| 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 <linux/module.h> |
| 41 | #include <net/arp.h> |
| 42 | #include <net/neighbour.h> |
| 43 | #include <net/route.h> |
| 44 | #include <net/netevent.h> |
| 45 | #include <net/addrconf.h> |
| 46 | #include <net/ip6_route.h> |
| 47 | #include <rdma/ib_addr.h> |
| 48 | |
| 49 | MODULE_AUTHOR("Sean Hefty"); |
| 50 | MODULE_DESCRIPTION("IB Address Translation"); |
| 51 | MODULE_LICENSE("Dual BSD/GPL"); |
| 52 | |
| 53 | struct addr_req { |
| 54 | struct list_head list; |
| 55 | struct sockaddr_storage src_addr; |
| 56 | struct sockaddr_storage dst_addr; |
| 57 | struct rdma_dev_addr *addr; |
| 58 | struct rdma_addr_client *client; |
| 59 | void *context; |
| 60 | void (*callback)(int status, struct sockaddr *src_addr, |
| 61 | struct rdma_dev_addr *addr, void *context); |
| 62 | unsigned long timeout; |
| 63 | int status; |
| 64 | }; |
| 65 | |
| 66 | static void process_req(struct work_struct *work); |
| 67 | |
| 68 | static DEFINE_MUTEX(lock); |
| 69 | static LIST_HEAD(req_list); |
| 70 | static DECLARE_DELAYED_WORK(work, process_req); |
| 71 | static struct workqueue_struct *addr_wq; |
| 72 | |
| 73 | void rdma_addr_register_client(struct rdma_addr_client *client) |
| 74 | { |
| 75 | atomic_set(&client->refcount, 1); |
| 76 | init_completion(&client->comp); |
| 77 | } |
| 78 | EXPORT_SYMBOL(rdma_addr_register_client); |
| 79 | |
| 80 | static inline void put_client(struct rdma_addr_client *client) |
| 81 | { |
| 82 | if (atomic_dec_and_test(&client->refcount)) |
| 83 | complete(&client->comp); |
| 84 | } |
| 85 | |
| 86 | void rdma_addr_unregister_client(struct rdma_addr_client *client) |
| 87 | { |
| 88 | put_client(client); |
| 89 | wait_for_completion(&client->comp); |
| 90 | } |
| 91 | EXPORT_SYMBOL(rdma_addr_unregister_client); |
| 92 | |
| 93 | int rdma_copy_addr(struct rdma_dev_addr *dev_addr, struct net_device *dev, |
| 94 | const unsigned char *dst_dev_addr) |
| 95 | { |
| 96 | dev_addr->dev_type = dev->type; |
| 97 | memcpy(dev_addr->src_dev_addr, dev->dev_addr, MAX_ADDR_LEN); |
| 98 | memcpy(dev_addr->broadcast, dev->broadcast, MAX_ADDR_LEN); |
| 99 | if (dst_dev_addr) |
| 100 | memcpy(dev_addr->dst_dev_addr, dst_dev_addr, MAX_ADDR_LEN); |
| 101 | dev_addr->bound_dev_if = dev->ifindex; |
| 102 | return 0; |
| 103 | } |
| 104 | EXPORT_SYMBOL(rdma_copy_addr); |
| 105 | |
| 106 | int rdma_translate_ip(struct sockaddr *addr, struct rdma_dev_addr *dev_addr) |
| 107 | { |
| 108 | struct net_device *dev; |
| 109 | int ret = -EADDRNOTAVAIL; |
| 110 | |
| 111 | if (dev_addr->bound_dev_if) { |
| 112 | dev = dev_get_by_index(&init_net, dev_addr->bound_dev_if); |
| 113 | if (!dev) |
| 114 | return -ENODEV; |
| 115 | ret = rdma_copy_addr(dev_addr, dev, NULL); |
| 116 | dev_put(dev); |
| 117 | return ret; |
| 118 | } |
| 119 | |
| 120 | switch (addr->sa_family) { |
| 121 | case AF_INET: |
| 122 | dev = ip_dev_find(&init_net, |
| 123 | ((struct sockaddr_in *) addr)->sin_addr.s_addr); |
| 124 | |
| 125 | if (!dev) |
| 126 | return ret; |
| 127 | |
| 128 | ret = rdma_copy_addr(dev_addr, dev, NULL); |
| 129 | dev_put(dev); |
| 130 | break; |
| 131 | |
| 132 | #if IS_ENABLED(CONFIG_IPV6) |
| 133 | case AF_INET6: |
| 134 | rcu_read_lock(); |
| 135 | for_each_netdev_rcu(&init_net, dev) { |
| 136 | if (ipv6_chk_addr(&init_net, |
| 137 | &((struct sockaddr_in6 *) addr)->sin6_addr, |
| 138 | dev, 1)) { |
| 139 | ret = rdma_copy_addr(dev_addr, dev, NULL); |
| 140 | break; |
| 141 | } |
| 142 | } |
| 143 | rcu_read_unlock(); |
| 144 | break; |
| 145 | #endif |
| 146 | } |
| 147 | return ret; |
| 148 | } |
| 149 | EXPORT_SYMBOL(rdma_translate_ip); |
| 150 | |
| 151 | static void set_timeout(unsigned long time) |
| 152 | { |
| 153 | unsigned long delay; |
| 154 | |
| 155 | delay = time - jiffies; |
| 156 | if ((long)delay <= 0) |
| 157 | delay = 1; |
| 158 | |
| 159 | mod_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 dst_fetch_ha(struct dst_entry *dst, struct rdma_dev_addr *dev_addr, void *daddr) |
| 180 | { |
| 181 | struct neighbour *n; |
| 182 | int ret; |
| 183 | |
| 184 | n = dst_neigh_lookup(dst, daddr); |
| 185 | |
| 186 | rcu_read_lock(); |
| 187 | if (!n || !(n->nud_state & NUD_VALID)) { |
| 188 | if (n) |
| 189 | neigh_event_send(n, NULL); |
| 190 | ret = -ENODATA; |
| 191 | } else { |
| 192 | ret = rdma_copy_addr(dev_addr, dst->dev, n->ha); |
| 193 | } |
| 194 | rcu_read_unlock(); |
| 195 | |
| 196 | if (n) |
| 197 | neigh_release(n); |
| 198 | |
| 199 | return ret; |
| 200 | } |
| 201 | |
| 202 | static int addr4_resolve(struct sockaddr_in *src_in, |
| 203 | struct sockaddr_in *dst_in, |
| 204 | struct rdma_dev_addr *addr) |
| 205 | { |
| 206 | __be32 src_ip = src_in->sin_addr.s_addr; |
| 207 | __be32 dst_ip = dst_in->sin_addr.s_addr; |
| 208 | struct rtable *rt; |
| 209 | struct flowi4 fl4; |
| 210 | int ret; |
| 211 | |
| 212 | memset(&fl4, 0, sizeof(fl4)); |
| 213 | fl4.daddr = dst_ip; |
| 214 | fl4.saddr = src_ip; |
| 215 | fl4.flowi4_oif = addr->bound_dev_if; |
| 216 | rt = ip_route_output_key(&init_net, &fl4); |
| 217 | if (IS_ERR(rt)) { |
| 218 | ret = PTR_ERR(rt); |
| 219 | goto out; |
| 220 | } |
| 221 | src_in->sin_family = AF_INET; |
| 222 | src_in->sin_addr.s_addr = fl4.saddr; |
| 223 | |
| 224 | if (rt->dst.dev->flags & IFF_LOOPBACK) { |
| 225 | ret = rdma_translate_ip((struct sockaddr *) dst_in, addr); |
| 226 | if (!ret) |
| 227 | memcpy(addr->dst_dev_addr, addr->src_dev_addr, MAX_ADDR_LEN); |
| 228 | goto put; |
| 229 | } |
| 230 | |
| 231 | /* If the device does ARP internally, return 'done' */ |
| 232 | if (rt->dst.dev->flags & IFF_NOARP) { |
| 233 | ret = rdma_copy_addr(addr, rt->dst.dev, NULL); |
| 234 | goto put; |
| 235 | } |
| 236 | |
| 237 | ret = dst_fetch_ha(&rt->dst, addr, &fl4.daddr); |
| 238 | put: |
| 239 | ip_rt_put(rt); |
| 240 | out: |
| 241 | return ret; |
| 242 | } |
| 243 | |
| 244 | #if IS_ENABLED(CONFIG_IPV6) |
| 245 | static int addr6_resolve(struct sockaddr_in6 *src_in, |
| 246 | struct sockaddr_in6 *dst_in, |
| 247 | struct rdma_dev_addr *addr) |
| 248 | { |
| 249 | struct flowi6 fl6; |
| 250 | struct dst_entry *dst; |
| 251 | int ret; |
| 252 | |
| 253 | memset(&fl6, 0, sizeof fl6); |
| 254 | fl6.daddr = dst_in->sin6_addr; |
| 255 | fl6.saddr = src_in->sin6_addr; |
| 256 | fl6.flowi6_oif = addr->bound_dev_if; |
| 257 | |
| 258 | dst = ip6_route_output(&init_net, NULL, &fl6); |
| 259 | if ((ret = dst->error)) |
| 260 | goto put; |
| 261 | |
| 262 | if (ipv6_addr_any(&fl6.saddr)) { |
| 263 | ret = ipv6_dev_get_saddr(&init_net, ip6_dst_idev(dst)->dev, |
| 264 | &fl6.daddr, 0, &fl6.saddr); |
| 265 | if (ret) |
| 266 | goto put; |
| 267 | |
| 268 | src_in->sin6_family = AF_INET6; |
| 269 | src_in->sin6_addr = fl6.saddr; |
| 270 | } |
| 271 | |
| 272 | if (dst->dev->flags & IFF_LOOPBACK) { |
| 273 | ret = rdma_translate_ip((struct sockaddr *) dst_in, addr); |
| 274 | if (!ret) |
| 275 | memcpy(addr->dst_dev_addr, addr->src_dev_addr, MAX_ADDR_LEN); |
| 276 | goto put; |
| 277 | } |
| 278 | |
| 279 | /* If the device does ARP internally, return 'done' */ |
| 280 | if (dst->dev->flags & IFF_NOARP) { |
| 281 | ret = rdma_copy_addr(addr, dst->dev, NULL); |
| 282 | goto put; |
| 283 | } |
| 284 | |
| 285 | ret = dst_fetch_ha(dst, addr, &fl6.daddr); |
| 286 | put: |
| 287 | dst_release(dst); |
| 288 | return ret; |
| 289 | } |
| 290 | #else |
| 291 | static int addr6_resolve(struct sockaddr_in6 *src_in, |
| 292 | struct sockaddr_in6 *dst_in, |
| 293 | struct rdma_dev_addr *addr) |
| 294 | { |
| 295 | return -EADDRNOTAVAIL; |
| 296 | } |
| 297 | #endif |
| 298 | |
| 299 | static int addr_resolve(struct sockaddr *src_in, |
| 300 | struct sockaddr *dst_in, |
| 301 | struct rdma_dev_addr *addr) |
| 302 | { |
| 303 | if (src_in->sa_family == AF_INET) { |
| 304 | return addr4_resolve((struct sockaddr_in *) src_in, |
| 305 | (struct sockaddr_in *) dst_in, addr); |
| 306 | } else |
| 307 | return addr6_resolve((struct sockaddr_in6 *) src_in, |
| 308 | (struct sockaddr_in6 *) dst_in, addr); |
| 309 | } |
| 310 | |
| 311 | static void process_req(struct work_struct *work) |
| 312 | { |
| 313 | struct addr_req *req, *temp_req; |
| 314 | struct sockaddr *src_in, *dst_in; |
| 315 | struct list_head done_list; |
| 316 | |
| 317 | INIT_LIST_HEAD(&done_list); |
| 318 | |
| 319 | mutex_lock(&lock); |
| 320 | list_for_each_entry_safe(req, temp_req, &req_list, list) { |
| 321 | if (req->status == -ENODATA) { |
| 322 | src_in = (struct sockaddr *) &req->src_addr; |
| 323 | dst_in = (struct sockaddr *) &req->dst_addr; |
| 324 | req->status = addr_resolve(src_in, dst_in, req->addr); |
| 325 | if (req->status && time_after_eq(jiffies, req->timeout)) |
| 326 | req->status = -ETIMEDOUT; |
| 327 | else if (req->status == -ENODATA) |
| 328 | continue; |
| 329 | } |
| 330 | list_move_tail(&req->list, &done_list); |
| 331 | } |
| 332 | |
| 333 | if (!list_empty(&req_list)) { |
| 334 | req = list_entry(req_list.next, struct addr_req, list); |
| 335 | set_timeout(req->timeout); |
| 336 | } |
| 337 | mutex_unlock(&lock); |
| 338 | |
| 339 | list_for_each_entry_safe(req, temp_req, &done_list, list) { |
| 340 | list_del(&req->list); |
| 341 | req->callback(req->status, (struct sockaddr *) &req->src_addr, |
| 342 | req->addr, req->context); |
| 343 | put_client(req->client); |
| 344 | kfree(req); |
| 345 | } |
| 346 | } |
| 347 | |
| 348 | int rdma_resolve_ip(struct rdma_addr_client *client, |
| 349 | struct sockaddr *src_addr, struct sockaddr *dst_addr, |
| 350 | struct rdma_dev_addr *addr, int timeout_ms, |
| 351 | void (*callback)(int status, struct sockaddr *src_addr, |
| 352 | struct rdma_dev_addr *addr, void *context), |
| 353 | void *context) |
| 354 | { |
| 355 | struct sockaddr *src_in, *dst_in; |
| 356 | struct addr_req *req; |
| 357 | int ret = 0; |
| 358 | |
| 359 | req = kzalloc(sizeof *req, GFP_KERNEL); |
| 360 | if (!req) |
| 361 | return -ENOMEM; |
| 362 | |
| 363 | src_in = (struct sockaddr *) &req->src_addr; |
| 364 | dst_in = (struct sockaddr *) &req->dst_addr; |
| 365 | |
| 366 | if (src_addr) { |
| 367 | if (src_addr->sa_family != dst_addr->sa_family) { |
| 368 | ret = -EINVAL; |
| 369 | goto err; |
| 370 | } |
| 371 | |
| 372 | memcpy(src_in, src_addr, ip_addr_size(src_addr)); |
| 373 | } else { |
| 374 | src_in->sa_family = dst_addr->sa_family; |
| 375 | } |
| 376 | |
| 377 | memcpy(dst_in, dst_addr, ip_addr_size(dst_addr)); |
| 378 | req->addr = addr; |
| 379 | req->callback = callback; |
| 380 | req->context = context; |
| 381 | req->client = client; |
| 382 | atomic_inc(&client->refcount); |
| 383 | |
| 384 | req->status = addr_resolve(src_in, dst_in, addr); |
| 385 | switch (req->status) { |
| 386 | case 0: |
| 387 | req->timeout = jiffies; |
| 388 | queue_req(req); |
| 389 | break; |
| 390 | case -ENODATA: |
| 391 | req->timeout = msecs_to_jiffies(timeout_ms) + jiffies; |
| 392 | queue_req(req); |
| 393 | break; |
| 394 | default: |
| 395 | ret = req->status; |
| 396 | atomic_dec(&client->refcount); |
| 397 | goto err; |
| 398 | } |
| 399 | return ret; |
| 400 | err: |
| 401 | kfree(req); |
| 402 | return ret; |
| 403 | } |
| 404 | EXPORT_SYMBOL(rdma_resolve_ip); |
| 405 | |
| 406 | void rdma_addr_cancel(struct rdma_dev_addr *addr) |
| 407 | { |
| 408 | struct addr_req *req, *temp_req; |
| 409 | |
| 410 | mutex_lock(&lock); |
| 411 | list_for_each_entry_safe(req, temp_req, &req_list, list) { |
| 412 | if (req->addr == addr) { |
| 413 | req->status = -ECANCELED; |
| 414 | req->timeout = jiffies; |
| 415 | list_move(&req->list, &req_list); |
| 416 | set_timeout(req->timeout); |
| 417 | break; |
| 418 | } |
| 419 | } |
| 420 | mutex_unlock(&lock); |
| 421 | } |
| 422 | EXPORT_SYMBOL(rdma_addr_cancel); |
| 423 | |
| 424 | static int netevent_callback(struct notifier_block *self, unsigned long event, |
| 425 | void *ctx) |
| 426 | { |
| 427 | if (event == NETEVENT_NEIGH_UPDATE) { |
| 428 | struct neighbour *neigh = ctx; |
| 429 | |
| 430 | if (neigh->nud_state & NUD_VALID) { |
| 431 | set_timeout(jiffies); |
| 432 | } |
| 433 | } |
| 434 | return 0; |
| 435 | } |
| 436 | |
| 437 | static struct notifier_block nb = { |
| 438 | .notifier_call = netevent_callback |
| 439 | }; |
| 440 | |
| 441 | static int __init addr_init(void) |
| 442 | { |
| 443 | addr_wq = create_singlethread_workqueue("ib_addr"); |
| 444 | if (!addr_wq) |
| 445 | return -ENOMEM; |
| 446 | |
| 447 | register_netevent_notifier(&nb); |
| 448 | return 0; |
| 449 | } |
| 450 | |
| 451 | static void __exit addr_cleanup(void) |
| 452 | { |
| 453 | unregister_netevent_notifier(&nb); |
| 454 | destroy_workqueue(addr_wq); |
| 455 | } |
| 456 | |
| 457 | module_init(addr_init); |
| 458 | module_exit(addr_cleanup); |