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
34static 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
55static ufs2_daddr_t badblk;
56static ufs2_daddr_t dupblk;
57static ino_t lastino;		/* last inode in use */
58
59static int checkinode(ino_t inumber, struct inodesc *, int rebuildcg);
60
61void
62pass1(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			dirty(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
243static int
244checkinode(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	intmax_t size, fixsize;
251	int j, ret, offset;
252
253	if ((dp = getnextinode(inumber, rebuildcg)) == NULL)
254		goto unknown;
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	(void)ckinode(dp, idesc);
382	if (sblock.fs_magic == FS_UFS2_MAGIC && dp->dp2.di_extsize > 0) {
383		idesc->id_type = ADDR;
384		ndb = howmany(dp->dp2.di_extsize, sblock.fs_bsize);
385		for (j = 0; j < UFS_NXADDR; j++) {
386			if (--ndb == 0 &&
387			    (offset = blkoff(&sblock, dp->dp2.di_extsize)) != 0)
388				idesc->id_numfrags = numfrags(&sblock,
389				    fragroundup(&sblock, offset));
390			else
391				idesc->id_numfrags = sblock.fs_frag;
392			if (dp->dp2.di_extb[j] == 0)
393				continue;
394			idesc->id_blkno = dp->dp2.di_extb[j];
395			ret = (*idesc->id_func)(idesc);
396			if (ret & STOP)
397				break;
398		}
399	}
400	if (sblock.fs_magic == FS_UFS2_MAGIC)
401		eascan(idesc, &dp->dp2);
402	idesc->id_entryno *= btodb(sblock.fs_fsize);
403	if (DIP(dp, di_blocks) != idesc->id_entryno) {
404		pwarn("INCORRECT BLOCK COUNT I=%lu (%ju should be %ju)",
405		    (u_long)inumber, (uintmax_t)DIP(dp, di_blocks),
406		    (uintmax_t)idesc->id_entryno);
407		if (preen)
408			printf(" (CORRECTED)\n");
409		else if (reply("CORRECT") == 0)
410			return (1);
411		if (bkgrdflag == 0) {
412			dp = ginode(inumber);
413			DIP_SET(dp, di_blocks, idesc->id_entryno);
414			inodirty(dp);
415		} else {
416			cmd.value = idesc->id_number;
417			cmd.size = idesc->id_entryno - DIP(dp, di_blocks);
418			if (debug)
419				printf("adjblkcnt ino %ju amount %lld\n",
420				    (uintmax_t)cmd.value, (long long)cmd.size);
421			if (sysctl(adjblkcnt, MIBSIZE, 0, 0,
422			    &cmd, sizeof cmd) == -1)
423				rwerror("ADJUST INODE BLOCK COUNT", cmd.value);
424		}
425	}
426	/*
427	 * UFS does not allow files to end with a hole; it requires that
428	 * the last block of a file be allocated. The last allocated block
429	 * in a file is tracked in id_lballoc. Here, we check for a size
430	 * past the last allocated block of the file and if that is found,
431	 * shorten the file to reference the last allocated block to avoid
432	 * having it reference a hole at its end.
433	 *
434	 * Soft updates will always ensure that the file size is correct
435	 * for files that contain only direct block pointers. However
436	 * soft updates does not roll back sizes for files with indirect
437	 * blocks that it has set to unallocated because their contents
438	 * have not yet been written to disk. Hence, the file can appear
439	 * to have a hole at its end because the block pointer has been
440	 * rolled back to zero. Thus finding a hole at the end of a file
441	 * that is located in an indirect block receives only a warning
442	 * while finding a hole at the end of a file in a direct block
443	 * receives a fatal error message.
444	 */
445	size = DIP(dp, di_size);
446	if (idesc->id_lballoc < lblkno(&sblock, size - 1) &&
447	    /* exclude embedded symbolic links */
448	    ((mode != IFLNK) || size >= sblock.fs_maxsymlinklen)) {
449 		fixsize = lblktosize(&sblock, idesc->id_lballoc + 1);
450		if (size > UFS_NDADDR * sblock.fs_bsize)
451			pwarn("INODE %lu: FILE SIZE %ju BEYOND END OF "
452			      "ALLOCATED FILE, SIZE SHOULD BE %ju",
453			      (u_long)inumber, size, fixsize);
454		else
455			pfatal("INODE %lu: FILE SIZE %ju BEYOND END OF "
456			      "ALLOCATED FILE, SIZE SHOULD BE %ju",
457			      (u_long)inumber, size, fixsize);
458		if (preen)
459			printf(" (ADJUSTED)\n");
460		else if (reply("ADJUST") == 0)
461			return (1);
462		if (bkgrdflag == 0) {
463			dp = ginode(inumber);
464			DIP_SET(dp, di_size, fixsize);
465			inodirty(dp);
466		} else {
467			cmd.value = idesc->id_number;
468			cmd.size = fixsize;
469			if (debug)
470				printf("setsize ino %ju size set to %ju\n",
471				    (uintmax_t)cmd.value, (uintmax_t)cmd.size);
472			if (sysctl(setsize, MIBSIZE, 0, 0,
473			    &cmd, sizeof cmd) == -1)
474				rwerror("SET INODE SIZE", cmd.value);
475		}
476
477	}
478	return (1);
479unknown:
480	pfatal("UNKNOWN FILE TYPE I=%lu", (u_long)inumber);
481	inoinfo(inumber)->ino_state = FCLEAR;
482	if (reply("CLEAR") == 1) {
483		inoinfo(inumber)->ino_state = USTATE;
484		dp = ginode(inumber);
485		clearinode(dp);
486		inodirty(dp);
487	}
488	return (1);
489}
490
491int
492pass1check(struct inodesc *idesc)
493{
494	int res = KEEPON;
495	int anyout, nfrags;
496	ufs2_daddr_t blkno = idesc->id_blkno;
497	struct dups *dlp;
498	struct dups *new;
499
500	if (idesc->id_type == SNAP) {
501		if (blkno == BLK_NOCOPY)
502			return (KEEPON);
503		if (idesc->id_number == cursnapshot) {
504			if (blkno == blkstofrags(&sblock, idesc->id_lbn))
505				return (KEEPON);
506			if (blkno == BLK_SNAP) {
507				blkno = blkstofrags(&sblock, idesc->id_lbn);
508				idesc->id_entryno -= idesc->id_numfrags;
509			}
510		} else {
511			if (blkno == BLK_SNAP)
512				return (KEEPON);
513		}
514	}
515	if ((anyout = chkrange(blkno, idesc->id_numfrags)) != 0) {
516		blkerror(idesc->id_number, "BAD", blkno);
517		if (badblk++ >= MAXBAD) {
518			pwarn("EXCESSIVE BAD BLKS I=%lu",
519			    (u_long)idesc->id_number);
520			if (preen)
521				printf(" (SKIPPING)\n");
522			else if (reply("CONTINUE") == 0) {
523				ckfini(0);
524				exit(EEXIT);
525			}
526			rerun = 1;
527			return (STOP);
528		}
529	}
530	for (nfrags = idesc->id_numfrags; nfrags > 0; blkno++, nfrags--) {
531		if (anyout && chkrange(blkno, 1)) {
532			res = SKIP;
533		} else if (!testbmap(blkno)) {
534			n_blks++;
535			setbmap(blkno);
536		} else {
537			blkerror(idesc->id_number, "DUP", blkno);
538			if (dupblk++ >= MAXDUP) {
539				pwarn("EXCESSIVE DUP BLKS I=%lu",
540					(u_long)idesc->id_number);
541				if (preen)
542					printf(" (SKIPPING)\n");
543				else if (reply("CONTINUE") == 0) {
544					ckfini(0);
545					exit(EEXIT);
546				}
547				rerun = 1;
548				return (STOP);
549			}
550			new = (struct dups *)Malloc(sizeof(struct dups));
551			if (new == NULL) {
552				pfatal("DUP TABLE OVERFLOW.");
553				if (reply("CONTINUE") == 0) {
554					ckfini(0);
555					exit(EEXIT);
556				}
557				rerun = 1;
558				return (STOP);
559			}
560			new->dup = blkno;
561			if (muldup == NULL) {
562				duplist = muldup = new;
563				new->next = NULL;
564			} else {
565				new->next = muldup->next;
566				muldup->next = new;
567			}
568			for (dlp = duplist; dlp != muldup; dlp = dlp->next)
569				if (dlp->dup == blkno)
570					break;
571			if (dlp == muldup && dlp->dup != blkno)
572				muldup = new;
573		}
574		/*
575		 * count the number of blocks found in id_entryno
576		 */
577		idesc->id_entryno++;
578	}
579	if (idesc->id_level == 0 && idesc->id_lballoc < idesc->id_lbn)
580		idesc->id_lballoc = idesc->id_lbn;
581	return (res);
582}
583