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 all ports on a device */ 44 static void rxe_cleanup_ports(struct rxe_dev *rxe) 45 { 46 kfree(rxe->port.pkey_tbl); 47 rxe->port.pkey_tbl = NULL; 48 49 } 50 51 /* free resources for a rxe device all objects created for this device must 52 * have been destroyed 53 */ 54 void rxe_dealloc(struct ib_device *ib_dev) 55 { 56 struct rxe_dev *rxe = container_of(ib_dev, struct rxe_dev, ib_dev); 57 58 rxe_pool_cleanup(&rxe->uc_pool); 59 rxe_pool_cleanup(&rxe->pd_pool); 60 rxe_pool_cleanup(&rxe->ah_pool); 61 rxe_pool_cleanup(&rxe->srq_pool); 62 rxe_pool_cleanup(&rxe->qp_pool); 63 rxe_pool_cleanup(&rxe->cq_pool); 64 rxe_pool_cleanup(&rxe->mr_pool); 65 rxe_pool_cleanup(&rxe->mw_pool); 66 rxe_pool_cleanup(&rxe->mc_grp_pool); 67 rxe_pool_cleanup(&rxe->mc_elem_pool); 68 69 rxe_cleanup_ports(rxe); 70 71 if (rxe->tfm) 72 crypto_free_shash(rxe->tfm); 73 } 74 75 /* initialize rxe device parameters */ 76 static void rxe_init_device_param(struct rxe_dev *rxe) 77 { 78 rxe->max_inline_data = RXE_MAX_INLINE_DATA; 79 80 rxe->attr.vendor_id = RXE_VENDOR_ID; 81 rxe->attr.max_mr_size = RXE_MAX_MR_SIZE; 82 rxe->attr.page_size_cap = RXE_PAGE_SIZE_CAP; 83 rxe->attr.max_qp = RXE_MAX_QP; 84 rxe->attr.max_qp_wr = RXE_MAX_QP_WR; 85 rxe->attr.device_cap_flags = RXE_DEVICE_CAP_FLAGS; 86 rxe->attr.max_send_sge = RXE_MAX_SGE; 87 rxe->attr.max_recv_sge = RXE_MAX_SGE; 88 rxe->attr.max_sge_rd = RXE_MAX_SGE_RD; 89 rxe->attr.max_cq = RXE_MAX_CQ; 90 rxe->attr.max_cqe = (1 << RXE_MAX_LOG_CQE) - 1; 91 rxe->attr.max_mr = RXE_MAX_MR; 92 rxe->attr.max_pd = RXE_MAX_PD; 93 rxe->attr.max_qp_rd_atom = RXE_MAX_QP_RD_ATOM; 94 rxe->attr.max_res_rd_atom = RXE_MAX_RES_RD_ATOM; 95 rxe->attr.max_qp_init_rd_atom = RXE_MAX_QP_INIT_RD_ATOM; 96 rxe->attr.atomic_cap = IB_ATOMIC_HCA; 97 rxe->attr.max_mcast_grp = RXE_MAX_MCAST_GRP; 98 rxe->attr.max_mcast_qp_attach = RXE_MAX_MCAST_QP_ATTACH; 99 rxe->attr.max_total_mcast_qp_attach = RXE_MAX_TOT_MCAST_QP_ATTACH; 100 rxe->attr.max_ah = RXE_MAX_AH; 101 rxe->attr.max_srq = RXE_MAX_SRQ; 102 rxe->attr.max_srq_wr = RXE_MAX_SRQ_WR; 103 rxe->attr.max_srq_sge = RXE_MAX_SRQ_SGE; 104 rxe->attr.max_fast_reg_page_list_len = RXE_MAX_FMR_PAGE_LIST_LEN; 105 rxe->attr.max_pkeys = RXE_MAX_PKEYS; 106 rxe->attr.local_ca_ack_delay = RXE_LOCAL_CA_ACK_DELAY; 107 addrconf_addr_eui48((unsigned char *)&rxe->attr.sys_image_guid, 108 rxe->ndev->dev_addr); 109 110 rxe->max_ucontext = RXE_MAX_UCONTEXT; 111 } 112 113 /* initialize port attributes */ 114 static int rxe_init_port_param(struct rxe_port *port) 115 { 116 port->attr.state = IB_PORT_DOWN; 117 port->attr.max_mtu = IB_MTU_4096; 118 port->attr.active_mtu = IB_MTU_256; 119 port->attr.gid_tbl_len = RXE_PORT_GID_TBL_LEN; 120 port->attr.port_cap_flags = RXE_PORT_PORT_CAP_FLAGS; 121 port->attr.max_msg_sz = RXE_PORT_MAX_MSG_SZ; 122 port->attr.bad_pkey_cntr = RXE_PORT_BAD_PKEY_CNTR; 123 port->attr.qkey_viol_cntr = RXE_PORT_QKEY_VIOL_CNTR; 124 port->attr.pkey_tbl_len = RXE_PORT_PKEY_TBL_LEN; 125 port->attr.lid = RXE_PORT_LID; 126 port->attr.sm_lid = RXE_PORT_SM_LID; 127 port->attr.lmc = RXE_PORT_LMC; 128 port->attr.max_vl_num = RXE_PORT_MAX_VL_NUM; 129 port->attr.sm_sl = RXE_PORT_SM_SL; 130 port->attr.subnet_timeout = RXE_PORT_SUBNET_TIMEOUT; 131 port->attr.init_type_reply = RXE_PORT_INIT_TYPE_REPLY; 132 port->attr.active_width = RXE_PORT_ACTIVE_WIDTH; 133 port->attr.active_speed = RXE_PORT_ACTIVE_SPEED; 134 port->attr.phys_state = RXE_PORT_PHYS_STATE; 135 port->mtu_cap = ib_mtu_enum_to_int(IB_MTU_256); 136 port->subnet_prefix = cpu_to_be64(RXE_PORT_SUBNET_PREFIX); 137 138 return 0; 139 } 140 141 /* initialize port state, note IB convention that HCA ports are always 142 * numbered from 1 143 */ 144 static int rxe_init_ports(struct rxe_dev *rxe) 145 { 146 struct rxe_port *port = &rxe->port; 147 148 rxe_init_port_param(port); 149 150 if (!port->attr.pkey_tbl_len || !port->attr.gid_tbl_len) 151 return -EINVAL; 152 153 port->pkey_tbl = kcalloc(port->attr.pkey_tbl_len, 154 sizeof(*port->pkey_tbl), GFP_KERNEL); 155 156 if (!port->pkey_tbl) 157 return -ENOMEM; 158 159 port->pkey_tbl[0] = 0xffff; 160 addrconf_addr_eui48((unsigned char *)&port->port_guid, 161 rxe->ndev->dev_addr); 162 163 spin_lock_init(&port->port_lock); 164 165 return 0; 166 } 167 168 /* init pools of managed objects */ 169 static int rxe_init_pools(struct rxe_dev *rxe) 170 { 171 int err; 172 173 err = rxe_pool_init(rxe, &rxe->uc_pool, RXE_TYPE_UC, 174 rxe->max_ucontext); 175 if (err) 176 goto err1; 177 178 err = rxe_pool_init(rxe, &rxe->pd_pool, RXE_TYPE_PD, 179 rxe->attr.max_pd); 180 if (err) 181 goto err2; 182 183 err = rxe_pool_init(rxe, &rxe->ah_pool, RXE_TYPE_AH, 184 rxe->attr.max_ah); 185 if (err) 186 goto err3; 187 188 err = rxe_pool_init(rxe, &rxe->srq_pool, RXE_TYPE_SRQ, 189 rxe->attr.max_srq); 190 if (err) 191 goto err4; 192 193 err = rxe_pool_init(rxe, &rxe->qp_pool, RXE_TYPE_QP, 194 rxe->attr.max_qp); 195 if (err) 196 goto err5; 197 198 err = rxe_pool_init(rxe, &rxe->cq_pool, RXE_TYPE_CQ, 199 rxe->attr.max_cq); 200 if (err) 201 goto err6; 202 203 err = rxe_pool_init(rxe, &rxe->mr_pool, RXE_TYPE_MR, 204 rxe->attr.max_mr); 205 if (err) 206 goto err7; 207 208 err = rxe_pool_init(rxe, &rxe->mw_pool, RXE_TYPE_MW, 209 rxe->attr.max_mw); 210 if (err) 211 goto err8; 212 213 err = rxe_pool_init(rxe, &rxe->mc_grp_pool, RXE_TYPE_MC_GRP, 214 rxe->attr.max_mcast_grp); 215 if (err) 216 goto err9; 217 218 err = rxe_pool_init(rxe, &rxe->mc_elem_pool, RXE_TYPE_MC_ELEM, 219 rxe->attr.max_total_mcast_qp_attach); 220 if (err) 221 goto err10; 222 223 return 0; 224 225 err10: 226 rxe_pool_cleanup(&rxe->mc_grp_pool); 227 err9: 228 rxe_pool_cleanup(&rxe->mw_pool); 229 err8: 230 rxe_pool_cleanup(&rxe->mr_pool); 231 err7: 232 rxe_pool_cleanup(&rxe->cq_pool); 233 err6: 234 rxe_pool_cleanup(&rxe->qp_pool); 235 err5: 236 rxe_pool_cleanup(&rxe->srq_pool); 237 err4: 238 rxe_pool_cleanup(&rxe->ah_pool); 239 err3: 240 rxe_pool_cleanup(&rxe->pd_pool); 241 err2: 242 rxe_pool_cleanup(&rxe->uc_pool); 243 err1: 244 return err; 245 } 246 247 /* initialize rxe device state */ 248 static int rxe_init(struct rxe_dev *rxe) 249 { 250 int err; 251 252 /* init default device parameters */ 253 rxe_init_device_param(rxe); 254 255 err = rxe_init_ports(rxe); 256 if (err) 257 goto err1; 258 259 err = rxe_init_pools(rxe); 260 if (err) 261 goto err2; 262 263 /* init pending mmap list */ 264 spin_lock_init(&rxe->mmap_offset_lock); 265 spin_lock_init(&rxe->pending_lock); 266 INIT_LIST_HEAD(&rxe->pending_mmaps); 267 268 mutex_init(&rxe->usdev_lock); 269 270 return 0; 271 272 err2: 273 rxe_cleanup_ports(rxe); 274 err1: 275 return err; 276 } 277 278 void rxe_set_mtu(struct rxe_dev *rxe, unsigned int ndev_mtu) 279 { 280 struct rxe_port *port = &rxe->port; 281 enum ib_mtu mtu; 282 283 mtu = eth_mtu_int_to_enum(ndev_mtu); 284 285 /* Make sure that new MTU in range */ 286 mtu = mtu ? min_t(enum ib_mtu, mtu, IB_MTU_4096) : IB_MTU_256; 287 288 port->attr.active_mtu = mtu; 289 port->mtu_cap = ib_mtu_enum_to_int(mtu); 290 } 291 292 /* called by ifc layer to create new rxe device. 293 * The caller should allocate memory for rxe by calling ib_alloc_device. 294 */ 295 int rxe_add(struct rxe_dev *rxe, unsigned int mtu, const char *ibdev_name) 296 { 297 int err; 298 299 err = rxe_init(rxe); 300 if (err) 301 return err; 302 303 rxe_set_mtu(rxe, mtu); 304 305 return rxe_register_device(rxe, ibdev_name); 306 } 307 308 static int rxe_newlink(const char *ibdev_name, struct net_device *ndev) 309 { 310 struct rxe_dev *exists; 311 int err = 0; 312 313 exists = rxe_get_dev_from_net(ndev); 314 if (exists) { 315 ib_device_put(&exists->ib_dev); 316 pr_err("already configured on %s\n", ndev->name); 317 err = -EEXIST; 318 goto err; 319 } 320 321 err = rxe_net_add(ibdev_name, ndev); 322 if (err) { 323 pr_err("failed to add %s\n", ndev->name); 324 goto err; 325 } 326 err: 327 return err; 328 } 329 330 static struct rdma_link_ops rxe_link_ops = { 331 .type = "rxe", 332 .newlink = rxe_newlink, 333 }; 334 335 static int __init rxe_module_init(void) 336 { 337 int err; 338 339 /* initialize slab caches for managed objects */ 340 err = rxe_cache_init(); 341 if (err) { 342 pr_err("unable to init object pools\n"); 343 return err; 344 } 345 346 err = rxe_net_init(); 347 if (err) 348 return err; 349 350 rdma_link_register(&rxe_link_ops); 351 pr_info("loaded\n"); 352 return 0; 353 } 354 355 static void __exit rxe_module_exit(void) 356 { 357 rdma_link_unregister(&rxe_link_ops); 358 ib_unregister_driver(RDMA_DRIVER_RXE); 359 rxe_net_exit(); 360 rxe_cache_exit(); 361 362 pr_info("unloaded\n"); 363 } 364 365 late_initcall(rxe_module_init); 366 module_exit(rxe_module_exit); 367 368 MODULE_ALIAS_RDMA_LINK("rxe"); 369