xref: /freebsd/sys/ufs/ffs/ffs_vfsops.c (revision bcd92649c9952c9c9e8845dbd34276a60dd16664)
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 1.41 1996/09/07 17:34:57 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 #include <vm/vm_extern.h>
71 
72 static int	ffs_sbupdate __P((struct ufsmount *, int));
73 static int	ffs_reload __P((struct mount *,struct ucred *,struct proc *));
74 static int	ffs_oldfscompat __P((struct fs *));
75 static int	ffs_mount __P((struct mount *,
76 	    char *, caddr_t, struct nameidata *, struct proc *));
77 
78 struct vfsops ufs_vfsops = {
79 	ffs_mount,
80 	ufs_start,
81 	ffs_unmount,
82 	ufs_root,
83 	ufs_quotactl,
84 	ffs_statfs,
85 	ffs_sync,
86 	ffs_vget,
87 	ffs_fhtovp,
88 	ffs_vptofh,
89 	ffs_init,
90 };
91 
92 VFS_SET(ufs_vfsops, ufs, MOUNT_UFS, 0);
93 
94 extern u_long nextgennumber;
95 
96 
97 /*
98  * ffs_mount
99  *
100  * Called when mounting local physical media
101  *
102  * PARAMETERS:
103  *		mountroot
104  *			mp	mount point structure
105  *			path	NULL (flag for root mount!!!)
106  *			data	<unused>
107  *			ndp	<unused>
108  *			p	process (user credentials check [statfs])
109  *
110  *		mount
111  *			mp	mount point structure
112  *			path	path to mount point
113  *			data	pointer to argument struct in user space
114  *			ndp	mount point namei() return (used for
115  *				credentials on reload), reused to look
116  *				up block device.
117  *			p	process (user credentials check)
118  *
119  * RETURNS:	0	Success
120  *		!0	error number (errno.h)
121  *
122  * LOCK STATE:
123  *
124  *		ENTRY
125  *			mount point is locked
126  *		EXIT
127  *			mount point is locked
128  *
129  * NOTES:
130  *		A NULL path can be used for a flag since the mount
131  *		system call will fail with EFAULT in copyinstr in
132  *		namei() if it is a genuine NULL from the user.
133  */
134 static int
135 ffs_mount( mp, path, data, ndp, p)
136         register struct mount	*mp;	/* mount struct pointer*/
137         char			*path;	/* path to mount point*/
138         caddr_t			data;	/* arguments to FS specific mount*/
139         struct nameidata	*ndp;	/* mount point credentials*/
140         struct proc		*p;	/* process requesting mount*/
141 {
142 	u_int		size;
143 	int		err = 0;
144 	struct vnode	*devvp;
145 
146 	struct ufs_args args;
147 	struct ufsmount *ump = 0;
148 	register struct fs *fs;
149 	int flags;
150 
151 	/*
152 	 * Use NULL path to flag a root mount
153 	 */
154 	if( path == NULL) {
155 		/*
156 		 ***
157 		 * Mounting root file system
158 		 ***
159 		 */
160 
161 		/* Get vnode for root device*/
162 		if( bdevvp( rootdev, &rootvp))
163 			panic("ffs_mountroot: can't setup bdevvp for root");
164 
165 		/*
166 		 * FS specific handling
167 		 */
168 		mp->mnt_flag |= MNT_RDONLY;	/* XXX globally applicable?*/
169 
170 		/*
171 		 * Attempt mount
172 		 */
173 		if( ( err = ffs_mountfs(rootvp, mp, p)) != 0) {
174 			/* fs specific cleanup (if any)*/
175 			goto error_1;
176 		}
177 
178 		goto dostatfs;		/* success*/
179 
180 	}
181 
182 	/*
183 	 ***
184 	 * Mounting non-root file system or updating a file system
185 	 ***
186 	 */
187 
188 	/* copy in user arguments*/
189 	err = copyin(data, (caddr_t)&args, sizeof (struct ufs_args));
190 	if (err)
191 		goto error_1;		/* can't get arguments*/
192 
193 	/*
194 	 * If updating, check whether changing from read-only to
195 	 * read/write; if there is no device name, that's all we do.
196 	 */
197 	if (mp->mnt_flag & MNT_UPDATE) {
198 		ump = VFSTOUFS(mp);
199 		fs = ump->um_fs;
200 		err = 0;
201 		if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
202 			flags = WRITECLOSE;
203 			if (mp->mnt_flag & MNT_FORCE)
204 				flags |= FORCECLOSE;
205 			if (vfs_busy(mp)) {
206 				err = EBUSY;
207 				goto error_1;
208 			}
209 			err = ffs_flushfiles(mp, flags, p);
210 			vfs_unbusy(mp);
211 		}
212 		if (!err && (mp->mnt_flag & MNT_RELOAD))
213 			err = ffs_reload(mp, ndp->ni_cnd.cn_cred, p);
214 		if (err) {
215 			goto error_1;
216 		}
217 		if (fs->fs_ronly && (mp->mnt_flag & MNT_WANTRDWR)) {
218 			if (!fs->fs_clean) {
219 				if (mp->mnt_flag & MNT_FORCE) {
220 					printf("WARNING: %s was not properly dismounted.\n",fs->fs_fsmnt);
221 				} else {
222 					printf("WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck.\n",
223 					    fs->fs_fsmnt);
224 					err = EPERM;
225 					goto error_1;
226 				}
227 			}
228 			fs->fs_ronly = 0;
229 		}
230 		if (fs->fs_ronly == 0) {
231 			fs->fs_clean = 0;
232 			ffs_sbupdate(ump, MNT_WAIT);
233 		}
234 		/* if not updating name...*/
235 		if (args.fspec == 0) {
236 			/*
237 			 * Process export requests.  Jumping to "success"
238 			 * will return the vfs_export() error code.
239 			 */
240 			err = vfs_export(mp, &ump->um_export, &args.export);
241 			goto success;
242 		}
243 	}
244 
245 	/*
246 	 * Not an update, or updating the name: look up the name
247 	 * and verify that it refers to a sensible block device.
248 	 */
249 	NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, p);
250 	err = namei(ndp);
251 	if (err) {
252 		/* can't get devvp!*/
253 		goto error_1;
254 	}
255 
256 	devvp = ndp->ni_vp;
257 
258 	if (devvp->v_type != VBLK) {
259 		err = ENOTBLK;
260 		goto error_2;
261 	}
262 	if (major(devvp->v_rdev) >= nblkdev) {
263 		err = ENXIO;
264 		goto error_2;
265 	}
266 	if (mp->mnt_flag & MNT_UPDATE) {
267 		/*
268 		 ********************
269 		 * UPDATE
270 		 ********************
271 		 */
272 
273 		if (devvp != ump->um_devvp)
274 			err = EINVAL;	/* needs translation */
275 		else
276 			vrele(devvp);
277 		/*
278 		 * Update device name only on success
279 		 */
280 		if( !err) {
281 			/* Save "mounted from" info for mount point (NULL pad)*/
282 			copyinstr(	args.fspec,
283 					mp->mnt_stat.f_mntfromname,
284 					MNAMELEN - 1,
285 					&size);
286 			bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
287 		}
288 	} else {
289 		/*
290 		 ********************
291 		 * NEW MOUNT
292 		 ********************
293 		 */
294 
295 		/*
296 		 * Since this is a new mount, we want the names for
297 		 * the device and the mount point copied in.  If an
298 		 * error occurs,  the mountpoint is discarded by the
299 		 * upper level code.
300 		 */
301 		/* Save "last mounted on" info for mount point (NULL pad)*/
302 		copyinstr(	path,				/* mount point*/
303 				mp->mnt_stat.f_mntonname,	/* save area*/
304 				MNAMELEN - 1,			/* max size*/
305 				&size);				/* real size*/
306 		bzero( mp->mnt_stat.f_mntonname + size, MNAMELEN - size);
307 
308 		/* Save "mounted from" info for mount point (NULL pad)*/
309 		copyinstr(	args.fspec,			/* device name*/
310 				mp->mnt_stat.f_mntfromname,	/* save area*/
311 				MNAMELEN - 1,			/* max size*/
312 				&size);				/* real size*/
313 		bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
314 
315 		err = ffs_mountfs(devvp, mp, p);
316 	}
317 	if (err) {
318 		goto error_2;
319 	}
320 
321 dostatfs:
322 	/*
323 	 * Initialize FS stat information in mount struct; uses both
324 	 * mp->mnt_stat.f_mntonname and mp->mnt_stat.f_mntfromname
325 	 *
326 	 * This code is common to root and non-root mounts
327 	 */
328 	(void)VFS_STATFS(mp, &mp->mnt_stat, p);
329 
330 	goto success;
331 
332 
333 error_2:	/* error with devvp held*/
334 
335 	/* release devvp before failing*/
336 	vrele(devvp);
337 
338 error_1:	/* no state to back out*/
339 
340 success:
341 	return( err);
342 }
343 
344 
345 /*
346  * Reload all incore data for a filesystem (used after running fsck on
347  * the root filesystem and finding things to fix). The filesystem must
348  * be mounted read-only.
349  *
350  * Things to do to update the mount:
351  *	1) invalidate all cached meta-data.
352  *	2) re-read superblock from disk.
353  *	3) re-read summary information from disk.
354  *	4) invalidate all inactive vnodes.
355  *	5) invalidate all cached file data.
356  *	6) re-read inode data for all active vnodes.
357  */
358 static int
359 ffs_reload(mp, cred, p)
360 	register struct mount *mp;
361 	struct ucred *cred;
362 	struct proc *p;
363 {
364 	register struct vnode *vp, *nvp, *devvp;
365 	struct inode *ip;
366 	struct csum *space;
367 	struct buf *bp;
368 	struct fs *fs;
369 	int i, blks, size, error;
370 
371 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
372 		return (EINVAL);
373 	/*
374 	 * Step 1: invalidate all cached meta-data.
375 	 */
376 	devvp = VFSTOUFS(mp)->um_devvp;
377 	if (vinvalbuf(devvp, 0, cred, p, 0, 0))
378 		panic("ffs_reload: dirty1");
379 	/*
380 	 * Step 2: re-read superblock from disk.
381 	 */
382 	error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp);
383 	if (error)
384 		return (error);
385 	fs = (struct fs *)bp->b_data;
386 	fs->fs_fmod = 0;
387 	if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE ||
388 	    fs->fs_bsize < sizeof(struct fs)) {
389 		brelse(bp);
390 		return (EIO);		/* XXX needs translation */
391 	}
392 	fs = VFSTOUFS(mp)->um_fs;
393 	bcopy(&fs->fs_csp[0], &((struct fs *)bp->b_data)->fs_csp[0],
394 	    sizeof(fs->fs_csp));
395 	bcopy(bp->b_data, fs, (u_int)fs->fs_sbsize);
396 	if (fs->fs_sbsize < SBSIZE)
397 		bp->b_flags |= B_INVAL;
398 	brelse(bp);
399 	ffs_oldfscompat(fs);
400 	/*
401 	 * Step 3: re-read summary information from disk.
402 	 */
403 	blks = howmany(fs->fs_cssize, fs->fs_fsize);
404 	space = fs->fs_csp[0];
405 	for (i = 0; i < blks; i += fs->fs_frag) {
406 		size = fs->fs_bsize;
407 		if (i + fs->fs_frag > blks)
408 			size = (blks - i) * fs->fs_fsize;
409 		error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
410 		    NOCRED, &bp);
411 		if (error)
412 			return (error);
413 		bcopy(bp->b_data, fs->fs_csp[fragstoblks(fs, i)], (u_int)size);
414 		brelse(bp);
415 	}
416 loop:
417 	for (vp = mp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) {
418 		nvp = vp->v_mntvnodes.le_next;
419 		/*
420 		 * Step 4: invalidate all inactive vnodes.
421 		 */
422 		if (vp->v_usecount == 0) {
423 			vgone(vp);
424 			continue;
425 		}
426 		/*
427 		 * Step 5: invalidate all cached file data.
428 		 */
429 		if (vget(vp, 1))
430 			goto loop;
431 		if (vinvalbuf(vp, 0, cred, p, 0, 0))
432 			panic("ffs_reload: dirty2");
433 		/*
434 		 * Step 6: re-read inode data for all active vnodes.
435 		 */
436 		ip = VTOI(vp);
437 		error =
438 		    bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
439 		    (int)fs->fs_bsize, NOCRED, &bp);
440 		if (error) {
441 			vput(vp);
442 			return (error);
443 		}
444 		ip->i_din = *((struct dinode *)bp->b_data +
445 		    ino_to_fsbo(fs, ip->i_number));
446 		brelse(bp);
447 		vput(vp);
448 		if (vp->v_mount != mp)
449 			goto loop;
450 	}
451 	return (0);
452 }
453 
454 /*
455  * Common code for mount and mountroot
456  */
457 int
458 ffs_mountfs(devvp, mp, p)
459 	register struct vnode *devvp;
460 	struct mount *mp;
461 	struct proc *p;
462 {
463 	register struct ufsmount *ump;
464 	struct buf *bp;
465 	register struct fs *fs;
466 	dev_t dev = devvp->v_rdev;
467 	struct partinfo dpart;
468 	caddr_t base, space;
469 	int havepart = 0, blks;
470 	int error, i, size;
471 	int ronly;
472 	u_int strsize;
473 	int ncount;
474 
475 	/*
476 	 * Disallow multiple mounts of the same device.
477 	 * Disallow mounting of a device that is currently in use
478 	 * (except for root, which might share swap device for miniroot).
479 	 * Flush out any old buffers remaining from a previous use.
480 	 */
481 	error = vfs_mountedon(devvp);
482 	if (error)
483 		return (error);
484 	ncount = vcount(devvp);
485 	if (devvp->v_object)
486 		ncount -= 1;
487 	if (ncount > 1 && devvp != rootvp)
488 		return (EBUSY);
489 	error = vinvalbuf(devvp, V_SAVE, p->p_ucred, p, 0, 0);
490 	if (error)
491 		return (error);
492 
493 	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
494 	error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p);
495 	if (error)
496 		return (error);
497 	if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, p) != 0)
498 		size = DEV_BSIZE;
499 	else {
500 		havepart = 1;
501 		size = dpart.disklab->d_secsize;
502 	}
503 
504 	bp = NULL;
505 	ump = NULL;
506 	error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp);
507 	if (error)
508 		goto out;
509 	fs = (struct fs *)bp->b_data;
510 	if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE ||
511 	    fs->fs_bsize < sizeof(struct fs)) {
512 		error = EINVAL;		/* XXX needs translation */
513 		goto out;
514 	}
515 	fs->fs_fmod = 0;
516 	if (!fs->fs_clean) {
517 		if (ronly || (mp->mnt_flag & MNT_FORCE)) {
518 			printf("WARNING: %s was not properly dismounted.\n",fs->fs_fsmnt);
519 		} else {
520 			printf("WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck.\n",fs->fs_fsmnt);
521 			error = EPERM;
522 			goto out;
523 		}
524 	}
525 	ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
526 	bzero((caddr_t)ump, sizeof *ump);
527 	ump->um_fs = malloc((u_long)fs->fs_sbsize, M_UFSMNT,
528 	    M_WAITOK);
529 	bcopy(bp->b_data, ump->um_fs, (u_int)fs->fs_sbsize);
530 	if (fs->fs_sbsize < SBSIZE)
531 		bp->b_flags |= B_INVAL;
532 	brelse(bp);
533 	bp = NULL;
534 	fs = ump->um_fs;
535 	fs->fs_ronly = ronly;
536 	if (ronly == 0) {
537 		fs->fs_fmod = 1;
538 		fs->fs_clean = 0;
539 	}
540 	blks = howmany(fs->fs_cssize, fs->fs_fsize);
541 	base = space = malloc((u_long)fs->fs_cssize, M_UFSMNT,
542 	    M_WAITOK);
543 	for (i = 0; i < blks; i += fs->fs_frag) {
544 		size = fs->fs_bsize;
545 		if (i + fs->fs_frag > blks)
546 			size = (blks - i) * fs->fs_fsize;
547 		error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
548 			NOCRED, &bp);
549 		if (error) {
550 			free(base, M_UFSMNT);
551 			goto out;
552 		}
553 		bcopy(bp->b_data, space, (u_int)size);
554 		fs->fs_csp[fragstoblks(fs, i)] = (struct csum *)space;
555 		space += size;
556 		brelse(bp);
557 		bp = NULL;
558 	}
559 	mp->mnt_data = (qaddr_t)ump;
560 	mp->mnt_stat.f_fsid.val[0] = (long)dev;
561 	mp->mnt_stat.f_fsid.val[1] = MOUNT_UFS;
562 	mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
563 	mp->mnt_flag |= MNT_LOCAL;
564 	ump->um_mountp = mp;
565 	ump->um_dev = dev;
566 	ump->um_devvp = devvp;
567 	ump->um_nindir = fs->fs_nindir;
568 	ump->um_bptrtodb = fs->fs_fsbtodb;
569 	ump->um_seqinc = fs->fs_frag;
570 	for (i = 0; i < MAXQUOTAS; i++)
571 		ump->um_quotas[i] = NULLVP;
572 	devvp->v_specflags |= SI_MOUNTEDON;
573 	ffs_oldfscompat(fs);
574 
575 	/*
576 	 * Set FS local "last mounted on" information (NULL pad)
577 	 */
578 	copystr(	mp->mnt_stat.f_mntonname,	/* mount point*/
579 			fs->fs_fsmnt,			/* copy area*/
580 			sizeof(fs->fs_fsmnt) - 1,	/* max size*/
581 			&strsize);			/* real size*/
582 	bzero( fs->fs_fsmnt + strsize, sizeof(fs->fs_fsmnt) - strsize);
583 
584 	if( mp->mnt_flag & MNT_ROOTFS) {
585 		/*
586 		 * Root mount; update timestamp in mount structure.
587 		 * this will be used by the common root mount code
588 		 * to update the system clock.
589 		 */
590 		mp->mnt_time = fs->fs_time;
591 	}
592 	if (ronly == 0)
593 		ffs_sbupdate(ump, MNT_WAIT);
594 	/*
595 	 * Only VMIO the backing device if the backing device is a real
596 	 * block device.  This excludes the original MFS implementation.
597 	 * Note that it is optional that the backing device be VMIOed.  This
598 	 * increases the opportunity for metadata caching.
599 	 */
600 	if ((devvp->v_type == VBLK) && (major(devvp->v_rdev) < nblkdev)) {
601 		vfs_object_create(devvp, p, p->p_ucred, 0);
602 	}
603 	return (0);
604 out:
605 	if (bp)
606 		brelse(bp);
607 	(void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, NOCRED, p);
608 	if (ump) {
609 		free(ump->um_fs, M_UFSMNT);
610 		free(ump, M_UFSMNT);
611 		mp->mnt_data = (qaddr_t)0;
612 	}
613 	return (error);
614 }
615 
616 /*
617  * Sanity checks for old file systems.
618  *
619  * XXX - goes away some day.
620  */
621 static int
622 ffs_oldfscompat(fs)
623 	struct fs *fs;
624 {
625 
626 	fs->fs_npsect = max(fs->fs_npsect, fs->fs_nsect);	/* XXX */
627 	fs->fs_interleave = max(fs->fs_interleave, 1);		/* XXX */
628 	if (fs->fs_postblformat == FS_42POSTBLFMT)		/* XXX */
629 		fs->fs_nrpos = 8;				/* XXX */
630 	if (fs->fs_inodefmt < FS_44INODEFMT) {			/* XXX */
631 #if 0
632 		int i;						/* XXX */
633 		quad_t sizepb = fs->fs_bsize;			/* XXX */
634 		fs->fs_maxfilesize = fs->fs_bsize * NDADDR - 1;	/* XXX */
635 		for (i = 0; i < NIADDR; i++) {			/* XXX */
636 			sizepb *= NINDIR(fs);			/* XXX */
637 			fs->fs_maxfilesize += sizepb;		/* XXX */
638 		}						/* XXX */
639 #endif
640 		fs->fs_maxfilesize = (u_quad_t) 1LL << 39;
641 		fs->fs_qbmask = ~fs->fs_bmask;			/* XXX */
642 		fs->fs_qfmask = ~fs->fs_fmask;			/* XXX */
643 	}							/* XXX */
644 	return (0);
645 }
646 
647 /*
648  * unmount system call
649  */
650 int
651 ffs_unmount(mp, mntflags, p)
652 	struct mount *mp;
653 	int mntflags;
654 	struct proc *p;
655 {
656 	register struct ufsmount *ump;
657 	register struct fs *fs;
658 	int error, flags, ronly;
659 
660 	flags = 0;
661 	if (mntflags & MNT_FORCE) {
662 		flags |= FORCECLOSE;
663 	}
664 	error = ffs_flushfiles(mp, flags, p);
665 	if (error)
666 		return (error);
667 	ump = VFSTOUFS(mp);
668 	fs = ump->um_fs;
669 	ronly = fs->fs_ronly;
670 	if (!ronly) {
671 		fs->fs_clean = 1;
672 		ffs_sbupdate(ump, MNT_WAIT);
673 	}
674 	ump->um_devvp->v_specflags &= ~SI_MOUNTEDON;
675 
676 	VOP_LOCK(ump->um_devvp);
677 	vnode_pager_uncache(ump->um_devvp);
678 	VOP_UNLOCK(ump->um_devvp);
679 
680 	error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD|FWRITE,
681 		NOCRED, p);
682 
683 	vrele(ump->um_devvp);
684 
685 	free(fs->fs_csp[0], M_UFSMNT);
686 	free(fs, M_UFSMNT);
687 	free(ump, M_UFSMNT);
688 	mp->mnt_data = (qaddr_t)0;
689 	mp->mnt_flag &= ~MNT_LOCAL;
690 	return (error);
691 }
692 
693 /*
694  * Flush out all the files in a filesystem.
695  */
696 int
697 ffs_flushfiles(mp, flags, p)
698 	register struct mount *mp;
699 	int flags;
700 	struct proc *p;
701 {
702 	register struct ufsmount *ump;
703 	int error;
704 
705 	if (!doforce)
706 		flags &= ~FORCECLOSE;
707 	ump = VFSTOUFS(mp);
708 #ifdef QUOTA
709 	if (mp->mnt_flag & MNT_QUOTA) {
710 		int i;
711 		error = vflush(mp, NULLVP, SKIPSYSTEM|flags);
712 		if (error)
713 			return (error);
714 		for (i = 0; i < MAXQUOTAS; i++) {
715 			if (ump->um_quotas[i] == NULLVP)
716 				continue;
717 			quotaoff(p, mp, i);
718 		}
719 		/*
720 		 * Here we fall through to vflush again to ensure
721 		 * that we have gotten rid of all the system vnodes.
722 		 */
723 	}
724 #endif
725 	error = vflush(mp, NULLVP, flags);
726 	return (error);
727 }
728 
729 /*
730  * Get file system statistics.
731  */
732 int
733 ffs_statfs(mp, sbp, p)
734 	struct mount *mp;
735 	register struct statfs *sbp;
736 	struct proc *p;
737 {
738 	register struct ufsmount *ump;
739 	register struct fs *fs;
740 
741 	ump = VFSTOUFS(mp);
742 	fs = ump->um_fs;
743 	if (fs->fs_magic != FS_MAGIC)
744 		panic("ffs_statfs");
745 	sbp->f_type = MOUNT_UFS;
746 	sbp->f_bsize = fs->fs_fsize;
747 	sbp->f_iosize = fs->fs_bsize;
748 	sbp->f_blocks = fs->fs_dsize;
749 	sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag +
750 		fs->fs_cstotal.cs_nffree;
751 	sbp->f_bavail = freespace(fs, fs->fs_minfree);
752 	sbp->f_files =  fs->fs_ncg * fs->fs_ipg - ROOTINO;
753 	sbp->f_ffree = fs->fs_cstotal.cs_nifree;
754 	if (sbp != &mp->mnt_stat) {
755 		bcopy((caddr_t)mp->mnt_stat.f_mntonname,
756 			(caddr_t)&sbp->f_mntonname[0], MNAMELEN);
757 		bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
758 			(caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
759 	}
760 	return (0);
761 }
762 
763 /*
764  * Go through the disk queues to initiate sandbagged IO;
765  * go through the inodes to write those that have been modified;
766  * initiate the writing of the super block if it has been modified.
767  *
768  * Note: we are always called with the filesystem marked `MPBUSY'.
769  */
770 int
771 ffs_sync(mp, waitfor, cred, p)
772 	struct mount *mp;
773 	int waitfor;
774 	struct ucred *cred;
775 	struct proc *p;
776 {
777 	register struct vnode *vp, *nvp;
778 	register struct inode *ip;
779 	register struct ufsmount *ump = VFSTOUFS(mp);
780 	register struct fs *fs;
781 	struct timeval tv;
782 	int error, allerror = 0;
783 
784 	fs = ump->um_fs;
785 	/*
786 	 * Write back modified superblock.
787 	 * Consistency check that the superblock
788 	 * is still in the buffer cache.
789 	 */
790 	if (fs->fs_fmod != 0) {
791 		if (fs->fs_ronly != 0) {		/* XXX */
792 			printf("fs = %s\n", fs->fs_fsmnt);
793 			panic("update: rofs mod");
794 		}
795 		fs->fs_fmod = 0;
796 		fs->fs_time = time.tv_sec;
797 		allerror = ffs_sbupdate(ump, waitfor);
798 	}
799 	/*
800 	 * Write back each (modified) inode.
801 	 */
802 loop:
803 	for (vp = mp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) {
804 		/*
805 		 * If the vnode that we are about to sync is no longer
806 		 * associated with this mount point, start over.
807 		 */
808 		if (vp->v_mount != mp)
809 			goto loop;
810 		nvp = vp->v_mntvnodes.le_next;
811 		if (VOP_ISLOCKED(vp))
812 			continue;
813 		ip = VTOI(vp);
814 		if ((((ip->i_flag &
815 		    (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0)) &&
816 		    vp->v_dirtyblkhd.lh_first == NULL)
817 			continue;
818 		if (vp->v_type != VCHR) {
819 			if (vget(vp, 1))
820 				goto loop;
821 			error = VOP_FSYNC(vp, cred, waitfor, p);
822 			if (error)
823 				allerror = error;
824 			vput(vp);
825 		} else {
826 			tv = time;
827 			/* VOP_UPDATE(vp, &tv, &tv, waitfor == MNT_WAIT); */
828 			VOP_UPDATE(vp, &tv, &tv, 0);
829 		}
830 	}
831 	/*
832 	 * Force stale file system control information to be flushed.
833 	 */
834 	error = VOP_FSYNC(ump->um_devvp, cred, waitfor, p);
835 	if (error)
836 		allerror = error;
837 #ifdef QUOTA
838 	qsync(mp);
839 #endif
840 	return (allerror);
841 }
842 
843 /*
844  * Look up a FFS dinode number to find its incore vnode, otherwise read it
845  * in from disk.  If it is in core, wait for the lock bit to clear, then
846  * return the inode locked.  Detection and handling of mount points must be
847  * done by the calling routine.
848  */
849 static int ffs_inode_hash_lock;
850 
851 int
852 ffs_vget(mp, ino, vpp)
853 	struct mount *mp;
854 	ino_t ino;
855 	struct vnode **vpp;
856 {
857 	register struct fs *fs;
858 	register struct inode *ip;
859 	struct ufsmount *ump;
860 	struct buf *bp;
861 	struct vnode *vp;
862 	dev_t dev;
863 	int type, error;
864 
865 	ump = VFSTOUFS(mp);
866 	dev = ump->um_dev;
867 restart:
868 	if ((*vpp = ufs_ihashget(dev, ino)) != NULL)
869 		return (0);
870 
871 	/*
872 	 * Lock out the creation of new entries in the FFS hash table in
873 	 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate
874 	 * may occur!
875 	 */
876 	if (ffs_inode_hash_lock) {
877 		while (ffs_inode_hash_lock) {
878 			ffs_inode_hash_lock = -1;
879 			tsleep(&ffs_inode_hash_lock, PVM, "ffsvgt", 0);
880 		}
881 		goto restart;
882 	}
883 	ffs_inode_hash_lock = 1;
884 
885 	/*
886 	 * If this MALLOC() is performed after the getnewvnode()
887 	 * it might block, leaving a vnode with a NULL v_data to be
888 	 * found by ffs_sync() if a sync happens to fire right then,
889 	 * which will cause a panic because ffs_sync() blindly
890 	 * dereferences vp->v_data (as well it should).
891 	 */
892 	type = ump->um_devvp->v_tag == VT_MFS ? M_MFSNODE : M_FFSNODE; /* XXX */
893 	MALLOC(ip, struct inode *, sizeof(struct inode), type, M_WAITOK);
894 
895 	/* Allocate a new vnode/inode. */
896 	error = getnewvnode(VT_UFS, mp, ffs_vnodeop_p, &vp);
897 	if (error) {
898 		if (ffs_inode_hash_lock < 0)
899 			wakeup(&ffs_inode_hash_lock);
900 		ffs_inode_hash_lock = 0;
901 		*vpp = NULL;
902 		FREE(ip, type);
903 		return (error);
904 	}
905 	bzero((caddr_t)ip, sizeof(struct inode));
906 	vp->v_data = ip;
907 	ip->i_vnode = vp;
908 	ip->i_fs = fs = ump->um_fs;
909 	ip->i_dev = dev;
910 	ip->i_number = ino;
911 #ifdef QUOTA
912 	{
913 		int i;
914 		for (i = 0; i < MAXQUOTAS; i++)
915 			ip->i_dquot[i] = NODQUOT;
916 	}
917 #endif
918 	/*
919 	 * Put it onto its hash chain and lock it so that other requests for
920 	 * this inode will block if they arrive while we are sleeping waiting
921 	 * for old data structures to be purged or for the contents of the
922 	 * disk portion of this inode to be read.
923 	 */
924 	ufs_ihashins(ip);
925 
926 	if (ffs_inode_hash_lock < 0)
927 		wakeup(&ffs_inode_hash_lock);
928 	ffs_inode_hash_lock = 0;
929 
930 	/* Read in the disk contents for the inode, copy into the inode. */
931 	error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
932 	    (int)fs->fs_bsize, NOCRED, &bp);
933 	if (error) {
934 		/*
935 		 * The inode does not contain anything useful, so it would
936 		 * be misleading to leave it on its hash chain. With mode
937 		 * still zero, it will be unlinked and returned to the free
938 		 * list by vput().
939 		 */
940 		brelse(bp);
941 		vput(vp);
942 		*vpp = NULL;
943 		return (error);
944 	}
945 	ip->i_din = *((struct dinode *)bp->b_data + ino_to_fsbo(fs, ino));
946 	bqrelse(bp);
947 
948 	/*
949 	 * Initialize the vnode from the inode, check for aliases.
950 	 * Note that the underlying vnode may have changed.
951 	 */
952 	error = ufs_vinit(mp, ffs_specop_p, ffs_fifoop_p, &vp);
953 	if (error) {
954 		vput(vp);
955 		*vpp = NULL;
956 		return (error);
957 	}
958 	/*
959 	 * Finish inode initialization now that aliasing has been resolved.
960 	 */
961 	ip->i_devvp = ump->um_devvp;
962 	VREF(ip->i_devvp);
963 	/*
964 	 * Set up a generation number for this inode if it does not
965 	 * already have one. This should only happen on old filesystems.
966 	 */
967 	if (ip->i_gen == 0) {
968 		if (++nextgennumber < (u_long)time.tv_sec)
969 			nextgennumber = time.tv_sec;
970 		ip->i_gen = nextgennumber;
971 		if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
972 			ip->i_flag |= IN_MODIFIED;
973 	}
974 	/*
975 	 * Ensure that uid and gid are correct. This is a temporary
976 	 * fix until fsck has been changed to do the update.
977 	 */
978 	if (fs->fs_inodefmt < FS_44INODEFMT) {		/* XXX */
979 		ip->i_uid = ip->i_din.di_ouid;		/* XXX */
980 		ip->i_gid = ip->i_din.di_ogid;		/* XXX */
981 	}						/* XXX */
982 
983 	*vpp = vp;
984 	return (0);
985 }
986 
987 /*
988  * File handle to vnode
989  *
990  * Have to be really careful about stale file handles:
991  * - check that the inode number is valid
992  * - call ffs_vget() to get the locked inode
993  * - check for an unallocated inode (i_mode == 0)
994  * - check that the given client host has export rights and return
995  *   those rights via. exflagsp and credanonp
996  */
997 int
998 ffs_fhtovp(mp, fhp, nam, vpp, exflagsp, credanonp)
999 	register struct mount *mp;
1000 	struct fid *fhp;
1001 	struct mbuf *nam;
1002 	struct vnode **vpp;
1003 	int *exflagsp;
1004 	struct ucred **credanonp;
1005 {
1006 	register struct ufid *ufhp;
1007 	struct fs *fs;
1008 
1009 	ufhp = (struct ufid *)fhp;
1010 	fs = VFSTOUFS(mp)->um_fs;
1011 	if (ufhp->ufid_ino < ROOTINO ||
1012 	    ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg)
1013 		return (ESTALE);
1014 	return (ufs_check_export(mp, ufhp, nam, vpp, exflagsp, credanonp));
1015 }
1016 
1017 /*
1018  * Vnode pointer to File handle
1019  */
1020 /* ARGSUSED */
1021 int
1022 ffs_vptofh(vp, fhp)
1023 	struct vnode *vp;
1024 	struct fid *fhp;
1025 {
1026 	register struct inode *ip;
1027 	register struct ufid *ufhp;
1028 
1029 	ip = VTOI(vp);
1030 	ufhp = (struct ufid *)fhp;
1031 	ufhp->ufid_len = sizeof(struct ufid);
1032 	ufhp->ufid_ino = ip->i_number;
1033 	ufhp->ufid_gen = ip->i_gen;
1034 	return (0);
1035 }
1036 
1037 /*
1038  * Write a superblock and associated information back to disk.
1039  */
1040 static int
1041 ffs_sbupdate(mp, waitfor)
1042 	struct ufsmount *mp;
1043 	int waitfor;
1044 {
1045 	register struct fs *fs = mp->um_fs;
1046 	register struct buf *bp;
1047 	int blks;
1048 	caddr_t space;
1049 	int i, size, error = 0;
1050 
1051 	bp = getblk(mp->um_devvp, SBLOCK, (int)fs->fs_sbsize, 0, 0);
1052 	bcopy((caddr_t)fs, bp->b_data, (u_int)fs->fs_sbsize);
1053 	/* Restore compatibility to old file systems.		   XXX */
1054 	if (fs->fs_postblformat == FS_42POSTBLFMT)		/* XXX */
1055 		((struct fs *)bp->b_data)->fs_nrpos = -1;	/* XXX */
1056 	if (waitfor == MNT_WAIT)
1057 		error = bwrite(bp);
1058 	else
1059 		bawrite(bp);
1060 	blks = howmany(fs->fs_cssize, fs->fs_fsize);
1061 	space = (caddr_t)fs->fs_csp[0];
1062 	for (i = 0; i < blks; i += fs->fs_frag) {
1063 		size = fs->fs_bsize;
1064 		if (i + fs->fs_frag > blks)
1065 			size = (blks - i) * fs->fs_fsize;
1066 		bp = getblk(mp->um_devvp, fsbtodb(fs, fs->fs_csaddr + i),
1067 		    size, 0, 0);
1068 		bcopy(space, bp->b_data, (u_int)size);
1069 		space += size;
1070 		if (waitfor == MNT_WAIT)
1071 			error = bwrite(bp);
1072 		else
1073 			bawrite(bp);
1074 	}
1075 	return (error);
1076 }
1077