| 1 | /* |
| 2 | * librpma_fio: librpma_apm and librpma_gpspm engines' common part. |
| 3 | * |
| 4 | * Copyright 2021, Intel Corporation |
| 5 | * |
| 6 | * This program is free software; you can redistribute it and/or |
| 7 | * modify it under the terms of the GNU General Public License, |
| 8 | * version 2 as published by the Free Software Foundation.. |
| 9 | * |
| 10 | * This program is distributed in the hope that it will be useful, |
| 11 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 12 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 13 | * GNU General Public License for more details. |
| 14 | */ |
| 15 | |
| 16 | #include "librpma_fio.h" |
| 17 | |
| 18 | #include <libpmem.h> |
| 19 | |
| 20 | struct fio_option librpma_fio_options[] = { |
| 21 | { |
| 22 | .name = "serverip", |
| 23 | .lname = "rpma_server_ip", |
| 24 | .type = FIO_OPT_STR_STORE, |
| 25 | .off1 = offsetof(struct librpma_fio_options_values, server_ip), |
| 26 | .help = "IP address the server is listening on", |
| 27 | .def = "", |
| 28 | .category = FIO_OPT_C_ENGINE, |
| 29 | .group = FIO_OPT_G_LIBRPMA, |
| 30 | }, |
| 31 | { |
| 32 | .name = "port", |
| 33 | .lname = "rpma_server port", |
| 34 | .type = FIO_OPT_STR_STORE, |
| 35 | .off1 = offsetof(struct librpma_fio_options_values, port), |
| 36 | .help = "port the server is listening on", |
| 37 | .def = "7204", |
| 38 | .category = FIO_OPT_C_ENGINE, |
| 39 | .group = FIO_OPT_G_LIBRPMA, |
| 40 | }, |
| 41 | { |
| 42 | .name = "direct_write_to_pmem", |
| 43 | .lname = "Direct Write to PMem (via RDMA) from the remote host is possible", |
| 44 | .type = FIO_OPT_BOOL, |
| 45 | .off1 = offsetof(struct librpma_fio_options_values, |
| 46 | direct_write_to_pmem), |
| 47 | .help = "Set to true ONLY when Direct Write to PMem from the remote host is possible (https://pmem.io/rpma/documentation/basic-direct-write-to-pmem.html)", |
| 48 | .def = "", |
| 49 | .category = FIO_OPT_C_ENGINE, |
| 50 | .group = FIO_OPT_G_LIBRPMA, |
| 51 | }, |
| 52 | { |
| 53 | .name = "busy_wait_polling", |
| 54 | .lname = "Set to 0 to wait for completion instead of busy-wait polling completion.", |
| 55 | .type = FIO_OPT_BOOL, |
| 56 | .off1 = offsetof(struct librpma_fio_options_values, |
| 57 | busy_wait_polling), |
| 58 | .help = "Set to false if you want to reduce CPU usage", |
| 59 | .def = "1", |
| 60 | .category = FIO_OPT_C_ENGINE, |
| 61 | .group = FIO_OPT_G_LIBRPMA, |
| 62 | }, |
| 63 | { |
| 64 | .name = NULL, |
| 65 | }, |
| 66 | }; |
| 67 | |
| 68 | int librpma_fio_td_port(const char *port_base_str, struct thread_data *td, |
| 69 | char *port_out) |
| 70 | { |
| 71 | unsigned long int port_ul = strtoul(port_base_str, NULL, 10); |
| 72 | unsigned int port_new; |
| 73 | |
| 74 | port_out[0] = '\0'; |
| 75 | |
| 76 | if (port_ul == ULONG_MAX) { |
| 77 | td_verror(td, errno, "strtoul"); |
| 78 | return -1; |
| 79 | } |
| 80 | port_ul += td->thread_number - 1; |
| 81 | if (port_ul >= UINT_MAX) { |
| 82 | log_err("[%u] port number (%lu) bigger than UINT_MAX\n", |
| 83 | td->thread_number, port_ul); |
| 84 | return -1; |
| 85 | } |
| 86 | |
| 87 | port_new = port_ul; |
| 88 | snprintf(port_out, LIBRPMA_FIO_PORT_STR_LEN_MAX - 1, "%u", port_new); |
| 89 | |
| 90 | return 0; |
| 91 | } |
| 92 | |
| 93 | char *librpma_fio_allocate_dram(struct thread_data *td, size_t size, |
| 94 | struct librpma_fio_mem *mem) |
| 95 | { |
| 96 | char *mem_ptr = NULL; |
| 97 | int ret; |
| 98 | |
| 99 | if ((ret = posix_memalign((void **)&mem_ptr, page_size, size))) { |
| 100 | log_err("fio: posix_memalign() failed\n"); |
| 101 | td_verror(td, ret, "posix_memalign"); |
| 102 | return NULL; |
| 103 | } |
| 104 | |
| 105 | mem->mem_ptr = mem_ptr; |
| 106 | mem->size_mmap = 0; |
| 107 | |
| 108 | return mem_ptr; |
| 109 | } |
| 110 | |
| 111 | char *librpma_fio_allocate_pmem(struct thread_data *td, struct fio_file *f, |
| 112 | size_t size, struct librpma_fio_mem *mem) |
| 113 | { |
| 114 | size_t size_mmap = 0; |
| 115 | char *mem_ptr = NULL; |
| 116 | int is_pmem = 0; |
| 117 | size_t ws_offset; |
| 118 | |
| 119 | if (size % page_size) { |
| 120 | log_err("fio: size (%zu) is not aligned to page size (%zu)\n", |
| 121 | size, page_size); |
| 122 | return NULL; |
| 123 | } |
| 124 | |
| 125 | if (f->filetype == FIO_TYPE_CHAR) { |
| 126 | /* Each thread uses a separate offset within DeviceDAX. */ |
| 127 | ws_offset = (td->thread_number - 1) * size; |
| 128 | } else { |
| 129 | /* Each thread uses a separate FileSystemDAX file. No offset is needed. */ |
| 130 | ws_offset = 0; |
| 131 | } |
| 132 | |
| 133 | if (!f->file_name) { |
| 134 | log_err("fio: filename is not set\n"); |
| 135 | return NULL; |
| 136 | } |
| 137 | |
| 138 | /* map the file */ |
| 139 | mem_ptr = pmem_map_file(f->file_name, 0 /* len */, 0 /* flags */, |
| 140 | 0 /* mode */, &size_mmap, &is_pmem); |
| 141 | if (mem_ptr == NULL) { |
| 142 | log_err("fio: pmem_map_file(%s) failed\n", f->file_name); |
| 143 | /* pmem_map_file() sets errno on failure */ |
| 144 | td_verror(td, errno, "pmem_map_file"); |
| 145 | return NULL; |
| 146 | } |
| 147 | |
| 148 | /* pmem is expected */ |
| 149 | if (!is_pmem) { |
| 150 | log_err("fio: %s is not located in persistent memory\n", |
| 151 | f->file_name); |
| 152 | goto err_unmap; |
| 153 | } |
| 154 | |
| 155 | /* check size of allocated persistent memory */ |
| 156 | if (size_mmap < ws_offset + size) { |
| 157 | log_err( |
| 158 | "fio: %s is too small to handle so many threads (%zu < %zu)\n", |
| 159 | f->file_name, size_mmap, ws_offset + size); |
| 160 | goto err_unmap; |
| 161 | } |
| 162 | |
| 163 | log_info("fio: size of memory mapped from the file %s: %zu\n", |
| 164 | f->file_name, size_mmap); |
| 165 | |
| 166 | mem->mem_ptr = mem_ptr; |
| 167 | mem->size_mmap = size_mmap; |
| 168 | |
| 169 | return mem_ptr + ws_offset; |
| 170 | |
| 171 | err_unmap: |
| 172 | (void) pmem_unmap(mem_ptr, size_mmap); |
| 173 | return NULL; |
| 174 | } |
| 175 | |
| 176 | void librpma_fio_free(struct librpma_fio_mem *mem) |
| 177 | { |
| 178 | if (mem->size_mmap) |
| 179 | (void) pmem_unmap(mem->mem_ptr, mem->size_mmap); |
| 180 | else |
| 181 | free(mem->mem_ptr); |
| 182 | } |
| 183 | |
| 184 | #define LIBRPMA_FIO_RETRY_MAX_NO 10 |
| 185 | #define LIBRPMA_FIO_RETRY_DELAY_S 5 |
| 186 | |
| 187 | int librpma_fio_client_init(struct thread_data *td, |
| 188 | struct rpma_conn_cfg *cfg) |
| 189 | { |
| 190 | struct librpma_fio_client_data *ccd; |
| 191 | struct librpma_fio_options_values *o = td->eo; |
| 192 | struct ibv_context *dev = NULL; |
| 193 | char port_td[LIBRPMA_FIO_PORT_STR_LEN_MAX]; |
| 194 | struct rpma_conn_req *req = NULL; |
| 195 | enum rpma_conn_event event; |
| 196 | struct rpma_conn_private_data pdata; |
| 197 | enum rpma_log_level log_level_aux = RPMA_LOG_LEVEL_WARNING; |
| 198 | int remote_flush_type; |
| 199 | int retry; |
| 200 | int ret; |
| 201 | |
| 202 | /* --debug=net sets RPMA_LOG_THRESHOLD_AUX to RPMA_LOG_LEVEL_INFO */ |
| 203 | #ifdef FIO_INC_DEBUG |
| 204 | if ((1UL << FD_NET) & fio_debug) |
| 205 | log_level_aux = RPMA_LOG_LEVEL_INFO; |
| 206 | #endif |
| 207 | |
| 208 | /* configure logging thresholds to see more details */ |
| 209 | rpma_log_set_threshold(RPMA_LOG_THRESHOLD, RPMA_LOG_LEVEL_INFO); |
| 210 | rpma_log_set_threshold(RPMA_LOG_THRESHOLD_AUX, log_level_aux); |
| 211 | |
| 212 | /* obtain an IBV context for a remote IP address */ |
| 213 | if ((ret = rpma_utils_get_ibv_context(o->server_ip, |
| 214 | RPMA_UTIL_IBV_CONTEXT_REMOTE, &dev))) { |
| 215 | librpma_td_verror(td, ret, "rpma_utils_get_ibv_context"); |
| 216 | return -1; |
| 217 | } |
| 218 | |
| 219 | /* allocate client's data */ |
| 220 | ccd = calloc(1, sizeof(*ccd)); |
| 221 | if (ccd == NULL) { |
| 222 | td_verror(td, errno, "calloc"); |
| 223 | return -1; |
| 224 | } |
| 225 | |
| 226 | /* allocate all in-memory queues */ |
| 227 | ccd->io_us_queued = calloc(td->o.iodepth, sizeof(*ccd->io_us_queued)); |
| 228 | if (ccd->io_us_queued == NULL) { |
| 229 | td_verror(td, errno, "calloc"); |
| 230 | goto err_free_ccd; |
| 231 | } |
| 232 | |
| 233 | ccd->io_us_flight = calloc(td->o.iodepth, sizeof(*ccd->io_us_flight)); |
| 234 | if (ccd->io_us_flight == NULL) { |
| 235 | td_verror(td, errno, "calloc"); |
| 236 | goto err_free_io_u_queues; |
| 237 | } |
| 238 | |
| 239 | ccd->io_us_completed = calloc(td->o.iodepth, |
| 240 | sizeof(*ccd->io_us_completed)); |
| 241 | if (ccd->io_us_completed == NULL) { |
| 242 | td_verror(td, errno, "calloc"); |
| 243 | goto err_free_io_u_queues; |
| 244 | } |
| 245 | |
| 246 | /* create a new peer object */ |
| 247 | if ((ret = rpma_peer_new(dev, &ccd->peer))) { |
| 248 | librpma_td_verror(td, ret, "rpma_peer_new"); |
| 249 | goto err_free_io_u_queues; |
| 250 | } |
| 251 | |
| 252 | /* create a connection request */ |
| 253 | if (librpma_fio_td_port(o->port, td, port_td)) |
| 254 | goto err_peer_delete; |
| 255 | |
| 256 | for (retry = 0; retry < LIBRPMA_FIO_RETRY_MAX_NO; retry++) { |
| 257 | if ((ret = rpma_conn_req_new(ccd->peer, o->server_ip, port_td, |
| 258 | cfg, &req))) { |
| 259 | librpma_td_verror(td, ret, "rpma_conn_req_new"); |
| 260 | goto err_peer_delete; |
| 261 | } |
| 262 | |
| 263 | /* |
| 264 | * Connect the connection request |
| 265 | * and obtain the connection object. |
| 266 | */ |
| 267 | if ((ret = rpma_conn_req_connect(&req, NULL, &ccd->conn))) { |
| 268 | librpma_td_verror(td, ret, "rpma_conn_req_connect"); |
| 269 | goto err_req_delete; |
| 270 | } |
| 271 | |
| 272 | /* wait for the connection to establish */ |
| 273 | if ((ret = rpma_conn_next_event(ccd->conn, &event))) { |
| 274 | librpma_td_verror(td, ret, "rpma_conn_next_event"); |
| 275 | goto err_conn_delete; |
| 276 | } else if (event == RPMA_CONN_ESTABLISHED) { |
| 277 | break; |
| 278 | } else if (event == RPMA_CONN_REJECTED) { |
| 279 | (void) rpma_conn_disconnect(ccd->conn); |
| 280 | (void) rpma_conn_delete(&ccd->conn); |
| 281 | if (retry < LIBRPMA_FIO_RETRY_MAX_NO - 1) { |
| 282 | log_err("Thread [%d]: Retrying (#%i) ...\n", |
| 283 | td->thread_number, retry + 1); |
| 284 | sleep(LIBRPMA_FIO_RETRY_DELAY_S); |
| 285 | } else { |
| 286 | log_err( |
| 287 | "Thread [%d]: The maximum number of retries exceeded. Closing.\n", |
| 288 | td->thread_number); |
| 289 | } |
| 290 | } else { |
| 291 | log_err( |
| 292 | "rpma_conn_next_event returned an unexptected event: (%s != RPMA_CONN_ESTABLISHED)\n", |
| 293 | rpma_utils_conn_event_2str(event)); |
| 294 | goto err_conn_delete; |
| 295 | } |
| 296 | } |
| 297 | |
| 298 | if (retry > 0) |
| 299 | log_err("Thread [%d]: Connected after retry #%i\n", |
| 300 | td->thread_number, retry); |
| 301 | |
| 302 | if (ccd->conn == NULL) |
| 303 | goto err_peer_delete; |
| 304 | |
| 305 | /* get the connection's main CQ */ |
| 306 | if ((ret = rpma_conn_get_cq(ccd->conn, &ccd->cq))) { |
| 307 | librpma_td_verror(td, ret, "rpma_conn_get_cq"); |
| 308 | goto err_conn_delete; |
| 309 | } |
| 310 | |
| 311 | /* get the connection's private data sent from the server */ |
| 312 | if ((ret = rpma_conn_get_private_data(ccd->conn, &pdata))) { |
| 313 | librpma_td_verror(td, ret, "rpma_conn_get_private_data"); |
| 314 | goto err_conn_delete; |
| 315 | } |
| 316 | |
| 317 | /* get the server's workspace representation */ |
| 318 | ccd->ws = pdata.ptr; |
| 319 | |
| 320 | /* create the server's memory representation */ |
| 321 | if ((ret = rpma_mr_remote_from_descriptor(&ccd->ws->descriptor[0], |
| 322 | ccd->ws->mr_desc_size, &ccd->server_mr))) { |
| 323 | librpma_td_verror(td, ret, "rpma_mr_remote_from_descriptor"); |
| 324 | goto err_conn_delete; |
| 325 | } |
| 326 | |
| 327 | /* get the total size of the shared server memory */ |
| 328 | if ((ret = rpma_mr_remote_get_size(ccd->server_mr, &ccd->ws_size))) { |
| 329 | librpma_td_verror(td, ret, "rpma_mr_remote_get_size"); |
| 330 | goto err_conn_delete; |
| 331 | } |
| 332 | |
| 333 | /* get flush type of the remote node */ |
| 334 | if ((ret = rpma_mr_remote_get_flush_type(ccd->server_mr, |
| 335 | &remote_flush_type))) { |
| 336 | librpma_td_verror(td, ret, "rpma_mr_remote_get_flush_type"); |
| 337 | goto err_conn_delete; |
| 338 | } |
| 339 | |
| 340 | ccd->server_mr_flush_type = |
| 341 | (remote_flush_type & RPMA_MR_USAGE_FLUSH_TYPE_PERSISTENT) ? |
| 342 | RPMA_FLUSH_TYPE_PERSISTENT : RPMA_FLUSH_TYPE_VISIBILITY; |
| 343 | |
| 344 | /* |
| 345 | * Assure an io_us buffer allocation is page-size-aligned which is required |
| 346 | * to register for RDMA. User-provided value is intentionally ignored. |
| 347 | */ |
| 348 | td->o.mem_align = page_size; |
| 349 | |
| 350 | td->io_ops_data = ccd; |
| 351 | |
| 352 | return 0; |
| 353 | |
| 354 | err_conn_delete: |
| 355 | (void) rpma_conn_disconnect(ccd->conn); |
| 356 | (void) rpma_conn_delete(&ccd->conn); |
| 357 | |
| 358 | err_req_delete: |
| 359 | (void) rpma_conn_req_delete(&req); |
| 360 | |
| 361 | err_peer_delete: |
| 362 | (void) rpma_peer_delete(&ccd->peer); |
| 363 | |
| 364 | err_free_io_u_queues: |
| 365 | free(ccd->io_us_queued); |
| 366 | free(ccd->io_us_flight); |
| 367 | free(ccd->io_us_completed); |
| 368 | |
| 369 | err_free_ccd: |
| 370 | free(ccd); |
| 371 | |
| 372 | return -1; |
| 373 | } |
| 374 | |
| 375 | void librpma_fio_client_cleanup(struct thread_data *td) |
| 376 | { |
| 377 | struct librpma_fio_client_data *ccd = td->io_ops_data; |
| 378 | enum rpma_conn_event ev; |
| 379 | int ret; |
| 380 | |
| 381 | if (ccd == NULL) |
| 382 | return; |
| 383 | |
| 384 | /* delete the iou's memory registration */ |
| 385 | if ((ret = rpma_mr_dereg(&ccd->orig_mr))) |
| 386 | librpma_td_verror(td, ret, "rpma_mr_dereg"); |
| 387 | /* delete the iou's memory registration */ |
| 388 | if ((ret = rpma_mr_remote_delete(&ccd->server_mr))) |
| 389 | librpma_td_verror(td, ret, "rpma_mr_remote_delete"); |
| 390 | /* initiate disconnection */ |
| 391 | if ((ret = rpma_conn_disconnect(ccd->conn))) |
| 392 | librpma_td_verror(td, ret, "rpma_conn_disconnect"); |
| 393 | /* wait for disconnection to end up */ |
| 394 | if ((ret = rpma_conn_next_event(ccd->conn, &ev))) { |
| 395 | librpma_td_verror(td, ret, "rpma_conn_next_event"); |
| 396 | } else if (ev != RPMA_CONN_CLOSED) { |
| 397 | log_err( |
| 398 | "client_cleanup received an unexpected event (%s != RPMA_CONN_CLOSED)\n", |
| 399 | rpma_utils_conn_event_2str(ev)); |
| 400 | } |
| 401 | /* delete the connection */ |
| 402 | if ((ret = rpma_conn_delete(&ccd->conn))) |
| 403 | librpma_td_verror(td, ret, "rpma_conn_delete"); |
| 404 | /* delete the peer */ |
| 405 | if ((ret = rpma_peer_delete(&ccd->peer))) |
| 406 | librpma_td_verror(td, ret, "rpma_peer_delete"); |
| 407 | /* free the software queues */ |
| 408 | free(ccd->io_us_queued); |
| 409 | free(ccd->io_us_flight); |
| 410 | free(ccd->io_us_completed); |
| 411 | free(ccd); |
| 412 | td->io_ops_data = NULL; /* zero ccd */ |
| 413 | } |
| 414 | |
| 415 | int librpma_fio_file_nop(struct thread_data *td, struct fio_file *f) |
| 416 | { |
| 417 | /* NOP */ |
| 418 | return 0; |
| 419 | } |
| 420 | |
| 421 | int librpma_fio_client_post_init(struct thread_data *td) |
| 422 | { |
| 423 | struct librpma_fio_client_data *ccd = td->io_ops_data; |
| 424 | size_t io_us_size; |
| 425 | int ret; |
| 426 | |
| 427 | /* |
| 428 | * td->orig_buffer is not aligned. The engine requires aligned io_us |
| 429 | * so FIO aligns up the address using the formula below. |
| 430 | */ |
| 431 | ccd->orig_buffer_aligned = PTR_ALIGN(td->orig_buffer, page_mask) + |
| 432 | td->o.mem_align; |
| 433 | |
| 434 | /* |
| 435 | * td->orig_buffer_size beside the space really consumed by io_us |
| 436 | * has paddings which can be omitted for the memory registration. |
| 437 | */ |
| 438 | io_us_size = (unsigned long long)td_max_bs(td) * |
| 439 | (unsigned long long)td->o.iodepth; |
| 440 | |
| 441 | if ((ret = rpma_mr_reg(ccd->peer, ccd->orig_buffer_aligned, io_us_size, |
| 442 | RPMA_MR_USAGE_READ_DST | RPMA_MR_USAGE_READ_SRC | |
| 443 | RPMA_MR_USAGE_WRITE_DST | RPMA_MR_USAGE_WRITE_SRC | |
| 444 | RPMA_MR_USAGE_FLUSH_TYPE_PERSISTENT, &ccd->orig_mr))) |
| 445 | librpma_td_verror(td, ret, "rpma_mr_reg"); |
| 446 | return ret; |
| 447 | } |
| 448 | |
| 449 | int librpma_fio_client_get_file_size(struct thread_data *td, |
| 450 | struct fio_file *f) |
| 451 | { |
| 452 | struct librpma_fio_client_data *ccd = td->io_ops_data; |
| 453 | |
| 454 | f->real_file_size = ccd->ws_size; |
| 455 | fio_file_set_size_known(f); |
| 456 | |
| 457 | return 0; |
| 458 | } |
| 459 | |
| 460 | static enum fio_q_status client_queue_sync(struct thread_data *td, |
| 461 | struct io_u *io_u) |
| 462 | { |
| 463 | struct librpma_fio_client_data *ccd = td->io_ops_data; |
| 464 | struct ibv_wc wc; |
| 465 | unsigned io_u_index; |
| 466 | int ret; |
| 467 | |
| 468 | /* execute io_u */ |
| 469 | if (io_u->ddir == DDIR_READ) { |
| 470 | /* post an RDMA read operation */ |
| 471 | if (librpma_fio_client_io_read(td, io_u, |
| 472 | RPMA_F_COMPLETION_ALWAYS)) |
| 473 | goto err; |
| 474 | } else if (io_u->ddir == DDIR_WRITE) { |
| 475 | /* post an RDMA write operation */ |
| 476 | if (librpma_fio_client_io_write(td, io_u)) |
| 477 | goto err; |
| 478 | if (ccd->flush(td, io_u, io_u, io_u->xfer_buflen)) |
| 479 | goto err; |
| 480 | } else { |
| 481 | log_err("unsupported IO mode: %s\n", io_ddir_name(io_u->ddir)); |
| 482 | goto err; |
| 483 | } |
| 484 | |
| 485 | do { |
| 486 | /* get a completion */ |
| 487 | ret = rpma_cq_get_wc(ccd->cq, 1, &wc, NULL); |
| 488 | if (ret == RPMA_E_NO_COMPLETION) { |
| 489 | /* lack of completion is not an error */ |
| 490 | continue; |
| 491 | } else if (ret != 0) { |
| 492 | /* an error occurred */ |
| 493 | librpma_td_verror(td, ret, "rpma_cq_get_wc"); |
| 494 | goto err; |
| 495 | } |
| 496 | |
| 497 | /* if io_us has completed with an error */ |
| 498 | if (wc.status != IBV_WC_SUCCESS) |
| 499 | goto err; |
| 500 | |
| 501 | if (wc.opcode == IBV_WC_SEND) |
| 502 | ++ccd->op_send_completed; |
| 503 | else { |
| 504 | if (wc.opcode == IBV_WC_RECV) |
| 505 | ++ccd->op_recv_completed; |
| 506 | |
| 507 | break; |
| 508 | } |
| 509 | } while (1); |
| 510 | |
| 511 | if (ccd->get_io_u_index(&wc, &io_u_index) != 1) |
| 512 | goto err; |
| 513 | |
| 514 | if (io_u->index != io_u_index) { |
| 515 | log_err( |
| 516 | "no matching io_u for received completion found (io_u_index=%u)\n", |
| 517 | io_u_index); |
| 518 | goto err; |
| 519 | } |
| 520 | |
| 521 | /* make sure all SENDs are completed before exit - clean up SQ */ |
| 522 | if (librpma_fio_client_io_complete_all_sends(td)) |
| 523 | goto err; |
| 524 | |
| 525 | return FIO_Q_COMPLETED; |
| 526 | |
| 527 | err: |
| 528 | io_u->error = -1; |
| 529 | return FIO_Q_COMPLETED; |
| 530 | } |
| 531 | |
| 532 | enum fio_q_status librpma_fio_client_queue(struct thread_data *td, |
| 533 | struct io_u *io_u) |
| 534 | { |
| 535 | struct librpma_fio_client_data *ccd = td->io_ops_data; |
| 536 | |
| 537 | if (ccd->io_u_queued_nr == (int)td->o.iodepth) |
| 538 | return FIO_Q_BUSY; |
| 539 | |
| 540 | if (td->o.sync_io) |
| 541 | return client_queue_sync(td, io_u); |
| 542 | |
| 543 | /* io_u -> queued[] */ |
| 544 | ccd->io_us_queued[ccd->io_u_queued_nr] = io_u; |
| 545 | ccd->io_u_queued_nr++; |
| 546 | |
| 547 | return FIO_Q_QUEUED; |
| 548 | } |
| 549 | |
| 550 | int librpma_fio_client_commit(struct thread_data *td) |
| 551 | { |
| 552 | struct librpma_fio_client_data *ccd = td->io_ops_data; |
| 553 | int flags = RPMA_F_COMPLETION_ON_ERROR; |
| 554 | struct timespec now; |
| 555 | bool fill_time; |
| 556 | int i; |
| 557 | struct io_u *flush_first_io_u = NULL; |
| 558 | unsigned long long int flush_len = 0; |
| 559 | |
| 560 | if (!ccd->io_us_queued) |
| 561 | return -1; |
| 562 | |
| 563 | /* execute all io_us from queued[] */ |
| 564 | for (i = 0; i < ccd->io_u_queued_nr; i++) { |
| 565 | struct io_u *io_u = ccd->io_us_queued[i]; |
| 566 | |
| 567 | if (io_u->ddir == DDIR_READ) { |
| 568 | if (i + 1 == ccd->io_u_queued_nr || |
| 569 | ccd->io_us_queued[i + 1]->ddir == DDIR_WRITE) |
| 570 | flags = RPMA_F_COMPLETION_ALWAYS; |
| 571 | /* post an RDMA read operation */ |
| 572 | if (librpma_fio_client_io_read(td, io_u, flags)) |
| 573 | return -1; |
| 574 | } else if (io_u->ddir == DDIR_WRITE) { |
| 575 | /* post an RDMA write operation */ |
| 576 | if (librpma_fio_client_io_write(td, io_u)) |
| 577 | return -1; |
| 578 | |
| 579 | /* cache the first io_u in the sequence */ |
| 580 | if (flush_first_io_u == NULL) |
| 581 | flush_first_io_u = io_u; |
| 582 | |
| 583 | /* |
| 584 | * the flush length is the sum of all io_u's creating |
| 585 | * the sequence |
| 586 | */ |
| 587 | flush_len += io_u->xfer_buflen; |
| 588 | |
| 589 | /* |
| 590 | * if io_u's are random the rpma_flush is required |
| 591 | * after each one of them |
| 592 | */ |
| 593 | if (!td_random(td)) { |
| 594 | /* |
| 595 | * When the io_u's are sequential and |
| 596 | * the current io_u is not the last one and |
| 597 | * the next one is also a write operation |
| 598 | * the flush can be postponed by one io_u and |
| 599 | * cover all of them which build a continuous |
| 600 | * sequence. |
| 601 | */ |
| 602 | if ((i + 1 < ccd->io_u_queued_nr) && |
| 603 | (ccd->io_us_queued[i + 1]->ddir == DDIR_WRITE)) |
| 604 | continue; |
| 605 | } |
| 606 | |
| 607 | /* flush all writes which build a continuous sequence */ |
| 608 | if (ccd->flush(td, flush_first_io_u, io_u, flush_len)) |
| 609 | return -1; |
| 610 | |
| 611 | /* |
| 612 | * reset the flush parameters in preparation for |
| 613 | * the next one |
| 614 | */ |
| 615 | flush_first_io_u = NULL; |
| 616 | flush_len = 0; |
| 617 | } else { |
| 618 | log_err("unsupported IO mode: %s\n", |
| 619 | io_ddir_name(io_u->ddir)); |
| 620 | return -1; |
| 621 | } |
| 622 | } |
| 623 | |
| 624 | if ((fill_time = fio_fill_issue_time(td))) |
| 625 | fio_gettime(&now, NULL); |
| 626 | |
| 627 | /* move executed io_us from queued[] to flight[] */ |
| 628 | for (i = 0; i < ccd->io_u_queued_nr; i++) { |
| 629 | struct io_u *io_u = ccd->io_us_queued[i]; |
| 630 | |
| 631 | /* FIO does not do this if the engine is asynchronous */ |
| 632 | if (fill_time) |
| 633 | memcpy(&io_u->issue_time, &now, sizeof(now)); |
| 634 | |
| 635 | /* move executed io_us from queued[] to flight[] */ |
| 636 | ccd->io_us_flight[ccd->io_u_flight_nr] = io_u; |
| 637 | ccd->io_u_flight_nr++; |
| 638 | |
| 639 | /* |
| 640 | * FIO says: |
| 641 | * If an engine has the commit hook |
| 642 | * it has to call io_u_queued() itself. |
| 643 | */ |
| 644 | io_u_queued(td, io_u); |
| 645 | } |
| 646 | |
| 647 | /* FIO does not do this if an engine has the commit hook. */ |
| 648 | io_u_mark_submit(td, ccd->io_u_queued_nr); |
| 649 | ccd->io_u_queued_nr = 0; |
| 650 | |
| 651 | return 0; |
| 652 | } |
| 653 | |
| 654 | /* |
| 655 | * RETURN VALUE |
| 656 | * - > 0 - a number of completed io_us |
| 657 | * - 0 - when no complicitions received |
| 658 | * - (-1) - when an error occurred |
| 659 | */ |
| 660 | static int client_getevent_process(struct thread_data *td) |
| 661 | { |
| 662 | struct librpma_fio_client_data *ccd = td->io_ops_data; |
| 663 | struct ibv_wc wc; |
| 664 | /* io_u->index of completed io_u (wc.wr_id) */ |
| 665 | unsigned int io_u_index; |
| 666 | /* # of completed io_us */ |
| 667 | int cmpl_num = 0; |
| 668 | /* helpers */ |
| 669 | struct io_u *io_u; |
| 670 | int i; |
| 671 | int ret; |
| 672 | |
| 673 | /* get a completion */ |
| 674 | if ((ret = rpma_cq_get_wc(ccd->cq, 1, &wc, NULL))) { |
| 675 | /* lack of completion is not an error */ |
| 676 | if (ret == RPMA_E_NO_COMPLETION) { |
| 677 | /* lack of completion is not an error */ |
| 678 | return 0; |
| 679 | } |
| 680 | |
| 681 | /* an error occurred */ |
| 682 | librpma_td_verror(td, ret, "rpma_cq_get_wc"); |
| 683 | return -1; |
| 684 | } |
| 685 | |
| 686 | /* if io_us has completed with an error */ |
| 687 | if (wc.status != IBV_WC_SUCCESS) { |
| 688 | td->error = wc.status; |
| 689 | return -1; |
| 690 | } |
| 691 | |
| 692 | if (wc.opcode == IBV_WC_SEND) |
| 693 | ++ccd->op_send_completed; |
| 694 | else if (wc.opcode == IBV_WC_RECV) |
| 695 | ++ccd->op_recv_completed; |
| 696 | |
| 697 | if ((ret = ccd->get_io_u_index(&wc, &io_u_index)) != 1) |
| 698 | return ret; |
| 699 | |
| 700 | /* look for an io_u being completed */ |
| 701 | for (i = 0; i < ccd->io_u_flight_nr; ++i) { |
| 702 | if (ccd->io_us_flight[i]->index == io_u_index) { |
| 703 | cmpl_num = i + 1; |
| 704 | break; |
| 705 | } |
| 706 | } |
| 707 | |
| 708 | /* if no matching io_u has been found */ |
| 709 | if (cmpl_num == 0) { |
| 710 | log_err( |
| 711 | "no matching io_u for received completion found (io_u_index=%u)\n", |
| 712 | io_u_index); |
| 713 | return -1; |
| 714 | } |
| 715 | |
| 716 | /* move completed io_us to the completed in-memory queue */ |
| 717 | for (i = 0; i < cmpl_num; ++i) { |
| 718 | /* get and prepare io_u */ |
| 719 | io_u = ccd->io_us_flight[i]; |
| 720 | |
| 721 | /* append to the queue */ |
| 722 | ccd->io_us_completed[ccd->io_u_completed_nr] = io_u; |
| 723 | ccd->io_u_completed_nr++; |
| 724 | } |
| 725 | |
| 726 | /* remove completed io_us from the flight queue */ |
| 727 | for (i = cmpl_num; i < ccd->io_u_flight_nr; ++i) |
| 728 | ccd->io_us_flight[i - cmpl_num] = ccd->io_us_flight[i]; |
| 729 | ccd->io_u_flight_nr -= cmpl_num; |
| 730 | |
| 731 | return cmpl_num; |
| 732 | } |
| 733 | |
| 734 | int librpma_fio_client_getevents(struct thread_data *td, unsigned int min, |
| 735 | unsigned int max, const struct timespec *t) |
| 736 | { |
| 737 | struct librpma_fio_client_data *ccd = td->io_ops_data; |
| 738 | /* total # of completed io_us */ |
| 739 | int cmpl_num_total = 0; |
| 740 | /* # of completed io_us from a single event */ |
| 741 | int cmpl_num; |
| 742 | |
| 743 | do { |
| 744 | cmpl_num = client_getevent_process(td); |
| 745 | if (cmpl_num > 0) { |
| 746 | /* new completions collected */ |
| 747 | cmpl_num_total += cmpl_num; |
| 748 | } else if (cmpl_num == 0) { |
| 749 | /* |
| 750 | * It is required to make sure that CQEs for SENDs |
| 751 | * will flow at least at the same pace as CQEs for RECVs. |
| 752 | */ |
| 753 | if (cmpl_num_total >= min && |
| 754 | ccd->op_send_completed >= ccd->op_recv_completed) |
| 755 | break; |
| 756 | |
| 757 | /* |
| 758 | * To reduce CPU consumption one can use |
| 759 | * the rpma_cq_wait() function. |
| 760 | * Note this greatly increase the latency |
| 761 | * and make the results less stable. |
| 762 | * The bandwidth stays more or less the same. |
| 763 | */ |
| 764 | } else { |
| 765 | /* an error occurred */ |
| 766 | return -1; |
| 767 | } |
| 768 | |
| 769 | /* |
| 770 | * The expected max can be exceeded if CQEs for RECVs will come up |
| 771 | * faster than CQEs for SENDs. But it is required to make sure CQEs for |
| 772 | * SENDs will flow at least at the same pace as CQEs for RECVs. |
| 773 | */ |
| 774 | } while (cmpl_num_total < max || |
| 775 | ccd->op_send_completed < ccd->op_recv_completed); |
| 776 | |
| 777 | /* |
| 778 | * All posted SENDs are completed and RECVs for them (responses) are |
| 779 | * completed. This is the initial situation so the counters are reset. |
| 780 | */ |
| 781 | if (ccd->op_send_posted == ccd->op_send_completed && |
| 782 | ccd->op_send_completed == ccd->op_recv_completed) { |
| 783 | ccd->op_send_posted = 0; |
| 784 | ccd->op_send_completed = 0; |
| 785 | ccd->op_recv_completed = 0; |
| 786 | } |
| 787 | |
| 788 | return cmpl_num_total; |
| 789 | } |
| 790 | |
| 791 | struct io_u *librpma_fio_client_event(struct thread_data *td, int event) |
| 792 | { |
| 793 | struct librpma_fio_client_data *ccd = td->io_ops_data; |
| 794 | struct io_u *io_u; |
| 795 | int i; |
| 796 | |
| 797 | /* get the first io_u from the queue */ |
| 798 | io_u = ccd->io_us_completed[0]; |
| 799 | |
| 800 | /* remove the first io_u from the queue */ |
| 801 | for (i = 1; i < ccd->io_u_completed_nr; ++i) |
| 802 | ccd->io_us_completed[i - 1] = ccd->io_us_completed[i]; |
| 803 | ccd->io_u_completed_nr--; |
| 804 | |
| 805 | dprint_io_u(io_u, "client_event"); |
| 806 | |
| 807 | return io_u; |
| 808 | } |
| 809 | |
| 810 | char *librpma_fio_client_errdetails(struct io_u *io_u) |
| 811 | { |
| 812 | /* get the string representation of an error */ |
| 813 | enum ibv_wc_status status = io_u->error; |
| 814 | const char *status_str = ibv_wc_status_str(status); |
| 815 | |
| 816 | char *details = strdup(status_str); |
| 817 | if (details == NULL) { |
| 818 | fprintf(stderr, "Error: %s\n", status_str); |
| 819 | fprintf(stderr, "Fatal error: out of memory. Aborting.\n"); |
| 820 | abort(); |
| 821 | } |
| 822 | |
| 823 | /* FIO frees the returned string when it becomes obsolete */ |
| 824 | return details; |
| 825 | } |
| 826 | |
| 827 | int librpma_fio_server_init(struct thread_data *td) |
| 828 | { |
| 829 | struct librpma_fio_options_values *o = td->eo; |
| 830 | struct librpma_fio_server_data *csd; |
| 831 | struct ibv_context *dev = NULL; |
| 832 | enum rpma_log_level log_level_aux = RPMA_LOG_LEVEL_WARNING; |
| 833 | int ret = -1; |
| 834 | |
| 835 | /* --debug=net sets RPMA_LOG_THRESHOLD_AUX to RPMA_LOG_LEVEL_INFO */ |
| 836 | #ifdef FIO_INC_DEBUG |
| 837 | if ((1UL << FD_NET) & fio_debug) |
| 838 | log_level_aux = RPMA_LOG_LEVEL_INFO; |
| 839 | #endif |
| 840 | |
| 841 | /* configure logging thresholds to see more details */ |
| 842 | rpma_log_set_threshold(RPMA_LOG_THRESHOLD, RPMA_LOG_LEVEL_INFO); |
| 843 | rpma_log_set_threshold(RPMA_LOG_THRESHOLD_AUX, log_level_aux); |
| 844 | |
| 845 | |
| 846 | /* obtain an IBV context for a remote IP address */ |
| 847 | if ((ret = rpma_utils_get_ibv_context(o->server_ip, |
| 848 | RPMA_UTIL_IBV_CONTEXT_LOCAL, &dev))) { |
| 849 | librpma_td_verror(td, ret, "rpma_utils_get_ibv_context"); |
| 850 | return -1; |
| 851 | } |
| 852 | |
| 853 | /* allocate server's data */ |
| 854 | csd = calloc(1, sizeof(*csd)); |
| 855 | if (csd == NULL) { |
| 856 | td_verror(td, errno, "calloc"); |
| 857 | return -1; |
| 858 | } |
| 859 | |
| 860 | /* create a new peer object */ |
| 861 | if ((ret = rpma_peer_new(dev, &csd->peer))) { |
| 862 | librpma_td_verror(td, ret, "rpma_peer_new"); |
| 863 | goto err_free_csd; |
| 864 | } |
| 865 | |
| 866 | td->io_ops_data = csd; |
| 867 | |
| 868 | return 0; |
| 869 | |
| 870 | err_free_csd: |
| 871 | free(csd); |
| 872 | |
| 873 | return -1; |
| 874 | } |
| 875 | |
| 876 | void librpma_fio_server_cleanup(struct thread_data *td) |
| 877 | { |
| 878 | struct librpma_fio_server_data *csd = td->io_ops_data; |
| 879 | int ret; |
| 880 | |
| 881 | if (csd == NULL) |
| 882 | return; |
| 883 | |
| 884 | /* free the peer */ |
| 885 | if ((ret = rpma_peer_delete(&csd->peer))) |
| 886 | librpma_td_verror(td, ret, "rpma_peer_delete"); |
| 887 | |
| 888 | free(csd); |
| 889 | } |
| 890 | |
| 891 | int librpma_fio_server_open_file(struct thread_data *td, struct fio_file *f, |
| 892 | struct rpma_conn_cfg *cfg) |
| 893 | { |
| 894 | struct librpma_fio_server_data *csd = td->io_ops_data; |
| 895 | struct librpma_fio_options_values *o = td->eo; |
| 896 | enum rpma_conn_event conn_event = RPMA_CONN_UNDEFINED; |
| 897 | struct librpma_fio_workspace ws = {0}; |
| 898 | struct rpma_conn_private_data pdata; |
| 899 | uint32_t max_msg_num; |
| 900 | struct rpma_conn_req *conn_req; |
| 901 | struct rpma_conn *conn; |
| 902 | struct rpma_mr_local *mr; |
| 903 | char port_td[LIBRPMA_FIO_PORT_STR_LEN_MAX]; |
| 904 | struct rpma_ep *ep; |
| 905 | size_t mem_size = td->o.size; |
| 906 | size_t mr_desc_size; |
| 907 | void *ws_ptr; |
| 908 | bool is_dram; |
| 909 | int usage_mem_type; |
| 910 | int ret; |
| 911 | |
| 912 | if (!f->file_name) { |
| 913 | log_err("fio: filename is not set\n"); |
| 914 | return -1; |
| 915 | } |
| 916 | |
| 917 | /* start a listening endpoint at addr:port */ |
| 918 | if (librpma_fio_td_port(o->port, td, port_td)) |
| 919 | return -1; |
| 920 | |
| 921 | if ((ret = rpma_ep_listen(csd->peer, o->server_ip, port_td, &ep))) { |
| 922 | librpma_td_verror(td, ret, "rpma_ep_listen"); |
| 923 | return -1; |
| 924 | } |
| 925 | |
| 926 | is_dram = !strcmp(f->file_name, "malloc"); |
| 927 | if (is_dram) { |
| 928 | /* allocation from DRAM using posix_memalign() */ |
| 929 | ws_ptr = librpma_fio_allocate_dram(td, mem_size, &csd->mem); |
| 930 | usage_mem_type = RPMA_MR_USAGE_FLUSH_TYPE_VISIBILITY; |
| 931 | } else { |
| 932 | /* allocation from PMEM using pmem_map_file() */ |
| 933 | ws_ptr = librpma_fio_allocate_pmem(td, f, mem_size, &csd->mem); |
| 934 | usage_mem_type = RPMA_MR_USAGE_FLUSH_TYPE_PERSISTENT; |
| 935 | } |
| 936 | |
| 937 | if (ws_ptr == NULL) |
| 938 | goto err_ep_shutdown; |
| 939 | |
| 940 | f->real_file_size = mem_size; |
| 941 | |
| 942 | if ((ret = rpma_mr_reg(csd->peer, ws_ptr, mem_size, |
| 943 | RPMA_MR_USAGE_READ_DST | RPMA_MR_USAGE_READ_SRC | |
| 944 | RPMA_MR_USAGE_WRITE_DST | RPMA_MR_USAGE_WRITE_SRC | |
| 945 | usage_mem_type, &mr))) { |
| 946 | librpma_td_verror(td, ret, "rpma_mr_reg"); |
| 947 | goto err_free; |
| 948 | } |
| 949 | |
| 950 | if (!is_dram && f->filetype == FIO_TYPE_FILE) { |
| 951 | ret = rpma_mr_advise(mr, 0, mem_size, |
| 952 | IBV_ADVISE_MR_ADVICE_PREFETCH_WRITE, |
| 953 | IBV_ADVISE_MR_FLAG_FLUSH); |
| 954 | if (ret) { |
| 955 | librpma_td_verror(td, ret, "rpma_mr_advise"); |
| 956 | /* an invalid argument is an error */ |
| 957 | if (ret == RPMA_E_INVAL) |
| 958 | goto err_mr_dereg; |
| 959 | |
| 960 | /* log_err used instead of log_info to avoid corruption of the JSON output */ |
| 961 | log_err("Note: having rpma_mr_advise(3) failed because of RPMA_E_NOSUPP or RPMA_E_PROVIDER may come with a performance penalty, but it is not a blocker for running the benchmark.\n"); |
| 962 | } |
| 963 | } |
| 964 | |
| 965 | /* get size of the memory region's descriptor */ |
| 966 | if ((ret = rpma_mr_get_descriptor_size(mr, &mr_desc_size))) { |
| 967 | librpma_td_verror(td, ret, "rpma_mr_get_descriptor_size"); |
| 968 | goto err_mr_dereg; |
| 969 | } |
| 970 | |
| 971 | /* verify size of the memory region's descriptor */ |
| 972 | if (mr_desc_size > LIBRPMA_FIO_DESCRIPTOR_MAX_SIZE) { |
| 973 | log_err( |
| 974 | "size of the memory region's descriptor is too big (max=%i)\n", |
| 975 | LIBRPMA_FIO_DESCRIPTOR_MAX_SIZE); |
| 976 | goto err_mr_dereg; |
| 977 | } |
| 978 | |
| 979 | /* get the memory region's descriptor */ |
| 980 | if ((ret = rpma_mr_get_descriptor(mr, &ws.descriptor[0]))) { |
| 981 | librpma_td_verror(td, ret, "rpma_mr_get_descriptor"); |
| 982 | goto err_mr_dereg; |
| 983 | } |
| 984 | |
| 985 | if (cfg != NULL) { |
| 986 | if ((ret = rpma_conn_cfg_get_rq_size(cfg, &max_msg_num))) { |
| 987 | librpma_td_verror(td, ret, "rpma_conn_cfg_get_rq_size"); |
| 988 | goto err_mr_dereg; |
| 989 | } |
| 990 | |
| 991 | /* verify whether iodepth fits into uint16_t */ |
| 992 | if (max_msg_num > UINT16_MAX) { |
| 993 | log_err("fio: iodepth too big (%u > %u)\n", |
| 994 | max_msg_num, UINT16_MAX); |
| 995 | return -1; |
| 996 | } |
| 997 | |
| 998 | ws.max_msg_num = max_msg_num; |
| 999 | } |
| 1000 | |
| 1001 | /* prepare a workspace description */ |
| 1002 | ws.direct_write_to_pmem = o->direct_write_to_pmem; |
| 1003 | ws.mr_desc_size = mr_desc_size; |
| 1004 | pdata.ptr = &ws; |
| 1005 | pdata.len = sizeof(ws); |
| 1006 | |
| 1007 | /* receive an incoming connection request */ |
| 1008 | if ((ret = rpma_ep_next_conn_req(ep, cfg, &conn_req))) { |
| 1009 | librpma_td_verror(td, ret, "rpma_ep_next_conn_req"); |
| 1010 | goto err_mr_dereg; |
| 1011 | } |
| 1012 | |
| 1013 | if (csd->prepare_connection && csd->prepare_connection(td, conn_req)) |
| 1014 | goto err_req_delete; |
| 1015 | |
| 1016 | /* accept the connection request and obtain the connection object */ |
| 1017 | if ((ret = rpma_conn_req_connect(&conn_req, &pdata, &conn))) { |
| 1018 | librpma_td_verror(td, ret, "rpma_conn_req_connect"); |
| 1019 | goto err_req_delete; |
| 1020 | } |
| 1021 | |
| 1022 | /* wait for the connection to be established */ |
| 1023 | if ((ret = rpma_conn_next_event(conn, &conn_event))) { |
| 1024 | librpma_td_verror(td, ret, "rpma_conn_next_event"); |
| 1025 | goto err_conn_delete; |
| 1026 | } else if (conn_event != RPMA_CONN_ESTABLISHED) { |
| 1027 | log_err("rpma_conn_next_event returned an unexptected event\n"); |
| 1028 | goto err_conn_delete; |
| 1029 | } |
| 1030 | |
| 1031 | /* end-point is no longer needed */ |
| 1032 | (void) rpma_ep_shutdown(&ep); |
| 1033 | |
| 1034 | csd->ws_mr = mr; |
| 1035 | csd->ws_ptr = ws_ptr; |
| 1036 | csd->conn = conn; |
| 1037 | |
| 1038 | /* get the connection's main CQ */ |
| 1039 | if ((ret = rpma_conn_get_cq(csd->conn, &csd->cq))) { |
| 1040 | librpma_td_verror(td, ret, "rpma_conn_get_cq"); |
| 1041 | goto err_conn_delete; |
| 1042 | } |
| 1043 | |
| 1044 | return 0; |
| 1045 | |
| 1046 | err_conn_delete: |
| 1047 | (void) rpma_conn_delete(&conn); |
| 1048 | |
| 1049 | err_req_delete: |
| 1050 | (void) rpma_conn_req_delete(&conn_req); |
| 1051 | |
| 1052 | err_mr_dereg: |
| 1053 | (void) rpma_mr_dereg(&mr); |
| 1054 | |
| 1055 | err_free: |
| 1056 | librpma_fio_free(&csd->mem); |
| 1057 | |
| 1058 | err_ep_shutdown: |
| 1059 | (void) rpma_ep_shutdown(&ep); |
| 1060 | |
| 1061 | return -1; |
| 1062 | } |
| 1063 | |
| 1064 | int librpma_fio_server_close_file(struct thread_data *td, struct fio_file *f) |
| 1065 | { |
| 1066 | struct librpma_fio_server_data *csd = td->io_ops_data; |
| 1067 | enum rpma_conn_event conn_event = RPMA_CONN_UNDEFINED; |
| 1068 | int rv = 0; |
| 1069 | int ret; |
| 1070 | |
| 1071 | /* wait for the connection to be closed */ |
| 1072 | ret = rpma_conn_next_event(csd->conn, &conn_event); |
| 1073 | if (!ret && conn_event != RPMA_CONN_CLOSED) { |
| 1074 | log_err("rpma_conn_next_event returned an unexptected event\n"); |
| 1075 | rv = -1; |
| 1076 | } |
| 1077 | |
| 1078 | if ((ret = rpma_conn_disconnect(csd->conn))) { |
| 1079 | librpma_td_verror(td, ret, "rpma_conn_disconnect"); |
| 1080 | rv = -1; |
| 1081 | } |
| 1082 | |
| 1083 | if ((ret = rpma_conn_delete(&csd->conn))) { |
| 1084 | librpma_td_verror(td, ret, "rpma_conn_delete"); |
| 1085 | rv = -1; |
| 1086 | } |
| 1087 | |
| 1088 | if ((ret = rpma_mr_dereg(&csd->ws_mr))) { |
| 1089 | librpma_td_verror(td, ret, "rpma_mr_dereg"); |
| 1090 | rv = -1; |
| 1091 | } |
| 1092 | |
| 1093 | librpma_fio_free(&csd->mem); |
| 1094 | |
| 1095 | return rv; |
| 1096 | } |