X-Git-Url: http://ftp.safe.ca/?a=blobdiff_plain;f=fs%2Fnfs%2Fnfs4state.c;h=16c9fbdf97b40baa1fb8c0e0042304b033c4268b;hb=19e03c570e6099ffaf24e5628d4fe1a8acbe820d;hp=17b187f2d77691ba9c6325e56526d57e6d027e3b;hpb=4ce79717ce32a9f88c1ddce4b9658556cb59d37a;p=safe%2Fjmp%2Flinux-2.6

diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c
index 17b187f..16c9fbd 100644
--- a/fs/nfs/nfs4state.c
+++ b/fs/nfs/nfs4state.c
@@ -38,219 +38,234 @@
  * subsequent patch.
  */
 
-#include <linux/config.h>
+#include <linux/kernel.h>
 #include <linux/slab.h>
 #include <linux/smp_lock.h>
 #include <linux/nfs_fs.h>
 #include <linux/nfs_idmap.h>
+#include <linux/kthread.h>
+#include <linux/module.h>
+#include <linux/random.h>
 #include <linux/workqueue.h>
 #include <linux/bitops.h>
 
 #include "nfs4_fs.h"
 #include "callback.h"
 #include "delegation.h"
+#include "internal.h"
 
 #define OPENOWNER_POOL_SIZE	8
 
 const nfs4_stateid zero_stateid;
 
-static DEFINE_SPINLOCK(state_spinlock);
 static LIST_HEAD(nfs4_clientid_list);
 
-static void nfs4_recover_state(void *);
-
-void
-init_nfsv4_state(struct nfs_server *server)
+static int nfs4_init_client(struct nfs_client *clp, struct rpc_cred *cred)
 {
-	server->nfs4_state = NULL;
-	INIT_LIST_HEAD(&server->nfs4_siblings);
+	int status = nfs4_proc_setclientid(clp, NFS4_CALLBACK,
+			nfs_callback_tcpport, cred);
+	if (status == 0)
+		status = nfs4_proc_setclientid_confirm(clp, cred);
+	if (status == 0)
+		nfs4_schedule_state_renewal(clp);
+	return status;
 }
 
-void
-destroy_nfsv4_state(struct nfs_server *server)
+static struct rpc_cred *nfs4_get_machine_cred_locked(struct nfs_client *clp)
 {
-	if (server->mnt_path) {
-		kfree(server->mnt_path);
-		server->mnt_path = NULL;
-	}
-	if (server->nfs4_state) {
-		nfs4_put_client(server->nfs4_state);
-		server->nfs4_state = NULL;
-	}
+	struct rpc_cred *cred = NULL;
+
+	if (clp->cl_machine_cred != NULL)
+		cred = get_rpccred(clp->cl_machine_cred);
+	return cred;
 }
 
-/*
- * nfs4_get_client(): returns an empty client structure
- * nfs4_put_client(): drops reference to client structure
- *
- * Since these are allocated/deallocated very rarely, we don't
- * bother putting them in a slab cache...
- */
-static struct nfs4_client *
-nfs4_alloc_client(struct in_addr *addr)
+static void nfs4_clear_machine_cred(struct nfs_client *clp)
 {
-	struct nfs4_client *clp;
+	struct rpc_cred *cred;
 
-	if (nfs_callback_up() < 0)
-		return NULL;
-	if ((clp = kmalloc(sizeof(*clp), GFP_KERNEL)) == NULL) {
-		nfs_callback_down();
-		return NULL;
-	}
-	memset(clp, 0, sizeof(*clp));
-	memcpy(&clp->cl_addr, addr, sizeof(clp->cl_addr));
-	init_rwsem(&clp->cl_sem);
-	INIT_LIST_HEAD(&clp->cl_delegations);
-	INIT_LIST_HEAD(&clp->cl_state_owners);
-	INIT_LIST_HEAD(&clp->cl_unused);
-	spin_lock_init(&clp->cl_lock);
-	atomic_set(&clp->cl_count, 1);
-	INIT_WORK(&clp->cl_recoverd, nfs4_recover_state, clp);
-	INIT_WORK(&clp->cl_renewd, nfs4_renew_state, clp);
-	INIT_LIST_HEAD(&clp->cl_superblocks);
-	init_waitqueue_head(&clp->cl_waitq);
-	rpc_init_wait_queue(&clp->cl_rpcwaitq, "NFS4 client");
-	clp->cl_boot_time = CURRENT_TIME;
-	clp->cl_state = 1 << NFS4CLNT_OK;
-	return clp;
+	spin_lock(&clp->cl_lock);
+	cred = clp->cl_machine_cred;
+	clp->cl_machine_cred = NULL;
+	spin_unlock(&clp->cl_lock);
+	if (cred != NULL)
+		put_rpccred(cred);
 }
 
-static void
-nfs4_free_client(struct nfs4_client *clp)
+struct rpc_cred *nfs4_get_renew_cred_locked(struct nfs_client *clp)
 {
 	struct nfs4_state_owner *sp;
+	struct rb_node *pos;
+	struct rpc_cred *cred = NULL;
 
-	while (!list_empty(&clp->cl_unused)) {
-		sp = list_entry(clp->cl_unused.next,
-				struct nfs4_state_owner,
-				so_list);
-		list_del(&sp->so_list);
-		kfree(sp);
-	}
-	BUG_ON(!list_empty(&clp->cl_state_owners));
-	if (clp->cl_cred)
-		put_rpccred(clp->cl_cred);
-	nfs_idmap_delete(clp);
-	if (clp->cl_rpcclient)
-		rpc_shutdown_client(clp->cl_rpcclient);
-	kfree(clp);
-	nfs_callback_down();
-}
-
-static struct nfs4_client *__nfs4_find_client(struct in_addr *addr)
-{
-	struct nfs4_client *clp;
-	list_for_each_entry(clp, &nfs4_clientid_list, cl_servers) {
-		if (memcmp(&clp->cl_addr, addr, sizeof(clp->cl_addr)) == 0) {
-			atomic_inc(&clp->cl_count);
-			return clp;
-		}
+	for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) {
+		sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
+		if (list_empty(&sp->so_states))
+			continue;
+		cred = get_rpccred(sp->so_cred);
+		break;
 	}
-	return NULL;
+	return cred;
 }
 
-struct nfs4_client *nfs4_find_client(struct in_addr *addr)
+static struct rpc_cred *nfs4_get_renew_cred(struct nfs_client *clp)
 {
-	struct nfs4_client *clp;
-	spin_lock(&state_spinlock);
-	clp = __nfs4_find_client(addr);
-	spin_unlock(&state_spinlock);
-	return clp;
+	struct rpc_cred *cred;
+
+	spin_lock(&clp->cl_lock);
+	cred = nfs4_get_renew_cred_locked(clp);
+	spin_unlock(&clp->cl_lock);
+	return cred;
 }
 
-struct nfs4_client *
-nfs4_get_client(struct in_addr *addr)
+static struct rpc_cred *nfs4_get_setclientid_cred(struct nfs_client *clp)
 {
-	struct nfs4_client *clp, *new = NULL;
+	struct nfs4_state_owner *sp;
+	struct rb_node *pos;
+	struct rpc_cred *cred;
+
+	spin_lock(&clp->cl_lock);
+	cred = nfs4_get_machine_cred_locked(clp);
+	if (cred != NULL)
+		goto out;
+	pos = rb_first(&clp->cl_state_owners);
+	if (pos != NULL) {
+		sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
+		cred = get_rpccred(sp->so_cred);
+	}
+out:
+	spin_unlock(&clp->cl_lock);
+	return cred;
+}
 
-	spin_lock(&state_spinlock);
+static void nfs_alloc_unique_id(struct rb_root *root, struct nfs_unique_id *new,
+		__u64 minval, int maxbits)
+{
+	struct rb_node **p, *parent;
+	struct nfs_unique_id *pos;
+	__u64 mask = ~0ULL;
+
+	if (maxbits < 64)
+		mask = (1ULL << maxbits) - 1ULL;
+
+	/* Ensure distribution is more or less flat */
+	get_random_bytes(&new->id, sizeof(new->id));
+	new->id &= mask;
+	if (new->id < minval)
+		new->id += minval;
+retry:
+	p = &root->rb_node;
+	parent = NULL;
+
+	while (*p != NULL) {
+		parent = *p;
+		pos = rb_entry(parent, struct nfs_unique_id, rb_node);
+
+		if (new->id < pos->id)
+			p = &(*p)->rb_left;
+		else if (new->id > pos->id)
+			p = &(*p)->rb_right;
+		else
+			goto id_exists;
+	}
+	rb_link_node(&new->rb_node, parent, p);
+	rb_insert_color(&new->rb_node, root);
+	return;
+id_exists:
 	for (;;) {
-		clp = __nfs4_find_client(addr);
-		if (clp != NULL)
-			break;
-		clp = new;
-		if (clp != NULL) {
-			list_add(&clp->cl_servers, &nfs4_clientid_list);
-			new = NULL;
+		new->id++;
+		if (new->id < minval || (new->id & mask) != new->id) {
+			new->id = minval;
 			break;
 		}
-		spin_unlock(&state_spinlock);
-		new = nfs4_alloc_client(addr);
-		spin_lock(&state_spinlock);
-		if (new == NULL)
+		parent = rb_next(parent);
+		if (parent == NULL)
+			break;
+		pos = rb_entry(parent, struct nfs_unique_id, rb_node);
+		if (new->id < pos->id)
 			break;
 	}
-	spin_unlock(&state_spinlock);
-	if (new)
-		nfs4_free_client(new);
-	return clp;
+	goto retry;
 }
 
-void
-nfs4_put_client(struct nfs4_client *clp)
+static void nfs_free_unique_id(struct rb_root *root, struct nfs_unique_id *id)
 {
-	if (!atomic_dec_and_lock(&clp->cl_count, &state_spinlock))
-		return;
-	list_del(&clp->cl_servers);
-	spin_unlock(&state_spinlock);
-	BUG_ON(!list_empty(&clp->cl_superblocks));
-	wake_up_all(&clp->cl_waitq);
-	rpc_wake_up(&clp->cl_rpcwaitq);
-	nfs4_kill_renewd(clp);
-	nfs4_free_client(clp);
+	rb_erase(&id->rb_node, root);
 }
 
-static int __nfs4_init_client(struct nfs4_client *clp)
+static struct nfs4_state_owner *
+nfs4_find_state_owner(struct nfs_server *server, struct rpc_cred *cred)
 {
-	int status = nfs4_proc_setclientid(clp, NFS4_CALLBACK, nfs_callback_tcpport);
-	if (status == 0)
-		status = nfs4_proc_setclientid_confirm(clp);
-	if (status == 0)
-		nfs4_schedule_state_renewal(clp);
-	return status;
-}
+	struct nfs_client *clp = server->nfs_client;
+	struct rb_node **p = &clp->cl_state_owners.rb_node,
+		       *parent = NULL;
+	struct nfs4_state_owner *sp, *res = NULL;
 
-int nfs4_init_client(struct nfs4_client *clp)
-{
-	return nfs4_map_errors(__nfs4_init_client(clp));
-}
+	while (*p != NULL) {
+		parent = *p;
+		sp = rb_entry(parent, struct nfs4_state_owner, so_client_node);
 
-u32
-nfs4_alloc_lockowner_id(struct nfs4_client *clp)
-{
-	return clp->cl_lockowner_id ++;
+		if (server < sp->so_server) {
+			p = &parent->rb_left;
+			continue;
+		}
+		if (server > sp->so_server) {
+			p = &parent->rb_right;
+			continue;
+		}
+		if (cred < sp->so_cred)
+			p = &parent->rb_left;
+		else if (cred > sp->so_cred)
+			p = &parent->rb_right;
+		else {
+			atomic_inc(&sp->so_count);
+			res = sp;
+			break;
+		}
+	}
+	return res;
 }
 
 static struct nfs4_state_owner *
-nfs4_client_grab_unused(struct nfs4_client *clp, struct rpc_cred *cred)
+nfs4_insert_state_owner(struct nfs_client *clp, struct nfs4_state_owner *new)
 {
-	struct nfs4_state_owner *sp = NULL;
+	struct rb_node **p = &clp->cl_state_owners.rb_node,
+		       *parent = NULL;
+	struct nfs4_state_owner *sp;
 
-	if (!list_empty(&clp->cl_unused)) {
-		sp = list_entry(clp->cl_unused.next, struct nfs4_state_owner, so_list);
-		atomic_inc(&sp->so_count);
-		sp->so_cred = cred;
-		list_move(&sp->so_list, &clp->cl_state_owners);
-		clp->cl_nunused--;
+	while (*p != NULL) {
+		parent = *p;
+		sp = rb_entry(parent, struct nfs4_state_owner, so_client_node);
+
+		if (new->so_server < sp->so_server) {
+			p = &parent->rb_left;
+			continue;
+		}
+		if (new->so_server > sp->so_server) {
+			p = &parent->rb_right;
+			continue;
+		}
+		if (new->so_cred < sp->so_cred)
+			p = &parent->rb_left;
+		else if (new->so_cred > sp->so_cred)
+			p = &parent->rb_right;
+		else {
+			atomic_inc(&sp->so_count);
+			return sp;
+		}
 	}
-	return sp;
+	nfs_alloc_unique_id(&clp->cl_openowner_id, &new->so_owner_id, 1, 64);
+	rb_link_node(&new->so_client_node, parent, p);
+	rb_insert_color(&new->so_client_node, &clp->cl_state_owners);
+	return new;
 }
 
-static struct nfs4_state_owner *
-nfs4_find_state_owner(struct nfs4_client *clp, struct rpc_cred *cred)
+static void
+nfs4_remove_state_owner(struct nfs_client *clp, struct nfs4_state_owner *sp)
 {
-	struct nfs4_state_owner *sp, *res = NULL;
-
-	list_for_each_entry(sp, &clp->cl_state_owners, so_list) {
-		if (sp->so_cred != cred)
-			continue;
-		atomic_inc(&sp->so_count);
-		/* Move to the head of the list */
-		list_move(&sp->so_list, &clp->cl_state_owners);
-		res = sp;
-		break;
-	}
-	return res;
+	if (!RB_EMPTY_NODE(&sp->so_client_node))
+		rb_erase(&sp->so_client_node, &clp->cl_state_owners);
+	nfs_free_unique_id(&clp->cl_openowner_id, &sp->so_owner_id);
 }
 
 /*
@@ -263,24 +278,31 @@ nfs4_alloc_state_owner(void)
 {
 	struct nfs4_state_owner *sp;
 
-	sp = kmalloc(sizeof(*sp),GFP_KERNEL);
+	sp = kzalloc(sizeof(*sp),GFP_KERNEL);
 	if (!sp)
 		return NULL;
-	init_MUTEX(&sp->so_sema);
-	sp->so_seqid = 0;                 /* arbitrary */
+	spin_lock_init(&sp->so_lock);
 	INIT_LIST_HEAD(&sp->so_states);
 	INIT_LIST_HEAD(&sp->so_delegations);
+	rpc_init_wait_queue(&sp->so_sequence.wait, "Seqid_waitqueue");
+	sp->so_seqid.sequence = &sp->so_sequence;
+	spin_lock_init(&sp->so_sequence.lock);
+	INIT_LIST_HEAD(&sp->so_sequence.list);
 	atomic_set(&sp->so_count, 1);
 	return sp;
 }
 
-void
+static void
 nfs4_drop_state_owner(struct nfs4_state_owner *sp)
 {
-	struct nfs4_client *clp = sp->so_client;
-	spin_lock(&clp->cl_lock);
-	list_del_init(&sp->so_list);
-	spin_unlock(&clp->cl_lock);
+	if (!RB_EMPTY_NODE(&sp->so_client_node)) {
+		struct nfs_client *clp = sp->so_client;
+
+		spin_lock(&clp->cl_lock);
+		rb_erase(&sp->so_client_node, &clp->cl_state_owners);
+		RB_CLEAR_NODE(&sp->so_client_node);
+		spin_unlock(&clp->cl_lock);
+	}
 }
 
 /*
@@ -289,30 +311,30 @@ nfs4_drop_state_owner(struct nfs4_state_owner *sp)
  */
 struct nfs4_state_owner *nfs4_get_state_owner(struct nfs_server *server, struct rpc_cred *cred)
 {
-	struct nfs4_client *clp = server->nfs4_state;
+	struct nfs_client *clp = server->nfs_client;
 	struct nfs4_state_owner *sp, *new;
 
-	get_rpccred(cred);
-	new = nfs4_alloc_state_owner();
 	spin_lock(&clp->cl_lock);
-	sp = nfs4_find_state_owner(clp, cred);
-	if (sp == NULL)
-		sp = nfs4_client_grab_unused(clp, cred);
-	if (sp == NULL && new != NULL) {
-		list_add(&new->so_list, &clp->cl_state_owners);
-		new->so_client = clp;
-		new->so_id = nfs4_alloc_lockowner_id(clp);
-		new->so_cred = cred;
-		sp = new;
-		new = NULL;
-	}
+	sp = nfs4_find_state_owner(server, cred);
 	spin_unlock(&clp->cl_lock);
-	if (new)
-		kfree(new);
 	if (sp != NULL)
 		return sp;
-	put_rpccred(cred);
-	return NULL;
+	new = nfs4_alloc_state_owner();
+	if (new == NULL)
+		return NULL;
+	new->so_client = clp;
+	new->so_server = server;
+	new->so_cred = cred;
+	spin_lock(&clp->cl_lock);
+	sp = nfs4_insert_state_owner(clp, new);
+	spin_unlock(&clp->cl_lock);
+	if (sp == new)
+		get_rpccred(cred);
+	else {
+		rpc_destroy_wait_queue(&new->so_sequence.wait);
+		kfree(new);
+	}
+	return sp;
 }
 
 /*
@@ -321,24 +343,14 @@ struct nfs4_state_owner *nfs4_get_state_owner(struct nfs_server *server, struct
  */
 void nfs4_put_state_owner(struct nfs4_state_owner *sp)
 {
-	struct nfs4_client *clp = sp->so_client;
+	struct nfs_client *clp = sp->so_client;
 	struct rpc_cred *cred = sp->so_cred;
 
 	if (!atomic_dec_and_lock(&sp->so_count, &clp->cl_lock))
 		return;
-	if (clp->cl_nunused >= OPENOWNER_POOL_SIZE)
-		goto out_free;
-	if (list_empty(&sp->so_list))
-		goto out_free;
-	list_move(&sp->so_list, &clp->cl_unused);
-	clp->cl_nunused++;
-	spin_unlock(&clp->cl_lock);
-	put_rpccred(cred);
-	cred = NULL;
-	return;
-out_free:
-	list_del(&sp->so_list);
+	nfs4_remove_state_owner(clp, sp);
 	spin_unlock(&clp->cl_lock);
+	rpc_destroy_wait_queue(&sp->so_sequence.wait);
 	put_rpccred(cred);
 	kfree(sp);
 }
@@ -348,45 +360,29 @@ nfs4_alloc_open_state(void)
 {
 	struct nfs4_state *state;
 
-	state = kmalloc(sizeof(*state), GFP_KERNEL);
+	state = kzalloc(sizeof(*state), GFP_KERNEL);
 	if (!state)
 		return NULL;
-	state->state = 0;
-	state->nreaders = 0;
-	state->nwriters = 0;
-	state->flags = 0;
-	memset(state->stateid.data, 0, sizeof(state->stateid.data));
 	atomic_set(&state->count, 1);
 	INIT_LIST_HEAD(&state->lock_states);
-	init_MUTEX(&state->lock_sema);
-	rwlock_init(&state->state_lock);
+	spin_lock_init(&state->state_lock);
+	seqlock_init(&state->seqlock);
 	return state;
 }
 
-static struct nfs4_state *
-__nfs4_find_state(struct inode *inode, struct rpc_cred *cred, mode_t mode)
+void
+nfs4_state_set_mode_locked(struct nfs4_state *state, mode_t mode)
 {
-	struct nfs_inode *nfsi = NFS_I(inode);
-	struct nfs4_state *state;
-
-	mode &= (FMODE_READ|FMODE_WRITE);
-	list_for_each_entry(state, &nfsi->open_states, inode_states) {
-		if (state->owner->so_cred != cred)
-			continue;
-		if ((mode & FMODE_READ) != 0 && state->nreaders == 0)
-			continue;
-		if ((mode & FMODE_WRITE) != 0 && state->nwriters == 0)
-			continue;
-		if ((state->state & mode) != mode)
-			continue;
-		atomic_inc(&state->count);
-		if (mode & FMODE_READ)
-			state->nreaders++;
+	if (state->state == mode)
+		return;
+	/* NB! List reordering - see the reclaim code for why.  */
+	if ((mode & FMODE_WRITE) != (state->state & FMODE_WRITE)) {
 		if (mode & FMODE_WRITE)
-			state->nwriters++;
-		return state;
+			list_move(&state->open_states, &state->owner->so_states);
+		else
+			list_move_tail(&state->open_states, &state->owner->so_states);
 	}
-	return NULL;
+	state->state = mode;
 }
 
 static struct nfs4_state *
@@ -396,28 +392,14 @@ __nfs4_find_state_byowner(struct inode *inode, struct nfs4_state_owner *owner)
 	struct nfs4_state *state;
 
 	list_for_each_entry(state, &nfsi->open_states, inode_states) {
-		/* Is this in the process of being freed? */
-		if (state->nreaders == 0 && state->nwriters == 0)
+		if (state->owner != owner)
 			continue;
-		if (state->owner == owner) {
-			atomic_inc(&state->count);
+		if (atomic_inc_not_zero(&state->count))
 			return state;
-		}
 	}
 	return NULL;
 }
 
-struct nfs4_state *
-nfs4_find_state(struct inode *inode, struct rpc_cred *cred, mode_t mode)
-{
-	struct nfs4_state *state;
-
-	spin_lock(&inode->i_lock);
-	state = __nfs4_find_state(inode, cred, mode);
-	spin_unlock(&inode->i_lock);
-	return state;
-}
-
 static void
 nfs4_free_open_state(struct nfs4_state *state)
 {
@@ -436,21 +418,23 @@ nfs4_get_open_state(struct inode *inode, struct nfs4_state_owner *owner)
 	if (state)
 		goto out;
 	new = nfs4_alloc_open_state();
+	spin_lock(&owner->so_lock);
 	spin_lock(&inode->i_lock);
 	state = __nfs4_find_state_byowner(inode, owner);
 	if (state == NULL && new != NULL) {
 		state = new;
-		/* Caller *must* be holding owner->so_sem */
-		/* Note: The reclaim code dictates that we add stateless
-		 * and read-only stateids to the end of the list */
-		list_add_tail(&state->open_states, &owner->so_states);
 		state->owner = owner;
 		atomic_inc(&owner->so_count);
 		list_add(&state->inode_states, &nfsi->open_states);
 		state->inode = igrab(inode);
 		spin_unlock(&inode->i_lock);
+		/* Note: The reclaim code dictates that we add stateless
+		 * and read-only stateids to the end of the list */
+		list_add_tail(&state->open_states, &owner->so_states);
+		spin_unlock(&owner->so_lock);
 	} else {
 		spin_unlock(&inode->i_lock);
+		spin_unlock(&owner->so_lock);
 		if (new)
 			nfs4_free_open_state(new);
 	}
@@ -460,68 +444,80 @@ out:
 
 /*
  * Beware! Caller must be holding exactly one
- * reference to clp->cl_sem and owner->so_sema!
+ * reference to clp->cl_sem!
  */
 void nfs4_put_open_state(struct nfs4_state *state)
 {
 	struct inode *inode = state->inode;
 	struct nfs4_state_owner *owner = state->owner;
 
-	if (!atomic_dec_and_lock(&state->count, &inode->i_lock))
+	if (!atomic_dec_and_lock(&state->count, &owner->so_lock))
 		return;
-	if (!list_empty(&state->inode_states))
-		list_del(&state->inode_states);
-	spin_unlock(&inode->i_lock);
+	spin_lock(&inode->i_lock);
+	list_del(&state->inode_states);
 	list_del(&state->open_states);
+	spin_unlock(&inode->i_lock);
+	spin_unlock(&owner->so_lock);
 	iput(inode);
-	BUG_ON (state->state != 0);
 	nfs4_free_open_state(state);
 	nfs4_put_state_owner(owner);
 }
 
 /*
- * Beware! Caller must be holding no references to clp->cl_sem!
- * of owner->so_sema!
+ * Close the current file.
  */
-void nfs4_close_state(struct nfs4_state *state, mode_t mode)
+static void __nfs4_close(struct path *path, struct nfs4_state *state, mode_t mode, int wait)
 {
-	struct inode *inode = state->inode;
 	struct nfs4_state_owner *owner = state->owner;
-	struct nfs4_client *clp = owner->so_client;
+	int call_close = 0;
 	int newstate;
 
 	atomic_inc(&owner->so_count);
-	down_read(&clp->cl_sem);
-	down(&owner->so_sema);
 	/* Protect against nfs4_find_state() */
-	spin_lock(&inode->i_lock);
-	if (mode & FMODE_READ)
-		state->nreaders--;
-	if (mode & FMODE_WRITE)
-		state->nwriters--;
-	if (state->nwriters == 0) {
-		if (state->nreaders == 0)
-			list_del_init(&state->inode_states);
-		/* See reclaim code */
-		list_move_tail(&state->open_states, &owner->so_states);
+	spin_lock(&owner->so_lock);
+	switch (mode & (FMODE_READ | FMODE_WRITE)) {
+		case FMODE_READ:
+			state->n_rdonly--;
+			break;
+		case FMODE_WRITE:
+			state->n_wronly--;
+			break;
+		case FMODE_READ|FMODE_WRITE:
+			state->n_rdwr--;
 	}
-	spin_unlock(&inode->i_lock);
-	newstate = 0;
-	if (state->state != 0) {
-		if (state->nreaders)
-			newstate |= FMODE_READ;
-		if (state->nwriters)
-			newstate |= FMODE_WRITE;
-		if (state->state == newstate)
-			goto out;
-		if (nfs4_do_close(inode, state, newstate) == -EINPROGRESS)
-			return;
+	newstate = FMODE_READ|FMODE_WRITE;
+	if (state->n_rdwr == 0) {
+		if (state->n_rdonly == 0) {
+			newstate &= ~FMODE_READ;
+			call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
+			call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
+		}
+		if (state->n_wronly == 0) {
+			newstate &= ~FMODE_WRITE;
+			call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
+			call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
+		}
+		if (newstate == 0)
+			clear_bit(NFS_DELEGATED_STATE, &state->flags);
 	}
-out:
-	nfs4_put_open_state(state);
-	up(&owner->so_sema);
-	nfs4_put_state_owner(owner);
-	up_read(&clp->cl_sem);
+	nfs4_state_set_mode_locked(state, newstate);
+	spin_unlock(&owner->so_lock);
+
+	if (!call_close) {
+		nfs4_put_open_state(state);
+		nfs4_put_state_owner(owner);
+	} else
+		nfs4_do_close(path, state, wait);
+}
+
+void nfs4_close_state(struct path *path, struct nfs4_state *state, mode_t mode)
+{
+	__nfs4_close(path, state, mode, 0);
+}
+
+void nfs4_close_sync(struct path *path, struct nfs4_state *state, mode_t mode)
+{
+	__nfs4_close(path, state, mode, 1);
 }
 
 /*
@@ -541,234 +537,327 @@ __nfs4_find_lock_state(struct nfs4_state *state, fl_owner_t fl_owner)
 	return NULL;
 }
 
-struct nfs4_lock_state *
-nfs4_find_lock_state(struct nfs4_state *state, fl_owner_t fl_owner)
-{
-	struct nfs4_lock_state *lsp;
-	read_lock(&state->state_lock);
-	lsp = __nfs4_find_lock_state(state, fl_owner);
-	read_unlock(&state->state_lock);
-	return lsp;
-}
-
 /*
  * Return a compatible lock_state. If no initialized lock_state structure
  * exists, return an uninitialized one.
  *
- * The caller must be holding state->lock_sema
  */
 static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, fl_owner_t fl_owner)
 {
 	struct nfs4_lock_state *lsp;
-	struct nfs4_client *clp = state->owner->so_client;
+	struct nfs_client *clp = state->owner->so_client;
 
-	lsp = kmalloc(sizeof(*lsp), GFP_KERNEL);
+	lsp = kzalloc(sizeof(*lsp), GFP_KERNEL);
 	if (lsp == NULL)
 		return NULL;
-	lsp->ls_flags = 0;
-	lsp->ls_seqid = 0;	/* arbitrary */
-	lsp->ls_id = -1; 
-	memset(lsp->ls_stateid.data, 0, sizeof(lsp->ls_stateid.data));
+	rpc_init_wait_queue(&lsp->ls_sequence.wait, "lock_seqid_waitqueue");
+	spin_lock_init(&lsp->ls_sequence.lock);
+	INIT_LIST_HEAD(&lsp->ls_sequence.list);
+	lsp->ls_seqid.sequence = &lsp->ls_sequence;
 	atomic_set(&lsp->ls_count, 1);
 	lsp->ls_owner = fl_owner;
-	INIT_LIST_HEAD(&lsp->ls_locks);
 	spin_lock(&clp->cl_lock);
-	lsp->ls_id = nfs4_alloc_lockowner_id(clp);
+	nfs_alloc_unique_id(&clp->cl_lockowner_id, &lsp->ls_id, 1, 64);
 	spin_unlock(&clp->cl_lock);
+	INIT_LIST_HEAD(&lsp->ls_locks);
 	return lsp;
 }
 
+static void nfs4_free_lock_state(struct nfs4_lock_state *lsp)
+{
+	struct nfs_client *clp = lsp->ls_state->owner->so_client;
+
+	spin_lock(&clp->cl_lock);
+	nfs_free_unique_id(&clp->cl_lockowner_id, &lsp->ls_id);
+	spin_unlock(&clp->cl_lock);
+	rpc_destroy_wait_queue(&lsp->ls_sequence.wait);
+	kfree(lsp);
+}
+
 /*
  * Return a compatible lock_state. If no initialized lock_state structure
  * exists, return an uninitialized one.
  *
- * The caller must be holding state->lock_sema and clp->cl_sem
+ * The caller must be holding clp->cl_sem
  */
-struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_owner_t owner)
+static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_owner_t owner)
 {
-	struct nfs4_lock_state * lsp;
+	struct nfs4_lock_state *lsp, *new = NULL;
 	
-	lsp = nfs4_find_lock_state(state, owner);
-	if (lsp == NULL)
-		lsp = nfs4_alloc_lock_state(state, owner);
+	for(;;) {
+		spin_lock(&state->state_lock);
+		lsp = __nfs4_find_lock_state(state, owner);
+		if (lsp != NULL)
+			break;
+		if (new != NULL) {
+			new->ls_state = state;
+			list_add(&new->ls_locks, &state->lock_states);
+			set_bit(LK_STATE_IN_USE, &state->flags);
+			lsp = new;
+			new = NULL;
+			break;
+		}
+		spin_unlock(&state->state_lock);
+		new = nfs4_alloc_lock_state(state, owner);
+		if (new == NULL)
+			return NULL;
+	}
+	spin_unlock(&state->state_lock);
+	if (new != NULL)
+		nfs4_free_lock_state(new);
 	return lsp;
 }
 
 /*
- * Byte-range lock aware utility to initialize the stateid of read/write
- * requests.
+ * Release reference to lock_state, and free it if we see that
+ * it is no longer in use
  */
-void
-nfs4_copy_stateid(nfs4_stateid *dst, struct nfs4_state *state, fl_owner_t fl_owner)
+void nfs4_put_lock_state(struct nfs4_lock_state *lsp)
 {
-	if (test_bit(LK_STATE_IN_USE, &state->flags)) {
-		struct nfs4_lock_state *lsp;
+	struct nfs4_state *state;
 
-		lsp = nfs4_find_lock_state(state, fl_owner);
-		if (lsp) {
-			memcpy(dst, &lsp->ls_stateid, sizeof(*dst));
-			nfs4_put_lock_state(lsp);
-			return;
-		}
-	}
-	memcpy(dst, &state->stateid, sizeof(*dst));
+	if (lsp == NULL)
+		return;
+	state = lsp->ls_state;
+	if (!atomic_dec_and_lock(&lsp->ls_count, &state->state_lock))
+		return;
+	list_del(&lsp->ls_locks);
+	if (list_empty(&state->lock_states))
+		clear_bit(LK_STATE_IN_USE, &state->flags);
+	spin_unlock(&state->state_lock);
+	nfs4_free_lock_state(lsp);
 }
 
-/*
-* Called with state->lock_sema and clp->cl_sem held.
-*/
-void nfs4_increment_lock_seqid(int status, struct nfs4_lock_state *lsp)
+static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src)
+{
+	struct nfs4_lock_state *lsp = src->fl_u.nfs4_fl.owner;
+
+	dst->fl_u.nfs4_fl.owner = lsp;
+	atomic_inc(&lsp->ls_count);
+}
+
+static void nfs4_fl_release_lock(struct file_lock *fl)
 {
-	if (status == NFS_OK || seqid_mutating_err(-status))
-		lsp->ls_seqid++;
+	nfs4_put_lock_state(fl->fl_u.nfs4_fl.owner);
 }
 
-/* 
-* Check to see if the request lock (type FL_UNLK) effects the fl lock.
-*
-* fl and request must have the same posix owner
-*
-* return: 
-* 0 -> fl not effected by request
-* 1 -> fl consumed by request
-*/
+static struct file_lock_operations nfs4_fl_lock_ops = {
+	.fl_copy_lock = nfs4_fl_copy_lock,
+	.fl_release_private = nfs4_fl_release_lock,
+};
 
-static int
-nfs4_check_unlock(struct file_lock *fl, struct file_lock *request)
+int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl)
 {
-	if (fl->fl_start >= request->fl_start && fl->fl_end <= request->fl_end)
-		return 1;
+	struct nfs4_lock_state *lsp;
+
+	if (fl->fl_ops != NULL)
+		return 0;
+	lsp = nfs4_get_lock_state(state, fl->fl_owner);
+	if (lsp == NULL)
+		return -ENOMEM;
+	fl->fl_u.nfs4_fl.owner = lsp;
+	fl->fl_ops = &nfs4_fl_lock_ops;
 	return 0;
 }
 
 /*
- * Post an initialized lock_state on the state->lock_states list.
+ * Byte-range lock aware utility to initialize the stateid of read/write
+ * requests.
  */
-void nfs4_notify_setlk(struct nfs4_state *state, struct file_lock *request, struct nfs4_lock_state *lsp)
+void nfs4_copy_stateid(nfs4_stateid *dst, struct nfs4_state *state, fl_owner_t fl_owner)
 {
-	if (!list_empty(&lsp->ls_locks))
+	struct nfs4_lock_state *lsp;
+	int seq;
+
+	do {
+		seq = read_seqbegin(&state->seqlock);
+		memcpy(dst, &state->stateid, sizeof(*dst));
+	} while (read_seqretry(&state->seqlock, seq));
+	if (test_bit(LK_STATE_IN_USE, &state->flags) == 0)
 		return;
-	atomic_inc(&lsp->ls_count);
-	write_lock(&state->state_lock);
-	list_add(&lsp->ls_locks, &state->lock_states);
-	set_bit(LK_STATE_IN_USE, &state->flags);
-	write_unlock(&state->state_lock);
+
+	spin_lock(&state->state_lock);
+	lsp = __nfs4_find_lock_state(state, fl_owner);
+	if (lsp != NULL && (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
+		memcpy(dst, &lsp->ls_stateid, sizeof(*dst));
+	spin_unlock(&state->state_lock);
+	nfs4_put_lock_state(lsp);
 }
 
-/* 
- * to decide to 'reap' lock state:
- * 1) search i_flock for file_locks with fl.lock_state = to ls.
- * 2) determine if unlock will consume found lock. 
- * 	if so, reap
- *
- * 	else, don't reap.
- *
- */
-void
-nfs4_notify_unlck(struct nfs4_state *state, struct file_lock *request, struct nfs4_lock_state *lsp)
+struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter)
 {
-	struct inode *inode = state->inode;
-	struct file_lock *fl;
+	struct nfs_seqid *new;
 
-	for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
-		if (!(fl->fl_flags & FL_POSIX))
-			continue;
-		if (fl->fl_owner != lsp->ls_owner)
-			continue;
-		/* Exit if we find at least one lock which is not consumed */
-		if (nfs4_check_unlock(fl,request) == 0)
-			return;
+	new = kmalloc(sizeof(*new), GFP_KERNEL);
+	if (new != NULL) {
+		new->sequence = counter;
+		INIT_LIST_HEAD(&new->list);
 	}
+	return new;
+}
 
-	write_lock(&state->state_lock);
-	list_del_init(&lsp->ls_locks);
-	if (list_empty(&state->lock_states))
-		clear_bit(LK_STATE_IN_USE, &state->flags);
-	write_unlock(&state->state_lock);
-	nfs4_put_lock_state(lsp);
+void nfs_free_seqid(struct nfs_seqid *seqid)
+{
+	if (!list_empty(&seqid->list)) {
+		struct rpc_sequence *sequence = seqid->sequence->sequence;
+
+		spin_lock(&sequence->lock);
+		list_del(&seqid->list);
+		spin_unlock(&sequence->lock);
+		rpc_wake_up(&sequence->wait);
+	}
+	kfree(seqid);
 }
 
 /*
- * Release reference to lock_state, and free it if we see that
- * it is no longer in use
+ * Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or
+ * failed with a seqid incrementing error -
+ * see comments nfs_fs.h:seqid_mutating_error()
  */
-void
-nfs4_put_lock_state(struct nfs4_lock_state *lsp)
+static void nfs_increment_seqid(int status, struct nfs_seqid *seqid)
 {
-	if (!atomic_dec_and_test(&lsp->ls_count))
-		return;
-	BUG_ON (!list_empty(&lsp->ls_locks));
-	kfree(lsp);
+	BUG_ON(list_first_entry(&seqid->sequence->sequence->list, struct nfs_seqid, list) != seqid);
+	switch (status) {
+		case 0:
+			break;
+		case -NFS4ERR_BAD_SEQID:
+			if (seqid->sequence->flags & NFS_SEQID_CONFIRMED)
+				return;
+			printk(KERN_WARNING "NFS: v4 server returned a bad"
+					" sequence-id error on an"
+					" unconfirmed sequence %p!\n",
+					seqid->sequence);
+		case -NFS4ERR_STALE_CLIENTID:
+		case -NFS4ERR_STALE_STATEID:
+		case -NFS4ERR_BAD_STATEID:
+		case -NFS4ERR_BADXDR:
+		case -NFS4ERR_RESOURCE:
+		case -NFS4ERR_NOFILEHANDLE:
+			/* Non-seqid mutating errors */
+			return;
+	};
+	/*
+	 * Note: no locking needed as we are guaranteed to be first
+	 * on the sequence list
+	 */
+	seqid->sequence->counter++;
 }
 
-/*
-* Called with sp->so_sema and clp->cl_sem held.
-*
-* Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or
-* failed with a seqid incrementing error -
-* see comments nfs_fs.h:seqid_mutating_error()
-*/
-void nfs4_increment_seqid(int status, struct nfs4_state_owner *sp)
-{
-	if (status == NFS_OK || seqid_mutating_err(-status))
-		sp->so_seqid++;
-	/* If the server returns BAD_SEQID, unhash state_owner here */
-	if (status == -NFS4ERR_BAD_SEQID)
+void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid)
+{
+	if (status == -NFS4ERR_BAD_SEQID) {
+		struct nfs4_state_owner *sp = container_of(seqid->sequence,
+				struct nfs4_state_owner, so_seqid);
 		nfs4_drop_state_owner(sp);
+	}
+	nfs_increment_seqid(status, seqid);
+}
+
+/*
+ * Increment the seqid if the LOCK/LOCKU succeeded, or
+ * failed with a seqid incrementing error -
+ * see comments nfs_fs.h:seqid_mutating_error()
+ */
+void nfs_increment_lock_seqid(int status, struct nfs_seqid *seqid)
+{
+	nfs_increment_seqid(status, seqid);
+}
+
+int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task)
+{
+	struct rpc_sequence *sequence = seqid->sequence->sequence;
+	int status = 0;
+
+	spin_lock(&sequence->lock);
+	if (list_empty(&seqid->list))
+		list_add_tail(&seqid->list, &sequence->list);
+	if (list_first_entry(&sequence->list, struct nfs_seqid, list) == seqid)
+		goto unlock;
+	rpc_sleep_on(&sequence->wait, task, NULL);
+	status = -EAGAIN;
+unlock:
+	spin_unlock(&sequence->lock);
+	return status;
 }
 
 static int reclaimer(void *);
-struct reclaimer_args {
-	struct nfs4_client *clp;
-	struct completion complete;
-};
+
+static inline void nfs4_clear_recover_bit(struct nfs_client *clp)
+{
+	smp_mb__before_clear_bit();
+	clear_bit(NFS4CLNT_STATE_RECOVER, &clp->cl_state);
+	smp_mb__after_clear_bit();
+	wake_up_bit(&clp->cl_state, NFS4CLNT_STATE_RECOVER);
+	rpc_wake_up(&clp->cl_rpcwaitq);
+}
 
 /*
  * State recovery routine
  */
-void
-nfs4_recover_state(void *data)
+static void nfs4_recover_state(struct nfs_client *clp)
 {
-	struct nfs4_client *clp = (struct nfs4_client *)data;
-	struct reclaimer_args args = {
-		.clp = clp,
-	};
-	might_sleep();
-
-	init_completion(&args.complete);
+	struct task_struct *task;
 
-	if (kernel_thread(reclaimer, &args, CLONE_KERNEL) < 0)
-		goto out_failed_clear;
-	wait_for_completion(&args.complete);
-	return;
-out_failed_clear:
-	set_bit(NFS4CLNT_OK, &clp->cl_state);
-	wake_up_all(&clp->cl_waitq);
-	rpc_wake_up(&clp->cl_rpcwaitq);
+	__module_get(THIS_MODULE);
+	atomic_inc(&clp->cl_count);
+	task = kthread_run(reclaimer, clp, "%s-reclaim",
+				rpc_peeraddr2str(clp->cl_rpcclient,
+							RPC_DISPLAY_ADDR));
+	if (!IS_ERR(task))
+		return;
+	nfs4_clear_recover_bit(clp);
+	nfs_put_client(clp);
+	module_put(THIS_MODULE);
 }
 
 /*
  * Schedule a state recovery attempt
  */
-void
-nfs4_schedule_state_recovery(struct nfs4_client *clp)
+void nfs4_schedule_state_recovery(struct nfs_client *clp)
 {
 	if (!clp)
 		return;
-	if (test_and_clear_bit(NFS4CLNT_OK, &clp->cl_state))
-		schedule_work(&clp->cl_recoverd);
+	if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
+		set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
+	if (test_and_set_bit(NFS4CLNT_STATE_RECOVER, &clp->cl_state) == 0)
+		nfs4_recover_state(clp);
+}
+
+static int nfs4_state_mark_reclaim_reboot(struct nfs_client *clp, struct nfs4_state *state)
+{
+
+	set_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
+	/* Don't recover state that expired before the reboot */
+	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags)) {
+		clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
+		return 0;
+	}
+	set_bit(NFS_OWNER_RECLAIM_REBOOT, &state->owner->so_flags);
+	set_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state);
+	return 1;
+}
+
+static int nfs4_state_mark_reclaim_nograce(struct nfs_client *clp, struct nfs4_state *state)
+{
+	set_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags);
+	clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags);
+	set_bit(NFS_OWNER_RECLAIM_NOGRACE, &state->owner->so_flags);
+	set_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
+	return 1;
 }
 
-static int nfs4_reclaim_locks(struct nfs4_state_recovery_ops *ops, struct nfs4_state *state)
+static int nfs4_reclaim_locks(struct nfs4_state *state, const struct nfs4_state_recovery_ops *ops)
 {
 	struct inode *inode = state->inode;
+	struct nfs_inode *nfsi = NFS_I(inode);
 	struct file_lock *fl;
 	int status = 0;
 
-	for (fl = inode->i_flock; fl != 0; fl = fl->fl_next) {
-		if (!(fl->fl_flags & FL_POSIX))
+	down_write(&nfsi->rwsem);
+	for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
+		if (!(fl->fl_flags & (FL_POSIX|FL_FLOCK)))
 			continue;
-		if (((struct nfs_open_context *)fl->fl_file->private_data)->state != state)
+		if (nfs_file_open_context(fl->fl_file)->state != state)
 			continue;
 		status = ops->recover_lock(state, fl);
 		if (status >= 0)
@@ -776,23 +865,25 @@ static int nfs4_reclaim_locks(struct nfs4_state_recovery_ops *ops, struct nfs4_s
 		switch (status) {
 			default:
 				printk(KERN_ERR "%s: unhandled error %d. Zeroing state\n",
-						__FUNCTION__, status);
+						__func__, status);
 			case -NFS4ERR_EXPIRED:
 			case -NFS4ERR_NO_GRACE:
 			case -NFS4ERR_RECLAIM_BAD:
 			case -NFS4ERR_RECLAIM_CONFLICT:
-				/* kill_proc(fl->fl_owner, SIGLOST, 1); */
+				/* kill_proc(fl->fl_pid, SIGLOST, 1); */
 				break;
 			case -NFS4ERR_STALE_CLIENTID:
 				goto out_err;
 		}
 	}
+	up_write(&nfsi->rwsem);
 	return 0;
 out_err:
+	up_write(&nfsi->rwsem);
 	return status;
 }
 
-static int nfs4_reclaim_open_state(struct nfs4_state_recovery_ops *ops, struct nfs4_state_owner *sp)
+static int nfs4_reclaim_open_state(struct nfs4_state_owner *sp, const struct nfs4_state_recovery_ops *ops)
 {
 	struct nfs4_state *state;
 	struct nfs4_lock_state *lock;
@@ -806,30 +897,34 @@ static int nfs4_reclaim_open_state(struct nfs4_state_recovery_ops *ops, struct n
 	 * recovering after a network partition or a reboot from a
 	 * server that doesn't support a grace period.
 	 */
+restart:
+	spin_lock(&sp->so_lock);
 	list_for_each_entry(state, &sp->so_states, open_states) {
+		if (!test_and_clear_bit(ops->state_flag_bit, &state->flags))
+			continue;
 		if (state->state == 0)
 			continue;
+		atomic_inc(&state->count);
+		spin_unlock(&sp->so_lock);
 		status = ops->recover_open(sp, state);
-		list_for_each_entry(lock, &state->lock_states, ls_locks)
-			lock->ls_flags &= ~NFS_LOCK_INITIALIZED;
 		if (status >= 0) {
-			status = nfs4_reclaim_locks(ops, state);
-			if (status < 0)
-				goto out_err;
-			list_for_each_entry(lock, &state->lock_states, ls_locks) {
-				if (!(lock->ls_flags & NFS_LOCK_INITIALIZED))
-					printk("%s: Lock reclaim failed!\n",
-							__FUNCTION__);
+			status = nfs4_reclaim_locks(state, ops);
+			if (status >= 0) {
+				list_for_each_entry(lock, &state->lock_states, ls_locks) {
+					if (!(lock->ls_flags & NFS_LOCK_INITIALIZED))
+						printk("%s: Lock reclaim failed!\n",
+							__func__);
+				}
+				nfs4_put_open_state(state);
+				goto restart;
 			}
-			continue;
 		}
 		switch (status) {
 			default:
 				printk(KERN_ERR "%s: unhandled error %d. Zeroing state\n",
-						__FUNCTION__, status);
+						__func__, status);
 			case -ENOENT:
-			case -NFS4ERR_RECLAIM_BAD:
-			case -NFS4ERR_RECLAIM_CONFLICT:
+			case -ESTALE:
 				/*
 				 * Open state on this file cannot be recovered
 				 * All we can do is revert to using the zero stateid.
@@ -839,83 +934,259 @@ static int nfs4_reclaim_open_state(struct nfs4_state_recovery_ops *ops, struct n
 				/* Mark the file as being 'closed' */
 				state->state = 0;
 				break;
+			case -NFS4ERR_RECLAIM_BAD:
+			case -NFS4ERR_RECLAIM_CONFLICT:
+				nfs4_state_mark_reclaim_nograce(sp->so_client, state);
+				break;
 			case -NFS4ERR_EXPIRED:
 			case -NFS4ERR_NO_GRACE:
+				nfs4_state_mark_reclaim_nograce(sp->so_client, state);
 			case -NFS4ERR_STALE_CLIENTID:
 				goto out_err;
 		}
+		nfs4_put_open_state(state);
+		goto restart;
 	}
+	spin_unlock(&sp->so_lock);
 	return 0;
 out_err:
+	nfs4_put_open_state(state);
 	return status;
 }
 
-static int reclaimer(void *ptr)
+static void nfs4_clear_open_state(struct nfs4_state *state)
+{
+	struct nfs4_lock_state *lock;
+
+	clear_bit(NFS_DELEGATED_STATE, &state->flags);
+	clear_bit(NFS_O_RDONLY_STATE, &state->flags);
+	clear_bit(NFS_O_WRONLY_STATE, &state->flags);
+	clear_bit(NFS_O_RDWR_STATE, &state->flags);
+	list_for_each_entry(lock, &state->lock_states, ls_locks) {
+		lock->ls_seqid.counter = 0;
+		lock->ls_seqid.flags = 0;
+		lock->ls_flags &= ~NFS_LOCK_INITIALIZED;
+	}
+}
+
+static void nfs4_state_mark_reclaim_helper(struct nfs_client *clp, int (*mark_reclaim)(struct nfs_client *clp, struct nfs4_state *state))
 {
-	struct reclaimer_args *args = (struct reclaimer_args *)ptr;
-	struct nfs4_client *clp = args->clp;
 	struct nfs4_state_owner *sp;
-	struct nfs4_state_recovery_ops *ops;
-	int status = 0;
+	struct rb_node *pos;
+	struct nfs4_state *state;
 
-	daemonize("%u.%u.%u.%u-reclaim", NIPQUAD(clp->cl_addr));
-	allow_signal(SIGKILL);
+	/* Reset all sequence ids to zero */
+	for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) {
+		sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
+		sp->so_seqid.counter = 0;
+		sp->so_seqid.flags = 0;
+		spin_lock(&sp->so_lock);
+		list_for_each_entry(state, &sp->so_states, open_states) {
+			if (mark_reclaim(clp, state))
+				nfs4_clear_open_state(state);
+		}
+		spin_unlock(&sp->so_lock);
+	}
+}
 
-	atomic_inc(&clp->cl_count);
-	complete(&args->complete);
+static void nfs4_state_start_reclaim_reboot(struct nfs_client *clp)
+{
+	/* Mark all delegations for reclaim */
+	nfs_delegation_mark_reclaim(clp);
+	nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_reboot);
+}
 
-	/* Ensure exclusive access to NFSv4 state */
-	lock_kernel();
-	down_write(&clp->cl_sem);
-	/* Are there any NFS mounts out there? */
-	if (list_empty(&clp->cl_superblocks))
-		goto out;
-restart_loop:
-	status = nfs4_proc_renew(clp);
-	switch (status) {
-		case 0:
+static void nfs4_state_end_reclaim_reboot(struct nfs_client *clp)
+{
+	struct nfs4_state_owner *sp;
+	struct rb_node *pos;
+	struct nfs4_state *state;
+
+	if (!test_and_clear_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
+		return;
+
+	for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) {
+		sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
+		spin_lock(&sp->so_lock);
+		list_for_each_entry(state, &sp->so_states, open_states) {
+			if (!test_and_clear_bit(NFS_STATE_RECLAIM_REBOOT, &state->flags))
+				continue;
+			nfs4_state_mark_reclaim_nograce(clp, state);
+		}
+		spin_unlock(&sp->so_lock);
+	}
+
+	nfs_delegation_reap_unclaimed(clp);
+}
+
+static void nfs_delegation_clear_all(struct nfs_client *clp)
+{
+	nfs_delegation_mark_reclaim(clp);
+	nfs_delegation_reap_unclaimed(clp);
+}
+
+static void nfs4_state_start_reclaim_nograce(struct nfs_client *clp)
+{
+	nfs_delegation_clear_all(clp);
+	nfs4_state_mark_reclaim_helper(clp, nfs4_state_mark_reclaim_nograce);
+}
+
+static void nfs4_state_end_reclaim_nograce(struct nfs_client *clp)
+{
+	clear_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
+}
+
+static void nfs4_recovery_handle_error(struct nfs_client *clp, int error)
+{
+	switch (error) {
 		case -NFS4ERR_CB_PATH_DOWN:
-			goto out;
+			set_bit(NFS4CLNT_CB_PATH_DOWN, &clp->cl_state);
+			break;
 		case -NFS4ERR_STALE_CLIENTID:
 		case -NFS4ERR_LEASE_MOVED:
-			ops = &nfs4_reboot_recovery_ops;
+			set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
+			nfs4_state_start_reclaim_reboot(clp);
 			break;
-		default:
-			ops = &nfs4_network_partition_recovery_ops;
-	};
-	status = __nfs4_init_client(clp);
-	if (status)
-		goto out_error;
-	/* Mark all delegations for reclaim */
-	nfs_delegation_mark_reclaim(clp);
-	/* Note: list is protected by exclusive lock on cl->cl_sem */
-	list_for_each_entry(sp, &clp->cl_state_owners, so_list) {
-		status = nfs4_reclaim_open_state(ops, sp);
+		case -NFS4ERR_EXPIRED:
+			set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
+			nfs4_state_start_reclaim_nograce(clp);
+	}
+}
+
+static int nfs4_do_reclaim(struct nfs_client *clp, const struct nfs4_state_recovery_ops *ops)
+{
+	struct rb_node *pos;
+	int status = 0;
+
+restart:
+	spin_lock(&clp->cl_lock);
+	for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) {
+		struct nfs4_state_owner *sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
+		if (!test_and_clear_bit(ops->owner_flag_bit, &sp->so_flags))
+			continue;
+		atomic_inc(&sp->so_count);
+		spin_unlock(&clp->cl_lock);
+		status = nfs4_reclaim_open_state(sp, ops);
 		if (status < 0) {
-			if (status == -NFS4ERR_NO_GRACE) {
-				ops = &nfs4_network_partition_recovery_ops;
-				status = nfs4_reclaim_open_state(ops, sp);
+			set_bit(ops->owner_flag_bit, &sp->so_flags);
+			nfs4_put_state_owner(sp);
+			nfs4_recovery_handle_error(clp, status);
+			return status;
+		}
+		nfs4_put_state_owner(sp);
+		goto restart;
+	}
+	spin_unlock(&clp->cl_lock);
+	return status;
+}
+
+static int nfs4_check_lease(struct nfs_client *clp)
+{
+	struct rpc_cred *cred;
+	int status = -NFS4ERR_EXPIRED;
+
+	/* Is the client already known to have an expired lease? */
+	if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state))
+		return 0;
+	cred = nfs4_get_renew_cred(clp);
+	if (cred == NULL) {
+		cred = nfs4_get_setclientid_cred(clp);
+		if (cred == NULL)
+			goto out;
+	}
+	status = nfs4_proc_renew(clp, cred);
+	put_rpccred(cred);
+out:
+	nfs4_recovery_handle_error(clp, status);
+	return status;
+}
+
+static int nfs4_reclaim_lease(struct nfs_client *clp)
+{
+	struct rpc_cred *cred;
+	int status = -ENOENT;
+
+	cred = nfs4_get_setclientid_cred(clp);
+	if (cred != NULL) {
+		status = nfs4_init_client(clp, cred);
+		put_rpccred(cred);
+		/* Handle case where the user hasn't set up machine creds */
+		if (status == -EACCES && cred == clp->cl_machine_cred) {
+			nfs4_clear_machine_cred(clp);
+			status = -EAGAIN;
+		}
+	}
+	return status;
+}
+
+static int reclaimer(void *ptr)
+{
+	struct nfs_client *clp = ptr;
+	int status = 0;
+
+	allow_signal(SIGKILL);
+
+	/* Ensure exclusive access to NFSv4 state */
+	down_write(&clp->cl_sem);
+	while (!list_empty(&clp->cl_superblocks)) {
+		if (test_and_clear_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state)) {
+			/* We're going to have to re-establish a clientid */
+			status = nfs4_reclaim_lease(clp);
+			if (status) {
+				set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
+				if (status == -EAGAIN)
+					continue;
+				goto out_error;
 			}
+			clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state);
+		}
+
+		if (test_and_clear_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state)) {
+			status = nfs4_check_lease(clp);
+			if (status != 0)
+				continue;
+		}
+
+		/* First recover reboot state... */
+		if (test_and_clear_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state)) {
+			/* Note: list is protected by exclusive lock on cl->cl_sem */
+			status = nfs4_do_reclaim(clp, &nfs4_reboot_recovery_ops);
 			if (status == -NFS4ERR_STALE_CLIENTID)
-				goto restart_loop;
-			if (status == -NFS4ERR_EXPIRED)
-				goto restart_loop;
+				continue;
+			nfs4_state_end_reclaim_reboot(clp);
+			continue;
 		}
+
+		/* Now recover expired state... */
+		if (test_and_clear_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state)) {
+			/* Note: list is protected by exclusive lock on cl->cl_sem */
+			status = nfs4_do_reclaim(clp, &nfs4_nograce_recovery_ops);
+			if (status < 0) {
+				set_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state);
+				if (status == -NFS4ERR_STALE_CLIENTID)
+					continue;
+				if (status == -NFS4ERR_EXPIRED)
+					continue;
+				goto out_error;
+			} else
+				nfs4_state_end_reclaim_nograce(clp);
+			continue;
+		}
+		break;
 	}
-	nfs_delegation_reap_unclaimed(clp);
 out:
-	set_bit(NFS4CLNT_OK, &clp->cl_state);
 	up_write(&clp->cl_sem);
-	unlock_kernel();
-	wake_up_all(&clp->cl_waitq);
-	rpc_wake_up(&clp->cl_rpcwaitq);
-	if (status == -NFS4ERR_CB_PATH_DOWN)
+	if (test_and_clear_bit(NFS4CLNT_CB_PATH_DOWN, &clp->cl_state))
 		nfs_handle_cb_pathdown(clp);
-	nfs4_put_client(clp);
+	nfs4_clear_recover_bit(clp);
+	nfs_put_client(clp);
+	module_put_and_exit(0);
 	return 0;
 out_error:
-	printk(KERN_WARNING "Error: state recovery failed on NFSv4 server %u.%u.%u.%u with error %d\n",
-				NIPQUAD(clp->cl_addr.s_addr), -status);
+	printk(KERN_WARNING "Error: state recovery failed on NFSv4 server %s"
+			" with error %d\n", clp->cl_hostname, -status);
+	if (test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state))
+		nfs4_state_end_reclaim_reboot(clp);
 	goto out;
 }