xref: /openbsd/sys/ufs/ffs/ffs_vfsops.c (revision 07ea8d15)
1 /*	$OpenBSD: ffs_vfsops.c,v 1.6 1996/06/27 06:42:06 downsj Exp $	*/
2 /*	$NetBSD: ffs_vfsops.c,v 1.19 1996/02/09 22:22:26 christos Exp $	*/
3 
4 /*
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  */
38 
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/namei.h>
42 #include <sys/proc.h>
43 #include <sys/kernel.h>
44 #include <sys/vnode.h>
45 #include <sys/socket.h>
46 #include <sys/mount.h>
47 #include <sys/buf.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 
55 #include <miscfs/specfs/specdev.h>
56 
57 #include <ufs/ufs/quota.h>
58 #include <ufs/ufs/ufsmount.h>
59 #include <ufs/ufs/inode.h>
60 #include <ufs/ufs/dir.h>
61 #include <ufs/ufs/ufs_extern.h>
62 
63 #include <ufs/ffs/fs.h>
64 #include <ufs/ffs/ffs_extern.h>
65 
66 int ffs_sbupdate __P((struct ufsmount *, int));
67 
68 struct vfsops ffs_vfsops = {
69 	MOUNT_FFS,
70 	ffs_mount,
71 	ufs_start,
72 	ffs_unmount,
73 	ufs_root,
74 	ufs_quotactl,
75 	ffs_statfs,
76 	ffs_sync,
77 	ffs_vget,
78 	ffs_fhtovp,
79 	ffs_vptofh,
80 	ffs_init,
81 };
82 
83 static struct ufs_dirops ffs_dirops = {
84 	ufs_dirremove,
85 	ufs_direnter,
86 	ufs_dirempty,
87 	ufs_dirrewrite,
88 	ufs_checkpath,
89 };
90 
91 extern u_long nextgennumber;
92 
93 /*
94  * Called by main() when ufs is going to be mounted as root.
95  *
96  * Name is updated by mount(8) after booting.
97  */
98 #define ROOTNAME	"root_device"
99 
100 int
101 ffs_mountroot()
102 {
103 	extern struct vnode *rootvp;
104 	register struct fs *fs;
105 	register struct mount *mp;
106 	struct proc *p = curproc;	/* XXX */
107 	struct ufsmount *ump;
108 	size_t size;
109 	int error;
110 
111 	/*
112 	 * Get vnodes for swapdev and rootdev.
113 	 */
114 	if (bdevvp(swapdev, &swapdev_vp) || bdevvp(rootdev, &rootvp))
115 		panic("ffs_mountroot: can't setup bdevvp's");
116 
117 	mp = malloc((u_long)sizeof(struct mount), M_MOUNT, M_WAITOK);
118 	bzero((char *)mp, (u_long)sizeof(struct mount));
119 	mp->mnt_op = &ffs_vfsops;
120 	mp->mnt_flag = MNT_RDONLY;
121 	if ((error = ffs_mountfs(rootvp, mp, p)) != 0) {
122 		free(mp, M_MOUNT);
123 		return (error);
124 	}
125 	if ((error = vfs_lock(mp)) != 0) {
126 		(void)ffs_unmount(mp, 0, p);
127 		free(mp, M_MOUNT);
128 		return (error);
129 	}
130 	CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
131 	mp->mnt_vnodecovered = NULLVP;
132 	ump = VFSTOUFS(mp);
133 	fs = ump->um_fs;
134 	bzero(fs->fs_fsmnt, sizeof(fs->fs_fsmnt));
135 	fs->fs_fsmnt[0] = '/';
136 	bcopy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname, MNAMELEN);
137 	(void) copystr(ROOTNAME, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
138 	    &size);
139 	bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
140 	(void)ffs_statfs(mp, &mp->mnt_stat, p);
141 	vfs_unlock(mp);
142 	inittodr(fs->fs_time);
143 	return (0);
144 }
145 
146 /*
147  * VFS Operations.
148  *
149  * mount system call
150  */
151 int
152 ffs_mount(mp, path, data, ndp, p)
153 	register struct mount *mp;
154 	char *path;
155 	caddr_t data;
156 	struct nameidata *ndp;
157 	struct proc *p;
158 {
159 	struct vnode *devvp;
160 	struct ufs_args args;
161 	struct ufsmount *ump = NULL;
162 	register struct fs *fs;
163 	size_t size;
164 	int error, flags;
165 	mode_t accessmode;
166 
167 	error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args));
168 	if (error)
169 		return (error);
170 	/*
171 	 * If updating, check whether changing from read-only to
172 	 * read/write; if there is no device name, that's all we do.
173 	 */
174 	if (mp->mnt_flag & MNT_UPDATE) {
175 		ump = VFSTOUFS(mp);
176 		fs = ump->um_fs;
177 		if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
178 			flags = WRITECLOSE;
179 			if (mp->mnt_flag & MNT_FORCE)
180 				flags |= FORCECLOSE;
181 			if (vfs_busy(mp))
182 				return (EBUSY);
183 			error = ffs_flushfiles(mp, flags, p);
184 			if (error == 0 &&
185 			    ffs_cgupdate(ump, MNT_WAIT) == 0 &&
186 			    fs->fs_clean & FS_WASCLEAN) {
187 				fs->fs_clean = FS_ISCLEAN;
188 				(void) ffs_sbupdate(ump, MNT_WAIT);
189 			}
190 			vfs_unbusy(mp);
191 			if (error)
192 				return (error);
193 			fs->fs_ronly = 1;
194 		}
195 		if (mp->mnt_flag & MNT_RELOAD) {
196 			error = ffs_reload(mp, ndp->ni_cnd.cn_cred, p);
197 			if (error)
198 				return (error);
199 		}
200 		if (fs->fs_ronly && (mp->mnt_flag & MNT_WANTRDWR)) {
201 			/*
202 			 * If upgrade to read-write by non-root, then verify
203 			 * that user has necessary permissions on the device.
204 			 */
205 			if (p->p_ucred->cr_uid != 0) {
206 				devvp = ump->um_devvp;
207 				VOP_LOCK(devvp);
208 				error = VOP_ACCESS(devvp, VREAD | VWRITE,
209 						   p->p_ucred, p);
210 				if (error) {
211 					VOP_UNLOCK(devvp);
212 					return (error);
213 				}
214 				VOP_UNLOCK(devvp);
215 			}
216 			fs->fs_ronly = 0;
217 			fs->fs_clean <<= 1;
218 			fs->fs_fmod = 1;
219 		}
220 		if (args.fspec == 0) {
221 			/*
222 			 * Process export requests.
223 			 */
224 			return (vfs_export(mp, &ump->um_export, &args.export));
225 		}
226 	}
227 	/*
228 	 * Not an update, or updating the name: look up the name
229 	 * and verify that it refers to a sensible block device.
230 	 */
231 	NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, p);
232 	if ((error = namei(ndp)) != 0)
233 		return (error);
234 	devvp = ndp->ni_vp;
235 
236 	if (devvp->v_type != VBLK) {
237 		vrele(devvp);
238 		return (ENOTBLK);
239 	}
240 	if (major(devvp->v_rdev) >= nblkdev) {
241 		vrele(devvp);
242 		return (ENXIO);
243 	}
244 	/*
245 	 * If mount by non-root, then verify that user has necessary
246 	 * permissions on the device.
247 	 */
248 	if (p->p_ucred->cr_uid != 0) {
249 		accessmode = VREAD;
250 		if ((mp->mnt_flag & MNT_RDONLY) == 0)
251 			accessmode |= VWRITE;
252 		VOP_LOCK(devvp);
253 		error = VOP_ACCESS(devvp, accessmode, p->p_ucred, p);
254 		if (error) {
255 			vput(devvp);
256 			return (error);
257 		}
258 		VOP_UNLOCK(devvp);
259 	}
260 	if ((mp->mnt_flag & MNT_UPDATE) == 0)
261 		error = ffs_mountfs(devvp, mp, p);
262 	else {
263 		if (devvp != ump->um_devvp)
264 			error = EINVAL;	/* needs translation */
265 		else
266 			vrele(devvp);
267 	}
268 	if (error) {
269 		vrele(devvp);
270 		return (error);
271 	}
272 	ump = VFSTOUFS(mp);
273 	fs = ump->um_fs;
274 	(void) copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size);
275 	bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size);
276 	bcopy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname, MNAMELEN);
277 	(void) copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
278 	    &size);
279 	bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
280 	if (fs->fs_fmod != 0) {	/* XXX */
281 		fs->fs_fmod = 0;
282 		if (fs->fs_clean & FS_WASCLEAN)
283 			fs->fs_time = time.tv_sec;
284 		else
285 			printf("%s: file system not clean; please fsck(8)\n",
286 			    mp->mnt_stat.f_mntfromname);
287 		(void) ffs_cgupdate(ump, MNT_WAIT);
288 	}
289 	return (0);
290 }
291 
292 /*
293  * Reload all incore data for a filesystem (used after running fsck on
294  * the root filesystem and finding things to fix). The filesystem must
295  * be mounted read-only.
296  *
297  * Things to do to update the mount:
298  *	1) invalidate all cached meta-data.
299  *	2) re-read superblock from disk.
300  *	3) re-read summary information from disk.
301  *	4) invalidate all inactive vnodes.
302  *	5) invalidate all cached file data.
303  *	6) re-read inode data for all active vnodes.
304  */
305 int
306 ffs_reload(mountp, cred, p)
307 	register struct mount *mountp;
308 	struct ucred *cred;
309 	struct proc *p;
310 {
311 	register struct vnode *vp, *nvp, *devvp;
312 	struct inode *ip;
313 	struct csum *space;
314 	struct buf *bp;
315 	struct fs *fs, *newfs;
316 	struct partinfo dpart;
317 	int i, blks, size, error;
318 	int32_t *lp;
319 
320 	if ((mountp->mnt_flag & MNT_RDONLY) == 0)
321 		return (EINVAL);
322 	/*
323 	 * Step 1: invalidate all cached meta-data.
324 	 */
325 	devvp = VFSTOUFS(mountp)->um_devvp;
326 	if (vinvalbuf(devvp, 0, cred, p, 0, 0))
327 		panic("ffs_reload: dirty1");
328 	/*
329 	 * Step 2: re-read superblock from disk.
330 	 */
331 	if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, p) != 0)
332 		size = DEV_BSIZE;
333 	else
334 		size = dpart.disklab->d_secsize;
335 	error = bread(devvp, (daddr_t)(SBOFF / size), SBSIZE, NOCRED, &bp);
336 	if (error)
337 		return (error);
338 	newfs = (struct fs *)bp->b_data;
339 	if (newfs->fs_magic != FS_MAGIC || newfs->fs_bsize > MAXBSIZE ||
340 	    newfs->fs_bsize < sizeof(struct fs)) {
341 		brelse(bp);
342 		return (EIO);		/* XXX needs translation */
343 	}
344 	fs = VFSTOUFS(mountp)->um_fs;
345 	/*
346 	 * Copy pointer fields back into superblock before copying in	XXX
347 	 * new superblock. These should really be in the ufsmount.	XXX
348 	 * Note that important parameters (eg fs_ncg) are unchanged.
349 	 */
350 	bcopy(&fs->fs_csp[0], &newfs->fs_csp[0], sizeof(fs->fs_csp));
351 	newfs->fs_maxcluster = fs->fs_maxcluster;
352 	bcopy(newfs, fs, (u_int)fs->fs_sbsize);
353 	if (fs->fs_sbsize < SBSIZE)
354 		bp->b_flags |= B_INVAL;
355 	brelse(bp);
356 	mountp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
357 	ffs_oldfscompat(fs);
358 	/*
359 	 * Step 3: re-read summary information from disk.
360 	 */
361 	blks = howmany(fs->fs_cssize, fs->fs_fsize);
362 	space = fs->fs_csp[0];
363 	for (i = 0; i < blks; i += fs->fs_frag) {
364 		size = fs->fs_bsize;
365 		if (i + fs->fs_frag > blks)
366 			size = (blks - i) * fs->fs_fsize;
367 		error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
368 			      NOCRED, &bp);
369 		if (error)
370 			return (error);
371 		bcopy(bp->b_data, fs->fs_csp[fragstoblks(fs, i)], (u_int)size);
372 		brelse(bp);
373 	}
374 	/*
375 	 * We no longer know anything about clusters per cylinder group.
376 	 */
377 	if (fs->fs_contigsumsize > 0) {
378 		lp = fs->fs_maxcluster;
379 		for (i = 0; i < fs->fs_ncg; i++)
380 			*lp++ = fs->fs_contigsumsize;
381 	}
382 
383 loop:
384 	for (vp = mountp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) {
385 		nvp = vp->v_mntvnodes.le_next;
386 		/*
387 		 * Step 4: invalidate all inactive vnodes.
388 		 */
389 		if (vp->v_usecount == 0) {
390 			vgone(vp);
391 			continue;
392 		}
393 		/*
394 		 * Step 5: invalidate all cached file data.
395 		 */
396 		if (vget(vp, 1))
397 			goto loop;
398 		if (vinvalbuf(vp, 0, cred, p, 0, 0))
399 			panic("ffs_reload: dirty2");
400 		/*
401 		 * Step 6: re-read inode data for all active vnodes.
402 		 */
403 		ip = VTOI(vp);
404 		error = bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
405 			      (int)fs->fs_bsize, NOCRED, &bp);
406 		if (error) {
407 			vput(vp);
408 			return (error);
409 		}
410 		ip->i_din = *((struct dinode *)bp->b_data +
411 		    ino_to_fsbo(fs, ip->i_number));
412 		brelse(bp);
413 		vput(vp);
414 		if (vp->v_mount != mountp)
415 			goto loop;
416 	}
417 	return (0);
418 }
419 
420 /*
421  * Common code for mount and mountroot
422  */
423 int
424 ffs_mountfs(devvp, mp, p)
425 	register struct vnode *devvp;
426 	struct mount *mp;
427 	struct proc *p;
428 {
429 	register struct ufsmount *ump;
430 	struct buf *bp;
431 	register struct fs *fs;
432 	dev_t dev;
433 	struct partinfo dpart;
434 	caddr_t base, space;
435 	int blks;
436 	int error, i, size, ronly;
437 	int32_t *lp;
438 	struct ucred *cred;
439 	extern struct vnode *rootvp;
440 	u_int64_t maxfilesize;					/* XXX */
441 
442 	dev = devvp->v_rdev;
443 	cred = p ? p->p_ucred : NOCRED;
444 	/*
445 	 * Disallow multiple mounts of the same device.
446 	 * Disallow mounting of a device that is currently in use
447 	 * (except for root, which might share swap device for miniroot).
448 	 * Flush out any old buffers remaining from a previous use.
449 	 */
450 	if ((error = vfs_mountedon(devvp)) != 0)
451 		return (error);
452 	if (vcount(devvp) > 1 && devvp != rootvp)
453 		return (EBUSY);
454 	if ((error = vinvalbuf(devvp, V_SAVE, cred, p, 0, 0)) != 0)
455 		return (error);
456 
457 	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
458 	error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p);
459 	if (error)
460 		return (error);
461 	if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, cred, p) != 0)
462 		size = DEV_BSIZE;
463 	else
464 		size = dpart.disklab->d_secsize;
465 
466 	bp = NULL;
467 	ump = NULL;
468 	error = bread(devvp, (daddr_t)(SBOFF / size), SBSIZE, cred, &bp);
469 	if (error)
470 		goto out;
471 	fs = (struct fs *)bp->b_data;
472 	if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE ||
473 	    fs->fs_bsize < sizeof(struct fs)) {
474 		error = EINVAL;		/* XXX needs translation */
475 		goto out;
476 	}
477 	/* XXX updating 4.2 FFS superblocks trashes rotational layout tables */
478 	if (fs->fs_postblformat == FS_42POSTBLFMT && !ronly) {
479 		error = EROFS;		/* XXX what should be returned? */
480 		goto out;
481 	}
482 	ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
483 	bzero((caddr_t)ump, sizeof *ump);
484 	ump->um_fs = malloc((u_long)fs->fs_sbsize, M_UFSMNT,
485 	    M_WAITOK);
486 	bcopy(bp->b_data, ump->um_fs, (u_int)fs->fs_sbsize);
487 	if (fs->fs_sbsize < SBSIZE)
488 		bp->b_flags |= B_INVAL;
489 	brelse(bp);
490 	bp = NULL;
491 	fs = ump->um_fs;
492 	fs->fs_ronly = ronly;
493 	if (ronly == 0) {
494 		fs->fs_clean <<= 1;
495 		fs->fs_fmod = 1;
496 	}
497 	size = fs->fs_cssize;
498 	blks = howmany(size, fs->fs_fsize);
499 	if (fs->fs_contigsumsize > 0)
500 		size += fs->fs_ncg * sizeof(int32_t);
501 	base = space = malloc((u_long)size, M_UFSMNT, M_WAITOK);
502 	for (i = 0; i < blks; i += fs->fs_frag) {
503 		size = fs->fs_bsize;
504 		if (i + fs->fs_frag > blks)
505 			size = (blks - i) * fs->fs_fsize;
506 		error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
507 			      cred, &bp);
508 		if (error) {
509 			free(base, M_UFSMNT);
510 			goto out;
511 		}
512 		bcopy(bp->b_data, space, (u_int)size);
513 		fs->fs_csp[fragstoblks(fs, i)] = (struct csum *)space;
514 		space += size;
515 		brelse(bp);
516 		bp = NULL;
517 	}
518 	if (fs->fs_contigsumsize > 0) {
519 		fs->fs_maxcluster = lp = (int32_t *)space;
520 		for (i = 0; i < fs->fs_ncg; i++)
521 			*lp++ = fs->fs_contigsumsize;
522 	}
523 	mp->mnt_data = (qaddr_t)ump;
524 	mp->mnt_stat.f_fsid.val[0] = (long)dev;
525 	mp->mnt_stat.f_fsid.val[1] = makefstype(MOUNT_FFS);
526 	mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
527 	mp->mnt_flag |= MNT_LOCAL;
528 	ump->um_mountp = mp;
529 	ump->um_dev = dev;
530 	ump->um_devvp = devvp;
531 	ump->um_nindir = fs->fs_nindir;
532 	ump->um_bptrtodb = fs->fs_fsbtodb;
533 	ump->um_seqinc = fs->fs_frag;
534 	ump->um_dirops = &ffs_dirops;
535 	for (i = 0; i < MAXQUOTAS; i++)
536 		ump->um_quotas[i] = NULLVP;
537 	devvp->v_specflags |= SI_MOUNTEDON;
538 	ffs_oldfscompat(fs);
539 	ump->um_savedmaxfilesize = fs->fs_maxfilesize;		/* XXX */
540 	maxfilesize = (u_int64_t)0x80000000 * fs->fs_bsize - 1;	/* XXX */
541 	if (fs->fs_maxfilesize > maxfilesize)			/* XXX */
542 		fs->fs_maxfilesize = maxfilesize;		/* XXX */
543 	return (0);
544 out:
545 	if (bp)
546 		brelse(bp);
547 	(void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, cred, p);
548 	if (ump) {
549 		free(ump->um_fs, M_UFSMNT);
550 		free(ump, M_UFSMNT);
551 		mp->mnt_data = (qaddr_t)0;
552 	}
553 	return (error);
554 }
555 
556 /*
557  * Sanity checks for old file systems.
558  *
559  * XXX - goes away some day.
560  */
561 int
562 ffs_oldfscompat(fs)
563 	struct fs *fs;
564 {
565 	int i;
566 
567 	fs->fs_npsect = max(fs->fs_npsect, fs->fs_nsect);	/* XXX */
568 	fs->fs_interleave = max(fs->fs_interleave, 1);		/* XXX */
569 	if (fs->fs_postblformat == FS_42POSTBLFMT)		/* XXX */
570 		fs->fs_nrpos = 8;				/* XXX */
571 	if (fs->fs_inodefmt < FS_44INODEFMT) {			/* XXX */
572 		u_int64_t sizepb = fs->fs_bsize;		/* XXX */
573 								/* XXX */
574 		fs->fs_maxfilesize = fs->fs_bsize * NDADDR - 1;	/* XXX */
575 		for (i = 0; i < NIADDR; i++) {			/* XXX */
576 			sizepb *= NINDIR(fs);			/* XXX */
577 			fs->fs_maxfilesize += sizepb;		/* XXX */
578 		}						/* XXX */
579 		fs->fs_qbmask = ~fs->fs_bmask;			/* XXX */
580 		fs->fs_qfmask = ~fs->fs_fmask;			/* XXX */
581 	}							/* XXX */
582 	return (0);
583 }
584 
585 /*
586  * unmount system call
587  */
588 int
589 ffs_unmount(mp, mntflags, p)
590 	struct mount *mp;
591 	int mntflags;
592 	struct proc *p;
593 {
594 	register struct ufsmount *ump;
595 	register struct fs *fs;
596 	int error, flags;
597 
598 	flags = 0;
599 	if (mntflags & MNT_FORCE)
600 		flags |= FORCECLOSE;
601 	if ((error = ffs_flushfiles(mp, flags, p)) != 0)
602 		return (error);
603 	ump = VFSTOUFS(mp);
604 	fs = ump->um_fs;
605 	if (fs->fs_ronly == 0 &&
606 	    ffs_cgupdate(ump, MNT_WAIT) == 0 &&
607 	    fs->fs_clean & FS_WASCLEAN) {
608 		fs->fs_clean = FS_ISCLEAN;
609 		(void) ffs_sbupdate(ump, MNT_WAIT);
610 	}
611 	ump->um_devvp->v_specflags &= ~SI_MOUNTEDON;
612 	error = VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD|FWRITE,
613 		NOCRED, p);
614 	vrele(ump->um_devvp);
615 	free(fs->fs_csp[0], M_UFSMNT);
616 	free(fs, M_UFSMNT);
617 	free(ump, M_UFSMNT);
618 	mp->mnt_data = (qaddr_t)0;
619 	mp->mnt_flag &= ~MNT_LOCAL;
620 	return (error);
621 }
622 
623 /*
624  * Flush out all the files in a filesystem.
625  */
626 int
627 ffs_flushfiles(mp, flags, p)
628 	register struct mount *mp;
629 	int flags;
630 	struct proc *p;
631 {
632 	extern int doforce;
633 	register struct ufsmount *ump;
634 	int error;
635 
636 	if (!doforce)
637 		flags &= ~FORCECLOSE;
638 	ump = VFSTOUFS(mp);
639 #ifdef QUOTA
640 	if (mp->mnt_flag & MNT_QUOTA) {
641 		int i;
642 		if ((error = vflush(mp, NULLVP, SKIPSYSTEM|flags)) != 0)
643 			return (error);
644 		for (i = 0; i < MAXQUOTAS; i++) {
645 			if (ump->um_quotas[i] == NULLVP)
646 				continue;
647 			quotaoff(p, mp, i);
648 		}
649 		/*
650 		 * Here we fall through to vflush again to ensure
651 		 * that we have gotten rid of all the system vnodes.
652 		 */
653 	}
654 #endif
655 	error = vflush(mp, NULLVP, flags);
656 	return (error);
657 }
658 
659 /*
660  * Get file system statistics.
661  */
662 int
663 ffs_statfs(mp, sbp, p)
664 	struct mount *mp;
665 	register struct statfs *sbp;
666 	struct proc *p;
667 {
668 	register struct ufsmount *ump;
669 	register struct fs *fs;
670 
671 	ump = VFSTOUFS(mp);
672 	fs = ump->um_fs;
673 	if (fs->fs_magic != FS_MAGIC)
674 		panic("ffs_statfs");
675 #ifdef COMPAT_09
676 	sbp->f_type = 1;
677 #else
678 	sbp->f_type = 0;
679 #endif
680 	sbp->f_bsize = fs->fs_fsize;
681 	sbp->f_iosize = fs->fs_bsize;
682 	sbp->f_blocks = fs->fs_dsize;
683 	sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag +
684 		fs->fs_cstotal.cs_nffree;
685 	sbp->f_bavail = (fs->fs_dsize * (100 - fs->fs_minfree) / 100) -
686 		(fs->fs_dsize - sbp->f_bfree);
687 	sbp->f_files =  fs->fs_ncg * fs->fs_ipg - ROOTINO;
688 	sbp->f_ffree = fs->fs_cstotal.cs_nifree;
689 	if (sbp != &mp->mnt_stat) {
690 		bcopy(mp->mnt_stat.f_mntonname, sbp->f_mntonname, MNAMELEN);
691 		bcopy(mp->mnt_stat.f_mntfromname, sbp->f_mntfromname, MNAMELEN);
692 	}
693 	strncpy(sbp->f_fstypename, mp->mnt_op->vfs_name, MFSNAMELEN);
694 	return (0);
695 }
696 
697 /*
698  * Go through the disk queues to initiate sandbagged IO;
699  * go through the inodes to write those that have been modified;
700  * initiate the writing of the super block if it has been modified.
701  *
702  * Note: we are always called with the filesystem marked `MPBUSY'.
703  */
704 int
705 ffs_sync(mp, waitfor, cred, p)
706 	struct mount *mp;
707 	int waitfor;
708 	struct ucred *cred;
709 	struct proc *p;
710 {
711 	register struct vnode *vp;
712 	register struct inode *ip;
713 	register struct ufsmount *ump = VFSTOUFS(mp);
714 	register struct fs *fs;
715 	int error, allerror = 0;
716 
717 	fs = ump->um_fs;
718 	/*
719 	 * Write back modified superblock.
720 	 * Consistency check that the superblock
721 	 * is still in the buffer cache.
722 	 */
723 	if (fs->fs_fmod != 0) {
724 		if (fs->fs_ronly != 0) {		/* XXX */
725 			printf("fs = %s\n", fs->fs_fsmnt);
726 			panic("update: rofs mod");
727 		}
728 		fs->fs_fmod = 0;
729 		fs->fs_time = time.tv_sec;
730 		allerror = ffs_cgupdate(ump, waitfor);
731 	}
732 	/*
733 	 * Write back each (modified) inode.
734 	 */
735 loop:
736 	for (vp = mp->mnt_vnodelist.lh_first;
737 	     vp != NULL;
738 	     vp = vp->v_mntvnodes.le_next) {
739 		/*
740 		 * If the vnode that we are about to sync is no longer
741 		 * associated with this mount point, start over.
742 		 */
743 		if (vp->v_mount != mp)
744 			goto loop;
745 		if (VOP_ISLOCKED(vp))
746 			continue;
747 		ip = VTOI(vp);
748 		if ((ip->i_flag &
749 		    (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
750 		    vp->v_dirtyblkhd.lh_first == NULL)
751 			continue;
752 		if (vget(vp, 1))
753 			goto loop;
754 		if ((error = VOP_FSYNC(vp, cred, waitfor, p)) != 0)
755 			allerror = error;
756 		vput(vp);
757 	}
758 	/*
759 	 * Force stale file system control information to be flushed.
760 	 */
761 	if ((error = VOP_FSYNC(ump->um_devvp, cred, waitfor, p)) != 0)
762 		allerror = error;
763 #ifdef QUOTA
764 	qsync(mp);
765 #endif
766 	return (allerror);
767 }
768 
769 /*
770  * Look up a FFS dinode number to find its incore vnode, otherwise read it
771  * in from disk.  If it is in core, wait for the lock bit to clear, then
772  * return the inode locked.  Detection and handling of mount points must be
773  * done by the calling routine.
774  */
775 int
776 ffs_vget(mp, ino, vpp)
777 	struct mount *mp;
778 	ino_t ino;
779 	struct vnode **vpp;
780 {
781 	register struct fs *fs;
782 	register struct inode *ip;
783 	struct ufsmount *ump;
784 	struct buf *bp;
785 	struct vnode *vp;
786 	dev_t dev;
787 	int type, error;
788 
789 	ump = VFSTOUFS(mp);
790 	dev = ump->um_dev;
791 	if ((*vpp = ufs_ihashget(dev, ino)) != NULL)
792 		return (0);
793 
794 	/* Allocate a new vnode/inode. */
795 	if ((error = getnewvnode(VT_UFS, mp, ffs_vnodeop_p, &vp)) != 0) {
796 		*vpp = NULL;
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 
823 	/* Read in the disk contents for the inode, copy into the inode. */
824 	error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
825 		      (int)fs->fs_bsize, NOCRED, &bp);
826 	if (error) {
827 		/*
828 		 * The inode does not contain anything useful, so it would
829 		 * be misleading to leave it on its hash chain. With mode
830 		 * still zero, it will be unlinked and returned to the free
831 		 * list by vput().
832 		 */
833 		vput(vp);
834 		brelse(bp);
835 		*vpp = NULL;
836 		return (error);
837 	}
838 	ip->i_din = *((struct dinode *)bp->b_data + ino_to_fsbo(fs, ino));
839 	brelse(bp);
840 
841 	/*
842 	 * Initialize the vnode from the inode, check for aliases.
843 	 * Note that the underlying vnode may have changed.
844 	 */
845 	error = ufs_vinit(mp, ffs_specop_p, FFS_FIFOOPS, &vp);
846 	if (error) {
847 		vput(vp);
848 		*vpp = NULL;
849 		return (error);
850 	}
851 	/*
852 	 * Finish inode initialization now that aliasing has been resolved.
853 	 */
854 	ip->i_devvp = ump->um_devvp;
855 	VREF(ip->i_devvp);
856 	/*
857 	 * Set up a generation number for this inode if it does not
858 	 * already have one. This should only happen on old filesystems.
859 	 */
860 	if (ip->i_gen == 0) {
861 		if (++nextgennumber < (u_long)time.tv_sec)
862 			nextgennumber = time.tv_sec;
863 		ip->i_gen = nextgennumber;
864 		if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
865 			ip->i_flag |= IN_MODIFIED;
866 	}
867 	/*
868 	 * Ensure that uid and gid are correct. This is a temporary
869 	 * fix until fsck has been changed to do the update.
870 	 */
871 	if (fs->fs_inodefmt < FS_44INODEFMT) {		/* XXX */
872 		ip->i_uid = ip->i_din.di_ouid;		/* XXX */
873 		ip->i_gid = ip->i_din.di_ogid;		/* XXX */
874 	}						/* XXX */
875 
876 	*vpp = vp;
877 	return (0);
878 }
879 
880 /*
881  * File handle to vnode
882  *
883  * Have to be really careful about stale file handles:
884  * - check that the inode number is valid
885  * - call ffs_vget() to get the locked inode
886  * - check for an unallocated inode (i_mode == 0)
887  * - check that the given client host has export rights and return
888  *   those rights via. exflagsp and credanonp
889  */
890 int
891 ffs_fhtovp(mp, fhp, nam, vpp, exflagsp, credanonp)
892 	register struct mount *mp;
893 	struct fid *fhp;
894 	struct mbuf *nam;
895 	struct vnode **vpp;
896 	int *exflagsp;
897 	struct ucred **credanonp;
898 {
899 	register struct ufid *ufhp;
900 	struct fs *fs;
901 
902 	ufhp = (struct ufid *)fhp;
903 	fs = VFSTOUFS(mp)->um_fs;
904 	if (ufhp->ufid_ino < ROOTINO ||
905 	    ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg)
906 		return (ESTALE);
907 	return (ufs_check_export(mp, ufhp, nam, vpp, exflagsp, credanonp));
908 }
909 
910 /*
911  * Vnode pointer to File handle
912  */
913 /* ARGSUSED */
914 int
915 ffs_vptofh(vp, fhp)
916 	struct vnode *vp;
917 	struct fid *fhp;
918 {
919 	register struct inode *ip;
920 	register struct ufid *ufhp;
921 
922 	ip = VTOI(vp);
923 	ufhp = (struct ufid *)fhp;
924 	ufhp->ufid_len = sizeof(struct ufid);
925 	ufhp->ufid_ino = ip->i_number;
926 	ufhp->ufid_gen = ip->i_gen;
927 	return (0);
928 }
929 
930 /*
931  * Write a superblock and associated information back to disk.
932  */
933 int
934 ffs_sbupdate(mp, waitfor)
935 	struct ufsmount *mp;
936 	int waitfor;
937 {
938 	register struct fs *dfs, *fs = mp->um_fs;
939 	register struct buf *bp;
940 	int i, error = 0;
941 
942 	bp = getblk(mp->um_devvp, SBOFF >> (fs->fs_fshift - fs->fs_fsbtodb),
943 	    (int)fs->fs_sbsize, 0, 0);
944 	bcopy((caddr_t)fs, bp->b_data, (u_int)fs->fs_sbsize);
945 	/* Restore compatibility to old file systems.		   XXX */
946 	dfs = (struct fs *)bp->b_data;				/* XXX */
947 	if (fs->fs_postblformat == FS_42POSTBLFMT)		/* XXX */
948 		dfs->fs_nrpos = -1;				/* XXX */
949 	if (fs->fs_inodefmt < FS_44INODEFMT) {			/* XXX */
950 		int32_t *lp, tmp;				/* XXX */
951 								/* XXX */
952 		lp = (int32_t *)&dfs->fs_qbmask;		/* XXX */
953 		tmp = lp[4];					/* XXX */
954 		for (i = 4; i > 0; i--)				/* XXX */
955 			lp[i] = lp[i-1];			/* XXX */
956 		lp[0] = tmp;					/* XXX */
957 	}							/* XXX */
958 	dfs->fs_maxfilesize = mp->um_savedmaxfilesize;		/* XXX */
959 	if (waitfor == MNT_WAIT)
960 		error = bwrite(bp);
961 	else
962 		bawrite(bp);
963 	return (error);
964 }
965 
966 int
967 ffs_cgupdate(mp, waitfor)
968 	struct ufsmount *mp;
969 	int waitfor;
970 {
971 	register struct fs *fs = mp->um_fs;
972 	register struct buf *bp;
973 	int blks;
974 	caddr_t space;
975 	int i, size, error = 0, allerror = 0;
976 
977 	allerror = ffs_sbupdate(mp, waitfor);
978 	blks = howmany(fs->fs_cssize, fs->fs_fsize);
979 	space = (caddr_t)fs->fs_csp[0];
980 	for (i = 0; i < blks; i += fs->fs_frag) {
981 		size = fs->fs_bsize;
982 		if (i + fs->fs_frag > blks)
983 			size = (blks - i) * fs->fs_fsize;
984 		bp = getblk(mp->um_devvp, fsbtodb(fs, fs->fs_csaddr + i),
985 		    size, 0, 0);
986 		bcopy(space, bp->b_data, (u_int)size);
987 		space += size;
988 		if (waitfor == MNT_WAIT)
989 			error = bwrite(bp);
990 		else
991 			bawrite(bp);
992 	}
993 	if (!allerror && error)
994 		allerror = error;
995 	return (allerror);
996 }
997