msdosfs_vfsops.c revision 171747
1251881Speter/* $FreeBSD: head/sys/fs/msdosfs/msdosfs_vfsops.c 171747 2007-08-07 01:07:16Z bde $ */
2251881Speter/*	$NetBSD: msdosfs_vfsops.c,v 1.51 1997/11/17 15:36:58 ws Exp $	*/
3251881Speter
4251881Speter/*-
5251881Speter * Copyright (C) 1994, 1995, 1997 Wolfgang Solfrank.
6251881Speter * Copyright (C) 1994, 1995, 1997 TooLs GmbH.
7251881Speter * All rights reserved.
8251881Speter * Original code by Paul Popelka (paulp@uts.amdahl.com) (see below).
9251881Speter *
10251881Speter * Redistribution and use in source and binary forms, with or without
11251881Speter * modification, are permitted provided that the following conditions
12251881Speter * are met:
13251881Speter * 1. Redistributions of source code must retain the above copyright
14251881Speter *    notice, this list of conditions and the following disclaimer.
15251881Speter * 2. Redistributions in binary form must reproduce the above copyright
16251881Speter *    notice, this list of conditions and the following disclaimer in the
17251881Speter *    documentation and/or other materials provided with the distribution.
18251881Speter * 3. All advertising materials mentioning features or use of this software
19251881Speter *    must display the following acknowledgement:
20251881Speter *	This product includes software developed by TooLs GmbH.
21251881Speter * 4. The name of TooLs GmbH may not be used to endorse or promote products
22251881Speter *    derived from this software without specific prior written permission.
23251881Speter *
24251881Speter * THIS SOFTWARE IS PROVIDED BY TOOLS GMBH ``AS IS'' AND ANY EXPRESS OR
25251881Speter * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
26251881Speter * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
27251881Speter * IN NO EVENT SHALL TOOLS GMBH BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
28251881Speter * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
29251881Speter * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;
30251881Speter * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
31251881Speter * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
32251881Speter * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
33251881Speter * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
34251881Speter */
35251881Speter/*-
36251881Speter * Written by Paul Popelka (paulp@uts.amdahl.com)
37251881Speter *
38251881Speter * You can do anything you want with this software, just don't say you wrote
39251881Speter * it, and don't remove this notice.
40251881Speter *
41251881Speter * This software is provided "as is".
42251881Speter *
43251881Speter * The author supplies this software to be publicly redistributed on the
44251881Speter * understanding that the author is not responsible for the correct
45251881Speter * functioning of this software in any circumstances and is not liable for
46251881Speter * any damages caused by this software.
47251881Speter *
48251881Speter * October 1992
49251881Speter */
50
51#include <sys/param.h>
52#include <sys/systm.h>
53#include <sys/conf.h>
54#include <sys/namei.h>
55#include <sys/priv.h>
56#include <sys/proc.h>
57#include <sys/kernel.h>
58#include <sys/vnode.h>
59#include <sys/mount.h>
60#include <sys/buf.h>
61#include <sys/malloc.h>
62#include <sys/stat.h> 				/* defines ALLPERMS */
63#include <sys/iconv.h>
64#include <sys/mutex.h>
65
66#include <fs/msdosfs/bpb.h>
67#include <fs/msdosfs/bootsect.h>
68#include <fs/msdosfs/msdosfsmount.h>
69#include <fs/msdosfs/direntry.h>
70#include <fs/msdosfs/denode.h>
71#include <fs/msdosfs/fat.h>
72
73#include <geom/geom.h>
74#include <geom/geom_vfs.h>
75
76/* List of mount options we support */
77static const char *msdosfs_opts[] = {
78	"from",
79	"atime", "export", "force", "sync",
80	"uid", "gid", "mask", "dirmask",
81	"shortname", "shortnames", "longname", "longnames", "nowin95", "win95",
82	"kiconv", "cs_win", "cs_dos", "cs_local", "large",
83	NULL
84};
85
86#if 1 /*def PC98*/
87/*
88 * XXX - The boot signature formatted by NEC PC-98 DOS looks like a
89 *       garbage or a random value :-{
90 *       If you want to use that broken-signatured media, define the
91 *       following symbol even though PC/AT.
92 *       (ex. mount PC-98 DOS formatted FD on PC/AT)
93 */
94#define	MSDOSFS_NOCHECKSIG
95#endif
96
97MALLOC_DEFINE(M_MSDOSFSMNT, "msdosfs_mount", "MSDOSFS mount structure");
98static MALLOC_DEFINE(M_MSDOSFSFAT, "msdosfs_fat", "MSDOSFS file allocation table");
99
100struct iconv_functions *msdosfs_iconv = NULL;
101
102static int	update_mp(struct mount *mp, struct thread *td);
103static int	mountmsdosfs(struct vnode *devvp, struct mount *mp,
104		    struct thread *td);
105static vfs_fhtovp_t	msdosfs_fhtovp;
106static vfs_mount_t	msdosfs_mount;
107static vfs_root_t	msdosfs_root;
108static vfs_statfs_t	msdosfs_statfs;
109static vfs_sync_t	msdosfs_sync;
110static vfs_unmount_t	msdosfs_unmount;
111
112/* Maximum length of a character set name (arbitrary). */
113#define	MAXCSLEN	64
114
115static int
116update_mp(struct mount *mp, struct thread *td)
117{
118	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
119	void *dos, *win, *local;
120	int error, v;
121
122	if (!vfs_getopt(mp->mnt_optnew, "kiconv", NULL, NULL)) {
123		if (msdosfs_iconv != NULL) {
124			error = vfs_getopt(mp->mnt_optnew,
125			    "cs_win", &win, NULL);
126			if (!error)
127				error = vfs_getopt(mp->mnt_optnew,
128				    "cs_local", &local, NULL);
129			if (!error)
130				error = vfs_getopt(mp->mnt_optnew,
131				    "cs_dos", &dos, NULL);
132			if (!error) {
133				msdosfs_iconv->open(win, local, &pmp->pm_u2w);
134				msdosfs_iconv->open(local, win, &pmp->pm_w2u);
135				msdosfs_iconv->open(dos, local, &pmp->pm_u2d);
136				msdosfs_iconv->open(local, dos, &pmp->pm_d2u);
137			}
138			if (error != 0)
139				return (error);
140		} else {
141			pmp->pm_w2u = NULL;
142			pmp->pm_u2w = NULL;
143			pmp->pm_d2u = NULL;
144			pmp->pm_u2d = NULL;
145		}
146	}
147
148	if (1 == vfs_scanopt(mp->mnt_optnew, "gid", "%d", &v))
149		pmp->pm_gid = v;
150	if (1 == vfs_scanopt(mp->mnt_optnew, "uid", "%d", &v))
151		pmp->pm_uid = v;
152	if (1 == vfs_scanopt(mp->mnt_optnew, "mask", "%d", &v))
153		pmp->pm_mask = v & ALLPERMS;
154	if (1 == vfs_scanopt(mp->mnt_optnew, "dirmask", "%d", &v))
155		pmp->pm_dirmask = v & ALLPERMS;
156	vfs_flagopt(mp->mnt_optnew, "shortname",
157	    &pmp->pm_flags, MSDOSFSMNT_SHORTNAME);
158	vfs_flagopt(mp->mnt_optnew, "shortnames",
159	    &pmp->pm_flags, MSDOSFSMNT_SHORTNAME);
160	vfs_flagopt(mp->mnt_optnew, "longname",
161	    &pmp->pm_flags, MSDOSFSMNT_LONGNAME);
162	vfs_flagopt(mp->mnt_optnew, "longnames",
163	    &pmp->pm_flags, MSDOSFSMNT_LONGNAME);
164	vfs_flagopt(mp->mnt_optnew, "kiconv",
165	    &pmp->pm_flags, MSDOSFSMNT_KICONV);
166
167	if (vfs_getopt(mp->mnt_optnew, "nowin95", NULL, NULL) == 0)
168		pmp->pm_flags |= MSDOSFSMNT_NOWIN95;
169	else
170		pmp->pm_flags &= ~MSDOSFSMNT_NOWIN95;
171
172	if (pmp->pm_flags & MSDOSFSMNT_NOWIN95)
173		pmp->pm_flags |= MSDOSFSMNT_SHORTNAME;
174	else if (!(pmp->pm_flags &
175	    (MSDOSFSMNT_SHORTNAME | MSDOSFSMNT_LONGNAME))) {
176		struct vnode *rootvp;
177
178		/*
179		 * Try to divine whether to support Win'95 long filenames
180		 */
181		if (FAT32(pmp))
182			pmp->pm_flags |= MSDOSFSMNT_LONGNAME;
183		else {
184			if ((error =
185			    msdosfs_root(mp, LK_EXCLUSIVE, &rootvp, td)) != 0)
186				return error;
187			pmp->pm_flags |= findwin95(VTODE(rootvp))
188				? MSDOSFSMNT_LONGNAME
189					: MSDOSFSMNT_SHORTNAME;
190			vput(rootvp);
191		}
192	}
193	return 0;
194}
195
196static int
197msdosfs_cmount(struct mntarg *ma, void *data, int flags, struct thread *td)
198{
199	struct msdosfs_args args;
200	int error;
201
202	if (data == NULL)
203		return (EINVAL);
204	error = copyin(data, &args, sizeof args);
205	if (error)
206		return (error);
207
208	ma = mount_argsu(ma, "from", args.fspec, MAXPATHLEN);
209	ma = mount_arg(ma, "export", &args.export, sizeof args.export);
210	ma = mount_argf(ma, "uid", "%d", args.uid);
211	ma = mount_argf(ma, "gid", "%d", args.gid);
212	ma = mount_argf(ma, "mask", "%d", args.mask);
213	ma = mount_argf(ma, "dirmask", "%d", args.dirmask);
214
215        ma = mount_argb(ma, args.flags & MSDOSFSMNT_SHORTNAME, "noshortname");
216        ma = mount_argb(ma, args.flags & MSDOSFSMNT_LONGNAME, "nolongname");
217        ma = mount_argb(ma, !(args.flags & MSDOSFSMNT_NOWIN95), "nowin95");
218        ma = mount_argb(ma, args.flags & MSDOSFSMNT_KICONV, "nokiconv");
219
220        ma = mount_argsu(ma, "cs_win", args.cs_win, MAXCSLEN);
221        ma = mount_argsu(ma, "cs_dos", args.cs_dos, MAXCSLEN);
222        ma = mount_argsu(ma, "cs_local", args.cs_local, MAXCSLEN);
223
224	error = kernel_mount(ma, flags);
225
226	return (error);
227}
228
229/*
230 * mp - path - addr in user space of mount point (ie /usr or whatever)
231 * data - addr in user space of mount params including the name of the block
232 * special file to treat as a filesystem.
233 */
234static int
235msdosfs_mount(struct mount *mp, struct thread *td)
236{
237	struct vnode *devvp;	  /* vnode for blk device to mount */
238	/* msdosfs specific mount control block */
239	struct msdosfsmount *pmp = NULL;
240	struct nameidata ndp;
241	int error, flags;
242	mode_t accessmode;
243	char *from;
244
245	if (vfs_filteropt(mp->mnt_optnew, msdosfs_opts))
246		return (EINVAL);
247
248	/*
249	 * If updating, check whether changing from read-only to
250	 * read/write; if there is no device name, that's all we do.
251	 */
252	if (mp->mnt_flag & MNT_UPDATE) {
253		int ro_to_rw = 0;
254		pmp = VFSTOMSDOSFS(mp);
255
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			DROP_GIANT();
279			g_topology_lock();
280			error = g_access(pmp->pm_cp, 0, -1, 0);
281			g_topology_unlock();
282			PICKUP_GIANT();
283			if (error)
284				return (error);
285
286			/* Now the volume is clean. Mark it. */
287			error = markvoldirty(pmp, 0);
288			if (error && (flags & FORCECLOSE) == 0)
289				return (error);
290		} else if ((pmp->pm_flags & MSDOSFSMNT_RONLY) &&
291		    !vfs_flagopt(mp->mnt_optnew, "ro", NULL, 0)) {
292			/*
293			 * If upgrade to read-write by non-root, then verify
294			 * that user has necessary permissions on the device.
295			 */
296			devvp = pmp->pm_devvp;
297			vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td);
298			error = VOP_ACCESS(devvp, VREAD | VWRITE,
299			   td->td_ucred, td);
300			if (error)
301				error = priv_check(td, PRIV_VFS_MOUNT_PERM);
302			if (error) {
303				VOP_UNLOCK(devvp, 0, td);
304				return (error);
305			}
306			VOP_UNLOCK(devvp, 0, td);
307			DROP_GIANT();
308			g_topology_lock();
309			error = g_access(pmp->pm_cp, 0, 1, 0);
310			g_topology_unlock();
311			PICKUP_GIANT();
312			if (error)
313				return (error);
314
315			ro_to_rw = 1;
316		}
317		vfs_flagopt(mp->mnt_optnew, "ro",
318		    &pmp->pm_flags, MSDOSFSMNT_RONLY);
319		vfs_flagopt(mp->mnt_optnew, "ro",
320		    &mp->mnt_flag, MNT_RDONLY);
321
322		if (ro_to_rw) {
323			/* Now that the volume is modifiable, mark it dirty. */
324			error = markvoldirty(pmp, 1);
325			if (error)
326				return (error);
327		}
328	}
329	/*
330	 * Not an update, or updating the name: look up the name
331	 * and verify that it refers to a sensible disk device.
332	 */
333	if (vfs_getopt(mp->mnt_optnew, "from", (void **)&from, NULL))
334		return (EINVAL);
335	NDINIT(&ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, from, td);
336	error = namei(&ndp);
337	if (error)
338		return (error);
339	devvp = ndp.ni_vp;
340	NDFREE(&ndp, NDF_ONLY_PNBUF);
341
342	if (!vn_isdisk(devvp, &error)) {
343		vput(devvp);
344		return (error);
345	}
346	/*
347	 * If mount by non-root, then verify that user has necessary
348	 * permissions on the device.
349	 */
350	accessmode = VREAD;
351	if ((mp->mnt_flag & MNT_RDONLY) == 0)
352		accessmode |= VWRITE;
353	error = VOP_ACCESS(devvp, accessmode, td->td_ucred, td);
354	if (error)
355		error = priv_check(td, PRIV_VFS_MOUNT_PERM);
356	if (error) {
357		vput(devvp);
358		return (error);
359	}
360	if ((mp->mnt_flag & MNT_UPDATE) == 0) {
361		error = mountmsdosfs(devvp, mp, td);
362#ifdef MSDOSFS_DEBUG		/* only needed for the printf below */
363		pmp = VFSTOMSDOSFS(mp);
364#endif
365	} else {
366		if (devvp != pmp->pm_devvp)
367			error = EINVAL;	/* XXX needs translation */
368		else
369			vput(devvp);
370	}
371	if (error) {
372		vrele(devvp);
373		return (error);
374	}
375
376	error = update_mp(mp, td);
377	if (error) {
378		if ((mp->mnt_flag & MNT_UPDATE) == 0)
379			msdosfs_unmount(mp, MNT_FORCE, td);
380		return error;
381	}
382
383	vfs_mountedfrom(mp, from);
384#ifdef MSDOSFS_DEBUG
385	printf("msdosfs_mount(): mp %p, pmp %p, inusemap %p\n", mp, pmp, pmp->pm_inusemap);
386#endif
387	return (0);
388}
389
390static int
391mountmsdosfs(struct vnode *devvp, struct mount *mp, struct thread *td)
392{
393	struct msdosfsmount *pmp;
394	struct buf *bp;
395	struct cdev *dev = devvp->v_rdev;
396	union bootsector *bsp;
397	struct byte_bpb33 *b33;
398	struct byte_bpb50 *b50;
399	struct byte_bpb710 *b710;
400	u_int8_t SecPerClust;
401	u_long clusters;
402	int	ronly, error;
403	struct g_consumer *cp;
404	struct bufobj *bo;
405
406	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
407	/* XXX: use VOP_ACCESS to check FS perms */
408	DROP_GIANT();
409	g_topology_lock();
410	error = g_vfs_open(devvp, &cp, "msdosfs", ronly ? 0 : 1);
411	g_topology_unlock();
412	PICKUP_GIANT();
413	VOP_UNLOCK(devvp, 0, td);
414	if (error)
415		return (error);
416
417	bo = &devvp->v_bufobj;
418	bp  = NULL; /* both used in error_exit */
419	pmp = NULL;
420
421	/*
422	 * Read the boot sector of the filesystem, and then check the
423	 * boot signature.  If not a dos boot sector then error out.
424	 *
425	 * NOTE: 8192 is a magic size that works for ffs.
426	 */
427	error = bread(devvp, 0, 8192, NOCRED, &bp);
428	if (error)
429		goto error_exit;
430	bp->b_flags |= B_AGE;
431	bsp = (union bootsector *)bp->b_data;
432	b33 = (struct byte_bpb33 *)bsp->bs33.bsBPB;
433	b50 = (struct byte_bpb50 *)bsp->bs50.bsBPB;
434	b710 = (struct byte_bpb710 *)bsp->bs710.bsBPB;
435
436#ifndef MSDOSFS_NOCHECKSIG
437	if (bsp->bs50.bsBootSectSig0 != BOOTSIG0
438	    || bsp->bs50.bsBootSectSig1 != BOOTSIG1) {
439		error = EINVAL;
440		goto error_exit;
441	}
442#endif
443
444	pmp = malloc(sizeof *pmp, M_MSDOSFSMNT, M_WAITOK | M_ZERO);
445	pmp->pm_mountp = mp;
446	pmp->pm_cp = cp;
447	pmp->pm_bo = bo;
448
449	/*
450	 * Initialize ownerships and permissions, since nothing else will
451	 * initialize them iff we are mounting root.
452	 */
453	pmp->pm_uid = UID_ROOT;
454	pmp->pm_gid = GID_WHEEL;
455	pmp->pm_mask = pmp->pm_dirmask = S_IXUSR | S_IXGRP | S_IXOTH |
456	    S_IRUSR | S_IRGRP | S_IROTH | S_IWUSR;
457
458	/*
459	 * Experimental support for large MS-DOS filesystems.
460	 * WARNING: This uses at least 32 bytes of kernel memory (which is not
461	 * reclaimed until the FS is unmounted) for each file on disk to map
462	 * between the 32-bit inode numbers used by VFS and the 64-bit
463	 * pseudo-inode numbers used internally by msdosfs. This is only
464	 * safe to use in certain controlled situations (e.g. read-only FS
465	 * with less than 1 million files).
466	 * Since the mappings do not persist across unmounts (or reboots), these
467	 * filesystems are not suitable for exporting through NFS, or any other
468	 * application that requires fixed inode numbers.
469	 */
470	vfs_flagopt(mp->mnt_optnew, "large", &pmp->pm_flags,
471	  MSDOSFS_LARGEFS);
472
473	/*
474	 * Compute several useful quantities from the bpb in the
475	 * bootsector.  Copy in the dos 5 variant of the bpb then fix up
476	 * the fields that are different between dos 5 and dos 3.3.
477	 */
478	SecPerClust = b50->bpbSecPerClust;
479	pmp->pm_BytesPerSec = getushort(b50->bpbBytesPerSec);
480	if (pmp->pm_BytesPerSec < DEV_BSIZE) {
481		error = EINVAL;
482		goto error_exit;
483	}
484	pmp->pm_ResSectors = getushort(b50->bpbResSectors);
485	pmp->pm_FATs = b50->bpbFATs;
486	pmp->pm_RootDirEnts = getushort(b50->bpbRootDirEnts);
487	pmp->pm_Sectors = getushort(b50->bpbSectors);
488	pmp->pm_FATsecs = getushort(b50->bpbFATsecs);
489	pmp->pm_SecPerTrack = getushort(b50->bpbSecPerTrack);
490	pmp->pm_Heads = getushort(b50->bpbHeads);
491	pmp->pm_Media = b50->bpbMedia;
492
493	/* calculate the ratio of sector size to DEV_BSIZE */
494	pmp->pm_BlkPerSec = pmp->pm_BytesPerSec / DEV_BSIZE;
495
496	/* XXX - We should probably check more values here */
497	if (!pmp->pm_BytesPerSec || !SecPerClust
498		|| !pmp->pm_Heads
499#ifdef PC98
500    		|| !pmp->pm_SecPerTrack || pmp->pm_SecPerTrack > 255) {
501#else
502		|| !pmp->pm_SecPerTrack || pmp->pm_SecPerTrack > 63) {
503#endif
504		error = EINVAL;
505		goto error_exit;
506	}
507
508	if (pmp->pm_Sectors == 0) {
509		pmp->pm_HiddenSects = getulong(b50->bpbHiddenSecs);
510		pmp->pm_HugeSectors = getulong(b50->bpbHugeSectors);
511	} else {
512		pmp->pm_HiddenSects = getushort(b33->bpbHiddenSecs);
513		pmp->pm_HugeSectors = pmp->pm_Sectors;
514	}
515	if (!(pmp->pm_flags & MSDOSFS_LARGEFS)) {
516		if (pmp->pm_HugeSectors > 0xffffffff /
517		    (pmp->pm_BytesPerSec / sizeof(struct direntry)) + 1) {
518			/*
519			 * We cannot deal currently with this size of disk
520			 * due to fileid limitations (see msdosfs_getattr and
521			 * msdosfs_readdir)
522			 */
523			error = EINVAL;
524			vfs_mount_error(mp,
525			    "Disk too big, try '-o large' mount option");
526			goto error_exit;
527		}
528	}
529
530	if (pmp->pm_RootDirEnts == 0) {
531		if (pmp->pm_Sectors
532		    || pmp->pm_FATsecs
533		    || getushort(b710->bpbFSVers)) {
534			error = EINVAL;
535			printf("mountmsdosfs(): bad FAT32 filesystem\n");
536			goto error_exit;
537		}
538		pmp->pm_fatmask = FAT32_MASK;
539		pmp->pm_fatmult = 4;
540		pmp->pm_fatdiv = 1;
541		pmp->pm_FATsecs = getulong(b710->bpbBigFATsecs);
542		if (getushort(b710->bpbExtFlags) & FATMIRROR)
543			pmp->pm_curfat = getushort(b710->bpbExtFlags) & FATNUM;
544		else
545			pmp->pm_flags |= MSDOSFS_FATMIRROR;
546	} else
547		pmp->pm_flags |= MSDOSFS_FATMIRROR;
548
549	/*
550	 * Check a few values (could do some more):
551	 * - logical sector size: power of 2, >= block size
552	 * - sectors per cluster: power of 2, >= 1
553	 * - number of sectors:   >= 1, <= size of partition
554	 * - number of FAT sectors: >= 1
555	 */
556	if ( (SecPerClust == 0)
557	  || (SecPerClust & (SecPerClust - 1))
558	  || (pmp->pm_BytesPerSec < DEV_BSIZE)
559	  || (pmp->pm_BytesPerSec & (pmp->pm_BytesPerSec - 1))
560	  || (pmp->pm_HugeSectors == 0)
561	  || (pmp->pm_FATsecs == 0)
562	) {
563		error = EINVAL;
564		goto error_exit;
565	}
566
567	pmp->pm_HugeSectors *= pmp->pm_BlkPerSec;
568	pmp->pm_HiddenSects *= pmp->pm_BlkPerSec; /* XXX not used? */
569	pmp->pm_FATsecs     *= pmp->pm_BlkPerSec;
570	SecPerClust         *= pmp->pm_BlkPerSec;
571
572	pmp->pm_fatblk = pmp->pm_ResSectors * pmp->pm_BlkPerSec;
573
574	if (FAT32(pmp)) {
575		pmp->pm_rootdirblk = getulong(b710->bpbRootClust);
576		pmp->pm_firstcluster = pmp->pm_fatblk
577			+ (pmp->pm_FATs * pmp->pm_FATsecs);
578		pmp->pm_fsinfo = getushort(b710->bpbFSInfo) * pmp->pm_BlkPerSec;
579	} else {
580		pmp->pm_rootdirblk = pmp->pm_fatblk +
581			(pmp->pm_FATs * pmp->pm_FATsecs);
582		pmp->pm_rootdirsize = (pmp->pm_RootDirEnts * sizeof(struct direntry)
583				       + DEV_BSIZE - 1)
584			/ DEV_BSIZE; /* in blocks */
585		pmp->pm_firstcluster = pmp->pm_rootdirblk + pmp->pm_rootdirsize;
586	}
587
588	pmp->pm_maxcluster = (pmp->pm_HugeSectors - pmp->pm_firstcluster) /
589	    SecPerClust + 1;
590	pmp->pm_fatsize = pmp->pm_FATsecs * DEV_BSIZE; /* XXX not used? */
591
592	if (pmp->pm_fatmask == 0) {
593		if (pmp->pm_maxcluster
594		    <= ((CLUST_RSRVD - CLUST_FIRST) & FAT12_MASK)) {
595			/*
596			 * This will usually be a floppy disk. This size makes
597			 * sure that one fat entry will not be split across
598			 * multiple blocks.
599			 */
600			pmp->pm_fatmask = FAT12_MASK;
601			pmp->pm_fatmult = 3;
602			pmp->pm_fatdiv = 2;
603		} else {
604			pmp->pm_fatmask = FAT16_MASK;
605			pmp->pm_fatmult = 2;
606			pmp->pm_fatdiv = 1;
607		}
608	}
609
610	clusters = (pmp->pm_fatsize / pmp->pm_fatmult) * pmp->pm_fatdiv;
611	if (pmp->pm_maxcluster >= clusters) {
612		printf("Warning: number of clusters (%ld) exceeds FAT "
613		    "capacity (%ld)\n", pmp->pm_maxcluster + 1, clusters);
614		pmp->pm_maxcluster = clusters - 1;
615	}
616
617	if (FAT12(pmp))
618		pmp->pm_fatblocksize = 3 * 512;
619	else
620		pmp->pm_fatblocksize = PAGE_SIZE;
621	pmp->pm_fatblocksize = roundup(pmp->pm_fatblocksize,
622	    pmp->pm_BytesPerSec);
623	pmp->pm_fatblocksec = pmp->pm_fatblocksize / DEV_BSIZE;
624	pmp->pm_bnshift = ffs(DEV_BSIZE) - 1;
625
626	/*
627	 * Compute mask and shift value for isolating cluster relative byte
628	 * offsets and cluster numbers from a file offset.
629	 */
630	pmp->pm_bpcluster = SecPerClust * DEV_BSIZE;
631	pmp->pm_crbomask = pmp->pm_bpcluster - 1;
632	pmp->pm_cnshift = ffs(pmp->pm_bpcluster) - 1;
633
634	/*
635	 * Check for valid cluster size
636	 * must be a power of 2
637	 */
638	if (pmp->pm_bpcluster ^ (1 << pmp->pm_cnshift)) {
639		error = EINVAL;
640		goto error_exit;
641	}
642
643	/*
644	 * Release the bootsector buffer.
645	 */
646	brelse(bp);
647	bp = NULL;
648
649	/*
650	 * Check the fsinfo sector if we have one.  Silently fix up our
651	 * in-core copy of fp->fsinxtfree if it is unknown (0xffffffff)
652	 * or too large.  Ignore fp->fsinfree for now, since we need to
653	 * read the entire FAT anyway to fill the inuse map.
654	 */
655	if (pmp->pm_fsinfo) {
656		struct fsinfo *fp;
657
658		if ((error = bread(devvp, pmp->pm_fsinfo, pmp->pm_BytesPerSec,
659		    NOCRED, &bp)) != 0)
660			goto error_exit;
661		fp = (struct fsinfo *)bp->b_data;
662		if (!bcmp(fp->fsisig1, "RRaA", 4)
663		    && !bcmp(fp->fsisig2, "rrAa", 4)
664		    && !bcmp(fp->fsisig3, "\0\0\125\252", 4)) {
665			pmp->pm_nxtfree = getulong(fp->fsinxtfree);
666			if (pmp->pm_nxtfree > pmp->pm_maxcluster)
667				pmp->pm_nxtfree = CLUST_FIRST;
668		} else
669			pmp->pm_fsinfo = 0;
670		brelse(bp);
671		bp = NULL;
672	}
673
674	/*
675	 * Finish initializing pmp->pm_nxtfree (just in case the first few
676	 * sectors aren't properly reserved in the FAT).  This completes
677	 * the fixup for fp->fsinxtfree, and fixes up the zero-initialized
678	 * value if there is no fsinfo.  We will use pmp->pm_nxtfree
679	 * internally even if there is no fsinfo.
680	 */
681	if (pmp->pm_nxtfree < CLUST_FIRST)
682		pmp->pm_nxtfree = CLUST_FIRST;
683
684	/*
685	 * Allocate memory for the bitmap of allocated clusters, and then
686	 * fill it in.
687	 */
688	pmp->pm_inusemap = malloc(howmany(pmp->pm_maxcluster + 1, N_INUSEBITS)
689				  * sizeof(*pmp->pm_inusemap),
690				  M_MSDOSFSFAT, M_WAITOK);
691
692	/*
693	 * fillinusemap() needs pm_devvp.
694	 */
695	pmp->pm_devvp = devvp;
696
697	/*
698	 * Have the inuse map filled in.
699	 */
700	if ((error = fillinusemap(pmp)) != 0)
701		goto error_exit;
702
703	/*
704	 * If they want fat updates to be synchronous then let them suffer
705	 * the performance degradation in exchange for the on disk copy of
706	 * the fat being correct just about all the time.  I suppose this
707	 * would be a good thing to turn on if the kernel is still flakey.
708	 */
709	if (mp->mnt_flag & MNT_SYNCHRONOUS)
710		pmp->pm_flags |= MSDOSFSMNT_WAITONFAT;
711
712	/*
713	 * Finish up.
714	 */
715	if (ronly)
716		pmp->pm_flags |= MSDOSFSMNT_RONLY;
717	else {
718		/* Mark the volume dirty while it is mounted read/write. */
719		if ((error = markvoldirty(pmp, 1)) != 0)
720			goto error_exit;
721		pmp->pm_fmod = 1;
722	}
723	mp->mnt_data = (qaddr_t) pmp;
724	mp->mnt_stat.f_fsid.val[0] = dev2udev(dev);
725	mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
726	MNT_ILOCK(mp);
727	mp->mnt_flag |= MNT_LOCAL;
728	MNT_IUNLOCK(mp);
729
730	if (pmp->pm_flags & MSDOSFS_LARGEFS)
731		msdosfs_fileno_init(mp);
732
733	return 0;
734
735error_exit:
736	if (bp)
737		brelse(bp);
738	if (cp != NULL) {
739		DROP_GIANT();
740		g_topology_lock();
741		g_vfs_close(cp, td);
742		g_topology_unlock();
743		PICKUP_GIANT();
744	}
745	if (pmp) {
746		if (pmp->pm_inusemap)
747			free(pmp->pm_inusemap, M_MSDOSFSFAT);
748		free(pmp, M_MSDOSFSMNT);
749		mp->mnt_data = (qaddr_t)0;
750	}
751	return (error);
752}
753
754/*
755 * Unmount the filesystem described by mp.
756 */
757static int
758msdosfs_unmount(struct mount *mp, int mntflags, struct thread *td)
759{
760	struct msdosfsmount *pmp;
761	int error, flags;
762
763	flags = 0;
764	if (mntflags & MNT_FORCE)
765		flags |= FORCECLOSE;
766	error = vflush(mp, 0, flags, td);
767	if (error)
768		return error;
769	pmp = VFSTOMSDOSFS(mp);
770	if (pmp->pm_flags & MSDOSFSMNT_KICONV && msdosfs_iconv) {
771		if (pmp->pm_w2u)
772			msdosfs_iconv->close(pmp->pm_w2u);
773		if (pmp->pm_u2w)
774			msdosfs_iconv->close(pmp->pm_u2w);
775		if (pmp->pm_d2u)
776			msdosfs_iconv->close(pmp->pm_d2u);
777		if (pmp->pm_u2d)
778			msdosfs_iconv->close(pmp->pm_u2d);
779	}
780
781	/* If the volume was mounted read/write, mark it clean now. */
782	if ((pmp->pm_flags & MSDOSFSMNT_RONLY) == 0) {
783		error = markvoldirty(pmp, 0);
784		if (error && (flags & FORCECLOSE) == 0)
785			return (error);
786	}
787#ifdef MSDOSFS_DEBUG
788	{
789		struct vnode *vp = pmp->pm_devvp;
790
791		VI_LOCK(vp);
792		vn_printf(vp,
793		    "msdosfs_umount(): just before calling VOP_CLOSE()\n");
794		printf("freef %p, freeb %p, mount %p\n",
795		    TAILQ_NEXT(vp, v_freelist), vp->v_freelist.tqe_prev,
796		    vp->v_mount);
797		printf("cleanblkhd %p, dirtyblkhd %p, numoutput %ld, type %d\n",
798		    TAILQ_FIRST(&vp->v_bufobj.bo_clean.bv_hd),
799		    TAILQ_FIRST(&vp->v_bufobj.bo_dirty.bv_hd),
800		    vp->v_bufobj.bo_numoutput, vp->v_type);
801		VI_UNLOCK(vp);
802	}
803#endif
804	DROP_GIANT();
805	g_topology_lock();
806	g_vfs_close(pmp->pm_cp, td);
807	g_topology_unlock();
808	PICKUP_GIANT();
809	vrele(pmp->pm_devvp);
810	free(pmp->pm_inusemap, M_MSDOSFSFAT);
811	if (pmp->pm_flags & MSDOSFS_LARGEFS) {
812		msdosfs_fileno_free(mp);
813	}
814	free(pmp, M_MSDOSFSMNT);
815	mp->mnt_data = (qaddr_t)0;
816	MNT_ILOCK(mp);
817	mp->mnt_flag &= ~MNT_LOCAL;
818	MNT_IUNLOCK(mp);
819	return (error);
820}
821
822static int
823msdosfs_root(struct mount *mp, int flags, struct vnode **vpp, struct thread *td)
824{
825	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
826	struct denode *ndep;
827	int error;
828
829#ifdef MSDOSFS_DEBUG
830	printf("msdosfs_root(); mp %p, pmp %p\n", mp, pmp);
831#endif
832	error = deget(pmp, MSDOSFSROOT, MSDOSFSROOT_OFS, &ndep);
833	if (error)
834		return (error);
835	*vpp = DETOV(ndep);
836	return (0);
837}
838
839static int
840msdosfs_statfs(struct mount *mp, struct statfs *sbp, struct thread *td)
841{
842	struct msdosfsmount *pmp;
843
844	pmp = VFSTOMSDOSFS(mp);
845	sbp->f_bsize = pmp->pm_bpcluster;
846	sbp->f_iosize = pmp->pm_bpcluster;
847	sbp->f_blocks = pmp->pm_maxcluster + 1;
848	sbp->f_bfree = pmp->pm_freeclustercount;
849	sbp->f_bavail = pmp->pm_freeclustercount;
850	sbp->f_files = pmp->pm_RootDirEnts;			/* XXX */
851	sbp->f_ffree = 0;	/* what to put in here? */
852	return (0);
853}
854
855static int
856msdosfs_sync(struct mount *mp, int waitfor, struct thread *td)
857{
858	struct vnode *vp, *nvp;
859	struct denode *dep;
860	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
861	int error, allerror = 0;
862
863	/*
864	 * If we ever switch to not updating all of the fats all the time,
865	 * this would be the place to update them from the first one.
866	 */
867	if (pmp->pm_fmod != 0) {
868		if (pmp->pm_flags & MSDOSFSMNT_RONLY)
869			panic("msdosfs_sync: rofs mod");
870		else {
871			/* update fats here */
872		}
873	}
874	/*
875	 * Write back each (modified) denode.
876	 */
877	MNT_ILOCK(mp);
878loop:
879	MNT_VNODE_FOREACH(vp, mp, nvp) {
880		VI_LOCK(vp);
881		if (vp->v_type == VNON || (vp->v_iflag & VI_DOOMED)) {
882			VI_UNLOCK(vp);
883			continue;
884		}
885		MNT_IUNLOCK(mp);
886		dep = VTODE(vp);
887		if ((dep->de_flag &
888		    (DE_ACCESS | DE_CREATE | DE_UPDATE | DE_MODIFIED)) == 0 &&
889		    (vp->v_bufobj.bo_dirty.bv_cnt == 0 ||
890		    waitfor == MNT_LAZY)) {
891			VI_UNLOCK(vp);
892			MNT_ILOCK(mp);
893			continue;
894		}
895		error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK, td);
896		if (error) {
897			MNT_ILOCK(mp);
898			if (error == ENOENT)
899				goto loop;
900			continue;
901		}
902		error = VOP_FSYNC(vp, waitfor, td);
903		if (error)
904			allerror = error;
905		VOP_UNLOCK(vp, 0, td);
906		vrele(vp);
907		MNT_ILOCK(mp);
908	}
909	MNT_IUNLOCK(mp);
910
911	/*
912	 * Flush filesystem control info.
913	 */
914	if (waitfor != MNT_LAZY) {
915		vn_lock(pmp->pm_devvp, LK_EXCLUSIVE | LK_RETRY, td);
916		error = VOP_FSYNC(pmp->pm_devvp, waitfor, td);
917		if (error)
918			allerror = error;
919		VOP_UNLOCK(pmp->pm_devvp, 0, td);
920	}
921	return (allerror);
922}
923
924static int
925msdosfs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
926{
927	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
928	struct defid *defhp = (struct defid *) fhp;
929	struct denode *dep;
930	int error;
931
932	error = deget(pmp, defhp->defid_dirclust, defhp->defid_dirofs, &dep);
933	if (error) {
934		*vpp = NULLVP;
935		return (error);
936	}
937	*vpp = DETOV(dep);
938	vnode_create_vobject(*vpp, dep->de_FileSize, curthread);
939	return (0);
940}
941
942static struct vfsops msdosfs_vfsops = {
943	.vfs_fhtovp =		msdosfs_fhtovp,
944	.vfs_mount =		msdosfs_mount,
945	.vfs_cmount =		msdosfs_cmount,
946	.vfs_root =		msdosfs_root,
947	.vfs_statfs =		msdosfs_statfs,
948	.vfs_sync =		msdosfs_sync,
949	.vfs_unmount =		msdosfs_unmount,
950};
951
952VFS_SET(msdosfs_vfsops, msdosfs, 0);
953MODULE_VERSION(msdosfs, 1);
954