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/bus.h> 34 #include <sys/sysctl.h> 35 36 #include "nvme_private.h" 37 38 #ifndef NVME_USE_NVD 39 #define NVME_USE_NVD 0 40 #endif 41 42 int nvme_use_nvd = NVME_USE_NVD; 43 bool nvme_verbose_cmd_dump = false; 44 45 SYSCTL_NODE(_hw, OID_AUTO, nvme, CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 46 "NVMe sysctl tunables"); 47 SYSCTL_INT(_hw_nvme, OID_AUTO, use_nvd, CTLFLAG_RDTUN, 48 &nvme_use_nvd, 1, "1 = Create NVD devices, 0 = Create NDA devices"); 49 SYSCTL_BOOL(_hw_nvme, OID_AUTO, verbose_cmd_dump, CTLFLAG_RWTUN, 50 &nvme_verbose_cmd_dump, 0, 51 "enable verbose command printing when a command fails"); 52 53 static void 54 nvme_dump_queue(struct nvme_qpair *qpair) 55 { 56 struct nvme_completion *cpl; 57 struct nvme_command *cmd; 58 int i; 59 60 printf("id:%04Xh phase:%d\n", qpair->id, qpair->phase); 61 62 printf("Completion queue:\n"); 63 for (i = 0; i < qpair->num_entries; i++) { 64 cpl = &qpair->cpl[i]; 65 printf("%05d: ", i); 66 nvme_qpair_print_completion(qpair, cpl); 67 } 68 69 printf("Submission queue:\n"); 70 for (i = 0; i < qpair->num_entries; i++) { 71 cmd = &qpair->cmd[i]; 72 printf("%05d: ", i); 73 nvme_qpair_print_command(qpair, cmd); 74 } 75 } 76 77 static int 78 nvme_sysctl_dump_debug(SYSCTL_HANDLER_ARGS) 79 { 80 struct nvme_qpair *qpair = arg1; 81 uint32_t val = 0; 82 83 int error = sysctl_handle_int(oidp, &val, 0, req); 84 85 if (error) 86 return (error); 87 88 if (val != 0) 89 nvme_dump_queue(qpair); 90 91 return (0); 92 } 93 94 static int 95 nvme_sysctl_int_coal_time(SYSCTL_HANDLER_ARGS) 96 { 97 struct nvme_controller *ctrlr = arg1; 98 uint32_t oldval = ctrlr->int_coal_time; 99 int error = sysctl_handle_int(oidp, &ctrlr->int_coal_time, 0, 100 req); 101 102 if (error) 103 return (error); 104 105 if (oldval != ctrlr->int_coal_time) 106 nvme_ctrlr_cmd_set_interrupt_coalescing(ctrlr, 107 ctrlr->int_coal_time, ctrlr->int_coal_threshold, NULL, 108 NULL); 109 110 return (0); 111 } 112 113 static int 114 nvme_sysctl_int_coal_threshold(SYSCTL_HANDLER_ARGS) 115 { 116 struct nvme_controller *ctrlr = arg1; 117 uint32_t oldval = ctrlr->int_coal_threshold; 118 int error = sysctl_handle_int(oidp, &ctrlr->int_coal_threshold, 0, 119 req); 120 121 if (error) 122 return (error); 123 124 if (oldval != ctrlr->int_coal_threshold) 125 nvme_ctrlr_cmd_set_interrupt_coalescing(ctrlr, 126 ctrlr->int_coal_time, ctrlr->int_coal_threshold, NULL, 127 NULL); 128 129 return (0); 130 } 131 132 static int 133 nvme_sysctl_timeout_period(SYSCTL_HANDLER_ARGS) 134 { 135 struct nvme_controller *ctrlr = arg1; 136 uint32_t newval = ctrlr->timeout_period; 137 int error = sysctl_handle_int(oidp, &newval, 0, req); 138 139 if (error || (req->newptr == NULL)) 140 return (error); 141 142 if (newval > NVME_MAX_TIMEOUT_PERIOD || 143 newval < NVME_MIN_TIMEOUT_PERIOD) { 144 return (EINVAL); 145 } else { 146 ctrlr->timeout_period = newval; 147 } 148 149 return (0); 150 } 151 152 static void 153 nvme_qpair_reset_stats(struct nvme_qpair *qpair) 154 { 155 156 /* 157 * Reset the values. Due to sanity checks in 158 * nvme_qpair_process_completions, we reset the number of interrupt 159 * calls to 1. 160 */ 161 qpair->num_cmds = 0; 162 qpair->num_intr_handler_calls = 1; 163 qpair->num_retries = 0; 164 qpair->num_failures = 0; 165 qpair->num_ignored = 0; 166 } 167 168 static int 169 nvme_sysctl_num_cmds(SYSCTL_HANDLER_ARGS) 170 { 171 struct nvme_controller *ctrlr = arg1; 172 int64_t num_cmds = 0; 173 int i; 174 175 num_cmds = ctrlr->adminq.num_cmds; 176 177 for (i = 0; i < ctrlr->num_io_queues; i++) 178 num_cmds += ctrlr->ioq[i].num_cmds; 179 180 return (sysctl_handle_64(oidp, &num_cmds, 0, req)); 181 } 182 183 static int 184 nvme_sysctl_num_intr_handler_calls(SYSCTL_HANDLER_ARGS) 185 { 186 struct nvme_controller *ctrlr = arg1; 187 int64_t num_intr_handler_calls = 0; 188 int i; 189 190 num_intr_handler_calls = ctrlr->adminq.num_intr_handler_calls; 191 192 for (i = 0; i < ctrlr->num_io_queues; i++) 193 num_intr_handler_calls += ctrlr->ioq[i].num_intr_handler_calls; 194 195 return (sysctl_handle_64(oidp, &num_intr_handler_calls, 0, req)); 196 } 197 198 static int 199 nvme_sysctl_num_retries(SYSCTL_HANDLER_ARGS) 200 { 201 struct nvme_controller *ctrlr = arg1; 202 int64_t num_retries = 0; 203 int i; 204 205 num_retries = ctrlr->adminq.num_retries; 206 207 for (i = 0; i < ctrlr->num_io_queues; i++) 208 num_retries += ctrlr->ioq[i].num_retries; 209 210 return (sysctl_handle_64(oidp, &num_retries, 0, req)); 211 } 212 213 static int 214 nvme_sysctl_num_failures(SYSCTL_HANDLER_ARGS) 215 { 216 struct nvme_controller *ctrlr = arg1; 217 int64_t num_failures = 0; 218 int i; 219 220 num_failures = ctrlr->adminq.num_failures; 221 222 for (i = 0; i < ctrlr->num_io_queues; i++) 223 num_failures += ctrlr->ioq[i].num_failures; 224 225 return (sysctl_handle_64(oidp, &num_failures, 0, req)); 226 } 227 228 static int 229 nvme_sysctl_num_ignored(SYSCTL_HANDLER_ARGS) 230 { 231 struct nvme_controller *ctrlr = arg1; 232 int64_t num_ignored = 0; 233 int i; 234 235 num_ignored = ctrlr->adminq.num_ignored; 236 237 for (i = 0; i < ctrlr->num_io_queues; i++) 238 num_ignored += ctrlr->ioq[i].num_ignored; 239 240 return (sysctl_handle_64(oidp, &num_ignored, 0, req)); 241 } 242 243 static int 244 nvme_sysctl_reset_stats(SYSCTL_HANDLER_ARGS) 245 { 246 struct nvme_controller *ctrlr = arg1; 247 uint32_t i, val = 0; 248 249 int error = sysctl_handle_int(oidp, &val, 0, req); 250 251 if (error) 252 return (error); 253 254 if (val != 0) { 255 nvme_qpair_reset_stats(&ctrlr->adminq); 256 257 for (i = 0; i < ctrlr->num_io_queues; i++) 258 nvme_qpair_reset_stats(&ctrlr->ioq[i]); 259 } 260 261 return (0); 262 } 263 264 static void 265 nvme_sysctl_initialize_queue(struct nvme_qpair *qpair, 266 struct sysctl_ctx_list *ctrlr_ctx, struct sysctl_oid *que_tree) 267 { 268 struct sysctl_oid_list *que_list = SYSCTL_CHILDREN(que_tree); 269 270 SYSCTL_ADD_UINT(ctrlr_ctx, que_list, OID_AUTO, "num_entries", 271 CTLFLAG_RD, &qpair->num_entries, 0, 272 "Number of entries in hardware queue"); 273 SYSCTL_ADD_UINT(ctrlr_ctx, que_list, OID_AUTO, "num_trackers", 274 CTLFLAG_RD, &qpair->num_trackers, 0, 275 "Number of trackers pre-allocated for this queue pair"); 276 SYSCTL_ADD_UINT(ctrlr_ctx, que_list, OID_AUTO, "sq_head", 277 CTLFLAG_RD, &qpair->sq_head, 0, 278 "Current head of submission queue (as observed by driver)"); 279 SYSCTL_ADD_UINT(ctrlr_ctx, que_list, OID_AUTO, "sq_tail", 280 CTLFLAG_RD, &qpair->sq_tail, 0, 281 "Current tail of submission queue (as observed by driver)"); 282 SYSCTL_ADD_UINT(ctrlr_ctx, que_list, OID_AUTO, "cq_head", 283 CTLFLAG_RD, &qpair->cq_head, 0, 284 "Current head of completion queue (as observed by driver)"); 285 286 SYSCTL_ADD_QUAD(ctrlr_ctx, que_list, OID_AUTO, "num_cmds", 287 CTLFLAG_RD, &qpair->num_cmds, "Number of commands submitted"); 288 SYSCTL_ADD_QUAD(ctrlr_ctx, que_list, OID_AUTO, "num_intr_handler_calls", 289 CTLFLAG_RD, &qpair->num_intr_handler_calls, 290 "Number of times interrupt handler was invoked (will typically be " 291 "less than number of actual interrupts generated due to " 292 "coalescing)"); 293 SYSCTL_ADD_QUAD(ctrlr_ctx, que_list, OID_AUTO, "num_retries", 294 CTLFLAG_RD, &qpair->num_retries, "Number of commands retried"); 295 SYSCTL_ADD_QUAD(ctrlr_ctx, que_list, OID_AUTO, "num_failures", 296 CTLFLAG_RD, &qpair->num_failures, 297 "Number of commands ending in failure after all retries"); 298 SYSCTL_ADD_QUAD(ctrlr_ctx, que_list, OID_AUTO, "num_ignored", 299 CTLFLAG_RD, &qpair->num_ignored, 300 "Number of interrupts posted, but were administratively ignored"); 301 302 SYSCTL_ADD_PROC(ctrlr_ctx, que_list, OID_AUTO, 303 "dump_debug", CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_MPSAFE, 304 qpair, 0, nvme_sysctl_dump_debug, "IU", "Dump debug data"); 305 } 306 307 void 308 nvme_sysctl_initialize_ctrlr(struct nvme_controller *ctrlr) 309 { 310 struct sysctl_ctx_list *ctrlr_ctx; 311 struct sysctl_oid *ctrlr_tree, *que_tree; 312 struct sysctl_oid_list *ctrlr_list; 313 #define QUEUE_NAME_LENGTH 16 314 char queue_name[QUEUE_NAME_LENGTH]; 315 int i; 316 317 ctrlr_ctx = device_get_sysctl_ctx(ctrlr->dev); 318 ctrlr_tree = device_get_sysctl_tree(ctrlr->dev); 319 ctrlr_list = SYSCTL_CHILDREN(ctrlr_tree); 320 321 SYSCTL_ADD_UINT(ctrlr_ctx, ctrlr_list, OID_AUTO, "num_io_queues", 322 CTLFLAG_RD, &ctrlr->num_io_queues, 0, 323 "Number of I/O queue pairs"); 324 325 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO, 326 "int_coal_time", CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_MPSAFE, 327 ctrlr, 0, nvme_sysctl_int_coal_time, "IU", 328 "Interrupt coalescing timeout (in microseconds)"); 329 330 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO, 331 "int_coal_threshold", 332 CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_MPSAFE, ctrlr, 0, 333 nvme_sysctl_int_coal_threshold, "IU", 334 "Interrupt coalescing threshold"); 335 336 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO, 337 "timeout_period", CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_MPSAFE, 338 ctrlr, 0, nvme_sysctl_timeout_period, "IU", 339 "Timeout period (in seconds)"); 340 341 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO, 342 "num_cmds", CTLTYPE_S64 | CTLFLAG_RD | CTLFLAG_MPSAFE, 343 ctrlr, 0, nvme_sysctl_num_cmds, "IU", 344 "Number of commands submitted"); 345 346 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO, 347 "num_intr_handler_calls", 348 CTLTYPE_S64 | CTLFLAG_RD | CTLFLAG_MPSAFE, ctrlr, 0, 349 nvme_sysctl_num_intr_handler_calls, "IU", 350 "Number of times interrupt handler was invoked (will " 351 "typically be less than number of actual interrupts " 352 "generated due to coalescing)"); 353 354 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO, 355 "num_retries", CTLTYPE_S64 | CTLFLAG_RD | CTLFLAG_MPSAFE, 356 ctrlr, 0, nvme_sysctl_num_retries, "IU", 357 "Number of commands retried"); 358 359 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO, 360 "num_failures", CTLTYPE_S64 | CTLFLAG_RD | CTLFLAG_MPSAFE, 361 ctrlr, 0, nvme_sysctl_num_failures, "IU", 362 "Number of commands ending in failure after all retries"); 363 364 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO, 365 "num_ignored", CTLTYPE_S64 | CTLFLAG_RD | CTLFLAG_MPSAFE, 366 ctrlr, 0, nvme_sysctl_num_ignored, "IU", 367 "Number of interrupts ignored administratively"); 368 369 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO, 370 "reset_stats", CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_MPSAFE, ctrlr, 371 0, nvme_sysctl_reset_stats, "IU", "Reset statistics to zero"); 372 373 SYSCTL_ADD_UINT(ctrlr_ctx, ctrlr_list, OID_AUTO, "cap_lo", 374 CTLFLAG_RD, &ctrlr->cap_lo, 0, 375 "Low 32-bits of capacities for the drive"); 376 377 SYSCTL_ADD_UINT(ctrlr_ctx, ctrlr_list, OID_AUTO, "cap_hi", 378 CTLFLAG_RD, &ctrlr->cap_hi, 0, 379 "Hi 32-bits of capacities for the drive"); 380 381 que_tree = SYSCTL_ADD_NODE(ctrlr_ctx, ctrlr_list, OID_AUTO, "adminq", 382 CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Admin Queue"); 383 384 nvme_sysctl_initialize_queue(&ctrlr->adminq, ctrlr_ctx, que_tree); 385 386 for (i = 0; i < ctrlr->num_io_queues; i++) { 387 snprintf(queue_name, QUEUE_NAME_LENGTH, "ioq%d", i); 388 que_tree = SYSCTL_ADD_NODE(ctrlr_ctx, ctrlr_list, OID_AUTO, 389 queue_name, CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "IO Queue"); 390 nvme_sysctl_initialize_queue(&ctrlr->ioq[i], ctrlr_ctx, 391 que_tree); 392 } 393 } 394