[PATCH v3 0/8] perf: Arm SPEv1.2 support

[PATCH v3 0/8] perf: Arm SPEv1.2 support

[Rob Herring]

This series adds support for Arm SPEv1.2 which is part of the
Armv8.7/Armv9.2 architecture. There's 2 new features that affect the 
kernel: a new event filter bit, branch 'not taken', and an inverted 
event filter register. 

Since this support adds new registers and fields, first the SPE register 
defines are converted to automatic generation.

Note that the 'config3' addition in sysfs format files causes SPE to 
break. A stable fix e552b7be12ed ("perf: Skip and warn on unknown format 
'configN' attrs") landed in v6.1-rc1.

The perf tool side changes are available here[1].

Tested on FVP.

[1] https://lore.kernel.org/all/20220914-arm-perf-tool-spe1-2-v2-v4-0-83c098e6212e@kernel.org/

Signed-off-by: Rob Herring <robh@kernel.org>
---
Changes in v3:
- Add some more missing SPE register fields and use Enums for some 
  fields
- Use the new PMSIDR_EL1 register Enum defines in the SPE driver
- Link to v2: https://lore.kernel.org/r/20220825-arm-spe-v8-7-v2-0-e37322d68ac0@kernel.org

Changes in v2:
- Convert the SPE register defines to automatic generation
- Fixed access to SYS_PMSNEVFR_EL1 when not present
- Rebase on v6.1-rc1
- Link to v1: https://lore.kernel.org/r/20220825-arm-spe-v8-7-v1-0-c75b8d92e692@kernel.org

---
Rob Herring (8):
      perf: arm_spe: Use feature numbering for PMSEVFR_EL1 defines
      arm64: Drop SYS_ from SPE register defines
      arm64/sysreg: Convert SPE registers to automatic generation
      perf: arm_spe: Drop BIT() and use FIELD_GET/PREP accessors
      perf: arm_spe: Use new PMSIDR_EL1 register enums
      perf: arm_spe: Support new SPEv1.2/v8.7 'not taken' event
      perf: Add perf_event_attr::config3
      perf: arm_spe: Add support for SPEv1.2 inverted event filtering

 arch/arm64/include/asm/el2_setup.h |   6 +-
 arch/arm64/include/asm/sysreg.h    |  99 +++--------------------
 arch/arm64/kvm/debug.c             |   2 +-
 arch/arm64/kvm/hyp/nvhe/debug-sr.c |   2 +-
 arch/arm64/tools/sysreg            | 139 +++++++++++++++++++++++++++++++++
 drivers/perf/arm_spe_pmu.c         | 156 ++++++++++++++++++++++++-------------
 include/uapi/linux/perf_event.h    |   3 +
 7 files changed, 257 insertions(+), 150 deletions(-)
---
base-commit: 9abf2313adc1ca1b6180c508c25f22f9395cc780
change-id: 20220825-arm-spe-v8-7-fedf04e16f23

Best regards,
-- 
Rob Herring <robh@kernel.org>

[PATCH v3 1/8] perf: arm_spe: Use feature numbering for PMSEVFR_EL1 defines

[Rob Herring]

Similar to commit 121a8fc088f1 ("arm64/sysreg: Use feature numbering for
PMU and SPE revisions") use feature numbering instead of architecture
versions for the PMSEVFR_EL1 Res0 defines.

Tested-by: James Clark <james.clark@arm.com>
Signed-off-by: Rob Herring <robh@kernel.org>
---
v3:
 - No change
v2:
 - New patch
---
 arch/arm64/include/asm/sysreg.h | 6 +++---
 drivers/perf/arm_spe_pmu.c      | 4 ++--
 2 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/arch/arm64/include/asm/sysreg.h b/arch/arm64/include/asm/sysreg.h
index 7d301700d1a9..9a4cf12e3e16 100644
--- a/arch/arm64/include/asm/sysreg.h
+++ b/arch/arm64/include/asm/sysreg.h
@@ -294,11 +294,11 @@
 #define SYS_PMSFCR_EL1_ST_SHIFT		18
 
 #define SYS_PMSEVFR_EL1			sys_reg(3, 0, 9, 9, 5)
-#define SYS_PMSEVFR_EL1_RES0_8_2	\
+#define PMSEVFR_EL1_RES0_IMP	\
 	(GENMASK_ULL(47, 32) | GENMASK_ULL(23, 16) | GENMASK_ULL(11, 8) |\
 	 BIT_ULL(6) | BIT_ULL(4) | BIT_ULL(2) | BIT_ULL(0))
-#define SYS_PMSEVFR_EL1_RES0_8_3	\
-	(SYS_PMSEVFR_EL1_RES0_8_2 & ~(BIT_ULL(18) | BIT_ULL(17) | BIT_ULL(11)))
+#define PMSEVFR_EL1_RES0_V1P1	\
+	(PMSEVFR_EL1_RES0_IMP & ~(BIT_ULL(18) | BIT_ULL(17) | BIT_ULL(11)))
 
 #define SYS_PMSLATFR_EL1		sys_reg(3, 0, 9, 9, 6)
 #define SYS_PMSLATFR_EL1_MINLAT_SHIFT	0
diff --git a/drivers/perf/arm_spe_pmu.c b/drivers/perf/arm_spe_pmu.c
index 00e3a637f7b6..65cf93dcc8ee 100644
--- a/drivers/perf/arm_spe_pmu.c
+++ b/drivers/perf/arm_spe_pmu.c
@@ -677,11 +677,11 @@ static u64 arm_spe_pmsevfr_res0(u16 pmsver)
 {
 	switch (pmsver) {
 	case ID_AA64DFR0_EL1_PMSVer_IMP:
-		return SYS_PMSEVFR_EL1_RES0_8_2;
+		return PMSEVFR_EL1_RES0_IMP;
 	case ID_AA64DFR0_EL1_PMSVer_V1P1:
 	/* Return the highest version we support in default */
 	default:
-		return SYS_PMSEVFR_EL1_RES0_8_3;
+		return PMSEVFR_EL1_RES0_V1P1;
 	}
 }
 

-- 
b4 0.11.0-dev

[PATCH v3 2/8] arm64: Drop SYS_ from SPE register defines

[Rob Herring]

We currently have a non-standard SYS_ prefix in the constants generated
for the SPE register bitfields. Drop this in preparation for automatic
register definition generation.

The SPE mask defines were unshifted, and the SPE register field
enumerations were shifted. The autogenerated defines are the opposite,
so make the necessary adjustments.

No functional changes.

Tested-by: James Clark <james.clark@arm.com>
Signed-off-by: Rob Herring <robh@kernel.org>
---
v3:
 - No change
v2:
 - New patch
---
 arch/arm64/include/asm/el2_setup.h |   6 +-
 arch/arm64/include/asm/sysreg.h    | 112 ++++++++++++++++++-------------------
 arch/arm64/kvm/debug.c             |   2 +-
 arch/arm64/kvm/hyp/nvhe/debug-sr.c |   2 +-
 drivers/perf/arm_spe_pmu.c         |  85 ++++++++++++++--------------
 5 files changed, 103 insertions(+), 104 deletions(-)

diff --git a/arch/arm64/include/asm/el2_setup.h b/arch/arm64/include/asm/el2_setup.h
index 668569adf4d3..f9da43e53cdb 100644
--- a/arch/arm64/include/asm/el2_setup.h
+++ b/arch/arm64/include/asm/el2_setup.h
@@ -53,10 +53,10 @@
 	cbz	x0, .Lskip_spe_\@		// Skip if SPE not present
 
 	mrs_s	x0, SYS_PMBIDR_EL1              // If SPE available at EL2,
-	and	x0, x0, #(1 << SYS_PMBIDR_EL1_P_SHIFT)
+	and	x0, x0, #(1 << PMBIDR_EL1_P_SHIFT)
 	cbnz	x0, .Lskip_spe_el2_\@		// then permit sampling of physical
-	mov	x0, #(1 << SYS_PMSCR_EL2_PCT_SHIFT | \
-		      1 << SYS_PMSCR_EL2_PA_SHIFT)
+	mov	x0, #(1 << PMSCR_EL2_PCT_SHIFT | \
+		      1 << PMSCR_EL2_PA_SHIFT)
 	msr_s	SYS_PMSCR_EL2, x0		// addresses and physical counter
 .Lskip_spe_el2_\@:
 	mov	x0, #(MDCR_EL2_E2PB_MASK << MDCR_EL2_E2PB_SHIFT)
diff --git a/arch/arm64/include/asm/sysreg.h b/arch/arm64/include/asm/sysreg.h
index 9a4cf12e3e16..8df8a0a51273 100644
--- a/arch/arm64/include/asm/sysreg.h
+++ b/arch/arm64/include/asm/sysreg.h
@@ -239,59 +239,59 @@
 /*** Statistical Profiling Extension ***/
 /* ID registers */
 #define SYS_PMSIDR_EL1			sys_reg(3, 0, 9, 9, 7)
-#define SYS_PMSIDR_EL1_FE_SHIFT		0
-#define SYS_PMSIDR_EL1_FT_SHIFT		1
-#define SYS_PMSIDR_EL1_FL_SHIFT		2
-#define SYS_PMSIDR_EL1_ARCHINST_SHIFT	3
-#define SYS_PMSIDR_EL1_LDS_SHIFT	4
-#define SYS_PMSIDR_EL1_ERND_SHIFT	5
-#define SYS_PMSIDR_EL1_INTERVAL_SHIFT	8
-#define SYS_PMSIDR_EL1_INTERVAL_MASK	0xfUL
-#define SYS_PMSIDR_EL1_MAXSIZE_SHIFT	12
-#define SYS_PMSIDR_EL1_MAXSIZE_MASK	0xfUL
-#define SYS_PMSIDR_EL1_COUNTSIZE_SHIFT	16
-#define SYS_PMSIDR_EL1_COUNTSIZE_MASK	0xfUL
+#define PMSIDR_EL1_FE_SHIFT		0
+#define PMSIDR_EL1_FT_SHIFT		1
+#define PMSIDR_EL1_FL_SHIFT		2
+#define PMSIDR_EL1_ARCHINST_SHIFT	3
+#define PMSIDR_EL1_LDS_SHIFT	4
+#define PMSIDR_EL1_ERND_SHIFT	5
+#define PMSIDR_EL1_INTERVAL_SHIFT	8
+#define PMSIDR_EL1_INTERVAL_MASK	GENMASK_ULL(11, 8)
+#define PMSIDR_EL1_MAXSIZE_SHIFT	12
+#define PMSIDR_EL1_MAXSIZE_MASK		GENMASK_ULL(15, 12)
+#define PMSIDR_EL1_COUNTSIZE_SHIFT	16
+#define PMSIDR_EL1_COUNTSIZE_MASK	GENMASK_ULL(19, 16)
 
 #define SYS_PMBIDR_EL1			sys_reg(3, 0, 9, 10, 7)
-#define SYS_PMBIDR_EL1_ALIGN_SHIFT	0
-#define SYS_PMBIDR_EL1_ALIGN_MASK	0xfU
-#define SYS_PMBIDR_EL1_P_SHIFT		4
-#define SYS_PMBIDR_EL1_F_SHIFT		5
+#define PMBIDR_EL1_ALIGN_SHIFT		0
+#define PMBIDR_EL1_ALIGN_MASK		0xfU
+#define PMBIDR_EL1_P_SHIFT		4
+#define PMBIDR_EL1_F_SHIFT		5
 
 /* Sampling controls */
 #define SYS_PMSCR_EL1			sys_reg(3, 0, 9, 9, 0)
-#define SYS_PMSCR_EL1_E0SPE_SHIFT	0
-#define SYS_PMSCR_EL1_E1SPE_SHIFT	1
-#define SYS_PMSCR_EL1_CX_SHIFT		3
-#define SYS_PMSCR_EL1_PA_SHIFT		4
-#define SYS_PMSCR_EL1_TS_SHIFT		5
-#define SYS_PMSCR_EL1_PCT_SHIFT		6
+#define PMSCR_EL1_E0SPE_SHIFT		0
+#define PMSCR_EL1_E1SPE_SHIFT		1
+#define PMSCR_EL1_CX_SHIFT		3
+#define PMSCR_EL1_PA_SHIFT		4
+#define PMSCR_EL1_TS_SHIFT		5
+#define PMSCR_EL1_PCT_SHIFT		6
 
 #define SYS_PMSCR_EL2			sys_reg(3, 4, 9, 9, 0)
-#define SYS_PMSCR_EL2_E0HSPE_SHIFT	0
-#define SYS_PMSCR_EL2_E2SPE_SHIFT	1
-#define SYS_PMSCR_EL2_CX_SHIFT		3
-#define SYS_PMSCR_EL2_PA_SHIFT		4
-#define SYS_PMSCR_EL2_TS_SHIFT		5
-#define SYS_PMSCR_EL2_PCT_SHIFT		6
+#define PMSCR_EL2_E0HSPE_SHIFT		0
+#define PMSCR_EL2_E2SPE_SHIFT		1
+#define PMSCR_EL2_CX_SHIFT		3
+#define PMSCR_EL2_PA_SHIFT		4
+#define PMSCR_EL2_TS_SHIFT		5
+#define PMSCR_EL2_PCT_SHIFT		6
 
 #define SYS_PMSICR_EL1			sys_reg(3, 0, 9, 9, 2)
 
 #define SYS_PMSIRR_EL1			sys_reg(3, 0, 9, 9, 3)
-#define SYS_PMSIRR_EL1_RND_SHIFT	0
-#define SYS_PMSIRR_EL1_INTERVAL_SHIFT	8
-#define SYS_PMSIRR_EL1_INTERVAL_MASK	0xffffffUL
+#define PMSIRR_EL1_RND_SHIFT		0
+#define PMSIRR_EL1_INTERVAL_SHIFT	8
+#define PMSIRR_EL1_INTERVAL_MASK	GENMASK_ULL(31, 8)
 
 /* Filtering controls */
 #define SYS_PMSNEVFR_EL1		sys_reg(3, 0, 9, 9, 1)
 
 #define SYS_PMSFCR_EL1			sys_reg(3, 0, 9, 9, 4)
-#define SYS_PMSFCR_EL1_FE_SHIFT		0
-#define SYS_PMSFCR_EL1_FT_SHIFT		1
-#define SYS_PMSFCR_EL1_FL_SHIFT		2
-#define SYS_PMSFCR_EL1_B_SHIFT		16
-#define SYS_PMSFCR_EL1_LD_SHIFT		17
-#define SYS_PMSFCR_EL1_ST_SHIFT		18
+#define PMSFCR_EL1_FE_SHIFT		0
+#define PMSFCR_EL1_FT_SHIFT		1
+#define PMSFCR_EL1_FL_SHIFT		2
+#define PMSFCR_EL1_B_SHIFT		16
+#define PMSFCR_EL1_LD_SHIFT		17
+#define PMSFCR_EL1_ST_SHIFT		18
 
 #define SYS_PMSEVFR_EL1			sys_reg(3, 0, 9, 9, 5)
 #define PMSEVFR_EL1_RES0_IMP	\
@@ -301,37 +301,37 @@
 	(PMSEVFR_EL1_RES0_IMP & ~(BIT_ULL(18) | BIT_ULL(17) | BIT_ULL(11)))
 
 #define SYS_PMSLATFR_EL1		sys_reg(3, 0, 9, 9, 6)
-#define SYS_PMSLATFR_EL1_MINLAT_SHIFT	0
+#define PMSLATFR_EL1_MINLAT_SHIFT	0
 
 /* Buffer controls */
 #define SYS_PMBLIMITR_EL1		sys_reg(3, 0, 9, 10, 0)
-#define SYS_PMBLIMITR_EL1_E_SHIFT	0
-#define SYS_PMBLIMITR_EL1_FM_SHIFT	1
-#define SYS_PMBLIMITR_EL1_FM_MASK	0x3UL
-#define SYS_PMBLIMITR_EL1_FM_STOP_IRQ	(0 << SYS_PMBLIMITR_EL1_FM_SHIFT)
+#define PMBLIMITR_EL1_E_SHIFT		0
+#define PMBLIMITR_EL1_FM_SHIFT		1
+#define PMBLIMITR_EL1_FM_MASK	GENMASK_ULL(2, 1)
+#define PMBLIMITR_EL1_FM_STOP_IRQ	0
 
 #define SYS_PMBPTR_EL1			sys_reg(3, 0, 9, 10, 1)
 
 /* Buffer error reporting */
 #define SYS_PMBSR_EL1			sys_reg(3, 0, 9, 10, 3)
-#define SYS_PMBSR_EL1_COLL_SHIFT	16
-#define SYS_PMBSR_EL1_S_SHIFT		17
-#define SYS_PMBSR_EL1_EA_SHIFT		18
-#define SYS_PMBSR_EL1_DL_SHIFT		19
-#define SYS_PMBSR_EL1_EC_SHIFT		26
-#define SYS_PMBSR_EL1_EC_MASK		0x3fUL
+#define PMBSR_EL1_COLL_SHIFT		16
+#define PMBSR_EL1_S_SHIFT		17
+#define PMBSR_EL1_EA_SHIFT		18
+#define PMBSR_EL1_DL_SHIFT		19
+#define PMBSR_EL1_EC_SHIFT		26
+#define PMBSR_EL1_EC_MASK		GENMASK_ULL(31, 26)
 
-#define SYS_PMBSR_EL1_EC_BUF		(0x0UL << SYS_PMBSR_EL1_EC_SHIFT)
-#define SYS_PMBSR_EL1_EC_FAULT_S1	(0x24UL << SYS_PMBSR_EL1_EC_SHIFT)
-#define SYS_PMBSR_EL1_EC_FAULT_S2	(0x25UL << SYS_PMBSR_EL1_EC_SHIFT)
+#define PMBSR_EL1_EC_BUF		0x0UL
+#define PMBSR_EL1_EC_FAULT_S1		0x24UL
+#define PMBSR_EL1_EC_FAULT_S2		0x25UL
 
-#define SYS_PMBSR_EL1_FAULT_FSC_SHIFT	0
-#define SYS_PMBSR_EL1_FAULT_FSC_MASK	0x3fUL
+#define PMBSR_EL1_FAULT_FSC_SHIFT	0
+#define PMBSR_EL1_FAULT_FSC_MASK	0x3fUL
 
-#define SYS_PMBSR_EL1_BUF_BSC_SHIFT	0
-#define SYS_PMBSR_EL1_BUF_BSC_MASK	0x3fUL
+#define PMBSR_EL1_BUF_BSC_SHIFT		0
+#define PMBSR_EL1_BUF_BSC_MASK		0x3fUL
 
-#define SYS_PMBSR_EL1_BUF_BSC_FULL	(0x1UL << SYS_PMBSR_EL1_BUF_BSC_SHIFT)
+#define PMBSR_EL1_BUF_BSC_FULL		0x1UL
 
 /*** End of Statistical Profiling Extension ***/
 
diff --git a/arch/arm64/kvm/debug.c b/arch/arm64/kvm/debug.c
index fccf9ec01813..55f80fb93925 100644
--- a/arch/arm64/kvm/debug.c
+++ b/arch/arm64/kvm/debug.c
@@ -328,7 +328,7 @@ void kvm_arch_vcpu_load_debug_state_flags(struct kvm_vcpu *vcpu)
 	 * we may need to check if the host state needs to be saved.
 	 */
 	if (cpuid_feature_extract_unsigned_field(dfr0, ID_AA64DFR0_EL1_PMSVer_SHIFT) &&
-	    !(read_sysreg_s(SYS_PMBIDR_EL1) & BIT(SYS_PMBIDR_EL1_P_SHIFT)))
+	    !(read_sysreg_s(SYS_PMBIDR_EL1) & BIT(PMBIDR_EL1_P_SHIFT)))
 		vcpu_set_flag(vcpu, DEBUG_STATE_SAVE_SPE);
 
 	/* Check if we have TRBE implemented and available at the host */
diff --git a/arch/arm64/kvm/hyp/nvhe/debug-sr.c b/arch/arm64/kvm/hyp/nvhe/debug-sr.c
index e17455773b98..2673bde62fad 100644
--- a/arch/arm64/kvm/hyp/nvhe/debug-sr.c
+++ b/arch/arm64/kvm/hyp/nvhe/debug-sr.c
@@ -27,7 +27,7 @@ static void __debug_save_spe(u64 *pmscr_el1)
 	 * Check if the host is actually using it ?
 	 */
 	reg = read_sysreg_s(SYS_PMBLIMITR_EL1);
-	if (!(reg & BIT(SYS_PMBLIMITR_EL1_E_SHIFT)))
+	if (!(reg & BIT(PMBLIMITR_EL1_E_SHIFT)))
 		return;
 
 	/* Yes; save the control register and disable data generation */
diff --git a/drivers/perf/arm_spe_pmu.c b/drivers/perf/arm_spe_pmu.c
index 65cf93dcc8ee..814ed18346b6 100644
--- a/drivers/perf/arm_spe_pmu.c
+++ b/drivers/perf/arm_spe_pmu.c
@@ -12,6 +12,7 @@
 #define DRVNAME					PMUNAME "_pmu"
 #define pr_fmt(fmt)				DRVNAME ": " fmt
 
+#include <linux/bitfield.h>
 #include <linux/bitops.h>
 #include <linux/bug.h>
 #include <linux/capability.h>
@@ -282,18 +283,18 @@ static u64 arm_spe_event_to_pmscr(struct perf_event *event)
 	struct perf_event_attr *attr = &event->attr;
 	u64 reg = 0;
 
-	reg |= ATTR_CFG_GET_FLD(attr, ts_enable) << SYS_PMSCR_EL1_TS_SHIFT;
-	reg |= ATTR_CFG_GET_FLD(attr, pa_enable) << SYS_PMSCR_EL1_PA_SHIFT;
-	reg |= ATTR_CFG_GET_FLD(attr, pct_enable) << SYS_PMSCR_EL1_PCT_SHIFT;
+	reg |= ATTR_CFG_GET_FLD(attr, ts_enable) << PMSCR_EL1_TS_SHIFT;
+	reg |= ATTR_CFG_GET_FLD(attr, pa_enable) << PMSCR_EL1_PA_SHIFT;
+	reg |= ATTR_CFG_GET_FLD(attr, pct_enable) << PMSCR_EL1_PCT_SHIFT;
 
 	if (!attr->exclude_user)
-		reg |= BIT(SYS_PMSCR_EL1_E0SPE_SHIFT);
+		reg |= BIT(PMSCR_EL1_E0SPE_SHIFT);
 
 	if (!attr->exclude_kernel)
-		reg |= BIT(SYS_PMSCR_EL1_E1SPE_SHIFT);
+		reg |= BIT(PMSCR_EL1_E1SPE_SHIFT);
 
 	if (get_spe_event_has_cx(event))
-		reg |= BIT(SYS_PMSCR_EL1_CX_SHIFT);
+		reg |= BIT(PMSCR_EL1_CX_SHIFT);
 
 	return reg;
 }
@@ -302,8 +303,7 @@ static void arm_spe_event_sanitise_period(struct perf_event *event)
 {
 	struct arm_spe_pmu *spe_pmu = to_spe_pmu(event->pmu);
 	u64 period = event->hw.sample_period;
-	u64 max_period = SYS_PMSIRR_EL1_INTERVAL_MASK
-			 << SYS_PMSIRR_EL1_INTERVAL_SHIFT;
+	u64 max_period = PMSIRR_EL1_INTERVAL_MASK;
 
 	if (period < spe_pmu->min_period)
 		period = spe_pmu->min_period;
@@ -322,7 +322,7 @@ static u64 arm_spe_event_to_pmsirr(struct perf_event *event)
 
 	arm_spe_event_sanitise_period(event);
 
-	reg |= ATTR_CFG_GET_FLD(attr, jitter) << SYS_PMSIRR_EL1_RND_SHIFT;
+	reg |= ATTR_CFG_GET_FLD(attr, jitter) << PMSIRR_EL1_RND_SHIFT;
 	reg |= event->hw.sample_period;
 
 	return reg;
@@ -333,18 +333,18 @@ static u64 arm_spe_event_to_pmsfcr(struct perf_event *event)
 	struct perf_event_attr *attr = &event->attr;
 	u64 reg = 0;
 
-	reg |= ATTR_CFG_GET_FLD(attr, load_filter) << SYS_PMSFCR_EL1_LD_SHIFT;
-	reg |= ATTR_CFG_GET_FLD(attr, store_filter) << SYS_PMSFCR_EL1_ST_SHIFT;
-	reg |= ATTR_CFG_GET_FLD(attr, branch_filter) << SYS_PMSFCR_EL1_B_SHIFT;
+	reg |= ATTR_CFG_GET_FLD(attr, load_filter) << PMSFCR_EL1_LD_SHIFT;
+	reg |= ATTR_CFG_GET_FLD(attr, store_filter) << PMSFCR_EL1_ST_SHIFT;
+	reg |= ATTR_CFG_GET_FLD(attr, branch_filter) << PMSFCR_EL1_B_SHIFT;
 
 	if (reg)
-		reg |= BIT(SYS_PMSFCR_EL1_FT_SHIFT);
+		reg |= BIT(PMSFCR_EL1_FT_SHIFT);
 
 	if (ATTR_CFG_GET_FLD(attr, event_filter))
-		reg |= BIT(SYS_PMSFCR_EL1_FE_SHIFT);
+		reg |= BIT(PMSFCR_EL1_FE_SHIFT);
 
 	if (ATTR_CFG_GET_FLD(attr, min_latency))
-		reg |= BIT(SYS_PMSFCR_EL1_FL_SHIFT);
+		reg |= BIT(PMSFCR_EL1_FL_SHIFT);
 
 	return reg;
 }
@@ -359,7 +359,7 @@ static u64 arm_spe_event_to_pmslatfr(struct perf_event *event)
 {
 	struct perf_event_attr *attr = &event->attr;
 	return ATTR_CFG_GET_FLD(attr, min_latency)
-	       << SYS_PMSLATFR_EL1_MINLAT_SHIFT;
+	       << PMSLATFR_EL1_MINLAT_SHIFT;
 }
 
 static void arm_spe_pmu_pad_buf(struct perf_output_handle *handle, int len)
@@ -511,7 +511,7 @@ static void arm_spe_perf_aux_output_begin(struct perf_output_handle *handle,
 	limit = buf->snapshot ? arm_spe_pmu_next_snapshot_off(handle)
 			      : arm_spe_pmu_next_off(handle);
 	if (limit)
-		limit |= BIT(SYS_PMBLIMITR_EL1_E_SHIFT);
+		limit |= BIT(PMBLIMITR_EL1_E_SHIFT);
 
 	limit += (u64)buf->base;
 	base = (u64)buf->base + PERF_IDX2OFF(handle->head, buf);
@@ -570,28 +570,28 @@ arm_spe_pmu_buf_get_fault_act(struct perf_output_handle *handle)
 
 	/* Service required? */
 	pmbsr = read_sysreg_s(SYS_PMBSR_EL1);
-	if (!(pmbsr & BIT(SYS_PMBSR_EL1_S_SHIFT)))
+	if (!(pmbsr & BIT(PMBSR_EL1_S_SHIFT)))
 		return SPE_PMU_BUF_FAULT_ACT_SPURIOUS;
 
 	/*
 	 * If we've lost data, disable profiling and also set the PARTIAL
 	 * flag to indicate that the last record is corrupted.
 	 */
-	if (pmbsr & BIT(SYS_PMBSR_EL1_DL_SHIFT))
+	if (pmbsr & BIT(PMBSR_EL1_DL_SHIFT))
 		perf_aux_output_flag(handle, PERF_AUX_FLAG_TRUNCATED |
 					     PERF_AUX_FLAG_PARTIAL);
 
 	/* Report collisions to userspace so that it can up the period */
-	if (pmbsr & BIT(SYS_PMBSR_EL1_COLL_SHIFT))
+	if (pmbsr & BIT(PMBSR_EL1_COLL_SHIFT))
 		perf_aux_output_flag(handle, PERF_AUX_FLAG_COLLISION);
 
 	/* We only expect buffer management events */
-	switch (pmbsr & (SYS_PMBSR_EL1_EC_MASK << SYS_PMBSR_EL1_EC_SHIFT)) {
-	case SYS_PMBSR_EL1_EC_BUF:
+	switch (FIELD_GET(PMBSR_EL1_EC_MASK, pmbsr)) {
+	case PMBSR_EL1_EC_BUF:
 		/* Handled below */
 		break;
-	case SYS_PMBSR_EL1_EC_FAULT_S1:
-	case SYS_PMBSR_EL1_EC_FAULT_S2:
+	case PMBSR_EL1_EC_FAULT_S1:
+	case PMBSR_EL1_EC_FAULT_S2:
 		err_str = "Unexpected buffer fault";
 		goto out_err;
 	default:
@@ -600,9 +600,8 @@ arm_spe_pmu_buf_get_fault_act(struct perf_output_handle *handle)
 	}
 
 	/* Buffer management event */
-	switch (pmbsr &
-		(SYS_PMBSR_EL1_BUF_BSC_MASK << SYS_PMBSR_EL1_BUF_BSC_SHIFT)) {
-	case SYS_PMBSR_EL1_BUF_BSC_FULL:
+	switch (FIELD_GET(PMBSR_EL1_BUF_BSC_MASK, pmbsr)) {
+	case PMBSR_EL1_BUF_BSC_FULL:
 		ret = SPE_PMU_BUF_FAULT_ACT_OK;
 		goto out_stop;
 	default:
@@ -717,23 +716,23 @@ static int arm_spe_pmu_event_init(struct perf_event *event)
 		return -EINVAL;
 
 	reg = arm_spe_event_to_pmsfcr(event);
-	if ((reg & BIT(SYS_PMSFCR_EL1_FE_SHIFT)) &&
+	if ((reg & BIT(PMSFCR_EL1_FE_SHIFT)) &&
 	    !(spe_pmu->features & SPE_PMU_FEAT_FILT_EVT))
 		return -EOPNOTSUPP;
 
-	if ((reg & BIT(SYS_PMSFCR_EL1_FT_SHIFT)) &&
+	if ((reg & BIT(PMSFCR_EL1_FT_SHIFT)) &&
 	    !(spe_pmu->features & SPE_PMU_FEAT_FILT_TYP))
 		return -EOPNOTSUPP;
 
-	if ((reg & BIT(SYS_PMSFCR_EL1_FL_SHIFT)) &&
+	if ((reg & BIT(PMSFCR_EL1_FL_SHIFT)) &&
 	    !(spe_pmu->features & SPE_PMU_FEAT_FILT_LAT))
 		return -EOPNOTSUPP;
 
 	set_spe_event_has_cx(event);
 	reg = arm_spe_event_to_pmscr(event);
 	if (!perfmon_capable() &&
-	    (reg & (BIT(SYS_PMSCR_EL1_PA_SHIFT) |
-		    BIT(SYS_PMSCR_EL1_PCT_SHIFT))))
+	    (reg & (BIT(PMSCR_EL1_PA_SHIFT) |
+		    BIT(PMSCR_EL1_PCT_SHIFT))))
 		return -EACCES;
 
 	return 0;
@@ -971,14 +970,14 @@ static void __arm_spe_pmu_dev_probe(void *info)
 
 	/* Read PMBIDR first to determine whether or not we have access */
 	reg = read_sysreg_s(SYS_PMBIDR_EL1);
-	if (reg & BIT(SYS_PMBIDR_EL1_P_SHIFT)) {
+	if (reg & BIT(PMBIDR_EL1_P_SHIFT)) {
 		dev_err(dev,
 			"profiling buffer owned by higher exception level\n");
 		return;
 	}
 
 	/* Minimum alignment. If it's out-of-range, then fail the probe */
-	fld = reg >> SYS_PMBIDR_EL1_ALIGN_SHIFT & SYS_PMBIDR_EL1_ALIGN_MASK;
+	fld = (reg & PMBIDR_EL1_ALIGN_MASK) >> PMBIDR_EL1_ALIGN_SHIFT;
 	spe_pmu->align = 1 << fld;
 	if (spe_pmu->align > SZ_2K) {
 		dev_err(dev, "unsupported PMBIDR.Align [%d] on CPU %d\n",
@@ -988,26 +987,26 @@ static void __arm_spe_pmu_dev_probe(void *info)
 
 	/* It's now safe to read PMSIDR and figure out what we've got */
 	reg = read_sysreg_s(SYS_PMSIDR_EL1);
-	if (reg & BIT(SYS_PMSIDR_EL1_FE_SHIFT))
+	if (reg & BIT(PMSIDR_EL1_FE_SHIFT))
 		spe_pmu->features |= SPE_PMU_FEAT_FILT_EVT;
 
-	if (reg & BIT(SYS_PMSIDR_EL1_FT_SHIFT))
+	if (reg & BIT(PMSIDR_EL1_FT_SHIFT))
 		spe_pmu->features |= SPE_PMU_FEAT_FILT_TYP;
 
-	if (reg & BIT(SYS_PMSIDR_EL1_FL_SHIFT))
+	if (reg & BIT(PMSIDR_EL1_FL_SHIFT))
 		spe_pmu->features |= SPE_PMU_FEAT_FILT_LAT;
 
-	if (reg & BIT(SYS_PMSIDR_EL1_ARCHINST_SHIFT))
+	if (reg & BIT(PMSIDR_EL1_ARCHINST_SHIFT))
 		spe_pmu->features |= SPE_PMU_FEAT_ARCH_INST;
 
-	if (reg & BIT(SYS_PMSIDR_EL1_LDS_SHIFT))
+	if (reg & BIT(PMSIDR_EL1_LDS_SHIFT))
 		spe_pmu->features |= SPE_PMU_FEAT_LDS;
 
-	if (reg & BIT(SYS_PMSIDR_EL1_ERND_SHIFT))
+	if (reg & BIT(PMSIDR_EL1_ERND_SHIFT))
 		spe_pmu->features |= SPE_PMU_FEAT_ERND;
 
 	/* This field has a spaced out encoding, so just use a look-up */
-	fld = reg >> SYS_PMSIDR_EL1_INTERVAL_SHIFT & SYS_PMSIDR_EL1_INTERVAL_MASK;
+	fld = (reg & PMSIDR_EL1_INTERVAL_MASK) >> PMSIDR_EL1_INTERVAL_SHIFT;
 	switch (fld) {
 	case 0:
 		spe_pmu->min_period = 256;
@@ -1039,7 +1038,7 @@ static void __arm_spe_pmu_dev_probe(void *info)
 	}
 
 	/* Maximum record size. If it's out-of-range, then fail the probe */
-	fld = reg >> SYS_PMSIDR_EL1_MAXSIZE_SHIFT & SYS_PMSIDR_EL1_MAXSIZE_MASK;
+	fld = (reg & PMSIDR_EL1_MAXSIZE_MASK) >> PMSIDR_EL1_MAXSIZE_SHIFT;
 	spe_pmu->max_record_sz = 1 << fld;
 	if (spe_pmu->max_record_sz > SZ_2K || spe_pmu->max_record_sz < 16) {
 		dev_err(dev, "unsupported PMSIDR_EL1.MaxSize [%d] on CPU %d\n",
@@ -1047,7 +1046,7 @@ static void __arm_spe_pmu_dev_probe(void *info)
 		return;
 	}
 
-	fld = reg >> SYS_PMSIDR_EL1_COUNTSIZE_SHIFT & SYS_PMSIDR_EL1_COUNTSIZE_MASK;
+	fld = (reg & PMSIDR_EL1_COUNTSIZE_MASK) >> PMSIDR_EL1_COUNTSIZE_SHIFT;
 	switch (fld) {
 	default:
 		dev_warn(dev, "unknown PMSIDR_EL1.CountSize [%d]; assuming 2\n",

-- 
b4 0.11.0-dev

[PATCH v3 3/8] arm64/sysreg: Convert SPE registers to automatic generation

[Rob Herring]

Convert all the SPE register defines to automatic generation. No
functional changes.

New registers and fields for SPEv1.2 are added with the conversion.

Some of the PMBSR MSS field defines are kept as the automatic generation
has no way to create multiple names for the same register bits. The
meaning of the MSS field depends on other bits.

Tested-by: James Clark <james.clark@arm.com>
Signed-off-by: Rob Herring <robh@kernel.org>
---
v3:
 - Make some fields enums and add some missing fields
v2:
 - New patch
---
 arch/arm64/include/asm/sysreg.h |  91 ++------------------------
 arch/arm64/tools/sysreg         | 139 ++++++++++++++++++++++++++++++++++++++++
 2 files changed, 144 insertions(+), 86 deletions(-)

diff --git a/arch/arm64/include/asm/sysreg.h b/arch/arm64/include/asm/sysreg.h
index 8df8a0a51273..d002dd00e53e 100644
--- a/arch/arm64/include/asm/sysreg.h
+++ b/arch/arm64/include/asm/sysreg.h
@@ -237,99 +237,18 @@
 #define SYS_PAR_EL1_FST			GENMASK(6, 1)
 
 /*** Statistical Profiling Extension ***/
-/* ID registers */
-#define SYS_PMSIDR_EL1			sys_reg(3, 0, 9, 9, 7)
-#define PMSIDR_EL1_FE_SHIFT		0
-#define PMSIDR_EL1_FT_SHIFT		1
-#define PMSIDR_EL1_FL_SHIFT		2
-#define PMSIDR_EL1_ARCHINST_SHIFT	3
-#define PMSIDR_EL1_LDS_SHIFT	4
-#define PMSIDR_EL1_ERND_SHIFT	5
-#define PMSIDR_EL1_INTERVAL_SHIFT	8
-#define PMSIDR_EL1_INTERVAL_MASK	GENMASK_ULL(11, 8)
-#define PMSIDR_EL1_MAXSIZE_SHIFT	12
-#define PMSIDR_EL1_MAXSIZE_MASK		GENMASK_ULL(15, 12)
-#define PMSIDR_EL1_COUNTSIZE_SHIFT	16
-#define PMSIDR_EL1_COUNTSIZE_MASK	GENMASK_ULL(19, 16)
-
-#define SYS_PMBIDR_EL1			sys_reg(3, 0, 9, 10, 7)
-#define PMBIDR_EL1_ALIGN_SHIFT		0
-#define PMBIDR_EL1_ALIGN_MASK		0xfU
-#define PMBIDR_EL1_P_SHIFT		4
-#define PMBIDR_EL1_F_SHIFT		5
-
-/* Sampling controls */
-#define SYS_PMSCR_EL1			sys_reg(3, 0, 9, 9, 0)
-#define PMSCR_EL1_E0SPE_SHIFT		0
-#define PMSCR_EL1_E1SPE_SHIFT		1
-#define PMSCR_EL1_CX_SHIFT		3
-#define PMSCR_EL1_PA_SHIFT		4
-#define PMSCR_EL1_TS_SHIFT		5
-#define PMSCR_EL1_PCT_SHIFT		6
-
-#define SYS_PMSCR_EL2			sys_reg(3, 4, 9, 9, 0)
-#define PMSCR_EL2_E0HSPE_SHIFT		0
-#define PMSCR_EL2_E2SPE_SHIFT		1
-#define PMSCR_EL2_CX_SHIFT		3
-#define PMSCR_EL2_PA_SHIFT		4
-#define PMSCR_EL2_TS_SHIFT		5
-#define PMSCR_EL2_PCT_SHIFT		6
-
-#define SYS_PMSICR_EL1			sys_reg(3, 0, 9, 9, 2)
-
-#define SYS_PMSIRR_EL1			sys_reg(3, 0, 9, 9, 3)
-#define PMSIRR_EL1_RND_SHIFT		0
-#define PMSIRR_EL1_INTERVAL_SHIFT	8
-#define PMSIRR_EL1_INTERVAL_MASK	GENMASK_ULL(31, 8)
-
-/* Filtering controls */
-#define SYS_PMSNEVFR_EL1		sys_reg(3, 0, 9, 9, 1)
-
-#define SYS_PMSFCR_EL1			sys_reg(3, 0, 9, 9, 4)
-#define PMSFCR_EL1_FE_SHIFT		0
-#define PMSFCR_EL1_FT_SHIFT		1
-#define PMSFCR_EL1_FL_SHIFT		2
-#define PMSFCR_EL1_B_SHIFT		16
-#define PMSFCR_EL1_LD_SHIFT		17
-#define PMSFCR_EL1_ST_SHIFT		18
-
-#define SYS_PMSEVFR_EL1			sys_reg(3, 0, 9, 9, 5)
 #define PMSEVFR_EL1_RES0_IMP	\
 	(GENMASK_ULL(47, 32) | GENMASK_ULL(23, 16) | GENMASK_ULL(11, 8) |\
 	 BIT_ULL(6) | BIT_ULL(4) | BIT_ULL(2) | BIT_ULL(0))
 #define PMSEVFR_EL1_RES0_V1P1	\
 	(PMSEVFR_EL1_RES0_IMP & ~(BIT_ULL(18) | BIT_ULL(17) | BIT_ULL(11)))
 
-#define SYS_PMSLATFR_EL1		sys_reg(3, 0, 9, 9, 6)
-#define PMSLATFR_EL1_MINLAT_SHIFT	0
-
-/* Buffer controls */
-#define SYS_PMBLIMITR_EL1		sys_reg(3, 0, 9, 10, 0)
-#define PMBLIMITR_EL1_E_SHIFT		0
-#define PMBLIMITR_EL1_FM_SHIFT		1
-#define PMBLIMITR_EL1_FM_MASK	GENMASK_ULL(2, 1)
-#define PMBLIMITR_EL1_FM_STOP_IRQ	0
-
-#define SYS_PMBPTR_EL1			sys_reg(3, 0, 9, 10, 1)
-
 /* Buffer error reporting */
-#define SYS_PMBSR_EL1			sys_reg(3, 0, 9, 10, 3)
-#define PMBSR_EL1_COLL_SHIFT		16
-#define PMBSR_EL1_S_SHIFT		17
-#define PMBSR_EL1_EA_SHIFT		18
-#define PMBSR_EL1_DL_SHIFT		19
-#define PMBSR_EL1_EC_SHIFT		26
-#define PMBSR_EL1_EC_MASK		GENMASK_ULL(31, 26)
-
-#define PMBSR_EL1_EC_BUF		0x0UL
-#define PMBSR_EL1_EC_FAULT_S1		0x24UL
-#define PMBSR_EL1_EC_FAULT_S2		0x25UL
-
-#define PMBSR_EL1_FAULT_FSC_SHIFT	0
-#define PMBSR_EL1_FAULT_FSC_MASK	0x3fUL
-
-#define PMBSR_EL1_BUF_BSC_SHIFT		0
-#define PMBSR_EL1_BUF_BSC_MASK		0x3fUL
+#define PMBSR_EL1_FAULT_FSC_SHIFT	PMBSR_EL1_MSS_SHIFT
+#define PMBSR_EL1_FAULT_FSC_MASK	PMBSR_EL1_MSS_MASK
+
+#define PMBSR_EL1_BUF_BSC_SHIFT		PMBSR_EL1_MSS_SHIFT
+#define PMBSR_EL1_BUF_BSC_MASK		PMBSR_EL1_MSS_MASK
 
 #define PMBSR_EL1_BUF_BSC_FULL		0x1UL
 
diff --git a/arch/arm64/tools/sysreg b/arch/arm64/tools/sysreg
index 384757a7eda9..04741f446c46 100644
--- a/arch/arm64/tools/sysreg
+++ b/arch/arm64/tools/sysreg
@@ -854,6 +854,130 @@ Sysreg	FAR_EL1	3	0	6	0	0
 Field	63:0	ADDR
 EndSysreg
 
+Sysreg	PMSCR_EL1	3	0	9	9	0
+Res0	63:8
+Field	7:6	PCT
+Field	5	TS
+Field	4	PA
+Field	3	CX
+Res0	2
+Field	1	E1SPE
+Field	0	E0SPE
+EndSysreg
+
+Sysreg	PMSNEVFR_EL1	3	0	9	9	1
+Field	63:0	E
+EndSysreg
+
+Sysreg	PMSICR_EL1	3	0	9	9	2
+Field	63:56	ECOUNT
+Res0	55:32
+Field	31:0	COUNT
+EndSysreg
+
+Sysreg	PMSIRR_EL1	3	0	9	9	3
+Res0	63:32
+Field	31:8	INTERVAL
+Res0	7:1
+Field	0	RND
+EndSysreg
+
+Sysreg	PMSFCR_EL1	3	0	9	9	4
+Res0	63:19
+Field	18	ST
+Field	17	LD
+Field	16	B
+Res0	15:4
+Field	3	FnE
+Field	2	FL
+Field	1	FT
+Field	0	FE
+EndSysreg
+
+Sysreg	PMSEVFR_EL1	3	0	9	9	5
+Field	63:0	E
+EndSysreg
+
+Sysreg	PMSLATFR_EL1	3	0	9	9	6
+Res0	63:16
+Field	15:0	MINLAT
+EndSysreg
+
+Sysreg	PMSIDR_EL1	3	0	9	9	7
+Res0	63:25
+Field	24	PBT
+Field	23:20	FORMAT
+Enum	19:16	COUNTSIZE
+	0b0010	12_BIT_SAT
+	0b0011	16_BIT_SAT
+EndEnum
+Field	15:12	MAXSIZE
+Enum	11:8	INTERVAL
+	0b0000	256
+	0b0010	512
+	0b0011	768
+	0b0100	1024
+	0b0101	1536
+	0b0110	2048
+	0b0111	3072
+	0b1000	4096
+EndEnum
+Res0	7
+Field	6	FnE
+Field	5	ERND
+Field	4	LDS
+Field	3	ARCHINST
+Field	2	FL
+Field	1	FT
+Field	0	FE
+EndSysreg
+
+Sysreg	PMBLIMITR_EL1	3	0	9	10	0
+Field	63:12	LIMIT
+Res0	11:6
+Field	5	PMFZ
+Res0	4:3
+Enum	2:1	FM
+	0b00	FILL
+	0b10	DISCARD
+EndEnum
+Field	0	E
+EndSysreg
+
+Sysreg	PMBPTR_EL1	3	0	9	10	1
+Field	63:0	PTR
+EndSysreg
+
+Sysreg	PMBSR_EL1	3	0	9	10	3
+Res0	63:32
+Enum	31:26	EC
+	0b000000	BUF
+	0b100100	FAULT_S1
+	0b100101	FAULT_S2
+	0b011110	FAULT_GPC
+	0b011111	IMP_DEF
+EndEnum
+Res0	25:20
+Field	19	DL
+Field	18	EA
+Field	17	S
+Field	16	COLL
+Field	15:0	MSS
+EndSysreg
+
+Sysreg	PMBIDR_EL1	3	0	9	10	7
+Res0	63:12
+Enum	11:8	EA
+	0b0000	NotDescribed
+	0b0001	Ignored
+	0b0010	SError
+EndEnum
+Res0	7:6
+Field	5	F
+Field	4	P
+Field	3:0	ALIGN
+EndSysreg
+
 SysregFields	CONTEXTIDR_ELx
 Res0	63:32
 Field	31:0	PROCID
@@ -1008,6 +1132,21 @@ Sysreg	FAR_EL2	3	4	6	0	0
 Field	63:0	ADDR
 EndSysreg
 
+Sysreg	PMSCR_EL2	3	4	9	9	0
+Res0	63:8
+Enum	7:6	PCT
+	0b00	VIRT
+	0b01	PHYS
+	0b11	GUEST
+EndEnum
+Field	5	TS
+Field	4	PA
+Field	3	CX
+Res0	2
+Field	1	E2SPE
+Field	0	E0HSPE
+EndSysreg
+
 Sysreg	CONTEXTIDR_EL2	3	4	13	0	1
 Fields	CONTEXTIDR_ELx
 EndSysreg

-- 
b4 0.11.0-dev

Re: [PATCH v3 3/8] arm64/sysreg: Convert SPE registers to automatic generation

[Mark Brown]

[-- Attachment #1: Type: text/plain, Size: 816 bytes --]

On Fri, Nov 04, 2022 at 10:55:03AM -0500, Rob Herring wrote:

> Convert all the SPE register defines to automatic generation. No
> functional changes.
> 
> New registers and fields for SPEv1.2 are added with the conversion.
> 
> Some of the PMBSR MSS field defines are kept as the automatic generation
> has no way to create multiple names for the same register bits. The
> meaning of the MSS field depends on other bits.

Reviewed-by: Mark Brown <broonie@kernel.org>

> +Sysreg	PMSNEVFR_EL1	3	0	9	9	1
> +Field	63:0	E
> +EndSysreg

JFTR as noted last time this looks nothing like the spec but is clearly
a sensible interpretation.

I do note that one advantage of doing this register by register rather
than en masse is that it makes it a lot easier to avoid re-reviewing the
same register...

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

[PATCH v3 4/8] perf: arm_spe: Drop BIT() and use FIELD_GET/PREP accessors

[Rob Herring]

Now that generated sysregs are in place, update the register field
accesses. The use of BIT() is no longer needed with the new defines. Use
FIELD_GET and FIELD_PREP instead of open coding masking and shifting.

No functional change.

Tested-by: James Clark <james.clark@arm.com>
Signed-off-by: Rob Herring <robh@kernel.org>
---
v3:
 - no change
---
 drivers/perf/arm_spe_pmu.c | 70 ++++++++++++++++++++++------------------------
 1 file changed, 34 insertions(+), 36 deletions(-)

diff --git a/drivers/perf/arm_spe_pmu.c b/drivers/perf/arm_spe_pmu.c
index 814ed18346b6..9b4bd72087ea 100644
--- a/drivers/perf/arm_spe_pmu.c
+++ b/drivers/perf/arm_spe_pmu.c
@@ -283,18 +283,18 @@ static u64 arm_spe_event_to_pmscr(struct perf_event *event)
 	struct perf_event_attr *attr = &event->attr;
 	u64 reg = 0;
 
-	reg |= ATTR_CFG_GET_FLD(attr, ts_enable) << PMSCR_EL1_TS_SHIFT;
-	reg |= ATTR_CFG_GET_FLD(attr, pa_enable) << PMSCR_EL1_PA_SHIFT;
-	reg |= ATTR_CFG_GET_FLD(attr, pct_enable) << PMSCR_EL1_PCT_SHIFT;
+	reg |= FIELD_PREP(PMSCR_EL1_TS, ATTR_CFG_GET_FLD(attr, ts_enable));
+	reg |= FIELD_PREP(PMSCR_EL1_PA, ATTR_CFG_GET_FLD(attr, pa_enable));
+	reg |= FIELD_PREP(PMSCR_EL1_PCT, ATTR_CFG_GET_FLD(attr, pct_enable));
 
 	if (!attr->exclude_user)
-		reg |= BIT(PMSCR_EL1_E0SPE_SHIFT);
+		reg |= PMSCR_EL1_E0SPE;
 
 	if (!attr->exclude_kernel)
-		reg |= BIT(PMSCR_EL1_E1SPE_SHIFT);
+		reg |= PMSCR_EL1_E1SPE;
 
 	if (get_spe_event_has_cx(event))
-		reg |= BIT(PMSCR_EL1_CX_SHIFT);
+		reg |= PMSCR_EL1_CX;
 
 	return reg;
 }
@@ -322,7 +322,7 @@ static u64 arm_spe_event_to_pmsirr(struct perf_event *event)
 
 	arm_spe_event_sanitise_period(event);
 
-	reg |= ATTR_CFG_GET_FLD(attr, jitter) << PMSIRR_EL1_RND_SHIFT;
+	reg |= FIELD_PREP(PMSIRR_EL1_RND, ATTR_CFG_GET_FLD(attr, jitter));
 	reg |= event->hw.sample_period;
 
 	return reg;
@@ -333,18 +333,18 @@ static u64 arm_spe_event_to_pmsfcr(struct perf_event *event)
 	struct perf_event_attr *attr = &event->attr;
 	u64 reg = 0;
 
-	reg |= ATTR_CFG_GET_FLD(attr, load_filter) << PMSFCR_EL1_LD_SHIFT;
-	reg |= ATTR_CFG_GET_FLD(attr, store_filter) << PMSFCR_EL1_ST_SHIFT;
-	reg |= ATTR_CFG_GET_FLD(attr, branch_filter) << PMSFCR_EL1_B_SHIFT;
+	reg |= FIELD_PREP(PMSFCR_EL1_LD, ATTR_CFG_GET_FLD(attr, load_filter));
+	reg |= FIELD_PREP(PMSFCR_EL1_ST, ATTR_CFG_GET_FLD(attr, store_filter));
+	reg |= FIELD_PREP(PMSFCR_EL1_B, ATTR_CFG_GET_FLD(attr, branch_filter));
 
 	if (reg)
-		reg |= BIT(PMSFCR_EL1_FT_SHIFT);
+		reg |= PMSFCR_EL1_FT;
 
 	if (ATTR_CFG_GET_FLD(attr, event_filter))
-		reg |= BIT(PMSFCR_EL1_FE_SHIFT);
+		reg |= PMSFCR_EL1_FE;
 
 	if (ATTR_CFG_GET_FLD(attr, min_latency))
-		reg |= BIT(PMSFCR_EL1_FL_SHIFT);
+		reg |= PMSFCR_EL1_FL;
 
 	return reg;
 }
@@ -358,8 +358,7 @@ static u64 arm_spe_event_to_pmsevfr(struct perf_event *event)
 static u64 arm_spe_event_to_pmslatfr(struct perf_event *event)
 {
 	struct perf_event_attr *attr = &event->attr;
-	return ATTR_CFG_GET_FLD(attr, min_latency)
-	       << PMSLATFR_EL1_MINLAT_SHIFT;
+	return FIELD_PREP(PMSLATFR_EL1_MINLAT, ATTR_CFG_GET_FLD(attr, min_latency));
 }
 
 static void arm_spe_pmu_pad_buf(struct perf_output_handle *handle, int len)
@@ -511,7 +510,7 @@ static void arm_spe_perf_aux_output_begin(struct perf_output_handle *handle,
 	limit = buf->snapshot ? arm_spe_pmu_next_snapshot_off(handle)
 			      : arm_spe_pmu_next_off(handle);
 	if (limit)
-		limit |= BIT(PMBLIMITR_EL1_E_SHIFT);
+		limit |= PMBLIMITR_EL1_E;
 
 	limit += (u64)buf->base;
 	base = (u64)buf->base + PERF_IDX2OFF(handle->head, buf);
@@ -570,23 +569,23 @@ arm_spe_pmu_buf_get_fault_act(struct perf_output_handle *handle)
 
 	/* Service required? */
 	pmbsr = read_sysreg_s(SYS_PMBSR_EL1);
-	if (!(pmbsr & BIT(PMBSR_EL1_S_SHIFT)))
+	if (!FIELD_GET(PMBSR_EL1_S, pmbsr))
 		return SPE_PMU_BUF_FAULT_ACT_SPURIOUS;
 
 	/*
 	 * If we've lost data, disable profiling and also set the PARTIAL
 	 * flag to indicate that the last record is corrupted.
 	 */
-	if (pmbsr & BIT(PMBSR_EL1_DL_SHIFT))
+	if (FIELD_GET(PMBSR_EL1_DL, pmbsr))
 		perf_aux_output_flag(handle, PERF_AUX_FLAG_TRUNCATED |
 					     PERF_AUX_FLAG_PARTIAL);
 
 	/* Report collisions to userspace so that it can up the period */
-	if (pmbsr & BIT(PMBSR_EL1_COLL_SHIFT))
+	if (FIELD_GET(PMBSR_EL1_COLL, pmbsr))
 		perf_aux_output_flag(handle, PERF_AUX_FLAG_COLLISION);
 
 	/* We only expect buffer management events */
-	switch (FIELD_GET(PMBSR_EL1_EC_MASK, pmbsr)) {
+	switch (FIELD_GET(PMBSR_EL1_EC, pmbsr)) {
 	case PMBSR_EL1_EC_BUF:
 		/* Handled below */
 		break;
@@ -716,23 +715,22 @@ static int arm_spe_pmu_event_init(struct perf_event *event)
 		return -EINVAL;
 
 	reg = arm_spe_event_to_pmsfcr(event);
-	if ((reg & BIT(PMSFCR_EL1_FE_SHIFT)) &&
+	if ((FIELD_GET(PMSFCR_EL1_FE, reg)) &&
 	    !(spe_pmu->features & SPE_PMU_FEAT_FILT_EVT))
 		return -EOPNOTSUPP;
 
-	if ((reg & BIT(PMSFCR_EL1_FT_SHIFT)) &&
+	if ((FIELD_GET(PMSFCR_EL1_FT, reg)) &&
 	    !(spe_pmu->features & SPE_PMU_FEAT_FILT_TYP))
 		return -EOPNOTSUPP;
 
-	if ((reg & BIT(PMSFCR_EL1_FL_SHIFT)) &&
+	if ((FIELD_GET(PMSFCR_EL1_FL, reg)) &&
 	    !(spe_pmu->features & SPE_PMU_FEAT_FILT_LAT))
 		return -EOPNOTSUPP;
 
 	set_spe_event_has_cx(event);
 	reg = arm_spe_event_to_pmscr(event);
 	if (!perfmon_capable() &&
-	    (reg & (BIT(PMSCR_EL1_PA_SHIFT) |
-		    BIT(PMSCR_EL1_PCT_SHIFT))))
+	    (reg & (PMSCR_EL1_PA | PMSCR_EL1_PCT)))
 		return -EACCES;
 
 	return 0;
@@ -970,14 +968,14 @@ static void __arm_spe_pmu_dev_probe(void *info)
 
 	/* Read PMBIDR first to determine whether or not we have access */
 	reg = read_sysreg_s(SYS_PMBIDR_EL1);
-	if (reg & BIT(PMBIDR_EL1_P_SHIFT)) {
+	if (FIELD_GET(PMBIDR_EL1_P, reg)) {
 		dev_err(dev,
 			"profiling buffer owned by higher exception level\n");
 		return;
 	}
 
 	/* Minimum alignment. If it's out-of-range, then fail the probe */
-	fld = (reg & PMBIDR_EL1_ALIGN_MASK) >> PMBIDR_EL1_ALIGN_SHIFT;
+	fld = FIELD_GET(PMBIDR_EL1_ALIGN, reg);
 	spe_pmu->align = 1 << fld;
 	if (spe_pmu->align > SZ_2K) {
 		dev_err(dev, "unsupported PMBIDR.Align [%d] on CPU %d\n",
@@ -987,26 +985,26 @@ static void __arm_spe_pmu_dev_probe(void *info)
 
 	/* It's now safe to read PMSIDR and figure out what we've got */
 	reg = read_sysreg_s(SYS_PMSIDR_EL1);
-	if (reg & BIT(PMSIDR_EL1_FE_SHIFT))
+	if (FIELD_GET(PMSIDR_EL1_FE, reg))
 		spe_pmu->features |= SPE_PMU_FEAT_FILT_EVT;
 
-	if (reg & BIT(PMSIDR_EL1_FT_SHIFT))
+	if (FIELD_GET(PMSIDR_EL1_FT, reg))
 		spe_pmu->features |= SPE_PMU_FEAT_FILT_TYP;
 
-	if (reg & BIT(PMSIDR_EL1_FL_SHIFT))
+	if (FIELD_GET(PMSIDR_EL1_FL, reg))
 		spe_pmu->features |= SPE_PMU_FEAT_FILT_LAT;
 
-	if (reg & BIT(PMSIDR_EL1_ARCHINST_SHIFT))
+	if (FIELD_GET(PMSIDR_EL1_ARCHINST, reg))
 		spe_pmu->features |= SPE_PMU_FEAT_ARCH_INST;
 
-	if (reg & BIT(PMSIDR_EL1_LDS_SHIFT))
+	if (FIELD_GET(PMSIDR_EL1_LDS, reg))
 		spe_pmu->features |= SPE_PMU_FEAT_LDS;
 
-	if (reg & BIT(PMSIDR_EL1_ERND_SHIFT))
+	if (FIELD_GET(PMSIDR_EL1_ERND, reg))
 		spe_pmu->features |= SPE_PMU_FEAT_ERND;
 
 	/* This field has a spaced out encoding, so just use a look-up */
-	fld = (reg & PMSIDR_EL1_INTERVAL_MASK) >> PMSIDR_EL1_INTERVAL_SHIFT;
+	fld = FIELD_GET(PMSIDR_EL1_INTERVAL, reg);
 	switch (fld) {
 	case 0:
 		spe_pmu->min_period = 256;
@@ -1038,7 +1036,7 @@ static void __arm_spe_pmu_dev_probe(void *info)
 	}
 
 	/* Maximum record size. If it's out-of-range, then fail the probe */
-	fld = (reg & PMSIDR_EL1_MAXSIZE_MASK) >> PMSIDR_EL1_MAXSIZE_SHIFT;
+	fld = FIELD_GET(PMSIDR_EL1_MAXSIZE, reg);
 	spe_pmu->max_record_sz = 1 << fld;
 	if (spe_pmu->max_record_sz > SZ_2K || spe_pmu->max_record_sz < 16) {
 		dev_err(dev, "unsupported PMSIDR_EL1.MaxSize [%d] on CPU %d\n",
@@ -1046,7 +1044,7 @@ static void __arm_spe_pmu_dev_probe(void *info)
 		return;
 	}
 
-	fld = (reg & PMSIDR_EL1_COUNTSIZE_MASK) >> PMSIDR_EL1_COUNTSIZE_SHIFT;
+	fld = FIELD_GET(PMSIDR_EL1_COUNTSIZE, reg);
 	switch (fld) {
 	default:
 		dev_warn(dev, "unknown PMSIDR_EL1.CountSize [%d]; assuming 2\n",

-- 
b4 0.11.0-dev

[PATCH v3 5/8] perf: arm_spe: Use new PMSIDR_EL1 register enums

[Rob Herring]

Now that the SPE register definitions include enums for some PMSIDR_EL1
fields, use them in the driver in place of magic values.

Signed-off-by: Rob Herring <robh@kernel.org>
---
v3: New patch
---
 drivers/perf/arm_spe_pmu.c | 20 ++++++++++----------
 1 file changed, 10 insertions(+), 10 deletions(-)

diff --git a/drivers/perf/arm_spe_pmu.c b/drivers/perf/arm_spe_pmu.c
index 9b4bd72087ea..af6d3867c3e7 100644
--- a/drivers/perf/arm_spe_pmu.c
+++ b/drivers/perf/arm_spe_pmu.c
@@ -1006,32 +1006,32 @@ static void __arm_spe_pmu_dev_probe(void *info)
 	/* This field has a spaced out encoding, so just use a look-up */
 	fld = FIELD_GET(PMSIDR_EL1_INTERVAL, reg);
 	switch (fld) {
-	case 0:
+	case PMSIDR_EL1_INTERVAL_256:
 		spe_pmu->min_period = 256;
 		break;
-	case 2:
+	case PMSIDR_EL1_INTERVAL_512:
 		spe_pmu->min_period = 512;
 		break;
-	case 3:
+	case PMSIDR_EL1_INTERVAL_768:
 		spe_pmu->min_period = 768;
 		break;
-	case 4:
+	case PMSIDR_EL1_INTERVAL_1024:
 		spe_pmu->min_period = 1024;
 		break;
-	case 5:
+	case PMSIDR_EL1_INTERVAL_1536:
 		spe_pmu->min_period = 1536;
 		break;
-	case 6:
+	case PMSIDR_EL1_INTERVAL_2048:
 		spe_pmu->min_period = 2048;
 		break;
-	case 7:
+	case PMSIDR_EL1_INTERVAL_3072:
 		spe_pmu->min_period = 3072;
 		break;
 	default:
 		dev_warn(dev, "unknown PMSIDR_EL1.Interval [%d]; assuming 8\n",
 			 fld);
 		fallthrough;
-	case 8:
+	case PMSIDR_EL1_INTERVAL_4096:
 		spe_pmu->min_period = 4096;
 	}
 
@@ -1050,10 +1050,10 @@ static void __arm_spe_pmu_dev_probe(void *info)
 		dev_warn(dev, "unknown PMSIDR_EL1.CountSize [%d]; assuming 2\n",
 			 fld);
 		fallthrough;
-	case 2:
+	case PMSIDR_EL1_COUNTSIZE_12_BIT_SAT:
 		spe_pmu->counter_sz = 12;
 		break;
-	case 3:
+	case PMSIDR_EL1_COUNTSIZE_16_BIT_SAT:
 		spe_pmu->counter_sz = 16;
 	}
 

-- 
b4 0.11.0-dev

[PATCH v3 6/8] perf: arm_spe: Support new SPEv1.2/v8.7 'not taken' event

[Rob Herring]

Arm SPEv1.2 (Armv8.7/v9.2) adds a new event, 'not taken', in bit 6 of
the PMSEVFR_EL1 register. Update arm_spe_pmsevfr_res0() to support the
additional event.

Tested-by: James Clark <james.clark@arm.com>
Signed-off-by: Rob Herring <robh@kernel.org>
---
v3:
 - No change
v2:
 - Update for v6.1 sysreg generated header changes
---
 arch/arm64/include/asm/sysreg.h | 2 ++
 drivers/perf/arm_spe_pmu.c      | 4 +++-
 2 files changed, 5 insertions(+), 1 deletion(-)

diff --git a/arch/arm64/include/asm/sysreg.h b/arch/arm64/include/asm/sysreg.h
index d002dd00e53e..06231e896832 100644
--- a/arch/arm64/include/asm/sysreg.h
+++ b/arch/arm64/include/asm/sysreg.h
@@ -242,6 +242,8 @@
 	 BIT_ULL(6) | BIT_ULL(4) | BIT_ULL(2) | BIT_ULL(0))
 #define PMSEVFR_EL1_RES0_V1P1	\
 	(PMSEVFR_EL1_RES0_IMP & ~(BIT_ULL(18) | BIT_ULL(17) | BIT_ULL(11)))
+#define PMSEVFR_EL1_RES0_V1P2	\
+	(PMSEVFR_EL1_RES0_V1P1 & ~BIT_ULL(6))
 
 /* Buffer error reporting */
 #define PMBSR_EL1_FAULT_FSC_SHIFT	PMBSR_EL1_MSS_SHIFT
diff --git a/drivers/perf/arm_spe_pmu.c b/drivers/perf/arm_spe_pmu.c
index af6d3867c3e7..82f67e941bc4 100644
--- a/drivers/perf/arm_spe_pmu.c
+++ b/drivers/perf/arm_spe_pmu.c
@@ -677,9 +677,11 @@ static u64 arm_spe_pmsevfr_res0(u16 pmsver)
 	case ID_AA64DFR0_EL1_PMSVer_IMP:
 		return PMSEVFR_EL1_RES0_IMP;
 	case ID_AA64DFR0_EL1_PMSVer_V1P1:
+		return PMSEVFR_EL1_RES0_V1P1;
+	case ID_AA64DFR0_EL1_PMSVer_V1P2:
 	/* Return the highest version we support in default */
 	default:
-		return PMSEVFR_EL1_RES0_V1P1;
+		return PMSEVFR_EL1_RES0_V1P2;
 	}
 }
 

-- 
b4 0.11.0-dev

[PATCH v3 7/8] perf: Add perf_event_attr::config3

[Rob Herring]

Arm SPEv1.2 adds another 64-bits of event filtering control. As the
existing perf_event_attr::configN fields are all used up for SPE PMU, an
additional field is needed. Add a new 'config3' field.

Tested-by: James Clark <james.clark@arm.com>
Signed-off-by: Rob Herring <robh@kernel.org>
---
v3:
 - No change
v2:
 - Drop tools/ side update
---
 include/uapi/linux/perf_event.h | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/include/uapi/linux/perf_event.h b/include/uapi/linux/perf_event.h
index 85be78e0e7f6..b2b1d7b54097 100644
--- a/include/uapi/linux/perf_event.h
+++ b/include/uapi/linux/perf_event.h
@@ -374,6 +374,7 @@ enum perf_event_read_format {
 #define PERF_ATTR_SIZE_VER5	112	/* add: aux_watermark */
 #define PERF_ATTR_SIZE_VER6	120	/* add: aux_sample_size */
 #define PERF_ATTR_SIZE_VER7	128	/* add: sig_data */
+#define PERF_ATTR_SIZE_VER8	136	/* add: config3 */
 
 /*
  * Hardware event_id to monitor via a performance monitoring event:
@@ -515,6 +516,8 @@ struct perf_event_attr {
 	 * truncated accordingly on 32 bit architectures.
 	 */
 	__u64	sig_data;
+
+	__u64	config3; /* extension of config2 */
 };
 
 /*

-- 
b4 0.11.0-dev

Re: [PATCH v3 7/8] perf: Add perf_event_attr::config3

[Will Deacon]

On Fri, Nov 04, 2022 at 10:55:07AM -0500, Rob Herring wrote:
> Arm SPEv1.2 adds another 64-bits of event filtering control. As the
> existing perf_event_attr::configN fields are all used up for SPE PMU, an
> additional field is needed. Add a new 'config3' field.
> 
> Tested-by: James Clark <james.clark@arm.com>
> Signed-off-by: Rob Herring <robh@kernel.org>
> ---
> v3:
>  - No change
> v2:
>  - Drop tools/ side update
> ---
>  include/uapi/linux/perf_event.h | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/include/uapi/linux/perf_event.h b/include/uapi/linux/perf_event.h
> index 85be78e0e7f6..b2b1d7b54097 100644
> --- a/include/uapi/linux/perf_event.h
> +++ b/include/uapi/linux/perf_event.h
> @@ -374,6 +374,7 @@ enum perf_event_read_format {
>  #define PERF_ATTR_SIZE_VER5	112	/* add: aux_watermark */
>  #define PERF_ATTR_SIZE_VER6	120	/* add: aux_sample_size */
>  #define PERF_ATTR_SIZE_VER7	128	/* add: sig_data */
> +#define PERF_ATTR_SIZE_VER8	136	/* add: config3 */
>  
>  /*
>   * Hardware event_id to monitor via a performance monitoring event:
> @@ -515,6 +516,8 @@ struct perf_event_attr {
>  	 * truncated accordingly on 32 bit architectures.
>  	 */
>  	__u64	sig_data;
> +
> +	__u64	config3; /* extension of config2 */

I need an ack from the perf core maintainers before I can take this.

Will

Re: [PATCH v3 7/8] perf: Add perf_event_attr::config3

[Rob Herring]

On Fri, Nov 18, 2022 at 10:49 AM Will Deacon <will@kernel.org> wrote:
>
> On Fri, Nov 04, 2022 at 10:55:07AM -0500, Rob Herring wrote:
> > Arm SPEv1.2 adds another 64-bits of event filtering control. As the
> > existing perf_event_attr::configN fields are all used up for SPE PMU, an
> > additional field is needed. Add a new 'config3' field.
> >
> > Tested-by: James Clark <james.clark@arm.com>
> > Signed-off-by: Rob Herring <robh@kernel.org>
> > ---
> > v3:
> >  - No change
> > v2:
> >  - Drop tools/ side update
> > ---
> >  include/uapi/linux/perf_event.h | 3 +++
> >  1 file changed, 3 insertions(+)
> >
> > diff --git a/include/uapi/linux/perf_event.h b/include/uapi/linux/perf_event.h
> > index 85be78e0e7f6..b2b1d7b54097 100644
> > --- a/include/uapi/linux/perf_event.h
> > +++ b/include/uapi/linux/perf_event.h
> > @@ -374,6 +374,7 @@ enum perf_event_read_format {
> >  #define PERF_ATTR_SIZE_VER5  112     /* add: aux_watermark */
> >  #define PERF_ATTR_SIZE_VER6  120     /* add: aux_sample_size */
> >  #define PERF_ATTR_SIZE_VER7  128     /* add: sig_data */
> > +#define PERF_ATTR_SIZE_VER8  136     /* add: config3 */
> >
> >  /*
> >   * Hardware event_id to monitor via a performance monitoring event:
> > @@ -515,6 +516,8 @@ struct perf_event_attr {
> >        * truncated accordingly on 32 bit architectures.
> >        */
> >       __u64   sig_data;
> > +
> > +     __u64   config3; /* extension of config2 */
>
> I need an ack from the perf core maintainers before I can take this.

Peter, Arnaldo, Ingo,

Can I get an ack on this please.

Rob

Re: [PATCH v3 7/8] perf: Add perf_event_attr::config3

[Alexander Shishkin]

Rob Herring <robh@kernel.org> writes:

> On Fri, Nov 18, 2022 at 10:49 AM Will Deacon <will@kernel.org> wrote:
>>
>> On Fri, Nov 04, 2022 at 10:55:07AM -0500, Rob Herring wrote:
>> > @@ -515,6 +516,8 @@ struct perf_event_attr {
>> >        * truncated accordingly on 32 bit architectures.
>> >        */
>> >       __u64   sig_data;
>> > +
>> > +     __u64   config3; /* extension of config2 */
>>
>> I need an ack from the perf core maintainers before I can take this.
>
> Peter, Arnaldo, Ingo,
>
> Can I get an ack on this please.

It appears that PMUs that don't use config{1,2} and now config3 allow
them to be whatever without any validation, whereas in reality we should
probably -EINVAL in those cases. Should something be done about that?

Regards,
--
Alex

Re: [PATCH v3 7/8] perf: Add perf_event_attr::config3

[Rob Herring]

On Mon, Nov 28, 2022 at 10:36 AM Alexander Shishkin
<alexander.shishkin@linux.intel.com> wrote:
>
> Rob Herring <robh@kernel.org> writes:
>
> > On Fri, Nov 18, 2022 at 10:49 AM Will Deacon <will@kernel.org> wrote:
> >>
> >> On Fri, Nov 04, 2022 at 10:55:07AM -0500, Rob Herring wrote:
> >> > @@ -515,6 +516,8 @@ struct perf_event_attr {
> >> >        * truncated accordingly on 32 bit architectures.
> >> >        */
> >> >       __u64   sig_data;
> >> > +
> >> > +     __u64   config3; /* extension of config2 */
> >>
> >> I need an ack from the perf core maintainers before I can take this.
> >
> > Peter, Arnaldo, Ingo,
> >
> > Can I get an ack on this please.
>
> It appears that PMUs that don't use config{1,2} and now config3 allow
> them to be whatever without any validation, whereas in reality we should
> probably -EINVAL in those cases. Should something be done about that?

Always the 3rd occurrence that gets to clean-up things. ;)

I think we'd have to add some capability flags for PMU drivers to set
to enable configN usage and then use those to validate configN is 0.
Wouldn't be too hard to do for config3 as we know there's exactly 1
user, but for 1,2 there's about 80 PMU drivers to check.

Rob

Re: [PATCH v3 7/8] perf: Add perf_event_attr::config3

[Mark Rutland]

Peter, it looks like this series is blocked on the below now; what would you
prefer out of:

(a) Take this as is, and look add adding additional validation on top.

(b) Add some flag to indicate a PMU driver supports config3, and have the core
    code check that, but leave the existing fields as-is for now (and hopefully
    follow up with further validation later for the existing fields).

(c) Go audit all the existing drivers, add flags to indicate support for
    existing fields, and have the core code check that. Atop that, add support
    for config3 with the same sort of flag check.

    I suspect that'd end up needing to go check more than config1/config2 given
    all the filter controls and so on that drivers aren't great at checking,
    and that might being fairly invasive.

(d) Something else?

I think we want to get to a point where drivers indicate what they actually
support and the core code rejects stuff drivers don't support or recognise, but
I think it'd be a little unreasonable to delay this series on cleaning up all
the existing issues.

I'm tempted to say (b) as that shouldn't introduce any regressions, should be a
relatively simple change to this series, and doesn't precluse making the rest
stricter as a follow-up. I'm happy to take a look at that (and IIUC Rob is
too).

What's your preference?

Thanks,
Mark.

On Mon, Nov 28, 2022 at 11:15:21AM -0600, Rob Herring wrote:
> On Mon, Nov 28, 2022 at 10:36 AM Alexander Shishkin
> <alexander.shishkin@linux.intel.com> wrote:
> >
> > Rob Herring <robh@kernel.org> writes:
> >
> > > On Fri, Nov 18, 2022 at 10:49 AM Will Deacon <will@kernel.org> wrote:
> > >>
> > >> On Fri, Nov 04, 2022 at 10:55:07AM -0500, Rob Herring wrote:
> > >> > @@ -515,6 +516,8 @@ struct perf_event_attr {
> > >> >        * truncated accordingly on 32 bit architectures.
> > >> >        */
> > >> >       __u64   sig_data;
> > >> > +
> > >> > +     __u64   config3; /* extension of config2 */
> > >>
> > >> I need an ack from the perf core maintainers before I can take this.
> > >
> > > Peter, Arnaldo, Ingo,
> > >
> > > Can I get an ack on this please.
> >
> > It appears that PMUs that don't use config{1,2} and now config3 allow
> > them to be whatever without any validation, whereas in reality we should
> > probably -EINVAL in those cases. Should something be done about that?
> 
> Always the 3rd occurrence that gets to clean-up things. ;)
> 
> I think we'd have to add some capability flags for PMU drivers to set
> to enable configN usage and then use those to validate configN is 0.
> Wouldn't be too hard to do for config3 as we know there's exactly 1
> user, but for 1,2 there's about 80 PMU drivers to check.
> 
> Rob
> 

Re: [PATCH v3 7/8] perf: Add perf_event_attr::config3

[Rob Herring]

On Tue, Dec 6, 2022 at 10:28 AM Mark Rutland <mark.rutland@arm.com> wrote:
>
> Peter, it looks like this series is blocked on the below now; what would you
> prefer out of:
>
> (a) Take this as is, and look add adding additional validation on top.
>
> (b) Add some flag to indicate a PMU driver supports config3, and have the core
>     code check that, but leave the existing fields as-is for now (and hopefully
>     follow up with further validation later for the existing fields).

That looks something like this:

diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index 853f64b6c8c2..845162b152ea 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -286,6 +286,7 @@ struct perf_event;
 #define PERF_PMU_CAP_NO_EXCLUDE                        0x0080
 #define PERF_PMU_CAP_AUX_OUTPUT                        0x0100
 #define PERF_PMU_CAP_EXTENDED_HW_TYPE          0x0200
+#define PERF_PMU_CAP_CONFIG3                   0x0400

 struct perf_output_handle;

diff --git a/kernel/events/core.c b/kernel/events/core.c
index aefc1e08e015..4414ae64432a 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -11314,6 +11314,9 @@ static int perf_try_init_event(struct pmu
*pmu, struct perf_event *event)
                    event_has_any_exclude_flag(event))
                        ret = -EINVAL;

+               if (!(pmu->capabilities & PERF_PMU_CAP_CONFIG3) &&
event->attr.config3)
+                       ret = -EINVAL;
+
                if (ret && event->destroy)
                        event->destroy(event);
        }

[PATCH v3 8/8] perf: arm_spe: Add support for SPEv1.2 inverted event filtering

[Rob Herring]

Arm SPEv1.2 (Arm v8.7/v9.2) adds a new feature called Inverted Event
Filter which excludes samples matching the event filter. The feature
mirrors the existing event filter in PMSEVFR_EL1 adding a new register,
PMSNEVFR_EL1, which has the same event bit assignments.

Tested-by: James Clark <james.clark@arm.com>
Signed-off-by: Rob Herring <robh@kernel.org>
---
v3:
 - No change
v2:
 - Update for auto generated register defines
 - Avoid accessing SYS_PMSNEVFR_EL1 on < v8.7
---
 drivers/perf/arm_spe_pmu.c | 45 +++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 45 insertions(+)

diff --git a/drivers/perf/arm_spe_pmu.c b/drivers/perf/arm_spe_pmu.c
index 82f67e941bc4..573db4211acd 100644
--- a/drivers/perf/arm_spe_pmu.c
+++ b/drivers/perf/arm_spe_pmu.c
@@ -85,6 +85,7 @@ struct arm_spe_pmu {
 #define SPE_PMU_FEAT_ARCH_INST			(1UL << 3)
 #define SPE_PMU_FEAT_LDS			(1UL << 4)
 #define SPE_PMU_FEAT_ERND			(1UL << 5)
+#define SPE_PMU_FEAT_INV_FILT_EVT		(1UL << 6)
 #define SPE_PMU_FEAT_DEV_PROBED			(1UL << 63)
 	u64					features;
 
@@ -202,6 +203,10 @@ static const struct attribute_group arm_spe_pmu_cap_group = {
 #define ATTR_CFG_FLD_min_latency_LO		0
 #define ATTR_CFG_FLD_min_latency_HI		11
 
+#define ATTR_CFG_FLD_inv_event_filter_CFG	config3	/* PMSNEVFR_EL1 */
+#define ATTR_CFG_FLD_inv_event_filter_LO	0
+#define ATTR_CFG_FLD_inv_event_filter_HI	63
+
 /* Why does everything I do descend into this? */
 #define __GEN_PMU_FORMAT_ATTR(cfg, lo, hi)				\
 	(lo) == (hi) ? #cfg ":" #lo "\n" : #cfg ":" #lo "-" #hi
@@ -232,6 +237,7 @@ GEN_PMU_FORMAT_ATTR(branch_filter);
 GEN_PMU_FORMAT_ATTR(load_filter);
 GEN_PMU_FORMAT_ATTR(store_filter);
 GEN_PMU_FORMAT_ATTR(event_filter);
+GEN_PMU_FORMAT_ATTR(inv_event_filter);
 GEN_PMU_FORMAT_ATTR(min_latency);
 
 static struct attribute *arm_spe_pmu_formats_attr[] = {
@@ -243,12 +249,27 @@ static struct attribute *arm_spe_pmu_formats_attr[] = {
 	&format_attr_load_filter.attr,
 	&format_attr_store_filter.attr,
 	&format_attr_event_filter.attr,
+	&format_attr_inv_event_filter.attr,
 	&format_attr_min_latency.attr,
 	NULL,
 };
 
+static umode_t arm_spe_pmu_format_attr_is_visible(struct kobject *kobj,
+						  struct attribute *attr,
+						  int unused)
+	{
+	struct device *dev = kobj_to_dev(kobj);
+	struct arm_spe_pmu *spe_pmu = dev_get_drvdata(dev);
+
+	if (attr == &format_attr_inv_event_filter.attr && !(spe_pmu->features & SPE_PMU_FEAT_INV_FILT_EVT))
+		return 0;
+
+	return attr->mode;
+}
+
 static const struct attribute_group arm_spe_pmu_format_group = {
 	.name	= "format",
+	.is_visible = arm_spe_pmu_format_attr_is_visible,
 	.attrs	= arm_spe_pmu_formats_attr,
 };
 
@@ -343,6 +364,9 @@ static u64 arm_spe_event_to_pmsfcr(struct perf_event *event)
 	if (ATTR_CFG_GET_FLD(attr, event_filter))
 		reg |= PMSFCR_EL1_FE;
 
+	if (ATTR_CFG_GET_FLD(attr, inv_event_filter))
+		reg |= PMSFCR_EL1_FnE;
+
 	if (ATTR_CFG_GET_FLD(attr, min_latency))
 		reg |= PMSFCR_EL1_FL;
 
@@ -355,6 +379,12 @@ static u64 arm_spe_event_to_pmsevfr(struct perf_event *event)
 	return ATTR_CFG_GET_FLD(attr, event_filter);
 }
 
+static u64 arm_spe_event_to_pmsnevfr(struct perf_event *event)
+{
+	struct perf_event_attr *attr = &event->attr;
+	return ATTR_CFG_GET_FLD(attr, inv_event_filter);
+}
+
 static u64 arm_spe_event_to_pmslatfr(struct perf_event *event)
 {
 	struct perf_event_attr *attr = &event->attr;
@@ -703,6 +733,9 @@ static int arm_spe_pmu_event_init(struct perf_event *event)
 	if (arm_spe_event_to_pmsevfr(event) & arm_spe_pmsevfr_res0(spe_pmu->pmsver))
 		return -EOPNOTSUPP;
 
+	if (arm_spe_event_to_pmsnevfr(event) & arm_spe_pmsevfr_res0(spe_pmu->pmsver))
+		return -EOPNOTSUPP;
+
 	if (attr->exclude_idle)
 		return -EOPNOTSUPP;
 
@@ -721,6 +754,10 @@ static int arm_spe_pmu_event_init(struct perf_event *event)
 	    !(spe_pmu->features & SPE_PMU_FEAT_FILT_EVT))
 		return -EOPNOTSUPP;
 
+	if ((FIELD_GET(PMSFCR_EL1_FnE, reg)) &&
+	    !(spe_pmu->features & SPE_PMU_FEAT_INV_FILT_EVT))
+		return -EOPNOTSUPP;
+
 	if ((FIELD_GET(PMSFCR_EL1_FT, reg)) &&
 	    !(spe_pmu->features & SPE_PMU_FEAT_FILT_TYP))
 		return -EOPNOTSUPP;
@@ -756,6 +793,11 @@ static void arm_spe_pmu_start(struct perf_event *event, int flags)
 	reg = arm_spe_event_to_pmsevfr(event);
 	write_sysreg_s(reg, SYS_PMSEVFR_EL1);
 
+	if (spe_pmu->features & SPE_PMU_FEAT_INV_FILT_EVT) {
+		reg = arm_spe_event_to_pmsnevfr(event);
+		write_sysreg_s(reg, SYS_PMSNEVFR_EL1);
+	}
+
 	reg = arm_spe_event_to_pmslatfr(event);
 	write_sysreg_s(reg, SYS_PMSLATFR_EL1);
 
@@ -990,6 +1032,9 @@ static void __arm_spe_pmu_dev_probe(void *info)
 	if (FIELD_GET(PMSIDR_EL1_FE, reg))
 		spe_pmu->features |= SPE_PMU_FEAT_FILT_EVT;
 
+	if (FIELD_GET(PMSIDR_EL1_FnE, reg))
+		spe_pmu->features |= SPE_PMU_FEAT_INV_FILT_EVT;
+
 	if (FIELD_GET(PMSIDR_EL1_FT, reg))
 		spe_pmu->features |= SPE_PMU_FEAT_FILT_TYP;
 

-- 
b4 0.11.0-dev

Re: [PATCH v3 8/8] perf: arm_spe: Add support for SPEv1.2 inverted event filtering

[Namhyung Kim]

Hello,

On Fri, Nov 4, 2022 at 8:55 AM Rob Herring <robh@kernel.org> wrote:
>
> Arm SPEv1.2 (Arm v8.7/v9.2) adds a new feature called Inverted Event
> Filter which excludes samples matching the event filter. The feature
> mirrors the existing event filter in PMSEVFR_EL1 adding a new register,
> PMSNEVFR_EL1, which has the same event bit assignments.

Just out of curiosity, is it possible to have the event filter and
inverted filter at the same time?  What if they have the same?

Thanks,
Namhyung

>
> Tested-by: James Clark <james.clark@arm.com>
> Signed-off-by: Rob Herring <robh@kernel.org>
> ---
> v3:
>  - No change
> v2:
>  - Update for auto generated register defines
>  - Avoid accessing SYS_PMSNEVFR_EL1 on < v8.7
> ---

Re: [PATCH v3 8/8] perf: arm_spe: Add support for SPEv1.2 inverted event filtering

[Rob Herring]

On Mon, Nov 7, 2022 at 1:48 PM Namhyung Kim <namhyung@kernel.org> wrote:
>
> Hello,
>
> On Fri, Nov 4, 2022 at 8:55 AM Rob Herring <robh@kernel.org> wrote:
> >
> > Arm SPEv1.2 (Arm v8.7/v9.2) adds a new feature called Inverted Event
> > Filter which excludes samples matching the event filter. The feature
> > mirrors the existing event filter in PMSEVFR_EL1 adding a new register,
> > PMSNEVFR_EL1, which has the same event bit assignments.
>
> Just out of curiosity, is it possible to have the event filter and
> inverted filter at the same time?

Yes, that's why config3 is needed. Samples with all the events set in
PMSEVFR_EL1 and none of the events set in PMSNEVFR_EL1 are recorded.
For example if an event bit is 0 in PMSEVFR_EL1, then recorded samples
may or may not have that event. If the same event bit is then set in
PMSNEVFR_EL1, then record samples will not have that event.

> What if they have the same?

You'd get no samples.

Rob

Re: [PATCH v3 0/8] perf: Arm SPEv1.2 support

[Rob Herring]

On Fri, Nov 4, 2022 at 10:55 AM Rob Herring <robh@kernel.org> wrote:
>
> This series adds support for Arm SPEv1.2 which is part of the
> Armv8.7/Armv9.2 architecture. There's 2 new features that affect the
> kernel: a new event filter bit, branch 'not taken', and an inverted
> event filter register.
>
> Since this support adds new registers and fields, first the SPE register
> defines are converted to automatic generation.
>
> Note that the 'config3' addition in sysfs format files causes SPE to
> break. A stable fix e552b7be12ed ("perf: Skip and warn on unknown format
> 'configN' attrs") landed in v6.1-rc1.
>
> The perf tool side changes are available here[1].
>
> Tested on FVP.
>
> [1] https://lore.kernel.org/all/20220914-arm-perf-tool-spe1-2-v2-v4-0-83c098e6212e@kernel.org/
>
> Signed-off-by: Rob Herring <robh@kernel.org>
> ---
> Changes in v3:
> - Add some more missing SPE register fields and use Enums for some
>   fields
> - Use the new PMSIDR_EL1 register Enum defines in the SPE driver
> - Link to v2: https://lore.kernel.org/r/20220825-arm-spe-v8-7-v2-0-e37322d68ac0@kernel.org
>
> Changes in v2:
> - Convert the SPE register defines to automatic generation
> - Fixed access to SYS_PMSNEVFR_EL1 when not present
> - Rebase on v6.1-rc1
> - Link to v1: https://lore.kernel.org/r/20220825-arm-spe-v8-7-v1-0-c75b8d92e692@kernel.org
>
> ---
> Rob Herring (8):
>       perf: arm_spe: Use feature numbering for PMSEVFR_EL1 defines
>       arm64: Drop SYS_ from SPE register defines
>       arm64/sysreg: Convert SPE registers to automatic generation
>       perf: arm_spe: Drop BIT() and use FIELD_GET/PREP accessors
>       perf: arm_spe: Use new PMSIDR_EL1 register enums
>       perf: arm_spe: Support new SPEv1.2/v8.7 'not taken' event
>       perf: Add perf_event_attr::config3
>       perf: arm_spe: Add support for SPEv1.2 inverted event filtering
>
>  arch/arm64/include/asm/el2_setup.h |   6 +-
>  arch/arm64/include/asm/sysreg.h    |  99 +++--------------------
>  arch/arm64/kvm/debug.c             |   2 +-
>  arch/arm64/kvm/hyp/nvhe/debug-sr.c |   2 +-
>  arch/arm64/tools/sysreg            | 139 +++++++++++++++++++++++++++++++++
>  drivers/perf/arm_spe_pmu.c         | 156 ++++++++++++++++++++++++-------------
>  include/uapi/linux/perf_event.h    |   3 +
>  7 files changed, 257 insertions(+), 150 deletions(-)

Will, any comments on this series?

Re: [PATCH v3 0/8] perf: Arm SPEv1.2 support

[Will Deacon]

On Thu, Nov 17, 2022 at 08:43:17AM -0600, Rob Herring wrote:
> On Fri, Nov 4, 2022 at 10:55 AM Rob Herring <robh@kernel.org> wrote:
> >
> > This series adds support for Arm SPEv1.2 which is part of the
> > Armv8.7/Armv9.2 architecture. There's 2 new features that affect the
> > kernel: a new event filter bit, branch 'not taken', and an inverted
> > event filter register.
> >
> > Since this support adds new registers and fields, first the SPE register
> > defines are converted to automatic generation.
> >
> > Note that the 'config3' addition in sysfs format files causes SPE to
> > break. A stable fix e552b7be12ed ("perf: Skip and warn on unknown format
> > 'configN' attrs") landed in v6.1-rc1.
> >
> > The perf tool side changes are available here[1].
> >
> > Tested on FVP.
> >
> > [1] https://lore.kernel.org/all/20220914-arm-perf-tool-spe1-2-v2-v4-0-83c098e6212e@kernel.org/
> >
> > Signed-off-by: Rob Herring <robh@kernel.org>
> > ---
> > Changes in v3:
> > - Add some more missing SPE register fields and use Enums for some
> >   fields
> > - Use the new PMSIDR_EL1 register Enum defines in the SPE driver
> > - Link to v2: https://lore.kernel.org/r/20220825-arm-spe-v8-7-v2-0-e37322d68ac0@kernel.org
> >
> > Changes in v2:
> > - Convert the SPE register defines to automatic generation
> > - Fixed access to SYS_PMSNEVFR_EL1 when not present
> > - Rebase on v6.1-rc1
> > - Link to v1: https://lore.kernel.org/r/20220825-arm-spe-v8-7-v1-0-c75b8d92e692@kernel.org
> >
> > ---
> > Rob Herring (8):
> >       perf: arm_spe: Use feature numbering for PMSEVFR_EL1 defines
> >       arm64: Drop SYS_ from SPE register defines
> >       arm64/sysreg: Convert SPE registers to automatic generation
> >       perf: arm_spe: Drop BIT() and use FIELD_GET/PREP accessors
> >       perf: arm_spe: Use new PMSIDR_EL1 register enums
> >       perf: arm_spe: Support new SPEv1.2/v8.7 'not taken' event
> >       perf: Add perf_event_attr::config3
> >       perf: arm_spe: Add support for SPEv1.2 inverted event filtering
> >
> >  arch/arm64/include/asm/el2_setup.h |   6 +-
> >  arch/arm64/include/asm/sysreg.h    |  99 +++--------------------
> >  arch/arm64/kvm/debug.c             |   2 +-
> >  arch/arm64/kvm/hyp/nvhe/debug-sr.c |   2 +-
> >  arch/arm64/tools/sysreg            | 139 +++++++++++++++++++++++++++++++++
> >  drivers/perf/arm_spe_pmu.c         | 156 ++++++++++++++++++++++++-------------
> >  include/uapi/linux/perf_event.h    |   3 +
> >  7 files changed, 257 insertions(+), 150 deletions(-)
> 
> Will, any comments on this series?

Looks fine to me. Happy to queue it once the uapi change has been acked.

Will