Merge tag 'jfs-6.8' of github.com:kleikamp/linux-shaggy
[linux-2.6-block.git] / net / rxrpc / local_object.c
CommitLineData
b4d0d230 1// SPDX-License-Identifier: GPL-2.0-or-later
87563616 2/* Local endpoint object management
17926a79 3 *
4f95dd78 4 * Copyright (C) 2016 Red Hat, Inc. All Rights Reserved.
17926a79 5 * Written by David Howells (dhowells@redhat.com)
17926a79
DH
6 */
7
9b6d5398
JP
8#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
9
17926a79
DH
10#include <linux/module.h>
11#include <linux/net.h>
12#include <linux/skbuff.h>
5a0e3ad6 13#include <linux/slab.h>
44ba0698
DH
14#include <linux/udp.h>
15#include <linux/ip.h>
4f95dd78 16#include <linux/hashtable.h>
17926a79 17#include <net/sock.h>
5271953c 18#include <net/udp.h>
5d30c626 19#include <net/udp_tunnel.h>
17926a79
DH
20#include <net/af_rxrpc.h>
21#include "ar-internal.h"
22
4f95dd78 23static void rxrpc_local_rcu(struct rcu_head *);
17926a79 24
b6c66c43
DH
25/*
26 * Handle an ICMP/ICMP6 error turning up at the tunnel. Push it through the
27 * usual mechanism so that it gets parsed and presented through the UDP
28 * socket's error_report().
29 */
30static void rxrpc_encap_err_rcv(struct sock *sk, struct sk_buff *skb, int err,
31 __be16 port, u32 info, u8 *payload)
32{
33 if (ip_hdr(skb)->version == IPVERSION)
34 return ip_icmp_error(sk, skb, err, port, info, payload);
41cf3a91
DH
35 if (IS_ENABLED(CONFIG_AF_RXRPC_IPV6))
36 return ipv6_icmp_error(sk, skb, err, port, info, payload);
b6c66c43
DH
37}
38
17926a79 39/*
4f95dd78
DH
40 * Compare a local to an address. Return -ve, 0 or +ve to indicate less than,
41 * same or greater than.
42 *
43 * We explicitly don't compare the RxRPC service ID as we want to reject
44 * conflicting uses by differing services. Further, we don't want to share
45 * addresses with different options (IPv6), so we don't compare those bits
46 * either.
17926a79 47 */
4f95dd78
DH
48static long rxrpc_local_cmp_key(const struct rxrpc_local *local,
49 const struct sockaddr_rxrpc *srx)
50{
51 long diff;
52
53 diff = ((local->srx.transport_type - srx->transport_type) ?:
54 (local->srx.transport_len - srx->transport_len) ?:
55 (local->srx.transport.family - srx->transport.family));
56 if (diff != 0)
57 return diff;
58
59 switch (srx->transport.family) {
60 case AF_INET:
61 /* If the choice of UDP port is left up to the transport, then
62 * the endpoint record doesn't match.
63 */
64 return ((u16 __force)local->srx.transport.sin.sin_port -
65 (u16 __force)srx->transport.sin.sin_port) ?:
66 memcmp(&local->srx.transport.sin.sin_addr,
67 &srx->transport.sin.sin_addr,
68 sizeof(struct in_addr));
d1912747 69#ifdef CONFIG_AF_RXRPC_IPV6
75b54cb5
DH
70 case AF_INET6:
71 /* If the choice of UDP6 port is left up to the transport, then
72 * the endpoint record doesn't match.
73 */
74 return ((u16 __force)local->srx.transport.sin6.sin6_port -
75 (u16 __force)srx->transport.sin6.sin6_port) ?:
76 memcmp(&local->srx.transport.sin6.sin6_addr,
77 &srx->transport.sin6.sin6_addr,
78 sizeof(struct in6_addr));
d1912747 79#endif
4f95dd78
DH
80 default:
81 BUG();
82 }
83}
84
0d6bf319
DH
85static void rxrpc_client_conn_reap_timeout(struct timer_list *timer)
86{
87 struct rxrpc_local *local =
88 container_of(timer, struct rxrpc_local, client_conn_reap_timer);
89
61e4a866 90 if (!local->kill_all_client_conns &&
0d6bf319
DH
91 test_and_set_bit(RXRPC_CLIENT_CONN_REAP_TIMER, &local->client_conn_flags))
92 rxrpc_wake_up_io_thread(local);
93}
94
4f95dd78
DH
95/*
96 * Allocate a new local endpoint.
97 */
8a758d98 98static struct rxrpc_local *rxrpc_alloc_local(struct net *net,
2baec2c3 99 const struct sockaddr_rxrpc *srx)
17926a79
DH
100{
101 struct rxrpc_local *local;
f06cb291 102 u32 tmp;
17926a79
DH
103
104 local = kzalloc(sizeof(struct rxrpc_local), GFP_KERNEL);
105 if (local) {
a0575429 106 refcount_set(&local->ref, 1);
730c5fd4 107 atomic_set(&local->active_users, 1);
8a758d98
DH
108 local->net = net;
109 local->rxnet = rxrpc_net(net);
33912c26 110 INIT_HLIST_NODE(&local->link);
8fbcc833 111 init_completion(&local->io_thread_ready);
af094824
DH
112#ifdef CONFIG_AF_RXRPC_INJECT_RX_DELAY
113 skb_queue_head_init(&local->rx_delay_queue);
114#endif
a275da62 115 skb_queue_head_init(&local->rx_queue);
f2cce89a 116 INIT_LIST_HEAD(&local->conn_attend_q);
15f661dc 117 INIT_LIST_HEAD(&local->call_attend_q);
0d6bf319 118
245500d8
DH
119 local->client_bundles = RB_ROOT;
120 spin_lock_init(&local->client_bundles_lock);
0d6bf319 121 local->kill_all_client_conns = false;
0d6bf319
DH
122 INIT_LIST_HEAD(&local->idle_client_conns);
123 timer_setup(&local->client_conn_reap_timer,
124 rxrpc_client_conn_reap_timeout, 0);
125
17926a79
DH
126 spin_lock_init(&local->lock);
127 rwlock_init(&local->services_lock);
17926a79
DH
128 local->debug_id = atomic_inc_return(&rxrpc_debug_id);
129 memcpy(&local->srx, srx, sizeof(*srx));
28036f44 130 local->srx.srx_service = 0;
f06cb291
DH
131 idr_init(&local->conn_ids);
132 get_random_bytes(&tmp, sizeof(tmp));
133 tmp &= 0x3fffffff;
134 if (tmp == 0)
135 tmp = 1;
136 idr_set_cursor(&local->conn_ids, tmp);
9d35d880
DH
137 INIT_LIST_HEAD(&local->new_client_calls);
138 spin_lock_init(&local->client_call_lock);
f06cb291 139
0fde882f 140 trace_rxrpc_local(local->debug_id, rxrpc_local_new, 1, 1);
17926a79
DH
141 }
142
143 _leave(" = %p", local);
144 return local;
145}
146
147/*
148 * create the local socket
4f95dd78 149 * - must be called with rxrpc_local_mutex locked
17926a79 150 */
2baec2c3 151static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net)
17926a79 152{
1a9b86c9
XL
153 struct udp_tunnel_sock_cfg tuncfg = {NULL};
154 struct sockaddr_rxrpc *srx = &local->srx;
155 struct udp_port_cfg udp_conf = {0};
a275da62 156 struct task_struct *io_thread;
5271953c 157 struct sock *usk;
fce93494 158 int ret;
17926a79 159
75b54cb5 160 _enter("%p{%d,%d}",
1a9b86c9
XL
161 local, srx->transport_type, srx->transport.family);
162
163 udp_conf.family = srx->transport.family;
39cb9faa 164 udp_conf.use_udp_checksums = true;
1a9b86c9
XL
165 if (udp_conf.family == AF_INET) {
166 udp_conf.local_ip = srx->transport.sin.sin_addr;
167 udp_conf.local_udp_port = srx->transport.sin.sin_port;
295f830e 168#if IS_ENABLED(CONFIG_AF_RXRPC_IPV6)
1a9b86c9
XL
169 } else {
170 udp_conf.local_ip6 = srx->transport.sin6.sin6_addr;
171 udp_conf.local_udp_port = srx->transport.sin6.sin6_port;
39cb9faa
DH
172 udp_conf.use_udp6_tx_checksums = true;
173 udp_conf.use_udp6_rx_checksums = true;
295f830e 174#endif
1a9b86c9
XL
175 }
176 ret = udp_sock_create(net, &udp_conf, &local->socket);
17926a79
DH
177 if (ret < 0) {
178 _leave(" = %d [socket]", ret);
179 return ret;
180 }
181
1a9b86c9 182 tuncfg.encap_type = UDP_ENCAP_RXRPC;
446b3e14 183 tuncfg.encap_rcv = rxrpc_encap_rcv;
ac56a0b4 184 tuncfg.encap_err_rcv = rxrpc_encap_err_rcv;
1a9b86c9
XL
185 tuncfg.sk_user_data = local;
186 setup_udp_tunnel_sock(net, local->socket, &tuncfg);
187
2cfa2271 188 /* set the socket up */
5271953c 189 usk = local->socket->sk;
5271953c 190 usk->sk_error_report = rxrpc_error_report;
2cfa2271 191
1a9b86c9 192 switch (srx->transport.family) {
37a675e7
DH
193 case AF_INET6:
194 /* we want to receive ICMPv6 errors */
1a9b86c9 195 ip6_sock_set_recverr(usk);
17926a79 196
37a675e7
DH
197 /* Fall through and set IPv4 options too otherwise we don't get
198 * errors from IPv4 packets sent through the IPv6 socket.
199 */
df561f66 200 fallthrough;
37a675e7 201 case AF_INET:
f2aeed3a 202 /* we want to receive ICMP errors */
1a9b86c9 203 ip_sock_set_recverr(usk);
f2aeed3a
DH
204
205 /* we want to set the don't fragment bit */
1a9b86c9 206 ip_sock_set_mtu_discover(usk, IP_PMTUDISC_DO);
b604dd98
DH
207
208 /* We want receive timestamps. */
1a9b86c9 209 sock_enable_timestamps(usk);
f2aeed3a
DH
210 break;
211
212 default:
213 BUG();
17926a79
DH
214 }
215
a275da62
DH
216 io_thread = kthread_run(rxrpc_io_thread, local,
217 "krxrpcio/%u", ntohs(udp_conf.local_udp_port));
218 if (IS_ERR(io_thread)) {
219 ret = PTR_ERR(io_thread);
220 goto error_sock;
221 }
222
8fbcc833 223 wait_for_completion(&local->io_thread_ready);
a275da62 224 local->io_thread = io_thread;
17926a79
DH
225 _leave(" = 0");
226 return 0;
a275da62
DH
227
228error_sock:
229 kernel_sock_shutdown(local->socket, SHUT_RDWR);
230 local->socket->sk->sk_user_data = NULL;
231 sock_release(local->socket);
232 local->socket = NULL;
233 return ret;
17926a79
DH
234}
235
236/*
4f95dd78 237 * Look up or create a new local endpoint using the specified local address.
17926a79 238 */
2baec2c3
DH
239struct rxrpc_local *rxrpc_lookup_local(struct net *net,
240 const struct sockaddr_rxrpc *srx)
17926a79
DH
241{
242 struct rxrpc_local *local;
2baec2c3 243 struct rxrpc_net *rxnet = rxrpc_net(net);
33912c26 244 struct hlist_node *cursor;
4f95dd78 245 long diff;
17926a79
DH
246 int ret;
247
75b54cb5
DH
248 _enter("{%d,%d,%pISp}",
249 srx->transport_type, srx->transport.family, &srx->transport);
17926a79 250
2baec2c3 251 mutex_lock(&rxnet->local_mutex);
17926a79 252
33912c26
DH
253 hlist_for_each(cursor, &rxnet->local_endpoints) {
254 local = hlist_entry(cursor, struct rxrpc_local, link);
17926a79 255
4f95dd78 256 diff = rxrpc_local_cmp_key(local, srx);
33912c26 257 if (diff != 0)
17926a79 258 continue;
4f95dd78
DH
259
260 /* Services aren't allowed to share transport sockets, so
261 * reject that here. It is possible that the object is dying -
262 * but it may also still have the local transport address that
263 * we want bound.
264 */
265 if (srx->srx_service) {
266 local = NULL;
267 goto addr_in_use;
268 }
17926a79 269
33912c26
DH
270 /* Found a match. We want to replace a dying object.
271 * Attempting to bind the transport socket may still fail if
272 * we're attempting to use a local address that the dying
273 * object is still using.
4f95dd78 274 */
0fde882f 275 if (!rxrpc_use_local(local, rxrpc_local_use_lookup))
4f95dd78 276 break;
17926a79 277
4f95dd78
DH
278 goto found;
279 }
17926a79 280
8a758d98 281 local = rxrpc_alloc_local(net, srx);
4f95dd78
DH
282 if (!local)
283 goto nomem;
17926a79 284
2baec2c3 285 ret = rxrpc_open_socket(local, net);
4f95dd78
DH
286 if (ret < 0)
287 goto sock_error;
288
33912c26
DH
289 if (cursor) {
290 hlist_replace_rcu(cursor, &local->link);
291 cursor->pprev = NULL;
292 } else {
293 hlist_add_head_rcu(&local->link, &rxnet->local_endpoints);
294 }
17926a79 295
4f95dd78 296found:
2baec2c3 297 mutex_unlock(&rxnet->local_mutex);
4f95dd78 298 _leave(" = %p", local);
17926a79
DH
299 return local;
300
4f95dd78
DH
301nomem:
302 ret = -ENOMEM;
303sock_error:
2baec2c3 304 mutex_unlock(&rxnet->local_mutex);
032be5f1
ED
305 if (local)
306 call_rcu(&local->rcu, rxrpc_local_rcu);
4f95dd78
DH
307 _leave(" = %d", ret);
308 return ERR_PTR(ret);
17926a79 309
4f95dd78 310addr_in_use:
2baec2c3 311 mutex_unlock(&rxnet->local_mutex);
4f95dd78
DH
312 _leave(" = -EADDRINUSE");
313 return ERR_PTR(-EADDRINUSE);
314}
17926a79 315
09d2bf59
DH
316/*
317 * Get a ref on a local endpoint.
318 */
0fde882f
DH
319struct rxrpc_local *rxrpc_get_local(struct rxrpc_local *local,
320 enum rxrpc_local_trace why)
09d2bf59 321{
0fde882f 322 int r, u;
09d2bf59 323
0fde882f 324 u = atomic_read(&local->active_users);
a0575429 325 __refcount_inc(&local->ref, &r);
0fde882f 326 trace_rxrpc_local(local->debug_id, why, r + 1, u);
09d2bf59
DH
327 return local;
328}
329
330/*
331 * Get a ref on a local endpoint unless its usage has already reached 0.
332 */
0fde882f
DH
333struct rxrpc_local *rxrpc_get_local_maybe(struct rxrpc_local *local,
334 enum rxrpc_local_trace why)
09d2bf59 335{
0fde882f 336 int r, u;
09d2bf59 337
0fde882f
DH
338 if (local && __refcount_inc_not_zero(&local->ref, &r)) {
339 u = atomic_read(&local->active_users);
340 trace_rxrpc_local(local->debug_id, why, r + 1, u);
341 return local;
09d2bf59 342 }
0fde882f
DH
343
344 return NULL;
09d2bf59
DH
345}
346
09d2bf59
DH
347/*
348 * Drop a ref on a local endpoint.
349 */
0fde882f 350void rxrpc_put_local(struct rxrpc_local *local, enum rxrpc_local_trace why)
09d2bf59 351{
fac20b9e 352 unsigned int debug_id;
a0575429 353 bool dead;
0fde882f 354 int r, u;
09d2bf59
DH
355
356 if (local) {
fac20b9e
DH
357 debug_id = local->debug_id;
358
0fde882f 359 u = atomic_read(&local->active_users);
a0575429 360 dead = __refcount_dec_and_test(&local->ref, &r);
0fde882f 361 trace_rxrpc_local(debug_id, why, r, u);
09d2bf59 362
a0575429 363 if (dead)
730c5fd4 364 call_rcu(&local->rcu, rxrpc_local_rcu);
09d2bf59
DH
365 }
366}
367
730c5fd4
DH
368/*
369 * Start using a local endpoint.
370 */
0fde882f
DH
371struct rxrpc_local *rxrpc_use_local(struct rxrpc_local *local,
372 enum rxrpc_local_trace why)
730c5fd4 373{
0fde882f 374 local = rxrpc_get_local_maybe(local, rxrpc_local_get_for_use);
730c5fd4
DH
375 if (!local)
376 return NULL;
377
0fde882f
DH
378 if (!__rxrpc_use_local(local, why)) {
379 rxrpc_put_local(local, rxrpc_local_put_for_use);
730c5fd4
DH
380 return NULL;
381 }
382
383 return local;
384}
385
386/*
387 * Cease using a local endpoint. Once the number of active users reaches 0, we
5e6ef4f1 388 * start the closure of the transport in the I/O thread..
730c5fd4 389 */
0fde882f 390void rxrpc_unuse_local(struct rxrpc_local *local, enum rxrpc_local_trace why)
730c5fd4 391{
eaa02390 392 unsigned int debug_id;
a2cf3264
DH
393 int r, u;
394
395 if (local) {
eaa02390 396 debug_id = local->debug_id;
a2cf3264
DH
397 r = refcount_read(&local->ref);
398 u = atomic_dec_return(&local->active_users);
399 trace_rxrpc_local(debug_id, why, r, u);
400 if (u == 0)
401 kthread_stop(local->io_thread);
402 }
730c5fd4
DH
403}
404
17926a79 405/*
4f95dd78
DH
406 * Destroy a local endpoint's socket and then hand the record to RCU to dispose
407 * of.
408 *
409 * Closing the socket cannot be done from bottom half context or RCU callback
410 * context because it might sleep.
17926a79 411 */
a275da62 412void rxrpc_destroy_local(struct rxrpc_local *local)
17926a79 413{
4f95dd78 414 struct socket *socket = local->socket;
2baec2c3 415 struct rxrpc_net *rxnet = local->rxnet;
17926a79 416
4f95dd78 417 _enter("%d", local->debug_id);
17926a79 418
d12040b6
DH
419 local->dead = true;
420
2baec2c3 421 mutex_lock(&rxnet->local_mutex);
33912c26 422 hlist_del_init_rcu(&local->link);
2baec2c3 423 mutex_unlock(&rxnet->local_mutex);
4f95dd78 424
d12040b6
DH
425 rxrpc_clean_up_local_conns(local);
426 rxrpc_service_connection_reaper(&rxnet->service_conn_reaper);
1e9e5c95 427 ASSERT(!local->service);
4f95dd78
DH
428
429 if (socket) {
430 local->socket = NULL;
431 kernel_sock_shutdown(socket, SHUT_RDWR);
432 socket->sk->sk_user_data = NULL;
433 sock_release(socket);
434 }
435
436 /* At this point, there should be no more packets coming in to the
437 * local endpoint.
438 */
af094824
DH
439#ifdef CONFIG_AF_RXRPC_INJECT_RX_DELAY
440 rxrpc_purge_queue(&local->rx_delay_queue);
441#endif
a275da62 442 rxrpc_purge_queue(&local->rx_queue);
9d35d880 443 rxrpc_purge_client_connections(local);
17926a79
DH
444}
445
4f95dd78
DH
446/*
447 * Destroy a local endpoint after the RCU grace period expires.
448 */
449static void rxrpc_local_rcu(struct rcu_head *rcu)
450{
451 struct rxrpc_local *local = container_of(rcu, struct rxrpc_local, rcu);
17926a79 452
0fde882f 453 rxrpc_see_local(local, rxrpc_local_free);
17926a79 454 kfree(local);
17926a79
DH
455}
456
457/*
4f95dd78 458 * Verify the local endpoint list is empty by this point.
17926a79 459 */
2baec2c3 460void rxrpc_destroy_all_locals(struct rxrpc_net *rxnet)
17926a79 461{
4f95dd78 462 struct rxrpc_local *local;
17926a79
DH
463
464 _enter("");
465
dee46364 466 flush_workqueue(rxrpc_workqueue);
17926a79 467
33912c26 468 if (!hlist_empty(&rxnet->local_endpoints)) {
2baec2c3 469 mutex_lock(&rxnet->local_mutex);
33912c26 470 hlist_for_each_entry(local, &rxnet->local_endpoints, link) {
dee46364 471 pr_err("AF_RXRPC: Leaked local %p {%d}\n",
a0575429 472 local, refcount_read(&local->ref));
dee46364 473 }
2baec2c3 474 mutex_unlock(&rxnet->local_mutex);
dee46364 475 BUG();
17926a79 476 }
17926a79 477}