mkfs.c revision 174675
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 174675 2007-12-16 19:41:31Z phk $");
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
104void
105mkfs(struct partition *pp, char *fsys)
106{
107	int fragsperinode, optimalfpg, origdensity, minfpg, lastminfpg;
108	long i, j, cylno, csfrags;
109	time_t utime;
110	quad_t sizepb;
111	int width;
112	char tmpbuf[100];	/* XXX this will break in about 2,500 years */
113	union {
114		struct fs fdummy;
115		char cdummy[SBLOCKSIZE];
116	} dummy;
117#define fsdummy dummy.fdummy
118#define chdummy dummy.cdummy
119
120	/*
121	 * Our blocks == sector size, and the version of UFS we are using is
122	 * specified by Oflag.
123	 */
124	disk.d_bsize = sectorsize;
125	disk.d_ufs = Oflag;
126	if (Rflag) {
127		utime = 1000000000;
128	} else {
129		time(&utime);
130		arc4random_stir();
131	}
132	sblock.fs_old_flags = FS_FLAGS_UPDATED;
133	sblock.fs_flags = 0;
134	if (Uflag)
135		sblock.fs_flags |= FS_DOSOFTDEP;
136	if (Lflag)
137		strlcpy(sblock.fs_volname, volumelabel, MAXVOLLEN);
138	if (Jflag)
139		sblock.fs_flags |= FS_GJOURNAL;
140	if (lflag)
141		sblock.fs_flags |= FS_MULTILABEL;
142	/*
143	 * Validate the given file system size.
144	 * Verify that its last block can actually be accessed.
145	 * Convert to file system fragment sized units.
146	 */
147	if (fssize <= 0) {
148		printf("preposterous size %jd\n", (intmax_t)fssize);
149		exit(13);
150	}
151	wtfs(fssize - (realsectorsize / DEV_BSIZE), realsectorsize,
152	    (char *)&sblock);
153	/*
154	 * collect and verify the file system density info
155	 */
156	sblock.fs_avgfilesize = avgfilesize;
157	sblock.fs_avgfpdir = avgfilesperdir;
158	if (sblock.fs_avgfilesize <= 0)
159		printf("illegal expected average file size %d\n",
160		    sblock.fs_avgfilesize), exit(14);
161	if (sblock.fs_avgfpdir <= 0)
162		printf("illegal expected number of files per directory %d\n",
163		    sblock.fs_avgfpdir), exit(15);
164	/*
165	 * collect and verify the block and fragment sizes
166	 */
167	sblock.fs_bsize = bsize;
168	sblock.fs_fsize = fsize;
169	if (!POWEROF2(sblock.fs_bsize)) {
170		printf("block size must be a power of 2, not %d\n",
171		    sblock.fs_bsize);
172		exit(16);
173	}
174	if (!POWEROF2(sblock.fs_fsize)) {
175		printf("fragment size must be a power of 2, not %d\n",
176		    sblock.fs_fsize);
177		exit(17);
178	}
179	if (sblock.fs_fsize < sectorsize) {
180		printf("increasing fragment size from %d to sector size (%d)\n",
181		    sblock.fs_fsize, sectorsize);
182		sblock.fs_fsize = sectorsize;
183	}
184	if (sblock.fs_bsize > MAXBSIZE) {
185		printf("decreasing block size from %d to maximum (%d)\n",
186		    sblock.fs_bsize, MAXBSIZE);
187		sblock.fs_bsize = MAXBSIZE;
188	}
189	if (sblock.fs_bsize < MINBSIZE) {
190		printf("increasing block size from %d to minimum (%d)\n",
191		    sblock.fs_bsize, MINBSIZE);
192		sblock.fs_bsize = MINBSIZE;
193	}
194	if (sblock.fs_fsize > MAXBSIZE) {
195		printf("decreasing fragment size from %d to maximum (%d)\n",
196		    sblock.fs_fsize, MAXBSIZE);
197		sblock.fs_fsize = MAXBSIZE;
198	}
199	if (sblock.fs_bsize < sblock.fs_fsize) {
200		printf("increasing block size from %d to fragment size (%d)\n",
201		    sblock.fs_bsize, sblock.fs_fsize);
202		sblock.fs_bsize = sblock.fs_fsize;
203	}
204	if (sblock.fs_fsize * MAXFRAG < sblock.fs_bsize) {
205		printf(
206		"increasing fragment size from %d to block size / %d (%d)\n",
207		    sblock.fs_fsize, MAXFRAG, sblock.fs_bsize / MAXFRAG);
208		sblock.fs_fsize = sblock.fs_bsize / MAXFRAG;
209	}
210	if (maxbsize < bsize || !POWEROF2(maxbsize)) {
211		sblock.fs_maxbsize = sblock.fs_bsize;
212		printf("Extent size set to %d\n", sblock.fs_maxbsize);
213	} else if (sblock.fs_maxbsize > FS_MAXCONTIG * sblock.fs_bsize) {
214		sblock.fs_maxbsize = FS_MAXCONTIG * sblock.fs_bsize;
215		printf("Extent size reduced to %d\n", sblock.fs_maxbsize);
216	} else {
217		sblock.fs_maxbsize = maxbsize;
218	}
219	sblock.fs_maxcontig = maxcontig;
220	if (sblock.fs_maxcontig < sblock.fs_maxbsize / sblock.fs_bsize) {
221		sblock.fs_maxcontig = sblock.fs_maxbsize / sblock.fs_bsize;
222		printf("Maxcontig raised to %d\n", sblock.fs_maxbsize);
223	}
224	if (sblock.fs_maxcontig > 1)
225		sblock.fs_contigsumsize = MIN(sblock.fs_maxcontig,FS_MAXCONTIG);
226	sblock.fs_bmask = ~(sblock.fs_bsize - 1);
227	sblock.fs_fmask = ~(sblock.fs_fsize - 1);
228	sblock.fs_qbmask = ~sblock.fs_bmask;
229	sblock.fs_qfmask = ~sblock.fs_fmask;
230	sblock.fs_bshift = ilog2(sblock.fs_bsize);
231	sblock.fs_fshift = ilog2(sblock.fs_fsize);
232	sblock.fs_frag = numfrags(&sblock, sblock.fs_bsize);
233	sblock.fs_fragshift = ilog2(sblock.fs_frag);
234	if (sblock.fs_frag > MAXFRAG) {
235		printf("fragment size %d is still too small (can't happen)\n",
236		    sblock.fs_bsize / MAXFRAG);
237		exit(21);
238	}
239	sblock.fs_fsbtodb = ilog2(sblock.fs_fsize / sectorsize);
240	sblock.fs_size = fssize = dbtofsb(&sblock, fssize);
241
242	/*
243	 * Before the filesystem is finally initialized, mark it
244	 * as incompletely initialized.
245	 */
246	sblock.fs_magic = FS_BAD_MAGIC;
247
248	if (Oflag == 1) {
249		sblock.fs_sblockloc = SBLOCK_UFS1;
250		sblock.fs_nindir = sblock.fs_bsize / sizeof(ufs1_daddr_t);
251		sblock.fs_inopb = sblock.fs_bsize / sizeof(struct ufs1_dinode);
252		sblock.fs_maxsymlinklen = ((NDADDR + NIADDR) *
253		    sizeof(ufs1_daddr_t));
254		sblock.fs_old_inodefmt = FS_44INODEFMT;
255		sblock.fs_old_cgoffset = 0;
256		sblock.fs_old_cgmask = 0xffffffff;
257		sblock.fs_old_size = sblock.fs_size;
258		sblock.fs_old_rotdelay = 0;
259		sblock.fs_old_rps = 60;
260		sblock.fs_old_nspf = sblock.fs_fsize / sectorsize;
261		sblock.fs_old_cpg = 1;
262		sblock.fs_old_interleave = 1;
263		sblock.fs_old_trackskew = 0;
264		sblock.fs_old_cpc = 0;
265		sblock.fs_old_postblformat = 1;
266		sblock.fs_old_nrpos = 1;
267	} else {
268		sblock.fs_sblockloc = SBLOCK_UFS2;
269		sblock.fs_nindir = sblock.fs_bsize / sizeof(ufs2_daddr_t);
270		sblock.fs_inopb = sblock.fs_bsize / sizeof(struct ufs2_dinode);
271		sblock.fs_maxsymlinklen = ((NDADDR + NIADDR) *
272		    sizeof(ufs2_daddr_t));
273	}
274	sblock.fs_sblkno =
275	    roundup(howmany(sblock.fs_sblockloc + SBLOCKSIZE, sblock.fs_fsize),
276		sblock.fs_frag);
277	sblock.fs_cblkno = sblock.fs_sblkno +
278	    roundup(howmany(SBLOCKSIZE, sblock.fs_fsize), sblock.fs_frag);
279	sblock.fs_iblkno = sblock.fs_cblkno + sblock.fs_frag;
280	sblock.fs_maxfilesize = sblock.fs_bsize * NDADDR - 1;
281	for (sizepb = sblock.fs_bsize, i = 0; i < NIADDR; i++) {
282		sizepb *= NINDIR(&sblock);
283		sblock.fs_maxfilesize += sizepb;
284	}
285
286	/*
287	 * It's impossible to create a snapshot in case that fs_maxfilesize
288	 * is smaller than the fssize.
289	 */
290	if (sblock.fs_maxfilesize < (u_quad_t)fssize) {
291		warnx("WARNING: You will be unable to create snapshots on this "
292		      "file system.  Correct by using a larger blocksize.");
293	}
294
295	/*
296	 * Calculate the number of blocks to put into each cylinder group.
297	 *
298	 * This algorithm selects the number of blocks per cylinder
299	 * group. The first goal is to have at least enough data blocks
300	 * in each cylinder group to meet the density requirement. Once
301	 * this goal is achieved we try to expand to have at least
302	 * MINCYLGRPS cylinder groups. Once this goal is achieved, we
303	 * pack as many blocks into each cylinder group map as will fit.
304	 *
305	 * We start by calculating the smallest number of blocks that we
306	 * can put into each cylinder group. If this is too big, we reduce
307	 * the density until it fits.
308	 */
309	origdensity = density;
310	for (;;) {
311		fragsperinode = MAX(numfrags(&sblock, density), 1);
312		minfpg = fragsperinode * INOPB(&sblock);
313		if (minfpg > sblock.fs_size)
314			minfpg = sblock.fs_size;
315		sblock.fs_ipg = INOPB(&sblock);
316		sblock.fs_fpg = roundup(sblock.fs_iblkno +
317		    sblock.fs_ipg / INOPF(&sblock), sblock.fs_frag);
318		if (sblock.fs_fpg < minfpg)
319			sblock.fs_fpg = minfpg;
320		sblock.fs_ipg = roundup(howmany(sblock.fs_fpg, fragsperinode),
321		    INOPB(&sblock));
322		sblock.fs_fpg = roundup(sblock.fs_iblkno +
323		    sblock.fs_ipg / INOPF(&sblock), sblock.fs_frag);
324		if (sblock.fs_fpg < minfpg)
325			sblock.fs_fpg = minfpg;
326		sblock.fs_ipg = roundup(howmany(sblock.fs_fpg, fragsperinode),
327		    INOPB(&sblock));
328		if (CGSIZE(&sblock) < (unsigned long)sblock.fs_bsize)
329			break;
330		density -= sblock.fs_fsize;
331	}
332	if (density != origdensity)
333		printf("density reduced from %d to %d\n", origdensity, density);
334	/*
335	 * Start packing more blocks into the cylinder group until
336	 * it cannot grow any larger, the number of cylinder groups
337	 * drops below MINCYLGRPS, or we reach the size requested.
338	 */
339	for ( ; sblock.fs_fpg < maxblkspercg; sblock.fs_fpg += sblock.fs_frag) {
340		sblock.fs_ipg = roundup(howmany(sblock.fs_fpg, fragsperinode),
341		    INOPB(&sblock));
342		if (sblock.fs_size / sblock.fs_fpg < MINCYLGRPS)
343			break;
344		if (CGSIZE(&sblock) < (unsigned long)sblock.fs_bsize)
345			continue;
346		if (CGSIZE(&sblock) == (unsigned long)sblock.fs_bsize)
347			break;
348		sblock.fs_fpg -= sblock.fs_frag;
349		sblock.fs_ipg = roundup(howmany(sblock.fs_fpg, fragsperinode),
350		    INOPB(&sblock));
351		break;
352	}
353	/*
354	 * Check to be sure that the last cylinder group has enough blocks
355	 * to be viable. If it is too small, reduce the number of blocks
356	 * per cylinder group which will have the effect of moving more
357	 * blocks into the last cylinder group.
358	 */
359	optimalfpg = sblock.fs_fpg;
360	for (;;) {
361		sblock.fs_ncg = howmany(sblock.fs_size, sblock.fs_fpg);
362		lastminfpg = roundup(sblock.fs_iblkno +
363		    sblock.fs_ipg / INOPF(&sblock), sblock.fs_frag);
364		if (sblock.fs_size < lastminfpg) {
365			printf("Filesystem size %jd < minimum size of %d\n",
366			    (intmax_t)sblock.fs_size, lastminfpg);
367			exit(28);
368		}
369		if (sblock.fs_size % sblock.fs_fpg >= lastminfpg ||
370		    sblock.fs_size % sblock.fs_fpg == 0)
371			break;
372		sblock.fs_fpg -= sblock.fs_frag;
373		sblock.fs_ipg = roundup(howmany(sblock.fs_fpg, fragsperinode),
374		    INOPB(&sblock));
375	}
376	if (optimalfpg != sblock.fs_fpg)
377		printf("Reduced frags per cylinder group from %d to %d %s\n",
378		   optimalfpg, sblock.fs_fpg, "to enlarge last cyl group");
379	sblock.fs_cgsize = fragroundup(&sblock, CGSIZE(&sblock));
380	sblock.fs_dblkno = sblock.fs_iblkno + sblock.fs_ipg / INOPF(&sblock);
381	if (Oflag == 1) {
382		sblock.fs_old_spc = sblock.fs_fpg * sblock.fs_old_nspf;
383		sblock.fs_old_nsect = sblock.fs_old_spc;
384		sblock.fs_old_npsect = sblock.fs_old_spc;
385		sblock.fs_old_ncyl = sblock.fs_ncg;
386	}
387	/*
388	 * fill in remaining fields of the super block
389	 */
390	sblock.fs_csaddr = cgdmin(&sblock, 0);
391	sblock.fs_cssize =
392	    fragroundup(&sblock, sblock.fs_ncg * sizeof(struct csum));
393	fscs = (struct csum *)calloc(1, sblock.fs_cssize);
394	if (fscs == NULL)
395		errx(31, "calloc failed");
396	sblock.fs_sbsize = fragroundup(&sblock, sizeof(struct fs));
397	if (sblock.fs_sbsize > SBLOCKSIZE)
398		sblock.fs_sbsize = SBLOCKSIZE;
399	sblock.fs_minfree = minfree;
400	sblock.fs_maxbpg = maxbpg;
401	sblock.fs_optim = opt;
402	sblock.fs_cgrotor = 0;
403	sblock.fs_pendingblocks = 0;
404	sblock.fs_pendinginodes = 0;
405	sblock.fs_fmod = 0;
406	sblock.fs_ronly = 0;
407	sblock.fs_state = 0;
408	sblock.fs_clean = 1;
409	sblock.fs_id[0] = (long)utime;
410	sblock.fs_id[1] = newfs_random();
411	sblock.fs_fsmnt[0] = '\0';
412	csfrags = howmany(sblock.fs_cssize, sblock.fs_fsize);
413	sblock.fs_dsize = sblock.fs_size - sblock.fs_sblkno -
414	    sblock.fs_ncg * (sblock.fs_dblkno - sblock.fs_sblkno);
415	sblock.fs_cstotal.cs_nbfree =
416	    fragstoblks(&sblock, sblock.fs_dsize) -
417	    howmany(csfrags, sblock.fs_frag);
418	sblock.fs_cstotal.cs_nffree =
419	    fragnum(&sblock, sblock.fs_size) +
420	    (fragnum(&sblock, csfrags) > 0 ?
421	     sblock.fs_frag - fragnum(&sblock, csfrags) : 0);
422	sblock.fs_cstotal.cs_nifree = sblock.fs_ncg * sblock.fs_ipg - ROOTINO;
423	sblock.fs_cstotal.cs_ndir = 0;
424	sblock.fs_dsize -= csfrags;
425	sblock.fs_time = utime;
426	if (Oflag == 1) {
427		sblock.fs_old_time = utime;
428		sblock.fs_old_dsize = sblock.fs_dsize;
429		sblock.fs_old_csaddr = sblock.fs_csaddr;
430		sblock.fs_old_cstotal.cs_ndir = sblock.fs_cstotal.cs_ndir;
431		sblock.fs_old_cstotal.cs_nbfree = sblock.fs_cstotal.cs_nbfree;
432		sblock.fs_old_cstotal.cs_nifree = sblock.fs_cstotal.cs_nifree;
433		sblock.fs_old_cstotal.cs_nffree = sblock.fs_cstotal.cs_nffree;
434	}
435
436	/*
437	 * Dump out summary information about file system.
438	 */
439#	define B2MBFACTOR (1 / (1024.0 * 1024.0))
440	printf("%s: %.1fMB (%jd sectors) block size %d, fragment size %d\n",
441	    fsys, (float)sblock.fs_size * sblock.fs_fsize * B2MBFACTOR,
442	    (intmax_t)fsbtodb(&sblock, sblock.fs_size), sblock.fs_bsize,
443	    sblock.fs_fsize);
444	printf("\tusing %d cylinder groups of %.2fMB, %d blks, %d inodes.\n",
445	    sblock.fs_ncg, (float)sblock.fs_fpg * sblock.fs_fsize * B2MBFACTOR,
446	    sblock.fs_fpg / sblock.fs_frag, sblock.fs_ipg);
447	if (sblock.fs_flags & FS_DOSOFTDEP)
448		printf("\twith soft updates\n");
449#	undef B2MBFACTOR
450
451	if (Eflag && !Nflag) {
452		printf("Erasing blocks [%jd...%jd[\n",
453		    sblock.fs_sblockloc / disk.d_bsize,
454		    sblock.fs_size * sblock.fs_fsize / disk.d_bsize);
455		berase(&disk, sblock.fs_sblockloc / disk.d_bsize,
456		    sblock.fs_size * sblock.fs_fsize - sblock.fs_sblockloc);
457	}
458	/*
459	 * Wipe out old UFS1 superblock(s) if necessary.
460	 */
461	if (!Nflag && Oflag != 1) {
462		i = bread(&disk, SBLOCK_UFS1 / disk.d_bsize, chdummy, SBLOCKSIZE);
463		if (i == -1)
464			err(1, "can't read old UFS1 superblock: %s", disk.d_error);
465
466		if (fsdummy.fs_magic == FS_UFS1_MAGIC) {
467			fsdummy.fs_magic = 0;
468			bwrite(&disk, SBLOCK_UFS1 / disk.d_bsize, chdummy, SBLOCKSIZE);
469			for (i = 0; i < fsdummy.fs_ncg; i++)
470				bwrite(&disk, fsbtodb(&fsdummy, cgsblock(&fsdummy, i)),
471	                    chdummy, SBLOCKSIZE);
472		}
473	}
474	if (!Nflag)
475		sbwrite(&disk, 0);
476	if (Xflag == 1) {
477		printf("** Exiting on Xflag 1\n");
478		exit(0);
479	}
480	if (Xflag == 2)
481		printf("** Leaving BAD MAGIC on Xflag 2\n");
482	else
483		sblock.fs_magic = (Oflag != 1) ? FS_UFS2_MAGIC : FS_UFS1_MAGIC;
484
485	/*
486	 * Now build the cylinders group blocks and
487	 * then print out indices of cylinder groups.
488	 */
489	printf("super-block backups (for fsck -b #) at:\n");
490	i = 0;
491	width = charsperline();
492	/*
493	 * allocate space for superblock, cylinder group map, and
494	 * two sets of inode blocks.
495	 */
496	if (sblock.fs_bsize < SBLOCKSIZE)
497		iobufsize = SBLOCKSIZE + 3 * sblock.fs_bsize;
498	else
499		iobufsize = 4 * sblock.fs_bsize;
500	if ((iobuf = malloc(iobufsize)) == 0) {
501		printf("Cannot allocate I/O buffer\n");
502		exit(38);
503	}
504	bzero(iobuf, iobufsize);
505	/*
506	 * Make a copy of the superblock into the buffer that we will be
507	 * writing out in each cylinder group.
508	 */
509	bcopy((char *)&sblock, iobuf, SBLOCKSIZE);
510	for (cylno = 0; cylno < sblock.fs_ncg; cylno++) {
511		initcg(cylno, utime);
512		j = snprintf(tmpbuf, sizeof(tmpbuf), " %jd%s",
513		    (intmax_t)fsbtodb(&sblock, cgsblock(&sblock, cylno)),
514		    cylno < (sblock.fs_ncg-1) ? "," : "");
515		if (j < 0)
516			tmpbuf[j = 0] = '\0';
517		if (i + j >= width) {
518			printf("\n");
519			i = 0;
520		}
521		i += j;
522		printf("%s", tmpbuf);
523		fflush(stdout);
524	}
525	printf("\n");
526	if (Nflag)
527		exit(0);
528	/*
529	 * Now construct the initial file system,
530	 * then write out the super-block.
531	 */
532	fsinit(utime);
533	if (Oflag == 1) {
534		sblock.fs_old_cstotal.cs_ndir = sblock.fs_cstotal.cs_ndir;
535		sblock.fs_old_cstotal.cs_nbfree = sblock.fs_cstotal.cs_nbfree;
536		sblock.fs_old_cstotal.cs_nifree = sblock.fs_cstotal.cs_nifree;
537		sblock.fs_old_cstotal.cs_nffree = sblock.fs_cstotal.cs_nffree;
538	}
539	if (Xflag == 3) {
540		printf("** Exiting on Xflag 3\n");
541		exit(0);
542	}
543	if (!Nflag)
544		sbwrite(&disk, 0);
545	for (i = 0; i < sblock.fs_cssize; i += sblock.fs_bsize)
546		wtfs(fsbtodb(&sblock, sblock.fs_csaddr + numfrags(&sblock, i)),
547			sblock.fs_cssize - i < sblock.fs_bsize ?
548			sblock.fs_cssize - i : sblock.fs_bsize,
549			((char *)fscs) + i);
550	/*
551	 * Update information about this partion in pack
552	 * label, to that it may be updated on disk.
553	 */
554	if (pp != NULL) {
555		pp->p_fstype = FS_BSDFFS;
556		pp->p_fsize = sblock.fs_fsize;
557		pp->p_frag = sblock.fs_frag;
558		pp->p_cpg = sblock.fs_fpg;
559	}
560}
561
562/*
563 * Initialize a cylinder group.
564 */
565void
566initcg(int cylno, time_t utime)
567{
568	long i, j, d, dlower, dupper, blkno, start;
569	ufs2_daddr_t cbase, dmax;
570	struct ufs1_dinode *dp1;
571	struct ufs2_dinode *dp2;
572	struct csum *cs;
573
574	/*
575	 * Determine block bounds for cylinder group.
576	 * Allow space for super block summary information in first
577	 * cylinder group.
578	 */
579	cbase = cgbase(&sblock, cylno);
580	dmax = cbase + sblock.fs_fpg;
581	if (dmax > sblock.fs_size)
582		dmax = sblock.fs_size;
583	dlower = cgsblock(&sblock, cylno) - cbase;
584	dupper = cgdmin(&sblock, cylno) - cbase;
585	if (cylno == 0)
586		dupper += howmany(sblock.fs_cssize, sblock.fs_fsize);
587	cs = &fscs[cylno];
588	memset(&acg, 0, sblock.fs_cgsize);
589	acg.cg_time = utime;
590	acg.cg_magic = CG_MAGIC;
591	acg.cg_cgx = cylno;
592	acg.cg_niblk = sblock.fs_ipg;
593	acg.cg_initediblk = sblock.fs_ipg < 2 * INOPB(&sblock) ?
594	    sblock.fs_ipg : 2 * INOPB(&sblock);
595	acg.cg_ndblk = dmax - cbase;
596	if (sblock.fs_contigsumsize > 0)
597		acg.cg_nclusterblks = acg.cg_ndblk / sblock.fs_frag;
598	start = &acg.cg_space[0] - (u_char *)(&acg.cg_firstfield);
599	if (Oflag == 2) {
600		acg.cg_iusedoff = start;
601	} else {
602		acg.cg_old_ncyl = sblock.fs_old_cpg;
603		acg.cg_old_time = acg.cg_time;
604		acg.cg_time = 0;
605		acg.cg_old_niblk = acg.cg_niblk;
606		acg.cg_niblk = 0;
607		acg.cg_initediblk = 0;
608		acg.cg_old_btotoff = start;
609		acg.cg_old_boff = acg.cg_old_btotoff +
610		    sblock.fs_old_cpg * sizeof(int32_t);
611		acg.cg_iusedoff = acg.cg_old_boff +
612		    sblock.fs_old_cpg * sizeof(u_int16_t);
613	}
614	acg.cg_freeoff = acg.cg_iusedoff + howmany(sblock.fs_ipg, CHAR_BIT);
615	acg.cg_nextfreeoff = acg.cg_freeoff + howmany(sblock.fs_fpg, CHAR_BIT);
616	if (sblock.fs_contigsumsize > 0) {
617		acg.cg_clustersumoff =
618		    roundup(acg.cg_nextfreeoff, sizeof(u_int32_t));
619		acg.cg_clustersumoff -= sizeof(u_int32_t);
620		acg.cg_clusteroff = acg.cg_clustersumoff +
621		    (sblock.fs_contigsumsize + 1) * sizeof(u_int32_t);
622		acg.cg_nextfreeoff = acg.cg_clusteroff +
623		    howmany(fragstoblks(&sblock, sblock.fs_fpg), CHAR_BIT);
624	}
625	if (acg.cg_nextfreeoff > sblock.fs_cgsize) {
626		printf("Panic: cylinder group too big\n");
627		exit(37);
628	}
629	acg.cg_cs.cs_nifree += sblock.fs_ipg;
630	if (cylno == 0)
631		for (i = 0; i < (long)ROOTINO; i++) {
632			setbit(cg_inosused(&acg), i);
633			acg.cg_cs.cs_nifree--;
634		}
635	if (cylno > 0) {
636		/*
637		 * In cylno 0, beginning space is reserved
638		 * for boot and super blocks.
639		 */
640		for (d = 0; d < dlower; d += sblock.fs_frag) {
641			blkno = d / sblock.fs_frag;
642			setblock(&sblock, cg_blksfree(&acg), blkno);
643			if (sblock.fs_contigsumsize > 0)
644				setbit(cg_clustersfree(&acg), blkno);
645			acg.cg_cs.cs_nbfree++;
646		}
647	}
648	if ((i = dupper % sblock.fs_frag)) {
649		acg.cg_frsum[sblock.fs_frag - i]++;
650		for (d = dupper + sblock.fs_frag - i; dupper < d; dupper++) {
651			setbit(cg_blksfree(&acg), dupper);
652			acg.cg_cs.cs_nffree++;
653		}
654	}
655	for (d = dupper; d + sblock.fs_frag <= acg.cg_ndblk;
656	     d += sblock.fs_frag) {
657		blkno = d / sblock.fs_frag;
658		setblock(&sblock, cg_blksfree(&acg), blkno);
659		if (sblock.fs_contigsumsize > 0)
660			setbit(cg_clustersfree(&acg), blkno);
661		acg.cg_cs.cs_nbfree++;
662	}
663	if (d < acg.cg_ndblk) {
664		acg.cg_frsum[acg.cg_ndblk - d]++;
665		for (; d < acg.cg_ndblk; d++) {
666			setbit(cg_blksfree(&acg), d);
667			acg.cg_cs.cs_nffree++;
668		}
669	}
670	if (sblock.fs_contigsumsize > 0) {
671		int32_t *sump = cg_clustersum(&acg);
672		u_char *mapp = cg_clustersfree(&acg);
673		int map = *mapp++;
674		int bit = 1;
675		int run = 0;
676
677		for (i = 0; i < acg.cg_nclusterblks; i++) {
678			if ((map & bit) != 0)
679				run++;
680			else if (run != 0) {
681				if (run > sblock.fs_contigsumsize)
682					run = sblock.fs_contigsumsize;
683				sump[run]++;
684				run = 0;
685			}
686			if ((i & (CHAR_BIT - 1)) != CHAR_BIT - 1)
687				bit <<= 1;
688			else {
689				map = *mapp++;
690				bit = 1;
691			}
692		}
693		if (run != 0) {
694			if (run > sblock.fs_contigsumsize)
695				run = sblock.fs_contigsumsize;
696			sump[run]++;
697		}
698	}
699	*cs = acg.cg_cs;
700	/*
701	 * Write out the duplicate super block, the cylinder group map
702	 * and two blocks worth of inodes in a single write.
703	 */
704	start = sblock.fs_bsize > SBLOCKSIZE ? sblock.fs_bsize : SBLOCKSIZE;
705	bcopy((char *)&acg, &iobuf[start], sblock.fs_cgsize);
706	start += sblock.fs_bsize;
707	dp1 = (struct ufs1_dinode *)(&iobuf[start]);
708	dp2 = (struct ufs2_dinode *)(&iobuf[start]);
709	for (i = 0; i < acg.cg_initediblk; i++) {
710		if (sblock.fs_magic == FS_UFS1_MAGIC) {
711			dp1->di_gen = newfs_random();
712			dp1++;
713		} else {
714			dp2->di_gen = newfs_random();
715			dp2++;
716		}
717	}
718	wtfs(fsbtodb(&sblock, cgsblock(&sblock, cylno)), iobufsize, iobuf);
719	/*
720	 * For the old file system, we have to initialize all the inodes.
721	 */
722	if (Oflag == 1) {
723		for (i = 2 * sblock.fs_frag;
724		     i < sblock.fs_ipg / INOPF(&sblock);
725		     i += sblock.fs_frag) {
726			dp1 = (struct ufs1_dinode *)(&iobuf[start]);
727			for (j = 0; j < INOPB(&sblock); j++) {
728				dp1->di_gen = newfs_random();
729				dp1++;
730			}
731			wtfs(fsbtodb(&sblock, cgimin(&sblock, cylno) + i),
732			    sblock.fs_bsize, &iobuf[start]);
733		}
734	}
735}
736
737/*
738 * initialize the file system
739 */
740#define ROOTLINKCNT 3
741
742struct direct root_dir[] = {
743	{ ROOTINO, sizeof(struct direct), DT_DIR, 1, "." },
744	{ ROOTINO, sizeof(struct direct), DT_DIR, 2, ".." },
745	{ ROOTINO + 1, sizeof(struct direct), DT_DIR, 5, ".snap" },
746};
747
748#define SNAPLINKCNT 2
749
750struct direct snap_dir[] = {
751	{ ROOTINO + 1, sizeof(struct direct), DT_DIR, 1, "." },
752	{ ROOTINO, sizeof(struct direct), DT_DIR, 2, ".." },
753};
754
755void
756fsinit(time_t utime)
757{
758	union dinode node;
759	struct group *grp;
760	gid_t gid;
761	int entries;
762
763	memset(&node, 0, sizeof node);
764	if ((grp = getgrnam("operator")) != NULL) {
765		gid = grp->gr_gid;
766	} else {
767		warnx("Cannot retrieve operator gid, using gid 0.");
768		gid = 0;
769	}
770	entries = (nflag) ? ROOTLINKCNT - 1: ROOTLINKCNT;
771	if (sblock.fs_magic == FS_UFS1_MAGIC) {
772		/*
773		 * initialize the node
774		 */
775		node.dp1.di_atime = utime;
776		node.dp1.di_mtime = utime;
777		node.dp1.di_ctime = utime;
778		/*
779		 * create the root directory
780		 */
781		node.dp1.di_mode = IFDIR | UMASK;
782		node.dp1.di_nlink = entries;
783		node.dp1.di_size = makedir(root_dir, entries);
784		node.dp1.di_db[0] = alloc(sblock.fs_fsize, node.dp1.di_mode);
785		node.dp1.di_blocks =
786		    btodb(fragroundup(&sblock, node.dp1.di_size));
787		wtfs(fsbtodb(&sblock, node.dp1.di_db[0]), sblock.fs_fsize,
788		    iobuf);
789		iput(&node, ROOTINO);
790		if (!nflag) {
791			/*
792			 * create the .snap directory
793			 */
794			node.dp1.di_mode |= 020;
795			node.dp1.di_gid = gid;
796			node.dp1.di_nlink = SNAPLINKCNT;
797			node.dp1.di_size = makedir(snap_dir, SNAPLINKCNT);
798				node.dp1.di_db[0] =
799				    alloc(sblock.fs_fsize, node.dp1.di_mode);
800			node.dp1.di_blocks =
801			    btodb(fragroundup(&sblock, node.dp1.di_size));
802				wtfs(fsbtodb(&sblock, node.dp1.di_db[0]),
803				    sblock.fs_fsize, iobuf);
804			iput(&node, ROOTINO + 1);
805		}
806	} else {
807		/*
808		 * initialize the node
809		 */
810		node.dp2.di_atime = utime;
811		node.dp2.di_mtime = utime;
812		node.dp2.di_ctime = utime;
813		node.dp2.di_birthtime = utime;
814		/*
815		 * create the root directory
816		 */
817		node.dp2.di_mode = IFDIR | UMASK;
818		node.dp2.di_nlink = entries;
819		node.dp2.di_size = makedir(root_dir, entries);
820		node.dp2.di_db[0] = alloc(sblock.fs_fsize, node.dp2.di_mode);
821		node.dp2.di_blocks =
822		    btodb(fragroundup(&sblock, node.dp2.di_size));
823		wtfs(fsbtodb(&sblock, node.dp2.di_db[0]), sblock.fs_fsize,
824		    iobuf);
825		iput(&node, ROOTINO);
826		if (!nflag) {
827			/*
828			 * create the .snap directory
829			 */
830			node.dp2.di_mode |= 020;
831			node.dp2.di_gid = gid;
832			node.dp2.di_nlink = SNAPLINKCNT;
833			node.dp2.di_size = makedir(snap_dir, SNAPLINKCNT);
834				node.dp2.di_db[0] =
835				    alloc(sblock.fs_fsize, node.dp2.di_mode);
836			node.dp2.di_blocks =
837			    btodb(fragroundup(&sblock, node.dp2.di_size));
838				wtfs(fsbtodb(&sblock, node.dp2.di_db[0]),
839				    sblock.fs_fsize, iobuf);
840			iput(&node, ROOTINO + 1);
841		}
842	}
843}
844
845/*
846 * construct a set of directory entries in "iobuf".
847 * return size of directory.
848 */
849int
850makedir(struct direct *protodir, int entries)
851{
852	char *cp;
853	int i, spcleft;
854
855	spcleft = DIRBLKSIZ;
856	memset(iobuf, 0, DIRBLKSIZ);
857	for (cp = iobuf, i = 0; i < entries - 1; i++) {
858		protodir[i].d_reclen = DIRSIZ(0, &protodir[i]);
859		memmove(cp, &protodir[i], protodir[i].d_reclen);
860		cp += protodir[i].d_reclen;
861		spcleft -= protodir[i].d_reclen;
862	}
863	protodir[i].d_reclen = spcleft;
864	memmove(cp, &protodir[i], DIRSIZ(0, &protodir[i]));
865	return (DIRBLKSIZ);
866}
867
868/*
869 * allocate a block or frag
870 */
871ufs2_daddr_t
872alloc(int size, int mode)
873{
874	int i, d, blkno, frag;
875
876	bread(&disk, fsbtodb(&sblock, cgtod(&sblock, 0)), (char *)&acg,
877	    sblock.fs_cgsize);
878	if (acg.cg_magic != CG_MAGIC) {
879		printf("cg 0: bad magic number\n");
880		exit(38);
881	}
882	if (acg.cg_cs.cs_nbfree == 0) {
883		printf("first cylinder group ran out of space\n");
884		exit(39);
885	}
886	for (d = 0; d < acg.cg_ndblk; d += sblock.fs_frag)
887		if (isblock(&sblock, cg_blksfree(&acg), d / sblock.fs_frag))
888			goto goth;
889	printf("internal error: can't find block in cyl 0\n");
890	exit(40);
891goth:
892	blkno = fragstoblks(&sblock, d);
893	clrblock(&sblock, cg_blksfree(&acg), blkno);
894	if (sblock.fs_contigsumsize > 0)
895		clrbit(cg_clustersfree(&acg), blkno);
896	acg.cg_cs.cs_nbfree--;
897	sblock.fs_cstotal.cs_nbfree--;
898	fscs[0].cs_nbfree--;
899	if (mode & IFDIR) {
900		acg.cg_cs.cs_ndir++;
901		sblock.fs_cstotal.cs_ndir++;
902		fscs[0].cs_ndir++;
903	}
904	if (size != sblock.fs_bsize) {
905		frag = howmany(size, sblock.fs_fsize);
906		fscs[0].cs_nffree += sblock.fs_frag - frag;
907		sblock.fs_cstotal.cs_nffree += sblock.fs_frag - frag;
908		acg.cg_cs.cs_nffree += sblock.fs_frag - frag;
909		acg.cg_frsum[sblock.fs_frag - frag]++;
910		for (i = frag; i < sblock.fs_frag; i++)
911			setbit(cg_blksfree(&acg), d + i);
912	}
913	/* XXX cgwrite(&disk, 0)??? */
914	wtfs(fsbtodb(&sblock, cgtod(&sblock, 0)), sblock.fs_cgsize,
915	    (char *)&acg);
916	return ((ufs2_daddr_t)d);
917}
918
919/*
920 * Allocate an inode on the disk
921 */
922void
923iput(union dinode *ip, ino_t ino)
924{
925	ufs2_daddr_t d;
926	int c;
927
928	c = ino_to_cg(&sblock, ino);
929	bread(&disk, fsbtodb(&sblock, cgtod(&sblock, 0)), (char *)&acg,
930	    sblock.fs_cgsize);
931	if (acg.cg_magic != CG_MAGIC) {
932		printf("cg 0: bad magic number\n");
933		exit(31);
934	}
935	acg.cg_cs.cs_nifree--;
936	setbit(cg_inosused(&acg), ino);
937	wtfs(fsbtodb(&sblock, cgtod(&sblock, 0)), sblock.fs_cgsize,
938	    (char *)&acg);
939	sblock.fs_cstotal.cs_nifree--;
940	fscs[0].cs_nifree--;
941	if (ino >= (unsigned long)sblock.fs_ipg * sblock.fs_ncg) {
942		printf("fsinit: inode value out of range (%d).\n", ino);
943		exit(32);
944	}
945	d = fsbtodb(&sblock, ino_to_fsba(&sblock, ino));
946	bread(&disk, d, (char *)iobuf, sblock.fs_bsize);
947	if (sblock.fs_magic == FS_UFS1_MAGIC)
948		((struct ufs1_dinode *)iobuf)[ino_to_fsbo(&sblock, ino)] =
949		    ip->dp1;
950	else
951		((struct ufs2_dinode *)iobuf)[ino_to_fsbo(&sblock, ino)] =
952		    ip->dp2;
953	wtfs(d, sblock.fs_bsize, (char *)iobuf);
954}
955
956/*
957 * possibly write to disk
958 */
959static void
960wtfs(ufs2_daddr_t bno, int size, char *bf)
961{
962	if (Nflag)
963		return;
964	if (bwrite(&disk, bno, bf, size) < 0)
965		err(36, "wtfs: %d bytes at sector %jd", size, (intmax_t)bno);
966}
967
968/*
969 * check if a block is available
970 */
971static int
972isblock(struct fs *fs, unsigned char *cp, int h)
973{
974	unsigned char mask;
975
976	switch (fs->fs_frag) {
977	case 8:
978		return (cp[h] == 0xff);
979	case 4:
980		mask = 0x0f << ((h & 0x1) << 2);
981		return ((cp[h >> 1] & mask) == mask);
982	case 2:
983		mask = 0x03 << ((h & 0x3) << 1);
984		return ((cp[h >> 2] & mask) == mask);
985	case 1:
986		mask = 0x01 << (h & 0x7);
987		return ((cp[h >> 3] & mask) == mask);
988	default:
989		fprintf(stderr, "isblock bad fs_frag %d\n", fs->fs_frag);
990		return (0);
991	}
992}
993
994/*
995 * take a block out of the map
996 */
997static void
998clrblock(struct fs *fs, unsigned char *cp, int h)
999{
1000	switch ((fs)->fs_frag) {
1001	case 8:
1002		cp[h] = 0;
1003		return;
1004	case 4:
1005		cp[h >> 1] &= ~(0x0f << ((h & 0x1) << 2));
1006		return;
1007	case 2:
1008		cp[h >> 2] &= ~(0x03 << ((h & 0x3) << 1));
1009		return;
1010	case 1:
1011		cp[h >> 3] &= ~(0x01 << (h & 0x7));
1012		return;
1013	default:
1014		fprintf(stderr, "clrblock bad fs_frag %d\n", fs->fs_frag);
1015		return;
1016	}
1017}
1018
1019/*
1020 * put a block into the map
1021 */
1022static void
1023setblock(struct fs *fs, unsigned char *cp, int h)
1024{
1025	switch (fs->fs_frag) {
1026	case 8:
1027		cp[h] = 0xff;
1028		return;
1029	case 4:
1030		cp[h >> 1] |= (0x0f << ((h & 0x1) << 2));
1031		return;
1032	case 2:
1033		cp[h >> 2] |= (0x03 << ((h & 0x3) << 1));
1034		return;
1035	case 1:
1036		cp[h >> 3] |= (0x01 << (h & 0x7));
1037		return;
1038	default:
1039		fprintf(stderr, "setblock bad fs_frag %d\n", fs->fs_frag);
1040		return;
1041	}
1042}
1043
1044/*
1045 * Determine the number of characters in a
1046 * single line.
1047 */
1048
1049static int
1050charsperline(void)
1051{
1052	int columns;
1053	char *cp;
1054	struct winsize ws;
1055
1056	columns = 0;
1057	if (ioctl(0, TIOCGWINSZ, &ws) != -1)
1058		columns = ws.ws_col;
1059	if (columns == 0 && (cp = getenv("COLUMNS")))
1060		columns = atoi(cp);
1061	if (columns == 0)
1062		columns = 80;	/* last resort */
1063	return (columns);
1064}
1065
1066static int
1067ilog2(int val)
1068{
1069	u_int n;
1070
1071	for (n = 0; n < sizeof(n) * CHAR_BIT; n++)
1072		if (1 << n == val)
1073			return (n);
1074	errx(1, "ilog2: %d is not a power of 2\n", val);
1075}
1076
1077/*
1078 * For the regression test, return predictable random values.
1079 * Otherwise use a true random number generator.
1080 */
1081static u_int32_t
1082newfs_random(void)
1083{
1084	static int nextnum = 1;
1085
1086	if (Rflag)
1087		return (nextnum++);
1088	return (arc4random());
1089}
1090