Home | History | Annotate | Line # | Download | only in ext2fs
ext2fs_vfsops.c revision 1.32
      1 /*	$NetBSD: ext2fs_vfsops.c,v 1.32 2000/01/28 16:00:23 bouyer 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 #if defined(_KERNEL) && !defined(_LKM)
     41 #include "opt_compat_netbsd.h"
     42 #endif
     43 
     44 #include <sys/param.h>
     45 #include <sys/systm.h>
     46 #include <sys/namei.h>
     47 #include <sys/proc.h>
     48 #include <sys/kernel.h>
     49 #include <sys/vnode.h>
     50 #include <sys/socket.h>
     51 #include <sys/mount.h>
     52 #include <sys/buf.h>
     53 #include <sys/device.h>
     54 #include <sys/mbuf.h>
     55 #include <sys/file.h>
     56 #include <sys/disklabel.h>
     57 #include <sys/ioctl.h>
     58 #include <sys/errno.h>
     59 #include <sys/malloc.h>
     60 #include <sys/pool.h>
     61 #include <sys/lock.h>
     62 
     63 #include <miscfs/specfs/specdev.h>
     64 
     65 #include <ufs/ufs/quota.h>
     66 #include <ufs/ufs/ufsmount.h>
     67 #include <ufs/ufs/inode.h>
     68 #include <ufs/ufs/dir.h>
     69 #include <ufs/ufs/ufs_extern.h>
     70 
     71 #include <ufs/ext2fs/ext2fs.h>
     72 #include <ufs/ext2fs/ext2fs_extern.h>
     73 
     74 extern struct lock ufs_hashlock;
     75 
     76 int ext2fs_sbupdate __P((struct ufsmount *, int));
     77 static int ext2fs_checksb __P((struct ext2fs *, int));
     78 
     79 extern struct vnodeopv_desc ext2fs_vnodeop_opv_desc;
     80 extern struct vnodeopv_desc ext2fs_specop_opv_desc;
     81 extern struct vnodeopv_desc ext2fs_fifoop_opv_desc;
     82 
     83 struct vnodeopv_desc *ext2fs_vnodeopv_descs[] = {
     84 	&ext2fs_vnodeop_opv_desc,
     85 	&ext2fs_specop_opv_desc,
     86 	&ext2fs_fifoop_opv_desc,
     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_sysctl,
    104 	ext2fs_mountroot,
    105 	ufs_check_export,
    106 	ext2fs_vnodeopv_descs,
    107 };
    108 
    109 struct pool ext2fs_inode_pool;
    110 
    111 extern u_long ext2gennumber;
    112 
    113 void
    114 ext2fs_init()
    115 {
    116 	ufs_init();
    117 
    118 	/*
    119 	 * XXX Same structure as FFS inodes?  Should we share a common pool?
    120 	 */
    121 	pool_init(&ext2fs_inode_pool, sizeof(struct inode), 0, 0, 0,
    122 	    "ext2fsinopl", 0, pool_page_alloc_nointr, pool_page_free_nointr,
    123 	    M_EXT2FSNODE);
    124 }
    125 
    126 
    127 /*
    128  * Called by main() when ext2fs is going to be mounted as root.
    129  *
    130  * Name is updated by mount(8) after booting.
    131  */
    132 #define ROOTNAME	"root_device"
    133 
    134 int
    135 ext2fs_mountroot()
    136 {
    137 	extern struct vnode *rootvp;
    138 	struct m_ext2fs *fs;
    139 	struct mount *mp;
    140 	struct proc *p = curproc;	/* XXX */
    141 	struct ufsmount *ump;
    142 	int error;
    143 
    144 	if (root_device->dv_class != DV_DISK)
    145 		return (ENODEV);
    146 
    147 	/*
    148 	 * Get vnodes for rootdev.
    149 	 */
    150 	if (bdevvp(rootdev, &rootvp))
    151 		panic("ext2fs_mountroot: can't setup bdevvp's");
    152 
    153 	if ((error = vfs_rootmountalloc(MOUNT_EXT2FS, "root_device", &mp))) {
    154 		vrele(rootvp);
    155 		return (error);
    156 	}
    157 
    158 	if ((error = ext2fs_mountfs(rootvp, mp, p)) != 0) {
    159 		mp->mnt_op->vfs_refcount--;
    160 		vfs_unbusy(mp);
    161 		free(mp, M_MOUNT);
    162 		vrele(rootvp);
    163 		return (error);
    164 	}
    165 	simple_lock(&mountlist_slock);
    166 	CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
    167 	simple_unlock(&mountlist_slock);
    168 	ump = VFSTOUFS(mp);
    169 	fs = ump->um_e2fs;
    170 	memset(fs->e2fs_fsmnt, 0, sizeof(fs->e2fs_fsmnt));
    171 	(void) copystr(mp->mnt_stat.f_mntonname, fs->e2fs_fsmnt,
    172 	    sizeof(fs->e2fs_fsmnt) - 1, 0);
    173 	if (fs->e2fs.e2fs_rev > E2FS_REV0) {
    174 		memset(fs->e2fs.e2fs_fsmnt, 0, sizeof(fs->e2fs.e2fs_fsmnt));
    175 		(void) copystr(mp->mnt_stat.f_mntonname, fs->e2fs.e2fs_fsmnt,
    176 		    sizeof(fs->e2fs.e2fs_fsmnt) - 1, 0);
    177 	}
    178 	(void)ext2fs_statfs(mp, &mp->mnt_stat, p);
    179 	vfs_unbusy(mp);
    180 	inittodr(fs->e2fs.e2fs_wtime);
    181 	return (0);
    182 }
    183 
    184 /*
    185  * VFS Operations.
    186  *
    187  * mount system call
    188  */
    189 int
    190 ext2fs_mount(mp, path, data, ndp, p)
    191 	register struct mount *mp;
    192 	const char *path;
    193 	void * data;
    194 	struct nameidata *ndp;
    195 	struct proc *p;
    196 {
    197 	struct vnode *devvp;
    198 	struct ufs_args args;
    199 	struct ufsmount *ump = NULL;
    200 	register struct m_ext2fs *fs;
    201 	size_t size;
    202 	int error, flags;
    203 	mode_t accessmode;
    204 
    205 	error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args));
    206 	if (error)
    207 		return (error);
    208 	/*
    209 	 * If updating, check whether changing from read-only to
    210 	 * read/write; if there is no device name, that's all we do.
    211 	 */
    212 	if (mp->mnt_flag & MNT_UPDATE) {
    213 		ump = VFSTOUFS(mp);
    214 		fs = ump->um_e2fs;
    215 		if (fs->e2fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
    216 			flags = WRITECLOSE;
    217 			if (mp->mnt_flag & MNT_FORCE)
    218 				flags |= FORCECLOSE;
    219 			error = ext2fs_flushfiles(mp, flags, p);
    220 			if (error == 0 &&
    221 				ext2fs_cgupdate(ump, MNT_WAIT) == 0 &&
    222 				(fs->e2fs.e2fs_state & E2FS_ERRORS) == 0) {
    223 				fs->e2fs.e2fs_state = E2FS_ISCLEAN;
    224 				(void) ext2fs_sbupdate(ump, MNT_WAIT);
    225 			}
    226 			if (error)
    227 				return (error);
    228 			fs->e2fs_ronly = 1;
    229 		}
    230 		if (mp->mnt_flag & MNT_RELOAD) {
    231 			error = ext2fs_reload(mp, ndp->ni_cnd.cn_cred, p);
    232 			if (error)
    233 				return (error);
    234 		}
    235 		if (fs->e2fs_ronly && (mp->mnt_flag & MNT_WANTRDWR)) {
    236 			/*
    237 			 * If upgrade to read-write by non-root, then verify
    238 			 * that user has necessary permissions on the device.
    239 			 */
    240 			if (p->p_ucred->cr_uid != 0) {
    241 				devvp = ump->um_devvp;
    242 				vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
    243 				error = VOP_ACCESS(devvp, VREAD | VWRITE,
    244 						   p->p_ucred, p);
    245 				VOP_UNLOCK(devvp, 0);
    246 				if (error)
    247 					return (error);
    248 			}
    249 			fs->e2fs_ronly = 0;
    250 			if (fs->e2fs.e2fs_state == E2FS_ISCLEAN)
    251 				fs->e2fs.e2fs_state = 0;
    252 			else
    253 				fs->e2fs.e2fs_state = E2FS_ERRORS;
    254 			fs->e2fs_fmod = 1;
    255 		}
    256 		if (args.fspec == 0) {
    257 			/*
    258 			 * Process export requests.
    259 			 */
    260 			return (vfs_export(mp, &ump->um_export, &args.export));
    261 		}
    262 	}
    263 	/*
    264 	 * Not an update, or updating the name: look up the name
    265 	 * and verify that it refers to a sensible block device.
    266 	 */
    267 	NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, p);
    268 	if ((error = namei(ndp)) != 0)
    269 		return (error);
    270 	devvp = ndp->ni_vp;
    271 
    272 	if (devvp->v_type != VBLK) {
    273 		vrele(devvp);
    274 		return (ENOTBLK);
    275 	}
    276 	if (major(devvp->v_rdev) >= nblkdev) {
    277 		vrele(devvp);
    278 		return (ENXIO);
    279 	}
    280 	/*
    281 	 * If mount by non-root, then verify that user has necessary
    282 	 * permissions on the device.
    283 	 */
    284 	if (p->p_ucred->cr_uid != 0) {
    285 		accessmode = VREAD;
    286 		if ((mp->mnt_flag & MNT_RDONLY) == 0)
    287 			accessmode |= VWRITE;
    288 		vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
    289 		error = VOP_ACCESS(devvp, accessmode, p->p_ucred, p);
    290 		VOP_UNLOCK(devvp, 0);
    291 		if (error) {
    292 			vrele(devvp);
    293 			return (error);
    294 		}
    295 	}
    296 	if ((mp->mnt_flag & MNT_UPDATE) == 0)
    297 		error = ext2fs_mountfs(devvp, mp, p);
    298 	else {
    299 		if (devvp != ump->um_devvp)
    300 			error = EINVAL;	/* needs translation */
    301 		else
    302 			vrele(devvp);
    303 	}
    304 	if (error) {
    305 		vrele(devvp);
    306 		return (error);
    307 	}
    308 	ump = VFSTOUFS(mp);
    309 	fs = ump->um_e2fs;
    310 	(void) copyinstr(path, fs->e2fs_fsmnt, sizeof(fs->e2fs_fsmnt) - 1,
    311 	    &size);
    312 	memset(fs->e2fs_fsmnt + size, 0, sizeof(fs->e2fs_fsmnt) - size);
    313 	if (fs->e2fs.e2fs_rev > E2FS_REV0) {
    314 		(void) copystr(mp->mnt_stat.f_mntonname, fs->e2fs.e2fs_fsmnt,
    315 		    sizeof(fs->e2fs.e2fs_fsmnt) - 1, &size);
    316 		memset(fs->e2fs.e2fs_fsmnt, 0,
    317 		    sizeof(fs->e2fs.e2fs_fsmnt) - size);
    318 	}
    319 	memcpy(mp->mnt_stat.f_mntonname, fs->e2fs_fsmnt, MNAMELEN);
    320 	(void) copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
    321 		&size);
    322 	memset(mp->mnt_stat.f_mntfromname + size, 0, MNAMELEN - size);
    323 	if (fs->e2fs_fmod != 0) {	/* XXX */
    324 		fs->e2fs_fmod = 0;
    325 		if (fs->e2fs.e2fs_state == 0)
    326 			fs->e2fs.e2fs_wtime = time.tv_sec;
    327 		else
    328 			printf("%s: file system not clean; please fsck(8)\n",
    329 				mp->mnt_stat.f_mntfromname);
    330 		(void) ext2fs_cgupdate(ump, MNT_WAIT);
    331 	}
    332 	return (0);
    333 }
    334 
    335 /*
    336  * Reload all incore data for a filesystem (used after running fsck on
    337  * the root filesystem and finding things to fix). The filesystem must
    338  * be mounted read-only.
    339  *
    340  * Things to do to update the mount:
    341  *	1) invalidate all cached meta-data.
    342  *	2) re-read superblock from disk.
    343  *	3) re-read summary information from disk.
    344  *	4) invalidate all inactive vnodes.
    345  *	5) invalidate all cached file data.
    346  *	6) re-read inode data for all active vnodes.
    347  */
    348 int
    349 ext2fs_reload(mountp, cred, p)
    350 	register struct mount *mountp;
    351 	struct ucred *cred;
    352 	struct proc *p;
    353 {
    354 	register struct vnode *vp, *nvp, *devvp;
    355 	struct inode *ip;
    356 	struct buf *bp;
    357 	struct m_ext2fs *fs;
    358 	struct ext2fs *newfs;
    359 	struct partinfo dpart;
    360 	int i, size, error;
    361 	caddr_t cp;
    362 
    363 	if ((mountp->mnt_flag & MNT_RDONLY) == 0)
    364 		return (EINVAL);
    365 	/*
    366 	 * Step 1: invalidate all cached meta-data.
    367 	 */
    368 	devvp = VFSTOUFS(mountp)->um_devvp;
    369 	if (vinvalbuf(devvp, 0, cred, p, 0, 0))
    370 		panic("ext2fs_reload: dirty1");
    371 	/*
    372 	 * Step 2: re-read superblock from disk.
    373 	 */
    374 	if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, p) != 0)
    375 		size = DEV_BSIZE;
    376 	else
    377 		size = dpart.disklab->d_secsize;
    378 	error = bread(devvp, (ufs_daddr_t)(SBOFF / size), SBSIZE, NOCRED, &bp);
    379 	if (error) {
    380 		brelse(bp);
    381 		return (error);
    382 	}
    383 	newfs = (struct ext2fs *)bp->b_data;
    384 	error = ext2fs_checksb(newfs, (mountp->mnt_flag & MNT_RDONLY) != 0);
    385 	if (error) {
    386 		brelse(bp);
    387 		return (error);
    388 	}
    389 
    390 	fs = VFSTOUFS(mountp)->um_e2fs;
    391 	/*
    392 	 * copy in new superblock, and compute in-memory values
    393 	 */
    394 	e2fs_sbload(newfs, &fs->e2fs);
    395 	fs->e2fs_ncg =
    396 	    howmany(fs->e2fs.e2fs_bcount - fs->e2fs.e2fs_first_dblock,
    397 	    fs->e2fs.e2fs_bpg);
    398 	/* XXX assume hw bsize = 512 */
    399 	fs->e2fs_fsbtodb = fs->e2fs.e2fs_log_bsize + 1;
    400 	fs->e2fs_bsize = 1024 << fs->e2fs.e2fs_log_bsize;
    401 	fs->e2fs_bshift = LOG_MINBSIZE + fs->e2fs.e2fs_log_bsize;
    402 	fs->e2fs_qbmask = fs->e2fs_bsize - 1;
    403 	fs->e2fs_bmask = ~fs->e2fs_qbmask;
    404 	fs->e2fs_ngdb = howmany(fs->e2fs_ncg,
    405 			fs->e2fs_bsize / sizeof(struct ext2_gd));
    406 	fs->e2fs_ipb = fs->e2fs_bsize / EXT2_DINODE_SIZE;
    407 	fs->e2fs_itpg = fs->e2fs.e2fs_ipg/fs->e2fs_ipb;
    408 
    409 	/*
    410 	 * Step 3: re-read summary information from disk.
    411 	 */
    412 
    413 	for (i=0; i < fs->e2fs_ngdb; i++) {
    414 		error = bread(devvp ,
    415 		    fsbtodb(fs, ((fs->e2fs_bsize>1024)? 0 : 1) + i + 1),
    416 		    fs->e2fs_bsize, NOCRED, &bp);
    417 		if (error) {
    418 			brelse(bp);
    419 			return (error);
    420 		}
    421 		e2fs_cgload((struct ext2_gd*)bp->b_data,
    422 		    &fs->e2fs_gd[i* fs->e2fs_bsize / sizeof(struct ext2_gd)],
    423 		    fs->e2fs_bsize);
    424 		brelse(bp);
    425 	}
    426 
    427 loop:
    428 	simple_lock(&mntvnode_slock);
    429 	for (vp = mountp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) {
    430 		if (vp->v_mount != mountp) {
    431 			simple_unlock(&mntvnode_slock);
    432 			goto loop;
    433 		}
    434 		nvp = vp->v_mntvnodes.le_next;
    435 		/*
    436 		 * Step 4: invalidate all inactive vnodes.
    437 		 */
    438 		if (vrecycle(vp, &mntvnode_slock, p))
    439 			goto loop;
    440 		/*
    441 		 * Step 5: invalidate all cached file data.
    442 		 */
    443 		simple_lock(&vp->v_interlock);
    444 		simple_unlock(&mntvnode_slock);
    445 		if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK))
    446 			goto loop;
    447 		if (vinvalbuf(vp, 0, cred, p, 0, 0))
    448 			panic("ext2fs_reload: dirty2");
    449 		/*
    450 		 * Step 6: re-read inode data for all active vnodes.
    451 		 */
    452 		ip = VTOI(vp);
    453 		error = bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
    454 				  (int)fs->e2fs_bsize, NOCRED, &bp);
    455 		if (error) {
    456 			vput(vp);
    457 			return (error);
    458 		}
    459 		cp = (caddr_t)bp->b_data +
    460 		    (ino_to_fsbo(fs, ip->i_number) * EXT2_DINODE_SIZE);
    461 		e2fs_iload((struct ext2fs_dinode *)cp, &ip->i_din.e2fs_din);
    462 		brelse(bp);
    463 		vput(vp);
    464 		simple_lock(&mntvnode_slock);
    465 	}
    466 	simple_unlock(&mntvnode_slock);
    467 	return (0);
    468 }
    469 
    470 /*
    471  * Common code for mount and mountroot
    472  */
    473 int
    474 ext2fs_mountfs(devvp, mp, p)
    475 	register struct vnode *devvp;
    476 	struct mount *mp;
    477 	struct proc *p;
    478 {
    479 	register struct ufsmount *ump;
    480 	struct buf *bp;
    481 	register struct ext2fs *fs;
    482 	register struct m_ext2fs *m_fs;
    483 	dev_t dev;
    484 	struct partinfo dpart;
    485 	int error, i, size, ronly;
    486 	struct ucred *cred;
    487 	extern struct vnode *rootvp;
    488 
    489 	dev = devvp->v_rdev;
    490 	cred = p ? p->p_ucred : NOCRED;
    491 	/*
    492 	 * Disallow multiple mounts of the same device.
    493 	 * Disallow mounting of a device that is currently in use
    494 	 * (except for root, which might share swap device for miniroot).
    495 	 * Flush out any old buffers remaining from a previous use.
    496 	 */
    497 	if ((error = vfs_mountedon(devvp)) != 0)
    498 		return (error);
    499 	if (vcount(devvp) > 1 && devvp != rootvp)
    500 		return (EBUSY);
    501 	if ((error = vinvalbuf(devvp, V_SAVE, cred, p, 0, 0)) != 0)
    502 		return (error);
    503 
    504 	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
    505 	error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p);
    506 	if (error)
    507 		return (error);
    508 	if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, cred, p) != 0)
    509 		size = DEV_BSIZE;
    510 	else
    511 		size = dpart.disklab->d_secsize;
    512 
    513 	bp = NULL;
    514 	ump = NULL;
    515 
    516 #ifdef DEBUG_EXT2
    517 	printf("sb size: %d ino size %d\n", sizeof(struct ext2fs),
    518 	    EXT2_DINODE_SIZE);
    519 #endif
    520 	error = bread(devvp, (SBOFF / DEV_BSIZE), SBSIZE, cred, &bp);
    521 	if (error)
    522 		goto out;
    523 	fs = (struct ext2fs *)bp->b_data;
    524 	if (fs2h16(fs->e2fs_magic) != E2FS_MAGIC) {
    525 		error = EINVAL;		/* XXX needs translation */
    526 		goto out;
    527 	}
    528 	error = ext2fs_checksb(fs, ronly);
    529 	if (error)
    530 		goto out;
    531 	ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
    532 	memset((caddr_t)ump, 0, sizeof *ump);
    533 	ump->um_e2fs = malloc(sizeof(struct m_ext2fs), M_UFSMNT, M_WAITOK);
    534 	memset((caddr_t)ump->um_e2fs, 0, sizeof(struct m_ext2fs));
    535 	e2fs_sbload((struct ext2fs*)bp->b_data, &ump->um_e2fs->e2fs);
    536 	brelse(bp);
    537 	bp = NULL;
    538 	m_fs = ump->um_e2fs;
    539 	m_fs->e2fs_ronly = ronly;
    540 	if (ronly == 0) {
    541 		if (m_fs->e2fs.e2fs_state == E2FS_ISCLEAN)
    542 			m_fs->e2fs.e2fs_state = 0;
    543 		else
    544 			m_fs->e2fs.e2fs_state = E2FS_ERRORS;
    545 		m_fs->e2fs_fmod = 1;
    546 	}
    547 
    548 	/* compute dynamic sb infos */
    549 	m_fs->e2fs_ncg =
    550 		howmany(m_fs->e2fs.e2fs_bcount - m_fs->e2fs.e2fs_first_dblock,
    551 		m_fs->e2fs.e2fs_bpg);
    552 	/* XXX assume hw bsize = 512 */
    553 	m_fs->e2fs_fsbtodb = m_fs->e2fs.e2fs_log_bsize + 1;
    554 	m_fs->e2fs_bsize = 1024 << m_fs->e2fs.e2fs_log_bsize;
    555 	m_fs->e2fs_bshift = LOG_MINBSIZE + m_fs->e2fs.e2fs_log_bsize;
    556 	m_fs->e2fs_qbmask = m_fs->e2fs_bsize - 1;
    557 	m_fs->e2fs_bmask = ~m_fs->e2fs_qbmask;
    558 	m_fs->e2fs_ngdb = howmany(m_fs->e2fs_ncg,
    559 		m_fs->e2fs_bsize / sizeof(struct ext2_gd));
    560 	m_fs->e2fs_ipb = m_fs->e2fs_bsize / EXT2_DINODE_SIZE;
    561 	m_fs->e2fs_itpg = m_fs->e2fs.e2fs_ipg/m_fs->e2fs_ipb;
    562 
    563 	m_fs->e2fs_gd = malloc(m_fs->e2fs_ngdb * m_fs->e2fs_bsize,
    564 		M_UFSMNT, M_WAITOK);
    565 	for (i=0; i < m_fs->e2fs_ngdb; i++) {
    566 		error = bread(devvp ,
    567 		    fsbtodb(m_fs, ((m_fs->e2fs_bsize>1024)? 0 : 1) + i + 1),
    568 		    m_fs->e2fs_bsize, NOCRED, &bp);
    569 		if (error) {
    570 			free(m_fs->e2fs_gd, M_UFSMNT);
    571 			goto out;
    572 		}
    573 		e2fs_cgload((struct ext2_gd*)bp->b_data,
    574 		    &m_fs->e2fs_gd[
    575 			i * m_fs->e2fs_bsize / sizeof(struct ext2_gd)],
    576 		    m_fs->e2fs_bsize);
    577 		brelse(bp);
    578 		bp = NULL;
    579 	}
    580 
    581 	mp->mnt_data = (qaddr_t)ump;
    582 	mp->mnt_stat.f_fsid.val[0] = (long)dev;
    583 	mp->mnt_stat.f_fsid.val[1] = makefstype(MOUNT_EXT2FS);
    584 	mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN;
    585 	mp->mnt_flag |= MNT_LOCAL;
    586 	ump->um_flags = 0;
    587 	ump->um_mountp = mp;
    588 	ump->um_dev = dev;
    589 	ump->um_devvp = devvp;
    590 	ump->um_nindir = NINDIR(m_fs);
    591 	ump->um_bptrtodb = m_fs->e2fs_fsbtodb;
    592 	ump->um_seqinc = 1; /* no frags */
    593 	devvp->v_specmountpoint = mp;
    594 	return (0);
    595 out:
    596 	if (bp)
    597 		brelse(bp);
    598 	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
    599 	(void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, cred, p);
    600 	VOP_UNLOCK(devvp, 0);
    601 	if (ump) {
    602 		free(ump->um_e2fs, M_UFSMNT);
    603 		free(ump, M_UFSMNT);
    604 		mp->mnt_data = (qaddr_t)0;
    605 	}
    606 	return (error);
    607 }
    608 
    609 /*
    610  * unmount system call
    611  */
    612 int
    613 ext2fs_unmount(mp, mntflags, p)
    614 	struct mount *mp;
    615 	int mntflags;
    616 	struct proc *p;
    617 {
    618 	register struct ufsmount *ump;
    619 	register struct m_ext2fs *fs;
    620 	int error, flags;
    621 
    622 	flags = 0;
    623 	if (mntflags & MNT_FORCE)
    624 		flags |= FORCECLOSE;
    625 	if ((error = ext2fs_flushfiles(mp, flags, p)) != 0)
    626 		return (error);
    627 	ump = VFSTOUFS(mp);
    628 	fs = ump->um_e2fs;
    629 	if (fs->e2fs_ronly == 0 &&
    630 		ext2fs_cgupdate(ump, MNT_WAIT) == 0 &&
    631 		(fs->e2fs.e2fs_state & E2FS_ERRORS) == 0) {
    632 		fs->e2fs.e2fs_state = E2FS_ISCLEAN;
    633 		(void) ext2fs_sbupdate(ump, MNT_WAIT);
    634 	}
    635 	if (ump->um_devvp->v_type != VBAD)
    636 		ump->um_devvp->v_specmountpoint = NULL;
    637 	vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
    638 	error = VOP_CLOSE(ump->um_devvp, fs->e2fs_ronly ? FREAD : FREAD|FWRITE,
    639 		NOCRED, p);
    640 	vput(ump->um_devvp);
    641 	free(fs->e2fs_gd, M_UFSMNT);
    642 	free(fs, M_UFSMNT);
    643 	free(ump, M_UFSMNT);
    644 	mp->mnt_data = (qaddr_t)0;
    645 	mp->mnt_flag &= ~MNT_LOCAL;
    646 	return (error);
    647 }
    648 
    649 /*
    650  * Flush out all the files in a filesystem.
    651  */
    652 int
    653 ext2fs_flushfiles(mp, flags, p)
    654 	register struct mount *mp;
    655 	int flags;
    656 	struct proc *p;
    657 {
    658 	extern int doforce;
    659 	register struct ufsmount *ump;
    660 	int error;
    661 
    662 	if (!doforce)
    663 		flags &= ~FORCECLOSE;
    664 	ump = VFSTOUFS(mp);
    665 	error = vflush(mp, NULLVP, flags);
    666 	return (error);
    667 }
    668 
    669 /*
    670  * Get file system statistics.
    671  */
    672 int
    673 ext2fs_statfs(mp, sbp, p)
    674 	struct mount *mp;
    675 	register struct statfs *sbp;
    676 	struct proc *p;
    677 {
    678 	register struct ufsmount *ump;
    679 	register struct m_ext2fs *fs;
    680 	u_int32_t overhead, overhead_per_group;
    681 	int i, ngroups;
    682 
    683 	ump = VFSTOUFS(mp);
    684 	fs = ump->um_e2fs;
    685 	if (fs->e2fs.e2fs_magic != E2FS_MAGIC)
    686 		panic("ext2fs_statfs");
    687 
    688 #ifdef COMPAT_09
    689 	sbp->f_type = 1;
    690 #else
    691 	sbp->f_type = 0;
    692 #endif
    693 
    694 	/*
    695 	 * Compute the overhead (FS structures)
    696 	 */
    697 	overhead_per_group = 1 /* block bitmap */ +
    698 				 1 /* inode bitmap */ +
    699 				 fs->e2fs_itpg;
    700 	overhead = fs->e2fs.e2fs_first_dblock +
    701 		   fs->e2fs_ncg * overhead_per_group;
    702 	if (fs->e2fs.e2fs_rev > E2FS_REV0 &&
    703 	    fs->e2fs.e2fs_features_rocompat & EXT2F_ROCOMPAT_SPARSESUPER) {
    704 		for (i = 0, ngroups = 0; i < fs->e2fs_ncg; i++) {
    705 			if (cg_has_sb(i))
    706 				ngroups++;
    707 		}
    708 	} else {
    709 		ngroups = fs->e2fs_ncg;
    710 	}
    711 	overhead += ngroups * (1 + fs->e2fs_ngdb);
    712 
    713 	sbp->f_bsize = fs->e2fs_bsize;
    714 	sbp->f_iosize = fs->e2fs_bsize;
    715 	sbp->f_blocks = fs->e2fs.e2fs_bcount - overhead;
    716 	sbp->f_bfree = fs->e2fs.e2fs_fbcount;
    717 	sbp->f_bavail = sbp->f_bfree - fs->e2fs.e2fs_rbcount;
    718 	sbp->f_files =  fs->e2fs.e2fs_icount;
    719 	sbp->f_ffree = fs->e2fs.e2fs_ficount;
    720 	if (sbp != &mp->mnt_stat) {
    721 		memcpy(sbp->f_mntonname, mp->mnt_stat.f_mntonname, MNAMELEN);
    722 		memcpy(sbp->f_mntfromname, mp->mnt_stat.f_mntfromname, MNAMELEN);
    723 	}
    724 	strncpy(sbp->f_fstypename, mp->mnt_op->vfs_name, MFSNAMELEN);
    725 	return (0);
    726 }
    727 
    728 /*
    729  * Go through the disk queues to initiate sandbagged IO;
    730  * go through the inodes to write those that have been modified;
    731  * initiate the writing of the super block if it has been modified.
    732  *
    733  * Note: we are always called with the filesystem marked `MPBUSY'.
    734  */
    735 int
    736 ext2fs_sync(mp, waitfor, cred, p)
    737 	struct mount *mp;
    738 	int waitfor;
    739 	struct ucred *cred;
    740 	struct proc *p;
    741 {
    742 	struct vnode *vp, *nvp;
    743 	struct inode *ip;
    744 	struct ufsmount *ump = VFSTOUFS(mp);
    745 	struct m_ext2fs *fs;
    746 	int error, allerror = 0;
    747 
    748 	fs = ump->um_e2fs;
    749 	if (fs->e2fs_ronly != 0) {		/* XXX */
    750 		printf("fs = %s\n", fs->e2fs_fsmnt);
    751 		panic("update: rofs mod");
    752 	}
    753 
    754 	/*
    755 	 * Write back each (modified) inode.
    756 	 */
    757 	simple_lock(&mntvnode_slock);
    758 loop:
    759 	for (vp = mp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) {
    760 		/*
    761 		 * If the vnode that we are about to sync is no longer
    762 		 * associated with this mount point, start over.
    763 		 */
    764 		if (vp->v_mount != mp)
    765 			goto loop;
    766 		simple_lock(&vp->v_interlock);
    767 		nvp = vp->v_mntvnodes.le_next;
    768 		ip = VTOI(vp);
    769 		if (vp->v_type == VNON || ((ip->i_flag &
    770 		    (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
    771 		    (vp->v_dirtyblkhd.lh_first == NULL || waitfor == MNT_LAZY)))
    772 		{
    773 			simple_unlock(&vp->v_interlock);
    774 			continue;
    775 		}
    776 		simple_unlock(&mntvnode_slock);
    777 		error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK);
    778 		if (error) {
    779 			simple_lock(&mntvnode_slock);
    780 			if (error == ENOENT)
    781 				goto loop;
    782 			continue;
    783 		}
    784 		if ((error = VOP_FSYNC(vp, cred,
    785 		    waitfor == MNT_WAIT ? FSYNC_WAIT : 0, p)) != 0)
    786 			allerror = error;
    787 		vput(vp);
    788 		simple_lock(&mntvnode_slock);
    789 	}
    790 	simple_unlock(&mntvnode_slock);
    791 	/*
    792 	 * Force stale file system control information to be flushed.
    793 	 */
    794 	if ((error = VOP_FSYNC(ump->um_devvp, cred,
    795 	    waitfor == MNT_WAIT ? FSYNC_WAIT : 0, p)) != 0)
    796 		allerror = error;
    797 	/*
    798 	 * Write back modified superblock.
    799 	 */
    800 	if (fs->e2fs_fmod != 0) {
    801 		fs->e2fs_fmod = 0;
    802 		fs->e2fs.e2fs_wtime = time.tv_sec;
    803 		if ((error = ext2fs_cgupdate(ump, waitfor)))
    804 			allerror = error;
    805 	}
    806 	return (allerror);
    807 }
    808 
    809 /*
    810  * Look up a EXT2FS dinode number to find its incore vnode, otherwise read it
    811  * in from disk.  If it is in core, wait for the lock bit to clear, then
    812  * return the inode locked.  Detection and handling of mount points must be
    813  * done by the calling routine.
    814  */
    815 int
    816 ext2fs_vget(mp, ino, vpp)
    817 	struct mount *mp;
    818 	ino_t ino;
    819 	struct vnode **vpp;
    820 {
    821 	struct m_ext2fs *fs;
    822 	struct inode *ip;
    823 	struct ufsmount *ump;
    824 	struct buf *bp;
    825 	struct vnode *vp;
    826 	dev_t dev;
    827 	int error;
    828 	caddr_t cp;
    829 
    830 	ump = VFSTOUFS(mp);
    831 	dev = ump->um_dev;
    832 	do {
    833 		if ((*vpp = ufs_ihashget(dev, ino, LK_EXCLUSIVE)) != NULL)
    834 			return (0);
    835 	} while (lockmgr(&ufs_hashlock, LK_EXCLUSIVE|LK_SLEEPFAIL, 0));
    836 
    837 	/* Allocate a new vnode/inode. */
    838 	if ((error = getnewvnode(VT_EXT2FS, mp, ext2fs_vnodeop_p, &vp)) != 0) {
    839 		*vpp = NULL;
    840 		lockmgr(&ufs_hashlock, LK_RELEASE, 0);
    841 		return (error);
    842 	}
    843 	ip = pool_get(&ext2fs_inode_pool, PR_WAITOK);
    844 	memset((caddr_t)ip, 0, sizeof(struct inode));
    845 	vp->v_data = ip;
    846 	ip->i_vnode = vp;
    847 	ip->i_e2fs = fs = ump->um_e2fs;
    848 	ip->i_dev = dev;
    849 	ip->i_number = ino;
    850 	ip->i_e2fs_last_lblk = 0;
    851 	ip->i_e2fs_last_blk = 0;
    852 
    853 	/*
    854 	 * Put it onto its hash chain and lock it so that other requests for
    855 	 * this inode will block if they arrive while we are sleeping waiting
    856 	 * for old data structures to be purged or for the contents of the
    857 	 * disk portion of this inode to be read.
    858 	 */
    859 	ufs_ihashins(ip);
    860 	lockmgr(&ufs_hashlock, LK_RELEASE, 0);
    861 
    862 	/* Read in the disk contents for the inode, copy into the inode. */
    863 	error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
    864 			  (int)fs->e2fs_bsize, NOCRED, &bp);
    865 	if (error) {
    866 		/*
    867 		 * The inode does not contain anything useful, so it would
    868 		 * be misleading to leave it on its hash chain. With mode
    869 		 * still zero, it will be unlinked and returned to the free
    870 		 * list by vput().
    871 		 */
    872 		vput(vp);
    873 		brelse(bp);
    874 		*vpp = NULL;
    875 		return (error);
    876 	}
    877 	cp = (caddr_t)bp->b_data +
    878 	    (ino_to_fsbo(fs, ino) * EXT2_DINODE_SIZE);
    879 	e2fs_iload((struct ext2fs_dinode *)cp, &ip->i_din.e2fs_din);
    880 	brelse(bp);
    881 
    882 	/* If the inode was deleted, reset all fields */
    883 	if (ip->i_e2fs_dtime != 0) {
    884 		ip->i_e2fs_mode = ip->i_e2fs_size = ip->i_e2fs_nblock = 0;
    885 		memset(ip->i_e2fs_blocks, 0, sizeof(ip->i_e2fs_blocks));
    886 	}
    887 
    888 	/*
    889 	 * Initialize the vnode from the inode, check for aliases.
    890 	 * Note that the underlying vnode may have changed.
    891 	 */
    892 	error = ext2fs_vinit(mp, ext2fs_specop_p, ext2fs_fifoop_p, &vp);
    893 	if (error) {
    894 		vput(vp);
    895 		*vpp = NULL;
    896 		return (error);
    897 	}
    898 	/*
    899 	 * Finish inode initialization now that aliasing has been resolved.
    900 	 */
    901 	ip->i_devvp = ump->um_devvp;
    902 	VREF(ip->i_devvp);
    903 	/*
    904 	 * Set up a generation number for this inode if it does not
    905 	 * already have one. This should only happen on old filesystems.
    906 	 */
    907 	if (ip->i_e2fs_gen == 0) {
    908 		if (++ext2gennumber < (u_long)time.tv_sec)
    909 			ext2gennumber = time.tv_sec;
    910 		ip->i_e2fs_gen = ext2gennumber;
    911 		if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
    912 			ip->i_flag |= IN_MODIFIED;
    913 	}
    914 
    915 	*vpp = vp;
    916 	return (0);
    917 }
    918 
    919 /*
    920  * File handle to vnode
    921  *
    922  * Have to be really careful about stale file handles:
    923  * - check that the inode number is valid
    924  * - call ext2fs_vget() to get the locked inode
    925  * - check for an unallocated inode (i_mode == 0)
    926  */
    927 int
    928 ext2fs_fhtovp(mp, fhp, vpp)
    929 	register struct mount *mp;
    930 	struct fid *fhp;
    931 	struct vnode **vpp;
    932 {
    933 	register struct inode *ip;
    934 	struct vnode *nvp;
    935 	int error;
    936 	register struct ufid *ufhp;
    937 	struct m_ext2fs *fs;
    938 
    939 	ufhp = (struct ufid *)fhp;
    940 	fs = VFSTOUFS(mp)->um_e2fs;
    941 	if ((ufhp->ufid_ino < EXT2_FIRSTINO && ufhp->ufid_ino != EXT2_ROOTINO) ||
    942 		ufhp->ufid_ino >= fs->e2fs_ncg * fs->e2fs.e2fs_ipg)
    943 		return (ESTALE);
    944 
    945 	if ((error = VFS_VGET(mp, ufhp->ufid_ino, &nvp)) != 0) {
    946 		*vpp = NULLVP;
    947 		return (error);
    948 	}
    949 	ip = VTOI(nvp);
    950 	if (ip->i_e2fs_mode == 0 || ip->i_e2fs_dtime != 0 ||
    951 		ip->i_e2fs_gen != ufhp->ufid_gen) {
    952 		vput(nvp);
    953 		*vpp = NULLVP;
    954 		return (ESTALE);
    955 	}
    956 	*vpp = nvp;
    957 	return (0);
    958 }
    959 
    960 /*
    961  * Vnode pointer to File handle
    962  */
    963 /* ARGSUSED */
    964 int
    965 ext2fs_vptofh(vp, fhp)
    966 	struct vnode *vp;
    967 	struct fid *fhp;
    968 {
    969 	register struct inode *ip;
    970 	register struct ufid *ufhp;
    971 
    972 	ip = VTOI(vp);
    973 	ufhp = (struct ufid *)fhp;
    974 	ufhp->ufid_len = sizeof(struct ufid);
    975 	ufhp->ufid_ino = ip->i_number;
    976 	ufhp->ufid_gen = ip->i_e2fs_gen;
    977 	return (0);
    978 }
    979 
    980 int
    981 ext2fs_sysctl(name, namelen, oldp, oldlenp, newp, newlen, p)
    982 	int *name;
    983 	u_int namelen;
    984 	void *oldp;
    985 	size_t *oldlenp;
    986 	void *newp;
    987 	size_t newlen;
    988 	struct proc *p;
    989 {
    990 	return (EOPNOTSUPP);
    991 }
    992 
    993 /*
    994  * Write a superblock and associated information back to disk.
    995  */
    996 int
    997 ext2fs_sbupdate(mp, waitfor)
    998 	struct ufsmount *mp;
    999 	int waitfor;
   1000 {
   1001 	register struct m_ext2fs *fs = mp->um_e2fs;
   1002 	register struct buf *bp;
   1003 	int error = 0;
   1004 
   1005 	bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0);
   1006 	e2fs_sbsave(&fs->e2fs, (struct ext2fs*)bp->b_data);
   1007 	if (waitfor == MNT_WAIT)
   1008 		error = bwrite(bp);
   1009 	else
   1010 		bawrite(bp);
   1011 	return (error);
   1012 }
   1013 
   1014 int
   1015 ext2fs_cgupdate(mp, waitfor)
   1016 	struct ufsmount *mp;
   1017 	int waitfor;
   1018 {
   1019 	register struct m_ext2fs *fs = mp->um_e2fs;
   1020 	register struct buf *bp;
   1021 	int i, error = 0, allerror = 0;
   1022 
   1023 	allerror = ext2fs_sbupdate(mp, waitfor);
   1024 	for (i = 0; i < fs->e2fs_ngdb; i++) {
   1025 		bp = getblk(mp->um_devvp, fsbtodb(fs, ((fs->e2fs_bsize>1024)?0:1)+i+1),
   1026 			fs->e2fs_bsize, 0, 0);
   1027 		e2fs_cgsave(&fs->e2fs_gd[i* fs->e2fs_bsize / sizeof(struct ext2_gd)],
   1028 				(struct ext2_gd*)bp->b_data, fs->e2fs_bsize);
   1029 		if (waitfor == MNT_WAIT)
   1030 			error = bwrite(bp);
   1031 		else
   1032 			bawrite(bp);
   1033 	}
   1034 
   1035 	if (!allerror && error)
   1036 		allerror = error;
   1037 	return (allerror);
   1038 }
   1039 
   1040 static int
   1041 ext2fs_checksb(fs, ronly)
   1042 	struct ext2fs *fs;
   1043 	int ronly;
   1044 {
   1045 	if (fs2h16(fs->e2fs_magic) != E2FS_MAGIC) {
   1046 		return (EIO);		/* XXX needs translation */
   1047 	}
   1048 	if (fs2h32(fs->e2fs_rev) > E2FS_REV1) {
   1049 #ifdef DIAGNOSTIC
   1050 		printf("Ext2 fs: unsupported revision number: %x\n",
   1051 					fs2h32(fs->e2fs_rev));
   1052 #endif
   1053 		return (EIO);		/* XXX needs translation */
   1054 	}
   1055 	if (fs2h32(fs->e2fs_log_bsize) > 2) { /* block size = 1024|2048|4096 */
   1056 #ifdef DIAGNOSTIC
   1057 		printf("Ext2 fs: bad block size: %d (expected <=2 for ext2 fs)\n",
   1058 			fs2h32(fs->e2fs_log_bsize));
   1059 #endif
   1060 		return (EIO);	   /* XXX needs translation */
   1061 	}
   1062 	if (fs2h32(fs->e2fs_rev) > E2FS_REV0) {
   1063 		if (fs2h32(fs->e2fs_features_incompat) &
   1064 		    ~EXT2F_INCOMPAT_SUPP) {
   1065 			printf("Ext2 fs: unsupported optionnal feature\n");
   1066 			return (EIO);      /* XXX needs translation */
   1067 		}
   1068 		if (!ronly && fs2h32(fs->e2fs_features_rocompat) &
   1069 		    ~EXT2F_ROCOMPAT_SUPP) {
   1070 			return (EROFS);      /* XXX needs translation */
   1071 		}
   1072 	}
   1073 	return (0);
   1074 }
   1075