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.
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:
13 * Redistribution and use in source and binary forms, with or
14 * without modification, are permitted provided that the following
17 * - Redistributions of source code must retain the above
18 * copyright notice, this list of conditions and the following
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.
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
36 #include <linux/mutex.h>
37 #include <linux/inetdevice.h>
38 #include <linux/workqueue.h>
39 #include <linux/if_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>
48 MODULE_AUTHOR("Sean Hefty");
49 MODULE_DESCRIPTION("IB Address Translation");
50 MODULE_LICENSE("Dual BSD/GPL");
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;
59 void (*callback)(int status, struct sockaddr *src_addr,
60 struct rdma_dev_addr *addr, void *context);
61 unsigned long timeout;
65 static void process_req(struct work_struct *work);
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;
72 void rdma_addr_register_client(struct rdma_addr_client *client)
74 atomic_set(&client->refcount, 1);
75 init_completion(&client->comp);
77 EXPORT_SYMBOL(rdma_addr_register_client);
79 static inline void put_client(struct rdma_addr_client *client)
81 if (atomic_dec_and_test(&client->refcount))
82 complete(&client->comp);
85 void rdma_addr_unregister_client(struct rdma_addr_client *client)
88 wait_for_completion(&client->comp);
90 EXPORT_SYMBOL(rdma_addr_unregister_client);
92 int rdma_copy_addr(struct rdma_dev_addr *dev_addr, struct net_device *dev,
93 const unsigned char *dst_dev_addr)
96 case ARPHRD_INFINIBAND:
97 dev_addr->dev_type = RDMA_NODE_IB_CA;
100 dev_addr->dev_type = RDMA_NODE_RNIC;
103 return -EADDRNOTAVAIL;
106 memcpy(dev_addr->src_dev_addr, dev->dev_addr, MAX_ADDR_LEN);
107 memcpy(dev_addr->broadcast, dev->broadcast, MAX_ADDR_LEN);
109 memcpy(dev_addr->dst_dev_addr, dst_dev_addr, MAX_ADDR_LEN);
110 dev_addr->bound_dev_if = dev->ifindex;
113 EXPORT_SYMBOL(rdma_copy_addr);
115 int rdma_translate_ip(struct sockaddr *addr, struct rdma_dev_addr *dev_addr)
117 struct net_device *dev;
118 int ret = -EADDRNOTAVAIL;
120 if (dev_addr->bound_dev_if) {
121 dev = dev_get_by_index(&init_net, dev_addr->bound_dev_if);
124 ret = rdma_copy_addr(dev_addr, dev, NULL);
129 switch (addr->sa_family) {
131 dev = ip_dev_find(&init_net,
132 ((struct sockaddr_in *) addr)->sin_addr.s_addr);
137 ret = rdma_copy_addr(dev_addr, dev, NULL);
141 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
143 read_lock(&dev_base_lock);
144 for_each_netdev(&init_net, dev) {
145 if (ipv6_chk_addr(&init_net,
146 &((struct sockaddr_in6 *) addr)->sin6_addr,
148 ret = rdma_copy_addr(dev_addr, dev, NULL);
152 read_unlock(&dev_base_lock);
158 EXPORT_SYMBOL(rdma_translate_ip);
160 static void set_timeout(unsigned long time)
164 cancel_delayed_work(&work);
166 delay = time - jiffies;
167 if ((long)delay <= 0)
170 queue_delayed_work(addr_wq, &work, delay);
173 static void queue_req(struct addr_req *req)
175 struct addr_req *temp_req;
178 list_for_each_entry_reverse(temp_req, &req_list, list) {
179 if (time_after_eq(req->timeout, temp_req->timeout))
183 list_add(&req->list, &temp_req->list);
185 if (req_list.next == &req->list)
186 set_timeout(req->timeout);
190 static void addr_send_arp(struct sockaddr *dst_in)
195 memset(&fl, 0, sizeof fl);
197 switch (dst_in->sa_family) {
199 fl.nl_u.ip4_u.daddr =
200 ((struct sockaddr_in *) dst_in)->sin_addr.s_addr;
202 if (ip_route_output_key(&init_net, &rt, &fl))
205 neigh_event_send(rt->u.dst.neighbour, NULL);
209 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
212 struct dst_entry *dst;
214 fl.nl_u.ip6_u.daddr =
215 ((struct sockaddr_in6 *) dst_in)->sin6_addr;
217 dst = ip6_route_output(&init_net, NULL, &fl);
221 neigh_event_send(dst->neighbour, NULL);
229 static int addr4_resolve_remote(struct sockaddr_in *src_in,
230 struct sockaddr_in *dst_in,
231 struct rdma_dev_addr *addr)
233 __be32 src_ip = src_in->sin_addr.s_addr;
234 __be32 dst_ip = dst_in->sin_addr.s_addr;
237 struct neighbour *neigh;
240 memset(&fl, 0, sizeof fl);
241 fl.nl_u.ip4_u.daddr = dst_ip;
242 fl.nl_u.ip4_u.saddr = src_ip;
243 fl.oif = addr->bound_dev_if;
245 ret = ip_route_output_key(&init_net, &rt, &fl);
249 /* If the device does ARP internally, return 'done' */
250 if (rt->idev->dev->flags & IFF_NOARP) {
251 rdma_copy_addr(addr, rt->idev->dev, NULL);
255 neigh = neigh_lookup(&arp_tbl, &rt->rt_gateway, rt->idev->dev);
261 if (!(neigh->nud_state & NUD_VALID)) {
267 src_in->sin_family = dst_in->sin_family;
268 src_in->sin_addr.s_addr = rt->rt_src;
271 ret = rdma_copy_addr(addr, neigh->dev, neigh->ha);
273 neigh_release(neigh);
280 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
281 static int addr6_resolve_remote(struct sockaddr_in6 *src_in,
282 struct sockaddr_in6 *dst_in,
283 struct rdma_dev_addr *addr)
286 struct neighbour *neigh;
287 struct dst_entry *dst;
290 memset(&fl, 0, sizeof fl);
291 fl.nl_u.ip6_u.daddr = dst_in->sin6_addr;
292 fl.nl_u.ip6_u.saddr = src_in->sin6_addr;
293 fl.oif = addr->bound_dev_if;
295 dst = ip6_route_output(&init_net, NULL, &fl);
299 if (dst->dev->flags & IFF_NOARP) {
300 ret = rdma_copy_addr(addr, dst->dev, NULL);
302 neigh = dst->neighbour;
303 if (neigh && (neigh->nud_state & NUD_VALID))
304 ret = rdma_copy_addr(addr, neigh->dev, neigh->ha);
311 static int addr6_resolve_remote(struct sockaddr_in6 *src_in,
312 struct sockaddr_in6 *dst_in,
313 struct rdma_dev_addr *addr)
315 return -EADDRNOTAVAIL;
319 static int addr_resolve_remote(struct sockaddr *src_in,
320 struct sockaddr *dst_in,
321 struct rdma_dev_addr *addr)
323 if (src_in->sa_family == AF_INET) {
324 return addr4_resolve_remote((struct sockaddr_in *) src_in,
325 (struct sockaddr_in *) dst_in, addr);
327 return addr6_resolve_remote((struct sockaddr_in6 *) src_in,
328 (struct sockaddr_in6 *) dst_in, addr);
331 static void process_req(struct work_struct *work)
333 struct addr_req *req, *temp_req;
334 struct sockaddr *src_in, *dst_in;
335 struct list_head done_list;
337 INIT_LIST_HEAD(&done_list);
340 list_for_each_entry_safe(req, temp_req, &req_list, list) {
341 if (req->status == -ENODATA) {
342 src_in = (struct sockaddr *) &req->src_addr;
343 dst_in = (struct sockaddr *) &req->dst_addr;
344 req->status = addr_resolve_remote(src_in, dst_in,
346 if (req->status && time_after_eq(jiffies, req->timeout))
347 req->status = -ETIMEDOUT;
348 else if (req->status == -ENODATA)
351 list_move_tail(&req->list, &done_list);
354 if (!list_empty(&req_list)) {
355 req = list_entry(req_list.next, struct addr_req, list);
356 set_timeout(req->timeout);
360 list_for_each_entry_safe(req, temp_req, &done_list, list) {
361 list_del(&req->list);
362 req->callback(req->status, (struct sockaddr *) &req->src_addr,
363 req->addr, req->context);
364 put_client(req->client);
369 static int addr_resolve_local(struct sockaddr *src_in,
370 struct sockaddr *dst_in,
371 struct rdma_dev_addr *addr)
373 struct net_device *dev;
376 switch (dst_in->sa_family) {
379 __be32 src_ip = ((struct sockaddr_in *) src_in)->sin_addr.s_addr;
380 __be32 dst_ip = ((struct sockaddr_in *) dst_in)->sin_addr.s_addr;
382 dev = ip_dev_find(&init_net, dst_ip);
384 return -EADDRNOTAVAIL;
386 if (ipv4_is_zeronet(src_ip)) {
387 src_in->sa_family = dst_in->sa_family;
388 ((struct sockaddr_in *) src_in)->sin_addr.s_addr = dst_ip;
389 ret = rdma_copy_addr(addr, dev, dev->dev_addr);
390 } else if (ipv4_is_loopback(src_ip)) {
391 ret = rdma_translate_ip(dst_in, addr);
393 memcpy(addr->dst_dev_addr, dev->dev_addr, MAX_ADDR_LEN);
395 ret = rdma_translate_ip(src_in, addr);
397 memcpy(addr->dst_dev_addr, dev->dev_addr, MAX_ADDR_LEN);
403 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
408 read_lock(&dev_base_lock);
409 for_each_netdev(&init_net, dev)
410 if (ipv6_chk_addr(&init_net,
411 &((struct sockaddr_in6 *) dst_in)->sin6_addr,
416 read_unlock(&dev_base_lock);
417 return -EADDRNOTAVAIL;
420 a = &((struct sockaddr_in6 *) src_in)->sin6_addr;
422 if (ipv6_addr_any(a)) {
423 src_in->sa_family = dst_in->sa_family;
424 ((struct sockaddr_in6 *) src_in)->sin6_addr =
425 ((struct sockaddr_in6 *) dst_in)->sin6_addr;
426 ret = rdma_copy_addr(addr, dev, dev->dev_addr);
427 } else if (ipv6_addr_loopback(a)) {
428 ret = rdma_translate_ip(dst_in, addr);
430 memcpy(addr->dst_dev_addr, dev->dev_addr, MAX_ADDR_LEN);
432 ret = rdma_translate_ip(src_in, addr);
434 memcpy(addr->dst_dev_addr, dev->dev_addr, MAX_ADDR_LEN);
436 read_unlock(&dev_base_lock);
442 ret = -EADDRNOTAVAIL;
449 int rdma_resolve_ip(struct rdma_addr_client *client,
450 struct sockaddr *src_addr, struct sockaddr *dst_addr,
451 struct rdma_dev_addr *addr, int timeout_ms,
452 void (*callback)(int status, struct sockaddr *src_addr,
453 struct rdma_dev_addr *addr, void *context),
456 struct sockaddr *src_in, *dst_in;
457 struct addr_req *req;
460 req = kzalloc(sizeof *req, GFP_KERNEL);
464 src_in = (struct sockaddr *) &req->src_addr;
465 dst_in = (struct sockaddr *) &req->dst_addr;
468 if (src_addr->sa_family != dst_addr->sa_family) {
473 memcpy(src_in, src_addr, ip_addr_size(src_addr));
475 src_in->sa_family = dst_addr->sa_family;
478 memcpy(dst_in, dst_addr, ip_addr_size(dst_addr));
480 req->callback = callback;
481 req->context = context;
482 req->client = client;
483 atomic_inc(&client->refcount);
485 req->status = addr_resolve_local(src_in, dst_in, addr);
486 if (req->status == -EADDRNOTAVAIL)
487 req->status = addr_resolve_remote(src_in, dst_in, addr);
489 switch (req->status) {
491 req->timeout = jiffies;
495 req->timeout = msecs_to_jiffies(timeout_ms) + jiffies;
497 addr_send_arp(dst_in);
501 atomic_dec(&client->refcount);
509 EXPORT_SYMBOL(rdma_resolve_ip);
511 void rdma_addr_cancel(struct rdma_dev_addr *addr)
513 struct addr_req *req, *temp_req;
516 list_for_each_entry_safe(req, temp_req, &req_list, list) {
517 if (req->addr == addr) {
518 req->status = -ECANCELED;
519 req->timeout = jiffies;
520 list_move(&req->list, &req_list);
521 set_timeout(req->timeout);
527 EXPORT_SYMBOL(rdma_addr_cancel);
529 static int netevent_callback(struct notifier_block *self, unsigned long event,
532 if (event == NETEVENT_NEIGH_UPDATE) {
533 struct neighbour *neigh = ctx;
535 if (neigh->nud_state & NUD_VALID) {
536 set_timeout(jiffies);
542 static struct notifier_block nb = {
543 .notifier_call = netevent_callback
546 static int __init addr_init(void)
548 addr_wq = create_singlethread_workqueue("ib_addr");
552 register_netevent_notifier(&nb);
556 static void __exit addr_cleanup(void)
558 unregister_netevent_notifier(&nb);
559 destroy_workqueue(addr_wq);
562 module_init(addr_init);
563 module_exit(addr_cleanup);