msdosfs_vfsops.c revision 204589
1/* $FreeBSD: head/sys/fs/msdosfs/msdosfs_vfsops.c 204589 2010-03-02 17:24:33Z 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
78static const char msdosfs_lock_msg[] = "fatlk";
79
80/* Mount options that we support. */
81static const char *msdosfs_opts[] = {
82	"async", "noatime", "noclusterr", "noclusterw",
83	"export", "force", "from", "sync",
84	"cs_dos", "cs_local", "cs_win", "dirmask",
85	"gid", "kiconv", "large", "longname",
86	"longnames", "mask", "shortname", "shortnames",
87	"uid", "win95", "nowin95",
88	NULL
89};
90
91#if 1 /*def PC98*/
92/*
93 * XXX - The boot signature formatted by NEC PC-98 DOS looks like a
94 *       garbage or a random value :-{
95 *       If you want to use that broken-signatured media, define the
96 *       following symbol even though PC/AT.
97 *       (ex. mount PC-98 DOS formatted FD on PC/AT)
98 */
99#define	MSDOSFS_NOCHECKSIG
100#endif
101
102MALLOC_DEFINE(M_MSDOSFSMNT, "msdosfs_mount", "MSDOSFS mount structure");
103static MALLOC_DEFINE(M_MSDOSFSFAT, "msdosfs_fat", "MSDOSFS file allocation table");
104
105struct iconv_functions *msdosfs_iconv;
106
107static int	update_mp(struct mount *mp, struct thread *td);
108static int	mountmsdosfs(struct vnode *devvp, struct mount *mp);
109static vfs_fhtovp_t	msdosfs_fhtovp;
110static vfs_mount_t	msdosfs_mount;
111static vfs_root_t	msdosfs_root;
112static vfs_statfs_t	msdosfs_statfs;
113static vfs_sync_t	msdosfs_sync;
114static vfs_unmount_t	msdosfs_unmount;
115
116/* Maximum length of a character set name (arbitrary). */
117#define	MAXCSLEN	64
118
119static int
120update_mp(struct mount *mp, struct thread *td)
121{
122	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
123	void *dos, *win, *local;
124	int error, v;
125
126	if (!vfs_getopt(mp->mnt_optnew, "kiconv", NULL, NULL)) {
127		if (msdosfs_iconv != NULL) {
128			error = vfs_getopt(mp->mnt_optnew,
129			    "cs_win", &win, NULL);
130			if (!error)
131				error = vfs_getopt(mp->mnt_optnew,
132				    "cs_local", &local, NULL);
133			if (!error)
134				error = vfs_getopt(mp->mnt_optnew,
135				    "cs_dos", &dos, NULL);
136			if (!error) {
137				msdosfs_iconv->open(win, local, &pmp->pm_u2w);
138				msdosfs_iconv->open(local, win, &pmp->pm_w2u);
139				msdosfs_iconv->open(dos, local, &pmp->pm_u2d);
140				msdosfs_iconv->open(local, dos, &pmp->pm_d2u);
141			}
142			if (error != 0)
143				return (error);
144		} else {
145			pmp->pm_w2u = NULL;
146			pmp->pm_u2w = NULL;
147			pmp->pm_d2u = NULL;
148			pmp->pm_u2d = NULL;
149		}
150	}
151
152	if (1 == vfs_scanopt(mp->mnt_optnew, "gid", "%d", &v))
153		pmp->pm_gid = v;
154	if (1 == vfs_scanopt(mp->mnt_optnew, "uid", "%d", &v))
155		pmp->pm_uid = v;
156	if (1 == vfs_scanopt(mp->mnt_optnew, "mask", "%d", &v))
157		pmp->pm_mask = v & ALLPERMS;
158	if (1 == vfs_scanopt(mp->mnt_optnew, "dirmask", "%d", &v))
159		pmp->pm_dirmask = v & ALLPERMS;
160	vfs_flagopt(mp->mnt_optnew, "shortname",
161	    &pmp->pm_flags, MSDOSFSMNT_SHORTNAME);
162	vfs_flagopt(mp->mnt_optnew, "shortnames",
163	    &pmp->pm_flags, MSDOSFSMNT_SHORTNAME);
164	vfs_flagopt(mp->mnt_optnew, "longname",
165	    &pmp->pm_flags, MSDOSFSMNT_LONGNAME);
166	vfs_flagopt(mp->mnt_optnew, "longnames",
167	    &pmp->pm_flags, MSDOSFSMNT_LONGNAME);
168	vfs_flagopt(mp->mnt_optnew, "kiconv",
169	    &pmp->pm_flags, MSDOSFSMNT_KICONV);
170
171	if (vfs_getopt(mp->mnt_optnew, "nowin95", NULL, NULL) == 0)
172		pmp->pm_flags |= MSDOSFSMNT_NOWIN95;
173	else
174		pmp->pm_flags &= ~MSDOSFSMNT_NOWIN95;
175
176	if (pmp->pm_flags & MSDOSFSMNT_NOWIN95)
177		pmp->pm_flags |= MSDOSFSMNT_SHORTNAME;
178	else if (!(pmp->pm_flags &
179	    (MSDOSFSMNT_SHORTNAME | MSDOSFSMNT_LONGNAME))) {
180		struct vnode *rootvp;
181
182		/*
183		 * Try to divine whether to support Win'95 long filenames
184		 */
185		if (FAT32(pmp))
186			pmp->pm_flags |= MSDOSFSMNT_LONGNAME;
187		else {
188			if ((error =
189			    msdosfs_root(mp, LK_EXCLUSIVE, &rootvp)) != 0)
190				return error;
191			pmp->pm_flags |= findwin95(VTODE(rootvp)) ?
192			    MSDOSFSMNT_LONGNAME : MSDOSFSMNT_SHORTNAME;
193			vput(rootvp);
194		}
195	}
196	return 0;
197}
198
199static int
200msdosfs_cmount(struct mntarg *ma, void *data, int flags)
201{
202	struct msdosfs_args args;
203	int error;
204
205	if (data == NULL)
206		return (EINVAL);
207	error = copyin(data, &args, sizeof args);
208	if (error)
209		return (error);
210
211	ma = mount_argsu(ma, "from", args.fspec, MAXPATHLEN);
212	ma = mount_arg(ma, "export", &args.export, sizeof args.export);
213	ma = mount_argf(ma, "uid", "%d", args.uid);
214	ma = mount_argf(ma, "gid", "%d", args.gid);
215	ma = mount_argf(ma, "mask", "%d", args.mask);
216	ma = mount_argf(ma, "dirmask", "%d", args.dirmask);
217
218	ma = mount_argb(ma, args.flags & MSDOSFSMNT_SHORTNAME, "noshortname");
219	ma = mount_argb(ma, args.flags & MSDOSFSMNT_LONGNAME, "nolongname");
220	ma = mount_argb(ma, !(args.flags & MSDOSFSMNT_NOWIN95), "nowin95");
221	ma = mount_argb(ma, args.flags & MSDOSFSMNT_KICONV, "nokiconv");
222
223	ma = mount_argsu(ma, "cs_win", args.cs_win, MAXCSLEN);
224	ma = mount_argsu(ma, "cs_dos", args.cs_dos, MAXCSLEN);
225	ma = mount_argsu(ma, "cs_local", args.cs_local, MAXCSLEN);
226
227	error = kernel_mount(ma, flags);
228
229	return (error);
230}
231
232/*
233 * mp - path - addr in user space of mount point (ie /usr or whatever)
234 * data - addr in user space of mount params including the name of the block
235 * special file to treat as a filesystem.
236 */
237static int
238msdosfs_mount(struct mount *mp)
239{
240	struct vnode *devvp;	  /* vnode for blk device to mount */
241	struct thread *td;
242	/* msdosfs specific mount control block */
243	struct msdosfsmount *pmp = NULL;
244	struct nameidata ndp;
245	int error, flags;
246	accmode_t accmode;
247	char *from;
248
249	td = curthread;
250	if (vfs_filteropt(mp->mnt_optnew, msdosfs_opts))
251		return (EINVAL);
252
253	/*
254	 * If updating, check whether changing from read-only to
255	 * read/write; if there is no device name, that's all we do.
256	 */
257	if (mp->mnt_flag & MNT_UPDATE) {
258		pmp = VFSTOMSDOSFS(mp);
259		if (vfs_flagopt(mp->mnt_optnew, "export", NULL, 0)) {
260			/*
261			 * Forbid export requests if filesystem has
262			 * MSDOSFS_LARGEFS flag set.
263			 */
264			if ((pmp->pm_flags & MSDOSFS_LARGEFS) != 0) {
265				vfs_mount_error(mp,
266				    "MSDOSFS_LARGEFS flag set, cannot export");
267				return (EOPNOTSUPP);
268			}
269		}
270		if (!(pmp->pm_flags & MSDOSFSMNT_RONLY) &&
271		    vfs_flagopt(mp->mnt_optnew, "ro", NULL, 0)) {
272			error = VFS_SYNC(mp, MNT_WAIT);
273			if (error)
274				return (error);
275			flags = WRITECLOSE;
276			if (mp->mnt_flag & MNT_FORCE)
277				flags |= FORCECLOSE;
278			error = vflush(mp, 0, flags, td);
279			if (error)
280				return (error);
281
282			/*
283			 * Now the volume is clean.  Mark it so while the
284			 * device is still rw.
285			 */
286			error = markvoldirty(pmp, 0);
287			if (error) {
288				(void)markvoldirty(pmp, 1);
289				return (error);
290			}
291
292			/* Downgrade the device from rw to ro. */
293			DROP_GIANT();
294			g_topology_lock();
295			error = g_access(pmp->pm_cp, 0, -1, 0);
296			g_topology_unlock();
297			PICKUP_GIANT();
298			if (error) {
299				(void)markvoldirty(pmp, 1);
300				return (error);
301			}
302
303			/*
304			 * Backing out after an error was painful in the
305			 * above.  Now we are committed to succeeding.
306			 */
307			pmp->pm_fmod = 0;
308			pmp->pm_flags |= MSDOSFSMNT_RONLY;
309			MNT_ILOCK(mp);
310			mp->mnt_flag |= MNT_RDONLY;
311			MNT_IUNLOCK(mp);
312		} else if ((pmp->pm_flags & MSDOSFSMNT_RONLY) &&
313		    !vfs_flagopt(mp->mnt_optnew, "ro", NULL, 0)) {
314			/*
315			 * If upgrade to read-write by non-root, then verify
316			 * that user has necessary permissions on the device.
317			 */
318			devvp = pmp->pm_devvp;
319			vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
320			error = VOP_ACCESS(devvp, VREAD | VWRITE,
321			    td->td_ucred, td);
322			if (error)
323				error = priv_check(td, PRIV_VFS_MOUNT_PERM);
324			if (error) {
325				VOP_UNLOCK(devvp, 0);
326				return (error);
327			}
328			VOP_UNLOCK(devvp, 0);
329			DROP_GIANT();
330			g_topology_lock();
331			error = g_access(pmp->pm_cp, 0, 1, 0);
332			g_topology_unlock();
333			PICKUP_GIANT();
334			if (error)
335				return (error);
336
337			pmp->pm_fmod = 1;
338			pmp->pm_flags &= ~MSDOSFSMNT_RONLY;
339			MNT_ILOCK(mp);
340			mp->mnt_flag &= ~MNT_RDONLY;
341			MNT_IUNLOCK(mp);
342
343			/* Now that the volume is modifiable, mark it dirty. */
344			error = markvoldirty(pmp, 1);
345			if (error)
346				return (error);
347		}
348	}
349	/*
350	 * Not an update, or updating the name: look up the name
351	 * and verify that it refers to a sensible disk device.
352	 */
353	if (vfs_getopt(mp->mnt_optnew, "from", (void **)&from, NULL))
354		return (EINVAL);
355	NDINIT(&ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, from, td);
356	error = namei(&ndp);
357	if (error)
358		return (error);
359	devvp = ndp.ni_vp;
360	NDFREE(&ndp, NDF_ONLY_PNBUF);
361
362	if (!vn_isdisk(devvp, &error)) {
363		vput(devvp);
364		return (error);
365	}
366	/*
367	 * If mount by non-root, then verify that user has necessary
368	 * permissions on the device.
369	 */
370	accmode = VREAD;
371	if ((mp->mnt_flag & MNT_RDONLY) == 0)
372		accmode |= VWRITE;
373	error = VOP_ACCESS(devvp, accmode, td->td_ucred, td);
374	if (error)
375		error = priv_check(td, PRIV_VFS_MOUNT_PERM);
376	if (error) {
377		vput(devvp);
378		return (error);
379	}
380	if ((mp->mnt_flag & MNT_UPDATE) == 0) {
381		error = mountmsdosfs(devvp, mp);
382#ifdef MSDOSFS_DEBUG		/* only needed for the printf below */
383		pmp = VFSTOMSDOSFS(mp);
384#endif
385	} else {
386		vput(devvp);
387		if (devvp != pmp->pm_devvp)
388			return (EINVAL);	/* XXX needs translation */
389	}
390	if (error) {
391		vrele(devvp);
392		return (error);
393	}
394
395	error = update_mp(mp, td);
396	if (error) {
397		if ((mp->mnt_flag & MNT_UPDATE) == 0)
398			msdosfs_unmount(mp, MNT_FORCE);
399		return error;
400	}
401
402	vfs_mountedfrom(mp, from);
403#ifdef MSDOSFS_DEBUG
404	printf("msdosfs_mount(): mp %p, pmp %p, inusemap %p\n", mp, pmp, pmp->pm_inusemap);
405#endif
406	return (0);
407}
408
409static int
410mountmsdosfs(struct vnode *devvp, struct mount *mp)
411{
412	struct msdosfsmount *pmp;
413	struct buf *bp;
414	struct cdev *dev;
415	union bootsector *bsp;
416	struct byte_bpb33 *b33;
417	struct byte_bpb50 *b50;
418	struct byte_bpb710 *b710;
419	u_int8_t SecPerClust;
420	u_long clusters;
421	int ronly, error;
422	struct g_consumer *cp;
423	struct bufobj *bo;
424
425	bp = NULL;		/* This and pmp both used in error_exit. */
426	pmp = NULL;
427	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
428
429	dev = devvp->v_rdev;
430	dev_ref(dev);
431	DROP_GIANT();
432	g_topology_lock();
433	error = g_vfs_open(devvp, &cp, "msdosfs", ronly ? 0 : 1);
434	g_topology_unlock();
435	PICKUP_GIANT();
436	VOP_UNLOCK(devvp, 0);
437	if (error)
438		goto error_exit;
439
440	bo = &devvp->v_bufobj;
441
442	/*
443	 * Read the boot sector of the filesystem, and then check the
444	 * boot signature.  If not a dos boot sector then error out.
445	 *
446	 * NOTE: 8192 is a magic size that works for ffs.
447	 */
448	error = bread(devvp, 0, 8192, NOCRED, &bp);
449	if (error)
450		goto error_exit;
451	bp->b_flags |= B_AGE;
452	bsp = (union bootsector *)bp->b_data;
453	b33 = (struct byte_bpb33 *)bsp->bs33.bsBPB;
454	b50 = (struct byte_bpb50 *)bsp->bs50.bsBPB;
455	b710 = (struct byte_bpb710 *)bsp->bs710.bsBPB;
456
457#ifndef MSDOSFS_NOCHECKSIG
458	if (bsp->bs50.bsBootSectSig0 != BOOTSIG0
459	    || bsp->bs50.bsBootSectSig1 != BOOTSIG1) {
460		error = EINVAL;
461		goto error_exit;
462	}
463#endif
464
465	pmp = malloc(sizeof *pmp, M_MSDOSFSMNT, M_WAITOK | M_ZERO);
466	pmp->pm_mountp = mp;
467	pmp->pm_cp = cp;
468	pmp->pm_bo = bo;
469
470	lockinit(&pmp->pm_fatlock, 0, msdosfs_lock_msg, 0, 0);
471
472	/*
473	 * Initialize ownerships and permissions, since nothing else will
474	 * initialize them iff we are mounting root.
475	 */
476	pmp->pm_uid = UID_ROOT;
477	pmp->pm_gid = GID_WHEEL;
478	pmp->pm_mask = pmp->pm_dirmask = S_IXUSR | S_IXGRP | S_IXOTH |
479	    S_IRUSR | S_IRGRP | S_IROTH | S_IWUSR;
480
481	/*
482	 * Experimental support for large MS-DOS filesystems.
483	 * WARNING: This uses at least 32 bytes of kernel memory (which is not
484	 * reclaimed until the FS is unmounted) for each file on disk to map
485	 * between the 32-bit inode numbers used by VFS and the 64-bit
486	 * pseudo-inode numbers used internally by msdosfs. This is only
487	 * safe to use in certain controlled situations (e.g. read-only FS
488	 * with less than 1 million files).
489	 * Since the mappings do not persist across unmounts (or reboots), these
490	 * filesystems are not suitable for exporting through NFS, or any other
491	 * application that requires fixed inode numbers.
492	 */
493	vfs_flagopt(mp->mnt_optnew, "large", &pmp->pm_flags, MSDOSFS_LARGEFS);
494
495	/*
496	 * Compute several useful quantities from the bpb in the
497	 * bootsector.  Copy in the dos 5 variant of the bpb then fix up
498	 * the fields that are different between dos 5 and dos 3.3.
499	 */
500	SecPerClust = b50->bpbSecPerClust;
501	pmp->pm_BytesPerSec = getushort(b50->bpbBytesPerSec);
502	if (pmp->pm_BytesPerSec < DEV_BSIZE) {
503		error = EINVAL;
504		goto error_exit;
505	}
506	pmp->pm_ResSectors = getushort(b50->bpbResSectors);
507	pmp->pm_FATs = b50->bpbFATs;
508	pmp->pm_RootDirEnts = getushort(b50->bpbRootDirEnts);
509	pmp->pm_Sectors = getushort(b50->bpbSectors);
510	pmp->pm_FATsecs = getushort(b50->bpbFATsecs);
511	pmp->pm_SecPerTrack = getushort(b50->bpbSecPerTrack);
512	pmp->pm_Heads = getushort(b50->bpbHeads);
513	pmp->pm_Media = b50->bpbMedia;
514
515	/* calculate the ratio of sector size to DEV_BSIZE */
516	pmp->pm_BlkPerSec = pmp->pm_BytesPerSec / DEV_BSIZE;
517
518	/*
519	 * We don't check pm_Heads nor pm_SecPerTrack, because
520	 * these may not be set for EFI file systems. We don't
521	 * use these anyway, so we're unaffected if they are
522	 * invalid.
523	 */
524	if (!pmp->pm_BytesPerSec || !SecPerClust) {
525		error = EINVAL;
526		goto error_exit;
527	}
528
529	if (pmp->pm_Sectors == 0) {
530		pmp->pm_HiddenSects = getulong(b50->bpbHiddenSecs);
531		pmp->pm_HugeSectors = getulong(b50->bpbHugeSectors);
532	} else {
533		pmp->pm_HiddenSects = getushort(b33->bpbHiddenSecs);
534		pmp->pm_HugeSectors = pmp->pm_Sectors;
535	}
536	if (!(pmp->pm_flags & MSDOSFS_LARGEFS)) {
537		if (pmp->pm_HugeSectors > 0xffffffff /
538		    (pmp->pm_BytesPerSec / sizeof(struct direntry)) + 1) {
539			/*
540			 * We cannot deal currently with this size of disk
541			 * due to fileid limitations (see msdosfs_getattr and
542			 * msdosfs_readdir)
543			 */
544			error = EINVAL;
545			vfs_mount_error(mp,
546			    "Disk too big, try '-o large' mount option");
547			goto error_exit;
548		}
549	}
550
551	if (pmp->pm_RootDirEnts == 0) {
552		if (pmp->pm_Sectors
553		    || pmp->pm_FATsecs
554		    || getushort(b710->bpbFSVers)) {
555			error = EINVAL;
556			printf("mountmsdosfs(): bad FAT32 filesystem\n");
557			goto error_exit;
558		}
559		pmp->pm_fatmask = FAT32_MASK;
560		pmp->pm_fatmult = 4;
561		pmp->pm_fatdiv = 1;
562		pmp->pm_FATsecs = getulong(b710->bpbBigFATsecs);
563		if (getushort(b710->bpbExtFlags) & FATMIRROR)
564			pmp->pm_curfat = getushort(b710->bpbExtFlags) & FATNUM;
565		else
566			pmp->pm_flags |= MSDOSFS_FATMIRROR;
567	} else
568		pmp->pm_flags |= MSDOSFS_FATMIRROR;
569
570	/*
571	 * Check a few values (could do some more):
572	 * - logical sector size: power of 2, >= block size
573	 * - sectors per cluster: power of 2, >= 1
574	 * - number of sectors:   >= 1, <= size of partition
575	 * - number of FAT sectors: >= 1
576	 */
577	if ( (SecPerClust == 0)
578	  || (SecPerClust & (SecPerClust - 1))
579	  || (pmp->pm_BytesPerSec < DEV_BSIZE)
580	  || (pmp->pm_BytesPerSec & (pmp->pm_BytesPerSec - 1))
581	  || (pmp->pm_HugeSectors == 0)
582	  || (pmp->pm_FATsecs == 0)
583	) {
584		error = EINVAL;
585		goto error_exit;
586	}
587
588	pmp->pm_HugeSectors *= pmp->pm_BlkPerSec;
589	pmp->pm_HiddenSects *= pmp->pm_BlkPerSec;	/* XXX not used? */
590	pmp->pm_FATsecs     *= pmp->pm_BlkPerSec;
591	SecPerClust         *= pmp->pm_BlkPerSec;
592
593	pmp->pm_fatblk = pmp->pm_ResSectors * pmp->pm_BlkPerSec;
594
595	if (FAT32(pmp)) {
596		pmp->pm_rootdirblk = getulong(b710->bpbRootClust);
597		pmp->pm_firstcluster = pmp->pm_fatblk
598			+ (pmp->pm_FATs * pmp->pm_FATsecs);
599		pmp->pm_fsinfo = getushort(b710->bpbFSInfo) * pmp->pm_BlkPerSec;
600	} else {
601		pmp->pm_rootdirblk = pmp->pm_fatblk +
602			(pmp->pm_FATs * pmp->pm_FATsecs);
603		pmp->pm_rootdirsize = (pmp->pm_RootDirEnts * sizeof(struct direntry)
604				       + DEV_BSIZE - 1)
605			/ DEV_BSIZE; /* in blocks */
606		pmp->pm_firstcluster = pmp->pm_rootdirblk + pmp->pm_rootdirsize;
607	}
608
609	pmp->pm_maxcluster = (pmp->pm_HugeSectors - pmp->pm_firstcluster) /
610	    SecPerClust + 1;
611	pmp->pm_fatsize = pmp->pm_FATsecs * DEV_BSIZE;	/* XXX not used? */
612
613	if (pmp->pm_fatmask == 0) {
614		if (pmp->pm_maxcluster
615		    <= ((CLUST_RSRVD - CLUST_FIRST) & FAT12_MASK)) {
616			/*
617			 * This will usually be a floppy disk. This size makes
618			 * sure that one fat entry will not be split across
619			 * multiple blocks.
620			 */
621			pmp->pm_fatmask = FAT12_MASK;
622			pmp->pm_fatmult = 3;
623			pmp->pm_fatdiv = 2;
624		} else {
625			pmp->pm_fatmask = FAT16_MASK;
626			pmp->pm_fatmult = 2;
627			pmp->pm_fatdiv = 1;
628		}
629	}
630
631	clusters = (pmp->pm_fatsize / pmp->pm_fatmult) * pmp->pm_fatdiv;
632	if (pmp->pm_maxcluster >= clusters) {
633		printf("Warning: number of clusters (%ld) exceeds FAT "
634		    "capacity (%ld)\n", pmp->pm_maxcluster + 1, clusters);
635		pmp->pm_maxcluster = clusters - 1;
636	}
637
638	if (FAT12(pmp))
639		pmp->pm_fatblocksize = 3 * 512;
640	else
641		pmp->pm_fatblocksize = PAGE_SIZE;
642	pmp->pm_fatblocksize = roundup(pmp->pm_fatblocksize,
643	    pmp->pm_BytesPerSec);
644	pmp->pm_fatblocksec = pmp->pm_fatblocksize / DEV_BSIZE;
645	pmp->pm_bnshift = ffs(DEV_BSIZE) - 1;
646
647	/*
648	 * Compute mask and shift value for isolating cluster relative byte
649	 * offsets and cluster numbers from a file offset.
650	 */
651	pmp->pm_bpcluster = SecPerClust * DEV_BSIZE;
652	pmp->pm_crbomask = pmp->pm_bpcluster - 1;
653	pmp->pm_cnshift = ffs(pmp->pm_bpcluster) - 1;
654
655	/*
656	 * Check for valid cluster size
657	 * must be a power of 2
658	 */
659	if (pmp->pm_bpcluster ^ (1 << pmp->pm_cnshift)) {
660		error = EINVAL;
661		goto error_exit;
662	}
663
664	/*
665	 * Release the bootsector buffer.
666	 */
667	brelse(bp);
668	bp = NULL;
669
670	/*
671	 * Check the fsinfo sector if we have one.  Silently fix up our
672	 * in-core copy of fp->fsinxtfree if it is unknown (0xffffffff)
673	 * or too large.  Ignore fp->fsinfree for now, since we need to
674	 * read the entire FAT anyway to fill the inuse map.
675	 */
676	if (pmp->pm_fsinfo) {
677		struct fsinfo *fp;
678
679		if ((error = bread(devvp, pmp->pm_fsinfo, pmp->pm_BytesPerSec,
680		    NOCRED, &bp)) != 0)
681			goto error_exit;
682		fp = (struct fsinfo *)bp->b_data;
683		if (!bcmp(fp->fsisig1, "RRaA", 4)
684		    && !bcmp(fp->fsisig2, "rrAa", 4)
685		    && !bcmp(fp->fsisig3, "\0\0\125\252", 4)) {
686			pmp->pm_nxtfree = getulong(fp->fsinxtfree);
687			if (pmp->pm_nxtfree > pmp->pm_maxcluster)
688				pmp->pm_nxtfree = CLUST_FIRST;
689		} else
690			pmp->pm_fsinfo = 0;
691		brelse(bp);
692		bp = NULL;
693	}
694
695	/*
696	 * Finish initializing pmp->pm_nxtfree (just in case the first few
697	 * sectors aren't properly reserved in the FAT).  This completes
698	 * the fixup for fp->fsinxtfree, and fixes up the zero-initialized
699	 * value if there is no fsinfo.  We will use pmp->pm_nxtfree
700	 * internally even if there is no fsinfo.
701	 */
702	if (pmp->pm_nxtfree < CLUST_FIRST)
703		pmp->pm_nxtfree = CLUST_FIRST;
704
705	/*
706	 * Allocate memory for the bitmap of allocated clusters, and then
707	 * fill it in.
708	 */
709	pmp->pm_inusemap = malloc(howmany(pmp->pm_maxcluster + 1, N_INUSEBITS)
710				  * sizeof(*pmp->pm_inusemap),
711				  M_MSDOSFSFAT, M_WAITOK);
712
713	/*
714	 * fillinusemap() needs pm_devvp.
715	 */
716	pmp->pm_devvp = devvp;
717	pmp->pm_dev = dev;
718
719	/*
720	 * Have the inuse map filled in.
721	 */
722	MSDOSFS_LOCK_MP(pmp);
723	error = fillinusemap(pmp);
724	MSDOSFS_UNLOCK_MP(pmp);
725	if (error != 0)
726		goto error_exit;
727
728	/*
729	 * If they want fat updates to be synchronous then let them suffer
730	 * the performance degradation in exchange for the on disk copy of
731	 * the fat being correct just about all the time.  I suppose this
732	 * would be a good thing to turn on if the kernel is still flakey.
733	 */
734	if (mp->mnt_flag & MNT_SYNCHRONOUS)
735		pmp->pm_flags |= MSDOSFSMNT_WAITONFAT;
736
737	/*
738	 * Finish up.
739	 */
740	if (ronly)
741		pmp->pm_flags |= MSDOSFSMNT_RONLY;
742	else {
743		if ((error = markvoldirty(pmp, 1)) != 0) {
744			(void)markvoldirty(pmp, 0);
745			goto error_exit;
746		}
747		pmp->pm_fmod = 1;
748	}
749	mp->mnt_data =  pmp;
750	mp->mnt_stat.f_fsid.val[0] = dev2udev(dev);
751	mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
752	MNT_ILOCK(mp);
753	mp->mnt_flag |= MNT_LOCAL;
754	mp->mnt_kern_flag |= MNTK_MPSAFE;
755	MNT_IUNLOCK(mp);
756
757	if (pmp->pm_flags & MSDOSFS_LARGEFS)
758		msdosfs_fileno_init(mp);
759
760	return 0;
761
762error_exit:
763	if (bp)
764		brelse(bp);
765	if (cp != NULL) {
766		DROP_GIANT();
767		g_topology_lock();
768		g_vfs_close(cp);
769		g_topology_unlock();
770		PICKUP_GIANT();
771	}
772	if (pmp) {
773		lockdestroy(&pmp->pm_fatlock);
774		if (pmp->pm_inusemap)
775			free(pmp->pm_inusemap, M_MSDOSFSFAT);
776		free(pmp, M_MSDOSFSMNT);
777		mp->mnt_data = NULL;
778	}
779	dev_rel(dev);
780	return (error);
781}
782
783/*
784 * Unmount the filesystem described by mp.
785 */
786static int
787msdosfs_unmount(struct mount *mp, int mntflags)
788{
789	struct msdosfsmount *pmp;
790	int error, flags;
791
792	flags = 0;
793	if (mntflags & MNT_FORCE)
794		flags |= FORCECLOSE;
795	error = vflush(mp, 0, flags, curthread);
796	if (error && error != ENXIO)
797		return error;
798	pmp = VFSTOMSDOSFS(mp);
799	if ((pmp->pm_flags & MSDOSFSMNT_RONLY) == 0) {
800		error = markvoldirty(pmp, 0);
801		if (error && error != ENXIO) {
802			(void)markvoldirty(pmp, 1);
803			return (error);
804		}
805	}
806	if (pmp->pm_flags & MSDOSFSMNT_KICONV && msdosfs_iconv) {
807		if (pmp->pm_w2u)
808			msdosfs_iconv->close(pmp->pm_w2u);
809		if (pmp->pm_u2w)
810			msdosfs_iconv->close(pmp->pm_u2w);
811		if (pmp->pm_d2u)
812			msdosfs_iconv->close(pmp->pm_d2u);
813		if (pmp->pm_u2d)
814			msdosfs_iconv->close(pmp->pm_u2d);
815	}
816
817#ifdef MSDOSFS_DEBUG
818	{
819		struct vnode *vp = pmp->pm_devvp;
820		struct bufobj *bo;
821
822		bo = &vp->v_bufobj;
823		BO_LOCK(bo);
824		VI_LOCK(vp);
825		vn_printf(vp,
826		    "msdosfs_umount(): just before calling VOP_CLOSE()\n");
827		printf("freef %p, freeb %p, mount %p\n",
828		    TAILQ_NEXT(vp, v_freelist), vp->v_freelist.tqe_prev,
829		    vp->v_mount);
830		printf("cleanblkhd %p, dirtyblkhd %p, numoutput %ld, type %d\n",
831		    TAILQ_FIRST(&vp->v_bufobj.bo_clean.bv_hd),
832		    TAILQ_FIRST(&vp->v_bufobj.bo_dirty.bv_hd),
833		    vp->v_bufobj.bo_numoutput, vp->v_type);
834		VI_UNLOCK(vp);
835		BO_UNLOCK(bo);
836	}
837#endif
838	DROP_GIANT();
839	g_topology_lock();
840	g_vfs_close(pmp->pm_cp);
841	g_topology_unlock();
842	PICKUP_GIANT();
843	vrele(pmp->pm_devvp);
844	dev_rel(pmp->pm_dev);
845	free(pmp->pm_inusemap, M_MSDOSFSFAT);
846	if (pmp->pm_flags & MSDOSFS_LARGEFS)
847		msdosfs_fileno_free(mp);
848	lockdestroy(&pmp->pm_fatlock);
849	free(pmp, M_MSDOSFSMNT);
850	mp->mnt_data = NULL;
851	MNT_ILOCK(mp);
852	mp->mnt_flag &= ~MNT_LOCAL;
853	MNT_IUNLOCK(mp);
854	return (error);
855}
856
857static int
858msdosfs_root(struct mount *mp, int flags, struct vnode **vpp)
859{
860	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
861	struct denode *ndep;
862	int error;
863
864#ifdef MSDOSFS_DEBUG
865	printf("msdosfs_root(); mp %p, pmp %p\n", mp, pmp);
866#endif
867	error = deget(pmp, MSDOSFSROOT, MSDOSFSROOT_OFS, &ndep);
868	if (error)
869		return (error);
870	*vpp = DETOV(ndep);
871	return (0);
872}
873
874static int
875msdosfs_statfs(struct mount *mp, struct statfs *sbp)
876{
877	struct msdosfsmount *pmp;
878
879	pmp = VFSTOMSDOSFS(mp);
880	sbp->f_bsize = pmp->pm_bpcluster;
881	sbp->f_iosize = pmp->pm_bpcluster;
882	sbp->f_blocks = pmp->pm_maxcluster + 1;
883	sbp->f_bfree = pmp->pm_freeclustercount;
884	sbp->f_bavail = pmp->pm_freeclustercount;
885	sbp->f_files = pmp->pm_RootDirEnts;	/* XXX */
886	sbp->f_ffree = 0;	/* what to put in here? */
887	return (0);
888}
889
890static int
891msdosfs_sync(struct mount *mp, int waitfor)
892{
893	struct vnode *vp, *nvp;
894	struct thread *td;
895	struct denode *dep;
896	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
897	int error, allerror = 0;
898
899	td = curthread;
900
901	/*
902	 * If we ever switch to not updating all of the fats all the time,
903	 * this would be the place to update them from the first one.
904	 */
905	if (pmp->pm_fmod != 0) {
906		if (pmp->pm_flags & MSDOSFSMNT_RONLY)
907			panic("msdosfs_sync: rofs mod");
908		else {
909			/* update fats here */
910		}
911	}
912	/*
913	 * Write back each (modified) denode.
914	 */
915	MNT_ILOCK(mp);
916loop:
917	MNT_VNODE_FOREACH(vp, mp, nvp) {
918		VI_LOCK(vp);
919		if (vp->v_type == VNON || (vp->v_iflag & VI_DOOMED)) {
920			VI_UNLOCK(vp);
921			continue;
922		}
923		MNT_IUNLOCK(mp);
924		dep = VTODE(vp);
925		if ((dep->de_flag &
926		    (DE_ACCESS | DE_CREATE | DE_UPDATE | DE_MODIFIED)) == 0 &&
927		    (vp->v_bufobj.bo_dirty.bv_cnt == 0 ||
928		    waitfor == MNT_LAZY)) {
929			VI_UNLOCK(vp);
930			MNT_ILOCK(mp);
931			continue;
932		}
933		error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK, td);
934		if (error) {
935			MNT_ILOCK(mp);
936			if (error == ENOENT)
937				goto loop;
938			continue;
939		}
940		error = VOP_FSYNC(vp, waitfor, td);
941		if (error)
942			allerror = error;
943		VOP_UNLOCK(vp, 0);
944		vrele(vp);
945		MNT_ILOCK(mp);
946	}
947	MNT_IUNLOCK(mp);
948
949	/*
950	 * Flush filesystem control info.
951	 */
952	if (waitfor != MNT_LAZY) {
953		vn_lock(pmp->pm_devvp, LK_EXCLUSIVE | LK_RETRY);
954		error = VOP_FSYNC(pmp->pm_devvp, waitfor, td);
955		if (error)
956			allerror = error;
957		VOP_UNLOCK(pmp->pm_devvp, 0);
958	}
959	return (allerror);
960}
961
962static int
963msdosfs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
964{
965	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
966	struct defid *defhp = (struct defid *) fhp;
967	struct denode *dep;
968	int error;
969
970	error = deget(pmp, defhp->defid_dirclust, defhp->defid_dirofs, &dep);
971	if (error) {
972		*vpp = NULLVP;
973		return (error);
974	}
975	*vpp = DETOV(dep);
976	vnode_create_vobject(*vpp, dep->de_FileSize, curthread);
977	return (0);
978}
979
980static struct vfsops msdosfs_vfsops = {
981	.vfs_fhtovp =		msdosfs_fhtovp,
982	.vfs_mount =		msdosfs_mount,
983	.vfs_cmount =		msdosfs_cmount,
984	.vfs_root =		msdosfs_root,
985	.vfs_statfs =		msdosfs_statfs,
986	.vfs_sync =		msdosfs_sync,
987	.vfs_unmount =		msdosfs_unmount,
988};
989
990VFS_SET(msdosfs_vfsops, msdosfs, 0);
991MODULE_VERSION(msdosfs, 1);
992