xref: /freebsd/sbin/fsck_ffs/pass1.c (revision 830940567b49bb0c08dfaed40418999e76616909)
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  * 4. 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 	ino_t inumber, inosused, mininos;
65 	ufs2_daddr_t i, cgd;
66 	u_int8_t *cp;
67 	int c, rebuildcg;
68 
69 	/*
70 	 * Set file system reserved blocks in used block map.
71 	 */
72 	for (c = 0; c < sblock.fs_ncg; c++) {
73 		cgd = cgdmin(&sblock, c);
74 		if (c == 0) {
75 			i = cgbase(&sblock, c);
76 		} else
77 			i = cgsblock(&sblock, c);
78 		for (; i < cgd; i++)
79 			setbmap(i);
80 	}
81 	i = sblock.fs_csaddr;
82 	cgd = i + howmany(sblock.fs_cssize, sblock.fs_fsize);
83 	for (; i < cgd; i++)
84 		setbmap(i);
85 
86 	/*
87 	 * Find all allocated blocks.
88 	 */
89 	memset(&idesc, 0, sizeof(struct inodesc));
90 	idesc.id_func = pass1check;
91 	n_files = n_blks = 0;
92 	for (c = 0; c < sblock.fs_ncg; c++) {
93 		inumber = c * sblock.fs_ipg;
94 		setinodebuf(inumber);
95 		getblk(&cgblk, cgtod(&sblock, c), sblock.fs_cgsize);
96 		rebuildcg = 0;
97 		if (!check_cgmagic(c, &cgrp))
98 			rebuildcg = 1;
99 		if (!rebuildcg && sblock.fs_magic == FS_UFS2_MAGIC) {
100 			inosused = cgrp.cg_initediblk;
101 			if (inosused > sblock.fs_ipg)
102 				inosused = sblock.fs_ipg;
103 		} else
104 			inosused = sblock.fs_ipg;
105 		if (got_siginfo) {
106 			printf("%s: phase 1: cyl group %d of %d (%d%%)\n",
107 			    cdevname, c, sblock.fs_ncg,
108 			    c * 100 / sblock.fs_ncg);
109 			got_siginfo = 0;
110 		}
111 		if (got_sigalarm) {
112 			setproctitle("%s p1 %d%%", cdevname,
113 			     c * 100 / sblock.fs_ncg);
114 			got_sigalarm = 0;
115 		}
116 		/*
117 		 * If we are using soft updates, then we can trust the
118 		 * cylinder group inode allocation maps to tell us which
119 		 * inodes are allocated. We will scan the used inode map
120 		 * to find the inodes that are really in use, and then
121 		 * read only those inodes in from disk.
122 		 */
123 		if ((preen || inoopt) && usedsoftdep && !rebuildcg) {
124 			cp = &cg_inosused(&cgrp)[(inosused - 1) / CHAR_BIT];
125 			for ( ; inosused > 0; inosused -= CHAR_BIT, cp--) {
126 				if (*cp == 0)
127 					continue;
128 				for (i = 1 << (CHAR_BIT - 1); i > 0; i >>= 1) {
129 					if (*cp & i)
130 						break;
131 					inosused--;
132 				}
133 				break;
134 			}
135 			if (inosused < 0)
136 				inosused = 0;
137 		}
138 		/*
139 		 * Allocate inoinfo structures for the allocated inodes.
140 		 */
141 		inostathead[c].il_numalloced = inosused;
142 		if (inosused == 0) {
143 			inostathead[c].il_stat = 0;
144 			continue;
145 		}
146 		info = calloc((unsigned)inosused, sizeof(struct inostat));
147 		if (info == NULL)
148 			errx(EEXIT, "cannot alloc %u bytes for inoinfo",
149 			    (unsigned)(sizeof(struct inostat) * inosused));
150 		inostathead[c].il_stat = info;
151 		/*
152 		 * Scan the allocated inodes.
153 		 */
154 		for (i = 0; i < inosused; i++, inumber++) {
155 			if (inumber < ROOTINO) {
156 				(void)getnextinode(inumber, rebuildcg);
157 				continue;
158 			}
159 			/*
160 			 * NULL return indicates probable end of allocated
161 			 * inodes during cylinder group rebuild attempt.
162 			 * We always keep trying until we get to the minimum
163 			 * valid number for this cylinder group.
164 			 */
165 			if (checkinode(inumber, &idesc, rebuildcg) == 0 &&
166 			    i > cgrp.cg_initediblk)
167 				break;
168 		}
169 		/*
170 		 * This optimization speeds up future runs of fsck
171 		 * by trimming down the number of inodes in cylinder
172 		 * groups that formerly had many inodes but now have
173 		 * fewer in use.
174 		 */
175 		mininos = roundup(inosused + INOPB(&sblock), INOPB(&sblock));
176 		if (inoopt && !preen && !rebuildcg &&
177 		    sblock.fs_magic == FS_UFS2_MAGIC &&
178 		    cgrp.cg_initediblk > 2 * INOPB(&sblock) &&
179 		    mininos < cgrp.cg_initediblk) {
180 			i = cgrp.cg_initediblk;
181 			if (mininos < 2 * INOPB(&sblock))
182 				cgrp.cg_initediblk = 2 * INOPB(&sblock);
183 			else
184 				cgrp.cg_initediblk = mininos;
185 			pwarn("CYLINDER GROUP %d: RESET FROM %ju TO %d %s\n",
186 			    c, i, cgrp.cg_initediblk, "VALID INODES");
187 			cgdirty();
188 		}
189 		if (inosused < sblock.fs_ipg)
190 			continue;
191 		lastino += 1;
192 		if (lastino < (c * sblock.fs_ipg))
193 			inosused = 0;
194 		else
195 			inosused = lastino - (c * sblock.fs_ipg);
196 		if (rebuildcg && inosused > cgrp.cg_initediblk &&
197 		    sblock.fs_magic == FS_UFS2_MAGIC) {
198 			cgrp.cg_initediblk = roundup(inosused, INOPB(&sblock));
199 			pwarn("CYLINDER GROUP %d: FOUND %d VALID INODES\n", c,
200 			    cgrp.cg_initediblk);
201 		}
202 		/*
203 		 * If we were not able to determine in advance which inodes
204 		 * were in use, then reduce the size of the inoinfo structure
205 		 * to the size necessary to describe the inodes that we
206 		 * really found.
207 		 */
208 		if (inumber == lastino)
209 			continue;
210 		inostathead[c].il_numalloced = inosused;
211 		if (inosused == 0) {
212 			free(inostathead[c].il_stat);
213 			inostathead[c].il_stat = 0;
214 			continue;
215 		}
216 		info = calloc((unsigned)inosused, sizeof(struct inostat));
217 		if (info == NULL)
218 			errx(EEXIT, "cannot alloc %u bytes for inoinfo",
219 			    (unsigned)(sizeof(struct inostat) * inosused));
220 		memmove(info, inostathead[c].il_stat, inosused * sizeof(*info));
221 		free(inostathead[c].il_stat);
222 		inostathead[c].il_stat = info;
223 	}
224 	freeinodebuf();
225 }
226 
227 static int
228 checkinode(ino_t inumber, struct inodesc *idesc, int rebuildcg)
229 {
230 	union dinode *dp;
231 	off_t kernmaxfilesize;
232 	ufs2_daddr_t ndb;
233 	mode_t mode;
234 	int j, ret, offset;
235 
236 	if ((dp = getnextinode(inumber, rebuildcg)) == NULL)
237 		return (0);
238 	mode = DIP(dp, di_mode) & IFMT;
239 	if (mode == 0) {
240 		if ((sblock.fs_magic == FS_UFS1_MAGIC &&
241 		     (memcmp(dp->dp1.di_db, ufs1_zino.di_db,
242 			NDADDR * sizeof(ufs1_daddr_t)) ||
243 		      memcmp(dp->dp1.di_ib, ufs1_zino.di_ib,
244 			NIADDR * sizeof(ufs1_daddr_t)) ||
245 		      dp->dp1.di_mode || dp->dp1.di_size)) ||
246 		    (sblock.fs_magic == FS_UFS2_MAGIC &&
247 		     (memcmp(dp->dp2.di_db, ufs2_zino.di_db,
248 			NDADDR * sizeof(ufs2_daddr_t)) ||
249 		      memcmp(dp->dp2.di_ib, ufs2_zino.di_ib,
250 			NIADDR * sizeof(ufs2_daddr_t)) ||
251 		      dp->dp2.di_mode || dp->dp2.di_size))) {
252 			pfatal("PARTIALLY ALLOCATED INODE I=%lu",
253 			    (u_long)inumber);
254 			if (reply("CLEAR") == 1) {
255 				dp = ginode(inumber);
256 				clearinode(dp);
257 				inodirty();
258 			}
259 		}
260 		inoinfo(inumber)->ino_state = USTATE;
261 		return (1);
262 	}
263 	lastino = inumber;
264 	/* This should match the file size limit in ffs_mountfs(). */
265 	if (sblock.fs_magic == FS_UFS1_MAGIC)
266 		kernmaxfilesize = (off_t)0x40000000 * sblock.fs_bsize - 1;
267 	else
268 		kernmaxfilesize = sblock.fs_maxfilesize;
269 	if (DIP(dp, di_size) > kernmaxfilesize ||
270 	    DIP(dp, di_size) > sblock.fs_maxfilesize ||
271 	    (mode == IFDIR && DIP(dp, di_size) > MAXDIRSIZE)) {
272 		if (debug)
273 			printf("bad size %ju:", (uintmax_t)DIP(dp, di_size));
274 		goto unknown;
275 	}
276 	if (!preen && mode == IFMT && reply("HOLD BAD BLOCK") == 1) {
277 		dp = ginode(inumber);
278 		DIP_SET(dp, di_size, sblock.fs_fsize);
279 		DIP_SET(dp, di_mode, IFREG|0600);
280 		inodirty();
281 	}
282 	if ((mode == IFBLK || mode == IFCHR || mode == IFIFO ||
283 	     mode == IFSOCK) && DIP(dp, di_size) != 0) {
284 		if (debug)
285 			printf("bad special-file size %ju:",
286 			    (uintmax_t)DIP(dp, di_size));
287 		goto unknown;
288 	}
289 	if ((mode == IFBLK || mode == IFCHR) &&
290 	    (dev_t)DIP(dp, di_rdev) == NODEV) {
291 		if (debug)
292 			printf("bad special-file rdev NODEV:");
293 		goto unknown;
294 	}
295 	ndb = howmany(DIP(dp, di_size), sblock.fs_bsize);
296 	if (ndb < 0) {
297 		if (debug)
298 			printf("bad size %ju ndb %ju:",
299 				(uintmax_t)DIP(dp, di_size), (uintmax_t)ndb);
300 		goto unknown;
301 	}
302 	if (mode == IFBLK || mode == IFCHR)
303 		ndb++;
304 	if (mode == IFLNK) {
305 		/*
306 		 * Fake ndb value so direct/indirect block checks below
307 		 * will detect any garbage after symlink string.
308 		 */
309 		if (DIP(dp, di_size) < (off_t)sblock.fs_maxsymlinklen) {
310 			if (sblock.fs_magic == FS_UFS1_MAGIC)
311 				ndb = howmany(DIP(dp, di_size),
312 				    sizeof(ufs1_daddr_t));
313 			else
314 				ndb = howmany(DIP(dp, di_size),
315 				    sizeof(ufs2_daddr_t));
316 			if (ndb > NDADDR) {
317 				j = ndb - NDADDR;
318 				for (ndb = 1; j > 1; j--)
319 					ndb *= NINDIR(&sblock);
320 				ndb += NDADDR;
321 			}
322 		}
323 	}
324 	for (j = ndb; ndb < NDADDR && j < NDADDR; j++)
325 		if (DIP(dp, di_db[j]) != 0) {
326 			if (debug)
327 				printf("bad direct addr[%d]: %ju\n", j,
328 				    (uintmax_t)DIP(dp, di_db[j]));
329 			goto unknown;
330 		}
331 	for (j = 0, ndb -= NDADDR; ndb > 0; j++)
332 		ndb /= NINDIR(&sblock);
333 	for (; j < NIADDR; j++)
334 		if (DIP(dp, di_ib[j]) != 0) {
335 			if (debug)
336 				printf("bad indirect addr: %ju\n",
337 				    (uintmax_t)DIP(dp, di_ib[j]));
338 			goto unknown;
339 		}
340 	if (ftypeok(dp) == 0)
341 		goto unknown;
342 	n_files++;
343 	inoinfo(inumber)->ino_linkcnt = DIP(dp, di_nlink);
344 	if (mode == IFDIR) {
345 		if (DIP(dp, di_size) == 0)
346 			inoinfo(inumber)->ino_state = DCLEAR;
347 		else if (DIP(dp, di_nlink) <= 0)
348 			inoinfo(inumber)->ino_state = DZLINK;
349 		else
350 			inoinfo(inumber)->ino_state = DSTATE;
351 		cacheino(dp, inumber);
352 		countdirs++;
353 	} else if (DIP(dp, di_nlink) <= 0)
354 		inoinfo(inumber)->ino_state = FZLINK;
355 	else
356 		inoinfo(inumber)->ino_state = FSTATE;
357 	inoinfo(inumber)->ino_type = IFTODT(mode);
358 	badblk = dupblk = 0;
359 	idesc->id_number = inumber;
360 	if (DIP(dp, di_flags) & SF_SNAPSHOT)
361 		idesc->id_type = SNAP;
362 	else
363 		idesc->id_type = ADDR;
364 	(void)ckinode(dp, idesc);
365 	if (sblock.fs_magic == FS_UFS2_MAGIC && dp->dp2.di_extsize > 0) {
366 		idesc->id_type = ADDR;
367 		ndb = howmany(dp->dp2.di_extsize, sblock.fs_bsize);
368 		for (j = 0; j < NXADDR; j++) {
369 			if (--ndb == 0 &&
370 			    (offset = blkoff(&sblock, dp->dp2.di_extsize)) != 0)
371 				idesc->id_numfrags = numfrags(&sblock,
372 				    fragroundup(&sblock, offset));
373 			else
374 				idesc->id_numfrags = sblock.fs_frag;
375 			if (dp->dp2.di_extb[j] == 0)
376 				continue;
377 			idesc->id_blkno = dp->dp2.di_extb[j];
378 			ret = (*idesc->id_func)(idesc);
379 			if (ret & STOP)
380 				break;
381 		}
382 	}
383 	if (sblock.fs_magic == FS_UFS2_MAGIC)
384 		eascan(idesc, &dp->dp2);
385 	idesc->id_entryno *= btodb(sblock.fs_fsize);
386 	if (DIP(dp, di_blocks) != idesc->id_entryno) {
387 		pwarn("INCORRECT BLOCK COUNT I=%lu (%ju should be %ju)",
388 		    (u_long)inumber, (uintmax_t)DIP(dp, di_blocks),
389 		    (uintmax_t)idesc->id_entryno);
390 		if (preen)
391 			printf(" (CORRECTED)\n");
392 		else if (reply("CORRECT") == 0)
393 			return (1);
394 		if (bkgrdflag == 0) {
395 			dp = ginode(inumber);
396 			DIP_SET(dp, di_blocks, idesc->id_entryno);
397 			inodirty();
398 		} else {
399 			cmd.value = idesc->id_number;
400 			cmd.size = idesc->id_entryno - DIP(dp, di_blocks);
401 			if (debug)
402 				printf("adjblkcnt ino %ju amount %lld\n",
403 				    (uintmax_t)cmd.value, (long long)cmd.size);
404 			if (sysctl(adjblkcnt, MIBSIZE, 0, 0,
405 			    &cmd, sizeof cmd) == -1)
406 				rwerror("ADJUST INODE BLOCK COUNT", cmd.value);
407 		}
408 	}
409 	return (1);
410 unknown:
411 	pfatal("UNKNOWN FILE TYPE I=%lu", (u_long)inumber);
412 	inoinfo(inumber)->ino_state = FCLEAR;
413 	if (reply("CLEAR") == 1) {
414 		inoinfo(inumber)->ino_state = USTATE;
415 		dp = ginode(inumber);
416 		clearinode(dp);
417 		inodirty();
418 	}
419 	return (1);
420 }
421 
422 int
423 pass1check(struct inodesc *idesc)
424 {
425 	int res = KEEPON;
426 	int anyout, nfrags;
427 	ufs2_daddr_t blkno = idesc->id_blkno;
428 	struct dups *dlp;
429 	struct dups *new;
430 
431 	if (idesc->id_type == SNAP) {
432 		if (blkno == BLK_NOCOPY)
433 			return (KEEPON);
434 		if (idesc->id_number == cursnapshot) {
435 			if (blkno == blkstofrags(&sblock, idesc->id_lbn))
436 				return (KEEPON);
437 			if (blkno == BLK_SNAP) {
438 				blkno = blkstofrags(&sblock, idesc->id_lbn);
439 				idesc->id_entryno -= idesc->id_numfrags;
440 			}
441 		} else {
442 			if (blkno == BLK_SNAP)
443 				return (KEEPON);
444 		}
445 	}
446 	if ((anyout = chkrange(blkno, idesc->id_numfrags)) != 0) {
447 		blkerror(idesc->id_number, "BAD", blkno);
448 		if (badblk++ >= MAXBAD) {
449 			pwarn("EXCESSIVE BAD BLKS I=%lu",
450 			    (u_long)idesc->id_number);
451 			if (preen)
452 				printf(" (SKIPPING)\n");
453 			else if (reply("CONTINUE") == 0) {
454 				ckfini(0);
455 				exit(EEXIT);
456 			}
457 			return (STOP);
458 		}
459 	}
460 	for (nfrags = idesc->id_numfrags; nfrags > 0; blkno++, nfrags--) {
461 		if (anyout && chkrange(blkno, 1)) {
462 			res = SKIP;
463 		} else if (!testbmap(blkno)) {
464 			n_blks++;
465 			setbmap(blkno);
466 		} else {
467 			blkerror(idesc->id_number, "DUP", blkno);
468 			if (dupblk++ >= MAXDUP) {
469 				pwarn("EXCESSIVE DUP BLKS I=%lu",
470 					(u_long)idesc->id_number);
471 				if (preen)
472 					printf(" (SKIPPING)\n");
473 				else if (reply("CONTINUE") == 0) {
474 					ckfini(0);
475 					exit(EEXIT);
476 				}
477 				return (STOP);
478 			}
479 			new = (struct dups *)malloc(sizeof(struct dups));
480 			if (new == NULL) {
481 				pfatal("DUP TABLE OVERFLOW.");
482 				if (reply("CONTINUE") == 0) {
483 					ckfini(0);
484 					exit(EEXIT);
485 				}
486 				return (STOP);
487 			}
488 			new->dup = blkno;
489 			if (muldup == 0) {
490 				duplist = muldup = new;
491 				new->next = 0;
492 			} else {
493 				new->next = muldup->next;
494 				muldup->next = new;
495 			}
496 			for (dlp = duplist; dlp != muldup; dlp = dlp->next)
497 				if (dlp->dup == blkno)
498 					break;
499 			if (dlp == muldup && dlp->dup != blkno)
500 				muldup = new;
501 		}
502 		/*
503 		 * count the number of blocks found in id_entryno
504 		 */
505 		idesc->id_entryno++;
506 	}
507 	return (res);
508 }
509