From: Christoph Hellwig <hch@lst.de>
Date: Tue, 8 Jul 2025 08:41:29 +0000 (+0200)
Subject: NFS: use a hash table for delegation lookup
X-Git-Url: https://www.infradead.org/git/?a=commitdiff_plain;h=refs%2Fheads%2Fnfs-delegation-hash;p=users%2Fhch%2Fmisc.git

NFS: use a hash table for delegation lookup

nfs_delegation_find_inode currently has to walk the entire list of
delegations per inode, which can become pretty large, and can become even
larger when increasing the delegation watermark.

Add a hash table to speed up the delegation lookup, sized as a fraction
of the delegation watermark.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---

diff --git a/fs/nfs/client.c b/fs/nfs/client.c
index c824511c0b23..7cf2acde347d 100644
--- a/fs/nfs/client.c
+++ b/fs/nfs/client.c
@@ -984,6 +984,7 @@ static DEFINE_IDA(s_sysfs_ids);
 struct nfs_server *nfs_alloc_server(void)
 {
 	struct nfs_server *server;
+	int delegation_buckets, i;
 
 	server = kzalloc(sizeof(struct nfs_server), GFP_KERNEL);
 	if (!server)
@@ -1009,11 +1010,18 @@ struct nfs_server *nfs_alloc_server(void)
 	atomic_set(&server->active, 0);
 	atomic_long_set(&server->nr_active_delegations, 0);
 
+	delegation_buckets = roundup_pow_of_two(nfs_delegation_watermark / 16);
+	server->delegation_hash_mask = delegation_buckets - 1;
+	server->delegation_hash_table = kmalloc_array(delegation_buckets,
+			sizeof(*server->delegation_hash_table), GFP_KERNEL);
+	if (!server->delegation_hash_table)
+		goto out_free_server;
+	for (i = 0; i < delegation_buckets; i++)
+		INIT_HLIST_HEAD(&server->delegation_hash_table[i]);
+
 	server->io_stats = nfs_alloc_iostats();
-	if (!server->io_stats) {
-		kfree(server);
-		return NULL;
-	}
+	if (!server->io_stats)
+		goto out_free_delegation_hash;
 
 	server->change_attr_type = NFS4_CHANGE_TYPE_IS_UNDEFINED;
 
@@ -1026,6 +1034,12 @@ struct nfs_server *nfs_alloc_server(void)
 	rpc_init_wait_queue(&server->uoc_rpcwaitq, "NFS UOC");
 
 	return server;
+
+out_free_delegation_hash:
+	kfree(server->delegation_hash_table);
+out_free_server:
+	kfree(server);
+	return NULL;
 }
 EXPORT_SYMBOL_GPL(nfs_alloc_server);
 
@@ -1034,6 +1048,7 @@ static void delayed_free(struct rcu_head *p)
 	struct nfs_server *server = container_of(p, struct nfs_server, rcu);
 
 	nfs_free_iostats(server->io_stats);
+	kfree(server->delegation_hash_table);
 	kfree(server);
 }
 
diff --git a/fs/nfs/delegation.c b/fs/nfs/delegation.c
index b10e47e762fb..ca97aa82ed94 100644
--- a/fs/nfs/delegation.c
+++ b/fs/nfs/delegation.c
@@ -27,9 +27,16 @@
 
 #define NFS_DEFAULT_DELEGATION_WATERMARK (5000U)
 
-static unsigned nfs_delegation_watermark = NFS_DEFAULT_DELEGATION_WATERMARK;
+unsigned nfs_delegation_watermark = NFS_DEFAULT_DELEGATION_WATERMARK;
 module_param_named(delegation_watermark, nfs_delegation_watermark, uint, 0644);
 
+static struct hlist_head *nfs_delegation_hash(struct nfs_server *server,
+		const struct nfs_fh *fhandle)
+{
+	return server->delegation_hash_table +
+		(nfs_fhandle_hash(fhandle) & server->delegation_hash_mask);
+}
+
 static void __nfs_free_delegation(struct nfs_delegation *delegation)
 {
 	put_cred(delegation->cred);
@@ -365,6 +372,7 @@ nfs_detach_delegation_locked(struct nfs_inode *nfsi,
 		spin_unlock(&delegation->lock);
 		return NULL;
 	}
+	hlist_del_init_rcu(&delegation->hash);
 	list_del_rcu(&delegation->super_list);
 	delegation->inode = NULL;
 	rcu_assign_pointer(nfsi->delegation, NULL);
@@ -527,6 +535,8 @@ add_new:
 	spin_unlock(&inode->i_lock);
 
 	list_add_tail_rcu(&delegation->super_list, &server->delegations);
+	hlist_add_head_rcu(&delegation->hash,
+			nfs_delegation_hash(server, &NFS_I(inode)->fh));
 	rcu_assign_pointer(nfsi->delegation, delegation);
 	delegation = NULL;
 
@@ -1162,11 +1172,12 @@ static struct inode *
 nfs_delegation_find_inode_server(struct nfs_server *server,
 				 const struct nfs_fh *fhandle)
 {
+	struct hlist_head *head = nfs_delegation_hash(server, fhandle);
 	struct nfs_delegation *delegation;
 	struct super_block *freeme = NULL;
 	struct inode *res = NULL;
 
-	list_for_each_entry_rcu(delegation, &server->delegations, super_list) {
+	hlist_for_each_entry_rcu(delegation, head, hash) {
 		spin_lock(&delegation->lock);
 		if (delegation->inode != NULL &&
 		    !test_bit(NFS_DELEGATION_REVOKED, &delegation->flags) &&
diff --git a/fs/nfs/delegation.h b/fs/nfs/delegation.h
index 8ff5ab9c5c25..9f1fb9b39c43 100644
--- a/fs/nfs/delegation.h
+++ b/fs/nfs/delegation.h
@@ -14,6 +14,7 @@
  * NFSv4 delegation
  */
 struct nfs_delegation {
+	struct hlist_node hash;
 	struct list_head super_list;
 	const struct cred *cred;
 	struct inode *inode;
@@ -123,4 +124,6 @@ static inline int nfs_have_delegated_mtime(struct inode *inode)
 						 NFS_DELEGATION_FLAG_TIME);
 }
 
+extern unsigned nfs_delegation_watermark;
+
 #endif
diff --git a/include/linux/nfs_fs_sb.h b/include/linux/nfs_fs_sb.h
index df73e97394ad..61f48ab250fb 100644
--- a/include/linux/nfs_fs_sb.h
+++ b/include/linux/nfs_fs_sb.h
@@ -258,6 +258,8 @@ struct nfs_server {
 	struct list_head	layouts;
 	struct list_head	delegations;
 	atomic_long_t		nr_active_delegations;
+	unsigned int		delegation_hash_mask;
+	struct hlist_head	*delegation_hash_table;
 	struct list_head	ss_copies;
 	struct list_head	ss_src_copies;