[PATCH RESEND 0/3] perf, amd: Support for Family 16h L2I Performance Counters

[PATCH RESEND 0/3] perf, amd: Support for Family 16h L2I Performance Counters

[Jacob Shin]

Upcoming AMD Family 16h Processors provide 4 new performance counters
to count L2 related events. Similar to northbridge counters, these new
counters are shared across multiple CPUs that share the same L2 cache.
This patchset adds support for these new counters and enforces sharing
by leveraging the existing sharing logic used for the northbridge
counters.

Jacob Shin (3):
  perf, amd: Further generalize NB event constraints handling logic
  perf, x86: Allow for multiple kfree_on_online pointers
  perf, amd: Enable L2I performance counters on AMD Family 16h

 arch/x86/include/asm/cpufeature.h      |    2 +
 arch/x86/include/asm/perf_event.h      |    4 +
 arch/x86/include/uapi/asm/msr-index.h  |    4 +
 arch/x86/kernel/cpu/perf_event.c       |    7 +-
 arch/x86/kernel/cpu/perf_event.h       |   11 +-
 arch/x86/kernel/cpu/perf_event_amd.c   |  235 +++++++++++++++++++++++++-------
 arch/x86/kernel/cpu/perf_event_intel.c |    2 +-
 7 files changed, 207 insertions(+), 58 deletions(-)

-- 
1.7.9.5

[PATCH RESEND 1/3] perf, amd: Further generalize NB event constraints handling logic

[Jacob Shin]

In preparation for enabling AMD L2I performance counters, we will
further generalize NB event constraints handling logic to now allow
any type of shared counters. This is just a code rework, there are
no functional changes.

Signed-off-by: Jacob Shin <jacob.shin@amd.com>
---
 arch/x86/kernel/cpu/perf_event.h     |    6 +--
 arch/x86/kernel/cpu/perf_event_amd.c |   70 +++++++++++++++++-----------------
 2 files changed, 38 insertions(+), 38 deletions(-)

diff --git a/arch/x86/kernel/cpu/perf_event.h b/arch/x86/kernel/cpu/perf_event.h
index ba9aadf..f092dfe 100644
--- a/arch/x86/kernel/cpu/perf_event.h
+++ b/arch/x86/kernel/cpu/perf_event.h
@@ -68,8 +68,8 @@ struct event_constraint {
 #define PERF_X86_EVENT_PEBS_LDLAT	0x1 /* ld+ldlat data address sampling */
 #define PERF_X86_EVENT_PEBS_ST		0x2 /* st data address sampling */
 
-struct amd_nb {
-	int nb_id;  /* NorthBridge id */
+struct amd_shared_regs {
+	int id;
 	int refcnt; /* reference count */
 	struct perf_event *owners[X86_PMC_IDX_MAX];
 	struct event_constraint event_constraints[X86_PMC_IDX_MAX];
@@ -170,7 +170,7 @@ struct cpu_hw_events {
 	/*
 	 * AMD specific bits
 	 */
-	struct amd_nb			*amd_nb;
+	struct amd_shared_regs		*amd_nb;
 	/* Inverted mask of bits to clear in the perf_ctr ctrl registers */
 	u64				perf_ctr_virt_mask;
 
diff --git a/arch/x86/kernel/cpu/perf_event_amd.c b/arch/x86/kernel/cpu/perf_event_amd.c
index dfdab42..23964a6 100644
--- a/arch/x86/kernel/cpu/perf_event_amd.c
+++ b/arch/x86/kernel/cpu/perf_event_amd.c
@@ -292,9 +292,9 @@ static inline int amd_is_perfctr_nb_event(struct hw_perf_event *hwc)
 
 static inline int amd_has_nb(struct cpu_hw_events *cpuc)
 {
-	struct amd_nb *nb = cpuc->amd_nb;
+	struct amd_shared_regs *nb = cpuc->amd_nb;
 
-	return nb && nb->nb_id != -1;
+	return nb && nb->id != -1;
 }
 
 static int amd_pmu_hw_config(struct perf_event *event)
@@ -321,10 +321,9 @@ static int amd_pmu_hw_config(struct perf_event *event)
 	return amd_core_hw_config(event);
 }
 
-static void __amd_put_nb_event_constraints(struct cpu_hw_events *cpuc,
-					   struct perf_event *event)
+static void amd_put_shared_event_constraints(struct amd_shared_regs *regs,
+					     struct perf_event *event)
 {
-	struct amd_nb *nb = cpuc->amd_nb;
 	int i;
 
 	/*
@@ -336,7 +335,7 @@ static void __amd_put_nb_event_constraints(struct cpu_hw_events *cpuc,
 	 * when we come here
 	 */
 	for (i = 0; i < x86_pmu.num_counters; i++) {
-		if (cmpxchg(nb->owners + i, event, NULL) == event)
+		if (cmpxchg(regs->owners + i, event, NULL) == event)
 			break;
 	}
 }
@@ -386,16 +385,17 @@ static void amd_nb_interrupt_hw_config(struct hw_perf_event *hwc)
   *
   * Given that resources are allocated (cmpxchg), they must be
   * eventually freed for others to use. This is accomplished by
-  * calling __amd_put_nb_event_constraints()
+  * calling amd_put_shared_event_constraints()
   *
   * Non NB events are not impacted by this restriction.
   */
 static struct event_constraint *
-__amd_get_nb_event_constraints(struct cpu_hw_events *cpuc, struct perf_event *event,
-			       struct event_constraint *c)
+amd_get_shared_event_constraints(struct cpu_hw_events *cpuc,
+				 struct amd_shared_regs *regs,
+				 struct perf_event *event,
+				 struct event_constraint *c)
 {
 	struct hw_perf_event *hwc = &event->hw;
-	struct amd_nb *nb = cpuc->amd_nb;
 	struct perf_event *old;
 	int idx, new = -1;
 
@@ -418,8 +418,8 @@ __amd_get_nb_event_constraints(struct cpu_hw_events *cpuc, struct perf_event *ev
 	for_each_set_bit(idx, c->idxmsk, x86_pmu.num_counters) {
 		if (new == -1 || hwc->idx == idx)
 			/* assign free slot, prefer hwc->idx */
-			old = cmpxchg(nb->owners + idx, NULL, event);
-		else if (nb->owners[idx] == event)
+			old = cmpxchg(regs->owners + idx, NULL, event);
+		else if (regs->owners[idx] == event)
 			/* event already present */
 			old = event;
 		else
@@ -430,7 +430,7 @@ __amd_get_nb_event_constraints(struct cpu_hw_events *cpuc, struct perf_event *ev
 
 		/* reassign to this slot */
 		if (new != -1)
-			cmpxchg(nb->owners + new, event, NULL);
+			cmpxchg(regs->owners + new, event, NULL);
 		new = idx;
 
 		/* already present, reuse */
@@ -444,29 +444,29 @@ __amd_get_nb_event_constraints(struct cpu_hw_events *cpuc, struct perf_event *ev
 	if (amd_is_perfctr_nb_event(hwc))
 		amd_nb_interrupt_hw_config(hwc);
 
-	return &nb->event_constraints[new];
+	return &regs->event_constraints[new];
 }
 
-static struct amd_nb *amd_alloc_nb(int cpu)
+static struct amd_shared_regs *amd_alloc_shared_regs(int cpu)
 {
-	struct amd_nb *nb;
+	struct amd_shared_regs *regs;
 	int i;
 
-	nb = kmalloc_node(sizeof(struct amd_nb), GFP_KERNEL | __GFP_ZERO,
-			  cpu_to_node(cpu));
-	if (!nb)
+	regs = kmalloc_node(sizeof(struct amd_shared_regs),
+			    GFP_KERNEL | __GFP_ZERO, cpu_to_node(cpu));
+	if (!regs)
 		return NULL;
 
-	nb->nb_id = -1;
+	regs->id = -1;
 
 	/*
-	 * initialize all possible NB constraints
+	 * initialize all possible constraints
 	 */
 	for (i = 0; i < x86_pmu.num_counters; i++) {
-		__set_bit(i, nb->event_constraints[i].idxmsk);
-		nb->event_constraints[i].weight = 1;
+		__set_bit(i, regs->event_constraints[i].idxmsk);
+		regs->event_constraints[i].weight = 1;
 	}
-	return nb;
+	return regs;
 }
 
 static int amd_pmu_cpu_prepare(int cpu)
@@ -478,7 +478,7 @@ static int amd_pmu_cpu_prepare(int cpu)
 	if (boot_cpu_data.x86_max_cores < 2)
 		return NOTIFY_OK;
 
-	cpuc->amd_nb = amd_alloc_nb(cpu);
+	cpuc->amd_nb = amd_alloc_shared_regs(cpu);
 	if (!cpuc->amd_nb)
 		return NOTIFY_BAD;
 
@@ -488,7 +488,7 @@ static int amd_pmu_cpu_prepare(int cpu)
 static void amd_pmu_cpu_starting(int cpu)
 {
 	struct cpu_hw_events *cpuc = &per_cpu(cpu_hw_events, cpu);
-	struct amd_nb *nb;
+	struct amd_shared_regs *nb;
 	int i, nb_id;
 
 	cpuc->perf_ctr_virt_mask = AMD64_EVENTSEL_HOSTONLY;
@@ -504,14 +504,14 @@ static void amd_pmu_cpu_starting(int cpu)
 		if (WARN_ON_ONCE(!nb))
 			continue;
 
-		if (nb->nb_id == nb_id) {
+		if (nb->id == nb_id) {
 			cpuc->kfree_on_online = cpuc->amd_nb;
 			cpuc->amd_nb = nb;
 			break;
 		}
 	}
 
-	cpuc->amd_nb->nb_id = nb_id;
+	cpuc->amd_nb->id = nb_id;
 	cpuc->amd_nb->refcnt++;
 }
 
@@ -525,9 +525,9 @@ static void amd_pmu_cpu_dead(int cpu)
 	cpuhw = &per_cpu(cpu_hw_events, cpu);
 
 	if (cpuhw->amd_nb) {
-		struct amd_nb *nb = cpuhw->amd_nb;
+		struct amd_shared_regs *nb = cpuhw->amd_nb;
 
-		if (nb->nb_id == -1 || --nb->refcnt == 0)
+		if (nb->id == -1 || --nb->refcnt == 0)
 			kfree(nb);
 
 		cpuhw->amd_nb = NULL;
@@ -543,15 +543,15 @@ amd_get_event_constraints(struct cpu_hw_events *cpuc, struct perf_event *event)
 	if (!(amd_has_nb(cpuc) && amd_is_nb_event(&event->hw)))
 		return &unconstrained;
 
-	return __amd_get_nb_event_constraints(cpuc, event,
-					      amd_nb_event_constraint);
+	return amd_get_shared_event_constraints(cpuc, cpuc->amd_nb, event,
+						amd_nb_event_constraint);
 }
 
 static void amd_put_event_constraints(struct cpu_hw_events *cpuc,
 				      struct perf_event *event)
 {
 	if (amd_has_nb(cpuc) && amd_is_nb_event(&event->hw))
-		__amd_put_nb_event_constraints(cpuc, event);
+		amd_put_shared_event_constraints(cpuc->amd_nb, event);
 }
 
 PMU_FORMAT_ATTR(event,	"config:0-7,32-35");
@@ -711,8 +711,8 @@ amd_get_event_constraints_f15h(struct cpu_hw_events *cpuc, struct perf_event *ev
 			return &amd_f15_PMC20;
 		}
 	case AMD_EVENT_NB:
-		return __amd_get_nb_event_constraints(cpuc, event,
-						      amd_nb_event_constraint);
+		return amd_get_shared_event_constraints(cpuc, cpuc->amd_nb,
+				event, amd_nb_event_constraint);
 	default:
 		return &emptyconstraint;
 	}
-- 
1.7.9.5

[PATCH RESEND 2/3] perf, x86: Allow for multiple kfree_on_online pointers

[Jacob Shin]

Currently only 1 pointer can be freed using kfree_on_online mechanism,
allow for multiple.

Signed-off-by: Jacob Shin <jacob.shin@amd.com>
---
 arch/x86/kernel/cpu/perf_event.c       |    7 +++++--
 arch/x86/kernel/cpu/perf_event.h       |    3 ++-
 arch/x86/kernel/cpu/perf_event_amd.c   |    2 +-
 arch/x86/kernel/cpu/perf_event_intel.c |    2 +-
 4 files changed, 9 insertions(+), 5 deletions(-)

diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c
index 5ed7a4c..62516cf 100644
--- a/arch/x86/kernel/cpu/perf_event.c
+++ b/arch/x86/kernel/cpu/perf_event.c
@@ -1258,10 +1258,12 @@ x86_pmu_notifier(struct notifier_block *self, unsigned long action, void *hcpu)
 	unsigned int cpu = (long)hcpu;
 	struct cpu_hw_events *cpuc = &per_cpu(cpu_hw_events, cpu);
 	int ret = NOTIFY_OK;
+	int i;
 
 	switch (action & ~CPU_TASKS_FROZEN) {
 	case CPU_UP_PREPARE:
-		cpuc->kfree_on_online = NULL;
+		for (i = 0; i < MAX_KFREE_ON_ONLINE; i++)
+			cpuc->kfree_on_online[i] = NULL;
 		if (x86_pmu.cpu_prepare)
 			ret = x86_pmu.cpu_prepare(cpu);
 		break;
@@ -1274,7 +1276,8 @@ x86_pmu_notifier(struct notifier_block *self, unsigned long action, void *hcpu)
 		break;
 
 	case CPU_ONLINE:
-		kfree(cpuc->kfree_on_online);
+		for (i = 0; i < MAX_KFREE_ON_ONLINE; i++)
+			kfree(cpuc->kfree_on_online[i]);
 		break;
 
 	case CPU_DYING:
diff --git a/arch/x86/kernel/cpu/perf_event.h b/arch/x86/kernel/cpu/perf_event.h
index f092dfe..9751201 100644
--- a/arch/x86/kernel/cpu/perf_event.h
+++ b/arch/x86/kernel/cpu/perf_event.h
@@ -118,6 +118,7 @@ struct intel_shared_regs {
 };
 
 #define MAX_LBR_ENTRIES		16
+#define MAX_KFREE_ON_ONLINE	4
 
 struct cpu_hw_events {
 	/*
@@ -174,7 +175,7 @@ struct cpu_hw_events {
 	/* Inverted mask of bits to clear in the perf_ctr ctrl registers */
 	u64				perf_ctr_virt_mask;
 
-	void				*kfree_on_online;
+	void				*kfree_on_online[MAX_KFREE_ON_ONLINE];
 };
 
 #define __EVENT_CONSTRAINT(c, n, m, w, o, f) {\
diff --git a/arch/x86/kernel/cpu/perf_event_amd.c b/arch/x86/kernel/cpu/perf_event_amd.c
index 23964a6..36b5162 100644
--- a/arch/x86/kernel/cpu/perf_event_amd.c
+++ b/arch/x86/kernel/cpu/perf_event_amd.c
@@ -505,7 +505,7 @@ static void amd_pmu_cpu_starting(int cpu)
 			continue;
 
 		if (nb->id == nb_id) {
-			cpuc->kfree_on_online = cpuc->amd_nb;
+			cpuc->kfree_on_online[0] = cpuc->amd_nb;
 			cpuc->amd_nb = nb;
 			break;
 		}
diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c
index e84c4ba..ef709d3 100644
--- a/arch/x86/kernel/cpu/perf_event_intel.c
+++ b/arch/x86/kernel/cpu/perf_event_intel.c
@@ -1741,7 +1741,7 @@ static void intel_pmu_cpu_starting(int cpu)
 
 			pc = per_cpu(cpu_hw_events, i).shared_regs;
 			if (pc && pc->core_id == core_id) {
-				cpuc->kfree_on_online = cpuc->shared_regs;
+				cpuc->kfree_on_online[0] = cpuc->shared_regs;
 				cpuc->shared_regs = pc;
 				break;
 			}
-- 
1.7.9.5

[PATCH RESEND 3/3] perf, amd: Enable L2I performance counters on AMD Family 16h

[Jacob Shin]

AMD Family 16h processors provide 4 new performance counters (in
addition to 4 legacy core counters, and 4 northbridge counters) for
monitoring L2 cache specific events (i.e. L2 cache misses). These 4
counters are shared between all CPUs that share the same L2 cache. We
will use the same existing event constraints handling logic to enforce
this sharing.

Signed-off-by: Jacob Shin <jacob.shin@amd.com>
---
 arch/x86/include/asm/cpufeature.h     |    2 +
 arch/x86/include/asm/perf_event.h     |    4 +
 arch/x86/include/uapi/asm/msr-index.h |    4 +
 arch/x86/kernel/cpu/perf_event.h      |    2 +
 arch/x86/kernel/cpu/perf_event_amd.c  |  167 +++++++++++++++++++++++++++++----
 5 files changed, 162 insertions(+), 17 deletions(-)

diff --git a/arch/x86/include/asm/cpufeature.h b/arch/x86/include/asm/cpufeature.h
index 93fe929..0f534af 100644
--- a/arch/x86/include/asm/cpufeature.h
+++ b/arch/x86/include/asm/cpufeature.h
@@ -168,6 +168,7 @@
 #define X86_FEATURE_TOPOEXT	(6*32+22) /* topology extensions CPUID leafs */
 #define X86_FEATURE_PERFCTR_CORE (6*32+23) /* core performance counter extensions */
 #define X86_FEATURE_PERFCTR_NB  (6*32+24) /* NB performance counter extensions */
+#define X86_FEATURE_PERFCTR_L2I (6*32+28) /* L2I performance counter extensions */
 
 /*
  * Auxiliary flags: Linux defined - For features scattered in various
@@ -311,6 +312,7 @@ extern const char * const x86_power_flags[32];
 #define cpu_has_pclmulqdq	boot_cpu_has(X86_FEATURE_PCLMULQDQ)
 #define cpu_has_perfctr_core	boot_cpu_has(X86_FEATURE_PERFCTR_CORE)
 #define cpu_has_perfctr_nb	boot_cpu_has(X86_FEATURE_PERFCTR_NB)
+#define cpu_has_perfctr_l2i	boot_cpu_has(X86_FEATURE_PERFCTR_L2I)
 #define cpu_has_cx8		boot_cpu_has(X86_FEATURE_CX8)
 #define cpu_has_cx16		boot_cpu_has(X86_FEATURE_CX16)
 #define cpu_has_eager_fpu	boot_cpu_has(X86_FEATURE_EAGER_FPU)
diff --git a/arch/x86/include/asm/perf_event.h b/arch/x86/include/asm/perf_event.h
index 57cb634..ed430ea 100644
--- a/arch/x86/include/asm/perf_event.h
+++ b/arch/x86/include/asm/perf_event.h
@@ -36,6 +36,9 @@
 #define AMD64_EVENTSEL_INT_CORE_SEL_SHIFT		37
 #define AMD64_EVENTSEL_INT_CORE_SEL_MASK		\
 	(0xFULL << AMD64_EVENTSEL_INT_CORE_SEL_SHIFT)
+#define AMD64_EVENTSEL_THREAD_MASK_SHIFT		56
+#define AMD64_EVENTSEL_THREAD_MASK_MASK			\
+	(0xFULL << AMD64_EVENTSEL_THREAD_MASK_SHIFT)
 
 #define AMD64_EVENTSEL_EVENT	\
 	(ARCH_PERFMON_EVENTSEL_EVENT | (0x0FULL << 32))
@@ -57,6 +60,7 @@
 #define AMD64_NUM_COUNTERS				4
 #define AMD64_NUM_COUNTERS_CORE				6
 #define AMD64_NUM_COUNTERS_NB				4
+#define AMD64_NUM_COUNTERS_L2I				4
 
 #define ARCH_PERFMON_UNHALTED_CORE_CYCLES_SEL		0x3c
 #define ARCH_PERFMON_UNHALTED_CORE_CYCLES_UMASK		(0x00 << 8)
diff --git a/arch/x86/include/uapi/asm/msr-index.h b/arch/x86/include/uapi/asm/msr-index.h
index bf7bb68..b575788 100644
--- a/arch/x86/include/uapi/asm/msr-index.h
+++ b/arch/x86/include/uapi/asm/msr-index.h
@@ -196,6 +196,10 @@
 #define MSR_AMD64_IBSBRTARGET		0xc001103b
 #define MSR_AMD64_IBS_REG_COUNT_MAX	8 /* includes MSR_AMD64_IBSBRTARGET */
 
+/* Fam 16h MSRs */
+#define MSR_F16H_L2I_PERF_CTL		0xc0010230
+#define MSR_F16H_L2I_PERF_CTR		0xc0010231
+
 /* Fam 15h MSRs */
 #define MSR_F15H_PERF_CTL		0xc0010200
 #define MSR_F15H_PERF_CTR		0xc0010201
diff --git a/arch/x86/kernel/cpu/perf_event.h b/arch/x86/kernel/cpu/perf_event.h
index 9751201..9297110 100644
--- a/arch/x86/kernel/cpu/perf_event.h
+++ b/arch/x86/kernel/cpu/perf_event.h
@@ -172,6 +172,8 @@ struct cpu_hw_events {
 	 * AMD specific bits
 	 */
 	struct amd_shared_regs		*amd_nb;
+	struct amd_shared_regs		*amd_l2i;
+
 	/* Inverted mask of bits to clear in the perf_ctr ctrl registers */
 	u64				perf_ctr_virt_mask;
 
diff --git a/arch/x86/kernel/cpu/perf_event_amd.c b/arch/x86/kernel/cpu/perf_event_amd.c
index 36b5162..e0fab88 100644
--- a/arch/x86/kernel/cpu/perf_event_amd.c
+++ b/arch/x86/kernel/cpu/perf_event_amd.c
@@ -132,7 +132,12 @@ static u64 amd_pmu_event_map(int hw_event)
 	return amd_perfmon_event_map[hw_event];
 }
 
+#define CONFIG1_CORE_EVENT 0
+#define CONFIG1_NB_EVENT   1
+#define CONFIG1_L2I_EVENT  2
+
 static struct event_constraint *amd_nb_event_constraint;
+static struct event_constraint *amd_l2i_event_constraint;
 
 /*
  * Previously calculated offsets
@@ -151,6 +156,9 @@ static unsigned int rdpmc_indexes[X86_PMC_IDX_MAX] __read_mostly;
  * CPUs with north bridge performance counter extensions:
  *   4 additional counters starting at 0xc0010240 each offset by 2
  *   (indexed right above either one of the above core counters)
+ *
+ * CPUs with L2I performance counter extensions:
+ *   4 additional counters starting at 0xc0010230 each offset by 2
  */
 static inline int amd_pmu_addr_offset(int index, bool eventsel)
 {
@@ -183,6 +191,18 @@ static inline int amd_pmu_addr_offset(int index, bool eventsel)
 			base = MSR_F15H_NB_PERF_CTR - x86_pmu.perfctr;
 
 		offset = base + ((index - first) << 1);
+	} else if (amd_l2i_event_constraint &&
+		   test_bit(index, amd_l2i_event_constraint->idxmsk)) {
+
+		first = find_first_bit(amd_l2i_event_constraint->idxmsk,
+				       X86_PMC_IDX_MAX);
+
+		if (eventsel)
+			base = MSR_F16H_L2I_PERF_CTL - x86_pmu.eventsel;
+		else
+			base = MSR_F16H_L2I_PERF_CTR - x86_pmu.perfctr;
+
+		offset = base + ((index - first) << 1);
 	} else if (!cpu_has_perfctr_core)
 		offset = index;
 	else
@@ -218,6 +238,13 @@ static inline int amd_pmu_rdpmc_index(int index)
 		first = find_first_bit(amd_nb_event_constraint->idxmsk,
 				       X86_PMC_IDX_MAX);
 		ret = index - first + 6;
+	} else if (amd_l2i_event_constraint &&
+	    test_bit(index, amd_l2i_event_constraint->idxmsk)) {
+
+		first = find_first_bit(amd_l2i_event_constraint->idxmsk,
+				       X86_PMC_IDX_MAX);
+
+		ret = index - first + 10;
 	} else
 		ret = index;
 
@@ -245,14 +272,14 @@ static int amd_core_hw_config(struct perf_event *event)
 }
 
 /*
- * NB counters do not support the following event select bits:
+ * NB and L2I counters do not support the following event select bits:
  *   Host/Guest only
  *   Counter mask
  *   Invert counter mask
  *   Edge detect
  *   OS/User mode
  */
-static int amd_nb_hw_config(struct perf_event *event)
+static int amd_shared_hw_config(struct perf_event *event)
 {
 	/* for NB, we only allow system wide counting mode */
 	if (is_sampling_event(event) || event->attach_state & PERF_ATTACH_TASK)
@@ -285,9 +312,22 @@ static inline int amd_is_nb_event(struct hw_perf_event *hwc)
 	return (hwc->config & 0xe0) == 0xe0;
 }
 
-static inline int amd_is_perfctr_nb_event(struct hw_perf_event *hwc)
+static inline int amd_is_perfctr_nb_event(struct perf_event *event)
 {
-	return amd_nb_event_constraint && amd_is_nb_event(hwc);
+	return amd_nb_event_constraint && amd_is_nb_event(&event->hw);
+}
+
+static inline int amd_is_perfctr_l2i_event(struct perf_event *event)
+{
+	unsigned int event_code = amd_get_event_code(&event->hw);
+
+	if (!amd_l2i_event_constraint)
+		return 0;
+
+	if (event_code >= 0x07d && event_code <= 0x07f)
+		return 1;
+
+	return event->attr.config1 == CONFIG1_L2I_EVENT;
 }
 
 static inline int amd_has_nb(struct cpu_hw_events *cpuc)
@@ -297,6 +337,13 @@ static inline int amd_has_nb(struct cpu_hw_events *cpuc)
 	return nb && nb->id != -1;
 }
 
+static inline int amd_has_l2i(struct cpu_hw_events *cpuc)
+{
+	struct amd_shared_regs *l2i = cpuc->amd_l2i;
+
+	return l2i && l2i->id != -1;
+}
+
 static int amd_pmu_hw_config(struct perf_event *event)
 {
 	int ret;
@@ -315,8 +362,8 @@ static int amd_pmu_hw_config(struct perf_event *event)
 	if (event->attr.type == PERF_TYPE_RAW)
 		event->hw.config |= event->attr.config & AMD64_RAW_EVENT_MASK;
 
-	if (amd_is_perfctr_nb_event(&event->hw))
-		return amd_nb_hw_config(event);
+	if (amd_is_perfctr_nb_event(event) || amd_is_perfctr_l2i_event(event))
+		return amd_shared_hw_config(event);
 
 	return amd_core_hw_config(event);
 }
@@ -340,8 +387,9 @@ static void amd_put_shared_event_constraints(struct amd_shared_regs *regs,
 	}
 }
 
-static void amd_nb_interrupt_hw_config(struct hw_perf_event *hwc)
+static void amd_shared_interrupt_hw_config(struct perf_event *event)
 {
+	struct hw_perf_event *hwc = &event->hw;
 	int core_id = cpu_data(smp_processor_id()).cpu_core_id;
 
 	/* deliver interrupts only to this core */
@@ -351,6 +399,13 @@ static void amd_nb_interrupt_hw_config(struct hw_perf_event *hwc)
 		hwc->config |= (u64)(core_id) <<
 			AMD64_EVENTSEL_INT_CORE_SEL_SHIFT;
 	}
+
+	/* mask out events from other cores */
+	if (amd_is_perfctr_l2i_event(event)) {
+		hwc->config |= AMD64_EVENTSEL_THREAD_MASK_MASK;
+		hwc->config &= ~(1ULL <<
+			(AMD64_EVENTSEL_THREAD_MASK_SHIFT + core_id));
+	}
 }
 
  /*
@@ -441,8 +496,8 @@ amd_get_shared_event_constraints(struct cpu_hw_events *cpuc,
 	if (new == -1)
 		return &emptyconstraint;
 
-	if (amd_is_perfctr_nb_event(hwc))
-		amd_nb_interrupt_hw_config(hwc);
+	if (amd_is_perfctr_nb_event(event) || amd_is_perfctr_l2i_event(event))
+		amd_shared_interrupt_hw_config(event);
 
 	return &regs->event_constraints[new];
 }
@@ -482,14 +537,18 @@ static int amd_pmu_cpu_prepare(int cpu)
 	if (!cpuc->amd_nb)
 		return NOTIFY_BAD;
 
+	cpuc->amd_l2i = amd_alloc_shared_regs(cpu);
+	if (!cpuc->amd_l2i)
+		return NOTIFY_BAD;
+
 	return NOTIFY_OK;
 }
 
 static void amd_pmu_cpu_starting(int cpu)
 {
 	struct cpu_hw_events *cpuc = &per_cpu(cpu_hw_events, cpu);
-	struct amd_shared_regs *nb;
-	int i, nb_id;
+	struct amd_shared_regs *nb, *l2i;
+	int i, nb_id, l2_id;
 
 	cpuc->perf_ctr_virt_mask = AMD64_EVENTSEL_HOSTONLY;
 
@@ -499,20 +558,44 @@ static void amd_pmu_cpu_starting(int cpu)
 	nb_id = amd_get_nb_id(cpu);
 	WARN_ON_ONCE(nb_id == BAD_APICID);
 
+	l2_id = cpu_data(cpu).compute_unit_id;
+
+	if (static_cpu_has(X86_FEATURE_TOPOEXT)) {
+		unsigned int eax, ebx, ecx, edx;
+		unsigned int nshared;
+		cpuid_count(0x8000001d, 2, &eax, &ebx, &ecx, &edx);
+		nshared = ((eax >> 14) & 0xfff) + 1;
+		l2_id = cpu_data(cpu).apicid - (cpu_data(cpu).apicid % nshared);
+	}
+
 	for_each_online_cpu(i) {
-		nb = per_cpu(cpu_hw_events, i).amd_nb;
-		if (WARN_ON_ONCE(!nb))
+		struct cpu_hw_events *other_cpuc = &per_cpu(cpu_hw_events, i);
+
+		nb = other_cpuc->amd_nb;
+		l2i = other_cpuc->amd_l2i;
+
+		if (WARN_ON_ONCE(!(nb && l2i)))
 			continue;
 
 		if (nb->id == nb_id) {
-			cpuc->kfree_on_online[0] = cpuc->amd_nb;
-			cpuc->amd_nb = nb;
-			break;
+			if (!cpuc->kfree_on_online[0]) {
+				cpuc->kfree_on_online[0] = cpuc->amd_nb;
+				cpuc->amd_nb = nb;
+			}
+
+			if (l2i->id == l2_id) {
+				cpuc->kfree_on_online[1] = cpuc->amd_l2i;
+				cpuc->amd_l2i = l2i;
+				break;
+			}
 		}
 	}
 
 	cpuc->amd_nb->id = nb_id;
 	cpuc->amd_nb->refcnt++;
+
+	cpuc->amd_l2i->id = l2_id;
+	cpuc->amd_l2i->refcnt++;
 }
 
 static void amd_pmu_cpu_dead(int cpu)
@@ -532,6 +615,15 @@ static void amd_pmu_cpu_dead(int cpu)
 
 		cpuhw->amd_nb = NULL;
 	}
+
+	if (cpuhw->amd_l2i) {
+		struct amd_shared_regs *l2i = cpuhw->amd_l2i;
+
+		if (l2i->id == -1 || --l2i->refcnt == 0)
+			kfree(l2i);
+
+		cpuhw->amd_l2i = NULL;
+	}
 }
 
 static struct event_constraint *
@@ -550,8 +642,12 @@ amd_get_event_constraints(struct cpu_hw_events *cpuc, struct perf_event *event)
 static void amd_put_event_constraints(struct cpu_hw_events *cpuc,
 				      struct perf_event *event)
 {
-	if (amd_has_nb(cpuc) && amd_is_nb_event(&event->hw))
+	struct hw_perf_event *hwc = &event->hw;
+
+	if (amd_has_nb(cpuc) && amd_is_nb_event(hwc))
 		amd_put_shared_event_constraints(cpuc->amd_nb, event);
+	else if (amd_has_l2i(cpuc) && amd_is_perfctr_l2i_event(event))
+		amd_put_shared_event_constraints(cpuc->amd_l2i, event);
 }
 
 PMU_FORMAT_ATTR(event,	"config:0-7,32-35");
@@ -718,6 +814,25 @@ amd_get_event_constraints_f15h(struct cpu_hw_events *cpuc, struct perf_event *ev
 	}
 }
 
+static struct event_constraint amd_f16_PMC30 = EVENT_CONSTRAINT(0, 0x0F, 0);
+
+static struct event_constraint amd_L2IPMC = EVENT_CONSTRAINT(0, 0xF00, 0);
+
+static struct event_constraint *
+amd_get_event_constraints_f16h(struct cpu_hw_events *cpuc,
+			       struct perf_event    *event)
+{
+	if (amd_is_perfctr_l2i_event(event))
+		return amd_get_shared_event_constraints(cpuc, cpuc->amd_l2i,
+				event, amd_l2i_event_constraint);
+
+	if (amd_is_perfctr_nb_event(event))
+		return amd_get_shared_event_constraints(cpuc, cpuc->amd_nb,
+				event, amd_nb_event_constraint);
+
+	return &amd_f16_PMC30;
+}
+
 static ssize_t amd_event_sysfs_show(char *page, u64 config)
 {
 	u64 event = (config & ARCH_PERFMON_EVENTSEL_EVENT) |
@@ -762,6 +877,9 @@ static int setup_event_constraints(void)
 {
 	if (boot_cpu_data.x86 == 0x15)
 		x86_pmu.get_event_constraints = amd_get_event_constraints_f15h;
+	else if (boot_cpu_data.x86 == 0x16)
+		x86_pmu.get_event_constraints = amd_get_event_constraints_f16h;
+
 	return 0;
 }
 
@@ -807,6 +925,20 @@ static int setup_perfctr_nb(void)
 	return 0;
 }
 
+static int setup_perfctr_l2i(void)
+{
+	if (!cpu_has_perfctr_l2i)
+		return -ENODEV;
+
+	x86_pmu.num_counters += AMD64_NUM_COUNTERS_L2I;
+
+	amd_l2i_event_constraint = &amd_L2IPMC;
+
+	printk(KERN_INFO "perf: AMD L2I performance counters detected\n");
+
+	return 0;
+}
+
 __init int amd_pmu_init(void)
 {
 	/* Performance-monitoring supported from K7 and later: */
@@ -818,6 +950,7 @@ __init int amd_pmu_init(void)
 	setup_event_constraints();
 	setup_perfctr_core();
 	setup_perfctr_nb();
+	setup_perfctr_l2i();
 
 	/* Events are common for all AMDs */
 	memcpy(hw_cache_event_ids, amd_hw_cache_event_ids,
-- 
1.7.9.5

Re: [PATCH RESEND 0/3] perf, amd: Support for Family 16h L2I Performance Counters

[Peter Zijlstra]

On Tue, 2013-04-09 at 10:23 -0500, Jacob Shin wrote:
> Upcoming AMD Family 16h Processors provide 4 new performance counters
> to count L2 related events. Similar to northbridge counters, these new
> counters are shared across multiple CPUs that share the same L2 cache.
> This patchset adds support for these new counters and enforces sharing
> by leveraging the existing sharing logic used for the northbridge
> counters.

If they're separate counters -- not shared with the regular cpu
counters like the 10h NB counters are, then they should have their own
PMU driver.

Similar to the 15h NB counters; which are a separate set of counters
and no longer overlay the normal counters.

Re: [PATCH RESEND 0/3] perf, amd: Support for Family 16h L2I Performance Counters

[Peter Zijlstra]

On Wed, 2013-04-10 at 11:41 +0200, Peter Zijlstra wrote:
> On Tue, 2013-04-09 at 10:23 -0500, Jacob Shin wrote:
> > Upcoming AMD Family 16h Processors provide 4 new performance counters
> > to count L2 related events. Similar to northbridge counters, these new
> > counters are shared across multiple CPUs that share the same L2 cache.
> > This patchset adds support for these new counters and enforces sharing
> > by leveraging the existing sharing logic used for the northbridge
> > counters.
> 
> If they're separate counters -- not shared with the regular cpu
> counters like the 10h NB counters are, then they should have their own
> PMU driver.
> 
> Similar to the 15h NB counters; which are a separate set of counters
> and no longer overlay the normal counters.

Related; where's the 16h BKDG?

Re: [PATCH RESEND 0/3] perf, amd: Support for Family 16h L2I Performance Counters

[Stephane Eranian]

On Wed, Apr 10, 2013 at 11:41 AM, Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:
>
> On Tue, 2013-04-09 at 10:23 -0500, Jacob Shin wrote:
> > Upcoming AMD Family 16h Processors provide 4 new performance counters
> > to count L2 related events. Similar to northbridge counters, these new
> > counters are shared across multiple CPUs that share the same L2 cache.
> > This patchset adds support for these new counters and enforces sharing
> > by leveraging the existing sharing logic used for the northbridge
> > counters.
>
> If they're separate counters -- not shared with the regular cpu
> counters like the 10h NB counters are, then they should have their own
> PMU driver.
>
> Similar to the 15h NB counters; which are a separate set of counters
> and no longer overlay the normal counters.
>
Well, that's how this was suggested but that's not how it's
implemented currently
and committed if I recall.

Re: [PATCH RESEND 0/3] perf, amd: Support for Family 16h L2I Performance Counters

[Peter Zijlstra]

On Wed, 2013-04-10 at 13:38 +0200, Stephane Eranian wrote:
> On Wed, Apr 10, 2013 at 11:41 AM, Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:
> >
> > On Tue, 2013-04-09 at 10:23 -0500, Jacob Shin wrote:
> > > Upcoming AMD Family 16h Processors provide 4 new performance counters
> > > to count L2 related events. Similar to northbridge counters, these new
> > > counters are shared across multiple CPUs that share the same L2 cache.
> > > This patchset adds support for these new counters and enforces sharing
> > > by leveraging the existing sharing logic used for the northbridge
> > > counters.
> >
> > If they're separate counters -- not shared with the regular cpu
> > counters like the 10h NB counters are, then they should have their own
> > PMU driver.
> >
> > Similar to the 15h NB counters; which are a separate set of counters
> > and no longer overlay the normal counters.
> >
> Well, that's how this was suggested but that's not how it's
> implemented currently
> and committed if I recall.

Hmm.. the 15h old interface bits got merged? I thought I kept telling
that should be done like the intel uncore stuff since the hardware
interface wasn't retarded anymore.

Re: [PATCH RESEND 0/3] perf, amd: Support for Family 16h L2I Performance Counters

[Stephane Eranian]

On Wed, Apr 10, 2013 at 1:49 PM, Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:
> On Wed, 2013-04-10 at 13:38 +0200, Stephane Eranian wrote:
>> On Wed, Apr 10, 2013 at 11:41 AM, Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:
>> >
>> > On Tue, 2013-04-09 at 10:23 -0500, Jacob Shin wrote:
>> > > Upcoming AMD Family 16h Processors provide 4 new performance counters
>> > > to count L2 related events. Similar to northbridge counters, these new
>> > > counters are shared across multiple CPUs that share the same L2 cache.
>> > > This patchset adds support for these new counters and enforces sharing
>> > > by leveraging the existing sharing logic used for the northbridge
>> > > counters.
>> >
>> > If they're separate counters -- not shared with the regular cpu
>> > counters like the 10h NB counters are, then they should have their own
>> > PMU driver.
>> >
>> > Similar to the 15h NB counters; which are a separate set of counters
>> > and no longer overlay the normal counters.
>> >
>> Well, that's how this was suggested but that's not how it's
>> implemented currently
>> and committed if I recall.
>
> Hmm.. the 15h old interface bits got merged? I thought I kept telling
> that should be done like the intel uncore stuff since the hardware
> interface wasn't retarded anymore.
>
I remember. I agreed with you on this. I don't recall the details as to
why Robert did not like that but I think it had to do with code duplication
from the perf_event_amd.c.

Re: [PATCH RESEND 0/3] perf, amd: Support for Family 16h L2I Performance Counters

[Peter Zijlstra]

On Wed, 2013-04-10 at 13:49 +0200, Peter Zijlstra wrote:

> Hmm.. the 15h old interface bits got merged? I thought I kept telling
> that should be done like the intel uncore stuff since the hardware
> interface wasn't retarded anymore.

Ah well, that crap seems to have slipped in in Feb when I was still a
near vegetable and not paying much attention.

/me curses a bit.

I don't suppose we can deprecate it and remove this stuff? 

Re: [PATCH RESEND 0/3] perf, amd: Support for Family 16h L2I Performance Counters

[Ingo Molnar]

* Peter Zijlstra <peterz@infradead.org> wrote:

> On Wed, 2013-04-10 at 13:49 +0200, Peter Zijlstra wrote:
> 
> > Hmm.. the 15h old interface bits got merged? I thought I kept telling
> > that should be done like the intel uncore stuff since the hardware
> > interface wasn't retarded anymore.
> 
> Ah well, that crap seems to have slipped in in Feb when I was still a
> near vegetable and not paying much attention.
> 
> /me curses a bit.
> 
> I don't suppose we can deprecate it and remove this stuff? 

I think we can - if there's a functional replacement.

Thanks,

	Ingo

Re: [PATCH RESEND 0/3] perf, amd: Support for Family 16h L2I Performance Counters

[Stephane Eranian]

On Wed, Apr 10, 2013 at 1:56 PM, Ingo Molnar <mingo@kernel.org> wrote:
>
> * Peter Zijlstra <peterz@infradead.org> wrote:
>
>> On Wed, 2013-04-10 at 13:49 +0200, Peter Zijlstra wrote:
>>
>> > Hmm.. the 15h old interface bits got merged? I thought I kept telling
>> > that should be done like the intel uncore stuff since the hardware
>> > interface wasn't retarded anymore.
>>
>> Ah well, that crap seems to have slipped in in Feb when I was still a
>> near vegetable and not paying much attention.
>>
>> /me curses a bit.
>>
>> I don't suppose we can deprecate it and remove this stuff?
>
> I think we can - if there's a functional replacement.
>
Does the existing code expose a type in sysfs?
If not then you cannot do this transparently, I am afraid
because the syntax would be different, i.e., not cpu/...

I reviewed the code and tested it. But at the time, I thought
you had agreed on the approach used.

Re: [PATCH RESEND 0/3] perf, amd: Support for Family 16h L2I Performance Counters

[Ingo Molnar]

* Stephane Eranian <eranian@google.com> wrote:

> On Wed, Apr 10, 2013 at 1:56 PM, Ingo Molnar <mingo@kernel.org> wrote:
> >
> > * Peter Zijlstra <peterz@infradead.org> wrote:
> >
> >> On Wed, 2013-04-10 at 13:49 +0200, Peter Zijlstra wrote:
> >>
> >> > Hmm.. the 15h old interface bits got merged? I thought I kept telling
> >> > that should be done like the intel uncore stuff since the hardware
> >> > interface wasn't retarded anymore.
> >>
> >> Ah well, that crap seems to have slipped in in Feb when I was still a
> >> near vegetable and not paying much attention.
> >>
> >> /me curses a bit.
> >>
> >> I don't suppose we can deprecate it and remove this stuff?
> >
> > I think we can - if there's a functional replacement.
> >
> Does the existing code expose a type in sysfs?
> If not then you cannot do this transparently, I am afraid
> because the syntax would be different, i.e., not cpu/...

That could be compatibility-bridged over in tooling?

I doubt these events are in heavy use.

> I reviewed the code and tested it. But at the time, I thought
> you had agreed on the approach used.

I did - but PeterZ has a point, so it would be nice if we could improve on that.

Thanks,

	Ingo

Re: [PATCH RESEND 0/3] perf, amd: Support for Family 16h L2I Performance Counters

[Borislav Petkov]

On Wed, Apr 10, 2013 at 02:12:42PM +0200, Stephane Eranian wrote:
> >> Ah well, that crap seems to have slipped in in Feb when I was still a
> >> near vegetable and not paying much attention.
> >>
> >> /me curses a bit.
> >>
> >> I don't suppose we can deprecate it and remove this stuff?
> >
> > I think we can - if there's a functional replacement.
> >
> Does the existing code expose a type in sysfs?
> If not then you cannot do this transparently, I am afraid
> because the syntax would be different, i.e., not cpu/...
> 
> I reviewed the code and tested it. But at the time, I thought
> you had agreed on the approach used.

Btw, Robert is away this week. I'll CC him - he might still read email.

-- 
Regards/Gruss,
    Boris.

Sent from a fat crate under my desk. Formatting is fine.
--

Re: [PATCH RESEND 0/3] perf, amd: Support for Family 16h L2I Performance Counters

[Stephane Eranian]

On Wed, Apr 10, 2013 at 2:28 PM, Ingo Molnar <mingo@kernel.org> wrote:
>
> * Stephane Eranian <eranian@google.com> wrote:
>
>> On Wed, Apr 10, 2013 at 1:56 PM, Ingo Molnar <mingo@kernel.org> wrote:
>> >
>> > * Peter Zijlstra <peterz@infradead.org> wrote:
>> >
>> >> On Wed, 2013-04-10 at 13:49 +0200, Peter Zijlstra wrote:
>> >>
>> >> > Hmm.. the 15h old interface bits got merged? I thought I kept telling
>> >> > that should be done like the intel uncore stuff since the hardware
>> >> > interface wasn't retarded anymore.
>> >>
>> >> Ah well, that crap seems to have slipped in in Feb when I was still a
>> >> near vegetable and not paying much attention.
>> >>
>> >> /me curses a bit.
>> >>
>> >> I don't suppose we can deprecate it and remove this stuff?
>> >
>> > I think we can - if there's a functional replacement.
>> >
>> Does the existing code expose a type in sysfs?
>> If not then you cannot do this transparently, I am afraid
>> because the syntax would be different, i.e., not cpu/...
>
> That could be compatibility-bridged over in tooling?
>
> I doubt these events are in heavy use.
>
I agree.

>> I reviewed the code and tested it. But at the time, I thought
>> you had agreed on the approach used.
>
> I did - but PeterZ has a point, so it would be nice if we could improve on that.
>
I think it is doable by cloning some of the functions into an amd_unc.c file.

Re: [PATCH RESEND 0/3] perf, amd: Support for Family 16h L2I Performance Counters

[Jacob Shin]

On Wed, Apr 10, 2013 at 02:29:44PM +0200, Stephane Eranian wrote:
> On Wed, Apr 10, 2013 at 2:28 PM, Ingo Molnar <mingo@kernel.org> wrote:
> >
> > * Stephane Eranian <eranian@google.com> wrote:
> >
> >> On Wed, Apr 10, 2013 at 1:56 PM, Ingo Molnar <mingo@kernel.org> wrote:
> >> >
> >> > * Peter Zijlstra <peterz@infradead.org> wrote:
> >> >
> >> >> On Wed, 2013-04-10 at 13:49 +0200, Peter Zijlstra wrote:
> >> >>
> >> >> > Hmm.. the 15h old interface bits got merged? I thought I kept telling
> >> >> > that should be done like the intel uncore stuff since the hardware
> >> >> > interface wasn't retarded anymore.
> >> >>
> >> >> Ah well, that crap seems to have slipped in in Feb when I was still a
> >> >> near vegetable and not paying much attention.
> >> >>
> >> >> /me curses a bit.
> >> >>
> >> >> I don't suppose we can deprecate it and remove this stuff?
> >> >
> >> > I think we can - if there's a functional replacement.
> >> >
> >> Does the existing code expose a type in sysfs?

It does not.

> >> If not then you cannot do this transparently, I am afraid
> >> because the syntax would be different, i.e., not cpu/...
> >
> > That could be compatibility-bridged over in tooling?
> >
> > I doubt these events are in heavy use.
> >
> I agree.

Right, I don't think they are.

> 
> >> I reviewed the code and tested it. But at the time, I thought
> >> you had agreed on the approach used.
> >
> > I did - but PeterZ has a point, so it would be nice if we could improve on that.
> >
> I think it is doable by cloning some of the functions into an amd_unc.c file.
> 

Sigh .. okay ....