[PATCH V6 0/3] perf tools: pmu event new style format fix

[PATCH V6 0/3] perf tools: pmu event new style format fix

[kan.liang]

From: Kan Liang <kan.liang@intel.com>

There are two types of pmu event stytle formats, "pmu_event_name"
or "cpu/pmu_event_name/". However, there is a bug on supporting these
two formats, especially when they mixed with other perf events.
The patch set intends to fix this issue.

The patch set has been tested on my haswell.
Here is the test script I used for this issue.
(Note: please make sure that your test system supports TSX and
L1-dcache-loads events. Otherwise, you may want to change the events 
to other pmu events.)

[lk@localhost ~]$ cat perf_style_test.sh
#hardware events + kernel pmu event with different style
perf stat -x, -e cycles,mem-stores,tx-start sleep 2
perf stat -x, -e cpu-cycles,cycles-ct,cycles-t sleep 2
perf stat -x, -e cycles,cpu/cycles-ct/,cpu/cycles-t/ sleep 2
perf stat -x, -e instructions,cpu/tx-start/ sleep 2
perf stat -x, -e '{cycles,tx-start}' sleep 2
perf stat -x, -e '{cycles,cpu/tx-start/}' sleep 2

#HW Cache event + kernel pmu event with different style
perf stat -x, -e L1-dcache-loads,cpu/mem-stores/,tx-start sleep 2
perf stat -x, -e L1-dcache-loads,mem-stores,cpu/tx-start/ sleep 2
perf stat -x, -e '{L1-dcache-loads,mem-stores}' sleep 2
perf stat -x, -e '{L1-dcache-loads,cpu/tx-start/}' sleep 2

#Raw event + kernel pmu event with different style:
perf stat -x, -e cpu/event=0xc0,umask=0x00/,mem-loads,cpu/mem-stores/ sleep 2
perf stat -x, -e cpu/event=0xc0,umask=0x00/,tx-start,cpu/el-start/ sleep 2
perf stat -x, -e '{cpu/event=0xc0,umask=0x00/,tx-start}' sleep 2

Changes since V1:
Read kernel PMU events from sysfs at runtime

Changes since V2:
Use strlcpy to replace strncpy

Changes since V3:
rebase to git.kernel.org/pub/scm/linux/kernel/git/acme/linux perf/core

Changes since V4:
scan kernel pmu events from sysfs only needed
rename the init/check/clenup functions and related struct.
allocate each symbol string separatelly
Use ALLOC_LIST

Changes since V5:
Using perf_pmu__find to instead of perf_pmu__scan
Don't scan all the time if the system doesn't support kernel pmu events

Kan Liang (3):
  Revert "perf tools: Default to cpu// for events v5"
  perf tools: parse the pmu event prefix and surfix
  perf tools: Add support to new style format of kernel PMU event

 tools/perf/util/include/linux/string.h |   1 -
 tools/perf/util/parse-events.c         | 128 +++++++++++++++++++++++++++------
 tools/perf/util/parse-events.h         |  14 ++++
 tools/perf/util/parse-events.l         |  30 +++++++-
 tools/perf/util/parse-events.y         |  40 +++++++++++
 tools/perf/util/pmu.c                  |  10 ---
 tools/perf/util/pmu.h                  |  10 +++
 tools/perf/util/string.c               |  24 -------
 8 files changed, 199 insertions(+), 58 deletions(-)

-- 
1.8.3.2

[PATCH V6 1/3] Revert "perf tools: Default to cpu// for events v5"

[kan.liang]

From: Kan Liang <kan.liang@intel.com>

This reverts commit 50e200f07948 ("perf tools: Default to cpu// for
events v5")

The fixup cannot handle the case that
new style format(which without //) mixed with
other different formats.

For example,
group events with new style format: {mem-stores,mem-loads}
some hardware event + new style event: cycles,mem-loads
Cache event + new style event: LLC-loads,mem-loads
Raw event + new style event:
cpu/event=0xc8,umask=0x08/,mem-loads
old style event and new stytle mixture: mem-stores,cpu/mem-loads/

Signed-off-by: Kan Liang <kan.liang@intel.com>
---
 tools/perf/util/include/linux/string.h |  1 -
 tools/perf/util/parse-events.c         | 30 +-----------------------------
 tools/perf/util/string.c               | 24 ------------------------
 3 files changed, 1 insertion(+), 54 deletions(-)

diff --git a/tools/perf/util/include/linux/string.h b/tools/perf/util/include/linux/string.h
index 97a8007..6f19c54 100644
--- a/tools/perf/util/include/linux/string.h
+++ b/tools/perf/util/include/linux/string.h
@@ -1,4 +1,3 @@
 #include <string.h>
 
 void *memdup(const void *src, size_t len);
-int str_append(char **s, int *len, const char *a);
diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c
index 61be3e6..b9174bc 100644
--- a/tools/perf/util/parse-events.c
+++ b/tools/perf/util/parse-events.c
@@ -6,7 +6,7 @@
 #include "parse-options.h"
 #include "parse-events.h"
 #include "exec_cmd.h"
-#include "linux/string.h"
+#include "string.h"
 #include "symbol.h"
 #include "cache.h"
 #include "header.h"
@@ -864,32 +864,6 @@ int parse_events_name(struct list_head *list, char *name)
 	return 0;
 }
 
-static int parse_events__scanner(const char *str, void *data, int start_token);
-
-static int parse_events_fixup(int ret, const char *str, void *data,
-			      int start_token)
-{
-	char *o = strdup(str);
-	char *s = NULL;
-	char *t = o;
-	char *p;
-	int len = 0;
-
-	if (!o)
-		return ret;
-	while ((p = strsep(&t, ",")) != NULL) {
-		if (s)
-			str_append(&s, &len, ",");
-		str_append(&s, &len, "cpu/");
-		str_append(&s, &len, p);
-		str_append(&s, &len, "/");
-	}
-	free(o);
-	if (!s)
-		return -ENOMEM;
-	return parse_events__scanner(s, data, start_token);
-}
-
 static int parse_events__scanner(const char *str, void *data, int start_token)
 {
 	YY_BUFFER_STATE buffer;
@@ -910,8 +884,6 @@ static int parse_events__scanner(const char *str, void *data, int start_token)
 	parse_events__flush_buffer(buffer, scanner);
 	parse_events__delete_buffer(buffer, scanner);
 	parse_events_lex_destroy(scanner);
-	if (ret && !strchr(str, '/'))
-		ret = parse_events_fixup(ret, str, data, start_token);
 	return ret;
 }
 
diff --git a/tools/perf/util/string.c b/tools/perf/util/string.c
index 2553e5b..4b0ff22 100644
--- a/tools/perf/util/string.c
+++ b/tools/perf/util/string.c
@@ -387,27 +387,3 @@ void *memdup(const void *src, size_t len)
 
 	return p;
 }
-
-/**
- * str_append - reallocate string and append another
- * @s: pointer to string pointer
- * @len: pointer to len (initialized)
- * @a: string to append.
- */
-int str_append(char **s, int *len, const char *a)
-{
-	int olen = *s ? strlen(*s) : 0;
-	int nlen = olen + strlen(a) + 1;
-	if (*len < nlen) {
-		*len = *len * 2;
-		if (*len < nlen)
-			*len = nlen;
-		*s = realloc(*s, *len);
-		if (!*s)
-			return -ENOMEM;
-		if (olen == 0)
-			**s = 0;
-	}
-	strcat(*s, a);
-	return 0;
-}
-- 
1.8.3.2

[PATCH V6 2/3] perf tools: parse the pmu event prefix and surfix

[kan.liang]

From: Kan Liang <kan.liang@intel.com>

There are two types of event formats for PMU events. E.g. el-abort OR
cpu/el-abort/. However, the lexer mistakenly recognizes the simple style
format as two events.

The parse_events_pmu_check function uses bsearch to search the name in
known pmu event list. It can tell the lexer that the name is a PE_NAME
or a PMU event name prefix or a PMU event name suffix. All these
information will be used for accurately parsing kernel PMU events.

The pmu events list will be read from sysfs at runtime.

Note: Currently, the patch only want to handle the PMU event name as
"a-b" and "a". The only exception, "stalled-cycles-frontend" and
"stalled-cycles-fronted", are already hardcoded in lexer.

Signed-off-by: Kan Liang <kan.liang@intel.com>
---
 tools/perf/util/parse-events.c | 112 +++++++++++++++++++++++++++++++++++++++++
 tools/perf/util/parse-events.h |  14 ++++++
 tools/perf/util/pmu.c          |  10 ----
 tools/perf/util/pmu.h          |  10 ++++
 4 files changed, 136 insertions(+), 10 deletions(-)

diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c
index b9174bc..20e01d1 100644
--- a/tools/perf/util/parse-events.c
+++ b/tools/perf/util/parse-events.c
@@ -30,6 +30,9 @@ extern int parse_events_debug;
 #endif
 int parse_events_parse(void *data, void *scanner);
 
+static struct perf_pmu_event_symbol *perf_pmu_events_list;
+static int perf_pmu_events_list_num;
+
 static struct event_symbol event_symbols_hw[PERF_COUNT_HW_MAX] = {
 	[PERF_COUNT_HW_CPU_CYCLES] = {
 		.symbol = "cpu-cycles",
@@ -864,6 +867,114 @@ int parse_events_name(struct list_head *list, char *name)
 	return 0;
 }
 
+static int
+comp_pmu(const void *p1, const void *p2)
+{
+	struct perf_pmu_event_symbol *pmu1 =
+			(struct perf_pmu_event_symbol *) p1;
+	struct perf_pmu_event_symbol *pmu2 =
+			(struct perf_pmu_event_symbol *) p2;
+
+	return strcmp(pmu1->symbol, pmu2->symbol);
+}
+
+/*
+ * Read the pmu events list from sysfs
+ * Save it into perf_pmu_events_list
+ */
+static void perf_pmu__parse_init(void)
+{
+
+	struct perf_pmu *pmu = NULL;
+	struct perf_pmu_alias *alias;
+	int len = 0;
+
+	pmu = perf_pmu__find("cpu");
+	if ((pmu == NULL) || list_empty(&pmu->aliases)) {
+		perf_pmu_events_list_num = -1;
+		return;
+	}
+	list_for_each_entry(alias, &pmu->aliases, list) {
+		if (strchr(alias->name, '-'))
+			len++;
+		len++;
+	}
+	perf_pmu_events_list =
+		malloc(sizeof(struct perf_pmu_event_symbol) * len);
+	perf_pmu_events_list_num = len;
+
+	len = 0;
+	pmu = perf_pmu__find("cpu");
+	list_for_each_entry(alias, &pmu->aliases, list) {
+		struct perf_pmu_event_symbol *p =
+			perf_pmu_events_list + len;
+		char *tmp = strchr(alias->name, '-');
+
+		if (tmp != NULL) {
+			p->symbol = malloc(tmp - alias->name + 1);
+			strlcpy(p->symbol, alias->name,
+				tmp - alias->name + 1);
+			p->type = KERNEL_PMU_EVENT_PREFIX;
+			tmp++;
+			p++;
+			p->symbol = malloc(strlen(tmp) + 1);
+			strcpy(p->symbol, tmp);
+			p->type = KERNEL_PMU_EVENT_SUFFIX;
+			len += 2;
+		} else {
+			p->symbol = malloc(strlen(alias->name) + 1);
+			strcpy(p->symbol, alias->name);
+			p->type = KERNEL_PMU_EVENT;
+			len++;
+		}
+	}
+	qsort(perf_pmu_events_list, len,
+		sizeof(struct perf_pmu_event_symbol), comp_pmu);
+}
+
+static void perf_pmu__parse_cleanup(void)
+{
+	if (perf_pmu_events_list_num > 0) {
+		struct perf_pmu_event_symbol *p;
+		int i;
+
+		for (i = 0; i < perf_pmu_events_list_num; i++) {
+			p = perf_pmu_events_list + i;
+			free(p->symbol);
+		}
+		free(perf_pmu_events_list);
+		perf_pmu_events_list = NULL;
+		perf_pmu_events_list_num = 0;
+	}
+}
+
+enum perf_pmu_event_symbol_type
+perf_pmu__parse_check(const char *name)
+{
+	struct perf_pmu_event_symbol p, *r;
+
+	/* scan kernel pmu events from sysfs if needed */
+	if (perf_pmu_events_list_num == 0)
+		perf_pmu__parse_init();
+	/*
+	 * name "cpu" could be prefix of cpu-cycles or cpu// events.
+	 * cpu-cycles has been handled by hardcode.
+	 * So it must be cpu// events, not kernel pmu event.
+	 */
+	if ((perf_pmu_events_list_num <= 0) || !strcmp(name, "cpu"))
+		return NONE_KERNEL_PMU_EVENT;
+
+	p.symbol = malloc(strlen(name) + 1);
+	strcpy(p.symbol, name);
+	r = bsearch(&p, perf_pmu_events_list,
+			(size_t) perf_pmu_events_list_num,
+			sizeof(struct perf_pmu_event_symbol), comp_pmu);
+	free(p.symbol);
+	if (r == NULL)
+		return NONE_KERNEL_PMU_EVENT;
+	return r->type;
+}
+
 static int parse_events__scanner(const char *str, void *data, int start_token)
 {
 	YY_BUFFER_STATE buffer;
@@ -918,6 +1029,7 @@ int parse_events(struct perf_evlist *evlist, const char *str)
 	int ret;
 
 	ret = parse_events__scanner(str, &data, PE_START_EVENTS);
+	perf_pmu__parse_cleanup();
 	if (!ret) {
 		int entries = data.idx - evlist->nr_entries;
 		perf_evlist__splice_list_tail(evlist, &data.list, entries);
diff --git a/tools/perf/util/parse-events.h b/tools/perf/util/parse-events.h
index df094b4..9f064e4 100644
--- a/tools/perf/util/parse-events.h
+++ b/tools/perf/util/parse-events.h
@@ -35,6 +35,18 @@ extern int parse_filter(const struct option *opt, const char *str, int unset);
 
 #define EVENTS_HELP_MAX (128*1024)
 
+enum perf_pmu_event_symbol_type {
+	NONE_KERNEL_PMU_EVENT,		/* not a PMU EVENT */
+	KERNEL_PMU_EVENT,		/* normal style PMU event */
+	KERNEL_PMU_EVENT_PREFIX,	/* prefix of pre-suf style event */
+	KERNEL_PMU_EVENT_SUFFIX,	/* suffix of pre-suf style event */
+};
+
+struct perf_pmu_event_symbol {
+	char	*symbol;
+	enum perf_pmu_event_symbol_type	type;
+};
+
 enum {
 	PARSE_EVENTS__TERM_TYPE_NUM,
 	PARSE_EVENTS__TERM_TYPE_STR,
@@ -95,6 +107,8 @@ int parse_events_add_breakpoint(struct list_head *list, int *idx,
 				void *ptr, char *type);
 int parse_events_add_pmu(struct list_head *list, int *idx,
 			 char *pmu , struct list_head *head_config);
+enum perf_pmu_event_symbol_type
+perf_pmu__parse_check(const char *name);
 void parse_events__set_leader(char *name, struct list_head *list);
 void parse_events_update_lists(struct list_head *list_event,
 			       struct list_head *list_all);
diff --git a/tools/perf/util/pmu.c b/tools/perf/util/pmu.c
index 22a4ad5..f358345 100644
--- a/tools/perf/util/pmu.c
+++ b/tools/perf/util/pmu.c
@@ -12,16 +12,6 @@
 #include "parse-events.h"
 #include "cpumap.h"
 
-#define UNIT_MAX_LEN	31 /* max length for event unit name */
-
-struct perf_pmu_alias {
-	char *name;
-	struct list_head terms; /* HEAD struct parse_events_term -> list */
-	struct list_head list;  /* ELEM */
-	char unit[UNIT_MAX_LEN+1];
-	double scale;
-};
-
 struct perf_pmu_format {
 	char *name;
 	int value;
diff --git a/tools/perf/util/pmu.h b/tools/perf/util/pmu.h
index 0f5c0a8..2020519 100644
--- a/tools/perf/util/pmu.h
+++ b/tools/perf/util/pmu.h
@@ -25,6 +25,16 @@ struct perf_pmu {
 	struct list_head list;    /* ELEM */
 };
 
+#define UNIT_MAX_LEN	31 /* max length for event unit name */
+
+struct perf_pmu_alias {
+	char *name;
+	struct list_head terms; /* HEAD struct parse_events_term -> list */
+	struct list_head list;  /* ELEM */
+	char unit[UNIT_MAX_LEN+1];
+	double scale;
+};
+
 struct perf_pmu *perf_pmu__find(const char *name);
 int perf_pmu__config(struct perf_pmu *pmu, struct perf_event_attr *attr,
 		     struct list_head *head_terms);
-- 
1.8.3.2

Re: [PATCH V6 2/3] perf tools: parse the pmu event prefix and surfix

[Jiri Olsa]

On Thu, Sep 11, 2014 at 03:08:58PM -0400, kan.liang@intel.com wrote:
> From: Kan Liang <kan.liang@intel.com>

typo in subject - s/surfix/suffix/

SNIP

> 
> diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c
> index b9174bc..20e01d1 100644
> --- a/tools/perf/util/parse-events.c
> +++ b/tools/perf/util/parse-events.c
> @@ -30,6 +30,9 @@ extern int parse_events_debug;
>  #endif
>  int parse_events_parse(void *data, void *scanner);
>  
> +static struct perf_pmu_event_symbol *perf_pmu_events_list;
> +static int perf_pmu_events_list_num;

please make some comment for the perf_pmu_events_list_num in here,
saying that 0 means 'ready to init', -1 means 'failed, dont try anymore',
and > 0 means 'initialized'

> +
>  static struct event_symbol event_symbols_hw[PERF_COUNT_HW_MAX] = {
>  	[PERF_COUNT_HW_CPU_CYCLES] = {
>  		.symbol = "cpu-cycles",
> @@ -864,6 +867,114 @@ int parse_events_name(struct list_head *list, char *name)
>  	return 0;
>  }
>  
> +static int
> +comp_pmu(const void *p1, const void *p2)
> +{
> +	struct perf_pmu_event_symbol *pmu1 =
> +			(struct perf_pmu_event_symbol *) p1;
> +	struct perf_pmu_event_symbol *pmu2 =
> +			(struct perf_pmu_event_symbol *) p2;

please keep it on one line, like:
	const struct perf_pmu_event_symbol *pmu1 = p1;
	const struct perf_pmu_event_symbol *pmu2 = p2;

> +
> +	return strcmp(pmu1->symbol, pmu2->symbol);
> +}
> +
> +/*
> + * Read the pmu events list from sysfs
> + * Save it into perf_pmu_events_list
> + */
> +static void perf_pmu__parse_init(void)
> +{
> +
> +	struct perf_pmu *pmu = NULL;
> +	struct perf_pmu_alias *alias;
> +	int len = 0;
> +
> +	pmu = perf_pmu__find("cpu");
> +	if ((pmu == NULL) || list_empty(&pmu->aliases)) {
> +		perf_pmu_events_list_num = -1;
> +		return;
> +	}
> +	list_for_each_entry(alias, &pmu->aliases, list) {
> +		if (strchr(alias->name, '-'))
> +			len++;
> +		len++;
> +	}
> +	perf_pmu_events_list =
> +		malloc(sizeof(struct perf_pmu_event_symbol) * len);

please keep above on one line

> +	perf_pmu_events_list_num = len;
> +
> +	len = 0;
> +	pmu = perf_pmu__find("cpu");

no need to lookup 'cpu' pmu again

> +	list_for_each_entry(alias, &pmu->aliases, list) {
> +		struct perf_pmu_event_symbol *p =
> +			perf_pmu_events_list + len;

please keep above on one line

> +		char *tmp = strchr(alias->name, '-');
> +
> +		if (tmp != NULL) {
> +			p->symbol = malloc(tmp - alias->name + 1);
> +			strlcpy(p->symbol, alias->name,
> +				tmp - alias->name + 1);
> +			p->type = KERNEL_PMU_EVENT_PREFIX;
> +			tmp++;
> +			p++;
> +			p->symbol = malloc(strlen(tmp) + 1);
> +			strcpy(p->symbol, tmp);
> +			p->type = KERNEL_PMU_EVENT_SUFFIX;
> +			len += 2;
> +		} else {
> +			p->symbol = malloc(strlen(alias->name) + 1);
> +			strcpy(p->symbol, alias->name);
> +			p->type = KERNEL_PMU_EVENT;
> +			len++;

I can see pattern above and missing check for failed malloc
how about using strdup and macro like:

...

#define SET_SYMBOL(str, type)	\
do {				\
	p->symbol = str;	\
	if (!p->symbol)		\
		goto err;	\
	p->type = type;		\
} while (0)

if (tmp != NULL) {
	SET_SYMBOL(strndup(alias->name, tmp - alias->name + 1), KERNEL_PMU_EVENT_PREFIX);
	p++;
	SET_SYMBOL(strdup(++tmp), KERNEL_PMU_EVENT_SUFFIX);
	len += 2;
} else {
	SET_SYMBOL(strdup(alias->name), KERNEL_PMU_EVENT);
}

...

err:
	perf_pmu__parse_cleanup

> +		}
> +	}
> +	qsort(perf_pmu_events_list, len,
> +		sizeof(struct perf_pmu_event_symbol), comp_pmu);
> +}
> +
> +static void perf_pmu__parse_cleanup(void)
> +{
> +	if (perf_pmu_events_list_num > 0) {
> +		struct perf_pmu_event_symbol *p;
> +		int i;
> +
> +		for (i = 0; i < perf_pmu_events_list_num; i++) {
> +			p = perf_pmu_events_list + i;
> +			free(p->symbol);
> +		}
> +		free(perf_pmu_events_list);
> +		perf_pmu_events_list = NULL;
> +		perf_pmu_events_list_num = 0;
> +	}
> +}
> +
> +enum perf_pmu_event_symbol_type
> +perf_pmu__parse_check(const char *name)
> +{
> +	struct perf_pmu_event_symbol p, *r;
> +
> +	/* scan kernel pmu events from sysfs if needed */
> +	if (perf_pmu_events_list_num == 0)
> +		perf_pmu__parse_init();
> +	/*
> +	 * name "cpu" could be prefix of cpu-cycles or cpu// events.
> +	 * cpu-cycles has been handled by hardcode.
> +	 * So it must be cpu// events, not kernel pmu event.
> +	 */
> +	if ((perf_pmu_events_list_num <= 0) || !strcmp(name, "cpu"))
> +		return NONE_KERNEL_PMU_EVENT;
> +
> +	p.symbol = malloc(strlen(name) + 1);
> +	strcpy(p.symbol, name);

please use strdup and check for error,
but I think you could use name directly no? like:
	p.symbol = name;

> +	r = bsearch(&p, perf_pmu_events_list,
> +			(size_t) perf_pmu_events_list_num,
> +			sizeof(struct perf_pmu_event_symbol), comp_pmu);
> +	free(p.symbol);
> +	if (r == NULL)
> +		return NONE_KERNEL_PMU_EVENT;
> +	return r->type;

	return r ? r->type : NONE_KERNEL_PMU_EVENT;

> +}
> +
>  static int parse_events__scanner(const char *str, void *data, int start_token)
>  {
>  	YY_BUFFER_STATE buffer;
> @@ -918,6 +1029,7 @@ int parse_events(struct perf_evlist *evlist, const char *str)
>  	int ret;
>  
>  	ret = parse_events__scanner(str, &data, PE_START_EVENTS);
> +	perf_pmu__parse_cleanup();
>  	if (!ret) {
>  		int entries = data.idx - evlist->nr_entries;
>  		perf_evlist__splice_list_tail(evlist, &data.list, entries);
> diff --git a/tools/perf/util/parse-events.h b/tools/perf/util/parse-events.h
> index df094b4..9f064e4 100644
> --- a/tools/perf/util/parse-events.h
> +++ b/tools/perf/util/parse-events.h
> @@ -35,6 +35,18 @@ extern int parse_filter(const struct option *opt, const char *str, int unset);
>  
>  #define EVENTS_HELP_MAX (128*1024)
>  
> +enum perf_pmu_event_symbol_type {
> +	NONE_KERNEL_PMU_EVENT,		/* not a PMU EVENT */
> +	KERNEL_PMU_EVENT,		/* normal style PMU event */
> +	KERNEL_PMU_EVENT_PREFIX,	/* prefix of pre-suf style event */
> +	KERNEL_PMU_EVENT_SUFFIX,	/* suffix of pre-suf style event */
> +};

maybe following enum names go better with the enum name:
	PMU_EVENT_SYMBOL_ERR
	PMU_EVENT_SYMBOL
	PMU_EVENT_SYMBOL_PREFIX
	PMU_EVENT_SYMBOL_SUFFIX

> +
> +struct perf_pmu_event_symbol {
> +	char	*symbol;
> +	enum perf_pmu_event_symbol_type	type;
> +};
> +
>  enum {
>  	PARSE_EVENTS__TERM_TYPE_NUM,
>  	PARSE_EVENTS__TERM_TYPE_STR,
> @@ -95,6 +107,8 @@ int parse_events_add_breakpoint(struct list_head *list, int *idx,
>  				void *ptr, char *type);
>  int parse_events_add_pmu(struct list_head *list, int *idx,
>  			 char *pmu , struct list_head *head_config);
> +enum perf_pmu_event_symbol_type
> +perf_pmu__parse_check(const char *name);
>  void parse_events__set_leader(char *name, struct list_head *list);
>  void parse_events_update_lists(struct list_head *list_event,
>  			       struct list_head *list_all);
> diff --git a/tools/perf/util/pmu.c b/tools/perf/util/pmu.c
> index 22a4ad5..f358345 100644
> --- a/tools/perf/util/pmu.c
> +++ b/tools/perf/util/pmu.c
> @@ -12,16 +12,6 @@
>  #include "parse-events.h"
>  #include "cpumap.h"
>  
> -#define UNIT_MAX_LEN	31 /* max length for event unit name */
> -
> -struct perf_pmu_alias {
> -	char *name;
> -	struct list_head terms; /* HEAD struct parse_events_term -> list */
> -	struct list_head list;  /* ELEM */
> -	char unit[UNIT_MAX_LEN+1];
> -	double scale;
> -};
> -
>  struct perf_pmu_format {
>  	char *name;
>  	int value;
> diff --git a/tools/perf/util/pmu.h b/tools/perf/util/pmu.h
> index 0f5c0a8..2020519 100644
> --- a/tools/perf/util/pmu.h
> +++ b/tools/perf/util/pmu.h
> @@ -25,6 +25,16 @@ struct perf_pmu {
>  	struct list_head list;    /* ELEM */
>  };
>  
> +#define UNIT_MAX_LEN	31 /* max length for event unit name */
> +
> +struct perf_pmu_alias {
> +	char *name;
> +	struct list_head terms; /* HEAD struct parse_events_term -> list */
> +	struct list_head list;  /* ELEM */
> +	char unit[UNIT_MAX_LEN+1];
> +	double scale;
> +};
> +
>  struct perf_pmu *perf_pmu__find(const char *name);
>  int perf_pmu__config(struct perf_pmu *pmu, struct perf_event_attr *attr,
>  		     struct list_head *head_terms);
> -- 
> 1.8.3.2
> 

RE: [PATCH V6 2/3] perf tools: parse the pmu event prefix and surfix

[Liang, Kan]

> >
> > +static int
> > +comp_pmu(const void *p1, const void *p2) {
> > +	struct perf_pmu_event_symbol *pmu1 =
> > +			(struct perf_pmu_event_symbol *) p1;
> > +	struct perf_pmu_event_symbol *pmu2 =
> > +			(struct perf_pmu_event_symbol *) p2;
> 
> please keep it on one line, like:
> 	const struct perf_pmu_event_symbol *pmu1 = p1;
> 	const struct perf_pmu_event_symbol *pmu2 = p2;
> 


Removing (struct perf_pmu_event_symbol *) will cause compiler error as below.

util/parse-events.c:877:39: error: initialization discards âconstâ qualifier from pointer target type [-Werror]
  struct perf_pmu_event_symbol *pmu1 = p1;

If we keep (struct perf_pmu_event_symbol *), we may not keep it on one line. The line will over 80 characters. There will be error from checkpatch.pl.


> > +	perf_pmu_events_list =
> > +		malloc(sizeof(struct perf_pmu_event_symbol) * len);
> 
> please keep above on one line

If so, the line will over 80 characters. There will be error from checkpatch.pl.

Kan

Re: [PATCH V6 2/3] perf tools: parse the pmu event prefix and surfix

[Jiri Olsa]

On Thu, Oct 02, 2014 at 05:33:07PM +0000, Liang, Kan wrote:
> 
> 
> > >
> > > +static int
> > > +comp_pmu(const void *p1, const void *p2) {
> > > +	struct perf_pmu_event_symbol *pmu1 =
> > > +			(struct perf_pmu_event_symbol *) p1;
> > > +	struct perf_pmu_event_symbol *pmu2 =
> > > +			(struct perf_pmu_event_symbol *) p2;
> > 
> > please keep it on one line, like:
> > 	const struct perf_pmu_event_symbol *pmu1 = p1;
> > 	const struct perf_pmu_event_symbol *pmu2 = p2;
> > 
> 
> 
> Removing (struct perf_pmu_event_symbol *) will cause compiler error as below.
> 
> util/parse-events.c:877:39: error: initialization discards âconstâ qualifier from pointer target type [-Werror]
>   struct perf_pmu_event_symbol *pmu1 = p1;
> 
> If we keep (struct perf_pmu_event_symbol *), we may not keep it on one line. The line will over 80 characters. There will be error from checkpatch.pl.
> 
> 
> > > +	perf_pmu_events_list =
> > > +		malloc(sizeof(struct perf_pmu_event_symbol) * len);
> > 
> > please keep above on one line
> 
> If so, the line will over 80 characters. There will be error from checkpatch.pl.

AFAIK we are not that strict on this warning.. I think the current
notion is to keep the code easy to read with reasonable breakages
of this rule

jirka

[PATCH V6 3/3] perf tools: Add support to new style format of kernel PMU event

[kan.liang]

From: Kan Liang <kan.liang@intel.com>

Add new rules for kernel PMU event.

Currently, the patch only want to handle the PMU event name as "a-b" and
"a".

event_pmu:
PE_KERNEL_PMU_EVENT
|
PE_PMU_EVENT_PRE '-' PE_PMU_EVENT_SUF

PE_KERNEL_PMU_EVENT token is for
cycles-ct/cycles-t/mem-loads/mem-stores.
The prefix cycles is mixed up with cpu-cycles.
loads and stores are mixed up with cache event
So they have to be hardcode in lex.

PE_PMU_EVENT_PRE and PE_PMU_EVENT_SUF tokens are for other PMU
events.
The lex looks generic identifier up in the table and return the matched
token. If there is no match, generic PE_NAME token will be return.

Using the rules, kernel PMU event could use new style format without //

so you can use

perf record -e mem-loads ...

instead of

perf record -e cpu/mem-loads/

Signed-off-by: Kan Liang <kan.liang@intel.com>
---
 tools/perf/util/parse-events.l | 30 +++++++++++++++++++++++++++++-
 tools/perf/util/parse-events.y | 40 ++++++++++++++++++++++++++++++++++++++++
 2 files changed, 69 insertions(+), 1 deletion(-)

diff --git a/tools/perf/util/parse-events.l b/tools/perf/util/parse-events.l
index 3432995..ea970cf 100644
--- a/tools/perf/util/parse-events.l
+++ b/tools/perf/util/parse-events.l
@@ -51,6 +51,24 @@ static int str(yyscan_t scanner, int token)
 	return token;
 }
 
+static int pmu_str_check(yyscan_t scanner)
+{
+	YYSTYPE *yylval = parse_events_get_lval(scanner);
+	char *text = parse_events_get_text(scanner);
+
+	yylval->str = strdup(text);
+	switch (perf_pmu__parse_check(text)) {
+		case KERNEL_PMU_EVENT_PREFIX:
+			return PE_PMU_EVENT_PRE;
+		case KERNEL_PMU_EVENT_SUFFIX:
+			return PE_PMU_EVENT_SUF;
+		case KERNEL_PMU_EVENT:
+			return PE_KERNEL_PMU_EVENT;
+		default:
+			return PE_NAME;
+	}
+}
+
 static int sym(yyscan_t scanner, int type, int config)
 {
 	YYSTYPE *yylval = parse_events_get_lval(scanner);
@@ -178,6 +196,16 @@ alignment-faults				{ return sym(yyscanner, PERF_TYPE_SOFTWARE, PERF_COUNT_SW_AL
 emulation-faults				{ return sym(yyscanner, PERF_TYPE_SOFTWARE, PERF_COUNT_SW_EMULATION_FAULTS); }
 dummy						{ return sym(yyscanner, PERF_TYPE_SOFTWARE, PERF_COUNT_SW_DUMMY); }
 
+	/*
+	 * We have to handle the kernel PMU event cycles-ct/cycles-t/mem-loads/mem-stores separately.
+	 * Because the prefix cycles is mixed up with cpu-cycles.
+	 * loads and stores are mixed up with cache event
+	 */
+cycles-ct					{ return str(yyscanner, PE_KERNEL_PMU_EVENT); }
+cycles-t					{ return str(yyscanner, PE_KERNEL_PMU_EVENT); }
+mem-loads					{ return str(yyscanner, PE_KERNEL_PMU_EVENT); }
+mem-stores					{ return str(yyscanner, PE_KERNEL_PMU_EVENT); }
+
 L1-dcache|l1-d|l1d|L1-data		|
 L1-icache|l1-i|l1i|L1-instruction	|
 LLC|L2					|
@@ -199,7 +227,7 @@ r{num_raw_hex}		{ return raw(yyscanner); }
 {num_hex}		{ return value(yyscanner, 16); }
 
 {modifier_event}	{ return str(yyscanner, PE_MODIFIER_EVENT); }
-{name}			{ return str(yyscanner, PE_NAME); }
+{name}			{ return pmu_str_check(yyscanner); }
 "/"			{ BEGIN(config); return '/'; }
 -			{ return '-'; }
 ,			{ BEGIN(event); return ','; }
diff --git a/tools/perf/util/parse-events.y b/tools/perf/util/parse-events.y
index 55fab6a..3f26c03 100644
--- a/tools/perf/util/parse-events.y
+++ b/tools/perf/util/parse-events.y
@@ -47,6 +47,7 @@ static inc_group_count(struct list_head *list,
 %token PE_NAME_CACHE_TYPE PE_NAME_CACHE_OP_RESULT
 %token PE_PREFIX_MEM PE_PREFIX_RAW PE_PREFIX_GROUP
 %token PE_ERROR
+%token PE_PMU_EVENT_PRE PE_PMU_EVENT_SUF PE_KERNEL_PMU_EVENT
 %type <num> PE_VALUE
 %type <num> PE_VALUE_SYM_HW
 %type <num> PE_VALUE_SYM_SW
@@ -58,6 +59,7 @@ static inc_group_count(struct list_head *list,
 %type <str> PE_MODIFIER_EVENT
 %type <str> PE_MODIFIER_BP
 %type <str> PE_EVENT_NAME
+%type <str> PE_PMU_EVENT_PRE PE_PMU_EVENT_SUF PE_KERNEL_PMU_EVENT
 %type <num> value_sym
 %type <head> event_config
 %type <term> event_term
@@ -220,6 +222,44 @@ PE_NAME '/' '/'
 	ABORT_ON(parse_events_add_pmu(list, &data->idx, $1, NULL));
 	$$ = list;
 }
+|
+PE_KERNEL_PMU_EVENT
+{
+	struct parse_events_evlist *data = _data;
+	struct list_head *head;
+	struct parse_events_term *term;
+	struct list_head *list;
+
+	ALLOC_LIST(head);
+	ABORT_ON(parse_events_term__num(&term, PARSE_EVENTS__TERM_TYPE_USER,
+					$1, 1));
+	list_add_tail(&term->list, head);
+
+	ALLOC_LIST(list);
+	ABORT_ON(parse_events_add_pmu(list, &data->idx, "cpu", head));
+	parse_events__free_terms(head);
+	$$ = list;
+}
+|
+PE_PMU_EVENT_PRE '-' PE_PMU_EVENT_SUF
+{
+	struct parse_events_evlist *data = _data;
+	struct list_head *head;
+	struct parse_events_term *term;
+	struct list_head *list;
+	char pmu_name[128];
+	snprintf(&pmu_name, 128, "%s-%s", $1, $3);
+
+	ALLOC_LIST(head);
+	ABORT_ON(parse_events_term__num(&term, PARSE_EVENTS__TERM_TYPE_USER,
+					&pmu_name, 1));
+	list_add_tail(&term->list, head);
+
+	ALLOC_LIST(list);
+	ABORT_ON(parse_events_add_pmu(list, &data->idx, "cpu", head));
+	parse_events__free_terms(head);
+	$$ = list;
+}
 
 value_sym:
 PE_VALUE_SYM_HW
-- 
1.8.3.2

Re: [PATCH V6 0/3] perf tools: pmu event new style format fix

[Jiri Olsa]

On Thu, Sep 11, 2014 at 03:08:56PM -0400, kan.liang@intel.com wrote:
> From: Kan Liang <kan.liang@intel.com>
> 
> There are two types of pmu event stytle formats, "pmu_event_name"
> or "cpu/pmu_event_name/". However, there is a bug on supporting these
> two formats, especially when they mixed with other perf events.
> The patch set intends to fix this issue.
> 
> The patch set has been tested on my haswell.
> Here is the test script I used for this issue.
> (Note: please make sure that your test system supports TSX and
> L1-dcache-loads events. Otherwise, you may want to change the events 
> to other pmu events.)

any chance changing this script to work for everyone?
and make it part of theautomated tests suite?

also how about something like in the attached change

jirka


---
diff --git a/tools/perf/tests/parse-events.c b/tools/perf/tests/parse-events.c
index 5941927a4b7f..2b7c48474761 100644
--- a/tools/perf/tests/parse-events.c
+++ b/tools/perf/tests/parse-events.c
@@ -1554,6 +1554,15 @@ static int test_pmu_events(void)
 		e.check = test__checkevent_pmu_events;
 
 		ret = test_event(&e);
+		if (ret)
+			break;
+
+		snprintf(name, MAX_NAME, "event=%s", ent->d_name);
+
+		e.name  = name;
+		e.check = test__checkevent_pmu_events;
+
+		ret = test_event(&e);
 #undef MAX_NAME
 	}

RE: [PATCH V6 0/3] perf tools: pmu event new style format fix

[Liang, Kan]

> 
> On Thu, Sep 11, 2014 at 03:08:56PM -0400, kan.liang@intel.com wrote:
> > From: Kan Liang <kan.liang@intel.com>
> >
> > There are two types of pmu event stytle formats, "pmu_event_name"
> > or "cpu/pmu_event_name/". However, there is a bug on supporting these
> > two formats, especially when they mixed with other perf events.
> > The patch set intends to fix this issue.
> >
> > The patch set has been tested on my haswell.
> > Here is the test script I used for this issue.
> > (Note: please make sure that your test system supports TSX and
> > L1-dcache-loads events. Otherwise, you may want to change the events
> > to other pmu events.)
> 
> any chance changing this script to work for everyone?
> and make it part of theautomated tests suite?
> 
> also how about something like in the attached change

It cannot trigger the issue, since current perf will automatically add cpu//  for pmu event.
We have to test at least two events with different style.
How about the change as below?

--- a/tools/perf/tests/parse-events.c
+++ b/tools/perf/tests/parse-events.c
@@ -457,6 +457,36 @@ static int test__checkevent_pmu_events(struct perf_evlist *evlist)
        return 0;
 }

+
+static int test__checkevent_pmu_events_mix(struct perf_evlist *evlist)
+{
+       struct perf_evsel *evsel = perf_evlist__first(evlist);
+
+       /* pmu-event:u */
+       TEST_ASSERT_VAL("wrong number of entries", 2 == evlist->nr_entries);
+       TEST_ASSERT_VAL("wrong exclude_user",
+                       !evsel->attr.exclude_user);
+       TEST_ASSERT_VAL("wrong exclude_kernel",
+                       evsel->attr.exclude_kernel);
+       TEST_ASSERT_VAL("wrong exclude_hv", evsel->attr.exclude_hv);
+       TEST_ASSERT_VAL("wrong precise_ip", !evsel->attr.precise_ip);
+       TEST_ASSERT_VAL("wrong pinned", !evsel->attr.pinned);
+
+       /* cpu/pmu-event/u*/
+       evsel = perf_evsel__next(evsel);
+       TEST_ASSERT_VAL("wrong number of entries", 2 == evlist->nr_entries);
+       TEST_ASSERT_VAL("wrong type", PERF_TYPE_RAW == evsel->attr.type);
+       TEST_ASSERT_VAL("wrong exclude_user",
+                       !evsel->attr.exclude_user);
+       TEST_ASSERT_VAL("wrong exclude_kernel",
+                       evsel->attr.exclude_kernel);
+       TEST_ASSERT_VAL("wrong exclude_hv", evsel->attr.exclude_hv);
+       TEST_ASSERT_VAL("wrong precise_ip", !evsel->attr.precise_ip);
+       TEST_ASSERT_VAL("wrong pinned", !evsel->attr.pinned);
+
+       return 0;
+}
+
 static int test__checkterms_simple(struct list_head *terms)
 {
        struct parse_events_term *term;
@@ -1554,6 +1584,12 @@ static int test_pmu_events(void)
                e.check = test__checkevent_pmu_events;

                ret = test_event(&e);
+               if (ret)
+                       break;
+               snprintf(name, MAX_NAME, "%s:u,cpu/event=%s/u", ent->d_name, ent->d_name);
+               e.name  = name;
+               e.check = test__checkevent_pmu_events_mix;
+               ret = test_event(&e);
 #undef MAX_NAME
        }


Thanks,
Kan


> 
> jirka
> 
> 
> ---
> diff --git a/tools/perf/tests/parse-events.c b/tools/perf/tests/parse-
> events.c index 5941927a4b7f..2b7c48474761 100644
> --- a/tools/perf/tests/parse-events.c
> +++ b/tools/perf/tests/parse-events.c
> @@ -1554,6 +1554,15 @@ static int test_pmu_events(void)
>  		e.check = test__checkevent_pmu_events;
> 
>  		ret = test_event(&e);
> +		if (ret)
> +			break;
> +
> +		snprintf(name, MAX_NAME, "event=%s", ent->d_name);
> +
> +		e.name  = name;
> +		e.check = test__checkevent_pmu_events;
> +
> +		ret = test_event(&e);
>  #undef MAX_NAME
>  	}

Re: [PATCH V6 0/3] perf tools: pmu event new style format fix

[Jiri Olsa]

On Thu, Oct 02, 2014 at 05:33:28PM +0000, Liang, Kan wrote:
> 
> 
> > 
> > On Thu, Sep 11, 2014 at 03:08:56PM -0400, kan.liang@intel.com wrote:
> > > From: Kan Liang <kan.liang@intel.com>
> > >
> > > There are two types of pmu event stytle formats, "pmu_event_name"
> > > or "cpu/pmu_event_name/". However, there is a bug on supporting these
> > > two formats, especially when they mixed with other perf events.
> > > The patch set intends to fix this issue.
> > >
> > > The patch set has been tested on my haswell.
> > > Here is the test script I used for this issue.
> > > (Note: please make sure that your test system supports TSX and
> > > L1-dcache-loads events. Otherwise, you may want to change the events
> > > to other pmu events.)
> > 
> > any chance changing this script to work for everyone?
> > and make it part of theautomated tests suite?
> > 
> > also how about something like in the attached change
> 
> It cannot trigger the issue, since current perf will automatically add cpu//  for pmu event.
> We have to test at least two events with different style.
> How about the change as below?

looks good,

jirka

> 
> --- a/tools/perf/tests/parse-events.c
> +++ b/tools/perf/tests/parse-events.c
> @@ -457,6 +457,36 @@ static int test__checkevent_pmu_events(struct perf_evlist *evlist)
>         return 0;
>  }
> 
> +
> +static int test__checkevent_pmu_events_mix(struct perf_evlist *evlist)
> +{
> +       struct perf_evsel *evsel = perf_evlist__first(evlist);
> +
> +       /* pmu-event:u */
> +       TEST_ASSERT_VAL("wrong number of entries", 2 == evlist->nr_entries);
> +       TEST_ASSERT_VAL("wrong exclude_user",
> +                       !evsel->attr.exclude_user);
> +       TEST_ASSERT_VAL("wrong exclude_kernel",
> +                       evsel->attr.exclude_kernel);
> +       TEST_ASSERT_VAL("wrong exclude_hv", evsel->attr.exclude_hv);
> +       TEST_ASSERT_VAL("wrong precise_ip", !evsel->attr.precise_ip);
> +       TEST_ASSERT_VAL("wrong pinned", !evsel->attr.pinned);
> +
> +       /* cpu/pmu-event/u*/
> +       evsel = perf_evsel__next(evsel);
> +       TEST_ASSERT_VAL("wrong number of entries", 2 == evlist->nr_entries);
> +       TEST_ASSERT_VAL("wrong type", PERF_TYPE_RAW == evsel->attr.type);
> +       TEST_ASSERT_VAL("wrong exclude_user",
> +                       !evsel->attr.exclude_user);
> +       TEST_ASSERT_VAL("wrong exclude_kernel",
> +                       evsel->attr.exclude_kernel);
> +       TEST_ASSERT_VAL("wrong exclude_hv", evsel->attr.exclude_hv);
> +       TEST_ASSERT_VAL("wrong precise_ip", !evsel->attr.precise_ip);
> +       TEST_ASSERT_VAL("wrong pinned", !evsel->attr.pinned);
> +
> +       return 0;
> +}
> +
>  static int test__checkterms_simple(struct list_head *terms)
>  {
>         struct parse_events_term *term;
> @@ -1554,6 +1584,12 @@ static int test_pmu_events(void)
>                 e.check = test__checkevent_pmu_events;
> 
>                 ret = test_event(&e);
> +               if (ret)
> +                       break;
> +               snprintf(name, MAX_NAME, "%s:u,cpu/event=%s/u", ent->d_name, ent->d_name);
> +               e.name  = name;
> +               e.check = test__checkevent_pmu_events_mix;
> +               ret = test_event(&e);
>  #undef MAX_NAME
>         }
> 
> 
> Thanks,
> Kan
> 
> 
> > 
> > jirka
> > 
> > 
> > ---
> > diff --git a/tools/perf/tests/parse-events.c b/tools/perf/tests/parse-
> > events.c index 5941927a4b7f..2b7c48474761 100644
> > --- a/tools/perf/tests/parse-events.c
> > +++ b/tools/perf/tests/parse-events.c
> > @@ -1554,6 +1554,15 @@ static int test_pmu_events(void)
> >  		e.check = test__checkevent_pmu_events;
> > 
> >  		ret = test_event(&e);
> > +		if (ret)
> > +			break;
> > +
> > +		snprintf(name, MAX_NAME, "event=%s", ent->d_name);
> > +
> > +		e.name  = name;
> > +		e.check = test__checkevent_pmu_events;
> > +
> > +		ret = test_event(&e);
> >  #undef MAX_NAME
> >  	}