1 /* This file is part of the Emulex RoCE Device Driver for 2 * RoCE (RDMA over Converged Ethernet) adapters. 3 * Copyright (C) 2012-2015 Emulex. All rights reserved. 4 * EMULEX and SLI are trademarks of Emulex. 5 * www.emulex.com 6 * 7 * This software is available to you under a choice of one of two licenses. 8 * You may choose to be licensed under the terms of the GNU General Public 9 * License (GPL) Version 2, available from the file COPYING in the main 10 * directory of this source tree, or the BSD license below: 11 * 12 * Redistribution and use in source and binary forms, with or without 13 * modification, are permitted provided that the following conditions 14 * are met: 15 * 16 * - Redistributions of source code must retain the above copyright notice, 17 * this list of conditions and the following disclaimer. 18 * 19 * - Redistributions in binary form must reproduce the above copyright 20 * notice, this list of conditions and the following disclaimer in 21 * the documentation and/or other materials provided with the distribution. 22 * 23 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 24 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,THE 25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 26 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE 27 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 28 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 29 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR 30 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, 31 * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR 32 * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF 33 * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 34 * 35 * Contact Information: 36 * linux-drivers@emulex.com 37 * 38 * Emulex 39 * 3333 Susan Street 40 * Costa Mesa, CA 92626 41 */ 42 43 #include <net/neighbour.h> 44 #include <net/netevent.h> 45 46 #include <rdma/ib_addr.h> 47 #include <rdma/ib_mad.h> 48 #include <rdma/ib_cache.h> 49 50 #include "ocrdma.h" 51 #include "ocrdma_verbs.h" 52 #include "ocrdma_ah.h" 53 #include "ocrdma_hw.h" 54 #include "ocrdma_stats.h" 55 56 #define OCRDMA_VID_PCP_SHIFT 0xD 57 58 static u16 ocrdma_hdr_type_to_proto_num(int devid, u8 hdr_type) 59 { 60 switch (hdr_type) { 61 case OCRDMA_L3_TYPE_IB_GRH: 62 return (u16)ETH_P_IBOE; 63 case OCRDMA_L3_TYPE_IPV4: 64 return (u16)0x0800; 65 case OCRDMA_L3_TYPE_IPV6: 66 return (u16)0x86dd; 67 default: 68 pr_err("ocrdma%d: Invalid network header\n", devid); 69 return 0; 70 } 71 } 72 73 static inline int set_av_attr(struct ocrdma_dev *dev, struct ocrdma_ah *ah, 74 struct ib_ah_attr *attr, union ib_gid *sgid, 75 int pdid, bool *isvlan, u16 vlan_tag) 76 { 77 int status; 78 struct ocrdma_eth_vlan eth; 79 struct ocrdma_grh grh; 80 int eth_sz; 81 u16 proto_num = 0; 82 u8 nxthdr = 0x11; 83 struct iphdr ipv4; 84 union { 85 struct sockaddr _sockaddr; 86 struct sockaddr_in _sockaddr_in; 87 struct sockaddr_in6 _sockaddr_in6; 88 } sgid_addr, dgid_addr; 89 90 memset(ð, 0, sizeof(eth)); 91 memset(&grh, 0, sizeof(grh)); 92 93 /* Protocol Number */ 94 proto_num = ocrdma_hdr_type_to_proto_num(dev->id, ah->hdr_type); 95 if (!proto_num) 96 return -EINVAL; 97 nxthdr = (proto_num == ETH_P_IBOE) ? 0x1b : 0x11; 98 /* VLAN */ 99 if (!vlan_tag || (vlan_tag > 0xFFF)) 100 vlan_tag = dev->pvid; 101 if (vlan_tag || dev->pfc_state) { 102 if (!vlan_tag) { 103 pr_err("ocrdma%d:Using VLAN with PFC is recommended\n", 104 dev->id); 105 pr_err("ocrdma%d:Using VLAN 0 for this connection\n", 106 dev->id); 107 } 108 eth.eth_type = cpu_to_be16(0x8100); 109 eth.roce_eth_type = cpu_to_be16(proto_num); 110 vlan_tag |= (dev->sl & 0x07) << OCRDMA_VID_PCP_SHIFT; 111 eth.vlan_tag = cpu_to_be16(vlan_tag); 112 eth_sz = sizeof(struct ocrdma_eth_vlan); 113 *isvlan = true; 114 } else { 115 eth.eth_type = cpu_to_be16(proto_num); 116 eth_sz = sizeof(struct ocrdma_eth_basic); 117 } 118 /* MAC */ 119 memcpy(ð.smac[0], &dev->nic_info.mac_addr[0], ETH_ALEN); 120 status = ocrdma_resolve_dmac(dev, attr, ð.dmac[0]); 121 if (status) 122 return status; 123 ah->sgid_index = attr->grh.sgid_index; 124 /* Eth HDR */ 125 memcpy(&ah->av->eth_hdr, ð, eth_sz); 126 if (ah->hdr_type == RDMA_NETWORK_IPV4) { 127 *((__be16 *)&ipv4) = htons((4 << 12) | (5 << 8) | 128 attr->grh.traffic_class); 129 ipv4.id = cpu_to_be16(pdid); 130 ipv4.frag_off = htons(IP_DF); 131 ipv4.tot_len = htons(0); 132 ipv4.ttl = attr->grh.hop_limit; 133 ipv4.protocol = nxthdr; 134 rdma_gid2ip(&sgid_addr._sockaddr, sgid); 135 ipv4.saddr = sgid_addr._sockaddr_in.sin_addr.s_addr; 136 rdma_gid2ip(&dgid_addr._sockaddr, &attr->grh.dgid); 137 ipv4.daddr = dgid_addr._sockaddr_in.sin_addr.s_addr; 138 memcpy((u8 *)ah->av + eth_sz, &ipv4, sizeof(struct iphdr)); 139 } else { 140 memcpy(&grh.sgid[0], sgid->raw, sizeof(union ib_gid)); 141 grh.tclass_flow = cpu_to_be32((6 << 28) | 142 (attr->grh.traffic_class << 24) | 143 attr->grh.flow_label); 144 memcpy(&grh.dgid[0], attr->grh.dgid.raw, 145 sizeof(attr->grh.dgid.raw)); 146 grh.pdid_hoplimit = cpu_to_be32((pdid << 16) | 147 (nxthdr << 8) | 148 attr->grh.hop_limit); 149 memcpy((u8 *)ah->av + eth_sz, &grh, sizeof(struct ocrdma_grh)); 150 } 151 if (*isvlan) 152 ah->av->valid |= OCRDMA_AV_VLAN_VALID; 153 ah->av->valid = cpu_to_le32(ah->av->valid); 154 return status; 155 } 156 157 struct ib_ah *ocrdma_create_ah(struct ib_pd *ibpd, struct ib_ah_attr *attr, 158 struct ib_udata *udata) 159 { 160 u32 *ahid_addr; 161 int status; 162 struct ocrdma_ah *ah; 163 bool isvlan = false; 164 u16 vlan_tag = 0xffff; 165 struct ib_gid_attr sgid_attr; 166 struct ocrdma_pd *pd = get_ocrdma_pd(ibpd); 167 struct ocrdma_dev *dev = get_ocrdma_dev(ibpd->device); 168 union ib_gid sgid; 169 170 if (!(attr->ah_flags & IB_AH_GRH)) 171 return ERR_PTR(-EINVAL); 172 173 if (atomic_cmpxchg(&dev->update_sl, 1, 0)) 174 ocrdma_init_service_level(dev); 175 176 ah = kzalloc(sizeof(*ah), GFP_ATOMIC); 177 if (!ah) 178 return ERR_PTR(-ENOMEM); 179 180 status = ocrdma_alloc_av(dev, ah); 181 if (status) 182 goto av_err; 183 184 status = ib_get_cached_gid(&dev->ibdev, 1, attr->grh.sgid_index, &sgid, 185 &sgid_attr); 186 if (status) { 187 pr_err("%s(): Failed to query sgid, status = %d\n", 188 __func__, status); 189 goto av_conf_err; 190 } 191 if (sgid_attr.ndev) { 192 if (is_vlan_dev(sgid_attr.ndev)) 193 vlan_tag = vlan_dev_vlan_id(sgid_attr.ndev); 194 dev_put(sgid_attr.ndev); 195 } 196 /* Get network header type for this GID */ 197 ah->hdr_type = ib_gid_to_network_type(sgid_attr.gid_type, &sgid); 198 199 if ((pd->uctx) && 200 (!rdma_is_multicast_addr((struct in6_addr *)attr->grh.dgid.raw)) && 201 (!rdma_link_local_addr((struct in6_addr *)attr->grh.dgid.raw))) { 202 status = rdma_addr_find_l2_eth_by_grh(&sgid, &attr->grh.dgid, 203 attr->dmac, &vlan_tag, 204 &sgid_attr.ndev->ifindex, 205 NULL); 206 if (status) { 207 pr_err("%s(): Failed to resolve dmac from gid." 208 "status = %d\n", __func__, status); 209 goto av_conf_err; 210 } 211 } 212 213 status = set_av_attr(dev, ah, attr, &sgid, pd->id, &isvlan, vlan_tag); 214 if (status) 215 goto av_conf_err; 216 217 /* if pd is for the user process, pass the ah_id to user space */ 218 if ((pd->uctx) && (pd->uctx->ah_tbl.va)) { 219 ahid_addr = pd->uctx->ah_tbl.va + attr->dlid; 220 *ahid_addr = 0; 221 *ahid_addr |= ah->id & OCRDMA_AH_ID_MASK; 222 if (ocrdma_is_udp_encap_supported(dev)) { 223 *ahid_addr |= ((u32)ah->hdr_type & 224 OCRDMA_AH_L3_TYPE_MASK) << 225 OCRDMA_AH_L3_TYPE_SHIFT; 226 } 227 if (isvlan) 228 *ahid_addr |= (OCRDMA_AH_VLAN_VALID_MASK << 229 OCRDMA_AH_VLAN_VALID_SHIFT); 230 } 231 232 return &ah->ibah; 233 234 av_conf_err: 235 ocrdma_free_av(dev, ah); 236 av_err: 237 kfree(ah); 238 return ERR_PTR(status); 239 } 240 241 int ocrdma_destroy_ah(struct ib_ah *ibah) 242 { 243 struct ocrdma_ah *ah = get_ocrdma_ah(ibah); 244 struct ocrdma_dev *dev = get_ocrdma_dev(ibah->device); 245 246 ocrdma_free_av(dev, ah); 247 kfree(ah); 248 return 0; 249 } 250 251 int ocrdma_query_ah(struct ib_ah *ibah, struct ib_ah_attr *attr) 252 { 253 struct ocrdma_ah *ah = get_ocrdma_ah(ibah); 254 struct ocrdma_av *av = ah->av; 255 struct ocrdma_grh *grh; 256 attr->ah_flags |= IB_AH_GRH; 257 if (ah->av->valid & OCRDMA_AV_VALID) { 258 grh = (struct ocrdma_grh *)((u8 *)ah->av + 259 sizeof(struct ocrdma_eth_vlan)); 260 attr->sl = be16_to_cpu(av->eth_hdr.vlan_tag) >> 13; 261 } else { 262 grh = (struct ocrdma_grh *)((u8 *)ah->av + 263 sizeof(struct ocrdma_eth_basic)); 264 attr->sl = 0; 265 } 266 memcpy(&attr->grh.dgid.raw[0], &grh->dgid[0], sizeof(grh->dgid)); 267 attr->grh.sgid_index = ah->sgid_index; 268 attr->grh.hop_limit = be32_to_cpu(grh->pdid_hoplimit) & 0xff; 269 attr->grh.traffic_class = be32_to_cpu(grh->tclass_flow) >> 24; 270 attr->grh.flow_label = be32_to_cpu(grh->tclass_flow) & 0x00ffffffff; 271 return 0; 272 } 273 274 int ocrdma_modify_ah(struct ib_ah *ibah, struct ib_ah_attr *attr) 275 { 276 /* modify_ah is unsupported */ 277 return -ENOSYS; 278 } 279 280 int ocrdma_process_mad(struct ib_device *ibdev, 281 int process_mad_flags, 282 u8 port_num, 283 const struct ib_wc *in_wc, 284 const struct ib_grh *in_grh, 285 const struct ib_mad_hdr *in, size_t in_mad_size, 286 struct ib_mad_hdr *out, size_t *out_mad_size, 287 u16 *out_mad_pkey_index) 288 { 289 int status; 290 struct ocrdma_dev *dev; 291 const struct ib_mad *in_mad = (const struct ib_mad *)in; 292 struct ib_mad *out_mad = (struct ib_mad *)out; 293 294 if (WARN_ON_ONCE(in_mad_size != sizeof(*in_mad) || 295 *out_mad_size != sizeof(*out_mad))) 296 return IB_MAD_RESULT_FAILURE; 297 298 switch (in_mad->mad_hdr.mgmt_class) { 299 case IB_MGMT_CLASS_PERF_MGMT: 300 dev = get_ocrdma_dev(ibdev); 301 if (!ocrdma_pma_counters(dev, out_mad)) 302 status = IB_MAD_RESULT_SUCCESS | IB_MAD_RESULT_REPLY; 303 else 304 status = IB_MAD_RESULT_SUCCESS; 305 break; 306 default: 307 status = IB_MAD_RESULT_SUCCESS; 308 break; 309 } 310 return status; 311 } 312