net/sunrpc/svc_xprt.c

/*
 * linux/net/sunrpc/svc_xprt.c
 *
 * Author: Tom Tucker <tom@opengridcomputing.com>
 */

#include <linux/sched.h>
#include <linux/errno.h>
#include <linux/fcntl.h>
#include <linux/net.h>
#include <linux/in.h>
#include <linux/inet.h>
#include <linux/udp.h>
#include <linux/tcp.h>
#include <linux/unistd.h>
#include <linux/slab.h>
#include <linux/netdevice.h>
#include <linux/skbuff.h>
#include <linux/file.h>
#include <linux/freezer.h>
#include <net/sock.h>
#include <net/checksum.h>
#include <net/ip.h>
#include <net/ipv6.h>
#include <net/tcp_states.h>
#include <linux/uaccess.h>
#include <asm/ioctls.h>

#include <linux/sunrpc/types.h>
#include <linux/sunrpc/clnt.h>
#include <linux/sunrpc/xdr.h>
#include <linux/sunrpc/stats.h>
#include <linux/sunrpc/svc_xprt.h>

#define RPCDBG_FACILITY	RPCDBG_SVCXPRT

static struct svc_deferred_req *svc_deferred_dequeue(struct svc_xprt *xprt);
static int svc_deferred_recv(struct svc_rqst *rqstp);
static struct cache_deferred_req *svc_defer(struct cache_req *req);
static void svc_age_temp_xprts(unsigned long closure);

/* apparently the "standard" is that clients close
 * idle connections after 5 minutes, servers after
 * 6 minutes
 *   http://www.connectathon.org/talks96/nfstcp.pdf
 */
static int svc_conn_age_period = 6*60;

/* List of registered transport classes */
static DEFINE_SPINLOCK(svc_xprt_class_lock);
static LIST_HEAD(svc_xprt_class_list);

/* SMP locking strategy:
 *
 *	svc_pool->sp_lock protects most of the fields of that pool.
 *	svc_serv->sv_lock protects sv_tempsocks, sv_permsocks, sv_tmpcnt.
 *	when both need to be taken (rare), svc_serv->sv_lock is first.
 *	BKL protects svc_serv->sv_nrthread.
 *	svc_sock->sk_lock protects the svc_sock->sk_deferred list
 *             and the ->sk_info_authunix cache.
 *
 *	The XPT_BUSY bit in xprt->xpt_flags prevents a transport being
 *	enqueued multiply. During normal transport processing this bit
 *	is set by svc_xprt_enqueue and cleared by svc_xprt_received.
 *	Providers should not manipulate this bit directly.
 *
 *	Some flags can be set to certain values at any time
 *	providing that certain rules are followed:
 *
 *	XPT_CONN, XPT_DATA:
 *		- Can be set or cleared at any time.
 *		- After a set, svc_xprt_enqueue must be called to enqueue
 *		  the transport for processing.
 *		- After a clear, the transport must be read/accepted.
 *		  If this succeeds, it must be set again.
 *	XPT_CLOSE:
 *		- Can set at any time. It is never cleared.
 *      XPT_DEAD:
 *		- Can only be set while XPT_BUSY is held which ensures
 *		  that no other thread will be using the transport or will
 *		  try to set XPT_DEAD.
 */

int svc_reg_xprt_class(struct svc_xprt_class *xcl)
{
	struct svc_xprt_class *cl;
	int res = -EEXIST;

	dprintk("svc: Adding svc transport class '%s'\n", xcl->xcl_name);

	INIT_LIST_HEAD(&xcl->xcl_list);
	spin_lock(&svc_xprt_class_lock);
	/* Make sure there isn't already a class with the same name */
	list_for_each_entry(cl, &svc_xprt_class_list, xcl_list) {
		if (strcmp(xcl->xcl_name, cl->xcl_name) == 0)
			goto out;
	}
	list_add_tail(&xcl->xcl_list, &svc_xprt_class_list);
	res = 0;
out:
	spin_unlock(&svc_xprt_class_lock);
	return res;
}
EXPORT_SYMBOL_GPL(svc_reg_xprt_class);

void svc_unreg_xprt_class(struct svc_xprt_class *xcl)
{
	dprintk("svc: Removing svc transport class '%s'\n", xcl->xcl_name);
	spin_lock(&svc_xprt_class_lock);
	list_del_init(&xcl->xcl_list);
	spin_unlock(&svc_xprt_class_lock);
}
EXPORT_SYMBOL_GPL(svc_unreg_xprt_class);

static void svc_xprt_free(struct kref *kref)
{
	struct svc_xprt *xprt =
		container_of(kref, struct svc_xprt, xpt_ref);
	struct module *owner = xprt->xpt_class->xcl_owner;
	if (test_bit(XPT_CACHE_AUTH, &xprt->xpt_flags)
	    && xprt->xpt_auth_cache != NULL)
		svcauth_unix_info_release(xprt->xpt_auth_cache);
	xprt->xpt_ops->xpo_free(xprt);
	module_put(owner);
}

void svc_xprt_put(struct svc_xprt *xprt)
{
	kref_put(&xprt->xpt_ref, svc_xprt_free);
}
EXPORT_SYMBOL_GPL(svc_xprt_put);

/*
 * Called by transport drivers to initialize the transport independent
 * portion of the transport instance.
 */
void svc_xprt_init(struct svc_xprt_class *xcl, struct svc_xprt *xprt,
		   struct svc_serv *serv)
{
	memset(xprt, 0, sizeof(*xprt));
	xprt->xpt_class = xcl;
	xprt->xpt_ops = xcl->xcl_ops;
	kref_init(&xprt->xpt_ref);
	xprt->xpt_server = serv;
	INIT_LIST_HEAD(&xprt->xpt_list);
	INIT_LIST_HEAD(&xprt->xpt_ready);
	INIT_LIST_HEAD(&xprt->xpt_deferred);
	mutex_init(&xprt->xpt_mutex);
	spin_lock_init(&xprt->xpt_lock);
	set_bit(XPT_BUSY, &xprt->xpt_flags);
}
EXPORT_SYMBOL_GPL(svc_xprt_init);

int svc_create_xprt(struct svc_serv *serv, char *xprt_name, unsigned short port,
		    int flags)
{
	struct svc_xprt_class *xcl;
	struct sockaddr_in sin = {
		.sin_family		= AF_INET,
		.sin_addr.s_addr	= INADDR_ANY,
		.sin_port		= htons(port),
	};
	dprintk("svc: creating transport %s[%d]\n", xprt_name, port);
	spin_lock(&svc_xprt_class_lock);
	list_for_each_entry(xcl, &svc_xprt_class_list, xcl_list) {
		struct svc_xprt *newxprt;

		if (strcmp(xprt_name, xcl->xcl_name))
			continue;

		if (!try_module_get(xcl->xcl_owner))
			goto err;

		spin_unlock(&svc_xprt_class_lock);
		newxprt = xcl->xcl_ops->
			xpo_create(serv, (struct sockaddr *)&sin, sizeof(sin),
				   flags);
		if (IS_ERR(newxprt)) {
			module_put(xcl->xcl_owner);
			return PTR_ERR(newxprt);
		}

		clear_bit(XPT_TEMP, &newxprt->xpt_flags);
		spin_lock_bh(&serv->sv_lock);
		list_add(&newxprt->xpt_list, &serv->sv_permsocks);
		spin_unlock_bh(&serv->sv_lock);
		clear_bit(XPT_BUSY, &newxprt->xpt_flags);
		return svc_xprt_local_port(newxprt);
	}
 err:
	spin_unlock(&svc_xprt_class_lock);
	dprintk("svc: transport %s not found\n", xprt_name);
	return -ENOENT;
}
EXPORT_SYMBOL_GPL(svc_create_xprt);

/*
 * Copy the local and remote xprt addresses to the rqstp structure
 */
void svc_xprt_copy_addrs(struct svc_rqst *rqstp, struct svc_xprt *xprt)
{
	struct sockaddr *sin;

	memcpy(&rqstp->rq_addr, &xprt->xpt_remote, xprt->xpt_remotelen);
	rqstp->rq_addrlen = xprt->xpt_remotelen;

	/*
	 * Destination address in request is needed for binding the
	 * source address in RPC replies/callbacks later.
	 */
	sin = (struct sockaddr *)&xprt->xpt_local;
	switch (sin->sa_family) {
	case AF_INET:
		rqstp->rq_daddr.addr = ((struct sockaddr_in *)sin)->sin_addr;
		break;
	case AF_INET6:
		rqstp->rq_daddr.addr6 = ((struct sockaddr_in6 *)sin)->sin6_addr;
		break;
	}
}
EXPORT_SYMBOL_GPL(svc_xprt_copy_addrs);

/**
 * svc_print_addr - Format rq_addr field for printing
 * @rqstp: svc_rqst struct containing address to print
 * @buf: target buffer for formatted address
 * @len: length of target buffer
 *
 */
char *svc_print_addr(struct svc_rqst *rqstp, char *buf, size_t len)
{
	return __svc_print_addr(svc_addr(rqstp), buf, len);
}
EXPORT_SYMBOL_GPL(svc_print_addr);

/*
 * Queue up an idle server thread.  Must have pool->sp_lock held.
 * Note: this is really a stack rather than a queue, so that we only
 * use as many different threads as we need, and the rest don't pollute
 * the cache.
 */
static void svc_thread_enqueue(struct svc_pool *pool, struct svc_rqst *rqstp)
{
	list_add(&rqstp->rq_list, &pool->sp_threads);
}

/*
 * Dequeue an nfsd thread.  Must have pool->sp_lock held.
 */
static void svc_thread_dequeue(struct svc_pool *pool, struct svc_rqst *rqstp)
{
	list_del(&rqstp->rq_list);
}

/*
 * Queue up a transport with data pending. If there are idle nfsd
 * processes, wake 'em up.
 *
 */
void svc_xprt_enqueue(struct svc_xprt *xprt)
{
	struct svc_serv	*serv = xprt->xpt_server;
	struct svc_pool *pool;
	struct svc_rqst	*rqstp;
	int cpu;

	if (!(xprt->xpt_flags &
	      ((1<<XPT_CONN)|(1<<XPT_DATA)|(1<<XPT_CLOSE)|(1<<XPT_DEFERRED))))
		return;
	if (test_bit(XPT_DEAD, &xprt->xpt_flags))
		return;

	cpu = get_cpu();
	pool = svc_pool_for_cpu(xprt->xpt_server, cpu);
	put_cpu();

	spin_lock_bh(&pool->sp_lock);

	if (!list_empty(&pool->sp_threads) &&
	    !list_empty(&pool->sp_sockets))
		printk(KERN_ERR
		       "svc_xprt_enqueue: "
		       "threads and transports both waiting??\n");

	if (test_bit(XPT_DEAD, &xprt->xpt_flags)) {
		/* Don't enqueue dead transports */
		dprintk("svc: transport %p is dead, not enqueued\n", xprt);
		goto out_unlock;
	}

	/* Mark transport as busy. It will remain in this state until
	 * the provider calls svc_xprt_received. We update XPT_BUSY
	 * atomically because it also guards against trying to enqueue
	 * the transport twice.
	 */
	if (test_and_set_bit(XPT_BUSY, &xprt->xpt_flags)) {
		/* Don't enqueue transport while already enqueued */
		dprintk("svc: transport %p busy, not enqueued\n", xprt);
		goto out_unlock;
	}
	BUG_ON(xprt->xpt_pool != NULL);
	xprt->xpt_pool = pool;

	/* Handle pending connection */
	if (test_bit(XPT_CONN, &xprt->xpt_flags))
		goto process;

	/* Handle close in-progress */
	if (test_bit(XPT_CLOSE, &xprt->xpt_flags))
		goto process;

	/* Check if we have space to reply to a request */
	if (!xprt->xpt_ops->xpo_has_wspace(xprt)) {
		/* Don't enqueue while not enough space for reply */
		dprintk("svc: no write space, transport %p  not enqueued\n",
			xprt);
		xprt->xpt_pool = NULL;
		clear_bit(XPT_BUSY, &xprt->xpt_flags);
		goto out_unlock;
	}

 process:
	if (!list_empty(&pool->sp_threads)) {
		rqstp = list_entry(pool->sp_threads.next,
				   struct svc_rqst,
				   rq_list);
		dprintk("svc: transport %p served by daemon %p\n",
			xprt, rqstp);
		svc_thread_dequeue(pool, rqstp);
		if (rqstp->rq_xprt)
			printk(KERN_ERR
				"svc_xprt_enqueue: server %p, rq_xprt=%p!\n",
				rqstp, rqstp->rq_xprt);
		rqstp->rq_xprt = xprt;
		svc_xprt_get(xprt);
		rqstp->rq_reserved = serv->sv_max_mesg;
		atomic_add(rqstp->rq_reserved, &xprt->xpt_reserved);
		BUG_ON(xprt->xpt_pool != pool);
		wake_up(&rqstp->rq_wait);
	} else {
		dprintk("svc: transport %p put into queue\n", xprt);
		list_add_tail(&xprt->xpt_ready, &pool->sp_sockets);
		BUG_ON(xprt->xpt_pool != pool);
	}

out_unlock:
	spin_unlock_bh(&pool->sp_lock);
}
EXPORT_SYMBOL_GPL(svc_xprt_enqueue);

/*
 * Dequeue the first transport.  Must be called with the pool->sp_lock held.
 */
static struct svc_xprt *svc_xprt_dequeue(struct svc_pool *pool)
{
	struct svc_xprt	*xprt;

	if (list_empty(&pool->sp_sockets))
		return NULL;

	xprt = list_entry(pool->sp_sockets.next,
			  struct svc_xprt, xpt_ready);
	list_del_init(&xprt->xpt_ready);

	dprintk("svc: transport %p dequeued, inuse=%d\n",
		xprt, atomic_read(&xprt->xpt_ref.refcount));

	return xprt;
}

/*
 * svc_xprt_received conditionally queues the transport for processing
 * by another thread. The caller must hold the XPT_BUSY bit and must
 * not thereafter touch transport data.
 *
 * Note: XPT_DATA only gets cleared when a read-attempt finds no (or
 * insufficient) data.
 */
void svc_xprt_received(struct svc_xprt *xprt)
{
	BUG_ON(!test_bit(XPT_BUSY, &xprt->xpt_flags));
	xprt->xpt_pool = NULL;
	clear_bit(XPT_BUSY, &xprt->xpt_flags);
	svc_xprt_enqueue(xprt);
}
EXPORT_SYMBOL_GPL(svc_xprt_received);

/**
 * svc_reserve - change the space reserved for the reply to a request.
 * @rqstp:  The request in question
 * @space: new max space to reserve
 *
 * Each request reserves some space on the output queue of the transport
 * to make sure the reply fits.  This function reduces that reserved
 * space to be the amount of space used already, plus @space.
 *
 */
void svc_reserve(struct svc_rqst *rqstp, int space)
{
	space += rqstp->rq_res.head[0].iov_len;

	if (space < rqstp->rq_reserved) {
		struct svc_xprt *xprt = rqstp->rq_xprt;
		atomic_sub((rqstp->rq_reserved - space), &xprt->xpt_reserved);
		rqstp->rq_reserved = space;

		svc_xprt_enqueue(xprt);
	}
}

static void svc_xprt_release(struct svc_rqst *rqstp)
{
	struct svc_xprt	*xprt = rqstp->rq_xprt;

	rqstp->rq_xprt->xpt_ops->xpo_release_rqst(rqstp);

	svc_free_res_pages(rqstp);
	rqstp->rq_res.page_len = 0;
	rqstp->rq_res.page_base = 0;

	/* Reset response buffer and release
	 * the reservation.
	 * But first, check that enough space was reserved
	 * for the reply, otherwise we have a bug!
	 */
	if ((rqstp->rq_res.len) >  rqstp->rq_reserved)
		printk(KERN_ERR "RPC request reserved %d but used %d\n",
		       rqstp->rq_reserved,
		       rqstp->rq_res.len);

	rqstp->rq_res.head[0].iov_len = 0;
	svc_reserve(rqstp, 0);
	rqstp->rq_xprt = NULL;

	svc_xprt_put(xprt);
}

/*
 * External function to wake up a server waiting for data
 * This really only makes sense for services like lockd
 * which have exactly one thread anyway.
 */
void svc_wake_up(struct svc_serv *serv)
{
	struct svc_rqst	*rqstp;
	unsigned int i;
	struct svc_pool *pool;

	for (i = 0; i < serv->sv_nrpools; i++) {
		pool = &serv->sv_pools[i];

		spin_lock_bh(&pool->sp_lock);
		if (!list_empty(&pool->sp_threads)) {
			rqstp = list_entry(pool->sp_threads.next,
					   struct svc_rqst,
					   rq_list);
			dprintk("svc: daemon %p woken up.\n", rqstp);
			/*
			svc_thread_dequeue(pool, rqstp);
			rqstp->rq_xprt = NULL;
			 */
			wake_up(&rqstp->rq_wait);
		}
		spin_unlock_bh(&pool->sp_lock);
	}
}

int svc_port_is_privileged(struct sockaddr *sin)
{
	switch (sin->sa_family) {
	case AF_INET:
		return ntohs(((struct sockaddr_in *)sin)->sin_port)
			< PROT_SOCK;
	case AF_INET6:
		return ntohs(((struct sockaddr_in6 *)sin)->sin6_port)
			< PROT_SOCK;
	default:
		return 0;
	}
}

/*
 * Make sure that we don't have too many active connections.  If we
 * have, something must be dropped.
 *
 * There's no point in trying to do random drop here for DoS
 * prevention. The NFS clients does 1 reconnect in 15 seconds. An
 * attacker can easily beat that.
 *
 * The only somewhat efficient mechanism would be if drop old
 * connections from the same IP first. But right now we don't even
 * record the client IP in svc_sock.
 */
static void svc_check_conn_limits(struct svc_serv *serv)
{
	if (serv->sv_tmpcnt > (serv->sv_nrthreads+3)*20) {
		struct svc_xprt *xprt = NULL;
		spin_lock_bh(&serv->sv_lock);
		if (!list_empty(&serv->sv_tempsocks)) {
			if (net_ratelimit()) {
				/* Try to help the admin */
				printk(KERN_NOTICE "%s: too many open  "
				       "connections, consider increasing the "
				       "number of nfsd threads\n",
				       serv->sv_name);
			}
			/*
			 * Always select the oldest connection. It's not fair,
			 * but so is life
			 */
			xprt = list_entry(serv->sv_tempsocks.prev,
					  struct svc_xprt,
					  xpt_list);
			set_bit(XPT_CLOSE, &xprt->xpt_flags);
			svc_xprt_get(xprt);
		}
		spin_unlock_bh(&serv->sv_lock);

		if (xprt) {
			svc_xprt_enqueue(xprt);
			svc_xprt_put(xprt);
		}
	}
}

/*
 * Receive the next request on any transport.  This code is carefully
 * organised not to touch any cachelines in the shared svc_serv
 * structure, only cachelines in the local svc_pool.
 */
int svc_recv(struct svc_rqst *rqstp, long timeout)
{
	struct svc_xprt		*xprt = NULL;
	struct svc_serv		*serv = rqstp->rq_server;
	struct svc_pool		*pool = rqstp->rq_pool;
	int			len, i;
	int			pages;
	struct xdr_buf		*arg;
	DECLARE_WAITQUEUE(wait, current);

	dprintk("svc: server %p waiting for data (to = %ld)\n",
		rqstp, timeout);

	if (rqstp->rq_xprt)
		printk(KERN_ERR
			"svc_recv: service %p, transport not NULL!\n",
			 rqstp);
	if (waitqueue_active(&rqstp->rq_wait))
		printk(KERN_ERR
			"svc_recv: service %p, wait queue active!\n",
			 rqstp);

	/* now allocate needed pages.  If we get a failure, sleep briefly */
	pages = (serv->sv_max_mesg + PAGE_SIZE) / PAGE_SIZE;
	for (i = 0; i < pages ; i++)
		while (rqstp->rq_pages[i] == NULL) {
			struct page *p = alloc_page(GFP_KERNEL);
			if (!p) {
				int j = msecs_to_jiffies(500);
				schedule_timeout_uninterruptible(j);
			}
			rqstp->rq_pages[i] = p;
		}
	rqstp->rq_pages[i++] = NULL; /* this might be seen in nfs_read_actor */
	BUG_ON(pages >= RPCSVC_MAXPAGES);

	/* Make arg->head point to first page and arg->pages point to rest */
	arg = &rqstp->rq_arg;
	arg->head[0].iov_base = page_address(rqstp->rq_pages[0]);
	arg->head[0].iov_len = PAGE_SIZE;
	arg->pages = rqstp->rq_pages + 1;
	arg->page_base = 0;
	/* save at least one page for response */
	arg->page_len = (pages-2)*PAGE_SIZE;
	arg->len = (pages-1)*PAGE_SIZE;
	arg->tail[0].iov_len = 0;

	try_to_freeze();
	cond_resched();
	if (signalled())
		return -EINTR;

	spin_lock_bh(&pool->sp_lock);
	xprt = svc_xprt_dequeue(pool);
	if (xprt) {
		rqstp->rq_xprt = xprt;
		svc_xprt_get(xprt);
		rqstp->rq_reserved = serv->sv_max_mesg;
		atomic_add(rqstp->rq_reserved, &xprt->xpt_reserved);
	} else {
		/* No data pending. Go to sleep */
		svc_thread_enqueue(pool, rqstp);

		/*
		 * We have to be able to interrupt this wait
		 * to bring down the daemons ...
		 */
		set_current_state(TASK_INTERRUPTIBLE);
		add_wait_queue(&rqstp->rq_wait, &wait);
		spin_unlock_bh(&pool->sp_lock);

		schedule_timeout(timeout);

		try_to_freeze();

		spin_lock_bh(&pool->sp_lock);
		remove_wait_queue(&rqstp->rq_wait, &wait);

		xprt = rqstp->rq_xprt;
		if (!xprt) {
			svc_thread_dequeue(pool, rqstp);
			spin_unlock_bh(&pool->sp_lock);
			dprintk("svc: server %p, no data yet\n", rqstp);
			return signalled()? -EINTR : -EAGAIN;
		}
	}
	spin_unlock_bh(&pool->sp_lock);

	len = 0;
	if (test_bit(XPT_CLOSE, &xprt->xpt_flags)) {
		dprintk("svc_recv: found XPT_CLOSE\n");
		svc_delete_xprt(xprt);
	} else if (test_bit(XPT_LISTENER, &xprt->xpt_flags)) {
		struct svc_xprt *newxpt;
		newxpt = xprt->xpt_ops->xpo_accept(xprt);
		if (newxpt) {
			/*
			 * We know this module_get will succeed because the
			 * listener holds a reference too
			 */
			__module_get(newxpt->xpt_class->xcl_owner);
			svc_check_conn_limits(xprt->xpt_server);
			spin_lock_bh(&serv->sv_lock);
			set_bit(XPT_TEMP, &newxpt->xpt_flags);
			list_add(&newxpt->xpt_list, &serv->sv_tempsocks);
			serv->sv_tmpcnt++;
			if (serv->sv_temptimer.function == NULL) {
				/* setup timer to age temp transports */
				setup_timer(&serv->sv_temptimer,
					    svc_age_temp_xprts,
					    (unsigned long)serv);
				mod_timer(&serv->sv_temptimer,
					  jiffies + svc_conn_age_period * HZ);
			}
			spin_unlock_bh(&serv->sv_lock);
			svc_xprt_received(newxpt);
		}
		svc_xprt_received(xprt);
	} else {
		dprintk("svc: server %p, pool %u, transport %p, inuse=%d\n",
			rqstp, pool->sp_id, xprt,
			atomic_read(&xprt->xpt_ref.refcount));
		rqstp->rq_deferred = svc_deferred_dequeue(xprt);
		if (rqstp->rq_deferred) {
			svc_xprt_received(xprt);
			len = svc_deferred_recv(rqstp);
		} else
			len = xprt->xpt_ops->xpo_recvfrom(rqstp);
		dprintk("svc: got len=%d\n", len);
	}

	/* No data, incomplete (TCP) read, or accept() */
	if (len == 0 || len == -EAGAIN) {
		rqstp->rq_res.len = 0;
		svc_xprt_release(rqstp);
		return -EAGAIN;
	}
	clear_bit(XPT_OLD, &xprt->xpt_flags);

	rqstp->rq_secure = svc_port_is_privileged(svc_addr(rqstp));
	rqstp->rq_chandle.defer = svc_defer;

	if (serv->sv_stats)
		serv->sv_stats->netcnt++;
	return len;
}

/*
 * Drop request
 */
void svc_drop(struct svc_rqst *rqstp)
{
	dprintk("svc: xprt %p dropped request\n", rqstp->rq_xprt);
	svc_xprt_release(rqstp);
}

/*
 * Return reply to client.
 */
int svc_send(struct svc_rqst *rqstp)
{
	struct svc_xprt	*xprt;
	int		len;
	struct xdr_buf	*xb;

	xprt = rqstp->rq_xprt;
	if (!xprt)
		return -EFAULT;

	/* release the receive skb before sending the reply */
	rqstp->rq_xprt->xpt_ops->xpo_release_rqst(rqstp);

	/* calculate over-all length */
	xb = &rqstp->rq_res;
	xb->len = xb->head[0].iov_len +
		xb->page_len +
		xb->tail[0].iov_len;

	/* Grab mutex to serialize outgoing data. */
	mutex_lock(&xprt->xpt_mutex);
	if (test_bit(XPT_DEAD, &xprt->xpt_flags))
		len = -ENOTCONN;
	else
		len = xprt->xpt_ops->xpo_sendto(rqstp);
	mutex_unlock(&xprt->xpt_mutex);
	svc_xprt_release(rqstp);

	if (len == -ECONNREFUSED || len == -ENOTCONN || len == -EAGAIN)
		return 0;
	return len;
}

/*
 * Timer function to close old temporary transports, using
 * a mark-and-sweep algorithm.
 */
static void svc_age_temp_xprts(unsigned long closure)
{
	struct svc_serv *serv = (struct svc_serv *)closure;
	struct svc_xprt *xprt;
	struct list_head *le, *next;
	LIST_HEAD(to_be_aged);

	dprintk("svc_age_temp_xprts\n");

	if (!spin_trylock_bh(&serv->sv_lock)) {
		/* busy, try again 1 sec later */
		dprintk("svc_age_temp_xprts: busy\n");
		mod_timer(&serv->sv_temptimer, jiffies + HZ);
		return;
	}

	list_for_each_safe(le, next, &serv->sv_tempsocks) {
		xprt = list_entry(le, struct svc_xprt, xpt_list);

		/* First time through, just mark it OLD. Second time
		 * through, close it. */
		if (!test_and_set_bit(XPT_OLD, &xprt->xpt_flags))
			continue;
		if (atomic_read(&xprt->xpt_ref.refcount) > 1
		    || test_bit(XPT_BUSY, &xprt->xpt_flags))
			continue;
		svc_xprt_get(xprt);
		list_move(le, &to_be_aged);
		set_bit(XPT_CLOSE, &xprt->xpt_flags);
		set_bit(XPT_DETACHED, &xprt->xpt_flags);
	}
	spin_unlock_bh(&serv->sv_lock);

	while (!list_empty(&to_be_aged)) {
		le = to_be_aged.next;
		/* fiddling the xpt_list node is safe 'cos we're XPT_DETACHED */
		list_del_init(le);
		xprt = list_entry(le, struct svc_xprt, xpt_list);

		dprintk("queuing xprt %p for closing\n", xprt);

		/* a thread will dequeue and close it soon */
		svc_xprt_enqueue(xprt);
		svc_xprt_put(xprt);
	}

	mod_timer(&serv->sv_temptimer, jiffies + svc_conn_age_period * HZ);
}

/*
 * Remove a dead transport
 */
void svc_delete_xprt(struct svc_xprt *xprt)
{
	struct svc_serv	*serv = xprt->xpt_server;

	dprintk("svc: svc_delete_xprt(%p)\n", xprt);
	xprt->xpt_ops->xpo_detach(xprt);

	spin_lock_bh(&serv->sv_lock);
	if (!test_and_set_bit(XPT_DETACHED, &xprt->xpt_flags))
		list_del_init(&xprt->xpt_list);
	/*
	 * We used to delete the transport from whichever list
	 * it's sk_xprt.xpt_ready node was on, but we don't actually
	 * need to.  This is because the only time we're called
	 * while still attached to a queue, the queue itself
	 * is about to be destroyed (in svc_destroy).
	 */
	if (!test_and_set_bit(XPT_DEAD, &xprt->xpt_flags)) {
		BUG_ON(atomic_read(&xprt->xpt_ref.refcount) < 2);
		if (test_bit(XPT_TEMP, &xprt->xpt_flags))
			serv->sv_tmpcnt--;
		svc_xprt_put(xprt);
	}
	spin_unlock_bh(&serv->sv_lock);
}

void svc_close_xprt(struct svc_xprt *xprt)
{
	set_bit(XPT_CLOSE, &xprt->xpt_flags);
	if (test_and_set_bit(XPT_BUSY, &xprt->xpt_flags))
		/* someone else will have to effect the close */
		return;

	svc_xprt_get(xprt);
	svc_delete_xprt(xprt);
	clear_bit(XPT_BUSY, &xprt->xpt_flags);
	svc_xprt_put(xprt);
}

void svc_close_all(struct list_head *xprt_list)
{
	struct svc_xprt *xprt;
	struct svc_xprt *tmp;

	list_for_each_entry_safe(xprt, tmp, xprt_list, xpt_list) {
		set_bit(XPT_CLOSE, &xprt->xpt_flags);
		if (test_bit(XPT_BUSY, &xprt->xpt_flags)) {
			/* Waiting to be processed, but no threads left,
			 * So just remove it from the waiting list
			 */
			list_del_init(&xprt->xpt_ready);
			clear_bit(XPT_BUSY, &xprt->xpt_flags);
		}
		svc_close_xprt(xprt);
	}
}

/*
 * Handle defer and revisit of requests
 */

static void svc_revisit(struct cache_deferred_req *dreq, int too_many)
{
	struct svc_deferred_req *dr =
		container_of(dreq, struct svc_deferred_req, handle);
	struct svc_xprt *xprt = dr->xprt;

	if (too_many) {
		svc_xprt_put(xprt);
		kfree(dr);
		return;
	}
	dprintk("revisit queued\n");
	dr->xprt = NULL;
	spin_lock(&xprt->xpt_lock);
	list_add(&dr->handle.recent, &xprt->xpt_deferred);
	spin_unlock(&xprt->xpt_lock);
	set_bit(XPT_DEFERRED, &xprt->xpt_flags);
	svc_xprt_enqueue(xprt);
	svc_xprt_put(xprt);
}

/*
 * Save the request off for later processing. The request buffer looks
 * like this:
 *
 * <xprt-header><rpc-header><rpc-pagelist><rpc-tail>
 *
 * This code can only handle requests that consist of an xprt-header
 * and rpc-header.
 */
static struct cache_deferred_req *svc_defer(struct cache_req *req)
{
	struct svc_rqst *rqstp = container_of(req, struct svc_rqst, rq_chandle);
	struct svc_deferred_req *dr;

	if (rqstp->rq_arg.page_len)
		return NULL; /* if more than a page, give up FIXME */
	if (rqstp->rq_deferred) {
		dr = rqstp->rq_deferred;
		rqstp->rq_deferred = NULL;
	} else {
		size_t skip;
		size_t size;
		/* FIXME maybe discard if size too large */
		size = sizeof(struct svc_deferred_req) + rqstp->rq_arg.len;
		dr = kmalloc(size, GFP_KERNEL);
		if (dr == NULL)
			return NULL;

		dr->handle.owner = rqstp->rq_server;
		dr->prot = rqstp->rq_prot;
		memcpy(&dr->addr, &rqstp->rq_addr, rqstp->rq_addrlen);
		dr->addrlen = rqstp->rq_addrlen;
		dr->daddr = rqstp->rq_daddr;
		dr->argslen = rqstp->rq_arg.len >> 2;
		dr->xprt_hlen = rqstp->rq_xprt_hlen;

		/* back up head to the start of the buffer and copy */
		skip = rqstp->rq_arg.len - rqstp->rq_arg.head[0].iov_len;
		memcpy(dr->args, rqstp->rq_arg.head[0].iov_base - skip,
		       dr->argslen << 2);
	}
	svc_xprt_get(rqstp->rq_xprt);
	dr->xprt = rqstp->rq_xprt;

	dr->handle.revisit = svc_revisit;
	return &dr->handle;
}

/*
 * recv data from a deferred request into an active one
 */
static int svc_deferred_recv(struct svc_rqst *rqstp)
{
	struct svc_deferred_req *dr = rqstp->rq_deferred;

	/* setup iov_base past transport header */
	rqstp->rq_arg.head[0].iov_base = dr->args + (dr->xprt_hlen>>2);
	/* The iov_len does not include the transport header bytes */
	rqstp->rq_arg.head[0].iov_len = (dr->argslen<<2) - dr->xprt_hlen;
	rqstp->rq_arg.page_len = 0;
	/* The rq_arg.len includes the transport header bytes */
	rqstp->rq_arg.len     = dr->argslen<<2;
	rqstp->rq_prot        = dr->prot;
	memcpy(&rqstp->rq_addr, &dr->addr, dr->addrlen);
	rqstp->rq_addrlen     = dr->addrlen;
	/* Save off transport header len in case we get deferred again */
	rqstp->rq_xprt_hlen   = dr->xprt_hlen;
	rqstp->rq_daddr       = dr->daddr;
	rqstp->rq_respages    = rqstp->rq_pages;
	return (dr->argslen<<2) - dr->xprt_hlen;
}


static struct svc_deferred_req *svc_deferred_dequeue(struct svc_xprt *xprt)
{
	struct svc_deferred_req *dr = NULL;

	if (!test_bit(XPT_DEFERRED, &xprt->xpt_flags))
		return NULL;
	spin_lock(&xprt->xpt_lock);
	clear_bit(XPT_DEFERRED, &xprt->xpt_flags);
	if (!list_empty(&xprt->xpt_deferred)) {
		dr = list_entry(xprt->xpt_deferred.next,
				struct svc_deferred_req,
				handle.recent);
		list_del_init(&dr->handle.recent);
		set_bit(XPT_DEFERRED, &xprt->xpt_flags);
	}
	spin_unlock(&xprt->xpt_lock);
	return dr;
}