[PATCH 0/5] arm_pmu: expose 'cpus' file in sysfs

[PATCH 0/5] arm_pmu: expose 'cpus' file in sysfs

[Mark Rutland]

Hi Will,

These are the arm_pmu patches previous posted as part of my arm_pmu/perf tools
series [1], fixed up as per your comments.

I've given this a spin on Juno, and everything seems to work as expected.

Thanks,
Mark.

[1] http://lists.infradead.org/pipermail/linux-arm-kernel/2016-September/455206.html

Mark Rutland (5):
  drivers/perf: arm_pmu: add common attr group fields
  arm64: perf: move to common attr_group fields
  arm: perf: move to common attr_group fields
  drivers/perf: arm_pmu: only use common attr_groups
  drivers/perf: arm_pmu: expose a cpumask in sysfs

 arch/arm/kernel/perf_event_v7.c | 47 ++++++++++++++++++++++++-----------------
 arch/arm64/kernel/perf_event.c  | 36 ++++++++++++++++++++-----------
 drivers/perf/arm_pmu.c          | 22 +++++++++++++++++++
 include/linux/perf/arm_pmu.h    | 11 +++++++++-
 4 files changed, 84 insertions(+), 32 deletions(-)

-- 
1.9.1

[PATCH 1/5] drivers/perf: arm_pmu: add common attr group fields

[Mark Rutland]

In preparation for adding common attribute groups, add an array of
attribute group pointers to arm_pmu, which will be used if the
backend hasn't already set pmu::attr_groups.

Subsequent patches will move backends over to using these, before adding
common fields.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
---
 drivers/perf/arm_pmu.c       |  3 +++
 include/linux/perf/arm_pmu.h | 10 +++++++++-
 2 files changed, 12 insertions(+), 1 deletion(-)

diff --git a/drivers/perf/arm_pmu.c b/drivers/perf/arm_pmu.c
index f5e1008..145caf4 100644
--- a/drivers/perf/arm_pmu.c
+++ b/drivers/perf/arm_pmu.c
@@ -1039,6 +1039,9 @@ int arm_pmu_device_probe(struct platform_device *pdev,
 		goto out_free;
 	}
 
+	if (!pmu->pmu.attr_groups)
+		pmu->pmu.attr_groups = pmu->attr_groups;
+
 	ret = cpu_pmu_init(pmu);
 	if (ret)
 		goto out_free;
diff --git a/include/linux/perf/arm_pmu.h b/include/linux/perf/arm_pmu.h
index e188438..268bc63 100644
--- a/include/linux/perf/arm_pmu.h
+++ b/include/linux/perf/arm_pmu.h
@@ -14,7 +14,7 @@
 
 #include <linux/interrupt.h>
 #include <linux/perf_event.h>
-
+#include <linux/sysfs.h>
 #include <asm/cputype.h>
 
 /*
@@ -77,6 +77,12 @@ struct pmu_hw_events {
 	struct arm_pmu		*percpu_pmu;
 };
 
+enum armpmu_attr_groups {
+	ARMPMU_ATTR_GROUP_EVENTS,
+	ARMPMU_ATTR_GROUP_FORMATS,
+	ARMPMU_NR_ATTR_GROUPS
+};
+
 struct arm_pmu {
 	struct pmu	pmu;
 	cpumask_t	active_irqs;
@@ -111,6 +117,8 @@ struct arm_pmu {
 	struct pmu_hw_events	__percpu *hw_events;
 	struct list_head	entry;
 	struct notifier_block	cpu_pm_nb;
+	/* the attr_groups array must be NULL-terminated */
+	const struct attribute_group *attr_groups[ARMPMU_NR_ATTR_GROUPS + 1];
 };
 
 #define to_arm_pmu(p) (container_of(p, struct arm_pmu, pmu))
-- 
1.9.1

[PATCH 2/5] arm64: perf: move to common attr_group fields

[Mark Rutland]

By using a common attr_groups array, the common arm_pmu code can set up
common files (e.g. cpumask) for us in subsequent patches.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
---
 arch/arm64/kernel/perf_event.c | 36 ++++++++++++++++++++++++------------
 1 file changed, 24 insertions(+), 12 deletions(-)

diff --git a/arch/arm64/kernel/perf_event.c b/arch/arm64/kernel/perf_event.c
index 838ccf1..bcc8dff 100644
--- a/arch/arm64/kernel/perf_event.c
+++ b/arch/arm64/kernel/perf_event.c
@@ -523,12 +523,6 @@ static struct attribute_group armv8_pmuv3_format_attr_group = {
 	.attrs = armv8_pmuv3_format_attrs,
 };
 
-static const struct attribute_group *armv8_pmuv3_attr_groups[] = {
-	&armv8_pmuv3_events_attr_group,
-	&armv8_pmuv3_format_attr_group,
-	NULL,
-};
-
 /*
  * Perf Events' indices
  */
@@ -985,7 +979,10 @@ static int armv8_pmuv3_init(struct arm_pmu *cpu_pmu)
 	armv8_pmu_init(cpu_pmu);
 	cpu_pmu->name			= "armv8_pmuv3";
 	cpu_pmu->map_event		= armv8_pmuv3_map_event;
-	cpu_pmu->pmu.attr_groups	= armv8_pmuv3_attr_groups;
+	cpu_pmu->attr_groups[ARMPMU_ATTR_GROUP_EVENTS] =
+		&armv8_pmuv3_events_attr_group;
+	cpu_pmu->attr_groups[ARMPMU_ATTR_GROUP_FORMATS] =
+		&armv8_pmuv3_format_attr_group;
 	return armv8pmu_probe_pmu(cpu_pmu);
 }
 
@@ -994,7 +991,10 @@ static int armv8_a53_pmu_init(struct arm_pmu *cpu_pmu)
 	armv8_pmu_init(cpu_pmu);
 	cpu_pmu->name			= "armv8_cortex_a53";
 	cpu_pmu->map_event		= armv8_a53_map_event;
-	cpu_pmu->pmu.attr_groups	= armv8_pmuv3_attr_groups;
+	cpu_pmu->attr_groups[ARMPMU_ATTR_GROUP_EVENTS] =
+		&armv8_pmuv3_events_attr_group;
+	cpu_pmu->attr_groups[ARMPMU_ATTR_GROUP_FORMATS] =
+		&armv8_pmuv3_format_attr_group;
 	return armv8pmu_probe_pmu(cpu_pmu);
 }
 
@@ -1003,7 +1003,10 @@ static int armv8_a57_pmu_init(struct arm_pmu *cpu_pmu)
 	armv8_pmu_init(cpu_pmu);
 	cpu_pmu->name			= "armv8_cortex_a57";
 	cpu_pmu->map_event		= armv8_a57_map_event;
-	cpu_pmu->pmu.attr_groups	= armv8_pmuv3_attr_groups;
+	cpu_pmu->attr_groups[ARMPMU_ATTR_GROUP_EVENTS] =
+		&armv8_pmuv3_events_attr_group;
+	cpu_pmu->attr_groups[ARMPMU_ATTR_GROUP_FORMATS] =
+		&armv8_pmuv3_format_attr_group;
 	return armv8pmu_probe_pmu(cpu_pmu);
 }
 
@@ -1012,7 +1015,10 @@ static int armv8_a72_pmu_init(struct arm_pmu *cpu_pmu)
 	armv8_pmu_init(cpu_pmu);
 	cpu_pmu->name			= "armv8_cortex_a72";
 	cpu_pmu->map_event		= armv8_a57_map_event;
-	cpu_pmu->pmu.attr_groups	= armv8_pmuv3_attr_groups;
+	cpu_pmu->attr_groups[ARMPMU_ATTR_GROUP_EVENTS] =
+		&armv8_pmuv3_events_attr_group;
+	cpu_pmu->attr_groups[ARMPMU_ATTR_GROUP_FORMATS] =
+		&armv8_pmuv3_format_attr_group;
 	return armv8pmu_probe_pmu(cpu_pmu);
 }
 
@@ -1021,7 +1027,10 @@ static int armv8_thunder_pmu_init(struct arm_pmu *cpu_pmu)
 	armv8_pmu_init(cpu_pmu);
 	cpu_pmu->name			= "armv8_cavium_thunder";
 	cpu_pmu->map_event		= armv8_thunder_map_event;
-	cpu_pmu->pmu.attr_groups	= armv8_pmuv3_attr_groups;
+	cpu_pmu->attr_groups[ARMPMU_ATTR_GROUP_EVENTS] =
+		&armv8_pmuv3_events_attr_group;
+	cpu_pmu->attr_groups[ARMPMU_ATTR_GROUP_FORMATS] =
+		&armv8_pmuv3_format_attr_group;
 	return armv8pmu_probe_pmu(cpu_pmu);
 }
 
@@ -1030,7 +1039,10 @@ static int armv8_vulcan_pmu_init(struct arm_pmu *cpu_pmu)
 	armv8_pmu_init(cpu_pmu);
 	cpu_pmu->name			= "armv8_brcm_vulcan";
 	cpu_pmu->map_event		= armv8_vulcan_map_event;
-	cpu_pmu->pmu.attr_groups	= armv8_pmuv3_attr_groups;
+	cpu_pmu->attr_groups[ARMPMU_ATTR_GROUP_EVENTS] =
+		&armv8_pmuv3_events_attr_group;
+	cpu_pmu->attr_groups[ARMPMU_ATTR_GROUP_FORMATS] =
+		&armv8_pmuv3_format_attr_group;
 	return armv8pmu_probe_pmu(cpu_pmu);
 }
 
-- 
1.9.1

[PATCH 3/5] arm: perf: move to common attr_group fields

[Mark Rutland]

By using a common attr_groups array, the common arm_pmu code can set up
common files (e.g. cpumask) for us in subsequent patches.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
---
 arch/arm/kernel/perf_event_v7.c | 47 ++++++++++++++++++++++++-----------------
 1 file changed, 28 insertions(+), 19 deletions(-)

diff --git a/arch/arm/kernel/perf_event_v7.c b/arch/arm/kernel/perf_event_v7.c
index 1506385..b942349 100644
--- a/arch/arm/kernel/perf_event_v7.c
+++ b/arch/arm/kernel/perf_event_v7.c
@@ -596,12 +596,6 @@ static struct attribute_group armv7_pmuv1_events_attr_group = {
 	.attrs = armv7_pmuv1_event_attrs,
 };
 
-static const struct attribute_group *armv7_pmuv1_attr_groups[] = {
-	&armv7_pmuv1_events_attr_group,
-	&armv7_pmu_format_attr_group,
-	NULL,
-};
-
 ARMV7_EVENT_ATTR(mem_access, ARMV7_PERFCTR_MEM_ACCESS);
 ARMV7_EVENT_ATTR(l1i_cache, ARMV7_PERFCTR_L1_ICACHE_ACCESS);
 ARMV7_EVENT_ATTR(l1d_cache_wb, ARMV7_PERFCTR_L1_DCACHE_WB);
@@ -653,12 +647,6 @@ static struct attribute_group armv7_pmuv2_events_attr_group = {
 	.attrs = armv7_pmuv2_event_attrs,
 };
 
-static const struct attribute_group *armv7_pmuv2_attr_groups[] = {
-	&armv7_pmuv2_events_attr_group,
-	&armv7_pmu_format_attr_group,
-	NULL,
-};
-
 /*
  * Perf Events' indices
  */
@@ -1208,7 +1196,10 @@ static int armv7_a8_pmu_init(struct arm_pmu *cpu_pmu)
 	armv7pmu_init(cpu_pmu);
 	cpu_pmu->name		= "armv7_cortex_a8";
 	cpu_pmu->map_event	= armv7_a8_map_event;
-	cpu_pmu->pmu.attr_groups = armv7_pmuv1_attr_groups;
+	cpu_pmu->attr_groups[ARMPMU_ATTR_GROUP_EVENTS] =
+		&armv7_pmuv1_events_attr_group;
+	cpu_pmu->attr_groups[ARMPMU_ATTR_GROUP_FORMATS] =
+		&armv7_pmu_format_attr_group;
 	return armv7_probe_num_events(cpu_pmu);
 }
 
@@ -1217,7 +1208,10 @@ static int armv7_a9_pmu_init(struct arm_pmu *cpu_pmu)
 	armv7pmu_init(cpu_pmu);
 	cpu_pmu->name		= "armv7_cortex_a9";
 	cpu_pmu->map_event	= armv7_a9_map_event;
-	cpu_pmu->pmu.attr_groups = armv7_pmuv1_attr_groups;
+	cpu_pmu->attr_groups[ARMPMU_ATTR_GROUP_EVENTS] =
+		&armv7_pmuv1_events_attr_group;
+	cpu_pmu->attr_groups[ARMPMU_ATTR_GROUP_FORMATS] =
+		&armv7_pmu_format_attr_group;
 	return armv7_probe_num_events(cpu_pmu);
 }
 
@@ -1226,7 +1220,10 @@ static int armv7_a5_pmu_init(struct arm_pmu *cpu_pmu)
 	armv7pmu_init(cpu_pmu);
 	cpu_pmu->name		= "armv7_cortex_a5";
 	cpu_pmu->map_event	= armv7_a5_map_event;
-	cpu_pmu->pmu.attr_groups = armv7_pmuv1_attr_groups;
+	cpu_pmu->attr_groups[ARMPMU_ATTR_GROUP_EVENTS] =
+		&armv7_pmuv1_events_attr_group;
+	cpu_pmu->attr_groups[ARMPMU_ATTR_GROUP_FORMATS] =
+		&armv7_pmu_format_attr_group;
 	return armv7_probe_num_events(cpu_pmu);
 }
 
@@ -1236,7 +1233,10 @@ static int armv7_a15_pmu_init(struct arm_pmu *cpu_pmu)
 	cpu_pmu->name		= "armv7_cortex_a15";
 	cpu_pmu->map_event	= armv7_a15_map_event;
 	cpu_pmu->set_event_filter = armv7pmu_set_event_filter;
-	cpu_pmu->pmu.attr_groups = armv7_pmuv2_attr_groups;
+	cpu_pmu->attr_groups[ARMPMU_ATTR_GROUP_EVENTS] =
+		&armv7_pmuv2_events_attr_group;
+	cpu_pmu->attr_groups[ARMPMU_ATTR_GROUP_FORMATS] =
+		&armv7_pmu_format_attr_group;
 	return armv7_probe_num_events(cpu_pmu);
 }
 
@@ -1246,7 +1246,10 @@ static int armv7_a7_pmu_init(struct arm_pmu *cpu_pmu)
 	cpu_pmu->name		= "armv7_cortex_a7";
 	cpu_pmu->map_event	= armv7_a7_map_event;
 	cpu_pmu->set_event_filter = armv7pmu_set_event_filter;
-	cpu_pmu->pmu.attr_groups = armv7_pmuv2_attr_groups;
+	cpu_pmu->attr_groups[ARMPMU_ATTR_GROUP_EVENTS] =
+		&armv7_pmuv2_events_attr_group;
+	cpu_pmu->attr_groups[ARMPMU_ATTR_GROUP_FORMATS] =
+		&armv7_pmu_format_attr_group;
 	return armv7_probe_num_events(cpu_pmu);
 }
 
@@ -1256,7 +1259,10 @@ static int armv7_a12_pmu_init(struct arm_pmu *cpu_pmu)
 	cpu_pmu->name		= "armv7_cortex_a12";
 	cpu_pmu->map_event	= armv7_a12_map_event;
 	cpu_pmu->set_event_filter = armv7pmu_set_event_filter;
-	cpu_pmu->pmu.attr_groups = armv7_pmuv2_attr_groups;
+	cpu_pmu->attr_groups[ARMPMU_ATTR_GROUP_EVENTS] =
+		&armv7_pmuv2_events_attr_group;
+	cpu_pmu->attr_groups[ARMPMU_ATTR_GROUP_FORMATS] =
+		&armv7_pmu_format_attr_group;
 	return armv7_probe_num_events(cpu_pmu);
 }
 
@@ -1264,7 +1270,10 @@ static int armv7_a17_pmu_init(struct arm_pmu *cpu_pmu)
 {
 	int ret = armv7_a12_pmu_init(cpu_pmu);
 	cpu_pmu->name = "armv7_cortex_a17";
-	cpu_pmu->pmu.attr_groups = armv7_pmuv2_attr_groups;
+	cpu_pmu->attr_groups[ARMPMU_ATTR_GROUP_EVENTS] =
+		&armv7_pmuv2_events_attr_group;
+	cpu_pmu->attr_groups[ARMPMU_ATTR_GROUP_FORMATS] =
+		&armv7_pmu_format_attr_group;
 	return ret;
 }
 
-- 
1.9.1

[PATCH 4/5] drivers/perf: arm_pmu: only use common attr_groups

[Mark Rutland]

Now that the 32-bit and 64-bit perf backends use the common groups
directly, remove the fallback and no longer allow the groups array to be
overridden.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
---
 drivers/perf/arm_pmu.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/perf/arm_pmu.c b/drivers/perf/arm_pmu.c
index 145caf4..ac83e1e 100644
--- a/drivers/perf/arm_pmu.c
+++ b/drivers/perf/arm_pmu.c
@@ -549,6 +549,7 @@ static void armpmu_init(struct arm_pmu *armpmu)
 		.stop		= armpmu_stop,
 		.read		= armpmu_read,
 		.filter_match	= armpmu_filter_match,
+		.attr_groups	= armpmu->attr_groups,
 	};
 }
 
@@ -1039,8 +1040,6 @@ int arm_pmu_device_probe(struct platform_device *pdev,
 		goto out_free;
 	}
 
-	if (!pmu->pmu.attr_groups)
-		pmu->pmu.attr_groups = pmu->attr_groups;
 
 	ret = cpu_pmu_init(pmu);
 	if (ret)
-- 
1.9.1

[PATCH 5/5] drivers/perf: arm_pmu: expose a cpumask in sysfs

[Mark Rutland]

In systems with heterogeneous CPUs, there are multiple logical CPU PMUs,
each of which covers a subset of CPUs in the system. In some cases
userspace needs to know which CPUs a given logical PMU covers, so we'd
like to expose a cpumask under sysfs, similar to what is done for uncore
PMUs.

Unfortunately, prior to commit 00e727bb389359c8 ("perf stat: Balance
opening and reading events"), perf stat only correctly handled a cpumask
holding a single CPU, and only when profiling in system-wide mode. In
other cases, the presence of a cpumask file could cause perf stat to
behave erratically.

Thus, exposing a cpumask file would break older perf binaries in cases
where they would otherwise work.

To avoid this issue while still providing userspace with the information
it needs, this patch exposes a differently-named file (cpus) under
sysfs. New tools can look for this and operate correctly, while older
tools will not be adversely affected by its presence.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
---
 drivers/perf/arm_pmu.c       | 20 ++++++++++++++++++++
 include/linux/perf/arm_pmu.h |  1 +
 2 files changed, 21 insertions(+)

diff --git a/drivers/perf/arm_pmu.c b/drivers/perf/arm_pmu.c
index ac83e1e..a484b51 100644
--- a/drivers/perf/arm_pmu.c
+++ b/drivers/perf/arm_pmu.c
@@ -534,6 +534,24 @@ static int armpmu_filter_match(struct perf_event *event)
 	return cpumask_test_cpu(cpu, &armpmu->supported_cpus);
 }
 
+static ssize_t armpmu_cpumask_show(struct device *dev,
+				   struct device_attribute *attr, char *buf)
+{
+	struct arm_pmu *armpmu = to_arm_pmu(dev_get_drvdata(dev));
+	return cpumap_print_to_pagebuf(true, buf, &armpmu->supported_cpus);
+}
+
+static DEVICE_ATTR(cpus, S_IRUGO, armpmu_cpumask_show, NULL);
+
+static struct attribute *armpmu_common_attrs[] = {
+	&dev_attr_cpus.attr,
+	NULL,
+};
+
+static struct attribute_group armpmu_common_attr_group = {
+	.attrs = armpmu_common_attrs,
+};
+
 static void armpmu_init(struct arm_pmu *armpmu)
 {
 	atomic_set(&armpmu->active_events, 0);
@@ -551,6 +569,8 @@ static void armpmu_init(struct arm_pmu *armpmu)
 		.filter_match	= armpmu_filter_match,
 		.attr_groups	= armpmu->attr_groups,
 	};
+	armpmu->attr_groups[ARMPMU_ATTR_GROUP_COMMON] =
+		&armpmu_common_attr_group;
 }
 
 /* Set at runtime when we know what CPU type we are. */
diff --git a/include/linux/perf/arm_pmu.h b/include/linux/perf/arm_pmu.h
index 268bc63..dc1f2f3 100644
--- a/include/linux/perf/arm_pmu.h
+++ b/include/linux/perf/arm_pmu.h
@@ -78,6 +78,7 @@ struct pmu_hw_events {
 };
 
 enum armpmu_attr_groups {
+	ARMPMU_ATTR_GROUP_COMMON,
 	ARMPMU_ATTR_GROUP_EVENTS,
 	ARMPMU_ATTR_GROUP_FORMATS,
 	ARMPMU_NR_ATTR_GROUPS
-- 
1.9.1