1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1980, 1986, 1993 5 * The Regents of the University of California. 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 32 #if 0 33 #ifndef lint 34 static const char sccsid[] = "@(#)pass1.c 8.6 (Berkeley) 4/28/95"; 35 #endif /* not lint */ 36 #endif 37 #include <sys/cdefs.h> 38 __FBSDID("$FreeBSD$"); 39 40 #include <sys/param.h> 41 #include <sys/stat.h> 42 #include <sys/sysctl.h> 43 44 #include <ufs/ufs/dinode.h> 45 #include <ufs/ufs/dir.h> 46 #include <ufs/ffs/fs.h> 47 48 #include <err.h> 49 #include <limits.h> 50 #include <stdint.h> 51 #include <string.h> 52 53 #include "fsck.h" 54 55 static ufs2_daddr_t badblk; 56 static ufs2_daddr_t dupblk; 57 static ino_t lastino; /* last inode in use */ 58 59 static int checkinode(ino_t inumber, struct inodesc *, int rebuildcg); 60 61 void 62 pass1(void) 63 { 64 struct inostat *info; 65 struct inodesc idesc; 66 struct bufarea *cgbp; 67 struct cg *cgp; 68 ino_t inumber, inosused, mininos; 69 ufs2_daddr_t i, cgd; 70 u_int8_t *cp; 71 int c, rebuildcg; 72 73 badblk = dupblk = lastino = 0; 74 75 /* 76 * Set file system reserved blocks in used block map. 77 */ 78 for (c = 0; c < sblock.fs_ncg; c++) { 79 cgd = cgdmin(&sblock, c); 80 if (c == 0) { 81 i = cgbase(&sblock, c); 82 } else 83 i = cgsblock(&sblock, c); 84 for (; i < cgd; i++) 85 setbmap(i); 86 } 87 i = sblock.fs_csaddr; 88 cgd = i + howmany(sblock.fs_cssize, sblock.fs_fsize); 89 for (; i < cgd; i++) 90 setbmap(i); 91 92 /* 93 * Find all allocated blocks. 94 */ 95 memset(&idesc, 0, sizeof(struct inodesc)); 96 idesc.id_func = pass1check; 97 n_files = n_blks = 0; 98 for (c = 0; c < sblock.fs_ncg; c++) { 99 inumber = c * sblock.fs_ipg; 100 setinodebuf(inumber); 101 cgbp = cglookup(c); 102 cgp = cgbp->b_un.b_cg; 103 rebuildcg = 0; 104 if (!check_cgmagic(c, cgbp)) 105 rebuildcg = 1; 106 if (!rebuildcg && sblock.fs_magic == FS_UFS2_MAGIC) { 107 inosused = cgp->cg_initediblk; 108 if (inosused > sblock.fs_ipg) { 109 pfatal( 110 "Too many initialized inodes (%ju > %d) in cylinder group %d\nReset to %d\n", 111 (uintmax_t)inosused, 112 sblock.fs_ipg, c, sblock.fs_ipg); 113 inosused = sblock.fs_ipg; 114 } 115 } else { 116 inosused = sblock.fs_ipg; 117 } 118 if (got_siginfo) { 119 printf("%s: phase 1: cyl group %d of %d (%d%%)\n", 120 cdevname, c, sblock.fs_ncg, 121 c * 100 / sblock.fs_ncg); 122 got_siginfo = 0; 123 } 124 if (got_sigalarm) { 125 setproctitle("%s p1 %d%%", cdevname, 126 c * 100 / sblock.fs_ncg); 127 got_sigalarm = 0; 128 } 129 /* 130 * If we are using soft updates, then we can trust the 131 * cylinder group inode allocation maps to tell us which 132 * inodes are allocated. We will scan the used inode map 133 * to find the inodes that are really in use, and then 134 * read only those inodes in from disk. 135 */ 136 if ((preen || inoopt) && usedsoftdep && !rebuildcg) { 137 cp = &cg_inosused(cgp)[(inosused - 1) / CHAR_BIT]; 138 for ( ; inosused != 0; cp--) { 139 if (*cp == 0) { 140 if (inosused > CHAR_BIT) 141 inosused -= CHAR_BIT; 142 else 143 inosused = 0; 144 continue; 145 } 146 for (i = 1 << (CHAR_BIT - 1); i > 0; i >>= 1) { 147 if (*cp & i) 148 break; 149 inosused--; 150 } 151 break; 152 } 153 } 154 /* 155 * Allocate inoinfo structures for the allocated inodes. 156 */ 157 inostathead[c].il_numalloced = inosused; 158 if (inosused == 0) { 159 inostathead[c].il_stat = NULL; 160 continue; 161 } 162 info = Calloc((unsigned)inosused, sizeof(struct inostat)); 163 if (info == NULL) 164 errx(EEXIT, "cannot alloc %u bytes for inoinfo", 165 (unsigned)(sizeof(struct inostat) * inosused)); 166 inostathead[c].il_stat = info; 167 /* 168 * Scan the allocated inodes. 169 */ 170 for (i = 0; i < inosused; i++, inumber++) { 171 if (inumber < UFS_ROOTINO) { 172 (void)getnextinode(inumber, rebuildcg); 173 continue; 174 } 175 /* 176 * NULL return indicates probable end of allocated 177 * inodes during cylinder group rebuild attempt. 178 * We always keep trying until we get to the minimum 179 * valid number for this cylinder group. 180 */ 181 if (checkinode(inumber, &idesc, rebuildcg) == 0 && 182 i > cgp->cg_initediblk) 183 break; 184 } 185 /* 186 * This optimization speeds up future runs of fsck 187 * by trimming down the number of inodes in cylinder 188 * groups that formerly had many inodes but now have 189 * fewer in use. 190 */ 191 mininos = roundup(inosused + INOPB(&sblock), INOPB(&sblock)); 192 if (inoopt && !preen && !rebuildcg && 193 sblock.fs_magic == FS_UFS2_MAGIC && 194 cgp->cg_initediblk > 2 * INOPB(&sblock) && 195 mininos < cgp->cg_initediblk) { 196 i = cgp->cg_initediblk; 197 if (mininos < 2 * INOPB(&sblock)) 198 cgp->cg_initediblk = 2 * INOPB(&sblock); 199 else 200 cgp->cg_initediblk = mininos; 201 pwarn("CYLINDER GROUP %d: RESET FROM %ju TO %d %s\n", 202 c, i, cgp->cg_initediblk, "VALID INODES"); 203 cgdirty(cgbp); 204 } 205 if (inosused < sblock.fs_ipg) 206 continue; 207 lastino += 1; 208 if (lastino < (c * sblock.fs_ipg)) 209 inosused = 0; 210 else 211 inosused = lastino - (c * sblock.fs_ipg); 212 if (rebuildcg && inosused > cgp->cg_initediblk && 213 sblock.fs_magic == FS_UFS2_MAGIC) { 214 cgp->cg_initediblk = roundup(inosused, INOPB(&sblock)); 215 pwarn("CYLINDER GROUP %d: FOUND %d VALID INODES\n", c, 216 cgp->cg_initediblk); 217 } 218 /* 219 * If we were not able to determine in advance which inodes 220 * were in use, then reduce the size of the inoinfo structure 221 * to the size necessary to describe the inodes that we 222 * really found. 223 */ 224 if (inumber == lastino) 225 continue; 226 inostathead[c].il_numalloced = inosused; 227 if (inosused == 0) { 228 free(inostathead[c].il_stat); 229 inostathead[c].il_stat = NULL; 230 continue; 231 } 232 info = Calloc((unsigned)inosused, sizeof(struct inostat)); 233 if (info == NULL) 234 errx(EEXIT, "cannot alloc %u bytes for inoinfo", 235 (unsigned)(sizeof(struct inostat) * inosused)); 236 memmove(info, inostathead[c].il_stat, inosused * sizeof(*info)); 237 free(inostathead[c].il_stat); 238 inostathead[c].il_stat = info; 239 } 240 freeinodebuf(); 241 } 242 243 static int 244 checkinode(ino_t inumber, struct inodesc *idesc, int rebuildcg) 245 { 246 union dinode *dp; 247 off_t kernmaxfilesize; 248 ufs2_daddr_t ndb; 249 mode_t mode; 250 uintmax_t fixsize; 251 int j, ret, offset; 252 253 if ((dp = getnextinode(inumber, rebuildcg)) == NULL) 254 return (0); 255 mode = DIP(dp, di_mode) & IFMT; 256 if (mode == 0) { 257 if ((sblock.fs_magic == FS_UFS1_MAGIC && 258 (memcmp(dp->dp1.di_db, ufs1_zino.di_db, 259 UFS_NDADDR * sizeof(ufs1_daddr_t)) || 260 memcmp(dp->dp1.di_ib, ufs1_zino.di_ib, 261 UFS_NIADDR * sizeof(ufs1_daddr_t)) || 262 dp->dp1.di_mode || dp->dp1.di_size)) || 263 (sblock.fs_magic == FS_UFS2_MAGIC && 264 (memcmp(dp->dp2.di_db, ufs2_zino.di_db, 265 UFS_NDADDR * sizeof(ufs2_daddr_t)) || 266 memcmp(dp->dp2.di_ib, ufs2_zino.di_ib, 267 UFS_NIADDR * sizeof(ufs2_daddr_t)) || 268 dp->dp2.di_mode || dp->dp2.di_size))) { 269 pfatal("PARTIALLY ALLOCATED INODE I=%lu", 270 (u_long)inumber); 271 if (reply("CLEAR") == 1) { 272 dp = ginode(inumber); 273 clearinode(dp); 274 inodirty(dp); 275 } 276 } 277 inoinfo(inumber)->ino_state = USTATE; 278 return (1); 279 } 280 lastino = inumber; 281 /* This should match the file size limit in ffs_mountfs(). */ 282 if (sblock.fs_magic == FS_UFS1_MAGIC) 283 kernmaxfilesize = (off_t)0x40000000 * sblock.fs_bsize - 1; 284 else 285 kernmaxfilesize = sblock.fs_maxfilesize; 286 if (DIP(dp, di_size) > kernmaxfilesize || 287 DIP(dp, di_size) > sblock.fs_maxfilesize || 288 (mode == IFDIR && DIP(dp, di_size) > MAXDIRSIZE)) { 289 if (debug) 290 printf("bad size %ju:", (uintmax_t)DIP(dp, di_size)); 291 goto unknown; 292 } 293 if (!preen && mode == IFMT && reply("HOLD BAD BLOCK") == 1) { 294 dp = ginode(inumber); 295 DIP_SET(dp, di_size, sblock.fs_fsize); 296 DIP_SET(dp, di_mode, IFREG|0600); 297 inodirty(dp); 298 } 299 if ((mode == IFBLK || mode == IFCHR || mode == IFIFO || 300 mode == IFSOCK) && DIP(dp, di_size) != 0) { 301 if (debug) 302 printf("bad special-file size %ju:", 303 (uintmax_t)DIP(dp, di_size)); 304 goto unknown; 305 } 306 if ((mode == IFBLK || mode == IFCHR) && 307 (dev_t)DIP(dp, di_rdev) == NODEV) { 308 if (debug) 309 printf("bad special-file rdev NODEV:"); 310 goto unknown; 311 } 312 ndb = howmany(DIP(dp, di_size), sblock.fs_bsize); 313 if (ndb < 0) { 314 if (debug) 315 printf("bad size %ju ndb %ju:", 316 (uintmax_t)DIP(dp, di_size), (uintmax_t)ndb); 317 goto unknown; 318 } 319 if (mode == IFBLK || mode == IFCHR) 320 ndb++; 321 if (mode == IFLNK) { 322 /* 323 * Fake ndb value so direct/indirect block checks below 324 * will detect any garbage after symlink string. 325 */ 326 if (DIP(dp, di_size) < (off_t)sblock.fs_maxsymlinklen) { 327 if (sblock.fs_magic == FS_UFS1_MAGIC) 328 ndb = howmany(DIP(dp, di_size), 329 sizeof(ufs1_daddr_t)); 330 else 331 ndb = howmany(DIP(dp, di_size), 332 sizeof(ufs2_daddr_t)); 333 if (ndb > UFS_NDADDR) { 334 j = ndb - UFS_NDADDR; 335 for (ndb = 1; j > 1; j--) 336 ndb *= NINDIR(&sblock); 337 ndb += UFS_NDADDR; 338 } 339 } 340 } 341 for (j = ndb; ndb < UFS_NDADDR && j < UFS_NDADDR; j++) 342 if (DIP(dp, di_db[j]) != 0) { 343 if (debug) 344 printf("bad direct addr[%d]: %ju\n", j, 345 (uintmax_t)DIP(dp, di_db[j])); 346 goto unknown; 347 } 348 for (j = 0, ndb -= UFS_NDADDR; ndb > 0; j++) 349 ndb /= NINDIR(&sblock); 350 for (; j < UFS_NIADDR; j++) 351 if (DIP(dp, di_ib[j]) != 0) { 352 if (debug) 353 printf("bad indirect addr: %ju\n", 354 (uintmax_t)DIP(dp, di_ib[j])); 355 goto unknown; 356 } 357 if (ftypeok(dp) == 0) 358 goto unknown; 359 n_files++; 360 inoinfo(inumber)->ino_linkcnt = DIP(dp, di_nlink); 361 if (mode == IFDIR) { 362 if (DIP(dp, di_size) == 0) 363 inoinfo(inumber)->ino_state = DCLEAR; 364 else if (DIP(dp, di_nlink) <= 0) 365 inoinfo(inumber)->ino_state = DZLINK; 366 else 367 inoinfo(inumber)->ino_state = DSTATE; 368 cacheino(dp, inumber); 369 countdirs++; 370 } else if (DIP(dp, di_nlink) <= 0) 371 inoinfo(inumber)->ino_state = FZLINK; 372 else 373 inoinfo(inumber)->ino_state = FSTATE; 374 inoinfo(inumber)->ino_type = IFTODT(mode); 375 badblk = dupblk = 0; 376 idesc->id_number = inumber; 377 if (DIP(dp, di_flags) & SF_SNAPSHOT) 378 idesc->id_type = SNAP; 379 else 380 idesc->id_type = ADDR; 381 idesc->id_lballoc = -1; 382 (void)ckinode(dp, idesc); 383 if (sblock.fs_magic == FS_UFS2_MAGIC && dp->dp2.di_extsize > 0) { 384 idesc->id_type = ADDR; 385 ndb = howmany(dp->dp2.di_extsize, sblock.fs_bsize); 386 for (j = 0; j < UFS_NXADDR; j++) { 387 if (--ndb == 0 && 388 (offset = blkoff(&sblock, dp->dp2.di_extsize)) != 0) 389 idesc->id_numfrags = numfrags(&sblock, 390 fragroundup(&sblock, offset)); 391 else 392 idesc->id_numfrags = sblock.fs_frag; 393 if (dp->dp2.di_extb[j] == 0) 394 continue; 395 idesc->id_blkno = dp->dp2.di_extb[j]; 396 ret = (*idesc->id_func)(idesc); 397 if (ret & STOP) 398 break; 399 } 400 } 401 if (sblock.fs_magic == FS_UFS2_MAGIC) 402 eascan(idesc, &dp->dp2); 403 idesc->id_entryno *= btodb(sblock.fs_fsize); 404 if (DIP(dp, di_blocks) != idesc->id_entryno) { 405 pwarn("INCORRECT BLOCK COUNT I=%lu (%ju should be %ju)", 406 (u_long)inumber, (uintmax_t)DIP(dp, di_blocks), 407 (uintmax_t)idesc->id_entryno); 408 if (preen) 409 printf(" (CORRECTED)\n"); 410 else if (reply("CORRECT") == 0) 411 return (1); 412 if (bkgrdflag == 0) { 413 dp = ginode(inumber); 414 DIP_SET(dp, di_blocks, idesc->id_entryno); 415 inodirty(dp); 416 } else { 417 cmd.value = idesc->id_number; 418 cmd.size = idesc->id_entryno - DIP(dp, di_blocks); 419 if (debug) 420 printf("adjblkcnt ino %ju amount %lld\n", 421 (uintmax_t)cmd.value, (long long)cmd.size); 422 if (sysctl(adjblkcnt, MIBSIZE, 0, 0, 423 &cmd, sizeof cmd) == -1) 424 rwerror("ADJUST INODE BLOCK COUNT", cmd.value); 425 } 426 } 427 /* 428 * Soft updates will always ensure that the file size is correct 429 * for files that contain only direct block pointers. However 430 * soft updates does not roll back sizes for files with indirect 431 * blocks that it has set to unallocated because their contents 432 * have not yet been written to disk. Hence, the file can appear 433 * to have a hole at its end because the block pointer has been 434 * rolled back to zero. Thus, id_lballoc tracks the last allocated 435 * block in the file. Here, for files that extend into indirect 436 * blocks, we check for a size past the last allocated block of 437 * the file and if that is found, shorten the file to reference 438 * the last allocated block to avoid having it reference a hole 439 * at its end. 440 */ 441 if (DIP(dp, di_size) > UFS_NDADDR * sblock.fs_bsize && 442 idesc->id_lballoc != lblkno(&sblock, DIP(dp, di_size) - 1)) { 443 fixsize = lblktosize(&sblock, idesc->id_lballoc + 1); 444 pwarn("INODE %lu: FILE SIZE %ju BEYOND END OF ALLOCATED FILE, " 445 "SIZE SHOULD BE %ju", (u_long)inumber, 446 (uintmax_t)DIP(dp, di_size), fixsize); 447 if (preen) 448 printf(" (ADJUSTED)\n"); 449 else if (reply("ADJUST") == 0) 450 return (1); 451 if (bkgrdflag == 0) { 452 dp = ginode(inumber); 453 DIP_SET(dp, di_size, fixsize); 454 inodirty(dp); 455 } else { 456 cmd.value = idesc->id_number; 457 cmd.size = fixsize; 458 if (debug) 459 printf("setsize ino %ju size set to %ju\n", 460 (uintmax_t)cmd.value, (uintmax_t)cmd.size); 461 if (sysctl(setsize, MIBSIZE, 0, 0, 462 &cmd, sizeof cmd) == -1) 463 rwerror("SET INODE SIZE", cmd.value); 464 } 465 466 } 467 return (1); 468 unknown: 469 pfatal("UNKNOWN FILE TYPE I=%lu", (u_long)inumber); 470 inoinfo(inumber)->ino_state = FCLEAR; 471 if (reply("CLEAR") == 1) { 472 inoinfo(inumber)->ino_state = USTATE; 473 dp = ginode(inumber); 474 clearinode(dp); 475 inodirty(dp); 476 } 477 return (1); 478 } 479 480 int 481 pass1check(struct inodesc *idesc) 482 { 483 int res = KEEPON; 484 int anyout, nfrags; 485 ufs2_daddr_t blkno = idesc->id_blkno; 486 struct dups *dlp; 487 struct dups *new; 488 489 if (idesc->id_type == SNAP) { 490 if (blkno == BLK_NOCOPY) 491 return (KEEPON); 492 if (idesc->id_number == cursnapshot) { 493 if (blkno == blkstofrags(&sblock, idesc->id_lbn)) 494 return (KEEPON); 495 if (blkno == BLK_SNAP) { 496 blkno = blkstofrags(&sblock, idesc->id_lbn); 497 idesc->id_entryno -= idesc->id_numfrags; 498 } 499 } else { 500 if (blkno == BLK_SNAP) 501 return (KEEPON); 502 } 503 } 504 if ((anyout = chkrange(blkno, idesc->id_numfrags)) != 0) { 505 blkerror(idesc->id_number, "BAD", blkno); 506 if (badblk++ >= MAXBAD) { 507 pwarn("EXCESSIVE BAD BLKS I=%lu", 508 (u_long)idesc->id_number); 509 if (preen) 510 printf(" (SKIPPING)\n"); 511 else if (reply("CONTINUE") == 0) { 512 ckfini(0); 513 exit(EEXIT); 514 } 515 rerun = 1; 516 return (STOP); 517 } 518 } 519 for (nfrags = idesc->id_numfrags; nfrags > 0; blkno++, nfrags--) { 520 if (anyout && chkrange(blkno, 1)) { 521 res = SKIP; 522 } else if (!testbmap(blkno)) { 523 n_blks++; 524 setbmap(blkno); 525 } else { 526 blkerror(idesc->id_number, "DUP", blkno); 527 if (dupblk++ >= MAXDUP) { 528 pwarn("EXCESSIVE DUP BLKS I=%lu", 529 (u_long)idesc->id_number); 530 if (preen) 531 printf(" (SKIPPING)\n"); 532 else if (reply("CONTINUE") == 0) { 533 ckfini(0); 534 exit(EEXIT); 535 } 536 rerun = 1; 537 return (STOP); 538 } 539 new = (struct dups *)Malloc(sizeof(struct dups)); 540 if (new == NULL) { 541 pfatal("DUP TABLE OVERFLOW."); 542 if (reply("CONTINUE") == 0) { 543 ckfini(0); 544 exit(EEXIT); 545 } 546 rerun = 1; 547 return (STOP); 548 } 549 new->dup = blkno; 550 if (muldup == NULL) { 551 duplist = muldup = new; 552 new->next = NULL; 553 } else { 554 new->next = muldup->next; 555 muldup->next = new; 556 } 557 for (dlp = duplist; dlp != muldup; dlp = dlp->next) 558 if (dlp->dup == blkno) 559 break; 560 if (dlp == muldup && dlp->dup != blkno) 561 muldup = new; 562 } 563 /* 564 * count the number of blocks found in id_entryno 565 */ 566 idesc->id_entryno++; 567 } 568 if (idesc->id_lballoc == -1 || idesc->id_lballoc < idesc->id_lbn) 569 idesc->id_lballoc = idesc->id_lbn; 570 return (res); 571 } 572