Home | History | Annotate | Line # | Download | only in ffs
ffs_vfsops.c revision 1.15
      1 /*	$NetBSD: ffs_vfsops.c,v 1.15 1995/04/12 21:21:00 mycroft Exp $	*/
      2 
      3 /*
      4  * Copyright (c) 1989, 1991, 1993, 1994
      5  *	The Regents of the University of California.  All rights reserved.
      6  *
      7  * Redistribution and use in source and binary forms, with or without
      8  * modification, are permitted provided that the following conditions
      9  * are met:
     10  * 1. Redistributions of source code must retain the above copyright
     11  *    notice, this list of conditions and the following disclaimer.
     12  * 2. Redistributions in binary form must reproduce the above copyright
     13  *    notice, this list of conditions and the following disclaimer in the
     14  *    documentation and/or other materials provided with the distribution.
     15  * 3. All advertising materials mentioning features or use of this software
     16  *    must display the following acknowledgement:
     17  *	This product includes software developed by the University of
     18  *	California, Berkeley and its contributors.
     19  * 4. Neither the name of the University nor the names of its contributors
     20  *    may be used to endorse or promote products derived from this software
     21  *    without specific prior written permission.
     22  *
     23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     33  * SUCH DAMAGE.
     34  *
     35  *	@(#)ffs_vfsops.c	8.14 (Berkeley) 11/28/94
     36  */
     37 
     38 #include <sys/param.h>
     39 #include <sys/systm.h>
     40 #include <sys/namei.h>
     41 #include <sys/proc.h>
     42 #include <sys/kernel.h>
     43 #include <sys/vnode.h>
     44 #include <sys/socket.h>
     45 #include <sys/mount.h>
     46 #include <sys/buf.h>
     47 #include <sys/mbuf.h>
     48 #include <sys/file.h>
     49 #include <sys/disklabel.h>
     50 #include <sys/ioctl.h>
     51 #include <sys/errno.h>
     52 #include <sys/malloc.h>
     53 
     54 #include <miscfs/specfs/specdev.h>
     55 
     56 #include <ufs/ufs/quota.h>
     57 #include <ufs/ufs/ufsmount.h>
     58 #include <ufs/ufs/inode.h>
     59 #include <ufs/ufs/ufs_extern.h>
     60 
     61 #include <ufs/ffs/fs.h>
     62 #include <ufs/ffs/ffs_extern.h>
     63 
     64 int ffs_sbupdate __P((struct ufsmount *, int));
     65 
     66 struct vfsops ufs_vfsops = {
     67 	MOUNT_UFS,
     68 	ffs_mount,
     69 	ufs_start,
     70 	ffs_unmount,
     71 	ufs_root,
     72 	ufs_quotactl,
     73 	ffs_statfs,
     74 	ffs_sync,
     75 	ffs_vget,
     76 	ffs_fhtovp,
     77 	ffs_vptofh,
     78 	ffs_init,
     79 };
     80 
     81 extern u_long nextgennumber;
     82 
     83 /*
     84  * Called by main() when ufs is going to be mounted as root.
     85  *
     86  * Name is updated by mount(8) after booting.
     87  */
     88 #define ROOTNAME	"root_device"
     89 
     90 ffs_mountroot()
     91 {
     92 	extern struct vnode *rootvp;
     93 	register struct fs *fs;
     94 	register struct mount *mp;
     95 	struct proc *p = curproc;	/* XXX */
     96 	struct ufsmount *ump;
     97 	size_t size;
     98 	int error;
     99 
    100 	/*
    101 	 * Get vnodes for swapdev and rootdev.
    102 	 */
    103 	if (bdevvp(swapdev, &swapdev_vp) || bdevvp(rootdev, &rootvp))
    104 		panic("ffs_mountroot: can't setup bdevvp's");
    105 
    106 	mp = malloc((u_long)sizeof(struct mount), M_MOUNT, M_WAITOK);
    107 	bzero((char *)mp, (u_long)sizeof(struct mount));
    108 	mp->mnt_op = &ufs_vfsops;
    109 	mp->mnt_flag = MNT_RDONLY;
    110 	if (error = ffs_mountfs(rootvp, mp, p)) {
    111 		free(mp, M_MOUNT);
    112 		return (error);
    113 	}
    114 	if (error = vfs_lock(mp)) {
    115 		(void)ffs_unmount(mp, 0, p);
    116 		free(mp, M_MOUNT);
    117 		return (error);
    118 	}
    119 	CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
    120 	mp->mnt_vnodecovered = NULLVP;
    121 	ump = VFSTOUFS(mp);
    122 	fs = ump->um_fs;
    123 	bzero(fs->fs_fsmnt, sizeof(fs->fs_fsmnt));
    124 	fs->fs_fsmnt[0] = '/';
    125 	bcopy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname, MNAMELEN);
    126 	(void) copystr(ROOTNAME, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
    127 	    &size);
    128 	bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
    129 	(void)ffs_statfs(mp, &mp->mnt_stat, p);
    130 	vfs_unlock(mp);
    131 	inittodr(fs->fs_time);
    132 	return (0);
    133 }
    134 
    135 /*
    136  * VFS Operations.
    137  *
    138  * mount system call
    139  */
    140 int
    141 ffs_mount(mp, path, data, ndp, p)
    142 	register struct mount *mp;
    143 	char *path;
    144 	caddr_t data;
    145 	struct nameidata *ndp;
    146 	struct proc *p;
    147 {
    148 	struct vnode *devvp;
    149 	struct ufs_args args;
    150 	struct ufsmount *ump;
    151 	register struct fs *fs;
    152 	size_t size;
    153 	int error, flags;
    154 	mode_t accessmode;
    155 
    156 	if (error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args)))
    157 		return (error);
    158 	/*
    159 	 * If updating, check whether changing from read-only to
    160 	 * read/write; if there is no device name, that's all we do.
    161 	 */
    162 	if (mp->mnt_flag & MNT_UPDATE) {
    163 		ump = VFSTOUFS(mp);
    164 		fs = ump->um_fs;
    165 		if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
    166 			flags = WRITECLOSE;
    167 			if (mp->mnt_flag & MNT_FORCE)
    168 				flags |= FORCECLOSE;
    169 			if (vfs_busy(mp))
    170 				return (EBUSY);
    171 			error = ffs_flushfiles(mp, flags, p);
    172 			if (error == 0 &&
    173 			    ffs_cgupdate(ump, MNT_WAIT) == 0 &&
    174 			    fs->fs_clean & FS_WASCLEAN) {
    175 				fs->fs_clean = FS_ISCLEAN;
    176 				(void) ffs_sbupdate(ump, MNT_WAIT);
    177 			}
    178 			vfs_unbusy(mp);
    179 			if (error)
    180 				return (error);
    181 			fs->fs_ronly = 1;
    182 		}
    183 		if (mp->mnt_flag & MNT_RELOAD) {
    184 			error = ffs_reload(mp, ndp->ni_cnd.cn_cred, p);
    185 			if (error)
    186 				return (error);
    187 		}
    188 		if (fs->fs_ronly && (mp->mnt_flag & MNT_WANTRDWR)) {
    189 			/*
    190 			 * If upgrade to read-write by non-root, then verify
    191 			 * that user has necessary permissions on the device.
    192 			 */
    193 			if (p->p_ucred->cr_uid != 0) {
    194 				devvp = ump->um_devvp;
    195 				VOP_LOCK(devvp);
    196 				if (error = VOP_ACCESS(devvp, VREAD | VWRITE,
    197 				    p->p_ucred, p)) {
    198 					VOP_UNLOCK(devvp);
    199 					return (error);
    200 				}
    201 				VOP_UNLOCK(devvp);
    202 			}
    203 			fs->fs_ronly = 0;
    204 			fs->fs_clean <<= 1;
    205 			fs->fs_fmod = 1;
    206 		}
    207 		if (args.fspec == 0) {
    208 			/*
    209 			 * Process export requests.
    210 			 */
    211 			return (vfs_export(mp, &ump->um_export, &args.export));
    212 		}
    213 	}
    214 	/*
    215 	 * Not an update, or updating the name: look up the name
    216 	 * and verify that it refers to a sensible block device.
    217 	 */
    218 	NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, p);
    219 	if (error = namei(ndp))
    220 		return (error);
    221 	devvp = ndp->ni_vp;
    222 
    223 	if (devvp->v_type != VBLK) {
    224 		vrele(devvp);
    225 		return (ENOTBLK);
    226 	}
    227 	if (major(devvp->v_rdev) >= nblkdev) {
    228 		vrele(devvp);
    229 		return (ENXIO);
    230 	}
    231 	/*
    232 	 * If mount by non-root, then verify that user has necessary
    233 	 * permissions on the device.
    234 	 */
    235 	if (p->p_ucred->cr_uid != 0) {
    236 		accessmode = VREAD;
    237 		if ((mp->mnt_flag & MNT_RDONLY) == 0)
    238 			accessmode |= VWRITE;
    239 		VOP_LOCK(devvp);
    240 		if (error = VOP_ACCESS(devvp, accessmode, p->p_ucred, p)) {
    241 			vput(devvp);
    242 			return (error);
    243 		}
    244 		VOP_UNLOCK(devvp);
    245 	}
    246 	if ((mp->mnt_flag & MNT_UPDATE) == 0)
    247 		error = ffs_mountfs(devvp, mp, p);
    248 	else {
    249 		if (devvp != ump->um_devvp)
    250 			error = EINVAL;	/* needs translation */
    251 		else
    252 			vrele(devvp);
    253 	}
    254 	if (error) {
    255 		vrele(devvp);
    256 		return (error);
    257 	}
    258 	ump = VFSTOUFS(mp);
    259 	fs = ump->um_fs;
    260 	(void) copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size);
    261 	bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size);
    262 	bcopy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname, MNAMELEN);
    263 	(void) copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
    264 	    &size);
    265 	bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
    266 	if (fs->fs_fmod != 0) {	/* XXX */
    267 		fs->fs_fmod = 0;
    268 		if (fs->fs_clean & FS_WASCLEAN)
    269 			fs->fs_time = time.tv_sec;
    270 		else
    271 			printf("%s: file system not clean; please fsck(8)\n",
    272 			    mp->mnt_stat.f_mntfromname);
    273 		(void) ffs_cgupdate(ump, MNT_WAIT);
    274 	}
    275 	return (0);
    276 }
    277 
    278 /*
    279  * Reload all incore data for a filesystem (used after running fsck on
    280  * the root filesystem and finding things to fix). The filesystem must
    281  * be mounted read-only.
    282  *
    283  * Things to do to update the mount:
    284  *	1) invalidate all cached meta-data.
    285  *	2) re-read superblock from disk.
    286  *	3) re-read summary information from disk.
    287  *	4) invalidate all inactive vnodes.
    288  *	5) invalidate all cached file data.
    289  *	6) re-read inode data for all active vnodes.
    290  */
    291 ffs_reload(mountp, cred, p)
    292 	register struct mount *mountp;
    293 	struct ucred *cred;
    294 	struct proc *p;
    295 {
    296 	register struct vnode *vp, *nvp, *devvp;
    297 	struct inode *ip;
    298 	struct csum *space;
    299 	struct buf *bp;
    300 	struct fs *fs;
    301 	struct partinfo dpart;
    302 	int i, blks, size, error;
    303 
    304 	if ((mountp->mnt_flag & MNT_RDONLY) == 0)
    305 		return (EINVAL);
    306 	/*
    307 	 * Step 1: invalidate all cached meta-data.
    308 	 */
    309 	devvp = VFSTOUFS(mountp)->um_devvp;
    310 	if (vinvalbuf(devvp, 0, cred, p, 0, 0))
    311 		panic("ffs_reload: dirty1");
    312 	/*
    313 	 * Step 2: re-read superblock from disk.
    314 	 */
    315 	if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, p) != 0)
    316 		size = DEV_BSIZE;
    317 	else
    318 		size = dpart.disklab->d_secsize;
    319 	if (error = bread(devvp, (daddr_t)(SBOFF / size), SBSIZE, NOCRED, &bp))
    320 		return (error);
    321 	fs = (struct fs *)bp->b_data;
    322 	if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE ||
    323 	    fs->fs_bsize < sizeof(struct fs)) {
    324 		brelse(bp);
    325 		return (EIO);		/* XXX needs translation */
    326 	}
    327 	fs = VFSTOUFS(mountp)->um_fs;
    328 	bcopy(&fs->fs_csp[0], &((struct fs *)bp->b_data)->fs_csp[0],
    329 	    sizeof(fs->fs_csp));
    330 	bcopy(bp->b_data, fs, (u_int)fs->fs_sbsize);
    331 	if (fs->fs_sbsize < SBSIZE)
    332 		bp->b_flags |= B_INVAL;
    333 	brelse(bp);
    334 	mountp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
    335 	ffs_oldfscompat(fs);
    336 	/*
    337 	 * Step 3: re-read summary information from disk.
    338 	 */
    339 	blks = howmany(fs->fs_cssize, fs->fs_fsize);
    340 	space = fs->fs_csp[0];
    341 	for (i = 0; i < blks; i += fs->fs_frag) {
    342 		size = fs->fs_bsize;
    343 		if (i + fs->fs_frag > blks)
    344 			size = (blks - i) * fs->fs_fsize;
    345 		if (error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
    346 		    NOCRED, &bp))
    347 			return (error);
    348 		bcopy(bp->b_data, fs->fs_csp[fragstoblks(fs, i)], (u_int)size);
    349 		brelse(bp);
    350 	}
    351 loop:
    352 	for (vp = mountp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) {
    353 		nvp = vp->v_mntvnodes.le_next;
    354 		/*
    355 		 * Step 4: invalidate all inactive vnodes.
    356 		 */
    357 		if (vp->v_usecount == 0) {
    358 			vgone(vp);
    359 			continue;
    360 		}
    361 		/*
    362 		 * Step 5: invalidate all cached file data.
    363 		 */
    364 		if (vget(vp, 1))
    365 			goto loop;
    366 		if (vinvalbuf(vp, 0, cred, p, 0, 0))
    367 			panic("ffs_reload: dirty2");
    368 		/*
    369 		 * Step 6: re-read inode data for all active vnodes.
    370 		 */
    371 		ip = VTOI(vp);
    372 		if (error =
    373 		    bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
    374 		    (int)fs->fs_bsize, NOCRED, &bp)) {
    375 			vput(vp);
    376 			return (error);
    377 		}
    378 		ip->i_din = *((struct dinode *)bp->b_data +
    379 		    ino_to_fsbo(fs, ip->i_number));
    380 		brelse(bp);
    381 		vput(vp);
    382 		if (vp->v_mount != mountp)
    383 			goto loop;
    384 	}
    385 	return (0);
    386 }
    387 
    388 /*
    389  * Common code for mount and mountroot
    390  */
    391 int
    392 ffs_mountfs(devvp, mp, p)
    393 	register struct vnode *devvp;
    394 	struct mount *mp;
    395 	struct proc *p;
    396 {
    397 	register struct ufsmount *ump;
    398 	struct buf *bp;
    399 	register struct fs *fs;
    400 	dev_t dev;
    401 	struct partinfo dpart;
    402 	caddr_t base, space;
    403 	int blks;
    404 	int error, i, size, ronly;
    405 	int32_t *lp;
    406 	struct ucred *cred;
    407 	extern struct vnode *rootvp;
    408 	u_int64_t maxfilesize;					/* XXX */
    409 
    410 	dev = devvp->v_rdev;
    411 	cred = p ? p->p_ucred : NOCRED;
    412 	/*
    413 	 * Disallow multiple mounts of the same device.
    414 	 * Disallow mounting of a device that is currently in use
    415 	 * (except for root, which might share swap device for miniroot).
    416 	 * Flush out any old buffers remaining from a previous use.
    417 	 */
    418 	if (error = vfs_mountedon(devvp))
    419 		return (error);
    420 	if (vcount(devvp) > 1 && devvp != rootvp)
    421 		return (EBUSY);
    422 	if (error = vinvalbuf(devvp, V_SAVE, cred, p, 0, 0))
    423 		return (error);
    424 
    425 	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
    426 	if (error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p))
    427 		return (error);
    428 	if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, cred, p) != 0)
    429 		size = DEV_BSIZE;
    430 	else
    431 		size = dpart.disklab->d_secsize;
    432 
    433 	bp = NULL;
    434 	ump = NULL;
    435 	if (error = bread(devvp, (daddr_t)(SBOFF / size), SBSIZE, cred, &bp))
    436 		goto out;
    437 	fs = (struct fs *)bp->b_data;
    438 	if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE ||
    439 	    fs->fs_bsize < sizeof(struct fs)) {
    440 		error = EINVAL;		/* XXX needs translation */
    441 		goto out;
    442 	}
    443 	/* XXX updating 4.2 FFS superblocks trashes rotational layout tables */
    444 	if (fs->fs_postblformat == FS_42POSTBLFMT && !ronly) {
    445 		error = EROFS;		/* XXX what should be returned? */
    446 		goto out;
    447 	}
    448 	ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
    449 	bzero((caddr_t)ump, sizeof *ump);
    450 	ump->um_fs = malloc((u_long)fs->fs_sbsize, M_UFSMNT,
    451 	    M_WAITOK);
    452 	bcopy(bp->b_data, ump->um_fs, (u_int)fs->fs_sbsize);
    453 	if (fs->fs_sbsize < SBSIZE)
    454 		bp->b_flags |= B_INVAL;
    455 	brelse(bp);
    456 	bp = NULL;
    457 	fs = ump->um_fs;
    458 	fs->fs_ronly = ronly;
    459 	if (ronly == 0) {
    460 		fs->fs_clean <<= 1;
    461 		fs->fs_fmod = 1;
    462 	}
    463 	size = fs->fs_cssize;
    464 	blks = howmany(size, fs->fs_fsize);
    465 	if (fs->fs_contigsumsize > 0)
    466 		size += fs->fs_ncg * sizeof(int32_t);
    467 	base = space = malloc((u_long)size, M_UFSMNT, M_WAITOK);
    468 	for (i = 0; i < blks; i += fs->fs_frag) {
    469 		size = fs->fs_bsize;
    470 		if (i + fs->fs_frag > blks)
    471 			size = (blks - i) * fs->fs_fsize;
    472 		if (error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
    473 		    cred, &bp)) {
    474 			free(base, M_UFSMNT);
    475 			goto out;
    476 		}
    477 		bcopy(bp->b_data, space, (u_int)size);
    478 		fs->fs_csp[fragstoblks(fs, i)] = (struct csum *)space;
    479 		space += size;
    480 		brelse(bp);
    481 		bp = NULL;
    482 	}
    483 	if (fs->fs_contigsumsize > 0) {
    484 		fs->fs_maxcluster = lp = (int32_t *)space;
    485 		for (i = 0; i < fs->fs_ncg; i++)
    486 			*lp++ = fs->fs_contigsumsize;
    487 	}
    488 	mp->mnt_data = (qaddr_t)ump;
    489 	mp->mnt_stat.f_fsid.val[0] = (long)dev;
    490 	mp->mnt_stat.f_fsid.val[1] = makefstype(MOUNT_UFS);
    491 	mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
    492 	mp->mnt_flag |= MNT_LOCAL;
    493 	ump->um_mountp = mp;
    494 	ump->um_dev = dev;
    495 	ump->um_devvp = devvp;
    496 	ump->um_nindir = fs->fs_nindir;
    497 	ump->um_bptrtodb = fs->fs_fsbtodb;
    498 	ump->um_seqinc = fs->fs_frag;
    499 	for (i = 0; i < MAXQUOTAS; i++)
    500 		ump->um_quotas[i] = NULLVP;
    501 	devvp->v_specflags |= SI_MOUNTEDON;
    502 	ffs_oldfscompat(fs);
    503 	ump->um_savedmaxfilesize = fs->fs_maxfilesize;		/* XXX */
    504 	maxfilesize = (u_int64_t)0x80000000 * fs->fs_bsize - 1;	/* XXX */
    505 	if (fs->fs_maxfilesize > maxfilesize)			/* XXX */
    506 		fs->fs_maxfilesize = maxfilesize;		/* XXX */
    507 	return (0);
    508 out:
    509 	if (bp)
    510 		brelse(bp);
    511 	(void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, cred, p);
    512 	if (ump) {
    513 		free(ump->um_fs, M_UFSMNT);
    514 		free(ump, M_UFSMNT);
    515 		mp->mnt_data = (qaddr_t)0;
    516 	}
    517 	return (error);
    518 }
    519 
    520 /*
    521  * Sanity checks for old file systems.
    522  *
    523  * XXX - goes away some day.
    524  */
    525 ffs_oldfscompat(fs)
    526 	struct fs *fs;
    527 {
    528 	int i;
    529 
    530 	fs->fs_npsect = max(fs->fs_npsect, fs->fs_nsect);	/* XXX */
    531 	fs->fs_interleave = max(fs->fs_interleave, 1);		/* XXX */
    532 	if (fs->fs_postblformat == FS_42POSTBLFMT)		/* XXX */
    533 		fs->fs_nrpos = 8;				/* XXX */
    534 	if (fs->fs_inodefmt < FS_44INODEFMT) {			/* XXX */
    535 		u_int64_t sizepb = fs->fs_bsize;		/* XXX */
    536 								/* XXX */
    537 		fs->fs_maxfilesize = fs->fs_bsize * NDADDR - 1;	/* XXX */
    538 		for (i = 0; i < NIADDR; i++) {			/* XXX */
    539 			sizepb *= NINDIR(fs);			/* XXX */
    540 			fs->fs_maxfilesize += sizepb;		/* XXX */
    541 		}						/* XXX */
    542 		fs->fs_qbmask = ~fs->fs_bmask;			/* XXX */
    543 		fs->fs_qfmask = ~fs->fs_fmask;			/* XXX */
    544 	}							/* XXX */
    545 	return (0);
    546 }
    547 
    548 /*
    549  * unmount system call
    550  */
    551 int
    552 ffs_unmount(mp, mntflags, p)
    553 	struct mount *mp;
    554 	int mntflags;
    555 	struct proc *p;
    556 {
    557 	register struct ufsmount *ump;
    558 	register struct fs *fs;
    559 	int error, flags;
    560 
    561 	flags = 0;
    562 	if (mntflags & MNT_FORCE)
    563 		flags |= FORCECLOSE;
    564 	if (error = ffs_flushfiles(mp, flags, p))
    565 		return (error);
    566 	ump = VFSTOUFS(mp);
    567 	fs = ump->um_fs;
    568 	if (fs->fs_ronly == 0 &&
    569 	    ffs_cgupdate(ump, MNT_WAIT) == 0 &&
    570 	    fs->fs_clean & FS_WASCLEAN) {
    571 		fs->fs_clean = FS_ISCLEAN;
    572 		(void) ffs_sbupdate(ump, MNT_WAIT);
    573 	}
    574 	ump->um_devvp->v_specflags &= ~SI_MOUNTEDON;
    575 	error = VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD|FWRITE,
    576 		NOCRED, p);
    577 	vrele(ump->um_devvp);
    578 	free(fs->fs_csp[0], M_UFSMNT);
    579 	free(fs, M_UFSMNT);
    580 	free(ump, M_UFSMNT);
    581 	mp->mnt_data = (qaddr_t)0;
    582 	mp->mnt_flag &= ~MNT_LOCAL;
    583 	return (error);
    584 }
    585 
    586 /*
    587  * Flush out all the files in a filesystem.
    588  */
    589 ffs_flushfiles(mp, flags, p)
    590 	register struct mount *mp;
    591 	int flags;
    592 	struct proc *p;
    593 {
    594 	extern int doforce;
    595 	register struct ufsmount *ump;
    596 	int i, error;
    597 
    598 	if (!doforce)
    599 		flags &= ~FORCECLOSE;
    600 	ump = VFSTOUFS(mp);
    601 #ifdef QUOTA
    602 	if (mp->mnt_flag & MNT_QUOTA) {
    603 		if (error = vflush(mp, NULLVP, SKIPSYSTEM|flags))
    604 			return (error);
    605 		for (i = 0; i < MAXQUOTAS; i++) {
    606 			if (ump->um_quotas[i] == NULLVP)
    607 				continue;
    608 			quotaoff(p, mp, i);
    609 		}
    610 		/*
    611 		 * Here we fall through to vflush again to ensure
    612 		 * that we have gotten rid of all the system vnodes.
    613 		 */
    614 	}
    615 #endif
    616 	error = vflush(mp, NULLVP, flags);
    617 	return (error);
    618 }
    619 
    620 /*
    621  * Get file system statistics.
    622  */
    623 int
    624 ffs_statfs(mp, sbp, p)
    625 	struct mount *mp;
    626 	register struct statfs *sbp;
    627 	struct proc *p;
    628 {
    629 	register struct ufsmount *ump;
    630 	register struct fs *fs;
    631 
    632 	ump = VFSTOUFS(mp);
    633 	fs = ump->um_fs;
    634 	if (fs->fs_magic != FS_MAGIC)
    635 		panic("ffs_statfs");
    636 #ifdef COMPAT_09
    637 	sbp->f_type = 1;
    638 #else
    639 	sbp->f_type = 0;
    640 #endif
    641 	sbp->f_bsize = fs->fs_fsize;
    642 	sbp->f_iosize = fs->fs_bsize;
    643 	sbp->f_blocks = fs->fs_dsize;
    644 	sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag +
    645 		fs->fs_cstotal.cs_nffree;
    646 	sbp->f_bavail = (fs->fs_dsize * (100 - fs->fs_minfree) / 100) -
    647 		(fs->fs_dsize - sbp->f_bfree);
    648 	sbp->f_files =  fs->fs_ncg * fs->fs_ipg - ROOTINO;
    649 	sbp->f_ffree = fs->fs_cstotal.cs_nifree;
    650 	if (sbp != &mp->mnt_stat) {
    651 		bcopy(mp->mnt_stat.f_mntonname, sbp->f_mntonname, MNAMELEN);
    652 		bcopy(mp->mnt_stat.f_mntfromname, sbp->f_mntfromname, MNAMELEN);
    653 	}
    654 	strncpy(sbp->f_fstypename, mp->mnt_op->vfs_name, MFSNAMELEN);
    655 	sbp->f_fstypename[MFSNAMELEN] = '\0';
    656 	return (0);
    657 }
    658 
    659 /*
    660  * Go through the disk queues to initiate sandbagged IO;
    661  * go through the inodes to write those that have been modified;
    662  * initiate the writing of the super block if it has been modified.
    663  *
    664  * Note: we are always called with the filesystem marked `MPBUSY'.
    665  */
    666 int
    667 ffs_sync(mp, waitfor, cred, p)
    668 	struct mount *mp;
    669 	int waitfor;
    670 	struct ucred *cred;
    671 	struct proc *p;
    672 {
    673 	register struct vnode *vp;
    674 	register struct inode *ip;
    675 	register struct ufsmount *ump = VFSTOUFS(mp);
    676 	register struct fs *fs;
    677 	int error, allerror = 0;
    678 
    679 	fs = ump->um_fs;
    680 	/*
    681 	 * Write back modified superblock.
    682 	 * Consistency check that the superblock
    683 	 * is still in the buffer cache.
    684 	 */
    685 	if (fs->fs_fmod != 0) {
    686 		if (fs->fs_ronly != 0) {		/* XXX */
    687 			printf("fs = %s\n", fs->fs_fsmnt);
    688 			panic("update: rofs mod");
    689 		}
    690 		fs->fs_fmod = 0;
    691 		fs->fs_time = time.tv_sec;
    692 		allerror = ffs_cgupdate(ump, waitfor);
    693 	}
    694 	/*
    695 	 * Write back each (modified) inode.
    696 	 */
    697 loop:
    698 	for (vp = mp->mnt_vnodelist.lh_first;
    699 	     vp != NULL;
    700 	     vp = vp->v_mntvnodes.le_next) {
    701 		/*
    702 		 * If the vnode that we are about to sync is no longer
    703 		 * associated with this mount point, start over.
    704 		 */
    705 		if (vp->v_mount != mp)
    706 			goto loop;
    707 		if (VOP_ISLOCKED(vp))
    708 			continue;
    709 		ip = VTOI(vp);
    710 		if ((ip->i_flag &
    711 		    (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
    712 		    vp->v_dirtyblkhd.lh_first == NULL)
    713 			continue;
    714 		if (vget(vp, 1))
    715 			goto loop;
    716 		if (error = VOP_FSYNC(vp, cred, waitfor, p))
    717 			allerror = error;
    718 		vput(vp);
    719 	}
    720 	/*
    721 	 * Force stale file system control information to be flushed.
    722 	 */
    723 	if (error = VOP_FSYNC(ump->um_devvp, cred, waitfor, p))
    724 		allerror = error;
    725 #ifdef QUOTA
    726 	qsync(mp);
    727 #endif
    728 	return (allerror);
    729 }
    730 
    731 /*
    732  * Look up a FFS dinode number to find its incore vnode, otherwise read it
    733  * in from disk.  If it is in core, wait for the lock bit to clear, then
    734  * return the inode locked.  Detection and handling of mount points must be
    735  * done by the calling routine.
    736  */
    737 int
    738 ffs_vget(mp, ino, vpp)
    739 	struct mount *mp;
    740 	ino_t ino;
    741 	struct vnode **vpp;
    742 {
    743 	register struct fs *fs;
    744 	register struct inode *ip;
    745 	struct ufsmount *ump;
    746 	struct buf *bp;
    747 	struct vnode *vp;
    748 	dev_t dev;
    749 	int i, type, error;
    750 
    751 	ump = VFSTOUFS(mp);
    752 	dev = ump->um_dev;
    753 	if ((*vpp = ufs_ihashget(dev, ino)) != NULL)
    754 		return (0);
    755 
    756 	/* Allocate a new vnode/inode. */
    757 	if (error = getnewvnode(VT_UFS, mp, ffs_vnodeop_p, &vp)) {
    758 		*vpp = NULL;
    759 		return (error);
    760 	}
    761 	type = ump->um_devvp->v_tag == VT_MFS ? M_MFSNODE : M_FFSNODE; /* XXX */
    762 	MALLOC(ip, struct inode *, sizeof(struct inode), type, M_WAITOK);
    763 	bzero((caddr_t)ip, sizeof(struct inode));
    764 	vp->v_data = ip;
    765 	ip->i_vnode = vp;
    766 	ip->i_fs = fs = ump->um_fs;
    767 	ip->i_dev = dev;
    768 	ip->i_number = ino;
    769 #ifdef QUOTA
    770 	for (i = 0; i < MAXQUOTAS; i++)
    771 		ip->i_dquot[i] = NODQUOT;
    772 #endif
    773 	/*
    774 	 * Put it onto its hash chain and lock it so that other requests for
    775 	 * this inode will block if they arrive while we are sleeping waiting
    776 	 * for old data structures to be purged or for the contents of the
    777 	 * disk portion of this inode to be read.
    778 	 */
    779 	ufs_ihashins(ip);
    780 
    781 	/* Read in the disk contents for the inode, copy into the inode. */
    782 	if (error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
    783 	    (int)fs->fs_bsize, NOCRED, &bp)) {
    784 		/*
    785 		 * The inode does not contain anything useful, so it would
    786 		 * be misleading to leave it on its hash chain. With mode
    787 		 * still zero, it will be unlinked and returned to the free
    788 		 * list by vput().
    789 		 */
    790 		vput(vp);
    791 		brelse(bp);
    792 		*vpp = NULL;
    793 		return (error);
    794 	}
    795 	ip->i_din = *((struct dinode *)bp->b_data + ino_to_fsbo(fs, ino));
    796 	brelse(bp);
    797 
    798 	/*
    799 	 * Initialize the vnode from the inode, check for aliases.
    800 	 * Note that the underlying vnode may have changed.
    801 	 */
    802 	if (error = ufs_vinit(mp, ffs_specop_p, FFS_FIFOOPS, &vp)) {
    803 		vput(vp);
    804 		*vpp = NULL;
    805 		return (error);
    806 	}
    807 	/*
    808 	 * Finish inode initialization now that aliasing has been resolved.
    809 	 */
    810 	ip->i_devvp = ump->um_devvp;
    811 	VREF(ip->i_devvp);
    812 	/*
    813 	 * Set up a generation number for this inode if it does not
    814 	 * already have one. This should only happen on old filesystems.
    815 	 */
    816 	if (ip->i_gen == 0) {
    817 		if (++nextgennumber < (u_long)time.tv_sec)
    818 			nextgennumber = time.tv_sec;
    819 		ip->i_gen = nextgennumber;
    820 		if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
    821 			ip->i_flag |= IN_MODIFIED;
    822 	}
    823 	/*
    824 	 * Ensure that uid and gid are correct. This is a temporary
    825 	 * fix until fsck has been changed to do the update.
    826 	 */
    827 	if (fs->fs_inodefmt < FS_44INODEFMT) {		/* XXX */
    828 		ip->i_uid = ip->i_din.di_ouid;		/* XXX */
    829 		ip->i_gid = ip->i_din.di_ogid;		/* XXX */
    830 	}						/* XXX */
    831 
    832 	*vpp = vp;
    833 	return (0);
    834 }
    835 
    836 /*
    837  * File handle to vnode
    838  *
    839  * Have to be really careful about stale file handles:
    840  * - check that the inode number is valid
    841  * - call ffs_vget() to get the locked inode
    842  * - check for an unallocated inode (i_mode == 0)
    843  * - check that the given client host has export rights and return
    844  *   those rights via. exflagsp and credanonp
    845  */
    846 int
    847 ffs_fhtovp(mp, fhp, nam, vpp, exflagsp, credanonp)
    848 	register struct mount *mp;
    849 	struct fid *fhp;
    850 	struct mbuf *nam;
    851 	struct vnode **vpp;
    852 	int *exflagsp;
    853 	struct ucred **credanonp;
    854 {
    855 	register struct ufid *ufhp;
    856 	struct fs *fs;
    857 
    858 	ufhp = (struct ufid *)fhp;
    859 	fs = VFSTOUFS(mp)->um_fs;
    860 	if (ufhp->ufid_ino < ROOTINO ||
    861 	    ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg)
    862 		return (ESTALE);
    863 	return (ufs_check_export(mp, ufhp, nam, vpp, exflagsp, credanonp));
    864 }
    865 
    866 /*
    867  * Vnode pointer to File handle
    868  */
    869 /* ARGSUSED */
    870 ffs_vptofh(vp, fhp)
    871 	struct vnode *vp;
    872 	struct fid *fhp;
    873 {
    874 	register struct inode *ip;
    875 	register struct ufid *ufhp;
    876 
    877 	ip = VTOI(vp);
    878 	ufhp = (struct ufid *)fhp;
    879 	ufhp->ufid_len = sizeof(struct ufid);
    880 	ufhp->ufid_ino = ip->i_number;
    881 	ufhp->ufid_gen = ip->i_gen;
    882 	return (0);
    883 }
    884 
    885 /*
    886  * Write a superblock and associated information back to disk.
    887  */
    888 int
    889 ffs_sbupdate(mp, waitfor)
    890 	struct ufsmount *mp;
    891 	int waitfor;
    892 {
    893 	register struct fs *dfs, *fs = mp->um_fs;
    894 	register struct buf *bp;
    895 	int i, error = 0;
    896 
    897 	bp = getblk(mp->um_devvp, SBOFF >> (fs->fs_fshift - fs->fs_fsbtodb),
    898 	    (int)fs->fs_sbsize, 0, 0);
    899 	bcopy((caddr_t)fs, bp->b_data, (u_int)fs->fs_sbsize);
    900 	/* Restore compatibility to old file systems.		   XXX */
    901 	dfs = (struct fs *)bp->b_data;				/* XXX */
    902 	if (fs->fs_postblformat == FS_42POSTBLFMT)		/* XXX */
    903 		dfs->fs_nrpos = -1;				/* XXX */
    904 	if (fs->fs_inodefmt < FS_44INODEFMT) {			/* XXX */
    905 		int32_t *lp, tmp;				/* XXX */
    906 								/* XXX */
    907 		lp = (int32_t *)&dfs->fs_qbmask;		/* XXX */
    908 		tmp = lp[4];					/* XXX */
    909 		for (i = 4; i > 0; i--)				/* XXX */
    910 			lp[i] = lp[i-1];			/* XXX */
    911 		lp[0] = tmp;					/* XXX */
    912 	}							/* XXX */
    913 	dfs->fs_maxfilesize = mp->um_savedmaxfilesize;		/* XXX */
    914 	if (waitfor == MNT_WAIT)
    915 		error = bwrite(bp);
    916 	else
    917 		bawrite(bp);
    918 	return (error);
    919 }
    920 
    921 int
    922 ffs_cgupdate(mp, waitfor)
    923 	struct ufsmount *mp;
    924 	int waitfor;
    925 {
    926 	register struct fs *fs = mp->um_fs;
    927 	register struct buf *bp;
    928 	int blks;
    929 	caddr_t space;
    930 	int i, size, error = 0, allerror = 0;
    931 
    932 	allerror = ffs_sbupdate(mp, waitfor);
    933 	blks = howmany(fs->fs_cssize, fs->fs_fsize);
    934 	space = (caddr_t)fs->fs_csp[0];
    935 	for (i = 0; i < blks; i += fs->fs_frag) {
    936 		size = fs->fs_bsize;
    937 		if (i + fs->fs_frag > blks)
    938 			size = (blks - i) * fs->fs_fsize;
    939 		bp = getblk(mp->um_devvp, fsbtodb(fs, fs->fs_csaddr + i),
    940 		    size, 0, 0);
    941 		bcopy(space, bp->b_data, (u_int)size);
    942 		space += size;
    943 		if (waitfor == MNT_WAIT)
    944 			error = bwrite(bp);
    945 		else
    946 			bawrite(bp);
    947 	}
    948 	if (!allerror && error)
    949 		allerror = error;
    950 	return (allerror);
    951 }
    952