We keep incrementing "spec" as we parse the args so we end up calling
kfree() on a modified of spec. It probably works or this would have
been caught in testing, but it looks weird.
Signed-off-by: Dan Carpenter <dan.carpenter@xxxxxxxxxx>
diff --git a/mm/vmpressure.c b/mm/vmpressure.c
index 0781b1363e0a..1225ec5d9596 100644
--- a/mm/vmpressure.c
+++ b/mm/vmpressure.c
@@ -386,11 +386,11 @@ int vmpressure_register_event(struct mem_cgroup *memcg,
struct vmpressure_event *ev;
enum vmpressure_modes mode = VMPRESSURE_NO_PASSTHROUGH;
enum vmpressure_levels level = -1;
- char *spec = NULL;
+ char *spec, *spec_orig;
char *token;
int ret = 0;
- spec = kzalloc(MAX_VMPRESSURE_ARGS_LEN + 1, GFP_KERNEL);
+ spec_orig = spec = kzalloc(MAX_VMPRESSURE_ARGS_LEN + 1, GFP_KERNEL);
if (!spec) {
ret = -ENOMEM;
goto out;
@@ -429,7 +429,7 @@ int vmpressure_register_event(struct mem_cgroup *memcg,
list_add(&ev->node, &vmpr->events);
mutex_unlock(&vmpr->events_lock);
out:
- kfree(spec);
+ kfree(spec_orig);
return ret;
}
--
To unsubscribe from this list: send the line "unsubscribe kernel-janitors" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
