xref: /freebsd/sys/fs/msdosfs/msdosfs_vfsops.c (revision b28624fde638caadd4a89f50c9b7e7da0f98c4d2)
1 /* $FreeBSD$ */
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 /* List of mount options we support */
78 static const char *msdosfs_opts[] = {
79 	"from",
80 	"noatime", "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 
98 MALLOC_DEFINE(M_MSDOSFSMNT, "msdosfs_mount", "MSDOSFS mount structure");
99 static MALLOC_DEFINE(M_MSDOSFSFAT, "msdosfs_fat", "MSDOSFS file allocation table");
100 
101 struct iconv_functions *msdosfs_iconv;
102 
103 static int	update_mp(struct mount *mp, struct thread *td);
104 static int	mountmsdosfs(struct vnode *devvp, struct mount *mp,
105 		    struct thread *td);
106 static vfs_fhtovp_t	msdosfs_fhtovp;
107 static vfs_mount_t	msdosfs_mount;
108 static vfs_root_t	msdosfs_root;
109 static vfs_statfs_t	msdosfs_statfs;
110 static vfs_sync_t	msdosfs_sync;
111 static vfs_unmount_t	msdosfs_unmount;
112 
113 /* Maximum length of a character set name (arbitrary). */
114 #define	MAXCSLEN	64
115 
116 static int
117 update_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 : MSDOSFSMNT_SHORTNAME;
190 			vput(rootvp);
191 		}
192 	}
193 	return 0;
194 }
195 
196 static int
197 msdosfs_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  */
234 static int
235 msdosfs_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 
255 		pmp = VFSTOMSDOSFS(mp);
256 		if (vfs_flagopt(mp->mnt_optnew, "export", NULL, 0)) {
257 			/*
258 			 * Forbid export requests if filesystem has
259 			 * MSDOSFS_LARGEFS flag set.
260 			 */
261 			if ((pmp->pm_flags & MSDOSFS_LARGEFS) != 0) {
262 				vfs_mount_error(mp,
263 				    "MSDOSFS_LARGEFS flag set, cannot export");
264 				return (EOPNOTSUPP);
265 			}
266 		}
267 		if (!(pmp->pm_flags & MSDOSFSMNT_RONLY) &&
268 		    vfs_flagopt(mp->mnt_optnew, "ro", NULL, 0)) {
269 			error = VFS_SYNC(mp, MNT_WAIT, td);
270 			if (error)
271 				return (error);
272 			flags = WRITECLOSE;
273 			if (mp->mnt_flag & MNT_FORCE)
274 				flags |= FORCECLOSE;
275 			error = vflush(mp, 0, flags, td);
276 			if (error)
277 				return (error);
278 			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 
390 static int
391 mountmsdosfs(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;		/* This and pmp 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, MSDOSFS_LARGEFS);
471 
472 	/*
473 	 * Compute several useful quantities from the bpb in the
474 	 * bootsector.  Copy in the dos 5 variant of the bpb then fix up
475 	 * the fields that are different between dos 5 and dos 3.3.
476 	 */
477 	SecPerClust = b50->bpbSecPerClust;
478 	pmp->pm_BytesPerSec = getushort(b50->bpbBytesPerSec);
479 	if (pmp->pm_BytesPerSec < DEV_BSIZE) {
480 		error = EINVAL;
481 		goto error_exit;
482 	}
483 	pmp->pm_ResSectors = getushort(b50->bpbResSectors);
484 	pmp->pm_FATs = b50->bpbFATs;
485 	pmp->pm_RootDirEnts = getushort(b50->bpbRootDirEnts);
486 	pmp->pm_Sectors = getushort(b50->bpbSectors);
487 	pmp->pm_FATsecs = getushort(b50->bpbFATsecs);
488 	pmp->pm_SecPerTrack = getushort(b50->bpbSecPerTrack);
489 	pmp->pm_Heads = getushort(b50->bpbHeads);
490 	pmp->pm_Media = b50->bpbMedia;
491 
492 	/* calculate the ratio of sector size to DEV_BSIZE */
493 	pmp->pm_BlkPerSec = pmp->pm_BytesPerSec / DEV_BSIZE;
494 
495 	/* XXX - We should probably check more values here */
496 	if (!pmp->pm_BytesPerSec || !SecPerClust
497 		|| !pmp->pm_Heads
498 #ifdef PC98
499     		|| !pmp->pm_SecPerTrack || pmp->pm_SecPerTrack > 255) {
500 #else
501 		|| !pmp->pm_SecPerTrack || pmp->pm_SecPerTrack > 63) {
502 #endif
503 		error = EINVAL;
504 		goto error_exit;
505 	}
506 
507 	if (pmp->pm_Sectors == 0) {
508 		pmp->pm_HiddenSects = getulong(b50->bpbHiddenSecs);
509 		pmp->pm_HugeSectors = getulong(b50->bpbHugeSectors);
510 	} else {
511 		pmp->pm_HiddenSects = getushort(b33->bpbHiddenSecs);
512 		pmp->pm_HugeSectors = pmp->pm_Sectors;
513 	}
514 	if (!(pmp->pm_flags & MSDOSFS_LARGEFS)) {
515 		if (pmp->pm_HugeSectors > 0xffffffff /
516 		    (pmp->pm_BytesPerSec / sizeof(struct direntry)) + 1) {
517 			/*
518 			 * We cannot deal currently with this size of disk
519 			 * due to fileid limitations (see msdosfs_getattr and
520 			 * msdosfs_readdir)
521 			 */
522 			error = EINVAL;
523 			vfs_mount_error(mp,
524 			    "Disk too big, try '-o large' mount option");
525 			goto error_exit;
526 		}
527 	}
528 
529 	if (pmp->pm_RootDirEnts == 0) {
530 		if (pmp->pm_Sectors
531 		    || pmp->pm_FATsecs
532 		    || getushort(b710->bpbFSVers)) {
533 			error = EINVAL;
534 			printf("mountmsdosfs(): bad FAT32 filesystem\n");
535 			goto error_exit;
536 		}
537 		pmp->pm_fatmask = FAT32_MASK;
538 		pmp->pm_fatmult = 4;
539 		pmp->pm_fatdiv = 1;
540 		pmp->pm_FATsecs = getulong(b710->bpbBigFATsecs);
541 		if (getushort(b710->bpbExtFlags) & FATMIRROR)
542 			pmp->pm_curfat = getushort(b710->bpbExtFlags) & FATNUM;
543 		else
544 			pmp->pm_flags |= MSDOSFS_FATMIRROR;
545 	} else
546 		pmp->pm_flags |= MSDOSFS_FATMIRROR;
547 
548 	/*
549 	 * Check a few values (could do some more):
550 	 * - logical sector size: power of 2, >= block size
551 	 * - sectors per cluster: power of 2, >= 1
552 	 * - number of sectors:   >= 1, <= size of partition
553 	 * - number of FAT sectors: >= 1
554 	 */
555 	if ( (SecPerClust == 0)
556 	  || (SecPerClust & (SecPerClust - 1))
557 	  || (pmp->pm_BytesPerSec < DEV_BSIZE)
558 	  || (pmp->pm_BytesPerSec & (pmp->pm_BytesPerSec - 1))
559 	  || (pmp->pm_HugeSectors == 0)
560 	  || (pmp->pm_FATsecs == 0)
561 	) {
562 		error = EINVAL;
563 		goto error_exit;
564 	}
565 
566 	pmp->pm_HugeSectors *= pmp->pm_BlkPerSec;
567 	pmp->pm_HiddenSects *= pmp->pm_BlkPerSec;	/* XXX not used? */
568 	pmp->pm_FATsecs     *= pmp->pm_BlkPerSec;
569 	SecPerClust         *= pmp->pm_BlkPerSec;
570 
571 	pmp->pm_fatblk = pmp->pm_ResSectors * pmp->pm_BlkPerSec;
572 
573 	if (FAT32(pmp)) {
574 		pmp->pm_rootdirblk = getulong(b710->bpbRootClust);
575 		pmp->pm_firstcluster = pmp->pm_fatblk
576 			+ (pmp->pm_FATs * pmp->pm_FATsecs);
577 		pmp->pm_fsinfo = getushort(b710->bpbFSInfo) * pmp->pm_BlkPerSec;
578 	} else {
579 		pmp->pm_rootdirblk = pmp->pm_fatblk +
580 			(pmp->pm_FATs * pmp->pm_FATsecs);
581 		pmp->pm_rootdirsize = (pmp->pm_RootDirEnts * sizeof(struct direntry)
582 				       + DEV_BSIZE - 1)
583 			/ DEV_BSIZE; /* in blocks */
584 		pmp->pm_firstcluster = pmp->pm_rootdirblk + pmp->pm_rootdirsize;
585 	}
586 
587 	pmp->pm_maxcluster = (pmp->pm_HugeSectors - pmp->pm_firstcluster) /
588 	    SecPerClust + 1;
589 	pmp->pm_fatsize = pmp->pm_FATsecs * DEV_BSIZE;	/* XXX not used? */
590 
591 	if (pmp->pm_fatmask == 0) {
592 		if (pmp->pm_maxcluster
593 		    <= ((CLUST_RSRVD - CLUST_FIRST) & FAT12_MASK)) {
594 			/*
595 			 * This will usually be a floppy disk. This size makes
596 			 * sure that one fat entry will not be split across
597 			 * multiple blocks.
598 			 */
599 			pmp->pm_fatmask = FAT12_MASK;
600 			pmp->pm_fatmult = 3;
601 			pmp->pm_fatdiv = 2;
602 		} else {
603 			pmp->pm_fatmask = FAT16_MASK;
604 			pmp->pm_fatmult = 2;
605 			pmp->pm_fatdiv = 1;
606 		}
607 	}
608 
609 	clusters = (pmp->pm_fatsize / pmp->pm_fatmult) * pmp->pm_fatdiv;
610 	if (pmp->pm_maxcluster >= clusters) {
611 		printf("Warning: number of clusters (%ld) exceeds FAT "
612 		    "capacity (%ld)\n", pmp->pm_maxcluster + 1, clusters);
613 		pmp->pm_maxcluster = clusters - 1;
614 	}
615 
616 	if (FAT12(pmp))
617 		pmp->pm_fatblocksize = 3 * 512;
618 	else
619 		pmp->pm_fatblocksize = PAGE_SIZE;
620 	pmp->pm_fatblocksize = roundup(pmp->pm_fatblocksize,
621 	    pmp->pm_BytesPerSec);
622 	pmp->pm_fatblocksec = pmp->pm_fatblocksize / DEV_BSIZE;
623 	pmp->pm_bnshift = ffs(DEV_BSIZE) - 1;
624 
625 	/*
626 	 * Compute mask and shift value for isolating cluster relative byte
627 	 * offsets and cluster numbers from a file offset.
628 	 */
629 	pmp->pm_bpcluster = SecPerClust * DEV_BSIZE;
630 	pmp->pm_crbomask = pmp->pm_bpcluster - 1;
631 	pmp->pm_cnshift = ffs(pmp->pm_bpcluster) - 1;
632 
633 	/*
634 	 * Check for valid cluster size
635 	 * must be a power of 2
636 	 */
637 	if (pmp->pm_bpcluster ^ (1 << pmp->pm_cnshift)) {
638 		error = EINVAL;
639 		goto error_exit;
640 	}
641 
642 	/*
643 	 * Release the bootsector buffer.
644 	 */
645 	brelse(bp);
646 	bp = NULL;
647 
648 	/*
649 	 * Check the fsinfo sector if we have one.  Silently fix up our
650 	 * in-core copy of fp->fsinxtfree if it is unknown (0xffffffff)
651 	 * or too large.  Ignore fp->fsinfree for now, since we need to
652 	 * read the entire FAT anyway to fill the inuse map.
653 	 */
654 	if (pmp->pm_fsinfo) {
655 		struct fsinfo *fp;
656 
657 		if ((error = bread(devvp, pmp->pm_fsinfo, pmp->pm_BytesPerSec,
658 		    NOCRED, &bp)) != 0)
659 			goto error_exit;
660 		fp = (struct fsinfo *)bp->b_data;
661 		if (!bcmp(fp->fsisig1, "RRaA", 4)
662 		    && !bcmp(fp->fsisig2, "rrAa", 4)
663 		    && !bcmp(fp->fsisig3, "\0\0\125\252", 4)) {
664 			pmp->pm_nxtfree = getulong(fp->fsinxtfree);
665 			if (pmp->pm_nxtfree > pmp->pm_maxcluster)
666 				pmp->pm_nxtfree = CLUST_FIRST;
667 		} else
668 			pmp->pm_fsinfo = 0;
669 		brelse(bp);
670 		bp = NULL;
671 	}
672 
673 	/*
674 	 * Finish initializing pmp->pm_nxtfree (just in case the first few
675 	 * sectors aren't properly reserved in the FAT).  This completes
676 	 * the fixup for fp->fsinxtfree, and fixes up the zero-initialized
677 	 * value if there is no fsinfo.  We will use pmp->pm_nxtfree
678 	 * internally even if there is no fsinfo.
679 	 */
680 	if (pmp->pm_nxtfree < CLUST_FIRST)
681 		pmp->pm_nxtfree = CLUST_FIRST;
682 
683 	/*
684 	 * Allocate memory for the bitmap of allocated clusters, and then
685 	 * fill it in.
686 	 */
687 	pmp->pm_inusemap = malloc(howmany(pmp->pm_maxcluster + 1, N_INUSEBITS)
688 				  * sizeof(*pmp->pm_inusemap),
689 				  M_MSDOSFSFAT, M_WAITOK);
690 
691 	/*
692 	 * fillinusemap() needs pm_devvp.
693 	 */
694 	pmp->pm_devvp = devvp;
695 
696 	/*
697 	 * Have the inuse map filled in.
698 	 */
699 	if ((error = fillinusemap(pmp)) != 0)
700 		goto error_exit;
701 
702 	/*
703 	 * If they want fat updates to be synchronous then let them suffer
704 	 * the performance degradation in exchange for the on disk copy of
705 	 * the fat being correct just about all the time.  I suppose this
706 	 * would be a good thing to turn on if the kernel is still flakey.
707 	 */
708 	if (mp->mnt_flag & MNT_SYNCHRONOUS)
709 		pmp->pm_flags |= MSDOSFSMNT_WAITONFAT;
710 
711 	/*
712 	 * Finish up.
713 	 */
714 	if (ronly)
715 		pmp->pm_flags |= MSDOSFSMNT_RONLY;
716 	else {
717 		/* Mark the volume dirty while it is mounted read/write. */
718 		if ((error = markvoldirty(pmp, 1)) != 0)
719 			goto error_exit;
720 		pmp->pm_fmod = 1;
721 	}
722 	mp->mnt_data = (qaddr_t) pmp;
723 	mp->mnt_stat.f_fsid.val[0] = dev2udev(dev);
724 	mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
725 	MNT_ILOCK(mp);
726 	mp->mnt_flag |= MNT_LOCAL;
727 	MNT_IUNLOCK(mp);
728 
729 	if (pmp->pm_flags & MSDOSFS_LARGEFS)
730 		msdosfs_fileno_init(mp);
731 
732 	return 0;
733 
734 error_exit:
735 	if (bp)
736 		brelse(bp);
737 	if (cp != NULL) {
738 		DROP_GIANT();
739 		g_topology_lock();
740 		g_vfs_close(cp, td);
741 		g_topology_unlock();
742 		PICKUP_GIANT();
743 	}
744 	if (pmp) {
745 		if (pmp->pm_inusemap)
746 			free(pmp->pm_inusemap, M_MSDOSFSFAT);
747 		free(pmp, M_MSDOSFSMNT);
748 		mp->mnt_data = (qaddr_t)0;
749 	}
750 	return (error);
751 }
752 
753 /*
754  * Unmount the filesystem described by mp.
755  */
756 static int
757 msdosfs_unmount(struct mount *mp, int mntflags, struct thread *td)
758 {
759 	struct msdosfsmount *pmp;
760 	int error, flags;
761 
762 	flags = 0;
763 	if (mntflags & MNT_FORCE)
764 		flags |= FORCECLOSE;
765 	error = vflush(mp, 0, flags, td);
766 	if (error)
767 		return error;
768 	pmp = VFSTOMSDOSFS(mp);
769 	if (pmp->pm_flags & MSDOSFSMNT_KICONV && msdosfs_iconv) {
770 		if (pmp->pm_w2u)
771 			msdosfs_iconv->close(pmp->pm_w2u);
772 		if (pmp->pm_u2w)
773 			msdosfs_iconv->close(pmp->pm_u2w);
774 		if (pmp->pm_d2u)
775 			msdosfs_iconv->close(pmp->pm_d2u);
776 		if (pmp->pm_u2d)
777 			msdosfs_iconv->close(pmp->pm_u2d);
778 	}
779 
780 	/* If the volume was mounted read/write, mark it clean now. */
781 	if ((pmp->pm_flags & MSDOSFSMNT_RONLY) == 0) {
782 		error = markvoldirty(pmp, 0);
783 		if (error && (flags & FORCECLOSE) == 0)
784 			return (error);
785 	}
786 #ifdef MSDOSFS_DEBUG
787 	{
788 		struct vnode *vp = pmp->pm_devvp;
789 
790 		VI_LOCK(vp);
791 		vn_printf(vp,
792 		    "msdosfs_umount(): just before calling VOP_CLOSE()\n");
793 		printf("freef %p, freeb %p, mount %p\n",
794 		    TAILQ_NEXT(vp, v_freelist), vp->v_freelist.tqe_prev,
795 		    vp->v_mount);
796 		printf("cleanblkhd %p, dirtyblkhd %p, numoutput %ld, type %d\n",
797 		    TAILQ_FIRST(&vp->v_bufobj.bo_clean.bv_hd),
798 		    TAILQ_FIRST(&vp->v_bufobj.bo_dirty.bv_hd),
799 		    vp->v_bufobj.bo_numoutput, vp->v_type);
800 		VI_UNLOCK(vp);
801 	}
802 #endif
803 	DROP_GIANT();
804 	g_topology_lock();
805 	g_vfs_close(pmp->pm_cp, td);
806 	g_topology_unlock();
807 	PICKUP_GIANT();
808 	vrele(pmp->pm_devvp);
809 	free(pmp->pm_inusemap, M_MSDOSFSFAT);
810 	if (pmp->pm_flags & MSDOSFS_LARGEFS) {
811 		msdosfs_fileno_free(mp);
812 	}
813 	free(pmp, M_MSDOSFSMNT);
814 	mp->mnt_data = (qaddr_t)0;
815 	MNT_ILOCK(mp);
816 	mp->mnt_flag &= ~MNT_LOCAL;
817 	MNT_IUNLOCK(mp);
818 	return (error);
819 }
820 
821 static int
822 msdosfs_root(struct mount *mp, int flags, struct vnode **vpp, struct thread *td)
823 {
824 	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
825 	struct denode *ndep;
826 	int error;
827 
828 #ifdef MSDOSFS_DEBUG
829 	printf("msdosfs_root(); mp %p, pmp %p\n", mp, pmp);
830 #endif
831 	error = deget(pmp, MSDOSFSROOT, MSDOSFSROOT_OFS, &ndep);
832 	if (error)
833 		return (error);
834 	*vpp = DETOV(ndep);
835 	return (0);
836 }
837 
838 static int
839 msdosfs_statfs(struct mount *mp, struct statfs *sbp, struct thread *td)
840 {
841 	struct msdosfsmount *pmp;
842 
843 	pmp = VFSTOMSDOSFS(mp);
844 	sbp->f_bsize = pmp->pm_bpcluster;
845 	sbp->f_iosize = pmp->pm_bpcluster;
846 	sbp->f_blocks = pmp->pm_maxcluster + 1;
847 	sbp->f_bfree = pmp->pm_freeclustercount;
848 	sbp->f_bavail = pmp->pm_freeclustercount;
849 	sbp->f_files = pmp->pm_RootDirEnts;	/* XXX */
850 	sbp->f_ffree = 0;	/* what to put in here? */
851 	return (0);
852 }
853 
854 static int
855 msdosfs_sync(struct mount *mp, int waitfor, struct thread *td)
856 {
857 	struct vnode *vp, *nvp;
858 	struct denode *dep;
859 	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
860 	int error, allerror = 0;
861 
862 	/*
863 	 * If we ever switch to not updating all of the fats all the time,
864 	 * this would be the place to update them from the first one.
865 	 */
866 	if (pmp->pm_fmod != 0) {
867 		if (pmp->pm_flags & MSDOSFSMNT_RONLY)
868 			panic("msdosfs_sync: rofs mod");
869 		else {
870 			/* update fats here */
871 		}
872 	}
873 	/*
874 	 * Write back each (modified) denode.
875 	 */
876 	MNT_ILOCK(mp);
877 loop:
878 	MNT_VNODE_FOREACH(vp, mp, nvp) {
879 		VI_LOCK(vp);
880 		if (vp->v_type == VNON || (vp->v_iflag & VI_DOOMED)) {
881 			VI_UNLOCK(vp);
882 			continue;
883 		}
884 		MNT_IUNLOCK(mp);
885 		dep = VTODE(vp);
886 		if ((dep->de_flag &
887 		    (DE_ACCESS | DE_CREATE | DE_UPDATE | DE_MODIFIED)) == 0 &&
888 		    (vp->v_bufobj.bo_dirty.bv_cnt == 0 ||
889 		    waitfor == MNT_LAZY)) {
890 			VI_UNLOCK(vp);
891 			MNT_ILOCK(mp);
892 			continue;
893 		}
894 		error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK, td);
895 		if (error) {
896 			MNT_ILOCK(mp);
897 			if (error == ENOENT)
898 				goto loop;
899 			continue;
900 		}
901 		error = VOP_FSYNC(vp, waitfor, td);
902 		if (error)
903 			allerror = error;
904 		VOP_UNLOCK(vp, 0, td);
905 		vrele(vp);
906 		MNT_ILOCK(mp);
907 	}
908 	MNT_IUNLOCK(mp);
909 
910 	/*
911 	 * Flush filesystem control info.
912 	 */
913 	if (waitfor != MNT_LAZY) {
914 		vn_lock(pmp->pm_devvp, LK_EXCLUSIVE | LK_RETRY, td);
915 		error = VOP_FSYNC(pmp->pm_devvp, waitfor, td);
916 		if (error)
917 			allerror = error;
918 		VOP_UNLOCK(pmp->pm_devvp, 0, td);
919 	}
920 	return (allerror);
921 }
922 
923 static int
924 msdosfs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
925 {
926 	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
927 	struct defid *defhp = (struct defid *) fhp;
928 	struct denode *dep;
929 	int error;
930 
931 	error = deget(pmp, defhp->defid_dirclust, defhp->defid_dirofs, &dep);
932 	if (error) {
933 		*vpp = NULLVP;
934 		return (error);
935 	}
936 	*vpp = DETOV(dep);
937 	vnode_create_vobject(*vpp, dep->de_FileSize, curthread);
938 	return (0);
939 }
940 
941 static struct vfsops msdosfs_vfsops = {
942 	.vfs_fhtovp =		msdosfs_fhtovp,
943 	.vfs_mount =		msdosfs_mount,
944 	.vfs_cmount =		msdosfs_cmount,
945 	.vfs_root =		msdosfs_root,
946 	.vfs_statfs =		msdosfs_statfs,
947 	.vfs_sync =		msdosfs_sync,
948 	.vfs_unmount =		msdosfs_unmount,
949 };
950 
951 VFS_SET(msdosfs_vfsops, msdosfs, 0);
952 MODULE_VERSION(msdosfs, 1);
953