xref: /freebsd/sbin/fsck_ffs/fsutil.c (revision c1cdf6a42f0d951ba720688dfc6ce07608b02f6e)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1980, 1986, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  */
31 
32 #if 0
33 #ifndef lint
34 static const char sccsid[] = "@(#)utilities.c	8.6 (Berkeley) 5/19/95";
35 #endif /* not lint */
36 #endif
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
39 
40 #include <sys/param.h>
41 #include <sys/time.h>
42 #include <sys/types.h>
43 #include <sys/sysctl.h>
44 #include <sys/disk.h>
45 #include <sys/disklabel.h>
46 #include <sys/ioctl.h>
47 #include <sys/stat.h>
48 
49 #include <ufs/ufs/dinode.h>
50 #include <ufs/ufs/dir.h>
51 #include <ufs/ffs/fs.h>
52 
53 #include <err.h>
54 #include <errno.h>
55 #include <string.h>
56 #include <ctype.h>
57 #include <fstab.h>
58 #include <stdint.h>
59 #include <stdio.h>
60 #include <stdlib.h>
61 #include <time.h>
62 #include <unistd.h>
63 #include <libufs.h>
64 
65 #include "fsck.h"
66 
67 static void slowio_start(void);
68 static void slowio_end(void);
69 static void printIOstats(void);
70 
71 static long diskreads, totaldiskreads, totalreads; /* Disk cache statistics */
72 static struct timespec startpass, finishpass;
73 struct timeval slowio_starttime;
74 int slowio_delay_usec = 10000;	/* Initial IO delay for background fsck */
75 int slowio_pollcnt;
76 static struct bufarea cgblk;	/* backup buffer for cylinder group blocks */
77 static TAILQ_HEAD(buflist, bufarea) bufhead;	/* head of buffer cache list */
78 static int numbufs;				/* size of buffer cache */
79 static char *buftype[BT_NUMBUFTYPES] = BT_NAMES;
80 static struct bufarea *cgbufs;	/* header for cylinder group cache */
81 static int flushtries;		/* number of tries to reclaim memory */
82 
83 void
84 fsutilinit(void)
85 {
86 	diskreads = totaldiskreads = totalreads = 0;
87 	bzero(&startpass, sizeof(struct timespec));
88 	bzero(&finishpass, sizeof(struct timespec));
89 	bzero(&slowio_starttime, sizeof(struct timeval));
90 	slowio_delay_usec = 10000;
91 	slowio_pollcnt = 0;
92 	bzero(&cgblk, sizeof(struct bufarea));
93 	TAILQ_INIT(&bufhead);
94 	numbufs = 0;
95 	/* buftype ? */
96 	cgbufs = NULL;
97 	flushtries = 0;
98 }
99 
100 int
101 ftypeok(union dinode *dp)
102 {
103 	switch (DIP(dp, di_mode) & IFMT) {
104 
105 	case IFDIR:
106 	case IFREG:
107 	case IFBLK:
108 	case IFCHR:
109 	case IFLNK:
110 	case IFSOCK:
111 	case IFIFO:
112 		return (1);
113 
114 	default:
115 		if (debug)
116 			printf("bad file type 0%o\n", DIP(dp, di_mode));
117 		return (0);
118 	}
119 }
120 
121 int
122 reply(const char *question)
123 {
124 	int persevere;
125 	char c;
126 
127 	if (preen)
128 		pfatal("INTERNAL ERROR: GOT TO reply()");
129 	persevere = !strcmp(question, "CONTINUE");
130 	printf("\n");
131 	if (!persevere && (nflag || (fswritefd < 0 && bkgrdflag == 0))) {
132 		printf("%s? no\n\n", question);
133 		resolved = 0;
134 		return (0);
135 	}
136 	if (yflag || (persevere && nflag)) {
137 		printf("%s? yes\n\n", question);
138 		return (1);
139 	}
140 	do	{
141 		printf("%s? [yn] ", question);
142 		(void) fflush(stdout);
143 		c = getc(stdin);
144 		while (c != '\n' && getc(stdin) != '\n') {
145 			if (feof(stdin)) {
146 				resolved = 0;
147 				return (0);
148 			}
149 		}
150 	} while (c != 'y' && c != 'Y' && c != 'n' && c != 'N');
151 	printf("\n");
152 	if (c == 'y' || c == 'Y')
153 		return (1);
154 	resolved = 0;
155 	return (0);
156 }
157 
158 /*
159  * Look up state information for an inode.
160  */
161 struct inostat *
162 inoinfo(ino_t inum)
163 {
164 	static struct inostat unallocated = { USTATE, 0, 0 };
165 	struct inostatlist *ilp;
166 	int iloff;
167 
168 	if (inum > maxino)
169 		errx(EEXIT, "inoinfo: inumber %ju out of range",
170 		    (uintmax_t)inum);
171 	ilp = &inostathead[inum / sblock.fs_ipg];
172 	iloff = inum % sblock.fs_ipg;
173 	if (iloff >= ilp->il_numalloced)
174 		return (&unallocated);
175 	return (&ilp->il_stat[iloff]);
176 }
177 
178 /*
179  * Malloc buffers and set up cache.
180  */
181 void
182 bufinit(void)
183 {
184 	struct bufarea *bp;
185 	long bufcnt, i;
186 	char *bufp;
187 
188 	pbp = pdirbp = (struct bufarea *)0;
189 	bufp = Malloc((unsigned int)sblock.fs_bsize);
190 	if (bufp == NULL)
191 		errx(EEXIT, "cannot allocate buffer pool");
192 	cgblk.b_un.b_buf = bufp;
193 	initbarea(&cgblk, BT_CYLGRP);
194 	TAILQ_INIT(&bufhead);
195 	bufcnt = MAXBUFS;
196 	if (bufcnt < MINBUFS)
197 		bufcnt = MINBUFS;
198 	for (i = 0; i < bufcnt; i++) {
199 		bp = (struct bufarea *)Malloc(sizeof(struct bufarea));
200 		bufp = Malloc((unsigned int)sblock.fs_bsize);
201 		if (bp == NULL || bufp == NULL) {
202 			if (i >= MINBUFS)
203 				break;
204 			errx(EEXIT, "cannot allocate buffer pool");
205 		}
206 		bp->b_un.b_buf = bufp;
207 		TAILQ_INSERT_HEAD(&bufhead, bp, b_list);
208 		initbarea(bp, BT_UNKNOWN);
209 	}
210 	numbufs = i;	/* save number of buffers */
211 	for (i = 0; i < BT_NUMBUFTYPES; i++) {
212 		readtime[i].tv_sec = totalreadtime[i].tv_sec = 0;
213 		readtime[i].tv_nsec = totalreadtime[i].tv_nsec = 0;
214 		readcnt[i] = totalreadcnt[i] = 0;
215 	}
216 }
217 
218 /*
219  * Manage cylinder group buffers.
220  */
221 static struct bufarea *cgbufs;	/* header for cylinder group cache */
222 static int flushtries;		/* number of tries to reclaim memory */
223 
224 struct bufarea *
225 cglookup(int cg)
226 {
227 	struct bufarea *cgbp;
228 	struct cg *cgp;
229 
230 	if (cgbufs == NULL) {
231 		cgbufs = calloc(sblock.fs_ncg, sizeof(struct bufarea));
232 		if (cgbufs == NULL)
233 			errx(EEXIT, "cannot allocate cylinder group buffers");
234 	}
235 	cgbp = &cgbufs[cg];
236 	if (cgbp->b_un.b_cg != NULL)
237 		return (cgbp);
238 	cgp = NULL;
239 	if (flushtries == 0)
240 		cgp = malloc((unsigned int)sblock.fs_cgsize);
241 	if (cgp == NULL) {
242 		getblk(&cgblk, cgtod(&sblock, cg), sblock.fs_cgsize);
243 		return (&cgblk);
244 	}
245 	cgbp->b_un.b_cg = cgp;
246 	initbarea(cgbp, BT_CYLGRP);
247 	getblk(cgbp, cgtod(&sblock, cg), sblock.fs_cgsize);
248 	return (cgbp);
249 }
250 
251 /*
252  * Attempt to flush a cylinder group cache entry.
253  * Return whether the flush was successful.
254  */
255 int
256 flushentry(void)
257 {
258 	struct bufarea *cgbp;
259 
260 	if (flushtries == sblock.fs_ncg || cgbufs == NULL)
261 		return (0);
262 	cgbp = &cgbufs[flushtries++];
263 	if (cgbp->b_un.b_cg == NULL)
264 		return (0);
265 	flush(fswritefd, cgbp);
266 	free(cgbp->b_un.b_buf);
267 	cgbp->b_un.b_buf = NULL;
268 	return (1);
269 }
270 
271 /*
272  * Manage a cache of directory blocks.
273  */
274 struct bufarea *
275 getdatablk(ufs2_daddr_t blkno, long size, int type)
276 {
277 	struct bufarea *bp;
278 
279 	TAILQ_FOREACH(bp, &bufhead, b_list)
280 		if (bp->b_bno == fsbtodb(&sblock, blkno))
281 			goto foundit;
282 	TAILQ_FOREACH_REVERSE(bp, &bufhead, buflist, b_list)
283 		if ((bp->b_flags & B_INUSE) == 0)
284 			break;
285 	if (bp == NULL)
286 		errx(EEXIT, "deadlocked buffer pool");
287 	bp->b_type = type;
288 	getblk(bp, blkno, size);
289 	/* fall through */
290 foundit:
291 	if (debug && bp->b_type != type)
292 		printf("Buffer type changed from %s to %s\n",
293 		    buftype[bp->b_type], buftype[type]);
294 	TAILQ_REMOVE(&bufhead, bp, b_list);
295 	TAILQ_INSERT_HEAD(&bufhead, bp, b_list);
296 	bp->b_flags |= B_INUSE;
297 	return (bp);
298 }
299 
300 void
301 getblk(struct bufarea *bp, ufs2_daddr_t blk, long size)
302 {
303 	ufs2_daddr_t dblk;
304 	struct timespec start, finish;
305 
306 	dblk = fsbtodb(&sblock, blk);
307 	if (bp->b_bno == dblk) {
308 		totalreads++;
309 	} else {
310 		flush(fswritefd, bp);
311 		if (debug) {
312 			readcnt[bp->b_type]++;
313 			clock_gettime(CLOCK_REALTIME_PRECISE, &start);
314 		}
315 		bp->b_errs = blread(fsreadfd, bp->b_un.b_buf, dblk, size);
316 		if (debug) {
317 			clock_gettime(CLOCK_REALTIME_PRECISE, &finish);
318 			timespecsub(&finish, &start, &finish);
319 			timespecadd(&readtime[bp->b_type], &finish,
320 			    &readtime[bp->b_type]);
321 		}
322 		bp->b_bno = dblk;
323 		bp->b_size = size;
324 	}
325 }
326 
327 void
328 flush(int fd, struct bufarea *bp)
329 {
330 
331 	if (!bp->b_dirty)
332 		return;
333 	bp->b_dirty = 0;
334 	if (fswritefd < 0) {
335 		pfatal("WRITING IN READ_ONLY MODE.\n");
336 		return;
337 	}
338 	if (bp->b_errs != 0)
339 		pfatal("WRITING %sZERO'ED BLOCK %lld TO DISK\n",
340 		    (bp->b_errs == bp->b_size / dev_bsize) ? "" : "PARTIALLY ",
341 		    (long long)bp->b_bno);
342 	bp->b_errs = 0;
343 	/*
344 	 * Write using the appropriate function.
345 	 */
346 	switch (bp->b_type) {
347 	case BT_SUPERBLK:
348 		if (bp != &sblk)
349 			pfatal("BUFFER %p DOES NOT MATCH SBLK %p\n",
350 			    bp, &sblk);
351 		if (sbput(fd, (struct fs *)bp->b_un.b_buf, 0) == 0)
352 			fsmodified = 1;
353 		break;
354 	case BT_CYLGRP:
355 		if (cgput(&disk, (struct cg *)bp->b_un.b_buf) == 0)
356 			fsmodified = 1;
357 		break;
358 	default:
359 		blwrite(fd, bp->b_un.b_buf, bp->b_bno, bp->b_size);
360 		break;
361 	}
362 }
363 
364 void
365 rwerror(const char *mesg, ufs2_daddr_t blk)
366 {
367 
368 	if (bkgrdcheck)
369 		exit(EEXIT);
370 	if (preen == 0)
371 		printf("\n");
372 	pfatal("CANNOT %s: %ld", mesg, (long)blk);
373 	if (reply("CONTINUE") == 0)
374 		exit(EEXIT);
375 }
376 
377 void
378 ckfini(int markclean)
379 {
380 	struct bufarea *bp, *nbp;
381 	int ofsmodified, cnt;
382 
383 	if (bkgrdflag) {
384 		unlink(snapname);
385 		if ((!(sblock.fs_flags & FS_UNCLEAN)) != markclean) {
386 			cmd.value = FS_UNCLEAN;
387 			cmd.size = markclean ? -1 : 1;
388 			if (sysctlbyname("vfs.ffs.setflags", 0, 0,
389 			    &cmd, sizeof cmd) == -1)
390 				rwerror("SET FILE SYSTEM FLAGS", FS_UNCLEAN);
391 			if (!preen) {
392 				printf("\n***** FILE SYSTEM MARKED %s *****\n",
393 				    markclean ? "CLEAN" : "DIRTY");
394 				if (!markclean)
395 					rerun = 1;
396 			}
397 		} else if (!preen && !markclean) {
398 			printf("\n***** FILE SYSTEM STILL DIRTY *****\n");
399 			rerun = 1;
400 		}
401 	}
402 	if (debug && totalreads > 0)
403 		printf("cache with %d buffers missed %ld of %ld (%d%%)\n",
404 		    numbufs, totaldiskreads, totalreads,
405 		    (int)(totaldiskreads * 100 / totalreads));
406 	if (fswritefd < 0) {
407 		(void)close(fsreadfd);
408 		return;
409 	}
410 	flush(fswritefd, &sblk);
411 	if (havesb && cursnapshot == 0 && sblock.fs_magic == FS_UFS2_MAGIC &&
412 	    sblk.b_bno != sblock.fs_sblockloc / dev_bsize &&
413 	    !preen && reply("UPDATE STANDARD SUPERBLOCK")) {
414 		/* Change the write destination to standard superblock */
415 		sblock.fs_sblockactualloc = sblock.fs_sblockloc;
416 		sblk.b_bno = sblock.fs_sblockloc / dev_bsize;
417 		sbdirty();
418 		flush(fswritefd, &sblk);
419 	}
420 	flush(fswritefd, &cgblk);
421 	free(cgblk.b_un.b_buf);
422 	cnt = 0;
423 	TAILQ_FOREACH_REVERSE_SAFE(bp, &bufhead, buflist, b_list, nbp) {
424 		TAILQ_REMOVE(&bufhead, bp, b_list);
425 		cnt++;
426 		flush(fswritefd, bp);
427 		free(bp->b_un.b_buf);
428 		free((char *)bp);
429 	}
430 	if (numbufs != cnt)
431 		errx(EEXIT, "panic: lost %d buffers", numbufs - cnt);
432 	if (cgbufs != NULL) {
433 		for (cnt = 0; cnt < sblock.fs_ncg; cnt++) {
434 			if (cgbufs[cnt].b_un.b_cg == NULL)
435 				continue;
436 			flush(fswritefd, &cgbufs[cnt]);
437 			free(cgbufs[cnt].b_un.b_cg);
438 		}
439 		free(cgbufs);
440 	}
441 	pbp = pdirbp = (struct bufarea *)0;
442 	if (cursnapshot == 0 && sblock.fs_clean != markclean) {
443 		if ((sblock.fs_clean = markclean) != 0) {
444 			sblock.fs_flags &= ~(FS_UNCLEAN | FS_NEEDSFSCK);
445 			sblock.fs_pendingblocks = 0;
446 			sblock.fs_pendinginodes = 0;
447 		}
448 		sbdirty();
449 		ofsmodified = fsmodified;
450 		flush(fswritefd, &sblk);
451 		fsmodified = ofsmodified;
452 		if (!preen) {
453 			printf("\n***** FILE SYSTEM MARKED %s *****\n",
454 			    markclean ? "CLEAN" : "DIRTY");
455 			if (!markclean)
456 				rerun = 1;
457 		}
458 	} else if (!preen) {
459 		if (markclean) {
460 			printf("\n***** FILE SYSTEM IS CLEAN *****\n");
461 		} else {
462 			printf("\n***** FILE SYSTEM STILL DIRTY *****\n");
463 			rerun = 1;
464 		}
465 	}
466 	(void)close(fsreadfd);
467 	(void)close(fswritefd);
468 }
469 
470 /*
471  * Print out I/O statistics.
472  */
473 void
474 IOstats(char *what)
475 {
476 	int i;
477 
478 	if (debug == 0)
479 		return;
480 	if (diskreads == 0) {
481 		printf("%s: no I/O\n\n", what);
482 		return;
483 	}
484 	if (startpass.tv_sec == 0)
485 		startpass = startprog;
486 	printf("%s: I/O statistics\n", what);
487 	printIOstats();
488 	totaldiskreads += diskreads;
489 	diskreads = 0;
490 	for (i = 0; i < BT_NUMBUFTYPES; i++) {
491 		timespecadd(&totalreadtime[i], &readtime[i], &totalreadtime[i]);
492 		totalreadcnt[i] += readcnt[i];
493 		readtime[i].tv_sec = readtime[i].tv_nsec = 0;
494 		readcnt[i] = 0;
495 	}
496 	clock_gettime(CLOCK_REALTIME_PRECISE, &startpass);
497 }
498 
499 void
500 finalIOstats(void)
501 {
502 	int i;
503 
504 	if (debug == 0)
505 		return;
506 	printf("Final I/O statistics\n");
507 	totaldiskreads += diskreads;
508 	diskreads = totaldiskreads;
509 	startpass = startprog;
510 	for (i = 0; i < BT_NUMBUFTYPES; i++) {
511 		timespecadd(&totalreadtime[i], &readtime[i], &totalreadtime[i]);
512 		totalreadcnt[i] += readcnt[i];
513 		readtime[i] = totalreadtime[i];
514 		readcnt[i] = totalreadcnt[i];
515 	}
516 	printIOstats();
517 }
518 
519 static void printIOstats(void)
520 {
521 	long long msec, totalmsec;
522 	int i;
523 
524 	clock_gettime(CLOCK_REALTIME_PRECISE, &finishpass);
525 	timespecsub(&finishpass, &startpass, &finishpass);
526 	printf("Running time: %jd.%03ld sec\n",
527 		(intmax_t)finishpass.tv_sec, finishpass.tv_nsec / 1000000);
528 	printf("buffer reads by type:\n");
529 	for (totalmsec = 0, i = 0; i < BT_NUMBUFTYPES; i++)
530 		totalmsec += readtime[i].tv_sec * 1000 +
531 		    readtime[i].tv_nsec / 1000000;
532 	if (totalmsec == 0)
533 		totalmsec = 1;
534 	for (i = 0; i < BT_NUMBUFTYPES; i++) {
535 		if (readcnt[i] == 0)
536 			continue;
537 		msec =
538 		    readtime[i].tv_sec * 1000 + readtime[i].tv_nsec / 1000000;
539 		printf("%21s:%8ld %2ld.%ld%% %4jd.%03ld sec %2lld.%lld%%\n",
540 		    buftype[i], readcnt[i], readcnt[i] * 100 / diskreads,
541 		    (readcnt[i] * 1000 / diskreads) % 10,
542 		    (intmax_t)readtime[i].tv_sec, readtime[i].tv_nsec / 1000000,
543 		    msec * 100 / totalmsec, (msec * 1000 / totalmsec) % 10);
544 	}
545 	printf("\n");
546 }
547 
548 int
549 blread(int fd, char *buf, ufs2_daddr_t blk, long size)
550 {
551 	char *cp;
552 	int i, errs;
553 	off_t offset;
554 
555 	offset = blk;
556 	offset *= dev_bsize;
557 	if (bkgrdflag)
558 		slowio_start();
559 	totalreads++;
560 	diskreads++;
561 	if (lseek(fd, offset, 0) < 0)
562 		rwerror("SEEK BLK", blk);
563 	else if (read(fd, buf, (int)size) == size) {
564 		if (bkgrdflag)
565 			slowio_end();
566 		return (0);
567 	}
568 
569 	/*
570 	 * This is handled specially here instead of in rwerror because
571 	 * rwerror is used for all sorts of errors, not just true read/write
572 	 * errors.  It should be refactored and fixed.
573 	 */
574 	if (surrender) {
575 		pfatal("CANNOT READ_BLK: %ld", (long)blk);
576 		errx(EEXIT, "ABORTING DUE TO READ ERRORS");
577 	} else
578 		rwerror("READ BLK", blk);
579 
580 	if (lseek(fd, offset, 0) < 0)
581 		rwerror("SEEK BLK", blk);
582 	errs = 0;
583 	memset(buf, 0, (size_t)size);
584 	printf("THE FOLLOWING DISK SECTORS COULD NOT BE READ:");
585 	for (cp = buf, i = 0; i < size; i += secsize, cp += secsize) {
586 		if (read(fd, cp, (int)secsize) != secsize) {
587 			(void)lseek(fd, offset + i + secsize, 0);
588 			if (secsize != dev_bsize && dev_bsize != 1)
589 				printf(" %jd (%jd),",
590 				    (intmax_t)(blk * dev_bsize + i) / secsize,
591 				    (intmax_t)blk + i / dev_bsize);
592 			else
593 				printf(" %jd,", (intmax_t)blk + i / dev_bsize);
594 			errs++;
595 		}
596 	}
597 	printf("\n");
598 	if (errs)
599 		resolved = 0;
600 	return (errs);
601 }
602 
603 void
604 blwrite(int fd, char *buf, ufs2_daddr_t blk, ssize_t size)
605 {
606 	int i;
607 	char *cp;
608 	off_t offset;
609 
610 	if (fd < 0)
611 		return;
612 	offset = blk;
613 	offset *= dev_bsize;
614 	if (lseek(fd, offset, 0) < 0)
615 		rwerror("SEEK BLK", blk);
616 	else if (write(fd, buf, size) == size) {
617 		fsmodified = 1;
618 		return;
619 	}
620 	resolved = 0;
621 	rwerror("WRITE BLK", blk);
622 	if (lseek(fd, offset, 0) < 0)
623 		rwerror("SEEK BLK", blk);
624 	printf("THE FOLLOWING SECTORS COULD NOT BE WRITTEN:");
625 	for (cp = buf, i = 0; i < size; i += dev_bsize, cp += dev_bsize)
626 		if (write(fd, cp, dev_bsize) != dev_bsize) {
627 			(void)lseek(fd, offset + i + dev_bsize, 0);
628 			printf(" %jd,", (intmax_t)blk + i / dev_bsize);
629 		}
630 	printf("\n");
631 	return;
632 }
633 
634 void
635 blerase(int fd, ufs2_daddr_t blk, long size)
636 {
637 	off_t ioarg[2];
638 
639 	if (fd < 0)
640 		return;
641 	ioarg[0] = blk * dev_bsize;
642 	ioarg[1] = size;
643 	ioctl(fd, DIOCGDELETE, ioarg);
644 	/* we don't really care if we succeed or not */
645 	return;
646 }
647 
648 /*
649  * Fill a contiguous region with all-zeroes.  Note ZEROBUFSIZE is by
650  * definition a multiple of dev_bsize.
651  */
652 void
653 blzero(int fd, ufs2_daddr_t blk, long size)
654 {
655 	static char *zero;
656 	off_t offset, len;
657 
658 	if (fd < 0)
659 		return;
660 	if (zero == NULL) {
661 		zero = calloc(ZEROBUFSIZE, 1);
662 		if (zero == NULL)
663 			errx(EEXIT, "cannot allocate buffer pool");
664 	}
665 	offset = blk * dev_bsize;
666 	if (lseek(fd, offset, 0) < 0)
667 		rwerror("SEEK BLK", blk);
668 	while (size > 0) {
669 		len = MIN(ZEROBUFSIZE, size);
670 		if (write(fd, zero, len) != len)
671 			rwerror("WRITE BLK", blk);
672 		blk += len / dev_bsize;
673 		size -= len;
674 	}
675 }
676 
677 /*
678  * Verify cylinder group's magic number and other parameters.  If the
679  * test fails, offer an option to rebuild the whole cylinder group.
680  */
681 int
682 check_cgmagic(int cg, struct bufarea *cgbp)
683 {
684 	struct cg *cgp = cgbp->b_un.b_cg;
685 
686 	/*
687 	 * Extended cylinder group checks.
688 	 */
689 	if (cg_chkmagic(cgp) &&
690 	    ((sblock.fs_magic == FS_UFS1_MAGIC &&
691 	      cgp->cg_old_niblk == sblock.fs_ipg &&
692 	      cgp->cg_ndblk <= sblock.fs_fpg &&
693 	      cgp->cg_old_ncyl <= sblock.fs_old_cpg) ||
694 	     (sblock.fs_magic == FS_UFS2_MAGIC &&
695 	      cgp->cg_niblk == sblock.fs_ipg &&
696 	      cgp->cg_ndblk <= sblock.fs_fpg &&
697 	      cgp->cg_initediblk <= sblock.fs_ipg))) {
698 		return (1);
699 	}
700 	pfatal("CYLINDER GROUP %d: BAD MAGIC NUMBER", cg);
701 	if (!reply("REBUILD CYLINDER GROUP")) {
702 		printf("YOU WILL NEED TO RERUN FSCK.\n");
703 		rerun = 1;
704 		return (1);
705 	}
706 	/*
707 	 * Zero out the cylinder group and then initialize critical fields.
708 	 * Bit maps and summaries will be recalculated by later passes.
709 	 */
710 	memset(cgp, 0, (size_t)sblock.fs_cgsize);
711 	cgp->cg_magic = CG_MAGIC;
712 	cgp->cg_cgx = cg;
713 	cgp->cg_niblk = sblock.fs_ipg;
714 	cgp->cg_initediblk = MIN(sblock.fs_ipg, 2 * INOPB(&sblock));
715 	if (cgbase(&sblock, cg) + sblock.fs_fpg < sblock.fs_size)
716 		cgp->cg_ndblk = sblock.fs_fpg;
717 	else
718 		cgp->cg_ndblk = sblock.fs_size - cgbase(&sblock, cg);
719 	cgp->cg_iusedoff = &cgp->cg_space[0] - (u_char *)(&cgp->cg_firstfield);
720 	if (sblock.fs_magic == FS_UFS1_MAGIC) {
721 		cgp->cg_niblk = 0;
722 		cgp->cg_initediblk = 0;
723 		cgp->cg_old_ncyl = sblock.fs_old_cpg;
724 		cgp->cg_old_niblk = sblock.fs_ipg;
725 		cgp->cg_old_btotoff = cgp->cg_iusedoff;
726 		cgp->cg_old_boff = cgp->cg_old_btotoff +
727 		    sblock.fs_old_cpg * sizeof(int32_t);
728 		cgp->cg_iusedoff = cgp->cg_old_boff +
729 		    sblock.fs_old_cpg * sizeof(u_int16_t);
730 	}
731 	cgp->cg_freeoff = cgp->cg_iusedoff + howmany(sblock.fs_ipg, CHAR_BIT);
732 	cgp->cg_nextfreeoff = cgp->cg_freeoff + howmany(sblock.fs_fpg,CHAR_BIT);
733 	if (sblock.fs_contigsumsize > 0) {
734 		cgp->cg_nclusterblks = cgp->cg_ndblk / sblock.fs_frag;
735 		cgp->cg_clustersumoff =
736 		    roundup(cgp->cg_nextfreeoff, sizeof(u_int32_t));
737 		cgp->cg_clustersumoff -= sizeof(u_int32_t);
738 		cgp->cg_clusteroff = cgp->cg_clustersumoff +
739 		    (sblock.fs_contigsumsize + 1) * sizeof(u_int32_t);
740 		cgp->cg_nextfreeoff = cgp->cg_clusteroff +
741 		    howmany(fragstoblks(&sblock, sblock.fs_fpg), CHAR_BIT);
742 	}
743 	dirty(cgbp);
744 	return (0);
745 }
746 
747 /*
748  * allocate a data block with the specified number of fragments
749  */
750 ufs2_daddr_t
751 allocblk(long frags)
752 {
753 	int i, j, k, cg, baseblk;
754 	struct bufarea *cgbp;
755 	struct cg *cgp;
756 
757 	if (frags <= 0 || frags > sblock.fs_frag)
758 		return (0);
759 	for (i = 0; i < maxfsblock - sblock.fs_frag; i += sblock.fs_frag) {
760 		for (j = 0; j <= sblock.fs_frag - frags; j++) {
761 			if (testbmap(i + j))
762 				continue;
763 			for (k = 1; k < frags; k++)
764 				if (testbmap(i + j + k))
765 					break;
766 			if (k < frags) {
767 				j += k;
768 				continue;
769 			}
770 			cg = dtog(&sblock, i + j);
771 			cgbp = cglookup(cg);
772 			cgp = cgbp->b_un.b_cg;
773 			if (!check_cgmagic(cg, cgbp))
774 				return (0);
775 			baseblk = dtogd(&sblock, i + j);
776 			for (k = 0; k < frags; k++) {
777 				setbmap(i + j + k);
778 				clrbit(cg_blksfree(cgp), baseblk + k);
779 			}
780 			n_blks += frags;
781 			if (frags == sblock.fs_frag)
782 				cgp->cg_cs.cs_nbfree--;
783 			else
784 				cgp->cg_cs.cs_nffree -= frags;
785 			dirty(cgbp);
786 			return (i + j);
787 		}
788 	}
789 	return (0);
790 }
791 
792 /*
793  * Free a previously allocated block
794  */
795 void
796 freeblk(ufs2_daddr_t blkno, long frags)
797 {
798 	struct inodesc idesc;
799 
800 	idesc.id_blkno = blkno;
801 	idesc.id_numfrags = frags;
802 	(void)pass4check(&idesc);
803 }
804 
805 /* Slow down IO so as to leave some disk bandwidth for other processes */
806 void
807 slowio_start()
808 {
809 
810 	/* Delay one in every 8 operations */
811 	slowio_pollcnt = (slowio_pollcnt + 1) & 7;
812 	if (slowio_pollcnt == 0) {
813 		gettimeofday(&slowio_starttime, NULL);
814 	}
815 }
816 
817 void
818 slowio_end()
819 {
820 	struct timeval tv;
821 	int delay_usec;
822 
823 	if (slowio_pollcnt != 0)
824 		return;
825 
826 	/* Update the slowdown interval. */
827 	gettimeofday(&tv, NULL);
828 	delay_usec = (tv.tv_sec - slowio_starttime.tv_sec) * 1000000 +
829 	    (tv.tv_usec - slowio_starttime.tv_usec);
830 	if (delay_usec < 64)
831 		delay_usec = 64;
832 	if (delay_usec > 2500000)
833 		delay_usec = 2500000;
834 	slowio_delay_usec = (slowio_delay_usec * 63 + delay_usec) >> 6;
835 	/* delay by 8 times the average IO delay */
836 	if (slowio_delay_usec > 64)
837 		usleep(slowio_delay_usec * 8);
838 }
839 
840 /*
841  * Find a pathname
842  */
843 void
844 getpathname(char *namebuf, ino_t curdir, ino_t ino)
845 {
846 	int len;
847 	char *cp;
848 	struct inodesc idesc;
849 	static int busy = 0;
850 
851 	if (curdir == ino && ino == UFS_ROOTINO) {
852 		(void)strcpy(namebuf, "/");
853 		return;
854 	}
855 	if (busy || !INO_IS_DVALID(curdir)) {
856 		(void)strcpy(namebuf, "?");
857 		return;
858 	}
859 	busy = 1;
860 	memset(&idesc, 0, sizeof(struct inodesc));
861 	idesc.id_type = DATA;
862 	idesc.id_fix = IGNORE;
863 	cp = &namebuf[MAXPATHLEN - 1];
864 	*cp = '\0';
865 	if (curdir != ino) {
866 		idesc.id_parent = curdir;
867 		goto namelookup;
868 	}
869 	while (ino != UFS_ROOTINO) {
870 		idesc.id_number = ino;
871 		idesc.id_func = findino;
872 		idesc.id_name = strdup("..");
873 		if ((ckinode(ginode(ino), &idesc) & FOUND) == 0)
874 			break;
875 	namelookup:
876 		idesc.id_number = idesc.id_parent;
877 		idesc.id_parent = ino;
878 		idesc.id_func = findname;
879 		idesc.id_name = namebuf;
880 		if ((ckinode(ginode(idesc.id_number), &idesc)&FOUND) == 0)
881 			break;
882 		len = strlen(namebuf);
883 		cp -= len;
884 		memmove(cp, namebuf, (size_t)len);
885 		*--cp = '/';
886 		if (cp < &namebuf[UFS_MAXNAMLEN])
887 			break;
888 		ino = idesc.id_number;
889 	}
890 	busy = 0;
891 	if (ino != UFS_ROOTINO)
892 		*--cp = '?';
893 	memmove(namebuf, cp, (size_t)(&namebuf[MAXPATHLEN] - cp));
894 }
895 
896 void
897 catch(int sig __unused)
898 {
899 
900 	ckfini(0);
901 	exit(12);
902 }
903 
904 /*
905  * When preening, allow a single quit to signal
906  * a special exit after file system checks complete
907  * so that reboot sequence may be interrupted.
908  */
909 void
910 catchquit(int sig __unused)
911 {
912 	printf("returning to single-user after file system check\n");
913 	returntosingle = 1;
914 	(void)signal(SIGQUIT, SIG_DFL);
915 }
916 
917 /*
918  * determine whether an inode should be fixed.
919  */
920 int
921 dofix(struct inodesc *idesc, const char *msg)
922 {
923 
924 	switch (idesc->id_fix) {
925 
926 	case DONTKNOW:
927 		if (idesc->id_type == DATA)
928 			direrror(idesc->id_number, msg);
929 		else
930 			pwarn("%s", msg);
931 		if (preen) {
932 			printf(" (SALVAGED)\n");
933 			idesc->id_fix = FIX;
934 			return (ALTERED);
935 		}
936 		if (reply("SALVAGE") == 0) {
937 			idesc->id_fix = NOFIX;
938 			return (0);
939 		}
940 		idesc->id_fix = FIX;
941 		return (ALTERED);
942 
943 	case FIX:
944 		return (ALTERED);
945 
946 	case NOFIX:
947 	case IGNORE:
948 		return (0);
949 
950 	default:
951 		errx(EEXIT, "UNKNOWN INODESC FIX MODE %d", idesc->id_fix);
952 	}
953 	/* NOTREACHED */
954 	return (0);
955 }
956 
957 #include <stdarg.h>
958 
959 /*
960  * An unexpected inconsistency occurred.
961  * Die if preening or file system is running with soft dependency protocol,
962  * otherwise just print message and continue.
963  */
964 void
965 pfatal(const char *fmt, ...)
966 {
967 	va_list ap;
968 	va_start(ap, fmt);
969 	if (!preen) {
970 		(void)vfprintf(stdout, fmt, ap);
971 		va_end(ap);
972 		if (usedsoftdep)
973 			(void)fprintf(stdout,
974 			    "\nUNEXPECTED SOFT UPDATE INCONSISTENCY\n");
975 		/*
976 		 * Force foreground fsck to clean up inconsistency.
977 		 */
978 		if (bkgrdflag) {
979 			cmd.value = FS_NEEDSFSCK;
980 			cmd.size = 1;
981 			if (sysctlbyname("vfs.ffs.setflags", 0, 0,
982 			    &cmd, sizeof cmd) == -1)
983 				pwarn("CANNOT SET FS_NEEDSFSCK FLAG\n");
984 			fprintf(stdout, "CANNOT RUN IN BACKGROUND\n");
985 			ckfini(0);
986 			exit(EEXIT);
987 		}
988 		return;
989 	}
990 	if (cdevname == NULL)
991 		cdevname = strdup("fsck");
992 	(void)fprintf(stdout, "%s: ", cdevname);
993 	(void)vfprintf(stdout, fmt, ap);
994 	(void)fprintf(stdout,
995 	    "\n%s: UNEXPECTED%sINCONSISTENCY; RUN fsck MANUALLY.\n",
996 	    cdevname, usedsoftdep ? " SOFT UPDATE " : " ");
997 	/*
998 	 * Force foreground fsck to clean up inconsistency.
999 	 */
1000 	if (bkgrdflag) {
1001 		cmd.value = FS_NEEDSFSCK;
1002 		cmd.size = 1;
1003 		if (sysctlbyname("vfs.ffs.setflags", 0, 0,
1004 		    &cmd, sizeof cmd) == -1)
1005 			pwarn("CANNOT SET FS_NEEDSFSCK FLAG\n");
1006 	}
1007 	ckfini(0);
1008 	exit(EEXIT);
1009 }
1010 
1011 /*
1012  * Pwarn just prints a message when not preening or running soft dependency
1013  * protocol, or a warning (preceded by filename) when preening.
1014  */
1015 void
1016 pwarn(const char *fmt, ...)
1017 {
1018 	va_list ap;
1019 	va_start(ap, fmt);
1020 	if (preen)
1021 		(void)fprintf(stdout, "%s: ", cdevname);
1022 	(void)vfprintf(stdout, fmt, ap);
1023 	va_end(ap);
1024 }
1025 
1026 /*
1027  * Stub for routines from kernel.
1028  */
1029 void
1030 panic(const char *fmt, ...)
1031 {
1032 	va_list ap;
1033 	va_start(ap, fmt);
1034 	pfatal("INTERNAL INCONSISTENCY:");
1035 	(void)vfprintf(stdout, fmt, ap);
1036 	va_end(ap);
1037 	exit(EEXIT);
1038 }
1039