inode.c revision 128073 
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[] = "@(#)inode.c	8.8 (Berkeley) 4/28/95";
33#endif /* not lint */
34#endif
35#include <sys/cdefs.h>
36__FBSDID("$FreeBSD: head/sbin/fsck_ffs/inode.c 128073 2004-04-09 19:58:40Z markm $");
37
38#include <sys/param.h>
39#include <sys/stdint.h>
40#include <sys/time.h>
41#include <sys/sysctl.h>
42
43#include <ufs/ufs/dinode.h>
44#include <ufs/ufs/dir.h>
45#include <ufs/ffs/fs.h>
46
47#include <err.h>
48#include <pwd.h>
49#include <string.h>
50
51#include "fsck.h"
52
53static ino_t startinum;
54
55static int iblock(struct inodesc *, long ilevel, off_t isize);
56
57int
58ckinode(union dinode *dp, struct inodesc *idesc)
59{
60	off_t remsize, sizepb;
61	int i, offset, ret;
62	union dinode dino;
63	ufs2_daddr_t ndb;
64	mode_t mode;
65	char pathbuf[MAXPATHLEN + 1];
66
67	if (idesc->id_fix != IGNORE)
68		idesc->id_fix = DONTKNOW;
69	idesc->id_lbn = -1;
70	idesc->id_entryno = 0;
71	idesc->id_filesize = DIP(dp, di_size);
72	mode = DIP(dp, di_mode) & IFMT;
73	if (mode == IFBLK || mode == IFCHR || (mode == IFLNK &&
74	    DIP(dp, di_size) < (unsigned)sblock.fs_maxsymlinklen))
75		return (KEEPON);
76	if (sblock.fs_magic == FS_UFS1_MAGIC)
77		dino.dp1 = dp->dp1;
78	else
79		dino.dp2 = dp->dp2;
80	ndb = howmany(DIP(&dino, di_size), sblock.fs_bsize);
81	for (i = 0; i < NDADDR; i++) {
82		idesc->id_lbn++;
83		if (--ndb == 0 &&
84		    (offset = blkoff(&sblock, DIP(&dino, di_size))) != 0)
85			idesc->id_numfrags =
86				numfrags(&sblock, fragroundup(&sblock, offset));
87		else
88			idesc->id_numfrags = sblock.fs_frag;
89		if (DIP(&dino, di_db[i]) == 0) {
90			if (idesc->id_type == DATA && ndb >= 0) {
91				/* An empty block in a directory XXX */
92				getpathname(pathbuf, idesc->id_number,
93						idesc->id_number);
94                        	pfatal("DIRECTORY %s: CONTAINS EMPTY BLOCKS",
95					pathbuf);
96                        	if (reply("ADJUST LENGTH") == 1) {
97					dp = ginode(idesc->id_number);
98                                	DIP(dp, di_size) = i * sblock.fs_bsize;
99					printf(
100					    "YOU MUST RERUN FSCK AFTERWARDS\n");
101					rerun = 1;
102                                	inodirty();
103
104                        	}
105			}
106			continue;
107		}
108		idesc->id_blkno = DIP(&dino, di_db[i]);
109		if (idesc->id_type != DATA)
110			ret = (*idesc->id_func)(idesc);
111		else
112			ret = dirscan(idesc);
113		if (ret & STOP)
114			return (ret);
115	}
116	idesc->id_numfrags = sblock.fs_frag;
117	remsize = DIP(&dino, di_size) - sblock.fs_bsize * NDADDR;
118	sizepb = sblock.fs_bsize;
119	for (i = 0; i < NIADDR; i++) {
120		sizepb *= NINDIR(&sblock);
121		if (DIP(&dino, di_ib[i])) {
122			idesc->id_blkno = DIP(&dino, di_ib[i]);
123			ret = iblock(idesc, i + 1, remsize);
124			if (ret & STOP)
125				return (ret);
126		} else {
127			idesc->id_lbn += sizepb / sblock.fs_bsize;
128			if (idesc->id_type == DATA && remsize > 0) {
129				/* An empty block in a directory XXX */
130				getpathname(pathbuf, idesc->id_number,
131						idesc->id_number);
132                        	pfatal("DIRECTORY %s: CONTAINS EMPTY BLOCKS",
133					pathbuf);
134                        	if (reply("ADJUST LENGTH") == 1) {
135					dp = ginode(idesc->id_number);
136                                	DIP(dp, di_size) -= remsize;
137					remsize = 0;
138					printf(
139					    "YOU MUST RERUN FSCK AFTERWARDS\n");
140					rerun = 1;
141                                	inodirty();
142					break;
143                        	}
144			}
145		}
146		remsize -= sizepb;
147	}
148	return (KEEPON);
149}
150
151static int
152iblock(struct inodesc *idesc, long ilevel, off_t isize)
153{
154	struct bufarea *bp;
155	int i, n, (*func)(struct inodesc *), nif;
156	off_t sizepb;
157	char buf[BUFSIZ];
158	char pathbuf[MAXPATHLEN + 1];
159	union dinode *dp;
160
161	if (idesc->id_type != DATA) {
162		func = idesc->id_func;
163		if (((n = (*func)(idesc)) & KEEPON) == 0)
164			return (n);
165	} else
166		func = dirscan;
167	if (chkrange(idesc->id_blkno, idesc->id_numfrags))
168		return (SKIP);
169	bp = getdatablk(idesc->id_blkno, sblock.fs_bsize);
170	ilevel--;
171	for (sizepb = sblock.fs_bsize, i = 0; i < ilevel; i++)
172		sizepb *= NINDIR(&sblock);
173	if (howmany(isize, sizepb) > NINDIR(&sblock))
174		nif = NINDIR(&sblock);
175	else
176		nif = howmany(isize, sizepb);
177	if (idesc->id_func == pass1check && nif < NINDIR(&sblock)) {
178		for (i = nif; i < NINDIR(&sblock); i++) {
179			if (IBLK(bp, i) == 0)
180				continue;
181			(void)sprintf(buf, "PARTIALLY TRUNCATED INODE I=%lu",
182			    (u_long)idesc->id_number);
183			if (preen) {
184				pfatal("%s", buf);
185			} else if (dofix(idesc, buf)) {
186				IBLK(bp, i) = 0;
187				dirty(bp);
188			}
189		}
190		flush(fswritefd, bp);
191	}
192	for (i = 0; i < nif; i++) {
193		if (ilevel == 0)
194			idesc->id_lbn++;
195		if (IBLK(bp, i)) {
196			idesc->id_blkno = IBLK(bp, i);
197			if (ilevel == 0)
198				n = (*func)(idesc);
199			else
200				n = iblock(idesc, ilevel, isize);
201			if (n & STOP) {
202				bp->b_flags &= ~B_INUSE;
203				return (n);
204			}
205		} else {
206			if (idesc->id_type == DATA && isize > 0) {
207				/* An empty block in a directory XXX */
208				getpathname(pathbuf, idesc->id_number,
209						idesc->id_number);
210                        	pfatal("DIRECTORY %s: CONTAINS EMPTY BLOCKS",
211					pathbuf);
212                        	if (reply("ADJUST LENGTH") == 1) {
213					dp = ginode(idesc->id_number);
214                                	DIP(dp, di_size) -= isize;
215					isize = 0;
216					printf(
217					    "YOU MUST RERUN FSCK AFTERWARDS\n");
218					rerun = 1;
219                                	inodirty();
220					bp->b_flags &= ~B_INUSE;
221					return(STOP);
222                        	}
223			}
224		}
225		isize -= sizepb;
226	}
227	bp->b_flags &= ~B_INUSE;
228	return (KEEPON);
229}
230
231/*
232 * Check that a block in a legal block number.
233 * Return 0 if in range, 1 if out of range.
234 */
235int
236chkrange(ufs2_daddr_t blk, int cnt)
237{
238	int c;
239
240	if (cnt <= 0 || blk <= 0 || blk > maxfsblock ||
241	    cnt - 1 > maxfsblock - blk)
242		return (1);
243	if (cnt > sblock.fs_frag ||
244	    fragnum(&sblock, blk) + cnt > sblock.fs_frag) {
245		if (debug)
246			printf("bad size: blk %ld, offset %i, size %d\n",
247			    (long)blk, (int)fragnum(&sblock, blk), cnt);
248		return (1);
249	}
250	c = dtog(&sblock, blk);
251	if (blk < cgdmin(&sblock, c)) {
252		if ((blk + cnt) > cgsblock(&sblock, c)) {
253			if (debug) {
254				printf("blk %ld < cgdmin %ld;",
255				    (long)blk, (long)cgdmin(&sblock, c));
256				printf(" blk + cnt %ld > cgsbase %ld\n",
257				    (long)(blk + cnt),
258				    (long)cgsblock(&sblock, c));
259			}
260			return (1);
261		}
262	} else {
263		if ((blk + cnt) > cgbase(&sblock, c+1)) {
264			if (debug)  {
265				printf("blk %ld >= cgdmin %ld;",
266				    (long)blk, (long)cgdmin(&sblock, c));
267				printf(" blk + cnt %ld > sblock.fs_fpg %ld\n",
268				    (long)(blk + cnt), (long)sblock.fs_fpg);
269			}
270			return (1);
271		}
272	}
273	return (0);
274}
275
276/*
277 * General purpose interface for reading inodes.
278 */
279union dinode *
280ginode(ino_t inumber)
281{
282	ufs2_daddr_t iblk;
283
284	if (inumber < ROOTINO || inumber > maxino)
285		errx(EEXIT, "bad inode number %d to ginode", inumber);
286	if (startinum == 0 ||
287	    inumber < startinum || inumber >= startinum + INOPB(&sblock)) {
288		iblk = ino_to_fsba(&sblock, inumber);
289		if (pbp != 0)
290			pbp->b_flags &= ~B_INUSE;
291		pbp = getdatablk(iblk, sblock.fs_bsize);
292		startinum = (inumber / INOPB(&sblock)) * INOPB(&sblock);
293	}
294	if (sblock.fs_magic == FS_UFS1_MAGIC)
295		return ((union dinode *)
296		    &pbp->b_un.b_dinode1[inumber % INOPB(&sblock)]);
297	return ((union dinode *)&pbp->b_un.b_dinode2[inumber % INOPB(&sblock)]);
298}
299
300/*
301 * Special purpose version of ginode used to optimize first pass
302 * over all the inodes in numerical order.
303 */
304static ino_t nextino, lastinum, lastvalidinum;
305static long readcnt, readpercg, fullcnt, inobufsize, partialcnt, partialsize;
306static caddr_t inodebuf;
307
308union dinode *
309getnextinode(ino_t inumber)
310{
311	long size;
312	ufs2_daddr_t dblk;
313	union dinode *dp;
314	static caddr_t nextinop;
315
316	if (inumber != nextino++ || inumber > lastvalidinum)
317		errx(EEXIT, "bad inode number %d to nextinode", inumber);
318	if (inumber >= lastinum) {
319		readcnt++;
320		dblk = fsbtodb(&sblock, ino_to_fsba(&sblock, lastinum));
321		if (readcnt % readpercg == 0) {
322			size = partialsize;
323			lastinum += partialcnt;
324		} else {
325			size = inobufsize;
326			lastinum += fullcnt;
327		}
328		/*
329		 * If bread returns an error, it will already have zeroed
330		 * out the buffer, so we do not need to do so here.
331		 */
332		(void)bread(fsreadfd, inodebuf, dblk, size);
333		nextinop = inodebuf;
334	}
335	dp = (union dinode *)nextinop;
336	if (sblock.fs_magic == FS_UFS1_MAGIC)
337		nextinop += sizeof(struct ufs1_dinode);
338	else
339		nextinop += sizeof(struct ufs2_dinode);
340	return (dp);
341}
342
343void
344setinodebuf(ino_t inum)
345{
346
347	if (inum % sblock.fs_ipg != 0)
348		errx(EEXIT, "bad inode number %d to setinodebuf", inum);
349	lastvalidinum = inum + sblock.fs_ipg - 1;
350	startinum = 0;
351	nextino = inum;
352	lastinum = inum;
353	readcnt = 0;
354	if (inodebuf != NULL)
355		return;
356	inobufsize = blkroundup(&sblock, INOBUFSIZE);
357	fullcnt = inobufsize / ((sblock.fs_magic == FS_UFS1_MAGIC) ?
358	    sizeof(struct ufs1_dinode) : sizeof(struct ufs2_dinode));
359	readpercg = sblock.fs_ipg / fullcnt;
360	partialcnt = sblock.fs_ipg % fullcnt;
361	partialsize = partialcnt * ((sblock.fs_magic == FS_UFS1_MAGIC) ?
362	    sizeof(struct ufs1_dinode) : sizeof(struct ufs2_dinode));
363	if (partialcnt != 0) {
364		readpercg++;
365	} else {
366		partialcnt = fullcnt;
367		partialsize = inobufsize;
368	}
369	if ((inodebuf = malloc((unsigned)inobufsize)) == NULL)
370		errx(EEXIT, "cannot allocate space for inode buffer");
371}
372
373void
374freeinodebuf(void)
375{
376
377	if (inodebuf != NULL)
378		free((char *)inodebuf);
379	inodebuf = NULL;
380}
381
382/*
383 * Routines to maintain information about directory inodes.
384 * This is built during the first pass and used during the
385 * second and third passes.
386 *
387 * Enter inodes into the cache.
388 */
389void
390cacheino(union dinode *dp, ino_t inumber)
391{
392	struct inoinfo *inp, **inpp;
393	int i, blks;
394
395	if (howmany(DIP(dp, di_size), sblock.fs_bsize) > NDADDR)
396		blks = NDADDR + NIADDR;
397	else
398		blks = howmany(DIP(dp, di_size), sblock.fs_bsize);
399	inp = (struct inoinfo *)
400		malloc(sizeof(*inp) + (blks - 1) * sizeof(ufs2_daddr_t));
401	if (inp == NULL)
402		errx(EEXIT, "cannot increase directory list");
403	inpp = &inphead[inumber % dirhash];
404	inp->i_nexthash = *inpp;
405	*inpp = inp;
406	inp->i_parent = inumber == ROOTINO ? ROOTINO : (ino_t)0;
407	inp->i_dotdot = (ino_t)0;
408	inp->i_number = inumber;
409	inp->i_isize = DIP(dp, di_size);
410	inp->i_numblks = blks;
411	for (i = 0; i < (blks < NDADDR ? blks : NDADDR); i++)
412		inp->i_blks[i] = DIP(dp, di_db[i]);
413	if (blks > NDADDR)
414		for (i = 0; i < NIADDR; i++)
415			inp->i_blks[NDADDR + i] = DIP(dp, di_ib[i]);
416	if (inplast == listmax) {
417		listmax += 100;
418		inpsort = (struct inoinfo **)realloc((char *)inpsort,
419		    (unsigned)listmax * sizeof(struct inoinfo *));
420		if (inpsort == NULL)
421			errx(EEXIT, "cannot increase directory list");
422	}
423	inpsort[inplast++] = inp;
424}
425
426/*
427 * Look up an inode cache structure.
428 */
429struct inoinfo *
430getinoinfo(ino_t inumber)
431{
432	struct inoinfo *inp;
433
434	for (inp = inphead[inumber % dirhash]; inp; inp = inp->i_nexthash) {
435		if (inp->i_number != inumber)
436			continue;
437		return (inp);
438	}
439	errx(EEXIT, "cannot find inode %d", inumber);
440	return ((struct inoinfo *)0);
441}
442
443/*
444 * Clean up all the inode cache structure.
445 */
446void
447inocleanup(void)
448{
449	struct inoinfo **inpp;
450
451	if (inphead == NULL)
452		return;
453	for (inpp = &inpsort[inplast - 1]; inpp >= inpsort; inpp--)
454		free((char *)(*inpp));
455	free((char *)inphead);
456	free((char *)inpsort);
457	inphead = inpsort = NULL;
458}
459
460void
461inodirty(void)
462{
463
464	dirty(pbp);
465}
466
467void
468clri(struct inodesc *idesc, const char *type, int flag)
469{
470	union dinode *dp;
471
472	dp = ginode(idesc->id_number);
473	if (flag == 1) {
474		pwarn("%s %s", type,
475		    (DIP(dp, di_mode) & IFMT) == IFDIR ? "DIR" : "FILE");
476		pinode(idesc->id_number);
477	}
478	if (preen || reply("CLEAR") == 1) {
479		if (preen)
480			printf(" (CLEARED)\n");
481		n_files--;
482		if (bkgrdflag == 0) {
483			(void)ckinode(dp, idesc);
484			inoinfo(idesc->id_number)->ino_state = USTATE;
485			clearinode(dp);
486			inodirty();
487		} else {
488			cmd.value = idesc->id_number;
489			cmd.size = -DIP(dp, di_nlink);
490			if (debug)
491				printf("adjrefcnt ino %ld amt %lld\n",
492				    (long)cmd.value, (long long)cmd.size);
493			if (sysctl(adjrefcnt, MIBSIZE, 0, 0,
494			    &cmd, sizeof cmd) == -1)
495				rwerror("ADJUST INODE", cmd.value);
496		}
497	}
498}
499
500int
501findname(struct inodesc *idesc)
502{
503	struct direct *dirp = idesc->id_dirp;
504
505	if (dirp->d_ino != idesc->id_parent || idesc->id_entryno < 2) {
506		idesc->id_entryno++;
507		return (KEEPON);
508	}
509	memmove(idesc->id_name, dirp->d_name, (size_t)dirp->d_namlen + 1);
510	return (STOP|FOUND);
511}
512
513int
514findino(struct inodesc *idesc)
515{
516	struct direct *dirp = idesc->id_dirp;
517
518	if (dirp->d_ino == 0)
519		return (KEEPON);
520	if (strcmp(dirp->d_name, idesc->id_name) == 0 &&
521	    dirp->d_ino >= ROOTINO && dirp->d_ino <= maxino) {
522		idesc->id_parent = dirp->d_ino;
523		return (STOP|FOUND);
524	}
525	return (KEEPON);
526}
527
528int
529clearentry(struct inodesc *idesc)
530{
531	struct direct *dirp = idesc->id_dirp;
532
533	if (dirp->d_ino != idesc->id_parent || idesc->id_entryno < 2) {
534		idesc->id_entryno++;
535		return (KEEPON);
536	}
537	dirp->d_ino = 0;
538	return (STOP|FOUND|ALTERED);
539}
540
541void
542pinode(ino_t ino)
543{
544	union dinode *dp;
545	char *p;
546	struct passwd *pw;
547	time_t t;
548
549	printf(" I=%lu ", (u_long)ino);
550	if (ino < ROOTINO || ino > maxino)
551		return;
552	dp = ginode(ino);
553	printf(" OWNER=");
554	if ((pw = getpwuid((int)DIP(dp, di_uid))) != 0)
555		printf("%s ", pw->pw_name);
556	else
557		printf("%u ", (unsigned)DIP(dp, di_uid));
558	printf("MODE=%o\n", DIP(dp, di_mode));
559	if (preen)
560		printf("%s: ", cdevname);
561	printf("SIZE=%ju ", (uintmax_t)DIP(dp, di_size));
562	t = DIP(dp, di_mtime);
563	p = ctime(&t);
564	printf("MTIME=%12.12s %4.4s ", &p[4], &p[20]);
565}
566
567void
568blkerror(ino_t ino, const char *type, ufs2_daddr_t blk)
569{
570
571	pfatal("%jd %s I=%ju", (intmax_t)blk, type, (uintmax_t)ino);
572	printf("\n");
573	switch (inoinfo(ino)->ino_state) {
574
575	case FSTATE:
576		inoinfo(ino)->ino_state = FCLEAR;
577		return;
578
579	case DSTATE:
580		inoinfo(ino)->ino_state = DCLEAR;
581		return;
582
583	case FCLEAR:
584	case DCLEAR:
585		return;
586
587	default:
588		errx(EEXIT, "BAD STATE %d TO BLKERR", inoinfo(ino)->ino_state);
589		/* NOTREACHED */
590	}
591}
592
593/*
594 * allocate an unused inode
595 */
596ino_t
597allocino(ino_t request, int type)
598{
599	ino_t ino;
600	union dinode *dp;
601	struct cg *cgp = &cgrp;
602	int cg;
603
604	if (request == 0)
605		request = ROOTINO;
606	else if (inoinfo(request)->ino_state != USTATE)
607		return (0);
608	for (ino = request; ino < maxino; ino++)
609		if (inoinfo(ino)->ino_state == USTATE)
610			break;
611	if (ino == maxino)
612		return (0);
613	cg = ino_to_cg(&sblock, ino);
614	getblk(&cgblk, cgtod(&sblock, cg), sblock.fs_cgsize);
615	if (!cg_chkmagic(cgp))
616		pfatal("CG %d: BAD MAGIC NUMBER\n", cg);
617	setbit(cg_inosused(cgp), ino % sblock.fs_ipg);
618	cgp->cg_cs.cs_nifree--;
619	switch (type & IFMT) {
620	case IFDIR:
621		inoinfo(ino)->ino_state = DSTATE;
622		cgp->cg_cs.cs_ndir++;
623		break;
624	case IFREG:
625	case IFLNK:
626		inoinfo(ino)->ino_state = FSTATE;
627		break;
628	default:
629		return (0);
630	}
631	cgdirty();
632	dp = ginode(ino);
633	DIP(dp, di_db[0]) = allocblk((long)1);
634	if (DIP(dp, di_db[0]) == 0) {
635		inoinfo(ino)->ino_state = USTATE;
636		return (0);
637	}
638	DIP(dp, di_mode) = type;
639	DIP(dp, di_flags) = 0;
640	DIP(dp, di_atime) = time(NULL);
641	DIP(dp, di_mtime) = DIP(dp, di_ctime) = DIP(dp, di_atime);
642	DIP(dp, di_mtimensec) = 0;
643	DIP(dp, di_ctimensec) = 0;
644	DIP(dp, di_atimensec) = 0;
645	DIP(dp, di_size) = sblock.fs_fsize;
646	DIP(dp, di_blocks) = btodb(sblock.fs_fsize);
647	n_files++;
648	inodirty();
649	inoinfo(ino)->ino_type = IFTODT(type);
650	return (ino);
651}
652
653/*
654 * deallocate an inode
655 */
656void
657freeino(ino_t ino)
658{
659	struct inodesc idesc;
660	union dinode *dp;
661
662	memset(&idesc, 0, sizeof(struct inodesc));
663	idesc.id_type = ADDR;
664	idesc.id_func = pass4check;
665	idesc.id_number = ino;
666	dp = ginode(ino);
667	(void)ckinode(dp, &idesc);
668	clearinode(dp);
669	inodirty();
670	inoinfo(ino)->ino_state = USTATE;
671	n_files--;
672}
673