svc_xprt.c 38.0 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0-only
T
Tom Tucker 已提交
2 3 4 5 6 7 8 9 10
/*
 * linux/net/sunrpc/svc_xprt.c
 *
 * Author: Tom Tucker <tom@opengridcomputing.com>
 */

#include <linux/sched.h>
#include <linux/errno.h>
#include <linux/freezer.h>
11
#include <linux/kthread.h>
12
#include <linux/slab.h>
T
Tom Tucker 已提交
13
#include <net/sock.h>
14
#include <linux/sunrpc/addr.h>
T
Tom Tucker 已提交
15 16
#include <linux/sunrpc/stats.h>
#include <linux/sunrpc/svc_xprt.h>
17
#include <linux/sunrpc/svcsock.h>
18
#include <linux/sunrpc/xprt.h>
19
#include <linux/module.h>
20
#include <linux/netdevice.h>
21
#include <trace/events/sunrpc.h>
T
Tom Tucker 已提交
22 23 24

#define RPCDBG_FACILITY	RPCDBG_SVCXPRT

25 26 27 28
static unsigned int svc_rpc_per_connection_limit __read_mostly;
module_param(svc_rpc_per_connection_limit, uint, 0644);


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);
32
static void svc_age_temp_xprts(struct timer_list *t);
33
static void svc_delete_xprt(struct svc_xprt *xprt);
34 35 36 37 38 39 40 41

/* 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;

T
Tom Tucker 已提交
42 43 44 45
/* List of registered transport classes */
static DEFINE_SPINLOCK(svc_xprt_class_lock);
static LIST_HEAD(svc_xprt_class_list);

46 47 48 49 50
/* 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.
51
 *	The "service mutex" protects svc_serv->sv_nrthread.
52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75
 *	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.
 */
T
Tom Tucker 已提交
76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106
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);

107 108 109 110 111
/*
 * Format the transport list for printing
 */
int svc_print_xprts(char *buf, int maxlen)
{
112
	struct svc_xprt_class *xcl;
113 114 115 116 117
	char tmpstr[80];
	int len = 0;
	buf[0] = '\0';

	spin_lock(&svc_xprt_class_lock);
118
	list_for_each_entry(xcl, &svc_xprt_class_list, xcl_list) {
119 120 121 122 123 124 125 126 127 128 129 130 131 132
		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;
}

T
Tom Tucker 已提交
133 134 135 136 137
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;
138 139
	if (test_bit(XPT_CACHE_AUTH, &xprt->xpt_flags))
		svcauth_unix_info_release(xprt);
140
	put_cred(xprt->xpt_cred);
P
Pavel Emelyanov 已提交
141
	put_net(xprt->xpt_net);
142 143 144
	/* See comment on corresponding get in xs_setup_bc_tcp(): */
	if (xprt->xpt_bc_xprt)
		xprt_put(xprt->xpt_bc_xprt);
145 146
	if (xprt->xpt_bc_xps)
		xprt_switch_put(xprt->xpt_bc_xps);
T
Tom Tucker 已提交
147 148 149 150 151 152 153 154 155 156
	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);

T
Tom Tucker 已提交
157 158 159 160
/*
 * Called by transport drivers to initialize the transport independent
 * portion of the transport instance.
 */
161 162
void svc_xprt_init(struct net *net, struct svc_xprt_class *xcl,
		   struct svc_xprt *xprt, struct svc_serv *serv)
T
Tom Tucker 已提交
163 164 165 166
{
	memset(xprt, 0, sizeof(*xprt));
	xprt->xpt_class = xcl;
	xprt->xpt_ops = xcl->xcl_ops;
T
Tom Tucker 已提交
167
	kref_init(&xprt->xpt_ref);
168
	xprt->xpt_server = serv;
169 170
	INIT_LIST_HEAD(&xprt->xpt_list);
	INIT_LIST_HEAD(&xprt->xpt_ready);
171
	INIT_LIST_HEAD(&xprt->xpt_deferred);
172
	INIT_LIST_HEAD(&xprt->xpt_users);
173
	mutex_init(&xprt->xpt_mutex);
174
	spin_lock_init(&xprt->xpt_lock);
175
	set_bit(XPT_BUSY, &xprt->xpt_flags);
176
	xprt->xpt_net = get_net(net);
177
	strcpy(xprt->xpt_remotebuf, "uninitialized");
T
Tom Tucker 已提交
178 179
}
EXPORT_SYMBOL_GPL(svc_xprt_init);
180

181 182
static struct svc_xprt *__svc_xpo_create(struct svc_xprt_class *xcl,
					 struct svc_serv *serv,
183
					 struct net *net,
184 185 186
					 const int family,
					 const unsigned short port,
					 int flags)
187 188 189
{
	struct sockaddr_in sin = {
		.sin_family		= AF_INET,
A
Al Viro 已提交
190
		.sin_addr.s_addr	= htonl(INADDR_ANY),
191 192
		.sin_port		= htons(port),
	};
E
Eric Dumazet 已提交
193
#if IS_ENABLED(CONFIG_IPV6)
194 195 196 197 198
	struct sockaddr_in6 sin6 = {
		.sin6_family		= AF_INET6,
		.sin6_addr		= IN6ADDR_ANY_INIT,
		.sin6_port		= htons(port),
	};
E
Eric Dumazet 已提交
199
#endif
200 201 202
	struct sockaddr *sap;
	size_t len;

203 204
	switch (family) {
	case PF_INET:
205 206 207
		sap = (struct sockaddr *)&sin;
		len = sizeof(sin);
		break;
E
Eric Dumazet 已提交
208
#if IS_ENABLED(CONFIG_IPV6)
209
	case PF_INET6:
210 211 212
		sap = (struct sockaddr *)&sin6;
		len = sizeof(sin6);
		break;
E
Eric Dumazet 已提交
213
#endif
214 215 216 217
	default:
		return ERR_PTR(-EAFNOSUPPORT);
	}

218
	return xcl->xcl_ops->xpo_create(serv, net, sap, len, flags);
219 220
}

221 222 223 224 225 226 227 228 229 230
/*
 * 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)
{
231 232
	if (!test_bit(XPT_BUSY, &xprt->xpt_flags)) {
		WARN_ONCE(1, "xprt=0x%p already busy!", xprt);
233
		return;
234 235
	}

236
	/* As soon as we clear busy, the xprt could be closed and
237
	 * 'put', so we need a reference to call svc_enqueue_xprt with:
238 239
	 */
	svc_xprt_get(xprt);
240
	smp_mb__before_atomic();
241
	clear_bit(XPT_BUSY, &xprt->xpt_flags);
242
	xprt->xpt_server->sv_ops->svo_enqueue_xprt(xprt);
243 244 245
	svc_xprt_put(xprt);
}

246 247 248 249 250 251 252 253 254
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);
}

255 256
static int _svc_create_xprt(struct svc_serv *serv, const char *xprt_name,
			    struct net *net, const int family,
257 258
			    const unsigned short port, int flags,
			    const struct cred *cred)
259 260 261
{
	struct svc_xprt_class *xcl;

262 263
	spin_lock(&svc_xprt_class_lock);
	list_for_each_entry(xcl, &svc_xprt_class_list, xcl_list) {
264
		struct svc_xprt *newxprt;
265
		unsigned short newport;
266 267 268 269 270 271 272 273

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

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

		spin_unlock(&svc_xprt_class_lock);
274
		newxprt = __svc_xpo_create(xcl, serv, net, family, port, flags);
275 276 277
		if (IS_ERR(newxprt)) {
			module_put(xcl->xcl_owner);
			return PTR_ERR(newxprt);
278
		}
279
		newxprt->xpt_cred = get_cred(cred);
280
		svc_add_new_perm_xprt(serv, newxprt);
281 282
		newport = svc_xprt_local_port(newxprt);
		return newport;
283
	}
284
 err:
285
	spin_unlock(&svc_xprt_class_lock);
286 287 288
	/* This errno is exposed to user space.  Provide a reasonable
	 * perror msg for a bad transport. */
	return -EPROTONOSUPPORT;
289
}
J
J. Bruce Fields 已提交
290 291 292

int svc_create_xprt(struct svc_serv *serv, const char *xprt_name,
		    struct net *net, const int family,
293 294
		    const unsigned short port, int flags,
		    const struct cred *cred)
J
J. Bruce Fields 已提交
295 296 297 298
{
	int err;

	dprintk("svc: creating transport %s[%d]\n", xprt_name, port);
299
	err = _svc_create_xprt(serv, xprt_name, net, family, port, flags, cred);
J
J. Bruce Fields 已提交
300 301
	if (err == -EPROTONOSUPPORT) {
		request_module("svc%s", xprt_name);
302
		err = _svc_create_xprt(serv, xprt_name, net, family, port, flags, cred);
J
J. Bruce Fields 已提交
303
	}
304
	if (err < 0)
J
J. Bruce Fields 已提交
305
		dprintk("svc: transport %s not found, err %d\n",
306
			xprt_name, -err);
J
J. Bruce Fields 已提交
307 308
	return err;
}
309
EXPORT_SYMBOL_GPL(svc_create_xprt);
310 311 312 313 314 315 316 317 318 319 320 321 322

/*
 * 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.
	 */
323 324
	memcpy(&rqstp->rq_daddr, &xprt->xpt_local, xprt->xpt_locallen);
	rqstp->rq_daddrlen = xprt->xpt_locallen;
325 326 327
}
EXPORT_SYMBOL_GPL(svc_xprt_copy_addrs);

328 329 330 331 332 333 334 335 336 337 338 339 340
/**
 * 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);

341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364
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);
365
		smp_wmb(); /* See smp_rmb() in svc_xprt_ready() */
366 367 368 369
		svc_xprt_enqueue(xprt);
	}
}

370
static bool svc_xprt_ready(struct svc_xprt *xprt)
J
J. Bruce Fields 已提交
371
{
372 373
	unsigned long xpt_flags;

374 375 376 377 378 379 380 381 382
	/*
	 * If another cpu has recently updated xpt_flags,
	 * sk_sock->flags, xpt_reserved, or xpt_nr_rqsts, we need to
	 * know about it; otherwise it's possible that both that cpu and
	 * this one could call svc_xprt_enqueue() without either
	 * svc_xprt_enqueue() recognizing that the conditions below
	 * are satisfied, and we could stall indefinitely:
	 */
	smp_rmb();
383 384 385
	xpt_flags = READ_ONCE(xprt->xpt_flags);

	if (xpt_flags & (BIT(XPT_CONN) | BIT(XPT_CLOSE)))
J
J. Bruce Fields 已提交
386
		return true;
387
	if (xpt_flags & (BIT(XPT_DATA) | BIT(XPT_DEFERRED))) {
388 389
		if (xprt->xpt_ops->xpo_has_wspace(xprt) &&
		    svc_xprt_slots_in_range(xprt))
390 391 392 393
			return true;
		trace_svc_xprt_no_write_space(xprt);
		return false;
	}
J
J. Bruce Fields 已提交
394 395 396
	return false;
}

397
void svc_xprt_do_enqueue(struct svc_xprt *xprt)
398 399
{
	struct svc_pool *pool;
400
	struct svc_rqst	*rqstp = NULL;
401 402
	int cpu;

403
	if (!svc_xprt_ready(xprt))
404
		return;
405 406 407 408 409 410

	/* 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.
	 */
411 412
	if (test_and_set_bit(XPT_BUSY, &xprt->xpt_flags))
		return;
413

414 415 416
	cpu = get_cpu();
	pool = svc_pool_for_cpu(xprt->xpt_server, cpu);

417
	atomic_long_inc(&pool->sp_stats.packets);
418

419 420 421 422 423
	spin_lock_bh(&pool->sp_lock);
	list_add_tail(&xprt->xpt_ready, &pool->sp_sockets);
	pool->sp_stats.sockets_queued++;
	spin_unlock_bh(&pool->sp_lock);

424 425 426
	/* find a thread for this xprt */
	rcu_read_lock();
	list_for_each_entry_rcu(rqstp, &pool->sp_all_threads, rq_all) {
427
		if (test_and_set_bit(RQ_BUSY, &rqstp->rq_flags))
428
			continue;
429
		atomic_long_inc(&pool->sp_stats.threads_woken);
C
Chuck Lever 已提交
430
		rqstp->rq_qtime = ktime_get();
431
		wake_up_process(rqstp->rq_task);
432
		goto out_unlock;
433
	}
434
	set_bit(SP_CONGESTED, &pool->sp_flags);
435
	rqstp = NULL;
436 437
out_unlock:
	rcu_read_unlock();
438
	put_cpu();
439
	trace_svc_xprt_do_enqueue(xprt, rqstp);
440
}
441
EXPORT_SYMBOL_GPL(svc_xprt_do_enqueue);
442 443 444 445 446 447 448 449 450 451

/*
 * 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;
452
	xprt->xpt_server->sv_ops->svo_enqueue_xprt(xprt);
453
}
454 455 456
EXPORT_SYMBOL_GPL(svc_xprt_enqueue);

/*
457
 * Dequeue the first transport, if there is one.
458 459 460
 */
static struct svc_xprt *svc_xprt_dequeue(struct svc_pool *pool)
{
461
	struct svc_xprt	*xprt = NULL;
462 463

	if (list_empty(&pool->sp_sockets))
464
		goto out;
465

466 467 468 469 470 471 472 473
	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);
	}
	spin_unlock_bh(&pool->sp_lock);
474
out:
475 476 477 478 479 480 481 482 483 484 485 486 487 488 489
	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)
{
490 491
	struct svc_xprt *xprt = rqstp->rq_xprt;

492 493
	space += rqstp->rq_res.head[0].iov_len;

494
	if (xprt && space < rqstp->rq_reserved) {
495 496
		atomic_sub((rqstp->rq_reserved - space), &xprt->xpt_reserved);
		rqstp->rq_reserved = space;
497
		smp_wmb(); /* See smp_rmb() in svc_xprt_ready() */
498 499 500
		svc_xprt_enqueue(xprt);
	}
}
501
EXPORT_SYMBOL_GPL(svc_reserve);
502 503 504 505 506

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

507
	xprt->xpt_ops->xpo_release_rqst(rqstp);
508

509 510 511
	kfree(rqstp->rq_deferred);
	rqstp->rq_deferred = NULL;

512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527
	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);
528
	svc_xprt_release_slot(rqstp);
529 530 531 532 533
	rqstp->rq_xprt = NULL;
	svc_xprt_put(xprt);
}

/*
534 535 536 537 538
 * 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.
539 540 541 542 543 544
 */
void svc_wake_up(struct svc_serv *serv)
{
	struct svc_rqst	*rqstp;
	struct svc_pool *pool;

545 546
	pool = &serv->sv_pools[0];

547 548 549 550 551 552
	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();
553
		wake_up_process(rqstp->rq_task);
554
		trace_svc_wake_up(rqstp->rq_task->pid);
555 556 557 558 559 560 561
		return;
	}
	rcu_read_unlock();

	/* No free entries available */
	set_bit(SP_TASK_PENDING, &pool->sp_flags);
	smp_wmb();
562
	trace_svc_wake_up(0);
563
}
564
EXPORT_SYMBOL_GPL(svc_wake_up);
565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580

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;
	}
}

/*
581 582 583 584
 * 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.
585 586 587 588 589 590 591 592
 *
 * 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.
593 594 595 596
 *
 * 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
597 598 599
 */
static void svc_check_conn_limits(struct svc_serv *serv)
{
600 601 602 603
	unsigned int limit = serv->sv_maxconn ? serv->sv_maxconn :
				(serv->sv_nrthreads+3) * 20;

	if (serv->sv_tmpcnt > limit) {
604 605 606
		struct svc_xprt *xprt = NULL;
		spin_lock_bh(&serv->sv_lock);
		if (!list_empty(&serv->sv_tempsocks)) {
607 608 609 610 611
			/* 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");
612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630
			/*
			 * 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);
		}
	}
}

631
static int svc_alloc_arg(struct svc_rqst *rqstp)
632
{
J
J. Bruce Fields 已提交
633 634 635 636
	struct svc_serv *serv = rqstp->rq_server;
	struct xdr_buf *arg;
	int pages;
	int i;
637 638

	/* now allocate needed pages.  If we get a failure, sleep briefly */
639 640 641 642
	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);
643
		/* use as many pages as possible */
644 645
		pages = RPCSVC_MAXPAGES;
	}
646 647 648 649
	for (i = 0; i < pages ; i++)
		while (rqstp->rq_pages[i] == NULL) {
			struct page *p = alloc_page(GFP_KERNEL);
			if (!p) {
650 651 652
				set_current_state(TASK_INTERRUPTIBLE);
				if (signalled() || kthread_should_stop()) {
					set_current_state(TASK_RUNNING);
653
					return -EINTR;
654 655
				}
				schedule_timeout(msecs_to_jiffies(500));
656 657 658
			}
			rqstp->rq_pages[i] = p;
		}
659
	rqstp->rq_page_end = &rqstp->rq_pages[i];
660 661 662 663 664 665 666 667 668 669 670 671
	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
J. Bruce Fields 已提交
672 673
	return 0;
}
674

675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698
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;
}

699
static struct svc_xprt *svc_get_next_xprt(struct svc_rqst *rqstp, long timeout)
J
J. Bruce Fields 已提交
700 701
{
	struct svc_pool		*pool = rqstp->rq_pool;
702
	long			time_left = 0;
703

704 705 706
	/* rq_xprt should be clear on entry */
	WARN_ON_ONCE(rqstp->rq_xprt);

707 708 709
	rqstp->rq_xprt = svc_xprt_dequeue(pool);
	if (rqstp->rq_xprt)
		goto out_found;
710

711 712 713 714 715
	/*
	 * We have to be able to interrupt this wait
	 * to bring down the daemons ...
	 */
	set_current_state(TASK_INTERRUPTIBLE);
716 717
	smp_mb__before_atomic();
	clear_bit(SP_CONGESTED, &pool->sp_flags);
718
	clear_bit(RQ_BUSY, &rqstp->rq_flags);
719
	smp_mb__after_atomic();
720

721 722 723 724
	if (likely(rqst_should_sleep(rqstp)))
		time_left = schedule_timeout(timeout);
	else
		__set_current_state(TASK_RUNNING);
725

726
	try_to_freeze();
727

728
	set_bit(RQ_BUSY, &rqstp->rq_flags);
729 730 731 732
	smp_mb__after_atomic();
	rqstp->rq_xprt = svc_xprt_dequeue(pool);
	if (rqstp->rq_xprt)
		goto out_found;
733 734 735 736 737 738 739

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

	if (signalled() || kthread_should_stop())
		return ERR_PTR(-EINTR);
	return ERR_PTR(-EAGAIN);
740 741 742 743 744 745 746 747
out_found:
	/* Normally we will wait up to 5 seconds for any required
	 * cache information to be provided.
	 */
	if (!test_bit(SP_CONGESTED, &pool->sp_flags))
		rqstp->rq_chandle.thread_wait = 5*HZ;
	else
		rqstp->rq_chandle.thread_wait = 1*HZ;
C
Chuck Lever 已提交
748
	trace_svc_xprt_dequeue(rqstp);
749
	return rqstp->rq_xprt;
J
J. Bruce Fields 已提交
750 751
}

752
static void svc_add_new_temp_xprt(struct svc_serv *serv, struct svc_xprt *newxpt)
753 754 755 756 757 758 759
{
	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 */
760
		serv->sv_temptimer.function = svc_age_temp_xprts;
761 762 763 764 765 766 767
		mod_timer(&serv->sv_temptimer,
			  jiffies + svc_conn_age_period * HZ);
	}
	spin_unlock_bh(&serv->sv_lock);
	svc_xprt_received(newxpt);
}

J
J. Bruce Fields 已提交
768 769 770 771
static int svc_handle_xprt(struct svc_rqst *rqstp, struct svc_xprt *xprt)
{
	struct svc_serv *serv = rqstp->rq_server;
	int len = 0;
772

773 774
	if (test_bit(XPT_CLOSE, &xprt->xpt_flags)) {
		dprintk("svc_recv: found XPT_CLOSE\n");
775 776
		if (test_and_clear_bit(XPT_KILL_TEMP, &xprt->xpt_flags))
			xprt->xpt_ops->xpo_kill_temp_xprt(xprt);
777
		svc_delete_xprt(xprt);
778
		/* Leave XPT_BUSY set on the dead xprt: */
779
		goto out;
780 781
	}
	if (test_bit(XPT_LISTENER, &xprt->xpt_flags)) {
782
		struct svc_xprt *newxpt;
783 784 785 786 787 788
		/*
		 * 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);
789
		newxpt = xprt->xpt_ops->xpo_accept(xprt);
790 791
		if (newxpt) {
			newxpt->xpt_cred = get_cred(xprt->xpt_cred);
792
			svc_add_new_temp_xprt(serv, newxpt);
793
		} else
794
			module_put(xprt->xpt_class->xcl_owner);
795
	} else if (svc_xprt_reserve_slot(rqstp, xprt)) {
J
J. Bruce Fields 已提交
796
		/* XPT_DATA|XPT_DEFERRED case: */
797
		dprintk("svc: server %p, pool %u, transport %p, inuse=%d\n",
J
J. Bruce Fields 已提交
798
			rqstp, rqstp->rq_pool->sp_id, xprt,
799
			kref_read(&xprt->xpt_ref));
800
		rqstp->rq_deferred = svc_deferred_dequeue(xprt);
801
		if (rqstp->rq_deferred)
802
			len = svc_deferred_recv(rqstp);
803
		else
804
			len = xprt->xpt_ops->xpo_recvfrom(rqstp);
805
		rqstp->rq_stime = ktime_get();
806 807
		rqstp->rq_reserved = serv->sv_max_mesg;
		atomic_add(rqstp->rq_reserved, &xprt->xpt_reserved);
808
	}
J
J. Bruce Fields 已提交
809
	/* clear XPT_BUSY: */
810
	svc_xprt_received(xprt);
811 812
out:
	trace_svc_handle_xprt(xprt, len);
J
J. Bruce Fields 已提交
813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833
	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);
834

J
J. Bruce Fields 已提交
835 836
	err = svc_alloc_arg(rqstp);
	if (err)
837
		goto out;
J
J. Bruce Fields 已提交
838 839 840

	try_to_freeze();
	cond_resched();
841
	err = -EINTR;
J
J. Bruce Fields 已提交
842
	if (signalled() || kthread_should_stop())
843
		goto out;
J
J. Bruce Fields 已提交
844 845

	xprt = svc_get_next_xprt(rqstp, timeout);
846 847 848 849
	if (IS_ERR(xprt)) {
		err = PTR_ERR(xprt);
		goto out;
	}
J
J. Bruce Fields 已提交
850 851

	len = svc_handle_xprt(rqstp, xprt);
852 853

	/* No data, incomplete (TCP) read, or accept() */
854
	err = -EAGAIN;
855
	if (len <= 0)
856
		goto out_release;
857

858 859
	clear_bit(XPT_OLD, &xprt->xpt_flags);

C
Chuck Lever 已提交
860
	xprt->xpt_ops->xpo_secure_port(rqstp);
861
	rqstp->rq_chandle.defer = svc_defer;
862
	rqstp->rq_xid = svc_getu32(&rqstp->rq_arg.head[0]);
863 864 865

	if (serv->sv_stats)
		serv->sv_stats->netcnt++;
866
	trace_svc_recv(rqstp, len);
867
	return len;
868
out_release:
869 870
	rqstp->rq_res.len = 0;
	svc_xprt_release(rqstp);
871 872
out:
	return err;
873
}
874
EXPORT_SYMBOL_GPL(svc_recv);
875 876 877 878 879 880

/*
 * Drop request
 */
void svc_drop(struct svc_rqst *rqstp)
{
881
	trace_svc_drop(rqstp);
882 883 884
	dprintk("svc: xprt %p dropped request\n", rqstp->rq_xprt);
	svc_xprt_release(rqstp);
}
885
EXPORT_SYMBOL_GPL(svc_drop);
886 887 888 889 890 891 892

/*
 * Return reply to client.
 */
int svc_send(struct svc_rqst *rqstp)
{
	struct svc_xprt	*xprt;
893
	int		len = -EFAULT;
894 895 896 897
	struct xdr_buf	*xb;

	xprt = rqstp->rq_xprt;
	if (!xprt)
898
		goto out;
899 900

	/* release the receive skb before sending the reply */
901
	xprt->xpt_ops->xpo_release_rqst(rqstp);
902 903 904 905 906 907 908 909 910

	/* 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);
911
	trace_svc_stats_latency(rqstp);
912 913
	if (test_bit(XPT_DEAD, &xprt->xpt_flags)
			|| test_bit(XPT_CLOSE, &xprt->xpt_flags))
914 915 916 917
		len = -ENOTCONN;
	else
		len = xprt->xpt_ops->xpo_sendto(rqstp);
	mutex_unlock(&xprt->xpt_mutex);
918
	trace_svc_send(rqstp, len);
919 920 921
	svc_xprt_release(rqstp);

	if (len == -ECONNREFUSED || len == -ENOTCONN || len == -EAGAIN)
922 923
		len = 0;
out:
924 925 926 927 928 929 930
	return len;
}

/*
 * Timer function to close old temporary transports, using
 * a mark-and-sweep algorithm.
 */
931
static void svc_age_temp_xprts(struct timer_list *t)
932
{
933
	struct svc_serv *serv = from_timer(serv, t, sv_temptimer);
934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952
	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;
953
		if (kref_read(&xprt->xpt_ref) > 1 ||
954
		    test_bit(XPT_BUSY, &xprt->xpt_flags))
955
			continue;
956
		list_del_init(le);
957 958 959 960 961 962
		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);
	}
963
	spin_unlock_bh(&serv->sv_lock);
964 965 966 967

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

968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994
/* 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);
995 996 997 998 999
		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);
1000 1001 1002 1003
	}
}
EXPORT_SYMBOL_GPL(svc_age_temp_xprts_now);

1004 1005 1006 1007 1008 1009 1010
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);
1011
		list_del_init(&u->list);
1012 1013 1014 1015 1016
		u->callback(u);
	}
	spin_unlock(&xprt->xpt_lock);
}

1017 1018 1019
/*
 * Remove a dead transport
 */
1020
static void svc_delete_xprt(struct svc_xprt *xprt)
1021 1022
{
	struct svc_serv	*serv = xprt->xpt_server;
1023 1024 1025 1026
	struct svc_deferred_req *dr;

	/* Only do this once */
	if (test_and_set_bit(XPT_DEAD, &xprt->xpt_flags))
1027
		BUG();
1028 1029 1030 1031 1032

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

	spin_lock_bh(&serv->sv_lock);
1033
	list_del_init(&xprt->xpt_list);
1034
	WARN_ON_ONCE(!list_empty(&xprt->xpt_ready));
1035 1036
	if (test_bit(XPT_TEMP, &xprt->xpt_flags))
		serv->sv_tmpcnt--;
1037
	spin_unlock_bh(&serv->sv_lock);
1038

1039
	while ((dr = svc_deferred_dequeue(xprt)) != NULL)
1040 1041
		kfree(dr);

1042
	call_xpt_users(xprt);
1043
	svc_xprt_put(xprt);
1044 1045 1046 1047 1048 1049 1050 1051
}

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;
J
J. Bruce Fields 已提交
1052 1053 1054 1055 1056 1057
	/*
	 * 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:
	 */
1058 1059
	svc_delete_xprt(xprt);
}
1060
EXPORT_SYMBOL_GPL(svc_close_xprt);
1061

1062
static int svc_close_list(struct svc_serv *serv, struct list_head *xprt_list, struct net *net)
1063 1064
{
	struct svc_xprt *xprt;
1065
	int ret = 0;
1066

1067
	spin_lock(&serv->sv_lock);
1068
	list_for_each_entry(xprt, xprt_list, xpt_list) {
1069 1070
		if (xprt->xpt_net != net)
			continue;
1071
		ret++;
1072
		set_bit(XPT_CLOSE, &xprt->xpt_flags);
1073
		svc_xprt_enqueue(xprt);
1074
	}
1075
	spin_unlock(&serv->sv_lock);
1076
	return ret;
1077 1078
}

1079
static struct svc_xprt *svc_dequeue_net(struct svc_serv *serv, struct net *net)
1080
{
1081
	struct svc_pool *pool;
1082 1083
	struct svc_xprt *xprt;
	struct svc_xprt *tmp;
1084 1085 1086 1087 1088 1089
	int i;

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

		spin_lock_bh(&pool->sp_lock);
1090
		list_for_each_entry_safe(xprt, tmp, &pool->sp_sockets, xpt_ready) {
1091 1092
			if (xprt->xpt_net != net)
				continue;
1093
			list_del_init(&xprt->xpt_ready);
1094 1095
			spin_unlock_bh(&pool->sp_lock);
			return xprt;
1096 1097 1098
		}
		spin_unlock_bh(&pool->sp_lock);
	}
1099
	return NULL;
1100 1101
}

1102
static void svc_clean_up_xprts(struct svc_serv *serv, struct net *net)
1103 1104
{
	struct svc_xprt *xprt;
1105

1106 1107
	while ((xprt = svc_dequeue_net(serv, net))) {
		set_bit(XPT_CLOSE, &xprt->xpt_flags);
1108
		svc_delete_xprt(xprt);
1109
	}
1110 1111
}

1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123
/*
 * 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.
 */
1124
void svc_close_net(struct svc_serv *serv, struct net *net)
1125
{
1126
	int delay = 0;
1127

1128 1129 1130 1131 1132 1133
	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++);
	}
1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145
}

/*
 * 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;

1146 1147 1148 1149 1150
	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");
1151
		svc_xprt_put(xprt);
1152
		trace_svc_drop_deferred(dr);
1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163
		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);
}

1164 1165 1166 1167 1168 1169 1170 1171 1172
/*
 * 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.
 */
1173 1174 1175 1176 1177
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;

1178
	if (rqstp->rq_arg.page_len || !test_bit(RQ_USEDEFERRAL, &rqstp->rq_flags))
1179 1180 1181 1182 1183
		return NULL; /* if more than a page, give up FIXME */
	if (rqstp->rq_deferred) {
		dr = rqstp->rq_deferred;
		rqstp->rq_deferred = NULL;
	} else {
1184 1185
		size_t skip;
		size_t size;
1186
		/* FIXME maybe discard if size too large */
1187
		size = sizeof(struct svc_deferred_req) + rqstp->rq_arg.len;
1188 1189 1190 1191 1192 1193 1194 1195 1196 1197
		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;
1198 1199 1200 1201 1202 1203
		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);
1204 1205 1206
	}
	svc_xprt_get(rqstp->rq_xprt);
	dr->xprt = rqstp->rq_xprt;
1207
	set_bit(RQ_DROPME, &rqstp->rq_flags);
1208 1209

	dr->handle.revisit = svc_revisit;
1210
	trace_svc_defer(rqstp);
1211 1212 1213 1214 1215 1216 1217 1218 1219 1220
	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;

1221 1222 1223 1224
	/* 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;
1225
	rqstp->rq_arg.page_len = 0;
1226 1227
	/* The rq_arg.len includes the transport header bytes */
	rqstp->rq_arg.len     = dr->argslen<<2;
1228 1229 1230
	rqstp->rq_prot        = dr->prot;
	memcpy(&rqstp->rq_addr, &dr->addr, dr->addrlen);
	rqstp->rq_addrlen     = dr->addrlen;
1231 1232
	/* Save off transport header len in case we get deferred again */
	rqstp->rq_xprt_hlen   = dr->xprt_hlen;
1233 1234
	rqstp->rq_daddr       = dr->daddr;
	rqstp->rq_respages    = rqstp->rq_pages;
1235
	return (dr->argslen<<2) - dr->xprt_hlen;
1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250
}


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);
1251
		trace_svc_revisit_deferred(dr);
1252 1253
	} else
		clear_bit(XPT_DEFERRED, &xprt->xpt_flags);
1254 1255 1256
	spin_unlock(&xprt->xpt_lock);
	return dr;
}
1257

1258 1259 1260 1261
/**
 * svc_find_xprt - find an RPC transport instance
 * @serv: pointer to svc_serv to search
 * @xcl_name: C string containing transport's class name
1262
 * @net: owner net pointer
1263 1264 1265
 * @af: Address family of transport's local address
 * @port: transport's IP port number
 *
1266 1267 1268 1269 1270 1271 1272 1273
 * 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.
 */
1274
struct svc_xprt *svc_find_xprt(struct svc_serv *serv, const char *xcl_name,
1275 1276
			       struct net *net, const sa_family_t af,
			       const unsigned short port)
1277 1278 1279 1280 1281
{
	struct svc_xprt *xprt;
	struct svc_xprt *found = NULL;

	/* Sanity check the args */
1282
	if (serv == NULL || xcl_name == NULL)
1283 1284 1285 1286
		return found;

	spin_lock_bh(&serv->sv_lock);
	list_for_each_entry(xprt, &serv->sv_permsocks, xpt_list) {
1287 1288
		if (xprt->xpt_net != net)
			continue;
1289 1290 1291 1292
		if (strcmp(xprt->xpt_class->xcl_name, xcl_name))
			continue;
		if (af != AF_UNSPEC && af != xprt->xpt_local.ss_family)
			continue;
1293
		if (port != 0 && port != svc_xprt_local_port(xprt))
1294 1295
			continue;
		found = xprt;
1296
		svc_xprt_get(xprt);
1297 1298 1299 1300 1301 1302
		break;
	}
	spin_unlock_bh(&serv->sv_lock);
	return found;
}
EXPORT_SYMBOL_GPL(svc_find_xprt);
1303

1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327
static int svc_one_xprt_name(const struct svc_xprt *xprt,
			     char *pos, int remaining)
{
	int len;

	len = snprintf(pos, remaining, "%s %u\n",
			xprt->xpt_class->xcl_name,
			svc_xprt_local_port(xprt));
	if (len >= remaining)
		return -ENAMETOOLONG;
	return len;
}

/**
 * svc_xprt_names - format a buffer with a list of transport names
 * @serv: pointer to an RPC service
 * @buf: pointer to a buffer to be filled in
 * @buflen: length of buffer to be filled in
 *
 * Fills in @buf with a string containing a list of transport names,
 * each name terminated with '\n'.
 *
 * Returns positive length of the filled-in string on success; otherwise
 * a negative errno value is returned if an error occurs.
1328
 */
1329
int svc_xprt_names(struct svc_serv *serv, char *buf, const int buflen)
1330 1331
{
	struct svc_xprt *xprt;
1332 1333
	int len, totlen;
	char *pos;
1334 1335 1336 1337 1338 1339

	/* Sanity check args */
	if (!serv)
		return 0;

	spin_lock_bh(&serv->sv_lock);
1340 1341 1342

	pos = buf;
	totlen = 0;
1343
	list_for_each_entry(xprt, &serv->sv_permsocks, xpt_list) {
1344 1345 1346 1347 1348 1349
		len = svc_one_xprt_name(xprt, pos, buflen - totlen);
		if (len < 0) {
			*buf = '\0';
			totlen = len;
		}
		if (len <= 0)
1350
			break;
1351 1352

		pos += len;
1353 1354
		totlen += len;
	}
1355

1356 1357 1358 1359
	spin_unlock_bh(&serv->sv_lock);
	return totlen;
}
EXPORT_SYMBOL_GPL(svc_xprt_names);
1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404


/*----------------------------------------------------------------------------*/

static void *svc_pool_stats_start(struct seq_file *m, loff_t *pos)
{
	unsigned int pidx = (unsigned int)*pos;
	struct svc_serv *serv = m->private;

	dprintk("svc_pool_stats_start, *pidx=%u\n", pidx);

	if (!pidx)
		return SEQ_START_TOKEN;
	return (pidx > serv->sv_nrpools ? NULL : &serv->sv_pools[pidx-1]);
}

static void *svc_pool_stats_next(struct seq_file *m, void *p, loff_t *pos)
{
	struct svc_pool *pool = p;
	struct svc_serv *serv = m->private;

	dprintk("svc_pool_stats_next, *pos=%llu\n", *pos);

	if (p == SEQ_START_TOKEN) {
		pool = &serv->sv_pools[0];
	} else {
		unsigned int pidx = (pool - &serv->sv_pools[0]);
		if (pidx < serv->sv_nrpools-1)
			pool = &serv->sv_pools[pidx+1];
		else
			pool = NULL;
	}
	++*pos;
	return pool;
}

static void svc_pool_stats_stop(struct seq_file *m, void *p)
{
}

static int svc_pool_stats_show(struct seq_file *m, void *p)
{
	struct svc_pool *pool = p;

	if (p == SEQ_START_TOKEN) {
1405
		seq_puts(m, "# pool packets-arrived sockets-enqueued threads-woken threads-timedout\n");
1406 1407 1408
		return 0;
	}

1409
	seq_printf(m, "%u %lu %lu %lu %lu\n",
1410
		pool->sp_id,
1411
		(unsigned long)atomic_long_read(&pool->sp_stats.packets),
1412
		pool->sp_stats.sockets_queued,
1413 1414
		(unsigned long)atomic_long_read(&pool->sp_stats.threads_woken),
		(unsigned long)atomic_long_read(&pool->sp_stats.threads_timedout));
1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437

	return 0;
}

static const struct seq_operations svc_pool_stats_seq_ops = {
	.start	= svc_pool_stats_start,
	.next	= svc_pool_stats_next,
	.stop	= svc_pool_stats_stop,
	.show	= svc_pool_stats_show,
};

int svc_pool_stats_open(struct svc_serv *serv, struct file *file)
{
	int err;

	err = seq_open(file, &svc_pool_stats_seq_ops);
	if (!err)
		((struct seq_file *) file->private_data)->private = serv;
	return err;
}
EXPORT_SYMBOL(svc_pool_stats_open);

/*----------------------------------------------------------------------------*/