[PATCH v7 00/13] perf/x86/amd: Add AMD Fam19h Branch Sampling support

[PATCH v7 00/13] perf/x86/amd: Add AMD Fam19h Branch Sampling support

[Stephane Eranian]

Add support for the AMD Fam19h 16-deep branch sampling feature as described
in the AMD PPR Fam19h Model 01h Revision B1 section 2.1.13. This is a model
specific extension. It is not an architected AMD feature.

The Branch Sampling Feature (BRS) provides the statistical taken branch
information necessary to enable autoFDO-style optimization by compilers,
i.e., basic block execution counts.

BRS operates with a 16-deep saturating buffer in MSR registers. There is no
hardware branch type filtering. All control flow changes are captured. BRS
relies on specific programming of the core PMU of Fam19h.  In particular,
the following requirements must be met:
 - the sampling period be greater than 16 (BRS depth)
 - the sampling period must use fixed and not frequency mode

BRS interacts with the NMI interrupt as well. Because enabling BRS is
expensive, it is only activated after P event occurrences, where P is the
desired sampling period. At P occurrences of the event, the counter
overflows, the CPU catches the NMI interrupt, activates BRS for 16 branches
until it saturates, and then delivers the NMI to the kernel. Between the
overflow and the time BRS activates more branches may be executed skewing the
period. All along, the sampling event keeps counting. The skid may be
attenuated by reducing the sampling period by 16.

BRS is integrated into perf_events seamlessly via the same
PERF_RECORD_BRANCH_STACK sample format. BRS generates branch
perf_branch_entry records in the sampling buffer. There is no prediction or
latency information supported. The branches are stored in reverse order of
execution.  The most recent branch is the first entry in each record.

Because BRS must be stopped when a CPU goes into low power mode, the series
includes patches to add callbacks on ACPI low power entry and exit which is
used on AMD processors.

Given that there is no privilege filterting with BRS, the kernel implements
filtering on privlege level.

This version adds a few simple modifications to perf record and report.
1. add the branch-brs event as a builtin such as it can used directly:
   perf record -e branch-brs ...
2. improve error handling for AMD IBS and is contributed by Kim Phillips.
3. use the better error handling to improve error handling for BRS.
4. add two new sort dimensions to help display the branch sampling
   information. Because there is no latency information associated with the
   branch sampling feature perf report would collapse all samples within a
   function into a single histogram entry. This is expected because the
   default sort mode for PERF_SAMPLE_BRANCH_STACK is symbol_from/symbol_to.
   This propagates to the annotation.

For more detailed view of the branch samples, the new sort dimensions
addr_from,addr_to can be used instead as follows:

$ perf report --sort=overhead,comm,dso,addr_from,addr_to 
# Overhead  Command    Shared Object     Source Address   Target Address
# ........  .......... ..............  ..............     ..............
#
     4.21%  test_prg   test_prg       [.] test_threa+0x3c [.] test_threa+0x4
     4.14%  test_prg   test_prg       [.] test_threa+0x3e [.] test_threa+0x2
     4.10%  test_prg   test_prg       [.] test_threa+0x4  [.] test_threa+0x3a
     4.07%  test_prg   test_prg       [.] test_threa+0x2  [.] test_threa+0x3c

Versus the default output:

$ perf report 
# Overhead  Command   Source Shared Object Source Symbol    Target Symbol      Basic Block Cycles
# ........  ......... .................... ................ .................  ..................
#
    99.52%  test_prg  test_prg             [.] test_thread  [.] test_thread    -                 

BRS can be used with any sampling event. However, it is recommended to use
the RETIRED_BRANCH event because it matches what the BRS captures. For
convenience, a pseudo event matching the branches captured by BRS is
exported by the kernel (branch-brs):

$ perf record -b -e cpu/branch-brs/ -c 1000037 test

$ perf report -D
56531696056126 0x193c000 [0x1a8]: PERF_RECORD_SAMPLE(IP, 0x2): 18122/18230: 0x401d24 period: 1000037 addr: 0
... branch stack: nr:16
.....  0: 0000000000401d24 -> 0000000000401d5a 0 cycles      0
.....  1: 0000000000401d5c -> 0000000000401d24 0 cycles      0
.....  2: 0000000000401d22 -> 0000000000401d5c 0 cycles      0
.....  3: 0000000000401d5e -> 0000000000401d22 0 cycles      0
.....  4: 0000000000401d20 -> 0000000000401d5e 0 cycles      0
.....  5: 0000000000401d3e -> 0000000000401d20 0 cycles      0
.....  6: 0000000000401d42 -> 0000000000401d3e 0 cycles      0
.....  7: 0000000000401d3c -> 0000000000401d42 0 cycles      0
.....  8: 0000000000401d44 -> 0000000000401d3c 0 cycles      0
.....  9: 0000000000401d3a -> 0000000000401d44 0 cycles      0
..... 10: 0000000000401d46 -> 0000000000401d3a 0 cycles      0
..... 11: 0000000000401d38 -> 0000000000401d46 0 cycles      0
..... 12: 0000000000401d48 -> 0000000000401d38 0 cycles      0
..... 13: 0000000000401d36 -> 0000000000401d48 0 cycles      0
..... 14: 0000000000401d4a -> 0000000000401d36 0 cycles      0
..... 15: 0000000000401d34 -> 0000000000401d4a 0 cycles      0
 ... thread: test:18230
 ...... dso: test

Special thanks to Kim Phillips @ AMD for the testing, reviews and
contributions.

V2 makes the following changes:
  - the low power callback code has be reworked completly. It is not
    impacting the generic perf_events code anymore. This is all handled
    via x86 code and only for ACPI low power driver which seems to be the
    default on AMD. The change in acpi_pad.c and processor_idle.c has no
    impact on non x86 architectures, on Intel x86 or AMD without BRS, a
    jump label is used to void the code unless necessary
  - BRS is an opt-in compile time option for the kernel
  - branch_stack bit clearing helper is introduced
  - As for the fact that BRS holds the NMI and that it may conflict with
    other sampling events and introduced skid, this is not really a problem
    because AMD PMI skid is already very large prompting special handling in
    amd_pmu_wait_on_overflow(), so adding a few cycles while the CPU executes
    at most 16 taken branches is not a problem.


V3 makes the following changes:
   - simplifies the handling of BRS enable/disable to mimic the Intel LBR code
     path more closely. That removes some callbacks in generic x86 code
   - add config option to compile BRS as an opt-in (off by default)
   - updated perf tool error reporting patch updates by Kim Phillips

V4 makes the following changes:
   - rebase to latest tip.git (commit 6f5ac142e5df)
   - integrate Kim Phillips latest perf tool error handling patches


V5 makes the following changes:
   - rebased to 5.17-rc1
   - integrated feedback from PeterZ about AMD perf_events callbacks
   - fix cpufeature macros name X86_FEATURE_BRS
   - integrated all perf tool error handling from kim Phillips

V6 makes the following changes:
   - rebase to 5.17-rc3
   - fix the typo in the Kconfig for BRS opt-in
   - reword all changelogs to be more aligned with standard

V7 makes the following changes:
   - rebased to 5.17-rc8
   - Restrict BRS to be used only with RETIRED_TAKEN_BRANCH event in order to
     avoid adding more skid to other events in case that are monitored
     concurrently
   - fixed bugs in the perf tool enhanced error handling for IBS and BRS

Kim Phillips (1):
  perf tools: Improve IBS error handling

Stephane Eranian (12):
  perf/core: add perf_clear_branch_entry_bitfields() helper
  x86/cpufeatures: add AMD Fam19h Branch Sampling feature
  perf/x86/amd: add AMD Fam19h Branch Sampling support
  perf/x86/amd: add branch-brs helper event for Fam19h BRS
  perf/x86/amd: enable branch sampling priv level filtering
  perf/x86/amd: add AMD branch sampling period adjustment
  perf/x86/amd: make Zen3 branch sampling opt-in
  ACPI: add perf low power callback
  perf/x86/amd: add idle hooks for branch sampling
  perf tools: fix NULL point in evsel__env()
  perf tools: Improve error handling of AMD Branch Sampling
  perf report: add addr_from/addr_to sort dimensions

 arch/x86/events/Kconfig            |   8 +
 arch/x86/events/amd/Makefile       |   1 +
 arch/x86/events/amd/brs.c          | 363 +++++++++++++++++++++++++++++
 arch/x86/events/amd/core.c         | 252 +++++++++++++++++++-
 arch/x86/events/core.c             |  17 +-
 arch/x86/events/intel/lbr.c        |  36 ++-
 arch/x86/events/perf_event.h       | 143 ++++++++++--
 arch/x86/include/asm/cpufeatures.h |   1 +
 arch/x86/include/asm/msr-index.h   |   4 +
 arch/x86/include/asm/perf_event.h  |  21 ++
 drivers/acpi/acpi_pad.c            |   6 +
 drivers/acpi/processor_idle.c      |   5 +
 include/linux/perf_event.h         |  22 ++
 tools/perf/util/evsel.c            |  48 +++-
 tools/perf/util/hist.c             |   2 +
 tools/perf/util/hist.h             |   2 +
 tools/perf/util/sort.c             | 128 ++++++++++
 tools/perf/util/sort.h             |   2 +
 18 files changed, 1020 insertions(+), 41 deletions(-)
 create mode 100644 arch/x86/events/amd/brs.c

-- 
2.35.1.894.gb6a874cedc-goog

[PATCH v7 01/13] perf/core: add perf_clear_branch_entry_bitfields() helper

[Stephane Eranian]

Make it simpler to reset all the info fields on the
perf_branch_entry by adding a helper inline function.

The goal is to centralize the initialization to avoid missing
a field in case more are added.

Signed-off-by: Stephane Eranian <eranian@google.com>
---
 arch/x86/events/intel/lbr.c | 36 +++++++++++++++++-------------------
 include/linux/perf_event.h  | 16 ++++++++++++++++
 2 files changed, 33 insertions(+), 19 deletions(-)

diff --git a/arch/x86/events/intel/lbr.c b/arch/x86/events/intel/lbr.c
index fe1742c4ca49..13179f31fe10 100644
--- a/arch/x86/events/intel/lbr.c
+++ b/arch/x86/events/intel/lbr.c
@@ -769,6 +769,7 @@ void intel_pmu_lbr_disable_all(void)
 void intel_pmu_lbr_read_32(struct cpu_hw_events *cpuc)
 {
 	unsigned long mask = x86_pmu.lbr_nr - 1;
+	struct perf_branch_entry *br = cpuc->lbr_entries;
 	u64 tos = intel_pmu_lbr_tos();
 	int i;
 
@@ -784,15 +785,11 @@ void intel_pmu_lbr_read_32(struct cpu_hw_events *cpuc)
 
 		rdmsrl(x86_pmu.lbr_from + lbr_idx, msr_lastbranch.lbr);
 
-		cpuc->lbr_entries[i].from	= msr_lastbranch.from;
-		cpuc->lbr_entries[i].to		= msr_lastbranch.to;
-		cpuc->lbr_entries[i].mispred	= 0;
-		cpuc->lbr_entries[i].predicted	= 0;
-		cpuc->lbr_entries[i].in_tx	= 0;
-		cpuc->lbr_entries[i].abort	= 0;
-		cpuc->lbr_entries[i].cycles	= 0;
-		cpuc->lbr_entries[i].type	= 0;
-		cpuc->lbr_entries[i].reserved	= 0;
+		perf_clear_branch_entry_bitfields(br);
+
+		br->from	= msr_lastbranch.from;
+		br->to		= msr_lastbranch.to;
+		br++;
 	}
 	cpuc->lbr_stack.nr = i;
 	cpuc->lbr_stack.hw_idx = tos;
@@ -807,6 +804,7 @@ void intel_pmu_lbr_read_64(struct cpu_hw_events *cpuc)
 {
 	bool need_info = false, call_stack = false;
 	unsigned long mask = x86_pmu.lbr_nr - 1;
+	struct perf_branch_entry *br = cpuc->lbr_entries;
 	u64 tos = intel_pmu_lbr_tos();
 	int i;
 	int out = 0;
@@ -878,15 +876,14 @@ void intel_pmu_lbr_read_64(struct cpu_hw_events *cpuc)
 		if (abort && x86_pmu.lbr_double_abort && out > 0)
 			out--;
 
-		cpuc->lbr_entries[out].from	 = from;
-		cpuc->lbr_entries[out].to	 = to;
-		cpuc->lbr_entries[out].mispred	 = mis;
-		cpuc->lbr_entries[out].predicted = pred;
-		cpuc->lbr_entries[out].in_tx	 = in_tx;
-		cpuc->lbr_entries[out].abort	 = abort;
-		cpuc->lbr_entries[out].cycles	 = cycles;
-		cpuc->lbr_entries[out].type	 = 0;
-		cpuc->lbr_entries[out].reserved	 = 0;
+		perf_clear_branch_entry_bitfields(br+out);
+		br[out].from	 = from;
+		br[out].to	 = to;
+		br[out].mispred	 = mis;
+		br[out].predicted = pred;
+		br[out].in_tx	 = in_tx;
+		br[out].abort	 = abort;
+		br[out].cycles	 = cycles;
 		out++;
 	}
 	cpuc->lbr_stack.nr = out;
@@ -951,6 +948,8 @@ static void intel_pmu_store_lbr(struct cpu_hw_events *cpuc,
 		to = rdlbr_to(i, lbr);
 		info = rdlbr_info(i, lbr);
 
+		perf_clear_branch_entry_bitfields(e);
+
 		e->from		= from;
 		e->to		= to;
 		e->mispred	= get_lbr_mispred(info);
@@ -959,7 +958,6 @@ static void intel_pmu_store_lbr(struct cpu_hw_events *cpuc,
 		e->abort	= !!(info & LBR_INFO_ABORT);
 		e->cycles	= get_lbr_cycles(info);
 		e->type		= get_lbr_br_type(info);
-		e->reserved	= 0;
 	}
 
 	cpuc->lbr_stack.nr = i;
diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index 733649184b27..496eb6aa6e54 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -1063,6 +1063,22 @@ static inline void perf_sample_data_init(struct perf_sample_data *data,
 	data->txn = 0;
 }
 
+/*
+ * Clear all bitfields in the perf_branch_entry.
+ * The to and from fields are not cleared because they are
+ * systematically modified by caller.
+ */
+static inline void perf_clear_branch_entry_bitfields(struct perf_branch_entry *br)
+{
+	br->mispred = 0;
+	br->predicted = 0;
+	br->in_tx = 0;
+	br->abort = 0;
+	br->cycles = 0;
+	br->type = 0;
+	br->reserved = 0;
+}
+
 extern void perf_output_sample(struct perf_output_handle *handle,
 			       struct perf_event_header *header,
 			       struct perf_sample_data *data,
-- 
2.35.1.894.gb6a874cedc-goog

[PATCH v7 02/13] x86/cpufeatures: add AMD Fam19h Branch Sampling feature

[Stephane Eranian]

Add a cpu feature for AMD Fam19h Branch Sampling feature as bit
31 of EBX on CPUID leaf function 0x80000008.

Signed-off-by: Stephane Eranian <eranian@google.com>
---
 arch/x86/include/asm/cpufeatures.h | 1 +
 1 file changed, 1 insertion(+)

diff --git a/arch/x86/include/asm/cpufeatures.h b/arch/x86/include/asm/cpufeatures.h
index 9a59e2e73bee..ff3ee4f4ebbb 100644
--- a/arch/x86/include/asm/cpufeatures.h
+++ b/arch/x86/include/asm/cpufeatures.h
@@ -315,6 +315,7 @@
 #define X86_FEATURE_VIRT_SSBD		(13*32+25) /* Virtualized Speculative Store Bypass Disable */
 #define X86_FEATURE_AMD_SSB_NO		(13*32+26) /* "" Speculative Store Bypass is fixed in hardware. */
 #define X86_FEATURE_CPPC		(13*32+27) /* Collaborative Processor Performance Control */
+#define X86_FEATURE_BRS			(13*32+31) /* Branch Sampling available */
 
 /* Thermal and Power Management Leaf, CPUID level 0x00000006 (EAX), word 14 */
 #define X86_FEATURE_DTHERM		(14*32+ 0) /* Digital Thermal Sensor */
-- 
2.35.1.894.gb6a874cedc-goog

[PATCH v7 03/13] perf/x86/amd: add AMD Fam19h Branch Sampling support

[Stephane Eranian]

Add support for the AMD Fam19h 16-deep branch sampling feature as
described in the AMD PPR Fam19h Model 01h Revision B1.  This is a model
specific extension. It is not an architected AMD feature.

The Branch Sampling (BRS) operates with a 16-deep saturating buffer in MSR
registers. There is no branch type filtering. All control flow changes are
captured. BRS relies on specific programming of the core PMU of Fam19h.  In
particular, the following requirements must be met:
 - the sampling period be greater than 16 (BRS depth)
 - the sampling period must use a fixed and not frequency mode

BRS interacts with the NMI interrupt as well. Because enabling BRS is
expensive, it is only activated after P event occurrences, where P is the
desired sampling period.  At P occurrences of the event, the counter
overflows, the CPU catches the interrupt, activates BRS for 16 branches until
it saturates, and then delivers the NMI to the kernel.  Between the overflow
and the time BRS activates more branches may be executed skewing the period.
All along, the sampling event keeps counting. The skid may be attenuated by
reducing the sampling period by 16 (subsequent patch).

BRS is integrated into perf_events seamlessly via the same
PERF_RECORD_BRANCH_STACK sample format. BRS generates perf_branch_entry
records in the sampling buffer. No prediction information is supported. The
branches are stored in reverse order of execution.  The most recent branch is
the first entry in each record.

No modification to the perf tool is necessary.

BRS can be used with any sampling event. However, it is recommended to use
the RETIRED_BRANCH_INSTRUCTIONS event because it matches what the BRS
captures.

$ perf record -b -c 1000037 -e cpu/event=0xc2,name=ret_br_instructions/ test

$ perf report -D
56531696056126 0x193c000 [0x1a8]: PERF_RECORD_SAMPLE(IP, 0x2): 18122/18230: 0x401d24 period: 1000037 addr: 0
... branch stack: nr:16
.....  0: 0000000000401d24 -> 0000000000401d5a 0 cycles      0
.....  1: 0000000000401d5c -> 0000000000401d24 0 cycles      0
.....  2: 0000000000401d22 -> 0000000000401d5c 0 cycles      0
.....  3: 0000000000401d5e -> 0000000000401d22 0 cycles      0
.....  4: 0000000000401d20 -> 0000000000401d5e 0 cycles      0
.....  5: 0000000000401d3e -> 0000000000401d20 0 cycles      0
.....  6: 0000000000401d42 -> 0000000000401d3e 0 cycles      0
.....  7: 0000000000401d3c -> 0000000000401d42 0 cycles      0
.....  8: 0000000000401d44 -> 0000000000401d3c 0 cycles      0
.....  9: 0000000000401d3a -> 0000000000401d44 0 cycles      0
..... 10: 0000000000401d46 -> 0000000000401d3a 0 cycles      0
..... 11: 0000000000401d38 -> 0000000000401d46 0 cycles      0
..... 12: 0000000000401d48 -> 0000000000401d38 0 cycles      0
..... 13: 0000000000401d36 -> 0000000000401d48 0 cycles      0
..... 14: 0000000000401d4a -> 0000000000401d36 0 cycles      0
..... 15: 0000000000401d34 -> 0000000000401d4a 0 cycles      0
 ... thread: test:18230
 ...... dso: test

Signed-off-by: Stephane Eranian <eranian@google.com>
---
 arch/x86/events/amd/Makefile     |   2 +-
 arch/x86/events/amd/brs.c        | 317 +++++++++++++++++++++++++++++++
 arch/x86/events/amd/core.c       | 233 ++++++++++++++++++++++-
 arch/x86/events/core.c           |  10 +-
 arch/x86/events/perf_event.h     | 101 ++++++++--
 arch/x86/include/asm/msr-index.h |   4 +
 6 files changed, 645 insertions(+), 22 deletions(-)
 create mode 100644 arch/x86/events/amd/brs.c

diff --git a/arch/x86/events/amd/Makefile b/arch/x86/events/amd/Makefile
index 6cbe38d5fd9d..cf323ffab5cd 100644
--- a/arch/x86/events/amd/Makefile
+++ b/arch/x86/events/amd/Makefile
@@ -1,5 +1,5 @@
 # SPDX-License-Identifier: GPL-2.0
-obj-$(CONFIG_CPU_SUP_AMD)		+= core.o
+obj-$(CONFIG_CPU_SUP_AMD)		+= core.o brs.o
 obj-$(CONFIG_PERF_EVENTS_AMD_POWER)	+= power.o
 obj-$(CONFIG_X86_LOCAL_APIC)		+= ibs.o
 obj-$(CONFIG_PERF_EVENTS_AMD_UNCORE)	+= amd-uncore.o
diff --git a/arch/x86/events/amd/brs.c b/arch/x86/events/amd/brs.c
new file mode 100644
index 000000000000..3c13c484c637
--- /dev/null
+++ b/arch/x86/events/amd/brs.c
@@ -0,0 +1,317 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Implement support for AMD Fam19h Branch Sampling feature
+ * Based on specifications published in AMD PPR Fam19 Model 01
+ *
+ * Copyright 2021 Google LLC
+ * Contributed by Stephane Eranian <eranian@google.com>
+ */
+#include <linux/kernel.h>
+#include <asm/msr.h>
+#include <asm/cpufeature.h>
+
+#include "../perf_event.h"
+
+#define BRS_POISON	0xFFFFFFFFFFFFFFFEULL /* mark limit of valid entries */
+
+/* Debug Extension Configuration register layout */
+union amd_debug_extn_cfg {
+	__u64 val;
+	struct {
+		__u64	rsvd0:2,  /* reserved */
+			brsmen:1, /* branch sample enable */
+			rsvd4_3:2,/* reserved - must be 0x3 */
+			vb:1,     /* valid branches recorded */
+			rsvd2:10, /* reserved */
+			msroff:4, /* index of next entry to write */
+			rsvd3:4,  /* reserved */
+			pmc:3,    /* #PMC holding the sampling event */
+			rsvd4:37; /* reserved */
+	};
+};
+
+static inline unsigned int brs_from(int idx)
+{
+	return MSR_AMD_SAMP_BR_FROM + 2 * idx;
+}
+
+static inline unsigned int brs_to(int idx)
+{
+	return MSR_AMD_SAMP_BR_FROM + 2 * idx + 1;
+}
+
+static inline void set_debug_extn_cfg(u64 val)
+{
+	/* bits[4:3] must always be set to 11b */
+	wrmsrl(MSR_AMD_DBG_EXTN_CFG, val | 3ULL << 3);
+}
+
+static inline u64 get_debug_extn_cfg(void)
+{
+	u64 val;
+
+	rdmsrl(MSR_AMD_DBG_EXTN_CFG, val);
+	return val;
+}
+
+static bool __init amd_brs_detect(void)
+{
+	if (!boot_cpu_has(X86_FEATURE_BRS))
+		return false;
+
+	switch (boot_cpu_data.x86) {
+	case 0x19: /* AMD Fam19h (Zen3) */
+		x86_pmu.lbr_nr = 16;
+
+		/* No hardware filtering supported */
+		x86_pmu.lbr_sel_map = NULL;
+		x86_pmu.lbr_sel_mask = 0;
+		break;
+	default:
+		return false;
+	}
+
+	return true;
+}
+
+/*
+ * Current BRS implementation does not support branch type or privilege level
+ * filtering. Therefore, this function simply enforces these limitations. No need for
+ * a br_sel_map. Software filtering is not supported because it would not correlate well
+ * with a sampling period.
+ */
+int amd_brs_setup_filter(struct perf_event *event)
+{
+	u64 type = event->attr.branch_sample_type;
+
+	/* No BRS support */
+	if (!x86_pmu.lbr_nr)
+		return -EOPNOTSUPP;
+
+	/* Can only capture all branches, i.e., no filtering */
+	if ((type & ~PERF_SAMPLE_BRANCH_PLM_ALL) != PERF_SAMPLE_BRANCH_ANY)
+		return -EINVAL;
+
+	/* can only capture at all priv levels due to the way BRS works */
+	if ((type & PERF_SAMPLE_BRANCH_PLM_ALL) != PERF_SAMPLE_BRANCH_PLM_ALL)
+		return -EINVAL;
+
+	return 0;
+}
+
+/* tos = top of stack, i.e., last valid entry written */
+static inline int amd_brs_get_tos(union amd_debug_extn_cfg *cfg)
+{
+	/*
+	 * msroff: index of next entry to write so top-of-stack is one off
+	 * if BRS is full then msroff is set back to 0.
+	 */
+	return (cfg->msroff ? cfg->msroff : x86_pmu.lbr_nr) - 1;
+}
+
+/*
+ * make sure we have a sane BRS offset to begin with
+ * especially with kexec
+ */
+void amd_brs_reset(void)
+{
+	/*
+	 * Reset config
+	 */
+	set_debug_extn_cfg(0);
+
+	/*
+	 * Mark first entry as poisoned
+	 */
+	wrmsrl(brs_to(0), BRS_POISON);
+}
+
+int __init amd_brs_init(void)
+{
+	if (!amd_brs_detect())
+		return -EOPNOTSUPP;
+
+	pr_cont("%d-deep BRS, ", x86_pmu.lbr_nr);
+
+	return 0;
+}
+
+void amd_brs_enable(void)
+{
+	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
+	union amd_debug_extn_cfg cfg;
+
+	/* Activate only on first user */
+	if (++cpuc->brs_active > 1)
+		return;
+
+	cfg.val    = 0; /* reset all fields */
+	cfg.brsmen = 1; /* enable branch sampling */
+
+	/* Set enable bit */
+	set_debug_extn_cfg(cfg.val);
+}
+
+void amd_brs_enable_all(void)
+{
+	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
+	if (cpuc->lbr_users)
+		amd_brs_enable();
+}
+
+void amd_brs_disable(void)
+{
+	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
+	union amd_debug_extn_cfg cfg;
+
+	/* Check if active (could be disabled via x86_pmu_disable_all()) */
+	if (!cpuc->brs_active)
+		return;
+
+	/* Only disable for last user */
+	if (--cpuc->brs_active)
+		return;
+
+	/*
+	 * Clear the brsmen bit but preserve the others as they contain
+	 * useful state such as vb and msroff
+	 */
+	cfg.val = get_debug_extn_cfg();
+
+	/*
+	 * When coming in on interrupt and BRS is full, then hw will have
+	 * already stopped BRS, no need to issue wrmsr again
+	 */
+	if (cfg.brsmen) {
+		cfg.brsmen = 0;
+		set_debug_extn_cfg(cfg.val);
+	}
+}
+
+void amd_brs_disable_all(void)
+{
+	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
+	if (cpuc->lbr_users)
+		amd_brs_disable();
+}
+
+/*
+ * Caller must ensure amd_brs_inuse() is true before calling
+ * return:
+ */
+void amd_brs_drain(void)
+{
+	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
+	struct perf_event *event = cpuc->events[0];
+	struct perf_branch_entry *br = cpuc->lbr_entries;
+	union amd_debug_extn_cfg cfg;
+	u32 i, nr = 0, num, tos, start;
+	u32 shift = 64 - boot_cpu_data.x86_virt_bits;
+
+	/*
+	 * BRS event forced on PMC0,
+	 * so check if there is an event.
+	 * It is possible to have lbr_users > 0 but the event
+	 * not yet scheduled due to long latency PMU irq
+	 */
+	if (!event)
+		goto empty;
+
+	cfg.val = get_debug_extn_cfg();
+
+	/* Sanity check [0-x86_pmu.lbr_nr] */
+	if (WARN_ON_ONCE(cfg.msroff >= x86_pmu.lbr_nr))
+		goto empty;
+
+	/* No valid branch */
+	if (cfg.vb == 0)
+		goto empty;
+
+	/*
+	 * msr.off points to next entry to be written
+	 * tos = most recent entry index = msr.off - 1
+	 * BRS register buffer saturates, so we know we have
+	 * start < tos and that we have to read from start to tos
+	 */
+	start = 0;
+	tos = amd_brs_get_tos(&cfg);
+
+	num = tos - start + 1;
+
+	/*
+	 * BRS is only one pass (saturation) from MSROFF to depth-1
+	 * MSROFF wraps to zero when buffer is full
+	 */
+	for (i = 0; i < num; i++) {
+		u32 brs_idx = tos - i;
+		u64 from, to;
+
+		rdmsrl(brs_to(brs_idx), to);
+
+		/* Entry does not belong to us (as marked by kernel) */
+		if (to == BRS_POISON)
+			break;
+
+		rdmsrl(brs_from(brs_idx), from);
+
+		/*
+		 * Sign-extend SAMP_BR_TO to 64 bits, bits 61-63 are reserved.
+		 * Necessary to generate proper virtual addresses suitable for
+		 * symbolization
+		 */
+		to = (u64)(((s64)to << shift) >> shift);
+
+		perf_clear_branch_entry_bitfields(br+nr);
+
+		br[nr].from = from;
+		br[nr].to   = to;
+
+		nr++;
+	}
+empty:
+	/* Record number of sampled branches */
+	cpuc->lbr_stack.nr = nr;
+}
+
+/*
+ * Poison most recent entry to prevent reuse by next task
+ * required because BRS entry are not tagged by PID
+ */
+static void amd_brs_poison_buffer(void)
+{
+	union amd_debug_extn_cfg cfg;
+	unsigned int idx;
+
+	/* Get current state */
+	cfg.val = get_debug_extn_cfg();
+
+	/* idx is most recently written entry */
+	idx = amd_brs_get_tos(&cfg);
+
+	/* Poison target of entry */
+	wrmsrl(brs_to(idx), BRS_POISON);
+}
+
+/*
+ * On context switch in, we need to make sure no samples from previous user
+ * are left in the BRS.
+ *
+ * On ctxswin, sched_in = true, called after the PMU has started
+ * On ctxswout, sched_in = false, called before the PMU is stopped
+ */
+void amd_pmu_brs_sched_task(struct perf_event_context *ctx, bool sched_in)
+{
+	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
+
+	/* no active users */
+	if (!cpuc->lbr_users)
+		return;
+
+	/*
+	 * On context switch in, we need to ensure we do not use entries
+	 * from previous BRS user on that CPU, so we poison the buffer as
+	 * a faster way compared to resetting all entries.
+	 */
+	if (sched_in)
+		amd_brs_poison_buffer();
+}
diff --git a/arch/x86/events/amd/core.c b/arch/x86/events/amd/core.c
index 9687a8aef01c..c7ac70d8ed9a 100644
--- a/arch/x86/events/amd/core.c
+++ b/arch/x86/events/amd/core.c
@@ -325,8 +325,16 @@ static inline bool amd_is_pair_event_code(struct hw_perf_event *hwc)
 	}
 }
 
+#define AMD_FAM19H_BRS_EVENT 0xc4 /* RETIRED_TAKEN_BRANCH_INSTRUCTIONS */
+static inline int amd_is_brs_event(struct perf_event *e)
+{
+	return (e->hw.config & AMD64_RAW_EVENT_MASK) == AMD_FAM19H_BRS_EVENT;
+}
+
 static int amd_core_hw_config(struct perf_event *event)
 {
+	int ret = 0;
+
 	if (event->attr.exclude_host && event->attr.exclude_guest)
 		/*
 		 * When HO == GO == 1 the hardware treats that as GO == HO == 0
@@ -343,7 +351,66 @@ static int amd_core_hw_config(struct perf_event *event)
 	if ((x86_pmu.flags & PMU_FL_PAIR) && amd_is_pair_event_code(&event->hw))
 		event->hw.flags |= PERF_X86_EVENT_PAIR;
 
-	return 0;
+	/*
+	 * if branch stack is requested
+	 */
+	if (has_branch_stack(event)) {
+		/*
+		 * Due to interrupt holding, BRS is not recommended in
+		 * counting mode.
+		 */
+		if (!is_sampling_event(event))
+			return -EINVAL;
+
+		/*
+		 * Due to the way BRS operates by holding the interrupt until
+		 * lbr_nr entries have been captured, it does not make sense
+		 * to allow sampling on BRS with an event that does not match
+		 * what BRS is capturing, i.e., retired taken branches.
+		 * Otherwise the correlation with the event's period is even
+		 * more loose:
+		 *
+		 * With retired taken branch:
+		 *   Effective P = P + 16 + X
+		 * With any other event:
+		 *   Effective P = P + Y + X
+		 *
+		 * Where X is the number of taken branches due to interrupt
+		 * skid. Skid is large.
+		 *
+		 * Where Y is the occurences of the event while BRS is
+		 * capturing the lbr_nr entries.
+		 *
+		 * By using retired taken branches, we limit the impact on the
+		 * Y variable. We know it cannot be more than the depth of
+		 * BRS.
+		 */
+		if (!amd_is_brs_event(event))
+			return -EINVAL;
+
+		/*
+		 * BRS implementation does not work with frequency mode
+		 * reprogramming of the period.
+		 */
+		if (event->attr.freq)
+			return -EINVAL;
+		/*
+		 * The kernel subtracts BRS depth from period, so it must
+		 * be big enough.
+		 */
+		if (event->attr.sample_period <= x86_pmu.lbr_nr)
+			return -EINVAL;
+
+		/*
+		 * Check if we can allow PERF_SAMPLE_BRANCH_STACK
+		 */
+		ret = amd_brs_setup_filter(event);
+
+		/* only set in case of success */
+		if (!ret)
+			event->hw.flags |= PERF_X86_EVENT_AMD_BRS;
+	}
+	return ret;
 }
 
 static inline int amd_is_nb_event(struct hw_perf_event *hwc)
@@ -366,7 +433,7 @@ static int amd_pmu_hw_config(struct perf_event *event)
 	if (event->attr.precise_ip && get_ibs_caps())
 		return -ENOENT;
 
-	if (has_branch_stack(event))
+	if (has_branch_stack(event) && !x86_pmu.lbr_nr)
 		return -EOPNOTSUPP;
 
 	ret = x86_pmu_hw_config(event);
@@ -555,6 +622,8 @@ static void amd_pmu_cpu_starting(int cpu)
 
 	cpuc->amd_nb->nb_id = nb_id;
 	cpuc->amd_nb->refcnt++;
+
+	amd_brs_reset();
 }
 
 static void amd_pmu_cpu_dead(int cpu)
@@ -610,6 +679,8 @@ static void amd_pmu_disable_all(void)
 	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
 	int idx;
 
+	amd_brs_disable_all();
+
 	x86_pmu_disable_all();
 
 	/*
@@ -634,6 +705,30 @@ static void amd_pmu_disable_all(void)
 	}
 }
 
+static void amd_pmu_enable_event(struct perf_event *event)
+{
+	x86_pmu_enable_event(event);
+}
+
+static void amd_pmu_enable_all(int added)
+{
+	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
+	struct hw_perf_event *hwc;
+	int idx;
+
+	amd_brs_enable_all();
+
+	for (idx = 0; idx < x86_pmu.num_counters; idx++) {
+		hwc = &cpuc->events[idx]->hw;
+
+		/* only activate events which are marked as active */
+		if (!test_bit(idx, cpuc->active_mask))
+			continue;
+
+		amd_pmu_enable_event(cpuc->events[idx]);
+	}
+}
+
 static void amd_pmu_disable_event(struct perf_event *event)
 {
 	x86_pmu_disable_event(event);
@@ -651,6 +746,18 @@ static void amd_pmu_disable_event(struct perf_event *event)
 	amd_pmu_wait_on_overflow(event->hw.idx);
 }
 
+static void amd_pmu_add_event(struct perf_event *event)
+{
+	if (needs_branch_stack(event))
+		amd_pmu_brs_add(event);
+}
+
+static void amd_pmu_del_event(struct perf_event *event)
+{
+	if (needs_branch_stack(event))
+		amd_pmu_brs_del(event);
+}
+
 /*
  * Because of NMI latency, if multiple PMC counters are active or other sources
  * of NMIs are received, the perf NMI handler can handle one or more overflowed
@@ -671,11 +778,31 @@ static void amd_pmu_disable_event(struct perf_event *event)
  */
 static int amd_pmu_handle_irq(struct pt_regs *regs)
 {
+	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
 	int handled;
+	int pmu_enabled;
+
+	/*
+	 * Save the PMU state.
+	 * It needs to be restored when leaving the handler.
+	 */
+	pmu_enabled = cpuc->enabled;
+	cpuc->enabled = 0;
+
+	/* stop everything (includes BRS) */
+	amd_pmu_disable_all();
+
+	/* Drain BRS is in use (could be inactive) */
+	if (cpuc->lbr_users)
+		amd_brs_drain();
 
 	/* Process any counter overflows */
 	handled = x86_pmu_handle_irq(regs);
 
+	cpuc->enabled = pmu_enabled;
+	if (pmu_enabled)
+		amd_pmu_enable_all(0);
+
 	/*
 	 * If a counter was handled, record a timestamp such that un-handled
 	 * NMIs will be claimed if arriving within that window.
@@ -897,6 +1024,51 @@ static void amd_put_event_constraints_f17h(struct cpu_hw_events *cpuc,
 		--cpuc->n_pair;
 }
 
+/*
+ * Because of the way BRS operates with an inactive and active phases, and
+ * the link to one counter, it is not possible to have two events using BRS
+ * scheduled at the same time. There would be an issue with enforcing the
+ * period of each one and given that the BRS saturates, it would not be possible
+ * to guarantee correlated content for all events. Therefore, in situations
+ * where multiple events want to use BRS, the kernel enforces mutual exclusion.
+ * Exclusion is enforced by chosing only one counter for events using BRS.
+ * The event scheduling logic will then automatically multiplex the
+ * events and ensure that at most one event is actively using BRS.
+ *
+ * The BRS counter could be any counter, but there is no constraint on Fam19h,
+ * therefore all counters are equal and thus we pick the first one: PMC0
+ */
+static struct event_constraint amd_fam19h_brs_cntr0_constraint =
+	EVENT_CONSTRAINT(0, 0x1, AMD64_RAW_EVENT_MASK);
+
+static struct event_constraint amd_fam19h_brs_pair_cntr0_constraint =
+	__EVENT_CONSTRAINT(0, 0x1, AMD64_RAW_EVENT_MASK, 1, 0, PERF_X86_EVENT_PAIR);
+
+static struct event_constraint *
+amd_get_event_constraints_f19h(struct cpu_hw_events *cpuc, int idx,
+			  struct perf_event *event)
+{
+	struct hw_perf_event *hwc = &event->hw;
+	bool has_brs = has_amd_brs(hwc);
+
+	/*
+	 * In case BRS is used with an event requiring a counter pair,
+	 * the kernel allows it but only on counter 0 & 1 to enforce
+	 * multiplexing requiring to protect BRS in case of multiple
+	 * BRS users
+	 */
+	if (amd_is_pair_event_code(hwc)) {
+		return has_brs ? &amd_fam19h_brs_pair_cntr0_constraint
+			       : &pair_constraint;
+	}
+
+	if (has_brs)
+		return &amd_fam19h_brs_cntr0_constraint;
+
+	return &unconstrained;
+}
+
+
 static ssize_t amd_event_sysfs_show(char *page, u64 config)
 {
 	u64 event = (config & ARCH_PERFMON_EVENTSEL_EVENT) |
@@ -905,12 +1077,19 @@ static ssize_t amd_event_sysfs_show(char *page, u64 config)
 	return x86_event_sysfs_show(page, config, event);
 }
 
+static void amd_pmu_sched_task(struct perf_event_context *ctx,
+				 bool sched_in)
+{
+	if (sched_in && x86_pmu.lbr_nr)
+		amd_pmu_brs_sched_task(ctx, sched_in);
+}
+
 static __initconst const struct x86_pmu amd_pmu = {
 	.name			= "AMD",
 	.handle_irq		= amd_pmu_handle_irq,
 	.disable_all		= amd_pmu_disable_all,
-	.enable_all		= x86_pmu_enable_all,
-	.enable			= x86_pmu_enable_event,
+	.enable_all		= amd_pmu_enable_all,
+	.enable			= amd_pmu_enable_event,
 	.disable		= amd_pmu_disable_event,
 	.hw_config		= amd_pmu_hw_config,
 	.schedule_events	= x86_schedule_events,
@@ -920,6 +1099,8 @@ static __initconst const struct x86_pmu amd_pmu = {
 	.event_map		= amd_pmu_event_map,
 	.max_events		= ARRAY_SIZE(amd_perfmon_event_map),
 	.num_counters		= AMD64_NUM_COUNTERS,
+	.add			= amd_pmu_add_event,
+	.del			= amd_pmu_del_event,
 	.cntval_bits		= 48,
 	.cntval_mask		= (1ULL << 48) - 1,
 	.apic			= 1,
@@ -938,6 +1119,37 @@ static __initconst const struct x86_pmu amd_pmu = {
 	.amd_nb_constraints	= 1,
 };
 
+static ssize_t branches_show(struct device *cdev,
+			      struct device_attribute *attr,
+			      char *buf)
+{
+	return snprintf(buf, PAGE_SIZE, "%d\n", x86_pmu.lbr_nr);
+}
+
+static DEVICE_ATTR_RO(branches);
+
+static struct attribute *amd_pmu_brs_attrs[] = {
+	&dev_attr_branches.attr,
+	NULL,
+};
+
+static umode_t
+amd_brs_is_visible(struct kobject *kobj, struct attribute *attr, int i)
+{
+	return x86_pmu.lbr_nr ? attr->mode : 0;
+}
+
+static struct attribute_group group_caps_amd_brs = {
+	.name  = "caps",
+	.attrs = amd_pmu_brs_attrs,
+	.is_visible = amd_brs_is_visible,
+};
+
+static const struct attribute_group *amd_attr_update[] = {
+	&group_caps_amd_brs,
+	NULL,
+};
+
 static int __init amd_core_pmu_init(void)
 {
 	u64 even_ctr_mask = 0ULL;
@@ -989,6 +1201,19 @@ static int __init amd_core_pmu_init(void)
 		x86_pmu.flags |= PMU_FL_PAIR;
 	}
 
+	/*
+	 * BRS requires special event constraints and flushing on ctxsw.
+	 */
+	if (boot_cpu_data.x86 >= 0x19 && !amd_brs_init()) {
+		x86_pmu.get_event_constraints = amd_get_event_constraints_f19h;
+		x86_pmu.sched_task = amd_pmu_sched_task;
+		/*
+		 * put_event_constraints callback same as Fam17h, set above
+		 */
+	}
+
+	x86_pmu.attr_update = amd_attr_update;
+
 	pr_cont("core perfctr, ");
 	return 0;
 }
diff --git a/arch/x86/events/core.c b/arch/x86/events/core.c
index e686c5e0537b..c2a890caeb0a 100644
--- a/arch/x86/events/core.c
+++ b/arch/x86/events/core.c
@@ -1338,6 +1338,10 @@ static void x86_pmu_enable(struct pmu *pmu)
 			if (hwc->state & PERF_HES_ARCH)
 				continue;
 
+			/*
+			 * if cpuc->enabled = 0, then no wrmsr as
+			 * per x86_pmu_enable_event()
+			 */
 			x86_pmu_start(event, PERF_EF_RELOAD);
 		}
 		cpuc->n_added = 0;
@@ -1704,11 +1708,15 @@ int x86_pmu_handle_irq(struct pt_regs *regs)
 		 * event overflow
 		 */
 		handled++;
-		perf_sample_data_init(&data, 0, event->hw.last_period);
 
 		if (!x86_perf_event_set_period(event))
 			continue;
 
+		perf_sample_data_init(&data, 0, event->hw.last_period);
+
+		if (has_branch_stack(event))
+			data.br_stack = &cpuc->lbr_stack;
+
 		if (perf_event_overflow(event, &data, regs))
 			x86_pmu_stop(event, 0);
 	}
diff --git a/arch/x86/events/perf_event.h b/arch/x86/events/perf_event.h
index 150261d929b9..3485a4cf0241 100644
--- a/arch/x86/events/perf_event.h
+++ b/arch/x86/events/perf_event.h
@@ -67,22 +67,23 @@ static inline bool constraint_match(struct event_constraint *c, u64 ecode)
 /*
  * struct hw_perf_event.flags flags
  */
-#define PERF_X86_EVENT_PEBS_LDLAT	0x0001 /* ld+ldlat data address sampling */
-#define PERF_X86_EVENT_PEBS_ST		0x0002 /* st data address sampling */
-#define PERF_X86_EVENT_PEBS_ST_HSW	0x0004 /* haswell style datala, store */
-#define PERF_X86_EVENT_PEBS_LD_HSW	0x0008 /* haswell style datala, load */
-#define PERF_X86_EVENT_PEBS_NA_HSW	0x0010 /* haswell style datala, unknown */
-#define PERF_X86_EVENT_EXCL		0x0020 /* HT exclusivity on counter */
-#define PERF_X86_EVENT_DYNAMIC		0x0040 /* dynamic alloc'd constraint */
-
-#define PERF_X86_EVENT_EXCL_ACCT	0x0100 /* accounted EXCL event */
-#define PERF_X86_EVENT_AUTO_RELOAD	0x0200 /* use PEBS auto-reload */
-#define PERF_X86_EVENT_LARGE_PEBS	0x0400 /* use large PEBS */
-#define PERF_X86_EVENT_PEBS_VIA_PT	0x0800 /* use PT buffer for PEBS */
-#define PERF_X86_EVENT_PAIR		0x1000 /* Large Increment per Cycle */
-#define PERF_X86_EVENT_LBR_SELECT	0x2000 /* Save/Restore MSR_LBR_SELECT */
-#define PERF_X86_EVENT_TOPDOWN		0x4000 /* Count Topdown slots/metrics events */
-#define PERF_X86_EVENT_PEBS_STLAT	0x8000 /* st+stlat data address sampling */
+#define PERF_X86_EVENT_PEBS_LDLAT	0x00001 /* ld+ldlat data address sampling */
+#define PERF_X86_EVENT_PEBS_ST		0x00002 /* st data address sampling */
+#define PERF_X86_EVENT_PEBS_ST_HSW	0x00004 /* haswell style datala, store */
+#define PERF_X86_EVENT_PEBS_LD_HSW	0x00008 /* haswell style datala, load */
+#define PERF_X86_EVENT_PEBS_NA_HSW	0x00010 /* haswell style datala, unknown */
+#define PERF_X86_EVENT_EXCL		0x00020 /* HT exclusivity on counter */
+#define PERF_X86_EVENT_DYNAMIC		0x00040 /* dynamic alloc'd constraint */
+
+#define PERF_X86_EVENT_EXCL_ACCT	0x00100 /* accounted EXCL event */
+#define PERF_X86_EVENT_AUTO_RELOAD	0x00200 /* use PEBS auto-reload */
+#define PERF_X86_EVENT_LARGE_PEBS	0x00400 /* use large PEBS */
+#define PERF_X86_EVENT_PEBS_VIA_PT	0x00800 /* use PT buffer for PEBS */
+#define PERF_X86_EVENT_PAIR		0x01000 /* Large Increment per Cycle */
+#define PERF_X86_EVENT_LBR_SELECT	0x02000 /* Save/Restore MSR_LBR_SELECT */
+#define PERF_X86_EVENT_TOPDOWN		0x04000 /* Count Topdown slots/metrics events */
+#define PERF_X86_EVENT_PEBS_STLAT	0x08000 /* st+stlat data address sampling */
+#define PERF_X86_EVENT_AMD_BRS		0x10000 /* AMD Branch Sampling */
 
 static inline bool is_topdown_count(struct perf_event *event)
 {
@@ -325,6 +326,8 @@ struct cpu_hw_events {
 	 * AMD specific bits
 	 */
 	struct amd_nb			*amd_nb;
+	int				brs_active; /* BRS is enabled */
+
 	/* Inverted mask of bits to clear in the perf_ctr ctrl registers */
 	u64				perf_ctr_virt_mask;
 	int				n_pair; /* Large increment events */
@@ -1105,6 +1108,11 @@ int x86_pmu_hw_config(struct perf_event *event);
 
 void x86_pmu_disable_all(void);
 
+static inline bool has_amd_brs(struct hw_perf_event *hwc)
+{
+	return hwc->flags & PERF_X86_EVENT_AMD_BRS;
+}
+
 static inline bool is_counter_pair(struct hw_perf_event *hwc)
 {
 	return hwc->flags & PERF_X86_EVENT_PAIR;
@@ -1210,6 +1218,50 @@ static inline bool fixed_counter_disabled(int i, struct pmu *pmu)
 #ifdef CONFIG_CPU_SUP_AMD
 
 int amd_pmu_init(void);
+int amd_brs_init(void);
+void amd_brs_disable(void);
+void amd_brs_enable(void);
+void amd_brs_enable_all(void);
+void amd_brs_disable_all(void);
+void amd_brs_drain(void);
+void amd_brs_disable_all(void);
+int amd_brs_setup_filter(struct perf_event *event);
+void amd_brs_reset(void);
+
+static inline void amd_pmu_brs_add(struct perf_event *event)
+{
+	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
+
+	perf_sched_cb_inc(event->ctx->pmu);
+	cpuc->lbr_users++;
+	/*
+	 * No need to reset BRS because it is reset
+	 * on brs_enable() and it is saturating
+	 */
+}
+
+static inline void amd_pmu_brs_del(struct perf_event *event)
+{
+	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
+
+	cpuc->lbr_users--;
+	WARN_ON_ONCE(cpuc->lbr_users < 0);
+
+	perf_sched_cb_dec(event->ctx->pmu);
+}
+
+void amd_pmu_brs_sched_task(struct perf_event_context *ctx, bool sched_in);
+
+/*
+ * check if BRS is activated on the CPU
+ * active defined as it has non-zero users and DBG_EXT_CFG.BRSEN=1
+ */
+static inline bool amd_brs_active(void)
+{
+	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
+
+	return cpuc->brs_active;
+}
 
 #else /* CONFIG_CPU_SUP_AMD */
 
@@ -1218,6 +1270,23 @@ static inline int amd_pmu_init(void)
 	return 0;
 }
 
+static inline int amd_brs_init(void)
+{
+	return 0;
+}
+
+static inline void amd_brs_drain(void)
+{
+}
+
+static inline void amd_brs_enable_all(void)
+{
+}
+
+static inline void amd_brs_disable_all(void)
+{
+}
+
 #endif /* CONFIG_CPU_SUP_AMD */
 
 static inline int is_pebs_pt(struct perf_event *event)
diff --git a/arch/x86/include/asm/msr-index.h b/arch/x86/include/asm/msr-index.h
index 87f80f0e5259..0179c72e829a 100644
--- a/arch/x86/include/asm/msr-index.h
+++ b/arch/x86/include/asm/msr-index.h
@@ -688,6 +688,10 @@
 #define MSR_IA32_PERF_CTL		0x00000199
 #define INTEL_PERF_CTL_MASK		0xffff
 
+/* AMD Branch Sampling configuration */
+#define MSR_AMD_DBG_EXTN_CFG		0xc000010f
+#define MSR_AMD_SAMP_BR_FROM		0xc0010300
+
 #define MSR_IA32_MPERF			0x000000e7
 #define MSR_IA32_APERF			0x000000e8
 
-- 
2.35.1.894.gb6a874cedc-goog

[PATCH v7 04/13] perf/x86/amd: add branch-brs helper event for Fam19h BRS

[Stephane Eranian]

Add a pseudo event called branch-brs to help use the FAM Fam19h
Branch Sampling feature (BRS). BRS samples taken branches, so it is best used
when sampling on a retired taken branch event (0xc4) which is what BRS
captures.  Instead of trying to remember the event code or actual event name,
users can simply do:

$ perf record -b -e cpu/branch-brs/ -c 1000037 .....

Signed-off-by: Stephane Eranian <eranian@google.com>
---
 arch/x86/events/amd/core.c | 15 +++++++++++++++
 1 file changed, 15 insertions(+)

diff --git a/arch/x86/events/amd/core.c b/arch/x86/events/amd/core.c
index c7ac70d8ed9a..f7bce8364fe4 100644
--- a/arch/x86/events/amd/core.c
+++ b/arch/x86/events/amd/core.c
@@ -1145,8 +1145,23 @@ static struct attribute_group group_caps_amd_brs = {
 	.is_visible = amd_brs_is_visible,
 };
 
+EVENT_ATTR_STR(branch-brs, amd_branch_brs,
+	       "event=" __stringify(AMD_FAM19H_BRS_EVENT)"\n");
+
+static struct attribute *amd_brs_events_attrs[] = {
+	EVENT_PTR(amd_branch_brs),
+	NULL,
+};
+
+static struct attribute_group group_events_amd_brs = {
+	.name       = "events",
+	.attrs      = amd_brs_events_attrs,
+	.is_visible = amd_brs_is_visible,
+};
+
 static const struct attribute_group *amd_attr_update[] = {
 	&group_caps_amd_brs,
+	&group_events_amd_brs,
 	NULL,
 };
 
-- 
2.35.1.894.gb6a874cedc-goog

[PATCH v7 05/13] perf/x86/amd: enable branch sampling priv level filtering

[Stephane Eranian]

The AMD Branch Sampling features does not provide hardware filtering by
privilege level. The associated PMU counter does but not the branch sampling
by itself. Given how BRS operates there is a possibility that BRS captures
kernel level branches even though the event is programmed to count only at
the user level.

Implement a workaround in software by removing the branches which belong to
the wrong privilege level. The privilege level is evaluated on the target of
the branch and not the source so as to be compatible with other architectures.
As a consequence of this patch, the number of entries in the
PERF_RECORD_BRANCH_STACK buffer may be less than the maximum (16).  It could
even be zero. Another consequence is that consecutive entries in the branch
stack may not reflect actual code path and may have discontinuities, in case
kernel branches were suppressed. But this is no different than what happens
on other architectures.

Signed-off-by: Stephane Eranian <eranian@google.com>
---
 arch/x86/events/amd/brs.c | 26 ++++++++++++++++++++------
 1 file changed, 20 insertions(+), 6 deletions(-)

diff --git a/arch/x86/events/amd/brs.c b/arch/x86/events/amd/brs.c
index 3c13c484c637..40461c3ce714 100644
--- a/arch/x86/events/amd/brs.c
+++ b/arch/x86/events/amd/brs.c
@@ -92,10 +92,6 @@ int amd_brs_setup_filter(struct perf_event *event)
 	if ((type & ~PERF_SAMPLE_BRANCH_PLM_ALL) != PERF_SAMPLE_BRANCH_ANY)
 		return -EINVAL;
 
-	/* can only capture at all priv levels due to the way BRS works */
-	if ((type & PERF_SAMPLE_BRANCH_PLM_ALL) != PERF_SAMPLE_BRANCH_PLM_ALL)
-		return -EINVAL;
-
 	return 0;
 }
 
@@ -195,6 +191,21 @@ void amd_brs_disable_all(void)
 		amd_brs_disable();
 }
 
+static bool amd_brs_match_plm(struct perf_event *event, u64 to)
+{
+	int type = event->attr.branch_sample_type;
+	int plm_k = PERF_SAMPLE_BRANCH_KERNEL | PERF_SAMPLE_BRANCH_HV;
+	int plm_u = PERF_SAMPLE_BRANCH_USER;
+
+	if (!(type & plm_k) && kernel_ip(to))
+		return 0;
+
+	if (!(type & plm_u) && !kernel_ip(to))
+		return 0;
+
+	return 1;
+}
+
 /*
  * Caller must ensure amd_brs_inuse() is true before calling
  * return:
@@ -252,8 +263,6 @@ void amd_brs_drain(void)
 		if (to == BRS_POISON)
 			break;
 
-		rdmsrl(brs_from(brs_idx), from);
-
 		/*
 		 * Sign-extend SAMP_BR_TO to 64 bits, bits 61-63 are reserved.
 		 * Necessary to generate proper virtual addresses suitable for
@@ -261,6 +270,11 @@ void amd_brs_drain(void)
 		 */
 		to = (u64)(((s64)to << shift) >> shift);
 
+		if (!amd_brs_match_plm(event, to))
+			continue;
+
+		rdmsrl(brs_from(brs_idx), from);
+
 		perf_clear_branch_entry_bitfields(br+nr);
 
 		br[nr].from = from;
-- 
2.35.1.894.gb6a874cedc-goog

[PATCH v7 06/13] perf/x86/amd: add AMD branch sampling period adjustment

[Stephane Eranian]

Add code to adjust the sampling event period when used with the Branch
Sampling feature (BRS). Given the depth of the BRS (16), the period is
reduced by that depth such that in the best case scenario, BRS saturates at
the desired sampling period. In practice, though, the processor may execute
more branches. Given a desired period P and a depth D, the kernel programs
the actual period at P - D. After P occurrences of the sampling event, the
counter overflows. It then may take X branches (skid) before the NMI is
caught and held by the hardware and BRS activates. Then, after D branches,
BRS saturates and the NMI is delivered.  With no skid, the effective period
would be (P - D) + D = P. In practice, however, it will likely be (P - D) +
X + D. There is no way to eliminate X or predict X.

Signed-off-by: Stephane Eranian <eranian@google.com>
---
 arch/x86/events/core.c       |  7 +++++++
 arch/x86/events/perf_event.h | 12 ++++++++++++
 2 files changed, 19 insertions(+)

diff --git a/arch/x86/events/core.c b/arch/x86/events/core.c
index c2a890caeb0a..ed285f640efe 100644
--- a/arch/x86/events/core.c
+++ b/arch/x86/events/core.c
@@ -1374,6 +1374,13 @@ int x86_perf_event_set_period(struct perf_event *event)
 	    x86_pmu.set_topdown_event_period)
 		return x86_pmu.set_topdown_event_period(event);
 
+	/*
+	 * decrease period by the depth of the BRS feature to get
+	 * the last N taken branches and approximate the desired period
+	 */
+	if (has_branch_stack(event))
+		period = amd_brs_adjust_period(period);
+
 	/*
 	 * If we are way outside a reasonable range then just skip forward:
 	 */
diff --git a/arch/x86/events/perf_event.h b/arch/x86/events/perf_event.h
index 3485a4cf0241..25b037b571e4 100644
--- a/arch/x86/events/perf_event.h
+++ b/arch/x86/events/perf_event.h
@@ -1263,6 +1263,14 @@ static inline bool amd_brs_active(void)
 	return cpuc->brs_active;
 }
 
+static inline s64 amd_brs_adjust_period(s64 period)
+{
+	if (period > x86_pmu.lbr_nr)
+		return period - x86_pmu.lbr_nr;
+
+	return period;
+}
+
 #else /* CONFIG_CPU_SUP_AMD */
 
 static inline int amd_pmu_init(void)
@@ -1287,6 +1295,10 @@ static inline void amd_brs_disable_all(void)
 {
 }
 
+static inline s64 amd_brs_adjust_period(s64 period)
+{
+	return period;
+}
 #endif /* CONFIG_CPU_SUP_AMD */
 
 static inline int is_pebs_pt(struct perf_event *event)
-- 
2.35.1.894.gb6a874cedc-goog

[PATCH v7 07/13] perf/x86/amd: make Zen3 branch sampling opt-in

[Stephane Eranian]

Add a kernel config option CONFIG_PERF_EVENTS_AMD_BRS
to make the support for AMD Zen3 Branch Sampling (BRS) an opt-in
compile time option.

Signed-off-by: Stephane Eranian <eranian@google.com>
---
 arch/x86/events/Kconfig      |  8 ++++++
 arch/x86/events/amd/Makefile |  3 ++-
 arch/x86/events/perf_event.h | 49 ++++++++++++++++++++++++++++--------
 3 files changed, 49 insertions(+), 11 deletions(-)

diff --git a/arch/x86/events/Kconfig b/arch/x86/events/Kconfig
index d6cdfe631674..09c56965750a 100644
--- a/arch/x86/events/Kconfig
+++ b/arch/x86/events/Kconfig
@@ -44,4 +44,12 @@ config PERF_EVENTS_AMD_UNCORE
 
 	  To compile this driver as a module, choose M here: the
 	  module will be called 'amd-uncore'.
+
+config PERF_EVENTS_AMD_BRS
+	depends on PERF_EVENTS && CPU_SUP_AMD
+	bool "AMD Zen3 Branch Sampling support"
+	help
+	  Enable AMD Zen3 branch sampling support (BRS) which samples up to
+	  16 consecutive taken branches in registers.
+
 endmenu
diff --git a/arch/x86/events/amd/Makefile b/arch/x86/events/amd/Makefile
index cf323ffab5cd..b9f5d4610256 100644
--- a/arch/x86/events/amd/Makefile
+++ b/arch/x86/events/amd/Makefile
@@ -1,5 +1,6 @@
 # SPDX-License-Identifier: GPL-2.0
-obj-$(CONFIG_CPU_SUP_AMD)		+= core.o brs.o
+obj-$(CONFIG_CPU_SUP_AMD)		+= core.o
+obj-$(CONFIG_PERF_EVENTS_AMD_BRS)	+= brs.o
 obj-$(CONFIG_PERF_EVENTS_AMD_POWER)	+= power.o
 obj-$(CONFIG_X86_LOCAL_APIC)		+= ibs.o
 obj-$(CONFIG_PERF_EVENTS_AMD_UNCORE)	+= amd-uncore.o
diff --git a/arch/x86/events/perf_event.h b/arch/x86/events/perf_event.h
index 25b037b571e4..4d050579dcbd 100644
--- a/arch/x86/events/perf_event.h
+++ b/arch/x86/events/perf_event.h
@@ -1218,6 +1218,8 @@ static inline bool fixed_counter_disabled(int i, struct pmu *pmu)
 #ifdef CONFIG_CPU_SUP_AMD
 
 int amd_pmu_init(void);
+
+#ifdef CONFIG_PERF_EVENTS_AMD_BRS
 int amd_brs_init(void);
 void amd_brs_disable(void);
 void amd_brs_enable(void);
@@ -1252,25 +1254,52 @@ static inline void amd_pmu_brs_del(struct perf_event *event)
 
 void amd_pmu_brs_sched_task(struct perf_event_context *ctx, bool sched_in);
 
-/*
- * check if BRS is activated on the CPU
- * active defined as it has non-zero users and DBG_EXT_CFG.BRSEN=1
- */
-static inline bool amd_brs_active(void)
+static inline s64 amd_brs_adjust_period(s64 period)
 {
-	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
+	if (period > x86_pmu.lbr_nr)
+		return period - x86_pmu.lbr_nr;
 
-	return cpuc->brs_active;
+	return period;
+}
+#else
+static inline int amd_brs_init(void)
+{
+	return 0;
 }
+static inline void amd_brs_disable(void) {}
+static inline void amd_brs_enable(void) {}
+static inline void amd_brs_drain(void) {}
+static inline void amd_brs_lopwr_init(void) {}
+static inline void amd_brs_disable_all(void) {}
+static inline int amd_brs_setup_filter(struct perf_event *event)
+{
+	return 0;
+}
+static inline void amd_brs_reset(void) {}
 
-static inline s64 amd_brs_adjust_period(s64 period)
+static inline void amd_pmu_brs_add(struct perf_event *event)
 {
-	if (period > x86_pmu.lbr_nr)
-		return period - x86_pmu.lbr_nr;
+}
+
+static inline void amd_pmu_brs_del(struct perf_event *event)
+{
+}
+
+static inline void amd_pmu_brs_sched_task(struct perf_event_context *ctx, bool sched_in)
+{
+}
 
+static inline s64 amd_brs_adjust_period(s64 period)
+{
 	return period;
 }
 
+static inline void amd_brs_enable_all(void)
+{
+}
+
+#endif
+
 #else /* CONFIG_CPU_SUP_AMD */
 
 static inline int amd_pmu_init(void)
-- 
2.35.1.894.gb6a874cedc-goog

[PATCH v7 08/13] ACPI: add perf low power callback

[Stephane Eranian]

Add an optional callback needed by some PMU features, e.g., AMD
BRS, to give a chance to the perf_events code to change its state before
a CPU goes to low power and after it comes back.

The callback is void when the PERF_NEEDS_LOPWR_CB flag is not set.
This flag must be set in arch specific perf_event.h header whenever needed.
When not set, there is no impact on the ACPI code.

Signed-off-by: Stephane Eranian <eranian@google.com>
---
 drivers/acpi/acpi_pad.c       | 6 ++++++
 drivers/acpi/processor_idle.c | 5 +++++
 include/linux/perf_event.h    | 6 ++++++
 3 files changed, 17 insertions(+)

diff --git a/drivers/acpi/acpi_pad.c b/drivers/acpi/acpi_pad.c
index f45979aa2d64..a306a07a60b5 100644
--- a/drivers/acpi/acpi_pad.c
+++ b/drivers/acpi/acpi_pad.c
@@ -164,6 +164,9 @@ static int power_saving_thread(void *data)
 				tsc_marked_unstable = 1;
 			}
 			local_irq_disable();
+
+			perf_lopwr_cb(true);
+
 			tick_broadcast_enable();
 			tick_broadcast_enter();
 			stop_critical_timings();
@@ -172,6 +175,9 @@ static int power_saving_thread(void *data)
 
 			start_critical_timings();
 			tick_broadcast_exit();
+
+			perf_lopwr_cb(false);
+
 			local_irq_enable();
 
 			if (time_before(expire_time, jiffies)) {
diff --git a/drivers/acpi/processor_idle.c b/drivers/acpi/processor_idle.c
index f8e9fa82cb9b..f83596960d9a 100644
--- a/drivers/acpi/processor_idle.c
+++ b/drivers/acpi/processor_idle.c
@@ -21,6 +21,7 @@
 #include <linux/cpuidle.h>
 #include <linux/cpu.h>
 #include <linux/minmax.h>
+#include <linux/perf_event.h>
 #include <acpi/processor.h>
 
 /*
@@ -549,6 +550,8 @@ static void wait_for_freeze(void)
  */
 static void __cpuidle acpi_idle_do_entry(struct acpi_processor_cx *cx)
 {
+	perf_lopwr_cb(true);
+
 	if (cx->entry_method == ACPI_CSTATE_FFH) {
 		/* Call into architectural FFH based C-state */
 		acpi_processor_ffh_cstate_enter(cx);
@@ -559,6 +562,8 @@ static void __cpuidle acpi_idle_do_entry(struct acpi_processor_cx *cx)
 		inb(cx->address);
 		wait_for_freeze();
 	}
+
+	perf_lopwr_cb(false);
 }
 
 /**
diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index 496eb6aa6e54..1b98e46588bc 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -1676,4 +1676,10 @@ typedef int (perf_snapshot_branch_stack_t)(struct perf_branch_entry *entries,
 					   unsigned int cnt);
 DECLARE_STATIC_CALL(perf_snapshot_branch_stack, perf_snapshot_branch_stack_t);
 
+#ifndef PERF_NEEDS_LOPWR_CB
+static inline void perf_lopwr_cb(bool mode)
+{
+}
+#endif
+
 #endif /* _LINUX_PERF_EVENT_H */
-- 
2.35.1.894.gb6a874cedc-goog

[PATCH v7 09/13] perf/x86/amd: add idle hooks for branch sampling

[Stephane Eranian]

On AMD Fam19h Zen3, the branch sampling (BRS) feature must be disabled before
entering low power and re-enabled (if was active) when returning from low
power. Otherwise, the NMI interrupt may be held up for too long and cause
problems. Stopping BRS will cause the NMI to be delivered if it was held up.

Define a perf_amd_brs_lopwr_cb() callback to stop/restart BRS.  The callback
is protected by a jump label which is enabled only when AMD BRS is detected.
In all other cases, the callback is never called.

Signed-off-by: Stephane Eranian <eranian@google.com>
---
 arch/x86/events/amd/brs.c         | 32 +++++++++++++++++++++++++++++++
 arch/x86/events/amd/core.c        |  4 ++++
 arch/x86/events/perf_event.h      |  1 +
 arch/x86/include/asm/perf_event.h | 21 ++++++++++++++++++++
 4 files changed, 58 insertions(+)

diff --git a/arch/x86/events/amd/brs.c b/arch/x86/events/amd/brs.c
index 40461c3ce714..185a58cea917 100644
--- a/arch/x86/events/amd/brs.c
+++ b/arch/x86/events/amd/brs.c
@@ -7,6 +7,7 @@
  * Contributed by Stephane Eranian <eranian@google.com>
  */
 #include <linux/kernel.h>
+#include <linux/jump_label.h>
 #include <asm/msr.h>
 #include <asm/cpufeature.h>
 
@@ -329,3 +330,34 @@ void amd_pmu_brs_sched_task(struct perf_event_context *ctx, bool sched_in)
 	if (sched_in)
 		amd_brs_poison_buffer();
 }
+
+DEFINE_STATIC_KEY_FALSE(perf_lopwr_needed);
+
+/*
+ * called from ACPI processor_idle.c or acpi_pad.c
+ * with interrupts disabled
+ */
+void perf_amd_brs_lopwr_cb(bool lopwr_in)
+{
+	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
+	union amd_debug_extn_cfg cfg;
+
+	/*
+	 * on mwait in, we may end up in non C0 state.
+	 * we must disable branch sampling to avoid holding the NMI
+	 * for too long. We disable it in hardware but we
+	 * keep the state in cpuc, so we can re-enable.
+	 *
+	 * The hardware will deliver the NMI if needed when brsmen cleared
+	 */
+	if (cpuc->brs_active) {
+		cfg.val = get_debug_extn_cfg();
+		cfg.brsmen = !lopwr_in;
+		set_debug_extn_cfg(cfg.val);
+	}
+}
+
+void __init amd_brs_lopwr_init(void)
+{
+	static_branch_enable(&perf_lopwr_needed);
+}
diff --git a/arch/x86/events/amd/core.c b/arch/x86/events/amd/core.c
index f7bce8364fe4..8e1e818f8195 100644
--- a/arch/x86/events/amd/core.c
+++ b/arch/x86/events/amd/core.c
@@ -1,5 +1,6 @@
 // SPDX-License-Identifier: GPL-2.0-only
 #include <linux/perf_event.h>
+#include <linux/jump_label.h>
 #include <linux/export.h>
 #include <linux/types.h>
 #include <linux/init.h>
@@ -1225,6 +1226,9 @@ static int __init amd_core_pmu_init(void)
 		/*
 		 * put_event_constraints callback same as Fam17h, set above
 		 */
+
+		/* branch sampling must be stopped when entering low power */
+		amd_brs_lopwr_init();
 	}
 
 	x86_pmu.attr_update = amd_attr_update;
diff --git a/arch/x86/events/perf_event.h b/arch/x86/events/perf_event.h
index 4d050579dcbd..2ed7bf5b51b1 100644
--- a/arch/x86/events/perf_event.h
+++ b/arch/x86/events/perf_event.h
@@ -1226,6 +1226,7 @@ void amd_brs_enable(void);
 void amd_brs_enable_all(void);
 void amd_brs_disable_all(void);
 void amd_brs_drain(void);
+void amd_brs_lopwr_init(void);
 void amd_brs_disable_all(void);
 int amd_brs_setup_filter(struct perf_event *event);
 void amd_brs_reset(void);
diff --git a/arch/x86/include/asm/perf_event.h b/arch/x86/include/asm/perf_event.h
index 58d9e4b1fa0a..42753a9dc3ed 100644
--- a/arch/x86/include/asm/perf_event.h
+++ b/arch/x86/include/asm/perf_event.h
@@ -513,6 +513,27 @@ static inline void intel_pt_handle_vmx(int on)
 #if defined(CONFIG_PERF_EVENTS) && defined(CONFIG_CPU_SUP_AMD)
  extern void amd_pmu_enable_virt(void);
  extern void amd_pmu_disable_virt(void);
+
+#if defined(CONFIG_PERF_EVENTS_AMD_BRS)
+
+#define PERF_NEEDS_LOPWR_CB 1
+
+/*
+ * architectural low power callback impacts
+ * drivers/acpi/processor_idle.c
+ * drivers/acpi/acpi_pad.c
+ */
+extern void perf_amd_brs_lopwr_cb(bool lopwr_in);
+DECLARE_STATIC_KEY_FALSE(perf_lopwr_needed);
+
+static inline void perf_lopwr_cb(bool mode)
+{
+	/* key enabled only when BRS is available */
+	if (static_branch_unlikely(&perf_lopwr_needed))
+		perf_amd_brs_lopwr_cb(mode);
+}
+#endif /* PERF_NEEDS_LOPWR_CB */
+
 #else
  static inline void amd_pmu_enable_virt(void) { }
  static inline void amd_pmu_disable_virt(void) { }
-- 
2.35.1.894.gb6a874cedc-goog

[PATCH v7 10/13] perf tools: fix NULL point in evsel__env()

[Stephane Eranian]

evsel_env() cannot return NULL. In case evsel->evlist->env is NULL default
to perf_env. Fixes crashes in evsel__open_strerror()

Signed-off-by: Ravi Bangoria <ravi.bangoria@amd.com>
Reviewed-by: Stephane Eranian <eranian@google.com>
---
 tools/perf/util/evsel.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
index 22d3267ce294..14b0e7ffa2c7 100644
--- a/tools/perf/util/evsel.c
+++ b/tools/perf/util/evsel.c
@@ -2965,7 +2965,7 @@ int evsel__open_strerror(struct evsel *evsel, struct target *target,
 
 struct perf_env *evsel__env(struct evsel *evsel)
 {
-	if (evsel && evsel->evlist)
+	if (evsel && evsel->evlist && evsel->evlist->env)
 		return evsel->evlist->env;
 	return &perf_env;
 }
-- 
2.35.1.894.gb6a874cedc-goog

[PATCH v7 11/13] perf tools: Improve IBS error handling

[Stephane Eranian]

From: Kim Phillips <kim.phillips@amd.com>

improve the error message returned on failed perf_event_open() on AMD when
using IBS.

Output of executing 'perf record -e ibs_op// true' BEFORE this patch:

The sys_perf_event_open() syscall returned with 22 (Invalid argument)for event (ibs_op//u).
/bin/dmesg | grep -i perf may provide additional information.

Output after:

AMD IBS cannot exclude kernel events.  Try running at a higher privilege level.

Output of executing 'sudo perf record -e ibs_op// true' BEFORE this patch:

Error:
The sys_perf_event_open() syscall returned with 22 (Invalid argument) for event (ibs_op//).
/bin/dmesg | grep -i perf may provide additional information.

Output after:

Error:
AMD IBS may only be available in system-wide/per-cpu mode.  Try using -a, or -C and workload affinity

Signed-off-by: Kim Phillips <kim.phillips@amd.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Joao Martins <joao.m.martins@oracle.com>
Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Robert Richter <robert.richter@amd.com>
Cc: Stephane Eranian <eranian@google.com>
---
 tools/perf/util/evsel.c | 25 +++++++++++++++++++++++++
 1 file changed, 25 insertions(+)

diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
index 14b0e7ffa2c7..e8ff7a4bd490 100644
--- a/tools/perf/util/evsel.c
+++ b/tools/perf/util/evsel.c
@@ -2847,9 +2847,23 @@ static bool find_process(const char *name)
 	return ret ? false : true;
 }
 
+static bool is_amd(const char *arch, const char *cpuid)
+{
+	return arch && !strcmp("x86", arch) && cpuid && strstarts(cpuid, "AuthenticAMD");
+}
+
+static bool is_amd_ibs(struct evsel *evsel)
+{
+	return evsel->core.attr.precise_ip
+	    || (evsel->pmu_name && !strncmp(evsel->pmu_name, "ibs", 3));
+}
+
 int evsel__open_strerror(struct evsel *evsel, struct target *target,
 			 int err, char *msg, size_t size)
 {
+	struct perf_env *env = evsel__env(evsel);
+	const char *arch = perf_env__arch(env);
+	const char *cpuid = perf_env__cpuid(env);
 	char sbuf[STRERR_BUFSIZE];
 	int printed = 0, enforced = 0;
 
@@ -2949,6 +2963,17 @@ int evsel__open_strerror(struct evsel *evsel, struct target *target,
 			return scnprintf(msg, size,
 	"Invalid event (%s) in per-thread mode, enable system wide with '-a'.",
 					evsel__name(evsel));
+		if (is_amd(arch, cpuid)) {
+			if (is_amd_ibs(evsel)) {
+				if (evsel->core.attr.exclude_kernel)
+					return scnprintf(msg, size,
+	"AMD IBS can't exclude kernel events.  Try running at a higher privilege level.");
+				if (!evsel->core.system_wide)
+					return scnprintf(msg, size,
+	"AMD IBS may only be available in system-wide/per-cpu mode.  Try using -a, or -C and workload affinity");
+			}
+		}
+
 		break;
 	case ENODATA:
 		return scnprintf(msg, size, "Cannot collect data source with the load latency event alone. "
-- 
2.35.1.894.gb6a874cedc-goog

[PATCH v7 12/13] perf tools: Improve error handling of AMD Branch Sampling

[Stephane Eranian]

Improve the error message printed by perf when perf_event_open() fails on
AMD Zen3 when using the branch sampling feature. In the case of EINVAL, there
are two main reasons: frequency mode or period is smaller than the depth of
the branch sampling buffer (16). The patch checks the parameters of the call
and tries to print a relevant message to explain the error:

$ perf record -b -e cpu/branch-brs/ -c 10 ls
Error:
AMD Branch Sampling does not support sampling period smaller than what is reported in /sys/devices/cpu/caps/branches.

$ perf record -b -e cpu/branch-brs/ ls
Error:
AMD Branch Sampling does not support frequency mode sampling, must pass a fixed sampling period via -c option or cpu/branch-brs,period=xxxx/.

Signed-off-by: Stephane Eranian <eranian@google.com>
[Rebased on commit 9fe8895a27a84 ("perf env: Add perf_env__cpuid, perf_env__{nr_}pmu_mappings")]
Signed-off-by: Kim Phillips <kim.phillips@amd.com>
---
 tools/perf/util/evsel.c | 43 ++++++++++++++++++++++++++++++-----------
 1 file changed, 32 insertions(+), 11 deletions(-)

diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
index e8ff7a4bd490..03f720a85764 100644
--- a/tools/perf/util/evsel.c
+++ b/tools/perf/util/evsel.c
@@ -2858,6 +2858,16 @@ static bool is_amd_ibs(struct evsel *evsel)
 	    || (evsel->pmu_name && !strncmp(evsel->pmu_name, "ibs", 3));
 }
 
+static bool is_amd_brs_event(struct evsel *evsel)
+{
+	return (evsel->core.attr.config & 0xff) == 0xc4;
+}
+
+static bool has_amd_brs(struct evsel *evsel)
+{
+	return (evsel->core.attr.sample_type & PERF_SAMPLE_BRANCH_STACK);
+}
+
 int evsel__open_strerror(struct evsel *evsel, struct target *target,
 			 int err, char *msg, size_t size)
 {
@@ -2947,6 +2957,28 @@ int evsel__open_strerror(struct evsel *evsel, struct target *target,
 	"We found oprofile daemon running, please stop it and try again.");
 		break;
 	case EINVAL:
+		if (is_amd(arch, cpuid)) {
+			if (is_amd_ibs(evsel)) {
+				if (evsel->core.attr.exclude_kernel)
+					return scnprintf(msg, size,
+	"AMD IBS can't exclude kernel events.  Try running at a higher privilege level.");
+				if (!evsel->core.system_wide)
+					return scnprintf(msg, size,
+	"AMD IBS may only be available in system-wide/per-cpu mode.  Try using -a, or -C and workload affinity");
+			}
+			if (has_amd_brs(evsel)) {
+				if (!is_amd_brs_event(evsel))
+					return scnprintf(msg, size,
+	"AMD Branch Sampling can only be used with the cpu/branch-brs,period=xxxx/ event or equivalent encodings.");
+				if (evsel->core.attr.freq)
+					return scnprintf(msg, size,
+	"AMD Branch Sampling does not support frequency mode sampling, must pass a fixed sampling period via -c option or cpu/branch-brs,period=xxxx/.");
+				/* another reason is that the period is too small */
+				return scnprintf(msg, size,
+	"AMD Branch Sampling does not support sampling period smaller than what is reported in /sys/devices/cpu/caps/branches.");
+			}
+		}
+
 		if (evsel->core.attr.sample_type & PERF_SAMPLE_CODE_PAGE_SIZE && perf_missing_features.code_page_size)
 			return scnprintf(msg, size, "Asking for the code page size isn't supported by this kernel.");
 		if (evsel->core.attr.sample_type & PERF_SAMPLE_DATA_PAGE_SIZE && perf_missing_features.data_page_size)
@@ -2963,17 +2995,6 @@ int evsel__open_strerror(struct evsel *evsel, struct target *target,
 			return scnprintf(msg, size,
 	"Invalid event (%s) in per-thread mode, enable system wide with '-a'.",
 					evsel__name(evsel));
-		if (is_amd(arch, cpuid)) {
-			if (is_amd_ibs(evsel)) {
-				if (evsel->core.attr.exclude_kernel)
-					return scnprintf(msg, size,
-	"AMD IBS can't exclude kernel events.  Try running at a higher privilege level.");
-				if (!evsel->core.system_wide)
-					return scnprintf(msg, size,
-	"AMD IBS may only be available in system-wide/per-cpu mode.  Try using -a, or -C and workload affinity");
-			}
-		}
-
 		break;
 	case ENODATA:
 		return scnprintf(msg, size, "Cannot collect data source with the load latency event alone. "
-- 
2.35.1.894.gb6a874cedc-goog

[PATCH v7 13/13] perf report: add addr_from/addr_to sort dimensions

[Stephane Eranian]

With the existing symbol_from/symbol_to, branches captured in the same
function would be collapsed into a single function if the latencies associated
with the each branch (cycles) were all the same.  That is the case on Intel
Broadwell, for instance. Since Intel Skylake, the latency is captured by
hardware and therefore is used to disambiguate branches.

Add addr_from/addr_to sort dimensions to sort branches based on their
addresses and not the function there are in. The output is still the function
name but the offset within the function is provided to uniquely identify each
branch.  These new sort dimensions also help with annotate because they create
different entries in the histogram which, in turn, generates proper branch
annotations.

Here is an example using AMD's branch sampling:

$ perf record -a -b -c 1000037 -e cpu/branch-brs/ test_prg

$ perf report
Samples: 6M of event 'cpu/branch-brs/', Event count (approx.): 6901276
Overhead  Command          Source Shared Object  Source Symbol                                   Target Symbol                                   Basic Block Cycle
  99.65%  test_prg	   test_prg              [.] test_thread                                 [.] test_thread                                 -
   0.02%  test_prg         [kernel.vmlinux]      [k] asm_sysvec_apic_timer_interrupt             [k] error_entry                                 -

$ perf report -F overhead,comm,dso,addr_from,addr_to
Samples: 6M of event 'cpu/branch-brs/', Event count (approx.): 6901276
Overhead  Command          Shared Object     Source Address          Target Address
   4.22%  test_prg         test_prg          [.] test_thread+0x3c    [.] test_thread+0x4
   4.13%  test_prg         test_prg          [.] test_thread+0x4     [.] test_thread+0x3a
   4.09%  test_prg         test_prg          [.] test_thread+0x3a    [.] test_thread+0x6
   4.08%  test_prg         test_prg          [.] test_thread+0x2     [.] test_thread+0x3c
   4.06%  test_prg         test_prg          [.] test_thread+0x3e    [.] test_thread+0x2
   3.87%  test_prg         test_prg          [.] test_thread+0x6     [.] test_thread+0x38
   3.84%  test_prg         test_prg          [.] test_thread         [.] test_thread+0x3e
   3.76%  test_prg         test_prg          [.] test_thread+0x1e    [.] test_thread
   3.76%  test_prg         test_prg          [.] test_thread+0x38    [.] test_thread+0x8
   3.56%  test_prg         test_prg          [.] test_thread+0x22    [.] test_thread+0x1e
   3.54%  test_prg         test_prg          [.] test_thread+0x8     [.] test_thread+0x36
   3.47%  test_prg         test_prg          [.] test_thread+0x1c    [.] test_thread+0x22
   3.45%  test_prg         test_prg          [.] test_thread+0x36    [.] test_thread+0xa
   3.28%  test_prg         test_prg          [.] test_thread+0x24    [.] test_thread+0x1c
   3.25%  test_prg         test_prg          [.] test_thread+0xa     [.] test_thread+0x34
   3.24%  test_prg         test_prg          [.] test_thread+0x1a    [.] test_thread+0x24
   3.20%  test_prg         test_prg          [.] test_thread+0x34    [.] test_thread+0xc
   3.04%  test_prg         test_prg          [.] test_thread+0x26    [.] test_thread+0x1a
   3.01%  test_prg         test_prg          [.] test_thread+0xc     [.] test_thread+0x32
   2.98%  test_prg         test_prg          [.] test_thread+0x18    [.] test_thread+0x26
   2.94%  test_prg         test_prg          [.] test_thread+0x32    [.] test_thread+0xe
   2.76%  test_prg         test_prg          [.] test_thread+0x28    [.] test_thread+0x18
   2.73%  test_prg         test_prg          [.] test_thread+0xe     [.] test_thread+0x30
   2.67%  test_prg         test_prg          [.] test_thread+0x30    [.] test_thread+0x10
   2.67%  test_prg         test_prg          [.] test_thread+0x16    [.] test_thread+0x28
   2.46%  test_prg         test_prg          [.] test_thread+0x10    [.] test_thread+0x2e
   2.44%  test_prg         test_prg          [.] test_thread+0x2a    [.] test_thread+0x16
   2.38%  test_prg         test_prg          [.] test_thread+0x14    [.] test_thread+0x2a
   2.32%  test_prg         test_prg          [.] test_thread+0x2e    [.] test_thread+0x12
   2.28%  test_prg         test_prg          [.] test_thread+0x12    [.] test_thread+0x2c
   2.16%  test_prg         test_prg          [.] test_thread+0x2c    [.] test_thread+0x14
   0.02%  test_prg         [kernel.vmlinux]  [k] asm_sysvec_apic_ti+0x5  [k] error_entry

Signed-off-by: Stephane Eranian <eranian@google.com>
---
 tools/perf/util/hist.c |   2 +
 tools/perf/util/hist.h |   2 +
 tools/perf/util/sort.c | 128 +++++++++++++++++++++++++++++++++++++++++
 tools/perf/util/sort.h |   2 +
 4 files changed, 134 insertions(+)

diff --git a/tools/perf/util/hist.c b/tools/perf/util/hist.c
index 0a8033b09e28..1c085ab56534 100644
--- a/tools/perf/util/hist.c
+++ b/tools/perf/util/hist.c
@@ -124,6 +124,7 @@ void hists__calc_col_len(struct hists *hists, struct hist_entry *h)
 		} else {
 			symlen = unresolved_col_width + 4 + 2;
 			hists__new_col_len(hists, HISTC_SYMBOL_FROM, symlen);
+			hists__new_col_len(hists, HISTC_ADDR_FROM, symlen);
 			hists__set_unres_dso_col_len(hists, HISTC_DSO_FROM);
 		}
 
@@ -138,6 +139,7 @@ void hists__calc_col_len(struct hists *hists, struct hist_entry *h)
 		} else {
 			symlen = unresolved_col_width + 4 + 2;
 			hists__new_col_len(hists, HISTC_SYMBOL_TO, symlen);
+			hists__new_col_len(hists, HISTC_ADDR_TO, symlen);
 			hists__set_unres_dso_col_len(hists, HISTC_DSO_TO);
 		}
 
diff --git a/tools/perf/util/hist.h b/tools/perf/util/hist.h
index 2a15e22fb89c..7ed4648d2fc2 100644
--- a/tools/perf/util/hist.h
+++ b/tools/perf/util/hist.h
@@ -77,6 +77,8 @@ enum hist_column {
 	HISTC_GLOBAL_INS_LAT,
 	HISTC_LOCAL_P_STAGE_CYC,
 	HISTC_GLOBAL_P_STAGE_CYC,
+	HISTC_ADDR_FROM,
+	HISTC_ADDR_TO,
 	HISTC_NR_COLS, /* Last entry */
 };
 
diff --git a/tools/perf/util/sort.c b/tools/perf/util/sort.c
index 2da081ef532b..6d5588e80935 100644
--- a/tools/perf/util/sort.c
+++ b/tools/perf/util/sort.c
@@ -990,6 +990,128 @@ struct sort_entry sort_sym_to = {
 	.se_width_idx	= HISTC_SYMBOL_TO,
 };
 
+static int _hist_entry__addr_snprintf(struct map_symbol *ms,
+				     u64 ip, char level, char *bf, size_t size,
+				     unsigned int width)
+{
+	struct symbol *sym = ms->sym;
+	struct map *map = ms->map;
+	size_t ret = 0, offs;
+
+	ret += repsep_snprintf(bf + ret, size - ret, "[%c] ", level);
+	if (sym && map) {
+		if (sym->type == STT_OBJECT) {
+			ret += repsep_snprintf(bf + ret, size - ret, "%s", sym->name);
+			ret += repsep_snprintf(bf + ret, size - ret, "+0x%llx",
+					ip - map->unmap_ip(map, sym->start));
+		} else {
+			ret += repsep_snprintf(bf + ret, size - ret, "%.*s",
+					       width - ret,
+					       sym->name);
+			offs = ip - sym->start;
+			if (offs)
+				ret += repsep_snprintf(bf + ret, size - ret, "+0x%llx", offs);
+		}
+	} else {
+		size_t len = BITS_PER_LONG / 4;
+		ret += repsep_snprintf(bf + ret, size - ret, "%-#.*llx",
+				       len, ip);
+	}
+
+	return ret;
+}
+
+static int hist_entry__addr_from_snprintf(struct hist_entry *he, char *bf,
+					 size_t size, unsigned int width)
+{
+	if (he->branch_info) {
+		struct addr_map_symbol *from = &he->branch_info->from;
+
+		return _hist_entry__addr_snprintf(&from->ms, from->al_addr,
+						 he->level, bf, size, width);
+	}
+
+	return repsep_snprintf(bf, size, "%-*.*s", width, width, "N/A");
+}
+
+static int hist_entry__addr_to_snprintf(struct hist_entry *he, char *bf,
+				       size_t size, unsigned int width)
+{
+	if (he->branch_info) {
+		struct addr_map_symbol *to = &he->branch_info->to;
+
+		return _hist_entry__addr_snprintf(&to->ms, to->al_addr,
+						 he->level, bf, size, width);
+	}
+
+	return repsep_snprintf(bf, size, "%-*.*s", width, width, "N/A");
+}
+
+static int64_t
+sort__addr_from_cmp(struct hist_entry *left, struct hist_entry *right)
+{
+	struct addr_map_symbol *from_l;
+	struct addr_map_symbol *from_r;
+	int64_t ret;
+
+	if (!left->branch_info || !right->branch_info)
+		return cmp_null(left->branch_info, right->branch_info);
+
+	from_l = &left->branch_info->from;
+	from_r = &right->branch_info->from;
+
+	/*
+	 * comparing symbol address alone is not enough since it's a
+	 * relative address within a dso.
+	 */
+	ret = _sort__dso_cmp(from_l->ms.map, from_r->ms.map);
+	if (ret != 0)
+		return ret;
+
+	return _sort__addr_cmp(from_l->addr, from_r->addr);
+}
+
+static int64_t
+sort__addr_to_cmp(struct hist_entry *left, struct hist_entry *right)
+{
+	struct addr_map_symbol *to_l;
+	struct addr_map_symbol *to_r;
+	int64_t ret;
+
+	if (!left->branch_info || !right->branch_info)
+		return cmp_null(left->branch_info, right->branch_info);
+
+	to_l = &left->branch_info->to;
+	to_r = &right->branch_info->to;
+
+	/*
+	 * comparing symbol address alone is not enough since it's a
+	 * relative address within a dso.
+	 */
+	ret = _sort__dso_cmp(to_l->ms.map, to_r->ms.map);
+	if (ret != 0)
+		return ret;
+
+	return _sort__addr_cmp(to_l->addr, to_r->addr);
+}
+
+struct sort_entry sort_addr_from = {
+	.se_header	= "Source Address",
+	.se_cmp		= sort__addr_from_cmp,
+	.se_snprintf	= hist_entry__addr_from_snprintf,
+	.se_filter	= hist_entry__sym_from_filter, /* shared with sym_from */
+	.se_width_idx	= HISTC_ADDR_FROM,
+};
+
+struct sort_entry sort_addr_to = {
+	.se_header	= "Target Address",
+	.se_cmp		= sort__addr_to_cmp,
+	.se_snprintf	= hist_entry__addr_to_snprintf,
+	.se_filter	= hist_entry__sym_to_filter, /* shared with sym_to */
+	.se_width_idx	= HISTC_ADDR_TO,
+};
+
+
 static int64_t
 sort__mispredict_cmp(struct hist_entry *left, struct hist_entry *right)
 {
@@ -1893,6 +2015,8 @@ static struct sort_dimension bstack_sort_dimensions[] = {
 	DIM(SORT_SRCLINE_FROM, "srcline_from", sort_srcline_from),
 	DIM(SORT_SRCLINE_TO, "srcline_to", sort_srcline_to),
 	DIM(SORT_SYM_IPC, "ipc_lbr", sort_sym_ipc),
+	DIM(SORT_ADDR_FROM, "addr_from", sort_addr_from),
+	DIM(SORT_ADDR_TO, "addr_to", sort_addr_to),
 };
 
 #undef DIM
@@ -3126,6 +3250,10 @@ static bool get_elide(int idx, FILE *output)
 		return __get_elide(symbol_conf.dso_from_list, "dso_from", output);
 	case HISTC_DSO_TO:
 		return __get_elide(symbol_conf.dso_to_list, "dso_to", output);
+	case HISTC_ADDR_FROM:
+		return __get_elide(symbol_conf.sym_from_list, "addr_from", output);
+	case HISTC_ADDR_TO:
+		return __get_elide(symbol_conf.sym_to_list, "addr_to", output);
 	default:
 		break;
 	}
diff --git a/tools/perf/util/sort.h b/tools/perf/util/sort.h
index f994261888e1..2ddc00d1c464 100644
--- a/tools/perf/util/sort.h
+++ b/tools/perf/util/sort.h
@@ -251,6 +251,8 @@ enum sort_type {
 	SORT_SRCLINE_FROM,
 	SORT_SRCLINE_TO,
 	SORT_SYM_IPC,
+	SORT_ADDR_FROM,
+	SORT_ADDR_TO,
 
 	/* memory mode specific sort keys */
 	__SORT_MEMORY_MODE,
-- 
2.35.1.894.gb6a874cedc-goog

Re: [PATCH v7 10/13] perf tools: fix NULL point in evsel__env()

[Kim Phillips]

On 3/22/22 5:15 PM, Stephane Eranian wrote:
> evsel_env() cannot return NULL. In case evsel->evlist->env is NULL default
> to perf_env. Fixes crashes in evsel__open_strerror()
> 
> Signed-off-by: Ravi Bangoria <ravi.bangoria@amd.com>
> Reviewed-by: Stephane Eranian <eranian@google.com>
> ---

Signed-off-by: Kim Phillips <kim.phillips@amd.com>

FWIW, here's a link to one of the original submissions:

https://lkml.org/lkml/2022/1/26/1220

Thanks,

Kim

Re: [PATCH v7 00/13] perf/x86/amd: Add AMD Fam19h Branch Sampling support

[Peter Zijlstra]

On Tue, Mar 22, 2022 at 03:15:04PM -0700, Stephane Eranian wrote:
> V7 makes the following changes:
>    - rebased to 5.17-rc8
>    - Restrict BRS to be used only with RETIRED_TAKEN_BRANCH event in order to
>      avoid adding more skid to other events in case that are monitored
>      concurrently
>    - fixed bugs in the perf tool enhanced error handling for IBS and BRS

But afaict you failed to pick up the build fixes I did in my tree :/ Let
me try and be careful preserving those while munging this in.

Re: [PATCH v7 00/13] perf/x86/amd: Add AMD Fam19h Branch Sampling support

[Stephane Eranian]

On Thu, Mar 24, 2022 at 7:59 AM Peter Zijlstra <peterz@infradead.org> wrote:
>
> On Tue, Mar 22, 2022 at 03:15:04PM -0700, Stephane Eranian wrote:
> > V7 makes the following changes:
> >    - rebased to 5.17-rc8
> >    - Restrict BRS to be used only with RETIRED_TAKEN_BRANCH event in order to
> >      avoid adding more skid to other events in case that are monitored
> >      concurrently
> >    - fixed bugs in the perf tool enhanced error handling for IBS and BRS
>
> But afaict you failed to pick up the build fixes I did in my tree :/ Let
> me try and be careful preserving those while munging this in.

Sorry. I missed that. I can resubmit if you prefer.

Re: [PATCH v7 00/13] perf/x86/amd: Add AMD Fam19h Branch Sampling support

[Peter Zijlstra]

On Thu, Mar 24, 2022 at 10:31:38PM -0700, Stephane Eranian wrote:
> On Thu, Mar 24, 2022 at 7:59 AM Peter Zijlstra <peterz@infradead.org> wrote:
> >
> > On Tue, Mar 22, 2022 at 03:15:04PM -0700, Stephane Eranian wrote:
> > > V7 makes the following changes:
> > >    - rebased to 5.17-rc8
> > >    - Restrict BRS to be used only with RETIRED_TAKEN_BRANCH event in order to
> > >      avoid adding more skid to other events in case that are monitored
> > >      concurrently
> > >    - fixed bugs in the perf tool enhanced error handling for IBS and BRS
> >
> > But afaict you failed to pick up the build fixes I did in my tree :/ Let
> > me try and be careful preserving those while munging this in.
> 
> Sorry. I missed that. I can resubmit if you prefer.

Nah, fixed it up again.. Even converted to static_call() this time.

Could you please test this branch?

  https://git.kernel.org/pub/scm/linux/kernel/git/peterz/queue.git/log/?h=perf/amd

I've since done more Changelog edits, and I should do a comment with the
static_call_mod() usage, but I think this ought to work.

At least it now actually builds...

Re: [PATCH v7 10/13] perf tools: fix NULL point in evsel__env()

[Arnaldo Carvalho de Melo]

Em Tue, Mar 22, 2022 at 03:15:14PM -0700, Stephane Eranian escreveu:
> evsel_env() cannot return NULL. In case evsel->evlist->env is NULL default
> to perf_env. Fixes crashes in evsel__open_strerror()
> 
> Signed-off-by: Ravi Bangoria <ravi.bangoria@amd.com>
> Reviewed-by: Stephane Eranian <eranian@google.com>

I have this already in from a patch by Kim Phillips.

- Arnaldo

> ---
>  tools/perf/util/evsel.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
> index 22d3267ce294..14b0e7ffa2c7 100644
> --- a/tools/perf/util/evsel.c
> +++ b/tools/perf/util/evsel.c
> @@ -2965,7 +2965,7 @@ int evsel__open_strerror(struct evsel *evsel, struct target *target,
>  
>  struct perf_env *evsel__env(struct evsel *evsel)
>  {
> -	if (evsel && evsel->evlist)
> +	if (evsel && evsel->evlist && evsel->evlist->env)
>  		return evsel->evlist->env;
>  	return &perf_env;
>  }
> -- 
> 2.35.1.894.gb6a874cedc-goog

-- 

- Arnaldo

Re: [PATCH v7 10/13] perf tools: fix NULL point in evsel__env()

[Stephane Eranian]

On Fri, Mar 25, 2022 at 11:27 AM Arnaldo Carvalho de Melo
<acme@kernel.org> wrote:
>
> Em Tue, Mar 22, 2022 at 03:15:14PM -0700, Stephane Eranian escreveu:
> > evsel_env() cannot return NULL. In case evsel->evlist->env is NULL default
> > to perf_env. Fixes crashes in evsel__open_strerror()
> >
> > Signed-off-by: Ravi Bangoria <ravi.bangoria@amd.com>
> > Reviewed-by: Stephane Eranian <eranian@google.com>
>
> I have this already in from a patch by Kim Phillips.
>
Ok, then. We are all set.

> - Arnaldo
>
> > ---
> >  tools/perf/util/evsel.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
> > index 22d3267ce294..14b0e7ffa2c7 100644
> > --- a/tools/perf/util/evsel.c
> > +++ b/tools/perf/util/evsel.c
> > @@ -2965,7 +2965,7 @@ int evsel__open_strerror(struct evsel *evsel, struct target *target,
> >
> >  struct perf_env *evsel__env(struct evsel *evsel)
> >  {
> > -     if (evsel && evsel->evlist)
> > +     if (evsel && evsel->evlist && evsel->evlist->env)
> >               return evsel->evlist->env;
> >       return &perf_env;
> >  }
> > --
> > 2.35.1.894.gb6a874cedc-goog
>
> --
>
> - Arnaldo

Re: [PATCH v7 11/13] perf tools: Improve IBS error handling

[Arnaldo Carvalho de Melo]

Em Tue, Mar 22, 2022 at 03:15:15PM -0700, Stephane Eranian escreveu:
> From: Kim Phillips <kim.phillips@amd.com>
> 
> improve the error message returned on failed perf_event_open() on AMD when
> using IBS.
> 
> Output of executing 'perf record -e ibs_op// true' BEFORE this patch:
> 
> The sys_perf_event_open() syscall returned with 22 (Invalid argument)for event (ibs_op//u).
> /bin/dmesg | grep -i perf may provide additional information.
> 
> Output after:
> 
> AMD IBS cannot exclude kernel events.  Try running at a higher privilege level.

So the error message don't match up the BEFORE part, that didn't have a
"u" at the end, and with this patch in:

  $ git log --oneline -1
  ca585f91ce13df4c (HEAD -> perf/core) perf evsel: Improve AMD IBS (Instruction-Based Sampling) error handling messages
  # perf -v
  perf version 5.17.gca585f91ce13
  #
  
On a:

  # grep -m1 "model name" /proc/cpuinfo
  model name	: AMD Ryzen 9 5950X 16-Core Processor
  #

If I try:

  # perf record -e ibs_op//u true
  Error:
  Invalid event (ibs_op//u) in per-thread mode, enable system wide with '-a'.
  # 

So now if I try:

  # perf record -a -e ibs_op//u true
  Error:
  AMD IBS can't exclude kernel events.  Try running at a higher privilege level.
  #

So the problem is with the patch description, I'm fixing it.

> Output of executing 'sudo perf record -e ibs_op// true' BEFORE this patch:
> 
> Error:
> The sys_perf_event_open() syscall returned with 22 (Invalid argument) for event (ibs_op//).
> /bin/dmesg | grep -i perf may provide additional information.
> 
> Output after:
> 
> Error:
> AMD IBS may only be available in system-wide/per-cpu mode.  Try using -a, or -C and workload affinity

But this one is never reached:

[root@five ~]# perf record -e ibs_op// true
Error:
Invalid event (ibs_op//) in per-thread mode, enable system wide with '-a'

> +                             if (!evsel->core.system_wide)
> +                                     return scnprintf(msg, size,
> +     "AMD IBS may only be available in system-wide/per-cpu mode.  Try using -a, or -C and workload affinity");


I'm applying this to make progress, the message now provides more clues.

- Arnaldo

 
> Signed-off-by: Kim Phillips <kim.phillips@amd.com>
> Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
> Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com>
> Cc: Ian Rogers <irogers@google.com>
> Cc: Ingo Molnar <mingo@redhat.com>
> Cc: Jiri Olsa <jolsa@redhat.com>
> Cc: Joao Martins <joao.m.martins@oracle.com>
> Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
> Cc: Mark Rutland <mark.rutland@arm.com>
> Cc: Michael Petlan <mpetlan@redhat.com>
> Cc: Namhyung Kim <namhyung@kernel.org>
> Cc: Peter Zijlstra <peterz@infradead.org>
> Cc: Robert Richter <robert.richter@amd.com>
> Cc: Stephane Eranian <eranian@google.com>
> ---
>  tools/perf/util/evsel.c | 25 +++++++++++++++++++++++++
>  1 file changed, 25 insertions(+)
> 
> diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
> index 14b0e7ffa2c7..e8ff7a4bd490 100644
> --- a/tools/perf/util/evsel.c
> +++ b/tools/perf/util/evsel.c
> @@ -2847,9 +2847,23 @@ static bool find_process(const char *name)
>  	return ret ? false : true;
>  }
>  
> +static bool is_amd(const char *arch, const char *cpuid)
> +{
> +	return arch && !strcmp("x86", arch) && cpuid && strstarts(cpuid, "AuthenticAMD");
> +}
> +
> +static bool is_amd_ibs(struct evsel *evsel)
> +{
> +	return evsel->core.attr.precise_ip
> +	    || (evsel->pmu_name && !strncmp(evsel->pmu_name, "ibs", 3));
> +}
> +
>  int evsel__open_strerror(struct evsel *evsel, struct target *target,
>  			 int err, char *msg, size_t size)
>  {
> +	struct perf_env *env = evsel__env(evsel);
> +	const char *arch = perf_env__arch(env);
> +	const char *cpuid = perf_env__cpuid(env);
>  	char sbuf[STRERR_BUFSIZE];
>  	int printed = 0, enforced = 0;
>  
> @@ -2949,6 +2963,17 @@ int evsel__open_strerror(struct evsel *evsel, struct target *target,
>  			return scnprintf(msg, size,
>  	"Invalid event (%s) in per-thread mode, enable system wide with '-a'.",
>  					evsel__name(evsel));
> +		if (is_amd(arch, cpuid)) {
> +			if (is_amd_ibs(evsel)) {
> +				if (evsel->core.attr.exclude_kernel)
> +					return scnprintf(msg, size,
> +	"AMD IBS can't exclude kernel events.  Try running at a higher privilege level.");
> +				if (!evsel->core.system_wide)
> +					return scnprintf(msg, size,
> +	"AMD IBS may only be available in system-wide/per-cpu mode.  Try using -a, or -C and workload affinity");
> +			}
> +		}
> +
>  		break;
>  	case ENODATA:
>  		return scnprintf(msg, size, "Cannot collect data source with the load latency event alone. "
> -- 
> 2.35.1.894.gb6a874cedc-goog

-- 

- Arnaldo

Re: [PATCH v7 11/13] perf tools: Improve IBS error handling

[Arnaldo Carvalho de Melo]

Em Fri, Mar 25, 2022 at 06:46:08PM -0300, Arnaldo Carvalho de Melo escreveu:
> Em Tue, Mar 22, 2022 at 03:15:15PM -0700, Stephane Eranian escreveu:
> > From: Kim Phillips <kim.phillips@amd.com>
> > 
> > improve the error message returned on failed perf_event_open() on AMD when
> > using IBS.
> > 
> > Output of executing 'perf record -e ibs_op// true' BEFORE this patch:
> > 
> > The sys_perf_event_open() syscall returned with 22 (Invalid argument)for event (ibs_op//u).
> > /bin/dmesg | grep -i perf may provide additional information.
> > 
> > Output after:
> > 
> > AMD IBS cannot exclude kernel events.  Try running at a higher privilege level.
> 
> So the error message don't match up the BEFORE part, that didn't have a
> "u" at the end, and with this patch in:

I see, the patch description doesn't mention that the BEFORE command is
run as a NON root user, when the 'u' suffix is auto-added by perf, I'm
ammending the comment log message with this.

Now it reads:

----------------------------------------------------------------------
perf evsel: Improve AMD IBS (Instruction-Based Sampling) error handling messages

Improve the error message returned on failed perf_event_open() on AMD
systems when using IBS (Instruction-Based Sampling).

Output of executing 'perf record -e ibs_op// true' as a non root user
BEFORE this patch (perf will add the 'u' modifier at the end to exclude
kernel/hypervisor sampling):

  The sys_perf_event_open() syscall returned with 22 (Invalid argument)for event (ibs_op//u).
  /bin/dmesg | grep -i perf may provide additional information.

Output after:

  AMD IBS can't exclude kernel events.  Try running at a higher privilege level.

Output of executing 'sudo perf record -e ibs_op// true' BEFORE this patch:

  Error:
  The sys_perf_event_open() syscall returned with 22 (Invalid argument) for event (ibs_op//).
  /bin/dmesg | grep -i perf may provide additional information.

Output after:

  Error:
  Invalid event (ibs_op//) in per-thread mode, enable system wide with '-a'.

Folowing the suggestion:

  $ sudo perf record -a -e ibs_op// true
  [ perf record: Woken up 1 times to write data ]
  [ perf record: Captured and wrote 1.664 MB perf.data (194 samples) ]
  $

Signed-off-by: Kim Phillips <kim.phillips@amd.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
----------------------------------------------------------------------

This is all with:

[root@five ~]# uname -a
Linux five 5.16.16-200.fc35.x86_64 #1 SMP PREEMPT Sat Mar 19 13:52:41 UTC 2022 x86_64 x86_64 x86_64 GNU/Linux

- Arnaldo
 
>   $ git log --oneline -1
>   ca585f91ce13df4c (HEAD -> perf/core) perf evsel: Improve AMD IBS (Instruction-Based Sampling) error handling messages
>   # perf -v
>   perf version 5.17.gca585f91ce13
>   #
>   
> On a:
> 
>   # grep -m1 "model name" /proc/cpuinfo
>   model name	: AMD Ryzen 9 5950X 16-Core Processor
>   #
> 
> If I try:
> 
>   # perf record -e ibs_op//u true
>   Error:
>   Invalid event (ibs_op//u) in per-thread mode, enable system wide with '-a'.
>   # 
> 
> So now if I try:
> 
>   # perf record -a -e ibs_op//u true
>   Error:
>   AMD IBS can't exclude kernel events.  Try running at a higher privilege level.
>   #
> 
> So the problem is with the patch description, I'm fixing it.
> 
> > Output of executing 'sudo perf record -e ibs_op// true' BEFORE this patch:
> > 
> > Error:
> > The sys_perf_event_open() syscall returned with 22 (Invalid argument) for event (ibs_op//).
> > /bin/dmesg | grep -i perf may provide additional information.
> > 
> > Output after:
> > 
> > Error:
> > AMD IBS may only be available in system-wide/per-cpu mode.  Try using -a, or -C and workload affinity
> 
> But this one is never reached:
> 
> [root@five ~]# perf record -e ibs_op// true
> Error:
> Invalid event (ibs_op//) in per-thread mode, enable system wide with '-a'
> 
> > +                             if (!evsel->core.system_wide)
> > +                                     return scnprintf(msg, size,
> > +     "AMD IBS may only be available in system-wide/per-cpu mode.  Try using -a, or -C and workload affinity");
> 
> 
> I'm applying this to make progress, the message now provides more clues.
> 
> - Arnaldo
> 
>  
> > Signed-off-by: Kim Phillips <kim.phillips@amd.com>
> > Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
> > Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
> > Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com>
> > Cc: Ian Rogers <irogers@google.com>
> > Cc: Ingo Molnar <mingo@redhat.com>
> > Cc: Jiri Olsa <jolsa@redhat.com>
> > Cc: Joao Martins <joao.m.martins@oracle.com>
> > Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
> > Cc: Mark Rutland <mark.rutland@arm.com>
> > Cc: Michael Petlan <mpetlan@redhat.com>
> > Cc: Namhyung Kim <namhyung@kernel.org>
> > Cc: Peter Zijlstra <peterz@infradead.org>
> > Cc: Robert Richter <robert.richter@amd.com>
> > Cc: Stephane Eranian <eranian@google.com>
> > ---
> >  tools/perf/util/evsel.c | 25 +++++++++++++++++++++++++
> >  1 file changed, 25 insertions(+)
> > 
> > diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
> > index 14b0e7ffa2c7..e8ff7a4bd490 100644
> > --- a/tools/perf/util/evsel.c
> > +++ b/tools/perf/util/evsel.c
> > @@ -2847,9 +2847,23 @@ static bool find_process(const char *name)
> >  	return ret ? false : true;
> >  }
> >  
> > +static bool is_amd(const char *arch, const char *cpuid)
> > +{
> > +	return arch && !strcmp("x86", arch) && cpuid && strstarts(cpuid, "AuthenticAMD");
> > +}
> > +
> > +static bool is_amd_ibs(struct evsel *evsel)
> > +{
> > +	return evsel->core.attr.precise_ip
> > +	    || (evsel->pmu_name && !strncmp(evsel->pmu_name, "ibs", 3));
> > +}
> > +
> >  int evsel__open_strerror(struct evsel *evsel, struct target *target,
> >  			 int err, char *msg, size_t size)
> >  {
> > +	struct perf_env *env = evsel__env(evsel);
> > +	const char *arch = perf_env__arch(env);
> > +	const char *cpuid = perf_env__cpuid(env);
> >  	char sbuf[STRERR_BUFSIZE];
> >  	int printed = 0, enforced = 0;
> >  
> > @@ -2949,6 +2963,17 @@ int evsel__open_strerror(struct evsel *evsel, struct target *target,
> >  			return scnprintf(msg, size,
> >  	"Invalid event (%s) in per-thread mode, enable system wide with '-a'.",
> >  					evsel__name(evsel));
> > +		if (is_amd(arch, cpuid)) {
> > +			if (is_amd_ibs(evsel)) {
> > +				if (evsel->core.attr.exclude_kernel)
> > +					return scnprintf(msg, size,
> > +	"AMD IBS can't exclude kernel events.  Try running at a higher privilege level.");
> > +				if (!evsel->core.system_wide)
> > +					return scnprintf(msg, size,
> > +	"AMD IBS may only be available in system-wide/per-cpu mode.  Try using -a, or -C and workload affinity");
> > +			}
> > +		}
> > +
> >  		break;
> >  	case ENODATA:
> >  		return scnprintf(msg, size, "Cannot collect data source with the load latency event alone. "
> > -- 
> > 2.35.1.894.gb6a874cedc-goog
> 
> -- 
> 
> - Arnaldo

-- 

- Arnaldo

Re: [PATCH v7 00/13] perf/x86/amd: Add AMD Fam19h Branch Sampling support

[Stephane Eranian]

On Fri, Mar 25, 2022 at 1:29 AM Peter Zijlstra <peterz@infradead.org> wrote:
>
> On Thu, Mar 24, 2022 at 10:31:38PM -0700, Stephane Eranian wrote:
> > On Thu, Mar 24, 2022 at 7:59 AM Peter Zijlstra <peterz@infradead.org> wrote:
> > >
> > > On Tue, Mar 22, 2022 at 03:15:04PM -0700, Stephane Eranian wrote:
> > > > V7 makes the following changes:
> > > >    - rebased to 5.17-rc8
> > > >    - Restrict BRS to be used only with RETIRED_TAKEN_BRANCH event in order to
> > > >      avoid adding more skid to other events in case that are monitored
> > > >      concurrently
> > > >    - fixed bugs in the perf tool enhanced error handling for IBS and BRS
> > >
> > > But afaict you failed to pick up the build fixes I did in my tree :/ Let
> > > me try and be careful preserving those while munging this in.
> >
> > Sorry. I missed that. I can resubmit if you prefer.
>
> Nah, fixed it up again.. Even converted to static_call() this time.
>
> Could you please test this branch?

Started testing. BRS does not come up so far. Let me debug this.

>
>   https://git.kernel.org/pub/scm/linux/kernel/git/peterz/queue.git/log/?h=perf/amd
>
> I've since done more Changelog edits, and I should do a comment with the
> static_call_mod() usage, but I think this ought to work.
>
> At least it now actually builds...

Re: [PATCH v7 00/13] perf/x86/amd: Add AMD Fam19h Branch Sampling support

[Stephane Eranian]

On Tue, Mar 29, 2022 at 2:58 PM Stephane Eranian <eranian@google.com> wrote:
>
> On Fri, Mar 25, 2022 at 1:29 AM Peter Zijlstra <peterz@infradead.org> wrote:
> >
> > On Thu, Mar 24, 2022 at 10:31:38PM -0700, Stephane Eranian wrote:
> > > On Thu, Mar 24, 2022 at 7:59 AM Peter Zijlstra <peterz@infradead.org> wrote:
> > > >
> > > > On Tue, Mar 22, 2022 at 03:15:04PM -0700, Stephane Eranian wrote:
> > > > > V7 makes the following changes:
> > > > >    - rebased to 5.17-rc8
> > > > >    - Restrict BRS to be used only with RETIRED_TAKEN_BRANCH event in order to
> > > > >      avoid adding more skid to other events in case that are monitored
> > > > >      concurrently
> > > > >    - fixed bugs in the perf tool enhanced error handling for IBS and BRS
> > > >
> > > > But afaict you failed to pick up the build fixes I did in my tree :/ Let
> > > > me try and be careful preserving those while munging this in.
> > >
> > > Sorry. I missed that. I can resubmit if you prefer.
> >
> > Nah, fixed it up again.. Even converted to static_call() this time.
> >
> > Could you please test this branch?
>
> Started testing. BRS does not come up so far. Let me debug this.
>
False alarm. I had not forced the config option. With the config
option enabled, it works as expected.
I just found a minor issue in case BRS is not enabled. To be
consistent, we should do:

--- a/arch/x86/events/perf_event.h
+++ b/arch/x86/events/perf_event.h
@@ -1265,7 +1265,7 @@ static inline s64 amd_brs_adjust_period(s64 period)
 #else
 static inline int amd_brs_init(void)
 {
-       return 0;
+       return -EOPNOTSUPP;
 }

Thanks.

> >
> >   https://git.kernel.org/pub/scm/linux/kernel/git/peterz/queue.git/log/?h=perf/amd
> >
> > I've since done more Changelog edits, and I should do a comment with the
> > static_call_mod() usage, but I think this ought to work.
> >
> > At least it now actually builds...

Re: [PATCH v7 00/13] perf/x86/amd: Add AMD Fam19h Branch Sampling support

[Peter Zijlstra]

On Tue, Mar 29, 2022 at 04:47:52PM -0700, Stephane Eranian wrote:
> False alarm. I had not forced the config option. With the config
> option enabled, it works as expected.

Great!

> I just found a minor issue in case BRS is not enabled. To be
> consistent, we should do:
> 
> --- a/arch/x86/events/perf_event.h
> +++ b/arch/x86/events/perf_event.h
> @@ -1265,7 +1265,7 @@ static inline s64 amd_brs_adjust_period(s64 period)
>  #else
>  static inline int amd_brs_init(void)
>  {
> -       return 0;
> +       return -EOPNOTSUPP;
>  }

Done!

[tip: perf/core] perf/x86/amd: Add idle hooks for branch sampling

[tip-bot2 for Stephane Eranian]

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

Commit-ID:     d5616bac7adadbf42a3b63b8717e75eb82a2cc2c
Gitweb:        https://git.kernel.org/tip/d5616bac7adadbf42a3b63b8717e75eb82a2cc2c
Author:        Stephane Eranian <eranian@google.com>
AuthorDate:    Tue, 22 Mar 2022 15:15:13 -07:00
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Tue, 05 Apr 2022 10:24:38 +02:00

perf/x86/amd: Add idle hooks for branch sampling

On AMD Fam19h Zen3, the branch sampling (BRS) feature must be disabled before
entering low power and re-enabled (if was active) when returning from low
power. Otherwise, the NMI interrupt may be held up for too long and cause
problems. Stopping BRS will cause the NMI to be delivered if it was held up.

Define a perf_amd_brs_lopwr_cb() callback to stop/restart BRS.  The callback
is protected by a jump label which is enabled only when AMD BRS is detected.
In all other cases, the callback is never called.

Signed-off-by: Stephane Eranian <eranian@google.com>
[peterz: static_call() and build fixes]
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20220322221517.2510440-10-eranian@google.com
---
 arch/x86/events/amd/brs.c         | 33 ++++++++++++++++++++++++++++++-
 arch/x86/events/amd/core.c        |  4 ++++-
 arch/x86/events/perf_event.h      |  1 +-
 arch/x86/include/asm/perf_event.h | 23 +++++++++++++++++++++-
 4 files changed, 61 insertions(+)

diff --git a/arch/x86/events/amd/brs.c b/arch/x86/events/amd/brs.c
index 40461c3..895c821 100644
--- a/arch/x86/events/amd/brs.c
+++ b/arch/x86/events/amd/brs.c
@@ -7,6 +7,7 @@
  * Contributed by Stephane Eranian <eranian@google.com>
  */
 #include <linux/kernel.h>
+#include <linux/jump_label.h>
 #include <asm/msr.h>
 #include <asm/cpufeature.h>
 
@@ -329,3 +330,35 @@ void amd_pmu_brs_sched_task(struct perf_event_context *ctx, bool sched_in)
 	if (sched_in)
 		amd_brs_poison_buffer();
 }
+
+/*
+ * called from ACPI processor_idle.c or acpi_pad.c
+ * with interrupts disabled
+ */
+void perf_amd_brs_lopwr_cb(bool lopwr_in)
+{
+	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
+	union amd_debug_extn_cfg cfg;
+
+	/*
+	 * on mwait in, we may end up in non C0 state.
+	 * we must disable branch sampling to avoid holding the NMI
+	 * for too long. We disable it in hardware but we
+	 * keep the state in cpuc, so we can re-enable.
+	 *
+	 * The hardware will deliver the NMI if needed when brsmen cleared
+	 */
+	if (cpuc->brs_active) {
+		cfg.val = get_debug_extn_cfg();
+		cfg.brsmen = !lopwr_in;
+		set_debug_extn_cfg(cfg.val);
+	}
+}
+
+DEFINE_STATIC_CALL_NULL(perf_lopwr_cb, perf_amd_brs_lopwr_cb);
+EXPORT_STATIC_CALL_TRAMP_GPL(perf_lopwr_cb);
+
+void __init amd_brs_lopwr_init(void)
+{
+	static_call_update(perf_lopwr_cb, perf_amd_brs_lopwr_cb);
+}
diff --git a/arch/x86/events/amd/core.c b/arch/x86/events/amd/core.c
index f7bce83..8e1e818 100644
--- a/arch/x86/events/amd/core.c
+++ b/arch/x86/events/amd/core.c
@@ -1,5 +1,6 @@
 // SPDX-License-Identifier: GPL-2.0-only
 #include <linux/perf_event.h>
+#include <linux/jump_label.h>
 #include <linux/export.h>
 #include <linux/types.h>
 #include <linux/init.h>
@@ -1225,6 +1226,9 @@ static int __init amd_core_pmu_init(void)
 		/*
 		 * put_event_constraints callback same as Fam17h, set above
 		 */
+
+		/* branch sampling must be stopped when entering low power */
+		amd_brs_lopwr_init();
 	}
 
 	x86_pmu.attr_update = amd_attr_update;
diff --git a/arch/x86/events/perf_event.h b/arch/x86/events/perf_event.h
index ef27aee..3b03245 100644
--- a/arch/x86/events/perf_event.h
+++ b/arch/x86/events/perf_event.h
@@ -1226,6 +1226,7 @@ void amd_brs_enable(void);
 void amd_brs_enable_all(void);
 void amd_brs_disable_all(void);
 void amd_brs_drain(void);
+void amd_brs_lopwr_init(void);
 void amd_brs_disable_all(void);
 int amd_brs_setup_filter(struct perf_event *event);
 void amd_brs_reset(void);
diff --git a/arch/x86/include/asm/perf_event.h b/arch/x86/include/asm/perf_event.h
index 58d9e4b..8199fc5 100644
--- a/arch/x86/include/asm/perf_event.h
+++ b/arch/x86/include/asm/perf_event.h
@@ -2,6 +2,8 @@
 #ifndef _ASM_X86_PERF_EVENT_H
 #define _ASM_X86_PERF_EVENT_H
 
+#include <linux/static_call.h>
+
 /*
  * Performance event hw details:
  */
@@ -513,6 +515,27 @@ static inline void intel_pt_handle_vmx(int on)
 #if defined(CONFIG_PERF_EVENTS) && defined(CONFIG_CPU_SUP_AMD)
  extern void amd_pmu_enable_virt(void);
  extern void amd_pmu_disable_virt(void);
+
+#if defined(CONFIG_PERF_EVENTS_AMD_BRS)
+
+#define PERF_NEEDS_LOPWR_CB 1
+
+/*
+ * architectural low power callback impacts
+ * drivers/acpi/processor_idle.c
+ * drivers/acpi/acpi_pad.c
+ */
+extern void perf_amd_brs_lopwr_cb(bool lopwr_in);
+
+DECLARE_STATIC_CALL(perf_lopwr_cb, perf_amd_brs_lopwr_cb);
+
+static inline void perf_lopwr_cb(bool lopwr_in)
+{
+	static_call_mod(perf_lopwr_cb)(lopwr_in);
+}
+
+#endif /* PERF_NEEDS_LOPWR_CB */
+
 #else
  static inline void amd_pmu_enable_virt(void) { }
  static inline void amd_pmu_disable_virt(void) { }

[tip: perf/core] ACPI: Add perf low power callback

[tip-bot2 for Stephane Eranian]

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

Commit-ID:     2a606a18cd672a16343d146a126721b34cc6adbd
Gitweb:        https://git.kernel.org/tip/2a606a18cd672a16343d146a126721b34cc6adbd
Author:        Stephane Eranian <eranian@google.com>
AuthorDate:    Tue, 22 Mar 2022 15:15:12 -07:00
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Tue, 05 Apr 2022 10:24:38 +02:00

ACPI: Add perf low power callback

Add an optional callback needed by some PMU features, e.g., AMD
BRS, to give a chance to the perf_events code to change its state before
a CPU goes to low power and after it comes back.

The callback is void when the PERF_NEEDS_LOPWR_CB flag is not set.
This flag must be set in arch specific perf_event.h header whenever needed.
When not set, there is no impact on the ACPI code.

Signed-off-by: Stephane Eranian <eranian@google.com>
[peterz: build fix]
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20220322221517.2510440-9-eranian@google.com
---
 drivers/acpi/acpi_pad.c       | 7 +++++++
 drivers/acpi/processor_idle.c | 5 +++++
 include/linux/perf_event.h    | 6 ++++++
 3 files changed, 18 insertions(+)

diff --git a/drivers/acpi/acpi_pad.c b/drivers/acpi/acpi_pad.c
index f45979a..ec0e22a 100644
--- a/drivers/acpi/acpi_pad.c
+++ b/drivers/acpi/acpi_pad.c
@@ -17,6 +17,7 @@
 #include <linux/tick.h>
 #include <linux/slab.h>
 #include <linux/acpi.h>
+#include <linux/perf_event.h>
 #include <asm/mwait.h>
 #include <xen/xen.h>
 
@@ -164,6 +165,9 @@ static int power_saving_thread(void *data)
 				tsc_marked_unstable = 1;
 			}
 			local_irq_disable();
+
+			perf_lopwr_cb(true);
+
 			tick_broadcast_enable();
 			tick_broadcast_enter();
 			stop_critical_timings();
@@ -172,6 +176,9 @@ static int power_saving_thread(void *data)
 
 			start_critical_timings();
 			tick_broadcast_exit();
+
+			perf_lopwr_cb(false);
+
 			local_irq_enable();
 
 			if (time_before(expire_time, jiffies)) {
diff --git a/drivers/acpi/processor_idle.c b/drivers/acpi/processor_idle.c
index 32b20ef..05dc0e1 100644
--- a/drivers/acpi/processor_idle.c
+++ b/drivers/acpi/processor_idle.c
@@ -21,6 +21,7 @@
 #include <linux/cpuidle.h>
 #include <linux/cpu.h>
 #include <linux/minmax.h>
+#include <linux/perf_event.h>
 #include <acpi/processor.h>
 
 /*
@@ -549,6 +550,8 @@ static void wait_for_freeze(void)
  */
 static void __cpuidle acpi_idle_do_entry(struct acpi_processor_cx *cx)
 {
+	perf_lopwr_cb(true);
+
 	if (cx->entry_method == ACPI_CSTATE_FFH) {
 		/* Call into architectural FFH based C-state */
 		acpi_processor_ffh_cstate_enter(cx);
@@ -559,6 +562,8 @@ static void __cpuidle acpi_idle_do_entry(struct acpi_processor_cx *cx)
 		inb(cx->address);
 		wait_for_freeze();
 	}
+
+	perf_lopwr_cb(false);
 }
 
 /**
diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index a411080..da75956 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -1676,4 +1676,10 @@ typedef int (perf_snapshot_branch_stack_t)(struct perf_branch_entry *entries,
 					   unsigned int cnt);
 DECLARE_STATIC_CALL(perf_snapshot_branch_stack, perf_snapshot_branch_stack_t);
 
+#ifndef PERF_NEEDS_LOPWR_CB
+static inline void perf_lopwr_cb(bool mode)
+{
+}
+#endif
+
 #endif /* _LINUX_PERF_EVENT_H */

[tip: perf/core] perf/x86/amd: Make Zen3 branch sampling opt-in

[tip-bot2 for Stephane Eranian]

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

Commit-ID:     cc37e520a236069c0de0e7ea455082fa11c73b12
Gitweb:        https://git.kernel.org/tip/cc37e520a236069c0de0e7ea455082fa11c73b12
Author:        Stephane Eranian <eranian@google.com>
AuthorDate:    Tue, 22 Mar 2022 15:15:11 -07:00
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Tue, 05 Apr 2022 10:24:38 +02:00

perf/x86/amd: Make Zen3 branch sampling opt-in

Add a kernel config option CONFIG_PERF_EVENTS_AMD_BRS
to make the support for AMD Zen3 Branch Sampling (BRS) an opt-in
compile time option.

Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20220322221517.2510440-8-eranian@google.com
---
 arch/x86/events/Kconfig      |  8 ++++++-
 arch/x86/events/amd/Makefile |  3 +-
 arch/x86/events/perf_event.h | 49 +++++++++++++++++++++++++++--------
 3 files changed, 49 insertions(+), 11 deletions(-)

diff --git a/arch/x86/events/Kconfig b/arch/x86/events/Kconfig
index d6cdfe6..09c5696 100644
--- a/arch/x86/events/Kconfig
+++ b/arch/x86/events/Kconfig
@@ -44,4 +44,12 @@ config PERF_EVENTS_AMD_UNCORE
 
 	  To compile this driver as a module, choose M here: the
 	  module will be called 'amd-uncore'.
+
+config PERF_EVENTS_AMD_BRS
+	depends on PERF_EVENTS && CPU_SUP_AMD
+	bool "AMD Zen3 Branch Sampling support"
+	help
+	  Enable AMD Zen3 branch sampling support (BRS) which samples up to
+	  16 consecutive taken branches in registers.
+
 endmenu
diff --git a/arch/x86/events/amd/Makefile b/arch/x86/events/amd/Makefile
index cf323ff..b9f5d46 100644
--- a/arch/x86/events/amd/Makefile
+++ b/arch/x86/events/amd/Makefile
@@ -1,5 +1,6 @@
 # SPDX-License-Identifier: GPL-2.0
-obj-$(CONFIG_CPU_SUP_AMD)		+= core.o brs.o
+obj-$(CONFIG_CPU_SUP_AMD)		+= core.o
+obj-$(CONFIG_PERF_EVENTS_AMD_BRS)	+= brs.o
 obj-$(CONFIG_PERF_EVENTS_AMD_POWER)	+= power.o
 obj-$(CONFIG_X86_LOCAL_APIC)		+= ibs.o
 obj-$(CONFIG_PERF_EVENTS_AMD_UNCORE)	+= amd-uncore.o
diff --git a/arch/x86/events/perf_event.h b/arch/x86/events/perf_event.h
index d91ff2c..ef27aee 100644
--- a/arch/x86/events/perf_event.h
+++ b/arch/x86/events/perf_event.h
@@ -1218,6 +1218,8 @@ static inline bool fixed_counter_disabled(int i, struct pmu *pmu)
 #ifdef CONFIG_CPU_SUP_AMD
 
 int amd_pmu_init(void);
+
+#ifdef CONFIG_PERF_EVENTS_AMD_BRS
 int amd_brs_init(void);
 void amd_brs_disable(void);
 void amd_brs_enable(void);
@@ -1252,25 +1254,52 @@ static inline void amd_pmu_brs_del(struct perf_event *event)
 
 void amd_pmu_brs_sched_task(struct perf_event_context *ctx, bool sched_in);
 
-/*
- * check if BRS is activated on the CPU
- * active defined as it has non-zero users and DBG_EXT_CFG.BRSEN=1
- */
-static inline bool amd_brs_active(void)
+static inline s64 amd_brs_adjust_period(s64 period)
 {
-	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
+	if (period > x86_pmu.lbr_nr)
+		return period - x86_pmu.lbr_nr;
 
-	return cpuc->brs_active;
+	return period;
+}
+#else
+static inline int amd_brs_init(void)
+{
+	return 0;
 }
+static inline void amd_brs_disable(void) {}
+static inline void amd_brs_enable(void) {}
+static inline void amd_brs_drain(void) {}
+static inline void amd_brs_lopwr_init(void) {}
+static inline void amd_brs_disable_all(void) {}
+static inline int amd_brs_setup_filter(struct perf_event *event)
+{
+	return 0;
+}
+static inline void amd_brs_reset(void) {}
 
-static inline s64 amd_brs_adjust_period(s64 period)
+static inline void amd_pmu_brs_add(struct perf_event *event)
 {
-	if (period > x86_pmu.lbr_nr)
-		return period - x86_pmu.lbr_nr;
+}
+
+static inline void amd_pmu_brs_del(struct perf_event *event)
+{
+}
+
+static inline void amd_pmu_brs_sched_task(struct perf_event_context *ctx, bool sched_in)
+{
+}
 
+static inline s64 amd_brs_adjust_period(s64 period)
+{
 	return period;
 }
 
+static inline void amd_brs_enable_all(void)
+{
+}
+
+#endif
+
 #else /* CONFIG_CPU_SUP_AMD */
 
 static inline int amd_pmu_init(void)

[tip: perf/core] perf/x86/amd: Add AMD branch sampling period adjustment

[tip-bot2 for Stephane Eranian]

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

Commit-ID:     ba2fe7500845a30fc845a72081999cf632051862
Gitweb:        https://git.kernel.org/tip/ba2fe7500845a30fc845a72081999cf632051862
Author:        Stephane Eranian <eranian@google.com>
AuthorDate:    Tue, 22 Mar 2022 15:15:10 -07:00
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Tue, 05 Apr 2022 10:24:37 +02:00

perf/x86/amd: Add AMD branch sampling period adjustment

Add code to adjust the sampling event period when used with the Branch
Sampling feature (BRS). Given the depth of the BRS (16), the period is
reduced by that depth such that in the best case scenario, BRS saturates at
the desired sampling period. In practice, though, the processor may execute
more branches. Given a desired period P and a depth D, the kernel programs
the actual period at P - D. After P occurrences of the sampling event, the
counter overflows. It then may take X branches (skid) before the NMI is
caught and held by the hardware and BRS activates. Then, after D branches,
BRS saturates and the NMI is delivered.  With no skid, the effective period
would be (P - D) + D = P. In practice, however, it will likely be (P - D) +
X + D. There is no way to eliminate X or predict X.

Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20220322221517.2510440-7-eranian@google.com
---
 arch/x86/events/core.c       |  7 +++++++
 arch/x86/events/perf_event.h | 12 ++++++++++++
 2 files changed, 19 insertions(+)

diff --git a/arch/x86/events/core.c b/arch/x86/events/core.c
index 7ada917..54f992e 100644
--- a/arch/x86/events/core.c
+++ b/arch/x86/events/core.c
@@ -1375,6 +1375,13 @@ int x86_perf_event_set_period(struct perf_event *event)
 		return x86_pmu.set_topdown_event_period(event);
 
 	/*
+	 * decrease period by the depth of the BRS feature to get
+	 * the last N taken branches and approximate the desired period
+	 */
+	if (has_branch_stack(event))
+		period = amd_brs_adjust_period(period);
+
+	/*
 	 * If we are way outside a reasonable range then just skip forward:
 	 */
 	if (unlikely(left <= -period)) {
diff --git a/arch/x86/events/perf_event.h b/arch/x86/events/perf_event.h
index 6f12651..d91ff2c 100644
--- a/arch/x86/events/perf_event.h
+++ b/arch/x86/events/perf_event.h
@@ -1263,6 +1263,14 @@ static inline bool amd_brs_active(void)
 	return cpuc->brs_active;
 }
 
+static inline s64 amd_brs_adjust_period(s64 period)
+{
+	if (period > x86_pmu.lbr_nr)
+		return period - x86_pmu.lbr_nr;
+
+	return period;
+}
+
 #else /* CONFIG_CPU_SUP_AMD */
 
 static inline int amd_pmu_init(void)
@@ -1287,6 +1295,10 @@ static inline void amd_brs_disable_all(void)
 {
 }
 
+static inline s64 amd_brs_adjust_period(s64 period)
+{
+	return period;
+}
 #endif /* CONFIG_CPU_SUP_AMD */
 
 static inline int is_pebs_pt(struct perf_event *event)

[tip: perf/core] perf/x86/amd: Enable branch sampling priv level filtering

[tip-bot2 for Stephane Eranian]

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

Commit-ID:     8910075d61a37e5b0d82e6c83ed9a0a31fe9ea08
Gitweb:        https://git.kernel.org/tip/8910075d61a37e5b0d82e6c83ed9a0a31fe9ea08
Author:        Stephane Eranian <eranian@google.com>
AuthorDate:    Tue, 22 Mar 2022 15:15:09 -07:00
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Tue, 05 Apr 2022 10:24:37 +02:00

perf/x86/amd: Enable branch sampling priv level filtering

The AMD Branch Sampling features does not provide hardware filtering by
privilege level. The associated PMU counter does but not the branch sampling
by itself. Given how BRS operates there is a possibility that BRS captures
kernel level branches even though the event is programmed to count only at
the user level.

Implement a workaround in software by removing the branches which belong to
the wrong privilege level. The privilege level is evaluated on the target of
the branch and not the source so as to be compatible with other architectures.
As a consequence of this patch, the number of entries in the
PERF_RECORD_BRANCH_STACK buffer may be less than the maximum (16).  It could
even be zero. Another consequence is that consecutive entries in the branch
stack may not reflect actual code path and may have discontinuities, in case
kernel branches were suppressed. But this is no different than what happens
on other architectures.

Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20220322221517.2510440-6-eranian@google.com
---
 arch/x86/events/amd/brs.c | 26 ++++++++++++++++++++------
 1 file changed, 20 insertions(+), 6 deletions(-)

diff --git a/arch/x86/events/amd/brs.c b/arch/x86/events/amd/brs.c
index 3c13c48..40461c3 100644
--- a/arch/x86/events/amd/brs.c
+++ b/arch/x86/events/amd/brs.c
@@ -92,10 +92,6 @@ int amd_brs_setup_filter(struct perf_event *event)
 	if ((type & ~PERF_SAMPLE_BRANCH_PLM_ALL) != PERF_SAMPLE_BRANCH_ANY)
 		return -EINVAL;
 
-	/* can only capture at all priv levels due to the way BRS works */
-	if ((type & PERF_SAMPLE_BRANCH_PLM_ALL) != PERF_SAMPLE_BRANCH_PLM_ALL)
-		return -EINVAL;
-
 	return 0;
 }
 
@@ -195,6 +191,21 @@ void amd_brs_disable_all(void)
 		amd_brs_disable();
 }
 
+static bool amd_brs_match_plm(struct perf_event *event, u64 to)
+{
+	int type = event->attr.branch_sample_type;
+	int plm_k = PERF_SAMPLE_BRANCH_KERNEL | PERF_SAMPLE_BRANCH_HV;
+	int plm_u = PERF_SAMPLE_BRANCH_USER;
+
+	if (!(type & plm_k) && kernel_ip(to))
+		return 0;
+
+	if (!(type & plm_u) && !kernel_ip(to))
+		return 0;
+
+	return 1;
+}
+
 /*
  * Caller must ensure amd_brs_inuse() is true before calling
  * return:
@@ -252,8 +263,6 @@ void amd_brs_drain(void)
 		if (to == BRS_POISON)
 			break;
 
-		rdmsrl(brs_from(brs_idx), from);
-
 		/*
 		 * Sign-extend SAMP_BR_TO to 64 bits, bits 61-63 are reserved.
 		 * Necessary to generate proper virtual addresses suitable for
@@ -261,6 +270,11 @@ void amd_brs_drain(void)
 		 */
 		to = (u64)(((s64)to << shift) >> shift);
 
+		if (!amd_brs_match_plm(event, to))
+			continue;
+
+		rdmsrl(brs_from(brs_idx), from);
+
 		perf_clear_branch_entry_bitfields(br+nr);
 
 		br[nr].from = from;

[tip: perf/core] perf/x86/amd: Add branch-brs helper event for Fam19h BRS

[tip-bot2 for Stephane Eranian]

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

Commit-ID:     44175993efbae04e8b2d7f7795ff512c3a726db0
Gitweb:        https://git.kernel.org/tip/44175993efbae04e8b2d7f7795ff512c3a726db0
Author:        Stephane Eranian <eranian@google.com>
AuthorDate:    Tue, 22 Mar 2022 15:15:08 -07:00
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Tue, 05 Apr 2022 10:24:37 +02:00

perf/x86/amd: Add branch-brs helper event for Fam19h BRS

Add a pseudo event called branch-brs to help use the FAM Fam19h
Branch Sampling feature (BRS). BRS samples taken branches, so it is best used
when sampling on a retired taken branch event (0xc4) which is what BRS
captures.  Instead of trying to remember the event code or actual event name,
users can simply do:

$ perf record -b -e cpu/branch-brs/ -c 1000037 .....

Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20220322221517.2510440-5-eranian@google.com
---
 arch/x86/events/amd/core.c | 15 +++++++++++++++
 1 file changed, 15 insertions(+)

diff --git a/arch/x86/events/amd/core.c b/arch/x86/events/amd/core.c
index c7ac70d..f7bce83 100644
--- a/arch/x86/events/amd/core.c
+++ b/arch/x86/events/amd/core.c
@@ -1145,8 +1145,23 @@ static struct attribute_group group_caps_amd_brs = {
 	.is_visible = amd_brs_is_visible,
 };
 
+EVENT_ATTR_STR(branch-brs, amd_branch_brs,
+	       "event=" __stringify(AMD_FAM19H_BRS_EVENT)"\n");
+
+static struct attribute *amd_brs_events_attrs[] = {
+	EVENT_PTR(amd_branch_brs),
+	NULL,
+};
+
+static struct attribute_group group_events_amd_brs = {
+	.name       = "events",
+	.attrs      = amd_brs_events_attrs,
+	.is_visible = amd_brs_is_visible,
+};
+
 static const struct attribute_group *amd_attr_update[] = {
 	&group_caps_amd_brs,
+	&group_events_amd_brs,
 	NULL,
 };
 

[tip: perf/core] perf/x86/amd: Add AMD Fam19h Branch Sampling support

[tip-bot2 for Stephane Eranian]

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

Commit-ID:     ada543459cab7f653dcacdaba4011a8bb19c627c
Gitweb:        https://git.kernel.org/tip/ada543459cab7f653dcacdaba4011a8bb19c627c
Author:        Stephane Eranian <eranian@google.com>
AuthorDate:    Tue, 22 Mar 2022 15:15:07 -07:00
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Tue, 05 Apr 2022 10:24:37 +02:00

perf/x86/amd: Add AMD Fam19h Branch Sampling support

Add support for the AMD Fam19h 16-deep branch sampling feature as
described in the AMD PPR Fam19h Model 01h Revision B1.  This is a model
specific extension. It is not an architected AMD feature.

The Branch Sampling (BRS) operates with a 16-deep saturating buffer in MSR
registers. There is no branch type filtering. All control flow changes are
captured. BRS relies on specific programming of the core PMU of Fam19h.  In
particular, the following requirements must be met:
 - the sampling period be greater than 16 (BRS depth)
 - the sampling period must use a fixed and not frequency mode

BRS interacts with the NMI interrupt as well. Because enabling BRS is
expensive, it is only activated after P event occurrences, where P is the
desired sampling period.  At P occurrences of the event, the counter
overflows, the CPU catches the interrupt, activates BRS for 16 branches until
it saturates, and then delivers the NMI to the kernel.  Between the overflow
and the time BRS activates more branches may be executed skewing the period.
All along, the sampling event keeps counting. The skid may be attenuated by
reducing the sampling period by 16 (subsequent patch).

BRS is integrated into perf_events seamlessly via the same
PERF_RECORD_BRANCH_STACK sample format. BRS generates perf_branch_entry
records in the sampling buffer. No prediction information is supported. The
branches are stored in reverse order of execution.  The most recent branch is
the first entry in each record.

No modification to the perf tool is necessary.

BRS can be used with any sampling event. However, it is recommended to use
the RETIRED_BRANCH_INSTRUCTIONS event because it matches what the BRS
captures.

$ perf record -b -c 1000037 -e cpu/event=0xc2,name=ret_br_instructions/ test

$ perf report -D
56531696056126 0x193c000 [0x1a8]: PERF_RECORD_SAMPLE(IP, 0x2): 18122/18230: 0x401d24 period: 1000037 addr: 0
... branch stack: nr:16
.....  0: 0000000000401d24 -> 0000000000401d5a 0 cycles      0
.....  1: 0000000000401d5c -> 0000000000401d24 0 cycles      0
.....  2: 0000000000401d22 -> 0000000000401d5c 0 cycles      0
.....  3: 0000000000401d5e -> 0000000000401d22 0 cycles      0
.....  4: 0000000000401d20 -> 0000000000401d5e 0 cycles      0
.....  5: 0000000000401d3e -> 0000000000401d20 0 cycles      0
.....  6: 0000000000401d42 -> 0000000000401d3e 0 cycles      0
.....  7: 0000000000401d3c -> 0000000000401d42 0 cycles      0
.....  8: 0000000000401d44 -> 0000000000401d3c 0 cycles      0
.....  9: 0000000000401d3a -> 0000000000401d44 0 cycles      0
..... 10: 0000000000401d46 -> 0000000000401d3a 0 cycles      0
..... 11: 0000000000401d38 -> 0000000000401d46 0 cycles      0
..... 12: 0000000000401d48 -> 0000000000401d38 0 cycles      0
..... 13: 0000000000401d36 -> 0000000000401d48 0 cycles      0
..... 14: 0000000000401d4a -> 0000000000401d36 0 cycles      0
..... 15: 0000000000401d34 -> 0000000000401d4a 0 cycles      0
 ... thread: test:18230
 ...... dso: test

Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20220322221517.2510440-4-eranian@google.com
---
 arch/x86/events/amd/Makefile     |   2 +-
 arch/x86/events/amd/brs.c        | 317 ++++++++++++++++++++++++++++++-
 arch/x86/events/amd/core.c       | 233 +++++++++++++++++++++-
 arch/x86/events/core.c           |  10 +-
 arch/x86/events/perf_event.h     | 101 ++++++++--
 arch/x86/include/asm/msr-index.h |   4 +-
 6 files changed, 645 insertions(+), 22 deletions(-)
 create mode 100644 arch/x86/events/amd/brs.c

diff --git a/arch/x86/events/amd/Makefile b/arch/x86/events/amd/Makefile
index 6cbe38d..cf323ff 100644
--- a/arch/x86/events/amd/Makefile
+++ b/arch/x86/events/amd/Makefile
@@ -1,5 +1,5 @@
 # SPDX-License-Identifier: GPL-2.0
-obj-$(CONFIG_CPU_SUP_AMD)		+= core.o
+obj-$(CONFIG_CPU_SUP_AMD)		+= core.o brs.o
 obj-$(CONFIG_PERF_EVENTS_AMD_POWER)	+= power.o
 obj-$(CONFIG_X86_LOCAL_APIC)		+= ibs.o
 obj-$(CONFIG_PERF_EVENTS_AMD_UNCORE)	+= amd-uncore.o
diff --git a/arch/x86/events/amd/brs.c b/arch/x86/events/amd/brs.c
new file mode 100644
index 0000000..3c13c48
--- /dev/null
+++ b/arch/x86/events/amd/brs.c
@@ -0,0 +1,317 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Implement support for AMD Fam19h Branch Sampling feature
+ * Based on specifications published in AMD PPR Fam19 Model 01
+ *
+ * Copyright 2021 Google LLC
+ * Contributed by Stephane Eranian <eranian@google.com>
+ */
+#include <linux/kernel.h>
+#include <asm/msr.h>
+#include <asm/cpufeature.h>
+
+#include "../perf_event.h"
+
+#define BRS_POISON	0xFFFFFFFFFFFFFFFEULL /* mark limit of valid entries */
+
+/* Debug Extension Configuration register layout */
+union amd_debug_extn_cfg {
+	__u64 val;
+	struct {
+		__u64	rsvd0:2,  /* reserved */
+			brsmen:1, /* branch sample enable */
+			rsvd4_3:2,/* reserved - must be 0x3 */
+			vb:1,     /* valid branches recorded */
+			rsvd2:10, /* reserved */
+			msroff:4, /* index of next entry to write */
+			rsvd3:4,  /* reserved */
+			pmc:3,    /* #PMC holding the sampling event */
+			rsvd4:37; /* reserved */
+	};
+};
+
+static inline unsigned int brs_from(int idx)
+{
+	return MSR_AMD_SAMP_BR_FROM + 2 * idx;
+}
+
+static inline unsigned int brs_to(int idx)
+{
+	return MSR_AMD_SAMP_BR_FROM + 2 * idx + 1;
+}
+
+static inline void set_debug_extn_cfg(u64 val)
+{
+	/* bits[4:3] must always be set to 11b */
+	wrmsrl(MSR_AMD_DBG_EXTN_CFG, val | 3ULL << 3);
+}
+
+static inline u64 get_debug_extn_cfg(void)
+{
+	u64 val;
+
+	rdmsrl(MSR_AMD_DBG_EXTN_CFG, val);
+	return val;
+}
+
+static bool __init amd_brs_detect(void)
+{
+	if (!boot_cpu_has(X86_FEATURE_BRS))
+		return false;
+
+	switch (boot_cpu_data.x86) {
+	case 0x19: /* AMD Fam19h (Zen3) */
+		x86_pmu.lbr_nr = 16;
+
+		/* No hardware filtering supported */
+		x86_pmu.lbr_sel_map = NULL;
+		x86_pmu.lbr_sel_mask = 0;
+		break;
+	default:
+		return false;
+	}
+
+	return true;
+}
+
+/*
+ * Current BRS implementation does not support branch type or privilege level
+ * filtering. Therefore, this function simply enforces these limitations. No need for
+ * a br_sel_map. Software filtering is not supported because it would not correlate well
+ * with a sampling period.
+ */
+int amd_brs_setup_filter(struct perf_event *event)
+{
+	u64 type = event->attr.branch_sample_type;
+
+	/* No BRS support */
+	if (!x86_pmu.lbr_nr)
+		return -EOPNOTSUPP;
+
+	/* Can only capture all branches, i.e., no filtering */
+	if ((type & ~PERF_SAMPLE_BRANCH_PLM_ALL) != PERF_SAMPLE_BRANCH_ANY)
+		return -EINVAL;
+
+	/* can only capture at all priv levels due to the way BRS works */
+	if ((type & PERF_SAMPLE_BRANCH_PLM_ALL) != PERF_SAMPLE_BRANCH_PLM_ALL)
+		return -EINVAL;
+
+	return 0;
+}
+
+/* tos = top of stack, i.e., last valid entry written */
+static inline int amd_brs_get_tos(union amd_debug_extn_cfg *cfg)
+{
+	/*
+	 * msroff: index of next entry to write so top-of-stack is one off
+	 * if BRS is full then msroff is set back to 0.
+	 */
+	return (cfg->msroff ? cfg->msroff : x86_pmu.lbr_nr) - 1;
+}
+
+/*
+ * make sure we have a sane BRS offset to begin with
+ * especially with kexec
+ */
+void amd_brs_reset(void)
+{
+	/*
+	 * Reset config
+	 */
+	set_debug_extn_cfg(0);
+
+	/*
+	 * Mark first entry as poisoned
+	 */
+	wrmsrl(brs_to(0), BRS_POISON);
+}
+
+int __init amd_brs_init(void)
+{
+	if (!amd_brs_detect())
+		return -EOPNOTSUPP;
+
+	pr_cont("%d-deep BRS, ", x86_pmu.lbr_nr);
+
+	return 0;
+}
+
+void amd_brs_enable(void)
+{
+	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
+	union amd_debug_extn_cfg cfg;
+
+	/* Activate only on first user */
+	if (++cpuc->brs_active > 1)
+		return;
+
+	cfg.val    = 0; /* reset all fields */
+	cfg.brsmen = 1; /* enable branch sampling */
+
+	/* Set enable bit */
+	set_debug_extn_cfg(cfg.val);
+}
+
+void amd_brs_enable_all(void)
+{
+	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
+	if (cpuc->lbr_users)
+		amd_brs_enable();
+}
+
+void amd_brs_disable(void)
+{
+	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
+	union amd_debug_extn_cfg cfg;
+
+	/* Check if active (could be disabled via x86_pmu_disable_all()) */
+	if (!cpuc->brs_active)
+		return;
+
+	/* Only disable for last user */
+	if (--cpuc->brs_active)
+		return;
+
+	/*
+	 * Clear the brsmen bit but preserve the others as they contain
+	 * useful state such as vb and msroff
+	 */
+	cfg.val = get_debug_extn_cfg();
+
+	/*
+	 * When coming in on interrupt and BRS is full, then hw will have
+	 * already stopped BRS, no need to issue wrmsr again
+	 */
+	if (cfg.brsmen) {
+		cfg.brsmen = 0;
+		set_debug_extn_cfg(cfg.val);
+	}
+}
+
+void amd_brs_disable_all(void)
+{
+	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
+	if (cpuc->lbr_users)
+		amd_brs_disable();
+}
+
+/*
+ * Caller must ensure amd_brs_inuse() is true before calling
+ * return:
+ */
+void amd_brs_drain(void)
+{
+	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
+	struct perf_event *event = cpuc->events[0];
+	struct perf_branch_entry *br = cpuc->lbr_entries;
+	union amd_debug_extn_cfg cfg;
+	u32 i, nr = 0, num, tos, start;
+	u32 shift = 64 - boot_cpu_data.x86_virt_bits;
+
+	/*
+	 * BRS event forced on PMC0,
+	 * so check if there is an event.
+	 * It is possible to have lbr_users > 0 but the event
+	 * not yet scheduled due to long latency PMU irq
+	 */
+	if (!event)
+		goto empty;
+
+	cfg.val = get_debug_extn_cfg();
+
+	/* Sanity check [0-x86_pmu.lbr_nr] */
+	if (WARN_ON_ONCE(cfg.msroff >= x86_pmu.lbr_nr))
+		goto empty;
+
+	/* No valid branch */
+	if (cfg.vb == 0)
+		goto empty;
+
+	/*
+	 * msr.off points to next entry to be written
+	 * tos = most recent entry index = msr.off - 1
+	 * BRS register buffer saturates, so we know we have
+	 * start < tos and that we have to read from start to tos
+	 */
+	start = 0;
+	tos = amd_brs_get_tos(&cfg);
+
+	num = tos - start + 1;
+
+	/*
+	 * BRS is only one pass (saturation) from MSROFF to depth-1
+	 * MSROFF wraps to zero when buffer is full
+	 */
+	for (i = 0; i < num; i++) {
+		u32 brs_idx = tos - i;
+		u64 from, to;
+
+		rdmsrl(brs_to(brs_idx), to);
+
+		/* Entry does not belong to us (as marked by kernel) */
+		if (to == BRS_POISON)
+			break;
+
+		rdmsrl(brs_from(brs_idx), from);
+
+		/*
+		 * Sign-extend SAMP_BR_TO to 64 bits, bits 61-63 are reserved.
+		 * Necessary to generate proper virtual addresses suitable for
+		 * symbolization
+		 */
+		to = (u64)(((s64)to << shift) >> shift);
+
+		perf_clear_branch_entry_bitfields(br+nr);
+
+		br[nr].from = from;
+		br[nr].to   = to;
+
+		nr++;
+	}
+empty:
+	/* Record number of sampled branches */
+	cpuc->lbr_stack.nr = nr;
+}
+
+/*
+ * Poison most recent entry to prevent reuse by next task
+ * required because BRS entry are not tagged by PID
+ */
+static void amd_brs_poison_buffer(void)
+{
+	union amd_debug_extn_cfg cfg;
+	unsigned int idx;
+
+	/* Get current state */
+	cfg.val = get_debug_extn_cfg();
+
+	/* idx is most recently written entry */
+	idx = amd_brs_get_tos(&cfg);
+
+	/* Poison target of entry */
+	wrmsrl(brs_to(idx), BRS_POISON);
+}
+
+/*
+ * On context switch in, we need to make sure no samples from previous user
+ * are left in the BRS.
+ *
+ * On ctxswin, sched_in = true, called after the PMU has started
+ * On ctxswout, sched_in = false, called before the PMU is stopped
+ */
+void amd_pmu_brs_sched_task(struct perf_event_context *ctx, bool sched_in)
+{
+	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
+
+	/* no active users */
+	if (!cpuc->lbr_users)
+		return;
+
+	/*
+	 * On context switch in, we need to ensure we do not use entries
+	 * from previous BRS user on that CPU, so we poison the buffer as
+	 * a faster way compared to resetting all entries.
+	 */
+	if (sched_in)
+		amd_brs_poison_buffer();
+}
diff --git a/arch/x86/events/amd/core.c b/arch/x86/events/amd/core.c
index 9687a8a..c7ac70d 100644
--- a/arch/x86/events/amd/core.c
+++ b/arch/x86/events/amd/core.c
@@ -325,8 +325,16 @@ static inline bool amd_is_pair_event_code(struct hw_perf_event *hwc)
 	}
 }
 
+#define AMD_FAM19H_BRS_EVENT 0xc4 /* RETIRED_TAKEN_BRANCH_INSTRUCTIONS */
+static inline int amd_is_brs_event(struct perf_event *e)
+{
+	return (e->hw.config & AMD64_RAW_EVENT_MASK) == AMD_FAM19H_BRS_EVENT;
+}
+
 static int amd_core_hw_config(struct perf_event *event)
 {
+	int ret = 0;
+
 	if (event->attr.exclude_host && event->attr.exclude_guest)
 		/*
 		 * When HO == GO == 1 the hardware treats that as GO == HO == 0
@@ -343,7 +351,66 @@ static int amd_core_hw_config(struct perf_event *event)
 	if ((x86_pmu.flags & PMU_FL_PAIR) && amd_is_pair_event_code(&event->hw))
 		event->hw.flags |= PERF_X86_EVENT_PAIR;
 
-	return 0;
+	/*
+	 * if branch stack is requested
+	 */
+	if (has_branch_stack(event)) {
+		/*
+		 * Due to interrupt holding, BRS is not recommended in
+		 * counting mode.
+		 */
+		if (!is_sampling_event(event))
+			return -EINVAL;
+
+		/*
+		 * Due to the way BRS operates by holding the interrupt until
+		 * lbr_nr entries have been captured, it does not make sense
+		 * to allow sampling on BRS with an event that does not match
+		 * what BRS is capturing, i.e., retired taken branches.
+		 * Otherwise the correlation with the event's period is even
+		 * more loose:
+		 *
+		 * With retired taken branch:
+		 *   Effective P = P + 16 + X
+		 * With any other event:
+		 *   Effective P = P + Y + X
+		 *
+		 * Where X is the number of taken branches due to interrupt
+		 * skid. Skid is large.
+		 *
+		 * Where Y is the occurences of the event while BRS is
+		 * capturing the lbr_nr entries.
+		 *
+		 * By using retired taken branches, we limit the impact on the
+		 * Y variable. We know it cannot be more than the depth of
+		 * BRS.
+		 */
+		if (!amd_is_brs_event(event))
+			return -EINVAL;
+
+		/*
+		 * BRS implementation does not work with frequency mode
+		 * reprogramming of the period.
+		 */
+		if (event->attr.freq)
+			return -EINVAL;
+		/*
+		 * The kernel subtracts BRS depth from period, so it must
+		 * be big enough.
+		 */
+		if (event->attr.sample_period <= x86_pmu.lbr_nr)
+			return -EINVAL;
+
+		/*
+		 * Check if we can allow PERF_SAMPLE_BRANCH_STACK
+		 */
+		ret = amd_brs_setup_filter(event);
+
+		/* only set in case of success */
+		if (!ret)
+			event->hw.flags |= PERF_X86_EVENT_AMD_BRS;
+	}
+	return ret;
 }
 
 static inline int amd_is_nb_event(struct hw_perf_event *hwc)
@@ -366,7 +433,7 @@ static int amd_pmu_hw_config(struct perf_event *event)
 	if (event->attr.precise_ip && get_ibs_caps())
 		return -ENOENT;
 
-	if (has_branch_stack(event))
+	if (has_branch_stack(event) && !x86_pmu.lbr_nr)
 		return -EOPNOTSUPP;
 
 	ret = x86_pmu_hw_config(event);
@@ -555,6 +622,8 @@ static void amd_pmu_cpu_starting(int cpu)
 
 	cpuc->amd_nb->nb_id = nb_id;
 	cpuc->amd_nb->refcnt++;
+
+	amd_brs_reset();
 }
 
 static void amd_pmu_cpu_dead(int cpu)
@@ -610,6 +679,8 @@ static void amd_pmu_disable_all(void)
 	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
 	int idx;
 
+	amd_brs_disable_all();
+
 	x86_pmu_disable_all();
 
 	/*
@@ -634,6 +705,30 @@ static void amd_pmu_disable_all(void)
 	}
 }
 
+static void amd_pmu_enable_event(struct perf_event *event)
+{
+	x86_pmu_enable_event(event);
+}
+
+static void amd_pmu_enable_all(int added)
+{
+	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
+	struct hw_perf_event *hwc;
+	int idx;
+
+	amd_brs_enable_all();
+
+	for (idx = 0; idx < x86_pmu.num_counters; idx++) {
+		hwc = &cpuc->events[idx]->hw;
+
+		/* only activate events which are marked as active */
+		if (!test_bit(idx, cpuc->active_mask))
+			continue;
+
+		amd_pmu_enable_event(cpuc->events[idx]);
+	}
+}
+
 static void amd_pmu_disable_event(struct perf_event *event)
 {
 	x86_pmu_disable_event(event);
@@ -651,6 +746,18 @@ static void amd_pmu_disable_event(struct perf_event *event)
 	amd_pmu_wait_on_overflow(event->hw.idx);
 }
 
+static void amd_pmu_add_event(struct perf_event *event)
+{
+	if (needs_branch_stack(event))
+		amd_pmu_brs_add(event);
+}
+
+static void amd_pmu_del_event(struct perf_event *event)
+{
+	if (needs_branch_stack(event))
+		amd_pmu_brs_del(event);
+}
+
 /*
  * Because of NMI latency, if multiple PMC counters are active or other sources
  * of NMIs are received, the perf NMI handler can handle one or more overflowed
@@ -671,11 +778,31 @@ static void amd_pmu_disable_event(struct perf_event *event)
  */
 static int amd_pmu_handle_irq(struct pt_regs *regs)
 {
+	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
 	int handled;
+	int pmu_enabled;
+
+	/*
+	 * Save the PMU state.
+	 * It needs to be restored when leaving the handler.
+	 */
+	pmu_enabled = cpuc->enabled;
+	cpuc->enabled = 0;
+
+	/* stop everything (includes BRS) */
+	amd_pmu_disable_all();
+
+	/* Drain BRS is in use (could be inactive) */
+	if (cpuc->lbr_users)
+		amd_brs_drain();
 
 	/* Process any counter overflows */
 	handled = x86_pmu_handle_irq(regs);
 
+	cpuc->enabled = pmu_enabled;
+	if (pmu_enabled)
+		amd_pmu_enable_all(0);
+
 	/*
 	 * If a counter was handled, record a timestamp such that un-handled
 	 * NMIs will be claimed if arriving within that window.
@@ -897,6 +1024,51 @@ static void amd_put_event_constraints_f17h(struct cpu_hw_events *cpuc,
 		--cpuc->n_pair;
 }
 
+/*
+ * Because of the way BRS operates with an inactive and active phases, and
+ * the link to one counter, it is not possible to have two events using BRS
+ * scheduled at the same time. There would be an issue with enforcing the
+ * period of each one and given that the BRS saturates, it would not be possible
+ * to guarantee correlated content for all events. Therefore, in situations
+ * where multiple events want to use BRS, the kernel enforces mutual exclusion.
+ * Exclusion is enforced by chosing only one counter for events using BRS.
+ * The event scheduling logic will then automatically multiplex the
+ * events and ensure that at most one event is actively using BRS.
+ *
+ * The BRS counter could be any counter, but there is no constraint on Fam19h,
+ * therefore all counters are equal and thus we pick the first one: PMC0
+ */
+static struct event_constraint amd_fam19h_brs_cntr0_constraint =
+	EVENT_CONSTRAINT(0, 0x1, AMD64_RAW_EVENT_MASK);
+
+static struct event_constraint amd_fam19h_brs_pair_cntr0_constraint =
+	__EVENT_CONSTRAINT(0, 0x1, AMD64_RAW_EVENT_MASK, 1, 0, PERF_X86_EVENT_PAIR);
+
+static struct event_constraint *
+amd_get_event_constraints_f19h(struct cpu_hw_events *cpuc, int idx,
+			  struct perf_event *event)
+{
+	struct hw_perf_event *hwc = &event->hw;
+	bool has_brs = has_amd_brs(hwc);
+
+	/*
+	 * In case BRS is used with an event requiring a counter pair,
+	 * the kernel allows it but only on counter 0 & 1 to enforce
+	 * multiplexing requiring to protect BRS in case of multiple
+	 * BRS users
+	 */
+	if (amd_is_pair_event_code(hwc)) {
+		return has_brs ? &amd_fam19h_brs_pair_cntr0_constraint
+			       : &pair_constraint;
+	}
+
+	if (has_brs)
+		return &amd_fam19h_brs_cntr0_constraint;
+
+	return &unconstrained;
+}
+
+
 static ssize_t amd_event_sysfs_show(char *page, u64 config)
 {
 	u64 event = (config & ARCH_PERFMON_EVENTSEL_EVENT) |
@@ -905,12 +1077,19 @@ static ssize_t amd_event_sysfs_show(char *page, u64 config)
 	return x86_event_sysfs_show(page, config, event);
 }
 
+static void amd_pmu_sched_task(struct perf_event_context *ctx,
+				 bool sched_in)
+{
+	if (sched_in && x86_pmu.lbr_nr)
+		amd_pmu_brs_sched_task(ctx, sched_in);
+}
+
 static __initconst const struct x86_pmu amd_pmu = {
 	.name			= "AMD",
 	.handle_irq		= amd_pmu_handle_irq,
 	.disable_all		= amd_pmu_disable_all,
-	.enable_all		= x86_pmu_enable_all,
-	.enable			= x86_pmu_enable_event,
+	.enable_all		= amd_pmu_enable_all,
+	.enable			= amd_pmu_enable_event,
 	.disable		= amd_pmu_disable_event,
 	.hw_config		= amd_pmu_hw_config,
 	.schedule_events	= x86_schedule_events,
@@ -920,6 +1099,8 @@ static __initconst const struct x86_pmu amd_pmu = {
 	.event_map		= amd_pmu_event_map,
 	.max_events		= ARRAY_SIZE(amd_perfmon_event_map),
 	.num_counters		= AMD64_NUM_COUNTERS,
+	.add			= amd_pmu_add_event,
+	.del			= amd_pmu_del_event,
 	.cntval_bits		= 48,
 	.cntval_mask		= (1ULL << 48) - 1,
 	.apic			= 1,
@@ -938,6 +1119,37 @@ static __initconst const struct x86_pmu amd_pmu = {
 	.amd_nb_constraints	= 1,
 };
 
+static ssize_t branches_show(struct device *cdev,
+			      struct device_attribute *attr,
+			      char *buf)
+{
+	return snprintf(buf, PAGE_SIZE, "%d\n", x86_pmu.lbr_nr);
+}
+
+static DEVICE_ATTR_RO(branches);
+
+static struct attribute *amd_pmu_brs_attrs[] = {
+	&dev_attr_branches.attr,
+	NULL,
+};
+
+static umode_t
+amd_brs_is_visible(struct kobject *kobj, struct attribute *attr, int i)
+{
+	return x86_pmu.lbr_nr ? attr->mode : 0;
+}
+
+static struct attribute_group group_caps_amd_brs = {
+	.name  = "caps",
+	.attrs = amd_pmu_brs_attrs,
+	.is_visible = amd_brs_is_visible,
+};
+
+static const struct attribute_group *amd_attr_update[] = {
+	&group_caps_amd_brs,
+	NULL,
+};
+
 static int __init amd_core_pmu_init(void)
 {
 	u64 even_ctr_mask = 0ULL;
@@ -989,6 +1201,19 @@ static int __init amd_core_pmu_init(void)
 		x86_pmu.flags |= PMU_FL_PAIR;
 	}
 
+	/*
+	 * BRS requires special event constraints and flushing on ctxsw.
+	 */
+	if (boot_cpu_data.x86 >= 0x19 && !amd_brs_init()) {
+		x86_pmu.get_event_constraints = amd_get_event_constraints_f19h;
+		x86_pmu.sched_task = amd_pmu_sched_task;
+		/*
+		 * put_event_constraints callback same as Fam17h, set above
+		 */
+	}
+
+	x86_pmu.attr_update = amd_attr_update;
+
 	pr_cont("core perfctr, ");
 	return 0;
 }
diff --git a/arch/x86/events/core.c b/arch/x86/events/core.c
index eef816f..7ada917 100644
--- a/arch/x86/events/core.c
+++ b/arch/x86/events/core.c
@@ -1338,6 +1338,10 @@ static void x86_pmu_enable(struct pmu *pmu)
 			if (hwc->state & PERF_HES_ARCH)
 				continue;
 
+			/*
+			 * if cpuc->enabled = 0, then no wrmsr as
+			 * per x86_pmu_enable_event()
+			 */
 			x86_pmu_start(event, PERF_EF_RELOAD);
 		}
 		cpuc->n_added = 0;
@@ -1704,11 +1708,15 @@ int x86_pmu_handle_irq(struct pt_regs *regs)
 		 * event overflow
 		 */
 		handled++;
-		perf_sample_data_init(&data, 0, event->hw.last_period);
 
 		if (!x86_perf_event_set_period(event))
 			continue;
 
+		perf_sample_data_init(&data, 0, event->hw.last_period);
+
+		if (has_branch_stack(event))
+			data.br_stack = &cpuc->lbr_stack;
+
 		if (perf_event_overflow(event, &data, regs))
 			x86_pmu_stop(event, 0);
 	}
diff --git a/arch/x86/events/perf_event.h b/arch/x86/events/perf_event.h
index 150261d..6f12651 100644
--- a/arch/x86/events/perf_event.h
+++ b/arch/x86/events/perf_event.h
@@ -67,22 +67,23 @@ static inline bool constraint_match(struct event_constraint *c, u64 ecode)
 /*
  * struct hw_perf_event.flags flags
  */
-#define PERF_X86_EVENT_PEBS_LDLAT	0x0001 /* ld+ldlat data address sampling */
-#define PERF_X86_EVENT_PEBS_ST		0x0002 /* st data address sampling */
-#define PERF_X86_EVENT_PEBS_ST_HSW	0x0004 /* haswell style datala, store */
-#define PERF_X86_EVENT_PEBS_LD_HSW	0x0008 /* haswell style datala, load */
-#define PERF_X86_EVENT_PEBS_NA_HSW	0x0010 /* haswell style datala, unknown */
-#define PERF_X86_EVENT_EXCL		0x0020 /* HT exclusivity on counter */
-#define PERF_X86_EVENT_DYNAMIC		0x0040 /* dynamic alloc'd constraint */
-
-#define PERF_X86_EVENT_EXCL_ACCT	0x0100 /* accounted EXCL event */
-#define PERF_X86_EVENT_AUTO_RELOAD	0x0200 /* use PEBS auto-reload */
-#define PERF_X86_EVENT_LARGE_PEBS	0x0400 /* use large PEBS */
-#define PERF_X86_EVENT_PEBS_VIA_PT	0x0800 /* use PT buffer for PEBS */
-#define PERF_X86_EVENT_PAIR		0x1000 /* Large Increment per Cycle */
-#define PERF_X86_EVENT_LBR_SELECT	0x2000 /* Save/Restore MSR_LBR_SELECT */
-#define PERF_X86_EVENT_TOPDOWN		0x4000 /* Count Topdown slots/metrics events */
-#define PERF_X86_EVENT_PEBS_STLAT	0x8000 /* st+stlat data address sampling */
+#define PERF_X86_EVENT_PEBS_LDLAT	0x00001 /* ld+ldlat data address sampling */
+#define PERF_X86_EVENT_PEBS_ST		0x00002 /* st data address sampling */
+#define PERF_X86_EVENT_PEBS_ST_HSW	0x00004 /* haswell style datala, store */
+#define PERF_X86_EVENT_PEBS_LD_HSW	0x00008 /* haswell style datala, load */
+#define PERF_X86_EVENT_PEBS_NA_HSW	0x00010 /* haswell style datala, unknown */
+#define PERF_X86_EVENT_EXCL		0x00020 /* HT exclusivity on counter */
+#define PERF_X86_EVENT_DYNAMIC		0x00040 /* dynamic alloc'd constraint */
+
+#define PERF_X86_EVENT_EXCL_ACCT	0x00100 /* accounted EXCL event */
+#define PERF_X86_EVENT_AUTO_RELOAD	0x00200 /* use PEBS auto-reload */
+#define PERF_X86_EVENT_LARGE_PEBS	0x00400 /* use large PEBS */
+#define PERF_X86_EVENT_PEBS_VIA_PT	0x00800 /* use PT buffer for PEBS */
+#define PERF_X86_EVENT_PAIR		0x01000 /* Large Increment per Cycle */
+#define PERF_X86_EVENT_LBR_SELECT	0x02000 /* Save/Restore MSR_LBR_SELECT */
+#define PERF_X86_EVENT_TOPDOWN		0x04000 /* Count Topdown slots/metrics events */
+#define PERF_X86_EVENT_PEBS_STLAT	0x08000 /* st+stlat data address sampling */
+#define PERF_X86_EVENT_AMD_BRS		0x10000 /* AMD Branch Sampling */
 
 static inline bool is_topdown_count(struct perf_event *event)
 {
@@ -325,6 +326,8 @@ struct cpu_hw_events {
 	 * AMD specific bits
 	 */
 	struct amd_nb			*amd_nb;
+	int				brs_active; /* BRS is enabled */
+
 	/* Inverted mask of bits to clear in the perf_ctr ctrl registers */
 	u64				perf_ctr_virt_mask;
 	int				n_pair; /* Large increment events */
@@ -1105,6 +1108,11 @@ int x86_pmu_hw_config(struct perf_event *event);
 
 void x86_pmu_disable_all(void);
 
+static inline bool has_amd_brs(struct hw_perf_event *hwc)
+{
+	return hwc->flags & PERF_X86_EVENT_AMD_BRS;
+}
+
 static inline bool is_counter_pair(struct hw_perf_event *hwc)
 {
 	return hwc->flags & PERF_X86_EVENT_PAIR;
@@ -1210,6 +1218,50 @@ static inline bool fixed_counter_disabled(int i, struct pmu *pmu)
 #ifdef CONFIG_CPU_SUP_AMD
 
 int amd_pmu_init(void);
+int amd_brs_init(void);
+void amd_brs_disable(void);
+void amd_brs_enable(void);
+void amd_brs_enable_all(void);
+void amd_brs_disable_all(void);
+void amd_brs_drain(void);
+void amd_brs_disable_all(void);
+int amd_brs_setup_filter(struct perf_event *event);
+void amd_brs_reset(void);
+
+static inline void amd_pmu_brs_add(struct perf_event *event)
+{
+	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
+
+	perf_sched_cb_inc(event->ctx->pmu);
+	cpuc->lbr_users++;
+	/*
+	 * No need to reset BRS because it is reset
+	 * on brs_enable() and it is saturating
+	 */
+}
+
+static inline void amd_pmu_brs_del(struct perf_event *event)
+{
+	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
+
+	cpuc->lbr_users--;
+	WARN_ON_ONCE(cpuc->lbr_users < 0);
+
+	perf_sched_cb_dec(event->ctx->pmu);
+}
+
+void amd_pmu_brs_sched_task(struct perf_event_context *ctx, bool sched_in);
+
+/*
+ * check if BRS is activated on the CPU
+ * active defined as it has non-zero users and DBG_EXT_CFG.BRSEN=1
+ */
+static inline bool amd_brs_active(void)
+{
+	struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
+
+	return cpuc->brs_active;
+}
 
 #else /* CONFIG_CPU_SUP_AMD */
 
@@ -1218,6 +1270,23 @@ static inline int amd_pmu_init(void)
 	return 0;
 }
 
+static inline int amd_brs_init(void)
+{
+	return -EOPNOTSUPP;
+}
+
+static inline void amd_brs_drain(void)
+{
+}
+
+static inline void amd_brs_enable_all(void)
+{
+}
+
+static inline void amd_brs_disable_all(void)
+{
+}
+
 #endif /* CONFIG_CPU_SUP_AMD */
 
 static inline int is_pebs_pt(struct perf_event *event)
diff --git a/arch/x86/include/asm/msr-index.h b/arch/x86/include/asm/msr-index.h
index 0eb90d2..8179ea3 100644
--- a/arch/x86/include/asm/msr-index.h
+++ b/arch/x86/include/asm/msr-index.h
@@ -688,6 +688,10 @@
 #define MSR_IA32_PERF_CTL		0x00000199
 #define INTEL_PERF_CTL_MASK		0xffff
 
+/* AMD Branch Sampling configuration */
+#define MSR_AMD_DBG_EXTN_CFG		0xc000010f
+#define MSR_AMD_SAMP_BR_FROM		0xc0010300
+
 #define MSR_IA32_MPERF			0x000000e7
 #define MSR_IA32_APERF			0x000000e8
 

[tip: perf/core] x86/cpufeatures: Add AMD Fam19h Branch Sampling feature

[tip-bot2 for Stephane Eranian]

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

Commit-ID:     a77d41ac3a0f41c80120ec5b8b08ab284fec950a
Gitweb:        https://git.kernel.org/tip/a77d41ac3a0f41c80120ec5b8b08ab284fec950a
Author:        Stephane Eranian <eranian@google.com>
AuthorDate:    Tue, 22 Mar 2022 15:15:06 -07:00
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Tue, 05 Apr 2022 10:24:36 +02:00

x86/cpufeatures: Add AMD Fam19h Branch Sampling feature

Add a cpu feature for AMD Fam19h Branch Sampling feature as bit
31 of EBX on CPUID leaf function 0x80000008.

Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20220322221517.2510440-3-eranian@google.com
---
 arch/x86/include/asm/cpufeatures.h | 1 +
 1 file changed, 1 insertion(+)

diff --git a/arch/x86/include/asm/cpufeatures.h b/arch/x86/include/asm/cpufeatures.h
index 73e643a..0d62afd 100644
--- a/arch/x86/include/asm/cpufeatures.h
+++ b/arch/x86/include/asm/cpufeatures.h
@@ -315,6 +315,7 @@
 #define X86_FEATURE_VIRT_SSBD		(13*32+25) /* Virtualized Speculative Store Bypass Disable */
 #define X86_FEATURE_AMD_SSB_NO		(13*32+26) /* "" Speculative Store Bypass is fixed in hardware. */
 #define X86_FEATURE_CPPC		(13*32+27) /* Collaborative Processor Performance Control */
+#define X86_FEATURE_BRS			(13*32+31) /* Branch Sampling available */
 
 /* Thermal and Power Management Leaf, CPUID level 0x00000006 (EAX), word 14 */
 #define X86_FEATURE_DTHERM		(14*32+ 0) /* Digital Thermal Sensor */

[tip: perf/core] perf/core: Add perf_clear_branch_entry_bitfields() helper

[tip-bot2 for Stephane Eranian]

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

Commit-ID:     bfe4daf850f45d92dcd3da477f0b0456620294c3
Gitweb:        https://git.kernel.org/tip/bfe4daf850f45d92dcd3da477f0b0456620294c3
Author:        Stephane Eranian <eranian@google.com>
AuthorDate:    Tue, 22 Mar 2022 15:15:05 -07:00
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Tue, 05 Apr 2022 10:24:36 +02:00

perf/core: Add perf_clear_branch_entry_bitfields() helper

Make it simpler to reset all the info fields on the
perf_branch_entry by adding a helper inline function.

The goal is to centralize the initialization to avoid missing
a field in case more are added.

Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20220322221517.2510440-2-eranian@google.com
---
 arch/x86/events/intel/lbr.c | 36 +++++++++++++++++-------------------
 include/linux/perf_event.h  | 16 ++++++++++++++++-
 2 files changed, 33 insertions(+), 19 deletions(-)

diff --git a/arch/x86/events/intel/lbr.c b/arch/x86/events/intel/lbr.c
index fe1742c..13179f3 100644
--- a/arch/x86/events/intel/lbr.c
+++ b/arch/x86/events/intel/lbr.c
@@ -769,6 +769,7 @@ void intel_pmu_lbr_disable_all(void)
 void intel_pmu_lbr_read_32(struct cpu_hw_events *cpuc)
 {
 	unsigned long mask = x86_pmu.lbr_nr - 1;
+	struct perf_branch_entry *br = cpuc->lbr_entries;
 	u64 tos = intel_pmu_lbr_tos();
 	int i;
 
@@ -784,15 +785,11 @@ void intel_pmu_lbr_read_32(struct cpu_hw_events *cpuc)
 
 		rdmsrl(x86_pmu.lbr_from + lbr_idx, msr_lastbranch.lbr);
 
-		cpuc->lbr_entries[i].from	= msr_lastbranch.from;
-		cpuc->lbr_entries[i].to		= msr_lastbranch.to;
-		cpuc->lbr_entries[i].mispred	= 0;
-		cpuc->lbr_entries[i].predicted	= 0;
-		cpuc->lbr_entries[i].in_tx	= 0;
-		cpuc->lbr_entries[i].abort	= 0;
-		cpuc->lbr_entries[i].cycles	= 0;
-		cpuc->lbr_entries[i].type	= 0;
-		cpuc->lbr_entries[i].reserved	= 0;
+		perf_clear_branch_entry_bitfields(br);
+
+		br->from	= msr_lastbranch.from;
+		br->to		= msr_lastbranch.to;
+		br++;
 	}
 	cpuc->lbr_stack.nr = i;
 	cpuc->lbr_stack.hw_idx = tos;
@@ -807,6 +804,7 @@ void intel_pmu_lbr_read_64(struct cpu_hw_events *cpuc)
 {
 	bool need_info = false, call_stack = false;
 	unsigned long mask = x86_pmu.lbr_nr - 1;
+	struct perf_branch_entry *br = cpuc->lbr_entries;
 	u64 tos = intel_pmu_lbr_tos();
 	int i;
 	int out = 0;
@@ -878,15 +876,14 @@ void intel_pmu_lbr_read_64(struct cpu_hw_events *cpuc)
 		if (abort && x86_pmu.lbr_double_abort && out > 0)
 			out--;
 
-		cpuc->lbr_entries[out].from	 = from;
-		cpuc->lbr_entries[out].to	 = to;
-		cpuc->lbr_entries[out].mispred	 = mis;
-		cpuc->lbr_entries[out].predicted = pred;
-		cpuc->lbr_entries[out].in_tx	 = in_tx;
-		cpuc->lbr_entries[out].abort	 = abort;
-		cpuc->lbr_entries[out].cycles	 = cycles;
-		cpuc->lbr_entries[out].type	 = 0;
-		cpuc->lbr_entries[out].reserved	 = 0;
+		perf_clear_branch_entry_bitfields(br+out);
+		br[out].from	 = from;
+		br[out].to	 = to;
+		br[out].mispred	 = mis;
+		br[out].predicted = pred;
+		br[out].in_tx	 = in_tx;
+		br[out].abort	 = abort;
+		br[out].cycles	 = cycles;
 		out++;
 	}
 	cpuc->lbr_stack.nr = out;
@@ -951,6 +948,8 @@ static void intel_pmu_store_lbr(struct cpu_hw_events *cpuc,
 		to = rdlbr_to(i, lbr);
 		info = rdlbr_info(i, lbr);
 
+		perf_clear_branch_entry_bitfields(e);
+
 		e->from		= from;
 		e->to		= to;
 		e->mispred	= get_lbr_mispred(info);
@@ -959,7 +958,6 @@ static void intel_pmu_store_lbr(struct cpu_hw_events *cpuc,
 		e->abort	= !!(info & LBR_INFO_ABORT);
 		e->cycles	= get_lbr_cycles(info);
 		e->type		= get_lbr_br_type(info);
-		e->reserved	= 0;
 	}
 
 	cpuc->lbr_stack.nr = i;
diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index af97dd4..a411080 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -1063,6 +1063,22 @@ static inline void perf_sample_data_init(struct perf_sample_data *data,
 	data->txn = 0;
 }
 
+/*
+ * Clear all bitfields in the perf_branch_entry.
+ * The to and from fields are not cleared because they are
+ * systematically modified by caller.
+ */
+static inline void perf_clear_branch_entry_bitfields(struct perf_branch_entry *br)
+{
+	br->mispred = 0;
+	br->predicted = 0;
+	br->in_tx = 0;
+	br->abort = 0;
+	br->cycles = 0;
+	br->type = 0;
+	br->reserved = 0;
+}
+
 extern void perf_output_sample(struct perf_output_handle *handle,
 			       struct perf_event_header *header,
 			       struct perf_sample_data *data,