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 1.10 2003/08/20 09:56:31 rob 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 static int ext2_fhtovp (struct mount *, struct fid *, struct vnode **);
74 static int ext2_flushfiles (struct mount *mp, int flags, struct thread *td);
75 static int ext2_mount (struct mount *,
76 char *, caddr_t, struct nameidata *, struct thread *);
77 static int ext2_mountfs (struct vnode *, struct mount *, struct thread *);
78 static int ext2_reload (struct mount *mountp, struct ucred *cred,
80 static int ext2_sbupdate (struct ufsmount *, int);
81 static int ext2_statfs (struct mount *, struct statfs *, struct thread *);
82 static int ext2_sync (struct mount *, int, struct thread *);
83 static int ext2_unmount (struct mount *, int, struct thread *);
84 static int ext2_vget (struct mount *, ino_t, struct vnode **);
85 static int ext2_vptofh (struct vnode *, struct fid *);
87 static MALLOC_DEFINE(M_EXT2NODE, "EXT2 node", "EXT2 vnode private part");
89 static struct vfsops ext2fs_vfsops = {
91 ufs_start, /* empty function */
93 ufs_root, /* root inode via vget */
94 ufs_quotactl, /* does operations associated with quotas */
106 VFS_SET(ext2fs_vfsops, ext2fs, 0);
107 #define bsd_malloc malloc
108 #define bsd_free free
110 static int ext2fs_inode_hash_lock;
112 static int ext2_check_sb_compat (struct ext2_super_block *es,
113 dev_t dev, int ronly);
114 static int compute_sb_data (struct vnode * devvp,
115 struct ext2_super_block * es,
116 struct ext2_sb_info * fs);
119 static int ext2_mountroot (void);
122 * Called by main() when ext2fs is going to be mounted as root.
124 * Name is updated by mount(8) after booting.
126 #define ROOTNAME "root_device"
131 struct ext2_sb_info *fs;
133 struct thread *td = curthread;
134 struct ufsmount *ump;
138 if ((error = bdevvp(rootdev, &rootvp))) {
139 printf("ext2_mountroot: can't find rootvp\n");
142 mp = bsd_malloc((u_long)sizeof(struct mount), M_MOUNT, M_WAITOK);
143 bzero((char *)mp, (u_long)sizeof(struct mount));
144 TAILQ_INIT(&mp->mnt_nvnodelist);
145 TAILQ_INIT(&mp->mnt_reservedvnlist);
146 mp->mnt_op = &ext2fs_vfsops;
147 mp->mnt_flag = MNT_RDONLY;
148 if (error = ext2_mountfs(rootvp, mp, td)) {
149 bsd_free(mp, M_MOUNT);
152 if (error = vfs_lock(mp)) {
153 (void)ext2_unmount(mp, 0, td);
154 bsd_free(mp, M_MOUNT);
157 TAILQ_INSERT_HEAD(&mountlist, mp, mnt_list);
158 mp->mnt_flag |= MNT_ROOTFS;
159 mp->mnt_vnodecovered = NULLVP;
162 bzero(fs->fs_fsmnt, sizeof(fs->fs_fsmnt));
163 fs->fs_fsmnt[0] = '/';
164 bcopy((caddr_t)fs->fs_fsmnt, (caddr_t)mp->mnt_stat.f_mntonname,
166 (void) copystr(ROOTNAME, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
168 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
169 (void)ext2_statfs(mp, &mp->mnt_stat, td);
171 inittodr(fs->s_es->s_wtime); /* this helps to set the time */
182 ext2_mount(mp, path, data, ndp, td)
185 caddr_t data; /* this is actually a (struct ufs_args *) */
186 struct nameidata *ndp;
190 struct ufs_args args;
191 struct ufsmount *ump = 0;
192 struct ext2_sb_info *fs;
198 if ((error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args))) != 0)
201 cred = td->td_proc->p_ucred;
203 * If updating, check whether changing from read-only to
204 * read/write; if there is no device name, that's all we do.
206 if (mp->mnt_flag & MNT_UPDATE) {
210 if (fs->s_rd_only == 0 && (mp->mnt_flag & MNT_RDONLY)) {
212 if (mp->mnt_flag & MNT_FORCE)
214 if (vfs_busy(mp, LK_NOWAIT, 0, td))
216 error = ext2_flushfiles(mp, flags, td);
218 if (!error && fs->s_wasvalid) {
219 fs->s_es->s_state |= EXT2_VALID_FS;
220 ext2_sbupdate(ump, MNT_WAIT);
224 if (!error && (mp->mnt_flag & MNT_RELOAD))
225 error = ext2_reload(mp, ndp->ni_cnd.cn_cred, td);
228 devvp = ump->um_devvp;
229 if (ext2_check_sb_compat(fs->s_es, devvp->v_rdev,
230 (mp->mnt_kern_flag & MNTK_WANTRDWR) == 0) != 0)
232 if (fs->s_rd_only && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
234 * If upgrade to read-write by non-root, then verify
235 * that user has necessary permissions on the device.
237 if (cred->cr_uid != 0) {
238 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
239 if ((error = VOP_ACCESS(devvp, VREAD | VWRITE,
241 VOP_UNLOCK(devvp, 0, td);
244 VOP_UNLOCK(devvp, 0, td);
247 if ((fs->s_es->s_state & EXT2_VALID_FS) == 0 ||
248 (fs->s_es->s_state & EXT2_ERROR_FS)) {
249 if (mp->mnt_flag & MNT_FORCE) {
251 "WARNING: %s was not properly dismounted\n",
255 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
260 fs->s_es->s_state &= ~EXT2_VALID_FS;
261 ext2_sbupdate(ump, MNT_WAIT);
264 if (args.fspec == 0) {
266 * Process export requests.
268 return (vfs_export(mp, &ump->um_export, &args.export));
272 * Not an update, or updating the name: look up the name
273 * and verify that it refers to a sensible block device.
275 NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, td);
276 if ((error = namei(ndp)) != 0)
278 NDFREE(ndp, NDF_ONLY_PNBUF);
281 if (!vn_isdisk(devvp, &error)) {
287 * If mount by non-root, then verify that user has necessary
288 * permissions on the device.
290 if (cred->cr_uid != 0) {
292 if ((mp->mnt_flag & MNT_RDONLY) == 0)
293 accessmode |= VWRITE;
294 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
295 if ((error = VOP_ACCESS(devvp, accessmode, cred, td)) != 0) {
299 VOP_UNLOCK(devvp, 0, td);
302 if ((mp->mnt_flag & MNT_UPDATE) == 0) {
303 error = ext2_mountfs(devvp, mp, td);
305 if (devvp != ump->um_devvp)
306 error = EINVAL; /* needs translation */
316 (void) copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size);
317 bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size);
318 bcopy((caddr_t)fs->fs_fsmnt, (caddr_t)mp->mnt_stat.f_mntonname,
320 (void) copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
322 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
323 (void)ext2_statfs(mp, &mp->mnt_stat, td);
328 * checks that the data in the descriptor blocks make sense
329 * this is taken from ext2/super.c
331 static int ext2_check_descriptors (struct ext2_sb_info * sb)
335 unsigned long block = sb->s_es->s_first_data_block;
336 struct ext2_group_desc * gdp = NULL;
338 /* ext2_debug ("Checking group descriptors"); */
340 for (i = 0; i < sb->s_groups_count; i++)
342 /* examine next descriptor block */
343 if ((i % EXT2_DESC_PER_BLOCK(sb)) == 0)
344 gdp = (struct ext2_group_desc *)
345 sb->s_group_desc[desc_block++]->b_data;
346 if (gdp->bg_block_bitmap < block ||
347 gdp->bg_block_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb))
349 printf ("ext2_check_descriptors: "
350 "Block bitmap for group %d"
351 " not in group (block %lu)!\n",
352 i, (unsigned long) gdp->bg_block_bitmap);
355 if (gdp->bg_inode_bitmap < block ||
356 gdp->bg_inode_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb))
358 printf ("ext2_check_descriptors: "
359 "Inode bitmap for group %d"
360 " not in group (block %lu)!\n",
361 i, (unsigned long) gdp->bg_inode_bitmap);
364 if (gdp->bg_inode_table < block ||
365 gdp->bg_inode_table + sb->s_itb_per_group >=
366 block + EXT2_BLOCKS_PER_GROUP(sb))
368 printf ("ext2_check_descriptors: "
369 "Inode table for group %d"
370 " not in group (block %lu)!\n",
371 i, (unsigned long) gdp->bg_inode_table);
374 block += EXT2_BLOCKS_PER_GROUP(sb);
381 ext2_check_sb_compat(es, dev, ronly)
382 struct ext2_super_block *es;
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
414 static int compute_sb_data(devvp, es, fs)
415 struct vnode * devvp;
416 struct ext2_super_block * es;
417 struct ext2_sb_info * fs;
421 int logic_sb_block = 1; /* XXX for now */
426 #define V(v) printf(#v"= %d\n", fs->v);
429 fs->s_blocksize = EXT2_MIN_BLOCK_SIZE << es->s_log_block_size;
431 fs->s_bshift = EXT2_MIN_BLOCK_LOG_SIZE + es->s_log_block_size;
433 fs->s_fsbtodb = es->s_log_block_size + 1;
435 fs->s_qbmask = fs->s_blocksize - 1;
437 fs->s_blocksize_bits = EXT2_BLOCK_SIZE_BITS(es);
439 fs->s_frag_size = EXT2_MIN_FRAG_SIZE << es->s_log_frag_size;
442 fs->s_frags_per_block = fs->s_blocksize / fs->s_frag_size;
444 fs->s_blocks_per_group = es->s_blocks_per_group;
445 V(s_blocks_per_group)
446 fs->s_frags_per_group = es->s_frags_per_group;
448 fs->s_inodes_per_group = es->s_inodes_per_group;
449 V(s_inodes_per_group)
450 fs->s_inodes_per_block = fs->s_blocksize / EXT2_INODE_SIZE;
451 V(s_inodes_per_block)
452 fs->s_itb_per_group = fs->s_inodes_per_group /fs->s_inodes_per_block;
454 fs->s_desc_per_block = fs->s_blocksize / sizeof (struct ext2_group_desc);
456 /* s_resuid / s_resgid ? */
457 fs->s_groups_count = (es->s_blocks_count -
458 es->s_first_data_block +
459 EXT2_BLOCKS_PER_GROUP(fs) - 1) /
460 EXT2_BLOCKS_PER_GROUP(fs);
462 db_count = (fs->s_groups_count + EXT2_DESC_PER_BLOCK(fs) - 1) /
463 EXT2_DESC_PER_BLOCK(fs);
464 fs->s_db_per_group = db_count;
467 fs->s_group_desc = bsd_malloc(db_count * sizeof (struct buf *),
470 /* adjust logic_sb_block */
471 if(fs->s_blocksize > SBSIZE)
472 /* Godmar thinks: if the blocksize is greater than 1024, then
473 the superblock is logically part of block zero.
477 for (i = 0; i < db_count; i++) {
478 error = bread(devvp , fsbtodb(fs, logic_sb_block + i + 1),
479 fs->s_blocksize, &fs->s_group_desc[i]);
481 for (j = 0; j < i; j++)
482 brelse(fs->s_group_desc[j]);
483 bsd_free(fs->s_group_desc, M_UFSMNT);
484 printf("EXT2-fs: unable to read group descriptors (%d)\n", error);
487 /* Set the B_LOCKED flag on the buffer, then brelse() it */
488 LCK_BUF(fs->s_group_desc[i])
490 if(!ext2_check_descriptors(fs)) {
491 for (j = 0; j < db_count; j++)
492 ULCK_BUF(fs->s_group_desc[j])
493 bsd_free(fs->s_group_desc, M_UFSMNT);
494 printf("EXT2-fs: (ext2_check_descriptors failure) "
495 "unable to read group descriptors\n");
499 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) {
500 fs->s_inode_bitmap_number[i] = 0;
501 fs->s_inode_bitmap[i] = NULL;
502 fs->s_block_bitmap_number[i] = 0;
503 fs->s_block_bitmap[i] = NULL;
505 fs->s_loaded_inode_bitmaps = 0;
506 fs->s_loaded_block_bitmaps = 0;
511 * Reload all incore data for a filesystem (used after running fsck on
512 * the root filesystem and finding things to fix). The filesystem must
513 * be mounted read-only.
515 * Things to do to update the mount:
516 * 1) invalidate all cached meta-data.
517 * 2) re-read superblock from disk.
518 * 3) re-read summary information from disk.
519 * 4) invalidate all inactive vnodes.
520 * 5) invalidate all cached file data.
521 * 6) re-read inode data for all active vnodes.
524 ext2_reload(mountp, cred, td)
525 struct mount *mountp;
529 struct vnode *vp, *nvp, *devvp;
532 struct ext2_super_block * es;
533 struct ext2_sb_info *fs;
537 if ((mountp->mnt_flag & MNT_RDONLY) == 0)
540 * Step 1: invalidate all cached meta-data.
542 devvp = VFSTOUFS(mountp)->um_devvp;
543 if (vinvalbuf(devvp, 0, td, 0, 0))
544 panic("ext2_reload: dirty1");
546 * Step 2: re-read superblock from disk.
547 * constants have been adjusted for ext2
549 if ((error = bread(devvp, SBLOCK, SBSIZE, &bp)) != 0)
551 es = (struct ext2_super_block *)bp->b_data;
552 if (ext2_check_sb_compat(es, devvp->v_rdev, 0) != 0) {
554 return (EIO); /* XXX needs translation */
556 fs = VFSTOUFS(mountp)->um_e2fs;
557 bcopy(bp->b_data, fs->s_es, sizeof(struct ext2_super_block));
559 if((error = compute_sb_data(devvp, es, fs)) != 0) {
564 if (fs->fs_sbsize < SBSIZE)
565 bp->b_flags |= B_INVAL;
569 gen = lwkt_gettoken(&mntvnode_token);
571 for (vp = TAILQ_FIRST(&mountp->mnt_nvnodelist); vp != NULL; vp = nvp) {
572 KKASSERT(vp->v_mount == mountp);
573 nvp = TAILQ_NEXT(vp, v_nmntvnodes);
575 * Step 4: invalidate all inactive vnodes.
577 if (vrecycle(vp, NULL, td)) {
578 lwkt_gentoken(&mntvnode_token, &gen);
582 * Step 5: invalidate all cached file data.
584 lwkt_gettoken(&vp->v_interlock);
585 if (lwkt_gentoken(&mntvnode_token, &gen)) {
586 lwkt_reltoken(&vp->v_interlock);
589 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td)) {
590 lwkt_gentoken(&mntvnode_token, &gen);
593 if (vinvalbuf(vp, 0, td, 0, 0))
594 panic("ext2_reload: dirty2");
596 * Step 6: re-read inode data for all active vnodes.
600 bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
601 (int)fs->s_blocksize, &bp);
604 lwkt_reltoken(&mntvnode_token);
607 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data +
608 EXT2_INODE_SIZE * ino_to_fsbo(fs, ip->i_number)),
612 if (lwkt_gentoken(&mntvnode_token, &gen))
615 lwkt_reltoken(&mntvnode_token);
620 * Common code for mount and mountroot
623 ext2_mountfs(devvp, mp, td)
628 struct ufsmount *ump;
630 struct ext2_sb_info *fs;
631 struct ext2_super_block * es;
632 dev_t dev = devvp->v_rdev;
633 struct partinfo dpart;
639 * Disallow multiple mounts of the same device.
640 * Disallow mounting of a device that is currently in use
641 * (except for root, which might share swap device for miniroot).
642 * Flush out any old buffers remaining from a previous use.
644 if ((error = vfs_mountedon(devvp)) != 0)
646 if (vcount(devvp) > 1 && devvp != rootvp)
648 if ((error = vinvalbuf(devvp, V_SAVE, td, 0, 0)) != 0)
651 /* turn on this to force it to be read-only */
652 mp->mnt_flag |= MNT_RDONLY;
655 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
656 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
657 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, td);
658 VOP_UNLOCK(devvp, 0, td);
661 if (devvp->v_rdev->si_iosize_max != 0)
662 mp->mnt_iosize_max = devvp->v_rdev->si_iosize_max;
663 if (mp->mnt_iosize_max > MAXPHYS)
664 mp->mnt_iosize_max = MAXPHYS;
665 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, td) != 0)
669 size = dpart.disklab->d_secsize;
674 if ((error = bread(devvp, SBLOCK, SBSIZE, &bp)) != 0)
676 es = (struct ext2_super_block *)bp->b_data;
677 if (ext2_check_sb_compat(es, dev, ronly) != 0) {
678 error = EINVAL; /* XXX needs translation */
681 if ((es->s_state & EXT2_VALID_FS) == 0 ||
682 (es->s_state & EXT2_ERROR_FS)) {
683 if (ronly || (mp->mnt_flag & MNT_FORCE)) {
685 "WARNING: Filesystem was not properly dismounted\n");
688 "WARNING: R/W mount denied. Filesystem is not clean - run fsck\n");
693 ump = bsd_malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
694 bzero((caddr_t)ump, sizeof *ump);
695 ump->um_malloctype = M_EXT2NODE;
696 ump->um_blkatoff = ext2_blkatoff;
697 ump->um_truncate = ext2_truncate;
698 ump->um_update = ext2_update;
699 ump->um_valloc = ext2_valloc;
700 ump->um_vfree = ext2_vfree;
701 /* I don't know whether this is the right strategy. Note that
702 we dynamically allocate both a ext2_sb_info and a ext2_super_block
703 while Linux keeps the super block in a locked buffer
705 ump->um_e2fs = bsd_malloc(sizeof(struct ext2_sb_info),
707 ump->um_e2fs->s_es = bsd_malloc(sizeof(struct ext2_super_block),
709 bcopy(es, ump->um_e2fs->s_es, (u_int)sizeof(struct ext2_super_block));
710 if ((error = compute_sb_data(devvp, ump->um_e2fs->s_es, ump->um_e2fs)))
713 * We don't free the group descriptors allocated by compute_sb_data()
714 * until ext2_unmount(). This is OK since the mount will succeed.
719 fs->s_rd_only = ronly; /* ronly is set according to mnt_flags */
720 /* if the fs is not mounted read-only, make sure the super block is
721 always written back on a sync()
723 fs->s_wasvalid = fs->s_es->s_state & EXT2_VALID_FS ? 1 : 0;
725 fs->s_dirt = 1; /* mark it modified */
726 fs->s_es->s_state &= ~EXT2_VALID_FS; /* set fs invalid */
728 mp->mnt_data = (qaddr_t)ump;
729 mp->mnt_stat.f_fsid.val[0] = dev2udev(dev);
730 mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
731 mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN;
732 mp->mnt_flag |= MNT_LOCAL;
735 ump->um_devvp = devvp;
736 /* setting those two parameters allows us to use
737 ufs_bmap w/o changse !
739 ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs);
740 ump->um_bptrtodb = fs->s_es->s_log_block_size + 1;
741 ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs);
742 for (i = 0; i < MAXQUOTAS; i++)
743 ump->um_quotas[i] = NULLVP;
744 devvp->v_specmountpoint = mp;
746 ext2_sbupdate(ump, MNT_WAIT);
751 (void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, td);
753 bsd_free(ump->um_e2fs->s_es, M_UFSMNT);
754 bsd_free(ump->um_e2fs, M_UFSMNT);
755 bsd_free(ump, M_UFSMNT);
756 mp->mnt_data = (qaddr_t)0;
762 * unmount system call
765 ext2_unmount(mp, mntflags, td)
770 struct ufsmount *ump;
771 struct ext2_sb_info *fs;
772 int error, flags, ronly, i;
775 if (mntflags & MNT_FORCE) {
776 if (mp->mnt_flag & MNT_ROOTFS)
780 if ((error = ext2_flushfiles(mp, flags, td)) != 0)
784 ronly = fs->s_rd_only;
787 fs->s_es->s_state |= EXT2_VALID_FS;
788 ext2_sbupdate(ump, MNT_WAIT);
791 /* release buffers containing group descriptors */
792 for(i = 0; i < fs->s_db_per_group; i++)
793 ULCK_BUF(fs->s_group_desc[i])
794 bsd_free(fs->s_group_desc, M_UFSMNT);
796 /* release cached inode/block bitmaps */
797 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
798 if (fs->s_inode_bitmap[i])
799 ULCK_BUF(fs->s_inode_bitmap[i])
801 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
802 if (fs->s_block_bitmap[i])
803 ULCK_BUF(fs->s_block_bitmap[i])
805 ump->um_devvp->v_specmountpoint = NULL;
806 error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD|FWRITE, td);
807 vrele(ump->um_devvp);
808 bsd_free(fs->s_es, M_UFSMNT);
809 bsd_free(fs, M_UFSMNT);
810 bsd_free(ump, M_UFSMNT);
811 mp->mnt_data = (qaddr_t)0;
812 mp->mnt_flag &= ~MNT_LOCAL;
817 * Flush out all the files in a filesystem.
820 ext2_flushfiles(mp, flags, td)
825 struct ufsmount *ump;
833 if (mp->mnt_flag & MNT_QUOTA) {
834 if ((error = vflush(mp, 0, SKIPSYSTEM|flags)) != 0)
836 for (i = 0; i < MAXQUOTAS; i++) {
837 if (ump->um_quotas[i] == NULLVP)
842 * Here we fall through to vflush again to ensure
843 * that we have gotten rid of all the system vnodes.
847 error = vflush(mp, 0, flags);
852 * Get file system statistics.
853 * taken from ext2/super.c ext2_statfs
856 ext2_statfs(mp, sbp, td)
861 unsigned long overhead;
862 struct ufsmount *ump;
863 struct ext2_sb_info *fs;
864 struct ext2_super_block *es;
871 if (es->s_magic != EXT2_SUPER_MAGIC)
872 panic("ext2_statfs - magic number spoiled");
875 * Compute the overhead (FS structures)
877 if (es->s_feature_ro_compat & EXT2_FEATURE_RO_COMPAT_SPARSE_SUPER) {
879 for (i = 0 ; i < fs->s_groups_count; i++)
880 if (ext2_group_sparse(i))
883 nsb = fs->s_groups_count;
884 overhead = es->s_first_data_block +
885 /* Superblocks and block group descriptors: */
886 nsb * (1 + fs->s_db_per_group) +
887 /* Inode bitmap, block bitmap, and inode table: */
888 fs->s_groups_count * (1 + 1 + fs->s_itb_per_group);
890 sbp->f_bsize = EXT2_FRAG_SIZE(fs);
891 sbp->f_iosize = EXT2_BLOCK_SIZE(fs);
892 sbp->f_blocks = es->s_blocks_count - overhead;
893 sbp->f_bfree = es->s_free_blocks_count;
894 sbp->f_bavail = sbp->f_bfree - es->s_r_blocks_count;
895 sbp->f_files = es->s_inodes_count;
896 sbp->f_ffree = es->s_free_inodes_count;
897 if (sbp != &mp->mnt_stat) {
898 sbp->f_type = mp->mnt_vfc->vfc_typenum;
899 bcopy((caddr_t)mp->mnt_stat.f_mntonname,
900 (caddr_t)&sbp->f_mntonname[0], MNAMELEN);
901 bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
902 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
908 * Go through the disk queues to initiate sandbagged IO;
909 * go through the inodes to write those that have been modified;
910 * initiate the writing of the super block if it has been modified.
912 * Note: we are always called with the filesystem marked `MPBUSY'.
915 ext2_sync(mp, waitfor, td)
920 struct vnode *nvp, *vp;
922 struct ufsmount *ump = VFSTOUFS(mp);
923 struct ext2_sb_info *fs;
924 int error, allerror = 0;
928 if (fs->s_dirt != 0 && fs->s_rd_only != 0) { /* XXX */
929 printf("fs = %s\n", fs->fs_fsmnt);
930 panic("ext2_sync: rofs mod");
933 * Write back each (modified) inode.
935 gen = lwkt_gettoken(&mntvnode_token);
937 for (vp = TAILQ_FIRST(&mp->mnt_nvnodelist); vp != NULL; vp = nvp) {
939 * If the vnode that we are about to sync is no longer
940 * associated with this mount point, start over.
942 if (vp->v_mount != mp) {
943 lwkt_gentoken(&mntvnode_token, &gen);
946 lwkt_gettoken(&vp->v_interlock);
947 if (lwkt_gentoken(&mntvnode_token, &gen)) {
948 lwkt_reltoken(&vp->v_interlock);
951 nvp = TAILQ_NEXT(vp, v_nmntvnodes);
953 if (vp->v_type == VNON ||
955 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
956 (TAILQ_EMPTY(&vp->v_dirtyblkhd) || waitfor == MNT_LAZY))) {
957 lwkt_reltoken(&vp->v_interlock);
960 error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK, td);
966 if ((error = VOP_FSYNC(vp, waitfor, td)) != 0)
968 VOP_UNLOCK(vp, 0, td);
970 lwkt_gentoken(&mntvnode_token, &gen);
972 lwkt_reltoken(&mntvnode_token);
974 * Force stale file system control information to be flushed.
976 if (waitfor != MNT_LAZY) {
977 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY, td);
978 if ((error = VOP_FSYNC(ump->um_devvp, waitfor, td)) != 0)
980 VOP_UNLOCK(ump->um_devvp, 0, td);
986 * Write back modified superblock.
988 if (fs->s_dirt != 0) {
990 fs->s_es->s_wtime = time_second;
991 if ((error = ext2_sbupdate(ump, waitfor)) != 0)
998 * Look up a EXT2FS dinode number to find its incore vnode, otherwise read it
999 * in from disk. If it is in core, wait for the lock bit to clear, then
1000 * return the inode locked. Detection and handling of mount points must be
1001 * done by the calling routine.
1004 ext2_vget(mp, ino, vpp)
1009 struct ext2_sb_info *fs;
1011 struct ufsmount *ump;
1021 if ((*vpp = ufs_ihashget(dev, ino)) != NULL)
1025 * Lock out the creation of new entries in the FFS hash table in
1026 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate
1029 if (ext2fs_inode_hash_lock) {
1030 while (ext2fs_inode_hash_lock) {
1031 ext2fs_inode_hash_lock = -1;
1032 tsleep(&ext2fs_inode_hash_lock, 0, "e2vget", 0);
1036 ext2fs_inode_hash_lock = 1;
1039 * If this MALLOC() is performed after the getnewvnode()
1040 * it might block, leaving a vnode with a NULL v_data to be
1041 * found by ext2_sync() if a sync happens to fire right then,
1042 * which will cause a panic because ext2_sync() blindly
1043 * dereferences vp->v_data (as well it should).
1045 MALLOC(ip, struct inode *, sizeof(struct inode), M_EXT2NODE, M_WAITOK);
1047 /* Allocate a new vnode/inode. */
1048 if ((error = getnewvnode(VT_UFS, mp, ext2_vnodeop_p, &vp)) != 0) {
1049 if (ext2fs_inode_hash_lock < 0)
1050 wakeup(&ext2fs_inode_hash_lock);
1051 ext2fs_inode_hash_lock = 0;
1053 FREE(ip, M_EXT2NODE);
1056 bzero((caddr_t)ip, sizeof(struct inode));
1057 lockinit(&ip->i_lock, 0, "ext2in", 0, 0);
1060 ip->i_e2fs = fs = ump->um_e2fs;
1064 for (i = 0; i < MAXQUOTAS; i++)
1065 ip->i_dquot[i] = NODQUOT;
1068 * Put it onto its hash chain and lock it so that other requests for
1069 * this inode will block if they arrive while we are sleeping waiting
1070 * for old data structures to be purged or for the contents of the
1071 * disk portion of this inode to be read.
1075 if (ext2fs_inode_hash_lock < 0)
1076 wakeup(&ext2fs_inode_hash_lock);
1077 ext2fs_inode_hash_lock = 0;
1079 /* Read in the disk contents for the inode, copy into the inode. */
1081 printf("ext2_vget(%d) dbn= %d ", ino, fsbtodb(fs, ino_to_fsba(fs, ino)));
1083 if ((error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
1084 (int)fs->s_blocksize, &bp)) != 0) {
1086 * The inode does not contain anything useful, so it would
1087 * be misleading to leave it on its hash chain. With mode
1088 * still zero, it will be unlinked and returned to the free
1096 /* convert ext2 inode to dinode */
1097 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data + EXT2_INODE_SIZE *
1098 ino_to_fsbo(fs, ino)), &ip->i_din);
1099 ip->i_block_group = ino_to_cg(fs, ino);
1100 ip->i_next_alloc_block = 0;
1101 ip->i_next_alloc_goal = 0;
1102 ip->i_prealloc_count = 0;
1103 ip->i_prealloc_block = 0;
1104 /* now we want to make sure that block pointers for unused
1105 blocks are zeroed out - ext2_balloc depends on this
1106 although for regular files and directories only
1108 if(S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode)) {
1109 used_blocks = (ip->i_size+fs->s_blocksize-1) / fs->s_blocksize;
1110 for(i = used_blocks; i < EXT2_NDIR_BLOCKS; i++)
1114 ext2_print_inode(ip);
1119 * Initialize the vnode from the inode, check for aliases.
1120 * Note that the underlying vnode may have changed.
1122 if ((error = ufs_vinit(mp, ext2_specop_p, ext2_fifoop_p, &vp)) != 0) {
1128 * Finish inode initialization now that aliasing has been resolved.
1130 ip->i_devvp = ump->um_devvp;
1133 * Set up a generation number for this inode if it does not
1134 * already have one. This should only happen on old filesystems.
1136 if (ip->i_gen == 0) {
1137 ip->i_gen = random() / 2 + 1;
1138 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1139 ip->i_flag |= IN_MODIFIED;
1146 * File handle to vnode
1148 * Have to be really careful about stale file handles:
1149 * - check that the inode number is valid
1150 * - call ext2_vget() to get the locked inode
1151 * - check for an unallocated inode (i_mode == 0)
1152 * - check that the given client host has export rights and return
1153 * those rights via. exflagsp and credanonp
1156 ext2_fhtovp(mp, fhp, vpp)
1162 struct ext2_sb_info *fs;
1164 ufhp = (struct ufid *)fhp;
1165 fs = VFSTOUFS(mp)->um_e2fs;
1166 if (ufhp->ufid_ino < ROOTINO ||
1167 ufhp->ufid_ino > fs->s_groups_count * fs->s_es->s_inodes_per_group)
1169 return (ufs_fhtovp(mp, ufhp, vpp));
1173 * Vnode pointer to File handle
1177 ext2_vptofh(vp, fhp)
1185 ufhp = (struct ufid *)fhp;
1186 ufhp->ufid_len = sizeof(struct ufid);
1187 ufhp->ufid_ino = ip->i_number;
1188 ufhp->ufid_gen = ip->i_gen;
1193 * Write a superblock and associated information back to disk.
1196 ext2_sbupdate(mp, waitfor)
1197 struct ufsmount *mp;
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