[PATCH 0/4] perf_counter bits

[PATCH 0/4] perf_counter bits

[Peter Zijlstra]

 - fixes a race in the output code
 - x86: fixes a hang in nmi_watchdog=2 vs perf_counters
 - teaches perf-report to handle 0-length files
 - updates the documentation
-- 

[PATCH 1/4] perf_counter: fix race in perf_output_*

[Peter Zijlstra]

[-- Attachment #1: perf_counter-output-race.patch --]
[-- Type: text/plain, Size: 6696 bytes --]

When two (or more) contexts output to the same buffer, it is possible
to observe half written output.

Suppose we have CPU0 doing perf_counter_mmap(), CPU1 doing
perf_counter_overflow(). If CPU1 does a wakeup and exposes head to
user-space, then CPU2 can observe the data CPU0 is still writing.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
---
 include/linux/perf_counter.h |    5 +
 kernel/perf_counter.c        |  130 +++++++++++++++++++++++++++++++++----------
 2 files changed, 105 insertions(+), 30 deletions(-)

Index: linux-2.6/include/linux/perf_counter.h
===================================================================
--- linux-2.6.orig/include/linux/perf_counter.h
+++ linux-2.6/include/linux/perf_counter.h
@@ -358,10 +358,13 @@ struct perf_mmap_data {
 	struct rcu_head			rcu_head;
 	int				nr_pages;	/* nr of data pages  */
 
-	atomic_t			wakeup;		/* POLL_ for wakeups */
+	atomic_t			poll;		/* POLL_ for wakeups */
 	atomic_t			head;		/* write position    */
 	atomic_t			events;		/* event limit       */
 
+	atomic_t			wakeup_head;	/* completed head    */
+	atomic_t			lock;		/* concurrent writes */
+
 	struct perf_counter_mmap_page   *user_page;
 	void 				*data_pages[0];
 };
Index: linux-2.6/kernel/perf_counter.c
===================================================================
--- linux-2.6.orig/kernel/perf_counter.c
+++ linux-2.6/kernel/perf_counter.c
@@ -1279,14 +1279,12 @@ static unsigned int perf_poll(struct fil
 {
 	struct perf_counter *counter = file->private_data;
 	struct perf_mmap_data *data;
-	unsigned int events;
+	unsigned int events = POLL_HUP;
 
 	rcu_read_lock();
 	data = rcu_dereference(counter->data);
 	if (data)
-		events = atomic_xchg(&data->wakeup, 0);
-	else
-		events = POLL_HUP;
+		events = atomic_xchg(&data->poll, 0);
 	rcu_read_unlock();
 
 	poll_wait(file, &counter->waitq, wait);
@@ -1568,22 +1566,6 @@ static const struct file_operations perf
 
 void perf_counter_wakeup(struct perf_counter *counter)
 {
-	struct perf_mmap_data *data;
-
-	rcu_read_lock();
-	data = rcu_dereference(counter->data);
-	if (data) {
-		atomic_set(&data->wakeup, POLL_IN);
-		/*
-		 * Ensure all data writes are issued before updating the
-		 * user-space data head information. The matching rmb()
-		 * will be in userspace after reading this value.
-		 */
-		smp_wmb();
-		data->user_page->data_head = atomic_read(&data->head);
-	}
-	rcu_read_unlock();
-
 	wake_up_all(&counter->waitq);
 
 	if (counter->pending_kill) {
@@ -1721,10 +1703,14 @@ struct perf_output_handle {
 	int			wakeup;
 	int			nmi;
 	int			overflow;
+	int			locked;
+	unsigned long		flags;
 };
 
-static inline void __perf_output_wakeup(struct perf_output_handle *handle)
+static void perf_output_wakeup(struct perf_output_handle *handle)
 {
+	atomic_set(&handle->data->poll, POLL_IN);
+
 	if (handle->nmi) {
 		handle->counter->pending_wakeup = 1;
 		perf_pending_queue(&handle->counter->pending,
@@ -1733,6 +1719,86 @@ static inline void __perf_output_wakeup(
 		perf_counter_wakeup(handle->counter);
 }
 
+/*
+ * Curious locking construct.
+ *
+ * We need to ensure a later event doesn't publish a head when a former
+ * event isn't done writing. However since we need to deal with NMIs we
+ * cannot fully serialize things.
+ *
+ * What we do is serialize between CPUs so we only have to deal with NMI
+ * nesting on a single CPU.
+ *
+ * We only publish the head (and generate a wakeup) when the outer-most
+ * event completes.
+ */
+static void perf_output_lock(struct perf_output_handle *handle)
+{
+	struct perf_mmap_data *data = handle->data;
+	int cpu;
+
+	handle->locked = 0;
+
+	local_irq_save(handle->flags);
+	cpu = smp_processor_id();
+
+	if (in_nmi() && atomic_read(&data->lock) == cpu)
+		return;
+
+	while (atomic_cmpxchg(&data->lock, 0, cpu) != 0)
+		cpu_relax();
+
+	handle->locked = 1;
+}
+
+static void perf_output_unlock(struct perf_output_handle *handle)
+{
+	struct perf_mmap_data *data = handle->data;
+	int head, cpu;
+
+	if (handle->wakeup)
+		data->wakeup_head = data->head;
+
+	if (!handle->locked)
+		goto out;
+
+again:
+	/*
+	 * The xchg implies a full barrier that ensures all writes are done
+	 * before we publish the new head, matched by a rmb() in userspace when
+	 * reading this position.
+	 */
+	while ((head = atomic_xchg(&data->wakeup_head, 0))) {
+		data->user_page->data_head = head;
+		handle->wakeup = 1;
+	}
+
+	/*
+	 * NMI can happen here, which means we can miss a wakeup_head update.
+	 */
+
+	cpu = atomic_xchg(&data->lock, 0);
+	WARN_ON_ONCE(cpu != smp_processor_id());
+
+	/*
+	 * Therefore we have to validate we did not indeed do so.
+	 */
+	if (unlikely(atomic_read(&data->wakeup_head))) {
+		/*
+		 * Since we had it locked, we can lock it again.
+		 */
+		while (atomic_cmpxchg(&data->lock, 0, cpu) != 0)
+			cpu_relax();
+
+		goto again;
+	}
+
+	if (handle->wakeup)
+		perf_output_wakeup(handle);
+out:
+	local_irq_restore(handle->flags);
+}
+
 static int perf_output_begin(struct perf_output_handle *handle,
 			     struct perf_counter *counter, unsigned int size,
 			     int nmi, int overflow)
@@ -1745,6 +1811,7 @@ static int perf_output_begin(struct perf
 	if (!data)
 		goto out;
 
+	handle->data	 = data;
 	handle->counter	 = counter;
 	handle->nmi	 = nmi;
 	handle->overflow = overflow;
@@ -1752,12 +1819,13 @@ static int perf_output_begin(struct perf
 	if (!data->nr_pages)
 		goto fail;
 
+	perf_output_lock(handle);
+
 	do {
 		offset = head = atomic_read(&data->head);
 		head += size;
 	} while (atomic_cmpxchg(&data->head, offset, head) != offset);
 
-	handle->data	= data;
 	handle->offset	= offset;
 	handle->head	= head;
 	handle->wakeup	= (offset >> PAGE_SHIFT) != (head >> PAGE_SHIFT);
@@ -1765,7 +1833,7 @@ static int perf_output_begin(struct perf
 	return 0;
 
 fail:
-	__perf_output_wakeup(handle);
+	perf_output_wakeup(handle);
 out:
 	rcu_read_unlock();
 
@@ -1809,16 +1877,20 @@ static void perf_output_copy(struct perf
 
 static void perf_output_end(struct perf_output_handle *handle)
 {
-	int wakeup_events = handle->counter->hw_event.wakeup_events;
+	struct perf_counter *counter = handle->counter;
+	struct perf_mmap_data *data = handle->data;
+
+	int wakeup_events = counter->hw_event.wakeup_events;
 
 	if (handle->overflow && wakeup_events) {
-		int events = atomic_inc_return(&handle->data->events);
+		int events = atomic_inc_return(&data->events);
 		if (events >= wakeup_events) {
-			atomic_sub(wakeup_events, &handle->data->events);
-			__perf_output_wakeup(handle);
+			atomic_sub(wakeup_events, &data->events);
+			handle->wakeup = 1;
 		}
-	} else if (handle->wakeup)
-		__perf_output_wakeup(handle);
+	}
+
+	perf_output_unlock(handle);
 	rcu_read_unlock();
 }
 

-- 

[tip:perfcounters/core] perf_counter: fix race in perf_output_*

[tip-bot for Peter Zijlstra]

Commit-ID:  c33a0bc4e41ef169d6e807d8abb9502544b518e5
Gitweb:     http://git.kernel.org/tip/c33a0bc4e41ef169d6e807d8abb9502544b518e5
Author:     Peter Zijlstra <a.p.zijlstra@chello.nl>
AuthorDate: Fri, 1 May 2009 12:23:16 +0200
Committer:  Ingo Molnar <mingo@elte.hu>
CommitDate: Fri, 1 May 2009 13:23:43 +0200

perf_counter: fix race in perf_output_*

When two (or more) contexts output to the same buffer, it is possible
to observe half written output.

Suppose we have CPU0 doing perf_counter_mmap(), CPU1 doing
perf_counter_overflow(). If CPU1 does a wakeup and exposes head to
user-space, then CPU2 can observe the data CPU0 is still writing.

[ Impact: fix occasionally corrupted profiling records ]

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
LKML-Reference: <20090501102533.007821627@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>


---
 include/linux/perf_counter.h |    5 +-
 kernel/perf_counter.c        |  130 ++++++++++++++++++++++++++++++++---------
 2 files changed, 105 insertions(+), 30 deletions(-)

diff --git a/include/linux/perf_counter.h b/include/linux/perf_counter.h
index 41aed42..f776851 100644
--- a/include/linux/perf_counter.h
+++ b/include/linux/perf_counter.h
@@ -358,10 +358,13 @@ struct perf_mmap_data {
 	struct rcu_head			rcu_head;
 	int				nr_pages;	/* nr of data pages  */
 
-	atomic_t			wakeup;		/* POLL_ for wakeups */
+	atomic_t			poll;		/* POLL_ for wakeups */
 	atomic_t			head;		/* write position    */
 	atomic_t			events;		/* event limit       */
 
+	atomic_t			wakeup_head;	/* completed head    */
+	atomic_t			lock;		/* concurrent writes */
+
 	struct perf_counter_mmap_page   *user_page;
 	void 				*data_pages[0];
 };
diff --git a/kernel/perf_counter.c b/kernel/perf_counter.c
index 75f2b6c..8660ae5 100644
--- a/kernel/perf_counter.c
+++ b/kernel/perf_counter.c
@@ -1279,14 +1279,12 @@ static unsigned int perf_poll(struct file *file, poll_table *wait)
 {
 	struct perf_counter *counter = file->private_data;
 	struct perf_mmap_data *data;
-	unsigned int events;
+	unsigned int events = POLL_HUP;
 
 	rcu_read_lock();
 	data = rcu_dereference(counter->data);
 	if (data)
-		events = atomic_xchg(&data->wakeup, 0);
-	else
-		events = POLL_HUP;
+		events = atomic_xchg(&data->poll, 0);
 	rcu_read_unlock();
 
 	poll_wait(file, &counter->waitq, wait);
@@ -1568,22 +1566,6 @@ static const struct file_operations perf_fops = {
 
 void perf_counter_wakeup(struct perf_counter *counter)
 {
-	struct perf_mmap_data *data;
-
-	rcu_read_lock();
-	data = rcu_dereference(counter->data);
-	if (data) {
-		atomic_set(&data->wakeup, POLL_IN);
-		/*
-		 * Ensure all data writes are issued before updating the
-		 * user-space data head information. The matching rmb()
-		 * will be in userspace after reading this value.
-		 */
-		smp_wmb();
-		data->user_page->data_head = atomic_read(&data->head);
-	}
-	rcu_read_unlock();
-
 	wake_up_all(&counter->waitq);
 
 	if (counter->pending_kill) {
@@ -1721,10 +1703,14 @@ struct perf_output_handle {
 	int			wakeup;
 	int			nmi;
 	int			overflow;
+	int			locked;
+	unsigned long		flags;
 };
 
-static inline void __perf_output_wakeup(struct perf_output_handle *handle)
+static void perf_output_wakeup(struct perf_output_handle *handle)
 {
+	atomic_set(&handle->data->poll, POLL_IN);
+
 	if (handle->nmi) {
 		handle->counter->pending_wakeup = 1;
 		perf_pending_queue(&handle->counter->pending,
@@ -1733,6 +1719,86 @@ static inline void __perf_output_wakeup(struct perf_output_handle *handle)
 		perf_counter_wakeup(handle->counter);
 }
 
+/*
+ * Curious locking construct.
+ *
+ * We need to ensure a later event doesn't publish a head when a former
+ * event isn't done writing. However since we need to deal with NMIs we
+ * cannot fully serialize things.
+ *
+ * What we do is serialize between CPUs so we only have to deal with NMI
+ * nesting on a single CPU.
+ *
+ * We only publish the head (and generate a wakeup) when the outer-most
+ * event completes.
+ */
+static void perf_output_lock(struct perf_output_handle *handle)
+{
+	struct perf_mmap_data *data = handle->data;
+	int cpu;
+
+	handle->locked = 0;
+
+	local_irq_save(handle->flags);
+	cpu = smp_processor_id();
+
+	if (in_nmi() && atomic_read(&data->lock) == cpu)
+		return;
+
+	while (atomic_cmpxchg(&data->lock, 0, cpu) != 0)
+		cpu_relax();
+
+	handle->locked = 1;
+}
+
+static void perf_output_unlock(struct perf_output_handle *handle)
+{
+	struct perf_mmap_data *data = handle->data;
+	int head, cpu;
+
+	if (handle->wakeup)
+		data->wakeup_head = data->head;
+
+	if (!handle->locked)
+		goto out;
+
+again:
+	/*
+	 * The xchg implies a full barrier that ensures all writes are done
+	 * before we publish the new head, matched by a rmb() in userspace when
+	 * reading this position.
+	 */
+	while ((head = atomic_xchg(&data->wakeup_head, 0))) {
+		data->user_page->data_head = head;
+		handle->wakeup = 1;
+	}
+
+	/*
+	 * NMI can happen here, which means we can miss a wakeup_head update.
+	 */
+
+	cpu = atomic_xchg(&data->lock, 0);
+	WARN_ON_ONCE(cpu != smp_processor_id());
+
+	/*
+	 * Therefore we have to validate we did not indeed do so.
+	 */
+	if (unlikely(atomic_read(&data->wakeup_head))) {
+		/*
+		 * Since we had it locked, we can lock it again.
+		 */
+		while (atomic_cmpxchg(&data->lock, 0, cpu) != 0)
+			cpu_relax();
+
+		goto again;
+	}
+
+	if (handle->wakeup)
+		perf_output_wakeup(handle);
+out:
+	local_irq_restore(handle->flags);
+}
+
 static int perf_output_begin(struct perf_output_handle *handle,
 			     struct perf_counter *counter, unsigned int size,
 			     int nmi, int overflow)
@@ -1745,6 +1811,7 @@ static int perf_output_begin(struct perf_output_handle *handle,
 	if (!data)
 		goto out;
 
+	handle->data	 = data;
 	handle->counter	 = counter;
 	handle->nmi	 = nmi;
 	handle->overflow = overflow;
@@ -1752,12 +1819,13 @@ static int perf_output_begin(struct perf_output_handle *handle,
 	if (!data->nr_pages)
 		goto fail;
 
+	perf_output_lock(handle);
+
 	do {
 		offset = head = atomic_read(&data->head);
 		head += size;
 	} while (atomic_cmpxchg(&data->head, offset, head) != offset);
 
-	handle->data	= data;
 	handle->offset	= offset;
 	handle->head	= head;
 	handle->wakeup	= (offset >> PAGE_SHIFT) != (head >> PAGE_SHIFT);
@@ -1765,7 +1833,7 @@ static int perf_output_begin(struct perf_output_handle *handle,
 	return 0;
 
 fail:
-	__perf_output_wakeup(handle);
+	perf_output_wakeup(handle);
 out:
 	rcu_read_unlock();
 
@@ -1809,16 +1877,20 @@ static void perf_output_copy(struct perf_output_handle *handle,
 
 static void perf_output_end(struct perf_output_handle *handle)
 {
-	int wakeup_events = handle->counter->hw_event.wakeup_events;
+	struct perf_counter *counter = handle->counter;
+	struct perf_mmap_data *data = handle->data;
+
+	int wakeup_events = counter->hw_event.wakeup_events;
 
 	if (handle->overflow && wakeup_events) {
-		int events = atomic_inc_return(&handle->data->events);
+		int events = atomic_inc_return(&data->events);
 		if (events >= wakeup_events) {
-			atomic_sub(wakeup_events, &handle->data->events);
-			__perf_output_wakeup(handle);
+			atomic_sub(wakeup_events, &data->events);
+			handle->wakeup = 1;
 		}
-	} else if (handle->wakeup)
-		__perf_output_wakeup(handle);
+	}
+
+	perf_output_unlock(handle);
 	rcu_read_unlock();
 }
 

[PATCH 2/4] perf_counter: fix nmi-watchdog interaction

[Peter Zijlstra]

[-- Attachment #1: perf_counter-fix-nmi.patch --]
[-- Type: text/plain, Size: 670 bytes --]

When we don't have any perf-counters active, don't act like we know
what the NMI is for.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
---
 arch/x86/kernel/cpu/perf_counter.c |    3 +++
 1 file changed, 3 insertions(+)

Index: linux-2.6/arch/x86/kernel/cpu/perf_counter.c
===================================================================
--- linux-2.6.orig/arch/x86/kernel/cpu/perf_counter.c
+++ linux-2.6/arch/x86/kernel/cpu/perf_counter.c
@@ -891,6 +891,9 @@ perf_counter_nmi_handler(struct notifier
 	struct pt_regs *regs;
 	int ret;
 
+	if (!atomic_read(&num_counters))
+		return NOTIFY_DONE;
+
 	switch (cmd) {
 	case DIE_NMI:
 	case DIE_NMI_IPI:

-- 

[tip:perfcounters/core] perf_counter: fix nmi-watchdog interaction

[tip-bot for Peter Zijlstra]

Commit-ID:  63a809a2dc53b91268dd915bbcbd425063893676
Gitweb:     http://git.kernel.org/tip/63a809a2dc53b91268dd915bbcbd425063893676
Author:     Peter Zijlstra <a.p.zijlstra@chello.nl>
AuthorDate: Fri, 1 May 2009 12:23:17 +0200
Committer:  Ingo Molnar <mingo@elte.hu>
CommitDate: Fri, 1 May 2009 13:23:44 +0200

perf_counter: fix nmi-watchdog interaction

When we don't have any perf-counters active, don't act like we know
what the NMI is for.

[ Impact: fix hard hang with nmi_watchdog=2 ]

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
LKML-Reference: <20090501102533.109867793@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>


---
 arch/x86/kernel/cpu/perf_counter.c |    3 +++
 1 files changed, 3 insertions(+), 0 deletions(-)

diff --git a/arch/x86/kernel/cpu/perf_counter.c b/arch/x86/kernel/cpu/perf_counter.c
index fc06f4d..d4c0cc9 100644
--- a/arch/x86/kernel/cpu/perf_counter.c
+++ b/arch/x86/kernel/cpu/perf_counter.c
@@ -871,6 +871,9 @@ perf_counter_nmi_handler(struct notifier_block *self,
 	struct pt_regs *regs;
 	int ret;
 
+	if (!atomic_read(&num_counters))
+		return NOTIFY_DONE;
+
 	switch (cmd) {
 	case DIE_NMI:
 	case DIE_NMI_IPI:

[PATCH 3/4] perf_counter: tool: handle 0-length data files

[Peter Zijlstra]

[-- Attachment #1: perf_counter-report-zero-file.patch --]
[-- Type: text/plain, Size: 650 bytes --]

Avoid perf-report barfing on 0-length data files.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
---
 Documentation/perf_counter/perf-report.cc |    5 +++++
 1 file changed, 5 insertions(+)

Index: linux-2.6/Documentation/perf_counter/perf-report.cc
===================================================================
--- linux-2.6.orig/Documentation/perf_counter/perf-report.cc
+++ linux-2.6/Documentation/perf_counter/perf-report.cc
@@ -402,6 +402,11 @@ int main(int argc, char *argv[])
 		exit(-1);
 	}
 
+	if (!stat.st_size) {
+		fprintf(stderr, "zero-sized file, nothing to do!\n");
+		exit(0);
+	}
+
 	load_kallsyms();
 
 remap:

-- 

[tip:perfcounters/core] perf_counter: tool: handle 0-length data files

[tip-bot for Peter Zijlstra]

Commit-ID:  585e3374d9d29376c2c37d821c8b7637dd48ca95
Gitweb:     http://git.kernel.org/tip/585e3374d9d29376c2c37d821c8b7637dd48ca95
Author:     Peter Zijlstra <a.p.zijlstra@chello.nl>
AuthorDate: Fri, 1 May 2009 12:23:18 +0200
Committer:  Ingo Molnar <mingo@elte.hu>
CommitDate: Fri, 1 May 2009 13:23:44 +0200

perf_counter: tool: handle 0-length data files

Avoid perf-report barfing on 0-length data files.

[ Impact: fix perf-report SIGBUS ]

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
LKML-Reference: <20090501102533.196245693@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>


---
 Documentation/perf_counter/perf-report.cc |    5 +++++
 1 files changed, 5 insertions(+), 0 deletions(-)

diff --git a/Documentation/perf_counter/perf-report.cc b/Documentation/perf_counter/perf-report.cc
index 933a075..911d7f3 100644
--- a/Documentation/perf_counter/perf-report.cc
+++ b/Documentation/perf_counter/perf-report.cc
@@ -402,6 +402,11 @@ int main(int argc, char *argv[])
 		exit(-1);
 	}
 
+	if (!stat.st_size) {
+		fprintf(stderr, "zero-sized file, nothing to do!\n");
+		exit(0);
+	}
+
 	load_kallsyms();
 
 remap:

[PATCH 4/4] perf_counter: documetation update

[Peter Zijlstra]

[-- Attachment #1: perf_counter-doc.patch --]
[-- Type: text/plain, Size: 13254 bytes --]

Update the documentation to reflect the current state of affairs

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
---
 Documentation/perf_counter/design.txt |  272 +++++++++++++++++++++++++++-------
 1 file changed, 219 insertions(+), 53 deletions(-)

Index: linux-2.6/Documentation/perf_counter/design.txt
===================================================================
--- linux-2.6.orig/Documentation/perf_counter/design.txt
+++ linux-2.6/Documentation/perf_counter/design.txt
@@ -34,41 +34,47 @@ can be poll()ed.
 
 When creating a new counter fd, 'perf_counter_hw_event' is:
 
-/*
- * Event to monitor via a performance monitoring counter:
- */
 struct perf_counter_hw_event {
-	__u64			event_config;
-
-	__u64			irq_period;
-	__u64			record_type;
-	__u64			read_format;
+        /*
+         * The MSB of the config word signifies if the rest contains cpu
+         * specific (raw) counter configuration data, if unset, the next
+         * 7 bits are an event type and the rest of the bits are the event
+         * identifier.
+         */
+        __u64                   config;
+
+        __u64                   irq_period;
+        __u32                   record_type;
+        __u32                   read_format;
+
+        __u64                   disabled       :  1, /* off by default        */
+                                nmi            :  1, /* NMI sampling          */
+                                inherit        :  1, /* children inherit it   */
+                                pinned         :  1, /* must always be on PMU */
+                                exclusive      :  1, /* only group on PMU     */
+                                exclude_user   :  1, /* don't count user      */
+                                exclude_kernel :  1, /* ditto kernel          */
+                                exclude_hv     :  1, /* ditto hypervisor      */
+                                exclude_idle   :  1, /* don't count when idle */
+                                mmap           :  1, /* include mmap data     */
+                                munmap         :  1, /* include munmap data   */
+                                comm           :  1, /* include comm data     */
 
-	__u64			disabled       :  1, /* off by default        */
-				nmi	       :  1, /* NMI sampling          */
-				inherit	       :  1, /* children inherit it   */
-				pinned	       :  1, /* must always be on PMU */
-				exclusive      :  1, /* only group on PMU     */
-				exclude_user   :  1, /* don't count user      */
-				exclude_kernel :  1, /* ditto kernel          */
-				exclude_hv     :  1, /* ditto hypervisor      */
-				exclude_idle   :  1, /* don't count when idle */
+                                __reserved_1   : 52;
 
-				__reserved_1   : 55;
+        __u32                   extra_config_len;
+        __u32                   wakeup_events;  /* wakeup every n events */
 
-	__u32			extra_config_len;
-
-	__u32			__reserved_4;
-	__u64			__reserved_2;
-	__u64			__reserved_3;
+        __u64                   __reserved_2;
+        __u64                   __reserved_3;
 };
 
-The 'event_config' field specifies what the counter should count.  It
+The 'config' field specifies what the counter should count.  It
 is divided into 3 bit-fields:
 
-raw_type: 1 bit (most significant bit)		0x8000_0000_0000_0000
-type:	  7 bits (next most significant)	0x7f00_0000_0000_0000
-event_id: 56 bits (least significant)		0x00ff_0000_0000_0000
+raw_type: 1 bit   (most significant bit)	0x8000_0000_0000_0000
+type:	  7 bits  (next most significant)	0x7f00_0000_0000_0000
+event_id: 56 bits (least significant)		0x00ff_ffff_ffff_ffff
 
 If 'raw_type' is 1, then the counter will count a hardware event
 specified by the remaining 63 bits of event_config.  The encoding is
@@ -134,41 +140,56 @@ enum sw_event_ids {
 	PERF_COUNT_PAGE_FAULTS_MAJ	= 6,
 };
 
+Counters of the type PERF_TYPE_TRACEPOINT are available when the ftrace event
+tracer is available, and event_id values can be obtained from
+/debug/tracing/events/*/*/id
+
+
 Counters come in two flavours: counting counters and sampling
 counters.  A "counting" counter is one that is used for counting the
 number of events that occur, and is characterised by having
-irq_period = 0 and record_type = PERF_RECORD_SIMPLE.  A read() on a
-counting counter simply returns the current value of the counter as
-an 8-byte number.
+irq_period = 0.
+
+
+A read() on a counter returns the current value of the counter and possible
+additional values as specified by 'read_format', each value is a u64 (8 bytes)
+in size.
+
+/*
+ * Bits that can be set in hw_event.read_format to request that
+ * reads on the counter should return the indicated quantities,
+ * in increasing order of bit value, after the counter value.
+ */
+enum perf_counter_read_format {
+        PERF_FORMAT_TOTAL_TIME_ENABLED  =  1,
+        PERF_FORMAT_TOTAL_TIME_RUNNING  =  2,
+};
+
+Using these additional values one can establish the overcommit ratio for a
+particular counter allowing one to take the round-robin scheduling effect
+into account.
+
 
 A "sampling" counter is one that is set up to generate an interrupt
 every N events, where N is given by 'irq_period'.  A sampling counter
-has irq_period > 0 and record_type != PERF_RECORD_SIMPLE.  The
-record_type controls what data is recorded on each interrupt, and the
-available values are currently:
+has irq_period > 0. The record_type controls what data is recorded on each
+interrupt:
 
 /*
- * IRQ-notification data record type:
+ * Bits that can be set in hw_event.record_type to request information
+ * in the overflow packets.
  */
-enum perf_counter_record_type {
-	PERF_RECORD_SIMPLE		= 0,
-	PERF_RECORD_IRQ			= 1,
-	PERF_RECORD_GROUP		= 2,
-};
-
-A record_type value of PERF_RECORD_IRQ will record the instruction
-pointer (IP) at which the interrupt occurred.  A record_type value of
-PERF_RECORD_GROUP will record the event_config and counter value of
-all of the other counters in the group, and should only be used on a
-group leader (see below).  Currently these two values are mutually
-exclusive, but record_type will become a bit-mask in future and
-support other values.
-
-A sampling counter has an event queue, into which an event is placed
-on each interrupt.  A read() on a sampling counter will read the next
-event from the event queue.  If the queue is empty, the read() will
-either block or return an EAGAIN error, depending on whether the fd
-has been set to non-blocking mode or not.
+enum perf_counter_record_format {
+        PERF_RECORD_IP          = 1U << 0,
+        PERF_RECORD_TID         = 1U << 1,
+        PERF_RECORD_TIME        = 1U << 2,
+        PERF_RECORD_ADDR        = 1U << 3,
+        PERF_RECORD_GROUP       = 1U << 4,
+        PERF_RECORD_CALLCHAIN   = 1U << 5,
+};
+
+Such (and other) events will be recorded in a ring-buffer, which is
+available to user-space using mmap() (see below).
 
 The 'disabled' bit specifies whether the counter starts out disabled
 or enabled.  If it is initially disabled, it can be enabled by ioctl
@@ -206,6 +227,13 @@ The 'exclude_user', 'exclude_kernel' and
 way to request that counting of events be restricted to times when the
 CPU is in user, kernel and/or hypervisor mode.
 
+The 'mmap' and 'munmap' bits allow recording of PROT_EXEC mmap/munmap
+operations, these can be used to relate userspace IP addresses to actual
+code, even after the mapping (or even the whole process) is gone,
+these events are recorded in the ring-buffer (see below).
+
+The 'comm' bit allows tracking of process comm data on process creation.
+This too is recorded in the ring-buffer (see below).
 
 The 'pid' parameter to the perf_counter_open() system call allows the
 counter to be specific to a task:
@@ -250,6 +278,138 @@ can be meaningfully compared, added, div
 with each other, since they have counted events for the same set of
 executed instructions.
 
+
+Like stated, asynchronous events, like counter overflow or PROT_EXEC mmap
+tracking are logged into a ring-buffer. This ring-buffer is created and
+accessed through mmap().
+
+The mmap size should be 1+2^n pages, where the first page is a meta-data page
+(struct perf_counter_mmap_page) that contains various bits of information such
+as where the ring-buffer head is.
+
+/*
+ * Structure of the page that can be mapped via mmap
+ */
+struct perf_counter_mmap_page {
+        __u32   version;                /* version number of this structure */
+        __u32   compat_version;         /* lowest version this is compat with */
+
+        /*
+         * Bits needed to read the hw counters in user-space.
+         *
+         *   u32 seq;
+         *   s64 count;
+         *
+         *   do {
+         *     seq = pc->lock;
+         *
+         *     barrier()
+         *     if (pc->index) {
+         *       count = pmc_read(pc->index - 1);
+         *       count += pc->offset;
+         *     } else
+         *       goto regular_read;
+         *
+         *     barrier();
+         *   } while (pc->lock != seq);
+         *
+         * NOTE: for obvious reason this only works on self-monitoring
+         *       processes.
+         */
+        __u32   lock;                   /* seqlock for synchronization */
+        __u32   index;                  /* hardware counter identifier */
+        __s64   offset;                 /* add to hardware counter value */
+
+        /*
+         * Control data for the mmap() data buffer.
+         *
+         * User-space reading this value should issue an rmb(), on SMP capable
+         * platforms, after reading this value -- see perf_counter_wakeup().
+         */
+        __u32   data_head;              /* head in the data section */
+};
+
+NOTE: the hw-counter userspace bits are arch specific and are currently only
+      implemented on powerpc.
+
+The following 2^n pages are the ring-buffer which contains events of the form:
+
+#define PERF_EVENT_MISC_KERNEL          (1 << 0)
+#define PERF_EVENT_MISC_USER            (1 << 1)
+#define PERF_EVENT_MISC_OVERFLOW        (1 << 2)
+
+struct perf_event_header {
+        __u32   type;
+        __u16   misc;
+        __u16   size;
+};
+
+enum perf_event_type {
+
+        /*
+         * The MMAP events record the PROT_EXEC mappings so that we can
+         * correlate userspace IPs to code. They have the following structure:
+         *
+         * struct {
+         *      struct perf_event_header        header;
+         *
+         *      u32                             pid, tid;
+         *      u64                             addr;
+         *      u64                             len;
+         *      u64                             pgoff;
+         *      char                            filename[];
+         * };
+         */
+        PERF_EVENT_MMAP                 = 1,
+        PERF_EVENT_MUNMAP               = 2,
+
+        /*
+         * struct {
+         *      struct perf_event_header        header;
+         *
+         *      u32                             pid, tid;
+         *      char                            comm[];
+         * };
+         */
+        PERF_EVENT_COMM                 = 3,
+
+        /*
+         * When header.misc & PERF_EVENT_MISC_OVERFLOW the event_type field
+         * will be PERF_RECORD_*
+         *
+         * struct {
+         *      struct perf_event_header        header;
+         *
+         *      { u64                   ip;       } && PERF_RECORD_IP
+         *      { u32                   pid, tid; } && PERF_RECORD_TID
+         *      { u64                   time;     } && PERF_RECORD_TIME
+         *      { u64                   addr;     } && PERF_RECORD_ADDR
+         *
+         *      { u64                   nr;
+         *        { u64 event, val; }   cnt[nr];  } && PERF_RECORD_GROUP
+         *
+         *      { u16                   nr,
+         *                              hv,
+         *                              kernel,
+         *                              user;
+         *        u64                   ips[nr];  } && PERF_RECORD_CALLCHAIN
+         * };
+         */
+};
+
+NOTE: PERF_RECORD_CALLCHAIN is arch specific and currently only implemented
+      on x86.
+
+Notification of new events is possible through poll()/select()/epoll() and
+fcntl() managing signals.
+
+Normally a notification is generated for every page filled, however one can
+additionally set perf_counter_hw_event.wakeup_events to generate one every
+so many counter overflow events.
+
+Future work will include a splice() interface to the ring-buffer.
+
+
 Counters can be enabled and disabled in two ways: via ioctl and via
 prctl.  When a counter is disabled, it doesn't count or generate
 events but does continue to exist and maintain its count value.
@@ -269,6 +429,12 @@ group other than the leader only affects
 non-leader stops that counter from counting but doesn't affect any
 other counter.
 
+Additionally, non-inherited overflow counters can use
+
+	ioctl(fd, PERF_COUNTER_IOC_REFRESH, nr);
+
+to enable a counter for 'nr' events, after which it gets disabled again.
+
 A process can enable or disable all the counter groups that are
 attached to it, using prctl:
 

-- 

[tip:perfcounters/core] perf_counter: documentation update

[tip-bot for Peter Zijlstra]

Commit-ID:  e5791a808ae91a9e7e1b65ea9b8de0f96a043d88
Gitweb:     http://git.kernel.org/tip/e5791a808ae91a9e7e1b65ea9b8de0f96a043d88
Author:     Peter Zijlstra <a.p.zijlstra@chello.nl>
AuthorDate: Fri, 1 May 2009 12:23:19 +0200
Committer:  Ingo Molnar <mingo@elte.hu>
CommitDate: Fri, 1 May 2009 13:23:45 +0200

perf_counter: documentation update

Update the documentation to reflect the current state of affairs

[ Impact: documentation update ]

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
LKML-Reference: <20090501102533.296727903@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>


---
 Documentation/perf_counter/design.txt |  274 ++++++++++++++++++++++++++-------
 1 files changed, 220 insertions(+), 54 deletions(-)

diff --git a/Documentation/perf_counter/design.txt b/Documentation/perf_counter/design.txt
index aaf105c..9930c4b 100644
--- a/Documentation/perf_counter/design.txt
+++ b/Documentation/perf_counter/design.txt
@@ -34,41 +34,47 @@ can be poll()ed.
 
 When creating a new counter fd, 'perf_counter_hw_event' is:
 
-/*
- * Event to monitor via a performance monitoring counter:
- */
 struct perf_counter_hw_event {
-	__u64			event_config;
-
-	__u64			irq_period;
-	__u64			record_type;
-	__u64			read_format;
-
-	__u64			disabled       :  1, /* off by default        */
-				nmi	       :  1, /* NMI sampling          */
-				inherit	       :  1, /* children inherit it   */
-				pinned	       :  1, /* must always be on PMU */
-				exclusive      :  1, /* only group on PMU     */
-				exclude_user   :  1, /* don't count user      */
-				exclude_kernel :  1, /* ditto kernel          */
-				exclude_hv     :  1, /* ditto hypervisor      */
-				exclude_idle   :  1, /* don't count when idle */
-
-				__reserved_1   : 55;
-
-	__u32			extra_config_len;
-
-	__u32			__reserved_4;
-	__u64			__reserved_2;
-	__u64			__reserved_3;
+        /*
+         * The MSB of the config word signifies if the rest contains cpu
+         * specific (raw) counter configuration data, if unset, the next
+         * 7 bits are an event type and the rest of the bits are the event
+         * identifier.
+         */
+        __u64                   config;
+
+        __u64                   irq_period;
+        __u32                   record_type;
+        __u32                   read_format;
+
+        __u64                   disabled       :  1, /* off by default        */
+                                nmi            :  1, /* NMI sampling          */
+                                inherit        :  1, /* children inherit it   */
+                                pinned         :  1, /* must always be on PMU */
+                                exclusive      :  1, /* only group on PMU     */
+                                exclude_user   :  1, /* don't count user      */
+                                exclude_kernel :  1, /* ditto kernel          */
+                                exclude_hv     :  1, /* ditto hypervisor      */
+                                exclude_idle   :  1, /* don't count when idle */
+                                mmap           :  1, /* include mmap data     */
+                                munmap         :  1, /* include munmap data   */
+                                comm           :  1, /* include comm data     */
+
+                                __reserved_1   : 52;
+
+        __u32                   extra_config_len;
+        __u32                   wakeup_events;  /* wakeup every n events */
+
+        __u64                   __reserved_2;
+        __u64                   __reserved_3;
 };
 
-The 'event_config' field specifies what the counter should count.  It
+The 'config' field specifies what the counter should count.  It
 is divided into 3 bit-fields:
 
-raw_type: 1 bit (most significant bit)		0x8000_0000_0000_0000
-type:	  7 bits (next most significant)	0x7f00_0000_0000_0000
-event_id: 56 bits (least significant)		0x00ff_0000_0000_0000
+raw_type: 1 bit   (most significant bit)	0x8000_0000_0000_0000
+type:	  7 bits  (next most significant)	0x7f00_0000_0000_0000
+event_id: 56 bits (least significant)		0x00ff_ffff_ffff_ffff
 
 If 'raw_type' is 1, then the counter will count a hardware event
 specified by the remaining 63 bits of event_config.  The encoding is
@@ -134,41 +140,56 @@ enum sw_event_ids {
 	PERF_COUNT_PAGE_FAULTS_MAJ	= 6,
 };
 
+Counters of the type PERF_TYPE_TRACEPOINT are available when the ftrace event
+tracer is available, and event_id values can be obtained from
+/debug/tracing/events/*/*/id
+
+
 Counters come in two flavours: counting counters and sampling
 counters.  A "counting" counter is one that is used for counting the
 number of events that occur, and is characterised by having
-irq_period = 0 and record_type = PERF_RECORD_SIMPLE.  A read() on a
-counting counter simply returns the current value of the counter as
-an 8-byte number.
+irq_period = 0.
+
+
+A read() on a counter returns the current value of the counter and possible
+additional values as specified by 'read_format', each value is a u64 (8 bytes)
+in size.
+
+/*
+ * Bits that can be set in hw_event.read_format to request that
+ * reads on the counter should return the indicated quantities,
+ * in increasing order of bit value, after the counter value.
+ */
+enum perf_counter_read_format {
+        PERF_FORMAT_TOTAL_TIME_ENABLED  =  1,
+        PERF_FORMAT_TOTAL_TIME_RUNNING  =  2,
+};
+
+Using these additional values one can establish the overcommit ratio for a
+particular counter allowing one to take the round-robin scheduling effect
+into account.
+
 
 A "sampling" counter is one that is set up to generate an interrupt
 every N events, where N is given by 'irq_period'.  A sampling counter
-has irq_period > 0 and record_type != PERF_RECORD_SIMPLE.  The
-record_type controls what data is recorded on each interrupt, and the
-available values are currently:
+has irq_period > 0. The record_type controls what data is recorded on each
+interrupt:
 
 /*
- * IRQ-notification data record type:
+ * Bits that can be set in hw_event.record_type to request information
+ * in the overflow packets.
  */
-enum perf_counter_record_type {
-	PERF_RECORD_SIMPLE		= 0,
-	PERF_RECORD_IRQ			= 1,
-	PERF_RECORD_GROUP		= 2,
+enum perf_counter_record_format {
+        PERF_RECORD_IP          = 1U << 0,
+        PERF_RECORD_TID         = 1U << 1,
+        PERF_RECORD_TIME        = 1U << 2,
+        PERF_RECORD_ADDR        = 1U << 3,
+        PERF_RECORD_GROUP       = 1U << 4,
+        PERF_RECORD_CALLCHAIN   = 1U << 5,
 };
 
-A record_type value of PERF_RECORD_IRQ will record the instruction
-pointer (IP) at which the interrupt occurred.  A record_type value of
-PERF_RECORD_GROUP will record the event_config and counter value of
-all of the other counters in the group, and should only be used on a
-group leader (see below).  Currently these two values are mutually
-exclusive, but record_type will become a bit-mask in future and
-support other values.
-
-A sampling counter has an event queue, into which an event is placed
-on each interrupt.  A read() on a sampling counter will read the next
-event from the event queue.  If the queue is empty, the read() will
-either block or return an EAGAIN error, depending on whether the fd
-has been set to non-blocking mode or not.
+Such (and other) events will be recorded in a ring-buffer, which is
+available to user-space using mmap() (see below).
 
 The 'disabled' bit specifies whether the counter starts out disabled
 or enabled.  If it is initially disabled, it can be enabled by ioctl
@@ -206,6 +227,13 @@ The 'exclude_user', 'exclude_kernel' and 'exclude_hv' bits provide a
 way to request that counting of events be restricted to times when the
 CPU is in user, kernel and/or hypervisor mode.
 
+The 'mmap' and 'munmap' bits allow recording of PROT_EXEC mmap/munmap
+operations, these can be used to relate userspace IP addresses to actual
+code, even after the mapping (or even the whole process) is gone,
+these events are recorded in the ring-buffer (see below).
+
+The 'comm' bit allows tracking of process comm data on process creation.
+This too is recorded in the ring-buffer (see below).
 
 The 'pid' parameter to the perf_counter_open() system call allows the
 counter to be specific to a task:
@@ -250,6 +278,138 @@ can be meaningfully compared, added, divided (to get ratios), etc.,
 with each other, since they have counted events for the same set of
 executed instructions.
 
+
+Like stated, asynchronous events, like counter overflow or PROT_EXEC mmap
+tracking are logged into a ring-buffer. This ring-buffer is created and
+accessed through mmap().
+
+The mmap size should be 1+2^n pages, where the first page is a meta-data page
+(struct perf_counter_mmap_page) that contains various bits of information such
+as where the ring-buffer head is.
+
+/*
+ * Structure of the page that can be mapped via mmap
+ */
+struct perf_counter_mmap_page {
+        __u32   version;                /* version number of this structure */
+        __u32   compat_version;         /* lowest version this is compat with */
+
+        /*
+         * Bits needed to read the hw counters in user-space.
+         *
+         *   u32 seq;
+         *   s64 count;
+         *
+         *   do {
+         *     seq = pc->lock;
+         *
+         *     barrier()
+         *     if (pc->index) {
+         *       count = pmc_read(pc->index - 1);
+         *       count += pc->offset;
+         *     } else
+         *       goto regular_read;
+         *
+         *     barrier();
+         *   } while (pc->lock != seq);
+         *
+         * NOTE: for obvious reason this only works on self-monitoring
+         *       processes.
+         */
+        __u32   lock;                   /* seqlock for synchronization */
+        __u32   index;                  /* hardware counter identifier */
+        __s64   offset;                 /* add to hardware counter value */
+
+        /*
+         * Control data for the mmap() data buffer.
+         *
+         * User-space reading this value should issue an rmb(), on SMP capable
+         * platforms, after reading this value -- see perf_counter_wakeup().
+         */
+        __u32   data_head;              /* head in the data section */
+};
+
+NOTE: the hw-counter userspace bits are arch specific and are currently only
+      implemented on powerpc.
+
+The following 2^n pages are the ring-buffer which contains events of the form:
+
+#define PERF_EVENT_MISC_KERNEL          (1 << 0)
+#define PERF_EVENT_MISC_USER            (1 << 1)
+#define PERF_EVENT_MISC_OVERFLOW        (1 << 2)
+
+struct perf_event_header {
+        __u32   type;
+        __u16   misc;
+        __u16   size;
+};
+
+enum perf_event_type {
+
+        /*
+         * The MMAP events record the PROT_EXEC mappings so that we can
+         * correlate userspace IPs to code. They have the following structure:
+         *
+         * struct {
+         *      struct perf_event_header        header;
+         *
+         *      u32                             pid, tid;
+         *      u64                             addr;
+         *      u64                             len;
+         *      u64                             pgoff;
+         *      char                            filename[];
+         * };
+         */
+        PERF_EVENT_MMAP                 = 1,
+        PERF_EVENT_MUNMAP               = 2,
+
+        /*
+         * struct {
+         *      struct perf_event_header        header;
+         *
+         *      u32                             pid, tid;
+         *      char                            comm[];
+         * };
+         */
+        PERF_EVENT_COMM                 = 3,
+
+        /*
+         * When header.misc & PERF_EVENT_MISC_OVERFLOW the event_type field
+         * will be PERF_RECORD_*
+         *
+         * struct {
+         *      struct perf_event_header        header;
+         *
+         *      { u64                   ip;       } && PERF_RECORD_IP
+         *      { u32                   pid, tid; } && PERF_RECORD_TID
+         *      { u64                   time;     } && PERF_RECORD_TIME
+         *      { u64                   addr;     } && PERF_RECORD_ADDR
+         *
+         *      { u64                   nr;
+         *        { u64 event, val; }   cnt[nr];  } && PERF_RECORD_GROUP
+         *
+         *      { u16                   nr,
+         *                              hv,
+         *                              kernel,
+         *                              user;
+         *        u64                   ips[nr];  } && PERF_RECORD_CALLCHAIN
+         * };
+         */
+};
+
+NOTE: PERF_RECORD_CALLCHAIN is arch specific and currently only implemented
+      on x86.
+
+Notification of new events is possible through poll()/select()/epoll() and
+fcntl() managing signals.
+
+Normally a notification is generated for every page filled, however one can
+additionally set perf_counter_hw_event.wakeup_events to generate one every
+so many counter overflow events.
+
+Future work will include a splice() interface to the ring-buffer.
+
+
 Counters can be enabled and disabled in two ways: via ioctl and via
 prctl.  When a counter is disabled, it doesn't count or generate
 events but does continue to exist and maintain its count value.
@@ -269,6 +429,12 @@ group other than the leader only affects that counter - disabling an
 non-leader stops that counter from counting but doesn't affect any
 other counter.
 
+Additionally, non-inherited overflow counters can use
+
+	ioctl(fd, PERF_COUNTER_IOC_REFRESH, nr);
+
+to enable a counter for 'nr' events, after which it gets disabled again.
+
 A process can enable or disable all the counter groups that are
 attached to it, using prctl:
 

[PATCH -tip] perf_counter: documentation update

[Yong Wang]

The 'nmi' bit is no longer there.

Signed-off-by: Yong Wang <yong.y.wang@intel.com>

---
 design.txt |    7 -------
 1 file changed, 7 deletions(-)

diff --git a/Documentation/perf_counter/design.txt b/Documentation/perf_counter/design.txt
index 9930c4b..d325076 100644
--- a/Documentation/perf_counter/design.txt
+++ b/Documentation/perf_counter/design.txt
@@ -48,7 +48,6 @@ struct perf_counter_hw_event {
         __u32                   read_format;
 
         __u64                   disabled       :  1, /* off by default        */
-                                nmi            :  1, /* NMI sampling          */
                                 inherit        :  1, /* children inherit it   */
                                 pinned         :  1, /* must always be on PMU */
                                 exclusive      :  1, /* only group on PMU     */
@@ -195,12 +194,6 @@ The 'disabled' bit specifies whether the counter starts out disabled
 or enabled.  If it is initially disabled, it can be enabled by ioctl
 or prctl (see below).
 
-The 'nmi' bit specifies, for hardware events, whether the counter
-should be set up to request non-maskable interrupts (NMIs) or normal
-interrupts.  This bit is ignored if the user doesn't have
-CAP_SYS_ADMIN privilege (i.e. is not root) or if the CPU doesn't
-generate NMIs from hardware counters.
-
 The 'inherit' bit, if set, specifies that this counter should count
 events on descendant tasks as well as the task specified.  This only
 applies to new descendents, not to any existing descendents at the

[tip:perfcounters/core] perf_counter: Documentation update

[tip-bot for Yong Wang]

Commit-ID:  3aff27ca84fa94311ae99189e54fed8d83b69fc1
Gitweb:     http://git.kernel.org/tip/3aff27ca84fa94311ae99189e54fed8d83b69fc1
Author:     Yong Wang <yong.y.wang@linux.intel.com>
AuthorDate: Wed, 3 Jun 2009 16:42:25 +0800
Committer:  Ingo Molnar <mingo@elte.hu>
CommitDate: Thu, 4 Jun 2009 13:20:12 +0200

perf_counter: Documentation update

The 'nmi' bit is no longer there.

Signed-off-by: Yong Wang <yong.y.wang@intel.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Arjan van de Ven <arjan@linux.intel.com>
LKML-Reference: <20090603084225.GA6553@ywang-moblin2.bj.intel.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>


---
 Documentation/perf_counter/design.txt |    7 -------
 1 files changed, 0 insertions(+), 7 deletions(-)

diff --git a/Documentation/perf_counter/design.txt b/Documentation/perf_counter/design.txt
index 9930c4b..d325076 100644
--- a/Documentation/perf_counter/design.txt
+++ b/Documentation/perf_counter/design.txt
@@ -48,7 +48,6 @@ struct perf_counter_hw_event {
         __u32                   read_format;
 
         __u64                   disabled       :  1, /* off by default        */
-                                nmi            :  1, /* NMI sampling          */
                                 inherit        :  1, /* children inherit it   */
                                 pinned         :  1, /* must always be on PMU */
                                 exclusive      :  1, /* only group on PMU     */
@@ -195,12 +194,6 @@ The 'disabled' bit specifies whether the counter starts out disabled
 or enabled.  If it is initially disabled, it can be enabled by ioctl
 or prctl (see below).
 
-The 'nmi' bit specifies, for hardware events, whether the counter
-should be set up to request non-maskable interrupts (NMIs) or normal
-interrupts.  This bit is ignored if the user doesn't have
-CAP_SYS_ADMIN privilege (i.e. is not root) or if the CPU doesn't
-generate NMIs from hardware counters.
-
 The 'inherit' bit, if set, specifies that this counter should count
 events on descendant tasks as well as the task specified.  This only
 applies to new descendents, not to any existing descendents at the