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 1.20 2005/09/10 21:01:20 swildner 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_strategy_t diskstrategy; 106 static d_open_t diskopen; 107 static d_close_t diskclose; 108 static d_ioctl_t diskioctl; 109 static d_psize_t diskpsize; 110 static d_clone_t diskclone; 111 static int disk_putport(lwkt_port_t port, lwkt_msg_t msg); 112 113 static LIST_HEAD(, disk) disklist = LIST_HEAD_INITIALIZER(&disklist); 114 115 /* 116 * Create a slice and unit managed disk. 117 * 118 * Our port layer will be responsible for assigning pblkno and handling 119 * high level partition operations, then forwarding the requests to the 120 * raw device. 121 * 122 * The disk_create() function clones the provided rawsw for creating a 123 * managed disk device. In addition, the cdevsw intercept port is 124 * changed to disk_putport, which is used to transform requests for the 125 * managed disk device. 126 * 127 * The raw device (based on rawsw) is returned to the caller, NOT the 128 * slice and unit managed cdev. The caller typically sets various 129 * driver parameters and IO limits on the returned rawdev which we must 130 * inherit when our managed device is opened. 131 */ 132 dev_t 133 disk_create(int unit, struct disk *dp, int flags, struct cdevsw *rawsw) 134 { 135 dev_t rawdev; 136 struct cdevsw *devsw; 137 138 /* 139 * Create the raw backing device 140 */ 141 compile_devsw(rawsw); 142 rawdev = make_dev(rawsw, dkmakeminor(unit, WHOLE_DISK_SLICE, RAW_PART), 143 UID_ROOT, GID_OPERATOR, 0640, 144 "%s%d", rawsw->d_name, unit); 145 146 /* 147 * Initialize our intercept port 148 */ 149 bzero(dp, sizeof(*dp)); 150 lwkt_initport(&dp->d_port, NULL); 151 dp->d_port.mp_putport = disk_putport; 152 dp->d_rawsw = rawsw; 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 devsw = cdevsw_add_override(rawdev, dkunitmask(), dkmakeunit(unit)); 160 devsw->d_port = &dp->d_port; 161 devsw->d_data = dp; 162 devsw->d_clone = diskclone; 163 dp->d_devsw = devsw; 164 dp->d_rawdev = rawdev; 165 dp->d_cdev = make_dev(devsw, 166 dkmakeminor(unit, WHOLE_DISK_SLICE, RAW_PART), 167 UID_ROOT, GID_OPERATOR, 0640, 168 "%s%d", devsw->d_name, unit); 169 170 dp->d_dsflags = flags; 171 LIST_INSERT_HEAD(&disklist, dp, d_list); 172 return (dp->d_rawdev); 173 } 174 175 /* 176 * This routine is called when an adapter detaches. The higher level 177 * managed disk device is destroyed while the lower level raw device is 178 * released. 179 */ 180 void 181 disk_destroy(struct disk *disk) 182 { 183 if (disk->d_devsw) { 184 cdevsw_remove(disk->d_devsw, dkunitmask(), 185 dkmakeunit(dkunit(disk->d_cdev))); 186 LIST_REMOVE(disk, d_list); 187 } 188 if (disk->d_rawsw) { 189 destroy_all_dev(disk->d_rawsw, dkunitmask(), 190 dkmakeunit(dkunit(disk->d_rawdev))); 191 } 192 bzero(disk, sizeof(*disk)); 193 } 194 195 int 196 disk_dumpcheck(dev_t dev, u_int *count, u_int *blkno, u_int *secsize) 197 { 198 struct disk *dp; 199 struct disklabel *dl; 200 u_int boff; 201 202 dp = dev->si_disk; 203 if (!dp) 204 return (ENXIO); 205 if (!dp->d_slice) 206 return (ENXIO); 207 dl = dsgetlabel(dev, dp->d_slice); 208 if (!dl) 209 return (ENXIO); 210 *count = Maxmem * (PAGE_SIZE / dl->d_secsize); 211 if (dumplo <= LABELSECTOR || 212 (dumplo + *count > dl->d_partitions[dkpart(dev)].p_size)) 213 return (EINVAL); 214 boff = dl->d_partitions[dkpart(dev)].p_offset + 215 dp->d_slice->dss_slices[dkslice(dev)].ds_offset; 216 *blkno = boff + dumplo; 217 *secsize = dl->d_secsize; 218 return (0); 219 220 } 221 222 void 223 disk_invalidate (struct disk *disk) 224 { 225 if (disk->d_slice) 226 dsgone(&disk->d_slice); 227 } 228 229 struct disk * 230 disk_enumerate(struct disk *disk) 231 { 232 if (!disk) 233 return (LIST_FIRST(&disklist)); 234 else 235 return (LIST_NEXT(disk, d_list)); 236 } 237 238 static 239 int 240 sysctl_disks(SYSCTL_HANDLER_ARGS) 241 { 242 struct disk *disk; 243 int error, first; 244 245 disk = NULL; 246 first = 1; 247 248 while ((disk = disk_enumerate(disk))) { 249 if (!first) { 250 error = SYSCTL_OUT(req, " ", 1); 251 if (error) 252 return error; 253 } else { 254 first = 0; 255 } 256 error = SYSCTL_OUT(req, disk->d_rawdev->si_name, 257 strlen(disk->d_rawdev->si_name)); 258 if (error) 259 return error; 260 } 261 error = SYSCTL_OUT(req, "", 1); 262 return error; 263 } 264 265 SYSCTL_PROC(_kern, OID_AUTO, disks, CTLTYPE_STRING | CTLFLAG_RD, 0, NULL, 266 sysctl_disks, "A", "names of available disks"); 267 268 /* 269 * The port intercept functions 270 */ 271 static 272 int 273 disk_putport(lwkt_port_t port, lwkt_msg_t lmsg) 274 { 275 struct disk *disk = (struct disk *)port; 276 cdevallmsg_t msg = (cdevallmsg_t)lmsg; 277 int error; 278 279 switch(msg->am_lmsg.ms_cmd.cm_op) { 280 case CDEV_CMD_OPEN: 281 error = diskopen( 282 msg->am_open.msg.dev, 283 msg->am_open.oflags, 284 msg->am_open.devtype, 285 msg->am_open.td); 286 break; 287 case CDEV_CMD_CLOSE: 288 error = diskclose( 289 msg->am_close.msg.dev, 290 msg->am_close.fflag, 291 msg->am_close.devtype, 292 msg->am_close.td); 293 break; 294 case CDEV_CMD_IOCTL: 295 error = diskioctl( 296 msg->am_ioctl.msg.dev, 297 msg->am_ioctl.cmd, 298 msg->am_ioctl.data, 299 msg->am_ioctl.fflag, 300 msg->am_ioctl.td); 301 break; 302 case CDEV_CMD_STRATEGY: 303 diskstrategy(msg->am_strategy.bp); 304 error = 0; 305 break; 306 case CDEV_CMD_PSIZE: 307 msg->am_psize.result = diskpsize(msg->am_psize.msg.dev); 308 error = 0; /* XXX */ 309 break; 310 case CDEV_CMD_READ: 311 error = physio(msg->am_read.msg.dev, 312 msg->am_read.uio, msg->am_read.ioflag); 313 break; 314 case CDEV_CMD_WRITE: 315 error = physio(msg->am_write.msg.dev, 316 msg->am_write.uio, msg->am_write.ioflag); 317 break; 318 case CDEV_CMD_POLL: 319 case CDEV_CMD_KQFILTER: 320 error = ENODEV; 321 case CDEV_CMD_MMAP: 322 error = -1; 323 break; 324 case CDEV_CMD_DUMP: 325 error = disk_dumpcheck(msg->am_dump.msg.dev, 326 &msg->am_dump.count, 327 &msg->am_dump.blkno, 328 &msg->am_dump.secsize); 329 if (error == 0) { 330 msg->am_dump.msg.dev = disk->d_rawdev; 331 error = lwkt_forwardmsg(disk->d_rawdev->si_port, 332 &msg->am_dump.msg.msg); 333 printf("error2 %d\n", error); 334 } 335 break; 336 default: 337 error = ENOTSUP; 338 break; 339 } 340 return(error); 341 } 342 343 /* 344 * When new device entries are instantiated, make sure they inherit our 345 * si_disk structure and block and iosize limits from the raw device. 346 * 347 * This routine is always called synchronously in the context of the 348 * client. 349 * 350 * XXX The various io and block size constraints are not always initialized 351 * properly by devices. 352 */ 353 static 354 int 355 diskclone(dev_t dev) 356 { 357 struct disk *dp; 358 359 dp = dev->si_devsw->d_data; 360 KKASSERT(dp != NULL); 361 dev->si_disk = dp; 362 dev->si_iosize_max = dp->d_rawdev->si_iosize_max; 363 dev->si_bsize_phys = dp->d_rawdev->si_bsize_phys; 364 dev->si_bsize_best = dp->d_rawdev->si_bsize_best; 365 return(0); 366 } 367 368 /* 369 * Open a disk device or partition. 370 */ 371 static 372 int 373 diskopen(dev_t dev, int oflags, int devtype, struct thread *td) 374 { 375 struct disk *dp; 376 int error; 377 378 /* 379 * dp can't be NULL here XXX. 380 */ 381 error = 0; 382 dp = dev->si_disk; 383 if (dp == NULL) 384 return (ENXIO); 385 386 /* 387 * Deal with open races 388 */ 389 while (dp->d_flags & DISKFLAG_LOCK) { 390 dp->d_flags |= DISKFLAG_WANTED; 391 error = tsleep(dp, PCATCH, "diskopen", hz); 392 if (error) 393 return (error); 394 } 395 dp->d_flags |= DISKFLAG_LOCK; 396 397 /* 398 * Open the underlying raw device. 399 */ 400 if (!dsisopen(dp->d_slice)) { 401 #if 0 402 if (!pdev->si_iosize_max) 403 pdev->si_iosize_max = dev->si_iosize_max; 404 #endif 405 error = dev_dopen(dp->d_rawdev, oflags, devtype, td); 406 } 407 408 /* 409 * Inherit properties from the underlying device now that it is 410 * open. 411 */ 412 diskclone(dev); 413 414 if (error) 415 goto out; 416 417 error = dsopen(dev, devtype, dp->d_dsflags, &dp->d_slice, &dp->d_label); 418 419 if (!dsisopen(dp->d_slice)) 420 dev_dclose(dp->d_rawdev, oflags, devtype, td); 421 out: 422 dp->d_flags &= ~DISKFLAG_LOCK; 423 if (dp->d_flags & DISKFLAG_WANTED) { 424 dp->d_flags &= ~DISKFLAG_WANTED; 425 wakeup(dp); 426 } 427 428 return(error); 429 } 430 431 /* 432 * Close a disk device or partition 433 */ 434 static 435 int 436 diskclose(dev_t dev, int fflag, int devtype, struct thread *td) 437 { 438 struct disk *dp; 439 int error; 440 441 error = 0; 442 dp = dev->si_disk; 443 444 dsclose(dev, devtype, dp->d_slice); 445 if (!dsisopen(dp->d_slice)) 446 error = dev_dclose(dp->d_rawdev, fflag, devtype, td); 447 return (error); 448 } 449 450 /* 451 * Execute strategy routine 452 */ 453 static 454 void 455 diskstrategy(struct buf *bp) 456 { 457 struct disk *dp; 458 459 dp = bp->b_dev->si_disk; 460 461 if (dp == NULL) { 462 bp->b_error = ENXIO; 463 bp->b_flags |= B_ERROR; 464 biodone(bp); 465 return; 466 } 467 KKASSERT(bp->b_dev->si_disk == dp); 468 469 /* 470 * The dscheck() function will also transform the slice relative 471 * block number i.e. bp->b_blkno into a block number that can be 472 * passed directly to the underlying raw device. 473 */ 474 if (dscheck(bp, dp->d_slice) <= 0) { 475 biodone(bp); 476 return; 477 } 478 bp->b_dev = dp->d_rawdev; 479 dev_dstrategy(dp->d_rawdev, bp); 480 } 481 482 /* 483 * First execute the ioctl on the disk device, and if it isn't supported 484 * try running it on the backing device. 485 */ 486 static 487 int 488 diskioctl(dev_t dev, u_long cmd, caddr_t data, int fflag, struct thread *td) 489 { 490 struct disk *dp; 491 int error; 492 493 dp = dev->si_disk; 494 if (dp == NULL) 495 return (ENXIO); 496 497 error = dsioctl(dev, cmd, data, fflag, &dp->d_slice); 498 if (error == ENOIOCTL) 499 error = dev_dioctl(dp->d_rawdev, cmd, data, fflag, td); 500 return (error); 501 } 502 503 /* 504 * 505 */ 506 static 507 int 508 diskpsize(dev_t dev) 509 { 510 struct disk *dp; 511 512 dp = dev->si_disk; 513 if (dp == NULL) 514 return (-1); 515 return(dssize(dev, &dp->d_slice)); 516 #if 0 517 if (dp != dev->si_disk) { 518 dev->si_drv1 = pdev->si_drv1; 519 dev->si_drv2 = pdev->si_drv2; 520 /* XXX: don't set bp->b_dev->si_disk (?) */ 521 } 522 #endif 523 } 524 525 SYSCTL_INT(_debug_sizeof, OID_AUTO, disklabel, CTLFLAG_RD, 526 0, sizeof(struct disklabel), "sizeof(struct disklabel)"); 527 528 SYSCTL_INT(_debug_sizeof, OID_AUTO, diskslices, CTLFLAG_RD, 529 0, sizeof(struct diskslices), "sizeof(struct diskslices)"); 530 531 SYSCTL_INT(_debug_sizeof, OID_AUTO, disk, CTLFLAG_RD, 532 0, sizeof(struct disk), "sizeof(struct disk)"); 533 534 535 /* 536 * Seek sort for disks. 537 * 538 * The buf_queue keep two queues, sorted in ascending block order. The first 539 * queue holds those requests which are positioned after the current block 540 * (in the first request); the second, which starts at queue->switch_point, 541 * holds requests which came in after their block number was passed. Thus 542 * we implement a one way scan, retracting after reaching the end of the drive 543 * to the first request on the second queue, at which time it becomes the 544 * first queue. 545 * 546 * A one-way scan is natural because of the way UNIX read-ahead blocks are 547 * allocated. 548 */ 549 void 550 bufqdisksort(struct buf_queue_head *bufq, struct buf *bp) 551 { 552 struct buf *bq; 553 struct buf *bn; 554 struct buf *be; 555 556 be = TAILQ_LAST(&bufq->queue, buf_queue); 557 /* 558 * If the queue is empty or we are an 559 * ordered transaction, then it's easy. 560 */ 561 if ((bq = bufq_first(bufq)) == NULL || 562 (bp->b_flags & B_ORDERED) != 0) { 563 bufq_insert_tail(bufq, bp); 564 return; 565 } else if (bufq->insert_point != NULL) { 566 567 /* 568 * A certain portion of the list is 569 * "locked" to preserve ordering, so 570 * we can only insert after the insert 571 * point. 572 */ 573 bq = bufq->insert_point; 574 } else { 575 576 /* 577 * If we lie before the last removed (currently active) 578 * request, and are not inserting ourselves into the 579 * "locked" portion of the list, then we must add ourselves 580 * to the second request list. 581 */ 582 if (bp->b_pblkno < bufq->last_pblkno) { 583 584 bq = bufq->switch_point; 585 /* 586 * If we are starting a new secondary list, 587 * then it's easy. 588 */ 589 if (bq == NULL) { 590 bufq->switch_point = bp; 591 bufq_insert_tail(bufq, bp); 592 return; 593 } 594 /* 595 * If we lie ahead of the current switch point, 596 * insert us before the switch point and move 597 * the switch point. 598 */ 599 if (bp->b_pblkno < bq->b_pblkno) { 600 bufq->switch_point = bp; 601 TAILQ_INSERT_BEFORE(bq, bp, b_act); 602 return; 603 } 604 } else { 605 if (bufq->switch_point != NULL) 606 be = TAILQ_PREV(bufq->switch_point, 607 buf_queue, b_act); 608 /* 609 * If we lie between last_pblkno and bq, 610 * insert before bq. 611 */ 612 if (bp->b_pblkno < bq->b_pblkno) { 613 TAILQ_INSERT_BEFORE(bq, bp, b_act); 614 return; 615 } 616 } 617 } 618 619 /* 620 * Request is at/after our current position in the list. 621 * Optimize for sequential I/O by seeing if we go at the tail. 622 */ 623 if (bp->b_pblkno > be->b_pblkno) { 624 TAILQ_INSERT_AFTER(&bufq->queue, be, bp, b_act); 625 return; 626 } 627 628 /* Otherwise, insertion sort */ 629 while ((bn = TAILQ_NEXT(bq, b_act)) != NULL) { 630 631 /* 632 * We want to go after the current request if it is the end 633 * of the first request list, or if the next request is a 634 * larger cylinder than our request. 635 */ 636 if (bn == bufq->switch_point 637 || bp->b_pblkno < bn->b_pblkno) 638 break; 639 bq = bn; 640 } 641 TAILQ_INSERT_AFTER(&bufq->queue, bq, bp, b_act); 642 } 643 644 645 /* 646 * Attempt to read a disk label from a device using the indicated strategy 647 * routine. The label must be partly set up before this: secpercyl, secsize 648 * and anything required in the strategy routine (e.g., dummy bounds for the 649 * partition containing the label) must be filled in before calling us. 650 * Returns NULL on success and an error string on failure. 651 */ 652 char * 653 readdisklabel(dev_t dev, struct disklabel *lp) 654 { 655 struct buf *bp; 656 struct disklabel *dlp; 657 char *msg = NULL; 658 659 bp = geteblk((int)lp->d_secsize); 660 bp->b_dev = dev; 661 bp->b_blkno = LABELSECTOR * ((int)lp->d_secsize/DEV_BSIZE); 662 bp->b_bcount = lp->d_secsize; 663 bp->b_flags &= ~B_INVAL; 664 bp->b_flags |= B_READ; 665 BUF_STRATEGY(bp, 1); 666 if (biowait(bp)) 667 msg = "I/O error"; 668 else for (dlp = (struct disklabel *)bp->b_data; 669 dlp <= (struct disklabel *)((char *)bp->b_data + 670 lp->d_secsize - sizeof(*dlp)); 671 dlp = (struct disklabel *)((char *)dlp + sizeof(long))) { 672 if (dlp->d_magic != DISKMAGIC || dlp->d_magic2 != DISKMAGIC) { 673 if (msg == NULL) 674 msg = "no disk label"; 675 } else if (dlp->d_npartitions > MAXPARTITIONS || 676 dkcksum(dlp) != 0) 677 msg = "disk label corrupted"; 678 else { 679 *lp = *dlp; 680 msg = NULL; 681 break; 682 } 683 } 684 bp->b_flags |= B_INVAL | B_AGE; 685 brelse(bp); 686 return (msg); 687 } 688 689 /* 690 * Check new disk label for sensibility before setting it. 691 */ 692 int 693 setdisklabel(struct disklabel *olp, struct disklabel *nlp, u_long openmask) 694 { 695 int i; 696 struct partition *opp, *npp; 697 698 /* 699 * Check it is actually a disklabel we are looking at. 700 */ 701 if (nlp->d_magic != DISKMAGIC || nlp->d_magic2 != DISKMAGIC || 702 dkcksum(nlp) != 0) 703 return (EINVAL); 704 /* 705 * For each partition that we think is open, 706 */ 707 while ((i = ffs((long)openmask)) != 0) { 708 i--; 709 /* 710 * Check it is not changing.... 711 */ 712 openmask &= ~(1 << i); 713 if (nlp->d_npartitions <= i) 714 return (EBUSY); 715 opp = &olp->d_partitions[i]; 716 npp = &nlp->d_partitions[i]; 717 if (npp->p_offset != opp->p_offset || npp->p_size < opp->p_size) 718 return (EBUSY); 719 /* 720 * Copy internally-set partition information 721 * if new label doesn't include it. XXX 722 * (If we are using it then we had better stay the same type) 723 * This is possibly dubious, as someone else noted (XXX) 724 */ 725 if (npp->p_fstype == FS_UNUSED && opp->p_fstype != FS_UNUSED) { 726 npp->p_fstype = opp->p_fstype; 727 npp->p_fsize = opp->p_fsize; 728 npp->p_frag = opp->p_frag; 729 npp->p_cpg = opp->p_cpg; 730 } 731 } 732 nlp->d_checksum = 0; 733 nlp->d_checksum = dkcksum(nlp); 734 *olp = *nlp; 735 return (0); 736 } 737 738 /* 739 * Write disk label back to device after modification. 740 */ 741 int 742 writedisklabel(dev_t dev, struct disklabel *lp) 743 { 744 struct buf *bp; 745 struct disklabel *dlp; 746 int error = 0; 747 748 if (lp->d_partitions[RAW_PART].p_offset != 0) 749 return (EXDEV); /* not quite right */ 750 bp = geteblk((int)lp->d_secsize); 751 bp->b_dev = dkmodpart(dev, RAW_PART); 752 bp->b_blkno = LABELSECTOR * ((int)lp->d_secsize/DEV_BSIZE); 753 bp->b_bcount = lp->d_secsize; 754 #if 1 755 /* 756 * We read the label first to see if it's there, 757 * in which case we will put ours at the same offset into the block.. 758 * (I think this is stupid [Julian]) 759 * Note that you can't write a label out over a corrupted label! 760 * (also stupid.. how do you write the first one? by raw writes?) 761 */ 762 bp->b_flags &= ~B_INVAL; 763 bp->b_flags |= B_READ; 764 BUF_STRATEGY(bp, 1); 765 error = biowait(bp); 766 if (error) 767 goto done; 768 for (dlp = (struct disklabel *)bp->b_data; 769 dlp <= (struct disklabel *) 770 ((char *)bp->b_data + lp->d_secsize - sizeof(*dlp)); 771 dlp = (struct disklabel *)((char *)dlp + sizeof(long))) { 772 if (dlp->d_magic == DISKMAGIC && dlp->d_magic2 == DISKMAGIC && 773 dkcksum(dlp) == 0) { 774 *dlp = *lp; 775 bp->b_flags &= ~(B_DONE | B_READ); 776 bp->b_flags |= B_WRITE; 777 bp->b_dev = dkmodpart(dev, RAW_PART); 778 BUF_STRATEGY(bp, 1); 779 error = biowait(bp); 780 goto done; 781 } 782 } 783 error = ESRCH; 784 done: 785 #else 786 bzero(bp->b_data, lp->d_secsize); 787 dlp = (struct disklabel *)bp->b_data; 788 *dlp = *lp; 789 bp->b_flags &= ~B_INVAL; 790 bp->b_flags |= B_WRITE; 791 BUF_STRATEGY(bp, 1); 792 error = biowait(bp); 793 #endif 794 bp->b_flags |= B_INVAL | B_AGE; 795 brelse(bp); 796 return (error); 797 } 798 799 /* 800 * Disk error is the preface to plaintive error messages 801 * about failing disk transfers. It prints messages of the form 802 803 hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d) 804 805 * if the offset of the error in the transfer and a disk label 806 * are both available. blkdone should be -1 if the position of the error 807 * is unknown; the disklabel pointer may be null from drivers that have not 808 * been converted to use them. The message is printed with printf 809 * if pri is LOG_PRINTF, otherwise it uses log at the specified priority. 810 * The message should be completed (with at least a newline) with printf 811 * or addlog, respectively. There is no trailing space. 812 */ 813 void 814 diskerr(struct buf *bp, dev_t dev, char *what, int pri, 815 int blkdone, struct disklabel *lp) 816 { 817 int unit = dkunit(dev); 818 int slice = dkslice(dev); 819 int part = dkpart(dev); 820 char partname[2]; 821 char *sname; 822 daddr_t sn; 823 824 sname = dsname(dev, unit, slice, part, partname); 825 printf("%s%s: %s %sing fsbn ", sname, partname, what, 826 bp->b_flags & B_READ ? "read" : "writ"); 827 sn = bp->b_blkno; 828 if (bp->b_bcount <= DEV_BSIZE) { 829 printf("%ld", (long)sn); 830 } else { 831 if (blkdone >= 0) { 832 sn += blkdone; 833 printf("%ld of ", (long)sn); 834 } 835 printf("%ld-%ld", (long)bp->b_blkno, 836 (long)(bp->b_blkno + (bp->b_bcount - 1) / DEV_BSIZE)); 837 } 838 if (lp && (blkdone >= 0 || bp->b_bcount <= lp->d_secsize)) { 839 sn += lp->d_partitions[part].p_offset; 840 /* 841 * XXX should add slice offset and not print the slice, 842 * but we don't know the slice pointer. 843 * XXX should print bp->b_pblkno so that this will work 844 * independent of slices, labels and bad sector remapping, 845 * but some drivers don't set bp->b_pblkno. 846 */ 847 printf(" (%s bn %ld; cn %ld", sname, (long)sn, 848 (long)(sn / lp->d_secpercyl)); 849 sn %= (long)lp->d_secpercyl; 850 printf(" tn %ld sn %ld)", (long)(sn / lp->d_nsectors), 851 (long)(sn % lp->d_nsectors)); 852 } 853 } 854