Merge branch 'perf-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip

* 'perf-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (184 commits)
  perf probe: Clean up probe_point_lazy_walker() return value
  tracing: Fix irqoff selftest expanding max buffer
  tracing: Align 4 byte ints together in struct tracer
  tracing: Export trace_set_clr_event()
  tracing: Explain about unstable clock on resume with ring buffer warning
  ftrace/graph: Trace function entry before updating index
  ftrace: Add .ref.text as one of the safe areas to trace
  tracing: Adjust conditional expression latency formatting.
  tracing: Fix event alignment: skb:kfree_skb
  tracing: Fix event alignment: mce:mce_record
  tracing: Fix event alignment: kvm:kvm_hv_hypercall
  tracing: Fix event alignment: module:module_request
  tracing: Fix event alignment: ftrace:context_switch and ftrace:wakeup
  tracing: Remove lock_depth from event entry
  perf header: Stop using 'self'
  perf session: Use evlist/evsel for managing perf.data attributes
  perf top: Don't let events to eat up whole header line
  perf top: Fix events overflow in top command
  ring-buffer: Remove unused #include <linux/trace_irq.h>
  tracing: Add an 'overwrite' trace_option.
  ...

15 files changed, 1904 insertions, 478 deletions

diff --git a/kernel/cgroup.c b/kernel/cgroup.c
index b24d7027b83..95362d15128 100644
--- a/kernel/cgroup.c
+++ b/kernel/cgroup.c
@@ -4230,20 +4230,8 @@ void cgroup_post_fork(struct task_struct *child)
  */
 void cgroup_exit(struct task_struct *tsk, int run_callbacks)
 {
-	int i;
 	struct css_set *cg;
-
-	if (run_callbacks && need_forkexit_callback) {
-		/*
-		 * modular subsystems can't use callbacks, so no need to lock
-		 * the subsys array
-		 */
-		for (i = 0; i < CGROUP_BUILTIN_SUBSYS_COUNT; i++) {
-			struct cgroup_subsys *ss = subsys[i];
-			if (ss->exit)
-				ss->exit(ss, tsk);
-		}
-	}
+	int i;
 
 	/*
 	 * Unlink from the css_set task list if necessary.
@@ -4261,7 +4249,24 @@ void cgroup_exit(struct task_struct *tsk, int run_callbacks)
 	task_lock(tsk);
 	cg = tsk->cgroups;
 	tsk->cgroups = &init_css_set;
+
+	if (run_callbacks && need_forkexit_callback) {
+		/*
+		 * modular subsystems can't use callbacks, so no need to lock
+		 * the subsys array
+		 */
+		for (i = 0; i < CGROUP_BUILTIN_SUBSYS_COUNT; i++) {
+			struct cgroup_subsys *ss = subsys[i];
+			if (ss->exit) {
+				struct cgroup *old_cgrp =
+					rcu_dereference_raw(cg->subsys[i])->cgroup;
+				struct cgroup *cgrp = task_cgroup(tsk, i);
+				ss->exit(ss, cgrp, old_cgrp, tsk);
+			}
+		}
+	}
 	task_unlock(tsk);
+
 	if (cg)
 		put_css_set_taskexit(cg);
 }
@@ -4813,6 +4818,29 @@ css_get_next(struct cgroup_subsys *ss, int id,
 	return ret;
 }
 
+/*
+ * get corresponding css from file open on cgroupfs directory
+ */
+struct cgroup_subsys_state *cgroup_css_from_dir(struct file *f, int id)
+{
+	struct cgroup *cgrp;
+	struct inode *inode;
+	struct cgroup_subsys_state *css;
+
+	inode = f->f_dentry->d_inode;
+	/* check in cgroup filesystem dir */
+	if (inode->i_op != &cgroup_dir_inode_operations)
+		return ERR_PTR(-EBADF);
+
+	if (id < 0 || id >= CGROUP_SUBSYS_COUNT)
+		return ERR_PTR(-EINVAL);
+
+	/* get cgroup */
+	cgrp = __d_cgrp(f->f_dentry);
+	css = cgrp->subsys[id];
+	return css ? css : ERR_PTR(-ENOENT);
+}
+
 #ifdef CONFIG_CGROUP_DEBUG
 static struct cgroup_subsys_state *debug_create(struct cgroup_subsys *ss,
 						   struct cgroup *cont)
diff --git a/kernel/perf_event.c b/kernel/perf_event.c
index 656222fcf76..ed253aa24ba 100644
--- a/kernel/perf_event.c
+++ b/kernel/perf_event.c
@@ -38,13 +38,96 @@
 
 #include <asm/irq_regs.h>
 
+struct remote_function_call {
+	struct task_struct *p;
+	int (*func)(void *info);
+	void *info;
+	int ret;
+};
+
+static void remote_function(void *data)
+{
+	struct remote_function_call *tfc = data;
+	struct task_struct *p = tfc->p;
+
+	if (p) {
+		tfc->ret = -EAGAIN;
+		if (task_cpu(p) != smp_processor_id() || !task_curr(p))
+			return;
+	}
+
+	tfc->ret = tfc->func(tfc->info);
+}
+
+/**
+ * task_function_call - call a function on the cpu on which a task runs
+ * @p:		the task to evaluate
+ * @func:	the function to be called
+ * @info:	the function call argument
+ *
+ * Calls the function @func when the task is currently running. This might
+ * be on the current CPU, which just calls the function directly
+ *
+ * returns: @func return value, or
+ *	    -ESRCH  - when the process isn't running
+ *	    -EAGAIN - when the process moved away
+ */
+static int
+task_function_call(struct task_struct *p, int (*func) (void *info), void *info)
+{
+	struct remote_function_call data = {
+		.p = p,
+		.func = func,
+		.info = info,
+		.ret = -ESRCH, /* No such (running) process */
+	};
+
+	if (task_curr(p))
+		smp_call_function_single(task_cpu(p), remote_function, &data, 1);
+
+	return data.ret;
+}
+
+/**
+ * cpu_function_call - call a function on the cpu
+ * @func:	the function to be called
+ * @info:	the function call argument
+ *
+ * Calls the function @func on the remote cpu.
+ *
+ * returns: @func return value or -ENXIO when the cpu is offline
+ */
+static int cpu_function_call(int cpu, int (*func) (void *info), void *info)
+{
+	struct remote_function_call data = {
+		.p = NULL,
+		.func = func,
+		.info = info,
+		.ret = -ENXIO, /* No such CPU */
+	};
+
+	smp_call_function_single(cpu, remote_function, &data, 1);
+
+	return data.ret;
+}
+
+#define PERF_FLAG_ALL (PERF_FLAG_FD_NO_GROUP |\
+		       PERF_FLAG_FD_OUTPUT  |\
+		       PERF_FLAG_PID_CGROUP)
+
 enum event_type_t {
 	EVENT_FLEXIBLE = 0x1,
 	EVENT_PINNED = 0x2,
 	EVENT_ALL = EVENT_FLEXIBLE | EVENT_PINNED,
 };
 
-atomic_t perf_task_events __read_mostly;
+/*
+ * perf_sched_events : >0 events exist
+ * perf_cgroup_events: >0 per-cpu cgroup events exist on this cpu
+ */
+atomic_t perf_sched_events __read_mostly;
+static DEFINE_PER_CPU(atomic_t, perf_cgroup_events);
+
 static atomic_t nr_mmap_events __read_mostly;
 static atomic_t nr_comm_events __read_mostly;
 static atomic_t nr_task_events __read_mostly;
@@ -67,7 +150,24 @@ 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;
+#define DEFAULT_MAX_SAMPLE_RATE 100000
+int sysctl_perf_event_sample_rate __read_mostly = DEFAULT_MAX_SAMPLE_RATE;
+static int max_samples_per_tick __read_mostly =
+	DIV_ROUND_UP(DEFAULT_MAX_SAMPLE_RATE, HZ);
+
+int perf_proc_update_handler(struct ctl_table *table, int write,
+		void __user *buffer, size_t *lenp,
+		loff_t *ppos)
+{
+	int ret = proc_dointvec(table, write, buffer, lenp, ppos);
+
+	if (ret || !write)
+		return ret;
+
+	max_samples_per_tick = DIV_ROUND_UP(sysctl_perf_event_sample_rate, HZ);
+
+	return 0;
+}
 
 static atomic64_t perf_event_id;
 
@@ -75,7 +175,11 @@ 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);
+			     enum event_type_t event_type,
+			     struct task_struct *task);
+
+static void update_context_time(struct perf_event_context *ctx);
+static u64 perf_event_time(struct perf_event *event);
 
 void __weak perf_event_print_debug(void)	{ }
 
@@ -89,6 +193,360 @@ static inline u64 perf_clock(void)
 	return local_clock();
 }
 
+static inline struct perf_cpu_context *
+__get_cpu_context(struct perf_event_context *ctx)
+{
+	return this_cpu_ptr(ctx->pmu->pmu_cpu_context);
+}
+
+#ifdef CONFIG_CGROUP_PERF
+
+/*
+ * Must ensure cgroup is pinned (css_get) before calling
+ * this function. In other words, we cannot call this function
+ * if there is no cgroup event for the current CPU context.
+ */
+static inline struct perf_cgroup *
+perf_cgroup_from_task(struct task_struct *task)
+{
+	return container_of(task_subsys_state(task, perf_subsys_id),
+			struct perf_cgroup, css);
+}
+
+static inline bool
+perf_cgroup_match(struct perf_event *event)
+{
+	struct perf_event_context *ctx = event->ctx;
+	struct perf_cpu_context *cpuctx = __get_cpu_context(ctx);
+
+	return !event->cgrp || event->cgrp == cpuctx->cgrp;
+}
+
+static inline void perf_get_cgroup(struct perf_event *event)
+{
+	css_get(&event->cgrp->css);
+}
+
+static inline void perf_put_cgroup(struct perf_event *event)
+{
+	css_put(&event->cgrp->css);
+}
+
+static inline void perf_detach_cgroup(struct perf_event *event)
+{
+	perf_put_cgroup(event);
+	event->cgrp = NULL;
+}
+
+static inline int is_cgroup_event(struct perf_event *event)
+{
+	return event->cgrp != NULL;
+}
+
+static inline u64 perf_cgroup_event_time(struct perf_event *event)
+{
+	struct perf_cgroup_info *t;
+
+	t = per_cpu_ptr(event->cgrp->info, event->cpu);
+	return t->time;
+}
+
+static inline void __update_cgrp_time(struct perf_cgroup *cgrp)
+{
+	struct perf_cgroup_info *info;
+	u64 now;
+
+	now = perf_clock();
+
+	info = this_cpu_ptr(cgrp->info);
+
+	info->time += now - info->timestamp;
+	info->timestamp = now;
+}
+
+static inline void update_cgrp_time_from_cpuctx(struct perf_cpu_context *cpuctx)
+{
+	struct perf_cgroup *cgrp_out = cpuctx->cgrp;
+	if (cgrp_out)
+		__update_cgrp_time(cgrp_out);
+}
+
+static inline void update_cgrp_time_from_event(struct perf_event *event)
+{
+	struct perf_cgroup *cgrp;
+
+	/*
+	 * ensure we access cgroup data only when needed and
+	 * when we know the cgroup is pinned (css_get)
+	 */
+	if (!is_cgroup_event(event))
+		return;
+
+	cgrp = perf_cgroup_from_task(current);
+	/*
+	 * Do not update time when cgroup is not active
+	 */
+	if (cgrp == event->cgrp)
+		__update_cgrp_time(event->cgrp);
+}
+
+static inline void
+perf_cgroup_set_timestamp(struct task_struct *task,
+			  struct perf_event_context *ctx)
+{
+	struct perf_cgroup *cgrp;
+	struct perf_cgroup_info *info;
+
+	/*
+	 * ctx->lock held by caller
+	 * ensure we do not access cgroup data
+	 * unless we have the cgroup pinned (css_get)
+	 */
+	if (!task || !ctx->nr_cgroups)
+		return;
+
+	cgrp = perf_cgroup_from_task(task);
+	info = this_cpu_ptr(cgrp->info);
+	info->timestamp = ctx->timestamp;
+}
+
+#define PERF_CGROUP_SWOUT	0x1 /* cgroup switch out every event */
+#define PERF_CGROUP_SWIN	0x2 /* cgroup switch in events based on task */
+
+/*
+ * reschedule events based on the cgroup constraint of task.
+ *
+ * mode SWOUT : schedule out everything
+ * mode SWIN : schedule in based on cgroup for next
+ */
+void perf_cgroup_switch(struct task_struct *task, int mode)
+{
+	struct perf_cpu_context *cpuctx;
+	struct pmu *pmu;
+	unsigned long flags;
+
+	/*
+	 * disable interrupts to avoid geting nr_cgroup
+	 * changes via __perf_event_disable(). Also
+	 * avoids preemption.
+	 */
+	local_irq_save(flags);
+
+	/*
+	 * we reschedule only in the presence of cgroup
+	 * constrained events.
+	 */
+	rcu_read_lock();
+
+	list_for_each_entry_rcu(pmu, &pmus, entry) {
+
+		cpuctx = this_cpu_ptr(pmu->pmu_cpu_context);
+
+		perf_pmu_disable(cpuctx->ctx.pmu);
+
+		/*
+		 * perf_cgroup_events says at least one
+		 * context on this CPU has cgroup events.
+		 *
+		 * ctx->nr_cgroups reports the number of cgroup
+		 * events for a context.
+		 */
+		if (cpuctx->ctx.nr_cgroups > 0) {
+
+			if (mode & PERF_CGROUP_SWOUT) {
+				cpu_ctx_sched_out(cpuctx, EVENT_ALL);
+				/*
+				 * must not be done before ctxswout due
+				 * to event_filter_match() in event_sched_out()
+				 */
+				cpuctx->cgrp = NULL;
+			}
+
+			if (mode & PERF_CGROUP_SWIN) {
+				/* set cgrp before ctxsw in to
+				 * allow event_filter_match() to not
+				 * have to pass task around
+				 */
+				cpuctx->cgrp = perf_cgroup_from_task(task);
+				cpu_ctx_sched_in(cpuctx, EVENT_ALL, task);
+			}
+		}
+
+		perf_pmu_enable(cpuctx->ctx.pmu);
+	}
+
+	rcu_read_unlock();
+
+	local_irq_restore(flags);
+}
+
+static inline void perf_cgroup_sched_out(struct task_struct *task)
+{
+	perf_cgroup_switch(task, PERF_CGROUP_SWOUT);
+}
+
+static inline void perf_cgroup_sched_in(struct task_struct *task)
+{
+	perf_cgroup_switch(task, PERF_CGROUP_SWIN);
+}
+
+static inline int perf_cgroup_connect(int fd, struct perf_event *event,
+				      struct perf_event_attr *attr,
+				      struct perf_event *group_leader)
+{
+	struct perf_cgroup *cgrp;
+	struct cgroup_subsys_state *css;
+	struct file *file;
+	int ret = 0, fput_needed;
+
+	file = fget_light(fd, &fput_needed);
+	if (!file)
+		return -EBADF;
+
+	css = cgroup_css_from_dir(file, perf_subsys_id);
+	if (IS_ERR(css)) {
+		ret = PTR_ERR(css);
+		goto out;
+	}
+
+	cgrp = container_of(css, struct perf_cgroup, css);
+	event->cgrp = cgrp;
+
+	/* must be done before we fput() the file */
+	perf_get_cgroup(event);
+
+	/*
+	 * all events in a group must monitor
+	 * the same cgroup because a task belongs
+	 * to only one perf cgroup at a time
+	 */
+	if (group_leader && group_leader->cgrp != cgrp) {
+		perf_detach_cgroup(event);
+		ret = -EINVAL;
+	}
+out:
+	fput_light(file, fput_needed);
+	return ret;
+}
+
+static inline void
+perf_cgroup_set_shadow_time(struct perf_event *event, u64 now)
+{
+	struct perf_cgroup_info *t;
+	t = per_cpu_ptr(event->cgrp->info, event->cpu);
+	event->shadow_ctx_time = now - t->timestamp;
+}
+
+static inline void
+perf_cgroup_defer_enabled(struct perf_event *event)
+{
+	/*
+	 * when the current task's perf cgroup does not match
+	 * the event's, we need to remember to call the
+	 * perf_mark_enable() function the first time a task with
+	 * a matching perf cgroup is scheduled in.
+	 */
+	if (is_cgroup_event(event) && !perf_cgroup_match(event))
+		event->cgrp_defer_enabled = 1;
+}
+
+static inline void
+perf_cgroup_mark_enabled(struct perf_event *event,
+			 struct perf_event_context *ctx)
+{
+	struct perf_event *sub;
+	u64 tstamp = perf_event_time(event);
+
+	if (!event->cgrp_defer_enabled)
+		return;
+
+	event->cgrp_defer_enabled = 0;
+
+	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;
+			sub->cgrp_defer_enabled = 0;
+		}
+	}
+}
+#else /* !CONFIG_CGROUP_PERF */
+
+static inline bool
+perf_cgroup_match(struct perf_event *event)
+{
+	return true;
+}
+
+static inline void perf_detach_cgroup(struct perf_event *event)
+{}
+
+static inline int is_cgroup_event(struct perf_event *event)
+{
+	return 0;
+}
+
+static inline u64 perf_cgroup_event_cgrp_time(struct perf_event *event)
+{
+	return 0;
+}
+
+static inline void update_cgrp_time_from_event(struct perf_event *event)
+{
+}
+
+static inline void update_cgrp_time_from_cpuctx(struct perf_cpu_context *cpuctx)
+{
+}
+
+static inline void perf_cgroup_sched_out(struct task_struct *task)
+{
+}
+
+static inline void perf_cgroup_sched_in(struct task_struct *task)
+{
+}
+
+static inline int perf_cgroup_connect(pid_t pid, struct perf_event *event,
+				      struct perf_event_attr *attr,
+				      struct perf_event *group_leader)
+{
+	return -EINVAL;
+}
+
+static inline void
+perf_cgroup_set_timestamp(struct task_struct *task,
+			  struct perf_event_context *ctx)
+{
+}
+
+void
+perf_cgroup_switch(struct task_struct *task, struct task_struct *next)
+{
+}
+
+static inline void
+perf_cgroup_set_shadow_time(struct perf_event *event, u64 now)
+{
+}
+
+static inline u64 perf_cgroup_event_time(struct perf_event *event)
+{
+	return 0;
+}
+
+static inline void
+perf_cgroup_defer_enabled(struct perf_event *event)
+{
+}
+
+static inline void
+perf_cgroup_mark_enabled(struct perf_event *event,
+			 struct perf_event_context *ctx)
+{
+}
+#endif
+
 void perf_pmu_disable(struct pmu *pmu)
 {
 	int *count = this_cpu_ptr(pmu->pmu_disable_count);
@@ -254,7 +712,6 @@ static void perf_unpin_context(struct perf_event_context *ctx)
 	raw_spin_lock_irqsave(&ctx->lock, flags);
 	--ctx->pin_count;
 	raw_spin_unlock_irqrestore(&ctx->lock, flags);
-	put_ctx(ctx);
 }
 
 /*
@@ -271,6 +728,10 @@ static void update_context_time(struct perf_event_context *ctx)
 static u64 perf_event_time(struct perf_event *event)
 {
 	struct perf_event_context *ctx = event->ctx;
+
+	if (is_cgroup_event(event))
+		return perf_cgroup_event_time(event);
+
 	return ctx ? ctx->time : 0;
 }
 
@@ -285,9 +746,20 @@ static void update_event_times(struct perf_event *event)
 	if (event->state < PERF_EVENT_STATE_INACTIVE ||
 	    event->group_leader->state < PERF_EVENT_STATE_INACTIVE)
 		return;
-
-	if (ctx->is_active)
+	/*
+	 * in cgroup mode, time_enabled represents
+	 * the time the event was enabled AND active
+	 * tasks were in the monitored cgroup. This is
+	 * independent of the activity of the context as
+	 * there may be a mix of cgroup and non-cgroup events.
+	 *
+	 * That is why we treat cgroup events differently
+	 * here.
+	 */
+	if (is_cgroup_event(event))
 		run_end = perf_event_time(event);
+	else if (ctx->is_active)
+		run_end = ctx->time;
 	else
 		run_end = event->tstamp_stopped;
 
@@ -299,6 +771,7 @@ static void update_event_times(struct perf_event *event)
 		run_end = perf_event_time(event);
 
 	event->total_time_running = run_end - event->tstamp_running;
+
 }
 
 /*
@@ -347,6 +820,9 @@ list_add_event(struct perf_event *event, struct perf_event_context *ctx)
 		list_add_tail(&event->group_entry, list);
 	}
 
+	if (is_cgroup_event(event))
+		ctx->nr_cgroups++;
+
 	list_add_rcu(&event->event_entry, &ctx->event_list);
 	if (!ctx->nr_events)
 		perf_pmu_rotate_start(ctx->pmu);
@@ -473,6 +949,9 @@ list_del_event(struct perf_event *event, struct perf_event_context *ctx)
 
 	event->attach_state &= ~PERF_ATTACH_CONTEXT;
 
+	if (is_cgroup_event(event))
+		ctx->nr_cgroups--;
+
 	ctx->nr_events--;
 	if (event->attr.inherit_stat)
 		ctx->nr_stat--;
@@ -544,7 +1023,8 @@ out:
 static inline int
 event_filter_match(struct perf_event *event)
 {
-	return event->cpu == -1 || event->cpu == smp_processor_id();
+	return (event->cpu == -1 || event->cpu == smp_processor_id())
+	    && perf_cgroup_match(event);
 }
 
 static void
@@ -562,7 +1042,7 @@ event_sched_out(struct perf_event *event,
 	 */
 	if (event->state == PERF_EVENT_STATE_INACTIVE
 	    && !event_filter_match(event)) {
-		delta = ctx->time - event->tstamp_stopped;
+		delta = tstamp - event->tstamp_stopped;
 		event->tstamp_running += delta;
 		event->tstamp_stopped = tstamp;
 	}
@@ -606,47 +1086,30 @@ group_sched_out(struct perf_event *group_event,
 		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)
+static int __perf_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);
+
+	return 0;
 }
 
 
 /*
  * 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.
  *
@@ -657,49 +1120,48 @@ static void __perf_event_remove_from_context(void *info)
  * When called from perf_event_exit_task, it's OK because the
  * context has been detached from its task.
  */
-static void perf_event_remove_from_context(struct perf_event *event)
+static void perf_remove_from_context(struct perf_event *event)
 {
 	struct perf_event_context *ctx = event->ctx;
 	struct task_struct *task = ctx->task;
 
+	lockdep_assert_held(&ctx->mutex);
+
 	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);
+		cpu_function_call(event->cpu, __perf_remove_from_context, event);
 		return;
 	}
 
 retry:
-	task_oncpu_function_call(task, __perf_event_remove_from_context,
-				 event);
+	if (!task_function_call(task, __perf_remove_from_context, event))
+		return;
 
 	raw_spin_lock_irq(&ctx->lock);
 	/*
-	 * If the context is active we need to retry the smp call.
+	 * If we failed to find a running task, but find the context active now
+	 * that we've acquired the ctx->lock, retry.
 	 */
-	if (ctx->nr_active && !list_empty(&event->group_entry)) {
+	if (ctx->is_active) {
 		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.
+	 * Since the task isn't running, its safe to remove the event, us
+	 * holding the ctx->lock ensures the task won't get scheduled in.
 	 */
-	if (!list_empty(&event->group_entry))
-		list_del_event(event, ctx);
+	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)
+static int __perf_event_disable(void *info)
 {
 	struct perf_event *event = info;
 	struct perf_event_context *ctx = event->ctx;
@@ -708,9 +1170,12 @@ static void __perf_event_disable(void *info)
 	/*
 	 * If this is a per-task event, need to check whether this
 	 * event's task is the current task on this cpu.
+	 *
+	 * Can trigger due to concurrent perf_event_context_sched_out()
+	 * flipping contexts around.
 	 */
 	if (ctx->task && cpuctx->task_ctx != ctx)
-		return;
+		return -EINVAL;
 
 	raw_spin_lock(&ctx->lock);
 
@@ -720,6 +1185,7 @@ static void __perf_event_disable(void *info)
 	 */
 	if (event->state >= PERF_EVENT_STATE_INACTIVE) {
 		update_context_time(ctx);
+		update_cgrp_time_from_event(event);
 		update_group_times(event);
 		if (event == event->group_leader)
 			group_sched_out(event, cpuctx, ctx);
@@ -729,6 +1195,8 @@ static void __perf_event_disable(void *info)
 	}
 
 	raw_spin_unlock(&ctx->lock);
+
+	return 0;
 }
 
 /*
@@ -753,13 +1221,13 @@ void perf_event_disable(struct perf_event *event)
 		/*
 		 * Disable the event on the cpu that it's on
 		 */
-		smp_call_function_single(event->cpu, __perf_event_disable,
-					 event, 1);
+		cpu_function_call(event->cpu, __perf_event_disable, event);
 		return;
 	}
 
 retry:
-	task_oncpu_function_call(task, __perf_event_disable, event);
+	if (!task_function_call(task, __perf_event_disable, event))
+		return;
 
 	raw_spin_lock_irq(&ctx->lock);
 	/*
@@ -767,6 +1235,11 @@ retry:
 	 */
 	if (event->state == PERF_EVENT_STATE_ACTIVE) {
 		raw_spin_unlock_irq(&ctx->lock);
+		/*
+		 * Reload the task pointer, it might have been changed by
+		 * a concurrent perf_event_context_sched_out().
+		 */
+		task = ctx->task;
 		goto retry;
 	}
 
@@ -778,10 +1251,44 @@ retry:
 		update_group_times(event);
 		event->state = PERF_EVENT_STATE_OFF;
 	}
-
 	raw_spin_unlock_irq(&ctx->lock);
 }
 
+static void perf_set_shadow_time(struct perf_event *event,
+				 struct perf_event_context *ctx,
+				 u64 tstamp)
+{
+	/*
+	 * use the correct time source for the time snapshot
+	 *
+	 * We could get by without this by leveraging the
+	 * fact that to get to this function, the caller
+	 * has most likely already called update_context_time()
+	 * and update_cgrp_time_xx() and thus both timestamp
+	 * are identical (or very close). Given that tstamp is,
+	 * already adjusted for cgroup, we could say that:
+	 *    tstamp - ctx->timestamp
+	 * is equivalent to
+	 *    tstamp - cgrp->timestamp.
+	 *
+	 * Then, in perf_output_read(), the calculation would
+	 * work with no changes because:
+	 * - event is guaranteed scheduled in
+	 * - no scheduled out in between
+	 * - thus the timestamp would be the same
+	 *
+	 * But this is a bit hairy.
+	 *
+	 * So instead, we have an explicit cgroup call to remain
+	 * within the time time source all along. We believe it
+	 * is cleaner and simpler to understand.
+	 */
+	if (is_cgroup_event(event))
+		perf_cgroup_set_shadow_time(event, tstamp);
+	else
+		event->shadow_ctx_time = tstamp - ctx->timestamp;
+}
+
 #define MAX_INTERRUPTS (~0ULL)
 
 static void perf_log_throttle(struct perf_event *event, int enable);
@@ -822,7 +1329,7 @@ event_sched_in(struct perf_event *event,
 
 	event->tstamp_running += tstamp - event->tstamp_stopped;
 
-	event->shadow_ctx_time = tstamp - ctx->timestamp;
+	perf_set_shadow_time(event, ctx, tstamp);
 
 	if (!is_software_event(event))
 		cpuctx->active_oncpu++;
@@ -943,12 +1450,15 @@ static void add_event_to_ctx(struct perf_event *event,
 	event->tstamp_stopped = tstamp;
 }
 
+static void perf_event_context_sched_in(struct perf_event_context *ctx,
+					struct task_struct *tsk);
+
 /*
  * 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)
+static int  __perf_install_in_context(void *info)
 {
 	struct perf_event *event = info;
 	struct perf_event_context *ctx = event->ctx;
@@ -957,21 +1467,22 @@ static void __perf_install_in_context(void *info)
 	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.
+	 * In case we're installing a new context to an already running task,
+	 * could also happen before perf_event_task_sched_in() on architectures
+	 * which do context switches with IRQs enabled.
 	 */
-	if (ctx->task && cpuctx->task_ctx != ctx) {
-		if (cpuctx->task_ctx || ctx->task != current)
-			return;
-		cpuctx->task_ctx = ctx;
-	}
+	if (ctx->task && !cpuctx->task_ctx)
+		perf_event_context_sched_in(ctx, ctx->task);
 
 	raw_spin_lock(&ctx->lock);
 	ctx->is_active = 1;
 	update_context_time(ctx);
+	/*
+	 * update cgrp time only if current cgrp
+	 * matches event->cgrp. Must be done before
+	 * calling add_event_to_ctx()
+	 */
+	update_cgrp_time_from_event(event);
 
 	add_event_to_ctx(event, ctx);
 
@@ -1012,6 +1523,8 @@ static void __perf_install_in_context(void *info)
 
 unlock:
 	raw_spin_unlock(&ctx->lock);
+
+	return 0;
 }
 
 /*
@@ -1023,8 +1536,6 @@ unlock:
  * 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,
@@ -1033,6 +1544,8 @@ perf_install_in_context(struct perf_event_context *ctx,
 {
 	struct task_struct *task = ctx->task;
 
+	lockdep_assert_held(&ctx->mutex);
+
 	event->ctx = ctx;
 
 	if (!task) {
@@ -1040,31 +1553,29 @@ perf_install_in_context(struct perf_event_context *ctx,
 		 * 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);
+		cpu_function_call(cpu, __perf_install_in_context, event);
 		return;
 	}
 
 retry:
-	task_oncpu_function_call(task, __perf_install_in_context,
-				 event);
+	if (!task_function_call(task, __perf_install_in_context, event))
+		return;
 
 	raw_spin_lock_irq(&ctx->lock);
 	/*
-	 * we need to retry the smp call.
+	 * If we failed to find a running task, but find the context active now
+	 * that we've acquired the ctx->lock, retry.
 	 */
-	if (ctx->is_active && list_empty(&event->group_entry)) {
+	if (ctx->is_active) {
 		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.
+	 * Since the task isn't running, its safe to add the event, us holding
+	 * the ctx->lock ensures the task won't get scheduled in.
 	 */
-	if (list_empty(&event->group_entry))
-		add_event_to_ctx(event, ctx);
+	add_event_to_ctx(event, ctx);
 	raw_spin_unlock_irq(&ctx->lock);
 }
 
@@ -1093,7 +1604,7 @@ static void __perf_event_mark_enabled(struct perf_event *event,
 /*
  * Cross CPU call to enable a performance event
  */
-static void __perf_event_enable(void *info)
+static int __perf_event_enable(void *info)
 {
 	struct perf_event *event = info;
 	struct perf_event_context *ctx = event->ctx;
@@ -1101,26 +1612,27 @@ static void __perf_event_enable(void *info)
 	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;
-	}
+	if (WARN_ON_ONCE(!ctx->is_activ