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 1.22 2004/11/12 00:09:30 dillon 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>
61 #include <vfs/ufs/quota.h>
62 #include <vfs/ufs/ufsmount.h>
63 #include <vfs/ufs/inode.h>
64 #include <vfs/ufs/ufs_extern.h>
66 #include <vm/vm_zone.h>
69 #include "ext2_extern.h"
71 #include "ext2_fs_sb.h"
73 extern struct vnodeopv_entry_desc ext2_vnodeop_entries[];
74 extern struct vnodeopv_entry_desc ext2_specop_entries[];
75 extern struct vnodeopv_entry_desc ext2_fifoop_entries[];
77 static int ext2_fhtovp (struct mount *, struct fid *, struct vnode **);
78 static int ext2_flushfiles (struct mount *mp, int flags, struct thread *td);
79 static int ext2_mount (struct mount *, char *, caddr_t, struct thread *);
80 static int ext2_mountfs (struct vnode *, struct mount *, struct thread *);
81 static int ext2_reload (struct mount *mountp, struct ucred *cred,
83 static int ext2_sbupdate (struct ufsmount *, int);
84 static int ext2_statfs (struct mount *, struct statfs *, struct thread *);
85 static int ext2_sync (struct mount *, int, struct thread *);
86 static int ext2_unmount (struct mount *, int, struct thread *);
87 static int ext2_vget (struct mount *, ino_t, struct vnode **);
88 static int ext2_vptofh (struct vnode *, struct fid *);
90 static MALLOC_DEFINE(M_EXT2NODE, "EXT2 node", "EXT2 vnode private part");
92 static struct vfsops ext2fs_vfsops = {
94 ufs_start, /* empty function */
96 ufs_root, /* root inode via vget */
97 ufs_quotactl, /* does operations associated with quotas */
109 VFS_SET(ext2fs_vfsops, ext2fs, 0);
110 #define bsd_malloc malloc
111 #define bsd_free free
113 static int ext2fs_inode_hash_lock;
115 static int ext2_check_sb_compat (struct ext2_super_block *es,
116 dev_t dev, int ronly);
117 static int compute_sb_data (struct vnode * devvp,
118 struct ext2_super_block * es,
119 struct ext2_sb_info * fs);
122 static int ext2_mountroot (void);
125 * Called by main() when ext2fs is going to be mounted as root.
127 * Name is updated by mount(8) after booting.
129 #define ROOTNAME "root_device"
134 struct ext2_sb_info *fs;
136 struct thread *td = curthread;
137 struct vnode *rootvp;
138 struct ufsmount *ump;
142 if ((error = bdevvp(rootdev, &rootvp))) {
143 printf("ext2_mountroot: can't find rootvp\n");
146 mp = bsd_malloc((u_long)sizeof(struct mount), M_MOUNT, M_WAITOK);
147 bzero((char *)mp, (u_long)sizeof(struct mount));
148 TAILQ_INIT(&mp->mnt_nvnodelist);
149 TAILQ_INIT(&mp->mnt_reservedvnlist);
150 mp->mnt_op = &ext2fs_vfsops;
151 mp->mnt_flag = MNT_RDONLY;
152 if (error = ext2_mountfs(rootvp, mp, td)) {
153 bsd_free(mp, M_MOUNT);
156 if (error = vfs_lock(mp)) {
157 (void)ext2_unmount(mp, 0, td);
158 bsd_free(mp, M_MOUNT);
161 TAILQ_INSERT_HEAD(&mountlist, mp, mnt_list);
162 mp->mnt_flag |= MNT_ROOTFS;
163 mp->mnt_vnodecovered = NULLVP;
166 bzero(fs->fs_fsmnt, sizeof(fs->fs_fsmnt));
167 fs->fs_fsmnt[0] = '/';
168 bcopy((caddr_t)fs->fs_fsmnt, (caddr_t)mp->mnt_stat.f_mntonname,
170 (void) copystr(ROOTNAME, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
172 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
173 (void)ext2_statfs(mp, &mp->mnt_stat, td);
175 inittodr(fs->s_es->s_wtime); /* this helps to set the time */
186 ext2_mount(struct mount *mp, char *path,
187 caddr_t data, /* this is actually a (struct ufs_args *) */
191 struct ufs_args args;
192 struct ufsmount *ump = 0;
193 struct ext2_sb_info *fs;
198 struct nlookupdata nd;
200 if ((error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args))) != 0)
203 cred = td->td_proc->p_ucred;
205 * If updating, check whether changing from read-only to
206 * read/write; if there is no device name, that's all we do.
208 if (mp->mnt_flag & MNT_UPDATE) {
212 if (fs->s_rd_only == 0 && (mp->mnt_flag & MNT_RDONLY)) {
214 if (mp->mnt_flag & MNT_FORCE)
216 if (vfs_busy(mp, LK_NOWAIT, NULL, td))
218 error = ext2_flushfiles(mp, flags, td);
220 if (!error && fs->s_wasvalid) {
221 fs->s_es->s_state |= EXT2_VALID_FS;
222 ext2_sbupdate(ump, MNT_WAIT);
226 if (!error && (mp->mnt_flag & MNT_RELOAD))
227 error = ext2_reload(mp, proc0.p_ucred, td);
230 devvp = ump->um_devvp;
231 if (ext2_check_sb_compat(fs->s_es, devvp->v_rdev,
232 (mp->mnt_kern_flag & MNTK_WANTRDWR) == 0) != 0)
234 if (fs->s_rd_only && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
236 * If upgrade to read-write by non-root, then verify
237 * that user has necessary permissions on the device.
239 if (cred->cr_uid != 0) {
240 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
241 if ((error = VOP_ACCESS(devvp, VREAD | VWRITE,
243 VOP_UNLOCK(devvp, 0, td);
246 VOP_UNLOCK(devvp, 0, td);
249 if ((fs->s_es->s_state & EXT2_VALID_FS) == 0 ||
250 (fs->s_es->s_state & EXT2_ERROR_FS)) {
251 if (mp->mnt_flag & MNT_FORCE) {
253 "WARNING: %s was not properly dismounted\n",
257 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
262 fs->s_es->s_state &= ~EXT2_VALID_FS;
263 ext2_sbupdate(ump, MNT_WAIT);
266 if (args.fspec == 0) {
268 * Process export requests.
270 return (vfs_export(mp, &ump->um_export, &args.export));
274 * Not an update, or updating the name: look up the name
275 * and verify that it refers to a sensible block device.
278 error = nlookup_init(&nd, args.fspec, UIO_USERSPACE, NLC_FOLLOW);
280 error = nlookup(&nd);
282 error = cache_vref(nd.nl_ncp, nd.nl_cred, &devvp);
287 if (!vn_isdisk(devvp, &error)) {
293 * If mount by non-root, then verify that user has necessary
294 * permissions on the device.
296 if (cred->cr_uid != 0) {
298 if ((mp->mnt_flag & MNT_RDONLY) == 0)
299 accessmode |= VWRITE;
300 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
301 if ((error = VOP_ACCESS(devvp, accessmode, cred, td)) != 0) {
305 VOP_UNLOCK(devvp, 0, td);
308 if ((mp->mnt_flag & MNT_UPDATE) == 0) {
309 error = ext2_mountfs(devvp, mp, td);
311 if (devvp != ump->um_devvp)
312 error = EINVAL; /* needs translation */
322 copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size);
323 bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size);
324 bcopy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname, MNAMELEN);
325 copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, &size);
326 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
327 ext2_statfs(mp, &mp->mnt_stat, td);
332 * checks that the data in the descriptor blocks make sense
333 * this is taken from ext2/super.c
335 static int ext2_check_descriptors(struct ext2_sb_info *sb)
339 unsigned long block = sb->s_es->s_first_data_block;
340 struct ext2_group_desc * gdp = NULL;
342 /* ext2_debug ("Checking group descriptors"); */
344 for (i = 0; i < sb->s_groups_count; i++)
346 /* examine next descriptor block */
347 if ((i % EXT2_DESC_PER_BLOCK(sb)) == 0)
348 gdp = (struct ext2_group_desc *)
349 sb->s_group_desc[desc_block++]->b_data;
350 if (gdp->bg_block_bitmap < block ||
351 gdp->bg_block_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb))
353 printf ("ext2_check_descriptors: "
354 "Block bitmap for group %d"
355 " not in group (block %lu)!\n",
356 i, (unsigned long) gdp->bg_block_bitmap);
359 if (gdp->bg_inode_bitmap < block ||
360 gdp->bg_inode_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb))
362 printf ("ext2_check_descriptors: "
363 "Inode bitmap for group %d"
364 " not in group (block %lu)!\n",
365 i, (unsigned long) gdp->bg_inode_bitmap);
368 if (gdp->bg_inode_table < block ||
369 gdp->bg_inode_table + sb->s_itb_per_group >=
370 block + EXT2_BLOCKS_PER_GROUP(sb))
372 printf ("ext2_check_descriptors: "
373 "Inode table for group %d"
374 " not in group (block %lu)!\n",
375 i, (unsigned long) gdp->bg_inode_table);
378 block += EXT2_BLOCKS_PER_GROUP(sb);
385 ext2_check_sb_compat(struct ext2_super_block *es, dev_t dev, int ronly)
387 if (es->s_magic != EXT2_SUPER_MAGIC) {
388 printf("ext2fs: %s: wrong magic number %#x (expected %#x)\n",
389 devtoname(dev), es->s_magic, EXT2_SUPER_MAGIC);
392 if (es->s_rev_level > EXT2_GOOD_OLD_REV) {
393 if (es->s_feature_incompat & ~EXT2_FEATURE_INCOMPAT_SUPP) {
395 "WARNING: mount of %s denied due to unsupported optional features\n",
400 (es->s_feature_ro_compat & ~EXT2_FEATURE_RO_COMPAT_SUPP)) {
402 "WARNING: R/W mount of %s denied due to unsupported optional features\n",
411 * this computes the fields of the ext2_sb_info structure from the
412 * data in the ext2_super_block structure read in
415 compute_sb_data(struct vnode *devvp, struct ext2_super_block *es,
416 struct ext2_sb_info *fs)
420 int logic_sb_block = 1; /* XXX for now */
425 #define V(v) printf(#v"= %d\n", fs->v);
428 fs->s_blocksize = EXT2_MIN_BLOCK_SIZE << es->s_log_block_size;
430 fs->s_bshift = EXT2_MIN_BLOCK_LOG_SIZE + es->s_log_block_size;
432 fs->s_fsbtodb = es->s_log_block_size + 1;
434 fs->s_qbmask = fs->s_blocksize - 1;
436 fs->s_blocksize_bits = EXT2_BLOCK_SIZE_BITS(es);
438 fs->s_frag_size = EXT2_MIN_FRAG_SIZE << es->s_log_frag_size;
441 fs->s_frags_per_block = fs->s_blocksize / fs->s_frag_size;
443 fs->s_blocks_per_group = es->s_blocks_per_group;
444 V(s_blocks_per_group)
445 fs->s_frags_per_group = es->s_frags_per_group;
447 fs->s_inodes_per_group = es->s_inodes_per_group;
448 V(s_inodes_per_group)
449 fs->s_inodes_per_block = fs->s_blocksize / EXT2_INODE_SIZE;
450 V(s_inodes_per_block)
451 fs->s_itb_per_group = fs->s_inodes_per_group /fs->s_inodes_per_block;
453 fs->s_desc_per_block = fs->s_blocksize / sizeof (struct ext2_group_desc);
455 /* s_resuid / s_resgid ? */
456 fs->s_groups_count = (es->s_blocks_count -
457 es->s_first_data_block +
458 EXT2_BLOCKS_PER_GROUP(fs) - 1) /
459 EXT2_BLOCKS_PER_GROUP(fs);
461 db_count = (fs->s_groups_count + EXT2_DESC_PER_BLOCK(fs) - 1) /
462 EXT2_DESC_PER_BLOCK(fs);
463 fs->s_db_per_group = db_count;
466 fs->s_group_desc = bsd_malloc(db_count * sizeof (struct buf *),
469 /* adjust logic_sb_block */
470 if(fs->s_blocksize > SBSIZE)
471 /* Godmar thinks: if the blocksize is greater than 1024, then
472 the superblock is logically part of block zero.
476 for (i = 0; i < db_count; i++) {
477 error = bread(devvp , fsbtodb(fs, logic_sb_block + i + 1),
478 fs->s_blocksize, &fs->s_group_desc[i]);
480 for (j = 0; j < i; j++)
481 brelse(fs->s_group_desc[j]);
482 bsd_free(fs->s_group_desc, M_UFSMNT);
483 printf("EXT2-fs: unable to read group descriptors (%d)\n", error);
486 /* Set the B_LOCKED flag on the buffer, then brelse() it */
487 LCK_BUF(fs->s_group_desc[i])
489 if(!ext2_check_descriptors(fs)) {
490 for (j = 0; j < db_count; j++)
491 ULCK_BUF(fs->s_group_desc[j])
492 bsd_free(fs->s_group_desc, M_UFSMNT);
493 printf("EXT2-fs: (ext2_check_descriptors failure) "
494 "unable to read group descriptors\n");
498 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) {
499 fs->s_inode_bitmap_number[i] = 0;
500 fs->s_inode_bitmap[i] = NULL;
501 fs->s_block_bitmap_number[i] = 0;
502 fs->s_block_bitmap[i] = NULL;
504 fs->s_loaded_inode_bitmaps = 0;
505 fs->s_loaded_block_bitmaps = 0;
510 * Reload all incore data for a filesystem (used after running fsck on
511 * the root filesystem and finding things to fix). The filesystem must
512 * be mounted read-only.
514 * Things to do to update the mount:
515 * 1) invalidate all cached meta-data.
516 * 2) re-read superblock from disk.
517 * 3) re-read summary information from disk.
518 * 4) invalidate all inactive vnodes.
519 * 5) invalidate all cached file data.
520 * 6) re-read inode data for all active vnodes.
522 static int ext2_reload_scan1(struct mount *mp, struct vnode *vp, void *rescan);
523 static int ext2_reload_scan2(struct mount *mp, struct vnode *vp, void *rescan);
531 struct ext2_sb_info *fs;
535 ext2_reload(struct mount *mountp, struct ucred *cred, struct thread *td)
539 struct ext2_super_block * es;
540 struct ext2_sb_info *fs;
542 struct scaninfo scaninfo;
544 if ((mountp->mnt_flag & MNT_RDONLY) == 0)
547 * Step 1: invalidate all cached meta-data.
549 devvp = VFSTOUFS(mountp)->um_devvp;
550 if (vinvalbuf(devvp, 0, td, 0, 0))
551 panic("ext2_reload: dirty1");
553 * Step 2: re-read superblock from disk.
554 * constants have been adjusted for ext2
556 if ((error = bread(devvp, SBLOCK, SBSIZE, &bp)) != 0)
558 es = (struct ext2_super_block *)bp->b_data;
559 if (ext2_check_sb_compat(es, devvp->v_rdev, 0) != 0) {
561 return (EIO); /* XXX needs translation */
563 fs = VFSTOUFS(mountp)->um_e2fs;
564 bcopy(bp->b_data, fs->s_es, sizeof(struct ext2_super_block));
566 if((error = compute_sb_data(devvp, es, fs)) != 0) {
571 if (fs->fs_sbsize < SBSIZE)
572 bp->b_flags |= B_INVAL;
578 scaninfo.devvp = devvp;
580 while (error == 0 && scaninfo.rescan) {
582 error = vmntvnodescan(mountp, VMSC_GETVX, ext2_reload_scan1,
583 ext2_reload_scan2, &scaninfo);
589 ext2_reload_scan1(struct mount *mp, struct vnode *vp, void *data)
591 /*struct scaninfo *info = data;*/
597 ext2_reload_scan2(struct mount *mp, struct vnode *vp, void *data)
599 struct scaninfo *info = data;
607 if (vrecycle(vp, curthread))
611 * Step 5: invalidate all cached file data.
613 if (vinvalbuf(vp, 0, info->td, 0, 0))
614 panic("ext2_reload: dirty2");
616 * Step 6: re-read inode data for all active vnodes.
619 error = bread(info->devvp, fsbtodb(info->fs, ino_to_fsba(info->fs, ip->i_number)),
620 (int)info->fs->s_blocksize, &bp);
623 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data +
624 EXT2_INODE_SIZE * ino_to_fsbo(info->fs, ip->i_number)),
631 * Common code for mount and mountroot
634 ext2_mountfs(struct vnode *devvp, struct mount *mp, struct thread *td)
636 struct ufsmount *ump;
638 struct ext2_sb_info *fs;
639 struct ext2_super_block * es;
641 struct partinfo dpart;
647 * Disallow multiple mounts of the same device.
648 * Disallow mounting of a device that is currently in use
649 * (except for root, which might share swap device for miniroot).
650 * Flush out any old buffers remaining from a previous use.
652 if ((error = vfs_mountedon(devvp)) != 0)
654 if (count_udev(devvp->v_udev) > 0)
656 if ((error = vinvalbuf(devvp, V_SAVE, td, 0, 0)) != 0)
659 /* turn on this to force it to be read-only */
660 mp->mnt_flag |= MNT_RDONLY;
663 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
664 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
665 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, NULL, td);
666 VOP_UNLOCK(devvp, 0, td);
670 if (dev->si_iosize_max != 0)
671 mp->mnt_iosize_max = dev->si_iosize_max;
672 if (mp->mnt_iosize_max > MAXPHYS)
673 mp->mnt_iosize_max = MAXPHYS;
674 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, td) != 0)
678 size = dpart.disklab->d_secsize;
683 if ((error = bread(devvp, SBLOCK, SBSIZE, &bp)) != 0)
685 es = (struct ext2_super_block *)bp->b_data;
686 if (ext2_check_sb_compat(es, dev, ronly) != 0) {
687 error = EINVAL; /* XXX needs translation */
690 if ((es->s_state & EXT2_VALID_FS) == 0 ||
691 (es->s_state & EXT2_ERROR_FS)) {
692 if (ronly || (mp->mnt_flag & MNT_FORCE)) {
694 "WARNING: Filesystem was not properly dismounted\n");
697 "WARNING: R/W mount denied. Filesystem is not clean - run fsck\n");
702 ump = bsd_malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
703 bzero((caddr_t)ump, sizeof *ump);
704 ump->um_malloctype = M_EXT2NODE;
705 ump->um_blkatoff = ext2_blkatoff;
706 ump->um_truncate = ext2_truncate;
707 ump->um_update = ext2_update;
708 ump->um_valloc = ext2_valloc;
709 ump->um_vfree = ext2_vfree;
710 /* I don't know whether this is the right strategy. Note that
711 we dynamically allocate both a ext2_sb_info and a ext2_super_block
712 while Linux keeps the super block in a locked buffer
714 ump->um_e2fs = bsd_malloc(sizeof(struct ext2_sb_info),
716 ump->um_e2fs->s_es = bsd_malloc(sizeof(struct ext2_super_block),
718 bcopy(es, ump->um_e2fs->s_es, (u_int)sizeof(struct ext2_super_block));
719 if ((error = compute_sb_data(devvp, ump->um_e2fs->s_es, ump->um_e2fs)))
722 * We don't free the group descriptors allocated by compute_sb_data()
723 * until ext2_unmount(). This is OK since the mount will succeed.
728 fs->s_rd_only = ronly; /* ronly is set according to mnt_flags */
729 /* if the fs is not mounted read-only, make sure the super block is
730 always written back on a sync()
732 fs->s_wasvalid = fs->s_es->s_state & EXT2_VALID_FS ? 1 : 0;
734 fs->s_dirt = 1; /* mark it modified */
735 fs->s_es->s_state &= ~EXT2_VALID_FS; /* set fs invalid */
737 mp->mnt_data = (qaddr_t)ump;
738 mp->mnt_stat.f_fsid.val[0] = dev2udev(dev);
739 mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
740 mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN;
741 mp->mnt_flag |= MNT_LOCAL;
744 ump->um_devvp = devvp;
745 /* setting those two parameters allows us to use
746 ufs_bmap w/o changse !
748 ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs);
749 ump->um_bptrtodb = fs->s_es->s_log_block_size + 1;
750 ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs);
751 for (i = 0; i < MAXQUOTAS; i++)
752 ump->um_quotas[i] = NULLVP;
753 dev->si_mountpoint = mp;
755 vfs_add_vnodeops(&mp->mnt_vn_ops, ext2_vnodeop_entries);
756 vfs_add_vnodeops(&mp->mnt_vn_spec_ops, ext2_specop_entries);
757 vfs_add_vnodeops(&mp->mnt_vn_fifo_ops, ext2_fifoop_entries);
760 ext2_sbupdate(ump, MNT_WAIT);
765 (void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, td);
767 bsd_free(ump->um_e2fs->s_es, M_UFSMNT);
768 bsd_free(ump->um_e2fs, M_UFSMNT);
769 bsd_free(ump, M_UFSMNT);
770 mp->mnt_data = (qaddr_t)0;
776 * unmount system call
779 ext2_unmount(struct mount *mp, int mntflags, struct thread *td)
781 struct ufsmount *ump;
782 struct ext2_sb_info *fs;
783 int error, flags, ronly, i;
786 if (mntflags & MNT_FORCE) {
787 if (mp->mnt_flag & MNT_ROOTFS)
791 if ((error = ext2_flushfiles(mp, flags, td)) != 0)
795 ronly = fs->s_rd_only;
798 fs->s_es->s_state |= EXT2_VALID_FS;
799 ext2_sbupdate(ump, MNT_WAIT);
802 /* release buffers containing group descriptors */
803 for(i = 0; i < fs->s_db_per_group; i++)
804 ULCK_BUF(fs->s_group_desc[i])
805 bsd_free(fs->s_group_desc, M_UFSMNT);
807 /* release cached inode/block bitmaps */
808 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
809 if (fs->s_inode_bitmap[i])
810 ULCK_BUF(fs->s_inode_bitmap[i])
812 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
813 if (fs->s_block_bitmap[i])
814 ULCK_BUF(fs->s_block_bitmap[i])
816 ump->um_devvp->v_rdev->si_mountpoint = NULL;
817 error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD|FWRITE, td);
818 vrele(ump->um_devvp);
819 bsd_free(fs->s_es, M_UFSMNT);
820 bsd_free(fs, M_UFSMNT);
821 bsd_free(ump, M_UFSMNT);
822 mp->mnt_data = (qaddr_t)0;
823 mp->mnt_flag &= ~MNT_LOCAL;
828 * Flush out all the files in a filesystem.
831 ext2_flushfiles(struct mount *mp, int flags, struct thread *td)
833 struct ufsmount *ump;
841 if (mp->mnt_flag & MNT_QUOTA) {
842 if ((error = vflush(mp, 0, SKIPSYSTEM|flags)) != 0)
844 for (i = 0; i < MAXQUOTAS; i++) {
845 if (ump->um_quotas[i] == NULLVP)
850 * Here we fall through to vflush again to ensure
851 * that we have gotten rid of all the system vnodes.
855 error = vflush(mp, 0, flags);
860 * Get file system statistics.
861 * taken from ext2/super.c ext2_statfs
864 ext2_statfs(struct mount *mp, struct statfs *sbp, struct thread *td)
866 unsigned long overhead;
867 struct ufsmount *ump;
868 struct ext2_sb_info *fs;
869 struct ext2_super_block *es;
876 if (es->s_magic != EXT2_SUPER_MAGIC)
877 panic("ext2_statfs - magic number spoiled");
880 * Compute the overhead (FS structures)
882 if (es->s_feature_ro_compat & EXT2_FEATURE_RO_COMPAT_SPARSE_SUPER) {
884 for (i = 0 ; i < fs->s_groups_count; i++)
885 if (ext2_group_sparse(i))
888 nsb = fs->s_groups_count;
889 overhead = es->s_first_data_block +
890 /* Superblocks and block group descriptors: */
891 nsb * (1 + fs->s_db_per_group) +
892 /* Inode bitmap, block bitmap, and inode table: */
893 fs->s_groups_count * (1 + 1 + fs->s_itb_per_group);
895 sbp->f_bsize = EXT2_FRAG_SIZE(fs);
896 sbp->f_iosize = EXT2_BLOCK_SIZE(fs);
897 sbp->f_blocks = es->s_blocks_count - overhead;
898 sbp->f_bfree = es->s_free_blocks_count;
899 sbp->f_bavail = sbp->f_bfree - es->s_r_blocks_count;
900 sbp->f_files = es->s_inodes_count;
901 sbp->f_ffree = es->s_free_inodes_count;
902 if (sbp != &mp->mnt_stat) {
903 sbp->f_type = mp->mnt_vfc->vfc_typenum;
904 bcopy((caddr_t)mp->mnt_stat.f_mntonname,
905 (caddr_t)&sbp->f_mntonname[0], MNAMELEN);
906 bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
907 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
913 * Go through the disk queues to initiate sandbagged IO;
914 * go through the inodes to write those that have been modified;
915 * initiate the writing of the super block if it has been modified.
917 * Note: we are always called with the filesystem marked `MPBUSY'.
920 static int ext2_sync_scan(struct mount *mp, struct vnode *vp, void *data);
923 ext2_sync(struct mount *mp, int waitfor, struct thread *td)
925 struct ufsmount *ump = VFSTOUFS(mp);
926 struct ext2_sb_info *fs;
927 struct scaninfo scaninfo;
931 if (fs->s_dirt != 0 && fs->s_rd_only != 0) { /* XXX */
932 printf("fs = %s\n", fs->fs_fsmnt);
933 panic("ext2_sync: rofs mod");
937 * Write back each (modified) inode.
939 scaninfo.allerror = 0;
941 scaninfo.waitfor = waitfor;
943 while (scaninfo.rescan) {
945 vmntvnodescan(mp, VMSC_GETVP|VMSC_NOWAIT,
946 NULL, ext2_sync_scan, &scaninfo);
950 * Force stale file system control information to be flushed.
952 if (waitfor != MNT_LAZY) {
953 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY, td);
954 if ((error = VOP_FSYNC(ump->um_devvp, waitfor, td)) != 0)
955 scaninfo.allerror = error;
956 VOP_UNLOCK(ump->um_devvp, 0, td);
962 * Write back modified superblock.
964 if (fs->s_dirt != 0) {
966 fs->s_es->s_wtime = time_second;
967 if ((error = ext2_sbupdate(ump, waitfor)) != 0)
968 scaninfo.allerror = error;
970 return (scaninfo.allerror);
974 ext2_sync_scan(struct mount *mp, struct vnode *vp, void *data)
976 struct scaninfo *info = data;
981 if (vp->v_type == VNON ||
983 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
984 (TAILQ_EMPTY(&vp->v_dirtyblkhd) || info->waitfor == MNT_LAZY))) {
987 if ((error = VOP_FSYNC(vp, info->waitfor, info->td)) != 0)
988 info->allerror = error;
993 * Look up a EXT2FS dinode number to find its incore vnode, otherwise read it
994 * in from disk. If it is in core, wait for the lock bit to clear, then
995 * return the inode locked. Detection and handling of mount points must be
996 * done by the calling routine.
999 ext2_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
1001 struct ext2_sb_info *fs;
1003 struct ufsmount *ump;
1013 if ((*vpp = ufs_ihashget(dev, ino)) != NULL)
1017 * Lock out the creation of new entries in the FFS hash table in
1018 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate
1021 if (ext2fs_inode_hash_lock) {
1022 while (ext2fs_inode_hash_lock) {
1023 ext2fs_inode_hash_lock = -1;
1024 tsleep(&ext2fs_inode_hash_lock, 0, "e2vget", 0);
1028 ext2fs_inode_hash_lock = 1;
1031 * If this MALLOC() is performed after the getnewvnode()
1032 * it might block, leaving a vnode with a NULL v_data to be
1033 * found by ext2_sync() if a sync happens to fire right then,
1034 * which will cause a panic because ext2_sync() blindly
1035 * dereferences vp->v_data (as well it should).
1037 MALLOC(ip, struct inode *, sizeof(struct inode), M_EXT2NODE, M_WAITOK);
1039 /* Allocate a new vnode/inode. */
1040 if ((error = getnewvnode(VT_UFS, mp, mp->mnt_vn_ops, &vp, 0, 0)) != 0) {
1041 if (ext2fs_inode_hash_lock < 0)
1042 wakeup(&ext2fs_inode_hash_lock);
1043 ext2fs_inode_hash_lock = 0;
1045 FREE(ip, M_EXT2NODE);
1048 bzero((caddr_t)ip, sizeof(struct inode));
1051 ip->i_e2fs = fs = ump->um_e2fs;
1055 for (i = 0; i < MAXQUOTAS; i++)
1056 ip->i_dquot[i] = NODQUOT;
1059 * Put it onto its hash chain. Since our vnode is locked, other
1060 * requests for this inode will block if they arrive while we are
1061 * sleeping waiting for old data structures to be purged or for the
1062 * contents of the disk portion of this inode to be read.
1066 if (ext2fs_inode_hash_lock < 0)
1067 wakeup(&ext2fs_inode_hash_lock);
1068 ext2fs_inode_hash_lock = 0;
1070 /* Read in the disk contents for the inode, copy into the inode. */
1072 printf("ext2_vget(%d) dbn= %d ", ino, fsbtodb(fs, ino_to_fsba(fs, ino)));
1074 if ((error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
1075 (int)fs->s_blocksize, &bp)) != 0) {
1077 * The inode does not contain anything useful, so it would
1078 * be misleading to leave it on its hash chain. With mode
1079 * still zero, it will be unlinked and returned to the free
1087 /* convert ext2 inode to dinode */
1088 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data + EXT2_INODE_SIZE *
1089 ino_to_fsbo(fs, ino)), &ip->i_din);
1090 ip->i_block_group = ino_to_cg(fs, ino);
1091 ip->i_next_alloc_block = 0;
1092 ip->i_next_alloc_goal = 0;
1093 ip->i_prealloc_count = 0;
1094 ip->i_prealloc_block = 0;
1095 /* now we want to make sure that block pointers for unused
1096 blocks are zeroed out - ext2_balloc depends on this
1097 although for regular files and directories only
1099 if(S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode)) {
1100 used_blocks = (ip->i_size+fs->s_blocksize-1) / fs->s_blocksize;
1101 for(i = used_blocks; i < EXT2_NDIR_BLOCKS; i++)
1105 ext2_print_inode(ip);
1110 * Initialize the vnode from the inode, check for aliases.
1111 * Note that the underlying vnode may have changed.
1113 if ((error = ufs_vinit(mp, &vp)) != 0) {
1119 * Finish inode initialization now that aliasing has been resolved.
1121 ip->i_devvp = ump->um_devvp;
1124 * Set up a generation number for this inode if it does not
1125 * already have one. This should only happen on old filesystems.
1127 if (ip->i_gen == 0) {
1128 ip->i_gen = random() / 2 + 1;
1129 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1130 ip->i_flag |= IN_MODIFIED;
1133 * Return the locked and refd vnode.
1140 * File handle to vnode
1142 * Have to be really careful about stale file handles:
1143 * - check that the inode number is valid
1144 * - call ext2_vget() to get the locked inode
1145 * - check for an unallocated inode (i_mode == 0)
1146 * - check that the given client host has export rights and return
1147 * those rights via. exflagsp and credanonp
1150 ext2_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
1153 struct ext2_sb_info *fs;
1155 ufhp = (struct ufid *)fhp;
1156 fs = VFSTOUFS(mp)->um_e2fs;
1157 if (ufhp->ufid_ino < ROOTINO ||
1158 ufhp->ufid_ino > fs->s_groups_count * fs->s_es->s_inodes_per_group)
1160 return (ufs_fhtovp(mp, ufhp, vpp));
1164 * Vnode pointer to File handle
1168 ext2_vptofh(struct vnode *vp, struct fid *fhp)
1174 ufhp = (struct ufid *)fhp;
1175 ufhp->ufid_len = sizeof(struct ufid);
1176 ufhp->ufid_ino = ip->i_number;
1177 ufhp->ufid_gen = ip->i_gen;
1182 * Write a superblock and associated information back to disk.
1185 ext2_sbupdate(struct ufsmount *mp, int waitfor)
1187 struct ext2_sb_info *fs = mp->um_e2fs;
1188 struct ext2_super_block *es = fs->s_es;
1192 printf("\nupdating superblock, waitfor=%s\n", waitfor == MNT_WAIT ? "yes":"no");
1194 bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0);
1195 bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2_super_block));
1196 if (waitfor == MNT_WAIT)
1202 * The buffers for group descriptors, inode bitmaps and block bitmaps
1203 * are not busy at this point and are (hopefully) written by the
1204 * usual sync mechanism. No need to write them here