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/workqueue.h> 39 #include <net/arp.h> 40 #include <net/neighbour.h> 41 #include <net/route.h> 42 #include <net/netevent.h> 43 #include <net/addrconf.h> 44 #include <net/ip6_route.h> 45 #include <rdma/ib_addr.h> 46 47 MODULE_AUTHOR("Sean Hefty"); 48 MODULE_DESCRIPTION("IB Address Translation"); 49 MODULE_LICENSE("Dual BSD/GPL"); 50 51 struct addr_req { 52 struct list_head list; 53 struct sockaddr_storage src_addr; 54 struct sockaddr_storage dst_addr; 55 struct rdma_dev_addr *addr; 56 struct rdma_addr_client *client; 57 void *context; 58 void (*callback)(int status, struct sockaddr *src_addr, 59 struct rdma_dev_addr *addr, void *context); 60 unsigned long timeout; 61 int status; 62 }; 63 64 static void process_req(struct work_struct *work); 65 66 static DEFINE_MUTEX(lock); 67 static LIST_HEAD(req_list); 68 static DECLARE_DELAYED_WORK(work, process_req); 69 static struct workqueue_struct *addr_wq; 70 71 void rdma_addr_register_client(struct rdma_addr_client *client) 72 { 73 atomic_set(&client->refcount, 1); 74 init_completion(&client->comp); 75 } 76 EXPORT_SYMBOL(rdma_addr_register_client); 77 78 static inline void put_client(struct rdma_addr_client *client) 79 { 80 if (atomic_dec_and_test(&client->refcount)) 81 complete(&client->comp); 82 } 83 84 void rdma_addr_unregister_client(struct rdma_addr_client *client) 85 { 86 put_client(client); 87 wait_for_completion(&client->comp); 88 } 89 EXPORT_SYMBOL(rdma_addr_unregister_client); 90 91 int rdma_copy_addr(struct rdma_dev_addr *dev_addr, struct net_device *dev, 92 const unsigned char *dst_dev_addr) 93 { 94 dev_addr->dev_type = dev->type; 95 memcpy(dev_addr->src_dev_addr, dev->dev_addr, MAX_ADDR_LEN); 96 memcpy(dev_addr->broadcast, dev->broadcast, MAX_ADDR_LEN); 97 if (dst_dev_addr) 98 memcpy(dev_addr->dst_dev_addr, dst_dev_addr, MAX_ADDR_LEN); 99 dev_addr->bound_dev_if = dev->ifindex; 100 return 0; 101 } 102 EXPORT_SYMBOL(rdma_copy_addr); 103 104 int rdma_translate_ip(struct sockaddr *addr, struct rdma_dev_addr *dev_addr) 105 { 106 struct net_device *dev; 107 int ret = -EADDRNOTAVAIL; 108 109 if (dev_addr->bound_dev_if) { 110 dev = dev_get_by_index(&init_net, dev_addr->bound_dev_if); 111 if (!dev) 112 return -ENODEV; 113 ret = rdma_copy_addr(dev_addr, dev, NULL); 114 dev_put(dev); 115 return ret; 116 } 117 118 switch (addr->sa_family) { 119 case AF_INET: 120 dev = ip_dev_find(&init_net, 121 ((struct sockaddr_in *) addr)->sin_addr.s_addr); 122 123 if (!dev) 124 return ret; 125 126 ret = rdma_copy_addr(dev_addr, dev, NULL); 127 dev_put(dev); 128 break; 129 130 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) 131 case AF_INET6: 132 read_lock(&dev_base_lock); 133 for_each_netdev(&init_net, dev) { 134 if (ipv6_chk_addr(&init_net, 135 &((struct sockaddr_in6 *) addr)->sin6_addr, 136 dev, 1)) { 137 ret = rdma_copy_addr(dev_addr, dev, NULL); 138 break; 139 } 140 } 141 read_unlock(&dev_base_lock); 142 break; 143 #endif 144 } 145 return ret; 146 } 147 EXPORT_SYMBOL(rdma_translate_ip); 148 149 static void set_timeout(unsigned long time) 150 { 151 unsigned long delay; 152 153 cancel_delayed_work(&work); 154 155 delay = time - jiffies; 156 if ((long)delay <= 0) 157 delay = 1; 158 159 queue_delayed_work(addr_wq, &work, delay); 160 } 161 162 static void queue_req(struct addr_req *req) 163 { 164 struct addr_req *temp_req; 165 166 mutex_lock(&lock); 167 list_for_each_entry_reverse(temp_req, &req_list, list) { 168 if (time_after_eq(req->timeout, temp_req->timeout)) 169 break; 170 } 171 172 list_add(&req->list, &temp_req->list); 173 174 if (req_list.next == &req->list) 175 set_timeout(req->timeout); 176 mutex_unlock(&lock); 177 } 178 179 static int addr4_resolve(struct sockaddr_in *src_in, 180 struct sockaddr_in *dst_in, 181 struct rdma_dev_addr *addr) 182 { 183 __be32 src_ip = src_in->sin_addr.s_addr; 184 __be32 dst_ip = dst_in->sin_addr.s_addr; 185 struct flowi fl; 186 struct rtable *rt; 187 struct neighbour *neigh; 188 int ret; 189 190 memset(&fl, 0, sizeof fl); 191 fl.nl_u.ip4_u.daddr = dst_ip; 192 fl.nl_u.ip4_u.saddr = src_ip; 193 fl.oif = addr->bound_dev_if; 194 195 ret = ip_route_output_key(&init_net, &rt, &fl); 196 if (ret) 197 goto out; 198 199 src_in->sin_family = AF_INET; 200 src_in->sin_addr.s_addr = rt->rt_src; 201 202 if (rt->idev->dev->flags & IFF_LOOPBACK) { 203 ret = rdma_translate_ip((struct sockaddr *) dst_in, addr); 204 if (!ret) 205 memcpy(addr->dst_dev_addr, addr->src_dev_addr, MAX_ADDR_LEN); 206 goto put; 207 } 208 209 /* If the device does ARP internally, return 'done' */ 210 if (rt->idev->dev->flags & IFF_NOARP) { 211 rdma_copy_addr(addr, rt->idev->dev, NULL); 212 goto put; 213 } 214 215 neigh = neigh_lookup(&arp_tbl, &rt->rt_gateway, rt->idev->dev); 216 if (!neigh || !(neigh->nud_state & NUD_VALID)) { 217 neigh_event_send(rt->u.dst.neighbour, NULL); 218 ret = -ENODATA; 219 if (neigh) 220 goto release; 221 goto put; 222 } 223 224 ret = rdma_copy_addr(addr, neigh->dev, neigh->ha); 225 release: 226 neigh_release(neigh); 227 put: 228 ip_rt_put(rt); 229 out: 230 return ret; 231 } 232 233 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) 234 static int addr6_resolve(struct sockaddr_in6 *src_in, 235 struct sockaddr_in6 *dst_in, 236 struct rdma_dev_addr *addr) 237 { 238 struct flowi fl; 239 struct neighbour *neigh; 240 struct dst_entry *dst; 241 int ret; 242 243 memset(&fl, 0, sizeof fl); 244 ipv6_addr_copy(&fl.fl6_dst, &dst_in->sin6_addr); 245 ipv6_addr_copy(&fl.fl6_src, &src_in->sin6_addr); 246 fl.oif = addr->bound_dev_if; 247 248 dst = ip6_route_output(&init_net, NULL, &fl); 249 if ((ret = dst->error)) 250 goto put; 251 252 if (ipv6_addr_any(&fl.fl6_src)) { 253 ret = ipv6_dev_get_saddr(&init_net, ip6_dst_idev(dst)->dev, 254 &fl.fl6_dst, 0, &fl.fl6_src); 255 if (ret) 256 goto put; 257 258 src_in->sin6_family = AF_INET6; 259 ipv6_addr_copy(&src_in->sin6_addr, &fl.fl6_src); 260 } 261 262 if (dst->dev->flags & IFF_LOOPBACK) { 263 ret = rdma_translate_ip((struct sockaddr *) dst_in, addr); 264 if (!ret) 265 memcpy(addr->dst_dev_addr, addr->src_dev_addr, MAX_ADDR_LEN); 266 goto put; 267 } 268 269 /* If the device does ARP internally, return 'done' */ 270 if (dst->dev->flags & IFF_NOARP) { 271 ret = rdma_copy_addr(addr, dst->dev, NULL); 272 goto put; 273 } 274 275 neigh = dst->neighbour; 276 if (!neigh || !(neigh->nud_state & NUD_VALID)) { 277 neigh_event_send(dst->neighbour, NULL); 278 ret = -ENODATA; 279 goto put; 280 } 281 282 ret = rdma_copy_addr(addr, dst->dev, neigh->ha); 283 put: 284 dst_release(dst); 285 return ret; 286 } 287 #else 288 static int addr6_resolve(struct sockaddr_in6 *src_in, 289 struct sockaddr_in6 *dst_in, 290 struct rdma_dev_addr *addr) 291 { 292 return -EADDRNOTAVAIL; 293 } 294 #endif 295 296 static int addr_resolve(struct sockaddr *src_in, 297 struct sockaddr *dst_in, 298 struct rdma_dev_addr *addr) 299 { 300 if (src_in->sa_family == AF_INET) { 301 return addr4_resolve((struct sockaddr_in *) src_in, 302 (struct sockaddr_in *) dst_in, addr); 303 } else 304 return addr6_resolve((struct sockaddr_in6 *) src_in, 305 (struct sockaddr_in6 *) dst_in, addr); 306 } 307 308 static void process_req(struct work_struct *work) 309 { 310 struct addr_req *req, *temp_req; 311 struct sockaddr *src_in, *dst_in; 312 struct list_head done_list; 313 314 INIT_LIST_HEAD(&done_list); 315 316 mutex_lock(&lock); 317 list_for_each_entry_safe(req, temp_req, &req_list, list) { 318 if (req->status == -ENODATA) { 319 src_in = (struct sockaddr *) &req->src_addr; 320 dst_in = (struct sockaddr *) &req->dst_addr; 321 req->status = addr_resolve(src_in, dst_in, req->addr); 322 if (req->status && time_after_eq(jiffies, req->timeout)) 323 req->status = -ETIMEDOUT; 324 else if (req->status == -ENODATA) 325 continue; 326 } 327 list_move_tail(&req->list, &done_list); 328 } 329 330 if (!list_empty(&req_list)) { 331 req = list_entry(req_list.next, struct addr_req, list); 332 set_timeout(req->timeout); 333 } 334 mutex_unlock(&lock); 335 336 list_for_each_entry_safe(req, temp_req, &done_list, list) { 337 list_del(&req->list); 338 req->callback(req->status, (struct sockaddr *) &req->src_addr, 339 req->addr, req->context); 340 put_client(req->client); 341 kfree(req); 342 } 343 } 344 345 int rdma_resolve_ip(struct rdma_addr_client *client, 346 struct sockaddr *src_addr, struct sockaddr *dst_addr, 347 struct rdma_dev_addr *addr, int timeout_ms, 348 void (*callback)(int status, struct sockaddr *src_addr, 349 struct rdma_dev_addr *addr, void *context), 350 void *context) 351 { 352 struct sockaddr *src_in, *dst_in; 353 struct addr_req *req; 354 int ret = 0; 355 356 req = kzalloc(sizeof *req, GFP_KERNEL); 357 if (!req) 358 return -ENOMEM; 359 360 src_in = (struct sockaddr *) &req->src_addr; 361 dst_in = (struct sockaddr *) &req->dst_addr; 362 363 if (src_addr) { 364 if (src_addr->sa_family != dst_addr->sa_family) { 365 ret = -EINVAL; 366 goto err; 367 } 368 369 memcpy(src_in, src_addr, ip_addr_size(src_addr)); 370 } else { 371 src_in->sa_family = dst_addr->sa_family; 372 } 373 374 memcpy(dst_in, dst_addr, ip_addr_size(dst_addr)); 375 req->addr = addr; 376 req->callback = callback; 377 req->context = context; 378 req->client = client; 379 atomic_inc(&client->refcount); 380 381 req->status = addr_resolve(src_in, dst_in, addr); 382 switch (req->status) { 383 case 0: 384 req->timeout = jiffies; 385 queue_req(req); 386 break; 387 case -ENODATA: 388 req->timeout = msecs_to_jiffies(timeout_ms) + jiffies; 389 queue_req(req); 390 break; 391 default: 392 ret = req->status; 393 atomic_dec(&client->refcount); 394 goto err; 395 } 396 return ret; 397 err: 398 kfree(req); 399 return ret; 400 } 401 EXPORT_SYMBOL(rdma_resolve_ip); 402 403 void rdma_addr_cancel(struct rdma_dev_addr *addr) 404 { 405 struct addr_req *req, *temp_req; 406 407 mutex_lock(&lock); 408 list_for_each_entry_safe(req, temp_req, &req_list, list) { 409 if (req->addr == addr) { 410 req->status = -ECANCELED; 411 req->timeout = jiffies; 412 list_move(&req->list, &req_list); 413 set_timeout(req->timeout); 414 break; 415 } 416 } 417 mutex_unlock(&lock); 418 } 419 EXPORT_SYMBOL(rdma_addr_cancel); 420 421 static int netevent_callback(struct notifier_block *self, unsigned long event, 422 void *ctx) 423 { 424 if (event == NETEVENT_NEIGH_UPDATE) { 425 struct neighbour *neigh = ctx; 426 427 if (neigh->nud_state & NUD_VALID) { 428 set_timeout(jiffies); 429 } 430 } 431 return 0; 432 } 433 434 static struct notifier_block nb = { 435 .notifier_call = netevent_callback 436 }; 437 438 static int __init addr_init(void) 439 { 440 addr_wq = create_singlethread_workqueue("ib_addr"); 441 if (!addr_wq) 442 return -ENOMEM; 443 444 register_netevent_notifier(&nb); 445 return 0; 446 } 447 448 static void __exit addr_cleanup(void) 449 { 450 unregister_netevent_notifier(&nb); 451 destroy_workqueue(addr_wq); 452 } 453 454 module_init(addr_init); 455 module_exit(addr_cleanup); 456