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