Home | History | Annotate | Line # | Download | only in ffs
ffs_vfsops.c revision 1.31
      1 /*	$NetBSD: ffs_vfsops.c,v 1.31 1997/10/16 18:29:11 mjacob 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/device.h>
     48 #include <sys/mbuf.h>
     49 #include <sys/file.h>
     50 #include <sys/disklabel.h>
     51 #include <sys/ioctl.h>
     52 #include <sys/errno.h>
     53 #include <sys/malloc.h>
     54 #include <sys/lock.h>
     55 
     56 #include <miscfs/specfs/specdev.h>
     57 
     58 #include <ufs/ufs/quota.h>
     59 #include <ufs/ufs/ufsmount.h>
     60 #include <ufs/ufs/inode.h>
     61 #include <ufs/ufs/dir.h>
     62 #include <ufs/ufs/ufs_extern.h>
     63 
     64 #include <ufs/ffs/fs.h>
     65 #include <ufs/ffs/ffs_extern.h>
     66 
     67 extern struct lock ufs_hashlock;
     68 
     69 int ffs_sbupdate __P((struct ufsmount *, int));
     70 
     71 struct vfsops ffs_vfsops = {
     72 	MOUNT_FFS,
     73 	ffs_mount,
     74 	ufs_start,
     75 	ffs_unmount,
     76 	ufs_root,
     77 	ufs_quotactl,
     78 	ffs_statfs,
     79 	ffs_sync,
     80 	ffs_vget,
     81 	ffs_fhtovp,
     82 	ffs_vptofh,
     83 	ffs_init,
     84 	ffs_mountroot,
     85 };
     86 
     87 /*
     88  * Called by main() when ufs is going to be mounted as root.
     89  *
     90  * Name is updated by mount(8) after booting.
     91  */
     92 #define ROOTNAME	"root_device"
     93 
     94 int
     95 ffs_mountroot()
     96 {
     97 	extern struct vnode *rootvp;
     98 	register struct fs *fs;
     99 	register struct mount *mp;
    100 	struct proc *p = curproc;	/* XXX */
    101 	struct ufsmount *ump;
    102 	size_t size;
    103 	int error;
    104 
    105 	if (root_device->dv_class != DV_DISK)
    106 		return (ENODEV);
    107 
    108 	/*
    109 	 * Get vnodes for rootdev.
    110 	 */
    111 	if (bdevvp(rootdev, &rootvp))
    112 		panic("ffs_mountroot: can't setup bdevvp's");
    113 
    114 	mp = malloc((u_long)sizeof(struct mount), M_MOUNT, M_WAITOK);
    115 	bzero((char *)mp, (u_long)sizeof(struct mount));
    116 	mp->mnt_op = &ffs_vfsops;
    117 	mp->mnt_flag = MNT_RDONLY;
    118 	if ((error = ffs_mountfs(rootvp, mp, p)) != 0) {
    119 		free(mp, M_MOUNT);
    120 		return (error);
    121 	}
    122 	if ((error = vfs_lock(mp)) != 0) {
    123 		(void)ffs_unmount(mp, 0, p);
    124 		free(mp, M_MOUNT);
    125 		return (error);
    126 	}
    127 	CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
    128 	mp->mnt_vnodecovered = NULLVP;
    129 	ump = VFSTOUFS(mp);
    130 	fs = ump->um_fs;
    131 	bzero(fs->fs_fsmnt, sizeof(fs->fs_fsmnt));
    132 	fs->fs_fsmnt[0] = '/';
    133 	bcopy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname, MNAMELEN);
    134 	(void) copystr(ROOTNAME, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
    135 	    &size);
    136 	bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
    137 	(void)ffs_statfs(mp, &mp->mnt_stat, p);
    138 	vfs_unlock(mp);
    139 	inittodr(fs->fs_time);
    140 	return (0);
    141 }
    142 
    143 /*
    144  * VFS Operations.
    145  *
    146  * mount system call
    147  */
    148 int
    149 ffs_mount(mp, path, data, ndp, p)
    150 	register struct mount *mp;
    151 	const char *path;
    152 	void *data;
    153 	struct nameidata *ndp;
    154 	struct proc *p;
    155 {
    156 	struct vnode *devvp;
    157 	struct ufs_args args;
    158 	struct ufsmount *ump = NULL;
    159 	register struct fs *fs;
    160 	size_t size;
    161 	int error, flags;
    162 	mode_t accessmode;
    163 
    164 	error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args));
    165 	if (error)
    166 		return (error);
    167 	/*
    168 	 * If updating, check whether changing from read-only to
    169 	 * read/write; if there is no device name, that's all we do.
    170 	 */
    171 	if (mp->mnt_flag & MNT_UPDATE) {
    172 		ump = VFSTOUFS(mp);
    173 		fs = ump->um_fs;
    174 		if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
    175 			flags = WRITECLOSE;
    176 			if (mp->mnt_flag & MNT_FORCE)
    177 				flags |= FORCECLOSE;
    178 			if (vfs_busy(mp))
    179 				return (EBUSY);
    180 			error = ffs_flushfiles(mp, flags, p);
    181 			if (error == 0 &&
    182 			    ffs_cgupdate(ump, MNT_WAIT) == 0 &&
    183 			    fs->fs_clean & FS_WASCLEAN) {
    184 				fs->fs_clean = FS_ISCLEAN;
    185 				(void) ffs_sbupdate(ump, MNT_WAIT);
    186 			}
    187 			vfs_unbusy(mp);
    188 			if (error)
    189 				return (error);
    190 			fs->fs_ronly = 1;
    191 		}
    192 		if (mp->mnt_flag & MNT_RELOAD) {
    193 			error = ffs_reload(mp, ndp->ni_cnd.cn_cred, p);
    194 			if (error)
    195 				return (error);
    196 		}
    197 		if (fs->fs_ronly && (mp->mnt_flag & MNT_WANTRDWR)) {
    198 			/*
    199 			 * If upgrade to read-write by non-root, then verify
    200 			 * that user has necessary permissions on the device.
    201 			 */
    202 			if (p->p_ucred->cr_uid != 0) {
    203 				devvp = ump->um_devvp;
    204 				VOP_LOCK(devvp);
    205 				error = VOP_ACCESS(devvp, VREAD | VWRITE,
    206 						   p->p_ucred, p);
    207 				if (error) {
    208 					VOP_UNLOCK(devvp);
    209 					return (error);
    210 				}
    211 				VOP_UNLOCK(devvp);
    212 			}
    213 			fs->fs_ronly = 0;
    214 			fs->fs_clean <<= 1;
    215 			fs->fs_fmod = 1;
    216 		}
    217 		if (args.fspec == 0) {
    218 			/*
    219 			 * Process export requests.
    220 			 */
    221 			return (vfs_export(mp, &ump->um_export, &args.export));
    222 		}
    223 	}
    224 	/*
    225 	 * Not an update, or updating the name: look up the name
    226 	 * and verify that it refers to a sensible block device.
    227 	 */
    228 	NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, p);
    229 	if ((error = namei(ndp)) != 0)
    230 		return (error);
    231 	devvp = ndp->ni_vp;
    232 
    233 	if (devvp->v_type != VBLK) {
    234 		vrele(devvp);
    235 		return (ENOTBLK);
    236 	}
    237 	if (major(devvp->v_rdev) >= nblkdev) {
    238 		vrele(devvp);
    239 		return (ENXIO);
    240 	}
    241 	/*
    242 	 * If mount by non-root, then verify that user has necessary
    243 	 * permissions on the device.
    244 	 */
    245 	if (p->p_ucred->cr_uid != 0) {
    246 		accessmode = VREAD;
    247 		if ((mp->mnt_flag & MNT_RDONLY) == 0)
    248 			accessmode |= VWRITE;
    249 		VOP_LOCK(devvp);
    250 		error = VOP_ACCESS(devvp, accessmode, p->p_ucred, p);
    251 		if (error) {
    252 			vput(devvp);
    253 			return (error);
    254 		}
    255 		VOP_UNLOCK(devvp);
    256 	}
    257 	if ((mp->mnt_flag & MNT_UPDATE) == 0)
    258 		error = ffs_mountfs(devvp, mp, p);
    259 	else {
    260 		if (devvp != ump->um_devvp)
    261 			error = EINVAL;	/* needs translation */
    262 		else
    263 			vrele(devvp);
    264 	}
    265 	if (error) {
    266 		vrele(devvp);
    267 		return (error);
    268 	}
    269 	ump = VFSTOUFS(mp);
    270 	fs = ump->um_fs;
    271 	(void) copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size);
    272 	bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size);
    273 	bcopy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname, MNAMELEN);
    274 	(void) copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
    275 	    &size);
    276 	bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
    277 	if (fs->fs_fmod != 0) {	/* XXX */
    278 		fs->fs_fmod = 0;
    279 		if (fs->fs_clean & FS_WASCLEAN)
    280 			fs->fs_time = time.tv_sec;
    281 		else
    282 			printf("%s: file system not clean; please fsck(8)\n",
    283 			    mp->mnt_stat.f_mntfromname);
    284 		(void) ffs_cgupdate(ump, MNT_WAIT);
    285 	}
    286 	return (0);
    287 }
    288 
    289 /*
    290  * Reload all incore data for a filesystem (used after running fsck on
    291  * the root filesystem and finding things to fix). The filesystem must
    292  * be mounted read-only.
    293  *
    294  * Things to do to update the mount:
    295  *	1) invalidate all cached meta-data.
    296  *	2) re-read superblock from disk.
    297  *	3) re-read summary information from disk.
    298  *	4) invalidate all inactive vnodes.
    299  *	5) invalidate all cached file data.
    300  *	6) re-read inode data for all active vnodes.
    301  */
    302 int
    303 ffs_reload(mountp, cred, p)
    304 	register struct mount *mountp;
    305 	struct ucred *cred;
    306 	struct proc *p;
    307 {
    308 	register struct vnode *vp, *nvp, *devvp;
    309 	struct inode *ip;
    310 	struct csum *space;
    311 	struct buf *bp;
    312 	struct fs *fs, *newfs;
    313 	struct partinfo dpart;
    314 	int i, blks, size, error;
    315 	int32_t *lp;
    316 
    317 	if ((mountp->mnt_flag & MNT_RDONLY) == 0)
    318 		return (EINVAL);
    319 	/*
    320 	 * Step 1: invalidate all cached meta-data.
    321 	 */
    322 	devvp = VFSTOUFS(mountp)->um_devvp;
    323 	if (vinvalbuf(devvp, 0, cred, p, 0, 0))
    324 		panic("ffs_reload: dirty1");
    325 	/*
    326 	 * Step 2: re-read superblock from disk.
    327 	 */
    328 	if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, p) != 0)
    329 		size = DEV_BSIZE;
    330 	else
    331 		size = dpart.disklab->d_secsize;
    332 	error = bread(devvp, (daddr_t)(SBOFF / size), SBSIZE, NOCRED, &bp);
    333 	if (error)
    334 		return (error);
    335 	newfs = (struct fs *)bp->b_data;
    336 	if (newfs->fs_magic != FS_MAGIC || newfs->fs_bsize > MAXBSIZE ||
    337 	    newfs->fs_bsize < sizeof(struct fs)) {
    338 		brelse(bp);
    339 		return (EIO);		/* XXX needs translation */
    340 	}
    341 	fs = VFSTOUFS(mountp)->um_fs;
    342 	/*
    343 	 * Copy pointer fields back into superblock before copying in	XXX
    344 	 * new superblock. These should really be in the ufsmount.	XXX
    345 	 * Note that important parameters (eg fs_ncg) are unchanged.
    346 	 */
    347 	bcopy(&fs->fs_csp[0], &newfs->fs_csp[0], sizeof(fs->fs_csp));
    348 	newfs->fs_maxcluster = fs->fs_maxcluster;
    349 	bcopy(newfs, fs, (u_int)fs->fs_sbsize);
    350 	if (fs->fs_sbsize < SBSIZE)
    351 		bp->b_flags |= B_INVAL;
    352 	brelse(bp);
    353 	mountp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
    354 	ffs_oldfscompat(fs);
    355 	/*
    356 	 * Step 3: re-read summary information from disk.
    357 	 */
    358 	blks = howmany(fs->fs_cssize, fs->fs_fsize);
    359 	space = fs->fs_csp[0];
    360 	for (i = 0; i < blks; i += fs->fs_frag) {
    361 		size = fs->fs_bsize;
    362 		if (i + fs->fs_frag > blks)
    363 			size = (blks - i) * fs->fs_fsize;
    364 		error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
    365 			      NOCRED, &bp);
    366 		if (error)
    367 			return (error);
    368 		bcopy(bp->b_data, fs->fs_csp[fragstoblks(fs, i)], (u_int)size);
    369 		brelse(bp);
    370 	}
    371 	/*
    372 	 * We no longer know anything about clusters per cylinder group.
    373 	 */
    374 	if (fs->fs_contigsumsize > 0) {
    375 		lp = fs->fs_maxcluster;
    376 		for (i = 0; i < fs->fs_ncg; i++)
    377 			*lp++ = fs->fs_contigsumsize;
    378 	}
    379 
    380 loop:
    381 	for (vp = mountp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) {
    382 		nvp = vp->v_mntvnodes.le_next;
    383 		/*
    384 		 * Step 4: invalidate all inactive vnodes.
    385 		 */
    386 		if (vp->v_usecount == 0) {
    387 			vgone(vp);
    388 			continue;
    389 		}
    390 		/*
    391 		 * Step 5: invalidate all cached file data.
    392 		 */
    393 		if (vget(vp, 1))
    394 			goto loop;
    395 		if (vinvalbuf(vp, 0, cred, p, 0, 0))
    396 			panic("ffs_reload: dirty2");
    397 		/*
    398 		 * Step 6: re-read inode data for all active vnodes.
    399 		 */
    400 		ip = VTOI(vp);
    401 		error = bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
    402 			      (int)fs->fs_bsize, NOCRED, &bp);
    403 		if (error) {
    404 			vput(vp);
    405 			return (error);
    406 		}
    407 		ip->i_din.ffs_din = *((struct dinode *)bp->b_data +
    408 		    ino_to_fsbo(fs, ip->i_number));
    409 		brelse(bp);
    410 		vput(vp);
    411 		if (vp->v_mount != mountp)
    412 			goto loop;
    413 	}
    414 	return (0);
    415 }
    416 
    417 /*
    418  * Common code for mount and mountroot
    419  */
    420 int
    421 ffs_mountfs(devvp, mp, p)
    422 	register struct vnode *devvp;
    423 	struct mount *mp;
    424 	struct proc *p;
    425 {
    426 	register struct ufsmount *ump;
    427 	struct buf *bp;
    428 	register struct fs *fs;
    429 	dev_t dev;
    430 	struct partinfo dpart;
    431 	caddr_t base, space;
    432 	int blks;
    433 	int error, i, size, ronly;
    434 	int32_t *lp;
    435 	struct ucred *cred;
    436 	extern struct vnode *rootvp;
    437 	u_int64_t maxfilesize;					/* XXX */
    438 
    439 	dev = devvp->v_rdev;
    440 	cred = p ? p->p_ucred : NOCRED;
    441 	/*
    442 	 * Disallow multiple mounts of the same device.
    443 	 * Disallow mounting of a device that is currently in use
    444 	 * (except for root, which might share swap device for miniroot).
    445 	 * Flush out any old buffers remaining from a previous use.
    446 	 */
    447 	if ((error = vfs_mountedon(devvp)) != 0)
    448 		return (error);
    449 	if (vcount(devvp) > 1 && devvp != rootvp)
    450 		return (EBUSY);
    451 	if ((error = vinvalbuf(devvp, V_SAVE, cred, p, 0, 0)) != 0)
    452 		return (error);
    453 
    454 	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
    455 	error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p);
    456 	if (error)
    457 		return (error);
    458 	if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, cred, p) != 0)
    459 		size = DEV_BSIZE;
    460 	else
    461 		size = dpart.disklab->d_secsize;
    462 
    463 	bp = NULL;
    464 	ump = NULL;
    465 	error = bread(devvp, (daddr_t)(SBOFF / size), SBSIZE, cred, &bp);
    466 	if (error)
    467 		goto out;
    468 	fs = (struct fs *)bp->b_data;
    469 	if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE ||
    470 	    fs->fs_bsize < sizeof(struct fs)) {
    471 		error = EINVAL;		/* XXX needs translation */
    472 		goto out;
    473 	}
    474 	/* XXX updating 4.2 FFS superblocks trashes rotational layout tables */
    475 	if (fs->fs_postblformat == FS_42POSTBLFMT && !ronly) {
    476 		error = EROFS;		/* XXX what should be returned? */
    477 		goto out;
    478 	}
    479 	ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
    480 	bzero((caddr_t)ump, sizeof *ump);
    481 	ump->um_fs = malloc((u_long)fs->fs_sbsize, M_UFSMNT,
    482 	    M_WAITOK);
    483 	bcopy(bp->b_data, ump->um_fs, (u_int)fs->fs_sbsize);
    484 	if (fs->fs_sbsize < SBSIZE)
    485 		bp->b_flags |= B_INVAL;
    486 	brelse(bp);
    487 	bp = NULL;
    488 	fs = ump->um_fs;
    489 	fs->fs_ronly = ronly;
    490 	if (ronly == 0) {
    491 		fs->fs_clean <<= 1;
    492 		fs->fs_fmod = 1;
    493 	}
    494 	size = fs->fs_cssize;
    495 	blks = howmany(size, fs->fs_fsize);
    496 	if (fs->fs_contigsumsize > 0)
    497 		size += fs->fs_ncg * sizeof(int32_t);
    498 	base = space = malloc((u_long)size, M_UFSMNT, M_WAITOK);
    499 	for (i = 0; i < blks; i += fs->fs_frag) {
    500 		size = fs->fs_bsize;
    501 		if (i + fs->fs_frag > blks)
    502 			size = (blks - i) * fs->fs_fsize;
    503 		error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
    504 			      cred, &bp);
    505 		if (error) {
    506 			free(base, M_UFSMNT);
    507 			goto out;
    508 		}
    509 		bcopy(bp->b_data, space, (u_int)size);
    510 		fs->fs_csp[fragstoblks(fs, i)] = (struct csum *)space;
    511 		space += size;
    512 		brelse(bp);
    513 		bp = NULL;
    514 	}
    515 	if (fs->fs_contigsumsize > 0) {
    516 		fs->fs_maxcluster = lp = (int32_t *)space;
    517 		for (i = 0; i < fs->fs_ncg; i++)
    518 			*lp++ = fs->fs_contigsumsize;
    519 	}
    520 	mp->mnt_data = (qaddr_t)ump;
    521 	mp->mnt_stat.f_fsid.val[0] = (long)dev;
    522 	mp->mnt_stat.f_fsid.val[1] = makefstype(MOUNT_FFS);
    523 	mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
    524 	mp->mnt_flag |= MNT_LOCAL;
    525 	ump->um_mountp = mp;
    526 	ump->um_dev = dev;
    527 	ump->um_devvp = devvp;
    528 	ump->um_nindir = fs->fs_nindir;
    529 	ump->um_bptrtodb = fs->fs_fsbtodb;
    530 	ump->um_seqinc = fs->fs_frag;
    531 	for (i = 0; i < MAXQUOTAS; i++)
    532 		ump->um_quotas[i] = NULLVP;
    533 	devvp->v_specflags |= SI_MOUNTEDON;
    534 	ffs_oldfscompat(fs);
    535 	ump->um_savedmaxfilesize = fs->fs_maxfilesize;		/* XXX */
    536 	maxfilesize = (u_int64_t)0x80000000 * fs->fs_bsize - 1;	/* XXX */
    537 	if (fs->fs_maxfilesize > maxfilesize)			/* XXX */
    538 		fs->fs_maxfilesize = maxfilesize;		/* XXX */
    539 	return (0);
    540 out:
    541 	if (bp)
    542 		brelse(bp);
    543 	(void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, cred, p);
    544 	if (ump) {
    545 		free(ump->um_fs, M_UFSMNT);
    546 		free(ump, M_UFSMNT);
    547 		mp->mnt_data = (qaddr_t)0;
    548 	}
    549 	return (error);
    550 }
    551 
    552 /*
    553  * Sanity checks for old file systems.
    554  *
    555  * XXX - goes away some day.
    556  */
    557 int
    558 ffs_oldfscompat(fs)
    559 	struct fs *fs;
    560 {
    561 	int i;
    562 
    563 	fs->fs_npsect = max(fs->fs_npsect, fs->fs_nsect);	/* XXX */
    564 	fs->fs_interleave = max(fs->fs_interleave, 1);		/* XXX */
    565 	if (fs->fs_postblformat == FS_42POSTBLFMT)		/* XXX */
    566 		fs->fs_nrpos = 8;				/* XXX */
    567 	if (fs->fs_inodefmt < FS_44INODEFMT) {			/* XXX */
    568 		u_int64_t sizepb = fs->fs_bsize;		/* XXX */
    569 								/* XXX */
    570 		fs->fs_maxfilesize = fs->fs_bsize * NDADDR - 1;	/* XXX */
    571 		for (i = 0; i < NIADDR; i++) {			/* XXX */
    572 			sizepb *= NINDIR(fs);			/* XXX */
    573 			fs->fs_maxfilesize += sizepb;		/* XXX */
    574 		}						/* XXX */
    575 		fs->fs_qbmask = ~fs->fs_bmask;			/* XXX */
    576 		fs->fs_qfmask = ~fs->fs_fmask;			/* XXX */
    577 	}							/* XXX */
    578 	return (0);
    579 }
    580 
    581 /*
    582  * unmount system call
    583  */
    584 int
    585 ffs_unmount(mp, mntflags, p)
    586 	struct mount *mp;
    587 	int mntflags;
    588 	struct proc *p;
    589 {
    590 	register struct ufsmount *ump;
    591 	register struct fs *fs;
    592 	int error, flags;
    593 
    594 	flags = 0;
    595 	if (mntflags & MNT_FORCE)
    596 		flags |= FORCECLOSE;
    597 	if ((error = ffs_flushfiles(mp, flags, p)) != 0)
    598 		return (error);
    599 	ump = VFSTOUFS(mp);
    600 	fs = ump->um_fs;
    601 	if (fs->fs_ronly == 0 &&
    602 	    ffs_cgupdate(ump, MNT_WAIT) == 0 &&
    603 	    fs->fs_clean & FS_WASCLEAN) {
    604 		fs->fs_clean = FS_ISCLEAN;
    605 		(void) ffs_sbupdate(ump, MNT_WAIT);
    606 	}
    607 	ump->um_devvp->v_specflags &= ~SI_MOUNTEDON;
    608 	error = VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD|FWRITE,
    609 		NOCRED, p);
    610 	vrele(ump->um_devvp);
    611 	free(fs->fs_csp[0], M_UFSMNT);
    612 	free(fs, M_UFSMNT);
    613 	free(ump, M_UFSMNT);
    614 	mp->mnt_data = (qaddr_t)0;
    615 	mp->mnt_flag &= ~MNT_LOCAL;
    616 	return (error);
    617 }
    618 
    619 /*
    620  * Flush out all the files in a filesystem.
    621  */
    622 int
    623 ffs_flushfiles(mp, flags, p)
    624 	register struct mount *mp;
    625 	int flags;
    626 	struct proc *p;
    627 {
    628 	extern int doforce;
    629 	register struct ufsmount *ump;
    630 	int error;
    631 
    632 	if (!doforce)
    633 		flags &= ~FORCECLOSE;
    634 	ump = VFSTOUFS(mp);
    635 #ifdef QUOTA
    636 	if (mp->mnt_flag & MNT_QUOTA) {
    637 		int i;
    638 		if ((error = vflush(mp, NULLVP, SKIPSYSTEM|flags)) != 0)
    639 			return (error);
    640 		for (i = 0; i < MAXQUOTAS; i++) {
    641 			if (ump->um_quotas[i] == NULLVP)
    642 				continue;
    643 			quotaoff(p, mp, i);
    644 		}
    645 		/*
    646 		 * Here we fall through to vflush again to ensure
    647 		 * that we have gotten rid of all the system vnodes.
    648 		 */
    649 	}
    650 #endif
    651 	error = vflush(mp, NULLVP, flags);
    652 	return (error);
    653 }
    654 
    655 /*
    656  * Get file system statistics.
    657  */
    658 int
    659 ffs_statfs(mp, sbp, p)
    660 	struct mount *mp;
    661 	register struct statfs *sbp;
    662 	struct proc *p;
    663 {
    664 	register struct ufsmount *ump;
    665 	register struct fs *fs;
    666 
    667 	ump = VFSTOUFS(mp);
    668 	fs = ump->um_fs;
    669 	if (fs->fs_magic != FS_MAGIC)
    670 		panic("ffs_statfs");
    671 #ifdef COMPAT_09
    672 	sbp->f_type = 1;
    673 #else
    674 	sbp->f_type = 0;
    675 #endif
    676 	sbp->f_bsize = fs->fs_fsize;
    677 	sbp->f_iosize = fs->fs_bsize;
    678 	sbp->f_blocks = fs->fs_dsize;
    679 	sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag +
    680 		fs->fs_cstotal.cs_nffree;
    681 	sbp->f_bavail = (long) (((u_int64_t) fs->fs_dsize * (u_int64_t)
    682 		(100 - fs->fs_minfree) / (u_int64_t) 100) -
    683 		(u_int64_t) (fs->fs_dsize - sbp->f_bfree));
    684 	sbp->f_files =  fs->fs_ncg * fs->fs_ipg - ROOTINO;
    685 	sbp->f_ffree = fs->fs_cstotal.cs_nifree;
    686 	if (sbp != &mp->mnt_stat) {
    687 		bcopy(mp->mnt_stat.f_mntonname, sbp->f_mntonname, MNAMELEN);
    688 		bcopy(mp->mnt_stat.f_mntfromname, sbp->f_mntfromname, MNAMELEN);
    689 	}
    690 	strncpy(sbp->f_fstypename, mp->mnt_op->vfs_name, MFSNAMELEN);
    691 	return (0);
    692 }
    693 
    694 /*
    695  * Go through the disk queues to initiate sandbagged IO;
    696  * go through the inodes to write those that have been modified;
    697  * initiate the writing of the super block if it has been modified.
    698  *
    699  * Note: we are always called with the filesystem marked `MPBUSY'.
    700  */
    701 int
    702 ffs_sync(mp, waitfor, cred, p)
    703 	struct mount *mp;
    704 	int waitfor;
    705 	struct ucred *cred;
    706 	struct proc *p;
    707 {
    708 	register struct vnode *vp;
    709 	register struct inode *ip;
    710 	register struct ufsmount *ump = VFSTOUFS(mp);
    711 	register struct fs *fs;
    712 	int error, allerror = 0;
    713 
    714 	fs = ump->um_fs;
    715 	/*
    716 	 * Write back modified superblock.
    717 	 * Consistency check that the superblock
    718 	 * is still in the buffer cache.
    719 	 */
    720 	if (fs->fs_fmod != 0) {
    721 		if (fs->fs_ronly != 0) {		/* XXX */
    722 			printf("fs = %s\n", fs->fs_fsmnt);
    723 			panic("update: rofs mod");
    724 		}
    725 		fs->fs_fmod = 0;
    726 		fs->fs_time = time.tv_sec;
    727 		allerror = ffs_cgupdate(ump, waitfor);
    728 	}
    729 	/*
    730 	 * Write back each (modified) inode.
    731 	 */
    732 loop:
    733 	for (vp = mp->mnt_vnodelist.lh_first;
    734 	     vp != NULL;
    735 	     vp = vp->v_mntvnodes.le_next) {
    736 		/*
    737 		 * If the vnode that we are about to sync is no longer
    738 		 * associated with this mount point, start over.
    739 		 */
    740 		if (vp->v_mount != mp)
    741 			goto loop;
    742 		if (VOP_ISLOCKED(vp))
    743 			continue;
    744 		ip = VTOI(vp);
    745 		if ((ip->i_flag &
    746 		    (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
    747 		    vp->v_dirtyblkhd.lh_first == NULL)
    748 			continue;
    749 		if (vget(vp, 1))
    750 			goto loop;
    751 		if ((error = VOP_FSYNC(vp, cred, waitfor, p)) != 0)
    752 			allerror = error;
    753 		vput(vp);
    754 	}
    755 	/*
    756 	 * Force stale file system control information to be flushed.
    757 	 */
    758 	if ((error = VOP_FSYNC(ump->um_devvp, cred, waitfor, p)) != 0)
    759 		allerror = error;
    760 #ifdef QUOTA
    761 	qsync(mp);
    762 #endif
    763 	return (allerror);
    764 }
    765 
    766 /*
    767  * Look up a FFS dinode number to find its incore vnode, otherwise read it
    768  * in from disk.  If it is in core, wait for the lock bit to clear, then
    769  * return the inode locked.  Detection and handling of mount points must be
    770  * done by the calling routine.
    771  */
    772 int
    773 ffs_vget(mp, ino, vpp)
    774 	struct mount *mp;
    775 	ino_t ino;
    776 	struct vnode **vpp;
    777 {
    778 	register struct fs *fs;
    779 	register struct inode *ip;
    780 	struct ufsmount *ump;
    781 	struct buf *bp;
    782 	struct vnode *vp;
    783 	dev_t dev;
    784 	int type, error;
    785 
    786 	ump = VFSTOUFS(mp);
    787 	dev = ump->um_dev;
    788 	do {
    789 		if ((*vpp = ufs_ihashget(dev, ino)) != NULL)
    790 			return (0);
    791 	} while (lockmgr(&ufs_hashlock, LK_EXCLUSIVE|LK_SLEEPFAIL, 0, curproc));
    792 
    793 	/* Allocate a new vnode/inode. */
    794 	if ((error = getnewvnode(VT_UFS, mp, ffs_vnodeop_p, &vp)) != 0) {
    795 		*vpp = NULL;
    796 		lockmgr(&ufs_hashlock, LK_RELEASE, 0, curproc);
    797 		return (error);
    798 	}
    799 	type = ump->um_devvp->v_tag == VT_MFS ? M_MFSNODE : M_FFSNODE; /* XXX */
    800 	MALLOC(ip, struct inode *, sizeof(struct inode), type, M_WAITOK);
    801 	bzero((caddr_t)ip, sizeof(struct inode));
    802 	vp->v_data = ip;
    803 	ip->i_vnode = vp;
    804 	ip->i_fs = fs = ump->um_fs;
    805 	ip->i_dev = dev;
    806 	ip->i_number = ino;
    807 #ifdef QUOTA
    808 	{
    809 		int i;
    810 
    811 		for (i = 0; i < MAXQUOTAS; i++)
    812 			ip->i_dquot[i] = NODQUOT;
    813 	}
    814 #endif
    815 	/*
    816 	 * Put it onto its hash chain and lock it so that other requests for
    817 	 * this inode will block if they arrive while we are sleeping waiting
    818 	 * for old data structures to be purged or for the contents of the
    819 	 * disk portion of this inode to be read.
    820 	 */
    821 	ufs_ihashins(ip);
    822 	lockmgr(&ufs_hashlock, LK_RELEASE, 0, curproc);
    823 
    824 	/* Read in the disk contents for the inode, copy into the inode. */
    825 	error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
    826 		      (int)fs->fs_bsize, NOCRED, &bp);
    827 	if (error) {
    828 		/*
    829 		 * The inode does not contain anything useful, so it would
    830 		 * be misleading to leave it on its hash chain. With mode
    831 		 * still zero, it will be unlinked and returned to the free
    832 		 * list by vput().
    833 		 */
    834 		vput(vp);
    835 		brelse(bp);
    836 		*vpp = NULL;
    837 		return (error);
    838 	}
    839 	ip->i_din.ffs_din = *((struct dinode *)bp->b_data + ino_to_fsbo(fs, ino));
    840 	brelse(bp);
    841 
    842 	/*
    843 	 * Initialize the vnode from the inode, check for aliases.
    844 	 * Note that the underlying vnode may have changed.
    845 	 */
    846 	error = ufs_vinit(mp, ffs_specop_p, FFS_FIFOOPS, &vp);
    847 	if (error) {
    848 		vput(vp);
    849 		*vpp = NULL;
    850 		return (error);
    851 	}
    852 	/*
    853 	 * Finish inode initialization now that aliasing has been resolved.
    854 	 */
    855 	ip->i_devvp = ump->um_devvp;
    856 	VREF(ip->i_devvp);
    857 	/*
    858 	 * Ensure that uid and gid are correct. This is a temporary
    859 	 * fix until fsck has been changed to do the update.
    860 	 */
    861 	if (fs->fs_inodefmt < FS_44INODEFMT) {		/* XXX */
    862 		ip->i_ffs_uid = ip->i_din.ffs_din.di_ouid;		/* XXX */
    863 		ip->i_ffs_gid = ip->i_din.ffs_din.di_ogid;		/* XXX */
    864 	}						/* XXX */
    865 
    866 	*vpp = vp;
    867 	return (0);
    868 }
    869 
    870 /*
    871  * File handle to vnode
    872  *
    873  * Have to be really careful about stale file handles:
    874  * - check that the inode number is valid
    875  * - call ffs_vget() to get the locked inode
    876  * - check for an unallocated inode (i_mode == 0)
    877  * - check that the given client host has export rights and return
    878  *   those rights via. exflagsp and credanonp
    879  */
    880 int
    881 ffs_fhtovp(mp, fhp, nam, vpp, exflagsp, credanonp)
    882 	register struct mount *mp;
    883 	struct fid *fhp;
    884 	struct mbuf *nam;
    885 	struct vnode **vpp;
    886 	int *exflagsp;
    887 	struct ucred **credanonp;
    888 {
    889 	register struct ufid *ufhp;
    890 	struct fs *fs;
    891 
    892 	ufhp = (struct ufid *)fhp;
    893 	fs = VFSTOUFS(mp)->um_fs;
    894 	if (ufhp->ufid_ino < ROOTINO ||
    895 	    ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg)
    896 		return (ESTALE);
    897 	return (ufs_check_export(mp, ufhp, nam, vpp, exflagsp, credanonp));
    898 }
    899 
    900 /*
    901  * Vnode pointer to File handle
    902  */
    903 /* ARGSUSED */
    904 int
    905 ffs_vptofh(vp, fhp)
    906 	struct vnode *vp;
    907 	struct fid *fhp;
    908 {
    909 	register struct inode *ip;
    910 	register struct ufid *ufhp;
    911 
    912 	ip = VTOI(vp);
    913 	ufhp = (struct ufid *)fhp;
    914 	ufhp->ufid_len = sizeof(struct ufid);
    915 	ufhp->ufid_ino = ip->i_number;
    916 	ufhp->ufid_gen = ip->i_ffs_gen;
    917 	return (0);
    918 }
    919 
    920 /*
    921  * Write a superblock and associated information back to disk.
    922  */
    923 int
    924 ffs_sbupdate(mp, waitfor)
    925 	struct ufsmount *mp;
    926 	int waitfor;
    927 {
    928 	register struct fs *dfs, *fs = mp->um_fs;
    929 	register struct buf *bp;
    930 	int i, error = 0;
    931 
    932 	bp = getblk(mp->um_devvp, SBOFF >> (fs->fs_fshift - fs->fs_fsbtodb),
    933 	    (int)fs->fs_sbsize, 0, 0);
    934 	bcopy((caddr_t)fs, bp->b_data, (u_int)fs->fs_sbsize);
    935 	/* Restore compatibility to old file systems.		   XXX */
    936 	dfs = (struct fs *)bp->b_data;				/* XXX */
    937 	if (fs->fs_postblformat == FS_42POSTBLFMT)		/* XXX */
    938 		dfs->fs_nrpos = -1;				/* XXX */
    939 	if (fs->fs_inodefmt < FS_44INODEFMT) {			/* XXX */
    940 		int32_t *lp, tmp;				/* XXX */
    941 								/* XXX */
    942 		lp = (int32_t *)&dfs->fs_qbmask;		/* XXX */
    943 		tmp = lp[4];					/* XXX */
    944 		for (i = 4; i > 0; i--)				/* XXX */
    945 			lp[i] = lp[i-1];			/* XXX */
    946 		lp[0] = tmp;					/* XXX */
    947 	}							/* XXX */
    948 	dfs->fs_maxfilesize = mp->um_savedmaxfilesize;		/* XXX */
    949 	if (waitfor == MNT_WAIT)
    950 		error = bwrite(bp);
    951 	else
    952 		bawrite(bp);
    953 	return (error);
    954 }
    955 
    956 int
    957 ffs_cgupdate(mp, waitfor)
    958 	struct ufsmount *mp;
    959 	int waitfor;
    960 {
    961 	register struct fs *fs = mp->um_fs;
    962 	register struct buf *bp;
    963 	int blks;
    964 	caddr_t space;
    965 	int i, size, error = 0, allerror = 0;
    966 
    967 	allerror = ffs_sbupdate(mp, waitfor);
    968 	blks = howmany(fs->fs_cssize, fs->fs_fsize);
    969 	space = (caddr_t)fs->fs_csp[0];
    970 	for (i = 0; i < blks; i += fs->fs_frag) {
    971 		size = fs->fs_bsize;
    972 		if (i + fs->fs_frag > blks)
    973 			size = (blks - i) * fs->fs_fsize;
    974 		bp = getblk(mp->um_devvp, fsbtodb(fs, fs->fs_csaddr + i),
    975 		    size, 0, 0);
    976 		bcopy(space, bp->b_data, (u_int)size);
    977 		space += size;
    978 		if (waitfor == MNT_WAIT)
    979 			error = bwrite(bp);
    980 		else
    981 			bawrite(bp);
    982 	}
    983 	if (!allerror && error)
    984 		allerror = error;
    985 	return (allerror);
    986 }
    987