Skip to contentSkip to conversation

arm64: PMU updates

PATCH Applied Robin Murphy 4 participants 15 messages Dec 14, 2021

Subsystems (4)
Device tree bindings perf events arm64 Arm PMU profiling and debugging
lore
Share & tools
lore

Robin Murphy

author
#0 lore raw

v2 cover arm64: PMU updates

Hi all,

Here's v2, with the DSU binding actually finished (oh the shame...) and the NVIDIA PMU patch pulled in to benefit from the cleanup and keep things manageably together. The initial binding patch from v1 is no longer included since Rob has picked it up separately.

Refactoring arm_pmu's probing to be more modular turns out to be a substantial yak to shave, so in the end I decided it *is* worth implementing Mark's macro suggestion now - after all, it can simply be converted from a function template to a structure template as and when the time comes.

Cheers,
Robin.

Robin Murphy (4):
arm64: perf: Simplify registration boilerplate
arm64: perf: Support new DT compatibles
dt-bindings: perf: Convert Arm DSU to schema
dt-bindings: perf: Add compatible for Arm DSU-110

Thierry Reding (1):
arm64: perf: Support Denver and Carmel PMUs

 .../devicetree/bindings/arm/arm-dsu-pmu.txt   |  27 -----
 .../devicetree/bindings/perf/arm,dsu-pmu.yaml |  45 +++++++
 arch/arm64/kernel/perf_event.c                | 111 +++++++-----------
 3 files changed, 88 insertions(+), 95 deletions(-)
 delete mode 100644 Documentation/devicetree/bindings/arm/arm-dsu-pmu.txt
 create mode 100644 Documentation/devicetree/bindings/perf/arm,dsu-pmu.yaml
Signature · 1 line
2.28.0.dirty
author
#1 lore raw

PATCH v2 1/5 arm64: perf: Support Denver and Carmel PMUs

From: Thierry Reding <[email protected]>

Add support for the NVIDIA Denver and Carmel PMUs using the generic PMUv3 event map for now.

Acked-by Mark Rutland <[email protected]>
Signed-off-by Thierry Reding <[email protected]>
[ rm: reorder entries alphabetically ]
Signed-off-by Robin Murphy <[email protected]>
---
 arch/arm64/kernel/perf_event.c | 14 ++++++++++++++
 1 file changed, 14 insertions(+)

v2: Newly picked up

arch/arm64/kernel/perf_event.c +14 -0
index b4044469527e..035da52e04bf 100644
@@ -1247,6 +1247,18 @@ static int armv8_vulcan_pmu_init(struct arm_pmu *cpu_pmu)
 				       armv8_vulcan_map_event);
 }
 
+static int armv8_carmel_pmu_init(struct arm_pmu *cpu_pmu)
+{
+	return armv8_pmu_init_nogroups(cpu_pmu, "armv8_nvidia_carmel",
+				       armv8_pmuv3_map_event);
+}
+
+static int armv8_denver_pmu_init(struct arm_pmu *cpu_pmu)
+{
+	return armv8_pmu_init_nogroups(cpu_pmu, "armv8_nvidia_denver",
+				       armv8_pmuv3_map_event);
+}
+
 static const struct of_device_id armv8_pmu_of_device_ids[] = {
 	{.compatible = "arm,armv8-pmuv3",	.data = armv8_pmuv3_init},
 	{.compatible = "arm,cortex-a34-pmu",	.data = armv8_a34_pmu_init},

@@ -1265,6 +1277,8 @@ static const struct of_device_id armv8_pmu_of_device_ids[] = {
 	{.compatible = "arm,neoverse-n1-pmu",	.data = armv8_n1_pmu_init},
 	{.compatible = "cavium,thunder-pmu",	.data = armv8_thunder_pmu_init},
 	{.compatible = "brcm,vulcan-pmu",	.data = armv8_vulcan_pmu_init},
+	{.compatible = "nvidia,carmel-pmu",	.data = armv8_carmel_pmu_init},
+	{.compatible = "nvidia,denver-pmu",	.data = armv8_denver_pmu_init},
 	{},
 };
Signature · 1 line
2.28.0.dirty
author
#2 lore raw

PATCH v2 2/5 arm64: perf: Simplify registration boilerplate

In reply to Robin Murphy

With the trend for per-core events moving to userspace JSON, registering names for PMUv3 implementations is increasingly a pure boilerplate exercise. Let's wrap things a step further so we can generate the basic PMUv3 init function with a macro invocation, and reduce further new addition to just 2 lines each.

Suggested-by Mark Rutland <[email protected]>
Signed-off-by Robin Murphy <[email protected]>
---
 arch/arm64/kernel/perf_event.c | 113 +++++++++------------------------
 1 file changed, 31 insertions(+), 82 deletions(-)

v2: Rewrite per Mark's suggestion, cover new NVIDIA PMUs too

arch/arm64/kernel/perf_event.c +31 -82
index 035da52e04bf..ed6d8cd2f88f 100644
@@ -1145,17 +1145,26 @@ static int armv8_pmu_init_nogroups(struct arm_pmu *cpu_pmu, char *name,
 	return armv8_pmu_init(cpu_pmu, name, map_event, NULL, NULL, NULL);
 }
 
-static int armv8_pmuv3_init(struct arm_pmu *cpu_pmu)
-{
-	return armv8_pmu_init_nogroups(cpu_pmu, "armv8_pmuv3",
-				       armv8_pmuv3_map_event);
+#define PMUV3_INIT_SIMPLE(name)						\
+static int name##_pmu_init(struct arm_pmu *cpu_pmu)			\
+{									\
+	return armv8_pmu_init_nogroups(cpu_pmu, #name, armv8_pmuv3_map_event);\
 }
 
-static int armv8_a34_pmu_init(struct arm_pmu *cpu_pmu)
-{
-	return armv8_pmu_init_nogroups(cpu_pmu, "armv8_cortex_a34",
-				       armv8_pmuv3_map_event);
-}
+PMUV3_INIT_SIMPLE(armv8_pmuv3)
+
+PMUV3_INIT_SIMPLE(armv8_cortex_a34)
+PMUV3_INIT_SIMPLE(armv8_cortex_a55)
+PMUV3_INIT_SIMPLE(armv8_cortex_a65)
+PMUV3_INIT_SIMPLE(armv8_cortex_a75)
+PMUV3_INIT_SIMPLE(armv8_cortex_a76)
+PMUV3_INIT_SIMPLE(armv8_cortex_a77)
+PMUV3_INIT_SIMPLE(armv8_cortex_a78)
+PMUV3_INIT_SIMPLE(armv8_neoverse_e1)
+PMUV3_INIT_SIMPLE(armv8_neoverse_n1)
+
+PMUV3_INIT_SIMPLE(armv8_nvidia_carmel)
+PMUV3_INIT_SIMPLE(armv8_nvidia_denver)
 
 static int armv8_a35_pmu_init(struct arm_pmu *cpu_pmu)
 {

@@ -1169,24 +1178,12 @@ static int armv8_a53_pmu_init(struct arm_pmu *cpu_pmu)
 				       armv8_a53_map_event);
 }
 
-static int armv8_a55_pmu_init(struct arm_pmu *cpu_pmu)
-{
-	return armv8_pmu_init_nogroups(cpu_pmu, "armv8_cortex_a55",
-				       armv8_pmuv3_map_event);
-}
-
 static int armv8_a57_pmu_init(struct arm_pmu *cpu_pmu)
 {
 	return armv8_pmu_init_nogroups(cpu_pmu, "armv8_cortex_a57",
 				       armv8_a57_map_event);
 }
 
-static int armv8_a65_pmu_init(struct arm_pmu *cpu_pmu)
-{
-	return armv8_pmu_init_nogroups(cpu_pmu, "armv8_cortex_a65",
-				       armv8_pmuv3_map_event);
-}
-
 static int armv8_a72_pmu_init(struct arm_pmu *cpu_pmu)
 {
 	return armv8_pmu_init_nogroups(cpu_pmu, "armv8_cortex_a72",

@@ -1199,42 +1196,6 @@ static int armv8_a73_pmu_init(struct arm_pmu *cpu_pmu)
 				       armv8_a73_map_event);
 }
 
-static int armv8_a75_pmu_init(struct arm_pmu *cpu_pmu)
-{
-	return armv8_pmu_init_nogroups(cpu_pmu, "armv8_cortex_a75",
-				       armv8_pmuv3_map_event);
-}
-
-static int armv8_a76_pmu_init(struct arm_pmu *cpu_pmu)
-{
-	return armv8_pmu_init_nogroups(cpu_pmu, "armv8_cortex_a76",
-				       armv8_pmuv3_map_event);
-}
-
-static int armv8_a77_pmu_init(struct arm_pmu *cpu_pmu)
-{
-	return armv8_pmu_init_nogroups(cpu_pmu, "armv8_cortex_a77",
-				       armv8_pmuv3_map_event);
-}
-
-static int armv8_a78_pmu_init(struct arm_pmu *cpu_pmu)
-{
-	return armv8_pmu_init_nogroups(cpu_pmu, "armv8_cortex_a78",
-				       armv8_pmuv3_map_event);
-}
-
-static int armv8_e1_pmu_init(struct arm_pmu *cpu_pmu)
-{
-	return armv8_pmu_init_nogroups(cpu_pmu, "armv8_neoverse_e1",
-				       armv8_pmuv3_map_event);
-}
-
-static int armv8_n1_pmu_init(struct arm_pmu *cpu_pmu)
-{
-	return armv8_pmu_init_nogroups(cpu_pmu, "armv8_neoverse_n1",
-				       armv8_pmuv3_map_event);
-}
-
 static int armv8_thunder_pmu_init(struct arm_pmu *cpu_pmu)
 {
 	return armv8_pmu_init_nogroups(cpu_pmu, "armv8_cavium_thunder",

@@ -1247,38 +1208,26 @@ static int armv8_vulcan_pmu_init(struct arm_pmu *cpu_pmu)
 				       armv8_vulcan_map_event);
 }
 
-static int armv8_carmel_pmu_init(struct arm_pmu *cpu_pmu)
-{
-	return armv8_pmu_init_nogroups(cpu_pmu, "armv8_nvidia_carmel",
-				       armv8_pmuv3_map_event);
-}
-
-static int armv8_denver_pmu_init(struct arm_pmu *cpu_pmu)
-{
-	return armv8_pmu_init_nogroups(cpu_pmu, "armv8_nvidia_denver",
-				       armv8_pmuv3_map_event);
-}
-
 static const struct of_device_id armv8_pmu_of_device_ids[] = {
-	{.compatible = "arm,armv8-pmuv3",	.data = armv8_pmuv3_init},
-	{.compatible = "arm,cortex-a34-pmu",	.data = armv8_a34_pmu_init},
+	{.compatible = "arm,armv8-pmuv3",	.data = armv8_pmuv3_pmu_init},
+	{.compatible = "arm,cortex-a34-pmu",	.data = armv8_cortex_a34_pmu_init},
 	{.compatible = "arm,cortex-a35-pmu",	.data = armv8_a35_pmu_init},
 	{.compatible = "arm,cortex-a53-pmu",	.data = armv8_a53_pmu_init},
-	{.compatible = "arm,cortex-a55-pmu",	.data = armv8_a55_pmu_init},
+	{.compatible = "arm,cortex-a55-pmu",	.data = armv8_cortex_a55_pmu_init},
 	{.compatible = "arm,cortex-a57-pmu",	.data = armv8_a57_pmu_init},
-	{.compatible = "arm,cortex-a65-pmu",	.data = armv8_a65_pmu_init},
+	{.compatible = "arm,cortex-a65-pmu",	.data = armv8_cortex_a65_pmu_init},
 	{.compatible = "arm,cortex-a72-pmu",	.data = armv8_a72_pmu_init},
 	{.compatible = "arm,cortex-a73-pmu",	.data = armv8_a73_pmu_init},
-	{.compatible = "arm,cortex-a75-pmu",	.data = armv8_a75_pmu_init},
-	{.compatible = "arm,cortex-a76-pmu",	.data = armv8_a76_pmu_init},
-	{.compatible = "arm,cortex-a77-pmu",	.data = armv8_a77_pmu_init},
-	{.compatible = "arm,cortex-a78-pmu",	.data = armv8_a78_pmu_init},
-	{.compatible = "arm,neoverse-e1-pmu",	.data = armv8_e1_pmu_init},
-	{.compatible = "arm,neoverse-n1-pmu",	.data = armv8_n1_pmu_init},
+	{.compatible = "arm,cortex-a75-pmu",	.data = armv8_cortex_a75_pmu_init},
+	{.compatible = "arm,cortex-a76-pmu",	.data = armv8_cortex_a76_pmu_init},
+	{.compatible = "arm,cortex-a77-pmu",	.data = armv8_cortex_a77_pmu_init},
+	{.compatible = "arm,cortex-a78-pmu",	.data = armv8_cortex_a78_pmu_init},
+	{.compatible = "arm,neoverse-e1-pmu",	.data = armv8_neoverse_e1_pmu_init},
+	{.compatible = "arm,neoverse-n1-pmu",	.data = armv8_neoverse_n1_pmu_init},
 	{.compatible = "cavium,thunder-pmu",	.data = armv8_thunder_pmu_init},
 	{.compatible = "brcm,vulcan-pmu",	.data = armv8_vulcan_pmu_init},
-	{.compatible = "nvidia,carmel-pmu",	.data = armv8_carmel_pmu_init},
-	{.compatible = "nvidia,denver-pmu",	.data = armv8_denver_pmu_init},
+	{.compatible = "nvidia,carmel-pmu",	.data = armv8_nvidia_carmel_pmu_init},
+	{.compatible = "nvidia,denver-pmu",	.data = armv8_nvidia_denver_pmu_init},
 	{},
 };
 

@@ -1301,7 +1250,7 @@ static int __init armv8_pmu_driver_init(void)
 	if (acpi_disabled)
 		return platform_driver_register(&armv8_pmu_driver);
 	else
-		return arm_pmu_acpi_probe(armv8_pmuv3_init);
+		return arm_pmu_acpi_probe(armv8_pmuv3_pmu_init);
 }
 device_initcall(armv8_pmu_driver_init)
Signature · 1 line
2.28.0.dirty
author
#3 lore raw

PATCH v2 3/5 arm64: perf: Support new DT compatibles

In reply to Robin Murphy

Wire up the new DT compatibles so we can present appropriate PMU names to userspace for the latest and greatest CPUs.

Signed-off-by Robin Murphy <[email protected]>

---

v2: Use new macro template

 arch/arm64/kernel/perf_event.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)
arch/arm64/kernel/perf_event.c +12 -0
index ed6d8cd2f88f..4c1698f63f34 100644
@@ -1160,8 +1160,14 @@ PMUV3_INIT_SIMPLE(armv8_cortex_a75)
 PMUV3_INIT_SIMPLE(armv8_cortex_a76)
 PMUV3_INIT_SIMPLE(armv8_cortex_a77)
 PMUV3_INIT_SIMPLE(armv8_cortex_a78)
+PMUV3_INIT_SIMPLE(armv9_cortex_a510)
+PMUV3_INIT_SIMPLE(armv9_cortex_a710)
+PMUV3_INIT_SIMPLE(armv8_cortex_x1)
+PMUV3_INIT_SIMPLE(armv9_cortex_x2)
 PMUV3_INIT_SIMPLE(armv8_neoverse_e1)
 PMUV3_INIT_SIMPLE(armv8_neoverse_n1)
+PMUV3_INIT_SIMPLE(armv9_neoverse_n2)
+PMUV3_INIT_SIMPLE(armv8_neoverse_v1)
 
 PMUV3_INIT_SIMPLE(armv8_nvidia_carmel)
 PMUV3_INIT_SIMPLE(armv8_nvidia_denver)

@@ -1222,8 +1228,14 @@ static const struct of_device_id armv8_pmu_of_device_ids[] = {
 	{.compatible = "arm,cortex-a76-pmu",	.data = armv8_cortex_a76_pmu_init},
 	{.compatible = "arm,cortex-a77-pmu",	.data = armv8_cortex_a77_pmu_init},
 	{.compatible = "arm,cortex-a78-pmu",	.data = armv8_cortex_a78_pmu_init},
+	{.compatible = "arm,cortex-a510-pmu",	.data = armv9_cortex_a510_pmu_init},
+	{.compatible = "arm,cortex-a710-pmu",	.data = armv9_cortex_a710_pmu_init},
+	{.compatible = "arm,cortex-x1-pmu",	.data = armv8_cortex_x1_pmu_init},
+	{.compatible = "arm,cortex-x2-pmu",	.data = armv9_cortex_x2_pmu_init},
 	{.compatible = "arm,neoverse-e1-pmu",	.data = armv8_neoverse_e1_pmu_init},
 	{.compatible = "arm,neoverse-n1-pmu",	.data = armv8_neoverse_n1_pmu_init},
+	{.compatible = "arm,neoverse-n2-pmu",	.data = armv9_neoverse_n2_pmu_init},
+	{.compatible = "arm,neoverse-v1-pmu",	.data = armv8_neoverse_v1_pmu_init},
 	{.compatible = "cavium,thunder-pmu",	.data = armv8_thunder_pmu_init},
 	{.compatible = "brcm,vulcan-pmu",	.data = armv8_vulcan_pmu_init},
 	{.compatible = "nvidia,carmel-pmu",	.data = armv8_nvidia_carmel_pmu_init},
Signature · 1 line
2.28.0.dirty
author
#4 lore raw

PATCH v2 4/5 dt-bindings: perf: Convert Arm DSU to schema

In reply to Robin Murphy

Convert the DSU binding to schema, as one does.

Signed-off-by Robin Murphy <[email protected]>

---

v2: Actually finish converting the thing to valid schema...

 .../devicetree/bindings/arm/arm-dsu-pmu.txt   | 27 ------------
 .../devicetree/bindings/perf/arm,dsu-pmu.yaml | 41 +++++++++++++++++++
 2 files changed, 41 insertions(+), 27 deletions(-)
 delete mode 100644 Documentation/devicetree/bindings/arm/arm-dsu-pmu.txt
 create mode 100644 Documentation/devicetree/bindings/perf/arm,dsu-pmu.yaml
Documentation/devicetree/bindings/arm/arm-dsu-pmu.txt deleted +0 -27
deleted file mode 100644
index 6efabba530f1..000000000000
@@ -1,27 +0,0 @@
-* ARM DynamIQ Shared Unit (DSU) Performance Monitor Unit (PMU)
-
-ARM DyanmIQ Shared Unit (DSU) integrates one or more CPU cores
-with a shared L3 memory system, control logic and external interfaces to
-form a multicore cluster. The PMU enables to gather various statistics on
-the operations of the DSU. The PMU provides independent 32bit counters that
-can count any of the supported events, along with a 64bit cycle counter.
-The PMU is accessed via CPU system registers and has no MMIO component.
-
-** DSU PMU required properties:
-
-- compatible	: should be one of :
-
-		"arm,dsu-pmu"
-
-- interrupts	: Exactly 1 SPI must be listed.
-
-- cpus		: List of phandles for the CPUs connected to this DSU instance.
-
-
-** Example:
-
-dsu-pmu-0 {
-	compatible = "arm,dsu-pmu";
-	interrupts = <GIC_SPI 02 IRQ_TYPE_LEVEL_HIGH>;
-	cpus = <&cpu_0>, <&cpu_1>;
-};
Documentation/devicetree/bindings/perf/arm,dsu-pmu.yaml new file +41 -0

41 lines not on this page · raw message

Signature · 1 line
2.28.0.dirty
author
#7 lore raw

PATCH v2 5/5 dt-bindings: perf: Add compatible for Arm DSU-110

In reply to Robin Murphy

DSU-110 is the newest and shiniest for Armv9. Its programmer's model is largely identical to the previous generation of DSUs, so we can treat it as compatible, but it does have a a handful of extra IMP-DEF PMU events to call its own. Thanks to the new notion of core complexes, the maximum number of supported CPUs goes up as well.

Signed-off-by Robin Murphy <[email protected]>

---

v2: Also make it complete and valid

 Documentation/devicetree/bindings/perf/arm,dsu-pmu.yaml | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)
Documentation/devicetree/bindings/perf/arm,dsu-pmu.yaml +6 -2
index 09ddeb6a3ccc..7f31bcecedb5 100644
@@ -21,7 +21,11 @@ description:
 
 properties:
   compatible:
-    const: arm,dsu-pmu
+    oneOf:
+      - const: arm,dsu-pmu
+      - items:
+          - const: arm,dsu-110-pmu
+          - const: arm,dsu-pmu
 
   interrupts:
     items:

@@ -30,7 +34,7 @@ properties:
   cpus:
     $ref: /schemas/types.yaml#/definitions/phandle-array
     minItems: 1
-    maxItems: 8
+    maxItems: 12
     description: List of phandles for the CPUs connected to this DSU instance.
 
 required:
Signature · 1 line
2.28.0.dirty
Acked-by on v2 4/5
#5 lore raw

Re: [PATCH v2 4/5] dt-bindings: perf: Convert Arm DSU to schema

In reply to Robin Murphy
Quoted 57 lines · Robin Murphy

Convert the DSU binding to schema, as one does.

Signed-off-by Robin Murphy <[email protected]>

---

v2: Actually finish converting the thing to valid schema...

  .../devicetree/bindings/arm/arm-dsu-pmu.txt   | 27 ------------
  .../devicetree/bindings/perf/arm,dsu-pmu.yaml | 41 +++++++++++++++++++
  2 files changed, 41 insertions(+), 27 deletions(-)

delete mode 100644 Documentation/devicetree/bindings/arm/arm-dsu-pmu.txt create mode 100644 Documentation/devicetree/bindings/perf/arm,dsu-pmu.yaml

Documentation/devicetree/bindings/arm/arm-dsu-pmu.txt deleted +0 -27
deleted file mode 100644
index 6efabba530f1..000000000000
@@ -1,27 +0,0 @@
-* ARM DynamIQ Shared Unit (DSU) Performance Monitor Unit (PMU)
-
-ARM DyanmIQ Shared Unit (DSU) integrates one or more CPU cores
-with a shared L3 memory system, control logic and external interfaces to
-form a multicore cluster. The PMU enables to gather various statistics on
-the operations of the DSU. The PMU provides independent 32bit counters that
-can count any of the supported events, along with a 64bit cycle counter.
-The PMU is accessed via CPU system registers and has no MMIO component.
-
-** DSU PMU required properties:
-
-- compatible	: should be one of :
-
-		"arm,dsu-pmu"
-
-- interrupts	: Exactly 1 SPI must be listed.
-
-- cpus		: List of phandles for the CPUs connected to this DSU instance.
-
-
-** Example:
-
-dsu-pmu-0 {
-	compatible = "arm,dsu-pmu";
-	interrupts = <GIC_SPI 02 IRQ_TYPE_LEVEL_HIGH>;
-	cpus = <&cpu_0>, <&cpu_1>;
-};
Documentation/devicetree/bindings/perf/arm,dsu-pmu.yaml new file +5 -0
new file mode 100644
index 000000000000..09ddeb6a3ccc
@@ -0,0 +1,41 @@
+# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+# Copyright 2021 Arm Ltd.
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/perf/arm,dsu-pmu.yaml#

Documentation/devicetree/bindings/perf/arm,dsu-pmu.yaml line 15

Quoted 10 lines
Documentation/devicetree/bindings/perf/arm,dsu-pmu.yaml new file +10 -0
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: ARM DynamIQ Shared Unit (DSU) Performance Monitor Unit (PMU)
+
+maintainers:
+  - Suzuki K Poulose <[email protected]>
+  - Robin Murphy <[email protected]>
+
+description:
+  ARM DyanmIQ Shared Unit (DSU) integrates one or more CPU cores with a shared

nit: DynamIQ. I understand I am the one to blame for ;-)

Rest looks good to me.

Acked-by Suzuki K Poulose <[email protected]>
#8 lore raw

Re: [PATCH v2 5/5] dt-bindings: perf: Add compatible for Arm DSU-110

In reply to Robin Murphy
Quoted 19 lines · Robin Murphy

DSU-110 is the newest and shiniest for Armv9. Its programmer's model is largely identical to the previous generation of DSUs, so we can treat it as compatible, but it does have a a handful of extra IMP-DEF PMU events to call its own. Thanks to the new notion of core complexes, the maximum number of supported CPUs goes up as well.

Signed-off-by Robin Murphy <[email protected]>

---

v2: Also make it complete and valid

  Documentation/devicetree/bindings/perf/arm,dsu-pmu.yaml | 8 ++++++--
  1 file changed, 6 insertions(+), 2 deletions(-)
Documentation/devicetree/bindings/perf/arm,dsu-pmu.yaml +0 -0
index 09ddeb6a3ccc..7f31bcecedb5 100644

Documentation/devicetree/bindings/perf/arm,dsu-pmu.yaml description: · line 28

Quoted 9 lines
Documentation/devicetree/bindings/perf/arm,dsu-pmu.yaml +5 -1
@@ -21,7 +21,11 @@ description:
 
  properties:
    compatible:
-    const: arm,dsu-pmu
+    oneOf:
+      - const: arm,dsu-pmu
+      - items:
+          - const: arm,dsu-110-pmu
+          - const: arm,dsu-pmu

Couldn't this be :

compatible:
oneOf:
- const: arm,dsu-110-pmu
- const: arm,dsu-pmu

Otherwise, looks good to me.

Suzuki

author
#9 lore raw

Re: [PATCH v2 5/5] dt-bindings: perf: Add compatible for Arm DSU-110

Quoted 31 lines · Suzuki K Poulose
Quoted 19 lines · Robin Murphy

DSU-110 is the newest and shiniest for Armv9. Its programmer's model is largely identical to the previous generation of DSUs, so we can treat it as compatible, but it does have a a handful of extra IMP-DEF PMU events to call its own. Thanks to the new notion of core complexes, the maximum number of supported CPUs goes up as well.

Signed-off-by Robin Murphy <[email protected]>

---

v2: Also make it complete and valid

  Documentation/devicetree/bindings/perf/arm,dsu-pmu.yaml | 8 ++++++--

  1 file changed, 6 insertions(+), 2 deletions(-)
Documentation/devicetree/bindings/perf/arm,dsu-pmu.yaml +0 -0
index 09ddeb6a3ccc..7f31bcecedb5 100644
@@ -21,7 +21,11 @@ description:
Quoted 8 lines

  properties:
    compatible:
-    const: arm,dsu-pmu
+    oneOf:
+      - const: arm,dsu-pmu
+      - items:
+          - const: arm,dsu-110-pmu
+          - const: arm,dsu-pmu

Couldn't this be :

Quoted 6 lines

    compatible:
      oneOf:
        - const: arm,dsu-110-pmu
        - const: arm,dsu-pmu

Otherwise, looks good to me.

My intent here was for the new case to use both strings and thus still permit a less-specific match to the original one (hence why there's no change to the actual driver included here). It's a true "compatible" in the sense that you can run the thing with the existing driver as if it's a 1st-gen DSU and get the expected results - at least that's the conclusion I drew from a side-by-side comparison of the TRMs, and the folks testing FPGAs haven't complained since I told them to try it months ago :)

Thanks,
Robin.

maintainer
#12 lore raw
In reply to Robin Murphy
Quoted 5 lines · Robin Murphy

Here's v2, with the DSU binding actually finished (oh the shame...) and the NVIDIA PMU patch pulled in to benefit from the cleanup and keep things manageably together. The initial binding patch from v1 is no longer included since Rob has picked it up separately.

Quoted 7 lines

Refactoring arm_pmu's probing to be more modular turns out to be a substantial yak to shave, so in the end I decided it *is* worth implementing Mark's macro suggestion now - after all, it can simply be converted from a function template to a structure template as and when the time comes.

[...]

Applied first three to will (for-next/perf-cpu), thanks!

[1/5] arm64: perf: Support Denver and Carmel PMUs

      https://git.kernel.org/will/c/d4c4844a9b47
[2/5] arm64: perf: Simplify registration boilerplate
      https://git.kernel.org/will/c/6ac9f30bd43b
[3/5] arm64: perf: Support new DT compatibles
      https://git.kernel.org/will/c/893c34b60a59

Cheers,

Signature · 5 lines
Will

https://fixes.arm64.dev
https://next.arm64.dev
https://will.arm64.dev
Acked-by
#10 lore raw

Re: [PATCH v2 5/5] dt-bindings: perf: Add compatible for Arm DSU-110

In reply to Robin Murphy
Quoted 32 lines · Robin Murphy
Quoted 31 lines · Suzuki K Poulose
Quoted 19 lines · Robin Murphy

DSU-110 is the newest and shiniest for Armv9. Its programmer's model is largely identical to the previous generation of DSUs, so we can treat it as compatible, but it does have a a handful of extra IMP-DEF PMU events to call its own. Thanks to the new notion of core complexes, the maximum number of supported CPUs goes up as well.

Signed-off-by Robin Murphy <[email protected]>

---

v2: Also make it complete and valid

  Documentation/devicetree/bindings/perf/arm,dsu-pmu.yaml | 8 ++++++--

  1 file changed, 6 insertions(+), 2 deletions(-)
Documentation/devicetree/bindings/perf/arm,dsu-pmu.yaml +0 -0
index 09ddeb6a3ccc..7f31bcecedb5 100644
@@ -21,7 +21,11 @@ description:
Quoted 8 lines

  properties:
    compatible:
-    const: arm,dsu-pmu
+    oneOf:
+      - const: arm,dsu-pmu
+      - items:
+          - const: arm,dsu-110-pmu
+          - const: arm,dsu-pmu

Couldn't this be :

Quoted 8 lines
Quoted 6 lines

     compatible:
       oneOf:
         - const: arm,dsu-110-pmu
         - const: arm,dsu-pmu

Otherwise, looks good to me.

My intent here was for the new case to use both strings and thus still permit a less-specific match to the original one (hence why there's no change to the actual driver included here). It's a true "compatible" in the sense that you can run the thing with the existing driver as if it's a 1st-gen DSU and get the expected results - at least that's the conclusion I drew from a side-by-side comparison of the TRMs, and the folks testing FPGAs haven't complained since I told them to try it months ago :)

Ah, right. I should have guessed it from the lack of update to the driver.

Acked-by Suzuki K Poulose <[email protected]>
author
#13 lore raw
In reply to Will Deacon
Quoted 14 lines · Will Deacon
Quoted 5 lines · Robin Murphy

Here's v2, with the DSU binding actually finished (oh the shame...) and the NVIDIA PMU patch pulled in to benefit from the cleanup and keep things manageably together. The initial binding patch from v1 is no longer included since Rob has picked it up separately.

Quoted 7 lines

Refactoring arm_pmu's probing to be more modular turns out to be a substantial yak to shave, so in the end I decided it *is* worth implementing Mark's macro suggestion now - after all, it can simply be converted from a function template to a structure template as and when the time comes.

[...]

Quoted 8 lines

Applied first three to will (for-next/perf-cpu), thanks!

[1/5] arm64: perf: Support Denver and Carmel PMUs

       https://git.kernel.org/will/c/d4c4844a9b47
[2/5] arm64: perf: Simplify registration boilerplate
       https://git.kernel.org/will/c/6ac9f30bd43b
[3/5] arm64: perf: Support new DT compatibles
       https://git.kernel.org/will/c/893c34b60a59

Thanks Will! Just to clarify, are you leaving the DSU binding patches for Rob to pick up, or waiting for the typo to get fixed? (and if the former - Rob, would you like me to resend a v3 with the typo fixed?)

Cheers,
Robin.

maintainer
#14 lore raw
Quoted 24 lines · Robin Murphy
Quoted 14 lines · Will Deacon
Quoted 5 lines · Robin Murphy

Here's v2, with the DSU binding actually finished (oh the shame...) and the NVIDIA PMU patch pulled in to benefit from the cleanup and keep things manageably together. The initial binding patch from v1 is no longer included since Rob has picked it up separately.

Quoted 7 lines

Refactoring arm_pmu's probing to be more modular turns out to be a substantial yak to shave, so in the end I decided it *is* worth implementing Mark's macro suggestion now - after all, it can simply be converted from a function template to a structure template as and when the time comes.

[...]

Quoted 8 lines

Applied first three to will (for-next/perf-cpu), thanks!

[1/5] arm64: perf: Support Denver and Carmel PMUs

       https://git.kernel.org/will/c/d4c4844a9b47
[2/5] arm64: perf: Simplify registration boilerplate
       https://git.kernel.org/will/c/6ac9f30bd43b
[3/5] arm64: perf: Support new DT compatibles
       https://git.kernel.org/will/c/893c34b60a59
Quoted 3 lines

Thanks Will! Just to clarify, are you leaving the DSU binding patches for Rob to pick up, or waiting for the typo to get fixed? (and if the former - Rob, would you like me to resend a v3 with the typo fixed?)

I'm about to disappear on holiday, so I figured it was best to take the first three patches and leave the other two (as they seem less urgent and also had comment from Suzuki). If somebody else grabs 'em when they're ready, so be it, otherwise I'll pick them next time around.

Cheers,

Will

maintainer Applied
#6 lore raw

Re: [PATCH v2 4/5] dt-bindings: perf: Convert Arm DSU to schema

In reply to Robin Murphy
Quoted 5 lines · Robin Murphy

Convert the DSU binding to schema, as one does.

Signed-off-by Robin Murphy <[email protected]>

---

Quoted 8 lines

v2: Actually finish converting the thing to valid schema...

 .../devicetree/bindings/arm/arm-dsu-pmu.txt   | 27 ------------
 .../devicetree/bindings/perf/arm,dsu-pmu.yaml | 41 +++++++++++++++++++
 2 files changed, 41 insertions(+), 27 deletions(-)
 delete mode 100644 Documentation/devicetree/bindings/arm/arm-dsu-pmu.txt
 create mode 100644 Documentation/devicetree/bindings/perf/arm,dsu-pmu.yaml

Applied, thanks!

maintainer Applied
#11 lore raw

Re: [PATCH v2 5/5] dt-bindings: perf: Add compatible for Arm DSU-110

In reply to Robin Murphy
Quoted 6 lines · Robin Murphy

DSU-110 is the newest and shiniest for Armv9. Its programmer's model is largely identical to the previous generation of DSUs, so we can treat it as compatible, but it does have a a handful of extra IMP-DEF PMU events to call its own. Thanks to the new notion of core complexes, the maximum number of supported CPUs goes up as well.

Quoted 8 lines
Signed-off-by Robin Murphy <[email protected]>

---

v2: Also make it complete and valid

 Documentation/devicetree/bindings/perf/arm,dsu-pmu.yaml | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

Applied, thanks!