xref: /freebsd/sys/ufs/ffs/ffs_vfsops.c (revision 5ebc7e6281887681c3a348a5a4c902e262ccd656)
1 /*
2  * Copyright (c) 1989, 1991, 1993, 1994
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by the University of
16  *	California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	@(#)ffs_vfsops.c	8.8 (Berkeley) 4/18/94
34  * $Id: ffs_vfsops.c,v 1.19 1995/05/15 08:39:37 davidg Exp $
35  */
36 
37 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/namei.h>
40 #include <sys/proc.h>
41 #include <sys/kernel.h>
42 #include <sys/vnode.h>
43 #include <sys/socket.h>
44 #include <sys/mount.h>
45 #include <sys/buf.h>
46 #include <sys/mbuf.h>
47 #include <sys/file.h>
48 #include <sys/disklabel.h>
49 #include <sys/ioctl.h>
50 #include <sys/errno.h>
51 #include <sys/malloc.h>
52 
53 #include <miscfs/specfs/specdev.h>
54 
55 #include <ufs/ufs/quota.h>
56 #include <ufs/ufs/ufsmount.h>
57 #include <ufs/ufs/inode.h>
58 #include <ufs/ufs/ufs_extern.h>
59 
60 #include <ufs/ffs/fs.h>
61 #include <ufs/ffs/ffs_extern.h>
62 
63 #include <vm/vm.h>
64 #include <vm/vm_page.h>
65 #include <vm/vm_object.h>
66 
67 int	ffs_sbupdate __P((struct ufsmount *, int));
68 int	ffs_reload __P((struct mount *,struct ucred *,struct proc *));
69 int	ffs_oldfscompat __P((struct fs *));
70 void ffs_vmlimits __P((struct fs *));
71 
72 struct vfsops ufs_vfsops = {
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 };
85 
86 VFS_SET(ufs_vfsops, ufs, MOUNT_UFS, 0);
87 
88 extern u_long nextgennumber;
89 
90 /*
91  * Called by main() when ufs is going to be mounted as root.
92  *
93  * Name is updated by mount(8) after booting.
94  */
95 #define ROOTNAME	"root_device"
96 
97 int
98 ffs_mountroot()
99 {
100 	register struct fs *fs;
101 	register struct mount *mp;
102 	struct proc *p = curproc;	/* XXX */
103 	struct ufsmount *ump;
104 	u_int size;
105 	int error;
106 
107 	/*
108 	 * Get vnode for rootdev.
109 	 */
110 	if (bdevvp(rootdev, &rootvp))
111 		panic("ffs_mountroot: can't setup bdevvp for root");
112 
113 	mp = malloc((u_long)sizeof(struct mount), M_MOUNT, M_WAITOK);
114 	bzero((char *)mp, (u_long)sizeof(struct mount));
115 	mp->mnt_op = &ufs_vfsops;
116 	mp->mnt_flag = MNT_RDONLY;
117 	error = ffs_mountfs(rootvp, mp, p);
118 	if (error) {
119 		free(mp, M_MOUNT);
120 		return (error);
121 	}
122 	error = vfs_lock(mp);
123 	if (error) {
124 		(void)ffs_unmount(mp, 0, p);
125 		free(mp, M_MOUNT);
126 		return (error);
127 	}
128 	TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
129 	mp->mnt_flag |= MNT_ROOTFS;
130 	mp->mnt_vnodecovered = NULLVP;
131 	ump = VFSTOUFS(mp);
132 	fs = ump->um_fs;
133 	bzero(fs->fs_fsmnt, sizeof(fs->fs_fsmnt));
134 	fs->fs_fsmnt[0] = '/';
135 	bcopy((caddr_t)fs->fs_fsmnt, (caddr_t)mp->mnt_stat.f_mntonname,
136 	    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 = 0;
162 	register struct fs *fs;
163 	u_int size;
164 	int error, flags;
165 
166 	error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args));
167 	if (error)
168 		return (error);
169 	/*
170 	 * If updating, check whether changing from read-only to
171 	 * read/write; if there is no device name, that's all we do.
172 	 */
173 	if (mp->mnt_flag & MNT_UPDATE) {
174 		ump = VFSTOUFS(mp);
175 		fs = ump->um_fs;
176 		error = 0;
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 			vfs_unbusy(mp);
185 		}
186 		if (!error && (mp->mnt_flag & MNT_RELOAD))
187 			error = ffs_reload(mp, ndp->ni_cnd.cn_cred, p);
188 		if (error)
189 			return (error);
190 		if (fs->fs_ronly && (mp->mnt_flag & MNT_WANTRDWR)) {
191 			if (!fs->fs_clean) {
192 				if (mp->mnt_flag & MNT_FORCE) {
193 					printf("WARNING: %s was not properly dismounted.\n",fs->fs_fsmnt);
194 				} else {
195 					printf("WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck.\n",
196 					    fs->fs_fsmnt);
197 					return (EPERM);
198 				}
199 			}
200 			fs->fs_ronly = 0;
201 		}
202 		if (fs->fs_ronly == 0) {
203 			fs->fs_clean = 0;
204 			ffs_sbupdate(ump, MNT_WAIT);
205 		}
206 		if (args.fspec == 0) {
207 			/*
208 			 * Process export requests.
209 			 */
210 			return (vfs_export(mp, &ump->um_export, &args.export));
211 		}
212 	}
213 	/*
214 	 * Not an update, or updating the name: look up the name
215 	 * and verify that it refers to a sensible block device.
216 	 */
217 	NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, p);
218 	error = namei(ndp);
219 	if (error)
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 	if ((mp->mnt_flag & MNT_UPDATE) == 0)
232 		error = ffs_mountfs(devvp, mp, p);
233 	else {
234 		if (devvp != ump->um_devvp)
235 			error = EINVAL;	/* needs translation */
236 		else
237 			vrele(devvp);
238 	}
239 	if (error) {
240 		vrele(devvp);
241 		return (error);
242 	}
243 	ump = VFSTOUFS(mp);
244 	fs = ump->um_fs;
245 	(void) copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size);
246 	bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size);
247 	bcopy((caddr_t)fs->fs_fsmnt, (caddr_t)mp->mnt_stat.f_mntonname,
248 	    MNAMELEN);
249 	(void) copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
250 	    &size);
251 	bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
252 	(void)ffs_statfs(mp, &mp->mnt_stat, p);
253 	return (0);
254 }
255 
256 /*
257  * Reload all incore data for a filesystem (used after running fsck on
258  * the root filesystem and finding things to fix). The filesystem must
259  * be mounted read-only.
260  *
261  * Things to do to update the mount:
262  *	1) invalidate all cached meta-data.
263  *	2) re-read superblock from disk.
264  *	3) re-read summary information from disk.
265  *	4) invalidate all inactive vnodes.
266  *	5) invalidate all cached file data.
267  *	6) re-read inode data for all active vnodes.
268  */
269 int
270 ffs_reload(mountp, cred, p)
271 	register struct mount *mountp;
272 	struct ucred *cred;
273 	struct proc *p;
274 {
275 	register struct vnode *vp, *nvp, *devvp;
276 	struct inode *ip;
277 	struct csum *space;
278 	struct buf *bp;
279 	struct fs *fs;
280 	int i, blks, size, error;
281 
282 	if ((mountp->mnt_flag & MNT_RDONLY) == 0)
283 		return (EINVAL);
284 	/*
285 	 * Step 1: invalidate all cached meta-data.
286 	 */
287 	devvp = VFSTOUFS(mountp)->um_devvp;
288 	if (vinvalbuf(devvp, 0, cred, p, 0, 0))
289 		panic("ffs_reload: dirty1");
290 	/*
291 	 * Step 2: re-read superblock from disk.
292 	 */
293 	error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp);
294 	if (error)
295 		return (error);
296 	fs = (struct fs *)bp->b_data;
297 	if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE ||
298 	    fs->fs_bsize < sizeof(struct fs)) {
299 		brelse(bp);
300 		return (EIO);		/* XXX needs translation */
301 	}
302 	fs = VFSTOUFS(mountp)->um_fs;
303 	bcopy(&fs->fs_csp[0], &((struct fs *)bp->b_data)->fs_csp[0],
304 	    sizeof(fs->fs_csp));
305 	bcopy(bp->b_data, fs, (u_int)fs->fs_sbsize);
306 	if (fs->fs_sbsize < SBSIZE)
307 		bp->b_flags |= B_INVAL;
308 	brelse(bp);
309 	ffs_oldfscompat(fs);
310 	ffs_vmlimits(fs);
311 	/*
312 	 * Step 3: re-read summary information from disk.
313 	 */
314 	blks = howmany(fs->fs_cssize, fs->fs_fsize);
315 	space = fs->fs_csp[0];
316 	for (i = 0; i < blks; i += fs->fs_frag) {
317 		size = fs->fs_bsize;
318 		if (i + fs->fs_frag > blks)
319 			size = (blks - i) * fs->fs_fsize;
320 		error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
321 		    NOCRED, &bp);
322 		if (error)
323 			return (error);
324 		bcopy(bp->b_data, fs->fs_csp[fragstoblks(fs, i)], (u_int)size);
325 		brelse(bp);
326 	}
327 loop:
328 	for (vp = mountp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) {
329 		nvp = vp->v_mntvnodes.le_next;
330 		/*
331 		 * Step 4: invalidate all inactive vnodes.
332 		 */
333 		if (vp->v_usecount == 0) {
334 			vgone(vp);
335 			continue;
336 		}
337 		/*
338 		 * Step 5: invalidate all cached file data.
339 		 */
340 		if (vget(vp, 1))
341 			goto loop;
342 		if (vinvalbuf(vp, 0, cred, p, 0, 0))
343 			panic("ffs_reload: dirty2");
344 		/*
345 		 * Step 6: re-read inode data for all active vnodes.
346 		 */
347 		ip = VTOI(vp);
348 		error =
349 		    bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
350 		    (int)fs->fs_bsize, NOCRED, &bp);
351 		if (error) {
352 			vput(vp);
353 			return (error);
354 		}
355 		ip->i_din = *((struct dinode *)bp->b_data +
356 		    ino_to_fsbo(fs, ip->i_number));
357 		brelse(bp);
358 		vput(vp);
359 		if (vp->v_mount != mountp)
360 			goto loop;
361 	}
362 	return (0);
363 }
364 
365 /*
366  * Common code for mount and mountroot
367  */
368 int
369 ffs_mountfs(devvp, mp, p)
370 	register struct vnode *devvp;
371 	struct mount *mp;
372 	struct proc *p;
373 {
374 	register struct ufsmount *ump;
375 	struct buf *bp;
376 	register struct fs *fs;
377 	dev_t dev = devvp->v_rdev;
378 	struct partinfo dpart;
379 	caddr_t base, space;
380 	int havepart = 0, blks;
381 	int error, i, size;
382 	int ronly;
383 
384 	/*
385 	 * Disallow multiple mounts of the same device.
386 	 * Disallow mounting of a device that is currently in use
387 	 * (except for root, which might share swap device for miniroot).
388 	 * Flush out any old buffers remaining from a previous use.
389 	 */
390 	error = vfs_mountedon(devvp);
391 	if (error)
392 		return (error);
393 	if (vcount(devvp) > 1 && devvp != rootvp)
394 		return (EBUSY);
395 	error = vinvalbuf(devvp, V_SAVE, p->p_ucred, p, 0, 0);
396 	if (error)
397 		return (error);
398 
399 	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
400 	error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p);
401 	if (error)
402 		return (error);
403 	if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, p) != 0)
404 		size = DEV_BSIZE;
405 	else {
406 		havepart = 1;
407 		size = dpart.disklab->d_secsize;
408 	}
409 
410 	bp = NULL;
411 	ump = NULL;
412 	error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp);
413 	if (error)
414 		goto out;
415 	fs = (struct fs *)bp->b_data;
416 	if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE ||
417 	    fs->fs_bsize < sizeof(struct fs)) {
418 		error = EINVAL;		/* XXX needs translation */
419 		goto out;
420 	}
421 	if (!fs->fs_clean) {
422 		if (ronly || (mp->mnt_flag & MNT_FORCE)) {
423 			printf("WARNING: %s was not properly dismounted.\n",fs->fs_fsmnt);
424 		} else {
425 			printf("WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck.\n",fs->fs_fsmnt);
426 			error = EPERM;
427 			goto out;
428 		}
429 	}
430 	ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
431 	bzero((caddr_t)ump, sizeof *ump);
432 	ump->um_fs = malloc((u_long)fs->fs_sbsize, M_UFSMNT,
433 	    M_WAITOK);
434 	bcopy(bp->b_data, ump->um_fs, (u_int)fs->fs_sbsize);
435 	if (fs->fs_sbsize < SBSIZE)
436 		bp->b_flags |= B_INVAL;
437 	brelse(bp);
438 	bp = NULL;
439 	fs = ump->um_fs;
440 	fs->fs_ronly = ronly;
441 	if (ronly == 0) {
442 		fs->fs_fmod = 1;
443 		fs->fs_clean = 0;
444 	}
445 	blks = howmany(fs->fs_cssize, fs->fs_fsize);
446 	base = space = malloc((u_long)fs->fs_cssize, M_UFSMNT,
447 	    M_WAITOK);
448 	for (i = 0; i < blks; i += fs->fs_frag) {
449 		size = fs->fs_bsize;
450 		if (i + fs->fs_frag > blks)
451 			size = (blks - i) * fs->fs_fsize;
452 		error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
453 			NOCRED, &bp);
454 		if (error) {
455 			free(base, M_UFSMNT);
456 			goto out;
457 		}
458 		bcopy(bp->b_data, space, (u_int)size);
459 		fs->fs_csp[fragstoblks(fs, i)] = (struct csum *)space;
460 		space += size;
461 		brelse(bp);
462 		bp = NULL;
463 	}
464 	mp->mnt_data = (qaddr_t)ump;
465 	mp->mnt_stat.f_fsid.val[0] = (long)dev;
466 	mp->mnt_stat.f_fsid.val[1] = MOUNT_UFS;
467 	mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
468 	mp->mnt_flag |= MNT_LOCAL;
469 	ump->um_mountp = mp;
470 	ump->um_dev = dev;
471 	ump->um_devvp = devvp;
472 	ump->um_nindir = fs->fs_nindir;
473 	ump->um_bptrtodb = fs->fs_fsbtodb;
474 	ump->um_seqinc = fs->fs_frag;
475 	for (i = 0; i < MAXQUOTAS; i++)
476 		ump->um_quotas[i] = NULLVP;
477 	devvp->v_specflags |= SI_MOUNTEDON;
478 	ffs_oldfscompat(fs);
479 	ffs_vmlimits(fs);
480 	if (ronly == 0)
481 		ffs_sbupdate(ump, MNT_WAIT);
482 	return (0);
483 out:
484 	if (bp)
485 		brelse(bp);
486 	(void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, NOCRED, p);
487 	if (ump) {
488 		free(ump->um_fs, M_UFSMNT);
489 		free(ump, M_UFSMNT);
490 		mp->mnt_data = (qaddr_t)0;
491 	}
492 	return (error);
493 }
494 
495 /*
496  * Sanity checks for old file systems.
497  *
498  * XXX - goes away some day.
499  */
500 int
501 ffs_oldfscompat(fs)
502 	struct fs *fs;
503 {
504 	int i;
505 
506 	fs->fs_npsect = max(fs->fs_npsect, fs->fs_nsect);	/* XXX */
507 	fs->fs_interleave = max(fs->fs_interleave, 1);		/* XXX */
508 	if (fs->fs_postblformat == FS_42POSTBLFMT)		/* XXX */
509 		fs->fs_nrpos = 8;				/* XXX */
510 	if (fs->fs_inodefmt < FS_44INODEFMT) {			/* XXX */
511 		quad_t sizepb = fs->fs_bsize;			/* XXX */
512 								/* XXX */
513 #if 0
514 		fs->fs_maxfilesize = fs->fs_bsize * NDADDR - 1;	/* XXX */
515 		for (i = 0; i < NIADDR; i++) {			/* XXX */
516 			sizepb *= NINDIR(fs);			/* XXX */
517 			fs->fs_maxfilesize += sizepb;		/* XXX */
518 		}						/* XXX */
519 #endif
520 		fs->fs_maxfilesize = (u_quad_t) 1 << 39;
521 		fs->fs_qbmask = ~fs->fs_bmask;			/* XXX */
522 		fs->fs_qfmask = ~fs->fs_fmask;			/* XXX */
523 	}							/* XXX */
524 	return (0);
525 }
526 
527 /*
528  * Sanity check for VM file size limits -- temporary until
529  * VM system can support > 32bit offsets
530  */
531 void
532 ffs_vmlimits(fs)
533 	struct fs *fs;
534 {
535 	if( fs->fs_maxfilesize > (((u_quad_t) 1 << 31) - 1))
536 		fs->fs_maxfilesize = ((u_quad_t) 1 << 31) - 1;
537 }
538 
539 /*
540  * unmount system call
541  */
542 int
543 ffs_unmount(mp, mntflags, p)
544 	struct mount *mp;
545 	int mntflags;
546 	struct proc *p;
547 {
548 	register struct ufsmount *ump;
549 	register struct fs *fs;
550 	int error, flags, ronly;
551 
552 	flags = 0;
553 	if (mntflags & MNT_FORCE) {
554 		flags |= FORCECLOSE;
555 	}
556 	error = ffs_flushfiles(mp, flags, p);
557 	if (error)
558 		return (error);
559 	ump = VFSTOUFS(mp);
560 	fs = ump->um_fs;
561 	ronly = fs->fs_ronly;
562 	if (!ronly) {
563 		fs->fs_clean = 1;
564 		ffs_sbupdate(ump, MNT_WAIT);
565 	}
566 	ump->um_devvp->v_specflags &= ~SI_MOUNTEDON;
567 	error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD|FWRITE,
568 		NOCRED, p);
569 	vrele(ump->um_devvp);
570 	free(fs->fs_csp[0], M_UFSMNT);
571 	free(fs, M_UFSMNT);
572 	free(ump, M_UFSMNT);
573 	mp->mnt_data = (qaddr_t)0;
574 	mp->mnt_flag &= ~MNT_LOCAL;
575 	return (error);
576 }
577 
578 /*
579  * Flush out all the files in a filesystem.
580  */
581 int
582 ffs_flushfiles(mp, flags, p)
583 	register struct mount *mp;
584 	int flags;
585 	struct proc *p;
586 {
587 	register struct ufsmount *ump;
588 	int error;
589 
590 	if (!doforce)
591 		flags &= ~FORCECLOSE;
592 	ump = VFSTOUFS(mp);
593 #ifdef QUOTA
594 	if (mp->mnt_flag & MNT_QUOTA) {
595 		int i;
596 		error = vflush(mp, NULLVP, SKIPSYSTEM|flags);
597 		if (error)
598 			return (error);
599 		for (i = 0; i < MAXQUOTAS; i++) {
600 			if (ump->um_quotas[i] == NULLVP)
601 				continue;
602 			quotaoff(p, mp, i);
603 		}
604 		/*
605 		 * Here we fall through to vflush again to ensure
606 		 * that we have gotten rid of all the system vnodes.
607 		 */
608 	}
609 #endif
610 	error = vflush(mp, NULLVP, flags);
611 	return (error);
612 }
613 
614 /*
615  * Get file system statistics.
616  */
617 int
618 ffs_statfs(mp, sbp, p)
619 	struct mount *mp;
620 	register struct statfs *sbp;
621 	struct proc *p;
622 {
623 	register struct ufsmount *ump;
624 	register struct fs *fs;
625 
626 	ump = VFSTOUFS(mp);
627 	fs = ump->um_fs;
628 	if (fs->fs_magic != FS_MAGIC)
629 		panic("ffs_statfs");
630 	sbp->f_type = MOUNT_UFS;
631 	sbp->f_bsize = fs->fs_fsize;
632 	sbp->f_iosize = fs->fs_bsize;
633 	sbp->f_blocks = fs->fs_dsize;
634 	sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag +
635 		fs->fs_cstotal.cs_nffree;
636 	sbp->f_bavail = (fs->fs_dsize * (100 - fs->fs_minfree) / 100) -
637 		(fs->fs_dsize - sbp->f_bfree);
638 	sbp->f_files =  fs->fs_ncg * fs->fs_ipg - ROOTINO;
639 	sbp->f_ffree = fs->fs_cstotal.cs_nifree;
640 	if (sbp != &mp->mnt_stat) {
641 		bcopy((caddr_t)mp->mnt_stat.f_mntonname,
642 			(caddr_t)&sbp->f_mntonname[0], MNAMELEN);
643 		bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
644 			(caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
645 	}
646 	return (0);
647 }
648 
649 /*
650  * Go through the disk queues to initiate sandbagged IO;
651  * go through the inodes to write those that have been modified;
652  * initiate the writing of the super block if it has been modified.
653  *
654  * Note: we are always called with the filesystem marked `MPBUSY'.
655  */
656 int
657 ffs_sync(mp, waitfor, cred, p)
658 	struct mount *mp;
659 	int waitfor;
660 	struct ucred *cred;
661 	struct proc *p;
662 {
663 	register struct vnode *vp;
664 	register struct inode *ip;
665 	register struct ufsmount *ump = VFSTOUFS(mp);
666 	register struct fs *fs;
667 	struct timeval tv;
668 	int error, allerror = 0;
669 
670 	fs = ump->um_fs;
671 	/*
672 	 * Write back modified superblock.
673 	 * Consistency check that the superblock
674 	 * is still in the buffer cache.
675 	 */
676 	if (fs->fs_fmod != 0) {
677 		if (fs->fs_ronly != 0) {		/* XXX */
678 			printf("fs = %s\n", fs->fs_fsmnt);
679 			panic("update: rofs mod");
680 		}
681 		fs->fs_fmod = 0;
682 		fs->fs_time = time.tv_sec;
683 		allerror = ffs_sbupdate(ump, waitfor);
684 	}
685 	/*
686 	 * Write back each (modified) inode.
687 	 */
688 loop:
689 	for (vp = mp->mnt_vnodelist.lh_first;
690 	     vp != NULL;
691 	     vp = vp->v_mntvnodes.le_next) {
692 		/*
693 		 * If the vnode that we are about to sync is no longer
694 		 * associated with this mount point, start over.
695 		 */
696 		if (vp->v_mount != mp)
697 			goto loop;
698 		if (VOP_ISLOCKED(vp))
699 			continue;
700 		ip = VTOI(vp);
701 		if (vp->v_vmdata &&
702 		   (((vm_object_t) vp->v_vmdata)->flags & OBJ_WRITEABLE)) {
703 			if (vget(vp, 1))
704 				goto loop;
705 			_vm_object_page_clean( (vm_object_t) vp->v_vmdata,
706 						0, 0, 0);
707 			vput(vp);
708 		}
709 
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 (vp->v_type != VCHR) {
715 			if (vget(vp, 1))
716 				goto loop;
717 			error = VOP_FSYNC(vp, cred, waitfor, p);
718 			if (error)
719 				allerror = error;
720 			vput(vp);
721 		} else {
722 			tv = time;
723 			VOP_UPDATE(vp, &tv, &tv, waitfor == MNT_WAIT);
724 		}
725 	}
726 	/*
727 	 * Force stale file system control information to be flushed.
728 	 */
729 	error = VOP_FSYNC(ump->um_devvp, cred, waitfor, p);
730 	if (error)
731 		allerror = error;
732 #ifdef QUOTA
733 	qsync(mp);
734 #endif
735 	return (allerror);
736 }
737 
738 /*
739  * Look up a FFS dinode number to find its incore vnode, otherwise read it
740  * in from disk.  If it is in core, wait for the lock bit to clear, then
741  * return the inode locked.  Detection and handling of mount points must be
742  * done by the calling routine.
743  */
744 int
745 ffs_vget(mp, ino, vpp)
746 	struct mount *mp;
747 	ino_t ino;
748 	struct vnode **vpp;
749 {
750 	register struct fs *fs;
751 	register struct inode *ip;
752 	struct ufsmount *ump;
753 	struct buf *bp;
754 	struct vnode *vp;
755 	dev_t dev;
756 	int type, error;
757 
758 	ump = VFSTOUFS(mp);
759 	dev = ump->um_dev;
760 	if ((*vpp = ufs_ihashget(dev, ino)) != NULL)
761 		return (0);
762 
763 	/* Allocate a new vnode/inode. */
764 	error = getnewvnode(VT_UFS, mp, ffs_vnodeop_p, &vp);
765 	if (error) {
766 		*vpp = NULL;
767 		return (error);
768 	}
769 	type = ump->um_devvp->v_tag == VT_MFS ? M_MFSNODE : M_FFSNODE; /* XXX */
770 	MALLOC(ip, struct inode *, sizeof(struct inode), type, M_WAITOK);
771 	bzero((caddr_t)ip, sizeof(struct inode));
772 	vp->v_data = ip;
773 	ip->i_vnode = vp;
774 	ip->i_fs = fs = ump->um_fs;
775 	ip->i_dev = dev;
776 	ip->i_number = ino;
777 #ifdef QUOTA
778 	{
779 	int i;
780 	for (i = 0; i < MAXQUOTAS; i++)
781 		ip->i_dquot[i] = NODQUOT;
782 	}
783 #endif
784 	/*
785 	 * Put it onto its hash chain and lock it so that other requests for
786 	 * this inode will block if they arrive while we are sleeping waiting
787 	 * for old data structures to be purged or for the contents of the
788 	 * disk portion of this inode to be read.
789 	 */
790 	ufs_ihashins(ip);
791 
792 	/* Read in the disk contents for the inode, copy into the inode. */
793 	error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
794 	    (int)fs->fs_bsize, NOCRED, &bp);
795 	if (error) {
796 		/*
797 		 * The inode does not contain anything useful, so it would
798 		 * be misleading to leave it on its hash chain. With mode
799 		 * still zero, it will be unlinked and returned to the free
800 		 * list by vput().
801 		 */
802 		vput(vp);
803 		brelse(bp);
804 		*vpp = NULL;
805 		return (error);
806 	}
807 	ip->i_din = *((struct dinode *)bp->b_data + ino_to_fsbo(fs, ino));
808 	brelse(bp);
809 
810 	/*
811 	 * Initialize the vnode from the inode, check for aliases.
812 	 * Note that the underlying vnode may have changed.
813 	 */
814 	error = ufs_vinit(mp, ffs_specop_p, FFS_FIFOOPS, &vp);
815 	if (error) {
816 		vput(vp);
817 		*vpp = NULL;
818 		return (error);
819 	}
820 	/*
821 	 * Finish inode initialization now that aliasing has been resolved.
822 	 */
823 	ip->i_devvp = ump->um_devvp;
824 	VREF(ip->i_devvp);
825 	/*
826 	 * Set up a generation number for this inode if it does not
827 	 * already have one. This should only happen on old filesystems.
828 	 */
829 	if (ip->i_gen == 0) {
830 		if (++nextgennumber < (u_long)time.tv_sec)
831 			nextgennumber = time.tv_sec;
832 		ip->i_gen = nextgennumber;
833 		if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
834 			ip->i_flag |= IN_MODIFIED;
835 	}
836 	/*
837 	 * Ensure that uid and gid are correct. This is a temporary
838 	 * fix until fsck has been changed to do the update.
839 	 */
840 	if (fs->fs_inodefmt < FS_44INODEFMT) {		/* XXX */
841 		ip->i_uid = ip->i_din.di_ouid;		/* XXX */
842 		ip->i_gid = ip->i_din.di_ogid;		/* XXX */
843 	}						/* XXX */
844 
845 	*vpp = vp;
846 	return (0);
847 }
848 
849 /*
850  * File handle to vnode
851  *
852  * Have to be really careful about stale file handles:
853  * - check that the inode number is valid
854  * - call ffs_vget() to get the locked inode
855  * - check for an unallocated inode (i_mode == 0)
856  * - check that the given client host has export rights and return
857  *   those rights via. exflagsp and credanonp
858  */
859 int
860 ffs_fhtovp(mp, fhp, nam, vpp, exflagsp, credanonp)
861 	register struct mount *mp;
862 	struct fid *fhp;
863 	struct mbuf *nam;
864 	struct vnode **vpp;
865 	int *exflagsp;
866 	struct ucred **credanonp;
867 {
868 	register struct ufid *ufhp;
869 	struct fs *fs;
870 
871 	ufhp = (struct ufid *)fhp;
872 	fs = VFSTOUFS(mp)->um_fs;
873 	if (ufhp->ufid_ino < ROOTINO ||
874 	    ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg)
875 		return (ESTALE);
876 	return (ufs_check_export(mp, ufhp, nam, vpp, exflagsp, credanonp));
877 }
878 
879 /*
880  * Vnode pointer to File handle
881  */
882 /* ARGSUSED */
883 int
884 ffs_vptofh(vp, fhp)
885 	struct vnode *vp;
886 	struct fid *fhp;
887 {
888 	register struct inode *ip;
889 	register struct ufid *ufhp;
890 
891 	ip = VTOI(vp);
892 	ufhp = (struct ufid *)fhp;
893 	ufhp->ufid_len = sizeof(struct ufid);
894 	ufhp->ufid_ino = ip->i_number;
895 	ufhp->ufid_gen = ip->i_gen;
896 	return (0);
897 }
898 
899 /*
900  * Write a superblock and associated information back to disk.
901  */
902 int
903 ffs_sbupdate(mp, waitfor)
904 	struct ufsmount *mp;
905 	int waitfor;
906 {
907 	register struct fs *fs = mp->um_fs;
908 	register struct buf *bp;
909 	int blks;
910 	caddr_t space;
911 	int i, size, error = 0;
912 
913 	bp = getblk(mp->um_devvp, SBLOCK, (int)fs->fs_sbsize, 0, 0);
914 	bcopy((caddr_t)fs, bp->b_data, (u_int)fs->fs_sbsize);
915 	/* Restore compatibility to old file systems.		   XXX */
916 	if (fs->fs_postblformat == FS_42POSTBLFMT)		/* XXX */
917 		((struct fs *)bp->b_data)->fs_nrpos = -1;	/* XXX */
918 	if (waitfor == MNT_WAIT)
919 		error = bwrite(bp);
920 	else
921 		bawrite(bp);
922 	blks = howmany(fs->fs_cssize, fs->fs_fsize);
923 	space = (caddr_t)fs->fs_csp[0];
924 	for (i = 0; i < blks; i += fs->fs_frag) {
925 		size = fs->fs_bsize;
926 		if (i + fs->fs_frag > blks)
927 			size = (blks - i) * fs->fs_fsize;
928 		bp = getblk(mp->um_devvp, fsbtodb(fs, fs->fs_csaddr + i),
929 		    size, 0, 0);
930 		bcopy(space, bp->b_data, (u_int)size);
931 		space += size;
932 		if (waitfor == MNT_WAIT)
933 			error = bwrite(bp);
934 		else
935 			bawrite(bp);
936 	}
937 	return (error);
938 }
939