fsutil.c revision 163845
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[] = "@(#)utilities.c	8.6 (Berkeley) 5/19/95";
33#endif /* not lint */
34#endif
35#include <sys/cdefs.h>
36__FBSDID("$FreeBSD: head/sbin/fsck_ffs/fsutil.c 163845 2006-10-31 22:06:56Z pjd $");
37
38#include <sys/param.h>
39#include <sys/time.h>
40#include <sys/types.h>
41#include <sys/sysctl.h>
42#include <sys/disklabel.h>
43#include <sys/stat.h>
44#include <sys/disklabel.h>
45
46#include <ufs/ufs/dinode.h>
47#include <ufs/ufs/dir.h>
48#include <ufs/ffs/fs.h>
49
50#include <err.h>
51#include <errno.h>
52#include <string.h>
53#include <ctype.h>
54#include <fstab.h>
55#include <stdint.h>
56#include <stdio.h>
57#include <stdlib.h>
58#include <unistd.h>
59
60#include "fsck.h"
61
62static void slowio_start(void);
63static void slowio_end(void);
64
65long	diskreads, totalreads;	/* Disk cache statistics */
66struct timeval slowio_starttime;
67int slowio_delay_usec = 10000;	/* Initial IO delay for background fsck */
68int slowio_pollcnt;
69
70int
71ftypeok(union dinode *dp)
72{
73	switch (DIP(dp, di_mode) & IFMT) {
74
75	case IFDIR:
76	case IFREG:
77	case IFBLK:
78	case IFCHR:
79	case IFLNK:
80	case IFSOCK:
81	case IFIFO:
82		return (1);
83
84	default:
85		if (debug)
86			printf("bad file type 0%o\n", DIP(dp, di_mode));
87		return (0);
88	}
89}
90
91int
92reply(const char *question)
93{
94	int persevere;
95	char c;
96
97	if (preen)
98		pfatal("INTERNAL ERROR: GOT TO reply()");
99	persevere = !strcmp(question, "CONTINUE");
100	printf("\n");
101	if (!persevere && (nflag || (fswritefd < 0 && bkgrdflag == 0))) {
102		printf("%s? no\n\n", question);
103		resolved = 0;
104		return (0);
105	}
106	if (yflag || (persevere && nflag)) {
107		printf("%s? yes\n\n", question);
108		return (1);
109	}
110	do	{
111		printf("%s? [yn] ", question);
112		(void) fflush(stdout);
113		c = getc(stdin);
114		while (c != '\n' && getc(stdin) != '\n') {
115			if (feof(stdin)) {
116				resolved = 0;
117				return (0);
118			}
119		}
120	} while (c != 'y' && c != 'Y' && c != 'n' && c != 'N');
121	printf("\n");
122	if (c == 'y' || c == 'Y')
123		return (1);
124	resolved = 0;
125	return (0);
126}
127
128/*
129 * Look up state information for an inode.
130 */
131struct inostat *
132inoinfo(ino_t inum)
133{
134	static struct inostat unallocated = { USTATE, 0, 0 };
135	struct inostatlist *ilp;
136	int iloff;
137
138	if (inum > maxino)
139		errx(EEXIT, "inoinfo: inumber %d out of range", inum);
140	ilp = &inostathead[inum / sblock.fs_ipg];
141	iloff = inum % sblock.fs_ipg;
142	if (iloff >= ilp->il_numalloced)
143		return (&unallocated);
144	return (&ilp->il_stat[iloff]);
145}
146
147/*
148 * Malloc buffers and set up cache.
149 */
150void
151bufinit(void)
152{
153	struct bufarea *bp;
154	long bufcnt, i;
155	char *bufp;
156
157	pbp = pdirbp = (struct bufarea *)0;
158	bufp = malloc((unsigned int)sblock.fs_bsize);
159	if (bufp == 0)
160		errx(EEXIT, "cannot allocate buffer pool");
161	cgblk.b_un.b_buf = bufp;
162	initbarea(&cgblk);
163	bufhead.b_next = bufhead.b_prev = &bufhead;
164	bufcnt = MAXBUFSPACE / sblock.fs_bsize;
165	if (bufcnt < MINBUFS)
166		bufcnt = MINBUFS;
167	for (i = 0; i < bufcnt; i++) {
168		bp = (struct bufarea *)malloc(sizeof(struct bufarea));
169		bufp = malloc((unsigned int)sblock.fs_bsize);
170		if (bp == NULL || bufp == NULL) {
171			if (i >= MINBUFS)
172				break;
173			errx(EEXIT, "cannot allocate buffer pool");
174		}
175		bp->b_un.b_buf = bufp;
176		bp->b_prev = &bufhead;
177		bp->b_next = bufhead.b_next;
178		bufhead.b_next->b_prev = bp;
179		bufhead.b_next = bp;
180		initbarea(bp);
181	}
182	bufhead.b_size = i;	/* save number of buffers */
183}
184
185/*
186 * Manage a cache of directory blocks.
187 */
188struct bufarea *
189getdatablk(ufs2_daddr_t blkno, long size)
190{
191	struct bufarea *bp;
192
193	for (bp = bufhead.b_next; bp != &bufhead; bp = bp->b_next)
194		if (bp->b_bno == fsbtodb(&sblock, blkno))
195			goto foundit;
196	for (bp = bufhead.b_prev; bp != &bufhead; bp = bp->b_prev)
197		if ((bp->b_flags & B_INUSE) == 0)
198			break;
199	if (bp == &bufhead)
200		errx(EEXIT, "deadlocked buffer pool");
201	getblk(bp, blkno, size);
202	/* fall through */
203foundit:
204	bp->b_prev->b_next = bp->b_next;
205	bp->b_next->b_prev = bp->b_prev;
206	bp->b_prev = &bufhead;
207	bp->b_next = bufhead.b_next;
208	bufhead.b_next->b_prev = bp;
209	bufhead.b_next = bp;
210	bp->b_flags |= B_INUSE;
211	return (bp);
212}
213
214void
215getblk(struct bufarea *bp, ufs2_daddr_t blk, long size)
216{
217	ufs2_daddr_t dblk;
218
219	totalreads++;
220	dblk = fsbtodb(&sblock, blk);
221	if (bp->b_bno != dblk) {
222		flush(fswritefd, bp);
223		diskreads++;
224		bp->b_errs = blread(fsreadfd, bp->b_un.b_buf, dblk, size);
225		bp->b_bno = dblk;
226		bp->b_size = size;
227	}
228}
229
230void
231flush(int fd, struct bufarea *bp)
232{
233	int i, j;
234
235	if (!bp->b_dirty)
236		return;
237	bp->b_dirty = 0;
238	if (fswritefd < 0) {
239		pfatal("WRITING IN READ_ONLY MODE.\n");
240		return;
241	}
242	if (bp->b_errs != 0)
243		pfatal("WRITING %sZERO'ED BLOCK %lld TO DISK\n",
244		    (bp->b_errs == bp->b_size / dev_bsize) ? "" : "PARTIALLY ",
245		    (long long)bp->b_bno);
246	bp->b_errs = 0;
247	blwrite(fd, bp->b_un.b_buf, bp->b_bno, (long)bp->b_size);
248	if (bp != &sblk)
249		return;
250	for (i = 0, j = 0; i < sblock.fs_cssize; i += sblock.fs_bsize, j++) {
251		blwrite(fswritefd, (char *)sblock.fs_csp + i,
252		    fsbtodb(&sblock, sblock.fs_csaddr + j * sblock.fs_frag),
253		    sblock.fs_cssize - i < sblock.fs_bsize ?
254		    sblock.fs_cssize - i : sblock.fs_bsize);
255	}
256}
257
258void
259rwerror(const char *mesg, ufs2_daddr_t blk)
260{
261
262	if (bkgrdcheck)
263		exit(EEXIT);
264	if (preen == 0)
265		printf("\n");
266	pfatal("CANNOT %s: %ld", mesg, (long)blk);
267	if (reply("CONTINUE") == 0)
268		exit(EEXIT);
269}
270
271void
272ckfini(int markclean)
273{
274	struct bufarea *bp, *nbp;
275	int ofsmodified, cnt = 0;
276
277	if (bkgrdflag) {
278		unlink(snapname);
279		if ((!(sblock.fs_flags & FS_UNCLEAN)) != markclean) {
280			cmd.value = FS_UNCLEAN;
281			cmd.size = markclean ? -1 : 1;
282			if (sysctlbyname("vfs.ffs.setflags", 0, 0,
283			    &cmd, sizeof cmd) == -1)
284				rwerror("SET FILE SYSTEM FLAGS", FS_UNCLEAN);
285			if (!preen) {
286				printf("\n***** FILE SYSTEM MARKED %s *****\n",
287				    markclean ? "CLEAN" : "DIRTY");
288				if (!markclean)
289					rerun = 1;
290			}
291		} else if (!preen && !markclean) {
292			printf("\n***** FILE SYSTEM STILL DIRTY *****\n");
293			rerun = 1;
294		}
295	}
296	if (fswritefd < 0) {
297		(void)close(fsreadfd);
298		return;
299	}
300	flush(fswritefd, &sblk);
301	if (havesb && cursnapshot == 0 && sblock.fs_magic == FS_UFS2_MAGIC &&
302	    sblk.b_bno != sblock.fs_sblockloc / dev_bsize &&
303	    !preen && reply("UPDATE STANDARD SUPERBLOCK")) {
304		sblk.b_bno = sblock.fs_sblockloc / dev_bsize;
305		sbdirty();
306		flush(fswritefd, &sblk);
307	}
308	flush(fswritefd, &cgblk);
309	free(cgblk.b_un.b_buf);
310	for (bp = bufhead.b_prev; bp && bp != &bufhead; bp = nbp) {
311		cnt++;
312		flush(fswritefd, bp);
313		nbp = bp->b_prev;
314		free(bp->b_un.b_buf);
315		free((char *)bp);
316	}
317	if (bufhead.b_size != cnt)
318		errx(EEXIT, "panic: lost %d buffers", bufhead.b_size - cnt);
319	pbp = pdirbp = (struct bufarea *)0;
320	if (cursnapshot == 0 && sblock.fs_clean != markclean) {
321		if ((sblock.fs_clean = markclean) != 0) {
322			sblock.fs_flags &= ~(FS_UNCLEAN | FS_NEEDSFSCK);
323			sblock.fs_pendingblocks = 0;
324			sblock.fs_pendinginodes = 0;
325		}
326		sbdirty();
327		ofsmodified = fsmodified;
328		flush(fswritefd, &sblk);
329		fsmodified = ofsmodified;
330		if (!preen) {
331			printf("\n***** FILE SYSTEM MARKED %s *****\n",
332			    markclean ? "CLEAN" : "DIRTY");
333			if (!markclean)
334				rerun = 1;
335		}
336	} else if (!preen && !markclean) {
337		printf("\n***** FILE SYSTEM STILL DIRTY *****\n");
338		rerun = 1;
339	}
340	if (debug && totalreads > 0)
341		printf("cache missed %ld of %ld (%d%%)\n", diskreads,
342		    totalreads, (int)(diskreads * 100 / totalreads));
343	(void)close(fsreadfd);
344	(void)close(fswritefd);
345}
346
347int
348blread(int fd, char *buf, ufs2_daddr_t blk, long size)
349{
350	char *cp;
351	int i, errs;
352	off_t offset;
353
354	offset = blk;
355	offset *= dev_bsize;
356	if (bkgrdflag)
357		slowio_start();
358	if (lseek(fd, offset, 0) < 0)
359		rwerror("SEEK BLK", blk);
360	else if (read(fd, buf, (int)size) == size) {
361		if (bkgrdflag)
362			slowio_end();
363		return (0);
364	}
365	rwerror("READ BLK", blk);
366	if (lseek(fd, offset, 0) < 0)
367		rwerror("SEEK BLK", blk);
368	errs = 0;
369	memset(buf, 0, (size_t)size);
370	printf("THE FOLLOWING DISK SECTORS COULD NOT BE READ:");
371	for (cp = buf, i = 0; i < size; i += secsize, cp += secsize) {
372		if (read(fd, cp, (int)secsize) != secsize) {
373			(void)lseek(fd, offset + i + secsize, 0);
374			if (secsize != dev_bsize && dev_bsize != 1)
375				printf(" %jd (%jd),",
376				    (intmax_t)(blk * dev_bsize + i) / secsize,
377				    (intmax_t)blk + i / dev_bsize);
378			else
379				printf(" %jd,", (intmax_t)blk + i / dev_bsize);
380			errs++;
381		}
382	}
383	printf("\n");
384	if (errs)
385		resolved = 0;
386	return (errs);
387}
388
389void
390blwrite(int fd, char *buf, ufs2_daddr_t blk, long size)
391{
392	int i;
393	char *cp;
394	off_t offset;
395
396	if (fd < 0)
397		return;
398	offset = blk;
399	offset *= dev_bsize;
400	if (lseek(fd, offset, 0) < 0)
401		rwerror("SEEK BLK", blk);
402	else if (write(fd, buf, (int)size) == size) {
403		fsmodified = 1;
404		return;
405	}
406	resolved = 0;
407	rwerror("WRITE BLK", blk);
408	if (lseek(fd, offset, 0) < 0)
409		rwerror("SEEK BLK", blk);
410	printf("THE FOLLOWING SECTORS COULD NOT BE WRITTEN:");
411	for (cp = buf, i = 0; i < size; i += dev_bsize, cp += dev_bsize)
412		if (write(fd, cp, (int)dev_bsize) != dev_bsize) {
413			(void)lseek(fd, offset + i + dev_bsize, 0);
414			printf(" %jd,", (intmax_t)blk + i / dev_bsize);
415		}
416	printf("\n");
417	return;
418}
419
420/*
421 * allocate a data block with the specified number of fragments
422 */
423ufs2_daddr_t
424allocblk(long frags)
425{
426	int i, j, k, cg, baseblk;
427	struct cg *cgp = &cgrp;
428
429	if (frags <= 0 || frags > sblock.fs_frag)
430		return (0);
431	for (i = 0; i < maxfsblock - sblock.fs_frag; i += sblock.fs_frag) {
432		for (j = 0; j <= sblock.fs_frag - frags; j++) {
433			if (testbmap(i + j))
434				continue;
435			for (k = 1; k < frags; k++)
436				if (testbmap(i + j + k))
437					break;
438			if (k < frags) {
439				j += k;
440				continue;
441			}
442			cg = dtog(&sblock, i + j);
443			getblk(&cgblk, cgtod(&sblock, cg), sblock.fs_cgsize);
444			if (!cg_chkmagic(cgp))
445				pfatal("CG %d: BAD MAGIC NUMBER\n", cg);
446			baseblk = dtogd(&sblock, i + j);
447			for (k = 0; k < frags; k++) {
448				setbmap(i + j + k);
449				clrbit(cg_blksfree(cgp), baseblk + k);
450			}
451			n_blks += frags;
452			if (frags == sblock.fs_frag)
453				cgp->cg_cs.cs_nbfree--;
454			else
455				cgp->cg_cs.cs_nffree -= frags;
456			cgdirty();
457			return (i + j);
458		}
459	}
460	return (0);
461}
462
463/*
464 * Free a previously allocated block
465 */
466void
467freeblk(ufs2_daddr_t blkno, long frags)
468{
469	struct inodesc idesc;
470
471	idesc.id_blkno = blkno;
472	idesc.id_numfrags = frags;
473	(void)pass4check(&idesc);
474}
475
476/* Slow down IO so as to leave some disk bandwidth for other processes */
477void
478slowio_start()
479{
480
481	/* Delay one in every 8 operations */
482	slowio_pollcnt = (slowio_pollcnt + 1) & 7;
483	if (slowio_pollcnt == 0) {
484		gettimeofday(&slowio_starttime, NULL);
485	}
486}
487
488void
489slowio_end()
490{
491	struct timeval tv;
492	int delay_usec;
493
494	if (slowio_pollcnt != 0)
495		return;
496
497	/* Update the slowdown interval. */
498	gettimeofday(&tv, NULL);
499	delay_usec = (tv.tv_sec - slowio_starttime.tv_sec) * 1000000 +
500	    (tv.tv_usec - slowio_starttime.tv_usec);
501	if (delay_usec < 64)
502		delay_usec = 64;
503	if (delay_usec > 2500000)
504		delay_usec = 2500000;
505	slowio_delay_usec = (slowio_delay_usec * 63 + delay_usec) >> 6;
506	/* delay by 8 times the average IO delay */
507	if (slowio_delay_usec > 64)
508		usleep(slowio_delay_usec * 8);
509}
510
511/*
512 * Find a pathname
513 */
514void
515getpathname(char *namebuf, ino_t curdir, ino_t ino)
516{
517	int len;
518	char *cp;
519	struct inodesc idesc;
520	static int busy = 0;
521
522	if (curdir == ino && ino == ROOTINO) {
523		(void)strcpy(namebuf, "/");
524		return;
525	}
526	if (busy || !INO_IS_DVALID(curdir)) {
527		(void)strcpy(namebuf, "?");
528		return;
529	}
530	busy = 1;
531	memset(&idesc, 0, sizeof(struct inodesc));
532	idesc.id_type = DATA;
533	idesc.id_fix = IGNORE;
534	cp = &namebuf[MAXPATHLEN - 1];
535	*cp = '\0';
536	if (curdir != ino) {
537		idesc.id_parent = curdir;
538		goto namelookup;
539	}
540	while (ino != ROOTINO) {
541		idesc.id_number = ino;
542		idesc.id_func = findino;
543		idesc.id_name = strdup("..");
544		if ((ckinode(ginode(ino), &idesc) & FOUND) == 0)
545			break;
546	namelookup:
547		idesc.id_number = idesc.id_parent;
548		idesc.id_parent = ino;
549		idesc.id_func = findname;
550		idesc.id_name = namebuf;
551		if ((ckinode(ginode(idesc.id_number), &idesc)&FOUND) == 0)
552			break;
553		len = strlen(namebuf);
554		cp -= len;
555		memmove(cp, namebuf, (size_t)len);
556		*--cp = '/';
557		if (cp < &namebuf[MAXNAMLEN])
558			break;
559		ino = idesc.id_number;
560	}
561	busy = 0;
562	if (ino != ROOTINO)
563		*--cp = '?';
564	memmove(namebuf, cp, (size_t)(&namebuf[MAXPATHLEN] - cp));
565}
566
567void
568catch(int sig __unused)
569{
570
571	ckfini(0);
572	exit(12);
573}
574
575/*
576 * When preening, allow a single quit to signal
577 * a special exit after file system checks complete
578 * so that reboot sequence may be interrupted.
579 */
580void
581catchquit(int sig __unused)
582{
583	printf("returning to single-user after file system check\n");
584	returntosingle = 1;
585	(void)signal(SIGQUIT, SIG_DFL);
586}
587
588/*
589 * determine whether an inode should be fixed.
590 */
591int
592dofix(struct inodesc *idesc, const char *msg)
593{
594
595	switch (idesc->id_fix) {
596
597	case DONTKNOW:
598		if (idesc->id_type == DATA)
599			direrror(idesc->id_number, msg);
600		else
601			pwarn("%s", msg);
602		if (preen) {
603			printf(" (SALVAGED)\n");
604			idesc->id_fix = FIX;
605			return (ALTERED);
606		}
607		if (reply("SALVAGE") == 0) {
608			idesc->id_fix = NOFIX;
609			return (0);
610		}
611		idesc->id_fix = FIX;
612		return (ALTERED);
613
614	case FIX:
615		return (ALTERED);
616
617	case NOFIX:
618	case IGNORE:
619		return (0);
620
621	default:
622		errx(EEXIT, "UNKNOWN INODESC FIX MODE %d", idesc->id_fix);
623	}
624	/* NOTREACHED */
625	return (0);
626}
627
628#include <stdarg.h>
629
630/*
631 * An unexpected inconsistency occured.
632 * Die if preening or file system is running with soft dependency protocol,
633 * otherwise just print message and continue.
634 */
635void
636pfatal(const char *fmt, ...)
637{
638	va_list ap;
639	va_start(ap, fmt);
640	if (!preen) {
641		(void)vfprintf(stdout, fmt, ap);
642		va_end(ap);
643		if (usedsoftdep)
644			(void)fprintf(stdout,
645			    "\nUNEXPECTED SOFT UPDATE INCONSISTENCY\n");
646		/*
647		 * Force foreground fsck to clean up inconsistency.
648		 */
649		if (bkgrdflag) {
650			cmd.value = FS_NEEDSFSCK;
651			cmd.size = 1;
652			if (sysctlbyname("vfs.ffs.setflags", 0, 0,
653			    &cmd, sizeof cmd) == -1)
654				pwarn("CANNOT SET FS_NEEDSFSCK FLAG\n");
655			fprintf(stdout, "CANNOT RUN IN BACKGROUND\n");
656			ckfini(0);
657			exit(EEXIT);
658		}
659		return;
660	}
661	if (cdevname == NULL)
662		cdevname = strdup("fsck");
663	(void)fprintf(stdout, "%s: ", cdevname);
664	(void)vfprintf(stdout, fmt, ap);
665	(void)fprintf(stdout,
666	    "\n%s: UNEXPECTED%sINCONSISTENCY; RUN fsck MANUALLY.\n",
667	    cdevname, usedsoftdep ? " SOFT UPDATE " : " ");
668	/*
669	 * Force foreground fsck to clean up inconsistency.
670	 */
671	if (bkgrdflag) {
672		cmd.value = FS_NEEDSFSCK;
673		cmd.size = 1;
674		if (sysctlbyname("vfs.ffs.setflags", 0, 0,
675		    &cmd, sizeof cmd) == -1)
676			pwarn("CANNOT SET FS_NEEDSFSCK FLAG\n");
677	}
678	ckfini(0);
679	exit(EEXIT);
680}
681
682/*
683 * Pwarn just prints a message when not preening or running soft dependency
684 * protocol, or a warning (preceded by filename) when preening.
685 */
686void
687pwarn(const char *fmt, ...)
688{
689	va_list ap;
690	va_start(ap, fmt);
691	if (preen)
692		(void)fprintf(stdout, "%s: ", cdevname);
693	(void)vfprintf(stdout, fmt, ap);
694	va_end(ap);
695}
696
697/*
698 * Stub for routines from kernel.
699 */
700void
701panic(const char *fmt, ...)
702{
703	va_list ap;
704	va_start(ap, fmt);
705	pfatal("INTERNAL INCONSISTENCY:");
706	(void)vfprintf(stdout, fmt, ap);
707	va_end(ap);
708	exit(EEXIT);
709}
710