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 1.14 2004/04/24 04:32:03 drhodus 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(struct mount *mp, char *path,
183 caddr_t data, /* this is actually a (struct ufs_args *) */
184 struct nameidata *ndp, struct thread *td)
187 struct ufs_args args;
188 struct ufsmount *ump = 0;
189 struct ext2_sb_info *fs;
195 if ((error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args))) != 0)
198 cred = td->td_proc->p_ucred;
200 * If updating, check whether changing from read-only to
201 * read/write; if there is no device name, that's all we do.
203 if (mp->mnt_flag & MNT_UPDATE) {
207 if (fs->s_rd_only == 0 && (mp->mnt_flag & MNT_RDONLY)) {
209 if (mp->mnt_flag & MNT_FORCE)
211 if (vfs_busy(mp, LK_NOWAIT, NULL, td))
213 error = ext2_flushfiles(mp, flags, td);
215 if (!error && fs->s_wasvalid) {
216 fs->s_es->s_state |= EXT2_VALID_FS;
217 ext2_sbupdate(ump, MNT_WAIT);
221 if (!error && (mp->mnt_flag & MNT_RELOAD))
222 error = ext2_reload(mp, ndp->ni_cnd.cn_cred, td);
225 devvp = ump->um_devvp;
226 if (ext2_check_sb_compat(fs->s_es, devvp->v_rdev,
227 (mp->mnt_kern_flag & MNTK_WANTRDWR) == 0) != 0)
229 if (fs->s_rd_only && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
231 * If upgrade to read-write by non-root, then verify
232 * that user has necessary permissions on the device.
234 if (cred->cr_uid != 0) {
235 vn_lock(devvp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
236 if ((error = VOP_ACCESS(devvp, VREAD | VWRITE,
238 VOP_UNLOCK(devvp, NULL, 0, td);
241 VOP_UNLOCK(devvp, NULL, 0, td);
244 if ((fs->s_es->s_state & EXT2_VALID_FS) == 0 ||
245 (fs->s_es->s_state & EXT2_ERROR_FS)) {
246 if (mp->mnt_flag & MNT_FORCE) {
248 "WARNING: %s was not properly dismounted\n",
252 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
257 fs->s_es->s_state &= ~EXT2_VALID_FS;
258 ext2_sbupdate(ump, MNT_WAIT);
261 if (args.fspec == 0) {
263 * Process export requests.
265 return (vfs_export(mp, &ump->um_export, &args.export));
269 * Not an update, or updating the name: look up the name
270 * and verify that it refers to a sensible block device.
272 NDINIT(ndp, NAMEI_LOOKUP, CNP_FOLLOW, UIO_USERSPACE, args.fspec, td);
273 if ((error = namei(ndp)) != 0)
275 NDFREE(ndp, NDF_ONLY_PNBUF);
278 if (!vn_isdisk(devvp, &error)) {
284 * If mount by non-root, then verify that user has necessary
285 * permissions on the device.
287 if (cred->cr_uid != 0) {
289 if ((mp->mnt_flag & MNT_RDONLY) == 0)
290 accessmode |= VWRITE;
291 vn_lock(devvp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
292 if ((error = VOP_ACCESS(devvp, accessmode, cred, td)) != 0) {
296 VOP_UNLOCK(devvp, NULL, 0, td);
299 if ((mp->mnt_flag & MNT_UPDATE) == 0) {
300 error = ext2_mountfs(devvp, mp, td);
302 if (devvp != ump->um_devvp)
303 error = EINVAL; /* needs translation */
313 (void) copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size);
314 bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size);
315 bcopy((caddr_t)fs->fs_fsmnt, (caddr_t)mp->mnt_stat.f_mntonname,
317 (void) copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
319 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
320 (void)ext2_statfs(mp, &mp->mnt_stat, td);
325 * checks that the data in the descriptor blocks make sense
326 * this is taken from ext2/super.c
328 static int ext2_check_descriptors(struct ext2_sb_info *sb)
332 unsigned long block = sb->s_es->s_first_data_block;
333 struct ext2_group_desc * gdp = NULL;
335 /* ext2_debug ("Checking group descriptors"); */
337 for (i = 0; i < sb->s_groups_count; i++)
339 /* examine next descriptor block */
340 if ((i % EXT2_DESC_PER_BLOCK(sb)) == 0)
341 gdp = (struct ext2_group_desc *)
342 sb->s_group_desc[desc_block++]->b_data;
343 if (gdp->bg_block_bitmap < block ||
344 gdp->bg_block_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb))
346 printf ("ext2_check_descriptors: "
347 "Block bitmap for group %d"
348 " not in group (block %lu)!\n",
349 i, (unsigned long) gdp->bg_block_bitmap);
352 if (gdp->bg_inode_bitmap < block ||
353 gdp->bg_inode_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb))
355 printf ("ext2_check_descriptors: "
356 "Inode bitmap for group %d"
357 " not in group (block %lu)!\n",
358 i, (unsigned long) gdp->bg_inode_bitmap);
361 if (gdp->bg_inode_table < block ||
362 gdp->bg_inode_table + sb->s_itb_per_group >=
363 block + EXT2_BLOCKS_PER_GROUP(sb))
365 printf ("ext2_check_descriptors: "
366 "Inode table for group %d"
367 " not in group (block %lu)!\n",
368 i, (unsigned long) gdp->bg_inode_table);
371 block += EXT2_BLOCKS_PER_GROUP(sb);
378 ext2_check_sb_compat(struct ext2_super_block *es, dev_t dev, int ronly)
380 if (es->s_magic != EXT2_SUPER_MAGIC) {
381 printf("ext2fs: %s: wrong magic number %#x (expected %#x)\n",
382 devtoname(dev), es->s_magic, EXT2_SUPER_MAGIC);
385 if (es->s_rev_level > EXT2_GOOD_OLD_REV) {
386 if (es->s_feature_incompat & ~EXT2_FEATURE_INCOMPAT_SUPP) {
388 "WARNING: mount of %s denied due to unsupported optional features\n",
393 (es->s_feature_ro_compat & ~EXT2_FEATURE_RO_COMPAT_SUPP)) {
395 "WARNING: R/W mount of %s denied due to unsupported optional features\n",
404 * this computes the fields of the ext2_sb_info structure from the
405 * data in the ext2_super_block structure read in
408 compute_sb_data(struct vnode *devvp, struct ext2_super_block *es,
409 struct ext2_sb_info *fs)
413 int logic_sb_block = 1; /* XXX for now */
418 #define V(v) printf(#v"= %d\n", fs->v);
421 fs->s_blocksize = EXT2_MIN_BLOCK_SIZE << es->s_log_block_size;
423 fs->s_bshift = EXT2_MIN_BLOCK_LOG_SIZE + es->s_log_block_size;
425 fs->s_fsbtodb = es->s_log_block_size + 1;
427 fs->s_qbmask = fs->s_blocksize - 1;
429 fs->s_blocksize_bits = EXT2_BLOCK_SIZE_BITS(es);
431 fs->s_frag_size = EXT2_MIN_FRAG_SIZE << es->s_log_frag_size;
434 fs->s_frags_per_block = fs->s_blocksize / fs->s_frag_size;
436 fs->s_blocks_per_group = es->s_blocks_per_group;
437 V(s_blocks_per_group)
438 fs->s_frags_per_group = es->s_frags_per_group;
440 fs->s_inodes_per_group = es->s_inodes_per_group;
441 V(s_inodes_per_group)
442 fs->s_inodes_per_block = fs->s_blocksize / EXT2_INODE_SIZE;
443 V(s_inodes_per_block)
444 fs->s_itb_per_group = fs->s_inodes_per_group /fs->s_inodes_per_block;
446 fs->s_desc_per_block = fs->s_blocksize / sizeof (struct ext2_group_desc);
448 /* s_resuid / s_resgid ? */
449 fs->s_groups_count = (es->s_blocks_count -
450 es->s_first_data_block +
451 EXT2_BLOCKS_PER_GROUP(fs) - 1) /
452 EXT2_BLOCKS_PER_GROUP(fs);
454 db_count = (fs->s_groups_count + EXT2_DESC_PER_BLOCK(fs) - 1) /
455 EXT2_DESC_PER_BLOCK(fs);
456 fs->s_db_per_group = db_count;
459 fs->s_group_desc = bsd_malloc(db_count * sizeof (struct buf *),
462 /* adjust logic_sb_block */
463 if(fs->s_blocksize > SBSIZE)
464 /* Godmar thinks: if the blocksize is greater than 1024, then
465 the superblock is logically part of block zero.
469 for (i = 0; i < db_count; i++) {
470 error = bread(devvp , fsbtodb(fs, logic_sb_block + i + 1),
471 fs->s_blocksize, &fs->s_group_desc[i]);
473 for (j = 0; j < i; j++)
474 brelse(fs->s_group_desc[j]);
475 bsd_free(fs->s_group_desc, M_UFSMNT);
476 printf("EXT2-fs: unable to read group descriptors (%d)\n", error);
479 /* Set the B_LOCKED flag on the buffer, then brelse() it */
480 LCK_BUF(fs->s_group_desc[i])
482 if(!ext2_check_descriptors(fs)) {
483 for (j = 0; j < db_count; j++)
484 ULCK_BUF(fs->s_group_desc[j])
485 bsd_free(fs->s_group_desc, M_UFSMNT);
486 printf("EXT2-fs: (ext2_check_descriptors failure) "
487 "unable to read group descriptors\n");
491 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) {
492 fs->s_inode_bitmap_number[i] = 0;
493 fs->s_inode_bitmap[i] = NULL;
494 fs->s_block_bitmap_number[i] = 0;
495 fs->s_block_bitmap[i] = NULL;
497 fs->s_loaded_inode_bitmaps = 0;
498 fs->s_loaded_block_bitmaps = 0;
503 * Reload all incore data for a filesystem (used after running fsck on
504 * the root filesystem and finding things to fix). The filesystem must
505 * be mounted read-only.
507 * Things to do to update the mount:
508 * 1) invalidate all cached meta-data.
509 * 2) re-read superblock from disk.
510 * 3) re-read summary information from disk.
511 * 4) invalidate all inactive vnodes.
512 * 5) invalidate all cached file data.
513 * 6) re-read inode data for all active vnodes.
515 static int ext2_reload_scan1(struct mount *mp, struct vnode *vp, void *rescan);
516 static int ext2_reload_scan2(struct mount *mp, struct vnode *vp,
517 lwkt_tokref_t vlock, void *rescan);
525 struct ext2_sb_info *fs;
529 ext2_reload(struct mount *mountp, struct ucred *cred, struct thread *td)
533 struct ext2_super_block * es;
534 struct ext2_sb_info *fs;
536 struct scaninfo scaninfo;
538 if ((mountp->mnt_flag & MNT_RDONLY) == 0)
541 * Step 1: invalidate all cached meta-data.
543 devvp = VFSTOUFS(mountp)->um_devvp;
544 if (vinvalbuf(devvp, 0, td, 0, 0))
545 panic("ext2_reload: dirty1");
547 * Step 2: re-read superblock from disk.
548 * constants have been adjusted for ext2
550 if ((error = bread(devvp, SBLOCK, SBSIZE, &bp)) != 0)
552 es = (struct ext2_super_block *)bp->b_data;
553 if (ext2_check_sb_compat(es, devvp->v_rdev, 0) != 0) {
555 return (EIO); /* XXX needs translation */
557 fs = VFSTOUFS(mountp)->um_e2fs;
558 bcopy(bp->b_data, fs->s_es, sizeof(struct ext2_super_block));
560 if((error = compute_sb_data(devvp, es, fs)) != 0) {
565 if (fs->fs_sbsize < SBSIZE)
566 bp->b_flags |= B_INVAL;
572 scaninfo.devvp = devvp;
574 while (error == 0 && scaninfo.rescan) {
576 error = vmntvnodescan(mountp, ext2_reload_scan1,
577 ext2_reload_scan2, &scaninfo);
583 ext2_reload_scan1(struct mount *mp, struct vnode *vp, void *data)
585 struct scaninfo *info = data;
588 * Step 4: invalidate all inactive vnodes.
590 if (vrecycle(vp, NULL, curthread)) {
592 return(-1); /* continue loop, do not call scan2 */
598 ext2_reload_scan2(struct mount *mp, struct vnode *vp, lwkt_tokref_t vlock, void *data)
600 struct scaninfo *info = data;
606 * Step 5: invalidate all cached file data.
608 if (vget(vp, vlock, LK_EXCLUSIVE | LK_INTERLOCK, info->td)) {
612 if (vinvalbuf(vp, 0, info->td, 0, 0))
613 panic("ext2_reload: dirty2");
615 * Step 6: re-read inode data for all active vnodes.
618 error = bread(info->devvp, fsbtodb(info->fs, ino_to_fsba(info->fs, ip->i_number)),
619 (int)info->fs->s_blocksize, &bp);
624 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data +
625 EXT2_INODE_SIZE * ino_to_fsbo(info->fs, ip->i_number)),
633 * Common code for mount and mountroot
636 ext2_mountfs(struct vnode *devvp, struct mount *mp, struct thread *td)
638 struct ufsmount *ump;
640 struct ext2_sb_info *fs;
641 struct ext2_super_block * es;
642 dev_t dev = devvp->v_rdev;
643 struct partinfo dpart;
649 * Disallow multiple mounts of the same device.
650 * Disallow mounting of a device that is currently in use
651 * (except for root, which might share swap device for miniroot).
652 * Flush out any old buffers remaining from a previous use.
654 if ((error = vfs_mountedon(devvp)) != 0)
656 if (vcount(devvp) > 1 && devvp != rootvp)
658 if ((error = vinvalbuf(devvp, V_SAVE, td, 0, 0)) != 0)
661 /* turn on this to force it to be read-only */
662 mp->mnt_flag |= MNT_RDONLY;
665 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
666 vn_lock(devvp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
667 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, td);
668 VOP_UNLOCK(devvp, NULL, 0, td);
671 if (devvp->v_rdev->si_iosize_max != 0)
672 mp->mnt_iosize_max = devvp->v_rdev->si_iosize_max;
673 if (mp->mnt_iosize_max > MAXPHYS)
674 mp->mnt_iosize_max = MAXPHYS;
675 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, td) != 0)
679 size = dpart.disklab->d_secsize;
684 if ((error = bread(devvp, SBLOCK, SBSIZE, &bp)) != 0)
686 es = (struct ext2_super_block *)bp->b_data;
687 if (ext2_check_sb_compat(es, dev, ronly) != 0) {
688 error = EINVAL; /* XXX needs translation */
691 if ((es->s_state & EXT2_VALID_FS) == 0 ||
692 (es->s_state & EXT2_ERROR_FS)) {
693 if (ronly || (mp->mnt_flag & MNT_FORCE)) {
695 "WARNING: Filesystem was not properly dismounted\n");
698 "WARNING: R/W mount denied. Filesystem is not clean - run fsck\n");
703 ump = bsd_malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
704 bzero((caddr_t)ump, sizeof *ump);
705 ump->um_malloctype = M_EXT2NODE;
706 ump->um_blkatoff = ext2_blkatoff;
707 ump->um_truncate = ext2_truncate;
708 ump->um_update = ext2_update;
709 ump->um_valloc = ext2_valloc;
710 ump->um_vfree = ext2_vfree;
711 /* I don't know whether this is the right strategy. Note that
712 we dynamically allocate both a ext2_sb_info and a ext2_super_block
713 while Linux keeps the super block in a locked buffer
715 ump->um_e2fs = bsd_malloc(sizeof(struct ext2_sb_info),
717 ump->um_e2fs->s_es = bsd_malloc(sizeof(struct ext2_super_block),
719 bcopy(es, ump->um_e2fs->s_es, (u_int)sizeof(struct ext2_super_block));
720 if ((error = compute_sb_data(devvp, ump->um_e2fs->s_es, ump->um_e2fs)))
723 * We don't free the group descriptors allocated by compute_sb_data()
724 * until ext2_unmount(). This is OK since the mount will succeed.
729 fs->s_rd_only = ronly; /* ronly is set according to mnt_flags */
730 /* if the fs is not mounted read-only, make sure the super block is
731 always written back on a sync()
733 fs->s_wasvalid = fs->s_es->s_state & EXT2_VALID_FS ? 1 : 0;
735 fs->s_dirt = 1; /* mark it modified */
736 fs->s_es->s_state &= ~EXT2_VALID_FS; /* set fs invalid */
738 mp->mnt_data = (qaddr_t)ump;
739 mp->mnt_stat.f_fsid.val[0] = dev2udev(dev);
740 mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
741 mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN;
742 mp->mnt_flag |= MNT_LOCAL;
745 ump->um_devvp = devvp;
746 /* setting those two parameters allows us to use
747 ufs_bmap w/o changse !
749 ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs);
750 ump->um_bptrtodb = fs->s_es->s_log_block_size + 1;
751 ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs);
752 for (i = 0; i < MAXQUOTAS; i++)
753 ump->um_quotas[i] = NULLVP;
754 devvp->v_specmountpoint = mp;
756 ext2_sbupdate(ump, MNT_WAIT);
761 (void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, td);
763 bsd_free(ump->um_e2fs->s_es, M_UFSMNT);
764 bsd_free(ump->um_e2fs, M_UFSMNT);
765 bsd_free(ump, M_UFSMNT);
766 mp->mnt_data = (qaddr_t)0;
772 * unmount system call
775 ext2_unmount(struct mount *mp, int mntflags, struct thread *td)
777 struct ufsmount *ump;
778 struct ext2_sb_info *fs;
779 int error, flags, ronly, i;
782 if (mntflags & MNT_FORCE) {
783 if (mp->mnt_flag & MNT_ROOTFS)
787 if ((error = ext2_flushfiles(mp, flags, td)) != 0)
791 ronly = fs->s_rd_only;
794 fs->s_es->s_state |= EXT2_VALID_FS;
795 ext2_sbupdate(ump, MNT_WAIT);
798 /* release buffers containing group descriptors */
799 for(i = 0; i < fs->s_db_per_group; i++)
800 ULCK_BUF(fs->s_group_desc[i])
801 bsd_free(fs->s_group_desc, M_UFSMNT);
803 /* release cached inode/block bitmaps */
804 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
805 if (fs->s_inode_bitmap[i])
806 ULCK_BUF(fs->s_inode_bitmap[i])
808 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
809 if (fs->s_block_bitmap[i])
810 ULCK_BUF(fs->s_block_bitmap[i])
812 ump->um_devvp->v_specmountpoint = NULL;
813 error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD|FWRITE, td);
814 vrele(ump->um_devvp);
815 bsd_free(fs->s_es, M_UFSMNT);
816 bsd_free(fs, M_UFSMNT);
817 bsd_free(ump, M_UFSMNT);
818 mp->mnt_data = (qaddr_t)0;
819 mp->mnt_flag &= ~MNT_LOCAL;
824 * Flush out all the files in a filesystem.
827 ext2_flushfiles(struct mount *mp, int flags, struct thread *td)
829 struct ufsmount *ump;
837 if (mp->mnt_flag & MNT_QUOTA) {
838 if ((error = vflush(mp, 0, SKIPSYSTEM|flags)) != 0)
840 for (i = 0; i < MAXQUOTAS; i++) {
841 if (ump->um_quotas[i] == NULLVP)
846 * Here we fall through to vflush again to ensure
847 * that we have gotten rid of all the system vnodes.
851 error = vflush(mp, 0, flags);
856 * Get file system statistics.
857 * taken from ext2/super.c ext2_statfs
860 ext2_statfs(struct mount *mp, struct statfs *sbp, struct thread *td)
862 unsigned long overhead;
863 struct ufsmount *ump;
864 struct ext2_sb_info *fs;
865 struct ext2_super_block *es;
872 if (es->s_magic != EXT2_SUPER_MAGIC)
873 panic("ext2_statfs - magic number spoiled");
876 * Compute the overhead (FS structures)
878 if (es->s_feature_ro_compat & EXT2_FEATURE_RO_COMPAT_SPARSE_SUPER) {
880 for (i = 0 ; i < fs->s_groups_count; i++)
881 if (ext2_group_sparse(i))
884 nsb = fs->s_groups_count;
885 overhead = es->s_first_data_block +
886 /* Superblocks and block group descriptors: */
887 nsb * (1 + fs->s_db_per_group) +
888 /* Inode bitmap, block bitmap, and inode table: */
889 fs->s_groups_count * (1 + 1 + fs->s_itb_per_group);
891 sbp->f_bsize = EXT2_FRAG_SIZE(fs);
892 sbp->f_iosize = EXT2_BLOCK_SIZE(fs);
893 sbp->f_blocks = es->s_blocks_count - overhead;
894 sbp->f_bfree = es->s_free_blocks_count;
895 sbp->f_bavail = sbp->f_bfree - es->s_r_blocks_count;
896 sbp->f_files = es->s_inodes_count;
897 sbp->f_ffree = es->s_free_inodes_count;
898 if (sbp != &mp->mnt_stat) {
899 sbp->f_type = mp->mnt_vfc->vfc_typenum;
900 bcopy((caddr_t)mp->mnt_stat.f_mntonname,
901 (caddr_t)&sbp->f_mntonname[0], MNAMELEN);
902 bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
903 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
909 * Go through the disk queues to initiate sandbagged IO;
910 * go through the inodes to write those that have been modified;
911 * initiate the writing of the super block if it has been modified.
913 * Note: we are always called with the filesystem marked `MPBUSY'.
916 static int ext2_sync_scan(struct mount *mp, struct vnode *vp,
917 lwkt_tokref_t vlock, void *data);
920 ext2_sync(struct mount *mp, int waitfor, struct thread *td)
922 struct ufsmount *ump = VFSTOUFS(mp);
923 struct ext2_sb_info *fs;
924 struct scaninfo scaninfo;
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");
934 * Write back each (modified) inode.
936 scaninfo.allerror = 0;
938 scaninfo.waitfor = waitfor;
940 while (scaninfo.rescan) {
942 vmntvnodescan(mp, NULL, ext2_sync_scan, &scaninfo);
946 * Force stale file system control information to be flushed.
948 if (waitfor != MNT_LAZY) {
949 vn_lock(ump->um_devvp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
950 if ((error = VOP_FSYNC(ump->um_devvp, waitfor, td)) != 0)
951 scaninfo.allerror = error;
952 VOP_UNLOCK(ump->um_devvp, NULL, 0, td);
958 * Write back modified superblock.
960 if (fs->s_dirt != 0) {
962 fs->s_es->s_wtime = time_second;
963 if ((error = ext2_sbupdate(ump, waitfor)) != 0)
964 scaninfo.allerror = error;
966 return (scaninfo.allerror);
970 ext2_sync_scan(struct mount *mp, struct vnode *vp,
971 lwkt_tokref_t vlock, void *data)
973 struct scaninfo *info = data;
978 if (vp->v_type == VNON ||
980 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
981 (TAILQ_EMPTY(&vp->v_dirtyblkhd) || info->waitfor == MNT_LAZY))) {
982 lwkt_reltoken(vlock);
985 error = vget(vp, vlock, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK, info->td);
991 if ((error = VOP_FSYNC(vp, info->waitfor, info->td)) != 0)
992 info->allerror = error;
993 VOP_UNLOCK(vp, NULL, 0, info->td);
999 * Look up a EXT2FS dinode number to find its incore vnode, otherwise read it
1000 * in from disk. If it is in core, wait for the lock bit to clear, then
1001 * return the inode locked. Detection and handling of mount points must be
1002 * done by the calling routine.
1005 ext2_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
1007 struct ext2_sb_info *fs;
1009 struct ufsmount *ump;
1019 if ((*vpp = ufs_ihashget(dev, ino)) != NULL)
1023 * Lock out the creation of new entries in the FFS hash table in
1024 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate
1027 if (ext2fs_inode_hash_lock) {
1028 while (ext2fs_inode_hash_lock) {
1029 ext2fs_inode_hash_lock = -1;
1030 tsleep(&ext2fs_inode_hash_lock, 0, "e2vget", 0);
1034 ext2fs_inode_hash_lock = 1;
1037 * If this MALLOC() is performed after the getnewvnode()
1038 * it might block, leaving a vnode with a NULL v_data to be
1039 * found by ext2_sync() if a sync happens to fire right then,
1040 * which will cause a panic because ext2_sync() blindly
1041 * dereferences vp->v_data (as well it should).
1043 MALLOC(ip, struct inode *, sizeof(struct inode), M_EXT2NODE, M_WAITOK);
1045 /* Allocate a new vnode/inode. */
1046 if ((error = getnewvnode(VT_UFS, mp, ext2_vnodeop_p, &vp)) != 0) {
1047 if (ext2fs_inode_hash_lock < 0)
1048 wakeup(&ext2fs_inode_hash_lock);
1049 ext2fs_inode_hash_lock = 0;
1051 FREE(ip, M_EXT2NODE);
1054 bzero((caddr_t)ip, sizeof(struct inode));
1055 lockinit(&ip->i_lock, 0, "ext2in", 0, 0);
1058 ip->i_e2fs = fs = ump->um_e2fs;
1062 for (i = 0; i < MAXQUOTAS; i++)
1063 ip->i_dquot[i] = NODQUOT;
1066 * Put it onto its hash chain and lock it so that other requests for
1067 * this inode will block if they arrive while we are sleeping waiting
1068 * for old data structures to be purged or for the contents of the
1069 * disk portion of this inode to be read.
1073 if (ext2fs_inode_hash_lock < 0)
1074 wakeup(&ext2fs_inode_hash_lock);
1075 ext2fs_inode_hash_lock = 0;
1077 /* Read in the disk contents for the inode, copy into the inode. */
1079 printf("ext2_vget(%d) dbn= %d ", ino, fsbtodb(fs, ino_to_fsba(fs, ino)));
1081 if ((error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
1082 (int)fs->s_blocksize, &bp)) != 0) {
1084 * The inode does not contain anything useful, so it would
1085 * be misleading to leave it on its hash chain. With mode
1086 * still zero, it will be unlinked and returned to the free
1094 /* convert ext2 inode to dinode */
1095 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data + EXT2_INODE_SIZE *
1096 ino_to_fsbo(fs, ino)), &ip->i_din);
1097 ip->i_block_group = ino_to_cg(fs, ino);
1098 ip->i_next_alloc_block = 0;
1099 ip->i_next_alloc_goal = 0;
1100 ip->i_prealloc_count = 0;
1101 ip->i_prealloc_block = 0;
1102 /* now we want to make sure that block pointers for unused
1103 blocks are zeroed out - ext2_balloc depends on this
1104 although for regular files and directories only
1106 if(S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode)) {
1107 used_blocks = (ip->i_size+fs->s_blocksize-1) / fs->s_blocksize;
1108 for(i = used_blocks; i < EXT2_NDIR_BLOCKS; i++)
1112 ext2_print_inode(ip);
1117 * Initialize the vnode from the inode, check for aliases.
1118 * Note that the underlying vnode may have changed.
1120 if ((error = ufs_vinit(mp, ext2_specop_p, ext2_fifoop_p, &vp)) != 0) {
1126 * Finish inode initialization now that aliasing has been resolved.
1128 ip->i_devvp = ump->um_devvp;
1131 * Set up a generation number for this inode if it does not
1132 * already have one. This should only happen on old filesystems.
1134 if (ip->i_gen == 0) {
1135 ip->i_gen = random() / 2 + 1;
1136 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1137 ip->i_flag |= IN_MODIFIED;
1144 * File handle to vnode
1146 * Have to be really careful about stale file handles:
1147 * - check that the inode number is valid
1148 * - call ext2_vget() to get the locked inode
1149 * - check for an unallocated inode (i_mode == 0)
1150 * - check that the given client host has export rights and return
1151 * those rights via. exflagsp and credanonp
1154 ext2_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
1157 struct ext2_sb_info *fs;
1159 ufhp = (struct ufid *)fhp;
1160 fs = VFSTOUFS(mp)->um_e2fs;
1161 if (ufhp->ufid_ino < ROOTINO ||
1162 ufhp->ufid_ino > fs->s_groups_count * fs->s_es->s_inodes_per_group)
1164 return (ufs_fhtovp(mp, ufhp, vpp));
1168 * Vnode pointer to File handle
1172 ext2_vptofh(struct vnode *vp, struct fid *fhp)
1178 ufhp = (struct ufid *)fhp;
1179 ufhp->ufid_len = sizeof(struct ufid);
1180 ufhp->ufid_ino = ip->i_number;
1181 ufhp->ufid_gen = ip->i_gen;
1186 * Write a superblock and associated information back to disk.
1189 ext2_sbupdate(struct ufsmount *mp, int waitfor)
1191 struct ext2_sb_info *fs = mp->um_e2fs;
1192 struct ext2_super_block *es = fs->s_es;
1196 printf("\nupdating superblock, waitfor=%s\n", waitfor == MNT_WAIT ? "yes":"no");
1198 bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0);
1199 bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2_super_block));
1200 if (waitfor == MNT_WAIT)
1206 * The buffers for group descriptors, inode bitmaps and block bitmaps
1207 * are not busy at this point and are (hopefully) written by the
1208 * usual sync mechanism. No need to write them here