1 /*-
2 * SPDX-License-Identifier: BSD-2-Clause OR GPL-2.0
3 *
4 * Copyright (c) 2005 Voltaire 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 #if !defined(IB_ADDR_H)
37 #define IB_ADDR_H
38
39 #include <linux/in.h>
40 #include <linux/in6.h>
41 #include <linux/if_arp.h>
42 #include <linux/netdevice.h>
43 #include <linux/socket.h>
44 #include <linux/if_vlan.h>
45 #include <net/ipv6.h>
46 #include <net/if_inet6.h>
47 #include <net/ip.h>
48 #include <rdma/ib_verbs.h>
49 #include <rdma/ib_pack.h>
50 #include <rdma/ib_addr_freebsd.h>
51
52 /* Linux netdevice.h but for working on an ifnet rather than a net_device. */
53 #define dev_hold(d) if_ref(d)
54 #define dev_put(d) if_rele(d)
55 #define dev_net(d) if_getvnet(d)
56 #define net_eq(a,b) ((a) == (b))
57
58
59 struct rdma_addr_client {
60 atomic_t refcount;
61 struct completion comp;
62 };
63
64 union rdma_sockaddr {
65 struct sockaddr _sockaddr;
66 struct sockaddr_in _sockaddr_in;
67 struct sockaddr_in6 _sockaddr_in6;
68 struct sockaddr_storage _sockaddr_ss;
69 };
70
71 /**
72 * rdma_addr_register_client - Register an address client.
73 */
74 void rdma_addr_register_client(struct rdma_addr_client *client);
75
76 /**
77 * rdma_addr_unregister_client - Deregister an address client.
78 * @client: Client object to deregister.
79 */
80 void rdma_addr_unregister_client(struct rdma_addr_client *client);
81
82 /**
83 * struct rdma_dev_addr - Contains resolved RDMA hardware addresses
84 * @src_dev_addr: Source MAC address.
85 * @dst_dev_addr: Destination MAC address.
86 * @broadcast: Broadcast address of the device.
87 * @dev_type: The interface hardware type of the device.
88 * @bound_dev_if: An optional device interface index.
89 * @transport: The transport type used.
90 * @net: Network namespace containing the bound_dev_if net_dev.
91 */
92 struct vnet;
93 struct rdma_dev_addr {
94 unsigned char src_dev_addr[MAX_ADDR_LEN];
95 unsigned char dst_dev_addr[MAX_ADDR_LEN];
96 unsigned char broadcast[MAX_ADDR_LEN];
97 unsigned short dev_type;
98 int bound_dev_if;
99 enum rdma_transport_type transport;
100 struct vnet *net;
101 enum rdma_network_type network;
102 int hoplimit;
103 };
104
105 /**
106 * rdma_translate_ip - Translate a local IP address to an RDMA hardware
107 * address.
108 *
109 * The dev_addr->net and dev_addr->bound_dev_if fields must be initialized.
110 */
111 int rdma_translate_ip(const struct sockaddr *addr,
112 struct rdma_dev_addr *dev_addr);
113
114 /**
115 * rdma_resolve_ip - Resolve source and destination IP addresses to
116 * RDMA hardware addresses.
117 * @client: Address client associated with request.
118 * @src_addr: An optional source address to use in the resolution. If a
119 * source address is not provided, a usable address will be returned via
120 * the callback.
121 * @dst_addr: The destination address to resolve.
122 * @addr: A reference to a data location that will receive the resolved
123 * addresses. The data location must remain valid until the callback has
124 * been invoked. The net field of the addr struct must be valid.
125 * @timeout_ms: Amount of time to wait for the address resolution to complete.
126 * @callback: Call invoked once address resolution has completed, timed out,
127 * or been canceled. A status of 0 indicates success.
128 * @context: User-specified context associated with the call.
129 */
130 int rdma_resolve_ip(struct rdma_addr_client *client,
131 struct sockaddr *src_addr, struct sockaddr *dst_addr,
132 struct rdma_dev_addr *addr, int timeout_ms,
133 void (*callback)(int status, struct sockaddr *src_addr,
134 struct rdma_dev_addr *addr, void *context),
135 void *context);
136
137 int rdma_resolve_ip_route(struct sockaddr *src_addr,
138 const struct sockaddr *dst_addr,
139 struct rdma_dev_addr *addr);
140
141 void rdma_addr_cancel(struct rdma_dev_addr *addr);
142
143 int rdma_copy_addr(struct rdma_dev_addr *dev_addr, if_t dev,
144 const unsigned char *dst_dev_addr);
145
146 int rdma_addr_size(struct sockaddr *addr);
147 int rdma_addr_size_in6(struct sockaddr_in6 *addr);
148 int rdma_addr_size_kss(struct sockaddr_storage *addr);
149
150 int rdma_addr_find_l2_eth_by_grh(const union ib_gid *sgid,
151 const union ib_gid *dgid,
152 u8 *smac, if_t dev,
153 int *hoplimit);
154
ib_addr_get_pkey(struct rdma_dev_addr * dev_addr)155 static inline u16 ib_addr_get_pkey(struct rdma_dev_addr *dev_addr)
156 {
157 return ((u16)dev_addr->broadcast[8] << 8) | (u16)dev_addr->broadcast[9];
158 }
159
ib_addr_set_pkey(struct rdma_dev_addr * dev_addr,u16 pkey)160 static inline void ib_addr_set_pkey(struct rdma_dev_addr *dev_addr, u16 pkey)
161 {
162 dev_addr->broadcast[8] = pkey >> 8;
163 dev_addr->broadcast[9] = (unsigned char) pkey;
164 }
165
ib_addr_get_mgid(struct rdma_dev_addr * dev_addr,union ib_gid * gid)166 static inline void ib_addr_get_mgid(struct rdma_dev_addr *dev_addr,
167 union ib_gid *gid)
168 {
169 memcpy(gid, dev_addr->broadcast + 4, sizeof *gid);
170 }
171
rdma_addr_gid_offset(struct rdma_dev_addr * dev_addr)172 static inline int rdma_addr_gid_offset(struct rdma_dev_addr *dev_addr)
173 {
174 return dev_addr->dev_type == ARPHRD_INFINIBAND ? 4 : 0;
175 }
176
rdma_vlan_dev_vlan_id(if_t dev)177 static inline u16 rdma_vlan_dev_vlan_id(if_t dev)
178 {
179 uint16_t tag;
180
181 if (if_gettype(dev) == IFT_ETHER && if_getpcp(dev) != IFNET_PCP_NONE)
182 return 0x0000; /* prio-tagged traffic */
183 if (VLAN_TAG(__DECONST(if_t, dev), &tag) != 0)
184 return 0xffff;
185 return tag;
186 }
187
rdma_ip2gid(const struct sockaddr * addr,union ib_gid * gid)188 static inline int rdma_ip2gid(const struct sockaddr *addr, union ib_gid *gid)
189 {
190 switch (addr->sa_family) {
191 case AF_INET:
192 ipv6_addr_set_v4mapped(((const struct sockaddr_in *)
193 addr)->sin_addr.s_addr,
194 (struct in6_addr *)gid);
195 break;
196 case AF_INET6:
197 memcpy(gid->raw, &((const struct sockaddr_in6 *)addr)->sin6_addr, 16);
198 /* make sure scope ID gets zeroed inside GID */
199 if (IN6_IS_SCOPE_LINKLOCAL((struct in6_addr *)gid->raw) ||
200 IN6_IS_ADDR_MC_INTFACELOCAL((struct in6_addr *)gid->raw)) {
201 gid->raw[2] = 0;
202 gid->raw[3] = 0;
203 }
204 break;
205 default:
206 return -EINVAL;
207 }
208 return 0;
209 }
210
211 /* Important - sockaddr should be a union of sockaddr_in and sockaddr_in6 */
rdma_gid2ip(struct sockaddr * out,const union ib_gid * gid)212 static inline void rdma_gid2ip(struct sockaddr *out, const union ib_gid *gid)
213 {
214 if (ipv6_addr_v4mapped((const struct in6_addr *)gid)) {
215 struct sockaddr_in *out_in = (struct sockaddr_in *)out;
216 memset(out_in, 0, sizeof(*out_in));
217 out_in->sin_len = sizeof(*out_in);
218 out_in->sin_family = AF_INET;
219 memcpy(&out_in->sin_addr.s_addr, gid->raw + 12, 4);
220 } else {
221 struct sockaddr_in6 *out_in = (struct sockaddr_in6 *)out;
222 memset(out_in, 0, sizeof(*out_in));
223 out_in->sin6_len = sizeof(*out_in);
224 out_in->sin6_family = AF_INET6;
225 memcpy(&out_in->sin6_addr.s6_addr, gid->raw, 16);
226 }
227 }
228
229 static u_int
_iboe_addr_get_sgid_ia_cb(void * arg,struct ifaddr * ifa,u_int count __unused)230 _iboe_addr_get_sgid_ia_cb(void *arg, struct ifaddr *ifa, u_int count __unused)
231 {
232 ipv6_addr_set_v4mapped(((struct sockaddr_in *)
233 ifa->ifa_addr)->sin_addr.s_addr,
234 (struct in6_addr *)arg);
235 return (0);
236 }
237
iboe_addr_get_sgid(struct rdma_dev_addr * dev_addr,union ib_gid * gid)238 static inline void iboe_addr_get_sgid(struct rdma_dev_addr *dev_addr,
239 union ib_gid *gid)
240 {
241 if_t dev;
242
243 #ifdef VIMAGE
244 if (dev_addr->net == NULL)
245 return;
246 #endif
247 dev = dev_get_by_index(dev_addr->net, dev_addr->bound_dev_if);
248 if (dev) {
249 if_foreach_addr_type(dev, AF_INET,
250 _iboe_addr_get_sgid_ia_cb, gid);
251 dev_put(dev);
252 }
253 }
254
rdma_addr_get_sgid(struct rdma_dev_addr * dev_addr,union ib_gid * gid)255 static inline void rdma_addr_get_sgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid)
256 {
257 if (dev_addr->transport == RDMA_TRANSPORT_IB &&
258 dev_addr->dev_type != ARPHRD_INFINIBAND)
259 iboe_addr_get_sgid(dev_addr, gid);
260 else
261 memcpy(gid, dev_addr->src_dev_addr +
262 rdma_addr_gid_offset(dev_addr), sizeof *gid);
263 }
264
rdma_addr_set_sgid(struct rdma_dev_addr * dev_addr,union ib_gid * gid)265 static inline void rdma_addr_set_sgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid)
266 {
267 memcpy(dev_addr->src_dev_addr + rdma_addr_gid_offset(dev_addr), gid, sizeof *gid);
268 }
269
rdma_addr_get_dgid(struct rdma_dev_addr * dev_addr,union ib_gid * gid)270 static inline void rdma_addr_get_dgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid)
271 {
272 memcpy(gid, dev_addr->dst_dev_addr + rdma_addr_gid_offset(dev_addr), sizeof *gid);
273 }
274
rdma_addr_set_dgid(struct rdma_dev_addr * dev_addr,union ib_gid * gid)275 static inline void rdma_addr_set_dgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid)
276 {
277 memcpy(dev_addr->dst_dev_addr + rdma_addr_gid_offset(dev_addr), gid, sizeof *gid);
278 }
279
iboe_get_mtu(int mtu)280 static inline enum ib_mtu iboe_get_mtu(int mtu)
281 {
282 /*
283 * reduce IB headers from effective IBoE MTU. 28 stands for
284 * atomic header which is the biggest possible header after BTH
285 */
286 mtu = mtu - IB_GRH_BYTES - IB_BTH_BYTES - 28;
287
288 if (mtu >= ib_mtu_enum_to_int(IB_MTU_4096))
289 return IB_MTU_4096;
290 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_2048))
291 return IB_MTU_2048;
292 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_1024))
293 return IB_MTU_1024;
294 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_512))
295 return IB_MTU_512;
296 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_256))
297 return IB_MTU_256;
298 else
299 return 0;
300 }
301
iboe_get_rate(if_t dev)302 static inline int iboe_get_rate(if_t dev)
303 {
304 uint64_t baudrate = if_getbaudrate(dev);
305 #ifdef if_baudrate_pf
306 int exp;
307 for (exp = dev->if_baudrate_pf; exp > 0; exp--)
308 baudrate *= 10;
309 #endif
310 if (baudrate >= IF_Gbps(40))
311 return IB_RATE_40_GBPS;
312 else if (baudrate >= IF_Gbps(30))
313 return IB_RATE_30_GBPS;
314 else if (baudrate >= IF_Gbps(20))
315 return IB_RATE_20_GBPS;
316 else if (baudrate >= IF_Gbps(10))
317 return IB_RATE_10_GBPS;
318 else
319 return IB_RATE_PORT_CURRENT;
320 }
321
rdma_link_local_addr(struct in6_addr * addr)322 static inline int rdma_link_local_addr(struct in6_addr *addr)
323 {
324 if (addr->s6_addr32[0] == htonl(0xfe800000) &&
325 addr->s6_addr32[1] == 0)
326 return 1;
327
328 return 0;
329 }
330
rdma_get_ll_mac(struct in6_addr * addr,u8 * mac)331 static inline void rdma_get_ll_mac(struct in6_addr *addr, u8 *mac)
332 {
333 memcpy(mac, &addr->s6_addr[8], 3);
334 memcpy(mac + 3, &addr->s6_addr[13], 3);
335 mac[0] ^= 2;
336 }
337
rdma_is_multicast_addr(struct in6_addr * addr)338 static inline int rdma_is_multicast_addr(struct in6_addr *addr)
339 {
340 __be32 ipv4_addr;
341
342 if (addr->s6_addr[0] == 0xff)
343 return 1;
344
345 ipv4_addr = addr->s6_addr32[3];
346 return (ipv6_addr_v4mapped(addr) && ipv4_is_multicast(ipv4_addr));
347 }
348
rdma_get_mcast_mac(struct in6_addr * addr,u8 * mac)349 static inline void rdma_get_mcast_mac(struct in6_addr *addr, u8 *mac)
350 {
351 int i;
352
353 mac[0] = 0x33;
354 mac[1] = 0x33;
355 for (i = 2; i < 6; ++i)
356 mac[i] = addr->s6_addr[i + 10];
357 }
358
rdma_get_vlan_id(union ib_gid * dgid)359 static inline u16 rdma_get_vlan_id(union ib_gid *dgid)
360 {
361 u16 vid;
362
363 vid = dgid->raw[11] << 8 | dgid->raw[12];
364 return vid < 0x1000 ? vid : 0xffff;
365 }
366
rdma_vlan_dev_real_dev(if_t dev)367 static inline if_t rdma_vlan_dev_real_dev(if_t dev)
368 {
369 struct epoch_tracker et;
370
371 NET_EPOCH_ENTER(et);
372 if (if_gettype(dev) != IFT_ETHER || if_getpcp(dev) == IFNET_PCP_NONE)
373 dev = VLAN_TRUNKDEV(dev); /* non prio-tagged traffic */
374 NET_EPOCH_EXIT(et);
375 return (dev);
376 }
377
378 #endif /* IB_ADDR_H */
379