parallel lookups machinery, part 2

We'll need to verify that there's neither a hashed nor in-lookup
dentry with desired parent/name before adding to in-lookup set.

One possible solution would be to hold the parent's ->d_lock through
both checks, but while the in-lookup set is relatively small at any
time, dcache is not.  And holding the parent's ->d_lock through
something like __d_lookup_rcu() would suck too badly.

So we leave the parent's ->d_lock alone, which means that we watch
out for the following scenario:
	* we verify that there's no hashed match
	* existing in-lookup match gets hashed by another process
	* we verify that there's no in-lookup matches and decide
that everything's fine.

Solution: per-directory kinda-sorta seqlock, bumped around the times
we hash something that used to be in-lookup or move (and hash)
something in place of in-lookup.  Then the above would turn into
	* read the counter
	* do dcache lookup
	* if no matches found, check for in-lookup matches
	* if there had been none of those either, check if the
counter has changed; repeat if it has.

The "kinda-sorta" part is due to the fact that we don't have much spare
space in inode.  There is a spare word (shared with i_bdev/i_cdev/i_pipe),
so the counter part is not a problem, but spinlock is a different story.

We could use the parent's ->d_lock, and it would be less painful in
terms of contention, for __d_add() it would be rather inconvenient to
grab; we could do that (using lock_parent()), but...

Fortunately, we can get serialization on the counter itself, and it
might be a good idea in general; we can use cmpxchg() in a loop to
get from even to odd and smp_store_release() from odd to even.

This commit adds the counter and updating logics; the readers will be
added in the next commit.

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

5 files changed, 44 insertions, 3 deletions

diff --git a/Documentation/filesystems/porting b/Documentation/filesystems/porting
index 57bb375..8810e23 100644
--- a/Documentation/filesystems/porting
+++ b/Documentation/filesystems/porting
@@ -531,3 +531,11 @@ in your dentry operations instead.
 	dentry might be yet to be attached to inode, so do _not_ use its ->d_inode
 	in the instances.  Rationale: !@#!@# security_d_instantiate() needs to be
 	called before we attach dentry to inode.
+--
+[mandatory]
+	symlinks are no longer the only inodes that do *not* have i_bdev/i_cdev/
+	i_pipe/i_link union zeroed out at inode eviction.  As the result, you can't
+	assume that non-NULL value in ->i_nlink at ->destroy_inode() implies that
+	it's a symlink.  Checking ->i_mode is really needed now.  In-tree we had
+	to fix shmem_destroy_callback() that used to take that kind of shortcut;
+	watch out, since that shortcut is no longer valid.
diff --git a/fs/dcache.c b/fs/dcache.c
index 0f1d938..10988f7 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -2364,6 +2364,22 @@ void d_rehash(struct dentry * entry)
 }
 EXPORT_SYMBOL(d_rehash);
 
+static inline unsigned start_dir_add(struct inode *dir)
+{
+
+	for (;;) {
+		unsigned n = dir->i_dir_seq;
+		if (!(n & 1) && cmpxchg(&dir->i_dir_seq, n, n + 1) == n)
+			return n;
+		cpu_relax();
+	}
+}
+
+static inline void end_dir_add(struct inode *dir, unsigned n)
+{
+	smp_store_release(&dir->i_dir_seq, n + 2);
+}
+
 void __d_lookup_done(struct dentry *dentry)
 {
 	dentry->d_flags &= ~DCACHE_PAR_LOOKUP;
@@ -2375,9 +2391,14 @@ EXPORT_SYMBOL(__d_lookup_done);
 
 static inline void __d_add(struct dentry *dentry, struct inode *inode)
 {
+	struct inode *dir = NULL;
+	unsigned n;
 	spin_lock(&dentry->d_lock);
-	if (unlikely(d_in_lookup(dentry)))
+	if (unlikely(d_in_lookup(dentry))) {
+		dir = dentry->d_parent->d_inode;
+		n = start_dir_add(dir);
 		__d_lookup_done(dentry);
+	}
 	if (inode) {
 		unsigned add_flags = d_flags_for_inode(inode);
 		hlist_add_head(&dentry->d_u.d_alias, &inode->i_dentry);
@@ -2387,6 +2408,8 @@ static inline void __d_add(struct dentry *dentry, struct inode *inode)
 		__fsnotify_d_instantiate(dentry);
 	}
 	_d_rehash(dentry);
+	if (dir)
+		end_dir_add(dir, n);
 	spin_unlock(&dentry->d_lock);
 	if (inode)
 		spin_unlock(&inode->i_lock);
@@ -2616,6 +2639,8 @@ static void dentry_unlock_for_move(struct dentry *dentry, struct dentry *target)
 static void __d_move(struct dentry *dentry, struct dentry *target,
 		     bool exchange)
 {
+	struct inode *dir = NULL;
+	unsigned n;
 	if (!dentry->d_inode)
 		printk(KERN_WARNING "VFS: moving negative dcache entry\n");
 
@@ -2623,8 +2648,11 @@ static void __d_move(struct dentry *dentry, struct dentry *target,
 	BUG_ON(d_ancestor(target, dentry));
 
 	dentry_lock_for_move(dentry, target);
-	if (unlikely(d_in_lookup(target)))
+	if (unlikely(d_in_lookup(target))) {
+		dir = target->d_parent->d_inode;
+		n = start_dir_add(dir);
 		__d_lookup_done(target);
+	}
 
 	write_seqcount_begin(&dentry->d_seq);
 	write_seqcount_begin_nested(&target->d_seq, DENTRY_D_LOCK_NESTED);
@@ -2674,6 +2702,8 @@ static void __d_move(struct dentry *dentry, struct dentry *target,
 	write_seqcount_end(&target->d_seq);
 	write_seqcount_end(&dentry->d_seq);
 
+	if (dir)
+		end_dir_add(dir, n);
 	dentry_unlock_for_move(dentry, target);
 }
 
diff --git a/fs/inode.c b/fs/inode.c
index 4202aac..4b884f7 100644
--- a/fs/inode.c
+++ b/fs/inode.c
@@ -151,6 +151,7 @@ int inode_init_always(struct super_block *sb, struct inode *inode)
 	inode->i_bdev = NULL;
 	inode->i_cdev = NULL;
 	inode->i_link = NULL;
+	inode->i_dir_seq = 0;
 	inode->i_rdev = 0;
 	inode->dirtied_when = 0;
 
diff --git a/include/linux/fs.h b/include/linux/fs.h
index 6d0fa91..00cecc5 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -684,6 +684,7 @@ struct inode {
 		struct block_device	*i_bdev;
 		struct cdev		*i_cdev;
 		char			*i_link;
+		unsigned		i_dir_seq;
 	};
 
 	__u32			i_generation;
diff --git a/mm/shmem.c b/mm/shmem.c
index 4640699..e684a91 100644
--- a/mm/shmem.c
+++ b/mm/shmem.c
@@ -3123,7 +3123,8 @@ static struct inode *shmem_alloc_inode(struct super_block *sb)
 static void shmem_destroy_callback(struct rcu_head *head)
 {
 	struct inode *inode = container_of(head, struct inode, i_rcu);
-	kfree(inode->i_link);
+	if (S_ISLNK(inode->i_mode))
+		kfree(inode->i_link);
 	kmem_cache_free(shmem_inode_cachep, SHMEM_I(inode));
 }