2 * linux/net/sunrpc/svc_xprt.c
4 * Author: Tom Tucker <tom@opengridcomputing.com>
7 #include <linux/sched.h>
8 #include <linux/errno.h>
9 #include <linux/fcntl.h>
10 #include <linux/net.h>
12 #include <linux/inet.h>
13 #include <linux/udp.h>
14 #include <linux/tcp.h>
15 #include <linux/unistd.h>
16 #include <linux/slab.h>
17 #include <linux/netdevice.h>
18 #include <linux/skbuff.h>
19 #include <linux/file.h>
20 #include <linux/freezer.h>
21 #include <linux/kthread.h>
23 #include <net/checksum.h>
26 #include <net/tcp_states.h>
27 #include <linux/uaccess.h>
28 #include <asm/ioctls.h>
30 #include <linux/sunrpc/types.h>
31 #include <linux/sunrpc/clnt.h>
32 #include <linux/sunrpc/xdr.h>
33 #include <linux/sunrpc/stats.h>
34 #include <linux/sunrpc/svc_xprt.h>
36 #define RPCDBG_FACILITY RPCDBG_SVCXPRT
38 static struct svc_deferred_req *svc_deferred_dequeue(struct svc_xprt *xprt);
39 static int svc_deferred_recv(struct svc_rqst *rqstp);
40 static struct cache_deferred_req *svc_defer(struct cache_req *req);
41 static void svc_age_temp_xprts(unsigned long closure);
43 /* apparently the "standard" is that clients close
44 * idle connections after 5 minutes, servers after
46 * http://www.connectathon.org/talks96/nfstcp.pdf
48 static int svc_conn_age_period = 6*60;
50 /* List of registered transport classes */
51 static DEFINE_SPINLOCK(svc_xprt_class_lock);
52 static LIST_HEAD(svc_xprt_class_list);
54 /* SMP locking strategy:
56 * svc_pool->sp_lock protects most of the fields of that pool.
57 * svc_serv->sv_lock protects sv_tempsocks, sv_permsocks, sv_tmpcnt.
58 * when both need to be taken (rare), svc_serv->sv_lock is first.
59 * BKL protects svc_serv->sv_nrthread.
60 * svc_sock->sk_lock protects the svc_sock->sk_deferred list
61 * and the ->sk_info_authunix cache.
63 * The XPT_BUSY bit in xprt->xpt_flags prevents a transport being
64 * enqueued multiply. During normal transport processing this bit
65 * is set by svc_xprt_enqueue and cleared by svc_xprt_received.
66 * Providers should not manipulate this bit directly.
68 * Some flags can be set to certain values at any time
69 * providing that certain rules are followed:
72 * - Can be set or cleared at any time.
73 * - After a set, svc_xprt_enqueue must be called to enqueue
74 * the transport for processing.
75 * - After a clear, the transport must be read/accepted.
76 * If this succeeds, it must be set again.
78 * - Can set at any time. It is never cleared.
80 * - Can only be set while XPT_BUSY is held which ensures
81 * that no other thread will be using the transport or will
82 * try to set XPT_DEAD.
85 int svc_reg_xprt_class(struct svc_xprt_class *xcl)
87 struct svc_xprt_class *cl;
90 dprintk("svc: Adding svc transport class '%s'\n", xcl->xcl_name);
92 INIT_LIST_HEAD(&xcl->xcl_list);
93 spin_lock(&svc_xprt_class_lock);
94 /* Make sure there isn't already a class with the same name */
95 list_for_each_entry(cl, &svc_xprt_class_list, xcl_list) {
96 if (strcmp(xcl->xcl_name, cl->xcl_name) == 0)
99 list_add_tail(&xcl->xcl_list, &svc_xprt_class_list);
102 spin_unlock(&svc_xprt_class_lock);
105 EXPORT_SYMBOL_GPL(svc_reg_xprt_class);
107 void svc_unreg_xprt_class(struct svc_xprt_class *xcl)
109 dprintk("svc: Removing svc transport class '%s'\n", xcl->xcl_name);
110 spin_lock(&svc_xprt_class_lock);
111 list_del_init(&xcl->xcl_list);
112 spin_unlock(&svc_xprt_class_lock);
114 EXPORT_SYMBOL_GPL(svc_unreg_xprt_class);
117 * Format the transport list for printing
119 int svc_print_xprts(char *buf, int maxlen)
121 struct list_head *le;
126 spin_lock(&svc_xprt_class_lock);
127 list_for_each(le, &svc_xprt_class_list) {
129 struct svc_xprt_class *xcl =
130 list_entry(le, struct svc_xprt_class, xcl_list);
132 sprintf(tmpstr, "%s %d\n", xcl->xcl_name, xcl->xcl_max_payload);
133 slen = strlen(tmpstr);
134 if (len + slen > maxlen)
139 spin_unlock(&svc_xprt_class_lock);
144 static void svc_xprt_free(struct kref *kref)
146 struct svc_xprt *xprt =
147 container_of(kref, struct svc_xprt, xpt_ref);
148 struct module *owner = xprt->xpt_class->xcl_owner;
149 if (test_bit(XPT_CACHE_AUTH, &xprt->xpt_flags)
150 && xprt->xpt_auth_cache != NULL)
151 svcauth_unix_info_release(xprt->xpt_auth_cache);
152 xprt->xpt_ops->xpo_free(xprt);
156 void svc_xprt_put(struct svc_xprt *xprt)
158 kref_put(&xprt->xpt_ref, svc_xprt_free);
160 EXPORT_SYMBOL_GPL(svc_xprt_put);
163 * Called by transport drivers to initialize the transport independent
164 * portion of the transport instance.
166 void svc_xprt_init(struct svc_xprt_class *xcl, struct svc_xprt *xprt,
167 struct svc_serv *serv)
169 memset(xprt, 0, sizeof(*xprt));
170 xprt->xpt_class = xcl;
171 xprt->xpt_ops = xcl->xcl_ops;
172 kref_init(&xprt->xpt_ref);
173 xprt->xpt_server = serv;
174 INIT_LIST_HEAD(&xprt->xpt_list);
175 INIT_LIST_HEAD(&xprt->xpt_ready);
176 INIT_LIST_HEAD(&xprt->xpt_deferred);
177 mutex_init(&xprt->xpt_mutex);
178 spin_lock_init(&xprt->xpt_lock);
179 set_bit(XPT_BUSY, &xprt->xpt_flags);
181 EXPORT_SYMBOL_GPL(svc_xprt_init);
183 int svc_create_xprt(struct svc_serv *serv, char *xprt_name, unsigned short port,
186 struct svc_xprt_class *xcl;
187 struct sockaddr_in sin = {
188 .sin_family = AF_INET,
189 .sin_addr.s_addr = htonl(INADDR_ANY),
190 .sin_port = htons(port),
192 dprintk("svc: creating transport %s[%d]\n", xprt_name, port);
193 spin_lock(&svc_xprt_class_lock);
194 list_for_each_entry(xcl, &svc_xprt_class_list, xcl_list) {
195 struct svc_xprt *newxprt;
197 if (strcmp(xprt_name, xcl->xcl_name))
200 if (!try_module_get(xcl->xcl_owner))
203 spin_unlock(&svc_xprt_class_lock);
204 newxprt = xcl->xcl_ops->
205 xpo_create(serv, (struct sockaddr *)&sin, sizeof(sin),
207 if (IS_ERR(newxprt)) {
208 module_put(xcl->xcl_owner);
209 return PTR_ERR(newxprt);
212 clear_bit(XPT_TEMP, &newxprt->xpt_flags);
213 spin_lock_bh(&serv->sv_lock);
214 list_add(&newxprt->xpt_list, &serv->sv_permsocks);
215 spin_unlock_bh(&serv->sv_lock);
216 clear_bit(XPT_BUSY, &newxprt->xpt_flags);
217 return svc_xprt_local_port(newxprt);
220 spin_unlock(&svc_xprt_class_lock);
221 dprintk("svc: transport %s not found\n", xprt_name);
224 EXPORT_SYMBOL_GPL(svc_create_xprt);
227 * Copy the local and remote xprt addresses to the rqstp structure
229 void svc_xprt_copy_addrs(struct svc_rqst *rqstp, struct svc_xprt *xprt)
231 struct sockaddr *sin;
233 memcpy(&rqstp->rq_addr, &xprt->xpt_remote, xprt->xpt_remotelen);
234 rqstp->rq_addrlen = xprt->xpt_remotelen;
237 * Destination address in request is needed for binding the
238 * source address in RPC replies/callbacks later.
240 sin = (struct sockaddr *)&xprt->xpt_local;
241 switch (sin->sa_family) {
243 rqstp->rq_daddr.addr = ((struct sockaddr_in *)sin)->sin_addr;
246 rqstp->rq_daddr.addr6 = ((struct sockaddr_in6 *)sin)->sin6_addr;
250 EXPORT_SYMBOL_GPL(svc_xprt_copy_addrs);
253 * svc_print_addr - Format rq_addr field for printing
254 * @rqstp: svc_rqst struct containing address to print
255 * @buf: target buffer for formatted address
256 * @len: length of target buffer
259 char *svc_print_addr(struct svc_rqst *rqstp, char *buf, size_t len)
261 return __svc_print_addr(svc_addr(rqstp), buf, len);
263 EXPORT_SYMBOL_GPL(svc_print_addr);
266 * Queue up an idle server thread. Must have pool->sp_lock held.
267 * Note: this is really a stack rather than a queue, so that we only
268 * use as many different threads as we need, and the rest don't pollute
271 static void svc_thread_enqueue(struct svc_pool *pool, struct svc_rqst *rqstp)
273 list_add(&rqstp->rq_list, &pool->sp_threads);
277 * Dequeue an nfsd thread. Must have pool->sp_lock held.
279 static void svc_thread_dequeue(struct svc_pool *pool, struct svc_rqst *rqstp)
281 list_del(&rqstp->rq_list);
285 * Queue up a transport with data pending. If there are idle nfsd
286 * processes, wake 'em up.
289 void svc_xprt_enqueue(struct svc_xprt *xprt)
291 struct svc_serv *serv = xprt->xpt_server;
292 struct svc_pool *pool;
293 struct svc_rqst *rqstp;
296 if (!(xprt->xpt_flags &
297 ((1<<XPT_CONN)|(1<<XPT_DATA)|(1<<XPT_CLOSE)|(1<<XPT_DEFERRED))))
299 if (test_bit(XPT_DEAD, &xprt->xpt_flags))
303 pool = svc_pool_for_cpu(xprt->xpt_server, cpu);
306 spin_lock_bh(&pool->sp_lock);
308 if (!list_empty(&pool->sp_threads) &&
309 !list_empty(&pool->sp_sockets))
312 "threads and transports both waiting??\n");
314 if (test_bit(XPT_DEAD, &xprt->xpt_flags)) {
315 /* Don't enqueue dead transports */
316 dprintk("svc: transport %p is dead, not enqueued\n", xprt);
320 /* Mark transport as busy. It will remain in this state until
321 * the provider calls svc_xprt_received. We update XPT_BUSY
322 * atomically because it also guards against trying to enqueue
323 * the transport twice.
325 if (test_and_set_bit(XPT_BUSY, &xprt->xpt_flags)) {
326 /* Don't enqueue transport while already enqueued */
327 dprintk("svc: transport %p busy, not enqueued\n", xprt);
330 BUG_ON(xprt->xpt_pool != NULL);
331 xprt->xpt_pool = pool;
333 /* Handle pending connection */
334 if (test_bit(XPT_CONN, &xprt->xpt_flags))
337 /* Handle close in-progress */
338 if (test_bit(XPT_CLOSE, &xprt->xpt_flags))
341 /* Check if we have space to reply to a request */
342 if (!xprt->xpt_ops->xpo_has_wspace(xprt)) {
343 /* Don't enqueue while not enough space for reply */
344 dprintk("svc: no write space, transport %p not enqueued\n",
346 xprt->xpt_pool = NULL;
347 clear_bit(XPT_BUSY, &xprt->xpt_flags);
352 if (!list_empty(&pool->sp_threads)) {
353 rqstp = list_entry(pool->sp_threads.next,
356 dprintk("svc: transport %p served by daemon %p\n",
358 svc_thread_dequeue(pool, rqstp);
361 "svc_xprt_enqueue: server %p, rq_xprt=%p!\n",
362 rqstp, rqstp->rq_xprt);
363 rqstp->rq_xprt = xprt;
365 rqstp->rq_reserved = serv->sv_max_mesg;
366 atomic_add(rqstp->rq_reserved, &xprt->xpt_reserved);
367 BUG_ON(xprt->xpt_pool != pool);
368 wake_up(&rqstp->rq_wait);
370 dprintk("svc: transport %p put into queue\n", xprt);
371 list_add_tail(&xprt->xpt_ready, &pool->sp_sockets);
372 BUG_ON(xprt->xpt_pool != pool);
376 spin_unlock_bh(&pool->sp_lock);
378 EXPORT_SYMBOL_GPL(svc_xprt_enqueue);
381 * Dequeue the first transport. Must be called with the pool->sp_lock held.
383 static struct svc_xprt *svc_xprt_dequeue(struct svc_pool *pool)
385 struct svc_xprt *xprt;
387 if (list_empty(&pool->sp_sockets))
390 xprt = list_entry(pool->sp_sockets.next,
391 struct svc_xprt, xpt_ready);
392 list_del_init(&xprt->xpt_ready);
394 dprintk("svc: transport %p dequeued, inuse=%d\n",
395 xprt, atomic_read(&xprt->xpt_ref.refcount));
401 * svc_xprt_received conditionally queues the transport for processing
402 * by another thread. The caller must hold the XPT_BUSY bit and must
403 * not thereafter touch transport data.
405 * Note: XPT_DATA only gets cleared when a read-attempt finds no (or
406 * insufficient) data.
408 void svc_xprt_received(struct svc_xprt *xprt)
410 BUG_ON(!test_bit(XPT_BUSY, &xprt->xpt_flags));
411 xprt->xpt_pool = NULL;
412 clear_bit(XPT_BUSY, &xprt->xpt_flags);
413 svc_xprt_enqueue(xprt);
415 EXPORT_SYMBOL_GPL(svc_xprt_received);
418 * svc_reserve - change the space reserved for the reply to a request.
419 * @rqstp: The request in question
420 * @space: new max space to reserve
422 * Each request reserves some space on the output queue of the transport
423 * to make sure the reply fits. This function reduces that reserved
424 * space to be the amount of space used already, plus @space.
427 void svc_reserve(struct svc_rqst *rqstp, int space)
429 space += rqstp->rq_res.head[0].iov_len;
431 if (space < rqstp->rq_reserved) {
432 struct svc_xprt *xprt = rqstp->rq_xprt;
433 atomic_sub((rqstp->rq_reserved - space), &xprt->xpt_reserved);
434 rqstp->rq_reserved = space;
436 svc_xprt_enqueue(xprt);
439 EXPORT_SYMBOL(svc_reserve);
441 static void svc_xprt_release(struct svc_rqst *rqstp)
443 struct svc_xprt *xprt = rqstp->rq_xprt;
445 rqstp->rq_xprt->xpt_ops->xpo_release_rqst(rqstp);
447 svc_free_res_pages(rqstp);
448 rqstp->rq_res.page_len = 0;
449 rqstp->rq_res.page_base = 0;
451 /* Reset response buffer and release
453 * But first, check that enough space was reserved
454 * for the reply, otherwise we have a bug!
456 if ((rqstp->rq_res.len) > rqstp->rq_reserved)
457 printk(KERN_ERR "RPC request reserved %d but used %d\n",
461 rqstp->rq_res.head[0].iov_len = 0;
462 svc_reserve(rqstp, 0);
463 rqstp->rq_xprt = NULL;
469 * External function to wake up a server waiting for data
470 * This really only makes sense for services like lockd
471 * which have exactly one thread anyway.
473 void svc_wake_up(struct svc_serv *serv)
475 struct svc_rqst *rqstp;
477 struct svc_pool *pool;
479 for (i = 0; i < serv->sv_nrpools; i++) {
480 pool = &serv->sv_pools[i];
482 spin_lock_bh(&pool->sp_lock);
483 if (!list_empty(&pool->sp_threads)) {
484 rqstp = list_entry(pool->sp_threads.next,
487 dprintk("svc: daemon %p woken up.\n", rqstp);
489 svc_thread_dequeue(pool, rqstp);
490 rqstp->rq_xprt = NULL;
492 wake_up(&rqstp->rq_wait);
494 spin_unlock_bh(&pool->sp_lock);
497 EXPORT_SYMBOL(svc_wake_up);
499 int svc_port_is_privileged(struct sockaddr *sin)
501 switch (sin->sa_family) {
503 return ntohs(((struct sockaddr_in *)sin)->sin_port)
506 return ntohs(((struct sockaddr_in6 *)sin)->sin6_port)
514 * Make sure that we don't have too many active connections. If we
515 * have, something must be dropped.
517 * There's no point in trying to do random drop here for DoS
518 * prevention. The NFS clients does 1 reconnect in 15 seconds. An
519 * attacker can easily beat that.
521 * The only somewhat efficient mechanism would be if drop old
522 * connections from the same IP first. But right now we don't even
523 * record the client IP in svc_sock.
525 static void svc_check_conn_limits(struct svc_serv *serv)
527 if (serv->sv_tmpcnt > (serv->sv_nrthreads+3)*20) {
528 struct svc_xprt *xprt = NULL;
529 spin_lock_bh(&serv->sv_lock);
530 if (!list_empty(&serv->sv_tempsocks)) {
531 if (net_ratelimit()) {
532 /* Try to help the admin */
533 printk(KERN_NOTICE "%s: too many open "
534 "connections, consider increasing the "
535 "number of nfsd threads\n",
539 * Always select the oldest connection. It's not fair,
542 xprt = list_entry(serv->sv_tempsocks.prev,
545 set_bit(XPT_CLOSE, &xprt->xpt_flags);
548 spin_unlock_bh(&serv->sv_lock);
551 svc_xprt_enqueue(xprt);
558 * Receive the next request on any transport. This code is carefully
559 * organised not to touch any cachelines in the shared svc_serv
560 * structure, only cachelines in the local svc_pool.
562 int svc_recv(struct svc_rqst *rqstp, long timeout)
564 struct svc_xprt *xprt = NULL;
565 struct svc_serv *serv = rqstp->rq_server;
566 struct svc_pool *pool = rqstp->rq_pool;
570 DECLARE_WAITQUEUE(wait, current);
572 dprintk("svc: server %p waiting for data (to = %ld)\n",
577 "svc_recv: service %p, transport not NULL!\n",
579 if (waitqueue_active(&rqstp->rq_wait))
581 "svc_recv: service %p, wait queue active!\n",
584 /* now allocate needed pages. If we get a failure, sleep briefly */
585 pages = (serv->sv_max_mesg + PAGE_SIZE) / PAGE_SIZE;
586 for (i = 0; i < pages ; i++)
587 while (rqstp->rq_pages[i] == NULL) {
588 struct page *p = alloc_page(GFP_KERNEL);
590 set_current_state(TASK_INTERRUPTIBLE);
591 if (signalled() || kthread_should_stop()) {
592 set_current_state(TASK_RUNNING);
595 schedule_timeout(msecs_to_jiffies(500));
597 rqstp->rq_pages[i] = p;
599 rqstp->rq_pages[i++] = NULL; /* this might be seen in nfs_read_actor */
600 BUG_ON(pages >= RPCSVC_MAXPAGES);
602 /* Make arg->head point to first page and arg->pages point to rest */
603 arg = &rqstp->rq_arg;
604 arg->head[0].iov_base = page_address(rqstp->rq_pages[0]);
605 arg->head[0].iov_len = PAGE_SIZE;
606 arg->pages = rqstp->rq_pages + 1;
608 /* save at least one page for response */
609 arg->page_len = (pages-2)*PAGE_SIZE;
610 arg->len = (pages-1)*PAGE_SIZE;
611 arg->tail[0].iov_len = 0;
615 if (signalled() || kthread_should_stop())
618 spin_lock_bh(&pool->sp_lock);
619 xprt = svc_xprt_dequeue(pool);
621 rqstp->rq_xprt = xprt;
623 rqstp->rq_reserved = serv->sv_max_mesg;
624 atomic_add(rqstp->rq_reserved, &xprt->xpt_reserved);
626 /* No data pending. Go to sleep */
627 svc_thread_enqueue(pool, rqstp);
630 * We have to be able to interrupt this wait
631 * to bring down the daemons ...
633 set_current_state(TASK_INTERRUPTIBLE);
636 * checking kthread_should_stop() here allows us to avoid
637 * locking and signalling when stopping kthreads that call
638 * svc_recv. If the thread has already been woken up, then
639 * we can exit here without sleeping. If not, then it
640 * it'll be woken up quickly during the schedule_timeout
642 if (kthread_should_stop()) {
643 set_current_state(TASK_RUNNING);
644 spin_unlock_bh(&pool->sp_lock);
648 add_wait_queue(&rqstp->rq_wait, &wait);
649 spin_unlock_bh(&pool->sp_lock);
651 schedule_timeout(timeout);
655 spin_lock_bh(&pool->sp_lock);
656 remove_wait_queue(&rqstp->rq_wait, &wait);
658 xprt = rqstp->rq_xprt;
660 svc_thread_dequeue(pool, rqstp);
661 spin_unlock_bh(&pool->sp_lock);
662 dprintk("svc: server %p, no data yet\n", rqstp);
663 if (signalled() || kthread_should_stop())
669 spin_unlock_bh(&pool->sp_lock);
672 if (test_bit(XPT_CLOSE, &xprt->xpt_flags)) {
673 dprintk("svc_recv: found XPT_CLOSE\n");
674 svc_delete_xprt(xprt);
675 } else if (test_bit(XPT_LISTENER, &xprt->xpt_flags)) {
676 struct svc_xprt *newxpt;
677 newxpt = xprt->xpt_ops->xpo_accept(xprt);
680 * We know this module_get will succeed because the
681 * listener holds a reference too
683 __module_get(newxpt->xpt_class->xcl_owner);
684 svc_check_conn_limits(xprt->xpt_server);
685 spin_lock_bh(&serv->sv_lock);
686 set_bit(XPT_TEMP, &newxpt->xpt_flags);
687 list_add(&newxpt->xpt_list, &serv->sv_tempsocks);
689 if (serv->sv_temptimer.function == NULL) {
690 /* setup timer to age temp transports */
691 setup_timer(&serv->sv_temptimer,
693 (unsigned long)serv);
694 mod_timer(&serv->sv_temptimer,
695 jiffies + svc_conn_age_period * HZ);
697 spin_unlock_bh(&serv->sv_lock);
698 svc_xprt_received(newxpt);
700 svc_xprt_received(xprt);
702 dprintk("svc: server %p, pool %u, transport %p, inuse=%d\n",
703 rqstp, pool->sp_id, xprt,
704 atomic_read(&xprt->xpt_ref.refcount));
705 rqstp->rq_deferred = svc_deferred_dequeue(xprt);
706 if (rqstp->rq_deferred) {
707 svc_xprt_received(xprt);
708 len = svc_deferred_recv(rqstp);
710 len = xprt->xpt_ops->xpo_recvfrom(rqstp);
711 dprintk("svc: got len=%d\n", len);
714 /* No data, incomplete (TCP) read, or accept() */
715 if (len == 0 || len == -EAGAIN) {
716 rqstp->rq_res.len = 0;
717 svc_xprt_release(rqstp);
720 clear_bit(XPT_OLD, &xprt->xpt_flags);
722 rqstp->rq_secure = svc_port_is_privileged(svc_addr(rqstp));
723 rqstp->rq_chandle.defer = svc_defer;
726 serv->sv_stats->netcnt++;
729 EXPORT_SYMBOL(svc_recv);
734 void svc_drop(struct svc_rqst *rqstp)
736 dprintk("svc: xprt %p dropped request\n", rqstp->rq_xprt);
737 svc_xprt_release(rqstp);
739 EXPORT_SYMBOL(svc_drop);
742 * Return reply to client.
744 int svc_send(struct svc_rqst *rqstp)
746 struct svc_xprt *xprt;
750 xprt = rqstp->rq_xprt;
754 /* release the receive skb before sending the reply */
755 rqstp->rq_xprt->xpt_ops->xpo_release_rqst(rqstp);
757 /* calculate over-all length */
759 xb->len = xb->head[0].iov_len +
763 /* Grab mutex to serialize outgoing data. */
764 mutex_lock(&xprt->xpt_mutex);
765 if (test_bit(XPT_DEAD, &xprt->xpt_flags))
768 len = xprt->xpt_ops->xpo_sendto(rqstp);
769 mutex_unlock(&xprt->xpt_mutex);
770 svc_xprt_release(rqstp);
772 if (len == -ECONNREFUSED || len == -ENOTCONN || len == -EAGAIN)
778 * Timer function to close old temporary transports, using
779 * a mark-and-sweep algorithm.
781 static void svc_age_temp_xprts(unsigned long closure)
783 struct svc_serv *serv = (struct svc_serv *)closure;
784 struct svc_xprt *xprt;
785 struct list_head *le, *next;
786 LIST_HEAD(to_be_aged);
788 dprintk("svc_age_temp_xprts\n");
790 if (!spin_trylock_bh(&serv->sv_lock)) {
791 /* busy, try again 1 sec later */
792 dprintk("svc_age_temp_xprts: busy\n");
793 mod_timer(&serv->sv_temptimer, jiffies + HZ);
797 list_for_each_safe(le, next, &serv->sv_tempsocks) {
798 xprt = list_entry(le, struct svc_xprt, xpt_list);
800 /* First time through, just mark it OLD. Second time
801 * through, close it. */
802 if (!test_and_set_bit(XPT_OLD, &xprt->xpt_flags))
804 if (atomic_read(&xprt->xpt_ref.refcount) > 1
805 || test_bit(XPT_BUSY, &xprt->xpt_flags))
808 list_move(le, &to_be_aged);
809 set_bit(XPT_CLOSE, &xprt->xpt_flags);
810 set_bit(XPT_DETACHED, &xprt->xpt_flags);
812 spin_unlock_bh(&serv->sv_lock);
814 while (!list_empty(&to_be_aged)) {
815 le = to_be_aged.next;
816 /* fiddling the xpt_list node is safe 'cos we're XPT_DETACHED */
818 xprt = list_entry(le, struct svc_xprt, xpt_list);
820 dprintk("queuing xprt %p for closing\n", xprt);
822 /* a thread will dequeue and close it soon */
823 svc_xprt_enqueue(xprt);
827 mod_timer(&serv->sv_temptimer, jiffies + svc_conn_age_period * HZ);
831 * Remove a dead transport
833 void svc_delete_xprt(struct svc_xprt *xprt)
835 struct svc_serv *serv = xprt->xpt_server;
837 dprintk("svc: svc_delete_xprt(%p)\n", xprt);
838 xprt->xpt_ops->xpo_detach(xprt);
840 spin_lock_bh(&serv->sv_lock);
841 if (!test_and_set_bit(XPT_DETACHED, &xprt->xpt_flags))
842 list_del_init(&xprt->xpt_list);
844 * We used to delete the transport from whichever list
845 * it's sk_xprt.xpt_ready node was on, but we don't actually
846 * need to. This is because the only time we're called
847 * while still attached to a queue, the queue itself
848 * is about to be destroyed (in svc_destroy).
850 if (!test_and_set_bit(XPT_DEAD, &xprt->xpt_flags)) {
851 BUG_ON(atomic_read(&xprt->xpt_ref.refcount) < 2);
852 if (test_bit(XPT_TEMP, &xprt->xpt_flags))
856 spin_unlock_bh(&serv->sv_lock);
859 void svc_close_xprt(struct svc_xprt *xprt)
861 set_bit(XPT_CLOSE, &xprt->xpt_flags);
862 if (test_and_set_bit(XPT_BUSY, &xprt->xpt_flags))
863 /* someone else will have to effect the close */
867 svc_delete_xprt(xprt);
868 clear_bit(XPT_BUSY, &xprt->xpt_flags);
871 EXPORT_SYMBOL_GPL(svc_close_xprt);
873 void svc_close_all(struct list_head *xprt_list)
875 struct svc_xprt *xprt;
876 struct svc_xprt *tmp;
878 list_for_each_entry_safe(xprt, tmp, xprt_list, xpt_list) {
879 set_bit(XPT_CLOSE, &xprt->xpt_flags);
880 if (test_bit(XPT_BUSY, &xprt->xpt_flags)) {
881 /* Waiting to be processed, but no threads left,
882 * So just remove it from the waiting list
884 list_del_init(&xprt->xpt_ready);
885 clear_bit(XPT_BUSY, &xprt->xpt_flags);
887 svc_close_xprt(xprt);
892 * Handle defer and revisit of requests
895 static void svc_revisit(struct cache_deferred_req *dreq, int too_many)
897 struct svc_deferred_req *dr =
898 container_of(dreq, struct svc_deferred_req, handle);
899 struct svc_xprt *xprt = dr->xprt;
906 dprintk("revisit queued\n");
908 spin_lock(&xprt->xpt_lock);
909 list_add(&dr->handle.recent, &xprt->xpt_deferred);
910 spin_unlock(&xprt->xpt_lock);
911 set_bit(XPT_DEFERRED, &xprt->xpt_flags);
912 svc_xprt_enqueue(xprt);
917 * Save the request off for later processing. The request buffer looks
920 * <xprt-header><rpc-header><rpc-pagelist><rpc-tail>
922 * This code can only handle requests that consist of an xprt-header
925 static struct cache_deferred_req *svc_defer(struct cache_req *req)
927 struct svc_rqst *rqstp = container_of(req, struct svc_rqst, rq_chandle);
928 struct svc_deferred_req *dr;
930 if (rqstp->rq_arg.page_len)
931 return NULL; /* if more than a page, give up FIXME */
932 if (rqstp->rq_deferred) {
933 dr = rqstp->rq_deferred;
934 rqstp->rq_deferred = NULL;
938 /* FIXME maybe discard if size too large */
939 size = sizeof(struct svc_deferred_req) + rqstp->rq_arg.len;
940 dr = kmalloc(size, GFP_KERNEL);
944 dr->handle.owner = rqstp->rq_server;
945 dr->prot = rqstp->rq_prot;
946 memcpy(&dr->addr, &rqstp->rq_addr, rqstp->rq_addrlen);
947 dr->addrlen = rqstp->rq_addrlen;
948 dr->daddr = rqstp->rq_daddr;
949 dr->argslen = rqstp->rq_arg.len >> 2;
950 dr->xprt_hlen = rqstp->rq_xprt_hlen;
952 /* back up head to the start of the buffer and copy */
953 skip = rqstp->rq_arg.len - rqstp->rq_arg.head[0].iov_len;
954 memcpy(dr->args, rqstp->rq_arg.head[0].iov_base - skip,
957 svc_xprt_get(rqstp->rq_xprt);
958 dr->xprt = rqstp->rq_xprt;
960 dr->handle.revisit = svc_revisit;
965 * recv data from a deferred request into an active one
967 static int svc_deferred_recv(struct svc_rqst *rqstp)
969 struct svc_deferred_req *dr = rqstp->rq_deferred;
971 /* setup iov_base past transport header */
972 rqstp->rq_arg.head[0].iov_base = dr->args + (dr->xprt_hlen>>2);
973 /* The iov_len does not include the transport header bytes */
974 rqstp->rq_arg.head[0].iov_len = (dr->argslen<<2) - dr->xprt_hlen;
975 rqstp->rq_arg.page_len = 0;
976 /* The rq_arg.len includes the transport header bytes */
977 rqstp->rq_arg.len = dr->argslen<<2;
978 rqstp->rq_prot = dr->prot;
979 memcpy(&rqstp->rq_addr, &dr->addr, dr->addrlen);
980 rqstp->rq_addrlen = dr->addrlen;
981 /* Save off transport header len in case we get deferred again */
982 rqstp->rq_xprt_hlen = dr->xprt_hlen;
983 rqstp->rq_daddr = dr->daddr;
984 rqstp->rq_respages = rqstp->rq_pages;
985 return (dr->argslen<<2) - dr->xprt_hlen;
989 static struct svc_deferred_req *svc_deferred_dequeue(struct svc_xprt *xprt)
991 struct svc_deferred_req *dr = NULL;
993 if (!test_bit(XPT_DEFERRED, &xprt->xpt_flags))
995 spin_lock(&xprt->xpt_lock);
996 clear_bit(XPT_DEFERRED, &xprt->xpt_flags);
997 if (!list_empty(&xprt->xpt_deferred)) {
998 dr = list_entry(xprt->xpt_deferred.next,
999 struct svc_deferred_req,
1001 list_del_init(&dr->handle.recent);
1002 set_bit(XPT_DEFERRED, &xprt->xpt_flags);
1004 spin_unlock(&xprt->xpt_lock);
1009 * Return the transport instance pointer for the endpoint accepting
1010 * connections/peer traffic from the specified transport class,
1011 * address family and port.
1013 * Specifying 0 for the address family or port is effectively a
1014 * wild-card, and will result in matching the first transport in the
1015 * service's list that has a matching class name.
1017 struct svc_xprt *svc_find_xprt(struct svc_serv *serv, char *xcl_name,
1020 struct svc_xprt *xprt;
1021 struct svc_xprt *found = NULL;
1023 /* Sanity check the args */
1024 if (!serv || !xcl_name)
1027 spin_lock_bh(&serv->sv_lock);
1028 list_for_each_entry(xprt, &serv->sv_permsocks, xpt_list) {
1029 if (strcmp(xprt->xpt_class->xcl_name, xcl_name))
1031 if (af != AF_UNSPEC && af != xprt->xpt_local.ss_family)
1033 if (port && port != svc_xprt_local_port(xprt))
1039 spin_unlock_bh(&serv->sv_lock);
1042 EXPORT_SYMBOL_GPL(svc_find_xprt);
1045 * Format a buffer with a list of the active transports. A zero for
1046 * the buflen parameter disables target buffer overflow checking.
1048 int svc_xprt_names(struct svc_serv *serv, char *buf, int buflen)
1050 struct svc_xprt *xprt;
1055 /* Sanity check args */
1059 spin_lock_bh(&serv->sv_lock);
1060 list_for_each_entry(xprt, &serv->sv_permsocks, xpt_list) {
1061 len = snprintf(xprt_str, sizeof(xprt_str),
1062 "%s %d\n", xprt->xpt_class->xcl_name,
1063 svc_xprt_local_port(xprt));
1064 /* If the string was truncated, replace with error string */
1065 if (len >= sizeof(xprt_str))
1066 strcpy(xprt_str, "name-too-long\n");
1067 /* Don't overflow buffer */
1068 len = strlen(xprt_str);
1069 if (buflen && (len + totlen >= buflen))
1071 strcpy(buf+totlen, xprt_str);
1074 spin_unlock_bh(&serv->sv_lock);
1077 EXPORT_SYMBOL_GPL(svc_xprt_names);