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 1.57 2008/09/17 21:44:21 dillon Exp $
44 #include "opt_quota.h"
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/nlookup.h>
51 #include <sys/kernel.h>
52 #include <sys/vnode.h>
53 #include <sys/mount.h>
56 #include <sys/fcntl.h>
57 #include <sys/diskslice.h>
58 #include <sys/malloc.h>
60 #include <vm/vm_zone.h>
63 #include <sys/thread2.h>
68 #include "ext2mount.h"
69 #include "ext2_extern.h"
73 #include "ext2_fs_sb.h"
75 extern struct vop_ops ext2_vnode_vops;
76 extern struct vop_ops ext2_spec_vops;
77 extern struct vop_ops ext2_fifo_vops;
79 static int ext2_fhtovp (struct mount *, struct vnode *,
80 struct fid *, struct vnode **);
81 static int ext2_flushfiles (struct mount *mp, int flags);
82 static int ext2_mount (struct mount *, char *, caddr_t, struct ucred *);
83 static int ext2_mountfs (struct vnode *, struct mount *, struct ucred *);
84 static int ext2_root(struct mount *, struct vnode **);
85 static int ext2_reload (struct mount *mountp, struct ucred *cred);
86 static int ext2_sbupdate (struct ext2mount *, int);
87 static int ext2_statfs (struct mount *, struct statfs *, struct ucred *);
88 static int ext2_sync (struct mount *, int);
89 static int ext2_unmount (struct mount *, int);
90 static int ext2_vget (struct mount *, ino_t, struct vnode **);
91 static int ext2_init(struct vfsconf *);
92 static int ext2_vptofh (struct vnode *, struct fid *);
94 static MALLOC_DEFINE(M_EXT2NODE, "EXT2 node", "EXT2 vnode private part");
95 MALLOC_DEFINE(M_EXT2MNT, "EXT2 mount", "EXT2 mount structure");
97 static struct vfsops ext2fs_vfsops = {
98 .vfs_mount = ext2_mount,
99 .vfs_unmount = ext2_unmount,
100 .vfs_root = ext2_root, /* root inode via vget */
101 .vfs_quotactl = ext2_quotactl, /* quota operations */
102 .vfs_statfs = ext2_statfs,
103 .vfs_sync = ext2_sync,
104 .vfs_vget = ext2_vget,
105 .vfs_fhtovp = ext2_fhtovp,
106 .vfs_checkexp = ext2_check_export,
107 .vfs_vptofh = ext2_vptofh,
108 .vfs_init = ext2_init,
109 .vfs_uninit = ext2_uninit
112 VFS_SET(ext2fs_vfsops, ext2fs, 0);
113 #define bsd_malloc kmalloc
114 #define bsd_free kfree
116 static int ext2fs_inode_hash_lock;
118 static int ext2_check_sb_compat (struct ext2_super_block *es,
119 cdev_t dev, int ronly);
120 static int compute_sb_data (struct vnode * devvp,
121 struct ext2_super_block * es,
122 struct ext2_sb_info * fs);
125 ext2_root(struct mount *mp, struct vnode **vpp)
130 error = VFS_VGET(mp, (ino_t)ROOTINO, &nvp);
138 * Do operations associated with quotas
141 ext2_quotactl(struct mount *mp, int cmds, uid_t uid, caddr_t arg,
147 int cmd, type, error;
149 type = cmds & SUBCMDMASK;
150 cmd = cmds >> SUBCMDSHIFT;
169 if (uid == cred->cr_ruid)
173 if ((error = priv_check_cred(cred, PRIV_ROOT, PRISON_ROOT)) != 0)
177 type = cmds & SUBCMDMASK;
178 if ((uint)type >= MAXQUOTAS)
180 if (vfs_busy(mp, LK_NOWAIT))
186 error = ext2_quotaon(cred, mp, type, arg);
190 error = ext2_quotaoff(mp, type);
194 error = ext2_setquota(mp, uid, type, arg);
198 error = ext2_setuse(mp, uid, type, arg);
202 error = ext2_getquota(mp, uid, type, arg);
206 error = ext2_qsync(mp);
219 * Initial UFS filesystems, done only once.
222 ext2_init(struct vfsconf *vfsp)
242 * data: this is actually a (struct ext2_args *)
245 ext2_mount(struct mount *mp, char *path, caddr_t data, struct ucred *cred)
248 struct ext2_args args;
249 struct ext2mount *ump = 0;
250 struct ext2_sb_info *fs;
254 struct nlookupdata nd;
256 if ((error = copyin(data, (caddr_t)&args, sizeof (struct ext2_args))) != 0)
260 * If updating, check whether changing from read-only to
261 * read/write; if there is no device name, that's all we do.
263 if (mp->mnt_flag & MNT_UPDATE) {
266 devvp = ump->um_devvp;
268 if (fs->s_rd_only == 0 && (mp->mnt_flag & MNT_RDONLY)) {
270 if (mp->mnt_flag & MNT_FORCE)
272 if (vfs_busy(mp, LK_NOWAIT))
274 error = ext2_flushfiles(mp, flags);
276 if (!error && fs->s_wasvalid) {
277 fs->s_es->s_state |= EXT2_VALID_FS;
278 ext2_sbupdate(ump, MNT_WAIT);
281 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
282 VOP_OPEN(devvp, FREAD, FSCRED, NULL);
283 VOP_CLOSE(devvp, FREAD|FWRITE);
286 if (!error && (mp->mnt_flag & MNT_RELOAD))
287 error = ext2_reload(mp, cred);
290 if (ext2_check_sb_compat(fs->s_es, devvp->v_rdev,
291 (mp->mnt_kern_flag & MNTK_WANTRDWR) == 0) != 0)
293 if (fs->s_rd_only && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
295 * If upgrade to read-write by non-root, then verify
296 * that user has necessary permissions on the device.
298 if (cred->cr_uid != 0) {
299 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
300 error = VOP_ACCESS(devvp, VREAD | VWRITE, cred);
308 if ((fs->s_es->s_state & EXT2_VALID_FS) == 0 ||
309 (fs->s_es->s_state & EXT2_ERROR_FS)) {
310 if (mp->mnt_flag & MNT_FORCE) {
312 "WARNING: %s was not properly dismounted\n",
316 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
321 fs->s_es->s_state &= ~EXT2_VALID_FS;
322 ext2_sbupdate(ump, MNT_WAIT);
324 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
325 VOP_OPEN(devvp, FREAD|FWRITE, FSCRED, NULL);
326 VOP_CLOSE(devvp, FREAD);
329 if (args.fspec == 0) {
331 * Process export requests.
333 return (vfs_export(mp, &ump->um_export, &args.export));
337 * Not an update, or updating the name: look up the name
338 * and verify that it refers to a sensible block device.
341 error = nlookup_init(&nd, args.fspec, UIO_USERSPACE, NLC_FOLLOW);
343 error = nlookup(&nd);
345 error = cache_vref(&nd.nl_nch, nd.nl_cred, &devvp);
350 if (!vn_isdisk(devvp, &error)) {
356 * If mount by non-root, then verify that user has necessary
357 * permissions on the device.
359 if (cred->cr_uid != 0) {
361 if ((mp->mnt_flag & MNT_RDONLY) == 0)
362 accessmode |= VWRITE;
363 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
364 if ((error = VOP_ACCESS(devvp, accessmode, cred)) != 0) {
371 if ((mp->mnt_flag & MNT_UPDATE) == 0) {
372 error = ext2_mountfs(devvp, mp, cred);
374 if (devvp != ump->um_devvp)
375 error = EINVAL; /* needs translation */
385 copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size);
386 bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size);
387 copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, &size);
388 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
389 ext2_statfs(mp, &mp->mnt_stat, cred);
394 * checks that the data in the descriptor blocks make sense
395 * this is taken from ext2/super.c
398 ext2_check_descriptors(struct ext2_sb_info *sb)
402 unsigned long block = sb->s_es->s_first_data_block;
403 struct ext2_group_desc * gdp = NULL;
405 /* ext2_debug ("Checking group descriptors"); */
407 for (i = 0; i < sb->s_groups_count; i++)
409 /* examine next descriptor block */
410 if ((i % EXT2_DESC_PER_BLOCK(sb)) == 0)
411 gdp = (struct ext2_group_desc *)
412 sb->s_group_desc[desc_block++]->b_data;
413 if (gdp->bg_block_bitmap < block ||
414 gdp->bg_block_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb))
416 kprintf ("ext2_check_descriptors: "
417 "Block bitmap for group %d"
418 " not in group (block %lu)!\n",
419 i, (unsigned long) gdp->bg_block_bitmap);
422 if (gdp->bg_inode_bitmap < block ||
423 gdp->bg_inode_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb))
425 kprintf ("ext2_check_descriptors: "
426 "Inode bitmap for group %d"
427 " not in group (block %lu)!\n",
428 i, (unsigned long) gdp->bg_inode_bitmap);
431 if (gdp->bg_inode_table < block ||
432 gdp->bg_inode_table + sb->s_itb_per_group >=
433 block + EXT2_BLOCKS_PER_GROUP(sb))
435 kprintf ("ext2_check_descriptors: "
436 "Inode table for group %d"
437 " not in group (block %lu)!\n",
438 i, (unsigned long) gdp->bg_inode_table);
441 block += EXT2_BLOCKS_PER_GROUP(sb);
448 ext2_check_sb_compat(struct ext2_super_block *es, cdev_t dev, int ronly)
450 if (es->s_magic != EXT2_SUPER_MAGIC) {
451 kprintf("ext2fs: %s: wrong magic number %#x (expected %#x)\n",
452 devtoname(dev), es->s_magic, EXT2_SUPER_MAGIC);
455 if (es->s_rev_level > EXT2_GOOD_OLD_REV) {
456 if (es->s_feature_incompat & ~EXT2_FEATURE_INCOMPAT_SUPP) {
458 "WARNING: mount of %s denied due to unsupported optional features\n",
463 (es->s_feature_ro_compat & ~EXT2_FEATURE_RO_COMPAT_SUPP)) {
465 "WARNING: R/W mount of %s denied due to unsupported optional features\n",
474 * this computes the fields of the ext2_sb_info structure from the
475 * data in the ext2_super_block structure read in
478 compute_sb_data(struct vnode *devvp, struct ext2_super_block *es,
479 struct ext2_sb_info *fs)
483 int logic_sb_block = 1; /* XXX for now */
488 #define V(v) kprintf(#v"= %d\n", fs->v);
491 fs->s_blocksize = EXT2_MIN_BLOCK_SIZE << es->s_log_block_size;
493 fs->s_bshift = EXT2_MIN_BLOCK_LOG_SIZE + es->s_log_block_size;
495 fs->s_fsbtodb = es->s_log_block_size + 1;
497 fs->s_qbmask = fs->s_blocksize - 1;
499 fs->s_blocksize_bits = EXT2_BLOCK_SIZE_BITS(es);
501 fs->s_frag_size = EXT2_MIN_FRAG_SIZE << es->s_log_frag_size;
504 fs->s_frags_per_block = fs->s_blocksize / fs->s_frag_size;
506 fs->s_blocks_per_group = es->s_blocks_per_group;
507 V(s_blocks_per_group)
508 fs->s_frags_per_group = es->s_frags_per_group;
510 fs->s_inodes_per_group = es->s_inodes_per_group;
511 V(s_inodes_per_group)
512 fs->s_inodes_per_block = fs->s_blocksize / EXT2_INODE_SIZE;
513 V(s_inodes_per_block)
514 fs->s_itb_per_group = fs->s_inodes_per_group /fs->s_inodes_per_block;
516 fs->s_desc_per_block = fs->s_blocksize / sizeof (struct ext2_group_desc);
518 /* s_resuid / s_resgid ? */
519 fs->s_groups_count = (es->s_blocks_count -
520 es->s_first_data_block +
521 EXT2_BLOCKS_PER_GROUP(fs) - 1) /
522 EXT2_BLOCKS_PER_GROUP(fs);
524 db_count = (fs->s_groups_count + EXT2_DESC_PER_BLOCK(fs) - 1) /
525 EXT2_DESC_PER_BLOCK(fs);
526 fs->s_db_per_group = db_count;
529 fs->s_group_desc = bsd_malloc(db_count * sizeof (struct buf *),
530 M_EXT2MNT, M_WAITOK);
532 /* adjust logic_sb_block */
533 if(fs->s_blocksize > SBSIZE)
534 /* Godmar thinks: if the blocksize is greater than 1024, then
535 the superblock is logically part of block zero.
539 for (i = 0; i < db_count; i++) {
540 error = bread(devvp, fsbtodoff(fs, logic_sb_block + i + 1),
541 fs->s_blocksize, &fs->s_group_desc[i]);
543 for (j = 0; j < i; j++)
544 brelse(fs->s_group_desc[j]);
545 bsd_free(fs->s_group_desc, M_EXT2MNT);
546 kprintf("EXT2-fs: unable to read group descriptors (%d)\n", error);
549 /* Set the B_LOCKED flag on the buffer, then brelse() it */
550 LCK_BUF(fs->s_group_desc[i])
552 if(!ext2_check_descriptors(fs)) {
553 for (j = 0; j < db_count; j++)
554 ULCK_BUF(fs->s_group_desc[j])
555 bsd_free(fs->s_group_desc, M_EXT2MNT);
556 kprintf("EXT2-fs: (ext2_check_descriptors failure) "
557 "unable to read group descriptors\n");
561 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) {
562 fs->s_inode_bitmap_number[i] = 0;
563 fs->s_inode_bitmap[i] = NULL;
564 fs->s_block_bitmap_number[i] = 0;
565 fs->s_block_bitmap[i] = NULL;
567 fs->s_loaded_inode_bitmaps = 0;
568 fs->s_loaded_block_bitmaps = 0;
573 * Reload all incore data for a filesystem (used after running fsck on
574 * the root filesystem and finding things to fix). The filesystem must
575 * be mounted read-only.
577 * Things to do to update the mount:
578 * 1) invalidate all cached meta-data.
579 * 2) re-read superblock from disk.
580 * 3) re-read summary information from disk.
581 * 4) invalidate all inactive vnodes.
582 * 5) invalidate all cached file data.
583 * 6) re-read inode data for all active vnodes.
585 static int ext2_reload_scan1(struct mount *mp, struct vnode *vp, void *rescan);
586 static int ext2_reload_scan2(struct mount *mp, struct vnode *vp, void *rescan);
593 struct ext2_sb_info *fs;
597 ext2_reload(struct mount *mountp, struct ucred *cred)
601 struct ext2_super_block * es;
602 struct ext2_sb_info *fs;
604 struct scaninfo scaninfo;
606 if ((mountp->mnt_flag & MNT_RDONLY) == 0)
609 * Step 1: invalidate all cached meta-data.
611 devvp = VFSTOEXT2(mountp)->um_devvp;
612 if (vinvalbuf(devvp, 0, 0, 0))
613 panic("ext2_reload: dirty1");
615 * Step 2: re-read superblock from disk.
616 * constants have been adjusted for ext2
618 if ((error = bread(devvp, SBOFF, SBSIZE, &bp)) != 0)
620 es = (struct ext2_super_block *)bp->b_data;
621 if (ext2_check_sb_compat(es, devvp->v_rdev, 0) != 0) {
623 return (EIO); /* XXX needs translation */
625 fs = VFSTOEXT2(mountp)->um_e2fs;
626 bcopy(bp->b_data, fs->s_es, sizeof(struct ext2_super_block));
628 if((error = compute_sb_data(devvp, es, fs)) != 0) {
633 if (fs->fs_sbsize < SBSIZE)
634 bp->b_flags |= B_INVAL;
639 scaninfo.devvp = devvp;
641 while (error == 0 && scaninfo.rescan) {
643 error = vmntvnodescan(mountp, VMSC_GETVX, ext2_reload_scan1,
644 ext2_reload_scan2, &scaninfo);
650 ext2_reload_scan1(struct mount *mp, struct vnode *vp, void *data)
652 /*struct scaninfo *info = data;*/
658 ext2_reload_scan2(struct mount *mp, struct vnode *vp, void *data)
660 struct scaninfo *info = data;
672 * Step 5: invalidate all cached file data.
674 if (vinvalbuf(vp, 0, 0, 0))
675 panic("ext2_reload: dirty2");
677 * Step 6: re-read inode data for all active vnodes.
680 error = bread(info->devvp,
681 fsbtodoff(info->fs, ino_to_fsba(info->fs, ip->i_number)),
682 (int)info->fs->s_blocksize, &bp);
685 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data +
686 EXT2_INODE_SIZE * ino_to_fsbo(info->fs, ip->i_number)),
693 * Common code for mount and mountroot
696 ext2_mountfs(struct vnode *devvp, struct mount *mp, struct ucred *cred)
698 struct ext2mount *ump;
700 struct ext2_sb_info *fs;
701 struct ext2_super_block * es;
703 struct partinfo dpart;
708 * Disallow multiple mounts of the same device.
709 * Disallow mounting of a device that is currently in use
710 * (except for root, which might share swap device for miniroot).
711 * Flush out any old buffers remaining from a previous use.
713 if ((error = vfs_mountedon(devvp)) != 0)
715 if (count_udev(devvp->v_umajor, devvp->v_uminor) > 0)
717 if ((error = vinvalbuf(devvp, V_SAVE, 0, 0)) != 0)
720 /* turn on this to force it to be read-only */
721 mp->mnt_flag |= MNT_RDONLY;
724 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
725 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
726 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, NULL);
731 if (dev->si_iosize_max != 0)
732 mp->mnt_iosize_max = dev->si_iosize_max;
733 if (mp->mnt_iosize_max > MAXPHYS)
734 mp->mnt_iosize_max = MAXPHYS;
735 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, cred) != 0) {
738 size = dpart.media_blksize;
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 | M_ZERO);
763 ump->um_malloctype = M_EXT2NODE;
764 ump->um_blkatoff = ext2_blkatoff;
765 ump->um_truncate = ext2_truncate;
766 ump->um_update = ext2_update;
767 ump->um_valloc = ext2_valloc;
768 ump->um_vfree = ext2_vfree;
769 /* I don't know whether this is the right strategy. Note that
770 we dynamically allocate both a ext2_sb_info and a ext2_super_block
771 while Linux keeps the super block in a locked buffer
773 ump->um_e2fs = bsd_malloc(sizeof(struct ext2_sb_info),
774 M_EXT2MNT, M_WAITOK);
775 ump->um_e2fs->s_es = bsd_malloc(sizeof(struct ext2_super_block),
776 M_EXT2MNT, M_WAITOK);
777 bcopy(es, ump->um_e2fs->s_es, (u_int)sizeof(struct ext2_super_block));
778 if ((error = compute_sb_data(devvp, ump->um_e2fs->s_es, ump->um_e2fs)))
781 * We don't free the group descriptors allocated by compute_sb_data()
782 * until ext2_unmount(). This is OK since the mount will succeed.
787 fs->s_rd_only = ronly; /* ronly is set according to mnt_flags */
788 /* if the fs is not mounted read-only, make sure the super block is
789 always written back on a sync()
791 fs->s_wasvalid = fs->s_es->s_state & EXT2_VALID_FS ? 1 : 0;
793 fs->s_dirt = 1; /* mark it modified */
794 fs->s_es->s_state &= ~EXT2_VALID_FS; /* set fs invalid */
796 mp->mnt_data = (qaddr_t)ump;
797 mp->mnt_stat.f_fsid.val[0] = dev2udev(dev);
798 mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
799 mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN;
800 mp->mnt_flag |= MNT_LOCAL;
803 ump->um_devvp = devvp;
804 /* setting those two parameters allows us to use
805 ext2_bmap w/o changse !
807 ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs);
808 ump->um_bptrtodb = fs->s_es->s_log_block_size + 1;
809 ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs);
810 for (i = 0; i < MAXQUOTAS; i++)
811 ump->um_quotas[i] = NULLVP;
812 dev->si_mountpoint = mp;
814 vfs_add_vnodeops(mp, &ext2_vnode_vops, &mp->mnt_vn_norm_ops);
815 vfs_add_vnodeops(mp, &ext2_spec_vops, &mp->mnt_vn_spec_ops);
816 vfs_add_vnodeops(mp, &ext2_fifo_vops, &mp->mnt_vn_fifo_ops);
819 ext2_sbupdate(ump, MNT_WAIT);
824 VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE);
826 bsd_free(ump->um_e2fs->s_es, M_EXT2MNT);
827 bsd_free(ump->um_e2fs, M_EXT2MNT);
828 bsd_free(ump, M_EXT2MNT);
829 mp->mnt_data = (qaddr_t)0;
835 * unmount system call
838 ext2_unmount(struct mount *mp, int mntflags)
840 struct ext2mount *ump;
841 struct ext2_sb_info *fs;
842 int error, flags, ronly, i;
845 if (mntflags & MNT_FORCE) {
846 if (mp->mnt_flag & MNT_ROOTFS)
850 if ((error = ext2_flushfiles(mp, flags)) != 0)
854 ronly = fs->s_rd_only;
857 fs->s_es->s_state |= EXT2_VALID_FS;
858 ext2_sbupdate(ump, MNT_WAIT);
861 /* release buffers containing group descriptors */
862 for(i = 0; i < fs->s_db_per_group; i++)
863 ULCK_BUF(fs->s_group_desc[i])
864 bsd_free(fs->s_group_desc, M_EXT2MNT);
866 /* release cached inode/block bitmaps */
867 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
868 if (fs->s_inode_bitmap[i])
869 ULCK_BUF(fs->s_inode_bitmap[i])
871 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
872 if (fs->s_block_bitmap[i])
873 ULCK_BUF(fs->s_block_bitmap[i])
875 ump->um_devvp->v_rdev->si_mountpoint = NULL;
876 error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD|FWRITE);
877 vrele(ump->um_devvp);
878 bsd_free(fs->s_es, M_EXT2MNT);
879 bsd_free(fs, M_EXT2MNT);
880 bsd_free(ump, M_EXT2MNT);
881 mp->mnt_data = (qaddr_t)0;
882 mp->mnt_flag &= ~MNT_LOCAL;
887 * Flush out all the files in a filesystem.
890 ext2_flushfiles(struct mount *mp, int flags)
892 struct ext2mount *ump;
900 if (mp->mnt_flag & MNT_QUOTA) {
901 if ((error = vflush(mp, 0, SKIPSYSTEM|flags)) != 0)
903 for (i = 0; i < MAXQUOTAS; i++) {
904 if (ump->um_quotas[i] == NULLVP)
906 ext2_quotaoff(mp, i);
909 * Here we fall through to vflush again to ensure
910 * that we have gotten rid of all the system vnodes.
914 error = vflush(mp, 0, flags);
919 * Get file system statistics.
920 * taken from ext2/super.c ext2_statfs
923 ext2_statfs(struct mount *mp, struct statfs *sbp, struct ucred *cred)
925 unsigned long overhead;
926 struct ext2mount *ump;
927 struct ext2_sb_info *fs;
928 struct ext2_super_block *es;
935 if (es->s_magic != EXT2_SUPER_MAGIC)
936 panic("ext2_statfs - magic number spoiled");
939 * Compute the overhead (FS structures)
941 if (es->s_feature_ro_compat & EXT2_FEATURE_RO_COMPAT_SPARSE_SUPER) {
943 for (i = 0 ; i < fs->s_groups_count; i++)
944 if (ext2_group_sparse(i))
947 nsb = fs->s_groups_count;
948 overhead = es->s_first_data_block +
949 /* Superblocks and block group descriptors: */
950 nsb * (1 + fs->s_db_per_group) +
951 /* Inode bitmap, block bitmap, and inode table: */
952 fs->s_groups_count * (1 + 1 + fs->s_itb_per_group);
954 sbp->f_bsize = EXT2_FRAG_SIZE(fs);
955 sbp->f_iosize = EXT2_BLOCK_SIZE(fs);
956 sbp->f_blocks = es->s_blocks_count - overhead;
957 sbp->f_bfree = es->s_free_blocks_count;
958 sbp->f_bavail = sbp->f_bfree - es->s_r_blocks_count;
959 sbp->f_files = es->s_inodes_count;
960 sbp->f_ffree = es->s_free_inodes_count;
961 if (sbp != &mp->mnt_stat) {
962 sbp->f_type = mp->mnt_vfc->vfc_typenum;
963 bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
964 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
970 * Go through the disk queues to initiate sandbagged IO;
971 * go through the inodes to write those that have been modified;
972 * initiate the writing of the super block if it has been modified.
974 * Note: we are always called with the filesystem marked `MPBUSY'.
977 static int ext2_sync_scan(struct mount *mp, struct vnode *vp, void *data);
980 ext2_sync(struct mount *mp, int waitfor)
982 struct ext2mount *ump = VFSTOEXT2(mp);
983 struct ext2_sb_info *fs;
984 struct scaninfo scaninfo;
988 if (fs->s_dirt != 0 && fs->s_rd_only != 0) { /* XXX */
989 kprintf("fs = %s\n", fs->fs_fsmnt);
990 panic("ext2_sync: rofs mod");
994 * Write back each (modified) inode.
996 scaninfo.allerror = 0;
998 scaninfo.waitfor = waitfor;
999 while (scaninfo.rescan) {
1000 scaninfo.rescan = 0;
1001 vmntvnodescan(mp, VMSC_GETVP|VMSC_NOWAIT,
1002 NULL, ext2_sync_scan, &scaninfo);
1006 * Force stale file system control information to be flushed.
1008 if (waitfor != MNT_LAZY) {
1009 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
1010 if ((error = VOP_FSYNC(ump->um_devvp, waitfor)) != 0)
1011 scaninfo.allerror = error;
1012 vn_unlock(ump->um_devvp);
1018 * Write back modified superblock.
1020 if (fs->s_dirt != 0) {
1022 fs->s_es->s_wtime = time_second;
1023 if ((error = ext2_sbupdate(ump, waitfor)) != 0)
1024 scaninfo.allerror = error;
1026 return (scaninfo.allerror);
1030 ext2_sync_scan(struct mount *mp, struct vnode *vp, void *data)
1032 struct scaninfo *info = data;
1037 if (vp->v_type == VNON ||
1039 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
1040 (RB_EMPTY(&vp->v_rbdirty_tree) || info->waitfor == MNT_LAZY))) {
1043 if ((error = VOP_FSYNC(vp, info->waitfor)) != 0)
1044 info->allerror = error;
1049 * Look up a EXT2FS dinode number to find its incore vnode, otherwise read it
1050 * in from disk. If it is in core, wait for the lock bit to clear, then
1051 * return the inode locked. Detection and handling of mount points must be
1052 * done by the calling routine.
1055 ext2_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
1057 struct ext2_sb_info *fs;
1059 struct ext2mount *ump;
1066 ump = VFSTOEXT2(mp);
1069 if ((*vpp = ext2_ihashget(dev, ino)) != NULL)
1073 * Lock out the creation of new entries in the FFS hash table in
1074 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate
1077 if (ext2fs_inode_hash_lock) {
1078 while (ext2fs_inode_hash_lock) {
1079 ext2fs_inode_hash_lock = -1;
1080 tsleep(&ext2fs_inode_hash_lock, 0, "e2vget", 0);
1084 ext2fs_inode_hash_lock = 1;
1087 * If this MALLOC() is performed after the getnewvnode()
1088 * it might block, leaving a vnode with a NULL v_data to be
1089 * found by ext2_sync() if a sync happens to fire right then,
1090 * which will cause a panic because ext2_sync() blindly
1091 * dereferences vp->v_data (as well it should).
1093 MALLOC(ip, struct inode *, sizeof(struct inode), M_EXT2NODE, M_WAITOK);
1095 /* Allocate a new vnode/inode. */
1096 if ((error = getnewvnode(VT_EXT2FS, mp, &vp, 0, LK_CANRECURSE)) != 0) {
1097 if (ext2fs_inode_hash_lock < 0)
1098 wakeup(&ext2fs_inode_hash_lock);
1099 ext2fs_inode_hash_lock = 0;
1101 FREE(ip, M_EXT2NODE);
1104 bzero((caddr_t)ip, sizeof(struct inode));
1107 ip->i_e2fs = fs = ump->um_e2fs;
1111 for (i = 0; i < MAXQUOTAS; i++)
1112 ip->i_dquot[i] = NODQUOT;
1115 * Put it onto its hash chain. Since our vnode is locked, other
1116 * requests for this inode will block if they arrive while we are
1117 * sleeping waiting for old data structures to be purged or for the
1118 * contents of the disk portion of this inode to be read.
1122 if (ext2fs_inode_hash_lock < 0)
1123 wakeup(&ext2fs_inode_hash_lock);
1124 ext2fs_inode_hash_lock = 0;
1126 /* Read in the disk contents for the inode, copy into the inode. */
1128 kprintf("ext2_vget(%d) dbn= %d ", ino, fsbtodb(fs, ino_to_fsba(fs, ino)));
1130 error = bread(ump->um_devvp, fsbtodoff(fs, ino_to_fsba(fs, ino)),
1131 (int)fs->s_blocksize, &bp);
1134 * The inode does not contain anything useful, so it would
1135 * be misleading to leave it on its hash chain. With mode
1136 * still zero, it will be unlinked and returned to the free
1144 /* convert ext2 inode to dinode */
1145 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data + EXT2_INODE_SIZE *
1146 ino_to_fsbo(fs, ino)), &ip->i_din);
1147 ip->i_block_group = ino_to_cg(fs, ino);
1148 ip->i_next_alloc_block = 0;
1149 ip->i_next_alloc_goal = 0;
1150 ip->i_prealloc_count = 0;
1151 ip->i_prealloc_block = 0;
1152 /* now we want to make sure that block pointers for unused
1153 blocks are zeroed out - ext2_balloc depends on this
1154 although for regular files and directories only
1156 if(S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode)) {
1157 used_blocks = (ip->i_size+fs->s_blocksize-1) / fs->s_blocksize;
1158 for(i = used_blocks; i < EXT2_NDIR_BLOCKS; i++)
1162 ext2_print_inode(ip);
1167 * Initialize the vnode from the inode, check for aliases.
1168 * Note that the underlying vnode may have changed.
1170 if ((error = ext2_vinit(mp, &vp)) != 0) {
1177 * Finish inode initialization now that aliasing has been resolved.
1179 ip->i_devvp = ump->um_devvp;
1182 * Set up a generation number for this inode if it does not
1183 * already have one. This should only happen on old filesystems.
1185 if (ip->i_gen == 0) {
1186 ip->i_gen = krandom() / 2 + 1;
1187 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1188 ip->i_flag |= IN_MODIFIED;
1191 * Return the locked and refd vnode.
1198 * File handle to vnode
1200 * Have to be really careful about stale file handles:
1201 * - check that the inode number is valid
1202 * - call ext2_vget() to get the locked inode
1203 * - check for an unallocated inode (i_mode == 0)
1204 * - check that the given client host has export rights and return
1205 * those rights via. exflagsp and credanonp
1208 ext2_fhtovp(struct mount *mp, struct vnode *rootvp,
1209 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