1 /*- 2 * Copyright (c) 2003 Silicon Graphics International Corp. 3 * Copyright (c) 2009-2011 Spectra Logic Corporation 4 * All rights reserved. 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 1. Redistributions of source code must retain the above copyright 10 * notice, this list of conditions, and the following disclaimer, 11 * without modification. 12 * 2. Redistributions in binary form must reproduce at minimum a disclaimer 13 * substantially similar to the "NO WARRANTY" disclaimer below 14 * ("Disclaimer") and any redistribution must be conditioned upon 15 * including a substantially similar Disclaimer requirement for further 16 * binary redistribution. 17 * 18 * NO WARRANTY 19 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 20 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 21 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTIBILITY AND FITNESS FOR 22 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 23 * HOLDERS OR CONTRIBUTORS BE LIABLE FOR SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, 27 * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING 28 * IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 29 * POSSIBILITY OF SUCH DAMAGES. 30 * 31 * $Id: //depot/users/kenm/FreeBSD-test2/sys/cam/ctl/ctl_backend_block.c#5 $ 32 */ 33 /* 34 * CAM Target Layer driver backend for block devices. 35 * 36 * Author: Ken Merry <ken@FreeBSD.org> 37 */ 38 #include <sys/cdefs.h> 39 __FBSDID("$FreeBSD$"); 40 41 #include <opt_kdtrace.h> 42 43 #include <sys/param.h> 44 #include <sys/systm.h> 45 #include <sys/kernel.h> 46 #include <sys/types.h> 47 #include <sys/kthread.h> 48 #include <sys/bio.h> 49 #include <sys/fcntl.h> 50 #include <sys/lock.h> 51 #include <sys/mutex.h> 52 #include <sys/condvar.h> 53 #include <sys/malloc.h> 54 #include <sys/conf.h> 55 #include <sys/ioccom.h> 56 #include <sys/queue.h> 57 #include <sys/sbuf.h> 58 #include <sys/endian.h> 59 #include <sys/uio.h> 60 #include <sys/buf.h> 61 #include <sys/taskqueue.h> 62 #include <sys/vnode.h> 63 #include <sys/namei.h> 64 #include <sys/mount.h> 65 #include <sys/disk.h> 66 #include <sys/fcntl.h> 67 #include <sys/filedesc.h> 68 #include <sys/proc.h> 69 #include <sys/pcpu.h> 70 #include <sys/module.h> 71 #include <sys/sdt.h> 72 #include <sys/devicestat.h> 73 #include <sys/sysctl.h> 74 75 #include <geom/geom.h> 76 77 #include <cam/cam.h> 78 #include <cam/scsi/scsi_all.h> 79 #include <cam/scsi/scsi_da.h> 80 #include <cam/ctl/ctl_io.h> 81 #include <cam/ctl/ctl.h> 82 #include <cam/ctl/ctl_backend.h> 83 #include <cam/ctl/ctl_frontend_internal.h> 84 #include <cam/ctl/ctl_ioctl.h> 85 #include <cam/ctl/ctl_scsi_all.h> 86 #include <cam/ctl/ctl_error.h> 87 88 /* 89 * The idea here is that we'll allocate enough S/G space to hold a 16MB 90 * I/O. If we get an I/O larger than that, we'll reject it. 91 */ 92 #define CTLBLK_MAX_IO_SIZE (16 * 1024 * 1024) 93 #define CTLBLK_MAX_SEGS (CTLBLK_MAX_IO_SIZE / MAXPHYS) + 1 94 95 #ifdef CTLBLK_DEBUG 96 #define DPRINTF(fmt, args...) \ 97 printf("cbb(%s:%d): " fmt, __FUNCTION__, __LINE__, ##args) 98 #else 99 #define DPRINTF(fmt, args...) do {} while(0) 100 #endif 101 102 SDT_PROVIDER_DEFINE(cbb); 103 104 typedef enum { 105 CTL_BE_BLOCK_LUN_UNCONFIGURED = 0x01, 106 CTL_BE_BLOCK_LUN_CONFIG_ERR = 0x02, 107 CTL_BE_BLOCK_LUN_WAITING = 0x04, 108 CTL_BE_BLOCK_LUN_MULTI_THREAD = 0x08 109 } ctl_be_block_lun_flags; 110 111 typedef enum { 112 CTL_BE_BLOCK_NONE, 113 CTL_BE_BLOCK_DEV, 114 CTL_BE_BLOCK_FILE 115 } ctl_be_block_type; 116 117 struct ctl_be_block_devdata { 118 struct cdev *cdev; 119 struct cdevsw *csw; 120 int dev_ref; 121 }; 122 123 struct ctl_be_block_filedata { 124 struct ucred *cred; 125 }; 126 127 union ctl_be_block_bedata { 128 struct ctl_be_block_devdata dev; 129 struct ctl_be_block_filedata file; 130 }; 131 132 struct ctl_be_block_io; 133 struct ctl_be_block_lun; 134 135 typedef void (*cbb_dispatch_t)(struct ctl_be_block_lun *be_lun, 136 struct ctl_be_block_io *beio); 137 138 /* 139 * Backend LUN structure. There is a 1:1 mapping between a block device 140 * and a backend block LUN, and between a backend block LUN and a CTL LUN. 141 */ 142 struct ctl_be_block_lun { 143 struct ctl_block_disk *disk; 144 char lunname[32]; 145 char *dev_path; 146 ctl_be_block_type dev_type; 147 struct vnode *vn; 148 union ctl_be_block_bedata backend; 149 cbb_dispatch_t dispatch; 150 cbb_dispatch_t lun_flush; 151 struct mtx lock; 152 uma_zone_t lun_zone; 153 uint64_t size_blocks; 154 uint64_t size_bytes; 155 uint32_t blocksize; 156 int blocksize_shift; 157 struct ctl_be_block_softc *softc; 158 struct devstat *disk_stats; 159 ctl_be_block_lun_flags flags; 160 STAILQ_ENTRY(ctl_be_block_lun) links; 161 struct ctl_be_lun ctl_be_lun; 162 struct taskqueue *io_taskqueue; 163 struct task io_task; 164 int num_threads; 165 STAILQ_HEAD(, ctl_io_hdr) input_queue; 166 STAILQ_HEAD(, ctl_io_hdr) config_write_queue; 167 STAILQ_HEAD(, ctl_io_hdr) datamove_queue; 168 }; 169 170 /* 171 * Overall softc structure for the block backend module. 172 */ 173 struct ctl_be_block_softc { 174 STAILQ_HEAD(, ctl_be_block_io) beio_free_queue; 175 struct mtx lock; 176 int prealloc_beio; 177 int num_disks; 178 STAILQ_HEAD(, ctl_block_disk) disk_list; 179 int num_luns; 180 STAILQ_HEAD(, ctl_be_block_lun) lun_list; 181 }; 182 183 static struct ctl_be_block_softc backend_block_softc; 184 185 /* 186 * Per-I/O information. 187 */ 188 struct ctl_be_block_io { 189 union ctl_io *io; 190 struct ctl_sg_entry sg_segs[CTLBLK_MAX_SEGS]; 191 struct iovec xiovecs[CTLBLK_MAX_SEGS]; 192 int bio_cmd; 193 int bio_flags; 194 int num_segs; 195 int num_bios_sent; 196 int num_bios_done; 197 int send_complete; 198 int num_errors; 199 struct bintime ds_t0; 200 devstat_tag_type ds_tag_type; 201 devstat_trans_flags ds_trans_type; 202 uint64_t io_len; 203 uint64_t io_offset; 204 struct ctl_be_block_softc *softc; 205 struct ctl_be_block_lun *lun; 206 STAILQ_ENTRY(ctl_be_block_io) links; 207 }; 208 209 static int cbb_num_threads = 14; 210 TUNABLE_INT("kern.cam.ctl.block.num_threads", &cbb_num_threads); 211 SYSCTL_NODE(_kern_cam_ctl, OID_AUTO, block, CTLFLAG_RD, 0, 212 "CAM Target Layer Block Backend"); 213 SYSCTL_INT(_kern_cam_ctl_block, OID_AUTO, num_threads, CTLFLAG_RW, 214 &cbb_num_threads, 0, "Number of threads per backing file"); 215 216 static struct ctl_be_block_io *ctl_alloc_beio(struct ctl_be_block_softc *softc); 217 static void ctl_free_beio(struct ctl_be_block_io *beio); 218 static int ctl_grow_beio(struct ctl_be_block_softc *softc, int count); 219 #if 0 220 static void ctl_shrink_beio(struct ctl_be_block_softc *softc); 221 #endif 222 static void ctl_complete_beio(struct ctl_be_block_io *beio); 223 static int ctl_be_block_move_done(union ctl_io *io); 224 static void ctl_be_block_biodone(struct bio *bio); 225 static void ctl_be_block_flush_file(struct ctl_be_block_lun *be_lun, 226 struct ctl_be_block_io *beio); 227 static void ctl_be_block_dispatch_file(struct ctl_be_block_lun *be_lun, 228 struct ctl_be_block_io *beio); 229 static void ctl_be_block_flush_dev(struct ctl_be_block_lun *be_lun, 230 struct ctl_be_block_io *beio); 231 static void ctl_be_block_dispatch_dev(struct ctl_be_block_lun *be_lun, 232 struct ctl_be_block_io *beio); 233 static void ctl_be_block_cw_dispatch(struct ctl_be_block_lun *be_lun, 234 union ctl_io *io); 235 static void ctl_be_block_dispatch(struct ctl_be_block_lun *be_lun, 236 union ctl_io *io); 237 static void ctl_be_block_worker(void *context, int pending); 238 static int ctl_be_block_submit(union ctl_io *io); 239 static int ctl_be_block_ioctl(struct cdev *dev, u_long cmd, caddr_t addr, 240 int flag, struct thread *td); 241 static int ctl_be_block_open_file(struct ctl_be_block_lun *be_lun, 242 struct ctl_lun_req *req); 243 static int ctl_be_block_open_dev(struct ctl_be_block_lun *be_lun, 244 struct ctl_lun_req *req); 245 static int ctl_be_block_close(struct ctl_be_block_lun *be_lun); 246 static int ctl_be_block_open(struct ctl_be_block_softc *softc, 247 struct ctl_be_block_lun *be_lun, 248 struct ctl_lun_req *req); 249 static int ctl_be_block_create(struct ctl_be_block_softc *softc, 250 struct ctl_lun_req *req); 251 static int ctl_be_block_rm(struct ctl_be_block_softc *softc, 252 struct ctl_lun_req *req); 253 static void ctl_be_block_lun_shutdown(void *be_lun); 254 static void ctl_be_block_lun_config_status(void *be_lun, 255 ctl_lun_config_status status); 256 static int ctl_be_block_config_write(union ctl_io *io); 257 static int ctl_be_block_config_read(union ctl_io *io); 258 static int ctl_be_block_lun_info(void *be_lun, struct sbuf *sb); 259 int ctl_be_block_init(void); 260 261 static struct ctl_backend_driver ctl_be_block_driver = 262 { 263 name: "block", 264 flags: CTL_BE_FLAG_HAS_CONFIG, 265 init: ctl_be_block_init, 266 data_submit: ctl_be_block_submit, 267 data_move_done: ctl_be_block_move_done, 268 config_read: ctl_be_block_config_read, 269 config_write: ctl_be_block_config_write, 270 ioctl: ctl_be_block_ioctl, 271 lun_info: ctl_be_block_lun_info 272 }; 273 274 MALLOC_DEFINE(M_CTLBLK, "ctlblk", "Memory used for CTL block backend"); 275 CTL_BACKEND_DECLARE(cbb, ctl_be_block_driver); 276 277 static struct ctl_be_block_io * 278 ctl_alloc_beio(struct ctl_be_block_softc *softc) 279 { 280 struct ctl_be_block_io *beio; 281 int count; 282 283 mtx_lock(&softc->lock); 284 285 beio = STAILQ_FIRST(&softc->beio_free_queue); 286 if (beio != NULL) { 287 STAILQ_REMOVE(&softc->beio_free_queue, beio, 288 ctl_be_block_io, links); 289 } 290 mtx_unlock(&softc->lock); 291 292 if (beio != NULL) { 293 bzero(beio, sizeof(*beio)); 294 beio->softc = softc; 295 return (beio); 296 } 297 298 for (;;) { 299 300 count = ctl_grow_beio(softc, /*count*/ 10); 301 302 /* 303 * This shouldn't be possible, since ctl_grow_beio() uses a 304 * blocking malloc. 305 */ 306 if (count == 0) 307 return (NULL); 308 309 /* 310 * Since we have to drop the lock when we're allocating beio 311 * structures, it's possible someone else can come along and 312 * allocate the beio's we've just allocated. 313 */ 314 mtx_lock(&softc->lock); 315 beio = STAILQ_FIRST(&softc->beio_free_queue); 316 if (beio != NULL) { 317 STAILQ_REMOVE(&softc->beio_free_queue, beio, 318 ctl_be_block_io, links); 319 } 320 mtx_unlock(&softc->lock); 321 322 if (beio != NULL) { 323 bzero(beio, sizeof(*beio)); 324 beio->softc = softc; 325 break; 326 } 327 } 328 return (beio); 329 } 330 331 static void 332 ctl_free_beio(struct ctl_be_block_io *beio) 333 { 334 struct ctl_be_block_softc *softc; 335 int duplicate_free; 336 int i; 337 338 softc = beio->softc; 339 duplicate_free = 0; 340 341 for (i = 0; i < beio->num_segs; i++) { 342 if (beio->sg_segs[i].addr == NULL) 343 duplicate_free++; 344 345 uma_zfree(beio->lun->lun_zone, beio->sg_segs[i].addr); 346 beio->sg_segs[i].addr = NULL; 347 } 348 349 if (duplicate_free > 0) { 350 printf("%s: %d duplicate frees out of %d segments\n", __func__, 351 duplicate_free, beio->num_segs); 352 } 353 mtx_lock(&softc->lock); 354 STAILQ_INSERT_TAIL(&softc->beio_free_queue, beio, links); 355 mtx_unlock(&softc->lock); 356 } 357 358 static int 359 ctl_grow_beio(struct ctl_be_block_softc *softc, int count) 360 { 361 int i; 362 363 for (i = 0; i < count; i++) { 364 struct ctl_be_block_io *beio; 365 366 beio = (struct ctl_be_block_io *)malloc(sizeof(*beio), 367 M_CTLBLK, 368 M_WAITOK | M_ZERO); 369 if (beio == NULL) 370 break; 371 372 bzero(beio, sizeof(*beio)); 373 beio->softc = softc; 374 mtx_lock(&softc->lock); 375 STAILQ_INSERT_TAIL(&softc->beio_free_queue, beio, links); 376 mtx_unlock(&softc->lock); 377 } 378 379 return (i); 380 } 381 382 #if 0 383 static void 384 ctl_shrink_beio(struct ctl_be_block_softc *softc) 385 { 386 struct ctl_be_block_io *beio, *beio_tmp; 387 388 mtx_lock(&softc->lock); 389 STAILQ_FOREACH_SAFE(beio, &softc->beio_free_queue, links, beio_tmp) { 390 STAILQ_REMOVE(&softc->beio_free_queue, beio, 391 ctl_be_block_io, links); 392 free(beio, M_CTLBLK); 393 } 394 mtx_unlock(&softc->lock); 395 } 396 #endif 397 398 static void 399 ctl_complete_beio(struct ctl_be_block_io *beio) 400 { 401 union ctl_io *io; 402 int io_len; 403 404 io = beio->io; 405 406 if ((io->io_hdr.status & CTL_STATUS_MASK) == CTL_SUCCESS) 407 io_len = beio->io_len; 408 else 409 io_len = 0; 410 411 devstat_end_transaction(beio->lun->disk_stats, 412 /*bytes*/ io_len, 413 beio->ds_tag_type, 414 beio->ds_trans_type, 415 /*now*/ NULL, 416 /*then*/&beio->ds_t0); 417 418 ctl_free_beio(beio); 419 ctl_done(io); 420 } 421 422 static int 423 ctl_be_block_move_done(union ctl_io *io) 424 { 425 struct ctl_be_block_io *beio; 426 struct ctl_be_block_lun *be_lun; 427 #ifdef CTL_TIME_IO 428 struct bintime cur_bt; 429 #endif 430 431 beio = (struct ctl_be_block_io *) 432 io->io_hdr.ctl_private[CTL_PRIV_BACKEND].ptr; 433 434 be_lun = beio->lun; 435 436 DPRINTF("entered\n"); 437 438 #ifdef CTL_TIME_IO 439 getbintime(&cur_bt); 440 bintime_sub(&cur_bt, &io->io_hdr.dma_start_bt); 441 bintime_add(&io->io_hdr.dma_bt, &cur_bt); 442 io->io_hdr.num_dmas++; 443 #endif 444 445 /* 446 * We set status at this point for read commands, and write 447 * commands with errors. 448 */ 449 if ((beio->bio_cmd == BIO_READ) 450 && (io->io_hdr.port_status == 0) 451 && ((io->io_hdr.flags & CTL_FLAG_ABORT) == 0) 452 && ((io->io_hdr.status & CTL_STATUS_MASK) == CTL_STATUS_NONE)) 453 ctl_set_success(&io->scsiio); 454 else if ((io->io_hdr.port_status != 0) 455 && ((io->io_hdr.flags & CTL_FLAG_ABORT) == 0) 456 && ((io->io_hdr.status & CTL_STATUS_MASK) == CTL_STATUS_NONE)) { 457 /* 458 * For hardware error sense keys, the sense key 459 * specific value is defined to be a retry count, 460 * but we use it to pass back an internal FETD 461 * error code. XXX KDM Hopefully the FETD is only 462 * using 16 bits for an error code, since that's 463 * all the space we have in the sks field. 464 */ 465 ctl_set_internal_failure(&io->scsiio, 466 /*sks_valid*/ 1, 467 /*retry_count*/ 468 io->io_hdr.port_status); 469 } 470 471 /* 472 * If this is a read, or a write with errors, it is done. 473 */ 474 if ((beio->bio_cmd == BIO_READ) 475 || ((io->io_hdr.flags & CTL_FLAG_ABORT) != 0) 476 || ((io->io_hdr.status & CTL_STATUS_MASK) != CTL_STATUS_NONE)) { 477 ctl_complete_beio(beio); 478 return (0); 479 } 480 481 /* 482 * At this point, we have a write and the DMA completed 483 * successfully. We now have to queue it to the task queue to 484 * execute the backend I/O. That is because we do blocking 485 * memory allocations, and in the file backing case, blocking I/O. 486 * This move done routine is generally called in the SIM's 487 * interrupt context, and therefore we cannot block. 488 */ 489 mtx_lock(&be_lun->lock); 490 /* 491 * XXX KDM make sure that links is okay to use at this point. 492 * Otherwise, we either need to add another field to ctl_io_hdr, 493 * or deal with resource allocation here. 494 */ 495 STAILQ_INSERT_TAIL(&be_lun->datamove_queue, &io->io_hdr, links); 496 mtx_unlock(&be_lun->lock); 497 498 taskqueue_enqueue(be_lun->io_taskqueue, &be_lun->io_task); 499 500 return (0); 501 } 502 503 static void 504 ctl_be_block_biodone(struct bio *bio) 505 { 506 struct ctl_be_block_io *beio; 507 struct ctl_be_block_lun *be_lun; 508 union ctl_io *io; 509 510 beio = bio->bio_caller1; 511 be_lun = beio->lun; 512 io = beio->io; 513 514 DPRINTF("entered\n"); 515 516 mtx_lock(&be_lun->lock); 517 if (bio->bio_error != 0) 518 beio->num_errors++; 519 520 beio->num_bios_done++; 521 522 /* 523 * XXX KDM will this cause WITNESS to complain? Holding a lock 524 * during the free might cause it to complain. 525 */ 526 g_destroy_bio(bio); 527 528 /* 529 * If the send complete bit isn't set, or we aren't the last I/O to 530 * complete, then we're done. 531 */ 532 if ((beio->send_complete == 0) 533 || (beio->num_bios_done < beio->num_bios_sent)) { 534 mtx_unlock(&be_lun->lock); 535 return; 536 } 537 538 /* 539 * At this point, we've verified that we are the last I/O to 540 * complete, so it's safe to drop the lock. 541 */ 542 mtx_unlock(&be_lun->lock); 543 544 /* 545 * If there are any errors from the backing device, we fail the 546 * entire I/O with a medium error. 547 */ 548 if (beio->num_errors > 0) { 549 if (beio->bio_cmd == BIO_FLUSH) { 550 /* XXX KDM is there is a better error here? */ 551 ctl_set_internal_failure(&io->scsiio, 552 /*sks_valid*/ 1, 553 /*retry_count*/ 0xbad2); 554 } else 555 ctl_set_medium_error(&io->scsiio); 556 ctl_complete_beio(beio); 557 return; 558 } 559 560 /* 561 * If this is a write or a flush, we're all done. 562 * If this is a read, we can now send the data to the user. 563 */ 564 if ((beio->bio_cmd == BIO_WRITE) 565 || (beio->bio_cmd == BIO_FLUSH)) { 566 ctl_set_success(&io->scsiio); 567 ctl_complete_beio(beio); 568 } else { 569 io->scsiio.be_move_done = ctl_be_block_move_done; 570 io->scsiio.kern_data_ptr = (uint8_t *)beio->sg_segs; 571 io->scsiio.kern_data_len = beio->io_len; 572 io->scsiio.kern_total_len = beio->io_len; 573 io->scsiio.kern_rel_offset = 0; 574 io->scsiio.kern_data_resid = 0; 575 io->scsiio.kern_sg_entries = beio->num_segs; 576 io->io_hdr.flags |= CTL_FLAG_ALLOCATED | CTL_FLAG_KDPTR_SGLIST; 577 #ifdef CTL_TIME_IO 578 getbintime(&io->io_hdr.dma_start_bt); 579 #endif 580 ctl_datamove(io); 581 } 582 } 583 584 static void 585 ctl_be_block_flush_file(struct ctl_be_block_lun *be_lun, 586 struct ctl_be_block_io *beio) 587 { 588 union ctl_io *io; 589 struct mount *mountpoint; 590 int vfs_is_locked, error, lock_flags; 591 592 DPRINTF("entered\n"); 593 594 io = beio->io; 595 596 vfs_is_locked = VFS_LOCK_GIANT(be_lun->vn->v_mount); 597 598 (void) vn_start_write(be_lun->vn, &mountpoint, V_WAIT); 599 600 if (MNT_SHARED_WRITES(mountpoint) 601 || ((mountpoint == NULL) 602 && MNT_SHARED_WRITES(be_lun->vn->v_mount))) 603 lock_flags = LK_SHARED; 604 else 605 lock_flags = LK_EXCLUSIVE; 606 607 vn_lock(be_lun->vn, lock_flags | LK_RETRY); 608 609 binuptime(&beio->ds_t0); 610 devstat_start_transaction(beio->lun->disk_stats, &beio->ds_t0); 611 612 error = VOP_FSYNC(be_lun->vn, MNT_WAIT, curthread); 613 VOP_UNLOCK(be_lun->vn, 0); 614 615 vn_finished_write(mountpoint); 616 617 VFS_UNLOCK_GIANT(vfs_is_locked); 618 619 if (error == 0) 620 ctl_set_success(&io->scsiio); 621 else { 622 /* XXX KDM is there is a better error here? */ 623 ctl_set_internal_failure(&io->scsiio, 624 /*sks_valid*/ 1, 625 /*retry_count*/ 0xbad1); 626 } 627 628 ctl_complete_beio(beio); 629 } 630 631 SDT_PROBE_DEFINE1(cbb, kernel, read, file_start, file_start, "uint64_t"); 632 SDT_PROBE_DEFINE1(cbb, kernel, write, file_start, file_start, "uint64_t"); 633 SDT_PROBE_DEFINE1(cbb, kernel, read, file_done, file_done,"uint64_t"); 634 SDT_PROBE_DEFINE1(cbb, kernel, write, file_done, file_done, "uint64_t"); 635 636 static void 637 ctl_be_block_dispatch_file(struct ctl_be_block_lun *be_lun, 638 struct ctl_be_block_io *beio) 639 { 640 struct ctl_be_block_filedata *file_data; 641 union ctl_io *io; 642 struct uio xuio; 643 struct iovec *xiovec; 644 int vfs_is_locked, flags; 645 int error, i; 646 647 DPRINTF("entered\n"); 648 649 file_data = &be_lun->backend.file; 650 io = beio->io; 651 flags = beio->bio_flags; 652 653 if (beio->bio_cmd == BIO_READ) { 654 SDT_PROBE(cbb, kernel, read, file_start, 0, 0, 0, 0, 0); 655 } else { 656 SDT_PROBE(cbb, kernel, write, file_start, 0, 0, 0, 0, 0); 657 } 658 659 bzero(&xuio, sizeof(xuio)); 660 if (beio->bio_cmd == BIO_READ) 661 xuio.uio_rw = UIO_READ; 662 else 663 xuio.uio_rw = UIO_WRITE; 664 665 xuio.uio_offset = beio->io_offset; 666 xuio.uio_resid = beio->io_len; 667 xuio.uio_segflg = UIO_SYSSPACE; 668 xuio.uio_iov = beio->xiovecs; 669 xuio.uio_iovcnt = beio->num_segs; 670 xuio.uio_td = curthread; 671 672 for (i = 0, xiovec = xuio.uio_iov; i < xuio.uio_iovcnt; i++, xiovec++) { 673 xiovec->iov_base = beio->sg_segs[i].addr; 674 xiovec->iov_len = beio->sg_segs[i].len; 675 } 676 677 vfs_is_locked = VFS_LOCK_GIANT(be_lun->vn->v_mount); 678 if (beio->bio_cmd == BIO_READ) { 679 vn_lock(be_lun->vn, LK_SHARED | LK_RETRY); 680 681 binuptime(&beio->ds_t0); 682 devstat_start_transaction(beio->lun->disk_stats, &beio->ds_t0); 683 684 /* 685 * UFS pays attention to IO_DIRECT for reads. If the 686 * DIRECTIO option is configured into the kernel, it calls 687 * ffs_rawread(). But that only works for single-segment 688 * uios with user space addresses. In our case, with a 689 * kernel uio, it still reads into the buffer cache, but it 690 * will just try to release the buffer from the cache later 691 * on in ffs_read(). 692 * 693 * ZFS does not pay attention to IO_DIRECT for reads. 694 * 695 * UFS does not pay attention to IO_SYNC for reads. 696 * 697 * ZFS pays attention to IO_SYNC (which translates into the 698 * Solaris define FRSYNC for zfs_read()) for reads. It 699 * attempts to sync the file before reading. 700 * 701 * So, to attempt to provide some barrier semantics in the 702 * BIO_ORDERED case, set both IO_DIRECT and IO_SYNC. 703 */ 704 error = VOP_READ(be_lun->vn, &xuio, (flags & BIO_ORDERED) ? 705 (IO_DIRECT|IO_SYNC) : 0, file_data->cred); 706 707 VOP_UNLOCK(be_lun->vn, 0); 708 } else { 709 struct mount *mountpoint; 710 int lock_flags; 711 712 (void)vn_start_write(be_lun->vn, &mountpoint, V_WAIT); 713 714 if (MNT_SHARED_WRITES(mountpoint) 715 || ((mountpoint == NULL) 716 && MNT_SHARED_WRITES(be_lun->vn->v_mount))) 717 lock_flags = LK_SHARED; 718 else 719 lock_flags = LK_EXCLUSIVE; 720 721 vn_lock(be_lun->vn, lock_flags | LK_RETRY); 722 723 binuptime(&beio->ds_t0); 724 devstat_start_transaction(beio->lun->disk_stats, &beio->ds_t0); 725 726 /* 727 * UFS pays attention to IO_DIRECT for writes. The write 728 * is done asynchronously. (Normally the write would just 729 * get put into cache. 730 * 731 * UFS pays attention to IO_SYNC for writes. It will 732 * attempt to write the buffer out synchronously if that 733 * flag is set. 734 * 735 * ZFS does not pay attention to IO_DIRECT for writes. 736 * 737 * ZFS pays attention to IO_SYNC (a.k.a. FSYNC or FRSYNC) 738 * for writes. It will flush the transaction from the 739 * cache before returning. 740 * 741 * So if we've got the BIO_ORDERED flag set, we want 742 * IO_SYNC in either the UFS or ZFS case. 743 */ 744 error = VOP_WRITE(be_lun->vn, &xuio, (flags & BIO_ORDERED) ? 745 IO_SYNC : 0, file_data->cred); 746 VOP_UNLOCK(be_lun->vn, 0); 747 748 vn_finished_write(mountpoint); 749 } 750 VFS_UNLOCK_GIANT(vfs_is_locked); 751 752 /* 753 * If we got an error, set the sense data to "MEDIUM ERROR" and 754 * return the I/O to the user. 755 */ 756 if (error != 0) { 757 char path_str[32]; 758 759 ctl_scsi_path_string(io, path_str, sizeof(path_str)); 760 /* 761 * XXX KDM ZFS returns ENOSPC when the underlying 762 * filesystem fills up. What kind of SCSI error should we 763 * return for that? 764 */ 765 printf("%s%s command returned errno %d\n", path_str, 766 (beio->bio_cmd == BIO_READ) ? "READ" : "WRITE", error); 767 ctl_set_medium_error(&io->scsiio); 768 ctl_complete_beio(beio); 769 return; 770 } 771 772 /* 773 * If this is a write, we're all done. 774 * If this is a read, we can now send the data to the user. 775 */ 776 if (beio->bio_cmd == BIO_WRITE) { 777 ctl_set_success(&io->scsiio); 778 SDT_PROBE(cbb, kernel, write, file_done, 0, 0, 0, 0, 0); 779 ctl_complete_beio(beio); 780 } else { 781 SDT_PROBE(cbb, kernel, read, file_done, 0, 0, 0, 0, 0); 782 io->scsiio.be_move_done = ctl_be_block_move_done; 783 io->scsiio.kern_data_ptr = (uint8_t *)beio->sg_segs; 784 io->scsiio.kern_data_len = beio->io_len; 785 io->scsiio.kern_total_len = beio->io_len; 786 io->scsiio.kern_rel_offset = 0; 787 io->scsiio.kern_data_resid = 0; 788 io->scsiio.kern_sg_entries = beio->num_segs; 789 io->io_hdr.flags |= CTL_FLAG_ALLOCATED | CTL_FLAG_KDPTR_SGLIST; 790 #ifdef CTL_TIME_IO 791 getbintime(&io->io_hdr.dma_start_bt); 792 #endif 793 ctl_datamove(io); 794 } 795 } 796 797 static void 798 ctl_be_block_flush_dev(struct ctl_be_block_lun *be_lun, 799 struct ctl_be_block_io *beio) 800 { 801 struct bio *bio; 802 union ctl_io *io; 803 struct ctl_be_block_devdata *dev_data; 804 805 dev_data = &be_lun->backend.dev; 806 io = beio->io; 807 808 DPRINTF("entered\n"); 809 810 /* This can't fail, it's a blocking allocation. */ 811 bio = g_alloc_bio(); 812 813 bio->bio_cmd = BIO_FLUSH; 814 bio->bio_flags |= BIO_ORDERED; 815 bio->bio_dev = dev_data->cdev; 816 bio->bio_offset = 0; 817 bio->bio_data = 0; 818 bio->bio_done = ctl_be_block_biodone; 819 bio->bio_caller1 = beio; 820 bio->bio_pblkno = 0; 821 822 /* 823 * We don't need to acquire the LUN lock here, because we are only 824 * sending one bio, and so there is no other context to synchronize 825 * with. 826 */ 827 beio->num_bios_sent = 1; 828 beio->send_complete = 1; 829 830 binuptime(&beio->ds_t0); 831 devstat_start_transaction(be_lun->disk_stats, &beio->ds_t0); 832 833 (*dev_data->csw->d_strategy)(bio); 834 } 835 836 static void 837 ctl_be_block_dispatch_dev(struct ctl_be_block_lun *be_lun, 838 struct ctl_be_block_io *beio) 839 { 840 int i; 841 struct bio *bio; 842 struct ctl_be_block_devdata *dev_data; 843 off_t cur_offset; 844 int max_iosize; 845 846 DPRINTF("entered\n"); 847 848 dev_data = &be_lun->backend.dev; 849 850 /* 851 * We have to limit our I/O size to the maximum supported by the 852 * backend device. Hopefully it is MAXPHYS. If the driver doesn't 853 * set it properly, use DFLTPHYS. 854 */ 855 max_iosize = dev_data->cdev->si_iosize_max; 856 if (max_iosize < PAGE_SIZE) 857 max_iosize = DFLTPHYS; 858 859 cur_offset = beio->io_offset; 860 861 /* 862 * XXX KDM need to accurately reflect the number of I/Os outstanding 863 * to a device. 864 */ 865 binuptime(&beio->ds_t0); 866 devstat_start_transaction(be_lun->disk_stats, &beio->ds_t0); 867 868 for (i = 0; i < beio->num_segs; i++) { 869 size_t cur_size; 870 uint8_t *cur_ptr; 871 872 cur_size = beio->sg_segs[i].len; 873 cur_ptr = beio->sg_segs[i].addr; 874 875 while (cur_size > 0) { 876 /* This can't fail, it's a blocking allocation. */ 877 bio = g_alloc_bio(); 878 879 KASSERT(bio != NULL, ("g_alloc_bio() failed!\n")); 880 881 bio->bio_cmd = beio->bio_cmd; 882 bio->bio_flags |= beio->bio_flags; 883 bio->bio_dev = dev_data->cdev; 884 bio->bio_caller1 = beio; 885 bio->bio_length = min(cur_size, max_iosize); 886 bio->bio_offset = cur_offset; 887 bio->bio_data = cur_ptr; 888 bio->bio_done = ctl_be_block_biodone; 889 bio->bio_pblkno = cur_offset / be_lun->blocksize; 890 891 cur_offset += bio->bio_length; 892 cur_ptr += bio->bio_length; 893 cur_size -= bio->bio_length; 894 895 /* 896 * Make sure we set the complete bit just before we 897 * issue the last bio so we don't wind up with a 898 * race. 899 * 900 * Use the LUN mutex here instead of a combination 901 * of atomic variables for simplicity. 902 * 903 * XXX KDM we could have a per-IO lock, but that 904 * would cause additional per-IO setup and teardown 905 * overhead. Hopefully there won't be too much 906 * contention on the LUN lock. 907 */ 908 mtx_lock(&be_lun->lock); 909 910 beio->num_bios_sent++; 911 912 if ((i == beio->num_segs - 1) 913 && (cur_size == 0)) 914 beio->send_complete = 1; 915 916 mtx_unlock(&be_lun->lock); 917 918 (*dev_data->csw->d_strategy)(bio); 919 } 920 } 921 } 922 923 static void 924 ctl_be_block_cw_dispatch(struct ctl_be_block_lun *be_lun, 925 union ctl_io *io) 926 { 927 struct ctl_be_block_io *beio; 928 struct ctl_be_block_softc *softc; 929 930 DPRINTF("entered\n"); 931 932 softc = be_lun->softc; 933 beio = ctl_alloc_beio(softc); 934 if (beio == NULL) { 935 /* 936 * This should not happen. ctl_alloc_beio() will call 937 * ctl_grow_beio() with a blocking malloc as needed. 938 * A malloc with M_WAITOK should not fail. 939 */ 940 ctl_set_busy(&io->scsiio); 941 ctl_done(io); 942 return; 943 } 944 945 beio->io = io; 946 beio->softc = softc; 947 beio->lun = be_lun; 948 io->io_hdr.ctl_private[CTL_PRIV_BACKEND].ptr = beio; 949 950 switch (io->scsiio.cdb[0]) { 951 case SYNCHRONIZE_CACHE: 952 case SYNCHRONIZE_CACHE_16: 953 beio->ds_trans_type = DEVSTAT_NO_DATA; 954 beio->ds_tag_type = DEVSTAT_TAG_ORDERED; 955 beio->io_len = 0; 956 be_lun->lun_flush(be_lun, beio); 957 break; 958 default: 959 panic("Unhandled CDB type %#x", io->scsiio.cdb[0]); 960 break; 961 } 962 } 963 964 SDT_PROBE_DEFINE1(cbb, kernel, read, start, start, "uint64_t"); 965 SDT_PROBE_DEFINE1(cbb, kernel, write, start, start, "uint64_t"); 966 SDT_PROBE_DEFINE1(cbb, kernel, read, alloc_done, alloc_done, "uint64_t"); 967 SDT_PROBE_DEFINE1(cbb, kernel, write, alloc_done, alloc_done, "uint64_t"); 968 969 static void 970 ctl_be_block_dispatch(struct ctl_be_block_lun *be_lun, 971 union ctl_io *io) 972 { 973 struct ctl_be_block_io *beio; 974 struct ctl_be_block_softc *softc; 975 struct ctl_lba_len lbalen; 976 uint64_t len_left, io_size_bytes; 977 int i; 978 979 softc = be_lun->softc; 980 981 DPRINTF("entered\n"); 982 983 if ((io->io_hdr.flags & CTL_FLAG_DATA_MASK) == CTL_FLAG_DATA_IN) { 984 SDT_PROBE(cbb, kernel, read, start, 0, 0, 0, 0, 0); 985 } else { 986 SDT_PROBE(cbb, kernel, write, start, 0, 0, 0, 0, 0); 987 } 988 989 memcpy(&lbalen, io->io_hdr.ctl_private[CTL_PRIV_LBA_LEN].bytes, 990 sizeof(lbalen)); 991 992 io_size_bytes = lbalen.len * be_lun->blocksize; 993 994 /* 995 * XXX KDM this is temporary, until we implement chaining of beio 996 * structures and multiple datamove calls to move all the data in 997 * or out. 998 */ 999 if (io_size_bytes > CTLBLK_MAX_IO_SIZE) { 1000 printf("%s: IO length %ju > max io size %u\n", __func__, 1001 io_size_bytes, CTLBLK_MAX_IO_SIZE); 1002 ctl_set_invalid_field(&io->scsiio, 1003 /*sks_valid*/ 0, 1004 /*command*/ 1, 1005 /*field*/ 0, 1006 /*bit_valid*/ 0, 1007 /*bit*/ 0); 1008 ctl_done(io); 1009 return; 1010 } 1011 1012 beio = ctl_alloc_beio(softc); 1013 if (beio == NULL) { 1014 /* 1015 * This should not happen. ctl_alloc_beio() will call 1016 * ctl_grow_beio() with a blocking malloc as needed. 1017 * A malloc with M_WAITOK should not fail. 1018 */ 1019 ctl_set_busy(&io->scsiio); 1020 ctl_done(io); 1021 return; 1022 } 1023 1024 beio->io = io; 1025 beio->softc = softc; 1026 beio->lun = be_lun; 1027 io->io_hdr.ctl_private[CTL_PRIV_BACKEND].ptr = beio; 1028 1029 /* 1030 * If the I/O came down with an ordered or head of queue tag, set 1031 * the BIO_ORDERED attribute. For head of queue tags, that's 1032 * pretty much the best we can do. 1033 * 1034 * XXX KDM we don't have a great way to easily know about the FUA 1035 * bit right now (it is decoded in ctl_read_write(), but we don't 1036 * pass that knowledge to the backend), and in any case we would 1037 * need to determine how to handle it. 1038 */ 1039 if ((io->scsiio.tag_type == CTL_TAG_ORDERED) 1040 || (io->scsiio.tag_type == CTL_TAG_HEAD_OF_QUEUE)) 1041 beio->bio_flags = BIO_ORDERED; 1042 1043 switch (io->scsiio.tag_type) { 1044 case CTL_TAG_ORDERED: 1045 beio->ds_tag_type = DEVSTAT_TAG_ORDERED; 1046 break; 1047 case CTL_TAG_HEAD_OF_QUEUE: 1048 beio->ds_tag_type = DEVSTAT_TAG_HEAD; 1049 break; 1050 case CTL_TAG_UNTAGGED: 1051 case CTL_TAG_SIMPLE: 1052 case CTL_TAG_ACA: 1053 default: 1054 beio->ds_tag_type = DEVSTAT_TAG_SIMPLE; 1055 break; 1056 } 1057 1058 /* 1059 * This path handles read and write only. The config write path 1060 * handles flush operations. 1061 */ 1062 if ((io->io_hdr.flags & CTL_FLAG_DATA_MASK) == CTL_FLAG_DATA_IN) { 1063 beio->bio_cmd = BIO_READ; 1064 beio->ds_trans_type = DEVSTAT_READ; 1065 } else { 1066 beio->bio_cmd = BIO_WRITE; 1067 beio->ds_trans_type = DEVSTAT_WRITE; 1068 } 1069 1070 beio->io_len = lbalen.len * be_lun->blocksize; 1071 beio->io_offset = lbalen.lba * be_lun->blocksize; 1072 1073 DPRINTF("%s at LBA %jx len %u\n", 1074 (beio->bio_cmd == BIO_READ) ? "READ" : "WRITE", 1075 (uintmax_t)lbalen.lba, lbalen.len); 1076 1077 for (i = 0, len_left = io_size_bytes; i < CTLBLK_MAX_SEGS && 1078 len_left > 0; i++) { 1079 1080 /* 1081 * Setup the S/G entry for this chunk. 1082 */ 1083 beio->sg_segs[i].len = min(MAXPHYS, len_left); 1084 beio->sg_segs[i].addr = uma_zalloc(be_lun->lun_zone, M_WAITOK); 1085 /* 1086 * uma_zalloc() can in theory return NULL even with M_WAITOK 1087 * if it can't pull more memory into the zone. 1088 */ 1089 if (beio->sg_segs[i].addr == NULL) { 1090 ctl_set_busy(&io->scsiio); 1091 ctl_complete_beio(beio); 1092 return; 1093 } 1094 1095 DPRINTF("segment %d addr %p len %zd\n", i, 1096 beio->sg_segs[i].addr, beio->sg_segs[i].len); 1097 1098 beio->num_segs++; 1099 len_left -= beio->sg_segs[i].len; 1100 } 1101 1102 /* 1103 * For the read case, we need to read the data into our buffers and 1104 * then we can send it back to the user. For the write case, we 1105 * need to get the data from the user first. 1106 */ 1107 if (beio->bio_cmd == BIO_READ) { 1108 SDT_PROBE(cbb, kernel, read, alloc_done, 0, 0, 0, 0, 0); 1109 be_lun->dispatch(be_lun, beio); 1110 } else { 1111 SDT_PROBE(cbb, kernel, write, alloc_done, 0, 0, 0, 0, 0); 1112 io->scsiio.be_move_done = ctl_be_block_move_done; 1113 io->scsiio.kern_data_ptr = (uint8_t *)beio->sg_segs; 1114 io->scsiio.kern_data_len = beio->io_len; 1115 io->scsiio.kern_total_len = beio->io_len; 1116 io->scsiio.kern_rel_offset = 0; 1117 io->scsiio.kern_data_resid = 0; 1118 io->scsiio.kern_sg_entries = beio->num_segs; 1119 io->io_hdr.flags |= CTL_FLAG_ALLOCATED | CTL_FLAG_KDPTR_SGLIST; 1120 #ifdef CTL_TIME_IO 1121 getbintime(&io->io_hdr.dma_start_bt); 1122 #endif 1123 ctl_datamove(io); 1124 } 1125 } 1126 1127 static void 1128 ctl_be_block_worker(void *context, int pending) 1129 { 1130 struct ctl_be_block_lun *be_lun; 1131 struct ctl_be_block_softc *softc; 1132 union ctl_io *io; 1133 1134 be_lun = (struct ctl_be_block_lun *)context; 1135 softc = be_lun->softc; 1136 1137 DPRINTF("entered\n"); 1138 1139 mtx_lock(&be_lun->lock); 1140 for (;;) { 1141 io = (union ctl_io *)STAILQ_FIRST(&be_lun->datamove_queue); 1142 if (io != NULL) { 1143 struct ctl_be_block_io *beio; 1144 1145 DPRINTF("datamove queue\n"); 1146 1147 STAILQ_REMOVE(&be_lun->datamove_queue, &io->io_hdr, 1148 ctl_io_hdr, links); 1149 1150 mtx_unlock(&be_lun->lock); 1151 1152 beio = (struct ctl_be_block_io *) 1153 io->io_hdr.ctl_private[CTL_PRIV_BACKEND].ptr; 1154 1155 be_lun->dispatch(be_lun, beio); 1156 1157 mtx_lock(&be_lun->lock); 1158 continue; 1159 } 1160 io = (union ctl_io *)STAILQ_FIRST(&be_lun->config_write_queue); 1161 if (io != NULL) { 1162 1163 DPRINTF("config write queue\n"); 1164 1165 STAILQ_REMOVE(&be_lun->config_write_queue, &io->io_hdr, 1166 ctl_io_hdr, links); 1167 1168 mtx_unlock(&be_lun->lock); 1169 1170 ctl_be_block_cw_dispatch(be_lun, io); 1171 1172 mtx_lock(&be_lun->lock); 1173 continue; 1174 } 1175 io = (union ctl_io *)STAILQ_FIRST(&be_lun->input_queue); 1176 if (io != NULL) { 1177 DPRINTF("input queue\n"); 1178 1179 STAILQ_REMOVE(&be_lun->input_queue, &io->io_hdr, 1180 ctl_io_hdr, links); 1181 mtx_unlock(&be_lun->lock); 1182 1183 /* 1184 * We must drop the lock, since this routine and 1185 * its children may sleep. 1186 */ 1187 ctl_be_block_dispatch(be_lun, io); 1188 1189 mtx_lock(&be_lun->lock); 1190 continue; 1191 } 1192 1193 /* 1194 * If we get here, there is no work left in the queues, so 1195 * just break out and let the task queue go to sleep. 1196 */ 1197 break; 1198 } 1199 mtx_unlock(&be_lun->lock); 1200 } 1201 1202 /* 1203 * Entry point from CTL to the backend for I/O. We queue everything to a 1204 * work thread, so this just puts the I/O on a queue and wakes up the 1205 * thread. 1206 */ 1207 static int 1208 ctl_be_block_submit(union ctl_io *io) 1209 { 1210 struct ctl_be_block_lun *be_lun; 1211 struct ctl_be_lun *ctl_be_lun; 1212 int retval; 1213 1214 DPRINTF("entered\n"); 1215 1216 retval = CTL_RETVAL_COMPLETE; 1217 1218 ctl_be_lun = (struct ctl_be_lun *)io->io_hdr.ctl_private[ 1219 CTL_PRIV_BACKEND_LUN].ptr; 1220 be_lun = (struct ctl_be_block_lun *)ctl_be_lun->be_lun; 1221 1222 /* 1223 * Make sure we only get SCSI I/O. 1224 */ 1225 KASSERT(io->io_hdr.io_type == CTL_IO_SCSI, ("Non-SCSI I/O (type " 1226 "%#x) encountered", io->io_hdr.io_type)); 1227 1228 mtx_lock(&be_lun->lock); 1229 /* 1230 * XXX KDM make sure that links is okay to use at this point. 1231 * Otherwise, we either need to add another field to ctl_io_hdr, 1232 * or deal with resource allocation here. 1233 */ 1234 STAILQ_INSERT_TAIL(&be_lun->input_queue, &io->io_hdr, links); 1235 mtx_unlock(&be_lun->lock); 1236 1237 taskqueue_enqueue(be_lun->io_taskqueue, &be_lun->io_task); 1238 1239 return (retval); 1240 } 1241 1242 static int 1243 ctl_be_block_ioctl(struct cdev *dev, u_long cmd, caddr_t addr, 1244 int flag, struct thread *td) 1245 { 1246 struct ctl_be_block_softc *softc; 1247 int error; 1248 1249 softc = &backend_block_softc; 1250 1251 error = 0; 1252 1253 switch (cmd) { 1254 case CTL_LUN_REQ: { 1255 struct ctl_lun_req *lun_req; 1256 1257 lun_req = (struct ctl_lun_req *)addr; 1258 1259 switch (lun_req->reqtype) { 1260 case CTL_LUNREQ_CREATE: 1261 error = ctl_be_block_create(softc, lun_req); 1262 break; 1263 case CTL_LUNREQ_RM: 1264 error = ctl_be_block_rm(softc, lun_req); 1265 break; 1266 default: 1267 lun_req->status = CTL_LUN_ERROR; 1268 snprintf(lun_req->error_str, sizeof(lun_req->error_str), 1269 "%s: invalid LUN request type %d", __func__, 1270 lun_req->reqtype); 1271 break; 1272 } 1273 break; 1274 } 1275 default: 1276 error = ENOTTY; 1277 break; 1278 } 1279 1280 return (error); 1281 } 1282 1283 static int 1284 ctl_be_block_open_file(struct ctl_be_block_lun *be_lun, struct ctl_lun_req *req) 1285 { 1286 struct ctl_be_block_filedata *file_data; 1287 struct ctl_lun_create_params *params; 1288 struct vattr vattr; 1289 int error; 1290 1291 error = 0; 1292 file_data = &be_lun->backend.file; 1293 params = &req->reqdata.create; 1294 1295 be_lun->dev_type = CTL_BE_BLOCK_FILE; 1296 be_lun->dispatch = ctl_be_block_dispatch_file; 1297 be_lun->lun_flush = ctl_be_block_flush_file; 1298 1299 error = VOP_GETATTR(be_lun->vn, &vattr, curthread->td_ucred); 1300 if (error != 0) { 1301 snprintf(req->error_str, sizeof(req->error_str), 1302 "error calling VOP_GETATTR() for file %s", 1303 be_lun->dev_path); 1304 return (error); 1305 } 1306 1307 /* 1308 * Verify that we have the ability to upgrade to exclusive 1309 * access on this file so we can trap errors at open instead 1310 * of reporting them during first access. 1311 */ 1312 if (VOP_ISLOCKED(be_lun->vn) != LK_EXCLUSIVE) { 1313 vn_lock(be_lun->vn, LK_UPGRADE | LK_RETRY); 1314 if (be_lun->vn->v_iflag & VI_DOOMED) { 1315 error = EBADF; 1316 snprintf(req->error_str, sizeof(req->error_str), 1317 "error locking file %s", be_lun->dev_path); 1318 return (error); 1319 } 1320 } 1321 1322 1323 file_data->cred = crhold(curthread->td_ucred); 1324 be_lun->size_bytes = vattr.va_size; 1325 /* 1326 * We set the multi thread flag for file operations because all 1327 * filesystems (in theory) are capable of allowing multiple readers 1328 * of a file at once. So we want to get the maximum possible 1329 * concurrency. 1330 */ 1331 be_lun->flags |= CTL_BE_BLOCK_LUN_MULTI_THREAD; 1332 1333 /* 1334 * XXX KDM vattr.va_blocksize may be larger than 512 bytes here. 1335 * With ZFS, it is 131072 bytes. Block sizes that large don't work 1336 * with disklabel and UFS on FreeBSD at least. Large block sizes 1337 * may not work with other OSes as well. So just export a sector 1338 * size of 512 bytes, which should work with any OS or 1339 * application. Since our backing is a file, any block size will 1340 * work fine for the backing store. 1341 */ 1342 #if 0 1343 be_lun->blocksize= vattr.va_blocksize; 1344 #endif 1345 if (params->blocksize_bytes != 0) 1346 be_lun->blocksize = params->blocksize_bytes; 1347 else 1348 be_lun->blocksize = 512; 1349 1350 /* 1351 * Sanity check. The media size has to be at least one 1352 * sector long. 1353 */ 1354 if (be_lun->size_bytes < be_lun->blocksize) { 1355 error = EINVAL; 1356 snprintf(req->error_str, sizeof(req->error_str), 1357 "file %s size %ju < block size %u", be_lun->dev_path, 1358 (uintmax_t)be_lun->size_bytes, be_lun->blocksize); 1359 } 1360 return (error); 1361 } 1362 1363 static int 1364 ctl_be_block_open_dev(struct ctl_be_block_lun *be_lun, struct ctl_lun_req *req) 1365 { 1366 struct ctl_lun_create_params *params; 1367 struct vattr vattr; 1368 struct cdev *dev; 1369 struct cdevsw *devsw; 1370 int error; 1371 1372 params = &req->reqdata.create; 1373 1374 be_lun->dev_type = CTL_BE_BLOCK_DEV; 1375 be_lun->dispatch = ctl_be_block_dispatch_dev; 1376 be_lun->lun_flush = ctl_be_block_flush_dev; 1377 be_lun->backend.dev.cdev = be_lun->vn->v_rdev; 1378 be_lun->backend.dev.csw = dev_refthread(be_lun->backend.dev.cdev, 1379 &be_lun->backend.dev.dev_ref); 1380 if (be_lun->backend.dev.csw == NULL) 1381 panic("Unable to retrieve device switch"); 1382 1383 error = VOP_GETATTR(be_lun->vn, &vattr, NOCRED); 1384 if (error) { 1385 snprintf(req->error_str, sizeof(req->error_str), 1386 "%s: error getting vnode attributes for device %s", 1387 __func__, be_lun->dev_path); 1388 return (error); 1389 } 1390 1391 dev = be_lun->vn->v_rdev; 1392 devsw = dev->si_devsw; 1393 if (!devsw->d_ioctl) { 1394 snprintf(req->error_str, sizeof(req->error_str), 1395 "%s: no d_ioctl for device %s!", __func__, 1396 be_lun->dev_path); 1397 return (ENODEV); 1398 } 1399 1400 error = devsw->d_ioctl(dev, DIOCGSECTORSIZE, 1401 (caddr_t)&be_lun->blocksize, FREAD, 1402 curthread); 1403 if (error) { 1404 snprintf(req->error_str, sizeof(req->error_str), 1405 "%s: error %d returned for DIOCGSECTORSIZE ioctl " 1406 "on %s!", __func__, error, be_lun->dev_path); 1407 return (error); 1408 } 1409 1410 /* 1411 * If the user has asked for a blocksize that is greater than the 1412 * backing device's blocksize, we can do it only if the blocksize 1413 * the user is asking for is an even multiple of the underlying 1414 * device's blocksize. 1415 */ 1416 if ((params->blocksize_bytes != 0) 1417 && (params->blocksize_bytes > be_lun->blocksize)) { 1418 uint32_t bs_multiple, tmp_blocksize; 1419 1420 bs_multiple = params->blocksize_bytes / be_lun->blocksize; 1421 1422 tmp_blocksize = bs_multiple * be_lun->blocksize; 1423 1424 if (tmp_blocksize == params->blocksize_bytes) { 1425 be_lun->blocksize = params->blocksize_bytes; 1426 } else { 1427 snprintf(req->error_str, sizeof(req->error_str), 1428 "%s: requested blocksize %u is not an even " 1429 "multiple of backing device blocksize %u", 1430 __func__, params->blocksize_bytes, 1431 be_lun->blocksize); 1432 return (EINVAL); 1433 1434 } 1435 } else if ((params->blocksize_bytes != 0) 1436 && (params->blocksize_bytes != be_lun->blocksize)) { 1437 snprintf(req->error_str, sizeof(req->error_str), 1438 "%s: requested blocksize %u < backing device " 1439 "blocksize %u", __func__, params->blocksize_bytes, 1440 be_lun->blocksize); 1441 return (EINVAL); 1442 } 1443 1444 error = devsw->d_ioctl(dev, DIOCGMEDIASIZE, 1445 (caddr_t)&be_lun->size_bytes, FREAD, 1446 curthread); 1447 if (error) { 1448 snprintf(req->error_str, sizeof(req->error_str), 1449 "%s: error %d returned for DIOCGMEDIASIZE ioctl " 1450 "on %s!", __func__, error, be_lun->dev_path); 1451 return (error); 1452 } 1453 1454 return (0); 1455 1456 } 1457 1458 1459 static int 1460 ctl_be_block_close(struct ctl_be_block_lun *be_lun) 1461 { 1462 DROP_GIANT(); 1463 if (be_lun->vn) { 1464 int flags = FREAD | FWRITE; 1465 int vfs_is_locked = 0; 1466 1467 switch (be_lun->dev_type) { 1468 case CTL_BE_BLOCK_DEV: 1469 if (be_lun->backend.dev.csw) { 1470 dev_relthread(be_lun->backend.dev.cdev, 1471 be_lun->backend.dev.dev_ref); 1472 be_lun->backend.dev.csw = NULL; 1473 be_lun->backend.dev.cdev = NULL; 1474 } 1475 break; 1476 case CTL_BE_BLOCK_FILE: 1477 vfs_is_locked = VFS_LOCK_GIANT(be_lun->vn->v_mount); 1478 break; 1479 case CTL_BE_BLOCK_NONE: 1480 default: 1481 panic("Unexpected backend type."); 1482 break; 1483 } 1484 1485 (void)vn_close(be_lun->vn, flags, NOCRED, curthread); 1486 be_lun->vn = NULL; 1487 1488 switch (be_lun->dev_type) { 1489 case CTL_BE_BLOCK_DEV: 1490 break; 1491 case CTL_BE_BLOCK_FILE: 1492 VFS_UNLOCK_GIANT(vfs_is_locked); 1493 if (be_lun->backend.file.cred != NULL) { 1494 crfree(be_lun->backend.file.cred); 1495 be_lun->backend.file.cred = NULL; 1496 } 1497 break; 1498 case CTL_BE_BLOCK_NONE: 1499 default: 1500 panic("Unexpected backend type."); 1501 break; 1502 } 1503 } 1504 PICKUP_GIANT(); 1505 1506 return (0); 1507 } 1508 1509 static int 1510 ctl_be_block_open(struct ctl_be_block_softc *softc, 1511 struct ctl_be_block_lun *be_lun, struct ctl_lun_req *req) 1512 { 1513 struct nameidata nd; 1514 int flags; 1515 int error; 1516 int vfs_is_locked; 1517 1518 /* 1519 * XXX KDM allow a read-only option? 1520 */ 1521 flags = FREAD | FWRITE; 1522 error = 0; 1523 1524 if (rootvnode == NULL) { 1525 snprintf(req->error_str, sizeof(req->error_str), 1526 "%s: Root filesystem is not mounted", __func__); 1527 return (1); 1528 } 1529 1530 if (!curthread->td_proc->p_fd->fd_cdir) { 1531 curthread->td_proc->p_fd->fd_cdir = rootvnode; 1532 VREF(rootvnode); 1533 } 1534 if (!curthread->td_proc->p_fd->fd_rdir) { 1535 curthread->td_proc->p_fd->fd_rdir = rootvnode; 1536 VREF(rootvnode); 1537 } 1538 if (!curthread->td_proc->p_fd->fd_jdir) { 1539 curthread->td_proc->p_fd->fd_jdir = rootvnode; 1540 VREF(rootvnode); 1541 } 1542 1543 again: 1544 NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, be_lun->dev_path, curthread); 1545 error = vn_open(&nd, &flags, 0, NULL); 1546 if (error) { 1547 /* 1548 * This is the only reasonable guess we can make as far as 1549 * path if the user doesn't give us a fully qualified path. 1550 * If they want to specify a file, they need to specify the 1551 * full path. 1552 */ 1553 if (be_lun->dev_path[0] != '/') { 1554 char *dev_path = "/dev/"; 1555 char *dev_name; 1556 1557 /* Try adding device path at beginning of name */ 1558 dev_name = malloc(strlen(be_lun->dev_path) 1559 + strlen(dev_path) + 1, 1560 M_CTLBLK, M_WAITOK); 1561 if (dev_name) { 1562 sprintf(dev_name, "%s%s", dev_path, 1563 be_lun->dev_path); 1564 free(be_lun->dev_path, M_CTLBLK); 1565 be_lun->dev_path = dev_name; 1566 goto again; 1567 } 1568 } 1569 snprintf(req->error_str, sizeof(req->error_str), 1570 "%s: error opening %s", __func__, be_lun->dev_path); 1571 return (error); 1572 } 1573 1574 vfs_is_locked = NDHASGIANT(&nd); 1575 1576 NDFREE(&nd, NDF_ONLY_PNBUF); 1577 1578 be_lun->vn = nd.ni_vp; 1579 1580 /* We only support disks and files. */ 1581 if (vn_isdisk(be_lun->vn, &error)) { 1582 error = ctl_be_block_open_dev(be_lun, req); 1583 } else if (be_lun->vn->v_type == VREG) { 1584 error = ctl_be_block_open_file(be_lun, req); 1585 } else { 1586 error = EINVAL; 1587 snprintf(req->error_str, sizeof(req->error_str), 1588 "%s is not a disk or file", be_lun->dev_path); 1589 } 1590 VOP_UNLOCK(be_lun->vn, 0); 1591 VFS_UNLOCK_GIANT(vfs_is_locked); 1592 1593 if (error != 0) { 1594 ctl_be_block_close(be_lun); 1595 return (error); 1596 } 1597 1598 be_lun->blocksize_shift = fls(be_lun->blocksize) - 1; 1599 be_lun->size_blocks = be_lun->size_bytes >> be_lun->blocksize_shift; 1600 1601 return (0); 1602 1603 } 1604 1605 static int 1606 ctl_be_block_mem_ctor(void *mem, int size, void *arg, int flags) 1607 { 1608 return (0); 1609 } 1610 1611 static void 1612 ctl_be_block_mem_dtor(void *mem, int size, void *arg) 1613 { 1614 bzero(mem, size); 1615 } 1616 1617 static int 1618 ctl_be_block_create(struct ctl_be_block_softc *softc, struct ctl_lun_req *req) 1619 { 1620 struct ctl_be_block_lun *be_lun; 1621 struct ctl_lun_create_params *params; 1622 struct ctl_be_arg *file_arg; 1623 char tmpstr[32]; 1624 int retval, num_threads; 1625 int i; 1626 1627 params = &req->reqdata.create; 1628 retval = 0; 1629 1630 num_threads = cbb_num_threads; 1631 1632 file_arg = NULL; 1633 1634 be_lun = malloc(sizeof(*be_lun), M_CTLBLK, M_ZERO | M_WAITOK); 1635 1636 if (be_lun == NULL) { 1637 snprintf(req->error_str, sizeof(req->error_str), 1638 "%s: error allocating %zd bytes", __func__, 1639 sizeof(*be_lun)); 1640 goto bailout_error; 1641 } 1642 1643 be_lun->softc = softc; 1644 STAILQ_INIT(&be_lun->input_queue); 1645 STAILQ_INIT(&be_lun->config_write_queue); 1646 STAILQ_INIT(&be_lun->datamove_queue); 1647 sprintf(be_lun->lunname, "cblk%d", softc->num_luns); 1648 mtx_init(&be_lun->lock, be_lun->lunname, NULL, MTX_DEF); 1649 1650 be_lun->lun_zone = uma_zcreate(be_lun->lunname, MAXPHYS, 1651 ctl_be_block_mem_ctor, ctl_be_block_mem_dtor, NULL, NULL, 1652 /*align*/ 0, /*flags*/0); 1653 1654 if (be_lun->lun_zone == NULL) { 1655 snprintf(req->error_str, sizeof(req->error_str), 1656 "%s: error allocating UMA zone", __func__); 1657 goto bailout_error; 1658 } 1659 1660 if (params->flags & CTL_LUN_FLAG_DEV_TYPE) 1661 be_lun->ctl_be_lun.lun_type = params->device_type; 1662 else 1663 be_lun->ctl_be_lun.lun_type = T_DIRECT; 1664 1665 if (be_lun->ctl_be_lun.lun_type == T_DIRECT) { 1666 for (i = 0; i < req->num_be_args; i++) { 1667 if (strcmp(req->kern_be_args[i].name, "file") == 0) { 1668 file_arg = &req->kern_be_args[i]; 1669 break; 1670 } 1671 } 1672 1673 if (file_arg == NULL) { 1674 snprintf(req->error_str, sizeof(req->error_str), 1675 "%s: no file argument specified", __func__); 1676 goto bailout_error; 1677 } 1678 1679 be_lun->dev_path = malloc(file_arg->vallen, M_CTLBLK, 1680 M_WAITOK | M_ZERO); 1681 if (be_lun->dev_path == NULL) { 1682 snprintf(req->error_str, sizeof(req->error_str), 1683 "%s: error allocating %d bytes", __func__, 1684 file_arg->vallen); 1685 goto bailout_error; 1686 } 1687 1688 strlcpy(be_lun->dev_path, (char *)file_arg->value, 1689 file_arg->vallen); 1690 1691 retval = ctl_be_block_open(softc, be_lun, req); 1692 if (retval != 0) { 1693 retval = 0; 1694 goto bailout_error; 1695 } 1696 1697 /* 1698 * Tell the user the size of the file/device. 1699 */ 1700 params->lun_size_bytes = be_lun->size_bytes; 1701 1702 /* 1703 * The maximum LBA is the size - 1. 1704 */ 1705 be_lun->ctl_be_lun.maxlba = be_lun->size_blocks - 1; 1706 } else { 1707 /* 1708 * For processor devices, we don't have any size. 1709 */ 1710 be_lun->blocksize = 0; 1711 be_lun->size_blocks = 0; 1712 be_lun->size_bytes = 0; 1713 be_lun->ctl_be_lun.maxlba = 0; 1714 params->lun_size_bytes = 0; 1715 1716 /* 1717 * Default to just 1 thread for processor devices. 1718 */ 1719 num_threads = 1; 1720 } 1721 1722 /* 1723 * XXX This searching loop might be refactored to be combined with 1724 * the loop above, 1725 */ 1726 for (i = 0; i < req->num_be_args; i++) { 1727 if (strcmp(req->kern_be_args[i].name, "num_threads") == 0) { 1728 struct ctl_be_arg *thread_arg; 1729 char num_thread_str[16]; 1730 int tmp_num_threads; 1731 1732 1733 thread_arg = &req->kern_be_args[i]; 1734 1735 strlcpy(num_thread_str, (char *)thread_arg->value, 1736 min(thread_arg->vallen, 1737 sizeof(num_thread_str))); 1738 1739 tmp_num_threads = strtol(num_thread_str, NULL, 0); 1740 1741 /* 1742 * We don't let the user specify less than one 1743 * thread, but hope he's clueful enough not to 1744 * specify 1000 threads. 1745 */ 1746 if (tmp_num_threads < 1) { 1747 snprintf(req->error_str, sizeof(req->error_str), 1748 "%s: invalid number of threads %s", 1749 __func__, num_thread_str); 1750 goto bailout_error; 1751 } 1752 1753 num_threads = tmp_num_threads; 1754 } 1755 } 1756 1757 be_lun->flags = CTL_BE_BLOCK_LUN_UNCONFIGURED; 1758 be_lun->ctl_be_lun.flags = CTL_LUN_FLAG_PRIMARY; 1759 be_lun->ctl_be_lun.be_lun = be_lun; 1760 be_lun->ctl_be_lun.blocksize = be_lun->blocksize; 1761 /* Tell the user the blocksize we ended up using */ 1762 params->blocksize_bytes = be_lun->blocksize; 1763 if (params->flags & CTL_LUN_FLAG_ID_REQ) { 1764 be_lun->ctl_be_lun.req_lun_id = params->req_lun_id; 1765 be_lun->ctl_be_lun.flags |= CTL_LUN_FLAG_ID_REQ; 1766 } else 1767 be_lun->ctl_be_lun.req_lun_id = 0; 1768 1769 be_lun->ctl_be_lun.lun_shutdown = ctl_be_block_lun_shutdown; 1770 be_lun->ctl_be_lun.lun_config_status = 1771 ctl_be_block_lun_config_status; 1772 be_lun->ctl_be_lun.be = &ctl_be_block_driver; 1773 1774 if ((params->flags & CTL_LUN_FLAG_SERIAL_NUM) == 0) { 1775 snprintf(tmpstr, sizeof(tmpstr), "MYSERIAL%4d", 1776 softc->num_luns); 1777 strncpy((char *)be_lun->ctl_be_lun.serial_num, tmpstr, 1778 ctl_min(sizeof(be_lun->ctl_be_lun.serial_num), 1779 sizeof(tmpstr))); 1780 1781 /* Tell the user what we used for a serial number */ 1782 strncpy((char *)params->serial_num, tmpstr, 1783 ctl_min(sizeof(params->serial_num), sizeof(tmpstr))); 1784 } else { 1785 strncpy((char *)be_lun->ctl_be_lun.serial_num, 1786 params->serial_num, 1787 ctl_min(sizeof(be_lun->ctl_be_lun.serial_num), 1788 sizeof(params->serial_num))); 1789 } 1790 if ((params->flags & CTL_LUN_FLAG_DEVID) == 0) { 1791 snprintf(tmpstr, sizeof(tmpstr), "MYDEVID%4d", softc->num_luns); 1792 strncpy((char *)be_lun->ctl_be_lun.device_id, tmpstr, 1793 ctl_min(sizeof(be_lun->ctl_be_lun.device_id), 1794 sizeof(tmpstr))); 1795 1796 /* Tell the user what we used for a device ID */ 1797 strncpy((char *)params->device_id, tmpstr, 1798 ctl_min(sizeof(params->device_id), sizeof(tmpstr))); 1799 } else { 1800 strncpy((char *)be_lun->ctl_be_lun.device_id, 1801 params->device_id, 1802 ctl_min(sizeof(be_lun->ctl_be_lun.device_id), 1803 sizeof(params->device_id))); 1804 } 1805 1806 TASK_INIT(&be_lun->io_task, /*priority*/0, ctl_be_block_worker, be_lun); 1807 1808 be_lun->io_taskqueue = taskqueue_create(be_lun->lunname, M_WAITOK, 1809 taskqueue_thread_enqueue, /*context*/&be_lun->io_taskqueue); 1810 1811 if (be_lun->io_taskqueue == NULL) { 1812 snprintf(req->error_str, sizeof(req->error_str), 1813 "%s: Unable to create taskqueue", __func__); 1814 goto bailout_error; 1815 } 1816 1817 /* 1818 * Note that we start the same number of threads by default for 1819 * both the file case and the block device case. For the file 1820 * case, we need multiple threads to allow concurrency, because the 1821 * vnode interface is designed to be a blocking interface. For the 1822 * block device case, ZFS zvols at least will block the caller's 1823 * context in many instances, and so we need multiple threads to 1824 * overcome that problem. Other block devices don't need as many 1825 * threads, but they shouldn't cause too many problems. 1826 * 1827 * If the user wants to just have a single thread for a block 1828 * device, he can specify that when the LUN is created, or change 1829 * the tunable/sysctl to alter the default number of threads. 1830 */ 1831 retval = taskqueue_start_threads(&be_lun->io_taskqueue, 1832 /*num threads*/num_threads, 1833 /*priority*/PWAIT, 1834 /*thread name*/ 1835 "%s taskq", be_lun->lunname); 1836 1837 if (retval != 0) 1838 goto bailout_error; 1839 1840 be_lun->num_threads = num_threads; 1841 1842 mtx_lock(&softc->lock); 1843 softc->num_luns++; 1844 STAILQ_INSERT_TAIL(&softc->lun_list, be_lun, links); 1845 1846 mtx_unlock(&softc->lock); 1847 1848 retval = ctl_add_lun(&be_lun->ctl_be_lun); 1849 if (retval != 0) { 1850 mtx_lock(&softc->lock); 1851 STAILQ_REMOVE(&softc->lun_list, be_lun, ctl_be_block_lun, 1852 links); 1853 softc->num_luns--; 1854 mtx_unlock(&softc->lock); 1855 snprintf(req->error_str, sizeof(req->error_str), 1856 "%s: ctl_add_lun() returned error %d, see dmesg for " 1857 "details", __func__, retval); 1858 retval = 0; 1859 goto bailout_error; 1860 } 1861 1862 mtx_lock(&softc->lock); 1863 1864 /* 1865 * Tell the config_status routine that we're waiting so it won't 1866 * clean up the LUN in the event of an error. 1867 */ 1868 be_lun->flags |= CTL_BE_BLOCK_LUN_WAITING; 1869 1870 while (be_lun->flags & CTL_BE_BLOCK_LUN_UNCONFIGURED) { 1871 retval = msleep(be_lun, &softc->lock, PCATCH, "ctlblk", 0); 1872 if (retval == EINTR) 1873 break; 1874 } 1875 be_lun->flags &= ~CTL_BE_BLOCK_LUN_WAITING; 1876 1877 if (be_lun->flags & CTL_BE_BLOCK_LUN_CONFIG_ERR) { 1878 snprintf(req->error_str, sizeof(req->error_str), 1879 "%s: LUN configuration error, see dmesg for details", 1880 __func__); 1881 STAILQ_REMOVE(&softc->lun_list, be_lun, ctl_be_block_lun, 1882 links); 1883 softc->num_luns--; 1884 mtx_unlock(&softc->lock); 1885 goto bailout_error; 1886 } else { 1887 params->req_lun_id = be_lun->ctl_be_lun.lun_id; 1888 } 1889 1890 mtx_unlock(&softc->lock); 1891 1892 be_lun->disk_stats = devstat_new_entry("cbb", params->req_lun_id, 1893 be_lun->blocksize, 1894 DEVSTAT_ALL_SUPPORTED, 1895 be_lun->ctl_be_lun.lun_type 1896 | DEVSTAT_TYPE_IF_OTHER, 1897 DEVSTAT_PRIORITY_OTHER); 1898 1899 1900 req->status = CTL_LUN_OK; 1901 1902 return (retval); 1903 1904 bailout_error: 1905 req->status = CTL_LUN_ERROR; 1906 1907 ctl_be_block_close(be_lun); 1908 1909 free(be_lun->dev_path, M_CTLBLK); 1910 free(be_lun, M_CTLBLK); 1911 1912 return (retval); 1913 } 1914 1915 static int 1916 ctl_be_block_rm(struct ctl_be_block_softc *softc, struct ctl_lun_req *req) 1917 { 1918 struct ctl_lun_rm_params *params; 1919 struct ctl_be_block_lun *be_lun; 1920 int retval; 1921 1922 params = &req->reqdata.rm; 1923 1924 mtx_lock(&softc->lock); 1925 1926 be_lun = NULL; 1927 1928 STAILQ_FOREACH(be_lun, &softc->lun_list, links) { 1929 if (be_lun->ctl_be_lun.lun_id == params->lun_id) 1930 break; 1931 } 1932 mtx_unlock(&softc->lock); 1933 1934 if (be_lun == NULL) { 1935 snprintf(req->error_str, sizeof(req->error_str), 1936 "%s: LUN %u is not managed by the block backend", 1937 __func__, params->lun_id); 1938 goto bailout_error; 1939 } 1940 1941 retval = ctl_disable_lun(&be_lun->ctl_be_lun); 1942 1943 if (retval != 0) { 1944 snprintf(req->error_str, sizeof(req->error_str), 1945 "%s: error %d returned from ctl_disable_lun() for " 1946 "LUN %d", __func__, retval, params->lun_id); 1947 goto bailout_error; 1948 1949 } 1950 1951 retval = ctl_invalidate_lun(&be_lun->ctl_be_lun); 1952 if (retval != 0) { 1953 snprintf(req->error_str, sizeof(req->error_str), 1954 "%s: error %d returned from ctl_invalidate_lun() for " 1955 "LUN %d", __func__, retval, params->lun_id); 1956 goto bailout_error; 1957 } 1958 1959 mtx_lock(&softc->lock); 1960 1961 be_lun->flags |= CTL_BE_BLOCK_LUN_WAITING; 1962 1963 while ((be_lun->flags & CTL_BE_BLOCK_LUN_UNCONFIGURED) == 0) { 1964 retval = msleep(be_lun, &softc->lock, PCATCH, "ctlblk", 0); 1965 if (retval == EINTR) 1966 break; 1967 } 1968 1969 be_lun->flags &= ~CTL_BE_BLOCK_LUN_WAITING; 1970 1971 if ((be_lun->flags & CTL_BE_BLOCK_LUN_UNCONFIGURED) == 0) { 1972 snprintf(req->error_str, sizeof(req->error_str), 1973 "%s: interrupted waiting for LUN to be freed", 1974 __func__); 1975 mtx_unlock(&softc->lock); 1976 goto bailout_error; 1977 } 1978 1979 STAILQ_REMOVE(&softc->lun_list, be_lun, ctl_be_block_lun, links); 1980 1981 softc->num_luns--; 1982 mtx_unlock(&softc->lock); 1983 1984 taskqueue_drain(be_lun->io_taskqueue, &be_lun->io_task); 1985 1986 taskqueue_free(be_lun->io_taskqueue); 1987 1988 ctl_be_block_close(be_lun); 1989 1990 if (be_lun->disk_stats != NULL) 1991 devstat_remove_entry(be_lun->disk_stats); 1992 1993 uma_zdestroy(be_lun->lun_zone); 1994 1995 free(be_lun->dev_path, M_CTLBLK); 1996 1997 free(be_lun, M_CTLBLK); 1998 1999 req->status = CTL_LUN_OK; 2000 2001 return (0); 2002 2003 bailout_error: 2004 2005 req->status = CTL_LUN_ERROR; 2006 2007 return (0); 2008 } 2009 2010 static void 2011 ctl_be_block_lun_shutdown(void *be_lun) 2012 { 2013 struct ctl_be_block_lun *lun; 2014 struct ctl_be_block_softc *softc; 2015 2016 lun = (struct ctl_be_block_lun *)be_lun; 2017 2018 softc = lun->softc; 2019 2020 mtx_lock(&softc->lock); 2021 lun->flags |= CTL_BE_BLOCK_LUN_UNCONFIGURED; 2022 if (lun->flags & CTL_BE_BLOCK_LUN_WAITING) 2023 wakeup(lun); 2024 mtx_unlock(&softc->lock); 2025 2026 } 2027 2028 static void 2029 ctl_be_block_lun_config_status(void *be_lun, ctl_lun_config_status status) 2030 { 2031 struct ctl_be_block_lun *lun; 2032 struct ctl_be_block_softc *softc; 2033 2034 lun = (struct ctl_be_block_lun *)be_lun; 2035 softc = lun->softc; 2036 2037 if (status == CTL_LUN_CONFIG_OK) { 2038 mtx_lock(&softc->lock); 2039 lun->flags &= ~CTL_BE_BLOCK_LUN_UNCONFIGURED; 2040 if (lun->flags & CTL_BE_BLOCK_LUN_WAITING) 2041 wakeup(lun); 2042 mtx_unlock(&softc->lock); 2043 2044 /* 2045 * We successfully added the LUN, attempt to enable it. 2046 */ 2047 if (ctl_enable_lun(&lun->ctl_be_lun) != 0) { 2048 printf("%s: ctl_enable_lun() failed!\n", __func__); 2049 if (ctl_invalidate_lun(&lun->ctl_be_lun) != 0) { 2050 printf("%s: ctl_invalidate_lun() failed!\n", 2051 __func__); 2052 } 2053 } 2054 2055 return; 2056 } 2057 2058 2059 mtx_lock(&softc->lock); 2060 lun->flags &= ~CTL_BE_BLOCK_LUN_UNCONFIGURED; 2061 lun->flags |= CTL_BE_BLOCK_LUN_CONFIG_ERR; 2062 wakeup(lun); 2063 mtx_unlock(&softc->lock); 2064 } 2065 2066 2067 static int 2068 ctl_be_block_config_write(union ctl_io *io) 2069 { 2070 struct ctl_be_block_lun *be_lun; 2071 struct ctl_be_lun *ctl_be_lun; 2072 int retval; 2073 2074 retval = 0; 2075 2076 DPRINTF("entered\n"); 2077 2078 ctl_be_lun = (struct ctl_be_lun *)io->io_hdr.ctl_private[ 2079 CTL_PRIV_BACKEND_LUN].ptr; 2080 be_lun = (struct ctl_be_block_lun *)ctl_be_lun->be_lun; 2081 2082 switch (io->scsiio.cdb[0]) { 2083 case SYNCHRONIZE_CACHE: 2084 case SYNCHRONIZE_CACHE_16: 2085 /* 2086 * The upper level CTL code will filter out any CDBs with 2087 * the immediate bit set and return the proper error. 2088 * 2089 * We don't really need to worry about what LBA range the 2090 * user asked to be synced out. When they issue a sync 2091 * cache command, we'll sync out the whole thing. 2092 */ 2093 mtx_lock(&be_lun->lock); 2094 STAILQ_INSERT_TAIL(&be_lun->config_write_queue, &io->io_hdr, 2095 links); 2096 mtx_unlock(&be_lun->lock); 2097 taskqueue_enqueue(be_lun->io_taskqueue, &be_lun->io_task); 2098 break; 2099 case START_STOP_UNIT: { 2100 struct scsi_start_stop_unit *cdb; 2101 2102 cdb = (struct scsi_start_stop_unit *)io->scsiio.cdb; 2103 2104 if (cdb->how & SSS_START) 2105 retval = ctl_start_lun(ctl_be_lun); 2106 else { 2107 retval = ctl_stop_lun(ctl_be_lun); 2108 /* 2109 * XXX KDM Copan-specific offline behavior. 2110 * Figure out a reasonable way to port this? 2111 */ 2112 #ifdef NEEDTOPORT 2113 if ((retval == 0) 2114 && (cdb->byte2 & SSS_ONOFFLINE)) 2115 retval = ctl_lun_offline(ctl_be_lun); 2116 #endif 2117 } 2118 2119 /* 2120 * In general, the above routines should not fail. They 2121 * just set state for the LUN. So we've got something 2122 * pretty wrong here if we can't start or stop the LUN. 2123 */ 2124 if (retval != 0) { 2125 ctl_set_internal_failure(&io->scsiio, 2126 /*sks_valid*/ 1, 2127 /*retry_count*/ 0xf051); 2128 retval = CTL_RETVAL_COMPLETE; 2129 } else { 2130 ctl_set_success(&io->scsiio); 2131 } 2132 ctl_config_write_done(io); 2133 break; 2134 } 2135 default: 2136 ctl_set_invalid_opcode(&io->scsiio); 2137 ctl_config_write_done(io); 2138 retval = CTL_RETVAL_COMPLETE; 2139 break; 2140 } 2141 2142 return (retval); 2143 2144 } 2145 2146 static int 2147 ctl_be_block_config_read(union ctl_io *io) 2148 { 2149 return (0); 2150 } 2151 2152 static int 2153 ctl_be_block_lun_info(void *be_lun, struct sbuf *sb) 2154 { 2155 struct ctl_be_block_lun *lun; 2156 int retval; 2157 2158 lun = (struct ctl_be_block_lun *)be_lun; 2159 retval = 0; 2160 2161 retval = sbuf_printf(sb, "<num_threads>"); 2162 2163 if (retval != 0) 2164 goto bailout; 2165 2166 retval = sbuf_printf(sb, "%d", lun->num_threads); 2167 2168 if (retval != 0) 2169 goto bailout; 2170 2171 retval = sbuf_printf(sb, "</num_threads>"); 2172 2173 /* 2174 * For processor devices, we don't have a path variable. 2175 */ 2176 if ((retval != 0) 2177 || (lun->dev_path == NULL)) 2178 goto bailout; 2179 2180 retval = sbuf_printf(sb, "<file>"); 2181 2182 if (retval != 0) 2183 goto bailout; 2184 2185 retval = ctl_sbuf_printf_esc(sb, lun->dev_path); 2186 2187 if (retval != 0) 2188 goto bailout; 2189 2190 retval = sbuf_printf(sb, "</file>\n"); 2191 2192 bailout: 2193 2194 return (retval); 2195 } 2196 2197 int 2198 ctl_be_block_init(void) 2199 { 2200 struct ctl_be_block_softc *softc; 2201 int retval; 2202 2203 softc = &backend_block_softc; 2204 retval = 0; 2205 2206 mtx_init(&softc->lock, "ctlblk", NULL, MTX_DEF); 2207 STAILQ_INIT(&softc->beio_free_queue); 2208 STAILQ_INIT(&softc->disk_list); 2209 STAILQ_INIT(&softc->lun_list); 2210 ctl_grow_beio(softc, 200); 2211 2212 return (retval); 2213 } 2214