libbpf: Report error when a negative kprobe offset is specified

In attach_kprobe(), the parsing logic uses sscanf() to extract the
target function name and offset from the section definition. Currently,
if a user specifies a negative offset (e.g., SEC("kprobe/func+-100")),
the input is not explicitly caught and reported as an error.

This commit updates the logic to explicitly notify the user when a
negative integer is provided. To facilitate this check, the offset
variable is changed from unsigned long to long so that sscanf()
can accurately capture a negative input for evaluation.

If a negative offset is detected, the loader will now print an
informative warning stating that the offset must be non-negative,
and return -EINVAL.

Additionally, free(func) is called in this new error path to prevent
a memory leak, as the function name string is dynamically allocated
by sscanf().

Fixes: e3f9bc35ea7e9 ("libbpf: Allow decimal offset for kprobes")
Signed-off-by: Aaron Tomlin <atomlin@atomlin.com>
Acked-by: Mykyta Yatsenko <yatsenko@meta.com>
Link: https://lore.kernel.org/bpf/20260419030944.1423642-1-atomlin@atomlin.com
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>

src/libbpf.c

@@ -12280,7 +12280,7 @@ error:
 static int attach_kprobe(const struct bpf_program *prog, long cookie, struct bpf_link **link)
 {
 	DECLARE_LIBBPF_OPTS(bpf_kprobe_opts, opts);
-	unsigned long offset = 0;
+	long offset = 0;
 	const char *func_name;
 	char *func;
 	int n;
@@ -12302,6 +12302,13 @@ static int attach_kprobe(const struct bpf_program *prog, long cookie, struct bpf
 		pr_warn("kprobe name is invalid: %s\n", func_name);
 		return -EINVAL;
 	}
+
+	if (offset < 0) {
+		free(func);
+		pr_warn("kprobe offset must be a non-negative integer: %li\n", offset);
+		return -EINVAL;
+	}
+
 	if (opts.retprobe && offset != 0) {
 		free(func);
 		pr_warn("kretprobes do not support offset specification\n");