[PATCH 00/19] perf intel-pt: Add support for efficient time interval filtering

[PATCH 00/19] perf intel-pt: Add support for efficient time interval filtering

[Adrian Hunter]

Hi

Here are some patches to add support for efficient time interval filtering.
First there are 3 patches to add perf time interval to the itrace options
structure.  Then changes to Intel PT to support "fast forwarding" to a
particular timestamp.  The filtering is added in patch "perf intel-pt: Add
support for efficient time interval filtering".  After that there are
patches to time-utils, leading up to adding a new test and adding support
multiple explicit time intervals.

The Intel PT changes make time filtering much faster because decoding is
limited to the minimal time ranges needed to support the time intervals.


Adrian Hunter (19):
      perf auxtrace: Add perf time interval to itrace_synth_ops
      perf script: Set perf time interval in itrace_synth_ops
      perf report: Set perf time interval in itrace_synth_ops
      perf intel-pt: Add lookahead callback
      perf intel-pt: Factor out intel_pt_8b_tsc()
      perf intel-pt: Factor out intel_pt_reposition()
      perf intel-pt: Add reposition parameter to intel_pt_get_data()
      perf intel-pt: Add intel_pt_fast_forward()
      perf intel-pt: Factor out intel_pt_get_buffer()
      perf intel-pt: Add support for lookahead
      perf intel-pt: Add support for efficient time interval filtering
      perf time-utils: Treat time ranges consistently
      perf time-utils: Factor out set_percent_time()
      perf time-utils: Prevent percentage time range overlap
      perf time-utils: Fix --time documentation
      perf time-utils: Simplify perf_time__parse_for_ranges() error paths slightly
      perf time-utils: Make perf_time__parse_for_ranges() more logical
      perf tests: Add a test for time-utils
      perf time-utils: Add support for multiple explicit time intervals

 tools/perf/Documentation/perf-diff.txt             |  14 +-
 tools/perf/Documentation/perf-report.txt           |   9 +-
 tools/perf/Documentation/perf-script.txt           |   9 +-
 tools/perf/builtin-report.c                        |   8 +-
 tools/perf/builtin-script.c                        |   8 +-
 tools/perf/tests/Build                             |   1 +
 tools/perf/tests/builtin-test.c                    |   4 +
 tools/perf/tests/tests.h                           |   1 +
 tools/perf/tests/time-utils-test.c                 | 251 ++++++++++++++++
 tools/perf/util/auxtrace.h                         |  34 +++
 .../perf/util/intel-pt-decoder/intel-pt-decoder.c  | 188 ++++++++++--
 .../perf/util/intel-pt-decoder/intel-pt-decoder.h  |   5 +
 tools/perf/util/intel-pt.c                         | 325 +++++++++++++++++++--
 tools/perf/util/time-utils.c                       | 132 ++++++---
 14 files changed, 894 insertions(+), 95 deletions(-)
 create mode 100644 tools/perf/tests/time-utils-test.c


Regards
Adrian

[PATCH 01/19] perf auxtrace: Add perf time interval to itrace_synth_ops

[Adrian Hunter]

Instruction trace decoders can optimize output based on what time intervals
will be filtered, so pass that information in itrace_synth_ops.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/util/auxtrace.h | 34 ++++++++++++++++++++++++++++++++++
 1 file changed, 34 insertions(+)

diff --git a/tools/perf/util/auxtrace.h b/tools/perf/util/auxtrace.h
index c69bcd9a3091..c80c58eb7f4d 100644
--- a/tools/perf/util/auxtrace.h
+++ b/tools/perf/util/auxtrace.h
@@ -83,6 +83,8 @@ enum itrace_period_type {
  * @period_type: 'instructions' events period type
  * @initial_skip: skip N events at the beginning.
  * @cpu_bitmap: CPUs for which to synthesize events, or NULL for all
+ * @ptime_range: time intervals to trace or NULL
+ * @range_num: number of time intervals to trace
  */
 struct itrace_synth_opts {
 	bool			set;
@@ -107,6 +109,8 @@ struct itrace_synth_opts {
 	enum itrace_period_type	period_type;
 	unsigned long		initial_skip;
 	unsigned long		*cpu_bitmap;
+	struct perf_time_interval *ptime_range;
+	int			range_num;
 };
 
 /**
@@ -599,6 +603,21 @@ static inline void auxtrace__free(struct perf_session *session)
 "				PERIOD[ns|us|ms|i|t]:   specify period to sample stream\n" \
 "				concatenate multiple options. Default is ibxwpe or cewp\n"
 
+static inline
+void itrace_synth_opts__set_time_range(struct itrace_synth_opts *opts,
+				       struct perf_time_interval *ptime_range,
+				       int range_num)
+{
+	opts->ptime_range = ptime_range;
+	opts->range_num = range_num;
+}
+
+static inline
+void itrace_synth_opts__clear_time_range(struct itrace_synth_opts *opts)
+{
+	opts->ptime_range = NULL;
+	opts->range_num = 0;
+}
 
 #else
 
@@ -742,6 +761,21 @@ void auxtrace_mmap_params__set_idx(struct auxtrace_mmap_params *mp,
 
 #define ITRACE_HELP ""
 
+static inline
+void itrace_synth_opts__set_time_range(struct itrace_synth_opts *opts
+				       __maybe_unused,
+				       struct perf_time_interval *ptime_range
+				       __maybe_unused,
+				       int range_num __maybe_unused)
+{
+}
+
+static inline
+void itrace_synth_opts__clear_time_range(struct itrace_synth_opts *opts
+					 __maybe_unused)
+{
+}
+
 #endif
 
 #endif
-- 
2.17.1

[tip:perf/core] perf auxtrace: Add perf time interval to itrace_synth_ops

[tip-bot for Adrian Hunter]

Commit-ID:  33526f362b019f0a17c6b522eb3b07017dba98a7
Gitweb:     https://git.kernel.org/tip/33526f362b019f0a17c6b522eb3b07017dba98a7
Author:     Adrian Hunter <adrian.hunter@intel.com>
AuthorDate: Tue, 4 Jun 2019 15:59:59 +0300
Committer:  Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Mon, 10 Jun 2019 16:20:11 -0300

perf auxtrace: Add perf time interval to itrace_synth_ops

Instruction trace decoders can optimize output based on what time
intervals will be filtered, so pass that information in
itrace_synth_ops.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-2-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/util/auxtrace.h | 34 ++++++++++++++++++++++++++++++++++
 1 file changed, 34 insertions(+)

diff --git a/tools/perf/util/auxtrace.h b/tools/perf/util/auxtrace.h
index c69bcd9a3091..c80c58eb7f4d 100644
--- a/tools/perf/util/auxtrace.h
+++ b/tools/perf/util/auxtrace.h
@@ -83,6 +83,8 @@ enum itrace_period_type {
  * @period_type: 'instructions' events period type
  * @initial_skip: skip N events at the beginning.
  * @cpu_bitmap: CPUs for which to synthesize events, or NULL for all
+ * @ptime_range: time intervals to trace or NULL
+ * @range_num: number of time intervals to trace
  */
 struct itrace_synth_opts {
 	bool			set;
@@ -107,6 +109,8 @@ struct itrace_synth_opts {
 	enum itrace_period_type	period_type;
 	unsigned long		initial_skip;
 	unsigned long		*cpu_bitmap;
+	struct perf_time_interval *ptime_range;
+	int			range_num;
 };
 
 /**
@@ -599,6 +603,21 @@ static inline void auxtrace__free(struct perf_session *session)
 "				PERIOD[ns|us|ms|i|t]:   specify period to sample stream\n" \
 "				concatenate multiple options. Default is ibxwpe or cewp\n"
 
+static inline
+void itrace_synth_opts__set_time_range(struct itrace_synth_opts *opts,
+				       struct perf_time_interval *ptime_range,
+				       int range_num)
+{
+	opts->ptime_range = ptime_range;
+	opts->range_num = range_num;
+}
+
+static inline
+void itrace_synth_opts__clear_time_range(struct itrace_synth_opts *opts)
+{
+	opts->ptime_range = NULL;
+	opts->range_num = 0;
+}
 
 #else
 
@@ -742,6 +761,21 @@ void auxtrace_mmap_params__set_idx(struct auxtrace_mmap_params *mp,
 
 #define ITRACE_HELP ""
 
+static inline
+void itrace_synth_opts__set_time_range(struct itrace_synth_opts *opts
+				       __maybe_unused,
+				       struct perf_time_interval *ptime_range
+				       __maybe_unused,
+				       int range_num __maybe_unused)
+{
+}
+
+static inline
+void itrace_synth_opts__clear_time_range(struct itrace_synth_opts *opts
+					 __maybe_unused)
+{
+}
+
 #endif
 
 #endif

[PATCH 02/19] perf script: Set perf time interval in itrace_synth_ops

[Adrian Hunter]

Instruction trace decoders can optimize output based on what time intervals
will be filtered, so pass that information in itrace_synth_ops.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/builtin-script.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/tools/perf/builtin-script.c b/tools/perf/builtin-script.c
index 80c722ade852..61f00055476a 100644
--- a/tools/perf/builtin-script.c
+++ b/tools/perf/builtin-script.c
@@ -3829,6 +3829,10 @@ int cmd_script(int argc, const char **argv)
 						  &script.range_num);
 		if (err < 0)
 			goto out_delete;
+
+		itrace_synth_opts__set_time_range(&itrace_synth_opts,
+						  script.ptime_range,
+						  script.range_num);
 	}
 
 	err = __cmd_script(&script);
@@ -3836,8 +3840,10 @@ int cmd_script(int argc, const char **argv)
 	flush_scripting();
 
 out_delete:
-	if (script.ptime_range)
+	if (script.ptime_range) {
+		itrace_synth_opts__clear_time_range(&itrace_synth_opts);
 		zfree(&script.ptime_range);
+	}
 
 	perf_evlist__free_stats(session->evlist);
 	perf_session__delete(session);
-- 
2.17.1

[tip:perf/core] perf script: Set perf time interval in itrace_synth_ops

[tip-bot for Adrian Hunter]

Commit-ID:  400ae9818fe64899cea921a89c7078e0df9e41ea
Gitweb:     https://git.kernel.org/tip/400ae9818fe64899cea921a89c7078e0df9e41ea
Author:     Adrian Hunter <adrian.hunter@intel.com>
AuthorDate: Tue, 4 Jun 2019 16:00:00 +0300
Committer:  Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Mon, 10 Jun 2019 16:20:11 -0300

perf script: Set perf time interval in itrace_synth_ops

Instruction trace decoders can optimize output based on what time
intervals will be filtered, so pass that information in
itrace_synth_ops.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-3-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/builtin-script.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/tools/perf/builtin-script.c b/tools/perf/builtin-script.c
index 80c722ade852..61f00055476a 100644
--- a/tools/perf/builtin-script.c
+++ b/tools/perf/builtin-script.c
@@ -3829,6 +3829,10 @@ int cmd_script(int argc, const char **argv)
 						  &script.range_num);
 		if (err < 0)
 			goto out_delete;
+
+		itrace_synth_opts__set_time_range(&itrace_synth_opts,
+						  script.ptime_range,
+						  script.range_num);
 	}
 
 	err = __cmd_script(&script);
@@ -3836,8 +3840,10 @@ int cmd_script(int argc, const char **argv)
 	flush_scripting();
 
 out_delete:
-	if (script.ptime_range)
+	if (script.ptime_range) {
+		itrace_synth_opts__clear_time_range(&itrace_synth_opts);
 		zfree(&script.ptime_range);
+	}
 
 	perf_evlist__free_stats(session->evlist);
 	perf_session__delete(session);

[PATCH 03/19] perf report: Set perf time interval in itrace_synth_ops

[Adrian Hunter]

Instruction trace decoders can optimize output based on what time intervals
will be filtered, so pass that information in itrace_synth_ops.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/builtin-report.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c
index 1ca533f06a4c..91c40808380d 100644
--- a/tools/perf/builtin-report.c
+++ b/tools/perf/builtin-report.c
@@ -1428,6 +1428,10 @@ int cmd_report(int argc, const char **argv)
 						  &report.range_num);
 		if (ret < 0)
 			goto error;
+
+		itrace_synth_opts__set_time_range(&itrace_synth_opts,
+						  report.ptime_range,
+						  report.range_num);
 	}
 
 	if (session->tevent.pevent &&
@@ -1449,8 +1453,10 @@ int cmd_report(int argc, const char **argv)
 		ret = 0;
 
 error:
-	if (report.ptime_range)
+	if (report.ptime_range) {
+		itrace_synth_opts__clear_time_range(&itrace_synth_opts);
 		zfree(&report.ptime_range);
+	}
 	zstd_fini(&(session->zstd_data));
 	perf_session__delete(session);
 	return ret;
-- 
2.17.1

[tip:perf/core] perf report: Set perf time interval in itrace_synth_ops

[tip-bot for Adrian Hunter]

Commit-ID:  4885c90c5e84926cfb083c58d8b6d70d1b1ac7cf
Gitweb:     https://git.kernel.org/tip/4885c90c5e84926cfb083c58d8b6d70d1b1ac7cf
Author:     Adrian Hunter <adrian.hunter@intel.com>
AuthorDate: Tue, 4 Jun 2019 16:00:01 +0300
Committer:  Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Mon, 10 Jun 2019 16:20:12 -0300

perf report: Set perf time interval in itrace_synth_ops

Instruction trace decoders can optimize output based on what time
intervals will be filtered, so pass that information in
itrace_synth_ops.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-4-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/builtin-report.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c
index 1ca533f06a4c..91c40808380d 100644
--- a/tools/perf/builtin-report.c
+++ b/tools/perf/builtin-report.c
@@ -1428,6 +1428,10 @@ repeat:
 						  &report.range_num);
 		if (ret < 0)
 			goto error;
+
+		itrace_synth_opts__set_time_range(&itrace_synth_opts,
+						  report.ptime_range,
+						  report.range_num);
 	}
 
 	if (session->tevent.pevent &&
@@ -1449,8 +1453,10 @@ repeat:
 		ret = 0;
 
 error:
-	if (report.ptime_range)
+	if (report.ptime_range) {
+		itrace_synth_opts__clear_time_range(&itrace_synth_opts);
 		zfree(&report.ptime_range);
+	}
 	zstd_fini(&(session->zstd_data));
 	perf_session__delete(session);
 	return ret;

[PATCH 04/19] perf intel-pt: Add lookahead callback

[Adrian Hunter]

Add a callback function to enable the decoder to lookahead at subsequent
trace buffers. This will be used to implement a "fast forward" facility
which will be needed to support efficient time interval filtering.

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

diff --git a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
index 9eb778f9c911..13123b195083 100644
--- a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
+++ b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
@@ -104,6 +104,7 @@ struct intel_pt_decoder {
 			 uint64_t *insn_cnt_ptr, uint64_t *ip, uint64_t to_ip,
 			 uint64_t max_insn_cnt, void *data);
 	bool (*pgd_ip)(uint64_t ip, void *data);
+	int (*lookahead)(void *data, intel_pt_lookahead_cb_t cb, void *cb_data);
 	void *data;
 	struct intel_pt_state state;
 	const unsigned char *buf;
@@ -233,6 +234,7 @@ struct intel_pt_decoder *intel_pt_decoder_new(struct intel_pt_params *params)
 	decoder->get_trace          = params->get_trace;
 	decoder->walk_insn          = params->walk_insn;
 	decoder->pgd_ip             = params->pgd_ip;
+	decoder->lookahead          = params->lookahead;
 	decoder->data               = params->data;
 	decoder->return_compression = params->return_compression;
 	decoder->branch_enable      = params->branch_enable;
diff --git a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.h b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.h
index 6a61773dc44b..de36254c6ac9 100644
--- a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.h
+++ b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.h
@@ -102,12 +102,15 @@ struct intel_pt_buffer {
 	uint64_t trace_nr;
 };
 
+typedef int (*intel_pt_lookahead_cb_t)(struct intel_pt_buffer *, void *);
+
 struct intel_pt_params {
 	int (*get_trace)(struct intel_pt_buffer *buffer, void *data);
 	int (*walk_insn)(struct intel_pt_insn *intel_pt_insn,
 			 uint64_t *insn_cnt_ptr, uint64_t *ip, uint64_t to_ip,
 			 uint64_t max_insn_cnt, void *data);
 	bool (*pgd_ip)(uint64_t ip, void *data);
+	int (*lookahead)(void *data, intel_pt_lookahead_cb_t cb, void *cb_data);
 	void *data;
 	bool return_compression;
 	bool branch_enable;
-- 
2.17.1

[tip:perf/core] perf intel-pt: Add lookahead callback

[tip-bot for Adrian Hunter]

Commit-ID:  4d678e9039b075f9418600dc87ec5e61cfb57115
Gitweb:     https://git.kernel.org/tip/4d678e9039b075f9418600dc87ec5e61cfb57115
Author:     Adrian Hunter <adrian.hunter@intel.com>
AuthorDate: Tue, 4 Jun 2019 16:00:02 +0300
Committer:  Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Mon, 10 Jun 2019 16:20:12 -0300

perf intel-pt: Add lookahead callback

Add a callback function to enable the decoder to lookahead at subsequent
trace buffers. This will be used to implement a "fast forward" facility
which will be needed to support efficient time interval filtering.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-5-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/util/intel-pt-decoder/intel-pt-decoder.c | 2 ++
 tools/perf/util/intel-pt-decoder/intel-pt-decoder.h | 3 +++
 2 files changed, 5 insertions(+)

diff --git a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
index 9eb778f9c911..13123b195083 100644
--- a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
+++ b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
@@ -104,6 +104,7 @@ struct intel_pt_decoder {
 			 uint64_t *insn_cnt_ptr, uint64_t *ip, uint64_t to_ip,
 			 uint64_t max_insn_cnt, void *data);
 	bool (*pgd_ip)(uint64_t ip, void *data);
+	int (*lookahead)(void *data, intel_pt_lookahead_cb_t cb, void *cb_data);
 	void *data;
 	struct intel_pt_state state;
 	const unsigned char *buf;
@@ -233,6 +234,7 @@ struct intel_pt_decoder *intel_pt_decoder_new(struct intel_pt_params *params)
 	decoder->get_trace          = params->get_trace;
 	decoder->walk_insn          = params->walk_insn;
 	decoder->pgd_ip             = params->pgd_ip;
+	decoder->lookahead          = params->lookahead;
 	decoder->data               = params->data;
 	decoder->return_compression = params->return_compression;
 	decoder->branch_enable      = params->branch_enable;
diff --git a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.h b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.h
index 6a61773dc44b..de36254c6ac9 100644
--- a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.h
+++ b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.h
@@ -102,12 +102,15 @@ struct intel_pt_buffer {
 	uint64_t trace_nr;
 };
 
+typedef int (*intel_pt_lookahead_cb_t)(struct intel_pt_buffer *, void *);
+
 struct intel_pt_params {
 	int (*get_trace)(struct intel_pt_buffer *buffer, void *data);
 	int (*walk_insn)(struct intel_pt_insn *intel_pt_insn,
 			 uint64_t *insn_cnt_ptr, uint64_t *ip, uint64_t to_ip,
 			 uint64_t max_insn_cnt, void *data);
 	bool (*pgd_ip)(uint64_t ip, void *data);
+	int (*lookahead)(void *data, intel_pt_lookahead_cb_t cb, void *cb_data);
 	void *data;
 	bool return_compression;
 	bool branch_enable;

[PATCH 05/19] perf intel-pt: Factor out intel_pt_8b_tsc()

[Adrian Hunter]

Factor out intel_pt_8b_tsc() so it can be reused.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 .../util/intel-pt-decoder/intel-pt-decoder.c  | 26 ++++++++++++-------
 1 file changed, 17 insertions(+), 9 deletions(-)

diff --git a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
index 13123b195083..c06dceb774e9 100644
--- a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
+++ b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
@@ -1369,6 +1369,21 @@ static int intel_pt_mode_tsx(struct intel_pt_decoder *decoder, bool *no_tip)
 	return 0;
 }
 
+static uint64_t intel_pt_8b_tsc(uint64_t timestamp, uint64_t ref_timestamp)
+{
+	timestamp |= (ref_timestamp & (0xffULL << 56));
+
+	if (timestamp < ref_timestamp) {
+		if (ref_timestamp - timestamp > (1ULL << 55))
+			timestamp += (1ULL << 56);
+	} else {
+		if (timestamp - ref_timestamp > (1ULL << 55))
+			timestamp -= (1ULL << 56);
+	}
+
+	return timestamp;
+}
+
 static void intel_pt_calc_tsc_timestamp(struct intel_pt_decoder *decoder)
 {
 	uint64_t timestamp;
@@ -1376,15 +1391,8 @@ static void intel_pt_calc_tsc_timestamp(struct intel_pt_decoder *decoder)
 	decoder->have_tma = false;
 
 	if (decoder->ref_timestamp) {
-		timestamp = decoder->packet.payload |
-			    (decoder->ref_timestamp & (0xffULL << 56));
-		if (timestamp < decoder->ref_timestamp) {
-			if (decoder->ref_timestamp - timestamp > (1ULL << 55))
-				timestamp += (1ULL << 56);
-		} else {
-			if (timestamp - decoder->ref_timestamp > (1ULL << 55))
-				timestamp -= (1ULL << 56);
-		}
+		timestamp = intel_pt_8b_tsc(decoder->packet.payload,
+					    decoder->ref_timestamp);
 		decoder->tsc_timestamp = timestamp;
 		decoder->timestamp = timestamp;
 		decoder->ref_timestamp = 0;
-- 
2.17.1

[tip:perf/core] perf intel-pt: Factor out intel_pt_8b_tsc()

[tip-bot for Adrian Hunter]

Commit-ID:  e72b52a2cfdea5cb0279b2d63a36d78b8c2134de
Gitweb:     https://git.kernel.org/tip/e72b52a2cfdea5cb0279b2d63a36d78b8c2134de
Author:     Adrian Hunter <adrian.hunter@intel.com>
AuthorDate: Tue, 4 Jun 2019 16:00:03 +0300
Committer:  Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Mon, 10 Jun 2019 16:20:12 -0300

perf intel-pt: Factor out intel_pt_8b_tsc()

Factor out intel_pt_8b_tsc() so it can be reused.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-6-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 .../perf/util/intel-pt-decoder/intel-pt-decoder.c  | 26 ++++++++++++++--------
 1 file changed, 17 insertions(+), 9 deletions(-)

diff --git a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
index 13123b195083..c06dceb774e9 100644
--- a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
+++ b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
@@ -1369,6 +1369,21 @@ static int intel_pt_mode_tsx(struct intel_pt_decoder *decoder, bool *no_tip)
 	return 0;
 }
 
+static uint64_t intel_pt_8b_tsc(uint64_t timestamp, uint64_t ref_timestamp)
+{
+	timestamp |= (ref_timestamp & (0xffULL << 56));
+
+	if (timestamp < ref_timestamp) {
+		if (ref_timestamp - timestamp > (1ULL << 55))
+			timestamp += (1ULL << 56);
+	} else {
+		if (timestamp - ref_timestamp > (1ULL << 55))
+			timestamp -= (1ULL << 56);
+	}
+
+	return timestamp;
+}
+
 static void intel_pt_calc_tsc_timestamp(struct intel_pt_decoder *decoder)
 {
 	uint64_t timestamp;
@@ -1376,15 +1391,8 @@ static void intel_pt_calc_tsc_timestamp(struct intel_pt_decoder *decoder)
 	decoder->have_tma = false;
 
 	if (decoder->ref_timestamp) {
-		timestamp = decoder->packet.payload |
-			    (decoder->ref_timestamp & (0xffULL << 56));
-		if (timestamp < decoder->ref_timestamp) {
-			if (decoder->ref_timestamp - timestamp > (1ULL << 55))
-				timestamp += (1ULL << 56);
-		} else {
-			if (timestamp - decoder->ref_timestamp > (1ULL << 55))
-				timestamp -= (1ULL << 56);
-		}
+		timestamp = intel_pt_8b_tsc(decoder->packet.payload,
+					    decoder->ref_timestamp);
 		decoder->tsc_timestamp = timestamp;
 		decoder->timestamp = timestamp;
 		decoder->ref_timestamp = 0;

[PATCH 06/19] perf intel-pt: Factor out intel_pt_reposition()

[Adrian Hunter]

Factor out intel_pt_reposition() so it can be reused.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/util/intel-pt-decoder/intel-pt-decoder.c | 13 +++++++++----
 1 file changed, 9 insertions(+), 4 deletions(-)

diff --git a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
index c06dceb774e9..70bff7bb79f3 100644
--- a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
+++ b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
@@ -494,6 +494,14 @@ static inline void intel_pt_update_sample_time(struct intel_pt_decoder *decoder)
 	decoder->sample_insn_cnt = decoder->timestamp_insn_cnt;
 }
 
+static void intel_pt_reposition(struct intel_pt_decoder *decoder)
+{
+	decoder->ip = 0;
+	decoder->pkt_state = INTEL_PT_STATE_NO_PSB;
+	decoder->timestamp = 0;
+	decoder->have_tma = false;
+}
+
 static int intel_pt_get_data(struct intel_pt_decoder *decoder)
 {
 	struct intel_pt_buffer buffer = { .buf = 0, };
@@ -512,11 +520,8 @@ static int intel_pt_get_data(struct intel_pt_decoder *decoder)
 		return -ENODATA;
 	}
 	if (!buffer.consecutive) {
-		decoder->ip = 0;
-		decoder->pkt_state = INTEL_PT_STATE_NO_PSB;
+		intel_pt_reposition(decoder);
 		decoder->ref_timestamp = buffer.ref_timestamp;
-		decoder->timestamp = 0;
-		decoder->have_tma = false;
 		decoder->state.trace_nr = buffer.trace_nr;
 		intel_pt_log("Reference timestamp 0x%" PRIx64 "\n",
 			     decoder->ref_timestamp);
-- 
2.17.1

[tip:perf/core] perf intel-pt: Factor out intel_pt_reposition()

[tip-bot for Adrian Hunter]

Commit-ID:  6492e5f013d9975d68528150edadead91e97a78a
Gitweb:     https://git.kernel.org/tip/6492e5f013d9975d68528150edadead91e97a78a
Author:     Adrian Hunter <adrian.hunter@intel.com>
AuthorDate: Tue, 4 Jun 2019 16:00:04 +0300
Committer:  Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Mon, 10 Jun 2019 16:20:12 -0300

perf intel-pt: Factor out intel_pt_reposition()

Factor out intel_pt_reposition() so it can be reused.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-7-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/util/intel-pt-decoder/intel-pt-decoder.c | 13 +++++++++----
 1 file changed, 9 insertions(+), 4 deletions(-)

diff --git a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
index c06dceb774e9..70bff7bb79f3 100644
--- a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
+++ b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
@@ -494,6 +494,14 @@ static inline void intel_pt_update_sample_time(struct intel_pt_decoder *decoder)
 	decoder->sample_insn_cnt = decoder->timestamp_insn_cnt;
 }
 
+static void intel_pt_reposition(struct intel_pt_decoder *decoder)
+{
+	decoder->ip = 0;
+	decoder->pkt_state = INTEL_PT_STATE_NO_PSB;
+	decoder->timestamp = 0;
+	decoder->have_tma = false;
+}
+
 static int intel_pt_get_data(struct intel_pt_decoder *decoder)
 {
 	struct intel_pt_buffer buffer = { .buf = 0, };
@@ -512,11 +520,8 @@ static int intel_pt_get_data(struct intel_pt_decoder *decoder)
 		return -ENODATA;
 	}
 	if (!buffer.consecutive) {
-		decoder->ip = 0;
-		decoder->pkt_state = INTEL_PT_STATE_NO_PSB;
+		intel_pt_reposition(decoder);
 		decoder->ref_timestamp = buffer.ref_timestamp;
-		decoder->timestamp = 0;
-		decoder->have_tma = false;
 		decoder->state.trace_nr = buffer.trace_nr;
 		intel_pt_log("Reference timestamp 0x%" PRIx64 "\n",
 			     decoder->ref_timestamp);

[PATCH 07/19] perf intel-pt: Add reposition parameter to intel_pt_get_data()

[Adrian Hunter]

When the decoder gets the next trace buffer, some state is reset if the
buffer is not consecutive to the previous buffer. Add a parameter
'reposition' so that can be done also to support a "fast forward" facility.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 .../util/intel-pt-decoder/intel-pt-decoder.c    | 17 +++++++++--------
 1 file changed, 9 insertions(+), 8 deletions(-)

diff --git a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
index 70bff7bb79f3..dde6a7a97a7a 100644
--- a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
+++ b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
@@ -502,7 +502,7 @@ static void intel_pt_reposition(struct intel_pt_decoder *decoder)
 	decoder->have_tma = false;
 }
 
-static int intel_pt_get_data(struct intel_pt_decoder *decoder)
+static int intel_pt_get_data(struct intel_pt_decoder *decoder, bool reposition)
 {
 	struct intel_pt_buffer buffer = { .buf = 0, };
 	int ret;
@@ -519,7 +519,7 @@ static int intel_pt_get_data(struct intel_pt_decoder *decoder)
 		intel_pt_log("No more data\n");
 		return -ENODATA;
 	}
-	if (!buffer.consecutive) {
+	if (!buffer.consecutive || reposition) {
 		intel_pt_reposition(decoder);
 		decoder->ref_timestamp = buffer.ref_timestamp;
 		decoder->state.trace_nr = buffer.trace_nr;
@@ -531,10 +531,11 @@ static int intel_pt_get_data(struct intel_pt_decoder *decoder)
 	return 0;
 }
 
-static int intel_pt_get_next_data(struct intel_pt_decoder *decoder)
+static int intel_pt_get_next_data(struct intel_pt_decoder *decoder,
+				  bool reposition)
 {
 	if (!decoder->next_buf)
-		return intel_pt_get_data(decoder);
+		return intel_pt_get_data(decoder, reposition);
 
 	decoder->buf = decoder->next_buf;
 	decoder->len = decoder->next_len;
@@ -553,7 +554,7 @@ static int intel_pt_get_split_packet(struct intel_pt_decoder *decoder)
 	len = decoder->len;
 	memcpy(buf, decoder->buf, len);
 
-	ret = intel_pt_get_data(decoder);
+	ret = intel_pt_get_data(decoder, false);
 	if (ret) {
 		decoder->pos += old_len;
 		return ret < 0 ? ret : -EINVAL;
@@ -879,7 +880,7 @@ static int intel_pt_get_next_packet(struct intel_pt_decoder *decoder)
 		decoder->len -= decoder->pkt_step;
 
 		if (!decoder->len) {
-			ret = intel_pt_get_next_data(decoder);
+			ret = intel_pt_get_next_data(decoder, false);
 			if (ret)
 				return ret;
 		}
@@ -2369,7 +2370,7 @@ static int intel_pt_get_split_psb(struct intel_pt_decoder *decoder,
 	decoder->pos += decoder->len;
 	decoder->len = 0;
 
-	ret = intel_pt_get_next_data(decoder);
+	ret = intel_pt_get_next_data(decoder, false);
 	if (ret)
 		return ret;
 
@@ -2395,7 +2396,7 @@ static int intel_pt_scan_for_psb(struct intel_pt_decoder *decoder)
 	intel_pt_log("Scanning for PSB\n");
 	while (1) {
 		if (!decoder->len) {
-			ret = intel_pt_get_next_data(decoder);
+			ret = intel_pt_get_next_data(decoder, false);
 			if (ret)
 				return ret;
 		}
-- 
2.17.1

[tip:perf/core] perf intel-pt: Add reposition parameter to intel_pt_get_data()

[tip-bot for Adrian Hunter]

Commit-ID:  6c1f0b18ac3361837dbe53e794e28096285fb4f0
Gitweb:     https://git.kernel.org/tip/6c1f0b18ac3361837dbe53e794e28096285fb4f0
Author:     Adrian Hunter <adrian.hunter@intel.com>
AuthorDate: Tue, 4 Jun 2019 16:00:05 +0300
Committer:  Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Mon, 10 Jun 2019 16:20:12 -0300

perf intel-pt: Add reposition parameter to intel_pt_get_data()

When the decoder gets the next trace buffer, some state is reset if the
buffer is not consecutive to the previous buffer. Add a parameter
'reposition' so that can be done also to support a "fast forward"
facility.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-8-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/util/intel-pt-decoder/intel-pt-decoder.c | 17 +++++++++--------
 1 file changed, 9 insertions(+), 8 deletions(-)

diff --git a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
index 70bff7bb79f3..dde6a7a97a7a 100644
--- a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
+++ b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
@@ -502,7 +502,7 @@ static void intel_pt_reposition(struct intel_pt_decoder *decoder)
 	decoder->have_tma = false;
 }
 
-static int intel_pt_get_data(struct intel_pt_decoder *decoder)
+static int intel_pt_get_data(struct intel_pt_decoder *decoder, bool reposition)
 {
 	struct intel_pt_buffer buffer = { .buf = 0, };
 	int ret;
@@ -519,7 +519,7 @@ static int intel_pt_get_data(struct intel_pt_decoder *decoder)
 		intel_pt_log("No more data\n");
 		return -ENODATA;
 	}
-	if (!buffer.consecutive) {
+	if (!buffer.consecutive || reposition) {
 		intel_pt_reposition(decoder);
 		decoder->ref_timestamp = buffer.ref_timestamp;
 		decoder->state.trace_nr = buffer.trace_nr;
@@ -531,10 +531,11 @@ static int intel_pt_get_data(struct intel_pt_decoder *decoder)
 	return 0;
 }
 
-static int intel_pt_get_next_data(struct intel_pt_decoder *decoder)
+static int intel_pt_get_next_data(struct intel_pt_decoder *decoder,
+				  bool reposition)
 {
 	if (!decoder->next_buf)
-		return intel_pt_get_data(decoder);
+		return intel_pt_get_data(decoder, reposition);
 
 	decoder->buf = decoder->next_buf;
 	decoder->len = decoder->next_len;
@@ -553,7 +554,7 @@ static int intel_pt_get_split_packet(struct intel_pt_decoder *decoder)
 	len = decoder->len;
 	memcpy(buf, decoder->buf, len);
 
-	ret = intel_pt_get_data(decoder);
+	ret = intel_pt_get_data(decoder, false);
 	if (ret) {
 		decoder->pos += old_len;
 		return ret < 0 ? ret : -EINVAL;
@@ -879,7 +880,7 @@ static int intel_pt_get_next_packet(struct intel_pt_decoder *decoder)
 		decoder->len -= decoder->pkt_step;
 
 		if (!decoder->len) {
-			ret = intel_pt_get_next_data(decoder);
+			ret = intel_pt_get_next_data(decoder, false);
 			if (ret)
 				return ret;
 		}
@@ -2369,7 +2370,7 @@ static int intel_pt_get_split_psb(struct intel_pt_decoder *decoder,
 	decoder->pos += decoder->len;
 	decoder->len = 0;
 
-	ret = intel_pt_get_next_data(decoder);
+	ret = intel_pt_get_next_data(decoder, false);
 	if (ret)
 		return ret;
 
@@ -2395,7 +2396,7 @@ static int intel_pt_scan_for_psb(struct intel_pt_decoder *decoder)
 	intel_pt_log("Scanning for PSB\n");
 	while (1) {
 		if (!decoder->len) {
-			ret = intel_pt_get_next_data(decoder);
+			ret = intel_pt_get_next_data(decoder, false);
 			if (ret)
 				return ret;
 		}

[PATCH 08/19] perf intel-pt: Add intel_pt_fast_forward()

[Adrian Hunter]

Intel PT decoding is done in time order. In order to support efficient time
interval filtering, add a facility to "fast forward" towards a particular
timestamp. That involves finding the right buffer, stepping to that buffer,
and then stepping forward PSBs. Because decoding must begin at a PSB,
"fast forward" stops at the last PSB that has a timestamp before the target
timestamp.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 .../util/intel-pt-decoder/intel-pt-decoder.c  | 130 ++++++++++++++++++
 .../util/intel-pt-decoder/intel-pt-decoder.h  |   2 +
 2 files changed, 132 insertions(+)

diff --git a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
index dde6a7a97a7a..c374a856e73f 100644
--- a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
+++ b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
@@ -126,6 +126,7 @@ struct intel_pt_decoder {
 	uint64_t timestamp;
 	uint64_t tsc_timestamp;
 	uint64_t ref_timestamp;
+	uint64_t buf_timestamp;
 	uint64_t sample_timestamp;
 	uint64_t ret_addr;
 	uint64_t ctc_timestamp;
@@ -519,6 +520,7 @@ static int intel_pt_get_data(struct intel_pt_decoder *decoder, bool reposition)
 		intel_pt_log("No more data\n");
 		return -ENODATA;
 	}
+	decoder->buf_timestamp = buffer.ref_timestamp;
 	if (!buffer.consecutive || reposition) {
 		intel_pt_reposition(decoder);
 		decoder->ref_timestamp = buffer.ref_timestamp;
@@ -2854,3 +2856,131 @@ unsigned char *intel_pt_find_overlap(unsigned char *buf_a, size_t len_a,
 			return buf_b; /* No overlap */
 	}
 }
+
+/**
+ * struct fast_forward_data - data used by intel_pt_ff_cb().
+ * @timestamp: timestamp to fast forward towards
+ * @buf_timestamp: buffer timestamp of last buffer with trace data earlier than
+ *                 the fast forward timestamp.
+ */
+struct fast_forward_data {
+	uint64_t timestamp;
+	uint64_t buf_timestamp;
+};
+
+/**
+ * intel_pt_ff_cb - fast forward lookahead callback.
+ * @buffer: Intel PT trace buffer
+ * @data: opaque pointer to fast forward data (struct fast_forward_data)
+ *
+ * Determine if @buffer trace is past the fast forward timestamp.
+ *
+ * Return: 1 (stop lookahead) if @buffer trace is past the fast forward
+ *         timestamp, and 0 otherwise.
+ */
+static int intel_pt_ff_cb(struct intel_pt_buffer *buffer, void *data)
+{
+	struct fast_forward_data *d = data;
+	unsigned char *buf;
+	uint64_t tsc;
+	size_t rem;
+	size_t len;
+
+	buf = (unsigned char *)buffer->buf;
+	len = buffer->len;
+
+	if (!intel_pt_next_psb(&buf, &len) ||
+	    !intel_pt_next_tsc(buf, len, &tsc, &rem))
+		return 0;
+
+	tsc = intel_pt_8b_tsc(tsc, buffer->ref_timestamp);
+
+	intel_pt_log("Buffer 1st timestamp " x64_fmt " ref timestamp " x64_fmt "\n",
+		     tsc, buffer->ref_timestamp);
+
+	/*
+	 * If the buffer contains a timestamp earlier that the fast forward
+	 * timestamp, then record it, else stop.
+	 */
+	if (tsc < d->timestamp)
+		d->buf_timestamp = buffer->ref_timestamp;
+	else
+		return 1;
+
+	return 0;
+}
+
+/**
+ * intel_pt_fast_forward - reposition decoder forwards.
+ * @decoder: Intel PT decoder
+ * @timestamp: timestamp to fast forward towards
+ *
+ * Reposition decoder at the last PSB with a timestamp earlier than @timestamp.
+ *
+ * Return: 0 on success or negative error code on failure.
+ */
+int intel_pt_fast_forward(struct intel_pt_decoder *decoder, uint64_t timestamp)
+{
+	struct fast_forward_data d = { .timestamp = timestamp };
+	unsigned char *buf;
+	size_t len;
+	int err;
+
+	intel_pt_log("Fast forward towards timestamp " x64_fmt "\n", timestamp);
+
+	/* Find buffer timestamp of buffer to fast forward to */
+	err = decoder->lookahead(decoder->data, intel_pt_ff_cb, &d);
+	if (err < 0)
+		return err;
+
+	/* Walk to buffer with same buffer timestamp */
+	if (d.buf_timestamp) {
+		do {
+			decoder->pos += decoder->len;
+			decoder->len = 0;
+			err = intel_pt_get_next_data(decoder, true);
+			/* -ENOLINK means non-consecutive trace */
+			if (err && err != -ENOLINK)
+				return err;
+		} while (decoder->buf_timestamp != d.buf_timestamp);
+	}
+
+	if (!decoder->buf)
+		return 0;
+
+	buf = (unsigned char *)decoder->buf;
+	len = decoder->len;
+
+	if (!intel_pt_next_psb(&buf, &len))
+		return 0;
+
+	/*
+	 * Walk PSBs while the PSB timestamp is less than the fast forward
+	 * timestamp.
+	 */
+	do {
+		uint64_t tsc;
+		size_t rem;
+
+		if (!intel_pt_next_tsc(buf, len, &tsc, &rem))
+			break;
+		tsc = intel_pt_8b_tsc(tsc, decoder->buf_timestamp);
+		/*
+		 * A TSC packet can slip past MTC packets but, after fast
+		 * forward, decoding starts at the TSC timestamp. That means
+		 * the timestamps may not be exactly the same as the timestamps
+		 * that would have been decoded without fast forward.
+		 */
+		if (tsc < timestamp) {
+			intel_pt_log("Fast forward to next PSB timestamp " x64_fmt "\n", tsc);
+			decoder->pos += decoder->len - len;
+			decoder->buf = buf;
+			decoder->len = len;
+			intel_pt_reposition(decoder);
+		} else {
+			break;
+		}
+	} while (intel_pt_step_psb(&buf, &len));
+
+	return 0;
+}
diff --git a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.h b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.h
index de36254c6ac9..e633fad2fd5d 100644
--- a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.h
+++ b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.h
@@ -130,6 +130,8 @@ void intel_pt_decoder_free(struct intel_pt_decoder *decoder);
 
 const struct intel_pt_state *intel_pt_decode(struct intel_pt_decoder *decoder);
 
+int intel_pt_fast_forward(struct intel_pt_decoder *decoder, uint64_t timestamp);
+
 unsigned char *intel_pt_find_overlap(unsigned char *buf_a, size_t len_a,
 				     unsigned char *buf_b, size_t len_b,
 				     bool have_tsc, bool *consecutive);
-- 
2.17.1

[tip:perf/core] perf intel-pt: Add intel_pt_fast_forward()

[tip-bot for Adrian Hunter]

Commit-ID:  a7fa19f5a255cc8970202d5c54092089a01fc042
Gitweb:     https://git.kernel.org/tip/a7fa19f5a255cc8970202d5c54092089a01fc042
Author:     Adrian Hunter <adrian.hunter@intel.com>
AuthorDate: Tue, 4 Jun 2019 16:00:06 +0300
Committer:  Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Mon, 10 Jun 2019 16:20:12 -0300

perf intel-pt: Add intel_pt_fast_forward()

Intel PT decoding is done in time order. In order to support efficient time
interval filtering, add a facility to "fast forward" towards a particular
timestamp. That involves finding the right buffer, stepping to that buffer,
and then stepping forward PSBs. Because decoding must begin at a PSB,
"fast forward" stops at the last PSB that has a timestamp before the target
timestamp.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-9-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 .../perf/util/intel-pt-decoder/intel-pt-decoder.c  | 130 +++++++++++++++++++++
 .../perf/util/intel-pt-decoder/intel-pt-decoder.h  |   2 +
 2 files changed, 132 insertions(+)

diff --git a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
index dde6a7a97a7a..c374a856e73f 100644
--- a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
+++ b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
@@ -126,6 +126,7 @@ struct intel_pt_decoder {
 	uint64_t timestamp;
 	uint64_t tsc_timestamp;
 	uint64_t ref_timestamp;
+	uint64_t buf_timestamp;
 	uint64_t sample_timestamp;
 	uint64_t ret_addr;
 	uint64_t ctc_timestamp;
@@ -519,6 +520,7 @@ static int intel_pt_get_data(struct intel_pt_decoder *decoder, bool reposition)
 		intel_pt_log("No more data\n");
 		return -ENODATA;
 	}
+	decoder->buf_timestamp = buffer.ref_timestamp;
 	if (!buffer.consecutive || reposition) {
 		intel_pt_reposition(decoder);
 		decoder->ref_timestamp = buffer.ref_timestamp;
@@ -2854,3 +2856,131 @@ unsigned char *intel_pt_find_overlap(unsigned char *buf_a, size_t len_a,
 			return buf_b; /* No overlap */
 	}
 }
+
+/**
+ * struct fast_forward_data - data used by intel_pt_ff_cb().
+ * @timestamp: timestamp to fast forward towards
+ * @buf_timestamp: buffer timestamp of last buffer with trace data earlier than
+ *                 the fast forward timestamp.
+ */
+struct fast_forward_data {
+	uint64_t timestamp;
+	uint64_t buf_timestamp;
+};
+
+/**
+ * intel_pt_ff_cb - fast forward lookahead callback.
+ * @buffer: Intel PT trace buffer
+ * @data: opaque pointer to fast forward data (struct fast_forward_data)
+ *
+ * Determine if @buffer trace is past the fast forward timestamp.
+ *
+ * Return: 1 (stop lookahead) if @buffer trace is past the fast forward
+ *         timestamp, and 0 otherwise.
+ */
+static int intel_pt_ff_cb(struct intel_pt_buffer *buffer, void *data)
+{
+	struct fast_forward_data *d = data;
+	unsigned char *buf;
+	uint64_t tsc;
+	size_t rem;
+	size_t len;
+
+	buf = (unsigned char *)buffer->buf;
+	len = buffer->len;
+
+	if (!intel_pt_next_psb(&buf, &len) ||
+	    !intel_pt_next_tsc(buf, len, &tsc, &rem))
+		return 0;
+
+	tsc = intel_pt_8b_tsc(tsc, buffer->ref_timestamp);
+
+	intel_pt_log("Buffer 1st timestamp " x64_fmt " ref timestamp " x64_fmt "\n",
+		     tsc, buffer->ref_timestamp);
+
+	/*
+	 * If the buffer contains a timestamp earlier that the fast forward
+	 * timestamp, then record it, else stop.
+	 */
+	if (tsc < d->timestamp)
+		d->buf_timestamp = buffer->ref_timestamp;
+	else
+		return 1;
+
+	return 0;
+}
+
+/**
+ * intel_pt_fast_forward - reposition decoder forwards.
+ * @decoder: Intel PT decoder
+ * @timestamp: timestamp to fast forward towards
+ *
+ * Reposition decoder at the last PSB with a timestamp earlier than @timestamp.
+ *
+ * Return: 0 on success or negative error code on failure.
+ */
+int intel_pt_fast_forward(struct intel_pt_decoder *decoder, uint64_t timestamp)
+{
+	struct fast_forward_data d = { .timestamp = timestamp };
+	unsigned char *buf;
+	size_t len;
+	int err;
+
+	intel_pt_log("Fast forward towards timestamp " x64_fmt "\n", timestamp);
+
+	/* Find buffer timestamp of buffer to fast forward to */
+	err = decoder->lookahead(decoder->data, intel_pt_ff_cb, &d);
+	if (err < 0)
+		return err;
+
+	/* Walk to buffer with same buffer timestamp */
+	if (d.buf_timestamp) {
+		do {
+			decoder->pos += decoder->len;
+			decoder->len = 0;
+			err = intel_pt_get_next_data(decoder, true);
+			/* -ENOLINK means non-consecutive trace */
+			if (err && err != -ENOLINK)
+				return err;
+		} while (decoder->buf_timestamp != d.buf_timestamp);
+	}
+
+	if (!decoder->buf)
+		return 0;
+
+	buf = (unsigned char *)decoder->buf;
+	len = decoder->len;
+
+	if (!intel_pt_next_psb(&buf, &len))
+		return 0;
+
+	/*
+	 * Walk PSBs while the PSB timestamp is less than the fast forward
+	 * timestamp.
+	 */
+	do {
+		uint64_t tsc;
+		size_t rem;
+
+		if (!intel_pt_next_tsc(buf, len, &tsc, &rem))
+			break;
+		tsc = intel_pt_8b_tsc(tsc, decoder->buf_timestamp);
+		/*
+		 * A TSC packet can slip past MTC packets but, after fast
+		 * forward, decoding starts at the TSC timestamp. That means
+		 * the timestamps may not be exactly the same as the timestamps
+		 * that would have been decoded without fast forward.
+		 */
+		if (tsc < timestamp) {
+			intel_pt_log("Fast forward to next PSB timestamp " x64_fmt "\n", tsc);
+			decoder->pos += decoder->len - len;
+			decoder->buf = buf;
+			decoder->len = len;
+			intel_pt_reposition(decoder);
+		} else {
+			break;
+		}
+	} while (intel_pt_step_psb(&buf, &len));
+
+	return 0;
+}
diff --git a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.h b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.h
index de36254c6ac9..e633fad2fd5d 100644
--- a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.h
+++ b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.h
@@ -130,6 +130,8 @@ void intel_pt_decoder_free(struct intel_pt_decoder *decoder);
 
 const struct intel_pt_state *intel_pt_decode(struct intel_pt_decoder *decoder);
 
+int intel_pt_fast_forward(struct intel_pt_decoder *decoder, uint64_t timestamp);
+
 unsigned char *intel_pt_find_overlap(unsigned char *buf_a, size_t len_a,
 				     unsigned char *buf_b, size_t len_b,
 				     bool have_tsc, bool *consecutive);

[PATCH 09/19] perf intel-pt: Factor out intel_pt_get_buffer()

[Adrian Hunter]

Factor out intel_pt_get_buffer() so it can be reused.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/util/intel-pt.c | 60 +++++++++++++++++++++++---------------
 1 file changed, 37 insertions(+), 23 deletions(-)

diff --git a/tools/perf/util/intel-pt.c b/tools/perf/util/intel-pt.c
index 3cff8fe2eaa0..4a61c73c9711 100644
--- a/tools/perf/util/intel-pt.c
+++ b/tools/perf/util/intel-pt.c
@@ -239,32 +239,13 @@ static int intel_pt_do_fix_overlap(struct intel_pt *pt, struct auxtrace_buffer *
 	return 0;
 }
 
-/* This function assumes data is processed sequentially only */
-static int intel_pt_get_trace(struct intel_pt_buffer *b, void *data)
+static int intel_pt_get_buffer(struct intel_pt_queue *ptq,
+			       struct auxtrace_buffer *buffer,
+			       struct auxtrace_buffer *old_buffer,
+			       struct intel_pt_buffer *b)
 {
-	struct intel_pt_queue *ptq = data;
-	struct auxtrace_buffer *buffer = ptq->buffer;
-	struct auxtrace_buffer *old_buffer = ptq->old_buffer;
-	struct auxtrace_queue *queue;
 	bool might_overlap;
 
-	if (ptq->stop) {
-		b->len = 0;
-		return 0;
-	}
-
-	queue = &ptq->pt->queues.queue_array[ptq->queue_nr];
-
-	buffer = auxtrace_buffer__next(queue, buffer);
-	if (!buffer) {
-		if (old_buffer)
-			auxtrace_buffer__drop_data(old_buffer);
-		b->len = 0;
-		return 0;
-	}
-
-	ptq->buffer = buffer;
-
 	if (!buffer->data) {
 		int fd = perf_data__fd(ptq->pt->session->data);
 
@@ -294,6 +275,39 @@ static int intel_pt_get_trace(struct intel_pt_buffer *b, void *data)
 		b->consecutive = true;
 	}
 
+	return 0;
+}
+
+/* This function assumes data is processed sequentially only */
+static int intel_pt_get_trace(struct intel_pt_buffer *b, void *data)
+{
+	struct intel_pt_queue *ptq = data;
+	struct auxtrace_buffer *buffer = ptq->buffer;
+	struct auxtrace_buffer *old_buffer = ptq->old_buffer;
+	struct auxtrace_queue *queue;
+	int err;
+
+	if (ptq->stop) {
+		b->len = 0;
+		return 0;
+	}
+
+	queue = &ptq->pt->queues.queue_array[ptq->queue_nr];
+
+	buffer = auxtrace_buffer__next(queue, buffer);
+	if (!buffer) {
+		if (old_buffer)
+			auxtrace_buffer__drop_data(old_buffer);
+		b->len = 0;
+		return 0;
+	}
+
+	ptq->buffer = buffer;
+
+	err = intel_pt_get_buffer(ptq, buffer, old_buffer, b);
+	if (err)
+		return err;
+
 	if (ptq->step_through_buffers)
 		ptq->stop = true;
 
-- 
2.17.1

[tip:perf/core] perf intel-pt: Factor out intel_pt_get_buffer()

[tip-bot for Adrian Hunter]

Commit-ID:  e96f7df8807615b96af59f8f8bc6263a7adc27b7
Gitweb:     https://git.kernel.org/tip/e96f7df8807615b96af59f8f8bc6263a7adc27b7
Author:     Adrian Hunter <adrian.hunter@intel.com>
AuthorDate: Tue, 4 Jun 2019 16:00:07 +0300
Committer:  Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Mon, 10 Jun 2019 16:20:12 -0300

perf intel-pt: Factor out intel_pt_get_buffer()

Factor out intel_pt_get_buffer() so it can be reused.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-10-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/util/intel-pt.c | 60 ++++++++++++++++++++++++++++------------------
 1 file changed, 37 insertions(+), 23 deletions(-)

diff --git a/tools/perf/util/intel-pt.c b/tools/perf/util/intel-pt.c
index 3cff8fe2eaa0..4a61c73c9711 100644
--- a/tools/perf/util/intel-pt.c
+++ b/tools/perf/util/intel-pt.c
@@ -239,32 +239,13 @@ static int intel_pt_do_fix_overlap(struct intel_pt *pt, struct auxtrace_buffer *
 	return 0;
 }
 
-/* This function assumes data is processed sequentially only */
-static int intel_pt_get_trace(struct intel_pt_buffer *b, void *data)
+static int intel_pt_get_buffer(struct intel_pt_queue *ptq,
+			       struct auxtrace_buffer *buffer,
+			       struct auxtrace_buffer *old_buffer,
+			       struct intel_pt_buffer *b)
 {
-	struct intel_pt_queue *ptq = data;
-	struct auxtrace_buffer *buffer = ptq->buffer;
-	struct auxtrace_buffer *old_buffer = ptq->old_buffer;
-	struct auxtrace_queue *queue;
 	bool might_overlap;
 
-	if (ptq->stop) {
-		b->len = 0;
-		return 0;
-	}
-
-	queue = &ptq->pt->queues.queue_array[ptq->queue_nr];
-
-	buffer = auxtrace_buffer__next(queue, buffer);
-	if (!buffer) {
-		if (old_buffer)
-			auxtrace_buffer__drop_data(old_buffer);
-		b->len = 0;
-		return 0;
-	}
-
-	ptq->buffer = buffer;
-
 	if (!buffer->data) {
 		int fd = perf_data__fd(ptq->pt->session->data);
 
@@ -294,6 +275,39 @@ static int intel_pt_get_trace(struct intel_pt_buffer *b, void *data)
 		b->consecutive = true;
 	}
 
+	return 0;
+}
+
+/* This function assumes data is processed sequentially only */
+static int intel_pt_get_trace(struct intel_pt_buffer *b, void *data)
+{
+	struct intel_pt_queue *ptq = data;
+	struct auxtrace_buffer *buffer = ptq->buffer;
+	struct auxtrace_buffer *old_buffer = ptq->old_buffer;
+	struct auxtrace_queue *queue;
+	int err;
+
+	if (ptq->stop) {
+		b->len = 0;
+		return 0;
+	}
+
+	queue = &ptq->pt->queues.queue_array[ptq->queue_nr];
+
+	buffer = auxtrace_buffer__next(queue, buffer);
+	if (!buffer) {
+		if (old_buffer)
+			auxtrace_buffer__drop_data(old_buffer);
+		b->len = 0;
+		return 0;
+	}
+
+	ptq->buffer = buffer;
+
+	err = intel_pt_get_buffer(ptq, buffer, old_buffer, b);
+	if (err)
+		return err;
+
 	if (ptq->step_through_buffers)
 		ptq->stop = true;
 

[PATCH 10/19] perf intel-pt: Add support for lookahead

[Adrian Hunter]

Implement the lookahead callback to let the decoder access subsequent
buffers. intel_pt_lookahead() manages the buffer lifetime and calls the
decoder for each buffer until the decoder returns a non-zero value.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/util/intel-pt.c | 59 +++++++++++++++++++++++++++++++++++++-
 1 file changed, 58 insertions(+), 1 deletion(-)

diff --git a/tools/perf/util/intel-pt.c b/tools/perf/util/intel-pt.c
index 4a61c73c9711..3e3a01318b76 100644
--- a/tools/perf/util/intel-pt.c
+++ b/tools/perf/util/intel-pt.c
@@ -278,7 +278,63 @@ static int intel_pt_get_buffer(struct intel_pt_queue *ptq,
 	return 0;
 }
 
-/* This function assumes data is processed sequentially only */
+/* Do not drop buffers with references - refer intel_pt_get_trace() */
+static void intel_pt_lookahead_drop_buffer(struct intel_pt_queue *ptq,
+					   struct auxtrace_buffer *buffer)
+{
+	if (!buffer || buffer == ptq->buffer || buffer == ptq->old_buffer)
+		return;
+
+	auxtrace_buffer__drop_data(buffer);
+}
+
+/* Must be serialized with respect to intel_pt_get_trace() */
+static int intel_pt_lookahead(void *data, intel_pt_lookahead_cb_t cb,
+			      void *cb_data)
+{
+	struct intel_pt_queue *ptq = data;
+	struct auxtrace_buffer *buffer = ptq->buffer;
+	struct auxtrace_buffer *old_buffer = ptq->old_buffer;
+	struct auxtrace_queue *queue;
+	int err = 0;
+
+	queue = &ptq->pt->queues.queue_array[ptq->queue_nr];
+
+	while (1) {
+		struct intel_pt_buffer b = { .len = 0 };
+
+		buffer = auxtrace_buffer__next(queue, buffer);
+		if (!buffer)
+			break;
+
+		err = intel_pt_get_buffer(ptq, buffer, old_buffer, &b);
+		if (err)
+			break;
+
+		if (b.len) {
+			intel_pt_lookahead_drop_buffer(ptq, old_buffer);
+			old_buffer = buffer;
+		} else {
+			intel_pt_lookahead_drop_buffer(ptq, buffer);
+			continue;
+		}
+
+		err = cb(&b, cb_data);
+		if (err)
+			break;
+	}
+
+	if (buffer != old_buffer)
+		intel_pt_lookahead_drop_buffer(ptq, buffer);
+	intel_pt_lookahead_drop_buffer(ptq, old_buffer);
+
+	return err;
+}
+
+/*
+ * This function assumes data is processed sequentially only.
+ * Must be serialized with respect to intel_pt_lookahead()
+ */
 static int intel_pt_get_trace(struct intel_pt_buffer *b, void *data)
 {
 	struct intel_pt_queue *ptq = data;
@@ -827,6 +883,7 @@ static struct intel_pt_queue *intel_pt_alloc_queue(struct intel_pt *pt,
 
 	params.get_trace = intel_pt_get_trace;
 	params.walk_insn = intel_pt_walk_next_insn;
+	params.lookahead = intel_pt_lookahead;
 	params.data = ptq;
 	params.return_compression = intel_pt_return_compression(pt);
 	params.branch_enable = intel_pt_branch_enable(pt);
-- 
2.17.1

[tip:perf/core] perf intel-pt: Add support for lookahead

[tip-bot for Adrian Hunter]

Commit-ID:  da9000ae35027fb7305b8cad0b37df71937ad578
Gitweb:     https://git.kernel.org/tip/da9000ae35027fb7305b8cad0b37df71937ad578
Author:     Adrian Hunter <adrian.hunter@intel.com>
AuthorDate: Tue, 4 Jun 2019 16:00:08 +0300
Committer:  Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Mon, 10 Jun 2019 16:20:12 -0300

perf intel-pt: Add support for lookahead

Implement the lookahead callback to let the decoder access subsequent
buffers. intel_pt_lookahead() manages the buffer lifetime and calls the
decoder for each buffer until the decoder returns a non-zero value.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-11-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/util/intel-pt.c | 59 +++++++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 58 insertions(+), 1 deletion(-)

diff --git a/tools/perf/util/intel-pt.c b/tools/perf/util/intel-pt.c
index 4a61c73c9711..3e3a01318b76 100644
--- a/tools/perf/util/intel-pt.c
+++ b/tools/perf/util/intel-pt.c
@@ -278,7 +278,63 @@ static int intel_pt_get_buffer(struct intel_pt_queue *ptq,
 	return 0;
 }
 
-/* This function assumes data is processed sequentially only */
+/* Do not drop buffers with references - refer intel_pt_get_trace() */
+static void intel_pt_lookahead_drop_buffer(struct intel_pt_queue *ptq,
+					   struct auxtrace_buffer *buffer)
+{
+	if (!buffer || buffer == ptq->buffer || buffer == ptq->old_buffer)
+		return;
+
+	auxtrace_buffer__drop_data(buffer);
+}
+
+/* Must be serialized with respect to intel_pt_get_trace() */
+static int intel_pt_lookahead(void *data, intel_pt_lookahead_cb_t cb,
+			      void *cb_data)
+{
+	struct intel_pt_queue *ptq = data;
+	struct auxtrace_buffer *buffer = ptq->buffer;
+	struct auxtrace_buffer *old_buffer = ptq->old_buffer;
+	struct auxtrace_queue *queue;
+	int err = 0;
+
+	queue = &ptq->pt->queues.queue_array[ptq->queue_nr];
+
+	while (1) {
+		struct intel_pt_buffer b = { .len = 0 };
+
+		buffer = auxtrace_buffer__next(queue, buffer);
+		if (!buffer)
+			break;
+
+		err = intel_pt_get_buffer(ptq, buffer, old_buffer, &b);
+		if (err)
+			break;
+
+		if (b.len) {
+			intel_pt_lookahead_drop_buffer(ptq, old_buffer);
+			old_buffer = buffer;
+		} else {
+			intel_pt_lookahead_drop_buffer(ptq, buffer);
+			continue;
+		}
+
+		err = cb(&b, cb_data);
+		if (err)
+			break;
+	}
+
+	if (buffer != old_buffer)
+		intel_pt_lookahead_drop_buffer(ptq, buffer);
+	intel_pt_lookahead_drop_buffer(ptq, old_buffer);
+
+	return err;
+}
+
+/*
+ * This function assumes data is processed sequentially only.
+ * Must be serialized with respect to intel_pt_lookahead()
+ */
 static int intel_pt_get_trace(struct intel_pt_buffer *b, void *data)
 {
 	struct intel_pt_queue *ptq = data;
@@ -827,6 +883,7 @@ static struct intel_pt_queue *intel_pt_alloc_queue(struct intel_pt *pt,
 
 	params.get_trace = intel_pt_get_trace;
 	params.walk_insn = intel_pt_walk_next_insn;
+	params.lookahead = intel_pt_lookahead;
 	params.data = ptq;
 	params.return_compression = intel_pt_return_compression(pt);
 	params.branch_enable = intel_pt_branch_enable(pt);

[PATCH 11/19] perf intel-pt: Add support for efficient time interval filtering

[Adrian Hunter]

Set up time ranges for efficient time interval filtering using the new
"fast forward" facility.

Because decoding is done in time order, intel_pt_time_filter() needs to
look only at the next start or end timestamp - refer intel_pt_next_time().

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/util/intel-pt.c | 208 +++++++++++++++++++++++++++++++++++++
 1 file changed, 208 insertions(+)

diff --git a/tools/perf/util/intel-pt.c b/tools/perf/util/intel-pt.c
index 3e3a01318b76..43ddc78a066e 100644
--- a/tools/perf/util/intel-pt.c
+++ b/tools/perf/util/intel-pt.c
@@ -42,6 +42,7 @@
 #include "tsc.h"
 #include "intel-pt.h"
 #include "config.h"
+#include "time-utils.h"
 
 #include "intel-pt-decoder/intel-pt-log.h"
 #include "intel-pt-decoder/intel-pt-decoder.h"
@@ -50,6 +51,11 @@
 
 #define MAX_TIMESTAMP (~0ULL)
 
+struct range {
+	u64 start;
+	u64 end;
+};
+
 struct intel_pt {
 	struct auxtrace auxtrace;
 	struct auxtrace_queues queues;
@@ -118,6 +124,9 @@ struct intel_pt {
 
 	char *filter;
 	struct addr_filters filts;
+
+	struct range *time_ranges;
+	unsigned int range_cnt;
 };
 
 enum switch_state {
@@ -154,6 +163,9 @@ struct intel_pt_queue {
 	bool have_sample;
 	u64 time;
 	u64 timestamp;
+	u64 sel_timestamp;
+	bool sel_start;
+	unsigned int sel_idx;
 	u32 flags;
 	u16 insn_len;
 	u64 last_insn_cnt;
@@ -1007,6 +1019,23 @@ static void intel_pt_sample_flags(struct intel_pt_queue *ptq)
 		ptq->flags |= PERF_IP_FLAG_TRACE_END;
 }
 
+static void intel_pt_setup_time_range(struct intel_pt *pt,
+				      struct intel_pt_queue *ptq)
+{
+	if (!pt->range_cnt)
+		return;
+
+	ptq->sel_timestamp = pt->time_ranges[0].start;
+	ptq->sel_idx = 0;
+
+	if (ptq->sel_timestamp) {
+		ptq->sel_start = true;
+	} else {
+		ptq->sel_timestamp = pt->time_ranges[0].end;
+		ptq->sel_start = false;
+	}
+}
+
 static int intel_pt_setup_queue(struct intel_pt *pt,
 				struct auxtrace_queue *queue,
 				unsigned int queue_nr)
@@ -1031,6 +1060,8 @@ static int intel_pt_setup_queue(struct intel_pt *pt,
 			ptq->step_through_buffers = true;
 
 		ptq->sync_switch = pt->sync_switch;
+
+		intel_pt_setup_time_range(pt, ptq);
 	}
 
 	if (!ptq->on_heap &&
@@ -1045,6 +1076,14 @@ static int intel_pt_setup_queue(struct intel_pt *pt,
 		intel_pt_log("queue %u getting timestamp\n", queue_nr);
 		intel_pt_log("queue %u decoding cpu %d pid %d tid %d\n",
 			     queue_nr, ptq->cpu, ptq->pid, ptq->tid);
+
+		if (ptq->sel_start && ptq->sel_timestamp) {
+			ret = intel_pt_fast_forward(ptq->decoder,
+						    ptq->sel_timestamp);
+			if (ret)
+				return ret;
+		}
+
 		while (1) {
 			state = intel_pt_decode(ptq->decoder);
 			if (state->err) {
@@ -1064,6 +1103,9 @@ static int intel_pt_setup_queue(struct intel_pt *pt,
 			     queue_nr, ptq->timestamp);
 		ptq->state = state;
 		ptq->have_sample = true;
+		if (ptq->sel_start && ptq->sel_timestamp &&
+		    ptq->timestamp < ptq->sel_timestamp)
+			ptq->have_sample = false;
 		intel_pt_sample_flags(ptq);
 		ret = auxtrace_heap__add(&pt->heap, queue_nr, ptq->timestamp);
 		if (ret)
@@ -1750,10 +1792,83 @@ static void intel_pt_enable_sync_switch(struct intel_pt *pt)
 	}
 }
 
+/*
+ * To filter against time ranges, it is only necessary to look at the next start
+ * or end time.
+ */
+static bool intel_pt_next_time(struct intel_pt_queue *ptq)
+{
+	struct intel_pt *pt = ptq->pt;
+
+	if (ptq->sel_start) {
+		/* Next time is an end time */
+		ptq->sel_start = false;
+		ptq->sel_timestamp = pt->time_ranges[ptq->sel_idx].end;
+		return true;
+	} else if (ptq->sel_idx + 1 < pt->range_cnt) {
+		/* Next time is a start time */
+		ptq->sel_start = true;
+		ptq->sel_idx += 1;
+		ptq->sel_timestamp = pt->time_ranges[ptq->sel_idx].start;
+		return true;
+	}
+
+	/* No next time */
+	return false;
+}
+
+static int intel_pt_time_filter(struct intel_pt_queue *ptq, u64 *ff_timestamp)
+{
+	int err;
+
+	while (1) {
+		if (ptq->sel_start) {
+			if (ptq->timestamp >= ptq->sel_timestamp) {
+				/* After start time, so consider next time */
+				intel_pt_next_time(ptq);
+				if (!ptq->sel_timestamp) {
+					/* No end time */
+					return 0;
+				}
+				/* Check against end time */
+				continue;
+			}
+			/* Before start time, so fast forward */
+			ptq->have_sample = false;
+			if (ptq->sel_timestamp > *ff_timestamp) {
+				if (ptq->sync_switch) {
+					intel_pt_next_tid(ptq->pt, ptq);
+					ptq->switch_state = INTEL_PT_SS_UNKNOWN;
+				}
+				*ff_timestamp = ptq->sel_timestamp;
+				err = intel_pt_fast_forward(ptq->decoder,
+							    ptq->sel_timestamp);
+				if (err)
+					return err;
+			}
+			return 0;
+		} else if (ptq->timestamp > ptq->sel_timestamp) {
+			/* After end time, so consider next time */
+			if (!intel_pt_next_time(ptq)) {
+				/* No next time range, so stop decoding */
+				ptq->have_sample = false;
+				ptq->switch_state = INTEL_PT_SS_NOT_TRACING;
+				return 1;
+			}
+			/* Check against next start time */
+			continue;
+		} else {
+			/* Before end time */
+			return 0;
+		}
+	}
+}
+
 static int intel_pt_run_decoder(struct intel_pt_queue *ptq, u64 *timestamp)
 {
 	const struct intel_pt_state *state = ptq->state;
 	struct intel_pt *pt = ptq->pt;
+	u64 ff_timestamp = 0;
 	int err;
 
 	if (!pt->kernel_start) {
@@ -1818,6 +1933,12 @@ static int intel_pt_run_decoder(struct intel_pt_queue *ptq, u64 *timestamp)
 			ptq->timestamp = state->timestamp;
 		}
 
+		if (ptq->sel_timestamp) {
+			err = intel_pt_time_filter(ptq, &ff_timestamp);
+			if (err)
+				return err;
+		}
+
 		if (!pt->timeless_decoding && ptq->timestamp >= *timestamp) {
 			*timestamp = ptq->timestamp;
 			return 0;
@@ -2223,6 +2344,7 @@ static void intel_pt_free(struct perf_session *session)
 	thread__put(pt->unknown_thread);
 	addr_filters__exit(&pt->filts);
 	zfree(&pt->filter);
+	zfree(&pt->time_ranges);
 	free(pt);
 }
 
@@ -2520,6 +2642,85 @@ static int intel_pt_perf_config(const char *var, const char *value, void *data)
 	return 0;
 }
 
+/* Find least TSC which converts to ns or later */
+static u64 intel_pt_tsc_start(u64 ns, struct intel_pt *pt)
+{
+	u64 tsc, tm;
+
+	tsc = perf_time_to_tsc(ns, &pt->tc);
+
+	while (1) {
+		tm = tsc_to_perf_time(tsc, &pt->tc);
+		if (tm < ns)
+			break;
+		tsc -= 1;
+	}
+
+	while (tm < ns)
+		tm = tsc_to_perf_time(++tsc, &pt->tc);
+
+	return tsc;
+}
+
+/* Find greatest TSC which converts to ns or earlier */
+static u64 intel_pt_tsc_end(u64 ns, struct intel_pt *pt)
+{
+	u64 tsc, tm;
+
+	tsc = perf_time_to_tsc(ns, &pt->tc);
+
+	while (1) {
+		tm = tsc_to_perf_time(tsc, &pt->tc);
+		if (tm > ns)
+			break;
+		tsc += 1;
+	}
+
+	while (tm > ns)
+		tm = tsc_to_perf_time(--tsc, &pt->tc);
+
+	return tsc;
+}
+
+static int intel_pt_setup_time_ranges(struct intel_pt *pt,
+				      struct itrace_synth_opts *opts)
+{
+	struct perf_time_interval *p = opts->ptime_range;
+	int n = opts->range_num;
+	int i;
+
+	if (!n || !p || pt->timeless_decoding)
+		return 0;
+
+	pt->time_ranges = calloc(n, sizeof(struct range));
+	if (!pt->time_ranges)
+		return -ENOMEM;
+
+	pt->range_cnt = n;
+
+	intel_pt_log("%s: %u range(s)\n", __func__, n);
+
+	for (i = 0; i < n; i++) {
+		struct range *r = &pt->time_ranges[i];
+		u64 ts = p[i].start;
+		u64 te = p[i].end;
+
+		/*
+		 * Take care to ensure the TSC range matches the perf-time range
+		 * when converted back to perf-time.
+		 */
+		r->start = ts ? intel_pt_tsc_start(ts, pt) : 0;
+		r->end   = te ? intel_pt_tsc_end(te, pt) : 0;
+
+		intel_pt_log("range %d: perf time interval: %"PRIu64" to %"PRIu64"\n",
+			     i, ts, te);
+		intel_pt_log("range %d: TSC time interval: %#"PRIx64" to %#"PRIx64"\n",
+			     i, r->start, r->end);
+	}
+
+	return 0;
+}
+
 static const char * const intel_pt_info_fmts[] = {
 	[INTEL_PT_PMU_TYPE]		= "  PMU Type            %"PRId64"\n",
 	[INTEL_PT_TIME_SHIFT]		= "  Time Shift          %"PRIu64"\n",
@@ -2752,6 +2953,12 @@ int intel_pt_process_auxtrace_info(union perf_event *event,
 		pt->cbr2khz = tsc_freq / pt->max_non_turbo_ratio / 1000;
 	}
 
+	if (session->itrace_synth_opts) {
+		err = intel_pt_setup_time_ranges(pt, session->itrace_synth_opts);
+		if (err)
+			goto err_delete_thread;
+	}
+
 	if (pt->synth_opts.calls)
 		pt->branches_filter |= PERF_IP_FLAG_CALL | PERF_IP_FLAG_ASYNC |
 				       PERF_IP_FLAG_TRACE_END;
@@ -2792,6 +2999,7 @@ int intel_pt_process_auxtrace_info(union perf_event *event,
 err_free:
 	addr_filters__exit(&pt->filts);
 	zfree(&pt->filter);
+	zfree(&pt->time_ranges);
 	free(pt);
 	return err;
 }
-- 
2.17.1

[tip:perf/core] perf intel-pt: Add support for efficient time interval filtering

[tip-bot for Adrian Hunter]

Commit-ID:  2c47db90ed71af9c12d5600dbcef864761d76b3d
Gitweb:     https://git.kernel.org/tip/2c47db90ed71af9c12d5600dbcef864761d76b3d
Author:     Adrian Hunter <adrian.hunter@intel.com>
AuthorDate: Tue, 4 Jun 2019 16:00:09 +0300
Committer:  Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Mon, 10 Jun 2019 16:20:12 -0300

perf intel-pt: Add support for efficient time interval filtering

Set up time ranges for efficient time interval filtering using the new
"fast forward" facility.

Because decoding is done in time order, intel_pt_time_filter() needs to
look only at the next start or end timestamp - refer intel_pt_next_time().

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-12-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/util/intel-pt.c | 208 +++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 208 insertions(+)

diff --git a/tools/perf/util/intel-pt.c b/tools/perf/util/intel-pt.c
index 3e3a01318b76..43ddc78a066e 100644
--- a/tools/perf/util/intel-pt.c
+++ b/tools/perf/util/intel-pt.c
@@ -42,6 +42,7 @@
 #include "tsc.h"
 #include "intel-pt.h"
 #include "config.h"
+#include "time-utils.h"
 
 #include "intel-pt-decoder/intel-pt-log.h"
 #include "intel-pt-decoder/intel-pt-decoder.h"
@@ -50,6 +51,11 @@
 
 #define MAX_TIMESTAMP (~0ULL)
 
+struct range {
+	u64 start;
+	u64 end;
+};
+
 struct intel_pt {
 	struct auxtrace auxtrace;
 	struct auxtrace_queues queues;
@@ -118,6 +124,9 @@ struct intel_pt {
 
 	char *filter;
 	struct addr_filters filts;
+
+	struct range *time_ranges;
+	unsigned int range_cnt;
 };
 
 enum switch_state {
@@ -154,6 +163,9 @@ struct intel_pt_queue {
 	bool have_sample;
 	u64 time;
 	u64 timestamp;
+	u64 sel_timestamp;
+	bool sel_start;
+	unsigned int sel_idx;
 	u32 flags;
 	u16 insn_len;
 	u64 last_insn_cnt;
@@ -1007,6 +1019,23 @@ static void intel_pt_sample_flags(struct intel_pt_queue *ptq)
 		ptq->flags |= PERF_IP_FLAG_TRACE_END;
 }
 
+static void intel_pt_setup_time_range(struct intel_pt *pt,
+				      struct intel_pt_queue *ptq)
+{
+	if (!pt->range_cnt)
+		return;
+
+	ptq->sel_timestamp = pt->time_ranges[0].start;
+	ptq->sel_idx = 0;
+
+	if (ptq->sel_timestamp) {
+		ptq->sel_start = true;
+	} else {
+		ptq->sel_timestamp = pt->time_ranges[0].end;
+		ptq->sel_start = false;
+	}
+}
+
 static int intel_pt_setup_queue(struct intel_pt *pt,
 				struct auxtrace_queue *queue,
 				unsigned int queue_nr)
@@ -1031,6 +1060,8 @@ static int intel_pt_setup_queue(struct intel_pt *pt,
 			ptq->step_through_buffers = true;
 
 		ptq->sync_switch = pt->sync_switch;
+
+		intel_pt_setup_time_range(pt, ptq);
 	}
 
 	if (!ptq->on_heap &&
@@ -1045,6 +1076,14 @@ static int intel_pt_setup_queue(struct intel_pt *pt,
 		intel_pt_log("queue %u getting timestamp\n", queue_nr);
 		intel_pt_log("queue %u decoding cpu %d pid %d tid %d\n",
 			     queue_nr, ptq->cpu, ptq->pid, ptq->tid);
+
+		if (ptq->sel_start && ptq->sel_timestamp) {
+			ret = intel_pt_fast_forward(ptq->decoder,
+						    ptq->sel_timestamp);
+			if (ret)
+				return ret;
+		}
+
 		while (1) {
 			state = intel_pt_decode(ptq->decoder);
 			if (state->err) {
@@ -1064,6 +1103,9 @@ static int intel_pt_setup_queue(struct intel_pt *pt,
 			     queue_nr, ptq->timestamp);
 		ptq->state = state;
 		ptq->have_sample = true;
+		if (ptq->sel_start && ptq->sel_timestamp &&
+		    ptq->timestamp < ptq->sel_timestamp)
+			ptq->have_sample = false;
 		intel_pt_sample_flags(ptq);
 		ret = auxtrace_heap__add(&pt->heap, queue_nr, ptq->timestamp);
 		if (ret)
@@ -1750,10 +1792,83 @@ static void intel_pt_enable_sync_switch(struct intel_pt *pt)
 	}
 }
 
+/*
+ * To filter against time ranges, it is only necessary to look at the next start
+ * or end time.
+ */
+static bool intel_pt_next_time(struct intel_pt_queue *ptq)
+{
+	struct intel_pt *pt = ptq->pt;
+
+	if (ptq->sel_start) {
+		/* Next time is an end time */
+		ptq->sel_start = false;
+		ptq->sel_timestamp = pt->time_ranges[ptq->sel_idx].end;
+		return true;
+	} else if (ptq->sel_idx + 1 < pt->range_cnt) {
+		/* Next time is a start time */
+		ptq->sel_start = true;
+		ptq->sel_idx += 1;
+		ptq->sel_timestamp = pt->time_ranges[ptq->sel_idx].start;
+		return true;
+	}
+
+	/* No next time */
+	return false;
+}
+
+static int intel_pt_time_filter(struct intel_pt_queue *ptq, u64 *ff_timestamp)
+{
+	int err;
+
+	while (1) {
+		if (ptq->sel_start) {
+			if (ptq->timestamp >= ptq->sel_timestamp) {
+				/* After start time, so consider next time */
+				intel_pt_next_time(ptq);
+				if (!ptq->sel_timestamp) {
+					/* No end time */
+					return 0;
+				}
+				/* Check against end time */
+				continue;
+			}
+			/* Before start time, so fast forward */
+			ptq->have_sample = false;
+			if (ptq->sel_timestamp > *ff_timestamp) {
+				if (ptq->sync_switch) {
+					intel_pt_next_tid(ptq->pt, ptq);
+					ptq->switch_state = INTEL_PT_SS_UNKNOWN;
+				}
+				*ff_timestamp = ptq->sel_timestamp;
+				err = intel_pt_fast_forward(ptq->decoder,
+							    ptq->sel_timestamp);
+				if (err)
+					return err;
+			}
+			return 0;
+		} else if (ptq->timestamp > ptq->sel_timestamp) {
+			/* After end time, so consider next time */
+			if (!intel_pt_next_time(ptq)) {
+				/* No next time range, so stop decoding */
+				ptq->have_sample = false;
+				ptq->switch_state = INTEL_PT_SS_NOT_TRACING;
+				return 1;
+			}
+			/* Check against next start time */
+			continue;
+		} else {
+			/* Before end time */
+			return 0;
+		}
+	}
+}
+
 static int intel_pt_run_decoder(struct intel_pt_queue *ptq, u64 *timestamp)
 {
 	const struct intel_pt_state *state = ptq->state;
 	struct intel_pt *pt = ptq->pt;
+	u64 ff_timestamp = 0;
 	int err;
 
 	if (!pt->kernel_start) {
@@ -1818,6 +1933,12 @@ static int intel_pt_run_decoder(struct intel_pt_queue *ptq, u64 *timestamp)
 			ptq->timestamp = state->timestamp;
 		}
 
+		if (ptq->sel_timestamp) {
+			err = intel_pt_time_filter(ptq, &ff_timestamp);
+			if (err)
+				return err;
+		}
+
 		if (!pt->timeless_decoding && ptq->timestamp >= *timestamp) {
 			*timestamp = ptq->timestamp;
 			return 0;
@@ -2223,6 +2344,7 @@ static void intel_pt_free(struct perf_session *session)
 	thread__put(pt->unknown_thread);
 	addr_filters__exit(&pt->filts);
 	zfree(&pt->filter);
+	zfree(&pt->time_ranges);
 	free(pt);
 }
 
@@ -2520,6 +2642,85 @@ static int intel_pt_perf_config(const char *var, const char *value, void *data)
 	return 0;
 }
 
+/* Find least TSC which converts to ns or later */
+static u64 intel_pt_tsc_start(u64 ns, struct intel_pt *pt)
+{
+	u64 tsc, tm;
+
+	tsc = perf_time_to_tsc(ns, &pt->tc);
+
+	while (1) {
+		tm = tsc_to_perf_time(tsc, &pt->tc);
+		if (tm < ns)
+			break;
+		tsc -= 1;
+	}
+
+	while (tm < ns)
+		tm = tsc_to_perf_time(++tsc, &pt->tc);
+
+	return tsc;
+}
+
+/* Find greatest TSC which converts to ns or earlier */
+static u64 intel_pt_tsc_end(u64 ns, struct intel_pt *pt)
+{
+	u64 tsc, tm;
+
+	tsc = perf_time_to_tsc(ns, &pt->tc);
+
+	while (1) {
+		tm = tsc_to_perf_time(tsc, &pt->tc);
+		if (tm > ns)
+			break;
+		tsc += 1;
+	}
+
+	while (tm > ns)
+		tm = tsc_to_perf_time(--tsc, &pt->tc);
+
+	return tsc;
+}
+
+static int intel_pt_setup_time_ranges(struct intel_pt *pt,
+				      struct itrace_synth_opts *opts)
+{
+	struct perf_time_interval *p = opts->ptime_range;
+	int n = opts->range_num;
+	int i;
+
+	if (!n || !p || pt->timeless_decoding)
+		return 0;
+
+	pt->time_ranges = calloc(n, sizeof(struct range));
+	if (!pt->time_ranges)
+		return -ENOMEM;
+
+	pt->range_cnt = n;
+
+	intel_pt_log("%s: %u range(s)\n", __func__, n);
+
+	for (i = 0; i < n; i++) {
+		struct range *r = &pt->time_ranges[i];
+		u64 ts = p[i].start;
+		u64 te = p[i].end;
+
+		/*
+		 * Take care to ensure the TSC range matches the perf-time range
+		 * when converted back to perf-time.
+		 */
+		r->start = ts ? intel_pt_tsc_start(ts, pt) : 0;
+		r->end   = te ? intel_pt_tsc_end(te, pt) : 0;
+
+		intel_pt_log("range %d: perf time interval: %"PRIu64" to %"PRIu64"\n",
+			     i, ts, te);
+		intel_pt_log("range %d: TSC time interval: %#"PRIx64" to %#"PRIx64"\n",
+			     i, r->start, r->end);
+	}
+
+	return 0;
+}
+
 static const char * const intel_pt_info_fmts[] = {
 	[INTEL_PT_PMU_TYPE]		= "  PMU Type            %"PRId64"\n",
 	[INTEL_PT_TIME_SHIFT]		= "  Time Shift          %"PRIu64"\n",
@@ -2752,6 +2953,12 @@ int intel_pt_process_auxtrace_info(union perf_event *event,
 		pt->cbr2khz = tsc_freq / pt->max_non_turbo_ratio / 1000;
 	}
 
+	if (session->itrace_synth_opts) {
+		err = intel_pt_setup_time_ranges(pt, session->itrace_synth_opts);
+		if (err)
+			goto err_delete_thread;
+	}
+
 	if (pt->synth_opts.calls)
 		pt->branches_filter |= PERF_IP_FLAG_CALL | PERF_IP_FLAG_ASYNC |
 				       PERF_IP_FLAG_TRACE_END;
@@ -2792,6 +2999,7 @@ err_free_queues:
 err_free:
 	addr_filters__exit(&pt->filts);
 	zfree(&pt->filter);
+	zfree(&pt->time_ranges);
 	free(pt);
 	return err;
 }

[PATCH 12/19] perf time-utils: Treat time ranges consistently

[Adrian Hunter]

Currently, options allow only 1 explicit (non-percentage) time range.
In preparation for adding support for multiple explicit time ranges,
treat time ranges consistently.

Instead of treating some time ranges as inclusive and some as excluding the
end time, treat all time ranges as inclusive. This is only a 1 nanosecond
change but is necessary to treat multiple explicit time ranges in a
consistent manner.

Note, there is a later patch that adds a test for time-utils.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/util/time-utils.c | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/tools/perf/util/time-utils.c b/tools/perf/util/time-utils.c
index 20663a460df3..1d67cf1216c7 100644
--- a/tools/perf/util/time-utils.c
+++ b/tools/perf/util/time-utils.c
@@ -389,13 +389,12 @@ bool perf_time__ranges_skip_sample(struct perf_time_interval *ptime_buf,
 		ptime = &ptime_buf[i];
 
 		if (timestamp >= ptime->start &&
-		    ((timestamp < ptime->end && i < num - 1) ||
-		     (timestamp <= ptime->end && i == num - 1))) {
-			break;
+		    (timestamp <= ptime->end || !ptime->end)) {
+			return false;
 		}
 	}
 
-	return (i == num) ? true : false;
+	return true;
 }
 
 int perf_time__parse_for_ranges(const char *time_str,
-- 
2.17.1

[tip:perf/core] perf time-utils: Treat time ranges consistently

[tip-bot for Adrian Hunter]

Commit-ID:  f79a7689d99366aee9f89d785bca6c52ed6b76eb
Gitweb:     https://git.kernel.org/tip/f79a7689d99366aee9f89d785bca6c52ed6b76eb
Author:     Adrian Hunter <adrian.hunter@intel.com>
AuthorDate: Tue, 4 Jun 2019 16:00:10 +0300
Committer:  Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Mon, 10 Jun 2019 16:20:12 -0300

perf time-utils: Treat time ranges consistently

Currently, options allow only 1 explicit (non-percentage) time range.
In preparation for adding support for multiple explicit time ranges,
treat time ranges consistently.

Instead of treating some time ranges as inclusive and some as excluding
the end time, treat all time ranges as inclusive. This is only a 1
nanosecond change but is necessary to treat multiple explicit time
ranges in a consistent manner.

Note, there is a later patch that adds a test for time-utils.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-13-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/util/time-utils.c | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/tools/perf/util/time-utils.c b/tools/perf/util/time-utils.c
index 20663a460df3..1d67cf1216c7 100644
--- a/tools/perf/util/time-utils.c
+++ b/tools/perf/util/time-utils.c
@@ -389,13 +389,12 @@ bool perf_time__ranges_skip_sample(struct perf_time_interval *ptime_buf,
 		ptime = &ptime_buf[i];
 
 		if (timestamp >= ptime->start &&
-		    ((timestamp < ptime->end && i < num - 1) ||
-		     (timestamp <= ptime->end && i == num - 1))) {
-			break;
+		    (timestamp <= ptime->end || !ptime->end)) {
+			return false;
 		}
 	}
 
-	return (i == num) ? true : false;
+	return true;
 }
 
 int perf_time__parse_for_ranges(const char *time_str,

[PATCH 13/19] perf time-utils: Factor out set_percent_time()

[Adrian Hunter]

Factor out set_percent_time() so it can be reused.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/util/time-utils.c | 39 +++++++++++++++++-------------------
 1 file changed, 18 insertions(+), 21 deletions(-)

diff --git a/tools/perf/util/time-utils.c b/tools/perf/util/time-utils.c
index 1d67cf1216c7..69441faab3d0 100644
--- a/tools/perf/util/time-utils.c
+++ b/tools/perf/util/time-utils.c
@@ -135,12 +135,27 @@ static int parse_percent(double *pcnt, char *str)
 	return 0;
 }
 
+static int set_percent_time(struct perf_time_interval *ptime, double start_pcnt,
+			    double end_pcnt, u64 start, u64 end)
+{
+	u64 total = end - start;
+
+	if (start_pcnt < 0.0 || start_pcnt > 1.0 ||
+	    end_pcnt < 0.0 || end_pcnt > 1.0) {
+		return -1;
+	}
+
+	ptime->start = start + round(start_pcnt * total);
+	ptime->end = start + round(end_pcnt * total);
+
+	return 0;
+}
+
 static int percent_slash_split(char *str, struct perf_time_interval *ptime,
 			       u64 start, u64 end)
 {
 	char *p, *end_str;
 	double pcnt, start_pcnt, end_pcnt;
-	u64 total = end - start;
 	int i;
 
 	/*
@@ -168,15 +183,7 @@ static int percent_slash_split(char *str, struct perf_time_interval *ptime,
 	start_pcnt = pcnt * (i - 1);
 	end_pcnt = pcnt * i;
 
-	if (start_pcnt < 0.0 || start_pcnt > 1.0 ||
-	    end_pcnt < 0.0 || end_pcnt > 1.0) {
-		return -1;
-	}
-
-	ptime->start = start + round(start_pcnt * total);
-	ptime->end = start + round(end_pcnt * total);
-
-	return 0;
+	return set_percent_time(ptime, start_pcnt, end_pcnt, start, end);
 }
 
 static int percent_dash_split(char *str, struct perf_time_interval *ptime,
@@ -184,7 +191,6 @@ static int percent_dash_split(char *str, struct perf_time_interval *ptime,
 {
 	char *start_str = NULL, *end_str;
 	double start_pcnt, end_pcnt;
-	u64 total = end - start;
 	int ret;
 
 	/*
@@ -203,16 +209,7 @@ static int percent_dash_split(char *str, struct perf_time_interval *ptime,
 
 	free(start_str);
 
-	if (start_pcnt < 0.0 || start_pcnt > 1.0 ||
-	    end_pcnt < 0.0 || end_pcnt > 1.0 ||
-	    start_pcnt > end_pcnt) {
-		return -1;
-	}
-
-	ptime->start = start + round(start_pcnt * total);
-	ptime->end = start + round(end_pcnt * total);
-
-	return 0;
+	return set_percent_time(ptime, start_pcnt, end_pcnt, start, end);
 }
 
 typedef int (*time_pecent_split)(char *, struct perf_time_interval *,
-- 
2.17.1

[tip:perf/core] perf time-utils: Factor out set_percent_time()

[tip-bot for Adrian Hunter]

Commit-ID:  c763242a5e742f8fefda0bb6cfdf6a5a34ae5e10
Gitweb:     https://git.kernel.org/tip/c763242a5e742f8fefda0bb6cfdf6a5a34ae5e10
Author:     Adrian Hunter <adrian.hunter@intel.com>
AuthorDate: Tue, 4 Jun 2019 16:00:11 +0300
Committer:  Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Mon, 10 Jun 2019 16:20:12 -0300

perf time-utils: Factor out set_percent_time()

Factor out set_percent_time() so it can be reused.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-14-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/util/time-utils.c | 39 ++++++++++++++++++---------------------
 1 file changed, 18 insertions(+), 21 deletions(-)

diff --git a/tools/perf/util/time-utils.c b/tools/perf/util/time-utils.c
index 1d67cf1216c7..69441faab3d0 100644
--- a/tools/perf/util/time-utils.c
+++ b/tools/perf/util/time-utils.c
@@ -135,12 +135,27 @@ static int parse_percent(double *pcnt, char *str)
 	return 0;
 }
 
+static int set_percent_time(struct perf_time_interval *ptime, double start_pcnt,
+			    double end_pcnt, u64 start, u64 end)
+{
+	u64 total = end - start;
+
+	if (start_pcnt < 0.0 || start_pcnt > 1.0 ||
+	    end_pcnt < 0.0 || end_pcnt > 1.0) {
+		return -1;
+	}
+
+	ptime->start = start + round(start_pcnt * total);
+	ptime->end = start + round(end_pcnt * total);
+
+	return 0;
+}
+
 static int percent_slash_split(char *str, struct perf_time_interval *ptime,
 			       u64 start, u64 end)
 {
 	char *p, *end_str;
 	double pcnt, start_pcnt, end_pcnt;
-	u64 total = end - start;
 	int i;
 
 	/*
@@ -168,15 +183,7 @@ static int percent_slash_split(char *str, struct perf_time_interval *ptime,
 	start_pcnt = pcnt * (i - 1);
 	end_pcnt = pcnt * i;
 
-	if (start_pcnt < 0.0 || start_pcnt > 1.0 ||
-	    end_pcnt < 0.0 || end_pcnt > 1.0) {
-		return -1;
-	}
-
-	ptime->start = start + round(start_pcnt * total);
-	ptime->end = start + round(end_pcnt * total);
-
-	return 0;
+	return set_percent_time(ptime, start_pcnt, end_pcnt, start, end);
 }
 
 static int percent_dash_split(char *str, struct perf_time_interval *ptime,
@@ -184,7 +191,6 @@ static int percent_dash_split(char *str, struct perf_time_interval *ptime,
 {
 	char *start_str = NULL, *end_str;
 	double start_pcnt, end_pcnt;
-	u64 total = end - start;
 	int ret;
 
 	/*
@@ -203,16 +209,7 @@ static int percent_dash_split(char *str, struct perf_time_interval *ptime,
 
 	free(start_str);
 
-	if (start_pcnt < 0.0 || start_pcnt > 1.0 ||
-	    end_pcnt < 0.0 || end_pcnt > 1.0 ||
-	    start_pcnt > end_pcnt) {
-		return -1;
-	}
-
-	ptime->start = start + round(start_pcnt * total);
-	ptime->end = start + round(end_pcnt * total);
-
-	return 0;
+	return set_percent_time(ptime, start_pcnt, end_pcnt, start, end);
 }
 
 typedef int (*time_pecent_split)(char *, struct perf_time_interval *,

[PATCH 14/19] perf time-utils: Prevent percentage time range overlap

[Adrian Hunter]

Prevent percentage time range overlap. This is only a 1 nanosecond change
but makes the results more logical e.g. a sample cannot be in both the
first 10% and the second 20%.

Note, there is a later patch that adds a test for time-utils.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/util/time-utils.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/tools/perf/util/time-utils.c b/tools/perf/util/time-utils.c
index 69441faab3d0..3e87c21c293c 100644
--- a/tools/perf/util/time-utils.c
+++ b/tools/perf/util/time-utils.c
@@ -148,6 +148,9 @@ static int set_percent_time(struct perf_time_interval *ptime, double start_pcnt,
 	ptime->start = start + round(start_pcnt * total);
 	ptime->end = start + round(end_pcnt * total);
 
+	if (ptime->end > ptime->start && ptime->end != end)
+		ptime->end -= 1;
+
 	return 0;
 }
 
-- 
2.17.1

[tip:perf/core] perf time-utils: Prevent percentage time range overlap

[tip-bot for Adrian Hunter]

Commit-ID:  b16bfeb3db1b50273e95f539953c337be759500d
Gitweb:     https://git.kernel.org/tip/b16bfeb3db1b50273e95f539953c337be759500d
Author:     Adrian Hunter <adrian.hunter@intel.com>
AuthorDate: Tue, 4 Jun 2019 16:00:12 +0300
Committer:  Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Mon, 10 Jun 2019 16:20:12 -0300

perf time-utils: Prevent percentage time range overlap

Prevent percentage time range overlap. This is only a 1 nanosecond
change but makes the results more logical e.g. a sample cannot be in
both the first 10% and the second 20%.

Note, there is a later patch that adds a test for time-utils.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-15-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/util/time-utils.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/tools/perf/util/time-utils.c b/tools/perf/util/time-utils.c
index 69441faab3d0..3e87c21c293c 100644
--- a/tools/perf/util/time-utils.c
+++ b/tools/perf/util/time-utils.c
@@ -148,6 +148,9 @@ static int set_percent_time(struct perf_time_interval *ptime, double start_pcnt,
 	ptime->start = start + round(start_pcnt * total);
 	ptime->end = start + round(end_pcnt * total);
 
+	if (ptime->end > ptime->start && ptime->end != end)
+		ptime->end -= 1;
+
 	return 0;
 }
 

[PATCH 15/19] perf time-utils: Fix --time documentation

[Adrian Hunter]

Correct some punctuation and spelling and correct the format to show that
the time resolution is nanoseconds not microseconds.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/Documentation/perf-diff.txt   | 6 +++---
 tools/perf/Documentation/perf-report.txt | 6 +++---
 tools/perf/Documentation/perf-script.txt | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/tools/perf/Documentation/perf-diff.txt b/tools/perf/Documentation/perf-diff.txt
index da7809b15cc9..5732f69580ab 100644
--- a/tools/perf/Documentation/perf-diff.txt
+++ b/tools/perf/Documentation/perf-diff.txt
@@ -142,9 +142,9 @@ OPTIONS
 	  perf diff --time 0%-10%,30%-40%
 
 	It also supports analyzing samples within a given time window
-	<start>,<stop>. Times have the format seconds.microseconds. If 'start'
-	is not given (i.e., time string is ',x.y') then analysis starts at
-	the beginning of the file. If stop time is not given (i.e, time
+	<start>,<stop>. Times have the format seconds.nanoseconds. If 'start'
+	is not given (i.e. time string is ',x.y') then analysis starts at
+	the beginning of the file. If stop time is not given (i.e. time
 	string is 'x.y,') then analysis goes to the end of the file. Time string is
 	'a1.b1,c1.d1:a2.b2,c2.d2'. Use ':' to separate timestamps for different
 	perf.data files.
diff --git a/tools/perf/Documentation/perf-report.txt b/tools/perf/Documentation/perf-report.txt
index f441baa794ce..3de029f6881d 100644
--- a/tools/perf/Documentation/perf-report.txt
+++ b/tools/perf/Documentation/perf-report.txt
@@ -412,12 +412,12 @@ OPTIONS
 
 --time::
 	Only analyze samples within given time window: <start>,<stop>. Times
-	have the format seconds.microseconds. If start is not given (i.e., time
+	have the format seconds.nanoseconds. If start is not given (i.e. time
 	string is ',x.y') then analysis starts at the beginning of the file. If
-	stop time is not given (i.e, time string is 'x.y,') then analysis goes
+	stop time is not given (i.e. time string is 'x.y,') then analysis goes
 	to end of file.
 
-	Also support time percent with multiple time range. Time string is
+	Also support time percent with multiple time ranges. Time string is
 	'a%/n,b%/m,...' or 'a%-b%,c%-%d,...'.
 
 	For example:
diff --git a/tools/perf/Documentation/perf-script.txt b/tools/perf/Documentation/perf-script.txt
index c59fd52e9e91..878349cce968 100644
--- a/tools/perf/Documentation/perf-script.txt
+++ b/tools/perf/Documentation/perf-script.txt
@@ -361,12 +361,12 @@ include::itrace.txt[]
 
 --time::
 	Only analyze samples within given time window: <start>,<stop>. Times
-	have the format seconds.microseconds. If start is not given (i.e., time
+	have the format seconds.nanoseconds. If start is not given (i.e. time
 	string is ',x.y') then analysis starts at the beginning of the file. If
-	stop time is not given (i.e, time string is 'x.y,') then analysis goes
+	stop time is not given (i.e. time string is 'x.y,') then analysis goes
 	to end of file.
 
-	Also support time percent with multipe time range. Time string is
+	Also support time percent with multiple time ranges. Time string is
 	'a%/n,b%/m,...' or 'a%-b%,c%-%d,...'.
 
 	For example:
-- 
2.17.1

[tip:perf/core] perf time-utils: Fix --time documentation

[tip-bot for Adrian Hunter]

Commit-ID:  0ccc69ba0af46e3371c8cefb506aaf9f0e4f554c
Gitweb:     https://git.kernel.org/tip/0ccc69ba0af46e3371c8cefb506aaf9f0e4f554c
Author:     Adrian Hunter <adrian.hunter@intel.com>
AuthorDate: Tue, 4 Jun 2019 16:00:13 +0300
Committer:  Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Mon, 10 Jun 2019 16:20:12 -0300

perf time-utils: Fix --time documentation

Correct some punctuation and spelling and correct the format to show
that the time resolution is nanoseconds not microseconds.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-16-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/Documentation/perf-diff.txt   | 6 +++---
 tools/perf/Documentation/perf-report.txt | 6 +++---
 tools/perf/Documentation/perf-script.txt | 6 +++---
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/tools/perf/Documentation/perf-diff.txt b/tools/perf/Documentation/perf-diff.txt
index da7809b15cc9..5732f69580ab 100644
--- a/tools/perf/Documentation/perf-diff.txt
+++ b/tools/perf/Documentation/perf-diff.txt
@@ -142,9 +142,9 @@ OPTIONS
 	  perf diff --time 0%-10%,30%-40%
 
 	It also supports analyzing samples within a given time window
-	<start>,<stop>. Times have the format seconds.microseconds. If 'start'
-	is not given (i.e., time string is ',x.y') then analysis starts at
-	the beginning of the file. If stop time is not given (i.e, time
+	<start>,<stop>. Times have the format seconds.nanoseconds. If 'start'
+	is not given (i.e. time string is ',x.y') then analysis starts at
+	the beginning of the file. If stop time is not given (i.e. time
 	string is 'x.y,') then analysis goes to the end of the file. Time string is
 	'a1.b1,c1.d1:a2.b2,c2.d2'. Use ':' to separate timestamps for different
 	perf.data files.
diff --git a/tools/perf/Documentation/perf-report.txt b/tools/perf/Documentation/perf-report.txt
index f441baa794ce..3de029f6881d 100644
--- a/tools/perf/Documentation/perf-report.txt
+++ b/tools/perf/Documentation/perf-report.txt
@@ -412,12 +412,12 @@ OPTIONS
 
 --time::
 	Only analyze samples within given time window: <start>,<stop>. Times
-	have the format seconds.microseconds. If start is not given (i.e., time
+	have the format seconds.nanoseconds. If start is not given (i.e. time
 	string is ',x.y') then analysis starts at the beginning of the file. If
-	stop time is not given (i.e, time string is 'x.y,') then analysis goes
+	stop time is not given (i.e. time string is 'x.y,') then analysis goes
 	to end of file.
 
-	Also support time percent with multiple time range. Time string is
+	Also support time percent with multiple time ranges. Time string is
 	'a%/n,b%/m,...' or 'a%-b%,c%-%d,...'.
 
 	For example:
diff --git a/tools/perf/Documentation/perf-script.txt b/tools/perf/Documentation/perf-script.txt
index c59fd52e9e91..878349cce968 100644
--- a/tools/perf/Documentation/perf-script.txt
+++ b/tools/perf/Documentation/perf-script.txt
@@ -361,12 +361,12 @@ include::itrace.txt[]
 
 --time::
 	Only analyze samples within given time window: <start>,<stop>. Times
-	have the format seconds.microseconds. If start is not given (i.e., time
+	have the format seconds.nanoseconds. If start is not given (i.e. time
 	string is ',x.y') then analysis starts at the beginning of the file. If
-	stop time is not given (i.e, time string is 'x.y,') then analysis goes
+	stop time is not given (i.e. time string is 'x.y,') then analysis goes
 	to end of file.
 
-	Also support time percent with multipe time range. Time string is
+	Also support time percent with multiple time ranges. Time string is
 	'a%/n,b%/m,...' or 'a%-b%,c%-%d,...'.
 
 	For example:

[PATCH 16/19] perf time-utils: Simplify perf_time__parse_for_ranges() error paths slightly

[Adrian Hunter]

Simplify perf_time__parse_for_ranges() error paths slightly.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/util/time-utils.c | 12 +++++-------
 1 file changed, 5 insertions(+), 7 deletions(-)

diff --git a/tools/perf/util/time-utils.c b/tools/perf/util/time-utils.c
index 3e87c21c293c..9a463752dba8 100644
--- a/tools/perf/util/time-utils.c
+++ b/tools/perf/util/time-utils.c
@@ -403,7 +403,7 @@ int perf_time__parse_for_ranges(const char *time_str,
 				int *range_size, int *range_num)
 {
 	struct perf_time_interval *ptime_range;
-	int size, num, ret;
+	int size, num, ret = -EINVAL;
 
 	ptime_range = perf_time__range_alloc(time_str, &size);
 	if (!ptime_range)
@@ -415,7 +415,6 @@ int perf_time__parse_for_ranges(const char *time_str,
 			pr_err("HINT: no first/last sample time found in perf data.\n"
 			       "Please use latest perf binary to execute 'perf record'\n"
 			       "(if '--buildid-all' is enabled, please set '--timestamp-boundary').\n");
-			ret = -EINVAL;
 			goto error;
 		}
 
@@ -425,11 +424,8 @@ int perf_time__parse_for_ranges(const char *time_str,
 				session->evlist->first_sample_time,
 				session->evlist->last_sample_time);
 
-		if (num < 0) {
-			pr_err("Invalid time string\n");
-			ret = -EINVAL;
-			goto error;
-		}
+		if (num < 0)
+			goto error_invalid;
 	} else {
 		num = 1;
 	}
@@ -439,6 +435,8 @@ int perf_time__parse_for_ranges(const char *time_str,
 	*ranges = ptime_range;
 	return 0;
 
+error_invalid:
+	pr_err("Invalid time string\n");
 error:
 	free(ptime_range);
 	return ret;
-- 
2.17.1

[tip:perf/core] perf time-utils: Simplify perf_time__parse_for_ranges() error paths slightly

[tip-bot for Adrian Hunter]

Commit-ID:  2a8afddc084a5f5f933382758dd2767ed8a69f77
Gitweb:     https://git.kernel.org/tip/2a8afddc084a5f5f933382758dd2767ed8a69f77
Author:     Adrian Hunter <adrian.hunter@intel.com>
AuthorDate: Tue, 4 Jun 2019 16:00:14 +0300
Committer:  Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Mon, 10 Jun 2019 16:20:12 -0300

perf time-utils: Simplify perf_time__parse_for_ranges() error paths slightly

Simplify perf_time__parse_for_ranges() error paths slightly.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-17-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/util/time-utils.c | 12 +++++-------
 1 file changed, 5 insertions(+), 7 deletions(-)

diff --git a/tools/perf/util/time-utils.c b/tools/perf/util/time-utils.c
index 3e87c21c293c..9a463752dba8 100644
--- a/tools/perf/util/time-utils.c
+++ b/tools/perf/util/time-utils.c
@@ -403,7 +403,7 @@ int perf_time__parse_for_ranges(const char *time_str,
 				int *range_size, int *range_num)
 {
 	struct perf_time_interval *ptime_range;
-	int size, num, ret;
+	int size, num, ret = -EINVAL;
 
 	ptime_range = perf_time__range_alloc(time_str, &size);
 	if (!ptime_range)
@@ -415,7 +415,6 @@ int perf_time__parse_for_ranges(const char *time_str,
 			pr_err("HINT: no first/last sample time found in perf data.\n"
 			       "Please use latest perf binary to execute 'perf record'\n"
 			       "(if '--buildid-all' is enabled, please set '--timestamp-boundary').\n");
-			ret = -EINVAL;
 			goto error;
 		}
 
@@ -425,11 +424,8 @@ int perf_time__parse_for_ranges(const char *time_str,
 				session->evlist->first_sample_time,
 				session->evlist->last_sample_time);
 
-		if (num < 0) {
-			pr_err("Invalid time string\n");
-			ret = -EINVAL;
-			goto error;
-		}
+		if (num < 0)
+			goto error_invalid;
 	} else {
 		num = 1;
 	}
@@ -439,6 +435,8 @@ int perf_time__parse_for_ranges(const char *time_str,
 	*ranges = ptime_range;
 	return 0;
 
+error_invalid:
+	pr_err("Invalid time string\n");
 error:
 	free(ptime_range);
 	return ret;

[PATCH 17/19] perf time-utils: Make perf_time__parse_for_ranges() more logical

[Adrian Hunter]

Explicit time ranges never contain a percent sign whereas percentage ranges
always do, so it is possible to call the correct parser.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/util/time-utils.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/tools/perf/util/time-utils.c b/tools/perf/util/time-utils.c
index 9a463752dba8..d942840356e3 100644
--- a/tools/perf/util/time-utils.c
+++ b/tools/perf/util/time-utils.c
@@ -402,6 +402,7 @@ int perf_time__parse_for_ranges(const char *time_str,
 				struct perf_time_interval **ranges,
 				int *range_size, int *range_num)
 {
+	bool has_percent = strchr(time_str, '%');
 	struct perf_time_interval *ptime_range;
 	int size, num, ret = -EINVAL;
 
@@ -409,7 +410,7 @@ int perf_time__parse_for_ranges(const char *time_str,
 	if (!ptime_range)
 		return -ENOMEM;
 
-	if (perf_time__parse_str(ptime_range, time_str) != 0) {
+	if (has_percent) {
 		if (session->evlist->first_sample_time == 0 &&
 		    session->evlist->last_sample_time == 0) {
 			pr_err("HINT: no first/last sample time found in perf data.\n"
@@ -427,6 +428,8 @@ int perf_time__parse_for_ranges(const char *time_str,
 		if (num < 0)
 			goto error_invalid;
 	} else {
+		if (perf_time__parse_str(ptime_range, time_str))
+			goto error_invalid;
 		num = 1;
 	}
 
-- 
2.17.1

[tip:perf/core] perf time-utils: Make perf_time__parse_for_ranges() more logical

[tip-bot for Adrian Hunter]

Commit-ID:  929afa0092d0ea6be2fbd0ac087319092595eba6
Gitweb:     https://git.kernel.org/tip/929afa0092d0ea6be2fbd0ac087319092595eba6
Author:     Adrian Hunter <adrian.hunter@intel.com>
AuthorDate: Tue, 4 Jun 2019 16:00:15 +0300
Committer:  Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Mon, 10 Jun 2019 16:20:12 -0300

perf time-utils: Make perf_time__parse_for_ranges() more logical

Explicit time ranges never contain a percent sign whereas percentage
ranges always do, so it is possible to call the correct parser.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-18-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/util/time-utils.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/tools/perf/util/time-utils.c b/tools/perf/util/time-utils.c
index 9a463752dba8..d942840356e3 100644
--- a/tools/perf/util/time-utils.c
+++ b/tools/perf/util/time-utils.c
@@ -402,6 +402,7 @@ int perf_time__parse_for_ranges(const char *time_str,
 				struct perf_time_interval **ranges,
 				int *range_size, int *range_num)
 {
+	bool has_percent = strchr(time_str, '%');
 	struct perf_time_interval *ptime_range;
 	int size, num, ret = -EINVAL;
 
@@ -409,7 +410,7 @@ int perf_time__parse_for_ranges(const char *time_str,
 	if (!ptime_range)
 		return -ENOMEM;
 
-	if (perf_time__parse_str(ptime_range, time_str) != 0) {
+	if (has_percent) {
 		if (session->evlist->first_sample_time == 0 &&
 		    session->evlist->last_sample_time == 0) {
 			pr_err("HINT: no first/last sample time found in perf data.\n"
@@ -427,6 +428,8 @@ int perf_time__parse_for_ranges(const char *time_str,
 		if (num < 0)
 			goto error_invalid;
 	} else {
+		if (perf_time__parse_str(ptime_range, time_str))
+			goto error_invalid;
 		num = 1;
 	}
 

[PATCH 18/19] perf tests: Add a test for time-utils

[Adrian Hunter]

Test time ranges work as expected.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/tests/Build             |   1 +
 tools/perf/tests/builtin-test.c    |   4 +
 tools/perf/tests/tests.h           |   1 +
 tools/perf/tests/time-utils-test.c | 234 +++++++++++++++++++++++++++++
 4 files changed, 240 insertions(+)
 create mode 100644 tools/perf/tests/time-utils-test.c

diff --git a/tools/perf/tests/Build b/tools/perf/tests/Build
index 4afb6319ed51..e3ba63cef01e 100644
--- a/tools/perf/tests/Build
+++ b/tools/perf/tests/Build
@@ -51,6 +51,7 @@ perf-y += clang.o
 perf-y += unit_number__scnprintf.o
 perf-y += mem2node.o
 perf-y += map_groups.o
+perf-y += time-utils-test.o
 
 $(OUTPUT)tests/llvm-src-base.c: tests/bpf-script-example.c tests/Build
 	$(call rule_mkdir)
diff --git a/tools/perf/tests/builtin-test.c b/tools/perf/tests/builtin-test.c
index 941c5456d625..cd72ff0f7658 100644
--- a/tools/perf/tests/builtin-test.c
+++ b/tools/perf/tests/builtin-test.c
@@ -289,6 +289,10 @@ static struct test generic_tests[] = {
 		.desc = "mem2node",
 		.func = test__mem2node,
 	},
+	{
+		.desc = "time utils",
+		.func = test__time_utils,
+	},
 	{
 		.desc = "map_groups__merge_in",
 		.func = test__map_groups__merge_in,
diff --git a/tools/perf/tests/tests.h b/tools/perf/tests/tests.h
index e5e3a57cd373..72912eb473cb 100644
--- a/tools/perf/tests/tests.h
+++ b/tools/perf/tests/tests.h
@@ -108,6 +108,7 @@ int test__clang_subtest_get_nr(void);
 int test__unit_number__scnprint(struct test *test, int subtest);
 int test__mem2node(struct test *t, int subtest);
 int test__map_groups__merge_in(struct test *t, int subtest);
+int test__time_utils(struct test *t, int subtest);
 
 bool test__bp_signal_is_supported(void);
 bool test__wp_is_supported(void);
diff --git a/tools/perf/tests/time-utils-test.c b/tools/perf/tests/time-utils-test.c
new file mode 100644
index 000000000000..7504046b111c
--- /dev/null
+++ b/tools/perf/tests/time-utils-test.c
@@ -0,0 +1,234 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <linux/compiler.h>
+#include <linux/time64.h>
+#include <inttypes.h>
+#include <string.h>
+#include "time-utils.h"
+#include "evlist.h"
+#include "session.h"
+#include "debug.h"
+#include "tests.h"
+
+static bool test__parse_nsec_time(const char *str, u64 expected)
+{
+	u64 ptime;
+	int err;
+
+	pr_debug("\nparse_nsec_time(\"%s\")\n", str);
+
+	err = parse_nsec_time(str, &ptime);
+	if (err) {
+		pr_debug("error %d\n", err);
+		return false;
+	}
+
+	if (ptime != expected) {
+		pr_debug("Failed. ptime %" PRIu64 " expected %" PRIu64 "\n",
+			 ptime, expected);
+		return false;
+	}
+
+	pr_debug("%" PRIu64 "\n", ptime);
+
+	return true;
+}
+
+static bool test__perf_time__parse_str(const char *ostr, u64 start, u64 end)
+{
+	struct perf_time_interval ptime;
+	int err;
+
+	pr_debug("\nperf_time__parse_str(\"%s\")\n", ostr);
+
+	err = perf_time__parse_str(&ptime, ostr);
+	if (err) {
+		pr_debug("Error %d\n", err);
+		return false;
+	}
+
+	if (ptime.start != start || ptime.end != end) {
+		pr_debug("Failed. Expected %" PRIu64 " to %" PRIu64 "\n",
+			 start, end);
+		return false;
+	}
+
+	return true;
+}
+
+#define TEST_MAX 64
+
+struct test_data {
+	const char *str;
+	u64 first;
+	u64 last;
+	struct perf_time_interval ptime[TEST_MAX];
+	int num;
+	u64 skip[TEST_MAX];
+	u64 noskip[TEST_MAX];
+};
+
+static bool test__perf_time__parse_for_ranges(struct test_data *d)
+{
+	struct perf_evlist evlist = {
+		.first_sample_time = d->first,
+		.last_sample_time = d->last,
+	};
+	struct perf_session session = { .evlist = &evlist };
+	struct perf_time_interval *ptime = NULL;
+	int range_size, range_num;
+	bool pass = false;
+	int i, err;
+
+	pr_debug("\nperf_time__parse_for_ranges(\"%s\")\n", d->str);
+
+	if (strchr(d->str, '%'))
+		pr_debug("first_sample_time %" PRIu64 " last_sample_time %" PRIu64 "\n",
+			 d->first, d->last);
+
+	err = perf_time__parse_for_ranges(d->str, &session, &ptime, &range_size,
+					  &range_num);
+	if (err) {
+		pr_debug("error %d\n", err);
+		goto out;
+	}
+
+	if (range_size < d->num || range_num != d->num) {
+		pr_debug("bad size: range_size %d range_num %d expected num %d\n",
+			 range_size, range_num, d->num);
+		goto out;
+	}
+
+	for (i = 0; i < d->num; i++) {
+		if (ptime[i].start != d->ptime[i].start ||
+		    ptime[i].end != d->ptime[i].end) {
+			pr_debug("bad range %d expected %" PRIu64 " to %" PRIu64 "\n",
+				 i, d->ptime[i].start, d->ptime[i].end);
+			goto out;
+		}
+	}
+
+	if (perf_time__ranges_skip_sample(ptime, d->num, 0)) {
+		pr_debug("failed to keep 0\n");
+		goto out;
+	}
+
+	for (i = 0; i < TEST_MAX; i++) {
+		if (d->skip[i] &&
+		    !perf_time__ranges_skip_sample(ptime, d->num, d->skip[i])) {
+			pr_debug("failed to skip %" PRIu64 "\n", d->skip[i]);
+			goto out;
+		}
+		if (d->noskip[i] &&
+		    perf_time__ranges_skip_sample(ptime, d->num, d->noskip[i])) {
+			pr_debug("failed to keep %" PRIu64 "\n", d->noskip[i]);
+			goto out;
+		}
+	}
+
+	pass = true;
+out:
+	free(ptime);
+	return pass;
+}
+
+int test__time_utils(struct test *t __maybe_unused, int subtest __maybe_unused)
+{
+	bool pass = true;
+
+	pass &= test__parse_nsec_time("0", 0);
+	pass &= test__parse_nsec_time("1", 1000000000ULL);
+	pass &= test__parse_nsec_time("0.000000001", 1);
+	pass &= test__parse_nsec_time("1.000000001", 1000000001ULL);
+	pass &= test__parse_nsec_time("123456.123456", 123456123456000ULL);
+	pass &= test__parse_nsec_time("1234567.123456789", 1234567123456789ULL);
+	pass &= test__parse_nsec_time("18446744073.709551615",
+				      0xFFFFFFFFFFFFFFFFULL);
+
+	pass &= test__perf_time__parse_str("1234567.123456789,1234567.123456789",
+					   1234567123456789ULL, 1234567123456789ULL);
+	pass &= test__perf_time__parse_str("1234567.123456789,1234567.123456790",
+					   1234567123456789ULL, 1234567123456790ULL);
+	pass &= test__perf_time__parse_str("1234567.123456789,",
+					   1234567123456789ULL, 0);
+	pass &= test__perf_time__parse_str(",1234567.123456789",
+					   0, 1234567123456789ULL);
+	pass &= test__perf_time__parse_str("0,1234567.123456789",
+					   0, 1234567123456789ULL);
+
+	{
+		u64 b = 1234567123456789ULL;
+		struct test_data d = {
+			.str   = "1234567.123456789,1234567.123456790",
+			.ptime = { {b, b + 1}, },
+			.num = 1,
+			.skip = { b - 1, b + 2, },
+			.noskip = { b, b + 1, },
+		};
+
+		pass &= test__perf_time__parse_for_ranges(&d);
+	}
+
+	{
+		u64 b = 7654321ULL * NSEC_PER_SEC;
+		struct test_data d = {
+			.str    = "10%/1",
+			.first  = b,
+			.last   = b + 100,
+			.ptime  = { {b, b + 9}, },
+			.num    = 1,
+			.skip   = { b - 1, b + 10, },
+			.noskip = { b, b + 9, },
+		};
+
+		pass &= test__perf_time__parse_for_ranges(&d);
+	}
+
+	{
+		u64 b = 7654321ULL * NSEC_PER_SEC;
+		struct test_data d = {
+			.str    = "10%/2",
+			.first  = b,
+			.last   = b + 100,
+			.ptime  = { {b + 10, b + 19}, },
+			.num    = 1,
+			.skip   = { b + 9, b + 20, },
+			.noskip = { b + 10, b + 19, },
+		};
+
+		pass &= test__perf_time__parse_for_ranges(&d);
+	}
+
+	{
+		u64 b = 11223344ULL * NSEC_PER_SEC;
+		struct test_data d = {
+			.str    = "10%/1,10%/2",
+			.first  = b,
+			.last   = b + 100,
+			.ptime  = { {b, b + 9}, {b + 10, b + 19}, },
+			.num    = 2,
+			.skip   = { b - 1, b + 20, },
+			.noskip = { b, b + 8, b + 9, b + 10, b + 11, b + 12, b + 19, },
+		};
+
+		pass &= test__perf_time__parse_for_ranges(&d);
+	}
+
+	{
+		u64 b = 11223344ULL * NSEC_PER_SEC;
+		struct test_data d = {
+			.str    = "10%/1,10%/3,10%/10",
+			.first  = b,
+			.last   = b + 100,
+			.ptime  = { {b, b + 9}, {b + 20, b + 29}, { b + 90, b + 100}, },
+			.num    = 3,
+			.skip   = { b - 1, b + 10, b + 19, b + 30, b + 89, b + 101 },
+			.noskip = { b, b + 9, b + 20, b + 29, b + 90, b + 100},
+		};
+
+		pass &= test__perf_time__parse_for_ranges(&d);
+	}
+
+	pr_debug("\n");
+
+	return pass ? 0 : TEST_FAIL;
+}
-- 
2.17.1

[tip:perf/core] perf tests: Add a test for time-utils

[tip-bot for Adrian Hunter]

Commit-ID:  e39a12cbd2496edb4cab0f99efb0d217c55ba273
Gitweb:     https://git.kernel.org/tip/e39a12cbd2496edb4cab0f99efb0d217c55ba273
Author:     Adrian Hunter <adrian.hunter@intel.com>
AuthorDate: Tue, 4 Jun 2019 16:00:16 +0300
Committer:  Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Mon, 10 Jun 2019 16:20:12 -0300

perf tests: Add a test for time-utils

Test time ranges work as expected.

Committer testing:

  $ perf test "time utils"
  59: time utils                                            : Ok
  $ perf test -v "time utils"
  59: time utils                                            :
  --- start ---
  test child forked, pid 31711

  parse_nsec_time("0")
  0

  parse_nsec_time("1")
  1000000000

  parse_nsec_time("0.000000001")
  1

  parse_nsec_time("1.000000001")
  1000000001

  parse_nsec_time("123456.123456")
  123456123456000

  parse_nsec_time("1234567.123456789")
  1234567123456789

  parse_nsec_time("18446744073.709551615")
  18446744073709551615

  perf_time__parse_str("1234567.123456789,1234567.123456789")
  start time 1234567123456789, end time 1234567123456789

  perf_time__parse_str("1234567.123456789,1234567.123456790")
  start time 1234567123456789, end time 1234567123456790

  perf_time__parse_str("1234567.123456789,")
  start time 1234567123456789, end time 0

  perf_time__parse_str(",1234567.123456789")
  start time 0, end time 1234567123456789

  perf_time__parse_str("0,1234567.123456789")
  start time 0, end time 1234567123456789

  perf_time__parse_for_ranges("1234567.123456789,1234567.123456790")
  start time 1234567123456789, end time 1234567123456790

  perf_time__parse_for_ranges("10%/1")
  first_sample_time 7654321000000000 last_sample_time 7654321000000100
  start time 0: 7654321000000000, end time 0: 7654321000000009

  perf_time__parse_for_ranges("10%/2")
  first_sample_time 7654321000000000 last_sample_time 7654321000000100
  start time 0: 7654321000000010, end time 0: 7654321000000019

  perf_time__parse_for_ranges("10%/1,10%/2")
  first_sample_time 11223344000000000 last_sample_time 11223344000000100
  start time 0: 11223344000000000, end time 0: 11223344000000009
  start time 1: 11223344000000010, end time 1: 11223344000000019

  perf_time__parse_for_ranges("10%/1,10%/3,10%/10")
  first_sample_time 11223344000000000 last_sample_time 11223344000000100
  start time 0: 11223344000000000, end time 0: 11223344000000009
  start time 1: 11223344000000020, end time 1: 11223344000000029
  start time 2: 11223344000000090, end time 2: 11223344000000100

  test child finished with 0
  ---- end ----
  time utils: Ok
  $

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-19-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/tests/Build             |   1 +
 tools/perf/tests/builtin-test.c    |   4 +
 tools/perf/tests/tests.h           |   1 +
 tools/perf/tests/time-utils-test.c | 234 +++++++++++++++++++++++++++++++++++++
 4 files changed, 240 insertions(+)

diff --git a/tools/perf/tests/Build b/tools/perf/tests/Build
index 4afb6319ed51..e3ba63cef01e 100644
--- a/tools/perf/tests/Build
+++ b/tools/perf/tests/Build
@@ -51,6 +51,7 @@ perf-y += clang.o
 perf-y += unit_number__scnprintf.o
 perf-y += mem2node.o
 perf-y += map_groups.o
+perf-y += time-utils-test.o
 
 $(OUTPUT)tests/llvm-src-base.c: tests/bpf-script-example.c tests/Build
 	$(call rule_mkdir)
diff --git a/tools/perf/tests/builtin-test.c b/tools/perf/tests/builtin-test.c
index 941c5456d625..cd72ff0f7658 100644
--- a/tools/perf/tests/builtin-test.c
+++ b/tools/perf/tests/builtin-test.c
@@ -289,6 +289,10 @@ static struct test generic_tests[] = {
 		.desc = "mem2node",
 		.func = test__mem2node,
 	},
+	{
+		.desc = "time utils",
+		.func = test__time_utils,
+	},
 	{
 		.desc = "map_groups__merge_in",
 		.func = test__map_groups__merge_in,
diff --git a/tools/perf/tests/tests.h b/tools/perf/tests/tests.h
index e5e3a57cd373..72912eb473cb 100644
--- a/tools/perf/tests/tests.h
+++ b/tools/perf/tests/tests.h
@@ -108,6 +108,7 @@ int test__clang_subtest_get_nr(void);
 int test__unit_number__scnprint(struct test *test, int subtest);
 int test__mem2node(struct test *t, int subtest);
 int test__map_groups__merge_in(struct test *t, int subtest);
+int test__time_utils(struct test *t, int subtest);
 
 bool test__bp_signal_is_supported(void);
 bool test__wp_is_supported(void);
diff --git a/tools/perf/tests/time-utils-test.c b/tools/perf/tests/time-utils-test.c
new file mode 100644
index 000000000000..7504046b111c
--- /dev/null
+++ b/tools/perf/tests/time-utils-test.c
@@ -0,0 +1,234 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <linux/compiler.h>
+#include <linux/time64.h>
+#include <inttypes.h>
+#include <string.h>
+#include "time-utils.h"
+#include "evlist.h"
+#include "session.h"
+#include "debug.h"
+#include "tests.h"
+
+static bool test__parse_nsec_time(const char *str, u64 expected)
+{
+	u64 ptime;
+	int err;
+
+	pr_debug("\nparse_nsec_time(\"%s\")\n", str);
+
+	err = parse_nsec_time(str, &ptime);
+	if (err) {
+		pr_debug("error %d\n", err);
+		return false;
+	}
+
+	if (ptime != expected) {
+		pr_debug("Failed. ptime %" PRIu64 " expected %" PRIu64 "\n",
+			 ptime, expected);
+		return false;
+	}
+
+	pr_debug("%" PRIu64 "\n", ptime);
+
+	return true;
+}
+
+static bool test__perf_time__parse_str(const char *ostr, u64 start, u64 end)
+{
+	struct perf_time_interval ptime;
+	int err;
+
+	pr_debug("\nperf_time__parse_str(\"%s\")\n", ostr);
+
+	err = perf_time__parse_str(&ptime, ostr);
+	if (err) {
+		pr_debug("Error %d\n", err);
+		return false;
+	}
+
+	if (ptime.start != start || ptime.end != end) {
+		pr_debug("Failed. Expected %" PRIu64 " to %" PRIu64 "\n",
+			 start, end);
+		return false;
+	}
+
+	return true;
+}
+
+#define TEST_MAX 64
+
+struct test_data {
+	const char *str;
+	u64 first;
+	u64 last;
+	struct perf_time_interval ptime[TEST_MAX];
+	int num;
+	u64 skip[TEST_MAX];
+	u64 noskip[TEST_MAX];
+};
+
+static bool test__perf_time__parse_for_ranges(struct test_data *d)
+{
+	struct perf_evlist evlist = {
+		.first_sample_time = d->first,
+		.last_sample_time = d->last,
+	};
+	struct perf_session session = { .evlist = &evlist };
+	struct perf_time_interval *ptime = NULL;
+	int range_size, range_num;
+	bool pass = false;
+	int i, err;
+
+	pr_debug("\nperf_time__parse_for_ranges(\"%s\")\n", d->str);
+
+	if (strchr(d->str, '%'))
+		pr_debug("first_sample_time %" PRIu64 " last_sample_time %" PRIu64 "\n",
+			 d->first, d->last);
+
+	err = perf_time__parse_for_ranges(d->str, &session, &ptime, &range_size,
+					  &range_num);
+	if (err) {
+		pr_debug("error %d\n", err);
+		goto out;
+	}
+
+	if (range_size < d->num || range_num != d->num) {
+		pr_debug("bad size: range_size %d range_num %d expected num %d\n",
+			 range_size, range_num, d->num);
+		goto out;
+	}
+
+	for (i = 0; i < d->num; i++) {
+		if (ptime[i].start != d->ptime[i].start ||
+		    ptime[i].end != d->ptime[i].end) {
+			pr_debug("bad range %d expected %" PRIu64 " to %" PRIu64 "\n",
+				 i, d->ptime[i].start, d->ptime[i].end);
+			goto out;
+		}
+	}
+
+	if (perf_time__ranges_skip_sample(ptime, d->num, 0)) {
+		pr_debug("failed to keep 0\n");
+		goto out;
+	}
+
+	for (i = 0; i < TEST_MAX; i++) {
+		if (d->skip[i] &&
+		    !perf_time__ranges_skip_sample(ptime, d->num, d->skip[i])) {
+			pr_debug("failed to skip %" PRIu64 "\n", d->skip[i]);
+			goto out;
+		}
+		if (d->noskip[i] &&
+		    perf_time__ranges_skip_sample(ptime, d->num, d->noskip[i])) {
+			pr_debug("failed to keep %" PRIu64 "\n", d->noskip[i]);
+			goto out;
+		}
+	}
+
+	pass = true;
+out:
+	free(ptime);
+	return pass;
+}
+
+int test__time_utils(struct test *t __maybe_unused, int subtest __maybe_unused)
+{
+	bool pass = true;
+
+	pass &= test__parse_nsec_time("0", 0);
+	pass &= test__parse_nsec_time("1", 1000000000ULL);
+	pass &= test__parse_nsec_time("0.000000001", 1);
+	pass &= test__parse_nsec_time("1.000000001", 1000000001ULL);
+	pass &= test__parse_nsec_time("123456.123456", 123456123456000ULL);
+	pass &= test__parse_nsec_time("1234567.123456789", 1234567123456789ULL);
+	pass &= test__parse_nsec_time("18446744073.709551615",
+				      0xFFFFFFFFFFFFFFFFULL);
+
+	pass &= test__perf_time__parse_str("1234567.123456789,1234567.123456789",
+					   1234567123456789ULL, 1234567123456789ULL);
+	pass &= test__perf_time__parse_str("1234567.123456789,1234567.123456790",
+					   1234567123456789ULL, 1234567123456790ULL);
+	pass &= test__perf_time__parse_str("1234567.123456789,",
+					   1234567123456789ULL, 0);
+	pass &= test__perf_time__parse_str(",1234567.123456789",
+					   0, 1234567123456789ULL);
+	pass &= test__perf_time__parse_str("0,1234567.123456789",
+					   0, 1234567123456789ULL);
+
+	{
+		u64 b = 1234567123456789ULL;
+		struct test_data d = {
+			.str   = "1234567.123456789,1234567.123456790",
+			.ptime = { {b, b + 1}, },
+			.num = 1,
+			.skip = { b - 1, b + 2, },
+			.noskip = { b, b + 1, },
+		};
+
+		pass &= test__perf_time__parse_for_ranges(&d);
+	}
+
+	{
+		u64 b = 7654321ULL * NSEC_PER_SEC;
+		struct test_data d = {
+			.str    = "10%/1",
+			.first  = b,
+			.last   = b + 100,
+			.ptime  = { {b, b + 9}, },
+			.num    = 1,
+			.skip   = { b - 1, b + 10, },
+			.noskip = { b, b + 9, },
+		};
+
+		pass &= test__perf_time__parse_for_ranges(&d);
+	}
+
+	{
+		u64 b = 7654321ULL * NSEC_PER_SEC;
+		struct test_data d = {
+			.str    = "10%/2",
+			.first  = b,
+			.last   = b + 100,
+			.ptime  = { {b + 10, b + 19}, },
+			.num    = 1,
+			.skip   = { b + 9, b + 20, },
+			.noskip = { b + 10, b + 19, },
+		};
+
+		pass &= test__perf_time__parse_for_ranges(&d);
+	}
+
+	{
+		u64 b = 11223344ULL * NSEC_PER_SEC;
+		struct test_data d = {
+			.str    = "10%/1,10%/2",
+			.first  = b,
+			.last   = b + 100,
+			.ptime  = { {b, b + 9}, {b + 10, b + 19}, },
+			.num    = 2,
+			.skip   = { b - 1, b + 20, },
+			.noskip = { b, b + 8, b + 9, b + 10, b + 11, b + 12, b + 19, },
+		};
+
+		pass &= test__perf_time__parse_for_ranges(&d);
+	}
+
+	{
+		u64 b = 11223344ULL * NSEC_PER_SEC;
+		struct test_data d = {
+			.str    = "10%/1,10%/3,10%/10",
+			.first  = b,
+			.last   = b + 100,
+			.ptime  = { {b, b + 9}, {b + 20, b + 29}, { b + 90, b + 100}, },
+			.num    = 3,
+			.skip   = { b - 1, b + 10, b + 19, b + 30, b + 89, b + 101 },
+			.noskip = { b, b + 9, b + 20, b + 29, b + 90, b + 100},
+		};
+
+		pass &= test__perf_time__parse_for_ranges(&d);
+	}
+
+	pr_debug("\n");
+
+	return pass ? 0 : TEST_FAIL;
+}

[PATCH 19/19] perf time-utils: Add support for multiple explicit time intervals

[Adrian Hunter]

Currently only a single explicit time range is accepted. Add support for
multiple ranges separated by spaces, which requires the string to be
quoted. Update the time utils test accordingly.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/Documentation/perf-diff.txt   |  8 ++-
 tools/perf/Documentation/perf-report.txt |  3 +-
 tools/perf/Documentation/perf-script.txt |  3 +-
 tools/perf/tests/time-utils-test.c       | 17 ++++++
 tools/perf/util/time-utils.c             | 74 ++++++++++++++++++++++--
 5 files changed, 94 insertions(+), 11 deletions(-)

diff --git a/tools/perf/Documentation/perf-diff.txt b/tools/perf/Documentation/perf-diff.txt
index 5732f69580ab..facd91e4e945 100644
--- a/tools/perf/Documentation/perf-diff.txt
+++ b/tools/perf/Documentation/perf-diff.txt
@@ -145,9 +145,11 @@ OPTIONS
 	<start>,<stop>. Times have the format seconds.nanoseconds. If 'start'
 	is not given (i.e. time string is ',x.y') then analysis starts at
 	the beginning of the file. If stop time is not given (i.e. time
-	string is 'x.y,') then analysis goes to the end of the file. Time string is
-	'a1.b1,c1.d1:a2.b2,c2.d2'. Use ':' to separate timestamps for different
-	perf.data files.
+	string is 'x.y,') then analysis goes to the end of the file.
+	Multiple ranges can be separated by spaces, which requires the argument
+	to be quoted e.g. --time "1234.567,1234.789 1235,"
+	Time string is'a1.b1,c1.d1:a2.b2,c2.d2'. Use ':' to separate timestamps
+	for different perf.data files.
 
 	For example, we get the timestamp information from 'perf script'.
 
diff --git a/tools/perf/Documentation/perf-report.txt b/tools/perf/Documentation/perf-report.txt
index 3de029f6881d..8c4372819e11 100644
--- a/tools/perf/Documentation/perf-report.txt
+++ b/tools/perf/Documentation/perf-report.txt
@@ -415,7 +415,8 @@ OPTIONS
 	have the format seconds.nanoseconds. If start is not given (i.e. time
 	string is ',x.y') then analysis starts at the beginning of the file. If
 	stop time is not given (i.e. time string is 'x.y,') then analysis goes
-	to end of file.
+	to end of file. Multiple ranges can be separated by spaces, which
+	requires the argument to be quoted e.g. --time "1234.567,1234.789 1235,"
 
 	Also support time percent with multiple time ranges. Time string is
 	'a%/n,b%/m,...' or 'a%-b%,c%-%d,...'.
diff --git a/tools/perf/Documentation/perf-script.txt b/tools/perf/Documentation/perf-script.txt
index 878349cce968..d4e2e18a5881 100644
--- a/tools/perf/Documentation/perf-script.txt
+++ b/tools/perf/Documentation/perf-script.txt
@@ -364,7 +364,8 @@ include::itrace.txt[]
 	have the format seconds.nanoseconds. If start is not given (i.e. time
 	string is ',x.y') then analysis starts at the beginning of the file. If
 	stop time is not given (i.e. time string is 'x.y,') then analysis goes
-	to end of file.
+	to end of file. Multiple ranges can be separated by spaces, which
+	requires the argument to be quoted e.g. --time "1234.567,1234.789 1235,"
 
 	Also support time percent with multiple time ranges. Time string is
 	'a%/n,b%/m,...' or 'a%-b%,c%-%d,...'.
diff --git a/tools/perf/tests/time-utils-test.c b/tools/perf/tests/time-utils-test.c
index 7504046b111c..4f53006233a1 100644
--- a/tools/perf/tests/time-utils-test.c
+++ b/tools/perf/tests/time-utils-test.c
@@ -168,6 +168,23 @@ int test__time_utils(struct test *t __maybe_unused, int subtest __maybe_unused)
 		pass &= test__perf_time__parse_for_ranges(&d);
 	}
 
+	{
+		u64 b = 1234567123456789ULL;
+		u64 c = 7654321987654321ULL;
+		u64 e = 8000000000000000ULL;
+		struct test_data d = {
+			.str   = "1234567.123456789,1234567.123456790 "
+				 "7654321.987654321,7654321.987654444 "
+				 "8000000,8000000.000000005",
+			.ptime = { {b, b + 1}, {c, c + 123}, {e, e + 5}, },
+			.num = 3,
+			.skip = { b - 1, b + 2, c - 1, c + 124, e - 1, e + 6 },
+			.noskip = { b, b + 1, c, c + 123, e, e + 5 },
+		};
+
+		pass &= test__perf_time__parse_for_ranges(&d);
+	}
+
 	{
 		u64 b = 7654321ULL * NSEC_PER_SEC;
 		struct test_data d = {
diff --git a/tools/perf/util/time-utils.c b/tools/perf/util/time-utils.c
index d942840356e3..2b48816a2d2e 100644
--- a/tools/perf/util/time-utils.c
+++ b/tools/perf/util/time-utils.c
@@ -7,6 +7,7 @@
 #include <errno.h>
 #include <inttypes.h>
 #include <math.h>
+#include <ctype.h>
 
 #include "perf.h"
 #include "debug.h"
@@ -116,6 +117,69 @@ int perf_time__parse_str(struct perf_time_interval *ptime, const char *ostr)
 	return rc;
 }
 
+static int perf_time__parse_strs(struct perf_time_interval *ptime,
+				 const char *ostr, int size)
+{
+	const char *cp;
+	char *str, *arg, *p;
+	int i, num = 0, rc = 0;
+
+	/* Count the commas */
+	for (cp = ostr; *cp; cp++)
+		num += !!(*cp == ',');
+
+	if (!num)
+		return -EINVAL;
+
+	BUG_ON(num > size);
+
+	str = strdup(ostr);
+	if (!str)
+		return -ENOMEM;
+
+	/* Split the string and parse each piece, except the last */
+	for (i = 0, p = str; i < num - 1; i++) {
+		arg = p;
+		/* Find next comma, there must be one */
+		p = strchr(p, ',') + 1;
+		/* Skip white space */
+		while (isspace(*p))
+			p++;
+		/* Skip the value, must not contain space or comma */
+		while (*p && !isspace(*p)) {
+			if (*p++ == ',') {
+				rc = -EINVAL;
+				goto out;
+			}
+		}
+		/* Split and parse */
+		if (*p)
+			*p++ = 0;
+		rc = perf_time__parse_str(ptime + i, arg);
+		if (rc < 0)
+			goto out;
+	}
+
+	/* Parse the last piece */
+	rc = perf_time__parse_str(ptime + i, p);
+	if (rc < 0)
+		goto out;
+
+	/* Check there is no overlap */
+	for (i = 0; i < num - 1; i++) {
+		if (ptime[i].end >= ptime[i + 1].start) {
+			rc = -EINVAL;
+			goto out;
+		}
+	}
+
+	rc = num;
+out:
+	free(str);
+
+	return rc;
+}
+
 static int parse_percent(double *pcnt, char *str)
 {
 	char *c, *endptr;
@@ -424,15 +488,13 @@ int perf_time__parse_for_ranges(const char *time_str,
 				time_str,
 				session->evlist->first_sample_time,
 				session->evlist->last_sample_time);
-
-		if (num < 0)
-			goto error_invalid;
 	} else {
-		if (perf_time__parse_str(ptime_range, time_str))
-			goto error_invalid;
-		num = 1;
+		num = perf_time__parse_strs(ptime_range, time_str, size);
 	}
 
+	if (num < 0)
+		goto error_invalid;
+
 	*range_size = size;
 	*range_num = num;
 	*ranges = ptime_range;
-- 
2.17.1

[tip:perf/core] perf time-utils: Add support for multiple explicit time intervals

[tip-bot for Adrian Hunter]

Commit-ID:  a77a05e2337df1347f4de96bfa313db7008fe8bd
Gitweb:     https://git.kernel.org/tip/a77a05e2337df1347f4de96bfa313db7008fe8bd
Author:     Adrian Hunter <adrian.hunter@intel.com>
AuthorDate: Tue, 4 Jun 2019 16:00:17 +0300
Committer:  Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Mon, 10 Jun 2019 16:20:13 -0300

perf time-utils: Add support for multiple explicit time intervals

Currently only a single explicit time range is accepted. Add support for
multiple ranges separated by spaces, which requires the string to be
quoted. Update the time utils test accordingly.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-20-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/Documentation/perf-diff.txt   |  8 ++--
 tools/perf/Documentation/perf-report.txt |  3 +-
 tools/perf/Documentation/perf-script.txt |  3 +-
 tools/perf/tests/time-utils-test.c       | 17 ++++++++
 tools/perf/util/time-utils.c             | 74 +++++++++++++++++++++++++++++---
 5 files changed, 94 insertions(+), 11 deletions(-)

diff --git a/tools/perf/Documentation/perf-diff.txt b/tools/perf/Documentation/perf-diff.txt
index 5732f69580ab..facd91e4e945 100644
--- a/tools/perf/Documentation/perf-diff.txt
+++ b/tools/perf/Documentation/perf-diff.txt
@@ -145,9 +145,11 @@ OPTIONS
 	<start>,<stop>. Times have the format seconds.nanoseconds. If 'start'
 	is not given (i.e. time string is ',x.y') then analysis starts at
 	the beginning of the file. If stop time is not given (i.e. time
-	string is 'x.y,') then analysis goes to the end of the file. Time string is
-	'a1.b1,c1.d1:a2.b2,c2.d2'. Use ':' to separate timestamps for different
-	perf.data files.
+	string is 'x.y,') then analysis goes to the end of the file.
+	Multiple ranges can be separated by spaces, which requires the argument
+	to be quoted e.g. --time "1234.567,1234.789 1235,"
+	Time string is'a1.b1,c1.d1:a2.b2,c2.d2'. Use ':' to separate timestamps
+	for different perf.data files.
 
 	For example, we get the timestamp information from 'perf script'.
 
diff --git a/tools/perf/Documentation/perf-report.txt b/tools/perf/Documentation/perf-report.txt
index 3de029f6881d..8c4372819e11 100644
--- a/tools/perf/Documentation/perf-report.txt
+++ b/tools/perf/Documentation/perf-report.txt
@@ -415,7 +415,8 @@ OPTIONS
 	have the format seconds.nanoseconds. If start is not given (i.e. time
 	string is ',x.y') then analysis starts at the beginning of the file. If
 	stop time is not given (i.e. time string is 'x.y,') then analysis goes
-	to end of file.
+	to end of file. Multiple ranges can be separated by spaces, which
+	requires the argument to be quoted e.g. --time "1234.567,1234.789 1235,"
 
 	Also support time percent with multiple time ranges. Time string is
 	'a%/n,b%/m,...' or 'a%-b%,c%-%d,...'.
diff --git a/tools/perf/Documentation/perf-script.txt b/tools/perf/Documentation/perf-script.txt
index 878349cce968..d4e2e18a5881 100644
--- a/tools/perf/Documentation/perf-script.txt
+++ b/tools/perf/Documentation/perf-script.txt
@@ -364,7 +364,8 @@ include::itrace.txt[]
 	have the format seconds.nanoseconds. If start is not given (i.e. time
 	string is ',x.y') then analysis starts at the beginning of the file. If
 	stop time is not given (i.e. time string is 'x.y,') then analysis goes
-	to end of file.
+	to end of file. Multiple ranges can be separated by spaces, which
+	requires the argument to be quoted e.g. --time "1234.567,1234.789 1235,"
 
 	Also support time percent with multiple time ranges. Time string is
 	'a%/n,b%/m,...' or 'a%-b%,c%-%d,...'.
diff --git a/tools/perf/tests/time-utils-test.c b/tools/perf/tests/time-utils-test.c
index 7504046b111c..4f53006233a1 100644
--- a/tools/perf/tests/time-utils-test.c
+++ b/tools/perf/tests/time-utils-test.c
@@ -168,6 +168,23 @@ int test__time_utils(struct test *t __maybe_unused, int subtest __maybe_unused)
 		pass &= test__perf_time__parse_for_ranges(&d);
 	}
 
+	{
+		u64 b = 1234567123456789ULL;
+		u64 c = 7654321987654321ULL;
+		u64 e = 8000000000000000ULL;
+		struct test_data d = {
+			.str   = "1234567.123456789,1234567.123456790 "
+				 "7654321.987654321,7654321.987654444 "
+				 "8000000,8000000.000000005",
+			.ptime = { {b, b + 1}, {c, c + 123}, {e, e + 5}, },
+			.num = 3,
+			.skip = { b - 1, b + 2, c - 1, c + 124, e - 1, e + 6 },
+			.noskip = { b, b + 1, c, c + 123, e, e + 5 },
+		};
+
+		pass &= test__perf_time__parse_for_ranges(&d);
+	}
+
 	{
 		u64 b = 7654321ULL * NSEC_PER_SEC;
 		struct test_data d = {
diff --git a/tools/perf/util/time-utils.c b/tools/perf/util/time-utils.c
index d942840356e3..2b48816a2d2e 100644
--- a/tools/perf/util/time-utils.c
+++ b/tools/perf/util/time-utils.c
@@ -7,6 +7,7 @@
 #include <errno.h>
 #include <inttypes.h>
 #include <math.h>
+#include <ctype.h>
 
 #include "perf.h"
 #include "debug.h"
@@ -116,6 +117,69 @@ int perf_time__parse_str(struct perf_time_interval *ptime, const char *ostr)
 	return rc;
 }
 
+static int perf_time__parse_strs(struct perf_time_interval *ptime,
+				 const char *ostr, int size)
+{
+	const char *cp;
+	char *str, *arg, *p;
+	int i, num = 0, rc = 0;
+
+	/* Count the commas */
+	for (cp = ostr; *cp; cp++)
+		num += !!(*cp == ',');
+
+	if (!num)
+		return -EINVAL;
+
+	BUG_ON(num > size);
+
+	str = strdup(ostr);
+	if (!str)
+		return -ENOMEM;
+
+	/* Split the string and parse each piece, except the last */
+	for (i = 0, p = str; i < num - 1; i++) {
+		arg = p;
+		/* Find next comma, there must be one */
+		p = strchr(p, ',') + 1;
+		/* Skip white space */
+		while (isspace(*p))
+			p++;
+		/* Skip the value, must not contain space or comma */
+		while (*p && !isspace(*p)) {
+			if (*p++ == ',') {
+				rc = -EINVAL;
+				goto out;
+			}
+		}
+		/* Split and parse */
+		if (*p)
+			*p++ = 0;
+		rc = perf_time__parse_str(ptime + i, arg);
+		if (rc < 0)
+			goto out;
+	}
+
+	/* Parse the last piece */
+	rc = perf_time__parse_str(ptime + i, p);
+	if (rc < 0)
+		goto out;
+
+	/* Check there is no overlap */
+	for (i = 0; i < num - 1; i++) {
+		if (ptime[i].end >= ptime[i + 1].start) {
+			rc = -EINVAL;
+			goto out;
+		}
+	}
+
+	rc = num;
+out:
+	free(str);
+
+	return rc;
+}
+
 static int parse_percent(double *pcnt, char *str)
 {
 	char *c, *endptr;
@@ -424,15 +488,13 @@ int perf_time__parse_for_ranges(const char *time_str,
 				time_str,
 				session->evlist->first_sample_time,
 				session->evlist->last_sample_time);
-
-		if (num < 0)
-			goto error_invalid;
 	} else {
-		if (perf_time__parse_str(ptime_range, time_str))
-			goto error_invalid;
-		num = 1;
+		num = perf_time__parse_strs(ptime_range, time_str, size);
 	}
 
+	if (num < 0)
+		goto error_invalid;
+
 	*range_size = size;
 	*range_num = num;
 	*ranges = ptime_range;

Re: [PATCH 00/19] perf intel-pt: Add support for efficient time interval filtering

[Arnaldo Carvalho de Melo]

Em Tue, Jun 04, 2019 at 03:59:58PM +0300, Adrian Hunter escreveu:
> Hi
> 
> Here are some patches to add support for efficient time interval filtering.
> First there are 3 patches to add perf time interval to the itrace options
> structure.  Then changes to Intel PT to support "fast forwarding" to a
> particular timestamp.  The filtering is added in patch "perf intel-pt: Add
> support for efficient time interval filtering".  After that there are
> patches to time-utils, leading up to adding a new test and adding support
> multiple explicit time intervals.
> 
> The Intel PT changes make time filtering much faster because decoding is
> limited to the minimal time ranges needed to support the time intervals.

Thanks, applied!

Will now go with the other patches in this batch to the container tests,
hopefully should hit Ingo's mail box soon :-)

- Arnaldo

- Arnaldo
 
> 
> Adrian Hunter (19):
>       perf auxtrace: Add perf time interval to itrace_synth_ops
>       perf script: Set perf time interval in itrace_synth_ops
>       perf report: Set perf time interval in itrace_synth_ops
>       perf intel-pt: Add lookahead callback
>       perf intel-pt: Factor out intel_pt_8b_tsc()
>       perf intel-pt: Factor out intel_pt_reposition()
>       perf intel-pt: Add reposition parameter to intel_pt_get_data()
>       perf intel-pt: Add intel_pt_fast_forward()
>       perf intel-pt: Factor out intel_pt_get_buffer()
>       perf intel-pt: Add support for lookahead
>       perf intel-pt: Add support for efficient time interval filtering
>       perf time-utils: Treat time ranges consistently
>       perf time-utils: Factor out set_percent_time()
>       perf time-utils: Prevent percentage time range overlap
>       perf time-utils: Fix --time documentation
>       perf time-utils: Simplify perf_time__parse_for_ranges() error paths slightly
>       perf time-utils: Make perf_time__parse_for_ranges() more logical
>       perf tests: Add a test for time-utils
>       perf time-utils: Add support for multiple explicit time intervals
> 
>  tools/perf/Documentation/perf-diff.txt             |  14 +-
>  tools/perf/Documentation/perf-report.txt           |   9 +-
>  tools/perf/Documentation/perf-script.txt           |   9 +-
>  tools/perf/builtin-report.c                        |   8 +-
>  tools/perf/builtin-script.c                        |   8 +-
>  tools/perf/tests/Build                             |   1 +
>  tools/perf/tests/builtin-test.c                    |   4 +
>  tools/perf/tests/tests.h                           |   1 +
>  tools/perf/tests/time-utils-test.c                 | 251 ++++++++++++++++
>  tools/perf/util/auxtrace.h                         |  34 +++
>  .../perf/util/intel-pt-decoder/intel-pt-decoder.c  | 188 ++++++++++--
>  .../perf/util/intel-pt-decoder/intel-pt-decoder.h  |   5 +
>  tools/perf/util/intel-pt.c                         | 325 +++++++++++++++++++--
>  tools/perf/util/time-utils.c                       | 132 ++++++---
>  14 files changed, 894 insertions(+), 95 deletions(-)
>  create mode 100644 tools/perf/tests/time-utils-test.c
> 
> 
> Regards
> Adrian

-- 

- Arnaldo