[PATCH next v3 0/2] fix potential memleak in perf events parser

[PATCH next v3 0/2] fix potential memleak in perf events parser

[Chen Wandun]

fix some memleaks in parse_events_term__sym_hw and parse_events_term__clone.
ls/perf

v1 ==> v2
1. split into two patches
2. add jump targets common exception handling
3. add Fixes tag

v2 ==> v3
add more commit log.

Chen Wandun (1):
  perf tools: fix potential memleak in perf events parser

Cheng Jian (1):
  perf tools: fix potential memleak in perf events parser

 tools/perf/util/parse-events.c | 51 ++++++++++++++++++++++++++++------
 1 file changed, 43 insertions(+), 8 deletions(-)

-- 
2.17.1

[PATCH next v3 1/2] perf tools: fix potential memleak in perf events parser

[Chen Wandun]

From: Cheng Jian <cj.chengjian@huawei.com>

Fix memory leak of in function parse_events_term__sym_hw()
and parse_events_term__clone() when string duplication failed.

Fixes: b6645a723595 ("perf parse: Ensure config and str in terms are unique")
Signed-off-by: Cheng Jian <cj.chengjian@huawei.com>
Signed-off-by: Chen Wandun <chenwandun@huawei.com>
---
 tools/perf/util/parse-events.c | 12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)

diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c
index 3decbb203846..6f4dc8a92817 100644
--- a/tools/perf/util/parse-events.c
+++ b/tools/perf/util/parse-events.c
@@ -2957,8 +2957,12 @@ int parse_events_term__sym_hw(struct parse_events_term **term,
 	sym = &event_symbols_hw[idx];
 
 	str = strdup(sym->symbol);
-	if (!str)
+	if (!str) {
+		if (!config)
+			free(temp.config);
 		return -ENOMEM;
+	}
+
 	return new_term(term, &temp, str, 0);
 }
 
@@ -2983,8 +2987,12 @@ int parse_events_term__clone(struct parse_events_term **new,
 		return new_term(new, &temp, NULL, term->val.num);
 
 	str = strdup(term->val.str);
-	if (!str)
+	if (!str) {
+		if (term->config)
+			free(temp.config);
 		return -ENOMEM;
+	}
+
 	return new_term(new, &temp, str, 0);
 }
 
-- 
2.17.1

Re: [PATCH v3 1/2] perf tools: Fix potential memory leaks in perf events parser

[Markus Elfring]

> Fix memory leak of in function parse_events_term__sym_hw()
> and parse_events_term__clone() when string duplication failed.

Can a wording like “Fix memory leaks in …” be more appropriate for
the final commit message?
Would you find any other description variant more pleasing?

Regards,
Markus

[PATCH next v3 2/2] perf tools: fix potential memleak in perf events parser

[Chen Wandun]

Fix potential memory leak. Function new_term may return error, so
it is need to free memory when the return value is negative.
What's more, add jump targets so that a configuration object and
a duplicated string are released after a call of the function
"strdup" or "new_term" failed.

Fixes: b6645a723595 ("perf parse: Ensure config and str in terms are unique")

Signed-off-by: Chen Wandun <chenwandun@huawei.com>
---
 tools/perf/util/parse-events.c | 47 ++++++++++++++++++++++++++--------
 1 file changed, 37 insertions(+), 10 deletions(-)

diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c
index 6f4dc8a92817..a9f32032af08 100644
--- a/tools/perf/util/parse-events.c
+++ b/tools/perf/util/parse-events.c
@@ -2947,6 +2947,7 @@ int parse_events_term__sym_hw(struct parse_events_term **term,
 		.type_term = PARSE_EVENTS__TERM_TYPE_USER,
 		.config    = config,
 	};
+	int ret;
 
 	if (!temp.config) {
 		temp.config = strdup("event");
@@ -2958,12 +2959,23 @@ int parse_events_term__sym_hw(struct parse_events_term **term,
 
 	str = strdup(sym->symbol);
 	if (!str) {
-		if (!config)
-			free(temp.config);
-		return -ENOMEM;
+		ret = -ENOMEM;
+		goto free_config;
 	}
 
-	return new_term(term, &temp, str, 0);
+	ret = new_term(term, &temp, str, 0);
+	if (ret)
+		goto free_str;
+
+	return 0;
+
+free_str:
+	free(str);
+free_config:
+	if (!config)
+		free(temp.config);
+
+	return ret;
 }
 
 int parse_events_term__clone(struct parse_events_term **new,
@@ -2977,23 +2989,38 @@ int parse_events_term__clone(struct parse_events_term **new,
 		.err_term  = term->err_term,
 		.err_val   = term->err_val,
 	};
+	int ret;
 
 	if (term->config) {
 		temp.config = strdup(term->config);
 		if (!temp.config)
 			return -ENOMEM;
 	}
-	if (term->type_val == PARSE_EVENTS__TERM_TYPE_NUM)
-		return new_term(new, &temp, NULL, term->val.num);
+	if (term->type_val == PARSE_EVENTS__TERM_TYPE_NUM) {
+		ret = new_term(new, &temp, NULL, term->val.num);
+		if (ret)
+			goto free_config;
+	}
 
 	str = strdup(term->val.str);
 	if (!str) {
-		if (term->config)
-			free(temp.config);
-		return -ENOMEM;
+		ret = -ENOMEM;
+		goto free_config;
 	}
 
-	return new_term(new, &temp, str, 0);
+	ret = new_term(new, &temp, str, 0);
+	if (ret)
+		goto free_str;
+
+	return 0;
+
+free_str:
+	free(str);
+free_config:
+	if (term->config)
+		free(temp.config);
+
+	return ret;
 }
 
 void parse_events_term__delete(struct parse_events_term *term)
-- 
2.17.1

Re: [PATCH v3 2/2] perf tools: Fix potential memory leaks in perf events parser

[Markus Elfring]

> Fix potential memory leak. Function new_term may return error, so
> it is need to free memory when the return value is negative.

I hope that a typo will be avoided for the final commit message.
Would you find any other description variant more pleasing?

Regards,
Markus

Re: [PATCH v3 0/2] Fixing memory leaks in perf events parser

[Markus Elfring]

> fix some memleaks in parse_events_term__sym_hw and parse_events_term__clone.

Can it be more appropriate to refer to the term “memory leak” in consistent ways?


> v1 ==> v2
> 1. split into two patches

Corresponding development consequences can become more interesting.


> v2 ==> v3
> add more commit log.
>
> Chen Wandun (1):
>   perf tools: fix potential memleak in perf events parser
>
> Cheng Jian (1):
>   perf tools: fix potential memleak in perf events parser
>
>  tools/perf/util/parse-events.c | 51 ++++++++++++++++++++++++++++------

Are there any chances to make the change distinction a bit easier
by adjusting such commit subjects?

Regards,
Markus