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,
331 int logic_sb_block = 1; /* XXX for now */
333 uint32_t e2fs_descpb;
335 fs->e2fs_bshift = EXT2_MIN_BLOCK_LOG_SIZE + es->e2fs_log_bsize;
336 fs->e2fs_bsize = 1U << fs->e2fs_bshift;
337 fs->e2fs_fsbtodb = es->e2fs_log_bsize + 1;
338 fs->e2fs_qbmask = fs->e2fs_bsize - 1;
339 fs->e2fs_fsize = EXT2_MIN_FRAG_SIZE << es->e2fs_log_fsize;
341 fs->e2fs_fpb = fs->e2fs_bsize / fs->e2fs_fsize;
342 fs->e2fs_bpg = es->e2fs_bpg;
343 fs->e2fs_fpg = es->e2fs_fpg;
344 fs->e2fs_ipg = es->e2fs_ipg;
345 if (es->e2fs_rev == E2FS_REV0) {
346 fs->e2fs_isize = E2FS_REV0_INODE_SIZE;
348 fs->e2fs_isize = es->e2fs_inode_size;
351 * Simple sanity check for superblock inode size value.
353 if (EXT2_INODE_SIZE(fs) < E2FS_REV0_INODE_SIZE ||
354 EXT2_INODE_SIZE(fs) > fs->e2fs_bsize ||
355 (fs->e2fs_isize & (fs->e2fs_isize - 1)) != 0) {
356 printf("ext2fs: invalid inode size %d\n",
361 /* Check for extra isize in big inodes. */
362 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_EXTRA_ISIZE) &&
363 EXT2_INODE_SIZE(fs) < sizeof(struct ext2fs_dinode)) {
364 printf("ext2fs: no space for extra inode timestamps\n");
367 /* Check for group descriptor size */
368 if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_64BIT) &&
369 (es->e3fs_desc_size != sizeof(struct ext2_gd))) {
370 printf("ext2fs: group descriptor size unsupported %d\n",
375 fs->e2fs_ipb = fs->e2fs_bsize / EXT2_INODE_SIZE(fs);
376 fs->e2fs_itpg = fs->e2fs_ipg / fs->e2fs_ipb;
377 /* s_resuid / s_resgid ? */
378 fs->e2fs_gcount = howmany(es->e2fs_bcount - es->e2fs_first_dblock,
379 EXT2_BLOCKS_PER_GROUP(fs));
380 e2fs_descpb = fs->e2fs_bsize / sizeof(struct ext2_gd);
381 db_count = howmany(fs->e2fs_gcount, e2fs_descpb);
382 fs->e2fs_gdbcount = db_count;
383 fs->e2fs_gd = malloc(db_count * fs->e2fs_bsize,
384 M_EXT2MNT, M_WAITOK);
385 fs->e2fs_contigdirs = malloc(fs->e2fs_gcount *
386 sizeof(*fs->e2fs_contigdirs), M_EXT2MNT, M_WAITOK | M_ZERO);
389 * Adjust logic_sb_block.
390 * Godmar thinks: if the blocksize is greater than 1024, then
391 * the superblock is logically part of block zero.
393 if (fs->e2fs_bsize > SBSIZE)
395 for (i = 0; i < db_count; i++) {
397 fsbtodb(fs, logic_sb_block + i + 1),
398 fs->e2fs_bsize, NOCRED, &bp);
400 free(fs->e2fs_contigdirs, M_EXT2MNT);
401 free(fs->e2fs_gd, M_EXT2MNT);
405 e2fs_cgload((struct ext2_gd *)bp->b_data,
407 i * fs->e2fs_bsize / sizeof(struct ext2_gd)],
413 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_GDT_CSUM)) {
414 error = ext2_gd_csum_verify(fs, devvp->v_rdev);
418 /* Initialization for the ext2 Orlov allocator variant. */
419 fs->e2fs_total_dir = 0;
420 for (i = 0; i < fs->e2fs_gcount; i++)
421 fs->e2fs_total_dir += fs->e2fs_gd[i].ext2bgd_ndirs;
423 if (es->e2fs_rev == E2FS_REV0 ||
424 !EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_LARGEFILE))
425 fs->e2fs_maxfilesize = 0x7fffffff;
427 fs->e2fs_maxfilesize = 0xffffffffffff;
428 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_HUGE_FILE))
429 fs->e2fs_maxfilesize = 0x7fffffffffffffff;
431 if (es->e4fs_flags & E2FS_UNSIGNED_HASH) {
433 } else if ((es->e4fs_flags & E2FS_SIGNED_HASH) == 0) {
434 #ifdef __CHAR_UNSIGNED__
435 es->e4fs_flags |= E2FS_UNSIGNED_HASH;
438 es->e4fs_flags |= E2FS_SIGNED_HASH;
446 * Reload all incore data for a filesystem (used after running fsck on
447 * the root filesystem and finding things to fix). The filesystem must
448 * be mounted read-only.
450 * Things to do to update the mount:
451 * 1) invalidate all cached meta-data.
452 * 2) re-read superblock from disk.
453 * 3) invalidate all cluster summary information.
454 * 4) invalidate all inactive vnodes.
455 * 5) invalidate all cached file data.
456 * 6) re-read inode data for all active vnodes.
457 * XXX we are missing some steps, in particular # 3, this has to be reviewed.
460 ext2_reload(struct mount *mp, struct thread *td)
462 struct vnode *vp, *mvp, *devvp;
471 if ((mp->mnt_flag & MNT_RDONLY) == 0)
474 * Step 1: invalidate all cached meta-data.
476 devvp = VFSTOEXT2(mp)->um_devvp;
477 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
478 if (vinvalbuf(devvp, 0, 0, 0) != 0)
479 panic("ext2_reload: dirty1");
480 VOP_UNLOCK(devvp, 0);
483 * Step 2: re-read superblock from disk.
484 * constants have been adjusted for ext2
486 if ((error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) != 0)
488 es = (struct ext2fs *)bp->b_data;
489 if (ext2_check_sb_compat(es, devvp->v_rdev, 0) != 0) {
491 return (EIO); /* XXX needs translation */
493 fs = VFSTOEXT2(mp)->um_e2fs;
494 bcopy(bp->b_data, fs->e2fs, sizeof(struct ext2fs));
496 if ((error = compute_sb_data(devvp, es, fs)) != 0) {
501 if (fs->fs_sbsize < SBSIZE)
502 bp->b_flags |= B_INVAL;
507 * Step 3: invalidate all cluster summary information.
509 if (fs->e2fs_contigsumsize > 0) {
510 lp = fs->e2fs_maxcluster;
511 sump = fs->e2fs_clustersum;
512 for (i = 0; i < fs->e2fs_gcount; i++, sump++) {
513 *lp++ = fs->e2fs_contigsumsize;
515 bzero(sump->cs_sum, fs->e2fs_contigsumsize + 1);
520 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
522 * Step 4: invalidate all cached file data.
524 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td)) {
525 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
528 if (vinvalbuf(vp, 0, 0, 0))
529 panic("ext2_reload: dirty2");
532 * Step 5: re-read inode data for all active vnodes.
535 error = bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
536 (int)fs->e2fs_bsize, NOCRED, &bp);
540 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
543 ext2_ei2i((struct ext2fs_dinode *)((char *)bp->b_data +
544 EXT2_INODE_SIZE(fs) * ino_to_fsbo(fs, ip->i_number)), ip);
553 * Common code for mount and mountroot.
556 ext2_mountfs(struct vnode *devvp, struct mount *mp)
558 struct ext2mount *ump;
562 struct cdev *dev = devvp->v_rdev;
563 struct g_consumer *cp;
571 int32_t e2fs_maxcontig;
573 ronly = vfs_flagopt(mp->mnt_optnew, "ro", NULL, 0);
574 /* XXX: use VOP_ACESS to check FS perms */
576 error = g_vfs_open(devvp, &cp, "ext2fs", ronly ? 0 : 1);
578 VOP_UNLOCK(devvp, 0);
582 /* XXX: should we check for some sectorsize or 512 instead? */
583 if (((SBSIZE % cp->provider->sectorsize) != 0) ||
584 (SBSIZE < cp->provider->sectorsize)) {
591 bo = &devvp->v_bufobj;
593 bo->bo_ops = g_vfs_bufops;
594 if (devvp->v_rdev->si_iosize_max != 0)
595 mp->mnt_iosize_max = devvp->v_rdev->si_iosize_max;
596 if (mp->mnt_iosize_max > MAXPHYS)
597 mp->mnt_iosize_max = MAXPHYS;
601 if ((error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) != 0)
603 es = (struct ext2fs *)bp->b_data;
604 if (ext2_check_sb_compat(es, dev, ronly) != 0) {
605 error = EINVAL; /* XXX needs translation */
608 if ((es->e2fs_state & E2FS_ISCLEAN) == 0 ||
609 (es->e2fs_state & E2FS_ERRORS)) {
610 if (ronly || (mp->mnt_flag & MNT_FORCE)) {
612 "WARNING: Filesystem was not properly dismounted\n");
615 "WARNING: R/W mount denied. Filesystem is not clean - run fsck\n");
620 ump = malloc(sizeof(*ump), M_EXT2MNT, M_WAITOK | M_ZERO);
623 * I don't know whether this is the right strategy. Note that
624 * we dynamically allocate both an m_ext2fs and an ext2fs
625 * while Linux keeps the super block in a locked buffer.
627 ump->um_e2fs = malloc(sizeof(struct m_ext2fs),
628 M_EXT2MNT, M_WAITOK | M_ZERO);
629 ump->um_e2fs->e2fs = malloc(sizeof(struct ext2fs),
630 M_EXT2MNT, M_WAITOK);
631 mtx_init(EXT2_MTX(ump), "EXT2FS", "EXT2FS Lock", MTX_DEF);
632 bcopy(es, ump->um_e2fs->e2fs, (u_int)sizeof(struct ext2fs));
633 if ((error = compute_sb_data(devvp, ump->um_e2fs->e2fs, ump->um_e2fs)))
637 * Calculate the maximum contiguous blocks and size of cluster summary
638 * array. In FFS this is done by newfs; however, the superblock
639 * in ext2fs doesn't have these variables, so we can calculate
642 e2fs_maxcontig = MAX(1, MAXPHYS / ump->um_e2fs->e2fs_bsize);
643 ump->um_e2fs->e2fs_contigsumsize = MIN(e2fs_maxcontig, EXT2_MAXCONTIG);
644 if (ump->um_e2fs->e2fs_contigsumsize > 0) {
645 size = ump->um_e2fs->e2fs_gcount * sizeof(int32_t);
646 ump->um_e2fs->e2fs_maxcluster = malloc(size, M_EXT2MNT, M_WAITOK);
647 size = ump->um_e2fs->e2fs_gcount * sizeof(struct csum);
648 ump->um_e2fs->e2fs_clustersum = malloc(size, M_EXT2MNT, M_WAITOK);
649 lp = ump->um_e2fs->e2fs_maxcluster;
650 sump = ump->um_e2fs->e2fs_clustersum;
651 for (i = 0; i < ump->um_e2fs->e2fs_gcount; i++, sump++) {
652 *lp++ = ump->um_e2fs->e2fs_contigsumsize;
654 sump->cs_sum = malloc((ump->um_e2fs->e2fs_contigsumsize + 1) *
655 sizeof(int32_t), M_EXT2MNT, M_WAITOK | M_ZERO);
662 fs->e2fs_ronly = ronly; /* ronly is set according to mnt_flags */
665 * If the fs is not mounted read-only, make sure the super block is
666 * always written back on a sync().
668 fs->e2fs_wasvalid = fs->e2fs->e2fs_state & E2FS_ISCLEAN ? 1 : 0;
670 fs->e2fs_fmod = 1; /* mark it modified */
671 fs->e2fs->e2fs_state &= ~E2FS_ISCLEAN; /* set fs invalid */
674 mp->mnt_stat.f_fsid.val[0] = dev2udev(dev);
675 mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
676 mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN;
678 mp->mnt_flag |= MNT_LOCAL;
682 ump->um_devvp = devvp;
683 ump->um_bo = &devvp->v_bufobj;
687 * Setting those two parameters allowed us to use
688 * ufs_bmap w/o changse!
690 ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs);
691 ump->um_bptrtodb = fs->e2fs->e2fs_log_bsize + 1;
692 ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs);
694 ext2_sbupdate(ump, MNT_WAIT);
696 * Initialize filesystem stat information in mount struct.
699 mp->mnt_kern_flag |= MNTK_LOOKUP_SHARED | MNTK_EXTENDED_SHARED |
712 mtx_destroy(EXT2_MTX(ump));
713 free(ump->um_e2fs->e2fs_gd, M_EXT2MNT);
714 free(ump->um_e2fs->e2fs_contigdirs, M_EXT2MNT);
715 free(ump->um_e2fs->e2fs, M_EXT2MNT);
716 free(ump->um_e2fs, M_EXT2MNT);
717 free(ump, M_EXT2MNT);
724 * Unmount system call.
727 ext2_unmount(struct mount *mp, int mntflags)
729 struct ext2mount *ump;
732 int error, flags, i, ronly;
735 if (mntflags & MNT_FORCE) {
736 if (mp->mnt_flag & MNT_ROOTFS)
740 if ((error = ext2_flushfiles(mp, flags, curthread)) != 0)
744 ronly = fs->e2fs_ronly;
745 if (ronly == 0 && ext2_cgupdate(ump, MNT_WAIT) == 0) {
746 if (fs->e2fs_wasvalid)
747 fs->e2fs->e2fs_state |= E2FS_ISCLEAN;
748 ext2_sbupdate(ump, MNT_WAIT);
752 g_vfs_close(ump->um_cp);
754 vrele(ump->um_devvp);
755 sump = fs->e2fs_clustersum;
756 for (i = 0; i < fs->e2fs_gcount; i++, sump++)
757 free(sump->cs_sum, M_EXT2MNT);
758 free(fs->e2fs_clustersum, M_EXT2MNT);
759 free(fs->e2fs_maxcluster, M_EXT2MNT);
760 free(fs->e2fs_gd, M_EXT2MNT);
761 free(fs->e2fs_contigdirs, M_EXT2MNT);
762 free(fs->e2fs, M_EXT2MNT);
764 free(ump, M_EXT2MNT);
767 mp->mnt_flag &= ~MNT_LOCAL;
773 * Flush out all the files in a filesystem.
776 ext2_flushfiles(struct mount *mp, int flags, struct thread *td)
780 error = vflush(mp, 0, flags, td);
785 * Get filesystem statistics.
788 ext2_statfs(struct mount *mp, struct statfs *sbp)
790 struct ext2mount *ump;
792 uint32_t overhead, overhead_per_group, ngdb;
797 if (fs->e2fs->e2fs_magic != E2FS_MAGIC)
798 panic("ext2_statfs");
801 * Compute the overhead (FS structures)
804 1 /* block bitmap */ +
805 1 /* inode bitmap */ +
807 overhead = fs->e2fs->e2fs_first_dblock +
808 fs->e2fs_gcount * overhead_per_group;
809 if (fs->e2fs->e2fs_rev > E2FS_REV0 &&
810 fs->e2fs->e2fs_features_rocompat & EXT2F_ROCOMPAT_SPARSESUPER) {
811 for (i = 0, ngroups = 0; i < fs->e2fs_gcount; i++) {
812 if (ext2_cg_has_sb(fs, i))
816 ngroups = fs->e2fs_gcount;
818 ngdb = fs->e2fs_gdbcount;
819 if (fs->e2fs->e2fs_rev > E2FS_REV0 &&
820 fs->e2fs->e2fs_features_compat & EXT2F_COMPAT_RESIZE)
821 ngdb += fs->e2fs->e2fs_reserved_ngdb;
822 overhead += ngroups * (1 /* superblock */ + ngdb);
824 sbp->f_bsize = EXT2_FRAG_SIZE(fs);
825 sbp->f_iosize = EXT2_BLOCK_SIZE(fs);
826 sbp->f_blocks = fs->e2fs->e2fs_bcount - overhead;
827 sbp->f_bfree = fs->e2fs->e2fs_fbcount;
828 sbp->f_bavail = sbp->f_bfree - fs->e2fs->e2fs_rbcount;
829 sbp->f_files = fs->e2fs->e2fs_icount;
830 sbp->f_ffree = fs->e2fs->e2fs_ficount;
835 * Go through the disk queues to initiate sandbagged IO;
836 * go through the inodes to write those that have been modified;
837 * initiate the writing of the super block if it has been modified.
839 * Note: we are always called with the filesystem marked `MPBUSY'.
842 ext2_sync(struct mount *mp, int waitfor)
844 struct vnode *mvp, *vp;
847 struct ext2mount *ump = VFSTOEXT2(mp);
849 int error, allerror = 0;
853 if (fs->e2fs_fmod != 0 && fs->e2fs_ronly != 0) { /* XXX */
854 printf("fs = %s\n", fs->e2fs_fsmnt);
855 panic("ext2_sync: rofs mod");
859 * Write back each (modified) inode.
862 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
863 if (vp->v_type == VNON) {
869 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
870 (vp->v_bufobj.bo_dirty.bv_cnt == 0 ||
871 waitfor == MNT_LAZY)) {
875 error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK, td);
877 if (error == ENOENT) {
878 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
883 if ((error = VOP_FSYNC(vp, waitfor, td)) != 0)
890 * Force stale filesystem control information to be flushed.
892 if (waitfor != MNT_LAZY) {
893 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
894 if ((error = VOP_FSYNC(ump->um_devvp, waitfor, td)) != 0)
896 VOP_UNLOCK(ump->um_devvp, 0);
900 * Write back modified superblock.
902 if (fs->e2fs_fmod != 0) {
904 fs->e2fs->e2fs_wtime = time_second;
905 if ((error = ext2_cgupdate(ump, waitfor)) != 0)
912 * Look up an EXT2FS dinode number to find its incore vnode, otherwise read it
913 * in from disk. If it is in core, wait for the lock bit to clear, then
914 * return the inode locked. Detection and handling of mount points must be
915 * done by the calling routine.
918 ext2_vget(struct mount *mp, ino_t ino, int flags, struct vnode **vpp)
922 struct ext2mount *ump;
930 error = vfs_hash_get(mp, ino, flags, td, vpp, NULL, NULL);
931 if (error || *vpp != NULL)
935 ip = malloc(sizeof(struct inode), M_EXT2NODE, M_WAITOK | M_ZERO);
937 /* Allocate a new vnode/inode. */
938 if ((error = getnewvnode("ext2fs", mp, &ext2_vnodeops, &vp)) != 0) {
940 free(ip, M_EXT2NODE);
945 ip->i_e2fs = fs = ump->um_e2fs;
949 lockmgr(vp->v_vnlock, LK_EXCLUSIVE, NULL);
950 error = insmntque(vp, mp);
952 free(ip, M_EXT2NODE);
956 error = vfs_hash_insert(vp, ino, flags, td, vpp, NULL, NULL);
957 if (error || *vpp != NULL)
960 /* Read in the disk contents for the inode, copy into the inode. */
961 if ((error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
962 (int)fs->e2fs_bsize, NOCRED, &bp)) != 0) {
964 * The inode does not contain anything useful, so it would
965 * be misleading to leave it on its hash chain. With mode
966 * still zero, it will be unlinked and returned to the free
974 /* convert ext2 inode to dinode */
975 ext2_ei2i((struct ext2fs_dinode *)((char *)bp->b_data + EXT2_INODE_SIZE(fs) *
976 ino_to_fsbo(fs, ino)), ip);
977 ip->i_block_group = ino_to_cg(fs, ino);
978 ip->i_next_alloc_block = 0;
979 ip->i_next_alloc_goal = 0;
982 * Now we want to make sure that block pointers for unused
983 * blocks are zeroed out - ext2_balloc depends on this
984 * although for regular files and directories only
986 * If IN_E4EXTENTS is enabled, unused blocks are not zeroed
987 * out because we could corrupt the extent tree.
989 if (!(ip->i_flag & IN_E4EXTENTS) &&
990 (S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode))) {
991 used_blocks = howmany(ip->i_size, fs->e2fs_bsize);
992 for (i = used_blocks; i < EXT2_NDIR_BLOCKS; i++)
996 ext2_print_inode(ip);
997 ext4_ext_print_extent_tree_status(ip);
1002 * Initialize the vnode from the inode, check for aliases.
1003 * Note that the underlying vnode may have changed.
1005 if ((error = ext2_vinit(mp, &ext2_fifoops, &vp)) != 0) {
1012 * Finish inode initialization.
1020 * File handle to vnode
1022 * Have to be really careful about stale file handles:
1023 * - check that the inode number is valid
1024 * - call ext2_vget() to get the locked inode
1025 * - check for an unallocated inode (i_mode == 0)
1026 * - check that the given client host has export rights and return
1027 * those rights via. exflagsp and credanonp
1030 ext2_fhtovp(struct mount *mp, struct fid *fhp, int flags, struct vnode **vpp)
1035 struct m_ext2fs *fs;
1038 ufhp = (struct ufid *)fhp;
1039 fs = VFSTOEXT2(mp)->um_e2fs;
1040 if (ufhp->ufid_ino < EXT2_ROOTINO ||
1041 ufhp->ufid_ino > fs->e2fs_gcount * fs->e2fs->e2fs_ipg)
1044 error = VFS_VGET(mp, ufhp->ufid_ino, LK_EXCLUSIVE, &nvp);
1050 if (ip->i_mode == 0 ||
1051 ip->i_gen != ufhp->ufid_gen || ip->i_nlink <= 0) {
1057 vnode_create_vobject(*vpp, 0, curthread);
1062 * Write a superblock and associated information back to disk.
1065 ext2_sbupdate(struct ext2mount *mp, int waitfor)
1067 struct m_ext2fs *fs = mp->um_e2fs;
1068 struct ext2fs *es = fs->e2fs;
1072 bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0, 0);
1073 bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2fs));
1074 if (waitfor == MNT_WAIT)
1080 * The buffers for group descriptors, inode bitmaps and block bitmaps
1081 * are not busy at this point and are (hopefully) written by the
1082 * usual sync mechanism. No need to write them here.
1087 ext2_cgupdate(struct ext2mount *mp, int waitfor)
1089 struct m_ext2fs *fs = mp->um_e2fs;
1091 int i, error = 0, allerror = 0;
1093 allerror = ext2_sbupdate(mp, waitfor);
1095 /* Update gd csums */
1096 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_GDT_CSUM))
1097 ext2_gd_csum_set(fs);
1099 for (i = 0; i < fs->e2fs_gdbcount; i++) {
1100 bp = getblk(mp->um_devvp, fsbtodb(fs,
1101 fs->e2fs->e2fs_first_dblock +
1102 1 /* superblock */ + i), fs->e2fs_bsize, 0, 0, 0);
1103 e2fs_cgsave(&fs->e2fs_gd[
1104 i * fs->e2fs_bsize / sizeof(struct ext2_gd)],
1105 (struct ext2_gd *)bp->b_data, fs->e2fs_bsize);
1106 if (waitfor == MNT_WAIT)
1112 if (!allerror && error)
1118 * Return the root of a filesystem.
1121 ext2_root(struct mount *mp, int flags, struct vnode **vpp)
1126 error = VFS_VGET(mp, EXT2_ROOTINO, LK_EXCLUSIVE, &nvp);