fsutil.c revision 253155
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: stable/9/sbin/fsck_ffs/fsutil.c 253155 2013-07-10 14:13:37Z des $");
37
38#include <sys/param.h>
39#include <sys/time.h>
40#include <sys/types.h>
41#include <sys/sysctl.h>
42#include <sys/disk.h>
43#include <sys/disklabel.h>
44#include <sys/ioctl.h>
45#include <sys/stat.h>
46
47#include <ufs/ufs/dinode.h>
48#include <ufs/ufs/dir.h>
49#include <ufs/ffs/fs.h>
50
51#include <err.h>
52#include <errno.h>
53#include <string.h>
54#include <ctype.h>
55#include <fstab.h>
56#include <stdint.h>
57#include <stdio.h>
58#include <stdlib.h>
59#include <time.h>
60#include <unistd.h>
61
62#include "fsck.h"
63
64static void slowio_start(void);
65static void slowio_end(void);
66static void printIOstats(void);
67
68static long diskreads, totaldiskreads, totalreads; /* Disk cache statistics */
69static struct timespec startpass, finishpass;
70struct timeval slowio_starttime;
71int slowio_delay_usec = 10000;	/* Initial IO delay for background fsck */
72int slowio_pollcnt;
73static struct bufarea cgblk;	/* backup buffer for cylinder group blocks */
74static TAILQ_HEAD(buflist, bufarea) bufhead;	/* head of buffer cache list */
75static int numbufs;				/* size of buffer cache */
76static char *buftype[BT_NUMBUFTYPES] = BT_NAMES;
77
78int
79ftypeok(union dinode *dp)
80{
81	switch (DIP(dp, di_mode) & IFMT) {
82
83	case IFDIR:
84	case IFREG:
85	case IFBLK:
86	case IFCHR:
87	case IFLNK:
88	case IFSOCK:
89	case IFIFO:
90		return (1);
91
92	default:
93		if (debug)
94			printf("bad file type 0%o\n", DIP(dp, di_mode));
95		return (0);
96	}
97}
98
99int
100reply(const char *question)
101{
102	int persevere;
103	char c;
104
105	if (preen)
106		pfatal("INTERNAL ERROR: GOT TO reply()");
107	persevere = !strcmp(question, "CONTINUE");
108	printf("\n");
109	if (!persevere && (nflag || (fswritefd < 0 && bkgrdflag == 0))) {
110		printf("%s? no\n\n", question);
111		resolved = 0;
112		return (0);
113	}
114	if (yflag || (persevere && nflag)) {
115		printf("%s? yes\n\n", question);
116		return (1);
117	}
118	do	{
119		printf("%s? [yn] ", question);
120		(void) fflush(stdout);
121		c = getc(stdin);
122		while (c != '\n' && getc(stdin) != '\n') {
123			if (feof(stdin)) {
124				resolved = 0;
125				return (0);
126			}
127		}
128	} while (c != 'y' && c != 'Y' && c != 'n' && c != 'N');
129	printf("\n");
130	if (c == 'y' || c == 'Y')
131		return (1);
132	resolved = 0;
133	return (0);
134}
135
136/*
137 * Look up state information for an inode.
138 */
139struct inostat *
140inoinfo(ino_t inum)
141{
142	static struct inostat unallocated = { USTATE, 0, 0 };
143	struct inostatlist *ilp;
144	int iloff;
145
146	if (inum > maxino)
147		errx(EEXIT, "inoinfo: inumber %d out of range", inum);
148	ilp = &inostathead[inum / sblock.fs_ipg];
149	iloff = inum % sblock.fs_ipg;
150	if (iloff >= ilp->il_numalloced)
151		return (&unallocated);
152	return (&ilp->il_stat[iloff]);
153}
154
155/*
156 * Malloc buffers and set up cache.
157 */
158void
159bufinit(void)
160{
161	struct bufarea *bp;
162	long bufcnt, i;
163	char *bufp;
164
165	pbp = pdirbp = (struct bufarea *)0;
166	bufp = Malloc((unsigned int)sblock.fs_bsize);
167	if (bufp == 0)
168		errx(EEXIT, "cannot allocate buffer pool");
169	cgblk.b_un.b_buf = bufp;
170	initbarea(&cgblk, BT_CYLGRP);
171	TAILQ_INIT(&bufhead);
172	bufcnt = MAXBUFS;
173	if (bufcnt < MINBUFS)
174		bufcnt = MINBUFS;
175	for (i = 0; i < bufcnt; i++) {
176		bp = (struct bufarea *)Malloc(sizeof(struct bufarea));
177		bufp = Malloc((unsigned int)sblock.fs_bsize);
178		if (bp == NULL || bufp == NULL) {
179			if (i >= MINBUFS)
180				break;
181			errx(EEXIT, "cannot allocate buffer pool");
182		}
183		bp->b_un.b_buf = bufp;
184		TAILQ_INSERT_HEAD(&bufhead, bp, b_list);
185		initbarea(bp, BT_UNKNOWN);
186	}
187	numbufs = i;	/* save number of buffers */
188	for (i = 0; i < BT_NUMBUFTYPES; i++) {
189		readtime[i].tv_sec = totalreadtime[i].tv_sec = 0;
190		readtime[i].tv_nsec = totalreadtime[i].tv_nsec = 0;
191		readcnt[i] = totalreadcnt[i] = 0;
192	}
193}
194
195/*
196 * Manage cylinder group buffers.
197 */
198static struct bufarea *cgbufs;	/* header for cylinder group cache */
199static int flushtries;		/* number of tries to reclaim memory */
200
201struct bufarea *
202cgget(int cg)
203{
204	struct bufarea *cgbp;
205	struct cg *cgp;
206
207	if (cgbufs == NULL) {
208		cgbufs = Calloc(sblock.fs_ncg, sizeof(struct bufarea));
209		if (cgbufs == NULL)
210			errx(EEXIT, "cannot allocate cylinder group buffers");
211	}
212	cgbp = &cgbufs[cg];
213	if (cgbp->b_un.b_cg != NULL)
214		return (cgbp);
215	cgp = NULL;
216	if (flushtries == 0)
217		cgp = malloc((unsigned int)sblock.fs_cgsize);
218	if (cgp == NULL) {
219		getblk(&cgblk, cgtod(&sblock, cg), sblock.fs_cgsize);
220		return (&cgblk);
221	}
222	cgbp->b_un.b_cg = cgp;
223	initbarea(cgbp, BT_CYLGRP);
224	getblk(cgbp, cgtod(&sblock, cg), sblock.fs_cgsize);
225	return (cgbp);
226}
227
228/*
229 * Attempt to flush a cylinder group cache entry.
230 * Return whether the flush was successful.
231 */
232int
233flushentry(void)
234{
235	struct bufarea *cgbp;
236
237	cgbp = &cgbufs[flushtries++];
238	if (cgbp->b_un.b_cg == NULL)
239		return (0);
240	flush(fswritefd, cgbp);
241	free(cgbp->b_un.b_buf);
242	cgbp->b_un.b_buf = NULL;
243	return (1);
244}
245
246/*
247 * Manage a cache of directory blocks.
248 */
249struct bufarea *
250getdatablk(ufs2_daddr_t blkno, long size, int type)
251{
252	struct bufarea *bp;
253
254	TAILQ_FOREACH(bp, &bufhead, b_list)
255		if (bp->b_bno == fsbtodb(&sblock, blkno))
256			goto foundit;
257	TAILQ_FOREACH_REVERSE(bp, &bufhead, buflist, b_list)
258		if ((bp->b_flags & B_INUSE) == 0)
259			break;
260	if (bp == NULL)
261		errx(EEXIT, "deadlocked buffer pool");
262	bp->b_type = type;
263	getblk(bp, blkno, size);
264	/* fall through */
265foundit:
266	if (debug && bp->b_type != type)
267		printf("Buffer type changed from %s to %s\n",
268		    buftype[bp->b_type], buftype[type]);
269	TAILQ_REMOVE(&bufhead, bp, b_list);
270	TAILQ_INSERT_HEAD(&bufhead, bp, b_list);
271	bp->b_flags |= B_INUSE;
272	return (bp);
273}
274
275/*
276 * Timespec operations (from <sys/time.h>).
277 */
278#define	timespecsub(vvp, uvp)						\
279	do {								\
280		(vvp)->tv_sec -= (uvp)->tv_sec;				\
281		(vvp)->tv_nsec -= (uvp)->tv_nsec;			\
282		if ((vvp)->tv_nsec < 0) {				\
283			(vvp)->tv_sec--;				\
284			(vvp)->tv_nsec += 1000000000;			\
285		}							\
286	} while (0)
287#define	timespecadd(vvp, uvp)						\
288	do {								\
289		(vvp)->tv_sec += (uvp)->tv_sec;				\
290		(vvp)->tv_nsec += (uvp)->tv_nsec;			\
291		if ((vvp)->tv_nsec >= 1000000000) {			\
292			(vvp)->tv_sec++;				\
293			(vvp)->tv_nsec -= 1000000000;			\
294		}							\
295	} while (0)
296
297void
298getblk(struct bufarea *bp, ufs2_daddr_t blk, long size)
299{
300	ufs2_daddr_t dblk;
301	struct timespec start, finish;
302
303	dblk = fsbtodb(&sblock, blk);
304	if (bp->b_bno == dblk) {
305		totalreads++;
306	} else {
307		flush(fswritefd, bp);
308		if (debug) {
309			readcnt[bp->b_type]++;
310			clock_gettime(CLOCK_REALTIME_PRECISE, &start);
311		}
312		bp->b_errs = blread(fsreadfd, bp->b_un.b_buf, dblk, size);
313		if (debug) {
314			clock_gettime(CLOCK_REALTIME_PRECISE, &finish);
315			timespecsub(&finish, &start);
316			timespecadd(&readtime[bp->b_type], &finish);
317		}
318		bp->b_bno = dblk;
319		bp->b_size = size;
320	}
321}
322
323void
324flush(int fd, struct bufarea *bp)
325{
326	int i, j;
327
328	if (!bp->b_dirty)
329		return;
330	bp->b_dirty = 0;
331	if (fswritefd < 0) {
332		pfatal("WRITING IN READ_ONLY MODE.\n");
333		return;
334	}
335	if (bp->b_errs != 0)
336		pfatal("WRITING %sZERO'ED BLOCK %lld TO DISK\n",
337		    (bp->b_errs == bp->b_size / dev_bsize) ? "" : "PARTIALLY ",
338		    (long long)bp->b_bno);
339	bp->b_errs = 0;
340	blwrite(fd, bp->b_un.b_buf, bp->b_bno, (long)bp->b_size);
341	if (bp != &sblk)
342		return;
343	for (i = 0, j = 0; i < sblock.fs_cssize; i += sblock.fs_bsize, j++) {
344		blwrite(fswritefd, (char *)sblock.fs_csp + i,
345		    fsbtodb(&sblock, sblock.fs_csaddr + j * sblock.fs_frag),
346		    sblock.fs_cssize - i < sblock.fs_bsize ?
347		    sblock.fs_cssize - i : sblock.fs_bsize);
348	}
349}
350
351void
352rwerror(const char *mesg, ufs2_daddr_t blk)
353{
354
355	if (bkgrdcheck)
356		exit(EEXIT);
357	if (preen == 0)
358		printf("\n");
359	pfatal("CANNOT %s: %ld", mesg, (long)blk);
360	if (reply("CONTINUE") == 0)
361		exit(EEXIT);
362}
363
364void
365ckfini(int markclean)
366{
367	struct bufarea *bp, *nbp;
368	int ofsmodified, cnt;
369
370	if (bkgrdflag) {
371		unlink(snapname);
372		if ((!(sblock.fs_flags & FS_UNCLEAN)) != markclean) {
373			cmd.value = FS_UNCLEAN;
374			cmd.size = markclean ? -1 : 1;
375			if (sysctlbyname("vfs.ffs.setflags", 0, 0,
376			    &cmd, sizeof cmd) == -1)
377				rwerror("SET FILE SYSTEM FLAGS", FS_UNCLEAN);
378			if (!preen) {
379				printf("\n***** FILE SYSTEM MARKED %s *****\n",
380				    markclean ? "CLEAN" : "DIRTY");
381				if (!markclean)
382					rerun = 1;
383			}
384		} else if (!preen && !markclean) {
385			printf("\n***** FILE SYSTEM STILL DIRTY *****\n");
386			rerun = 1;
387		}
388	}
389	if (debug && totalreads > 0)
390		printf("cache with %d buffers missed %ld of %ld (%d%%)\n",
391		    numbufs, totaldiskreads, totalreads,
392		    (int)(totaldiskreads * 100 / totalreads));
393	if (fswritefd < 0) {
394		(void)close(fsreadfd);
395		return;
396	}
397	flush(fswritefd, &sblk);
398	if (havesb && cursnapshot == 0 && sblock.fs_magic == FS_UFS2_MAGIC &&
399	    sblk.b_bno != sblock.fs_sblockloc / dev_bsize &&
400	    !preen && reply("UPDATE STANDARD SUPERBLOCK")) {
401		sblk.b_bno = sblock.fs_sblockloc / dev_bsize;
402		sbdirty();
403		flush(fswritefd, &sblk);
404	}
405	flush(fswritefd, &cgblk);
406	free(cgblk.b_un.b_buf);
407	cnt = 0;
408	TAILQ_FOREACH_REVERSE_SAFE(bp, &bufhead, buflist, b_list, nbp) {
409		TAILQ_REMOVE(&bufhead, bp, b_list);
410		cnt++;
411		flush(fswritefd, bp);
412		free(bp->b_un.b_buf);
413		free((char *)bp);
414	}
415	if (numbufs != cnt)
416		errx(EEXIT, "panic: lost %d buffers", numbufs - cnt);
417	for (cnt = 0; cnt < sblock.fs_ncg; cnt++) {
418		if (cgbufs[cnt].b_un.b_cg == NULL)
419			continue;
420		flush(fswritefd, &cgbufs[cnt]);
421		free(cgbufs[cnt].b_un.b_cg);
422	}
423	free(cgbufs);
424	pbp = pdirbp = (struct bufarea *)0;
425	if (cursnapshot == 0 && sblock.fs_clean != markclean) {
426		if ((sblock.fs_clean = markclean) != 0) {
427			sblock.fs_flags &= ~(FS_UNCLEAN | FS_NEEDSFSCK);
428			sblock.fs_pendingblocks = 0;
429			sblock.fs_pendinginodes = 0;
430		}
431		sbdirty();
432		ofsmodified = fsmodified;
433		flush(fswritefd, &sblk);
434		fsmodified = ofsmodified;
435		if (!preen) {
436			printf("\n***** FILE SYSTEM MARKED %s *****\n",
437			    markclean ? "CLEAN" : "DIRTY");
438			if (!markclean)
439				rerun = 1;
440		}
441	} else if (!preen) {
442		if (markclean) {
443			printf("\n***** FILE SYSTEM IS CLEAN *****\n");
444		} else {
445			printf("\n***** FILE SYSTEM STILL DIRTY *****\n");
446			rerun = 1;
447		}
448	}
449	(void)close(fsreadfd);
450	(void)close(fswritefd);
451}
452
453/*
454 * Print out I/O statistics.
455 */
456void
457IOstats(char *what)
458{
459	int i;
460
461	if (debug == 0)
462		return;
463	if (diskreads == 0) {
464		printf("%s: no I/O\n\n", what);
465		return;
466	}
467	if (startpass.tv_sec == 0)
468		startpass = startprog;
469	printf("%s: I/O statistics\n", what);
470	printIOstats();
471	totaldiskreads += diskreads;
472	diskreads = 0;
473	for (i = 0; i < BT_NUMBUFTYPES; i++) {
474		timespecadd(&totalreadtime[i], &readtime[i]);
475		totalreadcnt[i] += readcnt[i];
476		readtime[i].tv_sec = readtime[i].tv_nsec = 0;
477		readcnt[i] = 0;
478	}
479	clock_gettime(CLOCK_REALTIME_PRECISE, &startpass);
480}
481
482void
483finalIOstats(void)
484{
485	int i;
486
487	if (debug == 0)
488		return;
489	printf("Final I/O statistics\n");
490	totaldiskreads += diskreads;
491	diskreads = totaldiskreads;
492	startpass = startprog;
493	for (i = 0; i < BT_NUMBUFTYPES; i++) {
494		timespecadd(&totalreadtime[i], &readtime[i]);
495		totalreadcnt[i] += readcnt[i];
496		readtime[i] = totalreadtime[i];
497		readcnt[i] = totalreadcnt[i];
498	}
499	printIOstats();
500}
501
502static void printIOstats(void)
503{
504	long long msec, totalmsec;
505	int i;
506
507	clock_gettime(CLOCK_REALTIME_PRECISE, &finishpass);
508	timespecsub(&finishpass, &startpass);
509	printf("Running time: %jd.%03ld sec\n",
510		(intmax_t)finishpass.tv_sec, finishpass.tv_nsec / 1000000);
511	printf("buffer reads by type:\n");
512	for (totalmsec = 0, i = 0; i < BT_NUMBUFTYPES; i++)
513		totalmsec += readtime[i].tv_sec * 1000 +
514		    readtime[i].tv_nsec / 1000000;
515	if (totalmsec == 0)
516		totalmsec = 1;
517	for (i = 0; i < BT_NUMBUFTYPES; i++) {
518		if (readcnt[i] == 0)
519			continue;
520		msec =
521		    readtime[i].tv_sec * 1000 + readtime[i].tv_nsec / 1000000;
522		printf("%21s:%8ld %2ld.%ld%% %4jd.%03ld sec %2lld.%lld%%\n",
523		    buftype[i], readcnt[i], readcnt[i] * 100 / diskreads,
524		    (readcnt[i] * 1000 / diskreads) % 10,
525		    (intmax_t)readtime[i].tv_sec, readtime[i].tv_nsec / 1000000,
526		    msec * 100 / totalmsec, (msec * 1000 / totalmsec) % 10);
527	}
528	printf("\n");
529}
530
531int
532blread(int fd, char *buf, ufs2_daddr_t blk, long size)
533{
534	char *cp;
535	int i, errs;
536	off_t offset;
537
538	offset = blk;
539	offset *= dev_bsize;
540	if (bkgrdflag)
541		slowio_start();
542	totalreads++;
543	diskreads++;
544	if (lseek(fd, offset, 0) < 0)
545		rwerror("SEEK BLK", blk);
546	else if (read(fd, buf, (int)size) == size) {
547		if (bkgrdflag)
548			slowio_end();
549		return (0);
550	}
551	rwerror("READ BLK", blk);
552	if (lseek(fd, offset, 0) < 0)
553		rwerror("SEEK BLK", blk);
554	errs = 0;
555	memset(buf, 0, (size_t)size);
556	printf("THE FOLLOWING DISK SECTORS COULD NOT BE READ:");
557	for (cp = buf, i = 0; i < size; i += secsize, cp += secsize) {
558		if (read(fd, cp, (int)secsize) != secsize) {
559			(void)lseek(fd, offset + i + secsize, 0);
560			if (secsize != dev_bsize && dev_bsize != 1)
561				printf(" %jd (%jd),",
562				    (intmax_t)(blk * dev_bsize + i) / secsize,
563				    (intmax_t)blk + i / dev_bsize);
564			else
565				printf(" %jd,", (intmax_t)blk + i / dev_bsize);
566			errs++;
567		}
568	}
569	printf("\n");
570	if (errs)
571		resolved = 0;
572	return (errs);
573}
574
575void
576blwrite(int fd, char *buf, ufs2_daddr_t blk, long size)
577{
578	int i;
579	char *cp;
580	off_t offset;
581
582	if (fd < 0)
583		return;
584	offset = blk;
585	offset *= dev_bsize;
586	if (lseek(fd, offset, 0) < 0)
587		rwerror("SEEK BLK", blk);
588	else if (write(fd, buf, (int)size) == size) {
589		fsmodified = 1;
590		return;
591	}
592	resolved = 0;
593	rwerror("WRITE BLK", blk);
594	if (lseek(fd, offset, 0) < 0)
595		rwerror("SEEK BLK", blk);
596	printf("THE FOLLOWING SECTORS COULD NOT BE WRITTEN:");
597	for (cp = buf, i = 0; i < size; i += dev_bsize, cp += dev_bsize)
598		if (write(fd, cp, (int)dev_bsize) != dev_bsize) {
599			(void)lseek(fd, offset + i + dev_bsize, 0);
600			printf(" %jd,", (intmax_t)blk + i / dev_bsize);
601		}
602	printf("\n");
603	return;
604}
605
606void
607blerase(int fd, ufs2_daddr_t blk, long size)
608{
609	off_t ioarg[2];
610
611	if (fd < 0)
612		return;
613	ioarg[0] = blk * dev_bsize;
614	ioarg[1] = size;
615	ioctl(fd, DIOCGDELETE, ioarg);
616	/* we don't really care if we succeed or not */
617	return;
618}
619
620void
621blzero(int fd, ufs2_daddr_t blk, long size)
622{
623	static char *zero;
624	off_t offset, len;
625
626	if (fd < 0)
627		return;
628	len = ZEROBUFSIZE;
629	if (zero == NULL) {
630		zero = calloc(len, 1);
631		if (zero == NULL)
632			errx(EEXIT, "cannot allocate buffer pool");
633	}
634	offset = blk * dev_bsize;
635	if (lseek(fd, offset, 0) < 0)
636		rwerror("SEEK BLK", blk);
637	while (size > 0) {
638		if (size > len)
639			size = len;
640		else
641			len = size;
642		if (write(fd, zero, len) != len)
643			rwerror("WRITE BLK", blk);
644		blk += len / dev_bsize;
645		size -= len;
646	}
647}
648
649/*
650 * Verify cylinder group's magic number and other parameters.  If the
651 * test fails, offer an option to rebuild the whole cylinder group.
652 */
653int
654check_cgmagic(int cg, struct bufarea *cgbp)
655{
656	struct cg *cgp = cgbp->b_un.b_cg;
657
658	/*
659	 * Extended cylinder group checks.
660	 */
661	if (cg_chkmagic(cgp) &&
662	    ((sblock.fs_magic == FS_UFS1_MAGIC &&
663	      cgp->cg_old_niblk == sblock.fs_ipg &&
664	      cgp->cg_ndblk <= sblock.fs_fpg &&
665	      cgp->cg_old_ncyl <= sblock.fs_old_cpg) ||
666	     (sblock.fs_magic == FS_UFS2_MAGIC &&
667	      cgp->cg_niblk == sblock.fs_ipg &&
668	      cgp->cg_ndblk <= sblock.fs_fpg &&
669	      cgp->cg_initediblk <= sblock.fs_ipg))) {
670		return (1);
671	}
672	pfatal("CYLINDER GROUP %d: BAD MAGIC NUMBER", cg);
673	if (!reply("REBUILD CYLINDER GROUP")) {
674		printf("YOU WILL NEED TO RERUN FSCK.\n");
675		rerun = 1;
676		return (1);
677	}
678	/*
679	 * Zero out the cylinder group and then initialize critical fields.
680	 * Bit maps and summaries will be recalculated by later passes.
681	 */
682	memset(cgp, 0, (size_t)sblock.fs_cgsize);
683	cgp->cg_magic = CG_MAGIC;
684	cgp->cg_cgx = cg;
685	cgp->cg_niblk = sblock.fs_ipg;
686	cgp->cg_initediblk = sblock.fs_ipg < 2 * INOPB(&sblock) ?
687	    sblock.fs_ipg : 2 * INOPB(&sblock);
688	if (cgbase(&sblock, cg) + sblock.fs_fpg < sblock.fs_size)
689		cgp->cg_ndblk = sblock.fs_fpg;
690	else
691		cgp->cg_ndblk = sblock.fs_size - cgbase(&sblock, cg);
692	cgp->cg_iusedoff = &cgp->cg_space[0] - (u_char *)(&cgp->cg_firstfield);
693	if (sblock.fs_magic == FS_UFS1_MAGIC) {
694		cgp->cg_niblk = 0;
695		cgp->cg_initediblk = 0;
696		cgp->cg_old_ncyl = sblock.fs_old_cpg;
697		cgp->cg_old_niblk = sblock.fs_ipg;
698		cgp->cg_old_btotoff = cgp->cg_iusedoff;
699		cgp->cg_old_boff = cgp->cg_old_btotoff +
700		    sblock.fs_old_cpg * sizeof(int32_t);
701		cgp->cg_iusedoff = cgp->cg_old_boff +
702		    sblock.fs_old_cpg * sizeof(u_int16_t);
703	}
704	cgp->cg_freeoff = cgp->cg_iusedoff + howmany(sblock.fs_ipg, CHAR_BIT);
705	cgp->cg_nextfreeoff = cgp->cg_freeoff + howmany(sblock.fs_fpg,CHAR_BIT);
706	if (sblock.fs_contigsumsize > 0) {
707		cgp->cg_nclusterblks = cgp->cg_ndblk / sblock.fs_frag;
708		cgp->cg_clustersumoff =
709		    roundup(cgp->cg_nextfreeoff, sizeof(u_int32_t));
710		cgp->cg_clustersumoff -= sizeof(u_int32_t);
711		cgp->cg_clusteroff = cgp->cg_clustersumoff +
712		    (sblock.fs_contigsumsize + 1) * sizeof(u_int32_t);
713		cgp->cg_nextfreeoff = cgp->cg_clusteroff +
714		    howmany(fragstoblks(&sblock, sblock.fs_fpg), CHAR_BIT);
715	}
716	dirty(cgbp);
717	return (0);
718}
719
720/*
721 * allocate a data block with the specified number of fragments
722 */
723ufs2_daddr_t
724allocblk(long frags)
725{
726	int i, j, k, cg, baseblk;
727	struct bufarea *cgbp;
728	struct cg *cgp;
729
730	if (frags <= 0 || frags > sblock.fs_frag)
731		return (0);
732	for (i = 0; i < maxfsblock - sblock.fs_frag; i += sblock.fs_frag) {
733		for (j = 0; j <= sblock.fs_frag - frags; j++) {
734			if (testbmap(i + j))
735				continue;
736			for (k = 1; k < frags; k++)
737				if (testbmap(i + j + k))
738					break;
739			if (k < frags) {
740				j += k;
741				continue;
742			}
743			cg = dtog(&sblock, i + j);
744			cgbp = cgget(cg);
745			cgp = cgbp->b_un.b_cg;
746			if (!check_cgmagic(cg, cgbp))
747				return (0);
748			baseblk = dtogd(&sblock, i + j);
749			for (k = 0; k < frags; k++) {
750				setbmap(i + j + k);
751				clrbit(cg_blksfree(cgp), baseblk + k);
752			}
753			n_blks += frags;
754			if (frags == sblock.fs_frag)
755				cgp->cg_cs.cs_nbfree--;
756			else
757				cgp->cg_cs.cs_nffree -= frags;
758			dirty(cgbp);
759			return (i + j);
760		}
761	}
762	return (0);
763}
764
765/*
766 * Free a previously allocated block
767 */
768void
769freeblk(ufs2_daddr_t blkno, long frags)
770{
771	struct inodesc idesc;
772
773	idesc.id_blkno = blkno;
774	idesc.id_numfrags = frags;
775	(void)pass4check(&idesc);
776}
777
778/* Slow down IO so as to leave some disk bandwidth for other processes */
779void
780slowio_start()
781{
782
783	/* Delay one in every 8 operations */
784	slowio_pollcnt = (slowio_pollcnt + 1) & 7;
785	if (slowio_pollcnt == 0) {
786		gettimeofday(&slowio_starttime, NULL);
787	}
788}
789
790void
791slowio_end()
792{
793	struct timeval tv;
794	int delay_usec;
795
796	if (slowio_pollcnt != 0)
797		return;
798
799	/* Update the slowdown interval. */
800	gettimeofday(&tv, NULL);
801	delay_usec = (tv.tv_sec - slowio_starttime.tv_sec) * 1000000 +
802	    (tv.tv_usec - slowio_starttime.tv_usec);
803	if (delay_usec < 64)
804		delay_usec = 64;
805	if (delay_usec > 2500000)
806		delay_usec = 2500000;
807	slowio_delay_usec = (slowio_delay_usec * 63 + delay_usec) >> 6;
808	/* delay by 8 times the average IO delay */
809	if (slowio_delay_usec > 64)
810		usleep(slowio_delay_usec * 8);
811}
812
813/*
814 * Find a pathname
815 */
816void
817getpathname(char *namebuf, ino_t curdir, ino_t ino)
818{
819	int len;
820	char *cp;
821	struct inodesc idesc;
822	static int busy = 0;
823
824	if (curdir == ino && ino == ROOTINO) {
825		(void)strcpy(namebuf, "/");
826		return;
827	}
828	if (busy || !INO_IS_DVALID(curdir)) {
829		(void)strcpy(namebuf, "?");
830		return;
831	}
832	busy = 1;
833	memset(&idesc, 0, sizeof(struct inodesc));
834	idesc.id_type = DATA;
835	idesc.id_fix = IGNORE;
836	cp = &namebuf[MAXPATHLEN - 1];
837	*cp = '\0';
838	if (curdir != ino) {
839		idesc.id_parent = curdir;
840		goto namelookup;
841	}
842	while (ino != ROOTINO) {
843		idesc.id_number = ino;
844		idesc.id_func = findino;
845		idesc.id_name = strdup("..");
846		if ((ckinode(ginode(ino), &idesc) & FOUND) == 0)
847			break;
848	namelookup:
849		idesc.id_number = idesc.id_parent;
850		idesc.id_parent = ino;
851		idesc.id_func = findname;
852		idesc.id_name = namebuf;
853		if ((ckinode(ginode(idesc.id_number), &idesc)&FOUND) == 0)
854			break;
855		len = strlen(namebuf);
856		cp -= len;
857		memmove(cp, namebuf, (size_t)len);
858		*--cp = '/';
859		if (cp < &namebuf[MAXNAMLEN])
860			break;
861		ino = idesc.id_number;
862	}
863	busy = 0;
864	if (ino != ROOTINO)
865		*--cp = '?';
866	memmove(namebuf, cp, (size_t)(&namebuf[MAXPATHLEN] - cp));
867}
868
869void
870catch(int sig __unused)
871{
872
873	ckfini(0);
874	exit(12);
875}
876
877/*
878 * When preening, allow a single quit to signal
879 * a special exit after file system checks complete
880 * so that reboot sequence may be interrupted.
881 */
882void
883catchquit(int sig __unused)
884{
885	printf("returning to single-user after file system check\n");
886	returntosingle = 1;
887	(void)signal(SIGQUIT, SIG_DFL);
888}
889
890/*
891 * determine whether an inode should be fixed.
892 */
893int
894dofix(struct inodesc *idesc, const char *msg)
895{
896
897	switch (idesc->id_fix) {
898
899	case DONTKNOW:
900		if (idesc->id_type == DATA)
901			direrror(idesc->id_number, msg);
902		else
903			pwarn("%s", msg);
904		if (preen) {
905			printf(" (SALVAGED)\n");
906			idesc->id_fix = FIX;
907			return (ALTERED);
908		}
909		if (reply("SALVAGE") == 0) {
910			idesc->id_fix = NOFIX;
911			return (0);
912		}
913		idesc->id_fix = FIX;
914		return (ALTERED);
915
916	case FIX:
917		return (ALTERED);
918
919	case NOFIX:
920	case IGNORE:
921		return (0);
922
923	default:
924		errx(EEXIT, "UNKNOWN INODESC FIX MODE %d", idesc->id_fix);
925	}
926	/* NOTREACHED */
927	return (0);
928}
929
930#include <stdarg.h>
931
932/*
933 * An unexpected inconsistency occured.
934 * Die if preening or file system is running with soft dependency protocol,
935 * otherwise just print message and continue.
936 */
937void
938pfatal(const char *fmt, ...)
939{
940	va_list ap;
941	va_start(ap, fmt);
942	if (!preen) {
943		(void)vfprintf(stdout, fmt, ap);
944		va_end(ap);
945		if (usedsoftdep)
946			(void)fprintf(stdout,
947			    "\nUNEXPECTED SOFT UPDATE INCONSISTENCY\n");
948		/*
949		 * Force foreground fsck to clean up inconsistency.
950		 */
951		if (bkgrdflag) {
952			cmd.value = FS_NEEDSFSCK;
953			cmd.size = 1;
954			if (sysctlbyname("vfs.ffs.setflags", 0, 0,
955			    &cmd, sizeof cmd) == -1)
956				pwarn("CANNOT SET FS_NEEDSFSCK FLAG\n");
957			fprintf(stdout, "CANNOT RUN IN BACKGROUND\n");
958			ckfini(0);
959			exit(EEXIT);
960		}
961		return;
962	}
963	if (cdevname == NULL)
964		cdevname = strdup("fsck");
965	(void)fprintf(stdout, "%s: ", cdevname);
966	(void)vfprintf(stdout, fmt, ap);
967	(void)fprintf(stdout,
968	    "\n%s: UNEXPECTED%sINCONSISTENCY; RUN fsck MANUALLY.\n",
969	    cdevname, usedsoftdep ? " SOFT UPDATE " : " ");
970	/*
971	 * Force foreground fsck to clean up inconsistency.
972	 */
973	if (bkgrdflag) {
974		cmd.value = FS_NEEDSFSCK;
975		cmd.size = 1;
976		if (sysctlbyname("vfs.ffs.setflags", 0, 0,
977		    &cmd, sizeof cmd) == -1)
978			pwarn("CANNOT SET FS_NEEDSFSCK FLAG\n");
979	}
980	ckfini(0);
981	exit(EEXIT);
982}
983
984/*
985 * Pwarn just prints a message when not preening or running soft dependency
986 * protocol, or a warning (preceded by filename) when preening.
987 */
988void
989pwarn(const char *fmt, ...)
990{
991	va_list ap;
992	va_start(ap, fmt);
993	if (preen)
994		(void)fprintf(stdout, "%s: ", cdevname);
995	(void)vfprintf(stdout, fmt, ap);
996	va_end(ap);
997}
998
999/*
1000 * Stub for routines from kernel.
1001 */
1002void
1003panic(const char *fmt, ...)
1004{
1005	va_list ap;
1006	va_start(ap, fmt);
1007	pfatal("INTERNAL INCONSISTENCY:");
1008	(void)vfprintf(stdout, fmt, ap);
1009	va_end(ap);
1010	exit(EEXIT);
1011}
1012