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	/*
72	 * Set file system reserved blocks in used block map.
73	 */
74	for (c = 0; c < sblock.fs_ncg; c++) {
75		cgd = cgdmin(&sblock, c);
76		if (c == 0) {
77			i = cgbase(&sblock, c);
78		} else
79			i = cgsblock(&sblock, c);
80		for (; i < cgd; i++)
81			setbmap(i);
82	}
83	i = sblock.fs_csaddr;
84	cgd = i + howmany(sblock.fs_cssize, sblock.fs_fsize);
85	for (; i < cgd; i++)
86		setbmap(i);
87
88	/*
89	 * Find all allocated blocks.
90	 */
91	memset(&idesc, 0, sizeof(struct inodesc));
92	idesc.id_func = pass1check;
93	n_files = n_blks = 0;
94	for (c = 0; c < sblock.fs_ncg; c++) {
95		inumber = c * sblock.fs_ipg;
96		setinodebuf(inumber);
97		cgbp = cgget(c);
98		cgp = cgbp->b_un.b_cg;
99		rebuildcg = 0;
100		if (!check_cgmagic(c, cgbp))
101			rebuildcg = 1;
102		if (!rebuildcg && sblock.fs_magic == FS_UFS2_MAGIC) {
103			inosused = cgp->cg_initediblk;
104			if (inosused > sblock.fs_ipg) {
105				pfatal("%s (%d > %d) %s %d\nReset to %d\n",
106				    "Too many initialized inodes", inosused,
107				    sblock.fs_ipg, "in cylinder group", c,
108				    sblock.fs_ipg);
109				inosused = sblock.fs_ipg;
110			}
111		} else {
112			inosused = sblock.fs_ipg;
113		}
114		if (got_siginfo) {
115			printf("%s: phase 1: cyl group %d of %d (%d%%)\n",
116			    cdevname, c, sblock.fs_ncg,
117			    c * 100 / sblock.fs_ncg);
118			got_siginfo = 0;
119		}
120		if (got_sigalarm) {
121			setproctitle("%s p1 %d%%", cdevname,
122			     c * 100 / sblock.fs_ncg);
123			got_sigalarm = 0;
124		}
125		/*
126		 * If we are using soft updates, then we can trust the
127		 * cylinder group inode allocation maps to tell us which
128		 * inodes are allocated. We will scan the used inode map
129		 * to find the inodes that are really in use, and then
130		 * read only those inodes in from disk.
131		 */
132		if ((preen || inoopt) && usedsoftdep && !rebuildcg) {
133			cp = &cg_inosused(cgp)[(inosused - 1) / CHAR_BIT];
134			for ( ; inosused > 0; inosused -= CHAR_BIT, cp--) {
135				if (*cp == 0)
136					continue;
137				for (i = 1 << (CHAR_BIT - 1); i > 0; i >>= 1) {
138					if (*cp & i)
139						break;
140					inosused--;
141				}
142				break;
143			}
144			if (inosused < 0)
145				inosused = 0;
146		}
147		/*
148		 * Allocate inoinfo structures for the allocated inodes.
149		 */
150		inostathead[c].il_numalloced = inosused;
151		if (inosused == 0) {
152			inostathead[c].il_stat = 0;
153			continue;
154		}
155		info = Calloc((unsigned)inosused, sizeof(struct inostat));
156		if (info == NULL)
157			errx(EEXIT, "cannot alloc %u bytes for inoinfo",
158			    (unsigned)(sizeof(struct inostat) * inosused));
159		inostathead[c].il_stat = info;
160		/*
161		 * Scan the allocated inodes.
162		 */
163		for (i = 0; i < inosused; i++, inumber++) {
164			if (inumber < ROOTINO) {
165				(void)getnextinode(inumber, rebuildcg);
166				continue;
167			}
168			/*
169			 * NULL return indicates probable end of allocated
170			 * inodes during cylinder group rebuild attempt.
171			 * We always keep trying until we get to the minimum
172			 * valid number for this cylinder group.
173			 */
174			if (checkinode(inumber, &idesc, rebuildcg) == 0 &&
175			    i > cgp->cg_initediblk)
176				break;
177		}
178		/*
179		 * This optimization speeds up future runs of fsck
180		 * by trimming down the number of inodes in cylinder
181		 * groups that formerly had many inodes but now have
182		 * fewer in use.
183		 */
184		mininos = roundup(inosused + INOPB(&sblock), INOPB(&sblock));
185		if (inoopt && !preen && !rebuildcg &&
186		    sblock.fs_magic == FS_UFS2_MAGIC &&
187		    cgp->cg_initediblk > 2 * INOPB(&sblock) &&
188		    mininos < cgp->cg_initediblk) {
189			i = cgp->cg_initediblk;
190			if (mininos < 2 * INOPB(&sblock))
191				cgp->cg_initediblk = 2 * INOPB(&sblock);
192			else
193				cgp->cg_initediblk = mininos;
194			pwarn("CYLINDER GROUP %d: RESET FROM %ju TO %d %s\n",
195			    c, i, cgp->cg_initediblk, "VALID INODES");
196			dirty(cgbp);
197		}
198		if (inosused < sblock.fs_ipg)
199			continue;
200		lastino += 1;
201		if (lastino < (c * sblock.fs_ipg))
202			inosused = 0;
203		else
204			inosused = lastino - (c * sblock.fs_ipg);
205		if (rebuildcg && inosused > cgp->cg_initediblk &&
206		    sblock.fs_magic == FS_UFS2_MAGIC) {
207			cgp->cg_initediblk = roundup(inosused, INOPB(&sblock));
208			pwarn("CYLINDER GROUP %d: FOUND %d VALID INODES\n", c,
209			    cgp->cg_initediblk);
210		}
211		/*
212		 * If we were not able to determine in advance which inodes
213		 * were in use, then reduce the size of the inoinfo structure
214		 * to the size necessary to describe the inodes that we
215		 * really found.
216		 */
217		if (inumber == lastino)
218			continue;
219		inostathead[c].il_numalloced = inosused;
220		if (inosused == 0) {
221			free(inostathead[c].il_stat);
222			inostathead[c].il_stat = 0;
223			continue;
224		}
225		info = Calloc((unsigned)inosused, sizeof(struct inostat));
226		if (info == NULL)
227			errx(EEXIT, "cannot alloc %u bytes for inoinfo",
228			    (unsigned)(sizeof(struct inostat) * inosused));
229		memmove(info, inostathead[c].il_stat, inosused * sizeof(*info));
230		free(inostathead[c].il_stat);
231		inostathead[c].il_stat = info;
232	}
233	freeinodebuf();
234}
235
236static int
237checkinode(ino_t inumber, struct inodesc *idesc, int rebuildcg)
238{
239	union dinode *dp;
240	off_t kernmaxfilesize;
241	ufs2_daddr_t ndb;
242	mode_t mode;
243	int j, ret, offset;
244
245	if ((dp = getnextinode(inumber, rebuildcg)) == NULL)
246		return (0);
247	mode = DIP(dp, di_mode) & IFMT;
248	if (mode == 0) {
249		if ((sblock.fs_magic == FS_UFS1_MAGIC &&
250		     (memcmp(dp->dp1.di_db, ufs1_zino.di_db,
251			NDADDR * sizeof(ufs1_daddr_t)) ||
252		      memcmp(dp->dp1.di_ib, ufs1_zino.di_ib,
253			NIADDR * sizeof(ufs1_daddr_t)) ||
254		      dp->dp1.di_mode || dp->dp1.di_size)) ||
255		    (sblock.fs_magic == FS_UFS2_MAGIC &&
256		     (memcmp(dp->dp2.di_db, ufs2_zino.di_db,
257			NDADDR * sizeof(ufs2_daddr_t)) ||
258		      memcmp(dp->dp2.di_ib, ufs2_zino.di_ib,
259			NIADDR * sizeof(ufs2_daddr_t)) ||
260		      dp->dp2.di_mode || dp->dp2.di_size))) {
261			pfatal("PARTIALLY ALLOCATED INODE I=%lu",
262			    (u_long)inumber);
263			if (reply("CLEAR") == 1) {
264				dp = ginode(inumber);
265				clearinode(dp);
266				inodirty();
267			}
268		}
269		inoinfo(inumber)->ino_state = USTATE;
270		return (1);
271	}
272	lastino = inumber;
273	/* This should match the file size limit in ffs_mountfs(). */
274	if (sblock.fs_magic == FS_UFS1_MAGIC)
275		kernmaxfilesize = (off_t)0x40000000 * sblock.fs_bsize - 1;
276	else
277		kernmaxfilesize = sblock.fs_maxfilesize;
278	if (DIP(dp, di_size) > kernmaxfilesize ||
279	    DIP(dp, di_size) > sblock.fs_maxfilesize ||
280	    (mode == IFDIR && DIP(dp, di_size) > MAXDIRSIZE)) {
281		if (debug)
282			printf("bad size %ju:", (uintmax_t)DIP(dp, di_size));
283		goto unknown;
284	}
285	if (!preen && mode == IFMT && reply("HOLD BAD BLOCK") == 1) {
286		dp = ginode(inumber);
287		DIP_SET(dp, di_size, sblock.fs_fsize);
288		DIP_SET(dp, di_mode, IFREG|0600);
289		inodirty();
290	}
291	if ((mode == IFBLK || mode == IFCHR || mode == IFIFO ||
292	     mode == IFSOCK) && DIP(dp, di_size) != 0) {
293		if (debug)
294			printf("bad special-file size %ju:",
295			    (uintmax_t)DIP(dp, di_size));
296		goto unknown;
297	}
298	if ((mode == IFBLK || mode == IFCHR) &&
299	    (dev_t)DIP(dp, di_rdev) == NODEV) {
300		if (debug)
301			printf("bad special-file rdev NODEV:");
302		goto unknown;
303	}
304	ndb = howmany(DIP(dp, di_size), sblock.fs_bsize);
305	if (ndb < 0) {
306		if (debug)
307			printf("bad size %ju ndb %ju:",
308				(uintmax_t)DIP(dp, di_size), (uintmax_t)ndb);
309		goto unknown;
310	}
311	if (mode == IFBLK || mode == IFCHR)
312		ndb++;
313	if (mode == IFLNK) {
314		/*
315		 * Fake ndb value so direct/indirect block checks below
316		 * will detect any garbage after symlink string.
317		 */
318		if (DIP(dp, di_size) < (off_t)sblock.fs_maxsymlinklen) {
319			if (sblock.fs_magic == FS_UFS1_MAGIC)
320				ndb = howmany(DIP(dp, di_size),
321				    sizeof(ufs1_daddr_t));
322			else
323				ndb = howmany(DIP(dp, di_size),
324				    sizeof(ufs2_daddr_t));
325			if (ndb > NDADDR) {
326				j = ndb - NDADDR;
327				for (ndb = 1; j > 1; j--)
328					ndb *= NINDIR(&sblock);
329				ndb += NDADDR;
330			}
331		}
332	}
333	for (j = ndb; ndb < NDADDR && j < NDADDR; j++)
334		if (DIP(dp, di_db[j]) != 0) {
335			if (debug)
336				printf("bad direct addr[%d]: %ju\n", j,
337				    (uintmax_t)DIP(dp, di_db[j]));
338			goto unknown;
339		}
340	for (j = 0, ndb -= NDADDR; ndb > 0; j++)
341		ndb /= NINDIR(&sblock);
342	for (; j < NIADDR; j++)
343		if (DIP(dp, di_ib[j]) != 0) {
344			if (debug)
345				printf("bad indirect addr: %ju\n",
346				    (uintmax_t)DIP(dp, di_ib[j]));
347			goto unknown;
348		}
349	if (ftypeok(dp) == 0)
350		goto unknown;
351	n_files++;
352	inoinfo(inumber)->ino_linkcnt = DIP(dp, di_nlink);
353	if (mode == IFDIR) {
354		if (DIP(dp, di_size) == 0)
355			inoinfo(inumber)->ino_state = DCLEAR;
356		else if (DIP(dp, di_nlink) <= 0)
357			inoinfo(inumber)->ino_state = DZLINK;
358		else
359			inoinfo(inumber)->ino_state = DSTATE;
360		cacheino(dp, inumber);
361		countdirs++;
362	} else if (DIP(dp, di_nlink) <= 0)
363		inoinfo(inumber)->ino_state = FZLINK;
364	else
365		inoinfo(inumber)->ino_state = FSTATE;
366	inoinfo(inumber)->ino_type = IFTODT(mode);
367	badblk = dupblk = 0;
368	idesc->id_number = inumber;
369	if (DIP(dp, di_flags) & SF_SNAPSHOT)
370		idesc->id_type = SNAP;
371	else
372		idesc->id_type = ADDR;
373	(void)ckinode(dp, idesc);
374	if (sblock.fs_magic == FS_UFS2_MAGIC && dp->dp2.di_extsize > 0) {
375		idesc->id_type = ADDR;
376		ndb = howmany(dp->dp2.di_extsize, sblock.fs_bsize);
377		for (j = 0; j < NXADDR; j++) {
378			if (--ndb == 0 &&
379			    (offset = blkoff(&sblock, dp->dp2.di_extsize)) != 0)
380				idesc->id_numfrags = numfrags(&sblock,
381				    fragroundup(&sblock, offset));
382			else
383				idesc->id_numfrags = sblock.fs_frag;
384			if (dp->dp2.di_extb[j] == 0)
385				continue;
386			idesc->id_blkno = dp->dp2.di_extb[j];
387			ret = (*idesc->id_func)(idesc);
388			if (ret & STOP)
389				break;
390		}
391	}
392	if (sblock.fs_magic == FS_UFS2_MAGIC)
393		eascan(idesc, &dp->dp2);
394	idesc->id_entryno *= btodb(sblock.fs_fsize);
395	if (DIP(dp, di_blocks) != idesc->id_entryno) {
396		pwarn("INCORRECT BLOCK COUNT I=%lu (%ju should be %ju)",
397		    (u_long)inumber, (uintmax_t)DIP(dp, di_blocks),
398		    (uintmax_t)idesc->id_entryno);
399		if (preen)
400			printf(" (CORRECTED)\n");
401		else if (reply("CORRECT") == 0)
402			return (1);
403		if (bkgrdflag == 0) {
404			dp = ginode(inumber);
405			DIP_SET(dp, di_blocks, idesc->id_entryno);
406			inodirty();
407		} else {
408			cmd.value = idesc->id_number;
409			cmd.size = idesc->id_entryno - DIP(dp, di_blocks);
410			if (debug)
411				printf("adjblkcnt ino %ju amount %lld\n",
412				    (uintmax_t)cmd.value, (long long)cmd.size);
413			if (sysctl(adjblkcnt, MIBSIZE, 0, 0,
414			    &cmd, sizeof cmd) == -1)
415				rwerror("ADJUST INODE BLOCK COUNT", cmd.value);
416		}
417	}
418	return (1);
419unknown:
420	pfatal("UNKNOWN FILE TYPE I=%lu", (u_long)inumber);
421	inoinfo(inumber)->ino_state = FCLEAR;
422	if (reply("CLEAR") == 1) {
423		inoinfo(inumber)->ino_state = USTATE;
424		dp = ginode(inumber);
425		clearinode(dp);
426		inodirty();
427	}
428	return (1);
429}
430
431int
432pass1check(struct inodesc *idesc)
433{
434	int res = KEEPON;
435	int anyout, nfrags;
436	ufs2_daddr_t blkno = idesc->id_blkno;
437	struct dups *dlp;
438	struct dups *new;
439
440	if (idesc->id_type == SNAP) {
441		if (blkno == BLK_NOCOPY)
442			return (KEEPON);
443		if (idesc->id_number == cursnapshot) {
444			if (blkno == blkstofrags(&sblock, idesc->id_lbn))
445				return (KEEPON);
446			if (blkno == BLK_SNAP) {
447				blkno = blkstofrags(&sblock, idesc->id_lbn);
448				idesc->id_entryno -= idesc->id_numfrags;
449			}
450		} else {
451			if (blkno == BLK_SNAP)
452				return (KEEPON);
453		}
454	}
455	if ((anyout = chkrange(blkno, idesc->id_numfrags)) != 0) {
456		blkerror(idesc->id_number, "BAD", blkno);
457		if (badblk++ >= MAXBAD) {
458			pwarn("EXCESSIVE BAD BLKS I=%lu",
459			    (u_long)idesc->id_number);
460			if (preen)
461				printf(" (SKIPPING)\n");
462			else if (reply("CONTINUE") == 0) {
463				ckfini(0);
464				exit(EEXIT);
465			}
466			return (STOP);
467		}
468	}
469	for (nfrags = idesc->id_numfrags; nfrags > 0; blkno++, nfrags--) {
470		if (anyout && chkrange(blkno, 1)) {
471			res = SKIP;
472		} else if (!testbmap(blkno)) {
473			n_blks++;
474			setbmap(blkno);
475		} else {
476			blkerror(idesc->id_number, "DUP", blkno);
477			if (dupblk++ >= MAXDUP) {
478				pwarn("EXCESSIVE DUP BLKS I=%lu",
479					(u_long)idesc->id_number);
480				if (preen)
481					printf(" (SKIPPING)\n");
482				else if (reply("CONTINUE") == 0) {
483					ckfini(0);
484					exit(EEXIT);
485				}
486				return (STOP);
487			}
488			new = (struct dups *)Malloc(sizeof(struct dups));
489			if (new == NULL) {
490				pfatal("DUP TABLE OVERFLOW.");
491				if (reply("CONTINUE") == 0) {
492					ckfini(0);
493					exit(EEXIT);
494				}
495				return (STOP);
496			}
497			new->dup = blkno;
498			if (muldup == 0) {
499				duplist = muldup = new;
500				new->next = 0;
501			} else {
502				new->next = muldup->next;
503				muldup->next = new;
504			}
505			for (dlp = duplist; dlp != muldup; dlp = dlp->next)
506				if (dlp->dup == blkno)
507					break;
508			if (dlp == muldup && dlp->dup != blkno)
509				muldup = new;
510		}
511		/*
512		 * count the number of blocks found in id_entryno
513		 */
514		idesc->id_entryno++;
515	}
516	return (res);
517}
518