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/slab.h>
39 #include <linux/workqueue.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)
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);
99 memcpy(dev_addr->dst_dev_addr, dst_dev_addr, MAX_ADDR_LEN);
100 dev_addr->bound_dev_if = dev->ifindex;
103 EXPORT_SYMBOL(rdma_copy_addr);
105 int rdma_translate_ip(struct sockaddr *addr, struct rdma_dev_addr *dev_addr)
107 struct net_device *dev;
108 int ret = -EADDRNOTAVAIL;
110 if (dev_addr->bound_dev_if) {
111 dev = dev_get_by_index(&init_net, dev_addr->bound_dev_if);
114 ret = rdma_copy_addr(dev_addr, dev, NULL);
119 switch (addr->sa_family) {
121 dev = ip_dev_find(&init_net,
122 ((struct sockaddr_in *) addr)->sin_addr.s_addr);
127 ret = rdma_copy_addr(dev_addr, dev, NULL);
131 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
134 for_each_netdev_rcu(&init_net, dev) {
135 if (ipv6_chk_addr(&init_net,
136 &((struct sockaddr_in6 *) addr)->sin6_addr,
138 ret = rdma_copy_addr(dev_addr, dev, NULL);
148 EXPORT_SYMBOL(rdma_translate_ip);
150 static void set_timeout(unsigned long time)
154 cancel_delayed_work(&work);
156 delay = time - jiffies;
157 if ((long)delay <= 0)
160 queue_delayed_work(addr_wq, &work, delay);
163 static void queue_req(struct addr_req *req)
165 struct addr_req *temp_req;
168 list_for_each_entry_reverse(temp_req, &req_list, list) {
169 if (time_after_eq(req->timeout, temp_req->timeout))
173 list_add(&req->list, &temp_req->list);
175 if (req_list.next == &req->list)
176 set_timeout(req->timeout);
180 static int addr4_resolve(struct sockaddr_in *src_in,
181 struct sockaddr_in *dst_in,
182 struct rdma_dev_addr *addr)
184 __be32 src_ip = src_in->sin_addr.s_addr;
185 __be32 dst_ip = dst_in->sin_addr.s_addr;
187 struct neighbour *neigh;
191 memset(&fl4, 0, sizeof(fl4));
194 fl4.flowi4_oif = addr->bound_dev_if;
195 rt = ip_route_output_key(&init_net, &fl4);
200 src_in->sin_family = AF_INET;
201 src_in->sin_addr.s_addr = fl4.saddr;
203 if (rt->dst.dev->flags & IFF_LOOPBACK) {
204 ret = rdma_translate_ip((struct sockaddr *) dst_in, addr);
206 memcpy(addr->dst_dev_addr, addr->src_dev_addr, MAX_ADDR_LEN);
210 /* If the device does ARP internally, return 'done' */
211 if (rt->dst.dev->flags & IFF_NOARP) {
212 ret = rdma_copy_addr(addr, rt->dst.dev, NULL);
216 neigh = neigh_lookup(&arp_tbl, &rt->rt_gateway, rt->dst.dev);
217 if (!neigh || !(neigh->nud_state & NUD_VALID)) {
218 neigh_event_send(dst_get_neighbour(&rt->dst), NULL);
225 ret = rdma_copy_addr(addr, neigh->dev, neigh->ha);
227 neigh_release(neigh);
234 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
235 static int addr6_resolve(struct sockaddr_in6 *src_in,
236 struct sockaddr_in6 *dst_in,
237 struct rdma_dev_addr *addr)
240 struct neighbour *neigh;
241 struct dst_entry *dst;
244 memset(&fl6, 0, sizeof fl6);
245 ipv6_addr_copy(&fl6.daddr, &dst_in->sin6_addr);
246 ipv6_addr_copy(&fl6.saddr, &src_in->sin6_addr);
247 fl6.flowi6_oif = addr->bound_dev_if;
249 dst = ip6_route_output(&init_net, NULL, &fl6);
250 if ((ret = dst->error))
253 if (ipv6_addr_any(&fl6.saddr)) {
254 ret = ipv6_dev_get_saddr(&init_net, ip6_dst_idev(dst)->dev,
255 &fl6.daddr, 0, &fl6.saddr);
259 src_in->sin6_family = AF_INET6;
260 ipv6_addr_copy(&src_in->sin6_addr, &fl6.saddr);
263 if (dst->dev->flags & IFF_LOOPBACK) {
264 ret = rdma_translate_ip((struct sockaddr *) dst_in, addr);
266 memcpy(addr->dst_dev_addr, addr->src_dev_addr, MAX_ADDR_LEN);
270 /* If the device does ARP internally, return 'done' */
271 if (dst->dev->flags & IFF_NOARP) {
272 ret = rdma_copy_addr(addr, dst->dev, NULL);
276 neigh = dst_get_neighbour(dst);
277 if (!neigh || !(neigh->nud_state & NUD_VALID)) {
279 neigh_event_send(neigh, NULL);
284 ret = rdma_copy_addr(addr, dst->dev, neigh->ha);
290 static int addr6_resolve(struct sockaddr_in6 *src_in,
291 struct sockaddr_in6 *dst_in,
292 struct rdma_dev_addr *addr)
294 return -EADDRNOTAVAIL;
298 static int addr_resolve(struct sockaddr *src_in,
299 struct sockaddr *dst_in,
300 struct rdma_dev_addr *addr)
302 if (src_in->sa_family == AF_INET) {
303 return addr4_resolve((struct sockaddr_in *) src_in,
304 (struct sockaddr_in *) dst_in, addr);
306 return addr6_resolve((struct sockaddr_in6 *) src_in,
307 (struct sockaddr_in6 *) dst_in, addr);
310 static void process_req(struct work_struct *work)
312 struct addr_req *req, *temp_req;
313 struct sockaddr *src_in, *dst_in;
314 struct list_head done_list;
316 INIT_LIST_HEAD(&done_list);
319 list_for_each_entry_safe(req, temp_req, &req_list, list) {
320 if (req->status == -ENODATA) {
321 src_in = (struct sockaddr *) &req->src_addr;
322 dst_in = (struct sockaddr *) &req->dst_addr;
323 req->status = addr_resolve(src_in, dst_in, req->addr);
324 if (req->status && time_after_eq(jiffies, req->timeout))
325 req->status = -ETIMEDOUT;
326 else if (req->status == -ENODATA)
329 list_move_tail(&req->list, &done_list);
332 if (!list_empty(&req_list)) {
333 req = list_entry(req_list.next, struct addr_req, list);
334 set_timeout(req->timeout);
338 list_for_each_entry_safe(req, temp_req, &done_list, list) {
339 list_del(&req->list);
340 req->callback(req->status, (struct sockaddr *) &req->src_addr,
341 req->addr, req->context);
342 put_client(req->client);
347 int rdma_resolve_ip(struct rdma_addr_client *client,
348 struct sockaddr *src_addr, struct sockaddr *dst_addr,
349 struct rdma_dev_addr *addr, int timeout_ms,
350 void (*callback)(int status, struct sockaddr *src_addr,
351 struct rdma_dev_addr *addr, void *context),
354 struct sockaddr *src_in, *dst_in;
355 struct addr_req *req;
358 req = kzalloc(sizeof *req, GFP_KERNEL);
362 src_in = (struct sockaddr *) &req->src_addr;
363 dst_in = (struct sockaddr *) &req->dst_addr;
366 if (src_addr->sa_family != dst_addr->sa_family) {
371 memcpy(src_in, src_addr, ip_addr_size(src_addr));
373 src_in->sa_family = dst_addr->sa_family;
376 memcpy(dst_in, dst_addr, ip_addr_size(dst_addr));
378 req->callback = callback;
379 req->context = context;
380 req->client = client;
381 atomic_inc(&client->refcount);
383 req->status = addr_resolve(src_in, dst_in, addr);
384 switch (req->status) {
386 req->timeout = jiffies;
390 req->timeout = msecs_to_jiffies(timeout_ms) + jiffies;
395 atomic_dec(&client->refcount);
403 EXPORT_SYMBOL(rdma_resolve_ip);
405 void rdma_addr_cancel(struct rdma_dev_addr *addr)
407 struct addr_req *req, *temp_req;
410 list_for_each_entry_safe(req, temp_req, &req_list, list) {
411 if (req->addr == addr) {
412 req->status = -ECANCELED;
413 req->timeout = jiffies;
414 list_move(&req->list, &req_list);
415 set_timeout(req->timeout);
421 EXPORT_SYMBOL(rdma_addr_cancel);
423 static int netevent_callback(struct notifier_block *self, unsigned long event,
426 if (event == NETEVENT_NEIGH_UPDATE) {
427 struct neighbour *neigh = ctx;
429 if (neigh->nud_state & NUD_VALID) {
430 set_timeout(jiffies);
436 static struct notifier_block nb = {
437 .notifier_call = netevent_callback
440 static int __init addr_init(void)
442 addr_wq = create_singlethread_workqueue("ib_addr");
446 register_netevent_notifier(&nb);
450 static void __exit addr_cleanup(void)
452 unregister_netevent_notifier(&nb);
453 destroy_workqueue(addr_wq);
456 module_init(addr_init);
457 module_exit(addr_cleanup);