new inode method: ->free_inode()

A lot of ->destroy_inode() instances end with call_rcu() of a callback
that does RCU-delayed part of freeing.  Introduce a new method for
doing just that, with saner signature.

Rules:
->destroy_inode		->free_inode
	f			g		immediate call of f(),
						RCU-delayed call of g()
	f			NULL		immediate call of f(),
						no RCU-delayed calls
	NULL			g		RCU-delayed call of g()
	NULL			NULL		RCU-delayed default freeing

IOW, NULL ->free_inode gives the same behaviour as now.

Note that NULL, NULL is equivalent to NULL, free_inode_nonrcu; we could
mandate the latter form, but that would have very little benefit beyond
making rules a bit more symmetric.  It would break backwards compatibility,
require extra boilerplate and expected semantics for (NULL, NULL) pair
would have no use whatsoever...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

2 files changed, 27 insertions, 0 deletions

diff --git a/Documentation/filesystems/Locking b/Documentation/filesystems/Locking
index efea228ccd8a..7b20c385cc02 100644
--- a/Documentation/filesystems/Locking
+++ b/Documentation/filesystems/Locking
@@ -118,6 +118,7 @@ set:		exclusive
 --------------------------- super_operations ---------------------------
 prototypes:
 	struct inode *(*alloc_inode)(struct super_block *sb);
+	void (*free_inode)(struct inode *);
 	void (*destroy_inode)(struct inode *);
 	void (*dirty_inode) (struct inode *, int flags);
 	int (*write_inode) (struct inode *, struct writeback_control *wbc);
@@ -139,6 +140,7 @@ locking rules:
 	All may block [not true, see below]
 			s_umount
 alloc_inode:
+free_inode:				called from RCU callback
 destroy_inode:
 dirty_inode:
 write_inode:
diff --git a/Documentation/filesystems/porting b/Documentation/filesystems/porting
index cf43bc4dbf31..b8d3ddd8b8db 100644
--- a/Documentation/filesystems/porting
+++ b/Documentation/filesystems/porting
@@ -638,3 +638,28 @@ in your dentry operations instead.
 	inode to d_splice_alias() will also do the right thing (equivalent of
 	d_add(dentry, NULL); return NULL;), so that kind of special cases
 	also doesn't need a separate treatment.
+--
+[strongly recommended]
+	take the RCU-delayed parts of ->destroy_inode() into a new method -
+	->free_inode().  If ->destroy_inode() becomes empty - all the better,
+	just get rid of it.  Synchronous work (e.g. the stuff that can't
+	be done from an RCU callback, or any WARN_ON() where we want the
+	stack trace) *might* be movable to ->evict_inode(); however,
+	that goes only for the things that are not needed to balance something
+	done by ->alloc_inode().  IOW, if it's cleaning up the stuff that
+	might have accumulated over the life of in-core inode, ->evict_inode()
+	might be a fit.
+
+	Rules for inode destruction:
+		* if ->destroy_inode() is non-NULL, it gets called
+		* if ->free_inode() is non-NULL, it gets scheduled by call_rcu()
+		* combination of NULL ->destroy_inode and NULL ->free_inode is
+		  treated as NULL/free_inode_nonrcu, to preserve the compatibility.
+
+	Note that the callback (be it via ->free_inode() or explicit call_rcu()
+	in ->destroy_inode()) is *NOT* ordered wrt superblock destruction;
+	as the matter of fact, the superblock and all associated structures
+	might be already gone.  The filesystem driver is guaranteed to be still
+	there, but that's it.  Freeing memory in the callback is fine; doing
+	more than that is possible, but requires a lot of care and is best
+	avoided.