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 * Copyright (c) 1989, 1991, 1993, 1994
9 * The Regents of the University of California. All rights reserved.
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
19 * 3. All advertising materials mentioning features or use of this software
20 * must display the following acknowledgement:
21 * This product includes software developed by the University of
22 * California, Berkeley and its contributors.
23 * 4. Neither the name of the University nor the names of its contributors
24 * may be used to endorse or promote products derived from this software
25 * without specific prior written permission.
27 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
28 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
29 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
30 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
31 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
32 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
33 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
34 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
35 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
36 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
39 * @(#)ffs_vfsops.c 8.8 (Berkeley) 4/18/94
40 * $FreeBSD: src/sys/gnu/ext2fs/ext2_vfsops.c,v 1.63.2.7 2002/07/01 00:18:51 iedowse Exp $
43 #include "opt_quota.h"
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/nlookup.h>
50 #include <sys/kernel.h>
51 #include <sys/vnode.h>
52 #include <sys/mount.h>
55 #include <sys/fcntl.h>
56 #include <sys/diskslice.h>
57 #include <sys/malloc.h>
59 #include <vm/vm_zone.h>
62 #include <sys/thread2.h>
67 #include "ext2mount.h"
68 #include "ext2_extern.h"
72 #include "ext2_fs_sb.h"
74 extern struct vop_ops ext2_vnode_vops;
75 extern struct vop_ops ext2_spec_vops;
76 extern struct vop_ops ext2_fifo_vops;
78 static int ext2_fhtovp (struct mount *, struct vnode *,
79 struct fid *, struct vnode **);
80 static int ext2_flushfiles (struct mount *mp, int flags);
81 static int ext2_mount (struct mount *, char *, caddr_t, struct ucred *);
82 static int ext2_mountfs (struct vnode *, struct mount *, struct ucred *);
83 static int ext2_root(struct mount *, struct vnode **);
84 static int ext2_reload (struct mount *mountp, struct ucred *cred);
85 static int ext2_sbupdate (struct ext2mount *, int);
86 static int ext2_statfs (struct mount *, struct statfs *, struct ucred *);
87 static int ext2_sync (struct mount *, int);
88 static int ext2_unmount (struct mount *, int);
89 static int ext2_vget (struct mount *, struct vnode *, ino_t, struct vnode **);
90 static int ext2_init(struct vfsconf *);
91 static int ext2_vptofh (struct vnode *, struct fid *);
93 static MALLOC_DEFINE(M_EXT2NODE, "EXT2 node", "EXT2 vnode private part");
94 MALLOC_DEFINE(M_EXT2MNT, "EXT2 mount", "EXT2 mount structure");
96 static struct vfsops ext2fs_vfsops = {
97 .vfs_mount = ext2_mount,
98 .vfs_unmount = ext2_unmount,
99 .vfs_root = ext2_root, /* root inode via vget */
100 .vfs_quotactl = ext2_quotactl, /* quota operations */
101 .vfs_statfs = ext2_statfs,
102 .vfs_sync = ext2_sync,
103 .vfs_vget = ext2_vget,
104 .vfs_fhtovp = ext2_fhtovp,
105 .vfs_checkexp = ext2_check_export,
106 .vfs_vptofh = ext2_vptofh,
107 .vfs_init = ext2_init,
108 .vfs_uninit = ext2_uninit
111 VFS_SET(ext2fs_vfsops, ext2fs, 0);
112 MODULE_VERSION(ext2fs, 1);
114 #define bsd_malloc kmalloc
115 #define bsd_free kfree
117 static int ext2fs_inode_hash_lock;
119 static int ext2_check_sb_compat (struct ext2_super_block *es,
120 cdev_t dev, int ronly);
121 static int compute_sb_data (struct vnode * devvp,
122 struct ext2_super_block * es,
123 struct ext2_sb_info * fs);
126 ext2_root(struct mount *mp, struct vnode **vpp)
131 error = VFS_VGET(mp, NULL, (ino_t)ROOTINO, &nvp);
139 * Do operations associated with quotas
142 ext2_quotactl(struct mount *mp, int cmds, uid_t uid, caddr_t arg,
148 int cmd, type, error;
150 type = cmds & SUBCMDMASK;
151 cmd = cmds >> SUBCMDSHIFT;
172 error = priv_check_cred(cred, PRIV_UFS_QUOTAON, 0);
176 error = priv_check_cred(cred, PRIV_UFS_QUOTAOFF, 0);
180 error = priv_check_cred(cred, PRIV_VFS_SETQUOTA, 0);
184 error = priv_check_cred(cred, PRIV_UFS_SETUSE, 0);
188 if (uid == cred->cr_ruid)
191 error = priv_check_cred(cred, PRIV_VFS_GETQUOTA, 0);
207 if ((uint)type >= MAXQUOTAS)
209 if (vfs_busy(mp, LK_NOWAIT))
215 error = ext2_quotaon(cred, mp, type, arg);
219 error = ext2_quotaoff(mp, type);
223 error = ext2_setquota(mp, uid, type, arg);
227 error = ext2_setuse(mp, uid, type, arg);
231 error = ext2_getquota(mp, uid, type, arg);
235 error = ext2_qsync(mp);
248 * Initial UFS filesystems, done only once.
251 ext2_init(struct vfsconf *vfsp)
271 * data: this is actually a (struct ext2_args *)
274 ext2_mount(struct mount *mp, char *path, caddr_t data, struct ucred *cred)
277 struct ext2_args args;
278 struct ext2mount *ump = NULL;
279 struct ext2_sb_info *fs;
283 struct nlookupdata nd;
285 if ((error = copyin(data, (caddr_t)&args, sizeof (struct ext2_args))) != 0)
289 * If updating, check whether changing from read-only to
290 * read/write; if there is no device name, that's all we do.
292 if (mp->mnt_flag & MNT_UPDATE) {
295 devvp = ump->um_devvp;
297 if (fs->s_rd_only == 0 && (mp->mnt_flag & MNT_RDONLY)) {
299 if (mp->mnt_flag & MNT_FORCE)
301 if (vfs_busy(mp, LK_NOWAIT))
303 error = ext2_flushfiles(mp, flags);
305 if (!error && fs->s_wasvalid) {
306 fs->s_es->s_state |= EXT2_VALID_FS;
307 ext2_sbupdate(ump, MNT_WAIT);
310 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
311 VOP_OPEN(devvp, FREAD, FSCRED, NULL);
312 VOP_CLOSE(devvp, FREAD|FWRITE, NULL);
315 if (!error && (mp->mnt_flag & MNT_RELOAD))
316 error = ext2_reload(mp, cred);
319 if (ext2_check_sb_compat(fs->s_es, devvp->v_rdev,
320 (mp->mnt_kern_flag & MNTK_WANTRDWR) == 0) != 0)
322 if (fs->s_rd_only && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
324 * If upgrade to read-write by non-root, then verify
325 * that user has necessary permissions on the device.
327 if (cred->cr_uid != 0) {
328 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
329 error = VOP_EACCESS(devvp, VREAD | VWRITE, cred);
337 if ((fs->s_es->s_state & EXT2_VALID_FS) == 0 ||
338 (fs->s_es->s_state & EXT2_ERROR_FS)) {
339 if (mp->mnt_flag & MNT_FORCE) {
341 "WARNING: %s was not properly dismounted\n",
345 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
350 fs->s_es->s_state &= ~EXT2_VALID_FS;
351 ext2_sbupdate(ump, MNT_WAIT);
353 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
354 VOP_OPEN(devvp, FREAD|FWRITE, FSCRED, NULL);
355 VOP_CLOSE(devvp, FREAD, NULL);
358 if (args.fspec == NULL) {
360 * Process export requests.
362 return (vfs_export(mp, &ump->um_export, &args.export));
366 * Not an update, or updating the name: look up the name
367 * and verify that it refers to a sensible block device.
370 error = nlookup_init(&nd, args.fspec, UIO_USERSPACE, NLC_FOLLOW);
372 error = nlookup(&nd);
374 error = cache_vref(&nd.nl_nch, nd.nl_cred, &devvp);
379 if (!vn_isdisk(devvp, &error)) {
385 * If mount by non-root, then verify that user has necessary
386 * permissions on the device.
388 if (cred->cr_uid != 0) {
390 if ((mp->mnt_flag & MNT_RDONLY) == 0)
391 accessmode |= VWRITE;
392 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
393 if ((error = VOP_EACCESS(devvp, accessmode, cred)) != 0) {
400 if ((mp->mnt_flag & MNT_UPDATE) == 0) {
401 error = ext2_mountfs(devvp, mp, cred);
403 if (devvp != ump->um_devvp)
404 error = EINVAL; /* needs translation */
414 copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size);
415 bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size);
416 copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, &size);
417 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
418 ext2_statfs(mp, &mp->mnt_stat, cred);
423 * checks that the data in the descriptor blocks make sense
424 * this is taken from ext2/super.c
427 ext2_check_descriptors(struct ext2_sb_info *sb)
431 unsigned long block = sb->s_es->s_first_data_block;
432 struct ext2_group_desc * gdp = NULL;
434 /* ext2_debug ("Checking group descriptors"); */
436 for (i = 0; i < sb->s_groups_count; i++)
438 /* examine next descriptor block */
439 if ((i % EXT2_DESC_PER_BLOCK(sb)) == 0)
440 gdp = (struct ext2_group_desc *)
441 sb->s_group_desc[desc_block++]->b_data;
442 if (gdp->bg_block_bitmap < block ||
443 gdp->bg_block_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb))
445 kprintf ("ext2_check_descriptors: "
446 "Block bitmap for group %d"
447 " not in group (block %lu)!\n",
448 i, (unsigned long) gdp->bg_block_bitmap);
451 if (gdp->bg_inode_bitmap < block ||
452 gdp->bg_inode_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb))
454 kprintf ("ext2_check_descriptors: "
455 "Inode bitmap for group %d"
456 " not in group (block %lu)!\n",
457 i, (unsigned long) gdp->bg_inode_bitmap);
460 if (gdp->bg_inode_table < block ||
461 gdp->bg_inode_table + sb->s_itb_per_group >=
462 block + EXT2_BLOCKS_PER_GROUP(sb))
464 kprintf ("ext2_check_descriptors: "
465 "Inode table for group %d"
466 " not in group (block %lu)!\n",
467 i, (unsigned long) gdp->bg_inode_table);
470 block += EXT2_BLOCKS_PER_GROUP(sb);
477 ext2_check_sb_compat(struct ext2_super_block *es, cdev_t dev, int ronly)
479 if (es->s_magic != EXT2_SUPER_MAGIC) {
480 kprintf("ext2fs: %s: wrong magic number %#x (expected %#x)\n",
481 devtoname(dev), es->s_magic, EXT2_SUPER_MAGIC);
484 if (es->s_rev_level > EXT2_GOOD_OLD_REV) {
485 if (es->s_feature_incompat & ~EXT2_FEATURE_INCOMPAT_SUPP) {
487 "WARNING: mount of %s denied due to unsupported optional features\n",
492 (es->s_feature_ro_compat & ~EXT2_FEATURE_RO_COMPAT_SUPP)) {
494 "WARNING: R/W mount of %s denied due to unsupported optional features\n",
503 * this computes the fields of the ext2_sb_info structure from the
504 * data in the ext2_super_block structure read in
507 compute_sb_data(struct vnode *devvp, struct ext2_super_block *es,
508 struct ext2_sb_info *fs)
512 int logic_sb_block = 1; /* XXX for now */
517 #define V(v) kprintf(#v"= %d\n", fs->v);
520 fs->s_blocksize = EXT2_MIN_BLOCK_SIZE << es->s_log_block_size;
522 fs->s_bshift = EXT2_MIN_BLOCK_LOG_SIZE + es->s_log_block_size;
524 fs->s_fsbtodb = es->s_log_block_size + 1;
526 fs->s_qbmask = fs->s_blocksize - 1;
528 fs->s_blocksize_bits = EXT2_BLOCK_SIZE_BITS(es);
530 fs->s_frag_size = EXT2_MIN_FRAG_SIZE << es->s_log_frag_size;
533 fs->s_frags_per_block = fs->s_blocksize / fs->s_frag_size;
535 fs->s_blocks_per_group = es->s_blocks_per_group;
536 V(s_blocks_per_group)
537 fs->s_frags_per_group = es->s_frags_per_group;
539 fs->s_inodes_per_group = es->s_inodes_per_group;
540 V(s_inodes_per_group)
541 if (es->s_rev_level == EXT2_GOOD_OLD_REV) {
542 fs->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
543 fs->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
545 fs->s_first_ino = es->s_first_ino;
546 fs->s_inode_size = es->s_inode_size;
548 * Simple sanity check for superblock inode size value.
550 if (fs->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE ||
551 fs->s_inode_size > fs->s_blocksize ||
552 (fs->s_inode_size & (fs->s_inode_size - 1)) != 0) {
553 kprintf("EXT2-fs: invalid inode size %d\n",
560 fs->s_inodes_per_block = fs->s_blocksize / EXT2_INODE_SIZE(fs);
561 V(s_inodes_per_block)
562 fs->s_itb_per_group = fs->s_inodes_per_group /fs->s_inodes_per_block;
564 fs->s_desc_per_block = fs->s_blocksize / sizeof (struct ext2_group_desc);
566 /* s_resuid / s_resgid ? */
567 fs->s_groups_count = (es->s_blocks_count -
568 es->s_first_data_block +
569 EXT2_BLOCKS_PER_GROUP(fs) - 1) /
570 EXT2_BLOCKS_PER_GROUP(fs);
572 db_count = (fs->s_groups_count + EXT2_DESC_PER_BLOCK(fs) - 1) /
573 EXT2_DESC_PER_BLOCK(fs);
574 fs->s_db_per_group = db_count;
577 fs->s_group_desc = bsd_malloc(db_count * sizeof (struct buf *),
578 M_EXT2MNT, M_WAITOK);
580 /* adjust logic_sb_block */
581 if(fs->s_blocksize > SBSIZE)
582 /* Godmar thinks: if the blocksize is greater than 1024, then
583 the superblock is logically part of block zero.
587 for (i = 0; i < db_count; i++) {
588 error = bread(devvp, fsbtodoff(fs, logic_sb_block + i + 1),
589 fs->s_blocksize, &fs->s_group_desc[i]);
591 for (j = 0; j < i; j++)
592 brelse(fs->s_group_desc[j]);
593 bsd_free(fs->s_group_desc, M_EXT2MNT);
594 kprintf("EXT2-fs: unable to read group descriptors (%d)\n", error);
597 /* Set the B_LOCKED flag on the buffer, then brelse() it */
598 LCK_BUF(fs->s_group_desc[i])
600 if(!ext2_check_descriptors(fs)) {
601 for (j = 0; j < db_count; j++)
602 ULCK_BUF(fs->s_group_desc[j])
603 bsd_free(fs->s_group_desc, M_EXT2MNT);
604 kprintf("EXT2-fs: (ext2_check_descriptors failure) "
605 "unable to read group descriptors\n");
609 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) {
610 fs->s_inode_bitmap_number[i] = 0;
611 fs->s_inode_bitmap[i] = NULL;
612 fs->s_block_bitmap_number[i] = 0;
613 fs->s_block_bitmap[i] = NULL;
615 fs->s_loaded_inode_bitmaps = 0;
616 fs->s_loaded_block_bitmaps = 0;
621 * Reload all incore data for a filesystem (used after running fsck on
622 * the root filesystem and finding things to fix). The filesystem must
623 * be mounted read-only.
625 * Things to do to update the mount:
626 * 1) invalidate all cached meta-data.
627 * 2) re-read superblock from disk.
628 * 3) re-read summary information from disk.
629 * 4) invalidate all inactive vnodes.
630 * 5) invalidate all cached file data.
631 * 6) re-read inode data for all active vnodes.
633 static int ext2_reload_scan(struct mount *mp, struct vnode *vp, void *rescan);
640 struct ext2_sb_info *fs;
644 ext2_reload(struct mount *mountp, struct ucred *cred)
648 struct ext2_super_block * es;
649 struct ext2_sb_info *fs;
651 struct scaninfo scaninfo;
653 if ((mountp->mnt_flag & MNT_RDONLY) == 0)
656 * Step 1: invalidate all cached meta-data.
658 devvp = VFSTOEXT2(mountp)->um_devvp;
659 if (vinvalbuf(devvp, 0, 0, 0))
660 panic("ext2_reload: dirty1");
662 * Step 2: re-read superblock from disk.
663 * constants have been adjusted for ext2
665 if ((error = bread(devvp, SBOFF, SBSIZE, &bp)) != 0)
667 es = (struct ext2_super_block *)bp->b_data;
668 if (ext2_check_sb_compat(es, devvp->v_rdev, 0) != 0) {
670 return (EIO); /* XXX needs translation */
672 fs = VFSTOEXT2(mountp)->um_e2fs;
673 bcopy(bp->b_data, fs->s_es, sizeof(struct ext2_super_block));
675 if((error = compute_sb_data(devvp, es, fs)) != 0) {
680 if (fs->fs_sbsize < SBSIZE)
681 bp->b_flags |= B_INVAL;
686 scaninfo.devvp = devvp;
688 while (error == 0 && scaninfo.rescan) {
690 error = vmntvnodescan(mountp, VMSC_GETVX,
691 NULL, ext2_reload_scan, &scaninfo);
697 ext2_reload_scan(struct mount *mp, struct vnode *vp, void *data)
699 struct scaninfo *info = data;
711 * Step 5: invalidate all cached file data.
713 if (vinvalbuf(vp, 0, 0, 0))
714 panic("ext2_reload: dirty2");
716 * Step 6: re-read inode data for all active vnodes.
719 error = bread(info->devvp,
720 fsbtodoff(info->fs, ino_to_fsba(info->fs, ip->i_number)),
721 (int)info->fs->s_blocksize, &bp);
724 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data +
725 EXT2_INODE_SIZE(info->fs) * ino_to_fsbo(info->fs, ip->i_number)),
732 * Common code for mount and mountroot
735 ext2_mountfs(struct vnode *devvp, struct mount *mp, struct ucred *cred)
737 struct ext2mount *ump;
739 struct ext2_sb_info *fs;
740 struct ext2_super_block * es;
742 struct partinfo dpart;
747 * Disallow multiple mounts of the same device.
748 * Disallow mounting of a device that is currently in use
749 * (except for root, which might share swap device for miniroot).
750 * Flush out any old buffers remaining from a previous use.
752 if ((error = vfs_mountedon(devvp)) != 0)
754 if (vcount(devvp) > 0)
756 if ((error = vinvalbuf(devvp, V_SAVE, 0, 0)) != 0)
759 /* turn on this to force it to be read-only */
760 mp->mnt_flag |= MNT_RDONLY;
763 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
764 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
765 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, NULL);
770 if (dev->si_iosize_max != 0)
771 mp->mnt_iosize_max = dev->si_iosize_max;
772 if (mp->mnt_iosize_max > MAXPHYS)
773 mp->mnt_iosize_max = MAXPHYS;
774 VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, cred, NULL);
778 if ((error = bread(devvp, SBOFF, SBSIZE, &bp)) != 0)
780 es = (struct ext2_super_block *)bp->b_data;
781 if (ext2_check_sb_compat(es, dev, ronly) != 0) {
782 error = EINVAL; /* XXX needs translation */
785 if ((es->s_state & EXT2_VALID_FS) == 0 ||
786 (es->s_state & EXT2_ERROR_FS)) {
787 if (ronly || (mp->mnt_flag & MNT_FORCE)) {
789 "WARNING: Filesystem was not properly dismounted\n");
792 "WARNING: R/W mount denied. Filesystem is not clean - run fsck\n");
797 ump = bsd_malloc(sizeof *ump, M_EXT2MNT, M_WAITOK | M_ZERO);
798 ump->um_malloctype = M_EXT2NODE;
799 ump->um_blkatoff = ext2_blkatoff;
800 ump->um_truncate = ext2_truncate;
801 ump->um_update = ext2_update;
802 ump->um_valloc = ext2_valloc;
803 ump->um_vfree = ext2_vfree;
804 /* I don't know whether this is the right strategy. Note that
805 we dynamically allocate both a ext2_sb_info and a ext2_super_block
806 while Linux keeps the super block in a locked buffer
808 ump->um_e2fs = bsd_malloc(sizeof(struct ext2_sb_info),
809 M_EXT2MNT, M_WAITOK);
810 ump->um_e2fs->s_es = bsd_malloc(sizeof(struct ext2_super_block),
811 M_EXT2MNT, M_WAITOK);
812 bcopy(es, ump->um_e2fs->s_es, (u_int)sizeof(struct ext2_super_block));
813 if ((error = compute_sb_data(devvp, ump->um_e2fs->s_es, ump->um_e2fs)))
816 * We don't free the group descriptors allocated by compute_sb_data()
817 * until ext2_unmount(). This is OK since the mount will succeed.
822 fs->s_rd_only = ronly; /* ronly is set according to mnt_flags */
823 /* if the fs is not mounted read-only, make sure the super block is
824 always written back on a sync()
826 fs->s_wasvalid = fs->s_es->s_state & EXT2_VALID_FS ? 1 : 0;
828 fs->s_dirt = 1; /* mark it modified */
829 fs->s_es->s_state &= ~EXT2_VALID_FS; /* set fs invalid */
831 mp->mnt_data = (qaddr_t)ump;
832 mp->mnt_stat.f_fsid.val[0] = dev2udev(dev);
833 mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
834 mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN;
835 mp->mnt_flag |= MNT_LOCAL;
838 ump->um_devvp = devvp;
839 /* setting those two parameters allows us to use
840 ext2_bmap w/o changse !
842 ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs);
843 ump->um_bptrtodb = fs->s_es->s_log_block_size + 1;
844 ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs);
845 for (i = 0; i < MAXQUOTAS; i++)
846 ump->um_quotas[i] = NULLVP;
847 dev->si_mountpoint = mp;
849 vfs_add_vnodeops(mp, &ext2_vnode_vops, &mp->mnt_vn_norm_ops);
850 vfs_add_vnodeops(mp, &ext2_spec_vops, &mp->mnt_vn_spec_ops);
851 vfs_add_vnodeops(mp, &ext2_fifo_vops, &mp->mnt_vn_fifo_ops);
854 ext2_sbupdate(ump, MNT_WAIT);
859 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
860 VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, NULL);
863 bsd_free(ump->um_e2fs->s_es, M_EXT2MNT);
864 bsd_free(ump->um_e2fs, M_EXT2MNT);
865 bsd_free(ump, M_EXT2MNT);
866 mp->mnt_data = (qaddr_t)0;
872 * unmount system call
875 ext2_unmount(struct mount *mp, int mntflags)
877 struct ext2mount *ump;
878 struct ext2_sb_info *fs;
879 int error, flags, ronly, i;
882 if (mntflags & MNT_FORCE) {
883 if (mp->mnt_flag & MNT_ROOTFS)
887 if ((error = ext2_flushfiles(mp, flags)) != 0)
891 ronly = fs->s_rd_only;
894 fs->s_es->s_state |= EXT2_VALID_FS;
895 ext2_sbupdate(ump, MNT_WAIT);
898 /* release buffers containing group descriptors */
899 for(i = 0; i < fs->s_db_per_group; i++)
900 ULCK_BUF(fs->s_group_desc[i])
901 bsd_free(fs->s_group_desc, M_EXT2MNT);
903 /* release cached inode/block bitmaps */
904 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
905 if (fs->s_inode_bitmap[i])
906 ULCK_BUF(fs->s_inode_bitmap[i])
908 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
909 if (fs->s_block_bitmap[i])
910 ULCK_BUF(fs->s_block_bitmap[i])
912 ump->um_devvp->v_rdev->si_mountpoint = NULL;
914 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
915 error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD|FWRITE, NULL);
916 vn_unlock(ump->um_devvp);
918 vrele(ump->um_devvp);
919 bsd_free(fs->s_es, M_EXT2MNT);
920 bsd_free(fs, M_EXT2MNT);
921 bsd_free(ump, M_EXT2MNT);
922 mp->mnt_data = (qaddr_t)0;
923 mp->mnt_flag &= ~MNT_LOCAL;
928 * Flush out all the files in a filesystem.
931 ext2_flushfiles(struct mount *mp, int flags)
933 struct ext2mount *ump;
941 if (mp->mnt_flag & MNT_QUOTA) {
942 if ((error = vflush(mp, 0, SKIPSYSTEM|flags)) != 0)
944 for (i = 0; i < MAXQUOTAS; i++) {
945 if (ump->um_quotas[i] == NULLVP)
947 ext2_quotaoff(mp, i);
950 * Here we fall through to vflush again to ensure
951 * that we have gotten rid of all the system vnodes.
955 error = vflush(mp, 0, flags);
960 * Get file system statistics.
961 * taken from ext2/super.c ext2_statfs
964 ext2_statfs(struct mount *mp, struct statfs *sbp, struct ucred *cred)
966 unsigned long overhead;
967 struct ext2mount *ump;
968 struct ext2_sb_info *fs;
969 struct ext2_super_block *es;
976 if (es->s_magic != EXT2_SUPER_MAGIC)
977 panic("ext2_statfs - magic number spoiled");
980 * Compute the overhead (FS structures)
982 if (es->s_feature_ro_compat & EXT2_FEATURE_RO_COMPAT_SPARSE_SUPER) {
984 for (i = 0 ; i < fs->s_groups_count; i++)
985 if (ext2_group_sparse(i))
988 nsb = fs->s_groups_count;
989 overhead = es->s_first_data_block +
990 /* Superblocks and block group descriptors: */
991 nsb * (1 + fs->s_db_per_group) +
992 /* Inode bitmap, block bitmap, and inode table: */
993 fs->s_groups_count * (1 + 1 + fs->s_itb_per_group);
995 sbp->f_bsize = EXT2_FRAG_SIZE(fs);
996 sbp->f_iosize = EXT2_BLOCK_SIZE(fs);
997 sbp->f_blocks = es->s_blocks_count - overhead;
998 sbp->f_bfree = es->s_free_blocks_count;
999 sbp->f_bavail = sbp->f_bfree - es->s_r_blocks_count;
1000 sbp->f_files = es->s_inodes_count;
1001 sbp->f_ffree = es->s_free_inodes_count;
1002 if (sbp != &mp->mnt_stat) {
1003 sbp->f_type = mp->mnt_vfc->vfc_typenum;
1004 bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
1005 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
1011 * Go through the disk queues to initiate sandbagged IO;
1012 * go through the inodes to write those that have been modified;
1013 * initiate the writing of the super block if it has been modified.
1015 * Note: we are always called with the filesystem marked `MPBUSY'.
1018 static int ext2_sync_scan(struct mount *mp, struct vnode *vp, void *data);
1021 ext2_sync(struct mount *mp, int waitfor)
1023 struct ext2mount *ump = VFSTOEXT2(mp);
1024 struct ext2_sb_info *fs;
1025 struct scaninfo scaninfo;
1029 if (fs->s_dirt != 0 && fs->s_rd_only != 0) { /* XXX */
1030 kprintf("fs = %s\n", fs->fs_fsmnt);
1031 panic("ext2_sync: rofs mod");
1035 * Write back each (modified) inode.
1037 scaninfo.allerror = 0;
1038 scaninfo.rescan = 1;
1039 scaninfo.waitfor = waitfor;
1040 while (scaninfo.rescan) {
1041 scaninfo.rescan = 0;
1042 vmntvnodescan(mp, VMSC_GETVP|VMSC_NOWAIT,
1043 NULL, ext2_sync_scan, &scaninfo);
1047 * Force stale file system control information to be flushed.
1049 if ((waitfor & MNT_LAZY) == 0) {
1050 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
1051 if ((error = VOP_FSYNC(ump->um_devvp, waitfor, 0)) != 0)
1052 scaninfo.allerror = error;
1053 vn_unlock(ump->um_devvp);
1059 * Write back modified superblock.
1061 if (fs->s_dirt != 0) {
1063 fs->s_es->s_wtime = time_second;
1064 if ((error = ext2_sbupdate(ump, waitfor)) != 0)
1065 scaninfo.allerror = error;
1067 return (scaninfo.allerror);
1071 ext2_sync_scan(struct mount *mp, struct vnode *vp, void *data)
1073 struct scaninfo *info = data;
1078 if (vp->v_type == VNON ||
1080 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
1081 (RB_EMPTY(&vp->v_rbdirty_tree) || (info->waitfor & MNT_LAZY)))) {
1084 if ((error = VOP_FSYNC(vp, info->waitfor, 0)) != 0)
1085 info->allerror = error;
1090 * Look up a EXT2FS dinode number to find its incore vnode, otherwise read it
1091 * in from disk. If it is in core, wait for the lock bit to clear, then
1092 * return the inode locked. Detection and handling of mount points must be
1093 * done by the calling routine.
1096 ext2_vget(struct mount *mp, struct vnode *dvp, ino_t ino, struct vnode **vpp)
1098 struct ext2_sb_info *fs;
1100 struct ext2mount *ump;
1107 ump = VFSTOEXT2(mp);
1110 if ((*vpp = ext2_ihashget(dev, ino)) != NULL)
1114 * Lock out the creation of new entries in the FFS hash table in
1115 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate
1118 if (ext2fs_inode_hash_lock) {
1119 while (ext2fs_inode_hash_lock) {
1120 ext2fs_inode_hash_lock = -1;
1121 tsleep(&ext2fs_inode_hash_lock, 0, "e2vget", 0);
1125 ext2fs_inode_hash_lock = 1;
1128 * If this MALLOC() is performed after the getnewvnode()
1129 * it might block, leaving a vnode with a NULL v_data to be
1130 * found by ext2_sync() if a sync happens to fire right then,
1131 * which will cause a panic because ext2_sync() blindly
1132 * dereferences vp->v_data (as well it should).
1134 ip = kmalloc(sizeof(struct inode), M_EXT2NODE, M_WAITOK);
1136 /* Allocate a new vnode/inode. */
1137 if ((error = getnewvnode(VT_EXT2FS, mp, &vp, 0, LK_CANRECURSE)) != 0) {
1138 if (ext2fs_inode_hash_lock < 0)
1139 wakeup(&ext2fs_inode_hash_lock);
1140 ext2fs_inode_hash_lock = 0;
1142 kfree(ip, M_EXT2NODE);
1145 bzero((caddr_t)ip, sizeof(struct inode));
1148 ip->i_e2fs = fs = ump->um_e2fs;
1152 for (i = 0; i < MAXQUOTAS; i++)
1153 ip->i_dquot[i] = NODQUOT;
1156 * Put it onto its hash chain. Since our vnode is locked, other
1157 * requests for this inode will block if they arrive while we are
1158 * sleeping waiting for old data structures to be purged or for the
1159 * contents of the disk portion of this inode to be read.
1163 if (ext2fs_inode_hash_lock < 0)
1164 wakeup(&ext2fs_inode_hash_lock);
1165 ext2fs_inode_hash_lock = 0;
1167 /* Read in the disk contents for the inode, copy into the inode. */
1169 kprintf("ext2_vget(%d) dbn= %d ", ino, fsbtodb(fs, ino_to_fsba(fs, ino)));
1171 error = bread(ump->um_devvp, fsbtodoff(fs, ino_to_fsba(fs, ino)),
1172 (int)fs->s_blocksize, &bp);
1175 * The inode does not contain anything useful, so it would
1176 * be misleading to leave it on its hash chain. With mode
1177 * still zero, it will be unlinked and returned to the free
1185 /* convert ext2 inode to dinode */
1186 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data + EXT2_INODE_SIZE(fs) *
1187 ino_to_fsbo(fs, ino)), &ip->i_din);
1188 ip->i_block_group = ino_to_cg(fs, ino);
1189 ip->i_next_alloc_block = 0;
1190 ip->i_next_alloc_goal = 0;
1191 ip->i_prealloc_count = 0;
1192 ip->i_prealloc_block = 0;
1193 /* now we want to make sure that block pointers for unused
1194 blocks are zeroed out - ext2_balloc depends on this
1195 although for regular files and directories only
1197 if(S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode)) {
1198 used_blocks = (ip->i_size+fs->s_blocksize-1) / fs->s_blocksize;
1199 for(i = used_blocks; i < EXT2_NDIR_BLOCKS; i++)
1203 ext2_print_inode(ip);
1208 * Initialize the vnode from the inode, check for aliases.
1209 * Note that the underlying vnode may have changed.
1211 if ((error = ext2_vinit(mp, &vp)) != 0) {
1218 * Finish inode initialization now that aliasing has been resolved.
1220 ip->i_devvp = ump->um_devvp;
1223 * Set up a generation number for this inode if it does not
1224 * already have one. This should only happen on old filesystems.
1226 if (ip->i_gen == 0) {
1227 ip->i_gen = krandom() / 2 + 1;
1228 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1229 ip->i_flag |= IN_MODIFIED;
1232 * Return the locked and refd vnode.
1239 * File handle to vnode
1241 * Have to be really careful about stale file handles:
1242 * - check that the inode number is valid
1243 * - call ext2_vget() to get the locked inode
1244 * - check for an unallocated inode (i_mode == 0)
1245 * - check that the given client host has export rights and return
1246 * those rights via. exflagsp and credanonp
1249 ext2_fhtovp(struct mount *mp, struct vnode *rootvp,
1250 struct fid *fhp, struct vnode **vpp)
1253 struct ext2_sb_info *fs;
1258 ufhp = (struct ufid *)fhp;
1259 fs = VFSTOEXT2(mp)->um_e2fs;
1260 if (ufhp->ufid_ino < ROOTINO ||
1261 ufhp->ufid_ino > fs->s_groups_count * fs->s_es->s_inodes_per_group)
1264 error = VFS_VGET(mp, rootvp, ufhp->ufid_ino, &nvp);
1270 if (ip->i_mode == 0 ||
1271 ip->i_gen != ufhp->ufid_gen ||
1272 (VFSTOEXT2(mp)->um_i_effnlink_valid ? ip->i_effnlink :
1273 ip->i_nlink) <= 0) {
1283 * Vnode pointer to File handle
1287 ext2_vptofh(struct vnode *vp, struct fid *fhp)
1293 ufhp = (struct ufid *)fhp;
1294 ufhp->ufid_len = sizeof(struct ufid);
1295 ufhp->ufid_ino = ip->i_number;
1296 ufhp->ufid_gen = ip->i_gen;
1301 * This is the generic part of fhtovp called after the underlying
1302 * filesystem has validated the file handle.
1304 * Verify that a host should have access to a filesystem.
1307 ext2_check_export(struct mount *mp, struct sockaddr *nam, int *exflagsp,
1308 struct ucred **credanonp)
1311 struct ext2mount *ump;
1313 ump = VFSTOEXT2(mp);
1315 * Get the export permission structure for this <mp, client> tuple.
1317 np = vfs_export_lookup(mp, &ump->um_export, nam);
1321 *exflagsp = np->netc_exflags;
1322 *credanonp = &np->netc_anon;
1327 * Write a superblock and associated information back to disk.
1330 ext2_sbupdate(struct ext2mount *mp, int waitfor)
1332 struct ext2_sb_info *fs = mp->um_e2fs;
1333 struct ext2_super_block *es = fs->s_es;
1337 kprintf("\nupdating superblock, waitfor=%s\n", waitfor == MNT_WAIT ? "yes":"no");
1339 bp = getblk(mp->um_devvp, SBOFF, SBSIZE, 0, 0);
1340 bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2_super_block));
1341 if (waitfor == MNT_WAIT)
1347 * The buffers for group descriptors, inode bitmaps and block bitmaps
1348 * are not busy at this point and are (hopefully) written by the
1349 * usual sync mechanism. No need to write them here