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 1.29 2005/06/06 15:35:06 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>
60 #include <sys/thread2.h>
62 #include <vfs/ufs/quota.h>
63 #include <vfs/ufs/ufsmount.h>
64 #include <vfs/ufs/inode.h>
65 #include <vfs/ufs/ufs_extern.h>
67 #include <vm/vm_zone.h>
70 #include "ext2_extern.h"
72 #include "ext2_fs_sb.h"
74 extern struct vnodeopv_entry_desc ext2_vnodeop_entries[];
75 extern struct vnodeopv_entry_desc ext2_specop_entries[];
76 extern struct vnodeopv_entry_desc ext2_fifoop_entries[];
78 static int ext2_fhtovp (struct mount *, struct fid *, struct vnode **);
79 static int ext2_flushfiles (struct mount *mp, int flags, struct thread *td);
80 static int ext2_mount (struct mount *, char *, caddr_t, struct thread *);
81 static int ext2_mountfs (struct vnode *, struct mount *, struct thread *);
82 static int ext2_reload (struct mount *mountp, struct ucred *cred,
84 static int ext2_sbupdate (struct ufsmount *, int);
85 static int ext2_statfs (struct mount *, struct statfs *, struct thread *);
86 static int ext2_sync (struct mount *, int, struct thread *);
87 static int ext2_unmount (struct mount *, int, struct thread *);
88 static int ext2_vget (struct mount *, ino_t, struct vnode **);
89 static int ext2_vptofh (struct vnode *, struct fid *);
91 static MALLOC_DEFINE(M_EXT2NODE, "EXT2 node", "EXT2 vnode private part");
93 static struct vfsops ext2fs_vfsops = {
95 ufs_start, /* empty function */
97 ufs_root, /* root inode via vget */
98 ufs_quotactl, /* does operations associated with quotas */
110 VFS_SET(ext2fs_vfsops, ext2fs, 0);
111 #define bsd_malloc malloc
112 #define bsd_free free
114 static int ext2fs_inode_hash_lock;
116 static int ext2_check_sb_compat (struct ext2_super_block *es,
117 dev_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);
128 ext2_mount(struct mount *mp, char *path,
129 caddr_t data, /* this is actually a (struct ufs_args *) */
133 struct ufs_args args;
134 struct ufsmount *ump = 0;
135 struct ext2_sb_info *fs;
140 struct nlookupdata nd;
142 if ((error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args))) != 0)
145 cred = td->td_proc->p_ucred;
147 * If updating, check whether changing from read-only to
148 * read/write; if there is no device name, that's all we do.
150 if (mp->mnt_flag & MNT_UPDATE) {
154 if (fs->s_rd_only == 0 && (mp->mnt_flag & MNT_RDONLY)) {
156 if (mp->mnt_flag & MNT_FORCE)
158 if (vfs_busy(mp, LK_NOWAIT, td))
160 error = ext2_flushfiles(mp, flags, td);
162 if (!error && fs->s_wasvalid) {
163 fs->s_es->s_state |= EXT2_VALID_FS;
164 ext2_sbupdate(ump, MNT_WAIT);
168 if (!error && (mp->mnt_flag & MNT_RELOAD))
169 error = ext2_reload(mp, proc0.p_ucred, td);
172 devvp = ump->um_devvp;
173 if (ext2_check_sb_compat(fs->s_es, devvp->v_rdev,
174 (mp->mnt_kern_flag & MNTK_WANTRDWR) == 0) != 0)
176 if (fs->s_rd_only && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
178 * If upgrade to read-write by non-root, then verify
179 * that user has necessary permissions on the device.
181 if (cred->cr_uid != 0) {
182 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
183 if ((error = VOP_ACCESS(devvp, VREAD | VWRITE,
185 VOP_UNLOCK(devvp, 0, td);
188 VOP_UNLOCK(devvp, 0, td);
191 if ((fs->s_es->s_state & EXT2_VALID_FS) == 0 ||
192 (fs->s_es->s_state & EXT2_ERROR_FS)) {
193 if (mp->mnt_flag & MNT_FORCE) {
195 "WARNING: %s was not properly dismounted\n",
199 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
204 fs->s_es->s_state &= ~EXT2_VALID_FS;
205 ext2_sbupdate(ump, MNT_WAIT);
208 if (args.fspec == 0) {
210 * Process export requests.
212 return (vfs_export(mp, &ump->um_export, &args.export));
216 * Not an update, or updating the name: look up the name
217 * and verify that it refers to a sensible block device.
220 error = nlookup_init(&nd, args.fspec, UIO_USERSPACE, NLC_FOLLOW);
222 error = nlookup(&nd);
224 error = cache_vref(nd.nl_ncp, nd.nl_cred, &devvp);
229 if (!vn_isdisk(devvp, &error)) {
235 * If mount by non-root, then verify that user has necessary
236 * permissions on the device.
238 if (cred->cr_uid != 0) {
240 if ((mp->mnt_flag & MNT_RDONLY) == 0)
241 accessmode |= VWRITE;
242 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
243 if ((error = VOP_ACCESS(devvp, accessmode, cred, td)) != 0) {
247 VOP_UNLOCK(devvp, 0, td);
250 if ((mp->mnt_flag & MNT_UPDATE) == 0) {
251 error = ext2_mountfs(devvp, mp, td);
253 if (devvp != ump->um_devvp)
254 error = EINVAL; /* needs translation */
264 copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size);
265 bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size);
266 copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, &size);
267 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
268 ext2_statfs(mp, &mp->mnt_stat, td);
273 * checks that the data in the descriptor blocks make sense
274 * this is taken from ext2/super.c
276 static int ext2_check_descriptors(struct ext2_sb_info *sb)
280 unsigned long block = sb->s_es->s_first_data_block;
281 struct ext2_group_desc * gdp = NULL;
283 /* ext2_debug ("Checking group descriptors"); */
285 for (i = 0; i < sb->s_groups_count; i++)
287 /* examine next descriptor block */
288 if ((i % EXT2_DESC_PER_BLOCK(sb)) == 0)
289 gdp = (struct ext2_group_desc *)
290 sb->s_group_desc[desc_block++]->b_data;
291 if (gdp->bg_block_bitmap < block ||
292 gdp->bg_block_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb))
294 printf ("ext2_check_descriptors: "
295 "Block bitmap for group %d"
296 " not in group (block %lu)!\n",
297 i, (unsigned long) gdp->bg_block_bitmap);
300 if (gdp->bg_inode_bitmap < block ||
301 gdp->bg_inode_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb))
303 printf ("ext2_check_descriptors: "
304 "Inode bitmap for group %d"
305 " not in group (block %lu)!\n",
306 i, (unsigned long) gdp->bg_inode_bitmap);
309 if (gdp->bg_inode_table < block ||
310 gdp->bg_inode_table + sb->s_itb_per_group >=
311 block + EXT2_BLOCKS_PER_GROUP(sb))
313 printf ("ext2_check_descriptors: "
314 "Inode table for group %d"
315 " not in group (block %lu)!\n",
316 i, (unsigned long) gdp->bg_inode_table);
319 block += EXT2_BLOCKS_PER_GROUP(sb);
326 ext2_check_sb_compat(struct ext2_super_block *es, dev_t dev, int ronly)
328 if (es->s_magic != EXT2_SUPER_MAGIC) {
329 printf("ext2fs: %s: wrong magic number %#x (expected %#x)\n",
330 devtoname(dev), es->s_magic, EXT2_SUPER_MAGIC);
333 if (es->s_rev_level > EXT2_GOOD_OLD_REV) {
334 if (es->s_feature_incompat & ~EXT2_FEATURE_INCOMPAT_SUPP) {
336 "WARNING: mount of %s denied due to unsupported optional features\n",
341 (es->s_feature_ro_compat & ~EXT2_FEATURE_RO_COMPAT_SUPP)) {
343 "WARNING: R/W mount of %s denied due to unsupported optional features\n",
352 * this computes the fields of the ext2_sb_info structure from the
353 * data in the ext2_super_block structure read in
356 compute_sb_data(struct vnode *devvp, struct ext2_super_block *es,
357 struct ext2_sb_info *fs)
361 int logic_sb_block = 1; /* XXX for now */
366 #define V(v) printf(#v"= %d\n", fs->v);
369 fs->s_blocksize = EXT2_MIN_BLOCK_SIZE << es->s_log_block_size;
371 fs->s_bshift = EXT2_MIN_BLOCK_LOG_SIZE + es->s_log_block_size;
373 fs->s_fsbtodb = es->s_log_block_size + 1;
375 fs->s_qbmask = fs->s_blocksize - 1;
377 fs->s_blocksize_bits = EXT2_BLOCK_SIZE_BITS(es);
379 fs->s_frag_size = EXT2_MIN_FRAG_SIZE << es->s_log_frag_size;
382 fs->s_frags_per_block = fs->s_blocksize / fs->s_frag_size;
384 fs->s_blocks_per_group = es->s_blocks_per_group;
385 V(s_blocks_per_group)
386 fs->s_frags_per_group = es->s_frags_per_group;
388 fs->s_inodes_per_group = es->s_inodes_per_group;
389 V(s_inodes_per_group)
390 fs->s_inodes_per_block = fs->s_blocksize / EXT2_INODE_SIZE;
391 V(s_inodes_per_block)
392 fs->s_itb_per_group = fs->s_inodes_per_group /fs->s_inodes_per_block;
394 fs->s_desc_per_block = fs->s_blocksize / sizeof (struct ext2_group_desc);
396 /* s_resuid / s_resgid ? */
397 fs->s_groups_count = (es->s_blocks_count -
398 es->s_first_data_block +
399 EXT2_BLOCKS_PER_GROUP(fs) - 1) /
400 EXT2_BLOCKS_PER_GROUP(fs);
402 db_count = (fs->s_groups_count + EXT2_DESC_PER_BLOCK(fs) - 1) /
403 EXT2_DESC_PER_BLOCK(fs);
404 fs->s_db_per_group = db_count;
407 fs->s_group_desc = bsd_malloc(db_count * sizeof (struct buf *),
410 /* adjust logic_sb_block */
411 if(fs->s_blocksize > SBSIZE)
412 /* Godmar thinks: if the blocksize is greater than 1024, then
413 the superblock is logically part of block zero.
417 for (i = 0; i < db_count; i++) {
418 error = bread(devvp , fsbtodb(fs, logic_sb_block + i + 1),
419 fs->s_blocksize, &fs->s_group_desc[i]);
421 for (j = 0; j < i; j++)
422 brelse(fs->s_group_desc[j]);
423 bsd_free(fs->s_group_desc, M_UFSMNT);
424 printf("EXT2-fs: unable to read group descriptors (%d)\n", error);
427 /* Set the B_LOCKED flag on the buffer, then brelse() it */
428 LCK_BUF(fs->s_group_desc[i])
430 if(!ext2_check_descriptors(fs)) {
431 for (j = 0; j < db_count; j++)
432 ULCK_BUF(fs->s_group_desc[j])
433 bsd_free(fs->s_group_desc, M_UFSMNT);
434 printf("EXT2-fs: (ext2_check_descriptors failure) "
435 "unable to read group descriptors\n");
439 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) {
440 fs->s_inode_bitmap_number[i] = 0;
441 fs->s_inode_bitmap[i] = NULL;
442 fs->s_block_bitmap_number[i] = 0;
443 fs->s_block_bitmap[i] = NULL;
445 fs->s_loaded_inode_bitmaps = 0;
446 fs->s_loaded_block_bitmaps = 0;
451 * Reload all incore data for a filesystem (used after running fsck on
452 * the root filesystem and finding things to fix). The filesystem must
453 * be mounted read-only.
455 * Things to do to update the mount:
456 * 1) invalidate all cached meta-data.
457 * 2) re-read superblock from disk.
458 * 3) re-read summary information from disk.
459 * 4) invalidate all inactive vnodes.
460 * 5) invalidate all cached file data.
461 * 6) re-read inode data for all active vnodes.
463 static int ext2_reload_scan1(struct mount *mp, struct vnode *vp, void *rescan);
464 static int ext2_reload_scan2(struct mount *mp, struct vnode *vp, void *rescan);
472 struct ext2_sb_info *fs;
476 ext2_reload(struct mount *mountp, struct ucred *cred, struct thread *td)
480 struct ext2_super_block * es;
481 struct ext2_sb_info *fs;
483 struct scaninfo scaninfo;
485 if ((mountp->mnt_flag & MNT_RDONLY) == 0)
488 * Step 1: invalidate all cached meta-data.
490 devvp = VFSTOUFS(mountp)->um_devvp;
491 if (vinvalbuf(devvp, 0, td, 0, 0))
492 panic("ext2_reload: dirty1");
494 * Step 2: re-read superblock from disk.
495 * constants have been adjusted for ext2
497 if ((error = bread(devvp, SBLOCK, SBSIZE, &bp)) != 0)
499 es = (struct ext2_super_block *)bp->b_data;
500 if (ext2_check_sb_compat(es, devvp->v_rdev, 0) != 0) {
502 return (EIO); /* XXX needs translation */
504 fs = VFSTOUFS(mountp)->um_e2fs;
505 bcopy(bp->b_data, fs->s_es, sizeof(struct ext2_super_block));
507 if((error = compute_sb_data(devvp, es, fs)) != 0) {
512 if (fs->fs_sbsize < SBSIZE)
513 bp->b_flags |= B_INVAL;
519 scaninfo.devvp = devvp;
521 while (error == 0 && scaninfo.rescan) {
523 error = vmntvnodescan(mountp, VMSC_GETVX, ext2_reload_scan1,
524 ext2_reload_scan2, &scaninfo);
530 ext2_reload_scan1(struct mount *mp, struct vnode *vp, void *data)
532 /*struct scaninfo *info = data;*/
538 ext2_reload_scan2(struct mount *mp, struct vnode *vp, void *data)
540 struct scaninfo *info = data;
548 if (vrecycle(vp, curthread))
552 * Step 5: invalidate all cached file data.
554 if (vinvalbuf(vp, 0, info->td, 0, 0))
555 panic("ext2_reload: dirty2");
557 * Step 6: re-read inode data for all active vnodes.
560 error = bread(info->devvp, fsbtodb(info->fs, ino_to_fsba(info->fs, ip->i_number)),
561 (int)info->fs->s_blocksize, &bp);
564 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data +
565 EXT2_INODE_SIZE * ino_to_fsbo(info->fs, ip->i_number)),
572 * Common code for mount and mountroot
575 ext2_mountfs(struct vnode *devvp, struct mount *mp, struct thread *td)
577 struct ufsmount *ump;
579 struct ext2_sb_info *fs;
580 struct ext2_super_block * es;
582 struct partinfo dpart;
588 * Disallow multiple mounts of the same device.
589 * Disallow mounting of a device that is currently in use
590 * (except for root, which might share swap device for miniroot).
591 * Flush out any old buffers remaining from a previous use.
593 if ((error = vfs_mountedon(devvp)) != 0)
595 if (count_udev(devvp->v_udev) > 0)
597 if ((error = vinvalbuf(devvp, V_SAVE, td, 0, 0)) != 0)
600 /* turn on this to force it to be read-only */
601 mp->mnt_flag |= MNT_RDONLY;
604 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
605 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
606 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, NULL, td);
607 VOP_UNLOCK(devvp, 0, td);
611 if (dev->si_iosize_max != 0)
612 mp->mnt_iosize_max = dev->si_iosize_max;
613 if (mp->mnt_iosize_max > MAXPHYS)
614 mp->mnt_iosize_max = MAXPHYS;
615 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, td) != 0)
619 size = dpart.disklab->d_secsize;
624 if ((error = bread(devvp, SBLOCK, SBSIZE, &bp)) != 0)
626 es = (struct ext2_super_block *)bp->b_data;
627 if (ext2_check_sb_compat(es, dev, ronly) != 0) {
628 error = EINVAL; /* XXX needs translation */
631 if ((es->s_state & EXT2_VALID_FS) == 0 ||
632 (es->s_state & EXT2_ERROR_FS)) {
633 if (ronly || (mp->mnt_flag & MNT_FORCE)) {
635 "WARNING: Filesystem was not properly dismounted\n");
638 "WARNING: R/W mount denied. Filesystem is not clean - run fsck\n");
643 ump = bsd_malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
644 bzero((caddr_t)ump, sizeof *ump);
645 ump->um_malloctype = M_EXT2NODE;
646 ump->um_blkatoff = ext2_blkatoff;
647 ump->um_truncate = ext2_truncate;
648 ump->um_update = ext2_update;
649 ump->um_valloc = ext2_valloc;
650 ump->um_vfree = ext2_vfree;
651 /* I don't know whether this is the right strategy. Note that
652 we dynamically allocate both a ext2_sb_info and a ext2_super_block
653 while Linux keeps the super block in a locked buffer
655 ump->um_e2fs = bsd_malloc(sizeof(struct ext2_sb_info),
657 ump->um_e2fs->s_es = bsd_malloc(sizeof(struct ext2_super_block),
659 bcopy(es, ump->um_e2fs->s_es, (u_int)sizeof(struct ext2_super_block));
660 if ((error = compute_sb_data(devvp, ump->um_e2fs->s_es, ump->um_e2fs)))
663 * We don't free the group descriptors allocated by compute_sb_data()
664 * until ext2_unmount(). This is OK since the mount will succeed.
669 fs->s_rd_only = ronly; /* ronly is set according to mnt_flags */
670 /* if the fs is not mounted read-only, make sure the super block is
671 always written back on a sync()
673 fs->s_wasvalid = fs->s_es->s_state & EXT2_VALID_FS ? 1 : 0;
675 fs->s_dirt = 1; /* mark it modified */
676 fs->s_es->s_state &= ~EXT2_VALID_FS; /* set fs invalid */
678 mp->mnt_data = (qaddr_t)ump;
679 mp->mnt_stat.f_fsid.val[0] = dev2udev(dev);
680 mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
681 mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN;
682 mp->mnt_flag |= MNT_LOCAL;
685 ump->um_devvp = devvp;
686 /* setting those two parameters allows us to use
687 ufs_bmap w/o changse !
689 ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs);
690 ump->um_bptrtodb = fs->s_es->s_log_block_size + 1;
691 ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs);
692 for (i = 0; i < MAXQUOTAS; i++)
693 ump->um_quotas[i] = NULLVP;
694 dev->si_mountpoint = mp;
696 vfs_add_vnodeops(mp, &mp->mnt_vn_norm_ops, ext2_vnodeop_entries);
697 vfs_add_vnodeops(mp, &mp->mnt_vn_spec_ops, ext2_specop_entries);
698 vfs_add_vnodeops(mp, &mp->mnt_vn_fifo_ops, ext2_fifoop_entries);
701 ext2_sbupdate(ump, MNT_WAIT);
706 (void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, td);
708 bsd_free(ump->um_e2fs->s_es, M_UFSMNT);
709 bsd_free(ump->um_e2fs, M_UFSMNT);
710 bsd_free(ump, M_UFSMNT);
711 mp->mnt_data = (qaddr_t)0;
717 * unmount system call
720 ext2_unmount(struct mount *mp, int mntflags, struct thread *td)
722 struct ufsmount *ump;
723 struct ext2_sb_info *fs;
724 int error, flags, ronly, i;
727 if (mntflags & MNT_FORCE) {
728 if (mp->mnt_flag & MNT_ROOTFS)
732 if ((error = ext2_flushfiles(mp, flags, td)) != 0)
736 ronly = fs->s_rd_only;
739 fs->s_es->s_state |= EXT2_VALID_FS;
740 ext2_sbupdate(ump, MNT_WAIT);
743 /* release buffers containing group descriptors */
744 for(i = 0; i < fs->s_db_per_group; i++)
745 ULCK_BUF(fs->s_group_desc[i])
746 bsd_free(fs->s_group_desc, M_UFSMNT);
748 /* release cached inode/block bitmaps */
749 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
750 if (fs->s_inode_bitmap[i])
751 ULCK_BUF(fs->s_inode_bitmap[i])
753 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
754 if (fs->s_block_bitmap[i])
755 ULCK_BUF(fs->s_block_bitmap[i])
757 ump->um_devvp->v_rdev->si_mountpoint = NULL;
758 error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD|FWRITE, td);
759 vrele(ump->um_devvp);
760 bsd_free(fs->s_es, M_UFSMNT);
761 bsd_free(fs, M_UFSMNT);
762 bsd_free(ump, M_UFSMNT);
763 mp->mnt_data = (qaddr_t)0;
764 mp->mnt_flag &= ~MNT_LOCAL;
769 * Flush out all the files in a filesystem.
772 ext2_flushfiles(struct mount *mp, int flags, struct thread *td)
774 struct ufsmount *ump;
782 if (mp->mnt_flag & MNT_QUOTA) {
783 if ((error = vflush(mp, 0, SKIPSYSTEM|flags)) != 0)
785 for (i = 0; i < MAXQUOTAS; i++) {
786 if (ump->um_quotas[i] == NULLVP)
791 * Here we fall through to vflush again to ensure
792 * that we have gotten rid of all the system vnodes.
796 error = vflush(mp, 0, flags);
801 * Get file system statistics.
802 * taken from ext2/super.c ext2_statfs
805 ext2_statfs(struct mount *mp, struct statfs *sbp, struct thread *td)
807 unsigned long overhead;
808 struct ufsmount *ump;
809 struct ext2_sb_info *fs;
810 struct ext2_super_block *es;
817 if (es->s_magic != EXT2_SUPER_MAGIC)
818 panic("ext2_statfs - magic number spoiled");
821 * Compute the overhead (FS structures)
823 if (es->s_feature_ro_compat & EXT2_FEATURE_RO_COMPAT_SPARSE_SUPER) {
825 for (i = 0 ; i < fs->s_groups_count; i++)
826 if (ext2_group_sparse(i))
829 nsb = fs->s_groups_count;
830 overhead = es->s_first_data_block +
831 /* Superblocks and block group descriptors: */
832 nsb * (1 + fs->s_db_per_group) +
833 /* Inode bitmap, block bitmap, and inode table: */
834 fs->s_groups_count * (1 + 1 + fs->s_itb_per_group);
836 sbp->f_bsize = EXT2_FRAG_SIZE(fs);
837 sbp->f_iosize = EXT2_BLOCK_SIZE(fs);
838 sbp->f_blocks = es->s_blocks_count - overhead;
839 sbp->f_bfree = es->s_free_blocks_count;
840 sbp->f_bavail = sbp->f_bfree - es->s_r_blocks_count;
841 sbp->f_files = es->s_inodes_count;
842 sbp->f_ffree = es->s_free_inodes_count;
843 if (sbp != &mp->mnt_stat) {
844 sbp->f_type = mp->mnt_vfc->vfc_typenum;
845 bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
846 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
852 * Go through the disk queues to initiate sandbagged IO;
853 * go through the inodes to write those that have been modified;
854 * initiate the writing of the super block if it has been modified.
856 * Note: we are always called with the filesystem marked `MPBUSY'.
859 static int ext2_sync_scan(struct mount *mp, struct vnode *vp, void *data);
862 ext2_sync(struct mount *mp, int waitfor, struct thread *td)
864 struct ufsmount *ump = VFSTOUFS(mp);
865 struct ext2_sb_info *fs;
866 struct scaninfo scaninfo;
870 if (fs->s_dirt != 0 && fs->s_rd_only != 0) { /* XXX */
871 printf("fs = %s\n", fs->fs_fsmnt);
872 panic("ext2_sync: rofs mod");
876 * Write back each (modified) inode.
878 scaninfo.allerror = 0;
880 scaninfo.waitfor = waitfor;
882 while (scaninfo.rescan) {
884 vmntvnodescan(mp, VMSC_GETVP|VMSC_NOWAIT,
885 NULL, ext2_sync_scan, &scaninfo);
889 * Force stale file system control information to be flushed.
891 if (waitfor != MNT_LAZY) {
892 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY, td);
893 if ((error = VOP_FSYNC(ump->um_devvp, waitfor, td)) != 0)
894 scaninfo.allerror = error;
895 VOP_UNLOCK(ump->um_devvp, 0, td);
901 * Write back modified superblock.
903 if (fs->s_dirt != 0) {
905 fs->s_es->s_wtime = time_second;
906 if ((error = ext2_sbupdate(ump, waitfor)) != 0)
907 scaninfo.allerror = error;
909 return (scaninfo.allerror);
913 ext2_sync_scan(struct mount *mp, struct vnode *vp, void *data)
915 struct scaninfo *info = data;
920 if (vp->v_type == VNON ||
922 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
923 (RB_EMPTY(&vp->v_rbdirty_tree) || info->waitfor == MNT_LAZY))) {
926 if ((error = VOP_FSYNC(vp, info->waitfor, info->td)) != 0)
927 info->allerror = error;
932 * Look up a EXT2FS dinode number to find its incore vnode, otherwise read it
933 * in from disk. If it is in core, wait for the lock bit to clear, then
934 * return the inode locked. Detection and handling of mount points must be
935 * done by the calling routine.
938 ext2_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
940 struct ext2_sb_info *fs;
942 struct ufsmount *ump;
952 if ((*vpp = ufs_ihashget(dev, ino)) != NULL)
956 * Lock out the creation of new entries in the FFS hash table in
957 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate
960 if (ext2fs_inode_hash_lock) {
961 while (ext2fs_inode_hash_lock) {
962 ext2fs_inode_hash_lock = -1;
963 tsleep(&ext2fs_inode_hash_lock, 0, "e2vget", 0);
967 ext2fs_inode_hash_lock = 1;
970 * If this MALLOC() is performed after the getnewvnode()
971 * it might block, leaving a vnode with a NULL v_data to be
972 * found by ext2_sync() if a sync happens to fire right then,
973 * which will cause a panic because ext2_sync() blindly
974 * dereferences vp->v_data (as well it should).
976 MALLOC(ip, struct inode *, sizeof(struct inode), M_EXT2NODE, M_WAITOK);
978 /* Allocate a new vnode/inode. */
979 if ((error = getnewvnode(VT_UFS, mp, &vp, 0, LK_CANRECURSE)) != 0) {
980 if (ext2fs_inode_hash_lock < 0)
981 wakeup(&ext2fs_inode_hash_lock);
982 ext2fs_inode_hash_lock = 0;
984 FREE(ip, M_EXT2NODE);
987 bzero((caddr_t)ip, sizeof(struct inode));
990 ip->i_e2fs = fs = ump->um_e2fs;
994 for (i = 0; i < MAXQUOTAS; i++)
995 ip->i_dquot[i] = NODQUOT;
998 * Put it onto its hash chain. Since our vnode is locked, other
999 * requests for this inode will block if they arrive while we are
1000 * sleeping waiting for old data structures to be purged or for the
1001 * contents of the disk portion of this inode to be read.
1005 if (ext2fs_inode_hash_lock < 0)
1006 wakeup(&ext2fs_inode_hash_lock);
1007 ext2fs_inode_hash_lock = 0;
1009 /* Read in the disk contents for the inode, copy into the inode. */
1011 printf("ext2_vget(%d) dbn= %d ", ino, fsbtodb(fs, ino_to_fsba(fs, ino)));
1013 if ((error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
1014 (int)fs->s_blocksize, &bp)) != 0) {
1016 * The inode does not contain anything useful, so it would
1017 * be misleading to leave it on its hash chain. With mode
1018 * still zero, it will be unlinked and returned to the free
1026 /* convert ext2 inode to dinode */
1027 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data + EXT2_INODE_SIZE *
1028 ino_to_fsbo(fs, ino)), &ip->i_din);
1029 ip->i_block_group = ino_to_cg(fs, ino);
1030 ip->i_next_alloc_block = 0;
1031 ip->i_next_alloc_goal = 0;
1032 ip->i_prealloc_count = 0;
1033 ip->i_prealloc_block = 0;
1034 /* now we want to make sure that block pointers for unused
1035 blocks are zeroed out - ext2_balloc depends on this
1036 although for regular files and directories only
1038 if(S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode)) {
1039 used_blocks = (ip->i_size+fs->s_blocksize-1) / fs->s_blocksize;
1040 for(i = used_blocks; i < EXT2_NDIR_BLOCKS; i++)
1044 ext2_print_inode(ip);
1049 * Initialize the vnode from the inode, check for aliases.
1050 * Note that the underlying vnode may have changed.
1052 if ((error = ufs_vinit(mp, &vp)) != 0) {
1058 * Finish inode initialization now that aliasing has been resolved.
1060 ip->i_devvp = ump->um_devvp;
1063 * Set up a generation number for this inode if it does not
1064 * already have one. This should only happen on old filesystems.
1066 if (ip->i_gen == 0) {
1067 ip->i_gen = random() / 2 + 1;
1068 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1069 ip->i_flag |= IN_MODIFIED;
1072 * Return the locked and refd vnode.
1079 * File handle to vnode
1081 * Have to be really careful about stale file handles:
1082 * - check that the inode number is valid
1083 * - call ext2_vget() to get the locked inode
1084 * - check for an unallocated inode (i_mode == 0)
1085 * - check that the given client host has export rights and return
1086 * those rights via. exflagsp and credanonp
1089 ext2_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
1092 struct ext2_sb_info *fs;
1094 ufhp = (struct ufid *)fhp;
1095 fs = VFSTOUFS(mp)->um_e2fs;
1096 if (ufhp->ufid_ino < ROOTINO ||
1097 ufhp->ufid_ino > fs->s_groups_count * fs->s_es->s_inodes_per_group)
1099 return (ufs_fhtovp(mp, ufhp, vpp));
1103 * Vnode pointer to File handle
1107 ext2_vptofh(struct vnode *vp, struct fid *fhp)
1113 ufhp = (struct ufid *)fhp;
1114 ufhp->ufid_len = sizeof(struct ufid);
1115 ufhp->ufid_ino = ip->i_number;
1116 ufhp->ufid_gen = ip->i_gen;
1121 * Write a superblock and associated information back to disk.
1124 ext2_sbupdate(struct ufsmount *mp, int waitfor)
1126 struct ext2_sb_info *fs = mp->um_e2fs;
1127 struct ext2_super_block *es = fs->s_es;
1131 printf("\nupdating superblock, waitfor=%s\n", waitfor == MNT_WAIT ? "yes":"no");
1133 bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0);
1134 bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2_super_block));
1135 if (waitfor == MNT_WAIT)
1141 * The buffers for group descriptors, inode bitmaps and block bitmaps
1142 * are not busy at this point and are (hopefully) written by the
1143 * usual sync mechanism. No need to write them here