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 rdma_ah_attr *attr, const 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 const struct ib_global_route *ib_grh; 85 union { 86 struct sockaddr _sockaddr; 87 struct sockaddr_in _sockaddr_in; 88 struct sockaddr_in6 _sockaddr_in6; 89 } sgid_addr, dgid_addr; 90 91 memset(ð, 0, sizeof(eth)); 92 memset(&grh, 0, sizeof(grh)); 93 94 /* Protocol Number */ 95 proto_num = ocrdma_hdr_type_to_proto_num(dev->id, ah->hdr_type); 96 if (!proto_num) 97 return -EINVAL; 98 nxthdr = (proto_num == ETH_P_IBOE) ? 0x1b : 0x11; 99 /* VLAN */ 100 if (!vlan_tag || (vlan_tag > 0xFFF)) 101 vlan_tag = dev->pvid; 102 if (vlan_tag || dev->pfc_state) { 103 if (!vlan_tag) { 104 pr_err("ocrdma%d:Using VLAN with PFC is recommended\n", 105 dev->id); 106 pr_err("ocrdma%d:Using VLAN 0 for this connection\n", 107 dev->id); 108 } 109 eth.eth_type = cpu_to_be16(0x8100); 110 eth.roce_eth_type = cpu_to_be16(proto_num); 111 vlan_tag |= (dev->sl & 0x07) << OCRDMA_VID_PCP_SHIFT; 112 eth.vlan_tag = cpu_to_be16(vlan_tag); 113 eth_sz = sizeof(struct ocrdma_eth_vlan); 114 *isvlan = true; 115 } else { 116 eth.eth_type = cpu_to_be16(proto_num); 117 eth_sz = sizeof(struct ocrdma_eth_basic); 118 } 119 /* MAC */ 120 memcpy(ð.smac[0], &dev->nic_info.mac_addr[0], ETH_ALEN); 121 status = ocrdma_resolve_dmac(dev, attr, ð.dmac[0]); 122 if (status) 123 return status; 124 ib_grh = rdma_ah_read_grh(attr); 125 ah->sgid_index = ib_grh->sgid_index; 126 /* Eth HDR */ 127 memcpy(&ah->av->eth_hdr, ð, eth_sz); 128 if (ah->hdr_type == RDMA_NETWORK_IPV4) { 129 *((__be16 *)&ipv4) = htons((4 << 12) | (5 << 8) | 130 ib_grh->traffic_class); 131 ipv4.id = cpu_to_be16(pdid); 132 ipv4.frag_off = htons(IP_DF); 133 ipv4.tot_len = htons(0); 134 ipv4.ttl = ib_grh->hop_limit; 135 ipv4.protocol = nxthdr; 136 rdma_gid2ip(&sgid_addr._sockaddr, sgid); 137 ipv4.saddr = sgid_addr._sockaddr_in.sin_addr.s_addr; 138 rdma_gid2ip(&dgid_addr._sockaddr, &ib_grh->dgid); 139 ipv4.daddr = dgid_addr._sockaddr_in.sin_addr.s_addr; 140 memcpy((u8 *)ah->av + eth_sz, &ipv4, sizeof(struct iphdr)); 141 } else { 142 memcpy(&grh.sgid[0], sgid->raw, sizeof(union ib_gid)); 143 grh.tclass_flow = cpu_to_be32((6 << 28) | 144 (ib_grh->traffic_class << 24) | 145 ib_grh->flow_label); 146 memcpy(&grh.dgid[0], ib_grh->dgid.raw, 147 sizeof(ib_grh->dgid.raw)); 148 grh.pdid_hoplimit = cpu_to_be32((pdid << 16) | 149 (nxthdr << 8) | 150 ib_grh->hop_limit); 151 memcpy((u8 *)ah->av + eth_sz, &grh, sizeof(struct ocrdma_grh)); 152 } 153 if (*isvlan) 154 ah->av->valid |= OCRDMA_AV_VLAN_VALID; 155 ah->av->valid = cpu_to_le32(ah->av->valid); 156 return status; 157 } 158 159 struct ib_ah *ocrdma_create_ah(struct ib_pd *ibpd, struct rdma_ah_attr *attr, 160 u32 flags, struct ib_udata *udata) 161 { 162 u32 *ahid_addr; 163 int status; 164 struct ocrdma_ah *ah; 165 bool isvlan = false; 166 u16 vlan_tag = 0xffff; 167 const struct ib_gid_attr *sgid_attr; 168 struct ocrdma_pd *pd = get_ocrdma_pd(ibpd); 169 struct ocrdma_dev *dev = get_ocrdma_dev(ibpd->device); 170 171 if ((attr->type != RDMA_AH_ATTR_TYPE_ROCE) || 172 !(rdma_ah_get_ah_flags(attr) & IB_AH_GRH)) 173 return ERR_PTR(-EINVAL); 174 175 if (atomic_cmpxchg(&dev->update_sl, 1, 0)) 176 ocrdma_init_service_level(dev); 177 178 ah = kzalloc(sizeof(*ah), GFP_ATOMIC); 179 if (!ah) 180 return ERR_PTR(-ENOMEM); 181 182 status = ocrdma_alloc_av(dev, ah); 183 if (status) 184 goto av_err; 185 186 sgid_attr = attr->grh.sgid_attr; 187 if (is_vlan_dev(sgid_attr->ndev)) 188 vlan_tag = vlan_dev_vlan_id(sgid_attr->ndev); 189 190 /* Get network header type for this GID */ 191 ah->hdr_type = rdma_gid_attr_network_type(sgid_attr); 192 193 status = set_av_attr(dev, ah, attr, &sgid_attr->gid, pd->id, 194 &isvlan, vlan_tag); 195 if (status) 196 goto av_conf_err; 197 198 /* if pd is for the user process, pass the ah_id to user space */ 199 if ((pd->uctx) && (pd->uctx->ah_tbl.va)) { 200 ahid_addr = pd->uctx->ah_tbl.va + rdma_ah_get_dlid(attr); 201 *ahid_addr = 0; 202 *ahid_addr |= ah->id & OCRDMA_AH_ID_MASK; 203 if (ocrdma_is_udp_encap_supported(dev)) { 204 *ahid_addr |= ((u32)ah->hdr_type & 205 OCRDMA_AH_L3_TYPE_MASK) << 206 OCRDMA_AH_L3_TYPE_SHIFT; 207 } 208 if (isvlan) 209 *ahid_addr |= (OCRDMA_AH_VLAN_VALID_MASK << 210 OCRDMA_AH_VLAN_VALID_SHIFT); 211 } 212 213 return &ah->ibah; 214 215 av_conf_err: 216 ocrdma_free_av(dev, ah); 217 av_err: 218 kfree(ah); 219 return ERR_PTR(status); 220 } 221 222 int ocrdma_destroy_ah(struct ib_ah *ibah, u32 flags) 223 { 224 struct ocrdma_ah *ah = get_ocrdma_ah(ibah); 225 struct ocrdma_dev *dev = get_ocrdma_dev(ibah->device); 226 227 ocrdma_free_av(dev, ah); 228 kfree(ah); 229 return 0; 230 } 231 232 int ocrdma_query_ah(struct ib_ah *ibah, struct rdma_ah_attr *attr) 233 { 234 struct ocrdma_ah *ah = get_ocrdma_ah(ibah); 235 struct ocrdma_av *av = ah->av; 236 struct ocrdma_grh *grh; 237 238 attr->type = ibah->type; 239 if (ah->av->valid & OCRDMA_AV_VALID) { 240 grh = (struct ocrdma_grh *)((u8 *)ah->av + 241 sizeof(struct ocrdma_eth_vlan)); 242 rdma_ah_set_sl(attr, be16_to_cpu(av->eth_hdr.vlan_tag) >> 13); 243 } else { 244 grh = (struct ocrdma_grh *)((u8 *)ah->av + 245 sizeof(struct ocrdma_eth_basic)); 246 rdma_ah_set_sl(attr, 0); 247 } 248 rdma_ah_set_grh(attr, NULL, 249 be32_to_cpu(grh->tclass_flow) & 0xffffffff, 250 ah->sgid_index, 251 be32_to_cpu(grh->pdid_hoplimit) & 0xff, 252 be32_to_cpu(grh->tclass_flow) >> 24); 253 rdma_ah_set_dgid_raw(attr, &grh->dgid[0]); 254 return 0; 255 } 256 257 int ocrdma_process_mad(struct ib_device *ibdev, 258 int process_mad_flags, 259 u8 port_num, 260 const struct ib_wc *in_wc, 261 const struct ib_grh *in_grh, 262 const struct ib_mad_hdr *in, size_t in_mad_size, 263 struct ib_mad_hdr *out, size_t *out_mad_size, 264 u16 *out_mad_pkey_index) 265 { 266 int status; 267 struct ocrdma_dev *dev; 268 const struct ib_mad *in_mad = (const struct ib_mad *)in; 269 struct ib_mad *out_mad = (struct ib_mad *)out; 270 271 if (WARN_ON_ONCE(in_mad_size != sizeof(*in_mad) || 272 *out_mad_size != sizeof(*out_mad))) 273 return IB_MAD_RESULT_FAILURE; 274 275 switch (in_mad->mad_hdr.mgmt_class) { 276 case IB_MGMT_CLASS_PERF_MGMT: 277 dev = get_ocrdma_dev(ibdev); 278 if (!ocrdma_pma_counters(dev, out_mad)) 279 status = IB_MAD_RESULT_SUCCESS | IB_MAD_RESULT_REPLY; 280 else 281 status = IB_MAD_RESULT_SUCCESS; 282 break; 283 default: 284 status = IB_MAD_RESULT_SUCCESS; 285 break; 286 } 287 return status; 288 } 289