msdosfs_vfsops.c revision 171748
1/* $FreeBSD: head/sys/fs/msdosfs/msdosfs_vfsops.c 171748 2007-08-07 01:37:59Z bde $ */
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/conf.h>
54#include <sys/lock.h>
55#include <sys/mutex.h>
56#include <sys/namei.h>
57#include <sys/priv.h>
58#include <sys/proc.h>
59#include <sys/kernel.h>
60#include <sys/vnode.h>
61#include <sys/mount.h>
62#include <sys/buf.h>
63#include <sys/malloc.h>
64#include <sys/stat.h> 				/* defines ALLPERMS */
65#include <sys/iconv.h>
66
67#include <fs/msdosfs/bpb.h>
68#include <fs/msdosfs/bootsect.h>
69#include <fs/msdosfs/msdosfsmount.h>
70#include <fs/msdosfs/direntry.h>
71#include <fs/msdosfs/denode.h>
72#include <fs/msdosfs/fat.h>
73
74#include <geom/geom.h>
75#include <geom/geom_vfs.h>
76
77/* List of mount options we support */
78static const char *msdosfs_opts[] = {
79	"from",
80	"atime", "export", "force", "sync",
81	"uid", "gid", "mask", "dirmask",
82	"shortname", "shortnames", "longname", "longnames", "nowin95", "win95",
83	"kiconv", "cs_win", "cs_dos", "cs_local", "large",
84	NULL
85};
86
87#if 1 /*def PC98*/
88/*
89 * XXX - The boot signature formatted by NEC PC-98 DOS looks like a
90 *       garbage or a random value :-{
91 *       If you want to use that broken-signatured media, define the
92 *       following symbol even though PC/AT.
93 *       (ex. mount PC-98 DOS formatted FD on PC/AT)
94 */
95#define	MSDOSFS_NOCHECKSIG
96#endif
97
98MALLOC_DEFINE(M_MSDOSFSMNT, "msdosfs_mount", "MSDOSFS mount structure");
99static MALLOC_DEFINE(M_MSDOSFSFAT, "msdosfs_fat", "MSDOSFS file allocation table");
100
101struct iconv_functions *msdosfs_iconv = NULL;
102
103static int	update_mp(struct mount *mp, struct thread *td);
104static int	mountmsdosfs(struct vnode *devvp, struct mount *mp,
105		    struct thread *td);
106static vfs_fhtovp_t	msdosfs_fhtovp;
107static vfs_mount_t	msdosfs_mount;
108static vfs_root_t	msdosfs_root;
109static vfs_statfs_t	msdosfs_statfs;
110static vfs_sync_t	msdosfs_sync;
111static vfs_unmount_t	msdosfs_unmount;
112
113/* Maximum length of a character set name (arbitrary). */
114#define	MAXCSLEN	64
115
116static int
117update_mp(struct mount *mp, struct thread *td)
118{
119	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
120	void *dos, *win, *local;
121	int error, v;
122
123	if (!vfs_getopt(mp->mnt_optnew, "kiconv", NULL, NULL)) {
124		if (msdosfs_iconv != NULL) {
125			error = vfs_getopt(mp->mnt_optnew,
126			    "cs_win", &win, NULL);
127			if (!error)
128				error = vfs_getopt(mp->mnt_optnew,
129				    "cs_local", &local, NULL);
130			if (!error)
131				error = vfs_getopt(mp->mnt_optnew,
132				    "cs_dos", &dos, NULL);
133			if (!error) {
134				msdosfs_iconv->open(win, local, &pmp->pm_u2w);
135				msdosfs_iconv->open(local, win, &pmp->pm_w2u);
136				msdosfs_iconv->open(dos, local, &pmp->pm_u2d);
137				msdosfs_iconv->open(local, dos, &pmp->pm_d2u);
138			}
139			if (error != 0)
140				return (error);
141		} else {
142			pmp->pm_w2u = NULL;
143			pmp->pm_u2w = NULL;
144			pmp->pm_d2u = NULL;
145			pmp->pm_u2d = NULL;
146		}
147	}
148
149	if (1 == vfs_scanopt(mp->mnt_optnew, "gid", "%d", &v))
150		pmp->pm_gid = v;
151	if (1 == vfs_scanopt(mp->mnt_optnew, "uid", "%d", &v))
152		pmp->pm_uid = v;
153	if (1 == vfs_scanopt(mp->mnt_optnew, "mask", "%d", &v))
154		pmp->pm_mask = v & ALLPERMS;
155	if (1 == vfs_scanopt(mp->mnt_optnew, "dirmask", "%d", &v))
156		pmp->pm_dirmask = v & ALLPERMS;
157	vfs_flagopt(mp->mnt_optnew, "shortname",
158	    &pmp->pm_flags, MSDOSFSMNT_SHORTNAME);
159	vfs_flagopt(mp->mnt_optnew, "shortnames",
160	    &pmp->pm_flags, MSDOSFSMNT_SHORTNAME);
161	vfs_flagopt(mp->mnt_optnew, "longname",
162	    &pmp->pm_flags, MSDOSFSMNT_LONGNAME);
163	vfs_flagopt(mp->mnt_optnew, "longnames",
164	    &pmp->pm_flags, MSDOSFSMNT_LONGNAME);
165	vfs_flagopt(mp->mnt_optnew, "kiconv",
166	    &pmp->pm_flags, MSDOSFSMNT_KICONV);
167
168	if (vfs_getopt(mp->mnt_optnew, "nowin95", NULL, NULL) == 0)
169		pmp->pm_flags |= MSDOSFSMNT_NOWIN95;
170	else
171		pmp->pm_flags &= ~MSDOSFSMNT_NOWIN95;
172
173	if (pmp->pm_flags & MSDOSFSMNT_NOWIN95)
174		pmp->pm_flags |= MSDOSFSMNT_SHORTNAME;
175	else if (!(pmp->pm_flags &
176	    (MSDOSFSMNT_SHORTNAME | MSDOSFSMNT_LONGNAME))) {
177		struct vnode *rootvp;
178
179		/*
180		 * Try to divine whether to support Win'95 long filenames
181		 */
182		if (FAT32(pmp))
183			pmp->pm_flags |= MSDOSFSMNT_LONGNAME;
184		else {
185			if ((error =
186			    msdosfs_root(mp, LK_EXCLUSIVE, &rootvp, td)) != 0)
187				return error;
188			pmp->pm_flags |= findwin95(VTODE(rootvp))
189				? MSDOSFSMNT_LONGNAME
190					: 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		int ro_to_rw = 0;
255		pmp = VFSTOMSDOSFS(mp);
256
257		if (vfs_flagopt(mp->mnt_optnew, "export", NULL, 0)) {
258			/*
259			 * Forbid export requests if filesystem has
260			 * MSDOSFS_LARGEFS flag set.
261			 */
262			if ((pmp->pm_flags & MSDOSFS_LARGEFS) != 0) {
263				vfs_mount_error(mp,
264				    "MSDOSFS_LARGEFS flag set, cannot export");
265				return (EOPNOTSUPP);
266			}
267		}
268		if (!(pmp->pm_flags & MSDOSFSMNT_RONLY) &&
269		    vfs_flagopt(mp->mnt_optnew, "ro", NULL, 0)) {
270			error = VFS_SYNC(mp, MNT_WAIT, td);
271			if (error)
272				return (error);
273			flags = WRITECLOSE;
274			if (mp->mnt_flag & MNT_FORCE)
275				flags |= FORCECLOSE;
276			error = vflush(mp, 0, flags, td);
277			if (error)
278				return (error);
279			DROP_GIANT();
280			g_topology_lock();
281			error = g_access(pmp->pm_cp, 0, -1, 0);
282			g_topology_unlock();
283			PICKUP_GIANT();
284			if (error)
285				return (error);
286
287			/* Now the volume is clean. Mark it. */
288			error = markvoldirty(pmp, 0);
289			if (error && (flags & FORCECLOSE) == 0)
290				return (error);
291		} else if ((pmp->pm_flags & MSDOSFSMNT_RONLY) &&
292		    !vfs_flagopt(mp->mnt_optnew, "ro", NULL, 0)) {
293			/*
294			 * If upgrade to read-write by non-root, then verify
295			 * that user has necessary permissions on the device.
296			 */
297			devvp = pmp->pm_devvp;
298			vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
299			error = VOP_ACCESS(devvp, VREAD | VWRITE,
300			   td->td_ucred, td);
301			if (error)
302				error = priv_check(td, PRIV_VFS_MOUNT_PERM);
303			if (error) {
304				VOP_UNLOCK(devvp, 0, td);
305				return (error);
306			}
307			VOP_UNLOCK(devvp, 0, td);
308			DROP_GIANT();
309			g_topology_lock();
310			error = g_access(pmp->pm_cp, 0, 1, 0);
311			g_topology_unlock();
312			PICKUP_GIANT();
313			if (error)
314				return (error);
315
316			ro_to_rw = 1;
317		}
318		vfs_flagopt(mp->mnt_optnew, "ro",
319		    &pmp->pm_flags, MSDOSFSMNT_RONLY);
320		vfs_flagopt(mp->mnt_optnew, "ro",
321		    &mp->mnt_flag, MNT_RDONLY);
322
323		if (ro_to_rw) {
324			/* Now that the volume is modifiable, mark it dirty. */
325			error = markvoldirty(pmp, 1);
326			if (error)
327				return (error);
328		}
329	}
330	/*
331	 * Not an update, or updating the name: look up the name
332	 * and verify that it refers to a sensible disk device.
333	 */
334	if (vfs_getopt(mp->mnt_optnew, "from", (void **)&from, NULL))
335		return (EINVAL);
336	NDINIT(&ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, from, td);
337	error = namei(&ndp);
338	if (error)
339		return (error);
340	devvp = ndp.ni_vp;
341	NDFREE(&ndp, NDF_ONLY_PNBUF);
342
343	if (!vn_isdisk(devvp, &error)) {
344		vput(devvp);
345		return (error);
346	}
347	/*
348	 * If mount by non-root, then verify that user has necessary
349	 * permissions on the device.
350	 */
351	accessmode = VREAD;
352	if ((mp->mnt_flag & MNT_RDONLY) == 0)
353		accessmode |= VWRITE;
354	error = VOP_ACCESS(devvp, accessmode, td->td_ucred, td);
355	if (error)
356		error = priv_check(td, PRIV_VFS_MOUNT_PERM);
357	if (error) {
358		vput(devvp);
359		return (error);
360	}
361	if ((mp->mnt_flag & MNT_UPDATE) == 0) {
362		error = mountmsdosfs(devvp, mp, td);
363#ifdef MSDOSFS_DEBUG		/* only needed for the printf below */
364		pmp = VFSTOMSDOSFS(mp);
365#endif
366	} else {
367		if (devvp != pmp->pm_devvp)
368			error = EINVAL;	/* XXX needs translation */
369		else
370			vput(devvp);
371	}
372	if (error) {
373		vrele(devvp);
374		return (error);
375	}
376
377	error = update_mp(mp, td);
378	if (error) {
379		if ((mp->mnt_flag & MNT_UPDATE) == 0)
380			msdosfs_unmount(mp, MNT_FORCE, td);
381		return error;
382	}
383
384	vfs_mountedfrom(mp, from);
385#ifdef MSDOSFS_DEBUG
386	printf("msdosfs_mount(): mp %p, pmp %p, inusemap %p\n", mp, pmp, pmp->pm_inusemap);
387#endif
388	return (0);
389}
390
391static int
392mountmsdosfs(struct vnode *devvp, struct mount *mp, struct thread *td)
393{
394	struct msdosfsmount *pmp;
395	struct buf *bp;
396	struct cdev *dev = devvp->v_rdev;
397	union bootsector *bsp;
398	struct byte_bpb33 *b33;
399	struct byte_bpb50 *b50;
400	struct byte_bpb710 *b710;
401	u_int8_t SecPerClust;
402	u_long clusters;
403	int	ronly, error;
404	struct g_consumer *cp;
405	struct bufobj *bo;
406
407	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
408	/* XXX: use VOP_ACCESS to check FS perms */
409	DROP_GIANT();
410	g_topology_lock();
411	error = g_vfs_open(devvp, &cp, "msdosfs", ronly ? 0 : 1);
412	g_topology_unlock();
413	PICKUP_GIANT();
414	VOP_UNLOCK(devvp, 0, td);
415	if (error)
416		return (error);
417
418	bo = &devvp->v_bufobj;
419	bp  = NULL; /* both used in error_exit */
420	pmp = NULL;
421
422	/*
423	 * Read the boot sector of the filesystem, and then check the
424	 * boot signature.  If not a dos boot sector then error out.
425	 *
426	 * NOTE: 8192 is a magic size that works for ffs.
427	 */
428	error = bread(devvp, 0, 8192, NOCRED, &bp);
429	if (error)
430		goto error_exit;
431	bp->b_flags |= B_AGE;
432	bsp = (union bootsector *)bp->b_data;
433	b33 = (struct byte_bpb33 *)bsp->bs33.bsBPB;
434	b50 = (struct byte_bpb50 *)bsp->bs50.bsBPB;
435	b710 = (struct byte_bpb710 *)bsp->bs710.bsBPB;
436
437#ifndef MSDOSFS_NOCHECKSIG
438	if (bsp->bs50.bsBootSectSig0 != BOOTSIG0
439	    || bsp->bs50.bsBootSectSig1 != BOOTSIG1) {
440		error = EINVAL;
441		goto error_exit;
442	}
443#endif
444
445	pmp = malloc(sizeof *pmp, M_MSDOSFSMNT, M_WAITOK | M_ZERO);
446	pmp->pm_mountp = mp;
447	pmp->pm_cp = cp;
448	pmp->pm_bo = bo;
449
450	/*
451	 * Initialize ownerships and permissions, since nothing else will
452	 * initialize them iff we are mounting root.
453	 */
454	pmp->pm_uid = UID_ROOT;
455	pmp->pm_gid = GID_WHEEL;
456	pmp->pm_mask = pmp->pm_dirmask = S_IXUSR | S_IXGRP | S_IXOTH |
457	    S_IRUSR | S_IRGRP | S_IROTH | S_IWUSR;
458
459	/*
460	 * Experimental support for large MS-DOS filesystems.
461	 * WARNING: This uses at least 32 bytes of kernel memory (which is not
462	 * reclaimed until the FS is unmounted) for each file on disk to map
463	 * between the 32-bit inode numbers used by VFS and the 64-bit
464	 * pseudo-inode numbers used internally by msdosfs. This is only
465	 * safe to use in certain controlled situations (e.g. read-only FS
466	 * with less than 1 million files).
467	 * Since the mappings do not persist across unmounts (or reboots), these
468	 * filesystems are not suitable for exporting through NFS, or any other
469	 * application that requires fixed inode numbers.
470	 */
471	vfs_flagopt(mp->mnt_optnew, "large", &pmp->pm_flags,
472	  MSDOSFS_LARGEFS);
473
474	/*
475	 * Compute several useful quantities from the bpb in the
476	 * bootsector.  Copy in the dos 5 variant of the bpb then fix up
477	 * the fields that are different between dos 5 and dos 3.3.
478	 */
479	SecPerClust = b50->bpbSecPerClust;
480	pmp->pm_BytesPerSec = getushort(b50->bpbBytesPerSec);
481	if (pmp->pm_BytesPerSec < DEV_BSIZE) {
482		error = EINVAL;
483		goto error_exit;
484	}
485	pmp->pm_ResSectors = getushort(b50->bpbResSectors);
486	pmp->pm_FATs = b50->bpbFATs;
487	pmp->pm_RootDirEnts = getushort(b50->bpbRootDirEnts);
488	pmp->pm_Sectors = getushort(b50->bpbSectors);
489	pmp->pm_FATsecs = getushort(b50->bpbFATsecs);
490	pmp->pm_SecPerTrack = getushort(b50->bpbSecPerTrack);
491	pmp->pm_Heads = getushort(b50->bpbHeads);
492	pmp->pm_Media = b50->bpbMedia;
493
494	/* calculate the ratio of sector size to DEV_BSIZE */
495	pmp->pm_BlkPerSec = pmp->pm_BytesPerSec / DEV_BSIZE;
496
497	/* XXX - We should probably check more values here */
498	if (!pmp->pm_BytesPerSec || !SecPerClust
499		|| !pmp->pm_Heads
500#ifdef PC98
501    		|| !pmp->pm_SecPerTrack || pmp->pm_SecPerTrack > 255) {
502#else
503		|| !pmp->pm_SecPerTrack || pmp->pm_SecPerTrack > 63) {
504#endif
505		error = EINVAL;
506		goto error_exit;
507	}
508
509	if (pmp->pm_Sectors == 0) {
510		pmp->pm_HiddenSects = getulong(b50->bpbHiddenSecs);
511		pmp->pm_HugeSectors = getulong(b50->bpbHugeSectors);
512	} else {
513		pmp->pm_HiddenSects = getushort(b33->bpbHiddenSecs);
514		pmp->pm_HugeSectors = pmp->pm_Sectors;
515	}
516	if (!(pmp->pm_flags & MSDOSFS_LARGEFS)) {
517		if (pmp->pm_HugeSectors > 0xffffffff /
518		    (pmp->pm_BytesPerSec / sizeof(struct direntry)) + 1) {
519			/*
520			 * We cannot deal currently with this size of disk
521			 * due to fileid limitations (see msdosfs_getattr and
522			 * msdosfs_readdir)
523			 */
524			error = EINVAL;
525			vfs_mount_error(mp,
526			    "Disk too big, try '-o large' mount option");
527			goto error_exit;
528		}
529	}
530
531	if (pmp->pm_RootDirEnts == 0) {
532		if (pmp->pm_Sectors
533		    || pmp->pm_FATsecs
534		    || getushort(b710->bpbFSVers)) {
535			error = EINVAL;
536			printf("mountmsdosfs(): bad FAT32 filesystem\n");
537			goto error_exit;
538		}
539		pmp->pm_fatmask = FAT32_MASK;
540		pmp->pm_fatmult = 4;
541		pmp->pm_fatdiv = 1;
542		pmp->pm_FATsecs = getulong(b710->bpbBigFATsecs);
543		if (getushort(b710->bpbExtFlags) & FATMIRROR)
544			pmp->pm_curfat = getushort(b710->bpbExtFlags) & FATNUM;
545		else
546			pmp->pm_flags |= MSDOSFS_FATMIRROR;
547	} else
548		pmp->pm_flags |= MSDOSFS_FATMIRROR;
549
550	/*
551	 * Check a few values (could do some more):
552	 * - logical sector size: power of 2, >= block size
553	 * - sectors per cluster: power of 2, >= 1
554	 * - number of sectors:   >= 1, <= size of partition
555	 * - number of FAT sectors: >= 1
556	 */
557	if ( (SecPerClust == 0)
558	  || (SecPerClust & (SecPerClust - 1))
559	  || (pmp->pm_BytesPerSec < DEV_BSIZE)
560	  || (pmp->pm_BytesPerSec & (pmp->pm_BytesPerSec - 1))
561	  || (pmp->pm_HugeSectors == 0)
562	  || (pmp->pm_FATsecs == 0)
563	) {
564		error = EINVAL;
565		goto error_exit;
566	}
567
568	pmp->pm_HugeSectors *= pmp->pm_BlkPerSec;
569	pmp->pm_HiddenSects *= pmp->pm_BlkPerSec; /* XXX not used? */
570	pmp->pm_FATsecs     *= pmp->pm_BlkPerSec;
571	SecPerClust         *= pmp->pm_BlkPerSec;
572
573	pmp->pm_fatblk = pmp->pm_ResSectors * pmp->pm_BlkPerSec;
574
575	if (FAT32(pmp)) {
576		pmp->pm_rootdirblk = getulong(b710->bpbRootClust);
577		pmp->pm_firstcluster = pmp->pm_fatblk
578			+ (pmp->pm_FATs * pmp->pm_FATsecs);
579		pmp->pm_fsinfo = getushort(b710->bpbFSInfo) * pmp->pm_BlkPerSec;
580	} else {
581		pmp->pm_rootdirblk = pmp->pm_fatblk +
582			(pmp->pm_FATs * pmp->pm_FATsecs);
583		pmp->pm_rootdirsize = (pmp->pm_RootDirEnts * sizeof(struct direntry)
584				       + DEV_BSIZE - 1)
585			/ DEV_BSIZE; /* in blocks */
586		pmp->pm_firstcluster = pmp->pm_rootdirblk + pmp->pm_rootdirsize;
587	}
588
589	pmp->pm_maxcluster = (pmp->pm_HugeSectors - pmp->pm_firstcluster) /
590	    SecPerClust + 1;
591	pmp->pm_fatsize = pmp->pm_FATsecs * DEV_BSIZE; /* XXX not used? */
592
593	if (pmp->pm_fatmask == 0) {
594		if (pmp->pm_maxcluster
595		    <= ((CLUST_RSRVD - CLUST_FIRST) & FAT12_MASK)) {
596			/*
597			 * This will usually be a floppy disk. This size makes
598			 * sure that one fat entry will not be split across
599			 * multiple blocks.
600			 */
601			pmp->pm_fatmask = FAT12_MASK;
602			pmp->pm_fatmult = 3;
603			pmp->pm_fatdiv = 2;
604		} else {
605			pmp->pm_fatmask = FAT16_MASK;
606			pmp->pm_fatmult = 2;
607			pmp->pm_fatdiv = 1;
608		}
609	}
610
611	clusters = (pmp->pm_fatsize / pmp->pm_fatmult) * pmp->pm_fatdiv;
612	if (pmp->pm_maxcluster >= clusters) {
613		printf("Warning: number of clusters (%ld) exceeds FAT "
614		    "capacity (%ld)\n", pmp->pm_maxcluster + 1, clusters);
615		pmp->pm_maxcluster = clusters - 1;
616	}
617
618	if (FAT12(pmp))
619		pmp->pm_fatblocksize = 3 * 512;
620	else
621		pmp->pm_fatblocksize = PAGE_SIZE;
622	pmp->pm_fatblocksize = roundup(pmp->pm_fatblocksize,
623	    pmp->pm_BytesPerSec);
624	pmp->pm_fatblocksec = pmp->pm_fatblocksize / DEV_BSIZE;
625	pmp->pm_bnshift = ffs(DEV_BSIZE) - 1;
626
627	/*
628	 * Compute mask and shift value for isolating cluster relative byte
629	 * offsets and cluster numbers from a file offset.
630	 */
631	pmp->pm_bpcluster = SecPerClust * DEV_BSIZE;
632	pmp->pm_crbomask = pmp->pm_bpcluster - 1;
633	pmp->pm_cnshift = ffs(pmp->pm_bpcluster) - 1;
634
635	/*
636	 * Check for valid cluster size
637	 * must be a power of 2
638	 */
639	if (pmp->pm_bpcluster ^ (1 << pmp->pm_cnshift)) {
640		error = EINVAL;
641		goto error_exit;
642	}
643
644	/*
645	 * Release the bootsector buffer.
646	 */
647	brelse(bp);
648	bp = NULL;
649
650	/*
651	 * Check the fsinfo sector if we have one.  Silently fix up our
652	 * in-core copy of fp->fsinxtfree if it is unknown (0xffffffff)
653	 * or too large.  Ignore fp->fsinfree for now, since we need to
654	 * read the entire FAT anyway to fill the inuse map.
655	 */
656	if (pmp->pm_fsinfo) {
657		struct fsinfo *fp;
658
659		if ((error = bread(devvp, pmp->pm_fsinfo, pmp->pm_BytesPerSec,
660		    NOCRED, &bp)) != 0)
661			goto error_exit;
662		fp = (struct fsinfo *)bp->b_data;
663		if (!bcmp(fp->fsisig1, "RRaA", 4)
664		    && !bcmp(fp->fsisig2, "rrAa", 4)
665		    && !bcmp(fp->fsisig3, "\0\0\125\252", 4)) {
666			pmp->pm_nxtfree = getulong(fp->fsinxtfree);
667			if (pmp->pm_nxtfree > pmp->pm_maxcluster)
668				pmp->pm_nxtfree = CLUST_FIRST;
669		} else
670			pmp->pm_fsinfo = 0;
671		brelse(bp);
672		bp = NULL;
673	}
674
675	/*
676	 * Finish initializing pmp->pm_nxtfree (just in case the first few
677	 * sectors aren't properly reserved in the FAT).  This completes
678	 * the fixup for fp->fsinxtfree, and fixes up the zero-initialized
679	 * value if there is no fsinfo.  We will use pmp->pm_nxtfree
680	 * internally even if there is no fsinfo.
681	 */
682	if (pmp->pm_nxtfree < CLUST_FIRST)
683		pmp->pm_nxtfree = CLUST_FIRST;
684
685	/*
686	 * Allocate memory for the bitmap of allocated clusters, and then
687	 * fill it in.
688	 */
689	pmp->pm_inusemap = malloc(howmany(pmp->pm_maxcluster + 1, N_INUSEBITS)
690				  * sizeof(*pmp->pm_inusemap),
691				  M_MSDOSFSFAT, M_WAITOK);
692
693	/*
694	 * fillinusemap() needs pm_devvp.
695	 */
696	pmp->pm_devvp = devvp;
697
698	/*
699	 * Have the inuse map filled in.
700	 */
701	if ((error = fillinusemap(pmp)) != 0)
702		goto error_exit;
703
704	/*
705	 * If they want fat updates to be synchronous then let them suffer
706	 * the performance degradation in exchange for the on disk copy of
707	 * the fat being correct just about all the time.  I suppose this
708	 * would be a good thing to turn on if the kernel is still flakey.
709	 */
710	if (mp->mnt_flag & MNT_SYNCHRONOUS)
711		pmp->pm_flags |= MSDOSFSMNT_WAITONFAT;
712
713	/*
714	 * Finish up.
715	 */
716	if (ronly)
717		pmp->pm_flags |= MSDOSFSMNT_RONLY;
718	else {
719		/* Mark the volume dirty while it is mounted read/write. */
720		if ((error = markvoldirty(pmp, 1)) != 0)
721			goto error_exit;
722		pmp->pm_fmod = 1;
723	}
724	mp->mnt_data = (qaddr_t) pmp;
725	mp->mnt_stat.f_fsid.val[0] = dev2udev(dev);
726	mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
727	MNT_ILOCK(mp);
728	mp->mnt_flag |= MNT_LOCAL;
729	MNT_IUNLOCK(mp);
730
731	if (pmp->pm_flags & MSDOSFS_LARGEFS)
732		msdosfs_fileno_init(mp);
733
734	return 0;
735
736error_exit:
737	if (bp)
738		brelse(bp);
739	if (cp != NULL) {
740		DROP_GIANT();
741		g_topology_lock();
742		g_vfs_close(cp, td);
743		g_topology_unlock();
744		PICKUP_GIANT();
745	}
746	if (pmp) {
747		if (pmp->pm_inusemap)
748			free(pmp->pm_inusemap, M_MSDOSFSFAT);
749		free(pmp, M_MSDOSFSMNT);
750		mp->mnt_data = (qaddr_t)0;
751	}
752	return (error);
753}
754
755/*
756 * Unmount the filesystem described by mp.
757 */
758static int
759msdosfs_unmount(struct mount *mp, int mntflags, struct thread *td)
760{
761	struct msdosfsmount *pmp;
762	int error, flags;
763
764	flags = 0;
765	if (mntflags & MNT_FORCE)
766		flags |= FORCECLOSE;
767	error = vflush(mp, 0, flags, td);
768	if (error)
769		return error;
770	pmp = VFSTOMSDOSFS(mp);
771	if (pmp->pm_flags & MSDOSFSMNT_KICONV && msdosfs_iconv) {
772		if (pmp->pm_w2u)
773			msdosfs_iconv->close(pmp->pm_w2u);
774		if (pmp->pm_u2w)
775			msdosfs_iconv->close(pmp->pm_u2w);
776		if (pmp->pm_d2u)
777			msdosfs_iconv->close(pmp->pm_d2u);
778		if (pmp->pm_u2d)
779			msdosfs_iconv->close(pmp->pm_u2d);
780	}
781
782	/* If the volume was mounted read/write, mark it clean now. */
783	if ((pmp->pm_flags & MSDOSFSMNT_RONLY) == 0) {
784		error = markvoldirty(pmp, 0);
785		if (error && (flags & FORCECLOSE) == 0)
786			return (error);
787	}
788#ifdef MSDOSFS_DEBUG
789	{
790		struct vnode *vp = pmp->pm_devvp;
791
792		VI_LOCK(vp);
793		vn_printf(vp,
794		    "msdosfs_umount(): just before calling VOP_CLOSE()\n");
795		printf("freef %p, freeb %p, mount %p\n",
796		    TAILQ_NEXT(vp, v_freelist), vp->v_freelist.tqe_prev,
797		    vp->v_mount);
798		printf("cleanblkhd %p, dirtyblkhd %p, numoutput %ld, type %d\n",
799		    TAILQ_FIRST(&vp->v_bufobj.bo_clean.bv_hd),
800		    TAILQ_FIRST(&vp->v_bufobj.bo_dirty.bv_hd),
801		    vp->v_bufobj.bo_numoutput, vp->v_type);
802		VI_UNLOCK(vp);
803	}
804#endif
805	DROP_GIANT();
806	g_topology_lock();
807	g_vfs_close(pmp->pm_cp, td);
808	g_topology_unlock();
809	PICKUP_GIANT();
810	vrele(pmp->pm_devvp);
811	free(pmp->pm_inusemap, M_MSDOSFSFAT);
812	if (pmp->pm_flags & MSDOSFS_LARGEFS) {
813		msdosfs_fileno_free(mp);
814	}
815	free(pmp, M_MSDOSFSMNT);
816	mp->mnt_data = (qaddr_t)0;
817	MNT_ILOCK(mp);
818	mp->mnt_flag &= ~MNT_LOCAL;
819	MNT_IUNLOCK(mp);
820	return (error);
821}
822
823static int
824msdosfs_root(struct mount *mp, int flags, struct vnode **vpp, struct thread *td)
825{
826	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
827	struct denode *ndep;
828	int error;
829
830#ifdef MSDOSFS_DEBUG
831	printf("msdosfs_root(); mp %p, pmp %p\n", mp, pmp);
832#endif
833	error = deget(pmp, MSDOSFSROOT, MSDOSFSROOT_OFS, &ndep);
834	if (error)
835		return (error);
836	*vpp = DETOV(ndep);
837	return (0);
838}
839
840static int
841msdosfs_statfs(struct mount *mp, struct statfs *sbp, struct thread *td)
842{
843	struct msdosfsmount *pmp;
844
845	pmp = VFSTOMSDOSFS(mp);
846	sbp->f_bsize = pmp->pm_bpcluster;
847	sbp->f_iosize = pmp->pm_bpcluster;
848	sbp->f_blocks = pmp->pm_maxcluster + 1;
849	sbp->f_bfree = pmp->pm_freeclustercount;
850	sbp->f_bavail = pmp->pm_freeclustercount;
851	sbp->f_files = pmp->pm_RootDirEnts;			/* XXX */
852	sbp->f_ffree = 0;	/* what to put in here? */
853	return (0);
854}
855
856static int
857msdosfs_sync(struct mount *mp, int waitfor, struct thread *td)
858{
859	struct vnode *vp, *nvp;
860	struct denode *dep;
861	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
862	int error, allerror = 0;
863
864	/*
865	 * If we ever switch to not updating all of the fats all the time,
866	 * this would be the place to update them from the first one.
867	 */
868	if (pmp->pm_fmod != 0) {
869		if (pmp->pm_flags & MSDOSFSMNT_RONLY)
870			panic("msdosfs_sync: rofs mod");
871		else {
872			/* update fats here */
873		}
874	}
875	/*
876	 * Write back each (modified) denode.
877	 */
878	MNT_ILOCK(mp);
879loop:
880	MNT_VNODE_FOREACH(vp, mp, nvp) {
881		VI_LOCK(vp);
882		if (vp->v_type == VNON || (vp->v_iflag & VI_DOOMED)) {
883			VI_UNLOCK(vp);
884			continue;
885		}
886		MNT_IUNLOCK(mp);
887		dep = VTODE(vp);
888		if ((dep->de_flag &
889		    (DE_ACCESS | DE_CREATE | DE_UPDATE | DE_MODIFIED)) == 0 &&
890		    (vp->v_bufobj.bo_dirty.bv_cnt == 0 ||
891		    waitfor == MNT_LAZY)) {
892			VI_UNLOCK(vp);
893			MNT_ILOCK(mp);
894			continue;
895		}
896		error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK, td);
897		if (error) {
898			MNT_ILOCK(mp);
899			if (error == ENOENT)
900				goto loop;
901			continue;
902		}
903		error = VOP_FSYNC(vp, waitfor, td);
904		if (error)
905			allerror = error;
906		VOP_UNLOCK(vp, 0, td);
907		vrele(vp);
908		MNT_ILOCK(mp);
909	}
910	MNT_IUNLOCK(mp);
911
912	/*
913	 * Flush filesystem control info.
914	 */
915	if (waitfor != MNT_LAZY) {
916		vn_lock(pmp->pm_devvp, LK_EXCLUSIVE | LK_RETRY, td);
917		error = VOP_FSYNC(pmp->pm_devvp, waitfor, td);
918		if (error)
919			allerror = error;
920		VOP_UNLOCK(pmp->pm_devvp, 0, td);
921	}
922	return (allerror);
923}
924
925static int
926msdosfs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
927{
928	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
929	struct defid *defhp = (struct defid *) fhp;
930	struct denode *dep;
931	int error;
932
933	error = deget(pmp, defhp->defid_dirclust, defhp->defid_dirofs, &dep);
934	if (error) {
935		*vpp = NULLVP;
936		return (error);
937	}
938	*vpp = DETOV(dep);
939	vnode_create_vobject(*vpp, dep->de_FileSize, curthread);
940	return (0);
941}
942
943static struct vfsops msdosfs_vfsops = {
944	.vfs_fhtovp =		msdosfs_fhtovp,
945	.vfs_mount =		msdosfs_mount,
946	.vfs_cmount =		msdosfs_cmount,
947	.vfs_root =		msdosfs_root,
948	.vfs_statfs =		msdosfs_statfs,
949	.vfs_sync =		msdosfs_sync,
950	.vfs_unmount =		msdosfs_unmount,
951};
952
953VFS_SET(msdosfs_vfsops, msdosfs, 0);
954MODULE_VERSION(msdosfs, 1);
955