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 * $FreeBSD$ 36 */ 37 38 #if !defined(IB_ADDR_H) 39 #define IB_ADDR_H 40 41 #include <linux/in.h> 42 #include <linux/in6.h> 43 #include <linux/if_arp.h> 44 #include <linux/netdevice.h> 45 #include <linux/socket.h> 46 #include <linux/if_vlan.h> 47 #include <net/ipv6.h> 48 #include <net/if_inet6.h> 49 #include <net/ip.h> 50 #include <rdma/ib_verbs.h> 51 #include <rdma/ib_pack.h> 52 #include <rdma/ib_addr_freebsd.h> 53 54 /* Linux netdevice.h but for working on an ifnet rather than a net_device. */ 55 #define dev_hold(d) if_ref(d) 56 #define dev_put(d) if_rele(d) 57 #define dev_net(d) if_getvnet(d) 58 #define net_eq(a,b) ((a) == (b)) 59 60 61 struct rdma_addr_client { 62 atomic_t refcount; 63 struct completion comp; 64 }; 65 66 union rdma_sockaddr { 67 struct sockaddr _sockaddr; 68 struct sockaddr_in _sockaddr_in; 69 struct sockaddr_in6 _sockaddr_in6; 70 struct sockaddr_storage _sockaddr_ss; 71 }; 72 73 /** 74 * rdma_addr_register_client - Register an address client. 75 */ 76 void rdma_addr_register_client(struct rdma_addr_client *client); 77 78 /** 79 * rdma_addr_unregister_client - Deregister an address client. 80 * @client: Client object to deregister. 81 */ 82 void rdma_addr_unregister_client(struct rdma_addr_client *client); 83 84 /** 85 * struct rdma_dev_addr - Contains resolved RDMA hardware addresses 86 * @src_dev_addr: Source MAC address. 87 * @dst_dev_addr: Destination MAC address. 88 * @broadcast: Broadcast address of the device. 89 * @dev_type: The interface hardware type of the device. 90 * @bound_dev_if: An optional device interface index. 91 * @transport: The transport type used. 92 * @net: Network namespace containing the bound_dev_if net_dev. 93 */ 94 struct vnet; 95 struct rdma_dev_addr { 96 unsigned char src_dev_addr[MAX_ADDR_LEN]; 97 unsigned char dst_dev_addr[MAX_ADDR_LEN]; 98 unsigned char broadcast[MAX_ADDR_LEN]; 99 unsigned short dev_type; 100 int bound_dev_if; 101 enum rdma_transport_type transport; 102 struct vnet *net; 103 enum rdma_network_type network; 104 int hoplimit; 105 }; 106 107 /** 108 * rdma_translate_ip - Translate a local IP address to an RDMA hardware 109 * address. 110 * 111 * The dev_addr->net and dev_addr->bound_dev_if fields must be initialized. 112 */ 113 int rdma_translate_ip(const struct sockaddr *addr, 114 struct rdma_dev_addr *dev_addr); 115 116 /** 117 * rdma_resolve_ip - Resolve source and destination IP addresses to 118 * RDMA hardware addresses. 119 * @client: Address client associated with request. 120 * @src_addr: An optional source address to use in the resolution. If a 121 * source address is not provided, a usable address will be returned via 122 * the callback. 123 * @dst_addr: The destination address to resolve. 124 * @addr: A reference to a data location that will receive the resolved 125 * addresses. The data location must remain valid until the callback has 126 * been invoked. The net field of the addr struct must be valid. 127 * @timeout_ms: Amount of time to wait for the address resolution to complete. 128 * @callback: Call invoked once address resolution has completed, timed out, 129 * or been canceled. A status of 0 indicates success. 130 * @context: User-specified context associated with the call. 131 */ 132 int rdma_resolve_ip(struct rdma_addr_client *client, 133 struct sockaddr *src_addr, struct sockaddr *dst_addr, 134 struct rdma_dev_addr *addr, int timeout_ms, 135 void (*callback)(int status, struct sockaddr *src_addr, 136 struct rdma_dev_addr *addr, void *context), 137 void *context); 138 139 int rdma_resolve_ip_route(struct sockaddr *src_addr, 140 const struct sockaddr *dst_addr, 141 struct rdma_dev_addr *addr); 142 143 void rdma_addr_cancel(struct rdma_dev_addr *addr); 144 145 int rdma_copy_addr(struct rdma_dev_addr *dev_addr, if_t dev, 146 const unsigned char *dst_dev_addr); 147 148 int rdma_addr_size(struct sockaddr *addr); 149 int rdma_addr_size_in6(struct sockaddr_in6 *addr); 150 int rdma_addr_size_kss(struct sockaddr_storage *addr); 151 152 int rdma_addr_find_l2_eth_by_grh(const union ib_gid *sgid, 153 const union ib_gid *dgid, 154 u8 *smac, if_t dev, 155 int *hoplimit); 156 157 static inline u16 ib_addr_get_pkey(struct rdma_dev_addr *dev_addr) 158 { 159 return ((u16)dev_addr->broadcast[8] << 8) | (u16)dev_addr->broadcast[9]; 160 } 161 162 static inline void ib_addr_set_pkey(struct rdma_dev_addr *dev_addr, u16 pkey) 163 { 164 dev_addr->broadcast[8] = pkey >> 8; 165 dev_addr->broadcast[9] = (unsigned char) pkey; 166 } 167 168 static inline void ib_addr_get_mgid(struct rdma_dev_addr *dev_addr, 169 union ib_gid *gid) 170 { 171 memcpy(gid, dev_addr->broadcast + 4, sizeof *gid); 172 } 173 174 static inline int rdma_addr_gid_offset(struct rdma_dev_addr *dev_addr) 175 { 176 return dev_addr->dev_type == ARPHRD_INFINIBAND ? 4 : 0; 177 } 178 179 static inline u16 rdma_vlan_dev_vlan_id(if_t dev) 180 { 181 uint16_t tag; 182 183 if (if_gettype(dev) == IFT_ETHER && if_getpcp(dev) != IFNET_PCP_NONE) 184 return 0x0000; /* prio-tagged traffic */ 185 if (VLAN_TAG(__DECONST(if_t, dev), &tag) != 0) 186 return 0xffff; 187 return tag; 188 } 189 190 static inline int rdma_ip2gid(const struct sockaddr *addr, union ib_gid *gid) 191 { 192 switch (addr->sa_family) { 193 case AF_INET: 194 ipv6_addr_set_v4mapped(((const struct sockaddr_in *) 195 addr)->sin_addr.s_addr, 196 (struct in6_addr *)gid); 197 break; 198 case AF_INET6: 199 memcpy(gid->raw, &((const struct sockaddr_in6 *)addr)->sin6_addr, 16); 200 /* make sure scope ID gets zeroed inside GID */ 201 if (IN6_IS_SCOPE_LINKLOCAL((struct in6_addr *)gid->raw) || 202 IN6_IS_ADDR_MC_INTFACELOCAL((struct in6_addr *)gid->raw)) { 203 gid->raw[2] = 0; 204 gid->raw[3] = 0; 205 } 206 break; 207 default: 208 return -EINVAL; 209 } 210 return 0; 211 } 212 213 /* Important - sockaddr should be a union of sockaddr_in and sockaddr_in6 */ 214 static inline void rdma_gid2ip(struct sockaddr *out, const union ib_gid *gid) 215 { 216 if (ipv6_addr_v4mapped((const struct in6_addr *)gid)) { 217 struct sockaddr_in *out_in = (struct sockaddr_in *)out; 218 memset(out_in, 0, sizeof(*out_in)); 219 out_in->sin_len = sizeof(*out_in); 220 out_in->sin_family = AF_INET; 221 memcpy(&out_in->sin_addr.s_addr, gid->raw + 12, 4); 222 } else { 223 struct sockaddr_in6 *out_in = (struct sockaddr_in6 *)out; 224 memset(out_in, 0, sizeof(*out_in)); 225 out_in->sin6_len = sizeof(*out_in); 226 out_in->sin6_family = AF_INET6; 227 memcpy(&out_in->sin6_addr.s6_addr, gid->raw, 16); 228 } 229 } 230 231 static u_int 232 _iboe_addr_get_sgid_ia_cb(void *arg, struct ifaddr *ifa, u_int count __unused) 233 { 234 ipv6_addr_set_v4mapped(((struct sockaddr_in *) 235 ifa->ifa_addr)->sin_addr.s_addr, 236 (struct in6_addr *)arg); 237 return (0); 238 } 239 240 static inline void iboe_addr_get_sgid(struct rdma_dev_addr *dev_addr, 241 union ib_gid *gid) 242 { 243 if_t dev; 244 245 #ifdef VIMAGE 246 if (dev_addr->net == NULL) 247 return; 248 #endif 249 dev = dev_get_by_index(dev_addr->net, dev_addr->bound_dev_if); 250 if (dev) { 251 if_foreach_addr_type(dev, AF_INET, 252 _iboe_addr_get_sgid_ia_cb, gid); 253 dev_put(dev); 254 } 255 } 256 257 static inline void rdma_addr_get_sgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid) 258 { 259 if (dev_addr->transport == RDMA_TRANSPORT_IB && 260 dev_addr->dev_type != ARPHRD_INFINIBAND) 261 iboe_addr_get_sgid(dev_addr, gid); 262 else 263 memcpy(gid, dev_addr->src_dev_addr + 264 rdma_addr_gid_offset(dev_addr), sizeof *gid); 265 } 266 267 static inline void rdma_addr_set_sgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid) 268 { 269 memcpy(dev_addr->src_dev_addr + rdma_addr_gid_offset(dev_addr), gid, sizeof *gid); 270 } 271 272 static inline void rdma_addr_get_dgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid) 273 { 274 memcpy(gid, dev_addr->dst_dev_addr + rdma_addr_gid_offset(dev_addr), sizeof *gid); 275 } 276 277 static inline void rdma_addr_set_dgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid) 278 { 279 memcpy(dev_addr->dst_dev_addr + rdma_addr_gid_offset(dev_addr), gid, sizeof *gid); 280 } 281 282 static inline enum ib_mtu iboe_get_mtu(int mtu) 283 { 284 /* 285 * reduce IB headers from effective IBoE MTU. 28 stands for 286 * atomic header which is the biggest possible header after BTH 287 */ 288 mtu = mtu - IB_GRH_BYTES - IB_BTH_BYTES - 28; 289 290 if (mtu >= ib_mtu_enum_to_int(IB_MTU_4096)) 291 return IB_MTU_4096; 292 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_2048)) 293 return IB_MTU_2048; 294 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_1024)) 295 return IB_MTU_1024; 296 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_512)) 297 return IB_MTU_512; 298 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_256)) 299 return IB_MTU_256; 300 else 301 return 0; 302 } 303 304 static inline int iboe_get_rate(if_t dev) 305 { 306 uint64_t baudrate = if_getbaudrate(dev); 307 #ifdef if_baudrate_pf 308 int exp; 309 for (exp = dev->if_baudrate_pf; exp > 0; exp--) 310 baudrate *= 10; 311 #endif 312 if (baudrate >= IF_Gbps(40)) 313 return IB_RATE_40_GBPS; 314 else if (baudrate >= IF_Gbps(30)) 315 return IB_RATE_30_GBPS; 316 else if (baudrate >= IF_Gbps(20)) 317 return IB_RATE_20_GBPS; 318 else if (baudrate >= IF_Gbps(10)) 319 return IB_RATE_10_GBPS; 320 else 321 return IB_RATE_PORT_CURRENT; 322 } 323 324 static inline int rdma_link_local_addr(struct in6_addr *addr) 325 { 326 if (addr->s6_addr32[0] == htonl(0xfe800000) && 327 addr->s6_addr32[1] == 0) 328 return 1; 329 330 return 0; 331 } 332 333 static inline void rdma_get_ll_mac(struct in6_addr *addr, u8 *mac) 334 { 335 memcpy(mac, &addr->s6_addr[8], 3); 336 memcpy(mac + 3, &addr->s6_addr[13], 3); 337 mac[0] ^= 2; 338 } 339 340 static inline int rdma_is_multicast_addr(struct in6_addr *addr) 341 { 342 __be32 ipv4_addr; 343 344 if (addr->s6_addr[0] == 0xff) 345 return 1; 346 347 ipv4_addr = addr->s6_addr32[3]; 348 return (ipv6_addr_v4mapped(addr) && ipv4_is_multicast(ipv4_addr)); 349 } 350 351 static inline void rdma_get_mcast_mac(struct in6_addr *addr, u8 *mac) 352 { 353 int i; 354 355 mac[0] = 0x33; 356 mac[1] = 0x33; 357 for (i = 2; i < 6; ++i) 358 mac[i] = addr->s6_addr[i + 10]; 359 } 360 361 static inline u16 rdma_get_vlan_id(union ib_gid *dgid) 362 { 363 u16 vid; 364 365 vid = dgid->raw[11] << 8 | dgid->raw[12]; 366 return vid < 0x1000 ? vid : 0xffff; 367 } 368 369 static inline if_t rdma_vlan_dev_real_dev(if_t dev) 370 { 371 struct epoch_tracker et; 372 373 NET_EPOCH_ENTER(et); 374 if (if_gettype(dev) != IFT_ETHER || if_getpcp(dev) == IFNET_PCP_NONE) 375 dev = VLAN_TRUNKDEV(dev); /* non prio-tagged traffic */ 376 NET_EPOCH_EXIT(et); 377 return (dev); 378 } 379 380 #endif /* IB_ADDR_H */ 381