[PATCH V6 00/25] perf tools: Introduce an abstraction for AUX Area and Instruction Tracing

[PATCH V6 00/25] perf tools: Introduce an abstraction for AUX Area and Instruction Tracing

[Adrian Hunter]

Hi

Here is V6 of some more preparatory patches for Intel PT
that introduce an abstraction for using the AUX area and
Instruction tracing.

The master branch of the tree:

	git://git.infradead.org/users/ahunter/linux-perf.git

contains these patches and working Intel PT and Intel BTS.
Intel BTS can be used on most recent Intel CPUs. Intel PT
is available on Broadwell.

Examples:

	Trace 'ls' with Intel BTS userspace only

	perf record --per-thread -e intel_bts//u ls
	perf report
	perf script

	Trace 'ls' with Intel BTS kernel and userspace

	~/libexec/perf-core/perf-with-kcore record bts-ls --per-thread -e intel_bts// -- ls
	~/libexec/perf-core/perf-with-kcore report bts-ls
	~/libexec/perf-core/perf-with-kcore script bts-ls

	Trace 'ls' with Intel PT userspace only

	perf record -e intel_pt//u ls
	perf report
	perf script

	Trace 'ls' with Intel PT kernel and userspace

	~/libexec/perf-core/perf-with-kcore record pt-ls -e intel_pt// -- ls
	~/libexec/perf-core/perf-with-kcore report pt-ls
	~/libexec/perf-core/perf-with-kcore script pt-ls

Changes in V6:

	Renamed itrace to auxtrace to better reflect the use of the
	AUX area. In some cases itrace was retained where it seemed
	to relate exclusively to Instruction Tracing.
	
	Print warnings for separate auxtrace error types

	Re-based on Arnaldo's perf/core branch:

		107eb964d8c04417e8bce9e9ec2ed61d9261aec6
		Merge tag 'perf-core-for-mingo' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/core

Changes in V5:

	perf tools: Add a user event for Instruction Tracing errors
		Print error count from stats->nr_events

	perf session: Add hooks to allow transparent decoding of Instruction Tracing data
		Don't count Instruction Tracing errors (using stats->nr_events instead)

	perf itrace: Add helpers for Instruction Tracing errors
		Don't count Instruction Tracing errors (using stats->nr_events instead)

	perf itrace: Add option to synthesize events for transactions
		New patch

	perf script: Add field option 'flags' to print sample flags
		New patch

	Re-based on tip perf/core branch:

		94ac003b665fc04f13a7ab3b2be896b9b9503451
		Merge tag 'perf-core-for-mingo' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/core

Changes in V4:

	perf tools: Add build option NO_ITRACE to exclude Instruction Tracing
		New patch

	Re-based on Arnaldo's perf/core branch:

		3dd417d4010c8e141b0f32121cdc8d82aa4a9c6a
		perf tools: Remove some unused functions from color.c

Changes in V3:

	perf tools: Add support for Instruction Trace recording
		Added evsel as a parameter to itrace_record__init

	perf record: Add basic Instruction Tracing support
		Moved the call to itrace_record__init after parse
		options so that evsel could be passed and the
		selected events used to determine what kind of
		Instruction Tracing to use e.g. Intel PT vs BTS

	Re-based on Arnaldo's perf/core branch:

		41e950c033b7df997d4b38653efe6554be9b96a7
		Merge tag 'perf-core-for-mingo' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/urgent

Changes in V2:

	Dropped patches already applied.

	Re-based on Arnaldo's perf/core branch:

		a84808083688d82d7f1e5786ccf5df0ff7d448cb
		perf tools: Only override the default :tid comm entry


The abstraction has two separate aspects:
	1. recording AUX area data
	2. processing AUX area data

Recording consists of mmapping a separate buffer and copying
the data into the perf.data file.  The buffer is an AUX area
buffer although the details of the AUX area are not implemented
because the kernel support is pending.  The data is written
preceded by a new user event PERF_RECORD_AUXTRACE.  The data is
too big to fit in the event but follows immediately afterward.
Session processing has to skip to get to the next event header
in a similar fashion to the existing PERF_RECORD_HEADER_TRACING_DATA
event.  The main recording patches are:

      perf evlist: Add initial support for mmapping an AUX area buffer
      perf tools: Add user events for AUX area tracing
      perf tools: Add support for AUX area recording
      perf record: Add basic AUX area tracing support

Processing consists of providing hooks in session processing
to enable a decoder to see all the events and deliver synthesized
events transparently into the event stream.  The main processing
patch is:

      perf session: Add hooks to allow transparent decoding of AUX area tracing data


Adrian Hunter (25):
      perf header: Add AUX area tracing feature
      perf evlist: Add initial support for mmapping an AUX area buffer
      perf tools: Add user events for AUX area tracing
      perf tools: Add support for AUX area recording
      perf record: Add basic AUX area tracing support
      perf record: Extend -m option for AUX area tracing mmap pages
      perf tools: Add a user event for AUX area tracing errors
      perf session: Add hooks to allow transparent decoding of AUX area tracing data
      perf session: Add instruction tracing options
      perf auxtrace: Add helpers for AUX area tracing errors
      perf auxtrace: Add helpers for queuing AUX area tracing data
      perf auxtrace: Add a heap for sorting AUX area tracing queues
      perf auxtrace: Add processing for AUX area tracing events
      perf auxtrace: Add a hashtable for caching
      perf tools: Add member to struct dso for an instruction cache
      perf script: Add Instruction Tracing support
      perf script: Always allow fields 'addr' and 'cpu' for auxtrace
      perf report: Add Instruction Tracing support
      perf inject: Re-pipe AUX area tracing events
      perf inject: Add Instruction Tracing support
      perf tools: Add AUX area tracing index
      perf tools: Hit all build ids when AUX area tracing
      perf tools: Add build option NO_AUXTRACE to exclude AUX area tracing
      perf auxtrace: Add option to synthesize events for transactions
      perf script: Add field option 'flags' to print sample flags

 tools/perf/Documentation/perf-inject.txt |   27 +
 tools/perf/Documentation/perf-record.txt |    2 +
 tools/perf/Documentation/perf-report.txt |   29 +
 tools/perf/Documentation/perf-script.txt |   38 +-
 tools/perf/Makefile.perf                 |    2 +
 tools/perf/builtin-buildid-list.c        |    9 +
 tools/perf/builtin-inject.c              |  172 +++-
 tools/perf/builtin-record.c              |  167 +++-
 tools/perf/builtin-report.c              |   11 +
 tools/perf/builtin-script.c              |   74 +-
 tools/perf/config/Makefile               |    5 +
 tools/perf/perf.h                        |    2 +
 tools/perf/tests/make                    |    2 +
 tools/perf/util/Build                    |    1 +
 tools/perf/util/auxtrace.c               | 1293 ++++++++++++++++++++++++++++++
 tools/perf/util/auxtrace.h               |  575 +++++++++++++
 tools/perf/util/dso.c                    |    2 +
 tools/perf/util/dso.h                    |    3 +
 tools/perf/util/event.c                  |    3 +
 tools/perf/util/event.h                  |   46 ++
 tools/perf/util/evlist.c                 |   71 +-
 tools/perf/util/evlist.h                 |    6 +
 tools/perf/util/header.c                 |   37 +
 tools/perf/util/header.h                 |    1 +
 tools/perf/util/record.c                 |   11 +-
 tools/perf/util/session.c                |  169 +++-
 tools/perf/util/session.h                |    6 +
 tools/perf/util/tool.h                   |   10 +-
 28 files changed, 2720 insertions(+), 54 deletions(-)
 create mode 100644 tools/perf/util/auxtrace.c
 create mode 100644 tools/perf/util/auxtrace.h


Regards
Adrian

[PATCH V6 01/25] perf header: Add AUX area tracing feature

[Adrian Hunter]

Add a feature to indicate that a perf.data file
contains AUX area data.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/util/header.c | 14 ++++++++++++++
 tools/perf/util/header.h |  1 +
 2 files changed, 15 insertions(+)

diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c
index 1f407f7..9b06b5e 100644
--- a/tools/perf/util/header.c
+++ b/tools/perf/util/header.c
@@ -869,6 +869,13 @@ static int write_branch_stack(int fd __maybe_unused,
 	return 0;
 }
 
+static int write_auxtrace(int fd __maybe_unused,
+			  struct perf_header *h __maybe_unused,
+			  struct perf_evlist *evlist __maybe_unused)
+{
+	return 0;
+}
+
 static void print_hostname(struct perf_header *ph, int fd __maybe_unused,
 			   FILE *fp)
 {
@@ -1163,6 +1170,12 @@ static void print_branch_stack(struct perf_header *ph __maybe_unused,
 	fprintf(fp, "# contains samples with branch stack\n");
 }
 
+static void print_auxtrace(struct perf_header *ph __maybe_unused,
+			   int fd __maybe_unused, FILE *fp)
+{
+	fprintf(fp, "# contains AUX area data (e.g. instruction trace)\n");
+}
+
 static void print_pmu_mappings(struct perf_header *ph, int fd __maybe_unused,
 			       FILE *fp)
 {
@@ -1873,6 +1886,7 @@ static const struct feature_ops feat_ops[HEADER_LAST_FEATURE] = {
 	FEAT_OPA(HEADER_BRANCH_STACK,	branch_stack),
 	FEAT_OPP(HEADER_PMU_MAPPINGS,	pmu_mappings),
 	FEAT_OPP(HEADER_GROUP_DESC,	group_desc),
+	FEAT_OPA(HEADER_AUXTRACE,	auxtrace),
 };
 
 struct header_print_data {
diff --git a/tools/perf/util/header.h b/tools/perf/util/header.h
index 3bb90ac..d4d5796 100644
--- a/tools/perf/util/header.h
+++ b/tools/perf/util/header.h
@@ -30,6 +30,7 @@ enum {
 	HEADER_BRANCH_STACK,
 	HEADER_PMU_MAPPINGS,
 	HEADER_GROUP_DESC,
+	HEADER_AUXTRACE,
 	HEADER_LAST_FEATURE,
 	HEADER_FEAT_BITS	= 256,
 };
-- 
1.9.1

[PATCH V6 02/25] perf evlist: Add initial support for mmapping an AUX area buffer

[Adrian Hunter]

This patch anticipates the addition to the kernel
of an "aux" buffer that can be mmapped separately
from the perf-events buffer.

The expectation is that this buffer can be configured
to contain hardware-produced trace information.
The first implementation will support Intel BTS and
Intel PT.

One auxtrace buffer is mmapped per perf-events buffer.
If the requested auxtrace buffer size is zero, which
it will be until further support is added, then
no auxtrace mmapping is attempted.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/util/Build      |   1 +
 tools/perf/util/auxtrace.c | 103 +++++++++++++++++++++++++++++++++++++++++++++
 tools/perf/util/auxtrace.h |  97 ++++++++++++++++++++++++++++++++++++++++++
 tools/perf/util/evlist.c   |  61 +++++++++++++++++++++++++--
 tools/perf/util/evlist.h   |   5 +++
 5 files changed, 264 insertions(+), 3 deletions(-)
 create mode 100644 tools/perf/util/auxtrace.c
 create mode 100644 tools/perf/util/auxtrace.h

diff --git a/tools/perf/util/Build b/tools/perf/util/Build
index 972a6e0..bb55e26 100644
--- a/tools/perf/util/Build
+++ b/tools/perf/util/Build
@@ -74,6 +74,7 @@ libperf-y += data.o
 libperf-$(CONFIG_X86) += tsc.o
 libperf-y += cloexec.o
 libperf-y += thread-stack.o
+libperf-y += auxtrace.o
 
 libperf-$(CONFIG_LIBELF) += symbol-elf.o
 libperf-$(CONFIG_LIBELF) += probe-event.o
diff --git a/tools/perf/util/auxtrace.c b/tools/perf/util/auxtrace.c
new file mode 100644
index 0000000..4b17a80
--- /dev/null
+++ b/tools/perf/util/auxtrace.c
@@ -0,0 +1,103 @@
+/*
+ * auxtrace.c: AUX area trace support
+ * Copyright (c) 2013-2015, Intel Corporation.
+ *
+ * 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 <sys/types.h>
+#include <sys/mman.h>
+#include <stdbool.h>
+
+#include <linux/kernel.h>
+#include <linux/perf_event.h>
+#include <linux/types.h>
+#include <linux/bitops.h>
+#include <linux/log2.h>
+
+#include "../perf.h"
+#include "util.h"
+#include "evlist.h"
+#include "cpumap.h"
+#include "thread_map.h"
+#include "auxtrace.h"
+
+int auxtrace_mmap__mmap(struct auxtrace_mmap *mm,
+			struct auxtrace_mmap_params *mp,
+			void *userpg, int fd)
+{
+#if BITS_PER_LONG != 64 && !defined(HAVE_SYNC_COMPARE_AND_SWAP_SUPPORT)
+	pr_err("Cannot use AUX area tracing mmaps\n");
+	return -1;
+#endif
+
+	mm->userpg = userpg;
+	mm->mask = mp->mask;
+	mm->len = mp->len;
+	mm->prev = 0;
+	mm->idx = mp->idx;
+	mm->tid = mp->tid;
+	mm->cpu = mp->cpu;
+
+	if (!mp->len) {
+		mm->base = NULL;
+		return 0;
+	}
+
+	mm->base = mmap(NULL, mp->len, mp->prot, MAP_SHARED, fd, mp->offset);
+	if (mm->base == MAP_FAILED) {
+		pr_debug2("failed to mmap AUX area\n");
+		mm->base = NULL;
+		return -1;
+	}
+
+	return 0;
+}
+
+void auxtrace_mmap__munmap(struct auxtrace_mmap *mm)
+{
+	if (mm->base)
+		munmap(mm->base, mm->len);
+}
+
+void auxtrace_mmap_params__init(struct auxtrace_mmap_params *mp,
+				off_t auxtrace_offset,
+				unsigned int auxtrace_pages,
+				bool auxtrace_overwrite)
+{
+	if (auxtrace_pages) {
+		mp->offset = auxtrace_offset;
+		mp->len = auxtrace_pages * (size_t)page_size;
+		mp->mask = is_power_of_2(mp->len) ? mp->len - 1 : 0;
+		mp->prot = PROT_READ | (auxtrace_overwrite ? 0 : PROT_WRITE);
+		pr_debug2("AUX area mmap length %zu\n", mp->len);
+	} else {
+		mp->len = 0;
+	}
+}
+
+void auxtrace_mmap_params__set_idx(struct auxtrace_mmap_params *mp,
+				   struct perf_evlist *evlist, int idx,
+				   bool per_cpu)
+{
+	mp->idx = idx;
+
+	if (per_cpu) {
+		mp->cpu = evlist->cpus->map[idx];
+		if (evlist->threads)
+			mp->tid = evlist->threads->map[0];
+		else
+			mp->tid = -1;
+	} else {
+		mp->cpu = -1;
+		mp->tid = evlist->threads->map[idx];
+	}
+}
diff --git a/tools/perf/util/auxtrace.h b/tools/perf/util/auxtrace.h
new file mode 100644
index 0000000..735ca2a
--- /dev/null
+++ b/tools/perf/util/auxtrace.h
@@ -0,0 +1,97 @@
+/*
+ * auxtrace.h: AUX area trace support
+ * Copyright (c) 2013-2015, Intel Corporation.
+ *
+ * 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 __PERF_AUXTRACE_H
+#define __PERF_AUXTRACE_H
+
+#include <sys/types.h>
+#include <stdbool.h>
+
+#include <linux/perf_event.h>
+#include <linux/types.h>
+
+#include "../perf.h"
+
+struct perf_evlist;
+
+/**
+ * struct auxtrace_mmap - records an mmap of the auxtrace buffer.
+ * @base: address of mapped area
+ * @userpg: pointer to buffer's perf_event_mmap_page
+ * @mask: %0 if @len is not a power of two, otherwise (@len - %1)
+ * @len: size of mapped area
+ * @prev: previous aux_head
+ * @idx: index of this mmap
+ * @tid: tid for a per-thread mmap (also set if there is only 1 tid on a per-cpu
+ *       mmap) otherwise %0
+ * @cpu: cpu number for a per-cpu mmap otherwise %-1
+ */
+struct auxtrace_mmap {
+	void		*base;
+	void		*userpg;
+	size_t		mask;
+	size_t		len;
+	u64		prev;
+	int		idx;
+	pid_t		tid;
+	int		cpu;
+};
+
+/**
+ * struct auxtrace_mmap_params - parameters to set up struct auxtrace_mmap.
+ * @mask: %0 if @len is not a power of two, otherwise (@len - %1)
+ * @offset: file offset of mapped area
+ * @len: size of mapped area
+ * @prot: mmap memory protection
+ * @idx: index of this mmap
+ * @tid: tid for a per-thread mmap (also set if there is only 1 tid on a per-cpu
+ *       mmap) otherwise %0
+ * @cpu: cpu number for a per-cpu mmap otherwise %-1
+ */
+struct auxtrace_mmap_params {
+	size_t		mask;
+	off_t		offset;
+	size_t		len;
+	int		prot;
+	int		idx;
+	pid_t		tid;
+	int		cpu;
+};
+
+static inline u64 auxtrace_mmap__read_head(struct auxtrace_mmap *mm __maybe_unused)
+{
+	/* Not yet implemented */
+	return 0;
+}
+
+static inline void auxtrace_mmap__write_tail(struct auxtrace_mmap *mm __maybe_unused,
+					     u64 tail __maybe_unused)
+{
+	/* Not yet implemented */
+}
+
+int auxtrace_mmap__mmap(struct auxtrace_mmap *mm,
+			struct auxtrace_mmap_params *mp,
+			void *userpg, int fd);
+void auxtrace_mmap__munmap(struct auxtrace_mmap *mm);
+void auxtrace_mmap_params__init(struct auxtrace_mmap_params *mp,
+				off_t auxtrace_offset,
+				unsigned int auxtrace_pages,
+				bool auxtrace_overwrite);
+void auxtrace_mmap_params__set_idx(struct auxtrace_mmap_params *mp,
+				   struct perf_evlist *evlist, int idx,
+				   bool per_cpu);
+
+#endif
diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c
index 8d0b623..3cc7815 100644
--- a/tools/perf/util/evlist.c
+++ b/tools/perf/util/evlist.c
@@ -725,6 +725,34 @@ void perf_evlist__mmap_consume(struct perf_evlist *evlist, int idx)
 		perf_evlist__mmap_put(evlist, idx);
 }
 
+int __weak auxtrace_mmap__mmap(struct auxtrace_mmap *mm __maybe_unused,
+			       struct auxtrace_mmap_params *mp __maybe_unused,
+			       void *userpg __maybe_unused,
+			       int fd __maybe_unused)
+{
+	return 0;
+}
+
+void __weak auxtrace_mmap__munmap(struct auxtrace_mmap *mm __maybe_unused)
+{
+}
+
+void __weak auxtrace_mmap_params__init(
+			struct auxtrace_mmap_params *mp __maybe_unused,
+			off_t auxtrace_offset __maybe_unused,
+			unsigned int auxtrace_pages __maybe_unused,
+			bool auxtrace_overwrite __maybe_unused)
+{
+}
+
+void __weak auxtrace_mmap_params__set_idx(
+			struct auxtrace_mmap_params *mp __maybe_unused,
+			struct perf_evlist *evlist __maybe_unused,
+			int idx __maybe_unused,
+			bool per_cpu __maybe_unused)
+{
+}
+
 static void __perf_evlist__munmap(struct perf_evlist *evlist, int idx)
 {
 	if (evlist->mmap[idx].base != NULL) {
@@ -732,6 +760,7 @@ static void __perf_evlist__munmap(struct perf_evlist *evlist, int idx)
 		evlist->mmap[idx].base = NULL;
 		evlist->mmap[idx].refcnt = 0;
 	}
+	auxtrace_mmap__munmap(&evlist->mmap[idx].auxtrace_mmap);
 }
 
 void perf_evlist__munmap(struct perf_evlist *evlist)
@@ -759,6 +788,7 @@ static int perf_evlist__alloc_mmap(struct perf_evlist *evlist)
 struct mmap_params {
 	int prot;
 	int mask;
+	struct auxtrace_mmap_params auxtrace_mp;
 };
 
 static int __perf_evlist__mmap(struct perf_evlist *evlist, int idx,
@@ -789,6 +819,10 @@ static int __perf_evlist__mmap(struct perf_evlist *evlist, int idx,
 		return -1;
 	}
 
+	if (auxtrace_mmap__mmap(&evlist->mmap[idx].auxtrace_mmap,
+				&mp->auxtrace_mp, evlist->mmap[idx].base, fd))
+		return -1;
+
 	return 0;
 }
 
@@ -853,6 +887,9 @@ static int perf_evlist__mmap_per_cpu(struct perf_evlist *evlist,
 	for (cpu = 0; cpu < nr_cpus; cpu++) {
 		int output = -1;
 
+		auxtrace_mmap_params__set_idx(&mp->auxtrace_mp, evlist, cpu,
+					      true);
+
 		for (thread = 0; thread < nr_threads; thread++) {
 			if (perf_evlist__mmap_per_evsel(evlist, cpu, mp, cpu,
 							thread, &output))
@@ -878,6 +915,9 @@ static int perf_evlist__mmap_per_thread(struct perf_evlist *evlist,
 	for (thread = 0; thread < nr_threads; thread++) {
 		int output = -1;
 
+		auxtrace_mmap_params__set_idx(&mp->auxtrace_mp, evlist, thread,
+					      false);
+
 		if (perf_evlist__mmap_per_evsel(evlist, thread, mp, 0, thread,
 						&output))
 			goto out_unmap;
@@ -981,19 +1021,25 @@ int perf_evlist__parse_mmap_pages(const struct option *opt, const char *str,
 }
 
 /**
- * perf_evlist__mmap - Create mmaps to receive events.
+ * perf_evlist__mmap_ex - Create mmaps to receive events.
  * @evlist: list of events
  * @pages: map length in pages
  * @overwrite: overwrite older events?
+ * @auxtrace_pages - auxtrace map length in pages
+ * @auxtrace_overwrite - overwrite older auxtrace data?
  *
  * If @overwrite is %false the user needs to signal event consumption using
  * perf_mmap__write_tail().  Using perf_evlist__mmap_read() does this
  * automatically.
  *
+ * Similarly, if @auxtrace_overwrite is %false the user needs to signal data
+ * consumption using auxtrace_mmap__write_tail().
+ *
  * Return: %0 on success, negative error code otherwise.
  */
-int perf_evlist__mmap(struct perf_evlist *evlist, unsigned int pages,
-		      bool overwrite)
+int perf_evlist__mmap_ex(struct perf_evlist *evlist, unsigned int pages,
+			 bool overwrite, unsigned int auxtrace_pages,
+			 bool auxtrace_overwrite)
 {
 	struct perf_evsel *evsel;
 	const struct cpu_map *cpus = evlist->cpus;
@@ -1013,6 +1059,9 @@ int perf_evlist__mmap(struct perf_evlist *evlist, unsigned int pages,
 	pr_debug("mmap size %zuB\n", evlist->mmap_len);
 	mp.mask = evlist->mmap_len - page_size - 1;
 
+	auxtrace_mmap_params__init(&mp.auxtrace_mp, evlist->mmap_len,
+				   auxtrace_pages, auxtrace_overwrite);
+
 	evlist__for_each(evlist, evsel) {
 		if ((evsel->attr.read_format & PERF_FORMAT_ID) &&
 		    evsel->sample_id == NULL &&
@@ -1026,6 +1075,12 @@ int perf_evlist__mmap(struct perf_evlist *evlist, unsigned int pages,
 	return perf_evlist__mmap_per_cpu(evlist, &mp);
 }
 
+int perf_evlist__mmap(struct perf_evlist *evlist, unsigned int pages,
+		      bool overwrite)
+{
+	return perf_evlist__mmap_ex(evlist, pages, overwrite, 0, false);
+}
+
 int perf_evlist__create_maps(struct perf_evlist *evlist, struct target *target)
 {
 	evlist->threads = thread_map__new_str(target->pid, target->tid,
diff --git a/tools/perf/util/evlist.h b/tools/perf/util/evlist.h
index f07c984..fbb8063 100644
--- a/tools/perf/util/evlist.h
+++ b/tools/perf/util/evlist.h
@@ -8,6 +8,7 @@
 #include "event.h"
 #include "evsel.h"
 #include "util.h"
+#include "auxtrace.h"
 #include <unistd.h>
 
 struct pollfd;
@@ -28,6 +29,7 @@ struct perf_mmap {
 	int		 mask;
 	int		 refcnt;
 	unsigned int	 prev;
+	struct auxtrace_mmap auxtrace_mmap;
 	char		 event_copy[PERF_SAMPLE_MAX_SIZE] __attribute__((aligned(8)));
 };
 
@@ -126,6 +128,9 @@ int perf_evlist__parse_mmap_pages(const struct option *opt,
 				  const char *str,
 				  int unset);
 
+int perf_evlist__mmap_ex(struct perf_evlist *evlist, unsigned int pages,
+			 bool overwrite, unsigned int auxtrace_pages,
+			 bool auxtrace_overwrite);
 int perf_evlist__mmap(struct perf_evlist *evlist, unsigned int pages,
 		      bool overwrite);
 void perf_evlist__munmap(struct perf_evlist *evlist);
-- 
1.9.1

[PATCH V6 03/25] perf tools: Add user events for AUX area tracing

[Adrian Hunter]

Add two user events for AUX area tracing.

PERF_RECORD_AUXTRACE_INFO contains metadata,
consisting primarily the type of the
AUX area tracing data plus some amount
of architecture-specific information.
There should be only one
PERF_RECORD_AUXTRACE_INFO event.

PERF_RECORD_AUXTRACE identifies AUX area
tracing data copied from the mmapped
AUX area tracing region.  The actual
data is not part of the event but
immediately follows it.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/util/event.c   |  2 ++
 tools/perf/util/event.h   | 22 +++++++++++++++
 tools/perf/util/session.c | 69 +++++++++++++++++++++++++++++++++++++++++++++++
 tools/perf/util/tool.h    |  9 ++++++-
 4 files changed, 101 insertions(+), 1 deletion(-)

diff --git a/tools/perf/util/event.c b/tools/perf/util/event.c
index d5efa50..cbf3775 100644
--- a/tools/perf/util/event.c
+++ b/tools/perf/util/event.c
@@ -29,6 +29,8 @@ static const char *perf_event__names[] = {
 	[PERF_RECORD_HEADER_BUILD_ID]		= "BUILD_ID",
 	[PERF_RECORD_FINISHED_ROUND]		= "FINISHED_ROUND",
 	[PERF_RECORD_ID_INDEX]			= "ID_INDEX",
+	[PERF_RECORD_AUXTRACE_INFO]		= "AUXTRACE_INFO",
+	[PERF_RECORD_AUXTRACE]			= "AUXTRACE",
 };
 
 const char *perf_event__name(unsigned int id)
diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h
index c4ffe2b..5dc9329 100644
--- a/tools/perf/util/event.h
+++ b/tools/perf/util/event.h
@@ -215,6 +215,8 @@ enum perf_user_event_type { /* above any possible kernel type */
 	PERF_RECORD_HEADER_BUILD_ID		= 67,
 	PERF_RECORD_FINISHED_ROUND		= 68,
 	PERF_RECORD_ID_INDEX			= 69,
+	PERF_RECORD_AUXTRACE_INFO		= 70,
+	PERF_RECORD_AUXTRACE			= 71,
 	PERF_RECORD_HEADER_MAX
 };
 
@@ -281,6 +283,24 @@ struct id_index_event {
 	struct id_index_entry entries[0];
 };
 
+struct auxtrace_info_event {
+	struct perf_event_header header;
+	u32 type;
+	u32 reserved__; /* For alignment */
+	u64 priv[];
+};
+
+struct auxtrace_event {
+	struct perf_event_header header;
+	u64 size;
+	u64 offset;
+	u64 reference;
+	u32 idx;
+	u32 tid;
+	u32 cpu;
+	u32 reserved__; /* For alignment */
+};
+
 union perf_event {
 	struct perf_event_header	header;
 	struct mmap_event		mmap;
@@ -296,6 +316,8 @@ union perf_event {
 	struct tracing_data_event	tracing_data;
 	struct build_id_event		build_id;
 	struct id_index_event		id_index;
+	struct auxtrace_info_event	auxtrace_info;
+	struct auxtrace_event		auxtrace;
 };
 
 void perf_event__print_totals(void);
diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c
index adf0740..a8c67cb 100644
--- a/tools/perf/util/session.c
+++ b/tools/perf/util/session.c
@@ -254,6 +254,40 @@ static int process_id_index_stub(struct perf_tool *tool __maybe_unused,
 	return 0;
 }
 
+static int process_event_auxtrace_info_stub(struct perf_tool *tool __maybe_unused,
+				union perf_event *event __maybe_unused,
+				struct perf_session *session __maybe_unused)
+{
+	dump_printf(": unhandled!\n");
+	return 0;
+}
+
+static int skipn(int fd, off_t n)
+{
+	char buf[4096];
+	ssize_t ret;
+
+	while (n > 0) {
+		ret = read(fd, buf, MIN(n, sizeof(buf)));
+		if (ret <= 0)
+			return ret;
+		n -= ret;
+	}
+
+	return 0;
+}
+
+static s64 process_event_auxtrace_stub(struct perf_tool *tool __maybe_unused,
+				       union perf_event *event,
+				       struct perf_session *session
+				       __maybe_unused)
+{
+	dump_printf(": unhandled!\n");
+	if (perf_data_file__is_pipe(session->file))
+		skipn(perf_data_file__fd(session->file), event->auxtrace.size);
+	return event->auxtrace.size;
+}
+
 void perf_tool__fill_defaults(struct perf_tool *tool)
 {
 	if (tool->sample == NULL)
@@ -290,6 +324,10 @@ void perf_tool__fill_defaults(struct perf_tool *tool)
 	}
 	if (tool->id_index == NULL)
 		tool->id_index = process_id_index_stub;
+	if (tool->auxtrace_info == NULL)
+		tool->auxtrace_info = process_event_auxtrace_info_stub;
+	if (tool->auxtrace == NULL)
+		tool->auxtrace = process_event_auxtrace_stub;
 }
 
 static void swap_sample_id_all(union perf_event *event, void *data)
@@ -470,6 +508,29 @@ static void perf_event__tracing_data_swap(union perf_event *event,
 	event->tracing_data.size = bswap_32(event->tracing_data.size);
 }
 
+static void perf_event__auxtrace_info_swap(union perf_event *event,
+					   bool sample_id_all __maybe_unused)
+{
+	size_t size;
+
+	event->auxtrace_info.type = bswap_32(event->auxtrace_info.type);
+
+	size = event->header.size;
+	size -= (void *)&event->auxtrace_info.priv - (void *)event;
+	mem_bswap_64(event->auxtrace_info.priv, size);
+}
+
+static void perf_event__auxtrace_swap(union perf_event *event,
+				      bool sample_id_all __maybe_unused)
+{
+	event->auxtrace.size      = bswap_64(event->auxtrace.size);
+	event->auxtrace.offset    = bswap_64(event->auxtrace.offset);
+	event->auxtrace.reference = bswap_64(event->auxtrace.reference);
+	event->auxtrace.idx       = bswap_32(event->auxtrace.idx);
+	event->auxtrace.tid       = bswap_32(event->auxtrace.tid);
+	event->auxtrace.cpu       = bswap_32(event->auxtrace.cpu);
+}
+
 typedef void (*perf_event__swap_op)(union perf_event *event,
 				    bool sample_id_all);
 
@@ -489,6 +550,8 @@ static perf_event__swap_op perf_event__swap_ops[] = {
 	[PERF_RECORD_HEADER_TRACING_DATA] = perf_event__tracing_data_swap,
 	[PERF_RECORD_HEADER_BUILD_ID]	  = NULL,
 	[PERF_RECORD_ID_INDEX]		  = perf_event__all64_swap,
+	[PERF_RECORD_AUXTRACE_INFO]	  = perf_event__auxtrace_info_swap,
+	[PERF_RECORD_AUXTRACE]		  = perf_event__auxtrace_swap,
 	[PERF_RECORD_HEADER_MAX]	  = NULL,
 };
 
@@ -972,6 +1035,12 @@ static s64 perf_session__process_user_event(struct perf_session *session,
 		return tool->finished_round(tool, event, oe);
 	case PERF_RECORD_ID_INDEX:
 		return tool->id_index(tool, event, session);
+	case PERF_RECORD_AUXTRACE_INFO:
+		return tool->auxtrace_info(tool, event, session);
+	case PERF_RECORD_AUXTRACE:
+		/* setup for reading amidst mmap */
+		lseek(fd, file_offset + event->header.size, SEEK_SET);
+		return tool->auxtrace(tool, event, session);
 	default:
 		return -EINVAL;
 	}
diff --git a/tools/perf/util/tool.h b/tools/perf/util/tool.h
index 51d9e56..0146f8e 100644
--- a/tools/perf/util/tool.h
+++ b/tools/perf/util/tool.h
@@ -3,6 +3,8 @@
 
 #include <stdbool.h>
 
+#include <linux/types.h>
+
 struct perf_session;
 union perf_event;
 struct perf_evlist;
@@ -29,6 +31,9 @@ typedef int (*event_op2)(struct perf_tool *tool, union perf_event *event,
 typedef int (*event_oe)(struct perf_tool *tool, union perf_event *event,
 			struct ordered_events *oe);
 
+typedef s64 (*event_op3)(struct perf_tool *tool, union perf_event *event,
+			 struct perf_session *session);
+
 struct perf_tool {
 	event_sample	sample,
 			read;
@@ -44,7 +49,9 @@ struct perf_tool {
 	event_op2	tracing_data;
 	event_oe	finished_round;
 	event_op2	build_id,
-			id_index;
+			id_index,
+			auxtrace_info;
+	event_op3	auxtrace;
 	bool		ordered_events;
 	bool		ordering_requires_timestamps;
 };
-- 
1.9.1

[PATCH V6 04/25] perf tools: Add support for AUX area recording

[Adrian Hunter]

Add support for reading from the AUX area
tracing mmap and synthesizing AUX area
tracing events.

This patch introduces an abstraction for recording
AUX area data.  Recording is initialized
by auxtrace_record__init() which is a weak function
to be implemented by the architecture to provide
recording callbacks.  Recording is mainly handled
by auxtrace_mmap__read() and
perf_event__synthesize_auxtrace() but there are
callbacks for miscellaneous needs including
validating and processing user options, populating
private data in auxtrace_info_event, and freeing
the structure when finished.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/perf.h          |   2 +
 tools/perf/util/auxtrace.c | 196 +++++++++++++++++++++++++++++++++++++++++++++
 tools/perf/util/auxtrace.h |  60 +++++++++++++-
 tools/perf/util/record.c   |  11 ++-
 4 files changed, 267 insertions(+), 2 deletions(-)

diff --git a/tools/perf/perf.h b/tools/perf/perf.h
index 1caa70a..d635023 100644
--- a/tools/perf/perf.h
+++ b/tools/perf/perf.h
@@ -54,8 +54,10 @@ struct record_opts {
 	bool	     period;
 	bool	     sample_intr_regs;
 	bool	     running_time;
+	bool	     full_auxtrace;
 	unsigned int freq;
 	unsigned int mmap_pages;
+	unsigned int auxtrace_mmap_pages;
 	unsigned int user_freq;
 	u64          branch_stack;
 	u64	     default_interval;
diff --git a/tools/perf/util/auxtrace.c b/tools/perf/util/auxtrace.c
index 4b17a80..07718ae 100644
--- a/tools/perf/util/auxtrace.c
+++ b/tools/perf/util/auxtrace.c
@@ -23,6 +23,10 @@
 #include <linux/bitops.h>
 #include <linux/log2.h>
 
+#include <stdlib.h>
+#include <string.h>
+#include <errno.h>
+
 #include "../perf.h"
 #include "util.h"
 #include "evlist.h"
@@ -30,6 +34,9 @@
 #include "thread_map.h"
 #include "auxtrace.h"
 
+#include "event.h"
+#include "debug.h"
+
 int auxtrace_mmap__mmap(struct auxtrace_mmap *mm,
 			struct auxtrace_mmap_params *mp,
 			void *userpg, int fd)
@@ -101,3 +108,192 @@ void auxtrace_mmap_params__set_idx(struct auxtrace_mmap_params *mp,
 		mp->tid = evlist->threads->map[idx];
 	}
 }
+
+size_t auxtrace_record__info_priv_size(struct auxtrace_record *itr)
+{
+	if (itr)
+		return itr->info_priv_size(itr);
+	return 0;
+}
+
+static int auxtrace_not_supported(void)
+{
+	pr_err("AUX area tracing is not supported on this architecture\n");
+	return -EINVAL;
+}
+
+int auxtrace_record__info_fill(struct auxtrace_record *itr,
+			       struct perf_session *session,
+			       struct auxtrace_info_event *auxtrace_info,
+			       size_t priv_size)
+{
+	if (itr)
+		return itr->info_fill(itr, session, auxtrace_info, priv_size);
+	return auxtrace_not_supported();
+}
+
+void auxtrace_record__free(struct auxtrace_record *itr)
+{
+	if (itr)
+		itr->free(itr);
+}
+
+int auxtrace_record__options(struct auxtrace_record *itr,
+			     struct perf_evlist *evlist,
+			     struct record_opts *opts)
+{
+	if (itr)
+		return itr->recording_options(itr, evlist, opts);
+	return 0;
+}
+
+u64 auxtrace_record__reference(struct auxtrace_record *itr)
+{
+	if (itr)
+		return itr->reference(itr);
+	return 0;
+}
+
+struct auxtrace_record *__weak
+auxtrace_record__init(struct perf_evlist *evlist __maybe_unused, int *err)
+{
+	*err = 0;
+	return NULL;
+}
+
+int perf_event__synthesize_auxtrace_info(struct auxtrace_record *itr,
+					 struct perf_tool *tool,
+					 struct perf_session *session,
+					 perf_event__handler_t process)
+{
+	union perf_event *ev;
+	size_t priv_size;
+	int err;
+
+	pr_debug2("Synthesizing auxtrace information\n");
+	priv_size = auxtrace_record__info_priv_size(itr);
+	ev = zalloc(sizeof(struct auxtrace_info_event) + priv_size);
+	if (!ev)
+		return -ENOMEM;
+
+	ev->auxtrace_info.header.type = PERF_RECORD_AUXTRACE_INFO;
+	ev->auxtrace_info.header.size = sizeof(struct auxtrace_info_event) +
+					priv_size;
+	err = auxtrace_record__info_fill(itr, session, &ev->auxtrace_info,
+					 priv_size);
+	if (err)
+		goto out_free;
+
+	err = process(tool, ev, NULL, NULL);
+out_free:
+	free(ev);
+	return err;
+}
+
+int perf_event__synthesize_auxtrace(struct perf_tool *tool,
+				    perf_event__handler_t process,
+				    size_t size, u64 offset, u64 ref, int idx,
+				    u32 tid, u32 cpu)
+{
+	union perf_event ev;
+
+	memset(&ev, 0, sizeof(ev));
+	ev.auxtrace.header.type = PERF_RECORD_AUXTRACE;
+	ev.auxtrace.header.size = sizeof(ev.auxtrace);
+	ev.auxtrace.size = size;
+	ev.auxtrace.offset = offset;
+	ev.auxtrace.reference = ref;
+	ev.auxtrace.idx = idx;
+	ev.auxtrace.tid = tid;
+	ev.auxtrace.cpu = cpu;
+
+	return process(tool, &ev, NULL, NULL);
+}
+
+int auxtrace_mmap__read(struct auxtrace_mmap *mm, struct auxtrace_record *itr,
+			struct perf_tool *tool, process_auxtrace_t fn)
+{
+	u64 head = auxtrace_mmap__read_head(mm);
+	u64 old = mm->prev, offset, ref;
+	unsigned char *data = mm->base;
+	size_t size, head_off, old_off, len1, len2, padding;
+	union perf_event ev;
+	void *data1, *data2;
+
+	if (old == head)
+		return 0;
+
+	pr_debug3("auxtrace idx %d old %#"PRIx64" head %#"PRIx64" diff %#"PRIx64"\n",
+		  mm->idx, old, head, head - old);
+
+	if (mm->mask) {
+		head_off = head & mm->mask;
+		old_off = old & mm->mask;
+	} else {
+		head_off = head % mm->len;
+		old_off = old % mm->len;
+	}
+
+	if (head_off > old_off)
+		size = head_off - old_off;
+	else
+		size = mm->len - (old_off - head_off);
+
+	ref = auxtrace_record__reference(itr);
+
+	if (head > old || size <= head || mm->mask) {
+		offset = head - size;
+	} else {
+		/*
+		 * When the buffer size is not a power of 2, 'head' wraps at the
+		 * highest multiple of the buffer size, so we have to subtract
+		 * the remainder here.
+		 */
+		u64 rem = (0ULL - mm->len) % mm->len;
+
+		offset = head - size - rem;
+	}
+
+	if (size > head_off) {
+		len1 = size - head_off;
+		data1 = &data[mm->len - len1];
+		len2 = head_off;
+		data2 = &data[0];
+	} else {
+		len1 = size;
+		data1 = &data[head_off - len1];
+		len2 = 0;
+		data2 = NULL;
+	}
+
+	/* padding must be written by fn() e.g. record__process_auxtrace() */
+	padding = size & 7;
+	if (padding)
+		padding = 8 - padding;
+
+	memset(&ev, 0, sizeof(ev));
+	ev.auxtrace.header.type = PERF_RECORD_AUXTRACE;
+	ev.auxtrace.header.size = sizeof(ev.auxtrace);
+	ev.auxtrace.size = size + padding;
+	ev.auxtrace.offset = offset;
+	ev.auxtrace.reference = ref;
+	ev.auxtrace.idx = mm->idx;
+	ev.auxtrace.tid = mm->tid;
+	ev.auxtrace.cpu = mm->cpu;
+
+	if (fn(tool, &ev, data1, len1, data2, len2))
+		return -1;
+
+	mm->prev = head;
+
+	auxtrace_mmap__write_tail(mm, head);
+	if (itr->read_finish) {
+		int err;
+
+		err = itr->read_finish(itr, mm->idx);
+		if (err < 0)
+			return err;
+	}
+
+	return 1;
+}
diff --git a/tools/perf/util/auxtrace.h b/tools/perf/util/auxtrace.h
index 735ca2a..6b433e6 100644
--- a/tools/perf/util/auxtrace.h
+++ b/tools/perf/util/auxtrace.h
@@ -18,13 +18,18 @@
 
 #include <sys/types.h>
 #include <stdbool.h>
-
+#include <stddef.h>
 #include <linux/perf_event.h>
 #include <linux/types.h>
 
 #include "../perf.h"
 
+union perf_event;
+struct perf_session;
 struct perf_evlist;
+struct perf_tool;
+struct record_opts;
+struct auxtrace_info_event;
 
 /**
  * struct auxtrace_mmap - records an mmap of the auxtrace buffer.
@@ -70,6 +75,29 @@ struct auxtrace_mmap_params {
 	int		cpu;
 };
 
+/**
+ * struct auxtrace_record - callbacks for recording AUX area data.
+ * @recording_options: validate and process recording options
+ * @info_priv_size: return the size of the private data in auxtrace_info_event
+ * @info_fill: fill-in the private data in auxtrace_info_event
+ * @free: free this auxtrace record structure
+ * @reference: provide a 64-bit reference number for auxtrace_event
+ * @read_finish: called after reading from an auxtrace mmap
+ */
+struct auxtrace_record {
+	int (*recording_options)(struct auxtrace_record *itr,
+				 struct perf_evlist *evlist,
+				 struct record_opts *opts);
+	size_t (*info_priv_size)(struct auxtrace_record *itr);
+	int (*info_fill)(struct auxtrace_record *itr,
+			 struct perf_session *session,
+			 struct auxtrace_info_event *auxtrace_info,
+			 size_t priv_size);
+	void (*free)(struct auxtrace_record *itr);
+	u64 (*reference)(struct auxtrace_record *itr);
+	int (*read_finish)(struct auxtrace_record *itr, int idx);
+};
+
 static inline u64 auxtrace_mmap__read_head(struct auxtrace_mmap *mm __maybe_unused)
 {
 	/* Not yet implemented */
@@ -94,4 +122,34 @@ void auxtrace_mmap_params__set_idx(struct auxtrace_mmap_params *mp,
 				   struct perf_evlist *evlist, int idx,
 				   bool per_cpu);
 
+typedef int (*process_auxtrace_t)(struct perf_tool *tool,
+				  union perf_event *event, void *data1,
+				  size_t len1, void *data2, size_t len2);
+
+int auxtrace_mmap__read(struct auxtrace_mmap *mm, struct auxtrace_record *itr,
+			struct perf_tool *tool, process_auxtrace_t fn);
+
+struct auxtrace_record *auxtrace_record__init(struct perf_evlist *evlist,
+					      int *err);
+
+int auxtrace_record__options(struct auxtrace_record *itr,
+			     struct perf_evlist *evlist,
+			     struct record_opts *opts);
+size_t auxtrace_record__info_priv_size(struct auxtrace_record *itr);
+int auxtrace_record__info_fill(struct auxtrace_record *itr,
+			       struct perf_session *session,
+			       struct auxtrace_info_event *auxtrace_info,
+			       size_t priv_size);
+void auxtrace_record__free(struct auxtrace_record *itr);
+u64 auxtrace_record__reference(struct auxtrace_record *itr);
+
+int perf_event__synthesize_auxtrace_info(struct auxtrace_record *itr,
+					 struct perf_tool *tool,
+					 struct perf_session *session,
+					 perf_event__handler_t process);
+int perf_event__synthesize_auxtrace(struct perf_tool *tool,
+				    perf_event__handler_t process,
+				    size_t size, u64 offset, u64 ref, int idx,
+				    u32 tid, u32 cpu);
+
 #endif
diff --git a/tools/perf/util/record.c b/tools/perf/util/record.c
index 8acd0df..7553788 100644
--- a/tools/perf/util/record.c
+++ b/tools/perf/util/record.c
@@ -119,7 +119,16 @@ void perf_evlist__config(struct perf_evlist *evlist, struct record_opts *opts)
 			evsel->attr.comm_exec = 1;
 	}
 
-	if (evlist->nr_entries > 1) {
+	if (opts->full_auxtrace) {
+		/*
+		 * Need to be able to synthesize and parse selected events with
+		 * arbitrary sample types, which requires always being able to
+		 * match the id.
+		 */
+		use_sample_identifier = true;
+		evlist__for_each(evlist, evsel)
+			perf_evsel__set_sample_id(evsel, use_sample_identifier);
+	} else if (evlist->nr_entries > 1) {
 		struct perf_evsel *first = perf_evlist__first(evlist);
 
 		evlist__for_each(evlist, evsel) {
-- 
1.9.1

[PATCH V6 05/25] perf record: Add basic AUX area tracing support

[Adrian Hunter]

Amend the perf record tool to read the
AUX area tracing mmap and synthesize
AUX area tracing events.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/builtin-record.c | 80 ++++++++++++++++++++++++++++++++++++++++++---
 1 file changed, 76 insertions(+), 4 deletions(-)

diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
index 5a2ff51..8b3a50e 100644
--- a/tools/perf/builtin-record.c
+++ b/tools/perf/builtin-record.c
@@ -27,6 +27,7 @@
 #include "util/cpumap.h"
 #include "util/thread_map.h"
 #include "util/data.h"
+#include "util/auxtrace.h"
 
 #include <unistd.h>
 #include <sched.h>
@@ -38,6 +39,7 @@ struct record {
 	struct record_opts	opts;
 	u64			bytes_written;
 	struct perf_data_file	file;
+	struct auxtrace_record	*itr;
 	struct perf_evlist	*evlist;
 	struct perf_session	*session;
 	const char		*progname;
@@ -110,6 +112,43 @@ out:
 	return rc;
 }
 
+static int record__process_auxtrace(struct perf_tool *tool,
+				    union perf_event *event, void *data1,
+				    size_t len1, void *data2, size_t len2)
+{
+	struct record *rec = container_of(tool, struct record, tool);
+	size_t padding;
+	u8 pad[8] = {0};
+
+	/* event.auxtrace.size includes padding, see __auxtrace_mmap__read() */
+	padding = (len1 + len2) & 7;
+	if (padding)
+		padding = 8 - padding;
+
+	record__write(rec, event, event->header.size);
+	record__write(rec, data1, len1);
+	record__write(rec, data2, len2);
+	record__write(rec, &pad, padding);
+
+	return 0;
+}
+
+static int record__auxtrace_mmap_read(struct record *rec,
+				      struct auxtrace_mmap *mm)
+{
+	int ret;
+
+	ret = auxtrace_mmap__read(mm, rec->itr, &rec->tool,
+				  record__process_auxtrace);
+	if (ret < 0)
+		return ret;
+
+	if (ret)
+		rec->samples++;
+
+	return 0;
+}
+
 static volatile int done = 0;
 static volatile int signr = -1;
 static volatile int child_finished = 0;
@@ -168,13 +207,15 @@ try_again:
 		goto out;
 	}
 
-	if (perf_evlist__mmap(evlist, opts->mmap_pages, false) < 0) {
+	if (perf_evlist__mmap_ex(evlist, opts->mmap_pages, false,
+				 opts->auxtrace_mmap_pages, false) < 0) {
 		if (errno == EPERM) {
 			pr_err("Permission error mapping pages.\n"
 			       "Consider increasing "
 			       "/proc/sys/kernel/perf_event_mlock_kb,\n"
 			       "or try again with a smaller value of -m/--mmap_pages.\n"
-			       "(current value: %u)\n", opts->mmap_pages);
+			       "(current value: %u,%u)\n",
+			       opts->mmap_pages, opts->auxtrace_mmap_pages);
 			rc = -errno;
 		} else {
 			pr_err("failed to mmap with %d (%s)\n", errno,
@@ -269,12 +310,20 @@ static int record__mmap_read_all(struct record *rec)
 	int rc = 0;
 
 	for (i = 0; i < rec->evlist->nr_mmaps; i++) {
+		struct auxtrace_mmap *mm = &rec->evlist->mmap[i].auxtrace_mmap;
+
 		if (rec->evlist->mmap[i].base) {
 			if (record__mmap_read(rec, i) != 0) {
 				rc = -1;
 				goto out;
 			}
 		}
+
+		if (mm->base &&
+		    record__auxtrace_mmap_read(rec, mm) != 0) {
+			rc = -1;
+			goto out;
+		}
 	}
 
 	/*
@@ -304,6 +353,9 @@ static void record__init_features(struct record *rec)
 
 	if (!rec->opts.branch_stack)
 		perf_header__clear_feat(&session->header, HEADER_BRANCH_STACK);
+
+	if (!rec->opts.full_auxtrace)
+		perf_header__clear_feat(&session->header, HEADER_AUXTRACE);
 }
 
 static volatile int workload_exec_errno;
@@ -420,6 +472,13 @@ static int __cmd_record(struct record *rec, int argc, const char **argv)
 		}
 	}
 
+	if (rec->opts.full_auxtrace) {
+		err = perf_event__synthesize_auxtrace_info(rec->itr, tool,
+					session, process_synthesized_event);
+		if (err)
+			goto out_delete_session;
+	}
+
 	err = perf_event__synthesize_kernel_mmap(tool, process_synthesized_event,
 						 machine);
 	if (err < 0)
@@ -552,7 +611,7 @@ out_child:
 	if (!err && !quiet) {
 		char samples[128];
 
-		if (rec->samples)
+		if (rec->samples && !rec->opts.full_auxtrace)
 			scnprintf(samples, sizeof(samples),
 				  " (%" PRIu64 " samples)", rec->samples);
 		else
@@ -848,7 +907,7 @@ struct option *record_options = __record_options;
 
 int cmd_record(int argc, const char **argv, const char *prefix __maybe_unused)
 {
-	int err = -ENOMEM;
+	int err;
 	struct record *rec = &record;
 	char errbuf[BUFSIZ];
 
@@ -869,6 +928,14 @@ int cmd_record(int argc, const char **argv, const char *prefix __maybe_unused)
 		usage_with_options(record_usage, record_options);
 	}
 
+	if (!rec->itr) {
+		rec->itr = auxtrace_record__init(rec->evlist, &err);
+		if (err)
+			return err;
+	}
+
+	err = -ENOMEM;
+
 	symbol__init(NULL);
 
 	if (symbol_conf.kptr_restrict)
@@ -914,6 +981,10 @@ int cmd_record(int argc, const char **argv, const char *prefix __maybe_unused)
 	if (perf_evlist__create_maps(rec->evlist, &rec->opts.target) < 0)
 		usage_with_options(record_usage, record_options);
 
+	err = auxtrace_record__options(rec->itr, rec->evlist, &rec->opts);
+	if (err)
+		goto out_symbol_exit;
+
 	if (record_opts__config(&rec->opts)) {
 		err = -EINVAL;
 		goto out_symbol_exit;
@@ -923,5 +994,6 @@ int cmd_record(int argc, const char **argv, const char *prefix __maybe_unused)
 out_symbol_exit:
 	perf_evlist__delete(rec->evlist);
 	symbol__exit();
+	auxtrace_record__free(rec->itr);
 	return err;
 }
-- 
1.9.1

[PATCH V6 06/25] perf record: Extend -m option for AUX area tracing mmap pages

[Adrian Hunter]

Extend the -m option so that the number
of mmap pages for AUX area tracing
can be specified.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/Documentation/perf-record.txt |  2 ++
 tools/perf/builtin-record.c              | 49 ++++++++++++++++++++++++++++++--
 tools/perf/util/evlist.c                 | 10 +++++--
 tools/perf/util/evlist.h                 |  1 +
 4 files changed, 56 insertions(+), 6 deletions(-)

diff --git a/tools/perf/Documentation/perf-record.txt b/tools/perf/Documentation/perf-record.txt
index 355c4f5..e78dd80 100644
--- a/tools/perf/Documentation/perf-record.txt
+++ b/tools/perf/Documentation/perf-record.txt
@@ -108,6 +108,8 @@ OPTIONS
 	Number of mmap data pages (must be a power of two) or size
 	specification with appended unit character - B/K/M/G. The
 	size is rounded up to have nearest pages power of two value.
+	Also, by adding a comma, the number of mmap pages for AUX
+	area tracing can be specified.
 
 --group::
 	Put all events in a single event group.  This precedes the --event
diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
index 8b3a50e..d337896 100644
--- a/tools/perf/builtin-record.c
+++ b/tools/perf/builtin-record.c
@@ -769,6 +769,49 @@ static int perf_record_config(const char *var, const char *value, void *cb)
 	return perf_default_config(var, value, cb);
 }
 
+static int record__parse_mmap_pages(const struct option *opt,
+				    const char *str,
+				    int unset __maybe_unused)
+{
+	struct record_opts *opts = opt->value;
+	char *s, *p;
+	unsigned int mmap_pages;
+	int ret;
+
+	if (!str)
+		return -EINVAL;
+
+	s = strdup(str);
+	if (!s)
+		return -ENOMEM;
+
+	p = strchr(s, ',');
+	if (p)
+		*p = '\0';
+
+	if (*s) {
+		ret = __perf_evlist__parse_mmap_pages(&mmap_pages, s);
+		if (ret)
+			goto out_free;
+		opts->mmap_pages = mmap_pages;
+	}
+
+	if (!p) {
+		ret = 0;
+		goto out_free;
+	}
+
+	ret = __perf_evlist__parse_mmap_pages(&mmap_pages, p + 1);
+	if (ret)
+		goto out_free;
+
+	opts->auxtrace_mmap_pages = mmap_pages;
+
+out_free:
+	free(s);
+	return ret;
+}
+
 static const char * const __record_usage[] = {
 	"perf record [<options>] [<command>]",
 	"perf record [<options>] -- <command> [<options>]",
@@ -849,9 +892,9 @@ struct option __record_options[] = {
 			&record.opts.no_inherit_set,
 			"child tasks do not inherit counters"),
 	OPT_UINTEGER('F', "freq", &record.opts.user_freq, "profile at this frequency"),
-	OPT_CALLBACK('m', "mmap-pages", &record.opts.mmap_pages, "pages",
-		     "number of mmap data pages",
-		     perf_evlist__parse_mmap_pages),
+	OPT_CALLBACK('m', "mmap-pages", &record.opts, "pages[,pages]",
+		     "number of mmap data pages and AUX area tracing mmap pages",
+		     record__parse_mmap_pages),
 	OPT_BOOLEAN(0, "group", &record.opts.group,
 		    "put the counters into a counter group"),
 	OPT_CALLBACK_NOOPT('g', NULL, &record.opts,
diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c
index 3cc7815..4a42730 100644
--- a/tools/perf/util/evlist.c
+++ b/tools/perf/util/evlist.c
@@ -1000,10 +1000,8 @@ static long parse_pages_arg(const char *str, unsigned long min,
 	return pages;
 }
 
-int perf_evlist__parse_mmap_pages(const struct option *opt, const char *str,
-				  int unset __maybe_unused)
+int __perf_evlist__parse_mmap_pages(unsigned int *mmap_pages, const char *str)
 {
-	unsigned int *mmap_pages = opt->value;
 	unsigned long max = UINT_MAX;
 	long pages;
 
@@ -1020,6 +1018,12 @@ int perf_evlist__parse_mmap_pages(const struct option *opt, const char *str,
 	return 0;
 }
 
+int perf_evlist__parse_mmap_pages(const struct option *opt, const char *str,
+				  int unset __maybe_unused)
+{
+	return __perf_evlist__parse_mmap_pages(opt->value, str);
+}
+
 /**
  * perf_evlist__mmap_ex - Create mmaps to receive events.
  * @evlist: list of events
diff --git a/tools/perf/util/evlist.h b/tools/perf/util/evlist.h
index fbb8063..8228b58 100644
--- a/tools/perf/util/evlist.h
+++ b/tools/perf/util/evlist.h
@@ -124,6 +124,7 @@ int perf_evlist__start_workload(struct perf_evlist *evlist);
 
 struct option;
 
+int __perf_evlist__parse_mmap_pages(unsigned int *mmap_pages, const char *str);
 int perf_evlist__parse_mmap_pages(const struct option *opt,
 				  const char *str,
 				  int unset);
-- 
1.9.1

[PATCH V6 07/25] perf tools: Add a user event for AUX area tracing errors

[Adrian Hunter]

Errors encountered when decoding an AUX area
trace need to be reported to the user. However
the "user" might be a script or another tool,
so provide a new user event to capture those
errors.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/util/event.c   |  1 +
 tools/perf/util/event.h   | 16 ++++++++++++++++
 tools/perf/util/session.c | 25 +++++++++++++++++++++++++
 tools/perf/util/tool.h    |  3 ++-
 4 files changed, 44 insertions(+), 1 deletion(-)

diff --git a/tools/perf/util/event.c b/tools/perf/util/event.c
index cbf3775..6bde493 100644
--- a/tools/perf/util/event.c
+++ b/tools/perf/util/event.c
@@ -31,6 +31,7 @@ static const char *perf_event__names[] = {
 	[PERF_RECORD_ID_INDEX]			= "ID_INDEX",
 	[PERF_RECORD_AUXTRACE_INFO]		= "AUXTRACE_INFO",
 	[PERF_RECORD_AUXTRACE]			= "AUXTRACE",
+	[PERF_RECORD_AUXTRACE_ERROR]		= "AUXTRACE_ERROR",
 };
 
 const char *perf_event__name(unsigned int id)
diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h
index 5dc9329..02327bb 100644
--- a/tools/perf/util/event.h
+++ b/tools/perf/util/event.h
@@ -217,6 +217,7 @@ enum perf_user_event_type { /* above any possible kernel type */
 	PERF_RECORD_ID_INDEX			= 69,
 	PERF_RECORD_AUXTRACE_INFO		= 70,
 	PERF_RECORD_AUXTRACE			= 71,
+	PERF_RECORD_AUXTRACE_ERROR		= 72,
 	PERF_RECORD_HEADER_MAX
 };
 
@@ -301,6 +302,20 @@ struct auxtrace_event {
 	u32 reserved__; /* For alignment */
 };
 
+#define MAX_AUXTRACE_ERROR_MSG 64
+
+struct auxtrace_error_event {
+	struct perf_event_header header;
+	u32 type;
+	u32 code;
+	u32 cpu;
+	u32 pid;
+	u32 tid;
+	u32 reserved__; /* For alignment */
+	u64 ip;
+	char msg[MAX_AUXTRACE_ERROR_MSG];
+};
+
 union perf_event {
 	struct perf_event_header	header;
 	struct mmap_event		mmap;
@@ -318,6 +333,7 @@ union perf_event {
 	struct id_index_event		id_index;
 	struct auxtrace_info_event	auxtrace_info;
 	struct auxtrace_event		auxtrace;
+	struct auxtrace_error_event	auxtrace_error;
 };
 
 void perf_event__print_totals(void);
diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c
index a8c67cb..b4cac8b 100644
--- a/tools/perf/util/session.c
+++ b/tools/perf/util/session.c
@@ -288,6 +288,15 @@ static s64 process_event_auxtrace_stub(struct perf_tool *tool __maybe_unused,
 	return event->auxtrace.size;
 }
 
+static
+int process_event_auxtrace_error_stub(struct perf_tool *tool __maybe_unused,
+				      union perf_event *event __maybe_unused,
+				      struct perf_session *session __maybe_unused)
+{
+	dump_printf(": unhandled!\n");
+	return 0;
+}
+
 void perf_tool__fill_defaults(struct perf_tool *tool)
 {
 	if (tool->sample == NULL)
@@ -328,6 +337,8 @@ void perf_tool__fill_defaults(struct perf_tool *tool)
 		tool->auxtrace_info = process_event_auxtrace_info_stub;
 	if (tool->auxtrace == NULL)
 		tool->auxtrace = process_event_auxtrace_stub;
+	if (tool->auxtrace_error == NULL)
+		tool->auxtrace_error = process_event_auxtrace_error_stub;
 }
 
 static void swap_sample_id_all(union perf_event *event, void *data)
@@ -531,6 +542,17 @@ static void perf_event__auxtrace_swap(union perf_event *event,
 	event->auxtrace.cpu       = bswap_32(event->auxtrace.cpu);
 }
 
+static void perf_event__auxtrace_error_swap(union perf_event *event,
+					    bool sample_id_all __maybe_unused)
+{
+	event->auxtrace_error.type = bswap_32(event->auxtrace_error.type);
+	event->auxtrace_error.code = bswap_32(event->auxtrace_error.code);
+	event->auxtrace_error.cpu  = bswap_32(event->auxtrace_error.cpu);
+	event->auxtrace_error.pid  = bswap_32(event->auxtrace_error.pid);
+	event->auxtrace_error.tid  = bswap_32(event->auxtrace_error.tid);
+	event->auxtrace_error.ip   = bswap_64(event->auxtrace_error.ip);
+}
+
 typedef void (*perf_event__swap_op)(union perf_event *event,
 				    bool sample_id_all);
 
@@ -552,6 +574,7 @@ static perf_event__swap_op perf_event__swap_ops[] = {
 	[PERF_RECORD_ID_INDEX]		  = perf_event__all64_swap,
 	[PERF_RECORD_AUXTRACE_INFO]	  = perf_event__auxtrace_info_swap,
 	[PERF_RECORD_AUXTRACE]		  = perf_event__auxtrace_swap,
+	[PERF_RECORD_AUXTRACE_ERROR]	  = perf_event__auxtrace_error_swap,
 	[PERF_RECORD_HEADER_MAX]	  = NULL,
 };
 
@@ -1041,6 +1064,8 @@ static s64 perf_session__process_user_event(struct perf_session *session,
 		/* setup for reading amidst mmap */
 		lseek(fd, file_offset + event->header.size, SEEK_SET);
 		return tool->auxtrace(tool, event, session);
+	case PERF_RECORD_AUXTRACE_ERROR:
+		return tool->auxtrace_error(tool, event, session);
 	default:
 		return -EINVAL;
 	}
diff --git a/tools/perf/util/tool.h b/tools/perf/util/tool.h
index 0146f8e..20afe19 100644
--- a/tools/perf/util/tool.h
+++ b/tools/perf/util/tool.h
@@ -50,7 +50,8 @@ struct perf_tool {
 	event_oe	finished_round;
 	event_op2	build_id,
 			id_index,
-			auxtrace_info;
+			auxtrace_info,
+			auxtrace_error;
 	event_op3	auxtrace;
 	bool		ordered_events;
 	bool		ordering_requires_timestamps;
-- 
1.9.1

[PATCH V6 08/25] perf session: Add hooks to allow transparent decoding of AUX area tracing data

[Adrian Hunter]

Hook into session processing so that AUX area decoding can
synthesize events transparently to the tools.

The advantages of transparent decoding are that tools can be used
directly with perf.data files containing AUX area tracing data,
which is easier for the user and more efficient than having a
separate decoding tool.

This will work as follows:
1. Tools will feed auxtrace events to the decoder using
perf_tool->auxtrace() (support for that still to come).
2. The decoder can process side-band events as needed due
to the auxtrace->process_event() hook.
3. The decoder can deliver synthesized events into the
event stream using perf_session__deliver_synth_event().

Note the expectation is that decoding will work on data that is
time-ordered with respect to the per-cpu or per-thread contexts
that were recorded.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/util/auxtrace.h | 55 ++++++++++++++++++++++++++++++++++++
 tools/perf/util/session.c  | 70 +++++++++++++++++++++++++++++++++++-----------
 tools/perf/util/session.h  |  3 ++
 3 files changed, 111 insertions(+), 17 deletions(-)

diff --git a/tools/perf/util/auxtrace.h b/tools/perf/util/auxtrace.h
index 6b433e6..909df80 100644
--- a/tools/perf/util/auxtrace.h
+++ b/tools/perf/util/auxtrace.h
@@ -23,6 +23,7 @@
 #include <linux/types.h>
 
 #include "../perf.h"
+#include "session.h"
 
 union perf_event;
 struct perf_session;
@@ -32,6 +33,24 @@ struct record_opts;
 struct auxtrace_info_event;
 
 /**
+ * struct auxtrace - session callbacks to allow AUX area data decoding.
+ * @process_event: lets the decoder see all session events
+ * @flush_events: process any remaining data
+ * @free_events: free resources associated with event processing
+ * @free: free resources associated with the session
+ */
+struct auxtrace {
+	int (*process_event)(struct perf_session *session,
+			     union perf_event *event,
+			     struct perf_sample *sample,
+			     struct perf_tool *tool);
+	int (*flush_events)(struct perf_session *session,
+			    struct perf_tool *tool);
+	void (*free_events)(struct perf_session *session);
+	void (*free)(struct perf_session *session);
+};
+
+/**
  * struct auxtrace_mmap - records an mmap of the auxtrace buffer.
  * @base: address of mapped area
  * @userpg: pointer to buffer's perf_event_mmap_page
@@ -152,4 +171,40 @@ int perf_event__synthesize_auxtrace(struct perf_tool *tool,
 				    size_t size, u64 offset, u64 ref, int idx,
 				    u32 tid, u32 cpu);
 
+static inline int auxtrace__process_event(struct perf_session *session,
+					  union perf_event *event,
+					  struct perf_sample *sample,
+					  struct perf_tool *tool)
+{
+	if (!session->auxtrace)
+		return 0;
+
+	return session->auxtrace->process_event(session, event, sample, tool);
+}
+
+static inline int auxtrace__flush_events(struct perf_session *session,
+					 struct perf_tool *tool)
+{
+	if (!session->auxtrace)
+		return 0;
+
+	return session->auxtrace->flush_events(session, tool);
+}
+
+static inline void auxtrace__free_events(struct perf_session *session)
+{
+	if (!session->auxtrace)
+		return;
+
+	return session->auxtrace->free_events(session);
+}
+
+static inline void auxtrace__free(struct perf_session *session)
+{
+	if (!session->auxtrace)
+		return;
+
+	return session->auxtrace->free(session);
+}
+
 #endif
diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c
index b4cac8b..918554d 100644
--- a/tools/perf/util/session.c
+++ b/tools/perf/util/session.c
@@ -15,12 +15,7 @@
 #include "cpumap.h"
 #include "perf_regs.h"
 #include "asm/bug.h"
-
-static int machines__deliver_event(struct machines *machines,
-				   struct perf_evlist *evlist,
-				   union perf_event *event,
-				   struct perf_sample *sample,
-				   struct perf_tool *tool, u64 file_offset);
+#include "auxtrace.h"
 
 static int perf_session__open(struct perf_session *session)
 {
@@ -92,13 +87,9 @@ static void perf_session__set_comm_exec(struct perf_session *session)
 	machines__set_comm_exec(&session->machines, comm_exec);
 }
 
-static int ordered_events__deliver_event(struct ordered_events *oe,
-					 struct ordered_event *event,
-					 struct perf_sample *sample)
-{
-	return machines__deliver_event(oe->machines, oe->evlist, event->event,
-				       sample, oe->tool, event->file_offset);
-}
+static int perf_session__deliver_ordered_event(struct ordered_events *oe,
+					       struct ordered_event *event,
+					       struct perf_sample *sample);
 
 struct perf_session *perf_session__new(struct perf_data_file *file,
 				       bool repipe, struct perf_tool *tool)
@@ -141,7 +132,8 @@ struct perf_session *perf_session__new(struct perf_data_file *file,
 		tool->ordered_events = false;
 	} else {
 		ordered_events__init(&session->ordered_events, &session->machines,
-				     session->evlist, tool, ordered_events__deliver_event);
+				     session->evlist, tool,
+				     perf_session__deliver_ordered_event);
 	}
 
 	return session;
@@ -177,6 +169,7 @@ static void perf_session_env__delete(struct perf_session_env *env)
 
 void perf_session__delete(struct perf_session *session)
 {
+	auxtrace__free(session);
 	perf_session__destroy_kernel_maps(session);
 	perf_session__delete_threads(session);
 	perf_session_env__delete(&session->header.env);
@@ -1022,6 +1015,35 @@ static int machines__deliver_event(struct machines *machines,
 	}
 }
 
+static int perf_session__deliver_event(struct perf_session *session,
+				       union perf_event *event,
+				       struct perf_sample *sample,
+				       struct perf_tool *tool,
+				       u64 file_offset)
+{
+	int ret;
+
+	ret = auxtrace__process_event(session, event, sample, tool);
+	if (ret < 0)
+		return ret;
+	if (ret > 0)
+		return 0;
+
+	return machines__deliver_event(&session->machines, session->evlist,
+				       event, sample, tool, file_offset);
+}
+
+static int perf_session__deliver_ordered_event(struct ordered_events *oe,
+					       struct ordered_event *event,
+					       struct perf_sample *sample)
+{
+	struct perf_session *session =
+			container_of(oe, struct perf_session, ordered_events);
+
+	return perf_session__deliver_event(session, event->event, sample,
+					   oe->tool, event->file_offset);
+}
+
 static s64 perf_session__process_user_event(struct perf_session *session,
 					    union perf_event *event,
 					    u64 file_offset)
@@ -1182,8 +1204,8 @@ static s64 perf_session__process_event(struct perf_session *session,
 			return ret;
 	}
 
-	return machines__deliver_event(&session->machines, evlist, event,
-				       &sample, tool, file_offset);
+	return perf_session__deliver_event(session, event, &sample, tool,
+					   file_offset);
 }
 
 void perf_event_header__bswap(struct perf_event_header *hdr)
@@ -1340,10 +1362,14 @@ more:
 done:
 	/* do the final flush for ordered samples */
 	err = ordered_events__flush(oe, OE_FLUSH__FINAL);
+	if (err)
+		goto out_err;
+	err = auxtrace__flush_events(session, tool);
 out_err:
 	free(buf);
 	perf_tool__warn_about_errors(tool, &session->evlist->stats);
 	ordered_events__free(&session->ordered_events);
+	auxtrace__free_events(session);
 	return err;
 }
 
@@ -1486,10 +1512,14 @@ more:
 out:
 	/* do the final flush for ordered samples */
 	err = ordered_events__flush(oe, OE_FLUSH__FINAL);
+	if (err)
+		goto out_err;
+	err = auxtrace__flush_events(session, tool);
 out_err:
 	ui_progress__finish();
 	perf_tool__warn_about_errors(tool, &session->evlist->stats);
 	ordered_events__free(&session->ordered_events);
+	auxtrace__free_events(session);
 	session->one_mmap = false;
 	return err;
 }
@@ -1569,7 +1599,13 @@ size_t perf_session__fprintf_dsos_buildid(struct perf_session *session, FILE *fp
 
 size_t perf_session__fprintf_nr_events(struct perf_session *session, FILE *fp)
 {
-	size_t ret = fprintf(fp, "Aggregated stats:\n");
+	size_t ret;
+	const char *msg = "";
+
+	if (perf_header__has_feat(&session->header, HEADER_AUXTRACE))
+		msg = " (excludes AUX area (e.g. instruction trace) decoded / synthesized events)";
+
+	ret = fprintf(fp, "Aggregated stats:%s\n", msg);
 
 	ret += events_stats__fprintf(&session->evlist->stats, fp);
 	return ret;
diff --git a/tools/perf/util/session.h b/tools/perf/util/session.h
index 1310998..8575d6f 100644
--- a/tools/perf/util/session.h
+++ b/tools/perf/util/session.h
@@ -15,10 +15,13 @@
 struct ip_callchain;
 struct thread;
 
+struct auxtrace;
+
 struct perf_session {
 	struct perf_header	header;
 	struct machines		machines;
 	struct perf_evlist	*evlist;
+	struct auxtrace		*auxtrace;
 	struct trace_event	tevent;
 	bool			repipe;
 	bool			one_mmap;
-- 
1.9.1

[PATCH V6 09/25] perf session: Add instruction tracing options

[Adrian Hunter]

It is assumed that AUX area decoding will
synthesize events for consumption by other tools.
At this time, the main use of AUX area tracing will be
to capture instruction trace (aka processor trace) data.
The nature of instruction tracing suggests the
initial inclusion of options for "instructions"
and "branches" events, but more could be added
as needed.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/util/auxtrace.c | 131 +++++++++++++++++++++++++++++++++++++++++++++
 tools/perf/util/auxtrace.h |  43 +++++++++++++++
 tools/perf/util/session.h  |   2 +
 3 files changed, 176 insertions(+)

diff --git a/tools/perf/util/auxtrace.c b/tools/perf/util/auxtrace.c
index 07718ae..54f67db 100644
--- a/tools/perf/util/auxtrace.c
+++ b/tools/perf/util/auxtrace.c
@@ -36,6 +36,7 @@
 
 #include "event.h"
 #include "debug.h"
+#include "parse-options.h"
 
 int auxtrace_mmap__mmap(struct auxtrace_mmap *mm,
 			struct auxtrace_mmap_params *mp,
@@ -210,6 +211,136 @@ int perf_event__synthesize_auxtrace(struct perf_tool *tool,
 	return process(tool, &ev, NULL, NULL);
 }
 
+#define PERF_ITRACE_DEFAULT_PERIOD_TYPE		PERF_ITRACE_PERIOD_NANOSECS
+#define PERF_ITRACE_DEFAULT_PERIOD		100000
+#define PERF_ITRACE_DEFAULT_CALLCHAIN_SZ	16
+#define PERF_ITRACE_MAX_CALLCHAIN_SZ		1024
+
+void itrace_synth_opts__set_default(struct itrace_synth_opts *synth_opts)
+{
+	synth_opts->instructions = true;
+	synth_opts->branches = true;
+	synth_opts->errors = true;
+	synth_opts->period_type = PERF_ITRACE_DEFAULT_PERIOD_TYPE;
+	synth_opts->period = PERF_ITRACE_DEFAULT_PERIOD;
+	synth_opts->callchain_sz = PERF_ITRACE_DEFAULT_CALLCHAIN_SZ;
+}
+
+int itrace_parse_synth_opts(const struct option *opt, const char *str,
+			    int unset)
+{
+	struct itrace_synth_opts *synth_opts = opt->value;
+	const char *p;
+	char *endptr;
+
+	synth_opts->set = true;
+
+	if (unset) {
+		synth_opts->dont_decode = true;
+		return 0;
+	}
+
+	if (!str) {
+		itrace_synth_opts__set_default(synth_opts);
+		return 0;
+	}
+
+	for (p = str; *p;) {
+		switch (*p++) {
+		case 'i':
+			synth_opts->instructions = true;
+			while (*p == ' ' || *p == ',')
+				p += 1;
+			if (isdigit(*p)) {
+				synth_opts->period = strtoull(p, &endptr, 10);
+				p = endptr;
+				while (*p == ' ' || *p == ',')
+					p += 1;
+				switch (*p++) {
+				case 'i':
+					synth_opts->period_type =
+						PERF_ITRACE_PERIOD_INSTRUCTIONS;
+					break;
+				case 't':
+					synth_opts->period_type =
+						PERF_ITRACE_PERIOD_TICKS;
+					break;
+				case 'm':
+					synth_opts->period *= 1000;
+					/* Fall through */
+				case 'u':
+					synth_opts->period *= 1000;
+					/* Fall through */
+				case 'n':
+					if (*p++ != 's')
+						goto out_err;
+					synth_opts->period_type =
+						PERF_ITRACE_PERIOD_NANOSECS;
+					break;
+				case '\0':
+					goto out;
+				default:
+					goto out_err;
+				}
+			}
+			break;
+		case 'b':
+			synth_opts->branches = true;
+			break;
+		case 'e':
+			synth_opts->errors = true;
+			break;
+		case 'd':
+			synth_opts->log = true;
+			break;
+		case 'c':
+			synth_opts->branches = true;
+			synth_opts->calls = true;
+			break;
+		case 'r':
+			synth_opts->branches = true;
+			synth_opts->returns = true;
+			break;
+		case 'g':
+			synth_opts->instructions = true;
+			synth_opts->callchain = true;
+			synth_opts->callchain_sz =
+					PERF_ITRACE_DEFAULT_CALLCHAIN_SZ;
+			while (*p == ' ' || *p == ',')
+				p += 1;
+			if (isdigit(*p)) {
+				unsigned int val;
+
+				val = strtoul(p, &endptr, 10);
+				p = endptr;
+				if (!val || val > PERF_ITRACE_MAX_CALLCHAIN_SZ)
+					goto out_err;
+				synth_opts->callchain_sz = val;
+			}
+			break;
+		case ' ':
+		case ',':
+			break;
+		default:
+			goto out_err;
+		}
+	}
+out:
+	if (synth_opts->instructions) {
+		if (!synth_opts->period_type)
+			synth_opts->period_type =
+					PERF_ITRACE_DEFAULT_PERIOD_TYPE;
+		if (!synth_opts->period)
+			synth_opts->period = PERF_ITRACE_DEFAULT_PERIOD;
+	}
+
+	return 0;
+
+out_err:
+	pr_err("Bad Instruction Tracing options '%s'\n", str);
+	return -EINVAL;
+}
+
 int auxtrace_mmap__read(struct auxtrace_mmap *mm, struct auxtrace_record *itr,
 			struct perf_tool *tool, process_auxtrace_t fn)
 {
diff --git a/tools/perf/util/auxtrace.h b/tools/perf/util/auxtrace.h
index 909df80..4a7b7bf 100644
--- a/tools/perf/util/auxtrace.h
+++ b/tools/perf/util/auxtrace.h
@@ -29,9 +29,49 @@ union perf_event;
 struct perf_session;
 struct perf_evlist;
 struct perf_tool;
+struct option;
 struct record_opts;
 struct auxtrace_info_event;
 
+enum itrace_period_type {
+	PERF_ITRACE_PERIOD_INSTRUCTIONS,
+	PERF_ITRACE_PERIOD_TICKS,
+	PERF_ITRACE_PERIOD_NANOSECS,
+};
+
+/**
+ * struct itrace_synth_opts - AUX area tracing synthesis options.
+ * @set: indicates whether or not options have been set
+ * @inject: indicates the event (not just the sample) must be fully synthesized
+ *          because 'perf inject' will write it out
+ * @instructions: whether to synthesize 'instructions' events
+ * @branches: whether to synthesize 'branches' events
+ * @errors: whether to synthesize decoder error events
+ * @dont_decode: whether to skip decoding entirely
+ * @log: write a decoding log
+ * @calls: limit branch samples to calls (can be combined with @returns)
+ * @returns: limit branch samples to returns (can be combined with @calls)
+ * @callchain: add callchain to 'instructions' events
+ * @callchain_sz: maximum callchain size
+ * @period: 'instructions' events period
+ * @period_type: 'instructions' events period type
+ */
+struct itrace_synth_opts {
+	bool			set;
+	bool			inject;
+	bool			instructions;
+	bool			branches;
+	bool			errors;
+	bool			dont_decode;
+	bool			log;
+	bool			calls;
+	bool			returns;
+	bool			callchain;
+	unsigned int		callchain_sz;
+	unsigned long long	period;
+	enum itrace_period_type	period_type;
+};
+
 /**
  * struct auxtrace - session callbacks to allow AUX area data decoding.
  * @process_event: lets the decoder see all session events
@@ -170,6 +210,9 @@ int perf_event__synthesize_auxtrace(struct perf_tool *tool,
 				    perf_event__handler_t process,
 				    size_t size, u64 offset, u64 ref, int idx,
 				    u32 tid, u32 cpu);
+int itrace_parse_synth_opts(const struct option *opt, const char *str,
+			    int unset);
+void itrace_synth_opts__set_default(struct itrace_synth_opts *synth_opts);
 
 static inline int auxtrace__process_event(struct perf_session *session,
 					  union perf_event *event,
diff --git a/tools/perf/util/session.h b/tools/perf/util/session.h
index 8575d6f..d7d8225 100644
--- a/tools/perf/util/session.h
+++ b/tools/perf/util/session.h
@@ -16,12 +16,14 @@ struct ip_callchain;
 struct thread;
 
 struct auxtrace;
+struct itrace_synth_opts;
 
 struct perf_session {
 	struct perf_header	header;
 	struct machines		machines;
 	struct perf_evlist	*evlist;
 	struct auxtrace		*auxtrace;
+	struct itrace_synth_opts *itrace_synth_opts;
 	struct trace_event	tevent;
 	bool			repipe;
 	bool			one_mmap;
-- 
1.9.1

[PATCH V6 10/25] perf auxtrace: Add helpers for AUX area tracing errors

[Adrian Hunter]

Add functions to synthesize, count and print
AUX area tracing error events.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/util/auxtrace.c | 81 ++++++++++++++++++++++++++++++++++++++++++++++
 tools/perf/util/auxtrace.h | 14 ++++++++
 tools/perf/util/event.h    |  6 ++++
 tools/perf/util/session.c  |  3 ++
 4 files changed, 104 insertions(+)

diff --git a/tools/perf/util/auxtrace.c b/tools/perf/util/auxtrace.c
index 54f67db..cdcd6ae 100644
--- a/tools/perf/util/auxtrace.c
+++ b/tools/perf/util/auxtrace.c
@@ -24,6 +24,7 @@
 #include <linux/log2.h>
 
 #include <stdlib.h>
+#include <stdio.h>
 #include <string.h>
 #include <errno.h>
 
@@ -35,6 +36,7 @@
 #include "auxtrace.h"
 
 #include "event.h"
+#include "session.h"
 #include "debug.h"
 #include "parse-options.h"
 
@@ -162,6 +164,28 @@ auxtrace_record__init(struct perf_evlist *evlist __maybe_unused, int *err)
 	return NULL;
 }
 
+void auxtrace_synth_error(struct auxtrace_error_event *auxtrace_error, int type,
+			  int code, int cpu, pid_t pid, pid_t tid, u64 ip,
+			  const char *msg)
+{
+	size_t size;
+
+	memset(auxtrace_error, 0, sizeof(struct auxtrace_error_event));
+
+	auxtrace_error->header.type = PERF_RECORD_AUXTRACE_ERROR;
+	auxtrace_error->type = type;
+	auxtrace_error->code = code;
+	auxtrace_error->cpu = cpu;
+	auxtrace_error->pid = pid;
+	auxtrace_error->tid = tid;
+	auxtrace_error->ip = ip;
+	strlcpy(auxtrace_error->msg, msg, MAX_AUXTRACE_ERROR_MSG);
+
+	size = (void *)auxtrace_error->msg - (void *)auxtrace_error +
+	       strlen(auxtrace_error->msg) + 1;
+	auxtrace_error->header.size = PERF_ALIGN(size, sizeof(u64));
+}
+
 int perf_event__synthesize_auxtrace_info(struct auxtrace_record *itr,
 					 struct perf_tool *tool,
 					 struct perf_session *session,
@@ -341,6 +365,63 @@ out_err:
 	return -EINVAL;
 }
 
+static const char * const auxtrace_error_type_name[] = {
+	[PERF_AUXTRACE_ERROR_ITRACE] = "instruction trace",
+};
+
+static const char *auxtrace_error_name(int type)
+{
+	const char *error_type_name = NULL;
+
+	if (type < PERF_AUXTRACE_ERROR_MAX)
+		error_type_name = auxtrace_error_type_name[type];
+	if (!error_type_name)
+		error_type_name = "unknown AUX";
+	return error_type_name;
+}
+
+size_t perf_event__fprintf_auxtrace_error(union perf_event *event, FILE *fp)
+{
+	struct auxtrace_error_event *e = &event->auxtrace_error;
+	int ret;
+
+	ret = fprintf(fp, " %s error type %u",
+		      auxtrace_error_name(e->type), e->type);
+	ret += fprintf(fp, " cpu %d pid %d tid %d ip %#"PRIx64" code %u: %s\n",
+		       e->cpu, e->pid, e->tid, e->ip, e->code, e->msg);
+	return ret;
+}
+
+void perf_session__auxtrace_error_inc(struct perf_session *session,
+				      union perf_event *event)
+{
+	struct auxtrace_error_event *e = &event->auxtrace_error;
+
+	if (e->type < PERF_AUXTRACE_ERROR_MAX)
+		session->evlist->stats.nr_auxtrace_errors[e->type] += 1;
+}
+
+void events_stats__auxtrace_error_warn(const struct events_stats *stats)
+{
+	int i;
+
+	for (i = 0; i < PERF_AUXTRACE_ERROR_MAX; i++) {
+		if (!stats->nr_auxtrace_errors[i])
+			continue;
+		ui__warning("%u %s errors\n",
+			    stats->nr_auxtrace_errors[i],
+			    auxtrace_error_name(i));
+	}
+}
+
+int perf_event__process_auxtrace_error(struct perf_tool *tool __maybe_unused,
+				       union perf_event *event,
+				       struct perf_session *session __maybe_unused)
+{
+	perf_event__fprintf_auxtrace_error(event, stdout);
+	return 0;
+}
+
 int auxtrace_mmap__read(struct auxtrace_mmap *mm, struct auxtrace_record *itr,
 			struct perf_tool *tool, process_auxtrace_t fn)
 {
diff --git a/tools/perf/util/auxtrace.h b/tools/perf/util/auxtrace.h
index 4a7b7bf..05280f1 100644
--- a/tools/perf/util/auxtrace.h
+++ b/tools/perf/util/auxtrace.h
@@ -23,6 +23,7 @@
 #include <linux/types.h>
 
 #include "../perf.h"
+#include "event.h"
 #include "session.h"
 
 union perf_event;
@@ -32,6 +33,7 @@ struct perf_tool;
 struct option;
 struct record_opts;
 struct auxtrace_info_event;
+struct events_stats;
 
 enum itrace_period_type {
 	PERF_ITRACE_PERIOD_INSTRUCTIONS,
@@ -202,6 +204,10 @@ int auxtrace_record__info_fill(struct auxtrace_record *itr,
 void auxtrace_record__free(struct auxtrace_record *itr);
 u64 auxtrace_record__reference(struct auxtrace_record *itr);
 
+void auxtrace_synth_error(struct auxtrace_error_event *auxtrace_error, int type,
+			  int code, int cpu, pid_t pid, pid_t tid, u64 ip,
+			  const char *msg);
+
 int perf_event__synthesize_auxtrace_info(struct auxtrace_record *itr,
 					 struct perf_tool *tool,
 					 struct perf_session *session,
@@ -210,10 +216,18 @@ int perf_event__synthesize_auxtrace(struct perf_tool *tool,
 				    perf_event__handler_t process,
 				    size_t size, u64 offset, u64 ref, int idx,
 				    u32 tid, u32 cpu);
+int perf_event__process_auxtrace_error(struct perf_tool *tool,
+				       union perf_event *event,
+				       struct perf_session *session);
 int itrace_parse_synth_opts(const struct option *opt, const char *str,
 			    int unset);
 void itrace_synth_opts__set_default(struct itrace_synth_opts *synth_opts);
 
+size_t perf_event__fprintf_auxtrace_error(union perf_event *event, FILE *fp);
+void perf_session__auxtrace_error_inc(struct perf_session *session,
+				      union perf_event *event);
+void events_stats__auxtrace_error_warn(const struct events_stats *stats);
+
 static inline int auxtrace__process_event(struct perf_session *session,
 					  union perf_event *event,
 					  struct perf_sample *sample,
diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h
index 02327bb..473ad39 100644
--- a/tools/perf/util/event.h
+++ b/tools/perf/util/event.h
@@ -221,6 +221,11 @@ enum perf_user_event_type { /* above any possible kernel type */
 	PERF_RECORD_HEADER_MAX
 };
 
+enum auxtrace_error_type {
+	PERF_AUXTRACE_ERROR_ITRACE  = 1,
+	PERF_AUXTRACE_ERROR_MAX
+};
+
 /*
  * The kernel collects the number of events it couldn't send in a stretch and
  * when possible sends this number in a PERF_RECORD_LOST event. The number of
@@ -246,6 +251,7 @@ struct events_stats {
 	u32 nr_unknown_id;
 	u32 nr_unprocessable_samples;
 	u32 nr_unordered_events;
+	u32 nr_auxtrace_errors[PERF_AUXTRACE_ERROR_MAX];
 };
 
 struct attr_event {
diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c
index 918554d..e13654a 100644
--- a/tools/perf/util/session.c
+++ b/tools/perf/util/session.c
@@ -1087,6 +1087,7 @@ static s64 perf_session__process_user_event(struct perf_session *session,
 		lseek(fd, file_offset + event->header.size, SEEK_SET);
 		return tool->auxtrace(tool, event, session);
 	case PERF_RECORD_AUXTRACE_ERROR:
+		perf_session__auxtrace_error_inc(session, event);
 		return tool->auxtrace_error(tool, event, session);
 	default:
 		return -EINVAL;
@@ -1274,6 +1275,8 @@ static void perf_tool__warn_about_errors(const struct perf_tool *tool,
 
 	if (stats->nr_unordered_events != 0)
 		ui__warning("%u out of order events recorded.\n", stats->nr_unordered_events);
+
+	events_stats__auxtrace_error_warn(stats);
 }
 
 volatile int session_done;
-- 
1.9.1

[PATCH V6 11/25] perf auxtrace: Add helpers for queuing AUX area tracing data

[Adrian Hunter]

Provide functions to queue AUX area tracing data
buffers for processing.  A AUX area decoder
need not use the queues, however Intel BTS and Intel PT
will use them.

There is one queue for each of the mmap buffers that
were used for recording.  Because those mmaps were
associated with per-cpu or per-thread contexts, the
data is time-ordered with respect to those contexts.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/util/auxtrace.c | 305 +++++++++++++++++++++++++++++++++++++++++++++
 tools/perf/util/auxtrace.h |  87 +++++++++++++
 2 files changed, 392 insertions(+)

diff --git a/tools/perf/util/auxtrace.c b/tools/perf/util/auxtrace.c
index cdcd6ae..7e242b8 100644
--- a/tools/perf/util/auxtrace.c
+++ b/tools/perf/util/auxtrace.c
@@ -22,11 +22,15 @@
 #include <linux/types.h>
 #include <linux/bitops.h>
 #include <linux/log2.h>
+#include <linux/string.h>
 
+#include <sys/param.h>
 #include <stdlib.h>
 #include <stdio.h>
 #include <string.h>
+#include <limits.h>
 #include <errno.h>
+#include <linux/list.h>
 
 #include "../perf.h"
 #include "util.h"
@@ -112,6 +116,241 @@ void auxtrace_mmap_params__set_idx(struct auxtrace_mmap_params *mp,
 	}
 }
 
+#define AUXTRACE_INIT_NR_QUEUES	32
+
+static struct auxtrace_queue *auxtrace_alloc_queue_array(unsigned int nr_queues)
+{
+	struct auxtrace_queue *queue_array;
+	unsigned int max_nr_queues, i;
+
+	max_nr_queues = MIN(UINT_MAX, SIZE_MAX) / sizeof(struct auxtrace_queue);
+	if (nr_queues > max_nr_queues)
+		return NULL;
+
+	queue_array = calloc(nr_queues, sizeof(struct auxtrace_queue));
+	if (!queue_array)
+		return NULL;
+
+	for (i = 0; i < nr_queues; i++) {
+		INIT_LIST_HEAD(&queue_array[i].head);
+		queue_array[i].priv = NULL;
+	}
+
+	return queue_array;
+}
+
+int auxtrace_queues__init(struct auxtrace_queues *queues)
+{
+	queues->nr_queues = AUXTRACE_INIT_NR_QUEUES;
+	queues->queue_array = auxtrace_alloc_queue_array(queues->nr_queues);
+	if (!queues->queue_array)
+		return -ENOMEM;
+	return 0;
+}
+
+static int auxtrace_queues__grow(struct auxtrace_queues *queues,
+				 unsigned int new_nr_queues)
+{
+	unsigned int nr_queues = queues->nr_queues;
+	struct auxtrace_queue *queue_array;
+	unsigned int i;
+
+	if (!nr_queues)
+		nr_queues = AUXTRACE_INIT_NR_QUEUES;
+
+	while (nr_queues && nr_queues < new_nr_queues)
+		nr_queues <<= 1;
+
+	if (nr_queues < queues->nr_queues || nr_queues < new_nr_queues)
+		return -EINVAL;
+
+	queue_array = auxtrace_alloc_queue_array(nr_queues);
+	if (!queue_array)
+		return -ENOMEM;
+
+	for (i = 0; i < queues->nr_queues; i++) {
+		list_splice_tail(&queues->queue_array[i].head,
+				 &queue_array[i].head);
+		queue_array[i].priv = queues->queue_array[i].priv;
+	}
+
+	queues->nr_queues = nr_queues;
+	queues->queue_array = queue_array;
+
+	return 0;
+}
+
+static void *auxtrace_copy_data(u64 size, struct perf_session *session)
+{
+	int fd = perf_data_file__fd(session->file);
+	void *p;
+	ssize_t ret;
+
+	if (size > SSIZE_MAX)
+		return NULL;
+
+	p = malloc(size);
+	if (!p)
+		return NULL;
+
+	ret = readn(fd, p, size);
+	if (ret != (ssize_t)size) {
+		free(p);
+		return NULL;
+	}
+
+	return p;
+}
+
+static int auxtrace_queues__add_buffer(struct auxtrace_queues *queues,
+				       unsigned int idx,
+				       struct auxtrace_buffer *buffer)
+{
+	struct auxtrace_queue *queue;
+	int err;
+
+	if (idx >= queues->nr_queues) {
+		err = auxtrace_queues__grow(queues, idx + 1);
+		if (err)
+			return err;
+	}
+
+	queue = &queues->queue_array[idx];
+
+	if (!queue->set) {
+		queue->set = true;
+		queue->tid = buffer->tid;
+		queue->cpu = buffer->cpu;
+	} else if (buffer->cpu != queue->cpu || buffer->tid != queue->tid) {
+		pr_err("auxtrace queue conflict: cpu %d, tid %d vs cpu %d, tid %d\n",
+		       queue->cpu, queue->tid, buffer->cpu, buffer->tid);
+		return -EINVAL;
+	}
+
+	buffer->buffer_nr = queues->next_buffer_nr++;
+
+	list_add_tail(&buffer->list, &queue->head);
+
+	queues->new_data = true;
+	queues->populated = true;
+
+	return 0;
+}
+
+/* Limit buffers to 32MiB on 32-bit */
+#define BUFFER_LIMIT_FOR_32_BIT (32 * 1024 * 1024)
+
+static int auxtrace_queues__split_buffer(struct auxtrace_queues *queues,
+					 unsigned int idx,
+					 struct auxtrace_buffer *buffer)
+{
+	u64 sz = buffer->size;
+	bool consecutive = false;
+	struct auxtrace_buffer *b;
+	int err;
+
+	while (sz > BUFFER_LIMIT_FOR_32_BIT) {
+		b = memdup(buffer, sizeof(struct auxtrace_buffer));
+		if (!b)
+			return -ENOMEM;
+		b->size = BUFFER_LIMIT_FOR_32_BIT;
+		b->consecutive = consecutive;
+		err = auxtrace_queues__add_buffer(queues, idx, b);
+		if (err) {
+			auxtrace_buffer__free(b);
+			return err;
+		}
+		buffer->data_offset += BUFFER_LIMIT_FOR_32_BIT;
+		sz -= BUFFER_LIMIT_FOR_32_BIT;
+		consecutive = true;
+	}
+
+	buffer->size = sz;
+	buffer->consecutive = consecutive;
+
+	return 0;
+}
+
+static int auxtrace_queues__add_event_buffer(struct auxtrace_queues *queues,
+					     struct perf_session *session,
+					     unsigned int idx,
+					     struct auxtrace_buffer *buffer)
+{
+	if (session->one_mmap) {
+		buffer->data = buffer->data_offset - session->one_mmap_offset +
+			       session->one_mmap_addr;
+	} else if (perf_data_file__is_pipe(session->file)) {
+		buffer->data = auxtrace_copy_data(buffer->size, session);
+		if (!buffer->data)
+			return -ENOMEM;
+		buffer->data_needs_freeing = true;
+	} else if (BITS_PER_LONG == 32 &&
+		   buffer->size > BUFFER_LIMIT_FOR_32_BIT) {
+		int err;
+
+		err = auxtrace_queues__split_buffer(queues, idx, buffer);
+		if (err)
+			return err;
+	}
+
+	return auxtrace_queues__add_buffer(queues, idx, buffer);
+}
+
+int auxtrace_queues__add_event(struct auxtrace_queues *queues,
+			       struct perf_session *session,
+			       union perf_event *event, off_t data_offset,
+			       struct auxtrace_buffer **buffer_ptr)
+{
+	struct auxtrace_buffer *buffer;
+	unsigned int idx;
+	int err;
+
+	buffer = zalloc(sizeof(struct auxtrace_buffer));
+	if (!buffer)
+		return -ENOMEM;
+
+	buffer->pid = -1;
+	buffer->tid = event->auxtrace.tid;
+	buffer->cpu = event->auxtrace.cpu;
+	buffer->data_offset = data_offset;
+	buffer->offset = event->auxtrace.offset;
+	buffer->reference = event->auxtrace.reference;
+	buffer->size = event->auxtrace.size;
+	idx = event->auxtrace.idx;
+
+	err = auxtrace_queues__add_event_buffer(queues, session, idx, buffer);
+	if (err)
+		goto out_err;
+
+	if (buffer_ptr)
+		*buffer_ptr = buffer;
+
+	return 0;
+
+out_err:
+	auxtrace_buffer__free(buffer);
+	return err;
+}
+
+void auxtrace_queues__free(struct auxtrace_queues *queues)
+{
+	unsigned int i;
+
+	for (i = 0; i < queues->nr_queues; i++) {
+		while (!list_empty(&queues->queue_array[i].head)) {
+			struct auxtrace_buffer *buffer;
+
+			buffer = list_entry(queues->queue_array[i].head.next,
+					    struct auxtrace_buffer, list);
+			list_del(&buffer->list);
+			auxtrace_buffer__free(buffer);
+		}
+	}
+
+	zfree(&queues->queue_array);
+	queues->nr_queues = 0;
+}
+
 size_t auxtrace_record__info_priv_size(struct auxtrace_record *itr)
 {
 	if (itr)
@@ -164,6 +403,72 @@ auxtrace_record__init(struct perf_evlist *evlist __maybe_unused, int *err)
 	return NULL;
 }
 
+struct auxtrace_buffer *auxtrace_buffer__next(struct auxtrace_queue *queue,
+					      struct auxtrace_buffer *buffer)
+{
+	if (buffer) {
+		if (list_is_last(&buffer->list, &queue->head))
+			return NULL;
+		return list_entry(buffer->list.next, struct auxtrace_buffer,
+				  list);
+	} else {
+		if (list_empty(&queue->head))
+			return NULL;
+		return list_entry(queue->head.next, struct auxtrace_buffer,
+				  list);
+	}
+}
+
+void *auxtrace_buffer__get_data(struct auxtrace_buffer *buffer, int fd)
+{
+	size_t adj = buffer->data_offset & (page_size - 1);
+	size_t size = buffer->size + adj;
+	off_t file_offset = buffer->data_offset - adj;
+	void *addr;
+
+	if (buffer->data)
+		return buffer->data;
+
+	addr = mmap(NULL, size, PROT_READ, MAP_SHARED, fd, file_offset);
+	if (addr == MAP_FAILED)
+		return NULL;
+
+	buffer->mmap_addr = addr;
+	buffer->mmap_size = size;
+
+	buffer->data = addr + adj;
+
+	return buffer->data;
+}
+
+void auxtrace_buffer__put_data(struct auxtrace_buffer *buffer)
+{
+	if (!buffer->data || !buffer->mmap_addr)
+		return;
+	munmap(buffer->mmap_addr, buffer->mmap_size);
+	buffer->mmap_addr = NULL;
+	buffer->mmap_size = 0;
+	buffer->data = NULL;
+	buffer->use_data = NULL;
+}
+
+void auxtrace_buffer__drop_data(struct auxtrace_buffer *buffer)
+{
+	auxtrace_buffer__put_data(buffer);
+	if (buffer->data_needs_freeing) {
+		buffer->data_needs_freeing = false;
+		zfree(&buffer->data);
+		buffer->use_data = NULL;
+		buffer->size = 0;
+	}
+}
+
+void auxtrace_buffer__free(struct auxtrace_buffer *buffer)
+{
+	auxtrace_buffer__drop_data(buffer);
+	free(buffer);
+}
+
 void auxtrace_synth_error(struct auxtrace_error_event *auxtrace_error, int type,
 			  int code, int cpu, pid_t pid, pid_t tid, u64 ip,
 			  const char *msg)
diff --git a/tools/perf/util/auxtrace.h b/tools/perf/util/auxtrace.h
index 05280f1..3a093f7 100644
--- a/tools/perf/util/auxtrace.h
+++ b/tools/perf/util/auxtrace.h
@@ -19,6 +19,7 @@
 #include <sys/types.h>
 #include <stdbool.h>
 #include <stddef.h>
+#include <linux/list.h>
 #include <linux/perf_event.h>
 #include <linux/types.h>
 
@@ -93,6 +94,80 @@ struct auxtrace {
 };
 
 /**
+ * struct auxtrace_buffer - a buffer containing AUX area tracing data.
+ * @list: buffers are queued in a list held by struct auxtrace_queue
+ * @size: size of the buffer in bytes
+ * @pid: in per-thread mode, the pid this buffer is associated with
+ * @tid: in per-thread mode, the tid this buffer is associated with
+ * @cpu: in per-cpu mode, the cpu this buffer is associated with
+ * @data: actual buffer data (can be null if the data has not been loaded)
+ * @data_offset: file offset at which the buffer can be read
+ * @mmap_addr: mmap address at which the buffer can be read
+ * @mmap_size: size of the mmap at @mmap_addr
+ * @data_needs_freeing: @data was malloc'd so free it when it is no longer
+ *                      needed
+ * @consecutive: the original data was split up and this buffer is consecutive
+ *               to the previous buffer
+ * @offset: offset as determined by aux_head / aux_tail members of struct
+ *          perf_event_mmap_page
+ * @reference: an implementation-specific reference determined when the data is
+ *             recorded
+ * @buffer_nr: used to number each buffer
+ * @use_size: implementation actually only uses this number of bytes
+ * @use_data: implementation actually only uses data starting at this address
+ */
+struct auxtrace_buffer {
+	struct list_head	list;
+	size_t			size;
+	pid_t			pid;
+	pid_t			tid;
+	int			cpu;
+	void			*data;
+	off_t			data_offset;
+	void			*mmap_addr;
+	size_t			mmap_size;
+	bool			data_needs_freeing;
+	bool			consecutive;
+	u64			offset;
+	u64			reference;
+	u64			buffer_nr;
+	size_t			use_size;
+	void			*use_data;
+};
+
+/**
+ * struct auxtrace_queue - a queue of AUX area tracing data buffers.
+ * @head: head of buffer list
+ * @tid: in per-thread mode, the tid this queue is associated with
+ * @cpu: in per-cpu mode, the cpu this queue is associated with
+ * @set: %true once this queue has been dedicated to a specific thread or cpu
+ * @priv: implementation-specific data
+ */
+struct auxtrace_queue {
+	struct list_head	head;
+	pid_t			tid;
+	int			cpu;
+	bool			set;
+	void			*priv;
+};
+
+/**
+ * struct auxtrace_queues - an array of AUX area tracing queues.
+ * @queue_array: array of queues
+ * @nr_queues: number of queues
+ * @new_data: set whenever new data is queued
+ * @populated: queues have been fully populated using the auxtrace_index
+ * @next_buffer_nr: used to number each buffer
+ */
+struct auxtrace_queues {
+	struct auxtrace_queue	*queue_array;
+	unsigned int		nr_queues;
+	bool			new_data;
+	bool			populated;
+	u64			next_buffer_nr;
+};
+
+/**
  * struct auxtrace_mmap - records an mmap of the auxtrace buffer.
  * @base: address of mapped area
  * @userpg: pointer to buffer's perf_event_mmap_page
@@ -190,6 +265,18 @@ typedef int (*process_auxtrace_t)(struct perf_tool *tool,
 int auxtrace_mmap__read(struct auxtrace_mmap *mm, struct auxtrace_record *itr,
 			struct perf_tool *tool, process_auxtrace_t fn);
 
+int auxtrace_queues__init(struct auxtrace_queues *queues);
+int auxtrace_queues__add_event(struct auxtrace_queues *queues,
+			       struct perf_session *session,
+			       union perf_event *event, off_t data_offset,
+			       struct auxtrace_buffer **buffer_ptr);
+void auxtrace_queues__free(struct auxtrace_queues *queues);
+struct auxtrace_buffer *auxtrace_buffer__next(struct auxtrace_queue *queue,
+					      struct auxtrace_buffer *buffer);
+void *auxtrace_buffer__get_data(struct auxtrace_buffer *buffer, int fd);
+void auxtrace_buffer__put_data(struct auxtrace_buffer *buffer);
+void auxtrace_buffer__drop_data(struct auxtrace_buffer *buffer);
+void auxtrace_buffer__free(struct auxtrace_buffer *buffer);
 struct auxtrace_record *auxtrace_record__init(struct perf_evlist *evlist,
 					      int *err);
 
-- 
1.9.1

[PATCH V6 12/25] perf auxtrace: Add a heap for sorting AUX area tracing queues

[Adrian Hunter]

In order to process AUX area tracing
data in time order, the queue with data
with the lowest timestamp must be
processed first.  Provide a heap to
keep track of which queue that is.

As with the queues, a decoder does not have
to use the heap, but Intel BTS and Intel PT
will use it.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/util/auxtrace.c | 85 ++++++++++++++++++++++++++++++++++++++++++++++
 tools/perf/util/auxtrace.h | 29 ++++++++++++++++
 2 files changed, 114 insertions(+)

diff --git a/tools/perf/util/auxtrace.c b/tools/perf/util/auxtrace.c
index 7e242b8..225bf7e 100644
--- a/tools/perf/util/auxtrace.c
+++ b/tools/perf/util/auxtrace.c
@@ -351,6 +351,91 @@ void auxtrace_queues__free(struct auxtrace_queues *queues)
 	queues->nr_queues = 0;
 }
 
+static void auxtrace_heapify(struct auxtrace_heap_item *heap_array,
+			     unsigned int pos, unsigned int queue_nr,
+			     u64 ordinal)
+{
+	unsigned int parent;
+
+	while (pos) {
+		parent = (pos - 1) >> 1;
+		if (heap_array[parent].ordinal <= ordinal)
+			break;
+		heap_array[pos] = heap_array[parent];
+		pos = parent;
+	}
+	heap_array[pos].queue_nr = queue_nr;
+	heap_array[pos].ordinal = ordinal;
+}
+
+int auxtrace_heap__add(struct auxtrace_heap *heap, unsigned int queue_nr,
+		       u64 ordinal)
+{
+	struct auxtrace_heap_item *heap_array;
+
+	if (queue_nr >= heap->heap_sz) {
+		unsigned int heap_sz = AUXTRACE_INIT_NR_QUEUES;
+
+		while (heap_sz <= queue_nr)
+			heap_sz <<= 1;
+		heap_array = realloc(heap->heap_array,
+				     heap_sz * sizeof(struct auxtrace_heap_item));
+		if (!heap_array)
+			return -ENOMEM;
+		heap->heap_array = heap_array;
+		heap->heap_sz = heap_sz;
+	}
+
+	auxtrace_heapify(heap->heap_array, heap->heap_cnt++, queue_nr, ordinal);
+
+	return 0;
+}
+
+void auxtrace_heap__free(struct auxtrace_heap *heap)
+{
+	zfree(&heap->heap_array);
+	heap->heap_cnt = 0;
+	heap->heap_sz = 0;
+}
+
+void auxtrace_heap__pop(struct auxtrace_heap *heap)
+{
+	unsigned int pos, last, heap_cnt = heap->heap_cnt;
+	struct auxtrace_heap_item *heap_array;
+
+	if (!heap_cnt)
+		return;
+
+	heap->heap_cnt -= 1;
+
+	heap_array = heap->heap_array;
+
+	pos = 0;
+	while (1) {
+		unsigned int left, right;
+
+		left = (pos << 1) + 1;
+		if (left >= heap_cnt)
+			break;
+		right = left + 1;
+		if (right >= heap_cnt) {
+			heap_array[pos] = heap_array[left];
+			return;
+		}
+		if (heap_array[left].ordinal < heap_array[right].ordinal) {
+			heap_array[pos] = heap_array[left];
+			pos = left;
+		} else {
+			heap_array[pos] = heap_array[right];
+			pos = right;
+		}
+	}
+
+	last = heap_cnt - 1;
+	auxtrace_heapify(heap_array, pos, heap_array[last].queue_nr,
+			 heap_array[last].ordinal);
+}
+
 size_t auxtrace_record__info_priv_size(struct auxtrace_record *itr)
 {
 	if (itr)
diff --git a/tools/perf/util/auxtrace.h b/tools/perf/util/auxtrace.h
index 3a093f7..984ce57 100644
--- a/tools/perf/util/auxtrace.h
+++ b/tools/perf/util/auxtrace.h
@@ -168,6 +168,29 @@ struct auxtrace_queues {
 };
 
 /**
+ * struct auxtrace_heap_item - element of struct auxtrace_heap.
+ * @queue_nr: queue number
+ * @ordinal: value used for sorting (lowest ordinal is top of the heap) expected
+ *           to be a timestamp
+ */
+struct auxtrace_heap_item {
+	unsigned int		queue_nr;
+	u64			ordinal;
+};
+
+/**
+ * struct auxtrace_heap - a heap suitable for sorting AUX area tracing queues.
+ * @heap_array: the heap
+ * @heap_cnt: the number of elements in the heap
+ * @heap_sz: maximum number of elements (grows as needed)
+ */
+struct auxtrace_heap {
+	struct auxtrace_heap_item	*heap_array;
+	unsigned int		heap_cnt;
+	unsigned int		heap_sz;
+};
+
+/**
  * struct auxtrace_mmap - records an mmap of the auxtrace buffer.
  * @base: address of mapped area
  * @userpg: pointer to buffer's perf_event_mmap_page
@@ -277,6 +300,12 @@ void *auxtrace_buffer__get_data(struct auxtrace_buffer *buffer, int fd);
 void auxtrace_buffer__put_data(struct auxtrace_buffer *buffer);
 void auxtrace_buffer__drop_data(struct auxtrace_buffer *buffer);
 void auxtrace_buffer__free(struct auxtrace_buffer *buffer);
+
+int auxtrace_heap__add(struct auxtrace_heap *heap, unsigned int queue_nr,
+		       u64 ordinal);
+void auxtrace_heap__pop(struct auxtrace_heap *heap);
+void auxtrace_heap__free(struct auxtrace_heap *heap);
+
 struct auxtrace_record *auxtrace_record__init(struct perf_evlist *evlist,
 					      int *err);
 
-- 
1.9.1

[PATCH V6 13/25] perf auxtrace: Add processing for AUX area tracing events

[Adrian Hunter]

Provide hooks so that an AUX area
decoder can process AUX area tracing
events.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/util/auxtrace.c | 52 +++++++++++++++++++++++++++++++++++++++++++++-
 tools/perf/util/auxtrace.h | 13 ++++++++++++
 2 files changed, 64 insertions(+), 1 deletion(-)

diff --git a/tools/perf/util/auxtrace.c b/tools/perf/util/auxtrace.c
index 225bf7e..1c49a6c 100644
--- a/tools/perf/util/auxtrace.c
+++ b/tools/perf/util/auxtrace.c
@@ -605,6 +605,28 @@ out_free:
 	return err;
 }
 
+static bool auxtrace__dont_decode(struct perf_session *session)
+{
+	return !session->itrace_synth_opts ||
+	       session->itrace_synth_opts->dont_decode;
+}
+
+int perf_event__process_auxtrace_info(struct perf_tool *tool __maybe_unused,
+				      union perf_event *event,
+				      struct perf_session *session __maybe_unused)
+{
+	enum auxtrace_type type = event->auxtrace_info.type;
+
+	if (dump_trace)
+		fprintf(stdout, " type: %u\n", type);
+
+	switch (type) {
+	case PERF_AUXTRACE_UNKNOWN:
+	default:
+		return -EINVAL;
+	}
+}
+
 int perf_event__synthesize_auxtrace(struct perf_tool *tool,
 				    perf_event__handler_t process,
 				    size_t size, u64 offset, u64 ref, int idx,
@@ -625,6 +647,31 @@ int perf_event__synthesize_auxtrace(struct perf_tool *tool,
 	return process(tool, &ev, NULL, NULL);
 }
 
+s64 perf_event__process_auxtrace(struct perf_tool *tool,
+				 union perf_event *event,
+				 struct perf_session *session)
+{
+	s64 err;
+
+	if (dump_trace)
+		fprintf(stdout, " size: %#"PRIx64"  offset: %#"PRIx64"  ref: %#"PRIx64"  idx: %u  tid: %d  cpu: %d\n",
+			event->auxtrace.size, event->auxtrace.offset,
+			event->auxtrace.reference, event->auxtrace.idx,
+			event->auxtrace.tid, event->auxtrace.cpu);
+
+	if (auxtrace__dont_decode(session))
+		return event->auxtrace.size;
+
+	if (!session->auxtrace || event->header.type != PERF_RECORD_AUXTRACE)
+		return -EINVAL;
+
+	err = session->auxtrace->process_auxtrace_event(session, event, tool);
+	if (err < 0)
+		return err;
+
+	return event->auxtrace.size;
+}
+
 #define PERF_ITRACE_DEFAULT_PERIOD_TYPE		PERF_ITRACE_PERIOD_NANOSECS
 #define PERF_ITRACE_DEFAULT_PERIOD		100000
 #define PERF_ITRACE_DEFAULT_CALLCHAIN_SZ	16
@@ -806,8 +853,11 @@ void events_stats__auxtrace_error_warn(const struct events_stats *stats)
 
 int perf_event__process_auxtrace_error(struct perf_tool *tool __maybe_unused,
 				       union perf_event *event,
-				       struct perf_session *session __maybe_unused)
+				       struct perf_session *session)
 {
+	if (auxtrace__dont_decode(session))
+		return 0;
+
 	perf_event__fprintf_auxtrace_error(event, stdout);
 	return 0;
 }
diff --git a/tools/perf/util/auxtrace.h b/tools/perf/util/auxtrace.h
index 984ce57..882fb35 100644
--- a/tools/perf/util/auxtrace.h
+++ b/tools/perf/util/auxtrace.h
@@ -36,6 +36,10 @@ struct record_opts;
 struct auxtrace_info_event;
 struct events_stats;
 
+enum auxtrace_type {
+	PERF_AUXTRACE_UNKNOWN,
+};
+
 enum itrace_period_type {
 	PERF_ITRACE_PERIOD_INSTRUCTIONS,
 	PERF_ITRACE_PERIOD_TICKS,
@@ -87,6 +91,9 @@ struct auxtrace {
 			     union perf_event *event,
 			     struct perf_sample *sample,
 			     struct perf_tool *tool);
+	int (*process_auxtrace_event)(struct perf_session *session,
+				      union perf_event *event,
+				      struct perf_tool *tool);
 	int (*flush_events)(struct perf_session *session,
 			    struct perf_tool *tool);
 	void (*free_events)(struct perf_session *session);
@@ -328,10 +335,16 @@ int perf_event__synthesize_auxtrace_info(struct auxtrace_record *itr,
 					 struct perf_tool *tool,
 					 struct perf_session *session,
 					 perf_event__handler_t process);
+int perf_event__process_auxtrace_info(struct perf_tool *tool,
+				      union perf_event *event,
+				      struct perf_session *session);
 int perf_event__synthesize_auxtrace(struct perf_tool *tool,
 				    perf_event__handler_t process,
 				    size_t size, u64 offset, u64 ref, int idx,
 				    u32 tid, u32 cpu);
+s64 perf_event__process_auxtrace(struct perf_tool *tool,
+				 union perf_event *event,
+				 struct perf_session *session);
 int perf_event__process_auxtrace_error(struct perf_tool *tool,
 				       union perf_event *event,
 				       struct perf_session *session);
-- 
1.9.1

[PATCH V6 14/25] perf auxtrace: Add a hashtable for caching

[Adrian Hunter]

Decoding AUX area data may involve walking object
code.  Rather than repetitively decoding the same
instructions, a cache can be used to cache the results.

This patch implements a fairly generic hashtable
with a 32-bit key that could be used for other
purposes as well.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/util/auxtrace.c | 123 +++++++++++++++++++++++++++++++++++++++++++++
 tools/perf/util/auxtrace.h |  14 ++++++
 2 files changed, 137 insertions(+)

diff --git a/tools/perf/util/auxtrace.c b/tools/perf/util/auxtrace.c
index 1c49a6c..e1499c5 100644
--- a/tools/perf/util/auxtrace.c
+++ b/tools/perf/util/auxtrace.c
@@ -39,6 +39,8 @@
 #include "thread_map.h"
 #include "auxtrace.h"
 
+#include <linux/hash.h>
+
 #include "event.h"
 #include "session.h"
 #include "debug.h"
@@ -949,3 +951,124 @@ int auxtrace_mmap__read(struct auxtrace_mmap *mm, struct auxtrace_record *itr,
 
 	return 1;
 }
+
+/**
+ * struct auxtrace_cache - hash table to implement a cache
+ * @hashtable: the hashtable
+ * @sz: hashtable size (number of hlists)
+ * @entry_size: size of an entry
+ * @limit: limit the number of entries to this maximum, when reached the cache
+ *         is dropped and caching begins again with an empty cache
+ * @cnt: current number of entries
+ * @bits: hashtable size (@sz = 2^@bits)
+ */
+struct auxtrace_cache {
+	struct hlist_head *hashtable;
+	size_t sz;
+	size_t entry_size;
+	size_t limit;
+	size_t cnt;
+	unsigned int bits;
+};
+
+struct auxtrace_cache *auxtrace_cache__new(unsigned int bits, size_t entry_size,
+					   unsigned int limit_percent)
+{
+	struct auxtrace_cache *c;
+	struct hlist_head *ht;
+	size_t sz, i;
+
+	c = zalloc(sizeof(struct auxtrace_cache));
+	if (!c)
+		return NULL;
+
+	sz = 1UL << bits;
+
+	ht = calloc(sz, sizeof(struct hlist_head));
+	if (!ht)
+		goto out_free;
+
+	for (i = 0; i < sz; i++)
+		INIT_HLIST_HEAD(&ht[i]);
+
+	c->hashtable = ht;
+	c->sz = sz;
+	c->entry_size = entry_size;
+	c->limit = (c->sz * limit_percent) / 100;
+	c->bits = bits;
+
+	return c;
+
+out_free:
+	free(c);
+	return NULL;
+}
+
+static void auxtrace_cache__drop(struct auxtrace_cache *c)
+{
+	struct auxtrace_cache_entry *entry;
+	struct hlist_node *tmp;
+	size_t i;
+
+	if (!c)
+		return;
+
+	for (i = 0; i < c->sz; i++) {
+		hlist_for_each_entry_safe(entry, tmp, &c->hashtable[i], hash) {
+			hlist_del(&entry->hash);
+			auxtrace_cache__free_entry(c, entry);
+		}
+	}
+
+	c->cnt = 0;
+}
+
+void auxtrace_cache__free(struct auxtrace_cache *c)
+{
+	if (!c)
+		return;
+
+	auxtrace_cache__drop(c);
+	free(c->hashtable);
+	free(c);
+}
+
+void *auxtrace_cache__alloc_entry(struct auxtrace_cache *c)
+{
+	return malloc(c->entry_size);
+}
+
+void auxtrace_cache__free_entry(struct auxtrace_cache *c __maybe_unused,
+				void *entry)
+{
+	free(entry);
+}
+
+int auxtrace_cache__add(struct auxtrace_cache *c, u32 key,
+			struct auxtrace_cache_entry *entry)
+{
+	if (c->limit && ++c->cnt > c->limit)
+		auxtrace_cache__drop(c);
+
+	entry->key = key;
+	hlist_add_head(&entry->hash, &c->hashtable[hash_32(key, c->bits)]);
+
+	return 0;
+}
+
+void *auxtrace_cache__lookup(struct auxtrace_cache *c, u32 key)
+{
+	struct auxtrace_cache_entry *entry;
+	struct hlist_head *hlist;
+
+	if (!c)
+		return NULL;
+
+	hlist = &c->hashtable[hash_32(key, c->bits)];
+	hlist_for_each_entry(entry, hlist, hash) {
+		if (entry->key == key)
+			return entry;
+	}
+
+	return NULL;
+}
diff --git a/tools/perf/util/auxtrace.h b/tools/perf/util/auxtrace.h
index 882fb35..5dfa05a 100644
--- a/tools/perf/util/auxtrace.h
+++ b/tools/perf/util/auxtrace.h
@@ -313,6 +313,20 @@ int auxtrace_heap__add(struct auxtrace_heap *heap, unsigned int queue_nr,
 void auxtrace_heap__pop(struct auxtrace_heap *heap);
 void auxtrace_heap__free(struct auxtrace_heap *heap);
 
+struct auxtrace_cache_entry {
+	struct hlist_node hash;
+	u32 key;
+};
+
+struct auxtrace_cache *auxtrace_cache__new(unsigned int bits, size_t entry_size,
+					   unsigned int limit_percent);
+void auxtrace_cache__free(struct auxtrace_cache *auxtrace_cache);
+void *auxtrace_cache__alloc_entry(struct auxtrace_cache *c);
+void auxtrace_cache__free_entry(struct auxtrace_cache *c, void *entry);
+int auxtrace_cache__add(struct auxtrace_cache *c, u32 key,
+			struct auxtrace_cache_entry *entry);
+void *auxtrace_cache__lookup(struct auxtrace_cache *c, u32 key);
+
 struct auxtrace_record *auxtrace_record__init(struct perf_evlist *evlist,
 					      int *err);
 
-- 
1.9.1

[PATCH V6 15/25] perf tools: Add member to struct dso for an instruction cache

[Adrian Hunter]

Add a member to struct dso that can be used by Instruction
Trace implementations to hold a cache for decoded instructions.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/util/dso.c | 2 ++
 tools/perf/util/dso.h | 3 +++
 2 files changed, 5 insertions(+)

diff --git a/tools/perf/util/dso.c b/tools/perf/util/dso.c
index 814554d..247693df 100644
--- a/tools/perf/util/dso.c
+++ b/tools/perf/util/dso.c
@@ -4,6 +4,7 @@
 #include "symbol.h"
 #include "dso.h"
 #include "machine.h"
+#include "auxtrace.h"
 #include "util.h"
 #include "debug.h"
 
@@ -910,6 +911,7 @@ void dso__delete(struct dso *dso)
 	}
 
 	dso__data_close(dso);
+	auxtrace_cache__free(dso->auxtrace_cache);
 	dso_cache__free(&dso->data.cache);
 	dso__free_a2l(dso);
 	zfree(&dso->symsrc_filename);
diff --git a/tools/perf/util/dso.h b/tools/perf/util/dso.h
index ced9284..7ee28f9 100644
--- a/tools/perf/util/dso.h
+++ b/tools/perf/util/dso.h
@@ -101,6 +101,8 @@ struct dsos {
 	struct rb_root	 root;	/* rbtree root sorted by long name */
 };
 
+struct auxtrace_cache;
+
 struct dso {
 	struct list_head node;
 	struct rb_node	 rb_node;	/* rbtree node sorted by long name */
@@ -130,6 +132,7 @@ struct dso {
 	u16		 long_name_len;
 	u16		 short_name_len;
 	void		*dwfl;			/* DWARF debug info */
+	struct auxtrace_cache *auxtrace_cache;
 
 	/* dso data file */
 	struct {
-- 
1.9.1

[PATCH V6 16/25] perf script: Add Instruction Tracing support

[Adrian Hunter]

Add support for decoding an AUX area assuming
it contains instruction tracing data.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/Documentation/perf-script.txt | 28 ++++++++++++++++++++++++++++
 tools/perf/builtin-script.c              | 11 +++++++++++
 2 files changed, 39 insertions(+)

diff --git a/tools/perf/Documentation/perf-script.txt b/tools/perf/Documentation/perf-script.txt
index a21eec0..88da93a 100644
--- a/tools/perf/Documentation/perf-script.txt
+++ b/tools/perf/Documentation/perf-script.txt
@@ -215,6 +215,34 @@ OPTIONS
 --header-only
 	Show only perf.data header.
 
+-Z::
+--itrace::
+	Options for decoding instruction tracing data. The options are:
+
+		i	synthesize instructions events
+		b	synthesize branches events
+		c	synthesize branches events (calls only)
+		r	synthesize branches events (returns only)
+		e	synthesize error events
+		d	create a debug log
+		g	synthesize a call chain for instructions events
+
+	The default is all events i.e. the same as -Zibe
+
+	In addition, the period (default 100000) for instructions events
+	can be specified in units of:
+
+		i	instructions
+		t	ticks
+		ms	milliseconds
+		us	microseconds
+		ns	nanoseconds (default)
+
+	Also the call chain size (default 16, max. 1024) for instructions
+	events can be specified.
+
+	To disable decoding entirely, use --no-itrace.
+
 SEE ALSO
 --------
 linkperf:perf-record[1], linkperf:perf-script-perl[1],
diff --git a/tools/perf/builtin-script.c b/tools/perf/builtin-script.c
index c7e6750..93fde97 100644
--- a/tools/perf/builtin-script.c
+++ b/tools/perf/builtin-script.c
@@ -16,6 +16,7 @@
 #include "util/evsel.h"
 #include "util/sort.h"
 #include "util/data.h"
+#include "util/auxtrace.h"
 #include <linux/bitmap.h>
 
 static char const		*script_name;
@@ -1505,6 +1506,7 @@ int cmd_script(int argc, const char **argv, const char *prefix __maybe_unused)
 	char *rec_script_path = NULL;
 	char *rep_script_path = NULL;
 	struct perf_session *session;
+	struct itrace_synth_opts itrace_synth_opts = {0};
 	char *script_path = NULL;
 	const char **__argv;
 	int i, j, err = 0;
@@ -1519,6 +1521,10 @@ int cmd_script(int argc, const char **argv, const char *prefix __maybe_unused)
 			.attr		 = process_attr,
 			.tracing_data	 = perf_event__process_tracing_data,
 			.build_id	 = perf_event__process_build_id,
+			.id_index	 = perf_event__process_id_index,
+			.auxtrace_info	 = perf_event__process_auxtrace_info,
+			.auxtrace	 = perf_event__process_auxtrace,
+			.auxtrace_error	 = perf_event__process_auxtrace_error,
 			.ordered_events	 = true,
 			.ordering_requires_timestamps = true,
 		},
@@ -1570,6 +1576,9 @@ int cmd_script(int argc, const char **argv, const char *prefix __maybe_unused)
 		    "Show the fork/comm/exit events"),
 	OPT_BOOLEAN('\0', "show-mmap-events", &script.show_mmap_events,
 		    "Show the mmap events"),
+	OPT_CALLBACK_OPTARG('Z', "itrace", &itrace_synth_opts, NULL, "opts",
+			    "Instruction Tracing options",
+			    itrace_parse_synth_opts),
 	OPT_END()
 	};
 	const char * const script_usage[] = {
@@ -1767,6 +1776,8 @@ int cmd_script(int argc, const char **argv, const char *prefix __maybe_unused)
 
 	script.session = session;
 
+	session->itrace_synth_opts = &itrace_synth_opts;
+
 	if (cpu_list) {
 		err = perf_session__cpu_bitmap(session, cpu_list, cpu_bitmap);
 		if (err < 0)
-- 
1.9.1

[PATCH V6 17/25] perf script: Always allow fields 'addr' and 'cpu' for auxtrace

[Adrian Hunter]

If a file contains AUX area tracing data then always allow
fields 'addr' and 'cpu' to be selected as options for perf
script.  This is necessary because AUX area decoding
may synthesize events with that information.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/builtin-script.c | 29 ++++++++++++++++++++++-------
 1 file changed, 22 insertions(+), 7 deletions(-)

diff --git a/tools/perf/builtin-script.c b/tools/perf/builtin-script.c
index 93fde97..d73b55a 100644
--- a/tools/perf/builtin-script.c
+++ b/tools/perf/builtin-script.c
@@ -147,9 +147,10 @@ static const char *output_field2str(enum perf_output_field field)
 
 #define PRINT_FIELD(x)  (output[attr->type].fields & PERF_OUTPUT_##x)
 
-static int perf_evsel__check_stype(struct perf_evsel *evsel,
-				   u64 sample_type, const char *sample_msg,
-				   enum perf_output_field field)
+static int perf_evsel__do_check_stype(struct perf_evsel *evsel,
+				      u64 sample_type, const char *sample_msg,
+				      enum perf_output_field field,
+				      bool allow_user_set)
 {
 	struct perf_event_attr *attr = &evsel->attr;
 	int type = attr->type;
@@ -159,6 +160,8 @@ static int perf_evsel__check_stype(struct perf_evsel *evsel,
 		return 0;
 
 	if (output[type].user_set) {
+		if (allow_user_set)
+			return 0;
 		evname = perf_evsel__name(evsel);
 		pr_err("Samples for '%s' event do not have %s attribute set. "
 		       "Cannot print '%s' field.\n",
@@ -176,10 +179,22 @@ static int perf_evsel__check_stype(struct perf_evsel *evsel,
 	return 0;
 }
 
+static int perf_evsel__check_stype(struct perf_evsel *evsel,
+				   u64 sample_type, const char *sample_msg,
+				   enum perf_output_field field)
+{
+	return perf_evsel__do_check_stype(evsel, sample_type, sample_msg, field,
+					  false);
+}
+
 static int perf_evsel__check_attr(struct perf_evsel *evsel,
 				  struct perf_session *session)
 {
 	struct perf_event_attr *attr = &evsel->attr;
+	bool allow_user_set;
+
+	allow_user_set = perf_header__has_feat(&session->header,
+					       HEADER_AUXTRACE);
 
 	if (PRINT_FIELD(TRACE) &&
 		!perf_session__has_traces(session, "record -R"))
@@ -192,8 +207,8 @@ static int perf_evsel__check_attr(struct perf_evsel *evsel,
 	}
 
 	if (PRINT_FIELD(ADDR) &&
-		perf_evsel__check_stype(evsel, PERF_SAMPLE_ADDR, "ADDR",
-					PERF_OUTPUT_ADDR))
+		perf_evsel__do_check_stype(evsel, PERF_SAMPLE_ADDR, "ADDR",
+					   PERF_OUTPUT_ADDR, allow_user_set))
 		return -EINVAL;
 
 	if (PRINT_FIELD(SYM) && !PRINT_FIELD(IP) && !PRINT_FIELD(ADDR)) {
@@ -230,8 +245,8 @@ static int perf_evsel__check_attr(struct perf_evsel *evsel,
 		return -EINVAL;
 
 	if (PRINT_FIELD(CPU) &&
-		perf_evsel__check_stype(evsel, PERF_SAMPLE_CPU, "CPU",
-					PERF_OUTPUT_CPU))
+		perf_evsel__do_check_stype(evsel, PERF_SAMPLE_CPU, "CPU",
+					   PERF_OUTPUT_CPU, allow_user_set))
 		return -EINVAL;
 
 	if (PRINT_FIELD(PERIOD) &&
-- 
1.9.1

[PATCH V6 18/25] perf report: Add Instruction Tracing support

[Adrian Hunter]

Add support for decoding an AUX area assuming
it contains instruction tracing data.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/Documentation/perf-report.txt | 28 ++++++++++++++++++++++++++++
 tools/perf/builtin-report.c              | 11 +++++++++++
 2 files changed, 39 insertions(+)

diff --git a/tools/perf/Documentation/perf-report.txt b/tools/perf/Documentation/perf-report.txt
index dd7cccd..fa6f36b 100644
--- a/tools/perf/Documentation/perf-report.txt
+++ b/tools/perf/Documentation/perf-report.txt
@@ -318,6 +318,34 @@ OPTIONS
 --header-only::
 	Show only perf.data header (forces --stdio).
 
+-Z::
+--itrace::
+	Options for decoding AUX area tracing data. The options are:
+
+		i	synthesize instructions events
+		b	synthesize branches events
+		c	synthesize branches events (calls only)
+		r	synthesize branches events (returns only)
+		e	synthesize error events
+		d	create a debug log
+		g	synthesize a call chain for instructions events
+
+	The default is all events i.e. the same as -Zibe
+
+	In addition, the period (default 100000) for instructions events
+	can be specified in units of:
+
+		i	instructions
+		t	ticks
+		ms	milliseconds
+		us	microseconds
+		ns	nanoseconds (default)
+
+	Also the call chain size (default 16, max. 1024) for instructions
+	events can be specified.
+
+	To disable decoding entirely, use --no-itrace.
+
 SEE ALSO
 --------
 linkperf:perf-stat[1], linkperf:perf-annotate[1]
diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c
index 52f74e1..bf82b42 100644
--- a/tools/perf/builtin-report.c
+++ b/tools/perf/builtin-report.c
@@ -36,6 +36,8 @@
 #include "util/data.h"
 #include "arch/common.h"
 
+#include "util/auxtrace.h"
+
 #include <dlfcn.h>
 #include <linux/bitmap.h>
 
@@ -585,6 +587,7 @@ parse_percent_limit(const struct option *opt, const char *str,
 int cmd_report(int argc, const char **argv, const char *prefix __maybe_unused)
 {
 	struct perf_session *session;
+	struct itrace_synth_opts itrace_synth_opts = {0};
 	struct stat st;
 	bool has_br_stack = false;
 	int branch_mode = -1;
@@ -607,6 +610,9 @@ int cmd_report(int argc, const char **argv, const char *prefix __maybe_unused)
 			.attr		 = perf_event__process_attr,
 			.tracing_data	 = perf_event__process_tracing_data,
 			.build_id	 = perf_event__process_build_id,
+			.id_index	 = perf_event__process_id_index,
+			.auxtrace_info	 = perf_event__process_auxtrace_info,
+			.auxtrace	 = perf_event__process_auxtrace,
 			.ordered_events	 = true,
 			.ordering_requires_timestamps = true,
 		},
@@ -713,6 +719,9 @@ int cmd_report(int argc, const char **argv, const char *prefix __maybe_unused)
 		     "Don't show entries under that percent", parse_percent_limit),
 	OPT_CALLBACK(0, "percentage", NULL, "relative|absolute",
 		     "how to display percentage of filtered entries", parse_filter_percentage),
+	OPT_CALLBACK_OPTARG('Z', "itrace", &itrace_synth_opts, NULL, "opts",
+			    "Instruction Tracing options",
+			    itrace_parse_synth_opts),
 	OPT_END()
 	};
 	struct perf_data_file file = {
@@ -757,6 +766,8 @@ repeat:
 					       report.queue_size);
 	}
 
+	session->itrace_synth_opts = &itrace_synth_opts;
+
 	report.session = session;
 
 	has_br_stack = perf_header__has_feat(&session->header,
-- 
1.9.1

[PATCH V6 19/25] perf inject: Re-pipe AUX area tracing events

[Adrian Hunter]

New AUX area tracing events must be re-piped by default.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/builtin-inject.c | 63 +++++++++++++++++++++++++++++++++++++++++----
 1 file changed, 58 insertions(+), 5 deletions(-)

diff --git a/tools/perf/builtin-inject.c b/tools/perf/builtin-inject.c
index ea46df25..1f35e3d 100644
--- a/tools/perf/builtin-inject.c
+++ b/tools/perf/builtin-inject.c
@@ -38,14 +38,11 @@ struct event_entry {
 	union perf_event event[0];
 };
 
-static int perf_event__repipe_synth(struct perf_tool *tool,
-				    union perf_event *event)
+static int output_bytes(struct perf_inject *inject, void *buf, size_t sz)
 {
-	struct perf_inject *inject = container_of(tool, struct perf_inject, tool);
 	ssize_t size;
 
-	size = perf_data_file__write(&inject->output, event,
-				     event->header.size);
+	size = perf_data_file__write(&inject->output, buf, sz);
 	if (size < 0)
 		return -errno;
 
@@ -53,6 +50,34 @@ static int perf_event__repipe_synth(struct perf_tool *tool,
 	return 0;
 }
 
+static int copy_bytes(struct perf_inject *inject, int fd, off_t size)
+{
+	char buf[4096];
+	ssize_t ssz;
+	int ret;
+
+	while (size > 0) {
+		ssz = read(fd, buf, MIN(size, sizeof(buf)));
+		if (ssz < 0)
+			return -errno;
+		ret = output_bytes(inject, buf, ssz);
+		if (ret)
+			return ret;
+		size -= ssz;
+	}
+
+	return 0;
+}
+
+static int perf_event__repipe_synth(struct perf_tool *tool,
+				    union perf_event *event)
+{
+	struct perf_inject *inject = container_of(tool, struct perf_inject,
+						  tool);
+
+	return output_bytes(inject, event, event->header.size);
+}
+
 static int perf_event__repipe_oe_synth(struct perf_tool *tool,
 				       union perf_event *event,
 				       struct ordered_events *oe __maybe_unused)
@@ -86,6 +111,31 @@ static int perf_event__repipe_attr(struct perf_tool *tool,
 	return perf_event__repipe_synth(tool, event);
 }
 
+static s64 perf_event__repipe_auxtrace(struct perf_tool *tool,
+				       union perf_event *event,
+				       struct perf_session *session
+				       __maybe_unused)
+{
+	struct perf_inject *inject = container_of(tool, struct perf_inject,
+						  tool);
+	int ret;
+
+	if (perf_data_file__is_pipe(session->file) || !session->one_mmap) {
+		ret = output_bytes(inject, event, event->header.size);
+		if (ret < 0)
+			return ret;
+		ret = copy_bytes(inject, perf_data_file__fd(session->file),
+				 event->auxtrace.size);
+	} else {
+		ret = output_bytes(inject, event,
+				   event->header.size + event->auxtrace.size);
+	}
+	if (ret < 0)
+		return ret;
+
+	return event->auxtrace.size;
+}
+
 static int perf_event__repipe(struct perf_tool *tool,
 			      union perf_event *event,
 			      struct perf_sample *sample __maybe_unused,
@@ -413,6 +463,9 @@ int cmd_inject(int argc, const char **argv, const char *prefix __maybe_unused)
 			.unthrottle	= perf_event__repipe,
 			.attr		= perf_event__repipe_attr,
 			.tracing_data	= perf_event__repipe_op2_synth,
+			.auxtrace_info	= perf_event__repipe_op2_synth,
+			.auxtrace	= perf_event__repipe_auxtrace,
+			.auxtrace_error	= perf_event__repipe_op2_synth,
 			.finished_round	= perf_event__repipe_oe_synth,
 			.build_id	= perf_event__repipe_op2_synth,
 			.id_index	= perf_event__repipe_op2_synth,
-- 
1.9.1

[PATCH V6 20/25] perf inject: Add Instruction Tracing support

[Adrian Hunter]

Add support for decoding an AUX area assuming
it contains instruction tracing data.  The
AUX area tracing events are stripped and replaced
by synthesized events.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/Documentation/perf-inject.txt | 27 ++++++++++++
 tools/perf/builtin-inject.c              | 71 +++++++++++++++++++++++++++++++-
 2 files changed, 96 insertions(+), 2 deletions(-)

diff --git a/tools/perf/Documentation/perf-inject.txt b/tools/perf/Documentation/perf-inject.txt
index dc7442c..4465519 100644
--- a/tools/perf/Documentation/perf-inject.txt
+++ b/tools/perf/Documentation/perf-inject.txt
@@ -44,6 +44,33 @@ OPTIONS
 --kallsyms=<file>::
 	kallsyms pathname
 
+-Z::
+--itrace::
+	Decode Instruction Tracing data, replacing it with synthesized events.
+	Options are:
+
+		i	synthesize instructions events
+		b	synthesize branches events
+		c	synthesize branches events (calls only)
+		r	synthesize branches events (returns only)
+		e	synthesize error events
+		d	create a debug log
+		g	synthesize a call chain for instructions events
+
+	The default is all events i.e. the same as -Zibe
+
+	In addition, the period (default 100000) for instructions events
+	can be specified in units of:
+
+		i	instructions
+		t	ticks
+		ms	milliseconds
+		us	microseconds
+		ns	nanoseconds (default)
+
+	Also the call chain size (default 16, max. 1024) for instructions
+	events can be specified.
+
 SEE ALSO
 --------
 linkperf:perf-record[1], linkperf:perf-report[1], linkperf:perf-archive[1]
diff --git a/tools/perf/builtin-inject.c b/tools/perf/builtin-inject.c
index 1f35e3d..3a36c7b 100644
--- a/tools/perf/builtin-inject.c
+++ b/tools/perf/builtin-inject.c
@@ -16,6 +16,7 @@
 #include "util/debug.h"
 #include "util/build-id.h"
 #include "util/data.h"
+#include "util/auxtrace.h"
 
 #include "util/parse-options.h"
 
@@ -30,6 +31,7 @@ struct perf_inject {
 	struct perf_data_file	output;
 	u64			bytes_written;
 	struct list_head	samples;
+	struct itrace_synth_opts itrace_synth_opts;
 };
 
 struct event_entry {
@@ -205,6 +207,32 @@ static int perf_event__repipe_fork(struct perf_tool *tool,
 	return err;
 }
 
+static int perf_event__repipe_comm(struct perf_tool *tool,
+				   union perf_event *event,
+				   struct perf_sample *sample,
+				   struct machine *machine)
+{
+	int err;
+
+	err = perf_event__process_comm(tool, event, sample, machine);
+	perf_event__repipe(tool, event, sample, machine);
+
+	return err;
+}
+
+static int perf_event__repipe_exit(struct perf_tool *tool,
+				   union perf_event *event,
+				   struct perf_sample *sample,
+				   struct machine *machine)
+{
+	int err;
+
+	err = perf_event__process_exit(tool, event, sample, machine);
+	perf_event__repipe(tool, event, sample, machine);
+
+	return err;
+}
+
 static int perf_event__repipe_tracing_data(struct perf_tool *tool,
 					   union perf_event *event,
 					   struct perf_session *session)
@@ -217,6 +245,18 @@ static int perf_event__repipe_tracing_data(struct perf_tool *tool,
 	return err;
 }
 
+static int perf_event__repipe_id_index(struct perf_tool *tool,
+				       union perf_event *event,
+				       struct perf_session *session)
+{
+	int err;
+
+	perf_event__repipe_synth(tool, event);
+	err = perf_event__process_id_index(tool, event, session);
+
+	return err;
+}
+
 static int dso__read_build_id(struct dso *dso)
 {
 	if (dso->has_build_id)
@@ -401,16 +441,20 @@ static int __cmd_inject(struct perf_inject *inject)
 	struct perf_session *session = inject->session;
 	struct perf_data_file *file_out = &inject->output;
 	int fd = perf_data_file__fd(file_out);
+	u64 output_data_offset;
 
 	signal(SIGINT, sig_handler);
 
-	if (inject->build_ids || inject->sched_stat) {
+	if (inject->build_ids || inject->sched_stat ||
+	    inject->itrace_synth_opts.set) {
 		inject->tool.mmap	  = perf_event__repipe_mmap;
 		inject->tool.mmap2	  = perf_event__repipe_mmap2;
 		inject->tool.fork	  = perf_event__repipe_fork;
 		inject->tool.tracing_data = perf_event__repipe_tracing_data;
 	}
 
+	output_data_offset = session->header.data_offset;
+
 	if (inject->build_ids) {
 		inject->tool.sample = perf_event__inject_buildid;
 	} else if (inject->sched_stat) {
@@ -429,10 +473,22 @@ static int __cmd_inject(struct perf_inject *inject)
 			else if (!strncmp(name, "sched:sched_stat_", 17))
 				evsel->handler = perf_inject__sched_stat;
 		}
+	} else if (inject->itrace_synth_opts.set) {
+		session->itrace_synth_opts = &inject->itrace_synth_opts;
+		inject->itrace_synth_opts.inject = true;
+		inject->tool.comm	    = perf_event__repipe_comm;
+		inject->tool.exit	    = perf_event__repipe_exit;
+		inject->tool.id_index	    = perf_event__repipe_id_index;
+		inject->tool.auxtrace_info  = perf_event__process_auxtrace_info;
+		inject->tool.auxtrace	    = perf_event__process_auxtrace;
+		inject->tool.ordered_events = true;
+		inject->tool.ordering_requires_timestamps = true;
+		/* Allow space in the header for new attributes */
+		output_data_offset = 4096;
 	}
 
 	if (!file_out->is_pipe)
-		lseek(fd, session->header.data_offset, SEEK_SET);
+		lseek(fd, output_data_offset, SEEK_SET);
 
 	ret = perf_session__process_events(session);
 
@@ -440,6 +496,14 @@ static int __cmd_inject(struct perf_inject *inject)
 		if (inject->build_ids)
 			perf_header__set_feat(&session->header,
 					      HEADER_BUILD_ID);
+		/*
+		 * The AUX areas have been removed and replaced with
+		 * synthesized hardware events, so clear the feature flag.
+		 */
+		if (inject->itrace_synth_opts.set)
+			perf_header__clear_feat(&session->header,
+						HEADER_AUXTRACE);
+		session->header.data_offset = output_data_offset;
 		session->header.data_size = inject->bytes_written;
 		perf_session__write_header(session, session->evlist, fd, true);
 	}
@@ -496,6 +560,9 @@ int cmd_inject(int argc, const char **argv, const char *prefix __maybe_unused)
 			 "be more verbose (show build ids, etc)"),
 		OPT_STRING(0, "kallsyms", &symbol_conf.kallsyms_name, "file",
 			   "kallsyms pathname"),
+		OPT_CALLBACK_OPTARG('Z', "itrace", &inject.itrace_synth_opts,
+				    NULL, "opts", "Instruction Tracing options",
+				    itrace_parse_synth_opts),
 		OPT_END()
 	};
 	const char * const inject_usage[] = {
-- 
1.9.1

[PATCH V6 21/25] perf tools: Add AUX area tracing index

[Adrian Hunter]

Add an index of AUX area tracing events within
a perf.data file.

perf record uses a special user event
PERF_RECORD_FINISHED_ROUND to enable sorting of
events in chunks instead of having to sort all
events altogether.

AUX area tracing events contain data that can
span back to the very beginning of the recording
period. i.e. they do not obey the rules of
PERF_RECORD_FINISHED_ROUND.

By adding an index, AUX area tracing events
can be found in advance and the
PERF_RECORD_FINISHED_ROUND approach works as
usual.

The index is recorded with the auxtrace feature
in the perf.data file.  A session reads the index
but does not process it.  An AUX area
decoder can queue all the AUX area data
in advance using auxtrace_queues__process_index()
or otherwise process the index in some custom
manner.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/builtin-inject.c |  15 +++
 tools/perf/builtin-record.c |  15 +++
 tools/perf/util/auxtrace.c  | 216 ++++++++++++++++++++++++++++++++++++++++++++
 tools/perf/util/auxtrace.h  |  35 +++++++
 tools/perf/util/header.c    |  31 ++++++-
 tools/perf/util/session.c   |   2 +
 tools/perf/util/session.h   |   1 +
 7 files changed, 311 insertions(+), 4 deletions(-)

diff --git a/tools/perf/builtin-inject.c b/tools/perf/builtin-inject.c
index 3a36c7b..c7b429e 100644
--- a/tools/perf/builtin-inject.c
+++ b/tools/perf/builtin-inject.c
@@ -122,6 +122,18 @@ static s64 perf_event__repipe_auxtrace(struct perf_tool *tool,
 						  tool);
 	int ret;
 
+	if (!inject->output.is_pipe) {
+		off_t offset;
+
+		offset = lseek(inject->output.fd, 0, SEEK_CUR);
+		if (offset == -1)
+			return -errno;
+		ret = auxtrace_index__auxtrace_event(&session->auxtrace_index,
+						     event, offset);
+		if (ret < 0)
+			return ret;
+	}
+
 	if (perf_data_file__is_pipe(session->file) || !session->one_mmap) {
 		ret = output_bytes(inject, event, event->header.size);
 		if (ret < 0)
@@ -487,6 +499,9 @@ static int __cmd_inject(struct perf_inject *inject)
 		output_data_offset = 4096;
 	}
 
+	if (!inject->itrace_synth_opts.set)
+		auxtrace_index__free(&session->auxtrace_index);
+
 	if (!file_out->is_pipe)
 		lseek(fd, output_data_offset, SEEK_SET);
 
diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
index d337896..82f3049 100644
--- a/tools/perf/builtin-record.c
+++ b/tools/perf/builtin-record.c
@@ -117,9 +117,24 @@ static int record__process_auxtrace(struct perf_tool *tool,
 				    size_t len1, void *data2, size_t len2)
 {
 	struct record *rec = container_of(tool, struct record, tool);
+	struct perf_data_file *file = &rec->file;
 	size_t padding;
 	u8 pad[8] = {0};
 
+	if (!perf_data_file__is_pipe(file)) {
+		off_t file_offset;
+		int fd = perf_data_file__fd(file);
+		int err;
+
+		file_offset = lseek(fd, 0, SEEK_CUR);
+		if (file_offset == -1)
+			return -1;
+		err = auxtrace_index__auxtrace_event(&rec->session->auxtrace_index,
+						     event, file_offset);
+		if (err)
+			return err;
+	}
+
 	/* event.auxtrace.size includes padding, see __auxtrace_mmap__read() */
 	padding = (len1 + len2) & 7;
 	if (padding)
diff --git a/tools/perf/util/auxtrace.c b/tools/perf/util/auxtrace.c
index e1499c5..5a95767 100644
--- a/tools/perf/util/auxtrace.c
+++ b/tools/perf/util/auxtrace.c
@@ -334,6 +334,33 @@ out_err:
 	return err;
 }
 
+static int auxtrace_queues__add_indexed_event(struct auxtrace_queues *queues,
+					      struct perf_session *session,
+					      off_t file_offset, size_t sz)
+{
+	union perf_event *event;
+	int err;
+	char buf[PERF_SAMPLE_MAX_SIZE];
+
+	err = perf_session__peek_event(session, file_offset, buf,
+				       PERF_SAMPLE_MAX_SIZE, &event, NULL);
+	if (err)
+		return err;
+
+	if (event->header.type == PERF_RECORD_AUXTRACE) {
+		if (event->header.size != sizeof(struct auxtrace_event) ||
+		    event->header.size != sz) {
+			err = -EINVAL;
+			goto out;
+		}
+		file_offset += event->header.size;
+		err = auxtrace_queues__add_event(queues, session, event,
+						 file_offset, NULL);
+	}
+out:
+	return err;
+}
+
 void auxtrace_queues__free(struct auxtrace_queues *queues)
 {
 	unsigned int i;
@@ -490,6 +517,195 @@ auxtrace_record__init(struct perf_evlist *evlist __maybe_unused, int *err)
 	return NULL;
 }
 
+static int auxtrace_index__alloc(struct list_head *head)
+{
+	struct auxtrace_index *auxtrace_index;
+
+	auxtrace_index = malloc(sizeof(struct auxtrace_index));
+	if (!auxtrace_index)
+		return -ENOMEM;
+
+	auxtrace_index->nr = 0;
+	INIT_LIST_HEAD(&auxtrace_index->list);
+
+	list_add_tail(&auxtrace_index->list, head);
+
+	return 0;
+}
+
+void auxtrace_index__free(struct list_head *head)
+{
+	struct auxtrace_index *auxtrace_index, *n;
+
+	list_for_each_entry_safe(auxtrace_index, n, head, list) {
+		list_del(&auxtrace_index->list);
+		free(auxtrace_index);
+	}
+}
+
+static struct auxtrace_index *auxtrace_index__last(struct list_head *head)
+{
+	struct auxtrace_index *auxtrace_index;
+	int err;
+
+	if (list_empty(head)) {
+		err = auxtrace_index__alloc(head);
+		if (err)
+			return NULL;
+	}
+
+	auxtrace_index = list_entry(head->prev, struct auxtrace_index, list);
+
+	if (auxtrace_index->nr >= PERF_AUXTRACE_INDEX_ENTRY_COUNT) {
+		err = auxtrace_index__alloc(head);
+		if (err)
+			return NULL;
+		auxtrace_index = list_entry(head->prev, struct auxtrace_index,
+					    list);
+	}
+
+	return auxtrace_index;
+}
+
+int auxtrace_index__auxtrace_event(struct list_head *head,
+				   union perf_event *event, off_t file_offset)
+{
+	struct auxtrace_index *auxtrace_index;
+	size_t nr;
+
+	auxtrace_index = auxtrace_index__last(head);
+	if (!auxtrace_index)
+		return -ENOMEM;
+
+	nr = auxtrace_index->nr;
+	auxtrace_index->entries[nr].file_offset = file_offset;
+	auxtrace_index->entries[nr].sz = event->header.size;
+	auxtrace_index->nr += 1;
+
+	return 0;
+}
+
+static int auxtrace_index__do_write(int fd,
+				    struct auxtrace_index *auxtrace_index)
+{
+	struct auxtrace_index_entry index;
+	size_t i;
+
+	for (i = 0; i < auxtrace_index->nr; i++) {
+		index.file_offset = auxtrace_index->entries[i].file_offset;
+		index.sz = auxtrace_index->entries[i].sz;
+		if (writen(fd, &index, sizeof(index)) != sizeof(index))
+			return -errno;
+	}
+	return 0;
+}
+
+int auxtrace_index__write(int fd, struct list_head *head)
+{
+	struct auxtrace_index *auxtrace_index;
+	u64 total = 0;
+	int err;
+
+	list_for_each_entry(auxtrace_index, head, list)
+		total += auxtrace_index->nr;
+
+	if (writen(fd, &total, sizeof(total)) != sizeof(total))
+		return -errno;
+
+	list_for_each_entry(auxtrace_index, head, list) {
+		err = auxtrace_index__do_write(fd, auxtrace_index);
+		if (err)
+			return err;
+	}
+
+	return 0;
+}
+
+static int auxtrace_index__process_entry(int fd, struct list_head *head,
+					 bool needs_swap)
+{
+	struct auxtrace_index *auxtrace_index;
+	struct auxtrace_index_entry index;
+	size_t nr;
+
+	if (readn(fd, &index, sizeof(index)) != sizeof(index))
+		return -1;
+
+	auxtrace_index = auxtrace_index__last(head);
+	if (!auxtrace_index)
+		return -1;
+
+	nr = auxtrace_index->nr;
+	if (needs_swap) {
+		auxtrace_index->entries[nr].file_offset =
+						bswap_64(index.file_offset);
+		auxtrace_index->entries[nr].sz = bswap_64(index.sz);
+	} else {
+		auxtrace_index->entries[nr].file_offset = index.file_offset;
+		auxtrace_index->entries[nr].sz = index.sz;
+	}
+
+	auxtrace_index->nr = nr + 1;
+
+	return 0;
+}
+
+int auxtrace_index__process(int fd, u64 size, struct perf_session *session,
+			    bool needs_swap)
+{
+	struct list_head *head = &session->auxtrace_index;
+	u64 nr;
+
+	if (readn(fd, &nr, sizeof(u64)) != sizeof(u64))
+		return -1;
+
+	if (needs_swap)
+		nr = bswap_64(nr);
+
+	if (sizeof(u64) + nr * sizeof(struct auxtrace_index_entry) != size)
+		return -1;
+
+	while (nr--) {
+		int err;
+
+		err = auxtrace_index__process_entry(fd, head, needs_swap);
+		if (err)
+			return -1;
+	}
+
+	return 0;
+}
+
+static int auxtrace_queues__process_index_entry(struct auxtrace_queues *queues,
+						struct perf_session *session,
+						struct auxtrace_index_entry *index)
+{
+	return auxtrace_queues__add_indexed_event(queues, session,
+						  index->file_offset,
+						  index->sz);
+}
+
+int auxtrace_queues__process_index(struct auxtrace_queues *queues,
+				   struct perf_session *session)
+{
+	struct auxtrace_index *auxtrace_index;
+	struct auxtrace_index_entry *index;
+	size_t i;
+	int err;
+
+	list_for_each_entry(auxtrace_index, &session->auxtrace_index, list) {
+		for (i = 0; i < auxtrace_index->nr; i++) {
+			index = &auxtrace_index->entries[i];
+			err = auxtrace_queues__process_index_entry(queues,
+								   session,
+								   index);
+			if (err)
+				return err;
+		}
+	}
+	return 0;
+}
+
 struct auxtrace_buffer *auxtrace_buffer__next(struct auxtrace_queue *queue,
 					      struct auxtrace_buffer *buffer)
 {
diff --git a/tools/perf/util/auxtrace.h b/tools/perf/util/auxtrace.h
index 5dfa05a..17f9f64 100644
--- a/tools/perf/util/auxtrace.h
+++ b/tools/perf/util/auxtrace.h
@@ -80,6 +80,32 @@ struct itrace_synth_opts {
 };
 
 /**
+ * struct auxtrace_index_entry - indexes a AUX area tracing event within a
+ *                               perf.data file.
+ * @file_offset: offset within the perf.data file
+ * @sz: size of the event
+ */
+struct auxtrace_index_entry {
+	u64			file_offset;
+	u64			sz;
+};
+
+#define PERF_AUXTRACE_INDEX_ENTRY_COUNT 256
+
+/**
+ * struct auxtrace_index - index of AUX area tracing events within a perf.data
+ *                         file.
+ * @list: linking a number of arrays of entries
+ * @nr: number of entries
+ * @entries: array of entries
+ */
+struct auxtrace_index {
+	struct list_head	list;
+	size_t			nr;
+	struct auxtrace_index_entry entries[PERF_AUXTRACE_INDEX_ENTRY_COUNT];
+};
+
+/**
  * struct auxtrace - session callbacks to allow AUX area data decoding.
  * @process_event: lets the decoder see all session events
  * @flush_events: process any remaining data
@@ -301,6 +327,8 @@ int auxtrace_queues__add_event(struct auxtrace_queues *queues,
 			       union perf_event *event, off_t data_offset,
 			       struct auxtrace_buffer **buffer_ptr);
 void auxtrace_queues__free(struct auxtrace_queues *queues);
+int auxtrace_queues__process_index(struct auxtrace_queues *queues,
+				   struct perf_session *session);
 struct auxtrace_buffer *auxtrace_buffer__next(struct auxtrace_queue *queue,
 					      struct auxtrace_buffer *buffer);
 void *auxtrace_buffer__get_data(struct auxtrace_buffer *buffer, int fd);
@@ -341,6 +369,13 @@ int auxtrace_record__info_fill(struct auxtrace_record *itr,
 void auxtrace_record__free(struct auxtrace_record *itr);
 u64 auxtrace_record__reference(struct auxtrace_record *itr);
 
+int auxtrace_index__auxtrace_event(struct list_head *head, union perf_event *event,
+				   off_t file_offset);
+int auxtrace_index__write(int fd, struct list_head *head);
+int auxtrace_index__process(int fd, u64 size, struct perf_session *session,
+			    bool needs_swap);
+void auxtrace_index__free(struct list_head *head);
+
 void auxtrace_synth_error(struct auxtrace_error_event *auxtrace_error, int type,
 			  int code, int cpu, pid_t pid, pid_t tid, u64 ip,
 			  const char *msg);
diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c
index 9b06b5e..0e26a25 100644
--- a/tools/perf/util/header.c
+++ b/tools/perf/util/header.c
@@ -869,11 +869,18 @@ static int write_branch_stack(int fd __maybe_unused,
 	return 0;
 }
 
-static int write_auxtrace(int fd __maybe_unused,
-			  struct perf_header *h __maybe_unused,
+static int write_auxtrace(int fd, struct perf_header *h,
 			  struct perf_evlist *evlist __maybe_unused)
 {
-	return 0;
+	struct perf_session *session;
+	int err;
+
+	session = container_of(h, struct perf_session, header);
+
+	err = auxtrace_index__write(fd, &session->auxtrace_index);
+	if (err < 0)
+		pr_err("Failed to write auxtrace index\n");
+	return err;
 }
 
 static void print_hostname(struct perf_header *ph, int fd __maybe_unused,
@@ -1846,6 +1853,22 @@ out_free:
 	return ret;
 }
 
+static int process_auxtrace(struct perf_file_section *section,
+			    struct perf_header *ph, int fd,
+			    void *data __maybe_unused)
+{
+	struct perf_session *session;
+	int err;
+
+	session = container_of(ph, struct perf_session, header);
+
+	err = auxtrace_index__process(fd, section->size, session,
+				      ph->needs_swap);
+	if (err < 0)
+		pr_err("Failed to process auxtrace index\n");
+	return err;
+}
+
 struct feature_ops {
 	int (*write)(int fd, struct perf_header *h, struct perf_evlist *evlist);
 	void (*print)(struct perf_header *h, int fd, FILE *fp);
@@ -1886,7 +1909,7 @@ static const struct feature_ops feat_ops[HEADER_LAST_FEATURE] = {
 	FEAT_OPA(HEADER_BRANCH_STACK,	branch_stack),
 	FEAT_OPP(HEADER_PMU_MAPPINGS,	pmu_mappings),
 	FEAT_OPP(HEADER_GROUP_DESC,	group_desc),
-	FEAT_OPA(HEADER_AUXTRACE,	auxtrace),
+	FEAT_OPP(HEADER_AUXTRACE,	auxtrace),
 };
 
 struct header_print_data {
diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c
index e13654a..a5de160 100644
--- a/tools/perf/util/session.c
+++ b/tools/perf/util/session.c
@@ -100,6 +100,7 @@ struct perf_session *perf_session__new(struct perf_data_file *file,
 		goto out;
 
 	session->repipe = repipe;
+	INIT_LIST_HEAD(&session->auxtrace_index);
 	machines__init(&session->machines);
 
 	if (file) {
@@ -170,6 +171,7 @@ static void perf_session_env__delete(struct perf_session_env *env)
 void perf_session__delete(struct perf_session *session)
 {
 	auxtrace__free(session);
+	auxtrace_index__free(&session->auxtrace_index);
 	perf_session__destroy_kernel_maps(session);
 	perf_session__delete_threads(session);
 	perf_session_env__delete(&session->header.env);
diff --git a/tools/perf/util/session.h b/tools/perf/util/session.h
index d7d8225..08c975c 100644
--- a/tools/perf/util/session.h
+++ b/tools/perf/util/session.h
@@ -24,6 +24,7 @@ struct perf_session {
 	struct perf_evlist	*evlist;
 	struct auxtrace		*auxtrace;
 	struct itrace_synth_opts *itrace_synth_opts;
+	struct list_head	auxtrace_index;
 	struct trace_event	tevent;
 	bool			repipe;
 	bool			one_mmap;
-- 
1.9.1

[PATCH V6 22/25] perf tools: Hit all build ids when AUX area tracing

[Adrian Hunter]

We need to include all buildids when a perf.data
file contains AUX area tracing data because we
do not decode the trace for that purpose because
it would take too long.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/builtin-buildid-list.c |  9 +++++++++
 tools/perf/builtin-inject.c       |  8 +++++++-
 tools/perf/builtin-record.c       | 10 +++++++++-
 3 files changed, 25 insertions(+), 2 deletions(-)

diff --git a/tools/perf/builtin-buildid-list.c b/tools/perf/builtin-buildid-list.c
index feb420f..9fe93c8 100644
--- a/tools/perf/builtin-buildid-list.c
+++ b/tools/perf/builtin-buildid-list.c
@@ -69,6 +69,15 @@ static int perf_session__list_build_ids(bool force, bool with_hits)
 	session = perf_session__new(&file, false, &build_id__mark_dso_hit_ops);
 	if (session == NULL)
 		return -1;
+
+	/*
+	 * We take all buildids when the file contains AUX area tracing data
+	 * because we do not decode the trace because it would take too long.
+	 */
+	if (!perf_data_file__is_pipe(&file) &&
+	    perf_header__has_feat(&session->header, HEADER_AUXTRACE))
+		with_hits = false;
+
 	/*
 	 * in pipe-mode, the only way to get the buildids is to parse
 	 * the record stream. Buildids are stored as RECORD_HEADER_BUILD_ID
diff --git a/tools/perf/builtin-inject.c b/tools/perf/builtin-inject.c
index c7b429e..d26a052 100644
--- a/tools/perf/builtin-inject.c
+++ b/tools/perf/builtin-inject.c
@@ -27,6 +27,7 @@ struct perf_inject {
 	struct perf_session	*session;
 	bool			build_ids;
 	bool			sched_stat;
+	bool			have_auxtrace;
 	const char		*input_name;
 	struct perf_data_file	output;
 	u64			bytes_written;
@@ -122,6 +123,8 @@ static s64 perf_event__repipe_auxtrace(struct perf_tool *tool,
 						  tool);
 	int ret;
 
+	inject->have_auxtrace = true;
+
 	if (!inject->output.is_pipe) {
 		off_t offset;
 
@@ -508,9 +511,12 @@ static int __cmd_inject(struct perf_inject *inject)
 	ret = perf_session__process_events(session);
 
 	if (!file_out->is_pipe) {
-		if (inject->build_ids)
+		if (inject->build_ids) {
 			perf_header__set_feat(&session->header,
 					      HEADER_BUILD_ID);
+			if (inject->have_auxtrace)
+				dsos__hit_all(session);
+		}
 		/*
 		 * The AUX areas have been removed and replaced with
 		 * synthesized hardware events, so clear the feature flag.
diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
index 82f3049..1dd40b6 100644
--- a/tools/perf/builtin-record.c
+++ b/tools/perf/builtin-record.c
@@ -618,8 +618,16 @@ out_child:
 	if (!err && !file->is_pipe) {
 		rec->session->header.data_size += rec->bytes_written;
 
-		if (!rec->no_buildid)
+		if (!rec->no_buildid) {
 			process_buildids(rec);
+			/*
+			 * We take all buildids when the file contains
+			 * AUX area tracing data because we do not decode the
+			 * trace because it would take too long.
+			 */
+			if (rec->opts.full_auxtrace)
+				dsos__hit_all(rec->session);
+		}
 		perf_session__write_header(rec->session, rec->evlist, fd, true);
 	}
 
-- 
1.9.1

[PATCH V6 23/25] perf tools: Add build option NO_AUXTRACE to exclude AUX area tracing

[Adrian Hunter]

Add build option NO_AUXTRACE to exclude compiling support
for AUX area tracing. Support for both recording and
processing is excluded and by implication any future
additions such as Intel PT and Intel BTS will also not
be compiled in with this option.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/Makefile.perf    |   2 +
 tools/perf/builtin-inject.c |  53 +++++++++++-------
 tools/perf/builtin-record.c |  13 +++++
 tools/perf/config/Makefile  |   5 ++
 tools/perf/tests/make       |   2 +
 tools/perf/util/Build       |   2 +-
 tools/perf/util/auxtrace.h  | 128 ++++++++++++++++++++++++++++++++++++++++++++
 7 files changed, 185 insertions(+), 20 deletions(-)

diff --git a/tools/perf/Makefile.perf b/tools/perf/Makefile.perf
index 933d703..fa0a1e8 100644
--- a/tools/perf/Makefile.perf
+++ b/tools/perf/Makefile.perf
@@ -71,6 +71,8 @@ include config/utilities.mak
 #
 # Define NO_LIBBABELTRACE if you do not want libbabeltrace support
 # for CTF data format.
+#
+# Define NO_AUXTRACE if you do not want AUX area tracing support
 
 ifeq ($(srctree),)
 srctree := $(patsubst %/,%,$(dir $(shell pwd)))
diff --git a/tools/perf/builtin-inject.c b/tools/perf/builtin-inject.c
index d26a052..1639a83 100644
--- a/tools/perf/builtin-inject.c
+++ b/tools/perf/builtin-inject.c
@@ -53,25 +53,6 @@ static int output_bytes(struct perf_inject *inject, void *buf, size_t sz)
 	return 0;
 }
 
-static int copy_bytes(struct perf_inject *inject, int fd, off_t size)
-{
-	char buf[4096];
-	ssize_t ssz;
-	int ret;
-
-	while (size > 0) {
-		ssz = read(fd, buf, MIN(size, sizeof(buf)));
-		if (ssz < 0)
-			return -errno;
-		ret = output_bytes(inject, buf, ssz);
-		if (ret)
-			return ret;
-		size -= ssz;
-	}
-
-	return 0;
-}
-
 static int perf_event__repipe_synth(struct perf_tool *tool,
 				    union perf_event *event)
 {
@@ -114,6 +95,27 @@ static int perf_event__repipe_attr(struct perf_tool *tool,
 	return perf_event__repipe_synth(tool, event);
 }
 
+#ifdef HAVE_AUXTRACE_SUPPORT
+
+static int copy_bytes(struct perf_inject *inject, int fd, off_t size)
+{
+	char buf[4096];
+	ssize_t ssz;
+	int ret;
+
+	while (size > 0) {
+		ssz = read(fd, buf, MIN(size, sizeof(buf)));
+		if (ssz < 0)
+			return -errno;
+		ret = output_bytes(inject, buf, ssz);
+		if (ret)
+			return ret;
+		size -= ssz;
+	}
+
+	return 0;
+}
+
 static s64 perf_event__repipe_auxtrace(struct perf_tool *tool,
 				       union perf_event *event,
 				       struct perf_session *session
@@ -153,6 +155,19 @@ static s64 perf_event__repipe_auxtrace(struct perf_tool *tool,
 	return event->auxtrace.size;
 }
 
+#else
+
+static s64
+perf_event__repipe_auxtrace(struct perf_tool *tool __maybe_unused,
+			    union perf_event *event __maybe_unused,
+			    struct perf_session *session __maybe_unused)
+{
+	pr_err("AUX area tracing not supported\n");
+	return -EINVAL;
+}
+
+#endif
+
 static int perf_event__repipe(struct perf_tool *tool,
 			      union perf_event *event,
 			      struct perf_sample *sample __maybe_unused,
diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
index 1dd40b6..729e359 100644
--- a/tools/perf/builtin-record.c
+++ b/tools/perf/builtin-record.c
@@ -112,6 +112,8 @@ out:
 	return rc;
 }
 
+#ifdef HAVE_AUXTRACE_SUPPORT
+
 static int record__process_auxtrace(struct perf_tool *tool,
 				    union perf_event *event, void *data1,
 				    size_t len1, void *data2, size_t len2)
@@ -164,6 +166,17 @@ static int record__auxtrace_mmap_read(struct record *rec,
 	return 0;
 }
 
+#else
+
+static inline
+int record__auxtrace_mmap_read(struct record *rec __maybe_unused,
+			       struct auxtrace_mmap *mm __maybe_unused)
+{
+	return 0;
+}
+
+#endif
+
 static volatile int done = 0;
 static volatile int signr = -1;
 static volatile int child_finished = 0;
diff --git a/tools/perf/config/Makefile b/tools/perf/config/Makefile
index bd09718..bee7d3a 100644
--- a/tools/perf/config/Makefile
+++ b/tools/perf/config/Makefile
@@ -716,6 +716,11 @@ ifndef NO_LIBBABELTRACE
   endif
 endif
 
+ifndef NO_AUXTRACE
+  $(call detected,CONFIG_AUXTRACE)
+  CFLAGS += -DHAVE_AUXTRACE_SUPPORT
+endif
+
 # Among the variables below, these:
 #   perfexecdir
 #   template_dir
diff --git a/tools/perf/tests/make b/tools/perf/tests/make
index 75709d2..2d617c4 100644
--- a/tools/perf/tests/make
+++ b/tools/perf/tests/make
@@ -32,6 +32,7 @@ make_no_backtrace   := NO_BACKTRACE=1
 make_no_libnuma     := NO_LIBNUMA=1
 make_no_libaudit    := NO_LIBAUDIT=1
 make_no_libbionic   := NO_LIBBIONIC=1
+make_no_auxtrace      := NO_AUXTRACE=1
 make_tags           := tags
 make_cscope         := cscope
 make_help           := help
@@ -74,6 +75,7 @@ run += make_no_backtrace
 run += make_no_libnuma
 run += make_no_libaudit
 run += make_no_libbionic
+run += make_no_auxtrace
 run += make_help
 run += make_doc
 run += make_perf_o
diff --git a/tools/perf/util/Build b/tools/perf/util/Build
index bb55e26..e2d09c5 100644
--- a/tools/perf/util/Build
+++ b/tools/perf/util/Build
@@ -74,7 +74,7 @@ libperf-y += data.o
 libperf-$(CONFIG_X86) += tsc.o
 libperf-y += cloexec.o
 libperf-y += thread-stack.o
-libperf-y += auxtrace.o
+libperf-$(CONFIG_AUXTRACE) += auxtrace.o
 
 libperf-$(CONFIG_LIBELF) += symbol-elf.o
 libperf-$(CONFIG_LIBELF) += probe-event.o
diff --git a/tools/perf/util/auxtrace.h b/tools/perf/util/auxtrace.h
index 17f9f64..8441dee 100644
--- a/tools/perf/util/auxtrace.h
+++ b/tools/perf/util/auxtrace.h
@@ -26,6 +26,7 @@
 #include "../perf.h"
 #include "event.h"
 #include "session.h"
+#include "debug.h"
 
 union perf_event;
 struct perf_session;
@@ -290,6 +291,8 @@ struct auxtrace_record {
 	int (*read_finish)(struct auxtrace_record *itr, int idx);
 };
 
+#ifdef HAVE_AUXTRACE_SUPPORT
+
 static inline u64 auxtrace_mmap__read_head(struct auxtrace_mmap *mm __maybe_unused)
 {
 	/* Not yet implemented */
@@ -442,4 +445,129 @@ static inline void auxtrace__free(struct perf_session *session)
 	return session->auxtrace->free(session);
 }
 
+#else
+
+static inline struct auxtrace_record *
+auxtrace_record__init(struct perf_evlist *evlist __maybe_unused,
+		      int *err __maybe_unused)
+{
+	*err = 0;
+	return NULL;
+}
+
+static inline
+void auxtrace_record__free(struct auxtrace_record *itr __maybe_unused)
+{
+}
+
+static inline int
+perf_event__synthesize_auxtrace_info(struct auxtrace_record *itr __maybe_unused,
+				     struct perf_tool *tool __maybe_unused,
+				     struct perf_session *session __maybe_unused,
+				     perf_event__handler_t process __maybe_unused)
+{
+	return -EINVAL;
+}
+
+static inline
+int auxtrace_record__options(struct auxtrace_record *itr __maybe_unused,
+			     struct perf_evlist *evlist __maybe_unused,
+			     struct record_opts *opts __maybe_unused)
+{
+	return 0;
+}
+
+#define perf_event__process_auxtrace_info		0
+#define perf_event__process_auxtrace			0
+#define perf_event__process_auxtrace_error		0
+
+static inline
+void perf_session__auxtrace_error_inc(struct perf_session *session
+				      __maybe_unused,
+				      union perf_event *event
+				      __maybe_unused)
+{
+}
+
+static inline
+void events_stats__auxtrace_error_warn(const struct events_stats *stats
+				       __maybe_unused)
+{
+}
+
+static inline
+int itrace_parse_synth_opts(const struct option *opt __maybe_unused,
+			    const char *str __maybe_unused,
+			    int unset __maybe_unused)
+{
+	pr_err("AUX area tracing not supported\n");
+	return -EINVAL;
+}
+
+static inline
+int auxtrace__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 inline
+int auxtrace__flush_events(struct perf_session *session __maybe_unused,
+			   struct perf_tool *tool __maybe_unused)
+{
+	return 0;
+}
+
+static inline
+void auxtrace__free_events(struct perf_session *session __maybe_unused)
+{
+}
+
+static inline
+void auxtrace_cache__free(struct auxtrace_cache *auxtrace_cache __maybe_unused)
+{
+}
+
+static inline
+void auxtrace__free(struct perf_session *session __maybe_unused)
+{
+}
+
+static inline
+int auxtrace_index__write(int fd __maybe_unused,
+			  struct list_head *head __maybe_unused)
+{
+	return -EINVAL;
+}
+
+static inline
+int auxtrace_index__process(int fd __maybe_unused,
+			    u64 size __maybe_unused,
+			    struct perf_session *session __maybe_unused,
+			    bool needs_swap __maybe_unused)
+{
+	return -EINVAL;
+}
+
+static inline
+void auxtrace_index__free(struct list_head *head __maybe_unused)
+{
+}
+
+int auxtrace_mmap__mmap(struct auxtrace_mmap *mm,
+			struct auxtrace_mmap_params *mp,
+			void *userpg, int fd);
+void auxtrace_mmap__munmap(struct auxtrace_mmap *mm);
+void auxtrace_mmap_params__init(struct auxtrace_mmap_params *mp,
+				off_t auxtrace_offset,
+				unsigned int auxtrace_pages,
+				bool auxtrace_overwrite);
+void auxtrace_mmap_params__set_idx(struct auxtrace_mmap_params *mp,
+				   struct perf_evlist *evlist, int idx,
+				   bool per_cpu);
+
+#endif
+
 #endif
-- 
1.9.1

[PATCH V6 24/25] perf auxtrace: Add option to synthesize events for transactions

[Adrian Hunter]

Add AUX area tracing option 'x' to synthesize events for
transactions. This will be used by Intel PT to synthesize
an event record for each TSX start, commit or abort.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/Documentation/perf-report.txt | 9 +++++----
 tools/perf/Documentation/perf-script.txt | 9 +++++----
 tools/perf/util/auxtrace.c               | 5 ++++-
 tools/perf/util/auxtrace.h               | 2 ++
 4 files changed, 16 insertions(+), 9 deletions(-)

diff --git a/tools/perf/Documentation/perf-report.txt b/tools/perf/Documentation/perf-report.txt
index fa6f36b..dc41c93 100644
--- a/tools/perf/Documentation/perf-report.txt
+++ b/tools/perf/Documentation/perf-report.txt
@@ -326,11 +326,12 @@ OPTIONS
 		b	synthesize branches events
 		c	synthesize branches events (calls only)
 		r	synthesize branches events (returns only)
+		x	synthesize transactions events
 		e	synthesize error events
 		d	create a debug log
-		g	synthesize a call chain for instructions events
+		g	synthesize a call chain (use with i or x)
 
-	The default is all events i.e. the same as -Zibe
+	The default is all events i.e. the same as -Zibxe
 
 	In addition, the period (default 100000) for instructions events
 	can be specified in units of:
@@ -341,8 +342,8 @@ OPTIONS
 		us	microseconds
 		ns	nanoseconds (default)
 
-	Also the call chain size (default 16, max. 1024) for instructions
-	events can be specified.
+	Also the call chain size (default 16, max. 1024) for instructions or
+	transactions events can be specified.
 
 	To disable decoding entirely, use --no-itrace.
 
diff --git a/tools/perf/Documentation/perf-script.txt b/tools/perf/Documentation/perf-script.txt
index 88da93a..b46f443c 100644
--- a/tools/perf/Documentation/perf-script.txt
+++ b/tools/perf/Documentation/perf-script.txt
@@ -223,11 +223,12 @@ OPTIONS
 		b	synthesize branches events
 		c	synthesize branches events (calls only)
 		r	synthesize branches events (returns only)
+		x	synthesize transactions events
 		e	synthesize error events
 		d	create a debug log
-		g	synthesize a call chain for instructions events
+		g	synthesize a call chain (use with i or x)
 
-	The default is all events i.e. the same as -Zibe
+	The default is all events i.e. the same as -Zibxe
 
 	In addition, the period (default 100000) for instructions events
 	can be specified in units of:
@@ -238,8 +239,8 @@ OPTIONS
 		us	microseconds
 		ns	nanoseconds (default)
 
-	Also the call chain size (default 16, max. 1024) for instructions
-	events can be specified.
+	Also the call chain size (default 16, max. 1024) for instructions or
+	transactions events can be specified.
 
 	To disable decoding entirely, use --no-itrace.
 
diff --git a/tools/perf/util/auxtrace.c b/tools/perf/util/auxtrace.c
index 5a95767..2308c6e 100644
--- a/tools/perf/util/auxtrace.c
+++ b/tools/perf/util/auxtrace.c
@@ -899,6 +899,7 @@ void itrace_synth_opts__set_default(struct itrace_synth_opts *synth_opts)
 {
 	synth_opts->instructions = true;
 	synth_opts->branches = true;
+	synth_opts->transactions = true;
 	synth_opts->errors = true;
 	synth_opts->period_type = PERF_ITRACE_DEFAULT_PERIOD_TYPE;
 	synth_opts->period = PERF_ITRACE_DEFAULT_PERIOD;
@@ -966,6 +967,9 @@ int itrace_parse_synth_opts(const struct option *opt, const char *str,
 		case 'b':
 			synth_opts->branches = true;
 			break;
+		case 'x':
+			synth_opts->transactions = true;
+			break;
 		case 'e':
 			synth_opts->errors = true;
 			break;
@@ -981,7 +985,6 @@ int itrace_parse_synth_opts(const struct option *opt, const char *str,
 			synth_opts->returns = true;
 			break;
 		case 'g':
-			synth_opts->instructions = true;
 			synth_opts->callchain = true;
 			synth_opts->callchain_sz =
 					PERF_ITRACE_DEFAULT_CALLCHAIN_SZ;
diff --git a/tools/perf/util/auxtrace.h b/tools/perf/util/auxtrace.h
index 8441dee..f1137a5 100644
--- a/tools/perf/util/auxtrace.h
+++ b/tools/perf/util/auxtrace.h
@@ -54,6 +54,7 @@ enum itrace_period_type {
  *          because 'perf inject' will write it out
  * @instructions: whether to synthesize 'instructions' events
  * @branches: whether to synthesize 'branches' events
+ * @transactions: whether to synthesize events for transactions
  * @errors: whether to synthesize decoder error events
  * @dont_decode: whether to skip decoding entirely
  * @log: write a decoding log
@@ -69,6 +70,7 @@ struct itrace_synth_opts {
 	bool			inject;
 	bool			instructions;
 	bool			branches;
+	bool			transactions;
 	bool			errors;
 	bool			dont_decode;
 	bool			log;
-- 
1.9.1

[PATCH V6 25/25] perf script: Add field option 'flags' to print sample flags

[Adrian Hunter]

Instruction tracing will typically have access to information
about the instruction being executed for a particular ip sample.
Some of that information will be available in the 'flags' member
of struct perf_sample.

With the addition of transactions events synthesis to Instruction
Tracing options, there is a need to be able easily to see the
flags because they show whether the ip is at the start, commit or
abort of a tranasaction.

Consequently add an option to display the flags.

The flags are "bcrosyiABEx" which stand for branch, call, return,
conditional, system, asynchronous, interrupt, transaction abort,
trace begin, trace end, and in transaction, respectively.

Example using Intel PT:

perf script -fip,time,event,sym,addr,flags

...
 1288.721584105: branches:u:   bo              401146 main =>           401152 main
 1288.721584105: transactions:   x                   0           401164 main
 1288.721584105: branches:u:   bx              40117c main =>           40119b main
 1288.721584105: branches:u:   box             4011a4 main =>           40117e main
 1288.721584105: branches:u:   bcx             401187 main =>           401094 g
...
 1288.721591645: branches:u:   bx              4010c4 g =>           4010cb g
 1288.721591645: branches:u:   brx             4010cc g =>           401189 main
 1288.721591645: transactions:                       0           4011a6 main
 1288.721593199: branches:u:   b               4011a9 main =>           4011af main
 1288.721593199: branches:u:   bo              4011bc main =>           40113e main
 1288.721593199: branches:u:   b               401150 main =>           40115a main
 1288.721593199: transactions:   x                   0           401164 main
 1288.721593199: branches:u:   bx              40117c main =>           40119b main
 1288.721593199: branches:u:   box             4011a4 main =>           40117e main
 1288.721593199: branches:u:   bcx             401187 main =>           40105e f
...
 1288.722284747: branches:u:   brx             401093 f =>           401189 main
 1288.722284747: branches:u:   box             4011a4 main =>           40117e main
 1288.722284747: branches:u:   bcx             401187 main =>           40105e f
 1288.722285883: transactions:   bA                  0           401071 f
 1288.722285883: branches:u:   bA              401071 f =>           40116a main
 1288.722285883: branches:u:   bE              40116a main =>                0 [unknown]
 1288.722297174: branches:u:   bB                   0 [unknown] =>           40116a main
...

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/Documentation/perf-script.txt |  9 ++++++++-
 tools/perf/builtin-script.c              | 34 +++++++++++++++++++++++++++-----
 tools/perf/util/event.h                  |  2 ++
 3 files changed, 39 insertions(+), 6 deletions(-)

diff --git a/tools/perf/Documentation/perf-script.txt b/tools/perf/Documentation/perf-script.txt
index b46f443c..1598ddd 100644
--- a/tools/perf/Documentation/perf-script.txt
+++ b/tools/perf/Documentation/perf-script.txt
@@ -115,7 +115,8 @@ OPTIONS
 -f::
 --fields::
         Comma separated list of fields to print. Options are:
-        comm, tid, pid, time, cpu, event, trace, ip, sym, dso, addr, symoff, srcline, period.
+        comm, tid, pid, time, cpu, event, trace, ip, sym, dso, addr, symoff,
+	srcline, period, flags.
         Field list can be prepended with the type, trace, sw or hw,
         to indicate to which event type the field list applies.
         e.g., -f sw:comm,tid,time,ip,sym  and -f trace:time,cpu,trace
@@ -165,6 +166,12 @@ OPTIONS
 
 	At this point usage is displayed, and perf-script exits.
 
+	The flags field is synthesized and may have a value when Instruction
+	Trace decoding. The flags are "bcrosyiABEx" which stand for branch,
+	call, return, conditional, system, asynchronous, interrupt,
+	transaction abort, trace begin, trace end, and in transaction,
+	respectively.
+
 	Finally, a user may not set fields to none for all event types.
 	i.e., -f "" is not allowed.
 
diff --git a/tools/perf/builtin-script.c b/tools/perf/builtin-script.c
index d73b55a..7dfe585 100644
--- a/tools/perf/builtin-script.c
+++ b/tools/perf/builtin-script.c
@@ -27,6 +27,7 @@ static u64			nr_unordered;
 static bool			no_callchain;
 static bool			latency_format;
 static bool			system_wide;
+static bool			print_flags;
 static const char		*cpu_list;
 static DECLARE_BITMAP(cpu_bitmap, MAX_NR_CPUS);
 
@@ -461,6 +462,25 @@ static void print_sample_bts(union perf_event *event,
 	printf("\n");
 }
 
+static void print_sample_flags(u32 flags)
+{
+	const char *chars = PERF_IP_FLAG_CHARS;
+	const int n = strlen(PERF_IP_FLAG_CHARS);
+	char str[33];
+	int i, pos = 0;
+
+	for (i = 0; i < n; i++, flags >>= 1) {
+		if (flags & 1)
+			str[pos++] = chars[i];
+	}
+	for (; i < 32; i++, flags >>= 1) {
+		if (flags & 1)
+			str[pos++] = '?';
+	}
+	str[pos] = 0;
+	printf("  %-4s ", str);
+}
+
 static void process_event(union perf_event *event, struct perf_sample *sample,
 			  struct perf_evsel *evsel, struct thread *thread,
 			  struct addr_location *al)
@@ -480,6 +500,9 @@ static void process_event(union perf_event *event, struct perf_sample *sample,
 		printf("%s: ", evname ? evname : "[unknown]");
 	}
 
+	if (print_flags)
+		print_sample_flags(sample->flags);
+
 	if (is_bts_event(attr)) {
 		print_sample_bts(event, sample, evsel, thread, al);
 		return;
@@ -1023,12 +1046,15 @@ static int parse_output_fields(const struct option *opt __maybe_unused,
 		}
 	}
 
-	tok = strtok(tok, ",");
-	while (tok) {
+	for (tok = strtok(tok, ","); tok; tok = strtok(NULL, ",")) {
 		for (i = 0; i < imax; ++i) {
 			if (strcmp(tok, all_output_options[i].str) == 0)
 				break;
 		}
+		if (i == imax && strcmp(tok, "flags") == 0) {
+			print_flags = true;
+			continue;
+		}
 		if (i == imax) {
 			fprintf(stderr, "Invalid field requested.\n");
 			rc = -EINVAL;
@@ -1056,8 +1082,6 @@ static int parse_output_fields(const struct option *opt __maybe_unused,
 			}
 			output[type].fields |= all_output_options[i].field;
 		}
-
-		tok = strtok(NULL, ",");
 	}
 
 	if (type >= 0) {
@@ -1575,7 +1599,7 @@ int cmd_script(int argc, const char **argv, const char *prefix __maybe_unused)
 		     "comma separated output fields prepend with 'type:'. "
 		     "Valid types: hw,sw,trace,raw. "
 		     "Fields: comm,tid,pid,time,cpu,event,trace,ip,sym,dso,"
-		     "addr,symoff,period", parse_output_fields),
+		     "addr,symoff,period,flags", parse_output_fields),
 	OPT_BOOLEAN('a', "all-cpus", &system_wide,
 		    "system-wide collection from all CPUs"),
 	OPT_STRING('S', "symbols", &symbol_conf.sym_list_str, "symbol[,symbol...]",
diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h
index 473ad39..9919232 100644
--- a/tools/perf/util/event.h
+++ b/tools/perf/util/event.h
@@ -157,6 +157,8 @@ enum {
 	PERF_IP_FLAG_IN_TX		= 1ULL << 10,
 };
 
+#define PERF_IP_FLAG_CHARS "bcrosyiABEx"
+
 #define PERF_BRANCH_MASK		(\
 	PERF_IP_FLAG_BRANCH		|\
 	PERF_IP_FLAG_CALL		|\
-- 
1.9.1

Re: [PATCH V6 04/25] perf tools: Add support for AUX area recording

[Jiri Olsa]

On Mon, Mar 16, 2015 at 02:41:26PM +0200, Adrian Hunter wrote:

SNIP

> index 8acd0df..7553788 100644
> --- a/tools/perf/util/record.c
> +++ b/tools/perf/util/record.c
> @@ -119,7 +119,16 @@ void perf_evlist__config(struct perf_evlist *evlist, struct record_opts *opts)
>  			evsel->attr.comm_exec = 1;
>  	}
>  
> -	if (evlist->nr_entries > 1) {
> +	if (opts->full_auxtrace) {
> +		/*
> +		 * Need to be able to synthesize and parse selected events with
> +		 * arbitrary sample types, which requires always being able to
> +		 * match the id.
> +		 */
> +		use_sample_identifier = true;

no need to set use_sample_identifier, you can call
perf_evsel__set_sample_id with true


> +		evlist__for_each(evlist, evsel)
> +			perf_evsel__set_sample_id(evsel, use_sample_identifier);
> +	} else if (evlist->nr_entries > 1) {

but shouldn't you first detect the perf_can_sample_identifier bool same
as the other leg does and use it accordingly? maybe smth like:


	if (evlist->nr_entries > 1 || opts->full_auxtrace) {
		...
	}

	if (opts->full_auxtrace && !use_sample_identifier)
		err..


jirka

Re: [PATCH V6 05/25] perf record: Add basic AUX area tracing support

[Jiri Olsa]

On Mon, Mar 16, 2015 at 02:41:27PM +0200, Adrian Hunter wrote:
> Amend the perf record tool to read the
> AUX area tracing mmap and synthesize
> AUX area tracing events.
> 
> Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
> ---
>  tools/perf/builtin-record.c | 80 ++++++++++++++++++++++++++++++++++++++++++---
>  1 file changed, 76 insertions(+), 4 deletions(-)
> 
> diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
> index 5a2ff51..8b3a50e 100644
> --- a/tools/perf/builtin-record.c
> +++ b/tools/perf/builtin-record.c
> @@ -27,6 +27,7 @@
>  #include "util/cpumap.h"
>  #include "util/thread_map.h"
>  #include "util/data.h"
> +#include "util/auxtrace.h"
>  
>  #include <unistd.h>
>  #include <sched.h>
> @@ -38,6 +39,7 @@ struct record {
>  	struct record_opts	opts;
>  	u64			bytes_written;
>  	struct perf_data_file	file;
> +	struct auxtrace_record	*itr;
>  	struct perf_evlist	*evlist;
>  	struct perf_session	*session;
>  	const char		*progname;
> @@ -110,6 +112,43 @@ out:
>  	return rc;
>  }
>  
> +static int record__process_auxtrace(struct perf_tool *tool,
> +				    union perf_event *event, void *data1,
> +				    size_t len1, void *data2, size_t len2)
> +{
> +	struct record *rec = container_of(tool, struct record, tool);
> +	size_t padding;
> +	u8 pad[8] = {0};
> +
> +	/* event.auxtrace.size includes padding, see __auxtrace_mmap__read() */
> +	padding = (len1 + len2) & 7;
> +	if (padding)
> +		padding = 8 - padding;
> +
> +	record__write(rec, event, event->header.size);
> +	record__write(rec, data1, len1);
> +	record__write(rec, data2, len2);

record__write goes quite deep down the stack before it realizes
there's nothing to write (len2 == 0) which I'm guessing is the
likely case for data2/len2 (buffer wrap around) ?

maybe you want to check on that before calling that

jirka

Re: [PATCH V6 04/25] perf tools: Add support for AUX area recording

[Jiri Olsa]

On Mon, Mar 16, 2015 at 02:41:26PM +0200, Adrian Hunter wrote:

SNIP

> +
> +int perf_event__synthesize_auxtrace(struct perf_tool *tool,
> +				    perf_event__handler_t process,
> +				    size_t size, u64 offset, u64 ref, int idx,
> +				    u32 tid, u32 cpu)
> +{
> +	union perf_event ev;
> +
> +	memset(&ev, 0, sizeof(ev));
> +	ev.auxtrace.header.type = PERF_RECORD_AUXTRACE;
> +	ev.auxtrace.header.size = sizeof(ev.auxtrace);
> +	ev.auxtrace.size = size;
> +	ev.auxtrace.offset = offset;
> +	ev.auxtrace.reference = ref;
> +	ev.auxtrace.idx = idx;
> +	ev.auxtrace.tid = tid;
> +	ev.auxtrace.cpu = cpu;
> +
> +	return process(tool, &ev, NULL, NULL);
> +}

I dont see this one being used/called from any place in your branch,
was it meant to be called from auxtrace_mmap__read?

jirka

Re: [PATCH V6 02/25] perf evlist: Add initial support for mmapping an AUX area buffer

[Jiri Olsa]

On Mon, Mar 16, 2015 at 02:41:24PM +0200, Adrian Hunter wrote:

SNIP

> index 8d0b623..3cc7815 100644
> --- a/tools/perf/util/evlist.c
> +++ b/tools/perf/util/evlist.c
> @@ -725,6 +725,34 @@ void perf_evlist__mmap_consume(struct perf_evlist *evlist, int idx)
>  		perf_evlist__mmap_put(evlist, idx);
>  }
>  
> +int __weak auxtrace_mmap__mmap(struct auxtrace_mmap *mm __maybe_unused,
> +			       struct auxtrace_mmap_params *mp __maybe_unused,
> +			       void *userpg __maybe_unused,
> +			       int fd __maybe_unused)
> +{
> +	return 0;
> +}
> +
> +void __weak auxtrace_mmap__munmap(struct auxtrace_mmap *mm __maybe_unused)
> +{
> +}
> +
> +void __weak auxtrace_mmap_params__init(
> +			struct auxtrace_mmap_params *mp __maybe_unused,
> +			off_t auxtrace_offset __maybe_unused,
> +			unsigned int auxtrace_pages __maybe_unused,
> +			bool auxtrace_overwrite __maybe_unused)
> +{
> +}
> +
> +void __weak auxtrace_mmap_params__set_idx(
> +			struct auxtrace_mmap_params *mp __maybe_unused,
> +			struct perf_evlist *evlist __maybe_unused,
> +			int idx __maybe_unused,
> +			bool per_cpu __maybe_unused)
> +{
> +}

IMO it's better to place the weak auxtrace_* functions in auxtrace header

jirka

Re: [PATCH V6 02/25] perf evlist: Add initial support for mmapping an AUX area buffer

[Jiri Olsa]

On Mon, Mar 16, 2015 at 02:41:24PM +0200, Adrian Hunter wrote:

SNIP

> +#include <linux/bitops.h>
> +#include <linux/log2.h>
> +
> +#include "../perf.h"
> +#include "util.h"
> +#include "evlist.h"
> +#include "cpumap.h"
> +#include "thread_map.h"
> +#include "auxtrace.h"
> +
> +int auxtrace_mmap__mmap(struct auxtrace_mmap *mm,
> +			struct auxtrace_mmap_params *mp,
> +			void *userpg, int fd)
> +{
> +#if BITS_PER_LONG != 64 && !defined(HAVE_SYNC_COMPARE_AND_SWAP_SUPPORT)
> +	pr_err("Cannot use AUX area tracing mmaps\n");
> +	return -1;
> +#endif
> +
> +	mm->userpg = userpg;
> +	mm->mask = mp->mask;
> +	mm->len = mp->len;
> +	mm->prev = 0;
> +	mm->idx = mp->idx;
> +	mm->tid = mp->tid;
> +	mm->cpu = mp->cpu;
> +
> +	if (!mp->len) {
> +		mm->base = NULL;

is it just for safety, or could mm->base be != NULL in here?
maybe good place to use WARN_ON check?

jirka

Re: [PATCH V6 06/25] perf record: Extend -m option for AUX area tracing mmap pages

[Jiri Olsa]

On Mon, Mar 16, 2015 at 02:41:28PM +0200, Adrian Hunter wrote:
> Extend the -m option so that the number
> of mmap pages for AUX area tracing
> can be specified.

I know it's in the patch, but could you please
make it clear in the changelog what's the new
allowed syntax for -m option

jirka

Re: [PATCH V6 02/25] perf evlist: Add initial support for mmapping an AUX area buffer

[Adrian Hunter]

On 24/03/15 13:07, Jiri Olsa wrote:
> On Mon, Mar 16, 2015 at 02:41:24PM +0200, Adrian Hunter wrote:
> 
> SNIP
> 
>> index 8d0b623..3cc7815 100644
>> --- a/tools/perf/util/evlist.c
>> +++ b/tools/perf/util/evlist.c
>> @@ -725,6 +725,34 @@ void perf_evlist__mmap_consume(struct perf_evlist *evlist, int idx)
>>  		perf_evlist__mmap_put(evlist, idx);
>>  }
>>  
>> +int __weak auxtrace_mmap__mmap(struct auxtrace_mmap *mm __maybe_unused,
>> +			       struct auxtrace_mmap_params *mp __maybe_unused,
>> +			       void *userpg __maybe_unused,
>> +			       int fd __maybe_unused)
>> +{
>> +	return 0;
>> +}
>> +
>> +void __weak auxtrace_mmap__munmap(struct auxtrace_mmap *mm __maybe_unused)
>> +{
>> +}
>> +
>> +void __weak auxtrace_mmap_params__init(
>> +			struct auxtrace_mmap_params *mp __maybe_unused,
>> +			off_t auxtrace_offset __maybe_unused,
>> +			unsigned int auxtrace_pages __maybe_unused,
>> +			bool auxtrace_overwrite __maybe_unused)
>> +{
>> +}
>> +
>> +void __weak auxtrace_mmap_params__set_idx(
>> +			struct auxtrace_mmap_params *mp __maybe_unused,
>> +			struct perf_evlist *evlist __maybe_unused,
>> +			int idx __maybe_unused,
>> +			bool per_cpu __maybe_unused)
>> +{
>> +}
> 
> IMO it's better to place the weak auxtrace_* functions in auxtrace header

They are there for python extension linking.