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
43 #include "opt_quota.h"
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/namei.h>
49 #include <sys/kernel.h>
50 #include <sys/vnode.h>
51 #include <sys/mount.h>
54 #include <sys/fcntl.h>
55 #include <sys/disklabel.h>
56 #include <sys/malloc.h>
59 #include <ufs/ufs/quota.h>
60 #include <ufs/ufs/ufsmount.h>
61 #include <ufs/ufs/inode.h>
62 #include <ufs/ufs/ufs_extern.h>
64 #include <vm/vm_zone.h>
66 #include <gnu/ext2fs/fs.h>
67 #include <gnu/ext2fs/ext2_extern.h>
68 #include <gnu/ext2fs/ext2_fs.h>
69 #include <gnu/ext2fs/ext2_fs_sb.h>
71 static int ext2_fhtovp __P((struct mount *, struct fid *, struct vnode **));
72 static int ext2_flushfiles __P((struct mount *mp, int flags, struct proc *p));
73 static int ext2_mount __P((struct mount *,
74 char *, caddr_t, struct nameidata *, struct proc *));
75 static int ext2_mountfs __P((struct vnode *, struct mount *, struct proc *));
76 static int ext2_reload __P((struct mount *mountp, struct ucred *cred,
78 static int ext2_sbupdate __P((struct ufsmount *, int));
79 static int ext2_statfs __P((struct mount *, struct statfs *, struct proc *));
80 static int ext2_sync __P((struct mount *, int, struct ucred *, struct proc *));
81 static int ext2_unmount __P((struct mount *, int, struct proc *));
82 static int ext2_vget __P((struct mount *, ino_t, struct vnode **));
83 static int ext2_vptofh __P((struct vnode *, struct fid *));
85 static MALLOC_DEFINE(M_EXT2NODE, "EXT2 node", "EXT2 vnode private part");
87 static struct vfsops ext2fs_vfsops = {
89 ufs_start, /* empty function */
91 ufs_root, /* root inode via vget */
92 ufs_quotactl, /* does operations associated with quotas */
104 VFS_SET(ext2fs_vfsops, ext2fs, 0);
105 #define bsd_malloc malloc
106 #define bsd_free free
108 static int ext2fs_inode_hash_lock;
110 static int ext2_check_sb_compat __P((struct ext2_super_block *es,
111 dev_t dev, int ronly));
112 static int compute_sb_data __P((struct vnode * devvp,
113 struct ext2_super_block * es,
114 struct ext2_sb_info * fs));
117 static int ext2_mountroot __P((void));
120 * Called by main() when ext2fs is going to be mounted as root.
122 * Name is updated by mount(8) after booting.
124 #define ROOTNAME "root_device"
129 register struct ext2_sb_info *fs;
130 register struct mount *mp;
131 struct proc *p = curproc;
132 struct ufsmount *ump;
136 if ((error = bdevvp(rootdev, &rootvp))) {
137 printf("ext2_mountroot: can't find rootvp\n");
140 mp = bsd_malloc((u_long)sizeof(struct mount), M_MOUNT, M_WAITOK);
141 bzero((char *)mp, (u_long)sizeof(struct mount));
142 mp->mnt_op = &ext2fs_vfsops;
143 mp->mnt_flag = MNT_RDONLY;
144 if (error = ext2_mountfs(rootvp, mp, p)) {
145 bsd_free(mp, M_MOUNT);
148 if (error = vfs_lock(mp)) {
149 (void)ext2_unmount(mp, 0, p);
150 bsd_free(mp, M_MOUNT);
153 TAILQ_INSERT_HEAD(&mountlist, mp, mnt_list);
154 mp->mnt_flag |= MNT_ROOTFS;
155 mp->mnt_vnodecovered = NULLVP;
158 bzero(fs->fs_fsmnt, sizeof(fs->fs_fsmnt));
159 fs->fs_fsmnt[0] = '/';
160 bcopy((caddr_t)fs->fs_fsmnt, (caddr_t)mp->mnt_stat.f_mntonname,
162 (void) copystr(ROOTNAME, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
164 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
165 (void)ext2_statfs(mp, &mp->mnt_stat, p);
167 inittodr(fs->s_es->s_wtime); /* this helps to set the time */
178 ext2_mount(mp, path, data, ndp, p)
179 register struct mount *mp;
181 caddr_t data; /* this is actually a (struct ufs_args *) */
182 struct nameidata *ndp;
186 struct ufs_args args;
187 struct ufsmount *ump = 0;
188 register struct ext2_sb_info *fs;
193 if ((error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args))) != 0)
196 * If updating, check whether changing from read-only to
197 * read/write; if there is no device name, that's all we do.
199 if (mp->mnt_flag & MNT_UPDATE) {
203 if (fs->s_rd_only == 0 && (mp->mnt_flag & MNT_RDONLY)) {
205 if (mp->mnt_flag & MNT_FORCE)
207 if (vfs_busy(mp, LK_NOWAIT, 0, p))
209 error = ext2_flushfiles(mp, flags, p);
211 if (!error && fs->s_wasvalid) {
212 fs->s_es->s_state |= EXT2_VALID_FS;
213 ext2_sbupdate(ump, MNT_WAIT);
217 if (!error && (mp->mnt_flag & MNT_RELOAD))
218 error = ext2_reload(mp, ndp->ni_cnd.cn_cred, p);
221 if (fs->s_rd_only && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
223 * If upgrade to read-write by non-root, then verify
224 * that user has necessary permissions on the device.
226 if (p->p_ucred->cr_uid != 0) {
227 devvp = ump->um_devvp;
228 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p);
229 if ((error = VOP_ACCESS(devvp, VREAD | VWRITE,
230 p->p_ucred, p)) != 0) {
231 VOP_UNLOCK(devvp, 0, p);
234 VOP_UNLOCK(devvp, 0, p);
237 if ((fs->s_es->s_state & EXT2_VALID_FS) == 0 ||
238 (fs->s_es->s_state & EXT2_ERROR_FS)) {
239 if (mp->mnt_flag & MNT_FORCE) {
241 "WARNING: %s was not properly dismounted\n",
245 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
250 fs->s_es->s_state &= ~EXT2_VALID_FS;
251 ext2_sbupdate(ump, MNT_WAIT);
254 if (args.fspec == 0) {
256 * Process export requests.
258 return (vfs_export(mp, &ump->um_export, &args.export));
262 * Not an update, or updating the name: look up the name
263 * and verify that it refers to a sensible block device.
265 NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, p);
266 if ((error = namei(ndp)) != 0)
268 NDFREE(ndp, NDF_ONLY_PNBUF);
271 if (!vn_isdisk(devvp, &error)) {
277 * If mount by non-root, then verify that user has necessary
278 * permissions on the device.
280 if (p->p_ucred->cr_uid != 0) {
282 if ((mp->mnt_flag & MNT_RDONLY) == 0)
283 accessmode |= VWRITE;
284 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p);
285 if ((error = VOP_ACCESS(devvp, accessmode, p->p_ucred, p)) != 0) {
289 VOP_UNLOCK(devvp, 0, p);
292 if ((mp->mnt_flag & MNT_UPDATE) == 0) {
293 error = ext2_mountfs(devvp, mp, p);
295 if (devvp != ump->um_devvp)
296 error = EINVAL; /* needs translation */
306 (void) copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size);
307 bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size);
308 bcopy((caddr_t)fs->fs_fsmnt, (caddr_t)mp->mnt_stat.f_mntonname,
310 (void) copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
312 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
313 (void)ext2_statfs(mp, &mp->mnt_stat, p);
318 * checks that the data in the descriptor blocks make sense
319 * this is taken from ext2/super.c
321 static int ext2_check_descriptors (struct ext2_sb_info * sb)
325 unsigned long block = sb->s_es->s_first_data_block;
326 struct ext2_group_desc * gdp = NULL;
328 /* ext2_debug ("Checking group descriptors"); */
330 for (i = 0; i < sb->s_groups_count; i++)
332 /* examine next descriptor block */
333 if ((i % EXT2_DESC_PER_BLOCK(sb)) == 0)
334 gdp = (struct ext2_group_desc *)
335 sb->s_group_desc[desc_block++]->b_data;
336 if (gdp->bg_block_bitmap < block ||
337 gdp->bg_block_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb))
339 printf ("ext2_check_descriptors: "
340 "Block bitmap for group %d"
341 " not in group (block %lu)!\n",
342 i, (unsigned long) gdp->bg_block_bitmap);
345 if (gdp->bg_inode_bitmap < block ||
346 gdp->bg_inode_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb))
348 printf ("ext2_check_descriptors: "
349 "Inode bitmap for group %d"
350 " not in group (block %lu)!\n",
351 i, (unsigned long) gdp->bg_inode_bitmap);
354 if (gdp->bg_inode_table < block ||
355 gdp->bg_inode_table + sb->s_itb_per_group >=
356 block + EXT2_BLOCKS_PER_GROUP(sb))
358 printf ("ext2_check_descriptors: "
359 "Inode table for group %d"
360 " not in group (block %lu)!\n",
361 i, (unsigned long) gdp->bg_inode_table);
364 block += EXT2_BLOCKS_PER_GROUP(sb);
371 ext2_check_sb_compat(es, dev, ronly)
372 struct ext2_super_block *es;
377 if (es->s_magic != EXT2_SUPER_MAGIC) {
378 printf("ext2fs: %s: wrong magic number %#x (expected %#x)\n",
379 devtoname(dev), es->s_magic, EXT2_SUPER_MAGIC);
382 if (es->s_rev_level > EXT2_GOOD_OLD_REV) {
383 if (es->s_feature_incompat & ~EXT2_FEATURE_INCOMPAT_SUPP) {
385 "WARNING: mount of %s denied due to unsupported optional features\n",
390 (es->s_feature_ro_compat & ~EXT2_FEATURE_RO_COMPAT_SUPP)) {
392 "WARNING: R/W mount of %s denied due to unsupported optional features\n",
401 * this computes the fields of the ext2_sb_info structure from the
402 * data in the ext2_super_block structure read in
404 static int compute_sb_data(devvp, es, fs)
405 struct vnode * devvp;
406 struct ext2_super_block * es;
407 struct ext2_sb_info * fs;
411 int logic_sb_block = 1; /* XXX for now */
416 #define V(v) printf(#v"= %d\n", fs->v);
419 fs->s_blocksize = EXT2_MIN_BLOCK_SIZE << es->s_log_block_size;
421 fs->s_bshift = EXT2_MIN_BLOCK_LOG_SIZE + es->s_log_block_size;
423 fs->s_fsbtodb = es->s_log_block_size + 1;
425 fs->s_qbmask = fs->s_blocksize - 1;
427 fs->s_blocksize_bits = EXT2_BLOCK_SIZE_BITS(es);
429 fs->s_frag_size = EXT2_MIN_FRAG_SIZE << es->s_log_frag_size;
432 fs->s_frags_per_block = fs->s_blocksize / fs->s_frag_size;
434 fs->s_blocks_per_group = es->s_blocks_per_group;
435 V(s_blocks_per_group)
436 fs->s_frags_per_group = es->s_frags_per_group;
438 fs->s_inodes_per_group = es->s_inodes_per_group;
439 V(s_inodes_per_group)
440 fs->s_inodes_per_block = fs->s_blocksize / EXT2_INODE_SIZE;
441 V(s_inodes_per_block)
442 fs->s_itb_per_group = fs->s_inodes_per_group /fs->s_inodes_per_block;
444 fs->s_desc_per_block = fs->s_blocksize / sizeof (struct ext2_group_desc);
446 /* s_resuid / s_resgid ? */
447 fs->s_groups_count = (es->s_blocks_count -
448 es->s_first_data_block +
449 EXT2_BLOCKS_PER_GROUP(fs) - 1) /
450 EXT2_BLOCKS_PER_GROUP(fs);
452 db_count = (fs->s_groups_count + EXT2_DESC_PER_BLOCK(fs) - 1) /
453 EXT2_DESC_PER_BLOCK(fs);
454 fs->s_db_per_group = db_count;
457 fs->s_group_desc = bsd_malloc(db_count * sizeof (struct buf *),
460 /* adjust logic_sb_block */
461 if(fs->s_blocksize > SBSIZE)
462 /* Godmar thinks: if the blocksize is greater than 1024, then
463 the superblock is logically part of block zero.
467 for (i = 0; i < db_count; i++) {
468 error = bread(devvp , fsbtodb(fs, logic_sb_block + i + 1),
469 fs->s_blocksize, NOCRED, &fs->s_group_desc[i]);
471 for (j = 0; j < i; j++)
472 brelse(fs->s_group_desc[j]);
473 bsd_free(fs->s_group_desc, M_UFSMNT);
474 printf("EXT2-fs: unable to read group descriptors (%d)\n", error);
477 /* Set the B_LOCKED flag on the buffer, then brelse() it */
478 LCK_BUF(fs->s_group_desc[i])
480 if(!ext2_check_descriptors(fs)) {
481 for (j = 0; j < db_count; j++)
482 ULCK_BUF(fs->s_group_desc[j])
483 bsd_free(fs->s_group_desc, M_UFSMNT);
484 printf("EXT2-fs: (ext2_check_descriptors failure) "
485 "unable to read group descriptors\n");
489 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) {
490 fs->s_inode_bitmap_number[i] = 0;
491 fs->s_inode_bitmap[i] = NULL;
492 fs->s_block_bitmap_number[i] = 0;
493 fs->s_block_bitmap[i] = NULL;
495 fs->s_loaded_inode_bitmaps = 0;
496 fs->s_loaded_block_bitmaps = 0;
501 * Reload all incore data for a filesystem (used after running fsck on
502 * the root filesystem and finding things to fix). The filesystem must
503 * be mounted read-only.
505 * Things to do to update the mount:
506 * 1) invalidate all cached meta-data.
507 * 2) re-read superblock from disk.
508 * 3) re-read summary information from disk.
509 * 4) invalidate all inactive vnodes.
510 * 5) invalidate all cached file data.
511 * 6) re-read inode data for all active vnodes.
514 ext2_reload(mountp, cred, p)
515 register struct mount *mountp;
519 register struct vnode *vp, *nvp, *devvp;
522 struct ext2_super_block * es;
523 struct ext2_sb_info *fs;
526 if ((mountp->mnt_flag & MNT_RDONLY) == 0)
529 * Step 1: invalidate all cached meta-data.
531 devvp = VFSTOUFS(mountp)->um_devvp;
532 if (vinvalbuf(devvp, 0, cred, p, 0, 0))
533 panic("ext2_reload: dirty1");
535 * Step 2: re-read superblock from disk.
536 * constants have been adjusted for ext2
538 if ((error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) != 0)
540 es = (struct ext2_super_block *)bp->b_data;
541 if (ext2_check_sb_compat(es, devvp->v_rdev, 0) != 0) {
543 return (EIO); /* XXX needs translation */
545 fs = VFSTOUFS(mountp)->um_e2fs;
546 bcopy(bp->b_data, fs->s_es, sizeof(struct ext2_super_block));
548 if((error = compute_sb_data(devvp, es, fs)) != 0) {
553 if (fs->fs_sbsize < SBSIZE)
554 bp->b_flags |= B_INVAL;
559 simple_lock(&mntvnode_slock);
560 for (vp = mountp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) {
561 if (vp->v_mount != mountp) {
562 simple_unlock(&mntvnode_slock);
565 nvp = vp->v_mntvnodes.le_next;
567 * Step 4: invalidate all inactive vnodes.
569 if (vrecycle(vp, &mntvnode_slock, p))
572 * Step 5: invalidate all cached file data.
574 simple_lock(&vp->v_interlock);
575 simple_unlock(&mntvnode_slock);
576 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, p)) {
579 if (vinvalbuf(vp, 0, cred, p, 0, 0))
580 panic("ext2_reload: dirty2");
582 * Step 6: re-read inode data for all active vnodes.
586 bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
587 (int)fs->s_blocksize, NOCRED, &bp);
592 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data +
593 EXT2_INODE_SIZE * ino_to_fsbo(fs, ip->i_number)),
597 simple_lock(&mntvnode_slock);
599 simple_unlock(&mntvnode_slock);
604 * Common code for mount and mountroot
607 ext2_mountfs(devvp, mp, p)
608 register struct vnode *devvp;
612 register struct ufsmount *ump;
614 register struct ext2_sb_info *fs;
615 struct ext2_super_block * es;
616 dev_t dev = devvp->v_rdev;
617 struct partinfo dpart;
623 * Disallow multiple mounts of the same device.
624 * Disallow mounting of a device that is currently in use
625 * (except for root, which might share swap device for miniroot).
626 * Flush out any old buffers remaining from a previous use.
628 if ((error = vfs_mountedon(devvp)) != 0)
630 if (vcount(devvp) > 1 && devvp != rootvp)
632 if ((error = vinvalbuf(devvp, V_SAVE, p->p_ucred, p, 0, 0)) != 0)
635 /* turn on this to force it to be read-only */
636 mp->mnt_flag |= MNT_RDONLY;
639 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
640 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p);
641 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p);
642 VOP_UNLOCK(devvp, 0, p);
645 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, p) != 0)
649 size = dpart.disklab->d_secsize;
654 if ((error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) != 0)
656 es = (struct ext2_super_block *)bp->b_data;
657 if (ext2_check_sb_compat(es, dev, ronly) != 0) {
658 error = EINVAL; /* XXX needs translation */
661 if ((es->s_state & EXT2_VALID_FS) == 0 ||
662 (es->s_state & EXT2_ERROR_FS)) {
663 if (ronly || (mp->mnt_flag & MNT_FORCE)) {
665 "WARNING: Filesystem was not properly dismounted\n");
668 "WARNING: R/W mount denied. Filesystem is not clean - run fsck\n");
673 ump = bsd_malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
674 bzero((caddr_t)ump, sizeof *ump);
675 ump->um_malloctype = M_EXT2NODE;
676 ump->um_blkatoff = ext2_blkatoff;
677 ump->um_truncate = ext2_truncate;
678 ump->um_update = ext2_update;
679 ump->um_valloc = ext2_valloc;
680 ump->um_vfree = ext2_vfree;
681 /* I don't know whether this is the right strategy. Note that
682 we dynamically allocate both a ext2_sb_info and a ext2_super_block
683 while Linux keeps the super block in a locked buffer
685 ump->um_e2fs = bsd_malloc(sizeof(struct ext2_sb_info),
687 ump->um_e2fs->s_es = bsd_malloc(sizeof(struct ext2_super_block),
689 bcopy(es, ump->um_e2fs->s_es, (u_int)sizeof(struct ext2_super_block));
690 if ((error = compute_sb_data(devvp, ump->um_e2fs->s_es, ump->um_e2fs)))
693 * We don't free the group descriptors allocated by compute_sb_data()
694 * until ext2_unmount(). This is OK since the mount will succeed.
699 fs->s_rd_only = ronly; /* ronly is set according to mnt_flags */
700 /* if the fs is not mounted read-only, make sure the super block is
701 always written back on a sync()
703 fs->s_wasvalid = fs->s_es->s_state & EXT2_VALID_FS ? 1 : 0;
705 fs->s_dirt = 1; /* mark it modified */
706 fs->s_es->s_state &= ~EXT2_VALID_FS; /* set fs invalid */
708 mp->mnt_data = (qaddr_t)ump;
709 mp->mnt_stat.f_fsid.val[0] = dev2udev(dev);
710 mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
711 mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN;
712 mp->mnt_flag |= MNT_LOCAL;
715 ump->um_devvp = devvp;
716 /* setting those two parameters allows us to use
717 ufs_bmap w/o changse !
719 ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs);
720 ump->um_bptrtodb = fs->s_es->s_log_block_size + 1;
721 ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs);
722 for (i = 0; i < MAXQUOTAS; i++)
723 ump->um_quotas[i] = NULLVP;
724 devvp->v_specmountpoint = mp;
726 ext2_sbupdate(ump, MNT_WAIT);
731 (void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, NOCRED, p);
733 bsd_free(ump->um_e2fs->s_es, M_UFSMNT);
734 bsd_free(ump->um_e2fs, M_UFSMNT);
735 bsd_free(ump, M_UFSMNT);
736 mp->mnt_data = (qaddr_t)0;
742 * unmount system call
745 ext2_unmount(mp, mntflags, p)
750 register struct ufsmount *ump;
751 register struct ext2_sb_info *fs;
752 int error, flags, ronly, i;
755 if (mntflags & MNT_FORCE) {
756 if (mp->mnt_flag & MNT_ROOTFS)
760 if ((error = ext2_flushfiles(mp, flags, p)) != 0)
764 ronly = fs->s_rd_only;
767 fs->s_es->s_state |= EXT2_VALID_FS;
768 ext2_sbupdate(ump, MNT_WAIT);
771 /* release buffers containing group descriptors */
772 for(i = 0; i < fs->s_db_per_group; i++)
773 ULCK_BUF(fs->s_group_desc[i])
774 bsd_free(fs->s_group_desc, M_UFSMNT);
776 /* release cached inode/block bitmaps */
777 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
778 if (fs->s_inode_bitmap[i])
779 ULCK_BUF(fs->s_inode_bitmap[i])
781 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
782 if (fs->s_block_bitmap[i])
783 ULCK_BUF(fs->s_block_bitmap[i])
785 ump->um_devvp->v_specmountpoint = NULL;
786 error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD|FWRITE,
788 vrele(ump->um_devvp);
789 bsd_free(fs->s_es, M_UFSMNT);
790 bsd_free(fs, M_UFSMNT);
791 bsd_free(ump, M_UFSMNT);
792 mp->mnt_data = (qaddr_t)0;
793 mp->mnt_flag &= ~MNT_LOCAL;
798 * Flush out all the files in a filesystem.
801 ext2_flushfiles(mp, flags, p)
802 register struct mount *mp;
806 register struct ufsmount *ump;
814 if (mp->mnt_flag & MNT_QUOTA) {
815 if ((error = vflush(mp, NULLVP, SKIPSYSTEM|flags)) != 0)
817 for (i = 0; i < MAXQUOTAS; i++) {
818 if (ump->um_quotas[i] == NULLVP)
823 * Here we fall through to vflush again to ensure
824 * that we have gotten rid of all the system vnodes.
828 error = vflush(mp, NULLVP, flags);
833 * Get file system statistics.
834 * taken from ext2/super.c ext2_statfs
837 ext2_statfs(mp, sbp, p)
839 register struct statfs *sbp;
842 unsigned long overhead;
843 unsigned long overhead_per_group;
845 register struct ufsmount *ump;
846 register struct ext2_sb_info *fs;
847 register struct ext2_super_block *es;
853 if (es->s_magic != EXT2_SUPER_MAGIC)
854 panic("ext2_statfs - magic number spoiled");
857 * Compute the overhead (FS structures)
859 overhead_per_group = 1 /* super block */ +
861 1 /* block bitmap */ +
862 1 /* inode bitmap */ +
864 overhead = es->s_first_data_block +
865 fs->s_groups_count * overhead_per_group;
867 sbp->f_bsize = EXT2_FRAG_SIZE(fs);
868 sbp->f_iosize = EXT2_BLOCK_SIZE(fs);
869 sbp->f_blocks = es->s_blocks_count - overhead;
870 sbp->f_bfree = es->s_free_blocks_count;
871 sbp->f_bavail = sbp->f_bfree - es->s_r_blocks_count;
872 sbp->f_files = es->s_inodes_count;
873 sbp->f_ffree = es->s_free_inodes_count;
874 if (sbp != &mp->mnt_stat) {
875 sbp->f_type = mp->mnt_vfc->vfc_typenum;
876 bcopy((caddr_t)mp->mnt_stat.f_mntonname,
877 (caddr_t)&sbp->f_mntonname[0], MNAMELEN);
878 bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
879 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
885 * Go through the disk queues to initiate sandbagged IO;
886 * go through the inodes to write those that have been modified;
887 * initiate the writing of the super block if it has been modified.
889 * Note: we are always called with the filesystem marked `MPBUSY'.
892 ext2_sync(mp, waitfor, cred, p)
898 struct vnode *nvp, *vp;
900 struct ufsmount *ump = VFSTOUFS(mp);
901 struct ext2_sb_info *fs;
902 int error, allerror = 0;
905 if (fs->s_dirt != 0 && fs->s_rd_only != 0) { /* XXX */
906 printf("fs = %s\n", fs->fs_fsmnt);
907 panic("ext2_sync: rofs mod");
910 * Write back each (modified) inode.
912 simple_lock(&mntvnode_slock);
914 for (vp = mp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) {
916 * If the vnode that we are about to sync is no longer
917 * associated with this mount point, start over.
919 if (vp->v_mount != mp)
921 simple_lock(&vp->v_interlock);
922 nvp = vp->v_mntvnodes.le_next;
924 if (vp->v_type == VNON ||
926 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
927 (TAILQ_EMPTY(&vp->v_dirtyblkhd) || waitfor == MNT_LAZY))) {
928 simple_unlock(&vp->v_interlock);
931 simple_unlock(&mntvnode_slock);
932 error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK, p);
934 simple_lock(&mntvnode_slock);
939 if ((error = VOP_FSYNC(vp, cred, waitfor, p)) != 0)
941 VOP_UNLOCK(vp, 0, p);
943 simple_lock(&mntvnode_slock);
945 simple_unlock(&mntvnode_slock);
947 * Force stale file system control information to be flushed.
949 if (waitfor != MNT_LAZY) {
950 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY, p);
951 if ((error = VOP_FSYNC(ump->um_devvp, cred, waitfor, p)) != 0)
953 VOP_UNLOCK(ump->um_devvp, 0, p);
959 * Write back modified superblock.
961 if (fs->s_dirt != 0) {
963 fs->s_es->s_wtime = time_second;
964 if ((error = ext2_sbupdate(ump, waitfor)) != 0)
971 * Look up a EXT2FS dinode number to find its incore vnode, otherwise read it
972 * in from disk. If it is in core, wait for the lock bit to clear, then
973 * return the inode locked. Detection and handling of mount points must be
974 * done by the calling routine.
977 ext2_vget(mp, ino, vpp)
982 register struct ext2_sb_info *fs;
983 register struct inode *ip;
984 struct ufsmount *ump;
994 if ((*vpp = ufs_ihashget(dev, ino)) != NULL)
998 * Lock out the creation of new entries in the FFS hash table in
999 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate
1002 if (ext2fs_inode_hash_lock) {
1003 while (ext2fs_inode_hash_lock) {
1004 ext2fs_inode_hash_lock = -1;
1005 tsleep(&ext2fs_inode_hash_lock, PVM, "e2vget", 0);
1009 ext2fs_inode_hash_lock = 1;
1012 * If this MALLOC() is performed after the getnewvnode()
1013 * it might block, leaving a vnode with a NULL v_data to be
1014 * found by ext2_sync() if a sync happens to fire right then,
1015 * which will cause a panic because ext2_sync() blindly
1016 * dereferences vp->v_data (as well it should).
1018 MALLOC(ip, struct inode *, sizeof(struct inode), M_EXT2NODE, M_WAITOK);
1020 /* Allocate a new vnode/inode. */
1021 if ((error = getnewvnode(VT_UFS, mp, ext2_vnodeop_p, &vp)) != 0) {
1022 if (ext2fs_inode_hash_lock < 0)
1023 wakeup(&ext2fs_inode_hash_lock);
1024 ext2fs_inode_hash_lock = 0;
1026 FREE(ip, M_EXT2NODE);
1029 bzero((caddr_t)ip, sizeof(struct inode));
1030 lockinit(&ip->i_lock, PINOD, "ext2in", 0, 0);
1033 ip->i_e2fs = fs = ump->um_e2fs;
1037 for (i = 0; i < MAXQUOTAS; i++)
1038 ip->i_dquot[i] = NODQUOT;
1041 * Put it onto its hash chain and lock it so that other requests for
1042 * this inode will block if they arrive while we are sleeping waiting
1043 * for old data structures to be purged or for the contents of the
1044 * disk portion of this inode to be read.
1048 if (ext2fs_inode_hash_lock < 0)
1049 wakeup(&ext2fs_inode_hash_lock);
1050 ext2fs_inode_hash_lock = 0;
1052 /* Read in the disk contents for the inode, copy into the inode. */
1054 printf("ext2_vget(%d) dbn= %d ", ino, fsbtodb(fs, ino_to_fsba(fs, ino)));
1056 if ((error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
1057 (int)fs->s_blocksize, NOCRED, &bp)) != 0) {
1059 * The inode does not contain anything useful, so it would
1060 * be misleading to leave it on its hash chain. With mode
1061 * still zero, it will be unlinked and returned to the free
1069 /* convert ext2 inode to dinode */
1070 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data + EXT2_INODE_SIZE *
1071 ino_to_fsbo(fs, ino)), &ip->i_din);
1072 ip->i_block_group = ino_to_cg(fs, ino);
1073 ip->i_next_alloc_block = 0;
1074 ip->i_next_alloc_goal = 0;
1075 ip->i_prealloc_count = 0;
1076 ip->i_prealloc_block = 0;
1077 /* now we want to make sure that block pointers for unused
1078 blocks are zeroed out - ext2_balloc depends on this
1079 although for regular files and directories only
1081 if(S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode)) {
1082 used_blocks = (ip->i_size+fs->s_blocksize-1) / fs->s_blocksize;
1083 for(i = used_blocks; i < EXT2_NDIR_BLOCKS; i++)
1087 ext2_print_inode(ip);
1092 * Initialize the vnode from the inode, check for aliases.
1093 * Note that the underlying vnode may have changed.
1095 if ((error = ufs_vinit(mp, ext2_specop_p, ext2_fifoop_p, &vp)) != 0) {
1101 * Finish inode initialization now that aliasing has been resolved.
1103 ip->i_devvp = ump->um_devvp;
1106 * Set up a generation number for this inode if it does not
1107 * already have one. This should only happen on old filesystems.
1109 if (ip->i_gen == 0) {
1110 ip->i_gen = random() / 2 + 1;
1111 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1112 ip->i_flag |= IN_MODIFIED;
1119 * File handle to vnode
1121 * Have to be really careful about stale file handles:
1122 * - check that the inode number is valid
1123 * - call ext2_vget() to get the locked inode
1124 * - check for an unallocated inode (i_mode == 0)
1125 * - check that the given client host has export rights and return
1126 * those rights via. exflagsp and credanonp
1129 ext2_fhtovp(mp, fhp, vpp)
1130 register struct mount *mp;
1134 register struct ufid *ufhp;
1135 struct ext2_sb_info *fs;
1137 ufhp = (struct ufid *)fhp;
1138 fs = VFSTOUFS(mp)->um_e2fs;
1139 if (ufhp->ufid_ino < ROOTINO ||
1140 ufhp->ufid_ino >= fs->s_groups_count * fs->s_es->s_inodes_per_group)
1142 return (ufs_fhtovp(mp, ufhp, vpp));
1146 * Vnode pointer to File handle
1150 ext2_vptofh(vp, fhp)
1154 register struct inode *ip;
1155 register struct ufid *ufhp;
1158 ufhp = (struct ufid *)fhp;
1159 ufhp->ufid_len = sizeof(struct ufid);
1160 ufhp->ufid_ino = ip->i_number;
1161 ufhp->ufid_gen = ip->i_gen;
1166 * Write a superblock and associated information back to disk.
1169 ext2_sbupdate(mp, waitfor)
1170 struct ufsmount *mp;
1173 register struct ext2_sb_info *fs = mp->um_e2fs;
1174 register struct ext2_super_block *es = fs->s_es;
1175 register struct buf *bp;
1178 printf("\nupdating superblock, waitfor=%s\n", waitfor == MNT_WAIT ? "yes":"no");
1180 bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0);
1181 bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2_super_block));
1182 if (waitfor == MNT_WAIT)
1188 * The buffers for group descriptors, inode bitmaps and block bitmaps
1189 * are not busy at this point and are (hopefully) written by the
1190 * usual sync mechanism. No need to write them here