ffs_vfsops.c revision 1541
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 */
35
36#include <sys/param.h>
37#include <sys/systm.h>
38#include <sys/namei.h>
39#include <sys/proc.h>
40#include <sys/kernel.h>
41#include <sys/vnode.h>
42#include <sys/socket.h>
43#include <sys/mount.h>
44#include <sys/buf.h>
45#include <sys/mbuf.h>
46#include <sys/file.h>
47#include <sys/disklabel.h>
48#include <sys/ioctl.h>
49#include <sys/errno.h>
50#include <sys/malloc.h>
51
52#include <miscfs/specfs/specdev.h>
53
54#include <ufs/ufs/quota.h>
55#include <ufs/ufs/ufsmount.h>
56#include <ufs/ufs/inode.h>
57#include <ufs/ufs/ufs_extern.h>
58
59#include <ufs/ffs/fs.h>
60#include <ufs/ffs/ffs_extern.h>
61
62int ffs_sbupdate __P((struct ufsmount *, int));
63
64struct vfsops ufs_vfsops = {
65	ffs_mount,
66	ufs_start,
67	ffs_unmount,
68	ufs_root,
69	ufs_quotactl,
70	ffs_statfs,
71	ffs_sync,
72	ffs_vget,
73	ffs_fhtovp,
74	ffs_vptofh,
75	ffs_init,
76};
77
78extern u_long nextgennumber;
79
80/*
81 * Called by main() when ufs is going to be mounted as root.
82 *
83 * Name is updated by mount(8) after booting.
84 */
85#define ROOTNAME	"root_device"
86
87ffs_mountroot()
88{
89	extern struct vnode *rootvp;
90	register struct fs *fs;
91	register struct mount *mp;
92	struct proc *p = curproc;	/* XXX */
93	struct ufsmount *ump;
94	u_int size;
95	int error;
96
97	/*
98	 * Get vnodes for swapdev and rootdev.
99	 */
100	if (bdevvp(swapdev, &swapdev_vp) || bdevvp(rootdev, &rootvp))
101		panic("ffs_mountroot: can't setup bdevvp's");
102
103	mp = malloc((u_long)sizeof(struct mount), M_MOUNT, M_WAITOK);
104	bzero((char *)mp, (u_long)sizeof(struct mount));
105	mp->mnt_op = &ufs_vfsops;
106	mp->mnt_flag = MNT_RDONLY;
107	if (error = ffs_mountfs(rootvp, mp, p)) {
108		free(mp, M_MOUNT);
109		return (error);
110	}
111	if (error = vfs_lock(mp)) {
112		(void)ffs_unmount(mp, 0, p);
113		free(mp, M_MOUNT);
114		return (error);
115	}
116	TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
117	mp->mnt_flag |= MNT_ROOTFS;
118	mp->mnt_vnodecovered = NULLVP;
119	ump = VFSTOUFS(mp);
120	fs = ump->um_fs;
121	bzero(fs->fs_fsmnt, sizeof(fs->fs_fsmnt));
122	fs->fs_fsmnt[0] = '/';
123	bcopy((caddr_t)fs->fs_fsmnt, (caddr_t)mp->mnt_stat.f_mntonname,
124	    MNAMELEN);
125	(void) copystr(ROOTNAME, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
126	    &size);
127	bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
128	(void)ffs_statfs(mp, &mp->mnt_stat, p);
129	vfs_unlock(mp);
130	inittodr(fs->fs_time);
131	return (0);
132}
133
134/*
135 * VFS Operations.
136 *
137 * mount system call
138 */
139int
140ffs_mount(mp, path, data, ndp, p)
141	register struct mount *mp;
142	char *path;
143	caddr_t data;
144	struct nameidata *ndp;
145	struct proc *p;
146{
147	struct vnode *devvp;
148	struct ufs_args args;
149	struct ufsmount *ump;
150	register struct fs *fs;
151	u_int size;
152	int error, flags;
153
154	if (error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args)))
155		return (error);
156	/*
157	 * If updating, check whether changing from read-only to
158	 * read/write; if there is no device name, that's all we do.
159	 */
160	if (mp->mnt_flag & MNT_UPDATE) {
161		ump = VFSTOUFS(mp);
162		fs = ump->um_fs;
163		error = 0;
164		if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
165			flags = WRITECLOSE;
166			if (mp->mnt_flag & MNT_FORCE)
167				flags |= FORCECLOSE;
168			if (vfs_busy(mp))
169				return (EBUSY);
170			error = ffs_flushfiles(mp, flags, p);
171			vfs_unbusy(mp);
172		}
173		if (!error && (mp->mnt_flag & MNT_RELOAD))
174			error = ffs_reload(mp, ndp->ni_cnd.cn_cred, p);
175		if (error)
176			return (error);
177		if (fs->fs_ronly && (mp->mnt_flag & MNT_WANTRDWR))
178			fs->fs_ronly = 0;
179		if (args.fspec == 0) {
180			/*
181			 * Process export requests.
182			 */
183			return (vfs_export(mp, &ump->um_export, &args.export));
184		}
185	}
186	/*
187	 * Not an update, or updating the name: look up the name
188	 * and verify that it refers to a sensible block device.
189	 */
190	NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, p);
191	if (error = namei(ndp))
192		return (error);
193	devvp = ndp->ni_vp;
194
195	if (devvp->v_type != VBLK) {
196		vrele(devvp);
197		return (ENOTBLK);
198	}
199	if (major(devvp->v_rdev) >= nblkdev) {
200		vrele(devvp);
201		return (ENXIO);
202	}
203	if ((mp->mnt_flag & MNT_UPDATE) == 0)
204		error = ffs_mountfs(devvp, mp, p);
205	else {
206		if (devvp != ump->um_devvp)
207			error = EINVAL;	/* needs translation */
208		else
209			vrele(devvp);
210	}
211	if (error) {
212		vrele(devvp);
213		return (error);
214	}
215	ump = VFSTOUFS(mp);
216	fs = ump->um_fs;
217	(void) copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size);
218	bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size);
219	bcopy((caddr_t)fs->fs_fsmnt, (caddr_t)mp->mnt_stat.f_mntonname,
220	    MNAMELEN);
221	(void) copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
222	    &size);
223	bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
224	(void)ffs_statfs(mp, &mp->mnt_stat, p);
225	return (0);
226}
227
228/*
229 * Reload all incore data for a filesystem (used after running fsck on
230 * the root filesystem and finding things to fix). The filesystem must
231 * be mounted read-only.
232 *
233 * Things to do to update the mount:
234 *	1) invalidate all cached meta-data.
235 *	2) re-read superblock from disk.
236 *	3) re-read summary information from disk.
237 *	4) invalidate all inactive vnodes.
238 *	5) invalidate all cached file data.
239 *	6) re-read inode data for all active vnodes.
240 */
241ffs_reload(mountp, cred, p)
242	register struct mount *mountp;
243	struct ucred *cred;
244	struct proc *p;
245{
246	register struct vnode *vp, *nvp, *devvp;
247	struct inode *ip;
248	struct csum *space;
249	struct buf *bp;
250	struct fs *fs;
251	int i, blks, size, error;
252
253	if ((mountp->mnt_flag & MNT_RDONLY) == 0)
254		return (EINVAL);
255	/*
256	 * Step 1: invalidate all cached meta-data.
257	 */
258	devvp = VFSTOUFS(mountp)->um_devvp;
259	if (vinvalbuf(devvp, 0, cred, p, 0, 0))
260		panic("ffs_reload: dirty1");
261	/*
262	 * Step 2: re-read superblock from disk.
263	 */
264	if (error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp))
265		return (error);
266	fs = (struct fs *)bp->b_data;
267	if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE ||
268	    fs->fs_bsize < sizeof(struct fs)) {
269		brelse(bp);
270		return (EIO);		/* XXX needs translation */
271	}
272	fs = VFSTOUFS(mountp)->um_fs;
273	bcopy(&fs->fs_csp[0], &((struct fs *)bp->b_data)->fs_csp[0],
274	    sizeof(fs->fs_csp));
275	bcopy(bp->b_data, fs, (u_int)fs->fs_sbsize);
276	if (fs->fs_sbsize < SBSIZE)
277		bp->b_flags |= B_INVAL;
278	brelse(bp);
279	ffs_oldfscompat(fs);
280	/*
281	 * Step 3: re-read summary information from disk.
282	 */
283	blks = howmany(fs->fs_cssize, fs->fs_fsize);
284	space = fs->fs_csp[0];
285	for (i = 0; i < blks; i += fs->fs_frag) {
286		size = fs->fs_bsize;
287		if (i + fs->fs_frag > blks)
288			size = (blks - i) * fs->fs_fsize;
289		if (error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
290		    NOCRED, &bp))
291			return (error);
292		bcopy(bp->b_data, fs->fs_csp[fragstoblks(fs, i)], (u_int)size);
293		brelse(bp);
294	}
295loop:
296	for (vp = mountp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) {
297		nvp = vp->v_mntvnodes.le_next;
298		/*
299		 * Step 4: invalidate all inactive vnodes.
300		 */
301		if (vp->v_usecount == 0) {
302			vgone(vp);
303			continue;
304		}
305		/*
306		 * Step 5: invalidate all cached file data.
307		 */
308		if (vget(vp, 1))
309			goto loop;
310		if (vinvalbuf(vp, 0, cred, p, 0, 0))
311			panic("ffs_reload: dirty2");
312		/*
313		 * Step 6: re-read inode data for all active vnodes.
314		 */
315		ip = VTOI(vp);
316		if (error =
317		    bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
318		    (int)fs->fs_bsize, NOCRED, &bp)) {
319			vput(vp);
320			return (error);
321		}
322		ip->i_din = *((struct dinode *)bp->b_data +
323		    ino_to_fsbo(fs, ip->i_number));
324		brelse(bp);
325		vput(vp);
326		if (vp->v_mount != mountp)
327			goto loop;
328	}
329	return (0);
330}
331
332/*
333 * Common code for mount and mountroot
334 */
335int
336ffs_mountfs(devvp, mp, p)
337	register struct vnode *devvp;
338	struct mount *mp;
339	struct proc *p;
340{
341	register struct ufsmount *ump;
342	struct buf *bp;
343	register struct fs *fs;
344	dev_t dev = devvp->v_rdev;
345	struct partinfo dpart;
346	caddr_t base, space;
347	int havepart = 0, blks;
348	int error, i, size;
349	int ronly;
350	extern struct vnode *rootvp;
351
352	/*
353	 * Disallow multiple mounts of the same device.
354	 * Disallow mounting of a device that is currently in use
355	 * (except for root, which might share swap device for miniroot).
356	 * Flush out any old buffers remaining from a previous use.
357	 */
358	if (error = vfs_mountedon(devvp))
359		return (error);
360	if (vcount(devvp) > 1 && devvp != rootvp)
361		return (EBUSY);
362	if (error = vinvalbuf(devvp, V_SAVE, p->p_ucred, p, 0, 0))
363		return (error);
364
365	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
366	if (error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p))
367		return (error);
368	if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, p) != 0)
369		size = DEV_BSIZE;
370	else {
371		havepart = 1;
372		size = dpart.disklab->d_secsize;
373	}
374
375	bp = NULL;
376	ump = NULL;
377	if (error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp))
378		goto out;
379	fs = (struct fs *)bp->b_data;
380	if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE ||
381	    fs->fs_bsize < sizeof(struct fs)) {
382		error = EINVAL;		/* XXX needs translation */
383		goto out;
384	}
385	ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
386	bzero((caddr_t)ump, sizeof *ump);
387	ump->um_fs = malloc((u_long)fs->fs_sbsize, M_UFSMNT,
388	    M_WAITOK);
389	bcopy(bp->b_data, ump->um_fs, (u_int)fs->fs_sbsize);
390	if (fs->fs_sbsize < SBSIZE)
391		bp->b_flags |= B_INVAL;
392	brelse(bp);
393	bp = NULL;
394	fs = ump->um_fs;
395	fs->fs_ronly = ronly;
396	if (ronly == 0)
397		fs->fs_fmod = 1;
398	blks = howmany(fs->fs_cssize, fs->fs_fsize);
399	base = space = malloc((u_long)fs->fs_cssize, M_UFSMNT,
400	    M_WAITOK);
401	for (i = 0; i < blks; i += fs->fs_frag) {
402		size = fs->fs_bsize;
403		if (i + fs->fs_frag > blks)
404			size = (blks - i) * fs->fs_fsize;
405		error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
406			NOCRED, &bp);
407		if (error) {
408			free(base, M_UFSMNT);
409			goto out;
410		}
411		bcopy(bp->b_data, space, (u_int)size);
412		fs->fs_csp[fragstoblks(fs, i)] = (struct csum *)space;
413		space += size;
414		brelse(bp);
415		bp = NULL;
416	}
417	mp->mnt_data = (qaddr_t)ump;
418	mp->mnt_stat.f_fsid.val[0] = (long)dev;
419	mp->mnt_stat.f_fsid.val[1] = MOUNT_UFS;
420	mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
421	mp->mnt_flag |= MNT_LOCAL;
422	ump->um_mountp = mp;
423	ump->um_dev = dev;
424	ump->um_devvp = devvp;
425	ump->um_nindir = fs->fs_nindir;
426	ump->um_bptrtodb = fs->fs_fsbtodb;
427	ump->um_seqinc = fs->fs_frag;
428	for (i = 0; i < MAXQUOTAS; i++)
429		ump->um_quotas[i] = NULLVP;
430	devvp->v_specflags |= SI_MOUNTEDON;
431	ffs_oldfscompat(fs);
432	return (0);
433out:
434	if (bp)
435		brelse(bp);
436	(void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, NOCRED, p);
437	if (ump) {
438		free(ump->um_fs, M_UFSMNT);
439		free(ump, M_UFSMNT);
440		mp->mnt_data = (qaddr_t)0;
441	}
442	return (error);
443}
444
445/*
446 * Sanity checks for old file systems.
447 *
448 * XXX - goes away some day.
449 */
450ffs_oldfscompat(fs)
451	struct fs *fs;
452{
453	int i;
454
455	fs->fs_npsect = max(fs->fs_npsect, fs->fs_nsect);	/* XXX */
456	fs->fs_interleave = max(fs->fs_interleave, 1);		/* XXX */
457	if (fs->fs_postblformat == FS_42POSTBLFMT)		/* XXX */
458		fs->fs_nrpos = 8;				/* XXX */
459	if (fs->fs_inodefmt < FS_44INODEFMT) {			/* XXX */
460		quad_t sizepb = fs->fs_bsize;			/* XXX */
461								/* XXX */
462		fs->fs_maxfilesize = fs->fs_bsize * NDADDR - 1;	/* XXX */
463		for (i = 0; i < NIADDR; i++) {			/* XXX */
464			sizepb *= NINDIR(fs);			/* XXX */
465			fs->fs_maxfilesize += sizepb;		/* XXX */
466		}						/* XXX */
467		fs->fs_qbmask = ~fs->fs_bmask;			/* XXX */
468		fs->fs_qfmask = ~fs->fs_fmask;			/* XXX */
469	}							/* XXX */
470	return (0);
471}
472
473/*
474 * unmount system call
475 */
476int
477ffs_unmount(mp, mntflags, p)
478	struct mount *mp;
479	int mntflags;
480	struct proc *p;
481{
482	register struct ufsmount *ump;
483	register struct fs *fs;
484	int error, flags, ronly;
485
486	flags = 0;
487	if (mntflags & MNT_FORCE) {
488		if (mp->mnt_flag & MNT_ROOTFS)
489			return (EINVAL);
490		flags |= FORCECLOSE;
491	}
492	if (error = ffs_flushfiles(mp, flags, p))
493		return (error);
494	ump = VFSTOUFS(mp);
495	fs = ump->um_fs;
496	ronly = !fs->fs_ronly;
497	ump->um_devvp->v_specflags &= ~SI_MOUNTEDON;
498	error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD|FWRITE,
499		NOCRED, p);
500	vrele(ump->um_devvp);
501	free(fs->fs_csp[0], M_UFSMNT);
502	free(fs, M_UFSMNT);
503	free(ump, M_UFSMNT);
504	mp->mnt_data = (qaddr_t)0;
505	mp->mnt_flag &= ~MNT_LOCAL;
506	return (error);
507}
508
509/*
510 * Flush out all the files in a filesystem.
511 */
512ffs_flushfiles(mp, flags, p)
513	register struct mount *mp;
514	int flags;
515	struct proc *p;
516{
517	extern int doforce;
518	register struct ufsmount *ump;
519	int i, error;
520
521	if (!doforce)
522		flags &= ~FORCECLOSE;
523	ump = VFSTOUFS(mp);
524#ifdef QUOTA
525	if (mp->mnt_flag & MNT_QUOTA) {
526		if (error = vflush(mp, NULLVP, SKIPSYSTEM|flags))
527			return (error);
528		for (i = 0; i < MAXQUOTAS; i++) {
529			if (ump->um_quotas[i] == NULLVP)
530				continue;
531			quotaoff(p, mp, i);
532		}
533		/*
534		 * Here we fall through to vflush again to ensure
535		 * that we have gotten rid of all the system vnodes.
536		 */
537	}
538#endif
539	error = vflush(mp, NULLVP, flags);
540	return (error);
541}
542
543/*
544 * Get file system statistics.
545 */
546int
547ffs_statfs(mp, sbp, p)
548	struct mount *mp;
549	register struct statfs *sbp;
550	struct proc *p;
551{
552	register struct ufsmount *ump;
553	register struct fs *fs;
554
555	ump = VFSTOUFS(mp);
556	fs = ump->um_fs;
557	if (fs->fs_magic != FS_MAGIC)
558		panic("ffs_statfs");
559	sbp->f_type = MOUNT_UFS;
560	sbp->f_bsize = fs->fs_fsize;
561	sbp->f_iosize = fs->fs_bsize;
562	sbp->f_blocks = fs->fs_dsize;
563	sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag +
564		fs->fs_cstotal.cs_nffree;
565	sbp->f_bavail = (fs->fs_dsize * (100 - fs->fs_minfree) / 100) -
566		(fs->fs_dsize - sbp->f_bfree);
567	sbp->f_files =  fs->fs_ncg * fs->fs_ipg - ROOTINO;
568	sbp->f_ffree = fs->fs_cstotal.cs_nifree;
569	if (sbp != &mp->mnt_stat) {
570		bcopy((caddr_t)mp->mnt_stat.f_mntonname,
571			(caddr_t)&sbp->f_mntonname[0], MNAMELEN);
572		bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
573			(caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
574	}
575	return (0);
576}
577
578/*
579 * Go through the disk queues to initiate sandbagged IO;
580 * go through the inodes to write those that have been modified;
581 * initiate the writing of the super block if it has been modified.
582 *
583 * Note: we are always called with the filesystem marked `MPBUSY'.
584 */
585int
586ffs_sync(mp, waitfor, cred, p)
587	struct mount *mp;
588	int waitfor;
589	struct ucred *cred;
590	struct proc *p;
591{
592	register struct vnode *vp;
593	register struct inode *ip;
594	register struct ufsmount *ump = VFSTOUFS(mp);
595	register struct fs *fs;
596	int error, allerror = 0;
597
598	fs = ump->um_fs;
599	/*
600	 * Write back modified superblock.
601	 * Consistency check that the superblock
602	 * is still in the buffer cache.
603	 */
604	if (fs->fs_fmod != 0) {
605		if (fs->fs_ronly != 0) {		/* XXX */
606			printf("fs = %s\n", fs->fs_fsmnt);
607			panic("update: rofs mod");
608		}
609		fs->fs_fmod = 0;
610		fs->fs_time = time.tv_sec;
611		allerror = ffs_sbupdate(ump, waitfor);
612	}
613	/*
614	 * Write back each (modified) inode.
615	 */
616loop:
617	for (vp = mp->mnt_vnodelist.lh_first;
618	     vp != NULL;
619	     vp = vp->v_mntvnodes.le_next) {
620		/*
621		 * If the vnode that we are about to sync is no longer
622		 * associated with this mount point, start over.
623		 */
624		if (vp->v_mount != mp)
625			goto loop;
626		if (VOP_ISLOCKED(vp))
627			continue;
628		ip = VTOI(vp);
629		if ((ip->i_flag &
630		    (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
631		    vp->v_dirtyblkhd.lh_first == NULL)
632			continue;
633		if (vget(vp, 1))
634			goto loop;
635		if (error = VOP_FSYNC(vp, cred, waitfor, p))
636			allerror = error;
637		vput(vp);
638	}
639	/*
640	 * Force stale file system control information to be flushed.
641	 */
642	if (error = VOP_FSYNC(ump->um_devvp, cred, waitfor, p))
643		allerror = error;
644#ifdef QUOTA
645	qsync(mp);
646#endif
647	return (allerror);
648}
649
650/*
651 * Look up a FFS dinode number to find its incore vnode, otherwise read it
652 * in from disk.  If it is in core, wait for the lock bit to clear, then
653 * return the inode locked.  Detection and handling of mount points must be
654 * done by the calling routine.
655 */
656int
657ffs_vget(mp, ino, vpp)
658	struct mount *mp;
659	ino_t ino;
660	struct vnode **vpp;
661{
662	register struct fs *fs;
663	register struct inode *ip;
664	struct ufsmount *ump;
665	struct buf *bp;
666	struct vnode *vp;
667	dev_t dev;
668	int i, type, error;
669
670	ump = VFSTOUFS(mp);
671	dev = ump->um_dev;
672	if ((*vpp = ufs_ihashget(dev, ino)) != NULL)
673		return (0);
674
675	/* Allocate a new vnode/inode. */
676	if (error = getnewvnode(VT_UFS, mp, ffs_vnodeop_p, &vp)) {
677		*vpp = NULL;
678		return (error);
679	}
680	type = ump->um_devvp->v_tag == VT_MFS ? M_MFSNODE : M_FFSNODE; /* XXX */
681	MALLOC(ip, struct inode *, sizeof(struct inode), type, M_WAITOK);
682	bzero((caddr_t)ip, sizeof(struct inode));
683	vp->v_data = ip;
684	ip->i_vnode = vp;
685	ip->i_fs = fs = ump->um_fs;
686	ip->i_dev = dev;
687	ip->i_number = ino;
688#ifdef QUOTA
689	for (i = 0; i < MAXQUOTAS; i++)
690		ip->i_dquot[i] = NODQUOT;
691#endif
692	/*
693	 * Put it onto its hash chain and lock it so that other requests for
694	 * this inode will block if they arrive while we are sleeping waiting
695	 * for old data structures to be purged or for the contents of the
696	 * disk portion of this inode to be read.
697	 */
698	ufs_ihashins(ip);
699
700	/* Read in the disk contents for the inode, copy into the inode. */
701	if (error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
702	    (int)fs->fs_bsize, NOCRED, &bp)) {
703		/*
704		 * The inode does not contain anything useful, so it would
705		 * be misleading to leave it on its hash chain. With mode
706		 * still zero, it will be unlinked and returned to the free
707		 * list by vput().
708		 */
709		vput(vp);
710		brelse(bp);
711		*vpp = NULL;
712		return (error);
713	}
714	ip->i_din = *((struct dinode *)bp->b_data + ino_to_fsbo(fs, ino));
715	brelse(bp);
716
717	/*
718	 * Initialize the vnode from the inode, check for aliases.
719	 * Note that the underlying vnode may have changed.
720	 */
721	if (error = ufs_vinit(mp, ffs_specop_p, FFS_FIFOOPS, &vp)) {
722		vput(vp);
723		*vpp = NULL;
724		return (error);
725	}
726	/*
727	 * Finish inode initialization now that aliasing has been resolved.
728	 */
729	ip->i_devvp = ump->um_devvp;
730	VREF(ip->i_devvp);
731	/*
732	 * Set up a generation number for this inode if it does not
733	 * already have one. This should only happen on old filesystems.
734	 */
735	if (ip->i_gen == 0) {
736		if (++nextgennumber < (u_long)time.tv_sec)
737			nextgennumber = time.tv_sec;
738		ip->i_gen = nextgennumber;
739		if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
740			ip->i_flag |= IN_MODIFIED;
741	}
742	/*
743	 * Ensure that uid and gid are correct. This is a temporary
744	 * fix until fsck has been changed to do the update.
745	 */
746	if (fs->fs_inodefmt < FS_44INODEFMT) {		/* XXX */
747		ip->i_uid = ip->i_din.di_ouid;		/* XXX */
748		ip->i_gid = ip->i_din.di_ogid;		/* XXX */
749	}						/* XXX */
750
751	*vpp = vp;
752	return (0);
753}
754
755/*
756 * File handle to vnode
757 *
758 * Have to be really careful about stale file handles:
759 * - check that the inode number is valid
760 * - call ffs_vget() to get the locked inode
761 * - check for an unallocated inode (i_mode == 0)
762 * - check that the given client host has export rights and return
763 *   those rights via. exflagsp and credanonp
764 */
765int
766ffs_fhtovp(mp, fhp, nam, vpp, exflagsp, credanonp)
767	register struct mount *mp;
768	struct fid *fhp;
769	struct mbuf *nam;
770	struct vnode **vpp;
771	int *exflagsp;
772	struct ucred **credanonp;
773{
774	register struct ufid *ufhp;
775	struct fs *fs;
776
777	ufhp = (struct ufid *)fhp;
778	fs = VFSTOUFS(mp)->um_fs;
779	if (ufhp->ufid_ino < ROOTINO ||
780	    ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg)
781		return (ESTALE);
782	return (ufs_check_export(mp, ufhp, nam, vpp, exflagsp, credanonp));
783}
784
785/*
786 * Vnode pointer to File handle
787 */
788/* ARGSUSED */
789ffs_vptofh(vp, fhp)
790	struct vnode *vp;
791	struct fid *fhp;
792{
793	register struct inode *ip;
794	register struct ufid *ufhp;
795
796	ip = VTOI(vp);
797	ufhp = (struct ufid *)fhp;
798	ufhp->ufid_len = sizeof(struct ufid);
799	ufhp->ufid_ino = ip->i_number;
800	ufhp->ufid_gen = ip->i_gen;
801	return (0);
802}
803
804/*
805 * Write a superblock and associated information back to disk.
806 */
807int
808ffs_sbupdate(mp, waitfor)
809	struct ufsmount *mp;
810	int waitfor;
811{
812	register struct fs *fs = mp->um_fs;
813	register struct buf *bp;
814	int blks;
815	caddr_t space;
816	int i, size, error = 0;
817
818	bp = getblk(mp->um_devvp, SBLOCK, (int)fs->fs_sbsize, 0, 0);
819	bcopy((caddr_t)fs, bp->b_data, (u_int)fs->fs_sbsize);
820	/* Restore compatibility to old file systems.		   XXX */
821	if (fs->fs_postblformat == FS_42POSTBLFMT)		/* XXX */
822		((struct fs *)bp->b_data)->fs_nrpos = -1;	/* XXX */
823	if (waitfor == MNT_WAIT)
824		error = bwrite(bp);
825	else
826		bawrite(bp);
827	blks = howmany(fs->fs_cssize, fs->fs_fsize);
828	space = (caddr_t)fs->fs_csp[0];
829	for (i = 0; i < blks; i += fs->fs_frag) {
830		size = fs->fs_bsize;
831		if (i + fs->fs_frag > blks)
832			size = (blks - i) * fs->fs_fsize;
833		bp = getblk(mp->um_devvp, fsbtodb(fs, fs->fs_csaddr + i),
834		    size, 0, 0);
835		bcopy(space, bp->b_data, (u_int)size);
836		space += size;
837		if (waitfor == MNT_WAIT)
838			error = bwrite(bp);
839		else
840			bawrite(bp);
841	}
842	return (error);
843}
844