1 /* 2 * Copyright (c) 2003,2004,2009 The DragonFly Project. All rights reserved. 3 * 4 * This code is derived from software contributed to The DragonFly Project 5 * by Matthew Dillon <dillon@backplane.com> 6 * and Alex Hornung <ahornung@gmail.com> 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 12 * 1. Redistributions of source code must retain the above copyright 13 * notice, this list of conditions and the following disclaimer. 14 * 2. Redistributions in binary form must reproduce the above copyright 15 * notice, this list of conditions and the following disclaimer in 16 * the documentation and/or other materials provided with the 17 * distribution. 18 * 3. Neither the name of The DragonFly Project nor the names of its 19 * contributors may be used to endorse or promote products derived 20 * from this software without specific, prior written permission. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 23 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 24 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS 25 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE 26 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, 27 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING, 28 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; 29 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED 30 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 31 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT 32 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 33 * SUCH DAMAGE. 34 * 35 * ---------------------------------------------------------------------------- 36 * "THE BEER-WARE LICENSE" (Revision 42): 37 * <phk@FreeBSD.ORG> wrote this file. As long as you retain this notice you 38 * can do whatever you want with this stuff. If we meet some day, and you think 39 * this stuff is worth it, you can buy me a beer in return. Poul-Henning Kamp 40 * ---------------------------------------------------------------------------- 41 * 42 * Copyright (c) 1982, 1986, 1988, 1993 43 * The Regents of the University of California. All rights reserved. 44 * (c) UNIX System Laboratories, Inc. 45 * All or some portions of this file are derived from material licensed 46 * to the University of California by American Telephone and Telegraph 47 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 48 * the permission of UNIX System Laboratories, Inc. 49 * 50 * Redistribution and use in source and binary forms, with or without 51 * modification, are permitted provided that the following conditions 52 * are met: 53 * 1. Redistributions of source code must retain the above copyright 54 * notice, this list of conditions and the following disclaimer. 55 * 2. Redistributions in binary form must reproduce the above copyright 56 * notice, this list of conditions and the following disclaimer in the 57 * documentation and/or other materials provided with the distribution. 58 * 3. All advertising materials mentioning features or use of this software 59 * must display the following acknowledgement: 60 * This product includes software developed by the University of 61 * California, Berkeley and its contributors. 62 * 4. Neither the name of the University nor the names of its contributors 63 * may be used to endorse or promote products derived from this software 64 * without specific prior written permission. 65 * 66 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 67 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 68 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 69 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 70 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 71 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 72 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 73 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 74 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 75 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 76 * SUCH DAMAGE. 77 * 78 * @(#)ufs_disksubr.c 8.5 (Berkeley) 1/21/94 79 * $FreeBSD: src/sys/kern/subr_disk.c,v 1.20.2.6 2001/10/05 07:14:57 peter Exp $ 80 * $FreeBSD: src/sys/ufs/ufs/ufs_disksubr.c,v 1.44.2.3 2001/03/05 05:42:19 obrien Exp $ 81 */ 82 83 #include <sys/param.h> 84 #include <sys/systm.h> 85 #include <sys/kernel.h> 86 #include <sys/proc.h> 87 #include <sys/sysctl.h> 88 #include <sys/buf.h> 89 #include <sys/conf.h> 90 #include <sys/disklabel.h> 91 #include <sys/disklabel32.h> 92 #include <sys/disklabel64.h> 93 #include <sys/diskslice.h> 94 #include <sys/diskmbr.h> 95 #include <sys/disk.h> 96 #include <sys/kerneldump.h> 97 #include <sys/malloc.h> 98 #include <machine/md_var.h> 99 #include <sys/ctype.h> 100 #include <sys/syslog.h> 101 #include <sys/device.h> 102 #include <sys/msgport.h> 103 #include <sys/devfs.h> 104 #include <sys/thread.h> 105 #include <sys/dsched.h> 106 #include <sys/queue.h> 107 #include <sys/lock.h> 108 #include <sys/udev.h> 109 #include <sys/uuid.h> 110 111 #include <sys/buf2.h> 112 #include <sys/mplock2.h> 113 #include <sys/msgport2.h> 114 #include <sys/thread2.h> 115 116 static MALLOC_DEFINE(M_DISK, "disk", "disk data"); 117 static int disk_debug_enable = 0; 118 119 static void disk_msg_autofree_reply(lwkt_port_t, lwkt_msg_t); 120 static void disk_msg_core(void *); 121 static int disk_probe_slice(struct disk *dp, cdev_t dev, int slice, int reprobe); 122 static void disk_probe(struct disk *dp, int reprobe); 123 static void _setdiskinfo(struct disk *disk, struct disk_info *info); 124 static void bioqwritereorder(struct bio_queue_head *bioq); 125 static void disk_cleanserial(char *serno); 126 static int disk_debug(int, char *, ...) __printflike(2, 3); 127 static cdev_t _disk_create_named(const char *name, int unit, struct disk *dp, 128 struct dev_ops *raw_ops, int clone); 129 130 static d_open_t diskopen; 131 static d_close_t diskclose; 132 static d_ioctl_t diskioctl; 133 static d_strategy_t diskstrategy; 134 static d_psize_t diskpsize; 135 static d_dump_t diskdump; 136 137 static LIST_HEAD(, disk) disklist = LIST_HEAD_INITIALIZER(&disklist); 138 static struct lwkt_token disklist_token; 139 140 static struct dev_ops disk_ops = { 141 { "disk", 0, D_DISK | D_MPSAFE | D_TRACKCLOSE }, 142 .d_open = diskopen, 143 .d_close = diskclose, 144 .d_read = physread, 145 .d_write = physwrite, 146 .d_ioctl = diskioctl, 147 .d_strategy = diskstrategy, 148 .d_dump = diskdump, 149 .d_psize = diskpsize, 150 }; 151 152 static struct objcache *disk_msg_cache; 153 154 struct objcache_malloc_args disk_msg_malloc_args = { 155 sizeof(struct disk_msg), M_DISK }; 156 157 static struct lwkt_port disk_dispose_port; 158 static struct lwkt_port disk_msg_port; 159 160 static int 161 disk_debug(int level, char *fmt, ...) 162 { 163 __va_list ap; 164 165 __va_start(ap, fmt); 166 if (level <= disk_debug_enable) 167 kvprintf(fmt, ap); 168 __va_end(ap); 169 170 return 0; 171 } 172 173 static int 174 disk_probe_slice(struct disk *dp, cdev_t dev, int slice, int reprobe) 175 { 176 struct disk_info *info = &dp->d_info; 177 struct diskslice *sp = &dp->d_slice->dss_slices[slice]; 178 disklabel_ops_t ops; 179 struct partinfo part; 180 const char *msg; 181 char uuid_buf[128]; 182 cdev_t ndev; 183 int sno; 184 u_int i; 185 186 disk_debug(2, "disk_probe_slice (begin): %s (%s)\n", 187 dev->si_name, dp->d_cdev->si_name); 188 189 sno = slice ? slice - 1 : 0; 190 191 ops = &disklabel32_ops; 192 msg = ops->op_readdisklabel(dev, sp, &sp->ds_label, info); 193 if (msg && !strcmp(msg, "no disk label")) { 194 ops = &disklabel64_ops; 195 msg = ops->op_readdisklabel(dev, sp, &sp->ds_label, info); 196 } 197 198 if (msg == NULL) { 199 if (slice != WHOLE_DISK_SLICE) 200 ops->op_adjust_label_reserved(dp->d_slice, slice, sp); 201 else 202 sp->ds_reserved = 0; 203 204 sp->ds_ops = ops; 205 for (i = 0; i < ops->op_getnumparts(sp->ds_label); i++) { 206 ops->op_loadpartinfo(sp->ds_label, i, &part); 207 if (part.fstype) { 208 if (reprobe && 209 (ndev = devfs_find_device_by_name("%s%c", 210 dev->si_name, 'a' + i)) 211 ) { 212 /* 213 * Device already exists and 214 * is still valid. 215 */ 216 ndev->si_flags |= SI_REPROBE_TEST; 217 218 /* 219 * Destroy old UUID alias 220 */ 221 destroy_dev_alias(ndev, "part-by-uuid/*"); 222 223 /* Create UUID alias */ 224 if (!kuuid_is_nil(&part.storage_uuid)) { 225 snprintf_uuid(uuid_buf, 226 sizeof(uuid_buf), 227 &part.storage_uuid); 228 make_dev_alias(ndev, 229 "part-by-uuid/%s", 230 uuid_buf); 231 udev_dict_set_cstr(ndev, "uuid", uuid_buf); 232 } 233 } else { 234 ndev = make_dev_covering(&disk_ops, dp->d_rawdev->si_ops, 235 dkmakeminor(dkunit(dp->d_cdev), 236 slice, i), 237 UID_ROOT, GID_OPERATOR, 0640, 238 "%s%c", dev->si_name, 'a'+ i); 239 ndev->si_parent = dev; 240 ndev->si_iosize_max = dev->si_iosize_max; 241 ndev->si_disk = dp; 242 udev_dict_set_cstr(ndev, "subsystem", "disk"); 243 /* Inherit parent's disk type */ 244 if (dp->d_disktype) { 245 udev_dict_set_cstr(ndev, "disk-type", 246 __DECONST(char *, dp->d_disktype)); 247 } 248 249 /* Create serno alias */ 250 if (dp->d_info.d_serialno) { 251 make_dev_alias(ndev, 252 "serno/%s.s%d%c", 253 dp->d_info.d_serialno, 254 sno, 'a' + i); 255 } 256 257 /* Create UUID alias */ 258 if (!kuuid_is_nil(&part.storage_uuid)) { 259 snprintf_uuid(uuid_buf, 260 sizeof(uuid_buf), 261 &part.storage_uuid); 262 make_dev_alias(ndev, 263 "part-by-uuid/%s", 264 uuid_buf); 265 udev_dict_set_cstr(ndev, "uuid", uuid_buf); 266 } 267 ndev->si_flags |= SI_REPROBE_TEST; 268 } 269 } 270 } 271 } else if (info->d_dsflags & DSO_COMPATLABEL) { 272 msg = NULL; 273 if (sp->ds_size >= 0x100000000ULL) 274 ops = &disklabel64_ops; 275 else 276 ops = &disklabel32_ops; 277 sp->ds_label = ops->op_clone_label(info, sp); 278 } else { 279 if (sp->ds_type == DOSPTYP_386BSD || /* XXX */ 280 sp->ds_type == DOSPTYP_NETBSD || 281 sp->ds_type == DOSPTYP_OPENBSD) { 282 log(LOG_WARNING, "%s: cannot find label (%s)\n", 283 dev->si_name, msg); 284 } 285 286 if (sp->ds_label.opaque != NULL && sp->ds_ops != NULL) { 287 /* Clear out old label - it's not around anymore */ 288 disk_debug(2, 289 "disk_probe_slice: clear out old diskabel on %s\n", 290 dev->si_name); 291 292 sp->ds_ops->op_freedisklabel(&sp->ds_label); 293 sp->ds_ops = NULL; 294 } 295 } 296 297 if (msg == NULL) { 298 sp->ds_wlabel = FALSE; 299 } 300 301 return (msg ? EINVAL : 0); 302 } 303 304 /* 305 * This routine is only called for newly minted drives or to reprobe 306 * a drive with no open slices. disk_probe_slice() is called directly 307 * when reprobing partition changes within slices. 308 */ 309 static void 310 disk_probe(struct disk *dp, int reprobe) 311 { 312 struct disk_info *info = &dp->d_info; 313 cdev_t dev = dp->d_cdev; 314 cdev_t ndev; 315 int error, i, sno; 316 struct diskslices *osp; 317 struct diskslice *sp; 318 char uuid_buf[128]; 319 320 KKASSERT (info->d_media_blksize != 0); 321 322 osp = dp->d_slice; 323 dp->d_slice = dsmakeslicestruct(BASE_SLICE, info); 324 disk_debug(1, "disk_probe (begin): %s\n", dp->d_cdev->si_name); 325 326 error = mbrinit(dev, info, &(dp->d_slice)); 327 if (error) { 328 dsgone(&osp); 329 return; 330 } 331 332 for (i = 0; i < dp->d_slice->dss_nslices; i++) { 333 /* 334 * Ignore the whole-disk slice, it has already been created. 335 */ 336 if (i == WHOLE_DISK_SLICE) 337 continue; 338 339 #if 1 340 /* 341 * Ignore the compatibility slice s0 if it's a device mapper 342 * volume. 343 */ 344 if ((i == COMPATIBILITY_SLICE) && 345 (info->d_dsflags & DSO_DEVICEMAPPER)) 346 continue; 347 #endif 348 349 sp = &dp->d_slice->dss_slices[i]; 350 351 /* 352 * Handle s0. s0 is a compatibility slice if there are no 353 * other slices and it has not otherwise been set up, else 354 * we ignore it. 355 */ 356 if (i == COMPATIBILITY_SLICE) { 357 sno = 0; 358 if (sp->ds_type == 0 && 359 dp->d_slice->dss_nslices == BASE_SLICE) { 360 sp->ds_size = info->d_media_blocks; 361 sp->ds_reserved = 0; 362 } 363 } else { 364 sno = i - 1; 365 sp->ds_reserved = 0; 366 } 367 368 /* 369 * Ignore 0-length slices 370 */ 371 if (sp->ds_size == 0) 372 continue; 373 374 if (reprobe && 375 (ndev = devfs_find_device_by_name("%ss%d", 376 dev->si_name, sno))) { 377 /* 378 * Device already exists and is still valid 379 */ 380 ndev->si_flags |= SI_REPROBE_TEST; 381 382 /* 383 * Destroy old UUID alias 384 */ 385 destroy_dev_alias(ndev, "slice-by-uuid/*"); 386 387 /* Create UUID alias */ 388 if (!kuuid_is_nil(&sp->ds_stor_uuid)) { 389 snprintf_uuid(uuid_buf, sizeof(uuid_buf), 390 &sp->ds_stor_uuid); 391 make_dev_alias(ndev, "slice-by-uuid/%s", 392 uuid_buf); 393 } 394 } else { 395 /* 396 * Else create new device 397 */ 398 ndev = make_dev_covering(&disk_ops, dp->d_rawdev->si_ops, 399 dkmakewholeslice(dkunit(dev), i), 400 UID_ROOT, GID_OPERATOR, 0640, 401 (info->d_dsflags & DSO_DEVICEMAPPER)? 402 "%s.s%d" : "%ss%d", dev->si_name, sno); 403 ndev->si_parent = dev; 404 ndev->si_iosize_max = dev->si_iosize_max; 405 udev_dict_set_cstr(ndev, "subsystem", "disk"); 406 /* Inherit parent's disk type */ 407 if (dp->d_disktype) { 408 udev_dict_set_cstr(ndev, "disk-type", 409 __DECONST(char *, dp->d_disktype)); 410 } 411 412 /* Create serno alias */ 413 if (dp->d_info.d_serialno) { 414 make_dev_alias(ndev, "serno/%s.s%d", 415 dp->d_info.d_serialno, sno); 416 } 417 418 /* Create UUID alias */ 419 if (!kuuid_is_nil(&sp->ds_stor_uuid)) { 420 snprintf_uuid(uuid_buf, sizeof(uuid_buf), 421 &sp->ds_stor_uuid); 422 make_dev_alias(ndev, "slice-by-uuid/%s", 423 uuid_buf); 424 } 425 426 ndev->si_disk = dp; 427 ndev->si_flags |= SI_REPROBE_TEST; 428 } 429 sp->ds_dev = ndev; 430 431 /* 432 * Probe appropriate slices for a disklabel 433 * 434 * XXX slice type 1 used by our gpt probe code. 435 * XXX slice type 0 used by mbr compat slice. 436 */ 437 if (sp->ds_type == DOSPTYP_386BSD || 438 sp->ds_type == DOSPTYP_NETBSD || 439 sp->ds_type == DOSPTYP_OPENBSD || 440 sp->ds_type == 0 || 441 sp->ds_type == 1) { 442 if (dp->d_slice->dss_first_bsd_slice == 0) 443 dp->d_slice->dss_first_bsd_slice = i; 444 disk_probe_slice(dp, ndev, i, reprobe); 445 } 446 } 447 dsgone(&osp); 448 disk_debug(1, "disk_probe (end): %s\n", dp->d_cdev->si_name); 449 } 450 451 452 static void 453 disk_msg_core(void *arg) 454 { 455 struct disk *dp; 456 struct diskslice *sp; 457 disk_msg_t msg; 458 int run; 459 460 lwkt_gettoken(&disklist_token); 461 lwkt_initport_thread(&disk_msg_port, curthread); 462 wakeup(curthread); /* synchronous startup */ 463 lwkt_reltoken(&disklist_token); 464 465 get_mplock(); /* not mpsafe yet? */ 466 run = 1; 467 468 while (run) { 469 msg = (disk_msg_t)lwkt_waitport(&disk_msg_port, 0); 470 471 switch (msg->hdr.u.ms_result) { 472 case DISK_DISK_PROBE: 473 dp = (struct disk *)msg->load; 474 disk_debug(1, 475 "DISK_DISK_PROBE: %s\n", 476 dp->d_cdev->si_name); 477 disk_iocom_update(dp); 478 disk_probe(dp, 0); 479 break; 480 case DISK_DISK_DESTROY: 481 dp = (struct disk *)msg->load; 482 disk_debug(1, 483 "DISK_DISK_DESTROY: %s\n", 484 dp->d_cdev->si_name); 485 disk_iocom_uninit(dp); 486 devfs_destroy_related(dp->d_cdev); 487 destroy_dev(dp->d_cdev); 488 destroy_only_dev(dp->d_rawdev); 489 lwkt_gettoken(&disklist_token); 490 LIST_REMOVE(dp, d_list); 491 lwkt_reltoken(&disklist_token); 492 if (dp->d_info.d_serialno) { 493 kfree(dp->d_info.d_serialno, M_TEMP); 494 dp->d_info.d_serialno = NULL; 495 } 496 break; 497 case DISK_UNPROBE: 498 dp = (struct disk *)msg->load; 499 disk_debug(1, 500 "DISK_DISK_UNPROBE: %s\n", 501 dp->d_cdev->si_name); 502 devfs_destroy_related(dp->d_cdev); 503 break; 504 case DISK_SLICE_REPROBE: 505 dp = (struct disk *)msg->load; 506 sp = (struct diskslice *)msg->load2; 507 devfs_clr_related_flag(sp->ds_dev, 508 SI_REPROBE_TEST); 509 disk_debug(1, 510 "DISK_SLICE_REPROBE: %s\n", 511 sp->ds_dev->si_name); 512 disk_probe_slice(dp, sp->ds_dev, 513 dkslice(sp->ds_dev), 1); 514 devfs_destroy_related_without_flag( 515 sp->ds_dev, SI_REPROBE_TEST); 516 break; 517 case DISK_DISK_REPROBE: 518 dp = (struct disk *)msg->load; 519 devfs_clr_related_flag(dp->d_cdev, SI_REPROBE_TEST); 520 disk_debug(1, 521 "DISK_DISK_REPROBE: %s\n", 522 dp->d_cdev->si_name); 523 disk_probe(dp, 1); 524 devfs_destroy_related_without_flag( 525 dp->d_cdev, SI_REPROBE_TEST); 526 break; 527 case DISK_SYNC: 528 disk_debug(1, "DISK_SYNC\n"); 529 break; 530 default: 531 devfs_debug(DEVFS_DEBUG_WARNING, 532 "disk_msg_core: unknown message " 533 "received at core\n"); 534 break; 535 } 536 lwkt_replymsg(&msg->hdr, 0); 537 } 538 lwkt_exit(); 539 } 540 541 542 /* 543 * Acts as a message drain. Any message that is replied to here gets 544 * destroyed and the memory freed. 545 */ 546 static void 547 disk_msg_autofree_reply(lwkt_port_t port, lwkt_msg_t msg) 548 { 549 objcache_put(disk_msg_cache, msg); 550 } 551 552 553 void 554 disk_msg_send(uint32_t cmd, void *load, void *load2) 555 { 556 disk_msg_t disk_msg; 557 lwkt_port_t port = &disk_msg_port; 558 559 disk_msg = objcache_get(disk_msg_cache, M_WAITOK); 560 561 lwkt_initmsg(&disk_msg->hdr, &disk_dispose_port, 0); 562 563 disk_msg->hdr.u.ms_result = cmd; 564 disk_msg->load = load; 565 disk_msg->load2 = load2; 566 KKASSERT(port); 567 lwkt_sendmsg(port, &disk_msg->hdr); 568 } 569 570 void 571 disk_msg_send_sync(uint32_t cmd, void *load, void *load2) 572 { 573 struct lwkt_port rep_port; 574 disk_msg_t disk_msg; 575 lwkt_port_t port; 576 577 disk_msg = objcache_get(disk_msg_cache, M_WAITOK); 578 port = &disk_msg_port; 579 580 /* XXX could probably use curthread's built-in msgport */ 581 lwkt_initport_thread(&rep_port, curthread); 582 lwkt_initmsg(&disk_msg->hdr, &rep_port, 0); 583 584 disk_msg->hdr.u.ms_result = cmd; 585 disk_msg->load = load; 586 disk_msg->load2 = load2; 587 588 lwkt_sendmsg(port, &disk_msg->hdr); 589 lwkt_waitmsg(&disk_msg->hdr, 0); 590 objcache_put(disk_msg_cache, disk_msg); 591 } 592 593 /* 594 * Create a raw device for the dev_ops template (which is returned). Also 595 * create a slice and unit managed disk and overload the user visible 596 * device space with it. 597 * 598 * NOTE: The returned raw device is NOT a slice and unit managed device. 599 * It is an actual raw device representing the raw disk as specified by 600 * the passed dev_ops. The disk layer not only returns such a raw device, 601 * it also uses it internally when passing (modified) commands through. 602 */ 603 cdev_t 604 disk_create(int unit, struct disk *dp, struct dev_ops *raw_ops) 605 { 606 return _disk_create_named(NULL, unit, dp, raw_ops, 0); 607 } 608 609 cdev_t 610 disk_create_clone(int unit, struct disk *dp, 611 struct dev_ops *raw_ops) 612 { 613 return _disk_create_named(NULL, unit, dp, raw_ops, 1); 614 } 615 616 cdev_t 617 disk_create_named(const char *name, int unit, struct disk *dp, 618 struct dev_ops *raw_ops) 619 { 620 return _disk_create_named(name, unit, dp, raw_ops, 0); 621 } 622 623 cdev_t 624 disk_create_named_clone(const char *name, int unit, struct disk *dp, 625 struct dev_ops *raw_ops) 626 { 627 return _disk_create_named(name, unit, dp, raw_ops, 1); 628 } 629 630 static cdev_t 631 _disk_create_named(const char *name, int unit, struct disk *dp, 632 struct dev_ops *raw_ops, int clone) 633 { 634 cdev_t rawdev; 635 636 disk_debug(1, "disk_create (begin): %s%d\n", name, unit); 637 638 if (name) { 639 rawdev = make_only_dev(raw_ops, dkmakewholedisk(unit), 640 UID_ROOT, GID_OPERATOR, 0640, "%s", name); 641 } else { 642 rawdev = make_only_dev(raw_ops, dkmakewholedisk(unit), 643 UID_ROOT, GID_OPERATOR, 0640, 644 "%s%d", raw_ops->head.name, unit); 645 } 646 647 bzero(dp, sizeof(*dp)); 648 649 dp->d_rawdev = rawdev; 650 dp->d_raw_ops = raw_ops; 651 dp->d_dev_ops = &disk_ops; 652 653 if (name) { 654 if (clone) { 655 dp->d_cdev = make_only_dev_covering( 656 &disk_ops, dp->d_rawdev->si_ops, 657 dkmakewholedisk(unit), 658 UID_ROOT, GID_OPERATOR, 0640, 659 "%s", name); 660 } else { 661 dp->d_cdev = make_dev_covering( 662 &disk_ops, dp->d_rawdev->si_ops, 663 dkmakewholedisk(unit), 664 UID_ROOT, GID_OPERATOR, 0640, 665 "%s", name); 666 } 667 } else { 668 if (clone) { 669 dp->d_cdev = make_only_dev_covering( 670 &disk_ops, dp->d_rawdev->si_ops, 671 dkmakewholedisk(unit), 672 UID_ROOT, GID_OPERATOR, 0640, 673 "%s%d", raw_ops->head.name, unit); 674 } else { 675 dp->d_cdev = make_dev_covering( 676 &disk_ops, dp->d_rawdev->si_ops, 677 dkmakewholedisk(unit), 678 UID_ROOT, GID_OPERATOR, 0640, 679 "%s%d", raw_ops->head.name, unit); 680 } 681 } 682 683 udev_dict_set_cstr(dp->d_cdev, "subsystem", "disk"); 684 dp->d_cdev->si_disk = dp; 685 686 if (name) 687 dsched_disk_create_callback(dp, name, unit); 688 else 689 dsched_disk_create_callback(dp, raw_ops->head.name, unit); 690 691 lwkt_gettoken(&disklist_token); 692 LIST_INSERT_HEAD(&disklist, dp, d_list); 693 lwkt_reltoken(&disklist_token); 694 695 disk_iocom_init(dp); 696 697 disk_debug(1, "disk_create (end): %s%d\n", 698 (name != NULL)?(name):(raw_ops->head.name), unit); 699 700 return (dp->d_rawdev); 701 } 702 703 int 704 disk_setdisktype(struct disk *disk, const char *type) 705 { 706 int error; 707 708 KKASSERT(disk != NULL); 709 710 disk->d_disktype = type; 711 error = udev_dict_set_cstr(disk->d_cdev, "disk-type", 712 __DECONST(char *, type)); 713 return error; 714 } 715 716 int 717 disk_getopencount(struct disk *disk) 718 { 719 return disk->d_opencount; 720 } 721 722 static void 723 _setdiskinfo(struct disk *disk, struct disk_info *info) 724 { 725 char *oldserialno; 726 727 oldserialno = disk->d_info.d_serialno; 728 bcopy(info, &disk->d_info, sizeof(disk->d_info)); 729 info = &disk->d_info; 730 731 disk_debug(1, "_setdiskinfo: %s\n", disk->d_cdev->si_name); 732 733 /* 734 * The serial number is duplicated so the caller can throw 735 * their copy away. 736 */ 737 if (info->d_serialno && info->d_serialno[0] && 738 (info->d_serialno[0] != ' ' || strlen(info->d_serialno) > 1)) { 739 info->d_serialno = kstrdup(info->d_serialno, M_TEMP); 740 disk_cleanserial(info->d_serialno); 741 if (disk->d_cdev) { 742 make_dev_alias(disk->d_cdev, "serno/%s", 743 info->d_serialno); 744 } 745 } else { 746 info->d_serialno = NULL; 747 } 748 if (oldserialno) 749 kfree(oldserialno, M_TEMP); 750 751 dsched_disk_update_callback(disk, info); 752 753 /* 754 * The caller may set d_media_size or d_media_blocks and we 755 * calculate the other. 756 */ 757 KKASSERT(info->d_media_size == 0 || info->d_media_blocks == 0); 758 if (info->d_media_size == 0 && info->d_media_blocks) { 759 info->d_media_size = (u_int64_t)info->d_media_blocks * 760 info->d_media_blksize; 761 } else if (info->d_media_size && info->d_media_blocks == 0 && 762 info->d_media_blksize) { 763 info->d_media_blocks = info->d_media_size / 764 info->d_media_blksize; 765 } 766 767 /* 768 * The si_* fields for rawdev are not set until after the 769 * disk_create() call, so someone using the cooked version 770 * of the raw device (i.e. da0s0) will not get the right 771 * si_iosize_max unless we fix it up here. 772 */ 773 if (disk->d_cdev && disk->d_rawdev && 774 disk->d_cdev->si_iosize_max == 0) { 775 disk->d_cdev->si_iosize_max = disk->d_rawdev->si_iosize_max; 776 disk->d_cdev->si_bsize_phys = disk->d_rawdev->si_bsize_phys; 777 disk->d_cdev->si_bsize_best = disk->d_rawdev->si_bsize_best; 778 } 779 780 /* Add the serial number to the udev_dictionary */ 781 if (info->d_serialno) 782 udev_dict_set_cstr(disk->d_cdev, "serno", info->d_serialno); 783 } 784 785 /* 786 * Disk drivers must call this routine when media parameters are available 787 * or have changed. 788 */ 789 void 790 disk_setdiskinfo(struct disk *disk, struct disk_info *info) 791 { 792 _setdiskinfo(disk, info); 793 disk_msg_send(DISK_DISK_PROBE, disk, NULL); 794 disk_debug(1, "disk_setdiskinfo: sent probe for %s\n", 795 disk->d_cdev->si_name); 796 } 797 798 void 799 disk_setdiskinfo_sync(struct disk *disk, struct disk_info *info) 800 { 801 _setdiskinfo(disk, info); 802 disk_msg_send_sync(DISK_DISK_PROBE, disk, NULL); 803 disk_debug(1, "disk_setdiskinfo_sync: sent probe for %s\n", 804 disk->d_cdev->si_name); 805 } 806 807 /* 808 * This routine is called when an adapter detaches. The higher level 809 * managed disk device is destroyed while the lower level raw device is 810 * released. 811 */ 812 void 813 disk_destroy(struct disk *disk) 814 { 815 dsched_disk_destroy_callback(disk); 816 disk_msg_send_sync(DISK_DISK_DESTROY, disk, NULL); 817 return; 818 } 819 820 int 821 disk_dumpcheck(cdev_t dev, u_int64_t *size, 822 u_int64_t *blkno, u_int32_t *secsize) 823 { 824 struct partinfo pinfo; 825 int error; 826 827 bzero(&pinfo, sizeof(pinfo)); 828 error = dev_dioctl(dev, DIOCGPART, (void *)&pinfo, 0, 829 proc0.p_ucred, NULL); 830 if (error) 831 return (error); 832 833 if (pinfo.media_blksize == 0) 834 return (ENXIO); 835 836 if (blkno) /* XXX: make sure this reserved stuff is right */ 837 *blkno = pinfo.reserved_blocks + 838 pinfo.media_offset / pinfo.media_blksize; 839 if (secsize) 840 *secsize = pinfo.media_blksize; 841 if (size) 842 *size = (pinfo.media_blocks - pinfo.reserved_blocks); 843 844 return (0); 845 } 846 847 int 848 disk_dumpconf(cdev_t dev, u_int onoff) 849 { 850 struct dumperinfo di; 851 u_int64_t size, blkno; 852 u_int32_t secsize; 853 int error; 854 855 if (!onoff) 856 return set_dumper(NULL); 857 858 error = disk_dumpcheck(dev, &size, &blkno, &secsize); 859 860 if (error) 861 return ENXIO; 862 863 bzero(&di, sizeof(struct dumperinfo)); 864 di.dumper = diskdump; 865 di.priv = dev; 866 di.blocksize = secsize; 867 di.maxiosize = dev->si_iosize_max; 868 di.mediaoffset = blkno * DEV_BSIZE; 869 di.mediasize = size * DEV_BSIZE; 870 871 return set_dumper(&di); 872 } 873 874 void 875 disk_unprobe(struct disk *disk) 876 { 877 if (disk == NULL) 878 return; 879 880 disk_msg_send_sync(DISK_UNPROBE, disk, NULL); 881 } 882 883 void 884 disk_invalidate (struct disk *disk) 885 { 886 dsgone(&disk->d_slice); 887 } 888 889 struct disk * 890 disk_enumerate(struct disk *disk) 891 { 892 struct disk *dp; 893 894 lwkt_gettoken(&disklist_token); 895 if (!disk) 896 dp = (LIST_FIRST(&disklist)); 897 else 898 dp = (LIST_NEXT(disk, d_list)); 899 lwkt_reltoken(&disklist_token); 900 901 return dp; 902 } 903 904 static 905 int 906 sysctl_disks(SYSCTL_HANDLER_ARGS) 907 { 908 struct disk *disk; 909 int error, first; 910 911 disk = NULL; 912 first = 1; 913 914 while ((disk = disk_enumerate(disk))) { 915 if (!first) { 916 error = SYSCTL_OUT(req, " ", 1); 917 if (error) 918 return error; 919 } else { 920 first = 0; 921 } 922 error = SYSCTL_OUT(req, disk->d_rawdev->si_name, 923 strlen(disk->d_rawdev->si_name)); 924 if (error) 925 return error; 926 } 927 error = SYSCTL_OUT(req, "", 1); 928 return error; 929 } 930 931 SYSCTL_PROC(_kern, OID_AUTO, disks, CTLTYPE_STRING | CTLFLAG_RD, NULL, 0, 932 sysctl_disks, "A", "names of available disks"); 933 934 /* 935 * Open a disk device or partition. 936 */ 937 static 938 int 939 diskopen(struct dev_open_args *ap) 940 { 941 cdev_t dev = ap->a_head.a_dev; 942 struct disk *dp; 943 int error; 944 945 /* 946 * dp can't be NULL here XXX. 947 * 948 * d_slice will be NULL if setdiskinfo() has not been called yet. 949 * setdiskinfo() is typically called whether the disk is present 950 * or not (e.g. CD), but the base disk device is created first 951 * and there may be a race. 952 */ 953 dp = dev->si_disk; 954 if (dp == NULL || dp->d_slice == NULL) 955 return (ENXIO); 956 error = 0; 957 958 /* 959 * Deal with open races 960 */ 961 get_mplock(); 962 while (dp->d_flags & DISKFLAG_LOCK) { 963 dp->d_flags |= DISKFLAG_WANTED; 964 error = tsleep(dp, PCATCH, "diskopen", hz); 965 if (error) { 966 rel_mplock(); 967 return (error); 968 } 969 } 970 dp->d_flags |= DISKFLAG_LOCK; 971 972 /* 973 * Open the underlying raw device. 974 */ 975 if (!dsisopen(dp->d_slice)) { 976 #if 0 977 if (!pdev->si_iosize_max) 978 pdev->si_iosize_max = dev->si_iosize_max; 979 #endif 980 error = dev_dopen(dp->d_rawdev, ap->a_oflags, 981 ap->a_devtype, ap->a_cred); 982 } 983 984 if (error) 985 goto out; 986 error = dsopen(dev, ap->a_devtype, dp->d_info.d_dsflags, 987 &dp->d_slice, &dp->d_info); 988 if (!dsisopen(dp->d_slice)) { 989 dev_dclose(dp->d_rawdev, ap->a_oflags, ap->a_devtype); 990 } 991 out: 992 dp->d_flags &= ~DISKFLAG_LOCK; 993 if (dp->d_flags & DISKFLAG_WANTED) { 994 dp->d_flags &= ~DISKFLAG_WANTED; 995 wakeup(dp); 996 } 997 rel_mplock(); 998 999 KKASSERT(dp->d_opencount >= 0); 1000 /* If the open was successful, bump open count */ 1001 if (error == 0) 1002 atomic_add_int(&dp->d_opencount, 1); 1003 1004 return(error); 1005 } 1006 1007 /* 1008 * Close a disk device or partition 1009 */ 1010 static 1011 int 1012 diskclose(struct dev_close_args *ap) 1013 { 1014 cdev_t dev = ap->a_head.a_dev; 1015 struct disk *dp; 1016 int error; 1017 int lcount; 1018 1019 error = 0; 1020 dp = dev->si_disk; 1021 1022 /* 1023 * The cdev_t represents the disk/slice/part. The shared 1024 * dp structure governs all cdevs associated with the disk. 1025 * 1026 * As a safety only close the underlying raw device on the last 1027 * close the disk device if our tracking of the slices/partitions 1028 * also indicates nothing is open. 1029 */ 1030 KKASSERT(dp->d_opencount >= 1); 1031 lcount = atomic_fetchadd_int(&dp->d_opencount, -1); 1032 1033 get_mplock(); 1034 dsclose(dev, ap->a_devtype, dp->d_slice); 1035 if (lcount <= 1 && !dsisopen(dp->d_slice)) { 1036 error = dev_dclose(dp->d_rawdev, ap->a_fflag, ap->a_devtype); 1037 } 1038 rel_mplock(); 1039 return (error); 1040 } 1041 1042 /* 1043 * First execute the ioctl on the disk device, and if it isn't supported 1044 * try running it on the backing device. 1045 */ 1046 static 1047 int 1048 diskioctl(struct dev_ioctl_args *ap) 1049 { 1050 cdev_t dev = ap->a_head.a_dev; 1051 struct disk *dp; 1052 int error; 1053 u_int u; 1054 1055 dp = dev->si_disk; 1056 if (dp == NULL) 1057 return (ENXIO); 1058 1059 devfs_debug(DEVFS_DEBUG_DEBUG, 1060 "diskioctl: cmd is: %lx (name: %s)\n", 1061 ap->a_cmd, dev->si_name); 1062 devfs_debug(DEVFS_DEBUG_DEBUG, 1063 "diskioctl: &dp->d_slice is: %p, %p\n", 1064 &dp->d_slice, dp->d_slice); 1065 1066 if (ap->a_cmd == DIOCGKERNELDUMP) { 1067 u = *(u_int *)ap->a_data; 1068 return disk_dumpconf(dev, u); 1069 } 1070 1071 if (ap->a_cmd == DIOCRECLUSTER && dev == dp->d_cdev) { 1072 kprintf("RECLUSTER\n"); 1073 error = disk_iocom_ioctl(dp, ap->a_cmd, ap->a_data); 1074 return error; 1075 } 1076 1077 if (&dp->d_slice == NULL || dp->d_slice == NULL || 1078 ((dp->d_info.d_dsflags & DSO_DEVICEMAPPER) && 1079 dkslice(dev) == WHOLE_DISK_SLICE)) { 1080 error = ENOIOCTL; 1081 } else { 1082 get_mplock(); 1083 error = dsioctl(dev, ap->a_cmd, ap->a_data, ap->a_fflag, 1084 &dp->d_slice, &dp->d_info); 1085 rel_mplock(); 1086 } 1087 1088 if (error == ENOIOCTL) { 1089 error = dev_dioctl(dp->d_rawdev, ap->a_cmd, ap->a_data, 1090 ap->a_fflag, ap->a_cred, NULL); 1091 } 1092 return (error); 1093 } 1094 1095 /* 1096 * Execute strategy routine 1097 */ 1098 static 1099 int 1100 diskstrategy(struct dev_strategy_args *ap) 1101 { 1102 cdev_t dev = ap->a_head.a_dev; 1103 struct bio *bio = ap->a_bio; 1104 struct bio *nbio; 1105 struct disk *dp; 1106 1107 dp = dev->si_disk; 1108 1109 if (dp == NULL) { 1110 bio->bio_buf->b_error = ENXIO; 1111 bio->bio_buf->b_flags |= B_ERROR; 1112 biodone(bio); 1113 return(0); 1114 } 1115 KKASSERT(dev->si_disk == dp); 1116 1117 /* 1118 * The dscheck() function will also transform the slice relative 1119 * block number i.e. bio->bio_offset into a block number that can be 1120 * passed directly to the underlying raw device. If dscheck() 1121 * returns NULL it will have handled the bio for us (e.g. EOF 1122 * or error due to being beyond the device size). 1123 */ 1124 if ((nbio = dscheck(dev, bio, dp->d_slice)) != NULL) { 1125 dsched_queue(dp, nbio); 1126 } else { 1127 biodone(bio); 1128 } 1129 return(0); 1130 } 1131 1132 /* 1133 * Return the partition size in ?blocks? 1134 */ 1135 static 1136 int 1137 diskpsize(struct dev_psize_args *ap) 1138 { 1139 cdev_t dev = ap->a_head.a_dev; 1140 struct disk *dp; 1141 1142 dp = dev->si_disk; 1143 if (dp == NULL) 1144 return(ENODEV); 1145 1146 ap->a_result = dssize(dev, &dp->d_slice); 1147 1148 if ((ap->a_result == -1) && 1149 (dp->d_info.d_dsflags & DSO_RAWPSIZE)) { 1150 ap->a_head.a_dev = dp->d_rawdev; 1151 return dev_doperate(&ap->a_head); 1152 } 1153 return(0); 1154 } 1155 1156 int 1157 diskdump(struct dev_dump_args *ap) 1158 { 1159 cdev_t dev = ap->a_head.a_dev; 1160 struct disk *dp = dev->si_disk; 1161 u_int64_t size, offset; 1162 int error; 1163 1164 error = disk_dumpcheck(dev, &size, &ap->a_blkno, &ap->a_secsize); 1165 /* XXX: this should probably go in disk_dumpcheck somehow */ 1166 if (ap->a_length != 0) { 1167 size *= DEV_BSIZE; 1168 offset = ap->a_blkno * DEV_BSIZE; 1169 if ((ap->a_offset < offset) || 1170 (ap->a_offset + ap->a_length - offset > size)) { 1171 kprintf("Attempt to write outside dump " 1172 "device boundaries.\n"); 1173 error = ENOSPC; 1174 } 1175 } 1176 1177 if (error == 0) { 1178 ap->a_head.a_dev = dp->d_rawdev; 1179 error = dev_doperate(&ap->a_head); 1180 } 1181 1182 return(error); 1183 } 1184 1185 1186 SYSCTL_INT(_debug_sizeof, OID_AUTO, diskslices, CTLFLAG_RD, 1187 0, sizeof(struct diskslices), "sizeof(struct diskslices)"); 1188 1189 SYSCTL_INT(_debug_sizeof, OID_AUTO, disk, CTLFLAG_RD, 1190 0, sizeof(struct disk), "sizeof(struct disk)"); 1191 1192 /* 1193 * Reorder interval for burst write allowance and minor write 1194 * allowance. 1195 * 1196 * We always want to trickle some writes in to make use of the 1197 * disk's zone cache. Bursting occurs on a longer interval and only 1198 * runningbufspace is well over the hirunningspace limit. 1199 */ 1200 int bioq_reorder_burst_interval = 60; /* should be multiple of minor */ 1201 SYSCTL_INT(_kern, OID_AUTO, bioq_reorder_burst_interval, 1202 CTLFLAG_RW, &bioq_reorder_burst_interval, 0, ""); 1203 int bioq_reorder_minor_interval = 5; 1204 SYSCTL_INT(_kern, OID_AUTO, bioq_reorder_minor_interval, 1205 CTLFLAG_RW, &bioq_reorder_minor_interval, 0, ""); 1206 1207 int bioq_reorder_burst_bytes = 3000000; 1208 SYSCTL_INT(_kern, OID_AUTO, bioq_reorder_burst_bytes, 1209 CTLFLAG_RW, &bioq_reorder_burst_bytes, 0, ""); 1210 int bioq_reorder_minor_bytes = 262144; 1211 SYSCTL_INT(_kern, OID_AUTO, bioq_reorder_minor_bytes, 1212 CTLFLAG_RW, &bioq_reorder_minor_bytes, 0, ""); 1213 1214 1215 /* 1216 * Order I/Os. Generally speaking this code is designed to make better 1217 * use of drive zone caches. A drive zone cache can typically track linear 1218 * reads or writes for around 16 zones simultaniously. 1219 * 1220 * Read prioritization issues: It is possible for hundreds of megabytes worth 1221 * of writes to be queued asynchronously. This creates a huge bottleneck 1222 * for reads which reduce read bandwidth to a trickle. 1223 * 1224 * To solve this problem we generally reorder reads before writes. 1225 * 1226 * However, a large number of random reads can also starve writes and 1227 * make poor use of the drive zone cache so we allow writes to trickle 1228 * in every N reads. 1229 */ 1230 void 1231 bioqdisksort(struct bio_queue_head *bioq, struct bio *bio) 1232 { 1233 /* 1234 * The BIO wants to be ordered. Adding to the tail also 1235 * causes transition to be set to NULL, forcing the ordering 1236 * of all prior I/O's. 1237 */ 1238 if (bio->bio_buf->b_flags & B_ORDERED) { 1239 bioq_insert_tail(bioq, bio); 1240 return; 1241 } 1242 1243 switch(bio->bio_buf->b_cmd) { 1244 case BUF_CMD_READ: 1245 if (bioq->transition) { 1246 /* 1247 * Insert before the first write. Bleedover writes 1248 * based on reorder intervals to prevent starvation. 1249 */ 1250 TAILQ_INSERT_BEFORE(bioq->transition, bio, bio_act); 1251 ++bioq->reorder; 1252 if (bioq->reorder % bioq_reorder_minor_interval == 0) { 1253 bioqwritereorder(bioq); 1254 if (bioq->reorder >= 1255 bioq_reorder_burst_interval) { 1256 bioq->reorder = 0; 1257 } 1258 } 1259 } else { 1260 /* 1261 * No writes queued (or ordering was forced), 1262 * insert at tail. 1263 */ 1264 TAILQ_INSERT_TAIL(&bioq->queue, bio, bio_act); 1265 } 1266 break; 1267 case BUF_CMD_WRITE: 1268 /* 1269 * Writes are always appended. If no writes were previously 1270 * queued or an ordered tail insertion occured the transition 1271 * field will be NULL. 1272 */ 1273 TAILQ_INSERT_TAIL(&bioq->queue, bio, bio_act); 1274 if (bioq->transition == NULL) 1275 bioq->transition = bio; 1276 break; 1277 default: 1278 /* 1279 * All other request types are forced to be ordered. 1280 */ 1281 bioq_insert_tail(bioq, bio); 1282 break; 1283 } 1284 } 1285 1286 /* 1287 * Move the read-write transition point to prevent reads from 1288 * completely starving our writes. This brings a number of writes into 1289 * the fold every N reads. 1290 * 1291 * We bring a few linear writes into the fold on a minor interval 1292 * and we bring a non-linear burst of writes into the fold on a major 1293 * interval. Bursting only occurs if runningbufspace is really high 1294 * (typically from syncs, fsyncs, or HAMMER flushes). 1295 */ 1296 static 1297 void 1298 bioqwritereorder(struct bio_queue_head *bioq) 1299 { 1300 struct bio *bio; 1301 off_t next_offset; 1302 size_t left; 1303 size_t n; 1304 int check_off; 1305 1306 if (bioq->reorder < bioq_reorder_burst_interval || 1307 !buf_runningbufspace_severe()) { 1308 left = (size_t)bioq_reorder_minor_bytes; 1309 check_off = 1; 1310 } else { 1311 left = (size_t)bioq_reorder_burst_bytes; 1312 check_off = 0; 1313 } 1314 1315 next_offset = bioq->transition->bio_offset; 1316 while ((bio = bioq->transition) != NULL && 1317 (check_off == 0 || next_offset == bio->bio_offset) 1318 ) { 1319 n = bio->bio_buf->b_bcount; 1320 next_offset = bio->bio_offset + n; 1321 bioq->transition = TAILQ_NEXT(bio, bio_act); 1322 if (left < n) 1323 break; 1324 left -= n; 1325 } 1326 } 1327 1328 /* 1329 * Bounds checking against the media size, used for the raw partition. 1330 * secsize, mediasize and b_blkno must all be the same units. 1331 * Possibly this has to be DEV_BSIZE (512). 1332 */ 1333 int 1334 bounds_check_with_mediasize(struct bio *bio, int secsize, uint64_t mediasize) 1335 { 1336 struct buf *bp = bio->bio_buf; 1337 int64_t sz; 1338 1339 sz = howmany(bp->b_bcount, secsize); 1340 1341 if (bio->bio_offset/DEV_BSIZE + sz > mediasize) { 1342 sz = mediasize - bio->bio_offset/DEV_BSIZE; 1343 if (sz == 0) { 1344 /* If exactly at end of disk, return EOF. */ 1345 bp->b_resid = bp->b_bcount; 1346 return 0; 1347 } 1348 if (sz < 0) { 1349 /* If past end of disk, return EINVAL. */ 1350 bp->b_error = EINVAL; 1351 return 0; 1352 } 1353 /* Otherwise, truncate request. */ 1354 bp->b_bcount = sz * secsize; 1355 } 1356 1357 return 1; 1358 } 1359 1360 /* 1361 * Disk error is the preface to plaintive error messages 1362 * about failing disk transfers. It prints messages of the form 1363 1364 hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d) 1365 1366 * if the offset of the error in the transfer and a disk label 1367 * are both available. blkdone should be -1 if the position of the error 1368 * is unknown; the disklabel pointer may be null from drivers that have not 1369 * been converted to use them. The message is printed with kprintf 1370 * if pri is LOG_PRINTF, otherwise it uses log at the specified priority. 1371 * The message should be completed (with at least a newline) with kprintf 1372 * or log(-1, ...), respectively. There is no trailing space. 1373 */ 1374 void 1375 diskerr(struct bio *bio, cdev_t dev, const char *what, int pri, int donecnt) 1376 { 1377 struct buf *bp = bio->bio_buf; 1378 const char *term; 1379 1380 switch(bp->b_cmd) { 1381 case BUF_CMD_READ: 1382 term = "read"; 1383 break; 1384 case BUF_CMD_WRITE: 1385 term = "write"; 1386 break; 1387 default: 1388 term = "access"; 1389 break; 1390 } 1391 kprintf("%s: %s %sing ", dev->si_name, what, term); 1392 kprintf("offset %012llx for %d", 1393 (long long)bio->bio_offset, 1394 bp->b_bcount); 1395 1396 if (donecnt) 1397 kprintf(" (%d bytes completed)", donecnt); 1398 } 1399 1400 /* 1401 * Locate a disk device 1402 */ 1403 cdev_t 1404 disk_locate(const char *devname) 1405 { 1406 return devfs_find_device_by_name("%s", devname); 1407 } 1408 1409 void 1410 disk_config(void *arg) 1411 { 1412 disk_msg_send_sync(DISK_SYNC, NULL, NULL); 1413 } 1414 1415 static void 1416 disk_init(void) 1417 { 1418 struct thread* td_core; 1419 1420 disk_msg_cache = objcache_create("disk-msg-cache", 0, 0, 1421 NULL, NULL, NULL, 1422 objcache_malloc_alloc, 1423 objcache_malloc_free, 1424 &disk_msg_malloc_args); 1425 1426 lwkt_token_init(&disklist_token, "disks"); 1427 1428 /* 1429 * Initialize the reply-only port which acts as a message drain 1430 */ 1431 lwkt_initport_replyonly(&disk_dispose_port, disk_msg_autofree_reply); 1432 1433 lwkt_gettoken(&disklist_token); 1434 lwkt_create(disk_msg_core, /*args*/NULL, &td_core, NULL, 1435 0, -1, "disk_msg_core"); 1436 tsleep(td_core, 0, "diskcore", 0); 1437 lwkt_reltoken(&disklist_token); 1438 } 1439 1440 static void 1441 disk_uninit(void) 1442 { 1443 objcache_destroy(disk_msg_cache); 1444 } 1445 1446 /* 1447 * Clean out illegal characters in serial numbers. 1448 */ 1449 static void 1450 disk_cleanserial(char *serno) 1451 { 1452 char c; 1453 1454 while ((c = *serno) != 0) { 1455 if (c >= 'a' && c <= 'z') 1456 ; 1457 else if (c >= 'A' && c <= 'Z') 1458 ; 1459 else if (c >= '0' && c <= '9') 1460 ; 1461 else if (c == '-' || c == '@' || c == '+' || c == '.') 1462 ; 1463 else 1464 c = '_'; 1465 *serno++= c; 1466 } 1467 } 1468 1469 TUNABLE_INT("kern.disk_debug", &disk_debug_enable); 1470 SYSCTL_INT(_kern, OID_AUTO, disk_debug, CTLFLAG_RW, &disk_debug_enable, 1471 0, "Enable subr_disk debugging"); 1472 1473 SYSINIT(disk_register, SI_SUB_PRE_DRIVERS, SI_ORDER_FIRST, disk_init, NULL); 1474 SYSUNINIT(disk_register, SI_SUB_PRE_DRIVERS, SI_ORDER_ANY, disk_uninit, NULL); 1475