[PATCH v3] perf/profile: Report failure to create /proc/profile

[Yongxin Yang]

create_proc_profile() discards the result of proc_create(): when the
entry cannot be created, the function still returns 0 and profiling
setup silently pretends to have succeeded.  The local variable err is
never assigned anything but 0 and only hides that.

Report the failure and return -ENOMEM instead.

create_proc_profile() is a subsys_initcall and do_initcall_level()
discards the initcall return value, so the error message is what the
user will actually see.

Signed-off-by: Yongxin Yang <yangyongxin@xxxxxxxxxx>
---
Link: https://lore.kernel.org/all/20260824024251.279751-1-yang_yongxin123@xxxxxxx/

Changes in v3:
- Do not silently ignore a proc_create() failure: report it with pr_err()
  and return -ENOMEM, as suggested by Andrew Morton.  This is no longer a
  pure cleanup, so 'No functional change.' is dropped.
- Drop the Reviewed-by from Bradley Morgan: v3 changes the behaviour, so
  the tag no longer covers the code he reviewed on v1.  Bradley is Cc'd
  again for a fresh review.

 kernel/profile.c | 10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)

diff --git a/kernel/profile.c b/kernel/profile.c
index 13be6fc9fd00..818098cc094a 100644
--- a/kernel/profile.c
+++ b/kernel/profile.c
@@ -240,15 +240,17 @@ static const struct proc_ops profile_proc_ops = {
 int __ref create_proc_profile(void)
 {
 	struct proc_dir_entry *entry;
-	int err = 0;
 
 	if (!prof_on)
 		return 0;
 	entry = proc_create("profile", S_IWUSR | S_IRUGO,
 			    NULL, &profile_proc_ops);
-	if (entry)
-		proc_set_size(entry, (1 + prof_len) * sizeof(atomic_t));
-	return err;
+	if (!entry) {
+		pr_err("profile: failed to create /proc/profile\n");
+		return -ENOMEM;
+	}
+	proc_set_size(entry, (1 + prof_len) * sizeof(atomic_t));
+	return 0;
 }
 subsys_initcall(create_proc_profile);
 #endif /* CONFIG_PROC_FS */
-- 
2.25.1