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/fs.h>
62 #include <fs/ext2fs/ext2_mount.h>
63 #include <fs/ext2fs/inode.h>
65 #include <fs/ext2fs/ext2fs.h>
66 #include <fs/ext2fs/ext2_dinode.h>
67 #include <fs/ext2fs/ext2_extern.h>
68 #include <fs/ext2fs/ext2_extents.h>
71 static int ext2_flushfiles(struct mount *mp, int flags, struct thread *td);
72 static int ext2_mountfs(struct vnode *, struct mount *);
73 static int ext2_reload(struct mount *mp, struct thread *td);
74 static int ext2_sbupdate(struct ext2mount *, int);
75 static int ext2_cgupdate(struct ext2mount *, int);
76 static vfs_unmount_t ext2_unmount;
77 static vfs_root_t ext2_root;
78 static vfs_statfs_t ext2_statfs;
79 static vfs_sync_t ext2_sync;
80 static vfs_vget_t ext2_vget;
81 static vfs_fhtovp_t ext2_fhtovp;
82 static vfs_mount_t ext2_mount;
84 MALLOC_DEFINE(M_EXT2NODE, "ext2_node", "EXT2 vnode private part");
85 static MALLOC_DEFINE(M_EXT2MNT, "ext2_mount", "EXT2 mount structure");
87 static struct vfsops ext2fs_vfsops = {
88 .vfs_fhtovp = ext2_fhtovp,
89 .vfs_mount = ext2_mount,
90 .vfs_root = ext2_root, /* root inode via vget */
91 .vfs_statfs = ext2_statfs,
92 .vfs_sync = ext2_sync,
93 .vfs_unmount = ext2_unmount,
94 .vfs_vget = ext2_vget,
97 VFS_SET(ext2fs_vfsops, ext2fs, 0);
99 static int ext2_check_sb_compat(struct ext2fs *es, struct cdev *dev,
101 static int ext2_compute_sb_data(struct vnode * devvp,
102 struct ext2fs * es, struct m_ext2fs * fs);
104 static const char *ext2_opts[] = { "acls", "async", "noatime", "noclusterr",
105 "noclusterw", "noexec", "export", "force", "from", "multilabel",
106 "suiddir", "nosymfollow", "sync", "union", NULL };
114 ext2_mount(struct mount *mp)
116 struct vfsoptlist *opts;
119 struct ext2mount *ump = NULL;
121 struct nameidata nd, *ndp = &nd;
124 int error, flags, len;
127 opts = mp->mnt_optnew;
129 if (vfs_filteropt(opts, ext2_opts))
132 vfs_getopt(opts, "fspath", (void **)&path, NULL);
133 /* Double-check the length of path.. */
134 if (strlen(path) >= MAXMNTLEN)
135 return (ENAMETOOLONG);
138 error = vfs_getopt(opts, "from", (void **)&fspec, &len);
139 if (!error && fspec[len - 1] != '\0')
143 * If updating, check whether changing from read-only to
144 * read/write; if there is no device name, that's all we do.
146 if (mp->mnt_flag & MNT_UPDATE) {
150 if (fs->e2fs_ronly == 0 &&
151 vfs_flagopt(opts, "ro", NULL, 0)) {
152 error = VFS_SYNC(mp, MNT_WAIT);
156 if (mp->mnt_flag & MNT_FORCE)
158 error = ext2_flushfiles(mp, flags, td);
159 if (error == 0 && fs->e2fs_wasvalid &&
160 ext2_cgupdate(ump, MNT_WAIT) == 0) {
161 fs->e2fs->e2fs_state |= E2FS_ISCLEAN;
162 ext2_sbupdate(ump, MNT_WAIT);
165 vfs_flagopt(opts, "ro", &mp->mnt_flag, MNT_RDONLY);
167 g_access(ump->um_cp, 0, -1, 0);
170 if (!error && (mp->mnt_flag & MNT_RELOAD))
171 error = ext2_reload(mp, td);
174 devvp = ump->um_devvp;
175 if (fs->e2fs_ronly && !vfs_flagopt(opts, "ro", NULL, 0)) {
176 if (ext2_check_sb_compat(fs->e2fs, devvp->v_rdev, 0))
180 * If upgrade to read-write by non-root, then verify
181 * that user has necessary permissions on the device.
183 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
184 error = VOP_ACCESS(devvp, VREAD | VWRITE,
187 error = priv_check(td, PRIV_VFS_MOUNT_PERM);
189 VOP_UNLOCK(devvp, 0);
192 VOP_UNLOCK(devvp, 0);
194 error = g_access(ump->um_cp, 0, 1, 0);
199 if ((fs->e2fs->e2fs_state & E2FS_ISCLEAN) == 0 ||
200 (fs->e2fs->e2fs_state & E2FS_ERRORS)) {
201 if (mp->mnt_flag & MNT_FORCE) {
203 "WARNING: %s was not properly dismounted\n", fs->e2fs_fsmnt);
206 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
211 fs->e2fs->e2fs_state &= ~E2FS_ISCLEAN;
212 (void)ext2_cgupdate(ump, MNT_WAIT);
215 mp->mnt_flag &= ~MNT_RDONLY;
218 if (vfs_flagopt(opts, "export", NULL, 0)) {
219 /* Process export requests in vfs_mount.c. */
225 * Not an update, or updating the name: look up the name
226 * and verify that it refers to a sensible disk device.
230 NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, fspec, td);
231 if ((error = namei(ndp)) != 0)
233 NDFREE(ndp, NDF_ONLY_PNBUF);
236 if (!vn_isdisk(devvp, &error)) {
242 * If mount by non-root, then verify that user has necessary
243 * permissions on the device.
245 * XXXRW: VOP_ACCESS() enough?
248 if ((mp->mnt_flag & MNT_RDONLY) == 0)
250 error = VOP_ACCESS(devvp, accmode, td->td_ucred, td);
252 error = priv_check(td, PRIV_VFS_MOUNT_PERM);
258 if ((mp->mnt_flag & MNT_UPDATE) == 0) {
259 error = ext2_mountfs(devvp, mp);
261 if (devvp != ump->um_devvp) {
263 return (EINVAL); /* needs translation */
275 * Note that this strncpy() is ok because of a check at the start
278 strncpy(fs->e2fs_fsmnt, path, MAXMNTLEN);
279 fs->e2fs_fsmnt[MAXMNTLEN - 1] = '\0';
280 vfs_mountedfrom(mp, fspec);
285 ext2_check_sb_compat(struct ext2fs *es, struct cdev *dev, int ronly)
289 if (es->e2fs_magic != E2FS_MAGIC) {
290 printf("ext2fs: %s: wrong magic number %#x (expected %#x)\n",
291 devtoname(dev), es->e2fs_magic, E2FS_MAGIC);
294 if (es->e2fs_rev > E2FS_REV0) {
295 mask = es->e2fs_features_incompat & ~(EXT2F_INCOMPAT_SUPP);
297 printf("WARNING: mount of %s denied due to "
298 "unsupported optional features:\n", devtoname(dev));
300 i < sizeof(incompat)/sizeof(struct ext2_feature);
302 if (mask & incompat[i].mask)
303 printf("%s ", incompat[i].name);
307 mask = es->e2fs_features_rocompat & ~EXT2F_ROCOMPAT_SUPP;
308 if (!ronly && mask) {
309 printf("WARNING: R/W mount of %s denied due to "
310 "unsupported optional features:\n", devtoname(dev));
312 i < sizeof(ro_compat)/sizeof(struct ext2_feature);
314 if (mask & ro_compat[i].mask)
315 printf("%s ", ro_compat[i].name);
324 ext2_cg_location(struct m_ext2fs *fs, int number)
326 int cg, descpb, logical_sb, has_super = 0;
329 * Adjust logical superblock block number.
330 * Godmar thinks: if the blocksize is greater than 1024, then
331 * the superblock is logically part of block zero.
333 logical_sb = fs->e2fs_bsize > SBSIZE ? 0 : 1;
335 if (!EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_META_BG) ||
336 number < fs->e2fs->e3fs_first_meta_bg)
337 return (logical_sb + number + 1);
339 if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_64BIT))
340 descpb = fs->e2fs_bsize / sizeof(struct ext2_gd);
342 descpb = fs->e2fs_bsize / E2FS_REV0_GD_SIZE;
344 cg = descpb * number;
346 if (ext2_cg_has_sb(fs, cg))
349 return (has_super + cg * (e4fs_daddr_t)EXT2_BLOCKS_PER_GROUP(fs) +
350 fs->e2fs->e2fs_first_dblock);
354 ext2_cg_validate(struct m_ext2fs *fs)
359 uint64_t first_block, last_block, last_cg_block;
361 unsigned int i, cg_count;
363 first_block = fs->e2fs->e2fs_first_dblock;
364 last_cg_block = ext2_cg_number_gdb(fs, 0);
365 cg_count = fs->e2fs_gcount;
367 for (i = 0; i < fs->e2fs_gcount; i++) {
368 gd = &fs->e2fs_gd[i];
370 if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_FLEX_BG) ||
371 i == fs->e2fs_gcount - 1) {
372 last_block = fs->e2fs_bcount - 1;
374 last_block = first_block +
375 (EXT2_BLOCKS_PER_GROUP(fs) - 1);
378 if ((cg_count == fs->e2fs_gcount) &&
379 !(gd->ext4bgd_flags & EXT2_BG_INODE_ZEROED))
382 b_bitmap = e2fs_gd_get_b_bitmap(gd);
384 printf("ext2fs: cg %u: block bitmap is zero\n", i);
388 if (b_bitmap <= last_cg_block) {
389 printf("ext2fs: cg %u: block bitmap overlaps gds\n", i);
392 if (b_bitmap < first_block || b_bitmap > last_block) {
393 printf("ext2fs: cg %u: block bitmap not in group, blk=%ju\n",
398 i_bitmap = e2fs_gd_get_i_bitmap(gd);
400 printf("ext2fs: cg %u: inode bitmap is zero\n", i);
403 if (i_bitmap <= last_cg_block) {
404 printf("ext2fs: cg %u: inode bitmap overlaps gds\n", i);
407 if (i_bitmap < first_block || i_bitmap > last_block) {
408 printf("ext2fs: cg %u: inode bitmap not in group blk=%ju\n",
413 i_tables = e2fs_gd_get_i_tables(gd);
415 printf("ext2fs: cg %u: inode table is zero\n", i);
418 if (i_tables <= last_cg_block) {
419 printf("ext2fs: cg %u: inode talbes overlaps gds\n", i);
422 if (i_tables < first_block ||
423 i_tables + fs->e2fs_itpg - 1 > last_block) {
424 printf("ext2fs: cg %u: inode tables not in group blk=%ju\n",
429 if (!EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_FLEX_BG))
430 first_block += EXT2_BLOCKS_PER_GROUP(fs);
437 * This computes the fields of the m_ext2fs structure from the
438 * data in the ext2fs structure read in.
441 ext2_compute_sb_data(struct vnode *devvp, struct ext2fs *es,
445 uint32_t e2fs_descpb, e2fs_gdbcount_alloc;
450 /* Check checksum features */
451 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_GDT_CSUM) &&
452 EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_METADATA_CKSUM)) {
453 printf("ext2fs: incorrect checksum features combination\n");
457 /* Precompute checksum seed for all metadata */
458 ext2_sb_csum_set_seed(fs);
460 /* Verify sb csum if possible */
461 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_METADATA_CKSUM)) {
462 error = ext2_sb_csum_verify(fs);
468 /* Check for block size = 1K|2K|4K */
469 if (es->e2fs_log_bsize > 2) {
470 printf("ext2fs: bad block size: %d\n", es->e2fs_log_bsize);
474 fs->e2fs_bshift = EXT2_MIN_BLOCK_LOG_SIZE + es->e2fs_log_bsize;
475 fs->e2fs_bsize = 1U << fs->e2fs_bshift;
476 fs->e2fs_fsbtodb = es->e2fs_log_bsize + 1;
477 fs->e2fs_qbmask = fs->e2fs_bsize - 1;
479 /* Check for fragment size */
480 if (es->e2fs_log_fsize >
481 (EXT2_MAX_FRAG_LOG_SIZE - EXT2_MIN_BLOCK_LOG_SIZE)) {
482 printf("ext2fs: invalid log cluster size: %u\n",
487 fs->e2fs_fsize = EXT2_MIN_FRAG_SIZE << es->e2fs_log_fsize;
488 if (fs->e2fs_fsize != fs->e2fs_bsize) {
489 printf("ext2fs: fragment size (%u) != block size %u\n",
490 fs->e2fs_fsize, fs->e2fs_bsize);
494 fs->e2fs_fpb = fs->e2fs_bsize / fs->e2fs_fsize;
496 /* Check reserved gdt blocks for future filesystem expansion */
497 if (es->e2fs_reserved_ngdb > (fs->e2fs_bsize / 4)) {
498 printf("ext2fs: number of reserved GDT blocks too large: %u\n",
499 es->e2fs_reserved_ngdb);
503 if (es->e2fs_rev == E2FS_REV0) {
504 fs->e2fs_isize = E2FS_REV0_INODE_SIZE;
506 fs->e2fs_isize = es->e2fs_inode_size;
511 if (es->e2fs_first_ino < EXT2_FIRSTINO) {
512 printf("ext2fs: invalid first ino: %u\n",
518 * Simple sanity check for superblock inode size value.
520 if (EXT2_INODE_SIZE(fs) < E2FS_REV0_INODE_SIZE ||
521 EXT2_INODE_SIZE(fs) > fs->e2fs_bsize ||
522 (fs->e2fs_isize & (fs->e2fs_isize - 1)) != 0) {
523 printf("ext2fs: invalid inode size %u\n",
529 /* Check group descriptors */
530 if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_64BIT) &&
531 es->e3fs_desc_size != E2FS_64BIT_GD_SIZE) {
532 printf("ext2fs: unsupported 64bit descriptor size %u\n",
537 fs->e2fs_bpg = es->e2fs_bpg;
538 fs->e2fs_fpg = es->e2fs_fpg;
539 if (fs->e2fs_bpg == 0 || fs->e2fs_fpg == 0) {
540 printf("ext2fs: zero blocks/fragments per group\n");
543 if (fs->e2fs_bpg != fs->e2fs_bsize * 8) {
544 printf("ext2fs: non-standard group size unsupported %d\n",
549 fs->e2fs_ipb = fs->e2fs_bsize / EXT2_INODE_SIZE(fs);
550 if (fs->e2fs_ipb == 0 ||
551 fs->e2fs_ipb > fs->e2fs_bsize / E2FS_REV0_INODE_SIZE) {
552 printf("ext2fs: bad inodes per block size\n");
556 fs->e2fs_ipg = es->e2fs_ipg;
557 if (fs->e2fs_ipg < fs->e2fs_ipb || fs->e2fs_ipg > fs->e2fs_bsize * 8) {
558 printf("ext2fs: invalid inodes per group: %u\n", fs->e2fs_ipb);
562 fs->e2fs_itpg = fs->e2fs_ipg / fs->e2fs_ipb;
564 fs->e2fs_bcount = es->e2fs_bcount;
565 fs->e2fs_rbcount = es->e2fs_rbcount;
566 fs->e2fs_fbcount = es->e2fs_fbcount;
567 if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_64BIT)) {
568 fs->e2fs_bcount |= (uint64_t)(es->e4fs_bcount_hi) << 32;
569 fs->e2fs_rbcount |= (uint64_t)(es->e4fs_rbcount_hi) << 32;
570 fs->e2fs_fbcount |= (uint64_t)(es->e4fs_fbcount_hi) << 32;
572 if (fs->e2fs_rbcount > fs->e2fs_bcount ||
573 fs->e2fs_fbcount > fs->e2fs_bcount) {
574 printf("ext2fs: invalid block count\n");
577 if (es->e2fs_first_dblock >= fs->e2fs_bcount) {
578 printf("ext2fs: first data block out of range\n");
582 fs->e2fs_gcount = howmany(fs->e2fs_bcount - es->e2fs_first_dblock,
583 EXT2_BLOCKS_PER_GROUP(fs));
584 if (fs->e2fs_gcount > ((uint64_t)1 << 32) - EXT2_DESCS_PER_BLOCK(fs)) {
585 printf("ext2fs: groups count too large: %u\n", fs->e2fs_gcount);
589 /* Check for extra isize in big inodes. */
590 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_EXTRA_ISIZE) &&
591 EXT2_INODE_SIZE(fs) < sizeof(struct ext2fs_dinode)) {
592 printf("ext2fs: no space for extra inode timestamps\n");
596 /* s_resuid / s_resgid ? */
598 if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_64BIT)) {
599 e2fs_descpb = fs->e2fs_bsize / E2FS_64BIT_GD_SIZE;
600 e2fs_gdbcount_alloc = howmany(fs->e2fs_gcount, e2fs_descpb);
602 e2fs_descpb = fs->e2fs_bsize / E2FS_REV0_GD_SIZE;
603 e2fs_gdbcount_alloc = howmany(fs->e2fs_gcount,
604 fs->e2fs_bsize / sizeof(struct ext2_gd));
606 fs->e2fs_gdbcount = howmany(fs->e2fs_gcount, e2fs_descpb);
607 fs->e2fs_gd = malloc(e2fs_gdbcount_alloc * fs->e2fs_bsize,
608 M_EXT2MNT, M_WAITOK | M_ZERO);
609 fs->e2fs_contigdirs = malloc(fs->e2fs_gcount *
610 sizeof(*fs->e2fs_contigdirs), M_EXT2MNT, M_WAITOK | M_ZERO);
612 for (i = 0; i < fs->e2fs_gdbcount; i++) {
614 fsbtodb(fs, ext2_cg_location(fs, i)),
615 fs->e2fs_bsize, NOCRED, &bp);
618 * fs->e2fs_gd and fs->e2fs_contigdirs
619 * will be freed later by the caller,
620 * because this function could be called from
626 if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_64BIT)) {
628 i * fs->e2fs_bsize / sizeof(struct ext2_gd)],
629 bp->b_data, fs->e2fs_bsize);
631 for (j = 0; j < e2fs_descpb &&
632 g_count < fs->e2fs_gcount; j++, g_count++)
633 memcpy(&fs->e2fs_gd[g_count],
634 bp->b_data + j * E2FS_REV0_GD_SIZE,
641 /* Validate cgs consistency */
642 error = ext2_cg_validate(fs);
647 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_GDT_CSUM) ||
648 EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_METADATA_CKSUM)) {
649 error = ext2_gd_csum_verify(fs, devvp->v_rdev);
653 /* Initialization for the ext2 Orlov allocator variant. */
654 fs->e2fs_total_dir = 0;
655 for (i = 0; i < fs->e2fs_gcount; i++)
656 fs->e2fs_total_dir += e2fs_gd_get_ndirs(&fs->e2fs_gd[i]);
658 if (es->e2fs_rev == E2FS_REV0 ||
659 !EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_LARGEFILE))
660 fs->e2fs_maxfilesize = 0x7fffffff;
662 fs->e2fs_maxfilesize = 0xffffffffffff;
663 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_HUGE_FILE))
664 fs->e2fs_maxfilesize = 0x7fffffffffffffff;
666 if (es->e4fs_flags & E2FS_UNSIGNED_HASH) {
668 } else if ((es->e4fs_flags & E2FS_SIGNED_HASH) == 0) {
669 #ifdef __CHAR_UNSIGNED__
670 es->e4fs_flags |= E2FS_UNSIGNED_HASH;
673 es->e4fs_flags |= E2FS_SIGNED_HASH;
676 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_METADATA_CKSUM))
677 error = ext2_sb_csum_verify(fs);
683 * Reload all incore data for a filesystem (used after running fsck on
684 * the root filesystem and finding things to fix). The filesystem must
685 * be mounted read-only.
687 * Things to do to update the mount:
688 * 1) invalidate all cached meta-data.
689 * 2) re-read superblock from disk.
690 * 3) invalidate all cluster summary information.
691 * 4) invalidate all inactive vnodes.
692 * 5) invalidate all cached file data.
693 * 6) re-read inode data for all active vnodes.
694 * XXX we are missing some steps, in particular # 3, this has to be reviewed.
697 ext2_reload(struct mount *mp, struct thread *td)
699 struct vnode *vp, *mvp, *devvp;
708 if ((mp->mnt_flag & MNT_RDONLY) == 0)
711 * Step 1: invalidate all cached meta-data.
713 devvp = VFSTOEXT2(mp)->um_devvp;
714 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
715 if (vinvalbuf(devvp, 0, 0, 0) != 0)
716 panic("ext2_reload: dirty1");
717 VOP_UNLOCK(devvp, 0);
720 * Step 2: re-read superblock from disk.
721 * constants have been adjusted for ext2
723 if ((error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) != 0)
725 es = (struct ext2fs *)bp->b_data;
726 if (ext2_check_sb_compat(es, devvp->v_rdev, 0) != 0) {
728 return (EIO); /* XXX needs translation */
730 fs = VFSTOEXT2(mp)->um_e2fs;
731 bcopy(bp->b_data, fs->e2fs, sizeof(struct ext2fs));
733 if ((error = ext2_compute_sb_data(devvp, es, fs)) != 0) {
738 if (fs->fs_sbsize < SBSIZE)
739 bp->b_flags |= B_INVAL;
744 * Step 3: invalidate all cluster summary information.
746 if (fs->e2fs_contigsumsize > 0) {
747 lp = fs->e2fs_maxcluster;
748 sump = fs->e2fs_clustersum;
749 for (i = 0; i < fs->e2fs_gcount; i++, sump++) {
750 *lp++ = fs->e2fs_contigsumsize;
752 bzero(sump->cs_sum, fs->e2fs_contigsumsize + 1);
757 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
759 * Step 4: invalidate all cached file data.
761 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td)) {
762 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
765 if (vinvalbuf(vp, 0, 0, 0))
766 panic("ext2_reload: dirty2");
769 * Step 5: re-read inode data for all active vnodes.
772 error = bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
773 (int)fs->e2fs_bsize, NOCRED, &bp);
777 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
781 error = ext2_ei2i((struct ext2fs_dinode *)((char *)bp->b_data +
782 EXT2_INODE_SIZE(fs) * ino_to_fsbo(fs, ip->i_number)), ip);
789 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
797 * Common code for mount and mountroot.
800 ext2_mountfs(struct vnode *devvp, struct mount *mp)
802 struct ext2mount *ump;
806 struct cdev *dev = devvp->v_rdev;
807 struct g_consumer *cp;
815 int32_t e2fs_maxcontig;
817 ronly = vfs_flagopt(mp->mnt_optnew, "ro", NULL, 0);
818 /* XXX: use VOP_ACESS to check FS perms */
820 error = g_vfs_open(devvp, &cp, "ext2fs", ronly ? 0 : 1);
822 VOP_UNLOCK(devvp, 0);
826 /* XXX: should we check for some sectorsize or 512 instead? */
827 if (((SBSIZE % cp->provider->sectorsize) != 0) ||
828 (SBSIZE < cp->provider->sectorsize)) {
835 bo = &devvp->v_bufobj;
837 bo->bo_ops = g_vfs_bufops;
838 if (devvp->v_rdev->si_iosize_max != 0)
839 mp->mnt_iosize_max = devvp->v_rdev->si_iosize_max;
840 if (mp->mnt_iosize_max > MAXPHYS)
841 mp->mnt_iosize_max = MAXPHYS;
845 if ((error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) != 0)
847 es = (struct ext2fs *)bp->b_data;
848 if (ext2_check_sb_compat(es, dev, ronly) != 0) {
849 error = EINVAL; /* XXX needs translation */
852 if ((es->e2fs_state & E2FS_ISCLEAN) == 0 ||
853 (es->e2fs_state & E2FS_ERRORS)) {
854 if (ronly || (mp->mnt_flag & MNT_FORCE)) {
856 "WARNING: Filesystem was not properly dismounted\n");
859 "WARNING: R/W mount denied. Filesystem is not clean - run fsck\n");
864 ump = malloc(sizeof(*ump), M_EXT2MNT, M_WAITOK | M_ZERO);
867 * I don't know whether this is the right strategy. Note that
868 * we dynamically allocate both an m_ext2fs and an ext2fs
869 * while Linux keeps the super block in a locked buffer.
871 ump->um_e2fs = malloc(sizeof(struct m_ext2fs),
872 M_EXT2MNT, M_WAITOK | M_ZERO);
873 ump->um_e2fs->e2fs = malloc(sizeof(struct ext2fs),
874 M_EXT2MNT, M_WAITOK);
875 mtx_init(EXT2_MTX(ump), "EXT2FS", "EXT2FS Lock", MTX_DEF);
876 bcopy(es, ump->um_e2fs->e2fs, (u_int)sizeof(struct ext2fs));
877 if ((error = ext2_compute_sb_data(devvp, ump->um_e2fs->e2fs, ump->um_e2fs)))
881 * Calculate the maximum contiguous blocks and size of cluster summary
882 * array. In FFS this is done by newfs; however, the superblock
883 * in ext2fs doesn't have these variables, so we can calculate
886 e2fs_maxcontig = MAX(1, MAXPHYS / ump->um_e2fs->e2fs_bsize);
887 ump->um_e2fs->e2fs_contigsumsize = MIN(e2fs_maxcontig, EXT2_MAXCONTIG);
888 if (ump->um_e2fs->e2fs_contigsumsize > 0) {
889 size = ump->um_e2fs->e2fs_gcount * sizeof(int32_t);
890 ump->um_e2fs->e2fs_maxcluster = malloc(size, M_EXT2MNT, M_WAITOK);
891 size = ump->um_e2fs->e2fs_gcount * sizeof(struct csum);
892 ump->um_e2fs->e2fs_clustersum = malloc(size, M_EXT2MNT, M_WAITOK);
893 lp = ump->um_e2fs->e2fs_maxcluster;
894 sump = ump->um_e2fs->e2fs_clustersum;
895 for (i = 0; i < ump->um_e2fs->e2fs_gcount; i++, sump++) {
896 *lp++ = ump->um_e2fs->e2fs_contigsumsize;
898 sump->cs_sum = malloc((ump->um_e2fs->e2fs_contigsumsize + 1) *
899 sizeof(int32_t), M_EXT2MNT, M_WAITOK | M_ZERO);
906 fs->e2fs_ronly = ronly; /* ronly is set according to mnt_flags */
909 * If the fs is not mounted read-only, make sure the super block is
910 * always written back on a sync().
912 fs->e2fs_wasvalid = fs->e2fs->e2fs_state & E2FS_ISCLEAN ? 1 : 0;
914 fs->e2fs_fmod = 1; /* mark it modified */
915 fs->e2fs->e2fs_state &= ~E2FS_ISCLEAN; /* set fs invalid */
918 mp->mnt_stat.f_fsid.val[0] = dev2udev(dev);
919 mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
920 mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN;
922 mp->mnt_flag |= MNT_LOCAL;
926 ump->um_devvp = devvp;
927 ump->um_bo = &devvp->v_bufobj;
931 * Setting those two parameters allowed us to use
932 * ufs_bmap w/o changse!
934 ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs);
935 ump->um_bptrtodb = fs->e2fs->e2fs_log_bsize + 1;
936 ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs);
938 ext2_sbupdate(ump, MNT_WAIT);
940 * Initialize filesystem stat information in mount struct.
943 mp->mnt_kern_flag |= MNTK_LOOKUP_SHARED | MNTK_EXTENDED_SHARED |
956 mtx_destroy(EXT2_MTX(ump));
957 free(ump->um_e2fs->e2fs_gd, M_EXT2MNT);
958 free(ump->um_e2fs->e2fs_contigdirs, M_EXT2MNT);
959 free(ump->um_e2fs->e2fs, M_EXT2MNT);
960 free(ump->um_e2fs, M_EXT2MNT);
961 free(ump, M_EXT2MNT);
968 * Unmount system call.
971 ext2_unmount(struct mount *mp, int mntflags)
973 struct ext2mount *ump;
976 int error, flags, i, ronly;
979 if (mntflags & MNT_FORCE) {
980 if (mp->mnt_flag & MNT_ROOTFS)
984 if ((error = ext2_flushfiles(mp, flags, curthread)) != 0)
988 ronly = fs->e2fs_ronly;
989 if (ronly == 0 && ext2_cgupdate(ump, MNT_WAIT) == 0) {
990 if (fs->e2fs_wasvalid)
991 fs->e2fs->e2fs_state |= E2FS_ISCLEAN;
992 ext2_sbupdate(ump, MNT_WAIT);
996 g_vfs_close(ump->um_cp);
998 vrele(ump->um_devvp);
999 sump = fs->e2fs_clustersum;
1000 for (i = 0; i < fs->e2fs_gcount; i++, sump++)
1001 free(sump->cs_sum, M_EXT2MNT);
1002 free(fs->e2fs_clustersum, M_EXT2MNT);
1003 free(fs->e2fs_maxcluster, M_EXT2MNT);
1004 free(fs->e2fs_gd, M_EXT2MNT);
1005 free(fs->e2fs_contigdirs, M_EXT2MNT);
1006 free(fs->e2fs, M_EXT2MNT);
1007 free(fs, M_EXT2MNT);
1008 free(ump, M_EXT2MNT);
1009 mp->mnt_data = NULL;
1011 mp->mnt_flag &= ~MNT_LOCAL;
1017 * Flush out all the files in a filesystem.
1020 ext2_flushfiles(struct mount *mp, int flags, struct thread *td)
1024 error = vflush(mp, 0, flags, td);
1029 * Get filesystem statistics.
1032 ext2_statfs(struct mount *mp, struct statfs *sbp)
1034 struct ext2mount *ump;
1035 struct m_ext2fs *fs;
1036 uint32_t overhead, overhead_per_group, ngdb;
1039 ump = VFSTOEXT2(mp);
1041 if (fs->e2fs->e2fs_magic != E2FS_MAGIC)
1042 panic("ext2_statfs");
1045 * Compute the overhead (FS structures)
1047 overhead_per_group =
1048 1 /* block bitmap */ +
1049 1 /* inode bitmap */ +
1051 overhead = fs->e2fs->e2fs_first_dblock +
1052 fs->e2fs_gcount * overhead_per_group;
1053 if (fs->e2fs->e2fs_rev > E2FS_REV0 &&
1054 fs->e2fs->e2fs_features_rocompat & EXT2F_ROCOMPAT_SPARSESUPER) {
1055 for (i = 0, ngroups = 0; i < fs->e2fs_gcount; i++) {
1056 if (ext2_cg_has_sb(fs, i))
1060 ngroups = fs->e2fs_gcount;
1062 ngdb = fs->e2fs_gdbcount;
1063 if (fs->e2fs->e2fs_rev > E2FS_REV0 &&
1064 fs->e2fs->e2fs_features_compat & EXT2F_COMPAT_RESIZE)
1065 ngdb += fs->e2fs->e2fs_reserved_ngdb;
1066 overhead += ngroups * (1 /* superblock */ + ngdb);
1068 sbp->f_bsize = EXT2_FRAG_SIZE(fs);
1069 sbp->f_iosize = EXT2_BLOCK_SIZE(fs);
1070 sbp->f_blocks = fs->e2fs_bcount - overhead;
1071 sbp->f_bfree = fs->e2fs_fbcount;
1072 sbp->f_bavail = sbp->f_bfree - fs->e2fs_rbcount;
1073 sbp->f_files = fs->e2fs->e2fs_icount;
1074 sbp->f_ffree = fs->e2fs->e2fs_ficount;
1079 * Go through the disk queues to initiate sandbagged IO;
1080 * go through the inodes to write those that have been modified;
1081 * initiate the writing of the super block if it has been modified.
1083 * Note: we are always called with the filesystem marked `MPBUSY'.
1086 ext2_sync(struct mount *mp, int waitfor)
1088 struct vnode *mvp, *vp;
1091 struct ext2mount *ump = VFSTOEXT2(mp);
1092 struct m_ext2fs *fs;
1093 int error, allerror = 0;
1097 if (fs->e2fs_fmod != 0 && fs->e2fs_ronly != 0) { /* XXX */
1098 printf("fs = %s\n", fs->e2fs_fsmnt);
1099 panic("ext2_sync: rofs mod");
1103 * Write back each (modified) inode.
1106 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
1107 if (vp->v_type == VNON) {
1113 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
1114 (vp->v_bufobj.bo_dirty.bv_cnt == 0 ||
1115 waitfor == MNT_LAZY)) {
1119 error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK, td);
1121 if (error == ENOENT) {
1122 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
1127 if ((error = VOP_FSYNC(vp, waitfor, td)) != 0)
1134 * Force stale filesystem control information to be flushed.
1136 if (waitfor != MNT_LAZY) {
1137 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
1138 if ((error = VOP_FSYNC(ump->um_devvp, waitfor, td)) != 0)
1140 VOP_UNLOCK(ump->um_devvp, 0);
1144 * Write back modified superblock.
1146 if (fs->e2fs_fmod != 0) {
1148 fs->e2fs->e2fs_wtime = time_second;
1149 if ((error = ext2_cgupdate(ump, waitfor)) != 0)
1156 * Look up an EXT2FS dinode number to find its incore vnode, otherwise read it
1157 * in from disk. If it is in core, wait for the lock bit to clear, then
1158 * return the inode locked. Detection and handling of mount points must be
1159 * done by the calling routine.
1162 ext2_vget(struct mount *mp, ino_t ino, int flags, struct vnode **vpp)
1164 struct m_ext2fs *fs;
1166 struct ext2mount *ump;
1170 unsigned int i, used_blocks;
1174 error = vfs_hash_get(mp, ino, flags, td, vpp, NULL, NULL);
1175 if (error || *vpp != NULL)
1178 ump = VFSTOEXT2(mp);
1179 ip = malloc(sizeof(struct inode), M_EXT2NODE, M_WAITOK | M_ZERO);
1181 /* Allocate a new vnode/inode. */
1182 if ((error = getnewvnode("ext2fs", mp, &ext2_vnodeops, &vp)) != 0) {
1184 free(ip, M_EXT2NODE);
1189 ip->i_e2fs = fs = ump->um_e2fs;
1193 lockmgr(vp->v_vnlock, LK_EXCLUSIVE, NULL);
1194 error = insmntque(vp, mp);
1196 free(ip, M_EXT2NODE);
1200 error = vfs_hash_insert(vp, ino, flags, td, vpp, NULL, NULL);
1201 if (error || *vpp != NULL)
1204 /* Read in the disk contents for the inode, copy into the inode. */
1205 if ((error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
1206 (int)fs->e2fs_bsize, NOCRED, &bp)) != 0) {
1208 * The inode does not contain anything useful, so it would
1209 * be misleading to leave it on its hash chain. With mode
1210 * still zero, it will be unlinked and returned to the free
1218 /* convert ext2 inode to dinode */
1219 error = ext2_ei2i((struct ext2fs_dinode *)((char *)bp->b_data +
1220 EXT2_INODE_SIZE(fs) * ino_to_fsbo(fs, ino)), ip);
1227 ip->i_block_group = ino_to_cg(fs, ino);
1228 ip->i_next_alloc_block = 0;
1229 ip->i_next_alloc_goal = 0;
1232 * Now we want to make sure that block pointers for unused
1233 * blocks are zeroed out - ext2_balloc depends on this
1234 * although for regular files and directories only
1236 * If IN_E4EXTENTS is enabled, unused blocks are not zeroed
1237 * out because we could corrupt the extent tree.
1239 if (!(ip->i_flag & IN_E4EXTENTS) &&
1240 (S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode))) {
1241 used_blocks = howmany(ip->i_size, fs->e2fs_bsize);
1242 for (i = used_blocks; i < EXT2_NDIR_BLOCKS; i++)
1246 ext2_print_inode(ip);
1247 ext4_ext_print_extent_tree_status(ip);
1252 * Initialize the vnode from the inode, check for aliases.
1253 * Note that the underlying vnode may have changed.
1255 if ((error = ext2_vinit(mp, &ext2_fifoops, &vp)) != 0) {
1262 * Finish inode initialization.
1270 * File handle to vnode
1272 * Have to be really careful about stale file handles:
1273 * - check that the inode number is valid
1274 * - call ext2_vget() to get the locked inode
1275 * - check for an unallocated inode (i_mode == 0)
1276 * - check that the given client host has export rights and return
1277 * those rights via. exflagsp and credanonp
1280 ext2_fhtovp(struct mount *mp, struct fid *fhp, int flags, struct vnode **vpp)
1285 struct m_ext2fs *fs;
1288 ufhp = (struct ufid *)fhp;
1289 fs = VFSTOEXT2(mp)->um_e2fs;
1290 if (ufhp->ufid_ino < EXT2_ROOTINO ||
1291 ufhp->ufid_ino > fs->e2fs_gcount * fs->e2fs->e2fs_ipg)
1294 error = VFS_VGET(mp, ufhp->ufid_ino, LK_EXCLUSIVE, &nvp);
1300 if (ip->i_mode == 0 ||
1301 ip->i_gen != ufhp->ufid_gen || ip->i_nlink <= 0) {
1307 vnode_create_vobject(*vpp, 0, curthread);
1312 * Write a superblock and associated information back to disk.
1315 ext2_sbupdate(struct ext2mount *mp, int waitfor)
1317 struct m_ext2fs *fs = mp->um_e2fs;
1318 struct ext2fs *es = fs->e2fs;
1322 es->e2fs_bcount = fs->e2fs_bcount & 0xffffffff;
1323 es->e2fs_rbcount = fs->e2fs_rbcount & 0xffffffff;
1324 es->e2fs_fbcount = fs->e2fs_fbcount & 0xffffffff;
1325 if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_64BIT)) {
1326 es->e4fs_bcount_hi = fs->e2fs_bcount >> 32;
1327 es->e4fs_rbcount_hi = fs->e2fs_rbcount >> 32;
1328 es->e4fs_fbcount_hi = fs->e2fs_fbcount >> 32;
1331 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_METADATA_CKSUM))
1332 ext2_sb_csum_set(fs);
1334 bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0, 0);
1335 bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2fs));
1336 if (waitfor == MNT_WAIT)
1342 * The buffers for group descriptors, inode bitmaps and block bitmaps
1343 * are not busy at this point and are (hopefully) written by the
1344 * usual sync mechanism. No need to write them here.
1349 ext2_cgupdate(struct ext2mount *mp, int waitfor)
1351 struct m_ext2fs *fs = mp->um_e2fs;
1353 int i, j, g_count = 0, error = 0, allerror = 0;
1355 allerror = ext2_sbupdate(mp, waitfor);
1357 /* Update gd csums */
1358 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_GDT_CSUM) ||
1359 EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_METADATA_CKSUM))
1360 ext2_gd_csum_set(fs);
1362 for (i = 0; i < fs->e2fs_gdbcount; i++) {
1363 bp = getblk(mp->um_devvp, fsbtodb(fs,
1364 ext2_cg_location(fs, i)),
1365 fs->e2fs_bsize, 0, 0, 0);
1366 if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_64BIT)) {
1367 memcpy(bp->b_data, &fs->e2fs_gd[
1368 i * fs->e2fs_bsize / sizeof(struct ext2_gd)],
1371 for (j = 0; j < fs->e2fs_bsize / E2FS_REV0_GD_SIZE &&
1372 g_count < fs->e2fs_gcount; j++, g_count++)
1373 memcpy(bp->b_data + j * E2FS_REV0_GD_SIZE,
1374 &fs->e2fs_gd[g_count], E2FS_REV0_GD_SIZE);
1376 if (waitfor == MNT_WAIT)
1382 if (!allerror && error)
1388 * Return the root of a filesystem.
1391 ext2_root(struct mount *mp, int flags, struct vnode **vpp)
1396 error = VFS_VGET(mp, EXT2_ROOTINO, LK_EXCLUSIVE, &nvp);