mkfs.c revision 201399
1/*
2 * Copyright (c) 2002 Networks Associates Technology, Inc.
3 * All rights reserved.
4 *
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
9 * research program.
10 *
11 * Copyright (c) 1980, 1989, 1993
12 *	The Regents of the University of California.  All rights reserved.
13 *
14 * Redistribution and use in source and binary forms, with or without
15 * modification, are permitted provided that the following conditions
16 * are met:
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.
25 *
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
36 * SUCH DAMAGE.
37 */
38
39#if 0
40#ifndef lint
41static char sccsid[] = "@(#)mkfs.c	8.11 (Berkeley) 5/3/95";
42#endif /* not lint */
43#endif
44#include <sys/cdefs.h>
45__FBSDID("$FreeBSD: head/sbin/newfs/mkfs.c 201399 2010-01-02 17:32:40Z mbr $");
46
47#include <err.h>
48#include <grp.h>
49#include <limits.h>
50#include <signal.h>
51#include <stdlib.h>
52#include <string.h>
53#include <stdint.h>
54#include <stdio.h>
55#include <unistd.h>
56#include <sys/param.h>
57#include <sys/time.h>
58#include <sys/types.h>
59#include <sys/wait.h>
60#include <sys/resource.h>
61#include <sys/stat.h>
62#include <ufs/ufs/dinode.h>
63#include <ufs/ufs/dir.h>
64#include <ufs/ffs/fs.h>
65#include <sys/disklabel.h>
66#include <sys/file.h>
67#include <sys/mman.h>
68#include <sys/ioctl.h>
69#include "newfs.h"
70
71/*
72 * make file system for cylinder-group style file systems
73 */
74#define UMASK		0755
75#define POWEROF2(num)	(((num) & ((num) - 1)) == 0)
76
77static struct	csum *fscs;
78#define	sblock	disk.d_fs
79#define	acg	disk.d_cg
80
81union dinode {
82	struct ufs1_dinode dp1;
83	struct ufs2_dinode dp2;
84};
85#define DIP(dp, field) \
86	((sblock.fs_magic == FS_UFS1_MAGIC) ? \
87	(dp)->dp1.field : (dp)->dp2.field)
88
89static caddr_t iobuf;
90static long iobufsize;
91static ufs2_daddr_t alloc(int size, int mode);
92static int charsperline(void);
93static void clrblock(struct fs *, unsigned char *, int);
94static void fsinit(time_t);
95static int ilog2(int);
96static void initcg(int, time_t);
97static int isblock(struct fs *, unsigned char *, int);
98static void iput(union dinode *, ino_t);
99static int makedir(struct direct *, int);
100static void setblock(struct fs *, unsigned char *, int);
101static void wtfs(ufs2_daddr_t, int, char *);
102static u_int32_t newfs_random(void);
103
104static int
105do_sbwrite(struct uufsd *disk)
106{
107	if (!disk->d_sblock)
108		disk->d_sblock = disk->d_fs.fs_sblockloc / disk->d_bsize;
109	return (pwrite(disk->d_fd, &disk->d_fs, SBLOCKSIZE, (off_t)((part_ofs +
110	    disk->d_sblock) * disk->d_bsize)));
111}
112
113void
114mkfs(struct partition *pp, char *fsys)
115{
116	int fragsperinode, optimalfpg, origdensity, minfpg, lastminfpg;
117	long i, j, cylno, csfrags;
118	time_t utime;
119	quad_t sizepb;
120	int width;
121	char tmpbuf[100];	/* XXX this will break in about 2,500 years */
122	union {
123		struct fs fdummy;
124		char cdummy[SBLOCKSIZE];
125	} dummy;
126#define fsdummy dummy.fdummy
127#define chdummy dummy.cdummy
128
129	/*
130	 * Our blocks == sector size, and the version of UFS we are using is
131	 * specified by Oflag.
132	 */
133	disk.d_bsize = sectorsize;
134	disk.d_ufs = Oflag;
135	if (Rflag) {
136		utime = 1000000000;
137	} else {
138		time(&utime);
139		arc4random_stir();
140	}
141	sblock.fs_old_flags = FS_FLAGS_UPDATED;
142	sblock.fs_flags = 0;
143	if (Uflag)
144		sblock.fs_flags |= FS_DOSOFTDEP;
145	if (Lflag)
146		strlcpy(sblock.fs_volname, volumelabel, MAXVOLLEN);
147	if (Jflag)
148		sblock.fs_flags |= FS_GJOURNAL;
149	if (lflag)
150		sblock.fs_flags |= FS_MULTILABEL;
151	/*
152	 * Validate the given file system size.
153	 * Verify that its last block can actually be accessed.
154	 * Convert to file system fragment sized units.
155	 */
156	if (fssize <= 0) {
157		printf("preposterous size %jd\n", (intmax_t)fssize);
158		exit(13);
159	}
160	wtfs(fssize - (realsectorsize / DEV_BSIZE), realsectorsize,
161	    (char *)&sblock);
162	/*
163	 * collect and verify the file system density info
164	 */
165	sblock.fs_avgfilesize = avgfilesize;
166	sblock.fs_avgfpdir = avgfilesperdir;
167	if (sblock.fs_avgfilesize <= 0)
168		printf("illegal expected average file size %d\n",
169		    sblock.fs_avgfilesize), exit(14);
170	if (sblock.fs_avgfpdir <= 0)
171		printf("illegal expected number of files per directory %d\n",
172		    sblock.fs_avgfpdir), exit(15);
173	/*
174	 * collect and verify the block and fragment sizes
175	 */
176	sblock.fs_bsize = bsize;
177	sblock.fs_fsize = fsize;
178	if (!POWEROF2(sblock.fs_bsize)) {
179		printf("block size must be a power of 2, not %d\n",
180		    sblock.fs_bsize);
181		exit(16);
182	}
183	if (!POWEROF2(sblock.fs_fsize)) {
184		printf("fragment size must be a power of 2, not %d\n",
185		    sblock.fs_fsize);
186		exit(17);
187	}
188	if (sblock.fs_fsize < sectorsize) {
189		printf("increasing fragment size from %d to sector size (%d)\n",
190		    sblock.fs_fsize, sectorsize);
191		sblock.fs_fsize = sectorsize;
192	}
193	if (sblock.fs_bsize > MAXBSIZE) {
194		printf("decreasing block size from %d to maximum (%d)\n",
195		    sblock.fs_bsize, MAXBSIZE);
196		sblock.fs_bsize = MAXBSIZE;
197	}
198	if (sblock.fs_bsize < MINBSIZE) {
199		printf("increasing block size from %d to minimum (%d)\n",
200		    sblock.fs_bsize, MINBSIZE);
201		sblock.fs_bsize = MINBSIZE;
202	}
203	if (sblock.fs_fsize > MAXBSIZE) {
204		printf("decreasing fragment size from %d to maximum (%d)\n",
205		    sblock.fs_fsize, MAXBSIZE);
206		sblock.fs_fsize = MAXBSIZE;
207	}
208	if (sblock.fs_bsize < sblock.fs_fsize) {
209		printf("increasing block size from %d to fragment size (%d)\n",
210		    sblock.fs_bsize, sblock.fs_fsize);
211		sblock.fs_bsize = sblock.fs_fsize;
212	}
213	if (sblock.fs_fsize * MAXFRAG < sblock.fs_bsize) {
214		printf(
215		"increasing fragment size from %d to block size / %d (%d)\n",
216		    sblock.fs_fsize, MAXFRAG, sblock.fs_bsize / MAXFRAG);
217		sblock.fs_fsize = sblock.fs_bsize / MAXFRAG;
218	}
219	if (maxbsize < bsize || !POWEROF2(maxbsize)) {
220		sblock.fs_maxbsize = sblock.fs_bsize;
221		printf("Extent size set to %d\n", sblock.fs_maxbsize);
222	} else if (sblock.fs_maxbsize > FS_MAXCONTIG * sblock.fs_bsize) {
223		sblock.fs_maxbsize = FS_MAXCONTIG * sblock.fs_bsize;
224		printf("Extent size reduced to %d\n", sblock.fs_maxbsize);
225	} else {
226		sblock.fs_maxbsize = maxbsize;
227	}
228	sblock.fs_maxcontig = maxcontig;
229	if (sblock.fs_maxcontig < sblock.fs_maxbsize / sblock.fs_bsize) {
230		sblock.fs_maxcontig = sblock.fs_maxbsize / sblock.fs_bsize;
231		printf("Maxcontig raised to %d\n", sblock.fs_maxbsize);
232	}
233	if (sblock.fs_maxcontig > 1)
234		sblock.fs_contigsumsize = MIN(sblock.fs_maxcontig,FS_MAXCONTIG);
235	sblock.fs_bmask = ~(sblock.fs_bsize - 1);
236	sblock.fs_fmask = ~(sblock.fs_fsize - 1);
237	sblock.fs_qbmask = ~sblock.fs_bmask;
238	sblock.fs_qfmask = ~sblock.fs_fmask;
239	sblock.fs_bshift = ilog2(sblock.fs_bsize);
240	sblock.fs_fshift = ilog2(sblock.fs_fsize);
241	sblock.fs_frag = numfrags(&sblock, sblock.fs_bsize);
242	sblock.fs_fragshift = ilog2(sblock.fs_frag);
243	if (sblock.fs_frag > MAXFRAG) {
244		printf("fragment size %d is still too small (can't happen)\n",
245		    sblock.fs_bsize / MAXFRAG);
246		exit(21);
247	}
248	sblock.fs_fsbtodb = ilog2(sblock.fs_fsize / sectorsize);
249	sblock.fs_size = fssize = dbtofsb(&sblock, fssize);
250
251	/*
252	 * Before the filesystem is finally initialized, mark it
253	 * as incompletely initialized.
254	 */
255	sblock.fs_magic = FS_BAD_MAGIC;
256
257	if (Oflag == 1) {
258		sblock.fs_sblockloc = SBLOCK_UFS1;
259		sblock.fs_nindir = sblock.fs_bsize / sizeof(ufs1_daddr_t);
260		sblock.fs_inopb = sblock.fs_bsize / sizeof(struct ufs1_dinode);
261		sblock.fs_maxsymlinklen = ((NDADDR + NIADDR) *
262		    sizeof(ufs1_daddr_t));
263		sblock.fs_old_inodefmt = FS_44INODEFMT;
264		sblock.fs_old_cgoffset = 0;
265		sblock.fs_old_cgmask = 0xffffffff;
266		sblock.fs_old_size = sblock.fs_size;
267		sblock.fs_old_rotdelay = 0;
268		sblock.fs_old_rps = 60;
269		sblock.fs_old_nspf = sblock.fs_fsize / sectorsize;
270		sblock.fs_old_cpg = 1;
271		sblock.fs_old_interleave = 1;
272		sblock.fs_old_trackskew = 0;
273		sblock.fs_old_cpc = 0;
274		sblock.fs_old_postblformat = 1;
275		sblock.fs_old_nrpos = 1;
276	} else {
277		sblock.fs_sblockloc = SBLOCK_UFS2;
278		sblock.fs_nindir = sblock.fs_bsize / sizeof(ufs2_daddr_t);
279		sblock.fs_inopb = sblock.fs_bsize / sizeof(struct ufs2_dinode);
280		sblock.fs_maxsymlinklen = ((NDADDR + NIADDR) *
281		    sizeof(ufs2_daddr_t));
282	}
283	sblock.fs_sblkno =
284	    roundup(howmany(sblock.fs_sblockloc + SBLOCKSIZE, sblock.fs_fsize),
285		sblock.fs_frag);
286	sblock.fs_cblkno = sblock.fs_sblkno +
287	    roundup(howmany(SBLOCKSIZE, sblock.fs_fsize), sblock.fs_frag);
288	sblock.fs_iblkno = sblock.fs_cblkno + sblock.fs_frag;
289	sblock.fs_maxfilesize = sblock.fs_bsize * NDADDR - 1;
290	for (sizepb = sblock.fs_bsize, i = 0; i < NIADDR; i++) {
291		sizepb *= NINDIR(&sblock);
292		sblock.fs_maxfilesize += sizepb;
293	}
294
295	/*
296	 * It's impossible to create a snapshot in case that fs_maxfilesize
297	 * is smaller than the fssize.
298	 */
299	if (sblock.fs_maxfilesize < (u_quad_t)fssize) {
300		warnx("WARNING: You will be unable to create snapshots on this "
301		      "file system.  Correct by using a larger blocksize.");
302	}
303
304	/*
305	 * Calculate the number of blocks to put into each cylinder group.
306	 *
307	 * This algorithm selects the number of blocks per cylinder
308	 * group. The first goal is to have at least enough data blocks
309	 * in each cylinder group to meet the density requirement. Once
310	 * this goal is achieved we try to expand to have at least
311	 * MINCYLGRPS cylinder groups. Once this goal is achieved, we
312	 * pack as many blocks into each cylinder group map as will fit.
313	 *
314	 * We start by calculating the smallest number of blocks that we
315	 * can put into each cylinder group. If this is too big, we reduce
316	 * the density until it fits.
317	 */
318	origdensity = density;
319	for (;;) {
320		fragsperinode = MAX(numfrags(&sblock, density), 1);
321		minfpg = fragsperinode * INOPB(&sblock);
322		if (minfpg > sblock.fs_size)
323			minfpg = sblock.fs_size;
324		sblock.fs_ipg = INOPB(&sblock);
325		sblock.fs_fpg = roundup(sblock.fs_iblkno +
326		    sblock.fs_ipg / INOPF(&sblock), sblock.fs_frag);
327		if (sblock.fs_fpg < minfpg)
328			sblock.fs_fpg = minfpg;
329		sblock.fs_ipg = roundup(howmany(sblock.fs_fpg, fragsperinode),
330		    INOPB(&sblock));
331		sblock.fs_fpg = roundup(sblock.fs_iblkno +
332		    sblock.fs_ipg / INOPF(&sblock), sblock.fs_frag);
333		if (sblock.fs_fpg < minfpg)
334			sblock.fs_fpg = minfpg;
335		sblock.fs_ipg = roundup(howmany(sblock.fs_fpg, fragsperinode),
336		    INOPB(&sblock));
337		if (CGSIZE(&sblock) < (unsigned long)sblock.fs_bsize)
338			break;
339		density -= sblock.fs_fsize;
340	}
341	if (density != origdensity)
342		printf("density reduced from %d to %d\n", origdensity, density);
343	/*
344	 * Start packing more blocks into the cylinder group until
345	 * it cannot grow any larger, the number of cylinder groups
346	 * drops below MINCYLGRPS, or we reach the size requested.
347	 */
348	for ( ; sblock.fs_fpg < maxblkspercg; sblock.fs_fpg += sblock.fs_frag) {
349		sblock.fs_ipg = roundup(howmany(sblock.fs_fpg, fragsperinode),
350		    INOPB(&sblock));
351		if (sblock.fs_size / sblock.fs_fpg < MINCYLGRPS)
352			break;
353		if (CGSIZE(&sblock) < (unsigned long)sblock.fs_bsize)
354			continue;
355		if (CGSIZE(&sblock) == (unsigned long)sblock.fs_bsize)
356			break;
357		sblock.fs_fpg -= sblock.fs_frag;
358		sblock.fs_ipg = roundup(howmany(sblock.fs_fpg, fragsperinode),
359		    INOPB(&sblock));
360		break;
361	}
362	/*
363	 * Check to be sure that the last cylinder group has enough blocks
364	 * to be viable. If it is too small, reduce the number of blocks
365	 * per cylinder group which will have the effect of moving more
366	 * blocks into the last cylinder group.
367	 */
368	optimalfpg = sblock.fs_fpg;
369	for (;;) {
370		sblock.fs_ncg = howmany(sblock.fs_size, sblock.fs_fpg);
371		lastminfpg = roundup(sblock.fs_iblkno +
372		    sblock.fs_ipg / INOPF(&sblock), sblock.fs_frag);
373		if (sblock.fs_size < lastminfpg) {
374			printf("Filesystem size %jd < minimum size of %d\n",
375			    (intmax_t)sblock.fs_size, lastminfpg);
376			exit(28);
377		}
378		if (sblock.fs_size % sblock.fs_fpg >= lastminfpg ||
379		    sblock.fs_size % sblock.fs_fpg == 0)
380			break;
381		sblock.fs_fpg -= sblock.fs_frag;
382		sblock.fs_ipg = roundup(howmany(sblock.fs_fpg, fragsperinode),
383		    INOPB(&sblock));
384	}
385	if (optimalfpg != sblock.fs_fpg)
386		printf("Reduced frags per cylinder group from %d to %d %s\n",
387		   optimalfpg, sblock.fs_fpg, "to enlarge last cyl group");
388	sblock.fs_cgsize = fragroundup(&sblock, CGSIZE(&sblock));
389	sblock.fs_dblkno = sblock.fs_iblkno + sblock.fs_ipg / INOPF(&sblock);
390	if (Oflag == 1) {
391		sblock.fs_old_spc = sblock.fs_fpg * sblock.fs_old_nspf;
392		sblock.fs_old_nsect = sblock.fs_old_spc;
393		sblock.fs_old_npsect = sblock.fs_old_spc;
394		sblock.fs_old_ncyl = sblock.fs_ncg;
395	}
396	/*
397	 * fill in remaining fields of the super block
398	 */
399	sblock.fs_csaddr = cgdmin(&sblock, 0);
400	sblock.fs_cssize =
401	    fragroundup(&sblock, sblock.fs_ncg * sizeof(struct csum));
402	fscs = (struct csum *)calloc(1, sblock.fs_cssize);
403	if (fscs == NULL)
404		errx(31, "calloc failed");
405	sblock.fs_sbsize = fragroundup(&sblock, sizeof(struct fs));
406	if (sblock.fs_sbsize > SBLOCKSIZE)
407		sblock.fs_sbsize = SBLOCKSIZE;
408	sblock.fs_minfree = minfree;
409	sblock.fs_maxbpg = maxbpg;
410	sblock.fs_optim = opt;
411	sblock.fs_cgrotor = 0;
412	sblock.fs_pendingblocks = 0;
413	sblock.fs_pendinginodes = 0;
414	sblock.fs_fmod = 0;
415	sblock.fs_ronly = 0;
416	sblock.fs_state = 0;
417	sblock.fs_clean = 1;
418	sblock.fs_id[0] = (long)utime;
419	sblock.fs_id[1] = newfs_random();
420	sblock.fs_fsmnt[0] = '\0';
421	csfrags = howmany(sblock.fs_cssize, sblock.fs_fsize);
422	sblock.fs_dsize = sblock.fs_size - sblock.fs_sblkno -
423	    sblock.fs_ncg * (sblock.fs_dblkno - sblock.fs_sblkno);
424	sblock.fs_cstotal.cs_nbfree =
425	    fragstoblks(&sblock, sblock.fs_dsize) -
426	    howmany(csfrags, sblock.fs_frag);
427	sblock.fs_cstotal.cs_nffree =
428	    fragnum(&sblock, sblock.fs_size) +
429	    (fragnum(&sblock, csfrags) > 0 ?
430	     sblock.fs_frag - fragnum(&sblock, csfrags) : 0);
431	sblock.fs_cstotal.cs_nifree = sblock.fs_ncg * sblock.fs_ipg - ROOTINO;
432	sblock.fs_cstotal.cs_ndir = 0;
433	sblock.fs_dsize -= csfrags;
434	sblock.fs_time = utime;
435	if (Oflag == 1) {
436		sblock.fs_old_time = utime;
437		sblock.fs_old_dsize = sblock.fs_dsize;
438		sblock.fs_old_csaddr = sblock.fs_csaddr;
439		sblock.fs_old_cstotal.cs_ndir = sblock.fs_cstotal.cs_ndir;
440		sblock.fs_old_cstotal.cs_nbfree = sblock.fs_cstotal.cs_nbfree;
441		sblock.fs_old_cstotal.cs_nifree = sblock.fs_cstotal.cs_nifree;
442		sblock.fs_old_cstotal.cs_nffree = sblock.fs_cstotal.cs_nffree;
443	}
444
445	/*
446	 * Dump out summary information about file system.
447	 */
448#	define B2MBFACTOR (1 / (1024.0 * 1024.0))
449	printf("%s: %.1fMB (%jd sectors) block size %d, fragment size %d\n",
450	    fsys, (float)sblock.fs_size * sblock.fs_fsize * B2MBFACTOR,
451	    (intmax_t)fsbtodb(&sblock, sblock.fs_size), sblock.fs_bsize,
452	    sblock.fs_fsize);
453	printf("\tusing %d cylinder groups of %.2fMB, %d blks, %d inodes.\n",
454	    sblock.fs_ncg, (float)sblock.fs_fpg * sblock.fs_fsize * B2MBFACTOR,
455	    sblock.fs_fpg / sblock.fs_frag, sblock.fs_ipg);
456	if (sblock.fs_flags & FS_DOSOFTDEP)
457		printf("\twith soft updates\n");
458#	undef B2MBFACTOR
459
460	if (Eflag && !Nflag) {
461		printf("Erasing sectors [%jd...%jd]\n",
462		    sblock.fs_sblockloc / disk.d_bsize,
463		    fsbtodb(&sblock, sblock.fs_size) - 1);
464		berase(&disk, sblock.fs_sblockloc / disk.d_bsize,
465		    sblock.fs_size * sblock.fs_fsize - sblock.fs_sblockloc);
466	}
467	/*
468	 * Wipe out old UFS1 superblock(s) if necessary.
469	 */
470	if (!Nflag && Oflag != 1) {
471		i = bread(&disk, part_ofs + SBLOCK_UFS1 / disk.d_bsize, chdummy, SBLOCKSIZE);
472		if (i == -1)
473			err(1, "can't read old UFS1 superblock: %s", disk.d_error);
474
475		if (fsdummy.fs_magic == FS_UFS1_MAGIC) {
476			fsdummy.fs_magic = 0;
477			bwrite(&disk, part_ofs + SBLOCK_UFS1 / disk.d_bsize,
478			    chdummy, SBLOCKSIZE);
479			for (i = 0; i < fsdummy.fs_ncg; i++)
480				bwrite(&disk, part_ofs + fsbtodb(&fsdummy,
481				    cgsblock(&fsdummy, i)), chdummy, SBLOCKSIZE);
482		}
483	}
484	if (!Nflag)
485		do_sbwrite(&disk);
486	if (Xflag == 1) {
487		printf("** Exiting on Xflag 1\n");
488		exit(0);
489	}
490	if (Xflag == 2)
491		printf("** Leaving BAD MAGIC on Xflag 2\n");
492	else
493		sblock.fs_magic = (Oflag != 1) ? FS_UFS2_MAGIC : FS_UFS1_MAGIC;
494
495	/*
496	 * Now build the cylinders group blocks and
497	 * then print out indices of cylinder groups.
498	 */
499	printf("super-block backups (for fsck -b #) at:\n");
500	i = 0;
501	width = charsperline();
502	/*
503	 * allocate space for superblock, cylinder group map, and
504	 * two sets of inode blocks.
505	 */
506	if (sblock.fs_bsize < SBLOCKSIZE)
507		iobufsize = SBLOCKSIZE + 3 * sblock.fs_bsize;
508	else
509		iobufsize = 4 * sblock.fs_bsize;
510	if ((iobuf = calloc(1, iobufsize)) == 0) {
511		printf("Cannot allocate I/O buffer\n");
512		exit(38);
513	}
514	/*
515	 * Make a copy of the superblock into the buffer that we will be
516	 * writing out in each cylinder group.
517	 */
518	bcopy((char *)&sblock, iobuf, SBLOCKSIZE);
519	for (cylno = 0; cylno < sblock.fs_ncg; cylno++) {
520		initcg(cylno, utime);
521		j = snprintf(tmpbuf, sizeof(tmpbuf), " %jd%s",
522		    (intmax_t)fsbtodb(&sblock, cgsblock(&sblock, cylno)),
523		    cylno < (sblock.fs_ncg-1) ? "," : "");
524		if (j < 0)
525			tmpbuf[j = 0] = '\0';
526		if (i + j >= width) {
527			printf("\n");
528			i = 0;
529		}
530		i += j;
531		printf("%s", tmpbuf);
532		fflush(stdout);
533	}
534	printf("\n");
535	if (Nflag)
536		exit(0);
537	/*
538	 * Now construct the initial file system,
539	 * then write out the super-block.
540	 */
541	fsinit(utime);
542	if (Oflag == 1) {
543		sblock.fs_old_cstotal.cs_ndir = sblock.fs_cstotal.cs_ndir;
544		sblock.fs_old_cstotal.cs_nbfree = sblock.fs_cstotal.cs_nbfree;
545		sblock.fs_old_cstotal.cs_nifree = sblock.fs_cstotal.cs_nifree;
546		sblock.fs_old_cstotal.cs_nffree = sblock.fs_cstotal.cs_nffree;
547	}
548	if (Xflag == 3) {
549		printf("** Exiting on Xflag 3\n");
550		exit(0);
551	}
552	if (!Nflag)
553		do_sbwrite(&disk);
554	for (i = 0; i < sblock.fs_cssize; i += sblock.fs_bsize)
555		wtfs(fsbtodb(&sblock, sblock.fs_csaddr + numfrags(&sblock, i)),
556			sblock.fs_cssize - i < sblock.fs_bsize ?
557			sblock.fs_cssize - i : sblock.fs_bsize,
558			((char *)fscs) + i);
559	/*
560	 * Update information about this partition in pack
561	 * label, to that it may be updated on disk.
562	 */
563	if (pp != NULL) {
564		pp->p_fstype = FS_BSDFFS;
565		pp->p_fsize = sblock.fs_fsize;
566		pp->p_frag = sblock.fs_frag;
567		pp->p_cpg = sblock.fs_fpg;
568	}
569}
570
571/*
572 * Initialize a cylinder group.
573 */
574void
575initcg(int cylno, time_t utime)
576{
577	long i, j, d, dlower, dupper, blkno, start;
578	ufs2_daddr_t cbase, dmax;
579	struct ufs1_dinode *dp1;
580	struct ufs2_dinode *dp2;
581	struct csum *cs;
582
583	/*
584	 * Determine block bounds for cylinder group.
585	 * Allow space for super block summary information in first
586	 * cylinder group.
587	 */
588	cbase = cgbase(&sblock, cylno);
589	dmax = cbase + sblock.fs_fpg;
590	if (dmax > sblock.fs_size)
591		dmax = sblock.fs_size;
592	dlower = cgsblock(&sblock, cylno) - cbase;
593	dupper = cgdmin(&sblock, cylno) - cbase;
594	if (cylno == 0)
595		dupper += howmany(sblock.fs_cssize, sblock.fs_fsize);
596	cs = &fscs[cylno];
597	memset(&acg, 0, sblock.fs_cgsize);
598	acg.cg_time = utime;
599	acg.cg_magic = CG_MAGIC;
600	acg.cg_cgx = cylno;
601	acg.cg_niblk = sblock.fs_ipg;
602	acg.cg_initediblk = sblock.fs_ipg < 2 * INOPB(&sblock) ?
603	    sblock.fs_ipg : 2 * INOPB(&sblock);
604	acg.cg_ndblk = dmax - cbase;
605	if (sblock.fs_contigsumsize > 0)
606		acg.cg_nclusterblks = acg.cg_ndblk / sblock.fs_frag;
607	start = &acg.cg_space[0] - (u_char *)(&acg.cg_firstfield);
608	if (Oflag == 2) {
609		acg.cg_iusedoff = start;
610	} else {
611		acg.cg_old_ncyl = sblock.fs_old_cpg;
612		acg.cg_old_time = acg.cg_time;
613		acg.cg_time = 0;
614		acg.cg_old_niblk = acg.cg_niblk;
615		acg.cg_niblk = 0;
616		acg.cg_initediblk = 0;
617		acg.cg_old_btotoff = start;
618		acg.cg_old_boff = acg.cg_old_btotoff +
619		    sblock.fs_old_cpg * sizeof(int32_t);
620		acg.cg_iusedoff = acg.cg_old_boff +
621		    sblock.fs_old_cpg * sizeof(u_int16_t);
622	}
623	acg.cg_freeoff = acg.cg_iusedoff + howmany(sblock.fs_ipg, CHAR_BIT);
624	acg.cg_nextfreeoff = acg.cg_freeoff + howmany(sblock.fs_fpg, CHAR_BIT);
625	if (sblock.fs_contigsumsize > 0) {
626		acg.cg_clustersumoff =
627		    roundup(acg.cg_nextfreeoff, sizeof(u_int32_t));
628		acg.cg_clustersumoff -= sizeof(u_int32_t);
629		acg.cg_clusteroff = acg.cg_clustersumoff +
630		    (sblock.fs_contigsumsize + 1) * sizeof(u_int32_t);
631		acg.cg_nextfreeoff = acg.cg_clusteroff +
632		    howmany(fragstoblks(&sblock, sblock.fs_fpg), CHAR_BIT);
633	}
634	if (acg.cg_nextfreeoff > sblock.fs_cgsize) {
635		printf("Panic: cylinder group too big\n");
636		exit(37);
637	}
638	acg.cg_cs.cs_nifree += sblock.fs_ipg;
639	if (cylno == 0)
640		for (i = 0; i < (long)ROOTINO; i++) {
641			setbit(cg_inosused(&acg), i);
642			acg.cg_cs.cs_nifree--;
643		}
644	if (cylno > 0) {
645		/*
646		 * In cylno 0, beginning space is reserved
647		 * for boot and super blocks.
648		 */
649		for (d = 0; d < dlower; d += sblock.fs_frag) {
650			blkno = d / sblock.fs_frag;
651			setblock(&sblock, cg_blksfree(&acg), blkno);
652			if (sblock.fs_contigsumsize > 0)
653				setbit(cg_clustersfree(&acg), blkno);
654			acg.cg_cs.cs_nbfree++;
655		}
656	}
657	if ((i = dupper % sblock.fs_frag)) {
658		acg.cg_frsum[sblock.fs_frag - i]++;
659		for (d = dupper + sblock.fs_frag - i; dupper < d; dupper++) {
660			setbit(cg_blksfree(&acg), dupper);
661			acg.cg_cs.cs_nffree++;
662		}
663	}
664	for (d = dupper; d + sblock.fs_frag <= acg.cg_ndblk;
665	     d += sblock.fs_frag) {
666		blkno = d / sblock.fs_frag;
667		setblock(&sblock, cg_blksfree(&acg), blkno);
668		if (sblock.fs_contigsumsize > 0)
669			setbit(cg_clustersfree(&acg), blkno);
670		acg.cg_cs.cs_nbfree++;
671	}
672	if (d < acg.cg_ndblk) {
673		acg.cg_frsum[acg.cg_ndblk - d]++;
674		for (; d < acg.cg_ndblk; d++) {
675			setbit(cg_blksfree(&acg), d);
676			acg.cg_cs.cs_nffree++;
677		}
678	}
679	if (sblock.fs_contigsumsize > 0) {
680		int32_t *sump = cg_clustersum(&acg);
681		u_char *mapp = cg_clustersfree(&acg);
682		int map = *mapp++;
683		int bit = 1;
684		int run = 0;
685
686		for (i = 0; i < acg.cg_nclusterblks; i++) {
687			if ((map & bit) != 0)
688				run++;
689			else if (run != 0) {
690				if (run > sblock.fs_contigsumsize)
691					run = sblock.fs_contigsumsize;
692				sump[run]++;
693				run = 0;
694			}
695			if ((i & (CHAR_BIT - 1)) != CHAR_BIT - 1)
696				bit <<= 1;
697			else {
698				map = *mapp++;
699				bit = 1;
700			}
701		}
702		if (run != 0) {
703			if (run > sblock.fs_contigsumsize)
704				run = sblock.fs_contigsumsize;
705			sump[run]++;
706		}
707	}
708	*cs = acg.cg_cs;
709	/*
710	 * Write out the duplicate super block, the cylinder group map
711	 * and two blocks worth of inodes in a single write.
712	 */
713	start = sblock.fs_bsize > SBLOCKSIZE ? sblock.fs_bsize : SBLOCKSIZE;
714	bcopy((char *)&acg, &iobuf[start], sblock.fs_cgsize);
715	start += sblock.fs_bsize;
716	dp1 = (struct ufs1_dinode *)(&iobuf[start]);
717	dp2 = (struct ufs2_dinode *)(&iobuf[start]);
718	for (i = 0; i < acg.cg_initediblk; i++) {
719		if (sblock.fs_magic == FS_UFS1_MAGIC) {
720			dp1->di_gen = newfs_random();
721			dp1++;
722		} else {
723			dp2->di_gen = newfs_random();
724			dp2++;
725		}
726	}
727	wtfs(fsbtodb(&sblock, cgsblock(&sblock, cylno)), iobufsize, iobuf);
728	/*
729	 * For the old file system, we have to initialize all the inodes.
730	 */
731	if (Oflag == 1) {
732		for (i = 2 * sblock.fs_frag;
733		     i < sblock.fs_ipg / INOPF(&sblock);
734		     i += sblock.fs_frag) {
735			dp1 = (struct ufs1_dinode *)(&iobuf[start]);
736			for (j = 0; j < INOPB(&sblock); j++) {
737				dp1->di_gen = newfs_random();
738				dp1++;
739			}
740			wtfs(fsbtodb(&sblock, cgimin(&sblock, cylno) + i),
741			    sblock.fs_bsize, &iobuf[start]);
742		}
743	}
744}
745
746/*
747 * initialize the file system
748 */
749#define ROOTLINKCNT 3
750
751struct direct root_dir[] = {
752	{ ROOTINO, sizeof(struct direct), DT_DIR, 1, "." },
753	{ ROOTINO, sizeof(struct direct), DT_DIR, 2, ".." },
754	{ ROOTINO + 1, sizeof(struct direct), DT_DIR, 5, ".snap" },
755};
756
757#define SNAPLINKCNT 2
758
759struct direct snap_dir[] = {
760	{ ROOTINO + 1, sizeof(struct direct), DT_DIR, 1, "." },
761	{ ROOTINO, sizeof(struct direct), DT_DIR, 2, ".." },
762};
763
764void
765fsinit(time_t utime)
766{
767	union dinode node;
768	struct group *grp;
769	gid_t gid;
770	int entries;
771
772	memset(&node, 0, sizeof node);
773	if ((grp = getgrnam("operator")) != NULL) {
774		gid = grp->gr_gid;
775	} else {
776		warnx("Cannot retrieve operator gid, using gid 0.");
777		gid = 0;
778	}
779	entries = (nflag) ? ROOTLINKCNT - 1: ROOTLINKCNT;
780	if (sblock.fs_magic == FS_UFS1_MAGIC) {
781		/*
782		 * initialize the node
783		 */
784		node.dp1.di_atime = utime;
785		node.dp1.di_mtime = utime;
786		node.dp1.di_ctime = utime;
787		/*
788		 * create the root directory
789		 */
790		node.dp1.di_mode = IFDIR | UMASK;
791		node.dp1.di_nlink = entries;
792		node.dp1.di_size = makedir(root_dir, entries);
793		node.dp1.di_db[0] = alloc(sblock.fs_fsize, node.dp1.di_mode);
794		node.dp1.di_blocks =
795		    btodb(fragroundup(&sblock, node.dp1.di_size));
796		wtfs(fsbtodb(&sblock, node.dp1.di_db[0]), sblock.fs_fsize,
797		    iobuf);
798		iput(&node, ROOTINO);
799		if (!nflag) {
800			/*
801			 * create the .snap directory
802			 */
803			node.dp1.di_mode |= 020;
804			node.dp1.di_gid = gid;
805			node.dp1.di_nlink = SNAPLINKCNT;
806			node.dp1.di_size = makedir(snap_dir, SNAPLINKCNT);
807				node.dp1.di_db[0] =
808				    alloc(sblock.fs_fsize, node.dp1.di_mode);
809			node.dp1.di_blocks =
810			    btodb(fragroundup(&sblock, node.dp1.di_size));
811				wtfs(fsbtodb(&sblock, node.dp1.di_db[0]),
812				    sblock.fs_fsize, iobuf);
813			iput(&node, ROOTINO + 1);
814		}
815	} else {
816		/*
817		 * initialize the node
818		 */
819		node.dp2.di_atime = utime;
820		node.dp2.di_mtime = utime;
821		node.dp2.di_ctime = utime;
822		node.dp2.di_birthtime = utime;
823		/*
824		 * create the root directory
825		 */
826		node.dp2.di_mode = IFDIR | UMASK;
827		node.dp2.di_nlink = entries;
828		node.dp2.di_size = makedir(root_dir, entries);
829		node.dp2.di_db[0] = alloc(sblock.fs_fsize, node.dp2.di_mode);
830		node.dp2.di_blocks =
831		    btodb(fragroundup(&sblock, node.dp2.di_size));
832		wtfs(fsbtodb(&sblock, node.dp2.di_db[0]), sblock.fs_fsize,
833		    iobuf);
834		iput(&node, ROOTINO);
835		if (!nflag) {
836			/*
837			 * create the .snap directory
838			 */
839			node.dp2.di_mode |= 020;
840			node.dp2.di_gid = gid;
841			node.dp2.di_nlink = SNAPLINKCNT;
842			node.dp2.di_size = makedir(snap_dir, SNAPLINKCNT);
843				node.dp2.di_db[0] =
844				    alloc(sblock.fs_fsize, node.dp2.di_mode);
845			node.dp2.di_blocks =
846			    btodb(fragroundup(&sblock, node.dp2.di_size));
847				wtfs(fsbtodb(&sblock, node.dp2.di_db[0]),
848				    sblock.fs_fsize, iobuf);
849			iput(&node, ROOTINO + 1);
850		}
851	}
852}
853
854/*
855 * construct a set of directory entries in "iobuf".
856 * return size of directory.
857 */
858int
859makedir(struct direct *protodir, int entries)
860{
861	char *cp;
862	int i, spcleft;
863
864	spcleft = DIRBLKSIZ;
865	memset(iobuf, 0, DIRBLKSIZ);
866	for (cp = iobuf, i = 0; i < entries - 1; i++) {
867		protodir[i].d_reclen = DIRSIZ(0, &protodir[i]);
868		memmove(cp, &protodir[i], protodir[i].d_reclen);
869		cp += protodir[i].d_reclen;
870		spcleft -= protodir[i].d_reclen;
871	}
872	protodir[i].d_reclen = spcleft;
873	memmove(cp, &protodir[i], DIRSIZ(0, &protodir[i]));
874	return (DIRBLKSIZ);
875}
876
877/*
878 * allocate a block or frag
879 */
880ufs2_daddr_t
881alloc(int size, int mode)
882{
883	int i, d, blkno, frag;
884
885	bread(&disk, part_ofs + fsbtodb(&sblock, cgtod(&sblock, 0)), (char *)&acg,
886	    sblock.fs_cgsize);
887	if (acg.cg_magic != CG_MAGIC) {
888		printf("cg 0: bad magic number\n");
889		exit(38);
890	}
891	if (acg.cg_cs.cs_nbfree == 0) {
892		printf("first cylinder group ran out of space\n");
893		exit(39);
894	}
895	for (d = 0; d < acg.cg_ndblk; d += sblock.fs_frag)
896		if (isblock(&sblock, cg_blksfree(&acg), d / sblock.fs_frag))
897			goto goth;
898	printf("internal error: can't find block in cyl 0\n");
899	exit(40);
900goth:
901	blkno = fragstoblks(&sblock, d);
902	clrblock(&sblock, cg_blksfree(&acg), blkno);
903	if (sblock.fs_contigsumsize > 0)
904		clrbit(cg_clustersfree(&acg), blkno);
905	acg.cg_cs.cs_nbfree--;
906	sblock.fs_cstotal.cs_nbfree--;
907	fscs[0].cs_nbfree--;
908	if (mode & IFDIR) {
909		acg.cg_cs.cs_ndir++;
910		sblock.fs_cstotal.cs_ndir++;
911		fscs[0].cs_ndir++;
912	}
913	if (size != sblock.fs_bsize) {
914		frag = howmany(size, sblock.fs_fsize);
915		fscs[0].cs_nffree += sblock.fs_frag - frag;
916		sblock.fs_cstotal.cs_nffree += sblock.fs_frag - frag;
917		acg.cg_cs.cs_nffree += sblock.fs_frag - frag;
918		acg.cg_frsum[sblock.fs_frag - frag]++;
919		for (i = frag; i < sblock.fs_frag; i++)
920			setbit(cg_blksfree(&acg), d + i);
921	}
922	/* XXX cgwrite(&disk, 0)??? */
923	wtfs(fsbtodb(&sblock, cgtod(&sblock, 0)), sblock.fs_cgsize,
924	    (char *)&acg);
925	return ((ufs2_daddr_t)d);
926}
927
928/*
929 * Allocate an inode on the disk
930 */
931void
932iput(union dinode *ip, ino_t ino)
933{
934	ufs2_daddr_t d;
935	int c;
936
937	c = ino_to_cg(&sblock, ino);
938	bread(&disk, part_ofs + fsbtodb(&sblock, cgtod(&sblock, 0)), (char *)&acg,
939	    sblock.fs_cgsize);
940	if (acg.cg_magic != CG_MAGIC) {
941		printf("cg 0: bad magic number\n");
942		exit(31);
943	}
944	acg.cg_cs.cs_nifree--;
945	setbit(cg_inosused(&acg), ino);
946	wtfs(fsbtodb(&sblock, cgtod(&sblock, 0)), sblock.fs_cgsize,
947	    (char *)&acg);
948	sblock.fs_cstotal.cs_nifree--;
949	fscs[0].cs_nifree--;
950	if (ino >= (unsigned long)sblock.fs_ipg * sblock.fs_ncg) {
951		printf("fsinit: inode value out of range (%d).\n", ino);
952		exit(32);
953	}
954	d = fsbtodb(&sblock, ino_to_fsba(&sblock, ino));
955	bread(&disk, part_ofs + d, (char *)iobuf, sblock.fs_bsize);
956	if (sblock.fs_magic == FS_UFS1_MAGIC)
957		((struct ufs1_dinode *)iobuf)[ino_to_fsbo(&sblock, ino)] =
958		    ip->dp1;
959	else
960		((struct ufs2_dinode *)iobuf)[ino_to_fsbo(&sblock, ino)] =
961		    ip->dp2;
962	wtfs(d, sblock.fs_bsize, (char *)iobuf);
963}
964
965/*
966 * possibly write to disk
967 */
968static void
969wtfs(ufs2_daddr_t bno, int size, char *bf)
970{
971	if (Nflag)
972		return;
973	if (bwrite(&disk, part_ofs + bno, bf, size) < 0)
974		err(36, "wtfs: %d bytes at sector %jd", size, (intmax_t)bno);
975}
976
977/*
978 * check if a block is available
979 */
980static int
981isblock(struct fs *fs, unsigned char *cp, int h)
982{
983	unsigned char mask;
984
985	switch (fs->fs_frag) {
986	case 8:
987		return (cp[h] == 0xff);
988	case 4:
989		mask = 0x0f << ((h & 0x1) << 2);
990		return ((cp[h >> 1] & mask) == mask);
991	case 2:
992		mask = 0x03 << ((h & 0x3) << 1);
993		return ((cp[h >> 2] & mask) == mask);
994	case 1:
995		mask = 0x01 << (h & 0x7);
996		return ((cp[h >> 3] & mask) == mask);
997	default:
998		fprintf(stderr, "isblock bad fs_frag %d\n", fs->fs_frag);
999		return (0);
1000	}
1001}
1002
1003/*
1004 * take a block out of the map
1005 */
1006static void
1007clrblock(struct fs *fs, unsigned char *cp, int h)
1008{
1009	switch ((fs)->fs_frag) {
1010	case 8:
1011		cp[h] = 0;
1012		return;
1013	case 4:
1014		cp[h >> 1] &= ~(0x0f << ((h & 0x1) << 2));
1015		return;
1016	case 2:
1017		cp[h >> 2] &= ~(0x03 << ((h & 0x3) << 1));
1018		return;
1019	case 1:
1020		cp[h >> 3] &= ~(0x01 << (h & 0x7));
1021		return;
1022	default:
1023		fprintf(stderr, "clrblock bad fs_frag %d\n", fs->fs_frag);
1024		return;
1025	}
1026}
1027
1028/*
1029 * put a block into the map
1030 */
1031static void
1032setblock(struct fs *fs, unsigned char *cp, int h)
1033{
1034	switch (fs->fs_frag) {
1035	case 8:
1036		cp[h] = 0xff;
1037		return;
1038	case 4:
1039		cp[h >> 1] |= (0x0f << ((h & 0x1) << 2));
1040		return;
1041	case 2:
1042		cp[h >> 2] |= (0x03 << ((h & 0x3) << 1));
1043		return;
1044	case 1:
1045		cp[h >> 3] |= (0x01 << (h & 0x7));
1046		return;
1047	default:
1048		fprintf(stderr, "setblock bad fs_frag %d\n", fs->fs_frag);
1049		return;
1050	}
1051}
1052
1053/*
1054 * Determine the number of characters in a
1055 * single line.
1056 */
1057
1058static int
1059charsperline(void)
1060{
1061	int columns;
1062	char *cp;
1063	struct winsize ws;
1064
1065	columns = 0;
1066	if (ioctl(0, TIOCGWINSZ, &ws) != -1)
1067		columns = ws.ws_col;
1068	if (columns == 0 && (cp = getenv("COLUMNS")))
1069		columns = atoi(cp);
1070	if (columns == 0)
1071		columns = 80;	/* last resort */
1072	return (columns);
1073}
1074
1075static int
1076ilog2(int val)
1077{
1078	u_int n;
1079
1080	for (n = 0; n < sizeof(n) * CHAR_BIT; n++)
1081		if (1 << n == val)
1082			return (n);
1083	errx(1, "ilog2: %d is not a power of 2\n", val);
1084}
1085
1086/*
1087 * For the regression test, return predictable random values.
1088 * Otherwise use a true random number generator.
1089 */
1090static u_int32_t
1091newfs_random(void)
1092{
1093	static int nextnum = 1;
1094
1095	if (Rflag)
1096		return (nextnum++);
1097	return (arc4random());
1098}
1099