Loading...
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/socket.h>
42#include <linux/if_vlan.h>
43#include <rdma/ib_verbs.h>
44#include <rdma/ib_pack.h>
45
46struct rdma_addr_client {
47 atomic_t refcount;
48 struct completion comp;
49};
50
51/**
52 * rdma_addr_register_client - Register an address client.
53 */
54void rdma_addr_register_client(struct rdma_addr_client *client);
55
56/**
57 * rdma_addr_unregister_client - Deregister an address client.
58 * @client: Client object to deregister.
59 */
60void rdma_addr_unregister_client(struct rdma_addr_client *client);
61
62struct rdma_dev_addr {
63 unsigned char src_dev_addr[MAX_ADDR_LEN];
64 unsigned char dst_dev_addr[MAX_ADDR_LEN];
65 unsigned char broadcast[MAX_ADDR_LEN];
66 unsigned short dev_type;
67 int bound_dev_if;
68 enum rdma_transport_type transport;
69};
70
71/**
72 * rdma_translate_ip - Translate a local IP address to an RDMA hardware
73 * address.
74 */
75int rdma_translate_ip(struct sockaddr *addr, struct rdma_dev_addr *dev_addr);
76
77/**
78 * rdma_resolve_ip - Resolve source and destination IP addresses to
79 * RDMA hardware addresses.
80 * @client: Address client associated with request.
81 * @src_addr: An optional source address to use in the resolution. If a
82 * source address is not provided, a usable address will be returned via
83 * the callback.
84 * @dst_addr: The destination address to resolve.
85 * @addr: A reference to a data location that will receive the resolved
86 * addresses. The data location must remain valid until the callback has
87 * been invoked.
88 * @timeout_ms: Amount of time to wait for the address resolution to complete.
89 * @callback: Call invoked once address resolution has completed, timed out,
90 * or been canceled. A status of 0 indicates success.
91 * @context: User-specified context associated with the call.
92 */
93int rdma_resolve_ip(struct rdma_addr_client *client,
94 struct sockaddr *src_addr, struct sockaddr *dst_addr,
95 struct rdma_dev_addr *addr, int timeout_ms,
96 void (*callback)(int status, struct sockaddr *src_addr,
97 struct rdma_dev_addr *addr, void *context),
98 void *context);
99
100void rdma_addr_cancel(struct rdma_dev_addr *addr);
101
102int rdma_copy_addr(struct rdma_dev_addr *dev_addr, struct net_device *dev,
103 const unsigned char *dst_dev_addr);
104
105static inline int ip_addr_size(struct sockaddr *addr)
106{
107 return addr->sa_family == AF_INET6 ?
108 sizeof(struct sockaddr_in6) : sizeof(struct sockaddr_in);
109}
110
111static inline u16 ib_addr_get_pkey(struct rdma_dev_addr *dev_addr)
112{
113 return ((u16)dev_addr->broadcast[8] << 8) | (u16)dev_addr->broadcast[9];
114}
115
116static inline void ib_addr_set_pkey(struct rdma_dev_addr *dev_addr, u16 pkey)
117{
118 dev_addr->broadcast[8] = pkey >> 8;
119 dev_addr->broadcast[9] = (unsigned char) pkey;
120}
121
122static inline void ib_addr_get_mgid(struct rdma_dev_addr *dev_addr,
123 union ib_gid *gid)
124{
125 memcpy(gid, dev_addr->broadcast + 4, sizeof *gid);
126}
127
128static inline int rdma_addr_gid_offset(struct rdma_dev_addr *dev_addr)
129{
130 return dev_addr->dev_type == ARPHRD_INFINIBAND ? 4 : 0;
131}
132
133static inline void iboe_mac_vlan_to_ll(union ib_gid *gid, u8 *mac, u16 vid)
134{
135 memset(gid->raw, 0, 16);
136 *((__be32 *) gid->raw) = cpu_to_be32(0xfe800000);
137 if (vid < 0x1000) {
138 gid->raw[12] = vid & 0xff;
139 gid->raw[11] = vid >> 8;
140 } else {
141 gid->raw[12] = 0xfe;
142 gid->raw[11] = 0xff;
143 }
144 memcpy(gid->raw + 13, mac + 3, 3);
145 memcpy(gid->raw + 8, mac, 3);
146 gid->raw[8] ^= 2;
147}
148
149static inline u16 rdma_vlan_dev_vlan_id(const struct net_device *dev)
150{
151 return dev->priv_flags & IFF_802_1Q_VLAN ?
152 vlan_dev_vlan_id(dev) : 0xffff;
153}
154
155static inline void iboe_addr_get_sgid(struct rdma_dev_addr *dev_addr,
156 union ib_gid *gid)
157{
158 struct net_device *dev;
159 u16 vid = 0xffff;
160
161 dev = dev_get_by_index(&init_net, dev_addr->bound_dev_if);
162 if (dev) {
163 vid = rdma_vlan_dev_vlan_id(dev);
164 dev_put(dev);
165 }
166
167 iboe_mac_vlan_to_ll(gid, dev_addr->src_dev_addr, vid);
168}
169
170static inline void rdma_addr_get_sgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid)
171{
172 if (dev_addr->transport == RDMA_TRANSPORT_IB &&
173 dev_addr->dev_type != ARPHRD_INFINIBAND)
174 iboe_addr_get_sgid(dev_addr, gid);
175 else
176 memcpy(gid, dev_addr->src_dev_addr +
177 rdma_addr_gid_offset(dev_addr), sizeof *gid);
178}
179
180static inline void rdma_addr_set_sgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid)
181{
182 memcpy(dev_addr->src_dev_addr + rdma_addr_gid_offset(dev_addr), gid, sizeof *gid);
183}
184
185static inline void rdma_addr_get_dgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid)
186{
187 memcpy(gid, dev_addr->dst_dev_addr + rdma_addr_gid_offset(dev_addr), sizeof *gid);
188}
189
190static inline void rdma_addr_set_dgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid)
191{
192 memcpy(dev_addr->dst_dev_addr + rdma_addr_gid_offset(dev_addr), gid, sizeof *gid);
193}
194
195static inline enum ib_mtu iboe_get_mtu(int mtu)
196{
197 /*
198 * reduce IB headers from effective IBoE MTU. 28 stands for
199 * atomic header which is the biggest possible header after BTH
200 */
201 mtu = mtu - IB_GRH_BYTES - IB_BTH_BYTES - 28;
202
203 if (mtu >= ib_mtu_enum_to_int(IB_MTU_4096))
204 return IB_MTU_4096;
205 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_2048))
206 return IB_MTU_2048;
207 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_1024))
208 return IB_MTU_1024;
209 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_512))
210 return IB_MTU_512;
211 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_256))
212 return IB_MTU_256;
213 else
214 return 0;
215}
216
217static inline int iboe_get_rate(struct net_device *dev)
218{
219 struct ethtool_cmd cmd;
220 u32 speed;
221
222 if (dev_ethtool_get_settings(dev, &cmd))
223 return IB_RATE_PORT_CURRENT;
224
225 speed = ethtool_cmd_speed(&cmd);
226 if (speed >= 40000)
227 return IB_RATE_40_GBPS;
228 else if (speed >= 30000)
229 return IB_RATE_30_GBPS;
230 else if (speed >= 20000)
231 return IB_RATE_20_GBPS;
232 else if (speed >= 10000)
233 return IB_RATE_10_GBPS;
234 else
235 return IB_RATE_PORT_CURRENT;
236}
237
238static inline int rdma_link_local_addr(struct in6_addr *addr)
239{
240 if (addr->s6_addr32[0] == htonl(0xfe800000) &&
241 addr->s6_addr32[1] == 0)
242 return 1;
243
244 return 0;
245}
246
247static inline void rdma_get_ll_mac(struct in6_addr *addr, u8 *mac)
248{
249 memcpy(mac, &addr->s6_addr[8], 3);
250 memcpy(mac + 3, &addr->s6_addr[13], 3);
251 mac[0] ^= 2;
252}
253
254static inline int rdma_is_multicast_addr(struct in6_addr *addr)
255{
256 return addr->s6_addr[0] == 0xff;
257}
258
259static inline void rdma_get_mcast_mac(struct in6_addr *addr, u8 *mac)
260{
261 int i;
262
263 mac[0] = 0x33;
264 mac[1] = 0x33;
265 for (i = 2; i < 6; ++i)
266 mac[i] = addr->s6_addr[i + 10];
267}
268
269static inline u16 rdma_get_vlan_id(union ib_gid *dgid)
270{
271 u16 vid;
272
273 vid = dgid->raw[11] << 8 | dgid->raw[12];
274 return vid < 0x1000 ? vid : 0xffff;
275}
276
277static inline struct net_device *rdma_vlan_dev_real_dev(const struct net_device *dev)
278{
279 return dev->priv_flags & IFF_802_1Q_VLAN ?
280 vlan_dev_real_dev(dev) : 0;
281}
282
283#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/net_namespace.h>
50
51/**
52 * struct rdma_dev_addr - Contains resolved RDMA hardware addresses
53 * @src_dev_addr: Source MAC address.
54 * @dst_dev_addr: Destination MAC address.
55 * @broadcast: Broadcast address of the device.
56 * @dev_type: The interface hardware type of the device.
57 * @bound_dev_if: An optional device interface index.
58 * @transport: The transport type used.
59 * @net: Network namespace containing the bound_dev_if net_dev.
60 * @sgid_attr: GID attribute to use for identified SGID
61 */
62struct rdma_dev_addr {
63 unsigned char src_dev_addr[MAX_ADDR_LEN];
64 unsigned char dst_dev_addr[MAX_ADDR_LEN];
65 unsigned char broadcast[MAX_ADDR_LEN];
66 unsigned short dev_type;
67 int bound_dev_if;
68 enum rdma_transport_type transport;
69 struct net *net;
70 const struct ib_gid_attr *sgid_attr;
71 enum rdma_network_type network;
72 int hoplimit;
73};
74
75/**
76 * rdma_translate_ip - Translate a local IP address to an RDMA hardware
77 * address.
78 *
79 * The dev_addr->net field must be initialized.
80 */
81int rdma_translate_ip(const struct sockaddr *addr,
82 struct rdma_dev_addr *dev_addr);
83
84/**
85 * rdma_resolve_ip - Resolve source and destination IP addresses to
86 * RDMA hardware addresses.
87 * @src_addr: An optional source address to use in the resolution. If a
88 * source address is not provided, a usable address will be returned via
89 * the callback.
90 * @dst_addr: The destination address to resolve.
91 * @addr: A reference to a data location that will receive the resolved
92 * addresses. The data location must remain valid until the callback has
93 * been invoked. The net field of the addr struct must be valid.
94 * @timeout_ms: Amount of time to wait for the address resolution to complete.
95 * @callback: Call invoked once address resolution has completed, timed out,
96 * or been canceled. A status of 0 indicates success.
97 * @resolve_by_gid_attr: Resolve the ip based on the GID attribute from
98 * rdma_dev_addr.
99 * @context: User-specified context associated with the call.
100 */
101int rdma_resolve_ip(struct sockaddr *src_addr, const struct sockaddr *dst_addr,
102 struct rdma_dev_addr *addr, unsigned long timeout_ms,
103 void (*callback)(int status, struct sockaddr *src_addr,
104 struct rdma_dev_addr *addr, void *context),
105 bool resolve_by_gid_attr, void *context);
106
107void rdma_addr_cancel(struct rdma_dev_addr *addr);
108
109int rdma_addr_size(const struct sockaddr *addr);
110int rdma_addr_size_in6(struct sockaddr_in6 *addr);
111int rdma_addr_size_kss(struct __kernel_sockaddr_storage *addr);
112
113static inline u16 ib_addr_get_pkey(struct rdma_dev_addr *dev_addr)
114{
115 return ((u16)dev_addr->broadcast[8] << 8) | (u16)dev_addr->broadcast[9];
116}
117
118static inline void ib_addr_set_pkey(struct rdma_dev_addr *dev_addr, u16 pkey)
119{
120 dev_addr->broadcast[8] = pkey >> 8;
121 dev_addr->broadcast[9] = (unsigned char) pkey;
122}
123
124static inline void ib_addr_get_mgid(struct rdma_dev_addr *dev_addr,
125 union ib_gid *gid)
126{
127 memcpy(gid, dev_addr->broadcast + 4, sizeof *gid);
128}
129
130static inline int rdma_addr_gid_offset(struct rdma_dev_addr *dev_addr)
131{
132 return dev_addr->dev_type == ARPHRD_INFINIBAND ? 4 : 0;
133}
134
135static inline u16 rdma_vlan_dev_vlan_id(const struct net_device *dev)
136{
137 return is_vlan_dev(dev) ? vlan_dev_vlan_id(dev) : 0xffff;
138}
139
140static inline int rdma_ip2gid(struct sockaddr *addr, union ib_gid *gid)
141{
142 switch (addr->sa_family) {
143 case AF_INET:
144 ipv6_addr_set_v4mapped(((struct sockaddr_in *)
145 addr)->sin_addr.s_addr,
146 (struct in6_addr *)gid);
147 break;
148 case AF_INET6:
149 *(struct in6_addr *)&gid->raw =
150 ((struct sockaddr_in6 *)addr)->sin6_addr;
151 break;
152 default:
153 return -EINVAL;
154 }
155 return 0;
156}
157
158/* Important - sockaddr should be a union of sockaddr_in and sockaddr_in6 */
159static inline void rdma_gid2ip(struct sockaddr *out, const union ib_gid *gid)
160{
161 if (ipv6_addr_v4mapped((struct in6_addr *)gid)) {
162 struct sockaddr_in *out_in = (struct sockaddr_in *)out;
163 memset(out_in, 0, sizeof(*out_in));
164 out_in->sin_family = AF_INET;
165 memcpy(&out_in->sin_addr.s_addr, gid->raw + 12, 4);
166 } else {
167 struct sockaddr_in6 *out_in = (struct sockaddr_in6 *)out;
168 memset(out_in, 0, sizeof(*out_in));
169 out_in->sin6_family = AF_INET6;
170 memcpy(&out_in->sin6_addr.s6_addr, gid->raw, 16);
171 }
172}
173
174/*
175 * rdma_get/set_sgid/dgid() APIs are applicable to IB, and iWarp.
176 * They are not applicable to RoCE.
177 * RoCE GIDs are derived from the IP addresses.
178 */
179static inline void rdma_addr_get_sgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid)
180{
181 memcpy(gid, dev_addr->src_dev_addr + rdma_addr_gid_offset(dev_addr),
182 sizeof(*gid));
183}
184
185static inline void rdma_addr_set_sgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid)
186{
187 memcpy(dev_addr->src_dev_addr + rdma_addr_gid_offset(dev_addr), gid, sizeof *gid);
188}
189
190static inline void rdma_addr_get_dgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid)
191{
192 memcpy(gid, dev_addr->dst_dev_addr + rdma_addr_gid_offset(dev_addr), sizeof *gid);
193}
194
195static inline void rdma_addr_set_dgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid)
196{
197 memcpy(dev_addr->dst_dev_addr + rdma_addr_gid_offset(dev_addr), gid, sizeof *gid);
198}
199
200static inline enum ib_mtu iboe_get_mtu(int mtu)
201{
202 /*
203 * Reduce IB headers from effective IBoE MTU.
204 */
205 mtu = mtu - (IB_GRH_BYTES + IB_UDP_BYTES + IB_BTH_BYTES +
206 IB_EXT_XRC_BYTES + IB_EXT_ATOMICETH_BYTES +
207 IB_ICRC_BYTES);
208
209 if (mtu >= ib_mtu_enum_to_int(IB_MTU_4096))
210 return IB_MTU_4096;
211 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_2048))
212 return IB_MTU_2048;
213 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_1024))
214 return IB_MTU_1024;
215 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_512))
216 return IB_MTU_512;
217 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_256))
218 return IB_MTU_256;
219 else
220 return 0;
221}
222
223static inline int iboe_get_rate(struct net_device *dev)
224{
225 struct ethtool_link_ksettings cmd;
226 int err;
227
228 rtnl_lock();
229 err = __ethtool_get_link_ksettings(dev, &cmd);
230 rtnl_unlock();
231 if (err)
232 return IB_RATE_PORT_CURRENT;
233
234 if (cmd.base.speed >= 40000)
235 return IB_RATE_40_GBPS;
236 else if (cmd.base.speed >= 30000)
237 return IB_RATE_30_GBPS;
238 else if (cmd.base.speed >= 20000)
239 return IB_RATE_20_GBPS;
240 else if (cmd.base.speed >= 10000)
241 return IB_RATE_10_GBPS;
242 else
243 return IB_RATE_PORT_CURRENT;
244}
245
246static inline int rdma_link_local_addr(struct in6_addr *addr)
247{
248 if (addr->s6_addr32[0] == htonl(0xfe800000) &&
249 addr->s6_addr32[1] == 0)
250 return 1;
251
252 return 0;
253}
254
255static inline void rdma_get_ll_mac(struct in6_addr *addr, u8 *mac)
256{
257 memcpy(mac, &addr->s6_addr[8], 3);
258 memcpy(mac + 3, &addr->s6_addr[13], 3);
259 mac[0] ^= 2;
260}
261
262static inline int rdma_is_multicast_addr(struct in6_addr *addr)
263{
264 __be32 ipv4_addr;
265
266 if (addr->s6_addr[0] == 0xff)
267 return 1;
268
269 ipv4_addr = addr->s6_addr32[3];
270 return (ipv6_addr_v4mapped(addr) && ipv4_is_multicast(ipv4_addr));
271}
272
273static inline void rdma_get_mcast_mac(struct in6_addr *addr, u8 *mac)
274{
275 int i;
276
277 mac[0] = 0x33;
278 mac[1] = 0x33;
279 for (i = 2; i < 6; ++i)
280 mac[i] = addr->s6_addr[i + 10];
281}
282
283static inline u16 rdma_get_vlan_id(union ib_gid *dgid)
284{
285 u16 vid;
286
287 vid = dgid->raw[11] << 8 | dgid->raw[12];
288 return vid < 0x1000 ? vid : 0xffff;
289}
290
291static inline struct net_device *rdma_vlan_dev_real_dev(const struct net_device *dev)
292{
293 return is_vlan_dev(dev) ? vlan_dev_real_dev(dev) : NULL;
294}
295
296#endif /* IB_ADDR_H */