linux-arm-kernel.lists.infradead.org archive mirror
 help / color / mirror / Atom feed
* [RFC PATCH 09/15] ARM: perf: lock PMU registers per-CPU
@ 2011-04-28  9:17 Mark Rutland
  0 siblings, 0 replies; 2+ messages in thread
From: Mark Rutland @ 2011-04-28  9:17 UTC (permalink / raw)
  To: linux-arm-kernel

Currently, a single lock serialises access to CPU PMU registers. This
global locking is unnecessary as PMU registers are local to the CPU
they monitor.

This patch replaces the global lock with a per-CPU lock. As the lock is
in struct cpu_hw_events, PMUs providing a single cpu_hw_events instance
can be locked globally.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Will Deacon <will.deacon@arm.com>
---
 arch/arm/kernel/perf_event.c        |   17 +++++++++-----
 arch/arm/kernel/perf_event_v6.c     |   25 +++++++++++++--------
 arch/arm/kernel/perf_event_v7.c     |   20 ++++++++++-------
 arch/arm/kernel/perf_event_xscale.c |   40 +++++++++++++++++++++--------------
 4 files changed, 62 insertions(+), 40 deletions(-)

diff --git a/arch/arm/kernel/perf_event.c b/arch/arm/kernel/perf_event.c
index 5ce6c33..9331d57 100644
--- a/arch/arm/kernel/perf_event.c
+++ b/arch/arm/kernel/perf_event.c
@@ -27,12 +27,6 @@
 #include <asm/stacktrace.h>
 
 /*
- * Hardware lock to serialize accesses to PMU registers. Needed for the
- * read/modify/write sequences.
- */
-static DEFINE_RAW_SPINLOCK(pmu_lock);
-
-/*
  * ARMv6 supports a maximum of 3 events, starting from index 0. If we add
  * another platform that supports more, we need to increase this to be the
  * largest of all platforms.
@@ -55,6 +49,12 @@ struct cpu_hw_events {
 	 * an event. A 0 means that the counter can be used.
 	 */
 	unsigned long		used_mask[BITS_TO_LONGS(ARMPMU_MAX_HWEVENTS)];
+
+	/*
+	 * Hardware lock to serialize accesses to PMU registers. Needed for the
+	 * read/modify/write sequences.
+	 */
+	raw_spinlock_t		pmu_lock;
 };
 static DEFINE_PER_CPU(struct cpu_hw_events, cpu_hw_events);
 
@@ -685,6 +685,11 @@ static struct cpu_hw_events *armpmu_get_cpu_events(void)
 
 static void __init cpu_pmu_init(struct arm_pmu *armpmu)
 {
+	int cpu;
+	for_each_possible_cpu(cpu) {
+		struct cpu_hw_events *events = &per_cpu(cpu_hw_events, cpu);
+		raw_spin_lock_init(&events->pmu_lock);
+	}
 	armpmu->get_hw_events = armpmu_get_cpu_events;
 }
 
diff --git a/arch/arm/kernel/perf_event_v6.c b/arch/arm/kernel/perf_event_v6.c
index 8390128..68cf704 100644
--- a/arch/arm/kernel/perf_event_v6.c
+++ b/arch/arm/kernel/perf_event_v6.c
@@ -433,6 +433,7 @@ armv6pmu_enable_event(struct hw_perf_event *hwc,
 		      int idx)
 {
 	unsigned long val, mask, evt, flags;
+	struct cpu_hw_events *events = armpmu->get_hw_events();
 
 	if (ARMV6_CYCLE_COUNTER == idx) {
 		mask	= 0;
@@ -454,12 +455,12 @@ armv6pmu_enable_event(struct hw_perf_event *hwc,
 	 * Mask out the current event and set the counter to count the event
 	 * that we're interested in.
 	 */
-	raw_spin_lock_irqsave(&pmu_lock, flags);
+	raw_spin_lock_irqsave(&events->pmu_lock, flags);
 	val = armv6_pmcr_read();
 	val &= ~mask;
 	val |= evt;
 	armv6_pmcr_write(val);
-	raw_spin_unlock_irqrestore(&pmu_lock, flags);
+	raw_spin_unlock_irqrestore(&events->pmu_lock, flags);
 }
 
 static int counter_is_active(unsigned long pmcr, int idx)
@@ -544,24 +545,26 @@ static void
 armv6pmu_start(void)
 {
 	unsigned long flags, val;
+	struct cpu_hw_events *events = armpmu->get_hw_events();
 
-	raw_spin_lock_irqsave(&pmu_lock, flags);
+	raw_spin_lock_irqsave(&events->pmu_lock, flags);
 	val = armv6_pmcr_read();
 	val |= ARMV6_PMCR_ENABLE;
 	armv6_pmcr_write(val);
-	raw_spin_unlock_irqrestore(&pmu_lock, flags);
+	raw_spin_unlock_irqrestore(&events->pmu_lock, flags);
 }
 
 static void
 armv6pmu_stop(void)
 {
 	unsigned long flags, val;
+	struct cpu_hw_events *events = armpmu->get_hw_events();
 
-	raw_spin_lock_irqsave(&pmu_lock, flags);
+	raw_spin_lock_irqsave(&events->pmu_lock, flags);
 	val = armv6_pmcr_read();
 	val &= ~ARMV6_PMCR_ENABLE;
 	armv6_pmcr_write(val);
-	raw_spin_unlock_irqrestore(&pmu_lock, flags);
+	raw_spin_unlock_irqrestore(&events->pmu_lock, flags);
 }
 
 static int
@@ -595,6 +598,7 @@ armv6pmu_disable_event(struct hw_perf_event *hwc,
 		       int idx)
 {
 	unsigned long val, mask, evt, flags;
+	struct cpu_hw_events *events = armpmu->get_hw_events();
 
 	if (ARMV6_CYCLE_COUNTER == idx) {
 		mask	= ARMV6_PMCR_CCOUNT_IEN;
@@ -615,12 +619,12 @@ armv6pmu_disable_event(struct hw_perf_event *hwc,
 	 * of ETM bus signal assertion cycles. The external reporting should
 	 * be disabled and so this should never increment.
 	 */
-	raw_spin_lock_irqsave(&pmu_lock, flags);
+	raw_spin_lock_irqsave(&events->pmu_lock, flags);
 	val = armv6_pmcr_read();
 	val &= ~mask;
 	val |= evt;
 	armv6_pmcr_write(val);
-	raw_spin_unlock_irqrestore(&pmu_lock, flags);
+	raw_spin_unlock_irqrestore(&events->pmu_lock, flags);
 }
 
 static void
@@ -628,6 +632,7 @@ armv6mpcore_pmu_disable_event(struct hw_perf_event *hwc,
 			      int idx)
 {
 	unsigned long val, mask, flags, evt = 0;
+	struct cpu_hw_events *events = armpmu->get_hw_events();
 
 	if (ARMV6_CYCLE_COUNTER == idx) {
 		mask	= ARMV6_PMCR_CCOUNT_IEN;
@@ -644,12 +649,12 @@ armv6mpcore_pmu_disable_event(struct hw_perf_event *hwc,
 	 * Unlike UP ARMv6, we don't have a way of stopping the counters. We
 	 * simply disable the interrupt reporting.
 	 */
-	raw_spin_lock_irqsave(&pmu_lock, flags);
+	raw_spin_lock_irqsave(&events->pmu_lock, flags);
 	val = armv6_pmcr_read();
 	val &= ~mask;
 	val |= evt;
 	armv6_pmcr_write(val);
-	raw_spin_unlock_irqrestore(&pmu_lock, flags);
+	raw_spin_unlock_irqrestore(&events->pmu_lock, flags);
 }
 
 static struct arm_pmu armv6pmu = {
diff --git a/arch/arm/kernel/perf_event_v7.c b/arch/arm/kernel/perf_event_v7.c
index f4170fc..68ac522 100644
--- a/arch/arm/kernel/perf_event_v7.c
+++ b/arch/arm/kernel/perf_event_v7.c
@@ -936,12 +936,13 @@ static void armv7_pmnc_dump_regs(void)
 static void armv7pmu_enable_event(struct hw_perf_event *hwc, int idx)
 {
 	unsigned long flags;
+	struct cpu_hw_events *events = armpmu->get_hw_events();
 
 	/*
 	 * Enable counter and interrupt, and set the counter to count
 	 * the event that we're interested in.
 	 */
-	raw_spin_lock_irqsave(&pmu_lock, flags);
+	raw_spin_lock_irqsave(&events->pmu_lock, flags);
 
 	/*
 	 * Disable counter
@@ -966,17 +967,18 @@ static void armv7pmu_enable_event(struct hw_perf_event *hwc, int idx)
 	 */
 	armv7_pmnc_enable_counter(idx);
 
-	raw_spin_unlock_irqrestore(&pmu_lock, flags);
+	raw_spin_unlock_irqrestore(&events->pmu_lock, flags);
 }
 
 static void armv7pmu_disable_event(struct hw_perf_event *hwc, int idx)
 {
 	unsigned long flags;
+	struct cpu_hw_events *events = armpmu->get_hw_events();
 
 	/*
 	 * Disable counter and interrupt
 	 */
-	raw_spin_lock_irqsave(&pmu_lock, flags);
+	raw_spin_lock_irqsave(&events->pmu_lock, flags);
 
 	/*
 	 * Disable counter
@@ -988,7 +990,7 @@ static void armv7pmu_disable_event(struct hw_perf_event *hwc, int idx)
 	 */
 	armv7_pmnc_disable_intens(idx);
 
-	raw_spin_unlock_irqrestore(&pmu_lock, flags);
+	raw_spin_unlock_irqrestore(&events->pmu_lock, flags);
 }
 
 static irqreturn_t armv7pmu_handle_irq(int irq_num, void *dev)
@@ -1054,21 +1056,23 @@ static irqreturn_t armv7pmu_handle_irq(int irq_num, void *dev)
 static void armv7pmu_start(void)
 {
 	unsigned long flags;
+	struct cpu_hw_events *events = armpmu->get_hw_events();
 
-	raw_spin_lock_irqsave(&pmu_lock, flags);
+	raw_spin_lock_irqsave(&events->pmu_lock, flags);
 	/* Enable all counters */
 	armv7_pmnc_write(armv7_pmnc_read() | ARMV7_PMNC_E);
-	raw_spin_unlock_irqrestore(&pmu_lock, flags);
+	raw_spin_unlock_irqrestore(&events->pmu_lock, flags);
 }
 
 static void armv7pmu_stop(void)
 {
 	unsigned long flags;
+	struct cpu_hw_events *events = armpmu->get_hw_events();
 
-	raw_spin_lock_irqsave(&pmu_lock, flags);
+	raw_spin_lock_irqsave(&events->pmu_lock, flags);
 	/* Disable all counters */
 	armv7_pmnc_write(armv7_pmnc_read() & ~ARMV7_PMNC_E);
-	raw_spin_unlock_irqrestore(&pmu_lock, flags);
+	raw_spin_unlock_irqrestore(&events->pmu_lock, flags);
 }
 
 static int armv7pmu_get_event_idx(struct cpu_hw_events *cpuc,
diff --git a/arch/arm/kernel/perf_event_xscale.c b/arch/arm/kernel/perf_event_xscale.c
index ca89a06..18e4823 100644
--- a/arch/arm/kernel/perf_event_xscale.c
+++ b/arch/arm/kernel/perf_event_xscale.c
@@ -281,6 +281,7 @@ static void
 xscale1pmu_enable_event(struct hw_perf_event *hwc, int idx)
 {
 	unsigned long val, mask, evt, flags;
+	struct cpu_hw_events *events = armpmu->get_hw_events();
 
 	switch (idx) {
 	case XSCALE_CYCLE_COUNTER:
@@ -302,18 +303,19 @@ xscale1pmu_enable_event(struct hw_perf_event *hwc, int idx)
 		return;
 	}
 
-	raw_spin_lock_irqsave(&pmu_lock, flags);
+	raw_spin_lock_irqsave(&events->pmu_lock, flags);
 	val = xscale1pmu_read_pmnc();
 	val &= ~mask;
 	val |= evt;
 	xscale1pmu_write_pmnc(val);
-	raw_spin_unlock_irqrestore(&pmu_lock, flags);
+	raw_spin_unlock_irqrestore(&events->pmu_lock, flags);
 }
 
 static void
 xscale1pmu_disable_event(struct hw_perf_event *hwc, int idx)
 {
 	unsigned long val, mask, evt, flags;
+	struct cpu_hw_events *events = armpmu->get_hw_events();
 
 	switch (idx) {
 	case XSCALE_CYCLE_COUNTER:
@@ -333,12 +335,12 @@ xscale1pmu_disable_event(struct hw_perf_event *hwc, int idx)
 		return;
 	}
 
-	raw_spin_lock_irqsave(&pmu_lock, flags);
+	raw_spin_lock_irqsave(&events->pmu_lock, flags);
 	val = xscale1pmu_read_pmnc();
 	val &= ~mask;
 	val |= evt;
 	xscale1pmu_write_pmnc(val);
-	raw_spin_unlock_irqrestore(&pmu_lock, flags);
+	raw_spin_unlock_irqrestore(&events->pmu_lock, flags);
 }
 
 static int
@@ -365,24 +367,26 @@ static void
 xscale1pmu_start(void)
 {
 	unsigned long flags, val;
+	struct cpu_hw_events *events = armpmu->get_hw_events();
 
-	raw_spin_lock_irqsave(&pmu_lock, flags);
+	raw_spin_lock_irqsave(&events->pmu_lock, flags);
 	val = xscale1pmu_read_pmnc();
 	val |= XSCALE_PMU_ENABLE;
 	xscale1pmu_write_pmnc(val);
-	raw_spin_unlock_irqrestore(&pmu_lock, flags);
+	raw_spin_unlock_irqrestore(&events->pmu_lock, flags);
 }
 
 static void
 xscale1pmu_stop(void)
 {
 	unsigned long flags, val;
+	struct cpu_hw_events *events = armpmu->get_hw_events();
 
-	raw_spin_lock_irqsave(&pmu_lock, flags);
+	raw_spin_lock_irqsave(&events->pmu_lock, flags);
 	val = xscale1pmu_read_pmnc();
 	val &= ~XSCALE_PMU_ENABLE;
 	xscale1pmu_write_pmnc(val);
-	raw_spin_unlock_irqrestore(&pmu_lock, flags);
+	raw_spin_unlock_irqrestore(&events->pmu_lock, flags);
 }
 
 static inline u32
@@ -610,6 +614,7 @@ static void
 xscale2pmu_enable_event(struct hw_perf_event *hwc, int idx)
 {
 	unsigned long flags, ien, evtsel;
+	struct cpu_hw_events *events = armpmu->get_hw_events();
 
 	ien = xscale2pmu_read_int_enable();
 	evtsel = xscale2pmu_read_event_select();
@@ -643,16 +648,17 @@ xscale2pmu_enable_event(struct hw_perf_event *hwc, int idx)
 		return;
 	}
 
-	raw_spin_lock_irqsave(&pmu_lock, flags);
+	raw_spin_lock_irqsave(&events->pmu_lock, flags);
 	xscale2pmu_write_event_select(evtsel);
 	xscale2pmu_write_int_enable(ien);
-	raw_spin_unlock_irqrestore(&pmu_lock, flags);
+	raw_spin_unlock_irqrestore(&events->pmu_lock, flags);
 }
 
 static void
 xscale2pmu_disable_event(struct hw_perf_event *hwc, int idx)
 {
 	unsigned long flags, ien, evtsel;
+	struct cpu_hw_events *events = armpmu->get_hw_events();
 
 	ien = xscale2pmu_read_int_enable();
 	evtsel = xscale2pmu_read_event_select();
@@ -686,10 +692,10 @@ xscale2pmu_disable_event(struct hw_perf_event *hwc, int idx)
 		return;
 	}
 
-	raw_spin_lock_irqsave(&pmu_lock, flags);
+	raw_spin_lock_irqsave(&events->pmu_lock, flags);
 	xscale2pmu_write_event_select(evtsel);
 	xscale2pmu_write_int_enable(ien);
-	raw_spin_unlock_irqrestore(&pmu_lock, flags);
+	raw_spin_unlock_irqrestore(&events->pmu_lock, flags);
 }
 
 static int
@@ -712,24 +718,26 @@ static void
 xscale2pmu_start(void)
 {
 	unsigned long flags, val;
+	struct cpu_hw_events *events = armpmu->get_hw_events();
 
-	raw_spin_lock_irqsave(&pmu_lock, flags);
+	raw_spin_lock_irqsave(&events->pmu_lock, flags);
 	val = xscale2pmu_read_pmnc() & ~XSCALE_PMU_CNT64;
 	val |= XSCALE_PMU_ENABLE;
 	xscale2pmu_write_pmnc(val);
-	raw_spin_unlock_irqrestore(&pmu_lock, flags);
+	raw_spin_unlock_irqrestore(&events->pmu_lock, flags);
 }
 
 static void
 xscale2pmu_stop(void)
 {
 	unsigned long flags, val;
+	struct cpu_hw_events *events = armpmu->get_hw_events();
 
-	raw_spin_lock_irqsave(&pmu_lock, flags);
+	raw_spin_lock_irqsave(&events->pmu_lock, flags);
 	val = xscale2pmu_read_pmnc();
 	val &= ~XSCALE_PMU_ENABLE;
 	xscale2pmu_write_pmnc(val);
-	raw_spin_unlock_irqrestore(&pmu_lock, flags);
+	raw_spin_unlock_irqrestore(&events->pmu_lock, flags);
 }
 
 static inline u32
-- 
1.7.0.4

^ permalink raw reply related	[flat|nested] 2+ messages in thread
* [RFC PATCH 00/15] ARM: perf: support multiple PMUs
@ 2011-08-15 13:55 Mark Rutland
  2011-08-15 13:55 ` [RFC PATCH 09/15] ARM: perf: lock PMU registers per-CPU Mark Rutland
  0 siblings, 1 reply; 2+ messages in thread
From: Mark Rutland @ 2011-08-15 13:55 UTC (permalink / raw)
  To: linux-arm-kernel

System (AKA nest or uncore) PMUs exist on devices which are not affine
to a single CPU. They usually cannot be directly associated with
individual tasks and are asynchronous with respect to the current
execution. Examples of devices which could have system PMUs include L2
cache controllers, GPUs and memory buses.

The following patch series refactors the ARM PMU backend, enabling
new PMUs to reuse the existing code. This should allow for system PMUs
to be supported in future. Further work will be required to get perf to
fully understand system PMUs, but this provides something usable.

The framework is intended to be used by system PMUs which hang off core
platform components (e.g. L2 cache, AXI bus). If a device is complex
enough or separate enough from core functionality to have its own
driver, it should implement its own PMU handling using the core perf
API directly.

The first patch ("perf: provide PMU when initing events") is currently
sitting in the tip tree, but as it's required for event initialization
to function (and hence for the PMU to be usable), it's provided here
for convenience.

The series is based on Will Deacon's perf-updates branch at:
	git://linux-arm.org/linux-2.6-wd.git perf-updates

An example driver using the framework (supporting the PMU present in
L220/PL310 level 2 cache controllers) can be found at:
	git://linux-arm.org/linux-2.6-wd.git perf-l2x0-wip

Any comments would be welcome.

Thanks,
Mark.

Mark Rutland (15):
  perf: provide PMU when initing events
  ARM: perf: only register a CPU PMU when present
  ARM: perf: clean up event group validation
  ARM: perf: remove active_mask
  ARM: perf: move active_events into struct arm_pmu
  ARM: perf: move platform device to struct arm_pmu
  ARM: perf: indirect access to cpu_hw_events
  ARM: perf: remove unnecessary armpmu->stop
  ARM: perf: lock PMU registers per-CPU
  ARM: perf: add type field to struct arm_pmu
  ARM: perf: refactor event mapping
  ARM: perf: add support for multiple PMUs
  ARM: perf: remove event limit from pmu_hw_events
  ARM: perf: remove cpu-related misnomers
  ARM: perf: move arm_pmu into <asm/pmu.h>

 arch/arm/include/asm/pmu.h          |   64 +++++++
 arch/arm/kernel/perf_event.c        |  318 +++++++++++++++++------------------
 arch/arm/kernel/perf_event_v6.c     |   73 ++++++---
 arch/arm/kernel/perf_event_v7.c     |   74 +++++---
 arch/arm/kernel/perf_event_xscale.c |   76 +++++----
 kernel/events/core.c                |    4 +-
 6 files changed, 359 insertions(+), 250 deletions(-)

^ permalink raw reply	[flat|nested] 2+ messages in thread
end of thread, other threads:[~2011-08-15 13:55 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2011-04-28  9:17 [RFC PATCH 09/15] ARM: perf: lock PMU registers per-CPU Mark Rutland
  -- strict thread matches above, loose matches on Subject: below --
2011-08-15 13:55 [RFC PATCH 00/15] ARM: perf: support multiple PMUs Mark Rutland
2011-08-15 13:55 ` [RFC PATCH 09/15] ARM: perf: lock PMU registers per-CPU Mark Rutland

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).