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 union rdma_sockaddr {
60 struct sockaddr _sockaddr;
61 struct sockaddr_in _sockaddr_in;
62 struct sockaddr_in6 _sockaddr_in6;
63 struct sockaddr_storage _sockaddr_ss;
64 };
65
66 /**
67 * struct rdma_dev_addr - Contains resolved RDMA hardware addresses
68 * @src_dev_addr: Source MAC address.
69 * @dst_dev_addr: Destination MAC address.
70 * @broadcast: Broadcast address of the device.
71 * @dev_type: The interface hardware type of the device.
72 * @bound_dev_if: An optional device interface index.
73 * @transport: The transport type used.
74 * @net: Network namespace containing the bound_dev_if net_dev.
75 * @sgid_attr: GID attribute to use for identified SGID
76 */
77 struct vnet;
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 vnet *net;
86 const struct ib_gid_attr *sgid_attr;
87 enum rdma_network_type network;
88 int hoplimit;
89 };
90
91 /**
92 * rdma_translate_ip - Translate a local IP address to an RDMA hardware
93 * address.
94 *
95 * The dev_addr->net and dev_addr->bound_dev_if fields must be initialized.
96 */
97 int rdma_translate_ip(const struct sockaddr *addr,
98 struct rdma_dev_addr *dev_addr);
99
100 /**
101 * rdma_resolve_ip - Resolve source and destination IP addresses to
102 * RDMA hardware addresses.
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 * @resolve_by_gid_attr: Resolve the ip based on the GID attribute from
114 * rdma_dev_addr.
115 * @context: User-specified context associated with the call.
116 */
117 int rdma_resolve_ip(struct sockaddr *src_addr, const struct sockaddr *dst_addr,
118 struct rdma_dev_addr *addr, int timeout_ms,
119 void (*callback)(int status, struct sockaddr *src_addr,
120 struct rdma_dev_addr *addr, void *context),
121 bool resolve_by_gid_attr,
122 void *context);
123
124 void rdma_addr_cancel(struct rdma_dev_addr *addr);
125
126 int rdma_addr_size(const struct sockaddr *addr);
127 int rdma_addr_size_in6(struct sockaddr_in6 *addr);
128 int rdma_addr_size_kss(struct sockaddr_storage *addr);
129
ib_addr_get_pkey(struct rdma_dev_addr * dev_addr)130 static inline u16 ib_addr_get_pkey(struct rdma_dev_addr *dev_addr)
131 {
132 return ((u16)dev_addr->broadcast[8] << 8) | (u16)dev_addr->broadcast[9];
133 }
134
ib_addr_set_pkey(struct rdma_dev_addr * dev_addr,u16 pkey)135 static inline void ib_addr_set_pkey(struct rdma_dev_addr *dev_addr, u16 pkey)
136 {
137 dev_addr->broadcast[8] = pkey >> 8;
138 dev_addr->broadcast[9] = (unsigned char) pkey;
139 }
140
ib_addr_get_mgid(struct rdma_dev_addr * dev_addr,union ib_gid * gid)141 static inline void ib_addr_get_mgid(struct rdma_dev_addr *dev_addr,
142 union ib_gid *gid)
143 {
144 memcpy(gid, dev_addr->broadcast + 4, sizeof *gid);
145 }
146
rdma_addr_gid_offset(struct rdma_dev_addr * dev_addr)147 static inline int rdma_addr_gid_offset(struct rdma_dev_addr *dev_addr)
148 {
149 return dev_addr->dev_type == ARPHRD_INFINIBAND ? 4 : 0;
150 }
151
rdma_vlan_dev_vlan_id(if_t dev)152 static inline u16 rdma_vlan_dev_vlan_id(if_t dev)
153 {
154 uint16_t tag;
155
156 if (if_gettype(dev) == IFT_ETHER && if_getpcp(dev) != IFNET_PCP_NONE)
157 return 0x0000; /* prio-tagged traffic */
158 if (VLAN_TAG(__DECONST(if_t, dev), &tag) != 0)
159 return 0xffff;
160 return tag;
161 }
162
rdma_ip2gid(const struct sockaddr * addr,union ib_gid * gid)163 static inline int rdma_ip2gid(const struct sockaddr *addr, union ib_gid *gid)
164 {
165 switch (addr->sa_family) {
166 case AF_INET:
167 ipv6_addr_set_v4mapped(((const struct sockaddr_in *)
168 addr)->sin_addr.s_addr,
169 (struct in6_addr *)gid);
170 break;
171 case AF_INET6:
172 memcpy(gid->raw, &((const struct sockaddr_in6 *)addr)->sin6_addr, 16);
173 /* make sure scope ID gets zeroed inside GID */
174 if (IN6_IS_SCOPE_LINKLOCAL((struct in6_addr *)gid->raw) ||
175 IN6_IS_ADDR_MC_INTFACELOCAL((struct in6_addr *)gid->raw)) {
176 gid->raw[2] = 0;
177 gid->raw[3] = 0;
178 }
179 break;
180 default:
181 return -EINVAL;
182 }
183 return 0;
184 }
185
186 /* Important - sockaddr should be a union of sockaddr_in and sockaddr_in6 */
rdma_gid2ip(struct sockaddr * out,const union ib_gid * gid)187 static inline void rdma_gid2ip(struct sockaddr *out, const union ib_gid *gid)
188 {
189 if (ipv6_addr_v4mapped((const struct in6_addr *)gid)) {
190 struct sockaddr_in *out_in = (struct sockaddr_in *)out;
191 memset(out_in, 0, sizeof(*out_in));
192 out_in->sin_len = sizeof(*out_in);
193 out_in->sin_family = AF_INET;
194 memcpy(&out_in->sin_addr.s_addr, gid->raw + 12, 4);
195 } else {
196 struct sockaddr_in6 *out_in = (struct sockaddr_in6 *)out;
197 memset(out_in, 0, sizeof(*out_in));
198 out_in->sin6_len = sizeof(*out_in);
199 out_in->sin6_family = AF_INET6;
200 memcpy(&out_in->sin6_addr.s6_addr, gid->raw, 16);
201 }
202 }
203
204 /*
205 * rdma_get/set_sgid/dgid() APIs are applicable to IB, and iWarp.
206 * They are not applicable to RoCE.
207 * RoCE GIDs are derived from the IP addresses.
208 */
rdma_addr_get_sgid(struct rdma_dev_addr * dev_addr,union ib_gid * gid)209 static inline void rdma_addr_get_sgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid)
210 {
211 memcpy(gid, dev_addr->src_dev_addr + rdma_addr_gid_offset(dev_addr),
212 sizeof(*gid));
213 }
214
rdma_addr_set_sgid(struct rdma_dev_addr * dev_addr,union ib_gid * gid)215 static inline void rdma_addr_set_sgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid)
216 {
217 memcpy(dev_addr->src_dev_addr + rdma_addr_gid_offset(dev_addr), gid, sizeof *gid);
218 }
219
rdma_addr_get_dgid(struct rdma_dev_addr * dev_addr,union ib_gid * gid)220 static inline void rdma_addr_get_dgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid)
221 {
222 memcpy(gid, dev_addr->dst_dev_addr + rdma_addr_gid_offset(dev_addr), sizeof *gid);
223 }
224
rdma_addr_set_dgid(struct rdma_dev_addr * dev_addr,union ib_gid * gid)225 static inline void rdma_addr_set_dgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid)
226 {
227 memcpy(dev_addr->dst_dev_addr + rdma_addr_gid_offset(dev_addr), gid, sizeof *gid);
228 }
229
iboe_get_mtu(int mtu)230 static inline enum ib_mtu iboe_get_mtu(int mtu)
231 {
232 /*
233 * reduce IB headers from effective IBoE MTU. 28 stands for
234 * atomic header which is the biggest possible header after BTH
235 */
236 mtu = mtu - IB_GRH_BYTES - IB_BTH_BYTES - 28;
237
238 if (mtu >= ib_mtu_enum_to_int(IB_MTU_4096))
239 return IB_MTU_4096;
240 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_2048))
241 return IB_MTU_2048;
242 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_1024))
243 return IB_MTU_1024;
244 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_512))
245 return IB_MTU_512;
246 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_256))
247 return IB_MTU_256;
248 else
249 return 0;
250 }
251
iboe_get_rate(if_t dev)252 static inline int iboe_get_rate(if_t dev)
253 {
254 uint64_t baudrate = if_getbaudrate(dev);
255 #ifdef if_baudrate_pf
256 int exp;
257 for (exp = dev->if_baudrate_pf; exp > 0; exp--)
258 baudrate *= 10;
259 #endif
260 if (baudrate >= IF_Gbps(40))
261 return IB_RATE_40_GBPS;
262 else if (baudrate >= IF_Gbps(30))
263 return IB_RATE_30_GBPS;
264 else if (baudrate >= IF_Gbps(20))
265 return IB_RATE_20_GBPS;
266 else if (baudrate >= IF_Gbps(10))
267 return IB_RATE_10_GBPS;
268 else
269 return IB_RATE_PORT_CURRENT;
270 }
271
rdma_link_local_addr(struct in6_addr * addr)272 static inline int rdma_link_local_addr(struct in6_addr *addr)
273 {
274 if (addr->s6_addr32[0] == htonl(0xfe800000) &&
275 addr->s6_addr32[1] == 0)
276 return 1;
277
278 return 0;
279 }
280
rdma_get_ll_mac(struct in6_addr * addr,u8 * mac)281 static inline void rdma_get_ll_mac(struct in6_addr *addr, u8 *mac)
282 {
283 memcpy(mac, &addr->s6_addr[8], 3);
284 memcpy(mac + 3, &addr->s6_addr[13], 3);
285 mac[0] ^= 2;
286 }
287
rdma_is_multicast_addr(struct in6_addr * addr)288 static inline int rdma_is_multicast_addr(struct in6_addr *addr)
289 {
290 __be32 ipv4_addr;
291
292 if (addr->s6_addr[0] == 0xff)
293 return 1;
294
295 ipv4_addr = addr->s6_addr32[3];
296 return (ipv6_addr_v4mapped(addr) && ipv4_is_multicast(ipv4_addr));
297 }
298
rdma_get_mcast_mac(struct in6_addr * addr,u8 * mac)299 static inline void rdma_get_mcast_mac(struct in6_addr *addr, u8 *mac)
300 {
301 int i;
302
303 mac[0] = 0x33;
304 mac[1] = 0x33;
305 for (i = 2; i < 6; ++i)
306 mac[i] = addr->s6_addr[i + 10];
307 }
308
rdma_get_vlan_id(union ib_gid * dgid)309 static inline u16 rdma_get_vlan_id(union ib_gid *dgid)
310 {
311 u16 vid;
312
313 vid = dgid->raw[11] << 8 | dgid->raw[12];
314 return vid < 0x1000 ? vid : 0xffff;
315 }
316
rdma_vlan_dev_real_dev(if_t dev)317 static inline if_t rdma_vlan_dev_real_dev(if_t dev)
318 {
319 struct epoch_tracker et;
320
321 NET_EPOCH_ENTER(et);
322 if (if_gettype(dev) != IFT_ETHER || if_getpcp(dev) == IFNET_PCP_NONE)
323 dev = VLAN_TRUNKDEV(dev); /* non prio-tagged traffic */
324 NET_EPOCH_EXIT(et);
325 return (dev);
326 }
327
328 #endif /* IB_ADDR_H */
329