[PATCH v3 1/2] perf tools: Add ARM Statistical Profiling Extensions (SPE) support

Kim Phillips kim.phillips at arm.com
Fri Oct 27 16:37:43 PDT 2017


'perf record' and 'perf report --dump-raw-trace' supported in this
release.

Example usage:

$ ./perf record -e arm_spe_0/ts_enable=1,pa_enable=1/ \
		dd if=/dev/zero of=/dev/null count=10000

perf report --dump-raw-trace

Note that the perf.data file is portable, so the report can be run on
another architecture host if necessary.

Output will contain raw SPE data and its textual representation, such
as:

0x550 [0x30]: PERF_RECORD_AUXTRACE size: 0xc408  offset: 0  ref: 0x30005619  idx: 3  tid: 2109  cpu: 3
.
. ... ARM SPE data: size 50184 bytes
.  00000000:  49 00                                           LD
.  00000002:  b2 00 9c 7b 7a 00 80 ff ff                      VA 0xffff80007a7b9c00
.  0000000b:  9a 00 00                                        LAT 0 XLAT
.  0000000e:  42 16                                           EV RETIRED L1D-ACCESS TLB-ACCESS
.  00000010:  b0 b0 c9 15 08 00 00 ff ff                      PC 0xff00000815c9b0 el3 ns=1
.  00000019:  98 00 00                                        LAT 0 TOT
.  0000001c:  71 00 20 fa fd 16 00 00 00                      TS 98750308352
.  00000025:  49 01                                           ST
.  00000027:  b2 60 bc 0c 0f 00 00 ff ff                      VA 0xffff00000f0cbc60
.  00000030:  9a 00 00                                        LAT 0 XLAT
.  00000033:  42 16                                           EV RETIRED L1D-ACCESS TLB-ACCESS
.  00000035:  b0 48 cc 15 08 00 00 ff ff                      PC 0xff00000815cc48 el3 ns=1
.  0000003e:  98 00 00                                        LAT 0 TOT
.  00000041:  71 00 20 fa fd 16 00 00 00                      TS 98750308352
.  0000004a:  48 00                                           INSN-OTHER
.  0000004c:  42 02                                           EV RETIRED
.  0000004e:  b0 ac 47 0c 08 00 00 ff ff                      PC 0xff0000080c47ac el3 ns=1
.  00000057:  98 00 00                                        LAT 0 TOT
.  0000005a:  71 00 20 fa fd 16 00 00 00                      TS 98750308352
.  00000063:  49 00                                           LD
.  00000065:  b2 18 48 e5 7a 00 80 ff ff                      VA 0xffff80007ae54818
.  0000006e:  9a 00 00                                        LAT 0 XLAT
.  00000071:  42 16                                           EV RETIRED L1D-ACCESS TLB-ACCESS
.  00000073:  b0 08 f8 15 08 00 00 ff ff                      PC 0xff00000815f808 el3 ns=1
.  0000007c:  98 00 00                                        LAT 0 TOT
.  0000007f:  71 00 20 fa fd 16 00 00 00                      TS 98750308352
...

Other release notes:

- applies to acme's perf/{core,urgent} branches, likely elsewhere

- Report is self-contained within the tool.  Record requires Will's SPE
  driver:

  http://lists.infradead.org/pipermail/linux-arm-kernel/2017-October/536221.html

- the intel-bts implementation was used as a starting point; its
  min/default/max buffer sizes and power of 2 pages granularity need to be
  revisited for ARM SPE

- multiple SPE clusters/domains not supported

- snapshot support (record -S), and conversion to native perf events
  (e.g., via 'perf inject --itrace'), are also not supported

- technically both cs-etm and spe can be used simultaneously, however
  disabled for simplicity in this release

Signed-off-by: Kim Phillips <kim.phillips at arm.com>
---
v3: trying to address comments from v2:

- despite adding a find_all_arm_spe_pmus() function to scan for all
  arm_spe_<n> device instances, in order to ensure auxtrace_record__init
  successfully matches the evsel type with the correct arm_spe_pmu type,
  I am still having trouble running in multi-SPE PPI (heterogeneous)
  environments (mmap fails with EOPNOTSUPP, as does running with
  --per-thread on homogeneous systems).

- arm_spe_reference: use gettime instead of direct cntvct register access

- spe-decoder: add a comment for why SPE_EVENTS code sets packet->index.

- added arm_spe_pmu_default_config that accesses the driver
  caps/min_interval and sets the default sampling period to it.  This way
  users don't have to specify -c explicitly.  Also set is_uncore to false.

- set more sampling bits in the arm_spe and its tracking evsel.  Still
  unsure if too liberal, and not sure whether it needs another context
  switch tracking evsel.  Comments welcome!

v2: mostly addressing Mark Rutland's comments as much as possible without his
feedback to my feedback:

- decoder refactored with a get_payload, not extended to with-ext_len ones like
  get_addr,  named the constants

- 0x-ified %x output formats, but decided to not sign extend the addresses in
  the raw dump, rather do so if necessary in the synthesis stage:
  SPE implementations differ in this area, and raw dump should reflect that.

- CPU mask / new record behaviour bisected to commit e3ba76deef23064 "perf
  tools: Force uncore events to system wide monitoring".  Waiting to hear back
  on why driver can't do system wide monitoring, even across PPIs, by e.g.,
  sharing the SPE interrupts in one handler (SPE's don't differ in this record
  regard).

- addressed off-list comment from M. Williams:
  "Instruction Type" packet was renamed as "Operation Type".
   so in the spe packet decoder: INSN_TYPE -> OP_TYPE

- do_get_packet fixed to handle excessive, successive PADding from a new source
  of raw SPE data, so instead of:

	.  000011ae:  00                                              PAD
	.  000011af:  00                                              PAD
	.  000011b0:  00                                              PAD
	.  000011b1:  00                                              PAD
	.  000011b2:  00                                              PAD
	.  000011b3:  00                                              PAD
	.  000011b4:  00                                              PAD
	.  000011b5:  00                                              PAD
	.  000011b6:  00                                              PAD

  we now get:

	.  000011ae:  00 00 00 00 00 00 00 00 00                      PAD

- fixed 52 00 00 decoded with an empty events clause, adding 'EV' for all events
  clauses now.  parser writers can detect for empty event clauses by finding
  nothing after it.

 tools/perf/arch/arm/util/auxtrace.c   |  75 +++++-
 tools/perf/arch/arm/util/pmu.c        |   5 +-
 tools/perf/arch/arm64/util/Build      |   3 +-
 tools/perf/arch/arm64/util/arm-spe.c  | 235 +++++++++++++++++
 tools/perf/util/Build                 |   2 +
 tools/perf/util/arm-spe-pkt-decoder.c | 472 ++++++++++++++++++++++++++++++++++
 tools/perf/util/arm-spe-pkt-decoder.h |  52 ++++
 tools/perf/util/arm-spe.c             | 318 +++++++++++++++++++++++
 tools/perf/util/arm-spe.h             |  42 +++
 tools/perf/util/auxtrace.c            |   3 +
 tools/perf/util/auxtrace.h            |   1 +
 11 files changed, 1200 insertions(+), 8 deletions(-)
 create mode 100644 tools/perf/arch/arm64/util/arm-spe.c
 create mode 100644 tools/perf/util/arm-spe-pkt-decoder.c
 create mode 100644 tools/perf/util/arm-spe-pkt-decoder.h
 create mode 100644 tools/perf/util/arm-spe.c
 create mode 100644 tools/perf/util/arm-spe.h

diff --git a/tools/perf/arch/arm/util/auxtrace.c b/tools/perf/arch/arm/util/auxtrace.c
index 8edf2cb71564..8e7c1ad18224 100644
--- a/tools/perf/arch/arm/util/auxtrace.c
+++ b/tools/perf/arch/arm/util/auxtrace.c
@@ -22,6 +22,42 @@
 #include "../../util/evlist.h"
 #include "../../util/pmu.h"
 #include "cs-etm.h"
+#include "arm-spe.h"
+
+static struct perf_pmu **find_all_arm_spe_pmus(int *nr_spes, int *err)
+{
+	struct perf_pmu **arm_spe_pmus = NULL;
+	int ret, i, nr_cpus = sysconf(_SC_NPROCESSORS_CONF);
+	/* arm_spe_xxxxxxxxx\0 */
+	char arm_spe_pmu_name[sizeof(ARM_SPE_PMU_NAME) + 10];
+
+	arm_spe_pmus = zalloc(sizeof(struct perf_pmu *) * nr_cpus);
+	if (!arm_spe_pmus) {
+		pr_err("spes alloc failed\n");
+		*err = -ENOMEM;
+		return NULL;
+	}
+
+	for (i = 0; i < nr_cpus; i++) {
+		ret = sprintf(arm_spe_pmu_name, "%s%d", ARM_SPE_PMU_NAME, i);
+		if (ret < 0) {
+			pr_err("sprintf failed\n");
+			*err = -ENOMEM;
+			return NULL;
+		}
+
+		arm_spe_pmus[*nr_spes] = perf_pmu__find(arm_spe_pmu_name);
+		if (arm_spe_pmus[*nr_spes]) {
+			pr_debug2("%s %d: arm_spe_pmu %d type %d name %s\n",
+				 __func__, __LINE__, *nr_spes,
+				 arm_spe_pmus[*nr_spes]->type,
+				 arm_spe_pmus[*nr_spes]->name);
+			(*nr_spes)++;
+		}
+	}
+
+	return arm_spe_pmus;
+}
 
 struct auxtrace_record
 *auxtrace_record__init(struct perf_evlist *evlist, int *err)
@@ -29,22 +65,49 @@ struct auxtrace_record
 	struct perf_pmu	*cs_etm_pmu;
 	struct perf_evsel *evsel;
 	bool found_etm = false;
+	bool found_spe = false;
+	static struct perf_pmu **arm_spe_pmus = NULL;
+	static int nr_spes = 0;
+	int i;
+
+	if (!evlist)
+		return NULL;
 
 	cs_etm_pmu = perf_pmu__find(CORESIGHT_ETM_PMU_NAME);
 
-	if (evlist) {
-		evlist__for_each_entry(evlist, evsel) {
-			if (cs_etm_pmu &&
-			    evsel->attr.type == cs_etm_pmu->type)
-				found_etm = true;
+	if (!arm_spe_pmus)
+		arm_spe_pmus = find_all_arm_spe_pmus(&nr_spes, err);
+
+	evlist__for_each_entry(evlist, evsel) {
+		if (cs_etm_pmu &&
+		    evsel->attr.type == cs_etm_pmu->type)
+			found_etm = true;
+
+		if (!nr_spes)
+			continue;
+
+		for (i = 0; i < nr_spes; i++) {
+			if (evsel->attr.type == arm_spe_pmus[i]->type) {
+				found_spe = true;
+				break;
+			}
 		}
 	}
 
+	if (found_etm && found_spe) {
+		pr_err("Concurrent ARM Coresight ETM and SPE operation not currently supported\n");
+		*err = -EOPNOTSUPP;
+		return NULL;
+	}
+
 	if (found_etm)
 		return cs_etm_record_init(err);
 
+	if (found_spe)
+		return arm_spe_recording_init(err, arm_spe_pmus[i]);
+
 	/*
-	 * Clear 'err' even if we haven't found a cs_etm event - that way perf
+	 * Clear 'err' even if we haven't found an event - that way perf
 	 * record can still be used even if tracers aren't present.  The NULL
 	 * return value will take care of telling the infrastructure HW tracing
 	 * isn't available.
diff --git a/tools/perf/arch/arm/util/pmu.c b/tools/perf/arch/arm/util/pmu.c
index 98d67399a0d6..4c06a25ae6b1 100644
--- a/tools/perf/arch/arm/util/pmu.c
+++ b/tools/perf/arch/arm/util/pmu.c
@@ -20,6 +20,7 @@
 #include <linux/perf_event.h>
 
 #include "cs-etm.h"
+#include "arm-spe.h"
 #include "../../util/pmu.h"
 
 struct perf_event_attr
@@ -30,7 +31,9 @@ struct perf_event_attr
 		/* add ETM default config here */
 		pmu->selectable = true;
 		pmu->set_drv_config = cs_etm_set_drv_config;
-	}
+	} else
+		if (strstarts(pmu->name, ARM_SPE_PMU_NAME))
+			return arm_spe_pmu_default_config(pmu);
 #endif
 	return NULL;
 }
diff --git a/tools/perf/arch/arm64/util/Build b/tools/perf/arch/arm64/util/Build
index 8dee4aa31a68..a80a53f89cf2 100644
--- a/tools/perf/arch/arm64/util/Build
+++ b/tools/perf/arch/arm64/util/Build
@@ -5,4 +5,5 @@ libperf-$(CONFIG_LOCAL_LIBUNWIND) += unwind-libunwind.o
 
 libperf-$(CONFIG_AUXTRACE) += ../../arm/util/pmu.o \
 			      ../../arm/util/auxtrace.o \
-			      ../../arm/util/cs-etm.o
+			      ../../arm/util/cs-etm.o \
+			      arm-spe.o
diff --git a/tools/perf/arch/arm64/util/arm-spe.c b/tools/perf/arch/arm64/util/arm-spe.c
new file mode 100644
index 000000000000..ef576b52c850
--- /dev/null
+++ b/tools/perf/arch/arm64/util/arm-spe.c
@@ -0,0 +1,235 @@
+/*
+ * ARM Statistical Profiling Extensions (SPE) support
+ * Copyright (c) 2017, ARM Ltd.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ */
+
+#include <linux/kernel.h>
+#include <linux/types.h>
+#include <linux/bitops.h>
+#include <linux/log2.h>
+#include <time.h>
+
+#include "../../util/cpumap.h"
+#include "../../util/evsel.h"
+#include "../../util/evlist.h"
+#include "../../util/session.h"
+#include "../../util/util.h"
+#include "../../util/pmu.h"
+#include "../../util/debug.h"
+#include "../../util/tsc.h"
+#include "../../util/auxtrace.h"
+#include "../../util/arm-spe.h"
+
+#define KiB(x) ((x) * 1024)
+#define MiB(x) ((x) * 1024 * 1024)
+
+struct arm_spe_recording {
+	struct auxtrace_record		itr;
+	struct perf_pmu			*arm_spe_pmu;
+	struct perf_evlist		*evlist;
+};
+
+static size_t
+arm_spe_info_priv_size(struct auxtrace_record *itr __maybe_unused,
+		       struct perf_evlist *evlist __maybe_unused)
+{
+	return ARM_SPE_AUXTRACE_PRIV_SIZE;
+}
+
+static int arm_spe_info_fill(struct auxtrace_record *itr,
+			     struct perf_session *session,
+			     struct auxtrace_info_event *auxtrace_info,
+			     size_t priv_size)
+{
+	struct arm_spe_recording *sper =
+			container_of(itr, struct arm_spe_recording, itr);
+	struct perf_pmu *arm_spe_pmu = sper->arm_spe_pmu;
+
+	if (priv_size != ARM_SPE_AUXTRACE_PRIV_SIZE)
+		return -EINVAL;
+
+	if (!session->evlist->nr_mmaps)
+		return -EINVAL;
+
+	auxtrace_info->type = PERF_AUXTRACE_ARM_SPE;
+	auxtrace_info->priv[ARM_SPE_PMU_TYPE] = arm_spe_pmu->type;
+
+	return 0;
+}
+
+static int arm_spe_recording_options(struct auxtrace_record *itr,
+				     struct perf_evlist *evlist,
+				     struct record_opts *opts)
+{
+	struct arm_spe_recording *sper =
+			container_of(itr, struct arm_spe_recording, itr);
+	struct perf_pmu *arm_spe_pmu = sper->arm_spe_pmu;
+	struct perf_evsel *evsel, *arm_spe_evsel = NULL;
+	bool privileged = geteuid() == 0 || perf_event_paranoid() < 0;
+	struct perf_evsel *tracking_evsel;
+	int err;
+
+	sper->evlist = evlist;
+
+	evlist__for_each_entry(evlist, evsel) {
+		if (evsel->attr.type == arm_spe_pmu->type) {
+			if (arm_spe_evsel) {
+				pr_err("There may be only one " ARM_SPE_PMU_NAME "x event\n");
+				return -EINVAL;
+			}
+			evsel->attr.freq = 0;
+			evsel->attr.sample_period = 1;
+			arm_spe_evsel = evsel;
+			opts->full_auxtrace = true;
+		}
+	}
+
+	if (!opts->full_auxtrace)
+		return 0;
+
+	/* We are in full trace mode but '-m,xyz' wasn't specified */
+	if (opts->full_auxtrace && !opts->auxtrace_mmap_pages) {
+		if (privileged) {
+			opts->auxtrace_mmap_pages = MiB(4) / page_size;
+		} else {
+			opts->auxtrace_mmap_pages = KiB(128) / page_size;
+			if (opts->mmap_pages == UINT_MAX)
+				opts->mmap_pages = KiB(256) / page_size;
+		}
+	}
+
+	/* Validate auxtrace_mmap_pages */
+	if (opts->auxtrace_mmap_pages) {
+		size_t sz = opts->auxtrace_mmap_pages * (size_t)page_size;
+		size_t min_sz = KiB(8);
+
+		if (sz < min_sz || !is_power_of_2(sz)) {
+			pr_err("Invalid mmap size for ARM SPE: must be at least %zuKiB and a power of 2\n",
+			       min_sz / 1024);
+			return -EINVAL;
+		}
+	}
+
+
+	/*
+	 * To obtain the auxtrace buffer file descriptor, the auxtrace event
+	 * must come first.
+	 */
+	perf_evlist__to_front(evlist, arm_spe_evsel);
+
+	perf_evsel__set_sample_bit(arm_spe_evsel, CPU);
+	perf_evsel__set_sample_bit(arm_spe_evsel, TIME);
+	perf_evsel__set_sample_bit(arm_spe_evsel, TID);
+
+	/* Add dummy event to keep tracking */
+	err = parse_events(evlist, "dummy:u", NULL);
+	if (err)
+		return err;
+
+	tracking_evsel = perf_evlist__last(evlist);
+	perf_evlist__set_tracking_event(evlist, tracking_evsel);
+
+	tracking_evsel->attr.freq = 0;
+	tracking_evsel->attr.sample_period = 1;
+	perf_evsel__set_sample_bit(tracking_evsel, TIME);
+	perf_evsel__set_sample_bit(tracking_evsel, CPU);
+	perf_evsel__reset_sample_bit(tracking_evsel, BRANCH_STACK);
+
+	return 0;
+}
+
+static u64 arm_spe_reference(struct auxtrace_record *itr __maybe_unused)
+{
+	struct timespec ts;
+
+	clock_gettime(CLOCK_MONOTONIC_RAW, &ts);
+
+	return ts.tv_sec ^ ts.tv_nsec;
+}
+
+static void arm_spe_recording_free(struct auxtrace_record *itr)
+{
+	struct arm_spe_recording *sper =
+			container_of(itr, struct arm_spe_recording, itr);
+
+	free(sper);
+}
+
+static int arm_spe_read_finish(struct auxtrace_record *itr, int idx)
+{
+	struct arm_spe_recording *sper =
+			container_of(itr, struct arm_spe_recording, itr);
+	struct perf_evsel *evsel;
+
+	evlist__for_each_entry(sper->evlist, evsel) {
+		if (evsel->attr.type == sper->arm_spe_pmu->type)
+			return perf_evlist__enable_event_idx(sper->evlist,
+							     evsel, idx);
+	}
+	return -EINVAL;
+}
+
+struct auxtrace_record *arm_spe_recording_init(int *err,
+					       struct perf_pmu *arm_spe_pmu)
+{
+	struct arm_spe_recording *sper;
+
+	if (!arm_spe_pmu) {
+		*err = -ENODEV;
+		return NULL;
+	}
+
+	sper = zalloc(sizeof(struct arm_spe_recording));
+	if (!sper) {
+		*err = -ENOMEM;
+		return NULL;
+	}
+
+	sper->arm_spe_pmu = arm_spe_pmu;
+	sper->itr.recording_options = arm_spe_recording_options;
+	sper->itr.info_priv_size = arm_spe_info_priv_size;
+	sper->itr.info_fill = arm_spe_info_fill;
+	sper->itr.free = arm_spe_recording_free;
+	sper->itr.reference = arm_spe_reference;
+	sper->itr.read_finish = arm_spe_read_finish;
+	sper->itr.alignment = 0;
+
+	return &sper->itr;
+}
+
+struct perf_event_attr
+*arm_spe_pmu_default_config(struct perf_pmu *arm_spe_pmu)
+{
+	struct perf_event_attr *attr;
+
+	attr = zalloc(sizeof(struct perf_event_attr));
+	if (!attr) {
+		pr_err("arm_spe default config cannot allocate a perf_event_attr\n");
+		return NULL;
+	}
+
+	/*
+	 * If kernel driver doesn't advertise a minimum,
+	 * use max allowable by PMSIDR_EL1.INTERVAL
+	 */
+	if (perf_pmu__scan_file(arm_spe_pmu, "caps/min_interval", "%llu",
+				  &attr->sample_period) != 1) {
+		pr_debug("arm_spe driver doesn't advertise a min. interval. Using 4096\n");
+		attr->sample_period = 4096;
+	}
+
+	arm_spe_pmu->selectable = true;
+	arm_spe_pmu->is_uncore = false;
+
+	return attr;
+}
diff --git a/tools/perf/util/Build b/tools/perf/util/Build
index 94518c1bf8b6..e6cb696359a2 100644
--- a/tools/perf/util/Build
+++ b/tools/perf/util/Build
@@ -83,6 +83,8 @@ libperf-$(CONFIG_AUXTRACE) += auxtrace.o
 libperf-$(CONFIG_AUXTRACE) += intel-pt-decoder/
 libperf-$(CONFIG_AUXTRACE) += intel-pt.o
 libperf-$(CONFIG_AUXTRACE) += intel-bts.o
+libperf-$(CONFIG_AUXTRACE) += arm-spe.o
+libperf-$(CONFIG_AUXTRACE) += arm-spe-pkt-decoder.o
 libperf-y += parse-branch-options.o
 libperf-y += dump-insn.o
 libperf-y += parse-regs-options.o
diff --git a/tools/perf/util/arm-spe-pkt-decoder.c b/tools/perf/util/arm-spe-pkt-decoder.c
new file mode 100644
index 000000000000..833dad42988a
--- /dev/null
+++ b/tools/perf/util/arm-spe-pkt-decoder.c
@@ -0,0 +1,472 @@
+/*
+ * ARM Statistical Profiling Extensions (SPE) support
+ * Copyright (c) 2017, ARM Ltd.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ */
+
+#include <stdio.h>
+#include <string.h>
+#include <endian.h>
+#include <byteswap.h>
+
+#include "arm-spe-pkt-decoder.h"
+
+#define BIT(n)		(1ULL << (n))
+
+#define NS_FLAG		BIT(63)
+#define EL_FLAG		(BIT(62) | BIT(61))
+
+#define SPE_HEADER0_PAD			0x0
+#define SPE_HEADER0_END			0x1
+#define SPE_HEADER0_ADDRESS		0x30 /* address packet (short) */
+#define SPE_HEADER0_ADDRESS_MASK	0x38
+#define SPE_HEADER0_COUNTER		0x18 /* counter packet (short) */
+#define SPE_HEADER0_COUNTER_MASK	0x38
+#define SPE_HEADER0_TIMESTAMP		0x71
+#define SPE_HEADER0_TIMESTAMP		0x71
+#define SPE_HEADER0_EVENTS		0x2
+#define SPE_HEADER0_EVENTS_MASK		0xf
+#define SPE_HEADER0_SOURCE		0x3
+#define SPE_HEADER0_SOURCE_MASK		0xf
+#define SPE_HEADER0_CONTEXT		0x24
+#define SPE_HEADER0_CONTEXT_MASK	0x3c
+#define SPE_HEADER0_OP_TYPE		0x8
+#define SPE_HEADER0_OP_TYPE_MASK	0x3c
+#define SPE_HEADER1_ALIGNMENT		0x0
+#define SPE_HEADER1_ADDRESS		0xb0 /* address packet (extended) */
+#define SPE_HEADER1_ADDRESS_MASK	0xf8
+#define SPE_HEADER1_COUNTER		0x98 /* counter packet (extended) */
+#define SPE_HEADER1_COUNTER_MASK	0xf8
+
+#if __BYTE_ORDER == __BIG_ENDIAN
+#define le16_to_cpu bswap_16
+#define le32_to_cpu bswap_32
+#define le64_to_cpu bswap_64
+#define memcpy_le64(d, s, n) do { \
+	memcpy((d), (s), (n));    \
+	*(d) = le64_to_cpu(*(d)); \
+} while (0)
+#else
+#define le16_to_cpu
+#define le32_to_cpu
+#define le64_to_cpu
+#define memcpy_le64 memcpy
+#endif
+
+static const char * const arm_spe_packet_name[] = {
+	[ARM_SPE_PAD]		= "PAD",
+	[ARM_SPE_END]		= "END",
+	[ARM_SPE_TIMESTAMP]	= "TS",
+	[ARM_SPE_ADDRESS]	= "ADDR",
+	[ARM_SPE_COUNTER]	= "LAT",
+	[ARM_SPE_CONTEXT]	= "CONTEXT",
+	[ARM_SPE_OP_TYPE]	= "OP-TYPE",
+	[ARM_SPE_EVENTS]	= "EVENTS",
+	[ARM_SPE_DATA_SOURCE]	= "DATA-SOURCE",
+};
+
+const char *arm_spe_pkt_name(enum arm_spe_pkt_type type)
+{
+	return arm_spe_packet_name[type];
+}
+
+/* return ARM SPE payload size from its encoding,
+ * which is in bits 5:4 of the byte.
+ * 00 : byte
+ * 01 : halfword (2)
+ * 10 : word (4)
+ * 11 : doubleword (8)
+ */
+static int payloadlen(unsigned char byte)
+{
+	return 1 << ((byte & 0x30) >> 4);
+}
+
+static int arm_spe_get_payload(const unsigned char *buf, size_t len,
+			       struct arm_spe_pkt *packet)
+{
+	size_t payload_len = payloadlen(buf[0]);
+
+	if (len < 1 + payload_len)
+		return ARM_SPE_NEED_MORE_BYTES;
+
+	buf++;
+
+	switch (payload_len) {
+	case 1: packet->payload = *(uint8_t *)buf; break;
+	case 2: packet->payload = le16_to_cpu(*(uint16_t *)buf); break;
+	case 4: packet->payload = le32_to_cpu(*(uint32_t *)buf); break;
+	case 8: packet->payload = le64_to_cpu(*(uint64_t *)buf); break;
+	default: return ARM_SPE_BAD_PACKET;
+	}
+
+	return 1 + payload_len;
+}
+
+static int arm_spe_get_pad(struct arm_spe_pkt *packet)
+{
+	packet->type = ARM_SPE_PAD;
+	return 1;
+}
+
+static int arm_spe_get_alignment(const unsigned char *buf, size_t len,
+				 struct arm_spe_pkt *packet)
+{
+	unsigned int alignment = 1 << ((buf[0] & 0xf) + 1);
+
+	if (len < alignment)
+		return ARM_SPE_NEED_MORE_BYTES;
+
+	packet->type = ARM_SPE_PAD;
+	return alignment - (((uint64_t)buf) & (alignment - 1));
+}
+
+static int arm_spe_get_end(struct arm_spe_pkt *packet)
+{
+	packet->type = ARM_SPE_END;
+	return 1;
+}
+
+static int arm_spe_get_timestamp(const unsigned char *buf, size_t len,
+				 struct arm_spe_pkt *packet)
+{
+	packet->type = ARM_SPE_TIMESTAMP;
+	return arm_spe_get_payload(buf, len, packet);
+}
+
+static int arm_spe_get_events(const unsigned char *buf, size_t len,
+			      struct arm_spe_pkt *packet)
+{
+	int ret = arm_spe_get_payload(buf, len, packet);
+
+	packet->type = ARM_SPE_EVENTS;
+
+	/* we use index to identify Events with a less number of
+	 * comparisons in arm_spe_pkt_desc(): E.g., the LLC-ACCESS,
+	 * LLC-REFILL, and REMOTE-ACCESS events are identified iff
+	 * index > 1.
+	 */
+	packet->index = ret - 1;
+
+	return ret;
+}
+
+static int arm_spe_get_data_source(const unsigned char *buf, size_t len,
+				   struct arm_spe_pkt *packet)
+{
+	packet->type = ARM_SPE_DATA_SOURCE;
+	return arm_spe_get_payload(buf, len, packet);
+}
+
+static int arm_spe_get_context(const unsigned char *buf, size_t len,
+			       struct arm_spe_pkt *packet)
+{
+	packet->type = ARM_SPE_CONTEXT;
+	packet->index = buf[0] & 0x3;
+
+	return arm_spe_get_payload(buf, len, packet);
+}
+
+static int arm_spe_get_op_type(const unsigned char *buf, size_t len,
+			       struct arm_spe_pkt *packet)
+{
+	packet->type = ARM_SPE_OP_TYPE;
+	packet->index = buf[0] & 0x3;
+	return arm_spe_get_payload(buf, len, packet);
+}
+
+static int arm_spe_get_counter(const unsigned char *buf, size_t len,
+			       const unsigned char ext_hdr, struct arm_spe_pkt *packet)
+{
+	if (len < 2)
+		return ARM_SPE_NEED_MORE_BYTES;
+
+	packet->type = ARM_SPE_COUNTER;
+	if (ext_hdr)
+		packet->index = ((buf[0] & 0x3) << 3) | (buf[1] & 0x7);
+	else
+		packet->index = buf[0] & 0x7;
+
+	packet->payload = le16_to_cpu(*(uint16_t *)(buf + 1));
+
+	return 1 + ext_hdr + 2;
+}
+
+static int arm_spe_get_addr(const unsigned char *buf, size_t len,
+			    const unsigned char ext_hdr, struct arm_spe_pkt *packet)
+{
+	if (len < 8)
+		return ARM_SPE_NEED_MORE_BYTES;
+
+	packet->type = ARM_SPE_ADDRESS;
+	if (ext_hdr)
+		packet->index = ((buf[0] & 0x3) << 3) | (buf[1] & 0x7);
+	else
+		packet->index = buf[0] & 0x7;
+
+	memcpy_le64(&packet->payload, buf + 1, 8);
+
+	return 1 + ext_hdr + 8;
+}
+
+static int arm_spe_do_get_packet(const unsigned char *buf, size_t len,
+				 struct arm_spe_pkt *packet)
+{
+	unsigned int byte;
+
+	memset(packet, 0, sizeof(struct arm_spe_pkt));
+
+	if (!len)
+		return ARM_SPE_NEED_MORE_BYTES;
+
+	byte = buf[0];
+	if (byte == SPE_HEADER0_PAD)
+		return arm_spe_get_pad(packet);
+	else if (byte == SPE_HEADER0_END) /* no timestamp at end of record */
+		return arm_spe_get_end(packet);
+	else if (byte & 0xc0 /* 0y11xxxxxx */) {
+		if (byte & 0x80) {
+			if ((byte & SPE_HEADER0_ADDRESS_MASK) == SPE_HEADER0_ADDRESS)
+				return arm_spe_get_addr(buf, len, 0, packet);
+			if ((byte & SPE_HEADER0_COUNTER_MASK) == SPE_HEADER0_COUNTER)
+				return arm_spe_get_counter(buf, len, 0, packet);
+		} else
+			if (byte == SPE_HEADER0_TIMESTAMP)
+				return arm_spe_get_timestamp(buf, len, packet);
+			else if ((byte & SPE_HEADER0_EVENTS_MASK) == SPE_HEADER0_EVENTS)
+				return arm_spe_get_events(buf, len, packet);
+			else if ((byte & SPE_HEADER0_SOURCE_MASK) == SPE_HEADER0_SOURCE)
+				return arm_spe_get_data_source(buf, len, packet);
+			else if ((byte & SPE_HEADER0_CONTEXT_MASK) == SPE_HEADER0_CONTEXT)
+				return arm_spe_get_context(buf, len, packet);
+			else if ((byte & SPE_HEADER0_OP_TYPE_MASK) == SPE_HEADER0_OP_TYPE)
+				return arm_spe_get_op_type(buf, len, packet);
+	} else if ((byte & 0xe0) == 0x20 /* 0y001xxxxx */) {
+		/* 16-bit header */ 
+		byte = buf[1];
+		if (byte == SPE_HEADER1_ALIGNMENT)
+			return arm_spe_get_alignment(buf, len, packet);
+		else if ((byte & SPE_HEADER1_ADDRESS_MASK) == SPE_HEADER1_ADDRESS)
+			return arm_spe_get_addr(buf, len, 1, packet);
+		else if ((byte & SPE_HEADER1_COUNTER_MASK) == SPE_HEADER1_COUNTER)
+			return arm_spe_get_counter(buf, len, 1, packet);
+	}
+
+	return ARM_SPE_BAD_PACKET;
+}
+
+int arm_spe_get_packet(const unsigned char *buf, size_t len,
+		       struct arm_spe_pkt *packet)
+{
+	int ret;
+
+	ret = arm_spe_do_get_packet(buf, len, packet);
+	/* put multiple consecutive PADs on the same line, up to
+	 * the fixed-width output format of 16 bytes per line.
+	 */
+	if (ret > 0 && packet->type == ARM_SPE_PAD) {
+		while (ret < 16 && len > (size_t)ret && !buf[ret])
+			ret += 1;
+	}
+	return ret;
+}
+
+int arm_spe_pkt_desc(const struct arm_spe_pkt *packet, char *buf,
+		     size_t buf_len)
+{
+	int ret, ns, el, index = packet->index;
+	unsigned long long payload = packet->payload;
+	const char *name = arm_spe_pkt_name(packet->type);
+
+	switch (packet->type) {
+	case ARM_SPE_BAD:
+	case ARM_SPE_PAD:
+	case ARM_SPE_END:
+		return snprintf(buf, buf_len, "%s", name);
+	case ARM_SPE_EVENTS: {
+		size_t blen = buf_len;
+
+		ret = 0;
+		ret = snprintf(buf, buf_len, "EV");
+		buf += ret;
+		blen -= ret;
+		if (payload & 0x1) {
+			ret = snprintf(buf, buf_len, " EXCEPTION-GEN");
+			buf += ret;
+			blen -= ret;
+		}
+		if (payload & 0x2) {
+			ret = snprintf(buf, buf_len, " RETIRED");
+			buf += ret;
+			blen -= ret;
+		}
+		if (payload & 0x4) {
+			ret = snprintf(buf, buf_len, " L1D-ACCESS");
+			buf += ret;
+			blen -= ret;
+		}
+		if (payload & 0x8) {
+			ret = snprintf(buf, buf_len, " L1D-REFILL");
+			buf += ret;
+			blen -= ret;
+		}
+		if (payload & 0x10) {
+			ret = snprintf(buf, buf_len, " TLB-ACCESS");
+			buf += ret;
+			blen -= ret;
+		}
+		if (payload & 0x20) {
+			ret = snprintf(buf, buf_len, " TLB-REFILL");
+			buf += ret;
+			blen -= ret;
+		}
+		if (payload & 0x40) {
+			ret = snprintf(buf, buf_len, " NOT-TAKEN");
+			buf += ret;
+			blen -= ret;
+		}
+		if (payload & 0x80) {
+			ret = snprintf(buf, buf_len, " MISPRED");
+			buf += ret;
+			blen -= ret;
+		}
+		if (index > 1) {
+			if (payload & 0x100) {
+				ret = snprintf(buf, buf_len, " LLC-ACCESS");
+				buf += ret;
+				blen -= ret;
+			}
+			if (payload & 0x200) {
+				ret = snprintf(buf, buf_len, " LLC-REFILL");
+				buf += ret;
+				blen -= ret;
+			}
+			if (payload & 0x400) {
+				ret = snprintf(buf, buf_len, " REMOTE-ACCESS");
+				buf += ret;
+				blen -= ret;
+			}
+		}
+		if (ret < 0)
+			return ret;
+		blen -= ret;
+		return buf_len - blen;
+	}
+	case ARM_SPE_OP_TYPE:
+		switch (index) {
+		case 0:	return snprintf(buf, buf_len, "%s", payload & 0x1 ?
+					"COND-SELECT" : "INSN-OTHER");
+		case 1:	{
+			size_t blen = buf_len;
+
+			if (payload & 0x1)
+				ret = snprintf(buf, buf_len, "ST");
+			else
+				ret = snprintf(buf, buf_len, "LD");
+			buf += ret;
+			blen -= ret;
+			if (payload & 0x2) {
+				if (payload & 0x4) {
+					ret = snprintf(buf, buf_len, " AT");
+					buf += ret;
+					blen -= ret;
+				}
+				if (payload & 0x8) {
+					ret = snprintf(buf, buf_len, " EXCL");
+					buf += ret;
+					blen -= ret;
+				}
+				if (payload & 0x10) {
+					ret = snprintf(buf, buf_len, " AR");
+					buf += ret;
+					blen -= ret;
+				}
+			} else if (payload & 0x4) {
+				ret = snprintf(buf, buf_len, " SIMD-FP");
+				buf += ret;
+				blen -= ret;
+			}
+			if (ret < 0)
+				return ret;
+			blen -= ret;
+			return buf_len - blen;
+		}
+		case 2:	{
+			size_t blen = buf_len;
+
+			ret = snprintf(buf, buf_len, "B");
+			buf += ret;
+			blen -= ret;
+			if (payload & 0x1) {
+				ret = snprintf(buf, buf_len, " COND");
+				buf += ret;
+				blen -= ret;
+			}
+			if (payload & 0x2) {
+				ret = snprintf(buf, buf_len, " IND");
+				buf += ret;
+				blen -= ret;
+			}
+			if (ret < 0)
+				return ret;
+			blen -= ret;
+			return buf_len - blen;
+			}
+		default: return 0;
+		}
+	case ARM_SPE_DATA_SOURCE:
+	case ARM_SPE_TIMESTAMP:
+		return snprintf(buf, buf_len, "%s %lld", name, payload);
+	case ARM_SPE_ADDRESS:
+		switch (index) {
+		case 0:
+		case 1: ns = !!(packet->payload & NS_FLAG);
+			el = (packet->payload & EL_FLAG) >> 61;
+			payload &= ~(0xffULL << 56);
+			return snprintf(buf, buf_len, "%s 0x%llx el%d ns=%d",
+				        (index == 1) ? "TGT" : "PC", payload, el, ns);
+		case 2:	return snprintf(buf, buf_len, "VA 0x%llx", payload);
+		case 3:	ns = !!(packet->payload & NS_FLAG);
+			payload &= ~(0xffULL << 56);
+			return snprintf(buf, buf_len, "PA 0x%llx ns=%d",
+					payload, ns);
+		default: return 0;
+		}
+	case ARM_SPE_CONTEXT:
+		return snprintf(buf, buf_len, "%s 0x%lx el%d", name,
+				(unsigned long)payload, index + 1);
+	case ARM_SPE_COUNTER: {
+		size_t blen = buf_len;
+
+		ret = snprintf(buf, buf_len, "%s %d ", name,
+			       (unsigned short)payload);
+		buf += ret;
+		blen -= ret;
+		switch (index) {
+		case 0:	ret = snprintf(buf, buf_len, "TOT"); break;
+		case 1:	ret = snprintf(buf, buf_len, "ISSUE"); break;
+		case 2:	ret = snprintf(buf, buf_len, "XLAT"); break;
+		default: ret = 0;
+		}
+		if (ret < 0)
+			return ret;
+		blen -= ret;
+		return buf_len - blen;
+	}
+	default:
+		break;
+	}
+
+	return snprintf(buf, buf_len, "%s 0x%llx (%d)",
+			name, payload, packet->index);
+}
+
diff --git a/tools/perf/util/arm-spe-pkt-decoder.h b/tools/perf/util/arm-spe-pkt-decoder.h
new file mode 100644
index 000000000000..f146f4143447
--- /dev/null
+++ b/tools/perf/util/arm-spe-pkt-decoder.h
@@ -0,0 +1,52 @@
+/*
+ * ARM Statistical Profiling Extensions (SPE) support
+ * Copyright (c) 2017, ARM Ltd.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ */
+
+#ifndef INCLUDE__ARM_SPE_PKT_DECODER_H__
+#define INCLUDE__ARM_SPE_PKT_DECODER_H__
+
+#include <stddef.h>
+#include <stdint.h>
+
+#define ARM_SPE_PKT_DESC_MAX		256
+
+#define ARM_SPE_NEED_MORE_BYTES		-1
+#define ARM_SPE_BAD_PACKET		-2
+
+enum arm_spe_pkt_type {
+	ARM_SPE_BAD,
+	ARM_SPE_PAD,
+	ARM_SPE_END,
+	ARM_SPE_TIMESTAMP,
+	ARM_SPE_ADDRESS,
+	ARM_SPE_COUNTER,
+	ARM_SPE_CONTEXT,
+	ARM_SPE_OP_TYPE,
+	ARM_SPE_EVENTS,
+	ARM_SPE_DATA_SOURCE,
+};
+
+struct arm_spe_pkt {
+	enum arm_spe_pkt_type	type;
+	unsigned char		index;
+	uint64_t		payload;
+};
+
+const char *arm_spe_pkt_name(enum arm_spe_pkt_type);
+
+int arm_spe_get_packet(const unsigned char *buf, size_t len,
+		       struct arm_spe_pkt *packet);
+
+int arm_spe_pkt_desc(const struct arm_spe_pkt *packet, char *buf, size_t len);
+#endif
diff --git a/tools/perf/util/arm-spe.c b/tools/perf/util/arm-spe.c
new file mode 100644
index 000000000000..f3eccd73b54a
--- /dev/null
+++ b/tools/perf/util/arm-spe.c
@@ -0,0 +1,318 @@
+/*
+ * ARM Statistical Profiling Extensions (SPE) support
+ * Copyright (c) 2017, ARM Ltd.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ */
+
+#include <endian.h>
+#include <errno.h>
+#include <byteswap.h>
+#include <inttypes.h>
+#include <linux/kernel.h>
+#include <linux/types.h>
+#include <linux/bitops.h>
+#include <linux/log2.h>
+
+#include "cpumap.h"
+#include "color.h"
+#include "evsel.h"
+#include "evlist.h"
+#include "machine.h"
+#include "session.h"
+#include "util.h"
+#include "thread.h"
+#include "debug.h"
+#include "auxtrace.h"
+#include "arm-spe.h"
+#include "arm-spe-pkt-decoder.h"
+
+struct arm_spe {
+	struct auxtrace			auxtrace;
+	struct auxtrace_queues		queues;
+	struct auxtrace_heap		heap;
+	u32				auxtrace_type;
+	struct perf_session		*session;
+	struct machine			*machine;
+	u32				pmu_type;
+};
+
+struct arm_spe_queue {
+	struct arm_spe		*spe;
+	unsigned int		queue_nr;
+	struct auxtrace_buffer	*buffer;
+	bool			on_heap;
+	bool			done;
+	pid_t			pid;
+	pid_t			tid;
+	int			cpu;
+};
+
+static void arm_spe_dump(struct arm_spe *spe __maybe_unused,
+			   unsigned char *buf, size_t len)
+{
+	struct arm_spe_pkt packet;
+	size_t pos = 0;
+	int ret, pkt_len, i;
+	char desc[ARM_SPE_PKT_DESC_MAX];
+	const char *color = PERF_COLOR_BLUE;
+
+	color_fprintf(stdout, color,
+		      ". ... ARM SPE data: size %zu bytes\n",
+		      len);
+
+	while (len) {
+		ret = arm_spe_get_packet(buf, len, &packet);
+		if (ret > 0)
+			pkt_len = ret;
+		else
+			pkt_len = 1;
+		printf(".");
+		color_fprintf(stdout, color, "  %08x: ", pos);
+		for (i = 0; i < pkt_len; i++)
+			color_fprintf(stdout, color, " %02x", buf[i]);
+		for (; i < 16; i++)
+			color_fprintf(stdout, color, "   ");
+		if (ret > 0) {
+			ret = arm_spe_pkt_desc(&packet, desc,
+					       ARM_SPE_PKT_DESC_MAX);
+			if (ret > 0)
+				color_fprintf(stdout, color, " %s\n", desc);
+		} else {
+			color_fprintf(stdout, color, " Bad packet!\n");
+		}
+		pos += pkt_len;
+		buf += pkt_len;
+		len -= pkt_len;
+	}
+}
+
+static void arm_spe_dump_event(struct arm_spe *spe, unsigned char *buf,
+				 size_t len)
+{
+	printf(".\n");
+	arm_spe_dump(spe, buf, len);
+}
+
+static struct arm_spe_queue *arm_spe_alloc_queue(struct arm_spe *spe,
+						     unsigned int queue_nr)
+{
+	struct arm_spe_queue *speq;
+
+	speq = zalloc(sizeof(struct arm_spe_queue));
+	if (!speq)
+		return NULL;
+
+	speq->spe = spe;
+	speq->queue_nr = queue_nr;
+	speq->pid = -1;
+	speq->tid = -1;
+	speq->cpu = -1;
+
+	return speq;
+}
+
+static int arm_spe_setup_queue(struct arm_spe *spe,
+				 struct auxtrace_queue *queue,
+				 unsigned int queue_nr)
+{
+	struct arm_spe_queue *speq = queue->priv;
+
+	if (list_empty(&queue->head))
+		return 0;
+
+	if (!speq) {
+		speq = arm_spe_alloc_queue(spe, queue_nr);
+		if (!speq)
+			return -ENOMEM;
+		queue->priv = speq;
+
+		if (queue->cpu != -1)
+			speq->cpu = queue->cpu;
+		speq->tid = queue->tid;
+	}
+
+	if (!speq->on_heap && !speq->buffer) {
+		int ret;
+
+		speq->buffer = auxtrace_buffer__next(queue, NULL);
+		if (!speq->buffer)
+			return 0;
+
+		ret = auxtrace_heap__add(&spe->heap, queue_nr,
+					 speq->buffer->reference);
+		if (ret)
+			return ret;
+		speq->on_heap = true;
+	}
+
+	return 0;
+}
+
+static int arm_spe_setup_queues(struct arm_spe *spe)
+{
+	unsigned int i;
+	int ret;
+
+	for (i = 0; i < spe->queues.nr_queues; i++) {
+		ret = arm_spe_setup_queue(spe, &spe->queues.queue_array[i],
+					    i);
+		if (ret)
+			return ret;
+	}
+	return 0;
+}
+
+static inline int arm_spe_update_queues(struct arm_spe *spe)
+{
+	if (spe->queues.new_data) {
+		spe->queues.new_data = false;
+		return arm_spe_setup_queues(spe);
+	}
+	return 0;
+}
+
+static int arm_spe_process_event(struct perf_session *session __maybe_unused,
+				   union perf_event *event __maybe_unused,
+				   struct perf_sample *sample __maybe_unused,
+				   struct perf_tool *tool __maybe_unused)
+{
+	return 0;
+}
+
+static int arm_spe_process_auxtrace_event(struct perf_session *session,
+					    union perf_event *event,
+					    struct perf_tool *tool __maybe_unused)
+{
+	struct arm_spe *spe = container_of(session->auxtrace, struct arm_spe,
+					     auxtrace);
+	struct auxtrace_buffer *buffer;
+	off_t data_offset;
+	int fd = perf_data_file__fd(session->file);
+	int err;
+
+	if (perf_data_file__is_pipe(session->file)) {
+		data_offset = 0;
+	} else {
+		data_offset = lseek(fd, 0, SEEK_CUR);
+		if (data_offset == -1)
+			return -errno;
+	}
+
+	err = auxtrace_queues__add_event(&spe->queues, session, event,
+					 data_offset, &buffer);
+	if (err)
+		return err;
+
+	/* Dump here now we have copied a piped trace out of the pipe */
+	if (dump_trace) {
+		if (auxtrace_buffer__get_data(buffer, fd)) {
+			arm_spe_dump_event(spe, buffer->data,
+					     buffer->size);
+			auxtrace_buffer__put_data(buffer);
+		}
+	}
+
+	return 0;
+}
+
+static int arm_spe_flush(struct perf_session *session __maybe_unused,
+			   struct perf_tool *tool __maybe_unused)
+{
+	return 0;
+}
+
+static void arm_spe_free_queue(void *priv)
+{
+	struct arm_spe_queue *speq = priv;
+
+	if (!speq)
+		return;
+	free(speq);
+}
+
+static void arm_spe_free_events(struct perf_session *session)
+{
+	struct arm_spe *spe = container_of(session->auxtrace, struct arm_spe,
+					     auxtrace);
+	struct auxtrace_queues *queues = &spe->queues;
+	unsigned int i;
+
+	for (i = 0; i < queues->nr_queues; i++) {
+		arm_spe_free_queue(queues->queue_array[i].priv);
+		queues->queue_array[i].priv = NULL;
+	}
+	auxtrace_queues__free(queues);
+}
+
+static void arm_spe_free(struct perf_session *session)
+{
+	struct arm_spe *spe = container_of(session->auxtrace, struct arm_spe,
+					     auxtrace);
+
+	auxtrace_heap__free(&spe->heap);
+	arm_spe_free_events(session);
+	session->auxtrace = NULL;
+	free(spe);
+}
+
+static const char * const arm_spe_info_fmts[] = {
+	[ARM_SPE_PMU_TYPE]		= "  PMU Type           %"PRId64"\n",
+};
+
+static void arm_spe_print_info(u64 *arr)
+{
+	if (!dump_trace)
+		return;
+
+	fprintf(stdout, arm_spe_info_fmts[ARM_SPE_PMU_TYPE], arr[ARM_SPE_PMU_TYPE]);
+}
+
+int arm_spe_process_auxtrace_info(union perf_event *event,
+				    struct perf_session *session)
+{
+	struct auxtrace_info_event *auxtrace_info = &event->auxtrace_info;
+	size_t min_sz = sizeof(u64) * ARM_SPE_PMU_TYPE;
+	struct arm_spe *spe;
+	int err;
+
+	if (auxtrace_info->header.size < sizeof(struct auxtrace_info_event) +
+					min_sz)
+		return -EINVAL;
+
+	spe = zalloc(sizeof(struct arm_spe));
+	if (!spe)
+		return -ENOMEM;
+
+	err = auxtrace_queues__init(&spe->queues);
+	if (err)
+		goto err_free;
+
+	spe->session = session;
+	spe->machine = &session->machines.host; /* No kvm support */
+	spe->auxtrace_type = auxtrace_info->type;
+	spe->pmu_type = auxtrace_info->priv[ARM_SPE_PMU_TYPE];
+
+	spe->auxtrace.process_event = arm_spe_process_event;
+	spe->auxtrace.process_auxtrace_event = arm_spe_process_auxtrace_event;
+	spe->auxtrace.flush_events = arm_spe_flush;
+	spe->auxtrace.free_events = arm_spe_free_events;
+	spe->auxtrace.free = arm_spe_free;
+	session->auxtrace = &spe->auxtrace;
+
+	arm_spe_print_info(&auxtrace_info->priv[0]);
+
+	return 0;
+
+err_free:
+	free(spe);
+	return err;
+}
diff --git a/tools/perf/util/arm-spe.h b/tools/perf/util/arm-spe.h
new file mode 100644
index 000000000000..80752b20d850
--- /dev/null
+++ b/tools/perf/util/arm-spe.h
@@ -0,0 +1,42 @@
+/*
+ * ARM Statistical Profiling Extensions (SPE) support
+ * Copyright (c) 2017, ARM Ltd.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ */
+
+#ifndef INCLUDE__PERF_ARM_SPE_H__
+#define INCLUDE__PERF_ARM_SPE_H__
+
+#define ARM_SPE_PMU_NAME "arm_spe_"
+
+enum {
+	ARM_SPE_PMU_TYPE,
+	ARM_SPE_PER_CPU_MMAPS,
+	ARM_SPE_AUXTRACE_PRIV_MAX,
+};
+
+#define ARM_SPE_AUXTRACE_PRIV_SIZE (ARM_SPE_AUXTRACE_PRIV_MAX * sizeof(u64))
+
+struct auxtrace_record;
+struct perf_tool;
+union perf_event;
+struct perf_session;
+struct perf_pmu;
+
+struct auxtrace_record *arm_spe_recording_init(int *err,
+					       struct perf_pmu *arm_spe_pmu);
+
+int arm_spe_process_auxtrace_info(union perf_event *event,
+				  struct perf_session *session);
+
+struct perf_event_attr *arm_spe_pmu_default_config(struct perf_pmu *arm_spe_pmu);
+#endif
diff --git a/tools/perf/util/auxtrace.c b/tools/perf/util/auxtrace.c
index 5547457566a7..92da3981a761 100644
--- a/tools/perf/util/auxtrace.c
+++ b/tools/perf/util/auxtrace.c
@@ -57,6 +57,7 @@
 
 #include "intel-pt.h"
 #include "intel-bts.h"
+#include "arm-spe.h"
 
 #include "sane_ctype.h"
 #include "symbol/kallsyms.h"
@@ -913,6 +914,8 @@ int perf_event__process_auxtrace_info(struct perf_tool *tool __maybe_unused,
 		return intel_pt_process_auxtrace_info(event, session);
 	case PERF_AUXTRACE_INTEL_BTS:
 		return intel_bts_process_auxtrace_info(event, session);
+	case PERF_AUXTRACE_ARM_SPE:
+		return arm_spe_process_auxtrace_info(event, session);
 	case PERF_AUXTRACE_CS_ETM:
 	case PERF_AUXTRACE_UNKNOWN:
 	default:
diff --git a/tools/perf/util/auxtrace.h b/tools/perf/util/auxtrace.h
index 33b5e6cdf38c..fa1c8deffac4 100644
--- a/tools/perf/util/auxtrace.h
+++ b/tools/perf/util/auxtrace.h
@@ -43,6 +43,7 @@ enum auxtrace_type {
 	PERF_AUXTRACE_INTEL_PT,
 	PERF_AUXTRACE_INTEL_BTS,
 	PERF_AUXTRACE_CS_ETM,
+	PERF_AUXTRACE_ARM_SPE,
 };
 
 enum itrace_period_type {
-- 
2.14.2




More information about the linux-arm-kernel mailing list