Home | History | Annotate | Line # | Download | only in ext2fs
ext2fs_vfsops.c revision 1.8
      1 /*	$NetBSD: ext2fs_vfsops.c,v 1.8 1998/02/18 17:02:20 drochner Exp $	*/
      2 
      3 /*
      4  * Copyright (c) 1997 Manuel Bouyer.
      5  * Copyright (c) 1989, 1991, 1993, 1994
      6  *	The Regents of the University of California.  All rights reserved.
      7  *
      8  * Redistribution and use in source and binary forms, with or without
      9  * modification, are permitted provided that the following conditions
     10  * are met:
     11  * 1. Redistributions of source code must retain the above copyright
     12  *	notice, this list of conditions and the following disclaimer.
     13  * 2. Redistributions in binary form must reproduce the above copyright
     14  *	notice, this list of conditions and the following disclaimer in the
     15  *	documentation and/or other materials provided with the distribution.
     16  * 3. All advertising materials mentioning features or use of this software
     17  *	must display the following acknowledgement:
     18  *	This product includes software developed by the University of
     19  *	California, Berkeley and its contributors.
     20  * 4. Neither the name of the University nor the names of its contributors
     21  *	may be used to endorse or promote products derived from this software
     22  *	without specific prior written permission.
     23  *
     24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     34  * SUCH DAMAGE.
     35  *
     36  *	@(#)ffs_vfsops.c	8.14 (Berkeley) 11/28/94
     37  * Modified for ext2fs by Manuel Bouyer.
     38  */
     39 
     40 #include <sys/param.h>
     41 #include <sys/systm.h>
     42 #include <sys/namei.h>
     43 #include <sys/proc.h>
     44 #include <sys/kernel.h>
     45 #include <sys/vnode.h>
     46 #include <sys/socket.h>
     47 #include <sys/mount.h>
     48 #include <sys/buf.h>
     49 #include <sys/device.h>
     50 #include <sys/mbuf.h>
     51 #include <sys/file.h>
     52 #include <sys/disklabel.h>
     53 #include <sys/ioctl.h>
     54 #include <sys/errno.h>
     55 #include <sys/malloc.h>
     56 #include <sys/lock.h>
     57 
     58 #include <miscfs/specfs/specdev.h>
     59 
     60 #include <ufs/ufs/quota.h>
     61 #include <ufs/ufs/ufsmount.h>
     62 #include <ufs/ufs/inode.h>
     63 #include <ufs/ufs/dir.h>
     64 #include <ufs/ufs/ufs_extern.h>
     65 
     66 #include <ufs/ext2fs/ext2fs.h>
     67 #include <ufs/ext2fs/ext2fs_extern.h>
     68 
     69 extern struct lock ufs_hashlock;
     70 
     71 int ext2fs_sbupdate __P((struct ufsmount *, int));
     72 int ext2fs_check_export __P((struct mount *, struct ufid *, struct mbuf *,
     73 	struct vnode **, int *, struct ucred **));
     74 
     75 extern struct vnodeopv_desc ext2fs_vnodeop_opv_desc;
     76 extern struct vnodeopv_desc ext2fs_specop_opv_desc;
     77 #ifdef FIFO
     78 extern struct vnodeopv_desc ext2fs_fifoop_opv_desc;
     79 #endif
     80 
     81 struct vnodeopv_desc *ext2fs_vnodeopv_descs[] = {
     82 	&ext2fs_vnodeop_opv_desc,
     83 	&ext2fs_specop_opv_desc,
     84 #ifdef FIFO
     85 	&ext2fs_fifoop_opv_desc,
     86 #endif
     87 	NULL,
     88 };
     89 
     90 struct vfsops ext2fs_vfsops = {
     91 	MOUNT_EXT2FS,
     92 	ext2fs_mount,
     93 	ufs_start,
     94 	ext2fs_unmount,
     95 	ufs_root,
     96 	ufs_quotactl,
     97 	ext2fs_statfs,
     98 	ext2fs_sync,
     99 	ext2fs_vget,
    100 	ext2fs_fhtovp,
    101 	ext2fs_vptofh,
    102 	ext2fs_init,
    103 	ext2fs_mountroot,
    104 	ext2fs_vnodeopv_descs,
    105 };
    106 
    107 extern u_long ext2gennumber;
    108 
    109 /*
    110  * This is the generic part of fhtovp called after the underlying
    111  * filesystem has validated the file handle.
    112  *
    113  * Verify that a host should have access to a filesystem, and if so
    114  * return a vnode for the presented file handle.
    115  */
    116 int
    117 ext2fs_check_export(mp, ufhp, nam, vpp, exflagsp, credanonp)
    118 	register struct mount *mp;
    119 	struct ufid *ufhp;
    120 	struct mbuf *nam;
    121 	struct vnode **vpp;
    122 	int *exflagsp;
    123 	struct ucred **credanonp;
    124 {
    125 	register struct inode *ip;
    126 	register struct netcred *np;
    127 	register struct ufsmount *ump = VFSTOUFS(mp);
    128 	struct vnode *nvp;
    129 	int error;
    130 
    131 	/*
    132 	 * Get the export permission structure for this <mp, client> tuple.
    133 	 */
    134 	np = vfs_export_lookup(mp, &ump->um_export, nam);
    135 	if (np == NULL)
    136 		return (EACCES);
    137 
    138 	if ((error = VFS_VGET(mp, ufhp->ufid_ino, &nvp)) != 0) {
    139 		*vpp = NULLVP;
    140 		return (error);
    141 	}
    142 	ip = VTOI(nvp);
    143 	if (ip->i_e2fs_mode == 0 || ip->i_e2fs_dtime != 0 ||
    144 		ip->i_e2fs_gen != ufhp->ufid_gen) {
    145 		vput(nvp);
    146 		*vpp = NULLVP;
    147 		return (ESTALE);
    148 	}
    149 	*vpp = nvp;
    150 	*exflagsp = np->netc_exflags;
    151 	*credanonp = &np->netc_anon;
    152 	return (0);
    153 }
    154 
    155 
    156 /*
    157  * Called by main() when ext2fs is going to be mounted as root.
    158  *
    159  * Name is updated by mount(8) after booting.
    160  */
    161 #define ROOTNAME	"root_device"
    162 
    163 int
    164 ext2fs_mountroot()
    165 {
    166 	extern struct vnode *rootvp;
    167 	register struct m_ext2fs *fs;
    168 	register struct mount *mp;
    169 	struct proc *p = curproc;	/* XXX */
    170 	struct ufsmount *ump;
    171 	size_t size;
    172 	int error;
    173 
    174 	if (root_device->dv_class != DV_DISK)
    175 		return (ENODEV);
    176 
    177 	/*
    178 	 * Get vnodes for rootdev.
    179 	 */
    180 	if (bdevvp(rootdev, &rootvp))
    181 		panic("ext2fs_mountroot: can't setup bdevvp's");
    182 
    183 	mp = malloc(sizeof(struct mount), M_MOUNT, M_WAITOK);
    184 	bzero((char *)mp, sizeof(struct mount));
    185 	mp->mnt_op = &ext2fs_vfsops;
    186 	mp->mnt_flag = MNT_RDONLY;
    187 	if ((error = ext2fs_mountfs(rootvp, mp, p)) != 0) {
    188 		free(mp, M_MOUNT);
    189 		return (error);
    190 	}
    191 	if ((error = vfs_lock(mp)) != 0) {
    192 		(void)ext2fs_unmount(mp, 0, p);
    193 		free(mp, M_MOUNT);
    194 		return (error);
    195 	}
    196 	CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
    197 	mp->mnt_vnodecovered = NULLVP;
    198 	ump = VFSTOUFS(mp);
    199 	fs = ump->um_e2fs;
    200 	bzero(fs->e2fs_fsmnt, sizeof(fs->e2fs_fsmnt));
    201 	fs->e2fs_fsmnt[0] = '/';
    202 	bcopy(fs->e2fs_fsmnt, mp->mnt_stat.f_mntonname, MNAMELEN);
    203 	(void) copystr(ROOTNAME, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
    204 		&size);
    205 	bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
    206 	(void)ext2fs_statfs(mp, &mp->mnt_stat, p);
    207 	vfs_unlock(mp);
    208 	inittodr(fs->e2fs.e2fs_wtime);
    209 	return (0);
    210 }
    211 
    212 /*
    213  * VFS Operations.
    214  *
    215  * mount system call
    216  */
    217 int
    218 ext2fs_mount(mp, path, data, ndp, p)
    219 	register struct mount *mp;
    220 	const char *path;
    221 	void * data;
    222 	struct nameidata *ndp;
    223 	struct proc *p;
    224 {
    225 	struct vnode *devvp;
    226 	struct ufs_args args;
    227 	struct ufsmount *ump = NULL;
    228 	register struct m_ext2fs *fs;
    229 	size_t size;
    230 	int error, flags;
    231 	mode_t accessmode;
    232 
    233 	error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args));
    234 	if (error)
    235 		return (error);
    236 	/*
    237 	 * If updating, check whether changing from read-only to
    238 	 * read/write; if there is no device name, that's all we do.
    239 	 */
    240 	if (mp->mnt_flag & MNT_UPDATE) {
    241 		ump = VFSTOUFS(mp);
    242 		fs = ump->um_e2fs;
    243 		if (fs->e2fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
    244 			flags = WRITECLOSE;
    245 			if (mp->mnt_flag & MNT_FORCE)
    246 				flags |= FORCECLOSE;
    247 			if (vfs_busy(mp))
    248 				return (EBUSY);
    249 			error = ext2fs_flushfiles(mp, flags, p);
    250 			if (error == 0 &&
    251 				ext2fs_cgupdate(ump, MNT_WAIT) == 0 &&
    252 				(fs->e2fs.e2fs_state & E2FS_ERRORS) == 0) {
    253 				fs->e2fs.e2fs_state = E2FS_ISCLEAN;
    254 				(void) ext2fs_sbupdate(ump, MNT_WAIT);
    255 			}
    256 			vfs_unbusy(mp);
    257 			if (error)
    258 				return (error);
    259 			fs->e2fs_ronly = 1;
    260 		}
    261 		if (mp->mnt_flag & MNT_RELOAD) {
    262 			error = ext2fs_reload(mp, ndp->ni_cnd.cn_cred, p);
    263 			if (error)
    264 				return (error);
    265 		}
    266 		if (fs->e2fs_ronly && (mp->mnt_flag & MNT_WANTRDWR)) {
    267 			/*
    268 			 * If upgrade to read-write by non-root, then verify
    269 			 * that user has necessary permissions on the device.
    270 			 */
    271 			if (p->p_ucred->cr_uid != 0) {
    272 				devvp = ump->um_devvp;
    273 				VOP_LOCK(devvp);
    274 				error = VOP_ACCESS(devvp, VREAD | VWRITE,
    275 						   p->p_ucred, p);
    276 				if (error) {
    277 					VOP_UNLOCK(devvp);
    278 					return (error);
    279 				}
    280 				VOP_UNLOCK(devvp);
    281 			}
    282 			fs->e2fs_ronly = 0;
    283 			if (fs->e2fs.e2fs_state == E2FS_ISCLEAN)
    284 				fs->e2fs.e2fs_state = 0;
    285 			else
    286 				fs->e2fs.e2fs_state = E2FS_ERRORS;
    287 			fs->e2fs_fmod = 1;
    288 		}
    289 		if (args.fspec == 0) {
    290 			/*
    291 			 * Process export requests.
    292 			 */
    293 			return (vfs_export(mp, &ump->um_export, &args.export));
    294 		}
    295 	}
    296 	/*
    297 	 * Not an update, or updating the name: look up the name
    298 	 * and verify that it refers to a sensible block device.
    299 	 */
    300 	NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, p);
    301 	if ((error = namei(ndp)) != 0)
    302 		return (error);
    303 	devvp = ndp->ni_vp;
    304 
    305 	if (devvp->v_type != VBLK) {
    306 		vrele(devvp);
    307 		return (ENOTBLK);
    308 	}
    309 	if (major(devvp->v_rdev) >= nblkdev) {
    310 		vrele(devvp);
    311 		return (ENXIO);
    312 	}
    313 	/*
    314 	 * If mount by non-root, then verify that user has necessary
    315 	 * permissions on the device.
    316 	 */
    317 	if (p->p_ucred->cr_uid != 0) {
    318 		accessmode = VREAD;
    319 		if ((mp->mnt_flag & MNT_RDONLY) == 0)
    320 			accessmode |= VWRITE;
    321 		VOP_LOCK(devvp);
    322 		error = VOP_ACCESS(devvp, accessmode, p->p_ucred, p);
    323 		if (error) {
    324 			vput(devvp);
    325 			return (error);
    326 		}
    327 		VOP_UNLOCK(devvp);
    328 	}
    329 	if ((mp->mnt_flag & MNT_UPDATE) == 0)
    330 		error = ext2fs_mountfs(devvp, mp, p);
    331 	else {
    332 		if (devvp != ump->um_devvp)
    333 			error = EINVAL;	/* needs translation */
    334 		else
    335 			vrele(devvp);
    336 	}
    337 	if (error) {
    338 		vrele(devvp);
    339 		return (error);
    340 	}
    341 	ump = VFSTOUFS(mp);
    342 	fs = ump->um_e2fs;
    343 	(void) copyinstr(path, fs->e2fs_fsmnt, sizeof(fs->e2fs_fsmnt) - 1, &size);
    344 	bzero(fs->e2fs_fsmnt + size, sizeof(fs->e2fs_fsmnt) - size);
    345 	bcopy(fs->e2fs_fsmnt, mp->mnt_stat.f_mntonname, MNAMELEN);
    346 	(void) copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
    347 		&size);
    348 	bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
    349 	if (fs->e2fs_fmod != 0) {	/* XXX */
    350 		fs->e2fs_fmod = 0;
    351 		if (fs->e2fs.e2fs_state == 0)
    352 			fs->e2fs.e2fs_wtime = time.tv_sec;
    353 		else
    354 			printf("%s: file system not clean; please fsck(8)\n",
    355 				mp->mnt_stat.f_mntfromname);
    356 		(void) ext2fs_cgupdate(ump, MNT_WAIT);
    357 	}
    358 	return (0);
    359 }
    360 
    361 /*
    362  * Reload all incore data for a filesystem (used after running fsck on
    363  * the root filesystem and finding things to fix). The filesystem must
    364  * be mounted read-only.
    365  *
    366  * Things to do to update the mount:
    367  *	1) invalidate all cached meta-data.
    368  *	2) re-read superblock from disk.
    369  *	3) re-read summary information from disk.
    370  *	4) invalidate all inactive vnodes.
    371  *	5) invalidate all cached file data.
    372  *	6) re-read inode data for all active vnodes.
    373  */
    374 int
    375 ext2fs_reload(mountp, cred, p)
    376 	register struct mount *mountp;
    377 	struct ucred *cred;
    378 	struct proc *p;
    379 {
    380 	register struct vnode *vp, *nvp, *devvp;
    381 	struct inode *ip;
    382 	struct buf *bp;
    383 	struct m_ext2fs *fs;
    384 	struct ext2fs *newfs;
    385 	struct partinfo dpart;
    386 	int i, size, error;
    387 
    388 	if ((mountp->mnt_flag & MNT_RDONLY) == 0)
    389 		return (EINVAL);
    390 	/*
    391 	 * Step 1: invalidate all cached meta-data.
    392 	 */
    393 	devvp = VFSTOUFS(mountp)->um_devvp;
    394 	if (vinvalbuf(devvp, 0, cred, p, 0, 0))
    395 		panic("ext2fs_reload: dirty1");
    396 	/*
    397 	 * Step 2: re-read superblock from disk.
    398 	 */
    399 	if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, p) != 0)
    400 		size = DEV_BSIZE;
    401 	else
    402 		size = dpart.disklab->d_secsize;
    403 	error = bread(devvp, (daddr_t)(SBOFF / size), SBSIZE, NOCRED, &bp);
    404 	if (error)
    405 		return (error);
    406 	newfs = (struct ext2fs *)bp->b_data;
    407 	if (fs2h16(newfs->e2fs_magic) != E2FS_MAGIC ||
    408 		fs2h32(newfs->e2fs_rev) != E2FS_REV) {
    409 #ifdef DIAGNOSTIC
    410 		printf("Wrong magic number: %x (expected %x for ext2 fs)",
    411 					fs2h16(newfs->e2fs_magic), E2FS_MAGIC);
    412 		printf("or wrong revision number: %x (expected %x for ext2 fs)",
    413 					fs2h32(newfs->e2fs_rev), E2FS_REV);
    414 #endif
    415 		brelse(bp);
    416 		return (EIO);		/* XXX needs translation */
    417 	}
    418 	if (fs2h32(newfs->e2fs_log_bsize) > 2) { /* block size = 1024|2048|4096 */
    419 #ifdef DIAGNOSTIC
    420 		printf("wrong block size: %d (expected <=2 for ext2 fs)\n",
    421 			fs2h32(newfs->e2fs_log_bsize));
    422 #endif
    423 		brelse(bp);
    424 		return (EIO);	   /* XXX needs translation */
    425 	}
    426 
    427 
    428 	fs = VFSTOUFS(mountp)->um_e2fs;
    429 	/*
    430 	 * copy in new superblock, and compute in-memory values
    431 	 */
    432 	e2fs_sbload(newfs, &fs->e2fs);
    433 	fs->e2fs_ncg = howmany(fs->e2fs.e2fs_bcount - fs->e2fs.e2fs_first_dblock,
    434 		fs->e2fs.e2fs_bpg);
    435 	/* XXX assume hw bsize = 512 */
    436 	fs->e2fs_fsbtodb = fs->e2fs.e2fs_log_bsize + 1;
    437 	fs->e2fs_bsize = 1024 << fs->e2fs.e2fs_log_bsize;
    438 	fs->e2fs_bshift = LOG_MINBSIZE + fs->e2fs.e2fs_log_bsize;
    439 	fs->e2fs_qbmask = fs->e2fs_bsize - 1;
    440 	fs->e2fs_bmask = ~fs->e2fs_qbmask;
    441 	fs->e2fs_ngdb = howmany(fs->e2fs_ncg,
    442 			fs->e2fs_bsize / sizeof(struct ext2_gd));
    443 	fs->e2fs_ipb = fs->e2fs_bsize / sizeof(struct ext2fs_dinode);
    444 	fs->e2fs_itpg = fs->e2fs.e2fs_ipg/fs->e2fs_ipb;
    445 
    446 	/*
    447 	 * Step 3: re-read summary information from disk.
    448 	 */
    449 
    450 	for (i=0; i < fs->e2fs_ngdb; i++) {
    451 		error = bread(devvp , fsbtodb(fs, ((fs->e2fs_bsize>1024)?0:1)+i+1),
    452 				fs->e2fs_bsize, NOCRED, &bp);
    453 		if (error)
    454 			return (error);
    455 		e2fs_cgload((struct ext2_gd*)bp->b_data,
    456 			&fs->e2fs_gd[i* fs->e2fs_bsize / sizeof(struct ext2_gd)],
    457 			fs->e2fs_bsize);
    458 		brelse(bp);
    459 	}
    460 
    461 loop:
    462 	for (vp = mountp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) {
    463 		nvp = vp->v_mntvnodes.le_next;
    464 		/*
    465 		 * Step 4: invalidate all inactive vnodes.
    466 		 */
    467 		if (vp->v_usecount == 0) {
    468 			vgone(vp);
    469 			continue;
    470 		}
    471 		/*
    472 		 * Step 5: invalidate all cached file data.
    473 		 */
    474 		if (vget(vp, 1))
    475 			goto loop;
    476 		if (vinvalbuf(vp, 0, cred, p, 0, 0))
    477 			panic("ext2fs_reload: dirty2");
    478 		/*
    479 		 * Step 6: re-read inode data for all active vnodes.
    480 		 */
    481 		ip = VTOI(vp);
    482 		error = bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
    483 				  (int)fs->e2fs_bsize, NOCRED, &bp);
    484 		if (error) {
    485 			vput(vp);
    486 			return (error);
    487 		}
    488 		e2fs_iload((struct ext2fs_dinode *)bp->b_data +
    489 			ino_to_fsbo(fs, ip->i_number),
    490 					&ip->i_din.e2fs_din);
    491 		brelse(bp);
    492 		vput(vp);
    493 		if (vp->v_mount != mountp)
    494 			goto loop;
    495 	}
    496 	return (0);
    497 }
    498 
    499 /*
    500  * Common code for mount and mountroot
    501  */
    502 int
    503 ext2fs_mountfs(devvp, mp, p)
    504 	register struct vnode *devvp;
    505 	struct mount *mp;
    506 	struct proc *p;
    507 {
    508 	register struct ufsmount *ump;
    509 	struct buf *bp;
    510 	register struct ext2fs *fs;
    511 	register struct m_ext2fs *m_fs;
    512 	dev_t dev;
    513 	struct partinfo dpart;
    514 	int error, i, size, ronly;
    515 	struct ucred *cred;
    516 	extern struct vnode *rootvp;
    517 
    518 	dev = devvp->v_rdev;
    519 	cred = p ? p->p_ucred : NOCRED;
    520 	/*
    521 	 * Disallow multiple mounts of the same device.
    522 	 * Disallow mounting of a device that is currently in use
    523 	 * (except for root, which might share swap device for miniroot).
    524 	 * Flush out any old buffers remaining from a previous use.
    525 	 */
    526 	if ((error = vfs_mountedon(devvp)) != 0)
    527 		return (error);
    528 	if (vcount(devvp) > 1 && devvp != rootvp)
    529 		return (EBUSY);
    530 	if ((error = vinvalbuf(devvp, V_SAVE, cred, p, 0, 0)) != 0)
    531 		return (error);
    532 
    533 	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
    534 	error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p);
    535 	if (error)
    536 		return (error);
    537 	if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, cred, p) != 0)
    538 		size = DEV_BSIZE;
    539 	else
    540 		size = dpart.disklab->d_secsize;
    541 
    542 	bp = NULL;
    543 	ump = NULL;
    544 
    545 #ifdef DEBUG_EXT2
    546 	printf("sb size: %d ino size %d\n", sizeof(struct ext2fs),
    547 			sizeof(struct ext2fs_dinode));
    548 #endif
    549 	error = bread(devvp, (SBOFF / DEV_BSIZE), SBSIZE, cred, &bp);
    550 	if (error)
    551 		goto out;
    552 	fs = (struct ext2fs *)bp->b_data;
    553 	if (fs2h16(fs->e2fs_magic) != E2FS_MAGIC ||
    554 		fs2h32(fs->e2fs_rev) != E2FS_REV) {
    555 #ifdef DIAGNOSTIC
    556 		printf("Wrong magic number: %x (expected %x for ext2 fs)",
    557 			fs2h16(fs->e2fs_magic), E2FS_MAGIC);
    558 		printf(" or wrong revision number: %x (expected %x for ext2 fs)\n",
    559 			fs2h32(fs->e2fs_rev), E2FS_REV);
    560 #endif
    561 		error = EINVAL;		/* XXX needs translation */
    562 		goto out;
    563 	}
    564 
    565 	if (fs2h32(fs->e2fs_log_bsize) > 2) { /* block size = 1024|2048|4096 */
    566 #ifdef DIAGNOSTIC
    567 		printf("wrong block size: %d (expected <2 for ext2 fs)\n",
    568 			fs2h32(fs->e2fs_log_bsize));
    569 #endif
    570 		error = EINVAL;	 /* XXX needs translation */
    571 		goto out;
    572 	}
    573 
    574 	ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
    575 	bzero((caddr_t)ump, sizeof *ump);
    576 	ump->um_e2fs = malloc(sizeof(struct m_ext2fs), M_UFSMNT, M_WAITOK);
    577 	e2fs_sbload((struct ext2fs*)bp->b_data, &ump->um_e2fs->e2fs);
    578 	brelse(bp);
    579 	bp = NULL;
    580 	m_fs = ump->um_e2fs;
    581 	m_fs->e2fs_ronly = ronly;
    582 	if (ronly == 0) {
    583 		if (m_fs->e2fs.e2fs_state == E2FS_ISCLEAN)
    584 			m_fs->e2fs.e2fs_state = 0;
    585 		else
    586 			m_fs->e2fs.e2fs_state = E2FS_ERRORS;
    587 		m_fs->e2fs_fmod = 1;
    588 	}
    589 
    590 	/* compute dynamic sb infos */
    591 	m_fs->e2fs_ncg =
    592 		howmany(m_fs->e2fs.e2fs_bcount - m_fs->e2fs.e2fs_first_dblock,
    593 		m_fs->e2fs.e2fs_bpg);
    594 	/* XXX assume hw bsize = 512 */
    595 	m_fs->e2fs_fsbtodb = m_fs->e2fs.e2fs_log_bsize + 1;
    596 	m_fs->e2fs_bsize = 1024 << m_fs->e2fs.e2fs_log_bsize;
    597 	m_fs->e2fs_bshift = LOG_MINBSIZE + m_fs->e2fs.e2fs_log_bsize;
    598 	m_fs->e2fs_qbmask = m_fs->e2fs_bsize - 1;
    599 	m_fs->e2fs_bmask = ~m_fs->e2fs_qbmask;
    600 	m_fs->e2fs_ngdb = howmany(m_fs->e2fs_ncg,
    601 		m_fs->e2fs_bsize / sizeof(struct ext2_gd));
    602 	m_fs->e2fs_ipb = m_fs->e2fs_bsize / sizeof(struct ext2fs_dinode);
    603 	m_fs->e2fs_itpg = m_fs->e2fs.e2fs_ipg/m_fs->e2fs_ipb;
    604 
    605 	m_fs->e2fs_gd = malloc(m_fs->e2fs_ngdb * m_fs->e2fs_bsize,
    606 		M_UFSMNT, M_WAITOK);
    607 	for (i=0; i < m_fs->e2fs_ngdb; i++) {
    608 		error = bread(devvp , fsbtodb(m_fs, ((m_fs->e2fs_bsize>1024)?0:1)+i+1),
    609 			m_fs->e2fs_bsize, NOCRED, &bp);
    610 		if (error) {
    611 			free(m_fs->e2fs_gd, M_UFSMNT);
    612 			goto out;
    613 		}
    614 		e2fs_cgload((struct ext2_gd*)bp->b_data,
    615 			&m_fs->e2fs_gd[i* m_fs->e2fs_bsize / sizeof(struct ext2_gd)],
    616 			m_fs->e2fs_bsize);
    617 		brelse(bp);
    618 		bp = NULL;
    619 	}
    620 
    621 	mp->mnt_data = (qaddr_t)ump;
    622 	mp->mnt_stat.f_fsid.val[0] = (long)dev;
    623 	mp->mnt_stat.f_fsid.val[1] = makefstype(MOUNT_EXT2FS);
    624 	mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN;
    625 	mp->mnt_flag |= MNT_LOCAL;
    626 	ump->um_mountp = mp;
    627 	ump->um_dev = dev;
    628 	ump->um_devvp = devvp;
    629 	ump->um_nindir = NINDIR(m_fs);
    630 	ump->um_bptrtodb = m_fs->e2fs_fsbtodb;
    631 	ump->um_seqinc = 1; /* no frags */
    632 	devvp->v_specflags |= SI_MOUNTEDON;
    633 	return (0);
    634 out:
    635 	if (bp)
    636 		brelse(bp);
    637 	(void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, cred, p);
    638 	if (ump) {
    639 		free(ump->um_e2fs, M_UFSMNT);
    640 		free(ump, M_UFSMNT);
    641 		mp->mnt_data = (qaddr_t)0;
    642 	}
    643 	return (error);
    644 }
    645 
    646 /*
    647  * unmount system call
    648  */
    649 int
    650 ext2fs_unmount(mp, mntflags, p)
    651 	struct mount *mp;
    652 	int mntflags;
    653 	struct proc *p;
    654 {
    655 	register struct ufsmount *ump;
    656 	register struct m_ext2fs *fs;
    657 	int error, flags;
    658 
    659 	flags = 0;
    660 	if (mntflags & MNT_FORCE)
    661 		flags |= FORCECLOSE;
    662 	if ((error = ext2fs_flushfiles(mp, flags, p)) != 0)
    663 		return (error);
    664 	ump = VFSTOUFS(mp);
    665 	fs = ump->um_e2fs;
    666 	if (fs->e2fs_ronly == 0 &&
    667 		ext2fs_cgupdate(ump, MNT_WAIT) == 0 &&
    668 		(fs->e2fs.e2fs_state & E2FS_ERRORS) == 0) {
    669 		fs->e2fs.e2fs_state = E2FS_ISCLEAN;
    670 		(void) ext2fs_sbupdate(ump, MNT_WAIT);
    671 	}
    672 	ump->um_devvp->v_specflags &= ~SI_MOUNTEDON;
    673 	error = VOP_CLOSE(ump->um_devvp, fs->e2fs_ronly ? FREAD : FREAD|FWRITE,
    674 		NOCRED, p);
    675 	vrele(ump->um_devvp);
    676 	free(fs->e2fs_gd, M_UFSMNT);
    677 	free(fs, M_UFSMNT);
    678 	free(ump, M_UFSMNT);
    679 	mp->mnt_data = (qaddr_t)0;
    680 	mp->mnt_flag &= ~MNT_LOCAL;
    681 	return (error);
    682 }
    683 
    684 /*
    685  * Flush out all the files in a filesystem.
    686  */
    687 int
    688 ext2fs_flushfiles(mp, flags, p)
    689 	register struct mount *mp;
    690 	int flags;
    691 	struct proc *p;
    692 {
    693 	extern int doforce;
    694 	register struct ufsmount *ump;
    695 	int error;
    696 
    697 	if (!doforce)
    698 		flags &= ~FORCECLOSE;
    699 	ump = VFSTOUFS(mp);
    700 	error = vflush(mp, NULLVP, flags);
    701 	return (error);
    702 }
    703 
    704 /*
    705  * Get file system statistics.
    706  */
    707 int
    708 ext2fs_statfs(mp, sbp, p)
    709 	struct mount *mp;
    710 	register struct statfs *sbp;
    711 	struct proc *p;
    712 {
    713 	register struct ufsmount *ump;
    714 	register struct m_ext2fs *fs;
    715 	u_int32_t overhead, overhead_per_group;
    716 
    717 	ump = VFSTOUFS(mp);
    718 	fs = ump->um_e2fs;
    719 	if (fs->e2fs.e2fs_magic != E2FS_MAGIC)
    720 		panic("ext2fs_statfs");
    721 #ifdef COMPAT_09
    722 	sbp->f_type = 1;
    723 #else
    724 	sbp->f_type = 0;
    725 #endif
    726 
    727 	/*
    728 	 * Compute the overhead (FS structures)
    729 	 */
    730 	overhead_per_group = 1 /* super block */ +
    731 						 fs->e2fs_ngdb +
    732 						 1 /* block bitmap */ +
    733 						 1 /* inode bitmap */ +
    734 						 fs->e2fs_itpg;
    735 	overhead = fs->e2fs.e2fs_first_dblock +
    736 		   fs->e2fs_ncg * overhead_per_group;
    737 
    738 
    739 	sbp->f_bsize = fs->e2fs_bsize;
    740 	sbp->f_iosize = fs->e2fs_bsize;
    741 	sbp->f_blocks = fs->e2fs.e2fs_bcount - overhead;
    742 	sbp->f_bfree = fs->e2fs.e2fs_fbcount;
    743 	sbp->f_bavail = sbp->f_bfree - fs->e2fs.e2fs_rbcount;
    744 	sbp->f_files =  fs->e2fs.e2fs_icount;
    745 	sbp->f_ffree = fs->e2fs.e2fs_ficount;
    746 	if (sbp != &mp->mnt_stat) {
    747 		bcopy(mp->mnt_stat.f_mntonname, sbp->f_mntonname, MNAMELEN);
    748 		bcopy(mp->mnt_stat.f_mntfromname, sbp->f_mntfromname, MNAMELEN);
    749 	}
    750 	strncpy(sbp->f_fstypename, mp->mnt_op->vfs_name, MFSNAMELEN);
    751 	return (0);
    752 }
    753 
    754 /*
    755  * Go through the disk queues to initiate sandbagged IO;
    756  * go through the inodes to write those that have been modified;
    757  * initiate the writing of the super block if it has been modified.
    758  *
    759  * Note: we are always called with the filesystem marked `MPBUSY'.
    760  */
    761 int
    762 ext2fs_sync(mp, waitfor, cred, p)
    763 	struct mount *mp;
    764 	int waitfor;
    765 	struct ucred *cred;
    766 	struct proc *p;
    767 {
    768 	register struct vnode *vp;
    769 	register struct inode *ip;
    770 	register struct ufsmount *ump = VFSTOUFS(mp);
    771 	register struct m_ext2fs *fs;
    772 	int error, allerror = 0;
    773 
    774 	fs = ump->um_e2fs;
    775 	/*
    776 	 * Write back modified superblock.
    777 	 * Consistency check that the superblock
    778 	 * is still in the buffer cache.
    779 	 */
    780 	if (fs->e2fs_fmod != 0) {
    781 		if (fs->e2fs_ronly != 0) {		/* XXX */
    782 			printf("fs = %s\n", fs->e2fs_fsmnt);
    783 			panic("update: rofs mod");
    784 		}
    785 		fs->e2fs_fmod = 0;
    786 		fs->e2fs.e2fs_wtime = time.tv_sec;
    787 		allerror = ext2fs_cgupdate(ump, waitfor);
    788 	}
    789 	/*
    790 	 * Write back each (modified) inode.
    791 	 */
    792 loop:
    793 	for (vp = mp->mnt_vnodelist.lh_first;
    794 		 vp != NULL;
    795 		 vp = vp->v_mntvnodes.le_next) {
    796 		/*
    797 		 * If the vnode that we are about to sync is no longer
    798 		 * associated with this mount point, start over.
    799 		 */
    800 		if (vp->v_mount != mp)
    801 			goto loop;
    802 		if (VOP_ISLOCKED(vp))
    803 			continue;
    804 		ip = VTOI(vp);
    805 		if ((ip->i_flag &
    806 			(IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
    807 			vp->v_dirtyblkhd.lh_first == NULL)
    808 			continue;
    809 		if (vget(vp, 1))
    810 			goto loop;
    811 		if ((error = VOP_FSYNC(vp, cred, waitfor, p)) != 0)
    812 			allerror = error;
    813 		vput(vp);
    814 	}
    815 	/*
    816 	 * Force stale file system control information to be flushed.
    817 	 */
    818 	if ((error = VOP_FSYNC(ump->um_devvp, cred, waitfor, p)) != 0)
    819 		allerror = error;
    820 	return (allerror);
    821 }
    822 
    823 /*
    824  * Look up a EXT2FS dinode number to find its incore vnode, otherwise read it
    825  * in from disk.  If it is in core, wait for the lock bit to clear, then
    826  * return the inode locked.  Detection and handling of mount points must be
    827  * done by the calling routine.
    828  */
    829 int
    830 ext2fs_vget(mp, ino, vpp)
    831 	struct mount *mp;
    832 	ino_t ino;
    833 	struct vnode **vpp;
    834 {
    835 	register struct m_ext2fs *fs;
    836 	register struct inode *ip;
    837 	struct ufsmount *ump;
    838 	struct buf *bp;
    839 	struct vnode *vp;
    840 	dev_t dev;
    841 	int error;
    842 
    843 	ump = VFSTOUFS(mp);
    844 	dev = ump->um_dev;
    845 	do {
    846 		if ((*vpp = ufs_ihashget(dev, ino)) != NULL)
    847 			return (0);
    848 	} while (lockmgr(&ufs_hashlock, LK_EXCLUSIVE|LK_SLEEPFAIL, 0, curproc));
    849 
    850 	/* Allocate a new vnode/inode. */
    851 	if ((error = getnewvnode(VT_EXT2FS, mp, ext2fs_vnodeop_p, &vp)) != 0) {
    852 		*vpp = NULL;
    853 		lockmgr(&ufs_hashlock, LK_RELEASE, 0, curproc);
    854 		return (error);
    855 	}
    856 	MALLOC(ip, struct inode *, sizeof(struct inode), M_EXT2FSNODE, M_WAITOK);
    857 	bzero((caddr_t)ip, sizeof(struct inode));
    858 	vp->v_data = ip;
    859 	ip->i_vnode = vp;
    860 	ip->i_e2fs = fs = ump->um_e2fs;
    861 	ip->i_dev = dev;
    862 	ip->i_number = ino;
    863 	ip->i_e2fs_last_lblk = 0;
    864 	ip->i_e2fs_last_blk = 0;
    865 
    866 	/*
    867 	 * Put it onto its hash chain and lock it so that other requests for
    868 	 * this inode will block if they arrive while we are sleeping waiting
    869 	 * for old data structures to be purged or for the contents of the
    870 	 * disk portion of this inode to be read.
    871 	 */
    872 	ufs_ihashins(ip);
    873 	lockmgr(&ufs_hashlock, LK_RELEASE, 0, curproc);
    874 
    875 	/* Read in the disk contents for the inode, copy into the inode. */
    876 	error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
    877 			  (int)fs->e2fs_bsize, NOCRED, &bp);
    878 	if (error) {
    879 		/*
    880 		 * The inode does not contain anything useful, so it would
    881 		 * be misleading to leave it on its hash chain. With mode
    882 		 * still zero, it will be unlinked and returned to the free
    883 		 * list by vput().
    884 		 */
    885 		vput(vp);
    886 		brelse(bp);
    887 		*vpp = NULL;
    888 		return (error);
    889 	}
    890 	e2fs_iload((struct ext2fs_dinode*)bp->b_data + ino_to_fsbo(fs, ino),
    891 				&ip->i_din.e2fs_din);
    892 	brelse(bp);
    893 
    894 	/* If the inode was deleted, reset all fields */
    895 	if (ip->i_e2fs_dtime != 0) {
    896 		ip->i_e2fs_mode = ip->i_e2fs_size = ip->i_e2fs_nblock = 0;
    897 		bzero(ip->i_e2fs_blocks, sizeof(ip->i_e2fs_blocks));
    898 	}
    899 
    900 	/*
    901 	 * Initialize the vnode from the inode, check for aliases.
    902 	 * Note that the underlying vnode may have changed.
    903 	 */
    904 	error = ext2fs_vinit(mp, ext2fs_specop_p, EXT2FS_FIFOOPS, &vp);
    905 	if (error) {
    906 		vput(vp);
    907 		*vpp = NULL;
    908 		return (error);
    909 	}
    910 	/*
    911 	 * Finish inode initialization now that aliasing has been resolved.
    912 	 */
    913 	ip->i_devvp = ump->um_devvp;
    914 	VREF(ip->i_devvp);
    915 	/*
    916 	 * Set up a generation number for this inode if it does not
    917 	 * already have one. This should only happen on old filesystems.
    918 	 */
    919 	if (ip->i_e2fs_gen == 0) {
    920 		if (++ext2gennumber < (u_long)time.tv_sec)
    921 			ext2gennumber = time.tv_sec;
    922 		ip->i_e2fs_gen = ext2gennumber;
    923 		if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
    924 			ip->i_flag |= IN_MODIFIED;
    925 	}
    926 
    927 	*vpp = vp;
    928 	return (0);
    929 }
    930 
    931 /*
    932  * File handle to vnode
    933  *
    934  * Have to be really careful about stale file handles:
    935  * - check that the inode number is valid
    936  * - call ext2fs_vget() to get the locked inode
    937  * - check for an unallocated inode (i_mode == 0)
    938  * - check that the given client host has export rights and return
    939  *   those rights via. exflagsp and credanonp
    940  */
    941 int
    942 ext2fs_fhtovp(mp, fhp, nam, vpp, exflagsp, credanonp)
    943 	register struct mount *mp;
    944 	struct fid *fhp;
    945 	struct mbuf *nam;
    946 	struct vnode **vpp;
    947 	int *exflagsp;
    948 	struct ucred **credanonp;
    949 {
    950 	register struct ufid *ufhp;
    951 	struct m_ext2fs *fs;
    952 
    953 	ufhp = (struct ufid *)fhp;
    954 	fs = VFSTOUFS(mp)->um_e2fs;
    955 	if ((ufhp->ufid_ino < EXT2_FIRSTINO && ufhp->ufid_ino != EXT2_ROOTINO) ||
    956 		ufhp->ufid_ino >= fs->e2fs_ncg * fs->e2fs.e2fs_ipg)
    957 		return (ESTALE);
    958 	return (ext2fs_check_export(mp, ufhp, nam, vpp, exflagsp, credanonp));
    959 }
    960 
    961 /*
    962  * Vnode pointer to File handle
    963  */
    964 /* ARGSUSED */
    965 int
    966 ext2fs_vptofh(vp, fhp)
    967 	struct vnode *vp;
    968 	struct fid *fhp;
    969 {
    970 	register struct inode *ip;
    971 	register struct ufid *ufhp;
    972 
    973 	ip = VTOI(vp);
    974 	ufhp = (struct ufid *)fhp;
    975 	ufhp->ufid_len = sizeof(struct ufid);
    976 	ufhp->ufid_ino = ip->i_number;
    977 	ufhp->ufid_gen = ip->i_e2fs_gen;
    978 	return (0);
    979 }
    980 
    981 /*
    982  * Write a superblock and associated information back to disk.
    983  */
    984 int
    985 ext2fs_sbupdate(mp, waitfor)
    986 	struct ufsmount *mp;
    987 	int waitfor;
    988 {
    989 	register struct m_ext2fs *fs = mp->um_e2fs;
    990 	register struct buf *bp;
    991 	int error = 0;
    992 
    993 	bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0);
    994 	e2fs_sbsave(&fs->e2fs, (struct ext2fs*)bp->b_data);
    995 	if (waitfor == MNT_WAIT)
    996 		error = bwrite(bp);
    997 	else
    998 		bawrite(bp);
    999 	return (error);
   1000 }
   1001 
   1002 int
   1003 ext2fs_cgupdate(mp, waitfor)
   1004 	struct ufsmount *mp;
   1005 	int waitfor;
   1006 {
   1007 	register struct m_ext2fs *fs = mp->um_e2fs;
   1008 	register struct buf *bp;
   1009 	int i, error = 0, allerror = 0;
   1010 
   1011 	allerror = ext2fs_sbupdate(mp, waitfor);
   1012 	for (i = 0; i < fs->e2fs_ngdb; i++) {
   1013 		bp = getblk(mp->um_devvp, fsbtodb(fs, ((fs->e2fs_bsize>1024)?0:1)+i+1),
   1014 			fs->e2fs_bsize, 0, 0);
   1015 		e2fs_cgsave(&fs->e2fs_gd[i* fs->e2fs_bsize / sizeof(struct ext2_gd)],
   1016 				(struct ext2_gd*)bp->b_data, fs->e2fs_bsize);
   1017 		if (waitfor == MNT_WAIT)
   1018 			error = bwrite(bp);
   1019 		else
   1020 			bawrite(bp);
   1021 	}
   1022 
   1023 	if (!allerror && error)
   1024 		allerror = error;
   1025 	return (allerror);
   1026 }
   1027