1 /*- 2 * Copyright (C) 2012-2013 Intel Corporation 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 17 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 24 * SUCH DAMAGE. 25 */ 26 27 #include <sys/cdefs.h> 28 __FBSDID("$FreeBSD$"); 29 30 #include <sys/param.h> 31 #include <sys/bio.h> 32 #include <sys/kernel.h> 33 #include <sys/malloc.h> 34 #include <sys/module.h> 35 #include <sys/systm.h> 36 #include <sys/taskqueue.h> 37 38 #include <geom/geom.h> 39 #include <geom/geom_disk.h> 40 41 #include <dev/nvme/nvme.h> 42 43 #define NVD_STR "nvd" 44 45 struct nvd_disk; 46 47 static disk_ioctl_t nvd_ioctl; 48 static disk_strategy_t nvd_strategy; 49 50 static void *nvd_new_disk(struct nvme_namespace *ns, void *ctrlr); 51 static void destroy_geom_disk(struct nvd_disk *ndisk); 52 53 static void *nvd_new_controller(struct nvme_controller *ctrlr); 54 static void nvd_controller_fail(void *ctrlr); 55 56 static int nvd_load(void); 57 static void nvd_unload(void); 58 59 MALLOC_DEFINE(M_NVD, "nvd", "nvd(4) allocations"); 60 61 struct nvme_consumer *consumer_handle; 62 63 struct nvd_disk { 64 65 struct bio_queue_head bioq; 66 struct task bioqtask; 67 struct mtx bioqlock; 68 69 struct disk *disk; 70 struct taskqueue *tq; 71 struct nvme_namespace *ns; 72 73 uint32_t cur_depth; 74 75 TAILQ_ENTRY(nvd_disk) global_tailq; 76 TAILQ_ENTRY(nvd_disk) ctrlr_tailq; 77 }; 78 79 struct nvd_controller { 80 81 TAILQ_ENTRY(nvd_controller) tailq; 82 TAILQ_HEAD(, nvd_disk) disk_head; 83 }; 84 85 static TAILQ_HEAD(, nvd_controller) ctrlr_head; 86 static TAILQ_HEAD(disk_list, nvd_disk) disk_head; 87 88 static int nvd_modevent(module_t mod, int type, void *arg) 89 { 90 int error = 0; 91 92 switch (type) { 93 case MOD_LOAD: 94 error = nvd_load(); 95 break; 96 case MOD_UNLOAD: 97 nvd_unload(); 98 break; 99 default: 100 break; 101 } 102 103 return (error); 104 } 105 106 moduledata_t nvd_mod = { 107 NVD_STR, 108 (modeventhand_t)nvd_modevent, 109 0 110 }; 111 112 DECLARE_MODULE(nvd, nvd_mod, SI_SUB_DRIVERS, SI_ORDER_ANY); 113 MODULE_VERSION(nvd, 1); 114 MODULE_DEPEND(nvd, nvme, 1, 1, 1); 115 116 static int 117 nvd_load() 118 { 119 120 TAILQ_INIT(&ctrlr_head); 121 TAILQ_INIT(&disk_head); 122 123 consumer_handle = nvme_register_consumer(nvd_new_disk, 124 nvd_new_controller, NULL, nvd_controller_fail); 125 126 return (consumer_handle != NULL ? 0 : -1); 127 } 128 129 static void 130 nvd_unload() 131 { 132 struct nvd_controller *ctrlr; 133 struct nvd_disk *disk; 134 135 while (!TAILQ_EMPTY(&ctrlr_head)) { 136 ctrlr = TAILQ_FIRST(&ctrlr_head); 137 TAILQ_REMOVE(&ctrlr_head, ctrlr, tailq); 138 free(ctrlr, M_NVD); 139 } 140 141 while (!TAILQ_EMPTY(&disk_head)) { 142 disk = TAILQ_FIRST(&disk_head); 143 TAILQ_REMOVE(&disk_head, disk, global_tailq); 144 destroy_geom_disk(disk); 145 free(disk, M_NVD); 146 } 147 148 nvme_unregister_consumer(consumer_handle); 149 } 150 151 static void 152 nvd_strategy(struct bio *bp) 153 { 154 struct nvd_disk *ndisk; 155 156 ndisk = (struct nvd_disk *)bp->bio_disk->d_drv1; 157 158 mtx_lock(&ndisk->bioqlock); 159 bioq_insert_tail(&ndisk->bioq, bp); 160 mtx_unlock(&ndisk->bioqlock); 161 taskqueue_enqueue(ndisk->tq, &ndisk->bioqtask); 162 } 163 164 static int 165 nvd_ioctl(struct disk *ndisk, u_long cmd, void *data, int fflag, 166 struct thread *td) 167 { 168 int ret = 0; 169 170 switch (cmd) { 171 default: 172 ret = EIO; 173 } 174 175 return (ret); 176 } 177 178 static void 179 nvd_done(void *arg, const struct nvme_completion *cpl) 180 { 181 struct bio *bp; 182 struct nvd_disk *ndisk; 183 184 bp = (struct bio *)arg; 185 186 ndisk = bp->bio_disk->d_drv1; 187 188 atomic_add_int(&ndisk->cur_depth, -1); 189 190 biodone(bp); 191 } 192 193 static void 194 nvd_bioq_process(void *arg, int pending) 195 { 196 struct nvd_disk *ndisk = arg; 197 struct bio *bp; 198 int err; 199 200 for (;;) { 201 mtx_lock(&ndisk->bioqlock); 202 bp = bioq_takefirst(&ndisk->bioq); 203 mtx_unlock(&ndisk->bioqlock); 204 if (bp == NULL) 205 break; 206 207 #ifdef BIO_ORDERED 208 /* 209 * BIO_ORDERED flag dictates that all outstanding bios 210 * must be completed before processing the bio with 211 * BIO_ORDERED flag set. 212 */ 213 if (bp->bio_flags & BIO_ORDERED) { 214 while (ndisk->cur_depth > 0) { 215 pause("nvd flush", 1); 216 } 217 } 218 #endif 219 220 bp->bio_driver1 = NULL; 221 atomic_add_int(&ndisk->cur_depth, 1); 222 223 err = nvme_ns_bio_process(ndisk->ns, bp, nvd_done); 224 225 if (err) { 226 atomic_add_int(&ndisk->cur_depth, -1); 227 bp->bio_error = err; 228 bp->bio_flags |= BIO_ERROR; 229 bp->bio_resid = bp->bio_bcount; 230 biodone(bp); 231 } 232 233 #ifdef BIO_ORDERED 234 /* 235 * BIO_ORDERED flag dictates that the bio with BIO_ORDERED 236 * flag set must be completed before proceeding with 237 * additional bios. 238 */ 239 if (bp->bio_flags & BIO_ORDERED) { 240 while (ndisk->cur_depth > 0) { 241 pause("nvd flush", 1); 242 } 243 } 244 #endif 245 } 246 } 247 248 static void * 249 nvd_new_controller(struct nvme_controller *ctrlr) 250 { 251 struct nvd_controller *nvd_ctrlr; 252 253 nvd_ctrlr = malloc(sizeof(struct nvd_controller), M_NVD, 254 M_ZERO | M_WAITOK); 255 256 TAILQ_INIT(&nvd_ctrlr->disk_head); 257 TAILQ_INSERT_TAIL(&ctrlr_head, nvd_ctrlr, tailq); 258 259 return (nvd_ctrlr); 260 } 261 262 static void * 263 nvd_new_disk(struct nvme_namespace *ns, void *ctrlr_arg) 264 { 265 uint8_t descr[NVME_MODEL_NUMBER_LENGTH+1]; 266 struct nvd_disk *ndisk; 267 struct disk *disk; 268 struct nvd_controller *ctrlr = ctrlr_arg; 269 270 ndisk = malloc(sizeof(struct nvd_disk), M_NVD, M_ZERO | M_WAITOK); 271 272 disk = disk_alloc(); 273 disk->d_strategy = nvd_strategy; 274 disk->d_ioctl = nvd_ioctl; 275 disk->d_name = NVD_STR; 276 disk->d_drv1 = ndisk; 277 278 disk->d_maxsize = nvme_ns_get_max_io_xfer_size(ns); 279 disk->d_sectorsize = nvme_ns_get_sector_size(ns); 280 disk->d_mediasize = (off_t)nvme_ns_get_size(ns); 281 disk->d_delmaxsize = (off_t)nvme_ns_get_size(ns); 282 disk->d_stripesize = nvme_ns_get_stripesize(ns); 283 284 if (TAILQ_EMPTY(&disk_head)) 285 disk->d_unit = 0; 286 else 287 disk->d_unit = 288 TAILQ_LAST(&disk_head, disk_list)->disk->d_unit + 1; 289 290 disk->d_flags = 0; 291 292 if (nvme_ns_get_flags(ns) & NVME_NS_DEALLOCATE_SUPPORTED) 293 disk->d_flags |= DISKFLAG_CANDELETE; 294 295 if (nvme_ns_get_flags(ns) & NVME_NS_FLUSH_SUPPORTED) 296 disk->d_flags |= DISKFLAG_CANFLUSHCACHE; 297 298 /* ifdef used here to ease porting to stable branches at a later point. */ 299 #ifdef DISKFLAG_UNMAPPED_BIO 300 disk->d_flags |= DISKFLAG_UNMAPPED_BIO; 301 #endif 302 303 /* 304 * d_ident and d_descr are both far bigger than the length of either 305 * the serial or model number strings. 306 */ 307 nvme_strvis(disk->d_ident, nvme_ns_get_serial_number(ns), 308 sizeof(disk->d_ident), NVME_SERIAL_NUMBER_LENGTH); 309 310 nvme_strvis(descr, nvme_ns_get_model_number(ns), sizeof(descr), 311 NVME_MODEL_NUMBER_LENGTH); 312 313 #if __FreeBSD_version >= 900034 314 strlcpy(disk->d_descr, descr, sizeof(descr)); 315 #endif 316 317 ndisk->ns = ns; 318 ndisk->disk = disk; 319 ndisk->cur_depth = 0; 320 321 mtx_init(&ndisk->bioqlock, "NVD bioq lock", NULL, MTX_DEF); 322 bioq_init(&ndisk->bioq); 323 324 TASK_INIT(&ndisk->bioqtask, 0, nvd_bioq_process, ndisk); 325 ndisk->tq = taskqueue_create("nvd_taskq", M_WAITOK, 326 taskqueue_thread_enqueue, &ndisk->tq); 327 taskqueue_start_threads(&ndisk->tq, 1, PI_DISK, "nvd taskq"); 328 329 TAILQ_INSERT_TAIL(&disk_head, ndisk, global_tailq); 330 TAILQ_INSERT_TAIL(&ctrlr->disk_head, ndisk, ctrlr_tailq); 331 332 disk_create(disk, DISK_VERSION); 333 334 printf(NVD_STR"%u: <%s> NVMe namespace\n", disk->d_unit, descr); 335 printf(NVD_STR"%u: %juMB (%ju %u byte sectors)\n", disk->d_unit, 336 (uintmax_t)disk->d_mediasize / (1024*1024), 337 (uintmax_t)disk->d_mediasize / disk->d_sectorsize, 338 disk->d_sectorsize); 339 340 return (NULL); 341 } 342 343 static void 344 destroy_geom_disk(struct nvd_disk *ndisk) 345 { 346 struct bio *bp; 347 struct disk *disk; 348 uint32_t unit; 349 int cnt = 0; 350 351 disk = ndisk->disk; 352 unit = disk->d_unit; 353 taskqueue_free(ndisk->tq); 354 355 disk_destroy(ndisk->disk); 356 357 mtx_lock(&ndisk->bioqlock); 358 for (;;) { 359 bp = bioq_takefirst(&ndisk->bioq); 360 if (bp == NULL) 361 break; 362 bp->bio_error = EIO; 363 bp->bio_flags |= BIO_ERROR; 364 bp->bio_resid = bp->bio_bcount; 365 cnt++; 366 biodone(bp); 367 } 368 369 printf(NVD_STR"%u: lost device - %d outstanding\n", unit, cnt); 370 printf(NVD_STR"%u: removing device entry\n", unit); 371 372 mtx_unlock(&ndisk->bioqlock); 373 374 mtx_destroy(&ndisk->bioqlock); 375 } 376 377 static void 378 nvd_controller_fail(void *ctrlr_arg) 379 { 380 struct nvd_controller *ctrlr = ctrlr_arg; 381 struct nvd_disk *disk; 382 383 while (!TAILQ_EMPTY(&ctrlr->disk_head)) { 384 disk = TAILQ_FIRST(&ctrlr->disk_head); 385 TAILQ_REMOVE(&disk_head, disk, global_tailq); 386 TAILQ_REMOVE(&ctrlr->disk_head, disk, ctrlr_tailq); 387 destroy_geom_disk(disk); 388 free(disk, M_NVD); 389 } 390 391 TAILQ_REMOVE(&ctrlr_head, ctrlr, tailq); 392 free(ctrlr, M_NVD); 393 } 394 395