Skip to contentSkip to conversation

perf/core improvements and fixes

GIT PULL New Arnaldo Carvalho de Melo 2 participants 20 messages Dec 11, 2014 – Dec 12, 2014

Subsystems (3)
lore
Share & tools
lore

Arnaldo Carvalho de Melo

author maintainer
#0 lore raw

Hi Ingo,

Please consider pulling,

- Arnaldo

The following changes since commit cfa0bd52d0ba9b852f76c7b3f1055edd5e5c7846:

Merge tag 'perf-core-for-mingo-2' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/core (2014-12-08 07:45:45 +0100)

are available in the git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux.git tags/perf-core-for-mingo

for you to fetch changes up to e09b18d4907992d3d615b215c1abf585721b2810:

perf trace: Provide a better explanation when mmap fails (2014-12-11 18:04:10 -0300)

----------------------------------------------------------------
perf/core improvements and fixes:

User visible:

- Mark events as (x86 only) in help output for 'perf kvm stat live" (Alexander Yarygin)

- Provide a better explanation when mmap fails in 'trace' (Arnaldo Carvalho de Melo)

- Add --buildid-dir option to set cache directory, i.e. use:

      $ perf --buildid-dir /path/to/dir tool --tool-options

(Jiri Olsa)

- Fix memcpy/memset 'perf bench' output (Rabin Vicent)

- Fix 'perf test' attr tests size values to cope with machine state on interrupt ABI changes (Jiri Olsa)

- Fixup callchain type parameter handling error message (Kan Liang)

Infrastructure/cleanups:

- calloc/xcalloc: Fix argument order (Arjun Sreedharan)

- Move filename__read_int from tools/perf/ to tools/lib, add sysctl__read_int there and use it in place of ad-hoc copies (Arnaldo Carvalho de Melo)

- Use single strcmp call instead of two (Jiri Olsa)

- Remove extra debugdir variables in 'perf buildid-cache' (Jiri Olsa)

- Fix -a segfault related to kcore handling in 'perf buildid-cache' (Jiri Olsa)

- Move cpumode resolve code to add_callchain_ip (Kan Liang)

- Merge memset into memcpy 'perf bench' (Rabin Vincent)

- Change print format from %lu to %PRIu64 in the hists browser (Tom Huynh)

----------------------------------------------------------------
Alexander Yarygin (1):
perf kvm stat live: Mark events as (x86 only) in help output

Arjun Sreedharan (1):
calloc/xcalloc: Fix argument order

Arnaldo Carvalho de Melo (5):
tools lib fs: Adopt filename__read_int from tools/perf/ tools lib fs: Add sysctl__read_int helper
perf tools: Use sysctl__read_int instead of ad-hoc copies perf evlist: Introduce strerror_mmap method perf trace: Provide a better explanation when mmap fails

Jiri Olsa (5):
perf tools: Use single strcmp call instead of two
perf buildid-cache: Remove extra debugdir variables
perf buildid cache: Fix -a segfault related to kcore handling
perf tools: Add --buildid-dir option to set cache directory
perf tests: Fix attr tests size values to cope with machine state on interrupt ABI changes

Kan Liang (2):
perf callchain: Fixup parameter handling error message perf callchain: Move cpumode resolve code to add_callchain_ip

Rabin Vincent (3):
perf bench: Prepare memcpy for merge
perf bench: Merge memset into memcpy
perf bench: Fix memcpy/memset output

Tom Huynh (1):
perf hists browser: Change print format from %lu to %PRIu64

 scripts/kconfig/mconf.c            |   4 +-
 tools/lib/api/fs/fs.c              |  34 +++++
 tools/lib/api/fs/fs.h              |   3 +
 tools/perf/Documentation/perf.txt  |   4 +
 tools/perf/Makefile.perf           |   1 -
 tools/perf/bench/mem-memcpy.c      | 286 ++++++++++++++++++++++++----------
 tools/perf/bench/mem-memset.c      | 304 -------------------------------------
 tools/perf/builtin-buildid-cache.c |  13 +-
 tools/perf/builtin-kvm.c           |   3 +-
 tools/perf/builtin-trace.c         |  12 +-
 tools/perf/perf.c                  |  14 +-
 tools/perf/tests/attr/base-record  |   2 +-
 tools/perf/tests/attr/base-stat    |   2 +-
 tools/perf/ui/browsers/hists.c     |   2 +-
 tools/perf/ui/hist.c               |   4 +-
 tools/perf/util/build-id.c         |   9 +-
 tools/perf/util/callchain.c        |   2 +-
 tools/perf/util/config.c           |  10 +-
 tools/perf/util/evlist.c           |  23 +++
 tools/perf/util/evlist.h           |   1 +
 tools/perf/util/machine.c          |  72 +++++----
 tools/perf/util/record.c           |  11 +-
 tools/perf/util/util.c             |  26 +---
 tools/perf/util/util.h             |   3 +-
 tools/thermal/tmon/sysfs.c         |   6 +-
 25 files changed, 350 insertions(+), 501 deletions(-)
 delete mode 100644 tools/perf/bench/mem-memset.c
author maintainer
#1 lore raw

PATCH 1/18 perf bench: Prepare memcpy for merge

From: Rabin Vincent <[email protected]>

The memset benchmark is largely copy-pasted from the memcpy benchmark. Prepare the memcpy file for merge with memset by extracting out a generic function.

---
 tools/perf/bench/mem-memcpy.c | 182 ++++++++++++++++++++++++------------------
 1 file changed, 104 insertions(+), 78 deletions(-)
tools/perf/bench/mem-memcpy.c +104 -78
index 2465141b554b..4a4493a7a1f8 100644
@@ -52,16 +52,18 @@ typedef void *(*memcpy_t)(void *, const void *, size_t);
 struct routine {
 	const char *name;
 	const char *desc;
-	memcpy_t fn;
+	union {
+		memcpy_t memcpy;
+	} fn;
 };
 
-struct routine routines[] = {
-	{ "default",
-	  "Default memcpy() provided by glibc",
-	  memcpy },
+struct routine memcpy_routines[] = {
+	{ .name = "default",
+	  .desc = "Default memcpy() provided by glibc",
+	  .fn.memcpy = memcpy },
 #ifdef HAVE_ARCH_X86_64_SUPPORT
 
-#define MEMCPY_FN(fn, name, desc) { name, desc, fn },
+#define MEMCPY_FN(_fn, _name, _desc) {.name = _name, .desc = _desc, .fn.memcpy = _fn},
 #include "mem-memcpy-x86-64-asm-def.h"
 #undef MEMCPY_FN
 
@@ -69,7 +71,7 @@ struct routine routines[] = {
 
 	{ NULL,
 	  NULL,
-	  NULL   }
+	  {NULL}   }
 };
 
 static const char * const bench_mem_memcpy_usage[] = {
@@ -110,63 +112,6 @@ static double timeval2double(struct timeval *ts)
 		(double)ts->tv_usec / (double)1000000;
 }
 
-static void alloc_mem(void **dst, void **src, size_t length)
-{
-	*dst = zalloc(length);
-	if (!*dst)
-		die("memory allocation failed - maybe length is too large?\n");
-
-	*src = zalloc(length);
-	if (!*src)
-		die("memory allocation failed - maybe length is too large?\n");
-	/* Make sure to always replace the zero pages even if MMAP_THRESH is crossed */
-	memset(*src, 0, length);
-}
-
-static u64 do_memcpy_cycle(memcpy_t fn, size_t len, bool prefault)
-{
-	u64 cycle_start = 0ULL, cycle_end = 0ULL;
-	void *src = NULL, *dst = NULL;
-	int i;
-
-	alloc_mem(&src, &dst, len);
-
-	if (prefault)
-		fn(dst, src, len);
-
-	cycle_start = get_cycle();
-	for (i = 0; i < iterations; ++i)
-		fn(dst, src, len);
-	cycle_end = get_cycle();
-
-	free(src);
-	free(dst);
-	return cycle_end - cycle_start;
-}
-
-static double do_memcpy_gettimeofday(memcpy_t fn, size_t len, bool prefault)
-{
-	struct timeval tv_start, tv_end, tv_diff;
-	void *src = NULL, *dst = NULL;
-	int i;
-
-	alloc_mem(&src, &dst, len);
-
-	if (prefault)
-		fn(dst, src, len);
-
-	BUG_ON(gettimeofday(&tv_start, NULL));
-	for (i = 0; i < iterations; ++i)
-		fn(dst, src, len);
-	BUG_ON(gettimeofday(&tv_end, NULL));
-
-	timersub(&tv_end, &tv_start, &tv_diff);
-
-	free(src);
-	free(dst);
-	return (double)((double)len / timeval2double(&tv_diff));
-}
-
 #define pf (no_prefault ? 0 : 1)
 
 #define print_bps(x) do {					\
@@ -180,8 +125,16 @@ static double do_memcpy_gettimeofday(memcpy_t fn, size_t len, bool prefault)
 			printf(" %14lf GB/Sec", x / K / K / K); \
 	} while (0)
 
-int bench_mem_memcpy(int argc, const char **argv,
-		     const char *prefix __maybe_unused)
+struct bench_mem_info {
+	const struct routine *routines;
+	u64 (*do_cycle)(const struct routine *r, size_t len, bool prefault);
+	double (*do_gettimeofday)(const struct routine *r, size_t len, bool prefault);
+	const char *const *usage;
+};
+
+static int bench_mem_common(int argc, const char **argv,
+		     const char *prefix __maybe_unused,
+		     struct bench_mem_info *info)
 {
 	int i;
 	size_t len;
@@ -189,7 +142,7 @@ int bench_mem_memcpy(int argc, const char **argv,
 	u64 result_cycle[2];
 
 	argc = parse_options(argc, argv, options,
-			     bench_mem_memcpy_usage, 0);
+			     info->usage, 0);
 
 	if (no_prefault && only_prefault) {
 		fprintf(stderr, "Invalid options: -o and -n are mutually exclusive\n");
@@ -213,16 +166,16 @@ int bench_mem_memcpy(int argc, const char **argv,
 	if (only_prefault && no_prefault)
 		only_prefault = no_prefault = false;
 
-	for (i = 0; routines[i].name; i++) {
-		if (!strcmp(routines[i].name, routine))
+	for (i = 0; info->routines[i].name; i++) {
+		if (!strcmp(info->routines[i].name, routine))
 			break;
 	}
-	if (!routines[i].name) {
+	if (!info->routines[i].name) {
 		printf("Unknown routine:%s\n", routine);
 		printf("Available routines...\n");
-		for (i = 0; routines[i].name; i++) {
+		for (i = 0; info->routines[i].name; i++) {
 			printf("\t%s ... %s\n",
-			       routines[i].name, routines[i].desc);
+			       info->routines[i].name, info->routines[i].desc);
 		}
 		return 1;
 	}
@@ -234,25 +187,25 @@ int bench_mem_memcpy(int argc, const char **argv,
 		/* show both of results */
 		if (use_cycle) {
 			result_cycle[0] =
-				do_memcpy_cycle(routines[i].fn, len, false);
+				info->do_cycle(&info->routines[i], len, false);
 			result_cycle[1] =
-				do_memcpy_cycle(routines[i].fn, len, true);
+				info->do_cycle(&info->routines[i], len, true);
 		} else {
 			result_bps[0] =
-				do_memcpy_gettimeofday(routines[i].fn,
+				info->do_gettimeofday(&info->routines[i],
 						len, false);
 			result_bps[1] =
-				do_memcpy_gettimeofday(routines[i].fn,
+				info->do_gettimeofday(&info->routines[i],
 						len, true);
 		}
 	} else {
 		if (use_cycle) {
 			result_cycle[pf] =
-				do_memcpy_cycle(routines[i].fn,
+				info->do_cycle(&info->routines[i],
 						len, only_prefault);
 		} else {
 			result_bps[pf] =
-				do_memcpy_gettimeofday(routines[i].fn,
+				info->do_gettimeofday(&info->routines[i],
 						len, only_prefault);
 		}
 	}
@@ -310,3 +263,76 @@ int bench_mem_memcpy(int argc, const char **argv,
 
 	return 0;
 }
+
+static void memcpy_alloc_mem(void **dst, void **src, size_t length)
+{
+	*dst = zalloc(length);
+	if (!*dst)
+		die("memory allocation failed - maybe length is too large?\n");
+
+	*src = zalloc(length);
+	if (!*src)
+		die("memory allocation failed - maybe length is too large?\n");
+	/* Make sure to always replace the zero pages even if MMAP_THRESH is crossed */
+	memset(*src, 0, length);
+}
+
+static u64 do_memcpy_cycle(const struct routine *r, size_t len, bool prefault)
+{
+	u64 cycle_start = 0ULL, cycle_end = 0ULL;
+	void *src = NULL, *dst = NULL;
+	memcpy_t fn = r->fn.memcpy;
+	int i;
+
+	memcpy_alloc_mem(&src, &dst, len);
+
+	if (prefault)
+		fn(dst, src, len);
+
+	cycle_start = get_cycle();
+	for (i = 0; i < iterations; ++i)
+		fn(dst, src, len);
+	cycle_end = get_cycle();
+
+	free(src);
+	free(dst);
+	return cycle_end - cycle_start;
+}
+
+static double do_memcpy_gettimeofday(const struct routine *r, size_t len,
+				     bool prefault)
+{
+	struct timeval tv_start, tv_end, tv_diff;
+	memcpy_t fn = r->fn.memcpy;
+	void *src = NULL, *dst = NULL;
+	int i;
+
+	memcpy_alloc_mem(&src, &dst, len);
+
+	if (prefault)
+		fn(dst, src, len);
+
+	BUG_ON(gettimeofday(&tv_start, NULL));
+	for (i = 0; i < iterations; ++i)
+		fn(dst, src, len);
+	BUG_ON(gettimeofday(&tv_end, NULL));
+
+	timersub(&tv_end, &tv_start, &tv_diff);
+
+	free(src);
+	free(dst);
+	return (double)((double)len / timeval2double(&tv_diff));
+}
+
+int bench_mem_memcpy(int argc, const char **argv,
+		     const char *prefix __maybe_unused)
+{
+	struct bench_mem_info info = {
+		.routines = memcpy_routines,
+		.do_cycle = do_memcpy_cycle,
+		.do_gettimeofday = do_memcpy_gettimeofday,
+		.usage = bench_mem_memcpy_usage,
+	};
+
+	return bench_mem_common(argc, argv, prefix, &info);
+}
Signature · 1 line
1.9.3
author maintainer
#2 lore raw

PATCH 2/18 perf bench: Merge memset into memcpy

In reply to Arnaldo Carvalho de Melo

From: Rabin Vincent <[email protected]>

The memset benchmark is largely copy-pasted from the memcpy benchmark. Merge the two now that memcpy is made more generic.

---
 tools/perf/Makefile.perf      |   1 -
 tools/perf/bench/mem-memcpy.c |  90 +++++++++++++
 tools/perf/bench/mem-memset.c | 304 ------------------------------------------
 3 files changed, 90 insertions(+), 305 deletions(-)
 delete mode 100644 tools/perf/bench/mem-memset.c
tools/perf/Makefile.perf +0 -1
index 478efa9b2364..763e68fb5767 100644
@@ -458,7 +458,6 @@ BUILTIN_OBJS += $(OUTPUT)bench/mem-memcpy-x86-64-asm.o
 BUILTIN_OBJS += $(OUTPUT)bench/mem-memset-x86-64-asm.o
 endif
 BUILTIN_OBJS += $(OUTPUT)bench/mem-memcpy.o
-BUILTIN_OBJS += $(OUTPUT)bench/mem-memset.o
 BUILTIN_OBJS += $(OUTPUT)bench/futex-hash.o
 BUILTIN_OBJS += $(OUTPUT)bench/futex-wake.o
 BUILTIN_OBJS += $(OUTPUT)bench/futex-requeue.o
tools/perf/bench/mem-memcpy.c +90 -0

111 lines not on this page · raw message

tools/perf/bench/mem-memset.c deleted +0 -304

304 lines not on this page · raw message

Signature · 1 line
1.9.3
author maintainer
#3 lore raw

PATCH 3/18 perf bench: Fix memcpy/memset output

In reply to Arnaldo Carvalho de Melo

From: Rabin Vincent <[email protected]>

The memcpy and memset benchmarks return bogus results when iterations > 0 because the iterations value is not taken into account when calculating the final result:

 $ perf bench mem memset --only-prefault --length 1GB --iterations 1
 # Running 'mem/memset' benchmark:
 # Copying 1GB Bytes ...
       20.798669 GB/Sec (with prefault)
 $ perf bench mem memset --only-prefault --length 1GB --iterations 10
 # Running 'mem/memset' benchmark:
 # Copying 1GB Bytes ...
        2.086576 GB/Sec (with prefault)
 $ perf bench mem memset --only-prefault --length 1GB --iterations 100
 # Running 'mem/memset' benchmark:
 # Copying 1GB Bytes ...
      212.840917 MB/Sec (with prefault)

Fix this.

---
 tools/perf/bench/mem-memcpy.c | 18 ++++++++++--------
 1 file changed, 10 insertions(+), 8 deletions(-)
tools/perf/bench/mem-memcpy.c +10 -8
index e18be70c8a47..6c14afe8c1b1 100644
@@ -141,6 +141,7 @@ static int bench_mem_common(int argc, const char **argv,
 {
 	int i;
 	size_t len;
+	double totallen;
 	double result_bps[2];
 	u64 result_cycle[2];
 
@@ -156,6 +157,7 @@ static int bench_mem_common(int argc, const char **argv,
 		init_cycle();
 
 	len = (size_t)perf_atoll((char *)length_str);
+	totallen = (double)len * iterations;
 
 	result_cycle[0] = result_cycle[1] = 0ULL;
 	result_bps[0] = result_bps[1] = 0.0;
@@ -219,10 +221,10 @@ static int bench_mem_common(int argc, const char **argv,
 			if (use_cycle) {
 				printf(" %14lf Cycle/Byte\n",
 					(double)result_cycle[0]
-					/ (double)len);
+					/ totallen);
 				printf(" %14lf Cycle/Byte (with prefault)\n",
 					(double)result_cycle[1]
-					/ (double)len);
+					/ totallen);
 			} else {
 				print_bps(result_bps[0]);
 				printf("\n");
@@ -233,7 +235,7 @@ static int bench_mem_common(int argc, const char **argv,
 			if (use_cycle) {
 				printf(" %14lf Cycle/Byte",
 					(double)result_cycle[pf]
-					/ (double)len);
+					/ totallen);
 			} else
 				print_bps(result_bps[pf]);
 
@@ -244,8 +246,8 @@ static int bench_mem_common(int argc, const char **argv,
 		if (!only_prefault && !no_prefault) {
 			if (use_cycle) {
 				printf("%lf %lf\n",
-					(double)result_cycle[0] / (double)len,
-					(double)result_cycle[1] / (double)len);
+					(double)result_cycle[0] / totallen,
+					(double)result_cycle[1] / totallen);
 			} else {
 				printf("%lf %lf\n",
 					result_bps[0], result_bps[1]);
@@ -253,7 +255,7 @@ static int bench_mem_common(int argc, const char **argv,
 		} else {
 			if (use_cycle) {
 				printf("%lf\n", (double)result_cycle[pf]
-					/ (double)len);
+					/ totallen);
 			} else
 				printf("%lf\n", result_bps[pf]);
 		}
@@ -324,7 +326,7 @@ static double do_memcpy_gettimeofday(const struct routine *r, size_t len,
 
 	free(src);
 	free(dst);
-	return (double)((double)len / timeval2double(&tv_diff));
+	return (double)(((double)len * iterations) / timeval2double(&tv_diff));
 }
 
 int bench_mem_memcpy(int argc, const char **argv,
@@ -389,7 +391,7 @@ static double do_memset_gettimeofday(const struct routine *r, size_t len,
 	timersub(&tv_end, &tv_start, &tv_diff);
 
 	free(dst);
-	return (double)((double)len / timeval2double(&tv_diff));
+	return (double)(((double)len * iterations) / timeval2double(&tv_diff));
 }
 
 static const char * const bench_mem_memset_usage[] = {
Signature · 1 line
1.9.3
author maintainer
#4 lore raw

PATCH 4/18 perf hists browser: Change print format from %lu to %PRIu64

In reply to Arnaldo Carvalho de Melo

From: Tom Huynh <[email protected]>

The nr_events variable in tools/perf/ui/browsers/hists.c is of type u64, so the print format (%lu) causes 'perf report' to show 0 event count when running with 32-bit userspace without redirection.

This patch fixes that problem by printing nr_events as PRIu64.

---
 tools/perf/ui/browsers/hists.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/ui/browsers/hists.c +1 -1
index 502daff76ceb..e6bb04b5b09b 100644
@@ -1252,7 +1252,7 @@ static int hists__browser_title(struct hists *hists,
 
 	nr_samples = convert_unit(nr_samples, &unit);
 	printed = scnprintf(bf, size,
-			   "Samples: %lu%c of event '%s', Event count (approx.): %lu",
+			   "Samples: %lu%c of event '%s', Event count (approx.): %" PRIu64,
 			   nr_samples, unit, ev_name, nr_events);
 
 
Signature · 1 line
1.9.3
author maintainer
#5 lore raw

PATCH 5/18 perf tools: Use single strcmp call instead of two

In reply to Arnaldo Carvalho de Melo

From: Jiri Olsa <[email protected]>

There's no need to use 2 strcmp calls, one is enough.

Cc Corey Ashford <[email protected]>
Cc David Ahern <[email protected]>
Cc Frederic Weisbecker <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc Namhyung Kim <[email protected]>
Cc Paul Mackerras <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Stephane Eranian <[email protected]>
Cc Steven Rostedt <[email protected]>
---
 tools/perf/util/config.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/util/config.c +1 -1
index 57ff826f150b..c802236b2bfe 100644
@@ -522,7 +522,7 @@ static int buildid_dir_command_config(const char *var, const char *value,
 	const char *v;
 
 	/* same dir for all commands */
-	if (!prefixcmp(var, "buildid.") && !strcmp(var + 8, "dir")) {
+	if (!strcmp(var, "buildid.dir")) {
 		v = perf_config_dirname(var, value);
 		if (!v)
 			return -1;
Signature · 1 line
1.9.3
author maintainer
#6 lore raw

PATCH 6/18 perf buildid-cache: Remove extra debugdir variables

In reply to Arnaldo Carvalho de Melo

From: Jiri Olsa <[email protected]>

There's no need to copy over the buildid_dir into separate variable with no change.

This is leftover from commit:
45de34bbe3e1 perf buildid: add perfconfig option to specify buildid cache dir

that added global buildid_dir variable that holds cache directory, but did not cleanup the debugdir copies.

Cc Corey Ashford <[email protected]>
Cc David Ahern <[email protected]>
Cc Frederic Weisbecker <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc Namhyung Kim <[email protected]>
Cc Paul Mackerras <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Stephane Eranian <[email protected]>
Cc Steven Rostedt <[email protected]>
---
 tools/perf/builtin-buildid-cache.c | 11 ++++-------
 tools/perf/util/build-id.c         |  9 +++------
 2 files changed, 7 insertions(+), 13 deletions(-)
tools/perf/builtin-buildid-cache.c +4 -7
index 70385756da63..29f24c071bc6 100644
@@ -285,7 +285,6 @@ int cmd_buildid_cache(int argc, const char **argv,
 	struct str_node *pos;
 	int ret = 0;
 	bool force = false;
-	char debugdir[PATH_MAX];
 	char const *add_name_list_str = NULL,
 		   *remove_name_list_str = NULL,
 		   *missing_filename = NULL,
@@ -335,13 +334,11 @@ int cmd_buildid_cache(int argc, const char **argv,
 
 	setup_pager();
 
-	snprintf(debugdir, sizeof(debugdir), "%s", buildid_dir);
-
 	if (add_name_list_str) {
 		list = strlist__new(true, add_name_list_str);
 		if (list) {
 			strlist__for_each(pos, list)
-				if (build_id_cache__add_file(pos->s, debugdir)) {
+				if (build_id_cache__add_file(pos->s, buildid_dir)) {
 					if (errno == EEXIST) {
 						pr_debug("%s already in the cache\n",
 							 pos->s);
@@ -359,7 +356,7 @@ int cmd_buildid_cache(int argc, const char **argv,
 		list = strlist__new(true, remove_name_list_str);
 		if (list) {
 			strlist__for_each(pos, list)
-				if (build_id_cache__remove_file(pos->s, debugdir)) {
+				if (build_id_cache__remove_file(pos->s, buildid_dir)) {
 					if (errno == ENOENT) {
 						pr_debug("%s wasn't in the cache\n",
 							 pos->s);
@@ -380,7 +377,7 @@ int cmd_buildid_cache(int argc, const char **argv,
 		list = strlist__new(true, update_name_list_str);
 		if (list) {
 			strlist__for_each(pos, list)
-				if (build_id_cache__update_file(pos->s, debugdir)) {
+				if (build_id_cache__update_file(pos->s, buildid_dir)) {
 					if (errno == ENOENT) {
 						pr_debug("%s wasn't in the cache\n",
 							 pos->s);
@@ -395,7 +392,7 @@ int cmd_buildid_cache(int argc, const char **argv,
 	}
 
 	if (kcore_filename &&
-	    build_id_cache__add_kcore(kcore_filename, debugdir, force))
+	    build_id_cache__add_kcore(kcore_filename, buildid_dir, force))
 		pr_warning("Couldn't add %s\n", kcore_filename);
 
 out:
tools/perf/util/build-id.c +3 -6
index e8d79e5bfaf7..0c72680a977f 100644
@@ -410,21 +410,18 @@ int perf_session__cache_build_ids(struct perf_session *session)
 {
 	struct rb_node *nd;
 	int ret;
-	char debugdir[PATH_MAX];
 
 	if (no_buildid_cache)
 		return 0;
 
-	snprintf(debugdir, sizeof(debugdir), "%s", buildid_dir);
-
-	if (mkdir(debugdir, 0755) != 0 && errno != EEXIST)
+	if (mkdir(buildid_dir, 0755) != 0 && errno != EEXIST)
 		return -1;
 
-	ret = machine__cache_build_ids(&session->machines.host, debugdir);
+	ret = machine__cache_build_ids(&session->machines.host, buildid_dir);
 
 	for (nd = rb_first(&session->machines.guests); nd; nd = rb_next(nd)) {
 		struct machine *pos = rb_entry(nd, struct machine, rb_node);
-		ret |= machine__cache_build_ids(pos, debugdir);
+		ret |= machine__cache_build_ids(pos, buildid_dir);
 	}
 	return ret ? -1 : 0;
 }
Signature · 1 line
1.9.3
author maintainer
#7 lore raw

PATCH 7/18 perf buildid cache: Fix -a segfault related to kcore handling

In reply to Arnaldo Carvalho de Melo

From: Jiri Olsa <[email protected]>

The kcore_filename is uninitialized and trash value could trigger build_id_cache__add_kcore function ending up with segfault.

Cc Adrian Hunter <[email protected]>
Cc Corey Ashford <[email protected]>
Cc David Ahern <[email protected]>
Cc Frederic Weisbecker <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc Namhyung Kim <[email protected]>
Cc Paul Mackerras <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Stephane Eranian <[email protected]>
Cc Steven Rostedt <[email protected]>
---
 tools/perf/builtin-buildid-cache.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/builtin-buildid-cache.c +1 -1
index 29f24c071bc6..77d5cae54c6a 100644
@@ -289,7 +289,7 @@ int cmd_buildid_cache(int argc, const char **argv,
 		   *remove_name_list_str = NULL,
 		   *missing_filename = NULL,
 		   *update_name_list_str = NULL,
-		   *kcore_filename;
+		   *kcore_filename = NULL;
 	char sbuf[STRERR_BUFSIZE];
 
 	struct perf_data_file file = {
Signature · 1 line
1.9.3
author maintainer
#8 lore raw

PATCH 8/18 perf tools: Add --buildid-dir option to set cache directory

In reply to Arnaldo Carvalho de Melo

From: Jiri Olsa <[email protected]>

Adding --buildid-dir to be able to set specific cache directory. It's going to be handy for buildid tests coming in shortly.

Cc Corey Ashford <[email protected]>
Cc David Ahern <[email protected]>
Cc Frederic Weisbecker <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc Namhyung Kim <[email protected]>
Cc Paul Mackerras <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Stephane Eranian <[email protected]>
Cc Steven Rostedt <[email protected]>
---
 tools/perf/Documentation/perf.txt |  4 ++++
 tools/perf/perf.c                 | 14 ++++++++++++--
 tools/perf/util/config.c          |  8 +++++---
 tools/perf/util/util.h            |  2 +-
 4 files changed, 22 insertions(+), 6 deletions(-)
tools/perf/Documentation/perf.txt +4 -0
index d240bb2e5b22..1e8e400b4493 100644
@@ -18,6 +18,10 @@ OPTIONS
 	  --debug verbose   # sets verbose = 1
 	  --debug verbose=2 # sets verbose = 2
 
+--buildid-dir::
+	Setup buildid cache directory. It has higher priority than
+	buildid.dir config file option.
+
 DESCRIPTION
 -----------
 Performance counters for Linux are a new kernel-based subsystem
tools/perf/perf.c +12 -2
index 452a8474d29d..3700a7faca6c 100644
@@ -200,6 +200,16 @@ static int handle_options(const char ***argv, int *argc, int *envchanged)
 				*envchanged = 1;
 			(*argv)++;
 			(*argc)--;
+		} else if (!strcmp(cmd, "--buildid-dir")) {
+			if (*argc < 2) {
+				fprintf(stderr, "No directory given for --buildid-dir.\n");
+				usage(perf_usage_string);
+			}
+			set_buildid_dir((*argv)[1]);
+			if (envchanged)
+				*envchanged = 1;
+			(*argv)++;
+			(*argc)--;
 		} else if (!prefixcmp(cmd, CMD_DEBUGFS_DIR)) {
 			perf_debugfs_set_path(cmd + strlen(CMD_DEBUGFS_DIR));
 			fprintf(stderr, "dir: %s\n", debugfs_mountpoint);
@@ -499,7 +509,7 @@ int main(int argc, const char **argv)
 	}
 	if (!prefixcmp(cmd, "trace")) {
 #ifdef HAVE_LIBAUDIT_SUPPORT
-		set_buildid_dir();
+		set_buildid_dir(NULL);
 		setup_path();
 		argv[0] = "trace";
 		return cmd_trace(argc, argv, NULL);
@@ -514,7 +524,7 @@ int main(int argc, const char **argv)
 	argc--;
 	handle_options(&argv, &argc, NULL);
 	commit_pager_choice();
-	set_buildid_dir();
+	set_buildid_dir(NULL);
 
 	if (argc > 0) {
 		if (!prefixcmp(argv[0], "--"))
tools/perf/util/config.c +5 -3
index c802236b2bfe..e18f653cd7db 100644
@@ -539,12 +539,14 @@ static void check_buildid_dir_config(void)
 	perf_config(buildid_dir_command_config, &c);
 }
 
-void set_buildid_dir(void)
+void set_buildid_dir(const char *dir)
 {
-	buildid_dir[0] = '\0';
+	if (dir)
+		scnprintf(buildid_dir, MAXPATHLEN-1, "%s", dir);
 
 	/* try config file */
-	check_buildid_dir_config();
+	if (buildid_dir[0] == '\0')
+		check_buildid_dir_config();
 
 	/* default to $HOME/.debug */
 	if (buildid_dir[0] == '\0') {
tools/perf/util/util.h +1 -1
index 419bee030f83..abc445ee4f60 100644
@@ -153,7 +153,7 @@ extern void warning(const char *err, ...) __attribute__((format (printf, 1, 2)))
 extern void set_die_routine(void (*routine)(const char *err, va_list params) NORETURN);
 
 extern int prefixcmp(const char *str, const char *prefix);
-extern void set_buildid_dir(void);
+extern void set_buildid_dir(const char *dir);
 
 static inline const char *skip_prefix(const char *str, const char *prefix)
 {
Signature · 1 line
1.9.3
author maintainer
#9 lore raw

PATCH 9/18 perf callchain: Fixup parameter handling error message

In reply to Arnaldo Carvalho de Melo

From: Kan Liang <[email protected]>

Fix up parse_callchain_record_opt error message for 'fp', in the past using '-g fp' was a valid alternative to '--call-graph fp', which is not the case since:

commit 09b0fd45ff63413df94cbd832a765076b201edbb
Author: Jiri Olsa <[email protected]>
Date: Sat Oct 26 16:25:33 2013 +0200

  perf record: Split -g and --call-graph

I.e. -g means "use the configured unwind data collection method" which has as default 'fp', while --call-graph requires passing the method to use.

Cc Andi Kleen <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc Jiri Olsa <[email protected]>
Cc Namhyung Kim <[email protected]>
Cc Paul Mackerras <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Stephane Eranian <[email protected]>
[ split this from a larger patch related to LBR based unwinding ]
---
 tools/perf/util/callchain.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/util/callchain.c +1 -1
index cf524a35cc84..64b377e591e4 100644
@@ -77,7 +77,7 @@ int parse_callchain_record_opt(const char *arg)
 				ret = 0;
 			} else
 				pr_err("callchain: No more arguments "
-				       "needed for -g fp\n");
+				       "needed for --call-graph fp\n");
 			break;
 
 #ifdef HAVE_DWARF_UNWIND_SUPPORT
Signature · 1 line
1.9.3
author maintainer
#10 lore raw

PATCH 10/18 perf callchain: Move cpumode resolve code to add_callchain_ip

In reply to Arnaldo Carvalho de Melo

From: Kan Liang <[email protected]>

Using flag to distinguish between branch_history and normal callchain.

Move the cpumode to add_callchain_ip function.

No change in behavior.

---
 tools/perf/util/machine.c | 72 +++++++++++++++++++++++------------------------
 1 file changed, 35 insertions(+), 37 deletions(-)
tools/perf/util/machine.c +35 -37
index 15dd0a9691ce..94de3e48b490 100644
@@ -1385,19 +1385,46 @@ struct mem_info *sample__resolve_mem(struct perf_sample *sample,
 static int add_callchain_ip(struct thread *thread,
 			    struct symbol **parent,
 			    struct addr_location *root_al,
-			    int cpumode,
+			    bool branch_history,
 			    u64 ip)
 {
 	struct addr_location al;
 
 	al.filtered = 0;
 	al.sym = NULL;
-	if (cpumode == -1)
+	if (branch_history)
 		thread__find_cpumode_addr_location(thread, MAP__FUNCTION,
 						   ip, &al);
-	else
+	else {
+		u8 cpumode = PERF_RECORD_MISC_USER;
+
+		if (ip >= PERF_CONTEXT_MAX) {
+			switch (ip) {
+			case PERF_CONTEXT_HV:
+				cpumode = PERF_RECORD_MISC_HYPERVISOR;
+				break;
+			case PERF_CONTEXT_KERNEL:
+				cpumode = PERF_RECORD_MISC_KERNEL;
+				break;
+			case PERF_CONTEXT_USER:
+				cpumode = PERF_RECORD_MISC_USER;
+				break;
+			default:
+				pr_debug("invalid callchain context: "
+					 "%"PRId64"\n", (s64) ip);
+				/*
+				 * It seems the callchain is corrupted.
+				 * Discard all.
+				 */
+				callchain_cursor_reset(&callchain_cursor);
+				return 1;
+			}
+			return 0;
+		}
 		thread__find_addr_location(thread, cpumode, MAP__FUNCTION,
 				   ip, &al);
+	}
+
 	if (al.sym != NULL) {
 		if (sort__has_parent && !*parent &&
 		    symbol__match_regex(al.sym, &parent_regex))
@@ -1480,11 +1507,8 @@ static int thread__resolve_callchain_sample(struct thread *thread,
 					     struct addr_location *root_al,
 					     int max_stack)
 {
-	u8 cpumode = PERF_RECORD_MISC_USER;
 	int chain_nr = min(max_stack, (int)chain->nr);
-	int i;
-	int j;
-	int err;
+	int i, j, err;
 	int skip_idx = -1;
 	int first_call = 0;
 
@@ -1542,10 +1566,10 @@ static int thread__resolve_callchain_sample(struct thread *thread,
 
 		for (i = 0; i < nr; i++) {
 			err = add_callchain_ip(thread, parent, root_al,
-					       -1, be[i].to);
+					       true, be[i].to);
 			if (!err)
 				err = add_callchain_ip(thread, parent, root_al,
-						       -1, be[i].from);
+						       true, be[i].from);
 			if (err == -EINVAL)
 				break;
 			if (err)
@@ -1574,36 +1598,10 @@ check_calls:
 #endif
 		ip = chain->ips[j];
 
-		if (ip >= PERF_CONTEXT_MAX) {
-			switch (ip) {
-			case PERF_CONTEXT_HV:
-				cpumode = PERF_RECORD_MISC_HYPERVISOR;
-				break;
-			case PERF_CONTEXT_KERNEL:
-				cpumode = PERF_RECORD_MISC_KERNEL;
-				break;
-			case PERF_CONTEXT_USER:
-				cpumode = PERF_RECORD_MISC_USER;
-				break;
-			default:
-				pr_debug("invalid callchain context: "
-					 "%"PRId64"\n", (s64) ip);
-				/*
-				 * It seems the callchain is corrupted.
-				 * Discard all.
-				 */
-				callchain_cursor_reset(&callchain_cursor);
-				return 0;
-			}
-			continue;
-		}
+		err = add_callchain_ip(thread, parent, root_al, false, ip);
 
-		err = add_callchain_ip(thread, parent, root_al,
-				       cpumode, ip);
-		if (err == -EINVAL)
-			break;
 		if (err)
-			return err;
+			return (err < 0) ? err : 0;
 	}
 
 	return 0;
Signature · 1 line
1.9.3
author maintainer
#11 lore raw

PATCH 11/18 calloc/xcalloc: Fix argument order

In reply to Arnaldo Carvalho de Melo

From: Arjun Sreedharan <[email protected]>

The calloc() and xcalloc() functions takes @nmemb first and then @size. Fix all w/ pattern "calloc\s*(\s*sizeof".

---
 scripts/kconfig/mconf.c    | 4 ++--
 tools/perf/ui/hist.c       | 4 ++--
 tools/thermal/tmon/sysfs.c | 6 +++---
 3 files changed, 7 insertions(+), 7 deletions(-)
scripts/kconfig/mconf.c +2 -2
index 14cea7463a62..4dd37552abc2 100644
@@ -330,10 +330,10 @@ static void set_subtitle(void)
 	list_for_each_entry(sp, &trail, entries) {
 		if (sp->text) {
 			if (pos) {
-				pos->next = xcalloc(sizeof(*pos), 1);
+				pos->next = xcalloc(1, sizeof(*pos));
 				pos = pos->next;
 			} else {
-				subtitles = pos = xcalloc(sizeof(*pos), 1);
+				subtitles = pos = xcalloc(1, sizeof(*pos));
 			}
 			pos->text = sp->text;
 		}
tools/perf/ui/hist.c +2 -2
index 2af18376b077..dc0d095f318c 100644
@@ -162,8 +162,8 @@ static int __hpp__sort(struct hist_entry *a, struct hist_entry *b,
 		return ret;
 
 	nr_members = evsel->nr_members;
-	fields_a = calloc(sizeof(*fields_a), nr_members);
-	fields_b = calloc(sizeof(*fields_b), nr_members);
+	fields_a = calloc(nr_members, sizeof(*fields_a));
+	fields_b = calloc(nr_members, sizeof(*fields_b));
 
 	if (!fields_a || !fields_b)
 		goto out;
tools/thermal/tmon/sysfs.c +3 -3
index dfe454855cd2..1c12536f2081 100644
@@ -446,7 +446,7 @@ int probe_thermal_sysfs(void)
 		return -1;
 	}
 
-	ptdata.tzi = calloc(sizeof(struct tz_info), ptdata.max_tz_instance+1);
+	ptdata.tzi = calloc(ptdata.max_tz_instance+1, sizeof(struct tz_info));
 	if (!ptdata.tzi) {
 		fprintf(stderr, "Err: allocate tz_info\n");
 		return -1;
@@ -454,8 +454,8 @@ int probe_thermal_sysfs(void)
 
 	/* we still show thermal zone information if there is no cdev */
 	if (ptdata.nr_cooling_dev) {
-		ptdata.cdi = calloc(sizeof(struct cdev_info),
-				ptdata.max_cdev_instance + 1);
+		ptdata.cdi = calloc(ptdata.max_cdev_instance + 1,
+				sizeof(struct cdev_info));
 		if (!ptdata.cdi) {
 			free(ptdata.tzi);
 			fprintf(stderr, "Err: allocate cdev_info\n");
Signature · 1 line
1.9.3
author maintainer
#12 lore raw

PATCH 12/18 perf tests: Fix attr tests size values to cope with machine state on interrupt ABI changes

In reply to Arnaldo Carvalho de Melo

From: Jiri Olsa <[email protected]>

Following change adjusted 'struct perf_event_attr', but let the attr test's sizes untouched:
60e2364e60e8 perf: Add ability to sample machine state on interrupt

[jolsa@krava perf]$ ./perf test attr -vv
--- start ---
test child forked, pid 9719
running './tests/attr/test-stat-group1'
'PERF_TEST_ATTR=/tmp/tmp4drvul ./perf stat -o /tmp/tmp4drvul/perf.data -e '{cycles,instructions}' kill >/dev/null 2>&1' ret 1
expected size=96, got 104
FAILED './tests/attr/test-stat-group1' - match failure

Adjusting test size values for attr test.

---
 tools/perf/tests/attr/base-record | 2 +-
 tools/perf/tests/attr/base-stat   | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)
tools/perf/tests/attr/base-record +1 -1
index f710b92ccff6..d3095dafed36 100644
@@ -5,7 +5,7 @@ group_fd=-1
 flags=0|8
 cpu=*
 type=0|1
-size=96
+size=104
 config=0
 sample_period=4000
 sample_type=263
tools/perf/tests/attr/base-stat +1 -1
index dc3ada2470c0..872ed7e24c7c 100644
@@ -5,7 +5,7 @@ group_fd=-1
 flags=0|8
 cpu=*
 type=0
-size=96
+size=104
 config=0
 sample_period=0
 sample_type=0
Signature · 1 line
1.9.3
author maintainer
#13 lore raw

PATCH 13/18 perf kvm stat live: Mark events as (x86 only) in help output

In reply to Arnaldo Carvalho de Melo

From: Alexander Yarygin <[email protected]>

The mmio and ioport events are useful only on x86.

---
 tools/perf/builtin-kvm.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
tools/perf/builtin-kvm.c +2 -1
index 3c0f3d4fb021..0894a817f67e 100644
@@ -1293,7 +1293,8 @@ static int kvm_events_live(struct perf_kvm_stat *kvm,
 		OPT_UINTEGER('d', "display", &kvm->display_time,
 			"time in seconds between display updates"),
 		OPT_STRING(0, "event", &kvm->report_event, "report event",
-			"event for reporting: vmexit, mmio, ioport"),
+			"event for reporting: "
+			"vmexit, mmio (x86 only), ioport (x86 only)"),
 		OPT_INTEGER(0, "vcpu", &kvm->trace_vcpu,
 			"vcpu id to report"),
 		OPT_STRING('k', "key", &kvm->sort_key, "sort-key",
Signature · 1 line
1.9.3
author maintainer
#14 lore raw

PATCH 14/18 tools lib fs: Adopt filename__read_int from tools/perf/

In reply to Arnaldo Carvalho de Melo

From: Arnaldo Carvalho de Melo <[email protected]>

Will be useful for new helpers to read sysctl values.

Cc Adrian Hunter <[email protected]>
Cc Borislav Petkov <[email protected]>
Cc David Ahern <[email protected]>
Cc Don Zickus <[email protected]>
Cc Frederic Weisbecker <[email protected]>
Cc Jiri Olsa <[email protected]>
Cc Mike Galbraith <[email protected]>
Cc Namhyung Kim <[email protected]>
Cc Paul Mackerras <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Stephane Eranian <[email protected]>
---
 tools/lib/api/fs/fs.c  | 21 +++++++++++++++++++++
 tools/lib/api/fs/fs.h  |  2 ++
 tools/perf/util/util.c | 17 -----------------
 tools/perf/util/util.h |  1 -
 4 files changed, 23 insertions(+), 18 deletions(-)
tools/lib/api/fs/fs.c +21 -0
index c1b49c36a951..b8d0df8878d1 100644
@@ -7,6 +7,10 @@
 #include <stdlib.h>
 #include <string.h>
 #include <sys/vfs.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <fcntl.h>
+#include <unistd.h>
 
 #include "debugfs.h"
 #include "fs.h"
@@ -163,3 +167,20 @@ const char *name##__mountpoint(void)	\
 
 FS__MOUNTPOINT(sysfs,  FS__SYSFS);
 FS__MOUNTPOINT(procfs, FS__PROCFS);
+
+int filename__read_int(const char *filename, int *value)
+{
+	char line[64];
+	int fd = open(filename, O_RDONLY), err = -1;
+
+	if (fd < 0)
+		return -1;
+
+	if (read(fd, line, sizeof(line)) > 0) {
+		*value = atoi(line);
+		err = 0;
+	}
+
+	close(fd);
+	return err;
+}
tools/lib/api/fs/fs.h +2 -0
index cb7049551f33..4d0aef6e7163 100644
@@ -11,4 +11,6 @@
 
 const char *sysfs__mountpoint(void);
 const char *procfs__mountpoint(void);
+
+int filename__read_int(const char *filename, int *value);
 #endif /* __API_FS__ */
tools/perf/util/util.c +0 -17
index d5eab3f3323f..41125e5b011e 100644
@@ -442,23 +442,6 @@ unsigned long parse_tag_value(const char *str, struct parse_tag *tags)
 	return (unsigned long) -1;
 }
 
-int filename__read_int(const char *filename, int *value)
-{
-	char line[64];
-	int fd = open(filename, O_RDONLY), err = -1;
-
-	if (fd < 0)
-		return -1;
-
-	if (read(fd, line, sizeof(line)) > 0) {
-		*value = atoi(line);
-		err = 0;
-	}
-
-	close(fd);
-	return err;
-}
-
 int filename__read_str(const char *filename, char **buf, size_t *sizep)
 {
 	size_t size = 0, alloc_size = 0;
tools/perf/util/util.h +0 -1
index abc445ee4f60..008b361b1758 100644
@@ -343,7 +343,6 @@ char *get_srcline(struct dso *dso, unsigned long addr, struct symbol *sym,
 		  bool show_sym);
 void free_srcline(char *srcline);
 
-int filename__read_int(const char *filename, int *value);
 int filename__read_str(const char *filename, char **buf, size_t *sizep);
 int perf_event_paranoid(void);
 
Signature · 1 line
1.9.3
author maintainer
#15 lore raw

PATCH 15/18 tools lib fs: Add sysctl__read_int helper

In reply to Arnaldo Carvalho de Melo

From: Arnaldo Carvalho de Melo <[email protected]>

Will come in handy for tools, see next patches.

Cc Adrian Hunter <[email protected]>
Cc Borislav Petkov <[email protected]>
Cc David Ahern <[email protected]>
Cc Don Zickus <[email protected]>
Cc Frederic Weisbecker <[email protected]>
Cc Jiri Olsa <[email protected]>
Cc Mike Galbraith <[email protected]>
Cc Namhyung Kim <[email protected]>
Cc Paul Mackerras <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Stephane Eranian <[email protected]>
---
 tools/lib/api/fs/fs.c | 13 +++++++++++++
 tools/lib/api/fs/fs.h |  1 +
 2 files changed, 14 insertions(+)
tools/lib/api/fs/fs.c +13 -0
index b8d0df8878d1..65d9be3f9887 100644
@@ -184,3 +184,16 @@ int filename__read_int(const char *filename, int *value)
 	close(fd);
 	return err;
 }
+
+int sysctl__read_int(const char *sysctl, int *value)
+{
+	char path[PATH_MAX];
+	const char *procfs = procfs__mountpoint();
+
+	if (!procfs)
+		return -1;
+
+	snprintf(path, sizeof(path), "%s/sys/%s", procfs, sysctl);
+
+	return filename__read_int(path, value);
+}
tools/lib/api/fs/fs.h +1 -0
index 4d0aef6e7163..6caa2bbc6cec 100644
@@ -13,4 +13,5 @@ const char *sysfs__mountpoint(void);
 const char *procfs__mountpoint(void);
 
 int filename__read_int(const char *filename, int *value);
+int sysctl__read_int(const char *sysctl, int *value);
 #endif /* __API_FS__ */
Signature · 1 line
1.9.3
author maintainer
#16 lore raw

PATCH 16/18 perf tools: Use sysctl__read_int instead of ad-hoc copies

In reply to Arnaldo Carvalho de Melo

From: Arnaldo Carvalho de Melo <[email protected]>

Cc Adrian Hunter <[email protected]>
Cc Borislav Petkov <[email protected]>
Cc David Ahern <[email protected]>
Cc Don Zickus <[email protected]>
Cc Frederic Weisbecker <[email protected]>
Cc Jiri Olsa <[email protected]>
Cc Mike Galbraith <[email protected]>
Cc Namhyung Kim <[email protected]>
Cc Paul Mackerras <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Stephane Eranian <[email protected]>
---
 tools/perf/util/record.c | 11 +----------
 tools/perf/util/util.c   |  9 +--------
 2 files changed, 2 insertions(+), 18 deletions(-)
tools/perf/util/record.c +1 -10
index cf69325b985f..8acd0df88b5c 100644
@@ -137,16 +137,7 @@ void perf_evlist__config(struct perf_evlist *evlist, struct record_opts *opts)
 
 static int get_max_rate(unsigned int *rate)
 {
-	char path[PATH_MAX];
-	const char *procfs = procfs__mountpoint();
-
-	if (!procfs)
-		return -1;
-
-	snprintf(path, PATH_MAX,
-		 "%s/sys/kernel/perf_event_max_sample_rate", procfs);
-
-	return filename__read_int(path, (int *) rate);
+	return sysctl__read_int("kernel/perf_event_max_sample_rate", (int *)rate);
 }
 
 static int record_opts__config_freq(struct record_opts *opts)
tools/perf/util/util.c +1 -8
index 41125e5b011e..b86744f29eef 100644
@@ -506,16 +506,9 @@ const char *get_filename_for_perf_kvm(void)
 
 int perf_event_paranoid(void)
 {
-	char path[PATH_MAX];
-	const char *procfs = procfs__mountpoint();
 	int value;
 
-	if (!procfs)
-		return INT_MAX;
-
-	scnprintf(path, PATH_MAX, "%s/sys/kernel/perf_event_paranoid", procfs);
-
-	if (filename__read_int(path, &value))
+	if (sysctl__read_int("kernel/perf_event_paranoid", &value))
 		return INT_MAX;
 
 	return value;
Signature · 1 line
1.9.3
author maintainer
#17 lore raw

PATCH 17/18 perf evlist: Introduce strerror_mmap method

In reply to Arnaldo Carvalho de Melo

From: Arnaldo Carvalho de Melo <[email protected]>

To pretty print hints about perf_evlist__mmap errors. Will be used in 'trace' in the next patch.

Cc Adrian Hunter <[email protected]>
Cc Borislav Petkov <[email protected]>
Cc David Ahern <[email protected]>
Cc Don Zickus <[email protected]>
Cc Frederic Weisbecker <[email protected]>
Cc Jiri Olsa <[email protected]>
Cc Mike Galbraith <[email protected]>
Cc Namhyung Kim <[email protected]>
Cc Paul Mackerras <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Stephane Eranian <[email protected]>
---
 tools/perf/util/evlist.c | 23 +++++++++++++++++++++++
 tools/perf/util/evlist.h |  1 +
 2 files changed, 24 insertions(+)
tools/perf/util/evlist.c +23 -0
index cfbe2b99b9aa..bb5dfc5d1e75 100644
@@ -8,6 +8,7 @@
  */
 #include "util.h"
 #include <api/fs/debugfs.h>
+#include <api/fs/fs.h>
 #include <poll.h>
 #include "cpumap.h"
 #include "thread_map.h"
@@ -1483,6 +1484,28 @@ int perf_evlist__strerror_open(struct perf_evlist *evlist __maybe_unused,
 	return 0;
 }
 
+int perf_evlist__strerror_mmap(struct perf_evlist *evlist, int err, char *buf, size_t size)
+{
+	char sbuf[STRERR_BUFSIZE], *emsg = strerror_r(err, sbuf, sizeof(sbuf));
+	int value;
+
+	switch (err) {
+	case EPERM:
+		sysctl__read_int("kernel/perf_event_mlock_kb", &value);
+		scnprintf(buf, size, "Error:\t%s.\n"
+				     "Hint:\tCheck /proc/sys/kernel/perf_event_mlock_kb (%d kB) setting.\n"
+				     "Hint:\tTried using %zd kB.\n"
+				     "Hint:\tTry using a bigger -m/--mmap-pages value.",
+				     emsg, value, evlist->mmap_len / 1024);
+		break;
+	default:
+		scnprintf(buf, size, "%s", emsg);
+		break;
+	}
+
+	return 0;
+}
+
 void perf_evlist__to_front(struct perf_evlist *evlist,
 			   struct perf_evsel *move_evsel)
 {
tools/perf/util/evlist.h +1 -0
index 649b0c597283..0ba93f67ab94 100644
@@ -185,6 +185,7 @@ size_t perf_evlist__fprintf(struct perf_evlist *evlist, FILE *fp);
 
 int perf_evlist__strerror_tp(struct perf_evlist *evlist, int err, char *buf, size_t size);
 int perf_evlist__strerror_open(struct perf_evlist *evlist, int err, char *buf, size_t size);
+int perf_evlist__strerror_mmap(struct perf_evlist *evlist, int err, char *buf, size_t size);
 
 static inline unsigned int perf_mmap__read_head(struct perf_mmap *mm)
 {
Signature · 1 line
1.9.3
author maintainer
#18 lore raw

PATCH 18/18 perf trace: Provide a better explanation when mmap fails

In reply to Arnaldo Carvalho de Melo

From: Arnaldo Carvalho de Melo <[email protected]>

If we ask for a mmap lenght than the max configured via the relevant sysctl, provide a better warning, instead of just expanding the EPERM returned:

[acme@ssdandy ~]$ trace -m 256 -e nanosleep sleep 2
Error: Operation not permitted.
Hint: Check /proc/sys/kernel/perf_event_mlock_kb (516 kB) setting.
Hint: Tried using 1028 kB.
Hint: Try using a bigger -m/--mmap-pages value.

[acme@ssdandy ~]$ trace -m 128 -e nanosleep sleep 2
2001.280 (2000.403 ms): nanosleep(rqtp: 0x7fff89a8a7f0) = 0
[acme@ssdandy ~]$

An upcoming patch will autotune the request for non-root users when -m is not used.

Cc Adrian Hunter <[email protected]>
Cc Borislav Petkov <[email protected]>
Cc David Ahern <[email protected]>
Cc Don Zickus <[email protected]>
Cc Frederic Weisbecker <[email protected]>
Cc Jiri Olsa <[email protected]>
Cc Mike Galbraith <[email protected]>
Cc Namhyung Kim <[email protected]>
Cc Paul Mackerras <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Stephane Eranian <[email protected]>
---
 tools/perf/builtin-trace.c | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)
tools/perf/builtin-trace.c +6 -6
index 83a4835c8118..327541e43c7f 100644
@@ -2045,7 +2045,6 @@ static int trace__run(struct trace *trace, int argc, const char **argv)
 	unsigned long before;
 	const bool forks = argc > 0;
 	bool draining = false;
-	char sbuf[STRERR_BUFSIZE];
 
 	trace->live = true;
 
@@ -2106,11 +2105,8 @@ static int trace__run(struct trace *trace, int argc, const char **argv)
 		goto out_error_open;
 
 	err = perf_evlist__mmap(evlist, trace->opts.mmap_pages, false);
-	if (err < 0) {
-		fprintf(trace->output, "Couldn't mmap the events: %s\n",
-			strerror_r(errno, sbuf, sizeof(sbuf)));
-		goto out_delete_evlist;
-	}
+	if (err < 0)
+		goto out_error_mmap;
 
 	perf_evlist__enable(evlist);
 
@@ -2210,6 +2206,10 @@ out_error_tp:
 	perf_evlist__strerror_tp(evlist, errno, errbuf, sizeof(errbuf));
 	goto out_error;
 
+out_error_mmap:
+	perf_evlist__strerror_mmap(evlist, errno, errbuf, sizeof(errbuf));
+	goto out_error;
+
 out_error_open:
 	perf_evlist__strerror_open(evlist, errno, errbuf, sizeof(errbuf));
 
Signature · 1 line
1.9.3
#19 lore raw
In reply to Arnaldo Carvalho de Melo
Quoted 114 lines · Arnaldo Carvalho de Melo

Hi Ingo,

Please consider pulling,

- Arnaldo

The following changes since commit cfa0bd52d0ba9b852f76c7b3f1055edd5e5c7846:

Merge tag 'perf-core-for-mingo-2' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/core (2014-12-08 07:45:45 +0100)

are available in the git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux.git tags/perf-core-for-mingo

for you to fetch changes up to e09b18d4907992d3d615b215c1abf585721b2810:

perf trace: Provide a better explanation when mmap fails (2014-12-11 18:04:10 -0300)

----------------------------------------------------------------
perf/core improvements and fixes:

User visible:

- Mark events as (x86 only) in help output for 'perf kvm stat live" (Alexander Yarygin)

- Provide a better explanation when mmap fails in 'trace' (Arnaldo Carvalho de Melo)

- Add --buildid-dir option to set cache directory, i.e. use:

      $ perf --buildid-dir /path/to/dir tool --tool-options

(Jiri Olsa)

- Fix memcpy/memset 'perf bench' output (Rabin Vicent)

- Fix 'perf test' attr tests size values to cope with machine state on interrupt ABI changes (Jiri Olsa)

- Fixup callchain type parameter handling error message (Kan Liang)

Infrastructure/cleanups:

- calloc/xcalloc: Fix argument order (Arjun Sreedharan)

- Move filename__read_int from tools/perf/ to tools/lib, add sysctl__read_int there and use it in place of ad-hoc copies (Arnaldo Carvalho de Melo)

- Use single strcmp call instead of two (Jiri Olsa)

- Remove extra debugdir variables in 'perf buildid-cache' (Jiri Olsa)

- Fix -a segfault related to kcore handling in 'perf buildid-cache' (Jiri Olsa)

- Move cpumode resolve code to add_callchain_ip (Kan Liang)

- Merge memset into memcpy 'perf bench' (Rabin Vincent)

- Change print format from %lu to %PRIu64 in the hists browser (Tom Huynh)

----------------------------------------------------------------
Alexander Yarygin (1):
perf kvm stat live: Mark events as (x86 only) in help output

Arjun Sreedharan (1):
calloc/xcalloc: Fix argument order

Arnaldo Carvalho de Melo (5):
tools lib fs: Adopt filename__read_int from tools/perf/ tools lib fs: Add sysctl__read_int helper
perf tools: Use sysctl__read_int instead of ad-hoc copies perf evlist: Introduce strerror_mmap method perf trace: Provide a better explanation when mmap fails

Jiri Olsa (5):
perf tools: Use single strcmp call instead of two
perf buildid-cache: Remove extra debugdir variables
perf buildid cache: Fix -a segfault related to kcore handling
perf tools: Add --buildid-dir option to set cache directory
perf tests: Fix attr tests size values to cope with machine state on interrupt ABI changes

Kan Liang (2):
perf callchain: Fixup parameter handling error message perf callchain: Move cpumode resolve code to add_callchain_ip

Rabin Vincent (3):
perf bench: Prepare memcpy for merge
perf bench: Merge memset into memcpy
perf bench: Fix memcpy/memset output

Tom Huynh (1):
perf hists browser: Change print format from %lu to %PRIu64

 scripts/kconfig/mconf.c            |   4 +-
 tools/lib/api/fs/fs.c              |  34 +++++
 tools/lib/api/fs/fs.h              |   3 +
 tools/perf/Documentation/perf.txt  |   4 +
 tools/perf/Makefile.perf           |   1 -
 tools/perf/bench/mem-memcpy.c      | 286 ++++++++++++++++++++++++----------
 tools/perf/bench/mem-memset.c      | 304 -------------------------------------
 tools/perf/builtin-buildid-cache.c |  13 +-
 tools/perf/builtin-kvm.c           |   3 +-
 tools/perf/builtin-trace.c         |  12 +-
 tools/perf/perf.c                  |  14 +-
 tools/perf/tests/attr/base-record  |   2 +-
 tools/perf/tests/attr/base-stat    |   2 +-
 tools/perf/ui/browsers/hists.c     |   2 +-
 tools/perf/ui/hist.c               |   4 +-
 tools/perf/util/build-id.c         |   9 +-
 tools/perf/util/callchain.c        |   2 +-
 tools/perf/util/config.c           |  10 +-
 tools/perf/util/evlist.c           |  23 +++
Quoted 8 lines
 tools/perf/util/evlist.h           |   1 +
 tools/perf/util/machine.c          |  72 +++++----
 tools/perf/util/record.c           |  11 +-
 tools/perf/util/util.c             |  26 +---
 tools/perf/util/util.h             |   3 +-
 tools/thermal/tmon/sysfs.c         |   6 +-
 25 files changed, 350 insertions(+), 501 deletions(-)
 delete mode 100644 tools/perf/bench/mem-memset.c

Pulled, thanks a lot Arnaldo!

Ingo