2 * Copyright (c) 2002 Networks Associates Technology, Inc.
5 * This software was developed for the FreeBSD Project by Marshall
6 * Kirk McKusick and Network Associates Laboratories, the Security
7 * Research Division of Network Associates, Inc. under DARPA/SPAWAR
8 * contract N66001-01-C-8035 ("CBOSS"), as part of the DARPA CHATS
11 * Copyright (c) 1980, 1989, 1993
12 * The Regents of the University of California. All rights reserved.
14 * Redistribution and use in source and binary forms, with or without
15 * modification, are permitted provided that the following conditions
17 * 1. Redistributions of source code must retain the above copyright
18 * notice, this list of conditions and the following disclaimer.
19 * 2. Redistributions in binary form must reproduce the above copyright
20 * notice, this list of conditions and the following disclaimer in the
21 * documentation and/or other materials provided with the distribution.
22 * 4. Neither the name of the University nor the names of its contributors
23 * may be used to endorse or promote products derived from this software
24 * without specific prior written permission.
26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
41 static char sccsid[] = "@(#)mkfs.c 8.11 (Berkeley) 5/3/95";
44 #include <sys/cdefs.h>
45 __FBSDID("$FreeBSD$");
47 #include <sys/param.h>
48 #include <sys/disklabel.h>
50 #include <sys/ioctl.h>
52 #include <sys/resource.h>
65 #include <ufs/ufs/dinode.h>
66 #include <ufs/ufs/dir.h>
67 #include <ufs/ffs/fs.h>
71 * make file system for cylinder-group style file systems
74 #define POWEROF2(num) (((num) & ((num) - 1)) == 0)
76 static struct csum *fscs;
77 #define sblock disk.d_fs
81 struct ufs1_dinode dp1;
82 struct ufs2_dinode dp2;
84 #define DIP(dp, field) \
85 ((sblock.fs_magic == FS_UFS1_MAGIC) ? \
86 (dp)->dp1.field : (dp)->dp2.field)
89 static long iobufsize;
90 static ufs2_daddr_t alloc(int size, int mode);
91 static int charsperline(void);
92 static void clrblock(struct fs *, unsigned char *, int);
93 static void fsinit(time_t);
94 static int ilog2(int);
95 static void initcg(int, time_t);
96 static int isblock(struct fs *, unsigned char *, int);
97 static void iput(union dinode *, ino_t);
98 static int makedir(struct direct *, int);
99 static void setblock(struct fs *, unsigned char *, int);
100 static void wtfs(ufs2_daddr_t, int, char *);
101 static u_int32_t newfs_random(void);
104 do_sbwrite(struct uufsd *disk)
107 disk->d_sblock = disk->d_fs.fs_sblockloc / disk->d_bsize;
108 return (pwrite(disk->d_fd, &disk->d_fs, SBLOCKSIZE, (off_t)((part_ofs +
109 disk->d_sblock) * disk->d_bsize)));
113 mkfs(struct partition *pp, char *fsys)
115 int fragsperinode, optimalfpg, origdensity, minfpg, lastminfpg;
122 int minfragsperinode; /* minimum ratio of frags to inodes */
123 char tmpbuf[100]; /* XXX this will break in about 2,500 years */
124 struct fsrecovery fsr;
127 char cdummy[SBLOCKSIZE];
129 #define fsdummy dummy.fdummy
130 #define chdummy dummy.cdummy
133 * Our blocks == sector size, and the version of UFS we are using is
134 * specified by Oflag.
136 disk.d_bsize = sectorsize;
144 sblock.fs_old_flags = FS_FLAGS_UPDATED;
147 sblock.fs_flags |= FS_DOSOFTDEP;
149 strlcpy(sblock.fs_volname, volumelabel, MAXVOLLEN);
151 sblock.fs_flags |= FS_GJOURNAL;
153 sblock.fs_flags |= FS_MULTILABEL;
155 sblock.fs_flags |= FS_TRIM;
157 * Validate the given file system size.
158 * Verify that its last block can actually be accessed.
159 * Convert to file system fragment sized units.
162 printf("preposterous size %jd\n", (intmax_t)fssize);
165 wtfs(fssize - (realsectorsize / DEV_BSIZE), realsectorsize,
168 * collect and verify the file system density info
170 sblock.fs_avgfilesize = avgfilesize;
171 sblock.fs_avgfpdir = avgfilesperdir;
172 if (sblock.fs_avgfilesize <= 0)
173 printf("illegal expected average file size %d\n",
174 sblock.fs_avgfilesize), exit(14);
175 if (sblock.fs_avgfpdir <= 0)
176 printf("illegal expected number of files per directory %d\n",
177 sblock.fs_avgfpdir), exit(15);
181 * collect and verify the block and fragment sizes
183 sblock.fs_bsize = bsize;
184 sblock.fs_fsize = fsize;
185 if (!POWEROF2(sblock.fs_bsize)) {
186 printf("block size must be a power of 2, not %d\n",
190 if (!POWEROF2(sblock.fs_fsize)) {
191 printf("fragment size must be a power of 2, not %d\n",
195 if (sblock.fs_fsize < sectorsize) {
196 printf("increasing fragment size from %d to sector size (%d)\n",
197 sblock.fs_fsize, sectorsize);
198 sblock.fs_fsize = sectorsize;
200 if (sblock.fs_bsize > MAXBSIZE) {
201 printf("decreasing block size from %d to maximum (%d)\n",
202 sblock.fs_bsize, MAXBSIZE);
203 sblock.fs_bsize = MAXBSIZE;
205 if (sblock.fs_bsize < MINBSIZE) {
206 printf("increasing block size from %d to minimum (%d)\n",
207 sblock.fs_bsize, MINBSIZE);
208 sblock.fs_bsize = MINBSIZE;
210 if (sblock.fs_fsize > MAXBSIZE) {
211 printf("decreasing fragment size from %d to maximum (%d)\n",
212 sblock.fs_fsize, MAXBSIZE);
213 sblock.fs_fsize = MAXBSIZE;
215 if (sblock.fs_bsize < sblock.fs_fsize) {
216 printf("increasing block size from %d to fragment size (%d)\n",
217 sblock.fs_bsize, sblock.fs_fsize);
218 sblock.fs_bsize = sblock.fs_fsize;
220 if (sblock.fs_fsize * MAXFRAG < sblock.fs_bsize) {
222 "increasing fragment size from %d to block size / %d (%d)\n",
223 sblock.fs_fsize, MAXFRAG, sblock.fs_bsize / MAXFRAG);
224 sblock.fs_fsize = sblock.fs_bsize / MAXFRAG;
228 if (maxbsize < bsize || !POWEROF2(maxbsize)) {
229 sblock.fs_maxbsize = sblock.fs_bsize;
230 printf("Extent size set to %d\n", sblock.fs_maxbsize);
231 } else if (sblock.fs_maxbsize > FS_MAXCONTIG * sblock.fs_bsize) {
232 sblock.fs_maxbsize = FS_MAXCONTIG * sblock.fs_bsize;
233 printf("Extent size reduced to %d\n", sblock.fs_maxbsize);
235 sblock.fs_maxbsize = maxbsize;
238 * Maxcontig sets the default for the maximum number of blocks
239 * that may be allocated sequentially. With file system clustering
240 * it is possible to allocate contiguous blocks up to the maximum
241 * transfer size permitted by the controller or buffering.
244 maxcontig = MAX(1, MAXPHYS / bsize);
245 sblock.fs_maxcontig = maxcontig;
246 if (sblock.fs_maxcontig < sblock.fs_maxbsize / sblock.fs_bsize) {
247 sblock.fs_maxcontig = sblock.fs_maxbsize / sblock.fs_bsize;
248 printf("Maxcontig raised to %d\n", sblock.fs_maxbsize);
250 if (sblock.fs_maxcontig > 1)
251 sblock.fs_contigsumsize = MIN(sblock.fs_maxcontig,FS_MAXCONTIG);
252 sblock.fs_bmask = ~(sblock.fs_bsize - 1);
253 sblock.fs_fmask = ~(sblock.fs_fsize - 1);
254 sblock.fs_qbmask = ~sblock.fs_bmask;
255 sblock.fs_qfmask = ~sblock.fs_fmask;
256 sblock.fs_bshift = ilog2(sblock.fs_bsize);
257 sblock.fs_fshift = ilog2(sblock.fs_fsize);
258 sblock.fs_frag = numfrags(&sblock, sblock.fs_bsize);
259 sblock.fs_fragshift = ilog2(sblock.fs_frag);
260 if (sblock.fs_frag > MAXFRAG) {
261 printf("fragment size %d is still too small (can't happen)\n",
262 sblock.fs_bsize / MAXFRAG);
265 sblock.fs_fsbtodb = ilog2(sblock.fs_fsize / sectorsize);
266 sblock.fs_size = fssize = dbtofsb(&sblock, fssize);
267 sblock.fs_providersize = dbtofsb(&sblock, mediasize / sectorsize);
270 * Before the filesystem is finally initialized, mark it
271 * as incompletely initialized.
273 sblock.fs_magic = FS_BAD_MAGIC;
276 sblock.fs_sblockloc = SBLOCK_UFS1;
277 sblock.fs_nindir = sblock.fs_bsize / sizeof(ufs1_daddr_t);
278 sblock.fs_inopb = sblock.fs_bsize / sizeof(struct ufs1_dinode);
279 sblock.fs_maxsymlinklen = ((NDADDR + NIADDR) *
280 sizeof(ufs1_daddr_t));
281 sblock.fs_old_inodefmt = FS_44INODEFMT;
282 sblock.fs_old_cgoffset = 0;
283 sblock.fs_old_cgmask = 0xffffffff;
284 sblock.fs_old_size = sblock.fs_size;
285 sblock.fs_old_rotdelay = 0;
286 sblock.fs_old_rps = 60;
287 sblock.fs_old_nspf = sblock.fs_fsize / sectorsize;
288 sblock.fs_old_cpg = 1;
289 sblock.fs_old_interleave = 1;
290 sblock.fs_old_trackskew = 0;
291 sblock.fs_old_cpc = 0;
292 sblock.fs_old_postblformat = 1;
293 sblock.fs_old_nrpos = 1;
295 sblock.fs_sblockloc = SBLOCK_UFS2;
296 sblock.fs_nindir = sblock.fs_bsize / sizeof(ufs2_daddr_t);
297 sblock.fs_inopb = sblock.fs_bsize / sizeof(struct ufs2_dinode);
298 sblock.fs_maxsymlinklen = ((NDADDR + NIADDR) *
299 sizeof(ufs2_daddr_t));
302 roundup(howmany(sblock.fs_sblockloc + SBLOCKSIZE, sblock.fs_fsize),
304 sblock.fs_cblkno = sblock.fs_sblkno +
305 roundup(howmany(SBLOCKSIZE, sblock.fs_fsize), sblock.fs_frag);
306 sblock.fs_iblkno = sblock.fs_cblkno + sblock.fs_frag;
307 sblock.fs_maxfilesize = sblock.fs_bsize * NDADDR - 1;
308 for (sizepb = sblock.fs_bsize, i = 0; i < NIADDR; i++) {
309 sizepb *= NINDIR(&sblock);
310 sblock.fs_maxfilesize += sizepb;
314 * It's impossible to create a snapshot in case that fs_maxfilesize
315 * is smaller than the fssize.
317 if (sblock.fs_maxfilesize < (u_quad_t)fssize) {
318 warnx("WARNING: You will be unable to create snapshots on this "
319 "file system. Correct by using a larger blocksize.");
323 * Calculate the number of blocks to put into each cylinder group.
325 * This algorithm selects the number of blocks per cylinder
326 * group. The first goal is to have at least enough data blocks
327 * in each cylinder group to meet the density requirement. Once
328 * this goal is achieved we try to expand to have at least
329 * MINCYLGRPS cylinder groups. Once this goal is achieved, we
330 * pack as many blocks into each cylinder group map as will fit.
332 * We start by calculating the smallest number of blocks that we
333 * can put into each cylinder group. If this is too big, we reduce
334 * the density until it fits.
336 maxinum = (((int64_t)(1)) << 32) - INOPB(&sblock);
337 minfragsperinode = 1 + fssize / maxinum;
339 density = MAX(NFPI, minfragsperinode) * fsize;
340 } else if (density < minfragsperinode * fsize) {
341 origdensity = density;
342 density = minfragsperinode * fsize;
343 fprintf(stderr, "density increased from %d to %d\n",
344 origdensity, density);
346 origdensity = density;
348 fragsperinode = MAX(numfrags(&sblock, density), 1);
349 if (fragsperinode < minfragsperinode) {
352 printf("Block size too small for a file system %s %d\n",
353 "of this size. Increasing blocksize to", bsize);
356 minfpg = fragsperinode * INOPB(&sblock);
357 if (minfpg > sblock.fs_size)
358 minfpg = sblock.fs_size;
359 sblock.fs_ipg = INOPB(&sblock);
360 sblock.fs_fpg = roundup(sblock.fs_iblkno +
361 sblock.fs_ipg / INOPF(&sblock), sblock.fs_frag);
362 if (sblock.fs_fpg < minfpg)
363 sblock.fs_fpg = minfpg;
364 sblock.fs_ipg = roundup(howmany(sblock.fs_fpg, fragsperinode),
366 sblock.fs_fpg = roundup(sblock.fs_iblkno +
367 sblock.fs_ipg / INOPF(&sblock), sblock.fs_frag);
368 if (sblock.fs_fpg < minfpg)
369 sblock.fs_fpg = minfpg;
370 sblock.fs_ipg = roundup(howmany(sblock.fs_fpg, fragsperinode),
372 if (CGSIZE(&sblock) < (unsigned long)sblock.fs_bsize)
374 density -= sblock.fs_fsize;
376 if (density != origdensity)
377 printf("density reduced from %d to %d\n", origdensity, density);
379 * Start packing more blocks into the cylinder group until
380 * it cannot grow any larger, the number of cylinder groups
381 * drops below MINCYLGRPS, or we reach the size requested.
382 * For UFS1 inodes per cylinder group are stored in an int16_t
383 * so fs_ipg is limited to 2^15 - 1.
385 for ( ; sblock.fs_fpg < maxblkspercg; sblock.fs_fpg += sblock.fs_frag) {
386 sblock.fs_ipg = roundup(howmany(sblock.fs_fpg, fragsperinode),
388 if (Oflag > 1 || (Oflag == 1 && sblock.fs_ipg <= 0x7fff)) {
389 if (sblock.fs_size / sblock.fs_fpg < MINCYLGRPS)
391 if (CGSIZE(&sblock) < (unsigned long)sblock.fs_bsize)
393 if (CGSIZE(&sblock) == (unsigned long)sblock.fs_bsize)
396 sblock.fs_fpg -= sblock.fs_frag;
397 sblock.fs_ipg = roundup(howmany(sblock.fs_fpg, fragsperinode),
402 * Check to be sure that the last cylinder group has enough blocks
403 * to be viable. If it is too small, reduce the number of blocks
404 * per cylinder group which will have the effect of moving more
405 * blocks into the last cylinder group.
407 optimalfpg = sblock.fs_fpg;
409 sblock.fs_ncg = howmany(sblock.fs_size, sblock.fs_fpg);
410 lastminfpg = roundup(sblock.fs_iblkno +
411 sblock.fs_ipg / INOPF(&sblock), sblock.fs_frag);
412 if (sblock.fs_size < lastminfpg) {
413 printf("Filesystem size %jd < minimum size of %d\n",
414 (intmax_t)sblock.fs_size, lastminfpg);
417 if (sblock.fs_size % sblock.fs_fpg >= lastminfpg ||
418 sblock.fs_size % sblock.fs_fpg == 0)
420 sblock.fs_fpg -= sblock.fs_frag;
421 sblock.fs_ipg = roundup(howmany(sblock.fs_fpg, fragsperinode),
424 if (optimalfpg != sblock.fs_fpg)
425 printf("Reduced frags per cylinder group from %d to %d %s\n",
426 optimalfpg, sblock.fs_fpg, "to enlarge last cyl group");
427 sblock.fs_cgsize = fragroundup(&sblock, CGSIZE(&sblock));
428 sblock.fs_dblkno = sblock.fs_iblkno + sblock.fs_ipg / INOPF(&sblock);
430 sblock.fs_old_spc = sblock.fs_fpg * sblock.fs_old_nspf;
431 sblock.fs_old_nsect = sblock.fs_old_spc;
432 sblock.fs_old_npsect = sblock.fs_old_spc;
433 sblock.fs_old_ncyl = sblock.fs_ncg;
436 * fill in remaining fields of the super block
438 sblock.fs_csaddr = cgdmin(&sblock, 0);
440 fragroundup(&sblock, sblock.fs_ncg * sizeof(struct csum));
441 fscs = (struct csum *)calloc(1, sblock.fs_cssize);
443 errx(31, "calloc failed");
444 sblock.fs_sbsize = fragroundup(&sblock, sizeof(struct fs));
445 if (sblock.fs_sbsize > SBLOCKSIZE)
446 sblock.fs_sbsize = SBLOCKSIZE;
447 sblock.fs_minfree = minfree;
448 if (metaspace > 0 && metaspace < sblock.fs_fpg / 2)
449 sblock.fs_metaspace = blknum(&sblock, metaspace);
450 else if (metaspace != -1)
451 /* reserve half of minfree for metadata blocks */
452 sblock.fs_metaspace = blknum(&sblock,
453 (sblock.fs_fpg * minfree) / 200);
455 sblock.fs_maxbpg = MAXBLKPG(sblock.fs_bsize);
457 sblock.fs_maxbpg = maxbpg;
458 sblock.fs_optim = opt;
459 sblock.fs_cgrotor = 0;
460 sblock.fs_pendingblocks = 0;
461 sblock.fs_pendinginodes = 0;
466 sblock.fs_id[0] = (long)utime;
467 sblock.fs_id[1] = newfs_random();
468 sblock.fs_fsmnt[0] = '\0';
469 csfrags = howmany(sblock.fs_cssize, sblock.fs_fsize);
470 sblock.fs_dsize = sblock.fs_size - sblock.fs_sblkno -
471 sblock.fs_ncg * (sblock.fs_dblkno - sblock.fs_sblkno);
472 sblock.fs_cstotal.cs_nbfree =
473 fragstoblks(&sblock, sblock.fs_dsize) -
474 howmany(csfrags, sblock.fs_frag);
475 sblock.fs_cstotal.cs_nffree =
476 fragnum(&sblock, sblock.fs_size) +
477 (fragnum(&sblock, csfrags) > 0 ?
478 sblock.fs_frag - fragnum(&sblock, csfrags) : 0);
479 sblock.fs_cstotal.cs_nifree = sblock.fs_ncg * sblock.fs_ipg - ROOTINO;
480 sblock.fs_cstotal.cs_ndir = 0;
481 sblock.fs_dsize -= csfrags;
482 sblock.fs_time = utime;
484 sblock.fs_old_time = utime;
485 sblock.fs_old_dsize = sblock.fs_dsize;
486 sblock.fs_old_csaddr = sblock.fs_csaddr;
487 sblock.fs_old_cstotal.cs_ndir = sblock.fs_cstotal.cs_ndir;
488 sblock.fs_old_cstotal.cs_nbfree = sblock.fs_cstotal.cs_nbfree;
489 sblock.fs_old_cstotal.cs_nifree = sblock.fs_cstotal.cs_nifree;
490 sblock.fs_old_cstotal.cs_nffree = sblock.fs_cstotal.cs_nffree;
494 * Dump out summary information about file system.
496 # define B2MBFACTOR (1 / (1024.0 * 1024.0))
497 printf("%s: %.1fMB (%jd sectors) block size %d, fragment size %d\n",
498 fsys, (float)sblock.fs_size * sblock.fs_fsize * B2MBFACTOR,
499 (intmax_t)fsbtodb(&sblock, sblock.fs_size), sblock.fs_bsize,
501 printf("\tusing %d cylinder groups of %.2fMB, %d blks, %d inodes.\n",
502 sblock.fs_ncg, (float)sblock.fs_fpg * sblock.fs_fsize * B2MBFACTOR,
503 sblock.fs_fpg / sblock.fs_frag, sblock.fs_ipg);
504 if (sblock.fs_flags & FS_DOSOFTDEP)
505 printf("\twith soft updates\n");
508 if (Eflag && !Nflag) {
509 printf("Erasing sectors [%jd...%jd]\n",
510 sblock.fs_sblockloc / disk.d_bsize,
511 fsbtodb(&sblock, sblock.fs_size) - 1);
512 berase(&disk, sblock.fs_sblockloc / disk.d_bsize,
513 sblock.fs_size * sblock.fs_fsize - sblock.fs_sblockloc);
516 * Wipe out old UFS1 superblock(s) if necessary.
518 if (!Nflag && Oflag != 1) {
519 i = bread(&disk, part_ofs + SBLOCK_UFS1 / disk.d_bsize, chdummy, SBLOCKSIZE);
521 err(1, "can't read old UFS1 superblock: %s", disk.d_error);
523 if (fsdummy.fs_magic == FS_UFS1_MAGIC) {
524 fsdummy.fs_magic = 0;
525 bwrite(&disk, part_ofs + SBLOCK_UFS1 / disk.d_bsize,
526 chdummy, SBLOCKSIZE);
527 for (cg = 0; cg < fsdummy.fs_ncg; cg++) {
528 if (fsbtodb(&fsdummy, cgsblock(&fsdummy, cg)) > fssize)
530 bwrite(&disk, part_ofs + fsbtodb(&fsdummy,
531 cgsblock(&fsdummy, cg)), chdummy, SBLOCKSIZE);
538 printf("** Exiting on Xflag 1\n");
542 printf("** Leaving BAD MAGIC on Xflag 2\n");
544 sblock.fs_magic = (Oflag != 1) ? FS_UFS2_MAGIC : FS_UFS1_MAGIC;
547 * Now build the cylinders group blocks and
548 * then print out indices of cylinder groups.
550 printf("super-block backups (for fsck_ffs -b #) at:\n");
552 width = charsperline();
554 * allocate space for superblock, cylinder group map, and
555 * two sets of inode blocks.
557 if (sblock.fs_bsize < SBLOCKSIZE)
558 iobufsize = SBLOCKSIZE + 3 * sblock.fs_bsize;
560 iobufsize = 4 * sblock.fs_bsize;
561 if ((iobuf = calloc(1, iobufsize)) == 0) {
562 printf("Cannot allocate I/O buffer\n");
566 * Make a copy of the superblock into the buffer that we will be
567 * writing out in each cylinder group.
569 bcopy((char *)&sblock, iobuf, SBLOCKSIZE);
570 for (cg = 0; cg < sblock.fs_ncg; cg++) {
572 j = snprintf(tmpbuf, sizeof(tmpbuf), " %jd%s",
573 (intmax_t)fsbtodb(&sblock, cgsblock(&sblock, cg)),
574 cg < (sblock.fs_ncg-1) ? "," : "");
576 tmpbuf[j = 0] = '\0';
577 if (i + j >= width) {
582 printf("%s", tmpbuf);
589 * Now construct the initial file system,
590 * then write out the super-block.
594 sblock.fs_old_cstotal.cs_ndir = sblock.fs_cstotal.cs_ndir;
595 sblock.fs_old_cstotal.cs_nbfree = sblock.fs_cstotal.cs_nbfree;
596 sblock.fs_old_cstotal.cs_nifree = sblock.fs_cstotal.cs_nifree;
597 sblock.fs_old_cstotal.cs_nffree = sblock.fs_cstotal.cs_nffree;
600 printf("** Exiting on Xflag 3\n");
606 * For UFS1 filesystems with a blocksize of 64K, the first
607 * alternate superblock resides at the location used for
608 * the default UFS2 superblock. As there is a valid
609 * superblock at this location, the boot code will use
610 * it as its first choice. Thus we have to ensure that
611 * all of its statistcs on usage are correct.
613 if (Oflag == 1 && sblock.fs_bsize == 65536)
614 wtfs(fsbtodb(&sblock, cgsblock(&sblock, 0)),
615 sblock.fs_bsize, (char *)&sblock);
617 for (i = 0; i < sblock.fs_cssize; i += sblock.fs_bsize)
618 wtfs(fsbtodb(&sblock, sblock.fs_csaddr + numfrags(&sblock, i)),
619 sblock.fs_cssize - i < sblock.fs_bsize ?
620 sblock.fs_cssize - i : sblock.fs_bsize,
623 * Read the last sector of the boot block, replace the last
624 * 20 bytes with the recovery information, then write it back.
625 * The recovery information only works for UFS2 filesystems.
627 if (sblock.fs_magic == FS_UFS2_MAGIC) {
629 part_ofs + (SBLOCK_UFS2 - sizeof(fsr)) / disk.d_bsize,
630 (char *)&fsr, sizeof(fsr));
632 err(1, "can't read recovery area: %s", disk.d_error);
633 fsr.fsr_magic = sblock.fs_magic;
634 fsr.fsr_fpg = sblock.fs_fpg;
635 fsr.fsr_fsbtodb = sblock.fs_fsbtodb;
636 fsr.fsr_sblkno = sblock.fs_sblkno;
637 fsr.fsr_ncg = sblock.fs_ncg;
638 wtfs((SBLOCK_UFS2 - sizeof(fsr)) / disk.d_bsize, sizeof(fsr),
642 * Update information about this partition in pack
643 * label, to that it may be updated on disk.
646 pp->p_fstype = FS_BSDFFS;
647 pp->p_fsize = sblock.fs_fsize;
648 pp->p_frag = sblock.fs_frag;
649 pp->p_cpg = sblock.fs_fpg;
654 * Initialize a cylinder group.
657 initcg(int cylno, time_t utime)
660 uint i, j, d, dlower, dupper;
661 ufs2_daddr_t cbase, dmax;
662 struct ufs1_dinode *dp1;
663 struct ufs2_dinode *dp2;
667 * Determine block bounds for cylinder group.
668 * Allow space for super block summary information in first
671 cbase = cgbase(&sblock, cylno);
672 dmax = cbase + sblock.fs_fpg;
673 if (dmax > sblock.fs_size)
674 dmax = sblock.fs_size;
675 dlower = cgsblock(&sblock, cylno) - cbase;
676 dupper = cgdmin(&sblock, cylno) - cbase;
678 dupper += howmany(sblock.fs_cssize, sblock.fs_fsize);
680 memset(&acg, 0, sblock.fs_cgsize);
682 acg.cg_magic = CG_MAGIC;
684 acg.cg_niblk = sblock.fs_ipg;
685 acg.cg_initediblk = sblock.fs_ipg < 2 * INOPB(&sblock) ?
686 sblock.fs_ipg : 2 * INOPB(&sblock);
687 acg.cg_ndblk = dmax - cbase;
688 if (sblock.fs_contigsumsize > 0)
689 acg.cg_nclusterblks = acg.cg_ndblk / sblock.fs_frag;
690 start = &acg.cg_space[0] - (u_char *)(&acg.cg_firstfield);
692 acg.cg_iusedoff = start;
694 acg.cg_old_ncyl = sblock.fs_old_cpg;
695 acg.cg_old_time = acg.cg_time;
697 acg.cg_old_niblk = acg.cg_niblk;
699 acg.cg_initediblk = 0;
700 acg.cg_old_btotoff = start;
701 acg.cg_old_boff = acg.cg_old_btotoff +
702 sblock.fs_old_cpg * sizeof(int32_t);
703 acg.cg_iusedoff = acg.cg_old_boff +
704 sblock.fs_old_cpg * sizeof(u_int16_t);
706 acg.cg_freeoff = acg.cg_iusedoff + howmany(sblock.fs_ipg, CHAR_BIT);
707 acg.cg_nextfreeoff = acg.cg_freeoff + howmany(sblock.fs_fpg, CHAR_BIT);
708 if (sblock.fs_contigsumsize > 0) {
709 acg.cg_clustersumoff =
710 roundup(acg.cg_nextfreeoff, sizeof(u_int32_t));
711 acg.cg_clustersumoff -= sizeof(u_int32_t);
712 acg.cg_clusteroff = acg.cg_clustersumoff +
713 (sblock.fs_contigsumsize + 1) * sizeof(u_int32_t);
714 acg.cg_nextfreeoff = acg.cg_clusteroff +
715 howmany(fragstoblks(&sblock, sblock.fs_fpg), CHAR_BIT);
717 if (acg.cg_nextfreeoff > (unsigned)sblock.fs_cgsize) {
718 printf("Panic: cylinder group too big\n");
721 acg.cg_cs.cs_nifree += sblock.fs_ipg;
723 for (i = 0; i < (long)ROOTINO; i++) {
724 setbit(cg_inosused(&acg), i);
725 acg.cg_cs.cs_nifree--;
729 * In cylno 0, beginning space is reserved
730 * for boot and super blocks.
732 for (d = 0; d < dlower; d += sblock.fs_frag) {
733 blkno = d / sblock.fs_frag;
734 setblock(&sblock, cg_blksfree(&acg), blkno);
735 if (sblock.fs_contigsumsize > 0)
736 setbit(cg_clustersfree(&acg), blkno);
737 acg.cg_cs.cs_nbfree++;
740 if ((i = dupper % sblock.fs_frag)) {
741 acg.cg_frsum[sblock.fs_frag - i]++;
742 for (d = dupper + sblock.fs_frag - i; dupper < d; dupper++) {
743 setbit(cg_blksfree(&acg), dupper);
744 acg.cg_cs.cs_nffree++;
747 for (d = dupper; d + sblock.fs_frag <= acg.cg_ndblk;
748 d += sblock.fs_frag) {
749 blkno = d / sblock.fs_frag;
750 setblock(&sblock, cg_blksfree(&acg), blkno);
751 if (sblock.fs_contigsumsize > 0)
752 setbit(cg_clustersfree(&acg), blkno);
753 acg.cg_cs.cs_nbfree++;
755 if (d < acg.cg_ndblk) {
756 acg.cg_frsum[acg.cg_ndblk - d]++;
757 for (; d < acg.cg_ndblk; d++) {
758 setbit(cg_blksfree(&acg), d);
759 acg.cg_cs.cs_nffree++;
762 if (sblock.fs_contigsumsize > 0) {
763 int32_t *sump = cg_clustersum(&acg);
764 u_char *mapp = cg_clustersfree(&acg);
769 for (i = 0; i < acg.cg_nclusterblks; i++) {
770 if ((map & bit) != 0)
773 if (run > sblock.fs_contigsumsize)
774 run = sblock.fs_contigsumsize;
778 if ((i & (CHAR_BIT - 1)) != CHAR_BIT - 1)
786 if (run > sblock.fs_contigsumsize)
787 run = sblock.fs_contigsumsize;
793 * Write out the duplicate super block, the cylinder group map
794 * and two blocks worth of inodes in a single write.
796 start = sblock.fs_bsize > SBLOCKSIZE ? sblock.fs_bsize : SBLOCKSIZE;
797 bcopy((char *)&acg, &iobuf[start], sblock.fs_cgsize);
798 start += sblock.fs_bsize;
799 dp1 = (struct ufs1_dinode *)(&iobuf[start]);
800 dp2 = (struct ufs2_dinode *)(&iobuf[start]);
801 for (i = 0; i < acg.cg_initediblk; i++) {
802 if (sblock.fs_magic == FS_UFS1_MAGIC) {
803 dp1->di_gen = newfs_random();
806 dp2->di_gen = newfs_random();
810 wtfs(fsbtodb(&sblock, cgsblock(&sblock, cylno)), iobufsize, iobuf);
812 * For the old file system, we have to initialize all the inodes.
815 for (i = 2 * sblock.fs_frag;
816 i < sblock.fs_ipg / INOPF(&sblock);
817 i += sblock.fs_frag) {
818 dp1 = (struct ufs1_dinode *)(&iobuf[start]);
819 for (j = 0; j < INOPB(&sblock); j++) {
820 dp1->di_gen = newfs_random();
823 wtfs(fsbtodb(&sblock, cgimin(&sblock, cylno) + i),
824 sblock.fs_bsize, &iobuf[start]);
830 * initialize the file system
832 #define ROOTLINKCNT 3
834 static struct direct root_dir[] = {
835 { ROOTINO, sizeof(struct direct), DT_DIR, 1, "." },
836 { ROOTINO, sizeof(struct direct), DT_DIR, 2, ".." },
837 { ROOTINO + 1, sizeof(struct direct), DT_DIR, 5, ".snap" },
840 #define SNAPLINKCNT 2
842 static struct direct snap_dir[] = {
843 { ROOTINO + 1, sizeof(struct direct), DT_DIR, 1, "." },
844 { ROOTINO, sizeof(struct direct), DT_DIR, 2, ".." },
855 memset(&node, 0, sizeof node);
856 if ((grp = getgrnam("operator")) != NULL) {
859 warnx("Cannot retrieve operator gid, using gid 0.");
862 entries = (nflag) ? ROOTLINKCNT - 1: ROOTLINKCNT;
863 if (sblock.fs_magic == FS_UFS1_MAGIC) {
865 * initialize the node
867 node.dp1.di_atime = utime;
868 node.dp1.di_mtime = utime;
869 node.dp1.di_ctime = utime;
871 * create the root directory
873 node.dp1.di_mode = IFDIR | UMASK;
874 node.dp1.di_nlink = entries;
875 node.dp1.di_size = makedir(root_dir, entries);
876 node.dp1.di_db[0] = alloc(sblock.fs_fsize, node.dp1.di_mode);
878 btodb(fragroundup(&sblock, node.dp1.di_size));
879 wtfs(fsbtodb(&sblock, node.dp1.di_db[0]), sblock.fs_fsize,
881 iput(&node, ROOTINO);
884 * create the .snap directory
886 node.dp1.di_mode |= 020;
887 node.dp1.di_gid = gid;
888 node.dp1.di_nlink = SNAPLINKCNT;
889 node.dp1.di_size = makedir(snap_dir, SNAPLINKCNT);
891 alloc(sblock.fs_fsize, node.dp1.di_mode);
893 btodb(fragroundup(&sblock, node.dp1.di_size));
894 wtfs(fsbtodb(&sblock, node.dp1.di_db[0]),
895 sblock.fs_fsize, iobuf);
896 iput(&node, ROOTINO + 1);
900 * initialize the node
902 node.dp2.di_atime = utime;
903 node.dp2.di_mtime = utime;
904 node.dp2.di_ctime = utime;
905 node.dp2.di_birthtime = utime;
907 * create the root directory
909 node.dp2.di_mode = IFDIR | UMASK;
910 node.dp2.di_nlink = entries;
911 node.dp2.di_size = makedir(root_dir, entries);
912 node.dp2.di_db[0] = alloc(sblock.fs_fsize, node.dp2.di_mode);
914 btodb(fragroundup(&sblock, node.dp2.di_size));
915 wtfs(fsbtodb(&sblock, node.dp2.di_db[0]), sblock.fs_fsize,
917 iput(&node, ROOTINO);
920 * create the .snap directory
922 node.dp2.di_mode |= 020;
923 node.dp2.di_gid = gid;
924 node.dp2.di_nlink = SNAPLINKCNT;
925 node.dp2.di_size = makedir(snap_dir, SNAPLINKCNT);
927 alloc(sblock.fs_fsize, node.dp2.di_mode);
929 btodb(fragroundup(&sblock, node.dp2.di_size));
930 wtfs(fsbtodb(&sblock, node.dp2.di_db[0]),
931 sblock.fs_fsize, iobuf);
932 iput(&node, ROOTINO + 1);
938 * construct a set of directory entries in "iobuf".
939 * return size of directory.
942 makedir(struct direct *protodir, int entries)
948 memset(iobuf, 0, DIRBLKSIZ);
949 for (cp = iobuf, i = 0; i < entries - 1; i++) {
950 protodir[i].d_reclen = DIRSIZ(0, &protodir[i]);
951 memmove(cp, &protodir[i], protodir[i].d_reclen);
952 cp += protodir[i].d_reclen;
953 spcleft -= protodir[i].d_reclen;
955 protodir[i].d_reclen = spcleft;
956 memmove(cp, &protodir[i], DIRSIZ(0, &protodir[i]));
961 * allocate a block or frag
964 alloc(int size, int mode)
969 bread(&disk, part_ofs + fsbtodb(&sblock, cgtod(&sblock, 0)), (char *)&acg,
971 if (acg.cg_magic != CG_MAGIC) {
972 printf("cg 0: bad magic number\n");
975 if (acg.cg_cs.cs_nbfree == 0) {
976 printf("first cylinder group ran out of space\n");
979 for (d = 0; d < acg.cg_ndblk; d += sblock.fs_frag)
980 if (isblock(&sblock, cg_blksfree(&acg), d / sblock.fs_frag))
982 printf("internal error: can't find block in cyl 0\n");
985 blkno = fragstoblks(&sblock, d);
986 clrblock(&sblock, cg_blksfree(&acg), blkno);
987 if (sblock.fs_contigsumsize > 0)
988 clrbit(cg_clustersfree(&acg), blkno);
989 acg.cg_cs.cs_nbfree--;
990 sblock.fs_cstotal.cs_nbfree--;
994 sblock.fs_cstotal.cs_ndir++;
997 if (size != sblock.fs_bsize) {
998 frag = howmany(size, sblock.fs_fsize);
999 fscs[0].cs_nffree += sblock.fs_frag - frag;
1000 sblock.fs_cstotal.cs_nffree += sblock.fs_frag - frag;
1001 acg.cg_cs.cs_nffree += sblock.fs_frag - frag;
1002 acg.cg_frsum[sblock.fs_frag - frag]++;
1003 for (i = frag; i < sblock.fs_frag; i++)
1004 setbit(cg_blksfree(&acg), d + i);
1006 /* XXX cgwrite(&disk, 0)??? */
1007 wtfs(fsbtodb(&sblock, cgtod(&sblock, 0)), sblock.fs_cgsize,
1009 return ((ufs2_daddr_t)d);
1013 * Allocate an inode on the disk
1016 iput(union dinode *ip, ino_t ino)
1020 bread(&disk, part_ofs + fsbtodb(&sblock, cgtod(&sblock, 0)), (char *)&acg,
1022 if (acg.cg_magic != CG_MAGIC) {
1023 printf("cg 0: bad magic number\n");
1026 acg.cg_cs.cs_nifree--;
1027 setbit(cg_inosused(&acg), ino);
1028 wtfs(fsbtodb(&sblock, cgtod(&sblock, 0)), sblock.fs_cgsize,
1030 sblock.fs_cstotal.cs_nifree--;
1031 fscs[0].cs_nifree--;
1032 if (ino >= (unsigned long)sblock.fs_ipg * sblock.fs_ncg) {
1033 printf("fsinit: inode value out of range (%ju).\n",
1037 d = fsbtodb(&sblock, ino_to_fsba(&sblock, ino));
1038 bread(&disk, part_ofs + d, (char *)iobuf, sblock.fs_bsize);
1039 if (sblock.fs_magic == FS_UFS1_MAGIC)
1040 ((struct ufs1_dinode *)iobuf)[ino_to_fsbo(&sblock, ino)] =
1043 ((struct ufs2_dinode *)iobuf)[ino_to_fsbo(&sblock, ino)] =
1045 wtfs(d, sblock.fs_bsize, (char *)iobuf);
1049 * possibly write to disk
1052 wtfs(ufs2_daddr_t bno, int size, char *bf)
1056 if (bwrite(&disk, part_ofs + bno, bf, size) < 0)
1057 err(36, "wtfs: %d bytes at sector %jd", size, (intmax_t)bno);
1061 * check if a block is available
1064 isblock(struct fs *fs, unsigned char *cp, int h)
1068 switch (fs->fs_frag) {
1070 return (cp[h] == 0xff);
1072 mask = 0x0f << ((h & 0x1) << 2);
1073 return ((cp[h >> 1] & mask) == mask);
1075 mask = 0x03 << ((h & 0x3) << 1);
1076 return ((cp[h >> 2] & mask) == mask);
1078 mask = 0x01 << (h & 0x7);
1079 return ((cp[h >> 3] & mask) == mask);
1081 fprintf(stderr, "isblock bad fs_frag %d\n", fs->fs_frag);
1087 * take a block out of the map
1090 clrblock(struct fs *fs, unsigned char *cp, int h)
1092 switch ((fs)->fs_frag) {
1097 cp[h >> 1] &= ~(0x0f << ((h & 0x1) << 2));
1100 cp[h >> 2] &= ~(0x03 << ((h & 0x3) << 1));
1103 cp[h >> 3] &= ~(0x01 << (h & 0x7));
1106 fprintf(stderr, "clrblock bad fs_frag %d\n", fs->fs_frag);
1112 * put a block into the map
1115 setblock(struct fs *fs, unsigned char *cp, int h)
1117 switch (fs->fs_frag) {
1122 cp[h >> 1] |= (0x0f << ((h & 0x1) << 2));
1125 cp[h >> 2] |= (0x03 << ((h & 0x3) << 1));
1128 cp[h >> 3] |= (0x01 << (h & 0x7));
1131 fprintf(stderr, "setblock bad fs_frag %d\n", fs->fs_frag);
1137 * Determine the number of characters in a
1149 if (ioctl(0, TIOCGWINSZ, &ws) != -1)
1150 columns = ws.ws_col;
1151 if (columns == 0 && (cp = getenv("COLUMNS")))
1154 columns = 80; /* last resort */
1163 for (n = 0; n < sizeof(n) * CHAR_BIT; n++)
1166 errx(1, "ilog2: %d is not a power of 2\n", val);
1170 * For the regression test, return predictable random values.
1171 * Otherwise use a true random number generator.
1176 static int nextnum = 1;
1180 return (arc4random());