[tip:perf/urgent] perf report: Implement browsing of individual samples

[tip-bot for Andi Kleen]

Commit-ID:  4968ac8fb7c378e2bc40b7e9bd97768fa8c7aa32
Gitweb:     https://git.kernel.org/tip/4968ac8fb7c378e2bc40b7e9bd97768fa8c7aa32
Author:     Andi Kleen <ak@xxxxxxxxxxxxxxx>
AuthorDate: Mon, 11 Mar 2019 07:44:58 -0700
Committer:  Arnaldo Carvalho de Melo <acme@xxxxxxxxxx>
CommitDate: Mon, 11 Mar 2019 16:33:19 -0300

perf report: Implement browsing of individual samples

Now 'perf report' can show whole time periods with 'perf script', but
the user still has to find individual samples of interest manually.

It would be expensive and complicated to search for the right samples in
the whole perf file. Typically users only need to look at a small number
of samples for useful analysis.

Also the full scripts tend to show samples of all CPUs and all threads
mixed up, which can be very confusing on larger systems.

Add a new --samples option to save a small random number of samples per
hist entry.

Use a reservoir sample technique to select a representatve number of
samples.

Then allow browsing the samples using 'perf script' as part of the hist
entry context menu. This automatically adds the right filters, so only
the thread or cpu of the sample is displayed. Then we use less' search
functionality to directly jump the to the time stamp of the selected
sample.

It uses different menus for assembler and source display.  Assembler
needs xed installed and source needs debuginfo.

Currently it only supports as many samples as fit on the screen due to
some limitations in the slang ui code.

Signed-off-by: Andi Kleen <ak@xxxxxxxxxxxxxxx>
Tested-by: Arnaldo Carvalho de Melo <acme@xxxxxxxxxx>
Acked-by: Jiri Olsa <jolsa@xxxxxxxxxx>
Link: http://lkml.kernel.org/r/20190311174605.GA29294@xxxxxxxxxxxxxxxxxxxx
Signed-off-by: Arnaldo Carvalho de Melo <acme@xxxxxxxxxx>
---
 tools/perf/Documentation/perf-config.txt |  6 ++
 tools/perf/Documentation/perf-report.txt |  4 ++
 tools/perf/builtin-report.c              |  2 +
 tools/perf/ui/browsers/Build             |  1 +
 tools/perf/ui/browsers/hists.c           | 47 ++++++++++++++++
 tools/perf/ui/browsers/res_sample.c      | 94 ++++++++++++++++++++++++++++++++
 tools/perf/ui/browsers/scripts.c         |  2 +-
 tools/perf/util/hist.c                   | 39 +++++++++++++
 tools/perf/util/hist.h                   | 22 ++++++++
 tools/perf/util/sort.h                   |  8 +++
 tools/perf/util/symbol.c                 |  1 +
 tools/perf/util/symbol_conf.h            |  1 +
 12 files changed, 226 insertions(+), 1 deletion(-)

diff --git a/tools/perf/Documentation/perf-config.txt b/tools/perf/Documentation/perf-config.txt
index 86f3dcc15f83..2d0fb7613134 100644
--- a/tools/perf/Documentation/perf-config.txt
+++ b/tools/perf/Documentation/perf-config.txt
@@ -584,6 +584,12 @@ llvm.*::
 	llvm.opts::
 		Options passed to llc.
 
+samples.*::
+
+	samples.context::
+		Define how many ns worth of time to show
+		around samples in perf report sample context browser.
+
 SEE ALSO
 --------
 linkperf:perf[1]
diff --git a/tools/perf/Documentation/perf-report.txt b/tools/perf/Documentation/perf-report.txt
index 546d87221ad8..f441baa794ce 100644
--- a/tools/perf/Documentation/perf-report.txt
+++ b/tools/perf/Documentation/perf-report.txt
@@ -461,6 +461,10 @@ include::itrace.txt[]
 --socket-filter::
 	Only report the samples on the processor socket that match with this filter
 
+--samples=N::
+	Save N individual samples for each histogram entry to show context in perf
+	report tui browser.
+
 --raw-trace::
 	When displaying traceevent output, do not use print fmt or plugins.
 
diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c
index 05c8dd41106c..1921aaa9cece 100644
--- a/tools/perf/builtin-report.c
+++ b/tools/perf/builtin-report.c
@@ -1159,6 +1159,8 @@ int cmd_report(int argc, const char **argv)
 	OPT_BOOLEAN(0, "demangle-kernel", &symbol_conf.demangle_kernel,
 		    "Enable kernel symbol demangling"),
 	OPT_BOOLEAN(0, "mem-mode", &report.mem_mode, "mem access profile"),
+	OPT_INTEGER(0, "samples", &symbol_conf.res_sample,
+		    "Number of samples to save per histogram entry for individual browsing"),
 	OPT_CALLBACK(0, "percent-limit", &report, "percent",
 		     "Don't show entries under that percent", parse_percent_limit),
 	OPT_CALLBACK(0, "percentage", NULL, "relative|absolute",
diff --git a/tools/perf/ui/browsers/Build b/tools/perf/ui/browsers/Build
index 8fee56b46502..fdf86f7981ca 100644
--- a/tools/perf/ui/browsers/Build
+++ b/tools/perf/ui/browsers/Build
@@ -3,6 +3,7 @@ perf-y += hists.o
 perf-y += map.o
 perf-y += scripts.o
 perf-y += header.o
+perf-y += res_sample.o
 
 CFLAGS_annotate.o += -DENABLE_SLFUTURE_CONST
 CFLAGS_hists.o    += -DENABLE_SLFUTURE_CONST
diff --git a/tools/perf/ui/browsers/hists.c b/tools/perf/ui/browsers/hists.c
index fb4430f8982c..3421ecbdd3f0 100644
--- a/tools/perf/ui/browsers/hists.c
+++ b/tools/perf/ui/browsers/hists.c
@@ -1226,6 +1226,8 @@ void hist_browser__init_hpp(void)
 				hist_browser__hpp_color_overhead_guest_us;
 	perf_hpp__format[PERF_HPP__OVERHEAD_ACC].color =
 				hist_browser__hpp_color_overhead_acc;
+
+	res_sample_init();
 }
 
 static int hist_browser__show_entry(struct hist_browser *browser,
@@ -2345,6 +2347,7 @@ struct popup_action {
 	struct map_symbol 	ms;
 	int			socket;
 	struct perf_evsel	*evsel;
+	enum rstype		rstype;
 
 	int (*fn)(struct hist_browser *browser, struct popup_action *act);
 };
@@ -2572,6 +2575,17 @@ do_run_script(struct hist_browser *browser __maybe_unused,
 	return 0;
 }
 
+static int
+do_res_sample_script(struct hist_browser *browser __maybe_unused,
+		     struct popup_action *act)
+{
+	struct hist_entry *he;
+
+	he = hist_browser__selected_entry(browser);
+	res_sample_browse(he->res_samples, he->num_res, act->evsel, act->rstype);
+	return 0;
+}
+
 static int
 add_script_opt_2(struct hist_browser *browser __maybe_unused,
 	       struct popup_action *act, char **optstr,
@@ -2629,6 +2643,27 @@ add_script_opt(struct hist_browser *browser,
 	return n;
 }
 
+static int
+add_res_sample_opt(struct hist_browser *browser __maybe_unused,
+		   struct popup_action *act, char **optstr,
+		   struct res_sample *res_sample,
+		   struct perf_evsel *evsel,
+		   enum rstype type)
+{
+	if (!res_sample)
+		return 0;
+
+	if (asprintf(optstr, "Show context for individual samples %s",
+		type == A_ASM ? "with assembler" :
+		type == A_SOURCE ? "with source" : "") < 0)
+		return 0;
+
+	act->fn = do_res_sample_script;
+	act->evsel = evsel;
+	act->rstype = type;
+	return 1;
+}
+
 static int
 do_switch_data(struct hist_browser *browser __maybe_unused,
 	       struct popup_action *act __maybe_unused)
@@ -3115,6 +3150,18 @@ skip_annotation:
 		}
 		nr_options += add_script_opt(browser, &actions[nr_options],
 					     &options[nr_options], NULL, NULL, evsel);
+		nr_options += add_res_sample_opt(browser, &actions[nr_options],
+						 &options[nr_options],
+				 hist_browser__selected_entry(browser)->res_samples,
+				 evsel, A_NORMAL);
+		nr_options += add_res_sample_opt(browser, &actions[nr_options],
+						 &options[nr_options],
+				 hist_browser__selected_entry(browser)->res_samples,
+				 evsel, A_ASM);
+		nr_options += add_res_sample_opt(browser, &actions[nr_options],
+						 &options[nr_options],
+				 hist_browser__selected_entry(browser)->res_samples,
+				 evsel, A_SOURCE);
 		nr_options += add_switch_opt(browser, &actions[nr_options],
 					     &options[nr_options]);
 skip_scripting:
diff --git a/tools/perf/ui/browsers/res_sample.c b/tools/perf/ui/browsers/res_sample.c
new file mode 100644
index 000000000000..884ef2a92c15
--- /dev/null
+++ b/tools/perf/ui/browsers/res_sample.c
@@ -0,0 +1,94 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Display a menu with individual samples to browse with perf script */
+#include "util.h"
+#include "hist.h"
+#include "evsel.h"
+#include "hists.h"
+#include "sort.h"
+#include "config.h"
+#include "time-utils.h"
+#include <linux/time64.h>
+
+static u64 context_len = 10 * NSEC_PER_MSEC;
+
+static int res_sample_config(const char *var, const char *value, void *data __maybe_unused)
+{
+	if (!strcmp(var, "samples.context"))
+		return perf_config_u64(&context_len, var, value);
+	return 0;
+}
+
+void res_sample_init(void)
+{
+	perf_config(res_sample_config, NULL);
+}
+
+int res_sample_browse(struct res_sample *res_samples, int num_res,
+		      struct perf_evsel *evsel, enum rstype rstype)
+{
+	char **names;
+	int i, n;
+	int choice;
+	char *cmd;
+	char pbuf[256], tidbuf[32], cpubuf[32];
+	const char *perf = perf_exe(pbuf, sizeof pbuf);
+	char trange[128], tsample[64];
+	struct res_sample *r;
+	char extra_format[256];
+
+	/* For now since ui__popup_menu doesn't like lists that don't fit */
+	num_res = max(min(SLtt_Screen_Rows - 4, num_res), 0);
+
+	names = calloc(num_res, sizeof(char *));
+	if (!names)
+		return -1;
+	for (i = 0; i < num_res; i++) {
+		char tbuf[64];
+
+		timestamp__scnprintf_nsec(res_samples[i].time, tbuf, sizeof tbuf);
+		if (asprintf(&names[i], "%s: CPU %d tid %d", tbuf,
+			     res_samples[i].cpu, res_samples[i].tid) < 0) {
+			while (--i >= 0)
+				free(names[i]);
+			free(names);
+			return -1;
+		}
+	}
+	choice = ui__popup_menu(num_res, names);
+	for (i = 0; i < num_res; i++)
+		free(names[i]);
+	free(names);
+
+	if (choice < 0 || choice >= num_res)
+		return -1;
+	r = &res_samples[choice];
+
+	n = timestamp__scnprintf_nsec(r->time - context_len, trange, sizeof trange);
+	trange[n++] = ',';
+	timestamp__scnprintf_nsec(r->time + context_len, trange + n, sizeof trange - n);
+
+	timestamp__scnprintf_nsec(r->time, tsample, sizeof tsample);
+
+	attr_to_script(extra_format, &evsel->attr);
+
+	if (asprintf(&cmd, "%s script %s%s --time %s %s%s %s%s --ns %s %s %s %s %s | less +/%s",
+		     perf,
+		     input_name ? "-i " : "",
+		     input_name ? input_name : "",
+		     trange,
+		     r->cpu >= 0 ? "--cpu " : "",
+		     r->cpu >= 0 ? (sprintf(cpubuf, "%d", r->cpu), cpubuf) : "",
+		     r->tid ? "--tid " : "",
+		     r->tid ? (sprintf(tidbuf, "%d", r->tid), tidbuf) : "",
+		     extra_format,
+		     rstype == A_ASM ? "-F +insn --xed" :
+		     rstype == A_SOURCE ? "-F +srcline,+srccode" : "",
+		     symbol_conf.inline_name ? "--inline" : "",
+		     "--show-lost-events ",
+		     r->tid ? "--show-switch-events --show-task-events " : "",
+		     tsample) < 0)
+		return -1;
+	run_script(cmd);
+	free(cmd);
+	return 0;
+}
diff --git a/tools/perf/ui/browsers/scripts.c b/tools/perf/ui/browsers/scripts.c
index 9e5f87558af6..cdba58447b85 100644
--- a/tools/perf/ui/browsers/scripts.c
+++ b/tools/perf/ui/browsers/scripts.c
@@ -125,7 +125,7 @@ out:
 	return ret;
 }
 
-static void run_script(char *cmd)
+void run_script(char *cmd)
 {
 	pr_debug("Running %s\n", cmd);
 	SLang_reset_tty();
diff --git a/tools/perf/util/hist.c b/tools/perf/util/hist.c
index 34c0f00c68d1..1f230285d78a 100644
--- a/tools/perf/util/hist.c
+++ b/tools/perf/util/hist.c
@@ -436,6 +436,13 @@ static int hist_entry__init(struct hist_entry *he,
 			goto err_rawdata;
 	}
 
+	if (symbol_conf.res_sample) {
+		he->res_samples = calloc(sizeof(struct res_sample),
+					symbol_conf.res_sample);
+		if (!he->res_samples)
+			goto err_srcline;
+	}
+
 	INIT_LIST_HEAD(&he->pairs.node);
 	thread__get(he->thread);
 	he->hroot_in  = RB_ROOT_CACHED;
@@ -446,6 +453,9 @@ static int hist_entry__init(struct hist_entry *he,
 
 	return 0;
 
+err_srcline:
+	free(he->srcline);
+
 err_rawdata:
 	free(he->raw_data);
 
@@ -603,6 +613,32 @@ out:
 	return he;
 }
 
+static unsigned random_max(unsigned high)
+{
+	unsigned thresh = -high % high;
+	for (;;) {
+		unsigned r = random();
+		if (r >= thresh)
+			return r % high;
+	}
+}
+
+static void hists__res_sample(struct hist_entry *he, struct perf_sample *sample)
+{
+	struct res_sample *r;
+	int j;
+
+	if (he->num_res < symbol_conf.res_sample) {
+		j = he->num_res++;
+	} else {
+		j = random_max(symbol_conf.res_sample);
+	}
+	r = &he->res_samples[j];
+	r->time = sample->time;
+	r->cpu = sample->cpu;
+	r->tid = sample->tid;
+}
+
 static struct hist_entry*
 __hists__add_entry(struct hists *hists,
 		   struct addr_location *al,
@@ -650,6 +686,8 @@ __hists__add_entry(struct hists *hists,
 
 	if (!hists->has_callchains && he && he->callchain_size != 0)
 		hists->has_callchains = true;
+	if (he && symbol_conf.res_sample)
+		hists__res_sample(he, sample);
 	return he;
 }
 
@@ -1173,6 +1211,7 @@ void hist_entry__delete(struct hist_entry *he)
 		mem_info__zput(he->mem_info);
 	}
 
+	zfree(&he->res_samples);
 	zfree(&he->stat_acc);
 	free_srcline(he->srcline);
 	if (he->srcfile && he->srcfile[0])
diff --git a/tools/perf/util/hist.h b/tools/perf/util/hist.h
index 2113a6639cea..76ff6c6d03b8 100644
--- a/tools/perf/util/hist.h
+++ b/tools/perf/util/hist.h
@@ -433,6 +433,13 @@ struct hist_browser_timer {
 };
 
 struct annotation_options;
+struct res_sample;
+
+enum rstype {
+	A_NORMAL,
+	A_ASM,
+	A_SOURCE
+};
 
 #ifdef HAVE_SLANG_SUPPORT
 #include "../ui/keysyms.h"
@@ -454,6 +461,11 @@ int perf_evlist__tui_browse_hists(struct perf_evlist *evlist, const char *help,
 				  struct annotation_options *annotation_options);
 
 int script_browse(const char *script_opt, struct perf_evsel *evsel);
+
+void run_script(char *cmd);
+int res_sample_browse(struct res_sample *res_samples, int num_res,
+		      struct perf_evsel *evsel, enum rstype rstype);
+void res_sample_init(void);
 #else
 static inline
 int perf_evlist__tui_browse_hists(struct perf_evlist *evlist __maybe_unused,
@@ -488,6 +500,16 @@ static inline int script_browse(const char *script_opt __maybe_unused,
 	return 0;
 }
 
+static inline int res_sample_browse(struct res_sample *res_samples __maybe_unused,
+				    int num_res __maybe_unused,
+				    struct perf_evsel *evsel __maybe_unused,
+				    enum rstype rstype __maybe_unused)
+{
+	return 0;
+}
+
+static inline void res_sample_init(void) {}
+
 #define K_LEFT  -1000
 #define K_RIGHT -2000
 #define K_SWITCH_INPUT_DATA -3000
diff --git a/tools/perf/util/sort.h b/tools/perf/util/sort.h
index 19dceb7f6145..bb9442ab7a0c 100644
--- a/tools/perf/util/sort.h
+++ b/tools/perf/util/sort.h
@@ -47,6 +47,12 @@ extern struct sort_entry sort_srcline;
 extern enum sort_type sort__first_dimension;
 extern const char default_mem_sort_order[];
 
+struct res_sample {
+	u64 time;
+	int cpu;
+	int tid;
+};
+
 struct he_stat {
 	u64			period;
 	u64			period_sys;
@@ -140,6 +146,8 @@ struct hist_entry {
 	struct mem_info		*mem_info;
 	void			*raw_data;
 	u32			raw_size;
+	int			num_res;
+	struct res_sample	*res_samples;
 	void			*trace_output;
 	struct perf_hpp_list	*hpp_list;
 	struct hist_entry	*parent_he;
diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c
index 6b73a0eeb6a1..58442ca5e3c4 100644
--- a/tools/perf/util/symbol.c
+++ b/tools/perf/util/symbol.c
@@ -51,6 +51,7 @@ struct symbol_conf symbol_conf = {
 	.symfs			= "",
 	.event_group		= true,
 	.inline_name		= true,
+	.res_sample		= 0,
 };
 
 static enum dso_binary_type binary_type_symtab[] = {
diff --git a/tools/perf/util/symbol_conf.h b/tools/perf/util/symbol_conf.h
index a5684a71b78e..6c55fa6fccec 100644
--- a/tools/perf/util/symbol_conf.h
+++ b/tools/perf/util/symbol_conf.h
@@ -68,6 +68,7 @@ struct symbol_conf {
 	struct intlist	*pid_list,
 			*tid_list;
 	const char	*symfs;
+	int		res_sample;
 };
 
 extern struct symbol_conf symbol_conf;