Merge git://git.linux-nfs.org/pub/linux/nfs-2.6

* git://git.linux-nfs.org/pub/linux/nfs-2.6: (103 commits)
  SUNRPC,RPCSEC_GSS: spkm3--fix config dependencies
  SUNRPC,RPCSEC_GSS: spkm3: import contexts using NID_cast5_cbc
  LOCKD: Make nlmsvc_traverse_shares return void
  LOCKD: nlmsvc_traverse_blocks return is unused
  SUNRPC,RPCSEC_GSS: fix krb5 sequence numbers.
  NFSv4: Dont list system.nfs4_acl for filesystems that don't support it.
  SUNRPC,RPCSEC_GSS: remove unnecessary kmalloc of a checksum
  SUNRPC: Ensure rpc_call_async() always calls tk_ops->rpc_release()
  SUNRPC: Fix memory barriers for req->rq_received
  NFS: Fix a race in nfs_sync_inode()
  NFS: Clean up nfs_flush_list()
  NFS: Fix a race with PG_private and nfs_release_page()
  NFSv4: Ensure the callback daemon flushes signals
  SUNRPC: Fix a 'Busy inodes' error in rpc_pipefs
  NFS, NLM: Allow blocking locks to respect signals
  NFS: Make nfs_fhget() return appropriate error values
  NFSv4: Fix an oops in nfs4_fill_super
  lockd: blocks should hold a reference to the nlm_file
  NFSv4: SETCLIENTID_CONFIRM should handle NFS4ERR_DELAY/NFS4ERR_RESOURCE
  NFSv4: Send the delegation stateid for SETATTR calls
  ...

fs/Kconfig

@@ -1555,6 +1555,7 @@ config RPCSEC_GSS_SPKM3
 	select CRYPTO
 	select CRYPTO_MD5
 	select CRYPTO_DES
+	select CRYPTO_CAST5
 	help
 	  Provides for secure RPC calls by means of a gss-api
 	  mechanism based on the SPKM3 public-key mechanism.

fs/lockd/clntlock.c

@@ -44,32 +44,25 @@ static LIST_HEAD(nlm_blocked);
 /*
  * Queue up a lock for blocking so that the GRANTED request can see it
  */
-int nlmclnt_prepare_block(struct nlm_rqst *req, struct nlm_host *host, struct file_lock *fl)
+struct nlm_wait *nlmclnt_prepare_block(struct nlm_host *host, struct file_lock *fl)
 {
 	struct nlm_wait *block;
 
-	BUG_ON(req->a_block != NULL);
 	block = kmalloc(sizeof(*block), GFP_KERNEL);
-	if (block == NULL)
-		return -ENOMEM;
-	block->b_host = host;
-	block->b_lock = fl;
-	init_waitqueue_head(&block->b_wait);
-	block->b_status = NLM_LCK_BLOCKED;
-
-	list_add(&block->b_list, &nlm_blocked);
-	req->a_block = block;
-
-	return 0;
+	if (block != NULL) {
+		block->b_host = host;
+		block->b_lock = fl;
+		init_waitqueue_head(&block->b_wait);
+		block->b_status = NLM_LCK_BLOCKED;
+		list_add(&block->b_list, &nlm_blocked);
+	}
+	return block;
 }
 
-void nlmclnt_finish_block(struct nlm_rqst *req)
+void nlmclnt_finish_block(struct nlm_wait *block)
 {
-	struct nlm_wait *block = req->a_block;
-
 	if (block == NULL)
 		return;
-	req->a_block = NULL;
 	list_del(&block->b_list);
 	kfree(block);
 }
@@ -77,15 +70,14 @@ void nlmclnt_finish_block(struct nlm_rqst *req)
 /*
  * Block on a lock
  */
-long nlmclnt_block(struct nlm_rqst *req, long timeout)
+int nlmclnt_block(struct nlm_wait *block, struct nlm_rqst *req, long timeout)
 {
-	struct nlm_wait	*block = req->a_block;
 	long ret;
 
 	/* A borken server might ask us to block even if we didn't
 	 * request it. Just say no!
 	 */
-	if (!req->a_args.block)
+	if (block == NULL)
 		return -EAGAIN;
 
 	/* Go to sleep waiting for GRANT callback. Some servers seem
@@ -99,13 +91,10 @@ long nlmclnt_block(struct nlm_rqst *req, long timeout)
 	ret = wait_event_interruptible_timeout(block->b_wait,
 			block->b_status != NLM_LCK_BLOCKED,
 			timeout);
-
-	if (block->b_status != NLM_LCK_BLOCKED) {
-		req->a_res.status = block->b_status;
-		block->b_status = NLM_LCK_BLOCKED;
-	}
-
-	return ret;
+	if (ret < 0)
+		return -ERESTARTSYS;
+	req->a_res.status = block->b_status;
+	return 0;
 }
 
 /*
@@ -125,7 +114,15 @@ u32 nlmclnt_grant(const struct sockaddr_in *addr, const struct nlm_lock *lock)
 	list_for_each_entry(block, &nlm_blocked, b_list) {
 		struct file_lock *fl_blocked = block->b_lock;
 
-		if (!nlm_compare_locks(fl_blocked, fl))
+		if (fl_blocked->fl_start != fl->fl_start)
+			continue;
+		if (fl_blocked->fl_end != fl->fl_end)
+			continue;
+		/*
+		 * Careful! The NLM server will return the 32-bit "pid" that
+		 * we put on the wire: in this case the lockowner "pid".
+		 */
+		if (fl_blocked->fl_u.nfs_fl.owner->pid != lock->svid)
 			continue;
 		if (!nlm_cmp_addr(&block->b_host->h_addr, addr))
 			continue;
@@ -146,34 +143,6 @@ u32 nlmclnt_grant(const struct sockaddr_in *addr, const struct nlm_lock *lock)
  * server crash.
  */
 
-/*
- * Mark the locks for reclaiming.
- * FIXME: In 2.5 we don't want to iterate through any global file_lock_list.
- *        Maintain NLM lock reclaiming lists in the nlm_host instead.
- */
-static
-void nlmclnt_mark_reclaim(struct nlm_host *host)
-{
-	struct file_lock *fl;
-	struct inode *inode;
-	struct list_head *tmp;
-
-	list_for_each(tmp, &file_lock_list) {
-		fl = list_entry(tmp, struct file_lock, fl_link);
-
-		inode = fl->fl_file->f_dentry->d_inode;
-		if (inode->i_sb->s_magic != NFS_SUPER_MAGIC)
-			continue;
-		if (fl->fl_u.nfs_fl.owner == NULL)
-			continue;
-		if (fl->fl_u.nfs_fl.owner->host != host)
-			continue;
-		if (!(fl->fl_u.nfs_fl.flags & NFS_LCK_GRANTED))
-			continue;
-		fl->fl_u.nfs_fl.flags |= NFS_LCK_RECLAIM;
-	}
-}
-
 /*
  * Someone has sent us an SM_NOTIFY. Ensure we bind to the new port number,
  * that we mark locks for reclaiming, and that we bump the pseudo NSM state.
@@ -186,7 +155,12 @@ void nlmclnt_prepare_reclaim(struct nlm_host *host, u32 newstate)
 	host->h_state++;
 	host->h_nextrebind = 0;
 	nlm_rebind_host(host);
-	nlmclnt_mark_reclaim(host);
+
+	/*
+	 * Mark the locks for reclaiming.
+	 */
+	list_splice_init(&host->h_granted, &host->h_reclaim);
+
 	dprintk("NLM: reclaiming locks for host %s", host->h_name);
 }
 
@@ -215,9 +189,7 @@ reclaimer(void *ptr)
 {
 	struct nlm_host	  *host = (struct nlm_host *) ptr;
 	struct nlm_wait	  *block;
-	struct list_head *tmp;
-	struct file_lock *fl;
-	struct inode *inode;
+	struct file_lock *fl, *next;
 
 	daemonize("%s-reclaim", host->h_name);
 	allow_signal(SIGKILL);
@@ -229,23 +201,13 @@ reclaimer(void *ptr)
 
 	/* First, reclaim all locks that have been marked. */
 restart:
-	list_for_each(tmp, &file_lock_list) {
-		fl = list_entry(tmp, struct file_lock, fl_link);
+	list_for_each_entry_safe(fl, next, &host->h_reclaim, fl_u.nfs_fl.list) {
+		list_del_init(&fl->fl_u.nfs_fl.list);
 
-		inode = fl->fl_file->f_dentry->d_inode;
-		if (inode->i_sb->s_magic != NFS_SUPER_MAGIC)
-			continue;
-		if (fl->fl_u.nfs_fl.owner == NULL)
-			continue;
-		if (fl->fl_u.nfs_fl.owner->host != host)
-			continue;
-		if (!(fl->fl_u.nfs_fl.flags & NFS_LCK_RECLAIM))
-			continue;
-
-		fl->fl_u.nfs_fl.flags &= ~NFS_LCK_RECLAIM;
-		nlmclnt_reclaim(host, fl);
 		if (signalled())
-			break;
+			continue;
+		if (nlmclnt_reclaim(host, fl) == 0)
+			list_add_tail(&fl->fl_u.nfs_fl.list, &host->h_granted);
 		goto restart;
 	}
 

fs/lockd/clntproc.c

@@ -132,59 +132,18 @@ static void nlmclnt_setlockargs(struct nlm_rqst *req, struct file_lock *fl)
 	memcpy(&lock->fh, NFS_FH(fl->fl_file->f_dentry->d_inode), sizeof(struct nfs_fh));
 	lock->caller  = system_utsname.nodename;
 	lock->oh.data = req->a_owner;
-	lock->oh.len  = sprintf(req->a_owner, "%d@%s",
-				current->pid, system_utsname.nodename);
-	locks_copy_lock(&lock->fl, fl);
+	lock->oh.len  = snprintf(req->a_owner, sizeof(req->a_owner), "%u@%s",
+				(unsigned int)fl->fl_u.nfs_fl.owner->pid,
+				system_utsname.nodename);
+	lock->svid = fl->fl_u.nfs_fl.owner->pid;
+	lock->fl.fl_start = fl->fl_start;
+	lock->fl.fl_end = fl->fl_end;
+	lock->fl.fl_type = fl->fl_type;
 }
 
 static void nlmclnt_release_lockargs(struct nlm_rqst *req)
 {
-	struct file_lock *fl = &req->a_args.lock.fl;
-
-	if (fl->fl_ops && fl->fl_ops->fl_release_private)
-		fl->fl_ops->fl_release_private(fl);
-}
-
-/*
- * Initialize arguments for GRANTED call. The nlm_rqst structure
- * has been cleared already.
- */
-int
-nlmclnt_setgrantargs(struct nlm_rqst *call, struct nlm_lock *lock)
-{
-	locks_copy_lock(&call->a_args.lock.fl, &lock->fl);
-	memcpy(&call->a_args.lock.fh, &lock->fh, sizeof(call->a_args.lock.fh));
-	call->a_args.lock.caller = system_utsname.nodename;
-	call->a_args.lock.oh.len = lock->oh.len;
-
-	/* set default data area */
-	call->a_args.lock.oh.data = call->a_owner;
-
-	if (lock->oh.len > NLMCLNT_OHSIZE) {
-		void *data = kmalloc(lock->oh.len, GFP_KERNEL);
-		if (!data) {
-			nlmclnt_freegrantargs(call);
-			return 0;
-		}
-		call->a_args.lock.oh.data = (u8 *) data;
-	}
-
-	memcpy(call->a_args.lock.oh.data, lock->oh.data, lock->oh.len);
-	return 1;
-}
-
-void
-nlmclnt_freegrantargs(struct nlm_rqst *call)
-{
-	struct file_lock *fl = &call->a_args.lock.fl;
-	/*
-	 * Check whether we allocated memory for the owner.
-	 */
-	if (call->a_args.lock.oh.data != (u8 *) call->a_owner) {
-		kfree(call->a_args.lock.oh.data);
-	}
-	if (fl->fl_ops && fl->fl_ops->fl_release_private)
-		fl->fl_ops->fl_release_private(fl);
+	BUG_ON(req->a_args.lock.fl.fl_ops != NULL);
 }
 
 /*
@@ -193,9 +152,8 @@ nlmclnt_freegrantargs(struct nlm_rqst *call)
 int
 nlmclnt_proc(struct inode *inode, int cmd, struct file_lock *fl)
 {
-	struct nfs_server	*nfssrv = NFS_SERVER(inode);
 	struct nlm_host		*host;
-	struct nlm_rqst		reqst, *call = &reqst;
+	struct nlm_rqst		*call;
 	sigset_t		oldset;
 	unsigned long		flags;
 	int			status, proto, vers;
@@ -209,23 +167,17 @@ nlmclnt_proc(struct inode *inode, int cmd, struct file_lock *fl)
 	/* Retrieve transport protocol from NFS client */
 	proto = NFS_CLIENT(inode)->cl_xprt->prot;
 
-	if (!(host = nlmclnt_lookup_host(NFS_ADDR(inode), proto, vers)))
+	host = nlmclnt_lookup_host(NFS_ADDR(inode), proto, vers);
+	if (host == NULL)
 		return -ENOLCK;
 
-	/* Create RPC client handle if not there, and copy soft
-	 * and intr flags from NFS client. */
-	if (host->h_rpcclnt == NULL) {
-		struct rpc_clnt	*clnt;
+	call = nlm_alloc_call(host);
+	if (call == NULL)
+		return -ENOMEM;
 
-		/* Bind an rpc client to this host handle (does not
-		 * perform a portmapper lookup) */
-		if (!(clnt = nlm_bind_host(host))) {
-			status = -ENOLCK;
-			goto done;
-		}
-		clnt->cl_softrtry = nfssrv->client->cl_softrtry;
-		clnt->cl_intr = nfssrv->client->cl_intr;
-	}
+	nlmclnt_locks_init_private(fl, host);
+	/* Set up the argument struct */
+	nlmclnt_setlockargs(call, fl);
 
 	/* Keep the old signal mask */
 	spin_lock_irqsave(&current->sighand->siglock, flags);
@@ -238,26 +190,10 @@ nlmclnt_proc(struct inode *inode, int cmd, struct file_lock *fl)
 	    && (current->flags & PF_EXITING)) {
 		sigfillset(&current->blocked);	/* Mask all signals */
 		recalc_sigpending();
-		spin_unlock_irqrestore(&current->sighand->siglock, flags);
 
-		call = nlmclnt_alloc_call();
-		if (!call) {
-			status = -ENOMEM;
-			goto out_restore;
-		}
 		call->a_flags = RPC_TASK_ASYNC;
-	} else {
-		spin_unlock_irqrestore(&current->sighand->siglock, flags);
-		memset(call, 0, sizeof(*call));
-		locks_init_lock(&call->a_args.lock.fl);
-		locks_init_lock(&call->a_res.lock.fl);
 	}
-	call->a_host = host;
-
-	nlmclnt_locks_init_private(fl, host);
-
-	/* Set up the argument struct */
-	nlmclnt_setlockargs(call, fl);
+	spin_unlock_irqrestore(&current->sighand->siglock, flags);
 
 	if (IS_SETLK(cmd) || IS_SETLKW(cmd)) {
 		if (fl->fl_type != F_UNLCK) {
@@ -270,41 +206,58 @@ nlmclnt_proc(struct inode *inode, int cmd, struct file_lock *fl)
 	else
 		status = -EINVAL;
 
- out_restore:
+	fl->fl_ops->fl_release_private(fl);
+	fl->fl_ops = NULL;
+
 	spin_lock_irqsave(&current->sighand->siglock, flags);
 	current->blocked = oldset;
 	recalc_sigpending();
 	spin_unlock_irqrestore(&current->sighand->siglock, flags);
 
-done:
 	dprintk("lockd: clnt proc returns %d\n", status);
-	nlm_release_host(host);
 	return status;
 }
 EXPORT_SYMBOL(nlmclnt_proc);
 
 /*
  * Allocate an NLM RPC call struct
+ *
+ * Note: the caller must hold a reference to host. In case of failure,
+ * this reference will be released.
  */
-struct nlm_rqst *
-nlmclnt_alloc_call(void)
+struct nlm_rqst *nlm_alloc_call(struct nlm_host *host)
 {
 	struct nlm_rqst	*call;
 
-	while (!signalled()) {
-		call = (struct nlm_rqst *) kmalloc(sizeof(struct nlm_rqst), GFP_KERNEL);
-		if (call) {
-			memset(call, 0, sizeof(*call));
+	for(;;) {
+		call = kzalloc(sizeof(*call), GFP_KERNEL);
+		if (call != NULL) {
 			locks_init_lock(&call->a_args.lock.fl);
 			locks_init_lock(&call->a_res.lock.fl);
+			call->a_host = host;
 			return call;
 		}
-		printk("nlmclnt_alloc_call: failed, waiting for memory\n");
+		if (signalled())
+			break;
+		printk("nlm_alloc_call: failed, waiting for memory\n");
 		schedule_timeout_interruptible(5*HZ);
 	}
+	nlm_release_host(host);
 	return NULL;
 }
 
+void nlm_release_call(struct nlm_rqst *call)
+{
+	nlm_release_host(call->a_host);
+	nlmclnt_release_lockargs(call);
+	kfree(call);
+}
+
+static void nlmclnt_rpc_release(void *data)
+{
+	return nlm_release_call(data);
+}
+
 static int nlm_wait_on_grace(wait_queue_head_t *queue)
 {
 	DEFINE_WAIT(wait);
@@ -401,57 +354,45 @@ in_grace_period:
 /*
  * Generic NLM call, async version.
  */
-int nlmsvc_async_call(struct nlm_rqst *req, u32 proc, const struct rpc_call_ops *tk_ops)
+static int __nlm_async_call(struct nlm_rqst *req, u32 proc, struct rpc_message *msg, const struct rpc_call_ops *tk_ops)
 {
 	struct nlm_host	*host = req->a_host;
 	struct rpc_clnt	*clnt;
-	struct rpc_message msg = {
-		.rpc_argp	= &req->a_args,
-		.rpc_resp	= &req->a_res,
-	};
-	int		status;
+	int status = -ENOLCK;
 
 	dprintk("lockd: call procedure %d on %s (async)\n",
 			(int)proc, host->h_name);
 
 	/* If we have no RPC client yet, create one. */
-	if ((clnt = nlm_bind_host(host)) == NULL)
-		return -ENOLCK;
-	msg.rpc_proc = &clnt->cl_procinfo[proc];
+	clnt = nlm_bind_host(host);
+	if (clnt == NULL)
+		goto out_err;
+	msg->rpc_proc = &clnt->cl_procinfo[proc];
 
         /* bootstrap and kick off the async RPC call */
-        status = rpc_call_async(clnt, &msg, RPC_TASK_ASYNC, tk_ops, req);
-
+        status = rpc_call_async(clnt, msg, RPC_TASK_ASYNC, tk_ops, req);
+	if (status == 0)
+		return 0;
+out_err:
+	nlm_release_call(req);
 	return status;
 }
 
-static int nlmclnt_async_call(struct nlm_rqst *req, u32 proc, const struct rpc_call_ops *tk_ops)
+int nlm_async_call(struct nlm_rqst *req, u32 proc, const struct rpc_call_ops *tk_ops)
 {
-	struct nlm_host	*host = req->a_host;
-	struct rpc_clnt	*clnt;
-	struct nlm_args	*argp = &req->a_args;
-	struct nlm_res	*resp = &req->a_res;
 	struct rpc_message msg = {
-		.rpc_argp	= argp,
-		.rpc_resp	= resp,
+		.rpc_argp	= &req->a_args,
+		.rpc_resp	= &req->a_res,
 	};
-	int		status;
-
-	dprintk("lockd: call procedure %d on %s (async)\n",
-			(int)proc, host->h_name);
-
-	/* If we have no RPC client yet, create one. */
-	if ((clnt = nlm_bind_host(host)) == NULL)
-		return -ENOLCK;
-	msg.rpc_proc = &clnt->cl_procinfo[proc];
+	return __nlm_async_call(req, proc, &msg, tk_ops);
+}
 
-	/* Increment host refcount */
-	nlm_get_host(host);
-        /* bootstrap and kick off the async RPC call */
-        status = rpc_call_async(clnt, &msg, RPC_TASK_ASYNC, tk_ops, req);
-	if (status < 0)
-		nlm_release_host(host);
-	return status;
+int nlm_async_reply(struct nlm_rqst *req, u32 proc, const struct rpc_call_ops *tk_ops)
+{
+	struct rpc_message msg = {
+		.rpc_argp	= &req->a_res,
+	};
+	return __nlm_async_call(req, proc, &msg, tk_ops);
 }
 
 /*
@@ -463,36 +404,41 @@ nlmclnt_test(struct nlm_rqst *req, struct file_lock *fl)
 	int	status;
 
 	status = nlmclnt_call(req, NLMPROC_TEST);
-	nlmclnt_release_lockargs(req);
 	if (status < 0)
-		return status;
+		goto out;
 
-	status = req->a_res.status;
-	if (status == NLM_LCK_GRANTED) {
-		fl->fl_type = F_UNLCK;
-	} if (status == NLM_LCK_DENIED) {
-		/*
-		 * Report the conflicting lock back to the application.
-		 */
-		locks_copy_lock(fl, &req->a_res.lock.fl);
-		fl->fl_pid = 0;
-	} else {
-		return nlm_stat_to_errno(req->a_res.status);
+	switch (req->a_res.status) {
+		case NLM_LCK_GRANTED:
+			fl->fl_type = F_UNLCK;
+			break;
+		case NLM_LCK_DENIED:
+			/*
+			 * Report the conflicting lock back to the application.
+			 */
+			fl->fl_start = req->a_res.lock.fl.fl_start;
+			fl->fl_end = req->a_res.lock.fl.fl_start;
+			fl->fl_type = req->a_res.lock.fl.fl_type;
+			fl->fl_pid = 0;
+			break;
+		default:
+			status = nlm_stat_to_errno(req->a_res.status);
 	}
-
-	return 0;
+out:
+	nlm_release_call(req);
+	return status;
 }
 
 static void nlmclnt_locks_copy_lock(struct file_lock *new, struct file_lock *fl)
 {
-	memcpy(&new->fl_u.nfs_fl, &fl->fl_u.nfs_fl, sizeof(new->fl_u.nfs_fl));
-	nlm_get_lockowner(new->fl_u.nfs_fl.owner);
+	new->fl_u.nfs_fl.state = fl->fl_u.nfs_fl.state;
+	new->fl_u.nfs_fl.owner = nlm_get_lockowner(fl->fl_u.nfs_fl.owner);
+	list_add_tail(&new->fl_u.nfs_fl.list, &fl->fl_u.nfs_fl.owner->host->h_granted);
 }
 
 static void nlmclnt_locks_release_private(struct file_lock *fl)
 {
+	list_del(&fl->fl_u.nfs_fl.list);
 	nlm_put_lockowner(fl->fl_u.nfs_fl.owner);
-	fl->fl_ops = NULL;
 }
 
 static struct file_lock_operations nlmclnt_lock_ops = {
@@ -504,8 +450,8 @@ static void nlmclnt_locks_init_private(struct file_lock *fl, struct nlm_host *ho
 {
 	BUG_ON(fl->fl_ops != NULL);
 	fl->fl_u.nfs_fl.state = 0;
-	fl->fl_u.nfs_fl.flags = 0;
 	fl->fl_u.nfs_fl.owner = nlm_find_lockowner(host, fl->fl_owner);
+	INIT_LIST_HEAD(&fl->fl_u.nfs_fl.list);
 	fl->fl_ops = &nlmclnt_lock_ops;
 }
 
@@ -552,57 +498,52 @@ nlmclnt_lock(struct nlm_rqst *req, struct file_lock *fl)
 {
 	struct nlm_host	*host = req->a_host;
 	struct nlm_res	*resp = &req->a_res;
-	long timeout;
-	int status;
+	struct nlm_wait *block = NULL;
+	int status = -ENOLCK;
 
 	if (!host->h_monitored && nsm_monitor(host) < 0) {
 		printk(KERN_NOTICE "lockd: failed to monitor %s\n",
 					host->h_name);
-		status = -ENOLCK;
 		goto out;
 	}
 
-	if (req->a_args.block) {
-		status = nlmclnt_prepare_block(req, host, fl);
-		if (status < 0)
-			goto out;
-	}
+	block = nlmclnt_prepare_block(host, fl);
 	for(;;) {
 		status = nlmclnt_call(req, NLMPROC_LOCK);
 		if (status < 0)
 			goto out_unblock;
-		if (resp->status != NLM_LCK_BLOCKED)
+		if (!req->a_args.block)
 			break;
-		/* Wait on an NLM blocking lock */
-		timeout = nlmclnt_block(req, NLMCLNT_POLL_TIMEOUT);
 		/* Did a reclaimer thread notify us of a server reboot? */
 		if (resp->status ==  NLM_LCK_DENIED_GRACE_PERIOD)
 			continue;
 		if (resp->status != NLM_LCK_BLOCKED)
 			break;
-		if (timeout >= 0)
-			continue;
-		/* We were interrupted. Send a CANCEL request to the server
+		/* Wait on an NLM blocking lock */
+		status = nlmclnt_block(block, req, NLMCLNT_POLL_TIMEOUT);
+		/* if we were interrupted. Send a CANCEL request to the server
 		 * and exit
 		 */
-		status = (int)timeout;
-		goto out_unblock;
+		if (status < 0)
+			goto out_unblock;
+		if (resp->status != NLM_LCK_BLOCKED)
+			break;
 	}
 
 	if (resp->status == NLM_LCK_GRANTED) {
 		fl->fl_u.nfs_fl.state = host->h_state;
-		fl->fl_u.nfs_fl.flags |= NFS_LCK_GRANTED;
 		fl->fl_flags |= FL_SLEEP;
+		/* Ensure the resulting lock will get added to granted list */
 		do_vfs_lock(fl);
 	}
 	status = nlm_stat_to_errno(resp->status);
 out_unblock:
-	nlmclnt_finish_block(req);
+	nlmclnt_finish_block(block);
 	/* Cancel the blocked request if it is still pending */
 	if (resp->status == NLM_LCK_BLOCKED)
 		nlmclnt_cancel(host, req->a_args.block, fl);
 out:
-	nlmclnt_release_lockargs(req);
+	nlm_release_call(req);
 	return status;
 }
 
@@ -658,10 +599,6 @@ nlmclnt_unlock(struct nlm_rqst *req, struct file_lock *fl)
 	struct nlm_res	*resp = &req->a_res;
 	int		status;
 
-	/* Clean the GRANTED flag now so the lock doesn't get
-	 * reclaimed while we're stuck in the unlock call. */
-	fl->fl_u.nfs_fl.flags &= ~NFS_LCK_GRANTED;
-
 	/*
 	 * Note: the server is supposed to either grant us the unlock
 	 * request, or to deny it with NLM_LCK_DENIED_GRACE_PERIOD. In either
@@ -669,32 +606,24 @@ nlmclnt_unlock(struct nlm_rqst *req, struct file_lock *fl)
 	 */
 	do_vfs_lock(fl);
 
-	if (req->a_flags & RPC_TASK_ASYNC) {
-		status = nlmclnt_async_call(req, NLMPROC_UNLOCK,
-					&nlmclnt_unlock_ops);
-		/* Hrmf... Do the unlock early since locks_remove_posix()
-		 * really expects us to free the lock synchronously */
-		if (status < 0) {
-			nlmclnt_release_lockargs(req);
-			kfree(req);
-		}
-		return status;
-	}
+	if (req->a_flags & RPC_TASK_ASYNC)
+		return nlm_async_call(req, NLMPROC_UNLOCK, &nlmclnt_unlock_ops);
 
 	status = nlmclnt_call(req, NLMPROC_UNLOCK);
-	nlmclnt_release_lockargs(req);
 	if (status < 0)
-		return status;
+		goto out;
 
+	status = 0;
 	if (resp->status == NLM_LCK_GRANTED)
-		return 0;
+		goto out;
 
 	if (resp->status != NLM_LCK_DENIED_NOLOCKS)
 		printk("lockd: unexpected unlock status: %d\n", resp->status);
-
 	/* What to do now? I'm out of my depth... */
-
-	return -ENOLCK;
+	status = -ENOLCK;
+out:
+	nlm_release_call(req);
+	return status;
 }
 
 static void nlmclnt_unlock_callback(struct rpc_task *task, void *data)
@@ -716,9 +645,6 @@ static void nlmclnt_unlock_callback(struct rpc_task *task, void *data)
 	if (status != NLM_LCK_GRANTED)
 		printk(KERN_WARNING "lockd: unexpected unlock status: %d\n", status);
 die:
-	nlm_release_host(req->a_host);
-	nlmclnt_release_lockargs(req);
-	kfree(req);
 	return;
  retry_rebind:
 	nlm_rebind_host(req->a_host);
@@ -728,6 +654,7 @@ die:
 
 static const struct rpc_call_ops nlmclnt_unlock_ops = {
 	.rpc_call_done = nlmclnt_unlock_callback,
+	.rpc_release = nlmclnt_rpc_release,
 };
 
 /*
@@ -749,20 +676,15 @@ static int nlmclnt_cancel(struct nlm_host *host, int block, struct file_lock *fl
 	recalc_sigpending();
 	spin_unlock_irqrestore(&current->sighand->siglock, flags);
 
-	req = nlmclnt_alloc_call();
+	req = nlm_alloc_call(nlm_get_host(host));
 	if (!req)
 		return -ENOMEM;
-	req->a_host  = host;
 	req->a_flags = RPC_TASK_ASYNC;
 
 	nlmclnt_setlockargs(req, fl);
 	req->a_args.block = block;
 
-	status = nlmclnt_async_call(req, NLMPROC_CANCEL, &nlmclnt_cancel_ops);
-	if (status < 0) {
-		nlmclnt_release_lockargs(req);
-		kfree(req);
-	}
+	status = nlm_async_call(req, NLMPROC_CANCEL, &nlmclnt_cancel_ops);
 
 	spin_lock_irqsave(&current->sighand->siglock, flags);
 	current->blocked = oldset;
@@ -791,6 +713,7 @@ static void nlmclnt_cancel_callback(struct rpc_task *task, void *data)
 	switch (req->a_res.status) {
 	case NLM_LCK_GRANTED:
 	case NLM_LCK_DENIED_GRACE_PERIOD:
+	case NLM_LCK_DENIED:
 		/* Everything's good */
 		break;
 	case NLM_LCK_DENIED_NOLOCKS:
@@ -802,9 +725,6 @@ static void nlmclnt_cancel_callback(struct rpc_task *task, void *data)
 	}
 
 die:
-	nlm_release_host(req->a_host);
-	nlmclnt_release_lockargs(req);
-	kfree(req);
 	return;
 
 retry_cancel:
@@ -818,6 +738,7 @@ retry_cancel:
 
 static const struct rpc_call_ops nlmclnt_cancel_ops = {
 	.rpc_call_done = nlmclnt_cancel_callback,
+	.rpc_release = nlmclnt_rpc_release,
 };
 
 /*

fs/lockd/host.c

@@ -123,6 +123,8 @@ nlm_lookup_host(int server, struct sockaddr_in *sin,
 	nlm_hosts[hash]    = host;
 	INIT_LIST_HEAD(&host->h_lockowners);
 	spin_lock_init(&host->h_lock);
+	INIT_LIST_HEAD(&host->h_granted);
+	INIT_LIST_HEAD(&host->h_reclaim);
 
 	if (++nrhosts > NLM_HOST_MAX)
 		next_gc = 0;
@@ -191,11 +193,12 @@ nlm_bind_host(struct nlm_host *host)
 		xprt->resvport = 1;	/* NLM requires a reserved port */
 
 		/* Existing NLM servers accept AUTH_UNIX only */
-		clnt = rpc_create_client(xprt, host->h_name, &nlm_program,
+		clnt = rpc_new_client(xprt, host->h_name, &nlm_program,
 					host->h_version, RPC_AUTH_UNIX);
 		if (IS_ERR(clnt))
 			goto forgetit;
 		clnt->cl_autobind = 1;	/* turn on pmap queries */
+		clnt->cl_softrtry = 1; /* All queries are soft */
 
 		host->h_rpcclnt = clnt;
 	}
@@ -242,8 +245,12 @@ void nlm_release_host(struct nlm_host *host)
 {
 	if (host != NULL) {
 		dprintk("lockd: release host %s\n", host->h_name);
-		atomic_dec(&host->h_count);
 		BUG_ON(atomic_read(&host->h_count) < 0);
+		if (atomic_dec_and_test(&host->h_count)) {
+			BUG_ON(!list_empty(&host->h_lockowners));
+			BUG_ON(!list_empty(&host->h_granted));
+			BUG_ON(!list_empty(&host->h_reclaim));
+		}
 	}
 }
 
@@ -331,7 +338,6 @@ nlm_gc_hosts(void)
 					rpc_destroy_client(host->h_rpcclnt);
 				}
 			}
-			BUG_ON(!list_empty(&host->h_lockowners));
 			kfree(host);
 			nrhosts--;
 		}

fs/lockd/mon.c

@@ -35,6 +35,10 @@ nsm_mon_unmon(struct nlm_host *host, u32 proc, struct nsm_res *res)
 	struct rpc_clnt	*clnt;
 	int		status;
 	struct nsm_args	args;
+	struct rpc_message msg = {
+		.rpc_argp	= &args,
+		.rpc_resp	= res,
+	};
 
 	clnt = nsm_create();
 	if (IS_ERR(clnt)) {
@@ -49,7 +53,8 @@ nsm_mon_unmon(struct nlm_host *host, u32 proc, struct nsm_res *res)
 	args.proc = NLMPROC_NSM_NOTIFY;
 	memset(res, 0, sizeof(*res));
 
-	status = rpc_call(clnt, proc, &args, res, 0);
+	msg.rpc_proc = &clnt->cl_procinfo[proc];
+	status = rpc_call_sync(clnt, &msg, 0);
 	if (status < 0)
 		printk(KERN_DEBUG "nsm_mon_unmon: rpc failed, status=%d\n",
 			status);
@@ -214,12 +219,16 @@ static struct rpc_procinfo	nsm_procedures[] = {
 		.p_encode	= (kxdrproc_t) xdr_encode_mon,
 		.p_decode	= (kxdrproc_t) xdr_decode_stat_res,
 		.p_bufsiz	= MAX(SM_mon_sz, SM_monres_sz) << 2,
+		.p_statidx	= SM_MON,
+		.p_name		= "MONITOR",
 	},
 [SM_UNMON] = {
 		.p_proc		= SM_UNMON,
 		.p_encode	= (kxdrproc_t) xdr_encode_unmon,
 		.p_decode	= (kxdrproc_t) xdr_decode_stat,
 		.p_bufsiz	= MAX(SM_mon_id_sz, SM_unmonres_sz) << 2,
+		.p_statidx	= SM_UNMON,
+		.p_name		= "UNMONITOR",
 	},
 };
 

fs/lockd/svc4proc.c

@@ -21,10 +21,6 @@
 
 #define NLMDBG_FACILITY		NLMDBG_CLIENT
 
-static u32	nlm4svc_callback(struct svc_rqst *, u32, struct nlm_res *);
-
-static const struct rpc_call_ops nlm4svc_callback_ops;
-
 /*
  * Obtain client and file from arguments
  */
@@ -233,84 +229,90 @@ nlm4svc_proc_granted(struct svc_rqst *rqstp, struct nlm_args *argp,
 	return rpc_success;
 }
 
+/*
+ * This is the generic lockd callback for async RPC calls
+ */
+static void nlm4svc_callback_exit(struct rpc_task *task, void *data)
+{
+	dprintk("lockd: %4d callback returned %d\n", task->tk_pid,
+			-task->tk_status);
+}
+
+static void nlm4svc_callback_release(void *data)
+{
+	nlm_release_call(data);
+}
+
+static const struct rpc_call_ops nlm4svc_callback_ops = {
+	.rpc_call_done = nlm4svc_callback_exit,
+	.rpc_release = nlm4svc_callback_release,
+};
+
 /*
  * `Async' versions of the above service routines. They aren't really,
  * because we send the callback before the reply proper. I hope this
  * doesn't break any clients.
  */
-static int
-nlm4svc_proc_test_msg(struct svc_rqst *rqstp, struct nlm_args *argp,
-					     void	     *resp)
+static int nlm4svc_callback(struct svc_rqst *rqstp, u32 proc, struct nlm_args *argp,
+		int (*func)(struct svc_rqst *, struct nlm_args *, struct nlm_res  *))
 {
-	struct nlm_res	res;
-	u32		stat;
+	struct nlm_host	*host;
+	struct nlm_rqst	*call;
+	int stat;
 
-	dprintk("lockd: TEST_MSG      called\n");
-	memset(&res, 0, sizeof(res));
+	host = nlmsvc_lookup_host(rqstp);
+	if (host == NULL)
+		return rpc_system_err;
+
+	call = nlm_alloc_call(host);
+	if (call == NULL)
+		return rpc_system_err;
 
-	if ((stat = nlm4svc_proc_test(rqstp, argp, &res)) == 0)
-		stat = nlm4svc_callback(rqstp, NLMPROC_TEST_RES, &res);
-	return stat;
+	stat = func(rqstp, argp, &call->a_res);
+	if (stat != 0) {
+		nlm_release_call(call);
+		return stat;
+	}
+
+	call->a_flags = RPC_TASK_ASYNC;
+	if (nlm_async_reply(call, proc, &nlm4svc_callback_ops) < 0)
+		return rpc_system_err;
+	return rpc_success;
 }
 
-static int
-nlm4svc_proc_lock_msg(struct svc_rqst *rqstp, struct nlm_args *argp,
+static int nlm4svc_proc_test_msg(struct svc_rqst *rqstp, struct nlm_args *argp,
 					     void	     *resp)
 {
-	struct nlm_res	res;
-	u32		stat;
+	dprintk("lockd: TEST_MSG      called\n");
+	return nlm4svc_callback(rqstp, NLMPROC_TEST_RES, argp, nlm4svc_proc_test);
+}
 
+static int nlm4svc_proc_lock_msg(struct svc_rqst *rqstp, struct nlm_args *argp,
+					     void	     *resp)
+{
 	dprintk("lockd: LOCK_MSG      called\n");
-	memset(&res, 0, sizeof(res));
-
-	if ((stat = nlm4svc_proc_lock(rqstp, argp, &res)) == 0)
-		stat = nlm4svc_callback(rqstp, NLMPROC_LOCK_RES, &res);
-	return stat;
+	return nlm4svc_callback(rqstp, NLMPROC_LOCK_RES, argp, nlm4svc_proc_lock);
 }
 
-static int
-nlm4svc_proc_cancel_msg(struct svc_rqst *rqstp, struct nlm_args *argp,
+static int nlm4svc_proc_cancel_msg(struct svc_rqst *rqstp, struct nlm_args *argp,
 					       void	       *resp)
 {
-	struct nlm_res	res;
-	u32		stat;
-
 	dprintk("lockd: CANCEL_MSG    called\n");
-	memset(&res, 0, sizeof(res));
-
-	if ((stat = nlm4svc_proc_cancel(rqstp, argp, &res)) == 0)
-		stat = nlm4svc_callback(rqstp, NLMPROC_CANCEL_RES, &res);
-	return stat;
+	return nlm4svc_callback(rqstp, NLMPROC_CANCEL_RES, argp, nlm4svc_proc_cancel);
 }
 
-static int
-nlm4svc_proc_unlock_msg(struct svc_rqst *rqstp, struct nlm_args *argp,
+static int nlm4svc_proc_unlock_msg(struct svc_rqst *rqstp, struct nlm_args *argp,
                                                void            *resp)
 {
-	struct nlm_res	res;
-	u32		stat;
-
 	dprintk("lockd: UNLOCK_MSG    called\n");
-	memset(&res, 0, sizeof(res));
-
-	if ((stat = nlm4svc_proc_unlock(rqstp, argp, &res)) == 0)
-		stat = nlm4svc_callback(rqstp, NLMPROC_UNLOCK_RES, &res);
-	return stat;
+	return nlm4svc_callback(rqstp, NLMPROC_UNLOCK_RES, argp, nlm4svc_proc_unlock);
 }
 
-static int
-nlm4svc_proc_granted_msg(struct svc_rqst *rqstp, struct nlm_args *argp,
+static int nlm4svc_proc_granted_msg(struct svc_rqst *rqstp, struct nlm_args *argp,
                                                 void            *resp)
 {
-	struct nlm_res	res;
-	u32		stat;
-
 	dprintk("lockd: GRANTED_MSG   called\n");
-	memset(&res, 0, sizeof(res));
-
-	if ((stat = nlm4svc_proc_granted(rqstp, argp, &res)) == 0)
-		stat = nlm4svc_callback(rqstp, NLMPROC_GRANTED_RES, &res);
-	return stat;
+	return nlm4svc_callback(rqstp, NLMPROC_GRANTED_RES, argp, nlm4svc_proc_granted);
 }
 
 /*
@@ -471,55 +473,6 @@ nlm4svc_proc_granted_res(struct svc_rqst *rqstp, struct nlm_res  *argp,
 }
 
 
-/*
- * This is the generic lockd callback for async RPC calls
- */
-static u32
-nlm4svc_callback(struct svc_rqst *rqstp, u32 proc, struct nlm_res *resp)
-{
-	struct nlm_host	*host;
-	struct nlm_rqst	*call;
-
-	if (!(call = nlmclnt_alloc_call()))
-		return rpc_system_err;
-
-	host = nlmclnt_lookup_host(&rqstp->rq_addr,
-				rqstp->rq_prot, rqstp->rq_vers);
-	if (!host) {
-		kfree(call);
-		return rpc_system_err;
-	}
-
-	call->a_flags = RPC_TASK_ASYNC;
-	call->a_host  = host;
-	memcpy(&call->a_args, resp, sizeof(*resp));
-
-	if (nlmsvc_async_call(call, proc, &nlm4svc_callback_ops) < 0)
-		goto error;
-
-	return rpc_success;
- error:
-	kfree(call);
-	nlm_release_host(host);
-	return rpc_system_err;
-}
-
-static void nlm4svc_callback_exit(struct rpc_task *task, void *data)
-{
-	struct nlm_rqst	*call = data;
-
-	if (task->tk_status < 0) {
-		dprintk("lockd: %4d callback failed (errno = %d)\n",
-					task->tk_pid, -task->tk_status);
-	}
-	nlm_release_host(call->a_host);
-	kfree(call);
-}
-
-static const struct rpc_call_ops nlm4svc_callback_ops = {
-	.rpc_call_done = nlm4svc_callback_exit,
-};
-
 /*
  * NLM Server procedures.
  */

fs/lockd/svclock.c

@@ -39,9 +39,12 @@
 #define nlm_deadlock	nlm_lck_denied
 #endif
 
+static void nlmsvc_release_block(struct nlm_block *block);
 static void	nlmsvc_insert_block(struct nlm_block *block, unsigned long);
 static int	nlmsvc_remove_block(struct nlm_block *block);
 
+static int nlmsvc_setgrantargs(struct nlm_rqst *call, struct nlm_lock *lock);
+static void nlmsvc_freegrantargs(struct nlm_rqst *call);
 static const struct rpc_call_ops nlmsvc_grant_ops;
 
 /*
@@ -58,6 +61,7 @@ nlmsvc_insert_block(struct nlm_block *block, unsigned long when)
 	struct nlm_block **bp, *b;
 
 	dprintk("lockd: nlmsvc_insert_block(%p, %ld)\n", block, when);
+	kref_get(&block->b_count);
 	if (block->b_queued)
 		nlmsvc_remove_block(block);
 	bp = &nlm_blocked;
@@ -90,6 +94,7 @@ nlmsvc_remove_block(struct nlm_block *block)
 		if (b == block) {
 			*bp = block->b_next;
 			block->b_queued = 0;
+			nlmsvc_release_block(block);
 			return 1;
 		}
 	}
@@ -98,11 +103,10 @@ nlmsvc_remove_block(struct nlm_block *block)
 }
 
 /*
- * Find a block for a given lock and optionally remove it from
- * the list.
+ * Find a block for a given lock
  */
 static struct nlm_block *
-nlmsvc_lookup_block(struct nlm_file *file, struct nlm_lock *lock, int remove)
+nlmsvc_lookup_block(struct nlm_file *file, struct nlm_lock *lock)
 {
 	struct nlm_block	**head, *block;
 	struct file_lock	*fl;
@@ -112,17 +116,14 @@ nlmsvc_lookup_block(struct nlm_file *file, struct nlm_lock *lock, int remove)
 				(long long)lock->fl.fl_start,
 				(long long)lock->fl.fl_end, lock->fl.fl_type);
 	for (head = &nlm_blocked; (block = *head) != 0; head = &block->b_next) {
-		fl = &block->b_call.a_args.lock.fl;
+		fl = &block->b_call->a_args.lock.fl;
 		dprintk("lockd: check f=%p pd=%d %Ld-%Ld ty=%d cookie=%s\n",
 				block->b_file, fl->fl_pid,
 				(long long)fl->fl_start,
 				(long long)fl->fl_end, fl->fl_type,
-				nlmdbg_cookie2a(&block->b_call.a_args.cookie));
+				nlmdbg_cookie2a(&block->b_call->a_args.cookie));
 		if (block->b_file == file && nlm_compare_locks(fl, &lock->fl)) {
-			if (remove) {
-				*head = block->b_next;
-				block->b_queued = 0;
-			}
+			kref_get(&block->b_count);
 			return block;
 		}
 	}
@@ -150,11 +151,13 @@ nlmsvc_find_block(struct nlm_cookie *cookie,  struct sockaddr_in *sin)
 	for (block = nlm_blocked; block; block = block->b_next) {
 		dprintk("cookie: head of blocked queue %p, block %p\n", 
 			nlm_blocked, block);
-		if (nlm_cookie_match(&block->b_call.a_args.cookie,cookie)
+		if (nlm_cookie_match(&block->b_call->a_args.cookie,cookie)
 				&& nlm_cmp_addr(sin, &block->b_host->h_addr))
 			break;
 	}
 
+	if (block != NULL)
+		kref_get(&block->b_count);
 	return block;
 }
 
@@ -174,27 +177,30 @@ nlmsvc_create_block(struct svc_rqst *rqstp, struct nlm_file *file,
 {
 	struct nlm_block	*block;
 	struct nlm_host		*host;
-	struct nlm_rqst		*call;
+	struct nlm_rqst		*call = NULL;
 
 	/* Create host handle for callback */
-	host = nlmclnt_lookup_host(&rqstp->rq_addr,
-				rqstp->rq_prot, rqstp->rq_vers);
+	host = nlmsvc_lookup_host(rqstp);
 	if (host == NULL)
 		return NULL;
 
+	call = nlm_alloc_call(host);
+	if (call == NULL)
+		return NULL;
+
 	/* Allocate memory for block, and initialize arguments */
-	if (!(block = (struct nlm_block *) kmalloc(sizeof(*block), GFP_KERNEL)))
+	block = kzalloc(sizeof(*block), GFP_KERNEL);
+	if (block == NULL)
 		goto failed;
-	memset(block, 0, sizeof(*block));
-	locks_init_lock(&block->b_call.a_args.lock.fl);
-	locks_init_lock(&block->b_call.a_res.lock.fl);
+	kref_init(&block->b_count);
 
-	if (!nlmclnt_setgrantargs(&block->b_call, lock))
+	if (!nlmsvc_setgrantargs(call, lock))
 		goto failed_free;
 
 	/* Set notifier function for VFS, and init args */
-	block->b_call.a_args.lock.fl.fl_lmops = &nlmsvc_lock_operations;
-	block->b_call.a_args.cookie = *cookie;	/* see above */
+	call->a_args.lock.fl.fl_flags |= FL_SLEEP;
+	call->a_args.lock.fl.fl_lmops = &nlmsvc_lock_operations;
+	call->a_args.cookie = *cookie;	/* see above */
 
 	dprintk("lockd: created block %p...\n", block);
 
@@ -202,22 +208,23 @@ nlmsvc_create_block(struct svc_rqst *rqstp, struct nlm_file *file,
 	block->b_daemon = rqstp->rq_server;
 	block->b_host   = host;
 	block->b_file   = file;
+	file->f_count++;
 
 	/* Add to file's list of blocks */
 	block->b_fnext  = file->f_blocks;
 	file->f_blocks  = block;
 
 	/* Set up RPC arguments for callback */
-	call = &block->b_call;
-	call->a_host    = host;
+	block->b_call = call;
 	call->a_flags   = RPC_TASK_ASYNC;
+	call->a_block = block;
 
 	return block;
 
 failed_free:
 	kfree(block);
 failed:
-	nlm_release_host(host);
+	nlm_release_call(call);
 	return NULL;
 }
 
@@ -227,29 +234,26 @@ failed:
  * It is the caller's responsibility to check whether the file
  * can be closed hereafter.
  */
-static int
-nlmsvc_delete_block(struct nlm_block *block, int unlock)
+static int nlmsvc_unlink_block(struct nlm_block *block)
 {
-	struct file_lock	*fl = &block->b_call.a_args.lock.fl;
-	struct nlm_file		*file = block->b_file;
-	struct nlm_block	**bp;
-	int status = 0;
-
-	dprintk("lockd: deleting block %p...\n", block);
+	int status;
+	dprintk("lockd: unlinking block %p...\n", block);
 
 	/* Remove block from list */
+	status = posix_unblock_lock(block->b_file->f_file, &block->b_call->a_args.lock.fl);
 	nlmsvc_remove_block(block);
-	if (unlock)
-		status = posix_unblock_lock(file->f_file, fl);
+	return status;
+}
 
-	/* If the block is in the middle of a GRANT callback,
-	 * don't kill it yet. */
-	if (block->b_incall) {
-		nlmsvc_insert_block(block, NLM_NEVER);
-		block->b_done = 1;
-		return status;
-	}
+static void nlmsvc_free_block(struct kref *kref)
+{
+	struct nlm_block *block = container_of(kref, struct nlm_block, b_count);
+	struct nlm_file		*file = block->b_file;
+	struct nlm_block	**bp;
 
+	dprintk("lockd: freeing block %p...\n", block);
+
+	down(&file->f_sema);
 	/* Remove block from file's list of blocks */
 	for (bp = &file->f_blocks; *bp; bp = &(*bp)->b_fnext) {
 		if (*bp == block) {
@@ -257,36 +261,93 @@ nlmsvc_delete_block(struct nlm_block *block, int unlock)
 			break;
 		}
 	}
+	up(&file->f_sema);
 
-	if (block->b_host)
-		nlm_release_host(block->b_host);
-	nlmclnt_freegrantargs(&block->b_call);
+	nlmsvc_freegrantargs(block->b_call);
+	nlm_release_call(block->b_call);
+	nlm_release_file(block->b_file);
 	kfree(block);
-	return status;
+}
+
+static void nlmsvc_release_block(struct nlm_block *block)
+{
+	if (block != NULL)
+		kref_put(&block->b_count, nlmsvc_free_block);
+}
+
+static void nlmsvc_act_mark(struct nlm_host *host, struct nlm_file *file)
+{
+	struct nlm_block *block;
+
+	down(&file->f_sema);
+	for (block = file->f_blocks; block != NULL; block = block->b_fnext)
+		block->b_host->h_inuse = 1;
+	up(&file->f_sema);
+}
+
+static void nlmsvc_act_unlock(struct nlm_host *host, struct nlm_file *file)
+{
+	struct nlm_block *block;
+
+restart:
+	down(&file->f_sema);
+	for (block = file->f_blocks; block != NULL; block = block->b_fnext) {
+		if (host != NULL && host != block->b_host)
+			continue;
+		if (!block->b_queued)
+			continue;
+		kref_get(&block->b_count);
+		up(&file->f_sema);
+		nlmsvc_unlink_block(block);
+		nlmsvc_release_block(block);
+		goto restart;
+	}
+	up(&file->f_sema);
 }
 
 /*
  * Loop over all blocks and perform the action specified.
  * (NLM_ACT_CHECK handled by nlmsvc_inspect_file).
  */
-int
+void
 nlmsvc_traverse_blocks(struct nlm_host *host, struct nlm_file *file, int action)
 {
-	struct nlm_block	*block, *next;
-	/* XXX: Will everything get cleaned up if we don't unlock here? */
+	if (action == NLM_ACT_MARK)
+		nlmsvc_act_mark(host, file);
+	else
+		nlmsvc_act_unlock(host, file);
+}
 
-	down(&file->f_sema);
-	for (block = file->f_blocks; block; block = next) {
-		next = block->b_fnext;
-		if (action == NLM_ACT_MARK)
-			block->b_host->h_inuse = 1;
-		else if (action == NLM_ACT_UNLOCK) {
-			if (host == NULL || host == block->b_host)
-				nlmsvc_delete_block(block, 1);
-		}
+/*
+ * Initialize arguments for GRANTED call. The nlm_rqst structure
+ * has been cleared already.
+ */
+static int nlmsvc_setgrantargs(struct nlm_rqst *call, struct nlm_lock *lock)
+{
+	locks_copy_lock(&call->a_args.lock.fl, &lock->fl);
+	memcpy(&call->a_args.lock.fh, &lock->fh, sizeof(call->a_args.lock.fh));
+	call->a_args.lock.caller = system_utsname.nodename;
+	call->a_args.lock.oh.len = lock->oh.len;
+
+	/* set default data area */
+	call->a_args.lock.oh.data = call->a_owner;
+	call->a_args.lock.svid = lock->fl.fl_pid;
+
+	if (lock->oh.len > NLMCLNT_OHSIZE) {
+		void *data = kmalloc(lock->oh.len, GFP_KERNEL);
+		if (!data)
+			return 0;
+		call->a_args.lock.oh.data = (u8 *) data;
 	}
-	up(&file->f_sema);
-	return 0;
+
+	memcpy(call->a_args.lock.oh.data, lock->oh.data, lock->oh.len);
+	return 1;
+}
+
+static void nlmsvc_freegrantargs(struct nlm_rqst *call)
+{
+	if (call->a_args.lock.oh.data != call->a_owner)
+		kfree(call->a_args.lock.oh.data);
 }
 
 /*
@@ -297,9 +358,9 @@ u32
 nlmsvc_lock(struct svc_rqst *rqstp, struct nlm_file *file,
 			struct nlm_lock *lock, int wait, struct nlm_cookie *cookie)
 {
-	struct file_lock	*conflock;
-	struct nlm_block	*block;
+	struct nlm_block	*block, *newblock = NULL;
 	int			error;
+	u32			ret;
 
 	dprintk("lockd: nlmsvc_lock(%s/%ld, ty=%d, pi=%d, %Ld-%Ld, bl=%d)\n",
 				file->f_file->f_dentry->d_inode->i_sb->s_id,
@@ -310,69 +371,65 @@ nlmsvc_lock(struct svc_rqst *rqstp, struct nlm_file *file,
 				wait);
 
 
-	/* Get existing block (in case client is busy-waiting) */
-	block = nlmsvc_lookup_block(file, lock, 0);
-
-	lock->fl.fl_flags |= FL_LOCKD;
-
+	lock->fl.fl_flags &= ~FL_SLEEP;
 again:
 	/* Lock file against concurrent access */
 	down(&file->f_sema);
+	/* Get existing block (in case client is busy-waiting) */
+	block = nlmsvc_lookup_block(file, lock);
+	if (block == NULL) {
+		if (newblock != NULL)
+			lock = &newblock->b_call->a_args.lock;
+	} else
+		lock = &block->b_call->a_args.lock;
 
-	if (!(conflock = posix_test_lock(file->f_file, &lock->fl))) {
-		error = posix_lock_file(file->f_file, &lock->fl);
+	error = posix_lock_file(file->f_file, &lock->fl);
+	lock->fl.fl_flags &= ~FL_SLEEP;
 
-		if (block)
-			nlmsvc_delete_block(block, 0);
-		up(&file->f_sema);
+	dprintk("lockd: posix_lock_file returned %d\n", error);
 
-		dprintk("lockd: posix_lock_file returned %d\n", -error);
-		switch(-error) {
+	switch(error) {
 		case 0:
-			return nlm_granted;
-		case EDEADLK:
-			return nlm_deadlock;
-		case EAGAIN:
-			return nlm_lck_denied;
+			ret = nlm_granted;
+			goto out;
+		case -EAGAIN:
+			break;
+		case -EDEADLK:
+			ret = nlm_deadlock;
+			goto out;
 		default:			/* includes ENOLCK */
-			return nlm_lck_denied_nolocks;
-		}
+			ret = nlm_lck_denied_nolocks;
+			goto out;
 	}
 
-	if (!wait) {
-		up(&file->f_sema);
-		return nlm_lck_denied;
-	}
+	ret = nlm_lck_denied;
+	if (!wait)
+		goto out;
 
-	if (posix_locks_deadlock(&lock->fl, conflock)) {
-		up(&file->f_sema);
-		return nlm_deadlock;
-	}
+	ret = nlm_lck_blocked;
+	if (block != NULL)
+		goto out;
 
 	/* If we don't have a block, create and initialize it. Then
 	 * retry because we may have slept in kmalloc. */
 	/* We have to release f_sema as nlmsvc_create_block may try to
 	 * to claim it while doing host garbage collection */
-	if (block == NULL) {
+	if (newblock == NULL) {
 		up(&file->f_sema);
 		dprintk("lockd: blocking on this lock (allocating).\n");
-		if (!(block = nlmsvc_create_block(rqstp, file, lock, cookie)))
+		if (!(newblock = nlmsvc_create_block(rqstp, file, lock, cookie)))
 			return nlm_lck_denied_nolocks;
 		goto again;
 	}
 
 	/* Append to list of blocked */
-	nlmsvc_insert_block(block, NLM_NEVER);
-
-	if (list_empty(&block->b_call.a_args.lock.fl.fl_block)) {
-		/* Now add block to block list of the conflicting lock
-		   if we haven't done so. */
-		dprintk("lockd: blocking on this lock.\n");
-		posix_block_lock(conflock, &block->b_call.a_args.lock.fl);
-	}
-
+	nlmsvc_insert_block(newblock, NLM_NEVER);
+out:
 	up(&file->f_sema);
-	return nlm_lck_blocked;
+	nlmsvc_release_block(newblock);
+	nlmsvc_release_block(block);
+	dprintk("lockd: nlmsvc_lock returned %u\n", ret);
+	return ret;
 }
 
 /*
@@ -382,8 +439,6 @@ u32
 nlmsvc_testlock(struct nlm_file *file, struct nlm_lock *lock,
 				       struct nlm_lock *conflock)
 {
-	struct file_lock	*fl;
-
 	dprintk("lockd: nlmsvc_testlock(%s/%ld, ty=%d, %Ld-%Ld)\n",
 				file->f_file->f_dentry->d_inode->i_sb->s_id,
 				file->f_file->f_dentry->d_inode->i_ino,
@@ -391,13 +446,14 @@ nlmsvc_testlock(struct nlm_file *file, struct nlm_lock *lock,
 				(long long)lock->fl.fl_start,
 				(long long)lock->fl.fl_end);
 
-	if ((fl = posix_test_lock(file->f_file, &lock->fl)) != NULL) {
+	if (posix_test_lock(file->f_file, &lock->fl, &conflock->fl)) {
 		dprintk("lockd: conflicting lock(ty=%d, %Ld-%Ld)\n",
-				fl->fl_type, (long long)fl->fl_start,
-				(long long)fl->fl_end);
+				conflock->fl.fl_type,
+				(long long)conflock->fl.fl_start,
+				(long long)conflock->fl.fl_end);
 		conflock->caller = "somehost";	/* FIXME */
 		conflock->oh.len = 0;		/* don't return OH info */
-		conflock->fl = *fl;
+		conflock->svid = conflock->fl.fl_pid;
 		return nlm_lck_denied;
 	}
 
@@ -453,9 +509,12 @@ nlmsvc_cancel_blocked(struct nlm_file *file, struct nlm_lock *lock)
 				(long long)lock->fl.fl_end);
 
 	down(&file->f_sema);
-	if ((block = nlmsvc_lookup_block(file, lock, 1)) != NULL)
-		status = nlmsvc_delete_block(block, 1);
+	block = nlmsvc_lookup_block(file, lock);
 	up(&file->f_sema);
+	if (block != NULL) {
+		status = nlmsvc_unlink_block(block);
+		nlmsvc_release_block(block);
+	}
 	return status ? nlm_lck_denied : nlm_granted;
 }
 
@@ -473,7 +532,7 @@ nlmsvc_notify_blocked(struct file_lock *fl)
 
 	dprintk("lockd: VFS unblock notification for block %p\n", fl);
 	for (bp = &nlm_blocked; (block = *bp) != 0; bp = &block->b_next) {
-		if (nlm_compare_locks(&block->b_call.a_args.lock.fl, fl)) {
+		if (nlm_compare_locks(&block->b_call->a_args.lock.fl, fl)) {
 			nlmsvc_insert_block(block, 0);
 			svc_wake_up(block->b_daemon);
 			return;
@@ -508,17 +567,13 @@ static void
 nlmsvc_grant_blocked(struct nlm_block *block)
 {
 	struct nlm_file		*file = block->b_file;
-	struct nlm_lock		*lock = &block->b_call.a_args.lock;
-	struct file_lock	*conflock;
+	struct nlm_lock		*lock = &block->b_call->a_args.lock;
 	int			error;
 
 	dprintk("lockd: grant blocked lock %p\n", block);
 
-	/* First thing is lock the file */
-	down(&file->f_sema);
-
 	/* Unlink block request from list */
-	nlmsvc_remove_block(block);
+	nlmsvc_unlink_block(block);
 
 	/* If b_granted is true this means we've been here before.
 	 * Just retry the grant callback, possibly refreshing the RPC
@@ -529,24 +584,21 @@ nlmsvc_grant_blocked(struct nlm_block *block)
 	}
 
 	/* Try the lock operation again */
-	if ((conflock = posix_test_lock(file->f_file, &lock->fl)) != NULL) {
-		/* Bummer, we blocked again */
+	lock->fl.fl_flags |= FL_SLEEP;
+	error = posix_lock_file(file->f_file, &lock->fl);
+	lock->fl.fl_flags &= ~FL_SLEEP;
+
+	switch (error) {
+	case 0:
+		break;
+	case -EAGAIN:
 		dprintk("lockd: lock still blocked\n");
 		nlmsvc_insert_block(block, NLM_NEVER);
-		posix_block_lock(conflock, &lock->fl);
-		up(&file->f_sema);
 		return;
-	}
-
-	/* Alright, no conflicting lock. Now lock it for real. If the
-	 * following yields an error, this is most probably due to low
-	 * memory. Retry the lock in a few seconds.
-	 */
-	if ((error = posix_lock_file(file->f_file, &lock->fl)) < 0) {
+	default:
 		printk(KERN_WARNING "lockd: unexpected error %d in %s!\n",
 				-error, __FUNCTION__);
 		nlmsvc_insert_block(block, 10 * HZ);
-		up(&file->f_sema);
 		return;
 	}
 
@@ -554,17 +606,15 @@ callback:
 	/* Lock was granted by VFS. */
 	dprintk("lockd: GRANTing blocked lock.\n");
 	block->b_granted = 1;
-	block->b_incall  = 1;
 
 	/* Schedule next grant callback in 30 seconds */
 	nlmsvc_insert_block(block, 30 * HZ);
 
 	/* Call the client */
-	nlm_get_host(block->b_call.a_host);
-	if (nlmsvc_async_call(&block->b_call, NLMPROC_GRANTED_MSG,
+	kref_get(&block->b_count);
+	if (nlm_async_call(block->b_call, NLMPROC_GRANTED_MSG,
 						&nlmsvc_grant_ops) < 0)
-		nlm_release_host(block->b_call.a_host);
-	up(&file->f_sema);
+		nlmsvc_release_block(block);
 }
 
 /*
@@ -578,20 +628,10 @@ callback:
 static void nlmsvc_grant_callback(struct rpc_task *task, void *data)
 {
 	struct nlm_rqst		*call = data;
-	struct nlm_block	*block;
+	struct nlm_block	*block = call->a_block;
 	unsigned long		timeout;
-	struct sockaddr_in	*peer_addr = RPC_PEERADDR(task->tk_client);
 
 	dprintk("lockd: GRANT_MSG RPC callback\n");
-	dprintk("callback: looking for cookie %s, host (%u.%u.%u.%u)\n",
-		nlmdbg_cookie2a(&call->a_args.cookie),
-		NIPQUAD(peer_addr->sin_addr.s_addr));
-	if (!(block = nlmsvc_find_block(&call->a_args.cookie, peer_addr))) {
-		dprintk("lockd: no block for cookie %s, host (%u.%u.%u.%u)\n",
-			nlmdbg_cookie2a(&call->a_args.cookie),
-			NIPQUAD(peer_addr->sin_addr.s_addr));
-		return;
-	}
 
 	/* Technically, we should down the file semaphore here. Since we
 	 * move the block towards the head of the queue only, no harm
@@ -608,13 +648,18 @@ static void nlmsvc_grant_callback(struct rpc_task *task, void *data)
 	}
 	nlmsvc_insert_block(block, timeout);
 	svc_wake_up(block->b_daemon);
-	block->b_incall = 0;
+}
 
-	nlm_release_host(call->a_host);
+void nlmsvc_grant_release(void *data)
+{
+	struct nlm_rqst		*call = data;
+
+	nlmsvc_release_block(call->a_block);
 }
 
 static const struct rpc_call_ops nlmsvc_grant_ops = {
 	.rpc_call_done = nlmsvc_grant_callback,
+	.rpc_release = nlmsvc_grant_release,
 };
 
 /*
@@ -634,25 +679,17 @@ nlmsvc_grant_reply(struct svc_rqst *rqstp, struct nlm_cookie *cookie, u32 status
 		return;
 	file = block->b_file;
 
-	file->f_count++;
-	down(&file->f_sema);
-	block = nlmsvc_find_block(cookie, &rqstp->rq_addr);
 	if (block) {
 		if (status == NLM_LCK_DENIED_GRACE_PERIOD) {
 			/* Try again in a couple of seconds */
 			nlmsvc_insert_block(block, 10 * HZ);
-			up(&file->f_sema);
 		} else {
 			/* Lock is now held by client, or has been rejected.
 			 * In both cases, the block should be removed. */
-			up(&file->f_sema);
-			if (status == NLM_LCK_GRANTED)
-				nlmsvc_delete_block(block, 0);
-			else
-				nlmsvc_delete_block(block, 1);
+			nlmsvc_unlink_block(block);
 		}
 	}
-	nlm_release_file(file);
+	nlmsvc_release_block(block);
 }
 
 /*
@@ -675,10 +712,12 @@ nlmsvc_retry_blocked(void)
 			break;
 		dprintk("nlmsvc_retry_blocked(%p, when=%ld, done=%d)\n",
 			block, block->b_when, block->b_done);
+		kref_get(&block->b_count);
 		if (block->b_done)
-			nlmsvc_delete_block(block, 0);
+			nlmsvc_unlink_block(block);
 		else
 			nlmsvc_grant_blocked(block);
+		nlmsvc_release_block(block);
 	}
 
 	if ((block = nlm_blocked) && block->b_when != NLM_NEVER)

fs/lockd/svcproc.c

@@ -22,10 +22,6 @@
 
 #define NLMDBG_FACILITY		NLMDBG_CLIENT
 
-static u32	nlmsvc_callback(struct svc_rqst *, u32, struct nlm_res *);
-
-static const struct rpc_call_ops nlmsvc_callback_ops;
-
 #ifdef CONFIG_LOCKD_V4
 static u32
 cast_to_nlm(u32 status, u32 vers)
@@ -261,84 +257,92 @@ nlmsvc_proc_granted(struct svc_rqst *rqstp, struct nlm_args *argp,
 	return rpc_success;
 }
 
+/*
+ * This is the generic lockd callback for async RPC calls
+ */
+static void nlmsvc_callback_exit(struct rpc_task *task, void *data)
+{
+	dprintk("lockd: %4d callback returned %d\n", task->tk_pid,
+			-task->tk_status);
+}
+
+static void nlmsvc_callback_release(void *data)
+{
+	nlm_release_call(data);
+}
+
+static const struct rpc_call_ops nlmsvc_callback_ops = {
+	.rpc_call_done = nlmsvc_callback_exit,
+	.rpc_release = nlmsvc_callback_release,
+};
+
 /*
  * `Async' versions of the above service routines. They aren't really,
  * because we send the callback before the reply proper. I hope this
  * doesn't break any clients.
  */
-static int
-nlmsvc_proc_test_msg(struct svc_rqst *rqstp, struct nlm_args *argp,
-					     void	     *resp)
+static int nlmsvc_callback(struct svc_rqst *rqstp, u32 proc, struct nlm_args *argp,
+		int (*func)(struct svc_rqst *, struct nlm_args *, struct nlm_res  *))
 {
-	struct nlm_res	res;
-	u32		stat;
+	struct nlm_host	*host;
+	struct nlm_rqst	*call;
+	int stat;
 
-	dprintk("lockd: TEST_MSG      called\n");
-	memset(&res, 0, sizeof(res));
+	host = nlmsvc_lookup_host(rqstp);
+	if (host == NULL)
+		return rpc_system_err;
 
-	if ((stat = nlmsvc_proc_test(rqstp, argp, &res)) == 0)
-		stat = nlmsvc_callback(rqstp, NLMPROC_TEST_RES, &res);
-	return stat;
+	call = nlm_alloc_call(host);
+	if (call == NULL)
+		return rpc_system_err;
+
+	stat = func(rqstp, argp, &call->a_res);
+	if (stat != 0) {
+		nlm_release_call(call);
+		return stat;
+	}
+
+	call->a_flags = RPC_TASK_ASYNC;
+	if (nlm_async_reply(call, proc, &nlmsvc_callback_ops) < 0)
+		return rpc_system_err;
+	return rpc_success;
 }
 
-static int
-nlmsvc_proc_lock_msg(struct svc_rqst *rqstp, struct nlm_args *argp,
+static int nlmsvc_proc_test_msg(struct svc_rqst *rqstp, struct nlm_args *argp,
 					     void	     *resp)
 {
-	struct nlm_res	res;
-	u32		stat;
+	dprintk("lockd: TEST_MSG      called\n");
+	return nlmsvc_callback(rqstp, NLMPROC_TEST_RES, argp, nlmsvc_proc_test);
+}
 
+static int nlmsvc_proc_lock_msg(struct svc_rqst *rqstp, struct nlm_args *argp,
+					     void	     *resp)
+{
 	dprintk("lockd: LOCK_MSG      called\n");
-	memset(&res, 0, sizeof(res));
-
-	if ((stat = nlmsvc_proc_lock(rqstp, argp, &res)) == 0)
-		stat = nlmsvc_callback(rqstp, NLMPROC_LOCK_RES, &res);
-	return stat;
+	return nlmsvc_callback(rqstp, NLMPROC_LOCK_RES, argp, nlmsvc_proc_lock);
 }
 
-static int
-nlmsvc_proc_cancel_msg(struct svc_rqst *rqstp, struct nlm_args *argp,
+static int nlmsvc_proc_cancel_msg(struct svc_rqst *rqstp, struct nlm_args *argp,
 					       void	       *resp)
 {
-	struct nlm_res	res;
-	u32		stat;
-
 	dprintk("lockd: CANCEL_MSG    called\n");
-	memset(&res, 0, sizeof(res));
-
-	if ((stat = nlmsvc_proc_cancel(rqstp, argp, &res)) == 0)
-		stat = nlmsvc_callback(rqstp, NLMPROC_CANCEL_RES, &res);
-	return stat;
+	return nlmsvc_callback(rqstp, NLMPROC_CANCEL_RES, argp, nlmsvc_proc_cancel);
 }
 
 static int
 nlmsvc_proc_unlock_msg(struct svc_rqst *rqstp, struct nlm_args *argp,
                                                void            *resp)
 {
-	struct nlm_res	res;
-	u32		stat;
-
 	dprintk("lockd: UNLOCK_MSG    called\n");
-	memset(&res, 0, sizeof(res));
-
-	if ((stat = nlmsvc_proc_unlock(rqstp, argp, &res)) == 0)
-		stat = nlmsvc_callback(rqstp, NLMPROC_UNLOCK_RES, &res);
-	return stat;
+	return nlmsvc_callback(rqstp, NLMPROC_UNLOCK_RES, argp, nlmsvc_proc_unlock);
 }
 
 static int
 nlmsvc_proc_granted_msg(struct svc_rqst *rqstp, struct nlm_args *argp,
                                                 void            *resp)
 {
-	struct nlm_res	res;
-	u32		stat;
-
 	dprintk("lockd: GRANTED_MSG   called\n");
-	memset(&res, 0, sizeof(res));
-
-	if ((stat = nlmsvc_proc_granted(rqstp, argp, &res)) == 0)
-		stat = nlmsvc_callback(rqstp, NLMPROC_GRANTED_RES, &res);
-	return stat;
+	return nlmsvc_callback(rqstp, NLMPROC_GRANTED_RES, argp, nlmsvc_proc_granted);
 }
 
 /*
@@ -496,55 +500,6 @@ nlmsvc_proc_granted_res(struct svc_rqst *rqstp, struct nlm_res  *argp,
 	return rpc_success;
 }
 
-/*
- * This is the generic lockd callback for async RPC calls
- */
-static u32
-nlmsvc_callback(struct svc_rqst *rqstp, u32 proc, struct nlm_res *resp)
-{
-	struct nlm_host	*host;
-	struct nlm_rqst	*call;
-
-	if (!(call = nlmclnt_alloc_call()))
-		return rpc_system_err;
-
-	host = nlmclnt_lookup_host(&rqstp->rq_addr,
-				rqstp->rq_prot, rqstp->rq_vers);
-	if (!host) {
-		kfree(call);
-		return rpc_system_err;
-	}
-
-	call->a_flags = RPC_TASK_ASYNC;
-	call->a_host  = host;
-	memcpy(&call->a_args, resp, sizeof(*resp));
-
-	if (nlmsvc_async_call(call, proc, &nlmsvc_callback_ops) < 0)
-		goto error;
-
-	return rpc_success;
- error:
-	nlm_release_host(host);
-	kfree(call);
-	return rpc_system_err;
-}
-
-static void nlmsvc_callback_exit(struct rpc_task *task, void *data)
-{
-	struct nlm_rqst	*call = data;
-
-	if (task->tk_status < 0) {
-		dprintk("lockd: %4d callback failed (errno = %d)\n",
-					task->tk_pid, -task->tk_status);
-	}
-	nlm_release_host(call->a_host);
-	kfree(call);
-}
-
-static const struct rpc_call_ops nlmsvc_callback_ops = {
-	.rpc_call_done = nlmsvc_callback_exit,
-};
-
 /*
  * NLM Server procedures.
  */

fs/lockd/svcshare.c

@@ -88,7 +88,7 @@ nlmsvc_unshare_file(struct nlm_host *host, struct nlm_file *file,
  * Traverse all shares for a given file (and host).
  * NLM_ACT_CHECK is handled by nlmsvc_inspect_file.
  */
-int
+void
 nlmsvc_traverse_shares(struct nlm_host *host, struct nlm_file *file, int action)
 {
 	struct nlm_share	*share, **shpp;
@@ -106,6 +106,4 @@ nlmsvc_traverse_shares(struct nlm_host *host, struct nlm_file *file, int action)
 		}
 		shpp = &share->s_next;
 	}
-
-	return 0;
 }

fs/lockd/svcsubs.c

@@ -182,7 +182,7 @@ nlm_traverse_locks(struct nlm_host *host, struct nlm_file *file, int action)
 again:
 	file->f_locks = 0;
 	for (fl = inode->i_flock; fl; fl = fl->fl_next) {
-		if (!(fl->fl_flags & FL_LOCKD))
+		if (fl->fl_lmops != &nlmsvc_lock_operations)
 			continue;
 
 		/* update current lock count */
@@ -224,9 +224,8 @@ nlm_inspect_file(struct nlm_host *host, struct nlm_file *file, int action)
 		if (file->f_count || file->f_blocks || file->f_shares)
 			return 1;
 	} else {
-		if (nlmsvc_traverse_blocks(host, file, action)
-		 || nlmsvc_traverse_shares(host, file, action))
-			return 1;
+		nlmsvc_traverse_blocks(host, file, action);
+		nlmsvc_traverse_shares(host, file, action);
 	}
 	return nlm_traverse_locks(host, file, action);
 }

fs/lockd/xdr.c

@@ -131,10 +131,11 @@ nlm_decode_lock(u32 *p, struct nlm_lock *lock)
 	 || !(p = nlm_decode_fh(p, &lock->fh))
 	 || !(p = nlm_decode_oh(p, &lock->oh)))
 		return NULL;
+	lock->svid  = ntohl(*p++);
 
 	locks_init_lock(fl);
 	fl->fl_owner = current->files;
-	fl->fl_pid   = ntohl(*p++);
+	fl->fl_pid   = (pid_t)lock->svid;
 	fl->fl_flags = FL_POSIX;
 	fl->fl_type  = F_RDLCK;		/* as good as anything else */
 	start = ntohl(*p++);
@@ -174,7 +175,7 @@ nlm_encode_lock(u32 *p, struct nlm_lock *lock)
 	else
 		len = loff_t_to_s32(fl->fl_end - fl->fl_start + 1);
 
-	*p++ = htonl(fl->fl_pid);
+	*p++ = htonl(lock->svid);
 	*p++ = htonl(start);
 	*p++ = htonl(len);
 
@@ -197,7 +198,7 @@ nlm_encode_testres(u32 *p, struct nlm_res *resp)
 		struct file_lock	*fl = &resp->lock.fl;
 
 		*p++ = (fl->fl_type == F_RDLCK)? xdr_zero : xdr_one;
-		*p++ = htonl(fl->fl_pid);
+		*p++ = htonl(resp->lock.svid);
 
 		/* Encode owner handle. */
 		if (!(p = xdr_encode_netobj(p, &resp->lock.oh)))
@@ -298,7 +299,8 @@ nlmsvc_decode_shareargs(struct svc_rqst *rqstp, u32 *p, nlm_args *argp)
 
 	memset(lock, 0, sizeof(*lock));
 	locks_init_lock(&lock->fl);
-	lock->fl.fl_pid = ~(u32) 0;
+	lock->svid = ~(u32) 0;
+	lock->fl.fl_pid = (pid_t)lock->svid;
 
 	if (!(p = nlm_decode_cookie(p, &argp->cookie))
 	 || !(p = xdr_decode_string_inplace(p, &lock->caller,
@@ -415,7 +417,8 @@ nlmclt_decode_testres(struct rpc_rqst *req, u32 *p, struct nlm_res *resp)
 		memset(&resp->lock, 0, sizeof(resp->lock));
 		locks_init_lock(fl);
 		excl = ntohl(*p++);
-		fl->fl_pid = ntohl(*p++);
+		resp->lock.svid = ntohl(*p++);
+		fl->fl_pid = (pid_t)resp->lock.svid;
 		if (!(p = nlm_decode_oh(p, &resp->lock.oh)))
 			return -EIO;
 
@@ -543,7 +546,9 @@ nlmclt_decode_res(struct rpc_rqst *req, u32 *p, struct nlm_res *resp)
 	.p_proc      = NLMPROC_##proc,					\
 	.p_encode    = (kxdrproc_t) nlmclt_encode_##argtype,		\
 	.p_decode    = (kxdrproc_t) nlmclt_decode_##restype,		\
-	.p_bufsiz    = MAX(NLM_##argtype##_sz, NLM_##restype##_sz) << 2	\
+	.p_bufsiz    = MAX(NLM_##argtype##_sz, NLM_##restype##_sz) << 2,	\
+	.p_statidx   = NLMPROC_##proc,					\
+	.p_name      = #proc,						\
 	}
 
 static struct rpc_procinfo	nlm_procedures[] = {

fs/lockd/xdr4.c

@@ -130,10 +130,11 @@ nlm4_decode_lock(u32 *p, struct nlm_lock *lock)
 	 || !(p = nlm4_decode_fh(p, &lock->fh))
 	 || !(p = nlm4_decode_oh(p, &lock->oh)))
 		return NULL;
+	lock->svid  = ntohl(*p++);
 
 	locks_init_lock(fl);
 	fl->fl_owner = current->files;
-	fl->fl_pid   = ntohl(*p++);
+	fl->fl_pid   = (pid_t)lock->svid;
 	fl->fl_flags = FL_POSIX;
 	fl->fl_type  = F_RDLCK;		/* as good as anything else */
 	p = xdr_decode_hyper(p, &start);
@@ -167,7 +168,7 @@ nlm4_encode_lock(u32 *p, struct nlm_lock *lock)
 	 || (fl->fl_end > NLM4_OFFSET_MAX && fl->fl_end != OFFSET_MAX))
 		return NULL;
 
-	*p++ = htonl(fl->fl_pid);
+	*p++ = htonl(lock->svid);
 
 	start = loff_t_to_s64(fl->fl_start);
 	if (fl->fl_end == OFFSET_MAX)
@@ -198,7 +199,7 @@ nlm4_encode_testres(u32 *p, struct nlm_res *resp)
 		struct file_lock	*fl = &resp->lock.fl;
 
 		*p++ = (fl->fl_type == F_RDLCK)? xdr_zero : xdr_one;
-		*p++ = htonl(fl->fl_pid);
+		*p++ = htonl(resp->lock.svid);
 
 		/* Encode owner handle. */
 		if (!(p = xdr_encode_netobj(p, &resp->lock.oh)))
@@ -212,8 +213,8 @@ nlm4_encode_testres(u32 *p, struct nlm_res *resp)
 		
 		p = xdr_encode_hyper(p, start);
 		p = xdr_encode_hyper(p, len);
-		dprintk("xdr: encode_testres (status %d pid %d type %d start %Ld end %Ld)\n",
-			resp->status, fl->fl_pid, fl->fl_type,
+		dprintk("xdr: encode_testres (status %u pid %d type %d start %Ld end %Ld)\n",
+			resp->status, (int)resp->lock.svid, fl->fl_type,
 			(long long)fl->fl_start,  (long long)fl->fl_end);
 	}
 
@@ -303,7 +304,8 @@ nlm4svc_decode_shareargs(struct svc_rqst *rqstp, u32 *p, nlm_args *argp)
 
 	memset(lock, 0, sizeof(*lock));
 	locks_init_lock(&lock->fl);
-	lock->fl.fl_pid = ~(u32) 0;
+	lock->svid = ~(u32) 0;
+	lock->fl.fl_pid = (pid_t)lock->svid;
 
 	if (!(p = nlm4_decode_cookie(p, &argp->cookie))
 	 || !(p = xdr_decode_string_inplace(p, &lock->caller,
@@ -420,7 +422,8 @@ nlm4clt_decode_testres(struct rpc_rqst *req, u32 *p, struct nlm_res *resp)
 		memset(&resp->lock, 0, sizeof(resp->lock));
 		locks_init_lock(fl);
 		excl = ntohl(*p++);
-		fl->fl_pid = ntohl(*p++);
+		resp->lock.svid = ntohl(*p++);
+		fl->fl_pid = (pid_t)resp->lock.svid;
 		if (!(p = nlm4_decode_oh(p, &resp->lock.oh)))
 			return -EIO;
 
@@ -548,7 +551,9 @@ nlm4clt_decode_res(struct rpc_rqst *req, u32 *p, struct nlm_res *resp)
 	.p_proc      = NLMPROC_##proc,					\
 	.p_encode    = (kxdrproc_t) nlm4clt_encode_##argtype,		\
 	.p_decode    = (kxdrproc_t) nlm4clt_decode_##restype,		\
-	.p_bufsiz    = MAX(NLM4_##argtype##_sz, NLM4_##restype##_sz) << 2	\
+	.p_bufsiz    = MAX(NLM4_##argtype##_sz, NLM4_##restype##_sz) << 2,	\
+	.p_statidx   = NLMPROC_##proc,					\
+	.p_name      = #proc,						\
 	}
 
 static struct rpc_procinfo	nlm4_procedures[] = {

fs/locks.c

@@ -139,10 +139,7 @@ int lease_break_time = 45;
 #define for_each_lock(inode, lockp) \
 	for (lockp = &inode->i_flock; *lockp != NULL; lockp = &(*lockp)->fl_next)
 
-LIST_HEAD(file_lock_list);
-
-EXPORT_SYMBOL(file_lock_list);
-
+static LIST_HEAD(file_lock_list);
 static LIST_HEAD(blocked_list);
 
 static kmem_cache_t *filelock_cache;
@@ -153,6 +150,21 @@ static struct file_lock *locks_alloc_lock(void)
 	return kmem_cache_alloc(filelock_cache, SLAB_KERNEL);
 }
 
+static void locks_release_private(struct file_lock *fl)
+{
+	if (fl->fl_ops) {
+		if (fl->fl_ops->fl_release_private)
+			fl->fl_ops->fl_release_private(fl);
+		fl->fl_ops = NULL;
+	}
+	if (fl->fl_lmops) {
+		if (fl->fl_lmops->fl_release_private)
+			fl->fl_lmops->fl_release_private(fl);
+		fl->fl_lmops = NULL;
+	}
+
+}
+
 /* Free a lock which is not in use. */
 static void locks_free_lock(struct file_lock *fl)
 {
@@ -169,18 +181,7 @@ static void locks_free_lock(struct file_lock *fl)
 	if (!list_empty(&fl->fl_link))
 		panic("Attempting to free lock on active lock list");
 
-	if (fl->fl_ops) {
-		if (fl->fl_ops->fl_release_private)
-			fl->fl_ops->fl_release_private(fl);
-		fl->fl_ops = NULL;
-	}
-
-	if (fl->fl_lmops) {
-		if (fl->fl_lmops->fl_release_private)
-			fl->fl_lmops->fl_release_private(fl);
-		fl->fl_lmops = NULL;
-	}
-
+	locks_release_private(fl);
 	kmem_cache_free(filelock_cache, fl);
 }
 
@@ -218,24 +219,46 @@ static void init_once(void *foo, kmem_cache_t *cache, unsigned long flags)
 	locks_init_lock(lock);
 }
 
+static void locks_copy_private(struct file_lock *new, struct file_lock *fl)
+{
+	if (fl->fl_ops) {
+		if (fl->fl_ops->fl_copy_lock)
+			fl->fl_ops->fl_copy_lock(new, fl);
+		new->fl_ops = fl->fl_ops;
+	}
+	if (fl->fl_lmops) {
+		if (fl->fl_lmops->fl_copy_lock)
+			fl->fl_lmops->fl_copy_lock(new, fl);
+		new->fl_lmops = fl->fl_lmops;
+	}
+}
+
 /*
  * Initialize a new lock from an existing file_lock structure.
  */
-void locks_copy_lock(struct file_lock *new, struct file_lock *fl)
+static void __locks_copy_lock(struct file_lock *new, const struct file_lock *fl)
 {
 	new->fl_owner = fl->fl_owner;
 	new->fl_pid = fl->fl_pid;
-	new->fl_file = fl->fl_file;
+	new->fl_file = NULL;
 	new->fl_flags = fl->fl_flags;
 	new->fl_type = fl->fl_type;
 	new->fl_start = fl->fl_start;
 	new->fl_end = fl->fl_end;
+	new->fl_ops = NULL;
+	new->fl_lmops = NULL;
+}
+
+void locks_copy_lock(struct file_lock *new, struct file_lock *fl)
+{
+	locks_release_private(new);
+
+	__locks_copy_lock(new, fl);
+	new->fl_file = fl->fl_file;
 	new->fl_ops = fl->fl_ops;
 	new->fl_lmops = fl->fl_lmops;
-	if (fl->fl_ops && fl->fl_ops->fl_copy_lock)
-		fl->fl_ops->fl_copy_lock(new, fl);
-	if (fl->fl_lmops && fl->fl_lmops->fl_copy_lock)
-		fl->fl_lmops->fl_copy_lock(new, fl);
+
+	locks_copy_private(new, fl);
 }
 
 EXPORT_SYMBOL(locks_copy_lock);
@@ -654,8 +677,9 @@ static int locks_block_on_timeout(struct file_lock *blocker, struct file_lock *w
 	return result;
 }
 
-struct file_lock *
-posix_test_lock(struct file *filp, struct file_lock *fl)
+int
+posix_test_lock(struct file *filp, struct file_lock *fl,
+		struct file_lock *conflock)
 {
 	struct file_lock *cfl;
 
@@ -666,9 +690,13 @@ posix_test_lock(struct file *filp, struct file_lock *fl)
 		if (posix_locks_conflict(cfl, fl))
 			break;
 	}
+	if (cfl) {
+		__locks_copy_lock(conflock, cfl);
+		unlock_kernel();
+		return 1;
+	}
 	unlock_kernel();
-
-	return (cfl);
+	return 0;
 }
 
 EXPORT_SYMBOL(posix_test_lock);
@@ -904,7 +932,8 @@ static int __posix_lock_file(struct inode *inode, struct file_lock *request)
 				fl->fl_start = request->fl_start;
 				fl->fl_end = request->fl_end;
 				fl->fl_type = request->fl_type;
-				fl->fl_u = request->fl_u;
+				locks_release_private(fl);
+				locks_copy_private(fl, request);
 				request = fl;
 				added = 1;
 			}
@@ -1544,7 +1573,7 @@ asmlinkage long sys_flock(unsigned int fd, unsigned int cmd)
  */
 int fcntl_getlk(struct file *filp, struct flock __user *l)
 {
-	struct file_lock *fl, file_lock;
+	struct file_lock *fl, cfl, file_lock;
 	struct flock flock;
 	int error;
 
@@ -1568,7 +1597,7 @@ int fcntl_getlk(struct file *filp, struct flock __user *l)
 		else
 		  fl = (file_lock.fl_type == F_UNLCK ? NULL : &file_lock);
 	} else {
-		fl = posix_test_lock(filp, &file_lock);
+		fl = (posix_test_lock(filp, &file_lock, &cfl) ? &cfl : NULL);
 	}
  
 	flock.l_type = F_UNLCK;
@@ -1698,7 +1727,7 @@ out:
  */
 int fcntl_getlk64(struct file *filp, struct flock64 __user *l)
 {
-	struct file_lock *fl, file_lock;
+	struct file_lock *fl, cfl, file_lock;
 	struct flock64 flock;
 	int error;
 
@@ -1722,7 +1751,7 @@ int fcntl_getlk64(struct file *filp, struct flock64 __user *l)
 		else
 		  fl = (file_lock.fl_type == F_UNLCK ? NULL : &file_lock);
 	} else {
-		fl = posix_test_lock(filp, &file_lock);
+		fl = (posix_test_lock(filp, &file_lock, &cfl) ? &cfl : NULL);
 	}
  
 	flock.l_type = F_UNLCK;
@@ -1935,21 +1964,6 @@ void locks_remove_flock(struct file *filp)
 	unlock_kernel();
 }
 
-/**
- *	posix_block_lock - blocks waiting for a file lock
- *	@blocker: the lock which is blocking
- *	@waiter: the lock which conflicts and has to wait
- *
- * lockd needs to block waiting for locks.
- */
-void
-posix_block_lock(struct file_lock *blocker, struct file_lock *waiter)
-{
-	locks_insert_block(blocker, waiter);
-}
-
-EXPORT_SYMBOL(posix_block_lock);
-
 /**
  *	posix_unblock_lock - stop waiting for a file lock
  *      @filp:   how the file was opened

fs/namespace.c

@@ -399,6 +399,44 @@ struct seq_operations mounts_op = {
 	.show	= show_vfsmnt
 };
 
+static int show_vfsstat(struct seq_file *m, void *v)
+{
+	struct vfsmount *mnt = v;
+	int err = 0;
+
+	/* device */
+	if (mnt->mnt_devname) {
+		seq_puts(m, "device ");
+		mangle(m, mnt->mnt_devname);
+	} else
+		seq_puts(m, "no device");
+
+	/* mount point */
+	seq_puts(m, " mounted on ");
+	seq_path(m, mnt, mnt->mnt_root, " \t\n\\");
+	seq_putc(m, ' ');
+
+	/* file system type */
+	seq_puts(m, "with fstype ");
+	mangle(m, mnt->mnt_sb->s_type->name);
+
+	/* optional statistics */
+	if (mnt->mnt_sb->s_op->show_stats) {
+		seq_putc(m, ' ');
+		err = mnt->mnt_sb->s_op->show_stats(m, mnt);
+	}
+
+	seq_putc(m, '\n');
+	return err;
+}
+
+struct seq_operations mountstats_op = {
+	.start	= m_start,
+	.next	= m_next,
+	.stop	= m_stop,
+	.show	= show_vfsstat,
+};
+
 /**
  * may_umount_tree - check if a mount tree is busy
  * @mnt: root of mount tree

fs/nfs/callback.c

@@ -55,7 +55,12 @@ static void nfs_callback_svc(struct svc_rqst *rqstp)
 
 	complete(&nfs_callback_info.started);
 
-	while (nfs_callback_info.users != 0 || !signalled()) {
+	for(;;) {
+		if (signalled()) {
+			if (nfs_callback_info.users == 0)
+				break;
+			flush_signals(current);
+		}
 		/*
 		 * Listen for a request on the socket
 		 */
@@ -73,6 +78,7 @@ static void nfs_callback_svc(struct svc_rqst *rqstp)
 		svc_process(serv, rqstp);
 	}
 
+	svc_exit_thread(rqstp);
 	nfs_callback_info.pid = 0;
 	complete(&nfs_callback_info.stopped);
 	unlock_kernel();
@@ -134,11 +140,13 @@ int nfs_callback_down(void)
 
 	lock_kernel();
 	down(&nfs_callback_sema);
-	if (--nfs_callback_info.users || nfs_callback_info.pid == 0)
-		goto out;
-	kill_proc(nfs_callback_info.pid, SIGKILL, 1);
-	wait_for_completion(&nfs_callback_info.stopped);
-out:
+	nfs_callback_info.users--;
+	do {
+		if (nfs_callback_info.users != 0 || nfs_callback_info.pid == 0)
+			break;
+		if (kill_proc(nfs_callback_info.pid, SIGKILL, 1) < 0)
+			break;
+	} while (wait_for_completion_timeout(&nfs_callback_info.stopped, 5*HZ) == 0);
 	up(&nfs_callback_sema);
 	unlock_kernel();
 	return ret;

fs/nfs/callback_xdr.c

@@ -330,7 +330,7 @@ static unsigned encode_op_hdr(struct xdr_stream *xdr, uint32_t op, uint32_t res)
 
 static unsigned encode_getattr_res(struct svc_rqst *rqstp, struct xdr_stream *xdr, const struct cb_getattrres *res)
 {
-	uint32_t *savep;
+	uint32_t *savep = NULL;
 	unsigned status = res->status;
 	
 	if (unlikely(status != 0))
@@ -358,23 +358,26 @@ static unsigned process_op(struct svc_rqst *rqstp,
 		struct xdr_stream *xdr_in, void *argp,
 		struct xdr_stream *xdr_out, void *resp)
 {
-	struct callback_op *op;
-	unsigned int op_nr;
+	struct callback_op *op = &callback_ops[0];
+	unsigned int op_nr = OP_CB_ILLEGAL;
 	unsigned int status = 0;
 	long maxlen;
 	unsigned res;
 
 	dprintk("%s: start\n", __FUNCTION__);
 	status = decode_op_hdr(xdr_in, &op_nr);
-	if (unlikely(status != 0)) {
-		op_nr = OP_CB_ILLEGAL;
-		op = &callback_ops[0];
-	} else if (unlikely(op_nr != OP_CB_GETATTR && op_nr != OP_CB_RECALL)) {
-		op_nr = OP_CB_ILLEGAL;
-		op = &callback_ops[0];
-		status = htonl(NFS4ERR_OP_ILLEGAL);
-	} else
-		op = &callback_ops[op_nr];
+	if (likely(status == 0)) {
+		switch (op_nr) {
+			case OP_CB_GETATTR:
+			case OP_CB_RECALL:
+				op = &callback_ops[op_nr];
+				break;
+			default:
+				op_nr = OP_CB_ILLEGAL;
+				op = &callback_ops[0];
+				status = htonl(NFS4ERR_OP_ILLEGAL);
+		}
+	}
 
 	maxlen = xdr_out->end - xdr_out->p;
 	if (maxlen > 0 && maxlen < PAGE_SIZE) {
@@ -416,6 +419,7 @@ static int nfs4_callback_compound(struct svc_rqst *rqstp, void *argp, void *resp
 	decode_compound_hdr_arg(&xdr_in, &hdr_arg);
 	hdr_res.taglen = hdr_arg.taglen;
 	hdr_res.tag = hdr_arg.tag;
+	hdr_res.nops = NULL;
 	encode_compound_hdr_res(&xdr_out, &hdr_res);
 
 	for (;;) {

fs/nfs/delegation.c

@@ -421,3 +421,22 @@ void nfs_delegation_reap_unclaimed(struct nfs4_client *clp)
 		nfs_free_delegation(delegation);
 	}
 }
+
+int nfs4_copy_delegation_stateid(nfs4_stateid *dst, struct inode *inode)
+{
+	struct nfs4_client *clp = NFS_SERVER(inode)->nfs4_state;
+	struct nfs_inode *nfsi = NFS_I(inode);
+	struct nfs_delegation *delegation;
+	int res = 0;
+
+	if (nfsi->delegation_state == 0)
+		return 0;
+	spin_lock(&clp->cl_lock);
+	delegation = nfsi->delegation;
+	if (delegation != NULL) {
+		memcpy(dst->data, delegation->stateid.data, sizeof(dst->data));
+		res = 1;
+	}
+	spin_unlock(&clp->cl_lock);
+	return res;
+}

fs/nfs/delegation.h

@@ -41,6 +41,7 @@ void nfs_delegation_reap_unclaimed(struct nfs4_client *clp);
 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid);
 int nfs4_open_delegation_recall(struct dentry *dentry, struct nfs4_state *state);
 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl);
+int nfs4_copy_delegation_stateid(nfs4_stateid *dst, struct inode *inode);
 
 static inline int nfs_have_delegation(struct inode *inode, int flags)
 {

fs/nfs/dir.c

@@ -34,6 +34,7 @@
 
 #include "nfs4_fs.h"
 #include "delegation.h"
+#include "iostat.h"
 
 #define NFS_PARANOIA 1
 /* #define NFS_DEBUG_VERBOSE 1 */
@@ -129,6 +130,9 @@ nfs_opendir(struct inode *inode, struct file *filp)
 {
 	int res = 0;
 
+	dfprintk(VFS, "NFS: opendir(%s/%ld)\n",
+			inode->i_sb->s_id, inode->i_ino);
+
 	lock_kernel();
 	/* Call generic open code in order to cache credentials */
 	if (!res)
@@ -172,7 +176,9 @@ int nfs_readdir_filler(nfs_readdir_descriptor_t *desc, struct page *page)
 	unsigned long	timestamp;
 	int		error;
 
-	dfprintk(VFS, "NFS: nfs_readdir_filler() reading cookie %Lu into page %lu.\n", (long long)desc->entry->cookie, page->index);
+	dfprintk(DIRCACHE, "NFS: %s: reading cookie %Lu into page %lu\n",
+			__FUNCTION__, (long long)desc->entry->cookie,
+			page->index);
 
  again:
 	timestamp = jiffies;
@@ -244,7 +250,8 @@ int find_dirent(nfs_readdir_descriptor_t *desc)
 			status;
 
 	while((status = dir_decode(desc)) == 0) {
-		dfprintk(VFS, "NFS: found cookie %Lu\n", (unsigned long long)entry->cookie);
+		dfprintk(DIRCACHE, "NFS: %s: examining cookie %Lu\n",
+				__FUNCTION__, (unsigned long long)entry->cookie);
 		if (entry->prev_cookie == *desc->dir_cookie)
 			break;
 		if (loop_count++ > 200) {
@@ -252,7 +259,6 @@ int find_dirent(nfs_readdir_descriptor_t *desc)
 			schedule();
 		}
 	}
-	dfprintk(VFS, "NFS: find_dirent() returns %d\n", status);
 	return status;
 }
 
@@ -276,7 +282,8 @@ int find_dirent_index(nfs_readdir_descriptor_t *desc)
 		if (status)
 			break;
 
-		dfprintk(VFS, "NFS: found cookie %Lu at index %Ld\n", (unsigned long long)entry->cookie, desc->current_index);
+		dfprintk(DIRCACHE, "NFS: found cookie %Lu at index %Ld\n",
+				(unsigned long long)entry->cookie, desc->current_index);
 
 		if (desc->file->f_pos == desc->current_index) {
 			*desc->dir_cookie = entry->cookie;
@@ -288,7 +295,6 @@ int find_dirent_index(nfs_readdir_descriptor_t *desc)
 			schedule();
 		}
 	}
-	dfprintk(VFS, "NFS: find_dirent_index() returns %d\n", status);
 	return status;
 }
 
@@ -303,7 +309,9 @@ int find_dirent_page(nfs_readdir_descriptor_t *desc)
 	struct page	*page;
 	int		status;
 
-	dfprintk(VFS, "NFS: find_dirent_page() searching directory page %ld\n", desc->page_index);
+	dfprintk(DIRCACHE, "NFS: %s: searching page %ld for target %Lu\n",
+			__FUNCTION__, desc->page_index,
+			(long long) *desc->dir_cookie);
 
 	page = read_cache_page(inode->i_mapping, desc->page_index,
 			       (filler_t *)nfs_readdir_filler, desc);
@@ -324,7 +332,7 @@ int find_dirent_page(nfs_readdir_descriptor_t *desc)
 	if (status < 0)
 		dir_page_release(desc);
  out:
-	dfprintk(VFS, "NFS: find_dirent_page() returns %d\n", status);
+	dfprintk(DIRCACHE, "NFS: %s: returns %d\n", __FUNCTION__, status);
 	return status;
  read_error:
 	page_cache_release(page);
@@ -346,13 +354,15 @@ int readdir_search_pagecache(nfs_readdir_descriptor_t *desc)
 
 	/* Always search-by-index from the beginning of the cache */
 	if (*desc->dir_cookie == 0) {
-		dfprintk(VFS, "NFS: readdir_search_pagecache() searching for offset %Ld\n", (long long)desc->file->f_pos);
+		dfprintk(DIRCACHE, "NFS: readdir_search_pagecache() searching for offset %Ld\n",
+				(long long)desc->file->f_pos);
 		desc->page_index = 0;
 		desc->entry->cookie = desc->entry->prev_cookie = 0;
 		desc->entry->eof = 0;
 		desc->current_index = 0;
 	} else
-		dfprintk(VFS, "NFS: readdir_search_pagecache() searching for cookie %Lu\n", (unsigned long long)*desc->dir_cookie);
+		dfprintk(DIRCACHE, "NFS: readdir_search_pagecache() searching for cookie %Lu\n",
+				(unsigned long long)*desc->dir_cookie);
 
 	for (;;) {
 		res = find_dirent_page(desc);
@@ -365,7 +375,8 @@ int readdir_search_pagecache(nfs_readdir_descriptor_t *desc)
 			schedule();
 		}
 	}
-	dfprintk(VFS, "NFS: readdir_search_pagecache() returned %d\n", res);
+
+	dfprintk(DIRCACHE, "NFS: %s: returns %d\n", __FUNCTION__, res);
 	return res;
 }
 
@@ -390,7 +401,8 @@ int nfs_do_filldir(nfs_readdir_descriptor_t *desc, void *dirent,
 	int		loop_count = 0,
 			res;
 
-	dfprintk(VFS, "NFS: nfs_do_filldir() filling starting @ cookie %Lu\n", (long long)entry->cookie);
+	dfprintk(DIRCACHE, "NFS: nfs_do_filldir() filling starting @ cookie %Lu\n",
+			(unsigned long long)entry->cookie);
 
 	for(;;) {
 		unsigned d_type = DT_UNKNOWN;
@@ -427,7 +439,8 @@ int nfs_do_filldir(nfs_readdir_descriptor_t *desc, void *dirent,
 	dir_page_release(desc);
 	if (dentry != NULL)
 		dput(dentry);
-	dfprintk(VFS, "NFS: nfs_do_filldir() filling ended @ cookie %Lu; returning = %d\n", (unsigned long long)*desc->dir_cookie, res);
+	dfprintk(DIRCACHE, "NFS: nfs_do_filldir() filling ended @ cookie %Lu; returning = %d\n",
+			(unsigned long long)*desc->dir_cookie, res);
 	return res;
 }
 
@@ -453,7 +466,8 @@ int uncached_readdir(nfs_readdir_descriptor_t *desc, void *dirent,
 	struct page	*page = NULL;
 	int		status;
 
-	dfprintk(VFS, "NFS: uncached_readdir() searching for cookie %Lu\n", (unsigned long long)*desc->dir_cookie);
+	dfprintk(DIRCACHE, "NFS: uncached_readdir() searching for cookie %Lu\n",
+			(unsigned long long)*desc->dir_cookie);
 
 	page = alloc_page(GFP_HIGHUSER);
 	if (!page) {
@@ -485,7 +499,8 @@ int uncached_readdir(nfs_readdir_descriptor_t *desc, void *dirent,
 	desc->entry->cookie = desc->entry->prev_cookie = 0;
 	desc->entry->eof = 0;
  out:
-	dfprintk(VFS, "NFS: uncached_readdir() returns %d\n", status);
+	dfprintk(DIRCACHE, "NFS: %s: returns %d\n",
+			__FUNCTION__, status);
 	return status;
  out_release:
 	dir_page_release(desc);
@@ -507,6 +522,11 @@ static int nfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
 	struct nfs_fattr fattr;
 	long		res;
 
+	dfprintk(VFS, "NFS: readdir(%s/%s) starting at cookie %Lu\n",
+			dentry->d_parent->d_name.name, dentry->d_name.name,
+			(long long)filp->f_pos);
+	nfs_inc_stats(inode, NFSIOS_VFSGETDENTS);
+
 	lock_kernel();
 
 	res = nfs_revalidate_inode(NFS_SERVER(inode), inode);
@@ -566,9 +586,12 @@ static int nfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
 		}
 	}
 	unlock_kernel();
-	if (res < 0)
-		return res;
-	return 0;
+	if (res > 0)
+		res = 0;
+	dfprintk(VFS, "NFS: readdir(%s/%s) returns %ld\n",
+			dentry->d_parent->d_name.name, dentry->d_name.name,
+			res);
+	return res;
 }
 
 loff_t nfs_llseek_dir(struct file *filp, loff_t offset, int origin)
@@ -599,6 +622,10 @@ out:
  */
 int nfs_fsync_dir(struct file *filp, struct dentry *dentry, int datasync)
 {
+	dfprintk(VFS, "NFS: fsync_dir(%s/%s) datasync %d\n",
+			dentry->d_parent->d_name.name, dentry->d_name.name,
+			datasync);
+
 	return 0;
 }
 
@@ -713,6 +740,7 @@ static int nfs_lookup_revalidate(struct dentry * dentry, struct nameidata *nd)
 	parent = dget_parent(dentry);
 	lock_kernel();
 	dir = parent->d_inode;
+	nfs_inc_stats(dir, NFSIOS_DENTRYREVALIDATE);
 	inode = dentry->d_inode;
 
 	if (!inode) {
@@ -722,8 +750,9 @@ static int nfs_lookup_revalidate(struct dentry * dentry, struct nameidata *nd)
 	}
 
 	if (is_bad_inode(inode)) {
-		dfprintk(VFS, "nfs_lookup_validate: %s/%s has dud inode\n",
-			dentry->d_parent->d_name.name, dentry->d_name.name);
+		dfprintk(LOOKUPCACHE, "%s: %s/%s has dud inode\n",
+				__FUNCTION__, dentry->d_parent->d_name.name,
+				dentry->d_name.name);
 		goto out_bad;
 	}
 
@@ -755,6 +784,9 @@ static int nfs_lookup_revalidate(struct dentry * dentry, struct nameidata *nd)
  out_valid:
 	unlock_kernel();
 	dput(parent);
+	dfprintk(LOOKUPCACHE, "NFS: %s(%s/%s) is valid\n",
+			__FUNCTION__, dentry->d_parent->d_name.name,
+			dentry->d_name.name);
 	return 1;
 out_zap_parent:
 	nfs_zap_caches(dir);
@@ -771,6 +803,9 @@ out_zap_parent:
 	d_drop(dentry);
 	unlock_kernel();
 	dput(parent);
+	dfprintk(LOOKUPCACHE, "NFS: %s(%s/%s) is invalid\n",
+			__FUNCTION__, dentry->d_parent->d_name.name,
+			dentry->d_name.name);
 	return 0;
 }
 
@@ -844,6 +879,7 @@ static struct dentry *nfs_lookup(struct inode *dir, struct dentry * dentry, stru
 
 	dfprintk(VFS, "NFS: lookup(%s/%s)\n",
 		dentry->d_parent->d_name.name, dentry->d_name.name);
+	nfs_inc_stats(dir, NFSIOS_VFSLOOKUP);
 
 	res = ERR_PTR(-ENAMETOOLONG);
 	if (dentry->d_name.len > NFS_SERVER(dir)->namelen)
@@ -865,9 +901,9 @@ static struct dentry *nfs_lookup(struct inode *dir, struct dentry * dentry, stru
 		res = ERR_PTR(error);
 		goto out_unlock;
 	}
-	res = ERR_PTR(-EACCES);
 	inode = nfs_fhget(dentry->d_sb, &fhandle, &fattr);
-	if (!inode)
+	res = (struct dentry *)inode;
+	if (IS_ERR(res))
 		goto out_unlock;
 no_entry:
 	res = d_add_unique(dentry, inode);
@@ -912,6 +948,9 @@ static struct dentry *nfs_atomic_lookup(struct inode *dir, struct dentry *dentry
 	struct dentry *res = NULL;
 	int error;
 
+	dfprintk(VFS, "NFS: atomic_lookup(%s/%ld), %s\n",
+			dir->i_sb->s_id, dir->i_ino, dentry->d_name.name);
+
 	/* Check that we are indeed trying to open this file */
 	if (!is_atomic_open(dir, nd))
 		goto no_open;
@@ -1057,7 +1096,7 @@ static struct dentry *nfs_readdir_lookup(nfs_readdir_descriptor_t *desc)
 		return NULL;
 	dentry->d_op = NFS_PROTO(dir)->dentry_ops;
 	inode = nfs_fhget(dentry->d_sb, entry->fh, entry->fattr);
-	if (!inode) {
+	if (IS_ERR(inode)) {
 		dput(dentry);
 		return NULL;
 	}
@@ -1095,9 +1134,9 @@ int nfs_instantiate(struct dentry *dentry, struct nfs_fh *fhandle,
 		if (error < 0)
 			goto out_err;
 	}
-	error = -ENOMEM;
 	inode = nfs_fhget(dentry->d_sb, fhandle, fattr);
-	if (inode == NULL)
+	error = PTR_ERR(inode);
+	if (IS_ERR(inode))
 		goto out_err;
 	d_instantiate(dentry, inode);
 	return 0;
@@ -1119,8 +1158,8 @@ static int nfs_create(struct inode *dir, struct dentry *dentry, int mode,
 	int error;
 	int open_flags = 0;
 
-	dfprintk(VFS, "NFS: create(%s/%ld, %s\n", dir->i_sb->s_id, 
-		dir->i_ino, dentry->d_name.name);
+	dfprintk(VFS, "NFS: create(%s/%ld), %s\n",
+			dir->i_sb->s_id, dir->i_ino, dentry->d_name.name);
 
 	attr.ia_mode = mode;
 	attr.ia_valid = ATTR_MODE;
@@ -1153,8 +1192,8 @@ nfs_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t rdev)
 	struct iattr attr;
 	int status;
 
-	dfprintk(VFS, "NFS: mknod(%s/%ld, %s\n", dir->i_sb->s_id,
-		dir->i_ino, dentry->d_name.name);
+	dfprintk(VFS, "NFS: mknod(%s/%ld), %s\n",
+			dir->i_sb->s_id, dir->i_ino, dentry->d_name.name);
 
 	if (!new_valid_dev(rdev))
 		return -EINVAL;
@@ -1186,8 +1225,8 @@ static int nfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
 	struct iattr attr;
 	int error;
 
-	dfprintk(VFS, "NFS: mkdir(%s/%ld, %s\n", dir->i_sb->s_id,
-		dir->i_ino, dentry->d_name.name);
+	dfprintk(VFS, "NFS: mkdir(%s/%ld), %s\n",
+			dir->i_sb->s_id, dir->i_ino, dentry->d_name.name);
 
 	attr.ia_valid = ATTR_MODE;
 	attr.ia_mode = mode | S_IFDIR;
@@ -1212,8 +1251,8 @@ static int nfs_rmdir(struct inode *dir, struct dentry *dentry)
 {
 	int error;
 
-	dfprintk(VFS, "NFS: rmdir(%s/%ld, %s\n", dir->i_sb->s_id,
-		dir->i_ino, dentry->d_name.name);
+	dfprintk(VFS, "NFS: rmdir(%s/%ld), %s\n",
+			dir->i_sb->s_id, dir->i_ino, dentry->d_name.name);
 
 	lock_kernel();
 	nfs_begin_data_update(dir);
@@ -1241,6 +1280,7 @@ static int nfs_sillyrename(struct inode *dir, struct dentry *dentry)
 	dfprintk(VFS, "NFS: silly-rename(%s/%s, ct=%d)\n",
 		dentry->d_parent->d_name.name, dentry->d_name.name, 
 		atomic_read(&dentry->d_count));
+	nfs_inc_stats(dir, NFSIOS_SILLYRENAME);
 
 #ifdef NFS_PARANOIA
 if (!dentry->d_inode)
@@ -1268,8 +1308,8 @@ dentry->d_parent->d_name.name, dentry->d_name.name);
 		sillycounter++;
 		sprintf(suffix, "%*.*x", countersize, countersize, sillycounter);
 
-		dfprintk(VFS, "trying to rename %s to %s\n",
-			 dentry->d_name.name, silly);
+		dfprintk(VFS, "NFS: trying to rename %s to %s\n",
+				dentry->d_name.name, silly);
 		
 		sdentry = lookup_one_len(silly, dentry->d_parent, slen);
 		/*
@@ -1640,6 +1680,8 @@ int nfs_permission(struct inode *inode, int mask, struct nameidata *nd)
 	struct rpc_cred *cred;
 	int res = 0;
 
+	nfs_inc_stats(inode, NFSIOS_VFSACCESS);
+
 	if (mask == 0)
 		goto out;
 	/* Is this sys_access() ? */
@@ -1679,13 +1721,15 @@ force_lookup:
 		res = PTR_ERR(cred);
 	unlock_kernel();
 out:
+	dfprintk(VFS, "NFS: permission(%s/%ld), mask=0x%x, res=%d\n",
+		inode->i_sb->s_id, inode->i_ino, mask, res);
 	return res;
 out_notsup:
 	res = nfs_revalidate_inode(NFS_SERVER(inode), inode);
 	if (res == 0)
 		res = generic_permission(inode, mask, NULL);
 	unlock_kernel();
-	return res;
+	goto out;
 }
 
 /*

fs/nfs/direct.c

@@ -7,11 +7,11 @@
  *
  * There are important applications whose performance or correctness
  * depends on uncached access to file data.  Database clusters
- * (multiple copies of the same instance running on separate hosts) 
+ * (multiple copies of the same instance running on separate hosts)
  * implement their own cache coherency protocol that subsumes file
- * system cache protocols.  Applications that process datasets 
- * considerably larger than the client's memory do not always benefit 
- * from a local cache.  A streaming video server, for instance, has no 
+ * system cache protocols.  Applications that process datasets
+ * considerably larger than the client's memory do not always benefit
+ * from a local cache.  A streaming video server, for instance, has no
  * need to cache the contents of a file.
  *
  * When an application requests uncached I/O, all read and write requests
@@ -34,6 +34,7 @@
  * 08 Jun 2003	Port to 2.5 APIs  --cel
  * 31 Mar 2004	Handle direct I/O without VFS support  --cel
  * 15 Sep 2004	Parallel async reads  --cel
+ * 04 May 2005	support O_DIRECT with aio  --cel
  *
  */
 
@@ -54,10 +55,10 @@
 #include <asm/uaccess.h>
 #include <asm/atomic.h>
 
+#include "iostat.h"
+
 #define NFSDBG_FACILITY		NFSDBG_VFS
-#define MAX_DIRECTIO_SIZE	(4096UL << PAGE_SHIFT)
 
-static void nfs_free_user_pages(struct page **pages, int npages, int do_dirty);
 static kmem_cache_t *nfs_direct_cachep;
 
 /*
@@ -65,38 +66,78 @@ static kmem_cache_t *nfs_direct_cachep;
  */
 struct nfs_direct_req {
 	struct kref		kref;		/* release manager */
-	struct list_head	list;		/* nfs_read_data structs */
-	wait_queue_head_t	wait;		/* wait for i/o completion */
+
+	/* I/O parameters */
+	struct list_head	list,		/* nfs_read/write_data structs */
+				rewrite_list;	/* saved nfs_write_data structs */
+	struct nfs_open_context	*ctx;		/* file open context info */
+	struct kiocb *		iocb;		/* controlling i/o request */
+	struct inode *		inode;		/* target file of i/o */
+	unsigned long		user_addr;	/* location of user's buffer */
+	size_t			user_count;	/* total bytes to move */
+	loff_t			pos;		/* starting offset in file */
 	struct page **		pages;		/* pages in our buffer */
 	unsigned int		npages;		/* count of pages */
-	atomic_t		complete,	/* i/os we're waiting for */
-				count,		/* bytes actually processed */
+
+	/* completion state */
+	spinlock_t		lock;		/* protect completion state */
+	int			outstanding;	/* i/os we're waiting for */
+	ssize_t			count,		/* bytes actually processed */
 				error;		/* any reported error */
+	struct completion	completion;	/* wait for i/o completion */
+
+	/* commit state */
+	struct nfs_write_data *	commit_data;	/* special write_data for commits */
+	int			flags;
+#define NFS_ODIRECT_DO_COMMIT		(1)	/* an unstable reply was received */
+#define NFS_ODIRECT_RESCHED_WRITES	(2)	/* write verification failed */
+	struct nfs_writeverf	verf;		/* unstable write verifier */
 };
 
+static void nfs_direct_write_schedule(struct nfs_direct_req *dreq, int sync);
+static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode *inode);
 
 /**
- * nfs_get_user_pages - find and set up pages underlying user's buffer
- * rw: direction (read or write)
- * user_addr: starting address of this segment of user's buffer
- * count: size of this segment
- * @pages: returned array of page struct pointers underlying user's buffer
+ * nfs_direct_IO - NFS address space operation for direct I/O
+ * @rw: direction (read or write)
+ * @iocb: target I/O control block
+ * @iov: array of vectors that define I/O buffer
+ * @pos: offset in file to begin the operation
+ * @nr_segs: size of iovec array
+ *
+ * The presence of this routine in the address space ops vector means
+ * the NFS client supports direct I/O.  However, we shunt off direct
+ * read and write requests before the VFS gets them, so this method
+ * should never be called.
  */
-static inline int
-nfs_get_user_pages(int rw, unsigned long user_addr, size_t size,
-		struct page ***pages)
+ssize_t nfs_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov, loff_t pos, unsigned long nr_segs)
+{
+	struct dentry *dentry = iocb->ki_filp->f_dentry;
+
+	dprintk("NFS: nfs_direct_IO (%s) off/no(%Ld/%lu) EINVAL\n",
+			dentry->d_name.name, (long long) pos, nr_segs);
+
+	return -EINVAL;
+}
+
+static void nfs_free_user_pages(struct page **pages, int npages, int do_dirty)
+{
+	int i;
+	for (i = 0; i < npages; i++) {
+		struct page *page = pages[i];
+		if (do_dirty && !PageCompound(page))
+			set_page_dirty_lock(page);
+		page_cache_release(page);
+	}
+	kfree(pages);
+}
+
+static inline int nfs_get_user_pages(int rw, unsigned long user_addr, size_t size, struct page ***pages)
 {
 	int result = -ENOMEM;
 	unsigned long page_count;
 	size_t array_size;
 
-	/* set an arbitrary limit to prevent type overflow */
-	/* XXX: this can probably be as large as INT_MAX */
-	if (size > MAX_DIRECTIO_SIZE) {
-		*pages = NULL;
-		return -EFBIG;
-	}
-
 	page_count = (user_addr + size + PAGE_SIZE - 1) >> PAGE_SHIFT;
 	page_count -= user_addr >> PAGE_SHIFT;
 
@@ -108,75 +149,117 @@ nfs_get_user_pages(int rw, unsigned long user_addr, size_t size,
 					page_count, (rw == READ), 0,
 					*pages, NULL);
 		up_read(&current->mm->mmap_sem);
-		/*
-		 * If we got fewer pages than expected from get_user_pages(),
-		 * the user buffer runs off the end of a mapping; return EFAULT.
-		 */
-		if (result >= 0 && result < page_count) {
-			nfs_free_user_pages(*pages, result, 0);
+		if (result != page_count) {
+			/*
+			 * If we got fewer pages than expected from
+			 * get_user_pages(), the user buffer runs off the
+			 * end of a mapping; return EFAULT.
+			 */
+			if (result >= 0) {
+				nfs_free_user_pages(*pages, result, 0);
+				result = -EFAULT;
+			} else
+				kfree(*pages);
 			*pages = NULL;
-			result = -EFAULT;
 		}
 	}
 	return result;
 }
 
-/**
- * nfs_free_user_pages - tear down page struct array
- * @pages: array of page struct pointers underlying target buffer
- * @npages: number of pages in the array
- * @do_dirty: dirty the pages as we release them
- */
-static void
-nfs_free_user_pages(struct page **pages, int npages, int do_dirty)
+static inline struct nfs_direct_req *nfs_direct_req_alloc(void)
 {
-	int i;
-	for (i = 0; i < npages; i++) {
-		struct page *page = pages[i];
-		if (do_dirty && !PageCompound(page))
-			set_page_dirty_lock(page);
-		page_cache_release(page);
-	}
-	kfree(pages);
+	struct nfs_direct_req *dreq;
+
+	dreq = kmem_cache_alloc(nfs_direct_cachep, SLAB_KERNEL);
+	if (!dreq)
+		return NULL;
+
+	kref_init(&dreq->kref);
+	init_completion(&dreq->completion);
+	INIT_LIST_HEAD(&dreq->list);
+	INIT_LIST_HEAD(&dreq->rewrite_list);
+	dreq->iocb = NULL;
+	dreq->ctx = NULL;
+	spin_lock_init(&dreq->lock);
+	dreq->outstanding = 0;
+	dreq->count = 0;
+	dreq->error = 0;
+	dreq->flags = 0;
+
+	return dreq;
 }
 
-/**
- * nfs_direct_req_release - release  nfs_direct_req structure for direct read
- * @kref: kref object embedded in an nfs_direct_req structure
- *
- */
 static void nfs_direct_req_release(struct kref *kref)
 {
 	struct nfs_direct_req *dreq = container_of(kref, struct nfs_direct_req, kref);
+
+	if (dreq->ctx != NULL)
+		put_nfs_open_context(dreq->ctx);
 	kmem_cache_free(nfs_direct_cachep, dreq);
 }
 
-/**
- * nfs_direct_read_alloc - allocate nfs_read_data structures for direct read
- * @count: count of bytes for the read request
- * @rsize: local rsize setting
+/*
+ * Collects and returns the final error value/byte-count.
+ */
+static ssize_t nfs_direct_wait(struct nfs_direct_req *dreq)
+{
+	ssize_t result = -EIOCBQUEUED;
+
+	/* Async requests don't wait here */
+	if (dreq->iocb)
+		goto out;
+
+	result = wait_for_completion_interruptible(&dreq->completion);
+
+	if (!result)
+		result = dreq->error;
+	if (!result)
+		result = dreq->count;
+
+out:
+	kref_put(&dreq->kref, nfs_direct_req_release);
+	return (ssize_t) result;
+}
+
+/*
+ * We must hold a reference to all the pages in this direct read request
+ * until the RPCs complete.  This could be long *after* we are woken up in
+ * nfs_direct_wait (for instance, if someone hits ^C on a slow server).
  *
+ * In addition, synchronous I/O uses a stack-allocated iocb.  Thus we
+ * can't trust the iocb is still valid here if this is a synchronous
+ * request.  If the waiter is woken prematurely, the iocb is long gone.
+ */
+static void nfs_direct_complete(struct nfs_direct_req *dreq)
+{
+	nfs_free_user_pages(dreq->pages, dreq->npages, 1);
+
+	if (dreq->iocb) {
+		long res = (long) dreq->error;
+		if (!res)
+			res = (long) dreq->count;
+		aio_complete(dreq->iocb, res, 0);
+	}
+	complete_all(&dreq->completion);
+
+	kref_put(&dreq->kref, nfs_direct_req_release);
+}
+
+/*
  * Note we also set the number of requests we have in the dreq when we are
  * done.  This prevents races with I/O completion so we will always wait
  * until all requests have been dispatched and completed.
  */
-static struct nfs_direct_req *nfs_direct_read_alloc(size_t nbytes, unsigned int rsize)
+static struct nfs_direct_req *nfs_direct_read_alloc(size_t nbytes, size_t rsize)
 {
 	struct list_head *list;
 	struct nfs_direct_req *dreq;
-	unsigned int reads = 0;
 	unsigned int rpages = (rsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
 
-	dreq = kmem_cache_alloc(nfs_direct_cachep, SLAB_KERNEL);
+	dreq = nfs_direct_req_alloc();
 	if (!dreq)
 		return NULL;
 
-	kref_init(&dreq->kref);
-	init_waitqueue_head(&dreq->wait);
-	INIT_LIST_HEAD(&dreq->list);
-	atomic_set(&dreq->count, 0);
-	atomic_set(&dreq->error, 0);
-
 	list = &dreq->list;
 	for(;;) {
 		struct nfs_read_data *data = nfs_readdata_alloc(rpages);
@@ -196,72 +279,70 @@ static struct nfs_direct_req *nfs_direct_read_alloc(size_t nbytes, unsigned int
 		list_add(&data->pages, list);
 
 		data->req = (struct nfs_page *) dreq;
-		reads++;
+		dreq->outstanding++;
 		if (nbytes <= rsize)
 			break;
 		nbytes -= rsize;
 	}
 	kref_get(&dreq->kref);
-	atomic_set(&dreq->complete, reads);
 	return dreq;
 }
 
-/**
- * nfs_direct_read_result - handle a read reply for a direct read request
- * @data: address of NFS READ operation control block
- * @status: status of this NFS READ operation
- *
- * We must hold a reference to all the pages in this direct read request
- * until the RPCs complete.  This could be long *after* we are woken up in
- * nfs_direct_read_wait (for instance, if someone hits ^C on a slow server).
- */
-static void nfs_direct_read_result(struct nfs_read_data *data, int status)
+static void nfs_direct_read_result(struct rpc_task *task, void *calldata)
 {
+	struct nfs_read_data *data = calldata;
 	struct nfs_direct_req *dreq = (struct nfs_direct_req *) data->req;
 
-	if (likely(status >= 0))
-		atomic_add(data->res.count, &dreq->count);
+	if (nfs_readpage_result(task, data) != 0)
+		return;
+
+	spin_lock(&dreq->lock);
+
+	if (likely(task->tk_status >= 0))
+		dreq->count += data->res.count;
 	else
-		atomic_set(&dreq->error, status);
+		dreq->error = task->tk_status;
 
-	if (unlikely(atomic_dec_and_test(&dreq->complete))) {
-		nfs_free_user_pages(dreq->pages, dreq->npages, 1);
-		wake_up(&dreq->wait);
-		kref_put(&dreq->kref, nfs_direct_req_release);
+	if (--dreq->outstanding) {
+		spin_unlock(&dreq->lock);
+		return;
 	}
+
+	spin_unlock(&dreq->lock);
+	nfs_direct_complete(dreq);
 }
 
-/**
- * nfs_direct_read_schedule - dispatch NFS READ operations for a direct read
- * @dreq: address of nfs_direct_req struct for this request
- * @inode: target inode
- * @ctx: target file open context
- * @user_addr: starting address of this segment of user's buffer
- * @count: size of this segment
- * @file_offset: offset in file to begin the operation
- *
+static const struct rpc_call_ops nfs_read_direct_ops = {
+	.rpc_call_done = nfs_direct_read_result,
+	.rpc_release = nfs_readdata_release,
+};
+
+/*
  * For each nfs_read_data struct that was allocated on the list, dispatch
  * an NFS READ operation
  */
-static void nfs_direct_read_schedule(struct nfs_direct_req *dreq,
-		struct inode *inode, struct nfs_open_context *ctx,
-		unsigned long user_addr, size_t count, loff_t file_offset)
+static void nfs_direct_read_schedule(struct nfs_direct_req *dreq)
 {
+	struct nfs_open_context *ctx = dreq->ctx;
+	struct inode *inode = ctx->dentry->d_inode;
 	struct list_head *list = &dreq->list;
 	struct page **pages = dreq->pages;
+	size_t count = dreq->user_count;
+	loff_t pos = dreq->pos;
+	size_t rsize = NFS_SERVER(inode)->rsize;
 	unsigned int curpage, pgbase;
-	unsigned int rsize = NFS_SERVER(inode)->rsize;
 
 	curpage = 0;
-	pgbase = user_addr & ~PAGE_MASK;
+	pgbase = dreq->user_addr & ~PAGE_MASK;
 	do {
 		struct nfs_read_data *data;
-		unsigned int bytes;
+		size_t bytes;
 
 		bytes = rsize;
 		if (count < rsize)
 			bytes = count;
 
+		BUG_ON(list_empty(list));
 		data = list_entry(list->next, struct nfs_read_data, pages);
 		list_del_init(&data->pages);
 
@@ -269,7 +350,7 @@ static void nfs_direct_read_schedule(struct nfs_direct_req *dreq,
 		data->cred = ctx->cred;
 		data->args.fh = NFS_FH(inode);
 		data->args.context = ctx;
-		data->args.offset = file_offset;
+		data->args.offset = pos;
 		data->args.pgbase = pgbase;
 		data->args.pages = &pages[curpage];
 		data->args.count = bytes;
@@ -277,77 +358,38 @@ static void nfs_direct_read_schedule(struct nfs_direct_req *dreq,
 		data->res.eof = 0;
 		data->res.count = bytes;
 
+		rpc_init_task(&data->task, NFS_CLIENT(inode), RPC_TASK_ASYNC,
+				&nfs_read_direct_ops, data);
 		NFS_PROTO(inode)->read_setup(data);
 
 		data->task.tk_cookie = (unsigned long) inode;
-		data->complete = nfs_direct_read_result;
 
 		lock_kernel();
 		rpc_execute(&data->task);
 		unlock_kernel();
 
-		dfprintk(VFS, "NFS: %4d initiated direct read call (req %s/%Ld, %u bytes @ offset %Lu)\n",
+		dfprintk(VFS, "NFS: %5u initiated direct read call (req %s/%Ld, %zu bytes @ offset %Lu)\n",
 				data->task.tk_pid,
 				inode->i_sb->s_id,
 				(long long)NFS_FILEID(inode),
 				bytes,
 				(unsigned long long)data->args.offset);
 
-		file_offset += bytes;
+		pos += bytes;
 		pgbase += bytes;
 		curpage += pgbase >> PAGE_SHIFT;
 		pgbase &= ~PAGE_MASK;
 
 		count -= bytes;
 	} while (count != 0);
+	BUG_ON(!list_empty(list));
 }
 
-/**
- * nfs_direct_read_wait - wait for I/O completion for direct reads
- * @dreq: request on which we are to wait
- * @intr: whether or not this wait can be interrupted
- *
- * Collects and returns the final error value/byte-count.
- */
-static ssize_t nfs_direct_read_wait(struct nfs_direct_req *dreq, int intr)
-{
-	int result = 0;
-
-	if (intr) {
-		result = wait_event_interruptible(dreq->wait,
-					(atomic_read(&dreq->complete) == 0));
-	} else {
-		wait_event(dreq->wait, (atomic_read(&dreq->complete) == 0));
-	}
-
-	if (!result)
-		result = atomic_read(&dreq->error);
-	if (!result)
-		result = atomic_read(&dreq->count);
-
-	kref_put(&dreq->kref, nfs_direct_req_release);
-	return (ssize_t) result;
-}
-
-/**
- * nfs_direct_read_seg - Read in one iov segment.  Generate separate
- *                        read RPCs for each "rsize" bytes.
- * @inode: target inode
- * @ctx: target file open context
- * @user_addr: starting address of this segment of user's buffer
- * @count: size of this segment
- * @file_offset: offset in file to begin the operation
- * @pages: array of addresses of page structs defining user's buffer
- * @nr_pages: number of pages in the array
- *
- */
-static ssize_t nfs_direct_read_seg(struct inode *inode,
-		struct nfs_open_context *ctx, unsigned long user_addr,
-		size_t count, loff_t file_offset, struct page **pages,
-		unsigned int nr_pages)
+static ssize_t nfs_direct_read(struct kiocb *iocb, unsigned long user_addr, size_t count, loff_t pos, struct page **pages, unsigned int nr_pages)
 {
 	ssize_t result;
 	sigset_t oldset;
+	struct inode *inode = iocb->ki_filp->f_mapping->host;
 	struct rpc_clnt *clnt = NFS_CLIENT(inode);
 	struct nfs_direct_req *dreq;
 
@@ -355,284 +397,350 @@ static ssize_t nfs_direct_read_seg(struct inode *inode,
 	if (!dreq)
 		return -ENOMEM;
 
+	dreq->user_addr = user_addr;
+	dreq->user_count = count;
+	dreq->pos = pos;
 	dreq->pages = pages;
 	dreq->npages = nr_pages;
+	dreq->inode = inode;
+	dreq->ctx = get_nfs_open_context((struct nfs_open_context *)iocb->ki_filp->private_data);
+	if (!is_sync_kiocb(iocb))
+		dreq->iocb = iocb;
 
+	nfs_add_stats(inode, NFSIOS_DIRECTREADBYTES, count);
 	rpc_clnt_sigmask(clnt, &oldset);
-	nfs_direct_read_schedule(dreq, inode, ctx, user_addr, count,
-				 file_offset);
-	result = nfs_direct_read_wait(dreq, clnt->cl_intr);
+	nfs_direct_read_schedule(dreq);
+	result = nfs_direct_wait(dreq);
 	rpc_clnt_sigunmask(clnt, &oldset);
 
 	return result;
 }
 
-/**
- * nfs_direct_read - For each iov segment, map the user's buffer
- *                   then generate read RPCs.
- * @inode: target inode
- * @ctx: target file open context
- * @iov: array of vectors that define I/O buffer
- * file_offset: offset in file to begin the operation
- * nr_segs: size of iovec array
- *
- * We've already pushed out any non-direct writes so that this read
- * will see them when we read from the server.
- */
-static ssize_t
-nfs_direct_read(struct inode *inode, struct nfs_open_context *ctx,
-		const struct iovec *iov, loff_t file_offset,
-		unsigned long nr_segs)
+static void nfs_direct_free_writedata(struct nfs_direct_req *dreq)
 {
-	ssize_t tot_bytes = 0;
-	unsigned long seg = 0;
-
-	while ((seg < nr_segs) && (tot_bytes >= 0)) {
-		ssize_t result;
-		int page_count;
-		struct page **pages;
-		const struct iovec *vec = &iov[seg++];
-		unsigned long user_addr = (unsigned long) vec->iov_base;
-		size_t size = vec->iov_len;
-
-                page_count = nfs_get_user_pages(READ, user_addr, size, &pages);
-                if (page_count < 0) {
-                        nfs_free_user_pages(pages, 0, 0);
-			if (tot_bytes > 0)
-				break;
-                        return page_count;
-                }
+	list_splice_init(&dreq->rewrite_list, &dreq->list);
+	while (!list_empty(&dreq->list)) {
+		struct nfs_write_data *data = list_entry(dreq->list.next, struct nfs_write_data, pages);
+		list_del(&data->pages);
+		nfs_writedata_release(data);
+	}
+}
 
-		result = nfs_direct_read_seg(inode, ctx, user_addr, size,
-				file_offset, pages, page_count);
+#if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
+static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq)
+{
+	struct list_head *pos;
 
-		if (result <= 0) {
-			if (tot_bytes > 0)
-				break;
-			return result;
-		}
-		tot_bytes += result;
-		file_offset += result;
-		if (result < size)
-			break;
+	list_splice_init(&dreq->rewrite_list, &dreq->list);
+	list_for_each(pos, &dreq->list)
+		dreq->outstanding++;
+	dreq->count = 0;
+
+	nfs_direct_write_schedule(dreq, FLUSH_STABLE);
+}
+
+static void nfs_direct_commit_result(struct rpc_task *task, void *calldata)
+{
+	struct nfs_write_data *data = calldata;
+	struct nfs_direct_req *dreq = (struct nfs_direct_req *) data->req;
+
+	/* Call the NFS version-specific code */
+	if (NFS_PROTO(data->inode)->commit_done(task, data) != 0)
+		return;
+	if (unlikely(task->tk_status < 0)) {
+		dreq->error = task->tk_status;
+		dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
+	}
+	if (memcmp(&dreq->verf, &data->verf, sizeof(data->verf))) {
+		dprintk("NFS: %5u commit verify failed\n", task->tk_pid);
+		dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
 	}
 
-	return tot_bytes;
+	dprintk("NFS: %5u commit returned %d\n", task->tk_pid, task->tk_status);
+	nfs_direct_write_complete(dreq, data->inode);
 }
 
-/**
- * nfs_direct_write_seg - Write out one iov segment.  Generate separate
- *                        write RPCs for each "wsize" bytes, then commit.
- * @inode: target inode
- * @ctx: target file open context
- * user_addr: starting address of this segment of user's buffer
- * count: size of this segment
- * file_offset: offset in file to begin the operation
- * @pages: array of addresses of page structs defining user's buffer
- * nr_pages: size of pages array
- */
-static ssize_t nfs_direct_write_seg(struct inode *inode,
-		struct nfs_open_context *ctx, unsigned long user_addr,
-		size_t count, loff_t file_offset, struct page **pages,
-		int nr_pages)
+static const struct rpc_call_ops nfs_commit_direct_ops = {
+	.rpc_call_done = nfs_direct_commit_result,
+	.rpc_release = nfs_commit_release,
+};
+
+static void nfs_direct_commit_schedule(struct nfs_direct_req *dreq)
 {
-	const unsigned int wsize = NFS_SERVER(inode)->wsize;
-	size_t request;
-	int curpage, need_commit;
-	ssize_t result, tot_bytes;
-	struct nfs_writeverf first_verf;
-	struct nfs_write_data *wdata;
-
-	wdata = nfs_writedata_alloc(NFS_SERVER(inode)->wpages);
-	if (!wdata)
-		return -ENOMEM;
+	struct nfs_write_data *data = dreq->commit_data;
+	struct rpc_task *task = &data->task;
 
-	wdata->inode = inode;
-	wdata->cred = ctx->cred;
-	wdata->args.fh = NFS_FH(inode);
-	wdata->args.context = ctx;
-	wdata->args.stable = NFS_UNSTABLE;
-	if (IS_SYNC(inode) || NFS_PROTO(inode)->version == 2 || count <= wsize)
-		wdata->args.stable = NFS_FILE_SYNC;
-	wdata->res.fattr = &wdata->fattr;
-	wdata->res.verf = &wdata->verf;
+	data->inode = dreq->inode;
+	data->cred = dreq->ctx->cred;
 
-	nfs_begin_data_update(inode);
-retry:
-	need_commit = 0;
-	tot_bytes = 0;
-	curpage = 0;
-	request = count;
-	wdata->args.pgbase = user_addr & ~PAGE_MASK;
-	wdata->args.offset = file_offset;
-	do {
-		wdata->args.count = request;
-		if (wdata->args.count > wsize)
-			wdata->args.count = wsize;
-		wdata->args.pages = &pages[curpage];
+	data->args.fh = NFS_FH(data->inode);
+	data->args.offset = dreq->pos;
+	data->args.count = dreq->user_count;
+	data->res.count = 0;
+	data->res.fattr = &data->fattr;
+	data->res.verf = &data->verf;
 
-		dprintk("NFS: direct write: c=%u o=%Ld ua=%lu, pb=%u, cp=%u\n",
-			wdata->args.count, (long long) wdata->args.offset,
-			user_addr + tot_bytes, wdata->args.pgbase, curpage);
+	rpc_init_task(&data->task, NFS_CLIENT(dreq->inode), RPC_TASK_ASYNC,
+				&nfs_commit_direct_ops, data);
+	NFS_PROTO(data->inode)->commit_setup(data, 0);
 
-		lock_kernel();
-		result = NFS_PROTO(inode)->write(wdata);
-		unlock_kernel();
+	data->task.tk_priority = RPC_PRIORITY_NORMAL;
+	data->task.tk_cookie = (unsigned long)data->inode;
+	/* Note: task.tk_ops->rpc_release will free dreq->commit_data */
+	dreq->commit_data = NULL;
 
-		if (result <= 0) {
-			if (tot_bytes > 0)
-				break;
-			goto out;
-		}
+	dprintk("NFS: %5u initiated commit call\n", task->tk_pid);
 
-		if (tot_bytes == 0)
-			memcpy(&first_verf.verifier, &wdata->verf.verifier,
-						sizeof(first_verf.verifier));
-		if (wdata->verf.committed != NFS_FILE_SYNC) {
-			need_commit = 1;
-			if (memcmp(&first_verf.verifier, &wdata->verf.verifier,
-					sizeof(first_verf.verifier)))
-				goto sync_retry;
-		}
+	lock_kernel();
+	rpc_execute(&data->task);
+	unlock_kernel();
+}
 
-		tot_bytes += result;
+static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode *inode)
+{
+	int flags = dreq->flags;
 
-		/* in case of a short write: stop now, let the app recover */
-		if (result < wdata->args.count)
+	dreq->flags = 0;
+	switch (flags) {
+		case NFS_ODIRECT_DO_COMMIT:
+			nfs_direct_commit_schedule(dreq);
 			break;
+		case NFS_ODIRECT_RESCHED_WRITES:
+			nfs_direct_write_reschedule(dreq);
+			break;
+		default:
+			nfs_end_data_update(inode);
+			if (dreq->commit_data != NULL)
+				nfs_commit_free(dreq->commit_data);
+			nfs_direct_free_writedata(dreq);
+			nfs_direct_complete(dreq);
+	}
+}
 
-		wdata->args.offset += result;
-		wdata->args.pgbase += result;
-		curpage += wdata->args.pgbase >> PAGE_SHIFT;
-		wdata->args.pgbase &= ~PAGE_MASK;
-		request -= result;
-	} while (request != 0);
+static void nfs_alloc_commit_data(struct nfs_direct_req *dreq)
+{
+	dreq->commit_data = nfs_commit_alloc(0);
+	if (dreq->commit_data != NULL)
+		dreq->commit_data->req = (struct nfs_page *) dreq;
+}
+#else
+static inline void nfs_alloc_commit_data(struct nfs_direct_req *dreq)
+{
+	dreq->commit_data = NULL;
+}
 
-	/*
-	 * Commit data written so far, even in the event of an error
-	 */
-	if (need_commit) {
-		wdata->args.count = tot_bytes;
-		wdata->args.offset = file_offset;
+static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode *inode)
+{
+	nfs_end_data_update(inode);
+	nfs_direct_free_writedata(dreq);
+	nfs_direct_complete(dreq);
+}
+#endif
 
-		lock_kernel();
-		result = NFS_PROTO(inode)->commit(wdata);
-		unlock_kernel();
+static struct nfs_direct_req *nfs_direct_write_alloc(size_t nbytes, size_t wsize)
+{
+	struct list_head *list;
+	struct nfs_direct_req *dreq;
+	unsigned int wpages = (wsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
+
+	dreq = nfs_direct_req_alloc();
+	if (!dreq)
+		return NULL;
+
+	list = &dreq->list;
+	for(;;) {
+		struct nfs_write_data *data = nfs_writedata_alloc(wpages);
 
-		if (result < 0 || memcmp(&first_verf.verifier,
-					 &wdata->verf.verifier,
-					 sizeof(first_verf.verifier)) != 0)
-			goto sync_retry;
+		if (unlikely(!data)) {
+			while (!list_empty(list)) {
+				data = list_entry(list->next,
+						  struct nfs_write_data, pages);
+				list_del(&data->pages);
+				nfs_writedata_free(data);
+			}
+			kref_put(&dreq->kref, nfs_direct_req_release);
+			return NULL;
+		}
+
+		INIT_LIST_HEAD(&data->pages);
+		list_add(&data->pages, list);
+
+		data->req = (struct nfs_page *) dreq;
+		dreq->outstanding++;
+		if (nbytes <= wsize)
+			break;
+		nbytes -= wsize;
 	}
-	result = tot_bytes;
 
-out:
-	nfs_end_data_update(inode);
-	nfs_writedata_free(wdata);
-	return result;
+	nfs_alloc_commit_data(dreq);
 
-sync_retry:
-	wdata->args.stable = NFS_FILE_SYNC;
-	goto retry;
+	kref_get(&dreq->kref);
+	return dreq;
 }
 
-/**
- * nfs_direct_write - For each iov segment, map the user's buffer
- *                    then generate write and commit RPCs.
- * @inode: target inode
- * @ctx: target file open context
- * @iov: array of vectors that define I/O buffer
- * file_offset: offset in file to begin the operation
- * nr_segs: size of iovec array
- *
- * Upon return, generic_file_direct_IO invalidates any cached pages
- * that non-direct readers might access, so they will pick up these
- * writes immediately.
- */
-static ssize_t nfs_direct_write(struct inode *inode,
-		struct nfs_open_context *ctx, const struct iovec *iov,
-		loff_t file_offset, unsigned long nr_segs)
+static void nfs_direct_write_result(struct rpc_task *task, void *calldata)
 {
-	ssize_t tot_bytes = 0;
-	unsigned long seg = 0;
-
-	while ((seg < nr_segs) && (tot_bytes >= 0)) {
-		ssize_t result;
-		int page_count;
-		struct page **pages;
-		const struct iovec *vec = &iov[seg++];
-		unsigned long user_addr = (unsigned long) vec->iov_base;
-		size_t size = vec->iov_len;
-
-                page_count = nfs_get_user_pages(WRITE, user_addr, size, &pages);
-                if (page_count < 0) {
-                        nfs_free_user_pages(pages, 0, 0);
-			if (tot_bytes > 0)
-				break;
-                        return page_count;
-                }
+	struct nfs_write_data *data = calldata;
+	struct nfs_direct_req *dreq = (struct nfs_direct_req *) data->req;
+	int status = task->tk_status;
+
+	if (nfs_writeback_done(task, data) != 0)
+		return;
+
+	spin_lock(&dreq->lock);
 
-		result = nfs_direct_write_seg(inode, ctx, user_addr, size,
-				file_offset, pages, page_count);
-		nfs_free_user_pages(pages, page_count, 0);
+	if (likely(status >= 0))
+		dreq->count += data->res.count;
+	else
+		dreq->error = task->tk_status;
 
-		if (result <= 0) {
-			if (tot_bytes > 0)
+	if (data->res.verf->committed != NFS_FILE_SYNC) {
+		switch (dreq->flags) {
+			case 0:
+				memcpy(&dreq->verf, &data->verf, sizeof(dreq->verf));
+				dreq->flags = NFS_ODIRECT_DO_COMMIT;
 				break;
-			return result;
+			case NFS_ODIRECT_DO_COMMIT:
+				if (memcmp(&dreq->verf, &data->verf, sizeof(dreq->verf))) {
+					dprintk("NFS: %5u write verify failed\n", task->tk_pid);
+					dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
+				}
 		}
-		tot_bytes += result;
-		file_offset += result;
-		if (result < size)
-			break;
 	}
-	return tot_bytes;
+	/* In case we have to resend */
+	data->args.stable = NFS_FILE_SYNC;
+
+	spin_unlock(&dreq->lock);
 }
 
-/**
- * nfs_direct_IO - NFS address space operation for direct I/O
- * rw: direction (read or write)
- * @iocb: target I/O control block
- * @iov: array of vectors that define I/O buffer
- * file_offset: offset in file to begin the operation
- * nr_segs: size of iovec array
- *
+/*
+ * NB: Return the value of the first error return code.  Subsequent
+ *     errors after the first one are ignored.
  */
-ssize_t
-nfs_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov,
-		loff_t file_offset, unsigned long nr_segs)
+static void nfs_direct_write_release(void *calldata)
 {
-	ssize_t result = -EINVAL;
-	struct file *file = iocb->ki_filp;
-	struct nfs_open_context *ctx;
-	struct dentry *dentry = file->f_dentry;
-	struct inode *inode = dentry->d_inode;
+	struct nfs_write_data *data = calldata;
+	struct nfs_direct_req *dreq = (struct nfs_direct_req *) data->req;
 
-	/*
-	 * No support for async yet
-	 */
-	if (!is_sync_kiocb(iocb))
-		return result;
-
-	ctx = (struct nfs_open_context *)file->private_data;
-	switch (rw) {
-	case READ:
-		dprintk("NFS: direct_IO(read) (%s) off/no(%Lu/%lu)\n",
-				dentry->d_name.name, file_offset, nr_segs);
-
-		result = nfs_direct_read(inode, ctx, iov,
-						file_offset, nr_segs);
-		break;
-	case WRITE:
-		dprintk("NFS: direct_IO(write) (%s) off/no(%Lu/%lu)\n",
-				dentry->d_name.name, file_offset, nr_segs);
-
-		result = nfs_direct_write(inode, ctx, iov,
-						file_offset, nr_segs);
-		break;
-	default:
-		break;
+	spin_lock(&dreq->lock);
+	if (--dreq->outstanding) {
+		spin_unlock(&dreq->lock);
+		return;
 	}
+	spin_unlock(&dreq->lock);
+
+	nfs_direct_write_complete(dreq, data->inode);
+}
+
+static const struct rpc_call_ops nfs_write_direct_ops = {
+	.rpc_call_done = nfs_direct_write_result,
+	.rpc_release = nfs_direct_write_release,
+};
+
+/*
+ * For each nfs_write_data struct that was allocated on the list, dispatch
+ * an NFS WRITE operation
+ */
+static void nfs_direct_write_schedule(struct nfs_direct_req *dreq, int sync)
+{
+	struct nfs_open_context *ctx = dreq->ctx;
+	struct inode *inode = ctx->dentry->d_inode;
+	struct list_head *list = &dreq->list;
+	struct page **pages = dreq->pages;
+	size_t count = dreq->user_count;
+	loff_t pos = dreq->pos;
+	size_t wsize = NFS_SERVER(inode)->wsize;
+	unsigned int curpage, pgbase;
+
+	curpage = 0;
+	pgbase = dreq->user_addr & ~PAGE_MASK;
+	do {
+		struct nfs_write_data *data;
+		size_t bytes;
+
+		bytes = wsize;
+		if (count < wsize)
+			bytes = count;
+
+		BUG_ON(list_empty(list));
+		data = list_entry(list->next, struct nfs_write_data, pages);
+		list_move_tail(&data->pages, &dreq->rewrite_list);
+
+		data->inode = inode;
+		data->cred = ctx->cred;
+		data->args.fh = NFS_FH(inode);
+		data->args.context = ctx;
+		data->args.offset = pos;
+		data->args.pgbase = pgbase;
+		data->args.pages = &pages[curpage];
+		data->args.count = bytes;
+		data->res.fattr = &data->fattr;
+		data->res.count = bytes;
+		data->res.verf = &data->verf;
+
+		rpc_init_task(&data->task, NFS_CLIENT(inode), RPC_TASK_ASYNC,
+				&nfs_write_direct_ops, data);
+		NFS_PROTO(inode)->write_setup(data, sync);
+
+		data->task.tk_priority = RPC_PRIORITY_NORMAL;
+		data->task.tk_cookie = (unsigned long) inode;
+
+		lock_kernel();
+		rpc_execute(&data->task);
+		unlock_kernel();
+
+		dfprintk(VFS, "NFS: %5u initiated direct write call (req %s/%Ld, %zu bytes @ offset %Lu)\n",
+				data->task.tk_pid,
+				inode->i_sb->s_id,
+				(long long)NFS_FILEID(inode),
+				bytes,
+				(unsigned long long)data->args.offset);
+
+		pos += bytes;
+		pgbase += bytes;
+		curpage += pgbase >> PAGE_SHIFT;
+		pgbase &= ~PAGE_MASK;
+
+		count -= bytes;
+	} while (count != 0);
+	BUG_ON(!list_empty(list));
+}
+
+static ssize_t nfs_direct_write(struct kiocb *iocb, unsigned long user_addr, size_t count, loff_t pos, struct page **pages, int nr_pages)
+{
+	ssize_t result;
+	sigset_t oldset;
+	struct inode *inode = iocb->ki_filp->f_mapping->host;
+	struct rpc_clnt *clnt = NFS_CLIENT(inode);
+	struct nfs_direct_req *dreq;
+	size_t wsize = NFS_SERVER(inode)->wsize;
+	int sync = 0;
+
+	dreq = nfs_direct_write_alloc(count, wsize);
+	if (!dreq)
+		return -ENOMEM;
+	if (dreq->commit_data == NULL || count < wsize)
+		sync = FLUSH_STABLE;
+
+	dreq->user_addr = user_addr;
+	dreq->user_count = count;
+	dreq->pos = pos;
+	dreq->pages = pages;
+	dreq->npages = nr_pages;
+	dreq->inode = inode;
+	dreq->ctx = get_nfs_open_context((struct nfs_open_context *)iocb->ki_filp->private_data);
+	if (!is_sync_kiocb(iocb))
+		dreq->iocb = iocb;
+
+	nfs_add_stats(inode, NFSIOS_DIRECTWRITTENBYTES, count);
+
+	nfs_begin_data_update(inode);
+
+	rpc_clnt_sigmask(clnt, &oldset);
+	nfs_direct_write_schedule(dreq, sync);
+	result = nfs_direct_wait(dreq);
+	rpc_clnt_sigunmask(clnt, &oldset);
+
 	return result;
 }
 
@@ -640,49 +748,40 @@ nfs_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov,
  * nfs_file_direct_read - file direct read operation for NFS files
  * @iocb: target I/O control block
  * @buf: user's buffer into which to read data
- * count: number of bytes to read
- * pos: byte offset in file where reading starts
+ * @count: number of bytes to read
+ * @pos: byte offset in file where reading starts
  *
  * We use this function for direct reads instead of calling
  * generic_file_aio_read() in order to avoid gfar's check to see if
  * the request starts before the end of the file.  For that check
  * to work, we must generate a GETATTR before each direct read, and
  * even then there is a window between the GETATTR and the subsequent
- * READ where the file size could change.  So our preference is simply
+ * READ where the file size could change.  Our preference is simply
  * to do all reads the application wants, and the server will take
  * care of managing the end of file boundary.
- * 
+ *
  * This function also eliminates unnecessarily updating the file's
  * atime locally, as the NFS server sets the file's atime, and this
  * client must read the updated atime from the server back into its
  * cache.
  */
-ssize_t
-nfs_file_direct_read(struct kiocb *iocb, char __user *buf, size_t count, loff_t pos)
+ssize_t nfs_file_direct_read(struct kiocb *iocb, char __user *buf, size_t count, loff_t pos)
 {
 	ssize_t retval = -EINVAL;
-	loff_t *ppos = &iocb->ki_pos;
+	int page_count;
+	struct page **pages;
 	struct file *file = iocb->ki_filp;
-	struct nfs_open_context *ctx =
-			(struct nfs_open_context *) file->private_data;
 	struct address_space *mapping = file->f_mapping;
-	struct inode *inode = mapping->host;
-	struct iovec iov = {
-		.iov_base = buf,
-		.iov_len = count,
-	};
 
 	dprintk("nfs: direct read(%s/%s, %lu@%Ld)\n",
 		file->f_dentry->d_parent->d_name.name,
 		file->f_dentry->d_name.name,
 		(unsigned long) count, (long long) pos);
 
-	if (!is_sync_kiocb(iocb))
-		goto out;
 	if (count < 0)
 		goto out;
 	retval = -EFAULT;
-	if (!access_ok(VERIFY_WRITE, iov.iov_base, iov.iov_len))
+	if (!access_ok(VERIFY_WRITE, buf, count))
 		goto out;
 	retval = 0;
 	if (!count)
@@ -692,9 +791,16 @@ nfs_file_direct_read(struct kiocb *iocb, char __user *buf, size_t count, loff_t
 	if (retval)
 		goto out;
 
-	retval = nfs_direct_read(inode, ctx, &iov, pos, 1);
+	retval = nfs_get_user_pages(READ, (unsigned long) buf,
+						count, &pages);
+	if (retval < 0)
+		goto out;
+	page_count = retval;
+
+	retval = nfs_direct_read(iocb, (unsigned long) buf, count, pos,
+						pages, page_count);
 	if (retval > 0)
-		*ppos = pos + retval;
+		iocb->ki_pos = pos + retval;
 
 out:
 	return retval;
@@ -704,8 +810,8 @@ out:
  * nfs_file_direct_write - file direct write operation for NFS files
  * @iocb: target I/O control block
  * @buf: user's buffer from which to write data
- * count: number of bytes to write
- * pos: byte offset in file where writing starts
+ * @count: number of bytes to write
+ * @pos: byte offset in file where writing starts
  *
  * We use this function for direct writes instead of calling
  * generic_file_aio_write() in order to avoid taking the inode
@@ -725,28 +831,19 @@ out:
  * Note that O_APPEND is not supported for NFS direct writes, as there
  * is no atomic O_APPEND write facility in the NFS protocol.
  */
-ssize_t
-nfs_file_direct_write(struct kiocb *iocb, const char __user *buf, size_t count, loff_t pos)
+ssize_t nfs_file_direct_write(struct kiocb *iocb, const char __user *buf, size_t count, loff_t pos)
 {
 	ssize_t retval;
+	int page_count;
+	struct page **pages;
 	struct file *file = iocb->ki_filp;
-	struct nfs_open_context *ctx =
-			(struct nfs_open_context *) file->private_data;
 	struct address_space *mapping = file->f_mapping;
-	struct inode *inode = mapping->host;
-	struct iovec iov = {
-		.iov_base = (char __user *)buf,
-	};
 
 	dfprintk(VFS, "nfs: direct write(%s/%s, %lu@%Ld)\n",
 		file->f_dentry->d_parent->d_name.name,
 		file->f_dentry->d_name.name,
 		(unsigned long) count, (long long) pos);
 
-	retval = -EINVAL;
-	if (!is_sync_kiocb(iocb))
-		goto out;
-
 	retval = generic_write_checks(file, &pos, &count, 0);
 	if (retval)
 		goto out;
@@ -757,19 +854,35 @@ nfs_file_direct_write(struct kiocb *iocb, const char __user *buf, size_t count,
 	retval = 0;
 	if (!count)
 		goto out;
-	iov.iov_len = count,
 
 	retval = -EFAULT;
-	if (!access_ok(VERIFY_READ, iov.iov_base, iov.iov_len))
+	if (!access_ok(VERIFY_READ, buf, count))
 		goto out;
 
 	retval = nfs_sync_mapping(mapping);
 	if (retval)
 		goto out;
 
-	retval = nfs_direct_write(inode, ctx, &iov, pos, 1);
+	retval = nfs_get_user_pages(WRITE, (unsigned long) buf,
+						count, &pages);
+	if (retval < 0)
+		goto out;
+	page_count = retval;
+
+	retval = nfs_direct_write(iocb, (unsigned long) buf, count,
+					pos, pages, page_count);
+
+	/*
+	 * XXX: nfs_end_data_update() already ensures this file's
+	 *      cached data is subsequently invalidated.  Do we really
+	 *      need to call invalidate_inode_pages2() again here?
+	 *
+	 *      For aio writes, this invalidation will almost certainly
+	 *      occur before the writes complete.  Kind of racey.
+	 */
 	if (mapping->nrpages)
 		invalidate_inode_pages2(mapping);
+
 	if (retval > 0)
 		iocb->ki_pos = pos + retval;
 
@@ -777,6 +890,10 @@ out:
 	return retval;
 }
 
+/**
+ * nfs_init_directcache - create a slab cache for nfs_direct_req structures
+ *
+ */
 int nfs_init_directcache(void)
 {
 	nfs_direct_cachep = kmem_cache_create("nfs_direct_cache",
@@ -790,6 +907,10 @@ int nfs_init_directcache(void)
 	return 0;
 }
 
+/**
+ * nfs_init_directcache - destroy the slab cache for nfs_direct_req structures
+ *
+ */
 void nfs_destroy_directcache(void)
 {
 	if (kmem_cache_destroy(nfs_direct_cachep))

fs/nfs/file.c

@@ -32,6 +32,7 @@
 #include <asm/system.h>
 
 #include "delegation.h"
+#include "iostat.h"
 
 #define NFSDBG_FACILITY		NFSDBG_FILE
 
@@ -102,18 +103,15 @@ static int nfs_check_flags(int flags)
 static int
 nfs_file_open(struct inode *inode, struct file *filp)
 {
-	struct nfs_server *server = NFS_SERVER(inode);
-	int (*open)(struct inode *, struct file *);
 	int res;
 
 	res = nfs_check_flags(filp->f_flags);
 	if (res)
 		return res;
 
+	nfs_inc_stats(inode, NFSIOS_VFSOPEN);
 	lock_kernel();
-	/* Do NFSv4 open() call */
-	if ((open = server->rpc_ops->file_open) != NULL)
-		res = open(inode, filp);
+	res = NFS_SERVER(inode)->rpc_ops->file_open(inode, filp);
 	unlock_kernel();
 	return res;
 }
@@ -124,6 +122,7 @@ nfs_file_release(struct inode *inode, struct file *filp)
 	/* Ensure that dirty pages are flushed out with the right creds */
 	if (filp->f_mode & FMODE_WRITE)
 		filemap_fdatawrite(filp->f_mapping);
+	nfs_inc_stats(inode, NFSIOS_VFSRELEASE);
 	return NFS_PROTO(inode)->file_release(inode, filp);
 }
 
@@ -199,6 +198,7 @@ nfs_file_flush(struct file *file)
 
 	if ((file->f_mode & FMODE_WRITE) == 0)
 		return 0;
+	nfs_inc_stats(inode, NFSIOS_VFSFLUSH);
 	lock_kernel();
 	/* Ensure that data+attribute caches are up to date after close() */
 	status = nfs_wb_all(inode);
@@ -229,6 +229,7 @@ nfs_file_read(struct kiocb *iocb, char __user * buf, size_t count, loff_t pos)
 		(unsigned long) count, (unsigned long) pos);
 
 	result = nfs_revalidate_file(inode, iocb->ki_filp);
+	nfs_add_stats(inode, NFSIOS_NORMALREADBYTES, count);
 	if (!result)
 		result = generic_file_aio_read(iocb, buf, count, pos);
 	return result;
@@ -282,6 +283,7 @@ nfs_fsync(struct file *file, struct dentry *dentry, int datasync)
 
 	dfprintk(VFS, "nfs: fsync(%s/%ld)\n", inode->i_sb->s_id, inode->i_ino);
 
+	nfs_inc_stats(inode, NFSIOS_VFSFSYNC);
 	lock_kernel();
 	status = nfs_wb_all(inode);
 	if (!status) {
@@ -316,6 +318,17 @@ static int nfs_commit_write(struct file *file, struct page *page, unsigned offse
 	return status;
 }
 
+static int nfs_invalidate_page(struct page *page, unsigned long offset)
+{
+	/* FIXME: we really should cancel any unstarted writes on this page */
+	return 1;
+}
+
+static int nfs_release_page(struct page *page, gfp_t gfp)
+{
+	return !nfs_wb_page(page->mapping->host, page);
+}
+
 struct address_space_operations nfs_file_aops = {
 	.readpage = nfs_readpage,
 	.readpages = nfs_readpages,
@@ -324,6 +337,8 @@ struct address_space_operations nfs_file_aops = {
 	.writepages = nfs_writepages,
 	.prepare_write = nfs_prepare_write,
 	.commit_write = nfs_commit_write,
+	.invalidatepage = nfs_invalidate_page,
+	.releasepage = nfs_release_page,
 #ifdef CONFIG_NFS_DIRECTIO
 	.direct_IO = nfs_direct_IO,
 #endif
@@ -365,6 +380,7 @@ nfs_file_write(struct kiocb *iocb, const char __user *buf, size_t count, loff_t
 	if (!count)
 		goto out;
 
+	nfs_add_stats(inode, NFSIOS_NORMALWRITTENBYTES, count);
 	result = generic_file_aio_write(iocb, buf, count, pos);
 out:
 	return result;
@@ -376,15 +392,17 @@ out_swapfile:
 
 static int do_getlk(struct file *filp, int cmd, struct file_lock *fl)
 {
-	struct file_lock *cfl;
+	struct file_lock cfl;
 	struct inode *inode = filp->f_mapping->host;
 	int status = 0;
 
 	lock_kernel();
 	/* Try local locking first */
-	cfl = posix_test_lock(filp, fl);
-	if (cfl != NULL) {
-		locks_copy_lock(fl, cfl);
+	if (posix_test_lock(filp, fl, &cfl)) {
+		fl->fl_start = cfl.fl_start;
+		fl->fl_end = cfl.fl_end;
+		fl->fl_type = cfl.fl_type;
+		fl->fl_pid = cfl.fl_pid;
 		goto out;
 	}
 
@@ -425,10 +443,8 @@ static int do_vfs_lock(struct file *file, struct file_lock *fl)
 static int do_unlk(struct file *filp, int cmd, struct file_lock *fl)
 {
 	struct inode *inode = filp->f_mapping->host;
-	sigset_t oldset;
 	int status;
 
-	rpc_clnt_sigmask(NFS_CLIENT(inode), &oldset);
 	/*
 	 * Flush all pending writes before doing anything
 	 * with locks..
@@ -446,17 +462,14 @@ static int do_unlk(struct file *filp, int cmd, struct file_lock *fl)
 	else
 		status = do_vfs_lock(filp, fl);
 	unlock_kernel();
-	rpc_clnt_sigunmask(NFS_CLIENT(inode), &oldset);
 	return status;
 }
 
 static int do_setlk(struct file *filp, int cmd, struct file_lock *fl)
 {
 	struct inode *inode = filp->f_mapping->host;
-	sigset_t oldset;
 	int status;
 
-	rpc_clnt_sigmask(NFS_CLIENT(inode), &oldset);
 	/*
 	 * Flush all pending writes before doing anything
 	 * with locks..
@@ -489,7 +502,6 @@ static int do_setlk(struct file *filp, int cmd, struct file_lock *fl)
 	nfs_sync_mapping(filp->f_mapping);
 	nfs_zap_caches(inode);
 out:
-	rpc_clnt_sigunmask(NFS_CLIENT(inode), &oldset);
 	return status;
 }
 
@@ -504,9 +516,7 @@ static int nfs_lock(struct file *filp, int cmd, struct file_lock *fl)
 			inode->i_sb->s_id, inode->i_ino,
 			fl->fl_type, fl->fl_flags,
 			(long long)fl->fl_start, (long long)fl->fl_end);
-
-	if (!inode)
-		return -EINVAL;
+	nfs_inc_stats(inode, NFSIOS_VFSLOCK);
 
 	/* No mandatory locks over NFS */
 	if ((inode->i_mode & (S_ISGID | S_IXGRP)) == S_ISGID &&
@@ -531,9 +541,6 @@ static int nfs_flock(struct file *filp, int cmd, struct file_lock *fl)
 			inode->i_sb->s_id, inode->i_ino,
 			fl->fl_type, fl->fl_flags);
 
-	if (!inode)
-		return -EINVAL;
-
 	/*
 	 * No BSD flocks over NFS allowed.
 	 * Note: we could try to fake a POSIX lock request here by

fs/nfs/idmap.c

@@ -35,6 +35,7 @@
  */
 
 #include <linux/module.h>
+#include <linux/mutex.h>
 #include <linux/init.h>
 #include <linux/types.h>
 #include <linux/slab.h>
@@ -74,8 +75,8 @@ struct idmap {
 	struct dentry        *idmap_dentry;
 	wait_queue_head_t     idmap_wq;
 	struct idmap_msg      idmap_im;
-	struct semaphore      idmap_lock;    /* Serializes upcalls */
-	struct semaphore      idmap_im_lock; /* Protects the hashtable */
+	struct mutex          idmap_lock;    /* Serializes upcalls */
+	struct mutex          idmap_im_lock; /* Protects the hashtable */
 	struct idmap_hashtable idmap_user_hash;
 	struct idmap_hashtable idmap_group_hash;
 };
@@ -101,11 +102,9 @@ nfs_idmap_new(struct nfs4_client *clp)
 
 	if (clp->cl_idmap != NULL)
 		return;
-        if ((idmap = kmalloc(sizeof(*idmap), GFP_KERNEL)) == NULL)
+        if ((idmap = kzalloc(sizeof(*idmap), GFP_KERNEL)) == NULL)
                 return;
 
-	memset(idmap, 0, sizeof(*idmap));
-
 	snprintf(idmap->idmap_path, sizeof(idmap->idmap_path),
 	    "%s/idmap", clp->cl_rpcclient->cl_pathname);
 
@@ -116,8 +115,8 @@ nfs_idmap_new(struct nfs4_client *clp)
 		return;
 	}
 
-        init_MUTEX(&idmap->idmap_lock);
-        init_MUTEX(&idmap->idmap_im_lock);
+        mutex_init(&idmap->idmap_lock);
+        mutex_init(&idmap->idmap_im_lock);
 	init_waitqueue_head(&idmap->idmap_wq);
 	idmap->idmap_user_hash.h_type = IDMAP_TYPE_USER;
 	idmap->idmap_group_hash.h_type = IDMAP_TYPE_GROUP;
@@ -132,6 +131,8 @@ nfs_idmap_delete(struct nfs4_client *clp)
 
 	if (!idmap)
 		return;
+	dput(idmap->idmap_dentry);
+	idmap->idmap_dentry = NULL;
 	rpc_unlink(idmap->idmap_path);
 	clp->cl_idmap = NULL;
 	kfree(idmap);
@@ -232,8 +233,8 @@ nfs_idmap_id(struct idmap *idmap, struct idmap_hashtable *h,
 	if (namelen >= IDMAP_NAMESZ)
 		return -EINVAL;
 
-	down(&idmap->idmap_lock);
-	down(&idmap->idmap_im_lock);
+	mutex_lock(&idmap->idmap_lock);
+	mutex_lock(&idmap->idmap_im_lock);
 
 	he = idmap_lookup_name(h, name, namelen);
 	if (he != NULL) {
@@ -259,11 +260,11 @@ nfs_idmap_id(struct idmap *idmap, struct idmap_hashtable *h,
 	}
 
 	set_current_state(TASK_UNINTERRUPTIBLE);
-	up(&idmap->idmap_im_lock);
+	mutex_unlock(&idmap->idmap_im_lock);
 	schedule();
 	current->state = TASK_RUNNING;
 	remove_wait_queue(&idmap->idmap_wq, &wq);
-	down(&idmap->idmap_im_lock);
+	mutex_lock(&idmap->idmap_im_lock);
 
 	if (im->im_status & IDMAP_STATUS_SUCCESS) {
 		*id = im->im_id;
@@ -272,8 +273,8 @@ nfs_idmap_id(struct idmap *idmap, struct idmap_hashtable *h,
 
  out:
 	memset(im, 0, sizeof(*im));
-	up(&idmap->idmap_im_lock);
-	up(&idmap->idmap_lock);
+	mutex_unlock(&idmap->idmap_im_lock);
+	mutex_unlock(&idmap->idmap_lock);
 	return (ret);
 }
 
@@ -293,8 +294,8 @@ nfs_idmap_name(struct idmap *idmap, struct idmap_hashtable *h,
 
 	im = &idmap->idmap_im;
 
-	down(&idmap->idmap_lock);
-	down(&idmap->idmap_im_lock);
+	mutex_lock(&idmap->idmap_lock);
+	mutex_lock(&idmap->idmap_im_lock);
 
 	he = idmap_lookup_id(h, id);
 	if (he != 0) {
@@ -320,11 +321,11 @@ nfs_idmap_name(struct idmap *idmap, struct idmap_hashtable *h,
 	}
 
 	set_current_state(TASK_UNINTERRUPTIBLE);
-	up(&idmap->idmap_im_lock);
+	mutex_unlock(&idmap->idmap_im_lock);
 	schedule();
 	current->state = TASK_RUNNING;
 	remove_wait_queue(&idmap->idmap_wq, &wq);
-	down(&idmap->idmap_im_lock);
+	mutex_lock(&idmap->idmap_im_lock);
 
 	if (im->im_status & IDMAP_STATUS_SUCCESS) {
 		if ((len = strnlen(im->im_name, IDMAP_NAMESZ)) == 0)
@@ -335,8 +336,8 @@ nfs_idmap_name(struct idmap *idmap, struct idmap_hashtable *h,
 
  out:
 	memset(im, 0, sizeof(*im));
-	up(&idmap->idmap_im_lock);
-	up(&idmap->idmap_lock);
+	mutex_unlock(&idmap->idmap_im_lock);
+	mutex_unlock(&idmap->idmap_lock);
 	return ret;
 }
 
@@ -380,7 +381,7 @@ idmap_pipe_downcall(struct file *filp, const char __user *src, size_t mlen)
         if (copy_from_user(&im_in, src, mlen) != 0)
 		return (-EFAULT);
 
-	down(&idmap->idmap_im_lock);
+	mutex_lock(&idmap->idmap_im_lock);
 
 	ret = mlen;
 	im->im_status = im_in.im_status;
@@ -440,7 +441,7 @@ idmap_pipe_downcall(struct file *filp, const char __user *src, size_t mlen)
 		idmap_update_entry(he, im_in.im_name, namelen_in, im_in.im_id);
 	ret = mlen;
 out:
-	up(&idmap->idmap_im_lock);
+	mutex_unlock(&idmap->idmap_im_lock);
 	return ret;
 }
 
@@ -452,10 +453,10 @@ idmap_pipe_destroy_msg(struct rpc_pipe_msg *msg)
 
 	if (msg->errno >= 0)
 		return;
-	down(&idmap->idmap_im_lock);
+	mutex_lock(&idmap->idmap_im_lock);
 	im->im_status = IDMAP_STATUS_LOOKUPFAIL;
 	wake_up(&idmap->idmap_wq);
-	up(&idmap->idmap_im_lock);
+	mutex_unlock(&idmap->idmap_im_lock);
 }
 
 /* 

fs/nfs/inode.c

@@ -26,6 +26,7 @@
 #include <linux/unistd.h>
 #include <linux/sunrpc/clnt.h>
 #include <linux/sunrpc/stats.h>
+#include <linux/sunrpc/metrics.h>
 #include <linux/nfs_fs.h>
 #include <linux/nfs_mount.h>
 #include <linux/nfs4_mount.h>
@@ -42,6 +43,7 @@
 #include "nfs4_fs.h"
 #include "callback.h"
 #include "delegation.h"
+#include "iostat.h"
 
 #define NFSDBG_FACILITY		NFSDBG_VFS
 #define NFS_PARANOIA 1
@@ -65,6 +67,7 @@ static void nfs_clear_inode(struct inode *);
 static void nfs_umount_begin(struct super_block *);
 static int  nfs_statfs(struct super_block *, struct kstatfs *);
 static int  nfs_show_options(struct seq_file *, struct vfsmount *);
+static int  nfs_show_stats(struct seq_file *, struct vfsmount *);
 static void nfs_zap_acl_cache(struct inode *);
 
 static struct rpc_program	nfs_program;
@@ -78,6 +81,7 @@ static struct super_operations nfs_sops = {
 	.clear_inode	= nfs_clear_inode,
 	.umount_begin	= nfs_umount_begin,
 	.show_options	= nfs_show_options,
+	.show_stats	= nfs_show_stats,
 };
 
 /*
@@ -133,7 +137,7 @@ nfs_fattr_to_ino_t(struct nfs_fattr *fattr)
 static int
 nfs_write_inode(struct inode *inode, int sync)
 {
-	int flags = sync ? FLUSH_WAIT : 0;
+	int flags = sync ? FLUSH_SYNC : 0;
 	int ret;
 
 	ret = nfs_commit_inode(inode, flags);
@@ -237,7 +241,6 @@ static struct inode *
 nfs_get_root(struct super_block *sb, struct nfs_fh *rootfh, struct nfs_fsinfo *fsinfo)
 {
 	struct nfs_server	*server = NFS_SB(sb);
-	struct inode *rooti;
 	int			error;
 
 	error = server->rpc_ops->getroot(server, rootfh, fsinfo);
@@ -246,10 +249,7 @@ nfs_get_root(struct super_block *sb, struct nfs_fh *rootfh, struct nfs_fsinfo *f
 		return ERR_PTR(error);
 	}
 
-	rooti = nfs_fhget(sb, rootfh, fsinfo->fattr);
-	if (!rooti)
-		return ERR_PTR(-ENOMEM);
-	return rooti;
+	return nfs_fhget(sb, rootfh, fsinfo->fattr);
 }
 
 /*
@@ -277,6 +277,10 @@ nfs_sb_init(struct super_block *sb, rpc_authflavor_t authflavor)
 
 	sb->s_magic      = NFS_SUPER_MAGIC;
 
+	server->io_stats = nfs_alloc_iostats();
+	if (server->io_stats == NULL)
+		return -ENOMEM;
+
 	root_inode = nfs_get_root(sb, &server->fh, &fsinfo);
 	/* Did getting the root inode fail? */
 	if (IS_ERR(root_inode)) {
@@ -290,6 +294,9 @@ nfs_sb_init(struct super_block *sb, rpc_authflavor_t authflavor)
 	}
 	sb->s_root->d_op = server->rpc_ops->dentry_ops;
 
+	/* mount time stamp, in seconds */
+	server->mount_time = jiffies;
+
 	/* Get some general file system info */
 	if (server->namelen == 0 &&
 	    server->rpc_ops->pathconf(server, &server->fh, &pathinfo) >= 0)
@@ -396,6 +403,9 @@ nfs_create_client(struct nfs_server *server, const struct nfs_mount_data *data)
 
 	nfs_init_timeout_values(&timeparms, proto, data->timeo, data->retrans);
 
+	server->retrans_timeo = timeparms.to_initval;
+	server->retrans_count = timeparms.to_retries;
+
 	/* create transport and client */
 	xprt = xprt_create_proto(proto, &server->addr, &timeparms);
 	if (IS_ERR(xprt)) {
@@ -579,7 +589,7 @@ nfs_statfs(struct super_block *sb, struct kstatfs *buf)
 
 }
 
-static int nfs_show_options(struct seq_file *m, struct vfsmount *mnt)
+static void nfs_show_mount_options(struct seq_file *m, struct nfs_server *nfss, int showdefaults)
 {
 	static struct proc_nfs_info {
 		int flag;
@@ -588,28 +598,26 @@ static int nfs_show_options(struct seq_file *m, struct vfsmount *mnt)
 	} nfs_info[] = {
 		{ NFS_MOUNT_SOFT, ",soft", ",hard" },
 		{ NFS_MOUNT_INTR, ",intr", "" },
-		{ NFS_MOUNT_POSIX, ",posix", "" },
 		{ NFS_MOUNT_NOCTO, ",nocto", "" },
 		{ NFS_MOUNT_NOAC, ",noac", "" },
-		{ NFS_MOUNT_NONLM, ",nolock", ",lock" },
+		{ NFS_MOUNT_NONLM, ",nolock", "" },
 		{ NFS_MOUNT_NOACL, ",noacl", "" },
 		{ 0, NULL, NULL }
 	};
 	struct proc_nfs_info *nfs_infop;
-	struct nfs_server *nfss = NFS_SB(mnt->mnt_sb);
 	char buf[12];
 	char *proto;
 
-	seq_printf(m, ",v%d", nfss->rpc_ops->version);
+	seq_printf(m, ",vers=%d", nfss->rpc_ops->version);
 	seq_printf(m, ",rsize=%d", nfss->rsize);
 	seq_printf(m, ",wsize=%d", nfss->wsize);
-	if (nfss->acregmin != 3*HZ)
+	if (nfss->acregmin != 3*HZ || showdefaults)
 		seq_printf(m, ",acregmin=%d", nfss->acregmin/HZ);
-	if (nfss->acregmax != 60*HZ)
+	if (nfss->acregmax != 60*HZ || showdefaults)
 		seq_printf(m, ",acregmax=%d", nfss->acregmax/HZ);
-	if (nfss->acdirmin != 30*HZ)
+	if (nfss->acdirmin != 30*HZ || showdefaults)
 		seq_printf(m, ",acdirmin=%d", nfss->acdirmin/HZ);
-	if (nfss->acdirmax != 60*HZ)
+	if (nfss->acdirmax != 60*HZ || showdefaults)
 		seq_printf(m, ",acdirmax=%d", nfss->acdirmax/HZ);
 	for (nfs_infop = nfs_info; nfs_infop->flag; nfs_infop++) {
 		if (nfss->flags & nfs_infop->flag)
@@ -629,8 +637,96 @@ static int nfs_show_options(struct seq_file *m, struct vfsmount *mnt)
 			proto = buf;
 	}
 	seq_printf(m, ",proto=%s", proto);
+	seq_printf(m, ",timeo=%lu", 10U * nfss->retrans_timeo / HZ);
+	seq_printf(m, ",retrans=%u", nfss->retrans_count);
+}
+
+static int nfs_show_options(struct seq_file *m, struct vfsmount *mnt)
+{
+	struct nfs_server *nfss = NFS_SB(mnt->mnt_sb);
+
+	nfs_show_mount_options(m, nfss, 0);
+
 	seq_puts(m, ",addr=");
 	seq_escape(m, nfss->hostname, " \t\n\\");
+
+	return 0;
+}
+
+static int nfs_show_stats(struct seq_file *m, struct vfsmount *mnt)
+{
+	int i, cpu;
+	struct nfs_server *nfss = NFS_SB(mnt->mnt_sb);
+	struct rpc_auth *auth = nfss->client->cl_auth;
+	struct nfs_iostats totals = { };
+
+	seq_printf(m, "statvers=%s", NFS_IOSTAT_VERS);
+
+	/*
+	 * Display all mount option settings
+	 */
+	seq_printf(m, "\n\topts:\t");
+	seq_puts(m, mnt->mnt_sb->s_flags & MS_RDONLY ? "ro" : "rw");
+	seq_puts(m, mnt->mnt_sb->s_flags & MS_SYNCHRONOUS ? ",sync" : "");
+	seq_puts(m, mnt->mnt_sb->s_flags & MS_NOATIME ? ",noatime" : "");
+	seq_puts(m, mnt->mnt_sb->s_flags & MS_NODIRATIME ? ",nodiratime" : "");
+	nfs_show_mount_options(m, nfss, 1);
+
+	seq_printf(m, "\n\tage:\t%lu", (jiffies - nfss->mount_time) / HZ);
+
+	seq_printf(m, "\n\tcaps:\t");
+	seq_printf(m, "caps=0x%x", nfss->caps);
+	seq_printf(m, ",wtmult=%d", nfss->wtmult);
+	seq_printf(m, ",dtsize=%d", nfss->dtsize);
+	seq_printf(m, ",bsize=%d", nfss->bsize);
+	seq_printf(m, ",namelen=%d", nfss->namelen);
+
+#ifdef CONFIG_NFS_V4
+	if (nfss->rpc_ops->version == 4) {
+		seq_printf(m, "\n\tnfsv4:\t");
+		seq_printf(m, "bm0=0x%x", nfss->attr_bitmask[0]);
+		seq_printf(m, ",bm1=0x%x", nfss->attr_bitmask[1]);
+		seq_printf(m, ",acl=0x%x", nfss->acl_bitmask);
+	}
+#endif
+
+	/*
+	 * Display security flavor in effect for this mount
+	 */
+	seq_printf(m, "\n\tsec:\tflavor=%d", auth->au_ops->au_flavor);
+	if (auth->au_flavor)
+		seq_printf(m, ",pseudoflavor=%d", auth->au_flavor);
+
+	/*
+	 * Display superblock I/O counters
+	 */
+	for (cpu = 0; cpu < NR_CPUS; cpu++) {
+		struct nfs_iostats *stats;
+
+		if (!cpu_possible(cpu))
+			continue;
+
+		preempt_disable();
+		stats = per_cpu_ptr(nfss->io_stats, cpu);
+
+		for (i = 0; i < __NFSIOS_COUNTSMAX; i++)
+			totals.events[i] += stats->events[i];
+		for (i = 0; i < __NFSIOS_BYTESMAX; i++)
+			totals.bytes[i] += stats->bytes[i];
+
+		preempt_enable();
+	}
+
+	seq_printf(m, "\n\tevents:\t");
+	for (i = 0; i < __NFSIOS_COUNTSMAX; i++)
+		seq_printf(m, "%lu ", totals.events[i]);
+	seq_printf(m, "\n\tbytes:\t");
+	for (i = 0; i < __NFSIOS_BYTESMAX; i++)
+		seq_printf(m, "%Lu ", totals.bytes[i]);
+	seq_printf(m, "\n");
+
+	rpc_print_iostats(m, nfss->client);
+
 	return 0;
 }
 
@@ -660,6 +756,8 @@ static void nfs_zap_caches_locked(struct inode *inode)
 	struct nfs_inode *nfsi = NFS_I(inode);
 	int mode = inode->i_mode;
 
+	nfs_inc_stats(inode, NFSIOS_ATTRINVALIDATE);
+
 	NFS_ATTRTIMEO(inode) = NFS_MINATTRTIMEO(inode);
 	NFS_ATTRTIMEO_UPDATE(inode) = jiffies;
 
@@ -751,7 +849,7 @@ nfs_fhget(struct super_block *sb, struct nfs_fh *fh, struct nfs_fattr *fattr)
 		.fh	= fh,
 		.fattr	= fattr
 	};
-	struct inode *inode = NULL;
+	struct inode *inode = ERR_PTR(-ENOENT);
 	unsigned long hash;
 
 	if ((fattr->valid & NFS_ATTR_FATTR) == 0)
@@ -764,8 +862,11 @@ nfs_fhget(struct super_block *sb, struct nfs_fh *fh, struct nfs_fattr *fattr)
 
 	hash = nfs_fattr_to_ino_t(fattr);
 
-	if (!(inode = iget5_locked(sb, hash, nfs_find_actor, nfs_init_locked, &desc)))
+	inode = iget5_locked(sb, hash, nfs_find_actor, nfs_init_locked, &desc);
+	if (inode == NULL) {
+		inode = ERR_PTR(-ENOMEM);
 		goto out_no_inode;
+	}
 
 	if (inode->i_state & I_NEW) {
 		struct nfs_inode *nfsi = NFS_I(inode);
@@ -834,7 +935,7 @@ out:
 	return inode;
 
 out_no_inode:
-	printk("nfs_fhget: iget failed\n");
+	dprintk("nfs_fhget: iget failed with error %ld\n", PTR_ERR(inode));
 	goto out;
 }
 
@@ -847,6 +948,8 @@ nfs_setattr(struct dentry *dentry, struct iattr *attr)
 	struct nfs_fattr fattr;
 	int error;
 
+	nfs_inc_stats(inode, NFSIOS_VFSSETATTR);
+
 	if (attr->ia_valid & ATTR_SIZE) {
 		if (!S_ISREG(inode->i_mode) || attr->ia_size == i_size_read(inode))
 			attr->ia_valid &= ~ATTR_SIZE;
@@ -859,11 +962,9 @@ nfs_setattr(struct dentry *dentry, struct iattr *attr)
 
 	lock_kernel();
 	nfs_begin_data_update(inode);
-	/* Write all dirty data if we're changing file permissions or size */
-	if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID|ATTR_SIZE)) != 0) {
-		filemap_write_and_wait(inode->i_mapping);
-		nfs_wb_all(inode);
-	}
+	/* Write all dirty data */
+	filemap_write_and_wait(inode->i_mapping);
+	nfs_wb_all(inode);
 	/*
 	 * Return any delegations if we're going to change ACLs
 	 */
@@ -902,6 +1003,7 @@ void nfs_setattr_update_inode(struct inode *inode, struct iattr *attr)
 		spin_unlock(&inode->i_lock);
 	}
 	if ((attr->ia_valid & ATTR_SIZE) != 0) {
+		nfs_inc_stats(inode, NFSIOS_SETATTRTRUNC);
 		inode->i_size = attr->ia_size;
 		vmtruncate(inode, attr->ia_size);
 	}
@@ -949,7 +1051,7 @@ int nfs_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
 	int err;
 
 	/* Flush out writes to the server in order to update c/mtime */
-	nfs_sync_inode(inode, 0, 0, FLUSH_WAIT|FLUSH_NOCOMMIT);
+	nfs_sync_inode_wait(inode, 0, 0, FLUSH_NOCOMMIT);
 
 	/*
 	 * We may force a getattr if the user cares about atime.
@@ -973,7 +1075,7 @@ int nfs_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
 	return err;
 }
 
-struct nfs_open_context *alloc_nfs_open_context(struct dentry *dentry, struct rpc_cred *cred)
+static struct nfs_open_context *alloc_nfs_open_context(struct vfsmount *mnt, struct dentry *dentry, struct rpc_cred *cred)
 {
 	struct nfs_open_context *ctx;
 
@@ -981,6 +1083,7 @@ struct nfs_open_context *alloc_nfs_open_context(struct dentry *dentry, struct rp
 	if (ctx != NULL) {
 		atomic_set(&ctx->count, 1);
 		ctx->dentry = dget(dentry);
+		ctx->vfsmnt = mntget(mnt);
 		ctx->cred = get_rpccred(cred);
 		ctx->state = NULL;
 		ctx->lockowner = current->files;
@@ -1011,6 +1114,7 @@ void put_nfs_open_context(struct nfs_open_context *ctx)
 		if (ctx->cred != NULL)
 			put_rpccred(ctx->cred);
 		dput(ctx->dentry);
+		mntput(ctx->vfsmnt);
 		kfree(ctx);
 	}
 }
@@ -1019,7 +1123,7 @@ void put_nfs_open_context(struct nfs_open_context *ctx)
  * Ensure that mmap has a recent RPC credential for use when writing out
  * shared pages
  */
-void nfs_file_set_open_context(struct file *filp, struct nfs_open_context *ctx)
+static void nfs_file_set_open_context(struct file *filp, struct nfs_open_context *ctx)
 {
 	struct inode *inode = filp->f_dentry->d_inode;
 	struct nfs_inode *nfsi = NFS_I(inode);
@@ -1051,7 +1155,7 @@ struct nfs_open_context *nfs_find_open_context(struct inode *inode, struct rpc_c
 	return ctx;
 }
 
-void nfs_file_clear_open_context(struct file *filp)
+static void nfs_file_clear_open_context(struct file *filp)
 {
 	struct inode *inode = filp->f_dentry->d_inode;
 	struct nfs_open_context *ctx = (struct nfs_open_context *)filp->private_data;
@@ -1076,7 +1180,7 @@ int nfs_open(struct inode *inode, struct file *filp)
 	cred = rpcauth_lookupcred(NFS_CLIENT(inode)->cl_auth, 0);
 	if (IS_ERR(cred))
 		return PTR_ERR(cred);
-	ctx = alloc_nfs_open_context(filp->f_dentry, cred);
+	ctx = alloc_nfs_open_context(filp->f_vfsmnt, filp->f_dentry, cred);
 	put_rpccred(cred);
 	if (ctx == NULL)
 		return -ENOMEM;
@@ -1185,6 +1289,7 @@ int nfs_attribute_timeout(struct inode *inode)
  */
 int nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
 {
+	nfs_inc_stats(inode, NFSIOS_INODEREVALIDATE);
 	if (!(NFS_I(inode)->cache_validity & (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA))
 			&& !nfs_attribute_timeout(inode))
 		return NFS_STALE(inode) ? -ESTALE : 0;
@@ -1201,6 +1306,7 @@ void nfs_revalidate_mapping(struct inode *inode, struct address_space *mapping)
 	struct nfs_inode *nfsi = NFS_I(inode);
 
 	if (nfsi->cache_validity & NFS_INO_INVALID_DATA) {
+		nfs_inc_stats(inode, NFSIOS_DATAINVALIDATE);
 		if (S_ISREG(inode->i_mode))
 			nfs_sync_mapping(mapping);
 		invalidate_inode_pages2(mapping);
@@ -1299,39 +1405,37 @@ static int nfs_check_inode_attributes(struct inode *inode, struct nfs_fattr *fat
 	if ((fattr->valid & NFS_ATTR_FATTR) == 0)
 		return 0;
 
+	/* Has the inode gone and changed behind our back? */
+	if (nfsi->fileid != fattr->fileid
+			|| (inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT)) {
+		return -EIO;
+	}
+
 	/* Are we in the process of updating data on the server? */
 	data_unstable = nfs_caches_unstable(inode);
 
 	/* Do atomic weak cache consistency updates */
 	nfs_wcc_update_inode(inode, fattr);
 
-	if ((fattr->valid & NFS_ATTR_FATTR_V4) != 0 &&
-			nfsi->change_attr != fattr->change_attr) {
+	if ((fattr->valid & NFS_ATTR_FATTR_V4) != 0) {
+		if (nfsi->change_attr == fattr->change_attr)
+			goto out;
 		nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
 		if (!data_unstable)
 			nfsi->cache_validity |= NFS_INO_REVAL_PAGECACHE;
 	}
 
-	/* Has the inode gone and changed behind our back? */
-	if (nfsi->fileid != fattr->fileid
-			|| (inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT)) {
-		return -EIO;
-	}
-
-	cur_size = i_size_read(inode);
- 	new_isize = nfs_size_to_loff_t(fattr->size);
-
 	/* Verify a few of the more important attributes */
 	if (!timespec_equal(&inode->i_mtime, &fattr->mtime)) {
 		nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
 		if (!data_unstable)
 			nfsi->cache_validity |= NFS_INO_REVAL_PAGECACHE;
 	}
-	if (cur_size != new_isize) {
-		nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
-		if (nfsi->npages == 0)
-			nfsi->cache_validity |= NFS_INO_REVAL_PAGECACHE;
-	}
+
+	cur_size = i_size_read(inode);
+ 	new_isize = nfs_size_to_loff_t(fattr->size);
+	if (cur_size != new_isize && nfsi->npages == 0)
+		nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE;
 
 	/* Have any file permissions changed? */
 	if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO)
@@ -1343,6 +1447,7 @@ static int nfs_check_inode_attributes(struct inode *inode, struct nfs_fattr *fat
 	if (inode->i_nlink != fattr->nlink)
 		nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
 
+out:
 	if (!timespec_equal(&inode->i_atime, &fattr->atime))
 		nfsi->cache_validity |= NFS_INO_INVALID_ATIME;
 
@@ -1481,15 +1586,6 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr)
 		nfsi->cache_change_attribute = jiffies;
 	}
 
-	if ((fattr->valid & NFS_ATTR_FATTR_V4)
-	    && nfsi->change_attr != fattr->change_attr) {
-		dprintk("NFS: change_attr change on server for file %s/%ld\n",
-		       inode->i_sb->s_id, inode->i_ino);
-		nfsi->change_attr = fattr->change_attr;
-		invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
-		nfsi->cache_change_attribute = jiffies;
-	}
-
 	/* If ctime has changed we should definitely clear access+acl caches */
 	if (!timespec_equal(&inode->i_ctime, &fattr->ctime)) {
 		invalid |= NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
@@ -1519,8 +1615,20 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr)
  		inode->i_blksize = fattr->du.nfs2.blocksize;
  	}
 
+	if ((fattr->valid & NFS_ATTR_FATTR_V4)) {
+		if (nfsi->change_attr != fattr->change_attr) {
+			dprintk("NFS: change_attr change on server for file %s/%ld\n",
+					inode->i_sb->s_id, inode->i_ino);
+			nfsi->change_attr = fattr->change_attr;
+			invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
+			nfsi->cache_change_attribute = jiffies;
+		} else
+			invalid &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA);
+	}
+
 	/* Update attrtimeo value if we're out of the unstable period */
 	if (invalid & NFS_INO_INVALID_ATTR) {
+		nfs_inc_stats(inode, NFSIOS_ATTRINVALIDATE);
 		nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
 		nfsi->attrtimeo_timestamp = jiffies;
 	} else if (time_after(jiffies, nfsi->attrtimeo_timestamp+nfsi->attrtimeo)) {
@@ -1637,10 +1745,9 @@ static struct super_block *nfs_get_sb(struct file_system_type *fs_type,
 #endif /* CONFIG_NFS_V3 */
 
 	s = ERR_PTR(-ENOMEM);
-	server = kmalloc(sizeof(struct nfs_server), GFP_KERNEL);
+	server = kzalloc(sizeof(struct nfs_server), GFP_KERNEL);
 	if (!server)
 		goto out_err;
-	memset(server, 0, sizeof(struct nfs_server));
 	/* Zero out the NFS state stuff */
 	init_nfsv4_state(server);
 	server->client = server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
@@ -1712,6 +1819,7 @@ static void nfs_kill_super(struct super_block *s)
 
 	rpciod_down();		/* release rpciod */
 
+	nfs_free_iostats(server->io_stats);
 	kfree(server->hostname);
 	kfree(server);
 }
@@ -1738,6 +1846,7 @@ static struct super_operations nfs4_sops = {
 	.clear_inode	= nfs4_clear_inode,
 	.umount_begin	= nfs_umount_begin,
 	.show_options	= nfs_show_options,
+	.show_stats	= nfs_show_stats,
 };
 
 /*
@@ -1800,6 +1909,9 @@ static int nfs4_fill_super(struct super_block *sb, struct nfs4_mount_data *data,
 
 	nfs_init_timeout_values(&timeparms, data->proto, data->timeo, data->retrans);
 
+	server->retrans_timeo = timeparms.to_initval;
+	server->retrans_count = timeparms.to_retries;
+
 	clp = nfs4_get_client(&server->addr.sin_addr);
 	if (!clp) {
 		dprintk("%s: failed to create NFS4 client.\n", __FUNCTION__);
@@ -1941,10 +2053,9 @@ static struct super_block *nfs4_get_sb(struct file_system_type *fs_type,
 		return ERR_PTR(-EINVAL);
 	}
 
-	server = kmalloc(sizeof(struct nfs_server), GFP_KERNEL);
+	server = kzalloc(sizeof(struct nfs_server), GFP_KERNEL);
 	if (!server)
 		return ERR_PTR(-ENOMEM);
-	memset(server, 0, sizeof(struct nfs_server));
 	/* Zero out the NFS state stuff */
 	init_nfsv4_state(server);
 	server->client = server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
@@ -2024,10 +2135,12 @@ static void nfs4_kill_super(struct super_block *sb)
 
 	if (server->client != NULL && !IS_ERR(server->client))
 		rpc_shutdown_client(server->client);
-	rpciod_down();		/* release rpciod */
 
 	destroy_nfsv4_state(server);
 
+	rpciod_down();
+
+	nfs_free_iostats(server->io_stats);
 	kfree(server->hostname);
 	kfree(server);
 }

fs/nfs/iostat.h

@@ -0,0 +1,164 @@
+/*
+ *  linux/fs/nfs/iostat.h
+ *
+ *  Declarations for NFS client per-mount statistics
+ *
+ *  Copyright (C) 2005, 2006 Chuck Lever <cel@netapp.com>
+ *
+ *  NFS client per-mount statistics provide information about the health of
+ *  the NFS client and the health of each NFS mount point.  Generally these
+ *  are not for detailed problem diagnosis, but simply to indicate that there
+ *  is a problem.
+ *
+ *  These counters are not meant to be human-readable, but are meant to be
+ *  integrated into system monitoring tools such as "sar" and "iostat".  As
+ *  such, the counters are sampled by the tools over time, and are never
+ *  zeroed after a file system is mounted.  Moving averages can be computed
+ *  by the tools by taking the difference between two instantaneous samples
+ *  and dividing that by the time between the samples.
+ */
+
+#ifndef _NFS_IOSTAT
+#define _NFS_IOSTAT
+
+#define NFS_IOSTAT_VERS		"1.0"
+
+/*
+ * NFS byte counters
+ *
+ * 1.  SERVER - the number of payload bytes read from or written to the
+ *     server by the NFS client via an NFS READ or WRITE request.
+ *
+ * 2.  NORMAL - the number of bytes read or written by applications via
+ *     the read(2) and write(2) system call interfaces.
+ *
+ * 3.  DIRECT - the number of bytes read or written from files opened
+ *     with the O_DIRECT flag.
+ *
+ * These counters give a view of the data throughput into and out of the NFS
+ * client.  Comparing the number of bytes requested by an application with the
+ * number of bytes the client requests from the server can provide an
+ * indication of client efficiency (per-op, cache hits, etc).
+ *
+ * These counters can also help characterize which access methods are in
+ * use.  DIRECT by itself shows whether there is any O_DIRECT traffic.
+ * NORMAL + DIRECT shows how much data is going through the system call
+ * interface.  A large amount of SERVER traffic without much NORMAL or
+ * DIRECT traffic shows that applications are using mapped files.
+ *
+ * NFS page counters
+ *
+ * These count the number of pages read or written via nfs_readpage(),
+ * nfs_readpages(), or their write equivalents.
+ */
+enum nfs_stat_bytecounters {
+	NFSIOS_NORMALREADBYTES = 0,
+	NFSIOS_NORMALWRITTENBYTES,
+	NFSIOS_DIRECTREADBYTES,
+	NFSIOS_DIRECTWRITTENBYTES,
+	NFSIOS_SERVERREADBYTES,
+	NFSIOS_SERVERWRITTENBYTES,
+	NFSIOS_READPAGES,
+	NFSIOS_WRITEPAGES,
+	__NFSIOS_BYTESMAX,
+};
+
+/*
+ * NFS event counters
+ *
+ * These counters provide a low-overhead way of monitoring client activity
+ * without enabling NFS trace debugging.  The counters show the rate at
+ * which VFS requests are made, and how often the client invalidates its
+ * data and attribute caches.  This allows system administrators to monitor
+ * such things as how close-to-open is working, and answer questions such
+ * as "why are there so many GETATTR requests on the wire?"
+ *
+ * They also count anamolous events such as short reads and writes, silly
+ * renames due to close-after-delete, and operations that change the size
+ * of a file (such operations can often be the source of data corruption
+ * if applications aren't using file locking properly).
+ */
+enum nfs_stat_eventcounters {
+	NFSIOS_INODEREVALIDATE = 0,
+	NFSIOS_DENTRYREVALIDATE,
+	NFSIOS_DATAINVALIDATE,
+	NFSIOS_ATTRINVALIDATE,
+	NFSIOS_VFSOPEN,
+	NFSIOS_VFSLOOKUP,
+	NFSIOS_VFSACCESS,
+	NFSIOS_VFSUPDATEPAGE,
+	NFSIOS_VFSREADPAGE,
+	NFSIOS_VFSREADPAGES,
+	NFSIOS_VFSWRITEPAGE,
+	NFSIOS_VFSWRITEPAGES,
+	NFSIOS_VFSGETDENTS,
+	NFSIOS_VFSSETATTR,
+	NFSIOS_VFSFLUSH,
+	NFSIOS_VFSFSYNC,
+	NFSIOS_VFSLOCK,
+	NFSIOS_VFSRELEASE,
+	NFSIOS_CONGESTIONWAIT,
+	NFSIOS_SETATTRTRUNC,
+	NFSIOS_EXTENDWRITE,
+	NFSIOS_SILLYRENAME,
+	NFSIOS_SHORTREAD,
+	NFSIOS_SHORTWRITE,
+	NFSIOS_DELAY,
+	__NFSIOS_COUNTSMAX,
+};
+
+#ifdef __KERNEL__
+
+#include <linux/percpu.h>
+#include <linux/cache.h>
+
+struct nfs_iostats {
+	unsigned long long	bytes[__NFSIOS_BYTESMAX];
+	unsigned long		events[__NFSIOS_COUNTSMAX];
+} ____cacheline_aligned;
+
+static inline void nfs_inc_server_stats(struct nfs_server *server, enum nfs_stat_eventcounters stat)
+{
+	struct nfs_iostats *iostats;
+	int cpu;
+
+	cpu = get_cpu();
+	iostats = per_cpu_ptr(server->io_stats, cpu);
+	iostats->events[stat] ++;
+	put_cpu_no_resched();
+}
+
+static inline void nfs_inc_stats(struct inode *inode, enum nfs_stat_eventcounters stat)
+{
+	nfs_inc_server_stats(NFS_SERVER(inode), stat);
+}
+
+static inline void nfs_add_server_stats(struct nfs_server *server, enum nfs_stat_bytecounters stat, unsigned long addend)
+{
+	struct nfs_iostats *iostats;
+	int cpu;
+
+	cpu = get_cpu();
+	iostats = per_cpu_ptr(server->io_stats, cpu);
+	iostats->bytes[stat] += addend;
+	put_cpu_no_resched();
+}
+
+static inline void nfs_add_stats(struct inode *inode, enum nfs_stat_bytecounters stat, unsigned long addend)
+{
+	nfs_add_server_stats(NFS_SERVER(inode), stat, addend);
+}
+
+static inline struct nfs_iostats *nfs_alloc_iostats(void)
+{
+	return alloc_percpu(struct nfs_iostats);
+}
+
+static inline void nfs_free_iostats(struct nfs_iostats *stats)
+{
+	if (stats != NULL)
+		free_percpu(stats);
+}
+
+#endif
+#endif

fs/nfs/mount_clnt.c

@@ -49,9 +49,12 @@ nfsroot_mount(struct sockaddr_in *addr, char *path, struct nfs_fh *fh,
 	struct mnt_fhstatus	result = {
 		.fh		= fh
 	};
+	struct rpc_message msg	= {
+		.rpc_argp	= path,
+		.rpc_resp	= &result,
+	};
 	char			hostname[32];
 	int			status;
-	int			call;
 
 	dprintk("NFS:      nfs_mount(%08x:%s)\n",
 			(unsigned)ntohl(addr->sin_addr.s_addr), path);
@@ -61,8 +64,12 @@ nfsroot_mount(struct sockaddr_in *addr, char *path, struct nfs_fh *fh,
 	if (IS_ERR(mnt_clnt))
 		return PTR_ERR(mnt_clnt);
 
-	call = (version == NFS_MNT3_VERSION) ? MOUNTPROC3_MNT : MNTPROC_MNT;
-	status = rpc_call(mnt_clnt, call, path, &result, 0);
+	if (version == NFS_MNT3_VERSION)
+		msg.rpc_proc = &mnt_clnt->cl_procinfo[MOUNTPROC3_MNT];
+	else
+		msg.rpc_proc = &mnt_clnt->cl_procinfo[MNTPROC_MNT];
+
+	status = rpc_call_sync(mnt_clnt, &msg, 0);
 	return status < 0? status : (result.status? -EACCES : 0);
 }
 
@@ -137,6 +144,8 @@ static struct rpc_procinfo	mnt_procedures[] = {
 	  .p_encode		= (kxdrproc_t) xdr_encode_dirpath,	
 	  .p_decode		= (kxdrproc_t) xdr_decode_fhstatus,
 	  .p_bufsiz		= MNT_dirpath_sz << 2,
+	  .p_statidx		= MNTPROC_MNT,
+	  .p_name		= "MOUNT",
 	},
 };
 
@@ -146,6 +155,8 @@ static struct rpc_procinfo mnt3_procedures[] = {
 	  .p_encode		= (kxdrproc_t) xdr_encode_dirpath,
 	  .p_decode		= (kxdrproc_t) xdr_decode_fhstatus3,
 	  .p_bufsiz		= MNT_dirpath_sz << 2,
+	  .p_statidx		= MOUNTPROC3_MNT,
+	  .p_name		= "MOUNT",
 	},
 };
 

fs/nfs/nfs2xdr.c

@@ -682,7 +682,9 @@ nfs_stat_to_errno(int stat)
 	.p_encode   =  (kxdrproc_t) nfs_xdr_##argtype,			\
 	.p_decode   =  (kxdrproc_t) nfs_xdr_##restype,			\
 	.p_bufsiz   =  MAX(NFS_##argtype##_sz,NFS_##restype##_sz) << 2,	\
-	.p_timer    =  timer						\
+	.p_timer    =  timer,						\
+	.p_statidx  =  NFSPROC_##proc,					\
+	.p_name     =  #proc,						\
 	}
 struct rpc_procinfo	nfs_procedures[] = {
     PROC(GETATTR,	fhandle,	attrstat, 1),

fs/nfs/nfs3acl.c

@@ -190,6 +190,10 @@ struct posix_acl *nfs3_proc_getacl(struct inode *inode, int type)
 	struct nfs3_getaclres res = {
 		.fattr =	&fattr,
 	};
+	struct rpc_message msg = {
+		.rpc_argp	= &args,
+		.rpc_resp	= &res,
+	};
 	struct posix_acl *acl;
 	int status, count;
 
@@ -218,8 +222,8 @@ struct posix_acl *nfs3_proc_getacl(struct inode *inode, int type)
 		return NULL;
 
 	dprintk("NFS call getacl\n");
-	status = rpc_call(server->client_acl, ACLPROC3_GETACL,
-			  &args, &res, 0);
+	msg.rpc_proc = &server->client_acl->cl_procinfo[ACLPROC3_GETACL];
+	status = rpc_call_sync(server->client_acl, &msg, 0);
 	dprintk("NFS reply getacl: %d\n", status);
 
 	/* pages may have been allocated at the xdr layer. */
@@ -286,6 +290,10 @@ static int nfs3_proc_setacls(struct inode *inode, struct posix_acl *acl,
 		.acl_access = acl,
 		.pages = pages,
 	};
+	struct rpc_message msg = {
+		.rpc_argp	= &args,
+		.rpc_resp	= &fattr,
+	};
 	int status, count;
 
 	status = -EOPNOTSUPP;
@@ -306,8 +314,8 @@ static int nfs3_proc_setacls(struct inode *inode, struct posix_acl *acl,
 
 	dprintk("NFS call setacl\n");
 	nfs_begin_data_update(inode);
-	status = rpc_call(server->client_acl, ACLPROC3_SETACL,
-			  &args, &fattr, 0);
+	msg.rpc_proc = &server->client_acl->cl_procinfo[ACLPROC3_SETACL];
+	status = rpc_call_sync(server->client_acl, &msg, 0);
 	spin_lock(&inode->i_lock);
 	NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ACCESS;
 	spin_unlock(&inode->i_lock);

fs/nfs/nfs3proc.c

@@ -19,6 +19,8 @@
 #include <linux/smp_lock.h>
 #include <linux/nfs_mount.h>
 
+#include "iostat.h"
+
 #define NFSDBG_FACILITY		NFSDBG_PROC
 
 extern struct rpc_procinfo nfs3_procedures[];
@@ -41,27 +43,14 @@ nfs3_rpc_wrapper(struct rpc_clnt *clnt, struct rpc_message *msg, int flags)
 	return res;
 }
 
-static inline int
-nfs3_rpc_call_wrapper(struct rpc_clnt *clnt, u32 proc, void *argp, void *resp, int flags)
-{
-	struct rpc_message msg = {
-		.rpc_proc	= &clnt->cl_procinfo[proc],
-		.rpc_argp	= argp,
-		.rpc_resp	= resp,
-	};
-	return nfs3_rpc_wrapper(clnt, &msg, flags);
-}
-
-#define rpc_call(clnt, proc, argp, resp, flags) \
-		nfs3_rpc_call_wrapper(clnt, proc, argp, resp, flags)
-#define rpc_call_sync(clnt, msg, flags) \
-		nfs3_rpc_wrapper(clnt, msg, flags)
+#define rpc_call_sync(clnt, msg, flags)	nfs3_rpc_wrapper(clnt, msg, flags)
 
 static int
-nfs3_async_handle_jukebox(struct rpc_task *task)
+nfs3_async_handle_jukebox(struct rpc_task *task, struct inode *inode)
 {
 	if (task->tk_status != -EJUKEBOX)
 		return 0;
+	nfs_inc_stats(inode, NFSIOS_DELAY);
 	task->tk_status = 0;
 	rpc_restart_call(task);
 	rpc_delay(task, NFS_JUKEBOX_RETRY_TIME);
@@ -72,14 +61,21 @@ static int
 do_proc_get_root(struct rpc_clnt *client, struct nfs_fh *fhandle,
 		 struct nfs_fsinfo *info)
 {
+	struct rpc_message msg = {
+		.rpc_proc	= &nfs3_procedures[NFS3PROC_FSINFO],
+		.rpc_argp	= fhandle,
+		.rpc_resp	= info,
+	};
 	int	status;
 
 	dprintk("%s: call  fsinfo\n", __FUNCTION__);
 	nfs_fattr_init(info->fattr);
-	status = rpc_call(client, NFS3PROC_FSINFO, fhandle, info, 0);
+	status = rpc_call_sync(client, &msg, 0);
 	dprintk("%s: reply fsinfo: %d\n", __FUNCTION__, status);
 	if (!(info->fattr->valid & NFS_ATTR_FATTR)) {
-		status = rpc_call(client, NFS3PROC_GETATTR, fhandle, info->fattr, 0);
+		msg.rpc_proc = &nfs3_procedures[NFS3PROC_GETATTR];
+		msg.rpc_resp = info->fattr;
+		status = rpc_call_sync(client, &msg, 0);
 		dprintk("%s: reply getattr: %d\n", __FUNCTION__, status);
 	}
 	return status;
@@ -107,12 +103,16 @@ static int
 nfs3_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
 		struct nfs_fattr *fattr)
 {
+	struct rpc_message msg = {
+		.rpc_proc	= &nfs3_procedures[NFS3PROC_GETATTR],
+		.rpc_argp	= fhandle,
+		.rpc_resp	= fattr,
+	};
 	int	status;
 
 	dprintk("NFS call  getattr\n");
 	nfs_fattr_init(fattr);
-	status = rpc_call(server->client, NFS3PROC_GETATTR,
-			  fhandle, fattr, 0);
+	status = rpc_call_sync(server->client, &msg, 0);
 	dprintk("NFS reply getattr: %d\n", status);
 	return status;
 }
@@ -126,11 +126,16 @@ nfs3_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
 		.fh		= NFS_FH(inode),
 		.sattr		= sattr,
 	};
+	struct rpc_message msg = {
+		.rpc_proc	= &nfs3_procedures[NFS3PROC_SETATTR],
+		.rpc_argp	= &arg,
+		.rpc_resp	= fattr,
+	};
 	int	status;
 
 	dprintk("NFS call  setattr\n");
 	nfs_fattr_init(fattr);
-	status = rpc_call(NFS_CLIENT(inode), NFS3PROC_SETATTR, &arg, fattr, 0);
+	status = rpc_call_sync(NFS_CLIENT(inode), &msg, 0);
 	if (status == 0)
 		nfs_setattr_update_inode(inode, sattr);
 	dprintk("NFS reply setattr: %d\n", status);
@@ -152,15 +157,23 @@ nfs3_proc_lookup(struct inode *dir, struct qstr *name,
 		.fh		= fhandle,
 		.fattr		= fattr
 	};
+	struct rpc_message msg = {
+		.rpc_proc	= &nfs3_procedures[NFS3PROC_LOOKUP],
+		.rpc_argp	= &arg,
+		.rpc_resp	= &res,
+	};
 	int			status;
 
 	dprintk("NFS call  lookup %s\n", name->name);
 	nfs_fattr_init(&dir_attr);
 	nfs_fattr_init(fattr);
-	status = rpc_call(NFS_CLIENT(dir), NFS3PROC_LOOKUP, &arg, &res, 0);
-	if (status >= 0 && !(fattr->valid & NFS_ATTR_FATTR))
-		status = rpc_call(NFS_CLIENT(dir), NFS3PROC_GETATTR,
-			 fhandle, fattr, 0);
+	status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0);
+	if (status >= 0 && !(fattr->valid & NFS_ATTR_FATTR)) {
+		msg.rpc_proc = &nfs3_procedures[NFS3PROC_GETATTR];
+		msg.rpc_argp = fhandle;
+		msg.rpc_resp = fattr;
+		status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0);
+	}
 	dprintk("NFS reply lookup: %d\n", status);
 	if (status >= 0)
 		status = nfs_refresh_inode(dir, &dir_attr);
@@ -180,7 +193,7 @@ static int nfs3_proc_access(struct inode *inode, struct nfs_access_entry *entry)
 		.rpc_proc	= &nfs3_procedures[NFS3PROC_ACCESS],
 		.rpc_argp	= &arg,
 		.rpc_resp	= &res,
-		.rpc_cred	= entry->cred
+		.rpc_cred	= entry->cred,
 	};
 	int mode = entry->mask;
 	int status;
@@ -226,12 +239,16 @@ static int nfs3_proc_readlink(struct inode *inode, struct page *page,
 		.pglen		= pglen,
 		.pages		= &page
 	};
+	struct rpc_message msg = {
+		.rpc_proc	= &nfs3_procedures[NFS3PROC_READLINK],
+		.rpc_argp	= &args,
+		.rpc_resp	= &fattr,
+	};
 	int			status;
 
 	dprintk("NFS call  readlink\n");
 	nfs_fattr_init(&fattr);
-	status = rpc_call(NFS_CLIENT(inode), NFS3PROC_READLINK,
-			  &args, &fattr, 0);
+	status = rpc_call_sync(NFS_CLIENT(inode), &msg, 0);
 	nfs_refresh_inode(inode, &fattr);
 	dprintk("NFS reply readlink: %d\n", status);
 	return status;
@@ -327,6 +344,11 @@ nfs3_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
 		.fh		= &fhandle,
 		.fattr		= &fattr
 	};
+	struct rpc_message msg = {
+		.rpc_proc	= &nfs3_procedures[NFS3PROC_CREATE],
+		.rpc_argp	= &arg,
+		.rpc_resp	= &res,
+	};
 	mode_t mode = sattr->ia_mode;
 	int status;
 
@@ -343,8 +365,8 @@ nfs3_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
 again:
 	nfs_fattr_init(&dir_attr);
 	nfs_fattr_init(&fattr);
-	status = rpc_call(NFS_CLIENT(dir), NFS3PROC_CREATE, &arg, &res, 0);
-	nfs_post_op_update_inode(dir, &dir_attr);
+	status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0);
+	nfs_refresh_inode(dir, &dir_attr);
 
 	/* If the server doesn't support the exclusive creation semantics,
 	 * try again with simple 'guarded' mode. */
@@ -447,7 +469,7 @@ nfs3_proc_unlink_done(struct dentry *dir, struct rpc_task *task)
 	struct rpc_message *msg = &task->tk_msg;
 	struct nfs_fattr	*dir_attr;
 
-	if (nfs3_async_handle_jukebox(task))
+	if (nfs3_async_handle_jukebox(task, dir->d_inode))
 		return 1;
 	if (msg->rpc_argp) {
 		dir_attr = (struct nfs_fattr*)msg->rpc_resp;
@@ -474,12 +496,17 @@ nfs3_proc_rename(struct inode *old_dir, struct qstr *old_name,
 		.fromattr	= &old_dir_attr,
 		.toattr		= &new_dir_attr
 	};
+	struct rpc_message msg = {
+		.rpc_proc	= &nfs3_procedures[NFS3PROC_RENAME],
+		.rpc_argp	= &arg,
+		.rpc_resp	= &res,
+	};
 	int			status;
 
 	dprintk("NFS call  rename %s -> %s\n", old_name->name, new_name->name);
 	nfs_fattr_init(&old_dir_attr);
 	nfs_fattr_init(&new_dir_attr);
-	status = rpc_call(NFS_CLIENT(old_dir), NFS3PROC_RENAME, &arg, &res, 0);
+	status = rpc_call_sync(NFS_CLIENT(old_dir), &msg, 0);
 	nfs_post_op_update_inode(old_dir, &old_dir_attr);
 	nfs_post_op_update_inode(new_dir, &new_dir_attr);
 	dprintk("NFS reply rename: %d\n", status);
@@ -500,12 +527,17 @@ nfs3_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
 		.dir_attr	= &dir_attr,
 		.fattr		= &fattr
 	};
+	struct rpc_message msg = {
+		.rpc_proc	= &nfs3_procedures[NFS3PROC_LINK],
+		.rpc_argp	= &arg,
+		.rpc_resp	= &res,
+	};
 	int			status;
 
 	dprintk("NFS call  link %s\n", name->name);
 	nfs_fattr_init(&dir_attr);
 	nfs_fattr_init(&fattr);
-	status = rpc_call(NFS_CLIENT(inode), NFS3PROC_LINK, &arg, &res, 0);
+	status = rpc_call_sync(NFS_CLIENT(inode), &msg, 0);
 	nfs_post_op_update_inode(dir, &dir_attr);
 	nfs_post_op_update_inode(inode, &fattr);
 	dprintk("NFS reply link: %d\n", status);
@@ -531,6 +563,11 @@ nfs3_proc_symlink(struct inode *dir, struct qstr *name, struct qstr *path,
 		.fh		= fhandle,
 		.fattr		= fattr
 	};
+	struct rpc_message msg = {
+		.rpc_proc	= &nfs3_procedures[NFS3PROC_SYMLINK],
+		.rpc_argp	= &arg,
+		.rpc_resp	= &res,
+	};
 	int			status;
 
 	if (path->len > NFS3_MAXPATHLEN)
@@ -538,7 +575,7 @@ nfs3_proc_symlink(struct inode *dir, struct qstr *name, struct qstr *path,
 	dprintk("NFS call  symlink %s -> %s\n", name->name, path->name);
 	nfs_fattr_init(&dir_attr);
 	nfs_fattr_init(fattr);
-	status = rpc_call(NFS_CLIENT(dir), NFS3PROC_SYMLINK, &arg, &res, 0);
+	status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0);
 	nfs_post_op_update_inode(dir, &dir_attr);
 	dprintk("NFS reply symlink: %d\n", status);
 	return status;
@@ -560,6 +597,11 @@ nfs3_proc_mkdir(struct inode *dir, struct dentry *dentry, struct iattr *sattr)
 		.fh		= &fhandle,
 		.fattr		= &fattr
 	};
+	struct rpc_message msg = {
+		.rpc_proc	= &nfs3_procedures[NFS3PROC_MKDIR],
+		.rpc_argp	= &arg,
+		.rpc_resp	= &res,
+	};
 	int mode = sattr->ia_mode;
 	int status;
 
@@ -569,7 +611,7 @@ nfs3_proc_mkdir(struct inode *dir, struct dentry *dentry, struct iattr *sattr)
 
 	nfs_fattr_init(&dir_attr);
 	nfs_fattr_init(&fattr);
-	status = rpc_call(NFS_CLIENT(dir), NFS3PROC_MKDIR, &arg, &res, 0);
+	status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0);
 	nfs_post_op_update_inode(dir, &dir_attr);
 	if (status != 0)
 		goto out;
@@ -591,11 +633,16 @@ nfs3_proc_rmdir(struct inode *dir, struct qstr *name)
 		.name		= name->name,
 		.len		= name->len
 	};
+	struct rpc_message msg = {
+		.rpc_proc	= &nfs3_procedures[NFS3PROC_RMDIR],
+		.rpc_argp	= &arg,
+		.rpc_resp	= &dir_attr,
+	};
 	int			status;
 
 	dprintk("NFS call  rmdir %s\n", name->name);
 	nfs_fattr_init(&dir_attr);
-	status = rpc_call(NFS_CLIENT(dir), NFS3PROC_RMDIR, &arg, &dir_attr, 0);
+	status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0);
 	nfs_post_op_update_inode(dir, &dir_attr);
 	dprintk("NFS reply rmdir: %d\n", status);
 	return status;
@@ -672,6 +719,11 @@ nfs3_proc_mknod(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
 		.fh		= &fh,
 		.fattr		= &fattr
 	};
+	struct rpc_message msg = {
+		.rpc_proc	= &nfs3_procedures[NFS3PROC_MKNOD],
+		.rpc_argp	= &arg,
+		.rpc_resp	= &res,
+	};
 	mode_t mode = sattr->ia_mode;
 	int status;
 
@@ -690,7 +742,7 @@ nfs3_proc_mknod(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
 
 	nfs_fattr_init(&dir_attr);
 	nfs_fattr_init(&fattr);
-	status = rpc_call(NFS_CLIENT(dir), NFS3PROC_MKNOD, &arg, &res, 0);
+	status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0);
 	nfs_post_op_update_inode(dir, &dir_attr);
 	if (status != 0)
 		goto out;
@@ -707,11 +759,16 @@ static int
 nfs3_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
 		 struct nfs_fsstat *stat)
 {
+	struct rpc_message msg = {
+		.rpc_proc	= &nfs3_procedures[NFS3PROC_FSSTAT],
+		.rpc_argp	= fhandle,
+		.rpc_resp	= stat,
+	};
 	int	status;
 
 	dprintk("NFS call  fsstat\n");
 	nfs_fattr_init(stat->fattr);
-	status = rpc_call(server->client, NFS3PROC_FSSTAT, fhandle, stat, 0);
+	status = rpc_call_sync(server->client, &msg, 0);
 	dprintk("NFS reply statfs: %d\n", status);
 	return status;
 }
@@ -720,11 +777,16 @@ static int
 nfs3_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
 		 struct nfs_fsinfo *info)
 {
+	struct rpc_message msg = {
+		.rpc_proc	= &nfs3_procedures[NFS3PROC_FSINFO],
+		.rpc_argp	= fhandle,
+		.rpc_resp	= info,
+	};
 	int	status;
 
 	dprintk("NFS call  fsinfo\n");
 	nfs_fattr_init(info->fattr);
-	status = rpc_call(server->client_sys, NFS3PROC_FSINFO, fhandle, info, 0);
+	status = rpc_call_sync(server->client_sys, &msg, 0);
 	dprintk("NFS reply fsinfo: %d\n", status);
 	return status;
 }
@@ -733,40 +795,34 @@ static int
 nfs3_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
 		   struct nfs_pathconf *info)
 {
+	struct rpc_message msg = {
+		.rpc_proc	= &nfs3_procedures[NFS3PROC_PATHCONF],
+		.rpc_argp	= fhandle,
+		.rpc_resp	= info,
+	};
 	int	status;
 
 	dprintk("NFS call  pathconf\n");
 	nfs_fattr_init(info->fattr);
-	status = rpc_call(server->client, NFS3PROC_PATHCONF, fhandle, info, 0);
+	status = rpc_call_sync(server->client, &msg, 0);
 	dprintk("NFS reply pathconf: %d\n", status);
 	return status;
 }
 
 extern u32 *nfs3_decode_dirent(u32 *, struct nfs_entry *, int);
 
-static void nfs3_read_done(struct rpc_task *task, void *calldata)
+static int nfs3_read_done(struct rpc_task *task, struct nfs_read_data *data)
 {
-	struct nfs_read_data *data = calldata;
-
-	if (nfs3_async_handle_jukebox(task))
-		return;
+	if (nfs3_async_handle_jukebox(task, data->inode))
+		return -EAGAIN;
 	/* Call back common NFS readpage processing */
 	if (task->tk_status >= 0)
 		nfs_refresh_inode(data->inode, &data->fattr);
-	nfs_readpage_result(task, calldata);
+	return 0;
 }
 
-static const struct rpc_call_ops nfs3_read_ops = {
-	.rpc_call_done = nfs3_read_done,
-	.rpc_release = nfs_readdata_release,
-};
-
-static void
-nfs3_proc_read_setup(struct nfs_read_data *data)
+static void nfs3_proc_read_setup(struct nfs_read_data *data)
 {
-	struct rpc_task		*task = &data->task;
-	struct inode		*inode = data->inode;
-	int			flags;
 	struct rpc_message	msg = {
 		.rpc_proc	= &nfs3_procedures[NFS3PROC_READ],
 		.rpc_argp	= &data->args,
@@ -774,37 +830,20 @@ nfs3_proc_read_setup(struct nfs_read_data *data)
 		.rpc_cred	= data->cred,
 	};
 
-	/* N.B. Do we need to test? Never called for swapfile inode */
-	flags = RPC_TASK_ASYNC | (IS_SWAPFILE(inode)? NFS_RPC_SWAPFLAGS : 0);
-
-	/* Finalize the task. */
-	rpc_init_task(task, NFS_CLIENT(inode), flags, &nfs3_read_ops, data);
-	rpc_call_setup(task, &msg, 0);
+	rpc_call_setup(&data->task, &msg, 0);
 }
 
-static void nfs3_write_done(struct rpc_task *task, void *calldata)
+static int nfs3_write_done(struct rpc_task *task, struct nfs_write_data *data)
 {
-	struct nfs_write_data *data = calldata;
-
-	if (nfs3_async_handle_jukebox(task))
-		return;
+	if (nfs3_async_handle_jukebox(task, data->inode))
+		return -EAGAIN;
 	if (task->tk_status >= 0)
 		nfs_post_op_update_inode(data->inode, data->res.fattr);
-	nfs_writeback_done(task, calldata);
+	return 0;
 }
 
-static const struct rpc_call_ops nfs3_write_ops = {
-	.rpc_call_done = nfs3_write_done,
-	.rpc_release = nfs_writedata_release,
-};
-
-static void
-nfs3_proc_write_setup(struct nfs_write_data *data, int how)
+static void nfs3_proc_write_setup(struct nfs_write_data *data, int how)
 {
-	struct rpc_task		*task = &data->task;
-	struct inode		*inode = data->inode;
-	int			stable;
-	int			flags;
 	struct rpc_message	msg = {
 		.rpc_proc	= &nfs3_procedures[NFS3PROC_WRITE],
 		.rpc_argp	= &data->args,
@@ -812,45 +851,28 @@ nfs3_proc_write_setup(struct nfs_write_data *data, int how)
 		.rpc_cred	= data->cred,
 	};
 
+	data->args.stable = NFS_UNSTABLE;
 	if (how & FLUSH_STABLE) {
-		if (!NFS_I(inode)->ncommit)
-			stable = NFS_FILE_SYNC;
-		else
-			stable = NFS_DATA_SYNC;
-	} else
-		stable = NFS_UNSTABLE;
-	data->args.stable = stable;
-
-	/* Set the initial flags for the task.  */
-	flags = (how & FLUSH_SYNC) ? 0 : RPC_TASK_ASYNC;
+		data->args.stable = NFS_FILE_SYNC;
+		if (NFS_I(data->inode)->ncommit)
+			data->args.stable = NFS_DATA_SYNC;
+	}
 
 	/* Finalize the task. */
-	rpc_init_task(task, NFS_CLIENT(inode), flags, &nfs3_write_ops, data);
-	rpc_call_setup(task, &msg, 0);
+	rpc_call_setup(&data->task, &msg, 0);
 }
 
-static void nfs3_commit_done(struct rpc_task *task, void *calldata)
+static int nfs3_commit_done(struct rpc_task *task, struct nfs_write_data *data)
 {
-	struct nfs_write_data *data = calldata;
-
-	if (nfs3_async_handle_jukebox(task))
-		return;
+	if (nfs3_async_handle_jukebox(task, data->inode))
+		return -EAGAIN;
 	if (task->tk_status >= 0)
 		nfs_post_op_update_inode(data->inode, data->res.fattr);
-	nfs_commit_done(task, calldata);
+	return 0;
 }
 
-static const struct rpc_call_ops nfs3_commit_ops = {
-	.rpc_call_done = nfs3_commit_done,
-	.rpc_release = nfs_commit_release,
-};
-
-static void
-nfs3_proc_commit_setup(struct nfs_write_data *data, int how)
+static void nfs3_proc_commit_setup(struct nfs_write_data *data, int how)
 {
-	struct rpc_task		*task = &data->task;
-	struct inode		*inode = data->inode;
-	int			flags;
 	struct rpc_message	msg = {
 		.rpc_proc	= &nfs3_procedures[NFS3PROC_COMMIT],
 		.rpc_argp	= &data->args,
@@ -858,12 +880,7 @@ nfs3_proc_commit_setup(struct nfs_write_data *data, int how)
 		.rpc_cred	= data->cred,
 	};
 
-	/* Set the initial flags for the task.  */
-	flags = (how & FLUSH_SYNC) ? 0 : RPC_TASK_ASYNC;
-
-	/* Finalize the task. */
-	rpc_init_task(task, NFS_CLIENT(inode), flags, &nfs3_commit_ops, data);
-	rpc_call_setup(task, &msg, 0);
+	rpc_call_setup(&data->task, &msg, 0);
 }
 
 static int
@@ -902,8 +919,11 @@ struct nfs_rpc_ops	nfs_v3_clientops = {
 	.pathconf	= nfs3_proc_pathconf,
 	.decode_dirent	= nfs3_decode_dirent,
 	.read_setup	= nfs3_proc_read_setup,
+	.read_done	= nfs3_read_done,
 	.write_setup	= nfs3_proc_write_setup,
+	.write_done	= nfs3_write_done,
 	.commit_setup	= nfs3_proc_commit_setup,
+	.commit_done	= nfs3_commit_done,
 	.file_open	= nfs_open,
 	.file_release	= nfs_release,
 	.lock		= nfs3_proc_lock,

fs/nfs/nfs3xdr.c

@@ -1109,7 +1109,9 @@ nfs3_xdr_setaclres(struct rpc_rqst *req, u32 *p, struct nfs_fattr *fattr)
 	.p_encode    = (kxdrproc_t) nfs3_xdr_##argtype,			\
 	.p_decode    = (kxdrproc_t) nfs3_xdr_##restype,			\
 	.p_bufsiz    = MAX(NFS3_##argtype##_sz,NFS3_##restype##_sz) << 2,	\
-	.p_timer     = timer						\
+	.p_timer     = timer,						\
+	.p_statidx   = NFS3PROC_##proc,					\
+	.p_name      = #proc,						\
 	}
 
 struct rpc_procinfo	nfs3_procedures[] = {
@@ -1150,6 +1152,7 @@ static struct rpc_procinfo	nfs3_acl_procedures[] = {
 		.p_decode = (kxdrproc_t) nfs3_xdr_getaclres,
 		.p_bufsiz = MAX(ACL3_getaclargs_sz, ACL3_getaclres_sz) << 2,
 		.p_timer = 1,
+		.p_name = "GETACL",
 	},
 	[ACLPROC3_SETACL] = {
 		.p_proc = ACLPROC3_SETACL,
@@ -1157,6 +1160,7 @@ static struct rpc_procinfo	nfs3_acl_procedures[] = {
 		.p_decode = (kxdrproc_t) nfs3_xdr_setaclres,
 		.p_bufsiz = MAX(ACL3_setaclargs_sz, ACL3_setaclres_sz) << 2,
 		.p_timer = 0,
+		.p_name = "SETACL",
 	},
 };
 

fs/nfs/nfs4proc.c

@@ -51,6 +51,7 @@
 
 #include "nfs4_fs.h"
 #include "delegation.h"
+#include "iostat.h"
 
 #define NFSDBG_FACILITY		NFSDBG_PROC
 
@@ -335,7 +336,7 @@ static struct nfs4_state *nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data
 	if (!(data->f_attr.valid & NFS_ATTR_FATTR))
 		goto out;
 	inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
-	if (inode == NULL)
+	if (IS_ERR(inode))
 		goto out;
 	state = nfs4_get_open_state(inode, data->owner);
 	if (state == NULL)
@@ -604,11 +605,14 @@ static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
 	int status;
 
 	atomic_inc(&data->count);
+	/*
+	 * If rpc_run_task() ends up calling ->rpc_release(), we
+	 * want to ensure that it takes the 'error' code path.
+	 */
+	data->rpc_status = -ENOMEM;
 	task = rpc_run_task(server->client, RPC_TASK_ASYNC, &nfs4_open_confirm_ops, data);
-	if (IS_ERR(task)) {
-		nfs4_opendata_free(data);
+	if (IS_ERR(task))
 		return PTR_ERR(task);
-	}
 	status = nfs4_wait_for_completion_rpc_task(task);
 	if (status != 0) {
 		data->cancelled = 1;
@@ -707,11 +711,14 @@ static int _nfs4_proc_open(struct nfs4_opendata *data)
 	int status;
 
 	atomic_inc(&data->count);
+	/*
+	 * If rpc_run_task() ends up calling ->rpc_release(), we
+	 * want to ensure that it takes the 'error' code path.
+	 */
+	data->rpc_status = -ENOMEM;
 	task = rpc_run_task(server->client, RPC_TASK_ASYNC, &nfs4_open_ops, data);
-	if (IS_ERR(task)) {
-		nfs4_opendata_free(data);
+	if (IS_ERR(task))
 		return PTR_ERR(task);
-	}
 	status = nfs4_wait_for_completion_rpc_task(task);
 	if (status != 0) {
 		data->cancelled = 1;
@@ -908,7 +915,7 @@ out_put_state_owner:
 static struct nfs4_state *nfs4_open_delegated(struct inode *inode, int flags, struct rpc_cred *cred)
 {
 	struct nfs4_exception exception = { };
-	struct nfs4_state *res;
+	struct nfs4_state *res = ERR_PTR(-EIO);
 	int err;
 
 	do {
@@ -1017,12 +1024,12 @@ static struct nfs4_state *nfs4_do_open(struct inode *dir, struct dentry *dentry,
 	return res;
 }
 
-static int _nfs4_do_setattr(struct nfs_server *server, struct nfs_fattr *fattr,
-                struct nfs_fh *fhandle, struct iattr *sattr,
-                struct nfs4_state *state)
+static int _nfs4_do_setattr(struct inode *inode, struct nfs_fattr *fattr,
+                struct iattr *sattr, struct nfs4_state *state)
 {
+	struct nfs_server *server = NFS_SERVER(inode);
         struct nfs_setattrargs  arg = {
-                .fh             = fhandle,
+                .fh             = NFS_FH(inode),
                 .iap            = sattr,
 		.server		= server,
 		.bitmask = server->attr_bitmask,
@@ -1041,7 +1048,9 @@ static int _nfs4_do_setattr(struct nfs_server *server, struct nfs_fattr *fattr,
 
 	nfs_fattr_init(fattr);
 
-	if (state != NULL) {
+	if (nfs4_copy_delegation_stateid(&arg.stateid, inode)) {
+		/* Use that stateid */
+	} else if (state != NULL) {
 		msg.rpc_cred = state->owner->so_cred;
 		nfs4_copy_stateid(&arg.stateid, state, current->files);
 	} else
@@ -1053,16 +1062,15 @@ static int _nfs4_do_setattr(struct nfs_server *server, struct nfs_fattr *fattr,
 	return status;
 }
 
-static int nfs4_do_setattr(struct nfs_server *server, struct nfs_fattr *fattr,
-                struct nfs_fh *fhandle, struct iattr *sattr,
-                struct nfs4_state *state)
+static int nfs4_do_setattr(struct inode *inode, struct nfs_fattr *fattr,
+                struct iattr *sattr, struct nfs4_state *state)
 {
+	struct nfs_server *server = NFS_SERVER(inode);
 	struct nfs4_exception exception = { };
 	int err;
 	do {
 		err = nfs4_handle_exception(server,
-				_nfs4_do_setattr(server, fattr, fhandle, sattr,
-					state),
+				_nfs4_do_setattr(inode, fattr, sattr, state),
 				&exception);
 	} while (exception.retry);
 	return err;
@@ -1503,8 +1511,7 @@ nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
 	if (ctx != NULL)
 		state = ctx->state;
 
-	status = nfs4_do_setattr(NFS_SERVER(inode), fattr,
-			NFS_FH(inode), sattr, state);
+	status = nfs4_do_setattr(inode, fattr, sattr, state);
 	if (status == 0)
 		nfs_setattr_update_inode(inode, sattr);
 	if (ctx != NULL)
@@ -1823,8 +1830,7 @@ nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
 	d_instantiate(dentry, igrab(state->inode));
 	if (flags & O_EXCL) {
 		struct nfs_fattr fattr;
-		status = nfs4_do_setattr(NFS_SERVER(dir), &fattr,
-		                     NFS_FH(state->inode), sattr, state);
+		status = nfs4_do_setattr(state->inode, &fattr, sattr, state);
 		if (status == 0)
 			nfs_setattr_update_inode(state->inode, sattr);
 	}
@@ -2344,75 +2350,50 @@ static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
 	return err;
 }
 
-static void nfs4_read_done(struct rpc_task *task, void *calldata)
+static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
 {
-	struct nfs_read_data *data = calldata;
-	struct inode *inode = data->inode;
+	struct nfs_server *server = NFS_SERVER(data->inode);
 
-	if (nfs4_async_handle_error(task, NFS_SERVER(inode)) == -EAGAIN) {
+	if (nfs4_async_handle_error(task, server) == -EAGAIN) {
 		rpc_restart_call(task);
-		return;
+		return -EAGAIN;
 	}
 	if (task->tk_status > 0)
-		renew_lease(NFS_SERVER(inode), data->timestamp);
-	/* Call back common NFS readpage processing */
-	nfs_readpage_result(task, calldata);
+		renew_lease(server, data->timestamp);
+	return 0;
 }
 
-static const struct rpc_call_ops nfs4_read_ops = {
-	.rpc_call_done = nfs4_read_done,
-	.rpc_release = nfs_readdata_release,
-};
-
-static void
-nfs4_proc_read_setup(struct nfs_read_data *data)
+static void nfs4_proc_read_setup(struct nfs_read_data *data)
 {
-	struct rpc_task	*task = &data->task;
 	struct rpc_message msg = {
 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ],
 		.rpc_argp = &data->args,
 		.rpc_resp = &data->res,
 		.rpc_cred = data->cred,
 	};
-	struct inode *inode = data->inode;
-	int flags;
 
 	data->timestamp   = jiffies;
 
-	/* N.B. Do we need to test? Never called for swapfile inode */
-	flags = RPC_TASK_ASYNC | (IS_SWAPFILE(inode)? NFS_RPC_SWAPFLAGS : 0);
-
-	/* Finalize the task. */
-	rpc_init_task(task, NFS_CLIENT(inode), flags, &nfs4_read_ops, data);
-	rpc_call_setup(task, &msg, 0);
+	rpc_call_setup(&data->task, &msg, 0);
 }
 
-static void nfs4_write_done(struct rpc_task *task, void *calldata)
+static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
 {
-	struct nfs_write_data *data = calldata;
 	struct inode *inode = data->inode;
 	
 	if (nfs4_async_handle_error(task, NFS_SERVER(inode)) == -EAGAIN) {
 		rpc_restart_call(task);
-		return;
+		return -EAGAIN;
 	}
 	if (task->tk_status >= 0) {
 		renew_lease(NFS_SERVER(inode), data->timestamp);
 		nfs_post_op_update_inode(inode, data->res.fattr);
 	}
-	/* Call back common NFS writeback processing */
-	nfs_writeback_done(task, calldata);
+	return 0;
 }
 
-static const struct rpc_call_ops nfs4_write_ops = {
-	.rpc_call_done = nfs4_write_done,
-	.rpc_release = nfs_writedata_release,
-};
-
-static void
-nfs4_proc_write_setup(struct nfs_write_data *data, int how)
+static void nfs4_proc_write_setup(struct nfs_write_data *data, int how)
 {
-	struct rpc_task	*task = &data->task;
 	struct rpc_message msg = {
 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE],
 		.rpc_argp = &data->args,
@@ -2422,7 +2403,6 @@ nfs4_proc_write_setup(struct nfs_write_data *data, int how)
 	struct inode *inode = data->inode;
 	struct nfs_server *server = NFS_SERVER(inode);
 	int stable;
-	int flags;
 	
 	if (how & FLUSH_STABLE) {
 		if (!NFS_I(inode)->ncommit)
@@ -2437,57 +2417,37 @@ nfs4_proc_write_setup(struct nfs_write_data *data, int how)
 
 	data->timestamp   = jiffies;
 
-	/* Set the initial flags for the task.  */
-	flags = (how & FLUSH_SYNC) ? 0 : RPC_TASK_ASYNC;
-
 	/* Finalize the task. */
-	rpc_init_task(task, NFS_CLIENT(inode), flags, &nfs4_write_ops, data);
-	rpc_call_setup(task, &msg, 0);
+	rpc_call_setup(&data->task, &msg, 0);
 }
 
-static void nfs4_commit_done(struct rpc_task *task, void *calldata)
+static int nfs4_commit_done(struct rpc_task *task, struct nfs_write_data *data)
 {
-	struct nfs_write_data *data = calldata;
 	struct inode *inode = data->inode;
 	
 	if (nfs4_async_handle_error(task, NFS_SERVER(inode)) == -EAGAIN) {
 		rpc_restart_call(task);
-		return;
+		return -EAGAIN;
 	}
 	if (task->tk_status >= 0)
 		nfs_post_op_update_inode(inode, data->res.fattr);
-	/* Call back common NFS writeback processing */
-	nfs_commit_done(task, calldata);
+	return 0;
 }
 
-static const struct rpc_call_ops nfs4_commit_ops = {
-	.rpc_call_done = nfs4_commit_done,
-	.rpc_release = nfs_commit_release,
-};
-
-static void
-nfs4_proc_commit_setup(struct nfs_write_data *data, int how)
+static void nfs4_proc_commit_setup(struct nfs_write_data *data, int how)
 {
-	struct rpc_task	*task = &data->task;
 	struct rpc_message msg = {
 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
 		.rpc_argp = &data->args,
 		.rpc_resp = &data->res,
 		.rpc_cred = data->cred,
 	};	
-	struct inode *inode = data->inode;
-	struct nfs_server *server = NFS_SERVER(inode);
-	int flags;
+	struct nfs_server *server = NFS_SERVER(data->inode);
 	
 	data->args.bitmask = server->attr_bitmask;
 	data->res.server = server;
 
-	/* Set the initial flags for the task.  */
-	flags = (how & FLUSH_SYNC) ? 0 : RPC_TASK_ASYNC;
-
-	/* Finalize the task. */
-	rpc_init_task(task, NFS_CLIENT(inode), flags, &nfs4_commit_ops, data);
-	rpc_call_setup(task, &msg, 0);	
+	rpc_call_setup(&data->task, &msg, 0);
 }
 
 /*
@@ -2755,8 +2715,10 @@ nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server)
 				rpc_wake_up_task(task);
 			task->tk_status = 0;
 			return -EAGAIN;
-		case -NFS4ERR_GRACE:
 		case -NFS4ERR_DELAY:
+			nfs_inc_server_stats((struct nfs_server *) server,
+						NFSIOS_DELAY);
+		case -NFS4ERR_GRACE:
 			rpc_delay(task, NFS4_POLL_RETRY_MAX);
 			task->tk_status = 0;
 			return -EAGAIN;
@@ -2893,8 +2855,7 @@ int nfs4_proc_setclientid(struct nfs4_client *clp, u32 program, unsigned short p
 	return status;
 }
 
-int
-nfs4_proc_setclientid_confirm(struct nfs4_client *clp, struct rpc_cred *cred)
+static int _nfs4_proc_setclientid_confirm(struct nfs4_client *clp, struct rpc_cred *cred)
 {
 	struct nfs_fsinfo fsinfo;
 	struct rpc_message msg = {
@@ -2918,6 +2879,24 @@ nfs4_proc_setclientid_confirm(struct nfs4_client *clp, struct rpc_cred *cred)
 	return status;
 }
 
+int nfs4_proc_setclientid_confirm(struct nfs4_client *clp, struct rpc_cred *cred)
+{
+	long timeout;
+	int err;
+	do {
+		err = _nfs4_proc_setclientid_confirm(clp, cred);
+		switch (err) {
+			case 0:
+				return err;
+			case -NFS4ERR_RESOURCE:
+				/* The IBM lawyers misread another document! */
+			case -NFS4ERR_DELAY:
+				err = nfs4_delay(clp->cl_rpcclient, &timeout);
+		}
+	} while (err == 0);
+	return err;
+}
+
 struct nfs4_delegreturndata {
 	struct nfs4_delegreturnargs args;
 	struct nfs4_delegreturnres res;
@@ -2958,7 +2937,7 @@ static void nfs4_delegreturn_release(void *calldata)
 	kfree(calldata);
 }
 
-const static struct rpc_call_ops nfs4_delegreturn_ops = {
+static const struct rpc_call_ops nfs4_delegreturn_ops = {
 	.rpc_call_prepare = nfs4_delegreturn_prepare,
 	.rpc_call_done = nfs4_delegreturn_done,
 	.rpc_release = nfs4_delegreturn_release,
@@ -2986,10 +2965,8 @@ static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, co
 	data->rpc_status = 0;
 
 	task = rpc_run_task(NFS_CLIENT(inode), RPC_TASK_ASYNC, &nfs4_delegreturn_ops, data);
-	if (IS_ERR(task)) {
-		nfs4_delegreturn_release(data);
+	if (IS_ERR(task))
 		return PTR_ERR(task);
-	}
 	status = nfs4_wait_for_completion_rpc_task(task);
 	if (status == 0) {
 		status = data->rpc_status;
@@ -3209,7 +3186,6 @@ static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
 		struct nfs_seqid *seqid)
 {
 	struct nfs4_unlockdata *data;
-	struct rpc_task *task;
 
 	data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
 	if (data == NULL) {
@@ -3219,10 +3195,7 @@ static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
 
 	/* Unlock _before_ we do the RPC call */
 	do_vfs_lock(fl->fl_file, fl);
-	task = rpc_run_task(NFS_CLIENT(lsp->ls_state->inode), RPC_TASK_ASYNC, &nfs4_locku_ops, data);
-	if (IS_ERR(task))
-		nfs4_locku_release_calldata(data);
-	return task;
+	return rpc_run_task(NFS_CLIENT(lsp->ls_state->inode), RPC_TASK_ASYNC, &nfs4_locku_ops, data);
 }
 
 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
@@ -3403,10 +3376,8 @@ static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *f
 		data->arg.reclaim = 1;
 	task = rpc_run_task(NFS_CLIENT(state->inode), RPC_TASK_ASYNC,
 			&nfs4_lock_ops, data);
-	if (IS_ERR(task)) {
-		nfs4_lock_release(data);
+	if (IS_ERR(task))
 		return PTR_ERR(task);
-	}
 	ret = nfs4_wait_for_completion_rpc_task(task);
 	if (ret == 0) {
 		ret = data->rpc_status;
@@ -3588,6 +3559,8 @@ ssize_t nfs4_listxattr(struct dentry *dentry, char *buf, size_t buflen)
 {
 	size_t len = strlen(XATTR_NAME_NFSV4_ACL) + 1;
 
+	if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
+		return 0;
 	if (buf && buflen < len)
 		return -ERANGE;
 	if (buf)
@@ -3644,8 +3617,11 @@ struct nfs_rpc_ops	nfs_v4_clientops = {
 	.pathconf	= nfs4_proc_pathconf,
 	.decode_dirent	= nfs4_decode_dirent,
 	.read_setup	= nfs4_proc_read_setup,
+	.read_done	= nfs4_read_done,
 	.write_setup	= nfs4_proc_write_setup,
+	.write_done	= nfs4_write_done,
 	.commit_setup	= nfs4_proc_commit_setup,
+	.commit_done	= nfs4_commit_done,
 	.file_open      = nfs_open,
 	.file_release   = nfs_release,
 	.lock		= nfs4_proc_lock,

fs/nfs/nfs4state.c

@@ -977,6 +977,7 @@ out:
 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);
+	set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
 	goto out;
 }
 

fs/nfs/nfs4xdr.c

@@ -4344,6 +4344,8 @@ nfs_stat_to_errno(int stat)
 	.p_encode = (kxdrproc_t) nfs4_xdr_##argtype,		\
 	.p_decode = (kxdrproc_t) nfs4_xdr_##restype,		\
 	.p_bufsiz = MAX(NFS4_##argtype##_sz,NFS4_##restype##_sz) << 2,	\
+	.p_statidx = NFSPROC4_CLNT_##proc,			\
+	.p_name   = #proc,					\
     }
 
 struct rpc_procinfo	nfs4_procedures[] = {

fs/nfs/pagelist.c

@@ -85,6 +85,9 @@ nfs_create_request(struct nfs_open_context *ctx, struct inode *inode,
 	atomic_set(&req->wb_complete, 0);
 	req->wb_index	= page->index;
 	page_cache_get(page);
+	BUG_ON(PagePrivate(page));
+	BUG_ON(!PageLocked(page));
+	BUG_ON(page->mapping->host != inode);
 	req->wb_offset  = offset;
 	req->wb_pgbase	= offset;
 	req->wb_bytes   = count;
@@ -132,9 +135,11 @@ void nfs_clear_page_writeback(struct nfs_page *req)
 {
 	struct nfs_inode *nfsi = NFS_I(req->wb_context->dentry->d_inode);
 
-	spin_lock(&nfsi->req_lock);
-	radix_tree_tag_clear(&nfsi->nfs_page_tree, req->wb_index, NFS_PAGE_TAG_WRITEBACK);
-	spin_unlock(&nfsi->req_lock);
+	if (req->wb_page != NULL) {
+		spin_lock(&nfsi->req_lock);
+		radix_tree_tag_clear(&nfsi->nfs_page_tree, req->wb_index, NFS_PAGE_TAG_WRITEBACK);
+		spin_unlock(&nfsi->req_lock);
+	}
 	nfs_unlock_request(req);
 }
 
@@ -147,8 +152,9 @@ void nfs_clear_page_writeback(struct nfs_page *req)
  */
 void nfs_clear_request(struct nfs_page *req)
 {
-	if (req->wb_page) {
-		page_cache_release(req->wb_page);
+	struct page *page = req->wb_page;
+	if (page != NULL) {
+		page_cache_release(page);
 		req->wb_page = NULL;
 	}
 }

fs/nfs/proc.c

@@ -58,16 +58,23 @@ nfs_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle,
 {
 	struct nfs_fattr *fattr = info->fattr;
 	struct nfs2_fsstat fsinfo;
+	struct rpc_message msg = {
+		.rpc_proc	= &nfs_procedures[NFSPROC_GETATTR],
+		.rpc_argp	= fhandle,
+		.rpc_resp	= fattr,
+	};
 	int status;
 
 	dprintk("%s: call getattr\n", __FUNCTION__);
 	nfs_fattr_init(fattr);
-	status = rpc_call(server->client_sys, NFSPROC_GETATTR, fhandle, fattr, 0);
+	status = rpc_call_sync(server->client_sys, &msg, 0);
 	dprintk("%s: reply getattr: %d\n", __FUNCTION__, status);
 	if (status)
 		return status;
 	dprintk("%s: call statfs\n", __FUNCTION__);
-	status = rpc_call(server->client_sys, NFSPROC_STATFS, fhandle, &fsinfo, 0);
+	msg.rpc_proc = &nfs_procedures[NFSPROC_STATFS];
+	msg.rpc_resp = &fsinfo;
+	status = rpc_call_sync(server->client_sys, &msg, 0);
 	dprintk("%s: reply statfs: %d\n", __FUNCTION__, status);
 	if (status)
 		return status;
@@ -90,12 +97,16 @@ static int
 nfs_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
 		struct nfs_fattr *fattr)
 {
+	struct rpc_message msg = {
+		.rpc_proc	= &nfs_procedures[NFSPROC_GETATTR],
+		.rpc_argp	= fhandle,
+		.rpc_resp	= fattr,
+	};
 	int	status;
 
 	dprintk("NFS call  getattr\n");
 	nfs_fattr_init(fattr);
-	status = rpc_call(server->client, NFSPROC_GETATTR,
-				fhandle, fattr, 0);
+	status = rpc_call_sync(server->client, &msg, 0);
 	dprintk("NFS reply getattr: %d\n", status);
 	return status;
 }
@@ -109,6 +120,11 @@ nfs_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
 		.fh	= NFS_FH(inode),
 		.sattr	= sattr
 	};
+	struct rpc_message msg = {
+		.rpc_proc	= &nfs_procedures[NFSPROC_SETATTR],
+		.rpc_argp	= &arg,
+		.rpc_resp	= fattr,
+	};
 	int	status;
 
 	/* Mask out the non-modebit related stuff from attr->ia_mode */
@@ -116,7 +132,7 @@ nfs_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
 
 	dprintk("NFS call  setattr\n");
 	nfs_fattr_init(fattr);
-	status = rpc_call(NFS_CLIENT(inode), NFSPROC_SETATTR, &arg, fattr, 0);
+	status = rpc_call_sync(NFS_CLIENT(inode), &msg, 0);
 	if (status == 0)
 		nfs_setattr_update_inode(inode, sattr);
 	dprintk("NFS reply setattr: %d\n", status);
@@ -136,11 +152,16 @@ nfs_proc_lookup(struct inode *dir, struct qstr *name,
 		.fh		= fhandle,
 		.fattr		= fattr
 	};
+	struct rpc_message msg = {
+		.rpc_proc	= &nfs_procedures[NFSPROC_LOOKUP],
+		.rpc_argp	= &arg,
+		.rpc_resp	= &res,
+	};
 	int			status;
 
 	dprintk("NFS call  lookup %s\n", name->name);
 	nfs_fattr_init(fattr);
-	status = rpc_call(NFS_CLIENT(dir), NFSPROC_LOOKUP, &arg, &res, 0);
+	status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0);
 	dprintk("NFS reply lookup: %d\n", status);
 	return status;
 }
@@ -154,10 +175,14 @@ static int nfs_proc_readlink(struct inode *inode, struct page *page,
 		.pglen		= pglen,
 		.pages		= &page
 	};
+	struct rpc_message msg = {
+		.rpc_proc	= &nfs_procedures[NFSPROC_READLINK],
+		.rpc_argp	= &args,
+	};
 	int			status;
 
 	dprintk("NFS call  readlink\n");
-	status = rpc_call(NFS_CLIENT(inode), NFSPROC_READLINK, &args, NULL, 0);
+	status = rpc_call_sync(NFS_CLIENT(inode), &msg, 0);
 	dprintk("NFS reply readlink: %d\n", status);
 	return status;
 }
@@ -233,11 +258,16 @@ nfs_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
 		.fh		= &fhandle,
 		.fattr		= &fattr
 	};
+	struct rpc_message msg = {
+		.rpc_proc	= &nfs_procedures[NFSPROC_CREATE],
+		.rpc_argp	= &arg,
+		.rpc_resp	= &res,
+	};
 	int			status;
 
 	nfs_fattr_init(&fattr);
 	dprintk("NFS call  create %s\n", dentry->d_name.name);
-	status = rpc_call(NFS_CLIENT(dir), NFSPROC_CREATE, &arg, &res, 0);
+	status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0);
 	if (status == 0)
 		status = nfs_instantiate(dentry, &fhandle, &fattr);
 	dprintk("NFS reply create: %d\n", status);
@@ -263,6 +293,11 @@ nfs_proc_mknod(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
 		.fh		= &fhandle,
 		.fattr		= &fattr
 	};
+	struct rpc_message msg = {
+		.rpc_proc	= &nfs_procedures[NFSPROC_CREATE],
+		.rpc_argp	= &arg,
+		.rpc_resp	= &res,
+	};
 	int status, mode;
 
 	dprintk("NFS call  mknod %s\n", dentry->d_name.name);
@@ -277,13 +312,13 @@ nfs_proc_mknod(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
 	}
 
 	nfs_fattr_init(&fattr);
-	status = rpc_call(NFS_CLIENT(dir), NFSPROC_CREATE, &arg, &res, 0);
+	status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0);
 	nfs_mark_for_revalidate(dir);
 
 	if (status == -EINVAL && S_ISFIFO(mode)) {
 		sattr->ia_mode = mode;
 		nfs_fattr_init(&fattr);
-		status = rpc_call(NFS_CLIENT(dir), NFSPROC_CREATE, &arg, &res, 0);
+		status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0);
 	}
 	if (status == 0)
 		status = nfs_instantiate(dentry, &fhandle, &fattr);
@@ -302,8 +337,6 @@ nfs_proc_remove(struct inode *dir, struct qstr *name)
 	struct rpc_message	msg = { 
 		.rpc_proc	= &nfs_procedures[NFSPROC_REMOVE],
 		.rpc_argp	= &arg,
-		.rpc_resp	= NULL,
-		.rpc_cred	= NULL
 	};
 	int			status;
 
@@ -355,10 +388,14 @@ nfs_proc_rename(struct inode *old_dir, struct qstr *old_name,
 		.toname		= new_name->name,
 		.tolen		= new_name->len
 	};
+	struct rpc_message msg = {
+		.rpc_proc	= &nfs_procedures[NFSPROC_RENAME],
+		.rpc_argp	= &arg,
+	};
 	int			status;
 
 	dprintk("NFS call  rename %s -> %s\n", old_name->name, new_name->name);
-	status = rpc_call(NFS_CLIENT(old_dir), NFSPROC_RENAME, &arg, NULL, 0);
+	status = rpc_call_sync(NFS_CLIENT(old_dir), &msg, 0);
 	nfs_mark_for_revalidate(old_dir);
 	nfs_mark_for_revalidate(new_dir);
 	dprintk("NFS reply rename: %d\n", status);
@@ -374,10 +411,14 @@ nfs_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
 		.toname		= name->name,
 		.tolen		= name->len
 	};
+	struct rpc_message msg = {
+		.rpc_proc	= &nfs_procedures[NFSPROC_LINK],
+		.rpc_argp	= &arg,
+	};
 	int			status;
 
 	dprintk("NFS call  link %s\n", name->name);
-	status = rpc_call(NFS_CLIENT(inode), NFSPROC_LINK, &arg, NULL, 0);
+	status = rpc_call_sync(NFS_CLIENT(inode), &msg, 0);
 	nfs_mark_for_revalidate(inode);
 	nfs_mark_for_revalidate(dir);
 	dprintk("NFS reply link: %d\n", status);
@@ -397,6 +438,10 @@ nfs_proc_symlink(struct inode *dir, struct qstr *name, struct qstr *path,
 		.tolen		= path->len,
 		.sattr		= sattr
 	};
+	struct rpc_message msg = {
+		.rpc_proc	= &nfs_procedures[NFSPROC_SYMLINK],
+		.rpc_argp	= &arg,
+	};
 	int			status;
 
 	if (path->len > NFS2_MAXPATHLEN)
@@ -404,7 +449,7 @@ nfs_proc_symlink(struct inode *dir, struct qstr *name, struct qstr *path,
 	dprintk("NFS call  symlink %s -> %s\n", name->name, path->name);
 	nfs_fattr_init(fattr);
 	fhandle->size = 0;
-	status = rpc_call(NFS_CLIENT(dir), NFSPROC_SYMLINK, &arg, NULL, 0);
+	status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0);
 	nfs_mark_for_revalidate(dir);
 	dprintk("NFS reply symlink: %d\n", status);
 	return status;
@@ -425,11 +470,16 @@ nfs_proc_mkdir(struct inode *dir, struct dentry *dentry, struct iattr *sattr)
 		.fh		= &fhandle,
 		.fattr		= &fattr
 	};
+	struct rpc_message msg = {
+		.rpc_proc	= &nfs_procedures[NFSPROC_MKDIR],
+		.rpc_argp	= &arg,
+		.rpc_resp	= &res,
+	};
 	int			status;
 
 	dprintk("NFS call  mkdir %s\n", dentry->d_name.name);
 	nfs_fattr_init(&fattr);
-	status = rpc_call(NFS_CLIENT(dir), NFSPROC_MKDIR, &arg, &res, 0);
+	status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0);
 	nfs_mark_for_revalidate(dir);
 	if (status == 0)
 		status = nfs_instantiate(dentry, &fhandle, &fattr);
@@ -445,10 +495,14 @@ nfs_proc_rmdir(struct inode *dir, struct qstr *name)
 		.name		= name->name,
 		.len		= name->len
 	};
+	struct rpc_message msg = {
+		.rpc_proc	= &nfs_procedures[NFSPROC_RMDIR],
+		.rpc_argp	= &arg,
+	};
 	int			status;
 
 	dprintk("NFS call  rmdir %s\n", name->name);
-	status = rpc_call(NFS_CLIENT(dir), NFSPROC_RMDIR, &arg, NULL, 0);
+	status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0);
 	nfs_mark_for_revalidate(dir);
 	dprintk("NFS reply rmdir: %d\n", status);
 	return status;
@@ -470,13 +524,12 @@ nfs_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
 		.fh		= NFS_FH(dir),
 		.cookie		= cookie,
 		.count		= count,
-		.pages		= &page
+		.pages		= &page,
 	};
 	struct rpc_message	msg = {
 		.rpc_proc	= &nfs_procedures[NFSPROC_READDIR],
 		.rpc_argp	= &arg,
-		.rpc_resp	= NULL,
-		.rpc_cred	= cred
+		.rpc_cred	= cred,
 	};
 	int			status;
 
@@ -495,11 +548,16 @@ nfs_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
 			struct nfs_fsstat *stat)
 {
 	struct nfs2_fsstat fsinfo;
+	struct rpc_message msg = {
+		.rpc_proc	= &nfs_procedures[NFSPROC_STATFS],
+		.rpc_argp	= fhandle,
+		.rpc_resp	= &fsinfo,
+	};
 	int	status;
 
 	dprintk("NFS call  statfs\n");
 	nfs_fattr_init(stat->fattr);
-	status = rpc_call(server->client, NFSPROC_STATFS, fhandle, &fsinfo, 0);
+	status = rpc_call_sync(server->client, &msg, 0);
 	dprintk("NFS reply statfs: %d\n", status);
 	if (status)
 		goto out;
@@ -518,11 +576,16 @@ nfs_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
 			struct nfs_fsinfo *info)
 {
 	struct nfs2_fsstat fsinfo;
+	struct rpc_message msg = {
+		.rpc_proc	= &nfs_procedures[NFSPROC_STATFS],
+		.rpc_argp	= fhandle,
+		.rpc_resp	= &fsinfo,
+	};
 	int	status;
 
 	dprintk("NFS call  fsinfo\n");
 	nfs_fattr_init(info->fattr);
-	status = rpc_call(server->client, NFSPROC_STATFS, fhandle, &fsinfo, 0);
+	status = rpc_call_sync(server->client, &msg, 0);
 	dprintk("NFS reply fsinfo: %d\n", status);
 	if (status)
 		goto out;
@@ -550,10 +613,8 @@ nfs_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
 
 extern u32 * nfs_decode_dirent(u32 *, struct nfs_entry *, int);
 
-static void nfs_read_done(struct rpc_task *task, void *calldata)
+static int nfs_read_done(struct rpc_task *task, struct nfs_read_data *data)
 {
-	struct nfs_read_data *data = calldata;
-
 	if (task->tk_status >= 0) {
 		nfs_refresh_inode(data->inode, data->res.fattr);
 		/* Emulate the eof flag, which isn't normally needed in NFSv2
@@ -562,20 +623,11 @@ static void nfs_read_done(struct rpc_task *task, void *calldata)
 		if (data->args.offset + data->args.count >= data->res.fattr->size)
 			data->res.eof = 1;
 	}
-	nfs_readpage_result(task, calldata);
+	return 0;
 }
 
-static const struct rpc_call_ops nfs_read_ops = {
-	.rpc_call_done = nfs_read_done,
-	.rpc_release = nfs_readdata_release,
-};
-
-static void
-nfs_proc_read_setup(struct nfs_read_data *data)
+static void nfs_proc_read_setup(struct nfs_read_data *data)
 {
-	struct rpc_task		*task = &data->task;
-	struct inode		*inode = data->inode;
-	int			flags;
 	struct rpc_message	msg = {
 		.rpc_proc	= &nfs_procedures[NFSPROC_READ],
 		.rpc_argp	= &data->args,
@@ -583,34 +635,18 @@ nfs_proc_read_setup(struct nfs_read_data *data)
 		.rpc_cred	= data->cred,
 	};
 
-	/* N.B. Do we need to test? Never called for swapfile inode */
-	flags = RPC_TASK_ASYNC | (IS_SWAPFILE(inode)? NFS_RPC_SWAPFLAGS : 0);
-
-	/* Finalize the task. */
-	rpc_init_task(task, NFS_CLIENT(inode), flags, &nfs_read_ops, data);
-	rpc_call_setup(task, &msg, 0);
+	rpc_call_setup(&data->task, &msg, 0);
 }
 
-static void nfs_write_done(struct rpc_task *task, void *calldata)
+static int nfs_write_done(struct rpc_task *task, struct nfs_write_data *data)
 {
-	struct nfs_write_data *data = calldata;
-
 	if (task->tk_status >= 0)
 		nfs_post_op_update_inode(data->inode, data->res.fattr);
-	nfs_writeback_done(task, calldata);
+	return 0;
 }
 
-static const struct rpc_call_ops nfs_write_ops = {
-	.rpc_call_done = nfs_write_done,
-	.rpc_release = nfs_writedata_release,
-};
-
-static void
-nfs_proc_write_setup(struct nfs_write_data *data, int how)
+static void nfs_proc_write_setup(struct nfs_write_data *data, int how)
 {
-	struct rpc_task		*task = &data->task;
-	struct inode		*inode = data->inode;
-	int			flags;
 	struct rpc_message	msg = {
 		.rpc_proc	= &nfs_procedures[NFSPROC_WRITE],
 		.rpc_argp	= &data->args,
@@ -621,12 +657,8 @@ nfs_proc_write_setup(struct nfs_write_data *data, int how)
 	/* Note: NFSv2 ignores @stable and always uses NFS_FILE_SYNC */
 	data->args.stable = NFS_FILE_SYNC;
 
-	/* Set the initial flags for the task.  */
-	flags = (how & FLUSH_SYNC) ? 0 : RPC_TASK_ASYNC;
-
 	/* Finalize the task. */
-	rpc_init_task(task, NFS_CLIENT(inode), flags, &nfs_write_ops, data);
-	rpc_call_setup(task, &msg, 0);
+	rpc_call_setup(&data->task, &msg, 0);
 }
 
 static void
@@ -672,7 +704,9 @@ struct nfs_rpc_ops	nfs_v2_clientops = {
 	.pathconf	= nfs_proc_pathconf,
 	.decode_dirent	= nfs_decode_dirent,
 	.read_setup	= nfs_proc_read_setup,
+	.read_done	= nfs_read_done,
 	.write_setup	= nfs_proc_write_setup,
+	.write_done	= nfs_write_done,
 	.commit_setup	= nfs_proc_commit_setup,
 	.file_open	= nfs_open,
 	.file_release	= nfs_release,

fs/nfs/read.c

@@ -31,17 +31,49 @@
 
 #include <asm/system.h>
 
+#include "iostat.h"
+
 #define NFSDBG_FACILITY		NFSDBG_PAGECACHE
 
 static int nfs_pagein_one(struct list_head *, struct inode *);
-static void nfs_readpage_result_partial(struct nfs_read_data *, int);
-static void nfs_readpage_result_full(struct nfs_read_data *, int);
+static const struct rpc_call_ops nfs_read_partial_ops;
+static const struct rpc_call_ops nfs_read_full_ops;
 
 static kmem_cache_t *nfs_rdata_cachep;
-mempool_t *nfs_rdata_mempool;
+static mempool_t *nfs_rdata_mempool;
 
 #define MIN_POOL_READ	(32)
 
+struct nfs_read_data *nfs_readdata_alloc(unsigned int pagecount)
+{
+	struct nfs_read_data *p = mempool_alloc(nfs_rdata_mempool, SLAB_NOFS);
+
+	if (p) {
+		memset(p, 0, sizeof(*p));
+		INIT_LIST_HEAD(&p->pages);
+		if (pagecount < NFS_PAGEVEC_SIZE)
+			p->pagevec = &p->page_array[0];
+		else {
+			size_t size = ++pagecount * sizeof(struct page *);
+			p->pagevec = kmalloc(size, GFP_NOFS);
+			if (p->pagevec) {
+				memset(p->pagevec, 0, size);
+			} else {
+				mempool_free(p, nfs_rdata_mempool);
+				p = NULL;
+			}
+		}
+	}
+	return p;
+}
+
+void nfs_readdata_free(struct nfs_read_data *p)
+{
+	if (p && (p->pagevec != &p->page_array[0]))
+		kfree(p->pagevec);
+	mempool_free(p, nfs_rdata_mempool);
+}
+
 void nfs_readdata_release(void *data)
 {
         nfs_readdata_free(data);
@@ -133,6 +165,8 @@ static int nfs_readpage_sync(struct nfs_open_context *ctx, struct inode *inode,
 		}
 		count -= result;
 		rdata->args.pgbase += result;
+		nfs_add_stats(inode, NFSIOS_SERVERREADBYTES, result);
+
 		/* Note: result == 0 should only happen if we're caching
 		 * a write that extends the file and punches a hole.
 		 */
@@ -196,9 +230,11 @@ static void nfs_readpage_release(struct nfs_page *req)
  * Set up the NFS read request struct
  */
 static void nfs_read_rpcsetup(struct nfs_page *req, struct nfs_read_data *data,
+		const struct rpc_call_ops *call_ops,
 		unsigned int count, unsigned int offset)
 {
 	struct inode		*inode;
+	int flags;
 
 	data->req	  = req;
 	data->inode	  = inode = req->wb_context->dentry->d_inode;
@@ -216,6 +252,9 @@ static void nfs_read_rpcsetup(struct nfs_page *req, struct nfs_read_data *data,
 	data->res.eof     = 0;
 	nfs_fattr_init(&data->fattr);
 
+	/* Set up the initial task struct. */
+	flags = RPC_TASK_ASYNC | (IS_SWAPFILE(inode)? NFS_RPC_SWAPFLAGS : 0);
+	rpc_init_task(&data->task, NFS_CLIENT(inode), flags, call_ops, data);
 	NFS_PROTO(inode)->read_setup(data);
 
 	data->task.tk_cookie = (unsigned long)inode;
@@ -303,14 +342,15 @@ static int nfs_pagein_multi(struct list_head *head, struct inode *inode)
 		list_del_init(&data->pages);
 
 		data->pagevec[0] = page;
-		data->complete = nfs_readpage_result_partial;
 
 		if (nbytes > rsize) {
-			nfs_read_rpcsetup(req, data, rsize, offset);
+			nfs_read_rpcsetup(req, data, &nfs_read_partial_ops,
+					rsize, offset);
 			offset += rsize;
 			nbytes -= rsize;
 		} else {
-			nfs_read_rpcsetup(req, data, nbytes, offset);
+			nfs_read_rpcsetup(req, data, &nfs_read_partial_ops,
+					nbytes, offset);
 			nbytes = 0;
 		}
 		nfs_execute_read(data);
@@ -356,8 +396,7 @@ static int nfs_pagein_one(struct list_head *head, struct inode *inode)
 	}
 	req = nfs_list_entry(data->pages.next);
 
-	data->complete = nfs_readpage_result_full;
-	nfs_read_rpcsetup(req, data, count, 0);
+	nfs_read_rpcsetup(req, data, &nfs_read_full_ops, count, 0);
 
 	nfs_execute_read(data);
 	return 0;
@@ -391,12 +430,15 @@ nfs_pagein_list(struct list_head *head, int rpages)
 /*
  * Handle a read reply that fills part of a page.
  */
-static void nfs_readpage_result_partial(struct nfs_read_data *data, int status)
+static void nfs_readpage_result_partial(struct rpc_task *task, void *calldata)
 {
+	struct nfs_read_data *data = calldata;
 	struct nfs_page *req = data->req;
 	struct page *page = req->wb_page;
  
-	if (status >= 0) {
+	if (nfs_readpage_result(task, data) != 0)
+		return;
+	if (task->tk_status >= 0) {
 		unsigned int request = data->args.count;
 		unsigned int result = data->res.count;
 
@@ -415,20 +457,28 @@ static void nfs_readpage_result_partial(struct nfs_read_data *data, int status)
 	}
 }
 
+static const struct rpc_call_ops nfs_read_partial_ops = {
+	.rpc_call_done = nfs_readpage_result_partial,
+	.rpc_release = nfs_readdata_release,
+};
+
 /*
  * This is the callback from RPC telling us whether a reply was
  * received or some error occurred (timeout or socket shutdown).
  */
-static void nfs_readpage_result_full(struct nfs_read_data *data, int status)
+static void nfs_readpage_result_full(struct rpc_task *task, void *calldata)
 {
+	struct nfs_read_data *data = calldata;
 	unsigned int count = data->res.count;
 
+	if (nfs_readpage_result(task, data) != 0)
+		return;
 	while (!list_empty(&data->pages)) {
 		struct nfs_page *req = nfs_list_entry(data->pages.next);
 		struct page *page = req->wb_page;
 		nfs_list_remove_request(req);
 
-		if (status >= 0) {
+		if (task->tk_status >= 0) {
 			if (count < PAGE_CACHE_SIZE) {
 				if (count < req->wb_bytes)
 					memclear_highpage_flush(page,
@@ -444,22 +494,33 @@ static void nfs_readpage_result_full(struct nfs_read_data *data, int status)
 	}
 }
 
+static const struct rpc_call_ops nfs_read_full_ops = {
+	.rpc_call_done = nfs_readpage_result_full,
+	.rpc_release = nfs_readdata_release,
+};
+
 /*
  * This is the callback from RPC telling us whether a reply was
  * received or some error occurred (timeout or socket shutdown).
  */
-void nfs_readpage_result(struct rpc_task *task, void *calldata)
+int nfs_readpage_result(struct rpc_task *task, struct nfs_read_data *data)
 {
-	struct nfs_read_data *data = calldata;
 	struct nfs_readargs *argp = &data->args;
 	struct nfs_readres *resp = &data->res;
-	int status = task->tk_status;
+	int status;
 
 	dprintk("NFS: %4d nfs_readpage_result, (status %d)\n",
-		task->tk_pid, status);
+		task->tk_pid, task->tk_status);
+
+	status = NFS_PROTO(data->inode)->read_done(task, data);
+	if (status != 0)
+		return status;
+
+	nfs_add_stats(data->inode, NFSIOS_SERVERREADBYTES, resp->count);
 
 	/* Is this a short read? */
 	if (task->tk_status >= 0 && resp->count < argp->count && !resp->eof) {
+		nfs_inc_stats(data->inode, NFSIOS_SHORTREAD);
 		/* Has the server at least made some progress? */
 		if (resp->count != 0) {
 			/* Yes, so retry the read at the end of the data */
@@ -467,14 +528,14 @@ void nfs_readpage_result(struct rpc_task *task, void *calldata)
 			argp->pgbase += resp->count;
 			argp->count -= resp->count;
 			rpc_restart_call(task);
-			return;
+			return -EAGAIN;
 		}
 		task->tk_status = -EIO;
 	}
 	spin_lock(&data->inode->i_lock);
 	NFS_I(data->inode)->cache_validity |= NFS_INO_INVALID_ATIME;
 	spin_unlock(&data->inode->i_lock);
-	data->complete(data, status);
+	return 0;
 }
 
 /*
@@ -491,6 +552,9 @@ int nfs_readpage(struct file *file, struct page *page)
 
 	dprintk("NFS: nfs_readpage (%p %ld@%lu)\n",
 		page, PAGE_CACHE_SIZE, page->index);
+	nfs_inc_stats(inode, NFSIOS_VFSREADPAGE);
+	nfs_add_stats(inode, NFSIOS_READPAGES, 1);
+
 	/*
 	 * Try to flush any pending writes to the file..
 	 *
@@ -570,6 +634,7 @@ int nfs_readpages(struct file *filp, struct address_space *mapping,
 			inode->i_sb->s_id,
 			(long long)NFS_FILEID(inode),
 			nr_pages);
+	nfs_inc_stats(inode, NFSIOS_VFSREADPAGES);
 
 	if (filp == NULL) {
 		desc.ctx = nfs_find_open_context(inode, NULL, FMODE_READ);
@@ -582,6 +647,7 @@ int nfs_readpages(struct file *filp, struct address_space *mapping,
 	if (!list_empty(&head)) {
 		int err = nfs_pagein_list(&head, server->rpages);
 		if (!ret)
+			nfs_add_stats(inode, NFSIOS_READPAGES, err);
 			ret = err;
 	}
 	put_nfs_open_context(desc.ctx);

fs/nfs/unlink.c

@@ -163,10 +163,9 @@ nfs_async_unlink(struct dentry *dentry)
 	struct rpc_clnt	*clnt = NFS_CLIENT(dir->d_inode);
 	int		status = -ENOMEM;
 
-	data = kmalloc(sizeof(*data), GFP_KERNEL);
+	data = kzalloc(sizeof(*data), GFP_KERNEL);
 	if (!data)
 		goto out;
-	memset(data, 0, sizeof(*data));
 
 	data->cred = rpcauth_lookupcred(clnt->cl_auth, 0);
 	if (IS_ERR(data->cred)) {

fs/nfs/write.c

@@ -63,6 +63,7 @@
 #include <linux/smp_lock.h>
 
 #include "delegation.h"
+#include "iostat.h"
 
 #define NFSDBG_FACILITY		NFSDBG_PAGECACHE
 
@@ -76,23 +77,52 @@ static struct nfs_page * nfs_update_request(struct nfs_open_context*,
 					    struct inode *,
 					    struct page *,
 					    unsigned int, unsigned int);
-static void nfs_writeback_done_partial(struct nfs_write_data *, int);
-static void nfs_writeback_done_full(struct nfs_write_data *, int);
 static int nfs_wait_on_write_congestion(struct address_space *, int);
 static int nfs_wait_on_requests(struct inode *, unsigned long, unsigned int);
 static int nfs_flush_inode(struct inode *inode, unsigned long idx_start,
 			   unsigned int npages, int how);
+static const struct rpc_call_ops nfs_write_partial_ops;
+static const struct rpc_call_ops nfs_write_full_ops;
+static const struct rpc_call_ops nfs_commit_ops;
 
 static kmem_cache_t *nfs_wdata_cachep;
-mempool_t *nfs_wdata_mempool;
+static mempool_t *nfs_wdata_mempool;
 static mempool_t *nfs_commit_mempool;
 
 static DECLARE_WAIT_QUEUE_HEAD(nfs_write_congestion);
 
-static inline struct nfs_write_data *nfs_commit_alloc(unsigned int pagecount)
+struct nfs_write_data *nfs_commit_alloc(unsigned int pagecount)
 {
 	struct nfs_write_data *p = mempool_alloc(nfs_commit_mempool, SLAB_NOFS);
 
+	if (p) {
+		memset(p, 0, sizeof(*p));
+		INIT_LIST_HEAD(&p->pages);
+		if (pagecount < NFS_PAGEVEC_SIZE)
+			p->pagevec = &p->page_array[0];
+		else {
+			size_t size = ++pagecount * sizeof(struct page *);
+			p->pagevec = kzalloc(size, GFP_NOFS);
+			if (!p->pagevec) {
+				mempool_free(p, nfs_commit_mempool);
+				p = NULL;
+			}
+		}
+	}
+	return p;
+}
+
+void nfs_commit_free(struct nfs_write_data *p)
+{
+	if (p && (p->pagevec != &p->page_array[0]))
+		kfree(p->pagevec);
+	mempool_free(p, nfs_commit_mempool);
+}
+
+struct nfs_write_data *nfs_writedata_alloc(unsigned int pagecount)
+{
+	struct nfs_write_data *p = mempool_alloc(nfs_wdata_mempool, SLAB_NOFS);
+
 	if (p) {
 		memset(p, 0, sizeof(*p));
 		INIT_LIST_HEAD(&p->pages);
@@ -104,7 +134,7 @@ static inline struct nfs_write_data *nfs_commit_alloc(unsigned int pagecount)
 			if (p->pagevec) {
 				memset(p->pagevec, 0, size);
 			} else {
-				mempool_free(p, nfs_commit_mempool);
+				mempool_free(p, nfs_wdata_mempool);
 				p = NULL;
 			}
 		}
@@ -112,11 +142,11 @@ static inline struct nfs_write_data *nfs_commit_alloc(unsigned int pagecount)
 	return p;
 }
 
-static inline void nfs_commit_free(struct nfs_write_data *p)
+void nfs_writedata_free(struct nfs_write_data *p)
 {
 	if (p && (p->pagevec != &p->page_array[0]))
 		kfree(p->pagevec);
-	mempool_free(p, nfs_commit_mempool);
+	mempool_free(p, nfs_wdata_mempool);
 }
 
 void nfs_writedata_release(void *wdata)
@@ -136,6 +166,7 @@ static void nfs_grow_file(struct page *page, unsigned int offset, unsigned int c
 	end = ((loff_t)page->index << PAGE_CACHE_SHIFT) + ((loff_t)offset+count);
 	if (i_size >= end)
 		return;
+	nfs_inc_stats(inode, NFSIOS_EXTENDWRITE);
 	i_size_write(inode, end);
 }
 
@@ -225,6 +256,7 @@ static int nfs_writepage_sync(struct nfs_open_context *ctx, struct inode *inode,
 	        wdata->args.pgbase += result;
 		written += result;
 		count -= result;
+		nfs_add_stats(inode, NFSIOS_SERVERWRITTENBYTES, result);
 	} while (count);
 	/* Update file length */
 	nfs_grow_file(page, offset, written);
@@ -281,6 +313,9 @@ int nfs_writepage(struct page *page, struct writeback_control *wbc)
 	int priority = wb_priority(wbc);
 	int err;
 
+	nfs_inc_stats(inode, NFSIOS_VFSWRITEPAGE);
+	nfs_add_stats(inode, NFSIOS_WRITEPAGES, 1);
+
 	/*
 	 * Note: We need to ensure that we have a reference to the inode
 	 *       if we are to do asynchronous writes. If not, waiting
@@ -345,6 +380,8 @@ int nfs_writepages(struct address_space *mapping, struct writeback_control *wbc)
 	struct inode *inode = mapping->host;
 	int err;
 
+	nfs_inc_stats(inode, NFSIOS_VFSWRITEPAGES);
+
 	err = generic_writepages(mapping, wbc);
 	if (err)
 		return err;
@@ -356,6 +393,7 @@ int nfs_writepages(struct address_space *mapping, struct writeback_control *wbc)
 	err = nfs_flush_inode(inode, 0, 0, wb_priority(wbc));
 	if (err < 0)
 		goto out;
+	nfs_add_stats(inode, NFSIOS_WRITEPAGES, err);
 	wbc->nr_to_write -= err;
 	if (!wbc->nonblocking && wbc->sync_mode == WB_SYNC_ALL) {
 		err = nfs_wait_on_requests(inode, 0, 0);
@@ -391,6 +429,7 @@ static int nfs_inode_add_request(struct inode *inode, struct nfs_page *req)
 		if (nfs_have_delegation(inode, FMODE_WRITE))
 			nfsi->change_attr++;
 	}
+	SetPagePrivate(req->wb_page);
 	nfsi->npages++;
 	atomic_inc(&req->wb_count);
 	return 0;
@@ -407,6 +446,7 @@ static void nfs_inode_remove_request(struct nfs_page *req)
 	BUG_ON (!NFS_WBACK_BUSY(req));
 
 	spin_lock(&nfsi->req_lock);
+	ClearPagePrivate(req->wb_page);
 	radix_tree_delete(&nfsi->nfs_page_tree, req->wb_index);
 	nfsi->npages--;
 	if (!nfsi->npages) {
@@ -499,8 +539,7 @@ nfs_mark_request_commit(struct nfs_page *req)
  *
  * Interruptible by signals only if mounted with intr flag.
  */
-static int
-nfs_wait_on_requests(struct inode *inode, unsigned long idx_start, unsigned int npages)
+static int nfs_wait_on_requests_locked(struct inode *inode, unsigned long idx_start, unsigned int npages)
 {
 	struct nfs_inode *nfsi = NFS_I(inode);
 	struct nfs_page *req;
@@ -513,7 +552,6 @@ nfs_wait_on_requests(struct inode *inode, unsigned long idx_start, unsigned int
 	else
 		idx_end = idx_start + npages - 1;
 
-	spin_lock(&nfsi->req_lock);
 	next = idx_start;
 	while (radix_tree_gang_lookup_tag(&nfsi->nfs_page_tree, (void **)&req, next, 1, NFS_PAGE_TAG_WRITEBACK)) {
 		if (req->wb_index > idx_end)
@@ -526,15 +564,25 @@ nfs_wait_on_requests(struct inode *inode, unsigned long idx_start, unsigned int
 		spin_unlock(&nfsi->req_lock);
 		error = nfs_wait_on_request(req);
 		nfs_release_request(req);
+		spin_lock(&nfsi->req_lock);
 		if (error < 0)
 			return error;
-		spin_lock(&nfsi->req_lock);
 		res++;
 	}
-	spin_unlock(&nfsi->req_lock);
 	return res;
 }
 
+static int nfs_wait_on_requests(struct inode *inode, unsigned long idx_start, unsigned int npages)
+{
+	struct nfs_inode *nfsi = NFS_I(inode);
+	int ret;
+
+	spin_lock(&nfsi->req_lock);
+	ret = nfs_wait_on_requests_locked(inode, idx_start, npages);
+	spin_unlock(&nfsi->req_lock);
+	return ret;
+}
+
 /*
  * nfs_scan_dirty - Scan an inode for dirty requests
  * @inode: NFS inode to scan
@@ -586,6 +634,11 @@ nfs_scan_commit(struct inode *inode, struct list_head *dst, unsigned long idx_st
 	}
 	return res;
 }
+#else
+static inline int nfs_scan_commit(struct inode *inode, struct list_head *dst, unsigned long idx_start, unsigned int npages)
+{
+	return 0;
+}
 #endif
 
 static int nfs_wait_on_write_congestion(struct address_space *mapping, int intr)
@@ -598,6 +651,9 @@ static int nfs_wait_on_write_congestion(struct address_space *mapping, int intr)
 
 	if (!bdi_write_congested(bdi))
 		return 0;
+
+	nfs_inc_stats(mapping->host, NFSIOS_CONGESTIONWAIT);
+
 	if (intr) {
 		struct rpc_clnt *clnt = NFS_CLIENT(mapping->host);
 		sigset_t oldset;
@@ -653,8 +709,11 @@ static struct nfs_page * nfs_update_request(struct nfs_open_context* ctx,
 				spin_unlock(&nfsi->req_lock);
 				error = nfs_wait_on_request(req);
 				nfs_release_request(req);
-				if (error < 0)
+				if (error < 0) {
+					if (new)
+						nfs_release_request(new);
 					return ERR_PTR(error);
+				}
 				continue;
 			}
 			spin_unlock(&nfsi->req_lock);
@@ -748,6 +807,8 @@ int nfs_updatepage(struct file *file, struct page *page,
 	struct nfs_page	*req;
 	int		status = 0;
 
+	nfs_inc_stats(inode, NFSIOS_VFSUPDATEPAGE);
+
 	dprintk("NFS:      nfs_updatepage(%s/%s %d@%Ld)\n",
 		file->f_dentry->d_parent->d_name.name,
 		file->f_dentry->d_name.name, count,
@@ -857,10 +918,12 @@ static inline int flush_task_priority(int how)
  */
 static void nfs_write_rpcsetup(struct nfs_page *req,
 		struct nfs_write_data *data,
+		const struct rpc_call_ops *call_ops,
 		unsigned int count, unsigned int offset,
 		int how)
 {
 	struct inode		*inode;
+	int flags;
 
 	/* Set up the RPC argument and reply structs
 	 * NB: take care not to mess about with data->commit et al. */
@@ -881,6 +944,9 @@ static void nfs_write_rpcsetup(struct nfs_page *req,
 	data->res.verf    = &data->verf;
 	nfs_fattr_init(&data->fattr);
 
+	/* Set up the initial task struct.  */
+	flags = (how & FLUSH_SYNC) ? 0 : RPC_TASK_ASYNC;
+	rpc_init_task(&data->task, NFS_CLIENT(inode), flags, call_ops, data);
 	NFS_PROTO(inode)->write_setup(data, how);
 
 	data->task.tk_priority = flush_task_priority(how);
@@ -910,7 +976,7 @@ static void nfs_execute_write(struct nfs_write_data *data)
  * Generate multiple small requests to write out a single
  * contiguous dirty area on one page.
  */
-static int nfs_flush_multi(struct list_head *head, struct inode *inode, int how)
+static int nfs_flush_multi(struct inode *inode, struct list_head *head, int how)
 {
 	struct nfs_page *req = nfs_list_entry(head->next);
 	struct page *page = req->wb_page;
@@ -944,14 +1010,15 @@ static int nfs_flush_multi(struct list_head *head, struct inode *inode, int how)
 		list_del_init(&data->pages);
 
 		data->pagevec[0] = page;
-		data->complete = nfs_writeback_done_partial;
 
 		if (nbytes > wsize) {
-			nfs_write_rpcsetup(req, data, wsize, offset, how);
+			nfs_write_rpcsetup(req, data, &nfs_write_partial_ops,
+					wsize, offset, how);
 			offset += wsize;
 			nbytes -= wsize;
 		} else {
-			nfs_write_rpcsetup(req, data, nbytes, offset, how);
+			nfs_write_rpcsetup(req, data, &nfs_write_partial_ops,
+					nbytes, offset, how);
 			nbytes = 0;
 		}
 		nfs_execute_write(data);
@@ -978,16 +1045,13 @@ out_bad:
  * This is the case if nfs_updatepage detects a conflicting request
  * that has been written but not committed.
  */
-static int nfs_flush_one(struct list_head *head, struct inode *inode, int how)
+static int nfs_flush_one(struct inode *inode, struct list_head *head, int how)
 {
 	struct nfs_page		*req;
 	struct page		**pages;
 	struct nfs_write_data	*data;
 	unsigned int		count;
 
-	if (NFS_SERVER(inode)->wsize < PAGE_CACHE_SIZE)
-		return nfs_flush_multi(head, inode, how);
-
 	data = nfs_writedata_alloc(NFS_SERVER(inode)->wpages);
 	if (!data)
 		goto out_bad;
@@ -1005,9 +1069,8 @@ static int nfs_flush_one(struct list_head *head, struct inode *inode, int how)
 	}
 	req = nfs_list_entry(data->pages.next);
 
-	data->complete = nfs_writeback_done_full;
 	/* Set up the argument struct */
-	nfs_write_rpcsetup(req, data, count, 0, how);
+	nfs_write_rpcsetup(req, data, &nfs_write_full_ops, count, 0, how);
 
 	nfs_execute_write(data);
 	return 0;
@@ -1021,24 +1084,32 @@ static int nfs_flush_one(struct list_head *head, struct inode *inode, int how)
 	return -ENOMEM;
 }
 
-static int
-nfs_flush_list(struct list_head *head, int wpages, int how)
+static int nfs_flush_list(struct inode *inode, struct list_head *head, int npages, int how)
 {
 	LIST_HEAD(one_request);
-	struct nfs_page		*req;
-	int			error = 0;
-	unsigned int		pages = 0;
+	int (*flush_one)(struct inode *, struct list_head *, int);
+	struct nfs_page	*req;
+	int wpages = NFS_SERVER(inode)->wpages;
+	int wsize = NFS_SERVER(inode)->wsize;
+	int error;
 
-	while (!list_empty(head)) {
-		pages += nfs_coalesce_requests(head, &one_request, wpages);
+	flush_one = nfs_flush_one;
+	if (wsize < PAGE_CACHE_SIZE)
+		flush_one = nfs_flush_multi;
+	/* For single writes, FLUSH_STABLE is more efficient */
+	if (npages <= wpages && npages == NFS_I(inode)->npages
+			&& nfs_list_entry(head->next)->wb_bytes <= wsize)
+		how |= FLUSH_STABLE;
+
+	do {
+		nfs_coalesce_requests(head, &one_request, wpages);
 		req = nfs_list_entry(one_request.next);
-		error = nfs_flush_one(&one_request, req->wb_context->dentry->d_inode, how);
+		error = flush_one(inode, &one_request, how);
 		if (error < 0)
-			break;
-	}
-	if (error >= 0)
-		return pages;
-
+			goto out_err;
+	} while (!list_empty(head));
+	return 0;
+out_err:
 	while (!list_empty(head)) {
 		req = nfs_list_entry(head->next);
 		nfs_list_remove_request(req);
@@ -1051,8 +1122,9 @@ nfs_flush_list(struct list_head *head, int wpages, int how)
 /*
  * Handle a write reply that flushed part of a page.
  */
-static void nfs_writeback_done_partial(struct nfs_write_data *data, int status)
+static void nfs_writeback_done_partial(struct rpc_task *task, void *calldata)
 {
+	struct nfs_write_data	*data = calldata;
 	struct nfs_page		*req = data->req;
 	struct page		*page = req->wb_page;
 
@@ -1062,11 +1134,14 @@ static void nfs_writeback_done_partial(struct nfs_write_data *data, int status)
 		req->wb_bytes,
 		(long long)req_offset(req));
 
-	if (status < 0) {
+	if (nfs_writeback_done(task, data) != 0)
+		return;
+
+	if (task->tk_status < 0) {
 		ClearPageUptodate(page);
 		SetPageError(page);
-		req->wb_context->error = status;
-		dprintk(", error = %d\n", status);
+		req->wb_context->error = task->tk_status;
+		dprintk(", error = %d\n", task->tk_status);
 	} else {
 #if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
 		if (data->verf.committed < NFS_FILE_SYNC) {
@@ -1087,6 +1162,11 @@ static void nfs_writeback_done_partial(struct nfs_write_data *data, int status)
 		nfs_writepage_release(req);
 }
 
+static const struct rpc_call_ops nfs_write_partial_ops = {
+	.rpc_call_done = nfs_writeback_done_partial,
+	.rpc_release = nfs_writedata_release,
+};
+
 /*
  * Handle a write reply that flushes a whole page.
  *
@@ -1094,11 +1174,15 @@ static void nfs_writeback_done_partial(struct nfs_write_data *data, int status)
  *	  writebacks since the page->count is kept > 1 for as long
  *	  as the page has a write request pending.
  */
-static void nfs_writeback_done_full(struct nfs_write_data *data, int status)
+static void nfs_writeback_done_full(struct rpc_task *task, void *calldata)
 {
+	struct nfs_write_data	*data = calldata;
 	struct nfs_page		*req;
 	struct page		*page;
 
+	if (nfs_writeback_done(task, data) != 0)
+		return;
+
 	/* Update attributes as result of writeback. */
 	while (!list_empty(&data->pages)) {
 		req = nfs_list_entry(data->pages.next);
@@ -1111,13 +1195,13 @@ static void nfs_writeback_done_full(struct nfs_write_data *data, int status)
 			req->wb_bytes,
 			(long long)req_offset(req));
 
-		if (status < 0) {
+		if (task->tk_status < 0) {
 			ClearPageUptodate(page);
 			SetPageError(page);
-			req->wb_context->error = status;
+			req->wb_context->error = task->tk_status;
 			end_page_writeback(page);
 			nfs_inode_remove_request(req);
-			dprintk(", error = %d\n", status);
+			dprintk(", error = %d\n", task->tk_status);
 			goto next;
 		}
 		end_page_writeback(page);
@@ -1139,18 +1223,30 @@ static void nfs_writeback_done_full(struct nfs_write_data *data, int status)
 	}
 }
 
+static const struct rpc_call_ops nfs_write_full_ops = {
+	.rpc_call_done = nfs_writeback_done_full,
+	.rpc_release = nfs_writedata_release,
+};
+
+
 /*
  * This function is called when the WRITE call is complete.
  */
-void nfs_writeback_done(struct rpc_task *task, void *calldata)
+int nfs_writeback_done(struct rpc_task *task, struct nfs_write_data *data)
 {
-	struct nfs_write_data	*data = calldata;
 	struct nfs_writeargs	*argp = &data->args;
 	struct nfs_writeres	*resp = &data->res;
+	int status;
 
 	dprintk("NFS: %4d nfs_writeback_done (status %d)\n",
 		task->tk_pid, task->tk_status);
 
+	/* Call the NFS version-specific code */
+	status = NFS_PROTO(data->inode)->write_done(task, data);
+	if (status != 0)
+		return status;
+	nfs_add_stats(data->inode, NFSIOS_SERVERWRITTENBYTES, resp->count);
+
 #if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
 	if (resp->verf->committed < argp->stable && task->tk_status >= 0) {
 		/* We tried a write call, but the server did not
@@ -1176,6 +1272,8 @@ void nfs_writeback_done(struct rpc_task *task, void *calldata)
 	if (task->tk_status >= 0 && resp->count < argp->count) {
 		static unsigned long    complain;
 
+		nfs_inc_stats(data->inode, NFSIOS_SHORTWRITE);
+
 		/* Has the server at least made some progress? */
 		if (resp->count != 0) {
 			/* Was this an NFSv2 write or an NFSv3 stable write? */
@@ -1191,7 +1289,7 @@ void nfs_writeback_done(struct rpc_task *task, void *calldata)
 				argp->stable = NFS_FILE_SYNC;
 			}
 			rpc_restart_call(task);
-			return;
+			return -EAGAIN;
 		}
 		if (time_before(complain, jiffies)) {
 			printk(KERN_WARNING
@@ -1202,11 +1300,7 @@ void nfs_writeback_done(struct rpc_task *task, void *calldata)
 		/* Can't do anything about it except throw an error. */
 		task->tk_status = -EIO;
 	}
-
-	/*
-	 * Process the nfs_page list
-	 */
-	data->complete(data, task->tk_status);
+	return 0;
 }
 
 
@@ -1220,10 +1314,12 @@ void nfs_commit_release(void *wdata)
  * Set up the argument/result storage required for the RPC call.
  */
 static void nfs_commit_rpcsetup(struct list_head *head,
-		struct nfs_write_data *data, int how)
+		struct nfs_write_data *data,
+		int how)
 {
 	struct nfs_page		*first;
 	struct inode		*inode;
+	int flags;
 
 	/* Set up the RPC argument and reply structs
 	 * NB: take care not to mess about with data->commit et al. */
@@ -1243,7 +1339,10 @@ static void nfs_commit_rpcsetup(struct list_head *head,
 	data->res.fattr   = &data->fattr;
 	data->res.verf    = &data->verf;
 	nfs_fattr_init(&data->fattr);
-	
+
+	/* Set up the initial task struct.  */
+	flags = (how & FLUSH_SYNC) ? 0 : RPC_TASK_ASYNC;
+	rpc_init_task(&data->task, NFS_CLIENT(inode), flags, &nfs_commit_ops, data);
 	NFS_PROTO(inode)->commit_setup(data, how);
 
 	data->task.tk_priority = flush_task_priority(how);
@@ -1284,7 +1383,7 @@ nfs_commit_list(struct inode *inode, struct list_head *head, int how)
 /*
  * COMMIT call returned
  */
-void nfs_commit_done(struct rpc_task *task, void *calldata)
+static void nfs_commit_done(struct rpc_task *task, void *calldata)
 {
 	struct nfs_write_data	*data = calldata;
 	struct nfs_page		*req;
@@ -1293,6 +1392,10 @@ void nfs_commit_done(struct rpc_task *task, void *calldata)
         dprintk("NFS: %4d nfs_commit_done (status %d)\n",
                                 task->tk_pid, task->tk_status);
 
+	/* Call the NFS version-specific code */
+	if (NFS_PROTO(data->inode)->commit_done(task, data) != 0)
+		return;
+
 	while (!list_empty(&data->pages)) {
 		req = nfs_list_entry(data->pages.next);
 		nfs_list_remove_request(req);
@@ -1326,6 +1429,16 @@ void nfs_commit_done(struct rpc_task *task, void *calldata)
 	}
 	sub_page_state(nr_unstable,res);
 }
+
+static const struct rpc_call_ops nfs_commit_ops = {
+	.rpc_call_done = nfs_commit_done,
+	.rpc_release = nfs_commit_release,
+};
+#else
+static inline int nfs_commit_list(struct inode *inode, struct list_head *head, int how)
+{
+	return 0;
+}
 #endif
 
 static int nfs_flush_inode(struct inode *inode, unsigned long idx_start,
@@ -1333,24 +1446,16 @@ static int nfs_flush_inode(struct inode *inode, unsigned long idx_start,
 {
 	struct nfs_inode *nfsi = NFS_I(inode);
 	LIST_HEAD(head);
-	int			res,
-				error = 0;
+	int res;
 
 	spin_lock(&nfsi->req_lock);
 	res = nfs_scan_dirty(inode, &head, idx_start, npages);
 	spin_unlock(&nfsi->req_lock);
 	if (res) {
-		struct nfs_server *server = NFS_SERVER(inode);
-
-		/* For single writes, FLUSH_STABLE is more efficient */
-		if (res == nfsi->npages && nfsi->npages <= server->wpages) {
-			if (res > 1 || nfs_list_entry(head.next)->wb_bytes <= server->wsize)
-				how |= FLUSH_STABLE;
-		}
-		error = nfs_flush_list(&head, server->wpages, how);
+		int error = nfs_flush_list(inode, &head, res, how);
+		if (error < 0)
+			return error;
 	}
-	if (error < 0)
-		return error;
 	return res;
 }
 
@@ -1359,14 +1464,13 @@ int nfs_commit_inode(struct inode *inode, int how)
 {
 	struct nfs_inode *nfsi = NFS_I(inode);
 	LIST_HEAD(head);
-	int			res,
-				error = 0;
+	int res;
 
 	spin_lock(&nfsi->req_lock);
 	res = nfs_scan_commit(inode, &head, 0, 0);
 	spin_unlock(&nfsi->req_lock);
 	if (res) {
-		error = nfs_commit_list(inode, &head, how);
+		int error = nfs_commit_list(inode, &head, how);
 		if (error < 0)
 			return error;
 	}
@@ -1374,28 +1478,38 @@ int nfs_commit_inode(struct inode *inode, int how)
 }
 #endif
 
-int nfs_sync_inode(struct inode *inode, unsigned long idx_start,
-		  unsigned int npages, int how)
+int nfs_sync_inode_wait(struct inode *inode, unsigned long idx_start,
+		unsigned int npages, int how)
 {
+	struct nfs_inode *nfsi = NFS_I(inode);
+	LIST_HEAD(head);
 	int nocommit = how & FLUSH_NOCOMMIT;
-	int wait = how & FLUSH_WAIT;
-	int error;
-
-	how &= ~(FLUSH_WAIT|FLUSH_NOCOMMIT);
+	int pages, ret;
 
+	how &= ~FLUSH_NOCOMMIT;
+	spin_lock(&nfsi->req_lock);
 	do {
-		if (wait) {
-			error = nfs_wait_on_requests(inode, idx_start, npages);
-			if (error != 0)
-				continue;
-		}
-		error = nfs_flush_inode(inode, idx_start, npages, how);
-		if (error != 0)
+		ret = nfs_wait_on_requests_locked(inode, idx_start, npages);
+		if (ret != 0)
 			continue;
-		if (!nocommit)
-			error = nfs_commit_inode(inode, how);
-	} while (error > 0);
-	return error;
+		pages = nfs_scan_dirty(inode, &head, idx_start, npages);
+		if (pages != 0) {
+			spin_unlock(&nfsi->req_lock);
+			ret = nfs_flush_list(inode, &head, pages, how);
+			spin_lock(&nfsi->req_lock);
+			continue;
+		}
+		if (nocommit)
+			break;
+		pages = nfs_scan_commit(inode, &head, 0, 0);
+		if (pages == 0)
+			break;
+		spin_unlock(&nfsi->req_lock);
+		ret = nfs_commit_list(inode, &head, how);
+		spin_lock(&nfsi->req_lock);
+	} while (ret >= 0);
+	spin_unlock(&nfsi->req_lock);
+	return ret;
 }
 
 int nfs_init_writepagecache(void)

fs/nfsd/nfs4callback.c

@@ -326,6 +326,8 @@ out:
         .p_encode = (kxdrproc_t) nfs4_xdr_##argtype,                    \
         .p_decode = (kxdrproc_t) nfs4_xdr_##restype,                    \
         .p_bufsiz = MAX(NFS4_##argtype##_sz,NFS4_##restype##_sz) << 2,  \
+        .p_statidx = NFSPROC4_CB_##call,				\
+	.p_name   = #proc,                                              \
 }
 
 static struct rpc_procinfo     nfs4_cb_procedures[] = {

fs/nfsd/nfs4state.c

@@ -2639,7 +2639,7 @@ nfsd4_lock(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_lock
 	struct nfs4_stateid *lock_stp;
 	struct file *filp;
 	struct file_lock file_lock;
-	struct file_lock *conflock;
+	struct file_lock conflock;
 	int status = 0;
 	unsigned int strhashval;
 
@@ -2775,11 +2775,11 @@ conflicting_lock:
 	/* XXX There is a race here. Future patch needed to provide 
 	 * an atomic posix_lock_and_test_file
 	 */
-	if (!(conflock = posix_test_lock(filp, &file_lock))) {
+	if (!posix_test_lock(filp, &file_lock, &conflock)) {
 		status = nfserr_serverfault;
 		goto out;
 	}
-	nfs4_set_lock_denied(conflock, &lock->lk_denied);
+	nfs4_set_lock_denied(&conflock, &lock->lk_denied);
 out:
 	if (status && lock->lk_is_new && lock_sop)
 		release_stateowner(lock_sop);
@@ -2800,7 +2800,7 @@ nfsd4_lockt(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_lock
 	struct inode *inode;
 	struct file file;
 	struct file_lock file_lock;
-	struct file_lock *conflicting_lock;
+	struct file_lock conflock;
 	int status;
 
 	if (nfs4_in_grace())
@@ -2864,10 +2864,9 @@ nfsd4_lockt(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_lock
 	file.f_dentry = current_fh->fh_dentry;
 
 	status = nfs_ok;
-	conflicting_lock = posix_test_lock(&file, &file_lock);
-	if (conflicting_lock) {
+	if (posix_test_lock(&file, &file_lock, &conflock)) {
 		status = nfserr_denied;
-		nfs4_set_lock_denied(conflicting_lock, &lockt->lt_denied);
+		nfs4_set_lock_denied(&conflock, &lockt->lt_denied);
 	}
 out:
 	nfs4_unlock_state();

fs/proc/base.c

@@ -104,6 +104,7 @@ enum pid_directory_inos {
 	PROC_TGID_MAPS,
 	PROC_TGID_NUMA_MAPS,
 	PROC_TGID_MOUNTS,
+	PROC_TGID_MOUNTSTATS,
 	PROC_TGID_WCHAN,
 #ifdef CONFIG_MMU
 	PROC_TGID_SMAPS,
@@ -144,6 +145,7 @@ enum pid_directory_inos {
 	PROC_TID_MAPS,
 	PROC_TID_NUMA_MAPS,
 	PROC_TID_MOUNTS,
+	PROC_TID_MOUNTSTATS,
 	PROC_TID_WCHAN,
 #ifdef CONFIG_MMU
 	PROC_TID_SMAPS,
@@ -201,6 +203,7 @@ static struct pid_entry tgid_base_stuff[] = {
 	E(PROC_TGID_ROOT,      "root",    S_IFLNK|S_IRWXUGO),
 	E(PROC_TGID_EXE,       "exe",     S_IFLNK|S_IRWXUGO),
 	E(PROC_TGID_MOUNTS,    "mounts",  S_IFREG|S_IRUGO),
+	E(PROC_TGID_MOUNTSTATS, "mountstats", S_IFREG|S_IRUSR),
 #ifdef CONFIG_MMU
 	E(PROC_TGID_SMAPS,     "smaps",   S_IFREG|S_IRUGO),
 #endif
@@ -732,6 +735,38 @@ static struct file_operations proc_mounts_operations = {
 	.poll		= mounts_poll,
 };
 
+extern struct seq_operations mountstats_op;
+static int mountstats_open(struct inode *inode, struct file *file)
+{
+	struct task_struct *task = proc_task(inode);
+	int ret = seq_open(file, &mountstats_op);
+
+	if (!ret) {
+		struct seq_file *m = file->private_data;
+		struct namespace *namespace;
+		task_lock(task);
+		namespace = task->namespace;
+		if (namespace)
+			get_namespace(namespace);
+		task_unlock(task);
+
+		if (namespace)
+			m->private = namespace;
+		else {
+			seq_release(inode, file);
+			ret = -EINVAL;
+		}
+	}
+	return ret;
+}
+
+static struct file_operations proc_mountstats_operations = {
+	.open		= mountstats_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= mounts_release,
+};
+
 #define PROC_BLOCK_SIZE	(3*1024)		/* 4K page size but our output routines use some slack for overruns */
 
 static ssize_t proc_info_read(struct file * file, char __user * buf,
@@ -1730,6 +1765,10 @@ static struct dentry *proc_pident_lookup(struct inode *dir,
 			inode->i_fop = &proc_smaps_operations;
 			break;
 #endif
+		case PROC_TID_MOUNTSTATS:
+		case PROC_TGID_MOUNTSTATS:
+			inode->i_fop = &proc_mountstats_operations;
+			break;
 #ifdef CONFIG_SECURITY
 		case PROC_TID_ATTR:
 			inode->i_nlink = 2;

include/linux/fs.h

@@ -678,7 +678,6 @@ extern spinlock_t files_lock;
 #define FL_POSIX	1
 #define FL_FLOCK	2
 #define FL_ACCESS	8	/* not trying to lock, just looking */
-#define FL_LOCKD	16	/* lock held by rpc.lockd */
 #define FL_LEASE	32	/* lease held on this file */
 #define FL_SLEEP	128	/* A blocking lock */
 
@@ -742,8 +741,6 @@ struct file_lock {
 #define OFFT_OFFSET_MAX	INT_LIMIT(off_t)
 #endif
 
-extern struct list_head file_lock_list;
-
 #include <linux/fcntl.h>
 
 extern int fcntl_getlk(struct file *, struct flock __user *);
@@ -765,10 +762,9 @@ extern void locks_init_lock(struct file_lock *);
 extern void locks_copy_lock(struct file_lock *, struct file_lock *);
 extern void locks_remove_posix(struct file *, fl_owner_t);
 extern void locks_remove_flock(struct file *);
-extern struct file_lock *posix_test_lock(struct file *, struct file_lock *);
+extern int posix_test_lock(struct file *, struct file_lock *, struct file_lock *);
 extern int posix_lock_file(struct file *, struct file_lock *);
 extern int posix_lock_file_wait(struct file *, struct file_lock *);
-extern void posix_block_lock(struct file_lock *, struct file_lock *);
 extern int posix_unblock_lock(struct file *, struct file_lock *);
 extern int posix_locks_deadlock(struct file_lock *, struct file_lock *);
 extern int flock_lock_file_wait(struct file *filp, struct file_lock *fl);
@@ -1097,6 +1093,7 @@ struct super_operations {
 	void (*umount_begin) (struct super_block *);
 
 	int (*show_options)(struct seq_file *, struct vfsmount *);
+	int (*show_stats)(struct seq_file *, struct vfsmount *);
 
 	ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t);
 	ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t);

include/linux/lockd/lockd.h

@@ -14,6 +14,7 @@
 #include <linux/config.h>
 #include <linux/in.h>
 #include <linux/fs.h>
+#include <linux/kref.h>
 #include <linux/utsname.h>
 #include <linux/nfsd/nfsfh.h>
 #include <linux/lockd/bind.h>
@@ -58,6 +59,8 @@ struct nlm_host {
 	unsigned long		h_expires;	/* eligible for GC */
 	struct list_head	h_lockowners;	/* Lockowners for the client */
 	spinlock_t		h_lock;
+	struct list_head	h_granted;	/* Locks in GRANTED state */
+	struct list_head	h_reclaim;	/* Locks in RECLAIM state */
 };
 
 /*
@@ -83,9 +86,9 @@ struct nlm_rqst {
 	struct nlm_host *	a_host;		/* host handle */
 	struct nlm_args		a_args;		/* arguments */
 	struct nlm_res		a_res;		/* result */
-	struct nlm_wait *	a_block;
+	struct nlm_block *	a_block;
 	unsigned int		a_retries;	/* Retry count */
-	char			a_owner[NLMCLNT_OHSIZE];
+	u8			a_owner[NLMCLNT_OHSIZE];
 };
 
 /*
@@ -110,16 +113,16 @@ struct nlm_file {
  */
 #define NLM_NEVER		(~(unsigned long) 0)
 struct nlm_block {
+	struct kref		b_count;	/* Reference count */
 	struct nlm_block *	b_next;		/* linked list (all blocks) */
 	struct nlm_block *	b_fnext;	/* linked list (per file) */
-	struct nlm_rqst		b_call;		/* RPC args & callback info */
+	struct nlm_rqst	*	b_call;		/* RPC args & callback info */
 	struct svc_serv *	b_daemon;	/* NLM service */
 	struct nlm_host *	b_host;		/* host handle for RPC clnt */
 	unsigned long		b_when;		/* next re-xmit */
 	unsigned int		b_id;		/* block id */
 	unsigned char		b_queued;	/* re-queued */
 	unsigned char		b_granted;	/* VFS granted lock */
-	unsigned char		b_incall;	/* doing callback */
 	unsigned char		b_done;		/* callback complete */
 	struct nlm_file *	b_file;		/* file in question */
 };
@@ -145,15 +148,16 @@ extern unsigned long		nlmsvc_timeout;
 /*
  * Lockd client functions
  */
-struct nlm_rqst * nlmclnt_alloc_call(void);
-int		  nlmclnt_prepare_block(struct nlm_rqst *req, struct nlm_host *host, struct file_lock *fl);
-void		  nlmclnt_finish_block(struct nlm_rqst *req);
-long		  nlmclnt_block(struct nlm_rqst *req, long timeout);
+struct nlm_rqst * nlm_alloc_call(struct nlm_host *host);
+void		  nlm_release_call(struct nlm_rqst *);
+int		  nlm_async_call(struct nlm_rqst *, u32, const struct rpc_call_ops *);
+int		  nlm_async_reply(struct nlm_rqst *, u32, const struct rpc_call_ops *);
+struct nlm_wait * nlmclnt_prepare_block(struct nlm_host *host, struct file_lock *fl);
+void		  nlmclnt_finish_block(struct nlm_wait *block);
+int		  nlmclnt_block(struct nlm_wait *block, struct nlm_rqst *req, long timeout);
 u32		  nlmclnt_grant(const struct sockaddr_in *addr, const struct nlm_lock *);
 void		  nlmclnt_recovery(struct nlm_host *, u32);
 int		  nlmclnt_reclaim(struct nlm_host *, struct file_lock *);
-int		  nlmclnt_setgrantargs(struct nlm_rqst *, struct nlm_lock *);
-void		  nlmclnt_freegrantargs(struct nlm_rqst *);
 
 /*
  * Host cache
@@ -172,7 +176,6 @@ extern struct nlm_host *nlm_find_client(void);
 /*
  * Server-side lock handling
  */
-int		  nlmsvc_async_call(struct nlm_rqst *, u32, const struct rpc_call_ops *);
 u32		  nlmsvc_lock(struct svc_rqst *, struct nlm_file *,
 					struct nlm_lock *, int, struct nlm_cookie *);
 u32		  nlmsvc_unlock(struct nlm_file *, struct nlm_lock *);
@@ -180,7 +183,7 @@ u32		  nlmsvc_testlock(struct nlm_file *, struct nlm_lock *,
 					struct nlm_lock *);
 u32		  nlmsvc_cancel_blocked(struct nlm_file *, struct nlm_lock *);
 unsigned long	  nlmsvc_retry_blocked(void);
-int		  nlmsvc_traverse_blocks(struct nlm_host *, struct nlm_file *,
+void		  nlmsvc_traverse_blocks(struct nlm_host *, struct nlm_file *,
 					int action);
 void	  nlmsvc_grant_reply(struct svc_rqst *, struct nlm_cookie *, u32);
 

include/linux/lockd/share.h

@@ -25,6 +25,6 @@ u32	nlmsvc_share_file(struct nlm_host *, struct nlm_file *,
 					       struct nlm_args *);
 u32	nlmsvc_unshare_file(struct nlm_host *, struct nlm_file *,
 					       struct nlm_args *);
-int	nlmsvc_traverse_shares(struct nlm_host *, struct nlm_file *, int);
+void	nlmsvc_traverse_shares(struct nlm_host *, struct nlm_file *, int);
 
 #endif /* LINUX_LOCKD_SHARE_H */

include/linux/lockd/xdr.h

@@ -28,6 +28,7 @@ struct nlm_lock {
 	int			len; 	/* length of "caller" */
 	struct nfs_fh		fh;
 	struct xdr_netobj	oh;
+	u32			svid;
 	struct file_lock	fl;
 };
 

include/linux/nfs_fs.h

@@ -56,9 +56,7 @@
  * When flushing a cluster of dirty pages, there can be different
  * strategies:
  */
-#define FLUSH_AGING		0	/* only flush old buffers */
 #define FLUSH_SYNC		1	/* file being synced, or contention */
-#define FLUSH_WAIT		2	/* wait for completion */
 #define FLUSH_STABLE		4	/* commit to stable storage */
 #define FLUSH_LOWPRI		8	/* low priority background flush */
 #define FLUSH_HIGHPRI		16	/* high priority memory reclaim flush */
@@ -78,6 +76,7 @@ struct nfs_access_entry {
 struct nfs4_state;
 struct nfs_open_context {
 	atomic_t count;
+	struct vfsmount *vfsmnt;
 	struct dentry *dentry;
 	struct rpc_cred *cred;
 	struct nfs4_state *state;
@@ -118,8 +117,7 @@ struct nfs_inode {
 	unsigned long		cache_validity;		/* bit mask */
 
 	/*
-	 * read_cache_jiffies is when we started read-caching this inode,
-	 * and read_cache_mtime is the mtime of the inode at that time.
+	 * read_cache_jiffies is when we started read-caching this inode.
 	 * attrtimeo is for how long the cached information is assumed
 	 * to be valid. A successful attribute revalidation doubles
 	 * attrtimeo (up to acregmax/acdirmax), a failure resets it to
@@ -128,11 +126,6 @@ struct nfs_inode {
 	 * We need to revalidate the cached attrs for this inode if
 	 *
 	 *	jiffies - read_cache_jiffies > attrtimeo
-	 *
-	 * and invalidate any cached data/flush out any dirty pages if
-	 * we find that
-	 *
-	 *	mtime != read_cache_mtime
 	 */
 	unsigned long		read_cache_jiffies;
 	unsigned long		attrtimeo;
@@ -311,12 +304,9 @@ extern void nfs_begin_attr_update(struct inode *);
 extern void nfs_end_attr_update(struct inode *);
 extern void nfs_begin_data_update(struct inode *);
 extern void nfs_end_data_update(struct inode *);
-extern struct nfs_open_context *alloc_nfs_open_context(struct dentry *dentry, struct rpc_cred *cred);
 extern struct nfs_open_context *get_nfs_open_context(struct nfs_open_context *ctx);
 extern void put_nfs_open_context(struct nfs_open_context *ctx);
-extern void nfs_file_set_open_context(struct file *filp, struct nfs_open_context *ctx);
 extern struct nfs_open_context *nfs_find_open_context(struct inode *inode, struct rpc_cred *cred, int mode);
-extern void nfs_file_clear_open_context(struct file *filp);
 
 /* linux/net/ipv4/ipconfig.c: trims ip addr off front of name, too. */
 extern u32 root_nfs_parse_addr(char *name); /*__init*/
@@ -415,21 +405,22 @@ extern int  nfs_writepage(struct page *page, struct writeback_control *wbc);
 extern int  nfs_writepages(struct address_space *, struct writeback_control *);
 extern int  nfs_flush_incompatible(struct file *file, struct page *page);
 extern int  nfs_updatepage(struct file *, struct page *, unsigned int, unsigned int);
-extern void nfs_writeback_done(struct rpc_task *task, void *data);
-extern void nfs_writedata_release(void *data);
+extern int nfs_writeback_done(struct rpc_task *, struct nfs_write_data *);
+extern void nfs_writedata_release(void *);
 
 #if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
-extern void nfs_commit_done(struct rpc_task *, void *data);
-extern void nfs_commit_release(void *data);
+struct nfs_write_data *nfs_commit_alloc(unsigned int pagecount);
+void nfs_commit_free(struct nfs_write_data *p);
 #endif
 
 /*
  * Try to write back everything synchronously (but check the
  * return value!)
  */
-extern int  nfs_sync_inode(struct inode *, unsigned long, unsigned int, int);
+extern int  nfs_sync_inode_wait(struct inode *, unsigned long, unsigned int, int);
 #if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
 extern int  nfs_commit_inode(struct inode *, int);
+extern void nfs_commit_release(void *wdata);
 #else
 static inline int
 nfs_commit_inode(struct inode *inode, int how)
@@ -447,7 +438,7 @@ nfs_have_writebacks(struct inode *inode)
 static inline int
 nfs_wb_all(struct inode *inode)
 {
-	int error = nfs_sync_inode(inode, 0, 0, FLUSH_WAIT);
+	int error = nfs_sync_inode_wait(inode, 0, 0, 0);
 	return (error < 0) ? error : 0;
 }
 
@@ -456,8 +447,8 @@ nfs_wb_all(struct inode *inode)
  */
 static inline int nfs_wb_page_priority(struct inode *inode, struct page* page, int how)
 {
-	int error = nfs_sync_inode(inode, page->index, 1,
-			how | FLUSH_WAIT | FLUSH_STABLE);
+	int error = nfs_sync_inode_wait(inode, page->index, 1,
+			how | FLUSH_STABLE);
 	return (error < 0) ? error : 0;
 }
 
@@ -469,37 +460,8 @@ static inline int nfs_wb_page(struct inode *inode, struct page* page)
 /*
  * Allocate and free nfs_write_data structures
  */
-extern mempool_t *nfs_wdata_mempool;
-
-static inline struct nfs_write_data *nfs_writedata_alloc(unsigned int pagecount)
-{
-	struct nfs_write_data *p = mempool_alloc(nfs_wdata_mempool, SLAB_NOFS);
-
-	if (p) {
-		memset(p, 0, sizeof(*p));
-		INIT_LIST_HEAD(&p->pages);
-		if (pagecount < NFS_PAGEVEC_SIZE)
-			p->pagevec = &p->page_array[0];
-		else {
-			size_t size = ++pagecount * sizeof(struct page *);
-			p->pagevec = kmalloc(size, GFP_NOFS);
-			if (p->pagevec) {
-				memset(p->pagevec, 0, size);
-			} else {
-				mempool_free(p, nfs_wdata_mempool);
-				p = NULL;
-			}
-		}
-	}
-	return p;
-}
-
-static inline void nfs_writedata_free(struct nfs_write_data *p)
-{
-	if (p && (p->pagevec != &p->page_array[0]))
-		kfree(p->pagevec);
-	mempool_free(p, nfs_wdata_mempool);
-}
+extern struct nfs_write_data *nfs_writedata_alloc(unsigned int pagecount);
+extern void nfs_writedata_free(struct nfs_write_data *p);
 
 /*
  * linux/fs/nfs/read.c
@@ -507,44 +469,14 @@ static inline void nfs_writedata_free(struct nfs_write_data *p)
 extern int  nfs_readpage(struct file *, struct page *);
 extern int  nfs_readpages(struct file *, struct address_space *,
 		struct list_head *, unsigned);
-extern void nfs_readpage_result(struct rpc_task *, void *);
-extern void  nfs_readdata_release(void *data);
-
+extern int  nfs_readpage_result(struct rpc_task *, struct nfs_read_data *);
+extern void nfs_readdata_release(void *data);
 
 /*
  * Allocate and free nfs_read_data structures
  */
-extern mempool_t *nfs_rdata_mempool;
-
-static inline struct nfs_read_data *nfs_readdata_alloc(unsigned int pagecount)
-{
-	struct nfs_read_data *p = mempool_alloc(nfs_rdata_mempool, SLAB_NOFS);
-
-	if (p) {
-		memset(p, 0, sizeof(*p));
-		INIT_LIST_HEAD(&p->pages);
-		if (pagecount < NFS_PAGEVEC_SIZE)
-			p->pagevec = &p->page_array[0];
-		else {
-			size_t size = ++pagecount * sizeof(struct page *);
-			p->pagevec = kmalloc(size, GFP_NOFS);
-			if (p->pagevec) {
-				memset(p->pagevec, 0, size);
-			} else {
-				mempool_free(p, nfs_rdata_mempool);
-				p = NULL;
-			}
-		}
-	}
-	return p;
-}
-
-static inline void nfs_readdata_free(struct nfs_read_data *p)
-{
-	if (p && (p->pagevec != &p->page_array[0]))
-		kfree(p->pagevec);
-	mempool_free(p, nfs_rdata_mempool);
-}
+extern struct nfs_read_data *nfs_readdata_alloc(unsigned int pagecount);
+extern void nfs_readdata_free(struct nfs_read_data *p);
 
 /*
  * linux/fs/nfs3proc.c

include/linux/nfs_fs_i.h

@@ -12,8 +12,8 @@ struct nlm_lockowner;
  */
 struct nfs_lock_info {
 	u32		state;
-	u32		flags;
 	struct nlm_lockowner *owner;
+	struct list_head list;
 };
 
 struct nfs4_lock_state;
@@ -21,10 +21,4 @@ struct nfs4_lock_info {
 	struct nfs4_lock_state *owner;
 };
 
-/*
- * Lock flag values
- */
-#define NFS_LCK_GRANTED		0x0001		/* lock has been granted */
-#define NFS_LCK_RECLAIM		0x0002		/* lock marked for reclaiming */
-
 #endif

include/linux/nfs_fs_sb.h

@@ -4,6 +4,8 @@
 #include <linux/list.h>
 #include <linux/backing-dev.h>
 
+struct nfs_iostats;
+
 /*
  * NFS client parameters stored in the superblock.
  */
@@ -12,6 +14,7 @@ struct nfs_server {
 	struct rpc_clnt *	client_sys;	/* 2nd handle for FSINFO */
 	struct rpc_clnt *	client_acl;	/* ACL RPC client handle */
 	struct nfs_rpc_ops *	rpc_ops;	/* NFS protocol vector */
+	struct nfs_iostats *	io_stats;	/* I/O statistics */
 	struct backing_dev_info	backing_dev_info;
 	int			flags;		/* various flags */
 	unsigned int		caps;		/* server capabilities */
@@ -26,10 +29,13 @@ struct nfs_server {
 	unsigned int		acregmax;
 	unsigned int		acdirmin;
 	unsigned int		acdirmax;
+	unsigned long		retrans_timeo;	/* retransmit timeout */
+	unsigned int		retrans_count;	/* number of retransmit tries */
 	unsigned int		namelen;
 	char *			hostname;	/* remote hostname */
 	struct nfs_fh		fh;
 	struct sockaddr_in	addr;
+	unsigned long		mount_time;	/* when this fs was mounted */
 #ifdef CONFIG_NFS_V4
 	/* Our own IP address, as a null-terminated string.
 	 * This is used to generate the clientid, and the callback address.

include/linux/nfs_xdr.h

@@ -695,7 +695,6 @@ struct nfs_read_data {
 #ifdef CONFIG_NFS_V4
 	unsigned long		timestamp;	/* For lease renewal */
 #endif
-	void (*complete) (struct nfs_read_data *, int);
 	struct page		*page_array[NFS_PAGEVEC_SIZE + 1];
 };
 
@@ -714,7 +713,6 @@ struct nfs_write_data {
 #ifdef CONFIG_NFS_V4
 	unsigned long		timestamp;	/* For lease renewal */
 #endif
-	void (*complete) (struct nfs_write_data *, int);
 	struct page		*page_array[NFS_PAGEVEC_SIZE + 1];
 };
 
@@ -769,8 +767,11 @@ struct nfs_rpc_ops {
 			     struct nfs_pathconf *);
 	u32 *	(*decode_dirent)(u32 *, struct nfs_entry *, int plus);
 	void	(*read_setup)   (struct nfs_read_data *);
+	int	(*read_done)  (struct rpc_task *, struct nfs_read_data *);
 	void	(*write_setup)  (struct nfs_write_data *, int how);
+	int	(*write_done)  (struct rpc_task *, struct nfs_write_data *);
 	void	(*commit_setup) (struct nfs_write_data *, int how);
+	int	(*commit_done) (struct rpc_task *, struct nfs_write_data *);
 	int	(*file_open)   (struct inode *, struct file *);
 	int	(*file_release) (struct inode *, struct file *);
 	int	(*lock)(struct file *, int, struct file_lock *);

include/linux/sunrpc/clnt.h

@@ -45,7 +45,8 @@ struct rpc_clnt {
 	char *			cl_server;	/* server machine name */
 	char *			cl_protname;	/* protocol name */
 	struct rpc_auth *	cl_auth;	/* authenticator */
-	struct rpc_stat *	cl_stats;	/* statistics */
+	struct rpc_stat *	cl_stats;	/* per-program statistics */
+	struct rpc_iostats *	cl_metrics;	/* per-client statistics */
 
 	unsigned int		cl_softrtry : 1,/* soft timeouts */
 				cl_intr     : 1,/* interruptible */
@@ -59,6 +60,7 @@ struct rpc_clnt {
 	int			cl_nodelen;	/* nodename length */
 	char 			cl_nodename[UNX_MAXNODENAME];
 	char			cl_pathname[30];/* Path in rpc_pipe_fs */
+	struct vfsmount *	cl_vfsmnt;
 	struct dentry *		cl_dentry;	/* inode */
 	struct rpc_clnt *	cl_parent;	/* Points to parent of clones */
 	struct rpc_rtt		cl_rtt_default;
@@ -100,6 +102,8 @@ struct rpc_procinfo {
 	unsigned int		p_bufsiz;	/* req. buffer size */
 	unsigned int		p_count;	/* call count */
 	unsigned int		p_timer;	/* Which RTT timer to use */
+	u32			p_statidx;	/* Which procedure to account */
+	char *			p_name;		/* name of procedure */
 };
 
 #define RPC_CONGESTED(clnt)	(RPCXPRT_CONGESTED((clnt)->cl_xprt))
@@ -137,20 +141,6 @@ size_t		rpc_max_payload(struct rpc_clnt *);
 void		rpc_force_rebind(struct rpc_clnt *);
 int		rpc_ping(struct rpc_clnt *clnt, int flags);
 
-static __inline__
-int rpc_call(struct rpc_clnt *clnt, u32 proc, void *argp, void *resp, int flags)
-{
-	struct rpc_message msg = {
-		.rpc_proc	= &clnt->cl_procinfo[proc],
-		.rpc_argp	= argp,
-		.rpc_resp	= resp,
-		.rpc_cred	= NULL
-	};
-	return rpc_call_sync(clnt, &msg, flags);
-}
-		
-extern void rpciod_wake_up(void);
-
 /*
  * Helper function for NFSroot support
  */

include/linux/sunrpc/gss_krb5.h

@@ -53,6 +53,8 @@ struct krb5_ctx {
 	struct xdr_netobj	mech_used;
 };
 
+extern spinlock_t krb5_seq_lock;
+
 #define KG_TOK_MIC_MSG    0x0101
 #define KG_TOK_WRAP_MSG   0x0201
 

include/linux/sunrpc/metrics.h

@@ -0,0 +1,77 @@
+/*
+ *  linux/include/linux/sunrpc/metrics.h
+ *
+ *  Declarations for RPC client per-operation metrics
+ *
+ *  Copyright (C) 2005	Chuck Lever <cel@netapp.com>
+ *
+ *  RPC client per-operation statistics provide latency and retry
+ *  information about each type of RPC procedure in a given RPC program.
+ *  These statistics are not for detailed problem diagnosis, but simply
+ *  to indicate whether the problem is local or remote.
+ *
+ *  These counters are not meant to be human-readable, but are meant to be
+ *  integrated into system monitoring tools such as "sar" and "iostat".  As
+ *  such, the counters are sampled by the tools over time, and are never
+ *  zeroed after a file system is mounted.  Moving averages can be computed
+ *  by the tools by taking the difference between two instantaneous samples
+ *  and dividing that by the time between the samples.
+ *
+ *  The counters are maintained in a single array per RPC client, indexed
+ *  by procedure number.  There is no need to maintain separate counter
+ *  arrays per-CPU because these counters are always modified behind locks.
+ */
+
+#ifndef _LINUX_SUNRPC_METRICS_H
+#define _LINUX_SUNRPC_METRICS_H
+
+#include <linux/seq_file.h>
+
+#define RPC_IOSTATS_VERS	"1.0"
+
+struct rpc_iostats {
+	/*
+	 * These counters give an idea about how many request
+	 * transmissions are required, on average, to complete that
+	 * particular procedure.  Some procedures may require more
+	 * than one transmission because the server is unresponsive,
+	 * the client is retransmitting too aggressively, or the
+	 * requests are large and the network is congested.
+	 */
+	unsigned long		om_ops,		/* count of operations */
+				om_ntrans,	/* count of RPC transmissions */
+				om_timeouts;	/* count of major timeouts */
+
+	/*
+	 * These count how many bytes are sent and received for a
+	 * given RPC procedure type.  This indicates how much load a
+	 * particular procedure is putting on the network.  These
+	 * counts include the RPC and ULP headers, and the request
+	 * payload.
+	 */
+	unsigned long long      om_bytes_sent,	/* count of bytes out */
+				om_bytes_recv;	/* count of bytes in */
+
+	/*
+	 * The length of time an RPC request waits in queue before
+	 * transmission, the network + server latency of the request,
+	 * and the total time the request spent from init to release
+	 * are measured.
+	 */
+	unsigned long long	om_queue,	/* jiffies queued for xmit */
+				om_rtt,		/* jiffies for RPC RTT */
+				om_execute;	/* jiffies for RPC execution */
+} ____cacheline_aligned;
+
+struct rpc_task;
+struct rpc_clnt;
+
+/*
+ * EXPORTed functions for managing rpc_iostats structures
+ */
+struct rpc_iostats *	rpc_alloc_iostats(struct rpc_clnt *);
+void			rpc_count_iostats(struct rpc_task *);
+void			rpc_print_iostats(struct seq_file *, struct rpc_clnt *);
+void			rpc_free_iostats(struct rpc_iostats *);
+
+#endif /* _LINUX_SUNRPC_METRICS_H */

include/linux/sunrpc/rpc_pipe_fs.h

@@ -45,6 +45,8 @@ extern struct dentry *rpc_mkdir(char *, struct rpc_clnt *);
 extern int rpc_rmdir(char *);
 extern struct dentry *rpc_mkpipe(char *, void *, struct rpc_pipe_ops *, int flags);
 extern int rpc_unlink(char *);
+extern struct vfsmount *rpc_get_mount(void);
+extern void rpc_put_mount(void);
 
 #endif
 #endif

include/linux/sunrpc/sched.h

@@ -86,6 +86,12 @@ struct rpc_task {
 		struct work_struct	tk_work;	/* Async task work queue */
 		struct rpc_wait		tk_wait;	/* RPC wait */
 	} u;
+
+	unsigned short		tk_timeouts;	/* maj timeouts */
+	size_t			tk_bytes_sent;	/* total bytes sent */
+	unsigned long		tk_start;	/* RPC task init timestamp */
+	long			tk_rtt;		/* round-trip time (jiffies) */
+
 #ifdef RPC_DEBUG
 	unsigned short		tk_pid;		/* debugging aid */
 #endif
@@ -203,6 +209,7 @@ struct rpc_wait_queue {
 	unsigned char		priority;		/* current priority */
 	unsigned char		count;			/* # task groups remaining serviced so far */
 	unsigned char		nr;			/* # tasks remaining for cookie */
+	unsigned short		qlen;			/* total # tasks waiting in queue */
 #ifdef RPC_DEBUG
 	const char *		name;
 #endif
@@ -269,13 +276,13 @@ void *		rpc_malloc(struct rpc_task *, size_t);
 void		rpc_free(struct rpc_task *);
 int		rpciod_up(void);
 void		rpciod_down(void);
-void		rpciod_wake_up(void);
 int		__rpc_wait_for_completion_task(struct rpc_task *task, int (*)(void *));
 #ifdef RPC_DEBUG
 void		rpc_show_tasks(void);
 #endif
 int		rpc_init_mempool(void);
 void		rpc_destroy_mempool(void);
+extern struct workqueue_struct *rpciod_workqueue;
 
 static inline void rpc_exit(struct rpc_task *task, int status)
 {

include/linux/sunrpc/xprt.h

@@ -114,6 +114,7 @@ struct rpc_xprt_ops {
 	void		(*release_request)(struct rpc_task *task);
 	void		(*close)(struct rpc_xprt *xprt);
 	void		(*destroy)(struct rpc_xprt *xprt);
+	void		(*print_stats)(struct rpc_xprt *xprt, struct seq_file *seq);
 };
 
 struct rpc_xprt {
@@ -187,6 +188,18 @@ struct rpc_xprt {
 
 	struct list_head	recv;
 
+	struct {
+		unsigned long		bind_count,	/* total number of binds */
+					connect_count,	/* total number of connects */
+					connect_start,	/* connect start timestamp */
+					connect_time,	/* jiffies waiting for connect */
+					sends,		/* how many complete requests */
+					recvs,		/* how many complete requests */
+					bad_xids;	/* lookup_rqst didn't find XID */
+
+		unsigned long long	req_u,		/* average requests on the wire */
+					bklog_u;	/* backlog queue utilization */
+	} stat;
 
 	void			(*old_data_ready)(struct sock *, int);
 	void			(*old_state_change)(struct sock *);

net/sunrpc/auth.c

@@ -64,14 +64,26 @@ rpcauth_create(rpc_authflavor_t pseudoflavor, struct rpc_clnt *clnt)
 	struct rpc_authops	*ops;
 	u32			flavor = pseudoflavor_to_flavor(pseudoflavor);
 
-	if (flavor >= RPC_AUTH_MAXFLAVOR || !(ops = auth_flavors[flavor]))
-		return ERR_PTR(-EINVAL);
+	auth = ERR_PTR(-EINVAL);
+	if (flavor >= RPC_AUTH_MAXFLAVOR)
+		goto out;
+
+	/* FIXME - auth_flavors[] really needs an rw lock,
+	 * and module refcounting. */
+#ifdef CONFIG_KMOD
+	if ((ops = auth_flavors[flavor]) == NULL)
+		request_module("rpc-auth-%u", flavor);
+#endif
+	if ((ops = auth_flavors[flavor]) == NULL)
+		goto out;
 	auth = ops->create(clnt, pseudoflavor);
 	if (IS_ERR(auth))
 		return auth;
 	if (clnt->cl_auth)
 		rpcauth_destroy(clnt->cl_auth);
 	clnt->cl_auth = auth;
+
+out:
 	return auth;
 }
 

net/sunrpc/auth_gss/auth_gss.c

@@ -721,6 +721,8 @@ gss_destroy(struct rpc_auth *auth)
 
 	gss_auth = container_of(auth, struct gss_auth, rpc_auth);
 	rpc_unlink(gss_auth->path);
+	dput(gss_auth->dentry);
+	gss_auth->dentry = NULL;
 	gss_mech_put(gss_auth->mech);
 
 	rpcauth_free_credcache(auth);

net/sunrpc/auth_gss/gss_krb5_seal.c

@@ -70,15 +70,19 @@
 # define RPCDBG_FACILITY        RPCDBG_AUTH
 #endif
 
+spinlock_t krb5_seq_lock = SPIN_LOCK_UNLOCKED;
+
 u32
 gss_get_mic_kerberos(struct gss_ctx *gss_ctx, struct xdr_buf *text,
 		struct xdr_netobj *token)
 {
 	struct krb5_ctx		*ctx = gss_ctx->internal_ctx_id;
 	s32			checksum_type;
-	struct xdr_netobj	md5cksum = {.len = 0, .data = NULL};
+	char			cksumdata[16];
+	struct xdr_netobj	md5cksum = {.len = 0, .data = cksumdata};
 	unsigned char		*ptr, *krb5_hdr, *msg_start;
 	s32			now;
+	u32			seq_send;
 
 	dprintk("RPC:     gss_krb5_seal\n");
 
@@ -133,16 +137,15 @@ gss_get_mic_kerberos(struct gss_ctx *gss_ctx, struct xdr_buf *text,
 		BUG();
 	}
 
-	kfree(md5cksum.data);
+	spin_lock(&krb5_seq_lock);
+	seq_send = ctx->seq_send++;
+	spin_unlock(&krb5_seq_lock);
 
 	if ((krb5_make_seq_num(ctx->seq, ctx->initiate ? 0 : 0xff,
-			       ctx->seq_send, krb5_hdr + 16, krb5_hdr + 8)))
+			       seq_send, krb5_hdr + 16, krb5_hdr + 8)))
 		goto out_err;
 
-	ctx->seq_send++;
-
 	return ((ctx->endtime < now) ? GSS_S_CONTEXT_EXPIRED : GSS_S_COMPLETE);
 out_err:
-	kfree(md5cksum.data);
 	return GSS_S_FAILURE;
 }

net/sunrpc/auth_gss/gss_krb5_unseal.c

@@ -79,7 +79,8 @@ gss_verify_mic_kerberos(struct gss_ctx *gss_ctx,
 	int			signalg;
 	int			sealalg;
 	s32			checksum_type;
-	struct xdr_netobj	md5cksum = {.len = 0, .data = NULL};
+	char			cksumdata[16];
+	struct xdr_netobj	md5cksum = {.len = 0, .data = cksumdata};
 	s32			now;
 	int			direction;
 	s32			seqnum;
@@ -176,6 +177,5 @@ gss_verify_mic_kerberos(struct gss_ctx *gss_ctx,
 
 	ret = GSS_S_COMPLETE;
 out:
-	kfree(md5cksum.data);
 	return ret;
 }

net/sunrpc/auth_gss/gss_krb5_wrap.c

@@ -121,12 +121,14 @@ gss_wrap_kerberos(struct gss_ctx *ctx, int offset,
 {
 	struct krb5_ctx		*kctx = ctx->internal_ctx_id;
 	s32			checksum_type;
-	struct xdr_netobj	md5cksum = {.len = 0, .data = NULL};
+	char			cksumdata[16];
+	struct xdr_netobj	md5cksum = {.len = 0, .data = cksumdata};
 	int			blocksize = 0, plainlen;
 	unsigned char		*ptr, *krb5_hdr, *msg_start;
 	s32			now;
 	int			headlen;
 	struct page		**tmp_pages;
+	u32			seq_send;
 
 	dprintk("RPC:     gss_wrap_kerberos\n");
 
@@ -205,23 +207,22 @@ gss_wrap_kerberos(struct gss_ctx *ctx, int offset,
 		BUG();
 	}
 
-	kfree(md5cksum.data);
+	spin_lock(&krb5_seq_lock);
+	seq_send = kctx->seq_send++;
+	spin_unlock(&krb5_seq_lock);
 
 	/* XXX would probably be more efficient to compute checksum
 	 * and encrypt at the same time: */
 	if ((krb5_make_seq_num(kctx->seq, kctx->initiate ? 0 : 0xff,
-			       kctx->seq_send, krb5_hdr + 16, krb5_hdr + 8)))
+			       seq_send, krb5_hdr + 16, krb5_hdr + 8)))
 		goto out_err;
 
 	if (gss_encrypt_xdr_buf(kctx->enc, buf, offset + headlen - blocksize,
 									pages))
 		goto out_err;
 
-	kctx->seq_send++;
-
 	return ((kctx->endtime < now) ? GSS_S_CONTEXT_EXPIRED : GSS_S_COMPLETE);
 out_err:
-	if (md5cksum.data) kfree(md5cksum.data);
 	return GSS_S_FAILURE;
 }
 
@@ -232,7 +233,8 @@ gss_unwrap_kerberos(struct gss_ctx *ctx, int offset, struct xdr_buf *buf)
 	int			signalg;
 	int			sealalg;
 	s32			checksum_type;
-	struct xdr_netobj	md5cksum = {.len = 0, .data = NULL};
+	char			cksumdata[16];
+	struct xdr_netobj	md5cksum = {.len = 0, .data = cksumdata};
 	s32			now;
 	int			direction;
 	s32			seqnum;
@@ -358,6 +360,5 @@ gss_unwrap_kerberos(struct gss_ctx *ctx, int offset, struct xdr_buf *buf)
 
 	ret = GSS_S_COMPLETE;
 out:
-	if (md5cksum.data) kfree(md5cksum.data);
 	return ret;
 }

net/sunrpc/auth_gss/gss_spkm3_mech.c

@@ -102,6 +102,12 @@ get_key(const void *p, const void *end, struct crypto_tfm **res, int *resalg)
 			alg_mode = CRYPTO_TFM_MODE_CBC;
 			setkey = 1;
 			break;
+		case NID_cast5_cbc:
+			/* XXXX here in name only, not used */
+			alg_name = "cast5";
+			alg_mode = CRYPTO_TFM_MODE_CBC;
+			setkey = 0; /* XXX will need to set to 1 */
+			break;
 		case NID_md5:
 			if (key.len == 0) {
 				dprintk("RPC: SPKM3 get_key: NID_md5 zero Key length\n");

net/sunrpc/auth_gss/gss_spkm3_seal.c

@@ -57,7 +57,8 @@ spkm3_make_token(struct spkm3_ctx *ctx,
 {
 	s32			checksum_type;
 	char			tokhdrbuf[25];
-	struct xdr_netobj	md5cksum = {.len = 0, .data = NULL};
+	char			cksumdata[16];
+	struct xdr_netobj	md5cksum = {.len = 0, .data = cksumdata};
 	struct xdr_netobj	mic_hdr = {.len = 0, .data = tokhdrbuf};
 	int			tokenlen = 0;
 	unsigned char		*ptr;
@@ -115,13 +116,11 @@ spkm3_make_token(struct spkm3_ctx *ctx,
 		dprintk("RPC: gss_spkm3_seal: SPKM_WRAP_TOK not supported\n");
 		goto out_err;
 	}
-	kfree(md5cksum.data);
 
 	/* XXX need to implement sequence numbers, and ctx->expired */
 
 	return  GSS_S_COMPLETE;
 out_err:
-	kfree(md5cksum.data);
 	token->data = NULL;
 	token->len = 0;
 	return GSS_S_FAILURE;

net/sunrpc/auth_gss/gss_spkm3_unseal.c

@@ -56,7 +56,8 @@ spkm3_read_token(struct spkm3_ctx *ctx,
 {
 	s32			code;
 	struct xdr_netobj	wire_cksum = {.len =0, .data = NULL};
-	struct xdr_netobj	md5cksum = {.len = 0, .data = NULL};
+	char			cksumdata[16];
+	struct xdr_netobj	md5cksum = {.len = 0, .data = cksumdata};
 	unsigned char		*ptr = (unsigned char *)read_token->data;
 	unsigned char           *cksum;
 	int			bodysize, md5elen;
@@ -120,7 +121,6 @@ spkm3_read_token(struct spkm3_ctx *ctx,
 	/* XXX: need to add expiration and sequencing */
 	ret = GSS_S_COMPLETE;
 out:
-	kfree(md5cksum.data);
 	kfree(wire_cksum.data);
 	return ret;
 }

net/sunrpc/clnt.c

@@ -28,12 +28,11 @@
 #include <linux/mm.h>
 #include <linux/slab.h>
 #include <linux/utsname.h>
+#include <linux/workqueue.h>
 
 #include <linux/sunrpc/clnt.h>
-#include <linux/workqueue.h>
 #include <linux/sunrpc/rpc_pipe_fs.h>
-
-#include <linux/nfs.h>
+#include <linux/sunrpc/metrics.h>
 
 
 #define RPC_SLACK_SPACE		(1024)	/* total overkill */
@@ -71,8 +70,15 @@ rpc_setup_pipedir(struct rpc_clnt *clnt, char *dir_name)
 	static uint32_t clntid;
 	int error;
 
+	clnt->cl_vfsmnt = ERR_PTR(-ENOENT);
+	clnt->cl_dentry = ERR_PTR(-ENOENT);
 	if (dir_name == NULL)
 		return 0;
+
+	clnt->cl_vfsmnt = rpc_get_mount();
+	if (IS_ERR(clnt->cl_vfsmnt))
+		return PTR_ERR(clnt->cl_vfsmnt);
+
 	for (;;) {
 		snprintf(clnt->cl_pathname, sizeof(clnt->cl_pathname),
 				"%s/clnt%x", dir_name,
@@ -85,6 +91,7 @@ rpc_setup_pipedir(struct rpc_clnt *clnt, char *dir_name)
 		if (error != -EEXIST) {
 			printk(KERN_INFO "RPC: Couldn't create pipefs entry %s, error %d\n",
 					clnt->cl_pathname, error);
+			rpc_put_mount();
 			return error;
 		}
 	}
@@ -147,6 +154,7 @@ rpc_new_client(struct rpc_xprt *xprt, char *servname,
 	clnt->cl_vers     = version->number;
 	clnt->cl_prot     = xprt->prot;
 	clnt->cl_stats    = program->stats;
+	clnt->cl_metrics  = rpc_alloc_iostats(clnt);
 	rpc_init_wait_queue(&clnt->cl_pmap_default.pm_bindwait, "bindwait");
 
 	if (!clnt->cl_port)
@@ -175,7 +183,11 @@ rpc_new_client(struct rpc_xprt *xprt, char *servname,
 	return clnt;
 
 out_no_auth:
-	rpc_rmdir(clnt->cl_pathname);
+	if (!IS_ERR(clnt->cl_dentry)) {
+		rpc_rmdir(clnt->cl_pathname);
+		dput(clnt->cl_dentry);
+		rpc_put_mount();
+	}
 out_no_path:
 	if (clnt->cl_server != clnt->cl_inline_name)
 		kfree(clnt->cl_server);
@@ -240,11 +252,15 @@ rpc_clone_client(struct rpc_clnt *clnt)
 	new->cl_autobind = 0;
 	new->cl_oneshot = 0;
 	new->cl_dead = 0;
+	if (!IS_ERR(new->cl_dentry)) {
+		dget(new->cl_dentry);
+		rpc_get_mount();
+	}
 	rpc_init_rtt(&new->cl_rtt_default, clnt->cl_xprt->timeout.to_initval);
 	if (new->cl_auth)
 		atomic_inc(&new->cl_auth->au_count);
 	new->cl_pmap		= &new->cl_pmap_default;
-	rpc_init_wait_queue(&new->cl_pmap_default.pm_bindwait, "bindwait");
+	new->cl_metrics         = rpc_alloc_iostats(clnt);
 	return new;
 out_no_clnt:
 	printk(KERN_INFO "RPC: out of memory in %s\n", __FUNCTION__);
@@ -314,6 +330,12 @@ rpc_destroy_client(struct rpc_clnt *clnt)
 	if (clnt->cl_server != clnt->cl_inline_name)
 		kfree(clnt->cl_server);
 out_free:
+	rpc_free_iostats(clnt->cl_metrics);
+	clnt->cl_metrics = NULL;
+	if (!IS_ERR(clnt->cl_dentry)) {
+		dput(clnt->cl_dentry);
+		rpc_put_mount();
+	}
 	kfree(clnt);
 	return 0;
 }
@@ -473,15 +495,16 @@ rpc_call_async(struct rpc_clnt *clnt, struct rpc_message *msg, int flags,
 	int		status;
 
 	/* If this client is slain all further I/O fails */
+	status = -EIO;
 	if (clnt->cl_dead) 
-		return -EIO;
+		goto out_release;
 
 	flags |= RPC_TASK_ASYNC;
 
 	/* Create/initialize a new RPC task */
 	status = -ENOMEM;
 	if (!(task = rpc_new_task(clnt, flags, tk_ops, data)))
-		goto out;
+		goto out_release;
 
 	/* Mask signals on GSS_AUTH upcalls */
 	rpc_task_sigmask(task, &oldset);		
@@ -496,7 +519,10 @@ rpc_call_async(struct rpc_clnt *clnt, struct rpc_message *msg, int flags,
 		rpc_release_task(task);
 
 	rpc_restore_sigmask(&oldset);		
-out:
+	return status;
+out_release:
+	if (tk_ops->rpc_release != NULL)
+		tk_ops->rpc_release(data);
 	return status;
 }
 
@@ -993,6 +1019,8 @@ call_timeout(struct rpc_task *task)
 	}
 
 	dprintk("RPC: %4d call_timeout (major)\n", task->tk_pid);
+	task->tk_timeouts++;
+
 	if (RPC_IS_SOFT(task)) {
 		printk(KERN_NOTICE "%s: server %s not responding, timed out\n",
 				clnt->cl_protname, clnt->cl_server);
@@ -1045,6 +1073,11 @@ call_decode(struct rpc_task *task)
 		return;
 	}
 
+	/*
+	 * Ensure that we see all writes made by xprt_complete_rqst()
+	 * before it changed req->rq_received.
+	 */
+	smp_rmb();
 	req->rq_rcv_buf.len = req->rq_private_buf.len;
 
 	/* Check that the softirq receive buffer is valid */
@@ -1194,8 +1227,8 @@ call_verify(struct rpc_task *task)
 			task->tk_action = call_bind;
 			goto out_retry;
 		case RPC_AUTH_TOOWEAK:
-			printk(KERN_NOTICE "call_verify: server requires stronger "
-			       "authentication.\n");
+			printk(KERN_NOTICE "call_verify: server %s requires stronger "
+			       "authentication.\n", task->tk_client->cl_server);
 			break;
 		default:
 			printk(KERN_WARNING "call_verify: unknown auth error: %x\n", n);

net/sunrpc/pmap_clnt.c

@@ -82,6 +82,7 @@ rpc_getport(struct rpc_task *task, struct rpc_clnt *clnt)
 	rpc_call_setup(child, &msg, 0);
 
 	/* ... and run the child task */
+	task->tk_xprt->stat.bind_count++;
 	rpc_run_child(task, child, pmap_getport_done);
 	return;
 
@@ -103,6 +104,11 @@ rpc_getport_external(struct sockaddr_in *sin, __u32 prog, __u32 vers, int prot)
 		.pm_prot	= prot,
 		.pm_port	= 0
 	};
+	struct rpc_message msg = {
+		.rpc_proc	= &pmap_procedures[PMAP_GETPORT],
+		.rpc_argp	= &map,
+		.rpc_resp	= &map.pm_port,
+	};
 	struct rpc_clnt	*pmap_clnt;
 	char		hostname[32];
 	int		status;
@@ -116,7 +122,7 @@ rpc_getport_external(struct sockaddr_in *sin, __u32 prog, __u32 vers, int prot)
 		return PTR_ERR(pmap_clnt);
 
 	/* Setup the call info struct */
-	status = rpc_call(pmap_clnt, PMAP_GETPORT, &map, &map.pm_port, 0);
+	status = rpc_call_sync(pmap_clnt, &msg, 0);
 
 	if (status >= 0) {
 		if (map.pm_port != 0)
@@ -161,16 +167,27 @@ pmap_getport_done(struct rpc_task *task)
 int
 rpc_register(u32 prog, u32 vers, int prot, unsigned short port, int *okay)
 {
-	struct sockaddr_in	sin;
-	struct rpc_portmap	map;
+	struct sockaddr_in	sin = {
+		.sin_family	= AF_INET,
+		.sin_addr.s_addr = htonl(INADDR_LOOPBACK),
+	};
+	struct rpc_portmap	map = {
+		.pm_prog	= prog,
+		.pm_vers	= vers,
+		.pm_prot	= prot,
+		.pm_port	= port,
+	};
+	struct rpc_message msg = {
+		.rpc_proc	= &pmap_procedures[port ? PMAP_SET : PMAP_UNSET],
+		.rpc_argp	= &map,
+		.rpc_resp	= okay,
+	};
 	struct rpc_clnt		*pmap_clnt;
 	int error = 0;
 
 	dprintk("RPC: registering (%d, %d, %d, %d) with portmapper.\n",
 			prog, vers, prot, port);
 
-	sin.sin_family = AF_INET;
-	sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
 	pmap_clnt = pmap_create("localhost", &sin, IPPROTO_UDP, 1);
 	if (IS_ERR(pmap_clnt)) {
 		error = PTR_ERR(pmap_clnt);
@@ -178,13 +195,7 @@ rpc_register(u32 prog, u32 vers, int prot, unsigned short port, int *okay)
 		return error;
 	}
 
-	map.pm_prog = prog;
-	map.pm_vers = vers;
-	map.pm_prot = prot;
-	map.pm_port = port;
-
-	error = rpc_call(pmap_clnt, port? PMAP_SET : PMAP_UNSET,
-					&map, okay, 0);
+	error = rpc_call_sync(pmap_clnt, &msg, 0);
 
 	if (error < 0) {
 		printk(KERN_WARNING
@@ -260,6 +271,8 @@ static struct rpc_procinfo	pmap_procedures[] = {
 	  .p_decode		= (kxdrproc_t) xdr_decode_bool,
 	  .p_bufsiz		= 4,
 	  .p_count		= 1,
+	  .p_statidx		= PMAP_SET,
+	  .p_name		= "SET",
 	},
 [PMAP_UNSET] = {
 	  .p_proc		= PMAP_UNSET,
@@ -267,6 +280,8 @@ static struct rpc_procinfo	pmap_procedures[] = {
 	  .p_decode		= (kxdrproc_t) xdr_decode_bool,
 	  .p_bufsiz		= 4,
 	  .p_count		= 1,
+	  .p_statidx		= PMAP_UNSET,
+	  .p_name		= "UNSET",
 	},
 [PMAP_GETPORT] = {
 	  .p_proc		= PMAP_GETPORT,
@@ -274,6 +289,8 @@ static struct rpc_procinfo	pmap_procedures[] = {
 	  .p_decode		= (kxdrproc_t) xdr_decode_port,
 	  .p_bufsiz		= 4,
 	  .p_count		= 1,
+	  .p_statidx		= PMAP_GETPORT,
+	  .p_name		= "GETPORT",
 	},
 };
 

net/sunrpc/rpc_pipe.c

@@ -91,7 +91,8 @@ rpc_queue_upcall(struct inode *inode, struct rpc_pipe_msg *msg)
 		res = 0;
 	} else if (rpci->flags & RPC_PIPE_WAIT_FOR_OPEN) {
 		if (list_empty(&rpci->pipe))
-			schedule_delayed_work(&rpci->queue_timeout,
+			queue_delayed_work(rpciod_workqueue,
+					&rpci->queue_timeout,
 					RPC_UPCALL_TIMEOUT);
 		list_add_tail(&msg->list, &rpci->pipe);
 		rpci->pipelen += msg->len;
@@ -132,7 +133,7 @@ rpc_close_pipes(struct inode *inode)
 		if (ops->release_pipe)
 			ops->release_pipe(inode);
 		cancel_delayed_work(&rpci->queue_timeout);
-		flush_scheduled_work();
+		flush_workqueue(rpciod_workqueue);
 	}
 	rpc_inode_setowner(inode, NULL);
 	mutex_unlock(&inode->i_mutex);
@@ -434,14 +435,17 @@ static struct rpc_filelist authfiles[] = {
 	},
 };
 
-static int
-rpc_get_mount(void)
+struct vfsmount *rpc_get_mount(void)
 {
-	return simple_pin_fs("rpc_pipefs", &rpc_mount, &rpc_mount_count);
+	int err;
+
+	err = simple_pin_fs("rpc_pipefs", &rpc_mount, &rpc_mount_count);
+	if (err != 0)
+		return ERR_PTR(err);
+	return rpc_mount;
 }
 
-static void
-rpc_put_mount(void)
+void rpc_put_mount(void)
 {
 	simple_release_fs(&rpc_mount, &rpc_mount_count);
 }
@@ -451,12 +455,13 @@ rpc_lookup_parent(char *path, struct nameidata *nd)
 {
 	if (path[0] == '\0')
 		return -ENOENT;
-	if (rpc_get_mount()) {
+	nd->mnt = rpc_get_mount();
+	if (IS_ERR(nd->mnt)) {
 		printk(KERN_WARNING "%s: %s failed to mount "
 			       "pseudofilesystem \n", __FILE__, __FUNCTION__);
-		return -ENODEV;
+		return PTR_ERR(nd->mnt);
 	}
-	nd->mnt = mntget(rpc_mount);
+	mntget(nd->mnt);
 	nd->dentry = dget(rpc_mount->mnt_root);
 	nd->last_type = LAST_ROOT;
 	nd->flags = LOOKUP_PARENT;
@@ -593,7 +598,6 @@ __rpc_mkdir(struct inode *dir, struct dentry *dentry)
 	d_instantiate(dentry, inode);
 	dir->i_nlink++;
 	inode_dir_notify(dir, DN_CREATE);
-	rpc_get_mount();
 	return 0;
 out_err:
 	printk(KERN_WARNING "%s: %s failed to allocate inode for dentry %s\n",
@@ -614,7 +618,6 @@ __rpc_rmdir(struct inode *dir, struct dentry *dentry)
 	if (!error) {
 		inode_dir_notify(dir, DN_DELETE);
 		d_drop(dentry);
-		rpc_put_mount();
 	}
 	return 0;
 }
@@ -668,7 +671,7 @@ rpc_mkdir(char *path, struct rpc_clnt *rpc_client)
 out:
 	mutex_unlock(&dir->i_mutex);
 	rpc_release_path(&nd);
-	return dentry;
+	return dget(dentry);
 err_depopulate:
 	rpc_depopulate(dentry);
 	__rpc_rmdir(dir, dentry);
@@ -732,7 +735,7 @@ rpc_mkpipe(char *path, void *private, struct rpc_pipe_ops *ops, int flags)
 out:
 	mutex_unlock(&dir->i_mutex);
 	rpc_release_path(&nd);
-	return dentry;
+	return dget(dentry);
 err_dput:
 	dput(dentry);
 	dentry = ERR_PTR(-ENOMEM);

net/sunrpc/sched.c

@@ -65,7 +65,7 @@ static LIST_HEAD(all_tasks);
  */
 static DEFINE_MUTEX(rpciod_mutex);
 static unsigned int		rpciod_users;
-static struct workqueue_struct *rpciod_workqueue;
+struct workqueue_struct *rpciod_workqueue;
 
 /*
  * Spinlock for other critical sections of code.
@@ -182,6 +182,7 @@ static void __rpc_add_wait_queue(struct rpc_wait_queue *queue, struct rpc_task *
 	else
 		list_add_tail(&task->u.tk_wait.list, &queue->tasks[0]);
 	task->u.tk_wait.rpc_waitq = queue;
+	queue->qlen++;
 	rpc_set_queued(task);
 
 	dprintk("RPC: %4d added to queue %p \"%s\"\n",
@@ -216,6 +217,7 @@ static void __rpc_remove_wait_queue(struct rpc_task *task)
 		__rpc_remove_wait_queue_priority(task);
 	else
 		list_del(&task->u.tk_wait.list);
+	queue->qlen--;
 	dprintk("RPC: %4d removed from queue %p \"%s\"\n",
 				task->tk_pid, queue, rpc_qname(queue));
 }
@@ -816,6 +818,9 @@ void rpc_init_task(struct rpc_task *task, struct rpc_clnt *clnt, int flags, cons
 
 	BUG_ON(task->tk_ops == NULL);
 
+	/* starting timestamp */
+	task->tk_start = jiffies;
+
 	dprintk("RPC: %4d new task procpid %d\n", task->tk_pid,
 				current->pid);
 }
@@ -917,8 +922,11 @@ struct rpc_task *rpc_run_task(struct rpc_clnt *clnt, int flags,
 {
 	struct rpc_task *task;
 	task = rpc_new_task(clnt, flags, ops, data);
-	if (task == NULL)
+	if (task == NULL) {
+		if (ops->rpc_release != NULL)
+			ops->rpc_release(data);
 		return ERR_PTR(-ENOMEM);
+	}
 	atomic_inc(&task->tk_count);
 	rpc_execute(task);
 	return task;

net/sunrpc/stats.c

@@ -21,6 +21,7 @@
 #include <linux/seq_file.h>
 #include <linux/sunrpc/clnt.h>
 #include <linux/sunrpc/svcsock.h>
+#include <linux/sunrpc/metrics.h>
 
 #define RPCDBG_FACILITY	RPCDBG_MISC
 
@@ -106,6 +107,120 @@ void svc_seq_show(struct seq_file *seq, const struct svc_stat *statp) {
 	}
 }
 
+/**
+ * rpc_alloc_iostats - allocate an rpc_iostats structure
+ * @clnt: RPC program, version, and xprt
+ *
+ */
+struct rpc_iostats *rpc_alloc_iostats(struct rpc_clnt *clnt)
+{
+	unsigned int ops = clnt->cl_maxproc;
+	size_t size = ops * sizeof(struct rpc_iostats);
+	struct rpc_iostats *new;
+
+	new = kmalloc(size, GFP_KERNEL);
+	if (new)
+		memset(new, 0 , size);
+	return new;
+}
+EXPORT_SYMBOL(rpc_alloc_iostats);
+
+/**
+ * rpc_free_iostats - release an rpc_iostats structure
+ * @stats: doomed rpc_iostats structure
+ *
+ */
+void rpc_free_iostats(struct rpc_iostats *stats)
+{
+	kfree(stats);
+}
+EXPORT_SYMBOL(rpc_free_iostats);
+
+/**
+ * rpc_count_iostats - tally up per-task stats
+ * @task: completed rpc_task
+ *
+ * Relies on the caller for serialization.
+ */
+void rpc_count_iostats(struct rpc_task *task)
+{
+	struct rpc_rqst *req = task->tk_rqstp;
+	struct rpc_iostats *stats = task->tk_client->cl_metrics;
+	struct rpc_iostats *op_metrics;
+	long rtt, execute, queue;
+
+	if (!stats || !req)
+		return;
+	op_metrics = &stats[task->tk_msg.rpc_proc->p_statidx];
+
+	op_metrics->om_ops++;
+	op_metrics->om_ntrans += req->rq_ntrans;
+	op_metrics->om_timeouts += task->tk_timeouts;
+
+	op_metrics->om_bytes_sent += task->tk_bytes_sent;
+	op_metrics->om_bytes_recv += req->rq_received;
+
+	queue = (long)req->rq_xtime - task->tk_start;
+	if (queue < 0)
+		queue = -queue;
+	op_metrics->om_queue += queue;
+
+	rtt = task->tk_rtt;
+	if (rtt < 0)
+		rtt = -rtt;
+	op_metrics->om_rtt += rtt;
+
+	execute = (long)jiffies - task->tk_start;
+	if (execute < 0)
+		execute = -execute;
+	op_metrics->om_execute += execute;
+}
+
+void _print_name(struct seq_file *seq, unsigned int op, struct rpc_procinfo *procs)
+{
+	if (procs[op].p_name)
+		seq_printf(seq, "\t%12s: ", procs[op].p_name);
+	else if (op == 0)
+		seq_printf(seq, "\t        NULL: ");
+	else
+		seq_printf(seq, "\t%12u: ", op);
+}
+
+#define MILLISECS_PER_JIFFY	(1000 / HZ)
+
+void rpc_print_iostats(struct seq_file *seq, struct rpc_clnt *clnt)
+{
+	struct rpc_iostats *stats = clnt->cl_metrics;
+	struct rpc_xprt *xprt = clnt->cl_xprt;
+	unsigned int op, maxproc = clnt->cl_maxproc;
+
+	if (!stats)
+		return;
+
+	seq_printf(seq, "\tRPC iostats version: %s  ", RPC_IOSTATS_VERS);
+	seq_printf(seq, "p/v: %u/%u (%s)\n",
+			clnt->cl_prog, clnt->cl_vers, clnt->cl_protname);
+
+	if (xprt)
+		xprt->ops->print_stats(xprt, seq);
+
+	seq_printf(seq, "\tper-op statistics\n");
+	for (op = 0; op < maxproc; op++) {
+		struct rpc_iostats *metrics = &stats[op];
+		_print_name(seq, op, clnt->cl_procinfo);
+		seq_printf(seq, "%lu %lu %lu %Lu %Lu %Lu %Lu %Lu\n",
+				metrics->om_ops,
+				metrics->om_ntrans,
+				metrics->om_timeouts,
+				metrics->om_bytes_sent,
+				metrics->om_bytes_recv,
+				metrics->om_queue * MILLISECS_PER_JIFFY,
+				metrics->om_rtt * MILLISECS_PER_JIFFY,
+				metrics->om_execute * MILLISECS_PER_JIFFY);
+	}
+}
+EXPORT_SYMBOL(rpc_print_iostats);
+
 /*
  * Register/unregister RPC proc files
  */

net/sunrpc/xprt.c

@@ -44,13 +44,13 @@
 #include <linux/random.h>
 
 #include <linux/sunrpc/clnt.h>
+#include <linux/sunrpc/metrics.h>
 
 /*
  * Local variables
  */
 
 #ifdef RPC_DEBUG
-# undef  RPC_DEBUG_DATA
 # define RPCDBG_FACILITY	RPCDBG_XPRT
 #endif
 
@@ -548,6 +548,7 @@ void xprt_connect(struct rpc_task *task)
 
 		task->tk_timeout = xprt->connect_timeout;
 		rpc_sleep_on(&xprt->pending, task, xprt_connect_status, NULL);
+		xprt->stat.connect_start = jiffies;
 		xprt->ops->connect(task);
 	}
 	return;
@@ -558,6 +559,8 @@ static void xprt_connect_status(struct rpc_task *task)
 	struct rpc_xprt	*xprt = task->tk_xprt;
 
 	if (task->tk_status >= 0) {
+		xprt->stat.connect_count++;
+		xprt->stat.connect_time += (long)jiffies - xprt->stat.connect_start;
 		dprintk("RPC: %4d xprt_connect_status: connection established\n",
 				task->tk_pid);
 		return;
@@ -601,16 +604,14 @@ static void xprt_connect_status(struct rpc_task *task)
 struct rpc_rqst *xprt_lookup_rqst(struct rpc_xprt *xprt, u32 xid)
 {
 	struct list_head *pos;
-	struct rpc_rqst	*req = NULL;
 
 	list_for_each(pos, &xprt->recv) {
 		struct rpc_rqst *entry = list_entry(pos, struct rpc_rqst, rq_list);
-		if (entry->rq_xid == xid) {
-			req = entry;
-			break;
-		}
+		if (entry->rq_xid == xid)
+			return entry;
 	}
-	return req;
+	xprt->stat.bad_xids++;
+	return NULL;
 }
 
 /**
@@ -646,7 +647,12 @@ void xprt_complete_rqst(struct rpc_task *task, int copied)
 	dprintk("RPC: %5u xid %08x complete (%d bytes received)\n",
 			task->tk_pid, ntohl(req->rq_xid), copied);
 
+	task->tk_xprt->stat.recvs++;
+	task->tk_rtt = (long)jiffies - req->rq_xtime;
+
 	list_del_init(&req->rq_list);
+	/* Ensure all writes are done before we update req->rq_received */
+	smp_wmb();
 	req->rq_received = req->rq_private_buf.len = copied;
 	rpc_wake_up_task(task);
 }
@@ -723,7 +729,6 @@ void xprt_transmit(struct rpc_task *task)
 
 	dprintk("RPC: %4d xprt_transmit(%u)\n", task->tk_pid, req->rq_slen);
 
-	smp_rmb();
 	if (!req->rq_received) {
 		if (list_empty(&req->rq_list)) {
 			spin_lock_bh(&xprt->transport_lock);
@@ -744,12 +749,19 @@ void xprt_transmit(struct rpc_task *task)
 	if (status == 0) {
 		dprintk("RPC: %4d xmit complete\n", task->tk_pid);
 		spin_lock_bh(&xprt->transport_lock);
+
 		xprt->ops->set_retrans_timeout(task);
+
+		xprt->stat.sends++;
+		xprt->stat.req_u += xprt->stat.sends - xprt->stat.recvs;
+		xprt->stat.bklog_u += xprt->backlog.qlen;
+
 		/* Don't race with disconnect */
 		if (!xprt_connected(xprt))
 			task->tk_status = -ENOTCONN;
 		else if (!req->rq_received)
 			rpc_sleep_on(&xprt->pending, task, NULL, xprt_timer);
+
 		xprt->ops->release_xprt(xprt, task);
 		spin_unlock_bh(&xprt->transport_lock);
 		return;
@@ -848,6 +860,7 @@ void xprt_release(struct rpc_task *task)
 
 	if (!(req = task->tk_rqstp))
 		return;
+	rpc_count_iostats(task);
 	spin_lock_bh(&xprt->transport_lock);
 	xprt->ops->release_xprt(xprt, task);
 	if (xprt->ops->release_request)

net/sunrpc/xprtsock.c

@@ -382,6 +382,7 @@ static int xs_tcp_send_request(struct rpc_task *task)
 		/* If we've sent the entire packet, immediately
 		 * reset the count of bytes sent. */
 		req->rq_bytes_sent += status;
+		task->tk_bytes_sent += status;
 		if (likely(req->rq_bytes_sent >= req->rq_slen)) {
 			req->rq_bytes_sent = 0;
 			return 0;
@@ -1114,6 +1115,8 @@ static void xs_tcp_connect_worker(void *args)
 	}
 
 	/* Tell the socket layer to start connecting... */
+	xprt->stat.connect_count++;
+	xprt->stat.connect_start = jiffies;
 	status = sock->ops->connect(sock, (struct sockaddr *) &xprt->addr,
 			sizeof(xprt->addr), O_NONBLOCK);
 	dprintk("RPC: %p  connect status %d connected %d sock state %d\n",
@@ -1177,6 +1180,50 @@ static void xs_connect(struct rpc_task *task)
 	}
 }
 
+/**
+ * xs_udp_print_stats - display UDP socket-specifc stats
+ * @xprt: rpc_xprt struct containing statistics
+ * @seq: output file
+ *
+ */
+static void xs_udp_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
+{
+	seq_printf(seq, "\txprt:\tudp %u %lu %lu %lu %lu %Lu %Lu\n",
+			xprt->port,
+			xprt->stat.bind_count,
+			xprt->stat.sends,
+			xprt->stat.recvs,
+			xprt->stat.bad_xids,
+			xprt->stat.req_u,
+			xprt->stat.bklog_u);
+}
+
+/**
+ * xs_tcp_print_stats - display TCP socket-specifc stats
+ * @xprt: rpc_xprt struct containing statistics
+ * @seq: output file
+ *
+ */
+static void xs_tcp_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
+{
+	long idle_time = 0;
+
+	if (xprt_connected(xprt))
+		idle_time = (long)(jiffies - xprt->last_used) / HZ;
+
+	seq_printf(seq, "\txprt:\ttcp %u %lu %lu %lu %ld %lu %lu %lu %Lu %Lu\n",
+			xprt->port,
+			xprt->stat.bind_count,
+			xprt->stat.connect_count,
+			xprt->stat.connect_time,
+			idle_time,
+			xprt->stat.sends,
+			xprt->stat.recvs,
+			xprt->stat.bad_xids,
+			xprt->stat.req_u,
+			xprt->stat.bklog_u);
+}
+
 static struct rpc_xprt_ops xs_udp_ops = {
 	.set_buffer_size	= xs_udp_set_buffer_size,
 	.reserve_xprt		= xprt_reserve_xprt_cong,
@@ -1191,6 +1238,7 @@ static struct rpc_xprt_ops xs_udp_ops = {
 	.release_request	= xprt_release_rqst_cong,
 	.close			= xs_close,
 	.destroy		= xs_destroy,
+	.print_stats		= xs_udp_print_stats,
 };
 
 static struct rpc_xprt_ops xs_tcp_ops = {
@@ -1204,6 +1252,7 @@ static struct rpc_xprt_ops xs_tcp_ops = {
 	.set_retrans_timeout	= xprt_set_retrans_timeout_def,
 	.close			= xs_close,
 	.destroy		= xs_destroy,
+	.print_stats		= xs_tcp_print_stats,
 };
 
 /**