msdosfs_vfsops.c revision 173690
1/* $FreeBSD: head/sys/fs/msdosfs/msdosfs_vfsops.c 173690 2007-11-17 16:19:48Z maxim $ */
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/iconv.h>
56#include <sys/kernel.h>
57#include <sys/lock.h>
58#include <sys/malloc.h>
59#include <sys/mount.h>
60#include <sys/mutex.h>
61#include <sys/namei.h>
62#include <sys/priv.h>
63#include <sys/proc.h>
64#include <sys/stat.h>
65#include <sys/vnode.h>
66
67#include <geom/geom.h>
68#include <geom/geom_vfs.h>
69
70#include <fs/msdosfs/bootsect.h>
71#include <fs/msdosfs/bpb.h>
72#include <fs/msdosfs/direntry.h>
73#include <fs/msdosfs/denode.h>
74#include <fs/msdosfs/fat.h>
75#include <fs/msdosfs/msdosfsmount.h>
76
77/* Mount options that we support. */
78static const char *msdosfs_opts[] = {
79	"async", "noatime", "noclusterr", "noclusterw",
80	"export", "force", "from", "sync",
81	"cs_dos", "cs_local", "cs_win", "dirmask",
82	"gid", "kiconv", "large", "longname",
83	"longnames", "mask", "shortname", "shortnames",
84	"uid", "win95", "nowin95",
85	NULL
86};
87
88#if 1 /*def PC98*/
89/*
90 * XXX - The boot signature formatted by NEC PC-98 DOS looks like a
91 *       garbage or a random value :-{
92 *       If you want to use that broken-signatured media, define the
93 *       following symbol even though PC/AT.
94 *       (ex. mount PC-98 DOS formatted FD on PC/AT)
95 */
96#define	MSDOSFS_NOCHECKSIG
97#endif
98
99MALLOC_DEFINE(M_MSDOSFSMNT, "msdosfs_mount", "MSDOSFS mount structure");
100static MALLOC_DEFINE(M_MSDOSFSFAT, "msdosfs_fat", "MSDOSFS file allocation table");
101
102struct iconv_functions *msdosfs_iconv;
103
104static int	update_mp(struct mount *mp, struct thread *td);
105static int	mountmsdosfs(struct vnode *devvp, struct mount *mp,
106		    struct thread *td);
107static vfs_fhtovp_t	msdosfs_fhtovp;
108static vfs_mount_t	msdosfs_mount;
109static vfs_root_t	msdosfs_root;
110static vfs_statfs_t	msdosfs_statfs;
111static vfs_sync_t	msdosfs_sync;
112static vfs_unmount_t	msdosfs_unmount;
113
114/* Maximum length of a character set name (arbitrary). */
115#define	MAXCSLEN	64
116
117static int
118update_mp(struct mount *mp, struct thread *td)
119{
120	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
121	void *dos, *win, *local;
122	int error, v;
123
124	if (!vfs_getopt(mp->mnt_optnew, "kiconv", NULL, NULL)) {
125		if (msdosfs_iconv != NULL) {
126			error = vfs_getopt(mp->mnt_optnew,
127			    "cs_win", &win, NULL);
128			if (!error)
129				error = vfs_getopt(mp->mnt_optnew,
130				    "cs_local", &local, NULL);
131			if (!error)
132				error = vfs_getopt(mp->mnt_optnew,
133				    "cs_dos", &dos, NULL);
134			if (!error) {
135				msdosfs_iconv->open(win, local, &pmp->pm_u2w);
136				msdosfs_iconv->open(local, win, &pmp->pm_w2u);
137				msdosfs_iconv->open(dos, local, &pmp->pm_u2d);
138				msdosfs_iconv->open(local, dos, &pmp->pm_d2u);
139			}
140			if (error != 0)
141				return (error);
142		} else {
143			pmp->pm_w2u = NULL;
144			pmp->pm_u2w = NULL;
145			pmp->pm_d2u = NULL;
146			pmp->pm_u2d = NULL;
147		}
148	}
149
150	if (1 == vfs_scanopt(mp->mnt_optnew, "gid", "%d", &v))
151		pmp->pm_gid = v;
152	if (1 == vfs_scanopt(mp->mnt_optnew, "uid", "%d", &v))
153		pmp->pm_uid = v;
154	if (1 == vfs_scanopt(mp->mnt_optnew, "mask", "%d", &v))
155		pmp->pm_mask = v & ALLPERMS;
156	if (1 == vfs_scanopt(mp->mnt_optnew, "dirmask", "%d", &v))
157		pmp->pm_dirmask = v & ALLPERMS;
158	vfs_flagopt(mp->mnt_optnew, "shortname",
159	    &pmp->pm_flags, MSDOSFSMNT_SHORTNAME);
160	vfs_flagopt(mp->mnt_optnew, "shortnames",
161	    &pmp->pm_flags, MSDOSFSMNT_SHORTNAME);
162	vfs_flagopt(mp->mnt_optnew, "longname",
163	    &pmp->pm_flags, MSDOSFSMNT_LONGNAME);
164	vfs_flagopt(mp->mnt_optnew, "longnames",
165	    &pmp->pm_flags, MSDOSFSMNT_LONGNAME);
166	vfs_flagopt(mp->mnt_optnew, "kiconv",
167	    &pmp->pm_flags, MSDOSFSMNT_KICONV);
168
169	if (vfs_getopt(mp->mnt_optnew, "nowin95", NULL, NULL) == 0)
170		pmp->pm_flags |= MSDOSFSMNT_NOWIN95;
171	else
172		pmp->pm_flags &= ~MSDOSFSMNT_NOWIN95;
173
174	if (pmp->pm_flags & MSDOSFSMNT_NOWIN95)
175		pmp->pm_flags |= MSDOSFSMNT_SHORTNAME;
176	else if (!(pmp->pm_flags &
177	    (MSDOSFSMNT_SHORTNAME | MSDOSFSMNT_LONGNAME))) {
178		struct vnode *rootvp;
179
180		/*
181		 * Try to divine whether to support Win'95 long filenames
182		 */
183		if (FAT32(pmp))
184			pmp->pm_flags |= MSDOSFSMNT_LONGNAME;
185		else {
186			if ((error =
187			    msdosfs_root(mp, LK_EXCLUSIVE, &rootvp, td)) != 0)
188				return error;
189			pmp->pm_flags |= findwin95(VTODE(rootvp)) ?
190			    MSDOSFSMNT_LONGNAME : MSDOSFSMNT_SHORTNAME;
191			vput(rootvp);
192		}
193	}
194	return 0;
195}
196
197static int
198msdosfs_cmount(struct mntarg *ma, void *data, int flags, struct thread *td)
199{
200	struct msdosfs_args args;
201	int error;
202
203	if (data == NULL)
204		return (EINVAL);
205	error = copyin(data, &args, sizeof args);
206	if (error)
207		return (error);
208
209	ma = mount_argsu(ma, "from", args.fspec, MAXPATHLEN);
210	ma = mount_arg(ma, "export", &args.export, sizeof args.export);
211	ma = mount_argf(ma, "uid", "%d", args.uid);
212	ma = mount_argf(ma, "gid", "%d", args.gid);
213	ma = mount_argf(ma, "mask", "%d", args.mask);
214	ma = mount_argf(ma, "dirmask", "%d", args.dirmask);
215
216	ma = mount_argb(ma, args.flags & MSDOSFSMNT_SHORTNAME, "noshortname");
217	ma = mount_argb(ma, args.flags & MSDOSFSMNT_LONGNAME, "nolongname");
218	ma = mount_argb(ma, !(args.flags & MSDOSFSMNT_NOWIN95), "nowin95");
219	ma = mount_argb(ma, args.flags & MSDOSFSMNT_KICONV, "nokiconv");
220
221	ma = mount_argsu(ma, "cs_win", args.cs_win, MAXCSLEN);
222	ma = mount_argsu(ma, "cs_dos", args.cs_dos, MAXCSLEN);
223	ma = mount_argsu(ma, "cs_local", args.cs_local, MAXCSLEN);
224
225	error = kernel_mount(ma, flags);
226
227	return (error);
228}
229
230/*
231 * mp - path - addr in user space of mount point (ie /usr or whatever)
232 * data - addr in user space of mount params including the name of the block
233 * special file to treat as a filesystem.
234 */
235static int
236msdosfs_mount(struct mount *mp, struct thread *td)
237{
238	struct vnode *devvp;	  /* vnode for blk device to mount */
239	/* msdosfs specific mount control block */
240	struct msdosfsmount *pmp = NULL;
241	struct nameidata ndp;
242	int error, flags;
243	mode_t accessmode;
244	char *from;
245
246	if (vfs_filteropt(mp->mnt_optnew, msdosfs_opts))
247		return (EINVAL);
248
249	/*
250	 * If updating, check whether changing from read-only to
251	 * read/write; if there is no device name, that's all we do.
252	 */
253	if (mp->mnt_flag & MNT_UPDATE) {
254		pmp = VFSTOMSDOSFS(mp);
255		if (vfs_flagopt(mp->mnt_optnew, "export", NULL, 0)) {
256			/*
257			 * Forbid export requests if filesystem has
258			 * MSDOSFS_LARGEFS flag set.
259			 */
260			if ((pmp->pm_flags & MSDOSFS_LARGEFS) != 0) {
261				vfs_mount_error(mp,
262				    "MSDOSFS_LARGEFS flag set, cannot export");
263				return (EOPNOTSUPP);
264			}
265		}
266		if (!(pmp->pm_flags & MSDOSFSMNT_RONLY) &&
267		    vfs_flagopt(mp->mnt_optnew, "ro", NULL, 0)) {
268			error = VFS_SYNC(mp, MNT_WAIT, td);
269			if (error)
270				return (error);
271			flags = WRITECLOSE;
272			if (mp->mnt_flag & MNT_FORCE)
273				flags |= FORCECLOSE;
274			error = vflush(mp, 0, flags, td);
275			if (error)
276				return (error);
277
278			/*
279			 * Now the volume is clean.  Mark it so while the
280			 * device is still rw.
281			 */
282			error = markvoldirty(pmp, 0);
283			if (error) {
284				(void)markvoldirty(pmp, 1);
285				return (error);
286			}
287
288			/* Downgrade the device from rw to ro. */
289			DROP_GIANT();
290			g_topology_lock();
291			error = g_access(pmp->pm_cp, 0, -1, 0);
292			g_topology_unlock();
293			PICKUP_GIANT();
294			if (error) {
295				(void)markvoldirty(pmp, 1);
296				return (error);
297			}
298
299			/*
300			 * Backing out after an error was painful in the
301			 * above.  Now we are committed to succeeding.
302			 */
303			pmp->pm_fmod = 0;
304			pmp->pm_flags |= MSDOSFSMNT_RONLY;
305			MNT_ILOCK(mp);
306			mp->mnt_flag |= MNT_RDONLY;
307			MNT_IUNLOCK(mp);
308		} else if ((pmp->pm_flags & MSDOSFSMNT_RONLY) &&
309		    !vfs_flagopt(mp->mnt_optnew, "ro", NULL, 0)) {
310			/*
311			 * If upgrade to read-write by non-root, then verify
312			 * that user has necessary permissions on the device.
313			 */
314			devvp = pmp->pm_devvp;
315			vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
316			error = VOP_ACCESS(devvp, VREAD | VWRITE,
317			    td->td_ucred, td);
318			if (error)
319				error = priv_check(td, PRIV_VFS_MOUNT_PERM);
320			if (error) {
321				VOP_UNLOCK(devvp, 0, td);
322				return (error);
323			}
324			VOP_UNLOCK(devvp, 0, td);
325			DROP_GIANT();
326			g_topology_lock();
327			error = g_access(pmp->pm_cp, 0, 1, 0);
328			g_topology_unlock();
329			PICKUP_GIANT();
330			if (error)
331				return (error);
332
333			pmp->pm_fmod = 1;
334			pmp->pm_flags &= ~MSDOSFSMNT_RONLY;
335			MNT_ILOCK(mp);
336			mp->mnt_flag &= ~MNT_RDONLY;
337			MNT_IUNLOCK(mp);
338
339			/* Now that the volume is modifiable, mark it dirty. */
340			error = markvoldirty(pmp, 1);
341			if (error)
342				return (error);
343		}
344	}
345	/*
346	 * Not an update, or updating the name: look up the name
347	 * and verify that it refers to a sensible disk device.
348	 */
349	if (vfs_getopt(mp->mnt_optnew, "from", (void **)&from, NULL))
350		return (EINVAL);
351	NDINIT(&ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, from, td);
352	error = namei(&ndp);
353	if (error)
354		return (error);
355	devvp = ndp.ni_vp;
356	NDFREE(&ndp, NDF_ONLY_PNBUF);
357
358	if (!vn_isdisk(devvp, &error)) {
359		vput(devvp);
360		return (error);
361	}
362	/*
363	 * If mount by non-root, then verify that user has necessary
364	 * permissions on the device.
365	 */
366	accessmode = VREAD;
367	if ((mp->mnt_flag & MNT_RDONLY) == 0)
368		accessmode |= VWRITE;
369	error = VOP_ACCESS(devvp, accessmode, td->td_ucred, td);
370	if (error)
371		error = priv_check(td, PRIV_VFS_MOUNT_PERM);
372	if (error) {
373		vput(devvp);
374		return (error);
375	}
376	if ((mp->mnt_flag & MNT_UPDATE) == 0) {
377		error = mountmsdosfs(devvp, mp, td);
378#ifdef MSDOSFS_DEBUG		/* only needed for the printf below */
379		pmp = VFSTOMSDOSFS(mp);
380#endif
381	} else {
382		if (devvp != pmp->pm_devvp)
383			error = EINVAL;	/* XXX needs translation */
384		else
385			vput(devvp);
386	}
387	if (error) {
388		vrele(devvp);
389		return (error);
390	}
391
392	error = update_mp(mp, td);
393	if (error) {
394		if ((mp->mnt_flag & MNT_UPDATE) == 0)
395			msdosfs_unmount(mp, MNT_FORCE, td);
396		return error;
397	}
398
399	vfs_mountedfrom(mp, from);
400#ifdef MSDOSFS_DEBUG
401	printf("msdosfs_mount(): mp %p, pmp %p, inusemap %p\n", mp, pmp, pmp->pm_inusemap);
402#endif
403	return (0);
404}
405
406static int
407mountmsdosfs(struct vnode *devvp, struct mount *mp, struct thread *td)
408{
409	struct msdosfsmount *pmp;
410	struct buf *bp;
411	struct cdev *dev = devvp->v_rdev;
412	union bootsector *bsp;
413	struct byte_bpb33 *b33;
414	struct byte_bpb50 *b50;
415	struct byte_bpb710 *b710;
416	u_int8_t SecPerClust;
417	u_long clusters;
418	int ronly, error;
419	struct g_consumer *cp;
420	struct bufobj *bo;
421
422	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
423	/* XXX: use VOP_ACCESS to check FS perms */
424	DROP_GIANT();
425	g_topology_lock();
426	error = g_vfs_open(devvp, &cp, "msdosfs", ronly ? 0 : 1);
427	g_topology_unlock();
428	PICKUP_GIANT();
429	VOP_UNLOCK(devvp, 0, td);
430	if (error)
431		return (error);
432
433	bo = &devvp->v_bufobj;
434	bp = NULL;		/* This and pmp both used in error_exit. */
435	pmp = NULL;
436
437	/*
438	 * Read the boot sector of the filesystem, and then check the
439	 * boot signature.  If not a dos boot sector then error out.
440	 *
441	 * NOTE: 8192 is a magic size that works for ffs.
442	 */
443	error = bread(devvp, 0, 8192, NOCRED, &bp);
444	if (error)
445		goto error_exit;
446	bp->b_flags |= B_AGE;
447	bsp = (union bootsector *)bp->b_data;
448	b33 = (struct byte_bpb33 *)bsp->bs33.bsBPB;
449	b50 = (struct byte_bpb50 *)bsp->bs50.bsBPB;
450	b710 = (struct byte_bpb710 *)bsp->bs710.bsBPB;
451
452#ifndef MSDOSFS_NOCHECKSIG
453	if (bsp->bs50.bsBootSectSig0 != BOOTSIG0
454	    || bsp->bs50.bsBootSectSig1 != BOOTSIG1) {
455		error = EINVAL;
456		goto error_exit;
457	}
458#endif
459
460	pmp = malloc(sizeof *pmp, M_MSDOSFSMNT, M_WAITOK | M_ZERO);
461	pmp->pm_mountp = mp;
462	pmp->pm_cp = cp;
463	pmp->pm_bo = bo;
464
465	/*
466	 * Initialize ownerships and permissions, since nothing else will
467	 * initialize them if we are mounting root.
468	 */
469	pmp->pm_uid = UID_ROOT;
470	pmp->pm_gid = GID_WHEEL;
471	pmp->pm_mask = pmp->pm_dirmask = S_IXUSR | S_IXGRP | S_IXOTH |
472	    S_IRUSR | S_IRGRP | S_IROTH | S_IWUSR;
473
474	/*
475	 * Experimental support for large MS-DOS filesystems.
476	 * WARNING: This uses at least 32 bytes of kernel memory (which is not
477	 * reclaimed until the FS is unmounted) for each file on disk to map
478	 * between the 32-bit inode numbers used by VFS and the 64-bit
479	 * pseudo-inode numbers used internally by msdosfs. This is only
480	 * safe to use in certain controlled situations (e.g. read-only FS
481	 * with less than 1 million files).
482	 * Since the mappings do not persist across unmounts (or reboots), these
483	 * filesystems are not suitable for exporting through NFS, or any other
484	 * application that requires fixed inode numbers.
485	 */
486	vfs_flagopt(mp->mnt_optnew, "large", &pmp->pm_flags, MSDOSFS_LARGEFS);
487
488	/*
489	 * Compute several useful quantities from the bpb in the
490	 * bootsector.  Copy in the dos 5 variant of the bpb then fix up
491	 * the fields that are different between dos 5 and dos 3.3.
492	 */
493	SecPerClust = b50->bpbSecPerClust;
494	pmp->pm_BytesPerSec = getushort(b50->bpbBytesPerSec);
495	if (pmp->pm_BytesPerSec < DEV_BSIZE) {
496		error = EINVAL;
497		goto error_exit;
498	}
499	pmp->pm_ResSectors = getushort(b50->bpbResSectors);
500	pmp->pm_FATs = b50->bpbFATs;
501	pmp->pm_RootDirEnts = getushort(b50->bpbRootDirEnts);
502	pmp->pm_Sectors = getushort(b50->bpbSectors);
503	pmp->pm_FATsecs = getushort(b50->bpbFATsecs);
504	pmp->pm_SecPerTrack = getushort(b50->bpbSecPerTrack);
505	pmp->pm_Heads = getushort(b50->bpbHeads);
506	pmp->pm_Media = b50->bpbMedia;
507
508	/* calculate the ratio of sector size to DEV_BSIZE */
509	pmp->pm_BlkPerSec = pmp->pm_BytesPerSec / DEV_BSIZE;
510
511	/* XXX - We should probably check more values here */
512	if (!pmp->pm_BytesPerSec || !SecPerClust
513		|| !pmp->pm_Heads
514#ifdef PC98
515    		|| !pmp->pm_SecPerTrack || pmp->pm_SecPerTrack > 255) {
516#else
517		|| !pmp->pm_SecPerTrack || pmp->pm_SecPerTrack > 63) {
518#endif
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
808		VI_LOCK(vp);
809		vn_printf(vp,
810		    "msdosfs_umount(): just before calling VOP_CLOSE()\n");
811		printf("freef %p, freeb %p, mount %p\n",
812		    TAILQ_NEXT(vp, v_freelist), vp->v_freelist.tqe_prev,
813		    vp->v_mount);
814		printf("cleanblkhd %p, dirtyblkhd %p, numoutput %ld, type %d\n",
815		    TAILQ_FIRST(&vp->v_bufobj.bo_clean.bv_hd),
816		    TAILQ_FIRST(&vp->v_bufobj.bo_dirty.bv_hd),
817		    vp->v_bufobj.bo_numoutput, vp->v_type);
818		VI_UNLOCK(vp);
819	}
820#endif
821	DROP_GIANT();
822	g_topology_lock();
823	g_vfs_close(pmp->pm_cp, td);
824	g_topology_unlock();
825	PICKUP_GIANT();
826	vrele(pmp->pm_devvp);
827	free(pmp->pm_inusemap, M_MSDOSFSFAT);
828	if (pmp->pm_flags & MSDOSFS_LARGEFS)
829		msdosfs_fileno_free(mp);
830	free(pmp, M_MSDOSFSMNT);
831	mp->mnt_data = NULL;
832	MNT_ILOCK(mp);
833	mp->mnt_flag &= ~MNT_LOCAL;
834	MNT_IUNLOCK(mp);
835	return (0);
836}
837
838static int
839msdosfs_root(struct mount *mp, int flags, struct vnode **vpp, struct thread *td)
840{
841	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
842	struct denode *ndep;
843	int error;
844
845#ifdef MSDOSFS_DEBUG
846	printf("msdosfs_root(); mp %p, pmp %p\n", mp, pmp);
847#endif
848	error = deget(pmp, MSDOSFSROOT, MSDOSFSROOT_OFS, &ndep);
849	if (error)
850		return (error);
851	*vpp = DETOV(ndep);
852	return (0);
853}
854
855static int
856msdosfs_statfs(struct mount *mp, struct statfs *sbp, struct thread *td)
857{
858	struct msdosfsmount *pmp;
859
860	pmp = VFSTOMSDOSFS(mp);
861	sbp->f_bsize = pmp->pm_bpcluster;
862	sbp->f_iosize = pmp->pm_bpcluster;
863	sbp->f_blocks = pmp->pm_maxcluster + 1;
864	sbp->f_bfree = pmp->pm_freeclustercount;
865	sbp->f_bavail = pmp->pm_freeclustercount;
866	sbp->f_files = pmp->pm_RootDirEnts;	/* XXX */
867	sbp->f_ffree = 0;	/* what to put in here? */
868	return (0);
869}
870
871static int
872msdosfs_sync(struct mount *mp, int waitfor, struct thread *td)
873{
874	struct vnode *vp, *nvp;
875	struct denode *dep;
876	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
877	int error, allerror = 0;
878
879	/*
880	 * If we ever switch to not updating all of the fats all the time,
881	 * this would be the place to update them from the first one.
882	 */
883	if (pmp->pm_fmod != 0) {
884		if (pmp->pm_flags & MSDOSFSMNT_RONLY)
885			panic("msdosfs_sync: rofs mod");
886		else {
887			/* update fats here */
888		}
889	}
890	/*
891	 * Write back each (modified) denode.
892	 */
893	MNT_ILOCK(mp);
894loop:
895	MNT_VNODE_FOREACH(vp, mp, nvp) {
896		VI_LOCK(vp);
897		if (vp->v_type == VNON || (vp->v_iflag & VI_DOOMED)) {
898			VI_UNLOCK(vp);
899			continue;
900		}
901		MNT_IUNLOCK(mp);
902		dep = VTODE(vp);
903		if ((dep->de_flag &
904		    (DE_ACCESS | DE_CREATE | DE_UPDATE | DE_MODIFIED)) == 0 &&
905		    (vp->v_bufobj.bo_dirty.bv_cnt == 0 ||
906		    waitfor == MNT_LAZY)) {
907			VI_UNLOCK(vp);
908			MNT_ILOCK(mp);
909			continue;
910		}
911		error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK, td);
912		if (error) {
913			MNT_ILOCK(mp);
914			if (error == ENOENT)
915				goto loop;
916			continue;
917		}
918		error = VOP_FSYNC(vp, waitfor, td);
919		if (error)
920			allerror = error;
921		VOP_UNLOCK(vp, 0, td);
922		vrele(vp);
923		MNT_ILOCK(mp);
924	}
925	MNT_IUNLOCK(mp);
926
927	/*
928	 * Flush filesystem control info.
929	 */
930	if (waitfor != MNT_LAZY) {
931		vn_lock(pmp->pm_devvp, LK_EXCLUSIVE | LK_RETRY, td);
932		error = VOP_FSYNC(pmp->pm_devvp, waitfor, td);
933		if (error)
934			allerror = error;
935		VOP_UNLOCK(pmp->pm_devvp, 0, td);
936	}
937	return (allerror);
938}
939
940static int
941msdosfs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
942{
943	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
944	struct defid *defhp = (struct defid *) fhp;
945	struct denode *dep;
946	int error;
947
948	error = deget(pmp, defhp->defid_dirclust, defhp->defid_dirofs, &dep);
949	if (error) {
950		*vpp = NULLVP;
951		return (error);
952	}
953	*vpp = DETOV(dep);
954	vnode_create_vobject(*vpp, dep->de_FileSize, curthread);
955	return (0);
956}
957
958static struct vfsops msdosfs_vfsops = {
959	.vfs_fhtovp =		msdosfs_fhtovp,
960	.vfs_mount =		msdosfs_mount,
961	.vfs_cmount =		msdosfs_cmount,
962	.vfs_root =		msdosfs_root,
963	.vfs_statfs =		msdosfs_statfs,
964	.vfs_sync =		msdosfs_sync,
965	.vfs_unmount =		msdosfs_unmount,
966};
967
968VFS_SET(msdosfs_vfsops, msdosfs, 0);
969MODULE_VERSION(msdosfs, 1);
970