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: (123 commits)
  perf: Remove the nmi parameter from the oprofile_perf backend
  x86, perf: Make copy_from_user_nmi() a library function
  perf: Remove perf_event_attr::type check
  x86, perf: P4 PMU - Fix typos in comments and style cleanup
  perf tools: Make test use the preset debugfs path
  perf tools: Add automated tests for events parsing
  perf tools: De-opt the parse_events function
  perf script: Fix display of IP address for non-callchain path
  perf tools: Fix endian conversion reading event attr from file header
  perf tools: Add missing 'node' alias to the hw_cache[] array
  perf probe: Support adding probes on offline kernel modules
  perf probe: Add probed module in front of function
  perf probe: Introduce debuginfo to encapsulate dwarf information
  perf-probe: Move dwarf library routines to dwarf-aux.{c, h}
  perf probe: Remove redundant dwarf functions
  perf probe: Move strtailcmp to string.c
  perf probe: Rename DIE_FIND_CB_FOUND to DIE_FIND_CB_END
  tracing/kprobe: Update symbol reference when loading module
  tracing/kprobes: Support module init function probing
  kprobes: Return -ENOENT if probe point doesn't exist
  ...

25 files changed, 1655 insertions, 1182 deletions

diff --git a/kernel/async.c b/kernel/async.c
index cd9dbb913c77..d5fe7af0de2e 100644
--- a/kernel/async.c
+++ b/kernel/async.c
@@ -49,12 +49,13 @@ asynchronous and synchronous parts of the kernel.
 */
 
 #include <linux/async.h>
+#include <linux/atomic.h>
+#include <linux/ktime.h>
 #include <linux/module.h>
 #include <linux/wait.h>
 #include <linux/sched.h>
 #include <linux/slab.h>
 #include <linux/workqueue.h>
-#include <asm/atomic.h>
 
 static async_cookie_t next_cookie = 1;
 
@@ -128,7 +129,8 @@ static void async_run_entry_fn(struct work_struct *work)
 
 	/* 2) run (and print duration) */
 	if (initcall_debug && system_state == SYSTEM_BOOTING) {
-		printk("calling  %lli_%pF @ %i\n", (long long)entry->cookie,
+		printk(KERN_DEBUG "calling  %lli_%pF @ %i\n",
+			(long long)entry->cookie,
 			entry->func, task_pid_nr(current));
 		calltime = ktime_get();
 	}
@@ -136,7 +138,7 @@ static void async_run_entry_fn(struct work_struct *work)
 	if (initcall_debug && system_state == SYSTEM_BOOTING) {
 		rettime = ktime_get();
 		delta = ktime_sub(rettime, calltime);
-		printk("initcall %lli_%pF returned 0 after %lld usecs\n",
+		printk(KERN_DEBUG "initcall %lli_%pF returned 0 after %lld usecs\n",
 			(long long)entry->cookie,
 			entry->func,
 			(long long)ktime_to_ns(delta) >> 10);
@@ -270,7 +272,7 @@ void async_synchronize_cookie_domain(async_cookie_t cookie,
 	ktime_t starttime, delta, endtime;
 
 	if (initcall_debug && system_state == SYSTEM_BOOTING) {
-		printk("async_waiting @ %i\n", task_pid_nr(current));
+		printk(KERN_DEBUG "async_waiting @ %i\n", task_pid_nr(current));
 		starttime = ktime_get();
 	}
 
@@ -280,7 +282,7 @@ void async_synchronize_cookie_domain(async_cookie_t cookie,
 		endtime = ktime_get();
 		delta = ktime_sub(endtime, starttime);
 
-		printk("async_continuing @ %i after %lli usec\n",
+		printk(KERN_DEBUG "async_continuing @ %i after %lli usec\n",
 			task_pid_nr(current),
 			(long long)ktime_to_ns(delta) >> 10);
 	}
diff --git a/kernel/events/Makefile b/kernel/events/Makefile
index 1ce23d3d8394..89e5e8aa4c36 100644
--- a/kernel/events/Makefile
+++ b/kernel/events/Makefile
@@ -2,5 +2,5 @@ ifdef CONFIG_FUNCTION_TRACER
 CFLAGS_REMOVE_core.o = -pg
 endif
 
-obj-y := core.o
+obj-y := core.o ring_buffer.o
 obj-$(CONFIG_HAVE_HW_BREAKPOINT) += hw_breakpoint.o
diff --git a/kernel/events/core.c b/kernel/events/core.c
index 9efe7108ccaf..b8785e26ee1c 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -36,6 +36,8 @@
 #include <linux/ftrace_event.h>
 #include <linux/hw_breakpoint.h>
 
+#include "internal.h"
+
 #include <asm/irq_regs.h>
 
 struct remote_function_call {
@@ -200,6 +202,22 @@ __get_cpu_context(struct perf_event_context *ctx)
 	return this_cpu_ptr(ctx->pmu->pmu_cpu_context);
 }
 
+static void perf_ctx_lock(struct perf_cpu_context *cpuctx,
+			  struct perf_event_context *ctx)
+{
+	raw_spin_lock(&cpuctx->ctx.lock);
+	if (ctx)
+		raw_spin_lock(&ctx->lock);
+}
+
+static void perf_ctx_unlock(struct perf_cpu_context *cpuctx,
+			    struct perf_event_context *ctx)
+{
+	if (ctx)
+		raw_spin_unlock(&ctx->lock);
+	raw_spin_unlock(&cpuctx->ctx.lock);
+}
+
 #ifdef CONFIG_CGROUP_PERF
 
 /*
@@ -340,11 +358,8 @@ void perf_cgroup_switch(struct task_struct *task, int mode)
 	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.
@@ -353,6 +368,8 @@ void perf_cgroup_switch(struct task_struct *task, int mode)
 		 * events for a context.
 		 */
 		if (cpuctx->ctx.nr_cgroups > 0) {
+			perf_ctx_lock(cpuctx, cpuctx->task_ctx);
+			perf_pmu_disable(cpuctx->ctx.pmu);
 
 			if (mode & PERF_CGROUP_SWOUT) {
 				cpu_ctx_sched_out(cpuctx, EVENT_ALL);
@@ -372,9 +389,9 @@ void perf_cgroup_switch(struct task_struct *task, int mode)
 				cpuctx->cgrp = perf_cgroup_from_task(task);
 				cpu_ctx_sched_in(cpuctx, EVENT_ALL, task);
 			}
+			perf_pmu_enable(cpuctx->ctx.pmu);
+			perf_ctx_unlock(cpuctx, cpuctx->task_ctx);
 		}
-
-		perf_pmu_enable(cpuctx->ctx.pmu);
 	}
 
 	rcu_read_unlock();
@@ -731,6 +748,7 @@ static u64 perf_event_time(struct perf_event *event)
 
 /*
  * Update the total_time_enabled and total_time_running fields for a event.
+ * The caller of this function needs to hold the ctx->lock.
  */
 static void update_event_times(struct perf_event *event)
 {
@@ -1105,6 +1123,10 @@ static int __perf_remove_from_context(void *info)
 	raw_spin_lock(&ctx->lock);
 	event_sched_out(event, cpuctx, ctx);
 	list_del_event(event, ctx);
+	if (!ctx->nr_events && cpuctx->task_ctx == ctx) {
+		ctx->is_active = 0;
+		cpuctx->task_ctx = NULL;
+	}
 	raw_spin_unlock(&ctx->lock);
 
 	return 0;
@@ -1454,8 +1476,24 @@ 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);
+static void task_ctx_sched_out(struct perf_event_context *ctx);
+static void
+ctx_sched_in(struct perf_event_context *ctx,
+	     struct perf_cpu_context *cpuctx,
+	     enum event_type_t event_type,
+	     struct task_struct *task);
+
+static void perf_event_sched_in(struct perf_cpu_context *cpuctx,
+				struct perf_event_context *ctx,
+				struct task_struct *task)
+{
+	cpu_ctx_sched_in(cpuctx, EVENT_PINNED, task);
+	if (ctx)
+		ctx_sched_in(ctx, cpuctx, EVENT_PINNED, task);
+	cpu_ctx_sched_in(cpuctx, EVENT_FLEXIBLE, task);
+	if (ctx)
+		ctx_sched_in(ctx, cpuctx, EVENT_FLEXIBLE, task);
+}
 
 /*
  * Cross CPU call to install and enable a performance event
@@ -1466,20 +1504,37 @@ static int  __perf_install_in_context(void *info)
 {
 	struct perf_event *event = info;
 	struct perf_event_context *ctx = event->ctx;
-	struct perf_event *leader = event->group_leader;
 	struct perf_cpu_context *cpuctx = __get_cpu_context(ctx);
-	int err;
+	struct perf_event_context *task_ctx = cpuctx->task_ctx;
+	struct task_struct *task = current;
+
+	perf_ctx_lock(cpuctx, task_ctx);
+	perf_pmu_disable(cpuctx->ctx.pmu);
 
 	/*
-	 * 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 there was an active task_ctx schedule it out.
 	 */
-	if (ctx->task && !cpuctx->task_ctx)
-		perf_event_context_sched_in(ctx, ctx->task);
+	if (task_ctx)
+		task_ctx_sched_out(task_ctx);
+
+	/*
+	 * If the context we're installing events in is not the
+	 * active task_ctx, flip them.
+	 */
+	if (ctx->task && task_ctx != ctx) {
+		if (task_ctx)
+			raw_spin_unlock(&task_ctx->lock);
+		raw_spin_lock(&ctx->lock);
+		task_ctx = ctx;
+	}
+
+	if (task_ctx) {
+		cpuctx->task_ctx = task_ctx;
+		task = task_ctx->task;
+	}
+
+	cpu_ctx_sched_out(cpuctx, EVENT_ALL);
 
-	raw_spin_lock(&ctx->lock);
-	ctx->is_active = 1;
 	update_context_time(ctx);
 	/*
 	 * update cgrp time only if current cgrp
@@ -1490,43 +1545,13 @@ static int  __perf_install_in_context(void *info)
 
 	add_event_to_ctx(event, ctx);
 
-	if (!event_filter_match(event))
-		goto unlock;
-
-	/*
-	 * Don't put the event on if it is disabled or if
-	 * it is in a group and the group isn't on.
-	 */
-	if (event->state != PERF_EVENT_STATE_INACTIVE ||
-	    (leader != event && leader->state != PERF_EVENT_STATE_ACTIVE))
-		goto unlock;
-
 	/*
-	 * An exclusive event can't go on if there are already active
-	 * hardware events, and no hardware event can go on if there
-	 * is already an exclusive event on.
+	 * Schedule everything back in
 	 */
-	if (!group_can_go_on(event, cpuctx, 1))
-		err = -EEXIST;
-	else
-		err = event_sched_in(event, cpuctx, ctx);
-
-	if (err) {
-		/*
-		 * This event couldn't go on.  If it is in a group
-		 * then we have to pull the whole group off.
-		 * If the event group is pinned then put it in error state.
-		 */
-		if (leader != event)
-			group_sched_out(leader, cpuctx, ctx);
-		if (leader->attr.pinned) {
-			update_group_times(leader);
-			leader->state = PERF_EVENT_STATE_ERROR;
-		}
-	}
+	perf_event_sched_in(cpuctx, task_ctx, task);
 
-unlock:
-	raw_spin_unlock(&ctx->lock);
+	perf_pmu_enable(cpuctx->ctx.pmu);
+	perf_ctx_unlock(cpuctx, task_ctx);
 
 	return 0;
 }
@@ -1739,7 +1764,7 @@ out:
 	raw_spin_unlock_irq(&ctx->lock);
 }
 
-static int perf_event_refresh(struct perf_event *event, int refresh)
+int perf_event_refresh(struct perf_event *event, int refresh)
 {
 	/*
 	 * not supported on inherited events
@@ -1752,36 +1777,35 @@ static int perf_event_refresh(struct perf_event *event, int refresh)
 
 	return 0;
 }
+EXPORT_SYMBOL_GPL(perf_event_refresh);
 
 static void ctx_sched_out(struct perf_event_context *ctx,
 			  struct perf_cpu_context *cpuctx,
 			  enum event_type_t event_type)
 {
 	struct perf_event *event;
+	int is_active = ctx->is_active;
 
-	raw_spin_lock(&ctx->lock);
-	perf_pmu_disable(ctx->pmu);
-	ctx->is_active = 0;
+	ctx->is_active &= ~event_type;
 	if (likely(!ctx->nr_events))
-		goto out;
+		return;
+
 	update_context_time(ctx);
 	update_cgrp_time_from_cpuctx(cpuctx);
-
 	if (!ctx->nr_active)
-		goto out;
+		return;
 
-	if (event_type & EVENT_PINNED) {
+	perf_pmu_disable(ctx->pmu);
+	if ((is_active & EVENT_PINNED) && (event_type & EVENT_PINNED)) {
 		list_for_each_entry(event, &ctx->pinned_groups, group_entry)
 			group_sched_out(event, cpuctx, ctx);
 	}
 
-	if (event_type & EVENT_FLEXIBLE) {
+	if ((is_active & EVENT_FLEXIBLE) && (event_type & EVENT_FLEXIBLE)) {
 		list_for_each_entry(event, &ctx->flexible_groups, group_entry)
 			group_sched_out(event, cpuctx, ctx);
 	}
-out:
 	perf_pmu_enable(ctx->pmu);
-	raw_spin_unlock(&ctx->lock);
 }
 
 /*
@@ -1929,8 +1953,10 @@ static void perf_event_context_sched_out(struct task_struct *task, int ctxn,
 	rcu_read_unlock();
 
 	if (do_switch) {
+		raw_spin_lock(&ctx->lock);
 		ctx_sched_out(ctx, cpuctx, EVENT_ALL);
 		cpuctx->task_ctx = NULL;
+		raw_spin_unlock(&ctx->lock);
 	}
 }
 
@@ -1965,8 +1991,7 @@ void __perf_event_task_sched_out(struct task_struct *task,
 		perf_cgroup_sched_out(task);
 }
 
-static void task_ctx_sched_out(struct perf_event_context *ctx,
-			       enum event_type_t event_type)
+static void task_ctx_sched_out(struct perf_event_context *ctx)
 {
 	struct perf_cpu_context *cpuctx = __get_cpu_context(ctx);
 
@@ -1976,7 +2001,7 @@ static void task_ctx_sched_out(struct perf_event_context *ctx,
 	if (WARN_ON_ONCE(ctx != cpuctx->task_ctx))
 		return;
 
-	ctx_sched_out(ctx, cpuctx, event_type);
+	ctx_sched_out(ctx, cpuctx, EVENT_ALL);
 	cpuctx->task_ctx = NULL;
 }
 
@@ -2055,11 +2080,11 @@ ctx_sched_in(struct perf_event_context *ctx,
 	     struct task_struct *task)
 {
 	u64 now;
+	int is_active = ctx->is_active;
 
-	raw_spin_lock(&ctx->lock);
-	ctx->is_active = 1;
+	ctx->is_active |= event_type;
 	if (likely(!ctx->nr_events))
-		goto out;
+		return;
 
 	now = perf_clock();
 	ctx->timestamp = now;
@@ -2068,15 +2093,12 @@ ctx_sched_in(struct perf_event_context *ctx,
 	 * First go through the list and put on any pinned groups
 	 * in order to give them the best chance of going on.
 	 */
-	if (event_type & EVENT_PINNED)
+	if (!(is_active & EVENT_PINNED) && (event_type & EVENT_PINNED))
 		ctx_pinned_sched_in(ctx, cpuctx);
 
 	/* Then walk through the lower prio flexible groups */
-	if (event_type & EVENT_FLEXIBLE)
+	if (!(is_active & EVENT_FLEXIBLE) && (event_type & EVENT_FLEXIBLE))
 		ctx_flexible_sched_in(ctx, cpuctx);
-
-out:
-	raw_spin_unlock(&ctx->lock);
 }
 
 static void cpu_ctx_sched_in(struct perf_cpu_context *cpuctx,
@@ -2088,19 +2110,6 @@ static void cpu_ctx_sched_in(struct perf_cpu_context *cpuctx,
 	ctx_sched_in(ctx, cpuctx, event_type, task);
 }
 
-static void task_ctx_sched_in(struct perf_event_context *ctx,
-			      enum event_type_t event_type)
-{
-	struct perf_cpu_context *cpuctx;
-
-	cpuctx = __get_cpu_context(ctx);
-	if (cpuctx->task_ctx == ctx)
-		return;
-
-	ctx_sched_in(ctx, cpuctx, event_type, NULL);
-	cpuctx->task_ctx = ctx;
-}
-
 static void perf_event_context_sched_in(struct perf_event_context *ctx,
 					struct task_struct *task)
 {
@@ -2110,6 +2119,7 @@ static void perf_event_context_sched_in(struct perf_event_context *ctx,
 	if (cpuctx->task_ctx == ctx)
 		return;
 
+	perf_ctx_lock(cpuctx, ctx);
 	perf_pmu_disable(ctx->pmu);
 	/*
 	 * We want to keep the following priority order:
@@ -2118,18 +2128,18 @@ static void perf_event_context_sched_in(struct perf_event_context *ctx,
 	 */
 	cpu_ctx_sched_out(cpuctx, EVENT_FLEXIBLE);
 
-	ctx_sched_in(ctx, cpuctx, EVENT_PINNED, task);
-	cpu_ctx_sched_in(cpuctx, EVENT_FLEXIBLE, task);
-	ctx_sched_in(ctx, cpuctx, EVENT_FLEXIBLE, task);
+	perf_event_sched_in(cpuctx, ctx, task);
 
 	cpuctx->task_ctx = ctx;
 
+	perf_pmu_enable(ctx->pmu);
+	perf_ctx_unlock(cpuctx, ctx);
+
 	/*
 	 * Since these rotations are per-cpu, we need to ensure the
 	 * cpu-context we got scheduled on is actually rotating.
 	 */
 	perf_pmu_rotate_start(ctx->pmu);
-	perf_pmu_enable(ctx->pmu);
 }
 
 /*
@@ -2269,7 +2279,6 @@ static void perf_ctx_adjust_freq(struct perf_event_context *ctx, u64 period)
 	u64 interrupts, now;
 	s64 delta;
 
-	raw_spin_lock(&ctx->lock);
 	list_for_each_entry_rcu(event, &ctx->event_list, event_entry) {
 		if (event->state != PERF_EVENT_STATE_ACTIVE)
 			continue;
@@ -2301,7 +2310,6 @@ static void perf_ctx_adjust_freq(struct perf_event_context *ctx, u64 period)
 		if (delta > 0)
 			perf_adjust_period(event, period, delta);
 	}
-	raw_spin_unlock(&ctx->lock);
 }
 
 /*
@@ -2309,16 +2317,12 @@ static void perf_ctx_adjust_freq(struct perf_event_context *ctx, u64 period)
  */
 static void rotate_ctx(struct perf_event_context *ctx)
 {
-	raw_spin_lock(&ctx->lock);
-
 	/*
 	 * Rotate the first entry last of non-pinned groups. Rotation might be
 	 * disabled by the inheritance code.
 	 */
 	if (!ctx->rotate_disable)
 		list_rotate_left(&ctx->flexible_groups);
-
-	raw_spin_unlock(&ctx->lock);
 }
 
 /*
@@ -2345,6 +2349,7 @@ static void perf_rotate_context(struct perf_cpu_context *cpuctx)
 			rotate = 1;
 	}
 
+	perf_ctx_lock(cpuctx, cpuctx->task_ctx);
 	perf_pmu_disable(cpuctx->ctx.pmu);
 	perf_ctx_adjust_freq(&cpuctx->ctx, interval);
 	if (ctx)
@@ -2355,21 +2360,20 @@ static void perf_rotate_context(struct perf_cpu_context *cpuctx)
 
 	cpu_ctx_sched_out(cpuctx, EVENT_FLEXIBLE);
 	if (ctx)
-		task_ctx_sched_out(ctx, EVENT_FLEXIBLE);
+		ctx_sched_out(ctx, cpuctx, EVENT_FLEXIBLE);
 
 	rotate_ctx(&cpuctx->ctx);
 	if (ctx)
 		rotate_ctx(ctx);
 
-	cpu_ctx_sched_in(cpuctx, EVENT_FLEXIBLE, current);
-	if (ctx)
-		task_ctx_sched_in(ctx, EVENT_FLEXIBLE);
+	perf_event_sched_in(cpuctx, ctx, current);
 
 done:
 	if (remove)
 		list_del_init(&cpuctx->rotation_list);
 
 	perf_pmu_enable(cpuctx->ctx.pmu);
+	perf_ctx_unlock(cpuctx, cpuctx->task_ctx);
 }
 
 void perf_event_task_tick(void)
@@ -2424,9 +2428,9 @@ static void perf_event_enable_on_exec(struct perf_event_context *ctx)
 	 * in.
 	 */
 	perf_cgroup_sched_out(current);
-	task_ctx_sched_out(ctx, EVENT_ALL);
 
 	raw_spin_lock(&ctx->lock);
+	task_ctx_sched_out(ctx);
 
 	list_for_each_entry(event, &ctx->pinned_groups, group_entry) {
 		ret = event_enable_on_exec(event, ctx);
@@ -2835,16 +2839,12 @@ retry:
 		unclone_ctx(ctx);
 		++ctx->pin_count;
 		raw_spin_unlock_irqrestore(&ctx->lock, flags);
-	}
-
-	if (!ctx) {
+	} else {
 		ctx = alloc_perf_context(pmu, task);
 		err = -ENOMEM;
 		if (!ctx)
 			goto errout;
 
-		get_ctx(ctx);
-
 		err = 0;
 		mutex_lock(&task->perf_event_mutex);
 		/*
@@ -2856,14 +2856,14 @@ retry:
 		else if (task->perf_event_ctxp[ctxn])
 			err = -EAGAIN;
 		else {
+			get_ctx(ctx);
 			++ctx->pin_count;
 			rcu_assign_pointer(task->perf_event_ctxp[ctxn], ctx);
 		}
 		mutex_unlock(&task->perf_event_mutex);
 
 		if (unlikely(err)) {
-			put_task_struct(task);
-			kfree(ctx);
+			put_ctx(ctx);
 
 			if (err == -EAGAIN)
 				goto retry;
@@ -2890,7 +2890,7 @@ static void free_event_rcu(struct rcu_head *head)
 	kfree(event);
 }
 
-static void perf_buffer_put(struct perf_buffer *buffer);
+static void ring_buffer_put(struct ring_buffer *rb);
 
 static void free_event(struct perf_event *event)
 {
@@ -2913,9 +2913,9 @@ static void free_event(struct perf_event *event)
 		}
 	}
 
-	if (event->buffer) {
-		perf_buffer_put(event->buffer);
-		event->buffer = NULL;
+	if (event->rb) {
+		ring_buffer_put(event->rb);
+		event->rb = NULL;
 	}
 
 	if (is_cgroup_event(event))
@@ -2934,12 +2934,6 @@ int perf_event_release_kernel(struct perf_event *event)
 {
 	struct perf_event_context *ctx = event->ctx;
 
-	/*
-	 * Remove from the PMU, can't get re-enabled since we got
-	 * here because the last ref went.
-	 */
-	perf_event_disable(event);
-
 	WARN_ON_ONCE(ctx->parent_ctx);
 	/*
 	 * There are two ways this annotation is useful:
@@ -2956,8 +2950,8 @@ int perf_event_release_kernel(struct perf_event *event)
 	mutex_lock_nested(&ctx->mutex, SINGLE_DEPTH_NESTING);
 	raw_spin_lock_irq(&ctx->lock);
 	perf_group_detach(event);
-	list_del_event(event, ctx);
 	raw_spin_unlock_irq(&ctx->lock);
+	perf_remove_from_context(event);
 	mutex_unlock(&ctx->mutex);
 
 	free_event(event);
@@ -3149,13 +3143,13 @@ perf_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
 static unsigned int perf_poll(struct file *file, poll_table *wait)
 {
 	struct perf_event *event = file->private_data;
-	struct perf_buffer *buffer;
+	struct ring_buffer *rb;
 	unsigned int events = POLL_HUP;
 
 	rcu_read_lock();
-	buffer = rcu_dereference(event->buffer);
-	if (buffer)
-		events = atomic_xchg(&buffer->poll, 0);
+	rb = rcu_dereference(event->rb);
+	if (rb)
+		events = atomic_xchg(&rb->poll, 0);
 	rcu_read_unlock();
 
 	poll_wait(file, &event->waitq, wait);
@@ -3358,6 +3352,18 @@ static int perf_event_index(struct perf_event *event)
 	return event->hw.idx + 1 - PERF_EVENT_INDEX_OFFSET;
 }
 
+static void calc_timer_values(struct perf_event *event,
+				u64 *running,
+				u64 *enabled)
+{
+	u64 now, ctx_time;
+
+	now = perf_clock();
+	ctx_time = event->shadow_ctx_time + now;
+	*enabled = ctx_time - event->tstamp_enabled;
+	*running = ctx_time - event->tstamp_running;
+}
+
 /*
  * Callers need to ensure there can be no nesting of this function, otherwise
  * the seqlock logic goes bad. We can not serialize this because the arch
@@ -3366,14 +3372,25 @@ static int perf_event_index(struct perf_event *event)
 void perf_event_update_userpage(struct perf_event *event)
 {
 	struct perf_event_mmap_page *userpg;
-	struct perf_buffer *buffer;
+	struct ring_buffer *rb;
+	u64 enabled, running;
 
 	rcu_read_lock();
-	buffer = rcu_dereference(event->buffer);
-	if (!buffer)
+	/*
+	 * compute total_time_enabled, total_time_running
+	 * based on snapshot values taken when the event
+	 * was last scheduled in.
+	 *
+	 * we cannot simply called update_context_time()
+	 * because of locking issue as we can be called in
+	 * NMI context
+	 */
+	calc_timer_values(event, &enabled, &running);
+	rb = rcu_dereference(event->rb);
+	if (!rb)
 		goto unlock;
 
-	userpg = buffer->user_page;
+	userpg = rb->user_page;
 
 	/*
 	 * Disable preemption so as to not let the corresponding user-space
@@ -3387,10 +3404,10 @@ void perf_event_update_userpage(struct perf_event *event)
 	if (event->state == PERF_EVENT_STATE_ACTIVE)
 		userpg->offset -= local64_read(&event->hw.prev_count);
 
-	userpg->time_enabled = event->total_time_enabled +
+	userpg->time_enabled = enabled +
 			atomic64_read(&event->child_total_time_enabled);
 
-	userpg->time_running = event->total_time_running +
+	userpg->time_running = running +
 			atomic64_read(&event->child_total_time_running);
 
 	barrier();
@@ -3400,220 +3417,10 @@ unlock:
 	rcu_read_unlock();
 }
 
-static unsigned long perf_data_size(struct perf_buffer *buffer);
-
-static void
-perf_buffer_init(struct perf_buffer *buffer, long watermark, int flags)
-{
-	long max_size = perf_data_size(buffer);
-
-	if (watermark)
-		buffer->watermark = min(max_size, watermark);
-
-	if (!buffer->watermark)
-		buffer->watermark = max_size / 2;
-
-	if (flags & PERF_BUFFER_WRITABLE)
-		buffer->writable = 1;
-
-	atomic_set(&buffer->refcount, 1);
-}
-
-#ifndef CONFIG_PERF_USE_VMALLOC
-
-/*
- * Back perf_mmap() with regular GFP_KERNEL-0 pages.
- */
-
-static struct page *
-perf_mmap_to_page(struct perf_buffer *buffer, unsigned long pgoff)
-{
-	if (pgoff > buffer->nr_pages)
-		return NULL;
-
-	if (pgoff == 0)
-		return virt_to_page(buffer->user_page);
-
-	return virt_to_page(buffer->data_pages[pgoff - 1]);
-}
-
-static void *perf_mmap_alloc_page(int cpu)
-{
-	struct page *page;
-	int node;
-
-	node = (cpu == -1) ? cpu : cpu_to_node(cpu);
-	page = alloc_pages_node(node, GFP_KERNEL | __GFP_ZERO, 0);
-	if (!page)
-		return NULL;
-
-	return page_address(page);
-}
-
-static struct perf_buffer *
-perf_buffer_alloc(int nr_pages, long watermark, int cpu, int flags)
-{
-	struct perf_buffer *buffer;
-	unsigned long size;
-	int i;
-
-	size = sizeof(struct perf_buffer);
-	size += nr_pages * sizeof(void *);
-
-	buffer = kzalloc(size, GFP_KERNEL);
-	if (!buffer)
-		goto fail;
-
-	buffer->user_page = perf_mmap_alloc_page(cpu);
-	if (!buffer->user_page)
-		goto fail_user_page;
-
-	for (i = 0; i < nr_pages; i++) {
-		buffer->data_pages[i] = perf_mmap_alloc_page(cpu);
-		if (!buffer->data_pages[i])
-			goto fail_data_pages;
-	}
-
-	buffer->nr_pages = nr_pages;
-
-	perf_buffer_init(buffer, watermark, flags);
-
-	return buffer;
-
-fail_data_pages:
-	for (i--; i >= 0; i--)
-		free_page((unsigned long)buffer->data_pages[i]);
-
-	free_page((unsigned long)buffer->user_page);
-
-fail_user_page:
-	kfree(buffer);
-
-fail:
-	return NULL;
-}
-
-static void perf_mmap_free_page(unsigned long addr)
-{
-	struct page *page = virt_to_page((void *)addr);
-
-	page->mapping = NULL;
-	__free_page(page);
-}
-
-static void perf_buffer_free(struct perf_buffer *buffer)
-{
-	int i;
-
-	perf_mmap_free_page((unsigned long)buffer->user_page);
-	for (i = 0; i < buffer->nr_pages; i++)
-		perf_mmap_free_page((unsigned long)buffer->data_pages[i]);
-	kfree(buffer);
-}
-
-static inline int page_order(struct perf_buffer *buffer)
-{
-	return 0;
-}
-
-#else
-
-/*
- * Back perf_mmap() with vmalloc memory.
- *
- * Required for architectures that have d-cache aliasing issues.
- */
-
-static inline int page_order(struct perf_buffer *buffer)
-{
-	return buffer->page_order;
-}
-
-static struct page *
-perf_mmap_to_page(struct perf_buffer *buffer, unsigned long pgoff)
-{
-	if (pgoff > (1UL << page_order(buffer)))
-		return NULL;
-
-	return vmalloc_to_page((void *)buffer->user_page + pgoff * PAGE_SIZE);
-}
-
-static void perf_mmap_unmark_page(void *addr)
-{
-	struct page *page = vmalloc_to_page(addr);
-
-	page->mapping = NULL;
-}
-
-static void perf_buffer_free_work(struct work_struct *work)
-{
-	struct perf_buffer *buffer;
-	void *base;
-	int i, nr;
-
-	buffer = container_of(work, struct perf_buffer, work);
-	nr = 1 << page_order(buffer);
-
-	base = buffer->user_page;
-	for (i = 0; i < nr + 1; i++)
-		perf_mmap_unmark_page(base + (i * PAGE_SIZE));
-
-	vfree(base);
-	kfree(buffer);
-}
-
-static void perf_buffer_free(struct perf_buffer *buffer)
-{
-	schedule_work(&buffer->work);
-}
-
-static struct perf_buffer *
-perf_buffer_alloc(int nr_pages, long watermark, int cpu, int flags)
-{
-	struct perf_buffer *buffer;
-	unsigned long size;
-	void *all_buf;
-
-	size = sizeof(struct perf_buffer);
-	size += sizeof(void *);
-
-	buffer = kzalloc(size, GFP_KERNEL);
-	if (!buffer)
-		goto fail;
-
-	INIT_WORK(&buffer->work, perf_buffer_free_work);
-
-	all_buf = vmalloc_user((nr_pages + 1) * PAGE_SIZE);
-	if (!all_buf)
-		goto fail_all_buf;
-
-	buffer->user_page = all_buf;
-	buffer->data_pages[0] = all_buf + PAGE_SIZE;
-	buffer->page_order = ilog2(nr_pages);
-	buffer->nr_pages = 1;
-
-	perf_buffer_init(buffer, watermark, flags);
-
-	return buffer;
-
-fail_all_buf:
-	kfree(buffer);
-
-fail:
-	return NULL;
-}
-
-#endif
-
-static unsigned long perf_data_size(struct perf_buffer *buffer)
-{
-	return buffer->nr_pages << (PAGE_SHIFT + page_order(buffer));
-}
-
 static int perf_mmap_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
 {
 	struct perf_event *event = vma->vm_file->private_data;
-	struct perf_buffer *buffer;
+	struct ring_buffer *rb;
 	int ret = VM_FAULT_SIGBUS;
 
 	if (vmf->flags & FAULT_FLAG_MKWRITE) {
@@ -3623,14 +3430,14 @@ static int perf_mmap_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
 	}
 
 	rcu_read_lock();
-	buffer = rcu_dereference(event->buffer);
-	if (!buffer)
+	rb = rcu_dereference(event->rb);
+	if (!rb)
 		goto unlock;
 
 	if (vmf->pgoff && (vmf->flags & FAULT_FLAG_WRITE))
 		goto unlock;
 
-	vmf->page = perf_mmap_to_page(buffer, vmf->pgoff);
+	vmf->page = perf_mmap_to_page(rb, vmf->pgoff);
 	if (!vmf->page)
 		goto unlock;
 
@@ -3645,35 +3452,35 @@ unlock:
 	return ret;
 }
 
-static void perf_buffer_free_rcu(struct rcu_head *rcu_head)
+static void rb_free_rcu(struct rcu_head *rcu_head)
 {
-	struct perf_buffer *buffer;
+	struct ring_buffer *rb;
 
-	buffer = container_of(rcu_head, struct perf_buffer, rcu_head);
-	perf_buffer_free(buffer);
+	rb = container_of(rcu_head, struct ring_buffer, rcu_head);
+	rb_free(rb);
 }
 
-static struct perf_buffer *perf_buffer_get(struct perf_event *event)
+static struct ring_buffer *ring_buffer_get(struct perf_event *event)
 {
-	struct perf_buffer *buffer;
+	struct ring_buffer *rb;
 
 	rcu_read_lock();
-	buffer = rcu_dereference(event->buffer);
-	if (buffer) {
-		if (!atomic_inc_not_zero(&buffer->refcount))
-			buffer = NULL;
+	rb = rcu_dereference(event->rb);
+	if (rb) {