Skip to contentSkip to conversation

ethtool: fix potential NULL pointer dereference in find_option

PATCH In review Anton Moryakov 2 participants 2 messages Jul 9, 2025 – Jul 21, 2025

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 ethtool: fix potential NULL pointer dereference in find_option

Static analyzer reported a possible NULL pointer dereference:

- In main(), 'argp' is dereferenced and passed to find_option() without checking if *argp is NULL.

- The function find_option() does not validate its input argument, which may lead to undefined behavior when using strncmp() or strcspn() with a NULL pointer.

This patch adds proper NULL check in find_option() to prevent invalid memory access. Additionally, it improves robustness by making sure that the input argument is valid before passing it to find_option().

Found by Svace static analysis tool.

---
 ethtool.c | 3 +++
 1 file changed, 3 insertions(+)
ethtool.c +3 -0
index 0513a1a..4250add 100644
@@ -6395,6 +6395,9 @@ static int show_usage(struct cmd_context *ctx __maybe_unused)
 
 static int find_option(char *arg)
 {
+	if(!arg)
+		return -1;
+		
 	const char *opt;
 	size_t len;
 	int k;
Signature · 1 line
2.39.2
#1 lore raw
Quoted 4 lines · Anton Moryakov

Static analyzer reported a possible NULL pointer dereference:

- In main(), 'argp' is dereferenced and passed to find_option() without checking if *argp is NULL.

Not true. This is what the code looks like:

	if (!*argp)
		exit_bad_args();

	k = find_option(*argp); 

On the other hand, there is no such check when find_option() is called from do_perqueue(). But if we really want to address this, I would rather do it in a consistent way, i.e. the same way in both places.

However, I'm still not convinced the inconsistency between argc and argp[] you are trying to prevent can actually happen on Linux. AFAIK all exec* functions are only library wrappers for execve() which is passed only the argv[] array and the argc count is determined from it by finding the first null pointer. Unless I'm missing something important, this means that argp[i] can never be null for i < argc.

Do you have a scenario (not assuming a serious kernel bug) that would result in null pointer in argp[] before the argc index?

Quoted 2 lines

Additionally, it improves robustness by making sure that the input argument is valid before passing it to find_option().

The patch does not actually do this, it only changes the code inside find_option().

Quoted 13 lines

Found by Svace static analysis tool.

---
 ethtool.c | 3 +++
 1 file changed, 3 insertions(+)
ethtool.c +0 -0
index 0513a1a..4250add 100644

ethtool.c static int show_usage(struct cmd_context *ctx __maybe_unused) · line 6398

Quoted 4 lines
ethtool.c +1 -0
@@ -6395,6 +6395,9 @@ static int show_usage(struct cmd_context *ctx __maybe_unused)
 
 static int find_option(char *arg)
 {
+	if(!arg)

Missing space between "if" and "(".

ethtool.c static int show_usage(struct cmd_context *ctx __maybe_unused) · line 6400

Quoted 2 lines
ethtool.c +2 -0
+		return -1;
+		

You add whitespace on this empty line.

Michal

Quoted 6 lines
ethtool.c +0 -0
 	const char *opt;
 	size_t len;
 	int k;
Signature · 1 line
2.39.2