1 /*- 2 * BSD LICENSE 3 * 4 * Copyright (c) Intel Corporation. All rights reserved. 5 * Copyright (c) 2019 Mellanox Technologies LTD. All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 11 * * Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * * Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in 15 * the documentation and/or other materials provided with the 16 * distribution. 17 * * Neither the name of Intel Corporation nor the names of its 18 * contributors may be used to endorse or promote products derived 19 * from this software without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 22 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 24 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 25 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 26 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 27 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 28 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 29 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 30 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 31 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 32 */ 33 34 #include "spdk/stdinc.h" 35 36 #include "nvmf_internal.h" 37 38 #include "spdk/bdev.h" 39 #include "spdk/endian.h" 40 #include "spdk/thread.h" 41 #include "spdk/likely.h" 42 #include "spdk/nvme.h" 43 #include "spdk/nvmf_cmd.h" 44 #include "spdk/nvmf_spec.h" 45 #include "spdk/trace.h" 46 #include "spdk/scsi_spec.h" 47 #include "spdk/string.h" 48 #include "spdk/util.h" 49 50 #include "spdk/log.h" 51 52 static bool 53 nvmf_subsystem_bdev_io_type_supported(struct spdk_nvmf_subsystem *subsystem, 54 enum spdk_bdev_io_type io_type) 55 { 56 struct spdk_nvmf_ns *ns; 57 58 for (ns = spdk_nvmf_subsystem_get_first_ns(subsystem); ns != NULL; 59 ns = spdk_nvmf_subsystem_get_next_ns(subsystem, ns)) { 60 if (ns->bdev == NULL) { 61 continue; 62 } 63 64 if (!spdk_bdev_io_type_supported(ns->bdev, io_type)) { 65 SPDK_DEBUGLOG(nvmf, 66 "Subsystem %s namespace %u (%s) does not support io_type %d\n", 67 spdk_nvmf_subsystem_get_nqn(subsystem), 68 ns->opts.nsid, spdk_bdev_get_name(ns->bdev), (int)io_type); 69 return false; 70 } 71 } 72 73 SPDK_DEBUGLOG(nvmf, "All devices in Subsystem %s support io_type %d\n", 74 spdk_nvmf_subsystem_get_nqn(subsystem), (int)io_type); 75 return true; 76 } 77 78 bool 79 nvmf_ctrlr_dsm_supported(struct spdk_nvmf_ctrlr *ctrlr) 80 { 81 return nvmf_subsystem_bdev_io_type_supported(ctrlr->subsys, SPDK_BDEV_IO_TYPE_UNMAP); 82 } 83 84 bool 85 nvmf_ctrlr_write_zeroes_supported(struct spdk_nvmf_ctrlr *ctrlr) 86 { 87 return nvmf_subsystem_bdev_io_type_supported(ctrlr->subsys, SPDK_BDEV_IO_TYPE_WRITE_ZEROES); 88 } 89 90 static void 91 nvmf_bdev_ctrlr_complete_cmd(struct spdk_bdev_io *bdev_io, bool success, 92 void *cb_arg) 93 { 94 struct spdk_nvmf_request *req = cb_arg; 95 struct spdk_nvme_cpl *response = &req->rsp->nvme_cpl; 96 int first_sc = 0, first_sct = 0, second_sc = 0, second_sct = 0; 97 uint32_t cdw0 = 0; 98 struct spdk_nvmf_request *first_req = req->first_fused_req; 99 100 if (spdk_unlikely(first_req != NULL)) { 101 /* fused commands - get status for both operations */ 102 struct spdk_nvme_cpl *fused_response = &first_req->rsp->nvme_cpl; 103 104 spdk_bdev_io_get_nvme_fused_status(bdev_io, &cdw0, &second_sct, &second_sc, &first_sct, &first_sc); 105 fused_response->cdw0 = cdw0; 106 fused_response->status.sc = second_sc; 107 fused_response->status.sct = second_sct; 108 109 /* first request should be completed */ 110 spdk_nvmf_request_complete(first_req); 111 req->first_fused_req = NULL; 112 } else { 113 spdk_bdev_io_get_nvme_status(bdev_io, &cdw0, &first_sct, &first_sc); 114 } 115 116 response->cdw0 = cdw0; 117 response->status.sc = first_sc; 118 response->status.sct = first_sct; 119 120 spdk_nvmf_request_complete(req); 121 spdk_bdev_free_io(bdev_io); 122 } 123 124 static void 125 nvmf_bdev_ctrlr_complete_admin_cmd(struct spdk_bdev_io *bdev_io, bool success, 126 void *cb_arg) 127 { 128 struct spdk_nvmf_request *req = cb_arg; 129 130 if (req->cmd_cb_fn) { 131 req->cmd_cb_fn(req); 132 } 133 134 nvmf_bdev_ctrlr_complete_cmd(bdev_io, success, req); 135 } 136 137 void 138 nvmf_bdev_ctrlr_identify_ns(struct spdk_nvmf_ns *ns, struct spdk_nvme_ns_data *nsdata, 139 bool dif_insert_or_strip) 140 { 141 struct spdk_bdev *bdev = ns->bdev; 142 uint64_t num_blocks; 143 uint32_t phys_blocklen; 144 145 num_blocks = spdk_bdev_get_num_blocks(bdev); 146 147 nsdata->nsze = num_blocks; 148 nsdata->ncap = num_blocks; 149 nsdata->nuse = num_blocks; 150 nsdata->nlbaf = 0; 151 nsdata->flbas.format = 0; 152 nsdata->nacwu = spdk_bdev_get_acwu(bdev); 153 if (!dif_insert_or_strip) { 154 nsdata->lbaf[0].ms = spdk_bdev_get_md_size(bdev); 155 nsdata->lbaf[0].lbads = spdk_u32log2(spdk_bdev_get_block_size(bdev)); 156 if (nsdata->lbaf[0].ms != 0) { 157 nsdata->flbas.extended = 1; 158 nsdata->mc.extended = 1; 159 nsdata->mc.pointer = 0; 160 nsdata->dps.md_start = spdk_bdev_is_dif_head_of_md(bdev); 161 162 switch (spdk_bdev_get_dif_type(bdev)) { 163 case SPDK_DIF_TYPE1: 164 nsdata->dpc.pit1 = 1; 165 nsdata->dps.pit = SPDK_NVME_FMT_NVM_PROTECTION_TYPE1; 166 break; 167 case SPDK_DIF_TYPE2: 168 nsdata->dpc.pit2 = 1; 169 nsdata->dps.pit = SPDK_NVME_FMT_NVM_PROTECTION_TYPE2; 170 break; 171 case SPDK_DIF_TYPE3: 172 nsdata->dpc.pit3 = 1; 173 nsdata->dps.pit = SPDK_NVME_FMT_NVM_PROTECTION_TYPE3; 174 break; 175 default: 176 SPDK_DEBUGLOG(nvmf, "Protection Disabled\n"); 177 nsdata->dps.pit = SPDK_NVME_FMT_NVM_PROTECTION_DISABLE; 178 break; 179 } 180 } 181 } else { 182 nsdata->lbaf[0].ms = 0; 183 nsdata->lbaf[0].lbads = spdk_u32log2(spdk_bdev_get_data_block_size(bdev)); 184 } 185 186 phys_blocklen = spdk_bdev_get_physical_block_size(bdev); 187 assert(phys_blocklen > 0); 188 /* Linux driver uses min(nawupf, npwg) to set physical_block_size */ 189 nsdata->nsfeat.optperf = 1; 190 nsdata->nsfeat.ns_atomic_write_unit = 1; 191 nsdata->npwg = (phys_blocklen >> nsdata->lbaf[0].lbads) - 1; 192 nsdata->nawupf = nsdata->npwg; 193 194 nsdata->noiob = spdk_bdev_get_optimal_io_boundary(bdev); 195 nsdata->nmic.can_share = 1; 196 if (ns->ptpl_file != NULL) { 197 nsdata->nsrescap.rescap.persist = 1; 198 } 199 nsdata->nsrescap.rescap.write_exclusive = 1; 200 nsdata->nsrescap.rescap.exclusive_access = 1; 201 nsdata->nsrescap.rescap.write_exclusive_reg_only = 1; 202 nsdata->nsrescap.rescap.exclusive_access_reg_only = 1; 203 nsdata->nsrescap.rescap.write_exclusive_all_reg = 1; 204 nsdata->nsrescap.rescap.exclusive_access_all_reg = 1; 205 nsdata->nsrescap.rescap.ignore_existing_key = 1; 206 207 SPDK_STATIC_ASSERT(sizeof(nsdata->nguid) == sizeof(ns->opts.nguid), "size mismatch"); 208 memcpy(nsdata->nguid, ns->opts.nguid, sizeof(nsdata->nguid)); 209 210 SPDK_STATIC_ASSERT(sizeof(nsdata->eui64) == sizeof(ns->opts.eui64), "size mismatch"); 211 memcpy(&nsdata->eui64, ns->opts.eui64, sizeof(nsdata->eui64)); 212 } 213 214 static void 215 nvmf_bdev_ctrlr_get_rw_params(const struct spdk_nvme_cmd *cmd, uint64_t *start_lba, 216 uint64_t *num_blocks) 217 { 218 /* SLBA: CDW10 and CDW11 */ 219 *start_lba = from_le64(&cmd->cdw10); 220 221 /* NLB: CDW12 bits 15:00, 0's based */ 222 *num_blocks = (from_le32(&cmd->cdw12) & 0xFFFFu) + 1; 223 } 224 225 static bool 226 nvmf_bdev_ctrlr_lba_in_range(uint64_t bdev_num_blocks, uint64_t io_start_lba, 227 uint64_t io_num_blocks) 228 { 229 if (io_start_lba + io_num_blocks > bdev_num_blocks || 230 io_start_lba + io_num_blocks < io_start_lba) { 231 return false; 232 } 233 234 return true; 235 } 236 237 static void 238 nvmf_ctrlr_process_io_cmd_resubmit(void *arg) 239 { 240 struct spdk_nvmf_request *req = arg; 241 242 nvmf_ctrlr_process_io_cmd(req); 243 } 244 245 static void 246 nvmf_ctrlr_process_admin_cmd_resubmit(void *arg) 247 { 248 struct spdk_nvmf_request *req = arg; 249 250 nvmf_ctrlr_process_admin_cmd(req); 251 } 252 253 static void 254 nvmf_bdev_ctrl_queue_io(struct spdk_nvmf_request *req, struct spdk_bdev *bdev, 255 struct spdk_io_channel *ch, spdk_bdev_io_wait_cb cb_fn, void *cb_arg) 256 { 257 int rc; 258 259 req->bdev_io_wait.bdev = bdev; 260 req->bdev_io_wait.cb_fn = cb_fn; 261 req->bdev_io_wait.cb_arg = cb_arg; 262 263 rc = spdk_bdev_queue_io_wait(bdev, ch, &req->bdev_io_wait); 264 if (rc != 0) { 265 assert(false); 266 } 267 req->qpair->group->stat.pending_bdev_io++; 268 } 269 270 int 271 nvmf_bdev_ctrlr_read_cmd(struct spdk_bdev *bdev, struct spdk_bdev_desc *desc, 272 struct spdk_io_channel *ch, struct spdk_nvmf_request *req) 273 { 274 uint64_t bdev_num_blocks = spdk_bdev_get_num_blocks(bdev); 275 uint32_t block_size = spdk_bdev_get_block_size(bdev); 276 struct spdk_nvme_cmd *cmd = &req->cmd->nvme_cmd; 277 struct spdk_nvme_cpl *rsp = &req->rsp->nvme_cpl; 278 uint64_t start_lba; 279 uint64_t num_blocks; 280 int rc; 281 282 nvmf_bdev_ctrlr_get_rw_params(cmd, &start_lba, &num_blocks); 283 284 if (spdk_unlikely(!nvmf_bdev_ctrlr_lba_in_range(bdev_num_blocks, start_lba, num_blocks))) { 285 SPDK_ERRLOG("end of media\n"); 286 rsp->status.sct = SPDK_NVME_SCT_GENERIC; 287 rsp->status.sc = SPDK_NVME_SC_LBA_OUT_OF_RANGE; 288 return SPDK_NVMF_REQUEST_EXEC_STATUS_COMPLETE; 289 } 290 291 if (spdk_unlikely(num_blocks * block_size > req->length)) { 292 SPDK_ERRLOG("Read NLB %" PRIu64 " * block size %" PRIu32 " > SGL length %" PRIu32 "\n", 293 num_blocks, block_size, req->length); 294 rsp->status.sct = SPDK_NVME_SCT_GENERIC; 295 rsp->status.sc = SPDK_NVME_SC_DATA_SGL_LENGTH_INVALID; 296 return SPDK_NVMF_REQUEST_EXEC_STATUS_COMPLETE; 297 } 298 299 rc = spdk_bdev_readv_blocks(desc, ch, req->iov, req->iovcnt, start_lba, num_blocks, 300 nvmf_bdev_ctrlr_complete_cmd, req); 301 if (spdk_unlikely(rc)) { 302 if (rc == -ENOMEM) { 303 nvmf_bdev_ctrl_queue_io(req, bdev, ch, nvmf_ctrlr_process_io_cmd_resubmit, req); 304 return SPDK_NVMF_REQUEST_EXEC_STATUS_ASYNCHRONOUS; 305 } 306 rsp->status.sct = SPDK_NVME_SCT_GENERIC; 307 rsp->status.sc = SPDK_NVME_SC_INTERNAL_DEVICE_ERROR; 308 return SPDK_NVMF_REQUEST_EXEC_STATUS_COMPLETE; 309 } 310 311 return SPDK_NVMF_REQUEST_EXEC_STATUS_ASYNCHRONOUS; 312 } 313 314 int 315 nvmf_bdev_ctrlr_write_cmd(struct spdk_bdev *bdev, struct spdk_bdev_desc *desc, 316 struct spdk_io_channel *ch, struct spdk_nvmf_request *req) 317 { 318 uint64_t bdev_num_blocks = spdk_bdev_get_num_blocks(bdev); 319 uint32_t block_size = spdk_bdev_get_block_size(bdev); 320 struct spdk_nvme_cmd *cmd = &req->cmd->nvme_cmd; 321 struct spdk_nvme_cpl *rsp = &req->rsp->nvme_cpl; 322 uint64_t start_lba; 323 uint64_t num_blocks; 324 int rc; 325 326 nvmf_bdev_ctrlr_get_rw_params(cmd, &start_lba, &num_blocks); 327 328 if (spdk_unlikely(!nvmf_bdev_ctrlr_lba_in_range(bdev_num_blocks, start_lba, num_blocks))) { 329 SPDK_ERRLOG("end of media\n"); 330 rsp->status.sct = SPDK_NVME_SCT_GENERIC; 331 rsp->status.sc = SPDK_NVME_SC_LBA_OUT_OF_RANGE; 332 return SPDK_NVMF_REQUEST_EXEC_STATUS_COMPLETE; 333 } 334 335 if (spdk_unlikely(num_blocks * block_size > req->length)) { 336 SPDK_ERRLOG("Write NLB %" PRIu64 " * block size %" PRIu32 " > SGL length %" PRIu32 "\n", 337 num_blocks, block_size, req->length); 338 rsp->status.sct = SPDK_NVME_SCT_GENERIC; 339 rsp->status.sc = SPDK_NVME_SC_DATA_SGL_LENGTH_INVALID; 340 return SPDK_NVMF_REQUEST_EXEC_STATUS_COMPLETE; 341 } 342 343 rc = spdk_bdev_writev_blocks(desc, ch, req->iov, req->iovcnt, start_lba, num_blocks, 344 nvmf_bdev_ctrlr_complete_cmd, req); 345 if (spdk_unlikely(rc)) { 346 if (rc == -ENOMEM) { 347 nvmf_bdev_ctrl_queue_io(req, bdev, ch, nvmf_ctrlr_process_io_cmd_resubmit, req); 348 return SPDK_NVMF_REQUEST_EXEC_STATUS_ASYNCHRONOUS; 349 } 350 rsp->status.sct = SPDK_NVME_SCT_GENERIC; 351 rsp->status.sc = SPDK_NVME_SC_INTERNAL_DEVICE_ERROR; 352 return SPDK_NVMF_REQUEST_EXEC_STATUS_COMPLETE; 353 } 354 355 return SPDK_NVMF_REQUEST_EXEC_STATUS_ASYNCHRONOUS; 356 } 357 358 int 359 nvmf_bdev_ctrlr_compare_cmd(struct spdk_bdev *bdev, struct spdk_bdev_desc *desc, 360 struct spdk_io_channel *ch, struct spdk_nvmf_request *req) 361 { 362 uint64_t bdev_num_blocks = spdk_bdev_get_num_blocks(bdev); 363 uint32_t block_size = spdk_bdev_get_block_size(bdev); 364 struct spdk_nvme_cmd *cmd = &req->cmd->nvme_cmd; 365 struct spdk_nvme_cpl *rsp = &req->rsp->nvme_cpl; 366 uint64_t start_lba; 367 uint64_t num_blocks; 368 int rc; 369 370 nvmf_bdev_ctrlr_get_rw_params(cmd, &start_lba, &num_blocks); 371 372 if (spdk_unlikely(!nvmf_bdev_ctrlr_lba_in_range(bdev_num_blocks, start_lba, num_blocks))) { 373 SPDK_ERRLOG("end of media\n"); 374 rsp->status.sct = SPDK_NVME_SCT_GENERIC; 375 rsp->status.sc = SPDK_NVME_SC_LBA_OUT_OF_RANGE; 376 return SPDK_NVMF_REQUEST_EXEC_STATUS_COMPLETE; 377 } 378 379 if (spdk_unlikely(num_blocks * block_size > req->length)) { 380 SPDK_ERRLOG("Compare NLB %" PRIu64 " * block size %" PRIu32 " > SGL length %" PRIu32 "\n", 381 num_blocks, block_size, req->length); 382 rsp->status.sct = SPDK_NVME_SCT_GENERIC; 383 rsp->status.sc = SPDK_NVME_SC_DATA_SGL_LENGTH_INVALID; 384 return SPDK_NVMF_REQUEST_EXEC_STATUS_COMPLETE; 385 } 386 387 rc = spdk_bdev_comparev_blocks(desc, ch, req->iov, req->iovcnt, start_lba, num_blocks, 388 nvmf_bdev_ctrlr_complete_cmd, req); 389 if (spdk_unlikely(rc)) { 390 if (rc == -ENOMEM) { 391 nvmf_bdev_ctrl_queue_io(req, bdev, ch, nvmf_ctrlr_process_io_cmd_resubmit, req); 392 return SPDK_NVMF_REQUEST_EXEC_STATUS_ASYNCHRONOUS; 393 } 394 rsp->status.sct = SPDK_NVME_SCT_GENERIC; 395 rsp->status.sc = SPDK_NVME_SC_INTERNAL_DEVICE_ERROR; 396 return SPDK_NVMF_REQUEST_EXEC_STATUS_COMPLETE; 397 } 398 399 return SPDK_NVMF_REQUEST_EXEC_STATUS_ASYNCHRONOUS; 400 } 401 402 int 403 nvmf_bdev_ctrlr_compare_and_write_cmd(struct spdk_bdev *bdev, struct spdk_bdev_desc *desc, 404 struct spdk_io_channel *ch, struct spdk_nvmf_request *cmp_req, struct spdk_nvmf_request *write_req) 405 { 406 uint64_t bdev_num_blocks = spdk_bdev_get_num_blocks(bdev); 407 uint32_t block_size = spdk_bdev_get_block_size(bdev); 408 struct spdk_nvme_cmd *cmp_cmd = &cmp_req->cmd->nvme_cmd; 409 struct spdk_nvme_cmd *write_cmd = &write_req->cmd->nvme_cmd; 410 struct spdk_nvme_cpl *rsp = &write_req->rsp->nvme_cpl; 411 uint64_t write_start_lba, cmp_start_lba; 412 uint64_t write_num_blocks, cmp_num_blocks; 413 int rc; 414 415 nvmf_bdev_ctrlr_get_rw_params(cmp_cmd, &cmp_start_lba, &cmp_num_blocks); 416 nvmf_bdev_ctrlr_get_rw_params(write_cmd, &write_start_lba, &write_num_blocks); 417 418 if (spdk_unlikely(write_start_lba != cmp_start_lba || write_num_blocks != cmp_num_blocks)) { 419 SPDK_ERRLOG("Fused command start lba / num blocks mismatch\n"); 420 rsp->status.sct = SPDK_NVME_SCT_GENERIC; 421 rsp->status.sc = SPDK_NVME_SC_INVALID_FIELD; 422 return SPDK_NVMF_REQUEST_EXEC_STATUS_COMPLETE; 423 } 424 425 if (spdk_unlikely(!nvmf_bdev_ctrlr_lba_in_range(bdev_num_blocks, write_start_lba, 426 write_num_blocks))) { 427 SPDK_ERRLOG("end of media\n"); 428 rsp->status.sct = SPDK_NVME_SCT_GENERIC; 429 rsp->status.sc = SPDK_NVME_SC_LBA_OUT_OF_RANGE; 430 return SPDK_NVMF_REQUEST_EXEC_STATUS_COMPLETE; 431 } 432 433 if (spdk_unlikely(write_num_blocks * block_size > write_req->length)) { 434 SPDK_ERRLOG("Write NLB %" PRIu64 " * block size %" PRIu32 " > SGL length %" PRIu32 "\n", 435 write_num_blocks, block_size, write_req->length); 436 rsp->status.sct = SPDK_NVME_SCT_GENERIC; 437 rsp->status.sc = SPDK_NVME_SC_DATA_SGL_LENGTH_INVALID; 438 return SPDK_NVMF_REQUEST_EXEC_STATUS_COMPLETE; 439 } 440 441 rc = spdk_bdev_comparev_and_writev_blocks(desc, ch, cmp_req->iov, cmp_req->iovcnt, write_req->iov, 442 write_req->iovcnt, write_start_lba, write_num_blocks, nvmf_bdev_ctrlr_complete_cmd, write_req); 443 if (spdk_unlikely(rc)) { 444 if (rc == -ENOMEM) { 445 nvmf_bdev_ctrl_queue_io(cmp_req, bdev, ch, nvmf_ctrlr_process_io_cmd_resubmit, cmp_req); 446 nvmf_bdev_ctrl_queue_io(write_req, bdev, ch, nvmf_ctrlr_process_io_cmd_resubmit, write_req); 447 return SPDK_NVMF_REQUEST_EXEC_STATUS_ASYNCHRONOUS; 448 } 449 rsp->status.sct = SPDK_NVME_SCT_GENERIC; 450 rsp->status.sc = SPDK_NVME_SC_INTERNAL_DEVICE_ERROR; 451 return SPDK_NVMF_REQUEST_EXEC_STATUS_COMPLETE; 452 } 453 454 return SPDK_NVMF_REQUEST_EXEC_STATUS_ASYNCHRONOUS; 455 } 456 457 int 458 nvmf_bdev_ctrlr_write_zeroes_cmd(struct spdk_bdev *bdev, struct spdk_bdev_desc *desc, 459 struct spdk_io_channel *ch, struct spdk_nvmf_request *req) 460 { 461 uint64_t bdev_num_blocks = spdk_bdev_get_num_blocks(bdev); 462 struct spdk_nvme_cmd *cmd = &req->cmd->nvme_cmd; 463 struct spdk_nvme_cpl *rsp = &req->rsp->nvme_cpl; 464 uint64_t start_lba; 465 uint64_t num_blocks; 466 int rc; 467 468 nvmf_bdev_ctrlr_get_rw_params(cmd, &start_lba, &num_blocks); 469 470 if (spdk_unlikely(!nvmf_bdev_ctrlr_lba_in_range(bdev_num_blocks, start_lba, num_blocks))) { 471 SPDK_ERRLOG("end of media\n"); 472 rsp->status.sct = SPDK_NVME_SCT_GENERIC; 473 rsp->status.sc = SPDK_NVME_SC_LBA_OUT_OF_RANGE; 474 return SPDK_NVMF_REQUEST_EXEC_STATUS_COMPLETE; 475 } 476 477 rc = spdk_bdev_write_zeroes_blocks(desc, ch, start_lba, num_blocks, 478 nvmf_bdev_ctrlr_complete_cmd, req); 479 if (spdk_unlikely(rc)) { 480 if (rc == -ENOMEM) { 481 nvmf_bdev_ctrl_queue_io(req, bdev, ch, nvmf_ctrlr_process_io_cmd_resubmit, req); 482 return SPDK_NVMF_REQUEST_EXEC_STATUS_ASYNCHRONOUS; 483 } 484 rsp->status.sct = SPDK_NVME_SCT_GENERIC; 485 rsp->status.sc = SPDK_NVME_SC_INTERNAL_DEVICE_ERROR; 486 return SPDK_NVMF_REQUEST_EXEC_STATUS_COMPLETE; 487 } 488 489 return SPDK_NVMF_REQUEST_EXEC_STATUS_ASYNCHRONOUS; 490 } 491 492 int 493 nvmf_bdev_ctrlr_flush_cmd(struct spdk_bdev *bdev, struct spdk_bdev_desc *desc, 494 struct spdk_io_channel *ch, struct spdk_nvmf_request *req) 495 { 496 struct spdk_nvme_cpl *response = &req->rsp->nvme_cpl; 497 int rc; 498 499 /* As for NVMeoF controller, SPDK always set volatile write 500 * cache bit to 1, return success for those block devices 501 * which can't support FLUSH command. 502 */ 503 if (!spdk_bdev_io_type_supported(bdev, SPDK_BDEV_IO_TYPE_FLUSH)) { 504 response->status.sct = SPDK_NVME_SCT_GENERIC; 505 response->status.sc = SPDK_NVME_SC_SUCCESS; 506 return SPDK_NVMF_REQUEST_EXEC_STATUS_COMPLETE; 507 } 508 509 rc = spdk_bdev_flush_blocks(desc, ch, 0, spdk_bdev_get_num_blocks(bdev), 510 nvmf_bdev_ctrlr_complete_cmd, req); 511 if (spdk_unlikely(rc)) { 512 if (rc == -ENOMEM) { 513 nvmf_bdev_ctrl_queue_io(req, bdev, ch, nvmf_ctrlr_process_io_cmd_resubmit, req); 514 return SPDK_NVMF_REQUEST_EXEC_STATUS_ASYNCHRONOUS; 515 } 516 response->status.sc = SPDK_NVME_SC_INTERNAL_DEVICE_ERROR; 517 return SPDK_NVMF_REQUEST_EXEC_STATUS_COMPLETE; 518 } 519 return SPDK_NVMF_REQUEST_EXEC_STATUS_ASYNCHRONOUS; 520 } 521 522 struct nvmf_bdev_ctrlr_unmap { 523 struct spdk_nvmf_request *req; 524 uint32_t count; 525 struct spdk_bdev_desc *desc; 526 struct spdk_bdev *bdev; 527 struct spdk_io_channel *ch; 528 uint32_t range_index; 529 }; 530 531 static void 532 nvmf_bdev_ctrlr_unmap_cpl(struct spdk_bdev_io *bdev_io, bool success, 533 void *cb_arg) 534 { 535 struct nvmf_bdev_ctrlr_unmap *unmap_ctx = cb_arg; 536 struct spdk_nvmf_request *req = unmap_ctx->req; 537 struct spdk_nvme_cpl *response = &req->rsp->nvme_cpl; 538 int sc, sct; 539 uint32_t cdw0; 540 541 unmap_ctx->count--; 542 543 if (response->status.sct == SPDK_NVME_SCT_GENERIC && 544 response->status.sc == SPDK_NVME_SC_SUCCESS) { 545 spdk_bdev_io_get_nvme_status(bdev_io, &cdw0, &sct, &sc); 546 response->cdw0 = cdw0; 547 response->status.sc = sc; 548 response->status.sct = sct; 549 } 550 551 if (unmap_ctx->count == 0) { 552 spdk_nvmf_request_complete(req); 553 free(unmap_ctx); 554 } 555 spdk_bdev_free_io(bdev_io); 556 } 557 558 static int 559 nvmf_bdev_ctrlr_unmap(struct spdk_bdev *bdev, struct spdk_bdev_desc *desc, 560 struct spdk_io_channel *ch, struct spdk_nvmf_request *req, 561 struct nvmf_bdev_ctrlr_unmap *unmap_ctx); 562 static void 563 nvmf_bdev_ctrlr_unmap_resubmit(void *arg) 564 { 565 struct nvmf_bdev_ctrlr_unmap *unmap_ctx = arg; 566 struct spdk_nvmf_request *req = unmap_ctx->req; 567 struct spdk_bdev_desc *desc = unmap_ctx->desc; 568 struct spdk_bdev *bdev = unmap_ctx->bdev; 569 struct spdk_io_channel *ch = unmap_ctx->ch; 570 571 nvmf_bdev_ctrlr_unmap(bdev, desc, ch, req, unmap_ctx); 572 } 573 574 static int 575 nvmf_bdev_ctrlr_unmap(struct spdk_bdev *bdev, struct spdk_bdev_desc *desc, 576 struct spdk_io_channel *ch, struct spdk_nvmf_request *req, 577 struct nvmf_bdev_ctrlr_unmap *unmap_ctx) 578 { 579 uint16_t nr, i; 580 struct spdk_nvme_cmd *cmd = &req->cmd->nvme_cmd; 581 struct spdk_nvme_cpl *response = &req->rsp->nvme_cpl; 582 struct spdk_nvme_dsm_range *dsm_range; 583 uint64_t lba; 584 uint32_t lba_count; 585 int rc; 586 587 nr = cmd->cdw10_bits.dsm.nr + 1; 588 if (nr * sizeof(struct spdk_nvme_dsm_range) > req->length) { 589 SPDK_ERRLOG("Dataset Management number of ranges > SGL length\n"); 590 response->status.sc = SPDK_NVME_SC_DATA_SGL_LENGTH_INVALID; 591 return SPDK_NVMF_REQUEST_EXEC_STATUS_COMPLETE; 592 } 593 594 if (unmap_ctx == NULL) { 595 unmap_ctx = calloc(1, sizeof(*unmap_ctx)); 596 if (!unmap_ctx) { 597 response->status.sc = SPDK_NVME_SC_INTERNAL_DEVICE_ERROR; 598 return SPDK_NVMF_REQUEST_EXEC_STATUS_COMPLETE; 599 } 600 601 unmap_ctx->req = req; 602 unmap_ctx->desc = desc; 603 unmap_ctx->ch = ch; 604 unmap_ctx->bdev = bdev; 605 606 response->status.sct = SPDK_NVME_SCT_GENERIC; 607 response->status.sc = SPDK_NVME_SC_SUCCESS; 608 } else { 609 unmap_ctx->count--; /* dequeued */ 610 } 611 612 dsm_range = (struct spdk_nvme_dsm_range *)req->data; 613 for (i = unmap_ctx->range_index; i < nr; i++) { 614 lba = dsm_range[i].starting_lba; 615 lba_count = dsm_range[i].length; 616 617 unmap_ctx->count++; 618 619 rc = spdk_bdev_unmap_blocks(desc, ch, lba, lba_count, 620 nvmf_bdev_ctrlr_unmap_cpl, unmap_ctx); 621 if (rc) { 622 if (rc == -ENOMEM) { 623 nvmf_bdev_ctrl_queue_io(req, bdev, ch, nvmf_bdev_ctrlr_unmap_resubmit, unmap_ctx); 624 /* Unmap was not yet submitted to bdev */ 625 /* unmap_ctx->count will be decremented when the request is dequeued */ 626 return SPDK_NVMF_REQUEST_EXEC_STATUS_ASYNCHRONOUS; 627 } 628 response->status.sc = SPDK_NVME_SC_INTERNAL_DEVICE_ERROR; 629 unmap_ctx->count--; 630 /* We can't return here - we may have to wait for any other 631 * unmaps already sent to complete */ 632 break; 633 } 634 unmap_ctx->range_index++; 635 } 636 637 if (unmap_ctx->count == 0) { 638 free(unmap_ctx); 639 return SPDK_NVMF_REQUEST_EXEC_STATUS_COMPLETE; 640 } 641 642 return SPDK_NVMF_REQUEST_EXEC_STATUS_ASYNCHRONOUS; 643 } 644 645 int 646 nvmf_bdev_ctrlr_dsm_cmd(struct spdk_bdev *bdev, struct spdk_bdev_desc *desc, 647 struct spdk_io_channel *ch, struct spdk_nvmf_request *req) 648 { 649 struct spdk_nvme_cmd *cmd = &req->cmd->nvme_cmd; 650 struct spdk_nvme_cpl *response = &req->rsp->nvme_cpl; 651 652 if (cmd->cdw11_bits.dsm.ad) { 653 return nvmf_bdev_ctrlr_unmap(bdev, desc, ch, req, NULL); 654 } 655 656 response->status.sct = SPDK_NVME_SCT_GENERIC; 657 response->status.sc = SPDK_NVME_SC_SUCCESS; 658 return SPDK_NVMF_REQUEST_EXEC_STATUS_COMPLETE; 659 } 660 661 int 662 nvmf_bdev_ctrlr_nvme_passthru_io(struct spdk_bdev *bdev, struct spdk_bdev_desc *desc, 663 struct spdk_io_channel *ch, struct spdk_nvmf_request *req) 664 { 665 int rc; 666 667 rc = spdk_bdev_nvme_io_passthru(desc, ch, &req->cmd->nvme_cmd, req->data, req->length, 668 nvmf_bdev_ctrlr_complete_cmd, req); 669 if (spdk_unlikely(rc)) { 670 if (rc == -ENOMEM) { 671 nvmf_bdev_ctrl_queue_io(req, bdev, ch, nvmf_ctrlr_process_io_cmd_resubmit, req); 672 return SPDK_NVMF_REQUEST_EXEC_STATUS_ASYNCHRONOUS; 673 } 674 req->rsp->nvme_cpl.status.sct = SPDK_NVME_SCT_GENERIC; 675 req->rsp->nvme_cpl.status.sc = SPDK_NVME_SC_INVALID_OPCODE; 676 return SPDK_NVMF_REQUEST_EXEC_STATUS_COMPLETE; 677 } 678 679 return SPDK_NVMF_REQUEST_EXEC_STATUS_ASYNCHRONOUS; 680 } 681 682 int 683 spdk_nvmf_bdev_ctrlr_nvme_passthru_admin(struct spdk_bdev *bdev, struct spdk_bdev_desc *desc, 684 struct spdk_io_channel *ch, struct spdk_nvmf_request *req, 685 spdk_nvmf_nvme_passthru_cmd_cb cb_fn) 686 { 687 int rc; 688 689 req->cmd_cb_fn = cb_fn; 690 691 rc = spdk_bdev_nvme_admin_passthru(desc, ch, &req->cmd->nvme_cmd, req->data, req->length, 692 nvmf_bdev_ctrlr_complete_admin_cmd, req); 693 if (spdk_unlikely(rc)) { 694 if (rc == -ENOMEM) { 695 nvmf_bdev_ctrl_queue_io(req, bdev, ch, nvmf_ctrlr_process_admin_cmd_resubmit, req); 696 return SPDK_NVMF_REQUEST_EXEC_STATUS_ASYNCHRONOUS; 697 } 698 req->rsp->nvme_cpl.status.sct = SPDK_NVME_SCT_GENERIC; 699 req->rsp->nvme_cpl.status.sc = SPDK_NVME_SC_INTERNAL_DEVICE_ERROR; 700 return SPDK_NVMF_REQUEST_EXEC_STATUS_COMPLETE; 701 } 702 703 return SPDK_NVMF_REQUEST_EXEC_STATUS_ASYNCHRONOUS; 704 } 705 706 static void 707 nvmf_bdev_ctrlr_complete_abort_cmd(struct spdk_bdev_io *bdev_io, bool success, void *cb_arg) 708 { 709 struct spdk_nvmf_request *req = cb_arg; 710 711 if (success) { 712 req->rsp->nvme_cpl.cdw0 &= ~1U; 713 } 714 715 spdk_nvmf_request_complete(req); 716 spdk_bdev_free_io(bdev_io); 717 } 718 719 int 720 spdk_nvmf_bdev_ctrlr_abort_cmd(struct spdk_bdev *bdev, struct spdk_bdev_desc *desc, 721 struct spdk_io_channel *ch, struct spdk_nvmf_request *req, 722 struct spdk_nvmf_request *req_to_abort) 723 { 724 int rc; 725 726 assert((req->rsp->nvme_cpl.cdw0 & 1U) != 0); 727 728 rc = spdk_bdev_abort(desc, ch, req_to_abort, nvmf_bdev_ctrlr_complete_abort_cmd, req); 729 if (spdk_likely(rc == 0)) { 730 return SPDK_NVMF_REQUEST_EXEC_STATUS_ASYNCHRONOUS; 731 } else if (rc == -ENOMEM) { 732 nvmf_bdev_ctrl_queue_io(req, bdev, ch, nvmf_ctrlr_process_admin_cmd_resubmit, req); 733 return SPDK_NVMF_REQUEST_EXEC_STATUS_ASYNCHRONOUS; 734 } else { 735 return SPDK_NVMF_REQUEST_EXEC_STATUS_COMPLETE; 736 } 737 } 738 739 bool 740 nvmf_bdev_ctrlr_get_dif_ctx(struct spdk_bdev *bdev, struct spdk_nvme_cmd *cmd, 741 struct spdk_dif_ctx *dif_ctx) 742 { 743 uint32_t init_ref_tag, dif_check_flags = 0; 744 int rc; 745 746 if (spdk_bdev_get_md_size(bdev) == 0) { 747 return false; 748 } 749 750 /* Initial Reference Tag is the lower 32 bits of the start LBA. */ 751 init_ref_tag = (uint32_t)from_le64(&cmd->cdw10); 752 753 if (spdk_bdev_is_dif_check_enabled(bdev, SPDK_DIF_CHECK_TYPE_REFTAG)) { 754 dif_check_flags |= SPDK_DIF_FLAGS_REFTAG_CHECK; 755 } 756 757 if (spdk_bdev_is_dif_check_enabled(bdev, SPDK_DIF_CHECK_TYPE_GUARD)) { 758 dif_check_flags |= SPDK_DIF_FLAGS_GUARD_CHECK; 759 } 760 761 rc = spdk_dif_ctx_init(dif_ctx, 762 spdk_bdev_get_block_size(bdev), 763 spdk_bdev_get_md_size(bdev), 764 spdk_bdev_is_md_interleaved(bdev), 765 spdk_bdev_is_dif_head_of_md(bdev), 766 spdk_bdev_get_dif_type(bdev), 767 dif_check_flags, 768 init_ref_tag, 0, 0, 0, 0); 769 770 return (rc == 0) ? true : false; 771 } 772