xref: /netbsd-src/sys/fs/msdosfs/msdosfs_vfsops.c (revision b757af438b42b93f8c6571f026d8b8ef3eaf5fc9)
1 /*	$NetBSD: msdosfs_vfsops.c,v 1.94 2012/03/13 18:40:37 elad 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 1.94 2012/03/13 18:40:37 elad 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(a) uprintf a
92 #else
93 #define DPRINTF(a)
94 #endif
95 
96 #define MSDOSFS_NAMEMAX(pmp) \
97 	(pmp)->pm_flags & MSDOSFSMNT_LONGNAME ? WIN_MAXLEN : 12
98 
99 VFS_PROTOS(msdosfs);
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 #define ROOTNAME "root_device"
111 
112 static struct sysctllog *msdosfs_sysctl_log;
113 
114 extern const struct vnodeopv_desc msdosfs_vnodeop_opv_desc;
115 
116 const struct vnodeopv_desc * const msdosfs_vnodeopv_descs[] = {
117 	&msdosfs_vnodeop_opv_desc,
118 	NULL,
119 };
120 
121 struct vfsops msdosfs_vfsops = {
122 	MOUNT_MSDOS,
123 	sizeof (struct msdosfs_args),
124 	msdosfs_mount,
125 	msdosfs_start,
126 	msdosfs_unmount,
127 	msdosfs_root,
128 	(void *)eopnotsupp,		/* vfs_quotactl */
129 	msdosfs_statvfs,
130 	msdosfs_sync,
131 	msdosfs_vget,
132 	msdosfs_fhtovp,
133 	msdosfs_vptofh,
134 	msdosfs_init,
135 	msdosfs_reinit,
136 	msdosfs_done,
137 	msdosfs_mountroot,
138 	(int (*)(struct mount *, struct vnode *, struct timespec *)) eopnotsupp,
139 	vfs_stdextattrctl,
140 	msdosfs_suspendctl,
141 	genfs_renamelock_enter,
142 	genfs_renamelock_exit,
143 	(void *)eopnotsupp,
144 	msdosfs_vnodeopv_descs,
145 	0,
146 	{ NULL, NULL },
147 };
148 
149 static int
150 msdos_modcmd(modcmd_t cmd, void *arg)
151 {
152 	int error;
153 
154 	switch (cmd) {
155 	case MODULE_CMD_INIT:
156 		error = vfs_attach(&msdosfs_vfsops);
157 		if (error != 0)
158 			break;
159 		sysctl_createv(&msdosfs_sysctl_log, 0, NULL, NULL,
160 			       CTLFLAG_PERMANENT,
161 			       CTLTYPE_NODE, "vfs", NULL,
162 			       NULL, 0, NULL, 0,
163 			       CTL_VFS, CTL_EOL);
164 		sysctl_createv(&msdosfs_sysctl_log, 0, NULL, NULL,
165 			       CTLFLAG_PERMANENT,
166 			       CTLTYPE_NODE, "msdosfs",
167 			       SYSCTL_DESCR("MS-DOS file system"),
168 			       NULL, 0, NULL, 0,
169 			       CTL_VFS, 4, CTL_EOL);
170 		/*
171 		 * XXX the "4" above could be dynamic, thereby eliminating one
172 		 * more instance of the "number to vfs" mapping problem, but
173 		 * "4" is the order as taken from sys/mount.h
174 		 */
175 		break;
176 	case MODULE_CMD_FINI:
177 		error = vfs_detach(&msdosfs_vfsops);
178 		if (error != 0)
179 			break;
180 		sysctl_teardown(&msdosfs_sysctl_log);
181 		break;
182 	default:
183 		error = ENOTTY;
184 		break;
185 	}
186 
187 	return (error);
188 }
189 
190 static int
191 update_mp(struct mount *mp, struct msdosfs_args *argp)
192 {
193 	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
194 	int error;
195 
196 	pmp->pm_gid = argp->gid;
197 	pmp->pm_uid = argp->uid;
198 	pmp->pm_mask = argp->mask & ALLPERMS;
199 	pmp->pm_dirmask = argp->dirmask & ALLPERMS;
200 	pmp->pm_gmtoff = argp->gmtoff;
201 	pmp->pm_flags |= argp->flags & MSDOSFSMNT_MNTOPT;
202 
203 	/*
204 	 * GEMDOS knows nothing about win95 long filenames
205 	 */
206 	if (pmp->pm_flags & MSDOSFSMNT_GEMDOSFS)
207 		pmp->pm_flags |= MSDOSFSMNT_NOWIN95;
208 
209 	if (pmp->pm_flags & MSDOSFSMNT_NOWIN95)
210 		pmp->pm_flags |= MSDOSFSMNT_SHORTNAME;
211 	else if (!(pmp->pm_flags &
212 	    (MSDOSFSMNT_SHORTNAME | MSDOSFSMNT_LONGNAME))) {
213 		struct vnode *rtvp;
214 
215 		/*
216 		 * Try to divine whether to support Win'95 long filenames
217 		 */
218 		if (FAT32(pmp))
219 			pmp->pm_flags |= MSDOSFSMNT_LONGNAME;
220 		else {
221 			if ((error = msdosfs_root(mp, &rtvp)) != 0)
222 				return error;
223 			pmp->pm_flags |= findwin95(VTODE(rtvp))
224 				? MSDOSFSMNT_LONGNAME
225 					: MSDOSFSMNT_SHORTNAME;
226 			vput(rtvp);
227 		}
228 	}
229 
230 	mp->mnt_stat.f_namemax = MSDOSFS_NAMEMAX(pmp);
231 
232 	return 0;
233 }
234 
235 int
236 msdosfs_mountroot(void)
237 {
238 	struct mount *mp;
239 	struct lwp *l = curlwp;	/* XXX */
240 	int error;
241 	struct msdosfs_args args;
242 
243 	if (device_class(root_device) != DV_DISK)
244 		return (ENODEV);
245 
246 	if ((error = vfs_rootmountalloc(MOUNT_MSDOS, "root_device", &mp))) {
247 		vrele(rootvp);
248 		return (error);
249 	}
250 
251 	args.flags = MSDOSFSMNT_VERSIONED;
252 	args.uid = 0;
253 	args.gid = 0;
254 	args.mask = 0777;
255 	args.version = MSDOSFSMNT_VERSION;
256 	args.dirmask = 0777;
257 
258 	if ((error = msdosfs_mountfs(rootvp, mp, l, &args)) != 0) {
259 		vfs_unbusy(mp, false, NULL);
260 		vfs_destroy(mp);
261 		return (error);
262 	}
263 
264 	if ((error = update_mp(mp, &args)) != 0) {
265 		(void)msdosfs_unmount(mp, 0);
266 		vfs_unbusy(mp, false, NULL);
267 		vfs_destroy(mp);
268 		vrele(rootvp);
269 		return (error);
270 	}
271 
272 	mutex_enter(&mountlist_lock);
273 	CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
274 	mutex_exit(&mountlist_lock);
275 	(void)msdosfs_statvfs(mp, &mp->mnt_stat);
276 	vfs_unbusy(mp, false, NULL);
277 	return (0);
278 }
279 
280 /*
281  * mp - path - addr in user space of mount point (ie /usr or whatever)
282  * data - addr in user space of mount params including the name of the block
283  * special file to treat as a filesystem.
284  */
285 int
286 msdosfs_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
287 {
288 	struct lwp *l = curlwp;
289 	struct vnode *devvp;	  /* vnode for blk device to mount */
290 	struct msdosfs_args *args = data; /* holds data from mount request */
291 	/* msdosfs specific mount control block */
292 	struct msdosfsmount *pmp = NULL;
293 	int error, flags;
294 	mode_t accessmode;
295 
296 	if (*data_len < sizeof *args)
297 		return EINVAL;
298 
299 	if (mp->mnt_flag & MNT_GETARGS) {
300 		pmp = VFSTOMSDOSFS(mp);
301 		if (pmp == NULL)
302 			return EIO;
303 		args->fspec = NULL;
304 		args->uid = pmp->pm_uid;
305 		args->gid = pmp->pm_gid;
306 		args->mask = pmp->pm_mask;
307 		args->flags = pmp->pm_flags;
308 		args->version = MSDOSFSMNT_VERSION;
309 		args->dirmask = pmp->pm_dirmask;
310 		args->gmtoff = pmp->pm_gmtoff;
311 		*data_len = sizeof *args;
312 		return 0;
313 	}
314 
315 	/*
316 	 * If not versioned (i.e. using old mount_msdos(8)), fill in
317 	 * the additional structure items with suitable defaults.
318 	 */
319 	if ((args->flags & MSDOSFSMNT_VERSIONED) == 0) {
320 		args->version = 1;
321 		args->dirmask = args->mask;
322 	}
323 
324 	/*
325 	 * Reset GMT offset for pre-v3 mount structure args.
326 	 */
327 	if (args->version < 3)
328 		args->gmtoff = 0;
329 
330 	/*
331 	 * If updating, check whether changing from read-only to
332 	 * read/write; if there is no device name, that's all we do.
333 	 */
334 	if (mp->mnt_flag & MNT_UPDATE) {
335 		pmp = VFSTOMSDOSFS(mp);
336 		error = 0;
337 		if (!(pmp->pm_flags & MSDOSFSMNT_RONLY) &&
338 		    (mp->mnt_flag & MNT_RDONLY)) {
339 			flags = WRITECLOSE;
340 			if (mp->mnt_flag & MNT_FORCE)
341 				flags |= FORCECLOSE;
342 			error = vflush(mp, NULLVP, flags);
343 		}
344 		if (!error && (mp->mnt_flag & MNT_RELOAD))
345 			/* not yet implemented */
346 			error = EOPNOTSUPP;
347 		if (error) {
348 			DPRINTF(("vflush %d\n", error));
349 			return (error);
350 		}
351 		if ((pmp->pm_flags & MSDOSFSMNT_RONLY) &&
352 		    (mp->mnt_iflag & IMNT_WANTRDWR)) {
353 			/*
354 			 * If upgrade to read-write by non-root, then verify
355 			 * that user has necessary permissions on the device.
356 			 *
357 			 * Permission to update a mount is checked higher, so
358 			 * here we presume updating the mount is okay (for
359 			 * example, as far as securelevel goes) which leaves us
360 			 * with the normal check.
361 			 */
362 			devvp = pmp->pm_devvp;
363 			vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
364 			error = kauth_authorize_system(l->l_cred,
365 			    KAUTH_SYSTEM_MOUNT, KAUTH_REQ_SYSTEM_MOUNT_DEVICE,
366 			    mp, devvp, KAUTH_ARG(VREAD | VWRITE));
367 			VOP_UNLOCK(devvp);
368 			DPRINTF(("KAUTH_REQ_SYSTEM_MOUNT_DEVICE %d\n", error));
369 			if (error)
370 				return (error);
371 
372 			pmp->pm_flags &= ~MSDOSFSMNT_RONLY;
373 		}
374 		if (args->fspec == NULL) {
375 			DPRINTF(("missing fspec\n"));
376 			return EINVAL;
377 		}
378 	}
379 	/*
380 	 * Not an update, or updating the name: look up the name
381 	 * and verify that it refers to a sensible block device.
382 	 */
383 	error = namei_simple_user(args->fspec,
384 				NSM_FOLLOW_NOEMULROOT, &devvp);
385 	if (error != 0) {
386 		DPRINTF(("namei %d\n", error));
387 		return (error);
388 	}
389 
390 	if (devvp->v_type != VBLK) {
391 		DPRINTF(("not block\n"));
392 		vrele(devvp);
393 		return (ENOTBLK);
394 	}
395 	if (bdevsw_lookup(devvp->v_rdev) == NULL) {
396 		DPRINTF(("no block switch\n"));
397 		vrele(devvp);
398 		return (ENXIO);
399 	}
400 	/*
401 	 * If mount by non-root, then verify that user has necessary
402 	 * permissions on the device.
403 	 */
404 	accessmode = VREAD;
405 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
406 		accessmode |= VWRITE;
407 	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
408 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
409 	    KAUTH_REQ_SYSTEM_MOUNT_DEVICE, mp, devvp, KAUTH_ARG(accessmode));
410 	VOP_UNLOCK(devvp);
411 	if (error) {
412 		DPRINTF(("KAUTH_REQ_SYSTEM_MOUNT_DEVICE %d\n", error));
413 		vrele(devvp);
414 		return (error);
415 	}
416 	if ((mp->mnt_flag & MNT_UPDATE) == 0) {
417 		int xflags;
418 
419 		if (mp->mnt_flag & MNT_RDONLY)
420 			xflags = FREAD;
421 		else
422 			xflags = FREAD|FWRITE;
423 		vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
424 		error = VOP_OPEN(devvp, xflags, FSCRED);
425 		VOP_UNLOCK(devvp);
426 		if (error) {
427 			DPRINTF(("VOP_OPEN %d\n", error));
428 			goto fail;
429 		}
430 		error = msdosfs_mountfs(devvp, mp, l, args);
431 		if (error) {
432 			DPRINTF(("msdosfs_mountfs %d\n", error));
433 			vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
434 			(void) VOP_CLOSE(devvp, xflags, NOCRED);
435 			VOP_UNLOCK(devvp);
436 			goto fail;
437 		}
438 #ifdef MSDOSFS_DEBUG		/* only needed for the printf below */
439 		pmp = VFSTOMSDOSFS(mp);
440 #endif
441 	} else {
442 		vrele(devvp);
443 		if (devvp != pmp->pm_devvp) {
444 			DPRINTF(("devvp %p pmp %p\n",
445 			    devvp, pmp->pm_devvp));
446 			return (EINVAL);	/* needs translation */
447 		}
448 	}
449 	if ((error = update_mp(mp, args)) != 0) {
450 		msdosfs_unmount(mp, MNT_FORCE);
451 		DPRINTF(("update_mp %d\n", error));
452 		return error;
453 	}
454 
455 #ifdef MSDOSFS_DEBUG
456 	printf("msdosfs_mount(): mp %p, pmp %p, inusemap %p\n", mp, pmp, pmp->pm_inusemap);
457 #endif
458 	return set_statvfs_info(path, UIO_USERSPACE, args->fspec, UIO_USERSPACE,
459 	    mp->mnt_op->vfs_name, mp, l);
460 
461 fail:
462 	vrele(devvp);
463 	return (error);
464 }
465 
466 int
467 msdosfs_mountfs(struct vnode *devvp, struct mount *mp, struct lwp *l, struct msdosfs_args *argp)
468 {
469 	struct msdosfsmount *pmp;
470 	struct buf *bp;
471 	dev_t dev = devvp->v_rdev;
472 	union bootsector *bsp;
473 	struct byte_bpb33 *b33;
474 	struct byte_bpb50 *b50;
475 	struct byte_bpb710 *b710;
476 	uint8_t SecPerClust;
477 	int	ronly, error, tmp;
478 	int	bsize;
479 	uint64_t psize;
480 	unsigned secsize;
481 
482 	/* Flush out any old buffers remaining from a previous use. */
483 	if ((error = vinvalbuf(devvp, V_SAVE, l->l_cred, l, 0, 0)) != 0)
484 		return (error);
485 
486 	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
487 
488 	bp  = NULL; /* both used in error_exit */
489 	pmp = NULL;
490 
491 	error = fstrans_mount(mp);
492 	if (error)
493 		goto error_exit;
494 
495 	error = getdisksize(devvp, &psize, &secsize);
496 	if (error) {
497 		if (argp->flags & MSDOSFSMNT_GEMDOSFS)
498 			goto error_exit;
499 
500 		/* ok, so it failed.  we most likely don't need the info */
501 		secsize = DEV_BSIZE;
502 		psize = 0;
503 		error = 0;
504 	}
505 
506 	if (argp->flags & MSDOSFSMNT_GEMDOSFS) {
507 		bsize = secsize;
508 		if (bsize != 512) {
509 			DPRINTF(("Invalid block bsize %d for gemdos\n", bsize));
510 			error = EINVAL;
511 			goto error_exit;
512 		}
513 	} else
514 		bsize = 0;
515 
516 	/*
517 	 * Read the boot sector of the filesystem, and then check the
518 	 * boot signature.  If not a dos boot sector then error out.
519 	 */
520 	if ((error = bread(devvp, 0, secsize, NOCRED, 0, &bp)) != 0)
521 		goto error_exit;
522 	bsp = (union bootsector *)bp->b_data;
523 	b33 = (struct byte_bpb33 *)bsp->bs33.bsBPB;
524 	b50 = (struct byte_bpb50 *)bsp->bs50.bsBPB;
525 	b710 = (struct byte_bpb710 *)bsp->bs710.bsBPB;
526 
527 	if (!(argp->flags & MSDOSFSMNT_GEMDOSFS)) {
528 		if (bsp->bs50.bsBootSectSig0 != BOOTSIG0
529 		    || bsp->bs50.bsBootSectSig1 != BOOTSIG1) {
530 			DPRINTF(("bootsig0 %d bootsig1 %d\n",
531 			    bsp->bs50.bsBootSectSig0,
532 			    bsp->bs50.bsBootSectSig1));
533 			error = EINVAL;
534 			goto error_exit;
535 		}
536 	}
537 
538 	pmp = malloc(sizeof *pmp, M_MSDOSFSMNT, M_WAITOK);
539 	memset(pmp, 0, sizeof *pmp);
540 	pmp->pm_mountp = mp;
541 
542 	/*
543 	 * Compute several useful quantities from the bpb in the
544 	 * bootsector.  Copy in the dos 5 variant of the bpb then fix up
545 	 * the fields that are different between dos 5 and dos 3.3.
546 	 */
547 	SecPerClust = b50->bpbSecPerClust;
548 	pmp->pm_BytesPerSec = getushort(b50->bpbBytesPerSec);
549 	pmp->pm_ResSectors = getushort(b50->bpbResSectors);
550 	pmp->pm_FATs = b50->bpbFATs;
551 	pmp->pm_RootDirEnts = getushort(b50->bpbRootDirEnts);
552 	pmp->pm_Sectors = getushort(b50->bpbSectors);
553 	pmp->pm_FATsecs = getushort(b50->bpbFATsecs);
554 	pmp->pm_SecPerTrack = getushort(b50->bpbSecPerTrack);
555 	pmp->pm_Heads = getushort(b50->bpbHeads);
556 	pmp->pm_Media = b50->bpbMedia;
557 
558 	if (!(argp->flags & MSDOSFSMNT_GEMDOSFS)) {
559 		/* XXX - We should probably check more values here */
560     		if (!pmp->pm_BytesPerSec || !SecPerClust
561 	    		|| pmp->pm_SecPerTrack > 63) {
562 			DPRINTF(("bytespersec %d secperclust %d "
563 			    "secpertrack %d\n",
564 			    pmp->pm_BytesPerSec, SecPerClust,
565 			    pmp->pm_SecPerTrack));
566 			error = EINVAL;
567 			goto error_exit;
568 		}
569 	}
570 
571 	if (pmp->pm_Sectors == 0) {
572 		pmp->pm_HiddenSects = getulong(b50->bpbHiddenSecs);
573 		pmp->pm_HugeSectors = getulong(b50->bpbHugeSectors);
574 	} else {
575 		pmp->pm_HiddenSects = getushort(b33->bpbHiddenSecs);
576 		pmp->pm_HugeSectors = pmp->pm_Sectors;
577 	}
578 
579 	if (pmp->pm_RootDirEnts == 0) {
580 		unsigned short vers = getushort(b710->bpbFSVers);
581 		/*
582 		 * Some say that bsBootSectSig[23] must be zero, but
583 		 * Windows does not require this and some digital cameras
584 		 * do not set these to zero.  Therefore, do not insist.
585 		 */
586 		if (pmp->pm_Sectors || pmp->pm_FATsecs || vers) {
587 			DPRINTF(("sectors %d fatsecs %lu vers %d\n",
588 			    pmp->pm_Sectors, pmp->pm_FATsecs, vers));
589 			error = EINVAL;
590 			goto error_exit;
591 		}
592 		pmp->pm_fatmask = FAT32_MASK;
593 		pmp->pm_fatmult = 4;
594 		pmp->pm_fatdiv = 1;
595 		pmp->pm_FATsecs = getulong(b710->bpbBigFATsecs);
596 
597 		/* mirrorring is enabled if the FATMIRROR bit is not set */
598 		if ((getushort(b710->bpbExtFlags) & FATMIRROR) == 0)
599 			pmp->pm_flags |= MSDOSFS_FATMIRROR;
600 		else
601 			pmp->pm_curfat = getushort(b710->bpbExtFlags) & FATNUM;
602 	} else
603 		pmp->pm_flags |= MSDOSFS_FATMIRROR;
604 
605 	if (argp->flags & MSDOSFSMNT_GEMDOSFS) {
606 		if (FAT32(pmp)) {
607 			DPRINTF(("fat32 for gemdos\n"));
608 			/*
609 			 * GEMDOS doesn't know fat32.
610 			 */
611 			error = EINVAL;
612 			goto error_exit;
613 		}
614 
615 		/*
616 		 * Check a few values (could do some more):
617 		 * - logical sector size: power of 2, >= block size
618 		 * - sectors per cluster: power of 2, >= 1
619 		 * - number of sectors:   >= 1, <= size of partition
620 		 */
621 		if ( (SecPerClust == 0)
622 		  || (SecPerClust & (SecPerClust - 1))
623 		  || (pmp->pm_BytesPerSec < bsize)
624 		  || (pmp->pm_BytesPerSec & (pmp->pm_BytesPerSec - 1))
625 		  || (pmp->pm_HugeSectors == 0)
626 		  || (pmp->pm_HugeSectors * (pmp->pm_BytesPerSec / bsize)
627 		      > psize)) {
628 			DPRINTF(("consistency checks for gemdos\n"));
629 			error = EINVAL;
630 			goto error_exit;
631 		}
632 		/*
633 		 * XXX - Many parts of the msdos fs driver seem to assume that
634 		 * the number of bytes per logical sector (BytesPerSec) will
635 		 * always be the same as the number of bytes per disk block
636 		 * Let's pretend it is.
637 		 */
638 		tmp = pmp->pm_BytesPerSec / bsize;
639 		pmp->pm_BytesPerSec  = bsize;
640 		pmp->pm_HugeSectors *= tmp;
641 		pmp->pm_HiddenSects *= tmp;
642 		pmp->pm_ResSectors  *= tmp;
643 		pmp->pm_Sectors     *= tmp;
644 		pmp->pm_FATsecs     *= tmp;
645 		SecPerClust         *= tmp;
646 	}
647 
648 	/* Check that fs has nonzero FAT size */
649 	if (pmp->pm_FATsecs == 0) {
650 		DPRINTF(("FATsecs is 0\n"));
651 		error = EINVAL;
652 		goto error_exit;
653 	}
654 
655 	pmp->pm_fatblk = pmp->pm_ResSectors;
656 	if (FAT32(pmp)) {
657 		pmp->pm_rootdirblk = getulong(b710->bpbRootClust);
658 		pmp->pm_firstcluster = pmp->pm_fatblk
659 			+ (pmp->pm_FATs * pmp->pm_FATsecs);
660 		pmp->pm_fsinfo = getushort(b710->bpbFSInfo);
661 	} else {
662 		pmp->pm_rootdirblk = pmp->pm_fatblk +
663 			(pmp->pm_FATs * pmp->pm_FATsecs);
664 		pmp->pm_rootdirsize = (pmp->pm_RootDirEnts * sizeof(struct direntry)
665 				       + pmp->pm_BytesPerSec - 1)
666 			/ pmp->pm_BytesPerSec;/* in sectors */
667 		pmp->pm_firstcluster = pmp->pm_rootdirblk + pmp->pm_rootdirsize;
668 	}
669 
670 	pmp->pm_nmbrofclusters = (pmp->pm_HugeSectors - pmp->pm_firstcluster) /
671 	    SecPerClust;
672 	pmp->pm_maxcluster = pmp->pm_nmbrofclusters + 1;
673 	pmp->pm_fatsize = pmp->pm_FATsecs * pmp->pm_BytesPerSec;
674 
675 	if (argp->flags & MSDOSFSMNT_GEMDOSFS) {
676 		if (pmp->pm_nmbrofclusters <= (0xff0 - 2)) {
677 			pmp->pm_fatmask = FAT12_MASK;
678 			pmp->pm_fatmult = 3;
679 			pmp->pm_fatdiv = 2;
680 		} else {
681 			pmp->pm_fatmask = FAT16_MASK;
682 			pmp->pm_fatmult = 2;
683 			pmp->pm_fatdiv = 1;
684 		}
685 	} else if (pmp->pm_fatmask == 0) {
686 		if (pmp->pm_maxcluster
687 		    <= ((CLUST_RSRVD - CLUST_FIRST) & FAT12_MASK)) {
688 			/*
689 			 * This will usually be a floppy disk. This size makes
690 			 * sure that one fat entry will not be split across
691 			 * multiple blocks.
692 			 */
693 			pmp->pm_fatmask = FAT12_MASK;
694 			pmp->pm_fatmult = 3;
695 			pmp->pm_fatdiv = 2;
696 		} else {
697 			pmp->pm_fatmask = FAT16_MASK;
698 			pmp->pm_fatmult = 2;
699 			pmp->pm_fatdiv = 1;
700 		}
701 	}
702 	if (FAT12(pmp))
703 		pmp->pm_fatblocksize = 3 * pmp->pm_BytesPerSec;
704 	else
705 		pmp->pm_fatblocksize = MAXBSIZE;
706 
707 	pmp->pm_fatblocksec = pmp->pm_fatblocksize / pmp->pm_BytesPerSec;
708 	pmp->pm_bnshift = ffs(pmp->pm_BytesPerSec) - 1;
709 
710 	/*
711 	 * Compute mask and shift value for isolating cluster relative byte
712 	 * offsets and cluster numbers from a file offset.
713 	 */
714 	pmp->pm_bpcluster = SecPerClust * pmp->pm_BytesPerSec;
715 	pmp->pm_crbomask = pmp->pm_bpcluster - 1;
716 	pmp->pm_cnshift = ffs(pmp->pm_bpcluster) - 1;
717 
718 	/*
719 	 * Check for valid cluster size
720 	 * must be a power of 2
721 	 */
722 	if (pmp->pm_bpcluster ^ (1 << pmp->pm_cnshift)) {
723 		DPRINTF(("bpcluster %lu cnshift %lu\n",
724 		    pmp->pm_bpcluster, pmp->pm_cnshift));
725 		error = EINVAL;
726 		goto error_exit;
727 	}
728 
729 	/*
730 	 * Release the bootsector buffer.
731 	 */
732 	brelse(bp, BC_AGE);
733 	bp = NULL;
734 
735 	/*
736 	 * Check FSInfo.
737 	 */
738 	if (pmp->pm_fsinfo) {
739 		struct fsinfo *fp;
740 
741 		/*
742 		 * XXX	If the fsinfo block is stored on media with
743 		 *	2KB or larger sectors, is the fsinfo structure
744 		 *	padded at the end or in the middle?
745 		 */
746 		if ((error = bread(devvp, de_bn2kb(pmp, pmp->pm_fsinfo),
747 		    pmp->pm_BytesPerSec, NOCRED, 0, &bp)) != 0)
748 			goto error_exit;
749 		fp = (struct fsinfo *)bp->b_data;
750 		if (!memcmp(fp->fsisig1, "RRaA", 4)
751 		    && !memcmp(fp->fsisig2, "rrAa", 4)
752 		    && !memcmp(fp->fsisig3, "\0\0\125\252", 4)
753 		    && !memcmp(fp->fsisig4, "\0\0\125\252", 4))
754 			pmp->pm_nxtfree = getulong(fp->fsinxtfree);
755 		else
756 			pmp->pm_fsinfo = 0;
757 		brelse(bp, 0);
758 		bp = NULL;
759 	}
760 
761 	/*
762 	 * Check and validate (or perhaps invalidate?) the fsinfo structure?
763 	 * XXX
764 	 */
765 	if (pmp->pm_fsinfo) {
766 		if (pmp->pm_nxtfree == (u_long)-1)
767 			pmp->pm_fsinfo = 0;
768 	}
769 
770 	/*
771 	 * Allocate memory for the bitmap of allocated clusters, and then
772 	 * fill it in.
773 	 */
774 	pmp->pm_inusemap = malloc(((pmp->pm_maxcluster + N_INUSEBITS - 1)
775 				   / N_INUSEBITS)
776 				  * sizeof(*pmp->pm_inusemap),
777 				  M_MSDOSFSFAT, M_WAITOK);
778 
779 	/*
780 	 * fillinusemap() needs pm_devvp.
781 	 */
782 	pmp->pm_dev = dev;
783 	pmp->pm_devvp = devvp;
784 
785 	/*
786 	 * Have the inuse map filled in.
787 	 */
788 	if ((error = fillinusemap(pmp)) != 0) {
789 		DPRINTF(("fillinusemap %d\n", error));
790 		goto error_exit;
791 	}
792 
793 	/*
794 	 * If they want fat updates to be synchronous then let them suffer
795 	 * the performance degradation in exchange for the on disk copy of
796 	 * the fat being correct just about all the time.  I suppose this
797 	 * would be a good thing to turn on if the kernel is still flakey.
798 	 */
799 	if (mp->mnt_flag & MNT_SYNCHRONOUS)
800 		pmp->pm_flags |= MSDOSFSMNT_WAITONFAT;
801 
802 	/*
803 	 * Finish up.
804 	 */
805 	if (ronly)
806 		pmp->pm_flags |= MSDOSFSMNT_RONLY;
807 	else
808 		pmp->pm_fmod = 1;
809 	mp->mnt_data = pmp;
810 	mp->mnt_stat.f_fsidx.__fsid_val[0] = (long)dev;
811 	mp->mnt_stat.f_fsidx.__fsid_val[1] = makefstype(MOUNT_MSDOS);
812 	mp->mnt_stat.f_fsid = mp->mnt_stat.f_fsidx.__fsid_val[0];
813 	mp->mnt_stat.f_namemax = MSDOSFS_NAMEMAX(pmp);
814 	mp->mnt_flag |= MNT_LOCAL;
815 	mp->mnt_dev_bshift = pmp->pm_bnshift;
816 	mp->mnt_fs_bshift = pmp->pm_cnshift;
817 
818 	/*
819 	 * If we ever do quotas for DOS filesystems this would be a place
820 	 * to fill in the info in the msdosfsmount structure. You dolt,
821 	 * quotas on dos filesystems make no sense because files have no
822 	 * owners on dos filesystems. of course there is some empty space
823 	 * in the directory entry where we could put uid's and gid's.
824 	 */
825 
826 	devvp->v_specmountpoint = mp;
827 
828 	return (0);
829 
830 error_exit:
831 	fstrans_unmount(mp);
832 	if (bp)
833 		brelse(bp, BC_AGE);
834 	if (pmp) {
835 		if (pmp->pm_inusemap)
836 			free(pmp->pm_inusemap, M_MSDOSFSFAT);
837 		free(pmp, M_MSDOSFSMNT);
838 		mp->mnt_data = NULL;
839 	}
840 	return (error);
841 }
842 
843 int
844 msdosfs_start(struct mount *mp, int flags)
845 {
846 
847 	return (0);
848 }
849 
850 /*
851  * Unmount the filesystem described by mp.
852  */
853 int
854 msdosfs_unmount(struct mount *mp, int mntflags)
855 {
856 	struct msdosfsmount *pmp;
857 	int error, flags;
858 
859 	flags = 0;
860 	if (mntflags & MNT_FORCE)
861 		flags |= FORCECLOSE;
862 	if ((error = vflush(mp, NULLVP, flags)) != 0)
863 		return (error);
864 	pmp = VFSTOMSDOSFS(mp);
865 	if (pmp->pm_devvp->v_type != VBAD)
866 		pmp->pm_devvp->v_specmountpoint = NULL;
867 #ifdef MSDOSFS_DEBUG
868 	{
869 		struct vnode *vp = pmp->pm_devvp;
870 
871 		printf("msdosfs_umount(): just before calling VOP_CLOSE()\n");
872 		printf("flag %08x, usecount %d, writecount %d, holdcnt %d\n",
873 		    vp->v_vflag | vp->v_iflag | vp->v_uflag, vp->v_usecount,
874 		    vp->v_writecount, vp->v_holdcnt);
875 		printf("mount %p, op %p\n",
876 		    vp->v_mount, vp->v_op);
877 		printf("freef %p, freeb %p, mount %p\n",
878 		    vp->v_freelist.tqe_next, vp->v_freelist.tqe_prev,
879 		    vp->v_mount);
880 		printf("cleanblkhd %p, dirtyblkhd %p, numoutput %d, type %d\n",
881 		    vp->v_cleanblkhd.lh_first,
882 		    vp->v_dirtyblkhd.lh_first,
883 		    vp->v_numoutput, vp->v_type);
884 		printf("union %p, tag %d, data[0] %08x, data[1] %08x\n",
885 		    vp->v_socket, vp->v_tag,
886 		    ((u_int *)vp->v_data)[0],
887 		    ((u_int *)vp->v_data)[1]);
888 	}
889 #endif
890 	vn_lock(pmp->pm_devvp, LK_EXCLUSIVE | LK_RETRY);
891 	(void) VOP_CLOSE(pmp->pm_devvp,
892 	    pmp->pm_flags & MSDOSFSMNT_RONLY ? FREAD : FREAD|FWRITE, NOCRED);
893 	vput(pmp->pm_devvp);
894 	msdosfs_fh_destroy(pmp);
895 	free(pmp->pm_inusemap, M_MSDOSFSFAT);
896 	free(pmp, M_MSDOSFSMNT);
897 	mp->mnt_data = NULL;
898 	mp->mnt_flag &= ~MNT_LOCAL;
899 	fstrans_unmount(mp);
900 	return (0);
901 }
902 
903 int
904 msdosfs_root(struct mount *mp, struct vnode **vpp)
905 {
906 	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
907 	struct denode *ndep;
908 	int error;
909 
910 #ifdef MSDOSFS_DEBUG
911 	printf("msdosfs_root(); mp %p, pmp %p\n", mp, pmp);
912 #endif
913 	if ((error = deget(pmp, MSDOSFSROOT, MSDOSFSROOT_OFS, &ndep)) != 0)
914 		return (error);
915 	*vpp = DETOV(ndep);
916 	return (0);
917 }
918 
919 int
920 msdosfs_statvfs(struct mount *mp, struct statvfs *sbp)
921 {
922 	struct msdosfsmount *pmp;
923 
924 	pmp = VFSTOMSDOSFS(mp);
925 	sbp->f_bsize = pmp->pm_bpcluster;
926 	sbp->f_frsize = sbp->f_bsize;
927 	sbp->f_iosize = pmp->pm_bpcluster;
928 	sbp->f_blocks = pmp->pm_nmbrofclusters;
929 	sbp->f_bfree = pmp->pm_freeclustercount;
930 	sbp->f_bavail = pmp->pm_freeclustercount;
931 	sbp->f_bresvd = 0;
932 	sbp->f_files = pmp->pm_RootDirEnts;			/* XXX */
933 	sbp->f_ffree = 0;	/* what to put in here? */
934 	sbp->f_favail = 0;	/* what to put in here? */
935 	sbp->f_fresvd = 0;
936 	copy_statvfs_info(sbp, mp);
937 	return (0);
938 }
939 
940 int
941 msdosfs_sync(struct mount *mp, int waitfor, kauth_cred_t cred)
942 {
943 	struct vnode *vp, *mvp;
944 	struct denode *dep;
945 	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
946 	int error, allerror = 0;
947 
948 	/*
949 	 * If we ever switch to not updating all of the fats all the time,
950 	 * this would be the place to update them from the first one.
951 	 */
952 	if (pmp->pm_fmod != 0) {
953 		if (pmp->pm_flags & MSDOSFSMNT_RONLY)
954 			panic("msdosfs_sync: rofs mod");
955 		else {
956 			/* update fats here */
957 		}
958 	}
959 	/* Allocate a marker vnode. */
960 	mvp = vnalloc(mp);
961 	fstrans_start(mp, FSTRANS_SHARED);
962 	/*
963 	 * Write back each (modified) denode.
964 	 */
965 	mutex_enter(&mntvnode_lock);
966 loop:
967 	for (vp = TAILQ_FIRST(&mp->mnt_vnodelist); vp; vp = vunmark(mvp)) {
968 		vmark(mvp, vp);
969 		if (vp->v_mount != mp || vismarker(vp))
970 			continue;
971 		mutex_enter(vp->v_interlock);
972 		dep = VTODE(vp);
973 		if (waitfor == MNT_LAZY || vp->v_type == VNON ||
974 		    dep == NULL || (((dep->de_flag &
975 		    (DE_ACCESS | DE_CREATE | DE_UPDATE | DE_MODIFIED)) == 0) &&
976 		     (LIST_EMPTY(&vp->v_dirtyblkhd) &&
977 		      UVM_OBJ_IS_CLEAN(&vp->v_uobj)))) {
978 			mutex_exit(vp->v_interlock);
979 			continue;
980 		}
981 		mutex_exit(&mntvnode_lock);
982 		error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT);
983 		if (error) {
984 			mutex_enter(&mntvnode_lock);
985 			if (error == ENOENT) {
986 				(void)vunmark(mvp);
987 				goto loop;
988 			}
989 			continue;
990 		}
991 		if ((error = VOP_FSYNC(vp, cred,
992 		    waitfor == MNT_WAIT ? FSYNC_WAIT : 0, 0, 0)) != 0)
993 			allerror = error;
994 		vput(vp);
995 		mutex_enter(&mntvnode_lock);
996 	}
997 	mutex_exit(&mntvnode_lock);
998 	vnfree(mvp);
999 
1000 	/*
1001 	 * Force stale file system control information to be flushed.
1002 	 */
1003 	if ((error = VOP_FSYNC(pmp->pm_devvp, cred,
1004 	    waitfor == MNT_WAIT ? FSYNC_WAIT : 0, 0, 0)) != 0)
1005 		allerror = error;
1006 	fstrans_done(mp);
1007 	return (allerror);
1008 }
1009 
1010 int
1011 msdosfs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
1012 {
1013 	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
1014 	struct defid defh;
1015 	struct denode *dep;
1016 	uint32_t gen;
1017 	int error;
1018 
1019 	if (fhp->fid_len != sizeof(struct defid)) {
1020 		DPRINTF(("fid_len %d %zd\n", fhp->fid_len,
1021 		    sizeof(struct defid)));
1022 		return EINVAL;
1023 	}
1024 	memcpy(&defh, fhp, sizeof(defh));
1025 	error = msdosfs_fh_lookup(pmp, defh.defid_dirclust, defh.defid_dirofs,
1026 	    &gen);
1027 	if (error == 0 && gen != defh.defid_gen)
1028 		error = ESTALE;
1029 	if (error) {
1030 		*vpp = NULLVP;
1031 		return error;
1032 	}
1033 	error = deget(pmp, defh.defid_dirclust, defh.defid_dirofs, &dep);
1034 	if (error) {
1035 		DPRINTF(("deget %d\n", error));
1036 		*vpp = NULLVP;
1037 		return (error);
1038 	}
1039 	*vpp = DETOV(dep);
1040 	return (0);
1041 }
1042 
1043 int
1044 msdosfs_vptofh(struct vnode *vp, struct fid *fhp, size_t *fh_size)
1045 {
1046 	struct msdosfsmount *pmp = VFSTOMSDOSFS(vp->v_mount);
1047 	struct denode *dep;
1048 	struct defid defh;
1049 	int error;
1050 
1051 	if (*fh_size < sizeof(struct defid)) {
1052 		*fh_size = sizeof(struct defid);
1053 		return E2BIG;
1054 	}
1055 	*fh_size = sizeof(struct defid);
1056 	dep = VTODE(vp);
1057 	memset(&defh, 0, sizeof(defh));
1058 	defh.defid_len = sizeof(struct defid);
1059 	defh.defid_dirclust = dep->de_dirclust;
1060 	defh.defid_dirofs = dep->de_diroffset;
1061 	error = msdosfs_fh_enter(pmp, dep->de_dirclust, dep->de_diroffset,
1062 	     &defh.defid_gen);
1063 	if (error == 0)
1064 		memcpy(fhp, &defh, sizeof(defh));
1065 	return error;
1066 }
1067 
1068 int
1069 msdosfs_vget(struct mount *mp, ino_t ino,
1070     struct vnode **vpp)
1071 {
1072 
1073 	return (EOPNOTSUPP);
1074 }
1075 
1076 int
1077 msdosfs_suspendctl(struct mount *mp, int cmd)
1078 {
1079 	int error;
1080 	struct lwp *l = curlwp;
1081 
1082 	switch (cmd) {
1083 	case SUSPEND_SUSPEND:
1084 		if ((error = fstrans_setstate(mp, FSTRANS_SUSPENDING)) != 0)
1085 			return error;
1086 		error = msdosfs_sync(mp, MNT_WAIT, l->l_proc->p_cred);
1087 		if (error == 0)
1088 			error = fstrans_setstate(mp, FSTRANS_SUSPENDED);
1089 		if (error != 0) {
1090 			(void) fstrans_setstate(mp, FSTRANS_NORMAL);
1091 			return error;
1092 		}
1093 		return 0;
1094 
1095 	case SUSPEND_RESUME:
1096 		return fstrans_setstate(mp, FSTRANS_NORMAL);
1097 
1098 	default:
1099 		return EINVAL;
1100 	}
1101 }
1102