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 #define bsd_malloc kmalloc
113 #define bsd_free kfree
115 static int ext2fs_inode_hash_lock;
117 static int ext2_check_sb_compat (struct ext2_super_block *es,
118 cdev_t dev, int ronly);
119 static int compute_sb_data (struct vnode * devvp,
120 struct ext2_super_block * es,
121 struct ext2_sb_info * fs);
124 ext2_root(struct mount *mp, struct vnode **vpp)
129 error = VFS_VGET(mp, NULL, (ino_t)ROOTINO, &nvp);
137 * Do operations associated with quotas
140 ext2_quotactl(struct mount *mp, int cmds, uid_t uid, caddr_t arg,
146 int cmd, type, error;
148 type = cmds & SUBCMDMASK;
149 cmd = cmds >> SUBCMDSHIFT;
170 error = priv_check_cred(cred, PRIV_UFS_QUOTAON, 0);
174 error = priv_check_cred(cred, PRIV_UFS_QUOTAOFF, 0);
178 error = priv_check_cred(cred, PRIV_VFS_SETQUOTA, 0);
182 error = priv_check_cred(cred, PRIV_UFS_SETUSE, 0);
186 if (uid == cred->cr_ruid)
189 error = priv_check_cred(cred, PRIV_VFS_GETQUOTA, 0);
205 if ((uint)type >= MAXQUOTAS)
207 if (vfs_busy(mp, LK_NOWAIT))
213 error = ext2_quotaon(cred, mp, type, arg);
217 error = ext2_quotaoff(mp, type);
221 error = ext2_setquota(mp, uid, type, arg);
225 error = ext2_setuse(mp, uid, type, arg);
229 error = ext2_getquota(mp, uid, type, arg);
233 error = ext2_qsync(mp);
246 * Initial UFS filesystems, done only once.
249 ext2_init(struct vfsconf *vfsp)
269 * data: this is actually a (struct ext2_args *)
272 ext2_mount(struct mount *mp, char *path, caddr_t data, struct ucred *cred)
275 struct ext2_args args;
276 struct ext2mount *ump = 0;
277 struct ext2_sb_info *fs;
281 struct nlookupdata nd;
283 if ((error = copyin(data, (caddr_t)&args, sizeof (struct ext2_args))) != 0)
287 * If updating, check whether changing from read-only to
288 * read/write; if there is no device name, that's all we do.
290 if (mp->mnt_flag & MNT_UPDATE) {
293 devvp = ump->um_devvp;
295 if (fs->s_rd_only == 0 && (mp->mnt_flag & MNT_RDONLY)) {
297 if (mp->mnt_flag & MNT_FORCE)
299 if (vfs_busy(mp, LK_NOWAIT))
301 error = ext2_flushfiles(mp, flags);
303 if (!error && fs->s_wasvalid) {
304 fs->s_es->s_state |= EXT2_VALID_FS;
305 ext2_sbupdate(ump, MNT_WAIT);
308 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
309 VOP_OPEN(devvp, FREAD, FSCRED, NULL);
310 VOP_CLOSE(devvp, FREAD|FWRITE);
313 if (!error && (mp->mnt_flag & MNT_RELOAD))
314 error = ext2_reload(mp, cred);
317 if (ext2_check_sb_compat(fs->s_es, devvp->v_rdev,
318 (mp->mnt_kern_flag & MNTK_WANTRDWR) == 0) != 0)
320 if (fs->s_rd_only && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
322 * If upgrade to read-write by non-root, then verify
323 * that user has necessary permissions on the device.
325 if (cred->cr_uid != 0) {
326 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
327 error = VOP_EACCESS(devvp, VREAD | VWRITE, cred);
335 if ((fs->s_es->s_state & EXT2_VALID_FS) == 0 ||
336 (fs->s_es->s_state & EXT2_ERROR_FS)) {
337 if (mp->mnt_flag & MNT_FORCE) {
339 "WARNING: %s was not properly dismounted\n",
343 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
348 fs->s_es->s_state &= ~EXT2_VALID_FS;
349 ext2_sbupdate(ump, MNT_WAIT);
351 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
352 VOP_OPEN(devvp, FREAD|FWRITE, FSCRED, NULL);
353 VOP_CLOSE(devvp, FREAD);
356 if (args.fspec == 0) {
358 * Process export requests.
360 return (vfs_export(mp, &ump->um_export, &args.export));
364 * Not an update, or updating the name: look up the name
365 * and verify that it refers to a sensible block device.
368 error = nlookup_init(&nd, args.fspec, UIO_USERSPACE, NLC_FOLLOW);
370 error = nlookup(&nd);
372 error = cache_vref(&nd.nl_nch, nd.nl_cred, &devvp);
377 if (!vn_isdisk(devvp, &error)) {
383 * If mount by non-root, then verify that user has necessary
384 * permissions on the device.
386 if (cred->cr_uid != 0) {
388 if ((mp->mnt_flag & MNT_RDONLY) == 0)
389 accessmode |= VWRITE;
390 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
391 if ((error = VOP_EACCESS(devvp, accessmode, cred)) != 0) {
398 if ((mp->mnt_flag & MNT_UPDATE) == 0) {
399 error = ext2_mountfs(devvp, mp, cred);
401 if (devvp != ump->um_devvp)
402 error = EINVAL; /* needs translation */
412 copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size);
413 bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size);
414 copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, &size);
415 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
416 ext2_statfs(mp, &mp->mnt_stat, cred);
421 * checks that the data in the descriptor blocks make sense
422 * this is taken from ext2/super.c
425 ext2_check_descriptors(struct ext2_sb_info *sb)
429 unsigned long block = sb->s_es->s_first_data_block;
430 struct ext2_group_desc * gdp = NULL;
432 /* ext2_debug ("Checking group descriptors"); */
434 for (i = 0; i < sb->s_groups_count; i++)
436 /* examine next descriptor block */
437 if ((i % EXT2_DESC_PER_BLOCK(sb)) == 0)
438 gdp = (struct ext2_group_desc *)
439 sb->s_group_desc[desc_block++]->b_data;
440 if (gdp->bg_block_bitmap < block ||
441 gdp->bg_block_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb))
443 kprintf ("ext2_check_descriptors: "
444 "Block bitmap for group %d"
445 " not in group (block %lu)!\n",
446 i, (unsigned long) gdp->bg_block_bitmap);
449 if (gdp->bg_inode_bitmap < block ||
450 gdp->bg_inode_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb))
452 kprintf ("ext2_check_descriptors: "
453 "Inode bitmap for group %d"
454 " not in group (block %lu)!\n",
455 i, (unsigned long) gdp->bg_inode_bitmap);
458 if (gdp->bg_inode_table < block ||
459 gdp->bg_inode_table + sb->s_itb_per_group >=
460 block + EXT2_BLOCKS_PER_GROUP(sb))
462 kprintf ("ext2_check_descriptors: "
463 "Inode table for group %d"
464 " not in group (block %lu)!\n",
465 i, (unsigned long) gdp->bg_inode_table);
468 block += EXT2_BLOCKS_PER_GROUP(sb);
475 ext2_check_sb_compat(struct ext2_super_block *es, cdev_t dev, int ronly)
477 if (es->s_magic != EXT2_SUPER_MAGIC) {
478 kprintf("ext2fs: %s: wrong magic number %#x (expected %#x)\n",
479 devtoname(dev), es->s_magic, EXT2_SUPER_MAGIC);
482 if (es->s_rev_level > EXT2_GOOD_OLD_REV) {
483 if (es->s_feature_incompat & ~EXT2_FEATURE_INCOMPAT_SUPP) {
485 "WARNING: mount of %s denied due to unsupported optional features\n",
490 (es->s_feature_ro_compat & ~EXT2_FEATURE_RO_COMPAT_SUPP)) {
492 "WARNING: R/W mount of %s denied due to unsupported optional features\n",
501 * this computes the fields of the ext2_sb_info structure from the
502 * data in the ext2_super_block structure read in
505 compute_sb_data(struct vnode *devvp, struct ext2_super_block *es,
506 struct ext2_sb_info *fs)
510 int logic_sb_block = 1; /* XXX for now */
515 #define V(v) kprintf(#v"= %d\n", fs->v);
518 fs->s_blocksize = EXT2_MIN_BLOCK_SIZE << es->s_log_block_size;
520 fs->s_bshift = EXT2_MIN_BLOCK_LOG_SIZE + es->s_log_block_size;
522 fs->s_fsbtodb = es->s_log_block_size + 1;
524 fs->s_qbmask = fs->s_blocksize - 1;
526 fs->s_blocksize_bits = EXT2_BLOCK_SIZE_BITS(es);
528 fs->s_frag_size = EXT2_MIN_FRAG_SIZE << es->s_log_frag_size;
531 fs->s_frags_per_block = fs->s_blocksize / fs->s_frag_size;
533 fs->s_blocks_per_group = es->s_blocks_per_group;
534 V(s_blocks_per_group)
535 fs->s_frags_per_group = es->s_frags_per_group;
537 fs->s_inodes_per_group = es->s_inodes_per_group;
538 V(s_inodes_per_group)
539 if (es->s_rev_level == EXT2_GOOD_OLD_REV) {
540 fs->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
541 fs->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
543 fs->s_first_ino = es->s_first_ino;
544 fs->s_inode_size = es->s_inode_size;
546 * Simple sanity check for superblock inode size value.
548 if (fs->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE ||
549 fs->s_inode_size > fs->s_blocksize ||
550 (fs->s_inode_size & (fs->s_inode_size - 1)) != 0) {
551 kprintf("EXT2-fs: invalid inode size %d\n",
558 fs->s_inodes_per_block = fs->s_blocksize / EXT2_INODE_SIZE(fs);
559 V(s_inodes_per_block)
560 fs->s_itb_per_group = fs->s_inodes_per_group /fs->s_inodes_per_block;
562 fs->s_desc_per_block = fs->s_blocksize / sizeof (struct ext2_group_desc);
564 /* s_resuid / s_resgid ? */
565 fs->s_groups_count = (es->s_blocks_count -
566 es->s_first_data_block +
567 EXT2_BLOCKS_PER_GROUP(fs) - 1) /
568 EXT2_BLOCKS_PER_GROUP(fs);
570 db_count = (fs->s_groups_count + EXT2_DESC_PER_BLOCK(fs) - 1) /
571 EXT2_DESC_PER_BLOCK(fs);
572 fs->s_db_per_group = db_count;
575 fs->s_group_desc = bsd_malloc(db_count * sizeof (struct buf *),
576 M_EXT2MNT, M_WAITOK);
578 /* adjust logic_sb_block */
579 if(fs->s_blocksize > SBSIZE)
580 /* Godmar thinks: if the blocksize is greater than 1024, then
581 the superblock is logically part of block zero.
585 for (i = 0; i < db_count; i++) {
586 error = bread(devvp, fsbtodoff(fs, logic_sb_block + i + 1),
587 fs->s_blocksize, &fs->s_group_desc[i]);
589 for (j = 0; j < i; j++)
590 brelse(fs->s_group_desc[j]);
591 bsd_free(fs->s_group_desc, M_EXT2MNT);
592 kprintf("EXT2-fs: unable to read group descriptors (%d)\n", error);
595 /* Set the B_LOCKED flag on the buffer, then brelse() it */
596 LCK_BUF(fs->s_group_desc[i])
598 if(!ext2_check_descriptors(fs)) {
599 for (j = 0; j < db_count; j++)
600 ULCK_BUF(fs->s_group_desc[j])
601 bsd_free(fs->s_group_desc, M_EXT2MNT);
602 kprintf("EXT2-fs: (ext2_check_descriptors failure) "
603 "unable to read group descriptors\n");
607 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) {
608 fs->s_inode_bitmap_number[i] = 0;
609 fs->s_inode_bitmap[i] = NULL;
610 fs->s_block_bitmap_number[i] = 0;
611 fs->s_block_bitmap[i] = NULL;
613 fs->s_loaded_inode_bitmaps = 0;
614 fs->s_loaded_block_bitmaps = 0;
619 * Reload all incore data for a filesystem (used after running fsck on
620 * the root filesystem and finding things to fix). The filesystem must
621 * be mounted read-only.
623 * Things to do to update the mount:
624 * 1) invalidate all cached meta-data.
625 * 2) re-read superblock from disk.
626 * 3) re-read summary information from disk.
627 * 4) invalidate all inactive vnodes.
628 * 5) invalidate all cached file data.
629 * 6) re-read inode data for all active vnodes.
631 static int ext2_reload_scan1(struct mount *mp, struct vnode *vp, void *rescan);
632 static int ext2_reload_scan2(struct mount *mp, struct vnode *vp, void *rescan);
639 struct ext2_sb_info *fs;
643 ext2_reload(struct mount *mountp, struct ucred *cred)
647 struct ext2_super_block * es;
648 struct ext2_sb_info *fs;
650 struct scaninfo scaninfo;
652 if ((mountp->mnt_flag & MNT_RDONLY) == 0)
655 * Step 1: invalidate all cached meta-data.
657 devvp = VFSTOEXT2(mountp)->um_devvp;
658 if (vinvalbuf(devvp, 0, 0, 0))
659 panic("ext2_reload: dirty1");
661 * Step 2: re-read superblock from disk.
662 * constants have been adjusted for ext2
664 if ((error = bread(devvp, SBOFF, SBSIZE, &bp)) != 0)
666 es = (struct ext2_super_block *)bp->b_data;
667 if (ext2_check_sb_compat(es, devvp->v_rdev, 0) != 0) {
669 return (EIO); /* XXX needs translation */
671 fs = VFSTOEXT2(mountp)->um_e2fs;
672 bcopy(bp->b_data, fs->s_es, sizeof(struct ext2_super_block));
674 if((error = compute_sb_data(devvp, es, fs)) != 0) {
679 if (fs->fs_sbsize < SBSIZE)
680 bp->b_flags |= B_INVAL;
685 scaninfo.devvp = devvp;
687 while (error == 0 && scaninfo.rescan) {
689 error = vmntvnodescan(mountp, VMSC_GETVX, ext2_reload_scan1,
690 ext2_reload_scan2, &scaninfo);
696 ext2_reload_scan1(struct mount *mp, struct vnode *vp, void *data)
698 /*struct scaninfo *info = data;*/
704 ext2_reload_scan2(struct mount *mp, struct vnode *vp, void *data)
706 struct scaninfo *info = data;
718 * Step 5: invalidate all cached file data.
720 if (vinvalbuf(vp, 0, 0, 0))
721 panic("ext2_reload: dirty2");
723 * Step 6: re-read inode data for all active vnodes.
726 error = bread(info->devvp,
727 fsbtodoff(info->fs, ino_to_fsba(info->fs, ip->i_number)),
728 (int)info->fs->s_blocksize, &bp);
731 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data +
732 EXT2_INODE_SIZE(info->fs) * ino_to_fsbo(info->fs, ip->i_number)),
739 * Common code for mount and mountroot
742 ext2_mountfs(struct vnode *devvp, struct mount *mp, struct ucred *cred)
744 struct ext2mount *ump;
746 struct ext2_sb_info *fs;
747 struct ext2_super_block * es;
749 struct partinfo dpart;
754 * Disallow multiple mounts of the same device.
755 * Disallow mounting of a device that is currently in use
756 * (except for root, which might share swap device for miniroot).
757 * Flush out any old buffers remaining from a previous use.
759 if ((error = vfs_mountedon(devvp)) != 0)
761 if (vcount(devvp) > 0)
763 if ((error = vinvalbuf(devvp, V_SAVE, 0, 0)) != 0)
766 /* turn on this to force it to be read-only */
767 mp->mnt_flag |= MNT_RDONLY;
770 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
771 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
772 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, NULL);
777 if (dev->si_iosize_max != 0)
778 mp->mnt_iosize_max = dev->si_iosize_max;
779 if (mp->mnt_iosize_max > MAXPHYS)
780 mp->mnt_iosize_max = MAXPHYS;
781 VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, cred, NULL);
785 if ((error = bread(devvp, SBOFF, SBSIZE, &bp)) != 0)
787 es = (struct ext2_super_block *)bp->b_data;
788 if (ext2_check_sb_compat(es, dev, ronly) != 0) {
789 error = EINVAL; /* XXX needs translation */
792 if ((es->s_state & EXT2_VALID_FS) == 0 ||
793 (es->s_state & EXT2_ERROR_FS)) {
794 if (ronly || (mp->mnt_flag & MNT_FORCE)) {
796 "WARNING: Filesystem was not properly dismounted\n");
799 "WARNING: R/W mount denied. Filesystem is not clean - run fsck\n");
804 ump = bsd_malloc(sizeof *ump, M_EXT2MNT, M_WAITOK | M_ZERO);
805 ump->um_malloctype = M_EXT2NODE;
806 ump->um_blkatoff = ext2_blkatoff;
807 ump->um_truncate = ext2_truncate;
808 ump->um_update = ext2_update;
809 ump->um_valloc = ext2_valloc;
810 ump->um_vfree = ext2_vfree;
811 /* I don't know whether this is the right strategy. Note that
812 we dynamically allocate both a ext2_sb_info and a ext2_super_block
813 while Linux keeps the super block in a locked buffer
815 ump->um_e2fs = bsd_malloc(sizeof(struct ext2_sb_info),
816 M_EXT2MNT, M_WAITOK);
817 ump->um_e2fs->s_es = bsd_malloc(sizeof(struct ext2_super_block),
818 M_EXT2MNT, M_WAITOK);
819 bcopy(es, ump->um_e2fs->s_es, (u_int)sizeof(struct ext2_super_block));
820 if ((error = compute_sb_data(devvp, ump->um_e2fs->s_es, ump->um_e2fs)))
823 * We don't free the group descriptors allocated by compute_sb_data()
824 * until ext2_unmount(). This is OK since the mount will succeed.
829 fs->s_rd_only = ronly; /* ronly is set according to mnt_flags */
830 /* if the fs is not mounted read-only, make sure the super block is
831 always written back on a sync()
833 fs->s_wasvalid = fs->s_es->s_state & EXT2_VALID_FS ? 1 : 0;
835 fs->s_dirt = 1; /* mark it modified */
836 fs->s_es->s_state &= ~EXT2_VALID_FS; /* set fs invalid */
838 mp->mnt_data = (qaddr_t)ump;
839 mp->mnt_stat.f_fsid.val[0] = dev2udev(dev);
840 mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
841 mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN;
842 mp->mnt_flag |= MNT_LOCAL;
845 ump->um_devvp = devvp;
846 /* setting those two parameters allows us to use
847 ext2_bmap w/o changse !
849 ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs);
850 ump->um_bptrtodb = fs->s_es->s_log_block_size + 1;
851 ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs);
852 for (i = 0; i < MAXQUOTAS; i++)
853 ump->um_quotas[i] = NULLVP;
854 dev->si_mountpoint = mp;
856 vfs_add_vnodeops(mp, &ext2_vnode_vops, &mp->mnt_vn_norm_ops);
857 vfs_add_vnodeops(mp, &ext2_spec_vops, &mp->mnt_vn_spec_ops);
858 vfs_add_vnodeops(mp, &ext2_fifo_vops, &mp->mnt_vn_fifo_ops);
861 ext2_sbupdate(ump, MNT_WAIT);
866 VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE);
868 bsd_free(ump->um_e2fs->s_es, M_EXT2MNT);
869 bsd_free(ump->um_e2fs, M_EXT2MNT);
870 bsd_free(ump, M_EXT2MNT);
871 mp->mnt_data = (qaddr_t)0;
877 * unmount system call
880 ext2_unmount(struct mount *mp, int mntflags)
882 struct ext2mount *ump;
883 struct ext2_sb_info *fs;
884 int error, flags, ronly, i;
887 if (mntflags & MNT_FORCE) {
888 if (mp->mnt_flag & MNT_ROOTFS)
892 if ((error = ext2_flushfiles(mp, flags)) != 0)
896 ronly = fs->s_rd_only;
899 fs->s_es->s_state |= EXT2_VALID_FS;
900 ext2_sbupdate(ump, MNT_WAIT);
903 /* release buffers containing group descriptors */
904 for(i = 0; i < fs->s_db_per_group; i++)
905 ULCK_BUF(fs->s_group_desc[i])
906 bsd_free(fs->s_group_desc, M_EXT2MNT);
908 /* release cached inode/block bitmaps */
909 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
910 if (fs->s_inode_bitmap[i])
911 ULCK_BUF(fs->s_inode_bitmap[i])
913 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
914 if (fs->s_block_bitmap[i])
915 ULCK_BUF(fs->s_block_bitmap[i])
917 ump->um_devvp->v_rdev->si_mountpoint = NULL;
918 error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD|FWRITE);
919 vrele(ump->um_devvp);
920 bsd_free(fs->s_es, M_EXT2MNT);
921 bsd_free(fs, M_EXT2MNT);
922 bsd_free(ump, M_EXT2MNT);
923 mp->mnt_data = (qaddr_t)0;
924 mp->mnt_flag &= ~MNT_LOCAL;
929 * Flush out all the files in a filesystem.
932 ext2_flushfiles(struct mount *mp, int flags)
934 struct ext2mount *ump;
942 if (mp->mnt_flag & MNT_QUOTA) {
943 if ((error = vflush(mp, 0, SKIPSYSTEM|flags)) != 0)
945 for (i = 0; i < MAXQUOTAS; i++) {
946 if (ump->um_quotas[i] == NULLVP)
948 ext2_quotaoff(mp, i);
951 * Here we fall through to vflush again to ensure
952 * that we have gotten rid of all the system vnodes.
956 error = vflush(mp, 0, flags);
961 * Get file system statistics.
962 * taken from ext2/super.c ext2_statfs
965 ext2_statfs(struct mount *mp, struct statfs *sbp, struct ucred *cred)
967 unsigned long overhead;
968 struct ext2mount *ump;
969 struct ext2_sb_info *fs;
970 struct ext2_super_block *es;
977 if (es->s_magic != EXT2_SUPER_MAGIC)
978 panic("ext2_statfs - magic number spoiled");
981 * Compute the overhead (FS structures)
983 if (es->s_feature_ro_compat & EXT2_FEATURE_RO_COMPAT_SPARSE_SUPER) {
985 for (i = 0 ; i < fs->s_groups_count; i++)
986 if (ext2_group_sparse(i))
989 nsb = fs->s_groups_count;
990 overhead = es->s_first_data_block +
991 /* Superblocks and block group descriptors: */
992 nsb * (1 + fs->s_db_per_group) +
993 /* Inode bitmap, block bitmap, and inode table: */
994 fs->s_groups_count * (1 + 1 + fs->s_itb_per_group);
996 sbp->f_bsize = EXT2_FRAG_SIZE(fs);
997 sbp->f_iosize = EXT2_BLOCK_SIZE(fs);
998 sbp->f_blocks = es->s_blocks_count - overhead;
999 sbp->f_bfree = es->s_free_blocks_count;
1000 sbp->f_bavail = sbp->f_bfree - es->s_r_blocks_count;
1001 sbp->f_files = es->s_inodes_count;
1002 sbp->f_ffree = es->s_free_inodes_count;
1003 if (sbp != &mp->mnt_stat) {
1004 sbp->f_type = mp->mnt_vfc->vfc_typenum;
1005 bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
1006 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
1012 * Go through the disk queues to initiate sandbagged IO;
1013 * go through the inodes to write those that have been modified;
1014 * initiate the writing of the super block if it has been modified.
1016 * Note: we are always called with the filesystem marked `MPBUSY'.
1019 static int ext2_sync_scan(struct mount *mp, struct vnode *vp, void *data);
1022 ext2_sync(struct mount *mp, int waitfor)
1024 struct ext2mount *ump = VFSTOEXT2(mp);
1025 struct ext2_sb_info *fs;
1026 struct scaninfo scaninfo;
1030 if (fs->s_dirt != 0 && fs->s_rd_only != 0) { /* XXX */
1031 kprintf("fs = %s\n", fs->fs_fsmnt);
1032 panic("ext2_sync: rofs mod");
1036 * Write back each (modified) inode.
1038 scaninfo.allerror = 0;
1039 scaninfo.rescan = 1;
1040 scaninfo.waitfor = waitfor;
1041 while (scaninfo.rescan) {
1042 scaninfo.rescan = 0;
1043 vmntvnodescan(mp, VMSC_GETVP|VMSC_NOWAIT,
1044 NULL, ext2_sync_scan, &scaninfo);
1048 * Force stale file system control information to be flushed.
1050 if ((waitfor & MNT_LAZY) == 0) {
1051 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
1052 if ((error = VOP_FSYNC(ump->um_devvp, waitfor, 0)) != 0)
1053 scaninfo.allerror = error;
1054 vn_unlock(ump->um_devvp);
1060 * Write back modified superblock.
1062 if (fs->s_dirt != 0) {
1064 fs->s_es->s_wtime = time_second;
1065 if ((error = ext2_sbupdate(ump, waitfor)) != 0)
1066 scaninfo.allerror = error;
1068 return (scaninfo.allerror);
1072 ext2_sync_scan(struct mount *mp, struct vnode *vp, void *data)
1074 struct scaninfo *info = data;
1079 if (vp->v_type == VNON ||
1081 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
1082 (RB_EMPTY(&vp->v_rbdirty_tree) || (info->waitfor & MNT_LAZY)))) {
1085 if ((error = VOP_FSYNC(vp, info->waitfor, 0)) != 0)
1086 info->allerror = error;
1091 * Look up a EXT2FS dinode number to find its incore vnode, otherwise read it
1092 * in from disk. If it is in core, wait for the lock bit to clear, then
1093 * return the inode locked. Detection and handling of mount points must be
1094 * done by the calling routine.
1097 ext2_vget(struct mount *mp, struct vnode *dvp, ino_t ino, struct vnode **vpp)
1099 struct ext2_sb_info *fs;
1101 struct ext2mount *ump;
1108 ump = VFSTOEXT2(mp);
1111 if ((*vpp = ext2_ihashget(dev, ino)) != NULL)
1115 * Lock out the creation of new entries in the FFS hash table in
1116 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate
1119 if (ext2fs_inode_hash_lock) {
1120 while (ext2fs_inode_hash_lock) {
1121 ext2fs_inode_hash_lock = -1;
1122 tsleep(&ext2fs_inode_hash_lock, 0, "e2vget", 0);
1126 ext2fs_inode_hash_lock = 1;
1129 * If this MALLOC() is performed after the getnewvnode()
1130 * it might block, leaving a vnode with a NULL v_data to be
1131 * found by ext2_sync() if a sync happens to fire right then,
1132 * which will cause a panic because ext2_sync() blindly
1133 * dereferences vp->v_data (as well it should).
1135 MALLOC(ip, struct inode *, sizeof(struct inode), M_EXT2NODE, M_WAITOK);
1137 /* Allocate a new vnode/inode. */
1138 if ((error = getnewvnode(VT_EXT2FS, mp, &vp, 0, LK_CANRECURSE)) != 0) {
1139 if (ext2fs_inode_hash_lock < 0)
1140 wakeup(&ext2fs_inode_hash_lock);
1141 ext2fs_inode_hash_lock = 0;
1143 FREE(ip, M_EXT2NODE);
1146 bzero((caddr_t)ip, sizeof(struct inode));
1149 ip->i_e2fs = fs = ump->um_e2fs;
1153 for (i = 0; i < MAXQUOTAS; i++)
1154 ip->i_dquot[i] = NODQUOT;
1157 * Put it onto its hash chain. Since our vnode is locked, other
1158 * requests for this inode will block if they arrive while we are
1159 * sleeping waiting for old data structures to be purged or for the
1160 * contents of the disk portion of this inode to be read.
1164 if (ext2fs_inode_hash_lock < 0)
1165 wakeup(&ext2fs_inode_hash_lock);
1166 ext2fs_inode_hash_lock = 0;
1168 /* Read in the disk contents for the inode, copy into the inode. */
1170 kprintf("ext2_vget(%d) dbn= %d ", ino, fsbtodb(fs, ino_to_fsba(fs, ino)));
1172 error = bread(ump->um_devvp, fsbtodoff(fs, ino_to_fsba(fs, ino)),
1173 (int)fs->s_blocksize, &bp);
1176 * The inode does not contain anything useful, so it would
1177 * be misleading to leave it on its hash chain. With mode
1178 * still zero, it will be unlinked and returned to the free
1186 /* convert ext2 inode to dinode */
1187 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data + EXT2_INODE_SIZE(fs) *
1188 ino_to_fsbo(fs, ino)), &ip->i_din);
1189 ip->i_block_group = ino_to_cg(fs, ino);
1190 ip->i_next_alloc_block = 0;
1191 ip->i_next_alloc_goal = 0;
1192 ip->i_prealloc_count = 0;
1193 ip->i_prealloc_block = 0;
1194 /* now we want to make sure that block pointers for unused
1195 blocks are zeroed out - ext2_balloc depends on this
1196 although for regular files and directories only
1198 if(S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode)) {
1199 used_blocks = (ip->i_size+fs->s_blocksize-1) / fs->s_blocksize;
1200 for(i = used_blocks; i < EXT2_NDIR_BLOCKS; i++)
1204 ext2_print_inode(ip);
1209 * Initialize the vnode from the inode, check for aliases.
1210 * Note that the underlying vnode may have changed.
1212 if ((error = ext2_vinit(mp, &vp)) != 0) {
1219 * Finish inode initialization now that aliasing has been resolved.
1221 ip->i_devvp = ump->um_devvp;
1224 * Set up a generation number for this inode if it does not
1225 * already have one. This should only happen on old filesystems.
1227 if (ip->i_gen == 0) {
1228 ip->i_gen = krandom() / 2 + 1;
1229 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1230 ip->i_flag |= IN_MODIFIED;
1233 * Return the locked and refd vnode.
1240 * File handle to vnode
1242 * Have to be really careful about stale file handles:
1243 * - check that the inode number is valid
1244 * - call ext2_vget() to get the locked inode
1245 * - check for an unallocated inode (i_mode == 0)
1246 * - check that the given client host has export rights and return
1247 * those rights via. exflagsp and credanonp
1250 ext2_fhtovp(struct mount *mp, struct vnode *rootvp,
1251 struct fid *fhp, struct vnode **vpp)
1254 struct ext2_sb_info *fs;
1259 ufhp = (struct ufid *)fhp;
1260 fs = VFSTOEXT2(mp)->um_e2fs;
1261 if (ufhp->ufid_ino < ROOTINO ||
1262 ufhp->ufid_ino > fs->s_groups_count * fs->s_es->s_inodes_per_group)
1265 error = VFS_VGET(mp, rootvp, ufhp->ufid_ino, &nvp);
1271 if (ip->i_mode == 0 ||
1272 ip->i_gen != ufhp->ufid_gen ||
1273 (VFSTOEXT2(mp)->um_i_effnlink_valid ? ip->i_effnlink :
1274 ip->i_nlink) <= 0) {
1284 * Vnode pointer to File handle
1288 ext2_vptofh(struct vnode *vp, struct fid *fhp)
1294 ufhp = (struct ufid *)fhp;
1295 ufhp->ufid_len = sizeof(struct ufid);
1296 ufhp->ufid_ino = ip->i_number;
1297 ufhp->ufid_gen = ip->i_gen;
1302 * This is the generic part of fhtovp called after the underlying
1303 * filesystem has validated the file handle.
1305 * Verify that a host should have access to a filesystem.
1308 ext2_check_export(struct mount *mp, struct sockaddr *nam, int *exflagsp,
1309 struct ucred **credanonp)
1312 struct ext2mount *ump;
1314 ump = VFSTOEXT2(mp);
1316 * Get the export permission structure for this <mp, client> tuple.
1318 np = vfs_export_lookup(mp, &ump->um_export, nam);
1322 *exflagsp = np->netc_exflags;
1323 *credanonp = &np->netc_anon;
1328 * Write a superblock and associated information back to disk.
1331 ext2_sbupdate(struct ext2mount *mp, int waitfor)
1333 struct ext2_sb_info *fs = mp->um_e2fs;
1334 struct ext2_super_block *es = fs->s_es;
1338 kprintf("\nupdating superblock, waitfor=%s\n", waitfor == MNT_WAIT ? "yes":"no");
1340 bp = getblk(mp->um_devvp, SBOFF, SBSIZE, 0, 0);
1341 bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2_super_block));
1342 if (waitfor == MNT_WAIT)
1348 * The buffers for group descriptors, inode bitmaps and block bitmaps
1349 * are not busy at this point and are (hopefully) written by the
1350 * usual sync mechanism. No need to write them here