xref: /netbsd-src/sys/fs/msdosfs/msdosfs_vfsops.c (revision 181254a7b1bdde6873432bffef2d2decc4b5c22f)
1 /*	$NetBSD: msdosfs_vfsops.c,v 1.135 2020/04/13 19:23:17 ad 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 1.135 2020/04/13 19:23:17 ad 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 extern const struct vnodeopv_desc msdosfs_vnodeop_opv_desc;
110 
111 const struct vnodeopv_desc * const msdosfs_vnodeopv_descs[] = {
112 	&msdosfs_vnodeop_opv_desc,
113 	NULL,
114 };
115 
116 struct vfsops msdosfs_vfsops = {
117 	.vfs_name = MOUNT_MSDOS,
118 	.vfs_min_mount_data = sizeof (struct msdosfs_args),
119 	.vfs_mount = msdosfs_mount,
120 	.vfs_start = msdosfs_start,
121 	.vfs_unmount = msdosfs_unmount,
122 	.vfs_root = msdosfs_root,
123 	.vfs_quotactl = (void *)eopnotsupp,
124 	.vfs_statvfs = msdosfs_statvfs,
125 	.vfs_sync = msdosfs_sync,
126 	.vfs_vget = msdosfs_vget,
127 	.vfs_loadvnode = msdosfs_loadvnode,
128 	.vfs_fhtovp = msdosfs_fhtovp,
129 	.vfs_vptofh = msdosfs_vptofh,
130 	.vfs_init = msdosfs_init,
131 	.vfs_reinit = msdosfs_reinit,
132 	.vfs_done = msdosfs_done,
133 	.vfs_mountroot = msdosfs_mountroot,
134 	.vfs_snapshot = (void *)eopnotsupp,
135 	.vfs_extattrctl = vfs_stdextattrctl,
136 	.vfs_suspendctl = genfs_suspendctl,
137 	.vfs_renamelock_enter = genfs_renamelock_enter,
138 	.vfs_renamelock_exit = genfs_renamelock_exit,
139 	.vfs_fsync = (void *)eopnotsupp,
140 	.vfs_opv_descs = msdosfs_vnodeopv_descs
141 };
142 
143 SYSCTL_SETUP(msdosfs_sysctl_setup, "msdosfs sysctl")
144 {
145 	sysctl_createv(clog, 0, NULL, NULL,
146 		       CTLFLAG_PERMANENT,
147 		       CTLTYPE_NODE, "msdosfs",
148 		       SYSCTL_DESCR("MS-DOS file system"),
149 		       NULL, 0, NULL, 0,
150 		       CTL_VFS, 4, CTL_EOL);
151 	/*
152 	 * XXX the "4" above could be dynamic, thereby eliminating one
153 	 * more instance of the "number to vfs" mapping problem, but
154 	 * "4" is the order as taken from sys/mount.h
155 	 */
156 }
157 
158 static int
159 msdos_modcmd(modcmd_t cmd, void *arg)
160 {
161 	int error;
162 
163 	switch (cmd) {
164 	case MODULE_CMD_INIT:
165 		error = vfs_attach(&msdosfs_vfsops);
166 		if (error != 0)
167 			break;
168 		break;
169 	case MODULE_CMD_FINI:
170 		error = vfs_detach(&msdosfs_vfsops);
171 		if (error != 0)
172 			break;
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 			error = msdosfs_root(mp, LK_EXCLUSIVE, &rtvp);
214 			if (error != 0)
215 				return error;
216 			pmp->pm_flags |= findwin95(VTODE(rtvp))
217 				? MSDOSFSMNT_LONGNAME
218 					: MSDOSFSMNT_SHORTNAME;
219 			vput(rtvp);
220 		}
221 	}
222 
223 	mp->mnt_stat.f_namemax = MSDOSFS_NAMEMAX(pmp);
224 
225 	return 0;
226 }
227 
228 int
229 msdosfs_mountroot(void)
230 {
231 	struct mount *mp;
232 	struct lwp *l = curlwp;	/* XXX */
233 	int error;
234 	struct msdosfs_args args;
235 
236 	if (device_class(root_device) != DV_DISK)
237 		return (ENODEV);
238 
239 	if ((error = vfs_rootmountalloc(MOUNT_MSDOS, "root_device", &mp))) {
240 		vrele(rootvp);
241 		return (error);
242 	}
243 
244 	args.flags = MSDOSFSMNT_VERSIONED;
245 	args.uid = 0;
246 	args.gid = 0;
247 	args.mask = 0777;
248 	args.version = MSDOSFSMNT_VERSION;
249 	args.dirmask = 0777;
250 
251 	if ((error = msdosfs_mountfs(rootvp, mp, l, &args)) != 0) {
252 		vfs_unbusy(mp);
253 		vfs_rele(mp);
254 		return (error);
255 	}
256 
257 	if ((error = update_mp(mp, &args)) != 0) {
258 		(void)msdosfs_unmount(mp, 0);
259 		vfs_unbusy(mp);
260 		vfs_rele(mp);
261 		vrele(rootvp);
262 		return (error);
263 	}
264 
265 	mountlist_append(mp);
266 	(void)msdosfs_statvfs(mp, &mp->mnt_stat);
267 	vfs_unbusy(mp);
268 	return (0);
269 }
270 
271 /*
272  * mp - path - addr in user space of mount point (ie /usr or whatever)
273  * data - addr in user space of mount params including the name of the block
274  * special file to treat as a filesystem.
275  */
276 int
277 msdosfs_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
278 {
279 	struct lwp *l = curlwp;
280 	struct vnode *devvp;	  /* vnode for blk device to mount */
281 	struct msdosfs_args *args = data; /* holds data from mount request */
282 	/* msdosfs specific mount control block */
283 	struct msdosfsmount *pmp = NULL;
284 	int error, flags;
285 	mode_t accessmode;
286 
287 	if (args == NULL)
288 		return EINVAL;
289 	if (*data_len < sizeof *args)
290 		return EINVAL;
291 
292 	if (mp->mnt_flag & MNT_GETARGS) {
293 		pmp = VFSTOMSDOSFS(mp);
294 		if (pmp == NULL)
295 			return EIO;
296 		args->fspec = NULL;
297 		args->uid = pmp->pm_uid;
298 		args->gid = pmp->pm_gid;
299 		args->mask = pmp->pm_mask;
300 		args->flags = pmp->pm_flags;
301 		args->version = MSDOSFSMNT_VERSION;
302 		args->dirmask = pmp->pm_dirmask;
303 		args->gmtoff = pmp->pm_gmtoff;
304 		*data_len = sizeof *args;
305 		return 0;
306 	}
307 
308 	/*
309 	 * If not versioned (i.e. using old mount_msdos(8)), fill in
310 	 * the additional structure items with suitable defaults.
311 	 */
312 	if ((args->flags & MSDOSFSMNT_VERSIONED) == 0) {
313 		args->version = 1;
314 		args->dirmask = args->mask;
315 	}
316 
317 	/*
318 	 * Reset GMT offset for pre-v3 mount structure args.
319 	 */
320 	if (args->version < 3)
321 		args->gmtoff = 0;
322 
323 	/*
324 	 * If updating, check whether changing from read-only to
325 	 * read/write; if there is no device name, that's all we do.
326 	 */
327 	if (mp->mnt_flag & MNT_UPDATE) {
328 		pmp = VFSTOMSDOSFS(mp);
329 		error = 0;
330 		if (!(pmp->pm_flags & MSDOSFSMNT_RONLY) &&
331 		    (mp->mnt_flag & MNT_RDONLY)) {
332 			flags = WRITECLOSE;
333 			if (mp->mnt_flag & MNT_FORCE)
334 				flags |= FORCECLOSE;
335 			error = vflush(mp, NULLVP, flags);
336 		}
337 		if (!error && (mp->mnt_flag & MNT_RELOAD))
338 			/* not yet implemented */
339 			error = EOPNOTSUPP;
340 		if (error) {
341 			DPRINTF("vflush %d", error);
342 			return (error);
343 		}
344 		if ((pmp->pm_flags & MSDOSFSMNT_RONLY) &&
345 		    (mp->mnt_iflag & IMNT_WANTRDWR)) {
346 			/*
347 			 * If upgrade to read-write by non-root, then verify
348 			 * that user has necessary permissions on the device.
349 			 *
350 			 * Permission to update a mount is checked higher, so
351 			 * here we presume updating the mount is okay (for
352 			 * example, as far as securelevel goes) which leaves us
353 			 * with the normal check.
354 			 */
355 			devvp = pmp->pm_devvp;
356 			vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
357 			error = kauth_authorize_system(l->l_cred,
358 			    KAUTH_SYSTEM_MOUNT, KAUTH_REQ_SYSTEM_MOUNT_DEVICE,
359 			    mp, devvp, KAUTH_ARG(VREAD | VWRITE));
360 			VOP_UNLOCK(devvp);
361 			DPRINTF("KAUTH_REQ_SYSTEM_MOUNT_DEVICE %d", error);
362 			if (error)
363 				return (error);
364 
365 			pmp->pm_flags &= ~MSDOSFSMNT_RONLY;
366 		}
367 		if (args->fspec == NULL) {
368 			DPRINTF("missing fspec");
369 			return EINVAL;
370 		}
371 	}
372 	/*
373 	 * Not an update, or updating the name: look up the name
374 	 * and verify that it refers to a sensible block device.
375 	 */
376 	error = namei_simple_user(args->fspec,
377 				NSM_FOLLOW_NOEMULROOT, &devvp);
378 	if (error != 0) {
379 		DPRINTF("namei %d", error);
380 		return (error);
381 	}
382 
383 	if (devvp->v_type != VBLK) {
384 		DPRINTF("not block");
385 		vrele(devvp);
386 		return (ENOTBLK);
387 	}
388 	if (bdevsw_lookup(devvp->v_rdev) == NULL) {
389 		DPRINTF("no block switch");
390 		vrele(devvp);
391 		return (ENXIO);
392 	}
393 	/*
394 	 * If mount by non-root, then verify that user has necessary
395 	 * permissions on the device.
396 	 */
397 	accessmode = VREAD;
398 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
399 		accessmode |= VWRITE;
400 	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
401 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
402 	    KAUTH_REQ_SYSTEM_MOUNT_DEVICE, mp, devvp, KAUTH_ARG(accessmode));
403 	VOP_UNLOCK(devvp);
404 	if (error) {
405 		DPRINTF("KAUTH_REQ_SYSTEM_MOUNT_DEVICE %d", error);
406 		vrele(devvp);
407 		return (error);
408 	}
409 	if ((mp->mnt_flag & MNT_UPDATE) == 0) {
410 		int xflags;
411 
412 		if (mp->mnt_flag & MNT_RDONLY)
413 			xflags = FREAD;
414 		else
415 			xflags = FREAD|FWRITE;
416 		vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
417 		error = VOP_OPEN(devvp, xflags, FSCRED);
418 		VOP_UNLOCK(devvp);
419 		if (error) {
420 			DPRINTF("VOP_OPEN %d", error);
421 			goto fail;
422 		}
423 		error = msdosfs_mountfs(devvp, mp, l, args);
424 		if (error) {
425 			DPRINTF("msdosfs_mountfs %d", error);
426 			vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
427 			(void) VOP_CLOSE(devvp, xflags, NOCRED);
428 			VOP_UNLOCK(devvp);
429 			goto fail;
430 		}
431 #ifdef MSDOSFS_DEBUG		/* only needed for the printf below */
432 		pmp = VFSTOMSDOSFS(mp);
433 #endif
434 	} else {
435 		vrele(devvp);
436 		if (devvp != pmp->pm_devvp) {
437 			DPRINTF("devvp %p pmp %p", devvp, pmp->pm_devvp);
438 			return (EINVAL);	/* needs translation */
439 		}
440 	}
441 	if ((error = update_mp(mp, args)) != 0) {
442 		msdosfs_unmount(mp, MNT_FORCE);
443 		DPRINTF("update_mp %d", error);
444 		return error;
445 	}
446 
447 #ifdef MSDOSFS_DEBUG
448 	printf("msdosfs_mount(): mp %p, pmp %p, inusemap %p\n", mp, pmp, pmp->pm_inusemap);
449 #endif
450 	return set_statvfs_info(path, UIO_USERSPACE, args->fspec, UIO_USERSPACE,
451 	    mp->mnt_op->vfs_name, mp, l);
452 
453 fail:
454 	vrele(devvp);
455 	return (error);
456 }
457 
458 int
459 msdosfs_mountfs(struct vnode *devvp, struct mount *mp, struct lwp *l, struct msdosfs_args *argp)
460 {
461 	struct msdosfsmount *pmp;
462 	struct buf *bp;
463 	dev_t dev = devvp->v_rdev;
464 	union bootsector *bsp;
465 	struct byte_bpb33 *b33;
466 	struct byte_bpb50 *b50;
467 	struct byte_bpb710 *b710;
468 	uint8_t SecPerClust;
469 	int	ronly, error, BlkPerSec;
470 	uint64_t psize;
471 	unsigned secsize;
472 	u_long fatbytes, fatblocksecs;
473 
474 	/* Flush out any old buffers remaining from a previous use. */
475 	if ((error = vinvalbuf(devvp, V_SAVE, l->l_cred, l, 0, 0)) != 0)
476 		return (error);
477 
478 	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
479 
480 	bp  = NULL; /* both used in error_exit */
481 	pmp = NULL;
482 
483 	error = getdisksize(devvp, &psize, &secsize);
484 	if (error) {
485 		if (argp->flags & MSDOSFSMNT_GEMDOSFS)
486 			goto error_exit;
487 
488 		/* ok, so it failed.  we most likely don't need the info */
489 		secsize = DEV_BSIZE;
490 		psize = 0;
491 		error = 0;
492 	}
493 	if (secsize < DEV_BSIZE) {
494 		DPRINTF("Invalid block secsize (%d < DEV_BSIZE)", secsize);
495 		error = EINVAL;
496 		goto error_exit;
497 	}
498 
499 	if (argp->flags & MSDOSFSMNT_GEMDOSFS) {
500 		if (secsize != GEMDOSFS_BSIZE) {
501 			DPRINTF("Invalid block secsize %d for GEMDOS", secsize);
502 			error = EINVAL;
503 			goto error_exit;
504 		}
505 	}
506 
507 	/*
508 	 * Read the boot sector of the filesystem, and then check the
509 	 * boot signature.  If not a dos boot sector then error out.
510 	 */
511 	if (secsize < sizeof(*b50)) {
512 		DPRINTF("50 bootsec %u\n", secsize);
513 		error = EINVAL;
514 		goto error_exit;
515 	}
516 	if ((error = bread(devvp, 0, secsize, 0, &bp)) != 0)
517 		goto error_exit;
518 	bsp = (union bootsector *)bp->b_data;
519 	b33 = (struct byte_bpb33 *)bsp->bs33.bsBPB;
520 	b50 = (struct byte_bpb50 *)bsp->bs50.bsBPB;
521 	b710 = (struct byte_bpb710 *)bsp->bs710.bsBPB;
522 
523 	if (!(argp->flags & MSDOSFSMNT_GEMDOSFS)) {
524 		if (bsp->bs50.bsBootSectSig0 != BOOTSIG0
525 		    || bsp->bs50.bsBootSectSig1 != BOOTSIG1) {
526 			DPRINTF("bootsig0 %d bootsig1 %d",
527 			    bsp->bs50.bsBootSectSig0,
528 			    bsp->bs50.bsBootSectSig1);
529 			error = EINVAL;
530 			goto error_exit;
531 		}
532 	}
533 
534 	pmp = malloc(sizeof(*pmp), M_MSDOSFSMNT, M_WAITOK|M_ZERO);
535 	pmp->pm_mountp = mp;
536 
537 	/*
538 	 * Compute several useful quantities from the bpb in the
539 	 * bootsector.  Copy in the dos 5 variant of the bpb then fix up
540 	 * the fields that are different between dos 5 and dos 3.3.
541 	 */
542 	SecPerClust = b50->bpbSecPerClust;
543 	pmp->pm_BytesPerSec = getushort(b50->bpbBytesPerSec);
544 	pmp->pm_ResSectors = getushort(b50->bpbResSectors);
545 	pmp->pm_FATs = b50->bpbFATs;
546 	pmp->pm_RootDirEnts = getushort(b50->bpbRootDirEnts);
547 	pmp->pm_Sectors = getushort(b50->bpbSectors);
548 	pmp->pm_FATsecs = getushort(b50->bpbFATsecs);
549 	pmp->pm_SecPerTrack = getushort(b50->bpbSecPerTrack);
550 	pmp->pm_Heads = getushort(b50->bpbHeads);
551 	pmp->pm_Media = b50->bpbMedia;
552 
553 	if (pmp->pm_Sectors == 0) {
554 		pmp->pm_HiddenSects = getulong(b50->bpbHiddenSecs);
555 		pmp->pm_HugeSectors = getulong(b50->bpbHugeSectors);
556 	} else {
557 		if (secsize < sizeof(*b33)) {
558 			DPRINTF("33 bootsec %u\n", secsize);
559 			error = EINVAL;
560 			goto error_exit;
561 		}
562 		pmp->pm_HiddenSects = getushort(b33->bpbHiddenSecs);
563 		pmp->pm_HugeSectors = pmp->pm_Sectors;
564 	}
565 
566 	/*
567 	 * Sanity checks, from the FAT specification:
568 	 * - sectors per cluster: >= 1, power of 2
569 	 * - logical sector size: >= 1, power of 2
570 	 * - cluster size:        <= max FS block size
571 	 * - number of sectors:   >= 1
572 	 */
573 	if ((SecPerClust == 0) || !powerof2(SecPerClust) ||
574 	    (pmp->pm_BytesPerSec == 0) || !powerof2(pmp->pm_BytesPerSec) ||
575 	    (SecPerClust * pmp->pm_BytesPerSec > MAXBSIZE) ||
576 	    (pmp->pm_HugeSectors == 0)) {
577 		DPRINTF("consistency checks");
578 		error = EINVAL;
579 		goto error_exit;
580 	}
581 
582 	if (!(argp->flags & MSDOSFSMNT_GEMDOSFS) &&
583 	    (pmp->pm_SecPerTrack > 63)) {
584 		DPRINTF("SecPerTrack %d", pmp->pm_SecPerTrack);
585 		error = EINVAL;
586 		goto error_exit;
587 	}
588 
589 	if (pmp->pm_RootDirEnts == 0) {
590 		if (secsize < sizeof(*b710)) {
591 			DPRINTF("710 bootsec %u\n", secsize);
592 			error = EINVAL;
593 			goto error_exit;
594 		}
595 		unsigned short FSVers = getushort(b710->bpbFSVers);
596 		unsigned short ExtFlags = getushort(b710->bpbExtFlags);
597 		/*
598 		 * Some say that bsBootSectSig[23] must be zero, but
599 		 * Windows does not require this and some digital cameras
600 		 * do not set these to zero.  Therefore, do not insist.
601 		 */
602 		if (pmp->pm_Sectors || pmp->pm_FATsecs || FSVers) {
603 			DPRINTF("Sectors %d FATsecs %lu FSVers %d",
604 			    pmp->pm_Sectors, pmp->pm_FATsecs, FSVers);
605 			error = EINVAL;
606 			goto error_exit;
607 		}
608 		pmp->pm_fatmask = FAT32_MASK;
609 		pmp->pm_fatmult = 4;
610 		pmp->pm_fatdiv = 1;
611 		pmp->pm_FATsecs = getulong(b710->bpbBigFATsecs);
612 
613 		/* Mirroring is enabled if the FATMIRROR bit is not set. */
614 		if ((ExtFlags & FATMIRROR) == 0)
615 			pmp->pm_flags |= MSDOSFS_FATMIRROR;
616 		else
617 			pmp->pm_curfat = ExtFlags & FATNUM;
618 	} else
619 		pmp->pm_flags |= MSDOSFS_FATMIRROR;
620 
621 	if (argp->flags & MSDOSFSMNT_GEMDOSFS) {
622 		if (FAT32(pmp)) {
623 			/* GEMDOS doesn't know FAT32. */
624 			DPRINTF("FAT32 for GEMDOS");
625 			error = EINVAL;
626 			goto error_exit;
627 		}
628 
629 		/*
630 		 * Check a few values (could do some more):
631 		 * - logical sector size: >= block size
632 		 * - number of sectors:   <= size of partition
633 		 */
634 		if ((pmp->pm_BytesPerSec < GEMDOSFS_BSIZE) ||
635 		    (pmp->pm_HugeSectors *
636 		     (pmp->pm_BytesPerSec / GEMDOSFS_BSIZE) > psize)) {
637 			DPRINTF("consistency checks for GEMDOS");
638 			error = EINVAL;
639 			goto error_exit;
640 		}
641 		/*
642 		 * XXX - Many parts of the msdosfs driver seem to assume that
643 		 * the number of bytes per logical sector (BytesPerSec) will
644 		 * always be the same as the number of bytes per disk block
645 		 * Let's pretend it is.
646 		 */
647 		BlkPerSec = pmp->pm_BytesPerSec / GEMDOSFS_BSIZE;
648 		pmp->pm_BytesPerSec  = GEMDOSFS_BSIZE;
649 		pmp->pm_HugeSectors *= BlkPerSec;
650 		pmp->pm_HiddenSects *= BlkPerSec;
651 		pmp->pm_ResSectors  *= BlkPerSec;
652 		pmp->pm_Sectors     *= BlkPerSec;
653 		pmp->pm_FATsecs     *= BlkPerSec;
654 		SecPerClust         *= BlkPerSec;
655 	}
656 
657 	/* Check that fs has nonzero FAT size */
658 	if (pmp->pm_FATsecs == 0) {
659 		DPRINTF("FATsecs is 0");
660 		error = EINVAL;
661 		goto error_exit;
662 	}
663 
664 	pmp->pm_fatblk = pmp->pm_ResSectors;
665 	if (FAT32(pmp)) {
666 		if (secsize < sizeof(*b710)) {
667 			DPRINTF("710 bootsec %u\n", secsize);
668 			error = EINVAL;
669 			goto error_exit;
670 		}
671 		pmp->pm_rootdirblk = getulong(b710->bpbRootClust);
672 		pmp->pm_firstcluster = pmp->pm_fatblk
673 			+ (pmp->pm_FATs * pmp->pm_FATsecs);
674 		pmp->pm_fsinfo = getushort(b710->bpbFSInfo);
675 	} else {
676 		pmp->pm_rootdirblk = pmp->pm_fatblk +
677 			(pmp->pm_FATs * pmp->pm_FATsecs);
678 		pmp->pm_rootdirsize = (pmp->pm_RootDirEnts * sizeof(struct direntry)
679 				       + pmp->pm_BytesPerSec - 1)
680 			/ pmp->pm_BytesPerSec;/* in sectors */
681 		pmp->pm_firstcluster = pmp->pm_rootdirblk + pmp->pm_rootdirsize;
682 	}
683 
684 	pmp->pm_nmbrofclusters = (pmp->pm_HugeSectors - pmp->pm_firstcluster) /
685 	    SecPerClust;
686 	pmp->pm_maxcluster = pmp->pm_nmbrofclusters + 1;
687 	pmp->pm_fatsize = pmp->pm_FATsecs * pmp->pm_BytesPerSec;
688 
689 	if (argp->flags & MSDOSFSMNT_GEMDOSFS) {
690 		if (pmp->pm_nmbrofclusters <= (0xff0 - 2)) {
691 			pmp->pm_fatmask = FAT12_MASK;
692 			pmp->pm_fatmult = 3;
693 			pmp->pm_fatdiv = 2;
694 		} else {
695 			pmp->pm_fatmask = FAT16_MASK;
696 			pmp->pm_fatmult = 2;
697 			pmp->pm_fatdiv = 1;
698 		}
699 	} else if (pmp->pm_fatmask == 0) {
700 		if (pmp->pm_maxcluster
701 		    <= ((CLUST_RSRVD - CLUST_FIRST) & FAT12_MASK)) {
702 			/*
703 			 * This will usually be a floppy disk. This size makes
704 			 * sure that one FAT entry will not be split across
705 			 * multiple blocks.
706 			 */
707 			pmp->pm_fatmask = FAT12_MASK;
708 			pmp->pm_fatmult = 3;
709 			pmp->pm_fatdiv = 2;
710 		} else {
711 			pmp->pm_fatmask = FAT16_MASK;
712 			pmp->pm_fatmult = 2;
713 			pmp->pm_fatdiv = 1;
714 		}
715 	}
716 
717 	/* validate cluster count against FAT */
718 	if ((pmp->pm_maxcluster & pmp->pm_fatmask) != pmp->pm_maxcluster) {
719 		DPRINTF("maxcluster %lu outside of mask %#lx\n",
720 			pmp->pm_maxcluster, pmp->pm_fatmask);
721 		error = EINVAL;
722 		goto error_exit;
723 	}
724 
725 	/* validate FAT size */
726 	fatbytes = (pmp->pm_maxcluster+1) * pmp->pm_fatmult / pmp->pm_fatdiv;
727 	fatblocksecs = howmany(fatbytes, pmp->pm_BytesPerSec);
728 
729 	if (pmp->pm_FATsecs < fatblocksecs) {
730 		DPRINTF("FATsecs %lu < real %lu\n", pmp->pm_FATsecs,
731 			fatblocksecs);
732 		error = EINVAL;
733 		goto error_exit;
734 	}
735 
736 	if (FAT12(pmp)) {
737 		/*
738 		 * limit block size to what is needed to read a FAT block
739 		 * to not exceed MAXBSIZE
740 		 */
741 		pmp->pm_fatblocksec = uimin(3, fatblocksecs);
742 		pmp->pm_fatblocksize = pmp->pm_fatblocksec
743 			* pmp->pm_BytesPerSec;
744 	} else {
745 		pmp->pm_fatblocksize = MAXBSIZE;
746 		pmp->pm_fatblocksec = pmp->pm_fatblocksize
747 			/ pmp->pm_BytesPerSec;
748 	}
749 
750 	pmp->pm_bnshift = ffs(pmp->pm_BytesPerSec) - 1;
751 
752 	/*
753 	 * Compute mask and shift value for isolating cluster relative byte
754 	 * offsets and cluster numbers from a file offset.
755 	 */
756 	pmp->pm_bpcluster = SecPerClust * pmp->pm_BytesPerSec;
757 	pmp->pm_crbomask = pmp->pm_bpcluster - 1;
758 	pmp->pm_cnshift = ffs(pmp->pm_bpcluster) - 1;
759 
760 	/*
761 	 * Check for valid cluster size
762 	 * must be a power of 2
763 	 */
764 	if (pmp->pm_bpcluster ^ (1 << pmp->pm_cnshift)) {
765 		DPRINTF("bpcluster %lu cnshift %lu", pmp->pm_bpcluster,
766 		    pmp->pm_cnshift);
767 		error = EINVAL;
768 		goto error_exit;
769 	}
770 
771 	/*
772 	 * Cluster size must be within limit of MAXBSIZE.
773 	 * Many FAT filesystems will not have clusters larger than
774 	 * 32KiB due to limits in Windows versions before Vista.
775 	 */
776 	if (pmp->pm_bpcluster > MAXBSIZE) {
777 		DPRINTF("bpcluster %lu > MAXBSIZE %d",
778 		    pmp->pm_bpcluster, MAXBSIZE);
779 		error = EINVAL;
780 		goto error_exit;
781 	}
782 
783 	/*
784 	 * Release the bootsector buffer.
785 	 */
786 	brelse(bp, BC_AGE);
787 	bp = NULL;
788 
789 	/*
790 	 * Check FSInfo.
791 	 */
792 	if (pmp->pm_fsinfo) {
793 		struct fsinfo *fp;
794 		const int rdsz = roundup(sizeof(*fp), pmp->pm_BytesPerSec);
795 
796 		/*
797 		 * XXX	If the fsinfo block is stored on media with
798 		 *	2KB or larger sectors, is the fsinfo structure
799 		 *	padded at the end or in the middle?
800 		 */
801 		if ((error = bread(devvp, de_bn2kb(pmp, pmp->pm_fsinfo),
802 		    rdsz, 0, &bp)) != 0)
803 			goto error_exit;
804 		fp = (struct fsinfo *)bp->b_data;
805 		if (!memcmp(fp->fsisig1, "RRaA", 4)
806 		    && !memcmp(fp->fsisig2, "rrAa", 4)
807 		    && !memcmp(fp->fsisig3, "\0\0\125\252", 4)
808 		    && !memcmp(fp->fsisig4, "\0\0\125\252", 4))
809 			pmp->pm_nxtfree = getulong(fp->fsinxtfree);
810 		else
811 			pmp->pm_fsinfo = 0;
812 		brelse(bp, 0);
813 		bp = NULL;
814 	}
815 
816 	/*
817 	 * Check and validate (or perhaps invalidate?) the fsinfo structure?
818 	 * XXX
819 	 */
820 	if (pmp->pm_fsinfo) {
821 		if ((pmp->pm_nxtfree == 0xffffffffUL) ||
822 		    (pmp->pm_nxtfree > pmp->pm_maxcluster))
823 			pmp->pm_fsinfo = 0;
824 	}
825 
826 	/*
827 	 * Allocate memory for the bitmap of allocated clusters, and then
828 	 * fill it in.
829 	 */
830 	pmp->pm_inusemap = malloc(((pmp->pm_maxcluster + N_INUSEBITS)
831 				   / N_INUSEBITS)
832 				  * sizeof(*pmp->pm_inusemap),
833 				  M_MSDOSFSFAT, M_WAITOK);
834 
835 	/*
836 	 * fillinusemap() needs pm_devvp.
837 	 */
838 	pmp->pm_dev = dev;
839 	pmp->pm_devvp = devvp;
840 
841 	/*
842 	 * Have the inuse map filled in.
843 	 */
844 	if ((error = fillinusemap(pmp)) != 0) {
845 		DPRINTF("fillinusemap %d", error);
846 		goto error_exit;
847 	}
848 
849 	/*
850 	 * If they want FAT updates to be synchronous then let them suffer
851 	 * the performance degradation in exchange for the on disk copy of
852 	 * the FAT being correct just about all the time.  I suppose this
853 	 * would be a good thing to turn on if the kernel is still flakey.
854 	 */
855 	if (mp->mnt_flag & MNT_SYNCHRONOUS)
856 		pmp->pm_flags |= MSDOSFSMNT_WAITONFAT;
857 
858 	/*
859 	 * Finish up.
860 	 */
861 	if (ronly)
862 		pmp->pm_flags |= MSDOSFSMNT_RONLY;
863 	else
864 		pmp->pm_fmod = 1;
865 	mp->mnt_data = pmp;
866 	mp->mnt_stat.f_fsidx.__fsid_val[0] = (long)dev;
867 	mp->mnt_stat.f_fsidx.__fsid_val[1] = makefstype(MOUNT_MSDOS);
868 	mp->mnt_stat.f_fsid = mp->mnt_stat.f_fsidx.__fsid_val[0];
869 	mp->mnt_stat.f_namemax = MSDOSFS_NAMEMAX(pmp);
870 	mp->mnt_flag |= MNT_LOCAL;
871 	mp->mnt_iflag |= IMNT_SHRLOOKUP;
872 	mp->mnt_dev_bshift = pmp->pm_bnshift;
873 	mp->mnt_fs_bshift = pmp->pm_cnshift;
874 
875 	/*
876 	 * If we ever do quotas for DOS filesystems this would be a place
877 	 * to fill in the info in the msdosfsmount structure. You dolt,
878 	 * quotas on dos filesystems make no sense because files have no
879 	 * owners on dos filesystems. of course there is some empty space
880 	 * in the directory entry where we could put uid's and gid's.
881 	 */
882 
883 	spec_node_setmountedfs(devvp, mp);
884 
885 	return (0);
886 
887 error_exit:
888 	if (bp)
889 		brelse(bp, BC_AGE);
890 	if (pmp) {
891 		if (pmp->pm_inusemap)
892 			free(pmp->pm_inusemap, M_MSDOSFSFAT);
893 		free(pmp, M_MSDOSFSMNT);
894 		mp->mnt_data = NULL;
895 	}
896 	return (error);
897 }
898 
899 int
900 msdosfs_start(struct mount *mp, int flags)
901 {
902 
903 	return (0);
904 }
905 
906 /*
907  * Unmount the filesystem described by mp.
908  */
909 int
910 msdosfs_unmount(struct mount *mp, int mntflags)
911 {
912 	struct msdosfsmount *pmp;
913 	int error, flags;
914 
915 	flags = 0;
916 	if (mntflags & MNT_FORCE)
917 		flags |= FORCECLOSE;
918 	if ((error = vflush(mp, NULLVP, flags)) != 0)
919 		return (error);
920 	pmp = VFSTOMSDOSFS(mp);
921 	if (pmp->pm_devvp->v_type != VBAD)
922 		spec_node_setmountedfs(pmp->pm_devvp, NULL);
923 #ifdef MSDOSFS_DEBUG
924 	{
925 		struct vnode *vp = pmp->pm_devvp;
926 
927 		printf("msdosfs_umount(): just before calling VOP_CLOSE()\n");
928 		printf("flag %08x, usecount %d, writecount %d, holdcnt %d\n",
929 		    vp->v_vflag | vp->v_iflag | vp->v_uflag, vrefcnt(vp),
930 		    vp->v_writecount, vp->v_holdcnt);
931 		printf("mount %p, op %p\n",
932 		    vp->v_mount, vp->v_op);
933 		printf("cleanblkhd %p, dirtyblkhd %p, numoutput %d, type %d\n",
934 		    vp->v_cleanblkhd.lh_first,
935 		    vp->v_dirtyblkhd.lh_first,
936 		    vp->v_numoutput, vp->v_type);
937 		printf("union %p, tag %d, data[0] %08x, data[1] %08x\n",
938 		    vp->v_socket, vp->v_tag,
939 		    ((u_int *)vp->v_data)[0],
940 		    ((u_int *)vp->v_data)[1]);
941 	}
942 #endif
943 	vn_lock(pmp->pm_devvp, LK_EXCLUSIVE | LK_RETRY);
944 	(void) VOP_CLOSE(pmp->pm_devvp,
945 	    pmp->pm_flags & MSDOSFSMNT_RONLY ? FREAD : FREAD|FWRITE, NOCRED);
946 	vput(pmp->pm_devvp);
947 	msdosfs_fh_destroy(pmp);
948 	free(pmp->pm_inusemap, M_MSDOSFSFAT);
949 	free(pmp, M_MSDOSFSMNT);
950 	mp->mnt_data = NULL;
951 	mp->mnt_flag &= ~MNT_LOCAL;
952 	return (0);
953 }
954 
955 int
956 msdosfs_root(struct mount *mp, int lktype, struct vnode **vpp)
957 {
958 	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
959 	int error;
960 
961 #ifdef MSDOSFS_DEBUG
962 	printf("msdosfs_root(); mp %p, pmp %p\n", mp, pmp);
963 #endif
964 	if ((error = deget(pmp, MSDOSFSROOT, MSDOSFSROOT_OFS, vpp)) != 0)
965 		return error;
966 	error = vn_lock(*vpp, lktype);
967 	if (error) {
968 		vrele(*vpp);
969 		*vpp = NULL;
970 		return error;
971 	}
972 	return 0;
973 }
974 
975 int
976 msdosfs_statvfs(struct mount *mp, struct statvfs *sbp)
977 {
978 	struct msdosfsmount *pmp;
979 
980 	pmp = VFSTOMSDOSFS(mp);
981 	sbp->f_bsize = pmp->pm_bpcluster;
982 	sbp->f_frsize = sbp->f_bsize;
983 	sbp->f_iosize = pmp->pm_bpcluster;
984 	sbp->f_blocks = pmp->pm_nmbrofclusters;
985 	sbp->f_bfree = pmp->pm_freeclustercount;
986 	sbp->f_bavail = pmp->pm_freeclustercount;
987 	sbp->f_bresvd = 0;
988 	sbp->f_files = pmp->pm_RootDirEnts;			/* XXX */
989 	sbp->f_ffree = 0;	/* what to put in here? */
990 	sbp->f_favail = 0;	/* what to put in here? */
991 	sbp->f_fresvd = 0;
992 	copy_statvfs_info(sbp, mp);
993 	return (0);
994 }
995 
996 struct msdosfs_sync_ctx {
997 	int waitfor;
998 };
999 
1000 static bool
1001 msdosfs_sync_selector(void *cl, struct vnode *vp)
1002 {
1003 	struct msdosfs_sync_ctx *c = cl;
1004 	struct denode *dep;
1005 
1006 	KASSERT(mutex_owned(vp->v_interlock));
1007 
1008 	dep = VTODE(vp);
1009 	if (c->waitfor == MNT_LAZY || vp->v_type == VNON ||
1010 	    dep == NULL || (((dep->de_flag &
1011 	    (DE_ACCESS | DE_CREATE | DE_UPDATE | DE_MODIFIED)) == 0) &&
1012 	     (LIST_EMPTY(&vp->v_dirtyblkhd) &&
1013 	      (vp->v_iflag & VI_ONWORKLST) == 0)))
1014 		return false;
1015 	return true;
1016 }
1017 
1018 int
1019 msdosfs_sync(struct mount *mp, int waitfor, kauth_cred_t cred)
1020 {
1021 	struct vnode *vp;
1022 	struct vnode_iterator *marker;
1023 	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
1024 	int error, allerror = 0;
1025 	struct msdosfs_sync_ctx ctx;
1026 
1027 	/*
1028 	 * If we ever switch to not updating all of the FATs all the time,
1029 	 * this would be the place to update them from the first one.
1030 	 */
1031 	if (pmp->pm_fmod != 0) {
1032 		if (pmp->pm_flags & MSDOSFSMNT_RONLY)
1033 			panic("msdosfs_sync: rofs mod");
1034 		else {
1035 			/* update FATs here */
1036 		}
1037 	}
1038 	/*
1039 	 * Write back each (modified) denode.
1040 	 */
1041 	vfs_vnode_iterator_init(mp, &marker);
1042 	ctx.waitfor = waitfor;
1043 	while ((vp = vfs_vnode_iterator_next(marker, msdosfs_sync_selector,
1044 	    &ctx)))
1045 	{
1046 		error = vn_lock(vp, LK_EXCLUSIVE);
1047 		if (error) {
1048 			vrele(vp);
1049 			continue;
1050 		}
1051 		if ((error = VOP_FSYNC(vp, cred,
1052 		    waitfor == MNT_WAIT ? FSYNC_WAIT : 0, 0, 0)) != 0)
1053 			allerror = error;
1054 		vput(vp);
1055 	}
1056 	vfs_vnode_iterator_destroy(marker);
1057 
1058 	/*
1059 	 * Force stale file system control information to be flushed.
1060 	 */
1061 	vn_lock(pmp->pm_devvp, LK_EXCLUSIVE | LK_RETRY);
1062 	if ((error = VOP_FSYNC(pmp->pm_devvp, cred,
1063 	    waitfor == MNT_WAIT ? FSYNC_WAIT : 0, 0, 0)) != 0)
1064 		allerror = error;
1065 	VOP_UNLOCK(pmp->pm_devvp);
1066 	return (allerror);
1067 }
1068 
1069 int
1070 msdosfs_fhtovp(struct mount *mp, struct fid *fhp, int lktype, struct vnode **vpp)
1071 {
1072 	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
1073 	struct defid defh;
1074 	uint32_t gen;
1075 	int error;
1076 
1077 	if (fhp->fid_len != sizeof(struct defid)) {
1078 		DPRINTF("fid_len %d %zd", fhp->fid_len, sizeof(struct defid));
1079 		return EINVAL;
1080 	}
1081 	memcpy(&defh, fhp, sizeof(defh));
1082 	error = msdosfs_fh_lookup(pmp, defh.defid_dirclust, defh.defid_dirofs,
1083 	    &gen);
1084 	if (error == 0 && gen != defh.defid_gen)
1085 		error = ESTALE;
1086 	if (error) {
1087 		*vpp = NULLVP;
1088 		return error;
1089 	}
1090 	error = deget(pmp, defh.defid_dirclust, defh.defid_dirofs, vpp);
1091 	if (error) {
1092 		DPRINTF("deget %d", error);
1093 		*vpp = NULLVP;
1094 		return error;
1095 	}
1096 	error = vn_lock(*vpp, lktype);
1097 	if (error) {
1098 		vrele(*vpp);
1099 		*vpp = NULLVP;
1100 		return error;
1101 	}
1102 	return 0;
1103 }
1104 
1105 int
1106 msdosfs_vptofh(struct vnode *vp, struct fid *fhp, size_t *fh_size)
1107 {
1108 	struct msdosfsmount *pmp = VFSTOMSDOSFS(vp->v_mount);
1109 	struct denode *dep;
1110 	struct defid defh;
1111 	int error;
1112 
1113 	if (*fh_size < sizeof(struct defid)) {
1114 		*fh_size = sizeof(struct defid);
1115 		return E2BIG;
1116 	}
1117 	*fh_size = sizeof(struct defid);
1118 	dep = VTODE(vp);
1119 	memset(&defh, 0, sizeof(defh));
1120 	defh.defid_len = sizeof(struct defid);
1121 	defh.defid_dirclust = dep->de_dirclust;
1122 	defh.defid_dirofs = dep->de_diroffset;
1123 	error = msdosfs_fh_enter(pmp, dep->de_dirclust, dep->de_diroffset,
1124 	     &defh.defid_gen);
1125 	if (error == 0)
1126 		memcpy(fhp, &defh, sizeof(defh));
1127 	return error;
1128 }
1129 
1130 int
1131 msdosfs_vget(struct mount *mp, ino_t ino, int lktype,
1132     struct vnode **vpp)
1133 {
1134 
1135 	return (EOPNOTSUPP);
1136 }
1137