1 /* 2 * Copyright (c) 2005 Voltaire Inc. All rights reserved. 3 * Copyright (c) 2002-2005, Network Appliance, Inc. All rights reserved. 4 * Copyright (c) 1999-2005, Mellanox Technologies, Inc. All rights reserved. 5 * Copyright (c) 2005 Intel Corporation. All rights reserved. 6 * 7 * This software is available to you under a choice of one of two 8 * licenses. You may choose to be licensed under the terms of the GNU 9 * General Public License (GPL) Version 2, available from the file 10 * COPYING in the main directory of this source tree, or the 11 * OpenIB.org BSD license below: 12 * 13 * Redistribution and use in source and binary forms, with or 14 * without modification, are permitted provided that the following 15 * conditions are met: 16 * 17 * - Redistributions of source code must retain the above 18 * copyright notice, this list of conditions and the following 19 * disclaimer. 20 * 21 * - Redistributions in binary form must reproduce the above 22 * copyright notice, this list of conditions and the following 23 * disclaimer in the documentation and/or other materials 24 * provided with the distribution. 25 * 26 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, 27 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF 28 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND 29 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS 30 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN 31 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN 32 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE 33 * SOFTWARE. 34 */ 35 36 #include <linux/mutex.h> 37 #include <linux/inetdevice.h> 38 #include <linux/slab.h> 39 #include <linux/workqueue.h> 40 #include <net/arp.h> 41 #include <net/neighbour.h> 42 #include <net/route.h> 43 #include <net/netevent.h> 44 #include <net/addrconf.h> 45 #include <net/ip6_route.h> 46 #include <rdma/ib_addr.h> 47 48 MODULE_AUTHOR("Sean Hefty"); 49 MODULE_DESCRIPTION("IB Address Translation"); 50 MODULE_LICENSE("Dual BSD/GPL"); 51 52 struct addr_req { 53 struct list_head list; 54 struct sockaddr_storage src_addr; 55 struct sockaddr_storage dst_addr; 56 struct rdma_dev_addr *addr; 57 struct rdma_addr_client *client; 58 void *context; 59 void (*callback)(int status, struct sockaddr *src_addr, 60 struct rdma_dev_addr *addr, void *context); 61 unsigned long timeout; 62 int status; 63 }; 64 65 static void process_req(struct work_struct *work); 66 67 static DEFINE_MUTEX(lock); 68 static LIST_HEAD(req_list); 69 static DECLARE_DELAYED_WORK(work, process_req); 70 static struct workqueue_struct *addr_wq; 71 72 void rdma_addr_register_client(struct rdma_addr_client *client) 73 { 74 atomic_set(&client->refcount, 1); 75 init_completion(&client->comp); 76 } 77 EXPORT_SYMBOL(rdma_addr_register_client); 78 79 static inline void put_client(struct rdma_addr_client *client) 80 { 81 if (atomic_dec_and_test(&client->refcount)) 82 complete(&client->comp); 83 } 84 85 void rdma_addr_unregister_client(struct rdma_addr_client *client) 86 { 87 put_client(client); 88 wait_for_completion(&client->comp); 89 } 90 EXPORT_SYMBOL(rdma_addr_unregister_client); 91 92 int rdma_copy_addr(struct rdma_dev_addr *dev_addr, struct net_device *dev, 93 const unsigned char *dst_dev_addr) 94 { 95 dev_addr->dev_type = dev->type; 96 memcpy(dev_addr->src_dev_addr, dev->dev_addr, MAX_ADDR_LEN); 97 memcpy(dev_addr->broadcast, dev->broadcast, MAX_ADDR_LEN); 98 if (dst_dev_addr) 99 memcpy(dev_addr->dst_dev_addr, dst_dev_addr, MAX_ADDR_LEN); 100 dev_addr->bound_dev_if = dev->ifindex; 101 return 0; 102 } 103 EXPORT_SYMBOL(rdma_copy_addr); 104 105 int rdma_translate_ip(struct sockaddr *addr, struct rdma_dev_addr *dev_addr) 106 { 107 struct net_device *dev; 108 int ret = -EADDRNOTAVAIL; 109 110 if (dev_addr->bound_dev_if) { 111 dev = dev_get_by_index(&init_net, dev_addr->bound_dev_if); 112 if (!dev) 113 return -ENODEV; 114 ret = rdma_copy_addr(dev_addr, dev, NULL); 115 dev_put(dev); 116 return ret; 117 } 118 119 switch (addr->sa_family) { 120 case AF_INET: 121 dev = ip_dev_find(&init_net, 122 ((struct sockaddr_in *) addr)->sin_addr.s_addr); 123 124 if (!dev) 125 return ret; 126 127 ret = rdma_copy_addr(dev_addr, dev, NULL); 128 dev_put(dev); 129 break; 130 131 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) 132 case AF_INET6: 133 rcu_read_lock(); 134 for_each_netdev_rcu(&init_net, dev) { 135 if (ipv6_chk_addr(&init_net, 136 &((struct sockaddr_in6 *) addr)->sin6_addr, 137 dev, 1)) { 138 ret = rdma_copy_addr(dev_addr, dev, NULL); 139 break; 140 } 141 } 142 rcu_read_unlock(); 143 break; 144 #endif 145 } 146 return ret; 147 } 148 EXPORT_SYMBOL(rdma_translate_ip); 149 150 static void set_timeout(unsigned long time) 151 { 152 unsigned long delay; 153 154 cancel_delayed_work(&work); 155 156 delay = time - jiffies; 157 if ((long)delay <= 0) 158 delay = 1; 159 160 queue_delayed_work(addr_wq, &work, delay); 161 } 162 163 static void queue_req(struct addr_req *req) 164 { 165 struct addr_req *temp_req; 166 167 mutex_lock(&lock); 168 list_for_each_entry_reverse(temp_req, &req_list, list) { 169 if (time_after_eq(req->timeout, temp_req->timeout)) 170 break; 171 } 172 173 list_add(&req->list, &temp_req->list); 174 175 if (req_list.next == &req->list) 176 set_timeout(req->timeout); 177 mutex_unlock(&lock); 178 } 179 180 static int addr4_resolve(struct sockaddr_in *src_in, 181 struct sockaddr_in *dst_in, 182 struct rdma_dev_addr *addr) 183 { 184 __be32 src_ip = src_in->sin_addr.s_addr; 185 __be32 dst_ip = dst_in->sin_addr.s_addr; 186 struct rtable *rt; 187 struct neighbour *neigh; 188 int ret; 189 190 rt = ip_route_output(&init_net, dst_ip, src_ip, 0, addr->bound_dev_if); 191 if (IS_ERR(rt)) { 192 ret = PTR_ERR(rt); 193 goto out; 194 } 195 src_in->sin_family = AF_INET; 196 src_in->sin_addr.s_addr = rt->rt_src; 197 198 if (rt->dst.dev->flags & IFF_LOOPBACK) { 199 ret = rdma_translate_ip((struct sockaddr *) dst_in, addr); 200 if (!ret) 201 memcpy(addr->dst_dev_addr, addr->src_dev_addr, MAX_ADDR_LEN); 202 goto put; 203 } 204 205 /* If the device does ARP internally, return 'done' */ 206 if (rt->dst.dev->flags & IFF_NOARP) { 207 rdma_copy_addr(addr, rt->dst.dev, NULL); 208 goto put; 209 } 210 211 neigh = neigh_lookup(&arp_tbl, &rt->rt_gateway, rt->dst.dev); 212 if (!neigh || !(neigh->nud_state & NUD_VALID)) { 213 neigh_event_send(rt->dst.neighbour, NULL); 214 ret = -ENODATA; 215 if (neigh) 216 goto release; 217 goto put; 218 } 219 220 ret = rdma_copy_addr(addr, neigh->dev, neigh->ha); 221 release: 222 neigh_release(neigh); 223 put: 224 ip_rt_put(rt); 225 out: 226 return ret; 227 } 228 229 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) 230 static int addr6_resolve(struct sockaddr_in6 *src_in, 231 struct sockaddr_in6 *dst_in, 232 struct rdma_dev_addr *addr) 233 { 234 struct flowi6 fl6; 235 struct neighbour *neigh; 236 struct dst_entry *dst; 237 int ret; 238 239 memset(&fl6, 0, sizeof fl6); 240 ipv6_addr_copy(&fl6.daddr, &dst_in->sin6_addr); 241 ipv6_addr_copy(&fl6.saddr, &src_in->sin6_addr); 242 fl6.flowi6_oif = addr->bound_dev_if; 243 244 dst = ip6_route_output(&init_net, NULL, &fl6); 245 if ((ret = dst->error)) 246 goto put; 247 248 if (ipv6_addr_any(&fl6.saddr)) { 249 ret = ipv6_dev_get_saddr(&init_net, ip6_dst_idev(dst)->dev, 250 &fl6.daddr, 0, &fl6.saddr); 251 if (ret) 252 goto put; 253 254 src_in->sin6_family = AF_INET6; 255 ipv6_addr_copy(&src_in->sin6_addr, &fl6.saddr); 256 } 257 258 if (dst->dev->flags & IFF_LOOPBACK) { 259 ret = rdma_translate_ip((struct sockaddr *) dst_in, addr); 260 if (!ret) 261 memcpy(addr->dst_dev_addr, addr->src_dev_addr, MAX_ADDR_LEN); 262 goto put; 263 } 264 265 /* If the device does ARP internally, return 'done' */ 266 if (dst->dev->flags & IFF_NOARP) { 267 ret = rdma_copy_addr(addr, dst->dev, NULL); 268 goto put; 269 } 270 271 neigh = dst->neighbour; 272 if (!neigh || !(neigh->nud_state & NUD_VALID)) { 273 neigh_event_send(dst->neighbour, NULL); 274 ret = -ENODATA; 275 goto put; 276 } 277 278 ret = rdma_copy_addr(addr, dst->dev, neigh->ha); 279 put: 280 dst_release(dst); 281 return ret; 282 } 283 #else 284 static int addr6_resolve(struct sockaddr_in6 *src_in, 285 struct sockaddr_in6 *dst_in, 286 struct rdma_dev_addr *addr) 287 { 288 return -EADDRNOTAVAIL; 289 } 290 #endif 291 292 static int addr_resolve(struct sockaddr *src_in, 293 struct sockaddr *dst_in, 294 struct rdma_dev_addr *addr) 295 { 296 if (src_in->sa_family == AF_INET) { 297 return addr4_resolve((struct sockaddr_in *) src_in, 298 (struct sockaddr_in *) dst_in, addr); 299 } else 300 return addr6_resolve((struct sockaddr_in6 *) src_in, 301 (struct sockaddr_in6 *) dst_in, addr); 302 } 303 304 static void process_req(struct work_struct *work) 305 { 306 struct addr_req *req, *temp_req; 307 struct sockaddr *src_in, *dst_in; 308 struct list_head done_list; 309 310 INIT_LIST_HEAD(&done_list); 311 312 mutex_lock(&lock); 313 list_for_each_entry_safe(req, temp_req, &req_list, list) { 314 if (req->status == -ENODATA) { 315 src_in = (struct sockaddr *) &req->src_addr; 316 dst_in = (struct sockaddr *) &req->dst_addr; 317 req->status = addr_resolve(src_in, dst_in, req->addr); 318 if (req->status && time_after_eq(jiffies, req->timeout)) 319 req->status = -ETIMEDOUT; 320 else if (req->status == -ENODATA) 321 continue; 322 } 323 list_move_tail(&req->list, &done_list); 324 } 325 326 if (!list_empty(&req_list)) { 327 req = list_entry(req_list.next, struct addr_req, list); 328 set_timeout(req->timeout); 329 } 330 mutex_unlock(&lock); 331 332 list_for_each_entry_safe(req, temp_req, &done_list, list) { 333 list_del(&req->list); 334 req->callback(req->status, (struct sockaddr *) &req->src_addr, 335 req->addr, req->context); 336 put_client(req->client); 337 kfree(req); 338 } 339 } 340 341 int rdma_resolve_ip(struct rdma_addr_client *client, 342 struct sockaddr *src_addr, struct sockaddr *dst_addr, 343 struct rdma_dev_addr *addr, int timeout_ms, 344 void (*callback)(int status, struct sockaddr *src_addr, 345 struct rdma_dev_addr *addr, void *context), 346 void *context) 347 { 348 struct sockaddr *src_in, *dst_in; 349 struct addr_req *req; 350 int ret = 0; 351 352 req = kzalloc(sizeof *req, GFP_KERNEL); 353 if (!req) 354 return -ENOMEM; 355 356 src_in = (struct sockaddr *) &req->src_addr; 357 dst_in = (struct sockaddr *) &req->dst_addr; 358 359 if (src_addr) { 360 if (src_addr->sa_family != dst_addr->sa_family) { 361 ret = -EINVAL; 362 goto err; 363 } 364 365 memcpy(src_in, src_addr, ip_addr_size(src_addr)); 366 } else { 367 src_in->sa_family = dst_addr->sa_family; 368 } 369 370 memcpy(dst_in, dst_addr, ip_addr_size(dst_addr)); 371 req->addr = addr; 372 req->callback = callback; 373 req->context = context; 374 req->client = client; 375 atomic_inc(&client->refcount); 376 377 req->status = addr_resolve(src_in, dst_in, addr); 378 switch (req->status) { 379 case 0: 380 req->timeout = jiffies; 381 queue_req(req); 382 break; 383 case -ENODATA: 384 req->timeout = msecs_to_jiffies(timeout_ms) + jiffies; 385 queue_req(req); 386 break; 387 default: 388 ret = req->status; 389 atomic_dec(&client->refcount); 390 goto err; 391 } 392 return ret; 393 err: 394 kfree(req); 395 return ret; 396 } 397 EXPORT_SYMBOL(rdma_resolve_ip); 398 399 void rdma_addr_cancel(struct rdma_dev_addr *addr) 400 { 401 struct addr_req *req, *temp_req; 402 403 mutex_lock(&lock); 404 list_for_each_entry_safe(req, temp_req, &req_list, list) { 405 if (req->addr == addr) { 406 req->status = -ECANCELED; 407 req->timeout = jiffies; 408 list_move(&req->list, &req_list); 409 set_timeout(req->timeout); 410 break; 411 } 412 } 413 mutex_unlock(&lock); 414 } 415 EXPORT_SYMBOL(rdma_addr_cancel); 416 417 static int netevent_callback(struct notifier_block *self, unsigned long event, 418 void *ctx) 419 { 420 if (event == NETEVENT_NEIGH_UPDATE) { 421 struct neighbour *neigh = ctx; 422 423 if (neigh->nud_state & NUD_VALID) { 424 set_timeout(jiffies); 425 } 426 } 427 return 0; 428 } 429 430 static struct notifier_block nb = { 431 .notifier_call = netevent_callback 432 }; 433 434 static int __init addr_init(void) 435 { 436 addr_wq = create_singlethread_workqueue("ib_addr"); 437 if (!addr_wq) 438 return -ENOMEM; 439 440 register_netevent_notifier(&nb); 441 return 0; 442 } 443 444 static void __exit addr_cleanup(void) 445 { 446 unregister_netevent_notifier(&nb); 447 destroy_workqueue(addr_wq); 448 } 449 450 module_init(addr_init); 451 module_exit(addr_cleanup); 452