xref: /titanic_44/usr/src/uts/common/fs/hsfs/hsfs_vfsops.c (revision 643e2e74e1c00e6b3d1896a6a67dbdb7308135c3)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 /*
22  * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 
26 #pragma ident	"%Z%%M%	%I%	%E% SMI"
27 
28 /*
29  * VFS operations for High Sierra filesystem
30  */
31 
32 #include <sys/types.h>
33 #include <sys/isa_defs.h>
34 #include <sys/t_lock.h>
35 #include <sys/param.h>
36 #include <sys/systm.h>
37 #include <sys/sysmacros.h>
38 #include <sys/kmem.h>
39 #include <sys/signal.h>
40 #include <sys/user.h>
41 #include <sys/proc.h>
42 #include <sys/disp.h>
43 #include <sys/buf.h>
44 #include <sys/pathname.h>
45 #include <sys/vfs.h>
46 #include <sys/vfs_opreg.h>
47 #include <sys/vnode.h>
48 #include <sys/file.h>
49 #include <sys/uio.h>
50 #include <sys/conf.h>
51 #include <sys/policy.h>
52 
53 #include <vm/page.h>
54 
55 #include <sys/fs/snode.h>
56 #include <sys/fs/hsfs_spec.h>
57 #include <sys/fs/hsfs_isospec.h>
58 #include <sys/fs/hsfs_node.h>
59 #include <sys/fs/hsfs_impl.h>
60 #include <sys/fs/hsfs_susp.h>
61 #include <sys/fs/hsfs_rrip.h>
62 
63 #include <sys/statvfs.h>
64 #include <sys/mount.h>
65 #include <sys/mntent.h>
66 #include <sys/swap.h>
67 #include <sys/errno.h>
68 #include <sys/debug.h>
69 #include "fs/fs_subr.h"
70 #include <sys/cmn_err.h>
71 #include <sys/bootconf.h>
72 
73 #include <sys/sdt.h>
74 
75 /*
76  * These are needed for the CDROMREADOFFSET Code
77  */
78 #include <sys/cdio.h>
79 #include <sys/sunddi.h>
80 
81 #define	HSFS_CLKSET
82 
83 #include <sys/modctl.h>
84 
85 /*
86  * Options for mount.
87  */
88 #define	HOPT_GLOBAL	MNTOPT_GLOBAL
89 #define	HOPT_NOGLOBAL	MNTOPT_NOGLOBAL
90 #define	HOPT_MAPLCASE	"maplcase"
91 #define	HOPT_NOMAPLCASE	"nomaplcase"
92 #define	HOPT_NOTRAILDOT	"notraildot"
93 #define	HOPT_TRAILDOT	"traildot"
94 #define	HOPT_NRR	"nrr"
95 #define	HOPT_RR		"rr"
96 #define	HOPT_JOLIET	"joliet"
97 #define	HOPT_NOJOLIET	"nojoliet"
98 #define	HOPT_JOLIETLONG	"jolietlong"
99 #define	HOPT_VERS2	"vers2"
100 #define	HOPT_NOVERS2	"novers2"
101 #define	HOPT_RO		MNTOPT_RO
102 
103 static char *global_cancel[] = { HOPT_NOGLOBAL, NULL };
104 static char *noglobal_cancel[] = { HOPT_GLOBAL, NULL };
105 static char *mapl_cancel[] = { HOPT_NOMAPLCASE, NULL };
106 static char *nomapl_cancel[] = { HOPT_MAPLCASE, NULL };
107 static char *ro_cancel[] = { MNTOPT_RW, NULL };
108 static char *rr_cancel[] = { HOPT_NRR, NULL };
109 static char *nrr_cancel[] = { HOPT_RR, NULL };
110 static char *joliet_cancel[] = { HOPT_NOJOLIET, NULL };
111 static char *nojoliet_cancel[] = { HOPT_JOLIET, NULL };
112 static char *vers2_cancel[] = { HOPT_NOVERS2, NULL };
113 static char *novers2_cancel[] = { HOPT_VERS2, NULL };
114 static char *trail_cancel[] = { HOPT_NOTRAILDOT, NULL };
115 static char *notrail_cancel[] = { HOPT_TRAILDOT, NULL };
116 
117 static mntopt_t hsfs_options[] = {
118 	{ HOPT_GLOBAL, global_cancel, NULL, 0, NULL },
119 	{ HOPT_NOGLOBAL, noglobal_cancel, NULL, MO_DEFAULT, NULL },
120 	{ HOPT_MAPLCASE, mapl_cancel, NULL, MO_DEFAULT, NULL },
121 	{ HOPT_NOMAPLCASE, nomapl_cancel, NULL, 0, NULL },
122 	{ HOPT_RO, ro_cancel, NULL, MO_DEFAULT, NULL },
123 	{ HOPT_RR, rr_cancel, NULL, MO_DEFAULT, NULL },
124 	{ HOPT_NRR, nrr_cancel, NULL, 0, NULL },
125 	{ HOPT_JOLIET, joliet_cancel, NULL, 0, NULL },
126 	{ HOPT_NOJOLIET, nojoliet_cancel, NULL, 0, NULL },
127 	{ HOPT_JOLIETLONG, NULL, NULL, 0, NULL },
128 	{ HOPT_VERS2, vers2_cancel, NULL, 0, NULL },
129 	{ HOPT_NOVERS2, novers2_cancel, NULL, 0, NULL },
130 	{ HOPT_TRAILDOT, trail_cancel, NULL, MO_DEFAULT, NULL },
131 	{ HOPT_NOTRAILDOT, notrail_cancel, NULL, 0, NULL },
132 	{ "sector", NULL, "0", MO_HASVALUE, NULL},
133 };
134 
135 static mntopts_t hsfs_proto_opttbl = {
136 	sizeof (hsfs_options) / sizeof (mntopt_t),
137 	hsfs_options
138 };
139 
140 /*
141  * Indicates whether to enable the I/O scheduling and readahead logic
142  * 1 - Enable, 0 - Do not Enable.
143  * Debugging purposes.
144  */
145 int do_schedio = 1;
146 static int hsfsfstype;
147 static int hsfsinit(int, char *);
148 
149 static vfsdef_t vfw = {
150 	VFSDEF_VERSION,
151 	"hsfs",
152 	hsfsinit,
153 	/* We don't suppport remounting */
154 	VSW_HASPROTO|VSW_STATS|VSW_CANLOFI,
155 	&hsfs_proto_opttbl
156 };
157 
158 static struct modlfs modlfs = {
159 	&mod_fsops, "filesystem for HSFS", &vfw
160 };
161 
162 static struct modlinkage modlinkage = {
163 	MODREV_1, (void *)&modlfs, NULL
164 };
165 
166 char _depends_on[] = "fs/specfs";
167 
168 extern void hsched_init_caches(void);
169 extern void hsched_fini_caches(void);
170 
171 
172 int
173 _init(void)
174 {
175 	return (mod_install(&modlinkage));
176 }
177 
178 int
179 _fini(void)
180 {
181 	int	error;
182 
183 	error = mod_remove(&modlinkage);
184 
185 	DTRACE_PROBE1(mod_remove, int, error);
186 
187 	if (error)
188 		return (error);
189 
190 	mutex_destroy(&hs_mounttab_lock);
191 
192 	/*
193 	 * Tear down the operations vectors
194 	 */
195 	(void) vfs_freevfsops_by_type(hsfsfstype);
196 	vn_freevnodeops(hsfs_vnodeops);
197 
198 	hs_fini_hsnode_cache();
199 	hsched_fini_caches();
200 	return (0);
201 }
202 
203 int
204 _info(struct modinfo *modinfop)
205 {
206 	return (mod_info(&modlinkage, modinfop));
207 }
208 
209 #define	BDEVFLAG(dev)	((devopsp[getmajor(dev)])->devo_cb_ops->cb_flag)
210 
211 kmutex_t hs_mounttab_lock;
212 struct hsfs *hs_mounttab = NULL;
213 
214 /* default mode, uid, gid */
215 mode_t hsfs_default_mode = 0555;
216 uid_t hsfs_default_uid = 0;
217 gid_t hsfs_default_gid = 3;
218 
219 extern void hsched_init(struct hsfs *fsp, int fsid,
220 					struct modlinkage *modlinkage);
221 extern void hsched_fini(struct hsfs_queue *hqueue);
222 extern void hsfs_init_kstats(struct hsfs *fsp, int fsid);
223 extern void hsfs_fini_kstats(struct hsfs *fsp);
224 
225 static int hsfs_mount(struct vfs *vfsp, struct vnode *mvp,
226 	struct mounta *uap, struct cred *cr);
227 static int hsfs_unmount(struct vfs *vfsp, int, struct cred *cr);
228 static int hsfs_root(struct vfs *vfsp, struct vnode **vpp);
229 static int hsfs_statvfs(struct vfs *vfsp, struct statvfs64 *sbp);
230 static int hsfs_vget(struct vfs *vfsp, struct vnode **vpp, struct fid *fidp);
231 static int hsfs_mountroot(struct vfs *, enum whymountroot);
232 
233 static int hs_mountfs(struct vfs *vfsp, dev_t dev, char *path,
234 	mode_t mode, int flags, struct cred *cr, int isroot);
235 static int hs_getrootvp(struct vfs *vfsp, struct hsfs *fsp, size_t pathsize);
236 static int hs_findhsvol(struct hsfs *fsp, struct vnode *vp,
237 	struct hs_volume *hvp);
238 static int hs_parsehsvol(struct hsfs *fsp, uchar_t *volp,
239 	struct hs_volume *hvp);
240 static int hs_findisovol(struct hsfs *fsp, struct vnode *vp,
241 	struct hs_volume *hvp,
242 	struct hs_volume *svp,
243 	struct hs_volume *jvp);
244 static int hs_joliet_level(uchar_t *volp);
245 static int hs_parseisovol(struct hsfs *fsp, uchar_t *volp,
246 	struct hs_volume *hvp);
247 static void hs_copylabel(struct hs_volume *, unsigned char *, int);
248 static int hs_getmdev(struct vfs *, char *fspec, int flags, dev_t *pdev,
249 	mode_t *mode, cred_t *cr);
250 static int hs_findvoldesc(dev_t rdev, int desc_sec);
251 
252 static int
253 hsfsinit(int fstype, char *name)
254 {
255 	static const fs_operation_def_t hsfs_vfsops_template[] = {
256 		VFSNAME_MOUNT,		{ .vfs_mount = hsfs_mount },
257 		VFSNAME_UNMOUNT,	{ .vfs_unmount = hsfs_unmount },
258 		VFSNAME_ROOT,		{ .vfs_root = hsfs_root },
259 		VFSNAME_STATVFS,	{ .vfs_statvfs = hsfs_statvfs },
260 		VFSNAME_VGET,		{ .vfs_vget = hsfs_vget },
261 		VFSNAME_MOUNTROOT,	{ .vfs_mountroot = hsfs_mountroot },
262 		NULL,			NULL
263 	};
264 	int error;
265 
266 	error = vfs_setfsops(fstype, hsfs_vfsops_template, NULL);
267 	if (error != 0) {
268 		cmn_err(CE_WARN, "hsfsinit: bad vfs ops template");
269 		return (error);
270 	}
271 
272 	error = vn_make_ops(name, hsfs_vnodeops_template, &hsfs_vnodeops);
273 	if (error != 0) {
274 		(void) vfs_freevfsops_by_type(fstype);
275 		cmn_err(CE_WARN, "hsfsinit: bad vnode ops template");
276 		return (error);
277 	}
278 
279 	hsfsfstype = fstype;
280 	mutex_init(&hs_mounttab_lock, NULL, MUTEX_DEFAULT, NULL);
281 	hs_init_hsnode_cache();
282 	hsched_init_caches();
283 	return (0);
284 }
285 
286 /*ARGSUSED*/
287 static int
288 hsfs_mount(struct vfs *vfsp, struct vnode *mvp,
289     struct mounta *uap, struct cred *cr)
290 {
291 	int		vnode_busy;
292 	dev_t		dev;
293 	struct pathname dpn;
294 	int		error;
295 	mode_t		mode;
296 	int		flags;	/* this will hold the mount specific data */
297 
298 	if ((error = secpolicy_fs_mount(cr, mvp, vfsp)) != 0)
299 		return (error);
300 
301 	if (mvp->v_type != VDIR)
302 		return (ENOTDIR);
303 
304 	/* mount option must be read only, else mount will be rejected */
305 	if (!(uap->flags & MS_RDONLY))
306 		return (EROFS);
307 
308 	/*
309 	 * We already told the framework that we don't support remounting.
310 	 */
311 	ASSERT(!(uap->flags & MS_REMOUNT));
312 
313 	mutex_enter(&mvp->v_lock);
314 	vnode_busy = (mvp->v_count != 1) || (mvp->v_flag & VROOT);
315 	mutex_exit(&mvp->v_lock);
316 
317 	if ((uap->flags & MS_OVERLAY) == 0 && vnode_busy) {
318 		return (EBUSY);
319 	}
320 
321 	/*
322 	 * Check for the options that actually affect things
323 	 * at our level.
324 	 */
325 	flags = 0;
326 	if (vfs_optionisset(vfsp, HOPT_NOMAPLCASE, NULL))
327 		flags |= HSFSMNT_NOMAPLCASE;
328 	if (vfs_optionisset(vfsp, HOPT_NOTRAILDOT, NULL))
329 		flags |= HSFSMNT_NOTRAILDOT;
330 	if (vfs_optionisset(vfsp, HOPT_NRR, NULL))
331 		flags |= HSFSMNT_NORRIP;
332 	if (vfs_optionisset(vfsp, HOPT_NOJOLIET, NULL))
333 		flags |= HSFSMNT_NOJOLIET;
334 	if (vfs_optionisset(vfsp, HOPT_JOLIETLONG, NULL))
335 		flags |= HSFSMNT_JOLIETLONG;
336 	if (vfs_optionisset(vfsp, HOPT_NOVERS2, NULL))
337 		flags |= HSFSMNT_NOVERS2;
338 
339 	error = pn_get(uap->dir, (uap->flags & MS_SYSSPACE) ?
340 	    UIO_SYSSPACE : UIO_USERSPACE, &dpn);
341 	if (error)
342 		return (error);
343 
344 	error = hs_getmdev(vfsp, uap->spec, uap->flags, &dev, &mode, cr);
345 	if (error != 0) {
346 		pn_free(&dpn);
347 		return (error);
348 	}
349 
350 	/*
351 	 * If the device is a tape, return error
352 	 */
353 	if ((BDEVFLAG(dev) & D_TAPE) == D_TAPE)  {
354 		pn_free(&dpn);
355 		return (ENOTBLK);
356 	}
357 
358 	/*
359 	 * Mount the filesystem.
360 	 */
361 	error = hs_mountfs(vfsp, dev, dpn.pn_path, mode, flags, cr, 0);
362 	pn_free(&dpn);
363 	return (error);
364 }
365 
366 /*ARGSUSED*/
367 static int
368 hsfs_unmount(
369 	struct vfs *vfsp,
370 	int flag,
371 	struct cred *cr)
372 {
373 	struct hsfs **tspp;
374 	struct hsfs *fsp;
375 
376 	if (secpolicy_fs_unmount(cr, vfsp) != 0)
377 		return (EPERM);
378 
379 	/*
380 	 * forced unmount is not supported by this file system
381 	 * and thus, ENOTSUP is being returned.
382 	 */
383 	if (flag & MS_FORCE)
384 		return (ENOTSUP);
385 
386 	fsp = VFS_TO_HSFS(vfsp);
387 
388 	if (fsp->hsfs_rootvp->v_count != 1)
389 		return (EBUSY);
390 
391 	/* destroy all old pages and hsnodes for this vfs */
392 	if (hs_synchash(vfsp))
393 		return (EBUSY);
394 
395 	mutex_enter(&hs_mounttab_lock);
396 	for (tspp = &hs_mounttab; *tspp != NULL; tspp = &(*tspp)->hsfs_next) {
397 		if (*tspp == fsp)
398 			break;
399 	}
400 	if (*tspp == NULL) {
401 		mutex_exit(&hs_mounttab_lock);
402 		panic("hsfs_unmount: vfs not mounted?");
403 		/*NOTREACHED*/
404 	}
405 
406 	*tspp = fsp->hsfs_next;
407 
408 	mutex_exit(&hs_mounttab_lock);
409 
410 	hsfs_fini_kstats(fsp);
411 	(void) VOP_CLOSE(fsp->hsfs_devvp, FREAD, 1, (offset_t)0, cr, NULL);
412 	VN_RELE(fsp->hsfs_devvp);
413 	/* free path table space */
414 	if (fsp->hsfs_ptbl != NULL)
415 		kmem_free(fsp->hsfs_ptbl, (size_t)fsp->hsfs_vol.ptbl_len);
416 	/* free path table index table */
417 	if (fsp->hsfs_ptbl_idx != NULL)
418 		kmem_free(fsp->hsfs_ptbl_idx, (size_t)
419 		    (fsp->hsfs_ptbl_idx_size * sizeof (struct ptable_idx)));
420 
421 	/* free "mounted on" pathame */
422 	if (fsp->hsfs_fsmnt != NULL)
423 		kmem_free(fsp->hsfs_fsmnt, strlen(fsp->hsfs_fsmnt) + 1);
424 
425 	hsched_fini(fsp->hqueue);
426 	kmem_free(fsp->hqueue, sizeof (struct hsfs_queue));
427 
428 	mutex_destroy(&fsp->hsfs_free_lock);
429 	rw_destroy(&fsp->hsfs_hash_lock);
430 
431 	kmem_free(fsp, sizeof (*fsp));
432 	return (0);
433 }
434 
435 /*ARGSUSED*/
436 static int
437 hsfs_root(struct vfs *vfsp, struct vnode **vpp)
438 {
439 	*vpp = (VFS_TO_HSFS(vfsp))->hsfs_rootvp;
440 	VN_HOLD(*vpp);
441 	return (0);
442 }
443 
444 /*ARGSUSED*/
445 static int
446 hsfs_statvfs(struct vfs *vfsp, struct statvfs64 *sbp)
447 {
448 	struct hsfs *fsp;
449 	dev32_t d32;
450 
451 	fsp = VFS_TO_HSFS(vfsp);
452 	if (fsp->hsfs_magic != HSFS_MAGIC)
453 		return (EINVAL);
454 	bzero(sbp, sizeof (*sbp));
455 	sbp->f_bsize = vfsp->vfs_bsize;
456 	sbp->f_frsize = sbp->f_bsize; /* no fragment, same as block size */
457 	sbp->f_blocks = (fsblkcnt64_t)fsp->hsfs_vol.vol_size;
458 
459 	sbp->f_bfree = (fsblkcnt64_t)0;
460 	sbp->f_bavail = (fsblkcnt64_t)0;
461 	sbp->f_files = (fsfilcnt64_t)-1;
462 	sbp->f_ffree = (fsfilcnt64_t)0;
463 	sbp->f_favail = (fsfilcnt64_t)0;
464 	(void) cmpldev(&d32, vfsp->vfs_dev);
465 	sbp->f_fsid = d32;
466 	(void) strcpy(sbp->f_basetype, vfssw[vfsp->vfs_fstype].vsw_name);
467 	sbp->f_flag = vf_to_stf(vfsp->vfs_flag);
468 	sbp->f_namemax = fsp->hsfs_namemax;
469 	(void) strcpy(sbp->f_fstr, fsp->hsfs_vol.vol_id);
470 
471 	return (0);
472 }
473 
474 /*
475  * Previously nodeid was declared as uint32_t. This has been changed
476  * to conform better with the ISO9660 standard. The standard states that
477  * a LBN can be a 32 bit number, as the MAKE_NODEID macro shifts this
478  * LBN 11 places left (LBN_TO_BYTE) and then shifts the result 5 right
479  * (divide by 32) we are left with the potential of an overflow if
480  * confined to a 32 bit value.
481  */
482 
483 static int
484 hsfs_vget(struct vfs *vfsp, struct vnode **vpp, struct fid *fidp)
485 {
486 	struct hsfid *fid;
487 	struct hsfs *fsp;
488 	ino64_t nodeid;
489 	int error;
490 
491 	fsp = (struct hsfs *)VFS_TO_HSFS(vfsp);
492 	fid = (struct hsfid *)fidp;
493 
494 	/*
495 	 * Look for vnode on hashlist.
496 	 * If found, it's now active and the refcnt was incremented.
497 	 */
498 
499 	rw_enter(&fsp->hsfs_hash_lock, RW_READER);
500 
501 	nodeid = fid->hf_ino;
502 
503 	if ((*vpp = hs_findhash(nodeid, fid->hf_dir_lbn,
504 	    (uint_t)fid->hf_dir_off, vfsp)) == NULL) {
505 		/*
506 		 * Not in cache, so we need to remake it.
507 		 * hs_remakenode() will read the directory entry
508 		 * and then check again to see if anyone else has
509 		 * put it in the cache.
510 		 */
511 		rw_exit(&fsp->hsfs_hash_lock);
512 		error = hs_remakenode(fid->hf_dir_lbn, (uint_t)fid->hf_dir_off,
513 		    vfsp, vpp);
514 		return (error);
515 	}
516 	rw_exit(&fsp->hsfs_hash_lock);
517 	return (0);
518 }
519 
520 
521 #define	CHECKSUM_SIZE				(64 * 1024)
522 
523 /*
524  * Compute a CD-ROM fsid by checksumming the first 64K of data on the CD
525  * We use the 'fsp' argument to determine the location of the root
526  * directory entry, and we start reading from there.
527  */
528 static int
529 compute_cdrom_id(struct hsfs *fsp, vnode_t *devvp)
530 {
531 	uint_t		secno;
532 	struct hs_volume *hsvp = &fsp->hsfs_vol;
533 	struct buf	*bp;
534 	int		error;
535 	int		fsid;
536 
537 	secno = hsvp->root_dir.ext_lbn >> hsvp->lbn_secshift;
538 	bp = bread(devvp->v_rdev, secno * 4, CHECKSUM_SIZE);
539 	error = geterror(bp);
540 
541 	/*
542 	 * An error on read or a partial read means we asked
543 	 * for a nonexistant/corrupted piece of the device
544 	 * (including past-the-end of the media). Don't
545 	 * try to use the checksumming method then.
546 	 */
547 	if (!error && bp->b_bcount == CHECKSUM_SIZE) {
548 		int *ibuf = (int *)bp->b_un.b_addr;
549 		int i;
550 
551 		fsid = 0;
552 
553 		for (i = 0; i < CHECKSUM_SIZE / sizeof (int); i++)
554 			fsid ^= ibuf[ i ];
555 	} else {
556 		/*
557 		 * Fallback - use creation date
558 		 */
559 		fsid = hsvp->cre_date.tv_sec;
560 	}
561 
562 	brelse(bp);
563 
564 	return (fsid);
565 }
566 
567 
568 /*ARGSUSED*/
569 static int
570 hs_mountfs(
571 	struct vfs	*vfsp,
572 	dev_t		dev,
573 	char		*path,
574 	mode_t		mode,
575 	int		mount_flags,
576 	struct cred	*cr,
577 	int		isroot)
578 {
579 	struct vnode	*devvp;
580 	struct hsfs	*tsp;
581 	struct hsfs	*fsp = NULL;
582 	struct vattr	vap;
583 	struct hsnode	*hp;
584 	int		error;
585 	struct timeval	tv;
586 	int		fsid;
587 	int		use_rrip;
588 	int		use_vers2;
589 	int		use_joliet;
590 	int		has_rrip = 0;
591 	int		has_vers2 = 0;
592 	int		has_joliet = 0;
593 	int		force_rrip_off;
594 	int		force_vers2_off;
595 	int		force_joliet_off;
596 	size_t		pathbufsz = strlen(path) + 1;
597 	int		redo_rootvp;
598 
599 	struct hs_volume *svp = NULL;	/* Supplemental VD for ISO-9660:1999 */
600 	struct hs_volume *jvp = NULL;	/* Joliet VD */
601 
602 	/*
603 	 * The rules for which extension will be used are:
604 	 * 1. No specific mount options given:
605 	 *	- use rrip if available
606 	 *	- use ISO9660:1999 if available
607 	 *	- use joliet if available.
608 	 * 2. rrip/ISO9660:1999/joliet explicitly disabled via mount option:
609 	 *	- use next "lower" extension
610 	 * 3. joliet/ISO9660:1999/rrip explicitly requested via mount option:
611 	 *	- disable rrip support even if available
612 	 *	- disable IOS9660:1999 support even if available
613 	 *
614 	 * We need to adjust these flags as we discover the extensions
615 	 * present. See below. These are just the starting values.
616 	 */
617 	use_rrip = (mount_flags & HSFSMNT_NORRIP) == 0;
618 	use_vers2 = (mount_flags & HSFSMNT_NOVERS2) == 0;
619 	use_joliet = (mount_flags & HSFSMNT_NOJOLIET) == 0;
620 
621 	/*
622 	 * Open the device
623 	 */
624 	devvp = makespecvp(dev, VBLK);
625 	ASSERT(devvp != 0);
626 
627 	/*
628 	 * Open the target device (file) for read only.
629 	 */
630 	if (error = VOP_OPEN(&devvp, FREAD, cr, NULL)) {
631 		VN_RELE(devvp);
632 		return (error);
633 	}
634 
635 	/*
636 	 * Refuse to go any further if this
637 	 * device is being used for swapping
638 	 */
639 	if (IS_SWAPVP(common_specvp(devvp))) {
640 		error = EBUSY;
641 		goto cleanup;
642 	}
643 
644 	vap.va_mask = AT_SIZE;
645 	if ((error = VOP_GETATTR(devvp, &vap, ATTR_COMM, cr, NULL)) != 0) {
646 		cmn_err(CE_NOTE, "Cannot get attributes of the CD-ROM driver");
647 		goto cleanup;
648 	}
649 
650 	/*
651 	 * Make sure we have a nonzero size partition.
652 	 * The current version of the SD driver will *not* fail the open
653 	 * of such a partition so we have to check for it here.
654 	 */
655 	if (vap.va_size == 0) {
656 		error = ENXIO;
657 		goto cleanup;
658 	}
659 
660 	/*
661 	 * Init a new hsfs structure.
662 	 */
663 	fsp = kmem_zalloc(sizeof (*fsp), KM_SLEEP);
664 	svp = kmem_zalloc(sizeof (*svp), KM_SLEEP);
665 	jvp = kmem_zalloc(sizeof (*jvp), KM_SLEEP);
666 
667 	/* hardwire perms, uid, gid */
668 	fsp->hsfs_vol.vol_uid = hsfs_default_uid;
669 	fsp->hsfs_vol.vol_gid =  hsfs_default_gid;
670 	fsp->hsfs_vol.vol_prot = hsfs_default_mode;
671 	svp->vol_uid = hsfs_default_uid;
672 	svp->vol_gid =  hsfs_default_gid;
673 	svp->vol_prot = hsfs_default_mode;
674 	jvp->vol_uid = hsfs_default_uid;
675 	jvp->vol_gid =  hsfs_default_gid;
676 	jvp->vol_prot = hsfs_default_mode;
677 
678 	/*
679 	 * Look for a Standard File Structure Volume Descriptor,
680 	 * of which there must be at least one.
681 	 * If found, check for volume size consistency.
682 	 *
683 	 * If svp->lbn_size is != 0, we did find a ISO-9660:1999 SVD
684 	 * If jvp->lbn_size is != 0, we did find a Joliet SVD.
685 	 */
686 	fsp->hsfs_namemax = ISO_FILE_NAMELEN;
687 	fsp->hsfs_namelen = ISO_FILE_NAMELEN;
688 	error = hs_findisovol(fsp, devvp, &fsp->hsfs_vol, svp, jvp);
689 	if (error == EINVAL) /* no iso 9660 - try high sierra ... */
690 		error = hs_findhsvol(fsp, devvp, &fsp->hsfs_vol);
691 
692 	if (error)
693 		goto cleanup;
694 
695 	DTRACE_PROBE4(findvol,
696 	    struct hsfs *, fsp,
697 	    struct hs_volume *, &fsp->hsfs_vol,
698 	    struct hs_volume *, svp,
699 	    struct hs_volume *, jvp);
700 
701 	/*
702 	 * Generate a file system ID from the CD-ROM,
703 	 * and check it for uniqueness.
704 	 *
705 	 * What we are aiming for is some chance of integrity
706 	 * across disk change.  That is, if a client has an fhandle,
707 	 * it will be valid as long as the same disk is mounted.
708 	 */
709 	fsid = compute_cdrom_id(fsp, devvp);
710 
711 	mutex_enter(&hs_mounttab_lock);
712 
713 	if (fsid == 0 || fsid == -1) {
714 		uniqtime(&tv);
715 		fsid = tv.tv_sec;
716 	} else	/* make sure that the fsid is unique */
717 		for (tsp = hs_mounttab; tsp != NULL; tsp = tsp->hsfs_next) {
718 			if (fsid == tsp->hsfs_vfs->vfs_fsid.val[0]) {
719 				uniqtime(&tv);
720 				fsid = tv.tv_sec;
721 				break;
722 			}
723 		}
724 
725 	fsp->hsfs_next = hs_mounttab;
726 	hs_mounttab = fsp;
727 
728 	fsp->hsfs_devvp = devvp;
729 	fsp->hsfs_vfs = vfsp;
730 	fsp->hsfs_fsmnt = kmem_alloc(pathbufsz, KM_SLEEP);
731 	(void) strlcpy(fsp->hsfs_fsmnt, path, pathbufsz);
732 
733 	mutex_init(&fsp->hsfs_free_lock, NULL, MUTEX_DEFAULT, NULL);
734 	rw_init(&fsp->hsfs_hash_lock, NULL, RW_DEFAULT, NULL);
735 
736 	vfsp->vfs_data = (caddr_t)fsp;
737 	vfsp->vfs_dev = dev;
738 	vfsp->vfs_fstype = hsfsfstype;
739 	vfsp->vfs_bsize = fsp->hsfs_vol.lbn_size; /* %% */
740 	vfsp->vfs_fsid.val[0] = fsid;
741 	vfsp->vfs_fsid.val[1] =  hsfsfstype;
742 
743 	if (!hs_getrootvp(vfsp, fsp, pathbufsz)) {
744 		DTRACE_PROBE1(rootvp__failed, struct hsfs *, fsp);
745 		error = EINVAL;
746 		goto cleanup;
747 	}
748 	DTRACE_PROBE1(rootvp, struct hsfs *, fsp);
749 
750 	/*
751 	 * Attempt to discover a RR extension.
752 	 */
753 	if (use_rrip) {
754 		hp = VTOH(fsp->hsfs_rootvp);
755 		hs_check_root_dirent(fsp->hsfs_rootvp, &(hp->hs_dirent));
756 	}
757 
758 	has_rrip = IS_RRIP_IMPLEMENTED(fsp);
759 	has_vers2 = (svp->lbn_size != 0);
760 	has_joliet = (jvp->lbn_size != 0);
761 
762 	DTRACE_PROBE4(voltype__suggested, struct hsfs *, fsp,
763 	    int, use_rrip, int, use_vers2, int, use_joliet);
764 
765 	DTRACE_PROBE4(voltype__actual, struct hsfs *, fsp,
766 	    int, has_rrip, int, has_vers2, int, has_joliet);
767 
768 	DTRACE_PROBE4(findvol,
769 	    struct hsfs *, fsp,
770 	    struct hs_volume *, &fsp->hsfs_vol,
771 	    struct hs_volume *, svp,
772 	    struct hs_volume *, jvp);
773 
774 	force_rrip_off = !use_rrip ||
775 	    (vfs_optionisset(vfsp, HOPT_JOLIET, NULL) && has_joliet) ||
776 	    (vfs_optionisset(vfsp, HOPT_VERS2, NULL) && has_vers2);
777 
778 	force_vers2_off = !use_vers2 ||
779 	    (vfs_optionisset(vfsp, HOPT_JOLIET, NULL) && has_joliet);
780 
781 	force_joliet_off = !use_joliet;
782 
783 	DTRACE_PROBE4(voltype__force_off, struct hsfs *, fsp,
784 	    int, force_rrip_off, int, force_vers2_off, int, force_joliet_off);
785 
786 	/*
787 	 * At the moment, we have references of all three possible
788 	 * extensions (RR, ISO9660:1999/v2 and Joliet) if present.
789 	 *
790 	 * The "active" volume descriptor is RRIP (or ISO9660:1988).
791 	 * We now switch to the user-requested one.
792 	 */
793 	redo_rootvp = 0;
794 
795 	if (force_rrip_off || !has_rrip) {
796 		if (has_vers2 && !force_vers2_off) {
797 			VN_RELE(fsp->hsfs_rootvp);
798 			bcopy(svp, &fsp->hsfs_vol, sizeof (struct hs_volume));
799 			fsp->hsfs_vol_type = HS_VOL_TYPE_ISO_V2;
800 			vfsp->vfs_bsize = fsp->hsfs_vol.lbn_size;
801 			redo_rootvp = 1;
802 			has_joliet = 0;
803 		} else if (has_joliet && !force_joliet_off) {
804 			VN_RELE(fsp->hsfs_rootvp);
805 			bcopy(jvp, &fsp->hsfs_vol, sizeof (struct hs_volume));
806 			fsp->hsfs_vol_type = HS_VOL_TYPE_JOLIET;
807 			vfsp->vfs_bsize = fsp->hsfs_vol.lbn_size;
808 			redo_rootvp = 1;
809 			has_vers2 = 0;
810 		}
811 	}
812 
813 	if (redo_rootvp) {
814 		/*
815 		 * Make sure not to use Rock Ridge.
816 		 */
817 		UNSET_IMPL_BIT(fsp, RRIP_BIT);
818 		UNSET_SUSP_BIT(fsp);
819 		has_rrip = 0;
820 
821 		if (!hs_getrootvp(vfsp, fsp, pathbufsz)) {
822 			DTRACE_PROBE1(rootvp__failed, struct hsfs *, fsp);
823 			error = EINVAL;
824 			goto cleanup;
825 		}
826 		DTRACE_PROBE1(rootvp, struct hsfs *, fsp);
827 	}
828 	if (IS_RRIP_IMPLEMENTED(fsp)) {
829 		has_vers2 = 0;
830 		has_joliet = 0;
831 	}
832 	if (force_vers2_off)
833 		has_vers2 = 0;
834 	if (force_joliet_off)
835 		has_joliet = 0;
836 	DTRACE_PROBE4(voltype__taken, struct hsfs *, fsp,
837 	    int, has_rrip, int, has_vers2, int, has_joliet);
838 
839 	/*
840 	 * mark root node as VROOT
841 	 */
842 	fsp->hsfs_rootvp->v_flag |= VROOT;
843 
844 	/* Here we take care of some special case stuff for mountroot */
845 	if (isroot) {
846 		fsp->hsfs_rootvp->v_rdev = devvp->v_rdev;
847 		rootvp = fsp->hsfs_rootvp;
848 	}
849 
850 	if (IS_RRIP_IMPLEMENTED(fsp)) {
851 		/*
852 		 * if RRIP, don't copy NOMAPLCASE or NOTRAILDOT to hsfs_flags
853 		 */
854 		mount_flags &= ~(HSFSMNT_NOMAPLCASE | HSFSMNT_NOTRAILDOT);
855 
856 		fsp->hsfs_namemax = RRIP_FILE_NAMELEN;
857 		fsp->hsfs_namelen = RRIP_FILE_NAMELEN;
858 
859 		ASSERT(vfs_optionisset(vfsp, HOPT_RR, NULL));
860 		vfs_clearmntopt(vfsp, HOPT_VERS2);
861 		vfs_clearmntopt(vfsp, HOPT_JOLIET);
862 
863 	} else switch (fsp->hsfs_vol_type) {
864 
865 	case HS_VOL_TYPE_HS:
866 	case HS_VOL_TYPE_ISO:
867 	default:
868 		/*
869 		 * if iso v1, don't allow trailing spaces in iso file names
870 		 */
871 		mount_flags |= HSFSMNT_NOTRAILSPACE;
872 		fsp->hsfs_namemax = ISO_NAMELEN_V2_MAX;
873 		fsp->hsfs_namelen = ISO_FILE_NAMELEN;
874 		vfs_clearmntopt(vfsp, HOPT_RR);
875 		vfs_clearmntopt(vfsp, HOPT_VERS2);
876 		vfs_clearmntopt(vfsp, HOPT_JOLIET);
877 		break;
878 
879 	case HS_VOL_TYPE_ISO_V2:
880 		/*
881 		 * if iso v2, don't copy NOTRAILDOT to hsfs_flags
882 		 */
883 		mount_flags &= ~HSFSMNT_NOTRAILDOT;
884 		mount_flags |= HSFSMNT_NOMAPLCASE | HSFSMNT_NOVERSION;
885 		fsp->hsfs_namemax = ISO_NAMELEN_V2_MAX;
886 		fsp->hsfs_namelen = ISO_NAMELEN_V2;
887 		vfs_setmntopt(vfsp, HOPT_VERS2, NULL, 0);
888 		vfs_clearmntopt(vfsp, HOPT_RR);
889 		vfs_clearmntopt(vfsp, HOPT_JOLIET);
890 		break;
891 
892 	case HS_VOL_TYPE_JOLIET:
893 		/*
894 		 * if Joliet, don't copy NOMAPLCASE or NOTRAILDOT to hsfs_flags
895 		 */
896 		mount_flags &= ~(HSFSMNT_NOMAPLCASE | HSFSMNT_NOTRAILDOT);
897 		mount_flags |= HSFSMNT_NOMAPLCASE;
898 		if (mount_flags & HSFSMNT_JOLIETLONG)
899 			fsp->hsfs_namemax = JOLIET_NAMELEN_MAX*3; /* UTF-8 */
900 		else
901 			fsp->hsfs_namemax = MAXNAMELEN-1;
902 		fsp->hsfs_namelen = JOLIET_NAMELEN*2;
903 		vfs_setmntopt(vfsp, HOPT_JOLIET, NULL, 0);
904 		vfs_clearmntopt(vfsp, HOPT_RR);
905 		vfs_clearmntopt(vfsp, HOPT_VERS2);
906 		break;
907 	}
908 
909 	/*
910 	 * Add the HSFSMNT_INODE pseudo mount flag to the current mount flags.
911 	 */
912 	fsp->hsfs_flags = mount_flags | (fsp->hsfs_flags & HSFSMNT_INODE);
913 
914 	/*
915 	 * Setup I/O Scheduling structures
916 	 */
917 	if (do_schedio) {
918 		fsp->hqueue = kmem_alloc(sizeof (struct hsfs_queue), KM_SLEEP);
919 		hsched_init(fsp, fsid, &modlinkage);
920 	}
921 
922 	/*
923 	 * Setup kstats
924 	 */
925 	hsfs_init_kstats(fsp, fsid);
926 
927 	DTRACE_PROBE1(mount__done, struct hsfs *, fsp);
928 
929 	/*
930 	 * set the magic word
931 	 */
932 	fsp->hsfs_magic = HSFS_MAGIC;
933 	mutex_exit(&hs_mounttab_lock);
934 
935 	kmem_free(svp, sizeof (*svp));
936 	kmem_free(jvp, sizeof (*jvp));
937 
938 	return (0);
939 
940 cleanup:
941 	(void) VOP_CLOSE(devvp, FREAD, 1, (offset_t)0, cr, NULL);
942 	VN_RELE(devvp);
943 	if (fsp)
944 		kmem_free(fsp, sizeof (*fsp));
945 	if (svp)
946 		kmem_free(svp, sizeof (*svp));
947 	if (jvp)
948 		kmem_free(jvp, sizeof (*jvp));
949 	return (error);
950 }
951 
952 /*
953  * Get the rootvp associated with fsp->hsfs_vol
954  */
955 static int
956 hs_getrootvp(
957 	struct vfs	*vfsp,
958 	struct hsfs	*fsp,
959 	size_t		pathsize)
960 {
961 	struct hsnode	*hp;
962 
963 	ASSERT(pathsize == strlen(fsp->hsfs_fsmnt) + 1);
964 
965 	/*
966 	 * If the root directory does not appear to be
967 	 * valid, use what it points to as "." instead.
968 	 * Some Defense Mapping Agency disks are non-conformant
969 	 * in this way.
970 	 */
971 	if (!hsfs_valid_dir(&fsp->hsfs_vol.root_dir)) {
972 		hs_log_bogus_disk_warning(fsp, HSFS_ERR_BAD_ROOT_DIR, 0);
973 		if (hs_remakenode(fsp->hsfs_vol.root_dir.ext_lbn,
974 		    (uint_t)0, vfsp, &fsp->hsfs_rootvp)) {
975 			hs_mounttab = hs_mounttab->hsfs_next;
976 			mutex_destroy(&fsp->hsfs_free_lock);
977 			rw_destroy(&fsp->hsfs_hash_lock);
978 			kmem_free(fsp->hsfs_fsmnt, pathsize);
979 			mutex_exit(&hs_mounttab_lock);
980 			return (0);
981 		}
982 	} else {
983 		fsp->hsfs_rootvp = hs_makenode(&fsp->hsfs_vol.root_dir,
984 		    fsp->hsfs_vol.root_dir.ext_lbn, 0, vfsp);
985 	}
986 
987 	/* XXX - ignore the path table for now */
988 	fsp->hsfs_ptbl = NULL;
989 	hp = VTOH(fsp->hsfs_rootvp);
990 	hp->hs_ptbl_idx = NULL;
991 
992 	return (1);
993 }
994 
995 /*
996  * hs_findhsvol()
997  *
998  * Locate the Standard File Structure Volume Descriptor and
999  * parse it into an hs_volume structure.
1000  *
1001  * XXX - May someday want to look for Coded Character Set FSVD, too.
1002  */
1003 static int
1004 hs_findhsvol(struct hsfs *fsp, struct vnode *vp, struct hs_volume *hvp)
1005 {
1006 	struct buf *secbp;
1007 	int i;
1008 	int n;
1009 	uchar_t *volp;
1010 	int error;
1011 	uint_t secno;
1012 
1013 	secno = hs_findvoldesc(vp->v_rdev, HS_VOLDESC_SEC);
1014 	secbp = bread(vp->v_rdev, secno * 4, HS_SECTOR_SIZE);
1015 	error = geterror(secbp);
1016 
1017 	if (error != 0) {
1018 		cmn_err(CE_NOTE, "hs_findhsvol: bread: error=(%d)", error);
1019 		brelse(secbp);
1020 		return (error);
1021 	}
1022 
1023 	volp = (uchar_t *)secbp->b_un.b_addr;
1024 
1025 	/*
1026 	 * To avoid that we read the whole medium in case that someone prepares
1027 	 * a malicious "fs image", we read at most 32 blocks.
1028 	 */
1029 	for (n = 0; n < 32 &&
1030 	    HSV_DESC_TYPE(volp) != VD_EOV; n++) {
1031 		for (i = 0; i < HSV_ID_STRLEN; i++)
1032 			if (HSV_STD_ID(volp)[i] != HSV_ID_STRING[i])
1033 				goto cantfind;
1034 		if (HSV_STD_VER(volp) != HSV_ID_VER)
1035 			goto cantfind;
1036 		switch (HSV_DESC_TYPE(volp)) {
1037 		case VD_SFS:
1038 			/* Standard File Structure */
1039 			fsp->hsfs_vol_type = HS_VOL_TYPE_HS;
1040 			error = hs_parsehsvol(fsp, volp, hvp);
1041 			brelse(secbp);
1042 			return (error);
1043 
1044 		case VD_CCFS:
1045 			/* Coded Character File Structure */
1046 		case VD_BOOT:
1047 		case VD_UNSPEC:
1048 		case VD_EOV:
1049 			break;
1050 		}
1051 		brelse(secbp);
1052 		++secno;
1053 		secbp = bread(vp->v_rdev, secno * 4, HS_SECTOR_SIZE);
1054 
1055 		error = geterror(secbp);
1056 
1057 		if (error != 0) {
1058 			cmn_err(CE_NOTE, "hs_findhsvol: bread: error=(%d)",
1059 			    error);
1060 			brelse(secbp);
1061 			return (error);
1062 		}
1063 
1064 		volp = (uchar_t *)secbp->b_un.b_addr;
1065 	}
1066 cantfind:
1067 	brelse(secbp);
1068 	return (EINVAL);
1069 }
1070 
1071 /*
1072  * hs_parsehsvol
1073  *
1074  * Parse the Standard File Structure Volume Descriptor into
1075  * an hs_volume structure.  We can't just bcopy it into the
1076  * structure because of byte-ordering problems.
1077  *
1078  */
1079 static int
1080 hs_parsehsvol(struct hsfs *fsp, uchar_t *volp, struct hs_volume *hvp)
1081 {
1082 	hvp->vol_size = HSV_VOL_SIZE(volp);
1083 	hvp->lbn_size = HSV_BLK_SIZE(volp);
1084 	if (hvp->lbn_size == 0) {
1085 		cmn_err(CE_NOTE, "hs_parsehsvol: logical block size in the "
1086 		    "SFSVD is zero");
1087 		return (EINVAL);
1088 	}
1089 	hvp->lbn_shift = ffs((long)hvp->lbn_size) - 1;
1090 	hvp->lbn_secshift =
1091 	    ffs((long)howmany(HS_SECTOR_SIZE, (int)hvp->lbn_size)) - 1;
1092 	hvp->lbn_maxoffset = hvp->lbn_size - 1;
1093 	hs_parse_longdate(HSV_cre_date(volp), &hvp->cre_date);
1094 	hs_parse_longdate(HSV_mod_date(volp), &hvp->mod_date);
1095 	hvp->file_struct_ver = HSV_FILE_STRUCT_VER(volp);
1096 	hvp->ptbl_len = HSV_PTBL_SIZE(volp);
1097 	hvp->vol_set_size = (ushort_t)HSV_SET_SIZE(volp);
1098 	hvp->vol_set_seq = (ushort_t)HSV_SET_SEQ(volp);
1099 #if defined(_LITTLE_ENDIAN)
1100 	hvp->ptbl_lbn = HSV_PTBL_MAN_LS(volp);
1101 #else
1102 	hvp->ptbl_lbn = HSV_PTBL_MAN_MS(volp);
1103 #endif
1104 	hs_copylabel(hvp, HSV_VOL_ID(volp), 0);
1105 
1106 	/*
1107 	 * Make sure that lbn_size is a power of two and otherwise valid.
1108 	 */
1109 	if (hvp->lbn_size & ~(1 << hvp->lbn_shift)) {
1110 		cmn_err(CE_NOTE,
1111 		    "hsfs: %d-byte logical block size not supported",
1112 		    hvp->lbn_size);
1113 		return (EINVAL);
1114 	}
1115 	return (hs_parsedir(fsp, HSV_ROOT_DIR(volp), &hvp->root_dir,
1116 	    (char *)NULL, (int *)NULL, HDE_ROOT_DIR_REC_SIZE));
1117 }
1118 
1119 /*
1120  * hs_findisovol()
1121  *
1122  * Locate the Primary Volume Descriptor
1123  * parse it into an hs_volume structure.
1124  *
1125  * XXX - Partition not yet done
1126  *
1127  * Except for fsp->hsfs_vol_type, no fsp member may be modified.
1128  * fsp->hsfs_vol is modified indirectly via the *hvp argument.
1129  */
1130 static int
1131 hs_findisovol(struct hsfs *fsp, struct vnode *vp,
1132     struct hs_volume *hvp,
1133     struct hs_volume *svp,
1134     struct hs_volume *jvp)
1135 {
1136 	struct buf *secbp;
1137 	int i;
1138 	int n;
1139 	uchar_t *volp;
1140 	int error;
1141 	uint_t secno;
1142 	int foundpvd = 0;
1143 	int foundsvd = 0;
1144 	int foundjvd = 0;
1145 	int pvd_sum = 0;
1146 
1147 	secno = hs_findvoldesc(vp->v_rdev, ISO_VOLDESC_SEC);
1148 	secbp = bread(vp->v_rdev, secno * 4, ISO_SECTOR_SIZE);
1149 	error = geterror(secbp);
1150 
1151 	if (error != 0) {
1152 		cmn_err(CE_NOTE, "hs_findisovol: bread: error=(%d)", error);
1153 		brelse(secbp);
1154 		return (error);
1155 	}
1156 
1157 	volp = (uchar_t *)secbp->b_un.b_addr;
1158 
1159 	/*
1160 	 * To avoid that we read the whole medium in case that someone prepares
1161 	 * a malicious "fs image", we read at most 32 blocks.
1162 	 */
1163 	for (n = 0; n < 32 &&
1164 	    (enum iso_voldesc_type) ISO_DESC_TYPE(volp) != ISO_VD_EOV; n++) {
1165 		for (i = 0; i < ISO_ID_STRLEN; i++)
1166 			if (ISO_STD_ID(volp)[i] != ISO_ID_STRING[i])
1167 				goto cantfind;
1168 		switch (ISO_DESC_TYPE(volp)) {
1169 		case ISO_VD_PVD:
1170 			/* Standard File Structure */
1171 			if (ISO_STD_VER(volp) != ISO_ID_VER)
1172 				goto cantfind;
1173 			if (foundpvd != 1) {
1174 				fsp->hsfs_vol_type = HS_VOL_TYPE_ISO;
1175 				if (error = hs_parseisovol(fsp, volp, hvp)) {
1176 					brelse(secbp);
1177 					return (error);
1178 				}
1179 				foundpvd = 1;
1180 				for (i = 0; i < ISO_SECTOR_SIZE; i++)
1181 					pvd_sum += volp[i];
1182 			}
1183 			break;
1184 		case ISO_VD_SVD:
1185 			/* Supplementary Volume Descriptor */
1186 			if (ISO_STD_VER(volp) == ISO_ID_VER2 &&
1187 			    foundsvd != 1) {
1188 				fsp->hsfs_vol_type = HS_VOL_TYPE_ISO;
1189 				if (error = hs_parseisovol(fsp, volp, svp)) {
1190 					brelse(secbp);
1191 					return (error);
1192 				}
1193 				foundsvd = 1;
1194 			}
1195 			if (hs_joliet_level(volp) >= 1 && foundjvd != 1) {
1196 				fsp->hsfs_vol_type = HS_VOL_TYPE_ISO;
1197 				if (error = hs_parseisovol(fsp, volp, jvp)) {
1198 					brelse(secbp);
1199 					return (error);
1200 				}
1201 				foundjvd = 1;
1202 			}
1203 			break;
1204 		case ISO_VD_BOOT:
1205 			break;
1206 		case ISO_VD_VPD:
1207 			/* currently cannot handle partition */
1208 			break;
1209 		case VD_EOV:
1210 			break;
1211 		}
1212 		brelse(secbp);
1213 		++secno;
1214 		secbp = bread(vp->v_rdev, secno * 4, HS_SECTOR_SIZE);
1215 		error = geterror(secbp);
1216 
1217 		if (error != 0) {
1218 			cmn_err(CE_NOTE, "hs_findisovol: bread: error=(%d)",
1219 			    error);
1220 			brelse(secbp);
1221 			return (error);
1222 		}
1223 
1224 		volp = (uchar_t *)secbp->b_un.b_addr;
1225 	}
1226 	for (n = 0; n < 16; n++) {
1227 		brelse(secbp);
1228 		++secno;
1229 		secbp = bread(vp->v_rdev, secno * 4, HS_SECTOR_SIZE);
1230 		error = geterror(secbp);
1231 
1232 		if (error != 0) {
1233 			cmn_err(CE_NOTE, "hs_findisovol: bread: error=(%d)",
1234 			    error);
1235 			brelse(secbp);
1236 			return (error);
1237 		}
1238 
1239 		/*
1240 		 * Check for the signature from mkisofs that grants that
1241 		 * the current filesystem allows to use the extent lbn as
1242 		 * inode number even in pure ISO9660 mode.
1243 		 */
1244 		volp = (uchar_t *)secbp->b_un.b_addr;
1245 		if (strncmp((char *)volp, "MKI ", 4) == 0) {
1246 			int	sum;
1247 
1248 			sum  = volp[2045];
1249 			sum *= 256;
1250 			sum += volp[2046];
1251 			sum *= 256;
1252 			sum += volp[2047];
1253 			if (sum == pvd_sum)
1254 				fsp->hsfs_flags |= HSFSMNT_INODE;
1255 			break;
1256 		}
1257 	}
1258 	if (foundpvd) {
1259 		brelse(secbp);
1260 		return (0);
1261 	}
1262 cantfind:
1263 	brelse(secbp);
1264 	return (EINVAL);
1265 }
1266 
1267 /*
1268  * Return 0 if no Joliet is found
1269  * else return Joliet Level 1..3
1270  */
1271 static int
1272 hs_joliet_level(uchar_t *volp)
1273 {
1274 	if (ISO_std_ver(volp)[0] == ISO_ID_VER &&
1275 	    ISO_svd_esc(volp)[0] == '%' &&
1276 	    ISO_svd_esc(volp)[1] == '/') {
1277 
1278 		switch (ISO_svd_esc(volp)[2]) {
1279 
1280 		case '@':
1281 			return (1);
1282 		case 'C':
1283 			return (2);
1284 		case 'E':
1285 			return (3);
1286 		}
1287 	}
1288 	return (0);
1289 }
1290 
1291 /*
1292  * hs_parseisovol
1293  *
1294  * Parse the Primary Volume Descriptor into an hs_volume structure.
1295  *
1296  */
1297 static int
1298 hs_parseisovol(struct hsfs *fsp, uchar_t *volp, struct hs_volume *hvp)
1299 {
1300 	hvp->vol_size = ISO_VOL_SIZE(volp);
1301 	hvp->lbn_size = ISO_BLK_SIZE(volp);
1302 	if (hvp->lbn_size == 0) {
1303 		cmn_err(CE_NOTE, "hs_parseisovol: logical block size in the "
1304 		    "PVD is zero");
1305 		return (EINVAL);
1306 	}
1307 	hvp->lbn_shift = ffs((long)hvp->lbn_size) - 1;
1308 	hvp->lbn_secshift =
1309 	    ffs((long)howmany(ISO_SECTOR_SIZE, (int)hvp->lbn_size)) - 1;
1310 	hvp->lbn_maxoffset = hvp->lbn_size - 1;
1311 	hs_parse_longdate(ISO_cre_date(volp), &hvp->cre_date);
1312 	hs_parse_longdate(ISO_mod_date(volp), &hvp->mod_date);
1313 	hvp->file_struct_ver = ISO_FILE_STRUCT_VER(volp);
1314 	hvp->ptbl_len = ISO_PTBL_SIZE(volp);
1315 	hvp->vol_set_size = (ushort_t)ISO_SET_SIZE(volp);
1316 	hvp->vol_set_seq = (ushort_t)ISO_SET_SEQ(volp);
1317 #if defined(_LITTLE_ENDIAN)
1318 	hvp->ptbl_lbn = ISO_PTBL_MAN_LS(volp);
1319 #else
1320 	hvp->ptbl_lbn = ISO_PTBL_MAN_MS(volp);
1321 #endif
1322 	hs_copylabel(hvp, ISO_VOL_ID(volp), hs_joliet_level(volp) >= 1);
1323 
1324 	/*
1325 	 * Make sure that lbn_size is a power of two and otherwise valid.
1326 	 */
1327 	if (hvp->lbn_size & ~(1 << hvp->lbn_shift)) {
1328 		cmn_err(CE_NOTE,
1329 		    "hsfs: %d-byte logical block size not supported",
1330 		    hvp->lbn_size);
1331 		return (EINVAL);
1332 	}
1333 	return (hs_parsedir(fsp, ISO_ROOT_DIR(volp), &hvp->root_dir,
1334 	    (char *)NULL, (int *)NULL, IDE_ROOT_DIR_REC_SIZE));
1335 }
1336 
1337 /*
1338  * Common code for mount and umount.
1339  * Check that the user's argument is a reasonable
1340  * thing on which to mount, and return the device number if so.
1341  */
1342 static int
1343 hs_getmdev(struct vfs *vfsp, char *fspec, int flags, dev_t *pdev, mode_t *mode,
1344     cred_t *cr)
1345 {
1346 	int error;
1347 	struct vnode *svp = NULL;
1348 	struct vnode *lvp = NULL;
1349 	struct vnode *bvp;
1350 	struct vattr vap;
1351 	dev_t dev;
1352 	enum uio_seg fromspace = (flags & MS_SYSSPACE) ?
1353 	    UIO_SYSSPACE : UIO_USERSPACE;
1354 
1355 	/*
1356 	 * Look up the device/file to be mounted.
1357 	 */
1358 	error = lookupname(fspec, fromspace, FOLLOW, NULLVPP, &svp);
1359 	if (error) {
1360 		if (error == ENOENT)
1361 			error = ENODEV;
1362 		goto out;
1363 	}
1364 
1365 	error = vfs_get_lofi(vfsp, &lvp);
1366 
1367 	if (error > 0) {
1368 		if (error == ENOENT)
1369 			error = ENODEV;
1370 		goto out;
1371 	} else if (error == 0) {
1372 		bvp = lvp;
1373 	} else {
1374 		bvp = svp;
1375 
1376 		if (bvp->v_type != VBLK) {
1377 			error = ENOTBLK;
1378 			goto out;
1379 		}
1380 
1381 		if ((error = secpolicy_spec_open(cr, bvp, FREAD)) != 0)
1382 			goto out;
1383 	}
1384 
1385 	/*
1386 	 * Can we read from the device/file ?
1387 	 */
1388 	if ((error = VOP_ACCESS(svp, VREAD, 0, cr, NULL)) != 0)
1389 		goto out;
1390 
1391 	vap.va_mask = AT_MODE;		/* get protection mode */
1392 	(void) VOP_GETATTR(bvp, &vap, 0, CRED(), NULL);
1393 	*mode = vap.va_mode;
1394 
1395 	dev = *pdev = bvp->v_rdev;
1396 
1397 	error = EBUSY;
1398 
1399 	/*
1400 	 * Ensure that this device isn't already mounted,
1401 	 * unless this is a REMOUNT request or we are told to suppress
1402 	 * mount checks.
1403 	 */
1404 	if ((flags & MS_NOCHECK) == 0) {
1405 		if (vfs_devmounting(dev, vfsp))
1406 			goto out;
1407 		if (vfs_devismounted(dev) && !(flags & MS_REMOUNT))
1408 			goto out;
1409 	}
1410 
1411 	if (getmajor(*pdev) >= devcnt) {
1412 		error = ENXIO;
1413 		goto out;
1414 	}
1415 
1416 	error = 0;
1417 out:
1418 	if (svp != NULL)
1419 		VN_RELE(svp);
1420 	if (lvp != NULL)
1421 		VN_RELE(lvp);
1422 	return (error);
1423 }
1424 
1425 static void
1426 hs_copylabel(struct hs_volume *hvp, unsigned char *label, int isjoliet)
1427 {
1428 	char	lbuf[64];	/* hs_joliet_cp() creates 48 bytes at most */
1429 
1430 	if (isjoliet) {
1431 		/*
1432 		 * hs_joliet_cp() will output 16..48 bytes.
1433 		 * We need to clear 'lbuf' to avoid junk chars past byte 15.
1434 		 */
1435 		bzero(lbuf, sizeof (lbuf));
1436 		(void) hs_joliet_cp((char *)label, lbuf, 32);
1437 		label = (unsigned char *)lbuf;
1438 	}
1439 	/* cdrom volid is at most 32 bytes */
1440 	bcopy(label, hvp->vol_id, 32);
1441 	hvp->vol_id[31] = NULL;
1442 }
1443 
1444 /*
1445  * Mount root file system.
1446  * "why" is ROOT_INIT on initial call, ROOT_REMOUNT if called to
1447  * remount the root file system, and ROOT_UNMOUNT if called to
1448  * unmount the root (e.g., as part of a system shutdown).
1449  *
1450  * XXX - this may be partially machine-dependent; it, along with the VFS_SWAPVP
1451  * operation, goes along with auto-configuration.  A mechanism should be
1452  * provided by which machine-INdependent code in the kernel can say "get me the
1453  * right root file system" and "get me the right initial swap area", and have
1454  * that done in what may well be a machine-dependent fashion.
1455  * Unfortunately, it is also file-system-type dependent (NFS gets it via
1456  * bootparams calls, UFS gets it from various and sundry machine-dependent
1457  * mechanisms, as SPECFS does for swap).
1458  */
1459 static int
1460 hsfs_mountroot(struct vfs *vfsp, enum whymountroot why)
1461 {
1462 	int error;
1463 	struct hsfs *fsp;
1464 	struct hs_volume *fvolp;
1465 	static int hsfsrootdone = 0;
1466 	dev_t rootdev;
1467 	mode_t mode = 0;
1468 
1469 	if (why == ROOT_INIT) {
1470 		if (hsfsrootdone++)
1471 			return (EBUSY);
1472 		rootdev = getrootdev();
1473 		if (rootdev == (dev_t)NODEV)
1474 			return (ENODEV);
1475 		vfsp->vfs_dev = rootdev;
1476 		vfsp->vfs_flag |= VFS_RDONLY;
1477 	} else if (why == ROOT_REMOUNT) {
1478 		cmn_err(CE_NOTE, "hsfs_mountroot: ROOT_REMOUNT");
1479 		return (0);
1480 	} else if (why == ROOT_UNMOUNT) {
1481 		return (0);
1482 	}
1483 	error = vfs_lock(vfsp);
1484 	if (error) {
1485 		cmn_err(CE_NOTE, "hsfs_mountroot: couldn't get vfs_lock");
1486 		return (error);
1487 	}
1488 
1489 	error = hs_mountfs(vfsp, rootdev, "/", mode, 1, CRED(), 1);
1490 	/*
1491 	 * XXX - assumes root device is not indirect, because we don't set
1492 	 * rootvp.  Is rootvp used for anything?  If so, make another arg
1493 	 * to mountfs.
1494 	 */
1495 	if (error) {
1496 		vfs_unlock(vfsp);
1497 		if (rootvp) {
1498 			VN_RELE(rootvp);
1499 			rootvp = (struct vnode *)0;
1500 		}
1501 		return (error);
1502 	}
1503 	if (why == ROOT_INIT)
1504 		vfs_add((struct vnode *)0, vfsp,
1505 		    (vfsp->vfs_flag & VFS_RDONLY) ? MS_RDONLY : 0);
1506 	vfs_unlock(vfsp);
1507 	fsp = VFS_TO_HSFS(vfsp);
1508 	fvolp = &fsp->hsfs_vol;
1509 #ifdef HSFS_CLKSET
1510 	if (fvolp->cre_date.tv_sec == 0) {
1511 		cmn_err(CE_NOTE, "hsfs_mountroot: cre_date.tv_sec == 0");
1512 		if (fvolp->mod_date.tv_sec == 0) {
1513 			cmn_err(CE_NOTE,
1514 			    "hsfs_mountroot: mod_date.tv_sec == 0");
1515 			cmn_err(CE_NOTE, "hsfs_mountroot: clkset(-1L)");
1516 			clkset(-1L);
1517 		} else {
1518 			clkset(fvolp->mod_date.tv_sec);
1519 		}
1520 	} else {
1521 		clkset(fvolp->mod_date.tv_sec);
1522 	}
1523 #else	/* HSFS_CLKSET */
1524 	clkset(-1L);
1525 #endif	/* HSFS_CLKSET */
1526 	return (0);
1527 }
1528 
1529 /*
1530  * hs_findvoldesc()
1531  *
1532  * Return the sector where the volume descriptor lives.  This is
1533  * a fixed value for "normal" cd-rom's, but can change for
1534  * multisession cd's.
1535  *
1536  * desc_sec is the same for high-sierra and iso 9660 formats, why
1537  * there are two different #defines used in the code for this is
1538  * beyond me.  These are standards, cast in concrete, right?
1539  * To be general, however, this function supports passing in different
1540  * values.
1541  */
1542 static int
1543 hs_findvoldesc(dev_t rdev, int desc_sec)
1544 {
1545 	int secno;
1546 	int error;
1547 	int rval;	/* ignored */
1548 
1549 #ifdef CDROMREADOFFSET
1550 	/*
1551 	 * Issue the Read Offset ioctl directly to the
1552 	 * device. Ignore any errors and set starting
1553 	 * secno to the default, otherwise add the
1554 	 * VOLDESC sector number to the offset.
1555 	 */
1556 	error = cdev_ioctl(rdev, CDROMREADOFFSET, (intptr_t)&secno,
1557 	    FNATIVE|FKIOCTL|FREAD, CRED(), &rval);
1558 	if (error) {
1559 		secno = desc_sec;
1560 	} else {
1561 		secno += desc_sec;
1562 	}
1563 #else
1564 	secno = desc_sec;
1565 #endif
1566 
1567 	return (secno);
1568 }
1569