From fd13f6c85144bb2026c534a35be1d7cb7628a64a Mon Sep 17 00:00:00 2001
From: Robert Richter <robert.richter@amd.com>
Date: Sun, 19 Oct 2008 21:00:09 +0200
Subject: oprofile: comment cleanup

This fixes the coding style of some comments.

Signed-off-by: Robert Richter <robert.richter@amd.com>
---
 drivers/oprofile/cpu_buffer.c | 25 ++++++++++++++++---------
 1 file changed, 16 insertions(+), 9 deletions(-)

(limited to 'drivers/oprofile/cpu_buffer.c')

diff --git a/drivers/oprofile/cpu_buffer.c b/drivers/oprofile/cpu_buffer.c
index 01d38e78cde1..3958107723fb 100644
--- a/drivers/oprofile/cpu_buffer.c
+++ b/drivers/oprofile/cpu_buffer.c
@@ -127,9 +127,10 @@ void end_cpu_work(void)
 /* Resets the cpu buffer to a sane state. */
 void cpu_buffer_reset(struct oprofile_cpu_buffer *cpu_buf)
 {
-	/* reset these to invalid values; the next sample
-	 * collected will populate the buffer with proper
-	 * values to initialize the buffer
+	/*
+	 * reset these to invalid values; the next sample collected
+	 * will populate the buffer with proper values to initialize
+	 * the buffer
 	 */
 	cpu_buf->last_is_kernel = -1;
 	cpu_buf->last_task = NULL;
@@ -151,8 +152,10 @@ static void increment_head(struct oprofile_cpu_buffer *b)
 {
 	unsigned long new_head = b->head_pos + 1;
 
-	/* Ensure anything written to the slot before we
-	 * increment is visible */
+	/*
+	 * Ensure anything written to the slot before we increment is
+	 * visible
+	 */
 	wmb();
 
 	if (new_head < b->buffer_size)
@@ -253,8 +256,10 @@ void oprofile_add_ext_sample(unsigned long pc, struct pt_regs * const regs,
 	if (!oprofile_begin_trace(cpu_buf))
 		return;
 
-	/* if log_sample() fail we can't backtrace since we lost the source
-	 * of this event */
+	/*
+	 * if log_sample() fail we can't backtrace since we lost the
+	 * source of this event
+	 */
 	if (log_sample(cpu_buf, pc, is_kernel, event))
 		oprofile_ops.backtrace(regs, backtrace_depth);
 	oprofile_end_trace(cpu_buf);
@@ -338,8 +343,10 @@ void oprofile_add_trace(unsigned long pc)
 		return;
 	}
 
-	/* broken frame can give an eip with the same value as an escape code,
-	 * abort the trace if we get it */
+	/*
+	 * broken frame can give an eip with the same value as an
+	 * escape code, abort the trace if we get it
+	 */
 	if (pc == ESCAPE_CODE) {
 		cpu_buf->tracing = 0;
 		cpu_buf->backtrace_aborted++;
-- 
cgit v1.2.2


From cdc1834d1aa2e5b574a25e66f82625b44cdd0d8f Mon Sep 17 00:00:00 2001
From: Robert Richter <robert.richter@amd.com>
Date: Fri, 26 Sep 2008 22:18:44 -0400
Subject: oprofile: whitspace changes only

Signed-off-by: Robert Richter <robert.richter@amd.com>
---
 drivers/oprofile/cpu_buffer.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

(limited to 'drivers/oprofile/cpu_buffer.c')

diff --git a/drivers/oprofile/cpu_buffer.c b/drivers/oprofile/cpu_buffer.c
index 3958107723fb..2c4d54187b90 100644
--- a/drivers/oprofile/cpu_buffer.c
+++ b/drivers/oprofile/cpu_buffer.c
@@ -277,8 +277,8 @@ void oprofile_add_sample(struct pt_regs * const regs, unsigned long event)
 
 #define MAX_IBS_SAMPLE_SIZE 14
 
-void oprofile_add_ibs_sample(struct pt_regs *const regs,
-			     unsigned int *const ibs_sample, int ibs_code)
+void oprofile_add_ibs_sample(struct pt_regs * const regs,
+			     unsigned int * const ibs_sample, int ibs_code)
 {
 	int is_kernel = !user_mode(regs);
 	struct oprofile_cpu_buffer *cpu_buf = &__get_cpu_var(cpu_buffer);
-- 
cgit v1.2.2


From 7d468abee0f1a7e918b5e2f23120436a54ba9f33 Mon Sep 17 00:00:00 2001
From: Robert Richter <robert.richter@amd.com>
Date: Thu, 27 Nov 2008 10:57:09 +0100
Subject: oprofile: adding cpu buffer r/w access functions

This is in preparation for changes in the cpu buffer implementation.

Signed-off-by: Robert Richter <robert.richter@amd.com>
---
 drivers/oprofile/cpu_buffer.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

(limited to 'drivers/oprofile/cpu_buffer.c')

diff --git a/drivers/oprofile/cpu_buffer.c b/drivers/oprofile/cpu_buffer.c
index 2c4d54187b90..7e5e650e409f 100644
--- a/drivers/oprofile/cpu_buffer.c
+++ b/drivers/oprofile/cpu_buffer.c
@@ -168,7 +168,7 @@ static inline void
 add_sample(struct oprofile_cpu_buffer *cpu_buf,
 	   unsigned long pc, unsigned long event)
 {
-	struct op_sample *entry = &cpu_buf->buffer[cpu_buf->head_pos];
+	struct op_sample *entry = cpu_buffer_write_entry(cpu_buf);
 	entry->eip = pc;
 	entry->event = event;
 	increment_head(cpu_buf);
-- 
cgit v1.2.2


From 229234ae4a5ed9376b2e0524da04b0e5edadbf76 Mon Sep 17 00:00:00 2001
From: Robert Richter <robert.richter@amd.com>
Date: Thu, 27 Nov 2008 18:36:08 +0100
Subject: oprofile: adding cpu_buffer_write_commit()

This is in preparation for changes in the cpu buffer implementation.

Signed-off-by: Robert Richter <robert.richter@amd.com>
---
 drivers/oprofile/cpu_buffer.c | 18 +-----------------
 1 file changed, 1 insertion(+), 17 deletions(-)

(limited to 'drivers/oprofile/cpu_buffer.c')

diff --git a/drivers/oprofile/cpu_buffer.c b/drivers/oprofile/cpu_buffer.c
index 7e5e650e409f..d6f5de686363 100644
--- a/drivers/oprofile/cpu_buffer.c
+++ b/drivers/oprofile/cpu_buffer.c
@@ -148,22 +148,6 @@ static unsigned long nr_available_slots(struct oprofile_cpu_buffer const *b)
 	return tail + (b->buffer_size - head) - 1;
 }
 
-static void increment_head(struct oprofile_cpu_buffer *b)
-{
-	unsigned long new_head = b->head_pos + 1;
-
-	/*
-	 * Ensure anything written to the slot before we increment is
-	 * visible
-	 */
-	wmb();
-
-	if (new_head < b->buffer_size)
-		b->head_pos = new_head;
-	else
-		b->head_pos = 0;
-}
-
 static inline void
 add_sample(struct oprofile_cpu_buffer *cpu_buf,
 	   unsigned long pc, unsigned long event)
@@ -171,7 +155,7 @@ add_sample(struct oprofile_cpu_buffer *cpu_buf,
 	struct op_sample *entry = cpu_buffer_write_entry(cpu_buf);
 	entry->eip = pc;
 	entry->event = event;
-	increment_head(cpu_buf);
+	cpu_buffer_write_commit(cpu_buf);
 }
 
 static inline void
-- 
cgit v1.2.2


From fbc9bf9f0ed4f0fbc47dcb5b1c26c28c93b60e33 Mon Sep 17 00:00:00 2001
From: Robert Richter <robert.richter@amd.com>
Date: Thu, 4 Dec 2008 16:27:00 +0100
Subject: oprofile: moving cpu_buffer_reset() to cpu_buffer.h

This is in preparation for changes in the cpu buffer implementation.

Signed-off-by: Robert Richter <robert.richter@amd.com>
---
 drivers/oprofile/cpu_buffer.c | 12 ------------
 1 file changed, 12 deletions(-)

(limited to 'drivers/oprofile/cpu_buffer.c')

diff --git a/drivers/oprofile/cpu_buffer.c b/drivers/oprofile/cpu_buffer.c
index d6f5de686363..5cf7efe38e67 100644
--- a/drivers/oprofile/cpu_buffer.c
+++ b/drivers/oprofile/cpu_buffer.c
@@ -124,18 +124,6 @@ void end_cpu_work(void)
 	flush_scheduled_work();
 }
 
-/* Resets the cpu buffer to a sane state. */
-void cpu_buffer_reset(struct oprofile_cpu_buffer *cpu_buf)
-{
-	/*
-	 * reset these to invalid values; the next sample collected
-	 * will populate the buffer with proper values to initialize
-	 * the buffer
-	 */
-	cpu_buf->last_is_kernel = -1;
-	cpu_buf->last_task = NULL;
-}
-
 /* compute number of available slots in cpu_buffer queue */
 static unsigned long nr_available_slots(struct oprofile_cpu_buffer const *b)
 {
-- 
cgit v1.2.2


From 6dad828b76c7224a22ddc9ce7aa495d994f03b31 Mon Sep 17 00:00:00 2001
From: Robert Richter <robert.richter@amd.com>
Date: Tue, 9 Dec 2008 01:21:32 +0100
Subject: oprofile: port to the new ring_buffer

This patch replaces the current oprofile cpu buffer implementation
with the ring buffer provided by the tracing framework. The motivation
here is to leave the pain of implementing ring buffers to others. Oh,
no, there are more advantages. Main reason is the support of different
sample sizes that could be stored in the buffer. Use cases for this
are IBS and Cell spu profiling. Using the new ring buffer ensures
valid and complete samples and allows copying the cpu buffer stateless
without knowing its content. Second it will use generic kernel API and
also reduce code size. And hopefully, there are less bugs.

Since the new tracing ring buffer implementation uses spin locks to
protect the buffer during read/write access, it is difficult to use
the buffer in an NMI handler. In this case, writing to the buffer by
the NMI handler (x86) could occur also during critical sections when
reading the buffer. To avoid this, there are 2 buffers for independent
read and write access. Read access is in process context only, write
access only in the NMI handler. If the read buffer runs empty, both
buffers are swapped atomically. There is potentially a small window
during swapping where the buffers are disabled and samples could be
lost.

Using 2 buffers is a little bit overhead, but the solution is clear
and does not require changes in the ring buffer implementation. It can
be changed to a single buffer solution when the ring buffer access is
implemented as non-locking atomic code.

The new buffer requires more size to store the same amount of samples
because each sample includes an u32 header. Also, there is more code
to execute for buffer access. Nonetheless, the buffer implementation
is proven in the ftrace environment and worth to use also in oprofile.

Patches that changes the internal IBS buffer usage will follow.

Cc: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Robert Richter <robert.richter@amd.com>
---
 drivers/oprofile/cpu_buffer.c | 63 ++++++++++++++++++++++++++++++++-----------
 1 file changed, 48 insertions(+), 15 deletions(-)

(limited to 'drivers/oprofile/cpu_buffer.c')

diff --git a/drivers/oprofile/cpu_buffer.c b/drivers/oprofile/cpu_buffer.c
index 5cf7efe38e67..eb280ec96e24 100644
--- a/drivers/oprofile/cpu_buffer.c
+++ b/drivers/oprofile/cpu_buffer.c
@@ -28,6 +28,25 @@
 #include "buffer_sync.h"
 #include "oprof.h"
 
+#define OP_BUFFER_FLAGS	0
+
+/*
+ * Read and write access is using spin locking. Thus, writing to the
+ * buffer by NMI handler (x86) could occur also during critical
+ * sections when reading the buffer. To avoid this, there are 2
+ * buffers for independent read and write access. Read access is in
+ * process context only, write access only in the NMI handler. If the
+ * read buffer runs empty, both buffers are swapped atomically. There
+ * is potentially a small window during swapping where the buffers are
+ * disabled and samples could be lost.
+ *
+ * Using 2 buffers is a little bit overhead, but the solution is clear
+ * and does not require changes in the ring buffer implementation. It
+ * can be changed to a single buffer solution when the ring buffer
+ * access is implemented as non-locking atomic code.
+ */
+struct ring_buffer *op_ring_buffer_read;
+struct ring_buffer *op_ring_buffer_write;
 DEFINE_PER_CPU(struct oprofile_cpu_buffer, cpu_buffer);
 
 static void wq_sync_buffer(struct work_struct *work);
@@ -37,12 +56,12 @@ static int work_enabled;
 
 void free_cpu_buffers(void)
 {
-	int i;
-
-	for_each_possible_cpu(i) {
-		vfree(per_cpu(cpu_buffer, i).buffer);
-		per_cpu(cpu_buffer, i).buffer = NULL;
-	}
+	if (op_ring_buffer_read)
+		ring_buffer_free(op_ring_buffer_read);
+	op_ring_buffer_read = NULL;
+	if (op_ring_buffer_write)
+		ring_buffer_free(op_ring_buffer_write);
+	op_ring_buffer_write = NULL;
 }
 
 unsigned long oprofile_get_cpu_buffer_size(void)
@@ -64,14 +83,16 @@ int alloc_cpu_buffers(void)
 
 	unsigned long buffer_size = fs_cpu_buffer_size;
 
+	op_ring_buffer_read = ring_buffer_alloc(buffer_size, OP_BUFFER_FLAGS);
+	if (!op_ring_buffer_read)
+		goto fail;
+	op_ring_buffer_write = ring_buffer_alloc(buffer_size, OP_BUFFER_FLAGS);
+	if (!op_ring_buffer_write)
+		goto fail;
+
 	for_each_possible_cpu(i) {
 		struct oprofile_cpu_buffer *b = &per_cpu(cpu_buffer, i);
 
-		b->buffer = vmalloc_node(sizeof(struct op_sample) * buffer_size,
-			cpu_to_node(i));
-		if (!b->buffer)
-			goto fail;
-
 		b->last_task = NULL;
 		b->last_is_kernel = -1;
 		b->tracing = 0;
@@ -140,10 +161,22 @@ static inline void
 add_sample(struct oprofile_cpu_buffer *cpu_buf,
 	   unsigned long pc, unsigned long event)
 {
-	struct op_sample *entry = cpu_buffer_write_entry(cpu_buf);
-	entry->eip = pc;
-	entry->event = event;
-	cpu_buffer_write_commit(cpu_buf);
+	struct op_entry entry;
+
+	if (cpu_buffer_write_entry(&entry))
+		goto Error;
+
+	entry.sample->eip = pc;
+	entry.sample->event = event;
+
+	if (cpu_buffer_write_commit(&entry))
+		goto Error;
+
+	return;
+
+Error:
+	cpu_buf->sample_lost_overflow++;
+	return;
 }
 
 static inline void
-- 
cgit v1.2.2


From 1d7503b5dccf2b95babca050e4960e10d2633f2b Mon Sep 17 00:00:00 2001
From: Robert Richter <robert.richter@amd.com>
Date: Mon, 8 Dec 2008 11:59:52 +0100
Subject: oprofile: remove nr_available_slots()

This function is no longer available after the port to the new ring
buffer. Its removal can lead to incomplete sampling sequences since
IBS samples and backtraces are transfered in multiple samples. Due to
a full buffer, samples could be lost any time. The userspace daemon
has to live with such incomplete sampling sequences as long as the
data within one sample is consistent.

This will be fixed by changing the internal buffer data there all data
of one IBS sample or a backtrace is packed in a single ring buffer
entry. This is possible since the new ring buffer supports variable
data size.

Signed-off-by: Robert Richter <robert.richter@amd.com>
---
 drivers/oprofile/cpu_buffer.c | 34 ----------------------------------
 1 file changed, 34 deletions(-)

(limited to 'drivers/oprofile/cpu_buffer.c')

diff --git a/drivers/oprofile/cpu_buffer.c b/drivers/oprofile/cpu_buffer.c
index eb280ec96e24..7f7fc958297a 100644
--- a/drivers/oprofile/cpu_buffer.c
+++ b/drivers/oprofile/cpu_buffer.c
@@ -145,18 +145,6 @@ void end_cpu_work(void)
 	flush_scheduled_work();
 }
 
-/* compute number of available slots in cpu_buffer queue */
-static unsigned long nr_available_slots(struct oprofile_cpu_buffer const *b)
-{
-	unsigned long head = b->head_pos;
-	unsigned long tail = b->tail_pos;
-
-	if (tail > head)
-		return (tail - head) - 1;
-
-	return tail + (b->buffer_size - head) - 1;
-}
-
 static inline void
 add_sample(struct oprofile_cpu_buffer *cpu_buf,
 	   unsigned long pc, unsigned long event)
@@ -206,11 +194,6 @@ static int log_sample(struct oprofile_cpu_buffer *cpu_buf, unsigned long pc,
 		return 0;
 	}
 
-	if (nr_available_slots(cpu_buf) < 3) {
-		cpu_buf->sample_lost_overflow++;
-		return 0;
-	}
-
 	is_kernel = !!is_kernel;
 
 	task = current;
@@ -233,11 +216,6 @@ static int log_sample(struct oprofile_cpu_buffer *cpu_buf, unsigned long pc,
 
 static int oprofile_begin_trace(struct oprofile_cpu_buffer *cpu_buf)
 {
-	if (nr_available_slots(cpu_buf) < 4) {
-		cpu_buf->sample_lost_overflow++;
-		return 0;
-	}
-
 	add_code(cpu_buf, CPU_TRACE_BEGIN);
 	cpu_buf->tracing = 1;
 	return 1;
@@ -291,12 +269,6 @@ void oprofile_add_ibs_sample(struct pt_regs * const regs,
 
 	cpu_buf->sample_received++;
 
-	if (nr_available_slots(cpu_buf) < MAX_IBS_SAMPLE_SIZE) {
-		/* we can't backtrace since we lost the source of this event */
-		cpu_buf->sample_lost_overflow++;
-		return;
-	}
-
 	/* notice a switch from user->kernel or vice versa */
 	if (cpu_buf->last_is_kernel != is_kernel) {
 		cpu_buf->last_is_kernel = is_kernel;
@@ -342,12 +314,6 @@ void oprofile_add_trace(unsigned long pc)
 	if (!cpu_buf->tracing)
 		return;
 
-	if (nr_available_slots(cpu_buf) < 1) {
-		cpu_buf->tracing = 0;
-		cpu_buf->sample_lost_overflow++;
-		return;
-	}
-
 	/*
 	 * broken frame can give an eip with the same value as an
 	 * escape code, abort the trace if we get it
-- 
cgit v1.2.2


From 211117ff09b7d81d91b7857651587128ed8b13d9 Mon Sep 17 00:00:00 2001
From: Robert Richter <robert.richter@amd.com>
Date: Tue, 9 Dec 2008 02:13:25 +0100
Subject: oprofile: fix lost sample counter

The number of lost samples could be greater than the number of
received samples. This patches fixes this. The implementation
introduces return values for add_sample() and add_code().

Signed-off-by: Robert Richter <robert.richter@amd.com>
---
 drivers/oprofile/cpu_buffer.c | 83 +++++++++++++++++++++++++++----------------
 1 file changed, 53 insertions(+), 30 deletions(-)

(limited to 'drivers/oprofile/cpu_buffer.c')

diff --git a/drivers/oprofile/cpu_buffer.c b/drivers/oprofile/cpu_buffer.c
index 7f7fc958297a..61090969158f 100644
--- a/drivers/oprofile/cpu_buffer.c
+++ b/drivers/oprofile/cpu_buffer.c
@@ -145,32 +145,31 @@ void end_cpu_work(void)
 	flush_scheduled_work();
 }
 
-static inline void
+static inline int
 add_sample(struct oprofile_cpu_buffer *cpu_buf,
 	   unsigned long pc, unsigned long event)
 {
 	struct op_entry entry;
+	int ret;
 
-	if (cpu_buffer_write_entry(&entry))
-		goto Error;
+	ret = cpu_buffer_write_entry(&entry);
+	if (ret)
+		return ret;
 
 	entry.sample->eip = pc;
 	entry.sample->event = event;
 
-	if (cpu_buffer_write_commit(&entry))
-		goto Error;
+	ret = cpu_buffer_write_commit(&entry);
+	if (ret)
+		return ret;
 
-	return;
-
-Error:
-	cpu_buf->sample_lost_overflow++;
-	return;
+	return 0;
 }
 
-static inline void
+static inline int
 add_code(struct oprofile_cpu_buffer *buffer, unsigned long value)
 {
-	add_sample(buffer, ESCAPE_CODE, value);
+	return add_sample(buffer, ESCAPE_CODE, value);
 }
 
 /* This must be safe from any context. It's safe writing here
@@ -201,17 +200,25 @@ static int log_sample(struct oprofile_cpu_buffer *cpu_buf, unsigned long pc,
 	/* notice a switch from user->kernel or vice versa */
 	if (cpu_buf->last_is_kernel != is_kernel) {
 		cpu_buf->last_is_kernel = is_kernel;
-		add_code(cpu_buf, is_kernel);
+		if (add_code(cpu_buf, is_kernel))
+			goto fail;
 	}
 
 	/* notice a task switch */
 	if (cpu_buf->last_task != task) {
 		cpu_buf->last_task = task;
-		add_code(cpu_buf, (unsigned long)task);
+		if (add_code(cpu_buf, (unsigned long)task))
+			goto fail;
 	}
 
-	add_sample(cpu_buf, pc, event);
+	if (add_sample(cpu_buf, pc, event))
+		goto fail;
+
 	return 1;
+
+fail:
+	cpu_buf->sample_lost_overflow++;
+	return 0;
 }
 
 static int oprofile_begin_trace(struct oprofile_cpu_buffer *cpu_buf)
@@ -266,37 +273,49 @@ void oprofile_add_ibs_sample(struct pt_regs * const regs,
 	int is_kernel = !user_mode(regs);
 	struct oprofile_cpu_buffer *cpu_buf = &__get_cpu_var(cpu_buffer);
 	struct task_struct *task;
+	int fail = 0;
 
 	cpu_buf->sample_received++;
 
 	/* notice a switch from user->kernel or vice versa */
 	if (cpu_buf->last_is_kernel != is_kernel) {
+		if (add_code(cpu_buf, is_kernel))
+			goto fail;
 		cpu_buf->last_is_kernel = is_kernel;
-		add_code(cpu_buf, is_kernel);
 	}
 
 	/* notice a task switch */
 	if (!is_kernel) {
 		task = current;
 		if (cpu_buf->last_task != task) {
+			if (add_code(cpu_buf, (unsigned long)task))
+				goto fail;
 			cpu_buf->last_task = task;
-			add_code(cpu_buf, (unsigned long)task);
 		}
 	}
 
-	add_code(cpu_buf, ibs_code);
-	add_sample(cpu_buf, ibs_sample[0], ibs_sample[1]);
-	add_sample(cpu_buf, ibs_sample[2], ibs_sample[3]);
-	add_sample(cpu_buf, ibs_sample[4], ibs_sample[5]);
+	fail = fail || add_code(cpu_buf, ibs_code);
+	fail = fail || add_sample(cpu_buf, ibs_sample[0], ibs_sample[1]);
+	fail = fail || add_sample(cpu_buf, ibs_sample[2], ibs_sample[3]);
+	fail = fail || add_sample(cpu_buf, ibs_sample[4], ibs_sample[5]);
 
 	if (ibs_code == IBS_OP_BEGIN) {
-		add_sample(cpu_buf, ibs_sample[6], ibs_sample[7]);
-		add_sample(cpu_buf, ibs_sample[8], ibs_sample[9]);
-		add_sample(cpu_buf, ibs_sample[10], ibs_sample[11]);
+		fail = fail || add_sample(cpu_buf, ibs_sample[6], ibs_sample[7]);
+		fail = fail || add_sample(cpu_buf, ibs_sample[8], ibs_sample[9]);
+		fail = fail || add_sample(cpu_buf, ibs_sample[10], ibs_sample[11]);
 	}
 
+	if (fail)
+		goto fail;
+
 	if (backtrace_depth)
 		oprofile_ops.backtrace(regs, backtrace_depth);
+
+	return;
+
+fail:
+	cpu_buf->sample_lost_overflow++;
+	return;
 }
 
 #endif
@@ -318,13 +337,17 @@ void oprofile_add_trace(unsigned long pc)
 	 * broken frame can give an eip with the same value as an
 	 * escape code, abort the trace if we get it
 	 */
-	if (pc == ESCAPE_CODE) {
-		cpu_buf->tracing = 0;
-		cpu_buf->backtrace_aborted++;
-		return;
-	}
+	if (pc == ESCAPE_CODE)
+		goto fail;
+
+	if (add_sample(cpu_buf, pc, 0))
+		goto fail;
 
-	add_sample(cpu_buf, pc, 0);
+	return;
+fail:
+	cpu_buf->tracing = 0;
+	cpu_buf->backtrace_aborted++;
+	return;
 }
 
 /*
-- 
cgit v1.2.2


From bd2172f58094b3f8afa017e68f3f0b57577824e1 Mon Sep 17 00:00:00 2001
From: Robert Richter <robert.richter@amd.com>
Date: Tue, 16 Dec 2008 16:19:54 +0100
Subject: oprofile: rename kernel-wide identifiers

This patch renames kernel-wide identifiers to something more oprofile
specific names.

Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Robert Richter <robert.richter@amd.com>
---
 drivers/oprofile/cpu_buffer.c | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

(limited to 'drivers/oprofile/cpu_buffer.c')

diff --git a/drivers/oprofile/cpu_buffer.c b/drivers/oprofile/cpu_buffer.c
index 61090969158f..fcf96f608d86 100644
--- a/drivers/oprofile/cpu_buffer.c
+++ b/drivers/oprofile/cpu_buffer.c
@@ -66,7 +66,7 @@ void free_cpu_buffers(void)
 
 unsigned long oprofile_get_cpu_buffer_size(void)
 {
-	return fs_cpu_buffer_size;
+	return oprofile_cpu_buffer_size;
 }
 
 void oprofile_cpu_buffer_inc_smpl_lost(void)
@@ -81,7 +81,7 @@ int alloc_cpu_buffers(void)
 {
 	int i;
 
-	unsigned long buffer_size = fs_cpu_buffer_size;
+	unsigned long buffer_size = oprofile_cpu_buffer_size;
 
 	op_ring_buffer_read = ring_buffer_alloc(buffer_size, OP_BUFFER_FLAGS);
 	if (!op_ring_buffer_read)
@@ -238,7 +238,7 @@ void oprofile_add_ext_sample(unsigned long pc, struct pt_regs * const regs,
 {
 	struct oprofile_cpu_buffer *cpu_buf = &__get_cpu_var(cpu_buffer);
 
-	if (!backtrace_depth) {
+	if (!oprofile_backtrace_depth) {
 		log_sample(cpu_buf, pc, is_kernel, event);
 		return;
 	}
@@ -251,7 +251,7 @@ void oprofile_add_ext_sample(unsigned long pc, struct pt_regs * const regs,
 	 * source of this event
 	 */
 	if (log_sample(cpu_buf, pc, is_kernel, event))
-		oprofile_ops.backtrace(regs, backtrace_depth);
+		oprofile_ops.backtrace(regs, oprofile_backtrace_depth);
 	oprofile_end_trace(cpu_buf);
 }
 
@@ -308,8 +308,8 @@ void oprofile_add_ibs_sample(struct pt_regs * const regs,
 	if (fail)
 		goto fail;
 
-	if (backtrace_depth)
-		oprofile_ops.backtrace(regs, backtrace_depth);
+	if (oprofile_backtrace_depth)
+		oprofile_ops.backtrace(regs, oprofile_backtrace_depth);
 
 	return;
 
-- 
cgit v1.2.2


From 6d2c53f3cd81e33eec17aa99845d43e599986982 Mon Sep 17 00:00:00 2001
From: Robert Richter <robert.richter@amd.com>
Date: Wed, 24 Dec 2008 16:53:53 +0100
Subject: oprofile: rename cpu buffer functions

This patch renames cpu buffer functions to something more oprofile
specific names. Functions will be moved to the global name space.

Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Robert Richter <robert.richter@amd.com>
---
 drivers/oprofile/cpu_buffer.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

(limited to 'drivers/oprofile/cpu_buffer.c')

diff --git a/drivers/oprofile/cpu_buffer.c b/drivers/oprofile/cpu_buffer.c
index fcf96f608d86..e52c085cd186 100644
--- a/drivers/oprofile/cpu_buffer.c
+++ b/drivers/oprofile/cpu_buffer.c
@@ -152,14 +152,14 @@ add_sample(struct oprofile_cpu_buffer *cpu_buf,
 	struct op_entry entry;
 	int ret;
 
-	ret = cpu_buffer_write_entry(&entry);
+	ret = op_cpu_buffer_write_entry(&entry);
 	if (ret)
 		return ret;
 
 	entry.sample->eip = pc;
 	entry.sample->event = event;
 
-	ret = cpu_buffer_write_commit(&entry);
+	ret = op_cpu_buffer_write_commit(&entry);
 	if (ret)
 		return ret;
 
-- 
cgit v1.2.2


From 9966718daee592fbdc523703b2d8200009642506 Mon Sep 17 00:00:00 2001
From: Robert Richter <robert.richter@amd.com>
Date: Tue, 16 Dec 2008 16:19:54 +0100
Subject: oprofile: remove ring buffer inline functions in cpu_buffer.h

This patch moves ring buffer inline functions to cpu_buffer.c.

Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Robert Richter <robert.richter@amd.com>
---
 drivers/oprofile/cpu_buffer.c | 48 +++++++++++++++++++++++++++++++++++++++++--
 1 file changed, 46 insertions(+), 2 deletions(-)

(limited to 'drivers/oprofile/cpu_buffer.c')

diff --git a/drivers/oprofile/cpu_buffer.c b/drivers/oprofile/cpu_buffer.c
index e52c085cd186..cd67d4dd30b7 100644
--- a/drivers/oprofile/cpu_buffer.c
+++ b/drivers/oprofile/cpu_buffer.c
@@ -45,8 +45,8 @@
  * can be changed to a single buffer solution when the ring buffer
  * access is implemented as non-locking atomic code.
  */
-struct ring_buffer *op_ring_buffer_read;
-struct ring_buffer *op_ring_buffer_write;
+static struct ring_buffer *op_ring_buffer_read;
+static struct ring_buffer *op_ring_buffer_write;
 DEFINE_PER_CPU(struct oprofile_cpu_buffer, cpu_buffer);
 
 static void wq_sync_buffer(struct work_struct *work);
@@ -145,6 +145,50 @@ void end_cpu_work(void)
 	flush_scheduled_work();
 }
 
+int op_cpu_buffer_write_entry(struct op_entry *entry)
+{
+	entry->event = ring_buffer_lock_reserve(op_ring_buffer_write,
+						sizeof(struct op_sample),
+						&entry->irq_flags);
+	if (entry->event)
+		entry->sample = ring_buffer_event_data(entry->event);
+	else
+		entry->sample = NULL;
+
+	if (!entry->sample)
+		return -ENOMEM;
+
+	return 0;
+}
+
+int op_cpu_buffer_write_commit(struct op_entry *entry)
+{
+	return ring_buffer_unlock_commit(op_ring_buffer_write, entry->event,
+					 entry->irq_flags);
+}
+
+struct op_sample *op_cpu_buffer_read_entry(int cpu)
+{
+	struct ring_buffer_event *e;
+	e = ring_buffer_consume(op_ring_buffer_read, cpu, NULL);
+	if (e)
+		return ring_buffer_event_data(e);
+	if (ring_buffer_swap_cpu(op_ring_buffer_read,
+				 op_ring_buffer_write,
+				 cpu))
+		return NULL;
+	e = ring_buffer_consume(op_ring_buffer_read, cpu, NULL);
+	if (e)
+		return ring_buffer_event_data(e);
+	return NULL;
+}
+
+unsigned long op_cpu_buffer_entries(int cpu)
+{
+	return ring_buffer_entries_cpu(op_ring_buffer_read, cpu)
+		+ ring_buffer_entries_cpu(op_ring_buffer_write, cpu);
+}
+
 static inline int
 add_sample(struct oprofile_cpu_buffer *cpu_buf,
 	   unsigned long pc, unsigned long event)
-- 
cgit v1.2.2


From 300157768f050dabc73a99d958b504282088a132 Mon Sep 17 00:00:00 2001
From: Robert Richter <robert.richter@amd.com>
Date: Tue, 23 Dec 2008 01:35:12 +0100
Subject: oprofile: reordering some code in cpu_buffer.c

Reordering code to keep alloc/free functions together.

Signed-off-by: Robert Richter <robert.richter@amd.com>
---
 drivers/oprofile/cpu_buffer.c | 20 ++++++++++----------
 1 file changed, 10 insertions(+), 10 deletions(-)

(limited to 'drivers/oprofile/cpu_buffer.c')

diff --git a/drivers/oprofile/cpu_buffer.c b/drivers/oprofile/cpu_buffer.c
index cd67d4dd30b7..b353b19bd786 100644
--- a/drivers/oprofile/cpu_buffer.c
+++ b/drivers/oprofile/cpu_buffer.c
@@ -54,16 +54,6 @@ static void wq_sync_buffer(struct work_struct *work);
 #define DEFAULT_TIMER_EXPIRE (HZ / 10)
 static int work_enabled;
 
-void free_cpu_buffers(void)
-{
-	if (op_ring_buffer_read)
-		ring_buffer_free(op_ring_buffer_read);
-	op_ring_buffer_read = NULL;
-	if (op_ring_buffer_write)
-		ring_buffer_free(op_ring_buffer_write);
-	op_ring_buffer_write = NULL;
-}
-
 unsigned long oprofile_get_cpu_buffer_size(void)
 {
 	return oprofile_cpu_buffer_size;
@@ -77,6 +67,16 @@ void oprofile_cpu_buffer_inc_smpl_lost(void)
 	cpu_buf->sample_lost_overflow++;
 }
 
+void free_cpu_buffers(void)
+{
+	if (op_ring_buffer_read)
+		ring_buffer_free(op_ring_buffer_read);
+	op_ring_buffer_read = NULL;
+	if (op_ring_buffer_write)
+		ring_buffer_free(op_ring_buffer_write);
+	op_ring_buffer_write = NULL;
+}
+
 int alloc_cpu_buffers(void)
 {
 	int i;
-- 
cgit v1.2.2


From d45d23bed4bf7b25b7dcc336552a251db1aa1279 Mon Sep 17 00:00:00 2001
From: Robert Richter <robert.richter@amd.com>
Date: Tue, 16 Dec 2008 12:00:10 +0100
Subject: oprofile: add inline function __oprofile_add_ext_sample()

This patch adds the inline function __oprofile_add_ext_sample() to
cpu_buffer.c and thus reduces overhead when calling
oprofile_add_sample().

Signed-off-by: Robert Richter <robert.richter@amd.com>
---
 drivers/oprofile/cpu_buffer.c | 13 ++++++++++---
 1 file changed, 10 insertions(+), 3 deletions(-)

(limited to 'drivers/oprofile/cpu_buffer.c')

diff --git a/drivers/oprofile/cpu_buffer.c b/drivers/oprofile/cpu_buffer.c
index b353b19bd786..9e66c384e016 100644
--- a/drivers/oprofile/cpu_buffer.c
+++ b/drivers/oprofile/cpu_buffer.c
@@ -277,8 +277,9 @@ static void oprofile_end_trace(struct oprofile_cpu_buffer *cpu_buf)
 	cpu_buf->tracing = 0;
 }
 
-void oprofile_add_ext_sample(unsigned long pc, struct pt_regs * const regs,
-				unsigned long event, int is_kernel)
+static inline void
+__oprofile_add_ext_sample(unsigned long pc, struct pt_regs * const regs,
+			  unsigned long event, int is_kernel)
 {
 	struct oprofile_cpu_buffer *cpu_buf = &__get_cpu_var(cpu_buffer);
 
@@ -299,12 +300,18 @@ void oprofile_add_ext_sample(unsigned long pc, struct pt_regs * const regs,
 	oprofile_end_trace(cpu_buf);
 }
 
+void oprofile_add_ext_sample(unsigned long pc, struct pt_regs * const regs,
+			     unsigned long event, int is_kernel)
+{
+	__oprofile_add_ext_sample(pc, regs, event, is_kernel);
+}
+
 void oprofile_add_sample(struct pt_regs * const regs, unsigned long event)
 {
 	int is_kernel = !user_mode(regs);
 	unsigned long pc = profile_pc(regs);
 
-	oprofile_add_ext_sample(pc, regs, event, is_kernel);
+	__oprofile_add_ext_sample(pc, regs, event, is_kernel);
 }
 
 #ifdef CONFIG_OPROFILE_IBS
-- 
cgit v1.2.2


From 6352d92dec0c4b833c12a169e86762c05d0396f3 Mon Sep 17 00:00:00 2001
From: Robert Richter <robert.richter@amd.com>
Date: Thu, 18 Dec 2008 22:09:13 +0100
Subject: oprofile: simplify oprofile_begin_trace()

This patch removes the unused return parameter in
oprofile_begin_trace(). Also, oprofile_begin_trace() and
oprofile_end_trace() are inline now.

Signed-off-by: Robert Richter <robert.richter@amd.com>
---
 drivers/oprofile/cpu_buffer.c | 9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)

(limited to 'drivers/oprofile/cpu_buffer.c')

diff --git a/drivers/oprofile/cpu_buffer.c b/drivers/oprofile/cpu_buffer.c
index 9e66c384e016..435bd6e08d5b 100644
--- a/drivers/oprofile/cpu_buffer.c
+++ b/drivers/oprofile/cpu_buffer.c
@@ -265,14 +265,13 @@ fail:
 	return 0;
 }
 
-static int oprofile_begin_trace(struct oprofile_cpu_buffer *cpu_buf)
+static inline void oprofile_begin_trace(struct oprofile_cpu_buffer *cpu_buf)
 {
 	add_code(cpu_buf, CPU_TRACE_BEGIN);
 	cpu_buf->tracing = 1;
-	return 1;
 }
 
-static void oprofile_end_trace(struct oprofile_cpu_buffer *cpu_buf)
+static inline void oprofile_end_trace(struct oprofile_cpu_buffer *cpu_buf)
 {
 	cpu_buf->tracing = 0;
 }
@@ -288,8 +287,7 @@ __oprofile_add_ext_sample(unsigned long pc, struct pt_regs * const regs,
 		return;
 	}
 
-	if (!oprofile_begin_trace(cpu_buf))
-		return;
+	oprofile_begin_trace(cpu_buf);
 
 	/*
 	 * if log_sample() fail we can't backtrace since we lost the
@@ -297,6 +295,7 @@ __oprofile_add_ext_sample(unsigned long pc, struct pt_regs * const regs,
 	 */
 	if (log_sample(cpu_buf, pc, is_kernel, event))
 		oprofile_ops.backtrace(regs, oprofile_backtrace_depth);
+
 	oprofile_end_trace(cpu_buf);
 }
 
-- 
cgit v1.2.2


From 3967e93e063d7ee608f465cbccb65abb518e9d33 Mon Sep 17 00:00:00 2001
From: Robert Richter <robert.richter@amd.com>
Date: Tue, 30 Dec 2008 05:10:58 +0100
Subject: oprofile: simplify add_sample() in cpu_buffer.c

Signed-off-by: Robert Richter <robert.richter@amd.com>
---
 drivers/oprofile/cpu_buffer.c | 6 +-----
 1 file changed, 1 insertion(+), 5 deletions(-)

(limited to 'drivers/oprofile/cpu_buffer.c')

diff --git a/drivers/oprofile/cpu_buffer.c b/drivers/oprofile/cpu_buffer.c
index 435bd6e08d5b..d92f0020502e 100644
--- a/drivers/oprofile/cpu_buffer.c
+++ b/drivers/oprofile/cpu_buffer.c
@@ -203,11 +203,7 @@ add_sample(struct oprofile_cpu_buffer *cpu_buf,
 	entry.sample->eip = pc;
 	entry.sample->event = event;
 
-	ret = op_cpu_buffer_write_commit(&entry);
-	if (ret)
-		return ret;
-
-	return 0;
+	return op_cpu_buffer_write_commit(&entry);
 }
 
 static inline int
-- 
cgit v1.2.2


From 8d15df84a42b140a8262a325b987a283ef9f5f63 Mon Sep 17 00:00:00 2001
From: Robert Richter <robert.richter@amd.com>
Date: Wed, 24 Dec 2008 15:42:58 +0100
Subject: oprofile: remove unused components in struct oprofile_cpu_buffer

Signed-off-by: Robert Richter <robert.richter@amd.com>
---
 drivers/oprofile/cpu_buffer.c | 2 --
 1 file changed, 2 deletions(-)

(limited to 'drivers/oprofile/cpu_buffer.c')

diff --git a/drivers/oprofile/cpu_buffer.c b/drivers/oprofile/cpu_buffer.c
index d92f0020502e..b426ae8ad2e2 100644
--- a/drivers/oprofile/cpu_buffer.c
+++ b/drivers/oprofile/cpu_buffer.c
@@ -97,8 +97,6 @@ int alloc_cpu_buffers(void)
 		b->last_is_kernel = -1;
 		b->tracing = 0;
 		b->buffer_size = buffer_size;
-		b->tail_pos = 0;
-		b->head_pos = 0;
 		b->sample_received = 0;
 		b->sample_lost_overflow = 0;
 		b->backtrace_aborted = 0;
-- 
cgit v1.2.2


From f4ff2364417f0092e49f6a3aa474549a56697f2d Mon Sep 17 00:00:00 2001
From: Robert Richter <robert.richter@amd.com>
Date: Mon, 5 Jan 2009 11:27:52 +0100
Subject: oprofile: remove unused ibs macro

Signed-off-by: Robert Richter <robert.richter@amd.com>
---
 drivers/oprofile/cpu_buffer.c | 2 --
 1 file changed, 2 deletions(-)

(limited to 'drivers/oprofile/cpu_buffer.c')

diff --git a/drivers/oprofile/cpu_buffer.c b/drivers/oprofile/cpu_buffer.c
index b426ae8ad2e2..8ae37c9d0ec4 100644
--- a/drivers/oprofile/cpu_buffer.c
+++ b/drivers/oprofile/cpu_buffer.c
@@ -309,8 +309,6 @@ void oprofile_add_sample(struct pt_regs * const regs, unsigned long event)
 
 #ifdef CONFIG_OPROFILE_IBS
 
-#define MAX_IBS_SAMPLE_SIZE 14
-
 void oprofile_add_ibs_sample(struct pt_regs * const regs,
 			     unsigned int * const ibs_sample, int ibs_code)
 {
-- 
cgit v1.2.2


From 8350c78734e67ac1f8bfd4eb14b70ff4d01a9a12 Mon Sep 17 00:00:00 2001
From: Robert Richter <robert.richter@amd.com>
Date: Fri, 19 Dec 2008 12:59:28 +0100
Subject: oprofile: remove backtrace code for ibs

This code is broken since a TRACE_BEGIN_CODE is never sent to the
daemon. The data becomes corrupt since the backtrace is interpreted as
ibs sample.

Signed-off-by: Robert Richter <robert.richter@amd.com>
---
 drivers/oprofile/cpu_buffer.c | 10 ++--------
 1 file changed, 2 insertions(+), 8 deletions(-)

(limited to 'drivers/oprofile/cpu_buffer.c')

diff --git a/drivers/oprofile/cpu_buffer.c b/drivers/oprofile/cpu_buffer.c
index 8ae37c9d0ec4..92bf8c0d86fe 100644
--- a/drivers/oprofile/cpu_buffer.c
+++ b/drivers/oprofile/cpu_buffer.c
@@ -347,17 +347,11 @@ void oprofile_add_ibs_sample(struct pt_regs * const regs,
 		fail = fail || add_sample(cpu_buf, ibs_sample[10], ibs_sample[11]);
 	}
 
-	if (fail)
-		goto fail;
-
-	if (oprofile_backtrace_depth)
-		oprofile_ops.backtrace(regs, oprofile_backtrace_depth);
-
-	return;
+	if (!fail)
+		return;
 
 fail:
 	cpu_buf->sample_lost_overflow++;
-	return;
 }
 
 #endif
-- 
cgit v1.2.2


From d0e233846dcef56ae78f6d8fd0e0cba85a2a1489 Mon Sep 17 00:00:00 2001
From: Robert Richter <robert.richter@amd.com>
Date: Tue, 23 Dec 2008 04:03:05 +0100
Subject: oprofile: rename add_sample() in cpu_buffer.c

Rename the fucntion to op_add_sample() since there is a collision with
another one with the same name in buffer_sync.c.

Signed-off-by: Robert Richter <robert.richter@amd.com>
---
 drivers/oprofile/cpu_buffer.c | 22 +++++++++++-----------
 1 file changed, 11 insertions(+), 11 deletions(-)

(limited to 'drivers/oprofile/cpu_buffer.c')

diff --git a/drivers/oprofile/cpu_buffer.c b/drivers/oprofile/cpu_buffer.c
index 92bf8c0d86fe..ac79f6676033 100644
--- a/drivers/oprofile/cpu_buffer.c
+++ b/drivers/oprofile/cpu_buffer.c
@@ -188,8 +188,8 @@ unsigned long op_cpu_buffer_entries(int cpu)
 }
 
 static inline int
-add_sample(struct oprofile_cpu_buffer *cpu_buf,
-	   unsigned long pc, unsigned long event)
+op_add_sample(struct oprofile_cpu_buffer *cpu_buf,
+	      unsigned long pc, unsigned long event)
 {
 	struct op_entry entry;
 	int ret;
@@ -207,7 +207,7 @@ add_sample(struct oprofile_cpu_buffer *cpu_buf,
 static inline int
 add_code(struct oprofile_cpu_buffer *buffer, unsigned long value)
 {
-	return add_sample(buffer, ESCAPE_CODE, value);
+	return op_add_sample(buffer, ESCAPE_CODE, value);
 }
 
 /* This must be safe from any context. It's safe writing here
@@ -249,7 +249,7 @@ static int log_sample(struct oprofile_cpu_buffer *cpu_buf, unsigned long pc,
 			goto fail;
 	}
 
-	if (add_sample(cpu_buf, pc, event))
+	if (op_add_sample(cpu_buf, pc, event))
 		goto fail;
 
 	return 1;
@@ -337,14 +337,14 @@ void oprofile_add_ibs_sample(struct pt_regs * const regs,
 	}
 
 	fail = fail || add_code(cpu_buf, ibs_code);
-	fail = fail || add_sample(cpu_buf, ibs_sample[0], ibs_sample[1]);
-	fail = fail || add_sample(cpu_buf, ibs_sample[2], ibs_sample[3]);
-	fail = fail || add_sample(cpu_buf, ibs_sample[4], ibs_sample[5]);
+	fail = fail || op_add_sample(cpu_buf, ibs_sample[0], ibs_sample[1]);
+	fail = fail || op_add_sample(cpu_buf, ibs_sample[2], ibs_sample[3]);
+	fail = fail || op_add_sample(cpu_buf, ibs_sample[4], ibs_sample[5]);
 
 	if (ibs_code == IBS_OP_BEGIN) {
-		fail = fail || add_sample(cpu_buf, ibs_sample[6], ibs_sample[7]);
-		fail = fail || add_sample(cpu_buf, ibs_sample[8], ibs_sample[9]);
-		fail = fail || add_sample(cpu_buf, ibs_sample[10], ibs_sample[11]);
+		fail = fail || op_add_sample(cpu_buf, ibs_sample[6], ibs_sample[7]);
+		fail = fail || op_add_sample(cpu_buf, ibs_sample[8], ibs_sample[9]);
+		fail = fail || op_add_sample(cpu_buf, ibs_sample[10], ibs_sample[11]);
 	}
 
 	if (!fail)
@@ -376,7 +376,7 @@ void oprofile_add_trace(unsigned long pc)
 	if (pc == ESCAPE_CODE)
 		goto fail;
 
-	if (add_sample(cpu_buf, pc, 0))
+	if (op_add_sample(cpu_buf, pc, 0))
 		goto fail;
 
 	return;
-- 
cgit v1.2.2


From 2cc28b9f261dd28d69767a34682ce55a27d928ed Mon Sep 17 00:00:00 2001
From: Robert Richter <robert.richter@amd.com>
Date: Thu, 25 Dec 2008 17:26:07 +0100
Subject: oprofile: add op_cpu_buffer_write_reserve()

This function prepares the cpu buffer to write a sample.

Struct op_entry is used during operations on the ring buffer while
struct op_sample contains the data that is stored in the ring
buffer. Struct entry can be uninitialized. The function reserves a
data array that is specified by size. Use op_cpu_buffer_write_commit()
after preparing the sample. In case of errors a null pointer is
returned, otherwise the pointer to the sample.

Signed-off-by: Robert Richter <robert.richter@amd.com>
---
 drivers/oprofile/cpu_buffer.c | 43 ++++++++++++++++++++++++++++++-------------
 1 file changed, 30 insertions(+), 13 deletions(-)

(limited to 'drivers/oprofile/cpu_buffer.c')

diff --git a/drivers/oprofile/cpu_buffer.c b/drivers/oprofile/cpu_buffer.c
index ac79f6676033..934ff159e70e 100644
--- a/drivers/oprofile/cpu_buffer.c
+++ b/drivers/oprofile/cpu_buffer.c
@@ -1,11 +1,12 @@
 /**
  * @file cpu_buffer.c
  *
- * @remark Copyright 2002 OProfile authors
+ * @remark Copyright 2002-2009 OProfile authors
  * @remark Read the file COPYING
  *
  * @author John Levon <levon@movementarian.org>
  * @author Barry Kasindorf <barry.kasindorf@amd.com>
+ * @author Robert Richter <robert.richter@amd.com>
  *
  * Each CPU has a local buffer that stores PC value/event
  * pairs. We also log context switches when we notice them.
@@ -143,20 +144,36 @@ void end_cpu_work(void)
 	flush_scheduled_work();
 }
 
-int op_cpu_buffer_write_entry(struct op_entry *entry)
+/*
+ * This function prepares the cpu buffer to write a sample.
+ *
+ * Struct op_entry is used during operations on the ring buffer while
+ * struct op_sample contains the data that is stored in the ring
+ * buffer. Struct entry can be uninitialized. The function reserves a
+ * data array that is specified by size. Use
+ * op_cpu_buffer_write_commit() after preparing the sample. In case of
+ * errors a null pointer is returned, otherwise the pointer to the
+ * sample.
+ *
+ */
+struct op_sample
+*op_cpu_buffer_write_reserve(struct op_entry *entry, unsigned long size)
 {
-	entry->event = ring_buffer_lock_reserve(op_ring_buffer_write,
-						sizeof(struct op_sample),
-						&entry->irq_flags);
+	entry->event = ring_buffer_lock_reserve
+		(op_ring_buffer_write, sizeof(struct op_sample) +
+		 size * sizeof(entry->sample->data[0]), &entry->irq_flags);
 	if (entry->event)
 		entry->sample = ring_buffer_event_data(entry->event);
 	else
 		entry->sample = NULL;
 
 	if (!entry->sample)
-		return -ENOMEM;
+		return NULL;
 
-	return 0;
+	entry->size = size;
+	entry->data = entry->sample->data;
+
+	return entry->sample;
 }
 
 int op_cpu_buffer_write_commit(struct op_entry *entry)
@@ -192,14 +209,14 @@ op_add_sample(struct oprofile_cpu_buffer *cpu_buf,
 	      unsigned long pc, unsigned long event)
 {
 	struct op_entry entry;
-	int ret;
+	struct op_sample *sample;
 
-	ret = op_cpu_buffer_write_entry(&entry);
-	if (ret)
-		return ret;
+	sample = op_cpu_buffer_write_reserve(&entry, 0);
+	if (!sample)
+		return -ENOMEM;
 
-	entry.sample->eip = pc;
-	entry.sample->event = event;
+	sample->eip = pc;
+	sample->event = event;
 
 	return op_cpu_buffer_write_commit(&entry);
 }
-- 
cgit v1.2.2


From 2d87b14cf8d0b07720de26d90789d02124141616 Mon Sep 17 00:00:00 2001
From: Robert Richter <robert.richter@amd.com>
Date: Tue, 30 Dec 2008 04:10:46 +0100
Subject: oprofile: modify op_cpu_buffer_read_entry()

This implements the support of samples with attached data.

Signed-off-by: Robert Richter <robert.richter@amd.com>
---
 drivers/oprofile/cpu_buffer.c | 14 +++++++++++---
 1 file changed, 11 insertions(+), 3 deletions(-)

(limited to 'drivers/oprofile/cpu_buffer.c')

diff --git a/drivers/oprofile/cpu_buffer.c b/drivers/oprofile/cpu_buffer.c
index 934ff159e70e..400f7fcffdbe 100644
--- a/drivers/oprofile/cpu_buffer.c
+++ b/drivers/oprofile/cpu_buffer.c
@@ -182,20 +182,28 @@ int op_cpu_buffer_write_commit(struct op_entry *entry)
 					 entry->irq_flags);
 }
 
-struct op_sample *op_cpu_buffer_read_entry(int cpu)
+struct op_sample *op_cpu_buffer_read_entry(struct op_entry *entry, int cpu)
 {
 	struct ring_buffer_event *e;
 	e = ring_buffer_consume(op_ring_buffer_read, cpu, NULL);
 	if (e)
-		return ring_buffer_event_data(e);
+		goto event;
 	if (ring_buffer_swap_cpu(op_ring_buffer_read,
 				 op_ring_buffer_write,
 				 cpu))
 		return NULL;
 	e = ring_buffer_consume(op_ring_buffer_read, cpu, NULL);
 	if (e)
-		return ring_buffer_event_data(e);
+		goto event;
 	return NULL;
+
+event:
+	entry->event = e;
+	entry->sample = ring_buffer_event_data(e);
+	entry->size = (ring_buffer_event_length(e) - sizeof(struct op_sample))
+		/ sizeof(entry->sample->data[0]);
+	entry->data = entry->sample->data;
+	return entry->sample;
 }
 
 unsigned long op_cpu_buffer_entries(int cpu)
-- 
cgit v1.2.2


From ae735e9964b4584923f2997d98a8d80ae9c1a75c Mon Sep 17 00:00:00 2001
From: Robert Richter <robert.richter@amd.com>
Date: Thu, 25 Dec 2008 17:26:07 +0100
Subject: oprofile: rework implementation of cpu buffer events

Special events such as task or context switches are marked with an
escape code in the cpu buffer followed by an event code or a task
identifier. There is one escape code per event. To make escape
sequences also available for data samples the internal cpu buffer
format must be changed. The current implementation does not allow the
extension of event codes since this would lead to collisions with the
task identifiers. To avoid this, this patch introduces an event mask
that allows the storage of multiple events with one escape code. Now,
task identifiers are stored in the data section of the sample. The
implementation also allows the usage of custom data in a sample. As a
side effect the new code is much more readable and easier to
understand.

Signed-off-by: Robert Richter <robert.richter@amd.com>
---
 drivers/oprofile/cpu_buffer.c | 139 +++++++++++++++++++++++-------------------
 1 file changed, 75 insertions(+), 64 deletions(-)

(limited to 'drivers/oprofile/cpu_buffer.c')

diff --git a/drivers/oprofile/cpu_buffer.c b/drivers/oprofile/cpu_buffer.c
index 400f7fcffdbe..e859d23cfc57 100644
--- a/drivers/oprofile/cpu_buffer.c
+++ b/drivers/oprofile/cpu_buffer.c
@@ -212,6 +212,59 @@ unsigned long op_cpu_buffer_entries(int cpu)
 		+ ring_buffer_entries_cpu(op_ring_buffer_write, cpu);
 }
 
+static int
+op_add_code(struct oprofile_cpu_buffer *cpu_buf, unsigned long backtrace,
+	    int is_kernel, struct task_struct *task)
+{
+	struct op_entry entry;
+	struct op_sample *sample;
+	unsigned long flags;
+	int size;
+
+	flags = 0;
+
+	if (backtrace)
+		flags |= TRACE_BEGIN;
+
+	/* notice a switch from user->kernel or vice versa */
+	is_kernel = !!is_kernel;
+	if (cpu_buf->last_is_kernel != is_kernel) {
+		cpu_buf->last_is_kernel = is_kernel;
+		flags |= KERNEL_CTX_SWITCH;
+		if (is_kernel)
+			flags |= IS_KERNEL;
+	}
+
+	/* notice a task switch */
+	if (cpu_buf->last_task != task) {
+		cpu_buf->last_task = task;
+		flags |= USER_CTX_SWITCH;
+	}
+
+	if (!flags)
+		/* nothing to do */
+		return 0;
+
+	if (flags & USER_CTX_SWITCH)
+		size = 1;
+	else
+		size = 0;
+
+	sample = op_cpu_buffer_write_reserve(&entry, size);
+	if (!sample)
+		return -ENOMEM;
+
+	sample->eip = ESCAPE_CODE;
+	sample->event = flags;
+
+	if (size)
+		sample->data[0] = (unsigned long)task;
+
+	op_cpu_buffer_write_commit(&entry);
+
+	return 0;
+}
+
 static inline int
 op_add_sample(struct oprofile_cpu_buffer *cpu_buf,
 	      unsigned long pc, unsigned long event)
@@ -229,26 +282,18 @@ op_add_sample(struct oprofile_cpu_buffer *cpu_buf,
 	return op_cpu_buffer_write_commit(&entry);
 }
 
-static inline int
-add_code(struct oprofile_cpu_buffer *buffer, unsigned long value)
-{
-	return op_add_sample(buffer, ESCAPE_CODE, value);
-}
-
-/* This must be safe from any context. It's safe writing here
- * because of the head/tail separation of the writer and reader
- * of the CPU buffer.
+/*
+ * This must be safe from any context.
  *
  * is_kernel is needed because on some architectures you cannot
  * tell if you are in kernel or user space simply by looking at
  * pc. We tag this in the buffer by generating kernel enter/exit
  * events whenever is_kernel changes
  */
-static int log_sample(struct oprofile_cpu_buffer *cpu_buf, unsigned long pc,
-		      int is_kernel, unsigned long event)
+static int
+log_sample(struct oprofile_cpu_buffer *cpu_buf, unsigned long pc,
+	   unsigned long backtrace, int is_kernel, unsigned long event)
 {
-	struct task_struct *task;
-
 	cpu_buf->sample_received++;
 
 	if (pc == ESCAPE_CODE) {
@@ -256,23 +301,8 @@ static int log_sample(struct oprofile_cpu_buffer *cpu_buf, unsigned long pc,
 		return 0;
 	}
 
-	is_kernel = !!is_kernel;
-
-	task = current;
-
-	/* notice a switch from user->kernel or vice versa */
-	if (cpu_buf->last_is_kernel != is_kernel) {
-		cpu_buf->last_is_kernel = is_kernel;
-		if (add_code(cpu_buf, is_kernel))
-			goto fail;
-	}
-
-	/* notice a task switch */
-	if (cpu_buf->last_task != task) {
-		cpu_buf->last_task = task;
-		if (add_code(cpu_buf, (unsigned long)task))
-			goto fail;
-	}
+	if (op_add_code(cpu_buf, backtrace, is_kernel, current))
+		goto fail;
 
 	if (op_add_sample(cpu_buf, pc, event))
 		goto fail;
@@ -286,7 +316,6 @@ fail:
 
 static inline void oprofile_begin_trace(struct oprofile_cpu_buffer *cpu_buf)
 {
-	add_code(cpu_buf, CPU_TRACE_BEGIN);
 	cpu_buf->tracing = 1;
 }
 
@@ -300,21 +329,21 @@ __oprofile_add_ext_sample(unsigned long pc, struct pt_regs * const regs,
 			  unsigned long event, int is_kernel)
 {
 	struct oprofile_cpu_buffer *cpu_buf = &__get_cpu_var(cpu_buffer);
-
-	if (!oprofile_backtrace_depth) {
-		log_sample(cpu_buf, pc, is_kernel, event);
-		return;
-	}
-
-	oprofile_begin_trace(cpu_buf);
+	unsigned long backtrace = oprofile_backtrace_depth;
 
 	/*
 	 * if log_sample() fail we can't backtrace since we lost the
 	 * source of this event
 	 */
-	if (log_sample(cpu_buf, pc, is_kernel, event))
-		oprofile_ops.backtrace(regs, oprofile_backtrace_depth);
+	if (!log_sample(cpu_buf, pc, backtrace, is_kernel, event))
+		/* failed */
+		return;
 
+	if (!backtrace)
+		return;
+
+	oprofile_begin_trace(cpu_buf);
+	oprofile_ops.backtrace(regs, backtrace);
 	oprofile_end_trace(cpu_buf);
 }
 
@@ -339,29 +368,14 @@ void oprofile_add_ibs_sample(struct pt_regs * const regs,
 {
 	int is_kernel = !user_mode(regs);
 	struct oprofile_cpu_buffer *cpu_buf = &__get_cpu_var(cpu_buffer);
-	struct task_struct *task;
 	int fail = 0;
 
 	cpu_buf->sample_received++;
 
-	/* notice a switch from user->kernel or vice versa */
-	if (cpu_buf->last_is_kernel != is_kernel) {
-		if (add_code(cpu_buf, is_kernel))
-			goto fail;
-		cpu_buf->last_is_kernel = is_kernel;
-	}
+	/* backtraces disabled for ibs */
+	fail = fail || op_add_code(cpu_buf, 0, is_kernel, current);
 
-	/* notice a task switch */
-	if (!is_kernel) {
-		task = current;
-		if (cpu_buf->last_task != task) {
-			if (add_code(cpu_buf, (unsigned long)task))
-				goto fail;
-			cpu_buf->last_task = task;
-		}
-	}
-
-	fail = fail || add_code(cpu_buf, ibs_code);
+	fail = fail || op_add_sample(cpu_buf, ESCAPE_CODE,   ibs_code);
 	fail = fail || op_add_sample(cpu_buf, ibs_sample[0], ibs_sample[1]);
 	fail = fail || op_add_sample(cpu_buf, ibs_sample[2], ibs_sample[3]);
 	fail = fail || op_add_sample(cpu_buf, ibs_sample[4], ibs_sample[5]);
@@ -372,11 +386,8 @@ void oprofile_add_ibs_sample(struct pt_regs * const regs,
 		fail = fail || op_add_sample(cpu_buf, ibs_sample[10], ibs_sample[11]);
 	}
 
-	if (!fail)
-		return;
-
-fail:
-	cpu_buf->sample_lost_overflow++;
+	if (fail)
+		cpu_buf->sample_lost_overflow++;
 }
 
 #endif
@@ -384,7 +395,7 @@ fail:
 void oprofile_add_pc(unsigned long pc, int is_kernel, unsigned long event)
 {
 	struct oprofile_cpu_buffer *cpu_buf = &__get_cpu_var(cpu_buffer);
-	log_sample(cpu_buf, pc, is_kernel, event);
+	log_sample(cpu_buf, pc, 0, is_kernel, event);
 }
 
 void oprofile_add_trace(unsigned long pc)
-- 
cgit v1.2.2


From d9928c25a6960cf128c2078a89fe6f8e0180ff60 Mon Sep 17 00:00:00 2001
From: Robert Richter <robert.richter@amd.com>
Date: Thu, 25 Dec 2008 17:26:07 +0100
Subject: oprofile: add op_cpu_buffer_add_data()

This function can be used to attach data to a sample. It returns the
remaining free buffer size that has been reserved with
op_cpu_buffer_write_reserve().

Signed-off-by: Robert Richter <robert.richter@amd.com>
---
 drivers/oprofile/cpu_buffer.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

(limited to 'drivers/oprofile/cpu_buffer.c')

diff --git a/drivers/oprofile/cpu_buffer.c b/drivers/oprofile/cpu_buffer.c
index e859d23cfc57..1b6590746be4 100644
--- a/drivers/oprofile/cpu_buffer.c
+++ b/drivers/oprofile/cpu_buffer.c
@@ -258,7 +258,7 @@ op_add_code(struct oprofile_cpu_buffer *cpu_buf, unsigned long backtrace,
 	sample->event = flags;
 
 	if (size)
-		sample->data[0] = (unsigned long)task;
+		op_cpu_buffer_add_data(&entry, (unsigned long)task);
 
 	op_cpu_buffer_write_commit(&entry);
 
-- 
cgit v1.2.2


From 1acda878e20ea0cd3708ba66dca67d52eaafdd2b Mon Sep 17 00:00:00 2001
From: Robert Richter <robert.richter@amd.com>
Date: Mon, 5 Jan 2009 10:35:31 +0100
Subject: oprofile: use new data sample format for ibs

The new ring buffer implementation allows the storage of samples with
different size. This patch implements the usage of the new sample
format to store ibs samples in the cpu buffer. Until now, writing to
the cpu buffer could lead to incomplete sampling sequences since IBS
samples were transfered in multiple samples. Due to a full buffer,
data could be lost at any time. This can't happen any more since the
complete data is reserved in advance and then stored in a single
sample.

Signed-off-by: Robert Richter <robert.richter@amd.com>
---
 drivers/oprofile/cpu_buffer.c | 39 +++++++++++++++++++++++----------------
 1 file changed, 23 insertions(+), 16 deletions(-)

(limited to 'drivers/oprofile/cpu_buffer.c')

diff --git a/drivers/oprofile/cpu_buffer.c b/drivers/oprofile/cpu_buffer.c
index 1b6590746be4..ddba9d01f09b 100644
--- a/drivers/oprofile/cpu_buffer.c
+++ b/drivers/oprofile/cpu_buffer.c
@@ -363,31 +363,38 @@ void oprofile_add_sample(struct pt_regs * const regs, unsigned long event)
 
 #ifdef CONFIG_OPROFILE_IBS
 
-void oprofile_add_ibs_sample(struct pt_regs * const regs,
-			     unsigned int * const ibs_sample, int ibs_code)
+/*
+ * Add samples with data to the ring buffer.
+ *
+ * Use op_cpu_buffer_add_data(&entry, val) to add data and
+ * op_cpu_buffer_write_commit(&entry) to commit the sample.
+ */
+void oprofile_add_data(struct op_entry *entry, struct pt_regs * const regs,
+		       unsigned long pc, int code, int size)
 {
+	struct op_sample *sample;
 	int is_kernel = !user_mode(regs);
 	struct oprofile_cpu_buffer *cpu_buf = &__get_cpu_var(cpu_buffer);
-	int fail = 0;
 
 	cpu_buf->sample_received++;
 
-	/* backtraces disabled for ibs */
-	fail = fail || op_add_code(cpu_buf, 0, is_kernel, current);
+	/* no backtraces for samples with data */
+	if (op_add_code(cpu_buf, 0, is_kernel, current))
+		goto fail;
 
-	fail = fail || op_add_sample(cpu_buf, ESCAPE_CODE,   ibs_code);
-	fail = fail || op_add_sample(cpu_buf, ibs_sample[0], ibs_sample[1]);
-	fail = fail || op_add_sample(cpu_buf, ibs_sample[2], ibs_sample[3]);
-	fail = fail || op_add_sample(cpu_buf, ibs_sample[4], ibs_sample[5]);
+	sample = op_cpu_buffer_write_reserve(entry, size + 2);
+	if (!sample)
+		goto fail;
+	sample->eip = ESCAPE_CODE;
+	sample->event = 0;		/* no flags */
 
-	if (ibs_code == IBS_OP_BEGIN) {
-		fail = fail || op_add_sample(cpu_buf, ibs_sample[6], ibs_sample[7]);
-		fail = fail || op_add_sample(cpu_buf, ibs_sample[8], ibs_sample[9]);
-		fail = fail || op_add_sample(cpu_buf, ibs_sample[10], ibs_sample[11]);
-	}
+	op_cpu_buffer_add_data(entry, code);
+	op_cpu_buffer_add_data(entry, pc);
+
+	return;
 
-	if (fail)
-		cpu_buf->sample_lost_overflow++;
+fail:
+	cpu_buf->sample_lost_overflow++;
 }
 
 #endif
-- 
cgit v1.2.2


From ebf8d974e298018f0b4ee02b1b097bf5500d3d27 Mon Sep 17 00:00:00 2001
From: Robert Richter <robert.richter@amd.com>
Date: Wed, 7 Jan 2009 00:20:57 +0100
Subject: oprofile: remove #ifdef CONFIG_OPROFILE_IBS in non-ibs code

The ifdefs can be removed since the code is no longer ibs specific and
can be used for other purposes as well. IBS specific code is only in
op_model_amd.c.

Signed-off-by: Robert Richter <robert.richter@amd.com>
---
 drivers/oprofile/cpu_buffer.c | 4 ----
 1 file changed, 4 deletions(-)

(limited to 'drivers/oprofile/cpu_buffer.c')

diff --git a/drivers/oprofile/cpu_buffer.c b/drivers/oprofile/cpu_buffer.c
index ddba9d01f09b..b846af632c81 100644
--- a/drivers/oprofile/cpu_buffer.c
+++ b/drivers/oprofile/cpu_buffer.c
@@ -361,8 +361,6 @@ void oprofile_add_sample(struct pt_regs * const regs, unsigned long event)
 	__oprofile_add_ext_sample(pc, regs, event, is_kernel);
 }
 
-#ifdef CONFIG_OPROFILE_IBS
-
 /*
  * Add samples with data to the ring buffer.
  *
@@ -397,8 +395,6 @@ fail:
 	cpu_buf->sample_lost_overflow++;
 }
 
-#endif
-
 void oprofile_add_pc(unsigned long pc, int is_kernel, unsigned long event)
 {
 	struct oprofile_cpu_buffer *cpu_buf = &__get_cpu_var(cpu_buffer);
-- 
cgit v1.2.2


From 14f0ca8eaea42a5b5a69cfcb699665dd2618db5f Mon Sep 17 00:00:00 2001
From: Robert Richter <robert.richter@amd.com>
Date: Wed, 7 Jan 2009 21:50:22 +0100
Subject: oprofile: make new cpu buffer functions part of the api

This patch creates the new functions

 oprofile_write_reserve()
 oprofile_add_data()
 oprofile_write_commit()

and makes them part of the oprofile api.

Signed-off-by: Robert Richter <robert.richter@amd.com>
---
 drivers/oprofile/cpu_buffer.c | 17 ++++++++++++++---
 1 file changed, 14 insertions(+), 3 deletions(-)

(limited to 'drivers/oprofile/cpu_buffer.c')

diff --git a/drivers/oprofile/cpu_buffer.c b/drivers/oprofile/cpu_buffer.c
index b846af632c81..2e03b6d796d3 100644
--- a/drivers/oprofile/cpu_buffer.c
+++ b/drivers/oprofile/cpu_buffer.c
@@ -364,10 +364,11 @@ void oprofile_add_sample(struct pt_regs * const regs, unsigned long event)
 /*
  * Add samples with data to the ring buffer.
  *
- * Use op_cpu_buffer_add_data(&entry, val) to add data and
- * op_cpu_buffer_write_commit(&entry) to commit the sample.
+ * Use oprofile_add_data(&entry, val) to add data and
+ * oprofile_write_commit(&entry) to commit the sample.
  */
-void oprofile_add_data(struct op_entry *entry, struct pt_regs * const regs,
+void
+oprofile_write_reserve(struct op_entry *entry, struct pt_regs * const regs,
 		       unsigned long pc, int code, int size)
 {
 	struct op_sample *sample;
@@ -395,6 +396,16 @@ fail:
 	cpu_buf->sample_lost_overflow++;
 }
 
+int oprofile_add_data(struct op_entry *entry, unsigned long val)
+{
+	return op_cpu_buffer_add_data(entry, val);
+}
+
+int oprofile_write_commit(struct op_entry *entry)
+{
+	return op_cpu_buffer_write_commit(entry);
+}
+
 void oprofile_add_pc(unsigned long pc, int is_kernel, unsigned long event)
 {
 	struct oprofile_cpu_buffer *cpu_buf = &__get_cpu_var(cpu_buffer);
-- 
cgit v1.2.2