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 = 0;
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);
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);
358 if (args.fspec == 0) {
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_scan1(struct mount *mp, struct vnode *vp, void *rescan);
634 static int ext2_reload_scan2(struct mount *mp, struct vnode *vp, void *rescan);
641 struct ext2_sb_info *fs;
645 ext2_reload(struct mount *mountp, struct ucred *cred)
649 struct ext2_super_block * es;
650 struct ext2_sb_info *fs;
652 struct scaninfo scaninfo;
654 if ((mountp->mnt_flag & MNT_RDONLY) == 0)
657 * Step 1: invalidate all cached meta-data.
659 devvp = VFSTOEXT2(mountp)->um_devvp;
660 if (vinvalbuf(devvp, 0, 0, 0))
661 panic("ext2_reload: dirty1");
663 * Step 2: re-read superblock from disk.
664 * constants have been adjusted for ext2
666 if ((error = bread(devvp, SBOFF, SBSIZE, &bp)) != 0)
668 es = (struct ext2_super_block *)bp->b_data;
669 if (ext2_check_sb_compat(es, devvp->v_rdev, 0) != 0) {
671 return (EIO); /* XXX needs translation */
673 fs = VFSTOEXT2(mountp)->um_e2fs;
674 bcopy(bp->b_data, fs->s_es, sizeof(struct ext2_super_block));
676 if((error = compute_sb_data(devvp, es, fs)) != 0) {
681 if (fs->fs_sbsize < SBSIZE)
682 bp->b_flags |= B_INVAL;
687 scaninfo.devvp = devvp;
689 while (error == 0 && scaninfo.rescan) {
691 error = vmntvnodescan(mountp, VMSC_GETVX, ext2_reload_scan1,
692 ext2_reload_scan2, &scaninfo);
698 ext2_reload_scan1(struct mount *mp, struct vnode *vp, void *data)
700 /*struct scaninfo *info = data;*/
706 ext2_reload_scan2(struct mount *mp, struct vnode *vp, void *data)
708 struct scaninfo *info = data;
720 * Step 5: invalidate all cached file data.
722 if (vinvalbuf(vp, 0, 0, 0))
723 panic("ext2_reload: dirty2");
725 * Step 6: re-read inode data for all active vnodes.
728 error = bread(info->devvp,
729 fsbtodoff(info->fs, ino_to_fsba(info->fs, ip->i_number)),
730 (int)info->fs->s_blocksize, &bp);
733 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data +
734 EXT2_INODE_SIZE(info->fs) * ino_to_fsbo(info->fs, ip->i_number)),
741 * Common code for mount and mountroot
744 ext2_mountfs(struct vnode *devvp, struct mount *mp, struct ucred *cred)
746 struct ext2mount *ump;
748 struct ext2_sb_info *fs;
749 struct ext2_super_block * es;
751 struct partinfo dpart;
756 * Disallow multiple mounts of the same device.
757 * Disallow mounting of a device that is currently in use
758 * (except for root, which might share swap device for miniroot).
759 * Flush out any old buffers remaining from a previous use.
761 if ((error = vfs_mountedon(devvp)) != 0)
763 if (vcount(devvp) > 0)
765 if ((error = vinvalbuf(devvp, V_SAVE, 0, 0)) != 0)
768 /* turn on this to force it to be read-only */
769 mp->mnt_flag |= MNT_RDONLY;
772 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
773 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
774 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, NULL);
779 if (dev->si_iosize_max != 0)
780 mp->mnt_iosize_max = dev->si_iosize_max;
781 if (mp->mnt_iosize_max > MAXPHYS)
782 mp->mnt_iosize_max = MAXPHYS;
783 VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, cred, NULL);
787 if ((error = bread(devvp, SBOFF, SBSIZE, &bp)) != 0)
789 es = (struct ext2_super_block *)bp->b_data;
790 if (ext2_check_sb_compat(es, dev, ronly) != 0) {
791 error = EINVAL; /* XXX needs translation */
794 if ((es->s_state & EXT2_VALID_FS) == 0 ||
795 (es->s_state & EXT2_ERROR_FS)) {
796 if (ronly || (mp->mnt_flag & MNT_FORCE)) {
798 "WARNING: Filesystem was not properly dismounted\n");
801 "WARNING: R/W mount denied. Filesystem is not clean - run fsck\n");
806 ump = bsd_malloc(sizeof *ump, M_EXT2MNT, M_WAITOK | M_ZERO);
807 ump->um_malloctype = M_EXT2NODE;
808 ump->um_blkatoff = ext2_blkatoff;
809 ump->um_truncate = ext2_truncate;
810 ump->um_update = ext2_update;
811 ump->um_valloc = ext2_valloc;
812 ump->um_vfree = ext2_vfree;
813 /* I don't know whether this is the right strategy. Note that
814 we dynamically allocate both a ext2_sb_info and a ext2_super_block
815 while Linux keeps the super block in a locked buffer
817 ump->um_e2fs = bsd_malloc(sizeof(struct ext2_sb_info),
818 M_EXT2MNT, M_WAITOK);
819 ump->um_e2fs->s_es = bsd_malloc(sizeof(struct ext2_super_block),
820 M_EXT2MNT, M_WAITOK);
821 bcopy(es, ump->um_e2fs->s_es, (u_int)sizeof(struct ext2_super_block));
822 if ((error = compute_sb_data(devvp, ump->um_e2fs->s_es, ump->um_e2fs)))
825 * We don't free the group descriptors allocated by compute_sb_data()
826 * until ext2_unmount(). This is OK since the mount will succeed.
831 fs->s_rd_only = ronly; /* ronly is set according to mnt_flags */
832 /* if the fs is not mounted read-only, make sure the super block is
833 always written back on a sync()
835 fs->s_wasvalid = fs->s_es->s_state & EXT2_VALID_FS ? 1 : 0;
837 fs->s_dirt = 1; /* mark it modified */
838 fs->s_es->s_state &= ~EXT2_VALID_FS; /* set fs invalid */
840 mp->mnt_data = (qaddr_t)ump;
841 mp->mnt_stat.f_fsid.val[0] = dev2udev(dev);
842 mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
843 mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN;
844 mp->mnt_flag |= MNT_LOCAL;
847 ump->um_devvp = devvp;
848 /* setting those two parameters allows us to use
849 ext2_bmap w/o changse !
851 ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs);
852 ump->um_bptrtodb = fs->s_es->s_log_block_size + 1;
853 ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs);
854 for (i = 0; i < MAXQUOTAS; i++)
855 ump->um_quotas[i] = NULLVP;
856 dev->si_mountpoint = mp;
858 vfs_add_vnodeops(mp, &ext2_vnode_vops, &mp->mnt_vn_norm_ops);
859 vfs_add_vnodeops(mp, &ext2_spec_vops, &mp->mnt_vn_spec_ops);
860 vfs_add_vnodeops(mp, &ext2_fifo_vops, &mp->mnt_vn_fifo_ops);
863 ext2_sbupdate(ump, MNT_WAIT);
868 VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE);
870 bsd_free(ump->um_e2fs->s_es, M_EXT2MNT);
871 bsd_free(ump->um_e2fs, M_EXT2MNT);
872 bsd_free(ump, M_EXT2MNT);
873 mp->mnt_data = (qaddr_t)0;
879 * unmount system call
882 ext2_unmount(struct mount *mp, int mntflags)
884 struct ext2mount *ump;
885 struct ext2_sb_info *fs;
886 int error, flags, ronly, i;
889 if (mntflags & MNT_FORCE) {
890 if (mp->mnt_flag & MNT_ROOTFS)
894 if ((error = ext2_flushfiles(mp, flags)) != 0)
898 ronly = fs->s_rd_only;
901 fs->s_es->s_state |= EXT2_VALID_FS;
902 ext2_sbupdate(ump, MNT_WAIT);
905 /* release buffers containing group descriptors */
906 for(i = 0; i < fs->s_db_per_group; i++)
907 ULCK_BUF(fs->s_group_desc[i])
908 bsd_free(fs->s_group_desc, M_EXT2MNT);
910 /* release cached inode/block bitmaps */
911 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
912 if (fs->s_inode_bitmap[i])
913 ULCK_BUF(fs->s_inode_bitmap[i])
915 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
916 if (fs->s_block_bitmap[i])
917 ULCK_BUF(fs->s_block_bitmap[i])
919 ump->um_devvp->v_rdev->si_mountpoint = NULL;
920 error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD|FWRITE);
921 vrele(ump->um_devvp);
922 bsd_free(fs->s_es, M_EXT2MNT);
923 bsd_free(fs, M_EXT2MNT);
924 bsd_free(ump, M_EXT2MNT);
925 mp->mnt_data = (qaddr_t)0;
926 mp->mnt_flag &= ~MNT_LOCAL;
931 * Flush out all the files in a filesystem.
934 ext2_flushfiles(struct mount *mp, int flags)
936 struct ext2mount *ump;
944 if (mp->mnt_flag & MNT_QUOTA) {
945 if ((error = vflush(mp, 0, SKIPSYSTEM|flags)) != 0)
947 for (i = 0; i < MAXQUOTAS; i++) {
948 if (ump->um_quotas[i] == NULLVP)
950 ext2_quotaoff(mp, i);
953 * Here we fall through to vflush again to ensure
954 * that we have gotten rid of all the system vnodes.
958 error = vflush(mp, 0, flags);
963 * Get file system statistics.
964 * taken from ext2/super.c ext2_statfs
967 ext2_statfs(struct mount *mp, struct statfs *sbp, struct ucred *cred)
969 unsigned long overhead;
970 struct ext2mount *ump;
971 struct ext2_sb_info *fs;
972 struct ext2_super_block *es;
979 if (es->s_magic != EXT2_SUPER_MAGIC)
980 panic("ext2_statfs - magic number spoiled");
983 * Compute the overhead (FS structures)
985 if (es->s_feature_ro_compat & EXT2_FEATURE_RO_COMPAT_SPARSE_SUPER) {
987 for (i = 0 ; i < fs->s_groups_count; i++)
988 if (ext2_group_sparse(i))
991 nsb = fs->s_groups_count;
992 overhead = es->s_first_data_block +
993 /* Superblocks and block group descriptors: */
994 nsb * (1 + fs->s_db_per_group) +
995 /* Inode bitmap, block bitmap, and inode table: */
996 fs->s_groups_count * (1 + 1 + fs->s_itb_per_group);
998 sbp->f_bsize = EXT2_FRAG_SIZE(fs);
999 sbp->f_iosize = EXT2_BLOCK_SIZE(fs);
1000 sbp->f_blocks = es->s_blocks_count - overhead;
1001 sbp->f_bfree = es->s_free_blocks_count;
1002 sbp->f_bavail = sbp->f_bfree - es->s_r_blocks_count;
1003 sbp->f_files = es->s_inodes_count;
1004 sbp->f_ffree = es->s_free_inodes_count;
1005 if (sbp != &mp->mnt_stat) {
1006 sbp->f_type = mp->mnt_vfc->vfc_typenum;
1007 bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
1008 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
1014 * Go through the disk queues to initiate sandbagged IO;
1015 * go through the inodes to write those that have been modified;
1016 * initiate the writing of the super block if it has been modified.
1018 * Note: we are always called with the filesystem marked `MPBUSY'.
1021 static int ext2_sync_scan(struct mount *mp, struct vnode *vp, void *data);
1024 ext2_sync(struct mount *mp, int waitfor)
1026 struct ext2mount *ump = VFSTOEXT2(mp);
1027 struct ext2_sb_info *fs;
1028 struct scaninfo scaninfo;
1032 if (fs->s_dirt != 0 && fs->s_rd_only != 0) { /* XXX */
1033 kprintf("fs = %s\n", fs->fs_fsmnt);
1034 panic("ext2_sync: rofs mod");
1038 * Write back each (modified) inode.
1040 scaninfo.allerror = 0;
1041 scaninfo.rescan = 1;
1042 scaninfo.waitfor = waitfor;
1043 while (scaninfo.rescan) {
1044 scaninfo.rescan = 0;
1045 vmntvnodescan(mp, VMSC_GETVP|VMSC_NOWAIT,
1046 NULL, ext2_sync_scan, &scaninfo);
1050 * Force stale file system control information to be flushed.
1052 if ((waitfor & MNT_LAZY) == 0) {
1053 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
1054 if ((error = VOP_FSYNC(ump->um_devvp, waitfor, 0)) != 0)
1055 scaninfo.allerror = error;
1056 vn_unlock(ump->um_devvp);
1062 * Write back modified superblock.
1064 if (fs->s_dirt != 0) {
1066 fs->s_es->s_wtime = time_second;
1067 if ((error = ext2_sbupdate(ump, waitfor)) != 0)
1068 scaninfo.allerror = error;
1070 return (scaninfo.allerror);
1074 ext2_sync_scan(struct mount *mp, struct vnode *vp, void *data)
1076 struct scaninfo *info = data;
1081 if (vp->v_type == VNON ||
1083 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
1084 (RB_EMPTY(&vp->v_rbdirty_tree) || (info->waitfor & MNT_LAZY)))) {
1087 if ((error = VOP_FSYNC(vp, info->waitfor, 0)) != 0)
1088 info->allerror = error;
1093 * Look up a EXT2FS dinode number to find its incore vnode, otherwise read it
1094 * in from disk. If it is in core, wait for the lock bit to clear, then
1095 * return the inode locked. Detection and handling of mount points must be
1096 * done by the calling routine.
1099 ext2_vget(struct mount *mp, struct vnode *dvp, ino_t ino, struct vnode **vpp)
1101 struct ext2_sb_info *fs;
1103 struct ext2mount *ump;
1110 ump = VFSTOEXT2(mp);
1113 if ((*vpp = ext2_ihashget(dev, ino)) != NULL)
1117 * Lock out the creation of new entries in the FFS hash table in
1118 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate
1121 if (ext2fs_inode_hash_lock) {
1122 while (ext2fs_inode_hash_lock) {
1123 ext2fs_inode_hash_lock = -1;
1124 tsleep(&ext2fs_inode_hash_lock, 0, "e2vget", 0);
1128 ext2fs_inode_hash_lock = 1;
1131 * If this MALLOC() is performed after the getnewvnode()
1132 * it might block, leaving a vnode with a NULL v_data to be
1133 * found by ext2_sync() if a sync happens to fire right then,
1134 * which will cause a panic because ext2_sync() blindly
1135 * dereferences vp->v_data (as well it should).
1137 MALLOC(ip, struct inode *, sizeof(struct inode), M_EXT2NODE, M_WAITOK);
1139 /* Allocate a new vnode/inode. */
1140 if ((error = getnewvnode(VT_EXT2FS, mp, &vp, 0, LK_CANRECURSE)) != 0) {
1141 if (ext2fs_inode_hash_lock < 0)
1142 wakeup(&ext2fs_inode_hash_lock);
1143 ext2fs_inode_hash_lock = 0;
1145 FREE(ip, M_EXT2NODE);
1148 bzero((caddr_t)ip, sizeof(struct inode));
1151 ip->i_e2fs = fs = ump->um_e2fs;
1155 for (i = 0; i < MAXQUOTAS; i++)
1156 ip->i_dquot[i] = NODQUOT;
1159 * Put it onto its hash chain. Since our vnode is locked, other
1160 * requests for this inode will block if they arrive while we are
1161 * sleeping waiting for old data structures to be purged or for the
1162 * contents of the disk portion of this inode to be read.
1166 if (ext2fs_inode_hash_lock < 0)
1167 wakeup(&ext2fs_inode_hash_lock);
1168 ext2fs_inode_hash_lock = 0;
1170 /* Read in the disk contents for the inode, copy into the inode. */
1172 kprintf("ext2_vget(%d) dbn= %d ", ino, fsbtodb(fs, ino_to_fsba(fs, ino)));
1174 error = bread(ump->um_devvp, fsbtodoff(fs, ino_to_fsba(fs, ino)),
1175 (int)fs->s_blocksize, &bp);
1178 * The inode does not contain anything useful, so it would
1179 * be misleading to leave it on its hash chain. With mode
1180 * still zero, it will be unlinked and returned to the free
1188 /* convert ext2 inode to dinode */
1189 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data + EXT2_INODE_SIZE(fs) *
1190 ino_to_fsbo(fs, ino)), &ip->i_din);
1191 ip->i_block_group = ino_to_cg(fs, ino);
1192 ip->i_next_alloc_block = 0;
1193 ip->i_next_alloc_goal = 0;
1194 ip->i_prealloc_count = 0;
1195 ip->i_prealloc_block = 0;
1196 /* now we want to make sure that block pointers for unused
1197 blocks are zeroed out - ext2_balloc depends on this
1198 although for regular files and directories only
1200 if(S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode)) {
1201 used_blocks = (ip->i_size+fs->s_blocksize-1) / fs->s_blocksize;
1202 for(i = used_blocks; i < EXT2_NDIR_BLOCKS; i++)
1206 ext2_print_inode(ip);
1211 * Initialize the vnode from the inode, check for aliases.
1212 * Note that the underlying vnode may have changed.
1214 if ((error = ext2_vinit(mp, &vp)) != 0) {
1221 * Finish inode initialization now that aliasing has been resolved.
1223 ip->i_devvp = ump->um_devvp;
1226 * Set up a generation number for this inode if it does not
1227 * already have one. This should only happen on old filesystems.
1229 if (ip->i_gen == 0) {
1230 ip->i_gen = krandom() / 2 + 1;
1231 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1232 ip->i_flag |= IN_MODIFIED;
1235 * Return the locked and refd vnode.
1242 * File handle to vnode
1244 * Have to be really careful about stale file handles:
1245 * - check that the inode number is valid
1246 * - call ext2_vget() to get the locked inode
1247 * - check for an unallocated inode (i_mode == 0)
1248 * - check that the given client host has export rights and return
1249 * those rights via. exflagsp and credanonp
1252 ext2_fhtovp(struct mount *mp, struct vnode *rootvp,
1253 struct fid *fhp, struct vnode **vpp)
1256 struct ext2_sb_info *fs;
1261 ufhp = (struct ufid *)fhp;
1262 fs = VFSTOEXT2(mp)->um_e2fs;
1263 if (ufhp->ufid_ino < ROOTINO ||
1264 ufhp->ufid_ino > fs->s_groups_count * fs->s_es->s_inodes_per_group)
1267 error = VFS_VGET(mp, rootvp, ufhp->ufid_ino, &nvp);
1273 if (ip->i_mode == 0 ||
1274 ip->i_gen != ufhp->ufid_gen ||
1275 (VFSTOEXT2(mp)->um_i_effnlink_valid ? ip->i_effnlink :
1276 ip->i_nlink) <= 0) {
1286 * Vnode pointer to File handle
1290 ext2_vptofh(struct vnode *vp, struct fid *fhp)
1296 ufhp = (struct ufid *)fhp;
1297 ufhp->ufid_len = sizeof(struct ufid);
1298 ufhp->ufid_ino = ip->i_number;
1299 ufhp->ufid_gen = ip->i_gen;
1304 * This is the generic part of fhtovp called after the underlying
1305 * filesystem has validated the file handle.
1307 * Verify that a host should have access to a filesystem.
1310 ext2_check_export(struct mount *mp, struct sockaddr *nam, int *exflagsp,
1311 struct ucred **credanonp)
1314 struct ext2mount *ump;
1316 ump = VFSTOEXT2(mp);
1318 * Get the export permission structure for this <mp, client> tuple.
1320 np = vfs_export_lookup(mp, &ump->um_export, nam);
1324 *exflagsp = np->netc_exflags;
1325 *credanonp = &np->netc_anon;
1330 * Write a superblock and associated information back to disk.
1333 ext2_sbupdate(struct ext2mount *mp, int waitfor)
1335 struct ext2_sb_info *fs = mp->um_e2fs;
1336 struct ext2_super_block *es = fs->s_es;
1340 kprintf("\nupdating superblock, waitfor=%s\n", waitfor == MNT_WAIT ? "yes":"no");
1342 bp = getblk(mp->um_devvp, SBOFF, SBSIZE, 0, 0);
1343 bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2_super_block));
1344 if (waitfor == MNT_WAIT)
1350 * The buffers for group descriptors, inode bitmaps and block bitmaps
1351 * are not busy at this point and are (hopefully) written by the
1352 * usual sync mechanism. No need to write them here