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 43 int nvme_use_nvd = NVME_USE_NVD; 44 bool nvme_verbose_cmd_dump = false; 45 46 SYSCTL_NODE(_hw, OID_AUTO, nvme, CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 47 "NVMe sysctl tunables"); 48 SYSCTL_INT(_hw_nvme, OID_AUTO, use_nvd, CTLFLAG_RDTUN, 49 &nvme_use_nvd, 1, "1 = Create NVD devices, 0 = Create NDA devices"); 50 SYSCTL_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 54 static void 55 nvme_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 = NVME_SQE(qpair, i); 73 printf("%05d: ", i); 74 nvme_qpair_print_command(qpair, cmd); 75 } 76 } 77 78 static int 79 nvme_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 95 static int 96 nvme_sysctl_int_coal(SYSCTL_HANDLER_ARGS) 97 { 98 struct nvme_controller *ctrlr = arg1; 99 uint32_t *valp = (uint32_t *)((char *)ctrlr + (size_t)arg2); 100 uint32_t oldval = *valp; 101 int error = sysctl_handle_int(oidp, valp, 0, req); 102 103 if (error) 104 return (error); 105 106 if (oldval != *valp) 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 } 113 114 static int 115 nvme_sysctl_timeout_period(SYSCTL_HANDLER_ARGS) 116 { 117 uint32_t *ptr = arg1; 118 uint32_t newval = *ptr; 119 int error = sysctl_handle_int(oidp, &newval, 0, req); 120 121 if (error || (req->newptr == NULL)) 122 return (error); 123 124 if (newval > NVME_MAX_TIMEOUT_PERIOD || 125 newval < NVME_MIN_TIMEOUT_PERIOD) { 126 return (EINVAL); 127 } else { 128 *ptr = newval; 129 } 130 131 return (0); 132 } 133 134 static void 135 nvme_qpair_reset_stats(struct nvme_qpair *qpair) 136 { 137 /* 138 * Reset the values. Due to sanity checks in 139 * nvme_qpair_process_completions, we reset the number of interrupt 140 * calls to 1. 141 */ 142 qpair->num_cmds = 0; 143 qpair->num_intr_handler_calls = 1; 144 qpair->num_retries = 0; 145 qpair->num_failures = 0; 146 qpair->num_ignored = 0; 147 qpair->num_recovery_nolock = 0; 148 } 149 150 static int 151 nvme_sysctl_qpair_stat(SYSCTL_HANDLER_ARGS) 152 { 153 struct nvme_controller *ctrlr = arg1; 154 size_t off = arg2; 155 int64_t sum; 156 int i; 157 158 sum = *(int64_t *)((char *)&ctrlr->adminq + off); 159 160 if (ctrlr->ioq != NULL) { 161 for (i = 0; i < ctrlr->num_io_queues; i++) 162 sum += *(int64_t *)((char *)&ctrlr->ioq[i] + off); 163 } 164 165 return (sysctl_handle_64(oidp, &sum, 0, req)); 166 } 167 168 static int 169 nvme_sysctl_reset_stats(SYSCTL_HANDLER_ARGS) 170 { 171 struct nvme_controller *ctrlr = arg1; 172 uint32_t i, val = 0; 173 174 int error = sysctl_handle_int(oidp, &val, 0, req); 175 176 if (error) 177 return (error); 178 179 if (val != 0) { 180 nvme_qpair_reset_stats(&ctrlr->adminq); 181 182 if (ctrlr->ioq != NULL) { 183 for (i = 0; i < ctrlr->num_io_queues; i++) 184 nvme_qpair_reset_stats(&ctrlr->ioq[i]); 185 } 186 } 187 188 return (0); 189 } 190 191 static void 192 nvme_sysctl_initialize_queue(struct nvme_qpair *qpair, 193 struct sysctl_ctx_list *ctrlr_ctx, struct sysctl_oid *que_tree) 194 { 195 struct sysctl_oid_list *que_list = SYSCTL_CHILDREN(que_tree); 196 197 SYSCTL_ADD_UINT(ctrlr_ctx, que_list, OID_AUTO, "num_entries", 198 CTLFLAG_RD, &qpair->num_entries, 0, 199 "Number of entries in hardware queue"); 200 SYSCTL_ADD_UINT(ctrlr_ctx, que_list, OID_AUTO, "num_trackers", 201 CTLFLAG_RD, &qpair->num_trackers, 0, 202 "Number of trackers pre-allocated for this queue pair"); 203 SYSCTL_ADD_UINT(ctrlr_ctx, que_list, OID_AUTO, "sq_head", 204 CTLFLAG_RD, &qpair->sq_head, 0, 205 "Current head of submission queue (as observed by driver)"); 206 SYSCTL_ADD_UINT(ctrlr_ctx, que_list, OID_AUTO, "sq_tail", 207 CTLFLAG_RD, &qpair->sq_tail, 0, 208 "Current tail of submission queue (as observed by driver)"); 209 SYSCTL_ADD_UINT(ctrlr_ctx, que_list, OID_AUTO, "cq_head", 210 CTLFLAG_RD, &qpair->cq_head, 0, 211 "Current head of completion queue (as observed by driver)"); 212 213 SYSCTL_ADD_QUAD(ctrlr_ctx, que_list, OID_AUTO, "num_cmds", 214 CTLFLAG_RD, &qpair->num_cmds, "Number of commands submitted"); 215 SYSCTL_ADD_QUAD(ctrlr_ctx, que_list, OID_AUTO, "num_intr_handler_calls", 216 CTLFLAG_RD, &qpair->num_intr_handler_calls, 217 "Number of times interrupt handler was invoked (will typically be " 218 "less than number of actual interrupts generated due to " 219 "coalescing)"); 220 SYSCTL_ADD_QUAD(ctrlr_ctx, que_list, OID_AUTO, "num_retries", 221 CTLFLAG_RD, &qpair->num_retries, "Number of commands retried"); 222 SYSCTL_ADD_QUAD(ctrlr_ctx, que_list, OID_AUTO, "num_failures", 223 CTLFLAG_RD, &qpair->num_failures, 224 "Number of commands ending in failure after all retries"); 225 SYSCTL_ADD_QUAD(ctrlr_ctx, que_list, OID_AUTO, "num_ignored", 226 CTLFLAG_RD, &qpair->num_ignored, 227 "Number of interrupts posted, but were administratively ignored"); 228 SYSCTL_ADD_QUAD(ctrlr_ctx, que_list, OID_AUTO, "num_recovery_nolock", 229 CTLFLAG_RD, &qpair->num_recovery_nolock, 230 "Number of times that we failed to lock recovery in the ISR"); 231 232 SYSCTL_ADD_UINT(ctrlr_ctx, que_list, OID_AUTO, "recovery", 233 CTLFLAG_RW, &qpair->recovery_state, 0, 234 "Current recovery state of the queue"); 235 236 SYSCTL_ADD_PROC(ctrlr_ctx, que_list, OID_AUTO, 237 "dump_debug", CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_MPSAFE, 238 qpair, 0, nvme_sysctl_dump_debug, "IU", "Dump debug data"); 239 } 240 241 void 242 nvme_sysctl_initialize_ctrlr(struct nvme_controller *ctrlr) 243 { 244 struct sysctl_ctx_list *ctrlr_ctx; 245 struct sysctl_oid *ctrlr_tree, *que_tree, *ioq_tree; 246 struct sysctl_oid_list *ctrlr_list, *ioq_list; 247 #define QUEUE_NAME_LENGTH 16 248 char queue_name[QUEUE_NAME_LENGTH]; 249 int i; 250 251 ctrlr_ctx = device_get_sysctl_ctx(ctrlr->dev); 252 ctrlr_tree = device_get_sysctl_tree(ctrlr->dev); 253 ctrlr_list = SYSCTL_CHILDREN(ctrlr_tree); 254 255 SYSCTL_ADD_UINT(ctrlr_ctx, ctrlr_list, OID_AUTO, "num_io_queues", 256 CTLFLAG_RD, &ctrlr->num_io_queues, 0, 257 "Number of I/O queue pairs"); 258 259 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO, 260 "int_coal_time", CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_MPSAFE, 261 ctrlr, offsetof(struct nvme_controller, int_coal_time), 262 nvme_sysctl_int_coal, "IU", 263 "Interrupt coalescing timeout (in microseconds)"); 264 265 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO, 266 "int_coal_threshold", 267 CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_MPSAFE, ctrlr, 268 offsetof(struct nvme_controller, int_coal_threshold), 269 nvme_sysctl_int_coal, "IU", 270 "Interrupt coalescing threshold"); 271 272 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO, 273 "admin_timeout_period", CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_MPSAFE, 274 &ctrlr->admin_timeout_period, 0, nvme_sysctl_timeout_period, "IU", 275 "Timeout period for Admin queue (in seconds)"); 276 277 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO, 278 "timeout_period", CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_MPSAFE, 279 &ctrlr->timeout_period, 0, nvme_sysctl_timeout_period, "IU", 280 "Timeout period for I/O queues (in seconds)"); 281 282 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO, 283 "num_cmds", CTLTYPE_S64 | CTLFLAG_RD | CTLFLAG_MPSAFE, 284 ctrlr, offsetof(struct nvme_qpair, num_cmds), 285 nvme_sysctl_qpair_stat, "IU", 286 "Number of commands submitted"); 287 288 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO, 289 "num_intr_handler_calls", 290 CTLTYPE_S64 | CTLFLAG_RD | CTLFLAG_MPSAFE, ctrlr, 291 offsetof(struct nvme_qpair, num_intr_handler_calls), 292 nvme_sysctl_qpair_stat, "IU", 293 "Number of times interrupt handler was invoked (will " 294 "typically be less than number of actual interrupts " 295 "generated due to coalescing)"); 296 297 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO, 298 "num_retries", CTLTYPE_S64 | CTLFLAG_RD | CTLFLAG_MPSAFE, 299 ctrlr, offsetof(struct nvme_qpair, num_retries), 300 nvme_sysctl_qpair_stat, "IU", 301 "Number of commands retried"); 302 303 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO, 304 "num_failures", CTLTYPE_S64 | CTLFLAG_RD | CTLFLAG_MPSAFE, 305 ctrlr, offsetof(struct nvme_qpair, num_failures), 306 nvme_sysctl_qpair_stat, "IU", 307 "Number of commands ending in failure after all retries"); 308 309 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO, 310 "num_ignored", CTLTYPE_S64 | CTLFLAG_RD | CTLFLAG_MPSAFE, 311 ctrlr, offsetof(struct nvme_qpair, num_ignored), 312 nvme_sysctl_qpair_stat, "IU", 313 "Number of interrupts ignored administratively"); 314 315 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO, 316 "num_recovery_nolock", CTLTYPE_S64 | CTLFLAG_RD | CTLFLAG_MPSAFE, 317 ctrlr, offsetof(struct nvme_qpair, num_recovery_nolock), 318 nvme_sysctl_qpair_stat, "IU", 319 "Number of times that we failed to lock recovery in the ISR"); 320 321 SYSCTL_ADD_PROC(ctrlr_ctx, ctrlr_list, OID_AUTO, 322 "reset_stats", CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_MPSAFE, ctrlr, 323 0, nvme_sysctl_reset_stats, "IU", "Reset statistics to zero"); 324 325 SYSCTL_ADD_UINT(ctrlr_ctx, ctrlr_list, OID_AUTO, "cap_lo", 326 CTLFLAG_RD, &ctrlr->cap_lo, 0, 327 "Low 32-bits of capacities for the drive"); 328 329 SYSCTL_ADD_UINT(ctrlr_ctx, ctrlr_list, OID_AUTO, "cap_hi", 330 CTLFLAG_RD, &ctrlr->cap_hi, 0, 331 "Hi 32-bits of capacities for the drive"); 332 333 SYSCTL_ADD_BOOL(ctrlr_ctx, ctrlr_list, OID_AUTO, "fail_on_reset", 334 CTLFLAG_RD, &ctrlr->fail_on_reset, 0, 335 "Pretend the next reset fails and fail the controller"); 336 337 que_tree = SYSCTL_ADD_NODE(ctrlr_ctx, ctrlr_list, OID_AUTO, "adminq", 338 CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Admin Queue"); 339 340 nvme_sysctl_initialize_queue(&ctrlr->adminq, ctrlr_ctx, que_tree); 341 342 /* 343 * Make sure that we've constructed the I/O queues before setting up the 344 * sysctls. Failed controllers won't allocate it, but we want the rest 345 * of the sysctls to diagnose things. 346 */ 347 if (ctrlr->ioq != NULL) { 348 ioq_tree = SYSCTL_ADD_NODE(ctrlr_ctx, ctrlr_list, OID_AUTO, 349 "ioq", CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "I/O Queues"); 350 ioq_list = SYSCTL_CHILDREN(ioq_tree); 351 352 for (i = 0; i < ctrlr->num_io_queues; i++) { 353 snprintf(queue_name, QUEUE_NAME_LENGTH, "%d", i); 354 que_tree = SYSCTL_ADD_NODE(ctrlr_ctx, ioq_list, OID_AUTO, 355 queue_name, CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "IO Queue"); 356 nvme_sysctl_initialize_queue(&ctrlr->ioq[i], ctrlr_ctx, 357 que_tree); 358 } 359 } 360 361 SYSCTL_ADD_COUNTER_U64(ctrlr_ctx, ctrlr_list, OID_AUTO, "alignment_splits", 362 CTLFLAG_RD, &ctrlr->alignment_splits, 363 "Number of times we split the I/O alignment for drives with preferred alignment"); 364 } 365