msdosfs_vfsops.c revision 177785
1/* $FreeBSD: head/sys/fs/msdosfs/msdosfs_vfsops.c 177785 2008-03-31 12:01:21Z kib $ */
2/*	$NetBSD: msdosfs_vfsops.c,v 1.51 1997/11/17 15:36:58 ws Exp $	*/
3
4/*-
5 * Copyright (C) 1994, 1995, 1997 Wolfgang Solfrank.
6 * Copyright (C) 1994, 1995, 1997 TooLs GmbH.
7 * All rights reserved.
8 * Original code by Paul Popelka (paulp@uts.amdahl.com) (see below).
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 *    notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 *    notice, this list of conditions and the following disclaimer in the
17 *    documentation and/or other materials provided with the distribution.
18 * 3. All advertising materials mentioning features or use of this software
19 *    must display the following acknowledgement:
20 *	This product includes software developed by TooLs GmbH.
21 * 4. The name of TooLs GmbH may not be used to endorse or promote products
22 *    derived from this software without specific prior written permission.
23 *
24 * THIS SOFTWARE IS PROVIDED BY TOOLS GMBH ``AS IS'' AND ANY EXPRESS OR
25 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
26 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
27 * IN NO EVENT SHALL TOOLS GMBH BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
28 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
29 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;
30 * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
31 * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
32 * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
33 * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
34 */
35/*-
36 * Written by Paul Popelka (paulp@uts.amdahl.com)
37 *
38 * You can do anything you want with this software, just don't say you wrote
39 * it, and don't remove this notice.
40 *
41 * This software is provided "as is".
42 *
43 * The author supplies this software to be publicly redistributed on the
44 * understanding that the author is not responsible for the correct
45 * functioning of this software in any circumstances and is not liable for
46 * any damages caused by this software.
47 *
48 * October 1992
49 */
50
51#include <sys/param.h>
52#include <sys/systm.h>
53#include <sys/buf.h>
54#include <sys/conf.h>
55#include <sys/fcntl.h>
56#include <sys/iconv.h>
57#include <sys/kernel.h>
58#include <sys/lock.h>
59#include <sys/malloc.h>
60#include <sys/mount.h>
61#include <sys/mutex.h>
62#include <sys/namei.h>
63#include <sys/priv.h>
64#include <sys/proc.h>
65#include <sys/stat.h>
66#include <sys/vnode.h>
67
68#include <geom/geom.h>
69#include <geom/geom_vfs.h>
70
71#include <fs/msdosfs/bootsect.h>
72#include <fs/msdosfs/bpb.h>
73#include <fs/msdosfs/direntry.h>
74#include <fs/msdosfs/denode.h>
75#include <fs/msdosfs/fat.h>
76#include <fs/msdosfs/msdosfsmount.h>
77
78/* Mount options that we support. */
79static const char *msdosfs_opts[] = {
80	"async", "noatime", "noclusterr", "noclusterw",
81	"export", "force", "from", "sync",
82	"cs_dos", "cs_local", "cs_win", "dirmask",
83	"gid", "kiconv", "large", "longname",
84	"longnames", "mask", "shortname", "shortnames",
85	"uid", "win95", "nowin95",
86	NULL
87};
88
89#if 1 /*def PC98*/
90/*
91 * XXX - The boot signature formatted by NEC PC-98 DOS looks like a
92 *       garbage or a random value :-{
93 *       If you want to use that broken-signatured media, define the
94 *       following symbol even though PC/AT.
95 *       (ex. mount PC-98 DOS formatted FD on PC/AT)
96 */
97#define	MSDOSFS_NOCHECKSIG
98#endif
99
100MALLOC_DEFINE(M_MSDOSFSMNT, "msdosfs_mount", "MSDOSFS mount structure");
101static MALLOC_DEFINE(M_MSDOSFSFAT, "msdosfs_fat", "MSDOSFS file allocation table");
102
103struct iconv_functions *msdosfs_iconv;
104
105static int	update_mp(struct mount *mp, struct thread *td);
106static int	mountmsdosfs(struct vnode *devvp, struct mount *mp,
107		    struct thread *td);
108static vfs_fhtovp_t	msdosfs_fhtovp;
109static vfs_mount_t	msdosfs_mount;
110static vfs_root_t	msdosfs_root;
111static vfs_statfs_t	msdosfs_statfs;
112static vfs_sync_t	msdosfs_sync;
113static vfs_unmount_t	msdosfs_unmount;
114
115/* Maximum length of a character set name (arbitrary). */
116#define	MAXCSLEN	64
117
118static int
119update_mp(struct mount *mp, struct thread *td)
120{
121	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
122	void *dos, *win, *local;
123	int error, v;
124
125	if (!vfs_getopt(mp->mnt_optnew, "kiconv", NULL, NULL)) {
126		if (msdosfs_iconv != NULL) {
127			error = vfs_getopt(mp->mnt_optnew,
128			    "cs_win", &win, NULL);
129			if (!error)
130				error = vfs_getopt(mp->mnt_optnew,
131				    "cs_local", &local, NULL);
132			if (!error)
133				error = vfs_getopt(mp->mnt_optnew,
134				    "cs_dos", &dos, NULL);
135			if (!error) {
136				msdosfs_iconv->open(win, local, &pmp->pm_u2w);
137				msdosfs_iconv->open(local, win, &pmp->pm_w2u);
138				msdosfs_iconv->open(dos, local, &pmp->pm_u2d);
139				msdosfs_iconv->open(local, dos, &pmp->pm_d2u);
140			}
141			if (error != 0)
142				return (error);
143		} else {
144			pmp->pm_w2u = NULL;
145			pmp->pm_u2w = NULL;
146			pmp->pm_d2u = NULL;
147			pmp->pm_u2d = NULL;
148		}
149	}
150
151	if (1 == vfs_scanopt(mp->mnt_optnew, "gid", "%d", &v))
152		pmp->pm_gid = v;
153	if (1 == vfs_scanopt(mp->mnt_optnew, "uid", "%d", &v))
154		pmp->pm_uid = v;
155	if (1 == vfs_scanopt(mp->mnt_optnew, "mask", "%d", &v))
156		pmp->pm_mask = v & ALLPERMS;
157	if (1 == vfs_scanopt(mp->mnt_optnew, "dirmask", "%d", &v))
158		pmp->pm_dirmask = v & ALLPERMS;
159	vfs_flagopt(mp->mnt_optnew, "shortname",
160	    &pmp->pm_flags, MSDOSFSMNT_SHORTNAME);
161	vfs_flagopt(mp->mnt_optnew, "shortnames",
162	    &pmp->pm_flags, MSDOSFSMNT_SHORTNAME);
163	vfs_flagopt(mp->mnt_optnew, "longname",
164	    &pmp->pm_flags, MSDOSFSMNT_LONGNAME);
165	vfs_flagopt(mp->mnt_optnew, "longnames",
166	    &pmp->pm_flags, MSDOSFSMNT_LONGNAME);
167	vfs_flagopt(mp->mnt_optnew, "kiconv",
168	    &pmp->pm_flags, MSDOSFSMNT_KICONV);
169
170	if (vfs_getopt(mp->mnt_optnew, "nowin95", NULL, NULL) == 0)
171		pmp->pm_flags |= MSDOSFSMNT_NOWIN95;
172	else
173		pmp->pm_flags &= ~MSDOSFSMNT_NOWIN95;
174
175	if (pmp->pm_flags & MSDOSFSMNT_NOWIN95)
176		pmp->pm_flags |= MSDOSFSMNT_SHORTNAME;
177	else if (!(pmp->pm_flags &
178	    (MSDOSFSMNT_SHORTNAME | MSDOSFSMNT_LONGNAME))) {
179		struct vnode *rootvp;
180
181		/*
182		 * Try to divine whether to support Win'95 long filenames
183		 */
184		if (FAT32(pmp))
185			pmp->pm_flags |= MSDOSFSMNT_LONGNAME;
186		else {
187			if ((error =
188			    msdosfs_root(mp, LK_EXCLUSIVE, &rootvp, td)) != 0)
189				return error;
190			pmp->pm_flags |= findwin95(VTODE(rootvp)) ?
191			    MSDOSFSMNT_LONGNAME : MSDOSFSMNT_SHORTNAME;
192			vput(rootvp);
193		}
194	}
195	return 0;
196}
197
198static int
199msdosfs_cmount(struct mntarg *ma, void *data, int flags, struct thread *td)
200{
201	struct msdosfs_args args;
202	int error;
203
204	if (data == NULL)
205		return (EINVAL);
206	error = copyin(data, &args, sizeof args);
207	if (error)
208		return (error);
209
210	ma = mount_argsu(ma, "from", args.fspec, MAXPATHLEN);
211	ma = mount_arg(ma, "export", &args.export, sizeof args.export);
212	ma = mount_argf(ma, "uid", "%d", args.uid);
213	ma = mount_argf(ma, "gid", "%d", args.gid);
214	ma = mount_argf(ma, "mask", "%d", args.mask);
215	ma = mount_argf(ma, "dirmask", "%d", args.dirmask);
216
217	ma = mount_argb(ma, args.flags & MSDOSFSMNT_SHORTNAME, "noshortname");
218	ma = mount_argb(ma, args.flags & MSDOSFSMNT_LONGNAME, "nolongname");
219	ma = mount_argb(ma, !(args.flags & MSDOSFSMNT_NOWIN95), "nowin95");
220	ma = mount_argb(ma, args.flags & MSDOSFSMNT_KICONV, "nokiconv");
221
222	ma = mount_argsu(ma, "cs_win", args.cs_win, MAXCSLEN);
223	ma = mount_argsu(ma, "cs_dos", args.cs_dos, MAXCSLEN);
224	ma = mount_argsu(ma, "cs_local", args.cs_local, MAXCSLEN);
225
226	error = kernel_mount(ma, flags);
227
228	return (error);
229}
230
231/*
232 * mp - path - addr in user space of mount point (ie /usr or whatever)
233 * data - addr in user space of mount params including the name of the block
234 * special file to treat as a filesystem.
235 */
236static int
237msdosfs_mount(struct mount *mp, struct thread *td)
238{
239	struct vnode *devvp;	  /* vnode for blk device to mount */
240	/* msdosfs specific mount control block */
241	struct msdosfsmount *pmp = NULL;
242	struct nameidata ndp;
243	int error, flags;
244	mode_t accessmode;
245	char *from;
246
247	if (vfs_filteropt(mp->mnt_optnew, msdosfs_opts))
248		return (EINVAL);
249
250	/*
251	 * If updating, check whether changing from read-only to
252	 * read/write; if there is no device name, that's all we do.
253	 */
254	if (mp->mnt_flag & MNT_UPDATE) {
255		pmp = VFSTOMSDOSFS(mp);
256		if (vfs_flagopt(mp->mnt_optnew, "export", NULL, 0)) {
257			/*
258			 * Forbid export requests if filesystem has
259			 * MSDOSFS_LARGEFS flag set.
260			 */
261			if ((pmp->pm_flags & MSDOSFS_LARGEFS) != 0) {
262				vfs_mount_error(mp,
263				    "MSDOSFS_LARGEFS flag set, cannot export");
264				return (EOPNOTSUPP);
265			}
266		}
267		if (!(pmp->pm_flags & MSDOSFSMNT_RONLY) &&
268		    vfs_flagopt(mp->mnt_optnew, "ro", NULL, 0)) {
269			error = VFS_SYNC(mp, MNT_WAIT, td);
270			if (error)
271				return (error);
272			flags = WRITECLOSE;
273			if (mp->mnt_flag & MNT_FORCE)
274				flags |= FORCECLOSE;
275			error = vflush(mp, 0, flags, td);
276			if (error)
277				return (error);
278
279			/*
280			 * Now the volume is clean.  Mark it so while the
281			 * device is still rw.
282			 */
283			error = markvoldirty(pmp, 0);
284			if (error) {
285				(void)markvoldirty(pmp, 1);
286				return (error);
287			}
288
289			/* Downgrade the device from rw to ro. */
290			DROP_GIANT();
291			g_topology_lock();
292			error = g_access(pmp->pm_cp, 0, -1, 0);
293			g_topology_unlock();
294			PICKUP_GIANT();
295			if (error) {
296				(void)markvoldirty(pmp, 1);
297				return (error);
298			}
299
300			/*
301			 * Backing out after an error was painful in the
302			 * above.  Now we are committed to succeeding.
303			 */
304			pmp->pm_fmod = 0;
305			pmp->pm_flags |= MSDOSFSMNT_RONLY;
306			MNT_ILOCK(mp);
307			mp->mnt_flag |= MNT_RDONLY;
308			MNT_IUNLOCK(mp);
309		} else if ((pmp->pm_flags & MSDOSFSMNT_RONLY) &&
310		    !vfs_flagopt(mp->mnt_optnew, "ro", NULL, 0)) {
311			/*
312			 * If upgrade to read-write by non-root, then verify
313			 * that user has necessary permissions on the device.
314			 */
315			devvp = pmp->pm_devvp;
316			vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
317			error = VOP_ACCESS(devvp, VREAD | VWRITE,
318			    td->td_ucred, td);
319			if (error)
320				error = priv_check(td, PRIV_VFS_MOUNT_PERM);
321			if (error) {
322				VOP_UNLOCK(devvp, 0);
323				return (error);
324			}
325			VOP_UNLOCK(devvp, 0);
326			DROP_GIANT();
327			g_topology_lock();
328			error = g_access(pmp->pm_cp, 0, 1, 0);
329			g_topology_unlock();
330			PICKUP_GIANT();
331			if (error)
332				return (error);
333
334			pmp->pm_fmod = 1;
335			pmp->pm_flags &= ~MSDOSFSMNT_RONLY;
336			MNT_ILOCK(mp);
337			mp->mnt_flag &= ~MNT_RDONLY;
338			MNT_IUNLOCK(mp);
339
340			/* Now that the volume is modifiable, mark it dirty. */
341			error = markvoldirty(pmp, 1);
342			if (error)
343				return (error);
344		}
345	}
346	/*
347	 * Not an update, or updating the name: look up the name
348	 * and verify that it refers to a sensible disk device.
349	 */
350	if (vfs_getopt(mp->mnt_optnew, "from", (void **)&from, NULL))
351		return (EINVAL);
352	NDINIT(&ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, from, td);
353	error = namei(&ndp);
354	if (error)
355		return (error);
356	devvp = ndp.ni_vp;
357	NDFREE(&ndp, NDF_ONLY_PNBUF);
358
359	if (!vn_isdisk(devvp, &error)) {
360		vput(devvp);
361		return (error);
362	}
363	/*
364	 * If mount by non-root, then verify that user has necessary
365	 * permissions on the device.
366	 */
367	accessmode = VREAD;
368	if ((mp->mnt_flag & MNT_RDONLY) == 0)
369		accessmode |= VWRITE;
370	error = VOP_ACCESS(devvp, accessmode, td->td_ucred, td);
371	if (error)
372		error = priv_check(td, PRIV_VFS_MOUNT_PERM);
373	if (error) {
374		vput(devvp);
375		return (error);
376	}
377	if ((mp->mnt_flag & MNT_UPDATE) == 0) {
378		error = mountmsdosfs(devvp, mp, td);
379#ifdef MSDOSFS_DEBUG		/* only needed for the printf below */
380		pmp = VFSTOMSDOSFS(mp);
381#endif
382	} else {
383		if (devvp != pmp->pm_devvp)
384			error = EINVAL;	/* XXX needs translation */
385		else
386			vput(devvp);
387	}
388	if (error) {
389		vrele(devvp);
390		return (error);
391	}
392
393	error = update_mp(mp, td);
394	if (error) {
395		if ((mp->mnt_flag & MNT_UPDATE) == 0)
396			msdosfs_unmount(mp, MNT_FORCE, td);
397		return error;
398	}
399
400	vfs_mountedfrom(mp, from);
401#ifdef MSDOSFS_DEBUG
402	printf("msdosfs_mount(): mp %p, pmp %p, inusemap %p\n", mp, pmp, pmp->pm_inusemap);
403#endif
404	return (0);
405}
406
407static int
408mountmsdosfs(struct vnode *devvp, struct mount *mp, struct thread *td)
409{
410	struct msdosfsmount *pmp;
411	struct buf *bp;
412	struct cdev *dev = devvp->v_rdev;
413	union bootsector *bsp;
414	struct byte_bpb33 *b33;
415	struct byte_bpb50 *b50;
416	struct byte_bpb710 *b710;
417	u_int8_t SecPerClust;
418	u_long clusters;
419	int ronly, error;
420	struct g_consumer *cp;
421	struct bufobj *bo;
422
423	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
424	/* XXX: use VOP_ACCESS to check FS perms */
425	DROP_GIANT();
426	g_topology_lock();
427	error = g_vfs_open(devvp, &cp, "msdosfs", ronly ? 0 : 1);
428	g_topology_unlock();
429	PICKUP_GIANT();
430	VOP_UNLOCK(devvp, 0);
431	if (error)
432		return (error);
433
434	bo = &devvp->v_bufobj;
435	bp = NULL;		/* This and pmp both used in error_exit. */
436	pmp = NULL;
437
438	/*
439	 * Read the boot sector of the filesystem, and then check the
440	 * boot signature.  If not a dos boot sector then error out.
441	 *
442	 * NOTE: 8192 is a magic size that works for ffs.
443	 */
444	error = bread(devvp, 0, 8192, NOCRED, &bp);
445	if (error)
446		goto error_exit;
447	bp->b_flags |= B_AGE;
448	bsp = (union bootsector *)bp->b_data;
449	b33 = (struct byte_bpb33 *)bsp->bs33.bsBPB;
450	b50 = (struct byte_bpb50 *)bsp->bs50.bsBPB;
451	b710 = (struct byte_bpb710 *)bsp->bs710.bsBPB;
452
453#ifndef MSDOSFS_NOCHECKSIG
454	if (bsp->bs50.bsBootSectSig0 != BOOTSIG0
455	    || bsp->bs50.bsBootSectSig1 != BOOTSIG1) {
456		error = EINVAL;
457		goto error_exit;
458	}
459#endif
460
461	pmp = malloc(sizeof *pmp, M_MSDOSFSMNT, M_WAITOK | M_ZERO);
462	pmp->pm_mountp = mp;
463	pmp->pm_cp = cp;
464	pmp->pm_bo = bo;
465
466	/*
467	 * Initialize ownerships and permissions, since nothing else will
468	 * initialize them iff we are mounting root.
469	 */
470	pmp->pm_uid = UID_ROOT;
471	pmp->pm_gid = GID_WHEEL;
472	pmp->pm_mask = pmp->pm_dirmask = S_IXUSR | S_IXGRP | S_IXOTH |
473	    S_IRUSR | S_IRGRP | S_IROTH | S_IWUSR;
474
475	/*
476	 * Experimental support for large MS-DOS filesystems.
477	 * WARNING: This uses at least 32 bytes of kernel memory (which is not
478	 * reclaimed until the FS is unmounted) for each file on disk to map
479	 * between the 32-bit inode numbers used by VFS and the 64-bit
480	 * pseudo-inode numbers used internally by msdosfs. This is only
481	 * safe to use in certain controlled situations (e.g. read-only FS
482	 * with less than 1 million files).
483	 * Since the mappings do not persist across unmounts (or reboots), these
484	 * filesystems are not suitable for exporting through NFS, or any other
485	 * application that requires fixed inode numbers.
486	 */
487	vfs_flagopt(mp->mnt_optnew, "large", &pmp->pm_flags, MSDOSFS_LARGEFS);
488
489	/*
490	 * Compute several useful quantities from the bpb in the
491	 * bootsector.  Copy in the dos 5 variant of the bpb then fix up
492	 * the fields that are different between dos 5 and dos 3.3.
493	 */
494	SecPerClust = b50->bpbSecPerClust;
495	pmp->pm_BytesPerSec = getushort(b50->bpbBytesPerSec);
496	if (pmp->pm_BytesPerSec < DEV_BSIZE) {
497		error = EINVAL;
498		goto error_exit;
499	}
500	pmp->pm_ResSectors = getushort(b50->bpbResSectors);
501	pmp->pm_FATs = b50->bpbFATs;
502	pmp->pm_RootDirEnts = getushort(b50->bpbRootDirEnts);
503	pmp->pm_Sectors = getushort(b50->bpbSectors);
504	pmp->pm_FATsecs = getushort(b50->bpbFATsecs);
505	pmp->pm_SecPerTrack = getushort(b50->bpbSecPerTrack);
506	pmp->pm_Heads = getushort(b50->bpbHeads);
507	pmp->pm_Media = b50->bpbMedia;
508
509	/* calculate the ratio of sector size to DEV_BSIZE */
510	pmp->pm_BlkPerSec = pmp->pm_BytesPerSec / DEV_BSIZE;
511
512	/*
513	 * We don't check pm_Heads nor pm_SecPerTrack, because
514	 * these may not be set for EFI file systems. We don't
515	 * use these anyway, so we're unaffected if they are
516	 * invalid.
517	 */
518	if (!pmp->pm_BytesPerSec || !SecPerClust) {
519		error = EINVAL;
520		goto error_exit;
521	}
522
523	if (pmp->pm_Sectors == 0) {
524		pmp->pm_HiddenSects = getulong(b50->bpbHiddenSecs);
525		pmp->pm_HugeSectors = getulong(b50->bpbHugeSectors);
526	} else {
527		pmp->pm_HiddenSects = getushort(b33->bpbHiddenSecs);
528		pmp->pm_HugeSectors = pmp->pm_Sectors;
529	}
530	if (!(pmp->pm_flags & MSDOSFS_LARGEFS)) {
531		if (pmp->pm_HugeSectors > 0xffffffff /
532		    (pmp->pm_BytesPerSec / sizeof(struct direntry)) + 1) {
533			/*
534			 * We cannot deal currently with this size of disk
535			 * due to fileid limitations (see msdosfs_getattr and
536			 * msdosfs_readdir)
537			 */
538			error = EINVAL;
539			vfs_mount_error(mp,
540			    "Disk too big, try '-o large' mount option");
541			goto error_exit;
542		}
543	}
544
545	if (pmp->pm_RootDirEnts == 0) {
546		if (pmp->pm_Sectors
547		    || pmp->pm_FATsecs
548		    || getushort(b710->bpbFSVers)) {
549			error = EINVAL;
550			printf("mountmsdosfs(): bad FAT32 filesystem\n");
551			goto error_exit;
552		}
553		pmp->pm_fatmask = FAT32_MASK;
554		pmp->pm_fatmult = 4;
555		pmp->pm_fatdiv = 1;
556		pmp->pm_FATsecs = getulong(b710->bpbBigFATsecs);
557		if (getushort(b710->bpbExtFlags) & FATMIRROR)
558			pmp->pm_curfat = getushort(b710->bpbExtFlags) & FATNUM;
559		else
560			pmp->pm_flags |= MSDOSFS_FATMIRROR;
561	} else
562		pmp->pm_flags |= MSDOSFS_FATMIRROR;
563
564	/*
565	 * Check a few values (could do some more):
566	 * - logical sector size: power of 2, >= block size
567	 * - sectors per cluster: power of 2, >= 1
568	 * - number of sectors:   >= 1, <= size of partition
569	 * - number of FAT sectors: >= 1
570	 */
571	if ( (SecPerClust == 0)
572	  || (SecPerClust & (SecPerClust - 1))
573	  || (pmp->pm_BytesPerSec < DEV_BSIZE)
574	  || (pmp->pm_BytesPerSec & (pmp->pm_BytesPerSec - 1))
575	  || (pmp->pm_HugeSectors == 0)
576	  || (pmp->pm_FATsecs == 0)
577	) {
578		error = EINVAL;
579		goto error_exit;
580	}
581
582	pmp->pm_HugeSectors *= pmp->pm_BlkPerSec;
583	pmp->pm_HiddenSects *= pmp->pm_BlkPerSec;	/* XXX not used? */
584	pmp->pm_FATsecs     *= pmp->pm_BlkPerSec;
585	SecPerClust         *= pmp->pm_BlkPerSec;
586
587	pmp->pm_fatblk = pmp->pm_ResSectors * pmp->pm_BlkPerSec;
588
589	if (FAT32(pmp)) {
590		pmp->pm_rootdirblk = getulong(b710->bpbRootClust);
591		pmp->pm_firstcluster = pmp->pm_fatblk
592			+ (pmp->pm_FATs * pmp->pm_FATsecs);
593		pmp->pm_fsinfo = getushort(b710->bpbFSInfo) * pmp->pm_BlkPerSec;
594	} else {
595		pmp->pm_rootdirblk = pmp->pm_fatblk +
596			(pmp->pm_FATs * pmp->pm_FATsecs);
597		pmp->pm_rootdirsize = (pmp->pm_RootDirEnts * sizeof(struct direntry)
598				       + DEV_BSIZE - 1)
599			/ DEV_BSIZE; /* in blocks */
600		pmp->pm_firstcluster = pmp->pm_rootdirblk + pmp->pm_rootdirsize;
601	}
602
603	pmp->pm_maxcluster = (pmp->pm_HugeSectors - pmp->pm_firstcluster) /
604	    SecPerClust + 1;
605	pmp->pm_fatsize = pmp->pm_FATsecs * DEV_BSIZE;	/* XXX not used? */
606
607	if (pmp->pm_fatmask == 0) {
608		if (pmp->pm_maxcluster
609		    <= ((CLUST_RSRVD - CLUST_FIRST) & FAT12_MASK)) {
610			/*
611			 * This will usually be a floppy disk. This size makes
612			 * sure that one fat entry will not be split across
613			 * multiple blocks.
614			 */
615			pmp->pm_fatmask = FAT12_MASK;
616			pmp->pm_fatmult = 3;
617			pmp->pm_fatdiv = 2;
618		} else {
619			pmp->pm_fatmask = FAT16_MASK;
620			pmp->pm_fatmult = 2;
621			pmp->pm_fatdiv = 1;
622		}
623	}
624
625	clusters = (pmp->pm_fatsize / pmp->pm_fatmult) * pmp->pm_fatdiv;
626	if (pmp->pm_maxcluster >= clusters) {
627		printf("Warning: number of clusters (%ld) exceeds FAT "
628		    "capacity (%ld)\n", pmp->pm_maxcluster + 1, clusters);
629		pmp->pm_maxcluster = clusters - 1;
630	}
631
632	if (FAT12(pmp))
633		pmp->pm_fatblocksize = 3 * 512;
634	else
635		pmp->pm_fatblocksize = PAGE_SIZE;
636	pmp->pm_fatblocksize = roundup(pmp->pm_fatblocksize,
637	    pmp->pm_BytesPerSec);
638	pmp->pm_fatblocksec = pmp->pm_fatblocksize / DEV_BSIZE;
639	pmp->pm_bnshift = ffs(DEV_BSIZE) - 1;
640
641	/*
642	 * Compute mask and shift value for isolating cluster relative byte
643	 * offsets and cluster numbers from a file offset.
644	 */
645	pmp->pm_bpcluster = SecPerClust * DEV_BSIZE;
646	pmp->pm_crbomask = pmp->pm_bpcluster - 1;
647	pmp->pm_cnshift = ffs(pmp->pm_bpcluster) - 1;
648
649	/*
650	 * Check for valid cluster size
651	 * must be a power of 2
652	 */
653	if (pmp->pm_bpcluster ^ (1 << pmp->pm_cnshift)) {
654		error = EINVAL;
655		goto error_exit;
656	}
657
658	/*
659	 * Release the bootsector buffer.
660	 */
661	brelse(bp);
662	bp = NULL;
663
664	/*
665	 * Check the fsinfo sector if we have one.  Silently fix up our
666	 * in-core copy of fp->fsinxtfree if it is unknown (0xffffffff)
667	 * or too large.  Ignore fp->fsinfree for now, since we need to
668	 * read the entire FAT anyway to fill the inuse map.
669	 */
670	if (pmp->pm_fsinfo) {
671		struct fsinfo *fp;
672
673		if ((error = bread(devvp, pmp->pm_fsinfo, pmp->pm_BytesPerSec,
674		    NOCRED, &bp)) != 0)
675			goto error_exit;
676		fp = (struct fsinfo *)bp->b_data;
677		if (!bcmp(fp->fsisig1, "RRaA", 4)
678		    && !bcmp(fp->fsisig2, "rrAa", 4)
679		    && !bcmp(fp->fsisig3, "\0\0\125\252", 4)) {
680			pmp->pm_nxtfree = getulong(fp->fsinxtfree);
681			if (pmp->pm_nxtfree > pmp->pm_maxcluster)
682				pmp->pm_nxtfree = CLUST_FIRST;
683		} else
684			pmp->pm_fsinfo = 0;
685		brelse(bp);
686		bp = NULL;
687	}
688
689	/*
690	 * Finish initializing pmp->pm_nxtfree (just in case the first few
691	 * sectors aren't properly reserved in the FAT).  This completes
692	 * the fixup for fp->fsinxtfree, and fixes up the zero-initialized
693	 * value if there is no fsinfo.  We will use pmp->pm_nxtfree
694	 * internally even if there is no fsinfo.
695	 */
696	if (pmp->pm_nxtfree < CLUST_FIRST)
697		pmp->pm_nxtfree = CLUST_FIRST;
698
699	/*
700	 * Allocate memory for the bitmap of allocated clusters, and then
701	 * fill it in.
702	 */
703	pmp->pm_inusemap = malloc(howmany(pmp->pm_maxcluster + 1, N_INUSEBITS)
704				  * sizeof(*pmp->pm_inusemap),
705				  M_MSDOSFSFAT, M_WAITOK);
706
707	/*
708	 * fillinusemap() needs pm_devvp.
709	 */
710	pmp->pm_devvp = devvp;
711
712	/*
713	 * Have the inuse map filled in.
714	 */
715	if ((error = fillinusemap(pmp)) != 0)
716		goto error_exit;
717
718	/*
719	 * If they want fat updates to be synchronous then let them suffer
720	 * the performance degradation in exchange for the on disk copy of
721	 * the fat being correct just about all the time.  I suppose this
722	 * would be a good thing to turn on if the kernel is still flakey.
723	 */
724	if (mp->mnt_flag & MNT_SYNCHRONOUS)
725		pmp->pm_flags |= MSDOSFSMNT_WAITONFAT;
726
727	/*
728	 * Finish up.
729	 */
730	if (ronly)
731		pmp->pm_flags |= MSDOSFSMNT_RONLY;
732	else {
733		if ((error = markvoldirty(pmp, 1)) != 0) {
734			(void)markvoldirty(pmp, 0);
735			goto error_exit;
736		}
737		pmp->pm_fmod = 1;
738	}
739	mp->mnt_data =  pmp;
740	mp->mnt_stat.f_fsid.val[0] = dev2udev(dev);
741	mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
742	MNT_ILOCK(mp);
743	mp->mnt_flag |= MNT_LOCAL;
744	MNT_IUNLOCK(mp);
745
746	if (pmp->pm_flags & MSDOSFS_LARGEFS)
747		msdosfs_fileno_init(mp);
748
749	return 0;
750
751error_exit:
752	if (bp)
753		brelse(bp);
754	if (cp != NULL) {
755		DROP_GIANT();
756		g_topology_lock();
757		g_vfs_close(cp, td);
758		g_topology_unlock();
759		PICKUP_GIANT();
760	}
761	if (pmp) {
762		if (pmp->pm_inusemap)
763			free(pmp->pm_inusemap, M_MSDOSFSFAT);
764		free(pmp, M_MSDOSFSMNT);
765		mp->mnt_data = NULL;
766	}
767	return (error);
768}
769
770/*
771 * Unmount the filesystem described by mp.
772 */
773static int
774msdosfs_unmount(struct mount *mp, int mntflags, struct thread *td)
775{
776	struct msdosfsmount *pmp;
777	int error, flags;
778
779	flags = 0;
780	if (mntflags & MNT_FORCE)
781		flags |= FORCECLOSE;
782	error = vflush(mp, 0, flags, td);
783	if (error)
784		return error;
785	pmp = VFSTOMSDOSFS(mp);
786	if ((pmp->pm_flags & MSDOSFSMNT_RONLY) == 0) {
787		error = markvoldirty(pmp, 0);
788		if (error) {
789			(void)markvoldirty(pmp, 1);
790			return (error);
791		}
792	}
793	if (pmp->pm_flags & MSDOSFSMNT_KICONV && msdosfs_iconv) {
794		if (pmp->pm_w2u)
795			msdosfs_iconv->close(pmp->pm_w2u);
796		if (pmp->pm_u2w)
797			msdosfs_iconv->close(pmp->pm_u2w);
798		if (pmp->pm_d2u)
799			msdosfs_iconv->close(pmp->pm_d2u);
800		if (pmp->pm_u2d)
801			msdosfs_iconv->close(pmp->pm_u2d);
802	}
803
804#ifdef MSDOSFS_DEBUG
805	{
806		struct vnode *vp = pmp->pm_devvp;
807		struct bufobj *bo;
808
809		bo = &vp->v_bufobj;
810		BO_LOCK(bo);
811		VI_LOCK(vp);
812		vn_printf(vp,
813		    "msdosfs_umount(): just before calling VOP_CLOSE()\n");
814		printf("freef %p, freeb %p, mount %p\n",
815		    TAILQ_NEXT(vp, v_freelist), vp->v_freelist.tqe_prev,
816		    vp->v_mount);
817		printf("cleanblkhd %p, dirtyblkhd %p, numoutput %ld, type %d\n",
818		    TAILQ_FIRST(&vp->v_bufobj.bo_clean.bv_hd),
819		    TAILQ_FIRST(&vp->v_bufobj.bo_dirty.bv_hd),
820		    vp->v_bufobj.bo_numoutput, vp->v_type);
821		VI_UNLOCK(vp);
822		BO_UNLOCK(bo);
823	}
824#endif
825	DROP_GIANT();
826	g_topology_lock();
827	g_vfs_close(pmp->pm_cp, td);
828	g_topology_unlock();
829	PICKUP_GIANT();
830	vrele(pmp->pm_devvp);
831	free(pmp->pm_inusemap, M_MSDOSFSFAT);
832	if (pmp->pm_flags & MSDOSFS_LARGEFS)
833		msdosfs_fileno_free(mp);
834	free(pmp, M_MSDOSFSMNT);
835	mp->mnt_data = NULL;
836	MNT_ILOCK(mp);
837	mp->mnt_flag &= ~MNT_LOCAL;
838	MNT_IUNLOCK(mp);
839	return (0);
840}
841
842static int
843msdosfs_root(struct mount *mp, int flags, struct vnode **vpp, struct thread *td)
844{
845	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
846	struct denode *ndep;
847	int error;
848
849#ifdef MSDOSFS_DEBUG
850	printf("msdosfs_root(); mp %p, pmp %p\n", mp, pmp);
851#endif
852	error = deget(pmp, MSDOSFSROOT, MSDOSFSROOT_OFS, &ndep);
853	if (error)
854		return (error);
855	*vpp = DETOV(ndep);
856	return (0);
857}
858
859static int
860msdosfs_statfs(struct mount *mp, struct statfs *sbp, struct thread *td)
861{
862	struct msdosfsmount *pmp;
863
864	pmp = VFSTOMSDOSFS(mp);
865	sbp->f_bsize = pmp->pm_bpcluster;
866	sbp->f_iosize = pmp->pm_bpcluster;
867	sbp->f_blocks = pmp->pm_maxcluster + 1;
868	sbp->f_bfree = pmp->pm_freeclustercount;
869	sbp->f_bavail = pmp->pm_freeclustercount;
870	sbp->f_files = pmp->pm_RootDirEnts;	/* XXX */
871	sbp->f_ffree = 0;	/* what to put in here? */
872	return (0);
873}
874
875static int
876msdosfs_sync(struct mount *mp, int waitfor, struct thread *td)
877{
878	struct vnode *vp, *nvp;
879	struct denode *dep;
880	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
881	int error, allerror = 0;
882
883	/*
884	 * If we ever switch to not updating all of the fats all the time,
885	 * this would be the place to update them from the first one.
886	 */
887	if (pmp->pm_fmod != 0) {
888		if (pmp->pm_flags & MSDOSFSMNT_RONLY)
889			panic("msdosfs_sync: rofs mod");
890		else {
891			/* update fats here */
892		}
893	}
894	/*
895	 * Write back each (modified) denode.
896	 */
897	MNT_ILOCK(mp);
898loop:
899	MNT_VNODE_FOREACH(vp, mp, nvp) {
900		VI_LOCK(vp);
901		if (vp->v_type == VNON || (vp->v_iflag & VI_DOOMED)) {
902			VI_UNLOCK(vp);
903			continue;
904		}
905		MNT_IUNLOCK(mp);
906		dep = VTODE(vp);
907		if ((dep->de_flag &
908		    (DE_ACCESS | DE_CREATE | DE_UPDATE | DE_MODIFIED)) == 0 &&
909		    (vp->v_bufobj.bo_dirty.bv_cnt == 0 ||
910		    waitfor == MNT_LAZY)) {
911			VI_UNLOCK(vp);
912			MNT_ILOCK(mp);
913			continue;
914		}
915		error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK, td);
916		if (error) {
917			MNT_ILOCK(mp);
918			if (error == ENOENT)
919				goto loop;
920			continue;
921		}
922		error = VOP_FSYNC(vp, waitfor, td);
923		if (error)
924			allerror = error;
925		VOP_UNLOCK(vp, 0);
926		vrele(vp);
927		MNT_ILOCK(mp);
928	}
929	MNT_IUNLOCK(mp);
930
931	/*
932	 * Flush filesystem control info.
933	 */
934	if (waitfor != MNT_LAZY) {
935		vn_lock(pmp->pm_devvp, LK_EXCLUSIVE | LK_RETRY);
936		error = VOP_FSYNC(pmp->pm_devvp, waitfor, td);
937		if (error)
938			allerror = error;
939		VOP_UNLOCK(pmp->pm_devvp, 0);
940	}
941	return (allerror);
942}
943
944static int
945msdosfs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
946{
947	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
948	struct defid *defhp = (struct defid *) fhp;
949	struct denode *dep;
950	int error;
951
952	error = deget(pmp, defhp->defid_dirclust, defhp->defid_dirofs, &dep);
953	if (error) {
954		*vpp = NULLVP;
955		return (error);
956	}
957	*vpp = DETOV(dep);
958	vnode_create_vobject(*vpp, dep->de_FileSize, curthread);
959	return (0);
960}
961
962static struct vfsops msdosfs_vfsops = {
963	.vfs_fhtovp =		msdosfs_fhtovp,
964	.vfs_mount =		msdosfs_mount,
965	.vfs_cmount =		msdosfs_cmount,
966	.vfs_root =		msdosfs_root,
967	.vfs_statfs =		msdosfs_statfs,
968	.vfs_sync =		msdosfs_sync,
969	.vfs_unmount =		msdosfs_unmount,
970};
971
972VFS_SET(msdosfs_vfsops, msdosfs, 0);
973MODULE_VERSION(msdosfs, 1);
974