Home | History | Annotate | Line # | Download | only in ext2fs
ext2fs_vfsops.c revision 1.29
      1 /*	$NetBSD: ext2fs_vfsops.c,v 1.29 1999/10/20 14:32:11 enami 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 
     78 extern struct vnodeopv_desc ext2fs_vnodeop_opv_desc;
     79 extern struct vnodeopv_desc ext2fs_specop_opv_desc;
     80 extern struct vnodeopv_desc ext2fs_fifoop_opv_desc;
     81 
     82 struct vnodeopv_desc *ext2fs_vnodeopv_descs[] = {
     83 	&ext2fs_vnodeop_opv_desc,
     84 	&ext2fs_specop_opv_desc,
     85 	&ext2fs_fifoop_opv_desc,
     86 	NULL,
     87 };
     88 
     89 struct vfsops ext2fs_vfsops = {
     90 	MOUNT_EXT2FS,
     91 	ext2fs_mount,
     92 	ufs_start,
     93 	ext2fs_unmount,
     94 	ufs_root,
     95 	ufs_quotactl,
     96 	ext2fs_statfs,
     97 	ext2fs_sync,
     98 	ext2fs_vget,
     99 	ext2fs_fhtovp,
    100 	ext2fs_vptofh,
    101 	ext2fs_init,
    102 	ext2fs_sysctl,
    103 	ext2fs_mountroot,
    104 	ufs_check_export,
    105 	ext2fs_vnodeopv_descs,
    106 };
    107 
    108 struct pool ext2fs_inode_pool;
    109 
    110 extern u_long ext2gennumber;
    111 
    112 void
    113 ext2fs_init()
    114 {
    115 	ufs_init();
    116 
    117 	/*
    118 	 * XXX Same structure as FFS inodes?  Should we share a common pool?
    119 	 */
    120 	pool_init(&ext2fs_inode_pool, sizeof(struct inode), 0, 0, 0,
    121 	    "ext2fsinopl", 0, pool_page_alloc_nointr, pool_page_free_nointr,
    122 	    M_EXT2FSNODE);
    123 }
    124 
    125 
    126 /*
    127  * Called by main() when ext2fs is going to be mounted as root.
    128  *
    129  * Name is updated by mount(8) after booting.
    130  */
    131 #define ROOTNAME	"root_device"
    132 
    133 int
    134 ext2fs_mountroot()
    135 {
    136 	extern struct vnode *rootvp;
    137 	struct m_ext2fs *fs;
    138 	struct mount *mp;
    139 	struct proc *p = curproc;	/* XXX */
    140 	struct ufsmount *ump;
    141 	int error;
    142 
    143 	if (root_device->dv_class != DV_DISK)
    144 		return (ENODEV);
    145 
    146 	/*
    147 	 * Get vnodes for rootdev.
    148 	 */
    149 	if (bdevvp(rootdev, &rootvp))
    150 		panic("ext2fs_mountroot: can't setup bdevvp's");
    151 
    152 	if ((error = vfs_rootmountalloc(MOUNT_EXT2FS, "root_device", &mp))) {
    153 		vrele(rootvp);
    154 		return (error);
    155 	}
    156 
    157 	if ((error = ext2fs_mountfs(rootvp, mp, p)) != 0) {
    158 		mp->mnt_op->vfs_refcount--;
    159 		vfs_unbusy(mp);
    160 		free(mp, M_MOUNT);
    161 		vrele(rootvp);
    162 		return (error);
    163 	}
    164 	simple_lock(&mountlist_slock);
    165 	CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
    166 	simple_unlock(&mountlist_slock);
    167 	ump = VFSTOUFS(mp);
    168 	fs = ump->um_e2fs;
    169 	memset(fs->e2fs_fsmnt, 0, sizeof(fs->e2fs_fsmnt));
    170 	(void) copystr(mp->mnt_stat.f_mntonname, fs->e2fs_fsmnt,
    171 	    MNAMELEN - 1, 0);
    172 	(void)ext2fs_statfs(mp, &mp->mnt_stat, p);
    173 	vfs_unbusy(mp);
    174 	inittodr(fs->e2fs.e2fs_wtime);
    175 	return (0);
    176 }
    177 
    178 /*
    179  * VFS Operations.
    180  *
    181  * mount system call
    182  */
    183 int
    184 ext2fs_mount(mp, path, data, ndp, p)
    185 	register struct mount *mp;
    186 	const char *path;
    187 	void * data;
    188 	struct nameidata *ndp;
    189 	struct proc *p;
    190 {
    191 	struct vnode *devvp;
    192 	struct ufs_args args;
    193 	struct ufsmount *ump = NULL;
    194 	register struct m_ext2fs *fs;
    195 	size_t size;
    196 	int error, flags;
    197 	mode_t accessmode;
    198 
    199 	error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args));
    200 	if (error)
    201 		return (error);
    202 	/*
    203 	 * If updating, check whether changing from read-only to
    204 	 * read/write; if there is no device name, that's all we do.
    205 	 */
    206 	if (mp->mnt_flag & MNT_UPDATE) {
    207 		ump = VFSTOUFS(mp);
    208 		fs = ump->um_e2fs;
    209 		if (fs->e2fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
    210 			flags = WRITECLOSE;
    211 			if (mp->mnt_flag & MNT_FORCE)
    212 				flags |= FORCECLOSE;
    213 			error = ext2fs_flushfiles(mp, flags, p);
    214 			if (error == 0 &&
    215 				ext2fs_cgupdate(ump, MNT_WAIT) == 0 &&
    216 				(fs->e2fs.e2fs_state & E2FS_ERRORS) == 0) {
    217 				fs->e2fs.e2fs_state = E2FS_ISCLEAN;
    218 				(void) ext2fs_sbupdate(ump, MNT_WAIT);
    219 			}
    220 			if (error)
    221 				return (error);
    222 			fs->e2fs_ronly = 1;
    223 		}
    224 		if (mp->mnt_flag & MNT_RELOAD) {
    225 			error = ext2fs_reload(mp, ndp->ni_cnd.cn_cred, p);
    226 			if (error)
    227 				return (error);
    228 		}
    229 		if (fs->e2fs_ronly && (mp->mnt_flag & MNT_WANTRDWR)) {
    230 			/*
    231 			 * If upgrade to read-write by non-root, then verify
    232 			 * that user has necessary permissions on the device.
    233 			 */
    234 			if (p->p_ucred->cr_uid != 0) {
    235 				devvp = ump->um_devvp;
    236 				vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
    237 				error = VOP_ACCESS(devvp, VREAD | VWRITE,
    238 						   p->p_ucred, p);
    239 				VOP_UNLOCK(devvp, 0);
    240 				if (error)
    241 					return (error);
    242 			}
    243 			fs->e2fs_ronly = 0;
    244 			if (fs->e2fs.e2fs_state == E2FS_ISCLEAN)
    245 				fs->e2fs.e2fs_state = 0;
    246 			else
    247 				fs->e2fs.e2fs_state = E2FS_ERRORS;
    248 			fs->e2fs_fmod = 1;
    249 		}
    250 		if (args.fspec == 0) {
    251 			/*
    252 			 * Process export requests.
    253 			 */
    254 			return (vfs_export(mp, &ump->um_export, &args.export));
    255 		}
    256 	}
    257 	/*
    258 	 * Not an update, or updating the name: look up the name
    259 	 * and verify that it refers to a sensible block device.
    260 	 */
    261 	NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, p);
    262 	if ((error = namei(ndp)) != 0)
    263 		return (error);
    264 	devvp = ndp->ni_vp;
    265 
    266 	if (devvp->v_type != VBLK) {
    267 		vrele(devvp);
    268 		return (ENOTBLK);
    269 	}
    270 	if (major(devvp->v_rdev) >= nblkdev) {
    271 		vrele(devvp);
    272 		return (ENXIO);
    273 	}
    274 	/*
    275 	 * If mount by non-root, then verify that user has necessary
    276 	 * permissions on the device.
    277 	 */
    278 	if (p->p_ucred->cr_uid != 0) {
    279 		accessmode = VREAD;
    280 		if ((mp->mnt_flag & MNT_RDONLY) == 0)
    281 			accessmode |= VWRITE;
    282 		vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
    283 		error = VOP_ACCESS(devvp, accessmode, p->p_ucred, p);
    284 		VOP_UNLOCK(devvp, 0);
    285 		if (error) {
    286 			vrele(devvp);
    287 			return (error);
    288 		}
    289 	}
    290 	if ((mp->mnt_flag & MNT_UPDATE) == 0)
    291 		error = ext2fs_mountfs(devvp, mp, p);
    292 	else {
    293 		if (devvp != ump->um_devvp)
    294 			error = EINVAL;	/* needs translation */
    295 		else
    296 			vrele(devvp);
    297 	}
    298 	if (error) {
    299 		vrele(devvp);
    300 		return (error);
    301 	}
    302 	ump = VFSTOUFS(mp);
    303 	fs = ump->um_e2fs;
    304 	(void) copyinstr(path, fs->e2fs_fsmnt, sizeof(fs->e2fs_fsmnt) - 1, &size);
    305 	memset(fs->e2fs_fsmnt + size, 0, sizeof(fs->e2fs_fsmnt) - size);
    306 	memcpy(mp->mnt_stat.f_mntonname, fs->e2fs_fsmnt, MNAMELEN);
    307 	(void) copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
    308 		&size);
    309 	memset(mp->mnt_stat.f_mntfromname + size, 0, MNAMELEN - size);
    310 	if (fs->e2fs_fmod != 0) {	/* XXX */
    311 		fs->e2fs_fmod = 0;
    312 		if (fs->e2fs.e2fs_state == 0)
    313 			fs->e2fs.e2fs_wtime = time.tv_sec;
    314 		else
    315 			printf("%s: file system not clean; please fsck(8)\n",
    316 				mp->mnt_stat.f_mntfromname);
    317 		(void) ext2fs_cgupdate(ump, MNT_WAIT);
    318 	}
    319 	return (0);
    320 }
    321 
    322 /*
    323  * Reload all incore data for a filesystem (used after running fsck on
    324  * the root filesystem and finding things to fix). The filesystem must
    325  * be mounted read-only.
    326  *
    327  * Things to do to update the mount:
    328  *	1) invalidate all cached meta-data.
    329  *	2) re-read superblock from disk.
    330  *	3) re-read summary information from disk.
    331  *	4) invalidate all inactive vnodes.
    332  *	5) invalidate all cached file data.
    333  *	6) re-read inode data for all active vnodes.
    334  */
    335 int
    336 ext2fs_reload(mountp, cred, p)
    337 	register struct mount *mountp;
    338 	struct ucred *cred;
    339 	struct proc *p;
    340 {
    341 	register struct vnode *vp, *nvp, *devvp;
    342 	struct inode *ip;
    343 	struct buf *bp;
    344 	struct m_ext2fs *fs;
    345 	struct ext2fs *newfs;
    346 	struct partinfo dpart;
    347 	int i, size, error;
    348 	caddr_t cp;
    349 
    350 	if ((mountp->mnt_flag & MNT_RDONLY) == 0)
    351 		return (EINVAL);
    352 	/*
    353 	 * Step 1: invalidate all cached meta-data.
    354 	 */
    355 	devvp = VFSTOUFS(mountp)->um_devvp;
    356 	if (vinvalbuf(devvp, 0, cred, p, 0, 0))
    357 		panic("ext2fs_reload: dirty1");
    358 	/*
    359 	 * Step 2: re-read superblock from disk.
    360 	 */
    361 	if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, p) != 0)
    362 		size = DEV_BSIZE;
    363 	else
    364 		size = dpart.disklab->d_secsize;
    365 	error = bread(devvp, (ufs_daddr_t)(SBOFF / size), SBSIZE, NOCRED, &bp);
    366 	if (error) {
    367 		brelse(bp);
    368 		return (error);
    369 	}
    370 	newfs = (struct ext2fs *)bp->b_data;
    371 	if (fs2h16(newfs->e2fs_magic) != E2FS_MAGIC) {
    372 		brelse(bp);
    373 		return (EIO);		/* XXX needs translation */
    374 	}
    375 	if (fs2h32(newfs->e2fs_rev) != E2FS_REV) {
    376 #ifdef DIAGNOSTIC
    377 		printf("Ext2 fs: unsupported revision number: %x (expected %x)\n",
    378 					fs2h32(newfs->e2fs_rev), E2FS_REV);
    379 #endif
    380 		brelse(bp);
    381 		return (EIO);		/* XXX needs translation */
    382 	}
    383 	if (fs2h32(newfs->e2fs_log_bsize) > 2) { /* block size = 1024|2048|4096 */
    384 #ifdef DIAGNOSTIC
    385 		printf("Ext2 fs: bad block size: %d (expected <=2 for ext2 fs)\n",
    386 			fs2h32(newfs->e2fs_log_bsize));
    387 #endif
    388 		brelse(bp);
    389 		return (EIO);	   /* XXX needs translation */
    390 	}
    391 
    392 
    393 	fs = VFSTOUFS(mountp)->um_e2fs;
    394 	/*
    395 	 * copy in new superblock, and compute in-memory values
    396 	 */
    397 	e2fs_sbload(newfs, &fs->e2fs);
    398 	fs->e2fs_ncg =
    399 	    howmany(fs->e2fs.e2fs_bcount - fs->e2fs.e2fs_first_dblock,
    400 	    fs->e2fs.e2fs_bpg);
    401 	/* XXX assume hw bsize = 512 */
    402 	fs->e2fs_fsbtodb = fs->e2fs.e2fs_log_bsize + 1;
    403 	fs->e2fs_bsize = 1024 << fs->e2fs.e2fs_log_bsize;
    404 	fs->e2fs_bshift = LOG_MINBSIZE + fs->e2fs.e2fs_log_bsize;
    405 	fs->e2fs_qbmask = fs->e2fs_bsize - 1;
    406 	fs->e2fs_bmask = ~fs->e2fs_qbmask;
    407 	fs->e2fs_ngdb = howmany(fs->e2fs_ncg,
    408 			fs->e2fs_bsize / sizeof(struct ext2_gd));
    409 	fs->e2fs_ipb = fs->e2fs_bsize / EXT2_DINODE_SIZE;
    410 	fs->e2fs_itpg = fs->e2fs.e2fs_ipg/fs->e2fs_ipb;
    411 
    412 	/*
    413 	 * Step 3: re-read summary information from disk.
    414 	 */
    415 
    416 	for (i=0; i < fs->e2fs_ngdb; i++) {
    417 		error = bread(devvp ,
    418 		    fsbtodb(fs, ((fs->e2fs_bsize>1024)? 0 : 1) + i + 1),
    419 		    fs->e2fs_bsize, NOCRED, &bp);
    420 		if (error) {
    421 			brelse(bp);
    422 			return (error);
    423 		}
    424 		e2fs_cgload((struct ext2_gd*)bp->b_data,
    425 		    &fs->e2fs_gd[i* fs->e2fs_bsize / sizeof(struct ext2_gd)],
    426 		    fs->e2fs_bsize);
    427 		brelse(bp);
    428 	}
    429 
    430 loop:
    431 	simple_lock(&mntvnode_slock);
    432 	for (vp = mountp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) {
    433 		if (vp->v_mount != mountp) {
    434 			simple_unlock(&mntvnode_slock);
    435 			goto loop;
    436 		}
    437 		nvp = vp->v_mntvnodes.le_next;
    438 		/*
    439 		 * Step 4: invalidate all inactive vnodes.
    440 		 */
    441 		if (vrecycle(vp, &mntvnode_slock, p))
    442 			goto loop;
    443 		/*
    444 		 * Step 5: invalidate all cached file data.
    445 		 */
    446 		simple_lock(&vp->v_interlock);
    447 		simple_unlock(&mntvnode_slock);
    448 		if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK))
    449 			goto loop;
    450 		if (vinvalbuf(vp, 0, cred, p, 0, 0))
    451 			panic("ext2fs_reload: dirty2");
    452 		/*
    453 		 * Step 6: re-read inode data for all active vnodes.
    454 		 */
    455 		ip = VTOI(vp);
    456 		error = bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
    457 				  (int)fs->e2fs_bsize, NOCRED, &bp);
    458 		if (error) {
    459 			vput(vp);
    460 			return (error);
    461 		}
    462 		cp = (caddr_t)bp->b_data +
    463 		    (ino_to_fsbo(fs, ip->i_number) * EXT2_DINODE_SIZE);
    464 		e2fs_iload((struct ext2fs_dinode *)cp, &ip->i_din.e2fs_din);
    465 		brelse(bp);
    466 		vput(vp);
    467 		simple_lock(&mntvnode_slock);
    468 	}
    469 	simple_unlock(&mntvnode_slock);
    470 	return (0);
    471 }
    472 
    473 /*
    474  * Common code for mount and mountroot
    475  */
    476 int
    477 ext2fs_mountfs(devvp, mp, p)
    478 	register struct vnode *devvp;
    479 	struct mount *mp;
    480 	struct proc *p;
    481 {
    482 	register struct ufsmount *ump;
    483 	struct buf *bp;
    484 	register struct ext2fs *fs;
    485 	register struct m_ext2fs *m_fs;
    486 	dev_t dev;
    487 	struct partinfo dpart;
    488 	int error, i, size, ronly;
    489 	struct ucred *cred;
    490 	extern struct vnode *rootvp;
    491 
    492 	dev = devvp->v_rdev;
    493 	cred = p ? p->p_ucred : NOCRED;
    494 	/*
    495 	 * Disallow multiple mounts of the same device.
    496 	 * Disallow mounting of a device that is currently in use
    497 	 * (except for root, which might share swap device for miniroot).
    498 	 * Flush out any old buffers remaining from a previous use.
    499 	 */
    500 	if ((error = vfs_mountedon(devvp)) != 0)
    501 		return (error);
    502 	if (vcount(devvp) > 1 && devvp != rootvp)
    503 		return (EBUSY);
    504 	if ((error = vinvalbuf(devvp, V_SAVE, cred, p, 0, 0)) != 0)
    505 		return (error);
    506 
    507 	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
    508 	error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p);
    509 	if (error)
    510 		return (error);
    511 	if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, cred, p) != 0)
    512 		size = DEV_BSIZE;
    513 	else
    514 		size = dpart.disklab->d_secsize;
    515 
    516 	bp = NULL;
    517 	ump = NULL;
    518 
    519 #ifdef DEBUG_EXT2
    520 	printf("sb size: %d ino size %d\n", sizeof(struct ext2fs),
    521 	    EXT2_DINODE_SIZE);
    522 #endif
    523 	error = bread(devvp, (SBOFF / DEV_BSIZE), SBSIZE, cred, &bp);
    524 	if (error)
    525 		goto out;
    526 	fs = (struct ext2fs *)bp->b_data;
    527 	if (fs2h16(fs->e2fs_magic) != E2FS_MAGIC) {
    528 		error = EINVAL;		/* XXX needs translation */
    529 		goto out;
    530 	}
    531 	if (fs2h32(fs->e2fs_rev) != E2FS_REV) {
    532 #ifdef DIAGNOSTIC
    533 		printf("Ext2 fs: unsupported revision number: %x (expected %x)\n",
    534 					fs2h32(fs->e2fs_rev), E2FS_REV);
    535 #endif
    536 		error = EINVAL;		/* XXX needs translation */
    537 		goto out;
    538 	}
    539 	if (fs2h32(fs->e2fs_log_bsize) > 2) { /* block size = 1024|2048|4096 */
    540 #ifdef DIAGNOSTIC
    541 		printf("Ext2 fs: bad block size: %d (expected <=2 for ext2 fs)\n",
    542 			fs2h32(fs->e2fs_log_bsize));
    543 #endif
    544 		error = EINVAL;	 /* XXX needs translation */
    545 		goto out;
    546 	}
    547 
    548 	ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
    549 	memset((caddr_t)ump, 0, sizeof *ump);
    550 	ump->um_e2fs = malloc(sizeof(struct m_ext2fs), M_UFSMNT, M_WAITOK);
    551 	memset((caddr_t)ump->um_e2fs, 0, sizeof(struct m_ext2fs));
    552 	e2fs_sbload((struct ext2fs*)bp->b_data, &ump->um_e2fs->e2fs);
    553 	brelse(bp);
    554 	bp = NULL;
    555 	m_fs = ump->um_e2fs;
    556 	m_fs->e2fs_ronly = ronly;
    557 	if (ronly == 0) {
    558 		if (m_fs->e2fs.e2fs_state == E2FS_ISCLEAN)
    559 			m_fs->e2fs.e2fs_state = 0;
    560 		else
    561 			m_fs->e2fs.e2fs_state = E2FS_ERRORS;
    562 		m_fs->e2fs_fmod = 1;
    563 	}
    564 
    565 	/* compute dynamic sb infos */
    566 	m_fs->e2fs_ncg =
    567 		howmany(m_fs->e2fs.e2fs_bcount - m_fs->e2fs.e2fs_first_dblock,
    568 		m_fs->e2fs.e2fs_bpg);
    569 	/* XXX assume hw bsize = 512 */
    570 	m_fs->e2fs_fsbtodb = m_fs->e2fs.e2fs_log_bsize + 1;
    571 	m_fs->e2fs_bsize = 1024 << m_fs->e2fs.e2fs_log_bsize;
    572 	m_fs->e2fs_bshift = LOG_MINBSIZE + m_fs->e2fs.e2fs_log_bsize;
    573 	m_fs->e2fs_qbmask = m_fs->e2fs_bsize - 1;
    574 	m_fs->e2fs_bmask = ~m_fs->e2fs_qbmask;
    575 	m_fs->e2fs_ngdb = howmany(m_fs->e2fs_ncg,
    576 		m_fs->e2fs_bsize / sizeof(struct ext2_gd));
    577 	m_fs->e2fs_ipb = m_fs->e2fs_bsize / EXT2_DINODE_SIZE;
    578 	m_fs->e2fs_itpg = m_fs->e2fs.e2fs_ipg/m_fs->e2fs_ipb;
    579 
    580 	m_fs->e2fs_gd = malloc(m_fs->e2fs_ngdb * m_fs->e2fs_bsize,
    581 		M_UFSMNT, M_WAITOK);
    582 	for (i=0; i < m_fs->e2fs_ngdb; i++) {
    583 		error = bread(devvp ,
    584 		    fsbtodb(m_fs, ((m_fs->e2fs_bsize>1024)? 0 : 1) + i + 1),
    585 		    m_fs->e2fs_bsize, NOCRED, &bp);
    586 		if (error) {
    587 			free(m_fs->e2fs_gd, M_UFSMNT);
    588 			goto out;
    589 		}
    590 		e2fs_cgload((struct ext2_gd*)bp->b_data,
    591 		    &m_fs->e2fs_gd[
    592 			i * m_fs->e2fs_bsize / sizeof(struct ext2_gd)],
    593 		    m_fs->e2fs_bsize);
    594 		brelse(bp);
    595 		bp = NULL;
    596 	}
    597 
    598 	mp->mnt_data = (qaddr_t)ump;
    599 	mp->mnt_stat.f_fsid.val[0] = (long)dev;
    600 	mp->mnt_stat.f_fsid.val[1] = makefstype(MOUNT_EXT2FS);
    601 	mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN;
    602 	mp->mnt_flag |= MNT_LOCAL;
    603 	ump->um_flags = 0;
    604 	ump->um_mountp = mp;
    605 	ump->um_dev = dev;
    606 	ump->um_devvp = devvp;
    607 	ump->um_nindir = NINDIR(m_fs);
    608 	ump->um_bptrtodb = m_fs->e2fs_fsbtodb;
    609 	ump->um_seqinc = 1; /* no frags */
    610 	devvp->v_specflags |= SI_MOUNTEDON;
    611 	return (0);
    612 out:
    613 	if (bp)
    614 		brelse(bp);
    615 	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
    616 	(void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, cred, p);
    617 	VOP_UNLOCK(devvp, 0);
    618 	if (ump) {
    619 		free(ump->um_e2fs, M_UFSMNT);
    620 		free(ump, M_UFSMNT);
    621 		mp->mnt_data = (qaddr_t)0;
    622 	}
    623 	return (error);
    624 }
    625 
    626 /*
    627  * unmount system call
    628  */
    629 int
    630 ext2fs_unmount(mp, mntflags, p)
    631 	struct mount *mp;
    632 	int mntflags;
    633 	struct proc *p;
    634 {
    635 	register struct ufsmount *ump;
    636 	register struct m_ext2fs *fs;
    637 	int error, flags;
    638 
    639 	flags = 0;
    640 	if (mntflags & MNT_FORCE)
    641 		flags |= FORCECLOSE;
    642 	if ((error = ext2fs_flushfiles(mp, flags, p)) != 0)
    643 		return (error);
    644 	ump = VFSTOUFS(mp);
    645 	fs = ump->um_e2fs;
    646 	if (fs->e2fs_ronly == 0 &&
    647 		ext2fs_cgupdate(ump, MNT_WAIT) == 0 &&
    648 		(fs->e2fs.e2fs_state & E2FS_ERRORS) == 0) {
    649 		fs->e2fs.e2fs_state = E2FS_ISCLEAN;
    650 		(void) ext2fs_sbupdate(ump, MNT_WAIT);
    651 	}
    652 	if (ump->um_devvp->v_type != VBAD)
    653 		ump->um_devvp->v_specflags &= ~SI_MOUNTEDON;
    654 	vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
    655 	error = VOP_CLOSE(ump->um_devvp, fs->e2fs_ronly ? FREAD : FREAD|FWRITE,
    656 		NOCRED, p);
    657 	vput(ump->um_devvp);
    658 	free(fs->e2fs_gd, M_UFSMNT);
    659 	free(fs, M_UFSMNT);
    660 	free(ump, M_UFSMNT);
    661 	mp->mnt_data = (qaddr_t)0;
    662 	mp->mnt_flag &= ~MNT_LOCAL;
    663 	return (error);
    664 }
    665 
    666 /*
    667  * Flush out all the files in a filesystem.
    668  */
    669 int
    670 ext2fs_flushfiles(mp, flags, p)
    671 	register struct mount *mp;
    672 	int flags;
    673 	struct proc *p;
    674 {
    675 	extern int doforce;
    676 	register struct ufsmount *ump;
    677 	int error;
    678 
    679 	if (!doforce)
    680 		flags &= ~FORCECLOSE;
    681 	ump = VFSTOUFS(mp);
    682 	error = vflush(mp, NULLVP, flags);
    683 	return (error);
    684 }
    685 
    686 /*
    687  * Get file system statistics.
    688  */
    689 int
    690 ext2fs_statfs(mp, sbp, p)
    691 	struct mount *mp;
    692 	register struct statfs *sbp;
    693 	struct proc *p;
    694 {
    695 	register struct ufsmount *ump;
    696 	register struct m_ext2fs *fs;
    697 	u_int32_t overhead, overhead_per_group;
    698 
    699 	ump = VFSTOUFS(mp);
    700 	fs = ump->um_e2fs;
    701 	if (fs->e2fs.e2fs_magic != E2FS_MAGIC)
    702 		panic("ext2fs_statfs");
    703 
    704 #ifdef COMPAT_09
    705 	sbp->f_type = 1;
    706 #else
    707 	sbp->f_type = 0;
    708 #endif
    709 
    710 	/*
    711 	 * Compute the overhead (FS structures)
    712 	 */
    713 	overhead_per_group = 1 /* super block */ +
    714 						 fs->e2fs_ngdb +
    715 						 1 /* block bitmap */ +
    716 						 1 /* inode bitmap */ +
    717 						 fs->e2fs_itpg;
    718 	overhead = fs->e2fs.e2fs_first_dblock +
    719 		   fs->e2fs_ncg * overhead_per_group;
    720 
    721 
    722 	sbp->f_bsize = fs->e2fs_bsize;
    723 	sbp->f_iosize = fs->e2fs_bsize;
    724 	sbp->f_blocks = fs->e2fs.e2fs_bcount - overhead;
    725 	sbp->f_bfree = fs->e2fs.e2fs_fbcount;
    726 	sbp->f_bavail = sbp->f_bfree - fs->e2fs.e2fs_rbcount;
    727 	sbp->f_files =  fs->e2fs.e2fs_icount;
    728 	sbp->f_ffree = fs->e2fs.e2fs_ficount;
    729 	if (sbp != &mp->mnt_stat) {
    730 		memcpy(sbp->f_mntonname, mp->mnt_stat.f_mntonname, MNAMELEN);
    731 		memcpy(sbp->f_mntfromname, mp->mnt_stat.f_mntfromname, MNAMELEN);
    732 	}
    733 	strncpy(sbp->f_fstypename, mp->mnt_op->vfs_name, MFSNAMELEN);
    734 	return (0);
    735 }
    736 
    737 /*
    738  * Go through the disk queues to initiate sandbagged IO;
    739  * go through the inodes to write those that have been modified;
    740  * initiate the writing of the super block if it has been modified.
    741  *
    742  * Note: we are always called with the filesystem marked `MPBUSY'.
    743  */
    744 int
    745 ext2fs_sync(mp, waitfor, cred, p)
    746 	struct mount *mp;
    747 	int waitfor;
    748 	struct ucred *cred;
    749 	struct proc *p;
    750 {
    751 	struct vnode *vp, *nvp;
    752 	struct inode *ip;
    753 	struct ufsmount *ump = VFSTOUFS(mp);
    754 	struct m_ext2fs *fs;
    755 	int error, allerror = 0;
    756 
    757 	fs = ump->um_e2fs;
    758 	if (fs->e2fs_ronly != 0) {		/* XXX */
    759 		printf("fs = %s\n", fs->e2fs_fsmnt);
    760 		panic("update: rofs mod");
    761 	}
    762 
    763 	/*
    764 	 * Write back each (modified) inode.
    765 	 */
    766 	simple_lock(&mntvnode_slock);
    767 loop:
    768 	for (vp = mp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) {
    769 		/*
    770 		 * If the vnode that we are about to sync is no longer
    771 		 * associated with this mount point, start over.
    772 		 */
    773 		if (vp->v_mount != mp)
    774 			goto loop;
    775 		simple_lock(&vp->v_interlock);
    776 		nvp = vp->v_mntvnodes.le_next;
    777 		ip = VTOI(vp);
    778 		if ((ip->i_flag &
    779 		    (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
    780 			vp->v_dirtyblkhd.lh_first == NULL) {
    781 			simple_unlock(&vp->v_interlock);
    782 			continue;
    783 		}
    784 		simple_unlock(&mntvnode_slock);
    785 		error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK);
    786 		if (error) {
    787 			simple_lock(&mntvnode_slock);
    788 			if (error == ENOENT)
    789 				goto loop;
    790 			continue;
    791 		}
    792 		if ((error = VOP_FSYNC(vp, cred,
    793 		    waitfor == MNT_WAIT ? FSYNC_WAIT : 0, p)) != 0)
    794 			allerror = error;
    795 		vput(vp);
    796 		simple_lock(&mntvnode_slock);
    797 	}
    798 	simple_unlock(&mntvnode_slock);
    799 	/*
    800 	 * Force stale file system control information to be flushed.
    801 	 */
    802 	if ((error = VOP_FSYNC(ump->um_devvp, cred,
    803 	    waitfor == MNT_WAIT ? FSYNC_WAIT : 0, p)) != 0)
    804 		allerror = error;
    805 	/*
    806 	 * Write back modified superblock.
    807 	 */
    808 	if (fs->e2fs_fmod != 0) {
    809 		fs->e2fs_fmod = 0;
    810 		fs->e2fs.e2fs_wtime = time.tv_sec;
    811 		if ((error = ext2fs_cgupdate(ump, waitfor)))
    812 			allerror = error;
    813 	}
    814 	return (allerror);
    815 }
    816 
    817 /*
    818  * Look up a EXT2FS dinode number to find its incore vnode, otherwise read it
    819  * in from disk.  If it is in core, wait for the lock bit to clear, then
    820  * return the inode locked.  Detection and handling of mount points must be
    821  * done by the calling routine.
    822  */
    823 int
    824 ext2fs_vget(mp, ino, vpp)
    825 	struct mount *mp;
    826 	ino_t ino;
    827 	struct vnode **vpp;
    828 {
    829 	struct m_ext2fs *fs;
    830 	struct inode *ip;
    831 	struct ufsmount *ump;
    832 	struct buf *bp;
    833 	struct vnode *vp;
    834 	dev_t dev;
    835 	int error;
    836 	caddr_t cp;
    837 
    838 	ump = VFSTOUFS(mp);
    839 	dev = ump->um_dev;
    840 	do {
    841 		if ((*vpp = ufs_ihashget(dev, ino)) != NULL)
    842 			return (0);
    843 	} while (lockmgr(&ufs_hashlock, LK_EXCLUSIVE|LK_SLEEPFAIL, 0));
    844 
    845 	/* Allocate a new vnode/inode. */
    846 	if ((error = getnewvnode(VT_EXT2FS, mp, ext2fs_vnodeop_p, &vp)) != 0) {
    847 		*vpp = NULL;
    848 		lockmgr(&ufs_hashlock, LK_RELEASE, 0);
    849 		return (error);
    850 	}
    851 	ip = pool_get(&ext2fs_inode_pool, PR_WAITOK);
    852 	memset((caddr_t)ip, 0, sizeof(struct inode));
    853 	vp->v_data = ip;
    854 	ip->i_vnode = vp;
    855 	ip->i_e2fs = fs = ump->um_e2fs;
    856 	ip->i_dev = dev;
    857 	ip->i_number = ino;
    858 	ip->i_e2fs_last_lblk = 0;
    859 	ip->i_e2fs_last_blk = 0;
    860 
    861 	/*
    862 	 * Put it onto its hash chain and lock it so that other requests for
    863 	 * this inode will block if they arrive while we are sleeping waiting
    864 	 * for old data structures to be purged or for the contents of the
    865 	 * disk portion of this inode to be read.
    866 	 */
    867 	ufs_ihashins(ip);
    868 	lockmgr(&ufs_hashlock, LK_RELEASE, 0);
    869 
    870 	/* Read in the disk contents for the inode, copy into the inode. */
    871 	error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
    872 			  (int)fs->e2fs_bsize, NOCRED, &bp);
    873 	if (error) {
    874 		/*
    875 		 * The inode does not contain anything useful, so it would
    876 		 * be misleading to leave it on its hash chain. With mode
    877 		 * still zero, it will be unlinked and returned to the free
    878 		 * list by vput().
    879 		 */
    880 		vput(vp);
    881 		brelse(bp);
    882 		*vpp = NULL;
    883 		return (error);
    884 	}
    885 	cp = (caddr_t)bp->b_data +
    886 	    (ino_to_fsbo(fs, ino) * EXT2_DINODE_SIZE);
    887 	e2fs_iload((struct ext2fs_dinode *)cp, &ip->i_din.e2fs_din);
    888 	brelse(bp);
    889 
    890 	/* If the inode was deleted, reset all fields */
    891 	if (ip->i_e2fs_dtime != 0) {
    892 		ip->i_e2fs_mode = ip->i_e2fs_size = ip->i_e2fs_nblock = 0;
    893 		memset(ip->i_e2fs_blocks, 0, sizeof(ip->i_e2fs_blocks));
    894 	}
    895 
    896 	/*
    897 	 * Initialize the vnode from the inode, check for aliases.
    898 	 * Note that the underlying vnode may have changed.
    899 	 */
    900 	error = ext2fs_vinit(mp, ext2fs_specop_p, ext2fs_fifoop_p, &vp);
    901 	if (error) {
    902 		vput(vp);
    903 		*vpp = NULL;
    904 		return (error);
    905 	}
    906 	/*
    907 	 * Finish inode initialization now that aliasing has been resolved.
    908 	 */
    909 	ip->i_devvp = ump->um_devvp;
    910 	VREF(ip->i_devvp);
    911 	/*
    912 	 * Set up a generation number for this inode if it does not
    913 	 * already have one. This should only happen on old filesystems.
    914 	 */
    915 	if (ip->i_e2fs_gen == 0) {
    916 		if (++ext2gennumber < (u_long)time.tv_sec)
    917 			ext2gennumber = time.tv_sec;
    918 		ip->i_e2fs_gen = ext2gennumber;
    919 		if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
    920 			ip->i_flag |= IN_MODIFIED;
    921 	}
    922 
    923 	*vpp = vp;
    924 	return (0);
    925 }
    926 
    927 /*
    928  * File handle to vnode
    929  *
    930  * Have to be really careful about stale file handles:
    931  * - check that the inode number is valid
    932  * - call ext2fs_vget() to get the locked inode
    933  * - check for an unallocated inode (i_mode == 0)
    934  */
    935 int
    936 ext2fs_fhtovp(mp, fhp, vpp)
    937 	register struct mount *mp;
    938 	struct fid *fhp;
    939 	struct vnode **vpp;
    940 {
    941 	register struct inode *ip;
    942 	struct vnode *nvp;
    943 	int error;
    944 	register struct ufid *ufhp;
    945 	struct m_ext2fs *fs;
    946 
    947 	ufhp = (struct ufid *)fhp;
    948 	fs = VFSTOUFS(mp)->um_e2fs;
    949 	if ((ufhp->ufid_ino < EXT2_FIRSTINO && ufhp->ufid_ino != EXT2_ROOTINO) ||
    950 		ufhp->ufid_ino >= fs->e2fs_ncg * fs->e2fs.e2fs_ipg)
    951 		return (ESTALE);
    952 
    953 	if ((error = VFS_VGET(mp, ufhp->ufid_ino, &nvp)) != 0) {
    954 		*vpp = NULLVP;
    955 		return (error);
    956 	}
    957 	ip = VTOI(nvp);
    958 	if (ip->i_e2fs_mode == 0 || ip->i_e2fs_dtime != 0 ||
    959 		ip->i_e2fs_gen != ufhp->ufid_gen) {
    960 		vput(nvp);
    961 		*vpp = NULLVP;
    962 		return (ESTALE);
    963 	}
    964 	*vpp = nvp;
    965 	return (0);
    966 }
    967 
    968 /*
    969  * Vnode pointer to File handle
    970  */
    971 /* ARGSUSED */
    972 int
    973 ext2fs_vptofh(vp, fhp)
    974 	struct vnode *vp;
    975 	struct fid *fhp;
    976 {
    977 	register struct inode *ip;
    978 	register struct ufid *ufhp;
    979 
    980 	ip = VTOI(vp);
    981 	ufhp = (struct ufid *)fhp;
    982 	ufhp->ufid_len = sizeof(struct ufid);
    983 	ufhp->ufid_ino = ip->i_number;
    984 	ufhp->ufid_gen = ip->i_e2fs_gen;
    985 	return (0);
    986 }
    987 
    988 int
    989 ext2fs_sysctl(name, namelen, oldp, oldlenp, newp, newlen, p)
    990 	int *name;
    991 	u_int namelen;
    992 	void *oldp;
    993 	size_t *oldlenp;
    994 	void *newp;
    995 	size_t newlen;
    996 	struct proc *p;
    997 {
    998 	return (EOPNOTSUPP);
    999 }
   1000 
   1001 /*
   1002  * Write a superblock and associated information back to disk.
   1003  */
   1004 int
   1005 ext2fs_sbupdate(mp, waitfor)
   1006 	struct ufsmount *mp;
   1007 	int waitfor;
   1008 {
   1009 	register struct m_ext2fs *fs = mp->um_e2fs;
   1010 	register struct buf *bp;
   1011 	int error = 0;
   1012 
   1013 	bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0);
   1014 	e2fs_sbsave(&fs->e2fs, (struct ext2fs*)bp->b_data);
   1015 	if (waitfor == MNT_WAIT)
   1016 		error = bwrite(bp);
   1017 	else
   1018 		bawrite(bp);
   1019 	return (error);
   1020 }
   1021 
   1022 int
   1023 ext2fs_cgupdate(mp, waitfor)
   1024 	struct ufsmount *mp;
   1025 	int waitfor;
   1026 {
   1027 	register struct m_ext2fs *fs = mp->um_e2fs;
   1028 	register struct buf *bp;
   1029 	int i, error = 0, allerror = 0;
   1030 
   1031 	allerror = ext2fs_sbupdate(mp, waitfor);
   1032 	for (i = 0; i < fs->e2fs_ngdb; i++) {
   1033 		bp = getblk(mp->um_devvp, fsbtodb(fs, ((fs->e2fs_bsize>1024)?0:1)+i+1),
   1034 			fs->e2fs_bsize, 0, 0);
   1035 		e2fs_cgsave(&fs->e2fs_gd[i* fs->e2fs_bsize / sizeof(struct ext2_gd)],
   1036 				(struct ext2_gd*)bp->b_data, fs->e2fs_bsize);
   1037 		if (waitfor == MNT_WAIT)
   1038 			error = bwrite(bp);
   1039 		else
   1040 			bawrite(bp);
   1041 	}
   1042 
   1043 	if (!allerror && error)
   1044 		allerror = error;
   1045 	return (allerror);
   1046 }
   1047