Skip to contentSkip to conversation

: perf: reduce data loss when profiling highly parallel CPU bound workloads

PATCH Superseded Alexey Budankov 4 participants 14 messages Oct 15, 2018 – Nov 2, 2018 perf events

v14 of 15 Superseded Latest is v15 →

lore
Share & tools
lore
#0 lore raw

v14 cover : perf: reduce data loss when profiling highly parallel CPU bound workloads

Currently in record mode the tool implements trace writing serially. The algorithm loops over mapped per-cpu data buffers and stores ready data chunks into a trace file using write() system call.

At some circumstances the kernel may lack free space in a buffer because the other buffer's half is not yet written to disk due to some other buffer's data writing by the tool at the moment.

Thus serial trace writing implementation may cause the kernel to loose profiling data and that is what observed when profiling highly parallel CPU bound workloads on machines with big number of cores.

Experiment with profiling matrix multiplication code executing 128 threads on Intel Xeon Phi (KNM) with 272 cores, like below, demonstrates data loss metrics value of 98%:

/usr/bin/time perf record -o /tmp/perf-ser.data -a -N -B -T -R -g \
    --call-graph dwarf,1024 --user-regs=IP,SP,BP --switch-events \
    -e cycles,instructions,ref-cycles,software/period=1,name=cs,config=0x3/Duk -- \
    matrix.gcc

Data loss metrics is the ratio lost_time/elapsed_time where lost_time is the sum of time intervals containing PERF_RECORD_LOST records and elapsed_time is the elapsed application run time under profiling.

Applying asynchronous trace streaming thru Posix AIO API [1] lowers data loss metrics value providing 2x improvement (from 98% to ~1%)

Asynchronous trace streaming is currently limited to glibc linkage. musl libc [5] also provides Posix AIO API implementation, however the patchkit is not tested with it. There may be other libc libraries linked by Perf tool that currently lack Posix AIO API support [2], [3], [4] so NO_AIO define may be used to limit Perf tool binary to serial streaming only.

---
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

 tools/perf/Documentation/perf-record.txt |   5 +
 tools/perf/Makefile.config               |   5 +
 tools/perf/Makefile.perf                 |   7 +-
 tools/perf/builtin-record.c              | 252 ++++++++++++++++++++++++++++++-
 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 +++-
 9 files changed, 439 insertions(+), 11 deletions(-)

---
Changes in v14:
- implement default nr_cblocks_default variable
- fix --aio option handling
Changes in v13:
- named new functions with _aio_ word
- grouped aio functions under single #ifdef HAVE_AIO_SUPPORT
- moved perf_mmap__aio_push() stub into header
- removed trailed white space
Changes in v12:
- applied stub functions design for the whole patch kit
- grouped AIO related data into a struct under struct perf_mmap
- implemented record__aio_get/set_pos(), record__aio_enabled()
- implemented simple --aio option
- extended --aio option to --aio-cblocks=<n>
Changes in v11:
- replacing the both lseek() syscalls in every loop iteration by the only
two syscalls just before and after the loop at record__mmap_read_evlist() and advancing *in-flight* off file pos value at perf_mmap__aio_push()
Changes in v10:
- moved specific code to perf_mmap__aio_mmap(), perf_mmap__aio_munmap();
- adjusted error reporting by using %m
- avoided lseek() setting file pos back in case of record__aio_write() failure
- compacted code selecting between serial and AIO streaming
- optimized call places of record__mmap_read_sync()
- added description of aio-cblocks option into perf-record.txt
Changes in v9:
- enable AIO streaming only when --aio-cblocks option is specified explicitly
- enable AIO based implementation when linking with glibc only
- define NO_AIO to limit Perf binary to serial implementation
Changes in v8:
- run the whole thing thru checkpatch.pl and corrected found issues except lines longer than 80 symbols
- corrected comments alignment and formatting
- moved multi AIO implementation into 3rd patch in the series
- implemented explicit cblocks array allocation
- split AIO completion check into separate record__aio_complete()
- set nr_cblocks default to 1 and max allowed value to 4
Changes in v7:
- implemented handling record.aio setting from perfconfig file
Changes in v6:
- adjusted setting of priorities for cblocks;
- handled errno == EAGAIN case from aio_write() return;
Changes in v5:
- resolved livelock on perf record -e intel_pt// -- dd if=/dev/zero of=/dev/null count=100000
- data loss metrics decreased from 25% to 2x in trialed configuration;
- reshaped layout of data structures;
- implemented --aio option;
- avoided nanosleep() prior calling aio_suspend();
- switched to per-cpu aio multi buffer record__aio_sync();
- record_mmap_read_sync() now does global sync just before
switching trace file or collection stop;
Changes in v4:
- converted mmap()/munmap() to malloc()/free() for mmap->data buffer management
- converted void *bf to struct perf_mmap *md in signatures
- written comment in perf_mmap__push() just before perf_mmap__get();
- written comment in record__mmap_read_sync() on possible restarting
of aio_write() operation and releasing perf_mmap object after all;
- added perf_mmap__put() for the cases of failed aio_write();
Changes in v3:
- written comments about nanosleep(0.5ms) call prior aio_suspend()
to cope with intrusiveness of its implementation in glibc;
- written comments about rationale behind coping profiling data
into mmap->data buffer;
Changes in v2:
- converted zalloc() to calloc() for allocation of mmap_aio array,
- cleared typo and adjusted fallback branch code;

---

[1] http://man7.org/linux/man-pages/man7/aio.7.html
[2] https://android.googlesource.com/platform/bionic/+/master/docs/status.md
[3] https://www.uclibc.org/
[4] https://uclibc-ng.org/
[5] https://www.musl-libc.org/

#1 lore raw

PATCH v14 1/3 : perf util: map data buffer for preserving collected data

The map->data buffer is used to preserve map->base profiling data for writing to disk. AIO map->cblock is used to queue corresponding map->data buffer for asynchronous writing.

---
Changes in v13:
- grouped aio functions under single #ifdef HAVE_AIO_SUPPORT
Changes in v12:
- applied stub functions design for the whole patch kit
- grouped AIO related data into a struct under struct perf_mmap
Changes in v10:
- moved specific code to perf_mmap__aio_mmap(), perf_mmap__aio_munmap()
- adjusted error reporting by using %m
Changes in v9:
- implemented NO_AIO and HAVE_AIO_SUPPORT defines to cover cases of libc implementations without Posix AIO API support
Changes in v7:
- implemented handling record.aio setting from perfconfig file
Changes in v6:
- adjusted setting of priorities for cblocks;
Changes in v5:
- reshaped layout of data structures;
- implemented --aio option;
Changes in v4:
- converted mmap()/munmap() to malloc()/free() for mmap->data buffer management
Changes in v2:
- converted zalloc() to calloc() for allocation of mmap_aio array,
- cleared typo and adjusted fallback branch code;

---
 tools/perf/Makefile.config |  5 +++++
 tools/perf/Makefile.perf   |  7 ++++++-
 tools/perf/util/evlist.c   |  2 +-
 tools/perf/util/mmap.c     | 49 +++++++++++++++++++++++++++++++++++++++++++++-
 tools/perf/util/mmap.h     | 11 ++++++++++-
 5 files changed, 70 insertions(+), 4 deletions(-)
tools/perf/Makefile.config +5 -0
index f6d1a03c7523..2e90f4ce9214 100644
@@ -355,6 +355,11 @@ endif # NO_LIBELF
 
 ifeq ($(feature-glibc), 1)
   CFLAGS += -DHAVE_GLIBC_SUPPORT
+  ifndef NO_AIO
+    ifndef BIONIC
+        CFLAGS += -DHAVE_AIO_SUPPORT
+    endif
+  endif
 endif
 
 ifdef NO_DWARF
tools/perf/Makefile.perf +6 -1
index 92514fb3689f..7becc6a72cf2 100644
@@ -97,8 +97,13 @@ 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
tools/perf/util/evlist.c +1 -1
index be440df29615..1a83bf2c069c 100644
@@ -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 = 0 };
 
 	if (!evlist->mmap)
 		evlist->mmap = perf_evlist__alloc_mmap(evlist, false);
tools/perf/util/mmap.c +48 -1

63 lines not on this page · raw message

tools/perf/util/mmap.h +10 -1
index e603314dc792..a46dbdcdcc8a 100644
@@ -6,6 +6,9 @@
 #include <linux/types.h>
 #include <asm/barrier.h>
 #include <stdbool.h>
+#ifdef HAVE_AIO_SUPPORT
+#include <aio.h>
+#endif
 #include "auxtrace.h"
 #include "event.h"
 
@@ -26,6 +29,12 @@ 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	 cblock;
+	} aio;
+#endif
 };
 
 /*
@@ -57,7 +66,7 @@ enum bkw_mmap_state {
 };
 
 struct mmap_params {
-	int			    prot, mask;
+	int			    prot, mask, nr_cblocks;
 	struct auxtrace_mmap_params auxtrace_mp;
 };
 
Acked-by on v14 1/3
#2 lore raw

Re: [PATCH v14 1/3]: perf util: map data buffer for preserving collected data

Quoted 7 lines · Alexey Budankov

The map->data buffer is used to preserve map->base profiling data for writing to disk. AIO map->cblock is used to queue corresponding map->data buffer for asynchronous writing.

Shall we add libaio to tools/build/Makefile.feature? It will be great to see whether aio is enabled in "Auto-detecting system features:".

Thanks for implementing this!

Song

Quoted 185 lines

---
Changes in v13:
- grouped aio functions under single #ifdef HAVE_AIO_SUPPORT
Changes in v12:
- applied stub functions design for the whole patch kit
- grouped AIO related data into a struct under struct perf_mmap
Changes in v10:
- moved specific code to perf_mmap__aio_mmap(), perf_mmap__aio_munmap()
- adjusted error reporting by using %m
Changes in v9:
- implemented NO_AIO and HAVE_AIO_SUPPORT defines to cover cases of libc implementations without Posix AIO API support
Changes in v7:
- implemented handling record.aio setting from perfconfig file
Changes in v6:
- adjusted setting of priorities for cblocks;
Changes in v5:
- reshaped layout of data structures;
- implemented --aio option;
Changes in v4:
- converted mmap()/munmap() to malloc()/free() for mmap->data buffer management
Changes in v2:
- converted zalloc() to calloc() for allocation of mmap_aio array,
- cleared typo and adjusted fallback branch code;

---
 tools/perf/Makefile.config |  5 +++++
 tools/perf/Makefile.perf   |  7 ++++++-
 tools/perf/util/evlist.c   |  2 +-
 tools/perf/util/mmap.c     | 49 +++++++++++++++++++++++++++++++++++++++++++++-
 tools/perf/util/mmap.h     | 11 ++++++++++-
 5 files changed, 70 insertions(+), 4 deletions(-)
tools/perf/Makefile.config +5 -0
index f6d1a03c7523..2e90f4ce9214 100644
@@ -355,6 +355,11 @@ endif # NO_LIBELF
 
 ifeq ($(feature-glibc), 1)
   CFLAGS += -DHAVE_GLIBC_SUPPORT
+  ifndef NO_AIO
+    ifndef BIONIC
+        CFLAGS += -DHAVE_AIO_SUPPORT
+    endif
+  endif
 endif
 
 ifdef NO_DWARF
tools/perf/Makefile.perf +6 -1
index 92514fb3689f..7becc6a72cf2 100644
@@ -97,8 +97,13 @@ 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
tools/perf/util/evlist.c +1 -1
index be440df29615..1a83bf2c069c 100644
@@ -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 = 0 };
 
        if (!evlist->mmap)
                evlist->mmap = perf_evlist__alloc_mmap(evlist, false);
tools/perf/util/mmap.c +48 -1
index cdb95b3a1213..47cdc3ad6546 100644
@@ -153,8 +153,55 @@ 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;
+
+       if (mp->nr_cblocks) {
+               map->aio.data = malloc(perf_mmap__mmap_len(map));
+               if (!map->aio.data) {
+                       pr_debug2("failed to allocate data buffer, error %m\n");
+                       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.cblock.aio_fildes = -1;
+               /*
+                * Allocate cblock with max priority delta to
+                * have faster aio write system calls.
+                */
+               delta_max = sysconf(_SC_AIO_PRIO_DELTA_MAX);
+               map->aio.cblock.aio_reqprio = delta_max;
+       }
+
+       return 0;
+}
+
+static void perf_mmap__aio_munmap(struct perf_mmap *map)
+{
+       if (map->aio.data)
+               zfree(&map->aio.data);
+}
+#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 +244,7 @@ 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)
tools/perf/util/mmap.h +10 -1
index e603314dc792..a46dbdcdcc8a 100644
@@ -6,6 +6,9 @@
 #include <linux/types.h>
 #include <asm/barrier.h>
 #include <stdbool.h>
+#ifdef HAVE_AIO_SUPPORT
+#include <aio.h>
+#endif
 #include "auxtrace.h"
 #include "event.h"
 
@@ -26,6 +29,12 @@ 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     cblock;
+       } aio;
+#endif
 };
 
 /*
@@ -57,7 +66,7 @@ enum bkw_mmap_state {
 };
 
 struct mmap_params {
-       int                         prot, mask;
+       int                         prot, mask, nr_cblocks;
        struct auxtrace_mmap_params auxtrace_mp;
 };
 
#3 lore raw

Re: [PATCH v14 1/3]: perf util: map data buffer for preserving collected data

Hi,

Quoted 11 lines · Song Liu
Quoted 7 lines · Alexey Budankov

The map->data buffer is used to preserve map->base profiling data for writing to disk. AIO map->cblock is used to queue corresponding map->data buffer for asynchronous writing.

Thanks for review!

Quoted 3 lines

Shall we add libaio to tools/build/Makefile.feature? It will be great to see whether aio is enabled in "Auto-detecting system features:".

Yes. This makes sense. There are targeted libc implementations that currently lack Posix AIO API support required for this patch kit.

Thanks,
Alexey

Quoted 191 lines

Thanks for implementing this!

Song

Quoted 185 lines

---
Changes in v13:
- grouped aio functions under single #ifdef HAVE_AIO_SUPPORT
Changes in v12:
- applied stub functions design for the whole patch kit
- grouped AIO related data into a struct under struct perf_mmap
Changes in v10:
- moved specific code to perf_mmap__aio_mmap(), perf_mmap__aio_munmap()
- adjusted error reporting by using %m
Changes in v9:
- implemented NO_AIO and HAVE_AIO_SUPPORT defines to cover cases of libc implementations without Posix AIO API support
Changes in v7:
- implemented handling record.aio setting from perfconfig file
Changes in v6:
- adjusted setting of priorities for cblocks;
Changes in v5:
- reshaped layout of data structures;
- implemented --aio option;
Changes in v4:
- converted mmap()/munmap() to malloc()/free() for mmap->data buffer management
Changes in v2:
- converted zalloc() to calloc() for allocation of mmap_aio array,
- cleared typo and adjusted fallback branch code;

---
 tools/perf/Makefile.config |  5 +++++
 tools/perf/Makefile.perf   |  7 ++++++-
 tools/perf/util/evlist.c   |  2 +-
 tools/perf/util/mmap.c     | 49 +++++++++++++++++++++++++++++++++++++++++++++-
 tools/perf/util/mmap.h     | 11 ++++++++++-
 5 files changed, 70 insertions(+), 4 deletions(-)
tools/perf/Makefile.config +5 -0
index f6d1a03c7523..2e90f4ce9214 100644
@@ -355,6 +355,11 @@ endif # NO_LIBELF
 
 ifeq ($(feature-glibc), 1)
   CFLAGS += -DHAVE_GLIBC_SUPPORT
+  ifndef NO_AIO
+    ifndef BIONIC
+        CFLAGS += -DHAVE_AIO_SUPPORT
+    endif
+  endif
 endif
 
 ifdef NO_DWARF
tools/perf/Makefile.perf +6 -1
index 92514fb3689f..7becc6a72cf2 100644
@@ -97,8 +97,13 @@ 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
tools/perf/util/evlist.c +1 -1
index be440df29615..1a83bf2c069c 100644
@@ -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 = 0 };
 
        if (!evlist->mmap)
                evlist->mmap = perf_evlist__alloc_mmap(evlist, false);
tools/perf/util/mmap.c +48 -1
index cdb95b3a1213..47cdc3ad6546 100644
@@ -153,8 +153,55 @@ 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;
+
+       if (mp->nr_cblocks) {
+               map->aio.data = malloc(perf_mmap__mmap_len(map));
+               if (!map->aio.data) {
+                       pr_debug2("failed to allocate data buffer, error %m\n");
+                       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.cblock.aio_fildes = -1;
+               /*
+                * Allocate cblock with max priority delta to
+                * have faster aio write system calls.
+                */
+               delta_max = sysconf(_SC_AIO_PRIO_DELTA_MAX);
+               map->aio.cblock.aio_reqprio = delta_max;
+       }
+
+       return 0;
+}
+
+static void perf_mmap__aio_munmap(struct perf_mmap *map)
+{
+       if (map->aio.data)
+               zfree(&map->aio.data);
+}
+#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 +244,7 @@ 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)
tools/perf/util/mmap.h +10 -1
index e603314dc792..a46dbdcdcc8a 100644
@@ -6,6 +6,9 @@
 #include <linux/types.h>
 #include <asm/barrier.h>
 #include <stdbool.h>
+#ifdef HAVE_AIO_SUPPORT
+#include <aio.h>
+#endif
 #include "auxtrace.h"
 #include "event.h"
 
@@ -26,6 +29,12 @@ 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     cblock;
+       } aio;
+#endif
 };
 
 /*
@@ -57,7 +66,7 @@ enum bkw_mmap_state {
 };
 
 struct mmap_params {
-       int                         prot, mask;
+       int                         prot, mask, nr_cblocks;
        struct auxtrace_mmap_params auxtrace_mp;
 };
 
#4 lore raw

PATCH v14 2/3 : perf record: enable asynchronous trace writing

In reply to Alexey Budankov

Trace file offset is read once before mmaps iterating loop and written back after all performance data enqueued for aio writing. Trace file offset is incremented linearly after every successful aio write operation.

record__aio_sync() blocks till completion of started AIO operation and then proceeds.

record__aio_mmap_read_sync() implements a barrier for all incomplete aio write requests.

---
Changes in v14:
- implement default nr_cblocks_default variable
Changes in v13:
- named new functions with _aio_ word
- grouped aio functions under single #ifdef HAVE_AIO_SUPPORT
- moved perf_mmap__aio_push() stub into header
- removed trailed white space
Changes in v12:
- implemented record__aio_get/set_pos(), record__aio_enabled()
- implemented simple --aio option
Changes in v11:
- replacing the both lseek() syscalls in every loop iteration by the only
two syscalls just before and after the loop at record__mmap_read_evlist() and advancing *in-flight* off file pos value at perf_mmap__aio_push()
Changes in v10:
- avoided lseek() setting file pos back in case of record__aio_write() failure
- compacted code selecting between serial and AIO streaming
- optimized call places of record__mmap_read_sync()
Changes in v9:
- enable AIO streaming only when --aio-cblocks option is specified explicitly Changes in v8:
- split AIO completion check into separate record__aio_complete()
Changes in v6:
- handled errno == EAGAIN case from aio_write();
Changes in v5:
- data loss metrics decreased from 25% to 2x in trialed configuration;
- avoided nanosleep() prior calling aio_suspend();
- switched to per cpu multi record__aio_sync() aio
- record_mmap_read_sync() now does global barrier just before
switching trace file or collection stop;
- resolved livelock on perf record -e intel_pt// -- dd if=/dev/zero of=/dev/null count=100000 Changes in v4:
- converted void *bf to struct perf_mmap *md in signatures
- written comment in perf_mmap__push() just before perf_mmap__get();
- written comment in record__mmap_read_sync() on possible restarting
of aio_write() operation and releasing perf_mmap object after all;
- added perf_mmap__put() for the cases of failed aio_write();
Changes in v3:
- written comments about nanosleep(0.5ms) call prior aio_suspend()
to cope with intrusiveness of its implementation in glibc;
- written comments about rationale behind coping profiling data
into mmap->data buffer;

---
 tools/perf/Documentation/perf-record.txt |   5 +
 tools/perf/builtin-record.c              | 218 ++++++++++++++++++++++++++++++-
 tools/perf/perf.h                        |   1 +
 tools/perf/util/evlist.c                 |   6 +-
 tools/perf/util/evlist.h                 |   2 +-
 tools/perf/util/mmap.c                   |  77 ++++++++++-
 tools/perf/util/mmap.h                   |  14 ++
 7 files changed, 314 insertions(+), 9 deletions(-)
tools/perf/Documentation/perf-record.txt +5 -0
index 246dee081efd..7efb4af88a68 100644
@@ -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::
+Enable asynchronous (Posix AIO) trace writing mode.
+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.
 
tools/perf/builtin-record.c +214 -4

276 lines not on this page · raw message

tools/perf/perf.h +1 -0
index 21bf7f5a3cf5..0a1ae2ae567a 100644
@@ -82,6 +82,7 @@ struct record_opts {
 	bool         use_clockid;
 	clockid_t    clockid;
 	unsigned int proc_map_timeout;
+	int	     nr_cblocks;
 };
 
 struct option;
tools/perf/util/evlist.c +3 -3
index 1a83bf2c069c..6593bd0dc2af 100644
@@ -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 = { .nr_cblocks = 0 };
+	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)
tools/perf/util/evlist.h +1 -1
index dc66436add98..2464463879b4 100644
@@ -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);
 
tools/perf/util/mmap.c +76 -1

89 lines not on this page · raw message

tools/perf/util/mmap.h +14 -0
index a46dbdcdcc8a..9be06f18f17a 100644
@@ -12,6 +12,7 @@
 #include "auxtrace.h"
 #include "event.h"
 
+struct aiocb;
 /**
  * struct perf_mmap - perf's ring buffer mmap details
  *
@@ -33,6 +34,7 @@ struct perf_mmap {
 	struct {
 		void 		 *data;
 		struct aiocb	 cblock;
+		int		 nr_cblocks;
 	} aio;
 #endif
 };
@@ -103,6 +105,18 @@ 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 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 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);
 
Acked-by on v14 2/3
#5 lore raw

Re: [PATCH v14 2/3]: perf record: enable asynchronous trace writing

Quoted 362 lines · Alexey Budankov

Trace file offset is read once before mmaps iterating loop and written back after all performance data enqueued for aio writing. Trace file offset is incremented linearly after every successful aio write operation.

record__aio_sync() blocks till completion of started AIO operation and then proceeds.

record__aio_mmap_read_sync() implements a barrier for all incomplete aio write requests.

---
Changes in v14:
- implement default nr_cblocks_default variable
Changes in v13:
- named new functions with _aio_ word
- grouped aio functions under single #ifdef HAVE_AIO_SUPPORT
- moved perf_mmap__aio_push() stub into header
- removed trailed white space
Changes in v12:
- implemented record__aio_get/set_pos(), record__aio_enabled()
- implemented simple --aio option
Changes in v11:
- replacing the both lseek() syscalls in every loop iteration by the only
two syscalls just before and after the loop at record__mmap_read_evlist() and advancing *in-flight* off file pos value at perf_mmap__aio_push()
Changes in v10:
- avoided lseek() setting file pos back in case of record__aio_write() failure
- compacted code selecting between serial and AIO streaming
- optimized call places of record__mmap_read_sync()
Changes in v9:
- enable AIO streaming only when --aio-cblocks option is specified explicitly Changes in v8:
- split AIO completion check into separate record__aio_complete()
Changes in v6:
- handled errno == EAGAIN case from aio_write();
Changes in v5:
- data loss metrics decreased from 25% to 2x in trialed configuration;
- avoided nanosleep() prior calling aio_suspend();
- switched to per cpu multi record__aio_sync() aio
- record_mmap_read_sync() now does global barrier just before
switching trace file or collection stop;
- resolved livelock on perf record -e intel_pt// -- dd if=/dev/zero of=/dev/null count=100000 Changes in v4:
- converted void *bf to struct perf_mmap *md in signatures
- written comment in perf_mmap__push() just before perf_mmap__get();
- written comment in record__mmap_read_sync() on possible restarting
of aio_write() operation and releasing perf_mmap object after all;
- added perf_mmap__put() for the cases of failed aio_write();
Changes in v3:
- written comments about nanosleep(0.5ms) call prior aio_suspend()
to cope with intrusiveness of its implementation in glibc;
- written comments about rationale behind coping profiling data
into mmap->data buffer;

---
 tools/perf/Documentation/perf-record.txt |   5 +
 tools/perf/builtin-record.c              | 218 ++++++++++++++++++++++++++++++-
 tools/perf/perf.h                        |   1 +
 tools/perf/util/evlist.c                 |   6 +-
 tools/perf/util/evlist.h                 |   2 +-
 tools/perf/util/mmap.c                   |  77 ++++++++++-
 tools/perf/util/mmap.h                   |  14 ++
 7 files changed, 314 insertions(+), 9 deletions(-)
tools/perf/Documentation/perf-record.txt +5 -0
index 246dee081efd..7efb4af88a68 100644
@@ -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::
+Enable asynchronous (Posix AIO) trace writing mode.
+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.
 
tools/perf/builtin-record.c +211 -4
index 0980dfe3396b..0c6105860123 100644
@@ -124,6 +124,183 @@ 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 void record__aio_sync(struct perf_mmap *md)
+{
+       struct aiocb *cblock = &md->aio.cblock;
+       struct timespec timeout = { 0, 1000 * 1000  * 1 }; /* 1ms */
+
+       do {
+               if (cblock->aio_fildes == -1 || record__aio_complete(md, cblock))
+                       return;
+
+               while (aio_suspend((const struct aiocb**)&cblock, 1, &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);
+       }
+}
+
+static int nr_cblocks_default = 1;
+
+static int record__aio_parse(const struct option *opt,
+                            const char *str __maybe_unused,
+                            int unset)
+{
+       struct record_opts *opts = (struct record_opts *)opt->value;
+
+       if (unset)
+               opts->nr_cblocks = 0;
+       else
+               opts->nr_cblocks = nr_cblocks_default;
+
+       return 0;
+}
+#else /* HAVE_AIO_SUPPORT */
+static void record__aio_sync(struct perf_mmap *md __maybe_unused)
+{
+}
+
+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 +506,7 @@ 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 "
@@ -520,6 +697,8 @@ 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;
@@ -531,13 +710,29 @@ 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 {
+                               /*
+                                * Call record__aio_sync() to wait till map->data buffer
+                                * becomes available after previous aio write request.
+                                */
+                               record__aio_sync(map);
+                               if (perf_mmap__aio_push(map, rec, record__aio_pushfn, &off) != 0) {
+                                       record__aio_set_pos(trace_fd, off);
+                                       rc = -1;
+                                       goto out;
+                               }
                        }
                }
 
@@ -548,6 +743,9 @@ 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.
@@ -650,6 +848,8 @@ 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);
@@ -1157,6 +1357,8 @@ 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;
 
@@ -1681,6 +1883,11 @@ static struct option __record_options[] = {
                          "signal"),
        OPT_BOOLEAN(0, "dry-run", &dry_run,
                    "Parse options then exit"),
+#ifdef HAVE_AIO_SUPPORT
+       OPT_CALLBACK_NOOPT(0, "aio", &record.opts,
+                    NULL, "Enable asynchronous trace writing mode",
+                    record__aio_parse),
+#endif
        OPT_END()
 };
 

tools/perf/builtin-record.c int cmd_record(int argc, const char **argv) · line 2085

Quoted 6 lines
tools/perf/builtin-record.c +3 -0
@@ -1873,6 +2080,9 @@ int cmd_record(int argc, const char **argv)
                goto out;
        }
 
+       if (verbose > 0)
+               pr_info("nr_cblocks: %d\n", rec->opts.nr_cblocks);
+

nit: I guess we don't really need this pr_info()?

Other than this,

#6 lore raw

Re: [PATCH v14 2/3]: perf record: enable asynchronous trace writing

Hi,

Quoted 3 lines · Song Liu
Quoted 1 line · Alexey Budankov

<SNIP>

Quoted 5 lines

tools/perf/builtin-record.c int cmd_record(int argc, const char **argv) · line 2085

Quoted 3 lines
tools/perf/builtin-record.c +3 -0
+       if (verbose > 0)
+               pr_info("nr_cblocks: %d\n", rec->opts.nr_cblocks);
+

nit: I guess we don't really need this pr_info()?

Diagnostic aid.

Thanks,
Alexey

Quoted 5 lines

Other than this,

#7 lore raw

PATCH v14 3/3 : perf record: extend trace writing to multi AIO

In reply to Alexey Budankov

Multi AIO trace writing allows caching more kernel data into userspace memory postponing trace writing for the sake of overall profiling data thruput increase. It could be seen as kernel data buffer extension into userspace memory.

With aio option value different from 0, default value is 1, tool has capability to cache more and more data into user space along with delegating spill to AIO.

That allows avoiding suspend at record__aio_sync() between calls of record__mmap_read_evlist() and increase profiling data thruput for the cost of userspace memory.

---
Changes in v14:
- fix --aio option handling
Changes in v13:
- preserved --aio option name avoiding complication Changes in v12:
- extended --aio option to --aio-cblocks=<n> Changes in v10:
- added description of aio-cblocks option into perf-record.tx

---
 tools/perf/Documentation/perf-record.txt |  4 +-
 tools/perf/builtin-record.c              | 64 ++++++++++++++++++++++++--------
 tools/perf/util/mmap.c                   | 64 +++++++++++++++++++++-----------
 tools/perf/util/mmap.h                   |  9 +++--
 4 files changed, 99 insertions(+), 42 deletions(-)
tools/perf/Documentation/perf-record.txt +2 -2
index 7efb4af88a68..d232b13ea713 100644
@@ -435,8 +435,8 @@ Specify vmlinux path which has debuginfo.
 --buildid-all::
 Record build-id of all DSOs regardless whether it's actually hit or not.
 
---aio::
-Enable asynchronous (Posix AIO) trace writing mode.
+--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.
 
tools/perf/builtin-record.c +49 -15

125 lines not on this page · raw message

tools/perf/util/mmap.c +43 -21

112 lines not on this page · raw message

tools/perf/util/mmap.h +5 -4
index 9be06f18f17a..cb63d357a248 100644
@@ -32,8 +32,9 @@ struct perf_mmap {
 	char		 event_copy[PERF_SAMPLE_MAX_SIZE] __aligned(8);
 #ifdef HAVE_AIO_SUPPORT
 	struct {
-		void 		 *data;
-		struct aiocb	 cblock;
+		void		 **data;
+		struct aiocb	 *cblocks;
+		struct aiocb	 **aiocb;
 		int		 nr_cblocks;
 	} aio;
 #endif
@@ -106,11 +107,11 @@ 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 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,
+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)
 {
Acked-by on v14 3/3
#8 lore raw

Re: [PATCH v14 3/3]: perf record: extend trace writing to multi AIO

Quoted 174 lines · Alexey Budankov

Multi AIO trace writing allows caching more kernel data into userspace memory postponing trace writing for the sake of overall profiling data thruput increase. It could be seen as kernel data buffer extension into userspace memory.

With aio option value different from 0, default value is 1, tool has capability to cache more and more data into user space along with delegating spill to AIO.

That allows avoiding suspend at record__aio_sync() between calls of record__mmap_read_evlist() and increase profiling data thruput for the cost of userspace memory.

---
Changes in v14:
- fix --aio option handling
Changes in v13:
- preserved --aio option name avoiding complication Changes in v12:
- extended --aio option to --aio-cblocks=<n> Changes in v10:
- added description of aio-cblocks option into perf-record.tx

---
 tools/perf/Documentation/perf-record.txt |  4 +-
 tools/perf/builtin-record.c              | 64 ++++++++++++++++++++++++--------
 tools/perf/util/mmap.c                   | 64 +++++++++++++++++++++-----------
 tools/perf/util/mmap.h                   |  9 +++--
 4 files changed, 99 insertions(+), 42 deletions(-)
tools/perf/Documentation/perf-record.txt +2 -2
index 7efb4af88a68..d232b13ea713 100644
@@ -435,8 +435,8 @@ Specify vmlinux path which has debuginfo.
 --buildid-all::
 Record build-id of all DSOs regardless whether it's actually hit or not.
 
---aio::
-Enable asynchronous (Posix AIO) trace writing mode.
+--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.
 
tools/perf/builtin-record.c +47 -15
index 0c6105860123..c5d9288b8670 100644
@@ -196,16 +196,35 @@ static int record__aio_complete(struct perf_mmap *md, struct aiocb *cblock)
        return rc;
 }
 
-static void record__aio_sync(struct perf_mmap *md)
+static int record__aio_sync(struct perf_mmap *md, bool sync_all)
 {
-       struct aiocb *cblock = &md->aio.cblock;
+       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 {
-               if (cblock->aio_fildes == -1 || record__aio_complete(md, cblock))
-                       return;
+               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**)&cblock, 1, &timeout)) {
+               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");
                }
@@ -252,28 +271,33 @@ static void record__aio_mmap_read_sync(struct record *rec)
                struct perf_mmap *map = &maps[i];
 
                if (map->base)
-                       record__aio_sync(map);
+                       record__aio_sync(map, true);
        }
 }
 
 static int nr_cblocks_default = 1;
 
 static int record__aio_parse(const struct option *opt,
-                            const char *str __maybe_unused,
+                            const char *str,
                             int unset)
 {
        struct record_opts *opts = (struct record_opts *)opt->value;
 
-       if (unset)
+       if (unset) {
                opts->nr_cblocks = 0;
-       else
-               opts->nr_cblocks = nr_cblocks_default;
+       } 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 void record__aio_sync(struct perf_mmap *md __maybe_unused)
+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,
@@ -723,12 +747,13 @@ static int record__mmap_read_evlist(struct record *rec, struct perf_evlist *evli
                                        goto out;
                                }
                        } else {
+                               int idx;
                                /*
                                 * Call record__aio_sync() to wait till map->data buffer
                                 * becomes available after previous aio write request.
                                 */
-                               record__aio_sync(map);
-                               if (perf_mmap__aio_push(map, rec, record__aio_pushfn, &off) != 0) {
+                               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;
@@ -1492,6 +1517,13 @@ 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;
 }
@@ -1884,8 +1916,8 @@ static struct option __record_options[] = {
        OPT_BOOLEAN(0, "dry-run", &dry_run,
                    "Parse options then exit"),
 #ifdef HAVE_AIO_SUPPORT
-       OPT_CALLBACK_NOOPT(0, "aio", &record.opts,
-                    NULL, "Enable asynchronous trace writing mode",
+       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()

tools/perf/builtin-record.c int cmd_record(int argc, const char **argv) · line 2116

Quoted 5 lines
tools/perf/builtin-record.c +2 -0
@@ -2080,6 +2112,8 @@ int cmd_record(int argc, const char **argv)
                goto out;
        }
 
+       if (rec->opts.nr_cblocks > 4)
+               rec->opts.nr_cblocks = 4;

nit: I feel this logic belongs to record__aio_parse(). And we should have

static int nr_cblocks_max = 4;

Other than this

#9 lore raw

Re: [PATCH v14 3/3]: perf record: extend trace writing to multi AIO

Hi,

Quoted 3 lines · Song Liu
Quoted 1 line · Alexey Budankov

<SNIP>

Quoted 26 lines
Quoted 26 lines

static int nr_cblocks_default = 1;

static int record__aio_parse(const struct option *opt,

tools/perf/builtin-record.c +8 -4
-                            const char *str __maybe_unused,
+                            const char *str,
                             int unset)
 {
        struct record_opts *opts = (struct record_opts *)opt->value;
 
-       if (unset)
+       if (unset) {
                opts->nr_cblocks = 0;
-       else
-               opts->nr_cblocks = nr_cblocks_default;
+       } 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 */

<SNIP>

Quoted 29 lines
Quoted 29 lines
tools/perf/builtin-record.c +9 -2
@@ -1492,6 +1517,13 @@ 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;
 }
@@ -1884,8 +1916,8 @@ static struct option __record_options[] = {
        OPT_BOOLEAN(0, "dry-run", &dry_run,
                    "Parse options then exit"),
 #ifdef HAVE_AIO_SUPPORT
-       OPT_CALLBACK_NOOPT(0, "aio", &record.opts,
-                    NULL, "Enable asynchronous trace writing mode",
+       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()
@@ -2080,6 +2112,8 @@ int cmd_record(int argc, const char **argv)
                goto out;
        }
 
Quoted 5 lines

tools/perf/builtin-record.c int cmd_record(int argc, const char **argv) · line 2116

Quoted 2 lines
tools/perf/builtin-record.c +2 -0
+       if (rec->opts.nr_cblocks > 4)
+               rec->opts.nr_cblocks = 4;

nit: I feel this logic belongs to record__aio_parse(). And we should have

static int nr_cblocks_max = 4;

This looks reasonable.

Thanks,
Alexey

Quoted 5 lines

Other than this

Reviewed-by on the v14 series
#10 lore raw
In reply to Alexey Budankov
Quoted 14 lines · Alexey Budankov

Currently in record mode the tool implements trace writing serially. The algorithm loops over mapped per-cpu data buffers and stores ready data chunks into a trace file using write() system call.

At some circumstances the kernel may lack free space in a buffer because the other buffer's half is not yet written to disk due to some other buffer's data writing by the tool at the moment.

Thus serial trace writing implementation may cause the kernel to loose profiling data and that is what observed when profiling highly parallel CPU bound workloads on machines with big number of cores.

Quoted 8 lines

Experiment with profiling matrix multiplication code executing 128 threads on Intel Xeon Phi (KNM) with 272 cores, like below, demonstrates data loss metrics value of 98%:

/usr/bin/time perf record -o /tmp/perf-ser.data -a -N -B -T -R -g \
    --call-graph dwarf,1024 --user-regs=IP,SP,BP --switch-events \
    -e cycles,instructions,ref-cycles,software/period=1,name=cs,config=0x3/Duk -- \
    matrix.gcc

I ran above on 24 cpu server and could not see the gain, but I guess I'd need much bigger server to see that

anyway, the code is now nicely separated, and given the advertised results below I have no objections

thanks,
jirka

Quoted 108 lines

Data loss metrics is the ratio lost_time/elapsed_time where lost_time is the sum of time intervals containing PERF_RECORD_LOST records and elapsed_time is the elapsed application run time under profiling.

Applying asynchronous trace streaming thru Posix AIO API [1] lowers data loss metrics value providing 2x improvement (from 98% to ~1%)

Asynchronous trace streaming is currently limited to glibc linkage. musl libc [5] also provides Posix AIO API implementation, however the patchkit is not tested with it. There may be other libc libraries linked by Perf tool that currently lack Posix AIO API support [2], [3], [4] so NO_AIO define may be used to limit Perf tool binary to serial streaming only.

---
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

 tools/perf/Documentation/perf-record.txt |   5 +
 tools/perf/Makefile.config               |   5 +
 tools/perf/Makefile.perf                 |   7 +-
 tools/perf/builtin-record.c              | 252 ++++++++++++++++++++++++++++++-
 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 +++-
 9 files changed, 439 insertions(+), 11 deletions(-)

---
Changes in v14:
- implement default nr_cblocks_default variable
- fix --aio option handling
Changes in v13:
- named new functions with _aio_ word
- grouped aio functions under single #ifdef HAVE_AIO_SUPPORT
- moved perf_mmap__aio_push() stub into header
- removed trailed white space
Changes in v12:
- applied stub functions design for the whole patch kit
- grouped AIO related data into a struct under struct perf_mmap
- implemented record__aio_get/set_pos(), record__aio_enabled()
- implemented simple --aio option
- extended --aio option to --aio-cblocks=<n>
Changes in v11:
- replacing the both lseek() syscalls in every loop iteration by the only
two syscalls just before and after the loop at record__mmap_read_evlist() and advancing *in-flight* off file pos value at perf_mmap__aio_push()
Changes in v10:
- moved specific code to perf_mmap__aio_mmap(), perf_mmap__aio_munmap();
- adjusted error reporting by using %m
- avoided lseek() setting file pos back in case of record__aio_write() failure
- compacted code selecting between serial and AIO streaming
- optimized call places of record__mmap_read_sync()
- added description of aio-cblocks option into perf-record.txt
Changes in v9:
- enable AIO streaming only when --aio-cblocks option is specified explicitly
- enable AIO based implementation when linking with glibc only
- define NO_AIO to limit Perf binary to serial implementation
Changes in v8:
- run the whole thing thru checkpatch.pl and corrected found issues except lines longer than 80 symbols
- corrected comments alignment and formatting
- moved multi AIO implementation into 3rd patch in the series
- implemented explicit cblocks array allocation
- split AIO completion check into separate record__aio_complete()
- set nr_cblocks default to 1 and max allowed value to 4
Changes in v7:
- implemented handling record.aio setting from perfconfig file
Changes in v6:
- adjusted setting of priorities for cblocks;
- handled errno == EAGAIN case from aio_write() return;
Changes in v5:
- resolved livelock on perf record -e intel_pt// -- dd if=/dev/zero of=/dev/null count=100000
- data loss metrics decreased from 25% to 2x in trialed configuration;
- reshaped layout of data structures;
- implemented --aio option;
- avoided nanosleep() prior calling aio_suspend();
- switched to per-cpu aio multi buffer record__aio_sync();
- record_mmap_read_sync() now does global sync just before
switching trace file or collection stop;
Changes in v4:
- converted mmap()/munmap() to malloc()/free() for mmap->data buffer management
- converted void *bf to struct perf_mmap *md in signatures
- written comment in perf_mmap__push() just before perf_mmap__get();
- written comment in record__mmap_read_sync() on possible restarting
of aio_write() operation and releasing perf_mmap object after all;
- added perf_mmap__put() for the cases of failed aio_write();
Changes in v3:
- written comments about nanosleep(0.5ms) call prior aio_suspend()
to cope with intrusiveness of its implementation in glibc;
- written comments about rationale behind coping profiling data
into mmap->data buffer;
Changes in v2:
- converted zalloc() to calloc() for allocation of mmap_aio array,
- cleared typo and adjusted fallback branch code;

---

[1] http://man7.org/linux/man-pages/man7/aio.7.html
[2] https://android.googlesource.com/platform/bionic/+/master/docs/status.md
[3] https://www.uclibc.org/
[4] https://uclibc-ng.org/
[5] https://www.musl-libc.org/

#11 lore raw

Hi,

Quoted 24 lines · Jiri Olsa
Quoted 14 lines · Alexey Budankov

Currently in record mode the tool implements trace writing serially. The algorithm loops over mapped per-cpu data buffers and stores ready data chunks into a trace file using write() system call.

At some circumstances the kernel may lack free space in a buffer because the other buffer's half is not yet written to disk due to some other buffer's data writing by the tool at the moment.

Thus serial trace writing implementation may cause the kernel to loose profiling data and that is what observed when profiling highly parallel CPU bound workloads on machines with big number of cores.

Quoted 8 lines

Experiment with profiling matrix multiplication code executing 128 threads on Intel Xeon Phi (KNM) with 272 cores, like below, demonstrates data loss metrics value of 98%:

/usr/bin/time perf record -o /tmp/perf-ser.data -a -N -B -T -R -g \
    --call-graph dwarf,1024 --user-regs=IP,SP,BP --switch-events \
    -e cycles,instructions,ref-cycles,software/period=1,name=cs,config=0x3/Duk -- \
    matrix.gcc
Quoted 7 lines

I ran above on 24 cpu server and could not see the gain, but I guess I'd need much bigger server to see that

anyway, the code is now nicely separated, and given the advertised results below I have no objections

Is the plan Jiri mentioned earlier to have it as a stand alone patch kit or upstream the changes into mainline?

Thanks,
Alexey

Quoted 114 lines

thanks,
jirka

Quoted 108 lines

Data loss metrics is the ratio lost_time/elapsed_time where lost_time is the sum of time intervals containing PERF_RECORD_LOST records and elapsed_time is the elapsed application run time under profiling.

Applying asynchronous trace streaming thru Posix AIO API [1] lowers data loss metrics value providing 2x improvement (from 98% to ~1%)

Asynchronous trace streaming is currently limited to glibc linkage. musl libc [5] also provides Posix AIO API implementation, however the patchkit is not tested with it. There may be other libc libraries linked by Perf tool that currently lack Posix AIO API support [2], [3], [4] so NO_AIO define may be used to limit Perf tool binary to serial streaming only.

---
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

 tools/perf/Documentation/perf-record.txt |   5 +
 tools/perf/Makefile.config               |   5 +
 tools/perf/Makefile.perf                 |   7 +-
 tools/perf/builtin-record.c              | 252 ++++++++++++++++++++++++++++++-
 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 +++-
 9 files changed, 439 insertions(+), 11 deletions(-)

---
Changes in v14:
- implement default nr_cblocks_default variable
- fix --aio option handling
Changes in v13:
- named new functions with _aio_ word
- grouped aio functions under single #ifdef HAVE_AIO_SUPPORT
- moved perf_mmap__aio_push() stub into header
- removed trailed white space
Changes in v12:
- applied stub functions design for the whole patch kit
- grouped AIO related data into a struct under struct perf_mmap
- implemented record__aio_get/set_pos(), record__aio_enabled()
- implemented simple --aio option
- extended --aio option to --aio-cblocks=<n>
Changes in v11:
- replacing the both lseek() syscalls in every loop iteration by the only
two syscalls just before and after the loop at record__mmap_read_evlist() and advancing *in-flight* off file pos value at perf_mmap__aio_push()
Changes in v10:
- moved specific code to perf_mmap__aio_mmap(), perf_mmap__aio_munmap();
- adjusted error reporting by using %m
- avoided lseek() setting file pos back in case of record__aio_write() failure
- compacted code selecting between serial and AIO streaming
- optimized call places of record__mmap_read_sync()
- added description of aio-cblocks option into perf-record.txt
Changes in v9:
- enable AIO streaming only when --aio-cblocks option is specified explicitly
- enable AIO based implementation when linking with glibc only
- define NO_AIO to limit Perf binary to serial implementation
Changes in v8:
- run the whole thing thru checkpatch.pl and corrected found issues except lines longer than 80 symbols
- corrected comments alignment and formatting
- moved multi AIO implementation into 3rd patch in the series
- implemented explicit cblocks array allocation
- split AIO completion check into separate record__aio_complete()
- set nr_cblocks default to 1 and max allowed value to 4
Changes in v7:
- implemented handling record.aio setting from perfconfig file
Changes in v6:
- adjusted setting of priorities for cblocks;
- handled errno == EAGAIN case from aio_write() return;
Changes in v5:
- resolved livelock on perf record -e intel_pt// -- dd if=/dev/zero of=/dev/null count=100000
- data loss metrics decreased from 25% to 2x in trialed configuration;
- reshaped layout of data structures;
- implemented --aio option;
- avoided nanosleep() prior calling aio_suspend();
- switched to per-cpu aio multi buffer record__aio_sync();
- record_mmap_read_sync() now does global sync just before
switching trace file or collection stop;
Changes in v4:
- converted mmap()/munmap() to malloc()/free() for mmap->data buffer management
- converted void *bf to struct perf_mmap *md in signatures
- written comment in perf_mmap__push() just before perf_mmap__get();
- written comment in record__mmap_read_sync() on possible restarting
of aio_write() operation and releasing perf_mmap object after all;
- added perf_mmap__put() for the cases of failed aio_write();
Changes in v3:
- written comments about nanosleep(0.5ms) call prior aio_suspend()
to cope with intrusiveness of its implementation in glibc;
- written comments about rationale behind coping profiling data
into mmap->data buffer;
Changes in v2:
- converted zalloc() to calloc() for allocation of mmap_aio array,
- cleared typo and adjusted fallback branch code;

---

[1] http://man7.org/linux/man-pages/man7/aio.7.html
[2] https://android.googlesource.com/platform/bionic/+/master/docs/status.md
[3] https://www.uclibc.org/
[4] https://uclibc-ng.org/
[5] https://www.musl-libc.org/

#12 lore raw
Quoted 30 lines · Alexey Budankov

Hi,

Quoted 27 lines · Jiri Olsa
Quoted 14 lines · Alexey Budankov

Currently in record mode the tool implements trace writing serially. The algorithm loops over mapped per-cpu data buffers and stores ready data chunks into a trace file using write() system call.

At some circumstances the kernel may lack free space in a buffer because the other buffer's half is not yet written to disk due to some other buffer's data writing by the tool at the moment.

Thus serial trace writing implementation may cause the kernel to loose profiling data and that is what observed when profiling highly parallel CPU bound workloads on machines with big number of cores.

Quoted 8 lines

Experiment with profiling matrix multiplication code executing 128 threads on Intel Xeon Phi (KNM) with 272 cores, like below, demonstrates data loss metrics value of 98%:

/usr/bin/time perf record -o /tmp/perf-ser.data -a -N -B -T -R -g \
    --call-graph dwarf,1024 --user-regs=IP,SP,BP --switch-events \
    -e cycles,instructions,ref-cycles,software/period=1,name=cs,config=0x3/Duk -- \
    matrix.gcc

I ran above on 24 cpu server and could not see the gain, but I guess I'd need much bigger server to see that

Quoted 7 lines
Quoted 4 lines

anyway, the code is now nicely separated, and given the advertised results below I have no objections

Is the plan Jiri mentioned earlier to have it as a stand alone patch kit or upstream the changes into mainline?

I haven't heard from Arnaldo yet, but I'd like to have this merged in

jirka

#13 lore raw

Em Thu, Oct 25, 2018 at 10:54:39AM +0200, Jiri Olsa escreveu:

Quoted 34 lines · Alexey Budankov
Quoted 33 lines

Hi,

Quoted 30 lines · Jiri Olsa
Quoted 14 lines · Alexey Budankov

Currently in record mode the tool implements trace writing serially. The algorithm loops over mapped per-cpu data buffers and stores ready data chunks into a trace file using write() system call.

At some circumstances the kernel may lack free space in a buffer because the other buffer's half is not yet written to disk due to some other buffer's data writing by the tool at the moment.

Thus serial trace writing implementation may cause the kernel to loose profiling data and that is what observed when profiling highly parallel CPU bound workloads on machines with big number of cores.

Quoted 8 lines

Experiment with profiling matrix multiplication code executing 128 threads on Intel Xeon Phi (KNM) with 272 cores, like below, demonstrates data loss metrics value of 98%:

/usr/bin/time perf record -o /tmp/perf-ser.data -a -N -B -T -R -g \
    --call-graph dwarf,1024 --user-regs=IP,SP,BP --switch-events \
    -e cycles,instructions,ref-cycles,software/period=1,name=cs,config=0x3/Duk -- \
    matrix.gcc

I ran above on 24 cpu server and could not see the gain, but I guess I'd need much bigger server to see that

anyway, the code is now nicely separated, and given the advertised results below I have no objections

Quoted 6 lines
Quoted 4 lines
Quoted 1 line

Is the plan Jiri mentioned earlier to have it as a stand alone patch kit or upstream the changes into mainline?

I haven't heard from Arnaldo yet, but I'd like to have this merged in

I'll try and go over it today.

- Arnaldo