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 52 struct rdma_addr_client { 53 atomic_t refcount; 54 struct completion comp; 55 }; 56 57 /** 58 * rdma_addr_register_client - Register an address client. 59 */ 60 void 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 */ 66 void 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 */ 78 struct 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 */ 96 int rdma_translate_ip(const struct sockaddr *addr, 97 struct rdma_dev_addr *dev_addr); 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 */ 115 int 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); 121 122 int rdma_resolve_ip_route(struct sockaddr *src_addr, 123 const struct sockaddr *dst_addr, 124 struct rdma_dev_addr *addr); 125 126 void rdma_addr_cancel(struct rdma_dev_addr *addr); 127 128 void rdma_copy_addr(struct rdma_dev_addr *dev_addr, 129 const struct net_device *dev, 130 const unsigned char *dst_dev_addr); 131 132 int rdma_addr_size(struct sockaddr *addr); 133 int rdma_addr_size_in6(struct sockaddr_in6 *addr); 134 int rdma_addr_size_kss(struct __kernel_sockaddr_storage *addr); 135 136 int rdma_addr_find_l2_eth_by_grh(const union ib_gid *sgid, 137 const union ib_gid *dgid, 138 u8 *dmac, const struct net_device *ndev, 139 int *hoplimit); 140 141 static inline u16 ib_addr_get_pkey(struct rdma_dev_addr *dev_addr) 142 { 143 return ((u16)dev_addr->broadcast[8] << 8) | (u16)dev_addr->broadcast[9]; 144 } 145 146 static inline void ib_addr_set_pkey(struct rdma_dev_addr *dev_addr, u16 pkey) 147 { 148 dev_addr->broadcast[8] = pkey >> 8; 149 dev_addr->broadcast[9] = (unsigned char) pkey; 150 } 151 152 static inline void ib_addr_get_mgid(struct rdma_dev_addr *dev_addr, 153 union ib_gid *gid) 154 { 155 memcpy(gid, dev_addr->broadcast + 4, sizeof *gid); 156 } 157 158 static inline int rdma_addr_gid_offset(struct rdma_dev_addr *dev_addr) 159 { 160 return dev_addr->dev_type == ARPHRD_INFINIBAND ? 4 : 0; 161 } 162 163 static inline u16 rdma_vlan_dev_vlan_id(const struct net_device *dev) 164 { 165 return is_vlan_dev(dev) ? vlan_dev_vlan_id(dev) : 0xffff; 166 } 167 168 static inline int rdma_ip2gid(struct sockaddr *addr, union ib_gid *gid) 169 { 170 switch (addr->sa_family) { 171 case AF_INET: 172 ipv6_addr_set_v4mapped(((struct sockaddr_in *) 173 addr)->sin_addr.s_addr, 174 (struct in6_addr *)gid); 175 break; 176 case AF_INET6: 177 *(struct in6_addr *)&gid->raw = 178 ((struct sockaddr_in6 *)addr)->sin6_addr; 179 break; 180 default: 181 return -EINVAL; 182 } 183 return 0; 184 } 185 186 /* Important - sockaddr should be a union of sockaddr_in and sockaddr_in6 */ 187 static inline void rdma_gid2ip(struct sockaddr *out, const union ib_gid *gid) 188 { 189 if (ipv6_addr_v4mapped((struct in6_addr *)gid)) { 190 struct sockaddr_in *out_in = (struct sockaddr_in *)out; 191 memset(out_in, 0, sizeof(*out_in)); 192 out_in->sin_family = AF_INET; 193 memcpy(&out_in->sin_addr.s_addr, gid->raw + 12, 4); 194 } else { 195 struct sockaddr_in6 *out_in = (struct sockaddr_in6 *)out; 196 memset(out_in, 0, sizeof(*out_in)); 197 out_in->sin6_family = AF_INET6; 198 memcpy(&out_in->sin6_addr.s6_addr, gid->raw, 16); 199 } 200 } 201 202 /* 203 * rdma_get/set_sgid/dgid() APIs are applicable to IB, and iWarp. 204 * They are not applicable to RoCE. 205 * RoCE GIDs are derived from the IP addresses. 206 */ 207 static inline void rdma_addr_get_sgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid) 208 { 209 memcpy(gid, dev_addr->src_dev_addr + rdma_addr_gid_offset(dev_addr), 210 sizeof(*gid)); 211 } 212 213 static inline void rdma_addr_set_sgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid) 214 { 215 memcpy(dev_addr->src_dev_addr + rdma_addr_gid_offset(dev_addr), gid, sizeof *gid); 216 } 217 218 static inline void rdma_addr_get_dgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid) 219 { 220 memcpy(gid, dev_addr->dst_dev_addr + rdma_addr_gid_offset(dev_addr), sizeof *gid); 221 } 222 223 static inline void rdma_addr_set_dgid(struct rdma_dev_addr *dev_addr, union ib_gid *gid) 224 { 225 memcpy(dev_addr->dst_dev_addr + rdma_addr_gid_offset(dev_addr), gid, sizeof *gid); 226 } 227 228 static inline enum ib_mtu iboe_get_mtu(int mtu) 229 { 230 /* 231 * Reduce IB headers from effective IBoE MTU. 232 */ 233 mtu = mtu - (IB_GRH_BYTES + IB_UDP_BYTES + IB_BTH_BYTES + 234 IB_EXT_XRC_BYTES + IB_EXT_ATOMICETH_BYTES + 235 IB_ICRC_BYTES); 236 237 if (mtu >= ib_mtu_enum_to_int(IB_MTU_4096)) 238 return IB_MTU_4096; 239 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_2048)) 240 return IB_MTU_2048; 241 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_1024)) 242 return IB_MTU_1024; 243 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_512)) 244 return IB_MTU_512; 245 else if (mtu >= ib_mtu_enum_to_int(IB_MTU_256)) 246 return IB_MTU_256; 247 else 248 return 0; 249 } 250 251 static inline int iboe_get_rate(struct net_device *dev) 252 { 253 struct ethtool_link_ksettings cmd; 254 int err; 255 256 rtnl_lock(); 257 err = __ethtool_get_link_ksettings(dev, &cmd); 258 rtnl_unlock(); 259 if (err) 260 return IB_RATE_PORT_CURRENT; 261 262 if (cmd.base.speed >= 40000) 263 return IB_RATE_40_GBPS; 264 else if (cmd.base.speed >= 30000) 265 return IB_RATE_30_GBPS; 266 else if (cmd.base.speed >= 20000) 267 return IB_RATE_20_GBPS; 268 else if (cmd.base.speed >= 10000) 269 return IB_RATE_10_GBPS; 270 else 271 return IB_RATE_PORT_CURRENT; 272 } 273 274 static inline int rdma_link_local_addr(struct in6_addr *addr) 275 { 276 if (addr->s6_addr32[0] == htonl(0xfe800000) && 277 addr->s6_addr32[1] == 0) 278 return 1; 279 280 return 0; 281 } 282 283 static inline void rdma_get_ll_mac(struct in6_addr *addr, u8 *mac) 284 { 285 memcpy(mac, &addr->s6_addr[8], 3); 286 memcpy(mac + 3, &addr->s6_addr[13], 3); 287 mac[0] ^= 2; 288 } 289 290 static inline int rdma_is_multicast_addr(struct in6_addr *addr) 291 { 292 __be32 ipv4_addr; 293 294 if (addr->s6_addr[0] == 0xff) 295 return 1; 296 297 ipv4_addr = addr->s6_addr32[3]; 298 return (ipv6_addr_v4mapped(addr) && ipv4_is_multicast(ipv4_addr)); 299 } 300 301 static inline void rdma_get_mcast_mac(struct in6_addr *addr, u8 *mac) 302 { 303 int i; 304 305 mac[0] = 0x33; 306 mac[1] = 0x33; 307 for (i = 2; i < 6; ++i) 308 mac[i] = addr->s6_addr[i + 10]; 309 } 310 311 static inline u16 rdma_get_vlan_id(union ib_gid *dgid) 312 { 313 u16 vid; 314 315 vid = dgid->raw[11] << 8 | dgid->raw[12]; 316 return vid < 0x1000 ? vid : 0xffff; 317 } 318 319 static inline struct net_device *rdma_vlan_dev_real_dev(const struct net_device *dev) 320 { 321 return is_vlan_dev(dev) ? vlan_dev_real_dev(dev) : NULL; 322 } 323 324 #endif /* IB_ADDR_H */ 325