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_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 ib_grh = rdma_ah_read_grh(attr); 124 ah->sgid_index = ib_grh->sgid_index; 125 /* Eth HDR */ 126 memcpy(&ah->av->eth_hdr, ð, eth_sz); 127 if (ah->hdr_type == RDMA_NETWORK_IPV4) { 128 *((__be16 *)&ipv4) = htons((4 << 12) | (5 << 8) | 129 ib_grh->traffic_class); 130 ipv4.id = cpu_to_be16(pdid); 131 ipv4.frag_off = htons(IP_DF); 132 ipv4.tot_len = htons(0); 133 ipv4.ttl = ib_grh->hop_limit; 134 ipv4.protocol = nxthdr; 135 rdma_gid2ip((struct sockaddr *)&sgid_addr, sgid); 136 ipv4.saddr = sgid_addr._sockaddr_in.sin_addr.s_addr; 137 rdma_gid2ip((struct sockaddr*)&dgid_addr, &ib_grh->dgid); 138 ipv4.daddr = dgid_addr._sockaddr_in.sin_addr.s_addr; 139 memcpy((u8 *)ah->av + eth_sz, &ipv4, sizeof(struct iphdr)); 140 } else { 141 memcpy(&grh.sgid[0], sgid->raw, sizeof(union ib_gid)); 142 grh.tclass_flow = cpu_to_be32((6 << 28) | 143 (ib_grh->traffic_class << 24) | 144 ib_grh->flow_label); 145 memcpy(&grh.dgid[0], ib_grh->dgid.raw, 146 sizeof(ib_grh->dgid.raw)); 147 grh.pdid_hoplimit = cpu_to_be32((pdid << 16) | 148 (nxthdr << 8) | 149 ib_grh->hop_limit); 150 memcpy((u8 *)ah->av + eth_sz, &grh, sizeof(struct ocrdma_grh)); 151 } 152 if (*isvlan) 153 ah->av->valid |= OCRDMA_AV_VLAN_VALID; 154 ah->av->valid = cpu_to_le32(ah->av->valid); 155 return status; 156 } 157 158 int ocrdma_create_ah(struct ib_ah *ibah, struct rdma_ah_attr *attr, u32 flags, 159 struct ib_udata *udata) 160 { 161 u32 *ahid_addr; 162 int status; 163 struct ocrdma_ah *ah = get_ocrdma_ah(ibah); 164 bool isvlan = false; 165 u16 vlan_tag = 0xffff; 166 const struct ib_gid_attr *sgid_attr; 167 struct ocrdma_pd *pd = get_ocrdma_pd(ibah->pd); 168 struct ocrdma_dev *dev = get_ocrdma_dev(ibah->device); 169 170 if ((attr->type != RDMA_AH_ATTR_TYPE_ROCE) || 171 !(rdma_ah_get_ah_flags(attr) & IB_AH_GRH)) 172 return -EINVAL; 173 174 if (atomic_cmpxchg(&dev->update_sl, 1, 0)) 175 ocrdma_init_service_level(dev); 176 177 sgid_attr = attr->grh.sgid_attr; 178 status = rdma_read_gid_l2_fields(sgid_attr, &vlan_tag, NULL); 179 if (status) 180 return status; 181 182 status = ocrdma_alloc_av(dev, ah); 183 if (status) 184 goto av_err; 185 186 /* Get network header type for this GID */ 187 ah->hdr_type = rdma_gid_attr_network_type(sgid_attr); 188 189 status = set_av_attr(dev, ah, attr, &sgid_attr->gid, pd->id, 190 &isvlan, vlan_tag); 191 if (status) 192 goto av_conf_err; 193 194 /* if pd is for the user process, pass the ah_id to user space */ 195 if ((pd->uctx) && (pd->uctx->ah_tbl.va)) { 196 ahid_addr = pd->uctx->ah_tbl.va + rdma_ah_get_dlid(attr); 197 *ahid_addr = 0; 198 *ahid_addr |= ah->id & OCRDMA_AH_ID_MASK; 199 if (ocrdma_is_udp_encap_supported(dev)) { 200 *ahid_addr |= ((u32)ah->hdr_type & 201 OCRDMA_AH_L3_TYPE_MASK) << 202 OCRDMA_AH_L3_TYPE_SHIFT; 203 } 204 if (isvlan) 205 *ahid_addr |= (OCRDMA_AH_VLAN_VALID_MASK << 206 OCRDMA_AH_VLAN_VALID_SHIFT); 207 } 208 209 return 0; 210 211 av_conf_err: 212 ocrdma_free_av(dev, ah); 213 av_err: 214 return status; 215 } 216 217 void ocrdma_destroy_ah(struct ib_ah *ibah, u32 flags) 218 { 219 struct ocrdma_ah *ah = get_ocrdma_ah(ibah); 220 struct ocrdma_dev *dev = get_ocrdma_dev(ibah->device); 221 222 ocrdma_free_av(dev, ah); 223 } 224 225 int ocrdma_query_ah(struct ib_ah *ibah, struct rdma_ah_attr *attr) 226 { 227 struct ocrdma_ah *ah = get_ocrdma_ah(ibah); 228 struct ocrdma_av *av = ah->av; 229 struct ocrdma_grh *grh; 230 231 attr->type = ibah->type; 232 if (ah->av->valid & OCRDMA_AV_VALID) { 233 grh = (struct ocrdma_grh *)((u8 *)ah->av + 234 sizeof(struct ocrdma_eth_vlan)); 235 rdma_ah_set_sl(attr, be16_to_cpu(av->eth_hdr.vlan_tag) >> 13); 236 } else { 237 grh = (struct ocrdma_grh *)((u8 *)ah->av + 238 sizeof(struct ocrdma_eth_basic)); 239 rdma_ah_set_sl(attr, 0); 240 } 241 rdma_ah_set_grh(attr, NULL, 242 be32_to_cpu(grh->tclass_flow) & 0xffffffff, 243 ah->sgid_index, 244 be32_to_cpu(grh->pdid_hoplimit) & 0xff, 245 be32_to_cpu(grh->tclass_flow) >> 24); 246 rdma_ah_set_dgid_raw(attr, &grh->dgid[0]); 247 return 0; 248 } 249 250 int ocrdma_process_mad(struct ib_device *ibdev, int process_mad_flags, 251 u8 port_num, const struct ib_wc *in_wc, 252 const struct ib_grh *in_grh, const struct ib_mad *in, 253 struct ib_mad *out, size_t *out_mad_size, 254 u16 *out_mad_pkey_index) 255 { 256 int status = IB_MAD_RESULT_SUCCESS; 257 struct ocrdma_dev *dev; 258 259 if (in->mad_hdr.mgmt_class == IB_MGMT_CLASS_PERF_MGMT) { 260 dev = get_ocrdma_dev(ibdev); 261 ocrdma_pma_counters(dev, out); 262 status |= IB_MAD_RESULT_REPLY; 263 } 264 265 return status; 266 } 267