xref: /netbsd-src/sys/fs/msdosfs/msdosfs_vfsops.c (revision 80d9064ac03cbb6a4174695f0d5b237c8766d3d0)
1 /*	$NetBSD: msdosfs_vfsops.c,v 1.115 2014/07/18 17:24:34 maxv Exp $	*/
2 
3 /*-
4  * Copyright (C) 1994, 1995, 1997 Wolfgang Solfrank.
5  * Copyright (C) 1994, 1995, 1997 TooLs GmbH.
6  * All rights reserved.
7  * Original code by Paul Popelka (paulp@uts.amdahl.com) (see below).
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. All advertising materials mentioning features or use of this software
18  *    must display the following acknowledgement:
19  *	This product includes software developed by TooLs GmbH.
20  * 4. The name of TooLs GmbH may not be used to endorse or promote products
21  *    derived from this software without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY TOOLS GMBH ``AS IS'' AND ANY EXPRESS OR
24  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
25  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
26  * IN NO EVENT SHALL TOOLS GMBH BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
27  * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
28  * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;
29  * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
30  * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
31  * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
32  * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
33  */
34 /*
35  * Written by Paul Popelka (paulp@uts.amdahl.com)
36  *
37  * You can do anything you want with this software, just don't say you wrote
38  * it, and don't remove this notice.
39  *
40  * This software is provided "as is".
41  *
42  * The author supplies this software to be publicly redistributed on the
43  * understanding that the author is not responsible for the correct
44  * functioning of this software in any circumstances and is not liable for
45  * any damages caused by this software.
46  *
47  * October 1992
48  */
49 
50 #include <sys/cdefs.h>
51 __KERNEL_RCSID(0, "$NetBSD: msdosfs_vfsops.c,v 1.115 2014/07/18 17:24:34 maxv Exp $");
52 
53 #if defined(_KERNEL_OPT)
54 #include "opt_compat_netbsd.h"
55 #endif
56 
57 #include <sys/param.h>
58 #include <sys/systm.h>
59 #include <sys/sysctl.h>
60 #include <sys/namei.h>
61 #include <sys/proc.h>
62 #include <sys/kernel.h>
63 #include <sys/vnode.h>
64 #include <miscfs/genfs/genfs.h>
65 #include <miscfs/specfs/specdev.h> /* XXX */	/* defines v_rdev */
66 #include <sys/mount.h>
67 #include <sys/buf.h>
68 #include <sys/file.h>
69 #include <sys/device.h>
70 #include <sys/disklabel.h>
71 #include <sys/disk.h>
72 #include <sys/fstrans.h>
73 #include <sys/ioctl.h>
74 #include <sys/malloc.h>
75 #include <sys/dirent.h>
76 #include <sys/stat.h>
77 #include <sys/conf.h>
78 #include <sys/kauth.h>
79 #include <sys/module.h>
80 
81 #include <fs/msdosfs/bpb.h>
82 #include <fs/msdosfs/bootsect.h>
83 #include <fs/msdosfs/direntry.h>
84 #include <fs/msdosfs/denode.h>
85 #include <fs/msdosfs/msdosfsmount.h>
86 #include <fs/msdosfs/fat.h>
87 
88 MODULE(MODULE_CLASS_VFS, msdos, NULL);
89 
90 #ifdef MSDOSFS_DEBUG
91 #define DPRINTF(fmt, ...) uprintf("%s(): " fmt "\n", __func__, ##__VA_ARGS__)
92 #else
93 #define DPRINTF(fmt, ...)
94 #endif
95 
96 #define GEMDOSFS_BSIZE	512
97 
98 #define MSDOSFS_NAMEMAX(pmp) \
99 	(pmp)->pm_flags & MSDOSFSMNT_LONGNAME ? WIN_MAXLEN : 12
100 
101 int msdosfs_mountfs(struct vnode *, struct mount *, struct lwp *,
102     struct msdosfs_args *);
103 
104 static int update_mp(struct mount *, struct msdosfs_args *);
105 
106 MALLOC_JUSTDEFINE(M_MSDOSFSMNT, "MSDOSFS mount", "MSDOS FS mount structure");
107 MALLOC_JUSTDEFINE(M_MSDOSFSFAT, "MSDOSFS FAT", "MSDOS FS FAT table");
108 MALLOC_JUSTDEFINE(M_MSDOSFSTMP, "MSDOSFS temp", "MSDOS FS temp. structures");
109 
110 static struct sysctllog *msdosfs_sysctl_log;
111 
112 extern const struct vnodeopv_desc msdosfs_vnodeop_opv_desc;
113 
114 const struct vnodeopv_desc * const msdosfs_vnodeopv_descs[] = {
115 	&msdosfs_vnodeop_opv_desc,
116 	NULL,
117 };
118 
119 struct vfsops msdosfs_vfsops = {
120 	.vfs_name = MOUNT_MSDOS,
121 	.vfs_min_mount_data = sizeof (struct msdosfs_args),
122 	.vfs_mount = msdosfs_mount,
123 	.vfs_start = msdosfs_start,
124 	.vfs_unmount = msdosfs_unmount,
125 	.vfs_root = msdosfs_root,
126 	.vfs_quotactl = (void *)eopnotsupp,
127 	.vfs_statvfs = msdosfs_statvfs,
128 	.vfs_sync = msdosfs_sync,
129 	.vfs_vget = msdosfs_vget,
130 	.vfs_loadvnode = msdosfs_loadvnode,
131 	.vfs_fhtovp = msdosfs_fhtovp,
132 	.vfs_vptofh = msdosfs_vptofh,
133 	.vfs_init = msdosfs_init,
134 	.vfs_reinit = msdosfs_reinit,
135 	.vfs_done = msdosfs_done,
136 	.vfs_mountroot = msdosfs_mountroot,
137 	.vfs_snapshot = (void *)eopnotsupp,
138 	.vfs_extattrctl = vfs_stdextattrctl,
139 	.vfs_suspendctl = msdosfs_suspendctl,
140 	.vfs_renamelock_enter = genfs_renamelock_enter,
141 	.vfs_renamelock_exit = genfs_renamelock_exit,
142 	.vfs_fsync = (void *)eopnotsupp,
143 	.vfs_opv_descs = msdosfs_vnodeopv_descs
144 };
145 
146 static int
147 msdos_modcmd(modcmd_t cmd, void *arg)
148 {
149 	int error;
150 
151 	switch (cmd) {
152 	case MODULE_CMD_INIT:
153 		error = vfs_attach(&msdosfs_vfsops);
154 		if (error != 0)
155 			break;
156 		sysctl_createv(&msdosfs_sysctl_log, 0, NULL, NULL,
157 			       CTLFLAG_PERMANENT,
158 			       CTLTYPE_NODE, "msdosfs",
159 			       SYSCTL_DESCR("MS-DOS file system"),
160 			       NULL, 0, NULL, 0,
161 			       CTL_VFS, 4, CTL_EOL);
162 		/*
163 		 * XXX the "4" above could be dynamic, thereby eliminating one
164 		 * more instance of the "number to vfs" mapping problem, but
165 		 * "4" is the order as taken from sys/mount.h
166 		 */
167 		break;
168 	case MODULE_CMD_FINI:
169 		error = vfs_detach(&msdosfs_vfsops);
170 		if (error != 0)
171 			break;
172 		sysctl_teardown(&msdosfs_sysctl_log);
173 		break;
174 	default:
175 		error = ENOTTY;
176 		break;
177 	}
178 
179 	return (error);
180 }
181 
182 static int
183 update_mp(struct mount *mp, struct msdosfs_args *argp)
184 {
185 	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
186 	int error;
187 
188 	pmp->pm_gid = argp->gid;
189 	pmp->pm_uid = argp->uid;
190 	pmp->pm_mask = argp->mask & ALLPERMS;
191 	pmp->pm_dirmask = argp->dirmask & ALLPERMS;
192 	pmp->pm_gmtoff = argp->gmtoff;
193 	pmp->pm_flags |= argp->flags & MSDOSFSMNT_MNTOPT;
194 
195 	/*
196 	 * GEMDOS knows nothing about win95 long filenames
197 	 */
198 	if (pmp->pm_flags & MSDOSFSMNT_GEMDOSFS)
199 		pmp->pm_flags |= MSDOSFSMNT_NOWIN95;
200 
201 	if (pmp->pm_flags & MSDOSFSMNT_NOWIN95)
202 		pmp->pm_flags |= MSDOSFSMNT_SHORTNAME;
203 	else if (!(pmp->pm_flags &
204 	    (MSDOSFSMNT_SHORTNAME | MSDOSFSMNT_LONGNAME))) {
205 		struct vnode *rtvp;
206 
207 		/*
208 		 * Try to divine whether to support Win'95 long filenames
209 		 */
210 		if (FAT32(pmp))
211 			pmp->pm_flags |= MSDOSFSMNT_LONGNAME;
212 		else {
213 			if ((error = msdosfs_root(mp, &rtvp)) != 0)
214 				return error;
215 			pmp->pm_flags |= findwin95(VTODE(rtvp))
216 				? MSDOSFSMNT_LONGNAME
217 					: MSDOSFSMNT_SHORTNAME;
218 			vput(rtvp);
219 		}
220 	}
221 
222 	mp->mnt_stat.f_namemax = MSDOSFS_NAMEMAX(pmp);
223 
224 	return 0;
225 }
226 
227 int
228 msdosfs_mountroot(void)
229 {
230 	struct mount *mp;
231 	struct lwp *l = curlwp;	/* XXX */
232 	int error;
233 	struct msdosfs_args args;
234 
235 	if (device_class(root_device) != DV_DISK)
236 		return (ENODEV);
237 
238 	if ((error = vfs_rootmountalloc(MOUNT_MSDOS, "root_device", &mp))) {
239 		vrele(rootvp);
240 		return (error);
241 	}
242 
243 	args.flags = MSDOSFSMNT_VERSIONED;
244 	args.uid = 0;
245 	args.gid = 0;
246 	args.mask = 0777;
247 	args.version = MSDOSFSMNT_VERSION;
248 	args.dirmask = 0777;
249 
250 	if ((error = msdosfs_mountfs(rootvp, mp, l, &args)) != 0) {
251 		vfs_unbusy(mp, false, NULL);
252 		vfs_destroy(mp);
253 		return (error);
254 	}
255 
256 	if ((error = update_mp(mp, &args)) != 0) {
257 		(void)msdosfs_unmount(mp, 0);
258 		vfs_unbusy(mp, false, NULL);
259 		vfs_destroy(mp);
260 		vrele(rootvp);
261 		return (error);
262 	}
263 
264 	mountlist_append(mp);
265 	(void)msdosfs_statvfs(mp, &mp->mnt_stat);
266 	vfs_unbusy(mp, false, NULL);
267 	return (0);
268 }
269 
270 /*
271  * mp - path - addr in user space of mount point (ie /usr or whatever)
272  * data - addr in user space of mount params including the name of the block
273  * special file to treat as a filesystem.
274  */
275 int
276 msdosfs_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
277 {
278 	struct lwp *l = curlwp;
279 	struct vnode *devvp;	  /* vnode for blk device to mount */
280 	struct msdosfs_args *args = data; /* holds data from mount request */
281 	/* msdosfs specific mount control block */
282 	struct msdosfsmount *pmp = NULL;
283 	int error, flags;
284 	mode_t accessmode;
285 
286 	if (args == NULL)
287 		return EINVAL;
288 	if (*data_len < sizeof *args)
289 		return EINVAL;
290 
291 	if (mp->mnt_flag & MNT_GETARGS) {
292 		pmp = VFSTOMSDOSFS(mp);
293 		if (pmp == NULL)
294 			return EIO;
295 		args->fspec = NULL;
296 		args->uid = pmp->pm_uid;
297 		args->gid = pmp->pm_gid;
298 		args->mask = pmp->pm_mask;
299 		args->flags = pmp->pm_flags;
300 		args->version = MSDOSFSMNT_VERSION;
301 		args->dirmask = pmp->pm_dirmask;
302 		args->gmtoff = pmp->pm_gmtoff;
303 		*data_len = sizeof *args;
304 		return 0;
305 	}
306 
307 	/*
308 	 * If not versioned (i.e. using old mount_msdos(8)), fill in
309 	 * the additional structure items with suitable defaults.
310 	 */
311 	if ((args->flags & MSDOSFSMNT_VERSIONED) == 0) {
312 		args->version = 1;
313 		args->dirmask = args->mask;
314 	}
315 
316 	/*
317 	 * Reset GMT offset for pre-v3 mount structure args.
318 	 */
319 	if (args->version < 3)
320 		args->gmtoff = 0;
321 
322 	/*
323 	 * If updating, check whether changing from read-only to
324 	 * read/write; if there is no device name, that's all we do.
325 	 */
326 	if (mp->mnt_flag & MNT_UPDATE) {
327 		pmp = VFSTOMSDOSFS(mp);
328 		error = 0;
329 		if (!(pmp->pm_flags & MSDOSFSMNT_RONLY) &&
330 		    (mp->mnt_flag & MNT_RDONLY)) {
331 			flags = WRITECLOSE;
332 			if (mp->mnt_flag & MNT_FORCE)
333 				flags |= FORCECLOSE;
334 			error = vflush(mp, NULLVP, flags);
335 		}
336 		if (!error && (mp->mnt_flag & MNT_RELOAD))
337 			/* not yet implemented */
338 			error = EOPNOTSUPP;
339 		if (error) {
340 			DPRINTF("vflush %d", error);
341 			return (error);
342 		}
343 		if ((pmp->pm_flags & MSDOSFSMNT_RONLY) &&
344 		    (mp->mnt_iflag & IMNT_WANTRDWR)) {
345 			/*
346 			 * If upgrade to read-write by non-root, then verify
347 			 * that user has necessary permissions on the device.
348 			 *
349 			 * Permission to update a mount is checked higher, so
350 			 * here we presume updating the mount is okay (for
351 			 * example, as far as securelevel goes) which leaves us
352 			 * with the normal check.
353 			 */
354 			devvp = pmp->pm_devvp;
355 			vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
356 			error = kauth_authorize_system(l->l_cred,
357 			    KAUTH_SYSTEM_MOUNT, KAUTH_REQ_SYSTEM_MOUNT_DEVICE,
358 			    mp, devvp, KAUTH_ARG(VREAD | VWRITE));
359 			VOP_UNLOCK(devvp);
360 			DPRINTF("KAUTH_REQ_SYSTEM_MOUNT_DEVICE %d", error);
361 			if (error)
362 				return (error);
363 
364 			pmp->pm_flags &= ~MSDOSFSMNT_RONLY;
365 		}
366 		if (args->fspec == NULL) {
367 			DPRINTF("missing fspec");
368 			return EINVAL;
369 		}
370 	}
371 	/*
372 	 * Not an update, or updating the name: look up the name
373 	 * and verify that it refers to a sensible block device.
374 	 */
375 	error = namei_simple_user(args->fspec,
376 				NSM_FOLLOW_NOEMULROOT, &devvp);
377 	if (error != 0) {
378 		DPRINTF("namei %d", error);
379 		return (error);
380 	}
381 
382 	if (devvp->v_type != VBLK) {
383 		DPRINTF("not block");
384 		vrele(devvp);
385 		return (ENOTBLK);
386 	}
387 	if (bdevsw_lookup(devvp->v_rdev) == NULL) {
388 		DPRINTF("no block switch");
389 		vrele(devvp);
390 		return (ENXIO);
391 	}
392 	/*
393 	 * If mount by non-root, then verify that user has necessary
394 	 * permissions on the device.
395 	 */
396 	accessmode = VREAD;
397 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
398 		accessmode |= VWRITE;
399 	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
400 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
401 	    KAUTH_REQ_SYSTEM_MOUNT_DEVICE, mp, devvp, KAUTH_ARG(accessmode));
402 	VOP_UNLOCK(devvp);
403 	if (error) {
404 		DPRINTF("KAUTH_REQ_SYSTEM_MOUNT_DEVICE %d", error);
405 		vrele(devvp);
406 		return (error);
407 	}
408 	if ((mp->mnt_flag & MNT_UPDATE) == 0) {
409 		int xflags;
410 
411 		if (mp->mnt_flag & MNT_RDONLY)
412 			xflags = FREAD;
413 		else
414 			xflags = FREAD|FWRITE;
415 		vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
416 		error = VOP_OPEN(devvp, xflags, FSCRED);
417 		VOP_UNLOCK(devvp);
418 		if (error) {
419 			DPRINTF("VOP_OPEN %d", error);
420 			goto fail;
421 		}
422 		error = msdosfs_mountfs(devvp, mp, l, args);
423 		if (error) {
424 			DPRINTF("msdosfs_mountfs %d", error);
425 			vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
426 			(void) VOP_CLOSE(devvp, xflags, NOCRED);
427 			VOP_UNLOCK(devvp);
428 			goto fail;
429 		}
430 #ifdef MSDOSFS_DEBUG		/* only needed for the printf below */
431 		pmp = VFSTOMSDOSFS(mp);
432 #endif
433 	} else {
434 		vrele(devvp);
435 		if (devvp != pmp->pm_devvp) {
436 			DPRINTF("devvp %p pmp %p", devvp, pmp->pm_devvp);
437 			return (EINVAL);	/* needs translation */
438 		}
439 	}
440 	if ((error = update_mp(mp, args)) != 0) {
441 		msdosfs_unmount(mp, MNT_FORCE);
442 		DPRINTF("update_mp %d", error);
443 		return error;
444 	}
445 
446 #ifdef MSDOSFS_DEBUG
447 	printf("msdosfs_mount(): mp %p, pmp %p, inusemap %p\n", mp, pmp, pmp->pm_inusemap);
448 #endif
449 	return set_statvfs_info(path, UIO_USERSPACE, args->fspec, UIO_USERSPACE,
450 	    mp->mnt_op->vfs_name, mp, l);
451 
452 fail:
453 	vrele(devvp);
454 	return (error);
455 }
456 
457 int
458 msdosfs_mountfs(struct vnode *devvp, struct mount *mp, struct lwp *l, struct msdosfs_args *argp)
459 {
460 	struct msdosfsmount *pmp;
461 	struct buf *bp;
462 	dev_t dev = devvp->v_rdev;
463 	union bootsector *bsp;
464 	struct byte_bpb33 *b33;
465 	struct byte_bpb50 *b50;
466 	struct byte_bpb710 *b710;
467 	uint8_t SecPerClust;
468 	int	ronly, error, BlkPerSec;
469 	uint64_t psize;
470 	unsigned secsize;
471 
472 	/* Flush out any old buffers remaining from a previous use. */
473 	if ((error = vinvalbuf(devvp, V_SAVE, l->l_cred, l, 0, 0)) != 0)
474 		return (error);
475 
476 	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
477 
478 	bp  = NULL; /* both used in error_exit */
479 	pmp = NULL;
480 
481 	error = fstrans_mount(mp);
482 	if (error)
483 		goto error_exit;
484 
485 	error = getdisksize(devvp, &psize, &secsize);
486 	if (error) {
487 		if (argp->flags & MSDOSFSMNT_GEMDOSFS)
488 			goto error_exit;
489 
490 		/* ok, so it failed.  we most likely don't need the info */
491 		secsize = DEV_BSIZE;
492 		psize = 0;
493 		error = 0;
494 	}
495 	if (secsize < DEV_BSIZE) {
496 		DPRINTF("Invalid block secsize (%d < DEV_BSIZE)", secsize);
497 		error = EINVAL;
498 		goto error_exit;
499 	}
500 
501 	if (argp->flags & MSDOSFSMNT_GEMDOSFS) {
502 		if (secsize != GEMDOSFS_BSIZE) {
503 			DPRINTF("Invalid block secsize %d for GEMDOS", secsize);
504 			error = EINVAL;
505 			goto error_exit;
506 		}
507 	}
508 
509 	/*
510 	 * Read the boot sector of the filesystem, and then check the
511 	 * boot signature.  If not a dos boot sector then error out.
512 	 */
513 	if ((error = bread(devvp, 0, secsize, NOCRED, 0, &bp)) != 0)
514 		goto error_exit;
515 	bsp = (union bootsector *)bp->b_data;
516 	b33 = (struct byte_bpb33 *)bsp->bs33.bsBPB;
517 	b50 = (struct byte_bpb50 *)bsp->bs50.bsBPB;
518 	b710 = (struct byte_bpb710 *)bsp->bs710.bsBPB;
519 
520 	if (!(argp->flags & MSDOSFSMNT_GEMDOSFS)) {
521 		if (bsp->bs50.bsBootSectSig0 != BOOTSIG0
522 		    || bsp->bs50.bsBootSectSig1 != BOOTSIG1) {
523 			DPRINTF("bootsig0 %d bootsig1 %d",
524 			    bsp->bs50.bsBootSectSig0,
525 			    bsp->bs50.bsBootSectSig1);
526 			error = EINVAL;
527 			goto error_exit;
528 		}
529 	}
530 
531 	pmp = malloc(sizeof(*pmp), M_MSDOSFSMNT, M_WAITOK|M_ZERO);
532 	pmp->pm_mountp = mp;
533 
534 	/*
535 	 * Compute several useful quantities from the bpb in the
536 	 * bootsector.  Copy in the dos 5 variant of the bpb then fix up
537 	 * the fields that are different between dos 5 and dos 3.3.
538 	 */
539 	SecPerClust = b50->bpbSecPerClust;
540 	pmp->pm_BytesPerSec = getushort(b50->bpbBytesPerSec);
541 	pmp->pm_ResSectors = getushort(b50->bpbResSectors);
542 	pmp->pm_FATs = b50->bpbFATs;
543 	pmp->pm_RootDirEnts = getushort(b50->bpbRootDirEnts);
544 	pmp->pm_Sectors = getushort(b50->bpbSectors);
545 	pmp->pm_FATsecs = getushort(b50->bpbFATsecs);
546 	pmp->pm_SecPerTrack = getushort(b50->bpbSecPerTrack);
547 	pmp->pm_Heads = getushort(b50->bpbHeads);
548 	pmp->pm_Media = b50->bpbMedia;
549 
550 	if (pmp->pm_Sectors == 0) {
551 		pmp->pm_HiddenSects = getulong(b50->bpbHiddenSecs);
552 		pmp->pm_HugeSectors = getulong(b50->bpbHugeSectors);
553 	} else {
554 		pmp->pm_HiddenSects = getushort(b33->bpbHiddenSecs);
555 		pmp->pm_HugeSectors = pmp->pm_Sectors;
556 	}
557 
558 	/*
559 	 * Sanity checks, from the FAT specification:
560 	 * - sectors per cluster: >= 1, power of 2
561 	 * - logical sector size: >= 1, power of 2
562 	 * - cluster size:        <= max FS block size
563 	 * - number of sectors:   >= 1
564 	 */
565 	if ((SecPerClust == 0) || !powerof2(SecPerClust) ||
566 	    (pmp->pm_BytesPerSec == 0) || !powerof2(pmp->pm_BytesPerSec) ||
567 	    (SecPerClust * pmp->pm_BytesPerSec > MAXBSIZE) ||
568 	    (pmp->pm_HugeSectors == 0)) {
569 		DPRINTF("consistency checks");
570 		error = EINVAL;
571 		goto error_exit;
572 	}
573 
574 	if (!(argp->flags & MSDOSFSMNT_GEMDOSFS) &&
575 	    (pmp->pm_SecPerTrack > 63)) {
576 		DPRINTF("SecPerTrack %d", pmp->pm_SecPerTrack);
577 		error = EINVAL;
578 		goto error_exit;
579 	}
580 
581 	if (pmp->pm_RootDirEnts == 0) {
582 		unsigned short FSVers = getushort(b710->bpbFSVers);
583 		unsigned short ExtFlags = getushort(b710->bpbExtFlags);
584 		/*
585 		 * Some say that bsBootSectSig[23] must be zero, but
586 		 * Windows does not require this and some digital cameras
587 		 * do not set these to zero.  Therefore, do not insist.
588 		 */
589 		if (pmp->pm_Sectors || pmp->pm_FATsecs || FSVers) {
590 			DPRINTF("Sectors %d FATsecs %lu FSVers %d",
591 			    pmp->pm_Sectors, pmp->pm_FATsecs, FSVers);
592 			error = EINVAL;
593 			goto error_exit;
594 		}
595 		pmp->pm_fatmask = FAT32_MASK;
596 		pmp->pm_fatmult = 4;
597 		pmp->pm_fatdiv = 1;
598 		pmp->pm_FATsecs = getulong(b710->bpbBigFATsecs);
599 
600 		/* Mirroring is enabled if the FATMIRROR bit is not set. */
601 		if ((ExtFlags & FATMIRROR) == 0)
602 			pmp->pm_flags |= MSDOSFS_FATMIRROR;
603 		else
604 			pmp->pm_curfat = ExtFlags & FATNUM;
605 	} else
606 		pmp->pm_flags |= MSDOSFS_FATMIRROR;
607 
608 	if (argp->flags & MSDOSFSMNT_GEMDOSFS) {
609 		if (FAT32(pmp)) {
610 			/* GEMDOS doesn't know FAT32. */
611 			DPRINTF("FAT32 for GEMDOS");
612 			error = EINVAL;
613 			goto error_exit;
614 		}
615 
616 		/*
617 		 * Check a few values (could do some more):
618 		 * - logical sector size: >= block size
619 		 * - number of sectors:   <= size of partition
620 		 */
621 		if ((pmp->pm_BytesPerSec < GEMDOSFS_BSIZE) ||
622 		    (pmp->pm_HugeSectors *
623 		     (pmp->pm_BytesPerSec / GEMDOSFS_BSIZE) > psize)) {
624 			DPRINTF("consistency checks for GEMDOS");
625 			error = EINVAL;
626 			goto error_exit;
627 		}
628 		/*
629 		 * XXX - Many parts of the msdosfs driver seem to assume that
630 		 * the number of bytes per logical sector (BytesPerSec) will
631 		 * always be the same as the number of bytes per disk block
632 		 * Let's pretend it is.
633 		 */
634 		BlkPerSec = pmp->pm_BytesPerSec / GEMDOSFS_BSIZE;
635 		pmp->pm_BytesPerSec  = GEMDOSFS_BSIZE;
636 		pmp->pm_HugeSectors *= BlkPerSec;
637 		pmp->pm_HiddenSects *= BlkPerSec;
638 		pmp->pm_ResSectors  *= BlkPerSec;
639 		pmp->pm_Sectors     *= BlkPerSec;
640 		pmp->pm_FATsecs     *= BlkPerSec;
641 		SecPerClust         *= BlkPerSec;
642 	}
643 
644 	/* Check that fs has nonzero FAT size */
645 	if (pmp->pm_FATsecs == 0) {
646 		DPRINTF("FATsecs is 0");
647 		error = EINVAL;
648 		goto error_exit;
649 	}
650 
651 	pmp->pm_fatblk = pmp->pm_ResSectors;
652 	if (FAT32(pmp)) {
653 		pmp->pm_rootdirblk = getulong(b710->bpbRootClust);
654 		pmp->pm_firstcluster = pmp->pm_fatblk
655 			+ (pmp->pm_FATs * pmp->pm_FATsecs);
656 		pmp->pm_fsinfo = getushort(b710->bpbFSInfo);
657 	} else {
658 		pmp->pm_rootdirblk = pmp->pm_fatblk +
659 			(pmp->pm_FATs * pmp->pm_FATsecs);
660 		pmp->pm_rootdirsize = (pmp->pm_RootDirEnts * sizeof(struct direntry)
661 				       + pmp->pm_BytesPerSec - 1)
662 			/ pmp->pm_BytesPerSec;/* in sectors */
663 		pmp->pm_firstcluster = pmp->pm_rootdirblk + pmp->pm_rootdirsize;
664 	}
665 
666 	pmp->pm_nmbrofclusters = (pmp->pm_HugeSectors - pmp->pm_firstcluster) /
667 	    SecPerClust;
668 	pmp->pm_maxcluster = pmp->pm_nmbrofclusters + 1;
669 	pmp->pm_fatsize = pmp->pm_FATsecs * pmp->pm_BytesPerSec;
670 
671 	if (argp->flags & MSDOSFSMNT_GEMDOSFS) {
672 		if (pmp->pm_nmbrofclusters <= (0xff0 - 2)) {
673 			pmp->pm_fatmask = FAT12_MASK;
674 			pmp->pm_fatmult = 3;
675 			pmp->pm_fatdiv = 2;
676 		} else {
677 			pmp->pm_fatmask = FAT16_MASK;
678 			pmp->pm_fatmult = 2;
679 			pmp->pm_fatdiv = 1;
680 		}
681 	} else if (pmp->pm_fatmask == 0) {
682 		if (pmp->pm_maxcluster
683 		    <= ((CLUST_RSRVD - CLUST_FIRST) & FAT12_MASK)) {
684 			/*
685 			 * This will usually be a floppy disk. This size makes
686 			 * sure that one FAT entry will not be split across
687 			 * multiple blocks.
688 			 */
689 			pmp->pm_fatmask = FAT12_MASK;
690 			pmp->pm_fatmult = 3;
691 			pmp->pm_fatdiv = 2;
692 		} else {
693 			pmp->pm_fatmask = FAT16_MASK;
694 			pmp->pm_fatmult = 2;
695 			pmp->pm_fatdiv = 1;
696 		}
697 	}
698 	if (FAT12(pmp))
699 		pmp->pm_fatblocksize = 3 * pmp->pm_BytesPerSec;
700 	else
701 		pmp->pm_fatblocksize = MAXBSIZE;
702 
703 	pmp->pm_fatblocksec = pmp->pm_fatblocksize / pmp->pm_BytesPerSec;
704 	pmp->pm_bnshift = ffs(pmp->pm_BytesPerSec) - 1;
705 
706 	/*
707 	 * Compute mask and shift value for isolating cluster relative byte
708 	 * offsets and cluster numbers from a file offset.
709 	 */
710 	pmp->pm_bpcluster = SecPerClust * pmp->pm_BytesPerSec;
711 	pmp->pm_crbomask = pmp->pm_bpcluster - 1;
712 	pmp->pm_cnshift = ffs(pmp->pm_bpcluster) - 1;
713 
714 	/*
715 	 * Check for valid cluster size
716 	 * must be a power of 2
717 	 */
718 	if (pmp->pm_bpcluster ^ (1 << pmp->pm_cnshift)) {
719 		DPRINTF("bpcluster %lu cnshift %lu", pmp->pm_bpcluster,
720 		    pmp->pm_cnshift);
721 		error = EINVAL;
722 		goto error_exit;
723 	}
724 
725 	/*
726 	 * Cluster size must be within limit of MAXBSIZE.
727 	 * Many FAT filesystems will not have clusters larger than
728 	 * 32KiB due to limits in Windows versions before Vista.
729 	 */
730 	if (pmp->pm_bpcluster > MAXBSIZE) {
731 		DPRINTF("bpcluster %lu > MAXBSIZE %d",
732 		    pmp->pm_bpcluster, MAXBSIZE);
733 		error = EINVAL;
734 		goto error_exit;
735 	}
736 
737 	/*
738 	 * Release the bootsector buffer.
739 	 */
740 	brelse(bp, BC_AGE);
741 	bp = NULL;
742 
743 	/*
744 	 * Check FSInfo.
745 	 */
746 	if (pmp->pm_fsinfo) {
747 		struct fsinfo *fp;
748 		const int rdsz = roundup(sizeof(*fp), pmp->pm_BytesPerSec);
749 
750 		/*
751 		 * XXX	If the fsinfo block is stored on media with
752 		 *	2KB or larger sectors, is the fsinfo structure
753 		 *	padded at the end or in the middle?
754 		 */
755 		if ((error = bread(devvp, de_bn2kb(pmp, pmp->pm_fsinfo),
756 		    rdsz, NOCRED, 0, &bp)) != 0)
757 			goto error_exit;
758 		fp = (struct fsinfo *)bp->b_data;
759 		if (!memcmp(fp->fsisig1, "RRaA", 4)
760 		    && !memcmp(fp->fsisig2, "rrAa", 4)
761 		    && !memcmp(fp->fsisig3, "\0\0\125\252", 4)
762 		    && !memcmp(fp->fsisig4, "\0\0\125\252", 4))
763 			pmp->pm_nxtfree = getulong(fp->fsinxtfree);
764 		else
765 			pmp->pm_fsinfo = 0;
766 		brelse(bp, 0);
767 		bp = NULL;
768 	}
769 
770 	/*
771 	 * Check and validate (or perhaps invalidate?) the fsinfo structure?
772 	 * XXX
773 	 */
774 	if (pmp->pm_fsinfo) {
775 		if ((pmp->pm_nxtfree == 0xffffffffUL) ||
776 		    (pmp->pm_nxtfree > pmp->pm_maxcluster))
777 			pmp->pm_fsinfo = 0;
778 	}
779 
780 	/*
781 	 * Allocate memory for the bitmap of allocated clusters, and then
782 	 * fill it in.
783 	 */
784 	pmp->pm_inusemap = malloc(((pmp->pm_maxcluster + N_INUSEBITS)
785 				   / N_INUSEBITS)
786 				  * sizeof(*pmp->pm_inusemap),
787 				  M_MSDOSFSFAT, M_WAITOK);
788 
789 	/*
790 	 * fillinusemap() needs pm_devvp.
791 	 */
792 	pmp->pm_dev = dev;
793 	pmp->pm_devvp = devvp;
794 
795 	/*
796 	 * Have the inuse map filled in.
797 	 */
798 	if ((error = fillinusemap(pmp)) != 0) {
799 		DPRINTF("fillinusemap %d", error);
800 		goto error_exit;
801 	}
802 
803 	/*
804 	 * If they want FAT updates to be synchronous then let them suffer
805 	 * the performance degradation in exchange for the on disk copy of
806 	 * the FAT being correct just about all the time.  I suppose this
807 	 * would be a good thing to turn on if the kernel is still flakey.
808 	 */
809 	if (mp->mnt_flag & MNT_SYNCHRONOUS)
810 		pmp->pm_flags |= MSDOSFSMNT_WAITONFAT;
811 
812 	/*
813 	 * Finish up.
814 	 */
815 	if (ronly)
816 		pmp->pm_flags |= MSDOSFSMNT_RONLY;
817 	else
818 		pmp->pm_fmod = 1;
819 	mp->mnt_data = pmp;
820 	mp->mnt_stat.f_fsidx.__fsid_val[0] = (long)dev;
821 	mp->mnt_stat.f_fsidx.__fsid_val[1] = makefstype(MOUNT_MSDOS);
822 	mp->mnt_stat.f_fsid = mp->mnt_stat.f_fsidx.__fsid_val[0];
823 	mp->mnt_stat.f_namemax = MSDOSFS_NAMEMAX(pmp);
824 	mp->mnt_flag |= MNT_LOCAL;
825 	mp->mnt_dev_bshift = pmp->pm_bnshift;
826 	mp->mnt_fs_bshift = pmp->pm_cnshift;
827 
828 	/*
829 	 * If we ever do quotas for DOS filesystems this would be a place
830 	 * to fill in the info in the msdosfsmount structure. You dolt,
831 	 * quotas on dos filesystems make no sense because files have no
832 	 * owners on dos filesystems. of course there is some empty space
833 	 * in the directory entry where we could put uid's and gid's.
834 	 */
835 
836 	spec_node_setmountedfs(devvp, mp);
837 
838 	return (0);
839 
840 error_exit:
841 	fstrans_unmount(mp);
842 	if (bp)
843 		brelse(bp, BC_AGE);
844 	if (pmp) {
845 		if (pmp->pm_inusemap)
846 			free(pmp->pm_inusemap, M_MSDOSFSFAT);
847 		free(pmp, M_MSDOSFSMNT);
848 		mp->mnt_data = NULL;
849 	}
850 	return (error);
851 }
852 
853 int
854 msdosfs_start(struct mount *mp, int flags)
855 {
856 
857 	return (0);
858 }
859 
860 /*
861  * Unmount the filesystem described by mp.
862  */
863 int
864 msdosfs_unmount(struct mount *mp, int mntflags)
865 {
866 	struct msdosfsmount *pmp;
867 	int error, flags;
868 
869 	flags = 0;
870 	if (mntflags & MNT_FORCE)
871 		flags |= FORCECLOSE;
872 	if ((error = vflush(mp, NULLVP, flags)) != 0)
873 		return (error);
874 	pmp = VFSTOMSDOSFS(mp);
875 	if (pmp->pm_devvp->v_type != VBAD)
876 		spec_node_setmountedfs(pmp->pm_devvp, NULL);
877 #ifdef MSDOSFS_DEBUG
878 	{
879 		struct vnode *vp = pmp->pm_devvp;
880 
881 		printf("msdosfs_umount(): just before calling VOP_CLOSE()\n");
882 		printf("flag %08x, usecount %d, writecount %d, holdcnt %d\n",
883 		    vp->v_vflag | vp->v_iflag | vp->v_uflag, vp->v_usecount,
884 		    vp->v_writecount, vp->v_holdcnt);
885 		printf("mount %p, op %p\n",
886 		    vp->v_mount, vp->v_op);
887 		printf("freef %p, freeb %p, mount %p\n",
888 		    vp->v_freelist.tqe_next, vp->v_freelist.tqe_prev,
889 		    vp->v_mount);
890 		printf("cleanblkhd %p, dirtyblkhd %p, numoutput %d, type %d\n",
891 		    vp->v_cleanblkhd.lh_first,
892 		    vp->v_dirtyblkhd.lh_first,
893 		    vp->v_numoutput, vp->v_type);
894 		printf("union %p, tag %d, data[0] %08x, data[1] %08x\n",
895 		    vp->v_socket, vp->v_tag,
896 		    ((u_int *)vp->v_data)[0],
897 		    ((u_int *)vp->v_data)[1]);
898 	}
899 #endif
900 	vn_lock(pmp->pm_devvp, LK_EXCLUSIVE | LK_RETRY);
901 	(void) VOP_CLOSE(pmp->pm_devvp,
902 	    pmp->pm_flags & MSDOSFSMNT_RONLY ? FREAD : FREAD|FWRITE, NOCRED);
903 	vput(pmp->pm_devvp);
904 	msdosfs_fh_destroy(pmp);
905 	free(pmp->pm_inusemap, M_MSDOSFSFAT);
906 	free(pmp, M_MSDOSFSMNT);
907 	mp->mnt_data = NULL;
908 	mp->mnt_flag &= ~MNT_LOCAL;
909 	fstrans_unmount(mp);
910 	return (0);
911 }
912 
913 int
914 msdosfs_root(struct mount *mp, struct vnode **vpp)
915 {
916 	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
917 	int error;
918 
919 #ifdef MSDOSFS_DEBUG
920 	printf("msdosfs_root(); mp %p, pmp %p\n", mp, pmp);
921 #endif
922 	if ((error = deget(pmp, MSDOSFSROOT, MSDOSFSROOT_OFS, vpp)) != 0)
923 		return error;
924 	error = vn_lock(*vpp, LK_EXCLUSIVE);
925 	if (error) {
926 		vrele(*vpp);
927 		*vpp = NULL;
928 		return error;
929 	}
930 	return 0;
931 }
932 
933 int
934 msdosfs_statvfs(struct mount *mp, struct statvfs *sbp)
935 {
936 	struct msdosfsmount *pmp;
937 
938 	pmp = VFSTOMSDOSFS(mp);
939 	sbp->f_bsize = pmp->pm_bpcluster;
940 	sbp->f_frsize = sbp->f_bsize;
941 	sbp->f_iosize = pmp->pm_bpcluster;
942 	sbp->f_blocks = pmp->pm_nmbrofclusters;
943 	sbp->f_bfree = pmp->pm_freeclustercount;
944 	sbp->f_bavail = pmp->pm_freeclustercount;
945 	sbp->f_bresvd = 0;
946 	sbp->f_files = pmp->pm_RootDirEnts;			/* XXX */
947 	sbp->f_ffree = 0;	/* what to put in here? */
948 	sbp->f_favail = 0;	/* what to put in here? */
949 	sbp->f_fresvd = 0;
950 	copy_statvfs_info(sbp, mp);
951 	return (0);
952 }
953 
954 struct msdosfs_sync_ctx {
955 	int waitfor;
956 };
957 
958 static bool
959 msdosfs_sync_selector(void *cl, struct vnode *vp)
960 {
961 	struct msdosfs_sync_ctx *c = cl;
962 	struct denode *dep;
963 
964 	dep = VTODE(vp);
965 	if (c->waitfor == MNT_LAZY || vp->v_type == VNON ||
966 	    dep == NULL || (((dep->de_flag &
967 	    (DE_ACCESS | DE_CREATE | DE_UPDATE | DE_MODIFIED)) == 0) &&
968 	     (LIST_EMPTY(&vp->v_dirtyblkhd) &&
969 	      UVM_OBJ_IS_CLEAN(&vp->v_uobj))))
970 		return false;
971 	return true;
972 }
973 
974 int
975 msdosfs_sync(struct mount *mp, int waitfor, kauth_cred_t cred)
976 {
977 	struct vnode *vp;
978 	struct vnode_iterator *marker;
979 	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
980 	int error, allerror = 0;
981 	struct msdosfs_sync_ctx ctx;
982 
983 	/*
984 	 * If we ever switch to not updating all of the FATs all the time,
985 	 * this would be the place to update them from the first one.
986 	 */
987 	if (pmp->pm_fmod != 0) {
988 		if (pmp->pm_flags & MSDOSFSMNT_RONLY)
989 			panic("msdosfs_sync: rofs mod");
990 		else {
991 			/* update FATs here */
992 		}
993 	}
994 	fstrans_start(mp, FSTRANS_SHARED);
995 	/*
996 	 * Write back each (modified) denode.
997 	 */
998 	vfs_vnode_iterator_init(mp, &marker);
999 	ctx.waitfor = waitfor;
1000 	while ((vp = vfs_vnode_iterator_next(marker, msdosfs_sync_selector,
1001 	    &ctx)))
1002 	{
1003 		error = vn_lock(vp, LK_EXCLUSIVE);
1004 		if (error) {
1005 			vrele(vp);
1006 			continue;
1007 		}
1008 		if ((error = VOP_FSYNC(vp, cred,
1009 		    waitfor == MNT_WAIT ? FSYNC_WAIT : 0, 0, 0)) != 0)
1010 			allerror = error;
1011 		vput(vp);
1012 	}
1013 	vfs_vnode_iterator_destroy(marker);
1014 
1015 	/*
1016 	 * Force stale file system control information to be flushed.
1017 	 */
1018 	if ((error = VOP_FSYNC(pmp->pm_devvp, cred,
1019 	    waitfor == MNT_WAIT ? FSYNC_WAIT : 0, 0, 0)) != 0)
1020 		allerror = error;
1021 	fstrans_done(mp);
1022 	return (allerror);
1023 }
1024 
1025 int
1026 msdosfs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
1027 {
1028 	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
1029 	struct defid defh;
1030 	uint32_t gen;
1031 	int error;
1032 
1033 	if (fhp->fid_len != sizeof(struct defid)) {
1034 		DPRINTF("fid_len %d %zd", fhp->fid_len, sizeof(struct defid));
1035 		return EINVAL;
1036 	}
1037 	memcpy(&defh, fhp, sizeof(defh));
1038 	error = msdosfs_fh_lookup(pmp, defh.defid_dirclust, defh.defid_dirofs,
1039 	    &gen);
1040 	if (error == 0 && gen != defh.defid_gen)
1041 		error = ESTALE;
1042 	if (error) {
1043 		*vpp = NULLVP;
1044 		return error;
1045 	}
1046 	error = deget(pmp, defh.defid_dirclust, defh.defid_dirofs, vpp);
1047 	if (error) {
1048 		DPRINTF("deget %d", error);
1049 		*vpp = NULLVP;
1050 		return error;
1051 	}
1052 	error = vn_lock(*vpp, LK_EXCLUSIVE);
1053 	if (error) {
1054 		vrele(*vpp);
1055 		*vpp = NULLVP;
1056 		return error;
1057 	}
1058 	return 0;
1059 }
1060 
1061 int
1062 msdosfs_vptofh(struct vnode *vp, struct fid *fhp, size_t *fh_size)
1063 {
1064 	struct msdosfsmount *pmp = VFSTOMSDOSFS(vp->v_mount);
1065 	struct denode *dep;
1066 	struct defid defh;
1067 	int error;
1068 
1069 	if (*fh_size < sizeof(struct defid)) {
1070 		*fh_size = sizeof(struct defid);
1071 		return E2BIG;
1072 	}
1073 	*fh_size = sizeof(struct defid);
1074 	dep = VTODE(vp);
1075 	memset(&defh, 0, sizeof(defh));
1076 	defh.defid_len = sizeof(struct defid);
1077 	defh.defid_dirclust = dep->de_dirclust;
1078 	defh.defid_dirofs = dep->de_diroffset;
1079 	error = msdosfs_fh_enter(pmp, dep->de_dirclust, dep->de_diroffset,
1080 	     &defh.defid_gen);
1081 	if (error == 0)
1082 		memcpy(fhp, &defh, sizeof(defh));
1083 	return error;
1084 }
1085 
1086 int
1087 msdosfs_vget(struct mount *mp, ino_t ino,
1088     struct vnode **vpp)
1089 {
1090 
1091 	return (EOPNOTSUPP);
1092 }
1093 
1094 int
1095 msdosfs_suspendctl(struct mount *mp, int cmd)
1096 {
1097 	int error;
1098 	struct lwp *l = curlwp;
1099 
1100 	switch (cmd) {
1101 	case SUSPEND_SUSPEND:
1102 		if ((error = fstrans_setstate(mp, FSTRANS_SUSPENDING)) != 0)
1103 			return error;
1104 		error = msdosfs_sync(mp, MNT_WAIT, l->l_proc->p_cred);
1105 		if (error == 0)
1106 			error = fstrans_setstate(mp, FSTRANS_SUSPENDED);
1107 		if (error != 0) {
1108 			(void) fstrans_setstate(mp, FSTRANS_NORMAL);
1109 			return error;
1110 		}
1111 		return 0;
1112 
1113 	case SUSPEND_RESUME:
1114 		return fstrans_setstate(mp, FSTRANS_NORMAL);
1115 
1116 	default:
1117 		return EINVAL;
1118 	}
1119 }
1120