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 1.13 2004/07/16 05:51:10 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_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 raw device (based on rawsw) is returned to the caller, NOT the 123 * slice and unit managed cdev. The caller typically sets various 124 * driver parameters and IO limits on the returned rawdev which we must 125 * inherit when our managed device is opened. 126 */ 127 dev_t 128 disk_create(int unit, struct disk *dp, int flags, struct cdevsw *rawsw) 129 { 130 dev_t rawdev; 131 struct cdevsw *devsw; 132 133 /* 134 * Create the raw backing device 135 */ 136 compile_devsw(rawsw); 137 rawdev = make_dev(rawsw, dkmakeminor(unit, WHOLE_DISK_SLICE, RAW_PART), 138 UID_ROOT, GID_OPERATOR, 0640, 139 "%s%d", rawsw->d_name, unit); 140 141 /* 142 * Initialize our intercept port 143 */ 144 bzero(dp, sizeof(*dp)); 145 lwkt_initport(&dp->d_port, NULL); 146 dp->d_port.mp_putport = disk_putport; 147 dp->d_rawsw = rawsw; 148 149 /* 150 * We install a custom cdevsw rather then the passed cdevsw, 151 * and save our disk structure in d_data so we can get at it easily 152 * without any complex cloning code. 153 */ 154 devsw = cdevsw_add_override(rawdev, dkunitmask(), dkmakeunit(unit)); 155 devsw->d_port = &dp->d_port; 156 devsw->d_data = dp; 157 devsw->d_clone = diskclone; 158 dp->d_devsw = devsw; 159 dp->d_rawdev = rawdev; 160 dp->d_cdev = make_dev(devsw, 161 dkmakeminor(unit, WHOLE_DISK_SLICE, RAW_PART), 162 UID_ROOT, GID_OPERATOR, 0640, 163 "%s%d", devsw->d_name, unit); 164 165 dp->d_dsflags = flags; 166 LIST_INSERT_HEAD(&disklist, dp, d_list); 167 return (dp->d_rawdev); 168 } 169 170 /* 171 * This routine is called when an adapter detaches. The higher level 172 * managed disk device is destroyed while the lower level raw device is 173 * released. 174 */ 175 void 176 disk_destroy(struct disk *disk) 177 { 178 if (disk->d_devsw) { 179 cdevsw_remove(disk->d_devsw, dkunitmask(), dkunit(disk->d_cdev)); 180 LIST_REMOVE(disk, d_list); 181 } 182 if (disk->d_rawsw) 183 destroy_all_dev(disk->d_rawsw, dkunitmask(), dkunit(disk->d_rawdev)); 184 bzero(disk, sizeof(*disk)); 185 } 186 187 int 188 disk_dumpcheck(dev_t dev, u_int *count, u_int *blkno, u_int *secsize) 189 { 190 struct disk *dp; 191 struct disklabel *dl; 192 u_int boff; 193 194 dp = dev->si_disk; 195 if (!dp) 196 return (ENXIO); 197 if (!dp->d_slice) 198 return (ENXIO); 199 dl = dsgetlabel(dev, dp->d_slice); 200 if (!dl) 201 return (ENXIO); 202 *count = Maxmem * (PAGE_SIZE / dl->d_secsize); 203 if (dumplo <= LABELSECTOR || 204 (dumplo + *count > dl->d_partitions[dkpart(dev)].p_size)) 205 return (EINVAL); 206 boff = dl->d_partitions[dkpart(dev)].p_offset + 207 dp->d_slice->dss_slices[dkslice(dev)].ds_offset; 208 *blkno = boff + dumplo; 209 *secsize = dl->d_secsize; 210 return (0); 211 212 } 213 214 void 215 disk_invalidate (struct disk *disk) 216 { 217 if (disk->d_slice) 218 dsgone(&disk->d_slice); 219 } 220 221 struct disk * 222 disk_enumerate(struct disk *disk) 223 { 224 if (!disk) 225 return (LIST_FIRST(&disklist)); 226 else 227 return (LIST_NEXT(disk, d_list)); 228 } 229 230 static 231 int 232 sysctl_disks(SYSCTL_HANDLER_ARGS) 233 { 234 struct disk *disk; 235 int error, first; 236 237 disk = NULL; 238 first = 1; 239 240 while ((disk = disk_enumerate(disk))) { 241 if (!first) { 242 error = SYSCTL_OUT(req, " ", 1); 243 if (error) 244 return error; 245 } else { 246 first = 0; 247 } 248 error = SYSCTL_OUT(req, disk->d_rawdev->si_name, strlen(disk->d_rawdev->si_name)); 249 if (error) 250 return error; 251 } 252 error = SYSCTL_OUT(req, "", 1); 253 return error; 254 } 255 256 SYSCTL_PROC(_kern, OID_AUTO, disks, CTLTYPE_STRING | CTLFLAG_RD, 0, NULL, 257 sysctl_disks, "A", "names of available disks"); 258 259 /* 260 * The port intercept functions 261 */ 262 static 263 int 264 disk_putport(lwkt_port_t port, lwkt_msg_t lmsg) 265 { 266 struct disk *disk = (struct disk *)port; 267 cdevallmsg_t msg = (cdevallmsg_t)lmsg; 268 int error; 269 270 switch(msg->am_lmsg.ms_cmd.cm_op) { 271 case CDEV_CMD_OPEN: 272 error = diskopen( 273 msg->am_open.msg.dev, 274 msg->am_open.oflags, 275 msg->am_open.devtype, 276 msg->am_open.td); 277 break; 278 case CDEV_CMD_CLOSE: 279 error = diskclose( 280 msg->am_close.msg.dev, 281 msg->am_close.fflag, 282 msg->am_close.devtype, 283 msg->am_close.td); 284 break; 285 case CDEV_CMD_IOCTL: 286 error = diskioctl( 287 msg->am_ioctl.msg.dev, 288 msg->am_ioctl.cmd, 289 msg->am_ioctl.data, 290 msg->am_ioctl.fflag, 291 msg->am_ioctl.td); 292 break; 293 case CDEV_CMD_STRATEGY: 294 diskstrategy(msg->am_strategy.bp); 295 error = 0; 296 break; 297 case CDEV_CMD_PSIZE: 298 msg->am_psize.result = diskpsize(msg->am_psize.msg.dev); 299 error = 0; /* XXX */ 300 break; 301 case CDEV_CMD_READ: 302 error = physio(msg->am_read.msg.dev, 303 msg->am_read.uio, msg->am_read.ioflag); 304 break; 305 case CDEV_CMD_WRITE: 306 error = physio(msg->am_write.msg.dev, 307 msg->am_write.uio, msg->am_write.ioflag); 308 break; 309 case CDEV_CMD_POLL: 310 case CDEV_CMD_KQFILTER: 311 error = ENODEV; 312 case CDEV_CMD_MMAP: 313 error = -1; 314 break; 315 case CDEV_CMD_DUMP: 316 error = disk_dumpcheck(msg->am_dump.msg.dev, 317 &msg->am_dump.count, 318 &msg->am_dump.blkno, 319 &msg->am_dump.secsize); 320 if (error == 0) { 321 msg->am_dump.msg.dev = disk->d_rawdev; 322 error = lwkt_forwardmsg(disk->d_rawdev->si_port, 323 &msg->am_dump.msg.msg); 324 printf("error2 %d\n", error); 325 } 326 break; 327 default: 328 error = ENOTSUP; 329 break; 330 } 331 return(error); 332 } 333 334 /* 335 * When new device entries are instantiated, make sure they inherit our 336 * si_disk structure and block and iosize limits from the raw device. 337 * 338 * This routine is always called synchronously in the context of the 339 * client. 340 * 341 * XXX The various io and block size constraints are not always initialized 342 * properly by devices. 343 */ 344 static 345 int 346 diskclone(dev_t dev) 347 { 348 struct disk *dp; 349 350 dp = dev->si_devsw->d_data; 351 KKASSERT(dp != NULL); 352 dev->si_disk = dp; 353 dev->si_iosize_max = dp->d_rawdev->si_iosize_max; 354 dev->si_bsize_phys = dp->d_rawdev->si_bsize_phys; 355 dev->si_bsize_best = dp->d_rawdev->si_bsize_best; 356 return(0); 357 } 358 359 /* 360 * Open a disk device or partition. 361 */ 362 static 363 int 364 diskopen(dev_t dev, int oflags, int devtype, struct thread *td) 365 { 366 struct disk *dp; 367 int error; 368 369 /* 370 * dp can't be NULL here XXX. 371 */ 372 error = 0; 373 dp = dev->si_disk; 374 if (dp == NULL) 375 return (ENXIO); 376 377 /* 378 * Deal with open races 379 */ 380 while (dp->d_flags & DISKFLAG_LOCK) { 381 dp->d_flags |= DISKFLAG_WANTED; 382 error = tsleep(dp, PCATCH, "diskopen", hz); 383 if (error) 384 return (error); 385 } 386 dp->d_flags |= DISKFLAG_LOCK; 387 388 /* 389 * Open the underlying raw device. 390 */ 391 if (!dsisopen(dp->d_slice)) { 392 #if 0 393 if (!pdev->si_iosize_max) 394 pdev->si_iosize_max = dev->si_iosize_max; 395 #endif 396 error = dev_dopen(dp->d_rawdev, oflags, devtype, td); 397 } 398 399 /* 400 * Inherit properties from the underlying device now that it is 401 * open. 402 */ 403 diskclone(dev); 404 405 if (error) 406 goto out; 407 408 error = dsopen(dev, devtype, dp->d_dsflags, &dp->d_slice, &dp->d_label); 409 410 if (!dsisopen(dp->d_slice)) 411 dev_dclose(dp->d_rawdev, oflags, devtype, td); 412 out: 413 dp->d_flags &= ~DISKFLAG_LOCK; 414 if (dp->d_flags & DISKFLAG_WANTED) { 415 dp->d_flags &= ~DISKFLAG_WANTED; 416 wakeup(dp); 417 } 418 419 return(error); 420 } 421 422 /* 423 * Close a disk device or partition 424 */ 425 static 426 int 427 diskclose(dev_t dev, int fflag, int devtype, struct thread *td) 428 { 429 struct disk *dp; 430 int error; 431 432 error = 0; 433 dp = dev->si_disk; 434 435 dsclose(dev, devtype, dp->d_slice); 436 if (!dsisopen(dp->d_slice)) 437 error = dev_dclose(dp->d_rawdev, fflag, devtype, td); 438 return (error); 439 } 440 441 /* 442 * Execute strategy routine 443 */ 444 static 445 void 446 diskstrategy(struct buf *bp) 447 { 448 struct disk *dp; 449 450 dp = bp->b_dev->si_disk; 451 452 if (dp == NULL) { 453 bp->b_error = ENXIO; 454 bp->b_flags |= B_ERROR; 455 biodone(bp); 456 return; 457 } 458 KKASSERT(bp->b_dev->si_disk == dp); 459 460 if (dscheck(bp, dp->d_slice) <= 0) { 461 biodone(bp); 462 return; 463 } 464 bp->b_dev = dp->d_rawdev; 465 dev_dstrategy(dp->d_rawdev, bp); 466 } 467 468 /* 469 * First execute the ioctl on the disk device, and if it isn't supported 470 * try running it on the backing device. 471 */ 472 static 473 int 474 diskioctl(dev_t dev, u_long cmd, caddr_t data, int fflag, struct thread *td) 475 { 476 struct disk *dp; 477 int error; 478 479 dp = dev->si_disk; 480 if (dp == NULL) 481 return (ENXIO); 482 483 error = dsioctl(dev, cmd, data, fflag, &dp->d_slice); 484 if (error == ENOIOCTL) 485 error = dev_dioctl(dp->d_rawdev, cmd, data, fflag, td); 486 return (error); 487 } 488 489 /* 490 * 491 */ 492 static 493 int 494 diskpsize(dev_t dev) 495 { 496 struct disk *dp; 497 498 dp = dev->si_disk; 499 if (dp == NULL) 500 return (-1); 501 return(dssize(dev, &dp->d_slice)); 502 #if 0 503 if (dp != dev->si_disk) { 504 dev->si_drv1 = pdev->si_drv1; 505 dev->si_drv2 = pdev->si_drv2; 506 /* XXX: don't set bp->b_dev->si_disk (?) */ 507 } 508 #endif 509 } 510 511 SYSCTL_DECL(_debug_sizeof); 512 513 SYSCTL_INT(_debug_sizeof, OID_AUTO, disklabel, CTLFLAG_RD, 514 0, sizeof(struct disklabel), "sizeof(struct disklabel)"); 515 516 SYSCTL_INT(_debug_sizeof, OID_AUTO, diskslices, CTLFLAG_RD, 517 0, sizeof(struct diskslices), "sizeof(struct diskslices)"); 518 519 SYSCTL_INT(_debug_sizeof, OID_AUTO, disk, CTLFLAG_RD, 520 0, sizeof(struct disk), "sizeof(struct disk)"); 521 522 523 /* 524 * Seek sort for disks. 525 * 526 * The buf_queue keep two queues, sorted in ascending block order. The first 527 * queue holds those requests which are positioned after the current block 528 * (in the first request); the second, which starts at queue->switch_point, 529 * holds requests which came in after their block number was passed. Thus 530 * we implement a one way scan, retracting after reaching the end of the drive 531 * to the first request on the second queue, at which time it becomes the 532 * first queue. 533 * 534 * A one-way scan is natural because of the way UNIX read-ahead blocks are 535 * allocated. 536 */ 537 void 538 bufqdisksort(struct buf_queue_head *bufq, struct buf *bp) 539 { 540 struct buf *bq; 541 struct buf *bn; 542 struct buf *be; 543 544 be = TAILQ_LAST(&bufq->queue, buf_queue); 545 /* 546 * If the queue is empty or we are an 547 * ordered transaction, then it's easy. 548 */ 549 if ((bq = bufq_first(bufq)) == NULL || 550 (bp->b_flags & B_ORDERED) != 0) { 551 bufq_insert_tail(bufq, bp); 552 return; 553 } else if (bufq->insert_point != NULL) { 554 555 /* 556 * A certain portion of the list is 557 * "locked" to preserve ordering, so 558 * we can only insert after the insert 559 * point. 560 */ 561 bq = bufq->insert_point; 562 } else { 563 564 /* 565 * If we lie before the last removed (currently active) 566 * request, and are not inserting ourselves into the 567 * "locked" portion of the list, then we must add ourselves 568 * to the second request list. 569 */ 570 if (bp->b_pblkno < bufq->last_pblkno) { 571 572 bq = bufq->switch_point; 573 /* 574 * If we are starting a new secondary list, 575 * then it's easy. 576 */ 577 if (bq == NULL) { 578 bufq->switch_point = bp; 579 bufq_insert_tail(bufq, bp); 580 return; 581 } 582 /* 583 * If we lie ahead of the current switch point, 584 * insert us before the switch point and move 585 * the switch point. 586 */ 587 if (bp->b_pblkno < bq->b_pblkno) { 588 bufq->switch_point = bp; 589 TAILQ_INSERT_BEFORE(bq, bp, b_act); 590 return; 591 } 592 } else { 593 if (bufq->switch_point != NULL) 594 be = TAILQ_PREV(bufq->switch_point, 595 buf_queue, b_act); 596 /* 597 * If we lie between last_pblkno and bq, 598 * insert before bq. 599 */ 600 if (bp->b_pblkno < bq->b_pblkno) { 601 TAILQ_INSERT_BEFORE(bq, bp, b_act); 602 return; 603 } 604 } 605 } 606 607 /* 608 * Request is at/after our current position in the list. 609 * Optimize for sequential I/O by seeing if we go at the tail. 610 */ 611 if (bp->b_pblkno > be->b_pblkno) { 612 TAILQ_INSERT_AFTER(&bufq->queue, be, bp, b_act); 613 return; 614 } 615 616 /* Otherwise, insertion sort */ 617 while ((bn = TAILQ_NEXT(bq, b_act)) != NULL) { 618 619 /* 620 * We want to go after the current request if it is the end 621 * of the first request list, or if the next request is a 622 * larger cylinder than our request. 623 */ 624 if (bn == bufq->switch_point 625 || bp->b_pblkno < bn->b_pblkno) 626 break; 627 bq = bn; 628 } 629 TAILQ_INSERT_AFTER(&bufq->queue, bq, bp, b_act); 630 } 631 632 633 /* 634 * Attempt to read a disk label from a device using the indicated strategy 635 * routine. The label must be partly set up before this: secpercyl, secsize 636 * and anything required in the strategy routine (e.g., dummy bounds for the 637 * partition containing the label) must be filled in before calling us. 638 * Returns NULL on success and an error string on failure. 639 */ 640 char * 641 readdisklabel(dev_t dev, struct disklabel *lp) 642 { 643 struct buf *bp; 644 struct disklabel *dlp; 645 char *msg = NULL; 646 647 bp = geteblk((int)lp->d_secsize); 648 bp->b_dev = dev; 649 bp->b_blkno = LABELSECTOR * ((int)lp->d_secsize/DEV_BSIZE); 650 bp->b_bcount = lp->d_secsize; 651 bp->b_flags &= ~B_INVAL; 652 bp->b_flags |= B_READ; 653 BUF_STRATEGY(bp, 1); 654 if (biowait(bp)) 655 msg = "I/O error"; 656 else for (dlp = (struct disklabel *)bp->b_data; 657 dlp <= (struct disklabel *)((char *)bp->b_data + 658 lp->d_secsize - sizeof(*dlp)); 659 dlp = (struct disklabel *)((char *)dlp + sizeof(long))) { 660 if (dlp->d_magic != DISKMAGIC || dlp->d_magic2 != DISKMAGIC) { 661 if (msg == NULL) 662 msg = "no disk label"; 663 } else if (dlp->d_npartitions > MAXPARTITIONS || 664 dkcksum(dlp) != 0) 665 msg = "disk label corrupted"; 666 else { 667 *lp = *dlp; 668 msg = NULL; 669 break; 670 } 671 } 672 bp->b_flags |= B_INVAL | B_AGE; 673 brelse(bp); 674 return (msg); 675 } 676 677 /* 678 * Check new disk label for sensibility before setting it. 679 */ 680 int 681 setdisklabel(struct disklabel *olp, struct disklabel *nlp, u_long openmask) 682 { 683 int i; 684 struct partition *opp, *npp; 685 686 /* 687 * Check it is actually a disklabel we are looking at. 688 */ 689 if (nlp->d_magic != DISKMAGIC || nlp->d_magic2 != DISKMAGIC || 690 dkcksum(nlp) != 0) 691 return (EINVAL); 692 /* 693 * For each partition that we think is open, 694 */ 695 while ((i = ffs((long)openmask)) != 0) { 696 i--; 697 /* 698 * Check it is not changing.... 699 */ 700 openmask &= ~(1 << i); 701 if (nlp->d_npartitions <= i) 702 return (EBUSY); 703 opp = &olp->d_partitions[i]; 704 npp = &nlp->d_partitions[i]; 705 if (npp->p_offset != opp->p_offset || npp->p_size < opp->p_size) 706 return (EBUSY); 707 /* 708 * Copy internally-set partition information 709 * if new label doesn't include it. XXX 710 * (If we are using it then we had better stay the same type) 711 * This is possibly dubious, as someone else noted (XXX) 712 */ 713 if (npp->p_fstype == FS_UNUSED && opp->p_fstype != FS_UNUSED) { 714 npp->p_fstype = opp->p_fstype; 715 npp->p_fsize = opp->p_fsize; 716 npp->p_frag = opp->p_frag; 717 npp->p_cpg = opp->p_cpg; 718 } 719 } 720 nlp->d_checksum = 0; 721 nlp->d_checksum = dkcksum(nlp); 722 *olp = *nlp; 723 return (0); 724 } 725 726 /* 727 * Write disk label back to device after modification. 728 */ 729 int 730 writedisklabel(dev_t dev, struct disklabel *lp) 731 { 732 struct buf *bp; 733 struct disklabel *dlp; 734 int error = 0; 735 736 if (lp->d_partitions[RAW_PART].p_offset != 0) 737 return (EXDEV); /* not quite right */ 738 bp = geteblk((int)lp->d_secsize); 739 bp->b_dev = dkmodpart(dev, RAW_PART); 740 bp->b_blkno = LABELSECTOR * ((int)lp->d_secsize/DEV_BSIZE); 741 bp->b_bcount = lp->d_secsize; 742 #if 1 743 /* 744 * We read the label first to see if it's there, 745 * in which case we will put ours at the same offset into the block.. 746 * (I think this is stupid [Julian]) 747 * Note that you can't write a label out over a corrupted label! 748 * (also stupid.. how do you write the first one? by raw writes?) 749 */ 750 bp->b_flags &= ~B_INVAL; 751 bp->b_flags |= B_READ; 752 BUF_STRATEGY(bp, 1); 753 error = biowait(bp); 754 if (error) 755 goto done; 756 for (dlp = (struct disklabel *)bp->b_data; 757 dlp <= (struct disklabel *) 758 ((char *)bp->b_data + lp->d_secsize - sizeof(*dlp)); 759 dlp = (struct disklabel *)((char *)dlp + sizeof(long))) { 760 if (dlp->d_magic == DISKMAGIC && dlp->d_magic2 == DISKMAGIC && 761 dkcksum(dlp) == 0) { 762 *dlp = *lp; 763 bp->b_flags &= ~(B_DONE | B_READ); 764 bp->b_flags |= B_WRITE; 765 bp->b_dev = dkmodpart(dev, RAW_PART); 766 #ifdef __alpha__ 767 alpha_fix_srm_checksum(bp); 768 #endif 769 BUF_STRATEGY(bp, 1); 770 error = biowait(bp); 771 goto done; 772 } 773 } 774 error = ESRCH; 775 done: 776 #else 777 bzero(bp->b_data, lp->d_secsize); 778 dlp = (struct disklabel *)bp->b_data; 779 *dlp = *lp; 780 bp->b_flags &= ~B_INVAL; 781 bp->b_flags |= B_WRITE; 782 BUF_STRATEGY(bp, 1); 783 error = biowait(bp); 784 #endif 785 bp->b_flags |= B_INVAL | B_AGE; 786 brelse(bp); 787 return (error); 788 } 789 790 /* 791 * Disk error is the preface to plaintive error messages 792 * about failing disk transfers. It prints messages of the form 793 794 hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d) 795 796 * if the offset of the error in the transfer and a disk label 797 * are both available. blkdone should be -1 if the position of the error 798 * is unknown; the disklabel pointer may be null from drivers that have not 799 * been converted to use them. The message is printed with printf 800 * if pri is LOG_PRINTF, otherwise it uses log at the specified priority. 801 * The message should be completed (with at least a newline) with printf 802 * or addlog, respectively. There is no trailing space. 803 */ 804 void 805 diskerr(struct buf *bp, dev_t dev, char *what, int pri, 806 int blkdone, struct disklabel *lp) 807 { 808 int unit = dkunit(dev); 809 int slice = dkslice(dev); 810 int part = dkpart(dev); 811 char partname[2]; 812 char *sname; 813 daddr_t sn; 814 815 sname = dsname(dev, unit, slice, part, partname); 816 printf("%s%s: %s %sing fsbn ", sname, partname, what, 817 bp->b_flags & B_READ ? "read" : "writ"); 818 sn = bp->b_blkno; 819 if (bp->b_bcount <= DEV_BSIZE) { 820 printf("%ld", (long)sn); 821 } else { 822 if (blkdone >= 0) { 823 sn += blkdone; 824 printf("%ld of ", (long)sn); 825 } 826 printf("%ld-%ld", (long)bp->b_blkno, 827 (long)(bp->b_blkno + (bp->b_bcount - 1) / DEV_BSIZE)); 828 } 829 if (lp && (blkdone >= 0 || bp->b_bcount <= lp->d_secsize)) { 830 #ifdef tahoe 831 sn *= DEV_BSIZE / lp->d_secsize; /* XXX */ 832 #endif 833 sn += lp->d_partitions[part].p_offset; 834 /* 835 * XXX should add slice offset and not print the slice, 836 * but we don't know the slice pointer. 837 * XXX should print bp->b_pblkno so that this will work 838 * independent of slices, labels and bad sector remapping, 839 * but some drivers don't set bp->b_pblkno. 840 */ 841 printf(" (%s bn %ld; cn %ld", sname, (long)sn, 842 (long)(sn / lp->d_secpercyl)); 843 sn %= (long)lp->d_secpercyl; 844 printf(" tn %ld sn %ld)", (long)(sn / lp->d_nsectors), 845 (long)(sn % lp->d_nsectors)); 846 } 847 } 848