fsnotify: allow adding an inode mark without pinning inode

fsnotify_add_mark() and variants implicitly take a reference on inode
when attaching a mark to an inode.

Make that behavior opt-out with the mark flag FSNOTIFY_MARK_FLAG_NO_IREF.

Instead of taking the inode reference when attaching connector to inode
and dropping the inode reference when detaching connector from inode,
take the inode reference on attach of the first mark that wants to hold
an inode reference and drop the inode reference on detach of the last
mark that wants to hold an inode reference.

Backends can "upgrade" an existing mark to take an inode reference, but
cannot "downgrade" a mark with inode reference to release the refernce.

This leaves the choice to the backend whether or not to pin the inode
when adding an inode mark.

This is intended to be used when adding a mark with ignored mask that is
used for optimization in cases where group can afford getting unneeded
events and reinstate the mark with ignored mask when inode is accessed
again after being evicted.

Link: https://lore.kernel.org/r/[e-mail]
Signed-off-by: Amir Goldstein <[e-mail]>
Signed-off-by: Jan Kara <[e-mail]>

fs/notify/mark.c

···
 	return *fsnotify_conn_mask_p(conn);
 }
 
-static void __fsnotify_recalc_mask(struct fsnotify_mark_connector *conn)
+static void fsnotify_get_inode_ref(struct inode *inode)
+{
+	ihold(inode);
+	atomic_long_inc(&inode->i_sb->s_fsnotify_connectors);
+}
+
+/*
+ * Grab or drop inode reference for the connector if needed.
+ *
+ * When it's time to drop the reference, we only clear the HAS_IREF flag and
+ * return the inode object. fsnotify_drop_object() will be resonsible for doing
+ * iput() outside of spinlocks. This happens when last mark that wanted iref is
+ * detached.
+ */
+static struct inode *fsnotify_update_iref(struct fsnotify_mark_connector *conn,
+					  bool want_iref)
+{
+	bool has_iref = conn->flags & FSNOTIFY_CONN_FLAG_HAS_IREF;
+	struct inode *inode = NULL;
+
+	if (conn->type != FSNOTIFY_OBJ_TYPE_INODE ||
+	    want_iref == has_iref)
+		return NULL;
+
+	if (want_iref) {
+		/* Pin inode if any mark wants inode refcount held */
+		fsnotify_get_inode_ref(fsnotify_conn_inode(conn));
+		conn->flags |= FSNOTIFY_CONN_FLAG_HAS_IREF;
+	} else {
+		/* Unpin inode after detach of last mark that wanted iref */
+		inode = fsnotify_conn_inode(conn);
+		conn->flags &= ~FSNOTIFY_CONN_FLAG_HAS_IREF;
+	}
+
+	return inode;
+}
+
+static void *__fsnotify_recalc_mask(struct fsnotify_mark_connector *conn)
 {
 	u32 new_mask = 0;
+	bool want_iref = false;
 	struct fsnotify_mark *mark;
 
 	assert_spin_locked(&conn->lock);
 	/* We can get detached connector here when inode is getting unlinked. */
 	if (!fsnotify_valid_obj_type(conn->type))
-		return;
+		return NULL;
 	hlist_for_each_entry(mark, &conn->list, obj_list) {
-		if (mark->flags & FSNOTIFY_MARK_FLAG_ATTACHED)
-			new_mask |= fsnotify_calc_mask(mark);
+		if (!(mark->flags & FSNOTIFY_MARK_FLAG_ATTACHED))
+			continue;
+		new_mask |= fsnotify_calc_mask(mark);
+		if (conn->type == FSNOTIFY_OBJ_TYPE_INODE &&
+		    !(mark->flags & FSNOTIFY_MARK_FLAG_NO_IREF))
+			want_iref = true;
 	}
 	*fsnotify_conn_mask_p(conn) = new_mask;
+
+	return fsnotify_update_iref(conn, want_iref);
 }
 
 /*
···
 		conn = conn->destroy_next;
 		kmem_cache_free(fsnotify_mark_connector_cachep, free);
 	}
-}
-
-static void fsnotify_get_inode_ref(struct inode *inode)
-{
-	ihold(inode);
-	atomic_long_inc(&inode->i_sb->s_fsnotify_connectors);
 }
 
 static void fsnotify_put_inode_ref(struct inode *inode)
···
 	if (conn->type == FSNOTIFY_OBJ_TYPE_INODE) {
 		inode = fsnotify_conn_inode(conn);
 		inode->i_fsnotify_mask = 0;
+
+		/* Unpin inode when detaching from connector */
+		if (!(conn->flags & FSNOTIFY_CONN_FLAG_HAS_IREF))
+			inode = NULL;
 	} else if (conn->type == FSNOTIFY_OBJ_TYPE_VFSMOUNT) {
 		fsnotify_conn_mount(conn)->mnt_fsnotify_mask = 0;
 	} else if (conn->type == FSNOTIFY_OBJ_TYPE_SB) {
···
 		objp = fsnotify_detach_connector_from_object(conn, &type);
 		free_conn = true;
 	} else {
-		__fsnotify_recalc_mask(conn);
+		objp = __fsnotify_recalc_mask(conn);
+		type = conn->type;
 	}
 	WRITE_ONCE(mark->connector, NULL);
 	spin_unlock(&conn->lock);
···
 					       unsigned int obj_type,
 					       __kernel_fsid_t *fsid)
 {
-	struct inode *inode = NULL;
 	struct fsnotify_mark_connector *conn;
 
 	conn = kmem_cache_alloc(fsnotify_mark_connector_cachep, GFP_KERNEL);
···
 		return -ENOMEM;
 	spin_lock_init(&conn->lock);
 	INIT_HLIST_HEAD(&conn->list);
+	conn->flags = 0;
 	conn->type = obj_type;
 	conn->obj = connp;
 	/* Cache fsid of filesystem containing the object */
···
 		conn->fsid.val[0] = conn->fsid.val[1] = 0;
 		conn->flags = 0;
 	}
-	if (conn->type == FSNOTIFY_OBJ_TYPE_INODE) {
-		inode = fsnotify_conn_inode(conn);
-		fsnotify_get_inode_ref(inode);
-	}
 	fsnotify_get_sb_connectors(conn);
 
 	/*
···
 	 */
 	if (cmpxchg(connp, NULL, conn)) {
 		/* Someone else created list structure for us */
-		if (inode)
-			fsnotify_put_inode_ref(inode);
 		fsnotify_put_sb_connectors(conn);
 		kmem_cache_free(fsnotify_mark_connector_cachep, conn);
 	}
···
 	if (ret)
 		goto err;
 
-	if (mark->mask || mark->ignored_mask)
-		fsnotify_recalc_mask(mark->connector);
+	fsnotify_recalc_mask(mark->connector);
 
 	return ret;
 err:

include/linux/fsnotify_backend.h

···
 	spinlock_t lock;
 	unsigned short type;	/* Type of object [lock] */
 #define FSNOTIFY_CONN_FLAG_HAS_FSID	0x01
+#define FSNOTIFY_CONN_FLAG_HAS_IREF	0x02
 	unsigned short flags;	/* flags [lock] */
 	__kernel_fsid_t fsid;	/* fsid of filesystem containing object */
 	union {
···
 #define FSNOTIFY_MARK_FLAG_IN_ONESHOT		0x0020
 	/* fanotify mark flags */
 #define FSNOTIFY_MARK_FLAG_IGNORED_SURV_MODIFY	0x0100
+#define FSNOTIFY_MARK_FLAG_NO_IREF		0x0200
 	unsigned int flags;		/* flags [mark->lock] */
 };