NFS: Add a new ACCESS rpc call cache to the linux nfs client

The current access cache only allows one entry at a time to be cached for each
inode. Add a per-inode red-black tree in order to allow more than one to
be cached at a time.

Should significantly cut down the time spent in path traversal for shared
directories such as ${PATH}, /usr/share, etc.

Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>

3 files changed, 124 insertions, 27 deletions

diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
index e7ffb4deb3e..094afded2b1 100644
--- a/fs/nfs/dir.c
+++ b/fs/nfs/dir.c
@@ -1638,35 +1638,134 @@ out:
 	return error;
 }
 
-int nfs_access_get_cached(struct inode *inode, struct rpc_cred *cred, struct nfs_access_entry *res)
+static void nfs_access_free_entry(struct nfs_access_entry *entry)
+{
+	put_rpccred(entry->cred);
+	kfree(entry);
+}
+
+static void __nfs_access_zap_cache(struct inode *inode)
 {
 	struct nfs_inode *nfsi = NFS_I(inode);
-	struct nfs_access_entry *cache = &nfsi->cache_access;
+	struct rb_root *root_node = &nfsi->access_cache;
+	struct rb_node *n, *dispose = NULL;
+	struct nfs_access_entry *entry;
+
+	/* Unhook entries from the cache */
+	while ((n = rb_first(root_node)) != NULL) {
+		entry = rb_entry(n, struct nfs_access_entry, rb_node);
+		rb_erase(n, root_node);
+		n->rb_left = dispose;
+		dispose = n;
+	}
+	nfsi->cache_validity &= ~NFS_INO_INVALID_ACCESS;
+	spin_unlock(&inode->i_lock);
 
-	if (cache->cred != cred
-			|| time_after(jiffies, cache->jiffies + NFS_ATTRTIMEO(inode))
-			|| (nfsi->cache_validity & NFS_INO_INVALID_ACCESS))
-		return -ENOENT;
-	memcpy(res, cache, sizeof(*res));
-	return 0;
+	/* Now kill them all! */
+	while (dispose != NULL) {
+		n = dispose;
+		dispose = n->rb_left;
+		nfs_access_free_entry(rb_entry(n, struct nfs_access_entry, rb_node));
+	}
 }
 
-void nfs_access_add_cache(struct inode *inode, struct nfs_access_entry *set)
+void nfs_access_zap_cache(struct inode *inode)
 {
-	struct nfs_inode *nfsi = NFS_I(inode);
-	struct nfs_access_entry *cache = &nfsi->cache_access;
+	spin_lock(&inode->i_lock);
+	/* This will release the spinlock */
+	__nfs_access_zap_cache(inode);
+}
 
-	if (cache->cred != set->cred) {
-		if (cache->cred)
-			put_rpccred(cache->cred);
-		cache->cred = get_rpccred(set->cred);
+static struct nfs_access_entry *nfs_access_search_rbtree(struct inode *inode, struct rpc_cred *cred)
+{
+	struct rb_node *n = NFS_I(inode)->access_cache.rb_node;
+	struct nfs_access_entry *entry;
+
+	while (n != NULL) {
+		entry = rb_entry(n, struct nfs_access_entry, rb_node);
+
+		if (cred < entry->cred)
+			n = n->rb_left;
+		else if (cred > entry->cred)
+			n = n->rb_right;
+		else
+			return entry;
 	}
-	/* FIXME: replace current access_cache BKL reliance with inode->i_lock */
+	return NULL;
+}
+
+int nfs_access_get_cached(struct inode *inode, struct rpc_cred *cred, struct nfs_access_entry *res)
+{
+	struct nfs_inode *nfsi = NFS_I(inode);
+	struct nfs_access_entry *cache;
+	int err = -ENOENT;
+
 	spin_lock(&inode->i_lock);
-	nfsi->cache_validity &= ~NFS_INO_INVALID_ACCESS;
+	if (nfsi->cache_validity & NFS_INO_INVALID_ACCESS)
+		goto out_zap;
+	cache = nfs_access_search_rbtree(inode, cred);
+	if (cache == NULL)
+		goto out;
+	if (time_after(jiffies, cache->jiffies + NFS_ATTRTIMEO(inode)))
+		goto out_stale;
+	res->jiffies = cache->jiffies;
+	res->cred = cache->cred;
+	res->mask = cache->mask;
+	err = 0;
+out:
+	spin_unlock(&inode->i_lock);
+	return err;
+out_stale:
+	rb_erase(&cache->rb_node, &nfsi->access_cache);
+	spin_unlock(&inode->i_lock);
+	nfs_access_free_entry(cache);
+	return -ENOENT;
+out_zap:
+	/* This will release the spinlock */
+	__nfs_access_zap_cache(inode);
+	return -ENOENT;
+}
+
+static void nfs_access_add_rbtree(struct inode *inode, struct nfs_access_entry *set)
+{
+	struct rb_root *root_node = &NFS_I(inode)->access_cache;
+	struct rb_node **p = &root_node->rb_node;
+	struct rb_node *parent = NULL;
+	struct nfs_access_entry *entry;
+
+	spin_lock(&inode->i_lock);
+	while (*p != NULL) {
+		parent = *p;
+		entry = rb_entry(parent, struct nfs_access_entry, rb_node);
+
+		if (set->cred < entry->cred)
+			p = &parent->rb_left;
+		else if (set->cred > entry->cred)
+			p = &parent->rb_right;
+		else
+			goto found;
+	}
+	rb_link_node(&set->rb_node, parent, p);
+	rb_insert_color(&set->rb_node, root_node);
 	spin_unlock(&inode->i_lock);
+	return;
+found:
+	rb_replace_node(parent, &set->rb_node, root_node);
+	spin_unlock(&inode->i_lock);
+	nfs_access_free_entry(entry);
+}
+
+void nfs_access_add_cache(struct inode *inode, struct nfs_access_entry *set)
+{
+	struct nfs_access_entry *cache = kmalloc(sizeof(*cache), GFP_KERNEL);
+	if (cache == NULL)
+		return;
+	RB_CLEAR_NODE(&cache->rb_node);
 	cache->jiffies = set->jiffies;
+	cache->cred = get_rpccred(set->cred);
 	cache->mask = set->mask;
+
+	nfs_access_add_rbtree(inode, cache);
 }
 
 static int nfs_do_access(struct inode *inode, struct rpc_cred *cred, int mask)
diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c
index d349fb2245d..b94ab060bb1 100644
--- a/fs/nfs/inode.c
+++ b/fs/nfs/inode.c
@@ -76,19 +76,14 @@ int nfs_write_inode(struct inode *inode, int sync)
 
 void nfs_clear_inode(struct inode *inode)
 {
-	struct nfs_inode *nfsi = NFS_I(inode);
-	struct rpc_cred *cred;
-
 	/*
 	 * The following should never happen...
 	 */
 	BUG_ON(nfs_have_writebacks(inode));
-	BUG_ON (!list_empty(&nfsi->open_files));
+	BUG_ON(!list_empty(&NFS_I(inode)->open_files));
+	BUG_ON(atomic_read(&NFS_I(inode)->data_updates) != 0);
 	nfs_zap_acl_cache(inode);
-	cred = nfsi->cache_access.cred;
-	if (cred)
-		put_rpccred(cred);
-	BUG_ON(atomic_read(&nfsi->data_updates) != 0);
+	nfs_access_zap_cache(inode);
 }
 
 /**
@@ -290,7 +285,7 @@ nfs_fhget(struct super_block *sb, struct nfs_fh *fh, struct nfs_fattr *fattr)
 		nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
 		nfsi->attrtimeo_timestamp = jiffies;
 		memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
-		nfsi->cache_access.cred = NULL;
+		nfsi->access_cache = RB_ROOT;
 
 		unlock_new_inode(inode);
 	} else
diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h
index 6c2066caeaa..cc013ed2e52 100644
--- a/include/linux/nfs_fs.h
+++ b/include/linux/nfs_fs.h
@@ -42,6 +42,7 @@
 #include <linux/in.h>
 #include <linux/mm.h>
 #include <linux/pagemap.h>
+#include <linux/rbtree.h>
 #include <linux/rwsem.h>
 #include <linux/wait.h>
 
@@ -69,6 +70,7 @@
  * NFSv3/v4 Access mode cache entry
  */
 struct nfs_access_entry {
+	struct rb_node		rb_node;
 	unsigned long		jiffies;
 	struct rpc_cred *	cred;
 	int			mask;
@@ -145,7 +147,7 @@ struct nfs_inode {
 	 */
 	atomic_t		data_updates;
 
-	struct nfs_access_entry	cache_access;
+	struct rb_root		access_cache;
 #ifdef CONFIG_NFS_V3_ACL
 	struct posix_acl	*acl_access;
 	struct posix_acl	*acl_default;
@@ -297,6 +299,7 @@ extern int nfs_getattr(struct vfsmount *, struct dentry *, struct kstat *);
 extern int nfs_permission(struct inode *, int, struct nameidata *);
 extern int nfs_access_get_cached(struct inode *, struct rpc_cred *, struct nfs_access_entry *);
 extern void nfs_access_add_cache(struct inode *, struct nfs_access_entry *);
+extern void nfs_access_zap_cache(struct inode *inode);
 extern int nfs_open(struct inode *, struct file *);
 extern int nfs_release(struct inode *, struct file *);
 extern int nfs_attribute_timeout(struct inode *inode);