xref: /netbsd-src/sys/fs/msdosfs/msdosfs_vfsops.c (revision a0698ed9d41653d7a2378819ad501a285ca0d401)
1 /*	$NetBSD: msdosfs_vfsops.c,v 1.130 2018/09/03 16:29:34 riastradh 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 1.130 2018/09/03 16:29:34 riastradh 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/ioctl.h>
73 #include <sys/malloc.h>
74 #include <sys/dirent.h>
75 #include <sys/stat.h>
76 #include <sys/conf.h>
77 #include <sys/kauth.h>
78 #include <sys/module.h>
79 
80 #include <fs/msdosfs/bpb.h>
81 #include <fs/msdosfs/bootsect.h>
82 #include <fs/msdosfs/direntry.h>
83 #include <fs/msdosfs/denode.h>
84 #include <fs/msdosfs/msdosfsmount.h>
85 #include <fs/msdosfs/fat.h>
86 
87 MODULE(MODULE_CLASS_VFS, msdos, NULL);
88 
89 #ifdef MSDOSFS_DEBUG
90 #define DPRINTF(fmt, ...) uprintf("%s(): " fmt "\n", __func__, ##__VA_ARGS__)
91 #else
92 #define DPRINTF(fmt, ...)
93 #endif
94 
95 #define GEMDOSFS_BSIZE	512
96 
97 #define MSDOSFS_NAMEMAX(pmp) \
98 	(pmp)->pm_flags & MSDOSFSMNT_LONGNAME ? WIN_MAXLEN : 12
99 
100 int msdosfs_mountfs(struct vnode *, struct mount *, struct lwp *,
101     struct msdosfs_args *);
102 
103 static int update_mp(struct mount *, struct msdosfs_args *);
104 
105 MALLOC_JUSTDEFINE(M_MSDOSFSMNT, "MSDOSFS mount", "MSDOS FS mount structure");
106 MALLOC_JUSTDEFINE(M_MSDOSFSFAT, "MSDOSFS FAT", "MSDOS FS FAT table");
107 MALLOC_JUSTDEFINE(M_MSDOSFSTMP, "MSDOSFS temp", "MSDOS FS temp. structures");
108 
109 static struct sysctllog *msdosfs_sysctl_log;
110 
111 extern const struct vnodeopv_desc msdosfs_vnodeop_opv_desc;
112 
113 const struct vnodeopv_desc * const msdosfs_vnodeopv_descs[] = {
114 	&msdosfs_vnodeop_opv_desc,
115 	NULL,
116 };
117 
118 struct vfsops msdosfs_vfsops = {
119 	.vfs_name = MOUNT_MSDOS,
120 	.vfs_min_mount_data = sizeof (struct msdosfs_args),
121 	.vfs_mount = msdosfs_mount,
122 	.vfs_start = msdosfs_start,
123 	.vfs_unmount = msdosfs_unmount,
124 	.vfs_root = msdosfs_root,
125 	.vfs_quotactl = (void *)eopnotsupp,
126 	.vfs_statvfs = msdosfs_statvfs,
127 	.vfs_sync = msdosfs_sync,
128 	.vfs_vget = msdosfs_vget,
129 	.vfs_loadvnode = msdosfs_loadvnode,
130 	.vfs_fhtovp = msdosfs_fhtovp,
131 	.vfs_vptofh = msdosfs_vptofh,
132 	.vfs_init = msdosfs_init,
133 	.vfs_reinit = msdosfs_reinit,
134 	.vfs_done = msdosfs_done,
135 	.vfs_mountroot = msdosfs_mountroot,
136 	.vfs_snapshot = (void *)eopnotsupp,
137 	.vfs_extattrctl = vfs_stdextattrctl,
138 	.vfs_suspendctl = genfs_suspendctl,
139 	.vfs_renamelock_enter = genfs_renamelock_enter,
140 	.vfs_renamelock_exit = genfs_renamelock_exit,
141 	.vfs_fsync = (void *)eopnotsupp,
142 	.vfs_opv_descs = msdosfs_vnodeopv_descs
143 };
144 
145 static int
146 msdos_modcmd(modcmd_t cmd, void *arg)
147 {
148 	int error;
149 
150 	switch (cmd) {
151 	case MODULE_CMD_INIT:
152 		error = vfs_attach(&msdosfs_vfsops);
153 		if (error != 0)
154 			break;
155 		sysctl_createv(&msdosfs_sysctl_log, 0, NULL, NULL,
156 			       CTLFLAG_PERMANENT,
157 			       CTLTYPE_NODE, "msdosfs",
158 			       SYSCTL_DESCR("MS-DOS file system"),
159 			       NULL, 0, NULL, 0,
160 			       CTL_VFS, 4, CTL_EOL);
161 		/*
162 		 * XXX the "4" above could be dynamic, thereby eliminating one
163 		 * more instance of the "number to vfs" mapping problem, but
164 		 * "4" is the order as taken from sys/mount.h
165 		 */
166 		break;
167 	case MODULE_CMD_FINI:
168 		error = vfs_detach(&msdosfs_vfsops);
169 		if (error != 0)
170 			break;
171 		sysctl_teardown(&msdosfs_sysctl_log);
172 		break;
173 	default:
174 		error = ENOTTY;
175 		break;
176 	}
177 
178 	return (error);
179 }
180 
181 static int
182 update_mp(struct mount *mp, struct msdosfs_args *argp)
183 {
184 	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
185 	int error;
186 
187 	pmp->pm_gid = argp->gid;
188 	pmp->pm_uid = argp->uid;
189 	pmp->pm_mask = argp->mask & ALLPERMS;
190 	pmp->pm_dirmask = argp->dirmask & ALLPERMS;
191 	pmp->pm_gmtoff = argp->gmtoff;
192 	pmp->pm_flags |= argp->flags & MSDOSFSMNT_MNTOPT;
193 
194 	/*
195 	 * GEMDOS knows nothing about win95 long filenames
196 	 */
197 	if (pmp->pm_flags & MSDOSFSMNT_GEMDOSFS)
198 		pmp->pm_flags |= MSDOSFSMNT_NOWIN95;
199 
200 	if (pmp->pm_flags & MSDOSFSMNT_NOWIN95)
201 		pmp->pm_flags |= MSDOSFSMNT_SHORTNAME;
202 	else if (!(pmp->pm_flags &
203 	    (MSDOSFSMNT_SHORTNAME | MSDOSFSMNT_LONGNAME))) {
204 		struct vnode *rtvp;
205 
206 		/*
207 		 * Try to divine whether to support Win'95 long filenames
208 		 */
209 		if (FAT32(pmp))
210 			pmp->pm_flags |= MSDOSFSMNT_LONGNAME;
211 		else {
212 			if ((error = msdosfs_root(mp, &rtvp)) != 0)
213 				return error;
214 			pmp->pm_flags |= findwin95(VTODE(rtvp))
215 				? MSDOSFSMNT_LONGNAME
216 					: MSDOSFSMNT_SHORTNAME;
217 			vput(rtvp);
218 		}
219 	}
220 
221 	mp->mnt_stat.f_namemax = MSDOSFS_NAMEMAX(pmp);
222 
223 	return 0;
224 }
225 
226 int
227 msdosfs_mountroot(void)
228 {
229 	struct mount *mp;
230 	struct lwp *l = curlwp;	/* XXX */
231 	int error;
232 	struct msdosfs_args args;
233 
234 	if (device_class(root_device) != DV_DISK)
235 		return (ENODEV);
236 
237 	if ((error = vfs_rootmountalloc(MOUNT_MSDOS, "root_device", &mp))) {
238 		vrele(rootvp);
239 		return (error);
240 	}
241 
242 	args.flags = MSDOSFSMNT_VERSIONED;
243 	args.uid = 0;
244 	args.gid = 0;
245 	args.mask = 0777;
246 	args.version = MSDOSFSMNT_VERSION;
247 	args.dirmask = 0777;
248 
249 	if ((error = msdosfs_mountfs(rootvp, mp, l, &args)) != 0) {
250 		vfs_unbusy(mp);
251 		vfs_rele(mp);
252 		return (error);
253 	}
254 
255 	if ((error = update_mp(mp, &args)) != 0) {
256 		(void)msdosfs_unmount(mp, 0);
257 		vfs_unbusy(mp);
258 		vfs_rele(mp);
259 		vrele(rootvp);
260 		return (error);
261 	}
262 
263 	mountlist_append(mp);
264 	(void)msdosfs_statvfs(mp, &mp->mnt_stat);
265 	vfs_unbusy(mp);
266 	return (0);
267 }
268 
269 /*
270  * mp - path - addr in user space of mount point (ie /usr or whatever)
271  * data - addr in user space of mount params including the name of the block
272  * special file to treat as a filesystem.
273  */
274 int
275 msdosfs_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
276 {
277 	struct lwp *l = curlwp;
278 	struct vnode *devvp;	  /* vnode for blk device to mount */
279 	struct msdosfs_args *args = data; /* holds data from mount request */
280 	/* msdosfs specific mount control block */
281 	struct msdosfsmount *pmp = NULL;
282 	int error, flags;
283 	mode_t accessmode;
284 
285 	if (args == NULL)
286 		return EINVAL;
287 	if (*data_len < sizeof *args)
288 		return EINVAL;
289 
290 	if (mp->mnt_flag & MNT_GETARGS) {
291 		pmp = VFSTOMSDOSFS(mp);
292 		if (pmp == NULL)
293 			return EIO;
294 		args->fspec = NULL;
295 		args->uid = pmp->pm_uid;
296 		args->gid = pmp->pm_gid;
297 		args->mask = pmp->pm_mask;
298 		args->flags = pmp->pm_flags;
299 		args->version = MSDOSFSMNT_VERSION;
300 		args->dirmask = pmp->pm_dirmask;
301 		args->gmtoff = pmp->pm_gmtoff;
302 		*data_len = sizeof *args;
303 		return 0;
304 	}
305 
306 	/*
307 	 * If not versioned (i.e. using old mount_msdos(8)), fill in
308 	 * the additional structure items with suitable defaults.
309 	 */
310 	if ((args->flags & MSDOSFSMNT_VERSIONED) == 0) {
311 		args->version = 1;
312 		args->dirmask = args->mask;
313 	}
314 
315 	/*
316 	 * Reset GMT offset for pre-v3 mount structure args.
317 	 */
318 	if (args->version < 3)
319 		args->gmtoff = 0;
320 
321 	/*
322 	 * If updating, check whether changing from read-only to
323 	 * read/write; if there is no device name, that's all we do.
324 	 */
325 	if (mp->mnt_flag & MNT_UPDATE) {
326 		pmp = VFSTOMSDOSFS(mp);
327 		error = 0;
328 		if (!(pmp->pm_flags & MSDOSFSMNT_RONLY) &&
329 		    (mp->mnt_flag & MNT_RDONLY)) {
330 			flags = WRITECLOSE;
331 			if (mp->mnt_flag & MNT_FORCE)
332 				flags |= FORCECLOSE;
333 			error = vflush(mp, NULLVP, flags);
334 		}
335 		if (!error && (mp->mnt_flag & MNT_RELOAD))
336 			/* not yet implemented */
337 			error = EOPNOTSUPP;
338 		if (error) {
339 			DPRINTF("vflush %d", error);
340 			return (error);
341 		}
342 		if ((pmp->pm_flags & MSDOSFSMNT_RONLY) &&
343 		    (mp->mnt_iflag & IMNT_WANTRDWR)) {
344 			/*
345 			 * If upgrade to read-write by non-root, then verify
346 			 * that user has necessary permissions on the device.
347 			 *
348 			 * Permission to update a mount is checked higher, so
349 			 * here we presume updating the mount is okay (for
350 			 * example, as far as securelevel goes) which leaves us
351 			 * with the normal check.
352 			 */
353 			devvp = pmp->pm_devvp;
354 			vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
355 			error = kauth_authorize_system(l->l_cred,
356 			    KAUTH_SYSTEM_MOUNT, KAUTH_REQ_SYSTEM_MOUNT_DEVICE,
357 			    mp, devvp, KAUTH_ARG(VREAD | VWRITE));
358 			VOP_UNLOCK(devvp);
359 			DPRINTF("KAUTH_REQ_SYSTEM_MOUNT_DEVICE %d", error);
360 			if (error)
361 				return (error);
362 
363 			pmp->pm_flags &= ~MSDOSFSMNT_RONLY;
364 		}
365 		if (args->fspec == NULL) {
366 			DPRINTF("missing fspec");
367 			return EINVAL;
368 		}
369 	}
370 	/*
371 	 * Not an update, or updating the name: look up the name
372 	 * and verify that it refers to a sensible block device.
373 	 */
374 	error = namei_simple_user(args->fspec,
375 				NSM_FOLLOW_NOEMULROOT, &devvp);
376 	if (error != 0) {
377 		DPRINTF("namei %d", error);
378 		return (error);
379 	}
380 
381 	if (devvp->v_type != VBLK) {
382 		DPRINTF("not block");
383 		vrele(devvp);
384 		return (ENOTBLK);
385 	}
386 	if (bdevsw_lookup(devvp->v_rdev) == NULL) {
387 		DPRINTF("no block switch");
388 		vrele(devvp);
389 		return (ENXIO);
390 	}
391 	/*
392 	 * If mount by non-root, then verify that user has necessary
393 	 * permissions on the device.
394 	 */
395 	accessmode = VREAD;
396 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
397 		accessmode |= VWRITE;
398 	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
399 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
400 	    KAUTH_REQ_SYSTEM_MOUNT_DEVICE, mp, devvp, KAUTH_ARG(accessmode));
401 	VOP_UNLOCK(devvp);
402 	if (error) {
403 		DPRINTF("KAUTH_REQ_SYSTEM_MOUNT_DEVICE %d", error);
404 		vrele(devvp);
405 		return (error);
406 	}
407 	if ((mp->mnt_flag & MNT_UPDATE) == 0) {
408 		int xflags;
409 
410 		if (mp->mnt_flag & MNT_RDONLY)
411 			xflags = FREAD;
412 		else
413 			xflags = FREAD|FWRITE;
414 		vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
415 		error = VOP_OPEN(devvp, xflags, FSCRED);
416 		VOP_UNLOCK(devvp);
417 		if (error) {
418 			DPRINTF("VOP_OPEN %d", error);
419 			goto fail;
420 		}
421 		error = msdosfs_mountfs(devvp, mp, l, args);
422 		if (error) {
423 			DPRINTF("msdosfs_mountfs %d", error);
424 			vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
425 			(void) VOP_CLOSE(devvp, xflags, NOCRED);
426 			VOP_UNLOCK(devvp);
427 			goto fail;
428 		}
429 #ifdef MSDOSFS_DEBUG		/* only needed for the printf below */
430 		pmp = VFSTOMSDOSFS(mp);
431 #endif
432 	} else {
433 		vrele(devvp);
434 		if (devvp != pmp->pm_devvp) {
435 			DPRINTF("devvp %p pmp %p", devvp, pmp->pm_devvp);
436 			return (EINVAL);	/* needs translation */
437 		}
438 	}
439 	if ((error = update_mp(mp, args)) != 0) {
440 		msdosfs_unmount(mp, MNT_FORCE);
441 		DPRINTF("update_mp %d", error);
442 		return error;
443 	}
444 
445 #ifdef MSDOSFS_DEBUG
446 	printf("msdosfs_mount(): mp %p, pmp %p, inusemap %p\n", mp, pmp, pmp->pm_inusemap);
447 #endif
448 	return set_statvfs_info(path, UIO_USERSPACE, args->fspec, UIO_USERSPACE,
449 	    mp->mnt_op->vfs_name, mp, l);
450 
451 fail:
452 	vrele(devvp);
453 	return (error);
454 }
455 
456 int
457 msdosfs_mountfs(struct vnode *devvp, struct mount *mp, struct lwp *l, struct msdosfs_args *argp)
458 {
459 	struct msdosfsmount *pmp;
460 	struct buf *bp;
461 	dev_t dev = devvp->v_rdev;
462 	union bootsector *bsp;
463 	struct byte_bpb33 *b33;
464 	struct byte_bpb50 *b50;
465 	struct byte_bpb710 *b710;
466 	uint8_t SecPerClust;
467 	int	ronly, error, BlkPerSec;
468 	uint64_t psize;
469 	unsigned secsize;
470 	u_long fatbytes, fatblocksecs;
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 = getdisksize(devvp, &psize, &secsize);
482 	if (error) {
483 		if (argp->flags & MSDOSFSMNT_GEMDOSFS)
484 			goto error_exit;
485 
486 		/* ok, so it failed.  we most likely don't need the info */
487 		secsize = DEV_BSIZE;
488 		psize = 0;
489 		error = 0;
490 	}
491 	if (secsize < DEV_BSIZE) {
492 		DPRINTF("Invalid block secsize (%d < DEV_BSIZE)", secsize);
493 		error = EINVAL;
494 		goto error_exit;
495 	}
496 
497 	if (argp->flags & MSDOSFSMNT_GEMDOSFS) {
498 		if (secsize != GEMDOSFS_BSIZE) {
499 			DPRINTF("Invalid block secsize %d for GEMDOS", secsize);
500 			error = EINVAL;
501 			goto error_exit;
502 		}
503 	}
504 
505 	/*
506 	 * Read the boot sector of the filesystem, and then check the
507 	 * boot signature.  If not a dos boot sector then error out.
508 	 */
509 	if (secsize < sizeof(*b50)) {
510 		DPRINTF("50 bootsec %u\n", secsize);
511 		error = EINVAL;
512 		goto error_exit;
513 	}
514 	if ((error = bread(devvp, 0, secsize, 0, &bp)) != 0)
515 		goto error_exit;
516 	bsp = (union bootsector *)bp->b_data;
517 	b33 = (struct byte_bpb33 *)bsp->bs33.bsBPB;
518 	b50 = (struct byte_bpb50 *)bsp->bs50.bsBPB;
519 	b710 = (struct byte_bpb710 *)bsp->bs710.bsBPB;
520 
521 	if (!(argp->flags & MSDOSFSMNT_GEMDOSFS)) {
522 		if (bsp->bs50.bsBootSectSig0 != BOOTSIG0
523 		    || bsp->bs50.bsBootSectSig1 != BOOTSIG1) {
524 			DPRINTF("bootsig0 %d bootsig1 %d",
525 			    bsp->bs50.bsBootSectSig0,
526 			    bsp->bs50.bsBootSectSig1);
527 			error = EINVAL;
528 			goto error_exit;
529 		}
530 	}
531 
532 	pmp = malloc(sizeof(*pmp), M_MSDOSFSMNT, M_WAITOK|M_ZERO);
533 	pmp->pm_mountp = mp;
534 
535 	/*
536 	 * Compute several useful quantities from the bpb in the
537 	 * bootsector.  Copy in the dos 5 variant of the bpb then fix up
538 	 * the fields that are different between dos 5 and dos 3.3.
539 	 */
540 	SecPerClust = b50->bpbSecPerClust;
541 	pmp->pm_BytesPerSec = getushort(b50->bpbBytesPerSec);
542 	pmp->pm_ResSectors = getushort(b50->bpbResSectors);
543 	pmp->pm_FATs = b50->bpbFATs;
544 	pmp->pm_RootDirEnts = getushort(b50->bpbRootDirEnts);
545 	pmp->pm_Sectors = getushort(b50->bpbSectors);
546 	pmp->pm_FATsecs = getushort(b50->bpbFATsecs);
547 	pmp->pm_SecPerTrack = getushort(b50->bpbSecPerTrack);
548 	pmp->pm_Heads = getushort(b50->bpbHeads);
549 	pmp->pm_Media = b50->bpbMedia;
550 
551 	if (pmp->pm_Sectors == 0) {
552 		pmp->pm_HiddenSects = getulong(b50->bpbHiddenSecs);
553 		pmp->pm_HugeSectors = getulong(b50->bpbHugeSectors);
554 	} else {
555 		if (secsize < sizeof(*b33)) {
556 			DPRINTF("33 bootsec %u\n", secsize);
557 			error = EINVAL;
558 			goto error_exit;
559 		}
560 		pmp->pm_HiddenSects = getushort(b33->bpbHiddenSecs);
561 		pmp->pm_HugeSectors = pmp->pm_Sectors;
562 	}
563 
564 	/*
565 	 * Sanity checks, from the FAT specification:
566 	 * - sectors per cluster: >= 1, power of 2
567 	 * - logical sector size: >= 1, power of 2
568 	 * - cluster size:        <= max FS block size
569 	 * - number of sectors:   >= 1
570 	 */
571 	if ((SecPerClust == 0) || !powerof2(SecPerClust) ||
572 	    (pmp->pm_BytesPerSec == 0) || !powerof2(pmp->pm_BytesPerSec) ||
573 	    (SecPerClust * pmp->pm_BytesPerSec > MAXBSIZE) ||
574 	    (pmp->pm_HugeSectors == 0)) {
575 		DPRINTF("consistency checks");
576 		error = EINVAL;
577 		goto error_exit;
578 	}
579 
580 	if (!(argp->flags & MSDOSFSMNT_GEMDOSFS) &&
581 	    (pmp->pm_SecPerTrack > 63)) {
582 		DPRINTF("SecPerTrack %d", pmp->pm_SecPerTrack);
583 		error = EINVAL;
584 		goto error_exit;
585 	}
586 
587 	if (pmp->pm_RootDirEnts == 0) {
588 		if (secsize < sizeof(*b710)) {
589 			DPRINTF("710 bootsec %u\n", secsize);
590 			error = EINVAL;
591 			goto error_exit;
592 		}
593 		unsigned short FSVers = getushort(b710->bpbFSVers);
594 		unsigned short ExtFlags = getushort(b710->bpbExtFlags);
595 		/*
596 		 * Some say that bsBootSectSig[23] must be zero, but
597 		 * Windows does not require this and some digital cameras
598 		 * do not set these to zero.  Therefore, do not insist.
599 		 */
600 		if (pmp->pm_Sectors || pmp->pm_FATsecs || FSVers) {
601 			DPRINTF("Sectors %d FATsecs %lu FSVers %d",
602 			    pmp->pm_Sectors, pmp->pm_FATsecs, FSVers);
603 			error = EINVAL;
604 			goto error_exit;
605 		}
606 		pmp->pm_fatmask = FAT32_MASK;
607 		pmp->pm_fatmult = 4;
608 		pmp->pm_fatdiv = 1;
609 		pmp->pm_FATsecs = getulong(b710->bpbBigFATsecs);
610 
611 		/* Mirroring is enabled if the FATMIRROR bit is not set. */
612 		if ((ExtFlags & FATMIRROR) == 0)
613 			pmp->pm_flags |= MSDOSFS_FATMIRROR;
614 		else
615 			pmp->pm_curfat = ExtFlags & FATNUM;
616 	} else
617 		pmp->pm_flags |= MSDOSFS_FATMIRROR;
618 
619 	if (argp->flags & MSDOSFSMNT_GEMDOSFS) {
620 		if (FAT32(pmp)) {
621 			/* GEMDOS doesn't know FAT32. */
622 			DPRINTF("FAT32 for GEMDOS");
623 			error = EINVAL;
624 			goto error_exit;
625 		}
626 
627 		/*
628 		 * Check a few values (could do some more):
629 		 * - logical sector size: >= block size
630 		 * - number of sectors:   <= size of partition
631 		 */
632 		if ((pmp->pm_BytesPerSec < GEMDOSFS_BSIZE) ||
633 		    (pmp->pm_HugeSectors *
634 		     (pmp->pm_BytesPerSec / GEMDOSFS_BSIZE) > psize)) {
635 			DPRINTF("consistency checks for GEMDOS");
636 			error = EINVAL;
637 			goto error_exit;
638 		}
639 		/*
640 		 * XXX - Many parts of the msdosfs driver seem to assume that
641 		 * the number of bytes per logical sector (BytesPerSec) will
642 		 * always be the same as the number of bytes per disk block
643 		 * Let's pretend it is.
644 		 */
645 		BlkPerSec = pmp->pm_BytesPerSec / GEMDOSFS_BSIZE;
646 		pmp->pm_BytesPerSec  = GEMDOSFS_BSIZE;
647 		pmp->pm_HugeSectors *= BlkPerSec;
648 		pmp->pm_HiddenSects *= BlkPerSec;
649 		pmp->pm_ResSectors  *= BlkPerSec;
650 		pmp->pm_Sectors     *= BlkPerSec;
651 		pmp->pm_FATsecs     *= BlkPerSec;
652 		SecPerClust         *= BlkPerSec;
653 	}
654 
655 	/* Check that fs has nonzero FAT size */
656 	if (pmp->pm_FATsecs == 0) {
657 		DPRINTF("FATsecs is 0");
658 		error = EINVAL;
659 		goto error_exit;
660 	}
661 
662 	pmp->pm_fatblk = pmp->pm_ResSectors;
663 	if (FAT32(pmp)) {
664 		if (secsize < sizeof(*b710)) {
665 			DPRINTF("710 bootsec %u\n", secsize);
666 			error = EINVAL;
667 			goto error_exit;
668 		}
669 		pmp->pm_rootdirblk = getulong(b710->bpbRootClust);
670 		pmp->pm_firstcluster = pmp->pm_fatblk
671 			+ (pmp->pm_FATs * pmp->pm_FATsecs);
672 		pmp->pm_fsinfo = getushort(b710->bpbFSInfo);
673 	} else {
674 		pmp->pm_rootdirblk = pmp->pm_fatblk +
675 			(pmp->pm_FATs * pmp->pm_FATsecs);
676 		pmp->pm_rootdirsize = (pmp->pm_RootDirEnts * sizeof(struct direntry)
677 				       + pmp->pm_BytesPerSec - 1)
678 			/ pmp->pm_BytesPerSec;/* in sectors */
679 		pmp->pm_firstcluster = pmp->pm_rootdirblk + pmp->pm_rootdirsize;
680 	}
681 
682 	pmp->pm_nmbrofclusters = (pmp->pm_HugeSectors - pmp->pm_firstcluster) /
683 	    SecPerClust;
684 	pmp->pm_maxcluster = pmp->pm_nmbrofclusters + 1;
685 	pmp->pm_fatsize = pmp->pm_FATsecs * pmp->pm_BytesPerSec;
686 
687 	if (argp->flags & MSDOSFSMNT_GEMDOSFS) {
688 		if (pmp->pm_nmbrofclusters <= (0xff0 - 2)) {
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 	} else if (pmp->pm_fatmask == 0) {
698 		if (pmp->pm_maxcluster
699 		    <= ((CLUST_RSRVD - CLUST_FIRST) & FAT12_MASK)) {
700 			/*
701 			 * This will usually be a floppy disk. This size makes
702 			 * sure that one FAT entry will not be split across
703 			 * multiple blocks.
704 			 */
705 			pmp->pm_fatmask = FAT12_MASK;
706 			pmp->pm_fatmult = 3;
707 			pmp->pm_fatdiv = 2;
708 		} else {
709 			pmp->pm_fatmask = FAT16_MASK;
710 			pmp->pm_fatmult = 2;
711 			pmp->pm_fatdiv = 1;
712 		}
713 	}
714 
715 	/* validate cluster count against FAT */
716 	if ((pmp->pm_maxcluster & pmp->pm_fatmask) != pmp->pm_maxcluster) {
717 		DPRINTF("maxcluster %lu outside of mask %#lx\n",
718 			pmp->pm_maxcluster, pmp->pm_fatmask);
719 		error = EINVAL;
720 		goto error_exit;
721 	}
722 
723 	/* validate FAT size */
724 	fatbytes = (pmp->pm_maxcluster+1) * pmp->pm_fatmult / pmp->pm_fatdiv;
725 	fatblocksecs = howmany(fatbytes, pmp->pm_BytesPerSec);
726 
727 	if (pmp->pm_FATsecs < fatblocksecs) {
728 		DPRINTF("FATsecs %lu < real %lu\n", pmp->pm_FATsecs,
729 			fatblocksecs);
730 		error = EINVAL;
731 		goto error_exit;
732 	}
733 
734 	if (FAT12(pmp)) {
735 		/*
736 		 * limit block size to what is needed to read a FAT block
737 		 * to not exceed MAXBSIZE
738 		 */
739 		pmp->pm_fatblocksec = uimin(3, fatblocksecs);
740 		pmp->pm_fatblocksize = pmp->pm_fatblocksec
741 			* pmp->pm_BytesPerSec;
742 	} else {
743 		pmp->pm_fatblocksize = MAXBSIZE;
744 		pmp->pm_fatblocksec = pmp->pm_fatblocksize
745 			/ pmp->pm_BytesPerSec;
746 	}
747 
748 	pmp->pm_bnshift = ffs(pmp->pm_BytesPerSec) - 1;
749 
750 	/*
751 	 * Compute mask and shift value for isolating cluster relative byte
752 	 * offsets and cluster numbers from a file offset.
753 	 */
754 	pmp->pm_bpcluster = SecPerClust * pmp->pm_BytesPerSec;
755 	pmp->pm_crbomask = pmp->pm_bpcluster - 1;
756 	pmp->pm_cnshift = ffs(pmp->pm_bpcluster) - 1;
757 
758 	/*
759 	 * Check for valid cluster size
760 	 * must be a power of 2
761 	 */
762 	if (pmp->pm_bpcluster ^ (1 << pmp->pm_cnshift)) {
763 		DPRINTF("bpcluster %lu cnshift %lu", pmp->pm_bpcluster,
764 		    pmp->pm_cnshift);
765 		error = EINVAL;
766 		goto error_exit;
767 	}
768 
769 	/*
770 	 * Cluster size must be within limit of MAXBSIZE.
771 	 * Many FAT filesystems will not have clusters larger than
772 	 * 32KiB due to limits in Windows versions before Vista.
773 	 */
774 	if (pmp->pm_bpcluster > MAXBSIZE) {
775 		DPRINTF("bpcluster %lu > MAXBSIZE %d",
776 		    pmp->pm_bpcluster, MAXBSIZE);
777 		error = EINVAL;
778 		goto error_exit;
779 	}
780 
781 	/*
782 	 * Release the bootsector buffer.
783 	 */
784 	brelse(bp, BC_AGE);
785 	bp = NULL;
786 
787 	/*
788 	 * Check FSInfo.
789 	 */
790 	if (pmp->pm_fsinfo) {
791 		struct fsinfo *fp;
792 		const int rdsz = roundup(sizeof(*fp), pmp->pm_BytesPerSec);
793 
794 		/*
795 		 * XXX	If the fsinfo block is stored on media with
796 		 *	2KB or larger sectors, is the fsinfo structure
797 		 *	padded at the end or in the middle?
798 		 */
799 		if ((error = bread(devvp, de_bn2kb(pmp, pmp->pm_fsinfo),
800 		    rdsz, 0, &bp)) != 0)
801 			goto error_exit;
802 		fp = (struct fsinfo *)bp->b_data;
803 		if (!memcmp(fp->fsisig1, "RRaA", 4)
804 		    && !memcmp(fp->fsisig2, "rrAa", 4)
805 		    && !memcmp(fp->fsisig3, "\0\0\125\252", 4)
806 		    && !memcmp(fp->fsisig4, "\0\0\125\252", 4))
807 			pmp->pm_nxtfree = getulong(fp->fsinxtfree);
808 		else
809 			pmp->pm_fsinfo = 0;
810 		brelse(bp, 0);
811 		bp = NULL;
812 	}
813 
814 	/*
815 	 * Check and validate (or perhaps invalidate?) the fsinfo structure?
816 	 * XXX
817 	 */
818 	if (pmp->pm_fsinfo) {
819 		if ((pmp->pm_nxtfree == 0xffffffffUL) ||
820 		    (pmp->pm_nxtfree > pmp->pm_maxcluster))
821 			pmp->pm_fsinfo = 0;
822 	}
823 
824 	/*
825 	 * Allocate memory for the bitmap of allocated clusters, and then
826 	 * fill it in.
827 	 */
828 	pmp->pm_inusemap = malloc(((pmp->pm_maxcluster + N_INUSEBITS)
829 				   / N_INUSEBITS)
830 				  * sizeof(*pmp->pm_inusemap),
831 				  M_MSDOSFSFAT, M_WAITOK);
832 
833 	/*
834 	 * fillinusemap() needs pm_devvp.
835 	 */
836 	pmp->pm_dev = dev;
837 	pmp->pm_devvp = devvp;
838 
839 	/*
840 	 * Have the inuse map filled in.
841 	 */
842 	if ((error = fillinusemap(pmp)) != 0) {
843 		DPRINTF("fillinusemap %d", error);
844 		goto error_exit;
845 	}
846 
847 	/*
848 	 * If they want FAT updates to be synchronous then let them suffer
849 	 * the performance degradation in exchange for the on disk copy of
850 	 * the FAT being correct just about all the time.  I suppose this
851 	 * would be a good thing to turn on if the kernel is still flakey.
852 	 */
853 	if (mp->mnt_flag & MNT_SYNCHRONOUS)
854 		pmp->pm_flags |= MSDOSFSMNT_WAITONFAT;
855 
856 	/*
857 	 * Finish up.
858 	 */
859 	if (ronly)
860 		pmp->pm_flags |= MSDOSFSMNT_RONLY;
861 	else
862 		pmp->pm_fmod = 1;
863 	mp->mnt_data = pmp;
864 	mp->mnt_stat.f_fsidx.__fsid_val[0] = (long)dev;
865 	mp->mnt_stat.f_fsidx.__fsid_val[1] = makefstype(MOUNT_MSDOS);
866 	mp->mnt_stat.f_fsid = mp->mnt_stat.f_fsidx.__fsid_val[0];
867 	mp->mnt_stat.f_namemax = MSDOSFS_NAMEMAX(pmp);
868 	mp->mnt_flag |= MNT_LOCAL;
869 	mp->mnt_dev_bshift = pmp->pm_bnshift;
870 	mp->mnt_fs_bshift = pmp->pm_cnshift;
871 
872 	/*
873 	 * If we ever do quotas for DOS filesystems this would be a place
874 	 * to fill in the info in the msdosfsmount structure. You dolt,
875 	 * quotas on dos filesystems make no sense because files have no
876 	 * owners on dos filesystems. of course there is some empty space
877 	 * in the directory entry where we could put uid's and gid's.
878 	 */
879 
880 	spec_node_setmountedfs(devvp, mp);
881 
882 	return (0);
883 
884 error_exit:
885 	if (bp)
886 		brelse(bp, BC_AGE);
887 	if (pmp) {
888 		if (pmp->pm_inusemap)
889 			free(pmp->pm_inusemap, M_MSDOSFSFAT);
890 		free(pmp, M_MSDOSFSMNT);
891 		mp->mnt_data = NULL;
892 	}
893 	return (error);
894 }
895 
896 int
897 msdosfs_start(struct mount *mp, int flags)
898 {
899 
900 	return (0);
901 }
902 
903 /*
904  * Unmount the filesystem described by mp.
905  */
906 int
907 msdosfs_unmount(struct mount *mp, int mntflags)
908 {
909 	struct msdosfsmount *pmp;
910 	int error, flags;
911 
912 	flags = 0;
913 	if (mntflags & MNT_FORCE)
914 		flags |= FORCECLOSE;
915 	if ((error = vflush(mp, NULLVP, flags)) != 0)
916 		return (error);
917 	pmp = VFSTOMSDOSFS(mp);
918 	if (pmp->pm_devvp->v_type != VBAD)
919 		spec_node_setmountedfs(pmp->pm_devvp, NULL);
920 #ifdef MSDOSFS_DEBUG
921 	{
922 		struct vnode *vp = pmp->pm_devvp;
923 
924 		printf("msdosfs_umount(): just before calling VOP_CLOSE()\n");
925 		printf("flag %08x, usecount %d, writecount %d, holdcnt %d\n",
926 		    vp->v_vflag | vp->v_iflag | vp->v_uflag, vp->v_usecount,
927 		    vp->v_writecount, vp->v_holdcnt);
928 		printf("mount %p, op %p\n",
929 		    vp->v_mount, vp->v_op);
930 		printf("cleanblkhd %p, dirtyblkhd %p, numoutput %d, type %d\n",
931 		    vp->v_cleanblkhd.lh_first,
932 		    vp->v_dirtyblkhd.lh_first,
933 		    vp->v_numoutput, vp->v_type);
934 		printf("union %p, tag %d, data[0] %08x, data[1] %08x\n",
935 		    vp->v_socket, vp->v_tag,
936 		    ((u_int *)vp->v_data)[0],
937 		    ((u_int *)vp->v_data)[1]);
938 	}
939 #endif
940 	vn_lock(pmp->pm_devvp, LK_EXCLUSIVE | LK_RETRY);
941 	(void) VOP_CLOSE(pmp->pm_devvp,
942 	    pmp->pm_flags & MSDOSFSMNT_RONLY ? FREAD : FREAD|FWRITE, NOCRED);
943 	vput(pmp->pm_devvp);
944 	msdosfs_fh_destroy(pmp);
945 	free(pmp->pm_inusemap, M_MSDOSFSFAT);
946 	free(pmp, M_MSDOSFSMNT);
947 	mp->mnt_data = NULL;
948 	mp->mnt_flag &= ~MNT_LOCAL;
949 	return (0);
950 }
951 
952 int
953 msdosfs_root(struct mount *mp, struct vnode **vpp)
954 {
955 	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
956 	int error;
957 
958 #ifdef MSDOSFS_DEBUG
959 	printf("msdosfs_root(); mp %p, pmp %p\n", mp, pmp);
960 #endif
961 	if ((error = deget(pmp, MSDOSFSROOT, MSDOSFSROOT_OFS, vpp)) != 0)
962 		return error;
963 	error = vn_lock(*vpp, LK_EXCLUSIVE);
964 	if (error) {
965 		vrele(*vpp);
966 		*vpp = NULL;
967 		return error;
968 	}
969 	return 0;
970 }
971 
972 int
973 msdosfs_statvfs(struct mount *mp, struct statvfs *sbp)
974 {
975 	struct msdosfsmount *pmp;
976 
977 	pmp = VFSTOMSDOSFS(mp);
978 	sbp->f_bsize = pmp->pm_bpcluster;
979 	sbp->f_frsize = sbp->f_bsize;
980 	sbp->f_iosize = pmp->pm_bpcluster;
981 	sbp->f_blocks = pmp->pm_nmbrofclusters;
982 	sbp->f_bfree = pmp->pm_freeclustercount;
983 	sbp->f_bavail = pmp->pm_freeclustercount;
984 	sbp->f_bresvd = 0;
985 	sbp->f_files = pmp->pm_RootDirEnts;			/* XXX */
986 	sbp->f_ffree = 0;	/* what to put in here? */
987 	sbp->f_favail = 0;	/* what to put in here? */
988 	sbp->f_fresvd = 0;
989 	copy_statvfs_info(sbp, mp);
990 	return (0);
991 }
992 
993 struct msdosfs_sync_ctx {
994 	int waitfor;
995 };
996 
997 static bool
998 msdosfs_sync_selector(void *cl, struct vnode *vp)
999 {
1000 	struct msdosfs_sync_ctx *c = cl;
1001 	struct denode *dep;
1002 
1003 	KASSERT(mutex_owned(vp->v_interlock));
1004 
1005 	dep = VTODE(vp);
1006 	if (c->waitfor == MNT_LAZY || vp->v_type == VNON ||
1007 	    dep == NULL || (((dep->de_flag &
1008 	    (DE_ACCESS | DE_CREATE | DE_UPDATE | DE_MODIFIED)) == 0) &&
1009 	     (LIST_EMPTY(&vp->v_dirtyblkhd) &&
1010 	      UVM_OBJ_IS_CLEAN(&vp->v_uobj))))
1011 		return false;
1012 	return true;
1013 }
1014 
1015 int
1016 msdosfs_sync(struct mount *mp, int waitfor, kauth_cred_t cred)
1017 {
1018 	struct vnode *vp;
1019 	struct vnode_iterator *marker;
1020 	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
1021 	int error, allerror = 0;
1022 	struct msdosfs_sync_ctx ctx;
1023 
1024 	/*
1025 	 * If we ever switch to not updating all of the FATs all the time,
1026 	 * this would be the place to update them from the first one.
1027 	 */
1028 	if (pmp->pm_fmod != 0) {
1029 		if (pmp->pm_flags & MSDOSFSMNT_RONLY)
1030 			panic("msdosfs_sync: rofs mod");
1031 		else {
1032 			/* update FATs here */
1033 		}
1034 	}
1035 	/*
1036 	 * Write back each (modified) denode.
1037 	 */
1038 	vfs_vnode_iterator_init(mp, &marker);
1039 	ctx.waitfor = waitfor;
1040 	while ((vp = vfs_vnode_iterator_next(marker, msdosfs_sync_selector,
1041 	    &ctx)))
1042 	{
1043 		error = vn_lock(vp, LK_EXCLUSIVE);
1044 		if (error) {
1045 			vrele(vp);
1046 			continue;
1047 		}
1048 		if ((error = VOP_FSYNC(vp, cred,
1049 		    waitfor == MNT_WAIT ? FSYNC_WAIT : 0, 0, 0)) != 0)
1050 			allerror = error;
1051 		vput(vp);
1052 	}
1053 	vfs_vnode_iterator_destroy(marker);
1054 
1055 	/*
1056 	 * Force stale file system control information to be flushed.
1057 	 */
1058 	vn_lock(pmp->pm_devvp, LK_EXCLUSIVE | LK_RETRY);
1059 	if ((error = VOP_FSYNC(pmp->pm_devvp, cred,
1060 	    waitfor == MNT_WAIT ? FSYNC_WAIT : 0, 0, 0)) != 0)
1061 		allerror = error;
1062 	VOP_UNLOCK(pmp->pm_devvp);
1063 	return (allerror);
1064 }
1065 
1066 int
1067 msdosfs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
1068 {
1069 	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
1070 	struct defid defh;
1071 	uint32_t gen;
1072 	int error;
1073 
1074 	if (fhp->fid_len != sizeof(struct defid)) {
1075 		DPRINTF("fid_len %d %zd", fhp->fid_len, sizeof(struct defid));
1076 		return EINVAL;
1077 	}
1078 	memcpy(&defh, fhp, sizeof(defh));
1079 	error = msdosfs_fh_lookup(pmp, defh.defid_dirclust, defh.defid_dirofs,
1080 	    &gen);
1081 	if (error == 0 && gen != defh.defid_gen)
1082 		error = ESTALE;
1083 	if (error) {
1084 		*vpp = NULLVP;
1085 		return error;
1086 	}
1087 	error = deget(pmp, defh.defid_dirclust, defh.defid_dirofs, vpp);
1088 	if (error) {
1089 		DPRINTF("deget %d", error);
1090 		*vpp = NULLVP;
1091 		return error;
1092 	}
1093 	error = vn_lock(*vpp, LK_EXCLUSIVE);
1094 	if (error) {
1095 		vrele(*vpp);
1096 		*vpp = NULLVP;
1097 		return error;
1098 	}
1099 	return 0;
1100 }
1101 
1102 int
1103 msdosfs_vptofh(struct vnode *vp, struct fid *fhp, size_t *fh_size)
1104 {
1105 	struct msdosfsmount *pmp = VFSTOMSDOSFS(vp->v_mount);
1106 	struct denode *dep;
1107 	struct defid defh;
1108 	int error;
1109 
1110 	if (*fh_size < sizeof(struct defid)) {
1111 		*fh_size = sizeof(struct defid);
1112 		return E2BIG;
1113 	}
1114 	*fh_size = sizeof(struct defid);
1115 	dep = VTODE(vp);
1116 	memset(&defh, 0, sizeof(defh));
1117 	defh.defid_len = sizeof(struct defid);
1118 	defh.defid_dirclust = dep->de_dirclust;
1119 	defh.defid_dirofs = dep->de_diroffset;
1120 	error = msdosfs_fh_enter(pmp, dep->de_dirclust, dep->de_diroffset,
1121 	     &defh.defid_gen);
1122 	if (error == 0)
1123 		memcpy(fhp, &defh, sizeof(defh));
1124 	return error;
1125 }
1126 
1127 int
1128 msdosfs_vget(struct mount *mp, ino_t ino,
1129     struct vnode **vpp)
1130 {
1131 
1132 	return (EOPNOTSUPP);
1133 }
1134