounter_ctx_event(struct perf_counter_context *ctx,
                                      enum perf_type_id type,
                                      u32 event, u64 nr, int nmi,
@@ -3531,27 +3580,66 @@ void __perf_swcounter_event(u32 event, u64 nr, int nmi,
 
 static void perf_swcounter_read(struct perf_counter *counter)
 {
-        perf_swcounter_update(counter);
 }
 
 static int perf_swcounter_enable(struct perf_counter *counter)
 {
-        perf_swcounter_set_period(counter);
+        struct hw_perf_counter *hwc = &counter->hw;
+
+        if (hwc->sample_period) {
+                hwc->last_period = hwc->sample_period;
+                perf_swcounter_set_period(counter);
+        }
         return 0;
 }
 
 static void perf_swcounter_disable(struct perf_counter *counter)
 {
-        perf_swcounter_update(counter);
 }
 
 static const struct pmu perf_ops_generic = {
         .enable         = perf_swcounter_enable,
         .disable        = perf_swcounter_disable,
         .read           = perf_swcounter_read,
+        .unthrottle     = perf_swcounter_unthrottle,
 };
 
 /*
+ * hrtimer based swcounter callback
+ */
+
+static enum hrtimer_restart perf_swcounter_hrtimer(struct hrtimer *hrtimer)
+{
+        enum hrtimer_restart ret = HRTIMER_RESTART;
+        struct perf_sample_data data;
+        struct perf_counter *counter;
+        u64 period;
+
+        counter = container_of(hrtimer, struct perf_counter, hw.hrtimer);
+        counter->pmu->read(counter);
+
+        data.addr = 0;
+        data.regs = get_irq_regs();
+        /*
+         * In case we exclude kernel IPs or are somehow not in interrupt
+         * context, provide the next best thing, the user IP.
+         */
+        if ((counter->attr.exclude_kernel || !data.regs) &&
+                        !counter->attr.exclude_user)
+                data.regs = task_pt_regs(current);
+
+        if (data.regs) {
+                if (perf_counter_overflow(counter, 0, &data))
+                        ret = HRTIMER_NORESTART;
+        }
+
+        period = max_t(u64, 10000, counter->hw.sample_period);
+        hrtimer_forward_now(hrtimer, ns_to_ktime(period));
+
+        return ret;
+}
+
+/*
  * Software counter: cpu wall time clock
  */
 
@@ -3668,17 +3756,24 @@ static const struct pmu perf_ops_task_clock = {
 };
 
 #ifdef CONFIG_EVENT_PROFILE
-void perf_tpcounter_event(int event_id)
+void perf_tpcounter_event(int event_id, u64 addr, u64 count, void *record,
+                          int entry_size)
 {
+        struct perf_raw_record raw = {
+                .size = entry_size,
+                .data = record,
+        };
+
         struct perf_sample_data data = {
-                .regs = get_irq_regs();
-                .addr = 0,
+                .regs = get_irq_regs(),
+                .addr = addr,
+                .raw = &raw,
         };
 
         if (!data.regs)
                 data.regs = task_pt_regs(current);
 
-        do_perf_swcounter_event(PERF_TYPE_TRACEPOINT, event_id, 1, 1, &data);
+        do_perf_swcounter_event(PERF_TYPE_TRACEPOINT, event_id, count, 1, &data);
 }
 EXPORT_SYMBOL_GPL(perf_tpcounter_event);
 
@@ -3687,16 +3782,20 @@ extern void ftrace_profile_disable(int);
 
 static void tp_perf_counter_destroy(struct perf_counter *counter)
 {
-        ftrace_profile_disable(perf_event_id(&counter->attr));
+        ftrace_profile_disable(counter->attr.config);
 }
 
 static const struct pmu *tp_perf_counter_init(struct perf_counter *counter)
 {
-        int event_id = perf_event_id(&counter->attr);
-        int ret;
+        /*
+         * Raw tracepoint data is a severe data leak, only allow root to
+         * have these.
+         */
+        if ((counter->attr.sample_type & PERF_SAMPLE_RAW) &&
+                        !capable(CAP_SYS_ADMIN))
+                return ERR_PTR(-EPERM);
 
-        ret = ftrace_profile_enable(event_id);
-        if (ret)
+        if (ftrace_profile_enable(counter->attr.config))
                 return NULL;
 
         counter->destroy = tp_perf_counter_destroy;
@@ -3874,6 +3973,8 @@ done:
                         atomic_inc(&nr_mmap_counters);
                 if (counter->attr.comm)
                         atomic_inc(&nr_comm_counters);
+                if (counter->attr.task)
+                        atomic_inc(&nr_task_counters);
         }
 
         return counter;
@@ -4235,8 +4336,10 @@ void perf_counter_exit_task(struct task_struct *child)
         struct perf_counter_context *child_ctx;
         unsigned long flags;
 
-        if (likely(!child->perf_counter_ctxp))
+        if (likely(!child->perf_counter_ctxp)) {
+                perf_counter_task(child, NULL, 0);
                 return;
+        }
 
         local_irq_save(flags);
         /*
@@ -4255,17 +4358,20 @@ void perf_counter_exit_task(struct task_struct *child)
          */
         spin_lock(&child_ctx->lock);
         child->perf_counter_ctxp = NULL;
-        if (child_ctx->parent_ctx) {
-                /*
-                 * This context is a clone; unclone it so it can't get
-                 * swapped to another process while we're removing all
-                 * the counters from it.
-                 */
-                put_ctx(child_ctx->parent_ctx);
-                child_ctx->parent_ctx = NULL;
-        }
-        spin_unlock(&child_ctx->lock);
-        local_irq_restore(flags);
+        /*
+         * If this context is a clone; unclone it so it can't get
+         * swapped to another process while we're removing all
+         * the counters from it.
+         */
+        unclone_ctx(child_ctx);
+        spin_unlock_irqrestore(&child_ctx->lock, flags);
+
+        /*
+         * Report the task dead after unscheduling the counters so that we
+         * won't get any samples after PERF_EVENT_EXIT. We can however still
+         * get a few PERF_EVENT_READ events.
+         */
+        perf_counter_task(child, child_ctx, 0);
 
         /*
          * We can recurse on the same lock type through: