Re: [PATCH v15 0/3]: perf: reduce data loss when profiling highly parallel CPU bound workloads

[Alexey Budankov <alexey.budankov@linux.intel.com>]

Hi,
On 09.11.2018 11:22, Jiri Olsa wrote:
> On Tue, Nov 06, 2018 at 11:53:02AM +0300, Alexey Budankov wrote:
>>
<SNIP>
>>  Alexey Budankov (3):
>> 	perf util: map data buffer for preserving collected data
>> 	perf record: enable asynchronous trace writing
>> 	perf record: extend trace writing to multi AIO
> 
> FYI I was rebasing my threads branch on top of this and
> first 2 won't apply anymore on Arnaldo's perf/core

Thanks for the info. git apply -3 resolves rebasing issue.
Below is the whole patch for Arnaldo's perf/core.

-Alexey

---
 tools/build/Makefile.feature             |   6 +-
 tools/build/feature/Makefile             |   6 +-
 tools/build/feature/test-all.c           |   5 +
 tools/build/feature/test-libaio.c        |  16 ++
 tools/perf/Documentation/perf-record.txt |   5 +
 tools/perf/Makefile.config               |   6 +
 tools/perf/Makefile.perf                 |   7 +-
 tools/perf/builtin-record.c              | 255 ++++++++++++++++++++++++++++++-
 tools/perf/perf.h                        |   1 +
 tools/perf/util/evlist.c                 |   6 +-
 tools/perf/util/evlist.h                 |   2 +-
 tools/perf/util/mmap.c                   | 146 +++++++++++++++++-
 tools/perf/util/mmap.h                   |  26 +++-
 13 files changed, 473 insertions(+), 14 deletions(-)

diff --git a/tools/build/Makefile.feature b/tools/build/Makefile.feature
index f216b2f5c3d7..185b1932b25a 100644
--- a/tools/build/Makefile.feature
+++ b/tools/build/Makefile.feature
@@ -68,7 +68,8 @@ FEATURE_TESTS_BASIC :=                  \
         sched_getcpu			\
         sdt				\
         setns				\
-        libopencsd
+        libopencsd			\
+        libaio
 
 # FEATURE_TESTS_BASIC + FEATURE_TESTS_EXTRA is the complete list
 # of all feature tests
@@ -114,7 +115,8 @@ FEATURE_DISPLAY ?=              \
          zlib                   \
          lzma                   \
          get_cpuid              \
-         bpf
+         bpf			\
+         libaio
 
 # Set FEATURE_CHECK_(C|LD)FLAGS-all for all FEATURE_TESTS features.
 # If in the future we need per-feature checks/flags for features not
diff --git a/tools/build/feature/Makefile b/tools/build/feature/Makefile
index 0516259be70f..468817a6bc44 100644
--- a/tools/build/feature/Makefile
+++ b/tools/build/feature/Makefile
@@ -58,7 +58,8 @@ FILES=                                          \
          test-libopencsd.bin			\
          test-clang.bin				\
          test-llvm.bin				\
-         test-llvm-version.bin
+         test-llvm-version.bin			\
+         test-libaio.bin
 
 FILES := $(addprefix $(OUTPUT),$(FILES))
 
@@ -285,6 +286,9 @@ $(OUTPUT)test-clang.bin:
 
 -include $(OUTPUT)*.d
 
+$(OUTPUT)test-libaio.bin:
+	$(BUILD) -lrt
+
 ###############################
 
 clean:
diff --git a/tools/build/feature/test-all.c b/tools/build/feature/test-all.c
index 8dc20a61341f..19ca9c87cc61 100644
--- a/tools/build/feature/test-all.c
+++ b/tools/build/feature/test-all.c
@@ -166,6 +166,10 @@
 # include "test-libopencsd.c"
 #undef main
 
+#define main main_test_libaio
+# include "test-libaio.c"
+#undef main
+
 int main(int argc, char *argv[])
 {
 	main_test_libpython();
@@ -204,6 +208,7 @@ int main(int argc, char *argv[])
 	main_test_sdt();
 	main_test_setns();
 	main_test_libopencsd();
+	main_test_libaio();
 
 	return 0;
 }
diff --git a/tools/build/feature/test-libaio.c b/tools/build/feature/test-libaio.c
new file mode 100644
index 000000000000..932133c9a265
--- /dev/null
+++ b/tools/build/feature/test-libaio.c
@@ -0,0 +1,16 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <aio.h>
+
+int main(void)
+{
+	struct aiocb aiocb;
+
+	aiocb.aio_fildes  = 0;
+	aiocb.aio_offset  = 0;
+	aiocb.aio_buf     = 0;
+	aiocb.aio_nbytes  = 0;
+	aiocb.aio_reqprio = 0;
+	aiocb.aio_sigevent.sigev_notify = 1 /*SIGEV_NONE*/;
+
+	return (int)aio_return(&aiocb);
+}
diff --git a/tools/perf/Documentation/perf-record.txt b/tools/perf/Documentation/perf-record.txt
index 246dee081efd..d232b13ea713 100644
--- a/tools/perf/Documentation/perf-record.txt
+++ b/tools/perf/Documentation/perf-record.txt
@@ -435,6 +435,11 @@ Specify vmlinux path which has debuginfo.
 --buildid-all::
 Record build-id of all DSOs regardless whether it's actually hit or not.
 
+--aio[=n]::
+Use <n> control blocks in asynchronous (Posix AIO) trace writing mode (default: 1, max: 4).
+Asynchronous mode is supported only when linking Perf tool with libc library
+providing implementation for Posix AIO API.
+
 --all-kernel::
 Configure all used events to run in kernel space.
 
diff --git a/tools/perf/Makefile.config b/tools/perf/Makefile.config
index e30d20fb482d..cf8a35450026 100644
--- a/tools/perf/Makefile.config
+++ b/tools/perf/Makefile.config
@@ -357,6 +357,12 @@ ifeq ($(feature-glibc), 1)
   CFLAGS += -DHAVE_GLIBC_SUPPORT
 endif
 
+ifeq ($(feature-libaio), 1)
+  ifndef NO_AIO
+    CFLAGS += -DHAVE_AIO_SUPPORT
+  endif
+endif
+
 ifdef NO_DWARF
   NO_LIBDW_DWARF_UNWIND := 1
 endif
diff --git a/tools/perf/Makefile.perf b/tools/perf/Makefile.perf
index d95655489f7e..18c820c1b267 100644
--- a/tools/perf/Makefile.perf
+++ b/tools/perf/Makefile.perf
@@ -98,8 +98,13 @@ include ../scripts/utilities.mak
 # Define LIBCLANGLLVM if you DO want builtin clang and llvm support.
 # When selected, pass LLVM_CONFIG=/path/to/llvm-config to `make' if
 # llvm-config is not in $PATH.
-
+#
 # Define NO_CORESIGHT if you do not want support for CoreSight trace decoding.
+#
+# Define NO_AIO if you do not want support of Posix AIO based trace
+# streaming for record mode. Currently Posix AIO trace streaming is
+# supported only when linking with glibc.
+#
 
 # As per kernel Makefile, avoid funny character set dependencies
 unexport LC_ALL
diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
index 488779bc4c8d..4736dc96c4ca 100644
--- a/tools/perf/builtin-record.c
+++ b/tools/perf/builtin-record.c
@@ -124,6 +124,210 @@ static int record__write(struct record *rec, struct perf_mmap *map __maybe_unuse
 	return 0;
 }
 
+#ifdef HAVE_AIO_SUPPORT
+static int record__aio_write(struct aiocb *cblock, int trace_fd,
+		void *buf, size_t size, off_t off)
+{
+	int rc;
+
+	cblock->aio_fildes = trace_fd;
+	cblock->aio_buf    = buf;
+	cblock->aio_nbytes = size;
+	cblock->aio_offset = off;
+	cblock->aio_sigevent.sigev_notify = SIGEV_NONE;
+
+	do {
+		rc = aio_write(cblock);
+		if (rc == 0) {
+			break;
+		} else if (errno != EAGAIN) {
+			cblock->aio_fildes = -1;
+			pr_err("failed to queue perf data, error: %m\n");
+			break;
+		}
+	} while (1);
+
+	return rc;
+}
+
+static int record__aio_complete(struct perf_mmap *md, struct aiocb *cblock)
+{
+	void *rem_buf;
+	off_t rem_off;
+	size_t rem_size;
+	int rc, aio_errno;
+	ssize_t aio_ret, written;
+
+	aio_errno = aio_error(cblock);
+	if (aio_errno == EINPROGRESS)
+		return 0;
+
+	written = aio_ret = aio_return(cblock);
+	if (aio_ret < 0) {
+		if (aio_errno != EINTR)
+			pr_err("failed to write perf data, error: %m\n");
+		written = 0;
+	}
+
+	rem_size = cblock->aio_nbytes - written;
+
+	if (rem_size == 0) {
+		cblock->aio_fildes = -1;
+		/*
+		 * md->refcount is incremented in perf_mmap__push() for
+		 * every enqueued aio write request so decrement it because
+		 * the request is now complete.
+		 */
+		perf_mmap__put(md);
+		rc = 1;
+	} else {
+		/*
+		 * aio write request may require restart with the
+		 * reminder if the kernel didn't write whole
+		 * chunk at once.
+		 */
+		rem_off = cblock->aio_offset + written;
+		rem_buf = (void *)(cblock->aio_buf + written);
+		record__aio_write(cblock, cblock->aio_fildes,
+				rem_buf, rem_size, rem_off);
+		rc = 0;
+	}
+
+	return rc;
+}
+
+static int record__aio_sync(struct perf_mmap *md, bool sync_all)
+{
+	struct aiocb **aiocb = md->aio.aiocb;
+	struct aiocb *cblocks = md->aio.cblocks;
+	struct timespec timeout = { 0, 1000 * 1000  * 1 }; /* 1ms */
+	int i, do_suspend;
+
+	do {
+		do_suspend = 0;
+		for (i = 0; i < md->aio.nr_cblocks; ++i) {
+			if (cblocks[i].aio_fildes == -1 || record__aio_complete(md, &cblocks[i])) {
+				if (sync_all)
+					aiocb[i] = NULL;
+				else
+					return i;
+			} else {
+				/*
+				 * Started aio write is not complete yet
+				 * so it has to be waited before the
+				 * next allocation.
+				 */
+				aiocb[i] = &cblocks[i];
+				do_suspend = 1;
+			}
+		}
+		if (!do_suspend)
+			return -1;
+
+		while (aio_suspend((const struct aiocb **)aiocb, md->aio.nr_cblocks, &timeout)) {
+			if (!(errno == EAGAIN || errno == EINTR))
+				pr_err("failed to sync perf data, error: %m\n");
+		}
+	} while (1);
+}
+
+static int record__aio_pushfn(void *to, struct aiocb *cblock, void *bf, size_t size, off_t off)
+{
+	struct record *rec = to;
+	int ret, trace_fd = rec->session->data->file.fd;
+
+	rec->samples++;
+
+	ret = record__aio_write(cblock, trace_fd, bf, size, off);
+	if (!ret) {
+		rec->bytes_written += size;
+		if (switch_output_size(rec))
+			trigger_hit(&switch_output_trigger);
+	}
+
+	return ret;
+}
+
+static off_t record__aio_get_pos(int trace_fd)
+{
+	return lseek(trace_fd, 0, SEEK_CUR);
+}
+
+static void record__aio_set_pos(int trace_fd, off_t pos)
+{
+	lseek(trace_fd, pos, SEEK_SET);
+}
+
+static void record__aio_mmap_read_sync(struct record *rec)
+{
+	int i;
+	struct perf_evlist *evlist = rec->evlist;
+	struct perf_mmap *maps = evlist->mmap;
+
+	if (!rec->opts.nr_cblocks)
+		return;
+
+	for (i = 0; i < evlist->nr_mmaps; i++) {
+		struct perf_mmap *map = &maps[i];
+
+		if (map->base)
+			record__aio_sync(map, true);
+	}
+}
+
+static int nr_cblocks_default = 1;
+static int nr_cblocks_max = 4;
+
+static int record__aio_parse(const struct option *opt,
+			     const char *str,
+			     int unset)
+{
+	struct record_opts *opts = (struct record_opts *)opt->value;
+
+	if (unset) {
+		opts->nr_cblocks = 0;
+	} else {
+		if (str)
+			opts->nr_cblocks = strtol(str, NULL, 0);
+		if (!opts->nr_cblocks)
+			opts->nr_cblocks = nr_cblocks_default;
+	}
+
+	return 0;
+}
+#else /* HAVE_AIO_SUPPORT */
+static int nr_cblocks_max = 0;
+
+static int record__aio_sync(struct perf_mmap *md __maybe_unused, bool sync_all __maybe_unused)
+{
+	return -1;
+}
+
+static int record__aio_pushfn(void *to __maybe_unused, struct aiocb *cblock __maybe_unused,
+		void *bf __maybe_unused, size_t size __maybe_unused, off_t off __maybe_unused)
+{
+	return -1;
+}
+
+static off_t record__aio_get_pos(int trace_fd __maybe_unused)
+{
+	return -1;
+}
+
+static void record__aio_set_pos(int trace_fd __maybe_unused, off_t pos __maybe_unused)
+{
+}
+
+static void record__aio_mmap_read_sync(struct record *rec __maybe_unused)
+{
+}
+#endif
+
+static int record__aio_enabled(struct record *rec)
+{
+	return rec->opts.nr_cblocks > 0;
+}
+
 static int process_synthesized_event(struct perf_tool *tool,
 				     union perf_event *event,
 				     struct perf_sample *sample __maybe_unused,
@@ -329,7 +533,7 @@ static int record__mmap_evlist(struct record *rec,
 
 	if (perf_evlist__mmap_ex(evlist, opts->mmap_pages,
 				 opts->auxtrace_mmap_pages,
-				 opts->auxtrace_snapshot_mode) < 0) {
+				 opts->auxtrace_snapshot_mode, opts->nr_cblocks) < 0) {
 		if (errno == EPERM) {
 			pr_err("Permission error mapping pages.\n"
 			       "Consider increasing "
@@ -525,6 +729,8 @@ static int record__mmap_read_evlist(struct record *rec, struct perf_evlist *evli
 	int i;
 	int rc = 0;
 	struct perf_mmap *maps;
+	int trace_fd = rec->data.file.fd;
+	off_t off;
 
 	if (!evlist)
 		return 0;
@@ -536,13 +742,30 @@ static int record__mmap_read_evlist(struct record *rec, struct perf_evlist *evli
 	if (overwrite && evlist->bkw_mmap_state != BKW_MMAP_DATA_PENDING)
 		return 0;
 
+	if (record__aio_enabled(rec))
+		off = record__aio_get_pos(trace_fd);
+
 	for (i = 0; i < evlist->nr_mmaps; i++) {
 		struct perf_mmap *map = &maps[i];
 
 		if (map->base) {
-			if (perf_mmap__push(map, rec, record__pushfn) != 0) {
-				rc = -1;
-				goto out;
+			if (!record__aio_enabled(rec)) {
+				if (perf_mmap__push(map, rec, record__pushfn) != 0) {
+					rc = -1;
+					goto out;
+				}
+			} else {
+				int idx;
+				/*
+				 * Call record__aio_sync() to wait till map->data buffer
+				 * becomes available after previous aio write request.
+				 */
+				idx = record__aio_sync(map, false);
+				if (perf_mmap__aio_push(map, rec, idx, record__aio_pushfn, &off) != 0) {
+					record__aio_set_pos(trace_fd, off);
+					rc = -1;
+					goto out;
+				}
 			}
 		}
 
@@ -553,6 +776,9 @@ static int record__mmap_read_evlist(struct record *rec, struct perf_evlist *evli
 		}
 	}
 
+	if (record__aio_enabled(rec))
+		record__aio_set_pos(trace_fd, off);
+
 	/*
 	 * Mark the round finished in case we wrote
 	 * at least one event.
@@ -658,6 +884,8 @@ record__switch_output(struct record *rec, bool at_exit)
 	/* Same Size:      "2015122520103046"*/
 	char timestamp[] = "InvalidTimestamp";
 
+	record__aio_mmap_read_sync(rec);
+
 	record__synthesize(rec, true);
 	if (target__none(&rec->opts.target))
 		record__synthesize_workload(rec, true);
@@ -1168,6 +1396,8 @@ static int __cmd_record(struct record *rec, int argc, const char **argv)
 		record__synthesize_workload(rec, true);
 
 out_child:
+	record__aio_mmap_read_sync(rec);
+
 	if (forks) {
 		int exit_status;
 
@@ -1301,6 +1531,13 @@ static int perf_record_config(const char *var, const char *value, void *cb)
 		var = "call-graph.record-mode";
 		return perf_default_config(var, value, cb);
 	}
+#ifdef HAVE_AIO_SUPPORT
+	if (!strcmp(var, "record.aio")) {
+		rec->opts.nr_cblocks = strtol(value, NULL, 0);
+		if (!rec->opts.nr_cblocks)
+			rec->opts.nr_cblocks = nr_cblocks_default;
+	}
+#endif
 
 	return 0;
 }
@@ -1706,6 +1943,11 @@ static struct option __record_options[] = {
 			  "signal"),
 	OPT_BOOLEAN(0, "dry-run", &dry_run,
 		    "Parse options then exit"),
+#ifdef HAVE_AIO_SUPPORT
+	OPT_CALLBACK_OPTARG(0, "aio", &record.opts,
+		     &nr_cblocks_default, "n", "Use <n> control blocks in asynchronous trace writing mode (default: 1, max: 4)",
+		     record__aio_parse),
+#endif
 	OPT_END()
 };
 
@@ -1898,6 +2140,11 @@ int cmd_record(int argc, const char **argv)
 		goto out;
 	}
 
+	if (rec->opts.nr_cblocks > nr_cblocks_max)
+		rec->opts.nr_cblocks = nr_cblocks_max;
+	if (verbose > 0)
+		pr_info("nr_cblocks: %d\n", rec->opts.nr_cblocks);
+
 	err = __cmd_record(&record, argc, argv);
 out:
 	perf_evlist__delete(rec->evlist);
diff --git a/tools/perf/perf.h b/tools/perf/perf.h
index 0ed4a34c74c4..4d40baa45a5f 100644
--- a/tools/perf/perf.h
+++ b/tools/perf/perf.h
@@ -83,6 +83,7 @@ struct record_opts {
 	clockid_t    clockid;
 	u64          clockid_res_ns;
 	unsigned int proc_map_timeout;
+	int	     nr_cblocks;
 };
 
 struct option;
diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c
index 36526d229315..e90575192209 100644
--- a/tools/perf/util/evlist.c
+++ b/tools/perf/util/evlist.c
@@ -1018,7 +1018,7 @@ int perf_evlist__parse_mmap_pages(const struct option *opt, const char *str,
  */
 int perf_evlist__mmap_ex(struct perf_evlist *evlist, unsigned int pages,
 			 unsigned int auxtrace_pages,
-			 bool auxtrace_overwrite)
+			 bool auxtrace_overwrite, int nr_cblocks)
 {
 	struct perf_evsel *evsel;
 	const struct cpu_map *cpus = evlist->cpus;
@@ -1028,7 +1028,7 @@ int perf_evlist__mmap_ex(struct perf_evlist *evlist, unsigned int pages,
 	 * Its value is decided by evsel's write_backward.
 	 * So &mp should not be passed through const pointer.
 	 */
-	struct mmap_params mp;
+	struct mmap_params mp = { .nr_cblocks = nr_cblocks };
 
 	if (!evlist->mmap)
 		evlist->mmap = perf_evlist__alloc_mmap(evlist, false);
@@ -1060,7 +1060,7 @@ int perf_evlist__mmap_ex(struct perf_evlist *evlist, unsigned int pages,
 
 int perf_evlist__mmap(struct perf_evlist *evlist, unsigned int pages)
 {
-	return perf_evlist__mmap_ex(evlist, pages, 0, false);
+	return perf_evlist__mmap_ex(evlist, pages, 0, false, 0);
 }
 
 int perf_evlist__create_maps(struct perf_evlist *evlist, struct target *target)
diff --git a/tools/perf/util/evlist.h b/tools/perf/util/evlist.h
index d108d167eb36..868294491194 100644
--- a/tools/perf/util/evlist.h
+++ b/tools/perf/util/evlist.h
@@ -162,7 +162,7 @@ unsigned long perf_event_mlock_kb_in_pages(void);
 
 int perf_evlist__mmap_ex(struct perf_evlist *evlist, unsigned int pages,
 			 unsigned int auxtrace_pages,
-			 bool auxtrace_overwrite);
+			 bool auxtrace_overwrite, int nr_cblocks);
 int perf_evlist__mmap(struct perf_evlist *evlist, unsigned int pages);
 void perf_evlist__munmap(struct perf_evlist *evlist);
 
diff --git a/tools/perf/util/mmap.c b/tools/perf/util/mmap.c
index cdb95b3a1213..ab30555d2afc 100644
--- a/tools/perf/util/mmap.c
+++ b/tools/perf/util/mmap.c
@@ -153,8 +153,152 @@ void __weak auxtrace_mmap_params__set_idx(struct auxtrace_mmap_params *mp __mayb
 {
 }
 
+#ifdef HAVE_AIO_SUPPORT
+static int perf_mmap__aio_mmap(struct perf_mmap *map, struct mmap_params *mp)
+{
+	int delta_max, i, prio;
+
+	map->aio.nr_cblocks = mp->nr_cblocks;
+	if (map->aio.nr_cblocks) {
+		map->aio.aiocb = calloc(map->aio.nr_cblocks, sizeof(struct aiocb *));
+		if (!map->aio.aiocb) {
+			pr_debug2("failed to allocate aiocb for data buffer, error %m\n");
+			return -1;
+		}
+		map->aio.cblocks = calloc(map->aio.nr_cblocks, sizeof(struct aiocb));
+		if (!map->aio.cblocks) {
+			pr_debug2("failed to allocate cblocks for data buffer, error %m\n");
+			return -1;
+		}
+		map->aio.data = calloc(map->aio.nr_cblocks, sizeof(void *));
+		if (!map->aio.data) {
+			pr_debug2("failed to allocate data buffer, error %m\n");
+			return -1;
+		}
+		delta_max = sysconf(_SC_AIO_PRIO_DELTA_MAX);
+		for (i = 0; i < map->aio.nr_cblocks; ++i) {
+			map->aio.data[i] = malloc(perf_mmap__mmap_len(map));
+			if (!map->aio.data[i]) {
+				pr_debug2("failed to allocate data buffer area, error %m");
+				return -1;
+			}
+			/*
+			 * Use cblock.aio_fildes value different from -1
+			 * to denote started aio write operation on the
+			 * cblock so it requires explicit record__aio_sync()
+			 * call prior the cblock may be reused again.
+			 */
+			map->aio.cblocks[i].aio_fildes = -1;
+			/*
+			 * Allocate cblocks with priority delta to have
+			 * faster aio write system calls because queued requests
+			 * are kept in separate per-prio queues and adding
+			 * a new request will iterate thru shorter per-prio
+			 * list. Blocks with numbers higher than
+			 *  _SC_AIO_PRIO_DELTA_MAX go with priority 0.
+			 */
+			prio = delta_max - i;
+			map->aio.cblocks[i].aio_reqprio = prio >= 0 ? prio : 0;
+		}
+	}
+
+	return 0;
+}
+
+static void perf_mmap__aio_munmap(struct perf_mmap *map)
+{
+	if (map->aio.data)
+		zfree(&map->aio.data);
+}
+
+int perf_mmap__aio_push(struct perf_mmap *md, void *to, int idx,
+			int push(void *to, struct aiocb *cblock, void *buf, size_t size, off_t off),
+			off_t *off)
+{
+	u64 head = perf_mmap__read_head(md);
+	unsigned char *data = md->base + page_size;
+	unsigned long size, size0 = 0;
+	void *buf;
+	int rc = 0;
+
+	rc = perf_mmap__read_init(md);
+	if (rc < 0)
+		return (rc == -EAGAIN) ? 0 : -1;
+
+	/*
+	 * md->base data is copied into md->data[idx] buffer to
+	 * release space in the kernel buffer as fast as possible,
+	 * thru perf_mmap__consume() below.
+	 *
+	 * That lets the kernel to proceed with storing more
+	 * profiling data into the kernel buffer earlier than other
+	 * per-cpu kernel buffers are handled.
+	 *
+	 * Coping can be done in two steps in case the chunk of
+	 * profiling data crosses the upper bound of the kernel buffer.
+	 * In this case we first move part of data from md->start
+	 * till the upper bound and then the reminder from the
+	 * beginning of the kernel buffer till the end of
+	 * the data chunk.
+	 */
+
+	size = md->end - md->start;
+
+	if ((md->start & md->mask) + size != (md->end & md->mask)) {
+		buf = &data[md->start & md->mask];
+		size = md->mask + 1 - (md->start & md->mask);
+		md->start += size;
+		memcpy(md->aio.data[idx], buf, size);
+		size0 = size;
+	}
+
+	buf = &data[md->start & md->mask];
+	size = md->end - md->start;
+	md->start += size;
+	memcpy(md->aio.data[idx] + size0, buf, size);
+
+	/*
+	 * Increment md->refcount to guard md->data[idx] buffer
+	 * from premature deallocation because md object can be
+	 * released earlier than aio write request started
+	 * on mmap->data[idx] is complete.
+	 *
+	 * perf_mmap__put() is done at record__aio_complete()
+	 * after started request completion.
+	 */
+	perf_mmap__get(md);
+
+	md->prev = head;
+	perf_mmap__consume(md);
+
+	rc = push(to, &md->aio.cblocks[idx], md->aio.data[idx], size0 + size, *off);
+	if (!rc) {
+		*off += size0 + size;
+	} else {
+		/*
+		 * Decrement md->refcount back if aio write
+		 * operation failed to start.
+		 */
+		perf_mmap__put(md);
+	}
+
+	return rc;
+}
+#else
+static int perf_mmap__aio_mmap(struct perf_mmap *map __maybe_unused,
+			       struct mmap_params *mp __maybe_unused)
+{
+	return 0;
+}
+
+static void perf_mmap__aio_munmap(struct perf_mmap *map __maybe_unused)
+{
+}
+#endif
+
 void perf_mmap__munmap(struct perf_mmap *map)
 {
+	perf_mmap__aio_munmap(map);
 	if (map->base != NULL) {
 		munmap(map->base, perf_mmap__mmap_len(map));
 		map->base = NULL;
@@ -197,7 +341,7 @@ int perf_mmap__mmap(struct perf_mmap *map, struct mmap_params *mp, int fd, int c
 				&mp->auxtrace_mp, map->base, fd))
 		return -1;
 
-	return 0;
+	return perf_mmap__aio_mmap(map, mp);
 }
 
 static int overwrite_rb_find_range(void *buf, int mask, u64 *start, u64 *end)
diff --git a/tools/perf/util/mmap.h b/tools/perf/util/mmap.h
index cc5e2d6d17a9..aeb6942fdb00 100644
--- a/tools/perf/util/mmap.h
+++ b/tools/perf/util/mmap.h
@@ -6,9 +6,13 @@
 #include <linux/types.h>
 #include <linux/ring_buffer.h>
 #include <stdbool.h>
+#ifdef HAVE_AIO_SUPPORT
+#include <aio.h>
+#endif
 #include "auxtrace.h"
 #include "event.h"
 
+struct aiocb;
 /**
  * struct perf_mmap - perf's ring buffer mmap details
  *
@@ -26,6 +30,14 @@ struct perf_mmap {
 	bool		 overwrite;
 	struct auxtrace_mmap auxtrace_mmap;
 	char		 event_copy[PERF_SAMPLE_MAX_SIZE] __aligned(8);
+#ifdef HAVE_AIO_SUPPORT
+	struct {
+		void		 **data;
+		struct aiocb	 *cblocks;
+		struct aiocb	 **aiocb;
+		int		 nr_cblocks;
+	} aio;
+#endif
 };
 
 /*
@@ -57,7 +69,7 @@ enum bkw_mmap_state {
 };
 
 struct mmap_params {
-	int			    prot, mask;
+	int			    prot, mask, nr_cblocks;
 	struct auxtrace_mmap_params auxtrace_mp;
 };
 
@@ -85,6 +97,18 @@ union perf_event *perf_mmap__read_event(struct perf_mmap *map);
 
 int perf_mmap__push(struct perf_mmap *md, void *to,
 		    int push(struct perf_mmap *map, void *to, void *buf, size_t size));
+#ifdef HAVE_AIO_SUPPORT
+int perf_mmap__aio_push(struct perf_mmap *md, void *to, int idx,
+			int push(void *to, struct aiocb *cblock, void *buf, size_t size, off_t off),
+			off_t *off);
+#else
+static inline int perf_mmap__aio_push(struct perf_mmap *md __maybe_unused, void *to __maybe_unused, int idx __maybe_unused,
+	int push(void *to, struct aiocb *cblock, void *buf, size_t size, off_t off) __maybe_unused,
+	off_t *off __maybe_unused)
+{
+	return 0;
+}
+#endif
 
 size_t perf_mmap__mmap_len(struct perf_mmap *map);