xref: /freebsd/sbin/fsck_ffs/pass1.c (revision d6b92ffa)
1 /*
2  * Copyright (c) 1980, 1986, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  */
29 
30 #if 0
31 #ifndef lint
32 static const char sccsid[] = "@(#)pass1.c	8.6 (Berkeley) 4/28/95";
33 #endif /* not lint */
34 #endif
35 #include <sys/cdefs.h>
36 __FBSDID("$FreeBSD$");
37 
38 #include <sys/param.h>
39 #include <sys/stat.h>
40 #include <sys/sysctl.h>
41 
42 #include <ufs/ufs/dinode.h>
43 #include <ufs/ufs/dir.h>
44 #include <ufs/ffs/fs.h>
45 
46 #include <err.h>
47 #include <limits.h>
48 #include <stdint.h>
49 #include <string.h>
50 
51 #include "fsck.h"
52 
53 static ufs2_daddr_t badblk;
54 static ufs2_daddr_t dupblk;
55 static ino_t lastino;		/* last inode in use */
56 
57 static int checkinode(ino_t inumber, struct inodesc *, int rebuildcg);
58 
59 void
60 pass1(void)
61 {
62 	struct inostat *info;
63 	struct inodesc idesc;
64 	struct bufarea *cgbp;
65 	struct cg *cgp;
66 	ino_t inumber, inosused, mininos;
67 	ufs2_daddr_t i, cgd;
68 	u_int8_t *cp;
69 	int c, rebuildcg;
70 
71 	badblk = dupblk = lastino = 0;
72 
73 	/*
74 	 * Set file system reserved blocks in used block map.
75 	 */
76 	for (c = 0; c < sblock.fs_ncg; c++) {
77 		cgd = cgdmin(&sblock, c);
78 		if (c == 0) {
79 			i = cgbase(&sblock, c);
80 		} else
81 			i = cgsblock(&sblock, c);
82 		for (; i < cgd; i++)
83 			setbmap(i);
84 	}
85 	i = sblock.fs_csaddr;
86 	cgd = i + howmany(sblock.fs_cssize, sblock.fs_fsize);
87 	for (; i < cgd; i++)
88 		setbmap(i);
89 
90 	/*
91 	 * Find all allocated blocks.
92 	 */
93 	memset(&idesc, 0, sizeof(struct inodesc));
94 	idesc.id_func = pass1check;
95 	n_files = n_blks = 0;
96 	for (c = 0; c < sblock.fs_ncg; c++) {
97 		inumber = c * sblock.fs_ipg;
98 		setinodebuf(inumber);
99 		cgbp = cgget(c);
100 		cgp = cgbp->b_un.b_cg;
101 		rebuildcg = 0;
102 		if (!check_cgmagic(c, cgbp))
103 			rebuildcg = 1;
104 		if (!rebuildcg && sblock.fs_magic == FS_UFS2_MAGIC) {
105 			inosused = cgp->cg_initediblk;
106 			if (inosused > sblock.fs_ipg) {
107 				pfatal(
108 "Too many initialized inodes (%ju > %d) in cylinder group %d\nReset to %d\n",
109 				    (uintmax_t)inosused,
110 				    sblock.fs_ipg, c, sblock.fs_ipg);
111 				inosused = sblock.fs_ipg;
112 			}
113 		} else {
114 			inosused = sblock.fs_ipg;
115 		}
116 		if (got_siginfo) {
117 			printf("%s: phase 1: cyl group %d of %d (%d%%)\n",
118 			    cdevname, c, sblock.fs_ncg,
119 			    c * 100 / sblock.fs_ncg);
120 			got_siginfo = 0;
121 		}
122 		if (got_sigalarm) {
123 			setproctitle("%s p1 %d%%", cdevname,
124 			     c * 100 / sblock.fs_ncg);
125 			got_sigalarm = 0;
126 		}
127 		/*
128 		 * If we are using soft updates, then we can trust the
129 		 * cylinder group inode allocation maps to tell us which
130 		 * inodes are allocated. We will scan the used inode map
131 		 * to find the inodes that are really in use, and then
132 		 * read only those inodes in from disk.
133 		 */
134 		if ((preen || inoopt) && usedsoftdep && !rebuildcg) {
135 			cp = &cg_inosused(cgp)[(inosused - 1) / CHAR_BIT];
136 			for ( ; inosused != 0; cp--) {
137 				if (*cp == 0) {
138 					if (inosused > CHAR_BIT)
139 						inosused -= CHAR_BIT;
140 					else
141 						inosused = 0;
142 					continue;
143 				}
144 				for (i = 1 << (CHAR_BIT - 1); i > 0; i >>= 1) {
145 					if (*cp & i)
146 						break;
147 					inosused--;
148 				}
149 				break;
150 			}
151 		}
152 		/*
153 		 * Allocate inoinfo structures for the allocated inodes.
154 		 */
155 		inostathead[c].il_numalloced = inosused;
156 		if (inosused == 0) {
157 			inostathead[c].il_stat = NULL;
158 			continue;
159 		}
160 		info = Calloc((unsigned)inosused, sizeof(struct inostat));
161 		if (info == NULL)
162 			errx(EEXIT, "cannot alloc %u bytes for inoinfo",
163 			    (unsigned)(sizeof(struct inostat) * inosused));
164 		inostathead[c].il_stat = info;
165 		/*
166 		 * Scan the allocated inodes.
167 		 */
168 		for (i = 0; i < inosused; i++, inumber++) {
169 			if (inumber < UFS_ROOTINO) {
170 				(void)getnextinode(inumber, rebuildcg);
171 				continue;
172 			}
173 			/*
174 			 * NULL return indicates probable end of allocated
175 			 * inodes during cylinder group rebuild attempt.
176 			 * We always keep trying until we get to the minimum
177 			 * valid number for this cylinder group.
178 			 */
179 			if (checkinode(inumber, &idesc, rebuildcg) == 0 &&
180 			    i > cgp->cg_initediblk)
181 				break;
182 		}
183 		/*
184 		 * This optimization speeds up future runs of fsck
185 		 * by trimming down the number of inodes in cylinder
186 		 * groups that formerly had many inodes but now have
187 		 * fewer in use.
188 		 */
189 		mininos = roundup(inosused + INOPB(&sblock), INOPB(&sblock));
190 		if (inoopt && !preen && !rebuildcg &&
191 		    sblock.fs_magic == FS_UFS2_MAGIC &&
192 		    cgp->cg_initediblk > 2 * INOPB(&sblock) &&
193 		    mininos < cgp->cg_initediblk) {
194 			i = cgp->cg_initediblk;
195 			if (mininos < 2 * INOPB(&sblock))
196 				cgp->cg_initediblk = 2 * INOPB(&sblock);
197 			else
198 				cgp->cg_initediblk = mininos;
199 			pwarn("CYLINDER GROUP %d: RESET FROM %ju TO %d %s\n",
200 			    c, i, cgp->cg_initediblk, "VALID INODES");
201 			dirty(cgbp);
202 		}
203 		if (inosused < sblock.fs_ipg)
204 			continue;
205 		lastino += 1;
206 		if (lastino < (c * sblock.fs_ipg))
207 			inosused = 0;
208 		else
209 			inosused = lastino - (c * sblock.fs_ipg);
210 		if (rebuildcg && inosused > cgp->cg_initediblk &&
211 		    sblock.fs_magic == FS_UFS2_MAGIC) {
212 			cgp->cg_initediblk = roundup(inosused, INOPB(&sblock));
213 			pwarn("CYLINDER GROUP %d: FOUND %d VALID INODES\n", c,
214 			    cgp->cg_initediblk);
215 		}
216 		/*
217 		 * If we were not able to determine in advance which inodes
218 		 * were in use, then reduce the size of the inoinfo structure
219 		 * to the size necessary to describe the inodes that we
220 		 * really found.
221 		 */
222 		if (inumber == lastino)
223 			continue;
224 		inostathead[c].il_numalloced = inosused;
225 		if (inosused == 0) {
226 			free(inostathead[c].il_stat);
227 			inostathead[c].il_stat = NULL;
228 			continue;
229 		}
230 		info = Calloc((unsigned)inosused, sizeof(struct inostat));
231 		if (info == NULL)
232 			errx(EEXIT, "cannot alloc %u bytes for inoinfo",
233 			    (unsigned)(sizeof(struct inostat) * inosused));
234 		memmove(info, inostathead[c].il_stat, inosused * sizeof(*info));
235 		free(inostathead[c].il_stat);
236 		inostathead[c].il_stat = info;
237 	}
238 	freeinodebuf();
239 }
240 
241 static int
242 checkinode(ino_t inumber, struct inodesc *idesc, int rebuildcg)
243 {
244 	union dinode *dp;
245 	off_t kernmaxfilesize;
246 	ufs2_daddr_t ndb;
247 	mode_t mode;
248 	int j, ret, offset;
249 
250 	if ((dp = getnextinode(inumber, rebuildcg)) == NULL)
251 		return (0);
252 	mode = DIP(dp, di_mode) & IFMT;
253 	if (mode == 0) {
254 		if ((sblock.fs_magic == FS_UFS1_MAGIC &&
255 		     (memcmp(dp->dp1.di_db, ufs1_zino.di_db,
256 			UFS_NDADDR * sizeof(ufs1_daddr_t)) ||
257 		      memcmp(dp->dp1.di_ib, ufs1_zino.di_ib,
258 			UFS_NIADDR * sizeof(ufs1_daddr_t)) ||
259 		      dp->dp1.di_mode || dp->dp1.di_size)) ||
260 		    (sblock.fs_magic == FS_UFS2_MAGIC &&
261 		     (memcmp(dp->dp2.di_db, ufs2_zino.di_db,
262 			UFS_NDADDR * sizeof(ufs2_daddr_t)) ||
263 		      memcmp(dp->dp2.di_ib, ufs2_zino.di_ib,
264 			UFS_NIADDR * sizeof(ufs2_daddr_t)) ||
265 		      dp->dp2.di_mode || dp->dp2.di_size))) {
266 			pfatal("PARTIALLY ALLOCATED INODE I=%lu",
267 			    (u_long)inumber);
268 			if (reply("CLEAR") == 1) {
269 				dp = ginode(inumber);
270 				clearinode(dp);
271 				inodirty();
272 			}
273 		}
274 		inoinfo(inumber)->ino_state = USTATE;
275 		return (1);
276 	}
277 	lastino = inumber;
278 	/* This should match the file size limit in ffs_mountfs(). */
279 	if (sblock.fs_magic == FS_UFS1_MAGIC)
280 		kernmaxfilesize = (off_t)0x40000000 * sblock.fs_bsize - 1;
281 	else
282 		kernmaxfilesize = sblock.fs_maxfilesize;
283 	if (DIP(dp, di_size) > kernmaxfilesize ||
284 	    DIP(dp, di_size) > sblock.fs_maxfilesize ||
285 	    (mode == IFDIR && DIP(dp, di_size) > MAXDIRSIZE)) {
286 		if (debug)
287 			printf("bad size %ju:", (uintmax_t)DIP(dp, di_size));
288 		goto unknown;
289 	}
290 	if (!preen && mode == IFMT && reply("HOLD BAD BLOCK") == 1) {
291 		dp = ginode(inumber);
292 		DIP_SET(dp, di_size, sblock.fs_fsize);
293 		DIP_SET(dp, di_mode, IFREG|0600);
294 		inodirty();
295 	}
296 	if ((mode == IFBLK || mode == IFCHR || mode == IFIFO ||
297 	     mode == IFSOCK) && DIP(dp, di_size) != 0) {
298 		if (debug)
299 			printf("bad special-file size %ju:",
300 			    (uintmax_t)DIP(dp, di_size));
301 		goto unknown;
302 	}
303 	if ((mode == IFBLK || mode == IFCHR) &&
304 	    (dev_t)DIP(dp, di_rdev) == NODEV) {
305 		if (debug)
306 			printf("bad special-file rdev NODEV:");
307 		goto unknown;
308 	}
309 	ndb = howmany(DIP(dp, di_size), sblock.fs_bsize);
310 	if (ndb < 0) {
311 		if (debug)
312 			printf("bad size %ju ndb %ju:",
313 				(uintmax_t)DIP(dp, di_size), (uintmax_t)ndb);
314 		goto unknown;
315 	}
316 	if (mode == IFBLK || mode == IFCHR)
317 		ndb++;
318 	if (mode == IFLNK) {
319 		/*
320 		 * Fake ndb value so direct/indirect block checks below
321 		 * will detect any garbage after symlink string.
322 		 */
323 		if (DIP(dp, di_size) < (off_t)sblock.fs_maxsymlinklen) {
324 			if (sblock.fs_magic == FS_UFS1_MAGIC)
325 				ndb = howmany(DIP(dp, di_size),
326 				    sizeof(ufs1_daddr_t));
327 			else
328 				ndb = howmany(DIP(dp, di_size),
329 				    sizeof(ufs2_daddr_t));
330 			if (ndb > UFS_NDADDR) {
331 				j = ndb - UFS_NDADDR;
332 				for (ndb = 1; j > 1; j--)
333 					ndb *= NINDIR(&sblock);
334 				ndb += UFS_NDADDR;
335 			}
336 		}
337 	}
338 	for (j = ndb; ndb < UFS_NDADDR && j < UFS_NDADDR; j++)
339 		if (DIP(dp, di_db[j]) != 0) {
340 			if (debug)
341 				printf("bad direct addr[%d]: %ju\n", j,
342 				    (uintmax_t)DIP(dp, di_db[j]));
343 			goto unknown;
344 		}
345 	for (j = 0, ndb -= UFS_NDADDR; ndb > 0; j++)
346 		ndb /= NINDIR(&sblock);
347 	for (; j < UFS_NIADDR; j++)
348 		if (DIP(dp, di_ib[j]) != 0) {
349 			if (debug)
350 				printf("bad indirect addr: %ju\n",
351 				    (uintmax_t)DIP(dp, di_ib[j]));
352 			goto unknown;
353 		}
354 	if (ftypeok(dp) == 0)
355 		goto unknown;
356 	n_files++;
357 	inoinfo(inumber)->ino_linkcnt = DIP(dp, di_nlink);
358 	if (mode == IFDIR) {
359 		if (DIP(dp, di_size) == 0)
360 			inoinfo(inumber)->ino_state = DCLEAR;
361 		else if (DIP(dp, di_nlink) <= 0)
362 			inoinfo(inumber)->ino_state = DZLINK;
363 		else
364 			inoinfo(inumber)->ino_state = DSTATE;
365 		cacheino(dp, inumber);
366 		countdirs++;
367 	} else if (DIP(dp, di_nlink) <= 0)
368 		inoinfo(inumber)->ino_state = FZLINK;
369 	else
370 		inoinfo(inumber)->ino_state = FSTATE;
371 	inoinfo(inumber)->ino_type = IFTODT(mode);
372 	badblk = dupblk = 0;
373 	idesc->id_number = inumber;
374 	if (DIP(dp, di_flags) & SF_SNAPSHOT)
375 		idesc->id_type = SNAP;
376 	else
377 		idesc->id_type = ADDR;
378 	(void)ckinode(dp, idesc);
379 	if (sblock.fs_magic == FS_UFS2_MAGIC && dp->dp2.di_extsize > 0) {
380 		idesc->id_type = ADDR;
381 		ndb = howmany(dp->dp2.di_extsize, sblock.fs_bsize);
382 		for (j = 0; j < UFS_NXADDR; j++) {
383 			if (--ndb == 0 &&
384 			    (offset = blkoff(&sblock, dp->dp2.di_extsize)) != 0)
385 				idesc->id_numfrags = numfrags(&sblock,
386 				    fragroundup(&sblock, offset));
387 			else
388 				idesc->id_numfrags = sblock.fs_frag;
389 			if (dp->dp2.di_extb[j] == 0)
390 				continue;
391 			idesc->id_blkno = dp->dp2.di_extb[j];
392 			ret = (*idesc->id_func)(idesc);
393 			if (ret & STOP)
394 				break;
395 		}
396 	}
397 	if (sblock.fs_magic == FS_UFS2_MAGIC)
398 		eascan(idesc, &dp->dp2);
399 	idesc->id_entryno *= btodb(sblock.fs_fsize);
400 	if (DIP(dp, di_blocks) != idesc->id_entryno) {
401 		pwarn("INCORRECT BLOCK COUNT I=%lu (%ju should be %ju)",
402 		    (u_long)inumber, (uintmax_t)DIP(dp, di_blocks),
403 		    (uintmax_t)idesc->id_entryno);
404 		if (preen)
405 			printf(" (CORRECTED)\n");
406 		else if (reply("CORRECT") == 0)
407 			return (1);
408 		if (bkgrdflag == 0) {
409 			dp = ginode(inumber);
410 			DIP_SET(dp, di_blocks, idesc->id_entryno);
411 			inodirty();
412 		} else {
413 			cmd.value = idesc->id_number;
414 			cmd.size = idesc->id_entryno - DIP(dp, di_blocks);
415 			if (debug)
416 				printf("adjblkcnt ino %ju amount %lld\n",
417 				    (uintmax_t)cmd.value, (long long)cmd.size);
418 			if (sysctl(adjblkcnt, MIBSIZE, 0, 0,
419 			    &cmd, sizeof cmd) == -1)
420 				rwerror("ADJUST INODE BLOCK COUNT", cmd.value);
421 		}
422 	}
423 	return (1);
424 unknown:
425 	pfatal("UNKNOWN FILE TYPE I=%lu", (u_long)inumber);
426 	inoinfo(inumber)->ino_state = FCLEAR;
427 	if (reply("CLEAR") == 1) {
428 		inoinfo(inumber)->ino_state = USTATE;
429 		dp = ginode(inumber);
430 		clearinode(dp);
431 		inodirty();
432 	}
433 	return (1);
434 }
435 
436 int
437 pass1check(struct inodesc *idesc)
438 {
439 	int res = KEEPON;
440 	int anyout, nfrags;
441 	ufs2_daddr_t blkno = idesc->id_blkno;
442 	struct dups *dlp;
443 	struct dups *new;
444 
445 	if (idesc->id_type == SNAP) {
446 		if (blkno == BLK_NOCOPY)
447 			return (KEEPON);
448 		if (idesc->id_number == cursnapshot) {
449 			if (blkno == blkstofrags(&sblock, idesc->id_lbn))
450 				return (KEEPON);
451 			if (blkno == BLK_SNAP) {
452 				blkno = blkstofrags(&sblock, idesc->id_lbn);
453 				idesc->id_entryno -= idesc->id_numfrags;
454 			}
455 		} else {
456 			if (blkno == BLK_SNAP)
457 				return (KEEPON);
458 		}
459 	}
460 	if ((anyout = chkrange(blkno, idesc->id_numfrags)) != 0) {
461 		blkerror(idesc->id_number, "BAD", blkno);
462 		if (badblk++ >= MAXBAD) {
463 			pwarn("EXCESSIVE BAD BLKS I=%lu",
464 			    (u_long)idesc->id_number);
465 			if (preen)
466 				printf(" (SKIPPING)\n");
467 			else if (reply("CONTINUE") == 0) {
468 				ckfini(0);
469 				exit(EEXIT);
470 			}
471 			rerun = 1;
472 			return (STOP);
473 		}
474 	}
475 	for (nfrags = idesc->id_numfrags; nfrags > 0; blkno++, nfrags--) {
476 		if (anyout && chkrange(blkno, 1)) {
477 			res = SKIP;
478 		} else if (!testbmap(blkno)) {
479 			n_blks++;
480 			setbmap(blkno);
481 		} else {
482 			blkerror(idesc->id_number, "DUP", blkno);
483 			if (dupblk++ >= MAXDUP) {
484 				pwarn("EXCESSIVE DUP BLKS I=%lu",
485 					(u_long)idesc->id_number);
486 				if (preen)
487 					printf(" (SKIPPING)\n");
488 				else if (reply("CONTINUE") == 0) {
489 					ckfini(0);
490 					exit(EEXIT);
491 				}
492 				rerun = 1;
493 				return (STOP);
494 			}
495 			new = (struct dups *)Malloc(sizeof(struct dups));
496 			if (new == NULL) {
497 				pfatal("DUP TABLE OVERFLOW.");
498 				if (reply("CONTINUE") == 0) {
499 					ckfini(0);
500 					exit(EEXIT);
501 				}
502 				rerun = 1;
503 				return (STOP);
504 			}
505 			new->dup = blkno;
506 			if (muldup == NULL) {
507 				duplist = muldup = new;
508 				new->next = NULL;
509 			} else {
510 				new->next = muldup->next;
511 				muldup->next = new;
512 			}
513 			for (dlp = duplist; dlp != muldup; dlp = dlp->next)
514 				if (dlp->dup == blkno)
515 					break;
516 			if (dlp == muldup && dlp->dup != blkno)
517 				muldup = new;
518 		}
519 		/*
520 		 * count the number of blocks found in id_entryno
521 		 */
522 		idesc->id_entryno++;
523 	}
524 	return (res);
525 }
526