1/*- 2 * SPDX-License-Identifier: BSD-2-Clause 3 * 4 * Copyright (C) 2012-2016 Intel Corporation 5 * 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 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 26 * SUCH DAMAGE. 27 */ 28 29#include <sys/cdefs.h> 30#include "opt_nvme.h" 31 32#include <sys/param.h> 33#include <sys/systm.h> 34#include <sys/bus.h> 35#include <sys/sysctl.h> 36 37#include "nvme_private.h" 38 39#ifndef NVME_USE_NVD 40#define NVME_USE_NVD 0 41#endif 42 43int nvme_use_nvd = NVME_USE_NVD; 44bool nvme_verbose_cmd_dump = false; 45 46SYSCTL_NODE(_hw, OID_AUTO, nvme, CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 47 "NVMe sysctl tunables"); 48SYSCTL_INT(_hw_nvme, OID_AUTO, use_nvd, CTLFLAG_RDTUN, 49 &nvme_use_nvd, 1, "1 = Create NVD devices, 0 = Create NDA devices"); 50SYSCTL_BOOL(_hw_nvme, OID_AUTO, verbose_cmd_dump, CTLFLAG_RWTUN, 51 &nvme_verbose_cmd_dump, 0, 52 "enable verbose command printing when a command fails"); 53 54static void 55nvme_dump_queue(struct nvme_qpair *qpair) 56{ 57 struct nvme_completion *cpl; 58 struct nvme_command *cmd; 59 int i; 60 61 printf("id:%04Xh phase:%d\n", qpair->id, qpair->phase); 62 63 printf("Completion queue:\n"); 64 for (i = 0; i < qpair->num_entries; i++) { 65 cpl = &qpair->cpl[i]; 66 printf("%05d: ", i); 67 nvme_qpair_print_completion(qpair, cpl); 68 } 69 70 printf("Submission queue:\n"); 71 for (i = 0; i < qpair->num_entries; i++) { 72 cmd = &qpair->cmd[i]; 73 printf("%05d: ", i); 74 nvme_qpair_print_command(qpair, cmd); 75 } 76} 77 78static int 79nvme_sysctl_dump_debug(SYSCTL_HANDLER_ARGS) 80{ 81 struct nvme_qpair *qpair = arg1; 82 uint32_t val = 0; 83 84 int error = sysctl_handle_int(oidp, &val, 0, req); 85 86 if (error) 87 return (error); 88 89 if (val != 0) 90 nvme_dump_queue(qpair); 91 92 return (0); 93} 94 95static int 96nvme_sysctl_int_coal_time(SYSCTL_HANDLER_ARGS) 97{ 98 struct nvme_controller *ctrlr = arg1; 99 uint32_t oldval = ctrlr->int_coal_time; 100 int error = sysctl_handle_int(oidp, &ctrlr->int_coal_time, 0, 101 req); 102 103 if (error) 104 return (error); 105 106 if (oldval != ctrlr->int_coal_time) 107 nvme_ctrlr_cmd_set_interrupt_coalescing(ctrlr, 108 ctrlr->int_coal_time, ctrlr->int_coal_threshold, NULL, 109 NULL); 110 111 return (0); 112} 113 114static int 115nvme_sysctl_int_coal_threshold(SYSCTL_HANDLER_ARGS) 116{ 117 struct nvme_controller *ctrlr = arg1; 118 uint32_t oldval = ctrlr->int_coal_threshold; 119 int error = sysctl_handle_int(oidp, &ctrlr->int_coal_threshold, 0, 120 req); 121 122 if (error) 123 return (error); 124 125 if (oldval != ctrlr->int_coal_threshold) 126 nvme_ctrlr_cmd_set_interrupt_coalescing(ctrlr, 127 ctrlr->int_coal_time, ctrlr->int_coal_threshold, NULL, 128 NULL); 129 130 return (0); 131} 132 133static int 134nvme_sysctl_timeout_period(SYSCTL_HANDLER_ARGS) 135{ 136 uint32_t *ptr = arg1; 137 uint32_t newval = *ptr; 138 int error = sysctl_handle_int(oidp, &newval, 0, req); 139 140 if (error || (req->newptr == NULL)) 141 return (error); 142 143 if (newval > NVME_MAX_TIMEOUT_PERIOD || 144 newval < NVME_MIN_TIMEOUT_PERIOD) { 145 return (EINVAL); 146 } else { 147 *ptr = newval; 148 } 149 150 return (0); 151} 152 153static void 154nvme_qpair_reset_stats(struct nvme_qpair *qpair) 155{ 156 157 /* 158 * Reset the values. Due to sanity checks in 159 * nvme_qpair_process_completions, we reset the number of interrupt 160 * calls to 1. 161 */ 162 qpair->num_cmds = 0; 163 qpair->num_intr_handler_calls = 1; 164 qpair->num_retries = 0; 165 qpair->num_failures = 0; 166 qpair->num_ignored = 0; 167 qpair->num_recovery_nolock = 0; 168} 169 170static int 171nvme_sysctl_num_cmds(SYSCTL_HANDLER_ARGS) 172{ 173 struct nvme_controller *ctrlr = arg1; 174 int64_t num_cmds = 0; 175 int i; 176 177 num_cmds = ctrlr->adminq.num_cmds; 178 179 for (i = 0; i < ctrlr->num_io_queues; i++) 180 num_cmds += ctrlr->ioq[i].num_cmds; 181 182 return (sysctl_handle_64(oidp, &num_cmds, 0, req)); 183} 184 185static int 186nvme_sysctl_num_intr_handler_calls(SYSCTL_HANDLER_ARGS) 187{ 188 struct nvme_controller *ctrlr = arg1; 189 int64_t num_intr_handler_calls = 0; 190 int i; 191 192 num_intr_handler_calls = ctrlr->adminq.num_intr_handler_calls; 193 194 for (i = 0; i < ctrlr->num_io_queues; i++) 195 num_intr_handler_calls += ctrlr->ioq[i].num_intr_handler_calls; 196 197 return (sysctl_handle_64(oidp, &num_intr_handler_calls, 0, req)); 198} 199 200static int 201nvme_sysctl_num_retries(SYSCTL_HANDLER_ARGS) 202{ 203 struct nvme_controller *ctrlr = arg1; 204 int64_t num_retries = 0; 205 int i; 206 207 num_retries = ctrlr->adminq.num_retries; 208 209 for (i = 0; i < ctrlr->num_io_queues; i++) 210 num_retries += ctrlr->ioq[i].num_retries; 211 212 return (sysctl_handle_64(oidp, &num_retries, 0, req)); 213} 214 215static int 216nvme_sysctl_num_failures(SYSCTL_HANDLER_ARGS) 217{ 218 struct nvme_controller *ctrlr = arg1; 219 int64_t num_failures = 0; 220 int i; 221 222 num_failures = ctrlr->adminq.num_failures; 223 224 for (i = 0; i < ctrlr->num_io_queues; i++) 225 num_failures += ctrlr->ioq[i].num_failures; 226 227 return (sysctl_handle_64(oidp, &num_failures, 0, req)); 228} 229 230static int 231nvme_sysctl_num_ignored(SYSCTL_HANDLER_ARGS) 232{ 233 struct nvme_controller *ctrlr = arg1; 234 int64_t num_ignored = 0; 235 int i; 236 237 num_ignored = ctrlr->adminq.num_ignored; 238 239 for (i = 0; i < ctrlr->num_io_queues; i++) 240 num_ignored += ctrlr->ioq[i].num_ignored; 241 242 return (sysctl_handle_64(oidp, &num_ignored, 0, req)); 243} 244 245static int 246nvme_sysctl_num_recovery_nolock(SYSCTL_HANDLER_ARGS) 247{ 248 struct nvme_controller *ctrlr = arg1; 249 int64_t num; 250 int i; 251 252 num = ctrlr->adminq.num_recovery_nolock; 253 254 for (i = 0; i < ctrlr->num_io_queues; i++) 255 num += ctrlr->ioq[i].num_recovery_nolock; 256 257 return (sysctl_handle_64(oidp, &num, 0, req)); 258} 259 260static int 261nvme_sysctl_reset_stats(SYSCTL_HANDLER_ARGS) 262{ 263 struct nvme_controller *ctrlr = arg1; 264 uint32_t i, val = 0; 265 266 int error = sysctl_handle_int(oidp, &val, 0, req); 267 268 if (error) 269 return (error); 270 271 if (val != 0) { 272 nvme_qpair_reset_stats(&ctrlr->adminq); 273 274 for (i = 0; i < ctrlr->num_io_queues; i++) 275 nvme_qpair_reset_stats(&ctrlr->ioq[i]); 276 } 277 278 return (0); 279} 280 281static void 282nvme_sysctl_initialize_queue(struct nvme_qpair *qpair, 283 struct sysctl_ctx_list *ctrlr_ctx, struct sysctl_oid *que_tree) 284{ 285 struct sysctl_oid_list *que_list = SYSCTL_CHILDREN(que_tree); 286 287 SYSCTL_ADD_UINT(ctrlr_ctx, que_list, OID_AUTO, "num_entries", 288 CTLFLAG_RD, &qpair->num_entries, 0, 289 "Number of entries in hardware queue"); 290 SYSCTL_ADD_UINT(ctrlr_ctx, que_list, OID_AUTO, "num_trackers", 291 CTLFLAG_RD, &qpair->num_trackers, 0, 292 "Number of trackers pre-allocated for this queue pair"); 293 SYSCTL_ADD_UINT(ctrlr_ctx, que_list, OID_AUTO, "sq_head", 294 CTLFLAG_RD, &qpair->sq_head, 0, 295 "Current head of submission queue (as observed by driver)"); 296 SYSCTL_ADD_UINT(ctrlr_ctx, que_list, OID_AUTO, "sq_tail", 297 CTLFLAG_RD, &qpair->sq_tail, 0, 298 "Current tail of submission queue (as observed by driver)"); 299 SYSCTL_ADD_UINT(ctrlr_ctx, que_list, OID_AUTO, "cq_head", 300 CTLFLAG_RD, &qpair->cq_head, 0, 301 "Current head of completion queue (as observed by driver)"); 302 303 SYSCTL_ADD_QUAD(ctrlr_ctx, que_list, OID_AUTO, "num_cmds", 304 CTLFLAG_RD, &qpair->num_cmds, "Number of commands submitted"); 305 SYSCTL_ADD_QUAD(ctrlr_ctx, que_list, OID_AUTO, "num_intr_handler_calls", 306 CTLFLAG_RD, &qpair->num_intr_handler_calls, 307 "Number of times interrupt handler was invoked (will typically be " 308 "less than number of actual interrupts generated due to " 309 "coalescing)"); 310 SYSCTL_ADD_QUAD(ctrlr_ctx, que_list, OID_AUTO, "num_retries", 311 CTLFLAG_RD, &qpair->num_retries, "Number of commands retried"); 312 SYSCTL_ADD_QUAD(ctrlr_ctx, que_list, OID_AUTO, "num_failures", 313 CTLFLAG_RD, &qpair->num_failures, 314 "Number of commands ending in failure after all retries"); 315 SYSCTL_ADD_QUAD(ctrlr_ctx, que_list, OID_AUTO, "num_ignored", 316 CTLFLAG_RD, &qpair->num_ignored, 317 "Number of interrupts posted, but were administratively ignored"); 318 SYSCTL_ADD_QUAD(ctrlr_ctx, que_list, OID_AUTO, "num_recovery_nolock", 319 CTLFLAG_RD, &qpair->num_recovery_nolock, 320 "Number of times that we failed to lock recovery in the ISR"); 321 322 SYSCTL_ADD_PROC(ctrlr_ctx, que_list, OID_AUTO, 323 "dump_debug", CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_MPSAFE, 324 qpair, 0, nvme_sysctl_dump_debug, "IU", "Dump debug data"); 325} 326 327void 328nvme_sysctl_initialize_ctrlr(struct nvme_controller *ctrlr) 329{ 330 struct sysctl_ctx_list *ctrlr_ctx; 331 struct sysctl_oid *ctrlr_tree, *que_tree; 332 struct sysctl_oid_list *ctrlr_list; 333#define QUEUE_NAME_LENGTH 16 334 char queue_name[QUEUE_NAME_LENGTH]; 335 int i; 336 337 ctrlr_ctx = device_get_sysctl_ctx(ctrlr->dev); 338 ctrlr_tree = device_get_sysctl_tree(ctrlr->dev); 339 ctrlr_list = SYSCTL_CHILDREN(ctrlr_tree); 340 341 SYSCTL_ADD_UINT(ctrlr_ctx, ctrlr_list, OID_AUTO, "num_io_queues", 342 CTLFLAG_RD, &ctrlr->num_io_queues, 0, 343 "Number of I/O queue pairs"); 344 345 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO, 346 "int_coal_time", CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_MPSAFE, 347 ctrlr, 0, nvme_sysctl_int_coal_time, "IU", 348 "Interrupt coalescing timeout (in microseconds)"); 349 350 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO, 351 "int_coal_threshold", 352 CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_MPSAFE, ctrlr, 0, 353 nvme_sysctl_int_coal_threshold, "IU", 354 "Interrupt coalescing threshold"); 355 356 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO, 357 "admin_timeout_period", CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_MPSAFE, 358 &ctrlr->admin_timeout_period, 0, nvme_sysctl_timeout_period, "IU", 359 "Timeout period for Admin queue (in seconds)"); 360 361 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO, 362 "timeout_period", CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_MPSAFE, 363 &ctrlr->timeout_period, 0, nvme_sysctl_timeout_period, "IU", 364 "Timeout period for I/O queues (in seconds)"); 365 366 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO, 367 "num_cmds", CTLTYPE_S64 | CTLFLAG_RD | CTLFLAG_MPSAFE, 368 ctrlr, 0, nvme_sysctl_num_cmds, "IU", 369 "Number of commands submitted"); 370 371 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO, 372 "num_intr_handler_calls", 373 CTLTYPE_S64 | CTLFLAG_RD | CTLFLAG_MPSAFE, ctrlr, 0, 374 nvme_sysctl_num_intr_handler_calls, "IU", 375 "Number of times interrupt handler was invoked (will " 376 "typically be less than number of actual interrupts " 377 "generated due to coalescing)"); 378 379 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO, 380 "num_retries", CTLTYPE_S64 | CTLFLAG_RD | CTLFLAG_MPSAFE, 381 ctrlr, 0, nvme_sysctl_num_retries, "IU", 382 "Number of commands retried"); 383 384 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO, 385 "num_failures", CTLTYPE_S64 | CTLFLAG_RD | CTLFLAG_MPSAFE, 386 ctrlr, 0, nvme_sysctl_num_failures, "IU", 387 "Number of commands ending in failure after all retries"); 388 389 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO, 390 "num_ignored", CTLTYPE_S64 | CTLFLAG_RD | CTLFLAG_MPSAFE, 391 ctrlr, 0, nvme_sysctl_num_ignored, "IU", 392 "Number of interrupts ignored administratively"); 393 394 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO, 395 "num_recovery_nolock", CTLTYPE_S64 | CTLFLAG_RD | CTLFLAG_MPSAFE, 396 ctrlr, 0, nvme_sysctl_num_recovery_nolock, "IU", 397 "Number of times that we failed to lock recovery in the ISR"); 398 399 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO, 400 "reset_stats", CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_MPSAFE, ctrlr, 401 0, nvme_sysctl_reset_stats, "IU", "Reset statistics to zero"); 402 403 SYSCTL_ADD_UINT(ctrlr_ctx, ctrlr_list, OID_AUTO, "cap_lo", 404 CTLFLAG_RD, &ctrlr->cap_lo, 0, 405 "Low 32-bits of capacities for the drive"); 406 407 SYSCTL_ADD_UINT(ctrlr_ctx, ctrlr_list, OID_AUTO, "cap_hi", 408 CTLFLAG_RD, &ctrlr->cap_hi, 0, 409 "Hi 32-bits of capacities for the drive"); 410 411 que_tree = SYSCTL_ADD_NODE(ctrlr_ctx, ctrlr_list, OID_AUTO, "adminq", 412 CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Admin Queue"); 413 414 nvme_sysctl_initialize_queue(&ctrlr->adminq, ctrlr_ctx, que_tree); 415 416 for (i = 0; i < ctrlr->num_io_queues; i++) { 417 snprintf(queue_name, QUEUE_NAME_LENGTH, "ioq%d", i); 418 que_tree = SYSCTL_ADD_NODE(ctrlr_ctx, ctrlr_list, OID_AUTO, 419 queue_name, CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "IO Queue"); 420 nvme_sysctl_initialize_queue(&ctrlr->ioq[i], ctrlr_ctx, 421 que_tree); 422 } 423 424 SYSCTL_ADD_COUNTER_U64(ctrlr_ctx, ctrlr_list, OID_AUTO, "alignment_splits", 425 CTLFLAG_RD, &ctrlr->alignment_splits, 426 "Number of times we split the I/O alignment for drives with preferred alignment"); 427} 428