1 /* $NetBSD: subr_disk_mbr.c,v 1.17 2006/01/28 19:57:07 dsl Exp $ */ 2 3 /* 4 * Copyright (c) 1982, 1986, 1988 Regents of the University of California. 5 * All rights reserved. 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 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the University nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 * 31 * @(#)ufs_disksubr.c 7.16 (Berkeley) 5/4/91 32 */ 33 34 /* 35 * Code to find a NetBSD label on a disk that contains an i386 style MBR. 36 * The first NetBSD label found in the 2nd sector of a NetBSD partition 37 * is used. 38 * If we don't find a label searching the MBR, we look at the start of the 39 * disk, if that fails then a label is faked up from the MBR. 40 * 41 * If there isn't a disklabel or anything in the MBR then partition a 42 * is set to cover the whole disk. 43 * Useful for files that contain single filesystems (etc). 44 * 45 * This code will read host endian netbsd labels from little endian MBR. 46 * 47 * Based on the i386 disksubr.c 48 * 49 * Since the mbr only has 32bit fields for sector addresses, we do the same. 50 * 51 * XXX There are potential problems writing labels to disks where there 52 * is only space for 8 netbsd partitions but this code has been compiled 53 * with MAXPARTITIONS=16. 54 */ 55 56 #include <sys/cdefs.h> 57 __KERNEL_RCSID(0, "$NetBSD: subr_disk_mbr.c,v 1.17 2006/01/28 19:57:07 dsl Exp $"); 58 59 #include <sys/param.h> 60 #include <sys/systm.h> 61 #include <sys/buf.h> 62 #include <sys/disklabel.h> 63 #include <sys/disk.h> 64 #include <sys/syslog.h> 65 66 #include "opt_mbr.h" 67 68 typedef struct mbr_partition mbr_partition_t; 69 70 /* 71 * We allocate a buffer 2 sectors large, and look in both.... 72 * That means we find labels written by other ports with different offsets. 73 * LABELSECTOR and LABELOFFSET are only used if the disk doesn't have a label. 74 */ 75 #if LABELSECTOR > 1 || LABELOFFSET > 512 76 #error Invalid LABELSECTOR or LABELOFFSET 77 #endif 78 79 #define MBR_LABELSECTOR 1 80 81 #define SCAN_CONTINUE 0 82 #define SCAN_FOUND 1 83 #define SCAN_ERROR 2 84 85 typedef struct mbr_args { 86 struct disklabel *lp; 87 void (*strat)(struct buf *); 88 struct buf *bp; 89 const char *msg; 90 int error; 91 int written; /* number of times we wrote label */ 92 int found_mbr; /* set if disk has a valid mbr */ 93 uint label_sector; /* where we found the label */ 94 int action; 95 #define READ_LABEL 1 96 #define UPDATE_LABEL 2 97 #define WRITE_LABEL 3 98 } mbr_args_t; 99 100 static int validate_label(mbr_args_t *, uint); 101 static int look_netbsd_part(mbr_args_t *, mbr_partition_t *, int, uint); 102 static int write_netbsd_label(mbr_args_t *, mbr_partition_t *, int, uint); 103 104 static int 105 read_sector(mbr_args_t *a, uint sector, int count) 106 { 107 struct buf *bp = a->bp; 108 int error; 109 110 bp->b_blkno = sector; 111 bp->b_bcount = count * a->lp->d_secsize; 112 bp->b_flags = (bp->b_flags & ~(B_WRITE | B_DONE)) | B_READ; 113 bp->b_cylinder = sector / a->lp->d_secpercyl; 114 (*a->strat)(bp); 115 error = biowait(bp); 116 if (error != 0) 117 a->error = error; 118 return error; 119 } 120 121 /* 122 * Scan MBR for partitions, call 'action' routine for each. 123 */ 124 125 static int 126 scan_mbr(mbr_args_t *a, int (*actn)(mbr_args_t *, mbr_partition_t *, int, uint)) 127 { 128 mbr_partition_t ptns[MBR_PART_COUNT]; 129 mbr_partition_t *dp; 130 struct mbr_sector *mbr; 131 uint ext_base, this_ext, next_ext; 132 int rval; 133 int i; 134 #ifdef COMPAT_386BSD_MBRPART 135 int dp_386bsd = -1; 136 #endif 137 138 ext_base = 0; 139 this_ext = 0; 140 for (;;) { 141 if (read_sector(a, this_ext, 1)) { 142 a->msg = "dos partition I/O error"; 143 return SCAN_ERROR; 144 } 145 146 /* Note: Magic number is little-endian. */ 147 mbr = (void *)a->bp->b_data; 148 if (mbr->mbr_magic != htole16(MBR_MAGIC)) 149 return SCAN_CONTINUE; 150 151 /* Copy data out of buffer so action can use bp */ 152 memcpy(ptns, &mbr->mbr_parts, sizeof ptns); 153 154 /* Look for drivers and skip them */ 155 if (ext_base == 0 && ptns[0].mbrp_type == MBR_PTYPE_DM6_DDO) { 156 /* We've found a DM6 DDO partition type (used by 157 * the Ontrack Disk Manager drivers). 158 * 159 * Ensure that there are no other partitions in the 160 * MBR and jump to the real partition table (stored 161 * in the first sector of the second track). */ 162 boolean_t ok = TRUE; 163 164 for (i = 1; i < MBR_PART_COUNT; i++) 165 if (ptns[i].mbrp_type != MBR_PTYPE_UNUSED) 166 ok = FALSE; 167 168 if (ok) { 169 this_ext = le32toh(a->lp->d_secpercyl / 170 a->lp->d_ntracks); 171 continue; 172 } 173 } 174 175 /* look for NetBSD partition */ 176 next_ext = 0; 177 dp = ptns; 178 for (i = 0; i < MBR_PART_COUNT; i++, dp++) { 179 if (dp->mbrp_type == 0) 180 continue; 181 /* Check end of partition is inside disk limits */ 182 if ((uint64_t)ext_base + le32toh(dp->mbrp_start) + 183 le32toh(dp->mbrp_size) > a->lp->d_secperunit) { 184 /* This mbr doesn't look good.... */ 185 a->msg = "mbr partition exceeds disk size"; 186 /* ...but don't report this as an error (yet) */ 187 return SCAN_CONTINUE; 188 } 189 a->found_mbr = 1; 190 if (MBR_IS_EXTENDED(dp->mbrp_type)) { 191 next_ext = le32toh(dp->mbrp_start); 192 continue; 193 } 194 #ifdef COMPAT_386BSD_MBRPART 195 if (dp->mbrp_type == MBR_PTYPE_386BSD) { 196 /* 197 * If more than one matches, take last, 198 * as NetBSD install tool does. 199 */ 200 if (this_ext == 0) 201 dp_386bsd = i; 202 continue; 203 } 204 #endif 205 rval = (*actn)(a, dp, i, this_ext); 206 if (rval != SCAN_CONTINUE) 207 return rval; 208 } 209 if (next_ext == 0) 210 break; 211 if (ext_base == 0) { 212 ext_base = next_ext; 213 next_ext = 0; 214 } 215 next_ext += ext_base; 216 if (next_ext <= this_ext) 217 break; 218 this_ext = next_ext; 219 } 220 #ifdef COMPAT_386BSD_MBRPART 221 if (this_ext == 0 && dp_386bsd != -1) 222 return (*actn)(a, &ptns[dp_386bsd], dp_386bsd, 0); 223 #endif 224 return SCAN_CONTINUE; 225 } 226 227 /* 228 * Attempt to read a disk label from a device 229 * using the indicated strategy routine. 230 * The label must be partly set up before this: 231 * secpercyl, secsize and anything required for a block i/o read 232 * operation in the driver's strategy/start routines 233 * must be filled in before calling us. 234 * 235 * If dos partition table requested, attempt to load it and 236 * find disklabel inside a DOS partition. Also, if bad block 237 * table needed, attempt to extract it as well. Return buffer 238 * for use in signalling errors if requested. 239 * 240 * Returns null on success and an error string on failure. 241 */ 242 const char * 243 readdisklabel(dev_t dev, void (*strat)(struct buf *), struct disklabel *lp, 244 struct cpu_disklabel *osdep) 245 { 246 struct dkbad *bdp; 247 int rval; 248 int i; 249 mbr_args_t a; 250 251 memset(&a, 0, sizeof a); 252 a.lp = lp; 253 a.strat = strat; 254 a.action = READ_LABEL; 255 256 /* minimal requirements for architypal disk label */ 257 if (lp->d_secsize == 0) 258 lp->d_secsize = DEV_BSIZE; 259 if (lp->d_secperunit == 0) 260 lp->d_secperunit = 0x1fffffff; 261 lp->d_npartitions = RAW_PART + 1; 262 for (i = 0; i < RAW_PART; i++) { 263 lp->d_partitions[i].p_size = 0; 264 lp->d_partitions[i].p_offset = 0; 265 } 266 if (lp->d_partitions[RAW_PART].p_size == 0) 267 lp->d_partitions[RAW_PART].p_size = lp->d_secperunit; 268 lp->d_partitions[RAW_PART].p_offset = 0; 269 270 /* 271 * Set partition 'a' to be the whole disk. 272 * Cleared if we find an mbr or a netbsd label. 273 */ 274 lp->d_partitions[0].p_size = lp->d_partitions[RAW_PART].p_size; 275 lp->d_partitions[0].p_fstype = FS_BSDFFS; 276 277 /* get a buffer and initialize it */ 278 a.bp = geteblk(2 * (int)lp->d_secsize); 279 a.bp->b_dev = dev; 280 281 if (osdep) 282 /* 283 * Scan mbr searching for netbsd partition and saving 284 * bios partition information to use if the netbsd one 285 * is absent. 286 */ 287 rval = scan_mbr(&a, look_netbsd_part); 288 else 289 rval = SCAN_CONTINUE; 290 291 if (rval == SCAN_CONTINUE) { 292 /* Look at start of disk */ 293 rval = validate_label(&a, 0); 294 } 295 296 #if 0 297 /* 298 * Save sector where we found the label for the 'don't overwrite 299 * the label' check in bounds_check_with_label. 300 */ 301 if (rval == SCAN_FOUND) 302 xxx->label_sector = a.label_sector; 303 #endif 304 305 /* Obtain bad sector table if requested and present */ 306 if (rval == SCAN_FOUND && osdep && (lp->d_flags & D_BADSECT)) { 307 struct dkbad *db; 308 int blkno; 309 310 bdp = &osdep->bad; 311 i = 0; 312 rval = SCAN_ERROR; 313 do { 314 /* read a bad sector table */ 315 blkno = lp->d_secperunit - lp->d_nsectors + i; 316 if (lp->d_secsize > DEV_BSIZE) 317 blkno *= lp->d_secsize / DEV_BSIZE; 318 else 319 blkno /= DEV_BSIZE / lp->d_secsize; 320 /* if successful, validate, otherwise try another */ 321 if (read_sector(&a, blkno, 1)) { 322 a.msg = "bad sector table I/O error"; 323 continue; 324 } 325 db = (struct dkbad *)(a.bp->b_data); 326 #define DKBAD_MAGIC 0x4321 327 if (db->bt_mbz != 0 || db->bt_flag != DKBAD_MAGIC) { 328 a.msg = "bad sector table corrupted"; 329 continue; 330 } 331 rval = SCAN_FOUND; 332 *bdp = *db; 333 break; 334 } while ((a.bp->b_flags & B_ERROR) && (i += 2) < 10 && 335 i < lp->d_nsectors); 336 } 337 338 brelse(a.bp); 339 if (rval == SCAN_ERROR || rval == SCAN_CONTINUE) 340 return a.msg; 341 return NULL; 342 } 343 344 static int 345 look_netbsd_part(mbr_args_t *a, mbr_partition_t *dp, int slot, uint ext_base) 346 { 347 struct partition *pp; 348 int ptn_base = ext_base + le32toh(dp->mbrp_start); 349 int rval; 350 351 if ( 352 #ifdef COMPAT_386BSD_MBRPART 353 dp->mbrp_type == MBR_PTYPE_386BSD || 354 #endif 355 dp->mbrp_type == MBR_PTYPE_NETBSD) { 356 rval = validate_label(a, ptn_base); 357 358 #if RAW_PART == 3 359 /* Put actual location where we found the label into ptn 2 */ 360 if (rval == SCAN_FOUND || a->lp->d_partitions[2].p_size == 0) { 361 a->lp->d_partitions[2].p_size = le32toh(dp->mbrp_size); 362 a->lp->d_partitions[2].p_offset = ptn_base; 363 } 364 #endif 365 366 /* If we got a netbsd label look no further */ 367 if (rval == SCAN_FOUND) 368 return rval; 369 } 370 371 /* Install main partitions into e..h and extended into i+ */ 372 if (ext_base == 0) 373 slot += 4; 374 else { 375 slot = 4 + MBR_PART_COUNT; 376 pp = &a->lp->d_partitions[slot]; 377 for (; slot < MAXPARTITIONS; pp++, slot++) { 378 /* This gets called twice - avoid duplicates */ 379 if (pp->p_offset == ptn_base && 380 pp->p_size == le32toh(dp->mbrp_size)) 381 break; 382 if (pp->p_size == 0) 383 break; 384 } 385 } 386 387 if (slot < MAXPARTITIONS) { 388 /* Stop 'a' being the entire disk */ 389 a->lp->d_partitions[0].p_size = 0; 390 a->lp->d_partitions[0].p_fstype = 0; 391 392 /* save partition info */ 393 pp = &a->lp->d_partitions[slot]; 394 pp->p_offset = ptn_base; 395 pp->p_size = le32toh(dp->mbrp_size); 396 pp->p_fstype = xlat_mbr_fstype(dp->mbrp_type); 397 398 if (slot >= a->lp->d_npartitions) 399 a->lp->d_npartitions = slot + 1; 400 } 401 402 return SCAN_CONTINUE; 403 } 404 405 406 static int 407 validate_label(mbr_args_t *a, uint label_sector) 408 { 409 struct disklabel *dlp; 410 char *dlp_lim, *dlp_byte; 411 int error; 412 413 /* Next, dig out disk label */ 414 if (read_sector(a, label_sector, 2)) { 415 a->msg = "disk label read failed"; 416 return SCAN_ERROR; 417 } 418 419 /* Locate disk label within block and validate */ 420 /* 421 * XXX (dsl) This search may be a waste of time, a lot of other i386 422 * code assumes the label is at offset LABELOFFSET (=0) in the sector. 423 * 424 * If we want to support disks from other netbsd ports, then the 425 * code should also allow for a shorter label nearer the end of 426 * the disk sector, and (IIRC) labels within 8k of the disk start. 427 */ 428 dlp = (void *)a->bp->b_data; 429 dlp_lim = a->bp->b_data + a->bp->b_bcount - sizeof *dlp; 430 for (;; dlp = (void *)((char *)dlp + sizeof(long))) { 431 if ((char *)dlp > dlp_lim) { 432 if (a->action != WRITE_LABEL) 433 return SCAN_CONTINUE; 434 /* Write at arch. dependant default location */ 435 dlp_byte = a->bp->b_data + LABELOFFSET; 436 if (label_sector) 437 dlp_byte += MBR_LABELSECTOR * a->lp->d_secsize; 438 else 439 dlp_byte += LABELSECTOR * a->lp->d_secsize; 440 dlp = (void *)dlp_byte; 441 break; 442 } 443 if (dlp->d_magic != DISKMAGIC || dlp->d_magic2 != DISKMAGIC) 444 continue; 445 if (dlp->d_npartitions > MAXPARTITIONS || dkcksum(dlp) != 0) { 446 a->msg = "disk label corrupted"; 447 continue; 448 } 449 break; 450 } 451 452 switch (a->action) { 453 case READ_LABEL: 454 *a->lp = *dlp; 455 a->label_sector = label_sector; 456 return SCAN_FOUND; 457 case UPDATE_LABEL: 458 case WRITE_LABEL: 459 *dlp = *a->lp; 460 a->bp->b_flags &= ~(B_READ|B_DONE); 461 a->bp->b_flags |= B_WRITE; 462 (*a->strat)(a->bp); 463 error = biowait(a->bp); 464 if (error != 0) { 465 a->error = error; 466 a->msg = "disk label write failed"; 467 return SCAN_ERROR; 468 } 469 a->written++; 470 /* Write label to all mbr partitions */ 471 return SCAN_CONTINUE; 472 default: 473 return SCAN_ERROR; 474 } 475 } 476 477 /* 478 * Check new disk label for sensibility 479 * before setting it. 480 */ 481 int 482 setdisklabel(struct disklabel *olp, struct disklabel *nlp, u_long openmask, 483 struct cpu_disklabel *osdep) 484 { 485 int i; 486 struct partition *opp, *npp; 487 488 /* sanity clause */ 489 if (nlp->d_secpercyl == 0 || nlp->d_secsize == 0 490 || (nlp->d_secsize % DEV_BSIZE) != 0) 491 return (EINVAL); 492 493 /* special case to allow disklabel to be invalidated */ 494 if (nlp->d_magic == 0xffffffff) { 495 *olp = *nlp; 496 return (0); 497 } 498 499 if (nlp->d_magic != DISKMAGIC || nlp->d_magic2 != DISKMAGIC || 500 dkcksum(nlp) != 0) 501 return (EINVAL); 502 503 /* XXX missing check if other dos partitions will be overwritten */ 504 505 while (openmask != 0) { 506 i = ffs(openmask) - 1; 507 openmask &= ~(1 << i); 508 if (i > nlp->d_npartitions) 509 return (EBUSY); 510 opp = &olp->d_partitions[i]; 511 npp = &nlp->d_partitions[i]; 512 /* 513 * Copy internally-set partition information 514 * if new label doesn't include it. XXX 515 */ 516 if (npp->p_fstype == FS_UNUSED && opp->p_fstype != FS_UNUSED) { 517 *npp = *opp; 518 continue; 519 } 520 if (npp->p_offset != opp->p_offset || npp->p_size < opp->p_size) 521 return (EBUSY); 522 } 523 nlp->d_checksum = 0; 524 nlp->d_checksum = dkcksum(nlp); 525 *olp = *nlp; 526 return (0); 527 } 528 529 530 /* 531 * Write disk label back to device after modification. 532 */ 533 int 534 writedisklabel(dev_t dev, void (*strat)(struct buf *), struct disklabel *lp, 535 struct cpu_disklabel *osdep) 536 { 537 mbr_args_t a; 538 539 memset(&a, 0, sizeof a); 540 a.lp = lp; 541 a.strat = strat; 542 543 /* get a buffer and initialize it */ 544 a.bp = geteblk(2 * (int)lp->d_secsize); 545 a.bp->b_dev = dev; 546 547 /* osdep => we expect an mbr with label in netbsd ptn */ 548 a.action = osdep != NULL ? WRITE_LABEL : UPDATE_LABEL; 549 550 /* Write/update the label to every netbsd mbr partition */ 551 scan_mbr(&a, write_netbsd_label); 552 553 /* Old write the label at the start of the volume on disks that 554 * don't have a valid mbr (always update an existing one) */ 555 a.action = a.found_mbr ? UPDATE_LABEL : WRITE_LABEL; 556 validate_label(&a, 0); 557 558 if (a.written == 0 && a.error == 0) 559 a.error = ESRCH; 560 561 brelse(a.bp); 562 return a.error; 563 } 564 565 static int 566 write_netbsd_label(mbr_args_t *a, mbr_partition_t *dp, int slot, uint ext_base) 567 { 568 int ptn_base = ext_base + le32toh(dp->mbrp_start); 569 570 if (dp->mbrp_type != MBR_PTYPE_NETBSD) 571 return SCAN_CONTINUE; 572 573 return validate_label(a, ptn_base); 574 } 575 576 577 /* 578 * Determine the size of the transfer, and make sure it is 579 * within the boundaries of the partition. Adjust transfer 580 * if needed, and signal errors or early completion. 581 */ 582 int 583 bounds_check_with_label(struct disk *dk, struct buf *bp, int wlabel) 584 { 585 struct disklabel *lp = dk->dk_label; 586 struct partition *p = lp->d_partitions + DISKPART(bp->b_dev); 587 int labelsector = LABELSECTOR; 588 int64_t sz; 589 590 #if RAW_PART == 3 591 labelsector += lp->d_partitions[2].p_offset; 592 #endif 593 594 sz = howmany(bp->b_bcount, lp->d_secsize); 595 596 if (bp->b_blkno + sz > p->p_size) { 597 sz = p->p_size - bp->b_blkno; 598 if (sz == 0) { 599 /* If exactly at end of disk, return EOF. */ 600 bp->b_resid = bp->b_bcount; 601 return (0); 602 } 603 if (sz < 0) { 604 /* If past end of disk, return EINVAL. */ 605 bp->b_error = EINVAL; 606 goto bad; 607 } 608 /* Otherwise, truncate request. */ 609 bp->b_bcount = sz << DEV_BSHIFT; 610 } 611 612 /* Overwriting disk label? */ 613 if (bp->b_blkno + p->p_offset <= labelsector && 614 bp->b_blkno + p->p_offset + sz > labelsector && 615 (bp->b_flags & B_READ) == 0 && !wlabel) { 616 bp->b_error = EROFS; 617 goto bad; 618 } 619 620 /* calculate cylinder for disksort to order transfers with */ 621 bp->b_cylinder = (bp->b_blkno + p->p_offset) / 622 (lp->d_secsize / DEV_BSIZE) / lp->d_secpercyl; 623 return (1); 624 625 bad: 626 bp->b_flags |= B_ERROR; 627 return (-1); 628 } 629