Skip to contentSkip to conversation

perf tools: Fix pattern matching for same substring used in different pmu type

PATCH Superseded Jin Yao 3 participants 8 messages Jun 9, 2021 – Jun 30, 2021 perf events

v1 of 3 Superseded Latest is v3 →

lore
Share & tools
lore

Jin Yao

author
#0 lore raw

PATCH 1/1 perf tools: Fix pattern matching for same substring used in different pmu type

Some different pmu types may have same substring. For example, on Icelake server, we have pmu types "uncore_imc" and "uncore_imc_free_running". Both pmu types have substring "uncore_imc". But the parser would wrongly think they are the same pmu type.

We enable an imc event,
perf stat -e uncore_imc/event=0xe3/ -a -- sleep 1

Perf actually expands the event to:
uncore_imc_0/event=0xe3/
uncore_imc_1/event=0xe3/
uncore_imc_2/event=0xe3/
uncore_imc_3/event=0xe3/
uncore_imc_4/event=0xe3/
uncore_imc_5/event=0xe3/
uncore_imc_6/event=0xe3/
uncore_imc_7/event=0xe3/
uncore_imc_free_running_0/event=0xe3/
uncore_imc_free_running_1/event=0xe3/
uncore_imc_free_running_3/event=0xe3/
uncore_imc_free_running_4/event=0xe3/

That's because the "uncore_imc_free_running" matches the pattern "uncore_imc*".

Now we check that the last characters of pmu name is '_<digit>'.

Fixes b2b9d3a3f021 ("perf pmu: Support wildcards on pmu name in dynamic pmu events")
---
 tools/perf/util/parse-events.y |  2 ++
 tools/perf/util/pmu.c          | 25 ++++++++++++++++++++++++-
 tools/perf/util/pmu.h          |  1 +
 3 files changed, 27 insertions(+), 1 deletion(-)
tools/perf/util/parse-events.y +2 -0
index aba12a4d488e..7a694c7f7f1a 100644
@@ -317,6 +317,8 @@ event_pmu_name opt_pmu_config
 			    strncmp($1, "uncore_", 7))
 				name += 7;
 			if (!fnmatch(pattern, name, 0)) {
+				if (!perf_pmu__valid_suffix($1, name))
+					continue;
 				if (parse_events_copy_term_list(orig_terms, &terms))
 					CLEANUP_YYABORT;
 				if (!parse_events_add_pmu(_parse_state, list, pmu->name, terms, true, false))
tools/perf/util/pmu.c +24 -1
index 88c8ecdc60b0..78af01959830 100644
@@ -3,6 +3,7 @@
 #include <linux/compiler.h>
 #include <linux/string.h>
 #include <linux/zalloc.h>
+#include <linux/ctype.h>
 #include <subcmd/pager.h>
 #include <sys/types.h>
 #include <errno.h>
@@ -768,7 +769,7 @@ bool pmu_uncore_alias_match(const char *pmu_name, const char *name)
 	 */
 	for (; tok; name += strlen(tok), tok = strtok_r(NULL, ",", &tmp)) {
 		name = strstr(name, tok);
-		if (!name) {
+		if (!name || !perf_pmu__valid_suffix(tok, (char *)name)) {
 			res = false;
 			goto out;
 		}
@@ -1872,3 +1873,25 @@ bool perf_pmu__has_hybrid(void)
 
 	return !list_empty(&perf_pmu__hybrid_pmus);
 }
+
+bool perf_pmu__valid_suffix(char *tok, char *pmu_name)
+{
+	char *p;
+
+	/*
+	 * The pmu_name has substring tok. If the format of
+	 * pmu_name is <tok> or <tok>_<digit>, return true.
+	 */
+	p = pmu_name + strlen(tok);
+	if (*p == 0)
+		return true;
+
+	if (*p != '_')
+		return false;
+
+	++p;
+	if (*p == 0 || !isdigit(*p))
+		return false;
+
+	return true;
+}
tools/perf/util/pmu.h +1 -0
index a790ef758171..ebfd2b71532b 100644
@@ -133,5 +133,6 @@ void perf_pmu__warn_invalid_config(struct perf_pmu *pmu, __u64 config,
 				   char *name);
 
 bool perf_pmu__has_hybrid(void);
+bool perf_pmu__valid_suffix(char *tok, char *pmu_name);
 
 #endif /* __PMU_H */
Signature · 1 line
2.17.1
author
#1 lore raw

And, though this patch is to fix the uncore_imc/uncore_imc_free_running mismatching issue. We are also surprised to see it can solve another hybrid PMU issue on Alderlake.

On Alderlake,

# ./perf stat -e cpu/event=0xc2,umask=0x2/ true

Performance counter stats for 'true':

           702,246      cpu_core/event=0xc2,umask=0x2/
     <not counted>      cpu_atom/event=0xc2,umask=0x2/

It should error out with the wrong PMU rather than using 'cpu_core' and'cpu_atom' instead.

This is still the pattern matching issue. The pattern is "cpu*". Both "cpu_core" and "cpu_atom" can match the pattern "cpu*", so the parser wrongly thinks they are the same PMU type.

Now with this patch,

# ./perf stat -e cpu/cpu-cycles/ true
event syntax error: 'cpu/cpu-cycles/'
                     \___ Cannot find PMU `cpu'. Missing kernel support?
Run 'perf list' for a list of valid events

Thanks
Jin Yao

Quoted 108 lines · Jin Yao

Some different pmu types may have same substring. For example, on Icelake server, we have pmu types "uncore_imc" and "uncore_imc_free_running". Both pmu types have substring "uncore_imc". But the parser would wrongly think they are the same pmu type.

We enable an imc event,
perf stat -e uncore_imc/event=0xe3/ -a -- sleep 1

Perf actually expands the event to:
uncore_imc_0/event=0xe3/
uncore_imc_1/event=0xe3/
uncore_imc_2/event=0xe3/
uncore_imc_3/event=0xe3/
uncore_imc_4/event=0xe3/
uncore_imc_5/event=0xe3/
uncore_imc_6/event=0xe3/
uncore_imc_7/event=0xe3/
uncore_imc_free_running_0/event=0xe3/
uncore_imc_free_running_1/event=0xe3/
uncore_imc_free_running_3/event=0xe3/
uncore_imc_free_running_4/event=0xe3/

That's because the "uncore_imc_free_running" matches the pattern "uncore_imc*".

Now we check that the last characters of pmu name is '_<digit>'.

Fixes b2b9d3a3f021 ("perf pmu: Support wildcards on pmu name in dynamic pmu events")
---
  tools/perf/util/parse-events.y |  2 ++
  tools/perf/util/pmu.c          | 25 ++++++++++++++++++++++++-
  tools/perf/util/pmu.h          |  1 +
  3 files changed, 27 insertions(+), 1 deletion(-)
tools/perf/util/parse-events.y +2 -0
index aba12a4d488e..7a694c7f7f1a 100644
@@ -317,6 +317,8 @@ event_pmu_name opt_pmu_config
  			    strncmp($1, "uncore_", 7))
  				name += 7;
  			if (!fnmatch(pattern, name, 0)) {
+				if (!perf_pmu__valid_suffix($1, name))
+					continue;
  				if (parse_events_copy_term_list(orig_terms, &terms))
  					CLEANUP_YYABORT;
  				if (!parse_events_add_pmu(_parse_state, list, pmu->name, terms, true, false))
tools/perf/util/pmu.c +24 -1
index 88c8ecdc60b0..78af01959830 100644
@@ -3,6 +3,7 @@
  #include <linux/compiler.h>
  #include <linux/string.h>
  #include <linux/zalloc.h>
+#include <linux/ctype.h>
  #include <subcmd/pager.h>
  #include <sys/types.h>
  #include <errno.h>
@@ -768,7 +769,7 @@ bool pmu_uncore_alias_match(const char *pmu_name, const char *name)
  	 */
  	for (; tok; name += strlen(tok), tok = strtok_r(NULL, ",", &tmp)) {
  		name = strstr(name, tok);
-		if (!name) {
+		if (!name || !perf_pmu__valid_suffix(tok, (char *)name)) {
  			res = false;
  			goto out;
  		}
@@ -1872,3 +1873,25 @@ bool perf_pmu__has_hybrid(void)
 
  	return !list_empty(&perf_pmu__hybrid_pmus);
  }
+
+bool perf_pmu__valid_suffix(char *tok, char *pmu_name)
+{
+	char *p;
+
+	/*
+	 * The pmu_name has substring tok. If the format of
+	 * pmu_name is <tok> or <tok>_<digit>, return true.
+	 */
+	p = pmu_name + strlen(tok);
+	if (*p == 0)
+		return true;
+
+	if (*p != '_')
+		return false;
+
+	++p;
+	if (*p == 0 || !isdigit(*p))
+		return false;
+
+	return true;
+}
tools/perf/util/pmu.h +1 -0
index a790ef758171..ebfd2b71532b 100644
@@ -133,5 +133,6 @@ void perf_pmu__warn_invalid_config(struct perf_pmu *pmu, __u64 config,
  				   char *name);
 
  bool perf_pmu__has_hybrid(void);
+bool perf_pmu__valid_suffix(char *tok, char *pmu_name);
 
  #endif /* __PMU_H */
author
#2 lore raw

Hi Arnaldo, Jiri,

Any comments for this bug fix patch?

The issue does impact some uncore events and even some metrics.

Thanks
Jin Yao

Quoted 131 lines · Yao

And, though this patch is to fix the uncore_imc/uncore_imc_free_running mismatching issue. We are also surprised to see it can solve another hybrid PMU issue on Alderlake.

On Alderlake,

# ./perf stat -e cpu/event=0xc2,umask=0x2/ true

Performance counter stats for 'true':

           702,246      cpu_core/event=0xc2,umask=0x2/ <not counted>      cpu_atom/event=0xc2,umask=0x2/

It should error out with the wrong PMU rather than using 'cpu_core' and'cpu_atom' instead.

This is still the pattern matching issue. The pattern is "cpu*". Both "cpu_core" and "cpu_atom" can match the pattern "cpu*", so the parser wrongly thinks they are the same PMU type.

Now with this patch,

# ./perf stat -e cpu/cpu-cycles/ true
event syntax error: 'cpu/cpu-cycles/'
                      \___ Cannot find PMU `cpu'. Missing kernel support?
Run 'perf list' for a list of valid events

Thanks
Jin Yao

Quoted 105 lines · Jin Yao

Some different pmu types may have same substring. For example, on Icelake server, we have pmu types "uncore_imc" and "uncore_imc_free_running". Both pmu types have substring "uncore_imc". But the parser would wrongly think they are the same pmu type.

We enable an imc event,
perf stat -e uncore_imc/event=0xe3/ -a -- sleep 1

Perf actually expands the event to:
uncore_imc_0/event=0xe3/
uncore_imc_1/event=0xe3/
uncore_imc_2/event=0xe3/
uncore_imc_3/event=0xe3/
uncore_imc_4/event=0xe3/
uncore_imc_5/event=0xe3/
uncore_imc_6/event=0xe3/
uncore_imc_7/event=0xe3/
uncore_imc_free_running_0/event=0xe3/
uncore_imc_free_running_1/event=0xe3/
uncore_imc_free_running_3/event=0xe3/
uncore_imc_free_running_4/event=0xe3/

That's because the "uncore_imc_free_running" matches the pattern "uncore_imc*".

Now we check that the last characters of pmu name is '_<digit>'.

Fixes b2b9d3a3f021 ("perf pmu: Support wildcards on pmu name in dynamic pmu events")

---
  tools/perf/util/parse-events.y |  2 ++ tools/perf/util/pmu.c          | 25 ++++++++++++++++++++++++- tools/perf/util/pmu.h          |  1 +

  3 files changed, 27 insertions(+), 1 deletion(-)
tools/perf/util/parse-events.y +2 -0
index aba12a4d488e..7a694c7f7f1a 100644
@@ -317,6 +317,8 @@ event_pmu_name opt_pmu_config
                   strncmp($1, "uncore_", 7))
                   name += 7;
               if (!fnmatch(pattern, name, 0)) {
+                if (!perf_pmu__valid_suffix($1, name))
+                    continue;
                   if (parse_events_copy_term_list(orig_terms, &terms))
                       CLEANUP_YYABORT;
                   if (!parse_events_add_pmu(_parse_state, list, pmu->name, terms, true, false))
tools/perf/util/pmu.c +24 -1
index 88c8ecdc60b0..78af01959830 100644
@@ -3,6 +3,7 @@
   #include <linux/compiler.h>
   #include <linux/string.h>
   #include <linux/zalloc.h>
+#include <linux/ctype.h>
   #include <subcmd/pager.h>
   #include <sys/types.h>
   #include <errno.h>
@@ -768,7 +769,7 @@ bool pmu_uncore_alias_match(const char *pmu_name, const char *name)
        */
       for (; tok; name += strlen(tok), tok = strtok_r(NULL, ",", &tmp)) {
           name = strstr(name, tok);
-        if (!name) {
+        if (!name || !perf_pmu__valid_suffix(tok, (char *)name)) {
               res = false;
               goto out;
           }
@@ -1872,3 +1873,25 @@ bool perf_pmu__has_hybrid(void)
       return !list_empty(&perf_pmu__hybrid_pmus);
   }
+
+bool perf_pmu__valid_suffix(char *tok, char *pmu_name)
+{
+    char *p;
+
+    /*
+     * The pmu_name has substring tok. If the format of
+     * pmu_name is <tok> or <tok>_<digit>, return true.
+     */
+    p = pmu_name + strlen(tok);
+    if (*p == 0)
+        return true;
+
+    if (*p != '_')
+        return false;
+
+    ++p;
+    if (*p == 0 || !isdigit(*p))
+        return false;
+
+    return true;
+}
tools/perf/util/pmu.h +1 -0
index a790ef758171..ebfd2b71532b 100644
@@ -133,5 +133,6 @@ void perf_pmu__warn_invalid_config(struct perf_pmu *pmu, __u64 config,
                      char *name);
   bool perf_pmu__has_hybrid(void);
+bool perf_pmu__valid_suffix(char *tok, char *pmu_name);
   #endif /* __PMU_H */
 
#3 lore raw
Quoted 5 lines · Yao

Hi Arnaldo, Jiri,

Any comments for this bug fix patch?

The issue does impact some uncore events and even some metrics.

sry for delay

SNIP

Quoted 36 lines
Quoted 36 lines
Quoted 36 lines

Some different pmu types may have same substring. For example, on Icelake server, we have pmu types "uncore_imc" and "uncore_imc_free_running". Both pmu types have substring "uncore_imc". But the parser would wrongly think they are the same pmu type.

We enable an imc event,
perf stat -e uncore_imc/event=0xe3/ -a -- sleep 1

Perf actually expands the event to:
uncore_imc_0/event=0xe3/
uncore_imc_1/event=0xe3/
uncore_imc_2/event=0xe3/
uncore_imc_3/event=0xe3/
uncore_imc_4/event=0xe3/
uncore_imc_5/event=0xe3/
uncore_imc_6/event=0xe3/
uncore_imc_7/event=0xe3/
uncore_imc_free_running_0/event=0xe3/
uncore_imc_free_running_1/event=0xe3/
uncore_imc_free_running_3/event=0xe3/
uncore_imc_free_running_4/event=0xe3/

That's because the "uncore_imc_free_running" matches the pattern "uncore_imc*".

Now we check that the last characters of pmu name is '_<digit>'.

Fixes b2b9d3a3f021 ("perf pmu: Support wildcards on pmu name in dynamic pmu events")

---
  tools/perf/util/parse-events.y |  2 ++ tools/perf/util/pmu.c          | 25 ++++++++++++++++++++++++- tools/perf/util/pmu.h          |  1 +

  3 files changed, 27 insertions(+), 1 deletion(-)
Quoted 10 lines
Quoted 10 lines

tools/perf/util/parse-events.y event_pmu_name opt_pmu_config · line 321

Quoted 10 lines
tools/perf/util/parse-events.y +2 -0
index aba12a4d488e..7a694c7f7f1a 100644
@@ -317,6 +317,8 @@ event_pmu_name opt_pmu_config
                   strncmp($1, "uncore_", 7))
                   name += 7;
               if (!fnmatch(pattern, name, 0)) {
+                if (!perf_pmu__valid_suffix($1, name))
+                    continue;

could this be part of the fnmatch's pattern?

Quoted 41 lines
Quoted 41 lines
Quoted 41 lines
tools/perf/util/parse-events.y +0 -0
                   if (parse_events_copy_term_list(orig_terms, &terms))
                       CLEANUP_YYABORT;
                   if (!parse_events_add_pmu(_parse_state, list, pmu->name, terms, true, false))
tools/perf/util/pmu.c +16 -1
index 88c8ecdc60b0..78af01959830 100644
@@ -3,6 +3,7 @@
   #include <linux/compiler.h>
   #include <linux/string.h>
   #include <linux/zalloc.h>
+#include <linux/ctype.h>
   #include <subcmd/pager.h>
   #include <sys/types.h>
   #include <errno.h>
@@ -768,7 +769,7 @@ bool pmu_uncore_alias_match(const char *pmu_name, const char *name)
        */
       for (; tok; name += strlen(tok), tok = strtok_r(NULL, ",", &tmp)) {
           name = strstr(name, tok);
-        if (!name) {
+        if (!name || !perf_pmu__valid_suffix(tok, (char *)name)) {
               res = false;
               goto out;
           }
@@ -1872,3 +1873,25 @@ bool perf_pmu__has_hybrid(void)
       return !list_empty(&perf_pmu__hybrid_pmus);
   }
+
+bool perf_pmu__valid_suffix(char *tok, char *pmu_name)
+{
+    char *p;
+
+    /*
+     * The pmu_name has substring tok. If the format of
+     * pmu_name is <tok> or <tok>_<digit>, return true.
+     */
+    p = pmu_name + strlen(tok);
+    if (*p == 0)
+        return true;
+
+    if (*p != '_')
Quoted 8 lines
Quoted 8 lines

tools/perf/util/pmu.c bool perf_pmu__has_hybrid(void) · line 1897

Quoted 8 lines
tools/perf/util/pmu.c +8 -0
+        return false;
+
+    ++p;
+    if (*p == 0 || !isdigit(*p))
+        return false;
+
+    return true;
+}

hum, so we have pattern serch and then another function checking if that search was ok.. I understand that's convenient, because it's on 2 different places, but could we have some generic solution, line one function/search that returns/search for valid pmu name?

thanks,
jirka

author
#4 lore raw

Hi Jiri,

Quoted 47 lines · Jiri Olsa
Quoted 5 lines · Yao

Hi Arnaldo, Jiri,

Any comments for this bug fix patch?

The issue does impact some uncore events and even some metrics.

sry for delay

SNIP

Quoted 36 lines
Quoted 36 lines
Quoted 36 lines

Some different pmu types may have same substring. For example, on Icelake server, we have pmu types "uncore_imc" and "uncore_imc_free_running". Both pmu types have substring "uncore_imc". But the parser would wrongly think they are the same pmu type.

We enable an imc event,
perf stat -e uncore_imc/event=0xe3/ -a -- sleep 1

Perf actually expands the event to:
uncore_imc_0/event=0xe3/
uncore_imc_1/event=0xe3/
uncore_imc_2/event=0xe3/
uncore_imc_3/event=0xe3/
uncore_imc_4/event=0xe3/
uncore_imc_5/event=0xe3/
uncore_imc_6/event=0xe3/
uncore_imc_7/event=0xe3/
uncore_imc_free_running_0/event=0xe3/
uncore_imc_free_running_1/event=0xe3/
uncore_imc_free_running_3/event=0xe3/
uncore_imc_free_running_4/event=0xe3/

That's because the "uncore_imc_free_running" matches the pattern "uncore_imc*".

Now we check that the last characters of pmu name is '_<digit>'.

Fixes b2b9d3a3f021 ("perf pmu: Support wildcards on pmu name in dynamic pmu events")
---
   tools/perf/util/parse-events.y |  2 ++
   tools/perf/util/pmu.c          | 25 ++++++++++++++++++++++++-
   tools/perf/util/pmu.h          |  1 +
   3 files changed, 27 insertions(+), 1 deletion(-)
Quoted 13 lines
Quoted 10 lines
Quoted 10 lines

tools/perf/util/parse-events.y event_pmu_name opt_pmu_config · line 321

Quoted 10 lines
tools/perf/util/parse-events.y +2 -0
index aba12a4d488e..7a694c7f7f1a 100644
@@ -317,6 +317,8 @@ event_pmu_name opt_pmu_config
                   strncmp($1, "uncore_", 7))
                   name += 7;
               if (!fnmatch(pattern, name, 0)) {
+                if (!perf_pmu__valid_suffix($1, name))
+                    continue;

could this be part of the fnmatch's pattern?

Actually I had used the pattern "uncore_imc_[0-9]" before. But for some units, e.g., CHA, they have more than 10 units. So this simple pattern couldn't satisfy them.

And then I changed the pattern to "uncore_imc_[0-9]+$", which can match the string "uncore_imc_<integer id>". But unfortunately it didn't work for fnmatch.

I used regex, such as:

asprintf(&pattern, "%s_[0-9]+$", tok);
regcomp(&regex, pattern, REG_EXTENDED);
ret = regexec(&regex, name, 0, NULL, 0);

But the regex approach looks not very simple (a bit heavy), so finally I just keep using fnmatch and then just check the last character.

Quoted 44 lines
Quoted 44 lines
Quoted 44 lines
Quoted 44 lines
tools/perf/util/parse-events.y +0 -0
                   if (parse_events_copy_term_list(orig_terms, &terms))
                       CLEANUP_YYABORT;
                   if (!parse_events_add_pmu(_parse_state, list, pmu->name, terms, true, false))
tools/perf/util/pmu.c +19 -1
index 88c8ecdc60b0..78af01959830 100644
@@ -3,6 +3,7 @@
   #include <linux/compiler.h>
   #include <linux/string.h>
   #include <linux/zalloc.h>
+#include <linux/ctype.h>
   #include <subcmd/pager.h>
   #include <sys/types.h>
   #include <errno.h>
@@ -768,7 +769,7 @@ bool pmu_uncore_alias_match(const char *pmu_name, const char *name)
        */
       for (; tok; name += strlen(tok), tok = strtok_r(NULL, ",", &tmp)) {
           name = strstr(name, tok);
-        if (!name) {
+        if (!name || !perf_pmu__valid_suffix(tok, (char *)name)) {
               res = false;
               goto out;
           }
@@ -1872,3 +1873,25 @@ bool perf_pmu__has_hybrid(void)
       return !list_empty(&perf_pmu__hybrid_pmus);
   }
+
+bool perf_pmu__valid_suffix(char *tok, char *pmu_name)
+{
+    char *p;
+
+    /*
+     * The pmu_name has substring tok. If the format of
+     * pmu_name is <tok> or <tok>_<digit>, return true.
+     */
+    p = pmu_name + strlen(tok);
+    if (*p == 0)
+        return true;
+
+    if (*p != '_')
+        return false;
+
+    ++p;
Quoted 8 lines
Quoted 5 lines
Quoted 5 lines

tools/perf/util/pmu.c bool perf_pmu__has_hybrid(void) · line 1897

Quoted 5 lines
tools/perf/util/pmu.c +5 -0
+    if (*p == 0 || !isdigit(*p))
+        return false;
+
+    return true;
+}

hum, so we have pattern serch and then another function checking if that search was ok..

Yes, that's what this patch does.

I understand that's convenient, because

Quoted 1 line

it's on 2 different places

Yes, on pmu_uncore_alias_match() and on parse-events.y.

but could we have some generic solution,

Quoted 2 lines

line one function/search that returns/search for valid pmu name?

Sorry, I don't understand this idea well. Would you like to further explain?

Or can you accept the regex approach?

Quoted 3 lines

thanks,
jirka

#5 lore raw

On Mon, Jun 28, 2021 at 09:52:42AM +0800, Jin, Yao wrote:

SNIP

Quoted 28 lines
Quoted 12 lines
Quoted 12 lines
Quoted 12 lines
Quoted 12 lines

+    /*
+     * The pmu_name has substring tok. If the format of +     * pmu_name is <tok> or <tok>_<digit>, return true. +     */
+    p = pmu_name + strlen(tok);
+    if (*p == 0)
+        return true;

tools/perf/util/pmu.c +5 -0
+
+    if (*p != '_')
+        return false;
+
+    ++p;
Quoted 8 lines
Quoted 5 lines
Quoted 5 lines

tools/perf/util/pmu.c bool perf_pmu__has_hybrid(void) · line 1897

Quoted 5 lines

+    if (*p == 0 || !isdigit(*p))
+        return false;

tools/perf/util/pmu.c +3 -0
+
+    return true;
+}

hum, so we have pattern serch and then another function checking if that search was ok..

Yes, that's what this patch does.

I understand that's convenient, because

Quoted 1 line

it's on 2 different places

Yes, on pmu_uncore_alias_match() and on parse-events.y.

Quoted 7 lines

but could we have some generic solution,

Quoted 2 lines

line one function/search that returns/search for valid pmu name?

Sorry, I don't understand this idea well. Would you like to further explain?

Or can you accept the regex approach?

I don't really have any suggestion, just would be great to have this encapsulated in one function.. but if there's not any good one, we can go with this change

Andi, Kan, are you ok with this change?

thanks,
jirka

#6 lore raw
Quoted 36 lines · Jiri Olsa

On Mon, Jun 28, 2021 at 09:52:42AM +0800, Jin, Yao wrote:

SNIP

Quoted 32 lines
Quoted 12 lines
Quoted 12 lines
Quoted 12 lines
Quoted 12 lines

+    /*
+     * The pmu_name has substring tok. If the format of +     * pmu_name is <tok> or <tok>_<digit>, return true. +     */
+    p = pmu_name + strlen(tok);
+    if (*p == 0)
+        return true;

tools/perf/util/pmu.c +5 -0
+
+    if (*p != '_')
+        return false;
+
+    ++p;
Quoted 8 lines
Quoted 5 lines
Quoted 5 lines

tools/perf/util/pmu.c bool perf_pmu__has_hybrid(void) · line 1897

Quoted 5 lines

+    if (*p == 0 || !isdigit(*p))
+        return false;

tools/perf/util/pmu.c +3 -0
+
+    return true;
+}

hum, so we have pattern serch and then another function checking if that search was ok..

Yes, that's what this patch does.

I understand that's convenient, because

Quoted 1 line

it's on 2 different places

Yes, on pmu_uncore_alias_match() and on parse-events.y.

but could we have some generic solution,

Quoted 2 lines

line one function/search that returns/search for valid pmu name?

Quoted 6 lines
Quoted 3 lines

Sorry, I don't understand this idea well. Would you like to further explain?

Or can you accept the regex approach?

I don't really have any suggestion, just would be great to have this encapsulated in one function..

Yes, I agree. One function is better.

We just changed the design for the uncore PMU on SPR. There will be two PMU names for each uncore unit, a real name and an alias. The perf tool should handle both names. So we have to compare both names here. I think one generic function can facilitate the code rebase.

https://lore.kernel.org/lkml/[email protected]/

Thanks,
Kan

author
#7 lore raw

Hi Jiri, Hi Kan,

Quoted 48 lines · Kan
Quoted 36 lines · Jiri Olsa

On Mon, Jun 28, 2021 at 09:52:42AM +0800, Jin, Yao wrote:

SNIP

Quoted 32 lines
Quoted 12 lines
Quoted 12 lines
Quoted 12 lines
Quoted 12 lines

+    /*
+     * The pmu_name has substring tok. If the format of +     * pmu_name is <tok> or <tok>_<digit>, return true. +     */
+    p = pmu_name + strlen(tok);
+    if (*p == 0)
+        return true;

(quoted diff) +5 -0
+
+    if (*p != '_')
+        return false;
+
+    ++p;
Quoted 8 lines
Quoted 5 lines
Quoted 5 lines

Review

Quoted 5 lines

+    if (*p == 0 || !isdigit(*p))
+        return false;

(quoted diff) +3 -0
+
+    return true;
+}

hum, so we have pattern serch and then another function checking if that search was ok..

Yes, that's what this patch does.

I understand that's convenient, because

Quoted 1 line

it's on 2 different places

Yes, on pmu_uncore_alias_match() and on parse-events.y.

but could we have some generic solution,

Quoted 2 lines

line one function/search that returns/search for valid pmu name?

Quoted 6 lines
Quoted 3 lines

Sorry, I don't understand this idea well. Would you like to further explain?

Or can you accept the regex approach?

I don't really have any suggestion, just would be great to have this encapsulated in one function..

Yes, I agree. One function is better.

Quoted 8 lines

We just changed the design for the uncore PMU on SPR. There will be two PMU names for each uncore unit, a real name and an alias. The perf tool should handle both names. So we have to compare both names here. I think one generic function can facilitate the code rebase.

https://lore.kernel.org/lkml/[email protected]/

Thanks,
Kan

Thanks for the comments!

I'm now preparing v2. In v2 version, I will provide a new function 'perf_pmu__pattern_match which wraps the matching and checking. Something like:

while ((pmu = perf_pmu__scan(pmu)) != NULL) {
	if (!perf_pmu__pattern_match(pmu, pattern, $1)) {
		...
	}
}

I will post v2 soon.

Thanks
Jin Yao