1 /* 2 * Copyright (c) 2003,2004 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 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in 15 * the documentation and/or other materials provided with the 16 * distribution. 17 * 3. Neither the name of The DragonFly Project nor the names of its 18 * contributors may be used to endorse or promote products derived 19 * from this software without specific, prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 22 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS 24 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE 25 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, 26 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING, 27 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; 28 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED 29 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 30 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT 31 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32 * SUCH DAMAGE. 33 * 34 * ---------------------------------------------------------------------------- 35 * "THE BEER-WARE LICENSE" (Revision 42): 36 * <phk@FreeBSD.ORG> wrote this file. As long as you retain this notice you 37 * can do whatever you want with this stuff. If we meet some day, and you think 38 * this stuff is worth it, you can buy me a beer in return. Poul-Henning Kamp 39 * ---------------------------------------------------------------------------- 40 * 41 * Copyright (c) 1982, 1986, 1988, 1993 42 * The Regents of the University of California. All rights reserved. 43 * (c) UNIX System Laboratories, Inc. 44 * All or some portions of this file are derived from material licensed 45 * to the University of California by American Telephone and Telegraph 46 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 47 * the permission of UNIX System Laboratories, Inc. 48 * 49 * Redistribution and use in source and binary forms, with or without 50 * modification, are permitted provided that the following conditions 51 * are met: 52 * 1. Redistributions of source code must retain the above copyright 53 * notice, this list of conditions and the following disclaimer. 54 * 2. Redistributions in binary form must reproduce the above copyright 55 * notice, this list of conditions and the following disclaimer in the 56 * documentation and/or other materials provided with the distribution. 57 * 3. All advertising materials mentioning features or use of this software 58 * must display the following acknowledgement: 59 * This product includes software developed by the University of 60 * California, Berkeley and its contributors. 61 * 4. Neither the name of the University nor the names of its contributors 62 * may be used to endorse or promote products derived from this software 63 * without specific prior written permission. 64 * 65 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 66 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 67 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 68 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 69 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 70 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 71 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 72 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 73 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 74 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 75 * SUCH DAMAGE. 76 * 77 * @(#)ufs_disksubr.c 8.5 (Berkeley) 1/21/94 78 * $FreeBSD: src/sys/kern/subr_disk.c,v 1.20.2.6 2001/10/05 07:14:57 peter Exp $ 79 * $FreeBSD: src/sys/ufs/ufs/ufs_disksubr.c,v 1.44.2.3 2001/03/05 05:42:19 obrien Exp $ 80 * $DragonFly: src/sys/kern/subr_disk.c,v 1.33 2007/05/19 07:05:25 dillon 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/diskslice.h> 92 #include <sys/disk.h> 93 #include <sys/malloc.h> 94 #include <sys/sysctl.h> 95 #include <machine/md_var.h> 96 #include <sys/ctype.h> 97 #include <sys/syslog.h> 98 #include <sys/device.h> 99 #include <sys/msgport.h> 100 #include <sys/msgport2.h> 101 #include <sys/buf2.h> 102 103 static MALLOC_DEFINE(M_DISK, "disk", "disk data"); 104 105 static d_open_t diskopen; 106 static d_close_t diskclose; 107 static d_ioctl_t diskioctl; 108 static d_strategy_t diskstrategy; 109 static d_psize_t diskpsize; 110 static d_clone_t diskclone; 111 static d_dump_t diskdump; 112 113 static LIST_HEAD(, disk) disklist = LIST_HEAD_INITIALIZER(&disklist); 114 115 static struct dev_ops disk_ops = { 116 { "disk" }, 117 .d_open = diskopen, 118 .d_close = diskclose, 119 .d_read = physread, 120 .d_write = physwrite, 121 .d_ioctl = diskioctl, 122 .d_strategy = diskstrategy, 123 .d_dump = diskdump, 124 .d_psize = diskpsize, 125 .d_clone = diskclone 126 }; 127 128 /* 129 * Create a raw device for the dev_ops template (which is returned). Also 130 * create a slice and unit managed disk and overload the user visible 131 * device space with it. 132 * 133 * NOTE: The returned raw device is NOT a slice and unit managed device. 134 * It is an actual raw device representing the raw disk as specified by 135 * the passed dev_ops. The disk layer not only returns such a raw device, 136 * it also uses it internally when passing (modified) commands through. 137 */ 138 cdev_t 139 disk_create(int unit, struct disk *dp, struct dev_ops *raw_ops) 140 { 141 cdev_t rawdev; 142 struct dev_ops *dev_ops; 143 144 /* 145 * Create the raw backing device 146 */ 147 compile_dev_ops(raw_ops); 148 rawdev = make_dev(raw_ops, dkmakewholedisk(unit), 149 UID_ROOT, GID_OPERATOR, 0640, 150 "%s%d", raw_ops->head.name, unit); 151 152 bzero(dp, sizeof(*dp)); 153 154 /* 155 * We install a custom cdevsw rather then the passed cdevsw, 156 * and save our disk structure in d_data so we can get at it easily 157 * without any complex cloning code. 158 */ 159 dev_ops = dev_ops_add_override(rawdev, &disk_ops, 160 dkunitmask(), dkmakeunit(unit)); 161 dev_ops->head.data = dp; 162 163 dp->d_rawdev = rawdev; 164 dp->d_raw_ops = raw_ops; 165 dp->d_dev_ops = dev_ops; 166 dp->d_cdev = make_dev(dev_ops, 167 dkmakewholedisk(unit), 168 UID_ROOT, GID_OPERATOR, 0640, 169 "%s%d", dev_ops->head.name, unit); 170 171 LIST_INSERT_HEAD(&disklist, dp, d_list); 172 return (dp->d_rawdev); 173 } 174 175 /* 176 * Disk drivers must call this routine when media parameters are available 177 * or have changed. 178 */ 179 void 180 disk_setdiskinfo(struct disk *disk, struct disk_info *info) 181 { 182 bcopy(info, &disk->d_info, sizeof(disk->d_info)); 183 info = &disk->d_info; 184 185 KKASSERT(info->d_media_size == 0 || info->d_media_blksize == 0); 186 if (info->d_media_size == 0 && info->d_media_blocks) { 187 info->d_media_size = (u_int64_t)info->d_media_blocks * 188 info->d_media_blksize; 189 } else if (info->d_media_size && info->d_media_blocks == 0 && 190 info->d_media_blksize) { 191 info->d_media_blocks = info->d_media_size / 192 info->d_media_blksize; 193 } 194 } 195 196 /* 197 * This routine is called when an adapter detaches. The higher level 198 * managed disk device is destroyed while the lower level raw device is 199 * released. 200 */ 201 void 202 disk_destroy(struct disk *disk) 203 { 204 if (disk->d_dev_ops) { 205 dev_ops_remove(disk->d_dev_ops, dkunitmask(), 206 dkmakeunit(dkunit(disk->d_cdev))); 207 LIST_REMOVE(disk, d_list); 208 } 209 if (disk->d_raw_ops) { 210 destroy_all_devs(disk->d_raw_ops, dkunitmask(), 211 dkmakeunit(dkunit(disk->d_rawdev))); 212 } 213 bzero(disk, sizeof(*disk)); 214 } 215 216 int 217 disk_dumpcheck(cdev_t dev, u_int64_t *count, u_int64_t *blkno, u_int *secsize) 218 { 219 struct partinfo pinfo; 220 int error; 221 222 bzero(&pinfo, sizeof(pinfo)); 223 error = dev_dioctl(dev, DIOCGPART, (void *)&pinfo, 0, proc0.p_ucred); 224 if (error) 225 return (error); 226 if (pinfo.media_blksize == 0) 227 return (ENXIO); 228 *count = (u_int64_t)Maxmem * PAGE_SIZE / pinfo.media_blksize; 229 if (dumplo64 < pinfo.skip_bsdlabel || 230 dumplo64 + *count > pinfo.media_blocks) { 231 return (ENOSPC); 232 } 233 *blkno = dumplo64 + pinfo.media_offset / pinfo.media_blksize; 234 *secsize = pinfo.media_blksize; 235 return (0); 236 } 237 238 void 239 disk_invalidate (struct disk *disk) 240 { 241 if (disk->d_slice) 242 dsgone(&disk->d_slice); 243 } 244 245 struct disk * 246 disk_enumerate(struct disk *disk) 247 { 248 if (!disk) 249 return (LIST_FIRST(&disklist)); 250 else 251 return (LIST_NEXT(disk, d_list)); 252 } 253 254 static 255 int 256 sysctl_disks(SYSCTL_HANDLER_ARGS) 257 { 258 struct disk *disk; 259 int error, first; 260 261 disk = NULL; 262 first = 1; 263 264 while ((disk = disk_enumerate(disk))) { 265 if (!first) { 266 error = SYSCTL_OUT(req, " ", 1); 267 if (error) 268 return error; 269 } else { 270 first = 0; 271 } 272 error = SYSCTL_OUT(req, disk->d_rawdev->si_name, 273 strlen(disk->d_rawdev->si_name)); 274 if (error) 275 return error; 276 } 277 error = SYSCTL_OUT(req, "", 1); 278 return error; 279 } 280 281 SYSCTL_PROC(_kern, OID_AUTO, disks, CTLTYPE_STRING | CTLFLAG_RD, 0, NULL, 282 sysctl_disks, "A", "names of available disks"); 283 284 /* 285 * Open a disk device or partition. 286 */ 287 static 288 int 289 diskopen(struct dev_open_args *ap) 290 { 291 cdev_t dev = ap->a_head.a_dev; 292 struct disk *dp; 293 int error; 294 295 /* 296 * dp can't be NULL here XXX. 297 */ 298 dp = dev->si_disk; 299 if (dp == NULL) 300 return (ENXIO); 301 error = 0; 302 303 /* 304 * Deal with open races 305 */ 306 while (dp->d_flags & DISKFLAG_LOCK) { 307 dp->d_flags |= DISKFLAG_WANTED; 308 error = tsleep(dp, PCATCH, "diskopen", hz); 309 if (error) 310 return (error); 311 } 312 dp->d_flags |= DISKFLAG_LOCK; 313 314 /* 315 * Open the underlying raw device. 316 */ 317 if (!dsisopen(dp->d_slice)) { 318 #if 0 319 if (!pdev->si_iosize_max) 320 pdev->si_iosize_max = dev->si_iosize_max; 321 #endif 322 error = dev_dopen(dp->d_rawdev, ap->a_oflags, 323 ap->a_devtype, ap->a_cred); 324 } 325 326 /* 327 * Inherit properties from the underlying device now that it is 328 * open. 329 */ 330 dev_dclone(dev); 331 332 if (error) 333 goto out; 334 335 error = dsopen(dev, ap->a_devtype, dp->d_info.d_dsflags, 336 &dp->d_slice, &dp->d_info); 337 338 if (!dsisopen(dp->d_slice)) 339 dev_dclose(dp->d_rawdev, ap->a_oflags, ap->a_devtype); 340 out: 341 dp->d_flags &= ~DISKFLAG_LOCK; 342 if (dp->d_flags & DISKFLAG_WANTED) { 343 dp->d_flags &= ~DISKFLAG_WANTED; 344 wakeup(dp); 345 } 346 347 return(error); 348 } 349 350 /* 351 * Close a disk device or partition 352 */ 353 static 354 int 355 diskclose(struct dev_close_args *ap) 356 { 357 cdev_t dev = ap->a_head.a_dev; 358 struct disk *dp; 359 int error; 360 361 error = 0; 362 dp = dev->si_disk; 363 364 dsclose(dev, ap->a_devtype, dp->d_slice); 365 if (!dsisopen(dp->d_slice)) 366 error = dev_dclose(dp->d_rawdev, ap->a_fflag, ap->a_devtype); 367 return (error); 368 } 369 370 /* 371 * First execute the ioctl on the disk device, and if it isn't supported 372 * try running it on the backing device. 373 */ 374 static 375 int 376 diskioctl(struct dev_ioctl_args *ap) 377 { 378 cdev_t dev = ap->a_head.a_dev; 379 struct disk *dp; 380 int error; 381 382 dp = dev->si_disk; 383 if (dp == NULL) 384 return (ENXIO); 385 error = dsioctl(dev, ap->a_cmd, ap->a_data, ap->a_fflag, 386 &dp->d_slice, &dp->d_info); 387 if (error == ENOIOCTL) { 388 error = dev_dioctl(dp->d_rawdev, ap->a_cmd, ap->a_data, 389 ap->a_fflag, ap->a_cred); 390 } 391 return (error); 392 } 393 394 /* 395 * Execute strategy routine 396 */ 397 static 398 int 399 diskstrategy(struct dev_strategy_args *ap) 400 { 401 cdev_t dev = ap->a_head.a_dev; 402 struct bio *bio = ap->a_bio; 403 struct bio *nbio; 404 struct disk *dp; 405 406 dp = dev->si_disk; 407 408 if (dp == NULL) { 409 bio->bio_buf->b_error = ENXIO; 410 bio->bio_buf->b_flags |= B_ERROR; 411 biodone(bio); 412 return(0); 413 } 414 KKASSERT(dev->si_disk == dp); 415 416 /* 417 * The dscheck() function will also transform the slice relative 418 * block number i.e. bio->bio_offset into a block number that can be 419 * passed directly to the underlying raw device. If dscheck() 420 * returns NULL it will have handled the bio for us (e.g. EOF 421 * or error due to being beyond the device size). 422 */ 423 if ((nbio = dscheck(dev, bio, dp->d_slice)) != NULL) 424 dev_dstrategy(dp->d_rawdev, nbio); 425 else 426 biodone(bio); 427 return(0); 428 } 429 430 /* 431 * Return the partition size in ?blocks? 432 */ 433 static 434 int 435 diskpsize(struct dev_psize_args *ap) 436 { 437 cdev_t dev = ap->a_head.a_dev; 438 struct disk *dp; 439 440 dp = dev->si_disk; 441 if (dp == NULL) 442 return(ENODEV); 443 ap->a_result = dssize(dev, &dp->d_slice); 444 return(0); 445 } 446 447 /* 448 * When new device entries are instantiated, make sure they inherit our 449 * si_disk structure and block and iosize limits from the raw device. 450 * 451 * This routine is always called synchronously in the context of the 452 * client. 453 * 454 * XXX The various io and block size constraints are not always initialized 455 * properly by devices. 456 */ 457 static 458 int 459 diskclone(struct dev_clone_args *ap) 460 { 461 cdev_t dev = ap->a_head.a_dev; 462 struct disk *dp; 463 464 dp = dev->si_ops->head.data; 465 KKASSERT(dp != NULL); 466 dev->si_disk = dp; 467 dev->si_iosize_max = dp->d_rawdev->si_iosize_max; 468 dev->si_bsize_phys = dp->d_rawdev->si_bsize_phys; 469 dev->si_bsize_best = dp->d_rawdev->si_bsize_best; 470 return(0); 471 } 472 473 int 474 diskdump(struct dev_dump_args *ap) 475 { 476 cdev_t dev = ap->a_head.a_dev; 477 struct disk *dp = dev->si_ops->head.data; 478 int error; 479 480 error = disk_dumpcheck(dev, &ap->a_count, &ap->a_blkno, &ap->a_secsize); 481 if (error == 0) { 482 ap->a_head.a_dev = dp->d_rawdev; 483 error = dev_doperate(&ap->a_head); 484 } 485 486 return(error); 487 } 488 489 490 SYSCTL_INT(_debug_sizeof, OID_AUTO, diskslices, CTLFLAG_RD, 491 0, sizeof(struct diskslices), "sizeof(struct diskslices)"); 492 493 SYSCTL_INT(_debug_sizeof, OID_AUTO, disk, CTLFLAG_RD, 494 0, sizeof(struct disk), "sizeof(struct disk)"); 495 496 497 /* 498 * Seek sort for disks. 499 * 500 * The bio_queue keep two queues, sorted in ascending block order. The first 501 * queue holds those requests which are positioned after the current block 502 * (in the first request); the second, which starts at queue->switch_point, 503 * holds requests which came in after their block number was passed. Thus 504 * we implement a one way scan, retracting after reaching the end of the drive 505 * to the first request on the second queue, at which time it becomes the 506 * first queue. 507 * 508 * A one-way scan is natural because of the way UNIX read-ahead blocks are 509 * allocated. 510 */ 511 void 512 bioqdisksort(struct bio_queue_head *bioq, struct bio *bio) 513 { 514 struct bio *bq; 515 struct bio *bn; 516 struct bio *be; 517 518 be = TAILQ_LAST(&bioq->queue, bio_queue); 519 /* 520 * If the queue is empty or we are an 521 * ordered transaction, then it's easy. 522 */ 523 if ((bq = bioq_first(bioq)) == NULL || 524 (bio->bio_buf->b_flags & B_ORDERED) != 0) { 525 bioq_insert_tail(bioq, bio); 526 return; 527 } else if (bioq->insert_point != NULL) { 528 529 /* 530 * A certain portion of the list is 531 * "locked" to preserve ordering, so 532 * we can only insert after the insert 533 * point. 534 */ 535 bq = bioq->insert_point; 536 } else { 537 538 /* 539 * If we lie before the last removed (currently active) 540 * request, and are not inserting ourselves into the 541 * "locked" portion of the list, then we must add ourselves 542 * to the second request list. 543 */ 544 if (bio->bio_offset < bioq->last_offset) { 545 bq = bioq->switch_point; 546 /* 547 * If we are starting a new secondary list, 548 * then it's easy. 549 */ 550 if (bq == NULL) { 551 bioq->switch_point = bio; 552 bioq_insert_tail(bioq, bio); 553 return; 554 } 555 /* 556 * If we lie ahead of the current switch point, 557 * insert us before the switch point and move 558 * the switch point. 559 */ 560 if (bio->bio_offset < bq->bio_offset) { 561 bioq->switch_point = bio; 562 TAILQ_INSERT_BEFORE(bq, bio, bio_act); 563 return; 564 } 565 } else { 566 if (bioq->switch_point != NULL) 567 be = TAILQ_PREV(bioq->switch_point, 568 bio_queue, bio_act); 569 /* 570 * If we lie between last_offset and bq, 571 * insert before bq. 572 */ 573 if (bio->bio_offset < bq->bio_offset) { 574 TAILQ_INSERT_BEFORE(bq, bio, bio_act); 575 return; 576 } 577 } 578 } 579 580 /* 581 * Request is at/after our current position in the list. 582 * Optimize for sequential I/O by seeing if we go at the tail. 583 */ 584 if (bio->bio_offset > be->bio_offset) { 585 TAILQ_INSERT_AFTER(&bioq->queue, be, bio, bio_act); 586 return; 587 } 588 589 /* Otherwise, insertion sort */ 590 while ((bn = TAILQ_NEXT(bq, bio_act)) != NULL) { 591 592 /* 593 * We want to go after the current request if it is the end 594 * of the first request list, or if the next request is a 595 * larger cylinder than our request. 596 */ 597 if (bn == bioq->switch_point 598 || bio->bio_offset < bn->bio_offset) 599 break; 600 bq = bn; 601 } 602 TAILQ_INSERT_AFTER(&bioq->queue, bq, bio, bio_act); 603 } 604 605 606 /* 607 * Attempt to read a disk label from a device using the indicated strategy 608 * routine. The label must be partly set up before this: secpercyl, secsize 609 * and anything required in the strategy routine (e.g., dummy bounds for the 610 * partition containing the label) must be filled in before calling us. 611 * Returns NULL on success and an error string on failure. 612 */ 613 char * 614 readdisklabel(cdev_t dev, struct disklabel *lp) 615 { 616 struct buf *bp; 617 struct disklabel *dlp; 618 char *msg = NULL; 619 620 bp = geteblk((int)lp->d_secsize); 621 bp->b_bio1.bio_offset = (off_t)LABELSECTOR * lp->d_secsize; 622 bp->b_bcount = lp->d_secsize; 623 bp->b_flags &= ~B_INVAL; 624 bp->b_cmd = BUF_CMD_READ; 625 dev_dstrategy(dev, &bp->b_bio1); 626 if (biowait(bp)) 627 msg = "I/O error"; 628 else for (dlp = (struct disklabel *)bp->b_data; 629 dlp <= (struct disklabel *)((char *)bp->b_data + 630 lp->d_secsize - sizeof(*dlp)); 631 dlp = (struct disklabel *)((char *)dlp + sizeof(long))) { 632 if (dlp->d_magic != DISKMAGIC || dlp->d_magic2 != DISKMAGIC) { 633 if (msg == NULL) 634 msg = "no disk label"; 635 } else if (dlp->d_npartitions > MAXPARTITIONS || 636 dkcksum(dlp) != 0) 637 msg = "disk label corrupted"; 638 else { 639 *lp = *dlp; 640 msg = NULL; 641 break; 642 } 643 } 644 bp->b_flags |= B_INVAL | B_AGE; 645 brelse(bp); 646 return (msg); 647 } 648 649 /* 650 * Check new disk label for sensibility before setting it. 651 */ 652 int 653 setdisklabel(struct disklabel *olp, struct disklabel *nlp, u_long openmask) 654 { 655 int i; 656 struct partition *opp, *npp; 657 658 /* 659 * Check it is actually a disklabel we are looking at. 660 */ 661 if (nlp->d_magic != DISKMAGIC || nlp->d_magic2 != DISKMAGIC || 662 dkcksum(nlp) != 0) 663 return (EINVAL); 664 /* 665 * For each partition that we think is open, 666 */ 667 while ((i = ffs((long)openmask)) != 0) { 668 i--; 669 /* 670 * Check it is not changing.... 671 */ 672 openmask &= ~(1 << i); 673 if (nlp->d_npartitions <= i) 674 return (EBUSY); 675 opp = &olp->d_partitions[i]; 676 npp = &nlp->d_partitions[i]; 677 if (npp->p_offset != opp->p_offset || npp->p_size < opp->p_size) 678 return (EBUSY); 679 /* 680 * Copy internally-set partition information 681 * if new label doesn't include it. XXX 682 * (If we are using it then we had better stay the same type) 683 * This is possibly dubious, as someone else noted (XXX) 684 */ 685 if (npp->p_fstype == FS_UNUSED && opp->p_fstype != FS_UNUSED) { 686 npp->p_fstype = opp->p_fstype; 687 npp->p_fsize = opp->p_fsize; 688 npp->p_frag = opp->p_frag; 689 npp->p_cpg = opp->p_cpg; 690 } 691 } 692 nlp->d_checksum = 0; 693 nlp->d_checksum = dkcksum(nlp); 694 *olp = *nlp; 695 return (0); 696 } 697 698 /* 699 * Write disk label back to device after modification. 700 */ 701 int 702 writedisklabel(cdev_t dev, struct disklabel *lp) 703 { 704 struct buf *bp; 705 struct disklabel *dlp; 706 int error = 0; 707 708 if (lp->d_partitions[RAW_PART].p_offset != 0) 709 return (EXDEV); /* not quite right */ 710 bp = geteblk((int)lp->d_secsize); 711 bp->b_bio1.bio_offset = (off_t)LABELSECTOR * lp->d_secsize; 712 bp->b_bcount = lp->d_secsize; 713 #if 1 714 /* 715 * We read the label first to see if it's there, 716 * in which case we will put ours at the same offset into the block.. 717 * (I think this is stupid [Julian]) 718 * Note that you can't write a label out over a corrupted label! 719 * (also stupid.. how do you write the first one? by raw writes?) 720 */ 721 bp->b_flags &= ~B_INVAL; 722 bp->b_cmd = BUF_CMD_READ; 723 dev_dstrategy(dkmodpart(dev, WHOLE_SLICE_PART), &bp->b_bio1); 724 error = biowait(bp); 725 if (error) 726 goto done; 727 for (dlp = (struct disklabel *)bp->b_data; 728 dlp <= (struct disklabel *) 729 ((char *)bp->b_data + lp->d_secsize - sizeof(*dlp)); 730 dlp = (struct disklabel *)((char *)dlp + sizeof(long))) { 731 if (dlp->d_magic == DISKMAGIC && dlp->d_magic2 == DISKMAGIC && 732 dkcksum(dlp) == 0) { 733 *dlp = *lp; 734 bp->b_cmd = BUF_CMD_WRITE; 735 dev_dstrategy(dkmodpart(dev, WHOLE_SLICE_PART), &bp->b_bio1); 736 error = biowait(bp); 737 goto done; 738 } 739 } 740 error = ESRCH; 741 done: 742 #else 743 bzero(bp->b_data, lp->d_secsize); 744 dlp = (struct disklabel *)bp->b_data; 745 *dlp = *lp; 746 bp->b_flags &= ~B_INVAL; 747 bp->b_cmd = BUF_CMD_WRITE; 748 BUF_STRATEGY(bp, 1); 749 error = biowait(bp); 750 #endif 751 bp->b_flags |= B_INVAL | B_AGE; 752 brelse(bp); 753 return (error); 754 } 755 756 /* 757 * Disk error is the preface to plaintive error messages 758 * about failing disk transfers. It prints messages of the form 759 760 hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d) 761 762 * if the offset of the error in the transfer and a disk label 763 * are both available. blkdone should be -1 if the position of the error 764 * is unknown; the disklabel pointer may be null from drivers that have not 765 * been converted to use them. The message is printed with kprintf 766 * if pri is LOG_PRINTF, otherwise it uses log at the specified priority. 767 * The message should be completed (with at least a newline) with kprintf 768 * or log(-1, ...), respectively. There is no trailing space. 769 */ 770 void 771 diskerr(struct bio *bio, cdev_t dev, const char *what, int pri, int donecnt) 772 { 773 struct buf *bp = bio->bio_buf; 774 int unit = dkunit(dev); 775 int slice = dkslice(dev); 776 int part = dkpart(dev); 777 char partname[2]; 778 char *sname; 779 780 sname = dsname(dev, unit, slice, part, partname); 781 kprintf("%s%s: %s %sing ", sname, partname, what, 782 (bp->b_cmd == BUF_CMD_READ) ? "read" : "writ"); 783 kprintf("offset %012llx for %d", bio->bio_offset, bp->b_bcount); 784 if (donecnt) 785 kprintf(" (%d bytes completed)", donecnt); 786 } 787 788