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 $
41 * $DragonFly: src/sys/vfs/gnu/ext2fs/ext2_vfsops.c,v 1.52 2006/12/23 00:41:29 swildner Exp $
44 #include "opt_quota.h"
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #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/disklabel.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 fid *, struct vnode **);
79 static int ext2_flushfiles (struct mount *mp, int flags);
80 static int ext2_mount (struct mount *, char *, caddr_t, struct ucred *);
81 static int ext2_mountfs (struct vnode *, struct mount *, struct ucred *);
82 static int ext2_root(struct mount *, struct vnode **);
83 static int ext2_reload (struct mount *mountp, struct ucred *cred);
84 static int ext2_sbupdate (struct ext2mount *, int);
85 static int ext2_statfs (struct mount *, struct statfs *, struct ucred *);
86 static int ext2_sync (struct mount *, int);
87 static int ext2_unmount (struct mount *, int);
88 static int ext2_vget (struct mount *, ino_t, struct vnode **);
89 static int ext2_init(struct vfsconf *);
90 static int ext2_vptofh (struct vnode *, struct fid *);
92 static MALLOC_DEFINE(M_EXT2NODE, "EXT2 node", "EXT2 vnode private part");
93 MALLOC_DEFINE(M_EXT2MNT, "EXT2 mount", "EXT2 mount structure");
95 static struct vfsops ext2fs_vfsops = {
96 .vfs_mount = ext2_mount,
97 .vfs_unmount = ext2_unmount,
98 .vfs_root = ext2_root, /* root inode via vget */
99 .vfs_quotactl = ext2_quotactl, /* quota operations */
100 .vfs_statfs = ext2_statfs,
101 .vfs_sync = ext2_sync,
102 .vfs_vget = ext2_vget,
103 .vfs_fhtovp = ext2_fhtovp,
104 .vfs_checkexp = ext2_check_export,
105 .vfs_vptofh = ext2_vptofh,
106 .vfs_init = ext2_init,
107 .vfs_uninit = ext2_uninit
110 VFS_SET(ext2fs_vfsops, ext2fs, 0);
111 #define bsd_malloc kmalloc
112 #define bsd_free kfree
114 static int ext2fs_inode_hash_lock;
116 static int ext2_check_sb_compat (struct ext2_super_block *es,
117 cdev_t dev, int ronly);
118 static int compute_sb_data (struct vnode * devvp,
119 struct ext2_super_block * es,
120 struct ext2_sb_info * fs);
123 ext2_root(struct mount *mp, struct vnode **vpp)
128 error = VFS_VGET(mp, (ino_t)ROOTINO, &nvp);
136 * Do operations associated with quotas
139 ext2_quotactl(struct mount *mp, int cmds, uid_t uid, caddr_t arg,
145 int cmd, type, error;
147 type = cmds & SUBCMDMASK;
148 cmd = cmds >> SUBCMDSHIFT;
167 if (uid == cred->cr_ruid)
171 if ((error = suser_cred(cred, PRISON_ROOT)) != 0)
175 type = cmds & SUBCMDMASK;
176 if ((uint)type >= MAXQUOTAS)
178 if (vfs_busy(mp, LK_NOWAIT))
184 error = ext2_quotaon(cred, mp, type, arg);
188 error = ext2_quotaoff(mp, type);
192 error = ext2_setquota(mp, uid, type, arg);
196 error = ext2_setuse(mp, uid, type, arg);
200 error = ext2_getquota(mp, uid, type, arg);
204 error = ext2_qsync(mp);
217 * Initial UFS filesystems, done only once.
220 ext2_init(struct vfsconf *vfsp)
240 * data: this is actually a (struct ext2_args *)
243 ext2_mount(struct mount *mp, char *path, caddr_t data, struct ucred *cred)
246 struct ext2_args args;
247 struct ext2mount *ump = 0;
248 struct ext2_sb_info *fs;
252 struct nlookupdata nd;
254 if ((error = copyin(data, (caddr_t)&args, sizeof (struct ext2_args))) != 0)
258 * If updating, check whether changing from read-only to
259 * read/write; if there is no device name, that's all we do.
261 if (mp->mnt_flag & MNT_UPDATE) {
264 devvp = ump->um_devvp;
266 if (fs->s_rd_only == 0 && (mp->mnt_flag & MNT_RDONLY)) {
268 if (mp->mnt_flag & MNT_FORCE)
270 if (vfs_busy(mp, LK_NOWAIT))
272 error = ext2_flushfiles(mp, flags);
274 if (!error && fs->s_wasvalid) {
275 fs->s_es->s_state |= EXT2_VALID_FS;
276 ext2_sbupdate(ump, MNT_WAIT);
279 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
280 VOP_OPEN(devvp, FREAD, FSCRED, NULL);
281 VOP_CLOSE(devvp, FREAD|FWRITE);
284 if (!error && (mp->mnt_flag & MNT_RELOAD))
285 error = ext2_reload(mp, cred);
288 if (ext2_check_sb_compat(fs->s_es, devvp->v_rdev,
289 (mp->mnt_kern_flag & MNTK_WANTRDWR) == 0) != 0)
291 if (fs->s_rd_only && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
293 * If upgrade to read-write by non-root, then verify
294 * that user has necessary permissions on the device.
296 if (cred->cr_uid != 0) {
297 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
298 error = VOP_ACCESS(devvp, VREAD | VWRITE, cred);
306 if ((fs->s_es->s_state & EXT2_VALID_FS) == 0 ||
307 (fs->s_es->s_state & EXT2_ERROR_FS)) {
308 if (mp->mnt_flag & MNT_FORCE) {
310 "WARNING: %s was not properly dismounted\n",
314 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
319 fs->s_es->s_state &= ~EXT2_VALID_FS;
320 ext2_sbupdate(ump, MNT_WAIT);
322 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
323 VOP_OPEN(devvp, FREAD|FWRITE, FSCRED, NULL);
324 VOP_CLOSE(devvp, FREAD);
327 if (args.fspec == 0) {
329 * Process export requests.
331 return (vfs_export(mp, &ump->um_export, &args.export));
335 * Not an update, or updating the name: look up the name
336 * and verify that it refers to a sensible block device.
339 error = nlookup_init(&nd, args.fspec, UIO_USERSPACE, NLC_FOLLOW);
341 error = nlookup(&nd);
343 error = cache_vref(&nd.nl_nch, nd.nl_cred, &devvp);
348 if (!vn_isdisk(devvp, &error)) {
354 * If mount by non-root, then verify that user has necessary
355 * permissions on the device.
357 if (cred->cr_uid != 0) {
359 if ((mp->mnt_flag & MNT_RDONLY) == 0)
360 accessmode |= VWRITE;
361 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
362 if ((error = VOP_ACCESS(devvp, accessmode, cred)) != 0) {
369 if ((mp->mnt_flag & MNT_UPDATE) == 0) {
370 error = ext2_mountfs(devvp, mp, cred);
372 if (devvp != ump->um_devvp)
373 error = EINVAL; /* needs translation */
383 copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size);
384 bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size);
385 copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, &size);
386 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
387 ext2_statfs(mp, &mp->mnt_stat, cred);
392 * checks that the data in the descriptor blocks make sense
393 * this is taken from ext2/super.c
396 ext2_check_descriptors(struct ext2_sb_info *sb)
400 unsigned long block = sb->s_es->s_first_data_block;
401 struct ext2_group_desc * gdp = NULL;
403 /* ext2_debug ("Checking group descriptors"); */
405 for (i = 0; i < sb->s_groups_count; i++)
407 /* examine next descriptor block */
408 if ((i % EXT2_DESC_PER_BLOCK(sb)) == 0)
409 gdp = (struct ext2_group_desc *)
410 sb->s_group_desc[desc_block++]->b_data;
411 if (gdp->bg_block_bitmap < block ||
412 gdp->bg_block_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb))
414 kprintf ("ext2_check_descriptors: "
415 "Block bitmap for group %d"
416 " not in group (block %lu)!\n",
417 i, (unsigned long) gdp->bg_block_bitmap);
420 if (gdp->bg_inode_bitmap < block ||
421 gdp->bg_inode_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb))
423 kprintf ("ext2_check_descriptors: "
424 "Inode bitmap for group %d"
425 " not in group (block %lu)!\n",
426 i, (unsigned long) gdp->bg_inode_bitmap);
429 if (gdp->bg_inode_table < block ||
430 gdp->bg_inode_table + sb->s_itb_per_group >=
431 block + EXT2_BLOCKS_PER_GROUP(sb))
433 kprintf ("ext2_check_descriptors: "
434 "Inode table for group %d"
435 " not in group (block %lu)!\n",
436 i, (unsigned long) gdp->bg_inode_table);
439 block += EXT2_BLOCKS_PER_GROUP(sb);
446 ext2_check_sb_compat(struct ext2_super_block *es, cdev_t dev, int ronly)
448 if (es->s_magic != EXT2_SUPER_MAGIC) {
449 kprintf("ext2fs: %s: wrong magic number %#x (expected %#x)\n",
450 devtoname(dev), es->s_magic, EXT2_SUPER_MAGIC);
453 if (es->s_rev_level > EXT2_GOOD_OLD_REV) {
454 if (es->s_feature_incompat & ~EXT2_FEATURE_INCOMPAT_SUPP) {
456 "WARNING: mount of %s denied due to unsupported optional features\n",
461 (es->s_feature_ro_compat & ~EXT2_FEATURE_RO_COMPAT_SUPP)) {
463 "WARNING: R/W mount of %s denied due to unsupported optional features\n",
472 * this computes the fields of the ext2_sb_info structure from the
473 * data in the ext2_super_block structure read in
476 compute_sb_data(struct vnode *devvp, struct ext2_super_block *es,
477 struct ext2_sb_info *fs)
481 int logic_sb_block = 1; /* XXX for now */
486 #define V(v) kprintf(#v"= %d\n", fs->v);
489 fs->s_blocksize = EXT2_MIN_BLOCK_SIZE << es->s_log_block_size;
491 fs->s_bshift = EXT2_MIN_BLOCK_LOG_SIZE + es->s_log_block_size;
493 fs->s_fsbtodb = es->s_log_block_size + 1;
495 fs->s_qbmask = fs->s_blocksize - 1;
497 fs->s_blocksize_bits = EXT2_BLOCK_SIZE_BITS(es);
499 fs->s_frag_size = EXT2_MIN_FRAG_SIZE << es->s_log_frag_size;
502 fs->s_frags_per_block = fs->s_blocksize / fs->s_frag_size;
504 fs->s_blocks_per_group = es->s_blocks_per_group;
505 V(s_blocks_per_group)
506 fs->s_frags_per_group = es->s_frags_per_group;
508 fs->s_inodes_per_group = es->s_inodes_per_group;
509 V(s_inodes_per_group)
510 fs->s_inodes_per_block = fs->s_blocksize / EXT2_INODE_SIZE;
511 V(s_inodes_per_block)
512 fs->s_itb_per_group = fs->s_inodes_per_group /fs->s_inodes_per_block;
514 fs->s_desc_per_block = fs->s_blocksize / sizeof (struct ext2_group_desc);
516 /* s_resuid / s_resgid ? */
517 fs->s_groups_count = (es->s_blocks_count -
518 es->s_first_data_block +
519 EXT2_BLOCKS_PER_GROUP(fs) - 1) /
520 EXT2_BLOCKS_PER_GROUP(fs);
522 db_count = (fs->s_groups_count + EXT2_DESC_PER_BLOCK(fs) - 1) /
523 EXT2_DESC_PER_BLOCK(fs);
524 fs->s_db_per_group = db_count;
527 fs->s_group_desc = bsd_malloc(db_count * sizeof (struct buf *),
528 M_EXT2MNT, M_WAITOK);
530 /* adjust logic_sb_block */
531 if(fs->s_blocksize > SBSIZE)
532 /* Godmar thinks: if the blocksize is greater than 1024, then
533 the superblock is logically part of block zero.
537 for (i = 0; i < db_count; i++) {
538 error = bread(devvp, fsbtodoff(fs, logic_sb_block + i + 1),
539 fs->s_blocksize, &fs->s_group_desc[i]);
541 for (j = 0; j < i; j++)
542 brelse(fs->s_group_desc[j]);
543 bsd_free(fs->s_group_desc, M_EXT2MNT);
544 kprintf("EXT2-fs: unable to read group descriptors (%d)\n", error);
547 /* Set the B_LOCKED flag on the buffer, then brelse() it */
548 LCK_BUF(fs->s_group_desc[i])
550 if(!ext2_check_descriptors(fs)) {
551 for (j = 0; j < db_count; j++)
552 ULCK_BUF(fs->s_group_desc[j])
553 bsd_free(fs->s_group_desc, M_EXT2MNT);
554 kprintf("EXT2-fs: (ext2_check_descriptors failure) "
555 "unable to read group descriptors\n");
559 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) {
560 fs->s_inode_bitmap_number[i] = 0;
561 fs->s_inode_bitmap[i] = NULL;
562 fs->s_block_bitmap_number[i] = 0;
563 fs->s_block_bitmap[i] = NULL;
565 fs->s_loaded_inode_bitmaps = 0;
566 fs->s_loaded_block_bitmaps = 0;
571 * Reload all incore data for a filesystem (used after running fsck on
572 * the root filesystem and finding things to fix). The filesystem must
573 * be mounted read-only.
575 * Things to do to update the mount:
576 * 1) invalidate all cached meta-data.
577 * 2) re-read superblock from disk.
578 * 3) re-read summary information from disk.
579 * 4) invalidate all inactive vnodes.
580 * 5) invalidate all cached file data.
581 * 6) re-read inode data for all active vnodes.
583 static int ext2_reload_scan1(struct mount *mp, struct vnode *vp, void *rescan);
584 static int ext2_reload_scan2(struct mount *mp, struct vnode *vp, void *rescan);
591 struct ext2_sb_info *fs;
595 ext2_reload(struct mount *mountp, struct ucred *cred)
599 struct ext2_super_block * es;
600 struct ext2_sb_info *fs;
602 struct scaninfo scaninfo;
604 if ((mountp->mnt_flag & MNT_RDONLY) == 0)
607 * Step 1: invalidate all cached meta-data.
609 devvp = VFSTOEXT2(mountp)->um_devvp;
610 if (vinvalbuf(devvp, 0, 0, 0))
611 panic("ext2_reload: dirty1");
613 * Step 2: re-read superblock from disk.
614 * constants have been adjusted for ext2
616 if ((error = bread(devvp, SBOFF, SBSIZE, &bp)) != 0)
618 es = (struct ext2_super_block *)bp->b_data;
619 if (ext2_check_sb_compat(es, devvp->v_rdev, 0) != 0) {
621 return (EIO); /* XXX needs translation */
623 fs = VFSTOEXT2(mountp)->um_e2fs;
624 bcopy(bp->b_data, fs->s_es, sizeof(struct ext2_super_block));
626 if((error = compute_sb_data(devvp, es, fs)) != 0) {
631 if (fs->fs_sbsize < SBSIZE)
632 bp->b_flags |= B_INVAL;
637 scaninfo.devvp = devvp;
639 while (error == 0 && scaninfo.rescan) {
641 error = vmntvnodescan(mountp, VMSC_GETVX, ext2_reload_scan1,
642 ext2_reload_scan2, &scaninfo);
648 ext2_reload_scan1(struct mount *mp, struct vnode *vp, void *data)
650 /*struct scaninfo *info = data;*/
656 ext2_reload_scan2(struct mount *mp, struct vnode *vp, void *data)
658 struct scaninfo *info = data;
670 * Step 5: invalidate all cached file data.
672 if (vinvalbuf(vp, 0, 0, 0))
673 panic("ext2_reload: dirty2");
675 * Step 6: re-read inode data for all active vnodes.
678 error = bread(info->devvp,
679 fsbtodoff(info->fs, ino_to_fsba(info->fs, ip->i_number)),
680 (int)info->fs->s_blocksize, &bp);
683 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data +
684 EXT2_INODE_SIZE * ino_to_fsbo(info->fs, ip->i_number)),
691 * Common code for mount and mountroot
694 ext2_mountfs(struct vnode *devvp, struct mount *mp, struct ucred *cred)
696 struct ext2mount *ump;
698 struct ext2_sb_info *fs;
699 struct ext2_super_block * es;
701 struct partinfo dpart;
707 * Disallow multiple mounts of the same device.
708 * Disallow mounting of a device that is currently in use
709 * (except for root, which might share swap device for miniroot).
710 * Flush out any old buffers remaining from a previous use.
712 if ((error = vfs_mountedon(devvp)) != 0)
714 if (count_udev(devvp->v_udev) > 0)
716 if ((error = vinvalbuf(devvp, V_SAVE, 0, 0)) != 0)
719 /* turn on this to force it to be read-only */
720 mp->mnt_flag |= MNT_RDONLY;
723 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
724 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
725 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, NULL);
730 if (dev->si_iosize_max != 0)
731 mp->mnt_iosize_max = dev->si_iosize_max;
732 if (mp->mnt_iosize_max > MAXPHYS)
733 mp->mnt_iosize_max = MAXPHYS;
734 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, cred) != 0)
738 size = dpart.disklab->d_secsize;
743 if ((error = bread(devvp, SBOFF, SBSIZE, &bp)) != 0)
745 es = (struct ext2_super_block *)bp->b_data;
746 if (ext2_check_sb_compat(es, dev, ronly) != 0) {
747 error = EINVAL; /* XXX needs translation */
750 if ((es->s_state & EXT2_VALID_FS) == 0 ||
751 (es->s_state & EXT2_ERROR_FS)) {
752 if (ronly || (mp->mnt_flag & MNT_FORCE)) {
754 "WARNING: Filesystem was not properly dismounted\n");
757 "WARNING: R/W mount denied. Filesystem is not clean - run fsck\n");
762 ump = bsd_malloc(sizeof *ump, M_EXT2MNT, M_WAITOK);
763 bzero((caddr_t)ump, sizeof *ump);
764 ump->um_malloctype = M_EXT2NODE;
765 ump->um_blkatoff = ext2_blkatoff;
766 ump->um_truncate = ext2_truncate;
767 ump->um_update = ext2_update;
768 ump->um_valloc = ext2_valloc;
769 ump->um_vfree = ext2_vfree;
770 /* I don't know whether this is the right strategy. Note that
771 we dynamically allocate both a ext2_sb_info and a ext2_super_block
772 while Linux keeps the super block in a locked buffer
774 ump->um_e2fs = bsd_malloc(sizeof(struct ext2_sb_info),
775 M_EXT2MNT, M_WAITOK);
776 ump->um_e2fs->s_es = bsd_malloc(sizeof(struct ext2_super_block),
777 M_EXT2MNT, M_WAITOK);
778 bcopy(es, ump->um_e2fs->s_es, (u_int)sizeof(struct ext2_super_block));
779 if ((error = compute_sb_data(devvp, ump->um_e2fs->s_es, ump->um_e2fs)))
782 * We don't free the group descriptors allocated by compute_sb_data()
783 * until ext2_unmount(). This is OK since the mount will succeed.
788 fs->s_rd_only = ronly; /* ronly is set according to mnt_flags */
789 /* if the fs is not mounted read-only, make sure the super block is
790 always written back on a sync()
792 fs->s_wasvalid = fs->s_es->s_state & EXT2_VALID_FS ? 1 : 0;
794 fs->s_dirt = 1; /* mark it modified */
795 fs->s_es->s_state &= ~EXT2_VALID_FS; /* set fs invalid */
797 mp->mnt_data = (qaddr_t)ump;
798 mp->mnt_stat.f_fsid.val[0] = dev2udev(dev);
799 mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
800 mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN;
801 mp->mnt_flag |= MNT_LOCAL;
804 ump->um_devvp = devvp;
805 /* setting those two parameters allows us to use
806 ext2_bmap w/o changse !
808 ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs);
809 ump->um_bptrtodb = fs->s_es->s_log_block_size + 1;
810 ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs);
811 for (i = 0; i < MAXQUOTAS; i++)
812 ump->um_quotas[i] = NULLVP;
813 dev->si_mountpoint = mp;
815 vfs_add_vnodeops(mp, &ext2_vnode_vops, &mp->mnt_vn_norm_ops);
816 vfs_add_vnodeops(mp, &ext2_spec_vops, &mp->mnt_vn_spec_ops);
817 vfs_add_vnodeops(mp, &ext2_fifo_vops, &mp->mnt_vn_fifo_ops);
820 ext2_sbupdate(ump, MNT_WAIT);
825 VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE);
827 bsd_free(ump->um_e2fs->s_es, M_EXT2MNT);
828 bsd_free(ump->um_e2fs, M_EXT2MNT);
829 bsd_free(ump, M_EXT2MNT);
830 mp->mnt_data = (qaddr_t)0;
836 * unmount system call
839 ext2_unmount(struct mount *mp, int mntflags)
841 struct ext2mount *ump;
842 struct ext2_sb_info *fs;
843 int error, flags, ronly, i;
846 if (mntflags & MNT_FORCE) {
847 if (mp->mnt_flag & MNT_ROOTFS)
851 if ((error = ext2_flushfiles(mp, flags)) != 0)
855 ronly = fs->s_rd_only;
858 fs->s_es->s_state |= EXT2_VALID_FS;
859 ext2_sbupdate(ump, MNT_WAIT);
862 /* release buffers containing group descriptors */
863 for(i = 0; i < fs->s_db_per_group; i++)
864 ULCK_BUF(fs->s_group_desc[i])
865 bsd_free(fs->s_group_desc, M_EXT2MNT);
867 /* release cached inode/block bitmaps */
868 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
869 if (fs->s_inode_bitmap[i])
870 ULCK_BUF(fs->s_inode_bitmap[i])
872 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
873 if (fs->s_block_bitmap[i])
874 ULCK_BUF(fs->s_block_bitmap[i])
876 ump->um_devvp->v_rdev->si_mountpoint = NULL;
877 error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD|FWRITE);
878 vrele(ump->um_devvp);
879 bsd_free(fs->s_es, M_EXT2MNT);
880 bsd_free(fs, M_EXT2MNT);
881 bsd_free(ump, M_EXT2MNT);
882 mp->mnt_data = (qaddr_t)0;
883 mp->mnt_flag &= ~MNT_LOCAL;
888 * Flush out all the files in a filesystem.
891 ext2_flushfiles(struct mount *mp, int flags)
893 struct ext2mount *ump;
901 if (mp->mnt_flag & MNT_QUOTA) {
902 if ((error = vflush(mp, 0, SKIPSYSTEM|flags)) != 0)
904 for (i = 0; i < MAXQUOTAS; i++) {
905 if (ump->um_quotas[i] == NULLVP)
907 ext2_quotaoff(mp, i);
910 * Here we fall through to vflush again to ensure
911 * that we have gotten rid of all the system vnodes.
915 error = vflush(mp, 0, flags);
920 * Get file system statistics.
921 * taken from ext2/super.c ext2_statfs
924 ext2_statfs(struct mount *mp, struct statfs *sbp, struct ucred *cred)
926 unsigned long overhead;
927 struct ext2mount *ump;
928 struct ext2_sb_info *fs;
929 struct ext2_super_block *es;
936 if (es->s_magic != EXT2_SUPER_MAGIC)
937 panic("ext2_statfs - magic number spoiled");
940 * Compute the overhead (FS structures)
942 if (es->s_feature_ro_compat & EXT2_FEATURE_RO_COMPAT_SPARSE_SUPER) {
944 for (i = 0 ; i < fs->s_groups_count; i++)
945 if (ext2_group_sparse(i))
948 nsb = fs->s_groups_count;
949 overhead = es->s_first_data_block +
950 /* Superblocks and block group descriptors: */
951 nsb * (1 + fs->s_db_per_group) +
952 /* Inode bitmap, block bitmap, and inode table: */
953 fs->s_groups_count * (1 + 1 + fs->s_itb_per_group);
955 sbp->f_bsize = EXT2_FRAG_SIZE(fs);
956 sbp->f_iosize = EXT2_BLOCK_SIZE(fs);
957 sbp->f_blocks = es->s_blocks_count - overhead;
958 sbp->f_bfree = es->s_free_blocks_count;
959 sbp->f_bavail = sbp->f_bfree - es->s_r_blocks_count;
960 sbp->f_files = es->s_inodes_count;
961 sbp->f_ffree = es->s_free_inodes_count;
962 if (sbp != &mp->mnt_stat) {
963 sbp->f_type = mp->mnt_vfc->vfc_typenum;
964 bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
965 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
971 * Go through the disk queues to initiate sandbagged IO;
972 * go through the inodes to write those that have been modified;
973 * initiate the writing of the super block if it has been modified.
975 * Note: we are always called with the filesystem marked `MPBUSY'.
978 static int ext2_sync_scan(struct mount *mp, struct vnode *vp, void *data);
981 ext2_sync(struct mount *mp, int waitfor)
983 struct ext2mount *ump = VFSTOEXT2(mp);
984 struct ext2_sb_info *fs;
985 struct scaninfo scaninfo;
989 if (fs->s_dirt != 0 && fs->s_rd_only != 0) { /* XXX */
990 kprintf("fs = %s\n", fs->fs_fsmnt);
991 panic("ext2_sync: rofs mod");
995 * Write back each (modified) inode.
997 scaninfo.allerror = 0;
999 scaninfo.waitfor = waitfor;
1000 while (scaninfo.rescan) {
1001 scaninfo.rescan = 0;
1002 vmntvnodescan(mp, VMSC_GETVP|VMSC_NOWAIT,
1003 NULL, ext2_sync_scan, &scaninfo);
1007 * Force stale file system control information to be flushed.
1009 if (waitfor != MNT_LAZY) {
1010 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
1011 if ((error = VOP_FSYNC(ump->um_devvp, waitfor)) != 0)
1012 scaninfo.allerror = error;
1013 vn_unlock(ump->um_devvp);
1019 * Write back modified superblock.
1021 if (fs->s_dirt != 0) {
1023 fs->s_es->s_wtime = time_second;
1024 if ((error = ext2_sbupdate(ump, waitfor)) != 0)
1025 scaninfo.allerror = error;
1027 return (scaninfo.allerror);
1031 ext2_sync_scan(struct mount *mp, struct vnode *vp, void *data)
1033 struct scaninfo *info = data;
1038 if (vp->v_type == VNON ||
1040 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
1041 (RB_EMPTY(&vp->v_rbdirty_tree) || info->waitfor == MNT_LAZY))) {
1044 if ((error = VOP_FSYNC(vp, info->waitfor)) != 0)
1045 info->allerror = error;
1050 * Look up a EXT2FS dinode number to find its incore vnode, otherwise read it
1051 * in from disk. If it is in core, wait for the lock bit to clear, then
1052 * return the inode locked. Detection and handling of mount points must be
1053 * done by the calling routine.
1056 ext2_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
1058 struct ext2_sb_info *fs;
1060 struct ext2mount *ump;
1067 ump = VFSTOEXT2(mp);
1070 if ((*vpp = ext2_ihashget(dev, ino)) != NULL)
1074 * Lock out the creation of new entries in the FFS hash table in
1075 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate
1078 if (ext2fs_inode_hash_lock) {
1079 while (ext2fs_inode_hash_lock) {
1080 ext2fs_inode_hash_lock = -1;
1081 tsleep(&ext2fs_inode_hash_lock, 0, "e2vget", 0);
1085 ext2fs_inode_hash_lock = 1;
1088 * If this MALLOC() is performed after the getnewvnode()
1089 * it might block, leaving a vnode with a NULL v_data to be
1090 * found by ext2_sync() if a sync happens to fire right then,
1091 * which will cause a panic because ext2_sync() blindly
1092 * dereferences vp->v_data (as well it should).
1094 MALLOC(ip, struct inode *, sizeof(struct inode), M_EXT2NODE, M_WAITOK);
1096 /* Allocate a new vnode/inode. */
1097 if ((error = getnewvnode(VT_EXT2FS, mp, &vp, 0, LK_CANRECURSE)) != 0) {
1098 if (ext2fs_inode_hash_lock < 0)
1099 wakeup(&ext2fs_inode_hash_lock);
1100 ext2fs_inode_hash_lock = 0;
1102 FREE(ip, M_EXT2NODE);
1105 bzero((caddr_t)ip, sizeof(struct inode));
1108 ip->i_e2fs = fs = ump->um_e2fs;
1112 for (i = 0; i < MAXQUOTAS; i++)
1113 ip->i_dquot[i] = NODQUOT;
1116 * Put it onto its hash chain. Since our vnode is locked, other
1117 * requests for this inode will block if they arrive while we are
1118 * sleeping waiting for old data structures to be purged or for the
1119 * contents of the disk portion of this inode to be read.
1123 if (ext2fs_inode_hash_lock < 0)
1124 wakeup(&ext2fs_inode_hash_lock);
1125 ext2fs_inode_hash_lock = 0;
1127 /* Read in the disk contents for the inode, copy into the inode. */
1129 kprintf("ext2_vget(%d) dbn= %d ", ino, fsbtodb(fs, ino_to_fsba(fs, ino)));
1131 error = bread(ump->um_devvp, fsbtodoff(fs, ino_to_fsba(fs, ino)),
1132 (int)fs->s_blocksize, &bp);
1135 * The inode does not contain anything useful, so it would
1136 * be misleading to leave it on its hash chain. With mode
1137 * still zero, it will be unlinked and returned to the free
1145 /* convert ext2 inode to dinode */
1146 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data + EXT2_INODE_SIZE *
1147 ino_to_fsbo(fs, ino)), &ip->i_din);
1148 ip->i_block_group = ino_to_cg(fs, ino);
1149 ip->i_next_alloc_block = 0;
1150 ip->i_next_alloc_goal = 0;
1151 ip->i_prealloc_count = 0;
1152 ip->i_prealloc_block = 0;
1153 /* now we want to make sure that block pointers for unused
1154 blocks are zeroed out - ext2_balloc depends on this
1155 although for regular files and directories only
1157 if(S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode)) {
1158 used_blocks = (ip->i_size+fs->s_blocksize-1) / fs->s_blocksize;
1159 for(i = used_blocks; i < EXT2_NDIR_BLOCKS; i++)
1163 ext2_print_inode(ip);
1168 * Initialize the vnode from the inode, check for aliases.
1169 * Note that the underlying vnode may have changed.
1171 if ((error = ext2_vinit(mp, &vp)) != 0) {
1178 * Finish inode initialization now that aliasing has been resolved.
1180 ip->i_devvp = ump->um_devvp;
1183 * Set up a generation number for this inode if it does not
1184 * already have one. This should only happen on old filesystems.
1186 if (ip->i_gen == 0) {
1187 ip->i_gen = krandom() / 2 + 1;
1188 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1189 ip->i_flag |= IN_MODIFIED;
1192 * Return the locked and refd vnode.
1199 * File handle to vnode
1201 * Have to be really careful about stale file handles:
1202 * - check that the inode number is valid
1203 * - call ext2_vget() to get the locked inode
1204 * - check for an unallocated inode (i_mode == 0)
1205 * - check that the given client host has export rights and return
1206 * those rights via. exflagsp and credanonp
1209 ext2_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
1212 struct ext2_sb_info *fs;
1217 ufhp = (struct ufid *)fhp;
1218 fs = VFSTOEXT2(mp)->um_e2fs;
1219 if (ufhp->ufid_ino < ROOTINO ||
1220 ufhp->ufid_ino > fs->s_groups_count * fs->s_es->s_inodes_per_group)
1223 error = VFS_VGET(mp, ufhp->ufid_ino, &nvp);
1229 if (ip->i_mode == 0 ||
1230 ip->i_gen != ufhp->ufid_gen ||
1231 (VFSTOEXT2(mp)->um_i_effnlink_valid ? ip->i_effnlink :
1232 ip->i_nlink) <= 0) {
1242 * Vnode pointer to File handle
1246 ext2_vptofh(struct vnode *vp, struct fid *fhp)
1252 ufhp = (struct ufid *)fhp;
1253 ufhp->ufid_len = sizeof(struct ufid);
1254 ufhp->ufid_ino = ip->i_number;
1255 ufhp->ufid_gen = ip->i_gen;
1260 * This is the generic part of fhtovp called after the underlying
1261 * filesystem has validated the file handle.
1263 * Verify that a host should have access to a filesystem.
1266 ext2_check_export(struct mount *mp, struct sockaddr *nam, int *exflagsp,
1267 struct ucred **credanonp)
1270 struct ext2mount *ump;
1272 ump = VFSTOEXT2(mp);
1274 * Get the export permission structure for this <mp, client> tuple.
1276 np = vfs_export_lookup(mp, &ump->um_export, nam);
1280 *exflagsp = np->netc_exflags;
1281 *credanonp = &np->netc_anon;
1286 * Write a superblock and associated information back to disk.
1289 ext2_sbupdate(struct ext2mount *mp, int waitfor)
1291 struct ext2_sb_info *fs = mp->um_e2fs;
1292 struct ext2_super_block *es = fs->s_es;
1296 kprintf("\nupdating superblock, waitfor=%s\n", waitfor == MNT_WAIT ? "yes":"no");
1298 bp = getblk(mp->um_devvp, SBOFF, SBSIZE, 0, 0);
1299 bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2_super_block));
1300 if (waitfor == MNT_WAIT)
1306 * The buffers for group descriptors, inode bitmaps and block bitmaps
1307 * are not busy at this point and are (hopefully) written by the
1308 * usual sync mechanism. No need to write them here