fsutil.c revision 247212
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 247212 2013-02-24 06:44:29Z mckusick $");
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 TAILQ_HEAD(buflist, bufarea) bufhead;	/* head of buffer cache list */
74static int numbufs;				/* size of buffer cache */
75static char *buftype[BT_NUMBUFTYPES] = BT_NAMES;
76
77int
78ftypeok(union dinode *dp)
79{
80	switch (DIP(dp, di_mode) & IFMT) {
81
82	case IFDIR:
83	case IFREG:
84	case IFBLK:
85	case IFCHR:
86	case IFLNK:
87	case IFSOCK:
88	case IFIFO:
89		return (1);
90
91	default:
92		if (debug)
93			printf("bad file type 0%o\n", DIP(dp, di_mode));
94		return (0);
95	}
96}
97
98int
99reply(const char *question)
100{
101	int persevere;
102	char c;
103
104	if (preen)
105		pfatal("INTERNAL ERROR: GOT TO reply()");
106	persevere = !strcmp(question, "CONTINUE");
107	printf("\n");
108	if (!persevere && (nflag || (fswritefd < 0 && bkgrdflag == 0))) {
109		printf("%s? no\n\n", question);
110		resolved = 0;
111		return (0);
112	}
113	if (yflag || (persevere && nflag)) {
114		printf("%s? yes\n\n", question);
115		return (1);
116	}
117	do	{
118		printf("%s? [yn] ", question);
119		(void) fflush(stdout);
120		c = getc(stdin);
121		while (c != '\n' && getc(stdin) != '\n') {
122			if (feof(stdin)) {
123				resolved = 0;
124				return (0);
125			}
126		}
127	} while (c != 'y' && c != 'Y' && c != 'n' && c != 'N');
128	printf("\n");
129	if (c == 'y' || c == 'Y')
130		return (1);
131	resolved = 0;
132	return (0);
133}
134
135/*
136 * Look up state information for an inode.
137 */
138struct inostat *
139inoinfo(ino_t inum)
140{
141	static struct inostat unallocated = { USTATE, 0, 0 };
142	struct inostatlist *ilp;
143	int iloff;
144
145	if (inum > maxino)
146		errx(EEXIT, "inoinfo: inumber %ju out of range",
147		    (uintmax_t)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 a cache of directory blocks.
197 */
198struct bufarea *
199getdatablk(ufs2_daddr_t blkno, long size, int type)
200{
201	struct bufarea *bp;
202
203	TAILQ_FOREACH(bp, &bufhead, b_list)
204		if (bp->b_bno == fsbtodb(&sblock, blkno))
205			goto foundit;
206	TAILQ_FOREACH_REVERSE(bp, &bufhead, buflist, b_list)
207		if ((bp->b_flags & B_INUSE) == 0)
208			break;
209	if (bp == NULL)
210		errx(EEXIT, "deadlocked buffer pool");
211	bp->b_type = type;
212	getblk(bp, blkno, size);
213	/* fall through */
214foundit:
215	if (debug && bp->b_type != type)
216		printf("Buffer type changed from %s to %s\n",
217		    buftype[bp->b_type], buftype[type]);
218	TAILQ_REMOVE(&bufhead, bp, b_list);
219	TAILQ_INSERT_HEAD(&bufhead, bp, b_list);
220	bp->b_flags |= B_INUSE;
221	return (bp);
222}
223
224/*
225 * Timespec operations (from <sys/time.h>).
226 */
227#define	timespecsub(vvp, uvp)						\
228	do {								\
229		(vvp)->tv_sec -= (uvp)->tv_sec;				\
230		(vvp)->tv_nsec -= (uvp)->tv_nsec;			\
231		if ((vvp)->tv_nsec < 0) {				\
232			(vvp)->tv_sec--;				\
233			(vvp)->tv_nsec += 1000000000;			\
234		}							\
235	} while (0)
236#define	timespecadd(vvp, uvp)						\
237	do {								\
238		(vvp)->tv_sec += (uvp)->tv_sec;				\
239		(vvp)->tv_nsec += (uvp)->tv_nsec;			\
240		if ((vvp)->tv_nsec >= 1000000000) {			\
241			(vvp)->tv_sec++;				\
242			(vvp)->tv_nsec -= 1000000000;			\
243		}							\
244	} while (0)
245
246void
247getblk(struct bufarea *bp, ufs2_daddr_t blk, long size)
248{
249	ufs2_daddr_t dblk;
250	struct timespec start, finish;
251
252	dblk = fsbtodb(&sblock, blk);
253	if (bp->b_bno == dblk) {
254		totalreads++;
255	} else {
256		flush(fswritefd, bp);
257		if (debug) {
258			readcnt[bp->b_type]++;
259			clock_gettime(CLOCK_REALTIME_PRECISE, &start);
260		}
261		bp->b_errs = blread(fsreadfd, bp->b_un.b_buf, dblk, size);
262		if (debug) {
263			clock_gettime(CLOCK_REALTIME_PRECISE, &finish);
264			timespecsub(&finish, &start);
265			timespecadd(&readtime[bp->b_type], &finish);
266		}
267		bp->b_bno = dblk;
268		bp->b_size = size;
269	}
270}
271
272void
273flush(int fd, struct bufarea *bp)
274{
275	int i, j;
276
277	if (!bp->b_dirty)
278		return;
279	bp->b_dirty = 0;
280	if (fswritefd < 0) {
281		pfatal("WRITING IN READ_ONLY MODE.\n");
282		return;
283	}
284	if (bp->b_errs != 0)
285		pfatal("WRITING %sZERO'ED BLOCK %lld TO DISK\n",
286		    (bp->b_errs == bp->b_size / dev_bsize) ? "" : "PARTIALLY ",
287		    (long long)bp->b_bno);
288	bp->b_errs = 0;
289	blwrite(fd, bp->b_un.b_buf, bp->b_bno, bp->b_size);
290	if (bp != &sblk)
291		return;
292	for (i = 0, j = 0; i < sblock.fs_cssize; i += sblock.fs_bsize, j++) {
293		blwrite(fswritefd, (char *)sblock.fs_csp + i,
294		    fsbtodb(&sblock, sblock.fs_csaddr + j * sblock.fs_frag),
295		    sblock.fs_cssize - i < sblock.fs_bsize ?
296		    sblock.fs_cssize - i : sblock.fs_bsize);
297	}
298}
299
300void
301rwerror(const char *mesg, ufs2_daddr_t blk)
302{
303
304	if (bkgrdcheck)
305		exit(EEXIT);
306	if (preen == 0)
307		printf("\n");
308	pfatal("CANNOT %s: %ld", mesg, (long)blk);
309	if (reply("CONTINUE") == 0)
310		exit(EEXIT);
311}
312
313void
314ckfini(int markclean)
315{
316	struct bufarea *bp, *nbp;
317	int ofsmodified, cnt;
318
319	if (bkgrdflag) {
320		unlink(snapname);
321		if ((!(sblock.fs_flags & FS_UNCLEAN)) != markclean) {
322			cmd.value = FS_UNCLEAN;
323			cmd.size = markclean ? -1 : 1;
324			if (sysctlbyname("vfs.ffs.setflags", 0, 0,
325			    &cmd, sizeof cmd) == -1)
326				rwerror("SET FILE SYSTEM FLAGS", FS_UNCLEAN);
327			if (!preen) {
328				printf("\n***** FILE SYSTEM MARKED %s *****\n",
329				    markclean ? "CLEAN" : "DIRTY");
330				if (!markclean)
331					rerun = 1;
332			}
333		} else if (!preen && !markclean) {
334			printf("\n***** FILE SYSTEM STILL DIRTY *****\n");
335			rerun = 1;
336		}
337	}
338	if (debug && totalreads > 0)
339		printf("cache with %d buffers missed %ld of %ld (%d%%)\n",
340		    numbufs, totaldiskreads, totalreads,
341		    (int)(totaldiskreads * 100 / totalreads));
342	if (fswritefd < 0) {
343		(void)close(fsreadfd);
344		return;
345	}
346	flush(fswritefd, &sblk);
347	if (havesb && cursnapshot == 0 && sblock.fs_magic == FS_UFS2_MAGIC &&
348	    sblk.b_bno != sblock.fs_sblockloc / dev_bsize &&
349	    !preen && reply("UPDATE STANDARD SUPERBLOCK")) {
350		sblk.b_bno = sblock.fs_sblockloc / dev_bsize;
351		sbdirty();
352		flush(fswritefd, &sblk);
353	}
354	flush(fswritefd, &cgblk);
355	free(cgblk.b_un.b_buf);
356	cnt = 0;
357	TAILQ_FOREACH_REVERSE_SAFE(bp, &bufhead, buflist, b_list, nbp) {
358		TAILQ_REMOVE(&bufhead, bp, b_list);
359		cnt++;
360		flush(fswritefd, bp);
361		free(bp->b_un.b_buf);
362		free((char *)bp);
363	}
364	if (numbufs != cnt)
365		errx(EEXIT, "panic: lost %d buffers", numbufs - cnt);
366	pbp = pdirbp = (struct bufarea *)0;
367	if (cursnapshot == 0 && sblock.fs_clean != markclean) {
368		if ((sblock.fs_clean = markclean) != 0) {
369			sblock.fs_flags &= ~(FS_UNCLEAN | FS_NEEDSFSCK);
370			sblock.fs_pendingblocks = 0;
371			sblock.fs_pendinginodes = 0;
372		}
373		sbdirty();
374		ofsmodified = fsmodified;
375		flush(fswritefd, &sblk);
376		fsmodified = ofsmodified;
377		if (!preen) {
378			printf("\n***** FILE SYSTEM MARKED %s *****\n",
379			    markclean ? "CLEAN" : "DIRTY");
380			if (!markclean)
381				rerun = 1;
382		}
383	} else if (!preen) {
384		if (markclean) {
385			printf("\n***** FILE SYSTEM IS CLEAN *****\n");
386		} else {
387			printf("\n***** FILE SYSTEM STILL DIRTY *****\n");
388			rerun = 1;
389		}
390	}
391	(void)close(fsreadfd);
392	(void)close(fswritefd);
393}
394
395/*
396 * Print out I/O statistics.
397 */
398void
399IOstats(char *what)
400{
401	int i;
402
403	if (debug == 0)
404		return;
405	if (diskreads == 0) {
406		printf("%s: no I/O\n\n", what);
407		return;
408	}
409	if (startpass.tv_sec == 0)
410		startpass = startprog;
411	printf("%s: I/O statistics\n", what);
412	printIOstats();
413	totaldiskreads += diskreads;
414	diskreads = 0;
415	for (i = 0; i < BT_NUMBUFTYPES; i++) {
416		timespecadd(&totalreadtime[i], &readtime[i]);
417		totalreadcnt[i] += readcnt[i];
418		readtime[i].tv_sec = readtime[i].tv_nsec = 0;
419		readcnt[i] = 0;
420	}
421	clock_gettime(CLOCK_REALTIME_PRECISE, &startpass);
422}
423
424void
425finalIOstats(void)
426{
427	int i;
428
429	if (debug == 0)
430		return;
431	printf("Final I/O statistics\n");
432	totaldiskreads += diskreads;
433	diskreads = totaldiskreads;
434	startpass = startprog;
435	for (i = 0; i < BT_NUMBUFTYPES; i++) {
436		timespecadd(&totalreadtime[i], &readtime[i]);
437		totalreadcnt[i] += readcnt[i];
438		readtime[i] = totalreadtime[i];
439		readcnt[i] = totalreadcnt[i];
440	}
441	printIOstats();
442}
443
444static void printIOstats(void)
445{
446	long long msec, totalmsec;
447	int i;
448
449	clock_gettime(CLOCK_REALTIME_PRECISE, &finishpass);
450	timespecsub(&finishpass, &startpass);
451	printf("Running time: %ld msec\n",
452		finishpass.tv_sec * 1000 + finishpass.tv_nsec / 1000000);
453	printf("buffer reads by type:\n");
454	for (totalmsec = 0, i = 0; i < BT_NUMBUFTYPES; i++)
455		totalmsec += readtime[i].tv_sec * 1000 +
456		    readtime[i].tv_nsec / 1000000;
457	if (totalmsec == 0)
458		totalmsec = 1;
459	for (i = 0; i < BT_NUMBUFTYPES; i++) {
460		if (readcnt[i] == 0)
461			continue;
462		msec = readtime[i].tv_sec * 1000 + readtime[i].tv_nsec / 1000000;
463		printf("%21s:%8ld %2ld.%ld%% %8lld msec %2lld.%lld%%\n",
464		    buftype[i], readcnt[i], readcnt[i] * 100 / diskreads,
465		    (readcnt[i] * 1000 / diskreads) % 10, msec,
466		    msec * 100 / totalmsec, (msec * 1000 / totalmsec) % 10);
467	}
468	printf("\n");
469}
470
471int
472blread(int fd, char *buf, ufs2_daddr_t blk, long size)
473{
474	char *cp;
475	int i, errs;
476	off_t offset;
477
478	offset = blk;
479	offset *= dev_bsize;
480	if (bkgrdflag)
481		slowio_start();
482	totalreads++;
483	diskreads++;
484	if (lseek(fd, offset, 0) < 0)
485		rwerror("SEEK BLK", blk);
486	else if (read(fd, buf, (int)size) == size) {
487		if (bkgrdflag)
488			slowio_end();
489		return (0);
490	}
491	rwerror("READ BLK", blk);
492	if (lseek(fd, offset, 0) < 0)
493		rwerror("SEEK BLK", blk);
494	errs = 0;
495	memset(buf, 0, (size_t)size);
496	printf("THE FOLLOWING DISK SECTORS COULD NOT BE READ:");
497	for (cp = buf, i = 0; i < size; i += secsize, cp += secsize) {
498		if (read(fd, cp, (int)secsize) != secsize) {
499			(void)lseek(fd, offset + i + secsize, 0);
500			if (secsize != dev_bsize && dev_bsize != 1)
501				printf(" %jd (%jd),",
502				    (intmax_t)(blk * dev_bsize + i) / secsize,
503				    (intmax_t)blk + i / dev_bsize);
504			else
505				printf(" %jd,", (intmax_t)blk + i / dev_bsize);
506			errs++;
507		}
508	}
509	printf("\n");
510	if (errs)
511		resolved = 0;
512	return (errs);
513}
514
515void
516blwrite(int fd, char *buf, ufs2_daddr_t blk, ssize_t size)
517{
518	int i;
519	char *cp;
520	off_t offset;
521
522	if (fd < 0)
523		return;
524	offset = blk;
525	offset *= dev_bsize;
526	if (lseek(fd, offset, 0) < 0)
527		rwerror("SEEK BLK", blk);
528	else if (write(fd, buf, size) == size) {
529		fsmodified = 1;
530		return;
531	}
532	resolved = 0;
533	rwerror("WRITE BLK", blk);
534	if (lseek(fd, offset, 0) < 0)
535		rwerror("SEEK BLK", blk);
536	printf("THE FOLLOWING SECTORS COULD NOT BE WRITTEN:");
537	for (cp = buf, i = 0; i < size; i += dev_bsize, cp += dev_bsize)
538		if (write(fd, cp, dev_bsize) != dev_bsize) {
539			(void)lseek(fd, offset + i + dev_bsize, 0);
540			printf(" %jd,", (intmax_t)blk + i / dev_bsize);
541		}
542	printf("\n");
543	return;
544}
545
546void
547blerase(int fd, ufs2_daddr_t blk, long size)
548{
549	off_t ioarg[2];
550
551	if (fd < 0)
552		return;
553	ioarg[0] = blk * dev_bsize;
554	ioarg[1] = size;
555	ioctl(fd, DIOCGDELETE, ioarg);
556	/* we don't really care if we succeed or not */
557	return;
558}
559
560/*
561 * Verify cylinder group's magic number and other parameters.  If the
562 * test fails, offer an option to rebuild the whole cylinder group.
563 */
564int
565check_cgmagic(int cg, struct cg *cgp)
566{
567
568	/*
569	 * Extended cylinder group checks.
570	 */
571	if (cg_chkmagic(cgp) &&
572	    ((sblock.fs_magic == FS_UFS1_MAGIC &&
573	      cgp->cg_old_niblk == sblock.fs_ipg &&
574	      cgp->cg_ndblk <= sblock.fs_fpg &&
575	      cgp->cg_old_ncyl <= sblock.fs_old_cpg) ||
576	     (sblock.fs_magic == FS_UFS2_MAGIC &&
577	      cgp->cg_niblk == sblock.fs_ipg &&
578	      cgp->cg_ndblk <= sblock.fs_fpg &&
579	      cgp->cg_initediblk <= sblock.fs_ipg))) {
580		return (1);
581	}
582	pfatal("CYLINDER GROUP %d: BAD MAGIC NUMBER", cg);
583	if (!reply("REBUILD CYLINDER GROUP")) {
584		printf("YOU WILL NEED TO RERUN FSCK.\n");
585		rerun = 1;
586		return (1);
587	}
588	/*
589	 * Zero out the cylinder group and then initialize critical fields.
590	 * Bit maps and summaries will be recalculated by later passes.
591	 */
592	memset(cgp, 0, (size_t)sblock.fs_cgsize);
593	cgp->cg_magic = CG_MAGIC;
594	cgp->cg_cgx = cg;
595	cgp->cg_niblk = sblock.fs_ipg;
596	cgp->cg_initediblk = sblock.fs_ipg < 2 * INOPB(&sblock) ?
597	    sblock.fs_ipg : 2 * INOPB(&sblock);
598	if (cgbase(&sblock, cg) + sblock.fs_fpg < sblock.fs_size)
599		cgp->cg_ndblk = sblock.fs_fpg;
600	else
601		cgp->cg_ndblk = sblock.fs_size - cgbase(&sblock, cg);
602	cgp->cg_iusedoff = &cgp->cg_space[0] - (u_char *)(&cgp->cg_firstfield);
603	if (sblock.fs_magic == FS_UFS1_MAGIC) {
604		cgp->cg_niblk = 0;
605		cgp->cg_initediblk = 0;
606		cgp->cg_old_ncyl = sblock.fs_old_cpg;
607		cgp->cg_old_niblk = sblock.fs_ipg;
608		cgp->cg_old_btotoff = cgp->cg_iusedoff;
609		cgp->cg_old_boff = cgp->cg_old_btotoff +
610		    sblock.fs_old_cpg * sizeof(int32_t);
611		cgp->cg_iusedoff = cgp->cg_old_boff +
612		    sblock.fs_old_cpg * sizeof(u_int16_t);
613	}
614	cgp->cg_freeoff = cgp->cg_iusedoff + howmany(sblock.fs_ipg, CHAR_BIT);
615	cgp->cg_nextfreeoff = cgp->cg_freeoff + howmany(sblock.fs_fpg,CHAR_BIT);
616	if (sblock.fs_contigsumsize > 0) {
617		cgp->cg_nclusterblks = cgp->cg_ndblk / sblock.fs_frag;
618		cgp->cg_clustersumoff =
619		    roundup(cgp->cg_nextfreeoff, sizeof(u_int32_t));
620		cgp->cg_clustersumoff -= sizeof(u_int32_t);
621		cgp->cg_clusteroff = cgp->cg_clustersumoff +
622		    (sblock.fs_contigsumsize + 1) * sizeof(u_int32_t);
623		cgp->cg_nextfreeoff = cgp->cg_clusteroff +
624		    howmany(fragstoblks(&sblock, sblock.fs_fpg), CHAR_BIT);
625	}
626	cgdirty();
627	return (0);
628}
629
630/*
631 * allocate a data block with the specified number of fragments
632 */
633ufs2_daddr_t
634allocblk(long frags)
635{
636	int i, j, k, cg, baseblk;
637	struct cg *cgp = &cgrp;
638
639	if (frags <= 0 || frags > sblock.fs_frag)
640		return (0);
641	for (i = 0; i < maxfsblock - sblock.fs_frag; i += sblock.fs_frag) {
642		for (j = 0; j <= sblock.fs_frag - frags; j++) {
643			if (testbmap(i + j))
644				continue;
645			for (k = 1; k < frags; k++)
646				if (testbmap(i + j + k))
647					break;
648			if (k < frags) {
649				j += k;
650				continue;
651			}
652			cg = dtog(&sblock, i + j);
653			getblk(&cgblk, cgtod(&sblock, cg), sblock.fs_cgsize);
654			if (!check_cgmagic(cg, cgp))
655				return (0);
656			baseblk = dtogd(&sblock, i + j);
657			for (k = 0; k < frags; k++) {
658				setbmap(i + j + k);
659				clrbit(cg_blksfree(cgp), baseblk + k);
660			}
661			n_blks += frags;
662			if (frags == sblock.fs_frag)
663				cgp->cg_cs.cs_nbfree--;
664			else
665				cgp->cg_cs.cs_nffree -= frags;
666			cgdirty();
667			return (i + j);
668		}
669	}
670	return (0);
671}
672
673/*
674 * Free a previously allocated block
675 */
676void
677freeblk(ufs2_daddr_t blkno, long frags)
678{
679	struct inodesc idesc;
680
681	idesc.id_blkno = blkno;
682	idesc.id_numfrags = frags;
683	(void)pass4check(&idesc);
684}
685
686/* Slow down IO so as to leave some disk bandwidth for other processes */
687void
688slowio_start()
689{
690
691	/* Delay one in every 8 operations */
692	slowio_pollcnt = (slowio_pollcnt + 1) & 7;
693	if (slowio_pollcnt == 0) {
694		gettimeofday(&slowio_starttime, NULL);
695	}
696}
697
698void
699slowio_end()
700{
701	struct timeval tv;
702	int delay_usec;
703
704	if (slowio_pollcnt != 0)
705		return;
706
707	/* Update the slowdown interval. */
708	gettimeofday(&tv, NULL);
709	delay_usec = (tv.tv_sec - slowio_starttime.tv_sec) * 1000000 +
710	    (tv.tv_usec - slowio_starttime.tv_usec);
711	if (delay_usec < 64)
712		delay_usec = 64;
713	if (delay_usec > 2500000)
714		delay_usec = 2500000;
715	slowio_delay_usec = (slowio_delay_usec * 63 + delay_usec) >> 6;
716	/* delay by 8 times the average IO delay */
717	if (slowio_delay_usec > 64)
718		usleep(slowio_delay_usec * 8);
719}
720
721/*
722 * Find a pathname
723 */
724void
725getpathname(char *namebuf, ino_t curdir, ino_t ino)
726{
727	int len;
728	char *cp;
729	struct inodesc idesc;
730	static int busy = 0;
731
732	if (curdir == ino && ino == ROOTINO) {
733		(void)strcpy(namebuf, "/");
734		return;
735	}
736	if (busy || !INO_IS_DVALID(curdir)) {
737		(void)strcpy(namebuf, "?");
738		return;
739	}
740	busy = 1;
741	memset(&idesc, 0, sizeof(struct inodesc));
742	idesc.id_type = DATA;
743	idesc.id_fix = IGNORE;
744	cp = &namebuf[MAXPATHLEN - 1];
745	*cp = '\0';
746	if (curdir != ino) {
747		idesc.id_parent = curdir;
748		goto namelookup;
749	}
750	while (ino != ROOTINO) {
751		idesc.id_number = ino;
752		idesc.id_func = findino;
753		idesc.id_name = strdup("..");
754		if ((ckinode(ginode(ino), &idesc) & FOUND) == 0)
755			break;
756	namelookup:
757		idesc.id_number = idesc.id_parent;
758		idesc.id_parent = ino;
759		idesc.id_func = findname;
760		idesc.id_name = namebuf;
761		if ((ckinode(ginode(idesc.id_number), &idesc)&FOUND) == 0)
762			break;
763		len = strlen(namebuf);
764		cp -= len;
765		memmove(cp, namebuf, (size_t)len);
766		*--cp = '/';
767		if (cp < &namebuf[MAXNAMLEN])
768			break;
769		ino = idesc.id_number;
770	}
771	busy = 0;
772	if (ino != ROOTINO)
773		*--cp = '?';
774	memmove(namebuf, cp, (size_t)(&namebuf[MAXPATHLEN] - cp));
775}
776
777void
778catch(int sig __unused)
779{
780
781	ckfini(0);
782	exit(12);
783}
784
785/*
786 * When preening, allow a single quit to signal
787 * a special exit after file system checks complete
788 * so that reboot sequence may be interrupted.
789 */
790void
791catchquit(int sig __unused)
792{
793	printf("returning to single-user after file system check\n");
794	returntosingle = 1;
795	(void)signal(SIGQUIT, SIG_DFL);
796}
797
798/*
799 * determine whether an inode should be fixed.
800 */
801int
802dofix(struct inodesc *idesc, const char *msg)
803{
804
805	switch (idesc->id_fix) {
806
807	case DONTKNOW:
808		if (idesc->id_type == DATA)
809			direrror(idesc->id_number, msg);
810		else
811			pwarn("%s", msg);
812		if (preen) {
813			printf(" (SALVAGED)\n");
814			idesc->id_fix = FIX;
815			return (ALTERED);
816		}
817		if (reply("SALVAGE") == 0) {
818			idesc->id_fix = NOFIX;
819			return (0);
820		}
821		idesc->id_fix = FIX;
822		return (ALTERED);
823
824	case FIX:
825		return (ALTERED);
826
827	case NOFIX:
828	case IGNORE:
829		return (0);
830
831	default:
832		errx(EEXIT, "UNKNOWN INODESC FIX MODE %d", idesc->id_fix);
833	}
834	/* NOTREACHED */
835	return (0);
836}
837
838#include <stdarg.h>
839
840/*
841 * An unexpected inconsistency occurred.
842 * Die if preening or file system is running with soft dependency protocol,
843 * otherwise just print message and continue.
844 */
845void
846pfatal(const char *fmt, ...)
847{
848	va_list ap;
849	va_start(ap, fmt);
850	if (!preen) {
851		(void)vfprintf(stdout, fmt, ap);
852		va_end(ap);
853		if (usedsoftdep)
854			(void)fprintf(stdout,
855			    "\nUNEXPECTED SOFT UPDATE INCONSISTENCY\n");
856		/*
857		 * Force foreground fsck to clean up inconsistency.
858		 */
859		if (bkgrdflag) {
860			cmd.value = FS_NEEDSFSCK;
861			cmd.size = 1;
862			if (sysctlbyname("vfs.ffs.setflags", 0, 0,
863			    &cmd, sizeof cmd) == -1)
864				pwarn("CANNOT SET FS_NEEDSFSCK FLAG\n");
865			fprintf(stdout, "CANNOT RUN IN BACKGROUND\n");
866			ckfini(0);
867			exit(EEXIT);
868		}
869		return;
870	}
871	if (cdevname == NULL)
872		cdevname = strdup("fsck");
873	(void)fprintf(stdout, "%s: ", cdevname);
874	(void)vfprintf(stdout, fmt, ap);
875	(void)fprintf(stdout,
876	    "\n%s: UNEXPECTED%sINCONSISTENCY; RUN fsck MANUALLY.\n",
877	    cdevname, usedsoftdep ? " SOFT UPDATE " : " ");
878	/*
879	 * Force foreground fsck to clean up inconsistency.
880	 */
881	if (bkgrdflag) {
882		cmd.value = FS_NEEDSFSCK;
883		cmd.size = 1;
884		if (sysctlbyname("vfs.ffs.setflags", 0, 0,
885		    &cmd, sizeof cmd) == -1)
886			pwarn("CANNOT SET FS_NEEDSFSCK FLAG\n");
887	}
888	ckfini(0);
889	exit(EEXIT);
890}
891
892/*
893 * Pwarn just prints a message when not preening or running soft dependency
894 * protocol, or a warning (preceded by filename) when preening.
895 */
896void
897pwarn(const char *fmt, ...)
898{
899	va_list ap;
900	va_start(ap, fmt);
901	if (preen)
902		(void)fprintf(stdout, "%s: ", cdevname);
903	(void)vfprintf(stdout, fmt, ap);
904	va_end(ap);
905}
906
907/*
908 * Stub for routines from kernel.
909 */
910void
911panic(const char *fmt, ...)
912{
913	va_list ap;
914	va_start(ap, fmt);
915	pfatal("INTERNAL INCONSISTENCY:");
916	(void)vfprintf(stdout, fmt, ap);
917	va_end(ap);
918	exit(EEXIT);
919}
920