inode.c revision 41474 
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 * 3. All advertising materials mentioning features or use of this software
14 *    must display the following acknowledgement:
15 *	This product includes software developed by the University of
16 *	California, Berkeley and its contributors.
17 * 4. Neither the name of the University nor the names of its contributors
18 *    may be used to endorse or promote products derived from this software
19 *    without specific prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 */
33
34#ifndef lint
35static const char sccsid[] = "@(#)inode.c	8.8 (Berkeley) 4/28/95";
36#endif /* not lint */
37
38#include <sys/param.h>
39#include <sys/time.h>
40
41#include <ufs/ufs/dinode.h>
42#include <ufs/ufs/dir.h>
43#include <ufs/ffs/fs.h>
44
45#include <err.h>
46#include <pwd.h>
47#include <string.h>
48
49#include "fsck.h"
50
51static ino_t startinum;
52
53static int iblock __P((struct inodesc *, long ilevel, quad_t isize));
54
55int
56ckinode(dp, idesc)
57	struct dinode *dp;
58	register struct inodesc *idesc;
59{
60	ufs_daddr_t *ap;
61	int ret;
62	long n, ndb, offset;
63	struct dinode dino;
64	quad_t remsize, sizepb;
65	mode_t mode;
66	char pathbuf[MAXPATHLEN + 1];
67
68	if (idesc->id_fix != IGNORE)
69		idesc->id_fix = DONTKNOW;
70	idesc->id_entryno = 0;
71	idesc->id_filesize = dp->di_size;
72	mode = dp->di_mode & IFMT;
73	if (mode == IFBLK || mode == IFCHR || (mode == IFLNK &&
74	    dp->di_size < (unsigned)sblock.fs_maxsymlinklen))
75		return (KEEPON);
76	dino = *dp;
77	ndb = howmany(dino.di_size, sblock.fs_bsize);
78	for (ap = &dino.di_db[0]; ap < &dino.di_db[NDADDR]; ap++) {
79		if (--ndb == 0 && (offset = blkoff(&sblock, dino.di_size)) != 0)
80			idesc->id_numfrags =
81				numfrags(&sblock, fragroundup(&sblock, offset));
82		else
83			idesc->id_numfrags = sblock.fs_frag;
84		if (*ap == 0) {
85			if (idesc->id_type == DATA && ndb >= 0) {
86				/* An empty block in a directory XXX */
87				getpathname(pathbuf, idesc->id_number,
88						idesc->id_number);
89                        	pfatal("DIRECTORY %s: CONTAINS EMPTY BLOCKS",
90					pathbuf);
91                        	if (reply("ADJUST LENGTH") == 1) {
92					dp = ginode(idesc->id_number);
93                                	dp->di_size = (ap - &dino.di_db[0]) *
94					    sblock.fs_bsize;
95					printf(
96					    "YOU MUST RERUN FSCK AFTERWARDS\n");
97					rerun = 1;
98                                	inodirty();
99
100                        	}
101			}
102			continue;
103		}
104		idesc->id_blkno = *ap;
105		if (idesc->id_type == ADDR)
106			ret = (*idesc->id_func)(idesc);
107		else
108			ret = dirscan(idesc);
109		if (ret & STOP)
110			return (ret);
111	}
112	idesc->id_numfrags = sblock.fs_frag;
113	remsize = dino.di_size - sblock.fs_bsize * NDADDR;
114	sizepb = sblock.fs_bsize;
115	for (ap = &dino.di_ib[0], n = 1; n <= NIADDR; ap++, n++) {
116		if (*ap) {
117			idesc->id_blkno = *ap;
118			ret = iblock(idesc, n, remsize);
119			if (ret & STOP)
120				return (ret);
121		} else {
122			if (idesc->id_type == DATA && remsize > 0) {
123				/* An empty block in a directory XXX */
124				getpathname(pathbuf, idesc->id_number,
125						idesc->id_number);
126                        	pfatal("DIRECTORY %s: CONTAINS EMPTY BLOCKS",
127					pathbuf);
128                        	if (reply("ADJUST LENGTH") == 1) {
129					dp = ginode(idesc->id_number);
130                                	dp->di_size -= remsize;
131					remsize = 0;
132					printf(
133					    "YOU MUST RERUN FSCK AFTERWARDS\n");
134					rerun = 1;
135                                	inodirty();
136					break;
137                        	}
138			}
139		}
140		sizepb *= NINDIR(&sblock);
141		remsize -= sizepb;
142	}
143	return (KEEPON);
144}
145
146static int
147iblock(idesc, ilevel, isize)
148	struct inodesc *idesc;
149	long ilevel;
150	quad_t isize;
151{
152	ufs_daddr_t *ap;
153	ufs_daddr_t *aplim;
154	struct bufarea *bp;
155	int i, n, (*func)(), nif;
156	quad_t sizepb;
157	char buf[BUFSIZ];
158	char pathbuf[MAXPATHLEN + 1];
159	struct dinode *dp;
160
161	if (idesc->id_type == ADDR) {
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	nif = howmany(isize , sizepb);
174	if (nif > NINDIR(&sblock))
175		nif = NINDIR(&sblock);
176	if (idesc->id_func == pass1check && nif < NINDIR(&sblock)) {
177		aplim = &bp->b_un.b_indir[NINDIR(&sblock)];
178		for (ap = &bp->b_un.b_indir[nif]; ap < aplim; ap++) {
179			if (*ap == 0)
180				continue;
181			(void)sprintf(buf, "PARTIALLY TRUNCATED INODE I=%lu",
182			    (u_long)idesc->id_number);
183			if (dofix(idesc, buf)) {
184				*ap = 0;
185				dirty(bp);
186			}
187		}
188		flush(fswritefd, bp);
189	}
190	aplim = &bp->b_un.b_indir[nif];
191	for (ap = bp->b_un.b_indir; ap < aplim; ap++) {
192		if (*ap) {
193			idesc->id_blkno = *ap;
194			if (ilevel == 0)
195				n = (*func)(idesc);
196			else
197				n = iblock(idesc, ilevel, isize);
198			if (n & STOP) {
199				bp->b_flags &= ~B_INUSE;
200				return (n);
201			}
202		} else {
203			if (idesc->id_type == DATA && isize > 0) {
204				/* An empty block in a directory XXX */
205				getpathname(pathbuf, idesc->id_number,
206						idesc->id_number);
207                        	pfatal("DIRECTORY %s: CONTAINS EMPTY BLOCKS",
208					pathbuf);
209                        	if (reply("ADJUST LENGTH") == 1) {
210					dp = ginode(idesc->id_number);
211                                	dp->di_size -= isize;
212					isize = 0;
213					printf(
214					    "YOU MUST RERUN FSCK AFTERWARDS\n");
215					rerun = 1;
216                                	inodirty();
217					bp->b_flags &= ~B_INUSE;
218					return(STOP);
219                        	}
220			}
221		}
222		isize -= sizepb;
223	}
224	bp->b_flags &= ~B_INUSE;
225	return (KEEPON);
226}
227
228/*
229 * Check that a block in a legal block number.
230 * Return 0 if in range, 1 if out of range.
231 */
232int
233chkrange(blk, cnt)
234	ufs_daddr_t blk;
235	int cnt;
236{
237	register int c;
238
239	if (cnt <= 0 || blk <= 0 || blk > maxfsblock ||
240	    cnt - 1 > maxfsblock - blk)
241		return (1);
242	if (cnt > sblock.fs_frag ||
243	    fragnum(&sblock, blk) + cnt > sblock.fs_frag) {
244		if (debug)
245			printf("bad size: blk %ld, offset %ld, size %ld\n",
246				blk, fragnum(&sblock, blk), cnt);
247		return (1);
248	}
249	c = dtog(&sblock, blk);
250	if (blk < cgdmin(&sblock, c)) {
251		if ((blk + cnt) > cgsblock(&sblock, c)) {
252			if (debug) {
253				printf("blk %ld < cgdmin %ld;",
254				    (long)blk, (long)cgdmin(&sblock, c));
255				printf(" blk + cnt %ld > cgsbase %ld\n",
256				    (long)(blk + cnt),
257				    (long)cgsblock(&sblock, c));
258			}
259			return (1);
260		}
261	} else {
262		if ((blk + cnt) > cgbase(&sblock, c+1)) {
263			if (debug)  {
264				printf("blk %ld >= cgdmin %ld;",
265				    (long)blk, (long)cgdmin(&sblock, c));
266				printf(" blk + cnt %ld > sblock.fs_fpg %ld\n",
267				    (long)(blk + cnt), (long)sblock.fs_fpg);
268			}
269			return (1);
270		}
271	}
272	return (0);
273}
274
275/*
276 * General purpose interface for reading inodes.
277 */
278struct dinode *
279ginode(inumber)
280	ino_t inumber;
281{
282	ufs_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	return (&pbp->b_un.b_dinode[inumber % INOPB(&sblock)]);
295}
296
297/*
298 * Special purpose version of ginode used to optimize first pass
299 * over all the inodes in numerical order.
300 */
301ino_t nextino, lastinum;
302long readcnt, readpercg, fullcnt, inobufsize, partialcnt, partialsize;
303struct dinode *inodebuf;
304
305struct dinode *
306getnextinode(inumber)
307	ino_t inumber;
308{
309	long size;
310	ufs_daddr_t dblk;
311	static struct dinode *dp;
312
313	if (inumber != nextino++ || inumber > maxino)
314		errx(EEXIT, "bad inode number %d to nextinode", inumber);
315	if (inumber >= lastinum) {
316		readcnt++;
317		dblk = fsbtodb(&sblock, ino_to_fsba(&sblock, lastinum));
318		if (readcnt % readpercg == 0) {
319			size = partialsize;
320			lastinum += partialcnt;
321		} else {
322			size = inobufsize;
323			lastinum += fullcnt;
324		}
325		/*
326		 * If bread returns an error, it will already have zeroed
327		 * out the buffer, so we do not need to do so here.
328		 */
329		(void)bread(fsreadfd, (char *)inodebuf, dblk, size);
330		dp = inodebuf;
331	}
332	return (dp++);
333}
334
335void
336setinodebuf(inum)
337	ino_t inum;
338{
339
340	if (inum % sblock.fs_ipg != 0)
341		errx(EEXIT, "bad inode number %d to setinodebuf", inum);
342	startinum = 0;
343	nextino = inum;
344	lastinum = inum;
345	readcnt = 0;
346	if (inodebuf != NULL)
347		return;
348	inobufsize = blkroundup(&sblock, INOBUFSIZE);
349	fullcnt = inobufsize / sizeof(struct dinode);
350	readpercg = sblock.fs_ipg / fullcnt;
351	partialcnt = sblock.fs_ipg % fullcnt;
352	partialsize = partialcnt * sizeof(struct dinode);
353	if (partialcnt != 0) {
354		readpercg++;
355	} else {
356		partialcnt = fullcnt;
357		partialsize = inobufsize;
358	}
359	if ((inodebuf = (struct dinode *)malloc((unsigned)inobufsize)) == NULL)
360		errx(EEXIT, "cannot allocate space for inode buffer");
361}
362
363void
364freeinodebuf()
365{
366
367	if (inodebuf != NULL)
368		free((char *)inodebuf);
369	inodebuf = NULL;
370}
371
372/*
373 * Routines to maintain information about directory inodes.
374 * This is built during the first pass and used during the
375 * second and third passes.
376 *
377 * Enter inodes into the cache.
378 */
379void
380cacheino(dp, inumber)
381	register struct dinode *dp;
382	ino_t inumber;
383{
384	register struct inoinfo *inp;
385	struct inoinfo **inpp;
386	int blks;
387
388	blks = howmany(dp->di_size, sblock.fs_bsize);
389	if (blks > NDADDR)
390		blks = NDADDR + NIADDR;
391	inp = (struct inoinfo *)
392		malloc(sizeof(*inp) + (blks - 1) * sizeof(ufs_daddr_t));
393	if (inp == NULL)
394		errx(EEXIT, "cannot increase directory list");
395	inpp = &inphead[inumber % numdirs];
396	inp->i_nexthash = *inpp;
397	*inpp = inp;
398	inp->i_parent = inumber == ROOTINO ? ROOTINO : (ino_t)0;
399	inp->i_dotdot = (ino_t)0;
400	inp->i_number = inumber;
401	inp->i_isize = dp->di_size;
402	inp->i_numblks = blks * sizeof(ufs_daddr_t);
403	memmove(&inp->i_blks[0], &dp->di_db[0], (size_t)inp->i_numblks);
404	if (inplast == listmax) {
405		listmax += 100;
406		inpsort = (struct inoinfo **)realloc((char *)inpsort,
407		    (unsigned)listmax * sizeof(struct inoinfo *));
408		if (inpsort == NULL)
409			errx(EEXIT, "cannot increase directory list");
410	}
411	inpsort[inplast++] = inp;
412}
413
414/*
415 * Look up an inode cache structure.
416 */
417struct inoinfo *
418getinoinfo(inumber)
419	ino_t inumber;
420{
421	register struct inoinfo *inp;
422
423	for (inp = inphead[inumber % numdirs]; inp; inp = inp->i_nexthash) {
424		if (inp->i_number != inumber)
425			continue;
426		return (inp);
427	}
428	errx(EEXIT, "cannot find inode %d", inumber);
429	return ((struct inoinfo *)0);
430}
431
432/*
433 * Clean up all the inode cache structure.
434 */
435void
436inocleanup()
437{
438	register struct inoinfo **inpp;
439
440	if (inphead == NULL)
441		return;
442	for (inpp = &inpsort[inplast - 1]; inpp >= inpsort; inpp--)
443		free((char *)(*inpp));
444	free((char *)inphead);
445	free((char *)inpsort);
446	inphead = inpsort = NULL;
447}
448
449void
450inodirty()
451{
452
453	dirty(pbp);
454}
455
456void
457clri(idesc, type, flag)
458	register struct inodesc *idesc;
459	char *type;
460	int flag;
461{
462	register struct dinode *dp;
463
464	dp = ginode(idesc->id_number);
465	if (flag == 1) {
466		pwarn("%s %s", type,
467		    (dp->di_mode & IFMT) == IFDIR ? "DIR" : "FILE");
468		pinode(idesc->id_number);
469	}
470	if (preen || reply("CLEAR") == 1) {
471		if (preen)
472			printf(" (CLEARED)\n");
473		n_files--;
474		(void)ckinode(dp, idesc);
475		clearinode(dp);
476		inoinfo(idesc->id_number)->ino_state = USTATE;
477		inodirty();
478	}
479}
480
481int
482findname(idesc)
483	struct inodesc *idesc;
484{
485	register struct direct *dirp = idesc->id_dirp;
486
487	if (dirp->d_ino != idesc->id_parent || idesc->id_entryno < 2) {
488		idesc->id_entryno++;
489		return (KEEPON);
490	}
491	memmove(idesc->id_name, dirp->d_name, (size_t)dirp->d_namlen + 1);
492	return (STOP|FOUND);
493}
494
495int
496findino(idesc)
497	struct inodesc *idesc;
498{
499	register struct direct *dirp = idesc->id_dirp;
500
501	if (dirp->d_ino == 0)
502		return (KEEPON);
503	if (strcmp(dirp->d_name, idesc->id_name) == 0 &&
504	    dirp->d_ino >= ROOTINO && dirp->d_ino <= maxino) {
505		idesc->id_parent = dirp->d_ino;
506		return (STOP|FOUND);
507	}
508	return (KEEPON);
509}
510
511int
512clearentry(idesc)
513	struct inodesc *idesc;
514{
515	register struct direct *dirp = idesc->id_dirp;
516
517	if (dirp->d_ino != idesc->id_parent || idesc->id_entryno < 2) {
518		idesc->id_entryno++;
519		return (KEEPON);
520	}
521	dirp->d_ino = 0;
522	return (STOP|FOUND|ALTERED);
523}
524
525void
526pinode(ino)
527	ino_t ino;
528{
529	register struct dinode *dp;
530	register char *p;
531	struct passwd *pw;
532	time_t t;
533
534	printf(" I=%lu ", (u_long)ino);
535	if (ino < ROOTINO || ino > maxino)
536		return;
537	dp = ginode(ino);
538	printf(" OWNER=");
539	if ((pw = getpwuid((int)dp->di_uid)) != 0)
540		printf("%s ", pw->pw_name);
541	else
542		printf("%u ", (unsigned)dp->di_uid);
543	printf("MODE=%o\n", dp->di_mode);
544	if (preen)
545		printf("%s: ", cdevname);
546	printf("SIZE=%qu ", dp->di_size);
547	t = dp->di_mtime;
548	p = ctime(&t);
549	printf("MTIME=%12.12s %4.4s ", &p[4], &p[20]);
550}
551
552void
553blkerror(ino, type, blk)
554	ino_t ino;
555	char *type;
556	ufs_daddr_t blk;
557{
558
559	pfatal("%ld %s I=%lu", blk, type, ino);
560	printf("\n");
561	switch (inoinfo(ino)->ino_state) {
562
563	case FSTATE:
564		inoinfo(ino)->ino_state = FCLEAR;
565		return;
566
567	case DSTATE:
568		inoinfo(ino)->ino_state = DCLEAR;
569		return;
570
571	case FCLEAR:
572	case DCLEAR:
573		return;
574
575	default:
576		errx(EEXIT, "BAD STATE %d TO BLKERR", inoinfo(ino)->ino_state);
577		/* NOTREACHED */
578	}
579}
580
581/*
582 * allocate an unused inode
583 */
584ino_t
585allocino(request, type)
586	ino_t request;
587	int type;
588{
589	register ino_t ino;
590	register struct dinode *dp;
591	struct cg *cgp = &cgrp;
592	int cg;
593
594	if (request == 0)
595		request = ROOTINO;
596	else if (inoinfo(request)->ino_state != USTATE)
597		return (0);
598	for (ino = request; ino < maxino; ino++)
599		if (inoinfo(ino)->ino_state == USTATE)
600			break;
601	if (ino == maxino)
602		return (0);
603	cg = ino_to_cg(&sblock, ino);
604	getblk(&cgblk, cgtod(&sblock, cg), sblock.fs_cgsize);
605	if (!cg_chkmagic(cgp))
606		pfatal("CG %d: BAD MAGIC NUMBER\n", cg);
607	setbit(cg_inosused(cgp), ino % sblock.fs_ipg);
608	cgp->cg_cs.cs_nifree--;
609	switch (type & IFMT) {
610	case IFDIR:
611		inoinfo(ino)->ino_state = DSTATE;
612		cgp->cg_cs.cs_ndir++;
613		break;
614	case IFREG:
615	case IFLNK:
616		inoinfo(ino)->ino_state = FSTATE;
617		break;
618	default:
619		return (0);
620	}
621	cgdirty();
622	dp = ginode(ino);
623	dp->di_db[0] = allocblk((long)1);
624	if (dp->di_db[0] == 0) {
625		inoinfo(ino)->ino_state = USTATE;
626		return (0);
627	}
628	dp->di_mode = type;
629	dp->di_flags = 0;
630	dp->di_atime = time(NULL);
631	dp->di_mtime = dp->di_ctime = dp->di_atime;
632	dp->di_size = sblock.fs_fsize;
633	dp->di_blocks = btodb(sblock.fs_fsize);
634	n_files++;
635	inodirty();
636	if (newinofmt)
637		inoinfo(ino)->ino_type = IFTODT(type);
638	return (ino);
639}
640
641/*
642 * deallocate an inode
643 */
644void
645freeino(ino)
646	ino_t ino;
647{
648	struct inodesc idesc;
649	struct dinode *dp;
650
651	memset(&idesc, 0, sizeof(struct inodesc));
652	idesc.id_type = ADDR;
653	idesc.id_func = pass4check;
654	idesc.id_number = ino;
655	dp = ginode(ino);
656	(void)ckinode(dp, &idesc);
657	clearinode(dp);
658	inodirty();
659	inoinfo(ino)->ino_state = USTATE;
660	n_files--;
661}
662