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
32static 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
53static ufs2_daddr_t badblk;
54static ufs2_daddr_t dupblk;
55static ino_t lastino;		/* last inode in use */
56
57static int checkinode(ino_t inumber, struct inodesc *, int rebuildcg);
58
59void
60pass1(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; inosused -= CHAR_BIT, cp--) {
137				if (*cp == 0)
138					continue;
139				for (i = 1 << (CHAR_BIT - 1); i > 0; i >>= 1) {
140					if (*cp & i)
141						break;
142					inosused--;
143				}
144				break;
145			}
146			if (inosused < 0)
147				inosused = 0;
148		}
149		/*
150		 * Allocate inoinfo structures for the allocated inodes.
151		 */
152		inostathead[c].il_numalloced = inosused;
153		if (inosused == 0) {
154			inostathead[c].il_stat = 0;
155			continue;
156		}
157		info = Calloc((unsigned)inosused, sizeof(struct inostat));
158		if (info == NULL)
159			errx(EEXIT, "cannot alloc %u bytes for inoinfo",
160			    (unsigned)(sizeof(struct inostat) * inosused));
161		inostathead[c].il_stat = info;
162		/*
163		 * Scan the allocated inodes.
164		 */
165		for (i = 0; i < inosused; i++, inumber++) {
166			if (inumber < ROOTINO) {
167				(void)getnextinode(inumber, rebuildcg);
168				continue;
169			}
170			/*
171			 * NULL return indicates probable end of allocated
172			 * inodes during cylinder group rebuild attempt.
173			 * We always keep trying until we get to the minimum
174			 * valid number for this cylinder group.
175			 */
176			if (checkinode(inumber, &idesc, rebuildcg) == 0 &&
177			    i > cgp->cg_initediblk)
178				break;
179		}
180		/*
181		 * This optimization speeds up future runs of fsck
182		 * by trimming down the number of inodes in cylinder
183		 * groups that formerly had many inodes but now have
184		 * fewer in use.
185		 */
186		mininos = roundup(inosused + INOPB(&sblock), INOPB(&sblock));
187		if (inoopt && !preen && !rebuildcg &&
188		    sblock.fs_magic == FS_UFS2_MAGIC &&
189		    cgp->cg_initediblk > 2 * INOPB(&sblock) &&
190		    mininos < cgp->cg_initediblk) {
191			i = cgp->cg_initediblk;
192			if (mininos < 2 * INOPB(&sblock))
193				cgp->cg_initediblk = 2 * INOPB(&sblock);
194			else
195				cgp->cg_initediblk = mininos;
196			pwarn("CYLINDER GROUP %d: RESET FROM %ju TO %d %s\n",
197			    c, i, cgp->cg_initediblk, "VALID INODES");
198			dirty(cgbp);
199		}
200		if (inosused < sblock.fs_ipg)
201			continue;
202		lastino += 1;
203		if (lastino < (c * sblock.fs_ipg))
204			inosused = 0;
205		else
206			inosused = lastino - (c * sblock.fs_ipg);
207		if (rebuildcg && inosused > cgp->cg_initediblk &&
208		    sblock.fs_magic == FS_UFS2_MAGIC) {
209			cgp->cg_initediblk = roundup(inosused, INOPB(&sblock));
210			pwarn("CYLINDER GROUP %d: FOUND %d VALID INODES\n", c,
211			    cgp->cg_initediblk);
212		}
213		/*
214		 * If we were not able to determine in advance which inodes
215		 * were in use, then reduce the size of the inoinfo structure
216		 * to the size necessary to describe the inodes that we
217		 * really found.
218		 */
219		if (inumber == lastino)
220			continue;
221		inostathead[c].il_numalloced = inosused;
222		if (inosused == 0) {
223			free(inostathead[c].il_stat);
224			inostathead[c].il_stat = 0;
225			continue;
226		}
227		info = Calloc((unsigned)inosused, sizeof(struct inostat));
228		if (info == NULL)
229			errx(EEXIT, "cannot alloc %u bytes for inoinfo",
230			    (unsigned)(sizeof(struct inostat) * inosused));
231		memmove(info, inostathead[c].il_stat, inosused * sizeof(*info));
232		free(inostathead[c].il_stat);
233		inostathead[c].il_stat = info;
234	}
235	freeinodebuf();
236}
237
238static int
239checkinode(ino_t inumber, struct inodesc *idesc, int rebuildcg)
240{
241	union dinode *dp;
242	off_t kernmaxfilesize;
243	ufs2_daddr_t ndb;
244	mode_t mode;
245	int j, ret, offset;
246
247	if ((dp = getnextinode(inumber, rebuildcg)) == NULL)
248		return (0);
249	mode = DIP(dp, di_mode) & IFMT;
250	if (mode == 0) {
251		if ((sblock.fs_magic == FS_UFS1_MAGIC &&
252		     (memcmp(dp->dp1.di_db, ufs1_zino.di_db,
253			NDADDR * sizeof(ufs1_daddr_t)) ||
254		      memcmp(dp->dp1.di_ib, ufs1_zino.di_ib,
255			NIADDR * sizeof(ufs1_daddr_t)) ||
256		      dp->dp1.di_mode || dp->dp1.di_size)) ||
257		    (sblock.fs_magic == FS_UFS2_MAGIC &&
258		     (memcmp(dp->dp2.di_db, ufs2_zino.di_db,
259			NDADDR * sizeof(ufs2_daddr_t)) ||
260		      memcmp(dp->dp2.di_ib, ufs2_zino.di_ib,
261			NIADDR * sizeof(ufs2_daddr_t)) ||
262		      dp->dp2.di_mode || dp->dp2.di_size))) {
263			pfatal("PARTIALLY ALLOCATED INODE I=%lu",
264			    (u_long)inumber);
265			if (reply("CLEAR") == 1) {
266				dp = ginode(inumber);
267				clearinode(dp);
268				inodirty();
269			}
270		}
271		inoinfo(inumber)->ino_state = USTATE;
272		return (1);
273	}
274	lastino = inumber;
275	/* This should match the file size limit in ffs_mountfs(). */
276	if (sblock.fs_magic == FS_UFS1_MAGIC)
277		kernmaxfilesize = (off_t)0x40000000 * sblock.fs_bsize - 1;
278	else
279		kernmaxfilesize = sblock.fs_maxfilesize;
280	if (DIP(dp, di_size) > kernmaxfilesize ||
281	    DIP(dp, di_size) > sblock.fs_maxfilesize ||
282	    (mode == IFDIR && DIP(dp, di_size) > MAXDIRSIZE)) {
283		if (debug)
284			printf("bad size %ju:", (uintmax_t)DIP(dp, di_size));
285		goto unknown;
286	}
287	if (!preen && mode == IFMT && reply("HOLD BAD BLOCK") == 1) {
288		dp = ginode(inumber);
289		DIP_SET(dp, di_size, sblock.fs_fsize);
290		DIP_SET(dp, di_mode, IFREG|0600);
291		inodirty();
292	}
293	if ((mode == IFBLK || mode == IFCHR || mode == IFIFO ||
294	     mode == IFSOCK) && DIP(dp, di_size) != 0) {
295		if (debug)
296			printf("bad special-file size %ju:",
297			    (uintmax_t)DIP(dp, di_size));
298		goto unknown;
299	}
300	if ((mode == IFBLK || mode == IFCHR) &&
301	    (dev_t)DIP(dp, di_rdev) == NODEV) {
302		if (debug)
303			printf("bad special-file rdev NODEV:");
304		goto unknown;
305	}
306	ndb = howmany(DIP(dp, di_size), sblock.fs_bsize);
307	if (ndb < 0) {
308		if (debug)
309			printf("bad size %ju ndb %ju:",
310				(uintmax_t)DIP(dp, di_size), (uintmax_t)ndb);
311		goto unknown;
312	}
313	if (mode == IFBLK || mode == IFCHR)
314		ndb++;
315	if (mode == IFLNK) {
316		/*
317		 * Fake ndb value so direct/indirect block checks below
318		 * will detect any garbage after symlink string.
319		 */
320		if (DIP(dp, di_size) < (off_t)sblock.fs_maxsymlinklen) {
321			if (sblock.fs_magic == FS_UFS1_MAGIC)
322				ndb = howmany(DIP(dp, di_size),
323				    sizeof(ufs1_daddr_t));
324			else
325				ndb = howmany(DIP(dp, di_size),
326				    sizeof(ufs2_daddr_t));
327			if (ndb > NDADDR) {
328				j = ndb - NDADDR;
329				for (ndb = 1; j > 1; j--)
330					ndb *= NINDIR(&sblock);
331				ndb += NDADDR;
332			}
333		}
334	}
335	for (j = ndb; ndb < NDADDR && j < NDADDR; j++)
336		if (DIP(dp, di_db[j]) != 0) {
337			if (debug)
338				printf("bad direct addr[%d]: %ju\n", j,
339				    (uintmax_t)DIP(dp, di_db[j]));
340			goto unknown;
341		}
342	for (j = 0, ndb -= NDADDR; ndb > 0; j++)
343		ndb /= NINDIR(&sblock);
344	for (; j < NIADDR; j++)
345		if (DIP(dp, di_ib[j]) != 0) {
346			if (debug)
347				printf("bad indirect addr: %ju\n",
348				    (uintmax_t)DIP(dp, di_ib[j]));
349			goto unknown;
350		}
351	if (ftypeok(dp) == 0)
352		goto unknown;
353	n_files++;
354	inoinfo(inumber)->ino_linkcnt = DIP(dp, di_nlink);
355	if (mode == IFDIR) {
356		if (DIP(dp, di_size) == 0)
357			inoinfo(inumber)->ino_state = DCLEAR;
358		else if (DIP(dp, di_nlink) <= 0)
359			inoinfo(inumber)->ino_state = DZLINK;
360		else
361			inoinfo(inumber)->ino_state = DSTATE;
362		cacheino(dp, inumber);
363		countdirs++;
364	} else if (DIP(dp, di_nlink) <= 0)
365		inoinfo(inumber)->ino_state = FZLINK;
366	else
367		inoinfo(inumber)->ino_state = FSTATE;
368	inoinfo(inumber)->ino_type = IFTODT(mode);
369	badblk = dupblk = 0;
370	idesc->id_number = inumber;
371	if (DIP(dp, di_flags) & SF_SNAPSHOT)
372		idesc->id_type = SNAP;
373	else
374		idesc->id_type = ADDR;
375	(void)ckinode(dp, idesc);
376	if (sblock.fs_magic == FS_UFS2_MAGIC && dp->dp2.di_extsize > 0) {
377		idesc->id_type = ADDR;
378		ndb = howmany(dp->dp2.di_extsize, sblock.fs_bsize);
379		for (j = 0; j < NXADDR; j++) {
380			if (--ndb == 0 &&
381			    (offset = blkoff(&sblock, dp->dp2.di_extsize)) != 0)
382				idesc->id_numfrags = numfrags(&sblock,
383				    fragroundup(&sblock, offset));
384			else
385				idesc->id_numfrags = sblock.fs_frag;
386			if (dp->dp2.di_extb[j] == 0)
387				continue;
388			idesc->id_blkno = dp->dp2.di_extb[j];
389			ret = (*idesc->id_func)(idesc);
390			if (ret & STOP)
391				break;
392		}
393	}
394	if (sblock.fs_magic == FS_UFS2_MAGIC)
395		eascan(idesc, &dp->dp2);
396	idesc->id_entryno *= btodb(sblock.fs_fsize);
397	if (DIP(dp, di_blocks) != idesc->id_entryno) {
398		pwarn("INCORRECT BLOCK COUNT I=%lu (%ju should be %ju)",
399		    (u_long)inumber, (uintmax_t)DIP(dp, di_blocks),
400		    (uintmax_t)idesc->id_entryno);
401		if (preen)
402			printf(" (CORRECTED)\n");
403		else if (reply("CORRECT") == 0)
404			return (1);
405		if (bkgrdflag == 0) {
406			dp = ginode(inumber);
407			DIP_SET(dp, di_blocks, idesc->id_entryno);
408			inodirty();
409		} else {
410			cmd.value = idesc->id_number;
411			cmd.size = idesc->id_entryno - DIP(dp, di_blocks);
412			if (debug)
413				printf("adjblkcnt ino %ju amount %lld\n",
414				    (uintmax_t)cmd.value, (long long)cmd.size);
415			if (sysctl(adjblkcnt, MIBSIZE, 0, 0,
416			    &cmd, sizeof cmd) == -1)
417				rwerror("ADJUST INODE BLOCK COUNT", cmd.value);
418		}
419	}
420	return (1);
421unknown:
422	pfatal("UNKNOWN FILE TYPE I=%lu", (u_long)inumber);
423	inoinfo(inumber)->ino_state = FCLEAR;
424	if (reply("CLEAR") == 1) {
425		inoinfo(inumber)->ino_state = USTATE;
426		dp = ginode(inumber);
427		clearinode(dp);
428		inodirty();
429	}
430	return (1);
431}
432
433int
434pass1check(struct inodesc *idesc)
435{
436	int res = KEEPON;
437	int anyout, nfrags;
438	ufs2_daddr_t blkno = idesc->id_blkno;
439	struct dups *dlp;
440	struct dups *new;
441
442	if (idesc->id_type == SNAP) {
443		if (blkno == BLK_NOCOPY)
444			return (KEEPON);
445		if (idesc->id_number == cursnapshot) {
446			if (blkno == blkstofrags(&sblock, idesc->id_lbn))
447				return (KEEPON);
448			if (blkno == BLK_SNAP) {
449				blkno = blkstofrags(&sblock, idesc->id_lbn);
450				idesc->id_entryno -= idesc->id_numfrags;
451			}
452		} else {
453			if (blkno == BLK_SNAP)
454				return (KEEPON);
455		}
456	}
457	if ((anyout = chkrange(blkno, idesc->id_numfrags)) != 0) {
458		blkerror(idesc->id_number, "BAD", blkno);
459		if (badblk++ >= MAXBAD) {
460			pwarn("EXCESSIVE BAD BLKS I=%lu",
461			    (u_long)idesc->id_number);
462			if (preen)
463				printf(" (SKIPPING)\n");
464			else if (reply("CONTINUE") == 0) {
465				ckfini(0);
466				exit(EEXIT);
467			}
468			rerun = 1;
469			return (STOP);
470		}
471	}
472	for (nfrags = idesc->id_numfrags; nfrags > 0; blkno++, nfrags--) {
473		if (anyout && chkrange(blkno, 1)) {
474			res = SKIP;
475		} else if (!testbmap(blkno)) {
476			n_blks++;
477			setbmap(blkno);
478		} else {
479			blkerror(idesc->id_number, "DUP", blkno);
480			if (dupblk++ >= MAXDUP) {
481				pwarn("EXCESSIVE DUP BLKS I=%lu",
482					(u_long)idesc->id_number);
483				if (preen)
484					printf(" (SKIPPING)\n");
485				else if (reply("CONTINUE") == 0) {
486					ckfini(0);
487					exit(EEXIT);
488				}
489				rerun = 1;
490				return (STOP);
491			}
492			new = (struct dups *)Malloc(sizeof(struct dups));
493			if (new == NULL) {
494				pfatal("DUP TABLE OVERFLOW.");
495				if (reply("CONTINUE") == 0) {
496					ckfini(0);
497					exit(EEXIT);
498				}
499				rerun = 1;
500				return (STOP);
501			}
502			new->dup = blkno;
503			if (muldup == 0) {
504				duplist = muldup = new;
505				new->next = 0;
506			} else {
507				new->next = muldup->next;
508				muldup->next = new;
509			}
510			for (dlp = duplist; dlp != muldup; dlp = dlp->next)
511				if (dlp->dup == blkno)
512					break;
513			if (dlp == muldup && dlp->dup != blkno)
514				muldup = new;
515		}
516		/*
517		 * count the number of blocks found in id_entryno
518		 */
519		idesc->id_entryno++;
520	}
521	return (res);
522}
523