inotify: do not leak inode marks in inotify_add_watch

inotify_add_watch had a couple of problems.  The biggest being that if
inotify_add_watch was called on the same inode twice (to update or change the
event mask) a refence was taken on the original inode mark by
fsnotify_find_mark_entry but was not being dropped at the end of the
inotify_add_watch call.  Thus if inotify_rm_watch was called although the mark
was removed from the inode, the refcnt wouldn't hit zero and we would leak
memory.

Reported-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Eric Paris <eparis@redhat.com>

1 files changed, 10 insertions, 4 deletions

diff --git a/fs/notify/inotify/inotify_user.c b/fs/notify/inotify/inotify_user.c
index 1a870f9157b3..aff4214f16c3 100644
--- a/fs/notify/inotify/inotify_user.c
+++ b/fs/notify/inotify/inotify_user.c
@@ -463,9 +463,6 @@ retry:
                         goto out_err;
 
                 spin_lock(&group->inotify_data.idr_lock);
-                /* if entry is added to the idr we keep the reference obtained
-                 * through fsnotify_mark_add.  remember to drop this reference
-                 * when entry is removed from idr */
                 ret = idr_get_new_above(&group->inotify_data.idr, entry,
                                         ++group->inotify_data.last_wd,
                                         &ientry->wd);
@@ -476,8 +473,13 @@ retry:
                         goto out_err;
                 }
                 atomic_inc(&group->inotify_data.user->inotify_watches);
+
+                /* we put the mark on the idr, take a reference */
+                fsnotify_get_mark(entry);
         }
 
+        ret = ientry->wd;
+
         spin_lock(&entry->lock);
 
         old_mask = entry->mask;
@@ -508,7 +510,11 @@ retry:
                         fsnotify_recalc_group_mask(group);
         }
 
-        return ientry->wd;
+        /* this either matches fsnotify_find_mark_entry, or init_mark_entry
+         * depending on which path we took... */
+        fsnotify_put_mark(entry);
+
+        return ret;
 
 out_err:
         /* see this isn't supposed to happen, just kill the watch */