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 1.37 2006/04/09 20:07:43 dillon Exp $
44 #include "opt_quota.h"
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/nlookup.h>
50 #include <sys/kernel.h>
51 #include <sys/vnode.h>
52 #include <sys/mount.h>
55 #include <sys/fcntl.h>
56 #include <sys/disklabel.h>
57 #include <sys/malloc.h>
59 #include <vm/vm_zone.h>
62 #include <sys/thread2.h>
67 #include "ext2mount.h"
68 #include "ext2_extern.h"
72 #include "ext2_fs_sb.h"
74 extern struct vnodeopv_entry_desc ext2_vnodeop_entries[];
75 extern struct vnodeopv_entry_desc ext2_specop_entries[];
76 extern struct vnodeopv_entry_desc ext2_fifoop_entries[];
78 static int ext2_fhtovp (struct mount *, struct fid *, struct vnode **);
79 static int ext2_flushfiles (struct mount *mp, int flags, struct thread *td);
80 static int ext2_mount (struct mount *, char *, caddr_t, struct thread *);
81 static int ext2_mountfs (struct vnode *, struct mount *, struct thread *);
82 static int ext2_root(struct mount *, struct vnode **);
83 static int ext2_reload (struct mount *mountp, struct ucred *cred,
85 static int ext2_sbupdate (struct ext2mount *, int);
86 static int ext2_statfs (struct mount *, struct statfs *, struct thread *);
87 static int ext2_sync (struct mount *, int, struct thread *);
88 static int ext2_unmount (struct mount *, int, struct thread *);
89 static int ext2_vget (struct mount *, ino_t, struct vnode **);
90 static int ext2_init(struct vfsconf *);
91 static int ext2_vptofh (struct vnode *, struct fid *);
93 static MALLOC_DEFINE(M_EXT2NODE, "EXT2 node", "EXT2 vnode private part");
94 MALLOC_DEFINE(M_EXT2MNT, "EXT2 mount", "EXT2 mount structure");
96 static struct vfsops ext2fs_vfsops = {
97 .vfs_mount = ext2_mount,
98 .vfs_unmount = ext2_unmount,
99 .vfs_root = ext2_root, /* root inode via vget */
100 .vfs_quotactl = ext2_quotactl, /* quota operations */
101 .vfs_statfs = ext2_statfs,
102 .vfs_sync = ext2_sync,
103 .vfs_vget = ext2_vget,
104 .vfs_fhtovp = ext2_fhtovp,
105 .vfs_checkexp = ext2_check_export,
106 .vfs_vptofh = ext2_vptofh,
107 .vfs_init = ext2_init,
108 .vfs_uninit = ext2_uninit
111 VFS_SET(ext2fs_vfsops, ext2fs, 0);
112 #define bsd_malloc malloc
113 #define bsd_free free
115 static int ext2fs_inode_hash_lock;
117 static int ext2_check_sb_compat (struct ext2_super_block *es,
118 dev_t dev, int ronly);
119 static int compute_sb_data (struct vnode * devvp,
120 struct ext2_super_block * es,
121 struct ext2_sb_info * fs);
124 ext2_root(struct mount *mp, struct vnode **vpp)
129 error = VFS_VGET(mp, (ino_t)ROOTINO, &nvp);
137 * Do operations associated with quotas
140 ext2_quotactl(struct mount *mp, int cmds, uid_t uid, caddr_t arg,
146 struct proc *p = td->td_proc;
147 int cmd, type, error;
152 type = cmds & SUBCMDMASK;
153 cmd = cmds >> SUBCMDSHIFT;
158 uid = p->p_ucred->cr_ruid;
161 uid = p->p_ucred->cr_rgid;
172 if (uid == p->p_ucred->cr_ruid)
176 if ((error = suser_cred(p->p_ucred, PRISON_ROOT)) != 0)
180 type = cmds & SUBCMDMASK;
181 if ((uint)type >= MAXQUOTAS)
183 if (vfs_busy(mp, LK_NOWAIT, td))
189 error = ext2_quotaon(td, mp, type, arg);
193 error = ext2_quotaoff(td, mp, type);
197 error = ext2_setquota(mp, uid, type, arg);
201 error = ext2_setuse(mp, uid, type, arg);
205 error = ext2_getquota(mp, uid, type, arg);
209 error = ext2_qsync(mp);
222 * Initial UFS filesystems, done only once.
225 ext2_init(struct vfsconf *vfsp)
245 * data: this is actually a (struct ext2_args *)
248 ext2_mount(struct mount *mp, char *path, caddr_t data,
252 struct ext2_args args;
253 struct ext2mount *ump = 0;
254 struct ext2_sb_info *fs;
259 struct nlookupdata nd;
261 if ((error = copyin(data, (caddr_t)&args, sizeof (struct ext2_args))) != 0)
264 cred = td->td_proc->p_ucred;
266 * If updating, check whether changing from read-only to
267 * read/write; if there is no device name, that's all we do.
269 if (mp->mnt_flag & MNT_UPDATE) {
272 devvp = ump->um_devvp;
274 if (fs->s_rd_only == 0 && (mp->mnt_flag & MNT_RDONLY)) {
276 if (mp->mnt_flag & MNT_FORCE)
278 if (vfs_busy(mp, LK_NOWAIT, td))
280 error = ext2_flushfiles(mp, flags, td);
282 if (!error && fs->s_wasvalid) {
283 fs->s_es->s_state |= EXT2_VALID_FS;
284 ext2_sbupdate(ump, MNT_WAIT);
287 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
288 VOP_OPEN(devvp, FREAD, FSCRED, NULL, td);
289 VOP_CLOSE(devvp, FREAD|FWRITE, td);
290 VOP_UNLOCK(devvp, 0, td);
292 if (!error && (mp->mnt_flag & MNT_RELOAD))
293 error = ext2_reload(mp, proc0.p_ucred, td);
296 if (ext2_check_sb_compat(fs->s_es, devvp->v_rdev,
297 (mp->mnt_kern_flag & MNTK_WANTRDWR) == 0) != 0)
299 if (fs->s_rd_only && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
301 * If upgrade to read-write by non-root, then verify
302 * that user has necessary permissions on the device.
304 if (cred->cr_uid != 0) {
305 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
306 if ((error = VOP_ACCESS(devvp, VREAD | VWRITE,
308 VOP_UNLOCK(devvp, 0, td);
311 VOP_UNLOCK(devvp, 0, td);
314 if ((fs->s_es->s_state & EXT2_VALID_FS) == 0 ||
315 (fs->s_es->s_state & EXT2_ERROR_FS)) {
316 if (mp->mnt_flag & MNT_FORCE) {
318 "WARNING: %s was not properly dismounted\n",
322 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
327 fs->s_es->s_state &= ~EXT2_VALID_FS;
328 ext2_sbupdate(ump, MNT_WAIT);
330 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
331 VOP_OPEN(devvp, FREAD|FWRITE, FSCRED, NULL, td);
332 VOP_CLOSE(devvp, FREAD, td);
333 VOP_UNLOCK(devvp, 0, td);
335 if (args.fspec == 0) {
337 * Process export requests.
339 return (vfs_export(mp, &ump->um_export, &args.export));
343 * Not an update, or updating the name: look up the name
344 * and verify that it refers to a sensible block device.
347 error = nlookup_init(&nd, args.fspec, UIO_USERSPACE, NLC_FOLLOW);
349 error = nlookup(&nd);
351 error = cache_vref(nd.nl_ncp, nd.nl_cred, &devvp);
356 if (!vn_isdisk(devvp, &error)) {
362 * If mount by non-root, then verify that user has necessary
363 * permissions on the device.
365 if (cred->cr_uid != 0) {
367 if ((mp->mnt_flag & MNT_RDONLY) == 0)
368 accessmode |= VWRITE;
369 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
370 if ((error = VOP_ACCESS(devvp, accessmode, cred, td)) != 0) {
374 VOP_UNLOCK(devvp, 0, td);
377 if ((mp->mnt_flag & MNT_UPDATE) == 0) {
378 error = ext2_mountfs(devvp, mp, td);
380 if (devvp != ump->um_devvp)
381 error = EINVAL; /* needs translation */
391 copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size);
392 bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size);
393 copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, &size);
394 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
395 ext2_statfs(mp, &mp->mnt_stat, td);
400 * checks that the data in the descriptor blocks make sense
401 * this is taken from ext2/super.c
404 ext2_check_descriptors(struct ext2_sb_info *sb)
408 unsigned long block = sb->s_es->s_first_data_block;
409 struct ext2_group_desc * gdp = NULL;
411 /* ext2_debug ("Checking group descriptors"); */
413 for (i = 0; i < sb->s_groups_count; i++)
415 /* examine next descriptor block */
416 if ((i % EXT2_DESC_PER_BLOCK(sb)) == 0)
417 gdp = (struct ext2_group_desc *)
418 sb->s_group_desc[desc_block++]->b_data;
419 if (gdp->bg_block_bitmap < block ||
420 gdp->bg_block_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb))
422 printf ("ext2_check_descriptors: "
423 "Block bitmap for group %d"
424 " not in group (block %lu)!\n",
425 i, (unsigned long) gdp->bg_block_bitmap);
428 if (gdp->bg_inode_bitmap < block ||
429 gdp->bg_inode_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb))
431 printf ("ext2_check_descriptors: "
432 "Inode bitmap for group %d"
433 " not in group (block %lu)!\n",
434 i, (unsigned long) gdp->bg_inode_bitmap);
437 if (gdp->bg_inode_table < block ||
438 gdp->bg_inode_table + sb->s_itb_per_group >=
439 block + EXT2_BLOCKS_PER_GROUP(sb))
441 printf ("ext2_check_descriptors: "
442 "Inode table for group %d"
443 " not in group (block %lu)!\n",
444 i, (unsigned long) gdp->bg_inode_table);
447 block += EXT2_BLOCKS_PER_GROUP(sb);
454 ext2_check_sb_compat(struct ext2_super_block *es, dev_t dev, int ronly)
456 if (es->s_magic != EXT2_SUPER_MAGIC) {
457 printf("ext2fs: %s: wrong magic number %#x (expected %#x)\n",
458 devtoname(dev), es->s_magic, EXT2_SUPER_MAGIC);
461 if (es->s_rev_level > EXT2_GOOD_OLD_REV) {
462 if (es->s_feature_incompat & ~EXT2_FEATURE_INCOMPAT_SUPP) {
464 "WARNING: mount of %s denied due to unsupported optional features\n",
469 (es->s_feature_ro_compat & ~EXT2_FEATURE_RO_COMPAT_SUPP)) {
471 "WARNING: R/W mount of %s denied due to unsupported optional features\n",
480 * this computes the fields of the ext2_sb_info structure from the
481 * data in the ext2_super_block structure read in
484 compute_sb_data(struct vnode *devvp, struct ext2_super_block *es,
485 struct ext2_sb_info *fs)
489 int logic_sb_block = 1; /* XXX for now */
494 #define V(v) printf(#v"= %d\n", fs->v);
497 fs->s_blocksize = EXT2_MIN_BLOCK_SIZE << es->s_log_block_size;
499 fs->s_bshift = EXT2_MIN_BLOCK_LOG_SIZE + es->s_log_block_size;
501 fs->s_fsbtodb = es->s_log_block_size + 1;
503 fs->s_qbmask = fs->s_blocksize - 1;
505 fs->s_blocksize_bits = EXT2_BLOCK_SIZE_BITS(es);
507 fs->s_frag_size = EXT2_MIN_FRAG_SIZE << es->s_log_frag_size;
510 fs->s_frags_per_block = fs->s_blocksize / fs->s_frag_size;
512 fs->s_blocks_per_group = es->s_blocks_per_group;
513 V(s_blocks_per_group)
514 fs->s_frags_per_group = es->s_frags_per_group;
516 fs->s_inodes_per_group = es->s_inodes_per_group;
517 V(s_inodes_per_group)
518 fs->s_inodes_per_block = fs->s_blocksize / EXT2_INODE_SIZE;
519 V(s_inodes_per_block)
520 fs->s_itb_per_group = fs->s_inodes_per_group /fs->s_inodes_per_block;
522 fs->s_desc_per_block = fs->s_blocksize / sizeof (struct ext2_group_desc);
524 /* s_resuid / s_resgid ? */
525 fs->s_groups_count = (es->s_blocks_count -
526 es->s_first_data_block +
527 EXT2_BLOCKS_PER_GROUP(fs) - 1) /
528 EXT2_BLOCKS_PER_GROUP(fs);
530 db_count = (fs->s_groups_count + EXT2_DESC_PER_BLOCK(fs) - 1) /
531 EXT2_DESC_PER_BLOCK(fs);
532 fs->s_db_per_group = db_count;
535 fs->s_group_desc = bsd_malloc(db_count * sizeof (struct buf *),
536 M_EXT2MNT, M_WAITOK);
538 /* adjust logic_sb_block */
539 if(fs->s_blocksize > SBSIZE)
540 /* Godmar thinks: if the blocksize is greater than 1024, then
541 the superblock is logically part of block zero.
545 for (i = 0; i < db_count; i++) {
546 error = bread(devvp, fsbtodoff(fs, logic_sb_block + i + 1),
547 fs->s_blocksize, &fs->s_group_desc[i]);
549 for (j = 0; j < i; j++)
550 brelse(fs->s_group_desc[j]);
551 bsd_free(fs->s_group_desc, M_EXT2MNT);
552 printf("EXT2-fs: unable to read group descriptors (%d)\n", error);
555 /* Set the B_LOCKED flag on the buffer, then brelse() it */
556 LCK_BUF(fs->s_group_desc[i])
558 if(!ext2_check_descriptors(fs)) {
559 for (j = 0; j < db_count; j++)
560 ULCK_BUF(fs->s_group_desc[j])
561 bsd_free(fs->s_group_desc, M_EXT2MNT);
562 printf("EXT2-fs: (ext2_check_descriptors failure) "
563 "unable to read group descriptors\n");
567 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) {
568 fs->s_inode_bitmap_number[i] = 0;
569 fs->s_inode_bitmap[i] = NULL;
570 fs->s_block_bitmap_number[i] = 0;
571 fs->s_block_bitmap[i] = NULL;
573 fs->s_loaded_inode_bitmaps = 0;
574 fs->s_loaded_block_bitmaps = 0;
579 * Reload all incore data for a filesystem (used after running fsck on
580 * the root filesystem and finding things to fix). The filesystem must
581 * be mounted read-only.
583 * Things to do to update the mount:
584 * 1) invalidate all cached meta-data.
585 * 2) re-read superblock from disk.
586 * 3) re-read summary information from disk.
587 * 4) invalidate all inactive vnodes.
588 * 5) invalidate all cached file data.
589 * 6) re-read inode data for all active vnodes.
591 static int ext2_reload_scan1(struct mount *mp, struct vnode *vp, void *rescan);
592 static int ext2_reload_scan2(struct mount *mp, struct vnode *vp, void *rescan);
600 struct ext2_sb_info *fs;
604 ext2_reload(struct mount *mountp, struct ucred *cred, struct thread *td)
608 struct ext2_super_block * es;
609 struct ext2_sb_info *fs;
611 struct scaninfo scaninfo;
613 if ((mountp->mnt_flag & MNT_RDONLY) == 0)
616 * Step 1: invalidate all cached meta-data.
618 devvp = VFSTOEXT2(mountp)->um_devvp;
619 if (vinvalbuf(devvp, 0, td, 0, 0))
620 panic("ext2_reload: dirty1");
622 * Step 2: re-read superblock from disk.
623 * constants have been adjusted for ext2
625 if ((error = bread(devvp, SBOFF, SBSIZE, &bp)) != 0)
627 es = (struct ext2_super_block *)bp->b_data;
628 if (ext2_check_sb_compat(es, devvp->v_rdev, 0) != 0) {
630 return (EIO); /* XXX needs translation */
632 fs = VFSTOEXT2(mountp)->um_e2fs;
633 bcopy(bp->b_data, fs->s_es, sizeof(struct ext2_super_block));
635 if((error = compute_sb_data(devvp, es, fs)) != 0) {
640 if (fs->fs_sbsize < SBSIZE)
641 bp->b_flags |= B_INVAL;
647 scaninfo.devvp = devvp;
649 while (error == 0 && scaninfo.rescan) {
651 error = vmntvnodescan(mountp, VMSC_GETVX, ext2_reload_scan1,
652 ext2_reload_scan2, &scaninfo);
658 ext2_reload_scan1(struct mount *mp, struct vnode *vp, void *data)
660 /*struct scaninfo *info = data;*/
666 ext2_reload_scan2(struct mount *mp, struct vnode *vp, void *data)
668 struct scaninfo *info = data;
676 if (vrecycle(vp, curthread))
680 * Step 5: invalidate all cached file data.
682 if (vinvalbuf(vp, 0, info->td, 0, 0))
683 panic("ext2_reload: dirty2");
685 * Step 6: re-read inode data for all active vnodes.
688 error = bread(info->devvp,
689 fsbtodoff(info->fs, ino_to_fsba(info->fs, ip->i_number)),
690 (int)info->fs->s_blocksize, &bp);
693 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data +
694 EXT2_INODE_SIZE * ino_to_fsbo(info->fs, ip->i_number)),
701 * Common code for mount and mountroot
704 ext2_mountfs(struct vnode *devvp, struct mount *mp, struct thread *td)
706 struct ext2mount *ump;
708 struct ext2_sb_info *fs;
709 struct ext2_super_block * es;
711 struct partinfo dpart;
717 * Disallow multiple mounts of the same device.
718 * Disallow mounting of a device that is currently in use
719 * (except for root, which might share swap device for miniroot).
720 * Flush out any old buffers remaining from a previous use.
722 if ((error = vfs_mountedon(devvp)) != 0)
724 if (count_udev(devvp->v_udev) > 0)
726 if ((error = vinvalbuf(devvp, V_SAVE, td, 0, 0)) != 0)
729 /* turn on this to force it to be read-only */
730 mp->mnt_flag |= MNT_RDONLY;
733 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
734 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
735 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, NULL, td);
736 VOP_UNLOCK(devvp, 0, td);
740 if (dev->si_iosize_max != 0)
741 mp->mnt_iosize_max = dev->si_iosize_max;
742 if (mp->mnt_iosize_max > MAXPHYS)
743 mp->mnt_iosize_max = MAXPHYS;
744 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, td) != 0)
748 size = dpart.disklab->d_secsize;
753 if ((error = bread(devvp, SBOFF, SBSIZE, &bp)) != 0)
755 es = (struct ext2_super_block *)bp->b_data;
756 if (ext2_check_sb_compat(es, dev, ronly) != 0) {
757 error = EINVAL; /* XXX needs translation */
760 if ((es->s_state & EXT2_VALID_FS) == 0 ||
761 (es->s_state & EXT2_ERROR_FS)) {
762 if (ronly || (mp->mnt_flag & MNT_FORCE)) {
764 "WARNING: Filesystem was not properly dismounted\n");
767 "WARNING: R/W mount denied. Filesystem is not clean - run fsck\n");
772 ump = bsd_malloc(sizeof *ump, M_EXT2MNT, M_WAITOK);
773 bzero((caddr_t)ump, sizeof *ump);
774 ump->um_malloctype = M_EXT2NODE;
775 ump->um_blkatoff = ext2_blkatoff;
776 ump->um_truncate = ext2_truncate;
777 ump->um_update = ext2_update;
778 ump->um_valloc = ext2_valloc;
779 ump->um_vfree = ext2_vfree;
780 /* I don't know whether this is the right strategy. Note that
781 we dynamically allocate both a ext2_sb_info and a ext2_super_block
782 while Linux keeps the super block in a locked buffer
784 ump->um_e2fs = bsd_malloc(sizeof(struct ext2_sb_info),
785 M_EXT2MNT, M_WAITOK);
786 ump->um_e2fs->s_es = bsd_malloc(sizeof(struct ext2_super_block),
787 M_EXT2MNT, M_WAITOK);
788 bcopy(es, ump->um_e2fs->s_es, (u_int)sizeof(struct ext2_super_block));
789 if ((error = compute_sb_data(devvp, ump->um_e2fs->s_es, ump->um_e2fs)))
792 * We don't free the group descriptors allocated by compute_sb_data()
793 * until ext2_unmount(). This is OK since the mount will succeed.
798 fs->s_rd_only = ronly; /* ronly is set according to mnt_flags */
799 /* if the fs is not mounted read-only, make sure the super block is
800 always written back on a sync()
802 fs->s_wasvalid = fs->s_es->s_state & EXT2_VALID_FS ? 1 : 0;
804 fs->s_dirt = 1; /* mark it modified */
805 fs->s_es->s_state &= ~EXT2_VALID_FS; /* set fs invalid */
807 mp->mnt_data = (qaddr_t)ump;
808 mp->mnt_stat.f_fsid.val[0] = dev2udev(dev);
809 mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
810 mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN;
811 mp->mnt_flag |= MNT_LOCAL;
814 ump->um_devvp = devvp;
815 /* setting those two parameters allows us to use
816 ext2_bmap w/o changse !
818 ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs);
819 ump->um_bptrtodb = fs->s_es->s_log_block_size + 1;
820 ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs);
821 for (i = 0; i < MAXQUOTAS; i++)
822 ump->um_quotas[i] = NULLVP;
823 dev->si_mountpoint = mp;
825 vfs_add_vnodeops(mp, &mp->mnt_vn_norm_ops,
826 ext2_vnodeop_entries, 0);
827 vfs_add_vnodeops(mp, &mp->mnt_vn_spec_ops,
828 ext2_specop_entries, 0);
829 vfs_add_vnodeops(mp, &mp->mnt_vn_fifo_ops,
830 ext2_fifoop_entries, 0);
833 ext2_sbupdate(ump, MNT_WAIT);
838 VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, td);
840 bsd_free(ump->um_e2fs->s_es, M_EXT2MNT);
841 bsd_free(ump->um_e2fs, M_EXT2MNT);
842 bsd_free(ump, M_EXT2MNT);
843 mp->mnt_data = (qaddr_t)0;
849 * unmount system call
852 ext2_unmount(struct mount *mp, int mntflags, struct thread *td)
854 struct ext2mount *ump;
855 struct ext2_sb_info *fs;
856 int error, flags, ronly, i;
859 if (mntflags & MNT_FORCE) {
860 if (mp->mnt_flag & MNT_ROOTFS)
864 if ((error = ext2_flushfiles(mp, flags, td)) != 0)
868 ronly = fs->s_rd_only;
871 fs->s_es->s_state |= EXT2_VALID_FS;
872 ext2_sbupdate(ump, MNT_WAIT);
875 /* release buffers containing group descriptors */
876 for(i = 0; i < fs->s_db_per_group; i++)
877 ULCK_BUF(fs->s_group_desc[i])
878 bsd_free(fs->s_group_desc, M_EXT2MNT);
880 /* release cached inode/block bitmaps */
881 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
882 if (fs->s_inode_bitmap[i])
883 ULCK_BUF(fs->s_inode_bitmap[i])
885 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++)
886 if (fs->s_block_bitmap[i])
887 ULCK_BUF(fs->s_block_bitmap[i])
889 ump->um_devvp->v_rdev->si_mountpoint = NULL;
890 error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD|FWRITE, td);
891 vrele(ump->um_devvp);
892 bsd_free(fs->s_es, M_EXT2MNT);
893 bsd_free(fs, M_EXT2MNT);
894 bsd_free(ump, M_EXT2MNT);
895 mp->mnt_data = (qaddr_t)0;
896 mp->mnt_flag &= ~MNT_LOCAL;
901 * Flush out all the files in a filesystem.
904 ext2_flushfiles(struct mount *mp, int flags, struct thread *td)
906 struct ext2mount *ump;
914 if (mp->mnt_flag & MNT_QUOTA) {
915 if ((error = vflush(mp, 0, SKIPSYSTEM|flags)) != 0)
917 for (i = 0; i < MAXQUOTAS; i++) {
918 if (ump->um_quotas[i] == NULLVP)
920 ext2_quotaoff(td, mp, i);
923 * Here we fall through to vflush again to ensure
924 * that we have gotten rid of all the system vnodes.
928 error = vflush(mp, 0, flags);
933 * Get file system statistics.
934 * taken from ext2/super.c ext2_statfs
937 ext2_statfs(struct mount *mp, struct statfs *sbp, struct thread *td)
939 unsigned long overhead;
940 struct ext2mount *ump;
941 struct ext2_sb_info *fs;
942 struct ext2_super_block *es;
949 if (es->s_magic != EXT2_SUPER_MAGIC)
950 panic("ext2_statfs - magic number spoiled");
953 * Compute the overhead (FS structures)
955 if (es->s_feature_ro_compat & EXT2_FEATURE_RO_COMPAT_SPARSE_SUPER) {
957 for (i = 0 ; i < fs->s_groups_count; i++)
958 if (ext2_group_sparse(i))
961 nsb = fs->s_groups_count;
962 overhead = es->s_first_data_block +
963 /* Superblocks and block group descriptors: */
964 nsb * (1 + fs->s_db_per_group) +
965 /* Inode bitmap, block bitmap, and inode table: */
966 fs->s_groups_count * (1 + 1 + fs->s_itb_per_group);
968 sbp->f_bsize = EXT2_FRAG_SIZE(fs);
969 sbp->f_iosize = EXT2_BLOCK_SIZE(fs);
970 sbp->f_blocks = es->s_blocks_count - overhead;
971 sbp->f_bfree = es->s_free_blocks_count;
972 sbp->f_bavail = sbp->f_bfree - es->s_r_blocks_count;
973 sbp->f_files = es->s_inodes_count;
974 sbp->f_ffree = es->s_free_inodes_count;
975 if (sbp != &mp->mnt_stat) {
976 sbp->f_type = mp->mnt_vfc->vfc_typenum;
977 bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
978 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
984 * Go through the disk queues to initiate sandbagged IO;
985 * go through the inodes to write those that have been modified;
986 * initiate the writing of the super block if it has been modified.
988 * Note: we are always called with the filesystem marked `MPBUSY'.
991 static int ext2_sync_scan(struct mount *mp, struct vnode *vp, void *data);
994 ext2_sync(struct mount *mp, int waitfor, struct thread *td)
996 struct ext2mount *ump = VFSTOEXT2(mp);
997 struct ext2_sb_info *fs;
998 struct scaninfo scaninfo;
1002 if (fs->s_dirt != 0 && fs->s_rd_only != 0) { /* XXX */
1003 printf("fs = %s\n", fs->fs_fsmnt);
1004 panic("ext2_sync: rofs mod");
1008 * Write back each (modified) inode.
1010 scaninfo.allerror = 0;
1011 scaninfo.rescan = 1;
1012 scaninfo.waitfor = waitfor;
1014 while (scaninfo.rescan) {
1015 scaninfo.rescan = 0;
1016 vmntvnodescan(mp, VMSC_GETVP|VMSC_NOWAIT,
1017 NULL, ext2_sync_scan, &scaninfo);
1021 * Force stale file system control information to be flushed.
1023 if (waitfor != MNT_LAZY) {
1024 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY, td);
1025 if ((error = VOP_FSYNC(ump->um_devvp, waitfor, td)) != 0)
1026 scaninfo.allerror = error;
1027 VOP_UNLOCK(ump->um_devvp, 0, td);
1033 * Write back modified superblock.
1035 if (fs->s_dirt != 0) {
1037 fs->s_es->s_wtime = time_second;
1038 if ((error = ext2_sbupdate(ump, waitfor)) != 0)
1039 scaninfo.allerror = error;
1041 return (scaninfo.allerror);
1045 ext2_sync_scan(struct mount *mp, struct vnode *vp, void *data)
1047 struct scaninfo *info = data;
1052 if (vp->v_type == VNON ||
1054 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
1055 (RB_EMPTY(&vp->v_rbdirty_tree) || info->waitfor == MNT_LAZY))) {
1058 if ((error = VOP_FSYNC(vp, info->waitfor, info->td)) != 0)
1059 info->allerror = error;
1064 * Look up a EXT2FS dinode number to find its incore vnode, otherwise read it
1065 * in from disk. If it is in core, wait for the lock bit to clear, then
1066 * return the inode locked. Detection and handling of mount points must be
1067 * done by the calling routine.
1070 ext2_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
1072 struct ext2_sb_info *fs;
1074 struct ext2mount *ump;
1081 ump = VFSTOEXT2(mp);
1084 if ((*vpp = ext2_ihashget(dev, ino)) != NULL)
1088 * Lock out the creation of new entries in the FFS hash table in
1089 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate
1092 if (ext2fs_inode_hash_lock) {
1093 while (ext2fs_inode_hash_lock) {
1094 ext2fs_inode_hash_lock = -1;
1095 tsleep(&ext2fs_inode_hash_lock, 0, "e2vget", 0);
1099 ext2fs_inode_hash_lock = 1;
1102 * If this MALLOC() is performed after the getnewvnode()
1103 * it might block, leaving a vnode with a NULL v_data to be
1104 * found by ext2_sync() if a sync happens to fire right then,
1105 * which will cause a panic because ext2_sync() blindly
1106 * dereferences vp->v_data (as well it should).
1108 MALLOC(ip, struct inode *, sizeof(struct inode), M_EXT2NODE, M_WAITOK);
1110 /* Allocate a new vnode/inode. */
1111 if ((error = getnewvnode(VT_EXT2FS, mp, &vp, 0, LK_CANRECURSE)) != 0) {
1112 if (ext2fs_inode_hash_lock < 0)
1113 wakeup(&ext2fs_inode_hash_lock);
1114 ext2fs_inode_hash_lock = 0;
1116 FREE(ip, M_EXT2NODE);
1119 bzero((caddr_t)ip, sizeof(struct inode));
1122 ip->i_e2fs = fs = ump->um_e2fs;
1126 for (i = 0; i < MAXQUOTAS; i++)
1127 ip->i_dquot[i] = NODQUOT;
1130 * Put it onto its hash chain. Since our vnode is locked, other
1131 * requests for this inode will block if they arrive while we are
1132 * sleeping waiting for old data structures to be purged or for the
1133 * contents of the disk portion of this inode to be read.
1137 if (ext2fs_inode_hash_lock < 0)
1138 wakeup(&ext2fs_inode_hash_lock);
1139 ext2fs_inode_hash_lock = 0;
1141 /* Read in the disk contents for the inode, copy into the inode. */
1143 printf("ext2_vget(%d) dbn= %d ", ino, fsbtodb(fs, ino_to_fsba(fs, ino)));
1145 printf("BREAD MP %p FS %p INODE %d INO_TO_FSBA %d DOFF %lld\n",
1147 (int)ino, (int)ino_to_fsba(fs, ino),
1148 fsbtodoff(fs, ino_to_fsba(fs, ino)));
1150 error = bread(ump->um_devvp, fsbtodoff(fs, ino_to_fsba(fs, ino)),
1151 (int)fs->s_blocksize, &bp);
1154 * The inode does not contain anything useful, so it would
1155 * be misleading to leave it on its hash chain. With mode
1156 * still zero, it will be unlinked and returned to the free
1164 /* convert ext2 inode to dinode */
1165 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data + EXT2_INODE_SIZE *
1166 ino_to_fsbo(fs, ino)), &ip->i_din);
1167 ip->i_block_group = ino_to_cg(fs, ino);
1168 ip->i_next_alloc_block = 0;
1169 ip->i_next_alloc_goal = 0;
1170 ip->i_prealloc_count = 0;
1171 ip->i_prealloc_block = 0;
1172 /* now we want to make sure that block pointers for unused
1173 blocks are zeroed out - ext2_balloc depends on this
1174 although for regular files and directories only
1176 if(S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode)) {
1177 used_blocks = (ip->i_size+fs->s_blocksize-1) / fs->s_blocksize;
1178 for(i = used_blocks; i < EXT2_NDIR_BLOCKS; i++)
1181 ext2_print_inode(ip);
1185 * Initialize the vnode from the inode, check for aliases.
1186 * Note that the underlying vnode may have changed.
1188 if ((error = ext2_vinit(mp, &vp)) != 0) {
1195 * Finish inode initialization now that aliasing has been resolved.
1197 ip->i_devvp = ump->um_devvp;
1200 * Set up a generation number for this inode if it does not
1201 * already have one. This should only happen on old filesystems.
1203 if (ip->i_gen == 0) {
1204 ip->i_gen = random() / 2 + 1;
1205 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1206 ip->i_flag |= IN_MODIFIED;
1209 * Return the locked and refd vnode.
1216 * File handle to vnode
1218 * Have to be really careful about stale file handles:
1219 * - check that the inode number is valid
1220 * - call ext2_vget() to get the locked inode
1221 * - check for an unallocated inode (i_mode == 0)
1222 * - check that the given client host has export rights and return
1223 * those rights via. exflagsp and credanonp
1226 ext2_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
1229 struct ext2_sb_info *fs;
1234 ufhp = (struct ufid *)fhp;
1235 fs = VFSTOEXT2(mp)->um_e2fs;
1236 if (ufhp->ufid_ino < ROOTINO ||
1237 ufhp->ufid_ino > fs->s_groups_count * fs->s_es->s_inodes_per_group)
1240 error = VFS_VGET(mp, ufhp->ufid_ino, &nvp);
1246 if (ip->i_mode == 0 ||
1247 ip->i_gen != ufhp->ufid_gen ||
1248 (VFSTOEXT2(mp)->um_i_effnlink_valid ? ip->i_effnlink :
1249 ip->i_nlink) <= 0) {
1259 * Vnode pointer to File handle
1263 ext2_vptofh(struct vnode *vp, struct fid *fhp)
1269 ufhp = (struct ufid *)fhp;
1270 ufhp->ufid_len = sizeof(struct ufid);
1271 ufhp->ufid_ino = ip->i_number;
1272 ufhp->ufid_gen = ip->i_gen;
1277 * This is the generic part of fhtovp called after the underlying
1278 * filesystem has validated the file handle.
1280 * Verify that a host should have access to a filesystem.
1283 ext2_check_export(struct mount *mp, struct sockaddr *nam, int *exflagsp,
1284 struct ucred **credanonp)
1287 struct ext2mount *ump;;
1289 ump = VFSTOEXT2(mp);
1291 * Get the export permission structure for this <mp, client> tuple.
1293 np = vfs_export_lookup(mp, &ump->um_export, nam);
1297 *exflagsp = np->netc_exflags;
1298 *credanonp = &np->netc_anon;
1303 * Write a superblock and associated information back to disk.
1306 ext2_sbupdate(struct ext2mount *mp, int waitfor)
1308 struct ext2_sb_info *fs = mp->um_e2fs;
1309 struct ext2_super_block *es = fs->s_es;
1313 printf("\nupdating superblock, waitfor=%s\n", waitfor == MNT_WAIT ? "yes":"no");
1315 bp = getblk(mp->um_devvp, SBOFF, SBSIZE, 0, 0);
1316 bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2_super_block));
1317 if (waitfor == MNT_WAIT)
1323 * The buffers for group descriptors, inode bitmaps and block bitmaps
1324 * are not busy at this point and are (hopefully) written by the
1325 * usual sync mechanism. No need to write them here