2 * Copyright (c) 1980, 1986, 1993
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
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.
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
32 static const char sccsid[] = "@(#)utilities.c 8.6 (Berkeley) 5/19/95";
35 #include <sys/cdefs.h>
36 __FBSDID("$FreeBSD$");
38 #include <sys/param.h>
40 #include <sys/types.h>
41 #include <sys/sysctl.h>
43 #include <sys/disklabel.h>
44 #include <sys/ioctl.h>
47 #include <ufs/ufs/dinode.h>
48 #include <ufs/ufs/dir.h>
49 #include <ufs/ffs/fs.h>
63 static void slowio_start(void);
64 static void slowio_end(void);
66 long diskreads, totalreads; /* Disk cache statistics */
67 struct timeval slowio_starttime;
68 int slowio_delay_usec = 10000; /* Initial IO delay for background fsck */
70 static TAILQ_HEAD(buflist, bufarea) bufhead; /* head of buffer cache list */
71 static int numbufs; /* size of buffer cache */
74 ftypeok(union dinode *dp)
76 switch (DIP(dp, di_mode) & IFMT) {
89 printf("bad file type 0%o\n", DIP(dp, di_mode));
95 reply(const char *question)
101 pfatal("INTERNAL ERROR: GOT TO reply()");
102 persevere = !strcmp(question, "CONTINUE");
104 if (!persevere && (nflag || (fswritefd < 0 && bkgrdflag == 0))) {
105 printf("%s? no\n\n", question);
109 if (yflag || (persevere && nflag)) {
110 printf("%s? yes\n\n", question);
114 printf("%s? [yn] ", question);
115 (void) fflush(stdout);
117 while (c != '\n' && getc(stdin) != '\n') {
123 } while (c != 'y' && c != 'Y' && c != 'n' && c != 'N');
125 if (c == 'y' || c == 'Y')
132 * Look up state information for an inode.
137 static struct inostat unallocated = { USTATE, 0, 0 };
138 struct inostatlist *ilp;
142 errx(EEXIT, "inoinfo: inumber %d out of range", inum);
143 ilp = &inostathead[inum / sblock.fs_ipg];
144 iloff = inum % sblock.fs_ipg;
145 if (iloff >= ilp->il_numalloced)
146 return (&unallocated);
147 return (&ilp->il_stat[iloff]);
151 * Malloc buffers and set up cache.
160 pbp = pdirbp = (struct bufarea *)0;
161 bufp = malloc((unsigned int)sblock.fs_bsize);
163 errx(EEXIT, "cannot allocate buffer pool");
164 cgblk.b_un.b_buf = bufp;
166 TAILQ_INIT(&bufhead);
168 if (bufcnt < MINBUFS)
170 for (i = 0; i < bufcnt; i++) {
171 bp = (struct bufarea *)malloc(sizeof(struct bufarea));
172 bufp = malloc((unsigned int)sblock.fs_bsize);
173 if (bp == NULL || bufp == NULL) {
176 errx(EEXIT, "cannot allocate buffer pool");
178 bp->b_un.b_buf = bufp;
179 TAILQ_INSERT_HEAD(&bufhead, bp, b_list);
182 numbufs = i; /* save number of buffers */
186 * Manage a cache of directory blocks.
189 getdatablk(ufs2_daddr_t blkno, long size)
193 TAILQ_FOREACH(bp, &bufhead, b_list)
194 if (bp->b_bno == fsbtodb(&sblock, blkno))
196 TAILQ_FOREACH_REVERSE(bp, &bufhead, buflist, b_list)
197 if ((bp->b_flags & B_INUSE) == 0)
200 errx(EEXIT, "deadlocked buffer pool");
201 getblk(bp, blkno, size);
204 TAILQ_REMOVE(&bufhead, bp, b_list);
205 TAILQ_INSERT_HEAD(&bufhead, bp, b_list);
206 bp->b_flags |= B_INUSE;
211 getblk(struct bufarea *bp, ufs2_daddr_t blk, long size)
216 dblk = fsbtodb(&sblock, blk);
217 if (bp->b_bno != dblk) {
218 flush(fswritefd, bp);
220 bp->b_errs = blread(fsreadfd, bp->b_un.b_buf, dblk, size);
227 flush(int fd, struct bufarea *bp)
235 pfatal("WRITING IN READ_ONLY MODE.\n");
239 pfatal("WRITING %sZERO'ED BLOCK %lld TO DISK\n",
240 (bp->b_errs == bp->b_size / dev_bsize) ? "" : "PARTIALLY ",
241 (long long)bp->b_bno);
243 blwrite(fd, bp->b_un.b_buf, bp->b_bno, (long)bp->b_size);
246 for (i = 0, j = 0; i < sblock.fs_cssize; i += sblock.fs_bsize, j++) {
247 blwrite(fswritefd, (char *)sblock.fs_csp + i,
248 fsbtodb(&sblock, sblock.fs_csaddr + j * sblock.fs_frag),
249 sblock.fs_cssize - i < sblock.fs_bsize ?
250 sblock.fs_cssize - i : sblock.fs_bsize);
255 rwerror(const char *mesg, ufs2_daddr_t blk)
262 pfatal("CANNOT %s: %ld", mesg, (long)blk);
263 if (reply("CONTINUE") == 0)
268 ckfini(int markclean)
270 struct bufarea *bp, *nbp;
271 int ofsmodified, cnt;
275 if ((!(sblock.fs_flags & FS_UNCLEAN)) != markclean) {
276 cmd.value = FS_UNCLEAN;
277 cmd.size = markclean ? -1 : 1;
278 if (sysctlbyname("vfs.ffs.setflags", 0, 0,
279 &cmd, sizeof cmd) == -1)
280 rwerror("SET FILE SYSTEM FLAGS", FS_UNCLEAN);
282 printf("\n***** FILE SYSTEM MARKED %s *****\n",
283 markclean ? "CLEAN" : "DIRTY");
287 } else if (!preen && !markclean) {
288 printf("\n***** FILE SYSTEM STILL DIRTY *****\n");
292 if (debug && totalreads > 0)
293 printf("cache with %d buffers missed %ld of %ld (%d%%)\n",
294 numbufs, diskreads, totalreads,
295 (int)(diskreads * 100 / totalreads));
297 (void)close(fsreadfd);
300 flush(fswritefd, &sblk);
301 if (havesb && cursnapshot == 0 && sblock.fs_magic == FS_UFS2_MAGIC &&
302 sblk.b_bno != sblock.fs_sblockloc / dev_bsize &&
303 !preen && reply("UPDATE STANDARD SUPERBLOCK")) {
304 sblk.b_bno = sblock.fs_sblockloc / dev_bsize;
306 flush(fswritefd, &sblk);
308 flush(fswritefd, &cgblk);
309 free(cgblk.b_un.b_buf);
311 TAILQ_FOREACH_REVERSE_SAFE(bp, &bufhead, buflist, b_list, nbp) {
312 TAILQ_REMOVE(&bufhead, bp, b_list);
314 flush(fswritefd, bp);
315 free(bp->b_un.b_buf);
319 errx(EEXIT, "panic: lost %d buffers", numbufs - cnt);
320 pbp = pdirbp = (struct bufarea *)0;
321 if (cursnapshot == 0 && sblock.fs_clean != markclean) {
322 if ((sblock.fs_clean = markclean) != 0) {
323 sblock.fs_flags &= ~(FS_UNCLEAN | FS_NEEDSFSCK);
324 sblock.fs_pendingblocks = 0;
325 sblock.fs_pendinginodes = 0;
328 ofsmodified = fsmodified;
329 flush(fswritefd, &sblk);
330 fsmodified = ofsmodified;
332 printf("\n***** FILE SYSTEM MARKED %s *****\n",
333 markclean ? "CLEAN" : "DIRTY");
339 printf("\n***** FILE SYSTEM IS CLEAN *****\n");
341 printf("\n***** FILE SYSTEM STILL DIRTY *****\n");
345 (void)close(fsreadfd);
346 (void)close(fswritefd);
350 blread(int fd, char *buf, ufs2_daddr_t blk, long size)
360 if (lseek(fd, offset, 0) < 0)
361 rwerror("SEEK BLK", blk);
362 else if (read(fd, buf, (int)size) == size) {
367 rwerror("READ BLK", blk);
368 if (lseek(fd, offset, 0) < 0)
369 rwerror("SEEK BLK", blk);
371 memset(buf, 0, (size_t)size);
372 printf("THE FOLLOWING DISK SECTORS COULD NOT BE READ:");
373 for (cp = buf, i = 0; i < size; i += secsize, cp += secsize) {
374 if (read(fd, cp, (int)secsize) != secsize) {
375 (void)lseek(fd, offset + i + secsize, 0);
376 if (secsize != dev_bsize && dev_bsize != 1)
377 printf(" %jd (%jd),",
378 (intmax_t)(blk * dev_bsize + i) / secsize,
379 (intmax_t)blk + i / dev_bsize);
381 printf(" %jd,", (intmax_t)blk + i / dev_bsize);
392 blwrite(int fd, char *buf, ufs2_daddr_t blk, long size)
402 if (lseek(fd, offset, 0) < 0)
403 rwerror("SEEK BLK", blk);
404 else if (write(fd, buf, (int)size) == size) {
409 rwerror("WRITE BLK", blk);
410 if (lseek(fd, offset, 0) < 0)
411 rwerror("SEEK BLK", blk);
412 printf("THE FOLLOWING SECTORS COULD NOT BE WRITTEN:");
413 for (cp = buf, i = 0; i < size; i += dev_bsize, cp += dev_bsize)
414 if (write(fd, cp, (int)dev_bsize) != dev_bsize) {
415 (void)lseek(fd, offset + i + dev_bsize, 0);
416 printf(" %jd,", (intmax_t)blk + i / dev_bsize);
423 blerase(int fd, ufs2_daddr_t blk, long size)
429 ioarg[0] = blk * dev_bsize;
431 ioctl(fd, DIOCGDELETE, ioarg);
432 /* we don't really care if we succeed or not */
437 * Verify cylinder group's magic number and other parameters. If the
438 * test fails, offer an option to rebuild the whole cylinder group.
441 check_cgmagic(int cg, struct cg *cgp)
445 * Extended cylinder group checks.
447 if (cg_chkmagic(cgp) &&
448 ((sblock.fs_magic == FS_UFS1_MAGIC &&
449 cgp->cg_old_niblk == sblock.fs_ipg &&
450 cgp->cg_ndblk <= sblock.fs_fpg &&
451 cgp->cg_old_ncyl <= sblock.fs_old_cpg) ||
452 (sblock.fs_magic == FS_UFS2_MAGIC &&
453 cgp->cg_niblk == sblock.fs_ipg &&
454 cgp->cg_ndblk <= sblock.fs_fpg &&
455 cgp->cg_initediblk <= sblock.fs_ipg))) {
458 pfatal("CYLINDER GROUP %d: BAD MAGIC NUMBER", cg);
459 if (!reply("REBUILD CYLINDER GROUP")) {
460 printf("YOU WILL NEED TO RERUN FSCK.\n");
465 * Zero out the cylinder group and then initialize critical fields.
466 * Bit maps and summaries will be recalculated by later passes.
468 memset(cgp, 0, (size_t)sblock.fs_cgsize);
469 cgp->cg_magic = CG_MAGIC;
471 cgp->cg_niblk = sblock.fs_ipg;
472 cgp->cg_initediblk = sblock.fs_ipg < 2 * INOPB(&sblock) ?
473 sblock.fs_ipg : 2 * INOPB(&sblock);
474 if (cgbase(&sblock, cg) + sblock.fs_fpg < sblock.fs_size)
475 cgp->cg_ndblk = sblock.fs_fpg;
477 cgp->cg_ndblk = sblock.fs_size - cgbase(&sblock, cg);
478 cgp->cg_iusedoff = &cgp->cg_space[0] - (u_char *)(&cgp->cg_firstfield);
479 if (sblock.fs_magic == FS_UFS1_MAGIC) {
481 cgp->cg_initediblk = 0;
482 cgp->cg_old_ncyl = sblock.fs_old_cpg;
483 cgp->cg_old_niblk = sblock.fs_ipg;
484 cgp->cg_old_btotoff = cgp->cg_iusedoff;
485 cgp->cg_old_boff = cgp->cg_old_btotoff +
486 sblock.fs_old_cpg * sizeof(int32_t);
487 cgp->cg_iusedoff = cgp->cg_old_boff +
488 sblock.fs_old_cpg * sizeof(u_int16_t);
490 cgp->cg_freeoff = cgp->cg_iusedoff + howmany(sblock.fs_ipg, CHAR_BIT);
491 cgp->cg_nextfreeoff = cgp->cg_freeoff + howmany(sblock.fs_fpg,CHAR_BIT);
492 if (sblock.fs_contigsumsize > 0) {
493 cgp->cg_nclusterblks = cgp->cg_ndblk / sblock.fs_frag;
494 cgp->cg_clustersumoff =
495 roundup(cgp->cg_nextfreeoff, sizeof(u_int32_t));
496 cgp->cg_clustersumoff -= sizeof(u_int32_t);
497 cgp->cg_clusteroff = cgp->cg_clustersumoff +
498 (sblock.fs_contigsumsize + 1) * sizeof(u_int32_t);
499 cgp->cg_nextfreeoff = cgp->cg_clusteroff +
500 howmany(fragstoblks(&sblock, sblock.fs_fpg), CHAR_BIT);
507 * allocate a data block with the specified number of fragments
512 int i, j, k, cg, baseblk;
513 struct cg *cgp = &cgrp;
515 if (frags <= 0 || frags > sblock.fs_frag)
517 for (i = 0; i < maxfsblock - sblock.fs_frag; i += sblock.fs_frag) {
518 for (j = 0; j <= sblock.fs_frag - frags; j++) {
521 for (k = 1; k < frags; k++)
522 if (testbmap(i + j + k))
528 cg = dtog(&sblock, i + j);
529 getblk(&cgblk, cgtod(&sblock, cg), sblock.fs_cgsize);
530 if (!check_cgmagic(cg, cgp))
532 baseblk = dtogd(&sblock, i + j);
533 for (k = 0; k < frags; k++) {
535 clrbit(cg_blksfree(cgp), baseblk + k);
538 if (frags == sblock.fs_frag)
539 cgp->cg_cs.cs_nbfree--;
541 cgp->cg_cs.cs_nffree -= frags;
550 * Free a previously allocated block
553 freeblk(ufs2_daddr_t blkno, long frags)
555 struct inodesc idesc;
557 idesc.id_blkno = blkno;
558 idesc.id_numfrags = frags;
559 (void)pass4check(&idesc);
562 /* Slow down IO so as to leave some disk bandwidth for other processes */
567 /* Delay one in every 8 operations */
568 slowio_pollcnt = (slowio_pollcnt + 1) & 7;
569 if (slowio_pollcnt == 0) {
570 gettimeofday(&slowio_starttime, NULL);
580 if (slowio_pollcnt != 0)
583 /* Update the slowdown interval. */
584 gettimeofday(&tv, NULL);
585 delay_usec = (tv.tv_sec - slowio_starttime.tv_sec) * 1000000 +
586 (tv.tv_usec - slowio_starttime.tv_usec);
589 if (delay_usec > 2500000)
590 delay_usec = 2500000;
591 slowio_delay_usec = (slowio_delay_usec * 63 + delay_usec) >> 6;
592 /* delay by 8 times the average IO delay */
593 if (slowio_delay_usec > 64)
594 usleep(slowio_delay_usec * 8);
601 getpathname(char *namebuf, ino_t curdir, ino_t ino)
605 struct inodesc idesc;
608 if (curdir == ino && ino == ROOTINO) {
609 (void)strcpy(namebuf, "/");
612 if (busy || !INO_IS_DVALID(curdir)) {
613 (void)strcpy(namebuf, "?");
617 memset(&idesc, 0, sizeof(struct inodesc));
618 idesc.id_type = DATA;
619 idesc.id_fix = IGNORE;
620 cp = &namebuf[MAXPATHLEN - 1];
623 idesc.id_parent = curdir;
626 while (ino != ROOTINO) {
627 idesc.id_number = ino;
628 idesc.id_func = findino;
629 idesc.id_name = strdup("..");
630 if ((ckinode(ginode(ino), &idesc) & FOUND) == 0)
633 idesc.id_number = idesc.id_parent;
634 idesc.id_parent = ino;
635 idesc.id_func = findname;
636 idesc.id_name = namebuf;
637 if ((ckinode(ginode(idesc.id_number), &idesc)&FOUND) == 0)
639 len = strlen(namebuf);
641 memmove(cp, namebuf, (size_t)len);
643 if (cp < &namebuf[MAXNAMLEN])
645 ino = idesc.id_number;
650 memmove(namebuf, cp, (size_t)(&namebuf[MAXPATHLEN] - cp));
654 catch(int sig __unused)
662 * When preening, allow a single quit to signal
663 * a special exit after file system checks complete
664 * so that reboot sequence may be interrupted.
667 catchquit(int sig __unused)
669 printf("returning to single-user after file system check\n");
671 (void)signal(SIGQUIT, SIG_DFL);
675 * determine whether an inode should be fixed.
678 dofix(struct inodesc *idesc, const char *msg)
681 switch (idesc->id_fix) {
684 if (idesc->id_type == DATA)
685 direrror(idesc->id_number, msg);
689 printf(" (SALVAGED)\n");
693 if (reply("SALVAGE") == 0) {
694 idesc->id_fix = NOFIX;
708 errx(EEXIT, "UNKNOWN INODESC FIX MODE %d", idesc->id_fix);
717 * An unexpected inconsistency occured.
718 * Die if preening or file system is running with soft dependency protocol,
719 * otherwise just print message and continue.
722 pfatal(const char *fmt, ...)
727 (void)vfprintf(stdout, fmt, ap);
730 (void)fprintf(stdout,
731 "\nUNEXPECTED SOFT UPDATE INCONSISTENCY\n");
733 * Force foreground fsck to clean up inconsistency.
736 cmd.value = FS_NEEDSFSCK;
738 if (sysctlbyname("vfs.ffs.setflags", 0, 0,
739 &cmd, sizeof cmd) == -1)
740 pwarn("CANNOT SET FS_NEEDSFSCK FLAG\n");
741 fprintf(stdout, "CANNOT RUN IN BACKGROUND\n");
747 if (cdevname == NULL)
748 cdevname = strdup("fsck");
749 (void)fprintf(stdout, "%s: ", cdevname);
750 (void)vfprintf(stdout, fmt, ap);
751 (void)fprintf(stdout,
752 "\n%s: UNEXPECTED%sINCONSISTENCY; RUN fsck MANUALLY.\n",
753 cdevname, usedsoftdep ? " SOFT UPDATE " : " ");
755 * Force foreground fsck to clean up inconsistency.
758 cmd.value = FS_NEEDSFSCK;
760 if (sysctlbyname("vfs.ffs.setflags", 0, 0,
761 &cmd, sizeof cmd) == -1)
762 pwarn("CANNOT SET FS_NEEDSFSCK FLAG\n");
769 * Pwarn just prints a message when not preening or running soft dependency
770 * protocol, or a warning (preceded by filename) when preening.
773 pwarn(const char *fmt, ...)
778 (void)fprintf(stdout, "%s: ", cdevname);
779 (void)vfprintf(stdout, fmt, ap);
784 * Stub for routines from kernel.
787 panic(const char *fmt, ...)
791 pfatal("INTERNAL INCONSISTENCY:");
792 (void)vfprintf(stdout, fmt, ap);