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_done(void *arg, const struct nvme_completion *cpl); 51 52 static void *nvd_new_disk(struct nvme_namespace *ns, void *ctrlr); 53 static void destroy_geom_disk(struct nvd_disk *ndisk); 54 55 static void *nvd_new_controller(struct nvme_controller *ctrlr); 56 static void nvd_controller_fail(void *ctrlr); 57 58 static int nvd_load(void); 59 static void nvd_unload(void); 60 61 MALLOC_DEFINE(M_NVD, "nvd", "nvd(4) allocations"); 62 63 struct nvme_consumer *consumer_handle; 64 65 struct nvd_disk { 66 67 struct bio_queue_head bioq; 68 struct task bioqtask; 69 struct mtx bioqlock; 70 71 struct disk *disk; 72 struct taskqueue *tq; 73 struct nvme_namespace *ns; 74 75 uint32_t cur_depth; 76 77 TAILQ_ENTRY(nvd_disk) global_tailq; 78 TAILQ_ENTRY(nvd_disk) ctrlr_tailq; 79 }; 80 81 struct nvd_controller { 82 83 TAILQ_ENTRY(nvd_controller) tailq; 84 TAILQ_HEAD(, nvd_disk) disk_head; 85 }; 86 87 static TAILQ_HEAD(, nvd_controller) ctrlr_head; 88 static TAILQ_HEAD(disk_list, nvd_disk) disk_head; 89 90 static int nvd_modevent(module_t mod, int type, void *arg) 91 { 92 int error = 0; 93 94 switch (type) { 95 case MOD_LOAD: 96 error = nvd_load(); 97 break; 98 case MOD_UNLOAD: 99 nvd_unload(); 100 break; 101 default: 102 break; 103 } 104 105 return (error); 106 } 107 108 moduledata_t nvd_mod = { 109 NVD_STR, 110 (modeventhand_t)nvd_modevent, 111 0 112 }; 113 114 DECLARE_MODULE(nvd, nvd_mod, SI_SUB_DRIVERS, SI_ORDER_ANY); 115 MODULE_VERSION(nvd, 1); 116 MODULE_DEPEND(nvd, nvme, 1, 1, 1); 117 118 static int 119 nvd_load() 120 { 121 122 TAILQ_INIT(&ctrlr_head); 123 TAILQ_INIT(&disk_head); 124 125 consumer_handle = nvme_register_consumer(nvd_new_disk, 126 nvd_new_controller, NULL, nvd_controller_fail); 127 128 return (consumer_handle != NULL ? 0 : -1); 129 } 130 131 static void 132 nvd_unload() 133 { 134 struct nvd_controller *ctrlr; 135 struct nvd_disk *disk; 136 137 while (!TAILQ_EMPTY(&ctrlr_head)) { 138 ctrlr = TAILQ_FIRST(&ctrlr_head); 139 TAILQ_REMOVE(&ctrlr_head, ctrlr, tailq); 140 free(ctrlr, M_NVD); 141 } 142 143 while (!TAILQ_EMPTY(&disk_head)) { 144 disk = TAILQ_FIRST(&disk_head); 145 TAILQ_REMOVE(&disk_head, disk, global_tailq); 146 destroy_geom_disk(disk); 147 free(disk, M_NVD); 148 } 149 150 nvme_unregister_consumer(consumer_handle); 151 } 152 153 static int 154 nvd_bio_submit(struct nvd_disk *ndisk, struct bio *bp) 155 { 156 int err; 157 158 bp->bio_driver1 = NULL; 159 atomic_add_int(&ndisk->cur_depth, 1); 160 err = nvme_ns_bio_process(ndisk->ns, bp, nvd_done); 161 if (err) { 162 atomic_add_int(&ndisk->cur_depth, -1); 163 bp->bio_error = err; 164 bp->bio_flags |= BIO_ERROR; 165 bp->bio_resid = bp->bio_bcount; 166 biodone(bp); 167 return (-1); 168 } 169 170 return (0); 171 } 172 173 static void 174 nvd_strategy(struct bio *bp) 175 { 176 struct nvd_disk *ndisk; 177 178 ndisk = (struct nvd_disk *)bp->bio_disk->d_drv1; 179 180 mtx_lock(&ndisk->bioqlock); 181 bioq_insert_tail(&ndisk->bioq, bp); 182 mtx_unlock(&ndisk->bioqlock); 183 taskqueue_enqueue(ndisk->tq, &ndisk->bioqtask); 184 } 185 186 static int 187 nvd_ioctl(struct disk *ndisk, u_long cmd, void *data, int fflag, 188 struct thread *td) 189 { 190 int ret = 0; 191 192 switch (cmd) { 193 default: 194 ret = EIO; 195 } 196 197 return (ret); 198 } 199 200 static void 201 nvd_done(void *arg, const struct nvme_completion *cpl) 202 { 203 struct bio *bp; 204 struct nvd_disk *ndisk; 205 206 bp = (struct bio *)arg; 207 208 ndisk = bp->bio_disk->d_drv1; 209 210 atomic_add_int(&ndisk->cur_depth, -1); 211 212 biodone(bp); 213 } 214 215 static void 216 nvd_bioq_process(void *arg, int pending) 217 { 218 struct nvd_disk *ndisk = arg; 219 struct bio *bp; 220 221 for (;;) { 222 mtx_lock(&ndisk->bioqlock); 223 bp = bioq_takefirst(&ndisk->bioq); 224 mtx_unlock(&ndisk->bioqlock); 225 if (bp == NULL) 226 break; 227 228 if (nvd_bio_submit(ndisk, bp) != 0) { 229 continue; 230 } 231 232 #ifdef BIO_ORDERED 233 /* 234 * BIO_ORDERED flag dictates that the bio with BIO_ORDERED 235 * flag set must be completed before proceeding with 236 * additional bios. 237 */ 238 if (bp->bio_flags & BIO_ORDERED) { 239 while (ndisk->cur_depth > 0) { 240 pause("nvd flush", 1); 241 } 242 } 243 #endif 244 } 245 } 246 247 static void * 248 nvd_new_controller(struct nvme_controller *ctrlr) 249 { 250 struct nvd_controller *nvd_ctrlr; 251 252 nvd_ctrlr = malloc(sizeof(struct nvd_controller), M_NVD, 253 M_ZERO | M_WAITOK); 254 255 TAILQ_INIT(&nvd_ctrlr->disk_head); 256 TAILQ_INSERT_TAIL(&ctrlr_head, nvd_ctrlr, tailq); 257 258 return (nvd_ctrlr); 259 } 260 261 static void * 262 nvd_new_disk(struct nvme_namespace *ns, void *ctrlr_arg) 263 { 264 uint8_t descr[NVME_MODEL_NUMBER_LENGTH+1]; 265 struct nvd_disk *ndisk; 266 struct disk *disk; 267 struct nvd_controller *ctrlr = ctrlr_arg; 268 269 ndisk = malloc(sizeof(struct nvd_disk), M_NVD, M_ZERO | M_WAITOK); 270 271 disk = disk_alloc(); 272 disk->d_strategy = nvd_strategy; 273 disk->d_ioctl = nvd_ioctl; 274 disk->d_name = NVD_STR; 275 disk->d_drv1 = ndisk; 276 277 disk->d_maxsize = nvme_ns_get_max_io_xfer_size(ns); 278 disk->d_sectorsize = nvme_ns_get_sector_size(ns); 279 disk->d_mediasize = (off_t)nvme_ns_get_size(ns); 280 disk->d_delmaxsize = (off_t)nvme_ns_get_size(ns); 281 disk->d_stripesize = nvme_ns_get_optimal_sector_size(ns); 282 283 if (TAILQ_EMPTY(&disk_head)) 284 disk->d_unit = 0; 285 else 286 disk->d_unit = 287 TAILQ_LAST(&disk_head, disk_list)->disk->d_unit + 1; 288 289 disk->d_flags = DISKFLAG_DIRECT_COMPLETION; 290 291 if (nvme_ns_get_flags(ns) & NVME_NS_DEALLOCATE_SUPPORTED) 292 disk->d_flags |= DISKFLAG_CANDELETE; 293 294 if (nvme_ns_get_flags(ns) & NVME_NS_FLUSH_SUPPORTED) 295 disk->d_flags |= DISKFLAG_CANFLUSHCACHE; 296 297 /* ifdef used here to ease porting to stable branches at a later point. */ 298 #ifdef DISKFLAG_UNMAPPED_BIO 299 disk->d_flags |= DISKFLAG_UNMAPPED_BIO; 300 #endif 301 302 /* 303 * d_ident and d_descr are both far bigger than the length of either 304 * the serial or model number strings. 305 */ 306 nvme_strvis(disk->d_ident, nvme_ns_get_serial_number(ns), 307 sizeof(disk->d_ident), NVME_SERIAL_NUMBER_LENGTH); 308 309 nvme_strvis(descr, nvme_ns_get_model_number(ns), sizeof(descr), 310 NVME_MODEL_NUMBER_LENGTH); 311 312 #if __FreeBSD_version >= 900034 313 strlcpy(disk->d_descr, descr, sizeof(descr)); 314 #endif 315 316 ndisk->ns = ns; 317 ndisk->disk = disk; 318 ndisk->cur_depth = 0; 319 320 mtx_init(&ndisk->bioqlock, "NVD bioq lock", NULL, MTX_DEF); 321 bioq_init(&ndisk->bioq); 322 323 TASK_INIT(&ndisk->bioqtask, 0, nvd_bioq_process, ndisk); 324 ndisk->tq = taskqueue_create("nvd_taskq", M_WAITOK, 325 taskqueue_thread_enqueue, &ndisk->tq); 326 taskqueue_start_threads(&ndisk->tq, 1, PI_DISK, "nvd taskq"); 327 328 TAILQ_INSERT_TAIL(&disk_head, ndisk, global_tailq); 329 TAILQ_INSERT_TAIL(&ctrlr->disk_head, ndisk, ctrlr_tailq); 330 331 disk_create(disk, DISK_VERSION); 332 333 printf(NVD_STR"%u: <%s> NVMe namespace\n", disk->d_unit, descr); 334 printf(NVD_STR"%u: %juMB (%ju %u byte sectors)\n", disk->d_unit, 335 (uintmax_t)disk->d_mediasize / (1024*1024), 336 (uintmax_t)disk->d_mediasize / disk->d_sectorsize, 337 disk->d_sectorsize); 338 339 return (NULL); 340 } 341 342 static void 343 destroy_geom_disk(struct nvd_disk *ndisk) 344 { 345 struct bio *bp; 346 struct disk *disk; 347 uint32_t unit; 348 int cnt = 0; 349 350 disk = ndisk->disk; 351 unit = disk->d_unit; 352 taskqueue_free(ndisk->tq); 353 354 disk_destroy(ndisk->disk); 355 356 mtx_lock(&ndisk->bioqlock); 357 for (;;) { 358 bp = bioq_takefirst(&ndisk->bioq); 359 if (bp == NULL) 360 break; 361 bp->bio_error = EIO; 362 bp->bio_flags |= BIO_ERROR; 363 bp->bio_resid = bp->bio_bcount; 364 cnt++; 365 biodone(bp); 366 } 367 368 printf(NVD_STR"%u: lost device - %d outstanding\n", unit, cnt); 369 printf(NVD_STR"%u: removing device entry\n", unit); 370 371 mtx_unlock(&ndisk->bioqlock); 372 373 mtx_destroy(&ndisk->bioqlock); 374 } 375 376 static void 377 nvd_controller_fail(void *ctrlr_arg) 378 { 379 struct nvd_controller *ctrlr = ctrlr_arg; 380 struct nvd_disk *disk; 381 382 while (!TAILQ_EMPTY(&ctrlr->disk_head)) { 383 disk = TAILQ_FIRST(&ctrlr->disk_head); 384 TAILQ_REMOVE(&disk_head, disk, global_tailq); 385 TAILQ_REMOVE(&ctrlr->disk_head, disk, ctrlr_tailq); 386 destroy_geom_disk(disk); 387 free(disk, M_NVD); 388 } 389 390 TAILQ_REMOVE(&ctrlr_head, ctrlr, tailq); 391 free(ctrlr, M_NVD); 392 } 393 394