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