kernfs: Improve kernfs_drain() and always call on removal

__kernfs_remove() was skipping draining based on KERNFS_ACTIVATED - whether
the node has ever been activated since creation. Instead, update it to
always call kernfs_drain() which now drains or skips based on the precise
drain conditions. This ensures that the nodes will be deactivated and
drained regardless of their states.

This doesn't make meaningful difference now but will enable deactivating and
draining nodes dynamically by making removals safe when racing those
operations.

While at it, drop / update comments.

v2: Fix the inverted test on kernfs_should_drain_open_files() noted by
    Chengming. This was fixed by the next unrelated patch in the previous
    posting.

Cc: Chengming Zhou <zhouchengming@bytedance.com>
Tested-by: Chengming Zhou <zhouchengming@bytedance.com>
Reviewed-by: Chengming Zhou <zhouchengming@bytedance.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Link: https://lore.kernel.org/r/20220828050440.734579-6-tj@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

diff --git a/fs/kernfs/dir.c b/fs/kernfs/dir.c
index 8ae44db920d46149512e0d287601c60cc6d783a0..b3d2018a334da2fb9c1887477a6cfc780258b5de 100644 (file)
--- a/fs/kernfs/dir.c
+++ b/fs/kernfs/dir.c
@@ -472,6 +472,16 @@ static void kernfs_drain(struct kernfs_node *kn)
        lockdep_assert_held_write(&root->kernfs_rwsem);
        WARN_ON_ONCE(kernfs_active(kn));
 
+       /*
+        * Skip draining if already fully drained. This avoids draining and its
+        * lockdep annotations for nodes which have never been activated
+        * allowing embedding kernfs_remove() in create error paths without
+        * worrying about draining.
+        */
+       if (atomic_read(&kn->active) == KN_DEACTIVATED_BIAS &&
+           !kernfs_should_drain_open_files(kn))
+               return;
+
        up_write(&root->kernfs_rwsem);
 
        if (kernfs_lockdep(kn)) {
@@ -480,7 +490,6 @@ static void kernfs_drain(struct kernfs_node *kn)
                        lock_contended(&kn->dep_map, _RET_IP_);
        }
 
-       /* but everyone should wait for draining */
        wait_event(root->deactivate_waitq,
                   atomic_read(&kn->active) == KN_DEACTIVATED_BIAS);
 
@@ -1370,23 +1379,14 @@ static void __kernfs_remove(struct kernfs_node *kn)
                pos = kernfs_leftmost_descendant(kn);
 
                /*
-                * kernfs_drain() drops kernfs_rwsem temporarily and @pos's
+                * kernfs_drain() may drop kernfs_rwsem temporarily and @pos's
                 * base ref could have been put by someone else by the time
                 * the function returns.  Make sure it doesn't go away
                 * underneath us.
                 */
                kernfs_get(pos);
 
-               /*
-                * Drain iff @kn was activated.  This avoids draining and
-                * its lockdep annotations for nodes which have never been
-                * activated and allows embedding kernfs_remove() in create
-                * error paths without worrying about draining.
-                */
-               if (kn->flags & KERNFS_ACTIVATED)
-                       kernfs_drain(pos);
-               else
-                       WARN_ON_ONCE(atomic_read(&kn->active) != KN_DEACTIVATED_BIAS);
+               kernfs_drain(pos);
 
                /*
                 * kernfs_unlink_sibling() succeeds once per node.  Use it