svc_xprt.c 38.7 KB
Newer Older
Tom Tucker's avatar
Tom Tucker committed
1 2 3 4 5 6 7 8 9
/*
 * linux/net/sunrpc/svc_xprt.c
 *
 * Author: Tom Tucker <tom@opengridcomputing.com>
 */

#include <linux/sched.h>
#include <linux/errno.h>
#include <linux/freezer.h>
10
#include <linux/kthread.h>
11
#include <linux/slab.h>
Tom Tucker's avatar
Tom Tucker committed
12
#include <net/sock.h>
13
#include <linux/sunrpc/addr.h>
Tom Tucker's avatar
Tom Tucker committed
14 15
#include <linux/sunrpc/stats.h>
#include <linux/sunrpc/svc_xprt.h>
16
#include <linux/sunrpc/svcsock.h>
17
#include <linux/sunrpc/xprt.h>
18
#include <linux/module.h>
19
#include <linux/netdevice.h>
20
#include <trace/events/sunrpc.h>
Tom Tucker's avatar
Tom Tucker committed
21 22 23

#define RPCDBG_FACILITY	RPCDBG_SVCXPRT

24 25 26 27
static unsigned int svc_rpc_per_connection_limit __read_mostly;
module_param(svc_rpc_per_connection_limit, uint, 0644);


28 29 30 31
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);
32
static void svc_delete_xprt(struct svc_xprt *xprt);
33 34 35 36 37 38 39 40

/* 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;

Tom Tucker's avatar
Tom Tucker committed
41 42 43 44
/* List of registered transport classes */
static DEFINE_SPINLOCK(svc_xprt_class_lock);
static LIST_HEAD(svc_xprt_class_list);

45 46 47 48 49
/* 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.
50
 *	The "service mutex" protects svc_serv->sv_nrthread.
51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74
 *	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.
 */
Tom Tucker's avatar
Tom Tucker committed
75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105
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);

106 107 108 109 110
/*
 * Format the transport list for printing
 */
int svc_print_xprts(char *buf, int maxlen)
{
111
	struct svc_xprt_class *xcl;
112 113 114 115 116
	char tmpstr[80];
	int len = 0;
	buf[0] = '\0';

	spin_lock(&svc_xprt_class_lock);
117
	list_for_each_entry(xcl, &svc_xprt_class_list, xcl_list) {
118 119 120 121 122 123 124 125 126 127 128 129 130 131
		int slen;

		sprintf(tmpstr, "%s %d\n", xcl->xcl_name, xcl->xcl_max_payload);
		slen = strlen(tmpstr);
		if (len + slen > maxlen)
			break;
		len += slen;
		strcat(buf, tmpstr);
	}
	spin_unlock(&svc_xprt_class_lock);

	return len;
}

Tom Tucker's avatar
Tom Tucker committed
132 133 134 135 136
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;
137 138
	if (test_bit(XPT_CACHE_AUTH, &xprt->xpt_flags))
		svcauth_unix_info_release(xprt);
139
	put_net(xprt->xpt_net);
140 141 142
	/* See comment on corresponding get in xs_setup_bc_tcp(): */
	if (xprt->xpt_bc_xprt)
		xprt_put(xprt->xpt_bc_xprt);
143 144
	if (xprt->xpt_bc_xps)
		xprt_switch_put(xprt->xpt_bc_xps);
Tom Tucker's avatar
Tom Tucker committed
145 146 147 148 149 150 151 152 153 154
	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);

Tom Tucker's avatar
Tom Tucker committed
155 156 157 158
/*
 * Called by transport drivers to initialize the transport independent
 * portion of the transport instance.
 */
159 160
void svc_xprt_init(struct net *net, struct svc_xprt_class *xcl,
		   struct svc_xprt *xprt, struct svc_serv *serv)
Tom Tucker's avatar
Tom Tucker committed
161 162 163 164
{
	memset(xprt, 0, sizeof(*xprt));
	xprt->xpt_class = xcl;
	xprt->xpt_ops = xcl->xcl_ops;
Tom Tucker's avatar
Tom Tucker committed
165
	kref_init(&xprt->xpt_ref);
166
	xprt->xpt_server = serv;
167 168
	INIT_LIST_HEAD(&xprt->xpt_list);
	INIT_LIST_HEAD(&xprt->xpt_ready);
169
	INIT_LIST_HEAD(&xprt->xpt_deferred);
170
	INIT_LIST_HEAD(&xprt->xpt_users);
171
	mutex_init(&xprt->xpt_mutex);
172
	spin_lock_init(&xprt->xpt_lock);
173
	set_bit(XPT_BUSY, &xprt->xpt_flags);
174
	rpc_init_wait_queue(&xprt->xpt_bc_pending, "xpt_bc_pending");
175
	xprt->xpt_net = get_net(net);
Tom Tucker's avatar
Tom Tucker committed
176 177
}
EXPORT_SYMBOL_GPL(svc_xprt_init);
178

179 180
static struct svc_xprt *__svc_xpo_create(struct svc_xprt_class *xcl,
					 struct svc_serv *serv,
181
					 struct net *net,
182 183 184
					 const int family,
					 const unsigned short port,
					 int flags)
185 186 187
{
	struct sockaddr_in sin = {
		.sin_family		= AF_INET,
Al Viro's avatar
Al Viro committed
188
		.sin_addr.s_addr	= htonl(INADDR_ANY),
189 190
		.sin_port		= htons(port),
	};
191
#if IS_ENABLED(CONFIG_IPV6)
192 193 194 195 196
	struct sockaddr_in6 sin6 = {
		.sin6_family		= AF_INET6,
		.sin6_addr		= IN6ADDR_ANY_INIT,
		.sin6_port		= htons(port),
	};
197
#endif
198 199 200
	struct sockaddr *sap;
	size_t len;

201 202
	switch (family) {
	case PF_INET:
203 204 205
		sap = (struct sockaddr *)&sin;
		len = sizeof(sin);
		break;
206
#if IS_ENABLED(CONFIG_IPV6)
207
	case PF_INET6:
208 209 210
		sap = (struct sockaddr *)&sin6;
		len = sizeof(sin6);
		break;
211
#endif
212 213 214 215
	default:
		return ERR_PTR(-EAFNOSUPPORT);
	}

216
	return xcl->xcl_ops->xpo_create(serv, net, sap, len, flags);
217 218
}

219 220 221 222 223 224 225 226 227 228
/*
 * 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.
 */
static void svc_xprt_received(struct svc_xprt *xprt)
{
229 230
	if (!test_bit(XPT_BUSY, &xprt->xpt_flags)) {
		WARN_ONCE(1, "xprt=0x%p already busy!", xprt);
231
		return;
232 233
	}

234
	/* As soon as we clear busy, the xprt could be closed and
235
	 * 'put', so we need a reference to call svc_enqueue_xprt with:
236 237
	 */
	svc_xprt_get(xprt);
238
	smp_mb__before_atomic();
239
	clear_bit(XPT_BUSY, &xprt->xpt_flags);
240
	xprt->xpt_server->sv_ops->svo_enqueue_xprt(xprt);
241 242 243
	svc_xprt_put(xprt);
}

244 245 246 247 248 249 250 251 252
void svc_add_new_perm_xprt(struct svc_serv *serv, struct svc_xprt *new)
{
	clear_bit(XPT_TEMP, &new->xpt_flags);
	spin_lock_bh(&serv->sv_lock);
	list_add(&new->xpt_list, &serv->sv_permsocks);
	spin_unlock_bh(&serv->sv_lock);
	svc_xprt_received(new);
}

J. Bruce Fields's avatar
J. Bruce Fields committed
253
int _svc_create_xprt(struct svc_serv *serv, const char *xprt_name,
254 255
		    struct net *net, const int family,
		    const unsigned short port, int flags)
256 257 258
{
	struct svc_xprt_class *xcl;

259 260
	spin_lock(&svc_xprt_class_lock);
	list_for_each_entry(xcl, &svc_xprt_class_list, xcl_list) {
261
		struct svc_xprt *newxprt;
262
		unsigned short newport;
263 264 265 266 267 268 269 270

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

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

		spin_unlock(&svc_xprt_class_lock);
271
		newxprt = __svc_xpo_create(xcl, serv, net, family, port, flags);
272 273 274
		if (IS_ERR(newxprt)) {
			module_put(xcl->xcl_owner);
			return PTR_ERR(newxprt);
275
		}
276
		svc_add_new_perm_xprt(serv, newxprt);
277 278
		newport = svc_xprt_local_port(newxprt);
		return newport;
279
	}
280
 err:
281
	spin_unlock(&svc_xprt_class_lock);
282 283 284
	/* This errno is exposed to user space.  Provide a reasonable
	 * perror msg for a bad transport. */
	return -EPROTONOSUPPORT;
285
}
J. Bruce Fields's avatar
J. Bruce Fields committed
286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303

int svc_create_xprt(struct svc_serv *serv, const char *xprt_name,
		    struct net *net, const int family,
		    const unsigned short port, int flags)
{
	int err;

	dprintk("svc: creating transport %s[%d]\n", xprt_name, port);
	err = _svc_create_xprt(serv, xprt_name, net, family, port, flags);
	if (err == -EPROTONOSUPPORT) {
		request_module("svc%s", xprt_name);
		err = _svc_create_xprt(serv, xprt_name, net, family, port, flags);
	}
	if (err)
		dprintk("svc: transport %s not found, err %d\n",
			xprt_name, err);
	return err;
}
304
EXPORT_SYMBOL_GPL(svc_create_xprt);
305 306 307 308 309 310 311 312 313 314 315 316 317

/*
 * Copy the local and remote xprt addresses to the rqstp structure
 */
void svc_xprt_copy_addrs(struct svc_rqst *rqstp, struct svc_xprt *xprt)
{
	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.
	 */
318 319
	memcpy(&rqstp->rq_daddr, &xprt->xpt_local, xprt->xpt_locallen);
	rqstp->rq_daddrlen = xprt->xpt_locallen;
320 321 322
}
EXPORT_SYMBOL_GPL(svc_xprt_copy_addrs);

323 324 325 326 327 328 329 330 331 332 333 334 335
/**
 * 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);

336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363
static bool svc_xprt_slots_in_range(struct svc_xprt *xprt)
{
	unsigned int limit = svc_rpc_per_connection_limit;
	int nrqsts = atomic_read(&xprt->xpt_nr_rqsts);

	return limit == 0 || (nrqsts >= 0 && nrqsts < limit);
}

static bool svc_xprt_reserve_slot(struct svc_rqst *rqstp, struct svc_xprt *xprt)
{
	if (!test_bit(RQ_DATA, &rqstp->rq_flags)) {
		if (!svc_xprt_slots_in_range(xprt))
			return false;
		atomic_inc(&xprt->xpt_nr_rqsts);
		set_bit(RQ_DATA, &rqstp->rq_flags);
	}
	return true;
}

static void svc_xprt_release_slot(struct svc_rqst *rqstp)
{
	struct svc_xprt	*xprt = rqstp->rq_xprt;
	if (test_and_clear_bit(RQ_DATA, &rqstp->rq_flags)) {
		atomic_dec(&xprt->xpt_nr_rqsts);
		svc_xprt_enqueue(xprt);
	}
}

364 365 366 367
static bool svc_xprt_has_something_to_do(struct svc_xprt *xprt)
{
	if (xprt->xpt_flags & ((1<<XPT_CONN)|(1<<XPT_CLOSE)))
		return true;
368
	if (xprt->xpt_flags & ((1<<XPT_DATA)|(1<<XPT_DEFERRED))) {
369 370
		if (xprt->xpt_ops->xpo_has_wspace(xprt) &&
		    svc_xprt_slots_in_range(xprt))
371 372 373 374
			return true;
		trace_svc_xprt_no_write_space(xprt);
		return false;
	}
375 376 377
	return false;
}

378
void svc_xprt_do_enqueue(struct svc_xprt *xprt)
379 380
{
	struct svc_pool *pool;
381
	struct svc_rqst	*rqstp = NULL;
382
	int cpu;
383
	bool queued = false;
384

385
	if (!svc_xprt_has_something_to_do(xprt))
386
		goto out;
387 388 389 390 391 392 393 394 395

	/* 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);
396
		goto out;
397 398
	}

399 400 401
	cpu = get_cpu();
	pool = svc_pool_for_cpu(xprt->xpt_server, cpu);

402
	atomic_long_inc(&pool->sp_stats.packets);
403

404 405 406 407 408 409 410 411 412 413 414 415 416
redo_search:
	/* find a thread for this xprt */
	rcu_read_lock();
	list_for_each_entry_rcu(rqstp, &pool->sp_all_threads, rq_all) {
		/* Do a lockless check first */
		if (test_bit(RQ_BUSY, &rqstp->rq_flags))
			continue;

		/*
		 * Once the xprt has been queued, it can only be dequeued by
		 * the task that intends to service it. All we can do at that
		 * point is to try to wake this thread back up so that it can
		 * do so.
417
		 */
418 419 420 421 422 423 424 425 426 427 428 429 430 431 432
		if (!queued) {
			spin_lock_bh(&rqstp->rq_lock);
			if (test_and_set_bit(RQ_BUSY, &rqstp->rq_flags)) {
				/* already busy, move on... */
				spin_unlock_bh(&rqstp->rq_lock);
				continue;
			}

			/* this one will do */
			rqstp->rq_xprt = xprt;
			svc_xprt_get(xprt);
			spin_unlock_bh(&rqstp->rq_lock);
		}
		rcu_read_unlock();

433
		atomic_long_inc(&pool->sp_stats.threads_woken);
434 435
		wake_up_process(rqstp->rq_task);
		put_cpu();
436
		goto out;
437 438 439 440 441 442 443 444 445 446 447
	}
	rcu_read_unlock();

	/*
	 * We didn't find an idle thread to use, so we need to queue the xprt.
	 * Do so and then search again. If we find one, we can't hook this one
	 * up to it directly but we can wake the thread up in the hopes that it
	 * will pick it up once it searches for a xprt to service.
	 */
	if (!queued) {
		queued = true;
448
		dprintk("svc: transport %p put into queue\n", xprt);
449
		spin_lock_bh(&pool->sp_lock);
450
		list_add_tail(&xprt->xpt_ready, &pool->sp_sockets);
451
		pool->sp_stats.sockets_queued++;
452 453
		spin_unlock_bh(&pool->sp_lock);
		goto redo_search;
454
	}
455
	rqstp = NULL;
456
	put_cpu();
457 458
out:
	trace_svc_xprt_do_enqueue(xprt, rqstp);
459
}
460
EXPORT_SYMBOL_GPL(svc_xprt_do_enqueue);
461 462 463 464 465 466 467 468 469 470

/*
 * Queue up a transport with data pending. If there are idle nfsd
 * processes, wake 'em up.
 *
 */
void svc_xprt_enqueue(struct svc_xprt *xprt)
{
	if (test_bit(XPT_BUSY, &xprt->xpt_flags))
		return;
471
	xprt->xpt_server->sv_ops->svo_enqueue_xprt(xprt);
472
}
473 474 475
EXPORT_SYMBOL_GPL(svc_xprt_enqueue);

/*
476
 * Dequeue the first transport, if there is one.
477 478 479
 */
static struct svc_xprt *svc_xprt_dequeue(struct svc_pool *pool)
{
480
	struct svc_xprt	*xprt = NULL;
481 482

	if (list_empty(&pool->sp_sockets))
483
		goto out;
484

485 486 487 488 489 490
	spin_lock_bh(&pool->sp_lock);
	if (likely(!list_empty(&pool->sp_sockets))) {
		xprt = list_first_entry(&pool->sp_sockets,
					struct svc_xprt, xpt_ready);
		list_del_init(&xprt->xpt_ready);
		svc_xprt_get(xprt);
491

492
		dprintk("svc: transport %p dequeued, inuse=%d\n",
493
			xprt, kref_read(&xprt->xpt_ref));
494 495
	}
	spin_unlock_bh(&pool->sp_lock);
496 497
out:
	trace_svc_xprt_dequeue(xprt);
498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522
	return xprt;
}

/**
 * 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);
	}
}
523
EXPORT_SYMBOL_GPL(svc_reserve);
524 525 526 527 528 529 530

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);

531 532 533
	kfree(rqstp->rq_deferred);
	rqstp->rq_deferred = NULL;

534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549
	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);
550
	svc_xprt_release_slot(rqstp);
551 552 553 554 555
	rqstp->rq_xprt = NULL;
	svc_xprt_put(xprt);
}

/*
556 557 558 559 560
 * Some svc_serv's will have occasional work to do, even when a xprt is not
 * waiting to be serviced. This function is there to "kick" a task in one of
 * those services so that it can wake up and do that work. Note that we only
 * bother with pool 0 as we don't need to wake up more than one thread for
 * this purpose.
561 562 563 564 565 566
 */
void svc_wake_up(struct svc_serv *serv)
{
	struct svc_rqst	*rqstp;
	struct svc_pool *pool;

567 568
	pool = &serv->sv_pools[0];

569 570 571 572 573 574
	rcu_read_lock();
	list_for_each_entry_rcu(rqstp, &pool->sp_all_threads, rq_all) {
		/* skip any that aren't queued */
		if (test_bit(RQ_BUSY, &rqstp->rq_flags))
			continue;
		rcu_read_unlock();
575 576
		dprintk("svc: daemon %p woken up.\n", rqstp);
		wake_up_process(rqstp->rq_task);
577
		trace_svc_wake_up(rqstp->rq_task->pid);
578 579 580 581 582 583 584
		return;
	}
	rcu_read_unlock();

	/* No free entries available */
	set_bit(SP_TASK_PENDING, &pool->sp_flags);
	smp_wmb();
585
	trace_svc_wake_up(0);
586
}
587
EXPORT_SYMBOL_GPL(svc_wake_up);
588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603

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;
	}
}

/*
604 605 606 607
 * Make sure that we don't have too many active connections. If we have,
 * something must be dropped. It's not clear what will happen if we allow
 * "too many" connections, but when dealing with network-facing software,
 * we have to code defensively. Here we do that by imposing hard limits.
608 609 610 611 612 613 614 615
 *
 * 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.
616 617 618 619
 *
 * single-threaded services that expect a lot of clients will probably
 * need to set sv_maxconn to override the default value which is based
 * on the number of threads
620 621 622
 */
static void svc_check_conn_limits(struct svc_serv *serv)
{
623 624 625 626
	unsigned int limit = serv->sv_maxconn ? serv->sv_maxconn :
				(serv->sv_nrthreads+3) * 20;

	if (serv->sv_tmpcnt > limit) {
627 628 629
		struct svc_xprt *xprt = NULL;
		spin_lock_bh(&serv->sv_lock);
		if (!list_empty(&serv->sv_tempsocks)) {
630 631 632 633 634
			/* Try to help the admin */
			net_notice_ratelimited("%s: too many open connections, consider increasing the %s\n",
					       serv->sv_name, serv->sv_maxconn ?
					       "max number of connections" :
					       "number of threads");
635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653
			/*
			 * 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);
		}
	}
}

654
static int svc_alloc_arg(struct svc_rqst *rqstp)
655
{
J. Bruce Fields's avatar
J. Bruce Fields committed
656 657 658 659
	struct svc_serv *serv = rqstp->rq_server;
	struct xdr_buf *arg;
	int pages;
	int i;
660 661

	/* now allocate needed pages.  If we get a failure, sleep briefly */
662 663 664 665
	pages = (serv->sv_max_mesg + 2 * PAGE_SIZE) >> PAGE_SHIFT;
	if (pages > RPCSVC_MAXPAGES) {
		pr_warn_once("svc: warning: pages=%u > RPCSVC_MAXPAGES=%lu\n",
			     pages, RPCSVC_MAXPAGES);
666
		/* use as many pages as possible */
667 668
		pages = RPCSVC_MAXPAGES;
	}
669 670 671 672
	for (i = 0; i < pages ; i++)
		while (rqstp->rq_pages[i] == NULL) {
			struct page *p = alloc_page(GFP_KERNEL);
			if (!p) {
673 674 675
				set_current_state(TASK_INTERRUPTIBLE);
				if (signalled() || kthread_should_stop()) {
					set_current_state(TASK_RUNNING);
676
					return -EINTR;
677 678
				}
				schedule_timeout(msecs_to_jiffies(500));
679 680 681
			}
			rqstp->rq_pages[i] = p;
		}
682
	rqstp->rq_page_end = &rqstp->rq_pages[i];
683 684 685 686 687 688 689 690 691 692 693 694
	rqstp->rq_pages[i++] = NULL; /* this might be seen in nfs_read_actor */

	/* 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;
J. Bruce Fields's avatar
J. Bruce Fields committed
695 696
	return 0;
}
697

698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721
static bool
rqst_should_sleep(struct svc_rqst *rqstp)
{
	struct svc_pool		*pool = rqstp->rq_pool;

	/* did someone call svc_wake_up? */
	if (test_and_clear_bit(SP_TASK_PENDING, &pool->sp_flags))
		return false;

	/* was a socket queued? */
	if (!list_empty(&pool->sp_sockets))
		return false;

	/* are we shutting down? */
	if (signalled() || kthread_should_stop())
		return false;

	/* are we freezing? */
	if (freezing(current))
		return false;

	return true;
}

722
static struct svc_xprt *svc_get_next_xprt(struct svc_rqst *rqstp, long timeout)
J. Bruce Fields's avatar
J. Bruce Fields committed
723 724 725
{
	struct svc_xprt *xprt;
	struct svc_pool		*pool = rqstp->rq_pool;
726
	long			time_left = 0;
727

728 729 730
	/* rq_xprt should be clear on entry */
	WARN_ON_ONCE(rqstp->rq_xprt);

731 732 733 734 735
	/* Normally we will wait up to 5 seconds for any required
	 * cache information to be provided.
	 */
	rqstp->rq_chandle.thread_wait = 5*HZ;

736 737 738
	xprt = svc_xprt_dequeue(pool);
	if (xprt) {
		rqstp->rq_xprt = xprt;
739 740

		/* As there is a shortage of threads and this request
J. Bruce Fields's avatar
J. Bruce Fields committed
741
		 * had to be queued, don't allow the thread to wait so
742 743 744
		 * long for cache updates.
		 */
		rqstp->rq_chandle.thread_wait = 1*HZ;
745
		clear_bit(SP_TASK_PENDING, &pool->sp_flags);
746 747
		return xprt;
	}
748

749 750 751 752 753 754 755
	/*
	 * We have to be able to interrupt this wait
	 * to bring down the daemons ...
	 */
	set_current_state(TASK_INTERRUPTIBLE);
	clear_bit(RQ_BUSY, &rqstp->rq_flags);
	smp_mb();
756

757 758 759 760
	if (likely(rqst_should_sleep(rqstp)))
		time_left = schedule_timeout(timeout);
	else
		__set_current_state(TASK_RUNNING);
761

762
	try_to_freeze();
763

764 765 766
	spin_lock_bh(&rqstp->rq_lock);
	set_bit(RQ_BUSY, &rqstp->rq_flags);
	spin_unlock_bh(&rqstp->rq_lock);
767

768 769 770 771 772 773 774 775 776 777
	xprt = rqstp->rq_xprt;
	if (xprt != NULL)
		return xprt;

	if (!time_left)
		atomic_long_inc(&pool->sp_stats.threads_timedout);

	if (signalled() || kthread_should_stop())
		return ERR_PTR(-EINTR);
	return ERR_PTR(-EAGAIN);
J. Bruce Fields's avatar
J. Bruce Fields committed
778 779
}

780
static void svc_add_new_temp_xprt(struct svc_serv *serv, struct svc_xprt *newxpt)
781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796
{
	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);
}

J. Bruce Fields's avatar
J. Bruce Fields committed
797 798 799 800
static int svc_handle_xprt(struct svc_rqst *rqstp, struct svc_xprt *xprt)
{
	struct svc_serv *serv = rqstp->rq_server;
	int len = 0;
801

802 803
	if (test_bit(XPT_CLOSE, &xprt->xpt_flags)) {
		dprintk("svc_recv: found XPT_CLOSE\n");
804 805
		if (test_and_clear_bit(XPT_KILL_TEMP, &xprt->xpt_flags))
			xprt->xpt_ops->xpo_kill_temp_xprt(xprt);
806
		svc_delete_xprt(xprt);
807
		/* Leave XPT_BUSY set on the dead xprt: */
808
		goto out;
809 810
	}
	if (test_bit(XPT_LISTENER, &xprt->xpt_flags)) {
811
		struct svc_xprt *newxpt;
812 813 814 815 816 817
		/*
		 * We know this module_get will succeed because the
		 * listener holds a reference too
		 */
		__module_get(xprt->xpt_class->xcl_owner);
		svc_check_conn_limits(xprt->xpt_server);
818
		newxpt = xprt->xpt_ops->xpo_accept(xprt);
819 820
		if (newxpt)
			svc_add_new_temp_xprt(serv, newxpt);
821 822
		else
			module_put(xprt->xpt_class->xcl_owner);
823
	} else if (svc_xprt_reserve_slot(rqstp, xprt)) {
J. Bruce Fields's avatar
J. Bruce Fields committed
824
		/* XPT_DATA|XPT_DEFERRED case: */
825
		dprintk("svc: server %p, pool %u, transport %p, inuse=%d\n",
J. Bruce Fields's avatar
J. Bruce Fields committed
826
			rqstp, rqstp->rq_pool->sp_id, xprt,
827
			kref_read(&xprt->xpt_ref));
828
		rqstp->rq_deferred = svc_deferred_dequeue(xprt);
829
		if (rqstp->rq_deferred)
830
			len = svc_deferred_recv(rqstp);
831
		else
832 833
			len = xprt->xpt_ops->xpo_recvfrom(rqstp);
		dprintk("svc: got len=%d\n", len);
834 835
		rqstp->rq_reserved = serv->sv_max_mesg;
		atomic_add(rqstp->rq_reserved, &xprt->xpt_reserved);
836
	}
J. Bruce Fields's avatar
J. Bruce Fields committed
837
	/* clear XPT_BUSY: */
838
	svc_xprt_received(xprt);
839 840
out:
	trace_svc_handle_xprt(xprt, len);
J. Bruce Fields's avatar
J. Bruce Fields committed
841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861
	return len;
}

/*
 * 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;
	int			len, err;

	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);
862

J. Bruce Fields's avatar
J. Bruce Fields committed
863 864
	err = svc_alloc_arg(rqstp);
	if (err)
865
		goto out;
J. Bruce Fields's avatar
J. Bruce Fields committed
866 867 868

	try_to_freeze();
	cond_resched();
869
	err = -EINTR;
J. Bruce Fields's avatar
J. Bruce Fields committed
870
	if (signalled() || kthread_should_stop())
871
		goto out;
J. Bruce Fields's avatar
J. Bruce Fields committed
872 873

	xprt = svc_get_next_xprt(rqstp, timeout);
874 875 876 877
	if (IS_ERR(xprt)) {
		err = PTR_ERR(xprt);
		goto out;
	}
J. Bruce Fields's avatar
J. Bruce Fields committed
878 879

	len = svc_handle_xprt(rqstp, xprt);
880 881

	/* No data, incomplete (TCP) read, or accept() */
882
	err = -EAGAIN;
883
	if (len <= 0)
884
		goto out_release;
885

886 887
	clear_bit(XPT_OLD, &xprt->xpt_flags);

888 889 890 891
	if (xprt->xpt_ops->xpo_secure_port(rqstp))
		set_bit(RQ_SECURE, &rqstp->rq_flags);
	else
		clear_bit(RQ_SECURE, &rqstp->rq_flags);
892
	rqstp->rq_chandle.defer = svc_defer;
893
	rqstp->rq_xid = svc_getu32(&rqstp->rq_arg.head[0]);
894 895 896

	if (serv->sv_stats)
		serv->sv_stats->netcnt++;
897
	trace_svc_recv(rqstp, len);
898
	return len;
899
out_release:
900 901
	rqstp->rq_res.len = 0;
	svc_xprt_release(rqstp);
902 903 904
out:
	trace_svc_recv(rqstp, err);
	return err;
905
}
906
EXPORT_SYMBOL_GPL(svc_recv);
907 908 909 910 911 912

/*
 * Drop request
 */
void svc_drop(struct svc_rqst *rqstp)
{
913
	trace_svc_drop(rqstp);
914 915 916
	dprintk("svc: xprt %p dropped request\n", rqstp->rq_xprt);
	svc_xprt_release(rqstp);
}
917
EXPORT_SYMBOL_GPL(svc_drop);
918 919 920 921 922 923 924

/*
 * Return reply to client.
 */
int svc_send(struct svc_rqst *rqstp)
{
	struct svc_xprt	*xprt;
925
	int		len = -EFAULT;
926 927 928 929
	struct xdr_buf	*xb;

	xprt = rqstp->rq_xprt;
	if (!xprt)
930
		goto out;
931 932 933 934 935 936 937 938 939 940 941 942

	/* 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);
943 944
	if (test_bit(XPT_DEAD, &xprt->xpt_flags)
			|| test_bit(XPT_CLOSE, &xprt->xpt_flags))
945 946 947 948
		len = -ENOTCONN;
	else
		len = xprt->xpt_ops->xpo_sendto(rqstp);
	mutex_unlock(&xprt->xpt_mutex);
949
	rpc_wake_up(&xprt->xpt_bc_pending);
950 951 952
	svc_xprt_release(rqstp);

	if (len == -ECONNREFUSED || len == -ENOTCONN || len == -EAGAIN)
953 954 955
		len = 0;
out:
	trace_svc_send(rqstp, len);
956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984
	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;

	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;
985
		if (kref_read(&xprt->xpt_ref) > 1 ||
986
		    test_bit(XPT_BUSY, &xprt->xpt_flags))
987
			continue;
988
		list_del_init(le);
989 990 991 992 993 994
		set_bit(XPT_CLOSE, &xprt->xpt_flags);
		dprintk("queuing xprt %p for closing\n", xprt);

		/* a thread will dequeue and close it soon */
		svc_xprt_enqueue(xprt);
	}
995
	spin_unlock_bh(&serv->sv_lock);
996 997 998 999

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

1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026
/* Close temporary transports whose xpt_local matches server_addr immediately
 * instead of waiting for them to be picked up by the timer.
 *
 * This is meant to be called from a notifier_block that runs when an ip
 * address is deleted.
 */
void svc_age_temp_xprts_now(struct svc_serv *serv, struct sockaddr *server_addr)
{
	struct svc_xprt *xprt;
	struct list_head *le, *next;
	LIST_HEAD(to_be_closed);

	spin_lock_bh(&serv->sv_lock);
	list_for_each_safe(le, next, &serv->sv_tempsocks) {
		xprt = list_entry(le, struct svc_xprt, xpt_list);
		if (rpc_cmp_addr(server_addr, (struct sockaddr *)
				&xprt->xpt_local)) {
			dprintk("svc_age_temp_xprts_now: found %p\n", xprt);
			list_move(le, &to_be_closed);
		}
	}
	spin_unlock_bh(&serv->sv_lock);

	while (!list_empty(&to_be_closed)) {
		le = to_be_closed.next;
		list_del_init(le);
		xprt = list_entry(le, struct svc_xprt, xpt_list);
1027 1028 1029 1030 1031
		set_bit(XPT_CLOSE, &xprt->xpt_flags);
		set_bit(XPT_KILL_TEMP, &xprt->xpt_flags);
		dprintk("svc_age_temp_xprts_now: queuing xprt %p for closing\n",
				xprt);
		svc_xprt_enqueue(xprt);
1032 1033 1034 1035
	}
}
EXPORT_SYMBOL_GPL(svc_age_temp_xprts_now);

1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048
static void call_xpt_users(struct svc_xprt *xprt)
{
	struct svc_xpt_user *u;

	spin_lock(&xprt->xpt_lock);
	while (!list_empty(&xprt->xpt_users)) {
		u = list_first_entry(&xprt->xpt_users, struct svc_xpt_user, list);
		list_del(&u->list);
		u->callback(u);
	}
	spin_unlock(&xprt->xpt_lock);
}

1049 1050 1051
/*
 * Remove a dead transport
 */
1052
static void svc_delete_xprt(struct svc_xprt *xprt)
1053 1054
{
	struct svc_serv	*serv = xprt->xpt_server;
1055 1056 1057 1058
	struct svc_deferred_req *dr;

	/* Only do this once */
	if (test_and_set_bit(XPT_DEAD, &xprt->xpt_flags))
1059
		BUG();
1060 1061 1062 1063 1064

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

	spin_lock_bh(&serv->sv_lock);
1065
	list_del_init(&xprt->xpt_list);
1066
	WARN_ON_ONCE(!list_empty(&xprt->xpt_ready));
1067 1068
	if (test_bit(XPT_TEMP, &xprt->xpt_flags))
		serv->sv_tmpcnt--;
1069
	spin_unlock_bh(&serv->sv_lock);
1070

1071
	while ((dr = svc_deferred_dequeue(xprt)) != NULL)
1072 1073
		kfree(dr);

1074
	call_xpt_users(xprt);
1075
	svc_xprt_put(xprt);
1076 1077 1078 1079 1080 1081 1082 1083
}

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;
1084 1085 1086 1087 1088 1089
	/*
	 * We expect svc_close_xprt() to work even when no threads are
	 * running (e.g., while configuring the server before starting
	 * any threads), so if the transport isn't busy, we delete
	 * it ourself:
	 */
1090 1091
	svc_delete_xprt(xprt);
}
1092
EXPORT_SYMBOL_GPL(svc_close_xprt);
1093

1094
static int svc_close_list(struct svc_serv *serv, struct list_head *xprt_list, struct net *net)
1095 1096
{
	struct svc_xprt *xprt;
1097
	int ret = 0;
1098

1099
	spin_lock(&serv->sv_lock);
1100
	list_for_each_entry(xprt, xprt_list, xpt_list) {
1101 1102
		if (xprt->xpt_net != net)
			continue;
1103
		ret++;
1104
		set_bit(XPT_CLOSE, &xprt->xpt_flags);
1105
		svc_xprt_enqueue(xprt);
1106
	}
1107
	spin_unlock(&serv->sv_lock);
1108
	return ret;
1109 1110
}

1111
static struct svc_xprt *svc_dequeue_net(struct svc_serv *serv, struct net *net)
1112
{
1113
	struct svc_pool *pool;
1114 1115
	struct svc_xprt *xprt;
	struct svc_xprt *tmp;
1116 1117 1118 1119 1120 1121
	int i;

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

		spin_lock_bh(&pool->sp_lock);
1122
		list_for_each_entry_safe(xprt, tmp, &pool->sp_sockets, xpt_ready) {
1123 1124
			if (xprt->xpt_net != net)
				continue;
1125
			list_del_init(&xprt->xpt_ready);
1126 1127
			spin_unlock_bh(&pool->sp_lock);
			return xprt;
1128 1129 1130
		}
		spin_unlock_bh(&pool->sp_lock);
	}
1131
	return NULL;
1132 1133
}

1134
static void svc_clean_up_xprts(struct svc_serv *serv, struct net *net)
1135 1136
{
	struct svc_xprt *xprt;
1137

1138 1139
	while ((xprt = svc_dequeue_net(serv, net))) {
		set_bit(XPT_CLOSE, &xprt->xpt_flags);
1140
		svc_delete_xprt(xprt);
1141
	}
1142 1143
}

1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155
/*
 * Server threads may still be running (especially in the case where the
 * service is still running in other network namespaces).
 *
 * So we shut down sockets the same way we would on a running server, by
 * setting XPT_CLOSE, enqueuing, and letting a thread pick it up to do
 * the close.  In the case there are no such other threads,
 * threads running, svc_clean_up_xprts() does a simple version of a
 * server's main event loop, and in the case where there are other
 * threads, we may need to wait a little while and then check again to
 * see if they're done.
 */
1156
void svc_close_net(struct svc_serv *serv, struct net *net)
1157
{
1158
	int delay = 0;
1159

1160 1161 1162 1163 1164 1165
	while (svc_close_list(serv, &serv->sv_permsocks, net) +
	       svc_close_list(serv, &serv->sv_tempsocks, net)) {

		svc_clean_up_xprts(serv, net);
		msleep(delay++);
	}
1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177
}

/*
 * 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;

1178 1179 1180 1181 1182
	spin_lock(&xprt->xpt_lock);
	set_bit(XPT_DEFERRED, &xprt->xpt_flags);
	if (too_many || test_bit(XPT_DEAD, &xprt->xpt_flags)) {
		spin_unlock(&xprt->xpt_lock);
		dprintk("revisit canceled\n");
1183
		svc_xprt_put(xprt);
1184
		trace_svc_drop_deferred(dr);
1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195
		kfree(dr);
		return;
	}
	dprintk("revisit queued\n");
	dr->xprt = NULL;
	list_add(&dr->handle.recent, &xprt->xpt_deferred);
	spin_unlock(&xprt->xpt_lock);
	svc_xprt_enqueue(xprt);
	svc_xprt_put(xprt);
}

1196 1197 1198 1199 1200 1201 1202 1203 1204
/*
 * 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.
 */
1205 1206 1207 1208 1209
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;

1210
	if (rqstp->rq_arg.page_len || !test_bit(RQ_USEDEFERRAL, &rqstp->rq_flags))
1211 1212 1213 1214 1215
		return NULL; /* if more than a page, give up FIXME */
	if (rqstp->rq_deferred) {
		dr = rqstp->rq_deferred;
		rqstp->rq_deferred = NULL;
	} else {
1216 1217
		size_t skip;
		size_t size;
1218
		/* FIXME maybe discard if size too large */
1219
		size = sizeof(struct svc_deferred_req) + rqstp->rq_arg.len;
1220 1221 1222 1223 1224 1225 1226 1227 1228 1229
		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;
1230 1231 1232 1233 1234 1235
		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);
1236 1237 1238
	}
	svc_xprt_get(rqstp->rq_xprt);
	dr->xprt = rqstp->rq_xprt;
1239
	set_bit(RQ_DROPME, &rqstp->rq_flags);
1240 1241

	dr->handle.revisit = svc_revisit;
1242
	trace_svc_defer(rqstp);
1243 1244 1245 1246 1247 1248 1249 1250 1251 1252
	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;

1253 1254 1255 1256
	/* 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;
1257
	rqstp->rq_arg.page_len = 0;
1258 1259
	/* The rq_arg.len includes the transport header bytes */
	rqstp->rq_arg.len     = dr->argslen<<2;
1260 1261 1262
	rqstp->rq_prot        = dr->prot;
	memcpy(&rqstp->rq_addr, &dr->addr, dr->addrlen);
	rqstp->rq_addrlen     = dr->addrlen;
1263 1264
	/* Save off transport header len in case we get deferred again */
	rqstp->rq_xprt_hlen   = dr->xprt_hlen;
1265 1266
	rqstp->rq_daddr       = dr->daddr;
	rqstp->rq_respages    = rqstp->rq_pages;
1267
	return (dr->argslen<<2) - dr->xprt_hlen;
1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282
}


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);
	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);
1283
		trace_svc_revisit_deferred(dr);
1284 1285
	} else
		clear_bit(XPT_DEFERRED, &xprt->xpt_flags);
1286 1287 1288
	spin_unlock(&xprt->xpt_lock);
	return dr;
}
1289

1290 1291 1292 1293
/**
 * svc_find_xprt - find an RPC transport instance
 * @serv: pointer to svc_serv to search
 * @xcl_name: C string containing transport's class name
1294
 * @net: owner net pointer
1295 1296 1297
 * @af: Address family of transport's local address
 * @port: transport's IP port number
 *
1298 1299 1300 1301 1302 1303 1304 1305
 * Return the transport instance pointer for the endpoint accepting
 * connections/peer traffic from the specified transport class,
 * address family and port.
 *
 * Specifying 0 for the address family or port is effectively a
 * wild-card, and will result in matching the first transport in the
 * service's list that has a matching class name.
 */
1306
struct svc_xprt *svc_find_xprt(struct svc_serv *serv, const char *xcl_name,
1307 1308
			       struct net *net, const sa_family_t af,
			       const unsigned short port)
1309 1310 1311 1312 1313
{
	struct svc_xprt *xprt;
	struct svc_xprt *found = NULL;

	/* Sanity check the args */
1314
	if (serv == NULL