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 };
87
88 /**
89 * rdma_translate_ip - Translate a local IP address to an RDMA hardware
90 * address.
91 *
92 * The dev_addr->net field must be initialized.
93 */
94 int rdma_translate_ip(struct sockaddr *addr, struct rdma_dev_addr *dev_addr,
95 u16 *vlan_id);
96
97 /**
98 * rdma_resolve_ip - Resolve source and destination IP addresses to
99 * RDMA hardware addresses.
100 * @client: Address client associated with request.
101 * @src_addr: An optional source address to use in the resolution. If a
102 * source address is not provided, a usable address will be returned via
103 * the callback.
104 * @dst_addr: The destination address to resolve.
105 * @addr: A reference to a data location that will receive the resolved
106 * addresses. The data location must remain valid until the callback has
107 * been invoked. The net field of the addr struct must be valid.
108 * @timeout_ms: Amount of time to wait for the address resolution to complete.
109 * @callback: Call invoked once address resolution has completed, timed out,
110 * or been canceled. A status of 0 indicates success.
111 * @context: User-specified context associated with the call.
112 */
113 int rdma_resolve_ip(struct rdma_addr_client *client,
114 struct sockaddr *src_addr, struct sockaddr *dst_addr,
115 struct rdma_dev_addr *addr, int timeout_ms,
116 void (*callback)(int status, struct sockaddr *src_addr,
117 struct rdma_dev_addr *addr, void *context),
118 void *context);
119
120 void rdma_addr_cancel(struct rdma_dev_addr *addr);
121
122 int rdma_copy_addr(struct rdma_dev_addr *dev_addr, struct net_device *dev,
123 const unsigned char *dst_dev_addr);
124
125 int rdma_addr_size(struct sockaddr *addr);
126 int rdma_addr_size_in6(struct sockaddr_in6 *addr);
127 int rdma_addr_size_kss(struct __kernel_sockaddr_storage *addr);
128
129 int rdma_addr_find_smac_by_sgid(union ib_gid *sgid, u8 *smac, u16 *vlan_id);
130 int rdma_addr_find_dmac_by_grh(const union ib_gid *sgid, const union ib_gid *dgid,
131 u8 *smac, u16 *vlan_id, int if_index);
132
ib_addr_get_pkey(struct rdma_dev_addr * dev_addr)133 static inline u16 ib_addr_get_pkey(struct rdma_dev_addr *dev_addr)
134 {
135 return ((u16)dev_addr->broadcast[8] << 8) | (u16)dev_addr->broadcast[9];
136 }
137
ib_addr_set_pkey(struct rdma_dev_addr * dev_addr,u16 pkey)138 static inline void ib_addr_set_pkey(struct rdma_dev_addr *dev_addr, u16 pkey)
139 {
140 dev_addr->broadcast[8] = pkey >> 8;
141 dev_addr->broadcast[9] = (unsigned char) pkey;
142 }
143
ib_addr_get_mgid(struct rdma_dev_addr * dev_addr,union ib_gid * gid)144 static inline void ib_addr_get_mgid(struct rdma_dev_addr *dev_addr,
145 union ib_gid *gid)
146 {
147 memcpy(gid, dev_addr->broadcast + 4, sizeof *gid);
148 }
149
rdma_addr_gid_offset(struct rdma_dev_addr * dev_addr)150 static inline int rdma_addr_gid_offset(struct rdma_dev_addr *dev_addr)
151 {
152 return dev_addr->dev_type == ARPHRD_INFINIBAND ? 4 : 0;
153 }
154
rdma_vlan_dev_vlan_id(const struct net_device * dev)155 static inline u16 rdma_vlan_dev_vlan_id(const struct net_device *dev)
156 {
157 return dev->priv_flags & IFF_802_1Q_VLAN ?
158 vlan_dev_vlan_id(dev) : 0xffff;
159 }
160
rdma_ip2gid(struct sockaddr * addr,union ib_gid * gid)161 static inline int rdma_ip2gid(struct sockaddr *addr, union ib_gid *gid)
162 {
163 switch (addr->sa_family) {
164 case AF_INET:
165 ipv6_addr_set_v4mapped(((struct sockaddr_in *)
166 addr)->sin_addr.s_addr,
167 (struct in6_addr *)gid);
168 break;
169 case AF_INET6:
170 memcpy(gid->raw, &((struct sockaddr_in6 *)addr)->sin6_addr, 16);
171 break;
172 default:
173 return -EINVAL;
174 }
175 return 0;
176 }
177
178 /* Important - sockaddr should be a union of sockaddr_in and sockaddr_in6 */
rdma_gid2ip(struct sockaddr * out,const union ib_gid * gid)179 static inline void rdma_gid2ip(struct sockaddr *out, const union ib_gid *gid)
180 {
181 if (ipv6_addr_v4mapped((struct in6_addr *)gid)) {
182 struct sockaddr_in *out_in = (struct sockaddr_in *)out;
183 memset(out_in, 0, sizeof(*out_in));
184 out_in->sin_family = AF_INET;
185 memcpy(&out_in->sin_addr.s_addr, gid->raw + 12, 4);
186 } else {
187 struct sockaddr_in6 *out_in = (struct sockaddr_in6 *)out;
188 memset(out_in, 0, sizeof(*out_in));
189 out_in->sin6_family = AF_INET6;
190 memcpy(&out_in->sin6_addr.s6_addr, gid->raw, 16);
191 }
192 }
193
iboe_addr_get_sgid(struct rdma_dev_addr * dev_addr,union ib_gid * gid)194 static inline void iboe_addr_get_sgid(struct rdma_dev_addr *dev_addr,
195 union ib_gid *gid)
196 {
197 struct net_device *dev;
198 struct in_device *ip4;
199
200 dev = dev_get_by_index(&init_net, dev_addr->bound_dev_if);
201 if (dev) {
202 ip4 = in_dev_get(dev);
203 if (ip4 && ip4->ifa_list && ip4->ifa_list->ifa_address)
204 ipv6_addr_set_v4mapped(ip4->ifa_list->ifa_address,
205 (struct in6_addr *)gid);
206
207 if (ip4)
208 in_dev_put(ip4);
209
210 dev_put(dev);
211 }
212 }
213
rdma_addr_get_sgid(struct rdma_dev_addr * dev_addr,union ib_gid * gid)214 static inline void rdma_addr_get_sgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid)
215 {
216 if (dev_addr->transport == RDMA_TRANSPORT_IB &&
217 dev_addr->dev_type != ARPHRD_INFINIBAND)
218 iboe_addr_get_sgid(dev_addr, gid);
219 else
220 memcpy(gid, dev_addr->src_dev_addr +
221 rdma_addr_gid_offset(dev_addr), sizeof *gid);
222 }
223
rdma_addr_set_sgid(struct rdma_dev_addr * dev_addr,union ib_gid * gid)224 static inline void rdma_addr_set_sgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid)
225 {
226 memcpy(dev_addr->src_dev_addr + rdma_addr_gid_offset(dev_addr), gid, sizeof *gid);
227 }
228
rdma_addr_get_dgid(struct rdma_dev_addr * dev_addr,union ib_gid * gid)229 static inline void rdma_addr_get_dgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid)
230 {
231 memcpy(gid, dev_addr->dst_dev_addr + rdma_addr_gid_offset(dev_addr), sizeof *gid);
232 }
233
rdma_addr_set_dgid(struct rdma_dev_addr * dev_addr,union ib_gid * gid)234 static inline void rdma_addr_set_dgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid)
235 {
236 memcpy(dev_addr->dst_dev_addr + rdma_addr_gid_offset(dev_addr), gid, sizeof *gid);
237 }
238
iboe_get_mtu(int mtu)239 static inline enum ib_mtu iboe_get_mtu(int mtu)
240 {
241 /*
242 * reduce IB headers from effective IBoE MTU. 28 stands for
243 * atomic header which is the biggest possible header after BTH
244 */
245 mtu = mtu - IB_GRH_BYTES - IB_BTH_BYTES - 28;
246
247 if (mtu >= ib_mtu_enum_to_int(IB_MTU_4096))
248 return IB_MTU_4096;
249 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_2048))
250 return IB_MTU_2048;
251 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_1024))
252 return IB_MTU_1024;
253 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_512))
254 return IB_MTU_512;
255 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_256))
256 return IB_MTU_256;
257 else
258 return 0;
259 }
260
iboe_get_rate(struct net_device * dev)261 static inline int iboe_get_rate(struct net_device *dev)
262 {
263 struct ethtool_cmd cmd;
264 u32 speed;
265 int err;
266
267 rtnl_lock();
268 err = __ethtool_get_settings(dev, &cmd);
269 rtnl_unlock();
270 if (err)
271 return IB_RATE_PORT_CURRENT;
272
273 speed = ethtool_cmd_speed(&cmd);
274 if (speed >= 40000)
275 return IB_RATE_40_GBPS;
276 else if (speed >= 30000)
277 return IB_RATE_30_GBPS;
278 else if (speed >= 20000)
279 return IB_RATE_20_GBPS;
280 else if (speed >= 10000)
281 return IB_RATE_10_GBPS;
282 else
283 return IB_RATE_PORT_CURRENT;
284 }
285
rdma_link_local_addr(struct in6_addr * addr)286 static inline int rdma_link_local_addr(struct in6_addr *addr)
287 {
288 if (addr->s6_addr32[0] == htonl(0xfe800000) &&
289 addr->s6_addr32[1] == 0)
290 return 1;
291
292 return 0;
293 }
294
rdma_get_ll_mac(struct in6_addr * addr,u8 * mac)295 static inline void rdma_get_ll_mac(struct in6_addr *addr, u8 *mac)
296 {
297 memcpy(mac, &addr->s6_addr[8], 3);
298 memcpy(mac + 3, &addr->s6_addr[13], 3);
299 mac[0] ^= 2;
300 }
301
rdma_is_multicast_addr(struct in6_addr * addr)302 static inline int rdma_is_multicast_addr(struct in6_addr *addr)
303 {
304 return addr->s6_addr[0] == 0xff;
305 }
306
rdma_get_mcast_mac(struct in6_addr * addr,u8 * mac)307 static inline void rdma_get_mcast_mac(struct in6_addr *addr, u8 *mac)
308 {
309 int i;
310
311 mac[0] = 0x33;
312 mac[1] = 0x33;
313 for (i = 2; i < 6; ++i)
314 mac[i] = addr->s6_addr[i + 10];
315 }
316
rdma_get_vlan_id(union ib_gid * dgid)317 static inline u16 rdma_get_vlan_id(union ib_gid *dgid)
318 {
319 u16 vid;
320
321 vid = dgid->raw[11] << 8 | dgid->raw[12];
322 return vid < 0x1000 ? vid : 0xffff;
323 }
324
rdma_vlan_dev_real_dev(const struct net_device * dev)325 static inline struct net_device *rdma_vlan_dev_real_dev(const struct net_device *dev)
326 {
327 return dev->priv_flags & IFF_802_1Q_VLAN ?
328 vlan_dev_real_dev(dev) : NULL;
329 }
330
331 #endif /* IB_ADDR_H */
332