1 files changed, 398 insertions, 197 deletions

diff --git a/kernel/perf_counter.c b/kernel/perf_counter.c
index 868102172aa4..d7cbc579fc80 100644
--- a/kernel/perf_counter.c
+++ b/kernel/perf_counter.c
@@ -50,7 +50,7 @@ static atomic_t nr_task_counters __read_mostly;
  *  1 - disallow cpu counters to unpriv
  *  2 - disallow kernel profiling to unpriv
  */
-int sysctl_perf_counter_paranoid __read_mostly;
+int sysctl_perf_counter_paranoid __read_mostly = 1;
 
 static inline bool perf_paranoid_cpu(void)
 {
@@ -88,6 +88,7 @@ void __weak hw_perf_disable(void)		{ barrier(); }
 void __weak hw_perf_enable(void)                { barrier(); }
 
 void __weak hw_perf_counter_setup(int cpu)      { barrier(); }
+void __weak hw_perf_counter_setup_online(int cpu)       { barrier(); }
 
 int __weak
 hw_perf_group_sched_in(struct perf_counter *group_leader,
@@ -306,6 +307,10 @@ counter_sched_out(struct perf_counter *counter,
                 return;
 
         counter->state = PERF_COUNTER_STATE_INACTIVE;
+        if (counter->pending_disable) {
+                counter->pending_disable = 0;
+                counter->state = PERF_COUNTER_STATE_OFF;
+        }
         counter->tstamp_stopped = ctx->time;
         counter->pmu->disable(counter);
         counter->oncpu = -1;
@@ -1498,10 +1503,21 @@ static void perf_counter_enable_on_exec(struct task_struct *task)
  */
 static void __perf_counter_read(void *info)
 {
+        struct perf_cpu_context *cpuctx = &__get_cpu_var(perf_cpu_context);
         struct perf_counter *counter = info;
         struct perf_counter_context *ctx = counter->ctx;
         unsigned long flags;
 
+        /*
+         * 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
+         * counter->count would have been updated to a recent sample
+         * when the counter was scheduled out.
+         */
+        if (ctx->task && cpuctx->task_ctx != ctx)
+                return;
+
         local_irq_save(flags);
         if (ctx->is_active)
                 update_context_time(ctx);
@@ -1691,7 +1707,32 @@ static int perf_release(struct inode *inode, struct file *file)
         return 0;
 }
 
-static u64 perf_counter_read_tree(struct perf_counter *counter)
+static int perf_counter_read_size(struct perf_counter *counter)
+{
+        int entry = sizeof(u64); /* value */
+        int size = 0;
+        int nr = 1;
+
+        if (counter->attr.read_format & PERF_FORMAT_TOTAL_TIME_ENABLED)
+                size += sizeof(u64);
+
+        if (counter->attr.read_format & PERF_FORMAT_TOTAL_TIME_RUNNING)
+                size += sizeof(u64);
+
+        if (counter->attr.read_format & PERF_FORMAT_ID)
+                entry += sizeof(u64);
+
+        if (counter->attr.read_format & PERF_FORMAT_GROUP) {
+                nr += counter->group_leader->nr_siblings;
+                size += sizeof(u64);
+        }
+
+        size += entry * nr;
+
+        return size;
+}
+
+static u64 perf_counter_read_value(struct perf_counter *counter)
 {
         struct perf_counter *child;
         u64 total = 0;
@@ -1703,14 +1744,96 @@ static u64 perf_counter_read_tree(struct perf_counter *counter)
         return total;
 }
 
+static int perf_counter_read_entry(struct perf_counter *counter,
+                                   u64 read_format, char __user *buf)
+{
+        int n = 0, count = 0;
+        u64 values[2];
+
+        values[n++] = perf_counter_read_value(counter);
+        if (read_format & PERF_FORMAT_ID)
+                values[n++] = primary_counter_id(counter);
+
+        count = n * sizeof(u64);
+
+        if (copy_to_user(buf, values, count))
+                return -EFAULT;
+
+        return count;
+}
+
+static int perf_counter_read_group(struct perf_counter *counter,
+                                   u64 read_format, char __user *buf)
+{
+        struct perf_counter *leader = counter->group_leader, *sub;
+        int n = 0, size = 0, err = -EFAULT;
+        u64 values[3];
+
+        values[n++] = 1 + leader->nr_siblings;
+        if (read_format & PERF_FORMAT_TOTAL_TIME_ENABLED) {
+                values[n++] = leader->total_time_enabled +
+                        atomic64_read(&leader->child_total_time_enabled);
+        }
+        if (read_format & PERF_FORMAT_TOTAL_TIME_RUNNING) {
+                values[n++] = leader->total_time_running +
+                        atomic64_read(&leader->child_total_time_running);
+        }
+
+        size = n * sizeof(u64);
+
+        if (copy_to_user(buf, values, size))
+                return -EFAULT;
+
+        err = perf_counter_read_entry(leader, read_format, buf + size);
+        if (err < 0)
+                return err;
+
+        size += err;
+
+        list_for_each_entry(sub, &leader->sibling_list, list_entry) {
+                err = perf_counter_read_entry(sub, read_format,
+                                buf + size);
+                if (err < 0)
+                        return err;
+
+                size += err;
+        }
+
+        return size;
+}
+
+static int perf_counter_read_one(struct perf_counter *counter,
+                                 u64 read_format, char __user *buf)
+{
+        u64 values[4];
+        int n = 0;
+
+        values[n++] = perf_counter_read_value(counter);
+        if (read_format & PERF_FORMAT_TOTAL_TIME_ENABLED) {
+                values[n++] = counter->total_time_enabled +
+                        atomic64_read(&counter->child_total_time_enabled);
+        }
+        if (read_format & PERF_FORMAT_TOTAL_TIME_RUNNING) {
+                values[n++] = counter->total_time_running +
+                        atomic64_read(&counter->child_total_time_running);
+        }
+        if (read_format & PERF_FORMAT_ID)
+                values[n++] = primary_counter_id(counter);
+
+        if (copy_to_user(buf, values, n * sizeof(u64)))
+                return -EFAULT;
+
+        return n * sizeof(u64);
+}
+
 /*
  * Read the performance counter - simple non blocking version for now
  */
 static ssize_t
 perf_read_hw(struct perf_counter *counter, char __user *buf, size_t count)
 {
-        u64 values[4];
-        int n;
+        u64 read_format = counter->attr.read_format;
+        int ret;
 
         /*
          * Return end-of-file for a read on a counter that is in
@@ -1720,28 +1843,18 @@ perf_read_hw(struct perf_counter *counter, char __user *buf, size_t count)
         if (counter->state == PERF_COUNTER_STATE_ERROR)
                 return 0;
 
+        if (count < perf_counter_read_size(counter))
+                return -ENOSPC;
+
         WARN_ON_ONCE(counter->ctx->parent_ctx);
         mutex_lock(&counter->child_mutex);
-        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 +
-                        atomic64_read(&counter->child_total_time_enabled);
-        if (counter->attr.read_format & PERF_FORMAT_TOTAL_TIME_RUNNING)
-                values[n++] = counter->total_time_running +
-                        atomic64_read(&counter->child_total_time_running);
-        if (counter->attr.read_format & PERF_FORMAT_ID)
-                values[n++] = primary_counter_id(counter);
+        if (read_format & PERF_FORMAT_GROUP)
+                ret = perf_counter_read_group(counter, read_format, buf);
+        else
+                ret = perf_counter_read_one(counter, read_format, buf);
         mutex_unlock(&counter->child_mutex);
 
-        if (count < n * sizeof(u64))
-                return -EINVAL;
-        count = n * sizeof(u64);
-
-        if (copy_to_user(buf, values, count))
-                return -EFAULT;
-
-        return count;
+        return ret;
 }
 
 static ssize_t
@@ -1906,6 +2019,10 @@ int perf_counter_task_disable(void)
         return 0;
 }
 
+#ifndef PERF_COUNTER_INDEX_OFFSET
+# define PERF_COUNTER_INDEX_OFFSET 0
+#endif
+
 static int perf_counter_index(struct perf_counter *counter)
 {
         if (counter->state != PERF_COUNTER_STATE_ACTIVE)
@@ -2245,7 +2362,7 @@ static void perf_pending_counter(struct perf_pending_entry *entry)
 
         if (counter->pending_disable) {
                 counter->pending_disable = 0;
-                perf_counter_disable(counter);
+                __perf_counter_disable(counter);
         }
 
         if (counter->pending_wakeup) {
@@ -2630,7 +2747,80 @@ static u32 perf_counter_tid(struct perf_counter *counter, struct task_struct *p)
         return task_pid_nr_ns(p, counter->ns);
 }
 
-static void perf_counter_output(struct perf_counter *counter, int nmi,
+static void perf_output_read_one(struct perf_output_handle *handle,
+                                 struct perf_counter *counter)
+{
+        u64 read_format = counter->attr.read_format;
+        u64 values[4];
+        int n = 0;
+
+        values[n++] = atomic64_read(&counter->count);
+        if (read_format & PERF_FORMAT_TOTAL_TIME_ENABLED) {
+                values[n++] = counter->total_time_enabled +
+                        atomic64_read(&counter->child_total_time_enabled);
+        }
+        if (read_format & PERF_FORMAT_TOTAL_TIME_RUNNING) {
+                values[n++] = counter->total_time_running +
+                        atomic64_read(&counter->child_total_time_running);
+        }
+        if (read_format & PERF_FORMAT_ID)
+                values[n++] = primary_counter_id(counter);
+
+        perf_output_copy(handle, values, n * sizeof(u64));
+}
+
+/*
+ * XXX PERF_FORMAT_GROUP vs inherited counters seems difficult.
+ */
+static void perf_output_read_group(struct perf_output_handle *handle,
+                            struct perf_counter *counter)
+{
+        struct perf_counter *leader = counter->group_leader, *sub;
+        u64 read_format = counter->attr.read_format;
+        u64 values[5];
+        int n = 0;
+
+        values[n++] = 1 + leader->nr_siblings;
+
+        if (read_format & PERF_FORMAT_TOTAL_TIME_ENABLED)
+                values[n++] = leader->total_time_enabled;
+
+        if (read_format & PERF_FORMAT_TOTAL_TIME_RUNNING)
+                values[n++] = leader->total_time_running;
+
+        if (leader != counter)
+                leader->pmu->read(leader);
+
+        values[n++] = atomic64_read(&leader->count);
+        if (read_format & PERF_FORMAT_ID)
+                values[n++] = primary_counter_id(leader);
+
+        perf_output_copy(handle, values, n * sizeof(u64));
+
+        list_for_each_entry(sub, &leader->sibling_list, list_entry) {
+                n = 0;
+
+                if (sub != counter)
+                        sub->pmu->read(sub);
+
+                values[n++] = atomic64_read(&sub->count);
+                if (read_format & PERF_FORMAT_ID)
+                        values[n++] = primary_counter_id(sub);
+
+                perf_output_copy(handle, values, n * sizeof(u64));
+        }
+}
+
+static void perf_output_read(struct perf_output_handle *handle,
+                             struct perf_counter *counter)
+{
+        if (counter->attr.read_format & PERF_FORMAT_GROUP)
+                perf_output_read_group(handle, counter);
+        else
+                perf_output_read_one(handle, counter);
+}
+
+void perf_counter_output(struct perf_counter *counter, int nmi,
                                 struct perf_sample_data *data)
 {
         int ret;
@@ -2641,12 +2831,7 @@ static void perf_counter_output(struct perf_counter *counter, int nmi,
         struct {
                 u32 pid, tid;
         } tid_entry;
-        struct {
-                u64 id;
-                u64 counter;
-        } group_entry;
         struct perf_callchain_entry *callchain = NULL;
-        struct perf_tracepoint_record *tp;
         int callchain_size = 0;
         u64 time;
         struct {
@@ -2700,10 +2885,8 @@ static void perf_counter_output(struct perf_counter *counter, int nmi,
         if (sample_type & PERF_SAMPLE_PERIOD)
                 header.size += sizeof(u64);
 
-        if (sample_type & PERF_SAMPLE_GROUP) {
-                header.size += sizeof(u64) +
-                        counter->nr_siblings * sizeof(group_entry);
-        }
+        if (sample_type & PERF_SAMPLE_READ)
+                header.size += perf_counter_read_size(counter);
 
         if (sample_type & PERF_SAMPLE_CALLCHAIN) {
                 callchain = perf_callchain(data->regs);
@@ -2715,9 +2898,16 @@ static void perf_counter_output(struct perf_counter *counter, int nmi,
                         header.size += sizeof(u64);
         }
 
-        if (sample_type & PERF_SAMPLE_TP_RECORD) {
-                tp = data->private;
-                header.size += tp->size;
+        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);
@@ -2753,26 +2943,8 @@ static void perf_counter_output(struct perf_counter *counter, int nmi,
         if (sample_type & PERF_SAMPLE_PERIOD)
                 perf_output_put(&handle, data->period);
 
-        /*
-         * XXX PERF_SAMPLE_GROUP vs inherited counters seems difficult.
-         */
-        if (sample_type & PERF_SAMPLE_GROUP) {
-                struct perf_counter *leader, *sub;
-                u64 nr = counter->nr_siblings;
-
-                perf_output_put(&handle, nr);
-
-                leader = counter->group_leader;
-                list_for_each_entry(sub, &leader->sibling_list, list_entry) {
-                        if (sub != counter)
-                                sub->pmu->read(sub);
-
-                        group_entry.id = primary_counter_id(sub);
-                        group_entry.counter = atomic64_read(&sub->count);
-
-                        perf_output_put(&handle, group_entry);
-                }
-        }
+        if (sample_type & PERF_SAMPLE_READ)
+                perf_output_read(&handle, counter);
 
         if (sample_type & PERF_SAMPLE_CALLCHAIN) {
                 if (callchain)
@@ -2783,8 +2955,21 @@ static void perf_counter_output(struct perf_counter *counter, int nmi,
                 }
         }
 
-        if (sample_type & PERF_SAMPLE_TP_RECORD)
-                perf_output_copy(&handle, tp->record, tp->size);
+        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);
 }
@@ -2798,8 +2983,6 @@ struct perf_read_event {
 
         u32                             pid;
         u32                             tid;
-        u64                             value;
-        u64                             format[3];
 };
 
 static void
@@ -2811,34 +2994,20 @@ perf_counter_read_event(struct perf_counter *counter,
                 .header = {
                         .type = PERF_EVENT_READ,
                         .misc = 0,
-                        .size = sizeof(event) - sizeof(event.format),
+                        .size = sizeof(event) + perf_counter_read_size(counter),
                 },
                 .pid = perf_counter_pid(counter, task),
                 .tid = perf_counter_tid(counter, task),
-                .value = atomic64_read(&counter->count),
         };
-        int ret, i = 0;
-
-        if (counter->attr.read_format & PERF_FORMAT_TOTAL_TIME_ENABLED) {
-                event.header.size += sizeof(u64);
-                event.format[i++] = counter->total_time_enabled;
-        }
-
-        if (counter->attr.read_format & PERF_FORMAT_TOTAL_TIME_RUNNING) {
-                event.header.size += sizeof(u64);
-                event.format[i++] = counter->total_time_running;
-        }
-
-        if (counter->attr.read_format & PERF_FORMAT_ID) {
-                event.header.size += sizeof(u64);
-                event.format[i++] = primary_counter_id(counter);
-        }
+        int ret;
 
         ret = perf_output_begin(&handle, counter, event.header.size, 0, 0);
         if (ret)
                 return;
 
-        perf_output_copy(&handle, &event, event.header.size);
+        perf_output_put(&handle, event);
+        perf_output_read(&handle, counter);
+
         perf_output_end(&handle);
 }
 
@@ -2849,7 +3018,8 @@ perf_counter_read_event(struct perf_counter *counter,
  */
 
 struct perf_task_event {
-        struct task_struct      *task;
+        struct task_struct              *task;
+        struct perf_counter_context     *task_ctx;
 
         struct {
                 struct perf_event_header        header;
@@ -2873,10 +3043,10 @@ static void perf_counter_task_output(struct perf_counter *counter,
                 return;
 
         task_event->event.pid = perf_counter_pid(counter, task);
-        task_event->event.ppid = perf_counter_pid(counter, task->real_parent);
+        task_event->event.ppid = perf_counter_pid(counter, current);
 
         task_event->event.tid = perf_counter_tid(counter, task);
-        task_event->event.ptid = perf_counter_tid(counter, task->real_parent);
+        task_event->event.ptid = perf_counter_tid(counter, current);
 
         perf_output_put(&handle, task_event->event);
         perf_output_end(&handle);
@@ -2909,24 +3079,23 @@ static void perf_counter_task_ctx(struct perf_counter_context *ctx,
 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_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_task_ctx(ctx, task_event);
         rcu_read_unlock();
 }
 
-static void perf_counter_task(struct task_struct *task, int new)
+static void perf_counter_task(struct task_struct *task,
+                              struct perf_counter_context *task_ctx,
+                              int new)
 {
         struct perf_task_event task_event;
 
@@ -2936,8 +3105,9 @@ static void perf_counter_task(struct task_struct *task, int new)
                 return;
 
         task_event = (struct perf_task_event){
-                .task   = task,
-                .event  = {
+                .task     = task,
+                .task_ctx = task_ctx,
+                .event    = {
                         .header = {
                                 .type = new ? PERF_EVENT_FORK : PERF_EVENT_EXIT,
                                 .misc = 0,
@@ -2955,7 +3125,7 @@ static void perf_counter_task(struct task_struct *task, int new)
 
 void perf_counter_fork(struct task_struct *task)
 {
-        perf_counter_task(task, 1);
+        perf_counter_task(task, NULL, 1);
 }
 
 /*
@@ -3344,125 +3514,111 @@ 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)
 {
-        struct perf_counter_context *ctx;
-        unsigned long flags;
-        int count;
-
+        /*
+         * The counter is active, we're good!
+         */
         if (counter->state == PERF_COUNTER_STATE_ACTIVE)
                 return 1;
 
+        /*
+         * The counter is off/error, not counting.
+         */
         if (counter->state != PERF_COUNTER_STATE_INACTIVE)
                 return 0;
 
         /*
-         * If the counter is inactive, it could be just because
-         * its task is scheduled out, or because it's in a group
-         * which could not go on the PMU.  We want to count in
-         * the first case but not the second.  If the context is
-         * currently active then an inactive software counter must
-         * be the second case.  If it's not currently active then
-         * we need to know whether the counter was active when the
-         * context was last active, which we can determine by
-         * comparing counter->tstamp_stopped with ctx->time.
-         *
-         * We are within an RCU read-side critical section,
-         * which protects the existence of *ctx.
+         * The counter is inactive, if the context is active
+         * we're part of a group that didn't make it on the 'pmu',
+         * not counting.
          */
-        ctx = counter->ctx;
-        spin_lock_irqsave(&ctx->lock, flags);
-        count = 1;
-        /* Re-check state now we have the lock */
-        if (counter->state < PERF_COUNTER_STATE_INACTIVE ||
-            counter->ctx->is_active ||
-            counter->tstamp_stopped < ctx->time)
-                count = 0;
-        spin_unlock_irqrestore(&ctx->lock, flags);
-        return count;
+        if (counter->ctx->is_active)
+                return 0;
+
+        /*
+         * We're inactive and the context is too, this means the
+         * task is scheduled out, we're counting events that happen
+         * to us, like migration events.
+         */
+        return 1;
 }
 
 static int perf_swcounter_match(struct perf_counter *counter,
@@ -3488,15 +3644,6 @@ 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,
@@ -3575,27 +3722,66 @@ 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
  */
 
@@ -3715,15 +3901,15 @@ static const struct pmu perf_ops_task_clock = {
 void perf_tpcounter_event(int event_id, u64 addr, u64 count, void *record,
                           int entry_size)
 {
-        struct perf_tracepoint_record tp = {
+        struct perf_raw_record raw = {
                 .size = entry_size,
-                .record = record,
+                .data = record,
         };
 
         struct perf_sample_data data = {
                 .regs = get_irq_regs(),
                 .addr = addr,
-                .private = &tp,
+                .raw = &raw,
         };
 
         if (!data.regs)
@@ -3743,6 +3929,14 @@ static void tp_perf_counter_destroy(struct perf_counter *counter)
 
 static const struct pmu *tp_perf_counter_init(struct perf_counter *counter)
 {
+        /*
+         * 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);
+
         if (ftrace_profile_enable(counter->attr.config))
                 return NULL;
 
@@ -3872,13 +4066,14 @@ perf_counter_alloc(struct perf_counter_attr *attr,
         hwc->sample_period = attr->sample_period;
         if (attr->freq && attr->sample_freq)
                 hwc->sample_period = 1;
+        hwc->last_period = hwc->sample_period;
 
         atomic64_set(&hwc->period_left, hwc->sample_period);
 
         /*
-         * we currently do not support PERF_SAMPLE_GROUP on inherited counters
+         * we currently do not support PERF_FORMAT_GROUP on inherited counters
          */
-        if (attr->inherit && (attr->sample_type & PERF_SAMPLE_GROUP))
+        if (attr->inherit && (attr->read_format & PERF_FORMAT_GROUP))
                 goto done;
 
         switch (attr->type) {
@@ -4285,7 +4480,7 @@ void perf_counter_exit_task(struct task_struct *child)
         unsigned long flags;
 
         if (likely(!child->perf_counter_ctxp)) {
-                perf_counter_task(child, 0);
+                perf_counter_task(child, NULL, 0);
                 return;
         }
 
@@ -4305,6 +4500,7 @@ void perf_counter_exit_task(struct task_struct *child)
          * incremented the context's refcount before we do put_ctx below.
          */
         spin_lock(&child_ctx->lock);
+        child->perf_counter_ctxp = NULL;
         /*
          * If this context is a clone; unclone it so it can't get
          * swapped to another process while we're removing all
@@ -4318,9 +4514,7 @@ void perf_counter_exit_task(struct task_struct *child)
          * won't get any samples after PERF_EVENT_EXIT. We can however still
          * get a few PERF_EVENT_READ events.
          */
-        perf_counter_task(child, 0);
-
-        child->perf_counter_ctxp = NULL;
+        perf_counter_task(child, child_ctx, 0);
 
         /*
          * We can recurse on the same lock type through:
@@ -4541,6 +4735,11 @@ perf_cpu_notify(struct notifier_block *self, unsigned long action, void *hcpu)
                 perf_counter_init_cpu(cpu);
                 break;
 
+        case CPU_ONLINE:
+        case CPU_ONLINE_FROZEN:
+                hw_perf_counter_setup_online(cpu);
+                break;
+
         case CPU_DOWN_PREPARE:
         case CPU_DOWN_PREPARE_FROZEN:
                 perf_counter_exit_cpu(cpu);
@@ -4565,6 +4764,8 @@ void __init perf_counter_init(void)
 {
         perf_cpu_notify(&perf_cpu_nb, (unsigned long)CPU_UP_PREPARE,
                         (void *)(long)smp_processor_id());
+        perf_cpu_notify(&perf_cpu_nb, (unsigned long)CPU_ONLINE,
+                        (void *)(long)smp_processor_id());
         register_cpu_notifier(&perf_cpu_nb);
 }