1 /* 2 * Copyright (c) 2016 Mellanox Technologies Ltd. All rights reserved. 3 * Copyright (c) 2015 System Fabric Works, Inc. 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 #include <rdma/rdma_netlink.h> 35 #include <net/addrconf.h> 36 #include "rxe.h" 37 #include "rxe_loc.h" 38 39 MODULE_AUTHOR("Bob Pearson, Frank Zago, John Groves, Kamal Heib"); 40 MODULE_DESCRIPTION("Soft RDMA transport"); 41 MODULE_LICENSE("Dual BSD/GPL"); 42 43 /* free resources for a rxe device all objects created for this device must 44 * have been destroyed 45 */ 46 void rxe_dealloc(struct ib_device *ib_dev) 47 { 48 struct rxe_dev *rxe = container_of(ib_dev, struct rxe_dev, ib_dev); 49 50 rxe_pool_cleanup(&rxe->uc_pool); 51 rxe_pool_cleanup(&rxe->pd_pool); 52 rxe_pool_cleanup(&rxe->ah_pool); 53 rxe_pool_cleanup(&rxe->srq_pool); 54 rxe_pool_cleanup(&rxe->qp_pool); 55 rxe_pool_cleanup(&rxe->cq_pool); 56 rxe_pool_cleanup(&rxe->mr_pool); 57 rxe_pool_cleanup(&rxe->mw_pool); 58 rxe_pool_cleanup(&rxe->mc_grp_pool); 59 rxe_pool_cleanup(&rxe->mc_elem_pool); 60 61 if (rxe->tfm) 62 crypto_free_shash(rxe->tfm); 63 } 64 65 /* initialize rxe device parameters */ 66 static void rxe_init_device_param(struct rxe_dev *rxe) 67 { 68 rxe->max_inline_data = RXE_MAX_INLINE_DATA; 69 70 rxe->attr.vendor_id = RXE_VENDOR_ID; 71 rxe->attr.max_mr_size = RXE_MAX_MR_SIZE; 72 rxe->attr.page_size_cap = RXE_PAGE_SIZE_CAP; 73 rxe->attr.max_qp = RXE_MAX_QP; 74 rxe->attr.max_qp_wr = RXE_MAX_QP_WR; 75 rxe->attr.device_cap_flags = RXE_DEVICE_CAP_FLAGS; 76 rxe->attr.max_send_sge = RXE_MAX_SGE; 77 rxe->attr.max_recv_sge = RXE_MAX_SGE; 78 rxe->attr.max_sge_rd = RXE_MAX_SGE_RD; 79 rxe->attr.max_cq = RXE_MAX_CQ; 80 rxe->attr.max_cqe = (1 << RXE_MAX_LOG_CQE) - 1; 81 rxe->attr.max_mr = RXE_MAX_MR; 82 rxe->attr.max_pd = RXE_MAX_PD; 83 rxe->attr.max_qp_rd_atom = RXE_MAX_QP_RD_ATOM; 84 rxe->attr.max_res_rd_atom = RXE_MAX_RES_RD_ATOM; 85 rxe->attr.max_qp_init_rd_atom = RXE_MAX_QP_INIT_RD_ATOM; 86 rxe->attr.atomic_cap = IB_ATOMIC_HCA; 87 rxe->attr.max_mcast_grp = RXE_MAX_MCAST_GRP; 88 rxe->attr.max_mcast_qp_attach = RXE_MAX_MCAST_QP_ATTACH; 89 rxe->attr.max_total_mcast_qp_attach = RXE_MAX_TOT_MCAST_QP_ATTACH; 90 rxe->attr.max_ah = RXE_MAX_AH; 91 rxe->attr.max_srq = RXE_MAX_SRQ; 92 rxe->attr.max_srq_wr = RXE_MAX_SRQ_WR; 93 rxe->attr.max_srq_sge = RXE_MAX_SRQ_SGE; 94 rxe->attr.max_fast_reg_page_list_len = RXE_MAX_FMR_PAGE_LIST_LEN; 95 rxe->attr.max_pkeys = RXE_MAX_PKEYS; 96 rxe->attr.local_ca_ack_delay = RXE_LOCAL_CA_ACK_DELAY; 97 addrconf_addr_eui48((unsigned char *)&rxe->attr.sys_image_guid, 98 rxe->ndev->dev_addr); 99 100 rxe->max_ucontext = RXE_MAX_UCONTEXT; 101 } 102 103 /* initialize port attributes */ 104 static void rxe_init_port_param(struct rxe_port *port) 105 { 106 port->attr.state = IB_PORT_DOWN; 107 port->attr.max_mtu = IB_MTU_4096; 108 port->attr.active_mtu = IB_MTU_256; 109 port->attr.gid_tbl_len = RXE_PORT_GID_TBL_LEN; 110 port->attr.port_cap_flags = RXE_PORT_PORT_CAP_FLAGS; 111 port->attr.max_msg_sz = RXE_PORT_MAX_MSG_SZ; 112 port->attr.bad_pkey_cntr = RXE_PORT_BAD_PKEY_CNTR; 113 port->attr.qkey_viol_cntr = RXE_PORT_QKEY_VIOL_CNTR; 114 port->attr.pkey_tbl_len = RXE_PORT_PKEY_TBL_LEN; 115 port->attr.lid = RXE_PORT_LID; 116 port->attr.sm_lid = RXE_PORT_SM_LID; 117 port->attr.lmc = RXE_PORT_LMC; 118 port->attr.max_vl_num = RXE_PORT_MAX_VL_NUM; 119 port->attr.sm_sl = RXE_PORT_SM_SL; 120 port->attr.subnet_timeout = RXE_PORT_SUBNET_TIMEOUT; 121 port->attr.init_type_reply = RXE_PORT_INIT_TYPE_REPLY; 122 port->attr.active_width = RXE_PORT_ACTIVE_WIDTH; 123 port->attr.active_speed = RXE_PORT_ACTIVE_SPEED; 124 port->attr.phys_state = RXE_PORT_PHYS_STATE; 125 port->mtu_cap = ib_mtu_enum_to_int(IB_MTU_256); 126 port->subnet_prefix = cpu_to_be64(RXE_PORT_SUBNET_PREFIX); 127 } 128 129 /* initialize port state, note IB convention that HCA ports are always 130 * numbered from 1 131 */ 132 static void rxe_init_ports(struct rxe_dev *rxe) 133 { 134 struct rxe_port *port = &rxe->port; 135 136 rxe_init_port_param(port); 137 addrconf_addr_eui48((unsigned char *)&port->port_guid, 138 rxe->ndev->dev_addr); 139 spin_lock_init(&port->port_lock); 140 } 141 142 /* init pools of managed objects */ 143 static int rxe_init_pools(struct rxe_dev *rxe) 144 { 145 int err; 146 147 err = rxe_pool_init(rxe, &rxe->uc_pool, RXE_TYPE_UC, 148 rxe->max_ucontext); 149 if (err) 150 goto err1; 151 152 err = rxe_pool_init(rxe, &rxe->pd_pool, RXE_TYPE_PD, 153 rxe->attr.max_pd); 154 if (err) 155 goto err2; 156 157 err = rxe_pool_init(rxe, &rxe->ah_pool, RXE_TYPE_AH, 158 rxe->attr.max_ah); 159 if (err) 160 goto err3; 161 162 err = rxe_pool_init(rxe, &rxe->srq_pool, RXE_TYPE_SRQ, 163 rxe->attr.max_srq); 164 if (err) 165 goto err4; 166 167 err = rxe_pool_init(rxe, &rxe->qp_pool, RXE_TYPE_QP, 168 rxe->attr.max_qp); 169 if (err) 170 goto err5; 171 172 err = rxe_pool_init(rxe, &rxe->cq_pool, RXE_TYPE_CQ, 173 rxe->attr.max_cq); 174 if (err) 175 goto err6; 176 177 err = rxe_pool_init(rxe, &rxe->mr_pool, RXE_TYPE_MR, 178 rxe->attr.max_mr); 179 if (err) 180 goto err7; 181 182 err = rxe_pool_init(rxe, &rxe->mw_pool, RXE_TYPE_MW, 183 rxe->attr.max_mw); 184 if (err) 185 goto err8; 186 187 err = rxe_pool_init(rxe, &rxe->mc_grp_pool, RXE_TYPE_MC_GRP, 188 rxe->attr.max_mcast_grp); 189 if (err) 190 goto err9; 191 192 err = rxe_pool_init(rxe, &rxe->mc_elem_pool, RXE_TYPE_MC_ELEM, 193 rxe->attr.max_total_mcast_qp_attach); 194 if (err) 195 goto err10; 196 197 return 0; 198 199 err10: 200 rxe_pool_cleanup(&rxe->mc_grp_pool); 201 err9: 202 rxe_pool_cleanup(&rxe->mw_pool); 203 err8: 204 rxe_pool_cleanup(&rxe->mr_pool); 205 err7: 206 rxe_pool_cleanup(&rxe->cq_pool); 207 err6: 208 rxe_pool_cleanup(&rxe->qp_pool); 209 err5: 210 rxe_pool_cleanup(&rxe->srq_pool); 211 err4: 212 rxe_pool_cleanup(&rxe->ah_pool); 213 err3: 214 rxe_pool_cleanup(&rxe->pd_pool); 215 err2: 216 rxe_pool_cleanup(&rxe->uc_pool); 217 err1: 218 return err; 219 } 220 221 /* initialize rxe device state */ 222 static int rxe_init(struct rxe_dev *rxe) 223 { 224 int err; 225 226 /* init default device parameters */ 227 rxe_init_device_param(rxe); 228 229 rxe_init_ports(rxe); 230 231 err = rxe_init_pools(rxe); 232 if (err) 233 return err; 234 235 /* init pending mmap list */ 236 spin_lock_init(&rxe->mmap_offset_lock); 237 spin_lock_init(&rxe->pending_lock); 238 INIT_LIST_HEAD(&rxe->pending_mmaps); 239 240 mutex_init(&rxe->usdev_lock); 241 242 return 0; 243 } 244 245 void rxe_set_mtu(struct rxe_dev *rxe, unsigned int ndev_mtu) 246 { 247 struct rxe_port *port = &rxe->port; 248 enum ib_mtu mtu; 249 250 mtu = eth_mtu_int_to_enum(ndev_mtu); 251 252 /* Make sure that new MTU in range */ 253 mtu = mtu ? min_t(enum ib_mtu, mtu, IB_MTU_4096) : IB_MTU_256; 254 255 port->attr.active_mtu = mtu; 256 port->mtu_cap = ib_mtu_enum_to_int(mtu); 257 } 258 259 /* called by ifc layer to create new rxe device. 260 * The caller should allocate memory for rxe by calling ib_alloc_device. 261 */ 262 int rxe_add(struct rxe_dev *rxe, unsigned int mtu, const char *ibdev_name) 263 { 264 int err; 265 266 err = rxe_init(rxe); 267 if (err) 268 return err; 269 270 rxe_set_mtu(rxe, mtu); 271 272 return rxe_register_device(rxe, ibdev_name); 273 } 274 275 static int rxe_newlink(const char *ibdev_name, struct net_device *ndev) 276 { 277 struct rxe_dev *exists; 278 int err = 0; 279 280 exists = rxe_get_dev_from_net(ndev); 281 if (exists) { 282 ib_device_put(&exists->ib_dev); 283 pr_err("already configured on %s\n", ndev->name); 284 err = -EEXIST; 285 goto err; 286 } 287 288 err = rxe_net_add(ibdev_name, ndev); 289 if (err) { 290 pr_err("failed to add %s\n", ndev->name); 291 goto err; 292 } 293 err: 294 return err; 295 } 296 297 static struct rdma_link_ops rxe_link_ops = { 298 .type = "rxe", 299 .newlink = rxe_newlink, 300 }; 301 302 static int __init rxe_module_init(void) 303 { 304 int err; 305 306 /* initialize slab caches for managed objects */ 307 err = rxe_cache_init(); 308 if (err) { 309 pr_err("unable to init object pools\n"); 310 return err; 311 } 312 313 err = rxe_net_init(); 314 if (err) 315 return err; 316 317 rdma_link_register(&rxe_link_ops); 318 pr_info("loaded\n"); 319 return 0; 320 } 321 322 static void __exit rxe_module_exit(void) 323 { 324 rdma_link_unregister(&rxe_link_ops); 325 ib_unregister_driver(RDMA_DRIVER_RXE); 326 rxe_net_exit(); 327 rxe_cache_exit(); 328 329 pr_info("unloaded\n"); 330 } 331 332 late_initcall(rxe_module_init); 333 module_exit(rxe_module_exit); 334 335 MODULE_ALIAS_RDMA_LINK("rxe"); 336