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 1.19 2004/08/28 19:02:12 dillon Exp $
44 #include "opt_quota.h"
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/namei.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 *,
80 char *, caddr_t, struct nameidata *, 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);
123 static int ext2_mountroot (void);
126 * Called by main() when ext2fs is going to be mounted as root.
128 * Name is updated by mount(8) after booting.
130 #define ROOTNAME "root_device"
135 struct ext2_sb_info *fs;
137 struct thread *td = curthread;
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 *) */
188 struct nameidata *ndp, struct thread *td)
191 struct ufs_args args;
192 struct ufsmount *ump = 0;
193 struct ext2_sb_info *fs;
199 if ((error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args))) != 0)
202 cred = td->td_proc->p_ucred;
204 * If updating, check whether changing from read-only to
205 * read/write; if there is no device name, that's all we do.
207 if (mp->mnt_flag & MNT_UPDATE) {
211 if (fs->s_rd_only == 0 && (mp->mnt_flag & MNT_RDONLY)) {
213 if (mp->mnt_flag & MNT_FORCE)
215 if (vfs_busy(mp, LK_NOWAIT, NULL, td))
217 error = ext2_flushfiles(mp, flags, td);
219 if (!error && fs->s_wasvalid) {
220 fs->s_es->s_state |= EXT2_VALID_FS;
221 ext2_sbupdate(ump, MNT_WAIT);
225 if (!error && (mp->mnt_flag & MNT_RELOAD))
226 error = ext2_reload(mp, ndp->ni_cnd.cn_cred, td);
229 devvp = ump->um_devvp;
230 if (ext2_check_sb_compat(fs->s_es, devvp->v_rdev,
231 (mp->mnt_kern_flag & MNTK_WANTRDWR) == 0) != 0)
233 if (fs->s_rd_only && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
235 * If upgrade to read-write by non-root, then verify
236 * that user has necessary permissions on the device.
238 if (cred->cr_uid != 0) {
239 vn_lock(devvp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
240 if ((error = VOP_ACCESS(devvp, VREAD | VWRITE,
242 VOP_UNLOCK(devvp, NULL, 0, td);
245 VOP_UNLOCK(devvp, NULL, 0, td);
248 if ((fs->s_es->s_state & EXT2_VALID_FS) == 0 ||
249 (fs->s_es->s_state & EXT2_ERROR_FS)) {
250 if (mp->mnt_flag & MNT_FORCE) {
252 "WARNING: %s was not properly dismounted\n",
256 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
261 fs->s_es->s_state &= ~EXT2_VALID_FS;
262 ext2_sbupdate(ump, MNT_WAIT);
265 if (args.fspec == 0) {
267 * Process export requests.
269 return (vfs_export(mp, &ump->um_export, &args.export));
273 * Not an update, or updating the name: look up the name
274 * and verify that it refers to a sensible block device.
276 NDINIT(ndp, NAMEI_LOOKUP, CNP_FOLLOW, UIO_USERSPACE, args.fspec, td);
277 if ((error = namei(ndp)) != 0)
279 NDFREE(ndp, NDF_ONLY_PNBUF);
282 if (!vn_isdisk(devvp, &error)) {
288 * If mount by non-root, then verify that user has necessary
289 * permissions on the device.
291 if (cred->cr_uid != 0) {
293 if ((mp->mnt_flag & MNT_RDONLY) == 0)
294 accessmode |= VWRITE;
295 vn_lock(devvp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
296 if ((error = VOP_ACCESS(devvp, accessmode, cred, td)) != 0) {
300 VOP_UNLOCK(devvp, NULL, 0, td);
303 if ((mp->mnt_flag & MNT_UPDATE) == 0) {
304 error = ext2_mountfs(devvp, mp, td);
306 if (devvp != ump->um_devvp)
307 error = EINVAL; /* needs translation */
317 (void) copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size);
318 bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size);
319 bcopy((caddr_t)fs->fs_fsmnt, (caddr_t)mp->mnt_stat.f_mntonname,
321 (void) copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
323 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
324 (void)ext2_statfs(mp, &mp->mnt_stat, td);
329 * checks that the data in the descriptor blocks make sense
330 * this is taken from ext2/super.c
332 static int ext2_check_descriptors(struct ext2_sb_info *sb)
336 unsigned long block = sb->s_es->s_first_data_block;
337 struct ext2_group_desc * gdp = NULL;
339 /* ext2_debug ("Checking group descriptors"); */
341 for (i = 0; i < sb->s_groups_count; i++)
343 /* examine next descriptor block */
344 if ((i % EXT2_DESC_PER_BLOCK(sb)) == 0)
345 gdp = (struct ext2_group_desc *)
346 sb->s_group_desc[desc_block++]->b_data;
347 if (gdp->bg_block_bitmap < block ||
348 gdp->bg_block_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb))
350 printf ("ext2_check_descriptors: "
351 "Block bitmap for group %d"
352 " not in group (block %lu)!\n",
353 i, (unsigned long) gdp->bg_block_bitmap);
356 if (gdp->bg_inode_bitmap < block ||
357 gdp->bg_inode_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb))
359 printf ("ext2_check_descriptors: "
360 "Inode bitmap for group %d"
361 " not in group (block %lu)!\n",
362 i, (unsigned long) gdp->bg_inode_bitmap);
365 if (gdp->bg_inode_table < block ||
366 gdp->bg_inode_table + sb->s_itb_per_group >=
367 block + EXT2_BLOCKS_PER_GROUP(sb))
369 printf ("ext2_check_descriptors: "
370 "Inode table for group %d"
371 " not in group (block %lu)!\n",
372 i, (unsigned long) gdp->bg_inode_table);
375 block += EXT2_BLOCKS_PER_GROUP(sb);
382 ext2_check_sb_compat(struct ext2_super_block *es, dev_t dev, int ronly)
384 if (es->s_magic != EXT2_SUPER_MAGIC) {
385 printf("ext2fs: %s: wrong magic number %#x (expected %#x)\n",
386 devtoname(dev), es->s_magic, EXT2_SUPER_MAGIC);
389 if (es->s_rev_level > EXT2_GOOD_OLD_REV) {
390 if (es->s_feature_incompat & ~EXT2_FEATURE_INCOMPAT_SUPP) {
392 "WARNING: mount of %s denied due to unsupported optional features\n",
397 (es->s_feature_ro_compat & ~EXT2_FEATURE_RO_COMPAT_SUPP)) {
399 "WARNING: R/W mount of %s denied due to unsupported optional features\n",
408 * this computes the fields of the ext2_sb_info structure from the
409 * data in the ext2_super_block structure read in
412 compute_sb_data(struct vnode *devvp, struct ext2_super_block *es,
413 struct ext2_sb_info *fs)
417 int logic_sb_block = 1; /* XXX for now */
422 #define V(v) printf(#v"= %d\n", fs->v);
425 fs->s_blocksize = EXT2_MIN_BLOCK_SIZE << es->s_log_block_size;
427 fs->s_bshift = EXT2_MIN_BLOCK_LOG_SIZE + es->s_log_block_size;
429 fs->s_fsbtodb = es->s_log_block_size + 1;
431 fs->s_qbmask = fs->s_blocksize - 1;
433 fs->s_blocksize_bits = EXT2_BLOCK_SIZE_BITS(es);
435 fs->s_frag_size = EXT2_MIN_FRAG_SIZE << es->s_log_frag_size;
438 fs->s_frags_per_block = fs->s_blocksize / fs->s_frag_size;
440 fs->s_blocks_per_group = es->s_blocks_per_group;
441 V(s_blocks_per_group)
442 fs->s_frags_per_group = es->s_frags_per_group;
444 fs->s_inodes_per_group = es->s_inodes_per_group;
445 V(s_inodes_per_group)
446 fs->s_inodes_per_block = fs->s_blocksize / EXT2_INODE_SIZE;
447 V(s_inodes_per_block)
448 fs->s_itb_per_group = fs->s_inodes_per_group /fs->s_inodes_per_block;
450 fs->s_desc_per_block = fs->s_blocksize / sizeof (struct ext2_group_desc);
452 /* s_resuid / s_resgid ? */
453 fs->s_groups_count = (es->s_blocks_count -
454 es->s_first_data_block +
455 EXT2_BLOCKS_PER_GROUP(fs) - 1) /
456 EXT2_BLOCKS_PER_GROUP(fs);
458 db_count = (fs->s_groups_count + EXT2_DESC_PER_BLOCK(fs) - 1) /
459 EXT2_DESC_PER_BLOCK(fs);
460 fs->s_db_per_group = db_count;
463 fs->s_group_desc = bsd_malloc(db_count * sizeof (struct buf *),
466 /* adjust logic_sb_block */
467 if(fs->s_blocksize > SBSIZE)
468 /* Godmar thinks: if the blocksize is greater than 1024, then
469 the superblock is logically part of block zero.
473 for (i = 0; i < db_count; i++) {
474 error = bread(devvp , fsbtodb(fs, logic_sb_block + i + 1),
475 fs->s_blocksize, &fs->s_group_desc[i]);
477 for (j = 0; j < i; j++)
478 brelse(fs->s_group_desc[j]);
479 bsd_free(fs->s_group_desc, M_UFSMNT);
480 printf("EXT2-fs: unable to read group descriptors (%d)\n", error);
483 /* Set the B_LOCKED flag on the buffer, then brelse() it */
484 LCK_BUF(fs->s_group_desc[i])
486 if(!ext2_check_descriptors(fs)) {
487 for (j = 0; j < db_count; j++)
488 ULCK_BUF(fs->s_group_desc[j])
489 bsd_free(fs->s_group_desc, M_UFSMNT);
490 printf("EXT2-fs: (ext2_check_descriptors failure) "
491 "unable to read group descriptors\n");
495 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) {
496 fs->s_inode_bitmap_number[i] = 0;
497 fs->s_inode_bitmap[i] = NULL;
498 fs->s_block_bitmap_number[i] = 0;
499 fs->s_block_bitmap[i] = NULL;
501 fs->s_loaded_inode_bitmaps = 0;
502 fs->s_loaded_block_bitmaps = 0;
507 * Reload all incore data for a filesystem (used after running fsck on
508 * the root filesystem and finding things to fix). The filesystem must
509 * be mounted read-only.
511 * Things to do to update the mount:
512 * 1) invalidate all cached meta-data.
513 * 2) re-read superblock from disk.
514 * 3) re-read summary information from disk.
515 * 4) invalidate all inactive vnodes.
516 * 5) invalidate all cached file data.
517 * 6) re-read inode data for all active vnodes.
519 static int ext2_reload_scan1(struct mount *mp, struct vnode *vp, void *rescan);
520 static int ext2_reload_scan2(struct mount *mp, struct vnode *vp,
521 lwkt_tokref_t vlock, void *rescan);
529 struct ext2_sb_info *fs;
533 ext2_reload(struct mount *mountp, struct ucred *cred, struct thread *td)
537 struct ext2_super_block * es;
538 struct ext2_sb_info *fs;
540 struct scaninfo scaninfo;
542 if ((mountp->mnt_flag & MNT_RDONLY) == 0)
545 * Step 1: invalidate all cached meta-data.
547 devvp = VFSTOUFS(mountp)->um_devvp;
548 if (vinvalbuf(devvp, 0, td, 0, 0))
549 panic("ext2_reload: dirty1");
551 * Step 2: re-read superblock from disk.
552 * constants have been adjusted for ext2
554 if ((error = bread(devvp, SBLOCK, SBSIZE, &bp)) != 0)
556 es = (struct ext2_super_block *)bp->b_data;
557 if (ext2_check_sb_compat(es, devvp->v_rdev, 0) != 0) {
559 return (EIO); /* XXX needs translation */
561 fs = VFSTOUFS(mountp)->um_e2fs;
562 bcopy(bp->b_data, fs->s_es, sizeof(struct ext2_super_block));
564 if((error = compute_sb_data(devvp, es, fs)) != 0) {
569 if (fs->fs_sbsize < SBSIZE)
570 bp->b_flags |= B_INVAL;
576 scaninfo.devvp = devvp;
578 while (error == 0 && scaninfo.rescan) {
580 error = vmntvnodescan(mountp, ext2_reload_scan1,
581 ext2_reload_scan2, &scaninfo);
587 ext2_reload_scan1(struct mount *mp, struct vnode *vp, void *data)
589 struct scaninfo *info = data;
592 * Step 4: invalidate all inactive vnodes.
594 if (vrecycle(vp, NULL, curthread)) {
596 return(-1); /* continue loop, do not call scan2 */
602 ext2_reload_scan2(struct mount *mp, struct vnode *vp, lwkt_tokref_t vlock, void *data)
604 struct scaninfo *info = data;
610 * Step 5: invalidate all cached file data.
612 if (vget(vp, vlock, LK_EXCLUSIVE | LK_INTERLOCK, info->td)) {
616 if (vinvalbuf(vp, 0, info->td, 0, 0))
617 panic("ext2_reload: dirty2");
619 * Step 6: re-read inode data for all active vnodes.
622 error = bread(info->devvp, fsbtodb(info->fs, ino_to_fsba(info->fs, ip->i_number)),
623 (int)info->fs->s_blocksize, &bp);
628 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data +
629 EXT2_INODE_SIZE * ino_to_fsbo(info->fs, ip->i_number)),
637 * Common code for mount and mountroot
640 ext2_mountfs(struct vnode *devvp, struct mount *mp, struct thread *td)
642 struct ufsmount *ump;
644 struct ext2_sb_info *fs;
645 struct ext2_super_block * es;
647 struct partinfo dpart;
653 * Disallow multiple mounts of the same device.
654 * Disallow mounting of a device that is currently in use
655 * (except for root, which might share swap device for miniroot).
656 * Flush out any old buffers remaining from a previous use.
658 if ((error = vfs_mountedon(devvp)) != 0)
660 if (count_udev(devvp->v_udev) > 0 && devvp != rootvp)
662 if ((error = vinvalbuf(devvp, V_SAVE, td, 0, 0)) != 0)
665 /* turn on this to force it to be read-only */
666 mp->mnt_flag |= MNT_RDONLY;
669 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
670 vn_lock(devvp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
671 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, td);
672 VOP_UNLOCK(devvp, NULL, 0, td);
676 if (dev->si_iosize_max != 0)
677 mp->mnt_iosize_max = dev->si_iosize_max;
678 if (mp->mnt_iosize_max > MAXPHYS)
679 mp->mnt_iosize_max = MAXPHYS;
680 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, td) != 0)
684 size = dpart.disklab->d_secsize;
689 if ((error = bread(devvp, SBLOCK, SBSIZE, &bp)) != 0)
691 es = (struct ext2_super_block *)bp->b_data;
692 if (ext2_check_sb_compat(es, dev, ronly) != 0) {
693 error = EINVAL; /* XXX needs translation */
696 if ((es->s_state & EXT2_VALID_FS) == 0 ||
697 (es->s_state & EXT2_ERROR_FS)) {
698 if (ronly || (mp->mnt_flag & MNT_FORCE)) {
700 "WARNING: Filesystem was not properly dismounted\n");
703 "WARNING: R/W mount denied. Filesystem is not clean - run fsck\n");
708 ump = bsd_malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
709 bzero((caddr_t)ump, sizeof *ump);
710 ump->um_malloctype = M_EXT2NODE;
711 ump->um_blkatoff = ext2_blkatoff;
712 ump->um_truncate = ext2_truncate;
713 ump->um_update = ext2_update;
714 ump->um_valloc = ext2_valloc;
715 ump->um_vfree = ext2_vfree;
716 /* I don't know whether this is the right strategy. Note that
717 we dynamically allocate both a ext2_sb_info and a ext2_super_block
718 while Linux keeps the super block in a locked buffer
720 ump->um_e2fs = bsd_malloc(sizeof(struct ext2_sb_info),
722 ump->um_e2fs->s_es = bsd_malloc(sizeof(struct ext2_super_block),
724 bcopy(es, ump->um_e2fs->s_es, (u_int)sizeof(struct ext2_super_block));
725 if ((error = compute_sb_data(devvp, ump->um_e2fs->s_es, ump->um_e2fs)))
728 * We don't free the group descriptors allocated by compute_sb_data()
729 * until ext2_unmount(). This is OK since the mount will succeed.
734 fs->s_rd_only = ronly; /* ronly is set according to mnt_flags */
735 /* if the fs is not mounted read-only, make sure the super block is
736 always written back on a sync()
738 fs->s_wasvalid = fs->s_es->s_state & EXT2_VALID_FS ? 1 : 0;
740 fs->s_dirt = 1; /* mark it modified */
741 fs->s_es->s_state &= ~EXT2_VALID_FS; /* set fs invalid */
743 mp->mnt_data = (qaddr_t)ump;
744 mp->mnt_stat.f_fsid.val[0] = dev2udev(dev);
745 mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
746 mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN;
747 mp->mnt_flag |= MNT_LOCAL;
750 ump->um_devvp = devvp;
751 /* setting those two parameters allows us to use
752 ufs_bmap w/o changse !
754 ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs);
755 ump->um_bptrtodb = fs->s_es->s_log_block_size + 1;
756 ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs);
757 for (i = 0; i < MAXQUOTAS; i++)
758 ump->um_quotas[i] = NULLVP;
759 dev->si_mountpoint = mp;
761 vfs_add_vnodeops(&mp->mnt_vn_ops, ext2_vnodeop_entries);
762 vfs_add_vnodeops(&mp->mnt_vn_spec_ops, ext2_specop_entries);
763 vfs_add_vnodeops(&mp->mnt_vn_fifo_ops, ext2_fifoop_entries);
766 ext2_sbupdate(ump, MNT_WAIT);
771 (void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, td);
773 bsd_free(ump->um_e2fs->s_es, M_UFSMNT);
774 bsd_free(ump->um_e2fs, M_UFSMNT);
775 bsd_free(ump, M_UFSMNT);
776 mp->mnt_data = (qaddr_t)0;
782 * unmount system call
785 ext2_unmount(struct mount *mp, int mntflags, struct thread *td)
787 struct ufsmount *ump;
788 struct ext2_sb_info *fs;
789 int error, flags, ronly, i;
792 if (mntflags & MNT_FORCE) {
793 if (mp->mnt_flag & MNT_ROOTFS)
797 if ((error = ext2_flushfiles(mp, flags, td)) != 0)
801 ronly = fs->s_rd_only;
804 fs->s_es->s_state |= EXT2_VALID_FS;
805 ext2_sbupdate(ump, MNT_WAIT);
808 /* release buffers containing group descriptors */
809 for(i = 0; i < fs->s_db_per_group; i++)
810 ULCK_BUF(fs->s_group_desc[i])
811 bsd_free(fs->s_group_desc, M_UFSMNT);
813 /* release cached inode/block bitmaps */
814 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
815 if (fs->s_inode_bitmap[i])
816 ULCK_BUF(fs->s_inode_bitmap[i])
818 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
819 if (fs->s_block_bitmap[i])
820 ULCK_BUF(fs->s_block_bitmap[i])
822 ump->um_devvp->v_rdev->si_mountpoint = NULL;
823 error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD|FWRITE, td);
824 vrele(ump->um_devvp);
825 bsd_free(fs->s_es, M_UFSMNT);
826 bsd_free(fs, M_UFSMNT);
827 bsd_free(ump, M_UFSMNT);
828 mp->mnt_data = (qaddr_t)0;
829 mp->mnt_flag &= ~MNT_LOCAL;
834 * Flush out all the files in a filesystem.
837 ext2_flushfiles(struct mount *mp, int flags, struct thread *td)
839 struct ufsmount *ump;
847 if (mp->mnt_flag & MNT_QUOTA) {
848 if ((error = vflush(mp, 0, SKIPSYSTEM|flags)) != 0)
850 for (i = 0; i < MAXQUOTAS; i++) {
851 if (ump->um_quotas[i] == NULLVP)
856 * Here we fall through to vflush again to ensure
857 * that we have gotten rid of all the system vnodes.
861 error = vflush(mp, 0, flags);
866 * Get file system statistics.
867 * taken from ext2/super.c ext2_statfs
870 ext2_statfs(struct mount *mp, struct statfs *sbp, struct thread *td)
872 unsigned long overhead;
873 struct ufsmount *ump;
874 struct ext2_sb_info *fs;
875 struct ext2_super_block *es;
882 if (es->s_magic != EXT2_SUPER_MAGIC)
883 panic("ext2_statfs - magic number spoiled");
886 * Compute the overhead (FS structures)
888 if (es->s_feature_ro_compat & EXT2_FEATURE_RO_COMPAT_SPARSE_SUPER) {
890 for (i = 0 ; i < fs->s_groups_count; i++)
891 if (ext2_group_sparse(i))
894 nsb = fs->s_groups_count;
895 overhead = es->s_first_data_block +
896 /* Superblocks and block group descriptors: */
897 nsb * (1 + fs->s_db_per_group) +
898 /* Inode bitmap, block bitmap, and inode table: */
899 fs->s_groups_count * (1 + 1 + fs->s_itb_per_group);
901 sbp->f_bsize = EXT2_FRAG_SIZE(fs);
902 sbp->f_iosize = EXT2_BLOCK_SIZE(fs);
903 sbp->f_blocks = es->s_blocks_count - overhead;
904 sbp->f_bfree = es->s_free_blocks_count;
905 sbp->f_bavail = sbp->f_bfree - es->s_r_blocks_count;
906 sbp->f_files = es->s_inodes_count;
907 sbp->f_ffree = es->s_free_inodes_count;
908 if (sbp != &mp->mnt_stat) {
909 sbp->f_type = mp->mnt_vfc->vfc_typenum;
910 bcopy((caddr_t)mp->mnt_stat.f_mntonname,
911 (caddr_t)&sbp->f_mntonname[0], MNAMELEN);
912 bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
913 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
919 * Go through the disk queues to initiate sandbagged IO;
920 * go through the inodes to write those that have been modified;
921 * initiate the writing of the super block if it has been modified.
923 * Note: we are always called with the filesystem marked `MPBUSY'.
926 static int ext2_sync_scan(struct mount *mp, struct vnode *vp,
927 lwkt_tokref_t vlock, void *data);
930 ext2_sync(struct mount *mp, int waitfor, struct thread *td)
932 struct ufsmount *ump = VFSTOUFS(mp);
933 struct ext2_sb_info *fs;
934 struct scaninfo scaninfo;
938 if (fs->s_dirt != 0 && fs->s_rd_only != 0) { /* XXX */
939 printf("fs = %s\n", fs->fs_fsmnt);
940 panic("ext2_sync: rofs mod");
944 * Write back each (modified) inode.
946 scaninfo.allerror = 0;
948 scaninfo.waitfor = waitfor;
950 while (scaninfo.rescan) {
952 vmntvnodescan(mp, NULL, ext2_sync_scan, &scaninfo);
956 * Force stale file system control information to be flushed.
958 if (waitfor != MNT_LAZY) {
959 vn_lock(ump->um_devvp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
960 if ((error = VOP_FSYNC(ump->um_devvp, waitfor, td)) != 0)
961 scaninfo.allerror = error;
962 VOP_UNLOCK(ump->um_devvp, NULL, 0, td);
968 * Write back modified superblock.
970 if (fs->s_dirt != 0) {
972 fs->s_es->s_wtime = time_second;
973 if ((error = ext2_sbupdate(ump, waitfor)) != 0)
974 scaninfo.allerror = error;
976 return (scaninfo.allerror);
980 ext2_sync_scan(struct mount *mp, struct vnode *vp,
981 lwkt_tokref_t vlock, void *data)
983 struct scaninfo *info = data;
988 if (vp->v_type == VNON ||
990 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
991 (TAILQ_EMPTY(&vp->v_dirtyblkhd) || info->waitfor == MNT_LAZY))) {
992 lwkt_reltoken(vlock);
995 error = vget(vp, vlock, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK, info->td);
1001 if ((error = VOP_FSYNC(vp, info->waitfor, info->td)) != 0)
1002 info->allerror = error;
1003 VOP_UNLOCK(vp, NULL, 0, info->td);
1009 * Look up a EXT2FS dinode number to find its incore vnode, otherwise read it
1010 * in from disk. If it is in core, wait for the lock bit to clear, then
1011 * return the inode locked. Detection and handling of mount points must be
1012 * done by the calling routine.
1015 ext2_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
1017 struct ext2_sb_info *fs;
1019 struct ufsmount *ump;
1029 if ((*vpp = ufs_ihashget(dev, ino)) != NULL)
1033 * Lock out the creation of new entries in the FFS hash table in
1034 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate
1037 if (ext2fs_inode_hash_lock) {
1038 while (ext2fs_inode_hash_lock) {
1039 ext2fs_inode_hash_lock = -1;
1040 tsleep(&ext2fs_inode_hash_lock, 0, "e2vget", 0);
1044 ext2fs_inode_hash_lock = 1;
1047 * If this MALLOC() is performed after the getnewvnode()
1048 * it might block, leaving a vnode with a NULL v_data to be
1049 * found by ext2_sync() if a sync happens to fire right then,
1050 * which will cause a panic because ext2_sync() blindly
1051 * dereferences vp->v_data (as well it should).
1053 MALLOC(ip, struct inode *, sizeof(struct inode), M_EXT2NODE, M_WAITOK);
1055 /* Allocate a new vnode/inode. */
1056 if ((error = getnewvnode(VT_UFS, mp, mp->mnt_vn_ops, &vp, 0, 0)) != 0) {
1057 if (ext2fs_inode_hash_lock < 0)
1058 wakeup(&ext2fs_inode_hash_lock);
1059 ext2fs_inode_hash_lock = 0;
1061 FREE(ip, M_EXT2NODE);
1064 bzero((caddr_t)ip, sizeof(struct inode));
1067 ip->i_e2fs = fs = ump->um_e2fs;
1071 for (i = 0; i < MAXQUOTAS; i++)
1072 ip->i_dquot[i] = NODQUOT;
1075 * Put it onto its hash chain and lock it so that other requests for
1076 * this inode will block if they arrive while we are sleeping waiting
1077 * for old data structures to be purged or for the contents of the
1078 * disk portion of this inode to be read.
1082 if (ext2fs_inode_hash_lock < 0)
1083 wakeup(&ext2fs_inode_hash_lock);
1084 ext2fs_inode_hash_lock = 0;
1086 /* Read in the disk contents for the inode, copy into the inode. */
1088 printf("ext2_vget(%d) dbn= %d ", ino, fsbtodb(fs, ino_to_fsba(fs, ino)));
1090 if ((error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
1091 (int)fs->s_blocksize, &bp)) != 0) {
1093 * The inode does not contain anything useful, so it would
1094 * be misleading to leave it on its hash chain. With mode
1095 * still zero, it will be unlinked and returned to the free
1103 /* convert ext2 inode to dinode */
1104 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data + EXT2_INODE_SIZE *
1105 ino_to_fsbo(fs, ino)), &ip->i_din);
1106 ip->i_block_group = ino_to_cg(fs, ino);
1107 ip->i_next_alloc_block = 0;
1108 ip->i_next_alloc_goal = 0;
1109 ip->i_prealloc_count = 0;
1110 ip->i_prealloc_block = 0;
1111 /* now we want to make sure that block pointers for unused
1112 blocks are zeroed out - ext2_balloc depends on this
1113 although for regular files and directories only
1115 if(S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode)) {
1116 used_blocks = (ip->i_size+fs->s_blocksize-1) / fs->s_blocksize;
1117 for(i = used_blocks; i < EXT2_NDIR_BLOCKS; i++)
1121 ext2_print_inode(ip);
1126 * Initialize the vnode from the inode, check for aliases.
1127 * Note that the underlying vnode may have changed.
1129 if ((error = ufs_vinit(mp, &vp)) != 0) {
1135 * Finish inode initialization now that aliasing has been resolved.
1137 ip->i_devvp = ump->um_devvp;
1140 * Set up a generation number for this inode if it does not
1141 * already have one. This should only happen on old filesystems.
1143 if (ip->i_gen == 0) {
1144 ip->i_gen = random() / 2 + 1;
1145 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1146 ip->i_flag |= IN_MODIFIED;
1153 * File handle to vnode
1155 * Have to be really careful about stale file handles:
1156 * - check that the inode number is valid
1157 * - call ext2_vget() to get the locked inode
1158 * - check for an unallocated inode (i_mode == 0)
1159 * - check that the given client host has export rights and return
1160 * those rights via. exflagsp and credanonp
1163 ext2_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
1166 struct ext2_sb_info *fs;
1168 ufhp = (struct ufid *)fhp;
1169 fs = VFSTOUFS(mp)->um_e2fs;
1170 if (ufhp->ufid_ino < ROOTINO ||
1171 ufhp->ufid_ino > fs->s_groups_count * fs->s_es->s_inodes_per_group)
1173 return (ufs_fhtovp(mp, ufhp, vpp));
1177 * Vnode pointer to File handle
1181 ext2_vptofh(struct vnode *vp, struct fid *fhp)
1187 ufhp = (struct ufid *)fhp;
1188 ufhp->ufid_len = sizeof(struct ufid);
1189 ufhp->ufid_ino = ip->i_number;
1190 ufhp->ufid_gen = ip->i_gen;
1195 * Write a superblock and associated information back to disk.
1198 ext2_sbupdate(struct ufsmount *mp, int waitfor)
1200 struct ext2_sb_info *fs = mp->um_e2fs;
1201 struct ext2_super_block *es = fs->s_es;
1205 printf("\nupdating superblock, waitfor=%s\n", waitfor == MNT_WAIT ? "yes":"no");
1207 bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0);
1208 bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2_super_block));
1209 if (waitfor == MNT_WAIT)
1215 * The buffers for group descriptors, inode bitmaps and block bitmaps
1216 * are not busy at this point and are (hopefully) written by the
1217 * usual sync mechanism. No need to write them here