2 * modified for EXT2FS support in Lites 1.1
4 * Aug 1995, Godmar Back (gback@cs.utah.edu)
5 * University of Utah, Department of Computer Science
8 * SPDX-License-Identifier: BSD-3-Clause
10 * Copyright (c) 1989, 1991, 1993, 1994
11 * The Regents of the University of California. All rights reserved.
13 * Redistribution and use in source and binary forms, with or without
14 * modification, are permitted provided that the following conditions
16 * 1. Redistributions of source code must retain the above copyright
17 * notice, this list of conditions and the following disclaimer.
18 * 2. Redistributions in binary form must reproduce the above copyright
19 * notice, this list of conditions and the following disclaimer in the
20 * documentation and/or other materials provided with the distribution.
21 * 3. Neither the name of the University nor the names of its contributors
22 * may be used to endorse or promote products derived from this software
23 * without specific prior written permission.
25 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
37 * @(#)ffs_vfsops.c 8.8 (Berkeley) 4/18/94
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/namei.h>
46 #include <sys/kernel.h>
47 #include <sys/vnode.h>
48 #include <sys/mount.h>
52 #include <sys/endian.h>
53 #include <sys/fcntl.h>
54 #include <sys/malloc.h>
56 #include <sys/mutex.h>
58 #include <geom/geom.h>
59 #include <geom/geom_vfs.h>
61 #include <fs/ext2fs/ext2_mount.h>
62 #include <fs/ext2fs/inode.h>
64 #include <fs/ext2fs/fs.h>
65 #include <fs/ext2fs/ext2fs.h>
66 #include <fs/ext2fs/ext2_dinode.h>
67 #include <fs/ext2fs/ext2_extern.h>
69 static int ext2_flushfiles(struct mount *mp, int flags, struct thread *td);
70 static int ext2_mountfs(struct vnode *, struct mount *);
71 static int ext2_reload(struct mount *mp, struct thread *td);
72 static int ext2_sbupdate(struct ext2mount *, int);
73 static int ext2_cgupdate(struct ext2mount *, int);
74 static vfs_unmount_t ext2_unmount;
75 static vfs_root_t ext2_root;
76 static vfs_statfs_t ext2_statfs;
77 static vfs_sync_t ext2_sync;
78 static vfs_vget_t ext2_vget;
79 static vfs_fhtovp_t ext2_fhtovp;
80 static vfs_mount_t ext2_mount;
82 MALLOC_DEFINE(M_EXT2NODE, "ext2_node", "EXT2 vnode private part");
83 static MALLOC_DEFINE(M_EXT2MNT, "ext2_mount", "EXT2 mount structure");
85 static struct vfsops ext2fs_vfsops = {
86 .vfs_fhtovp = ext2_fhtovp,
87 .vfs_mount = ext2_mount,
88 .vfs_root = ext2_root, /* root inode via vget */
89 .vfs_statfs = ext2_statfs,
90 .vfs_sync = ext2_sync,
91 .vfs_unmount = ext2_unmount,
92 .vfs_vget = ext2_vget,
95 VFS_SET(ext2fs_vfsops, ext2fs, 0);
97 static int ext2_check_sb_compat(struct ext2fs *es, struct cdev *dev,
99 static int compute_sb_data(struct vnode * devvp,
100 struct ext2fs * es, struct m_ext2fs * fs);
102 static const char *ext2_opts[] = { "acls", "async", "noatime", "noclusterr",
103 "noclusterw", "noexec", "export", "force", "from", "multilabel",
104 "suiddir", "nosymfollow", "sync", "union", NULL };
112 ext2_mount(struct mount *mp)
114 struct vfsoptlist *opts;
117 struct ext2mount *ump = NULL;
119 struct nameidata nd, *ndp = &nd;
122 int error, flags, len;
125 opts = mp->mnt_optnew;
127 if (vfs_filteropt(opts, ext2_opts))
130 vfs_getopt(opts, "fspath", (void **)&path, NULL);
131 /* Double-check the length of path.. */
132 if (strlen(path) >= MAXMNTLEN)
133 return (ENAMETOOLONG);
136 error = vfs_getopt(opts, "from", (void **)&fspec, &len);
137 if (!error && fspec[len - 1] != '\0')
141 * If updating, check whether changing from read-only to
142 * read/write; if there is no device name, that's all we do.
144 if (mp->mnt_flag & MNT_UPDATE) {
148 if (fs->e2fs_ronly == 0 &&
149 vfs_flagopt(opts, "ro", NULL, 0)) {
150 error = VFS_SYNC(mp, MNT_WAIT);
154 if (mp->mnt_flag & MNT_FORCE)
156 error = ext2_flushfiles(mp, flags, td);
157 if (error == 0 && fs->e2fs_wasvalid && ext2_cgupdate(ump, MNT_WAIT) == 0) {
158 fs->e2fs->e2fs_state |= E2FS_ISCLEAN;
159 ext2_sbupdate(ump, MNT_WAIT);
162 vfs_flagopt(opts, "ro", &mp->mnt_flag, MNT_RDONLY);
164 g_access(ump->um_cp, 0, -1, 0);
167 if (!error && (mp->mnt_flag & MNT_RELOAD))
168 error = ext2_reload(mp, td);
171 devvp = ump->um_devvp;
172 if (fs->e2fs_ronly && !vfs_flagopt(opts, "ro", NULL, 0)) {
173 if (ext2_check_sb_compat(fs->e2fs, devvp->v_rdev, 0))
177 * If upgrade to read-write by non-root, then verify
178 * that user has necessary permissions on the device.
180 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
181 error = VOP_ACCESS(devvp, VREAD | VWRITE,
184 error = priv_check(td, PRIV_VFS_MOUNT_PERM);
186 VOP_UNLOCK(devvp, 0);
189 VOP_UNLOCK(devvp, 0);
191 error = g_access(ump->um_cp, 0, 1, 0);
196 if ((fs->e2fs->e2fs_state & E2FS_ISCLEAN) == 0 ||
197 (fs->e2fs->e2fs_state & E2FS_ERRORS)) {
198 if (mp->mnt_flag & MNT_FORCE) {
200 "WARNING: %s was not properly dismounted\n", fs->e2fs_fsmnt);
203 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
208 fs->e2fs->e2fs_state &= ~E2FS_ISCLEAN;
209 (void)ext2_cgupdate(ump, MNT_WAIT);
212 mp->mnt_flag &= ~MNT_RDONLY;
215 if (vfs_flagopt(opts, "export", NULL, 0)) {
216 /* Process export requests in vfs_mount.c. */
222 * Not an update, or updating the name: look up the name
223 * and verify that it refers to a sensible disk device.
227 NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, fspec, td);
228 if ((error = namei(ndp)) != 0)
230 NDFREE(ndp, NDF_ONLY_PNBUF);
233 if (!vn_isdisk(devvp, &error)) {
239 * If mount by non-root, then verify that user has necessary
240 * permissions on the device.
242 * XXXRW: VOP_ACCESS() enough?
245 if ((mp->mnt_flag & MNT_RDONLY) == 0)
247 error = VOP_ACCESS(devvp, accmode, td->td_ucred, td);
249 error = priv_check(td, PRIV_VFS_MOUNT_PERM);
255 if ((mp->mnt_flag & MNT_UPDATE) == 0) {
256 error = ext2_mountfs(devvp, mp);
258 if (devvp != ump->um_devvp) {
260 return (EINVAL); /* needs translation */
272 * Note that this strncpy() is ok because of a check at the start
275 strncpy(fs->e2fs_fsmnt, path, MAXMNTLEN);
276 fs->e2fs_fsmnt[MAXMNTLEN - 1] = '\0';
277 vfs_mountedfrom(mp, fspec);
282 ext2_check_sb_compat(struct ext2fs *es, struct cdev *dev, int ronly)
286 if (es->e2fs_magic != E2FS_MAGIC) {
287 printf("ext2fs: %s: wrong magic number %#x (expected %#x)\n",
288 devtoname(dev), es->e2fs_magic, E2FS_MAGIC);
291 if (es->e2fs_rev > E2FS_REV0) {
292 mask = es->e2fs_features_incompat & ~(EXT2F_INCOMPAT_SUPP |
293 EXT4F_RO_INCOMPAT_SUPP);
295 printf("WARNING: mount of %s denied due to "
296 "unsupported optional features:\n", devtoname(dev));
298 i < sizeof(incompat)/sizeof(struct ext2_feature);
300 if (mask & incompat[i].mask)
301 printf("%s ", incompat[i].name);
305 mask = es->e2fs_features_rocompat & ~EXT2F_ROCOMPAT_SUPP;
306 if (!ronly && mask) {
307 printf("WARNING: R/W mount of %s denied due to "
308 "unsupported optional features:\n", devtoname(dev));
310 i < sizeof(ro_compat)/sizeof(struct ext2_feature);
312 if (mask & ro_compat[i].mask)
313 printf("%s ", ro_compat[i].name);
322 * This computes the fields of the m_ext2fs structure from the
323 * data in the ext2fs structure read in.
326 compute_sb_data(struct vnode *devvp, struct ext2fs *es,
329 int g_count = 0, error;
331 int logic_sb_block = 1; /* XXX for now */
333 uint32_t e2fs_descpb, e2fs_gdbcount_alloc;
335 fs->e2fs_bcount = es->e2fs_bcount;
336 fs->e2fs_rbcount = es->e2fs_rbcount;
337 fs->e2fs_fbcount = es->e2fs_fbcount;
338 if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_64BIT)) {
339 fs->e2fs_bcount |= (uint64_t)(es->e4fs_bcount_hi) << 32;
340 fs->e2fs_rbcount |= (uint64_t)(es->e4fs_rbcount_hi) << 32;
341 fs->e2fs_fbcount |= (uint64_t)(es->e4fs_fbcount_hi) << 32;
343 fs->e2fs_bshift = EXT2_MIN_BLOCK_LOG_SIZE + es->e2fs_log_bsize;
344 fs->e2fs_bsize = 1U << fs->e2fs_bshift;
345 fs->e2fs_fsbtodb = es->e2fs_log_bsize + 1;
346 fs->e2fs_qbmask = fs->e2fs_bsize - 1;
347 fs->e2fs_fsize = EXT2_MIN_FRAG_SIZE << es->e2fs_log_fsize;
349 fs->e2fs_fpb = fs->e2fs_bsize / fs->e2fs_fsize;
350 fs->e2fs_bpg = es->e2fs_bpg;
351 fs->e2fs_fpg = es->e2fs_fpg;
352 fs->e2fs_ipg = es->e2fs_ipg;
353 if (es->e2fs_rev == E2FS_REV0) {
354 fs->e2fs_isize = E2FS_REV0_INODE_SIZE;
356 fs->e2fs_isize = es->e2fs_inode_size;
359 * Simple sanity check for superblock inode size value.
361 if (EXT2_INODE_SIZE(fs) < E2FS_REV0_INODE_SIZE ||
362 EXT2_INODE_SIZE(fs) > fs->e2fs_bsize ||
363 (fs->e2fs_isize & (fs->e2fs_isize - 1)) != 0) {
364 printf("ext2fs: invalid inode size %d\n",
369 /* Check for extra isize in big inodes. */
370 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_EXTRA_ISIZE) &&
371 EXT2_INODE_SIZE(fs) < sizeof(struct ext2fs_dinode)) {
372 printf("ext2fs: no space for extra inode timestamps\n");
375 /* Check checksum features */
376 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_GDT_CSUM) &&
377 EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_METADATA_CKSUM)) {
378 printf("ext2fs: incorrect checksum features combination\n");
381 /* Check for group descriptor size */
382 if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_64BIT) &&
383 (es->e3fs_desc_size != sizeof(struct ext2_gd))) {
384 printf("ext2fs: group descriptor size unsupported %d\n",
389 fs->e2fs_ipb = fs->e2fs_bsize / EXT2_INODE_SIZE(fs);
390 fs->e2fs_itpg = fs->e2fs_ipg / fs->e2fs_ipb;
391 /* s_resuid / s_resgid ? */
392 fs->e2fs_gcount = howmany(fs->e2fs_bcount - es->e2fs_first_dblock,
393 EXT2_BLOCKS_PER_GROUP(fs));
394 if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_64BIT)) {
395 e2fs_descpb = fs->e2fs_bsize / sizeof(struct ext2_gd);
396 e2fs_gdbcount_alloc = howmany(fs->e2fs_gcount, e2fs_descpb);
398 e2fs_descpb = fs->e2fs_bsize / E2FS_REV0_GD_SIZE;
399 e2fs_gdbcount_alloc = howmany(fs->e2fs_gcount,
400 fs->e2fs_bsize / sizeof(struct ext2_gd));
402 fs->e2fs_gdbcount = howmany(fs->e2fs_gcount, e2fs_descpb);
403 fs->e2fs_gd = mallocarray(e2fs_gdbcount_alloc, fs->e2fs_bsize,
404 M_EXT2MNT, M_WAITOK | M_ZERO);
405 fs->e2fs_contigdirs = mallocarray(fs->e2fs_gcount,
406 sizeof(*fs->e2fs_contigdirs), M_EXT2MNT, M_WAITOK | M_ZERO);
409 * Adjust logic_sb_block.
410 * Godmar thinks: if the blocksize is greater than 1024, then
411 * the superblock is logically part of block zero.
413 if (fs->e2fs_bsize > SBSIZE)
415 for (i = 0; i < fs->e2fs_gdbcount; i++) {
417 fsbtodb(fs, logic_sb_block + i + 1),
418 fs->e2fs_bsize, NOCRED, &bp);
420 free(fs->e2fs_contigdirs, M_EXT2MNT);
421 free(fs->e2fs_gd, M_EXT2MNT);
425 if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_64BIT)) {
427 i * fs->e2fs_bsize / sizeof(struct ext2_gd)],
428 bp->b_data, fs->e2fs_bsize);
430 for (j = 0; j < e2fs_descpb &&
431 g_count < fs->e2fs_gcount; j++, g_count++)
432 memcpy(&fs->e2fs_gd[g_count],
433 bp->b_data + j * E2FS_REV0_GD_SIZE,
439 /* Precompute checksum seed for all metadata */
440 ext2_sb_csum_set_seed(fs);
442 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_GDT_CSUM) ||
443 EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_METADATA_CKSUM)) {
444 error = ext2_gd_csum_verify(fs, devvp->v_rdev);
448 /* Initialization for the ext2 Orlov allocator variant. */
449 fs->e2fs_total_dir = 0;
450 for (i = 0; i < fs->e2fs_gcount; i++)
451 fs->e2fs_total_dir += e2fs_gd_get_ndirs(&fs->e2fs_gd[i]);
453 if (es->e2fs_rev == E2FS_REV0 ||
454 !EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_LARGEFILE))
455 fs->e2fs_maxfilesize = 0x7fffffff;
457 fs->e2fs_maxfilesize = 0xffffffffffff;
458 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_HUGE_FILE))
459 fs->e2fs_maxfilesize = 0x7fffffffffffffff;
461 if (es->e4fs_flags & E2FS_UNSIGNED_HASH) {
463 } else if ((es->e4fs_flags & E2FS_SIGNED_HASH) == 0) {
464 #ifdef __CHAR_UNSIGNED__
465 es->e4fs_flags |= E2FS_UNSIGNED_HASH;
468 es->e4fs_flags |= E2FS_SIGNED_HASH;
471 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_METADATA_CKSUM))
472 error = ext2_sb_csum_verify(fs);
478 * Reload all incore data for a filesystem (used after running fsck on
479 * the root filesystem and finding things to fix). The filesystem must
480 * be mounted read-only.
482 * Things to do to update the mount:
483 * 1) invalidate all cached meta-data.
484 * 2) re-read superblock from disk.
485 * 3) invalidate all cluster summary information.
486 * 4) invalidate all inactive vnodes.
487 * 5) invalidate all cached file data.
488 * 6) re-read inode data for all active vnodes.
489 * XXX we are missing some steps, in particular # 3, this has to be reviewed.
492 ext2_reload(struct mount *mp, struct thread *td)
494 struct vnode *vp, *mvp, *devvp;
503 if ((mp->mnt_flag & MNT_RDONLY) == 0)
506 * Step 1: invalidate all cached meta-data.
508 devvp = VFSTOEXT2(mp)->um_devvp;
509 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
510 if (vinvalbuf(devvp, 0, 0, 0) != 0)
511 panic("ext2_reload: dirty1");
512 VOP_UNLOCK(devvp, 0);
515 * Step 2: re-read superblock from disk.
516 * constants have been adjusted for ext2
518 if ((error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) != 0)
520 es = (struct ext2fs *)bp->b_data;
521 if (ext2_check_sb_compat(es, devvp->v_rdev, 0) != 0) {
523 return (EIO); /* XXX needs translation */
525 fs = VFSTOEXT2(mp)->um_e2fs;
526 bcopy(bp->b_data, fs->e2fs, sizeof(struct ext2fs));
528 if ((error = compute_sb_data(devvp, es, fs)) != 0) {
533 if (fs->fs_sbsize < SBSIZE)
534 bp->b_flags |= B_INVAL;
539 * Step 3: invalidate all cluster summary information.
541 if (fs->e2fs_contigsumsize > 0) {
542 lp = fs->e2fs_maxcluster;
543 sump = fs->e2fs_clustersum;
544 for (i = 0; i < fs->e2fs_gcount; i++, sump++) {
545 *lp++ = fs->e2fs_contigsumsize;
547 bzero(sump->cs_sum, fs->e2fs_contigsumsize + 1);
552 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
554 * Step 4: invalidate all cached file data.
556 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td)) {
557 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
560 if (vinvalbuf(vp, 0, 0, 0))
561 panic("ext2_reload: dirty2");
564 * Step 5: re-read inode data for all active vnodes.
567 error = bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
568 (int)fs->e2fs_bsize, NOCRED, &bp);
572 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
575 ext2_ei2i((struct ext2fs_dinode *)((char *)bp->b_data +
576 EXT2_INODE_SIZE(fs) * ino_to_fsbo(fs, ip->i_number)), ip);
585 * Common code for mount and mountroot.
588 ext2_mountfs(struct vnode *devvp, struct mount *mp)
590 struct ext2mount *ump;
594 struct cdev *dev = devvp->v_rdev;
595 struct g_consumer *cp;
603 int32_t e2fs_maxcontig;
605 ronly = vfs_flagopt(mp->mnt_optnew, "ro", NULL, 0);
606 /* XXX: use VOP_ACESS to check FS perms */
608 error = g_vfs_open(devvp, &cp, "ext2fs", ronly ? 0 : 1);
610 VOP_UNLOCK(devvp, 0);
614 /* XXX: should we check for some sectorsize or 512 instead? */
615 if (((SBSIZE % cp->provider->sectorsize) != 0) ||
616 (SBSIZE < cp->provider->sectorsize)) {
623 bo = &devvp->v_bufobj;
625 bo->bo_ops = g_vfs_bufops;
626 if (devvp->v_rdev->si_iosize_max != 0)
627 mp->mnt_iosize_max = devvp->v_rdev->si_iosize_max;
628 if (mp->mnt_iosize_max > MAXPHYS)
629 mp->mnt_iosize_max = MAXPHYS;
633 if ((error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) != 0)
635 es = (struct ext2fs *)bp->b_data;
636 if (ext2_check_sb_compat(es, dev, ronly) != 0) {
637 error = EINVAL; /* XXX needs translation */
640 if ((es->e2fs_state & E2FS_ISCLEAN) == 0 ||
641 (es->e2fs_state & E2FS_ERRORS)) {
642 if (ronly || (mp->mnt_flag & MNT_FORCE)) {
644 "WARNING: Filesystem was not properly dismounted\n");
647 "WARNING: R/W mount denied. Filesystem is not clean - run fsck\n");
652 ump = malloc(sizeof(*ump), M_EXT2MNT, M_WAITOK | M_ZERO);
655 * I don't know whether this is the right strategy. Note that
656 * we dynamically allocate both an m_ext2fs and an ext2fs
657 * while Linux keeps the super block in a locked buffer.
659 ump->um_e2fs = malloc(sizeof(struct m_ext2fs),
660 M_EXT2MNT, M_WAITOK | M_ZERO);
661 ump->um_e2fs->e2fs = malloc(sizeof(struct ext2fs),
662 M_EXT2MNT, M_WAITOK);
663 mtx_init(EXT2_MTX(ump), "EXT2FS", "EXT2FS Lock", MTX_DEF);
664 bcopy(es, ump->um_e2fs->e2fs, (u_int)sizeof(struct ext2fs));
665 if ((error = compute_sb_data(devvp, ump->um_e2fs->e2fs, ump->um_e2fs)))
669 * Calculate the maximum contiguous blocks and size of cluster summary
670 * array. In FFS this is done by newfs; however, the superblock
671 * in ext2fs doesn't have these variables, so we can calculate
674 e2fs_maxcontig = MAX(1, MAXPHYS / ump->um_e2fs->e2fs_bsize);
675 ump->um_e2fs->e2fs_contigsumsize = MIN(e2fs_maxcontig, EXT2_MAXCONTIG);
676 if (ump->um_e2fs->e2fs_contigsumsize > 0) {
677 size = ump->um_e2fs->e2fs_gcount * sizeof(int32_t);
678 ump->um_e2fs->e2fs_maxcluster = malloc(size, M_EXT2MNT, M_WAITOK);
679 size = ump->um_e2fs->e2fs_gcount * sizeof(struct csum);
680 ump->um_e2fs->e2fs_clustersum = malloc(size, M_EXT2MNT, M_WAITOK);
681 lp = ump->um_e2fs->e2fs_maxcluster;
682 sump = ump->um_e2fs->e2fs_clustersum;
683 for (i = 0; i < ump->um_e2fs->e2fs_gcount; i++, sump++) {
684 *lp++ = ump->um_e2fs->e2fs_contigsumsize;
686 sump->cs_sum = mallocarray(
687 ump->um_e2fs->e2fs_contigsumsize + 1,
688 sizeof(int32_t), M_EXT2MNT, M_WAITOK | M_ZERO);
695 fs->e2fs_ronly = ronly; /* ronly is set according to mnt_flags */
698 * If the fs is not mounted read-only, make sure the super block is
699 * always written back on a sync().
701 fs->e2fs_wasvalid = fs->e2fs->e2fs_state & E2FS_ISCLEAN ? 1 : 0;
703 fs->e2fs_fmod = 1; /* mark it modified */
704 fs->e2fs->e2fs_state &= ~E2FS_ISCLEAN; /* set fs invalid */
707 mp->mnt_stat.f_fsid.val[0] = dev2udev(dev);
708 mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
709 mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN;
711 mp->mnt_flag |= MNT_LOCAL;
715 ump->um_devvp = devvp;
716 ump->um_bo = &devvp->v_bufobj;
720 * Setting those two parameters allowed us to use
721 * ufs_bmap w/o changse!
723 ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs);
724 ump->um_bptrtodb = fs->e2fs->e2fs_log_bsize + 1;
725 ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs);
727 ext2_sbupdate(ump, MNT_WAIT);
729 * Initialize filesystem stat information in mount struct.
732 mp->mnt_kern_flag |= MNTK_LOOKUP_SHARED | MNTK_EXTENDED_SHARED |
745 mtx_destroy(EXT2_MTX(ump));
746 free(ump->um_e2fs->e2fs_gd, M_EXT2MNT);
747 free(ump->um_e2fs->e2fs_contigdirs, M_EXT2MNT);
748 free(ump->um_e2fs->e2fs, M_EXT2MNT);
749 free(ump->um_e2fs, M_EXT2MNT);
750 free(ump, M_EXT2MNT);
757 * Unmount system call.
760 ext2_unmount(struct mount *mp, int mntflags)
762 struct ext2mount *ump;
765 int error, flags, i, ronly;
768 if (mntflags & MNT_FORCE) {
769 if (mp->mnt_flag & MNT_ROOTFS)
773 if ((error = ext2_flushfiles(mp, flags, curthread)) != 0)
777 ronly = fs->e2fs_ronly;
778 if (ronly == 0 && ext2_cgupdate(ump, MNT_WAIT) == 0) {
779 if (fs->e2fs_wasvalid)
780 fs->e2fs->e2fs_state |= E2FS_ISCLEAN;
781 ext2_sbupdate(ump, MNT_WAIT);
785 g_vfs_close(ump->um_cp);
787 vrele(ump->um_devvp);
788 sump = fs->e2fs_clustersum;
789 for (i = 0; i < fs->e2fs_gcount; i++, sump++)
790 free(sump->cs_sum, M_EXT2MNT);
791 free(fs->e2fs_clustersum, M_EXT2MNT);
792 free(fs->e2fs_maxcluster, M_EXT2MNT);
793 free(fs->e2fs_gd, M_EXT2MNT);
794 free(fs->e2fs_contigdirs, M_EXT2MNT);
795 free(fs->e2fs, M_EXT2MNT);
797 free(ump, M_EXT2MNT);
800 mp->mnt_flag &= ~MNT_LOCAL;
806 * Flush out all the files in a filesystem.
809 ext2_flushfiles(struct mount *mp, int flags, struct thread *td)
813 error = vflush(mp, 0, flags, td);
818 * Get filesystem statistics.
821 ext2_statfs(struct mount *mp, struct statfs *sbp)
823 struct ext2mount *ump;
825 uint32_t overhead, overhead_per_group, ngdb;
830 if (fs->e2fs->e2fs_magic != E2FS_MAGIC)
831 panic("ext2_statfs");
834 * Compute the overhead (FS structures)
837 1 /* block bitmap */ +
838 1 /* inode bitmap */ +
840 overhead = fs->e2fs->e2fs_first_dblock +
841 fs->e2fs_gcount * overhead_per_group;
842 if (fs->e2fs->e2fs_rev > E2FS_REV0 &&
843 fs->e2fs->e2fs_features_rocompat & EXT2F_ROCOMPAT_SPARSESUPER) {
844 for (i = 0, ngroups = 0; i < fs->e2fs_gcount; i++) {
845 if (ext2_cg_has_sb(fs, i))
849 ngroups = fs->e2fs_gcount;
851 ngdb = fs->e2fs_gdbcount;
852 if (fs->e2fs->e2fs_rev > E2FS_REV0 &&
853 fs->e2fs->e2fs_features_compat & EXT2F_COMPAT_RESIZE)
854 ngdb += fs->e2fs->e2fs_reserved_ngdb;
855 overhead += ngroups * (1 /* superblock */ + ngdb);
857 sbp->f_bsize = EXT2_FRAG_SIZE(fs);
858 sbp->f_iosize = EXT2_BLOCK_SIZE(fs);
859 sbp->f_blocks = fs->e2fs_bcount - overhead;
860 sbp->f_bfree = fs->e2fs_fbcount;
861 sbp->f_bavail = sbp->f_bfree - fs->e2fs_rbcount;
862 sbp->f_files = fs->e2fs->e2fs_icount;
863 sbp->f_ffree = fs->e2fs->e2fs_ficount;
868 * Go through the disk queues to initiate sandbagged IO;
869 * go through the inodes to write those that have been modified;
870 * initiate the writing of the super block if it has been modified.
872 * Note: we are always called with the filesystem marked `MPBUSY'.
875 ext2_sync(struct mount *mp, int waitfor)
877 struct vnode *mvp, *vp;
880 struct ext2mount *ump = VFSTOEXT2(mp);
882 int error, allerror = 0;
886 if (fs->e2fs_fmod != 0 && fs->e2fs_ronly != 0) { /* XXX */
887 printf("fs = %s\n", fs->e2fs_fsmnt);
888 panic("ext2_sync: rofs mod");
892 * Write back each (modified) inode.
895 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
896 if (vp->v_type == VNON) {
902 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
903 (vp->v_bufobj.bo_dirty.bv_cnt == 0 ||
904 waitfor == MNT_LAZY)) {
908 error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK, td);
910 if (error == ENOENT) {
911 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
916 if ((error = VOP_FSYNC(vp, waitfor, td)) != 0)
923 * Force stale filesystem control information to be flushed.
925 if (waitfor != MNT_LAZY) {
926 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
927 if ((error = VOP_FSYNC(ump->um_devvp, waitfor, td)) != 0)
929 VOP_UNLOCK(ump->um_devvp, 0);
933 * Write back modified superblock.
935 if (fs->e2fs_fmod != 0) {
937 fs->e2fs->e2fs_wtime = time_second;
938 if ((error = ext2_cgupdate(ump, waitfor)) != 0)
945 * Look up an EXT2FS dinode number to find its incore vnode, otherwise read it
946 * in from disk. If it is in core, wait for the lock bit to clear, then
947 * return the inode locked. Detection and handling of mount points must be
948 * done by the calling routine.
951 ext2_vget(struct mount *mp, ino_t ino, int flags, struct vnode **vpp)
955 struct ext2mount *ump;
963 error = vfs_hash_get(mp, ino, flags, td, vpp, NULL, NULL);
964 if (error || *vpp != NULL)
968 ip = malloc(sizeof(struct inode), M_EXT2NODE, M_WAITOK | M_ZERO);
970 /* Allocate a new vnode/inode. */
971 if ((error = getnewvnode("ext2fs", mp, &ext2_vnodeops, &vp)) != 0) {
973 free(ip, M_EXT2NODE);
978 ip->i_e2fs = fs = ump->um_e2fs;
982 lockmgr(vp->v_vnlock, LK_EXCLUSIVE, NULL);
983 error = insmntque(vp, mp);
985 free(ip, M_EXT2NODE);
989 error = vfs_hash_insert(vp, ino, flags, td, vpp, NULL, NULL);
990 if (error || *vpp != NULL)
993 /* Read in the disk contents for the inode, copy into the inode. */
994 if ((error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
995 (int)fs->e2fs_bsize, NOCRED, &bp)) != 0) {
997 * The inode does not contain anything useful, so it would
998 * be misleading to leave it on its hash chain. With mode
999 * still zero, it will be unlinked and returned to the free
1007 /* convert ext2 inode to dinode */
1008 error = ext2_ei2i((struct ext2fs_dinode *)((char *)bp->b_data +
1009 EXT2_INODE_SIZE(fs) * ino_to_fsbo(fs, ino)), ip);
1011 printf("ext2fs: Bad inode %lu csum - run fsck\n",
1012 (unsigned long)ino);
1018 ip->i_block_group = ino_to_cg(fs, ino);
1019 ip->i_next_alloc_block = 0;
1020 ip->i_next_alloc_goal = 0;
1023 * Now we want to make sure that block pointers for unused
1024 * blocks are zeroed out - ext2_balloc depends on this
1025 * although for regular files and directories only
1027 * If IN_E4EXTENTS is enabled, unused blocks are not zeroed
1028 * out because we could corrupt the extent tree.
1030 if (!(ip->i_flag & IN_E4EXTENTS) &&
1031 (S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode))) {
1032 used_blocks = howmany(ip->i_size, fs->e2fs_bsize);
1033 for (i = used_blocks; i < EXT2_NDIR_BLOCKS; i++)
1037 ext2_print_inode(ip);
1038 ext4_ext_print_extent_tree_status(ip);
1043 * Initialize the vnode from the inode, check for aliases.
1044 * Note that the underlying vnode may have changed.
1046 if ((error = ext2_vinit(mp, &ext2_fifoops, &vp)) != 0) {
1053 * Finish inode initialization.
1061 * File handle to vnode
1063 * Have to be really careful about stale file handles:
1064 * - check that the inode number is valid
1065 * - call ext2_vget() to get the locked inode
1066 * - check for an unallocated inode (i_mode == 0)
1067 * - check that the given client host has export rights and return
1068 * those rights via. exflagsp and credanonp
1071 ext2_fhtovp(struct mount *mp, struct fid *fhp, int flags, struct vnode **vpp)
1076 struct m_ext2fs *fs;
1079 ufhp = (struct ufid *)fhp;
1080 fs = VFSTOEXT2(mp)->um_e2fs;
1081 if (ufhp->ufid_ino < EXT2_ROOTINO ||
1082 ufhp->ufid_ino > fs->e2fs_gcount * fs->e2fs->e2fs_ipg)
1085 error = VFS_VGET(mp, ufhp->ufid_ino, LK_EXCLUSIVE, &nvp);
1091 if (ip->i_mode == 0 ||
1092 ip->i_gen != ufhp->ufid_gen || ip->i_nlink <= 0) {
1098 vnode_create_vobject(*vpp, 0, curthread);
1103 * Write a superblock and associated information back to disk.
1106 ext2_sbupdate(struct ext2mount *mp, int waitfor)
1108 struct m_ext2fs *fs = mp->um_e2fs;
1109 struct ext2fs *es = fs->e2fs;
1113 es->e2fs_bcount = fs->e2fs_bcount & 0xffffffff;
1114 es->e2fs_rbcount = fs->e2fs_rbcount & 0xffffffff;
1115 es->e2fs_fbcount = fs->e2fs_fbcount & 0xffffffff;
1116 if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_64BIT)) {
1117 es->e4fs_bcount_hi = fs->e2fs_bcount >> 32;
1118 es->e4fs_rbcount_hi = fs->e2fs_rbcount >> 32;
1119 es->e4fs_fbcount_hi = fs->e2fs_fbcount >> 32;
1122 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_METADATA_CKSUM))
1123 ext2_sb_csum_set(fs);
1125 bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0, 0);
1126 bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2fs));
1127 if (waitfor == MNT_WAIT)
1133 * The buffers for group descriptors, inode bitmaps and block bitmaps
1134 * are not busy at this point and are (hopefully) written by the
1135 * usual sync mechanism. No need to write them here.
1140 ext2_cgupdate(struct ext2mount *mp, int waitfor)
1142 struct m_ext2fs *fs = mp->um_e2fs;
1144 int i, j, g_count = 0, error = 0, allerror = 0;
1146 allerror = ext2_sbupdate(mp, waitfor);
1148 /* Update gd csums */
1149 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_GDT_CSUM) ||
1150 EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_METADATA_CKSUM))
1151 ext2_gd_csum_set(fs);
1153 for (i = 0; i < fs->e2fs_gdbcount; i++) {
1154 bp = getblk(mp->um_devvp, fsbtodb(fs,
1155 fs->e2fs->e2fs_first_dblock +
1156 1 /* superblock */ + i), fs->e2fs_bsize, 0, 0, 0);
1157 if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_64BIT)) {
1158 memcpy(bp->b_data, &fs->e2fs_gd[
1159 i * fs->e2fs_bsize / sizeof(struct ext2_gd)],
1162 for (j = 0; j < fs->e2fs_bsize / E2FS_REV0_GD_SIZE &&
1163 g_count < fs->e2fs_gcount; j++, g_count++)
1164 memcpy(bp->b_data + j * E2FS_REV0_GD_SIZE,
1165 &fs->e2fs_gd[g_count], E2FS_REV0_GD_SIZE);
1167 if (waitfor == MNT_WAIT)
1173 if (!allerror && error)
1179 * Return the root of a filesystem.
1182 ext2_root(struct mount *mp, int flags, struct vnode **vpp)
1187 error = VFS_VGET(mp, EXT2_ROOTINO, LK_EXCLUSIVE, &nvp);