xref: /freebsd/sys/ufs/ffs/ffs_vfsops.c (revision ef5d438ed4bc17ad7ece3e40fe4d1f9baf3aadf7)
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 1.35 1996/01/19 03:59:14 dyson Exp $
35  */
36 
37 #include "opt_quota.h"
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/ufs_extern.h>
61 
62 #include <ufs/ffs/fs.h>
63 #include <ufs/ffs/ffs_extern.h>
64 
65 #include <vm/vm.h>
66 #include <vm/vm_param.h>
67 #include <vm/vm_prot.h>
68 #include <vm/vm_page.h>
69 #include <vm/vm_object.h>
70 
71 static int	ffs_sbupdate __P((struct ufsmount *, int));
72 static int	ffs_reload __P((struct mount *,struct ucred *,struct proc *));
73 static int	ffs_oldfscompat __P((struct fs *));
74 static int	ffs_mount __P((struct mount *,
75 	    char *, caddr_t, struct nameidata *, struct proc *));
76 
77 struct vfsops ufs_vfsops = {
78 	ffs_mount,
79 	ufs_start,
80 	ffs_unmount,
81 	ufs_root,
82 	ufs_quotactl,
83 	ffs_statfs,
84 	ffs_sync,
85 	ffs_vget,
86 	ffs_fhtovp,
87 	ffs_vptofh,
88 	ffs_init,
89 };
90 
91 VFS_SET(ufs_vfsops, ufs, MOUNT_UFS, 0);
92 
93 extern u_long nextgennumber;
94 
95 
96 /*
97  * ffs_mount
98  *
99  * Called when mounting local physical media
100  *
101  * PARAMETERS:
102  *		mountroot
103  *			mp	mount point structure
104  *			path	NULL (flag for root mount!!!)
105  *			data	<unused>
106  *			ndp	<unused>
107  *			p	process (user credentials check [statfs])
108  *
109  *		mount
110  *			mp	mount point structure
111  *			path	path to mount point
112  *			data	pointer to argument struct in user space
113  *			ndp	mount point namei() return (used for
114  *				credentials on reload), reused to look
115  *				up block device.
116  *			p	process (user credentials check)
117  *
118  * RETURNS:	0	Success
119  *		!0	error number (errno.h)
120  *
121  * LOCK STATE:
122  *
123  *		ENTRY
124  *			mount point is locked
125  *		EXIT
126  *			mount point is locked
127  *
128  * NOTES:
129  *		A NULL path can be used for a flag since the mount
130  *		system call will fail with EFAULT in copyinstr in
131  *		namei() if it is a genuine NULL from the user.
132  */
133 static int
134 ffs_mount( mp, path, data, ndp, p)
135         register struct mount	*mp;	/* mount struct pointer*/
136         char			*path;	/* path to mount point*/
137         caddr_t			data;	/* arguments to FS specific mount*/
138         struct nameidata	*ndp;	/* mount point credentials*/
139         struct proc		*p;	/* process requesting mount*/
140 {
141 	u_int		size;
142 	int		err = 0;
143 	struct vnode	*devvp;
144 
145 	struct ufs_args args;
146 	struct ufsmount *ump = 0;
147 	register struct fs *fs;
148 	int flags;
149 
150 	/*
151 	 * Use NULL path to flag a root mount
152 	 */
153 	if( path == NULL) {
154 		/*
155 		 ***
156 		 * Mounting root file system
157 		 ***
158 		 */
159 
160 		/* Get vnode for root device*/
161 		if( bdevvp( rootdev, &rootvp))
162 			panic("ffs_mountroot: can't setup bdevvp for root");
163 
164 		/*
165 		 * FS specific handling
166 		 */
167 		mp->mnt_flag |= MNT_RDONLY;	/* XXX globally applicable?*/
168 
169 		/*
170 		 * Attempt mount
171 		 */
172 		if( ( err = ffs_mountfs(rootvp, mp, p)) != 0) {
173 			/* fs specific cleanup (if any)*/
174 			goto error_1;
175 		}
176 
177 		goto dostatfs;		/* success*/
178 
179 	}
180 
181 	/*
182 	 ***
183 	 * Mounting non-root file system or updating a file system
184 	 ***
185 	 */
186 
187 	/* copy in user arguments*/
188 	err = copyin(data, (caddr_t)&args, sizeof (struct ufs_args));
189 	if (err)
190 		goto error_1;		/* can't get arguments*/
191 
192 	/*
193 	 * If updating, check whether changing from read-only to
194 	 * read/write; if there is no device name, that's all we do.
195 	 */
196 	if (mp->mnt_flag & MNT_UPDATE) {
197 		ump = VFSTOUFS(mp);
198 		fs = ump->um_fs;
199 		err = 0;
200 		if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
201 			flags = WRITECLOSE;
202 			if (mp->mnt_flag & MNT_FORCE)
203 				flags |= FORCECLOSE;
204 			if (vfs_busy(mp)) {
205 				err = EBUSY;
206 				goto error_1;
207 			}
208 			err = ffs_flushfiles(mp, flags, p);
209 			vfs_unbusy(mp);
210 		}
211 		if (!err && (mp->mnt_flag & MNT_RELOAD))
212 			err = ffs_reload(mp, ndp->ni_cnd.cn_cred, p);
213 		if (err) {
214 			goto error_1;
215 		}
216 		if (fs->fs_ronly && (mp->mnt_flag & MNT_WANTRDWR)) {
217 			if (!fs->fs_clean) {
218 				if (mp->mnt_flag & MNT_FORCE) {
219 					printf("WARNING: %s was not properly dismounted.\n",fs->fs_fsmnt);
220 				} else {
221 					printf("WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck.\n",
222 					    fs->fs_fsmnt);
223 					err = EPERM;
224 					goto error_1;
225 				}
226 			}
227 			fs->fs_ronly = 0;
228 		}
229 		if (fs->fs_ronly == 0) {
230 			fs->fs_clean = 0;
231 			ffs_sbupdate(ump, MNT_WAIT);
232 		}
233 		/* if not updating name...*/
234 		if (args.fspec == 0) {
235 			/*
236 			 * Process export requests.  Jumping to "success"
237 			 * will return the vfs_export() error code.
238 			 */
239 			err = vfs_export(mp, &ump->um_export, &args.export);
240 			goto success;
241 		}
242 	}
243 
244 	/*
245 	 * Not an update, or updating the name: look up the name
246 	 * and verify that it refers to a sensible block device.
247 	 */
248 	NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, p);
249 	err = namei(ndp);
250 	if (err) {
251 		/* can't get devvp!*/
252 		goto error_1;
253 	}
254 
255 	devvp = ndp->ni_vp;
256 
257 	if (devvp->v_type != VBLK) {
258 		err = ENOTBLK;
259 		goto error_2;
260 	}
261 	if (major(devvp->v_rdev) >= nblkdev) {
262 		err = ENXIO;
263 		goto error_2;
264 	}
265 	if (mp->mnt_flag & MNT_UPDATE) {
266 		/*
267 		 ********************
268 		 * UPDATE
269 		 ********************
270 		 */
271 
272 		if (devvp != ump->um_devvp)
273 			err = EINVAL;	/* needs translation */
274 		else
275 			vrele(devvp);
276 		/*
277 		 * Update device name only on success
278 		 */
279 		if( !err) {
280 			/* Save "mounted from" info for mount point (NULL pad)*/
281 			copyinstr(	args.fspec,
282 					mp->mnt_stat.f_mntfromname,
283 					MNAMELEN - 1,
284 					&size);
285 			bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
286 		}
287 	} else {
288 		/*
289 		 ********************
290 		 * NEW MOUNT
291 		 ********************
292 		 */
293 
294 		/*
295 		 * Since this is a new mount, we want the names for
296 		 * the device and the mount point copied in.  If an
297 		 * error occurs,  the mountpoint is discarded by the
298 		 * upper level code.
299 		 */
300 		/* Save "last mounted on" info for mount point (NULL pad)*/
301 		copyinstr(	path,				/* mount point*/
302 				mp->mnt_stat.f_mntonname,	/* save area*/
303 				MNAMELEN - 1,			/* max size*/
304 				&size);				/* real size*/
305 		bzero( mp->mnt_stat.f_mntonname + size, MNAMELEN - size);
306 
307 		/* Save "mounted from" info for mount point (NULL pad)*/
308 		copyinstr(	args.fspec,			/* device name*/
309 				mp->mnt_stat.f_mntfromname,	/* save area*/
310 				MNAMELEN - 1,			/* max size*/
311 				&size);				/* real size*/
312 		bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
313 
314 		err = ffs_mountfs(devvp, mp, p);
315 	}
316 	if (err) {
317 		goto error_2;
318 	}
319 
320 dostatfs:
321 	/*
322 	 * Initialize FS stat information in mount struct; uses both
323 	 * mp->mnt_stat.f_mntonname and mp->mnt_stat.f_mntfromname
324 	 *
325 	 * This code is common to root and non-root mounts
326 	 */
327 	(void)VFS_STATFS(mp, &mp->mnt_stat, p);
328 
329 	goto success;
330 
331 
332 error_2:	/* error with devvp held*/
333 
334 	/* release devvp before failing*/
335 	vrele(devvp);
336 
337 error_1:	/* no state to back out*/
338 
339 success:
340 	return( err);
341 }
342 
343 
344 /*
345  * Reload all incore data for a filesystem (used after running fsck on
346  * the root filesystem and finding things to fix). The filesystem must
347  * be mounted read-only.
348  *
349  * Things to do to update the mount:
350  *	1) invalidate all cached meta-data.
351  *	2) re-read superblock from disk.
352  *	3) re-read summary information from disk.
353  *	4) invalidate all inactive vnodes.
354  *	5) invalidate all cached file data.
355  *	6) re-read inode data for all active vnodes.
356  */
357 static int
358 ffs_reload(mp, cred, p)
359 	register struct mount *mp;
360 	struct ucred *cred;
361 	struct proc *p;
362 {
363 	register struct vnode *vp, *nvp, *devvp;
364 	struct inode *ip;
365 	struct csum *space;
366 	struct buf *bp;
367 	struct fs *fs;
368 	int i, blks, size, error;
369 
370 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
371 		return (EINVAL);
372 	/*
373 	 * Step 1: invalidate all cached meta-data.
374 	 */
375 	devvp = VFSTOUFS(mp)->um_devvp;
376 	if (vinvalbuf(devvp, 0, cred, p, 0, 0))
377 		panic("ffs_reload: dirty1");
378 	/*
379 	 * Step 2: re-read superblock from disk.
380 	 */
381 	error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp);
382 	if (error)
383 		return (error);
384 	fs = (struct fs *)bp->b_data;
385 	if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE ||
386 	    fs->fs_bsize < sizeof(struct fs)) {
387 		brelse(bp);
388 		return (EIO);		/* XXX needs translation */
389 	}
390 	fs = VFSTOUFS(mp)->um_fs;
391 	bcopy(&fs->fs_csp[0], &((struct fs *)bp->b_data)->fs_csp[0],
392 	    sizeof(fs->fs_csp));
393 	bcopy(bp->b_data, fs, (u_int)fs->fs_sbsize);
394 	if (fs->fs_sbsize < SBSIZE)
395 		bp->b_flags |= B_INVAL;
396 	brelse(bp);
397 	ffs_oldfscompat(fs);
398 	/*
399 	 * Step 3: re-read summary information from disk.
400 	 */
401 	blks = howmany(fs->fs_cssize, fs->fs_fsize);
402 	space = fs->fs_csp[0];
403 	for (i = 0; i < blks; i += fs->fs_frag) {
404 		size = fs->fs_bsize;
405 		if (i + fs->fs_frag > blks)
406 			size = (blks - i) * fs->fs_fsize;
407 		error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
408 		    NOCRED, &bp);
409 		if (error)
410 			return (error);
411 		bcopy(bp->b_data, fs->fs_csp[fragstoblks(fs, i)], (u_int)size);
412 		brelse(bp);
413 	}
414 loop:
415 	for (vp = mp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) {
416 		nvp = vp->v_mntvnodes.le_next;
417 		/*
418 		 * Step 4: invalidate all inactive vnodes.
419 		 */
420 		if (vp->v_usecount == 0) {
421 			vgone(vp);
422 			continue;
423 		}
424 		/*
425 		 * Step 5: invalidate all cached file data.
426 		 */
427 		if (vget(vp, 1))
428 			goto loop;
429 		if (vinvalbuf(vp, 0, cred, p, 0, 0))
430 			panic("ffs_reload: dirty2");
431 		/*
432 		 * Step 6: re-read inode data for all active vnodes.
433 		 */
434 		ip = VTOI(vp);
435 		error =
436 		    bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
437 		    (int)fs->fs_bsize, NOCRED, &bp);
438 		if (error) {
439 			vput(vp);
440 			return (error);
441 		}
442 		ip->i_din = *((struct dinode *)bp->b_data +
443 		    ino_to_fsbo(fs, ip->i_number));
444 		brelse(bp);
445 		vput(vp);
446 		if (vp->v_mount != mp)
447 			goto loop;
448 	}
449 	return (0);
450 }
451 
452 /*
453  * Common code for mount and mountroot
454  */
455 int
456 ffs_mountfs(devvp, mp, p)
457 	register struct vnode *devvp;
458 	struct mount *mp;
459 	struct proc *p;
460 {
461 	register struct ufsmount *ump;
462 	struct buf *bp;
463 	register struct fs *fs;
464 	dev_t dev = devvp->v_rdev;
465 	struct partinfo dpart;
466 	caddr_t base, space;
467 	int havepart = 0, blks;
468 	int error, i, size;
469 	int ronly;
470 	u_int strsize;
471 
472 	/*
473 	 * Disallow multiple mounts of the same device.
474 	 * Disallow mounting of a device that is currently in use
475 	 * (except for root, which might share swap device for miniroot).
476 	 * Flush out any old buffers remaining from a previous use.
477 	 */
478 	error = vfs_mountedon(devvp);
479 	if (error)
480 		return (error);
481 	if (vcount(devvp) > 1 && devvp != rootvp)
482 		return (EBUSY);
483 	error = vinvalbuf(devvp, V_SAVE, p->p_ucred, p, 0, 0);
484 	if (error)
485 		return (error);
486 
487 	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
488 	error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p);
489 	if (error)
490 		return (error);
491 	if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, p) != 0)
492 		size = DEV_BSIZE;
493 	else {
494 		havepart = 1;
495 		size = dpart.disklab->d_secsize;
496 	}
497 
498 	bp = NULL;
499 	ump = NULL;
500 	error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp);
501 	if (error)
502 		goto out;
503 	fs = (struct fs *)bp->b_data;
504 	if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE ||
505 	    fs->fs_bsize < sizeof(struct fs)) {
506 		error = EINVAL;		/* XXX needs translation */
507 		goto out;
508 	}
509 	if (!fs->fs_clean) {
510 		if (ronly || (mp->mnt_flag & MNT_FORCE)) {
511 			printf("WARNING: %s was not properly dismounted.\n",fs->fs_fsmnt);
512 		} else {
513 			printf("WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck.\n",fs->fs_fsmnt);
514 			error = EPERM;
515 			goto out;
516 		}
517 	}
518 	ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
519 	bzero((caddr_t)ump, sizeof *ump);
520 	ump->um_fs = malloc((u_long)fs->fs_sbsize, M_UFSMNT,
521 	    M_WAITOK);
522 	bcopy(bp->b_data, ump->um_fs, (u_int)fs->fs_sbsize);
523 	if (fs->fs_sbsize < SBSIZE)
524 		bp->b_flags |= B_INVAL;
525 	brelse(bp);
526 	bp = NULL;
527 	fs = ump->um_fs;
528 	fs->fs_ronly = ronly;
529 	if (ronly == 0) {
530 		fs->fs_fmod = 1;
531 		fs->fs_clean = 0;
532 	}
533 	blks = howmany(fs->fs_cssize, fs->fs_fsize);
534 	base = space = malloc((u_long)fs->fs_cssize, M_UFSMNT,
535 	    M_WAITOK);
536 	for (i = 0; i < blks; i += fs->fs_frag) {
537 		size = fs->fs_bsize;
538 		if (i + fs->fs_frag > blks)
539 			size = (blks - i) * fs->fs_fsize;
540 		error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
541 			NOCRED, &bp);
542 		if (error) {
543 			free(base, M_UFSMNT);
544 			goto out;
545 		}
546 		bcopy(bp->b_data, space, (u_int)size);
547 		fs->fs_csp[fragstoblks(fs, i)] = (struct csum *)space;
548 		space += size;
549 		brelse(bp);
550 		bp = NULL;
551 	}
552 	mp->mnt_data = (qaddr_t)ump;
553 	mp->mnt_stat.f_fsid.val[0] = (long)dev;
554 	mp->mnt_stat.f_fsid.val[1] = MOUNT_UFS;
555 	mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
556 	mp->mnt_flag |= MNT_LOCAL;
557 	ump->um_mountp = mp;
558 	ump->um_dev = dev;
559 	ump->um_devvp = devvp;
560 	ump->um_nindir = fs->fs_nindir;
561 	ump->um_bptrtodb = fs->fs_fsbtodb;
562 	ump->um_seqinc = fs->fs_frag;
563 	for (i = 0; i < MAXQUOTAS; i++)
564 		ump->um_quotas[i] = NULLVP;
565 	devvp->v_specflags |= SI_MOUNTEDON;
566 	ffs_oldfscompat(fs);
567 
568 	/*
569 	 * Set FS local "last mounted on" information (NULL pad)
570 	 */
571 	copystr(	mp->mnt_stat.f_mntonname,	/* mount point*/
572 			fs->fs_fsmnt,			/* copy area*/
573 			sizeof(fs->fs_fsmnt) - 1,	/* max size*/
574 			&strsize);			/* real size*/
575 	bzero( fs->fs_fsmnt + strsize, sizeof(fs->fs_fsmnt) - strsize);
576 
577 	if( mp->mnt_flag & MNT_ROOTFS) {
578 		/*
579 		 * Root mount; update timestamp in mount structure.
580 		 * this will be used by the common root mount code
581 		 * to update the system clock.
582 		 */
583 		mp->mnt_time = fs->fs_time;
584 	}
585 	if (ronly == 0)
586 		ffs_sbupdate(ump, MNT_WAIT);
587 	return (0);
588 out:
589 	if (bp)
590 		brelse(bp);
591 	(void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, NOCRED, p);
592 	if (ump) {
593 		free(ump->um_fs, M_UFSMNT);
594 		free(ump, M_UFSMNT);
595 		mp->mnt_data = (qaddr_t)0;
596 	}
597 	return (error);
598 }
599 
600 /*
601  * Sanity checks for old file systems.
602  *
603  * XXX - goes away some day.
604  */
605 static int
606 ffs_oldfscompat(fs)
607 	struct fs *fs;
608 {
609 
610 	fs->fs_npsect = max(fs->fs_npsect, fs->fs_nsect);	/* XXX */
611 	fs->fs_interleave = max(fs->fs_interleave, 1);		/* XXX */
612 	if (fs->fs_postblformat == FS_42POSTBLFMT)		/* XXX */
613 		fs->fs_nrpos = 8;				/* XXX */
614 	if (fs->fs_inodefmt < FS_44INODEFMT) {			/* XXX */
615 #if 0
616 		int i;						/* XXX */
617 		quad_t sizepb = fs->fs_bsize;			/* XXX */
618 		fs->fs_maxfilesize = fs->fs_bsize * NDADDR - 1;	/* XXX */
619 		for (i = 0; i < NIADDR; i++) {			/* XXX */
620 			sizepb *= NINDIR(fs);			/* XXX */
621 			fs->fs_maxfilesize += sizepb;		/* XXX */
622 		}						/* XXX */
623 #endif
624 		fs->fs_maxfilesize = (u_quad_t) 1LL << 39;
625 		fs->fs_qbmask = ~fs->fs_bmask;			/* XXX */
626 		fs->fs_qfmask = ~fs->fs_fmask;			/* XXX */
627 	}							/* XXX */
628 	return (0);
629 }
630 
631 /*
632  * unmount system call
633  */
634 int
635 ffs_unmount(mp, mntflags, p)
636 	struct mount *mp;
637 	int mntflags;
638 	struct proc *p;
639 {
640 	register struct ufsmount *ump;
641 	register struct fs *fs;
642 	int error, flags, ronly;
643 
644 	flags = 0;
645 	if (mntflags & MNT_FORCE) {
646 		flags |= FORCECLOSE;
647 	}
648 	error = ffs_flushfiles(mp, flags, p);
649 	if (error)
650 		return (error);
651 	ump = VFSTOUFS(mp);
652 	fs = ump->um_fs;
653 	ronly = fs->fs_ronly;
654 	if (!ronly) {
655 		fs->fs_clean = 1;
656 		ffs_sbupdate(ump, MNT_WAIT);
657 	}
658 	ump->um_devvp->v_specflags &= ~SI_MOUNTEDON;
659 	error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD|FWRITE,
660 		NOCRED, p);
661 /*
662 	vrele(ump->um_devvp);
663 */
664 	vn_vmio_close(ump->um_devvp);
665 	free(fs->fs_csp[0], M_UFSMNT);
666 	free(fs, M_UFSMNT);
667 	free(ump, M_UFSMNT);
668 	mp->mnt_data = (qaddr_t)0;
669 	mp->mnt_flag &= ~MNT_LOCAL;
670 	return (error);
671 }
672 
673 /*
674  * Flush out all the files in a filesystem.
675  */
676 int
677 ffs_flushfiles(mp, flags, p)
678 	register struct mount *mp;
679 	int flags;
680 	struct proc *p;
681 {
682 	register struct ufsmount *ump;
683 	int error;
684 
685 	if (!doforce)
686 		flags &= ~FORCECLOSE;
687 	ump = VFSTOUFS(mp);
688 #ifdef QUOTA
689 	if (mp->mnt_flag & MNT_QUOTA) {
690 		int i;
691 		error = vflush(mp, NULLVP, SKIPSYSTEM|flags);
692 		if (error)
693 			return (error);
694 		for (i = 0; i < MAXQUOTAS; i++) {
695 			if (ump->um_quotas[i] == NULLVP)
696 				continue;
697 			quotaoff(p, mp, i);
698 		}
699 		/*
700 		 * Here we fall through to vflush again to ensure
701 		 * that we have gotten rid of all the system vnodes.
702 		 */
703 	}
704 #endif
705 	error = vflush(mp, NULLVP, flags);
706 	return (error);
707 }
708 
709 /*
710  * Get file system statistics.
711  */
712 int
713 ffs_statfs(mp, sbp, p)
714 	struct mount *mp;
715 	register struct statfs *sbp;
716 	struct proc *p;
717 {
718 	register struct ufsmount *ump;
719 	register struct fs *fs;
720 
721 	ump = VFSTOUFS(mp);
722 	fs = ump->um_fs;
723 	if (fs->fs_magic != FS_MAGIC)
724 		panic("ffs_statfs");
725 	sbp->f_type = MOUNT_UFS;
726 	sbp->f_bsize = fs->fs_fsize;
727 	sbp->f_iosize = fs->fs_bsize;
728 	sbp->f_blocks = fs->fs_dsize;
729 	sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag +
730 		fs->fs_cstotal.cs_nffree;
731 	sbp->f_bavail = freespace(fs, fs->fs_minfree);
732 	sbp->f_files =  fs->fs_ncg * fs->fs_ipg - ROOTINO;
733 	sbp->f_ffree = fs->fs_cstotal.cs_nifree;
734 	if (sbp != &mp->mnt_stat) {
735 		bcopy((caddr_t)mp->mnt_stat.f_mntonname,
736 			(caddr_t)&sbp->f_mntonname[0], MNAMELEN);
737 		bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
738 			(caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
739 	}
740 	return (0);
741 }
742 
743 /*
744  * Go through the disk queues to initiate sandbagged IO;
745  * go through the inodes to write those that have been modified;
746  * initiate the writing of the super block if it has been modified.
747  *
748  * Note: we are always called with the filesystem marked `MPBUSY'.
749  */
750 int
751 ffs_sync(mp, waitfor, cred, p)
752 	struct mount *mp;
753 	int waitfor;
754 	struct ucred *cred;
755 	struct proc *p;
756 {
757 	register struct vnode *vp, *nvp;
758 	register struct inode *ip;
759 	register struct ufsmount *ump = VFSTOUFS(mp);
760 	register struct fs *fs;
761 	struct timeval tv;
762 	int error, allerror = 0;
763 
764 	fs = ump->um_fs;
765 	/*
766 	 * Write back modified superblock.
767 	 * Consistency check that the superblock
768 	 * is still in the buffer cache.
769 	 */
770 	if (fs->fs_fmod != 0) {
771 		if (fs->fs_ronly != 0) {		/* XXX */
772 			printf("fs = %s\n", fs->fs_fsmnt);
773 			panic("update: rofs mod");
774 		}
775 		fs->fs_fmod = 0;
776 		fs->fs_time = time.tv_sec;
777 		allerror = ffs_sbupdate(ump, waitfor);
778 	}
779 	/*
780 	 * Write back each (modified) inode.
781 	 */
782 loop:
783 	for (vp = mp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) {
784 		/*
785 		 * If the vnode that we are about to sync is no longer
786 		 * associated with this mount point, start over.
787 		 */
788 		if (vp->v_mount != mp)
789 			goto loop;
790 		nvp = vp->v_mntvnodes.le_next;
791 		if (VOP_ISLOCKED(vp))
792 			continue;
793 		ip = VTOI(vp);
794 		if ((((ip->i_flag &
795 		    (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0)) &&
796 		    vp->v_dirtyblkhd.lh_first == NULL)
797 			continue;
798 		if (vp->v_type != VCHR) {
799 			if (vget(vp, 1))
800 				goto loop;
801 			error = VOP_FSYNC(vp, cred, waitfor, p);
802 			if (error)
803 				allerror = error;
804 			vput(vp);
805 		} else {
806 			tv = time;
807 			/* VOP_UPDATE(vp, &tv, &tv, waitfor == MNT_WAIT); */
808 			VOP_UPDATE(vp, &tv, &tv, 0);
809 		}
810 	}
811 	/*
812 	 * Force stale file system control information to be flushed.
813 	 */
814 	error = VOP_FSYNC(ump->um_devvp, cred, waitfor, p);
815 	if (error)
816 		allerror = error;
817 #ifdef QUOTA
818 	qsync(mp);
819 #endif
820 	return (allerror);
821 }
822 
823 /*
824  * Look up a FFS dinode number to find its incore vnode, otherwise read it
825  * in from disk.  If it is in core, wait for the lock bit to clear, then
826  * return the inode locked.  Detection and handling of mount points must be
827  * done by the calling routine.
828  */
829 static int ffs_inode_hash_lock;
830 
831 int
832 ffs_vget(mp, ino, vpp)
833 	struct mount *mp;
834 	ino_t ino;
835 	struct vnode **vpp;
836 {
837 	register struct fs *fs;
838 	register struct inode *ip;
839 	struct ufsmount *ump;
840 	struct buf *bp;
841 	struct vnode *vp;
842 	dev_t dev;
843 	int type, error;
844 
845 	ump = VFSTOUFS(mp);
846 	dev = ump->um_dev;
847 restart:
848 	if ((*vpp = ufs_ihashget(dev, ino)) != NULL)
849 		return (0);
850 
851 	/*
852 	 * Lock out the creation of new entries in the FFS hash table in
853 	 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate
854 	 * may occur!
855 	 */
856 	if (ffs_inode_hash_lock) {
857 		while (ffs_inode_hash_lock) {
858 			ffs_inode_hash_lock = -1;
859 			tsleep(&ffs_inode_hash_lock, PVM, "ffsvgt", 0);
860 		}
861 		goto restart;
862 	}
863 	ffs_inode_hash_lock = 1;
864 
865 	/* Allocate a new vnode/inode. */
866 	error = getnewvnode(VT_UFS, mp, ffs_vnodeop_p, &vp);
867 	if (error) {
868 		if (ffs_inode_hash_lock < 0)
869 			wakeup(&ffs_inode_hash_lock);
870 		ffs_inode_hash_lock = 0;
871 		*vpp = NULL;
872 		return (error);
873 	}
874 	type = ump->um_devvp->v_tag == VT_MFS ? M_MFSNODE : M_FFSNODE; /* XXX */
875 	MALLOC(ip, struct inode *, sizeof(struct inode), type, M_WAITOK);
876 	bzero((caddr_t)ip, sizeof(struct inode));
877 	vp->v_data = ip;
878 	ip->i_vnode = vp;
879 	ip->i_fs = fs = ump->um_fs;
880 	ip->i_dev = dev;
881 	ip->i_number = ino;
882 #ifdef QUOTA
883 	{
884 		int i;
885 		for (i = 0; i < MAXQUOTAS; i++)
886 			ip->i_dquot[i] = NODQUOT;
887 	}
888 #endif
889 	/*
890 	 * Put it onto its hash chain and lock it so that other requests for
891 	 * this inode will block if they arrive while we are sleeping waiting
892 	 * for old data structures to be purged or for the contents of the
893 	 * disk portion of this inode to be read.
894 	 */
895 	ufs_ihashins(ip);
896 
897 	if (ffs_inode_hash_lock < 0)
898 		wakeup(&ffs_inode_hash_lock);
899 	ffs_inode_hash_lock = 0;
900 
901 	/* Read in the disk contents for the inode, copy into the inode. */
902 	error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
903 	    (int)fs->fs_bsize, NOCRED, &bp);
904 	if (error) {
905 		/*
906 		 * The inode does not contain anything useful, so it would
907 		 * be misleading to leave it on its hash chain. With mode
908 		 * still zero, it will be unlinked and returned to the free
909 		 * list by vput().
910 		 */
911 		brelse(bp);
912 		vput(vp);
913 		*vpp = NULL;
914 		return (error);
915 	}
916 	ip->i_din = *((struct dinode *)bp->b_data + ino_to_fsbo(fs, ino));
917 	bqrelse(bp);
918 
919 	/*
920 	 * Initialize the vnode from the inode, check for aliases.
921 	 * Note that the underlying vnode may have changed.
922 	 */
923 	error = ufs_vinit(mp, ffs_specop_p, ffs_fifoop_p, &vp);
924 	if (error) {
925 		vput(vp);
926 		*vpp = NULL;
927 		return (error);
928 	}
929 	/*
930 	 * Finish inode initialization now that aliasing has been resolved.
931 	 */
932 	ip->i_devvp = ump->um_devvp;
933 	VREF(ip->i_devvp);
934 	/*
935 	 * Set up a generation number for this inode if it does not
936 	 * already have one. This should only happen on old filesystems.
937 	 */
938 	if (ip->i_gen == 0) {
939 		if (++nextgennumber < (u_long)time.tv_sec)
940 			nextgennumber = time.tv_sec;
941 		ip->i_gen = nextgennumber;
942 		if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
943 			ip->i_flag |= IN_MODIFIED;
944 	}
945 	/*
946 	 * Ensure that uid and gid are correct. This is a temporary
947 	 * fix until fsck has been changed to do the update.
948 	 */
949 	if (fs->fs_inodefmt < FS_44INODEFMT) {		/* XXX */
950 		ip->i_uid = ip->i_din.di_ouid;		/* XXX */
951 		ip->i_gid = ip->i_din.di_ogid;		/* XXX */
952 	}						/* XXX */
953 
954 	*vpp = vp;
955 	return (0);
956 }
957 
958 /*
959  * File handle to vnode
960  *
961  * Have to be really careful about stale file handles:
962  * - check that the inode number is valid
963  * - call ffs_vget() to get the locked inode
964  * - check for an unallocated inode (i_mode == 0)
965  * - check that the given client host has export rights and return
966  *   those rights via. exflagsp and credanonp
967  */
968 int
969 ffs_fhtovp(mp, fhp, nam, vpp, exflagsp, credanonp)
970 	register struct mount *mp;
971 	struct fid *fhp;
972 	struct mbuf *nam;
973 	struct vnode **vpp;
974 	int *exflagsp;
975 	struct ucred **credanonp;
976 {
977 	register struct ufid *ufhp;
978 	struct fs *fs;
979 
980 	ufhp = (struct ufid *)fhp;
981 	fs = VFSTOUFS(mp)->um_fs;
982 	if (ufhp->ufid_ino < ROOTINO ||
983 	    ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg)
984 		return (ESTALE);
985 	return (ufs_check_export(mp, ufhp, nam, vpp, exflagsp, credanonp));
986 }
987 
988 /*
989  * Vnode pointer to File handle
990  */
991 /* ARGSUSED */
992 int
993 ffs_vptofh(vp, fhp)
994 	struct vnode *vp;
995 	struct fid *fhp;
996 {
997 	register struct inode *ip;
998 	register struct ufid *ufhp;
999 
1000 	ip = VTOI(vp);
1001 	ufhp = (struct ufid *)fhp;
1002 	ufhp->ufid_len = sizeof(struct ufid);
1003 	ufhp->ufid_ino = ip->i_number;
1004 	ufhp->ufid_gen = ip->i_gen;
1005 	return (0);
1006 }
1007 
1008 /*
1009  * Write a superblock and associated information back to disk.
1010  */
1011 static int
1012 ffs_sbupdate(mp, waitfor)
1013 	struct ufsmount *mp;
1014 	int waitfor;
1015 {
1016 	register struct fs *fs = mp->um_fs;
1017 	register struct buf *bp;
1018 	int blks;
1019 	caddr_t space;
1020 	int i, size, error = 0;
1021 
1022 	bp = getblk(mp->um_devvp, SBLOCK, (int)fs->fs_sbsize, 0, 0);
1023 	bcopy((caddr_t)fs, bp->b_data, (u_int)fs->fs_sbsize);
1024 	/* Restore compatibility to old file systems.		   XXX */
1025 	if (fs->fs_postblformat == FS_42POSTBLFMT)		/* XXX */
1026 		((struct fs *)bp->b_data)->fs_nrpos = -1;	/* XXX */
1027 	if (waitfor == MNT_WAIT)
1028 		error = bwrite(bp);
1029 	else
1030 		bawrite(bp);
1031 	blks = howmany(fs->fs_cssize, fs->fs_fsize);
1032 	space = (caddr_t)fs->fs_csp[0];
1033 	for (i = 0; i < blks; i += fs->fs_frag) {
1034 		size = fs->fs_bsize;
1035 		if (i + fs->fs_frag > blks)
1036 			size = (blks - i) * fs->fs_fsize;
1037 		bp = getblk(mp->um_devvp, fsbtodb(fs, fs->fs_csaddr + i),
1038 		    size, 0, 0);
1039 		bcopy(space, bp->b_data, (u_int)size);
1040 		space += size;
1041 		if (waitfor == MNT_WAIT)
1042 			error = bwrite(bp);
1043 		else
1044 			bawrite(bp);
1045 	}
1046 	return (error);
1047 }
1048