Loading...
1/* SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB */
2/*
3 * Copyright (c) 2005 Voltaire Inc. All rights reserved.
4 * Copyright (c) 2005 Intel Corporation. All rights reserved.
5 */
6
7#ifndef IB_ADDR_H
8#define IB_ADDR_H
9
10#include <linux/ethtool.h>
11#include <linux/in.h>
12#include <linux/in6.h>
13#include <linux/if_arp.h>
14#include <linux/netdevice.h>
15#include <linux/inetdevice.h>
16#include <linux/socket.h>
17#include <linux/if_vlan.h>
18#include <net/ipv6.h>
19#include <net/if_inet6.h>
20#include <net/ip.h>
21#include <rdma/ib_verbs.h>
22#include <rdma/ib_pack.h>
23#include <net/net_namespace.h>
24
25/**
26 * struct rdma_dev_addr - Contains resolved RDMA hardware addresses
27 * @src_dev_addr: Source MAC address.
28 * @dst_dev_addr: Destination MAC address.
29 * @broadcast: Broadcast address of the device.
30 * @dev_type: The interface hardware type of the device.
31 * @bound_dev_if: An optional device interface index.
32 * @transport: The transport type used.
33 * @net: Network namespace containing the bound_dev_if net_dev.
34 * @sgid_attr: GID attribute to use for identified SGID
35 */
36struct rdma_dev_addr {
37 unsigned char src_dev_addr[MAX_ADDR_LEN];
38 unsigned char dst_dev_addr[MAX_ADDR_LEN];
39 unsigned char broadcast[MAX_ADDR_LEN];
40 unsigned short dev_type;
41 int bound_dev_if;
42 enum rdma_transport_type transport;
43 struct net *net;
44 const struct ib_gid_attr *sgid_attr;
45 enum rdma_network_type network;
46 int hoplimit;
47};
48
49/**
50 * rdma_translate_ip - Translate a local IP address to an RDMA hardware
51 * address.
52 *
53 * The dev_addr->net field must be initialized.
54 */
55int rdma_translate_ip(const struct sockaddr *addr,
56 struct rdma_dev_addr *dev_addr);
57
58/**
59 * rdma_resolve_ip - Resolve source and destination IP addresses to
60 * RDMA hardware addresses.
61 * @src_addr: An optional source address to use in the resolution. If a
62 * source address is not provided, a usable address will be returned via
63 * the callback.
64 * @dst_addr: The destination address to resolve.
65 * @addr: A reference to a data location that will receive the resolved
66 * addresses. The data location must remain valid until the callback has
67 * been invoked. The net field of the addr struct must be valid.
68 * @timeout_ms: Amount of time to wait for the address resolution to complete.
69 * @callback: Call invoked once address resolution has completed, timed out,
70 * or been canceled. A status of 0 indicates success.
71 * @resolve_by_gid_attr: Resolve the ip based on the GID attribute from
72 * rdma_dev_addr.
73 * @context: User-specified context associated with the call.
74 */
75int rdma_resolve_ip(struct sockaddr *src_addr, const struct sockaddr *dst_addr,
76 struct rdma_dev_addr *addr, unsigned long timeout_ms,
77 void (*callback)(int status, struct sockaddr *src_addr,
78 struct rdma_dev_addr *addr, void *context),
79 bool resolve_by_gid_attr, void *context);
80
81void rdma_addr_cancel(struct rdma_dev_addr *addr);
82
83int rdma_addr_size(const struct sockaddr *addr);
84int rdma_addr_size_in6(struct sockaddr_in6 *addr);
85int rdma_addr_size_kss(struct __kernel_sockaddr_storage *addr);
86
87static inline u16 ib_addr_get_pkey(struct rdma_dev_addr *dev_addr)
88{
89 return ((u16)dev_addr->broadcast[8] << 8) | (u16)dev_addr->broadcast[9];
90}
91
92static inline void ib_addr_set_pkey(struct rdma_dev_addr *dev_addr, u16 pkey)
93{
94 dev_addr->broadcast[8] = pkey >> 8;
95 dev_addr->broadcast[9] = (unsigned char) pkey;
96}
97
98static inline void ib_addr_get_mgid(struct rdma_dev_addr *dev_addr,
99 union ib_gid *gid)
100{
101 memcpy(gid, dev_addr->broadcast + 4, sizeof *gid);
102}
103
104static inline int rdma_addr_gid_offset(struct rdma_dev_addr *dev_addr)
105{
106 return dev_addr->dev_type == ARPHRD_INFINIBAND ? 4 : 0;
107}
108
109static inline u16 rdma_vlan_dev_vlan_id(const struct net_device *dev)
110{
111 return is_vlan_dev(dev) ? vlan_dev_vlan_id(dev) : 0xffff;
112}
113
114static inline int rdma_ip2gid(struct sockaddr *addr, union ib_gid *gid)
115{
116 switch (addr->sa_family) {
117 case AF_INET:
118 ipv6_addr_set_v4mapped(((struct sockaddr_in *)
119 addr)->sin_addr.s_addr,
120 (struct in6_addr *)gid);
121 break;
122 case AF_INET6:
123 *(struct in6_addr *)&gid->raw =
124 ((struct sockaddr_in6 *)addr)->sin6_addr;
125 break;
126 default:
127 return -EINVAL;
128 }
129 return 0;
130}
131
132/* Important - sockaddr should be a union of sockaddr_in and sockaddr_in6 */
133static inline void rdma_gid2ip(struct sockaddr *out, const union ib_gid *gid)
134{
135 if (ipv6_addr_v4mapped((struct in6_addr *)gid)) {
136 struct sockaddr_in *out_in = (struct sockaddr_in *)out;
137 memset(out_in, 0, sizeof(*out_in));
138 out_in->sin_family = AF_INET;
139 memcpy(&out_in->sin_addr.s_addr, gid->raw + 12, 4);
140 } else {
141 struct sockaddr_in6 *out_in = (struct sockaddr_in6 *)out;
142 memset(out_in, 0, sizeof(*out_in));
143 out_in->sin6_family = AF_INET6;
144 memcpy(&out_in->sin6_addr.s6_addr, gid->raw, 16);
145 }
146}
147
148/*
149 * rdma_get/set_sgid/dgid() APIs are applicable to IB, and iWarp.
150 * They are not applicable to RoCE.
151 * RoCE GIDs are derived from the IP addresses.
152 */
153static inline void rdma_addr_get_sgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid)
154{
155 memcpy(gid, dev_addr->src_dev_addr + rdma_addr_gid_offset(dev_addr),
156 sizeof(*gid));
157}
158
159static inline void rdma_addr_set_sgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid)
160{
161 memcpy(dev_addr->src_dev_addr + rdma_addr_gid_offset(dev_addr), gid, sizeof *gid);
162}
163
164static inline void rdma_addr_get_dgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid)
165{
166 memcpy(gid, dev_addr->dst_dev_addr + rdma_addr_gid_offset(dev_addr), sizeof *gid);
167}
168
169static inline void rdma_addr_set_dgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid)
170{
171 memcpy(dev_addr->dst_dev_addr + rdma_addr_gid_offset(dev_addr), gid, sizeof *gid);
172}
173
174static inline enum ib_mtu iboe_get_mtu(int mtu)
175{
176 /*
177 * Reduce IB headers from effective IBoE MTU.
178 */
179 mtu = mtu - (IB_GRH_BYTES + IB_UDP_BYTES + IB_BTH_BYTES +
180 IB_EXT_XRC_BYTES + IB_EXT_ATOMICETH_BYTES +
181 IB_ICRC_BYTES);
182
183 if (mtu >= ib_mtu_enum_to_int(IB_MTU_4096))
184 return IB_MTU_4096;
185 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_2048))
186 return IB_MTU_2048;
187 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_1024))
188 return IB_MTU_1024;
189 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_512))
190 return IB_MTU_512;
191 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_256))
192 return IB_MTU_256;
193 else
194 return 0;
195}
196
197static inline int rdma_link_local_addr(struct in6_addr *addr)
198{
199 if (addr->s6_addr32[0] == htonl(0xfe800000) &&
200 addr->s6_addr32[1] == 0)
201 return 1;
202
203 return 0;
204}
205
206static inline void rdma_get_ll_mac(struct in6_addr *addr, u8 *mac)
207{
208 memcpy(mac, &addr->s6_addr[8], 3);
209 memcpy(mac + 3, &addr->s6_addr[13], 3);
210 mac[0] ^= 2;
211}
212
213static inline int rdma_is_multicast_addr(struct in6_addr *addr)
214{
215 __be32 ipv4_addr;
216
217 if (addr->s6_addr[0] == 0xff)
218 return 1;
219
220 ipv4_addr = addr->s6_addr32[3];
221 return (ipv6_addr_v4mapped(addr) && ipv4_is_multicast(ipv4_addr));
222}
223
224static inline void rdma_get_mcast_mac(struct in6_addr *addr, u8 *mac)
225{
226 int i;
227
228 mac[0] = 0x33;
229 mac[1] = 0x33;
230 for (i = 2; i < 6; ++i)
231 mac[i] = addr->s6_addr[i + 10];
232}
233
234static inline u16 rdma_get_vlan_id(union ib_gid *dgid)
235{
236 u16 vid;
237
238 vid = dgid->raw[11] << 8 | dgid->raw[12];
239 return vid < 0x1000 ? vid : 0xffff;
240}
241
242static inline struct net_device *rdma_vlan_dev_real_dev(const struct net_device *dev)
243{
244 return is_vlan_dev(dev) ? vlan_dev_real_dev(dev) : NULL;
245}
246
247#endif /* IB_ADDR_H */
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
52struct rdma_addr_client {
53 atomic_t refcount;
54 struct completion comp;
55};
56
57/**
58 * rdma_addr_register_client - Register an address client.
59 */
60void 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 */
66void 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 */
78struct 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 */
96int 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 */
115int 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
122int rdma_resolve_ip_route(struct sockaddr *src_addr,
123 const struct sockaddr *dst_addr,
124 struct rdma_dev_addr *addr);
125
126void rdma_addr_cancel(struct rdma_dev_addr *addr);
127
128int rdma_copy_addr(struct rdma_dev_addr *dev_addr, struct net_device *dev,
129 const unsigned char *dst_dev_addr);
130
131int rdma_addr_size(struct sockaddr *addr);
132
133int rdma_addr_find_smac_by_sgid(union ib_gid *sgid, u8 *smac, u16 *vlan_id);
134int 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
139static 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
144static 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
150static 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
156static 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
161static inline u16 rdma_vlan_dev_vlan_id(const struct net_device *dev)
162{
163 return dev->priv_flags & IFF_802_1Q_VLAN ?
164 vlan_dev_vlan_id(dev) : 0xffff;
165}
166
167static inline int rdma_ip2gid(struct sockaddr *addr, union ib_gid *gid)
168{
169 switch (addr->sa_family) {
170 case AF_INET:
171 ipv6_addr_set_v4mapped(((struct sockaddr_in *)
172 addr)->sin_addr.s_addr,
173 (struct in6_addr *)gid);
174 break;
175 case AF_INET6:
176 memcpy(gid->raw, &((struct sockaddr_in6 *)addr)->sin6_addr, 16);
177 break;
178 default:
179 return -EINVAL;
180 }
181 return 0;
182}
183
184/* Important - sockaddr should be a union of sockaddr_in and sockaddr_in6 */
185static inline void rdma_gid2ip(struct sockaddr *out, const union ib_gid *gid)
186{
187 if (ipv6_addr_v4mapped((struct in6_addr *)gid)) {
188 struct sockaddr_in *out_in = (struct sockaddr_in *)out;
189 memset(out_in, 0, sizeof(*out_in));
190 out_in->sin_family = AF_INET;
191 memcpy(&out_in->sin_addr.s_addr, gid->raw + 12, 4);
192 } else {
193 struct sockaddr_in6 *out_in = (struct sockaddr_in6 *)out;
194 memset(out_in, 0, sizeof(*out_in));
195 out_in->sin6_family = AF_INET6;
196 memcpy(&out_in->sin6_addr.s6_addr, gid->raw, 16);
197 }
198}
199
200static inline void iboe_addr_get_sgid(struct rdma_dev_addr *dev_addr,
201 union ib_gid *gid)
202{
203 struct net_device *dev;
204 struct in_device *ip4;
205
206 dev = dev_get_by_index(&init_net, dev_addr->bound_dev_if);
207 if (dev) {
208 ip4 = (struct in_device *)dev->ip_ptr;
209 if (ip4 && ip4->ifa_list && ip4->ifa_list->ifa_address)
210 ipv6_addr_set_v4mapped(ip4->ifa_list->ifa_address,
211 (struct in6_addr *)gid);
212 dev_put(dev);
213 }
214}
215
216static inline void rdma_addr_get_sgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid)
217{
218 if (dev_addr->transport == RDMA_TRANSPORT_IB &&
219 dev_addr->dev_type != ARPHRD_INFINIBAND)
220 iboe_addr_get_sgid(dev_addr, gid);
221 else
222 memcpy(gid, dev_addr->src_dev_addr +
223 rdma_addr_gid_offset(dev_addr), sizeof *gid);
224}
225
226static inline void rdma_addr_set_sgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid)
227{
228 memcpy(dev_addr->src_dev_addr + rdma_addr_gid_offset(dev_addr), gid, sizeof *gid);
229}
230
231static inline void rdma_addr_get_dgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid)
232{
233 memcpy(gid, dev_addr->dst_dev_addr + rdma_addr_gid_offset(dev_addr), sizeof *gid);
234}
235
236static inline void rdma_addr_set_dgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid)
237{
238 memcpy(dev_addr->dst_dev_addr + rdma_addr_gid_offset(dev_addr), gid, sizeof *gid);
239}
240
241static inline enum ib_mtu iboe_get_mtu(int mtu)
242{
243 /*
244 * reduce IB headers from effective IBoE MTU. 28 stands for
245 * atomic header which is the biggest possible header after BTH
246 */
247 mtu = mtu - IB_GRH_BYTES - IB_BTH_BYTES - 28;
248
249 if (mtu >= ib_mtu_enum_to_int(IB_MTU_4096))
250 return IB_MTU_4096;
251 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_2048))
252 return IB_MTU_2048;
253 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_1024))
254 return IB_MTU_1024;
255 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_512))
256 return IB_MTU_512;
257 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_256))
258 return IB_MTU_256;
259 else
260 return 0;
261}
262
263static inline int iboe_get_rate(struct net_device *dev)
264{
265 struct ethtool_link_ksettings cmd;
266 int err;
267
268 rtnl_lock();
269 err = __ethtool_get_link_ksettings(dev, &cmd);
270 rtnl_unlock();
271 if (err)
272 return IB_RATE_PORT_CURRENT;
273
274 if (cmd.base.speed >= 40000)
275 return IB_RATE_40_GBPS;
276 else if (cmd.base.speed >= 30000)
277 return IB_RATE_30_GBPS;
278 else if (cmd.base.speed >= 20000)
279 return IB_RATE_20_GBPS;
280 else if (cmd.base.speed >= 10000)
281 return IB_RATE_10_GBPS;
282 else
283 return IB_RATE_PORT_CURRENT;
284}
285
286static 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
295static 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
302static inline int rdma_is_multicast_addr(struct in6_addr *addr)
303{
304 return addr->s6_addr[0] == 0xff;
305}
306
307static 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
317static 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
325static 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 */