xref: /linux/include/rdma/ib_addr.h (revision e5c86679d5e864947a52fb31e45a425dea3e7fa9)
1 /*
2  * Copyright (c) 2005 Voltaire Inc.  All rights reserved.
3  * Copyright (c) 2005 Intel Corporation.  All rights reserved.
4  *
5  * This software is available to you under a choice of one of two
6  * licenses.  You may choose to be licensed under the terms of the GNU
7  * General Public License (GPL) Version 2, available from the file
8  * COPYING in the main directory of this source tree, or the
9  * OpenIB.org BSD license below:
10  *
11  *     Redistribution and use in source and binary forms, with or
12  *     without modification, are permitted provided that the following
13  *     conditions are met:
14  *
15  *      - Redistributions of source code must retain the above
16  *        copyright notice, this list of conditions and the following
17  *        disclaimer.
18  *
19  *      - Redistributions in binary form must reproduce the above
20  *        copyright notice, this list of conditions and the following
21  *        disclaimer in the documentation and/or other materials
22  *        provided with the distribution.
23  *
24  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
25  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
26  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
27  * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
28  * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
29  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
30  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
31  * SOFTWARE.
32  */
33 
34 #if !defined(IB_ADDR_H)
35 #define IB_ADDR_H
36 
37 #include <linux/in.h>
38 #include <linux/in6.h>
39 #include <linux/if_arp.h>
40 #include <linux/netdevice.h>
41 #include <linux/inetdevice.h>
42 #include <linux/socket.h>
43 #include <linux/if_vlan.h>
44 #include <net/ipv6.h>
45 #include <net/if_inet6.h>
46 #include <net/ip.h>
47 #include <rdma/ib_verbs.h>
48 #include <rdma/ib_pack.h>
49 #include <net/ipv6.h>
50 #include <net/net_namespace.h>
51 
52 struct rdma_addr_client {
53 	atomic_t refcount;
54 	struct completion comp;
55 };
56 
57 /**
58  * rdma_addr_register_client - Register an address client.
59  */
60 void rdma_addr_register_client(struct rdma_addr_client *client);
61 
62 /**
63  * rdma_addr_unregister_client - Deregister an address client.
64  * @client: Client object to deregister.
65  */
66 void rdma_addr_unregister_client(struct rdma_addr_client *client);
67 
68 /**
69  * struct rdma_dev_addr - Contains resolved RDMA hardware addresses
70  * @src_dev_addr:	Source MAC address.
71  * @dst_dev_addr:	Destination MAC address.
72  * @broadcast:		Broadcast address of the device.
73  * @dev_type:		The interface hardware type of the device.
74  * @bound_dev_if:	An optional device interface index.
75  * @transport:		The transport type used.
76  * @net:		Network namespace containing the bound_dev_if net_dev.
77  */
78 struct rdma_dev_addr {
79 	unsigned char src_dev_addr[MAX_ADDR_LEN];
80 	unsigned char dst_dev_addr[MAX_ADDR_LEN];
81 	unsigned char broadcast[MAX_ADDR_LEN];
82 	unsigned short dev_type;
83 	int bound_dev_if;
84 	enum rdma_transport_type transport;
85 	struct net *net;
86 	enum rdma_network_type network;
87 	int hoplimit;
88 };
89 
90 /**
91  * rdma_translate_ip - Translate a local IP address to an RDMA hardware
92  *   address.
93  *
94  * The dev_addr->net field must be initialized.
95  */
96 int rdma_translate_ip(const struct sockaddr *addr,
97 		      struct rdma_dev_addr *dev_addr, u16 *vlan_id);
98 
99 /**
100  * rdma_resolve_ip - Resolve source and destination IP addresses to
101  *   RDMA hardware addresses.
102  * @client: Address client associated with request.
103  * @src_addr: An optional source address to use in the resolution.  If a
104  *   source address is not provided, a usable address will be returned via
105  *   the callback.
106  * @dst_addr: The destination address to resolve.
107  * @addr: A reference to a data location that will receive the resolved
108  *   addresses.  The data location must remain valid until the callback has
109  *   been invoked. The net field of the addr struct must be valid.
110  * @timeout_ms: Amount of time to wait for the address resolution to complete.
111  * @callback: Call invoked once address resolution has completed, timed out,
112  *   or been canceled.  A status of 0 indicates success.
113  * @context: User-specified context associated with the call.
114  */
115 int rdma_resolve_ip(struct rdma_addr_client *client,
116 		    struct sockaddr *src_addr, struct sockaddr *dst_addr,
117 		    struct rdma_dev_addr *addr, int timeout_ms,
118 		    void (*callback)(int status, struct sockaddr *src_addr,
119 				     struct rdma_dev_addr *addr, void *context),
120 		    void *context);
121 
122 int rdma_resolve_ip_route(struct sockaddr *src_addr,
123 			  const struct sockaddr *dst_addr,
124 			  struct rdma_dev_addr *addr);
125 
126 void rdma_addr_cancel(struct rdma_dev_addr *addr);
127 
128 int rdma_copy_addr(struct rdma_dev_addr *dev_addr, struct net_device *dev,
129 	      const unsigned char *dst_dev_addr);
130 
131 int rdma_addr_size(struct sockaddr *addr);
132 
133 int rdma_addr_find_smac_by_sgid(union ib_gid *sgid, u8 *smac, u16 *vlan_id);
134 int rdma_addr_find_l2_eth_by_grh(const union ib_gid *sgid,
135 				 const union ib_gid *dgid,
136 				 u8 *smac, u16 *vlan_id, int *if_index,
137 				 int *hoplimit);
138 
139 static inline u16 ib_addr_get_pkey(struct rdma_dev_addr *dev_addr)
140 {
141 	return ((u16)dev_addr->broadcast[8] << 8) | (u16)dev_addr->broadcast[9];
142 }
143 
144 static inline void ib_addr_set_pkey(struct rdma_dev_addr *dev_addr, u16 pkey)
145 {
146 	dev_addr->broadcast[8] = pkey >> 8;
147 	dev_addr->broadcast[9] = (unsigned char) pkey;
148 }
149 
150 static inline void ib_addr_get_mgid(struct rdma_dev_addr *dev_addr,
151 				    union ib_gid *gid)
152 {
153 	memcpy(gid, dev_addr->broadcast + 4, sizeof *gid);
154 }
155 
156 static inline int rdma_addr_gid_offset(struct rdma_dev_addr *dev_addr)
157 {
158 	return dev_addr->dev_type == ARPHRD_INFINIBAND ? 4 : 0;
159 }
160 
161 static inline u16 rdma_vlan_dev_vlan_id(const struct net_device *dev)
162 {
163 	return is_vlan_dev(dev) ? vlan_dev_vlan_id(dev) : 0xffff;
164 }
165 
166 static inline int rdma_ip2gid(struct sockaddr *addr, union ib_gid *gid)
167 {
168 	switch (addr->sa_family) {
169 	case AF_INET:
170 		ipv6_addr_set_v4mapped(((struct sockaddr_in *)
171 					addr)->sin_addr.s_addr,
172 				       (struct in6_addr *)gid);
173 		break;
174 	case AF_INET6:
175 		memcpy(gid->raw, &((struct sockaddr_in6 *)addr)->sin6_addr, 16);
176 		break;
177 	default:
178 		return -EINVAL;
179 	}
180 	return 0;
181 }
182 
183 /* Important - sockaddr should be a union of sockaddr_in and sockaddr_in6 */
184 static inline void rdma_gid2ip(struct sockaddr *out, const union ib_gid *gid)
185 {
186 	if (ipv6_addr_v4mapped((struct in6_addr *)gid)) {
187 		struct sockaddr_in *out_in = (struct sockaddr_in *)out;
188 		memset(out_in, 0, sizeof(*out_in));
189 		out_in->sin_family = AF_INET;
190 		memcpy(&out_in->sin_addr.s_addr, gid->raw + 12, 4);
191 	} else {
192 		struct sockaddr_in6 *out_in = (struct sockaddr_in6 *)out;
193 		memset(out_in, 0, sizeof(*out_in));
194 		out_in->sin6_family = AF_INET6;
195 		memcpy(&out_in->sin6_addr.s6_addr, gid->raw, 16);
196 	}
197 }
198 
199 static inline void iboe_addr_get_sgid(struct rdma_dev_addr *dev_addr,
200 				      union ib_gid *gid)
201 {
202 	struct net_device *dev;
203 	struct in_device *ip4;
204 
205 	dev = dev_get_by_index(&init_net, dev_addr->bound_dev_if);
206 	if (dev) {
207 		ip4 = in_dev_get(dev);
208 		if (ip4 && ip4->ifa_list && ip4->ifa_list->ifa_address) {
209 			ipv6_addr_set_v4mapped(ip4->ifa_list->ifa_address,
210 					       (struct in6_addr *)gid);
211 			in_dev_put(ip4);
212 		}
213 		dev_put(dev);
214 	}
215 }
216 
217 static inline void rdma_addr_get_sgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid)
218 {
219 	if (dev_addr->transport == RDMA_TRANSPORT_IB &&
220 	    dev_addr->dev_type != ARPHRD_INFINIBAND)
221 		iboe_addr_get_sgid(dev_addr, gid);
222 	else
223 		memcpy(gid, dev_addr->src_dev_addr +
224 		       rdma_addr_gid_offset(dev_addr), sizeof *gid);
225 }
226 
227 static inline void rdma_addr_set_sgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid)
228 {
229 	memcpy(dev_addr->src_dev_addr + rdma_addr_gid_offset(dev_addr), gid, sizeof *gid);
230 }
231 
232 static inline void rdma_addr_get_dgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid)
233 {
234 	memcpy(gid, dev_addr->dst_dev_addr + rdma_addr_gid_offset(dev_addr), sizeof *gid);
235 }
236 
237 static inline void rdma_addr_set_dgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid)
238 {
239 	memcpy(dev_addr->dst_dev_addr + rdma_addr_gid_offset(dev_addr), gid, sizeof *gid);
240 }
241 
242 static inline enum ib_mtu iboe_get_mtu(int mtu)
243 {
244 	/*
245 	 * reduce IB headers from effective IBoE MTU. 28 stands for
246 	 * atomic header which is the biggest possible header after BTH
247 	 */
248 	mtu = mtu - IB_GRH_BYTES - IB_BTH_BYTES - 28;
249 
250 	if (mtu >= ib_mtu_enum_to_int(IB_MTU_4096))
251 		return IB_MTU_4096;
252 	else if (mtu >= ib_mtu_enum_to_int(IB_MTU_2048))
253 		return IB_MTU_2048;
254 	else if (mtu >= ib_mtu_enum_to_int(IB_MTU_1024))
255 		return IB_MTU_1024;
256 	else if (mtu >= ib_mtu_enum_to_int(IB_MTU_512))
257 		return IB_MTU_512;
258 	else if (mtu >= ib_mtu_enum_to_int(IB_MTU_256))
259 		return IB_MTU_256;
260 	else
261 		return 0;
262 }
263 
264 static inline int iboe_get_rate(struct net_device *dev)
265 {
266 	struct ethtool_link_ksettings cmd;
267 	int err;
268 
269 	rtnl_lock();
270 	err = __ethtool_get_link_ksettings(dev, &cmd);
271 	rtnl_unlock();
272 	if (err)
273 		return IB_RATE_PORT_CURRENT;
274 
275 	if (cmd.base.speed >= 40000)
276 		return IB_RATE_40_GBPS;
277 	else if (cmd.base.speed >= 30000)
278 		return IB_RATE_30_GBPS;
279 	else if (cmd.base.speed >= 20000)
280 		return IB_RATE_20_GBPS;
281 	else if (cmd.base.speed >= 10000)
282 		return IB_RATE_10_GBPS;
283 	else
284 		return IB_RATE_PORT_CURRENT;
285 }
286 
287 static inline int rdma_link_local_addr(struct in6_addr *addr)
288 {
289 	if (addr->s6_addr32[0] == htonl(0xfe800000) &&
290 	    addr->s6_addr32[1] == 0)
291 		return 1;
292 
293 	return 0;
294 }
295 
296 static inline void rdma_get_ll_mac(struct in6_addr *addr, u8 *mac)
297 {
298 	memcpy(mac, &addr->s6_addr[8], 3);
299 	memcpy(mac + 3, &addr->s6_addr[13], 3);
300 	mac[0] ^= 2;
301 }
302 
303 static inline int rdma_is_multicast_addr(struct in6_addr *addr)
304 {
305 	return addr->s6_addr[0] == 0xff;
306 }
307 
308 static inline void rdma_get_mcast_mac(struct in6_addr *addr, u8 *mac)
309 {
310 	int i;
311 
312 	mac[0] = 0x33;
313 	mac[1] = 0x33;
314 	for (i = 2; i < 6; ++i)
315 		mac[i] = addr->s6_addr[i + 10];
316 }
317 
318 static inline u16 rdma_get_vlan_id(union ib_gid *dgid)
319 {
320 	u16 vid;
321 
322 	vid = dgid->raw[11] << 8 | dgid->raw[12];
323 	return vid < 0x1000 ? vid : 0xffff;
324 }
325 
326 static inline struct net_device *rdma_vlan_dev_real_dev(const struct net_device *dev)
327 {
328 	return is_vlan_dev(dev) ? vlan_dev_real_dev(dev) : NULL;
329 }
330 
331 #endif /* IB_ADDR_H */
332