Skip to contentSkip to conversation

perf/x86/intel: Fix PEBS memory access info encoding for ADL

PATCH Applied [email protected] 2 participants 4 messages Jun 29, 2022 – Jul 6, 2022

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH v2 1/2 perf/x86/intel: Fix PEBS memory access info encoding for ADL

From: Kan Liang <[email protected]>

The PEBS memory access latency encoding for the e-core is slightly different from the p-core. The bit 4 is Lock, while the bit 5 is TLB access.

Add a new flag to indicate the load/store latency event on a hybrid platform.
Add a new function pointer to retrieve the latency data for a hybrid platform. Only implement the new flag and function for the e-core on ADL. Still use the existing PERF_X86_EVENT_PEBS_LDLAT/STLAT flag for the p-core on ADL.

Factor out pebs_set_tlb_lock() to set the generic memory data source information of the TLB access and lock for both load and store latency.

Move the intel_get_event_constraints() to ahead of the :ppp check, otherwise the new flag never gets a chance to be set for the :ppp events.

Fixes f83d2f91d259 ("perf/x86/intel: Add Alder Lake Hybrid support")

---

The V1 can be found at
https://lore.kernel.org/lkml/[email protected]/

Changes since V1:
- Fix the issue that the flag of :ppp load and store latency event is overwritten
- Add a new flag and a new function pointer to indicate and handle the load/store latency event on a hybrid platform. The e-core and p-eore have different data source encoding. I think it should be a cleaner way to handle them separately, rather than add multiple if...else in a single function in the V1 series.
It's also easier to be extended if the data source format is changed again later.
- Add Reviewed-by from Andi

 arch/x86/events/intel/core.c |  5 ++-
 arch/x86/events/intel/ds.c   | 80 ++++++++++++++++++++++--------------
 arch/x86/events/perf_event.h |  8 ++++
 3 files changed, 60 insertions(+), 33 deletions(-)
arch/x86/events/intel/core.c +3 -2
index 759ef244786a..b8788ab30159 100644
@@ -4141,6 +4141,8 @@ tnt_get_event_constraints(struct cpu_hw_events *cpuc, int idx,
 {
 	struct event_constraint *c;
 
+	c = intel_get_event_constraints(cpuc, idx, event);
+
 	/*
 	 * :ppp means to do reduced skid PEBS,
 	 * which is available on PMC0 and fixed counter 0.
@@ -4153,8 +4155,6 @@ tnt_get_event_constraints(struct cpu_hw_events *cpuc, int idx,
 		return &counter0_constraint;
 	}
 
-	c = intel_get_event_constraints(cpuc, idx, event);
-
 	return c;
 }
 
@@ -6242,6 +6242,7 @@ __init int intel_pmu_init(void)
 		x86_pmu.flags |= PMU_FL_MEM_LOADS_AUX;
 		x86_pmu.lbr_pt_coexist = true;
 		intel_pmu_pebs_data_source_skl(false);
+		x86_pmu.pebs_latency_data = adl_latency_data_small;
 		x86_pmu.num_topdown_events = 8;
 		x86_pmu.update_topdown_event = adl_update_topdown_event;
 		x86_pmu.set_topdown_event_period = adl_set_topdown_event_period;
arch/x86/events/intel/ds.c +49 -31

111 lines not on this page · raw message

arch/x86/events/perf_event.h +8 -0
index 21a5482bcf84..ff6dd189739e 100644
@@ -84,6 +84,7 @@ static inline bool constraint_match(struct event_constraint *c, u64 ecode)
 #define PERF_X86_EVENT_TOPDOWN		0x04000 /* Count Topdown slots/metrics events */
 #define PERF_X86_EVENT_PEBS_STLAT	0x08000 /* st+stlat data address sampling */
 #define PERF_X86_EVENT_AMD_BRS		0x10000 /* AMD Branch Sampling */
+#define PERF_X86_EVENT_PEBS_LAT_HYBRID	0x20000 /* ld and st lat for hybrid */
 
 static inline bool is_topdown_count(struct perf_event *event)
 {
@@ -460,6 +461,10 @@ struct cpu_hw_events {
 	__EVENT_CONSTRAINT(c, n, INTEL_ARCH_EVENT_MASK|X86_ALL_EVENT_FLAGS, \
 			  HWEIGHT(n), 0, PERF_X86_EVENT_PEBS_ST)
 
+#define INTEL_HYBRID_LAT_CONSTRAINT(c, n)	\
+	__EVENT_CONSTRAINT(c, n, INTEL_ARCH_EVENT_MASK|X86_ALL_EVENT_FLAGS, \
+			  HWEIGHT(n), 0, PERF_X86_EVENT_PEBS_LAT_HYBRID)
+
 /* Event constraint, but match on all event flags too. */
 #define INTEL_FLAGS_EVENT_CONSTRAINT(c, n) \
 	EVENT_CONSTRAINT(c, n, ARCH_PERFMON_EVENTSEL_EVENT|X86_ALL_EVENT_FLAGS)
@@ -825,6 +830,7 @@ struct x86_pmu {
 	void		(*drain_pebs)(struct pt_regs *regs, struct perf_sample_data *data);
 	struct event_constraint *pebs_constraints;
 	void		(*pebs_aliases)(struct perf_event *event);
+	u64		(*pebs_latency_data)(struct perf_event *event, u64 status);
 	unsigned long	large_pebs_flags;
 	u64		rtm_abort_event;
 
@@ -1392,6 +1398,8 @@ void intel_pmu_disable_bts(void);
 
 int intel_pmu_drain_bts_buffer(void);
 
+u64 adl_latency_data_small(struct perf_event *event, u64 status);
+
 extern struct event_constraint intel_core2_pebs_event_constraints[];
 
 extern struct event_constraint intel_atom_pebs_event_constraints[];
Signature · 1 line
2.35.1
#1 lore raw

PATCH v2 2/2 perf/x86/intel: Fix PEBS data source encoding for ADL

From: Kan Liang <[email protected]>

The PEBS data source encoding for the e-core is different from the p-core.

Add the pebs_data_source[] in the struct x86_hybrid_pmu to store the data source encoding for each type of the core.

Add intel_pmu_pebs_data_source_grt() for the e-core. There is nothing changed for the data source encoding of the p-core, which still reuse the intel_pmu_pebs_data_source_skl().

Fixes f83d2f91d259 ("perf/x86/intel: Add Alder Lake Hybrid support")

---

The V1 can be found at
https://lore.kernel.org/lkml/[email protected]/

Changes since V1:
- Rebase on patch 1 accordingly
- Add Reviewed-by from Andi

 arch/x86/events/intel/core.c |  2 +-
 arch/x86/events/intel/ds.c   | 51 +++++++++++++++++++++++++++---------
 arch/x86/events/perf_event.h |  6 +++++
 3 files changed, 45 insertions(+), 14 deletions(-)
arch/x86/events/intel/core.c +1 -1
index b8788ab30159..7638ed3672fa 100644
@@ -6241,7 +6241,7 @@ __init int intel_pmu_init(void)
 		x86_pmu.flags |= PMU_FL_INSTR_LATENCY;
 		x86_pmu.flags |= PMU_FL_MEM_LOADS_AUX;
 		x86_pmu.lbr_pt_coexist = true;
-		intel_pmu_pebs_data_source_skl(false);
+		intel_pmu_pebs_data_source_adl();
 		x86_pmu.pebs_latency_data = adl_latency_data_small;
 		x86_pmu.num_topdown_events = 8;
 		x86_pmu.update_topdown_event = adl_update_topdown_event;
arch/x86/events/intel/ds.c +38 -13

97 lines not on this page · raw message

arch/x86/events/perf_event.h +6 -0
index ff6dd189739e..821098aebf78 100644
@@ -643,6 +643,8 @@ enum {
 	x86_lbr_exclusive_max,
 };
 
+#define PERF_PEBS_DATA_SOURCE_MAX	0x10
+
 struct x86_hybrid_pmu {
 	struct pmu			pmu;
 	const char			*name;
@@ -670,6 +672,8 @@ struct x86_hybrid_pmu {
 	unsigned int			late_ack	:1,
 					mid_ack		:1,
 					enabled_ack	:1;
+
+	u64				pebs_data_source[PERF_PEBS_DATA_SOURCE_MAX];
 };
 
 static __always_inline struct x86_hybrid_pmu *hybrid_pmu(struct pmu *pmu)
@@ -1507,6 +1511,8 @@ void intel_pmu_pebs_data_source_nhm(void);
 
 void intel_pmu_pebs_data_source_skl(bool pmem);
 
+void intel_pmu_pebs_data_source_adl(void);
+
 int intel_pmu_setup_lbr_filter(struct perf_event *event);
 
 void intel_pt_interrupt(void);
Signature · 1 line
2.35.1
merge bot Applied
#2 lore raw

[tip: perf/core] perf/x86/intel: Fix PEBS data source encoding for ADL

The following commit has been merged into the perf/core branch of tip:

The following commit has been merged into the perf/core branch of tip:

Commit-ID:     ccf170e9d8fdacfe435bbe3749c897c7d86d32f8
Gitweb:        https://git.kernel.org/tip/ccf170e9d8fdacfe435bbe3749c897c7d86d32f8
Author:        Kan Liang <[email protected]>
AuthorDate:    Wed, 29 Jun 2022 08:08:40 -07:00
Committer:     Peter Zijlstra <[email protected]>
CommitterDate: Mon, 04 Jul 2022 09:23:09 +02:00

perf/x86/intel: Fix PEBS data source encoding for ADL

The PEBS data source encoding for the e-core is different from the p-core.

Add the pebs_data_source[] in the struct x86_hybrid_pmu to store the data source encoding for each type of the core.

Add intel_pmu_pebs_data_source_grt() for the e-core. There is nothing changed for the data source encoding of the p-core, which still reuse the intel_pmu_pebs_data_source_skl().

Fixes f83d2f91d259 ("perf/x86/intel: Add Alder Lake Hybrid support")
---
 arch/x86/events/intel/core.c |  2 +-
 arch/x86/events/intel/ds.c   | 51 ++++++++++++++++++++++++++---------
 arch/x86/events/perf_event.h |  6 ++++-
 3 files changed, 45 insertions(+), 14 deletions(-)
arch/x86/events/intel/core.c +1 -1
index 07d4a5f..bd8b988 100644
@@ -6241,7 +6241,7 @@ __init int intel_pmu_init(void)
 		x86_pmu.flags |= PMU_FL_INSTR_LATENCY;
 		x86_pmu.flags |= PMU_FL_MEM_LOADS_AUX;
 		x86_pmu.lbr_pt_coexist = true;
-		intel_pmu_pebs_data_source_skl(false);
+		intel_pmu_pebs_data_source_adl();
 		x86_pmu.pebs_latency_data = adl_latency_data_small;
 		x86_pmu.num_topdown_events = 8;
 		x86_pmu.update_topdown_event = adl_update_topdown_event;
arch/x86/events/intel/ds.c +38 -13

97 lines not on this page · raw message

arch/x86/events/perf_event.h +6 -0
index 2d11445..ca2f8bf 100644
@@ -644,6 +644,8 @@ enum {
 	x86_lbr_exclusive_max,
 };
 
+#define PERF_PEBS_DATA_SOURCE_MAX	0x10
+
 struct x86_hybrid_pmu {
 	struct pmu			pmu;
 	const char			*name;
@@ -671,6 +673,8 @@ struct x86_hybrid_pmu {
 	unsigned int			late_ack	:1,
 					mid_ack		:1,
 					enabled_ack	:1;
+
+	u64				pebs_data_source[PERF_PEBS_DATA_SOURCE_MAX];
 };
 
 static __always_inline struct x86_hybrid_pmu *hybrid_pmu(struct pmu *pmu)
@@ -1508,6 +1512,8 @@ void intel_pmu_pebs_data_source_nhm(void);
 
 void intel_pmu_pebs_data_source_skl(bool pmem);
 
+void intel_pmu_pebs_data_source_adl(void);
+
 int intel_pmu_setup_lbr_filter(struct perf_event *event);
 
 void intel_pt_interrupt(void);
merge bot Applied
#3 lore raw
In reply to [email protected]
The following commit has been merged into the perf/core branch of tip:

The following commit has been merged into the perf/core branch of tip:

Commit-ID:     39a41278f041e4b7ee6c83caefac845c9b19fc61
Gitweb:        https://git.kernel.org/tip/39a41278f041e4b7ee6c83caefac845c9b19fc61
Author:        Kan Liang <[email protected]>
AuthorDate:    Wed, 29 Jun 2022 08:08:39 -07:00
Committer:     Peter Zijlstra <[email protected]>
CommitterDate: Mon, 04 Jul 2022 09:23:09 +02:00

perf/x86/intel: Fix PEBS memory access info encoding for ADL

The PEBS memory access latency encoding for the e-core is slightly different from the p-core. The bit 4 is Lock, while the bit 5 is TLB access.

Add a new flag to indicate the load/store latency event on a hybrid platform.
Add a new function pointer to retrieve the latency data for a hybrid platform. Only implement the new flag and function for the e-core on ADL. Still use the existing PERF_X86_EVENT_PEBS_LDLAT/STLAT flag for the p-core on ADL.

Factor out pebs_set_tlb_lock() to set the generic memory data source information of the TLB access and lock for both load and store latency.

Move the intel_get_event_constraints() to ahead of the :ppp check, otherwise the new flag never gets a chance to be set for the :ppp events.

Fixes f83d2f91d259 ("perf/x86/intel: Add Alder Lake Hybrid support")
---
 arch/x86/events/intel/core.c |  5 +-
 arch/x86/events/intel/ds.c   | 80 +++++++++++++++++++++--------------
 arch/x86/events/perf_event.h |  8 ++++-
 3 files changed, 60 insertions(+), 33 deletions(-)
arch/x86/events/intel/core.c +3 -2
index 45024ab..07d4a5f 100644
@@ -4141,6 +4141,8 @@ tnt_get_event_constraints(struct cpu_hw_events *cpuc, int idx,
 {
 	struct event_constraint *c;
 
+	c = intel_get_event_constraints(cpuc, idx, event);
+
 	/*
 	 * :ppp means to do reduced skid PEBS,
 	 * which is available on PMC0 and fixed counter 0.
@@ -4153,8 +4155,6 @@ tnt_get_event_constraints(struct cpu_hw_events *cpuc, int idx,
 		return &counter0_constraint;
 	}
 
-	c = intel_get_event_constraints(cpuc, idx, event);
-
 	return c;
 }
 
@@ -6242,6 +6242,7 @@ __init int intel_pmu_init(void)
 		x86_pmu.flags |= PMU_FL_MEM_LOADS_AUX;
 		x86_pmu.lbr_pt_coexist = true;
 		intel_pmu_pebs_data_source_skl(false);
+		x86_pmu.pebs_latency_data = adl_latency_data_small;
 		x86_pmu.num_topdown_events = 8;
 		x86_pmu.update_topdown_event = adl_update_topdown_event;
 		x86_pmu.set_topdown_event_period = adl_set_topdown_event_period;
arch/x86/events/intel/ds.c +49 -31

111 lines not on this page · raw message

arch/x86/events/perf_event.h +8 -0
index 1ca6200..2d11445 100644
@@ -84,6 +84,7 @@ static inline bool constraint_match(struct event_constraint *c, u64 ecode)
 #define PERF_X86_EVENT_TOPDOWN		0x04000 /* Count Topdown slots/metrics events */
 #define PERF_X86_EVENT_PEBS_STLAT	0x08000 /* st+stlat data address sampling */
 #define PERF_X86_EVENT_AMD_BRS		0x10000 /* AMD Branch Sampling */
+#define PERF_X86_EVENT_PEBS_LAT_HYBRID	0x20000 /* ld and st lat for hybrid */
 
 static inline bool is_topdown_count(struct perf_event *event)
 {
@@ -461,6 +462,10 @@ struct cpu_hw_events {
 	__EVENT_CONSTRAINT(c, n, INTEL_ARCH_EVENT_MASK|X86_ALL_EVENT_FLAGS, \
 			  HWEIGHT(n), 0, PERF_X86_EVENT_PEBS_ST)
 
+#define INTEL_HYBRID_LAT_CONSTRAINT(c, n)	\
+	__EVENT_CONSTRAINT(c, n, INTEL_ARCH_EVENT_MASK|X86_ALL_EVENT_FLAGS, \
+			  HWEIGHT(n), 0, PERF_X86_EVENT_PEBS_LAT_HYBRID)
+
 /* Event constraint, but match on all event flags too. */
 #define INTEL_FLAGS_EVENT_CONSTRAINT(c, n) \
 	EVENT_CONSTRAINT(c, n, ARCH_PERFMON_EVENTSEL_EVENT|X86_ALL_EVENT_FLAGS)
@@ -826,6 +831,7 @@ struct x86_pmu {
 	void		(*drain_pebs)(struct pt_regs *regs, struct perf_sample_data *data);
 	struct event_constraint *pebs_constraints;
 	void		(*pebs_aliases)(struct perf_event *event);
+	u64		(*pebs_latency_data)(struct perf_event *event, u64 status);
 	unsigned long	large_pebs_flags;
 	u64		rtm_abort_event;
 
@@ -1393,6 +1399,8 @@ void intel_pmu_disable_bts(void);
 
 int intel_pmu_drain_bts_buffer(void);
 
+u64 adl_latency_data_small(struct perf_event *event, u64 status);
+
 extern struct event_constraint intel_core2_pebs_event_constraints[];
 
 extern struct event_constraint intel_atom_pebs_event_constraints[];