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 1.16 2005/08/07 03:17:37 hmp 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(), 180 dkmakeunit(dkunit(disk->d_cdev))); 181 LIST_REMOVE(disk, d_list); 182 } 183 if (disk->d_rawsw) { 184 destroy_all_dev(disk->d_rawsw, dkunitmask(), 185 dkmakeunit(dkunit(disk->d_rawdev))); 186 } 187 bzero(disk, sizeof(*disk)); 188 } 189 190 int 191 disk_dumpcheck(dev_t dev, u_int *count, u_int *blkno, u_int *secsize) 192 { 193 struct disk *dp; 194 struct disklabel *dl; 195 u_int boff; 196 197 dp = dev->si_disk; 198 if (!dp) 199 return (ENXIO); 200 if (!dp->d_slice) 201 return (ENXIO); 202 dl = dsgetlabel(dev, dp->d_slice); 203 if (!dl) 204 return (ENXIO); 205 *count = Maxmem * (PAGE_SIZE / dl->d_secsize); 206 if (dumplo <= LABELSECTOR || 207 (dumplo + *count > dl->d_partitions[dkpart(dev)].p_size)) 208 return (EINVAL); 209 boff = dl->d_partitions[dkpart(dev)].p_offset + 210 dp->d_slice->dss_slices[dkslice(dev)].ds_offset; 211 *blkno = boff + dumplo; 212 *secsize = dl->d_secsize; 213 return (0); 214 215 } 216 217 void 218 disk_invalidate (struct disk *disk) 219 { 220 if (disk->d_slice) 221 dsgone(&disk->d_slice); 222 } 223 224 struct disk * 225 disk_enumerate(struct disk *disk) 226 { 227 if (!disk) 228 return (LIST_FIRST(&disklist)); 229 else 230 return (LIST_NEXT(disk, d_list)); 231 } 232 233 static 234 int 235 sysctl_disks(SYSCTL_HANDLER_ARGS) 236 { 237 struct disk *disk; 238 int error, first; 239 240 disk = NULL; 241 first = 1; 242 243 while ((disk = disk_enumerate(disk))) { 244 if (!first) { 245 error = SYSCTL_OUT(req, " ", 1); 246 if (error) 247 return error; 248 } else { 249 first = 0; 250 } 251 error = SYSCTL_OUT(req, disk->d_rawdev->si_name, strlen(disk->d_rawdev->si_name)); 252 if (error) 253 return error; 254 } 255 error = SYSCTL_OUT(req, "", 1); 256 return error; 257 } 258 259 SYSCTL_PROC(_kern, OID_AUTO, disks, CTLTYPE_STRING | CTLFLAG_RD, 0, NULL, 260 sysctl_disks, "A", "names of available disks"); 261 262 /* 263 * The port intercept functions 264 */ 265 static 266 int 267 disk_putport(lwkt_port_t port, lwkt_msg_t lmsg) 268 { 269 struct disk *disk = (struct disk *)port; 270 cdevallmsg_t msg = (cdevallmsg_t)lmsg; 271 int error; 272 273 switch(msg->am_lmsg.ms_cmd.cm_op) { 274 case CDEV_CMD_OPEN: 275 error = diskopen( 276 msg->am_open.msg.dev, 277 msg->am_open.oflags, 278 msg->am_open.devtype, 279 msg->am_open.td); 280 break; 281 case CDEV_CMD_CLOSE: 282 error = diskclose( 283 msg->am_close.msg.dev, 284 msg->am_close.fflag, 285 msg->am_close.devtype, 286 msg->am_close.td); 287 break; 288 case CDEV_CMD_IOCTL: 289 error = diskioctl( 290 msg->am_ioctl.msg.dev, 291 msg->am_ioctl.cmd, 292 msg->am_ioctl.data, 293 msg->am_ioctl.fflag, 294 msg->am_ioctl.td); 295 break; 296 case CDEV_CMD_STRATEGY: 297 diskstrategy(msg->am_strategy.bp); 298 error = 0; 299 break; 300 case CDEV_CMD_PSIZE: 301 msg->am_psize.result = diskpsize(msg->am_psize.msg.dev); 302 error = 0; /* XXX */ 303 break; 304 case CDEV_CMD_READ: 305 error = physio(msg->am_read.msg.dev, 306 msg->am_read.uio, msg->am_read.ioflag); 307 break; 308 case CDEV_CMD_WRITE: 309 error = physio(msg->am_write.msg.dev, 310 msg->am_write.uio, msg->am_write.ioflag); 311 break; 312 case CDEV_CMD_POLL: 313 case CDEV_CMD_KQFILTER: 314 error = ENODEV; 315 case CDEV_CMD_MMAP: 316 error = -1; 317 break; 318 case CDEV_CMD_DUMP: 319 error = disk_dumpcheck(msg->am_dump.msg.dev, 320 &msg->am_dump.count, 321 &msg->am_dump.blkno, 322 &msg->am_dump.secsize); 323 if (error == 0) { 324 msg->am_dump.msg.dev = disk->d_rawdev; 325 error = lwkt_forwardmsg(disk->d_rawdev->si_port, 326 &msg->am_dump.msg.msg); 327 printf("error2 %d\n", error); 328 } 329 break; 330 default: 331 error = ENOTSUP; 332 break; 333 } 334 return(error); 335 } 336 337 /* 338 * When new device entries are instantiated, make sure they inherit our 339 * si_disk structure and block and iosize limits from the raw device. 340 * 341 * This routine is always called synchronously in the context of the 342 * client. 343 * 344 * XXX The various io and block size constraints are not always initialized 345 * properly by devices. 346 */ 347 static 348 int 349 diskclone(dev_t dev) 350 { 351 struct disk *dp; 352 353 dp = dev->si_devsw->d_data; 354 KKASSERT(dp != NULL); 355 dev->si_disk = dp; 356 dev->si_iosize_max = dp->d_rawdev->si_iosize_max; 357 dev->si_bsize_phys = dp->d_rawdev->si_bsize_phys; 358 dev->si_bsize_best = dp->d_rawdev->si_bsize_best; 359 return(0); 360 } 361 362 /* 363 * Open a disk device or partition. 364 */ 365 static 366 int 367 diskopen(dev_t dev, int oflags, int devtype, struct thread *td) 368 { 369 struct disk *dp; 370 int error; 371 372 /* 373 * dp can't be NULL here XXX. 374 */ 375 error = 0; 376 dp = dev->si_disk; 377 if (dp == NULL) 378 return (ENXIO); 379 380 /* 381 * Deal with open races 382 */ 383 while (dp->d_flags & DISKFLAG_LOCK) { 384 dp->d_flags |= DISKFLAG_WANTED; 385 error = tsleep(dp, PCATCH, "diskopen", hz); 386 if (error) 387 return (error); 388 } 389 dp->d_flags |= DISKFLAG_LOCK; 390 391 /* 392 * Open the underlying raw device. 393 */ 394 if (!dsisopen(dp->d_slice)) { 395 #if 0 396 if (!pdev->si_iosize_max) 397 pdev->si_iosize_max = dev->si_iosize_max; 398 #endif 399 error = dev_dopen(dp->d_rawdev, oflags, devtype, td); 400 } 401 402 /* 403 * Inherit properties from the underlying device now that it is 404 * open. 405 */ 406 diskclone(dev); 407 408 if (error) 409 goto out; 410 411 error = dsopen(dev, devtype, dp->d_dsflags, &dp->d_slice, &dp->d_label); 412 413 if (!dsisopen(dp->d_slice)) 414 dev_dclose(dp->d_rawdev, oflags, devtype, td); 415 out: 416 dp->d_flags &= ~DISKFLAG_LOCK; 417 if (dp->d_flags & DISKFLAG_WANTED) { 418 dp->d_flags &= ~DISKFLAG_WANTED; 419 wakeup(dp); 420 } 421 422 return(error); 423 } 424 425 /* 426 * Close a disk device or partition 427 */ 428 static 429 int 430 diskclose(dev_t dev, int fflag, int devtype, struct thread *td) 431 { 432 struct disk *dp; 433 int error; 434 435 error = 0; 436 dp = dev->si_disk; 437 438 dsclose(dev, devtype, dp->d_slice); 439 if (!dsisopen(dp->d_slice)) 440 error = dev_dclose(dp->d_rawdev, fflag, devtype, td); 441 return (error); 442 } 443 444 /* 445 * Execute strategy routine 446 */ 447 static 448 void 449 diskstrategy(struct buf *bp) 450 { 451 struct disk *dp; 452 453 dp = bp->b_dev->si_disk; 454 455 if (dp == NULL) { 456 bp->b_error = ENXIO; 457 bp->b_flags |= B_ERROR; 458 biodone(bp); 459 return; 460 } 461 KKASSERT(bp->b_dev->si_disk == dp); 462 463 if (dscheck(bp, dp->d_slice) <= 0) { 464 biodone(bp); 465 return; 466 } 467 bp->b_dev = dp->d_rawdev; 468 dev_dstrategy(dp->d_rawdev, bp); 469 } 470 471 /* 472 * First execute the ioctl on the disk device, and if it isn't supported 473 * try running it on the backing device. 474 */ 475 static 476 int 477 diskioctl(dev_t dev, u_long cmd, caddr_t data, int fflag, struct thread *td) 478 { 479 struct disk *dp; 480 int error; 481 482 dp = dev->si_disk; 483 if (dp == NULL) 484 return (ENXIO); 485 486 error = dsioctl(dev, cmd, data, fflag, &dp->d_slice); 487 if (error == ENOIOCTL) 488 error = dev_dioctl(dp->d_rawdev, cmd, data, fflag, td); 489 return (error); 490 } 491 492 /* 493 * 494 */ 495 static 496 int 497 diskpsize(dev_t dev) 498 { 499 struct disk *dp; 500 501 dp = dev->si_disk; 502 if (dp == NULL) 503 return (-1); 504 return(dssize(dev, &dp->d_slice)); 505 #if 0 506 if (dp != dev->si_disk) { 507 dev->si_drv1 = pdev->si_drv1; 508 dev->si_drv2 = pdev->si_drv2; 509 /* XXX: don't set bp->b_dev->si_disk (?) */ 510 } 511 #endif 512 } 513 514 SYSCTL_INT(_debug_sizeof, OID_AUTO, disklabel, CTLFLAG_RD, 515 0, sizeof(struct disklabel), "sizeof(struct disklabel)"); 516 517 SYSCTL_INT(_debug_sizeof, OID_AUTO, diskslices, CTLFLAG_RD, 518 0, sizeof(struct diskslices), "sizeof(struct diskslices)"); 519 520 SYSCTL_INT(_debug_sizeof, OID_AUTO, disk, CTLFLAG_RD, 521 0, sizeof(struct disk), "sizeof(struct disk)"); 522 523 524 /* 525 * Seek sort for disks. 526 * 527 * The buf_queue keep two queues, sorted in ascending block order. The first 528 * queue holds those requests which are positioned after the current block 529 * (in the first request); the second, which starts at queue->switch_point, 530 * holds requests which came in after their block number was passed. Thus 531 * we implement a one way scan, retracting after reaching the end of the drive 532 * to the first request on the second queue, at which time it becomes the 533 * first queue. 534 * 535 * A one-way scan is natural because of the way UNIX read-ahead blocks are 536 * allocated. 537 */ 538 void 539 bufqdisksort(struct buf_queue_head *bufq, struct buf *bp) 540 { 541 struct buf *bq; 542 struct buf *bn; 543 struct buf *be; 544 545 be = TAILQ_LAST(&bufq->queue, buf_queue); 546 /* 547 * If the queue is empty or we are an 548 * ordered transaction, then it's easy. 549 */ 550 if ((bq = bufq_first(bufq)) == NULL || 551 (bp->b_flags & B_ORDERED) != 0) { 552 bufq_insert_tail(bufq, bp); 553 return; 554 } else if (bufq->insert_point != NULL) { 555 556 /* 557 * A certain portion of the list is 558 * "locked" to preserve ordering, so 559 * we can only insert after the insert 560 * point. 561 */ 562 bq = bufq->insert_point; 563 } else { 564 565 /* 566 * If we lie before the last removed (currently active) 567 * request, and are not inserting ourselves into the 568 * "locked" portion of the list, then we must add ourselves 569 * to the second request list. 570 */ 571 if (bp->b_pblkno < bufq->last_pblkno) { 572 573 bq = bufq->switch_point; 574 /* 575 * If we are starting a new secondary list, 576 * then it's easy. 577 */ 578 if (bq == NULL) { 579 bufq->switch_point = bp; 580 bufq_insert_tail(bufq, bp); 581 return; 582 } 583 /* 584 * If we lie ahead of the current switch point, 585 * insert us before the switch point and move 586 * the switch point. 587 */ 588 if (bp->b_pblkno < bq->b_pblkno) { 589 bufq->switch_point = bp; 590 TAILQ_INSERT_BEFORE(bq, bp, b_act); 591 return; 592 } 593 } else { 594 if (bufq->switch_point != NULL) 595 be = TAILQ_PREV(bufq->switch_point, 596 buf_queue, b_act); 597 /* 598 * If we lie between last_pblkno and bq, 599 * insert before bq. 600 */ 601 if (bp->b_pblkno < bq->b_pblkno) { 602 TAILQ_INSERT_BEFORE(bq, bp, b_act); 603 return; 604 } 605 } 606 } 607 608 /* 609 * Request is at/after our current position in the list. 610 * Optimize for sequential I/O by seeing if we go at the tail. 611 */ 612 if (bp->b_pblkno > be->b_pblkno) { 613 TAILQ_INSERT_AFTER(&bufq->queue, be, bp, b_act); 614 return; 615 } 616 617 /* Otherwise, insertion sort */ 618 while ((bn = TAILQ_NEXT(bq, b_act)) != NULL) { 619 620 /* 621 * We want to go after the current request if it is the end 622 * of the first request list, or if the next request is a 623 * larger cylinder than our request. 624 */ 625 if (bn == bufq->switch_point 626 || bp->b_pblkno < bn->b_pblkno) 627 break; 628 bq = bn; 629 } 630 TAILQ_INSERT_AFTER(&bufq->queue, bq, bp, b_act); 631 } 632 633 634 /* 635 * Attempt to read a disk label from a device using the indicated strategy 636 * routine. The label must be partly set up before this: secpercyl, secsize 637 * and anything required in the strategy routine (e.g., dummy bounds for the 638 * partition containing the label) must be filled in before calling us. 639 * Returns NULL on success and an error string on failure. 640 */ 641 char * 642 readdisklabel(dev_t dev, struct disklabel *lp) 643 { 644 struct buf *bp; 645 struct disklabel *dlp; 646 char *msg = NULL; 647 648 bp = geteblk((int)lp->d_secsize); 649 bp->b_dev = dev; 650 bp->b_blkno = LABELSECTOR * ((int)lp->d_secsize/DEV_BSIZE); 651 bp->b_bcount = lp->d_secsize; 652 bp->b_flags &= ~B_INVAL; 653 bp->b_flags |= B_READ; 654 BUF_STRATEGY(bp, 1); 655 if (biowait(bp)) 656 msg = "I/O error"; 657 else for (dlp = (struct disklabel *)bp->b_data; 658 dlp <= (struct disklabel *)((char *)bp->b_data + 659 lp->d_secsize - sizeof(*dlp)); 660 dlp = (struct disklabel *)((char *)dlp + sizeof(long))) { 661 if (dlp->d_magic != DISKMAGIC || dlp->d_magic2 != DISKMAGIC) { 662 if (msg == NULL) 663 msg = "no disk label"; 664 } else if (dlp->d_npartitions > MAXPARTITIONS || 665 dkcksum(dlp) != 0) 666 msg = "disk label corrupted"; 667 else { 668 *lp = *dlp; 669 msg = NULL; 670 break; 671 } 672 } 673 bp->b_flags |= B_INVAL | B_AGE; 674 brelse(bp); 675 return (msg); 676 } 677 678 /* 679 * Check new disk label for sensibility before setting it. 680 */ 681 int 682 setdisklabel(struct disklabel *olp, struct disklabel *nlp, u_long openmask) 683 { 684 int i; 685 struct partition *opp, *npp; 686 687 /* 688 * Check it is actually a disklabel we are looking at. 689 */ 690 if (nlp->d_magic != DISKMAGIC || nlp->d_magic2 != DISKMAGIC || 691 dkcksum(nlp) != 0) 692 return (EINVAL); 693 /* 694 * For each partition that we think is open, 695 */ 696 while ((i = ffs((long)openmask)) != 0) { 697 i--; 698 /* 699 * Check it is not changing.... 700 */ 701 openmask &= ~(1 << i); 702 if (nlp->d_npartitions <= i) 703 return (EBUSY); 704 opp = &olp->d_partitions[i]; 705 npp = &nlp->d_partitions[i]; 706 if (npp->p_offset != opp->p_offset || npp->p_size < opp->p_size) 707 return (EBUSY); 708 /* 709 * Copy internally-set partition information 710 * if new label doesn't include it. XXX 711 * (If we are using it then we had better stay the same type) 712 * This is possibly dubious, as someone else noted (XXX) 713 */ 714 if (npp->p_fstype == FS_UNUSED && opp->p_fstype != FS_UNUSED) { 715 npp->p_fstype = opp->p_fstype; 716 npp->p_fsize = opp->p_fsize; 717 npp->p_frag = opp->p_frag; 718 npp->p_cpg = opp->p_cpg; 719 } 720 } 721 nlp->d_checksum = 0; 722 nlp->d_checksum = dkcksum(nlp); 723 *olp = *nlp; 724 return (0); 725 } 726 727 /* 728 * Write disk label back to device after modification. 729 */ 730 int 731 writedisklabel(dev_t dev, struct disklabel *lp) 732 { 733 struct buf *bp; 734 struct disklabel *dlp; 735 int error = 0; 736 737 if (lp->d_partitions[RAW_PART].p_offset != 0) 738 return (EXDEV); /* not quite right */ 739 bp = geteblk((int)lp->d_secsize); 740 bp->b_dev = dkmodpart(dev, RAW_PART); 741 bp->b_blkno = LABELSECTOR * ((int)lp->d_secsize/DEV_BSIZE); 742 bp->b_bcount = lp->d_secsize; 743 #if 1 744 /* 745 * We read the label first to see if it's there, 746 * in which case we will put ours at the same offset into the block.. 747 * (I think this is stupid [Julian]) 748 * Note that you can't write a label out over a corrupted label! 749 * (also stupid.. how do you write the first one? by raw writes?) 750 */ 751 bp->b_flags &= ~B_INVAL; 752 bp->b_flags |= B_READ; 753 BUF_STRATEGY(bp, 1); 754 error = biowait(bp); 755 if (error) 756 goto done; 757 for (dlp = (struct disklabel *)bp->b_data; 758 dlp <= (struct disklabel *) 759 ((char *)bp->b_data + lp->d_secsize - sizeof(*dlp)); 760 dlp = (struct disklabel *)((char *)dlp + sizeof(long))) { 761 if (dlp->d_magic == DISKMAGIC && dlp->d_magic2 == DISKMAGIC && 762 dkcksum(dlp) == 0) { 763 *dlp = *lp; 764 bp->b_flags &= ~(B_DONE | B_READ); 765 bp->b_flags |= B_WRITE; 766 bp->b_dev = dkmodpart(dev, RAW_PART); 767 BUF_STRATEGY(bp, 1); 768 error = biowait(bp); 769 goto done; 770 } 771 } 772 error = ESRCH; 773 done: 774 #else 775 bzero(bp->b_data, lp->d_secsize); 776 dlp = (struct disklabel *)bp->b_data; 777 *dlp = *lp; 778 bp->b_flags &= ~B_INVAL; 779 bp->b_flags |= B_WRITE; 780 BUF_STRATEGY(bp, 1); 781 error = biowait(bp); 782 #endif 783 bp->b_flags |= B_INVAL | B_AGE; 784 brelse(bp); 785 return (error); 786 } 787 788 /* 789 * Disk error is the preface to plaintive error messages 790 * about failing disk transfers. It prints messages of the form 791 792 hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d) 793 794 * if the offset of the error in the transfer and a disk label 795 * are both available. blkdone should be -1 if the position of the error 796 * is unknown; the disklabel pointer may be null from drivers that have not 797 * been converted to use them. The message is printed with printf 798 * if pri is LOG_PRINTF, otherwise it uses log at the specified priority. 799 * The message should be completed (with at least a newline) with printf 800 * or addlog, respectively. There is no trailing space. 801 */ 802 void 803 diskerr(struct buf *bp, dev_t dev, char *what, int pri, 804 int blkdone, struct disklabel *lp) 805 { 806 int unit = dkunit(dev); 807 int slice = dkslice(dev); 808 int part = dkpart(dev); 809 char partname[2]; 810 char *sname; 811 daddr_t sn; 812 813 sname = dsname(dev, unit, slice, part, partname); 814 printf("%s%s: %s %sing fsbn ", sname, partname, what, 815 bp->b_flags & B_READ ? "read" : "writ"); 816 sn = bp->b_blkno; 817 if (bp->b_bcount <= DEV_BSIZE) { 818 printf("%ld", (long)sn); 819 } else { 820 if (blkdone >= 0) { 821 sn += blkdone; 822 printf("%ld of ", (long)sn); 823 } 824 printf("%ld-%ld", (long)bp->b_blkno, 825 (long)(bp->b_blkno + (bp->b_bcount - 1) / DEV_BSIZE)); 826 } 827 if (lp && (blkdone >= 0 || bp->b_bcount <= lp->d_secsize)) { 828 #ifdef tahoe 829 sn *= DEV_BSIZE / lp->d_secsize; /* XXX */ 830 #endif 831 sn += lp->d_partitions[part].p_offset; 832 /* 833 * XXX should add slice offset and not print the slice, 834 * but we don't know the slice pointer. 835 * XXX should print bp->b_pblkno so that this will work 836 * independent of slices, labels and bad sector remapping, 837 * but some drivers don't set bp->b_pblkno. 838 */ 839 printf(" (%s bn %ld; cn %ld", sname, (long)sn, 840 (long)(sn / lp->d_secpercyl)); 841 sn %= (long)lp->d_secpercyl; 842 printf(" tn %ld sn %ld)", (long)(sn / lp->d_nsectors), 843 (long)(sn % lp->d_nsectors)); 844 } 845 } 846