Skip to contentSkip to conversation

Generalise ARM perf-events backend for oprofile

PATCH Superseded Matt Fleming 5 participants 43 messages Oct 4, 2010 – Oct 6, 2010

Subsystems (5)

v4 of 5 Superseded Latest is v5 →

lore
Share & tools
lore
#0 lore raw

v4 cover Generalise ARM perf-events backend for oprofile

The perf-events backend for OProfile that Will Deacon wrote in 8c1fc96f6fd1f361428ba805103af0d0eee65179 ("ARM: 6072/1: oprofile: use perf-events framework as backend") is of use to more architectures than just ARM. Move the code into drivers/oprofile/ so that SH can use it instead of the nearly identical copy of its OProfile code.

The benefit of the backend is that it becomes necessary to only maintain one copy of the PMU accessor functions for each architecture, with bug fixes and new features benefiting both OProfile and perf.

Note that I haven't been able to test these patches on an ARM board to see if I've caused any regressions. If anyone else could do that I'd appreciate it. Though, I have been able to compile this version of the series.

This patch series is based on,

git://git.kernel.org/pub/scm/linux/kernel/git/rric/oprofile.git core

These patches can also be found at,

git://git.kernel.org/pub/scm/linux/kernel/git/mfleming/sh-2.6.git perf-oprofile

Robert, I haven't rebased this work on your oprofile updates for 2.6.37. If you need me to do that, just shout.

Changes from v3:
- New perf_pmu_name() function
- Sprinkled an EXPORT_SYMBOL_GPL()
- Don't repeatedly call perf_num_counters()
- Change name of oprofile_driver to "oprofile-perf"
- Hard overwrite the oprofile_operations in oprofile_perf_init()
- Dropped the patch that made oprofile depend on HW_PERF_EVENTS for ARM

Changes from v2:
- Rebased against Robert's oprofile core branch
- Moved even more of the ARM code into the generic oprofile code
- Broke the patches up into more logical steps

Changes from v1:
- Prefix the new functons with "oprofile_" instead of "op_".
- Fix ARM compilation errors
- Move all the oprofile-perf logic into oprofile_perf.c
- Include cleanup patch from Will

Matt Fleming (7):
perf: Add helper function to return number of counters
perf: New helper function for pmu name
ARM: oprofile: Rename op_arm to oprofile_perf
ARM: oprofile: Move non-ARM code into separate init/exit
oprofile: Abstract the perf-events backend
sh: oprofile: Use perf-events oprofile backend
sh: Annotate oprofile_arch_exit() with __exit marker

 arch/arm/kernel/perf_event.c     |   29 ++++
 arch/arm/oprofile/Makefile       |    4 +
 arch/arm/oprofile/common.c       |  327 +-------------------------------------
 arch/sh/Kconfig                  |   13 ++
 arch/sh/kernel/perf_event.c      |   23 +++
 arch/sh/oprofile/Makefile        |    4 +
 arch/sh/oprofile/common.c        |  109 ++------------
 arch/sh/oprofile/op_impl.h       |   33 ----
 drivers/oprofile/oprofile_perf.c |  326 +++++++++++++++++++++++++++++++++++++
 include/linux/oprofile.h         |    6 +
 include/linux/perf_event.h       |    2 +
 kernel/perf_event.c              |    5 +
 12 files changed, 425 insertions(+), 456 deletions(-)
 delete mode 100644 arch/sh/oprofile/op_impl.h
 create mode 100644 drivers/oprofile/oprofile_perf.c
#1 lore raw

PATCH v4 1/7 perf: Add helper function to return number of counters

The number of counters for the registered pmu is needed in a few places so provide a helper function that returns this number.

---
 arch/arm/kernel/perf_event.c |    6 ++++++
 arch/arm/oprofile/common.c   |   31 ++++++++++++++++++-------------
 arch/sh/kernel/perf_event.c  |    9 +++++++++
 include/linux/perf_event.h   |    1 +
 4 files changed, 34 insertions(+), 13 deletions(-)
arch/arm/kernel/perf_event.c +6 -0
index ecbb028..ef3bc33 100644
@@ -123,6 +123,12 @@ armpmu_get_max_events(void)
 }
 EXPORT_SYMBOL_GPL(armpmu_get_max_events);
 
+int perf_num_counters(void)
+{
+	return armpmu_get_max_events();
+}
+EXPORT_SYMBOL_GPL(perf_num_counters);
+
 #define HW_OP_UNSUPPORTED		0xFFFF
 
 #define C(_x) \
arch/arm/oprofile/common.c +18 -13
index d660cb8..23f18a0 100644
@@ -43,7 +43,7 @@ static DEFINE_MUTEX(op_arm_mutex);
 
 static struct op_counter_config *counter_config;
 static struct perf_event **perf_events[nr_cpumask_bits];
-static int perf_num_counters;
+static int perf_nr_counters;
 
 /*
  * Overflow callback for oprofile.
@@ -54,11 +54,11 @@ static void op_overflow_handler(struct perf_event *event, int unused,
 	int id;
 	u32 cpu = smp_processor_id();
 
-	for (id = 0; id < perf_num_counters; ++id)
+	for (id = 0; id < perf_nr_counters; ++id)
 		if (perf_events[cpu][id] == event)
 			break;
 
-	if (id != perf_num_counters)
+	if (id != perf_nr_counters)
 		oprofile_add_sample(regs, id);
 	else
 		pr_warning("oprofile: ignoring spurious overflow "
@@ -76,7 +76,7 @@ static void op_perf_setup(void)
 	u32 size = sizeof(struct perf_event_attr);
 	struct perf_event_attr *attr;
 
-	for (i = 0; i < perf_num_counters; ++i) {
+	for (i = 0; i < perf_nr_counters; ++i) {
 		attr = &counter_config[i].attr;
 		memset(attr, 0, size);
 		attr->type		= PERF_TYPE_RAW;
@@ -131,7 +131,7 @@ static int op_perf_start(void)
 	int cpu, event, ret = 0;
 
 	for_each_online_cpu(cpu) {
-		for (event = 0; event < perf_num_counters; ++event) {
+		for (event = 0; event < perf_nr_counters; ++event) {
 			ret = op_create_counter(cpu, event);
 			if (ret)
 				goto out;
@@ -150,7 +150,7 @@ static void op_perf_stop(void)
 	int cpu, event;
 
 	for_each_online_cpu(cpu)
-		for (event = 0; event < perf_num_counters; ++event)
+		for (event = 0; event < perf_nr_counters; ++event)
 			op_destroy_counter(cpu, event);
 }
 
@@ -179,7 +179,7 @@ static int op_arm_create_files(struct super_block *sb, struct dentry *root)
 {
 	unsigned int i;
 
-	for (i = 0; i < perf_num_counters; i++) {
+	for (i = 0; i < perf_nr_counters; i++) {
 		struct dentry *dir;
 		char buf[4];
 
@@ -353,14 +353,19 @@ int __init oprofile_arch_init(struct oprofile_operations *ops)
 
 	memset(&perf_events, 0, sizeof(perf_events));
 
-	perf_num_counters = armpmu_get_max_events();
+	perf_nr_counters = perf_num_counters();
+	if (perf_nr_counters <= 0) {
+		pr_info("oprofile: no performance counters\n");
+		ret = -ENODEV;
+		goto out;
+	}
 
-	counter_config = kcalloc(perf_num_counters,
+	counter_config = kcalloc(perf_nr_counters,
 			sizeof(struct op_counter_config), GFP_KERNEL);
 
 	if (!counter_config) {
 		pr_info("oprofile: failed to allocate %d "
-				"counters\n", perf_num_counters);
+				"counters\n", perf_nr_counters);
 		ret = -ENOMEM;
 		goto out;
 	}
@@ -370,11 +375,11 @@ int __init oprofile_arch_init(struct oprofile_operations *ops)
 		goto out;
 
 	for_each_possible_cpu(cpu) {
-		perf_events[cpu] = kcalloc(perf_num_counters,
+		perf_events[cpu] = kcalloc(perf_nr_counters,
 				sizeof(struct perf_event *), GFP_KERNEL);
 		if (!perf_events[cpu]) {
 			pr_info("oprofile: failed to allocate %d perf events "
-					"for cpu %d\n", perf_num_counters, cpu);
+					"for cpu %d\n", perf_nr_counters, cpu);
 			ret = -ENOMEM;
 			goto out;
 		}
@@ -409,7 +414,7 @@ void __exit oprofile_arch_exit(void)
 	struct perf_event *event;
 
 	for_each_possible_cpu(cpu) {
-		for (id = 0; id < perf_num_counters; ++id) {
+		for (id = 0; id < perf_nr_counters; ++id) {
 			event = perf_events[cpu][id];
 			if (event)
 				perf_event_release_kernel(event);
arch/sh/kernel/perf_event.c +9 -0
index 7a3dc35..2cb9ad5 100644
@@ -59,6 +59,15 @@ static inline int sh_pmu_initialized(void)
 	return !!sh_pmu;
 }
 
+int perf_num_counters(void)
+{
+	if (!sh_pmu)
+		return 0;
+
+	return sh_pmu->num_events;
+}
+EXPORT_SYMBOL_GPL(perf_num_counters);
+
 /*
  * Release the PMU if this is the last perf_event.
  */
include/linux/perf_event.h +1 -0
index 716f99b..1a02192 100644
@@ -849,6 +849,7 @@ extern int perf_max_events;
 
 extern const struct pmu *hw_perf_event_init(struct perf_event *event);
 
+extern int perf_num_counters(void);
 extern void perf_event_task_sched_in(struct task_struct *task);
 extern void perf_event_task_sched_out(struct task_struct *task, struct task_struct *next);
 extern void perf_event_task_tick(struct task_struct *task);
Signature · 1 line
1.7.1
#3 lore raw

Re: [PATCH 1/7] perf: Add helper function to return number of counters

In reply to Matt Fleming
Quoted 17 lines · Matt Fleming

The number of counters for the registered pmu is needed in a few places so provide a helper function that returns this number.

---
 arch/arm/kernel/perf_event.c |    6 ++++++
 arch/arm/oprofile/common.c   |   31 ++++++++++++++++++-------------
 arch/sh/kernel/perf_event.c  |    9 +++++++++
 include/linux/perf_event.h   |    1 +
 4 files changed, 34 insertions(+), 13 deletions(-)
arch/arm/kernel/perf_event.c +0 -0
index ecbb028..ef3bc33 100644

arch/arm/kernel/perf_event.c armpmu_get_max_events(void) · line 130

Quoted 8 lines
arch/arm/kernel/perf_event.c +5 -0
@@ -123,6 +123,12 @@ armpmu_get_max_events(void)
 }
 EXPORT_SYMBOL_GPL(armpmu_get_max_events);
 
+int perf_num_counters(void)
+{
+	return armpmu_get_max_events();
+}
+EXPORT_SYMBOL_GPL(perf_num_counters);

Peter,

can you please ack this interface?

Thanks,

-Robert

Quoted 149 lines
arch/arm/kernel/perf_event.c +1 -0
+
 #define HW_OP_UNSUPPORTED		0xFFFF
 
 #define C(_x) \
arch/arm/oprofile/common.c +18 -13
index d660cb8..23f18a0 100644
@@ -43,7 +43,7 @@ static DEFINE_MUTEX(op_arm_mutex);
 
 static struct op_counter_config *counter_config;
 static struct perf_event **perf_events[nr_cpumask_bits];
-static int perf_num_counters;
+static int perf_nr_counters;
 
 /*
  * Overflow callback for oprofile.
@@ -54,11 +54,11 @@ static void op_overflow_handler(struct perf_event *event, int unused,
 	int id;
 	u32 cpu = smp_processor_id();
 
-	for (id = 0; id < perf_num_counters; ++id)
+	for (id = 0; id < perf_nr_counters; ++id)
 		if (perf_events[cpu][id] == event)
 			break;
 
-	if (id != perf_num_counters)
+	if (id != perf_nr_counters)
 		oprofile_add_sample(regs, id);
 	else
 		pr_warning("oprofile: ignoring spurious overflow "
@@ -76,7 +76,7 @@ static void op_perf_setup(void)
 	u32 size = sizeof(struct perf_event_attr);
 	struct perf_event_attr *attr;
 
-	for (i = 0; i < perf_num_counters; ++i) {
+	for (i = 0; i < perf_nr_counters; ++i) {
 		attr = &counter_config[i].attr;
 		memset(attr, 0, size);
 		attr->type		= PERF_TYPE_RAW;
@@ -131,7 +131,7 @@ static int op_perf_start(void)
 	int cpu, event, ret = 0;
 
 	for_each_online_cpu(cpu) {
-		for (event = 0; event < perf_num_counters; ++event) {
+		for (event = 0; event < perf_nr_counters; ++event) {
 			ret = op_create_counter(cpu, event);
 			if (ret)
 				goto out;
@@ -150,7 +150,7 @@ static void op_perf_stop(void)
 	int cpu, event;
 
 	for_each_online_cpu(cpu)
-		for (event = 0; event < perf_num_counters; ++event)
+		for (event = 0; event < perf_nr_counters; ++event)
 			op_destroy_counter(cpu, event);
 }
 
@@ -179,7 +179,7 @@ static int op_arm_create_files(struct super_block *sb, struct dentry *root)
 {
 	unsigned int i;
 
-	for (i = 0; i < perf_num_counters; i++) {
+	for (i = 0; i < perf_nr_counters; i++) {
 		struct dentry *dir;
 		char buf[4];
 
@@ -353,14 +353,19 @@ int __init oprofile_arch_init(struct oprofile_operations *ops)
 
 	memset(&perf_events, 0, sizeof(perf_events));
 
-	perf_num_counters = armpmu_get_max_events();
+	perf_nr_counters = perf_num_counters();
+	if (perf_nr_counters <= 0) {
+		pr_info("oprofile: no performance counters\n");
+		ret = -ENODEV;
+		goto out;
+	}
 
-	counter_config = kcalloc(perf_num_counters,
+	counter_config = kcalloc(perf_nr_counters,
 			sizeof(struct op_counter_config), GFP_KERNEL);
 
 	if (!counter_config) {
 		pr_info("oprofile: failed to allocate %d "
-				"counters\n", perf_num_counters);
+				"counters\n", perf_nr_counters);
 		ret = -ENOMEM;
 		goto out;
 	}
@@ -370,11 +375,11 @@ int __init oprofile_arch_init(struct oprofile_operations *ops)
 		goto out;
 
 	for_each_possible_cpu(cpu) {
-		perf_events[cpu] = kcalloc(perf_num_counters,
+		perf_events[cpu] = kcalloc(perf_nr_counters,
 				sizeof(struct perf_event *), GFP_KERNEL);
 		if (!perf_events[cpu]) {
 			pr_info("oprofile: failed to allocate %d perf events "
-					"for cpu %d\n", perf_num_counters, cpu);
+					"for cpu %d\n", perf_nr_counters, cpu);
 			ret = -ENOMEM;
 			goto out;
 		}
@@ -409,7 +414,7 @@ void __exit oprofile_arch_exit(void)
 	struct perf_event *event;
 
 	for_each_possible_cpu(cpu) {
-		for (id = 0; id < perf_num_counters; ++id) {
+		for (id = 0; id < perf_nr_counters; ++id) {
 			event = perf_events[cpu][id];
 			if (event)
 				perf_event_release_kernel(event);
arch/sh/kernel/perf_event.c +9 -0
index 7a3dc35..2cb9ad5 100644
@@ -59,6 +59,15 @@ static inline int sh_pmu_initialized(void)
 	return !!sh_pmu;
 }
 
+int perf_num_counters(void)
+{
+	if (!sh_pmu)
+		return 0;
+
+	return sh_pmu->num_events;
+}
+EXPORT_SYMBOL_GPL(perf_num_counters);
+
 /*
  * Release the PMU if this is the last perf_event.
  */
include/linux/perf_event.h +1 -0
index 716f99b..1a02192 100644
@@ -849,6 +849,7 @@ extern int perf_max_events;
 
 extern const struct pmu *hw_perf_event_init(struct perf_event *event);
 
+extern int perf_num_counters(void);
 extern void perf_event_task_sched_in(struct task_struct *task);
 extern void perf_event_task_sched_out(struct task_struct *task, struct task_struct *next);
 extern void perf_event_task_tick(struct task_struct *task);
Signature · 1 line
1.7.1
Signature · 2 lines
Advanced Micro Devices, Inc.
Operating System Research Center
#4 lore raw

Re: [PATCH 1/7] perf: Add helper function to return number of counters

Quoted 19 lines · Robert Richter
Quoted 18 lines · Matt Fleming

The number of counters for the registered pmu is needed in a few places so provide a helper function that returns this number.

---
 arch/arm/kernel/perf_event.c |    6 ++++++
 arch/arm/oprofile/common.c   |   31 ++++++++++++++++++-------------
 arch/sh/kernel/perf_event.c  |    9 +++++++++
 include/linux/perf_event.h   |    1 +
 4 files changed, 34 insertions(+), 13 deletions(-)
arch/arm/oprofile/common.c +0 -0
index d660cb8..23f18a0 100644
@@ -43,7 +43,7 @@ static DEFINE_MUTEX(op_arm_mutex);
 
Quoted 4 lines

arch/arm/oprofile/common.c static DEFINE_MUTEX(op_arm_mutex); · line 46

Quoted 4 lines
arch/arm/oprofile/common.c +1 -1
 static struct op_counter_config *counter_config;
 static struct perf_event **perf_events[nr_cpumask_bits];
-static int perf_num_counters;
+static int perf_nr_counters;

Matt,

as we will have one more version with this patch set, please use num_counters instead of perf_nr_counters. The same name is used also for x86.

Thanks,

-Robert

Quoted 3 lines
Quoted 3 lines
arch/arm/oprofile/common.c +0 -0
 /*
  * Overflow callback for oprofile.
Signature · 2 lines
Advanced Micro Devices, Inc.
Operating System Research Center
maintainer Acked-by
#5 lore raw

Re: [PATCH 1/7] perf: Add helper function to return number of counters

In reply to Robert Richter
Quoted 9 lines · Robert Richter

arch/arm/kernel/perf_event.c armpmu_get_max_events(void) · line 130

Quoted 5 lines
arch/arm/kernel/perf_event.c +5 -0
+int perf_num_counters(void)
+{
+     return armpmu_get_max_events();
+}
+EXPORT_SYMBOL_GPL(perf_num_counters);

Peter,

can you please ack this interface?

It does seem like the best way forward for this, so

#6 lore raw

PATCH v4 2/7 perf: New helper function for pmu name

In reply to Matt Fleming

Introduce perf_pmu_name() helper function that returns the name of the pmu. This gives us a generic way to get the name of a pmu regardless of how an architecture identifies it internally, e.g. ARM uses an id whereas SH currently uses a string.

---
 arch/arm/kernel/perf_event.c |   23 +++++++++++++++++++++++
 arch/arm/oprofile/common.c   |   22 +---------------------
 arch/sh/kernel/perf_event.c  |   14 ++++++++++++++
 include/linux/perf_event.h   |    1 +
 kernel/perf_event.c          |    5 +++++
 5 files changed, 44 insertions(+), 21 deletions(-)
arch/arm/kernel/perf_event.c +23 -0
index ef3bc33..3bff24d 100644
@@ -123,6 +123,29 @@ armpmu_get_max_events(void)
 }
 EXPORT_SYMBOL_GPL(armpmu_get_max_events);
 
+const char *perf_pmu_name(void)
+{
+	enum arm_perf_pmu_ids id = armpmu_get_pmu_id();
+
+	switch (id) {
+	case ARM_PERF_PMU_ID_XSCALE1:
+		return "arm/xscale1";
+	case ARM_PERF_PMU_ID_XSCALE2:
+		return "arm/xscale2";
+	case ARM_PERF_PMU_ID_V6:
+		return "arm/armv6";
+	case ARM_PERF_PMU_ID_V6MP:
+		return "arm/mpcore";
+	case ARM_PERF_PMU_ID_CA8:
+		return "arm/armv7";
+	case ARM_PERF_PMU_ID_CA9:
+		return "arm/armv7-ca9";
+	default:
+		return NULL;
+	}
+}
+EXPORT_SYMBOL_GPL(perf_pmu_name);
+
 int perf_num_counters(void)
 {
 	return armpmu_get_max_events();
arch/arm/oprofile/common.c +1 -21
index 23f18a0..cb224ee 100644
@@ -155,26 +155,6 @@ static void op_perf_stop(void)
 }
 
 
-static char *op_name_from_perf_id(enum arm_perf_pmu_ids id)
-{
-	switch (id) {
-	case ARM_PERF_PMU_ID_XSCALE1:
-		return "arm/xscale1";
-	case ARM_PERF_PMU_ID_XSCALE2:
-		return "arm/xscale2";
-	case ARM_PERF_PMU_ID_V6:
-		return "arm/armv6";
-	case ARM_PERF_PMU_ID_V6MP:
-		return "arm/mpcore";
-	case ARM_PERF_PMU_ID_CA8:
-		return "arm/armv7";
-	case ARM_PERF_PMU_ID_CA9:
-		return "arm/armv7-ca9";
-	default:
-		return NULL;
-	}
-}
-
 static int op_arm_create_files(struct super_block *sb, struct dentry *root)
 {
 	unsigned int i;
@@ -391,7 +371,7 @@ int __init oprofile_arch_init(struct oprofile_operations *ops)
 	ops->start		= op_arm_start;
 	ops->stop		= op_arm_stop;
 	ops->shutdown		= op_arm_stop;
-	ops->cpu_type		= op_name_from_perf_id(armpmu_get_pmu_id());
+	ops->cpu_type		= perf_pmu_name();
 
 	if (!ops->cpu_type)
 		ret = -ENODEV;
arch/sh/kernel/perf_event.c +14 -0
index 2cb9ad5..e065a1d 100644
@@ -59,6 +59,20 @@ static inline int sh_pmu_initialized(void)
 	return !!sh_pmu;
 }
 
+const char *perf_pmu_name(void)
+{
+	if (!sh_pmu)
+		return NULL;
+
+	if (!strcmp(sh_pmu->name, "SH7750"))
+		return "sh/sh7750";
+	if (!strcmp(sh_pmu->name, "SH-4A"))
+		return "sh/sh4a";
+
+	return NULL;
+}
+EXPORT_SYMBOL_GPL(perf_pmu_name);
+
 int perf_num_counters(void)
 {
 	if (!sh_pmu)
include/linux/perf_event.h +1 -0
index 1a02192..33f08da 100644
@@ -850,6 +850,7 @@ extern int perf_max_events;
 extern const struct pmu *hw_perf_event_init(struct perf_event *event);
 
 extern int perf_num_counters(void);
+extern const char *perf_pmu_name(void);
 extern void perf_event_task_sched_in(struct task_struct *task);
 extern void perf_event_task_sched_out(struct task_struct *task, struct task_struct *next);
 extern void perf_event_task_tick(struct task_struct *task);
kernel/perf_event.c +5 -0
index db5b560..fc51268 100644
@@ -85,6 +85,11 @@ void __weak hw_perf_enable(void)		{ barrier(); }
 
 void __weak perf_event_print_debug(void)	{ }
 
+extern __weak const char *perf_pmu_name(void)
+{
+	return "pmu";
+}
+
 static DEFINE_PER_CPU(int, perf_disable_count);
 
 void perf_disable(void)
Signature · 1 line
1.7.1
Acked-by on v4 2/7
#7 lore raw

Re: [PATCH 2/7] perf: New helper function for pmu name

Quoted 6 lines · Matt Fleming

Introduce perf_pmu_name() helper function that returns the name of the pmu. This gives us a generic way to get the name of a pmu regardless of how an architecture identifies it internally, e.g. ARM uses an id whereas SH currently uses a string.

This looks fine too. Later we can see about switching the SH case over to CPU subtype IDs, but it doesn't matter for now.

#8 lore raw

Re: [PATCH 2/7] perf: New helper function for pmu name

In reply to Matt Fleming
Quoted 65 lines · Matt Fleming

Introduce perf_pmu_name() helper function that returns the name of the pmu. This gives us a generic way to get the name of a pmu regardless of how an architecture identifies it internally, e.g. ARM uses an id whereas SH currently uses a string.

---
 arch/arm/kernel/perf_event.c |   23 +++++++++++++++++++++++
 arch/arm/oprofile/common.c   |   22 +---------------------
 arch/sh/kernel/perf_event.c  |   14 ++++++++++++++
 include/linux/perf_event.h   |    1 +
 kernel/perf_event.c          |    5 +++++
 5 files changed, 44 insertions(+), 21 deletions(-)
arch/arm/kernel/perf_event.c +23 -0
index ef3bc33..3bff24d 100644
@@ -123,6 +123,29 @@ armpmu_get_max_events(void)
 }
 EXPORT_SYMBOL_GPL(armpmu_get_max_events);
 
+const char *perf_pmu_name(void)
+{
+	enum arm_perf_pmu_ids id = armpmu_get_pmu_id();
+
+	switch (id) {
+	case ARM_PERF_PMU_ID_XSCALE1:
+		return "arm/xscale1";
+	case ARM_PERF_PMU_ID_XSCALE2:
+		return "arm/xscale2";
+	case ARM_PERF_PMU_ID_V6:
+		return "arm/armv6";
+	case ARM_PERF_PMU_ID_V6MP:
+		return "arm/mpcore";
+	case ARM_PERF_PMU_ID_CA8:
+		return "arm/armv7";
+	case ARM_PERF_PMU_ID_CA9:
+		return "arm/armv7-ca9";
+	default:
+		return NULL;
+	}
+}
+EXPORT_SYMBOL_GPL(perf_pmu_name);
+
 int perf_num_counters(void)
 {
 	return armpmu_get_max_events();
arch/arm/oprofile/common.c +0 -9
index 23f18a0..cb224ee 100644
@@ -155,26 +155,6 @@ static void op_perf_stop(void)
 }
 
 
-static char *op_name_from_perf_id(enum arm_perf_pmu_ids id)
-{
-	switch (id) {
-	case ARM_PERF_PMU_ID_XSCALE1:
-		return "arm/xscale1";
-	case ARM_PERF_PMU_ID_XSCALE2:
-		return "arm/xscale2";
-	case ARM_PERF_PMU_ID_V6:
-		return "arm/armv6";

arch/arm/oprofile/common.c static void op_perf_stop(void) · line 176

Quoted 10 lines
arch/arm/oprofile/common.c +0 -10
-	case ARM_PERF_PMU_ID_V6MP:
-		return "arm/mpcore";
-	case ARM_PERF_PMU_ID_CA8:
-		return "arm/armv7";
-	case ARM_PERF_PMU_ID_CA9:
-		return "arm/armv7-ca9";
-	default:
-		return NULL;
-	}
-}

I don't like moving this oprofile names out to perf. These strings don't have much to do with the pmu names and are hints for the oprofile daemon to detect the pmu type with no other use. Of course this is more human readable than an enum or so, but it is of no worth other than oprofile. So it should be kept inside oprofile.

Quoted 24 lines
arch/arm/oprofile/common.c +1 -2
-
 static int op_arm_create_files(struct super_block *sb, struct dentry *root)
 {
 	unsigned int i;
@@ -391,7 +371,7 @@ int __init oprofile_arch_init(struct oprofile_operations *ops)
 	ops->start		= op_arm_start;
 	ops->stop		= op_arm_stop;
 	ops->shutdown		= op_arm_stop;
-	ops->cpu_type		= op_name_from_perf_id(armpmu_get_pmu_id());
+	ops->cpu_type		= perf_pmu_name();
 
 	if (!ops->cpu_type)
 		ret = -ENODEV;
arch/sh/kernel/perf_event.c +3 -0
index 2cb9ad5..e065a1d 100644
@@ -59,6 +59,20 @@ static inline int sh_pmu_initialized(void)
 	return !!sh_pmu;
 }
 
+const char *perf_pmu_name(void)
+{
+	if (!sh_pmu)

arch/sh/kernel/perf_event.c static inline int sh_pmu_initialized(void) · line 74

Quoted 10 lines
arch/sh/kernel/perf_event.c +10 -0
+		return NULL;
+
+	if (!strcmp(sh_pmu->name, "SH7750"))
+		return "sh/sh7750";
+	if (!strcmp(sh_pmu->name, "SH-4A"))
+		return "sh/sh4a";
+
+	return NULL;
+}
+EXPORT_SYMBOL_GPL(perf_pmu_name);

I rather want use here the solution we discussed earlier, simply including <asm/perf_event.h> and then access sh_pmu->name directly from oprofile.

-Robert

Quoted 36 lines
arch/sh/kernel/perf_event.c +1 -0
+
 int perf_num_counters(void)
 {
 	if (!sh_pmu)
include/linux/perf_event.h +1 -0
index 1a02192..33f08da 100644
@@ -850,6 +850,7 @@ extern int perf_max_events;
 extern const struct pmu *hw_perf_event_init(struct perf_event *event);
 
 extern int perf_num_counters(void);
+extern const char *perf_pmu_name(void);
 extern void perf_event_task_sched_in(struct task_struct *task);
 extern void perf_event_task_sched_out(struct task_struct *task, struct task_struct *next);
 extern void perf_event_task_tick(struct task_struct *task);
kernel/perf_event.c +5 -0
index db5b560..fc51268 100644
@@ -85,6 +85,11 @@ void __weak hw_perf_enable(void)		{ barrier(); }
 
 void __weak perf_event_print_debug(void)	{ }
 
+extern __weak const char *perf_pmu_name(void)
+{
+	return "pmu";
+}
+
 static DEFINE_PER_CPU(int, perf_disable_count);
 
 void perf_disable(void)
Signature · 1 line
1.7.1
Signature · 2 lines
Advanced Micro Devices, Inc.
Operating System Research Center
#9 lore raw

Re: [PATCH 2/7] perf: New helper function for pmu name

Quoted 10 lines · Robert Richter
Quoted 4 lines · Matt Fleming

Introduce perf_pmu_name() helper function that returns the name of the pmu. This gives us a generic way to get the name of a pmu regardless of how an architecture identifies it internally, e.g. ARM uses an id whereas SH currently uses a string.

I rather want use here the solution we discussed earlier, simply including <asm/perf_event.h> and then access sh_pmu->name directly from oprofile.

No. Exposing sh_pmu generically is unacceptable. This is already centrally managed through perf, and if oprofile needs any additional information then it needs to get that through the perf layer. We already have the situation that effectively every architecture with perf support implements a name string already, so making this part of the perf API hardly seems like that big of a stretch. If the perf people are violently opposed to this, then of course we can look at alternatives.

sh_pmu is created for use solely by the perf events code, we are not going to have oprofile poking around in data structures it has no place knowing anything about when 99% of everything else it is doing is already abstracted cleanly through the perf events interfaces. Likewise for special oprofile-specific APIs.

#10 lore raw

Re: [PATCH 2/7] perf: New helper function for pmu name

(cc'ing Peter)

Quoted 7 lines · Paul Mundt
Quoted 6 lines · Robert Richter
Quoted 4 lines · Matt Fleming

Introduce perf_pmu_name() helper function that returns the name of the pmu. This gives us a generic way to get the name of a pmu regardless of how an architecture identifies it internally, e.g. ARM uses an id whereas SH currently uses a string.

Quoted 5 lines
Quoted 4 lines

I rather want use here the solution we discussed earlier, simply including <asm/perf_event.h> and then access sh_pmu->name directly from oprofile.

No. Exposing sh_pmu generically is unacceptable. This is already

I don't want to be the perf_pmu_name() discussion a show stopper for this patch set. We don't have a generic perf interface available now that allows us to detect the pmu type for sh cpus. I also don't see a name string as a final solution for pmu detection. So it will be more than just adding perf_pmu_name().

Until then, as we currently would only need the name for SH, I don't see something wrong here to include architectural interfaces from <asm/perf_event.h> directly by architectural code until a general solution is available. If you run a 'git grep include..asm/perf' this is common for other architectures.

Quoted 6 lines

centrally managed through perf, and if oprofile needs any additional information then it needs to get that through the perf layer. We already have the situation that effectively every architecture with perf support implements a name string already, so making this part of the perf API hardly seems like that big of a stretch. If the perf people are violently opposed to this, then of course we can look at alternatives.

I am not against adding the pmu name to the perf API. But the oprofile cpu_type strings are oprofile centric esp. for the userland. So these strings will remain part of oprofile. Also I don't think we want to polute the perf pmu names with it.

Quoted 5 lines

sh_pmu is created for use solely by the perf events code, we are not going to have oprofile poking around in data structures it has no place knowing anything about when 99% of everything else it is doing is already abstracted cleanly through the perf events interfaces. Likewise for special oprofile-specific APIs.

So, I am also fine with implementing a generic perf_pmu_name() for sh and then derive the oprofile cpu_type string from it in the oprofile code.

-Robert

Signature · 2 lines
Advanced Micro Devices, Inc.
Operating System Research Center
#11 lore raw

Re: [PATCH 2/7] perf: New helper function for pmu name

Quoted 6 lines · Robert Richter

I don't want to be the perf_pmu_name() discussion a show stopper for this patch set. We don't have a generic perf interface available now that allows us to detect the pmu type for sh cpus. I also don't see a name string as a final solution for pmu detection. So it will be more than just adding perf_pmu_name().

Whether you see it as a show stopper or not is not the issue, I will not expose sh_pmu generically. Nor do I imagine that the x86 people or any other architecture wishing to tie in to the oprofile wrapper are interested in doing the equivalent there. We have a registration interface to do, and I will not provide oprofile-specific workarounds for it hidden in the arch code.

Quoted 6 lines

Until then, as we currently would only need the name for SH, I don't see something wrong here to include architectural interfaces from <asm/perf_event.h> directly by architectural code until a general solution is available. If you run a 'git grep include..asm/perf' this is common for other architectures.

This is nonsense. Every architecture that needs to support the oprofile userspace tools needs to support the string interface. At the same time, everyone supporting hardware perf events in the kernel today already has a string to pmu mapping somewhere in their data structures. This can easily be exposed generically, and platforms that have special mangling they need to do before exposing the string to oprofile can do so if the need to.

Quoted 5 lines

I am not against adding the pmu name to the perf API. But the oprofile cpu_type strings are oprofile centric esp. for the userland. So these strings will remain part of oprofile. Also I don't think we want to polute the perf pmu names with it.

And those architectures that have opted to use different strings for perf events are free to mangle them however they want for the oprofile case. It doesn't change the fact that strings are still being managed by all of the architectures. The perf PMU names aren't presently locked in to an ABI, whereas the oprofile strings are, so it seems fairly straightforward to develop standard mangling rules for preventing an oprofile-facing string, or to simply reuse the strings verbatim.

Quoted 10 lines
Quoted 5 lines

sh_pmu is created for use solely by the perf events code, we are not going to have oprofile poking around in data structures it has no place knowing anything about when 99% of everything else it is doing is already abstracted cleanly through the perf events interfaces. Likewise for special oprofile-specific APIs.

So, I am also fine with implementing a generic perf_pmu_name() for sh and then derive the oprofile cpu_type string from it in the oprofile code.

Again, this is unacceptable.

#12 lore raw

Re: [PATCH 2/7] perf: New helper function for pmu name

Quoted 6 lines · Paul Mundt
Quoted 5 lines · Robert Richter

I am not against adding the pmu name to the perf API. But the oprofile cpu_type strings are oprofile centric esp. for the userland. So these strings will remain part of oprofile. Also I don't think we want to polute the perf pmu names with it.

Quoted 8 lines

And those architectures that have opted to use different strings for perf events are free to mangle them however they want for the oprofile case. It doesn't change the fact that strings are still being managed by all of the architectures. The perf PMU names aren't presently locked in to an ABI, whereas the oprofile strings are, so it seems fairly straightforward to develop standard mangling rules for preventing an oprofile-facing string, or to simply reuse the strings verbatim.

So to add a bit of context here, I was just looking at the oprofile tools. The naming format here is one of:

	<arch>/<pmu>

if there were a generic perf to oprofile pmu name mangler that did this it would cover almost all of the ARM cases already, the SH strings I'm happy to convert to work this way, and a good chunk of the PowerPC PMUs would work fine, too. PowerPC already has an oprofile CPU string in its CPU spec, so this would be even more trivial to wire up there if such a generic interface were to exist.

This would just leave x86 as the odd one out, but I suppose if x86 were to move to the oprofile perf wrapper in the future then a bit of id to name mangling as an override wouldn't be too much work.

#13 lore raw

Re: [PATCH 2/7] perf: New helper function for pmu name

Quoted 15 lines · Paul Mundt
Quoted 6 lines · Paul Mundt
Quoted 5 lines · Robert Richter

I am not against adding the pmu name to the perf API. But the oprofile cpu_type strings are oprofile centric esp. for the userland. So these strings will remain part of oprofile. Also I don't think we want to polute the perf pmu names with it.

Quoted 8 lines

And those architectures that have opted to use different strings for perf events are free to mangle them however they want for the oprofile case. It doesn't change the fact that strings are still being managed by all of the architectures. The perf PMU names aren't presently locked in to an ABI, whereas the oprofile strings are, so it seems fairly straightforward to develop standard mangling rules for preventing an oprofile-facing string, or to simply reuse the strings verbatim.

Quoted 4 lines

So to add a bit of context here, I was just looking at the oprofile tools. The naming format here is one of:

	<arch>/<pmu>

Do you really want the oprofile name scheme move to perf?

Quoted 2 lines

if there were a generic perf to oprofile pmu name mangler that did this

Or, do you mean here to derive an oprofile name from the pmu string in a generic way? I was suggesting this for SH when commenting on version 3 of this patch set. We dropped this idea to keep changes simple for this initial patch set, because it was much easier to implement it with strcmp() and the pmu strings are not expected to be changed for sh in the near future.

So, can't we use op_name_from_perf_name() from [PATCH -v3 6/6] for SH here and implement a perf_pmu_name() function for sh that is part of perf's generic interface?

Later we can add a generic function for sh ...

Quoted 9 lines

it would cover almost all of the ARM cases already, the SH strings I'm happy to convert to work this way, and a good chunk of the PowerPC PMUs would work fine, too. PowerPC already has an oprofile CPU string in its CPU spec, so this would be even more trivial to wire up there if such a generic interface were to exist.

This would just leave x86 as the odd one out, but I suppose if x86 were to move to the oprofile perf wrapper in the future then a bit of id to name mangling as an override wouldn't be too much work.

... and also other architectures on top of these patches.

Thanks,

-Robert

Signature · 2 lines
Advanced Micro Devices, Inc.
Operating System Research Center
#14 lore raw

Re: [PATCH 2/7] perf: New helper function for pmu name

Quoted 10 lines · Robert Richter
Quoted 1 line · Paul Mundt

if there were a generic perf to oprofile pmu name mangler that did this

Or, do you mean here to derive an oprofile name from the pmu string in a generic way? I was suggesting this for SH when commenting on version 3 of this patch set. We dropped this idea to keep changes simple for this initial patch set, because it was much easier to implement it with strcmp() and the pmu strings are not expected to be changed for sh in the near future.

Yes, this is what I meant, so it seems we basically have a consensus after all.

Quoted 4 lines

So, can't we use op_name_from_perf_name() from [PATCH -v3 6/6] for SH here and implement a perf_pmu_name() function for sh that is part of perf's generic interface?

As long as perf_pmu_name() is generically provided and we just have architecture overrides as necessary, then this is fine. If perf_pmu_name() is __weak then we can provide an override in the SH code that returns the PMU name without issue, and we don't have to expose sh_pmu generically, so this is the solution I prefer.

That is assuming that perf folks are ok with a generic perf_pmu_name() anyways!

Once that is provided, we can simply have the oprofile wrapper provide a generic op_name_from_perf_name() that does the arch/perf_pmu_name() string contstruction for the oprofile case, and then start killing off the special cases.

Quoted 8 lines

Later we can add a generic function for sh ...

Quoted 6 lines

it would cover almost all of the ARM cases already, the SH strings I'm happy to convert to work this way, and a good chunk of the PowerPC PMUs would work fine, too. PowerPC already has an oprofile CPU string in its CPU spec, so this would be even more trivial to wire up there if such a generic interface were to exist.

Quoted 6 lines
Quoted 3 lines

This would just leave x86 as the odd one out, but I suppose if x86 were to move to the oprofile perf wrapper in the future then a bit of id to name mangling as an override wouldn't be too much work.

... and also other architectures on top of these patches.

That sounds like a plan to me.

#15 lore raw

Re: [PATCH 2/7] perf: New helper function for pmu name

In reply to Paul Mundt
Quoted 6 lines · Paul Mundt
Quoted 4 lines · Robert Richter

So, I am also fine with implementing a generic perf_pmu_name() for sh and then derive the oprofile cpu_type string from it in the oprofile code.

Again, this is unacceptable.

Paul,

maybe I misunderstood you, but isn't this you preferred solution? We make perf_pmu_name() part of the generic i/f and then derive the oprofile name from the pmu name provided by perf.

-Robert

Signature · 2 lines
Advanced Micro Devices, Inc.
Operating System Research Center
#16 lore raw

Re: [PATCH 2/7] perf: New helper function for pmu name

Quoted 6 lines · Robert Richter
Quoted 5 lines · Paul Mundt
Quoted 4 lines · Robert Richter

So, I am also fine with implementing a generic perf_pmu_name() for sh and then derive the oprofile cpu_type string from it in the oprofile code.

Quoted 8 lines
Quoted 1 line

Again, this is unacceptable.

Paul,

maybe I misunderstood you, but isn't this you preferred solution? We make perf_pmu_name() part of the generic i/f and then derive the oprofile name from the pmu name provided by perf.

Perhaps we've misunderstood each other. I thought what you meant is that we would carry a special API deviation in the architecture code or expose the sh_pmu structure to generic code, both of which I have no interest in. My preferred solution is that we provide a perf_pmu_name() in the generic perf code as a __weak and then override it in the sh code, where we have sh_pmu visibility. This way we aren't providing any special APIs and sh_pmu remains constrained to the sh perf events code.

#17 lore raw

Re: [PATCH 2/7] perf: New helper function for pmu name

In reply to Matt Fleming
Quoted 5 lines · Matt Fleming

Introduce perf_pmu_name() helper function that returns the name of the pmu. This gives us a generic way to get the name of a pmu regardless of how an architecture identifies it internally, e.g. ARM uses an id whereas SH currently uses a string.

Quoted 9 lines
---
 arch/arm/kernel/perf_event.c |   23 +++++++++++++++++++++++
 arch/arm/oprofile/common.c   |   22 +---------------------
 arch/sh/kernel/perf_event.c  |   14 ++++++++++++++
 include/linux/perf_event.h   |    1 +
 kernel/perf_event.c          |    5 +++++
 5 files changed, 44 insertions(+), 21 deletions(-)

Matt,

as an outcome of the discussion we have had in this thread I think we agree on the following:

We extend the generic perf interface with perf_pmu_name() implenting a default with the __weak attribute. On sh we override this function and will use it for perf/oprofile string translation. This will be implemented in the oprofile function op_name_from_perf_name().

Later we implement a generic oprofile function that derives the cpu_type from perf_pmu_name() in a generic way and add other architectures. Thus, we will drop the ARM changes with this patch set.

Paul,

please correct me if I am wrong with the above.

Peter,

please ack the perf_pmu_name() interface introduced below.

See also my comments below. If there are any other objections, please let me know.

Thanks,

-Robert

Quoted 66 lines
arch/arm/kernel/perf_event.c +23 -0
index ef3bc33..3bff24d 100644
@@ -123,6 +123,29 @@ armpmu_get_max_events(void)
 }
 EXPORT_SYMBOL_GPL(armpmu_get_max_events);
 
+const char *perf_pmu_name(void)
+{
+	enum arm_perf_pmu_ids id = armpmu_get_pmu_id();
+
+	switch (id) {
+	case ARM_PERF_PMU_ID_XSCALE1:
+		return "arm/xscale1";
+	case ARM_PERF_PMU_ID_XSCALE2:
+		return "arm/xscale2";
+	case ARM_PERF_PMU_ID_V6:
+		return "arm/armv6";
+	case ARM_PERF_PMU_ID_V6MP:
+		return "arm/mpcore";
+	case ARM_PERF_PMU_ID_CA8:
+		return "arm/armv7";
+	case ARM_PERF_PMU_ID_CA9:
+		return "arm/armv7-ca9";
+	default:
+		return NULL;
+	}
+}
+EXPORT_SYMBOL_GPL(perf_pmu_name);
+
 int perf_num_counters(void)
 {
 	return armpmu_get_max_events();
arch/arm/oprofile/common.c +0 -20
index 23f18a0..cb224ee 100644
@@ -155,26 +155,6 @@ static void op_perf_stop(void)
 }
 
 
-static char *op_name_from_perf_id(enum arm_perf_pmu_ids id)
-{
-	switch (id) {
-	case ARM_PERF_PMU_ID_XSCALE1:
-		return "arm/xscale1";
-	case ARM_PERF_PMU_ID_XSCALE2:
-		return "arm/xscale2";
-	case ARM_PERF_PMU_ID_V6:
-		return "arm/armv6";
-	case ARM_PERF_PMU_ID_V6MP:
-		return "arm/mpcore";
-	case ARM_PERF_PMU_ID_CA8:
-		return "arm/armv7";
-	case ARM_PERF_PMU_ID_CA9:
-		return "arm/armv7-ca9";
-	default:
-		return NULL;
-	}
-}
-
 static int op_arm_create_files(struct super_block *sb, struct dentry *root)
 {
 	unsigned int i;

arch/arm/oprofile/common.c int __init oprofile_arch_init(struct oprofile_operations *ops) · line 374

Quoted 5 lines
arch/arm/oprofile/common.c +1 -1
@@ -391,7 +371,7 @@ int __init oprofile_arch_init(struct oprofile_operations *ops)
 	ops->start		= op_arm_start;
 	ops->stop		= op_arm_stop;
 	ops->shutdown		= op_arm_stop;
-	ops->cpu_type		= op_name_from_perf_id(armpmu_get_pmu_id());
+	ops->cpu_type		= perf_pmu_name();

Please drop all arm changes in this patch. We will do the change after a generic patch for perf/oprofile string translation is available.

Quoted 14 lines
arch/arm/oprofile/common.c +0 -0
 	if (!ops->cpu_type)
 		ret = -ENODEV;
arch/sh/kernel/perf_event.c +3 -0
index 2cb9ad5..e065a1d 100644
@@ -59,6 +59,20 @@ static inline int sh_pmu_initialized(void)
 	return !!sh_pmu;
 }
 
+const char *perf_pmu_name(void)
+{
+	if (!sh_pmu)

arch/sh/kernel/perf_event.c static inline int sh_pmu_initialized(void) · line 74

Quoted 10 lines
arch/sh/kernel/perf_event.c +10 -0
+		return NULL;
+
+	if (!strcmp(sh_pmu->name, "SH7750"))
+		return "sh/sh7750";
+	if (!strcmp(sh_pmu->name, "SH-4A"))
+		return "sh/sh4a";
+
+	return NULL;
+}
+EXPORT_SYMBOL_GPL(perf_pmu_name);

The implementation should be moved to function op_name_from_perf_name() in sh oprofile code. perf_pmu_name() should simply return sh_pmu->name.

Quoted 36 lines
arch/sh/kernel/perf_event.c +1 -0
+
 int perf_num_counters(void)
 {
 	if (!sh_pmu)
include/linux/perf_event.h +1 -0
index 1a02192..33f08da 100644
@@ -850,6 +850,7 @@ extern int perf_max_events;
 extern const struct pmu *hw_perf_event_init(struct perf_event *event);
 
 extern int perf_num_counters(void);
+extern const char *perf_pmu_name(void);
 extern void perf_event_task_sched_in(struct task_struct *task);
 extern void perf_event_task_sched_out(struct task_struct *task, struct task_struct *next);
 extern void perf_event_task_tick(struct task_struct *task);
kernel/perf_event.c +5 -0
index db5b560..fc51268 100644
@@ -85,6 +85,11 @@ void __weak hw_perf_enable(void)		{ barrier(); }
 
 void __weak perf_event_print_debug(void)	{ }
 
+extern __weak const char *perf_pmu_name(void)
+{
+	return "pmu";
+}
+
 static DEFINE_PER_CPU(int, perf_disable_count);
 
 void perf_disable(void)
Signature · 1 line
1.7.1
Signature · 2 lines
Advanced Micro Devices, Inc.
Operating System Research Center
#18 lore raw

Re: [PATCH 2/7] perf: New helper function for pmu name

Quoted 17 lines · Robert Richter
Quoted 5 lines · Matt Fleming

Introduce perf_pmu_name() helper function that returns the name of the pmu. This gives us a generic way to get the name of a pmu regardless of how an architecture identifies it internally, e.g. ARM uses an id whereas SH currently uses a string.

Matt,

as an outcome of the discussion we have had in this thread I think we agree on the following:

We extend the generic perf interface with perf_pmu_name() implenting a default with the __weak attribute. On sh we override this function and will use it for perf/oprofile string translation. This will be implemented in the oprofile function op_name_from_perf_name().

Quoted 8 lines

Later we implement a generic oprofile function that derives the cpu_type from perf_pmu_name() in a generic way and add other architectures. Thus, we will drop the ARM changes with this patch set.

Paul,

please correct me if I am wrong with the above.

That all sounds fine to me, thanks for summarizing!

#19 lore raw

PATCH v4 3/7 ARM: oprofile: Rename op_arm to oprofile_perf

In reply to Matt Fleming

In preparation for moving the generic functions out of this file, give the functions more general names (e.g. remove "arm" from the names).

---
 arch/arm/oprofile/common.c |   68 ++++++++++++++++++++++----------------------
 1 files changed, 34 insertions(+), 34 deletions(-)
arch/arm/oprofile/common.c +34 -34
index cb224ee..86df75f 100644
@@ -38,8 +38,8 @@ struct op_counter_config {
 	struct perf_event_attr attr;
 };
 
-static int op_arm_enabled;
-static DEFINE_MUTEX(op_arm_mutex);
+static int oprofile_perf_enabled;
+static DEFINE_MUTEX(oprofile_perf_mutex);
 
 static struct op_counter_config *counter_config;
 static struct perf_event **perf_events[nr_cpumask_bits];
@@ -66,7 +66,7 @@ static void op_overflow_handler(struct perf_event *event, int unused,
 }
 
 /*
- * Called by op_arm_setup to create perf attributes to mirror the oprofile
+ * Called by oprofile_perf_setup to create perf attributes to mirror the oprofile
  * settings in counter_config. Attributes are created as `pinned' events and
  * so are permanently scheduled on the PMU.
  */
@@ -123,7 +123,7 @@ static void op_destroy_counter(int cpu, int event)
 }
 
 /*
- * Called by op_arm_start to create active perf events based on the
+ * Called by oprofile_perf_start to create active perf events based on the
  * perviously configured attributes.
  */
 static int op_perf_start(void)
@@ -143,7 +143,7 @@ out:
 }
 
 /*
- * Called by op_arm_stop at the end of a profiling run.
+ * Called by oprofile_perf_stop at the end of a profiling run.
  */
 static void op_perf_stop(void)
 {
@@ -155,7 +155,7 @@ static void op_perf_stop(void)
 }
 
 
-static int op_arm_create_files(struct super_block *sb, struct dentry *root)
+static int oprofile_perf_create_files(struct super_block *sb, struct dentry *root)
 {
 	unsigned int i;
 
@@ -176,7 +176,7 @@ static int op_arm_create_files(struct super_block *sb, struct dentry *root)
 	return 0;
 }
 
-static int op_arm_setup(void)
+static int oprofile_perf_setup(void)
 {
 	spin_lock(&oprofilefs_lock);
 	op_perf_setup();
@@ -184,54 +184,54 @@ static int op_arm_setup(void)
 	return 0;
 }
 
-static int op_arm_start(void)
+static int oprofile_perf_start(void)
 {
 	int ret = -EBUSY;
 
-	mutex_lock(&op_arm_mutex);
-	if (!op_arm_enabled) {
+	mutex_lock(&oprofile_perf_mutex);
+	if (!oprofile_perf_enabled) {
 		ret = 0;
 		op_perf_start();
-		op_arm_enabled = 1;
+		oprofile_perf_enabled = 1;
 	}
-	mutex_unlock(&op_arm_mutex);
+	mutex_unlock(&oprofile_perf_mutex);
 	return ret;
 }
 
-static void op_arm_stop(void)
+static void oprofile_perf_stop(void)
 {
-	mutex_lock(&op_arm_mutex);
-	if (op_arm_enabled)
+	mutex_lock(&oprofile_perf_mutex);
+	if (oprofile_perf_enabled)
 		op_perf_stop();
-	op_arm_enabled = 0;
-	mutex_unlock(&op_arm_mutex);
+	oprofile_perf_enabled = 0;
+	mutex_unlock(&oprofile_perf_mutex);
 }
 
 #ifdef CONFIG_PM
-static int op_arm_suspend(struct platform_device *dev, pm_message_t state)
+static int oprofile_perf_suspend(struct platform_device *dev, pm_message_t state)
 {
-	mutex_lock(&op_arm_mutex);
-	if (op_arm_enabled)
+	mutex_lock(&oprofile_perf_mutex);
+	if (oprofile_perf_enabled)
 		op_perf_stop();
-	mutex_unlock(&op_arm_mutex);
+	mutex_unlock(&oprofile_perf_mutex);
 	return 0;
 }
 
-static int op_arm_resume(struct platform_device *dev)
+static int oprofile_perf_resume(struct platform_device *dev)
 {
-	mutex_lock(&op_arm_mutex);
-	if (op_arm_enabled && op_perf_start())
-		op_arm_enabled = 0;
-	mutex_unlock(&op_arm_mutex);
+	mutex_lock(&oprofile_perf_mutex);
+	if (oprofile_perf_enabled && op_perf_start())
+		oprofile_perf_enabled = 0;
+	mutex_unlock(&oprofile_perf_mutex);
 	return 0;
 }
 
 static struct platform_driver oprofile_driver = {
 	.driver		= {
-		.name		= "arm-oprofile",
+		.name		= "oprofile-perf",
 	},
-	.resume		= op_arm_resume,
-	.suspend	= op_arm_suspend,
+	.resume		= oprofile_perf_resume,
+	.suspend	= oprofile_perf_suspend,
 };
 
 static struct platform_device *oprofile_pdev;
@@ -366,11 +366,11 @@ int __init oprofile_arch_init(struct oprofile_operations *ops)
 	}
 
 	ops->backtrace		= arm_backtrace;
-	ops->create_files	= op_arm_create_files;
-	ops->setup		= op_arm_setup;
-	ops->start		= op_arm_start;
-	ops->stop		= op_arm_stop;
-	ops->shutdown		= op_arm_stop;
+	ops->create_files	= oprofile_perf_create_files;
+	ops->setup		= oprofile_perf_setup;
+	ops->start		= oprofile_perf_start;
+	ops->stop		= oprofile_perf_stop;
+	ops->shutdown		= oprofile_perf_stop;
 	ops->cpu_type		= perf_pmu_name();
 
 	if (!ops->cpu_type)
Signature · 1 line
1.7.1
#20 lore raw

Re: [PATCH 3/7] ARM: oprofile: Rename op_arm to oprofile_perf

Quoted 8 lines · Matt Fleming

In preparation for moving the generic functions out of this file, give the functions more general names (e.g. remove "arm" from the names).

---
 arch/arm/oprofile/common.c |   68 ++++++++++++++++++++++----------------------
 1 files changed, 34 insertions(+), 34 deletions(-)

This patch looks good.

-Robert

Signature · 2 lines
Advanced Micro Devices, Inc.
Operating System Research Center
#21 lore raw

PATCH v4 4/7 ARM: oprofile: Move non-ARM code into separate init/exit

In reply to Matt Fleming

In preparation for moving the majority of this oprofile code into an architecture-neutral place separate the architecture-independent code into oprofile_perf_init() and oprofile_perf_exit().

---
 arch/arm/oprofile/common.c |   17 ++++++++++++++---
 1 files changed, 14 insertions(+), 3 deletions(-)
arch/arm/oprofile/common.c +14 -3
index 86df75f..e4cd332 100644
@@ -327,7 +327,7 @@ static void arm_backtrace(struct pt_regs * const regs, unsigned int depth)
 		tail = user_backtrace(tail);
 }
 
-int __init oprofile_arch_init(struct oprofile_operations *ops)
+int __init oprofile_perf_init(struct oprofile_operations *ops)
 {
 	int cpu, ret = 0;
 
@@ -365,7 +365,6 @@ int __init oprofile_arch_init(struct oprofile_operations *ops)
 		}
 	}
 
-	ops->backtrace		= arm_backtrace;
 	ops->create_files	= oprofile_perf_create_files;
 	ops->setup		= oprofile_perf_setup;
 	ops->start		= oprofile_perf_start;
@@ -388,7 +387,14 @@ out:
 	return ret;
 }
 
-void __exit oprofile_arch_exit(void)
+int __init oprofile_arch_init(struct oprofile_operations *ops)
+{
+	ops->backtrace		= arm_backtrace;
+
+	return oprofile_perf_init(ops);
+}
+
+void __exit oprofile_perf_exit(void)
 {
 	int cpu, id;
 	struct perf_event *event;
@@ -406,6 +412,11 @@ void __exit oprofile_arch_exit(void)
 	kfree(counter_config);
 	exit_driverfs();
 }
+
+void __exit oprofile_arch_exit(void)
+{
+	oprofile_perf_exit();
+}
 #else
 int __init oprofile_arch_init(struct oprofile_operations *ops)
 {
Signature · 1 line
1.7.1
#22 lore raw

Re: [PATCH 4/7] ARM: oprofile: Move non-ARM code into separate init/exit

Quoted 32 lines · Matt Fleming

In preparation for moving the majority of this oprofile code into an architecture-neutral place separate the architecture-independent code into oprofile_perf_init() and oprofile_perf_exit().

---
 arch/arm/oprofile/common.c |   17 ++++++++++++++---
 1 files changed, 14 insertions(+), 3 deletions(-)
arch/arm/oprofile/common.c +1 -2
index 86df75f..e4cd332 100644
@@ -327,7 +327,7 @@ static void arm_backtrace(struct pt_regs * const regs, unsigned int depth)
 		tail = user_backtrace(tail);
 }
 
-int __init oprofile_arch_init(struct oprofile_operations *ops)
+int __init oprofile_perf_init(struct oprofile_operations *ops)
 {
 	int cpu, ret = 0;
 
@@ -365,7 +365,6 @@ int __init oprofile_arch_init(struct oprofile_operations *ops)
 		}
 	}
 
-	ops->backtrace		= arm_backtrace;
 	ops->create_files	= oprofile_perf_create_files;
 	ops->setup		= oprofile_perf_setup;
 	ops->start		= oprofile_perf_start;

arch/arm/oprofile/common.c out: · line 391

Quoted 6 lines
arch/arm/oprofile/common.c +2 -1
@@ -388,7 +387,14 @@ out:
 	return ret;
 }
 
-void __exit oprofile_arch_exit(void)
+int __init oprofile_arch_init(struct oprofile_operations *ops)
+{

We should do here a:

	memset(ops, 0, sizeof(*ops));

Otherwise it looks good.

-Robert

Quoted 25 lines
arch/arm/oprofile/common.c +11 -0
+	ops->backtrace		= arm_backtrace;
+
+	return oprofile_perf_init(ops);
+}
+
+void __exit oprofile_perf_exit(void)
 {
 	int cpu, id;
 	struct perf_event *event;
@@ -406,6 +412,11 @@ void __exit oprofile_arch_exit(void)
 	kfree(counter_config);
 	exit_driverfs();
 }
+
+void __exit oprofile_arch_exit(void)
+{
+	oprofile_perf_exit();
+}
 #else
 int __init oprofile_arch_init(struct oprofile_operations *ops)
 {
Signature · 1 line
1.7.1
Signature · 2 lines
Advanced Micro Devices, Inc.
Operating System Research Center
#23 lore raw

Re: [PATCH 4/7] ARM: oprofile: Move non-ARM code into separate init/exit

Quoted 11 lines · Robert Richter

arch/arm/oprofile/common.c out: · line 391

Quoted 7 lines · Matt Fleming
arch/arm/oprofile/common.c +2 -1
@@ -388,7 +387,14 @@ out:
 	return ret;
 }
 
-void __exit oprofile_arch_exit(void)
+int __init oprofile_arch_init(struct oprofile_operations *ops)
+{

We should do here a:

Quoted 4 lines
	memset(ops, 0, sizeof(*ops));

Otherwise it looks good.

What kind of crap interface is oprofile peddling, seriously?

You expect every single architecture to clear this structure that gets passed in from the oprofile code without having any idea if there are defaults assigned or not, instead of just having the oprofile code clear it out before passing it along?

#24 lore raw

Re: [PATCH 4/7] ARM: oprofile: Move non-ARM code into separate init/exit

In reply to Robert Richter
Quoted 36 lines · Robert Richter
Quoted 35 lines · Matt Fleming

In preparation for moving the majority of this oprofile code into an architecture-neutral place separate the architecture-independent code into oprofile_perf_init() and oprofile_perf_exit().

---
 arch/arm/oprofile/common.c |   17 ++++++++++++++---
 1 files changed, 14 insertions(+), 3 deletions(-)
arch/arm/oprofile/common.c +1 -2
index 86df75f..e4cd332 100644
@@ -327,7 +327,7 @@ static void arm_backtrace(struct pt_regs * const regs, unsigned int depth)
 		tail = user_backtrace(tail);
 }
 
-int __init oprofile_arch_init(struct oprofile_operations *ops)
+int __init oprofile_perf_init(struct oprofile_operations *ops)
 {
 	int cpu, ret = 0;
 
@@ -365,7 +365,6 @@ int __init oprofile_arch_init(struct oprofile_operations *ops)
 		}
 	}
 
-	ops->backtrace		= arm_backtrace;
 	ops->create_files	= oprofile_perf_create_files;
 	ops->setup		= oprofile_perf_setup;
 	ops->start		= oprofile_perf_start;
@@ -388,7 +387,14 @@ out:
 	return ret;
 }
 
Quoted 7 lines

arch/arm/oprofile/common.c out: · line 391

Quoted 3 lines
arch/arm/oprofile/common.c +2 -1
-void __exit oprofile_arch_exit(void)
+int __init oprofile_arch_init(struct oprofile_operations *ops)
+{

We should do here a:

	memset(ops, 0, sizeof(*ops));

Matt,

an alternative would be to move this to oprofile_perf_init() and modify ops->backtrace afterwards, but then we will lose the function's tail call. Not sure that is better. Paul, what do you think?

-Robert

Quoted 28 lines

Otherwise it looks good.

-Robert

Quoted 23 lines
arch/arm/oprofile/common.c +11 -0
+	ops->backtrace		= arm_backtrace;
+
+	return oprofile_perf_init(ops);
+}
+
+void __exit oprofile_perf_exit(void)
 {
 	int cpu, id;
 	struct perf_event *event;
@@ -406,6 +412,11 @@ void __exit oprofile_arch_exit(void)
 	kfree(counter_config);
 	exit_driverfs();
 }
+
+void __exit oprofile_arch_exit(void)
+{
+	oprofile_perf_exit();
+}
 #else
 int __init oprofile_arch_init(struct oprofile_operations *ops)
 {
Signature · 1 line
1.7.1
Signature · 2 lines
Advanced Micro Devices, Inc.
Operating System Research Center
#25 lore raw

Re: [PATCH 4/7] ARM: oprofile: Move non-ARM code into separate init/exit

Quoted 8 lines · Robert Richter
Quoted 7 lines · Robert Richter

Review

Quoted 3 lines · Matt Fleming
(quoted diff) +2 -1
-void __exit oprofile_arch_exit(void)
+int __init oprofile_arch_init(struct oprofile_operations *ops)
+{

We should do here a:

Quoted 8 lines
Quoted 1 line
	memset(ops, 0, sizeof(*ops));

Matt,

an alternative would be to move this to oprofile_perf_init() and modify ops->backtrace afterwards, but then we will lose the function's tail call. Not sure that is better. Paul, what do you think?

I'm a bit confused as to why we just don't do:

---

drivers/oprofile/oprof.c +2 -0
index b336cd9..76d97a5 100644
@@ -252,6 +252,8 @@ static int __init oprofile_init(void)
 {
 	int err;
 
+	memset(&oprofile_ops, 0, sizeof(struct oprofile_operations));
+
 	err = oprofile_arch_init(&oprofile_ops);
 	if (err < 0 || timer) {
 		printk(KERN_INFO "oprofile: using timer interrupt.\n");
#26 lore raw

Re: [PATCH 4/7] ARM: oprofile: Move non-ARM code into separate init/exit

Quoted 26 lines · Paul Mundt
Quoted 8 lines · Robert Richter
Quoted 7 lines · Robert Richter

Review

Quoted 3 lines · Matt Fleming
(quoted diff) +2 -1
-void __exit oprofile_arch_exit(void)
+int __init oprofile_arch_init(struct oprofile_operations *ops)
+{

We should do here a:

Quoted 8 lines
Quoted 1 line
	memset(ops, 0, sizeof(*ops));

Matt,

an alternative would be to move this to oprofile_perf_init() and modify ops->backtrace afterwards, but then we will lose the function's tail call. Not sure that is better. Paul, what do you think?

I'm a bit confused as to why we just don't do:

---

drivers/oprofile/oprof.c +0 -0
index b336cd9..76d97a5 100644

drivers/oprofile/oprof.c static int __init oprofile_init(void) · line 256

Quoted 5 lines
drivers/oprofile/oprof.c +2 -0
@@ -252,6 +252,8 @@ static int __init oprofile_init(void)
 {
 	int err;
 
+	memset(&oprofile_ops, 0, sizeof(struct oprofile_operations));
+

Yes, this looks much better. Thanks Paul.

-Robert

Quoted 4 lines
drivers/oprofile/oprof.c +0 -0
 	err = oprofile_arch_init(&oprofile_ops);
 	if (err < 0 || timer) {
 		printk(KERN_INFO "oprofile: using timer interrupt.\n");
Signature · 2 lines
Advanced Micro Devices, Inc.
Operating System Research Center
#27 lore raw

Re: [PATCH 4/7] ARM: oprofile: Move non-ARM code into separate init/exit

Quoted 30 lines · Robert Richter
Quoted 29 lines · Paul Mundt
Quoted 8 lines · Robert Richter
Quoted 7 lines · Robert Richter

Review

Quoted 3 lines · Matt Fleming
(quoted diff) +2 -1
-void __exit oprofile_arch_exit(void)
+int __init oprofile_arch_init(struct oprofile_operations *ops)
+{

We should do here a:

Quoted 8 lines
Quoted 1 line
	memset(ops, 0, sizeof(*ops));

Matt,

an alternative would be to move this to oprofile_perf_init() and modify ops->backtrace afterwards, but then we will lose the function's tail call. Not sure that is better. Paul, what do you think?

I'm a bit confused as to why we just don't do:

---

drivers/oprofile/oprof.c +0 -0
index b336cd9..76d97a5 100644
@@ -252,6 +252,8 @@ static int __init oprofile_init(void)
 {
 	int err;
 
Quoted 6 lines

drivers/oprofile/oprof.c static int __init oprofile_init(void) · line 256

Quoted 2 lines
drivers/oprofile/oprof.c +2 -0
+	memset(&oprofile_ops, 0, sizeof(struct oprofile_operations));
+

Yes, this looks much better. Thanks Paul.

In which case, if you want to take it:

#28 lore raw

Re: [PATCH 4/7] ARM: oprofile: Move non-ARM code into separate init/exit

In reply to Robert Richter
Quoted 30 lines · Robert Richter
Quoted 29 lines · Paul Mundt
Quoted 8 lines · Robert Richter
Quoted 7 lines · Robert Richter

Review

Quoted 3 lines · Matt Fleming
(quoted diff) +2 -1
-void __exit oprofile_arch_exit(void)
+int __init oprofile_arch_init(struct oprofile_operations *ops)
+{

We should do here a:

Quoted 8 lines
Quoted 1 line

  memset(ops, 0, sizeof(*ops));

Matt,

an alternative would be to move this to oprofile_perf_init() and modify ops->backtrace afterwards, but then we will lose the function's tail call. Not sure that is better. Paul, what do you think?

I'm a bit confused as to why we just don't do:

---

drivers/oprofile/oprof.c +0 -0
index b336cd9..76d97a5 100644
@@ -252,6 +252,8 @@ static int __init oprofile_init(void)
  {
       int err;
 
Quoted 4 lines

drivers/oprofile/oprof.c static int __init oprofile_init(void) · line 256

Quoted 2 lines
drivers/oprofile/oprof.c +2 -0
+     memset(&oprofile_ops, 0, sizeof(struct oprofile_operations));
+

Yes, this looks much better. Thanks Paul.

oprofile_ops is allocated in bss and so is already zeroed. No need to do it explicitly.

g.

#29 lore raw

Re: [PATCH 4/7] ARM: oprofile: Move non-ARM code into separate init/exit

Quoted 8 lines · Grant Likely
Quoted 8 lines
Quoted 8 lines
drivers/oprofile/oprof.c +0 -0
index b336cd9..76d97a5 100644
@@ -252,6 +252,8 @@ static int __init oprofile_init(void)
 ?{
 ? ? ? int err;
 
Quoted 7 lines
Quoted 4 lines

drivers/oprofile/oprof.c static int __init oprofile_init(void) · line 256

Quoted 2 lines
drivers/oprofile/oprof.c +2 -0
+ ? ? memset(&oprofile_ops, 0, sizeof(struct oprofile_operations));
+

Yes, this looks much better. Thanks Paul.

oprofile_ops is allocated in bss and so is already zeroed. No need to do it explicitly.

Before oprofile_arch_init() it isn't necessarily needed, but we should definitely add it before oprofile_timer_init(). For code readability I would rather tend to let the first in too.

-Robert

Signature · 2 lines
Advanced Micro Devices, Inc.
Operating System Research Center
#30 lore raw

Re: [PATCH 4/7] ARM: oprofile: Move non-ARM code into separate init/exit

Quoted 14 lines · Robert Richter
Quoted 13 lines · Grant Likely
Quoted 12 lines

drivers/oprofile/oprof.c static int __init oprofile_init(void) · line 256

Quoted 10 lines
drivers/oprofile/oprof.c +2 -0
index b336cd9..76d97a5 100644
@@ -252,6 +252,8 @@ static int __init oprofile_init(void)
 ?{
 ? ? ? int err;
 
+ ? ? memset(&oprofile_ops, 0, sizeof(struct oprofile_operations));
+

Yes, this looks much better. Thanks Paul.

Quoted 7 lines
Quoted 2 lines

oprofile_ops is allocated in bss and so is already zeroed. No need to do it explicitly.

Before oprofile_arch_init() it isn't necessarily needed, but we should definitely add it before oprofile_timer_init(). For code readability I would rather tend to let the first in too.

No, it was intentionally not added in between because architectures could set a backtrace op even if they had no actual counter information to plug in. oprofile_timer_init() is pretty careful to clear out all of the unused fields manually, while leaving the backtrace setting untouched.

sh, ia64, and s390 all depend on this behaviour.

#31 lore raw

Re: [PATCH 4/7] ARM: oprofile: Move non-ARM code into separate init/exit

In reply to Robert Richter
Quoted 14 lines · Robert Richter
Quoted 13 lines · Grant Likely
Quoted 12 lines

drivers/oprofile/oprof.c static int __init oprofile_init(void) · line 256

Quoted 10 lines
drivers/oprofile/oprof.c +2 -0
index b336cd9..76d97a5 100644
@@ -252,6 +252,8 @@ static int __init oprofile_init(void)
  {
       int err;
 
+     memset(&oprofile_ops, 0, sizeof(struct oprofile_operations));
+

Yes, this looks much better. Thanks Paul.

Quoted 6 lines
Quoted 2 lines

oprofile_ops is allocated in bss and so is already zeroed.  No need to do it explicitly.

Before oprofile_arch_init() it isn't necessarily needed, but we should definitely add it before oprofile_timer_init(). For code readability I would rather tend to let the first in too.

? Sorry, I don't understand what you're trying to say. Because oprofile_ops is statically allocated in bss; it will be zeroed before *any* of the init hooks is called.

g.

Quoted 8 lines

-Robert

--
Advanced Micro Devices, Inc.
Operating System Research Center

Signature · 2 lines
Grant Likely, B.Sc., P.Eng.
Secret Lab Technologies Ltd.
#32 lore raw

Re: [PATCH 4/7] ARM: oprofile: Move non-ARM code into separate init/exit

Quoted 18 lines · Grant Likely
Quoted 17 lines · Robert Richter
Quoted 15 lines · Grant Likely
Quoted 8 lines
Quoted 8 lines
drivers/oprofile/oprof.c +0 -0
index b336cd9..76d97a5 100644
@@ -252,6 +252,8 @@ static int __init oprofile_init(void)
 ?{
 ? ? ? int err;
 
Quoted 4 lines

drivers/oprofile/oprof.c static int __init oprofile_init(void) · line 256

Quoted 2 lines
drivers/oprofile/oprof.c +2 -0
+ ? ? memset(&oprofile_ops, 0, sizeof(struct oprofile_operations));
+

Yes, this looks much better. Thanks Paul.

oprofile_ops is allocated in bss and so is already zeroed. ?No need to do it explicitly.

Quoted 7 lines
Quoted 3 lines

Before oprofile_arch_init() it isn't necessarily needed, but we should definitely add it before oprofile_timer_init(). For code readability I would rather tend to let the first in too.

? Sorry, I don't understand what you're trying to say. Because oprofile_ops is statically allocated in bss; it will be zeroed before *any* of the init hooks is called.

The first call to oprofile_arch_init() might leave a dirty struct, but as Paul just pointed out the function is aware of it and cleans up the struct during init where necessary. The backtrace callback setup will be reused if arch init fails (seems to be an ugly hack). So it is better not to zero the struct then and keep everything unchanged.

-Robert

Signature · 2 lines
Advanced Micro Devices, Inc.
Operating System Research Center
#33 lore raw

PATCH v4 5/7 oprofile: Abstract the perf-events backend

In reply to Matt Fleming

Move the perf-events backend from arch/arm/oprofile into drivers/oprofile so that the code can be shared between architectures.

This allows each architecture to maintain only a single copy of the PMU accessor functions instead of one for both perf and OProfile. It also becomes possible for other architectures to delete much of their OProfile code in favour of the common code now available in drivers/oprofile/oprofile_perf.c.

---
 arch/arm/oprofile/Makefile       |    4 +
 arch/arm/oprofile/common.c       |  319 -------------------------------------
 drivers/oprofile/oprofile_perf.c |  326 ++++++++++++++++++++++++++++++++++++++
 include/linux/oprofile.h         |    6 +
 4 files changed, 336 insertions(+), 319 deletions(-)
 create mode 100644 drivers/oprofile/oprofile_perf.c
arch/arm/oprofile/Makefile +4 -0
index e666eaf..b2215c6 100644
@@ -6,4 +6,8 @@ DRIVER_OBJS = $(addprefix ../../../drivers/oprofile/, \
 		oprofilefs.o oprofile_stats.o \
 		timer_int.o )
 
+ifeq ($(CONFIG_HW_PERF_EVENTS),y)
+DRIVER_OBJS += $(addprefix ../../../drivers/oprofile/, oprofile_perf.o)
+endif
+
 oprofile-y				:= $(DRIVER_OBJS) common.o
arch/arm/oprofile/common.c +0 -319

337 lines not on this page · raw message

drivers/oprofile/oprofile_perf.c new file +326 -0

326 lines not on this page · raw message

include/linux/oprofile.h +6 -0
index 5171639..a53bfc0 100644
@@ -15,6 +15,7 @@
 
 #include <linux/types.h>
 #include <linux/spinlock.h>
+#include <linux/init.h>
 #include <asm/atomic.h>
  
 /* Each escaped entry is prefixed by ESCAPE_CODE
@@ -185,4 +186,9 @@ int oprofile_add_data(struct op_entry *entry, unsigned long val);
 int oprofile_add_data64(struct op_entry *entry, u64 val);
 int oprofile_write_commit(struct op_entry *entry);
 
+#ifdef CONFIG_PERF_EVENTS
+int __init oprofile_perf_init(struct oprofile_operations *ops);
+void __exit oprofile_perf_exit(void);
+#endif /* CONFIG_PERF_EVENTS */
+
 #endif /* OPROFILE_H */
Signature · 1 line
1.7.1
#34 lore raw

Re: [PATCH 5/7] oprofile: Abstract the perf-events backend

Quoted 11 lines · Matt Fleming

Move the perf-events backend from arch/arm/oprofile into drivers/oprofile so that the code can be shared between architectures.

This allows each architecture to maintain only a single copy of the PMU accessor functions instead of one for both perf and OProfile. It also becomes possible for other architectures to delete much of their OProfile code in favour of the common code now available in drivers/oprofile/oprofile_perf.c.

Quoted 7 lines
---
 arch/arm/oprofile/Makefile       |    4 +
 arch/arm/oprofile/common.c       |  319 -------------------------------------
 drivers/oprofile/oprofile_perf.c |  326 ++++++++++++++++++++++++++++++++++++++
 include/linux/oprofile.h         |    6 +
 4 files changed, 336 insertions(+), 319 deletions(-)
 create mode 100644 drivers/oprofile/oprofile_perf.c

This patch looks good.

-Robert

Signature · 2 lines
Advanced Micro Devices, Inc.
Operating System Research Center
#35 lore raw

PATCH v4 6/7 sh: oprofile: Use perf-events oprofile backend

In reply to Matt Fleming

Now that we've got a generic perf-events based oprofile backend we might as well make use of it seeing as SH doesn't do anything special with its oprofile backend. Also introduce a new CONFIG_HW_PERF_EVENTS symbol so that we can fallback to using the timer interrupt for oprofile if the CPU doesn't support perf events.

---
 arch/sh/Kconfig            |   13 +++++
 arch/sh/oprofile/Makefile  |    4 ++
 arch/sh/oprofile/common.c  |  109 ++++---------------------------------------
 arch/sh/oprofile/op_impl.h |   33 -------------
 4 files changed, 28 insertions(+), 131 deletions(-)
 delete mode 100644 arch/sh/oprofile/op_impl.h
arch/sh/Kconfig +13 -0
index 33990fa..35b6c3f 100644
@@ -249,6 +249,11 @@ config ARCH_SHMOBILE
 	select PM
 	select PM_RUNTIME
 
+config CPU_HAS_PMU
+       depends on CPU_SH4 || CPU_SH4A
+       default y
+       bool
+
 if SUPERH32
 
 choice
@@ -738,6 +743,14 @@ config GUSA_RB
 	  LLSC, this should be more efficient than the other alternative of
 	  disabling interrupts around the atomic sequence.
 
+config HW_PERF_EVENTS
+	bool "Enable hardware performance counter support for perf events"
+	depends on PERF_EVENTS && CPU_HAS_PMU
+	default y
+	help
+	  Enable hardware performance counter support for perf events. If
+	  disabled, perf events will use software events only.
+
 source "drivers/sh/Kconfig"
 
 endmenu
arch/sh/oprofile/Makefile +4 -0
index 4886c5c..e85aae7 100644
@@ -6,4 +6,8 @@ DRIVER_OBJS = $(addprefix ../../../drivers/oprofile/, \
 		oprofilefs.o oprofile_stats.o \
 		timer_int.o )
 
+ifeq ($(CONFIG_HW_PERF_EVENTS),y)
+DRIVER_OBJS += $(addprefix ../../../drivers/oprofile/, oprofile_perf.o)
+endif
+
 oprofile-y	:= $(DRIVER_OBJS) common.o backtrace.o
arch/sh/oprofile/common.c +11 -98

124 lines not on this page · raw message

arch/sh/oprofile/op_impl.h deleted +0 -33
deleted file mode 100644
index 1244479..0000000
@@ -1,33 +0,0 @@
-#ifndef __OP_IMPL_H
-#define __OP_IMPL_H
-
-/* Per-counter configuration as set via oprofilefs.  */
-struct op_counter_config {
-	unsigned long enabled;
-	unsigned long event;
-
-	unsigned long count;
-
-	/* Dummy values for userspace tool compliance */
-	unsigned long kernel;
-	unsigned long user;
-	unsigned long unit_mask;
-};
-
-/* Per-architecture configury and hooks.  */
-struct op_sh_model {
-	void (*reg_setup)(struct op_counter_config *);
-	int (*create_files)(struct super_block *sb, struct dentry *dir);
-	void (*cpu_setup)(void *dummy);
-	int (*init)(void);
-	void (*exit)(void);
-	void (*cpu_start)(void *args);
-	void (*cpu_stop)(void *args);
-	char *cpu_type;
-	unsigned char num_counters;
-};
-
-/* arch/sh/oprofile/common.c */
-extern void sh_backtrace(struct pt_regs * const regs, unsigned int depth);
-
-#endif /* __OP_IMPL_H */
Signature · 1 line
1.7.1
Acked-by on v4 6/7
#36 lore raw

Re: [PATCH 6/7] sh: oprofile: Use perf-events oprofile backend

Quoted 7 lines · Matt Fleming

Now that we've got a generic perf-events based oprofile backend we might as well make use of it seeing as SH doesn't do anything special with its oprofile backend. Also introduce a new CONFIG_HW_PERF_EVENTS symbol so that we can fallback to using the timer interrupt for oprofile if the CPU doesn't support perf events.

#37 lore raw

Re: [PATCH 6/7] sh: oprofile: Use perf-events oprofile backend

In reply to Matt Fleming
Quoted 6 lines · Matt Fleming

Now that we've got a generic perf-events based oprofile backend we might as well make use of it seeing as SH doesn't do anything special with its oprofile backend. Also introduce a new CONFIG_HW_PERF_EVENTS symbol so that we can fallback to using the timer interrupt for oprofile if the CPU doesn't support perf events.

Quoted 8 lines
---
 arch/sh/Kconfig            |   13 +++++
 arch/sh/oprofile/Makefile  |    4 ++
 arch/sh/oprofile/common.c  |  109 ++++---------------------------------------
 arch/sh/oprofile/op_impl.h |   33 -------------
 4 files changed, 28 insertions(+), 131 deletions(-)
 delete mode 100644 arch/sh/oprofile/op_impl.h

This patch looks good.

We might merge it with patch 7/7.

-Robert

Signature · 2 lines
Advanced Micro Devices, Inc.
Operating System Research Center
#38 lore raw

Re: [PATCH 6/7] sh: oprofile: Use perf-events oprofile backend

Quoted 16 lines · Robert Richter
Quoted 6 lines · Matt Fleming

Now that we've got a generic perf-events based oprofile backend we might as well make use of it seeing as SH doesn't do anything special with its oprofile backend. Also introduce a new CONFIG_HW_PERF_EVENTS symbol so that we can fallback to using the timer interrupt for oprofile if the CPU doesn't support perf events.

Quoted 8 lines
---
 arch/sh/Kconfig            |   13 +++++
 arch/sh/oprofile/Makefile  |    4 ++
 arch/sh/oprofile/common.c  |  109 ++++---------------------------------------
 arch/sh/oprofile/op_impl.h |   33 -------------
 4 files changed, 28 insertions(+), 131 deletions(-)
 delete mode 100644 arch/sh/oprofile/op_impl.h
Quoted 3 lines

This patch looks good.

We might merge it with patch 7/7.

This patch was broken out for Paul's benefit to ease review. If he's happy for me to merge this with 7/7 then that's fine with me.

#39 lore raw

Re: [PATCH 6/7] sh: oprofile: Use perf-events oprofile backend

Quoted 17 lines · Matt Fleming
Quoted 16 lines · Robert Richter
Quoted 14 lines · Matt Fleming

Now that we've got a generic perf-events based oprofile backend we might as well make use of it seeing as SH doesn't do anything special with its oprofile backend. Also introduce a new CONFIG_HW_PERF_EVENTS symbol so that we can fallback to using the timer interrupt for oprofile if the CPU doesn't support perf events.

---
 arch/sh/Kconfig            |   13 +++++
 arch/sh/oprofile/Makefile  |    4 ++
 arch/sh/oprofile/common.c  |  109 ++++---------------------------------------
 arch/sh/oprofile/op_impl.h |   33 -------------
 4 files changed, 28 insertions(+), 131 deletions(-)
 delete mode 100644 arch/sh/oprofile/op_impl.h
Quoted 6 lines
Quoted 3 lines

This patch looks good.

We might merge it with patch 7/7.

This patch was broken out for Paul's benefit to ease review. If he's happy for me to merge this with 7/7 then that's fine with me.

Yes, no objections from me.

#40 lore raw

PATCH v4 7/7 sh: Annotate oprofile_arch_exit() with __exit marker

In reply to Matt Fleming

In order to avoid a section mismatch we need to annotate oprofile_arch_exit() with an __exit marker.

---
 arch/sh/oprofile/common.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
arch/sh/oprofile/common.c +1 -1
index 449f842..93033d8 100644
@@ -29,7 +29,7 @@ int __init oprofile_arch_init(struct oprofile_operations *ops)
 	return oprofile_perf_init(ops);
 }
 
-void oprofile_arch_exit(void)
+void __exit oprofile_arch_exit(void)
 {
 	oprofile_perf_exit();
 }
Signature · 1 line
1.7.1
#42 lore raw

Re: [PATCH 7/7] sh: Annotate oprofile_arch_exit() with __exit marker

In reply to Matt Fleming
Quoted 4 lines · Matt Fleming

In order to avoid a section mismatch we need to annotate oprofile_arch_exit() with an __exit marker.

Looks good.

-Robert

Quoted 21 lines
---
 arch/sh/oprofile/common.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
arch/sh/oprofile/common.c +1 -1
index 449f842..93033d8 100644
@@ -29,7 +29,7 @@ int __init oprofile_arch_init(struct oprofile_operations *ops)
 	return oprofile_perf_init(ops);
 }
 
-void oprofile_arch_exit(void)
+void __exit oprofile_arch_exit(void)
 {
 	oprofile_perf_exit();
 }
Signature · 1 line
1.7.1
Signature · 2 lines
Advanced Micro Devices, Inc.
Operating System Research Center