perf: Arm system PMU and BMU drivers

Previously:
https://lore.kernel.org/linux-perf-users/222d98feab5ca2836fdef69fee9113f6f860167d.1791209460.git.robin.murphy@arm.com/
https://lore.kernel.org/linux-perf-users/c936505c43275f434bce924b7657e42bfb25579c.1789652832.git.robin.murphy@arm.com/

As promised, this a resend of the new drivers which missed the boat last cycle. No material changes since the respective v3s, just rebasing, and I decided I may as well post them together as a series to pre-resolve the Kconfig/Makefile conflicts. I've also revived the BMU DT patches from v2, as some interest has been expressed from within Arm, but again they can still be considered somewhat optional as it's only the first two which the AGI CPU folks are clamouring for right now.

Cheers,
Robin.

Robin Murphy (4):
perf/arm_cspmu: Add Arm system PMU
perf: Add Arm Bus Monitor Unit driver
dt-bindings: perf: Add Arm Bus Monitor Unit
perf/arm-bmu: Add Devicetree support

 .../bindings/perf/arm,bus-monitor-unit.yaml   |  41 ++
 drivers/perf/arm_cspmu/Kconfig                |  13 +
 drivers/perf/arm_cspmu/Makefile               |   3 +
 drivers/perf/arm_cspmu/arm-bmu.c              | 541 ++++++++++++++++++
 drivers/perf/arm_cspmu/arm_cspmu.c            |   7 +
 drivers/perf/arm_cspmu/arm_cspmu.h            |   1 +
 drivers/perf/arm_cspmu/arm_syspmu.c           |  67 +++
 7 files changed, 673 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/perf/arm,bus-monitor-unit.yaml
 create mode 100644 drivers/perf/arm_cspmu/arm-bmu.c
 create mode 100644 drivers/perf/arm_cspmu/arm_syspmu.c

2.54.0.dirty

Arm's latest Compute Subsystems and SoCs are using a common PMU block in various places which can mostly just be driven as a generic 64-bit CoreSight PMU, but does have an IMP-DEF register for additional filtering. The details of event and filter encodings depend on the particular PMU instance, so are left to the higher level Arm Telemetry Solution tooling.

---
 drivers/perf/arm_cspmu/Kconfig      |  7 +++
 drivers/perf/arm_cspmu/Makefile     |  1 +
 drivers/perf/arm_cspmu/arm_cspmu.c  |  7 +++
 drivers/perf/arm_cspmu/arm_cspmu.h  |  1 +
 drivers/perf/arm_cspmu/arm_syspmu.c | 67 +++++++++++++++++++++++++++++
 5 files changed, 83 insertions(+)
 create mode 100644 drivers/perf/arm_cspmu/arm_syspmu.c

index 6f4e28fc84a2..25bd55f036a4 100644
@@ -27,3 +27,10 @@ config AMPERE_CORESIGHT_PMU_ARCH_SYSTEM_PMU
 
 	  In the first phase, the driver enables support on MCU PMU used in
 	  AmpereOne SoC family.
+
+config ARM_SYSTEM_PMU
+	tristate "Arm System PMU"
+	depends on ARM_CORESIGHT_PMU_ARCH_SYSTEM_PMU
+	help
+	  Support additional event filtering attributes for the CoreSight-based
+	  PMU blocks found in various parts of Arm Compute Subsystems and SoCs.

index 220a734efd54..90c3cd710157 100644
@@ -8,3 +8,4 @@ arm_cspmu_module-y := arm_cspmu.o
 
 obj-$(CONFIG_NVIDIA_CORESIGHT_PMU_ARCH_SYSTEM_PMU) += nvidia_cspmu.o
 obj-$(CONFIG_AMPERE_CORESIGHT_PMU_ARCH_SYSTEM_PMU) += ampere_cspmu.o
+obj-$(CONFIG_ARM_SYSTEM_PMU) += arm_syspmu.o

index 912f03ce3ef5..9f298ea1a974 100644
@@ -358,6 +358,13 @@ static struct arm_cspmu_impl_match impl_match[] = {
 		.module		= NULL,
 		.impl_init_ops	= NULL,
 	},
+	{
+		.module_name	= "arm_syspmu",
+		.pmiidr_val	= ARM_CSPMU_IMPL_ID_ARM,
+		.pmiidr_mask	= PMIIDR_IMPLEMENTER,
+		.module		= NULL,
+		.impl_init_ops	= NULL,
+	},
 	{
 		.module_name	= "ampere_cspmu",
 		.pmiidr_val	= ARM_CSPMU_IMPL_ID_AMPERE,

index 93d5b062de8b..86b62c236a4b 100644
@@ -149,6 +149,7 @@
 
 /* JEDEC-assigned JEP106 identification code */
 #define ARM_CSPMU_IMPL_ID_NVIDIA	0x36B
+#define ARM_CSPMU_IMPL_ID_ARM		0x43B
 #define ARM_CSPMU_IMPL_ID_AMPERE	0xA16
 
 /* PMDEVARCH */

2.54.0.dirty

Sure, let me know what other Arm Ltd. PMU implementations those are, and I'll get back to you with how much I care...

Plus there's nothing here that's any different from the existing Ampere and NVIDIA cases. Perhaps we could #ifdef the impl_match entries if anyone really thought it mattered, but that would be a general matter for the arm_cspmu framework itself, not this patch.

Thanks,
Robin.

Arm's Bus Monitor Unit is a low-level performance analysis tool for matching and counting transactions at interconnect interfaces. Each BMU consists of a number of "interface monitoring units", with some global controls to synchronise them and so permit accurate calculation of cross-interface metrics. Each IMU contains, among other things, its own PMU largely based on the CoreSight PMU architecture, but with some functional incompatibilities due to the nature of how much per-event filtering needs to be crammed in to categorise AXI/CHI transactions at the protocol level. As such it gets an honorary home in arm_cspmu/ in order to share the common register definitions, but for simplicity does not try to bend the functional abstraction to fit.

---
 drivers/perf/arm_cspmu/Kconfig   |   6 +
 drivers/perf/arm_cspmu/Makefile  |   2 +
 drivers/perf/arm_cspmu/arm-bmu.c | 528 +++++++++++++++++++++++++++++++
 3 files changed, 536 insertions(+)
 create mode 100644 drivers/perf/arm_cspmu/arm-bmu.c

index 25bd55f036a4..48c2a3ac7fab 100644
@@ -34,3 +34,9 @@ config ARM_SYSTEM_PMU
 	help
 	  Support additional event filtering attributes for the CoreSight-based
 	  PMU blocks found in various parts of Arm Compute Subsystems and SoCs.
+
+config ARM_BMU
+	tristate "ARM Bus Monitor Unit"
+	depends on (ARM64 && ACPI) || COMPILE_TEST
+	help
+	  Support perf event counting on Arm Bus Monitor Unit devices.

index 90c3cd710157..4e0b420a2b55 100644
@@ -9,3 +9,5 @@ arm_cspmu_module-y := arm_cspmu.o
 obj-$(CONFIG_NVIDIA_CORESIGHT_PMU_ARCH_SYSTEM_PMU) += nvidia_cspmu.o
 obj-$(CONFIG_AMPERE_CORESIGHT_PMU_ARCH_SYSTEM_PMU) += ampere_cspmu.o
 obj-$(CONFIG_ARM_SYSTEM_PMU) += arm_syspmu.o
+
+obj-$(CONFIG_ARM_BMU) += arm-bmu.o

2.54.0.dirty

Unlikely to be an issue in practice for the moment, as the users who want arm-bmu also want arm_syspmu anyway (did you notice I sent them together as a series?). However the existing double-conditional compilation does seem a bit silly in general, so I've sent a quick patch to clean that up separately.

Thanks,
Robin.

Arm's Bus Monitor Unit is a low-level performance analysis tool for matching and counting transactions at interconnect interfaces. Typical platforms will have multiple BMUs each measuring multiple interfaces in various parts of the system, hence for the sake of usability we make it mandatory for the platform to label each instance such that the end user can correlate with platform documentation to know what's what.

---
 .../bindings/perf/arm,bus-monitor-unit.yaml   | 41 +++++++++++++++++++
 1 file changed, 41 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/perf/arm,bus-monitor-unit.yaml

new file mode 100644
index 000000000000..3da59e5447e0
@@ -0,0 +1,41 @@
+# SPDX-License-Identifier: GPL-2.0-only OR BSD-2-Clause
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/perf/arm,bus-monitor-unit.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: Arm Bus Monitor Unit
+
+maintainers:
+  - Robin Murphy <[email protected]>
+
+properties:
+  compatible:
+    const: arm,bus-monitor-unit
+
+  reg:
+    items:
+      - description: Full BMU register space
+
+  interrupts:
+    items:
+      - description: PMU overflow interrupt
+
+  label:
+    description: BMU instance identifier (same as ACPI _UID), per the
+      platform optimisation guide
+
+required:
+  - compatible
+  - reg
+  - label
+
+additionalProperties: false
+
+examples:
+  - |
+    pmu@12340000 {
+        compatible = "arm,bus-monitor-unit";
+        reg = <0x12340000 0x40000>;
+        label = "ddr_ch_1_2";
+    };

2.54.0.dirty

Initial production BMU users will be ACPI-based, but DT consumers will almost certainly come along sooner or later.

---
 drivers/perf/arm_cspmu/Kconfig   |  2 +-
 drivers/perf/arm_cspmu/arm-bmu.c | 15 ++++++++++++++-
 2 files changed, 15 insertions(+), 2 deletions(-)

index 48c2a3ac7fab..8ce05bbe6b9a 100644
@@ -37,6 +37,6 @@ config ARM_SYSTEM_PMU
 
 config ARM_BMU
 	tristate "ARM Bus Monitor Unit"
-	depends on (ARM64 && ACPI) || COMPILE_TEST
+	depends on ARM64 || COMPILE_TEST
 	help
 	  Support perf event counting on Arm Bus Monitor Unit devices.

index fe25954ac836..aa4ccbdae7a2 100644
@@ -6,6 +6,7 @@
 #include <linux/interrupt.h>
 #include <linux/io-64-nonatomic-lo-hi.h>
 #include <linux/module.h>
+#include <linux/of.h>
 #include <linux/perf_event.h>
 #include <linux/platform_device.h>
 #include <linux/property.h>

@@ -409,7 +410,10 @@ static int arm_bmu_probe(struct platform_device *pdev)
 		.read = arm_bmu_event_read,
 	};
 
-	name = acpi_device_uid(ACPI_COMPANION(dev));
+	if (has_acpi_companion(dev))
+		name = acpi_device_uid(ACPI_COMPANION(dev));
+	else
+		device_property_read_string(dev, "label", &name);
 
 	if (name)
 		name = devm_kasprintf(dev, GFP_KERNEL, "arm_bmu_%s", name);

@@ -440,6 +444,14 @@ static void arm_bmu_remove(struct platform_device *pdev)
 	cpuhp_state_remove_instance_nocalls(arm_bmu_cpuhp_state, &bmu->cpuhp_node);
 }
 
+#ifdef CONFIG_OF
+static const struct of_device_id arm_bmu_of_match[] = {
+	{ .compatible = "arm,bus-monitor-unit" },
+	{}
+};
+MODULE_DEVICE_TABLE(of, arm_bmu_of_match);
+#endif
+
 #ifdef CONFIG_ACPI
 static const struct acpi_device_id arm_bmu_acpi_match[] = {
 	{ "ARMHB001" },

@@ -451,6 +463,7 @@ MODULE_DEVICE_TABLE(acpi, arm_bmu_acpi_match);
 static struct platform_driver arm_bmu_driver = {
 	.driver = {
 		.name = "arm-bmu",
+		.of_match_table = of_match_ptr(arm_bmu_of_match),
 		.acpi_match_table = ACPI_PTR(arm_bmu_acpi_match),
 		.suppress_bind_attrs = true,
 	},

2.54.0.dirty