Merge tag 'perf-core-for-mingo-5.7-20200310' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/core

Pull perf/core improvements and fixes from Arnaldo Carvalho de Melo:

perf stat:

  Jin Yao:

  - Show percore counts in per CPU output.

perf report:

  Jin Yao:

  - Allow selecting which block info columns to report and its order.

  - Support color ops to print block percents in color.

  - Fix wrong block address comparison in block_info__cmp().

perf annotate:

  Ravi Bangoria:

  - Get rid of annotation->nr_jumps, unused.

expr:

  Jiri Olsa:

  - Move expr lexer to flex.

llvm:

  Arnaldo Carvalho de Melo:

  - Add debug hint message about missing kernel-devel package.

core:

  Kan Liang:

  - Initial patches to support the recently added PERF_SAMPLE_BRANCH_HW_INDEX
    kernel feature.

  - Add check for unexpected use of reserved membrs in event attr, so that in
    the future older perf tools will complain instead of silently try to process
    unknown features.

libapi:

  Namhyung Kim:

  - Adopt cgroupsfs_find_mountpoint() from tools/perf/util/.

libperf:

  Michael Petlan:

  - Add counting example.

libtraceevent:

   Steven Rostedt (VMware):

  - Remove extra '\n' in print_event_time().

Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
Signed-off-by: Ingo Molnar <[email protected]>

Author: Ingo Molnar

tools/include/uapi/linux/perf_event.h

@@ -181,6 +181,8 @@ enum perf_branch_sample_type_shift {
 
 	PERF_SAMPLE_BRANCH_TYPE_SAVE_SHIFT	= 16, /* save branch type */
 
+	PERF_SAMPLE_BRANCH_HW_INDEX_SHIFT	= 17, /* save low level index of raw branch records */
+
 	PERF_SAMPLE_BRANCH_MAX_SHIFT		/* non-ABI */
 };
 
@@ -208,6 +210,8 @@ enum perf_branch_sample_type {
 	PERF_SAMPLE_BRANCH_TYPE_SAVE	=
 		1U << PERF_SAMPLE_BRANCH_TYPE_SAVE_SHIFT,
 
+	PERF_SAMPLE_BRANCH_HW_INDEX	= 1U << PERF_SAMPLE_BRANCH_HW_INDEX_SHIFT,
+
 	PERF_SAMPLE_BRANCH_MAX		= 1U << PERF_SAMPLE_BRANCH_MAX_SHIFT,
 };
 
@@ -853,7 +857,9 @@ enum perf_event_type {
 	 *	  char                  data[size];}&& PERF_SAMPLE_RAW
 	 *
 	 *	{ u64                   nr;
-	 *        { u64 from, to, flags } lbr[nr];} && PERF_SAMPLE_BRANCH_STACK
+	 *	  { u64	hw_idx; } && PERF_SAMPLE_BRANCH_HW_INDEX
+	 *        { u64 from, to, flags } lbr[nr];
+	 *      } && PERF_SAMPLE_BRANCH_STACK
 	 *
 	 * 	{ u64			abi; # enum perf_sample_regs_abi
 	 * 	  u64			regs[weight(mask)]; } && PERF_SAMPLE_REGS_USER

tools/lib/api/fs/Build

@@ -1,2 +1,3 @@
 libapi-y += fs.o
 libapi-y += tracing_path.o
+libapi-y += cgroup.o

tools/lib/api/fs/cgroup.c

@@ -0,0 +1,67 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <linux/stringify.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <fcntl.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include "fs.h"
+
+int cgroupfs_find_mountpoint(char *buf, size_t maxlen, const char *subsys)
+{
+	FILE *fp;
+	char mountpoint[PATH_MAX + 1], tokens[PATH_MAX + 1], type[PATH_MAX + 1];
+	char path_v1[PATH_MAX + 1], path_v2[PATH_MAX + 2], *path;
+	char *token, *saved_ptr = NULL;
+
+	fp = fopen("/proc/mounts", "r");
+	if (!fp)
+		return -1;
+
+	/*
+	 * in order to handle split hierarchy, we need to scan /proc/mounts
+	 * and inspect every cgroupfs mount point to find one that has
+	 * perf_event subsystem
+	 */
+	path_v1[0] = '\0';
+	path_v2[0] = '\0';
+
+	while (fscanf(fp, "%*s %"__stringify(PATH_MAX)"s %"__stringify(PATH_MAX)"s %"
+				__stringify(PATH_MAX)"s %*d %*d\n",
+				mountpoint, type, tokens) == 3) {
+
+		if (!path_v1[0] && !strcmp(type, "cgroup")) {
+
+			token = strtok_r(tokens, ",", &saved_ptr);
+
+			while (token != NULL) {
+				if (subsys && !strcmp(token, subsys)) {
+					strcpy(path_v1, mountpoint);
+					break;
+				}
+				token = strtok_r(NULL, ",", &saved_ptr);
+			}
+		}
+
+		if (!path_v2[0] && !strcmp(type, "cgroup2"))
+			strcpy(path_v2, mountpoint);
+
+		if (path_v1[0] && path_v2[0])
+			break;
+	}
+	fclose(fp);
+
+	if (path_v1[0])
+		path = path_v1;
+	else if (path_v2[0])
+		path = path_v2;
+	else
+		return -1;
+
+	if (strlen(path) < maxlen) {
+		strcpy(buf, path);
+		return 0;
+	}
+	return -1;
+}

tools/lib/api/fs/fs.h

@@ -28,6 +28,8 @@ FS(bpf_fs)
 #undef FS
 
 
+int cgroupfs_find_mountpoint(char *buf, size_t maxlen, const char *subsys);
+
 int filename__read_int(const char *filename, int *value);
 int filename__read_ull(const char *filename, unsigned long long *value);
 int filename__read_xll(const char *filename, unsigned long long *value);

tools/lib/perf/Documentation/examples/counting.c

@@ -0,0 +1,83 @@
+#include <linux/perf_event.h>
+#include <perf/evlist.h>
+#include <perf/evsel.h>
+#include <perf/cpumap.h>
+#include <perf/threadmap.h>
+#include <perf/mmap.h>
+#include <perf/core.h>
+#include <perf/event.h>
+#include <stdio.h>
+#include <unistd.h>
+
+static int libperf_print(enum libperf_print_level level,
+                         const char *fmt, va_list ap)
+{
+	return vfprintf(stderr, fmt, ap);
+}
+
+int main(int argc, char **argv)
+{
+	int count = 100000, err = 0;
+	struct perf_evlist *evlist;
+	struct perf_evsel *evsel;
+	struct perf_thread_map *threads;
+	struct perf_counts_values counts;
+
+	struct perf_event_attr attr1 = {
+		.type        = PERF_TYPE_SOFTWARE,
+		.config      = PERF_COUNT_SW_CPU_CLOCK,
+		.read_format = PERF_FORMAT_TOTAL_TIME_ENABLED|PERF_FORMAT_TOTAL_TIME_RUNNING,
+		.disabled    = 1,
+	};
+	struct perf_event_attr attr2 = {
+		.type        = PERF_TYPE_SOFTWARE,
+		.config      = PERF_COUNT_SW_TASK_CLOCK,
+		.read_format = PERF_FORMAT_TOTAL_TIME_ENABLED|PERF_FORMAT_TOTAL_TIME_RUNNING,
+		.disabled    = 1,
+	};
+
+	libperf_init(libperf_print);
+	threads = perf_thread_map__new_dummy();
+	if (!threads) {
+		fprintf(stderr, "failed to create threads\n");
+		return -1;
+	}
+	perf_thread_map__set_pid(threads, 0, 0);
+	evlist = perf_evlist__new();
+	if (!evlist) {
+		fprintf(stderr, "failed to create evlist\n");
+		goto out_threads;
+	}
+	evsel = perf_evsel__new(&attr1);
+	if (!evsel) {
+		fprintf(stderr, "failed to create evsel1\n");
+		goto out_evlist;
+	}
+	perf_evlist__add(evlist, evsel);
+	evsel = perf_evsel__new(&attr2);
+	if (!evsel) {
+		fprintf(stderr, "failed to create evsel2\n");
+		goto out_evlist;
+	}
+	perf_evlist__add(evlist, evsel);
+	perf_evlist__set_maps(evlist, NULL, threads);
+	err = perf_evlist__open(evlist);
+	if (err) {
+		fprintf(stderr, "failed to open evsel\n");
+		goto out_evlist;
+	}
+	perf_evlist__enable(evlist);
+	while (count--);
+	perf_evlist__disable(evlist);
+	perf_evlist__for_each_evsel(evlist, evsel) {
+		perf_evsel__read(evsel, 0, 0, &counts);
+		fprintf(stdout, "count %llu, enabled %llu, run %llu\n",
+				counts.val, counts.ena, counts.run);
+	}
+	perf_evlist__close(evlist);
+out_evlist:
+	perf_evlist__delete(evlist);
+out_threads:
+	perf_thread_map__put(threads);
+	return err;
+}

tools/lib/traceevent/event-parse.c

@@ -5541,7 +5541,7 @@ static void print_event_time(struct tep_handle *tep, struct trace_seq *s,
 	if (p10 > 1 && p10 < time)
 		trace_seq_printf(s, "%5llu.%0*llu", time / p10, prec, time % p10);
 	else
-		trace_seq_printf(s, "%12llu\n", time);
+		trace_seq_printf(s, "%12llu", time);
 }
 
 struct print_event_type {

tools/perf/Documentation/perf-stat.txt

@@ -334,6 +334,15 @@ Configure all used events to run in kernel space.
 --all-user::
 Configure all used events to run in user space.
 
+--percore-show-thread::
+The event modifier "percore" has supported to sum up the event counts
+for all hardware threads in a core and show the counts per core.
+
+This option with event modifier "percore" enabled also sums up the event
+counts for all hardware threads in a core but show the sum counts per
+hardware thread. This is essentially a replacement for the any bit and
+convenient for post processing.
+
 EXAMPLES
 --------
 

tools/perf/builtin-diff.c

@@ -572,37 +572,20 @@ static void init_block_hist(struct block_hist *bh)
 	bh->valid = true;
 }
 
-static int block_pair_cmp(struct hist_entry *a, struct hist_entry *b)
-{
-	struct block_info *bi_a = a->block_info;
-	struct block_info *bi_b = b->block_info;
-	int cmp;
-
-	if (!bi_a->sym || !bi_b->sym)
-		return -1;
-
-	cmp = strcmp(bi_a->sym->name, bi_b->sym->name);
-
-	if ((!cmp) && (bi_a->start == bi_b->start) && (bi_a->end == bi_b->end))
-		return 0;
-
-	return -1;
-}
-
 static struct hist_entry *get_block_pair(struct hist_entry *he,
 					 struct hists *hists_pair)
 {
 	struct rb_root_cached *root = hists_pair->entries_in;
 	struct rb_node *next = rb_first_cached(root);
-	int cmp;
+	int64_t cmp;
 
 	while (next != NULL) {
 		struct hist_entry *he_pair = rb_entry(next, struct hist_entry,
 						      rb_node_in);
 
 		next = rb_next(&he_pair->rb_node_in);
 
-		cmp = block_pair_cmp(he_pair, he);
+		cmp = __block_info__cmp(he_pair, he);
 		if (!cmp)
 			return he_pair;
 	}
@@ -1312,7 +1295,8 @@ static int cycles_printf(struct hist_entry *he, struct hist_entry *pair,
 	end_line = map__srcline(he->ms.map, bi->sym->start + bi->end,
 				he->ms.sym);
 
-	if ((start_line != SRCLINE_UNKNOWN) && (end_line != SRCLINE_UNKNOWN)) {
+	if ((strncmp(start_line, SRCLINE_UNKNOWN, strlen(SRCLINE_UNKNOWN)) != 0) &&
+	    (strncmp(end_line, SRCLINE_UNKNOWN, strlen(SRCLINE_UNKNOWN)) != 0)) {
 		scnprintf(buf, sizeof(buf), "[%s -> %s] %4ld",
 			  start_line, end_line, block_he->diff.cycles);
 	} else {

tools/perf/builtin-report.c

@@ -104,6 +104,7 @@ struct report {
 	bool			symbol_ipc;
 	bool			total_cycles_mode;
 	struct block_report	*block_reports;
+	int			nr_block_reports;
 };
 
 static int report__config(const char *var, const char *value, void *cb)
@@ -966,8 +967,19 @@ static int __cmd_report(struct report *rep)
 	report__output_resort(rep);
 
 	if (rep->total_cycles_mode) {
+		int block_hpps[6] = {
+			PERF_HPP_REPORT__BLOCK_TOTAL_CYCLES_PCT,
+			PERF_HPP_REPORT__BLOCK_LBR_CYCLES,
+			PERF_HPP_REPORT__BLOCK_CYCLES_PCT,
+			PERF_HPP_REPORT__BLOCK_AVG_CYCLES,
+			PERF_HPP_REPORT__BLOCK_RANGE,
+			PERF_HPP_REPORT__BLOCK_DSO,
+		};
+
 		rep->block_reports = block_info__create_report(session->evlist,
-							       rep->total_cycles);
+							       rep->total_cycles,
+							       block_hpps, 6,
+							       &rep->nr_block_reports);
 		if (!rep->block_reports)
 			return -1;
 	}
@@ -1551,8 +1563,11 @@ int cmd_report(int argc, const char **argv)
 		zfree(&report.ptime_range);
 	}
 
-	if (report.block_reports)
-		zfree(&report.block_reports);
+	if (report.block_reports) {
+		block_info__free_report(report.block_reports,
+					report.nr_block_reports);
+		report.block_reports = NULL;
+	}
 
 	zstd_fini(&(session->zstd_data));
 	perf_session__delete(session);