diff --git a/kernel/perf_counter.c b/kernel/perf_counter.c
index 534e20d14d63..36f65e2b8b57 100644
--- a/kernel/perf_counter.c
+++ b/kernel/perf_counter.c
@@ -1503,10 +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);
@@ -2008,6 +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)
diff --git a/tools/perf/Documentation/Makefile b/tools/perf/Documentation/Makefile
index 5457192e1b41..bdd3b7ecad0a 100644
--- a/tools/perf/Documentation/Makefile
+++ b/tools/perf/Documentation/Makefile
@@ -35,7 +35,7 @@ man7dir=$(mandir)/man7
 # DESTDIR=
 
 ASCIIDOC=asciidoc
-ASCIIDOC_EXTRA =
+ASCIIDOC_EXTRA = --unsafe
 MANPAGE_XSL = manpage-normal.xsl
 XMLTO_EXTRA =
 INSTALL?=install
diff --git a/tools/perf/Documentation/perf-examples.txt b/tools/perf/Documentation/examples.txt
similarity index 100%
rename from tools/perf/Documentation/perf-examples.txt
rename to tools/perf/Documentation/examples.txt
diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c
index 1dba568e1941..343e7b14bf01 100644
--- a/tools/perf/builtin-annotate.c
+++ b/tools/perf/builtin-annotate.c
@@ -980,6 +980,13 @@ process_fork_event(event_t *event, unsigned long offset, unsigned long head)
 		(void *)(long)(event->header.size),
 		event->fork.pid, event->fork.ppid);
 
+	/*
+	 * A thread clone will have the same PID for both
+	 * parent and child.
+	 */
+	if (thread == parent)
+		return 0;
+
 	if (!thread || !parent || thread__fork(thread, parent)) {
 		dprintf("problem processing PERF_EVENT_FORK, skipping event.\n");
 		return -1;
diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
index 3d051b9cf25f..89a5ddcd1ded 100644
--- a/tools/perf/builtin-record.c
+++ b/tools/perf/builtin-record.c
@@ -219,7 +219,7 @@ static pid_t pid_synthesize_comm_event(pid_t pid, int full)
 	snprintf(filename, sizeof(filename), "/proc/%d/status", pid);
 
 	fp = fopen(filename, "r");
-	if (fd == NULL) {
+	if (fp == NULL) {
 		/*
 		 * We raced with a task exiting - just return:
 		 */