local_object.c 11.5 KB
Newer Older
1
/* Local endpoint object management
2
 *
3
 * Copyright (C) 2016 Red Hat, Inc. All Rights Reserved.
4 5 6
 * Written by David Howells (dhowells@redhat.com)
 *
 * This program is free software; you can redistribute it and/or
7
 * modify it under the terms of the GNU General Public Licence
8
 * as published by the Free Software Foundation; either version
9
 * 2 of the Licence, or (at your option) any later version.
10 11
 */

12 13
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt

14 15 16
#include <linux/module.h>
#include <linux/net.h>
#include <linux/skbuff.h>
17
#include <linux/slab.h>
18 19
#include <linux/udp.h>
#include <linux/ip.h>
20
#include <linux/hashtable.h>
21 22 23 24
#include <net/sock.h>
#include <net/af_rxrpc.h>
#include "ar-internal.h"

25 26
static void rxrpc_local_processor(struct work_struct *);
static void rxrpc_local_rcu(struct rcu_head *);
27 28

/*
29 30 31 32 33 34 35
 * Compare a local to an address.  Return -ve, 0 or +ve to indicate less than,
 * same or greater than.
 *
 * We explicitly don't compare the RxRPC service ID as we want to reject
 * conflicting uses by differing services.  Further, we don't want to share
 * addresses with different options (IPv6), so we don't compare those bits
 * either.
36
 */
37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57
static long rxrpc_local_cmp_key(const struct rxrpc_local *local,
				const struct sockaddr_rxrpc *srx)
{
	long diff;

	diff = ((local->srx.transport_type - srx->transport_type) ?:
		(local->srx.transport_len - srx->transport_len) ?:
		(local->srx.transport.family - srx->transport.family));
	if (diff != 0)
		return diff;

	switch (srx->transport.family) {
	case AF_INET:
		/* If the choice of UDP port is left up to the transport, then
		 * the endpoint record doesn't match.
		 */
		return ((u16 __force)local->srx.transport.sin.sin_port -
			(u16 __force)srx->transport.sin.sin_port) ?:
			memcmp(&local->srx.transport.sin.sin_addr,
			       &srx->transport.sin.sin_addr,
			       sizeof(struct in_addr));
58
#ifdef CONFIG_AF_RXRPC_IPV6
D
David Howells 已提交
59 60 61 62 63 64 65 66 67
	case AF_INET6:
		/* If the choice of UDP6 port is left up to the transport, then
		 * the endpoint record doesn't match.
		 */
		return ((u16 __force)local->srx.transport.sin6.sin6_port -
			(u16 __force)srx->transport.sin6.sin6_port) ?:
			memcmp(&local->srx.transport.sin6.sin6_addr,
			       &srx->transport.sin6.sin6_addr,
			       sizeof(struct in6_addr));
68
#endif
69 70 71 72 73 74 75 76
	default:
		BUG();
	}
}

/*
 * Allocate a new local endpoint.
 */
77 78
static struct rxrpc_local *rxrpc_alloc_local(struct rxrpc_net *rxnet,
					     const struct sockaddr_rxrpc *srx)
79 80 81 82 83
{
	struct rxrpc_local *local;

	local = kzalloc(sizeof(struct rxrpc_local), GFP_KERNEL);
	if (local) {
84
		atomic_set(&local->usage, 1);
85
		local->rxnet = rxnet;
86
		INIT_LIST_HEAD(&local->link);
87
		INIT_WORK(&local->processor, rxrpc_local_processor);
88 89
		init_rwsem(&local->defrag_sem);
		skb_queue_head_init(&local->reject_queue);
90
		skb_queue_head_init(&local->event_queue);
91 92
		local->client_conns = RB_ROOT;
		spin_lock_init(&local->client_conns_lock);
93 94 95 96
		spin_lock_init(&local->lock);
		rwlock_init(&local->services_lock);
		local->debug_id = atomic_inc_return(&rxrpc_debug_id);
		memcpy(&local->srx, srx, sizeof(*srx));
97
		local->srx.srx_service = 0;
98
		trace_rxrpc_local(local, rxrpc_local_new, 1, NULL);
99 100 101 102 103 104 105 106
	}

	_leave(" = %p", local);
	return local;
}

/*
 * create the local socket
107
 * - must be called with rxrpc_local_mutex locked
108
 */
109
static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net)
110 111 112 113
{
	struct sock *sock;
	int ret, opt;

D
David Howells 已提交
114 115
	_enter("%p{%d,%d}",
	       local, local->srx.transport_type, local->srx.transport.family);
116 117

	/* create a socket to represent the local endpoint */
118
	ret = sock_create_kern(net, local->srx.transport.family,
119
			       local->srx.transport_type, 0, &local->socket);
120 121 122 123 124 125 126 127 128
	if (ret < 0) {
		_leave(" = %d [socket]", ret);
		return ret;
	}

	/* if a local address was supplied then bind it */
	if (local->srx.transport_len > sizeof(sa_family_t)) {
		_debug("bind");
		ret = kernel_bind(local->socket,
129
				  (struct sockaddr *)&local->srx.transport,
130 131
				  local->srx.transport_len);
		if (ret < 0) {
132
			_debug("bind failed %d", ret);
133 134 135 136
			goto error;
		}
	}

137 138 139 140 141 142 143 144 145 146
	switch (local->srx.transport.family) {
	case AF_INET:
		/* we want to receive ICMP errors */
		opt = 1;
		ret = kernel_setsockopt(local->socket, SOL_IP, IP_RECVERR,
					(char *) &opt, sizeof(opt));
		if (ret < 0) {
			_debug("setsockopt failed");
			goto error;
		}
147

148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175
		/* we want to set the don't fragment bit */
		opt = IP_PMTUDISC_DO;
		ret = kernel_setsockopt(local->socket, SOL_IP, IP_MTU_DISCOVER,
					(char *) &opt, sizeof(opt));
		if (ret < 0) {
			_debug("setsockopt failed");
			goto error;
		}
		break;

	case AF_INET6:
		/* we want to receive ICMP errors */
		opt = 1;
		ret = kernel_setsockopt(local->socket, SOL_IPV6, IPV6_RECVERR,
					(char *) &opt, sizeof(opt));
		if (ret < 0) {
			_debug("setsockopt failed");
			goto error;
		}

		/* we want to set the don't fragment bit */
		opt = IPV6_PMTUDISC_DO;
		ret = kernel_setsockopt(local->socket, SOL_IPV6, IPV6_MTU_DISCOVER,
					(char *) &opt, sizeof(opt));
		if (ret < 0) {
			_debug("setsockopt failed");
			goto error;
		}
D
David Howells 已提交
176 177 178 179 180 181 182 183 184

		/* We want receive timestamps. */
		opt = 1;
		ret = kernel_setsockopt(local->socket, SOL_SOCKET, SO_TIMESTAMPNS,
					(char *)&opt, sizeof(opt));
		if (ret < 0) {
			_debug("setsockopt failed");
			goto error;
		}
185 186 187 188
		break;

	default:
		BUG();
189 190 191 192 193 194
	}

	/* set the socket up */
	sock = local->socket->sk;
	sock->sk_user_data	= local;
	sock->sk_data_ready	= rxrpc_data_ready;
195
	sock->sk_error_report	= rxrpc_error_report;
196 197 198 199
	_leave(" = 0");
	return 0;

error:
200
	kernel_sock_shutdown(local->socket, SHUT_RDWR);
201 202 203 204 205 206 207 208 209
	local->socket->sk->sk_user_data = NULL;
	sock_release(local->socket);
	local->socket = NULL;

	_leave(" = %d", ret);
	return ret;
}

/*
210
 * Look up or create a new local endpoint using the specified local address.
211
 */
212 213
struct rxrpc_local *rxrpc_lookup_local(struct net *net,
				       const struct sockaddr_rxrpc *srx)
214 215
{
	struct rxrpc_local *local;
216
	struct rxrpc_net *rxnet = rxrpc_net(net);
217 218 219
	struct list_head *cursor;
	const char *age;
	long diff;
220 221
	int ret;

D
David Howells 已提交
222 223
	_enter("{%d,%d,%pISp}",
	       srx->transport_type, srx->transport.family, &srx->transport);
224

225
	mutex_lock(&rxnet->local_mutex);
226

227 228
	for (cursor = rxnet->local_endpoints.next;
	     cursor != &rxnet->local_endpoints;
229 230
	     cursor = cursor->next) {
		local = list_entry(cursor, struct rxrpc_local, link);
231

232 233
		diff = rxrpc_local_cmp_key(local, srx);
		if (diff < 0)
234
			continue;
235 236 237 238 239 240 241 242 243 244 245 246
		if (diff > 0)
			break;

		/* Services aren't allowed to share transport sockets, so
		 * reject that here.  It is possible that the object is dying -
		 * but it may also still have the local transport address that
		 * we want bound.
		 */
		if (srx->srx_service) {
			local = NULL;
			goto addr_in_use;
		}
247

248 249 250 251
		/* Found a match.  We replace a dying object.  Attempting to
		 * bind the transport socket may still fail if we're attempting
		 * to use a local address that the dying object is still using.
		 */
252
		if (!rxrpc_get_local_maybe(local)) {
253 254 255
			cursor = cursor->next;
			list_del_init(&local->link);
			break;
256 257
		}

258 259 260
		age = "old";
		goto found;
	}
261

262
	local = rxrpc_alloc_local(rxnet, srx);
263 264
	if (!local)
		goto nomem;
265

266
	ret = rxrpc_open_socket(local, net);
267 268 269 270 271
	if (ret < 0)
		goto sock_error;

	list_add_tail(&local->link, cursor);
	age = "new";
272

273
found:
274
	mutex_unlock(&rxnet->local_mutex);
275

D
David Howells 已提交
276 277
	_net("LOCAL %s %d {%pISp}",
	     age, local->debug_id, &local->srx.transport);
278

279
	_leave(" = %p", local);
280 281
	return local;

282 283 284
nomem:
	ret = -ENOMEM;
sock_error:
285
	mutex_unlock(&rxnet->local_mutex);
286 287 288
	kfree(local);
	_leave(" = %d", ret);
	return ERR_PTR(ret);
289

290
addr_in_use:
291
	mutex_unlock(&rxnet->local_mutex);
292 293 294
	_leave(" = -EADDRINUSE");
	return ERR_PTR(-EADDRINUSE);
}
295

296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316
/*
 * Get a ref on a local endpoint.
 */
struct rxrpc_local *rxrpc_get_local(struct rxrpc_local *local)
{
	const void *here = __builtin_return_address(0);
	int n;

	n = atomic_inc_return(&local->usage);
	trace_rxrpc_local(local, rxrpc_local_got, n, here);
	return local;
}

/*
 * Get a ref on a local endpoint unless its usage has already reached 0.
 */
struct rxrpc_local *rxrpc_get_local_maybe(struct rxrpc_local *local)
{
	const void *here = __builtin_return_address(0);

	if (local) {
317
		int n = atomic_fetch_add_unless(&local->usage, 1, 0);
318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337
		if (n > 0)
			trace_rxrpc_local(local, rxrpc_local_got, n + 1, here);
		else
			local = NULL;
	}
	return local;
}

/*
 * Queue a local endpoint.
 */
void rxrpc_queue_local(struct rxrpc_local *local)
{
	const void *here = __builtin_return_address(0);

	if (rxrpc_queue_work(&local->processor))
		trace_rxrpc_local(local, rxrpc_local_queued,
				  atomic_read(&local->usage), here);
}

338 339 340
/*
 * A local endpoint reached its end of life.
 */
341
static void __rxrpc_put_local(struct rxrpc_local *local)
342 343 344
{
	_enter("%d", local->debug_id);
	rxrpc_queue_work(&local->processor);
345 346
}

347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363
/*
 * Drop a ref on a local endpoint.
 */
void rxrpc_put_local(struct rxrpc_local *local)
{
	const void *here = __builtin_return_address(0);
	int n;

	if (local) {
		n = atomic_dec_return(&local->usage);
		trace_rxrpc_local(local, rxrpc_local_put, n, here);

		if (n == 0)
			__rxrpc_put_local(local);
	}
}

364
/*
365 366 367 368 369
 * Destroy a local endpoint's socket and then hand the record to RCU to dispose
 * of.
 *
 * Closing the socket cannot be done from bottom half context or RCU callback
 * context because it might sleep.
370
 */
371
static void rxrpc_local_destroyer(struct rxrpc_local *local)
372
{
373
	struct socket *socket = local->socket;
374
	struct rxrpc_net *rxnet = local->rxnet;
375

376
	_enter("%d", local->debug_id);
377

378 379 380 381 382 383 384
	/* We can get a race between an incoming call packet queueing the
	 * processor again and the work processor starting the destruction
	 * process which will shut down the UDP socket.
	 */
	if (local->dead) {
		_leave(" [already dead]");
		return;
385
	}
386 387
	local->dead = true;

388
	mutex_lock(&rxnet->local_mutex);
389
	list_del_init(&local->link);
390
	mutex_unlock(&rxnet->local_mutex);
391

392
	ASSERT(RB_EMPTY_ROOT(&local->client_conns));
393
	ASSERT(!local->service);
394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409

	if (socket) {
		local->socket = NULL;
		kernel_sock_shutdown(socket, SHUT_RDWR);
		socket->sk->sk_user_data = NULL;
		sock_release(socket);
	}

	/* At this point, there should be no more packets coming in to the
	 * local endpoint.
	 */
	rxrpc_purge_queue(&local->reject_queue);
	rxrpc_purge_queue(&local->event_queue);

	_debug("rcu local %d", local->debug_id);
	call_rcu(&local->rcu, rxrpc_local_rcu);
410 411 412
}

/*
413
 * Process events on an endpoint
414
 */
415
static void rxrpc_local_processor(struct work_struct *work)
416 417
{
	struct rxrpc_local *local =
418 419
		container_of(work, struct rxrpc_local, processor);
	bool again;
420

421 422
	trace_rxrpc_local(local, rxrpc_local_processing,
			  atomic_read(&local->usage), NULL);
423

424 425 426 427
	do {
		again = false;
		if (atomic_read(&local->usage) == 0)
			return rxrpc_local_destroyer(local);
428

429 430 431 432
		if (!skb_queue_empty(&local->reject_queue)) {
			rxrpc_reject_packets(local);
			again = true;
		}
433

434 435 436 437 438 439
		if (!skb_queue_empty(&local->event_queue)) {
			rxrpc_process_local_events(local);
			again = true;
		}
	} while (again);
}
440

441 442 443 444 445 446
/*
 * Destroy a local endpoint after the RCU grace period expires.
 */
static void rxrpc_local_rcu(struct rcu_head *rcu)
{
	struct rxrpc_local *local = container_of(rcu, struct rxrpc_local, rcu);
447

448
	_enter("%d", local->debug_id);
449

450
	ASSERT(!work_pending(&local->processor));
451 452 453 454 455 456 457

	_net("DESTROY LOCAL %d", local->debug_id);
	kfree(local);
	_leave("");
}

/*
458
 * Verify the local endpoint list is empty by this point.
459
 */
460
void rxrpc_destroy_all_locals(struct rxrpc_net *rxnet)
461
{
462
	struct rxrpc_local *local;
463 464 465

	_enter("");

466
	flush_workqueue(rxrpc_workqueue);
467

468 469 470
	if (!list_empty(&rxnet->local_endpoints)) {
		mutex_lock(&rxnet->local_mutex);
		list_for_each_entry(local, &rxnet->local_endpoints, link) {
471 472 473
			pr_err("AF_RXRPC: Leaked local %p {%d}\n",
			       local, atomic_read(&local->usage));
		}
474
		mutex_unlock(&rxnet->local_mutex);
475
		BUG();
476 477
	}
}