umount: getopt return value is int, not char
[minix.git] / sbin / newfs_ext2fs / mke2fs.c
blob5ecb882f4a97ff6b00c82e4bf53668fc6d7161ce
1 /* $NetBSD: mke2fs.c,v 1.14 2010/09/10 15:51:20 tsutsui Exp $ */
3 /*-
4 * Copyright (c) 2007 Izumi Tsutsui. All rights reserved.
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
8 * are met:
9 * 1. Redistributions of source code must retain the above copyright
10 * notice, this list of conditions and the following disclaimer.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
16 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
17 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
18 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
19 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
20 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
21 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
22 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
23 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
24 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
28 * Copyright (c) 1980, 1989, 1993
29 * The Regents of the University of California. All rights reserved.
31 * Redistribution and use in source and binary forms, with or without
32 * modification, are permitted provided that the following conditions
33 * are met:
34 * 1. Redistributions of source code must retain the above copyright
35 * notice, this list of conditions and the following disclaimer.
36 * 2. Redistributions in binary form must reproduce the above copyright
37 * notice, this list of conditions and the following disclaimer in the
38 * documentation and/or other materials provided with the distribution.
39 * 3. Neither the name of the University nor the names of its contributors
40 * may be used to endorse or promote products derived from this software
41 * without specific prior written permission.
43 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
44 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
45 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
46 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
47 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
48 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
49 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
50 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
51 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
52 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
53 * SUCH DAMAGE.
57 * Copyright (c) 1997 Manuel Bouyer.
59 * Redistribution and use in source and binary forms, with or without
60 * modification, are permitted provided that the following conditions
61 * are met:
62 * 1. Redistributions of source code must retain the above copyright
63 * notice, this list of conditions and the following disclaimer.
64 * 2. Redistributions in binary form must reproduce the above copyright
65 * notice, this list of conditions and the following disclaimer in the
66 * documentation and/or other materials provided with the distribution.
68 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
69 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
70 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
71 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
72 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
73 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
74 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
75 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
76 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
77 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
80 /* mmap/munmap are used in this file just to allocate/free memory
81 * so these functions are ok.
83 #define mmap minix_mmap
84 #define munmap minix_munmap
87 * mke2fs.c: "re-invent (dumb but non-GPLed) wheel as a fun project"
89 * In spite of this name, there is no piece of code
90 * derived from GPLed e2fsprogs written for Linux.
91 * I referred them only to see how each structure
92 * member should be initialized.
94 * Reference:
95 * - All NetBSD sources under src/sys/ufs/ext2fs and src/sbin/fsck_ext2fs
96 * - Ext2fs Home Page
97 * http://e2fsprogs.sourceforge.net/ext2.html
98 * - Design and Implementation of the Second Extended Filesystem
99 * http://e2fsprogs.sourceforge.net/ext2intro.html
100 * - Linux Documentation "The Second Extended Filesystem"
101 * http://www.kernel.org/doc/Documentation/filesystems/ext2.txt
104 #include <sys/cdefs.h>
105 #ifndef lint
106 #if 0
107 static char sccsid[] = "@(#)mkfs.c 8.11 (Berkeley) 5/3/95";
108 #else
109 __RCSID("$NetBSD: mke2fs.c,v 1.14 2010/09/10 15:51:20 tsutsui Exp $");
110 #endif
111 #endif /* not lint */
113 #include <sys/param.h>
114 #include <sys/mman.h>
115 #include <sys/time.h>
116 #include <ufs/ext2fs/ext2fs_dinode.h>
117 #include <ufs/ext2fs/ext2fs_dir.h>
118 #include <ufs/ext2fs/ext2fs.h>
119 #include <sys/ioctl.h>
121 #define _MINIX
122 #include <err.h>
123 #include <errno.h>
124 #include <string.h>
125 #include <unistd.h>
126 #include <stdlib.h>
127 #include <stddef.h>
128 #include <stdio.h>
129 #include <uuid.h>
130 #include <assert.h>
131 #include <minix/termios.h>
133 #include "extern.h"
135 static void initcg(uint);
136 static void zap_old_sblock(daddr_t);
137 static uint cgoverhead(uint);
138 static int fsinit(const struct timeval *);
139 static int makedir(struct ext2fs_direct *, int);
140 static void copy_dir(struct ext2fs_direct *, struct ext2fs_direct *);
141 static void init_resizeino(const struct timeval *);
142 static uint32_t alloc(uint32_t, uint16_t);
143 static void iput(struct ext2fs_dinode *, ino_t);
144 static void rdfs(daddr_t, int, void *);
145 static void wtfs(daddr_t, int, void *);
146 static int ilog2(uint);
147 static int skpc(int, size_t, uint8_t *);
149 /* XXX: some of these macro should be into <ufs/ext2fs/ext2fs.h>? */
150 #define EXT2_DEF_MAX_MNT_COUNT 20
151 #define EXT2_DEF_FSCKINTV (180 * 24 * 60 * 60) /* 180 days */
152 #define EXT2_RESERVED_INODES (EXT2_FIRSTINO - 1)
153 #define EXT2_UMASK 0755
155 #define EXT2_INO_INDEX(ino) ((ino) - 1) /* no inode zero */
157 #define EXT2_LOSTFOUNDSIZE 16384
158 #define EXT2_LOSTFOUNDINO EXT2_FIRSTINO /* XXX: not quite */
159 #define EXT2_LOSTFOUNDUMASK 0700
161 #define EXT2_RESIZEINOUMASK 0600
163 #define NBLOCK_SUPERBLOCK 1
164 #define NBLOCK_BLOCK_BITMAP 1
165 #define NBLOCK_INODE_BITMAP 1
167 #define cgbase(fs, c) \
168 ((fs)->e2fs.e2fs_first_dblock + (fs)->e2fs.e2fs_bpg * (c))
172 * ext2fs super block and group descriptor structures
174 * We don't have to use or setup whole in-memory m_ext2fs structure,
175 * but prepare it to use several macro defined in kernel headers.
177 union {
178 struct m_ext2fs m_ext2fs;
179 char pad[SBSIZE];
180 } ext2fsun;
181 #define sblock ext2fsun.m_ext2fs
182 #define gd ext2fsun.m_ext2fs.e2fs_gd
184 static uint8_t *iobuf; /* for superblock and group descriptors */
185 static int iobufsize;
187 static uint8_t buf[MAXBSIZE]; /* for initcg() and makedir() ops */
189 static int fsi = -1, fso = -1;
191 void
192 mke2fs(const char *fsys, int fi, int fo)
194 struct timeval tv;
195 int64_t minfssize;
196 uint bcount, fbcount, ficount;
197 uint blocks_gd, blocks_per_cg, inodes_per_cg, iblocks_per_cg;
198 uint minblocks_per_cg, blocks_lastcg;
199 uint ncg, cylno, sboff;
200 uuid_t uuid;
201 uint32_t uustat;
202 int i, len, col, delta, fld_width, max_cols;
203 struct winsize winsize;
205 gettimeofday(&tv, NULL);
206 fsi = fi;
207 fso = fo;
210 * collect and verify the block and fragment sizes
212 if (!powerof2(bsize)) {
213 errx(EXIT_FAILURE,
214 "block size must be a power of 2, not %u\n",
215 bsize);
217 if (!powerof2(fsize)) {
218 errx(EXIT_FAILURE,
219 "fragment size must be a power of 2, not %u\n",
220 fsize);
222 if (fsize < sectorsize) {
223 errx(EXIT_FAILURE,
224 "fragment size %u is too small, minimum is %u\n",
225 fsize, sectorsize);
227 if (bsize < MINBSIZE) {
228 errx(EXIT_FAILURE,
229 "block size %u is too small, minimum is %u\n",
230 bsize, MINBSIZE);
232 if (bsize > EXT2_MAXBSIZE) {
233 errx(EXIT_FAILURE,
234 "block size %u is too large, maximum is %u\n",
235 bsize, MAXBSIZE);
237 if (bsize != fsize) {
239 * There is no fragment support on current ext2fs (yet?),
240 * but some kernel code refers fsize or fpg as bsize or bpg
241 * and Linux seems to set the same values to them.
243 errx(EXIT_FAILURE,
244 "block size (%u) can't be different from "
245 "fragment size (%u)\n",
246 bsize, fsize);
249 /* variable inodesize is REV1 feature */
250 if (Oflag == 0 && inodesize != EXT2_REV0_DINODE_SIZE) {
251 errx(EXIT_FAILURE, "GOOD_OLD_REV file system format"
252 " doesn't support %d byte inode\n", inodesize);
255 sblock.e2fs.e2fs_log_bsize = ilog2(bsize) - LOG_MINBSIZE;
256 /* Umm, why not e2fs_log_fsize? */
257 sblock.e2fs.e2fs_fsize = ilog2(fsize) - LOG_MINBSIZE;
259 sblock.e2fs_bsize = bsize;
260 sblock.e2fs_bshift = sblock.e2fs.e2fs_log_bsize + LOG_MINBSIZE;
261 sblock.e2fs_qbmask = sblock.e2fs_bsize - 1;
262 sblock.e2fs_bmask = ~sblock.e2fs_qbmask;
263 sblock.e2fs_fsbtodb = ilog2(sblock.e2fs_bsize) - ilog2(sectorsize);
264 sblock.e2fs_ipb = sblock.e2fs_bsize / inodesize;
267 * Ext2fs preserves BBSIZE (1024 bytes) space at the top for
268 * bootloader (though it is not enough at all for our bootloader).
269 * If bsize == BBSIZE we have to preserve one block.
270 * If bsize > BBSIZE, the first block already contains BBSIZE space
271 * before superblock because superblock is allocated at SBOFF and
272 * bsize is a power of two (i.e. 2048 bytes or more).
274 sblock.e2fs.e2fs_first_dblock = (sblock.e2fs_bsize > BBSIZE) ? 0 : 1;
275 minfssize = fsbtodb(&sblock,
276 sblock.e2fs.e2fs_first_dblock +
277 NBLOCK_SUPERBLOCK +
278 1 /* at least one group descriptor */ +
279 NBLOCK_BLOCK_BITMAP +
280 NBLOCK_INODE_BITMAP +
281 1 /* at least one inode table block */ +
282 1 /* at least one data block for rootdir */ +
283 1 /* at least one data block for data */
284 ); /* XXX and more? */
286 if (fssize < minfssize)
287 errx(EXIT_FAILURE, "Filesystem size %" PRId64
288 " < minimum size of %" PRId64 "\n", fssize, minfssize);
290 bcount = dbtofsb(&sblock, fssize);
293 * While many people claim that ext2fs is a (bad) clone of ufs/ffs,
294 * it isn't actual ffs so maybe we should call it "block group"
295 * as their native name rather than ffs derived "cylinder group."
296 * But we'll use the latter here since other kernel sources use it.
297 * (I also agree "cylinder" based allocation is obsolete though)
300 /* maybe "simple is the best" */
301 blocks_per_cg = sblock.e2fs_bsize * NBBY;
303 ncg = howmany(bcount - sblock.e2fs.e2fs_first_dblock, blocks_per_cg);
304 blocks_gd = howmany(sizeof(struct ext2_gd) * ncg, bsize);
306 /* check range of inode number */
307 if (num_inodes < EXT2_FIRSTINO)
308 num_inodes = EXT2_FIRSTINO; /* needs reserved inodes + 1 */
309 if (num_inodes > UINT16_MAX * ncg)
310 num_inodes = UINT16_MAX * ncg; /* ext2bgd_nifree is uint16_t */
312 inodes_per_cg = num_inodes / ncg;
313 iblocks_per_cg = howmany(inodesize * inodes_per_cg, bsize);
315 /* Check that the last cylinder group has enough space for inodes */
316 minblocks_per_cg =
317 NBLOCK_BLOCK_BITMAP +
318 NBLOCK_INODE_BITMAP +
319 iblocks_per_cg +
320 1; /* at least one data block */
321 if (Oflag == 0 || cg_has_sb(ncg - 1) != 0)
322 minblocks_per_cg += NBLOCK_SUPERBLOCK + blocks_gd;
324 blocks_lastcg = bcount - sblock.e2fs.e2fs_first_dblock -
325 blocks_per_cg * (ncg - 1);
326 if (blocks_lastcg < minblocks_per_cg) {
328 * Since we make all the cylinder groups the same size, the
329 * last will only be small if there are more than one
330 * cylinder groups. If the last one is too small to store
331 * filesystem data, just kill it.
333 * XXX: Does fsck_ext2fs(8) properly handle this case?
335 bcount -= blocks_lastcg;
336 ncg--;
337 blocks_lastcg = blocks_per_cg;
338 blocks_gd = howmany(sizeof(struct ext2_gd) * ncg, bsize);
339 inodes_per_cg = num_inodes / ncg;
341 /* roundup inodes_per_cg to make it use whole inode table blocks */
342 inodes_per_cg = roundup(inodes_per_cg, sblock.e2fs_ipb);
343 num_inodes = inodes_per_cg * ncg;
344 iblocks_per_cg = inodes_per_cg / sblock.e2fs_ipb;
346 /* XXX: probably we should check these adjusted values again */
348 sblock.e2fs.e2fs_bcount = bcount;
349 sblock.e2fs.e2fs_icount = num_inodes;
351 sblock.e2fs_ncg = ncg;
352 sblock.e2fs_ngdb = blocks_gd;
353 sblock.e2fs_itpg = iblocks_per_cg;
355 sblock.e2fs.e2fs_rbcount = sblock.e2fs.e2fs_bcount * minfree / 100;
356 /* e2fs_fbcount will be accounted later */
357 /* e2fs_ficount will be accounted later */
359 sblock.e2fs.e2fs_bpg = blocks_per_cg;
360 sblock.e2fs.e2fs_fpg = blocks_per_cg;
362 sblock.e2fs.e2fs_ipg = inodes_per_cg;
364 sblock.e2fs.e2fs_mtime = 0;
365 sblock.e2fs.e2fs_wtime = tv.tv_sec;
366 sblock.e2fs.e2fs_mnt_count = 0;
367 /* XXX: should add some entropy to avoid checking all fs at once? */
368 sblock.e2fs.e2fs_max_mnt_count = EXT2_DEF_MAX_MNT_COUNT;
370 sblock.e2fs.e2fs_magic = E2FS_MAGIC;
371 sblock.e2fs.e2fs_state = E2FS_ISCLEAN;
372 sblock.e2fs.e2fs_beh = E2FS_BEH_DEFAULT;
373 sblock.e2fs.e2fs_minrev = 0;
374 sblock.e2fs.e2fs_lastfsck = tv.tv_sec;
375 sblock.e2fs.e2fs_fsckintv = EXT2_DEF_FSCKINTV;
378 * Maybe we can use E2FS_OS_FREEBSD here and it would be more proper,
379 * but the purpose of this newfs_ext2fs(8) command is to provide
380 * a filesystem which can be recognized by firmware on some
381 * Linux based appliances that can load bootstrap files only from
382 * (their native) ext2fs, and anyway we will (and should) try to
383 * act like them as much as possible.
385 * Anyway, I hope that all newer such boxes will keep their support
386 * for the "GOOD_OLD_REV" ext2fs.
388 sblock.e2fs.e2fs_creator = E2FS_OS_LINUX;
390 if (Oflag == 0) {
391 sblock.e2fs.e2fs_rev = E2FS_REV0;
392 sblock.e2fs.e2fs_features_compat = 0;
393 sblock.e2fs.e2fs_features_incompat = 0;
394 sblock.e2fs.e2fs_features_rocompat = 0;
395 } else {
396 sblock.e2fs.e2fs_rev = E2FS_REV1;
398 * e2fsprogs say "REV1" is "dynamic" so
399 * it isn't quite a version and maybe it means
400 * "extended from REV0 so check compat features."
402 * XXX: We don't have any native tool to activate
403 * the EXT2F_COMPAT_RESIZE feature and
404 * fsck_ext2fs(8) might not fix structures for it.
406 sblock.e2fs.e2fs_features_compat = EXT2F_COMPAT_RESIZE;
407 sblock.e2fs.e2fs_features_incompat = EXT2F_INCOMPAT_FTYPE;
408 sblock.e2fs.e2fs_features_rocompat =
409 EXT2F_ROCOMPAT_SPARSESUPER | EXT2F_ROCOMPAT_LARGEFILE;
412 sblock.e2fs.e2fs_ruid = geteuid();
413 sblock.e2fs.e2fs_rgid = getegid();
415 sblock.e2fs.e2fs_first_ino = EXT2_FIRSTINO;
416 sblock.e2fs.e2fs_inode_size = inodesize;
418 /* e2fs_block_group_nr is set on writing superblock to each group */
420 uuid_create(&uuid, &uustat);
421 if (uustat != uuid_s_ok)
422 errx(EXIT_FAILURE, "Failed to generate uuid\n");
423 uuid_enc_le(sblock.e2fs.e2fs_uuid, &uuid);
424 if (volname != NULL) {
425 if (strlen(volname) > sizeof(sblock.e2fs.e2fs_vname))
426 errx(EXIT_FAILURE, "Volume name is too long");
427 strlcpy(sblock.e2fs.e2fs_vname, volname,
428 sizeof(sblock.e2fs.e2fs_vname));
431 sblock.e2fs.e2fs_fsmnt[0] = '\0';
432 sblock.e2fs_fsmnt[0] = '\0';
434 sblock.e2fs.e2fs_algo = 0; /* XXX unsupported? */
435 sblock.e2fs.e2fs_prealloc = 0; /* XXX unsupported? */
436 sblock.e2fs.e2fs_dir_prealloc = 0; /* XXX unsupported? */
438 /* calculate blocks for reserved group descriptors for resize */
439 sblock.e2fs.e2fs_reserved_ngdb = 0;
440 if (sblock.e2fs.e2fs_rev > E2FS_REV0 &&
441 (sblock.e2fs.e2fs_features_compat & EXT2F_COMPAT_RESIZE) != 0) {
442 uint64_t target_blocks;
443 uint target_ncg, target_ngdb, reserved_ngdb;
445 /* reserve descriptors for size as 1024 times as current */
446 target_blocks =
447 (sblock.e2fs.e2fs_bcount - sblock.e2fs.e2fs_first_dblock)
448 * 1024ULL;
449 /* number of blocks must be in uint32_t */
450 if (target_blocks > UINT32_MAX)
451 target_blocks = UINT32_MAX;
452 target_ncg = howmany(target_blocks, sblock.e2fs.e2fs_bpg);
453 target_ngdb = howmany(sizeof(struct ext2_gd) * target_ncg,
454 sblock.e2fs_bsize);
456 * Reserved group descriptor blocks are preserved as
457 * the second level double indirect reference blocks in
458 * the EXT2_RESIZEINO inode, so the maximum number of
459 * the blocks is NINDIR(fs).
460 * (see also descriptions in init_resizeino() function)
462 * We check a number including current e2fs_ngdb here
463 * because they will be moved into reserved gdb on
464 * possible future size shrink, though e2fsprogs don't
465 * seem to care about it.
467 if (target_ngdb > NINDIR(&sblock))
468 target_ngdb = NINDIR(&sblock);
470 reserved_ngdb = target_ngdb - sblock.e2fs_ngdb;
472 /* make sure reserved_ngdb fits in the last cg */
473 if (reserved_ngdb >= blocks_lastcg - cgoverhead(ncg - 1))
474 reserved_ngdb = blocks_lastcg - cgoverhead(ncg - 1);
475 if (reserved_ngdb == 0) {
476 /* if no space for reserved gdb, disable the feature */
477 sblock.e2fs.e2fs_features_compat &=
478 ~EXT2F_COMPAT_RESIZE;
480 sblock.e2fs.e2fs_reserved_ngdb = reserved_ngdb;
484 * Initialize group descriptors
486 gd = malloc(sblock.e2fs_ngdb * bsize);
487 if (gd == NULL)
488 errx(EXIT_FAILURE, "Can't allocate descriptors buffer");
489 memset(gd, 0, sblock.e2fs_ngdb * bsize);
491 fbcount = 0;
492 ficount = 0;
493 for (cylno = 0; cylno < ncg; cylno++) {
494 uint boffset;
496 boffset = cgbase(&sblock, cylno);
497 if (sblock.e2fs.e2fs_rev == E2FS_REV0 ||
498 (sblock.e2fs.e2fs_features_rocompat &
499 EXT2F_ROCOMPAT_SPARSESUPER) == 0 ||
500 cg_has_sb(cylno)) {
501 boffset += NBLOCK_SUPERBLOCK + sblock.e2fs_ngdb;
502 if (sblock.e2fs.e2fs_rev > E2FS_REV0 &&
503 (sblock.e2fs.e2fs_features_compat &
504 EXT2F_COMPAT_RESIZE) != 0)
505 boffset += sblock.e2fs.e2fs_reserved_ngdb;
507 gd[cylno].ext2bgd_b_bitmap = boffset;
508 boffset += NBLOCK_BLOCK_BITMAP;
509 gd[cylno].ext2bgd_i_bitmap = boffset;
510 boffset += NBLOCK_INODE_BITMAP;
511 gd[cylno].ext2bgd_i_tables = boffset;
512 if (cylno == (ncg - 1))
513 gd[cylno].ext2bgd_nbfree =
514 blocks_lastcg - cgoverhead(cylno);
515 else
516 gd[cylno].ext2bgd_nbfree =
517 sblock.e2fs.e2fs_bpg - cgoverhead(cylno);
518 fbcount += gd[cylno].ext2bgd_nbfree;
519 gd[cylno].ext2bgd_nifree = sblock.e2fs.e2fs_ipg;
520 if (cylno == 0) {
521 /* take reserved inodes off nifree */
522 gd[cylno].ext2bgd_nifree -= EXT2_RESERVED_INODES;
524 ficount += gd[cylno].ext2bgd_nifree;
525 gd[cylno].ext2bgd_ndirs = 0;
527 sblock.e2fs.e2fs_fbcount = fbcount;
528 sblock.e2fs.e2fs_ficount = ficount;
531 * Dump out summary information about file system.
533 if (verbosity > 0) {
534 printf("%s: %u.%1uMB (%" PRId64 " sectors) "
535 "block size %u, fragment size %u\n",
536 fsys,
537 (uint)(((uint64_t)bcount * bsize) / (1024 * 1024)),
538 (uint)((uint64_t)bcount * bsize -
539 rounddown((uint64_t)bcount * bsize, 1024 * 1024))
540 / 1024 / 100,
541 fssize, bsize, fsize);
542 printf("\tusing %u block groups of %u.0MB, %u blks, "
543 "%u inodes.\n",
544 ncg, bsize * sblock.e2fs.e2fs_bpg / (1024 * 1024),
545 sblock.e2fs.e2fs_bpg, sblock.e2fs.e2fs_ipg);
549 * allocate space for superblock and group descriptors
551 iobufsize = (NBLOCK_SUPERBLOCK + sblock.e2fs_ngdb) * sblock.e2fs_bsize;
552 iobuf = mmap(0, iobufsize, PROT_READ|PROT_WRITE,
553 MAP_ANON|MAP_PRIVATE, -1, 0);
554 if (iobuf == NULL)
555 errx(EXIT_FAILURE, "Cannot allocate I/O buffer\n");
556 memset(iobuf, 0, iobufsize);
559 * We now start writing to the filesystem
562 if (!Nflag) {
563 static const uint pbsize[] = { 1024, 2048, 4096, 0 };
564 uint pblock, epblock;
566 * Validate the given file system size.
567 * Verify that its last block can actually be accessed.
568 * Convert to file system fragment sized units.
570 if (fssize <= 0)
571 errx(EXIT_FAILURE, "Preposterous size %" PRId64 "\n",
572 fssize);
573 wtfs(fssize - 1, sectorsize, iobuf);
576 * Ensure there is nothing that looks like a filesystem
577 * superblock anywhere other than where ours will be.
578 * If fsck_ext2fs finds the wrong one all hell breaks loose!
580 * XXX: needs to check how fsck_ext2fs programs even
581 * on other OSes determine alternate superblocks
583 for (i = 0; pbsize[i] != 0; i++) {
584 epblock = (uint64_t)bcount * bsize / pbsize[i];
585 for (pblock = ((pbsize[i] == SBSIZE) ? 1 : 0);
586 pblock < epblock;
587 pblock += pbsize[i] * NBBY /* bpg */)
588 zap_old_sblock((daddr_t)pblock *
589 pbsize[i] / sectorsize);
593 if (verbosity >= 3)
594 printf("super-block backups (for fsck_ext2fs -b #) at:\n");
595 /* If we are printing more than one line of numbers, line up columns */
596 fld_width = verbosity < 4 ? 1 : snprintf(NULL, 0, "%" PRIu64,
597 (uint64_t)cgbase(&sblock, ncg - 1));
598 /* Get terminal width */
599 if (ioctl(fileno(stdout), TIOCGWINSZ, &winsize) == 0)
600 max_cols = winsize.ws_col;
601 else
602 max_cols = 80;
603 if (Nflag && verbosity == 3)
604 /* Leave space to add " ..." after one row of numbers */
605 max_cols -= 4;
606 #define BASE 0x10000 /* For some fixed-point maths */
607 col = 0;
608 delta = verbosity > 2 ? 0 : max_cols * BASE / ncg;
609 for (cylno = 0; cylno < ncg; cylno++) {
610 fflush(stdout);
611 initcg(cylno);
612 if (verbosity < 2)
613 continue;
614 /* the first one is a master, not backup */
615 if (cylno == 0)
616 continue;
617 /* skip if this cylinder doesn't have a backup */
618 if (sblock.e2fs.e2fs_rev > E2FS_REV0 &&
619 (sblock.e2fs.e2fs_features_rocompat &
620 EXT2F_ROCOMPAT_SPARSESUPER) != 0 &&
621 cg_has_sb(cylno) == 0)
622 continue;
624 if (delta > 0) {
625 if (Nflag)
626 /* No point doing dots for -N */
627 break;
628 /* Print dots scaled to end near RH margin */
629 for (col += delta; col > BASE; col -= BASE)
630 printf(".");
631 continue;
633 /* Print superblock numbers */
634 len = printf(" %*" PRIu64 "," + !col, fld_width,
635 (uint64_t)cgbase(&sblock, cylno));
636 col += len;
637 if (col + len < max_cols)
638 /* Next number fits */
639 continue;
640 /* Next number won't fit, need a newline */
641 if (verbosity <= 3) {
642 /* Print dots for subsequent cylinder groups */
643 delta = sblock.e2fs_ncg - cylno - 1;
644 if (delta != 0) {
645 if (Nflag) {
646 printf(" ...");
647 break;
649 delta = max_cols * BASE / delta;
652 col = 0;
653 printf("\n");
655 #undef BASE
656 if (col > 0)
657 printf("\n");
658 if (Nflag)
659 return;
662 * Now construct the initial file system,
664 if (fsinit(&tv) == 0)
665 errx(EXIT_FAILURE, "Error making filesystem");
667 * Write out the superblock and group descriptors
669 sblock.e2fs.e2fs_block_group_nr = 0;
670 sboff = 0;
671 if (cgbase(&sblock, 0) == 0) {
673 * If the first block contains the boot block sectors,
674 * (i.e. in case of sblock.e2fs.e2fs_bsize > BBSIZE)
675 * we have to preserve data in it.
677 sboff = SBOFF;
679 e2fs_sbsave(&sblock.e2fs, (struct ext2fs *)(iobuf + sboff));
680 e2fs_cgsave(gd, (struct ext2_gd *)(iobuf + sblock.e2fs_bsize),
681 sizeof(struct ext2_gd) * sblock.e2fs_ncg);
682 wtfs(fsbtodb(&sblock, cgbase(&sblock, 0)) + sboff / sectorsize,
683 iobufsize - sboff, iobuf + sboff);
685 munmap(iobuf, iobufsize);
689 * Initialize a cylinder (block) group.
691 void
692 initcg(uint cylno)
694 uint nblcg, i, j, sboff;
695 struct ext2fs_dinode *dp;
698 * Make a copy of the superblock and group descriptors.
700 if (sblock.e2fs.e2fs_rev == E2FS_REV0 ||
701 (sblock.e2fs.e2fs_features_rocompat &
702 EXT2F_ROCOMPAT_SPARSESUPER) == 0 ||
703 cg_has_sb(cylno)) {
704 sblock.e2fs.e2fs_block_group_nr = cylno;
705 sboff = 0;
706 if (cgbase(&sblock, cylno) == 0) {
707 /* preserve data in bootblock in cg0 */
708 sboff = SBOFF;
710 e2fs_sbsave(&sblock.e2fs, (struct ext2fs *)(iobuf + sboff));
711 e2fs_cgsave(gd, (struct ext2_gd *)(iobuf +
712 sblock.e2fs_bsize * NBLOCK_SUPERBLOCK),
713 sizeof(struct ext2_gd) * sblock.e2fs_ncg);
714 /* write superblock and group descriptor backups */
715 wtfs(fsbtodb(&sblock, cgbase(&sblock, cylno)) +
716 sboff / sectorsize, iobufsize - sboff, iobuf + sboff);
720 * Initialize block bitmap.
722 memset(buf, 0, sblock.e2fs_bsize);
723 if (cylno == (sblock.e2fs_ncg - 1)) {
724 /* The last group could have less blocks than e2fs_bpg. */
725 nblcg = sblock.e2fs.e2fs_bcount -
726 cgbase(&sblock, sblock.e2fs_ncg - 1);
727 for (i = nblcg; i < roundup(nblcg, NBBY); i++)
728 setbit(buf, i);
729 memset(&buf[i / NBBY], ~0U, sblock.e2fs.e2fs_bpg - i);
731 /* set overhead (superblock, group descriptor etc.) blocks used */
732 for (i = 0; i < cgoverhead(cylno) / NBBY; i++)
733 buf[i] = ~0;
734 i = i * NBBY;
735 for (; i < cgoverhead(cylno); i++)
736 setbit(buf, i);
737 wtfs(fsbtodb(&sblock, gd[cylno].ext2bgd_b_bitmap), sblock.e2fs_bsize,
738 buf);
741 * Initialize inode bitmap.
743 * Assume e2fs_ipg is a multiple of NBBY since
744 * it's a multiple of e2fs_ipb (as we did above).
745 * Note even (possibly smaller) the last group has the same e2fs_ipg.
747 assert(!(sblock.e2fs.e2fs_ipg % NBBY));
748 i = sblock.e2fs.e2fs_ipg / NBBY;
749 memset(buf, 0, i);
750 assert(sblock.e2fs_bsize >= i);
751 memset(buf + i, ~0U, sblock.e2fs_bsize - i);
752 if (cylno == 0) {
753 /* mark reserved inodes */
754 for (i = 1; i < EXT2_FIRSTINO; i++)
755 setbit(buf, EXT2_INO_INDEX(i));
757 wtfs(fsbtodb(&sblock, gd[cylno].ext2bgd_i_bitmap), sblock.e2fs_bsize,
758 buf);
761 * Initialize inode tables.
763 * Just initialize generation numbers for NFS security.
764 * XXX: sys/ufs/ext2fs/ext2fs_alloc.c:ext2fs_valloc() seems
765 * to override these generated numbers.
767 memset(buf, 0, sblock.e2fs_bsize);
768 for (i = 0; i < sblock.e2fs_itpg; i++) {
769 for (j = 0; j < sblock.e2fs_ipb; j++) {
770 dp = (struct ext2fs_dinode *)(buf + inodesize * j);
771 /* h2fs32() just for consistency */
772 dp->e2di_gen = h2fs32(arc4random());
774 wtfs(fsbtodb(&sblock, gd[cylno].ext2bgd_i_tables + i),
775 sblock.e2fs_bsize, buf);
780 * Zap possible lingering old superblock data
782 static void
783 zap_old_sblock(daddr_t sec)
785 static daddr_t cg0_data;
786 uint32_t oldfs[SBSIZE / sizeof(uint32_t)];
787 static const struct fsm {
788 uint32_t offset;
789 uint32_t magic;
790 uint32_t mask;
791 } fs_magics[] = {
792 {offsetof(struct ext2fs, e2fs_magic) / 4, E2FS_MAGIC, 0xffff},
793 {offsetof(struct ext2fs, e2fs_magic) / 4,
794 E2FS_MAGIC << 16, 0xffff0000},
795 {14, 0xef530000, 0xffff0000}, /* EXT2FS (big) */
796 {0x55c / 4, 0x00011954, ~0U}, /* FS_UFS1_MAGIC */
797 {0x55c / 4, 0x19540119, ~0U}, /* FS_UFS2_MAGIC */
798 {0, 0x70162, ~0U}, /* LFS_MAGIC */
799 {.offset = ~0U},
801 const struct fsm *fsm;
803 if (Nflag)
804 return;
806 /* don't override data before superblock */
807 if (sec < SBOFF / sectorsize)
808 return;
810 if (cg0_data == 0) {
811 cg0_data =
812 ((daddr_t)sblock.e2fs.e2fs_first_dblock + cgoverhead(0)) *
813 sblock.e2fs_bsize / sectorsize;
816 /* Ignore anything that is beyond our filesystem */
817 if (sec >= fssize)
818 return;
819 /* Zero anything inside our filesystem... */
820 if (sec >= sblock.e2fs.e2fs_first_dblock * bsize / sectorsize) {
821 /* ...unless we will write that area anyway */
822 if (sec >= cg0_data)
823 /* assume iobuf is zero'ed here */
824 wtfs(sec, roundup(SBSIZE, sectorsize), iobuf);
825 return;
829 * The sector might contain boot code, so we must validate it
831 * XXX: ext2fs won't preserve data after SBOFF,
832 * but first_dblock could have a different value.
834 rdfs(sec, sizeof(oldfs), &oldfs);
835 for (fsm = fs_magics;; fsm++) {
836 uint32_t v;
837 if (fsm->mask == 0)
838 return;
839 v = oldfs[fsm->offset];
840 if ((v & fsm->mask) == fsm->magic ||
841 (bswap32(v) & fsm->mask) == fsm->magic)
842 break;
845 /* Just zap the magic number */
846 oldfs[fsm->offset] = 0;
847 wtfs(sec, sizeof(oldfs), &oldfs);
851 * uint cgoverhead(uint c)
853 * Return a number of reserved blocks on the specified group.
854 * XXX: should be shared with src/sbin/fsck_ext2fs/setup.c
856 uint
857 cgoverhead(uint c)
859 uint overh;
861 overh = NBLOCK_BLOCK_BITMAP + NBLOCK_INODE_BITMAP + sblock.e2fs_itpg;
863 if (sblock.e2fs.e2fs_rev == E2FS_REV0 ||
864 (sblock.e2fs.e2fs_features_rocompat &
865 EXT2F_ROCOMPAT_SPARSESUPER) == 0 ||
866 cg_has_sb(c) != 0) {
867 overh += NBLOCK_SUPERBLOCK + sblock.e2fs_ngdb;
869 if (sblock.e2fs.e2fs_rev > E2FS_REV0 &&
870 (sblock.e2fs.e2fs_features_compat &
871 EXT2F_COMPAT_RESIZE) != 0)
872 overh += sblock.e2fs.e2fs_reserved_ngdb;
875 return overh;
879 * Initialize the file system
882 #define LOSTDIR /* e2fsck complains if there is no lost+found */
884 #define PREDEFDIR 2
886 #ifdef LOSTDIR
887 #define PREDEFROOTDIR (PREDEFDIR + 1)
888 #else
889 #define PREDEFROOTDIR PREDEFDIR
890 #endif
892 struct ext2fs_direct root_dir[] = {
893 { EXT2_ROOTINO, 0, 1, 0, "." },
894 { EXT2_ROOTINO, 0, 2, 0, ".." },
895 #ifdef LOSTDIR
896 { EXT2_LOSTFOUNDINO, 0, 10, 0, "lost+found" },
897 #endif
900 #ifdef LOSTDIR
901 struct ext2fs_direct lost_found_dir[] = {
902 { EXT2_LOSTFOUNDINO, 0, 1, 0, "." },
903 { EXT2_ROOTINO, 0, 2, 0, ".." },
905 struct ext2fs_direct pad_dir = { 0, sizeof(struct ext2fs_direct), 0, 0, "" };
906 #endif
909 fsinit(const struct timeval *tv)
911 struct ext2fs_dinode node;
912 #ifdef LOSTDIR
913 uint i, nblks_lostfound, blk;
914 #endif
917 * Initialize the inode for the resizefs feature
919 if (sblock.e2fs.e2fs_rev > E2FS_REV0 &&
920 (sblock.e2fs.e2fs_features_compat & EXT2F_COMPAT_RESIZE) != 0)
921 init_resizeino(tv);
924 * Initialize the node
927 #ifdef LOSTDIR
929 * Create the lost+found directory
931 if (sblock.e2fs.e2fs_rev > E2FS_REV0 &&
932 sblock.e2fs.e2fs_features_incompat & EXT2F_INCOMPAT_FTYPE) {
933 lost_found_dir[0].e2d_type = EXT2_FT_DIR;
934 lost_found_dir[1].e2d_type = EXT2_FT_DIR;
936 (void)makedir(lost_found_dir, __arraycount(lost_found_dir));
938 /* prepare a bit large directory for preserved files */
939 nblks_lostfound = EXT2_LOSTFOUNDSIZE / sblock.e2fs_bsize;
940 /* ...but only with direct blocks */
941 if (nblks_lostfound > NDADDR)
942 nblks_lostfound = NDADDR;
944 memset(&node, 0, sizeof(node));
945 node.e2di_mode = EXT2_IFDIR | EXT2_LOSTFOUNDUMASK;
946 node.e2di_uid = geteuid();
947 node.e2di_size = sblock.e2fs_bsize * nblks_lostfound;
948 node.e2di_atime = tv->tv_sec;
949 node.e2di_ctime = tv->tv_sec;
950 node.e2di_mtime = tv->tv_sec;
951 node.e2di_gid = getegid();
952 node.e2di_nlink = PREDEFDIR;
953 /* e2di_nblock is a number of disk blocks, not ext2fs blocks */
954 node.e2di_nblock = fsbtodb(&sblock, nblks_lostfound);
955 node.e2di_blocks[0] = alloc(sblock.e2fs_bsize, node.e2di_mode);
956 if (node.e2di_blocks[0] == 0) {
957 printf("%s: can't allocate block for lost+found\n", __func__);
958 return 0;
960 for (i = 1; i < nblks_lostfound; i++) {
961 blk = alloc(sblock.e2fs_bsize, 0);
962 if (blk == 0) {
963 printf("%s: can't allocate blocks for lost+found\n",
964 __func__);
965 return 0;
967 node.e2di_blocks[i] = blk;
969 wtfs(fsbtodb(&sblock, node.e2di_blocks[0]), sblock.e2fs_bsize, buf);
970 pad_dir.e2d_reclen = sblock.e2fs_bsize;
971 for (i = 1; i < nblks_lostfound; i++) {
972 memset(buf, 0, sblock.e2fs_bsize);
973 copy_dir(&pad_dir, (struct ext2fs_direct *)buf);
974 wtfs(fsbtodb(&sblock, node.e2di_blocks[i]), sblock.e2fs_bsize,
975 buf);
977 iput(&node, EXT2_LOSTFOUNDINO);
978 #endif
980 * create the root directory
982 memset(&node, 0, sizeof(node));
983 if (sblock.e2fs.e2fs_rev > E2FS_REV0 &&
984 sblock.e2fs.e2fs_features_incompat & EXT2F_INCOMPAT_FTYPE) {
985 root_dir[0].e2d_type = EXT2_FT_DIR;
986 root_dir[1].e2d_type = EXT2_FT_DIR;
987 #ifdef LOSTDIR
988 root_dir[2].e2d_type = EXT2_FT_DIR;
989 #endif
991 node.e2di_mode = EXT2_IFDIR | EXT2_UMASK;
992 node.e2di_uid = geteuid();
993 node.e2di_size = makedir(root_dir, __arraycount(root_dir));
994 node.e2di_atime = tv->tv_sec;
995 node.e2di_ctime = tv->tv_sec;
996 node.e2di_mtime = tv->tv_sec;
997 node.e2di_gid = getegid();
998 node.e2di_nlink = PREDEFROOTDIR;
999 /* e2di_nblock is a number of disk block, not ext2fs block */
1000 node.e2di_nblock = fsbtodb(&sblock, 1);
1001 node.e2di_blocks[0] = alloc(node.e2di_size, node.e2di_mode);
1002 if (node.e2di_blocks[0] == 0) {
1003 printf("%s: can't allocate block for root dir\n", __func__);
1004 return 0;
1006 wtfs(fsbtodb(&sblock, node.e2di_blocks[0]), sblock.e2fs_bsize, buf);
1007 iput(&node, EXT2_ROOTINO);
1008 return 1;
1012 * Construct a set of directory entries in "buf".
1013 * return size of directory.
1016 makedir(struct ext2fs_direct *protodir, int entries)
1018 uint8_t *cp;
1019 uint i, spcleft;
1020 uint dirblksiz;
1022 dirblksiz = sblock.e2fs_bsize;
1023 memset(buf, 0, dirblksiz);
1024 spcleft = dirblksiz;
1025 for (cp = buf, i = 0; i < entries - 1; i++) {
1026 protodir[i].e2d_reclen = EXT2FS_DIRSIZ(protodir[i].e2d_namlen);
1027 copy_dir(&protodir[i], (struct ext2fs_direct *)cp);
1028 cp += protodir[i].e2d_reclen;
1029 spcleft -= protodir[i].e2d_reclen;
1031 protodir[i].e2d_reclen = spcleft;
1032 copy_dir(&protodir[i], (struct ext2fs_direct *)cp);
1033 return dirblksiz;
1037 * Copy a direntry to a buffer, in fs byte order
1039 static void
1040 copy_dir(struct ext2fs_direct *dir, struct ext2fs_direct *dbuf)
1043 memcpy(dbuf, dir, EXT2FS_DIRSIZ(dir->e2d_namlen));
1044 dbuf->e2d_ino = h2fs32(dir->e2d_ino);
1045 dbuf->e2d_reclen = h2fs16(dir->e2d_reclen);
1049 * void init_resizeino(const struct timeval *tv);
1051 * Initialize the EXT2_RESEIZE_INO inode to preserve
1052 * reserved group descriptor blocks for future growth of this ext2fs.
1054 void
1055 init_resizeino(const struct timeval *tv)
1057 struct ext2fs_dinode node;
1058 uint64_t isize;
1059 uint32_t *dindir_block, *reserved_gdb;
1060 uint nblock, i, cylno, n;
1062 memset(&node, 0, sizeof(node));
1065 * Note this function only prepares required structures for
1066 * future resize. It's a quite different work to implement
1067 * a utility like resize_ext2fs(8) which handles actual
1068 * resize ops even on offline.
1070 * Anyway, I'm not sure if there is any documentation about
1071 * this resize ext2fs feature and related data structures,
1072 * and I've written this function based on things what I see
1073 * on some existing implementation and real file system data
1074 * created by existing tools. To be honest, they are not
1075 * so easy to read, so I will try to implement it here without
1076 * any dumb optimization for people who would eventually
1077 * work on "yet another wheel" like resize_ext2fs(8).
1081 * I'm not sure what type is appropriate for this inode.
1082 * The release notes of e2fsprogs says they changed e2fsck to allow
1083 * IFREG for RESIZEINO since a certain resize tool used it. Hmm.
1085 node.e2di_mode = EXT2_IFREG | EXT2_RESIZEINOUMASK;
1086 node.e2di_uid = geteuid();
1087 node.e2di_atime = tv->tv_sec;
1088 node.e2di_ctime = tv->tv_sec;
1089 node.e2di_mtime = tv->tv_sec;
1090 node.e2di_gid = getegid();
1091 node.e2di_nlink = 1;
1094 * To preserve the reserved group descriptor blocks,
1095 * EXT2_RESIZEINO uses only double indirect reference
1096 * blocks in its inode entries.
1098 * All entries for direct, single indirect and triple
1099 * indirect references are left zero'ed. Maybe it's safe
1100 * because no write operation will happen with this inode.
1102 * We have to allocate a block for the first level double
1103 * indirect reference block. Indexes of inode entries in
1104 * this first level dindirect block are corresponding to
1105 * indexes of group descriptors including both used (e2fs_ngdb)
1106 * and reserved (e2fs_reserved_ngdb) group descriptor blocks.
1108 * Inode entries of indexes for used (e2fs_ngdb) descriptors are
1109 * left zero'ed. Entries for reserved (e2fs_reserved_ngdb) ones
1110 * have block numbers of actual reserved group descriptors
1111 * allocated at block group zero. This means e2fs_reserved_ngdb
1112 * blocks are reserved as the second level dindirect reference
1113 * blocks, and they actually contain block numbers of indirect
1114 * references. It may be safe since they don't have to keep any
1115 * data yet.
1117 * Each these second dindirect blocks (i.e. reserved group
1118 * descriptor blocks in the first block group) should have
1119 * block numbers of its backups in all other block groups.
1120 * I.e. reserved_ngdb[0] block in block group 0 contains block
1121 * numbers of resreved_ngdb[0] from group 1 through (e2fs_ncg - 1).
1122 * The number of backups can be determined by the
1123 * EXT2_ROCOMPAT_SPARSESUPER feature and cg_has_sb() macro
1124 * as done in the above initcg() function.
1127 /* set e2di_size which occupies whole blocks through DINDIR blocks */
1128 isize = (uint64_t)sblock.e2fs_bsize * NDADDR +
1129 (uint64_t)sblock.e2fs_bsize * NINDIR(&sblock) +
1130 (uint64_t)sblock.e2fs_bsize * NINDIR(&sblock) * NINDIR(&sblock);
1131 if (isize > UINT32_MAX &&
1132 (sblock.e2fs.e2fs_features_rocompat &
1133 EXT2F_ROCOMPAT_LARGEFILE) == 0) {
1134 /* XXX should enable it here and update all backups? */
1135 errx(EXIT_FAILURE, "%s: large_file rocompat feature is "
1136 "required to enable resize feature for this filesystem\n",
1137 __func__);
1139 /* upper 32bit is stored into e2di_dacl on REV1 feature */
1140 node.e2di_size = isize & UINT32_MAX;
1141 node.e2di_dacl = isize >> 32;
1143 #define SINGLE 0 /* index of single indirect block */
1144 #define DOUBLE 1 /* index of double indirect block */
1145 #define TRIPLE 2 /* index of triple indirect block */
1147 /* zero out entries for direct references */
1148 for (i = 0; i < NDADDR; i++)
1149 node.e2di_blocks[i] = 0;
1150 /* also zero out entries for single and triple indirect references */
1151 node.e2di_blocks[NDADDR + SINGLE] = 0;
1152 node.e2di_blocks[NDADDR + TRIPLE] = 0;
1154 /* allocate a block for the first level double indirect reference */
1155 node.e2di_blocks[NDADDR + DOUBLE] =
1156 alloc(sblock.e2fs_bsize, node.e2di_mode);
1157 if (node.e2di_blocks[NDADDR + DOUBLE] == 0)
1158 errx(EXIT_FAILURE, "%s: Can't allocate a dindirect block",
1159 __func__);
1161 /* account this first block */
1162 nblock = fsbtodb(&sblock, 1);
1164 /* allocate buffer to set data in the dindirect block */
1165 dindir_block = malloc(sblock.e2fs_bsize);
1166 if (dindir_block == NULL)
1167 errx(EXIT_FAILURE,
1168 "%s: Can't allocate buffer for a dindirect block",
1169 __func__);
1171 /* allocate buffer to set data in the group descriptor blocks */
1172 reserved_gdb = malloc(sblock.e2fs_bsize);
1173 if (reserved_gdb == NULL)
1174 errx(EXIT_FAILURE,
1175 "%s: Can't allocate buffer for group descriptor blocks",
1176 __func__);
1179 * Setup block entries in the first level dindirect blocks
1181 for (i = 0; i < sblock.e2fs_ngdb; i++) {
1182 /* no need to handle used group descriptor blocks */
1183 dindir_block[i] = 0;
1185 for (; i < sblock.e2fs_ngdb + sblock.e2fs.e2fs_reserved_ngdb; i++) {
1187 * point reserved group descriptor block in the first
1188 * (i.e. master) block group
1190 * XXX: e2fsprogs seem to use "(i % NINDIR(&sblock))" here
1191 * to store maximum NINDIR(&sblock) reserved gdbs.
1192 * I'm not sure what will be done on future filesystem
1193 * shrink in that case on their way.
1195 if (i >= NINDIR(&sblock))
1196 errx(EXIT_FAILURE, "%s: too many reserved "
1197 "group descriptors (%u) for resize inode",
1198 __func__, sblock.e2fs.e2fs_reserved_ngdb);
1199 dindir_block[i] =
1200 h2fs32(cgbase(&sblock, 0) + NBLOCK_SUPERBLOCK + i);
1203 * Setup block entries in the second dindirect blocks
1204 * (which are primary reserved group descriptor blocks)
1205 * to point their backups.
1207 for (n = 0, cylno = 1; cylno < sblock.e2fs_ncg; cylno++) {
1208 /* skip block groups without backup */
1209 if ((sblock.e2fs.e2fs_features_rocompat &
1210 EXT2F_ROCOMPAT_SPARSESUPER) != 0 &&
1211 cg_has_sb(cylno) == 0)
1212 continue;
1214 if (n >= NINDIR(&sblock))
1215 errx(EXIT_FAILURE, "%s: too many block groups "
1216 "for the resize feature", __func__);
1218 * These blocks are already reserved in
1219 * initcg() so no need to use alloc() here.
1221 reserved_gdb[n++] = h2fs32(cgbase(&sblock, cylno) +
1222 NBLOCK_SUPERBLOCK + i);
1223 nblock += fsbtodb(&sblock, 1);
1225 for (; n < NINDIR(&sblock); n++)
1226 reserved_gdb[n] = 0;
1228 /* write group descriptor block as the second dindirect refs */
1229 wtfs(fsbtodb(&sblock, fs2h32(dindir_block[i])),
1230 sblock.e2fs_bsize, reserved_gdb);
1231 nblock += fsbtodb(&sblock, 1);
1233 for (; i < NINDIR(&sblock); i++) {
1234 /* leave trailing entries unallocated */
1235 dindir_block[i] = 0;
1237 free(reserved_gdb);
1239 /* finally write the first level dindirect block */
1240 wtfs(fsbtodb(&sblock, node.e2di_blocks[NDADDR + DOUBLE]),
1241 sblock.e2fs_bsize, dindir_block);
1242 free(dindir_block);
1244 node.e2di_nblock = nblock;
1245 iput(&node, EXT2_RESIZEINO);
1249 * uint32_t alloc(uint32_t size, uint16_t mode)
1251 * Allocate a block (from cylinder group 0)
1252 * Reference: src/sys/ufs/ext2fs/ext2fs_alloc.c:ext2fs_alloccg()
1254 uint32_t
1255 alloc(uint32_t size, uint16_t mode)
1257 uint32_t loc, bno;
1258 uint8_t *bbp;
1259 uint len, map, i;
1261 if (gd[0].ext2bgd_nbfree == 0)
1262 return 0;
1264 if (size > sblock.e2fs_bsize)
1265 return 0;
1267 bbp = malloc(sblock.e2fs_bsize);
1268 if (bbp == NULL)
1269 return 0;
1270 rdfs(fsbtodb(&sblock, gd[0].ext2bgd_b_bitmap), sblock.e2fs_bsize, bbp);
1272 /* XXX: kernel uses e2fs_fpg here */
1273 len = sblock.e2fs.e2fs_bpg / NBBY;
1275 #if 0 /* no need block allocation for root or lost+found dir */
1276 for (loc = 0; loc < len; loc++) {
1277 if (bbp[loc] == 0) {
1278 bno = loc * NBBY;
1279 goto gotit;
1282 #endif
1284 loc = skpc(~0U, len, bbp);
1285 if (loc == 0)
1286 return 0;
1287 loc = len - loc;
1288 map = bbp[loc];
1289 bno = loc * NBBY;
1290 for (i = 0; i < NBBY; i++, bno++) {
1291 if ((map & (1 << i)) == 0)
1292 goto gotit;
1294 return 0;
1296 gotit:
1297 if (isset(bbp, bno))
1298 errx(EXIT_FAILURE, "%s: inconsistent bitmap\n", __func__);
1300 setbit(bbp, bno);
1301 wtfs(fsbtodb(&sblock, gd[0].ext2bgd_b_bitmap), sblock.e2fs_bsize, bbp);
1302 free(bbp);
1303 /* XXX: modified group descriptors won't be written into backups */
1304 gd[0].ext2bgd_nbfree--;
1305 if ((mode & EXT2_IFDIR) != 0)
1306 gd[0].ext2bgd_ndirs++;
1307 sblock.e2fs.e2fs_fbcount--;
1309 return sblock.e2fs.e2fs_first_dblock + bno;
1313 * void iput(struct ext2fs_dinode *ip, ino_t ino)
1315 * Put an inode entry into the corresponding table.
1317 static void
1318 iput(struct ext2fs_dinode *ip, ino_t ino)
1320 daddr_t d;
1321 uint c, i;
1322 struct ext2fs_dinode *dp;
1323 uint8_t *bp;
1325 bp = malloc(sblock.e2fs_bsize);
1326 if (bp == NULL)
1327 errx(EXIT_FAILURE, "%s: can't allocate buffer for inode\n",
1328 __func__);
1331 * Reserved inodes are allocated and accounted in initcg()
1332 * so skip checks of the bitmap and allocation for them.
1334 if (ino >= EXT2_FIRSTINO) {
1335 c = ino_to_cg(&sblock, ino);
1337 /* sanity check */
1338 if (gd[c].ext2bgd_nifree == 0)
1339 errx(EXIT_FAILURE,
1340 "%s: no free inode %" PRIu64 " in block group %u\n",
1341 __func__, (uint64_t)ino, c);
1343 /* update inode bitmap */
1344 rdfs(fsbtodb(&sblock, gd[0].ext2bgd_i_bitmap),
1345 sblock.e2fs_bsize, bp);
1347 /* more sanity */
1348 if (isset(bp, EXT2_INO_INDEX(ino)))
1349 errx(EXIT_FAILURE, "%s: inode %" PRIu64
1350 " already in use\n", __func__, (uint64_t)ino);
1351 setbit(bp, EXT2_INO_INDEX(ino));
1352 wtfs(fsbtodb(&sblock, gd[0].ext2bgd_i_bitmap),
1353 sblock.e2fs_bsize, bp);
1354 gd[c].ext2bgd_nifree--;
1355 sblock.e2fs.e2fs_ficount--;
1358 if (ino >= sblock.e2fs.e2fs_ipg * sblock.e2fs_ncg)
1359 errx(EXIT_FAILURE, "%s: inode value out of range (%" PRIu64
1360 ").\n", __func__, (uint64_t)ino);
1362 /* update an inode entry in the table */
1363 d = fsbtodb(&sblock, ino_to_fsba(&sblock, ino));
1364 rdfs(d, sblock.e2fs_bsize, bp);
1366 dp = (struct ext2fs_dinode *)(bp +
1367 inodesize * ino_to_fsbo(&sblock, ino));
1368 e2fs_isave(ip, dp);
1369 /* e2fs_i_bswap() doesn't swap e2di_blocks addrs */
1370 if ((ip->e2di_mode & EXT2_IFMT) != EXT2_IFLNK) {
1371 for (i = 0; i < NDADDR + NIADDR; i++)
1372 dp->e2di_blocks[i] = h2fs32(ip->e2di_blocks[i]);
1374 /* h2fs32() just for consistency */
1375 dp->e2di_gen = h2fs32(arc4random());
1377 wtfs(d, sblock.e2fs_bsize, bp);
1378 free(bp);
1382 * Read a block from the file system
1384 void
1385 rdfs(daddr_t bno, int size, void *bf)
1387 int n;
1388 #ifndef __minix
1389 off_t offset;
1390 #else
1391 u64_t offset;
1392 #endif
1394 offset = bno;
1395 #ifndef __minix
1396 n = pread(fsi, bf, size, offset * sectorsize);
1397 #else
1398 n = pread64(fsi, bf, size, offset * sectorsize);
1399 #endif
1400 if (n != size)
1401 err(EXIT_FAILURE, "%s: read error for sector %" PRId64,
1402 __func__, (int64_t)bno);
1406 * Write a block to the file system
1408 void
1409 wtfs(daddr_t bno, int size, void *bf)
1411 int n;
1412 #ifndef __minix
1413 off_t offset;
1414 #else
1415 u64_t offset;
1416 #endif
1418 if (Nflag)
1419 return;
1420 offset = bno;
1421 errno = 0;
1422 #ifndef __minix
1423 n = pwrite(fso, bf, size, offset * sectorsize);
1424 #else
1425 n = pwrite64(fso, bf, size, offset * sectorsize);
1426 #endif
1427 if (n != size)
1428 err(EXIT_FAILURE, "%s: write error for sector %" PRId64,
1429 __func__, (int64_t)bno);
1433 ilog2(uint val)
1436 if (val == 0 || !powerof2(val))
1437 errx(EXIT_FAILURE, "%s: %u is not a power of 2\n",
1438 __func__, val);
1440 return ffs(val) - 1;
1444 * int skpc(int mask, size_t size, uint8_t *cp)
1446 * Locate an unsigned character of value mask inside cp[].
1447 * (from src/sys/lib/libkern/skpc.c)
1450 skpc(int mask, size_t size, uint8_t *cp)
1452 uint8_t *end;
1454 end = &cp[size];
1455 while (cp < end && *cp == (uint8_t)mask)
1456 cp++;
1458 return end - cp;