vfs: use fast_list for superblock's inode listfast_list

Use the new fast_list for super_block.s_inodes.

This gives similar performance to Dave's dlock list approach [1]; lock
contention is now moved to the lru_list locks.

Iteration is now fully lockless - instead we iterate using
rcu_read_lock(), which means we must take care for racing with removal.

Generally this is already handled - code that iterates over s_inodes
takes i_lock and checks i_state, skipping inodes that are
I_WILL_FREE|I_FREEING. However, code may also check for nonzero
i_sb_list_idx if it wishes to iterate over precisely the inodes that are
on the s_inodes list.

[1]: https://lore.kernel.org/linux-fsdevel/20231206060629.2827226-4-david@fromorbit.com/

Cc: Christian Brauner <brauner@kernel.org>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Waiman Long <longman@redhat.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

1 files changed, 13 insertions, 12 deletions

diff --git a/fs/notify/fsnotify.c b/fs/notify/fsnotify.c
index 272c8a1dab3c..9a8da7a02a6f 100644
--- a/fs/notify/fsnotify.c
+++ b/fs/notify/fsnotify.c
@@ -33,14 +33,19 @@ void __fsnotify_vfsmount_delete(struct vfsmount *mnt)
  * @sb: superblock being unmounted.
  *
  * Called during unmount with no locks held, so needs to be safe against
- * concurrent modifiers. We temporarily drop sb->s_inode_list_lock and CAN block.
+ * concurrent modifiers. Can block.
  */
 static void fsnotify_unmount_inodes(struct super_block *sb)
 {
-	struct inode *inode, *iput_inode = NULL;
+	struct genradix_iter iter;
+	void **i;
+
+	rcu_read_lock();
+	genradix_for_each(&sb->s_inodes.items, iter, i) {
+		struct inode *inode = *((struct inode **) i);
+		if (!inode)
+			continue;
 
-	spin_lock(&sb->s_inode_list_lock);
-	list_for_each_entry(inode, &sb->s_inodes, i_sb_list) {
 		/*
 		 * We cannot __iget() an inode in state I_FREEING,
 		 * I_WILL_FREE, or I_NEW which is fine because by that point
@@ -68,23 +73,19 @@ static void fsnotify_unmount_inodes(struct super_block *sb)
 
 		__iget(inode);
 		spin_unlock(&inode->i_lock);
-		spin_unlock(&sb->s_inode_list_lock);
-
-		iput(iput_inode);
+		rcu_read_unlock();
 
 		/* for each watch, send FS_UNMOUNT and then remove it */
 		fsnotify_inode(inode, FS_UNMOUNT);
 
 		fsnotify_inode_delete(inode);
 
-		iput_inode = inode;
+		iput(inode);
 
 		cond_resched();
-		spin_lock(&sb->s_inode_list_lock);
+		rcu_read_lock();
 	}
-	spin_unlock(&sb->s_inode_list_lock);
-
-	iput(iput_inode);
+	rcu_read_unlock();
 }
 
 void fsnotify_sb_delete(struct super_block *sb)