xref: /netbsd-src/sys/fs/msdosfs/msdosfs_vfsops.c (revision b1c86f5f087524e68db12794ee9c3e3da1ab17a0)
1 /*	$NetBSD: msdosfs_vfsops.c,v 1.87 2010/07/21 17:52:10 hannken 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 1.87 2010/07/21 17:52:10 hannken 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 = genfs_can_mount(devvp, VREAD | VWRITE,
365 			    l->l_cred);
366 			VOP_UNLOCK(devvp);
367 			DPRINTF(("genfs_can_mount %d\n", error));
368 			if (error)
369 				return (error);
370 
371 			pmp->pm_flags &= ~MSDOSFSMNT_RONLY;
372 		}
373 		if (args->fspec == NULL) {
374 			DPRINTF(("missing fspec\n"));
375 			return EINVAL;
376 		}
377 	}
378 	/*
379 	 * Not an update, or updating the name: look up the name
380 	 * and verify that it refers to a sensible block device.
381 	 */
382 	error = namei_simple_user(args->fspec,
383 				NSM_FOLLOW_NOEMULROOT, &devvp);
384 	if (error != 0) {
385 		DPRINTF(("namei %d\n", error));
386 		return (error);
387 	}
388 
389 	if (devvp->v_type != VBLK) {
390 		DPRINTF(("not block\n"));
391 		vrele(devvp);
392 		return (ENOTBLK);
393 	}
394 	if (bdevsw_lookup(devvp->v_rdev) == NULL) {
395 		DPRINTF(("no block switch\n"));
396 		vrele(devvp);
397 		return (ENXIO);
398 	}
399 	/*
400 	 * If mount by non-root, then verify that user has necessary
401 	 * permissions on the device.
402 	 */
403 	accessmode = VREAD;
404 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
405 		accessmode |= VWRITE;
406 	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
407 	error = genfs_can_mount(devvp, accessmode, l->l_cred);
408 	VOP_UNLOCK(devvp);
409 	if (error) {
410 		DPRINTF(("genfs_can_mount %d\n", error));
411 		vrele(devvp);
412 		return (error);
413 	}
414 	if ((mp->mnt_flag & MNT_UPDATE) == 0) {
415 		int xflags;
416 
417 		if (mp->mnt_flag & MNT_RDONLY)
418 			xflags = FREAD;
419 		else
420 			xflags = FREAD|FWRITE;
421 		error = VOP_OPEN(devvp, xflags, FSCRED);
422 		if (error) {
423 			DPRINTF(("VOP_OPEN %d\n", error));
424 			goto fail;
425 		}
426 		error = msdosfs_mountfs(devvp, mp, l, args);
427 		if (error) {
428 			DPRINTF(("msdosfs_mountfs %d\n", error));
429 			vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
430 			(void) VOP_CLOSE(devvp, xflags, NOCRED);
431 			VOP_UNLOCK(devvp);
432 			goto fail;
433 		}
434 #ifdef MSDOSFS_DEBUG		/* only needed for the printf below */
435 		pmp = VFSTOMSDOSFS(mp);
436 #endif
437 	} else {
438 		vrele(devvp);
439 		if (devvp != pmp->pm_devvp) {
440 			DPRINTF(("devvp %p pmp %p\n",
441 			    devvp, pmp->pm_devvp));
442 			return (EINVAL);	/* needs translation */
443 		}
444 	}
445 	if ((error = update_mp(mp, args)) != 0) {
446 		msdosfs_unmount(mp, MNT_FORCE);
447 		DPRINTF(("update_mp %d\n", error));
448 		return error;
449 	}
450 
451 #ifdef MSDOSFS_DEBUG
452 	printf("msdosfs_mount(): mp %p, pmp %p, inusemap %p\n", mp, pmp, pmp->pm_inusemap);
453 #endif
454 	return set_statvfs_info(path, UIO_USERSPACE, args->fspec, UIO_USERSPACE,
455 	    mp->mnt_op->vfs_name, mp, l);
456 
457 fail:
458 	vrele(devvp);
459 	return (error);
460 }
461 
462 int
463 msdosfs_mountfs(struct vnode *devvp, struct mount *mp, struct lwp *l, struct msdosfs_args *argp)
464 {
465 	struct msdosfsmount *pmp;
466 	struct buf *bp;
467 	dev_t dev = devvp->v_rdev;
468 	union bootsector *bsp;
469 	struct byte_bpb33 *b33;
470 	struct byte_bpb50 *b50;
471 	struct byte_bpb710 *b710;
472 	uint8_t SecPerClust;
473 	int	ronly, error, tmp;
474 	int	bsize;
475 	uint64_t psize;
476 	unsigned secsize;
477 
478 	/* Flush out any old buffers remaining from a previous use. */
479 	if ((error = vinvalbuf(devvp, V_SAVE, l->l_cred, l, 0, 0)) != 0)
480 		return (error);
481 
482 	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
483 
484 	bp  = NULL; /* both used in error_exit */
485 	pmp = NULL;
486 
487 	error = fstrans_mount(mp);
488 	if (error)
489 		goto error_exit;
490 
491 	error = getdisksize(devvp, &psize, &secsize);
492 	if (error) {
493 		if (argp->flags & MSDOSFSMNT_GEMDOSFS)
494 			goto error_exit;
495 
496 		/* ok, so it failed.  we most likely don't need the info */
497 		secsize = DEV_BSIZE;
498 		psize = 0;
499 		error = 0;
500 	}
501 
502 	if (argp->flags & MSDOSFSMNT_GEMDOSFS) {
503 		bsize = secsize;
504 		if (bsize != 512) {
505 			DPRINTF(("Invalid block bsize %d for gemdos\n", bsize));
506 			error = EINVAL;
507 			goto error_exit;
508 		}
509 	} else
510 		bsize = 0;
511 
512 	/*
513 	 * Read the boot sector of the filesystem, and then check the
514 	 * boot signature.  If not a dos boot sector then error out.
515 	 */
516 	if ((error = bread(devvp, 0, secsize, NOCRED, 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\n",
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);
535 	memset(pmp, 0, sizeof *pmp);
536 	pmp->pm_mountp = mp;
537 
538 	/*
539 	 * Compute several useful quantities from the bpb in the
540 	 * bootsector.  Copy in the dos 5 variant of the bpb then fix up
541 	 * the fields that are different between dos 5 and dos 3.3.
542 	 */
543 	SecPerClust = b50->bpbSecPerClust;
544 	pmp->pm_BytesPerSec = getushort(b50->bpbBytesPerSec);
545 	pmp->pm_ResSectors = getushort(b50->bpbResSectors);
546 	pmp->pm_FATs = b50->bpbFATs;
547 	pmp->pm_RootDirEnts = getushort(b50->bpbRootDirEnts);
548 	pmp->pm_Sectors = getushort(b50->bpbSectors);
549 	pmp->pm_FATsecs = getushort(b50->bpbFATsecs);
550 	pmp->pm_SecPerTrack = getushort(b50->bpbSecPerTrack);
551 	pmp->pm_Heads = getushort(b50->bpbHeads);
552 	pmp->pm_Media = b50->bpbMedia;
553 
554 	if (!(argp->flags & MSDOSFSMNT_GEMDOSFS)) {
555 		/* XXX - We should probably check more values here */
556     		if (!pmp->pm_BytesPerSec || !SecPerClust
557 	    		|| pmp->pm_SecPerTrack > 63) {
558 			DPRINTF(("bytespersec %d secperclust %d "
559 			    "secpertrack %d\n",
560 			    pmp->pm_BytesPerSec, SecPerClust,
561 			    pmp->pm_SecPerTrack));
562 			error = EINVAL;
563 			goto error_exit;
564 		}
565 	}
566 
567 	if (pmp->pm_Sectors == 0) {
568 		pmp->pm_HiddenSects = getulong(b50->bpbHiddenSecs);
569 		pmp->pm_HugeSectors = getulong(b50->bpbHugeSectors);
570 	} else {
571 		pmp->pm_HiddenSects = getushort(b33->bpbHiddenSecs);
572 		pmp->pm_HugeSectors = pmp->pm_Sectors;
573 	}
574 
575 	if (pmp->pm_RootDirEnts == 0) {
576 		unsigned short vers = getushort(b710->bpbFSVers);
577 		/*
578 		 * Some say that bsBootSectSig[23] must be zero, but
579 		 * Windows does not require this and some digital cameras
580 		 * do not set these to zero.  Therefore, do not insist.
581 		 */
582 		if (pmp->pm_Sectors || pmp->pm_FATsecs || vers) {
583 			DPRINTF(("sectors %d fatsecs %lu vers %d\n",
584 			    pmp->pm_Sectors, pmp->pm_FATsecs, vers));
585 			error = EINVAL;
586 			goto error_exit;
587 		}
588 		pmp->pm_fatmask = FAT32_MASK;
589 		pmp->pm_fatmult = 4;
590 		pmp->pm_fatdiv = 1;
591 		pmp->pm_FATsecs = getulong(b710->bpbBigFATsecs);
592 
593 		/* mirrorring is enabled if the FATMIRROR bit is not set */
594 		if ((getushort(b710->bpbExtFlags) & FATMIRROR) == 0)
595 			pmp->pm_flags |= MSDOSFS_FATMIRROR;
596 		else
597 			pmp->pm_curfat = getushort(b710->bpbExtFlags) & FATNUM;
598 	} else
599 		pmp->pm_flags |= MSDOSFS_FATMIRROR;
600 
601 	if (argp->flags & MSDOSFSMNT_GEMDOSFS) {
602 		if (FAT32(pmp)) {
603 			DPRINTF(("fat32 for gemdos\n"));
604 			/*
605 			 * GEMDOS doesn't know fat32.
606 			 */
607 			error = EINVAL;
608 			goto error_exit;
609 		}
610 
611 		/*
612 		 * Check a few values (could do some more):
613 		 * - logical sector size: power of 2, >= block size
614 		 * - sectors per cluster: power of 2, >= 1
615 		 * - number of sectors:   >= 1, <= size of partition
616 		 */
617 		if ( (SecPerClust == 0)
618 		  || (SecPerClust & (SecPerClust - 1))
619 		  || (pmp->pm_BytesPerSec < bsize)
620 		  || (pmp->pm_BytesPerSec & (pmp->pm_BytesPerSec - 1))
621 		  || (pmp->pm_HugeSectors == 0)
622 		  || (pmp->pm_HugeSectors * (pmp->pm_BytesPerSec / bsize)
623 		      > psize)) {
624 			DPRINTF(("consistency checks for gemdos\n"));
625 			error = EINVAL;
626 			goto error_exit;
627 		}
628 		/*
629 		 * XXX - Many parts of the msdos fs driver seem to assume that
630 		 * the number of bytes per logical sector (BytesPerSec) will
631 		 * always be the same as the number of bytes per disk block
632 		 * Let's pretend it is.
633 		 */
634 		tmp = pmp->pm_BytesPerSec / bsize;
635 		pmp->pm_BytesPerSec  = bsize;
636 		pmp->pm_HugeSectors *= tmp;
637 		pmp->pm_HiddenSects *= tmp;
638 		pmp->pm_ResSectors  *= tmp;
639 		pmp->pm_Sectors     *= tmp;
640 		pmp->pm_FATsecs     *= tmp;
641 		SecPerClust         *= tmp;
642 	}
643 
644 	/* Check that fs has nonzero FAT size */
645 	if (pmp->pm_FATsecs == 0) {
646 		DPRINTF(("FATsecs is 0\n"));
647 		error = EINVAL;
648 		goto error_exit;
649 	}
650 
651 	pmp->pm_fatblk = pmp->pm_ResSectors;
652 	if (FAT32(pmp)) {
653 		pmp->pm_rootdirblk = getulong(b710->bpbRootClust);
654 		pmp->pm_firstcluster = pmp->pm_fatblk
655 			+ (pmp->pm_FATs * pmp->pm_FATsecs);
656 		pmp->pm_fsinfo = getushort(b710->bpbFSInfo);
657 	} else {
658 		pmp->pm_rootdirblk = pmp->pm_fatblk +
659 			(pmp->pm_FATs * pmp->pm_FATsecs);
660 		pmp->pm_rootdirsize = (pmp->pm_RootDirEnts * sizeof(struct direntry)
661 				       + pmp->pm_BytesPerSec - 1)
662 			/ pmp->pm_BytesPerSec;/* in sectors */
663 		pmp->pm_firstcluster = pmp->pm_rootdirblk + pmp->pm_rootdirsize;
664 	}
665 
666 	pmp->pm_nmbrofclusters = (pmp->pm_HugeSectors - pmp->pm_firstcluster) /
667 	    SecPerClust;
668 	pmp->pm_maxcluster = pmp->pm_nmbrofclusters + 1;
669 	pmp->pm_fatsize = pmp->pm_FATsecs * pmp->pm_BytesPerSec;
670 
671 	if (argp->flags & MSDOSFSMNT_GEMDOSFS) {
672 		if (pmp->pm_nmbrofclusters <= (0xff0 - 2)) {
673 			pmp->pm_fatmask = FAT12_MASK;
674 			pmp->pm_fatmult = 3;
675 			pmp->pm_fatdiv = 2;
676 		} else {
677 			pmp->pm_fatmask = FAT16_MASK;
678 			pmp->pm_fatmult = 2;
679 			pmp->pm_fatdiv = 1;
680 		}
681 	} else if (pmp->pm_fatmask == 0) {
682 		if (pmp->pm_maxcluster
683 		    <= ((CLUST_RSRVD - CLUST_FIRST) & FAT12_MASK)) {
684 			/*
685 			 * This will usually be a floppy disk. This size makes
686 			 * sure that one fat entry will not be split across
687 			 * multiple blocks.
688 			 */
689 			pmp->pm_fatmask = FAT12_MASK;
690 			pmp->pm_fatmult = 3;
691 			pmp->pm_fatdiv = 2;
692 		} else {
693 			pmp->pm_fatmask = FAT16_MASK;
694 			pmp->pm_fatmult = 2;
695 			pmp->pm_fatdiv = 1;
696 		}
697 	}
698 	if (FAT12(pmp))
699 		pmp->pm_fatblocksize = 3 * pmp->pm_BytesPerSec;
700 	else
701 		pmp->pm_fatblocksize = MAXBSIZE;
702 
703 	pmp->pm_fatblocksec = pmp->pm_fatblocksize / pmp->pm_BytesPerSec;
704 	pmp->pm_bnshift = ffs(pmp->pm_BytesPerSec) - 1;
705 
706 	/*
707 	 * Compute mask and shift value for isolating cluster relative byte
708 	 * offsets and cluster numbers from a file offset.
709 	 */
710 	pmp->pm_bpcluster = SecPerClust * pmp->pm_BytesPerSec;
711 	pmp->pm_crbomask = pmp->pm_bpcluster - 1;
712 	pmp->pm_cnshift = ffs(pmp->pm_bpcluster) - 1;
713 
714 	/*
715 	 * Check for valid cluster size
716 	 * must be a power of 2
717 	 */
718 	if (pmp->pm_bpcluster ^ (1 << pmp->pm_cnshift)) {
719 		DPRINTF(("bpcluster %lu cnshift %lu\n",
720 		    pmp->pm_bpcluster, pmp->pm_cnshift));
721 		error = EINVAL;
722 		goto error_exit;
723 	}
724 
725 	/*
726 	 * Release the bootsector buffer.
727 	 */
728 	brelse(bp, BC_AGE);
729 	bp = NULL;
730 
731 	/*
732 	 * Check FSInfo.
733 	 */
734 	if (pmp->pm_fsinfo) {
735 		struct fsinfo *fp;
736 
737 		/*
738 		 * XXX	If the fsinfo block is stored on media with
739 		 *	2KB or larger sectors, is the fsinfo structure
740 		 *	padded at the end or in the middle?
741 		 */
742 		if ((error = bread(devvp, de_bn2kb(pmp, pmp->pm_fsinfo),
743 		    pmp->pm_BytesPerSec, NOCRED, 0, &bp)) != 0)
744 			goto error_exit;
745 		fp = (struct fsinfo *)bp->b_data;
746 		if (!memcmp(fp->fsisig1, "RRaA", 4)
747 		    && !memcmp(fp->fsisig2, "rrAa", 4)
748 		    && !memcmp(fp->fsisig3, "\0\0\125\252", 4)
749 		    && !memcmp(fp->fsisig4, "\0\0\125\252", 4))
750 			pmp->pm_nxtfree = getulong(fp->fsinxtfree);
751 		else
752 			pmp->pm_fsinfo = 0;
753 		brelse(bp, 0);
754 		bp = NULL;
755 	}
756 
757 	/*
758 	 * Check and validate (or perhaps invalidate?) the fsinfo structure?
759 	 * XXX
760 	 */
761 	if (pmp->pm_fsinfo) {
762 		if (pmp->pm_nxtfree == (u_long)-1)
763 			pmp->pm_fsinfo = 0;
764 	}
765 
766 	/*
767 	 * Allocate memory for the bitmap of allocated clusters, and then
768 	 * fill it in.
769 	 */
770 	pmp->pm_inusemap = malloc(((pmp->pm_maxcluster + N_INUSEBITS - 1)
771 				   / N_INUSEBITS)
772 				  * sizeof(*pmp->pm_inusemap),
773 				  M_MSDOSFSFAT, M_WAITOK);
774 
775 	/*
776 	 * fillinusemap() needs pm_devvp.
777 	 */
778 	pmp->pm_dev = dev;
779 	pmp->pm_devvp = devvp;
780 
781 	/*
782 	 * Have the inuse map filled in.
783 	 */
784 	if ((error = fillinusemap(pmp)) != 0) {
785 		DPRINTF(("fillinusemap %d\n", error));
786 		goto error_exit;
787 	}
788 
789 	/*
790 	 * If they want fat updates to be synchronous then let them suffer
791 	 * the performance degradation in exchange for the on disk copy of
792 	 * the fat being correct just about all the time.  I suppose this
793 	 * would be a good thing to turn on if the kernel is still flakey.
794 	 */
795 	if (mp->mnt_flag & MNT_SYNCHRONOUS)
796 		pmp->pm_flags |= MSDOSFSMNT_WAITONFAT;
797 
798 	/*
799 	 * Finish up.
800 	 */
801 	if (ronly)
802 		pmp->pm_flags |= MSDOSFSMNT_RONLY;
803 	else
804 		pmp->pm_fmod = 1;
805 	mp->mnt_data = pmp;
806 	mp->mnt_stat.f_fsidx.__fsid_val[0] = (long)dev;
807 	mp->mnt_stat.f_fsidx.__fsid_val[1] = makefstype(MOUNT_MSDOS);
808 	mp->mnt_stat.f_fsid = mp->mnt_stat.f_fsidx.__fsid_val[0];
809 	mp->mnt_stat.f_namemax = MSDOSFS_NAMEMAX(pmp);
810 	mp->mnt_flag |= MNT_LOCAL;
811 	mp->mnt_dev_bshift = pmp->pm_bnshift;
812 	mp->mnt_fs_bshift = pmp->pm_cnshift;
813 
814 	/*
815 	 * If we ever do quotas for DOS filesystems this would be a place
816 	 * to fill in the info in the msdosfsmount structure. You dolt,
817 	 * quotas on dos filesystems make no sense because files have no
818 	 * owners on dos filesystems. of course there is some empty space
819 	 * in the directory entry where we could put uid's and gid's.
820 	 */
821 
822 	devvp->v_specmountpoint = mp;
823 
824 	return (0);
825 
826 error_exit:
827 	fstrans_unmount(mp);
828 	if (bp)
829 		brelse(bp, BC_AGE);
830 	if (pmp) {
831 		if (pmp->pm_inusemap)
832 			free(pmp->pm_inusemap, M_MSDOSFSFAT);
833 		free(pmp, M_MSDOSFSMNT);
834 		mp->mnt_data = NULL;
835 	}
836 	return (error);
837 }
838 
839 int
840 msdosfs_start(struct mount *mp, int flags)
841 {
842 
843 	return (0);
844 }
845 
846 /*
847  * Unmount the filesystem described by mp.
848  */
849 int
850 msdosfs_unmount(struct mount *mp, int mntflags)
851 {
852 	struct msdosfsmount *pmp;
853 	int error, flags;
854 
855 	flags = 0;
856 	if (mntflags & MNT_FORCE)
857 		flags |= FORCECLOSE;
858 	if ((error = vflush(mp, NULLVP, flags)) != 0)
859 		return (error);
860 	pmp = VFSTOMSDOSFS(mp);
861 	if (pmp->pm_devvp->v_type != VBAD)
862 		pmp->pm_devvp->v_specmountpoint = NULL;
863 #ifdef MSDOSFS_DEBUG
864 	{
865 		struct vnode *vp = pmp->pm_devvp;
866 
867 		printf("msdosfs_umount(): just before calling VOP_CLOSE()\n");
868 		printf("flag %08x, usecount %d, writecount %d, holdcnt %d\n",
869 		    vp->v_vflag | vp->v_iflag | vp->v_uflag, vp->v_usecount,
870 		    vp->v_writecount, vp->v_holdcnt);
871 		printf("mount %p, op %p\n",
872 		    vp->v_mount, vp->v_op);
873 		printf("freef %p, freeb %p, mount %p\n",
874 		    vp->v_freelist.tqe_next, vp->v_freelist.tqe_prev,
875 		    vp->v_mount);
876 		printf("cleanblkhd %p, dirtyblkhd %p, numoutput %d, type %d\n",
877 		    vp->v_cleanblkhd.lh_first,
878 		    vp->v_dirtyblkhd.lh_first,
879 		    vp->v_numoutput, vp->v_type);
880 		printf("union %p, tag %d, data[0] %08x, data[1] %08x\n",
881 		    vp->v_socket, vp->v_tag,
882 		    ((u_int *)vp->v_data)[0],
883 		    ((u_int *)vp->v_data)[1]);
884 	}
885 #endif
886 	vn_lock(pmp->pm_devvp, LK_EXCLUSIVE | LK_RETRY);
887 	(void) VOP_CLOSE(pmp->pm_devvp,
888 	    pmp->pm_flags & MSDOSFSMNT_RONLY ? FREAD : FREAD|FWRITE, NOCRED);
889 	vput(pmp->pm_devvp);
890 	free(pmp->pm_inusemap, M_MSDOSFSFAT);
891 	free(pmp, M_MSDOSFSMNT);
892 	mp->mnt_data = NULL;
893 	mp->mnt_flag &= ~MNT_LOCAL;
894 	fstrans_unmount(mp);
895 	return (0);
896 }
897 
898 int
899 msdosfs_root(struct mount *mp, struct vnode **vpp)
900 {
901 	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
902 	struct denode *ndep;
903 	int error;
904 
905 #ifdef MSDOSFS_DEBUG
906 	printf("msdosfs_root(); mp %p, pmp %p\n", mp, pmp);
907 #endif
908 	if ((error = deget(pmp, MSDOSFSROOT, MSDOSFSROOT_OFS, &ndep)) != 0)
909 		return (error);
910 	*vpp = DETOV(ndep);
911 	return (0);
912 }
913 
914 int
915 msdosfs_statvfs(struct mount *mp, struct statvfs *sbp)
916 {
917 	struct msdosfsmount *pmp;
918 
919 	pmp = VFSTOMSDOSFS(mp);
920 	sbp->f_bsize = pmp->pm_bpcluster;
921 	sbp->f_frsize = sbp->f_bsize;
922 	sbp->f_iosize = pmp->pm_bpcluster;
923 	sbp->f_blocks = pmp->pm_nmbrofclusters;
924 	sbp->f_bfree = pmp->pm_freeclustercount;
925 	sbp->f_bavail = pmp->pm_freeclustercount;
926 	sbp->f_bresvd = 0;
927 	sbp->f_files = pmp->pm_RootDirEnts;			/* XXX */
928 	sbp->f_ffree = 0;	/* what to put in here? */
929 	sbp->f_favail = 0;	/* what to put in here? */
930 	sbp->f_fresvd = 0;
931 	copy_statvfs_info(sbp, mp);
932 	return (0);
933 }
934 
935 int
936 msdosfs_sync(struct mount *mp, int waitfor, kauth_cred_t cred)
937 {
938 	struct vnode *vp, *mvp;
939 	struct denode *dep;
940 	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
941 	int lk_flags, error, allerror = 0;
942 	bool is_suspending;
943 
944 	/*
945 	 * If we ever switch to not updating all of the fats all the time,
946 	 * this would be the place to update them from the first one.
947 	 */
948 	if (pmp->pm_fmod != 0) {
949 		if (pmp->pm_flags & MSDOSFSMNT_RONLY)
950 			panic("msdosfs_sync: rofs mod");
951 		else {
952 			/* update fats here */
953 		}
954 	}
955 	/* Allocate a marker vnode. */
956 	if ((mvp = vnalloc(mp)) == NULL)
957 		return ENOMEM;
958 	fstrans_start(mp, FSTRANS_SHARED);
959 	is_suspending = (fstrans_getstate(mp) == FSTRANS_SUSPENDING);
960 	/*
961 	 * We can't lock vnodes while the file system is suspending because
962 	 * threads waiting on fstrans may have locked vnodes.
963 	 */
964 	if (is_suspending)
965 		lk_flags = 0;
966 	else
967 		lk_flags = LK_EXCLUSIVE | LK_NOWAIT;
968 	/*
969 	 * Write back each (modified) denode.
970 	 */
971 	mutex_enter(&mntvnode_lock);
972 loop:
973 	for (vp = TAILQ_FIRST(&mp->mnt_vnodelist); vp; vp = vunmark(mvp)) {
974 		vmark(mvp, vp);
975 		if (vp->v_mount != mp || vismarker(vp))
976 			continue;
977 		mutex_enter(&vp->v_interlock);
978 		dep = VTODE(vp);
979 		if (waitfor == MNT_LAZY || vp->v_type == VNON ||
980 		    (((dep->de_flag &
981 		    (DE_ACCESS | DE_CREATE | DE_UPDATE | DE_MODIFIED)) == 0) &&
982 		     (LIST_EMPTY(&vp->v_dirtyblkhd) &&
983 		      UVM_OBJ_IS_CLEAN(&vp->v_uobj)))) {
984 			mutex_exit(&vp->v_interlock);
985 			continue;
986 		}
987 		mutex_exit(&mntvnode_lock);
988 		error = vget(vp, lk_flags);
989 		if (error) {
990 			mutex_enter(&mntvnode_lock);
991 			if (error == ENOENT) {
992 				(void)vunmark(mvp);
993 				goto loop;
994 			}
995 			continue;
996 		}
997 		if ((error = VOP_FSYNC(vp, cred,
998 		    waitfor == MNT_WAIT ? FSYNC_WAIT : 0, 0, 0)) != 0)
999 			allerror = error;
1000 		if (is_suspending)
1001 			vrele(vp);
1002 		else
1003 			vput(vp);
1004 		mutex_enter(&mntvnode_lock);
1005 	}
1006 	mutex_exit(&mntvnode_lock);
1007 	vnfree(mvp);
1008 
1009 	/*
1010 	 * Force stale file system control information to be flushed.
1011 	 */
1012 	if ((error = VOP_FSYNC(pmp->pm_devvp, cred,
1013 	    waitfor == MNT_WAIT ? FSYNC_WAIT : 0, 0, 0)) != 0)
1014 		allerror = error;
1015 	fstrans_done(mp);
1016 	return (allerror);
1017 }
1018 
1019 int
1020 msdosfs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
1021 {
1022 	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
1023 	struct defid defh;
1024 	struct denode *dep;
1025 	int error;
1026 
1027 	if (fhp->fid_len != sizeof(struct defid)) {
1028 		DPRINTF(("fid_len %d %zd\n", fhp->fid_len,
1029 		    sizeof(struct defid)));
1030 		return EINVAL;
1031 	}
1032 
1033 	memcpy(&defh, fhp, sizeof(defh));
1034 	error = deget(pmp, defh.defid_dirclust, defh.defid_dirofs, &dep);
1035 	if (error) {
1036 		DPRINTF(("deget %d\n", error));
1037 		*vpp = NULLVP;
1038 		return (error);
1039 	}
1040 	*vpp = DETOV(dep);
1041 	return (0);
1042 }
1043 
1044 int
1045 msdosfs_vptofh(struct vnode *vp, struct fid *fhp, size_t *fh_size)
1046 {
1047 	struct denode *dep;
1048 	struct defid defh;
1049 
1050 	if (*fh_size < sizeof(struct defid)) {
1051 		*fh_size = sizeof(struct defid);
1052 		return E2BIG;
1053 	}
1054 	*fh_size = sizeof(struct defid);
1055 	dep = VTODE(vp);
1056 	memset(&defh, 0, sizeof(defh));
1057 	defh.defid_len = sizeof(struct defid);
1058 	defh.defid_dirclust = dep->de_dirclust;
1059 	defh.defid_dirofs = dep->de_diroffset;
1060 	/* defh.defid_gen = dep->de_gen; */
1061 	memcpy(fhp, &defh, sizeof(defh));
1062 	return (0);
1063 }
1064 
1065 int
1066 msdosfs_vget(struct mount *mp, ino_t ino,
1067     struct vnode **vpp)
1068 {
1069 
1070 	return (EOPNOTSUPP);
1071 }
1072 
1073 int
1074 msdosfs_suspendctl(struct mount *mp, int cmd)
1075 {
1076 	int error;
1077 	struct lwp *l = curlwp;
1078 
1079 	switch (cmd) {
1080 	case SUSPEND_SUSPEND:
1081 		if ((error = fstrans_setstate(mp, FSTRANS_SUSPENDING)) != 0)
1082 			return error;
1083 		error = msdosfs_sync(mp, MNT_WAIT, l->l_proc->p_cred);
1084 		if (error == 0)
1085 			error = fstrans_setstate(mp, FSTRANS_SUSPENDED);
1086 		if (error != 0) {
1087 			(void) fstrans_setstate(mp, FSTRANS_NORMAL);
1088 			return error;
1089 		}
1090 		return 0;
1091 
1092 	case SUSPEND_RESUME:
1093 		return fstrans_setstate(mp, FSTRANS_NORMAL);
1094 
1095 	default:
1096 		return EINVAL;
1097 	}
1098 }
1099