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
32 static const char sccsid[] = "@(#)utilities.c 8.6 (Berkeley) 5/19/95";
33 #endif /* not lint */
34 #endif
35 #include <sys/cdefs.h>
36 __FBSDID("$FreeBSD$");
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
64 static void slowio_start(void);
65 static void slowio_end(void);
66 static void printIOstats(void);
67
68 static long diskreads, totaldiskreads, totalreads; /* Disk cache statistics */
69 static struct timespec startpass, finishpass;
70 struct timeval slowio_starttime;
71 int slowio_delay_usec = 10000; /* Initial IO delay for background fsck */
72 int slowio_pollcnt;
73 static struct bufarea cgblk; /* backup buffer for cylinder group blocks */
74 static TAILQ_HEAD(buflist, bufarea) bufhead; /* head of buffer cache list */
75 static int numbufs; /* size of buffer cache */
76 static char *buftype[BT_NUMBUFTYPES] = BT_NAMES;
77 static struct bufarea *cgbufs; /* header for cylinder group cache */
78 static int flushtries; /* number of tries to reclaim memory */
79
80 void
fsutilinit(void)81 fsutilinit(void)
82 {
83 diskreads = totaldiskreads = totalreads = 0;
84 bzero(&startpass, sizeof(struct timespec));
85 bzero(&finishpass, sizeof(struct timespec));
86 bzero(&slowio_starttime, sizeof(struct timeval));
87 slowio_delay_usec = 10000;
88 slowio_pollcnt = 0;
89 bzero(&cgblk, sizeof(struct bufarea));
90 TAILQ_INIT(&bufhead);
91 numbufs = 0;
92 /* buftype ? */
93 cgbufs = NULL;
94 flushtries = 0;
95 }
96
97 int
ftypeok(union dinode * dp)98 ftypeok(union dinode *dp)
99 {
100 switch (DIP(dp, di_mode) & IFMT) {
101
102 case IFDIR:
103 case IFREG:
104 case IFBLK:
105 case IFCHR:
106 case IFLNK:
107 case IFSOCK:
108 case IFIFO:
109 return (1);
110
111 default:
112 if (debug)
113 printf("bad file type 0%o\n", DIP(dp, di_mode));
114 return (0);
115 }
116 }
117
118 int
reply(const char * question)119 reply(const char *question)
120 {
121 int persevere;
122 char c;
123
124 if (preen)
125 pfatal("INTERNAL ERROR: GOT TO reply()");
126 persevere = !strcmp(question, "CONTINUE");
127 printf("\n");
128 if (!persevere && (nflag || (fswritefd < 0 && bkgrdflag == 0))) {
129 printf("%s? no\n\n", question);
130 resolved = 0;
131 return (0);
132 }
133 if (yflag || (persevere && nflag)) {
134 printf("%s? yes\n\n", question);
135 return (1);
136 }
137 do {
138 printf("%s? [yn] ", question);
139 (void) fflush(stdout);
140 c = getc(stdin);
141 while (c != '\n' && getc(stdin) != '\n') {
142 if (feof(stdin)) {
143 resolved = 0;
144 return (0);
145 }
146 }
147 } while (c != 'y' && c != 'Y' && c != 'n' && c != 'N');
148 printf("\n");
149 if (c == 'y' || c == 'Y')
150 return (1);
151 resolved = 0;
152 return (0);
153 }
154
155 /*
156 * Look up state information for an inode.
157 */
158 struct inostat *
inoinfo(ino_t inum)159 inoinfo(ino_t inum)
160 {
161 static struct inostat unallocated = { USTATE, 0, 0 };
162 struct inostatlist *ilp;
163 int iloff;
164
165 if (inum > maxino)
166 errx(EEXIT, "inoinfo: inumber %ju out of range",
167 (uintmax_t)inum);
168 ilp = &inostathead[inum / sblock.fs_ipg];
169 iloff = inum % sblock.fs_ipg;
170 if (iloff >= ilp->il_numalloced)
171 return (&unallocated);
172 return (&ilp->il_stat[iloff]);
173 }
174
175 /*
176 * Malloc buffers and set up cache.
177 */
178 void
bufinit(void)179 bufinit(void)
180 {
181 struct bufarea *bp;
182 long bufcnt, i;
183 char *bufp;
184
185 pbp = pdirbp = (struct bufarea *)0;
186 bufp = Malloc((unsigned int)sblock.fs_bsize);
187 if (bufp == 0)
188 errx(EEXIT, "cannot allocate buffer pool");
189 cgblk.b_un.b_buf = bufp;
190 initbarea(&cgblk, BT_CYLGRP);
191 TAILQ_INIT(&bufhead);
192 bufcnt = MAXBUFS;
193 if (bufcnt < MINBUFS)
194 bufcnt = MINBUFS;
195 for (i = 0; i < bufcnt; i++) {
196 bp = (struct bufarea *)Malloc(sizeof(struct bufarea));
197 bufp = Malloc((unsigned int)sblock.fs_bsize);
198 if (bp == NULL || bufp == NULL) {
199 if (i >= MINBUFS)
200 break;
201 errx(EEXIT, "cannot allocate buffer pool");
202 }
203 bp->b_un.b_buf = bufp;
204 TAILQ_INSERT_HEAD(&bufhead, bp, b_list);
205 initbarea(bp, BT_UNKNOWN);
206 }
207 numbufs = i; /* save number of buffers */
208 for (i = 0; i < BT_NUMBUFTYPES; i++) {
209 readtime[i].tv_sec = totalreadtime[i].tv_sec = 0;
210 readtime[i].tv_nsec = totalreadtime[i].tv_nsec = 0;
211 readcnt[i] = totalreadcnt[i] = 0;
212 }
213 }
214
215 /*
216 * Manage cylinder group buffers.
217 */
218 static struct bufarea *cgbufs; /* header for cylinder group cache */
219 static int flushtries; /* number of tries to reclaim memory */
220
221 struct bufarea *
cgget(int cg)222 cgget(int cg)
223 {
224 struct bufarea *cgbp;
225 struct cg *cgp;
226
227 if (cgbufs == NULL) {
228 cgbufs = calloc(sblock.fs_ncg, sizeof(struct bufarea));
229 if (cgbufs == NULL)
230 errx(EEXIT, "cannot allocate cylinder group buffers");
231 }
232 cgbp = &cgbufs[cg];
233 if (cgbp->b_un.b_cg != NULL)
234 return (cgbp);
235 cgp = NULL;
236 if (flushtries == 0)
237 cgp = malloc((unsigned int)sblock.fs_cgsize);
238 if (cgp == NULL) {
239 getblk(&cgblk, cgtod(&sblock, cg), sblock.fs_cgsize);
240 return (&cgblk);
241 }
242 cgbp->b_un.b_cg = cgp;
243 initbarea(cgbp, BT_CYLGRP);
244 getblk(cgbp, cgtod(&sblock, cg), sblock.fs_cgsize);
245 return (cgbp);
246 }
247
248 /*
249 * Attempt to flush a cylinder group cache entry.
250 * Return whether the flush was successful.
251 */
252 int
flushentry(void)253 flushentry(void)
254 {
255 struct bufarea *cgbp;
256
257 if (flushtries == sblock.fs_ncg || cgbufs == NULL)
258 return (0);
259 cgbp = &cgbufs[flushtries++];
260 if (cgbp->b_un.b_cg == NULL)
261 return (0);
262 flush(fswritefd, cgbp);
263 free(cgbp->b_un.b_buf);
264 cgbp->b_un.b_buf = NULL;
265 return (1);
266 }
267
268 /*
269 * Manage a cache of directory blocks.
270 */
271 struct bufarea *
getdatablk(ufs2_daddr_t blkno,long size,int type)272 getdatablk(ufs2_daddr_t blkno, long size, int type)
273 {
274 struct bufarea *bp;
275
276 TAILQ_FOREACH(bp, &bufhead, b_list)
277 if (bp->b_bno == fsbtodb(&sblock, blkno))
278 goto foundit;
279 TAILQ_FOREACH_REVERSE(bp, &bufhead, buflist, b_list)
280 if ((bp->b_flags & B_INUSE) == 0)
281 break;
282 if (bp == NULL)
283 errx(EEXIT, "deadlocked buffer pool");
284 bp->b_type = type;
285 getblk(bp, blkno, size);
286 /* fall through */
287 foundit:
288 if (debug && bp->b_type != type)
289 printf("Buffer type changed from %s to %s\n",
290 buftype[bp->b_type], buftype[type]);
291 TAILQ_REMOVE(&bufhead, bp, b_list);
292 TAILQ_INSERT_HEAD(&bufhead, bp, b_list);
293 bp->b_flags |= B_INUSE;
294 return (bp);
295 }
296
297 /*
298 * Timespec operations (from <sys/time.h>).
299 */
300 #define timespecsub(vvp, uvp) \
301 do { \
302 (vvp)->tv_sec -= (uvp)->tv_sec; \
303 (vvp)->tv_nsec -= (uvp)->tv_nsec; \
304 if ((vvp)->tv_nsec < 0) { \
305 (vvp)->tv_sec--; \
306 (vvp)->tv_nsec += 1000000000; \
307 } \
308 } while (0)
309 #define timespecadd(vvp, uvp) \
310 do { \
311 (vvp)->tv_sec += (uvp)->tv_sec; \
312 (vvp)->tv_nsec += (uvp)->tv_nsec; \
313 if ((vvp)->tv_nsec >= 1000000000) { \
314 (vvp)->tv_sec++; \
315 (vvp)->tv_nsec -= 1000000000; \
316 } \
317 } while (0)
318
319 void
getblk(struct bufarea * bp,ufs2_daddr_t blk,long size)320 getblk(struct bufarea *bp, ufs2_daddr_t blk, long size)
321 {
322 ufs2_daddr_t dblk;
323 struct timespec start, finish;
324
325 dblk = fsbtodb(&sblock, blk);
326 if (bp->b_bno == dblk) {
327 totalreads++;
328 } else {
329 flush(fswritefd, bp);
330 if (debug) {
331 readcnt[bp->b_type]++;
332 clock_gettime(CLOCK_REALTIME_PRECISE, &start);
333 }
334 bp->b_errs = blread(fsreadfd, bp->b_un.b_buf, dblk, size);
335 if (debug) {
336 clock_gettime(CLOCK_REALTIME_PRECISE, &finish);
337 timespecsub(&finish, &start);
338 timespecadd(&readtime[bp->b_type], &finish);
339 }
340 bp->b_bno = dblk;
341 bp->b_size = size;
342 }
343 }
344
345 void
flush(int fd,struct bufarea * bp)346 flush(int fd, struct bufarea *bp)
347 {
348 int i, j;
349
350 if (!bp->b_dirty)
351 return;
352 bp->b_dirty = 0;
353 if (fswritefd < 0) {
354 pfatal("WRITING IN READ_ONLY MODE.\n");
355 return;
356 }
357 if (bp->b_errs != 0)
358 pfatal("WRITING %sZERO'ED BLOCK %lld TO DISK\n",
359 (bp->b_errs == bp->b_size / dev_bsize) ? "" : "PARTIALLY ",
360 (long long)bp->b_bno);
361 bp->b_errs = 0;
362 blwrite(fd, bp->b_un.b_buf, bp->b_bno, bp->b_size);
363 if (bp != &sblk)
364 return;
365 for (i = 0, j = 0; i < sblock.fs_cssize; i += sblock.fs_bsize, j++) {
366 blwrite(fswritefd, (char *)sblock.fs_csp + i,
367 fsbtodb(&sblock, sblock.fs_csaddr + j * sblock.fs_frag),
368 sblock.fs_cssize - i < sblock.fs_bsize ?
369 sblock.fs_cssize - i : sblock.fs_bsize);
370 }
371 }
372
373 void
rwerror(const char * mesg,ufs2_daddr_t blk)374 rwerror(const char *mesg, ufs2_daddr_t blk)
375 {
376
377 if (bkgrdcheck)
378 exit(EEXIT);
379 if (preen == 0)
380 printf("\n");
381 pfatal("CANNOT %s: %ld", mesg, (long)blk);
382 if (reply("CONTINUE") == 0)
383 exit(EEXIT);
384 }
385
386 void
ckfini(int markclean)387 ckfini(int markclean)
388 {
389 struct bufarea *bp, *nbp;
390 int ofsmodified, cnt;
391
392 if (bkgrdflag) {
393 unlink(snapname);
394 if ((!(sblock.fs_flags & FS_UNCLEAN)) != markclean) {
395 cmd.value = FS_UNCLEAN;
396 cmd.size = markclean ? -1 : 1;
397 if (sysctlbyname("vfs.ffs.setflags", 0, 0,
398 &cmd, sizeof cmd) == -1)
399 rwerror("SET FILE SYSTEM FLAGS", FS_UNCLEAN);
400 if (!preen) {
401 printf("\n***** FILE SYSTEM MARKED %s *****\n",
402 markclean ? "CLEAN" : "DIRTY");
403 if (!markclean)
404 rerun = 1;
405 }
406 } else if (!preen && !markclean) {
407 printf("\n***** FILE SYSTEM STILL DIRTY *****\n");
408 rerun = 1;
409 }
410 }
411 if (debug && totalreads > 0)
412 printf("cache with %d buffers missed %ld of %ld (%d%%)\n",
413 numbufs, totaldiskreads, totalreads,
414 (int)(totaldiskreads * 100 / totalreads));
415 if (fswritefd < 0) {
416 (void)close(fsreadfd);
417 return;
418 }
419 flush(fswritefd, &sblk);
420 if (havesb && cursnapshot == 0 && sblock.fs_magic == FS_UFS2_MAGIC &&
421 sblk.b_bno != sblock.fs_sblockloc / dev_bsize &&
422 !preen && reply("UPDATE STANDARD SUPERBLOCK")) {
423 sblk.b_bno = sblock.fs_sblockloc / dev_bsize;
424 sbdirty();
425 flush(fswritefd, &sblk);
426 }
427 flush(fswritefd, &cgblk);
428 free(cgblk.b_un.b_buf);
429 cnt = 0;
430 TAILQ_FOREACH_REVERSE_SAFE(bp, &bufhead, buflist, b_list, nbp) {
431 TAILQ_REMOVE(&bufhead, bp, b_list);
432 cnt++;
433 flush(fswritefd, bp);
434 free(bp->b_un.b_buf);
435 free((char *)bp);
436 }
437 if (numbufs != cnt)
438 errx(EEXIT, "panic: lost %d buffers", numbufs - cnt);
439 if (cgbufs != NULL) {
440 for (cnt = 0; cnt < sblock.fs_ncg; cnt++) {
441 if (cgbufs[cnt].b_un.b_cg == NULL)
442 continue;
443 flush(fswritefd, &cgbufs[cnt]);
444 free(cgbufs[cnt].b_un.b_cg);
445 }
446 free(cgbufs);
447 }
448 pbp = pdirbp = (struct bufarea *)0;
449 if (cursnapshot == 0 && sblock.fs_clean != markclean) {
450 if ((sblock.fs_clean = markclean) != 0) {
451 sblock.fs_flags &= ~(FS_UNCLEAN | FS_NEEDSFSCK);
452 sblock.fs_pendingblocks = 0;
453 sblock.fs_pendinginodes = 0;
454 }
455 sbdirty();
456 ofsmodified = fsmodified;
457 flush(fswritefd, &sblk);
458 fsmodified = ofsmodified;
459 if (!preen) {
460 printf("\n***** FILE SYSTEM MARKED %s *****\n",
461 markclean ? "CLEAN" : "DIRTY");
462 if (!markclean)
463 rerun = 1;
464 }
465 } else if (!preen) {
466 if (markclean) {
467 printf("\n***** FILE SYSTEM IS CLEAN *****\n");
468 } else {
469 printf("\n***** FILE SYSTEM STILL DIRTY *****\n");
470 rerun = 1;
471 }
472 }
473 (void)close(fsreadfd);
474 (void)close(fswritefd);
475 }
476
477 /*
478 * Print out I/O statistics.
479 */
480 void
IOstats(char * what)481 IOstats(char *what)
482 {
483 int i;
484
485 if (debug == 0)
486 return;
487 if (diskreads == 0) {
488 printf("%s: no I/O\n\n", what);
489 return;
490 }
491 if (startpass.tv_sec == 0)
492 startpass = startprog;
493 printf("%s: I/O statistics\n", what);
494 printIOstats();
495 totaldiskreads += diskreads;
496 diskreads = 0;
497 for (i = 0; i < BT_NUMBUFTYPES; i++) {
498 timespecadd(&totalreadtime[i], &readtime[i]);
499 totalreadcnt[i] += readcnt[i];
500 readtime[i].tv_sec = readtime[i].tv_nsec = 0;
501 readcnt[i] = 0;
502 }
503 clock_gettime(CLOCK_REALTIME_PRECISE, &startpass);
504 }
505
506 void
finalIOstats(void)507 finalIOstats(void)
508 {
509 int i;
510
511 if (debug == 0)
512 return;
513 printf("Final I/O statistics\n");
514 totaldiskreads += diskreads;
515 diskreads = totaldiskreads;
516 startpass = startprog;
517 for (i = 0; i < BT_NUMBUFTYPES; i++) {
518 timespecadd(&totalreadtime[i], &readtime[i]);
519 totalreadcnt[i] += readcnt[i];
520 readtime[i] = totalreadtime[i];
521 readcnt[i] = totalreadcnt[i];
522 }
523 printIOstats();
524 }
525
printIOstats(void)526 static void printIOstats(void)
527 {
528 long long msec, totalmsec;
529 int i;
530
531 clock_gettime(CLOCK_REALTIME_PRECISE, &finishpass);
532 timespecsub(&finishpass, &startpass);
533 printf("Running time: %jd.%03ld sec\n",
534 (intmax_t)finishpass.tv_sec, finishpass.tv_nsec / 1000000);
535 printf("buffer reads by type:\n");
536 for (totalmsec = 0, i = 0; i < BT_NUMBUFTYPES; i++)
537 totalmsec += readtime[i].tv_sec * 1000 +
538 readtime[i].tv_nsec / 1000000;
539 if (totalmsec == 0)
540 totalmsec = 1;
541 for (i = 0; i < BT_NUMBUFTYPES; i++) {
542 if (readcnt[i] == 0)
543 continue;
544 msec =
545 readtime[i].tv_sec * 1000 + readtime[i].tv_nsec / 1000000;
546 printf("%21s:%8ld %2ld.%ld%% %4jd.%03ld sec %2lld.%lld%%\n",
547 buftype[i], readcnt[i], readcnt[i] * 100 / diskreads,
548 (readcnt[i] * 1000 / diskreads) % 10,
549 (intmax_t)readtime[i].tv_sec, readtime[i].tv_nsec / 1000000,
550 msec * 100 / totalmsec, (msec * 1000 / totalmsec) % 10);
551 }
552 printf("\n");
553 }
554
555 int
blread(int fd,char * buf,ufs2_daddr_t blk,long size)556 blread(int fd, char *buf, ufs2_daddr_t blk, long size)
557 {
558 char *cp;
559 int i, errs;
560 off_t offset;
561
562 offset = blk;
563 offset *= dev_bsize;
564 if (bkgrdflag)
565 slowio_start();
566 totalreads++;
567 diskreads++;
568 if (lseek(fd, offset, 0) < 0)
569 rwerror("SEEK BLK", blk);
570 else if (read(fd, buf, (int)size) == size) {
571 if (bkgrdflag)
572 slowio_end();
573 return (0);
574 }
575
576 /*
577 * This is handled specially here instead of in rwerror because
578 * rwerror is used for all sorts of errors, not just true read/write
579 * errors. It should be refactored and fixed.
580 */
581 if (surrender) {
582 pfatal("CANNOT READ_BLK: %ld", (long)blk);
583 errx(EEXIT, "ABORTING DUE TO READ ERRORS");
584 } else
585 rwerror("READ BLK", blk);
586
587 if (lseek(fd, offset, 0) < 0)
588 rwerror("SEEK BLK", blk);
589 errs = 0;
590 memset(buf, 0, (size_t)size);
591 printf("THE FOLLOWING DISK SECTORS COULD NOT BE READ:");
592 for (cp = buf, i = 0; i < size; i += secsize, cp += secsize) {
593 if (read(fd, cp, (int)secsize) != secsize) {
594 (void)lseek(fd, offset + i + secsize, 0);
595 if (secsize != dev_bsize && dev_bsize != 1)
596 printf(" %jd (%jd),",
597 (intmax_t)(blk * dev_bsize + i) / secsize,
598 (intmax_t)blk + i / dev_bsize);
599 else
600 printf(" %jd,", (intmax_t)blk + i / dev_bsize);
601 errs++;
602 }
603 }
604 printf("\n");
605 if (errs)
606 resolved = 0;
607 return (errs);
608 }
609
610 void
blwrite(int fd,char * buf,ufs2_daddr_t blk,ssize_t size)611 blwrite(int fd, char *buf, ufs2_daddr_t blk, ssize_t size)
612 {
613 int i;
614 char *cp;
615 off_t offset;
616
617 if (fd < 0)
618 return;
619 offset = blk;
620 offset *= dev_bsize;
621 if (lseek(fd, offset, 0) < 0)
622 rwerror("SEEK BLK", blk);
623 else if (write(fd, buf, size) == size) {
624 fsmodified = 1;
625 return;
626 }
627 resolved = 0;
628 rwerror("WRITE BLK", blk);
629 if (lseek(fd, offset, 0) < 0)
630 rwerror("SEEK BLK", blk);
631 printf("THE FOLLOWING SECTORS COULD NOT BE WRITTEN:");
632 for (cp = buf, i = 0; i < size; i += dev_bsize, cp += dev_bsize)
633 if (write(fd, cp, dev_bsize) != dev_bsize) {
634 (void)lseek(fd, offset + i + dev_bsize, 0);
635 printf(" %jd,", (intmax_t)blk + i / dev_bsize);
636 }
637 printf("\n");
638 return;
639 }
640
641 void
blerase(int fd,ufs2_daddr_t blk,long size)642 blerase(int fd, ufs2_daddr_t blk, long size)
643 {
644 off_t ioarg[2];
645
646 if (fd < 0)
647 return;
648 ioarg[0] = blk * dev_bsize;
649 ioarg[1] = size;
650 ioctl(fd, DIOCGDELETE, ioarg);
651 /* we don't really care if we succeed or not */
652 return;
653 }
654
655 /*
656 * Fill a contiguous region with all-zeroes. Note ZEROBUFSIZE is by
657 * definition a multiple of dev_bsize.
658 */
659 void
blzero(int fd,ufs2_daddr_t blk,long size)660 blzero(int fd, ufs2_daddr_t blk, long size)
661 {
662 static char *zero;
663 off_t offset, len;
664
665 if (fd < 0)
666 return;
667 if (zero == NULL) {
668 zero = calloc(ZEROBUFSIZE, 1);
669 if (zero == NULL)
670 errx(EEXIT, "cannot allocate buffer pool");
671 }
672 offset = blk * dev_bsize;
673 if (lseek(fd, offset, 0) < 0)
674 rwerror("SEEK BLK", blk);
675 while (size > 0) {
676 len = size > ZEROBUFSIZE ? ZEROBUFSIZE : size;
677 if (write(fd, zero, len) != len)
678 rwerror("WRITE BLK", blk);
679 blk += len / dev_bsize;
680 size -= len;
681 }
682 }
683
684 /*
685 * Verify cylinder group's magic number and other parameters. If the
686 * test fails, offer an option to rebuild the whole cylinder group.
687 */
688 int
check_cgmagic(int cg,struct bufarea * cgbp)689 check_cgmagic(int cg, struct bufarea *cgbp)
690 {
691 struct cg *cgp = cgbp->b_un.b_cg;
692
693 /*
694 * Extended cylinder group checks.
695 */
696 if (cg_chkmagic(cgp) &&
697 ((sblock.fs_magic == FS_UFS1_MAGIC &&
698 cgp->cg_old_niblk == sblock.fs_ipg &&
699 cgp->cg_ndblk <= sblock.fs_fpg &&
700 cgp->cg_old_ncyl <= sblock.fs_old_cpg) ||
701 (sblock.fs_magic == FS_UFS2_MAGIC &&
702 cgp->cg_niblk == sblock.fs_ipg &&
703 cgp->cg_ndblk <= sblock.fs_fpg &&
704 cgp->cg_initediblk <= sblock.fs_ipg))) {
705 return (1);
706 }
707 pfatal("CYLINDER GROUP %d: BAD MAGIC NUMBER", cg);
708 if (!reply("REBUILD CYLINDER GROUP")) {
709 printf("YOU WILL NEED TO RERUN FSCK.\n");
710 rerun = 1;
711 return (1);
712 }
713 /*
714 * Zero out the cylinder group and then initialize critical fields.
715 * Bit maps and summaries will be recalculated by later passes.
716 */
717 memset(cgp, 0, (size_t)sblock.fs_cgsize);
718 cgp->cg_magic = CG_MAGIC;
719 cgp->cg_cgx = cg;
720 cgp->cg_niblk = sblock.fs_ipg;
721 cgp->cg_initediblk = sblock.fs_ipg < 2 * INOPB(&sblock) ?
722 sblock.fs_ipg : 2 * INOPB(&sblock);
723 if (cgbase(&sblock, cg) + sblock.fs_fpg < sblock.fs_size)
724 cgp->cg_ndblk = sblock.fs_fpg;
725 else
726 cgp->cg_ndblk = sblock.fs_size - cgbase(&sblock, cg);
727 cgp->cg_iusedoff = &cgp->cg_space[0] - (u_char *)(&cgp->cg_firstfield);
728 if (sblock.fs_magic == FS_UFS1_MAGIC) {
729 cgp->cg_niblk = 0;
730 cgp->cg_initediblk = 0;
731 cgp->cg_old_ncyl = sblock.fs_old_cpg;
732 cgp->cg_old_niblk = sblock.fs_ipg;
733 cgp->cg_old_btotoff = cgp->cg_iusedoff;
734 cgp->cg_old_boff = cgp->cg_old_btotoff +
735 sblock.fs_old_cpg * sizeof(int32_t);
736 cgp->cg_iusedoff = cgp->cg_old_boff +
737 sblock.fs_old_cpg * sizeof(u_int16_t);
738 }
739 cgp->cg_freeoff = cgp->cg_iusedoff + howmany(sblock.fs_ipg, CHAR_BIT);
740 cgp->cg_nextfreeoff = cgp->cg_freeoff + howmany(sblock.fs_fpg,CHAR_BIT);
741 if (sblock.fs_contigsumsize > 0) {
742 cgp->cg_nclusterblks = cgp->cg_ndblk / sblock.fs_frag;
743 cgp->cg_clustersumoff =
744 roundup(cgp->cg_nextfreeoff, sizeof(u_int32_t));
745 cgp->cg_clustersumoff -= sizeof(u_int32_t);
746 cgp->cg_clusteroff = cgp->cg_clustersumoff +
747 (sblock.fs_contigsumsize + 1) * sizeof(u_int32_t);
748 cgp->cg_nextfreeoff = cgp->cg_clusteroff +
749 howmany(fragstoblks(&sblock, sblock.fs_fpg), CHAR_BIT);
750 }
751 dirty(cgbp);
752 return (0);
753 }
754
755 /*
756 * allocate a data block with the specified number of fragments
757 */
758 ufs2_daddr_t
allocblk(long frags)759 allocblk(long frags)
760 {
761 int i, j, k, cg, baseblk;
762 struct bufarea *cgbp;
763 struct cg *cgp;
764
765 if (frags <= 0 || frags > sblock.fs_frag)
766 return (0);
767 for (i = 0; i < maxfsblock - sblock.fs_frag; i += sblock.fs_frag) {
768 for (j = 0; j <= sblock.fs_frag - frags; j++) {
769 if (testbmap(i + j))
770 continue;
771 for (k = 1; k < frags; k++)
772 if (testbmap(i + j + k))
773 break;
774 if (k < frags) {
775 j += k;
776 continue;
777 }
778 cg = dtog(&sblock, i + j);
779 cgbp = cgget(cg);
780 cgp = cgbp->b_un.b_cg;
781 if (!check_cgmagic(cg, cgbp))
782 return (0);
783 baseblk = dtogd(&sblock, i + j);
784 for (k = 0; k < frags; k++) {
785 setbmap(i + j + k);
786 clrbit(cg_blksfree(cgp), baseblk + k);
787 }
788 n_blks += frags;
789 if (frags == sblock.fs_frag)
790 cgp->cg_cs.cs_nbfree--;
791 else
792 cgp->cg_cs.cs_nffree -= frags;
793 dirty(cgbp);
794 return (i + j);
795 }
796 }
797 return (0);
798 }
799
800 /*
801 * Free a previously allocated block
802 */
803 void
freeblk(ufs2_daddr_t blkno,long frags)804 freeblk(ufs2_daddr_t blkno, long frags)
805 {
806 struct inodesc idesc;
807
808 idesc.id_blkno = blkno;
809 idesc.id_numfrags = frags;
810 (void)pass4check(&idesc);
811 }
812
813 /* Slow down IO so as to leave some disk bandwidth for other processes */
814 void
slowio_start()815 slowio_start()
816 {
817
818 /* Delay one in every 8 operations */
819 slowio_pollcnt = (slowio_pollcnt + 1) & 7;
820 if (slowio_pollcnt == 0) {
821 gettimeofday(&slowio_starttime, NULL);
822 }
823 }
824
825 void
slowio_end()826 slowio_end()
827 {
828 struct timeval tv;
829 int delay_usec;
830
831 if (slowio_pollcnt != 0)
832 return;
833
834 /* Update the slowdown interval. */
835 gettimeofday(&tv, NULL);
836 delay_usec = (tv.tv_sec - slowio_starttime.tv_sec) * 1000000 +
837 (tv.tv_usec - slowio_starttime.tv_usec);
838 if (delay_usec < 64)
839 delay_usec = 64;
840 if (delay_usec > 2500000)
841 delay_usec = 2500000;
842 slowio_delay_usec = (slowio_delay_usec * 63 + delay_usec) >> 6;
843 /* delay by 8 times the average IO delay */
844 if (slowio_delay_usec > 64)
845 usleep(slowio_delay_usec * 8);
846 }
847
848 /*
849 * Find a pathname
850 */
851 void
getpathname(char * namebuf,ino_t curdir,ino_t ino)852 getpathname(char *namebuf, ino_t curdir, ino_t ino)
853 {
854 int len;
855 char *cp;
856 struct inodesc idesc;
857 static int busy = 0;
858
859 if (curdir == ino && ino == ROOTINO) {
860 (void)strcpy(namebuf, "/");
861 return;
862 }
863 if (busy || !INO_IS_DVALID(curdir)) {
864 (void)strcpy(namebuf, "?");
865 return;
866 }
867 busy = 1;
868 memset(&idesc, 0, sizeof(struct inodesc));
869 idesc.id_type = DATA;
870 idesc.id_fix = IGNORE;
871 cp = &namebuf[MAXPATHLEN - 1];
872 *cp = '\0';
873 if (curdir != ino) {
874 idesc.id_parent = curdir;
875 goto namelookup;
876 }
877 while (ino != ROOTINO) {
878 idesc.id_number = ino;
879 idesc.id_func = findino;
880 idesc.id_name = strdup("..");
881 if ((ckinode(ginode(ino), &idesc) & FOUND) == 0)
882 break;
883 namelookup:
884 idesc.id_number = idesc.id_parent;
885 idesc.id_parent = ino;
886 idesc.id_func = findname;
887 idesc.id_name = namebuf;
888 if ((ckinode(ginode(idesc.id_number), &idesc)&FOUND) == 0)
889 break;
890 len = strlen(namebuf);
891 cp -= len;
892 memmove(cp, namebuf, (size_t)len);
893 *--cp = '/';
894 if (cp < &namebuf[MAXNAMLEN])
895 break;
896 ino = idesc.id_number;
897 }
898 busy = 0;
899 if (ino != ROOTINO)
900 *--cp = '?';
901 memmove(namebuf, cp, (size_t)(&namebuf[MAXPATHLEN] - cp));
902 }
903
904 void
catch(int sig __unused)905 catch(int sig __unused)
906 {
907
908 ckfini(0);
909 exit(12);
910 }
911
912 /*
913 * When preening, allow a single quit to signal
914 * a special exit after file system checks complete
915 * so that reboot sequence may be interrupted.
916 */
917 void
catchquit(int sig __unused)918 catchquit(int sig __unused)
919 {
920 printf("returning to single-user after file system check\n");
921 returntosingle = 1;
922 (void)signal(SIGQUIT, SIG_DFL);
923 }
924
925 /*
926 * determine whether an inode should be fixed.
927 */
928 int
dofix(struct inodesc * idesc,const char * msg)929 dofix(struct inodesc *idesc, const char *msg)
930 {
931
932 switch (idesc->id_fix) {
933
934 case DONTKNOW:
935 if (idesc->id_type == DATA)
936 direrror(idesc->id_number, msg);
937 else
938 pwarn("%s", msg);
939 if (preen) {
940 printf(" (SALVAGED)\n");
941 idesc->id_fix = FIX;
942 return (ALTERED);
943 }
944 if (reply("SALVAGE") == 0) {
945 idesc->id_fix = NOFIX;
946 return (0);
947 }
948 idesc->id_fix = FIX;
949 return (ALTERED);
950
951 case FIX:
952 return (ALTERED);
953
954 case NOFIX:
955 case IGNORE:
956 return (0);
957
958 default:
959 errx(EEXIT, "UNKNOWN INODESC FIX MODE %d", idesc->id_fix);
960 }
961 /* NOTREACHED */
962 return (0);
963 }
964
965 #include <stdarg.h>
966
967 /*
968 * An unexpected inconsistency occurred.
969 * Die if preening or file system is running with soft dependency protocol,
970 * otherwise just print message and continue.
971 */
972 void
pfatal(const char * fmt,...)973 pfatal(const char *fmt, ...)
974 {
975 va_list ap;
976 va_start(ap, fmt);
977 if (!preen) {
978 (void)vfprintf(stdout, fmt, ap);
979 va_end(ap);
980 if (usedsoftdep)
981 (void)fprintf(stdout,
982 "\nUNEXPECTED SOFT UPDATE INCONSISTENCY\n");
983 /*
984 * Force foreground fsck to clean up inconsistency.
985 */
986 if (bkgrdflag) {
987 cmd.value = FS_NEEDSFSCK;
988 cmd.size = 1;
989 if (sysctlbyname("vfs.ffs.setflags", 0, 0,
990 &cmd, sizeof cmd) == -1)
991 pwarn("CANNOT SET FS_NEEDSFSCK FLAG\n");
992 fprintf(stdout, "CANNOT RUN IN BACKGROUND\n");
993 ckfini(0);
994 exit(EEXIT);
995 }
996 return;
997 }
998 if (cdevname == NULL)
999 cdevname = strdup("fsck");
1000 (void)fprintf(stdout, "%s: ", cdevname);
1001 (void)vfprintf(stdout, fmt, ap);
1002 (void)fprintf(stdout,
1003 "\n%s: UNEXPECTED%sINCONSISTENCY; RUN fsck MANUALLY.\n",
1004 cdevname, usedsoftdep ? " SOFT UPDATE " : " ");
1005 /*
1006 * Force foreground fsck to clean up inconsistency.
1007 */
1008 if (bkgrdflag) {
1009 cmd.value = FS_NEEDSFSCK;
1010 cmd.size = 1;
1011 if (sysctlbyname("vfs.ffs.setflags", 0, 0,
1012 &cmd, sizeof cmd) == -1)
1013 pwarn("CANNOT SET FS_NEEDSFSCK FLAG\n");
1014 }
1015 ckfini(0);
1016 exit(EEXIT);
1017 }
1018
1019 /*
1020 * Pwarn just prints a message when not preening or running soft dependency
1021 * protocol, or a warning (preceded by filename) when preening.
1022 */
1023 void
pwarn(const char * fmt,...)1024 pwarn(const char *fmt, ...)
1025 {
1026 va_list ap;
1027 va_start(ap, fmt);
1028 if (preen)
1029 (void)fprintf(stdout, "%s: ", cdevname);
1030 (void)vfprintf(stdout, fmt, ap);
1031 va_end(ap);
1032 }
1033
1034 /*
1035 * Stub for routines from kernel.
1036 */
1037 void
panic(const char * fmt,...)1038 panic(const char *fmt, ...)
1039 {
1040 va_list ap;
1041 va_start(ap, fmt);
1042 pfatal("INTERNAL INCONSISTENCY:");
1043 (void)vfprintf(stdout, fmt, ap);
1044 va_end(ap);
1045 exit(EEXIT);
1046 }
1047