fs: dcache scale subdirs

Protect d_subdirs and d_child with d_lock, except in filesystems that aren't
using dcache_lock for these anyway (eg. using i_mutex).

Note: if we change the locking rule in future so that ->d_child protection is
provided only with ->d_parent->d_lock, it may allow us to reduce some locking.
But it would be an exception to an otherwise regular locking scheme, so we'd
have to see some good results. Probably not worthwhile.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

1 files changed, 174 insertions, 74 deletions

diff --git a/fs/dcache.c b/fs/dcache.c
index ee127f9ab274..a661247a20d5 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -47,6 +47,8 @@
  *   - d_lru
  *   - d_count
  *   - d_unhashed()
+ *   - d_parent and d_subdirs
+ *   - childrens' d_child and d_parent
  *
  * Ordering:
  * dcache_lock
@@ -223,24 +225,22 @@ static void dentry_lru_move_tail(struct dentry *dentry)
  *
  * If this is the root of the dentry tree, return NULL.
  *
- * dcache_lock and d_lock must be held by caller, are dropped by d_kill.
+ * dcache_lock and d_lock and d_parent->d_lock must be held by caller, and
+ * are dropped by d_kill.
  */
-static struct dentry *d_kill(struct dentry *dentry)
+static struct dentry *d_kill(struct dentry *dentry, struct dentry *parent)
 	__releases(dentry->d_lock)
+	__releases(parent->d_lock)
 	__releases(dcache_lock)
 {
-	struct dentry *parent;
-
 	list_del(&dentry->d_u.d_child);
+	if (parent)
+		spin_unlock(&parent->d_lock);
 	dentry_iput(dentry);
 	/*
 	 * dentry_iput drops the locks, at which point nobody (except
 	 * transient RCU lookups) can reach this dentry.
 	 */
-	if (IS_ROOT(dentry))
-		parent = NULL;
-	else
-		parent = dentry->d_parent;
 	d_free(dentry);
 	return parent;
 }
@@ -312,6 +312,7 @@ EXPORT_SYMBOL(d_drop);
 
 void dput(struct dentry *dentry)
 {
+	struct dentry *parent;
 	if (!dentry)
 		return;
 
@@ -319,6 +320,10 @@ repeat:
 	if (dentry->d_count == 1)
 		might_sleep();
 	spin_lock(&dentry->d_lock);
+	if (IS_ROOT(dentry))
+		parent = NULL;
+	else
+		parent = dentry->d_parent;
 	if (dentry->d_count == 1) {
 		if (!spin_trylock(&dcache_lock)) {
 			/*
@@ -330,10 +335,17 @@ repeat:
 			spin_unlock(&dentry->d_lock);
 			goto repeat;
 		}
+		if (parent && !spin_trylock(&parent->d_lock)) {
+			spin_unlock(&dentry->d_lock);
+			spin_unlock(&dcache_lock);
+			goto repeat;
+		}
 	}
 	dentry->d_count--;
 	if (dentry->d_count) {
 		spin_unlock(&dentry->d_lock);
+		if (parent)
+			spin_unlock(&parent->d_lock);
 		spin_unlock(&dcache_lock);
 		return;
 	}
@@ -355,6 +367,8 @@ repeat:
 	dentry_lru_add(dentry);
 
  	spin_unlock(&dentry->d_lock);
+	if (parent)
+		spin_unlock(&parent->d_lock);
 	spin_unlock(&dcache_lock);
 	return;
 
@@ -363,7 +377,7 @@ unhash_it:
 kill_it:
 	/* if dentry was on the d_lru list delete it from there */
 	dentry_lru_del(dentry);
-	dentry = d_kill(dentry);
+	dentry = d_kill(dentry, parent);
 	if (dentry)
 		goto repeat;
 }
@@ -584,12 +598,13 @@ EXPORT_SYMBOL(d_prune_aliases);
  * quadratic behavior of shrink_dcache_parent(), but is also expected
  * to be beneficial in reducing dentry cache fragmentation.
  */
-static void prune_one_dentry(struct dentry * dentry)
+static void prune_one_dentry(struct dentry *dentry, struct dentry *parent)
 	__releases(dentry->d_lock)
+	__releases(parent->d_lock)
 	__releases(dcache_lock)
 {
 	__d_drop(dentry);
-	dentry = d_kill(dentry);
+	dentry = d_kill(dentry, parent);
 
 	/*
 	 * Prune ancestors.  Locking is simpler than in dput(),
@@ -597,9 +612,20 @@ static void prune_one_dentry(struct dentry * dentry)
 	 */
 	while (dentry) {
 		spin_lock(&dcache_lock);
+again:
 		spin_lock(&dentry->d_lock);
+		if (IS_ROOT(dentry))
+			parent = NULL;
+		else
+			parent = dentry->d_parent;
+		if (parent && !spin_trylock(&parent->d_lock)) {
+			spin_unlock(&dentry->d_lock);
+			goto again;
+		}
 		dentry->d_count--;
 		if (dentry->d_count) {
+			if (parent)
+				spin_unlock(&parent->d_lock);
 			spin_unlock(&dentry->d_lock);
 			spin_unlock(&dcache_lock);
 			return;
@@ -607,7 +633,7 @@ static void prune_one_dentry(struct dentry * dentry)
 
 		dentry_lru_del(dentry);
 		__d_drop(dentry);
-		dentry = d_kill(dentry);
+		dentry = d_kill(dentry, parent);
 	}
 }
 
@@ -616,29 +642,40 @@ static void shrink_dentry_list(struct list_head *list)
 	struct dentry *dentry;
 
 	while (!list_empty(list)) {
+		struct dentry *parent;
+
 		dentry = list_entry(list->prev, struct dentry, d_lru);
 
 		if (!spin_trylock(&dentry->d_lock)) {
+relock:
 			spin_unlock(&dcache_lru_lock);
 			cpu_relax();
 			spin_lock(&dcache_lru_lock);
 			continue;
 		}
 
-		__dentry_lru_del(dentry);
-
 		/*
 		 * We found an inuse dentry which was not removed from
 		 * the LRU because of laziness during lookup.  Do not free
 		 * it - just keep it off the LRU list.
 		 */
 		if (dentry->d_count) {
+			__dentry_lru_del(dentry);
 			spin_unlock(&dentry->d_lock);
 			continue;
 		}
+		if (IS_ROOT(dentry))
+			parent = NULL;
+		else
+			parent = dentry->d_parent;
+		if (parent && !spin_trylock(&parent->d_lock)) {
+			spin_unlock(&dentry->d_lock);
+			goto relock;
+		}
+		__dentry_lru_del(dentry);
 		spin_unlock(&dcache_lru_lock);
 
-		prune_one_dentry(dentry);
+		prune_one_dentry(dentry, parent);
 		/* dcache_lock and dentry->d_lock dropped */
 		spin_lock(&dcache_lock);
 		spin_lock(&dcache_lru_lock);
@@ -833,14 +870,16 @@ static void shrink_dcache_for_umount_subtree(struct dentry *dentry)
 			/* this is a branch with children - detach all of them
 			 * from the system in one go */
 			spin_lock(&dcache_lock);
+			spin_lock(&dentry->d_lock);
 			list_for_each_entry(loop, &dentry->d_subdirs,
 					    d_u.d_child) {
-				spin_lock(&loop->d_lock);
+				spin_lock_nested(&loop->d_lock,
+						DENTRY_D_LOCK_NESTED);
 				dentry_lru_del(loop);
 				__d_drop(loop);
 				spin_unlock(&loop->d_lock);
-				cond_resched_lock(&dcache_lock);
 			}
+			spin_unlock(&dentry->d_lock);
 			spin_unlock(&dcache_lock);
 
 			/* move to the first child */
@@ -868,16 +907,17 @@ static void shrink_dcache_for_umount_subtree(struct dentry *dentry)
 				BUG();
 			}
 
-			if (IS_ROOT(dentry))
+			if (IS_ROOT(dentry)) {
 				parent = NULL;
-			else {
+				list_del(&dentry->d_u.d_child);
+			} else {
 				parent = dentry->d_parent;
 				spin_lock(&parent->d_lock);
 				parent->d_count--;
+				list_del(&dentry->d_u.d_child);
 				spin_unlock(&parent->d_lock);
 			}
 
-			list_del(&dentry->d_u.d_child);
 			detached++;
 
 			inode = dentry->d_inode;
@@ -958,6 +998,7 @@ int have_submounts(struct dentry *parent)
 	spin_lock(&dcache_lock);
 	if (d_mountpoint(parent))
 		goto positive;
+	spin_lock(&this_parent->d_lock);
 repeat:
 	next = this_parent->d_subdirs.next;
 resume:
@@ -965,22 +1006,34 @@ resume:
 		struct list_head *tmp = next;
 		struct dentry *dentry = list_entry(tmp, struct dentry, d_u.d_child);
 		next = tmp->next;
+
+		spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
 		/* Have we found a mount point ? */
-		if (d_mountpoint(dentry))
+		if (d_mountpoint(dentry)) {
+			spin_unlock(&dentry->d_lock);
+			spin_unlock(&this_parent->d_lock);
 			goto positive;
+		}
 		if (!list_empty(&dentry->d_subdirs)) {
+			spin_unlock(&this_parent->d_lock);
+			spin_release(&dentry->d_lock.dep_map, 1, _RET_IP_);
 			this_parent = dentry;
+			spin_acquire(&this_parent->d_lock.dep_map, 0, 1, _RET_IP_);
 			goto repeat;
 		}
+		spin_unlock(&dentry->d_lock);
 	}
 	/*
 	 * All done at this level ... ascend and resume the search.
 	 */
 	if (this_parent != parent) {
 		next = this_parent->d_u.d_child.next;
+		spin_unlock(&this_parent->d_lock);
 		this_parent = this_parent->d_parent;
+		spin_lock(&this_parent->d_lock);
 		goto resume;
 	}
+	spin_unlock(&this_parent->d_lock);
 	spin_unlock(&dcache_lock);
 	return 0; /* No mount points found in tree */
 positive:
@@ -1010,6 +1063,7 @@ static int select_parent(struct dentry * parent)
 	int found = 0;
 
 	spin_lock(&dcache_lock);
+	spin_lock(&this_parent->d_lock);
 repeat:
 	next = this_parent->d_subdirs.next;
 resume:
@@ -1017,8 +1071,9 @@ resume:
 		struct list_head *tmp = next;
 		struct dentry *dentry = list_entry(tmp, struct dentry, d_u.d_child);
 		next = tmp->next;
+		BUG_ON(this_parent == dentry);
 
-		spin_lock(&dentry->d_lock);
+		spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
 
 		/* 
 		 * move only zero ref count dentries to the end 
@@ -1031,33 +1086,44 @@ resume:
 			dentry_lru_del(dentry);
 		}
 
-		spin_unlock(&dentry->d_lock);
-
 		/*
 		 * We can return to the caller if we have found some (this
 		 * ensures forward progress). We'll be coming back to find
 		 * the rest.
 		 */
-		if (found && need_resched())
+		if (found && need_resched()) {
+			spin_unlock(&dentry->d_lock);
 			goto out;
+		}
 
 		/*
 		 * Descend a level if the d_subdirs list is non-empty.
 		 */
 		if (!list_empty(&dentry->d_subdirs)) {
+			spin_unlock(&this_parent->d_lock);
+			spin_release(&dentry->d_lock.dep_map, 1, _RET_IP_);
 			this_parent = dentry;
+			spin_acquire(&this_parent->d_lock.dep_map, 0, 1, _RET_IP_);
 			goto repeat;
 		}
+
+		spin_unlock(&dentry->d_lock);
 	}
 	/*
 	 * All done at this level ... ascend and resume the search.
 	 */
 	if (this_parent != parent) {
+		struct dentry *tmp;
 		next = this_parent->d_u.d_child.next;
-		this_parent = this_parent->d_parent;
+		tmp = this_parent->d_parent;
+		spin_unlock(&this_parent->d_lock);
+		BUG_ON(tmp == this_parent);
+		this_parent = tmp;
+		spin_lock(&this_parent->d_lock);
 		goto resume;
 	}
 out:
+	spin_unlock(&this_parent->d_lock);
 	spin_unlock(&dcache_lock);
 	return found;
 }
@@ -1155,18 +1221,19 @@ struct dentry *d_alloc(struct dentry * parent, const struct qstr *name)
 	INIT_LIST_HEAD(&dentry->d_lru);
 	INIT_LIST_HEAD(&dentry->d_subdirs);
 	INIT_LIST_HEAD(&dentry->d_alias);
+	INIT_LIST_HEAD(&dentry->d_u.d_child);
 
 	if (parent) {
-		dentry->d_parent = dget(parent);
+		spin_lock(&dcache_lock);
+		spin_lock(&parent->d_lock);
+		spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
+		dentry->d_parent = dget_dlock(parent);
 		dentry->d_sb = parent->d_sb;
-	} else {
-		INIT_LIST_HEAD(&dentry->d_u.d_child);
-	}
-
-	spin_lock(&dcache_lock);
-	if (parent)
 		list_add(&dentry->d_u.d_child, &parent->d_subdirs);
-	spin_unlock(&dcache_lock);
+		spin_unlock(&dentry->d_lock);
+		spin_unlock(&parent->d_lock);
+		spin_unlock(&dcache_lock);
+	}
 
 	this_cpu_inc(nr_dentry);
 
@@ -1684,13 +1751,18 @@ int d_validate(struct dentry *dentry, struct dentry *dparent)
 	struct dentry *child;
 
 	spin_lock(&dcache_lock);
+	spin_lock(&dparent->d_lock);
 	list_for_each_entry(child, &dparent->d_subdirs, d_u.d_child) {
 		if (dentry == child) {
-			__dget_locked(dentry);
+			spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
+			__dget_locked_dlock(dentry);
+			spin_unlock(&dentry->d_lock);
+			spin_unlock(&dparent->d_lock);
 			spin_unlock(&dcache_lock);
 			return 1;
 		}
 	}
+	spin_unlock(&dparent->d_lock);
 	spin_unlock(&dcache_lock);
 
 	return 0;
@@ -1802,17 +1874,6 @@ void dentry_update_name_case(struct dentry *dentry, struct qstr *name)
 }
 EXPORT_SYMBOL(dentry_update_name_case);
 
-/*
- * When switching names, the actual string doesn't strictly have to
- * be preserved in the target - because we're dropping the target
- * anyway. As such, we can just do a simple memcpy() to copy over
- * the new name before we switch.
- *
- * Note that we have to be a lot more careful about getting the hash
- * switched - we have to switch the hash value properly even if it
- * then no longer matches the actual (corrupted) string of the target.
- * The hash value has to match the hash queue that the dentry is on..
- */
 static void switch_names(struct dentry *dentry, struct dentry *target)
 {
 	if (dname_external(target)) {
@@ -1854,18 +1915,53 @@ static void switch_names(struct dentry *dentry, struct dentry *target)
 	swap(dentry->d_name.len, target->d_name.len);
 }
 
+static void dentry_lock_for_move(struct dentry *dentry, struct dentry *target)
+{
+	/*
+	 * XXXX: do we really need to take target->d_lock?
+	 */
+	if (IS_ROOT(dentry) || dentry->d_parent == target->d_parent)
+		spin_lock(&target->d_parent->d_lock);
+	else {
+		if (d_ancestor(dentry->d_parent, target->d_parent)) {
+			spin_lock(&dentry->d_parent->d_lock);
+			spin_lock_nested(&target->d_parent->d_lock,
+						DENTRY_D_LOCK_NESTED);
+		} else {
+			spin_lock(&target->d_parent->d_lock);
+			spin_lock_nested(&dentry->d_parent->d_lock,
+						DENTRY_D_LOCK_NESTED);
+		}
+	}
+	if (target < dentry) {
+		spin_lock_nested(&target->d_lock, 2);
+		spin_lock_nested(&dentry->d_lock, 3);
+	} else {
+		spin_lock_nested(&dentry->d_lock, 2);
+		spin_lock_nested(&target->d_lock, 3);
+	}
+}
+
+static void dentry_unlock_parents_for_move(struct dentry *dentry,
+					struct dentry *target)
+{
+	if (target->d_parent != dentry->d_parent)
+		spin_unlock(&dentry->d_parent->d_lock);
+	if (target->d_parent != target)
+		spin_unlock(&target->d_parent->d_lock);
+}
+
 /*
- * We cannibalize "target" when moving dentry on top of it,
- * because it's going to be thrown away anyway. We could be more
- * polite about it, though.
- *
- * This forceful removal will result in ugly /proc output if
- * somebody holds a file open that got deleted due to a rename.
- * We could be nicer about the deleted file, and let it show
- * up under the name it had before it was deleted rather than
- * under the original name of the file that was moved on top of it.
+ * When switching names, the actual string doesn't strictly have to
+ * be preserved in the target - because we're dropping the target
+ * anyway. As such, we can just do a simple memcpy() to copy over
+ * the new name before we switch.
+ *
+ * Note that we have to be a lot more careful about getting the hash
+ * switched - we have to switch the hash value properly even if it
+ * then no longer matches the actual (corrupted) string of the target.
+ * The hash value has to match the hash queue that the dentry is on..
  */
- 
 /*
  * d_move_locked - move a dentry
  * @dentry: entry to move
@@ -1879,20 +1975,12 @@ static void d_move_locked(struct dentry * dentry, struct dentry * target)
 	if (!dentry->d_inode)
 		printk(KERN_WARNING "VFS: moving negative dcache entry\n");
 
+	BUG_ON(d_ancestor(dentry, target));
+	BUG_ON(d_ancestor(target, dentry));
+
 	write_seqlock(&rename_lock);
-	/*
-	 * XXXX: do we really need to take target->d_lock?
-	 */
-	if (d_ancestor(dentry, target)) {
-		spin_lock(&dentry->d_lock);
-		spin_lock_nested(&target->d_lock, DENTRY_D_LOCK_NESTED);
-	} else if (d_ancestor(target, dentry) || target < dentry) {
-		spin_lock(&target->d_lock);
-		spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
-	} else {
-		spin_lock(&dentry->d_lock);
-		spin_lock_nested(&target->d_lock, DENTRY_D_LOCK_NESTED);
-	}
+
+	dentry_lock_for_move(dentry, target);
 
 	/* Move the dentry to the target hash queue, if on different bucket */
 	spin_lock(&dcache_hash_lock);
@@ -1924,6 +2012,8 @@ static void d_move_locked(struct dentry * dentry, struct dentry * target)
 	}
 
 	list_add(&dentry->d_u.d_child, &dentry->d_parent->d_subdirs);
+
+	dentry_unlock_parents_for_move(dentry, target);
 	spin_unlock(&target->d_lock);
 	fsnotify_d_move(dentry);
 	spin_unlock(&dentry->d_lock);
@@ -2013,17 +2103,20 @@ out_err:
 /*
  * Prepare an anonymous dentry for life in the superblock's dentry tree as a
  * named dentry in place of the dentry to be replaced.
+ * returns with anon->d_lock held!
  */
 static void __d_materialise_dentry(struct dentry *dentry, struct dentry *anon)
 {
 	struct dentry *dparent, *aparent;
 
-	switch_names(dentry, anon);
-	swap(dentry->d_name.hash, anon->d_name.hash);
+	dentry_lock_for_move(anon, dentry);
 
 	dparent = dentry->d_parent;
 	aparent = anon->d_parent;
 
+	switch_names(dentry, anon);
+	swap(dentry->d_name.hash, anon->d_name.hash);
+
 	dentry->d_parent = (aparent == anon) ? dentry : aparent;
 	list_del(&dentry->d_u.d_child);
 	if (!IS_ROOT(dentry))
@@ -2038,6 +2131,10 @@ static void __d_materialise_dentry(struct dentry *dentry, struct dentry *anon)
 	else
 		INIT_LIST_HEAD(&anon->d_u.d_child);
 
+	dentry_unlock_parents_for_move(anon, dentry);
+	spin_unlock(&dentry->d_lock);
+
+	/* anon->d_lock still locked, returns locked */
 	anon->d_flags &= ~DCACHE_DISCONNECTED;
 }
 
@@ -2073,7 +2170,6 @@ struct dentry *d_materialise_unique(struct dentry *dentry, struct inode *inode)
 			/* Is this an anonymous mountpoint that we could splice
 			 * into our tree? */
 			if (IS_ROOT(alias)) {
-				spin_lock(&alias->d_lock);
 				__d_materialise_dentry(dentry, alias);
 				__d_drop(alias);
 				goto found;
@@ -2558,6 +2654,7 @@ void d_genocide(struct dentry *root)
 	struct list_head *next;
 
 	spin_lock(&dcache_lock);
+	spin_lock(&this_parent->d_lock);
 repeat:
 	next = this_parent->d_subdirs.next;
 resume:
@@ -2571,8 +2668,10 @@ resume:
 			continue;
 		}
 		if (!list_empty(&dentry->d_subdirs)) {
-			spin_unlock(&dentry->d_lock);
+			spin_unlock(&this_parent->d_lock);
+			spin_release(&dentry->d_lock.dep_map, 1, _RET_IP_);
 			this_parent = dentry;
+			spin_acquire(&this_parent->d_lock.dep_map, 0, 1, _RET_IP_);
 			goto repeat;
 		}
 		dentry->d_count--;
@@ -2580,12 +2679,13 @@ resume:
 	}
 	if (this_parent != root) {
 		next = this_parent->d_u.d_child.next;
-		spin_lock(&this_parent->d_lock);
 		this_parent->d_count--;
 		spin_unlock(&this_parent->d_lock);
 		this_parent = this_parent->d_parent;
+		spin_lock(&this_parent->d_lock);
 		goto resume;
 	}
+	spin_unlock(&this_parent->d_lock);
 	spin_unlock(&dcache_lock);
 }