ffs_vfsops.c revision 18104
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 1.40 1996/08/21 21:56:09 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
72static int	ffs_sbupdate __P((struct ufsmount *, int));
73static int	ffs_reload __P((struct mount *,struct ucred *,struct proc *));
74static int	ffs_oldfscompat __P((struct fs *));
75static int	ffs_mount __P((struct mount *,
76	    char *, caddr_t, struct nameidata *, struct proc *));
77
78struct 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
92VFS_SET(ufs_vfsops, ufs, MOUNT_UFS, 0);
93
94extern 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 */
134static int
135ffs_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
321dostatfs:
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
333error_2:	/* error with devvp held*/
334
335	/* release devvp before failing*/
336	vrele(devvp);
337
338error_1:	/* no state to back out*/
339
340success:
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 */
358static int
359ffs_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	if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE ||
387	    fs->fs_bsize < sizeof(struct fs)) {
388		brelse(bp);
389		return (EIO);		/* XXX needs translation */
390	}
391	fs = VFSTOUFS(mp)->um_fs;
392	bcopy(&fs->fs_csp[0], &((struct fs *)bp->b_data)->fs_csp[0],
393	    sizeof(fs->fs_csp));
394	bcopy(bp->b_data, fs, (u_int)fs->fs_sbsize);
395	if (fs->fs_sbsize < SBSIZE)
396		bp->b_flags |= B_INVAL;
397	brelse(bp);
398	ffs_oldfscompat(fs);
399	/*
400	 * Step 3: re-read summary information from disk.
401	 */
402	blks = howmany(fs->fs_cssize, fs->fs_fsize);
403	space = fs->fs_csp[0];
404	for (i = 0; i < blks; i += fs->fs_frag) {
405		size = fs->fs_bsize;
406		if (i + fs->fs_frag > blks)
407			size = (blks - i) * fs->fs_fsize;
408		error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
409		    NOCRED, &bp);
410		if (error)
411			return (error);
412		bcopy(bp->b_data, fs->fs_csp[fragstoblks(fs, i)], (u_int)size);
413		brelse(bp);
414	}
415loop:
416	for (vp = mp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) {
417		nvp = vp->v_mntvnodes.le_next;
418		/*
419		 * Step 4: invalidate all inactive vnodes.
420		 */
421		if (vp->v_usecount == 0) {
422			vgone(vp);
423			continue;
424		}
425		/*
426		 * Step 5: invalidate all cached file data.
427		 */
428		if (vget(vp, 1))
429			goto loop;
430		if (vinvalbuf(vp, 0, cred, p, 0, 0))
431			panic("ffs_reload: dirty2");
432		/*
433		 * Step 6: re-read inode data for all active vnodes.
434		 */
435		ip = VTOI(vp);
436		error =
437		    bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
438		    (int)fs->fs_bsize, NOCRED, &bp);
439		if (error) {
440			vput(vp);
441			return (error);
442		}
443		ip->i_din = *((struct dinode *)bp->b_data +
444		    ino_to_fsbo(fs, ip->i_number));
445		brelse(bp);
446		vput(vp);
447		if (vp->v_mount != mp)
448			goto loop;
449	}
450	return (0);
451}
452
453/*
454 * Common code for mount and mountroot
455 */
456int
457ffs_mountfs(devvp, mp, p)
458	register struct vnode *devvp;
459	struct mount *mp;
460	struct proc *p;
461{
462	register struct ufsmount *ump;
463	struct buf *bp;
464	register struct fs *fs;
465	dev_t dev = devvp->v_rdev;
466	struct partinfo dpart;
467	caddr_t base, space;
468	int havepart = 0, blks;
469	int error, i, size;
470	int ronly;
471	u_int strsize;
472	int ncount;
473
474	/*
475	 * Disallow multiple mounts of the same device.
476	 * Disallow mounting of a device that is currently in use
477	 * (except for root, which might share swap device for miniroot).
478	 * Flush out any old buffers remaining from a previous use.
479	 */
480	error = vfs_mountedon(devvp);
481	if (error)
482		return (error);
483	ncount = vcount(devvp);
484	if (devvp->v_object)
485		ncount -= 1;
486	if (ncount > 1 && devvp != rootvp)
487		return (EBUSY);
488	error = vinvalbuf(devvp, V_SAVE, p->p_ucred, p, 0, 0);
489	if (error)
490		return (error);
491
492	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
493	error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p);
494	if (error)
495		return (error);
496	if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, p) != 0)
497		size = DEV_BSIZE;
498	else {
499		havepart = 1;
500		size = dpart.disklab->d_secsize;
501	}
502
503	bp = NULL;
504	ump = NULL;
505	error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp);
506	if (error)
507		goto out;
508	fs = (struct fs *)bp->b_data;
509	if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE ||
510	    fs->fs_bsize < sizeof(struct fs)) {
511		error = EINVAL;		/* XXX needs translation */
512		goto out;
513	}
514	if (!fs->fs_clean) {
515		if (ronly || (mp->mnt_flag & MNT_FORCE)) {
516			printf("WARNING: %s was not properly dismounted.\n",fs->fs_fsmnt);
517		} else {
518			printf("WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck.\n",fs->fs_fsmnt);
519			error = EPERM;
520			goto out;
521		}
522	}
523	ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
524	bzero((caddr_t)ump, sizeof *ump);
525	ump->um_fs = malloc((u_long)fs->fs_sbsize, M_UFSMNT,
526	    M_WAITOK);
527	bcopy(bp->b_data, ump->um_fs, (u_int)fs->fs_sbsize);
528	if (fs->fs_sbsize < SBSIZE)
529		bp->b_flags |= B_INVAL;
530	brelse(bp);
531	bp = NULL;
532	fs = ump->um_fs;
533	fs->fs_ronly = ronly;
534	if (ronly == 0) {
535		fs->fs_fmod = 1;
536		fs->fs_clean = 0;
537	}
538	blks = howmany(fs->fs_cssize, fs->fs_fsize);
539	base = space = malloc((u_long)fs->fs_cssize, M_UFSMNT,
540	    M_WAITOK);
541	for (i = 0; i < blks; i += fs->fs_frag) {
542		size = fs->fs_bsize;
543		if (i + fs->fs_frag > blks)
544			size = (blks - i) * fs->fs_fsize;
545		error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
546			NOCRED, &bp);
547		if (error) {
548			free(base, M_UFSMNT);
549			goto out;
550		}
551		bcopy(bp->b_data, space, (u_int)size);
552		fs->fs_csp[fragstoblks(fs, i)] = (struct csum *)space;
553		space += size;
554		brelse(bp);
555		bp = NULL;
556	}
557	mp->mnt_data = (qaddr_t)ump;
558	mp->mnt_stat.f_fsid.val[0] = (long)dev;
559	mp->mnt_stat.f_fsid.val[1] = MOUNT_UFS;
560	mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
561	mp->mnt_flag |= MNT_LOCAL;
562	ump->um_mountp = mp;
563	ump->um_dev = dev;
564	ump->um_devvp = devvp;
565	ump->um_nindir = fs->fs_nindir;
566	ump->um_bptrtodb = fs->fs_fsbtodb;
567	ump->um_seqinc = fs->fs_frag;
568	for (i = 0; i < MAXQUOTAS; i++)
569		ump->um_quotas[i] = NULLVP;
570	devvp->v_specflags |= SI_MOUNTEDON;
571	ffs_oldfscompat(fs);
572
573	/*
574	 * Set FS local "last mounted on" information (NULL pad)
575	 */
576	copystr(	mp->mnt_stat.f_mntonname,	/* mount point*/
577			fs->fs_fsmnt,			/* copy area*/
578			sizeof(fs->fs_fsmnt) - 1,	/* max size*/
579			&strsize);			/* real size*/
580	bzero( fs->fs_fsmnt + strsize, sizeof(fs->fs_fsmnt) - strsize);
581
582	if( mp->mnt_flag & MNT_ROOTFS) {
583		/*
584		 * Root mount; update timestamp in mount structure.
585		 * this will be used by the common root mount code
586		 * to update the system clock.
587		 */
588		mp->mnt_time = fs->fs_time;
589	}
590	if (ronly == 0)
591		ffs_sbupdate(ump, MNT_WAIT);
592	/*
593	 * Only VMIO the backing device if the backing device is a real
594	 * block device.  This excludes the original MFS implementation.
595	 * Note that it is optional that the backing device be VMIOed.  This
596	 * increases the opportunity for metadata caching.
597	 */
598	if ((devvp->v_type == VBLK) && (major(devvp->v_rdev) < nblkdev)) {
599		vfs_object_create(devvp, p, p->p_ucred, 0);
600	}
601	return (0);
602out:
603	if (bp)
604		brelse(bp);
605	(void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, NOCRED, p);
606	if (ump) {
607		free(ump->um_fs, M_UFSMNT);
608		free(ump, M_UFSMNT);
609		mp->mnt_data = (qaddr_t)0;
610	}
611	return (error);
612}
613
614/*
615 * Sanity checks for old file systems.
616 *
617 * XXX - goes away some day.
618 */
619static int
620ffs_oldfscompat(fs)
621	struct fs *fs;
622{
623
624	fs->fs_npsect = max(fs->fs_npsect, fs->fs_nsect);	/* XXX */
625	fs->fs_interleave = max(fs->fs_interleave, 1);		/* XXX */
626	if (fs->fs_postblformat == FS_42POSTBLFMT)		/* XXX */
627		fs->fs_nrpos = 8;				/* XXX */
628	if (fs->fs_inodefmt < FS_44INODEFMT) {			/* XXX */
629#if 0
630		int i;						/* XXX */
631		quad_t sizepb = fs->fs_bsize;			/* XXX */
632		fs->fs_maxfilesize = fs->fs_bsize * NDADDR - 1;	/* XXX */
633		for (i = 0; i < NIADDR; i++) {			/* XXX */
634			sizepb *= NINDIR(fs);			/* XXX */
635			fs->fs_maxfilesize += sizepb;		/* XXX */
636		}						/* XXX */
637#endif
638		fs->fs_maxfilesize = (u_quad_t) 1LL << 39;
639		fs->fs_qbmask = ~fs->fs_bmask;			/* XXX */
640		fs->fs_qfmask = ~fs->fs_fmask;			/* XXX */
641	}							/* XXX */
642	return (0);
643}
644
645/*
646 * unmount system call
647 */
648int
649ffs_unmount(mp, mntflags, p)
650	struct mount *mp;
651	int mntflags;
652	struct proc *p;
653{
654	register struct ufsmount *ump;
655	register struct fs *fs;
656	int error, flags, ronly;
657
658	flags = 0;
659	if (mntflags & MNT_FORCE) {
660		flags |= FORCECLOSE;
661	}
662	error = ffs_flushfiles(mp, flags, p);
663	if (error)
664		return (error);
665	ump = VFSTOUFS(mp);
666	fs = ump->um_fs;
667	ronly = fs->fs_ronly;
668	if (!ronly) {
669		fs->fs_clean = 1;
670		ffs_sbupdate(ump, MNT_WAIT);
671	}
672	ump->um_devvp->v_specflags &= ~SI_MOUNTEDON;
673
674	VOP_LOCK(ump->um_devvp);
675	vnode_pager_uncache(ump->um_devvp);
676	VOP_UNLOCK(ump->um_devvp);
677
678	error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD|FWRITE,
679		NOCRED, p);
680
681	vrele(ump->um_devvp);
682
683	free(fs->fs_csp[0], M_UFSMNT);
684	free(fs, M_UFSMNT);
685	free(ump, M_UFSMNT);
686	mp->mnt_data = (qaddr_t)0;
687	mp->mnt_flag &= ~MNT_LOCAL;
688	return (error);
689}
690
691/*
692 * Flush out all the files in a filesystem.
693 */
694int
695ffs_flushfiles(mp, flags, p)
696	register struct mount *mp;
697	int flags;
698	struct proc *p;
699{
700	register struct ufsmount *ump;
701	int error;
702
703	if (!doforce)
704		flags &= ~FORCECLOSE;
705	ump = VFSTOUFS(mp);
706#ifdef QUOTA
707	if (mp->mnt_flag & MNT_QUOTA) {
708		int i;
709		error = vflush(mp, NULLVP, SKIPSYSTEM|flags);
710		if (error)
711			return (error);
712		for (i = 0; i < MAXQUOTAS; i++) {
713			if (ump->um_quotas[i] == NULLVP)
714				continue;
715			quotaoff(p, mp, i);
716		}
717		/*
718		 * Here we fall through to vflush again to ensure
719		 * that we have gotten rid of all the system vnodes.
720		 */
721	}
722#endif
723	error = vflush(mp, NULLVP, flags);
724	return (error);
725}
726
727/*
728 * Get file system statistics.
729 */
730int
731ffs_statfs(mp, sbp, p)
732	struct mount *mp;
733	register struct statfs *sbp;
734	struct proc *p;
735{
736	register struct ufsmount *ump;
737	register struct fs *fs;
738
739	ump = VFSTOUFS(mp);
740	fs = ump->um_fs;
741	if (fs->fs_magic != FS_MAGIC)
742		panic("ffs_statfs");
743	sbp->f_type = MOUNT_UFS;
744	sbp->f_bsize = fs->fs_fsize;
745	sbp->f_iosize = fs->fs_bsize;
746	sbp->f_blocks = fs->fs_dsize;
747	sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag +
748		fs->fs_cstotal.cs_nffree;
749	sbp->f_bavail = freespace(fs, fs->fs_minfree);
750	sbp->f_files =  fs->fs_ncg * fs->fs_ipg - ROOTINO;
751	sbp->f_ffree = fs->fs_cstotal.cs_nifree;
752	if (sbp != &mp->mnt_stat) {
753		bcopy((caddr_t)mp->mnt_stat.f_mntonname,
754			(caddr_t)&sbp->f_mntonname[0], MNAMELEN);
755		bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
756			(caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
757	}
758	return (0);
759}
760
761/*
762 * Go through the disk queues to initiate sandbagged IO;
763 * go through the inodes to write those that have been modified;
764 * initiate the writing of the super block if it has been modified.
765 *
766 * Note: we are always called with the filesystem marked `MPBUSY'.
767 */
768int
769ffs_sync(mp, waitfor, cred, p)
770	struct mount *mp;
771	int waitfor;
772	struct ucred *cred;
773	struct proc *p;
774{
775	register struct vnode *vp, *nvp;
776	register struct inode *ip;
777	register struct ufsmount *ump = VFSTOUFS(mp);
778	register struct fs *fs;
779	struct timeval tv;
780	int error, allerror = 0;
781
782	fs = ump->um_fs;
783	/*
784	 * Write back modified superblock.
785	 * Consistency check that the superblock
786	 * is still in the buffer cache.
787	 */
788	if (fs->fs_fmod != 0) {
789		if (fs->fs_ronly != 0) {		/* XXX */
790			printf("fs = %s\n", fs->fs_fsmnt);
791			panic("update: rofs mod");
792		}
793		fs->fs_fmod = 0;
794		fs->fs_time = time.tv_sec;
795		allerror = ffs_sbupdate(ump, waitfor);
796	}
797	/*
798	 * Write back each (modified) inode.
799	 */
800loop:
801	for (vp = mp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) {
802		/*
803		 * If the vnode that we are about to sync is no longer
804		 * associated with this mount point, start over.
805		 */
806		if (vp->v_mount != mp)
807			goto loop;
808		nvp = vp->v_mntvnodes.le_next;
809		if (VOP_ISLOCKED(vp))
810			continue;
811		ip = VTOI(vp);
812		if ((((ip->i_flag &
813		    (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0)) &&
814		    vp->v_dirtyblkhd.lh_first == NULL)
815			continue;
816		if (vp->v_type != VCHR) {
817			if (vget(vp, 1))
818				goto loop;
819			error = VOP_FSYNC(vp, cred, waitfor, p);
820			if (error)
821				allerror = error;
822			vput(vp);
823		} else {
824			tv = time;
825			/* VOP_UPDATE(vp, &tv, &tv, waitfor == MNT_WAIT); */
826			VOP_UPDATE(vp, &tv, &tv, 0);
827		}
828	}
829	/*
830	 * Force stale file system control information to be flushed.
831	 */
832	error = VOP_FSYNC(ump->um_devvp, cred, waitfor, p);
833	if (error)
834		allerror = error;
835#ifdef QUOTA
836	qsync(mp);
837#endif
838	return (allerror);
839}
840
841/*
842 * Look up a FFS dinode number to find its incore vnode, otherwise read it
843 * in from disk.  If it is in core, wait for the lock bit to clear, then
844 * return the inode locked.  Detection and handling of mount points must be
845 * done by the calling routine.
846 */
847static int ffs_inode_hash_lock;
848
849int
850ffs_vget(mp, ino, vpp)
851	struct mount *mp;
852	ino_t ino;
853	struct vnode **vpp;
854{
855	register struct fs *fs;
856	register struct inode *ip;
857	struct ufsmount *ump;
858	struct buf *bp;
859	struct vnode *vp;
860	dev_t dev;
861	int type, error;
862
863	ump = VFSTOUFS(mp);
864	dev = ump->um_dev;
865restart:
866	if ((*vpp = ufs_ihashget(dev, ino)) != NULL)
867		return (0);
868
869	/*
870	 * Lock out the creation of new entries in the FFS hash table in
871	 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate
872	 * may occur!
873	 */
874	if (ffs_inode_hash_lock) {
875		while (ffs_inode_hash_lock) {
876			ffs_inode_hash_lock = -1;
877			tsleep(&ffs_inode_hash_lock, PVM, "ffsvgt", 0);
878		}
879		goto restart;
880	}
881	ffs_inode_hash_lock = 1;
882
883	/*
884	 * If this MALLOC() is performed after the getnewvnode()
885	 * it might block, leaving a vnode with a NULL v_data to be
886	 * found by ffs_sync() if a sync happens to fire right then,
887	 * which will cause a panic because ffs_sync() blindly
888	 * dereferences vp->v_data (as well it should).
889	 */
890	type = ump->um_devvp->v_tag == VT_MFS ? M_MFSNODE : M_FFSNODE; /* XXX */
891	MALLOC(ip, struct inode *, sizeof(struct inode), type, M_WAITOK);
892
893	/* Allocate a new vnode/inode. */
894	error = getnewvnode(VT_UFS, mp, ffs_vnodeop_p, &vp);
895	if (error) {
896		if (ffs_inode_hash_lock < 0)
897			wakeup(&ffs_inode_hash_lock);
898		ffs_inode_hash_lock = 0;
899		*vpp = NULL;
900		FREE(ip, type);
901		return (error);
902	}
903	bzero((caddr_t)ip, sizeof(struct inode));
904	vp->v_data = ip;
905	ip->i_vnode = vp;
906	ip->i_fs = fs = ump->um_fs;
907	ip->i_dev = dev;
908	ip->i_number = ino;
909#ifdef QUOTA
910	{
911		int i;
912		for (i = 0; i < MAXQUOTAS; i++)
913			ip->i_dquot[i] = NODQUOT;
914	}
915#endif
916	/*
917	 * Put it onto its hash chain and lock it so that other requests for
918	 * this inode will block if they arrive while we are sleeping waiting
919	 * for old data structures to be purged or for the contents of the
920	 * disk portion of this inode to be read.
921	 */
922	ufs_ihashins(ip);
923
924	if (ffs_inode_hash_lock < 0)
925		wakeup(&ffs_inode_hash_lock);
926	ffs_inode_hash_lock = 0;
927
928	/* Read in the disk contents for the inode, copy into the inode. */
929	error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
930	    (int)fs->fs_bsize, NOCRED, &bp);
931	if (error) {
932		/*
933		 * The inode does not contain anything useful, so it would
934		 * be misleading to leave it on its hash chain. With mode
935		 * still zero, it will be unlinked and returned to the free
936		 * list by vput().
937		 */
938		brelse(bp);
939		vput(vp);
940		*vpp = NULL;
941		return (error);
942	}
943	ip->i_din = *((struct dinode *)bp->b_data + ino_to_fsbo(fs, ino));
944	bqrelse(bp);
945
946	/*
947	 * Initialize the vnode from the inode, check for aliases.
948	 * Note that the underlying vnode may have changed.
949	 */
950	error = ufs_vinit(mp, ffs_specop_p, ffs_fifoop_p, &vp);
951	if (error) {
952		vput(vp);
953		*vpp = NULL;
954		return (error);
955	}
956	/*
957	 * Finish inode initialization now that aliasing has been resolved.
958	 */
959	ip->i_devvp = ump->um_devvp;
960	VREF(ip->i_devvp);
961	/*
962	 * Set up a generation number for this inode if it does not
963	 * already have one. This should only happen on old filesystems.
964	 */
965	if (ip->i_gen == 0) {
966		if (++nextgennumber < (u_long)time.tv_sec)
967			nextgennumber = time.tv_sec;
968		ip->i_gen = nextgennumber;
969		if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
970			ip->i_flag |= IN_MODIFIED;
971	}
972	/*
973	 * Ensure that uid and gid are correct. This is a temporary
974	 * fix until fsck has been changed to do the update.
975	 */
976	if (fs->fs_inodefmt < FS_44INODEFMT) {		/* XXX */
977		ip->i_uid = ip->i_din.di_ouid;		/* XXX */
978		ip->i_gid = ip->i_din.di_ogid;		/* XXX */
979	}						/* XXX */
980
981	*vpp = vp;
982	return (0);
983}
984
985/*
986 * File handle to vnode
987 *
988 * Have to be really careful about stale file handles:
989 * - check that the inode number is valid
990 * - call ffs_vget() to get the locked inode
991 * - check for an unallocated inode (i_mode == 0)
992 * - check that the given client host has export rights and return
993 *   those rights via. exflagsp and credanonp
994 */
995int
996ffs_fhtovp(mp, fhp, nam, vpp, exflagsp, credanonp)
997	register struct mount *mp;
998	struct fid *fhp;
999	struct mbuf *nam;
1000	struct vnode **vpp;
1001	int *exflagsp;
1002	struct ucred **credanonp;
1003{
1004	register struct ufid *ufhp;
1005	struct fs *fs;
1006
1007	ufhp = (struct ufid *)fhp;
1008	fs = VFSTOUFS(mp)->um_fs;
1009	if (ufhp->ufid_ino < ROOTINO ||
1010	    ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg)
1011		return (ESTALE);
1012	return (ufs_check_export(mp, ufhp, nam, vpp, exflagsp, credanonp));
1013}
1014
1015/*
1016 * Vnode pointer to File handle
1017 */
1018/* ARGSUSED */
1019int
1020ffs_vptofh(vp, fhp)
1021	struct vnode *vp;
1022	struct fid *fhp;
1023{
1024	register struct inode *ip;
1025	register struct ufid *ufhp;
1026
1027	ip = VTOI(vp);
1028	ufhp = (struct ufid *)fhp;
1029	ufhp->ufid_len = sizeof(struct ufid);
1030	ufhp->ufid_ino = ip->i_number;
1031	ufhp->ufid_gen = ip->i_gen;
1032	return (0);
1033}
1034
1035/*
1036 * Write a superblock and associated information back to disk.
1037 */
1038static int
1039ffs_sbupdate(mp, waitfor)
1040	struct ufsmount *mp;
1041	int waitfor;
1042{
1043	register struct fs *fs = mp->um_fs;
1044	register struct buf *bp;
1045	int blks;
1046	caddr_t space;
1047	int i, size, error = 0;
1048
1049	bp = getblk(mp->um_devvp, SBLOCK, (int)fs->fs_sbsize, 0, 0);
1050	bcopy((caddr_t)fs, bp->b_data, (u_int)fs->fs_sbsize);
1051	/* Restore compatibility to old file systems.		   XXX */
1052	if (fs->fs_postblformat == FS_42POSTBLFMT)		/* XXX */
1053		((struct fs *)bp->b_data)->fs_nrpos = -1;	/* XXX */
1054	if (waitfor == MNT_WAIT)
1055		error = bwrite(bp);
1056	else
1057		bawrite(bp);
1058	blks = howmany(fs->fs_cssize, fs->fs_fsize);
1059	space = (caddr_t)fs->fs_csp[0];
1060	for (i = 0; i < blks; i += fs->fs_frag) {
1061		size = fs->fs_bsize;
1062		if (i + fs->fs_frag > blks)
1063			size = (blks - i) * fs->fs_fsize;
1064		bp = getblk(mp->um_devvp, fsbtodb(fs, fs->fs_csaddr + i),
1065		    size, 0, 0);
1066		bcopy(space, bp->b_data, (u_int)size);
1067		space += size;
1068		if (waitfor == MNT_WAIT)
1069			error = bwrite(bp);
1070		else
1071			bawrite(bp);
1072	}
1073	return (error);
1074}
1075