xref: /titanic_51/usr/src/uts/common/fs/hsfs/hsfs_vfsops.c (revision 05ead181677a01a3a118f8b89ce79361113e34cf)
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 (c) 1990, 2010, Oracle and/or its affiliates. All rights reserved.
23  */
24 
25 /*
26  * VFS operations for High Sierra filesystem
27  */
28 
29 #include <sys/types.h>
30 #include <sys/isa_defs.h>
31 #include <sys/t_lock.h>
32 #include <sys/param.h>
33 #include <sys/systm.h>
34 #include <sys/sysmacros.h>
35 #include <sys/kmem.h>
36 #include <sys/signal.h>
37 #include <sys/user.h>
38 #include <sys/proc.h>
39 #include <sys/disp.h>
40 #include <sys/buf.h>
41 #include <sys/pathname.h>
42 #include <sys/vfs.h>
43 #include <sys/vfs_opreg.h>
44 #include <sys/vnode.h>
45 #include <sys/file.h>
46 #include <sys/uio.h>
47 #include <sys/conf.h>
48 #include <sys/policy.h>
49 
50 #include <vm/page.h>
51 
52 #include <sys/fs/snode.h>
53 #include <sys/fs/hsfs_spec.h>
54 #include <sys/fs/hsfs_isospec.h>
55 #include <sys/fs/hsfs_node.h>
56 #include <sys/fs/hsfs_impl.h>
57 #include <sys/fs/hsfs_susp.h>
58 #include <sys/fs/hsfs_rrip.h>
59 
60 #include <sys/statvfs.h>
61 #include <sys/mount.h>
62 #include <sys/mntent.h>
63 #include <sys/swap.h>
64 #include <sys/errno.h>
65 #include <sys/debug.h>
66 #include "fs/fs_subr.h"
67 #include <sys/cmn_err.h>
68 #include <sys/bootconf.h>
69 
70 #include <sys/sdt.h>
71 
72 /*
73  * These are needed for the CDROMREADOFFSET Code
74  */
75 #include <sys/cdio.h>
76 #include <sys/sunddi.h>
77 
78 #define	HSFS_CLKSET
79 
80 #include <sys/modctl.h>
81 
82 /*
83  * Options for mount.
84  */
85 #define	HOPT_GLOBAL	MNTOPT_GLOBAL
86 #define	HOPT_NOGLOBAL	MNTOPT_NOGLOBAL
87 #define	HOPT_MAPLCASE	"maplcase"
88 #define	HOPT_NOMAPLCASE	"nomaplcase"
89 #define	HOPT_NOTRAILDOT	"notraildot"
90 #define	HOPT_TRAILDOT	"traildot"
91 #define	HOPT_NRR	"nrr"
92 #define	HOPT_RR		"rr"
93 #define	HOPT_JOLIET	"joliet"
94 #define	HOPT_NOJOLIET	"nojoliet"
95 #define	HOPT_JOLIETLONG	"jolietlong"
96 #define	HOPT_VERS2	"vers2"
97 #define	HOPT_NOVERS2	"novers2"
98 #define	HOPT_RO		MNTOPT_RO
99 
100 static char *global_cancel[] = { HOPT_NOGLOBAL, NULL };
101 static char *noglobal_cancel[] = { HOPT_GLOBAL, NULL };
102 static char *mapl_cancel[] = { HOPT_NOMAPLCASE, NULL };
103 static char *nomapl_cancel[] = { HOPT_MAPLCASE, NULL };
104 static char *ro_cancel[] = { MNTOPT_RW, NULL };
105 static char *rr_cancel[] = { HOPT_NRR, NULL };
106 static char *nrr_cancel[] = { HOPT_RR, NULL };
107 static char *joliet_cancel[] = { HOPT_NOJOLIET, NULL };
108 static char *nojoliet_cancel[] = { HOPT_JOLIET, NULL };
109 static char *vers2_cancel[] = { HOPT_NOVERS2, NULL };
110 static char *novers2_cancel[] = { HOPT_VERS2, NULL };
111 static char *trail_cancel[] = { HOPT_NOTRAILDOT, NULL };
112 static char *notrail_cancel[] = { HOPT_TRAILDOT, NULL };
113 
114 static mntopt_t hsfs_options[] = {
115 	{ HOPT_GLOBAL, global_cancel, NULL, 0, NULL },
116 	{ HOPT_NOGLOBAL, noglobal_cancel, NULL, MO_DEFAULT, NULL },
117 	{ HOPT_MAPLCASE, mapl_cancel, NULL, MO_DEFAULT, NULL },
118 	{ HOPT_NOMAPLCASE, nomapl_cancel, NULL, 0, NULL },
119 	{ HOPT_RO, ro_cancel, NULL, MO_DEFAULT, NULL },
120 	{ HOPT_RR, rr_cancel, NULL, MO_DEFAULT, NULL },
121 	{ HOPT_NRR, nrr_cancel, NULL, 0, NULL },
122 	{ HOPT_JOLIET, joliet_cancel, NULL, 0, NULL },
123 	{ HOPT_NOJOLIET, nojoliet_cancel, NULL, 0, NULL },
124 	{ HOPT_JOLIETLONG, NULL, NULL, 0, NULL },
125 	{ HOPT_VERS2, vers2_cancel, NULL, 0, NULL },
126 	{ HOPT_NOVERS2, novers2_cancel, NULL, 0, NULL },
127 	{ HOPT_TRAILDOT, trail_cancel, NULL, MO_DEFAULT, NULL },
128 	{ HOPT_NOTRAILDOT, notrail_cancel, NULL, 0, NULL },
129 	{ "sector", NULL, "0", MO_HASVALUE, NULL},
130 };
131 
132 static mntopts_t hsfs_proto_opttbl = {
133 	sizeof (hsfs_options) / sizeof (mntopt_t),
134 	hsfs_options
135 };
136 
137 /*
138  * Indicates whether to enable the I/O scheduling and readahead logic
139  * 1 - Enable, 0 - Do not Enable.
140  * Debugging purposes.
141  */
142 int do_schedio = 1;
143 static int hsfsfstype;
144 static int hsfsinit(int, char *);
145 
146 static vfsdef_t vfw = {
147 	VFSDEF_VERSION,
148 	"hsfs",
149 	hsfsinit,
150 	/* We don't suppport remounting */
151 	VSW_HASPROTO|VSW_STATS|VSW_CANLOFI|VSW_ZMOUNT,
152 	&hsfs_proto_opttbl
153 };
154 
155 static struct modlfs modlfs = {
156 	&mod_fsops, "filesystem for HSFS", &vfw
157 };
158 
159 static struct modlinkage modlinkage = {
160 	MODREV_1, (void *)&modlfs, NULL
161 };
162 
163 char _depends_on[] = "fs/specfs";
164 
165 extern void hsched_init_caches(void);
166 extern void hsched_fini_caches(void);
167 
168 
169 int
170 _init(void)
171 {
172 	return (mod_install(&modlinkage));
173 }
174 
175 int
176 _fini(void)
177 {
178 	int	error;
179 
180 	error = mod_remove(&modlinkage);
181 
182 	DTRACE_PROBE1(mod_remove, int, error);
183 
184 	if (error)
185 		return (error);
186 
187 	mutex_destroy(&hs_mounttab_lock);
188 
189 	/*
190 	 * Tear down the operations vectors
191 	 */
192 	(void) vfs_freevfsops_by_type(hsfsfstype);
193 	vn_freevnodeops(hsfs_vnodeops);
194 
195 	hs_fini_hsnode_cache();
196 	hsched_fini_caches();
197 	return (0);
198 }
199 
200 int
201 _info(struct modinfo *modinfop)
202 {
203 	return (mod_info(&modlinkage, modinfop));
204 }
205 
206 #define	BDEVFLAG(dev)	((devopsp[getmajor(dev)])->devo_cb_ops->cb_flag)
207 
208 kmutex_t hs_mounttab_lock;
209 struct hsfs *hs_mounttab = NULL;
210 
211 /* default mode, uid, gid */
212 mode_t hsfs_default_mode = 0555;
213 uid_t hsfs_default_uid = 0;
214 gid_t hsfs_default_gid = 3;
215 
216 extern void hsched_init(struct hsfs *fsp, int fsid,
217 					struct modlinkage *modlinkage);
218 extern void hsched_fini(struct hsfs_queue *hqueue);
219 extern void hsfs_init_kstats(struct hsfs *fsp, int fsid);
220 extern void hsfs_fini_kstats(struct hsfs *fsp);
221 
222 static int hsfs_mount(struct vfs *vfsp, struct vnode *mvp,
223 	struct mounta *uap, struct cred *cr);
224 static int hsfs_unmount(struct vfs *vfsp, int, struct cred *cr);
225 static int hsfs_root(struct vfs *vfsp, struct vnode **vpp);
226 static int hsfs_statvfs(struct vfs *vfsp, struct statvfs64 *sbp);
227 static int hsfs_vget(struct vfs *vfsp, struct vnode **vpp, struct fid *fidp);
228 static int hsfs_mountroot(struct vfs *, enum whymountroot);
229 
230 static int hs_mountfs(struct vfs *vfsp, dev_t dev, char *path,
231 	mode_t mode, int flags, struct cred *cr, int isroot);
232 static int hs_getrootvp(struct vfs *vfsp, struct hsfs *fsp, size_t pathsize);
233 static int hs_findhsvol(struct hsfs *fsp, struct vnode *vp,
234 	struct hs_volume *hvp);
235 static int hs_parsehsvol(struct hsfs *fsp, uchar_t *volp,
236 	struct hs_volume *hvp);
237 static int hs_findisovol(struct hsfs *fsp, struct vnode *vp,
238 	struct hs_volume *hvp,
239 	struct hs_volume *svp,
240 	struct hs_volume *jvp);
241 static int hs_joliet_level(uchar_t *volp);
242 static int hs_parseisovol(struct hsfs *fsp, uchar_t *volp,
243 	struct hs_volume *hvp);
244 static void hs_copylabel(struct hs_volume *, unsigned char *, int);
245 static int hs_getmdev(struct vfs *, char *fspec, int flags, dev_t *pdev,
246 	mode_t *mode, cred_t *cr);
247 static int hs_findvoldesc(dev_t rdev, int desc_sec);
248 
249 static int
250 hsfsinit(int fstype, char *name)
251 {
252 	static const fs_operation_def_t hsfs_vfsops_template[] = {
253 		VFSNAME_MOUNT,		{ .vfs_mount = hsfs_mount },
254 		VFSNAME_UNMOUNT,	{ .vfs_unmount = hsfs_unmount },
255 		VFSNAME_ROOT,		{ .vfs_root = hsfs_root },
256 		VFSNAME_STATVFS,	{ .vfs_statvfs = hsfs_statvfs },
257 		VFSNAME_VGET,		{ .vfs_vget = hsfs_vget },
258 		VFSNAME_MOUNTROOT,	{ .vfs_mountroot = hsfs_mountroot },
259 		NULL,			NULL
260 	};
261 	int error;
262 
263 	error = vfs_setfsops(fstype, hsfs_vfsops_template, NULL);
264 	if (error != 0) {
265 		cmn_err(CE_WARN, "hsfsinit: bad vfs ops template");
266 		return (error);
267 	}
268 
269 	error = vn_make_ops(name, hsfs_vnodeops_template, &hsfs_vnodeops);
270 	if (error != 0) {
271 		(void) vfs_freevfsops_by_type(fstype);
272 		cmn_err(CE_WARN, "hsfsinit: bad vnode ops template");
273 		return (error);
274 	}
275 
276 	hsfsfstype = fstype;
277 	mutex_init(&hs_mounttab_lock, NULL, MUTEX_DEFAULT, NULL);
278 	hs_init_hsnode_cache();
279 	hsched_init_caches();
280 	return (0);
281 }
282 
283 /*ARGSUSED*/
284 static int
285 hsfs_mount(struct vfs *vfsp, struct vnode *mvp,
286     struct mounta *uap, struct cred *cr)
287 {
288 	int		vnode_busy;
289 	dev_t		dev;
290 	struct pathname dpn;
291 	int		error;
292 	mode_t		mode;
293 	int		flags;	/* this will hold the mount specific data */
294 
295 	if ((error = secpolicy_fs_mount(cr, mvp, vfsp)) != 0)
296 		return (error);
297 
298 	if (mvp->v_type != VDIR)
299 		return (ENOTDIR);
300 
301 	/* mount option must be read only, else mount will be rejected */
302 	if (!(uap->flags & MS_RDONLY))
303 		return (EROFS);
304 
305 	/*
306 	 * We already told the framework that we don't support remounting.
307 	 */
308 	ASSERT(!(uap->flags & MS_REMOUNT));
309 
310 	mutex_enter(&mvp->v_lock);
311 	vnode_busy = (mvp->v_count != 1) || (mvp->v_flag & VROOT);
312 	mutex_exit(&mvp->v_lock);
313 
314 	if ((uap->flags & MS_OVERLAY) == 0 && vnode_busy) {
315 		return (EBUSY);
316 	}
317 
318 	/*
319 	 * Check for the options that actually affect things
320 	 * at our level.
321 	 */
322 	flags = 0;
323 	if (vfs_optionisset(vfsp, HOPT_NOMAPLCASE, NULL))
324 		flags |= HSFSMNT_NOMAPLCASE;
325 	if (vfs_optionisset(vfsp, HOPT_NOTRAILDOT, NULL))
326 		flags |= HSFSMNT_NOTRAILDOT;
327 	if (vfs_optionisset(vfsp, HOPT_NRR, NULL))
328 		flags |= HSFSMNT_NORRIP;
329 	if (vfs_optionisset(vfsp, HOPT_NOJOLIET, NULL))
330 		flags |= HSFSMNT_NOJOLIET;
331 	if (vfs_optionisset(vfsp, HOPT_JOLIETLONG, NULL))
332 		flags |= HSFSMNT_JOLIETLONG;
333 	if (vfs_optionisset(vfsp, HOPT_NOVERS2, NULL))
334 		flags |= HSFSMNT_NOVERS2;
335 
336 	error = pn_get(uap->dir, (uap->flags & MS_SYSSPACE) ?
337 	    UIO_SYSSPACE : UIO_USERSPACE, &dpn);
338 	if (error)
339 		return (error);
340 
341 	error = hs_getmdev(vfsp, uap->spec, uap->flags, &dev, &mode, cr);
342 	if (error != 0) {
343 		pn_free(&dpn);
344 		return (error);
345 	}
346 
347 	/*
348 	 * If the device is a tape, return error
349 	 */
350 	if ((BDEVFLAG(dev) & D_TAPE) == D_TAPE)  {
351 		pn_free(&dpn);
352 		return (ENOTBLK);
353 	}
354 
355 	/*
356 	 * Mount the filesystem.
357 	 */
358 	error = hs_mountfs(vfsp, dev, dpn.pn_path, mode, flags, cr, 0);
359 	pn_free(&dpn);
360 	return (error);
361 }
362 
363 /*ARGSUSED*/
364 static int
365 hsfs_unmount(
366 	struct vfs *vfsp,
367 	int flag,
368 	struct cred *cr)
369 {
370 	struct hsfs **tspp;
371 	struct hsfs *fsp;
372 
373 	if (secpolicy_fs_unmount(cr, vfsp) != 0)
374 		return (EPERM);
375 
376 	/*
377 	 * forced unmount is not supported by this file system
378 	 * and thus, ENOTSUP is being returned.
379 	 */
380 	if (flag & MS_FORCE)
381 		return (ENOTSUP);
382 
383 	fsp = VFS_TO_HSFS(vfsp);
384 
385 	if (fsp->hsfs_rootvp->v_count != 1)
386 		return (EBUSY);
387 
388 	/* destroy all old pages and hsnodes for this vfs */
389 	if (hs_synchash(vfsp))
390 		return (EBUSY);
391 
392 	mutex_enter(&hs_mounttab_lock);
393 	for (tspp = &hs_mounttab; *tspp != NULL; tspp = &(*tspp)->hsfs_next) {
394 		if (*tspp == fsp)
395 			break;
396 	}
397 	if (*tspp == NULL) {
398 		mutex_exit(&hs_mounttab_lock);
399 		panic("hsfs_unmount: vfs not mounted?");
400 		/*NOTREACHED*/
401 	}
402 
403 	*tspp = fsp->hsfs_next;
404 
405 	mutex_exit(&hs_mounttab_lock);
406 
407 	hsfs_fini_kstats(fsp);
408 	(void) VOP_CLOSE(fsp->hsfs_devvp, FREAD, 1, (offset_t)0, cr, NULL);
409 	VN_RELE(fsp->hsfs_devvp);
410 	/* free path table space */
411 	if (fsp->hsfs_ptbl != NULL)
412 		kmem_free(fsp->hsfs_ptbl, (size_t)fsp->hsfs_vol.ptbl_len);
413 	/* free path table index table */
414 	if (fsp->hsfs_ptbl_idx != NULL)
415 		kmem_free(fsp->hsfs_ptbl_idx, (size_t)
416 		    (fsp->hsfs_ptbl_idx_size * sizeof (struct ptable_idx)));
417 
418 	/* free "mounted on" pathame */
419 	if (fsp->hsfs_fsmnt != NULL)
420 		kmem_free(fsp->hsfs_fsmnt, strlen(fsp->hsfs_fsmnt) + 1);
421 
422 	hsched_fini(fsp->hqueue);
423 	kmem_free(fsp->hqueue, sizeof (struct hsfs_queue));
424 
425 	mutex_destroy(&fsp->hsfs_free_lock);
426 	rw_destroy(&fsp->hsfs_hash_lock);
427 
428 	kmem_free(fsp, sizeof (*fsp));
429 	return (0);
430 }
431 
432 /*ARGSUSED*/
433 static int
434 hsfs_root(struct vfs *vfsp, struct vnode **vpp)
435 {
436 	*vpp = (VFS_TO_HSFS(vfsp))->hsfs_rootvp;
437 	VN_HOLD(*vpp);
438 	return (0);
439 }
440 
441 /*ARGSUSED*/
442 static int
443 hsfs_statvfs(struct vfs *vfsp, struct statvfs64 *sbp)
444 {
445 	struct hsfs *fsp;
446 	dev32_t d32;
447 
448 	fsp = VFS_TO_HSFS(vfsp);
449 	if (fsp->hsfs_magic != HSFS_MAGIC)
450 		return (EINVAL);
451 	bzero(sbp, sizeof (*sbp));
452 	sbp->f_bsize = vfsp->vfs_bsize;
453 	sbp->f_frsize = sbp->f_bsize; /* no fragment, same as block size */
454 	sbp->f_blocks = (fsblkcnt64_t)fsp->hsfs_vol.vol_size;
455 
456 	sbp->f_bfree = (fsblkcnt64_t)0;
457 	sbp->f_bavail = (fsblkcnt64_t)0;
458 	sbp->f_files = (fsfilcnt64_t)-1;
459 	sbp->f_ffree = (fsfilcnt64_t)0;
460 	sbp->f_favail = (fsfilcnt64_t)0;
461 	(void) cmpldev(&d32, vfsp->vfs_dev);
462 	sbp->f_fsid = d32;
463 	(void) strcpy(sbp->f_basetype, vfssw[vfsp->vfs_fstype].vsw_name);
464 	sbp->f_flag = vf_to_stf(vfsp->vfs_flag);
465 	sbp->f_namemax = fsp->hsfs_namemax;
466 	(void) strcpy(sbp->f_fstr, fsp->hsfs_vol.vol_id);
467 
468 	return (0);
469 }
470 
471 /*
472  * Previously nodeid was declared as uint32_t. This has been changed
473  * to conform better with the ISO9660 standard. The standard states that
474  * a LBN can be a 32 bit number, as the MAKE_NODEID macro shifts this
475  * LBN 11 places left (LBN_TO_BYTE) and then shifts the result 5 right
476  * (divide by 32) we are left with the potential of an overflow if
477  * confined to a 32 bit value.
478  */
479 
480 static int
481 hsfs_vget(struct vfs *vfsp, struct vnode **vpp, struct fid *fidp)
482 {
483 	struct hsfid *fid;
484 	struct hsfs *fsp;
485 	ino64_t nodeid;
486 	int error;
487 
488 	fsp = (struct hsfs *)VFS_TO_HSFS(vfsp);
489 	fid = (struct hsfid *)fidp;
490 
491 	/*
492 	 * Look for vnode on hashlist.
493 	 * If found, it's now active and the refcnt was incremented.
494 	 */
495 
496 	rw_enter(&fsp->hsfs_hash_lock, RW_READER);
497 
498 	nodeid = fid->hf_ino;
499 
500 	if ((*vpp = hs_findhash(nodeid, fid->hf_dir_lbn,
501 	    (uint_t)fid->hf_dir_off, vfsp)) == NULL) {
502 		/*
503 		 * Not in cache, so we need to remake it.
504 		 * hs_remakenode() will read the directory entry
505 		 * and then check again to see if anyone else has
506 		 * put it in the cache.
507 		 */
508 		rw_exit(&fsp->hsfs_hash_lock);
509 		error = hs_remakenode(fid->hf_dir_lbn, (uint_t)fid->hf_dir_off,
510 		    vfsp, vpp);
511 		return (error);
512 	}
513 	rw_exit(&fsp->hsfs_hash_lock);
514 	return (0);
515 }
516 
517 
518 #define	CHECKSUM_SIZE				(64 * 1024)
519 
520 /*
521  * Compute a CD-ROM fsid by checksumming the first 64K of data on the CD
522  * We use the 'fsp' argument to determine the location of the root
523  * directory entry, and we start reading from there.
524  */
525 static int
526 compute_cdrom_id(struct hsfs *fsp, vnode_t *devvp)
527 {
528 	uint_t		secno;
529 	struct hs_volume *hsvp = &fsp->hsfs_vol;
530 	struct buf	*bp;
531 	int		error;
532 	int		fsid;
533 
534 	secno = hsvp->root_dir.ext_lbn >> hsvp->lbn_secshift;
535 	bp = bread(devvp->v_rdev, secno * 4, CHECKSUM_SIZE);
536 	error = geterror(bp);
537 
538 	/*
539 	 * An error on read or a partial read means we asked
540 	 * for a nonexistant/corrupted piece of the device
541 	 * (including past-the-end of the media). Don't
542 	 * try to use the checksumming method then.
543 	 */
544 	if (!error && bp->b_bcount == CHECKSUM_SIZE) {
545 		int *ibuf = (int *)bp->b_un.b_addr;
546 		int i;
547 
548 		fsid = 0;
549 
550 		for (i = 0; i < CHECKSUM_SIZE / sizeof (int); i++)
551 			fsid ^= ibuf[ i ];
552 	} else {
553 		/*
554 		 * Fallback - use creation date
555 		 */
556 		fsid = hsvp->cre_date.tv_sec;
557 	}
558 
559 	brelse(bp);
560 
561 	return (fsid);
562 }
563 
564 
565 /*ARGSUSED*/
566 static int
567 hs_mountfs(
568 	struct vfs	*vfsp,
569 	dev_t		dev,
570 	char		*path,
571 	mode_t		mode,
572 	int		mount_flags,
573 	struct cred	*cr,
574 	int		isroot)
575 {
576 	struct vnode	*devvp;
577 	struct hsfs	*tsp;
578 	struct hsfs	*fsp = NULL;
579 	struct vattr	vap;
580 	struct hsnode	*hp;
581 	int		error;
582 	struct timeval	tv;
583 	int		fsid;
584 	int		use_rrip;
585 	int		use_vers2;
586 	int		use_joliet;
587 	int		has_rrip = 0;
588 	int		has_vers2 = 0;
589 	int		has_joliet = 0;
590 	int		force_rrip_off;
591 	int		force_vers2_off;
592 	int		force_joliet_off;
593 	size_t		pathbufsz = strlen(path) + 1;
594 	int		redo_rootvp;
595 
596 	struct hs_volume *svp = NULL;	/* Supplemental VD for ISO-9660:1999 */
597 	struct hs_volume *jvp = NULL;	/* Joliet VD */
598 
599 	/*
600 	 * The rules for which extension will be used are:
601 	 * 1. No specific mount options given:
602 	 *	- use rrip if available
603 	 *	- use ISO9660:1999 if available
604 	 *	- use joliet if available.
605 	 * 2. rrip/ISO9660:1999/joliet explicitly disabled via mount option:
606 	 *	- use next "lower" extension
607 	 * 3. joliet/ISO9660:1999/rrip explicitly requested via mount option:
608 	 *	- disable rrip support even if available
609 	 *	- disable IOS9660:1999 support even if available
610 	 *
611 	 * We need to adjust these flags as we discover the extensions
612 	 * present. See below. These are just the starting values.
613 	 */
614 	use_rrip = (mount_flags & HSFSMNT_NORRIP) == 0;
615 	use_vers2 = (mount_flags & HSFSMNT_NOVERS2) == 0;
616 	use_joliet = (mount_flags & HSFSMNT_NOJOLIET) == 0;
617 
618 	/*
619 	 * Open the device
620 	 */
621 	devvp = makespecvp(dev, VBLK);
622 	ASSERT(devvp != 0);
623 
624 	/*
625 	 * Open the target device (file) for read only.
626 	 */
627 	if (error = VOP_OPEN(&devvp, FREAD, cr, NULL)) {
628 		VN_RELE(devvp);
629 		return (error);
630 	}
631 
632 	/*
633 	 * Refuse to go any further if this
634 	 * device is being used for swapping
635 	 */
636 	if (IS_SWAPVP(common_specvp(devvp))) {
637 		error = EBUSY;
638 		goto cleanup;
639 	}
640 
641 	vap.va_mask = AT_SIZE;
642 	if ((error = VOP_GETATTR(devvp, &vap, ATTR_COMM, cr, NULL)) != 0) {
643 		cmn_err(CE_NOTE, "Cannot get attributes of the CD-ROM driver");
644 		goto cleanup;
645 	}
646 
647 	/*
648 	 * Make sure we have a nonzero size partition.
649 	 * The current version of the SD driver will *not* fail the open
650 	 * of such a partition so we have to check for it here.
651 	 */
652 	if (vap.va_size == 0) {
653 		error = ENXIO;
654 		goto cleanup;
655 	}
656 
657 	/*
658 	 * Init a new hsfs structure.
659 	 */
660 	fsp = kmem_zalloc(sizeof (*fsp), KM_SLEEP);
661 	svp = kmem_zalloc(sizeof (*svp), KM_SLEEP);
662 	jvp = kmem_zalloc(sizeof (*jvp), KM_SLEEP);
663 
664 	/* hardwire perms, uid, gid */
665 	fsp->hsfs_vol.vol_uid = hsfs_default_uid;
666 	fsp->hsfs_vol.vol_gid =  hsfs_default_gid;
667 	fsp->hsfs_vol.vol_prot = hsfs_default_mode;
668 	svp->vol_uid = hsfs_default_uid;
669 	svp->vol_gid =  hsfs_default_gid;
670 	svp->vol_prot = hsfs_default_mode;
671 	jvp->vol_uid = hsfs_default_uid;
672 	jvp->vol_gid =  hsfs_default_gid;
673 	jvp->vol_prot = hsfs_default_mode;
674 
675 	/*
676 	 * Look for a Standard File Structure Volume Descriptor,
677 	 * of which there must be at least one.
678 	 * If found, check for volume size consistency.
679 	 *
680 	 * If svp->lbn_size is != 0, we did find a ISO-9660:1999 SVD
681 	 * If jvp->lbn_size is != 0, we did find a Joliet SVD.
682 	 */
683 	fsp->hsfs_namemax = ISO_FILE_NAMELEN;
684 	fsp->hsfs_namelen = ISO_FILE_NAMELEN;
685 	error = hs_findisovol(fsp, devvp, &fsp->hsfs_vol, svp, jvp);
686 	if (error == EINVAL) /* no iso 9660 - try high sierra ... */
687 		error = hs_findhsvol(fsp, devvp, &fsp->hsfs_vol);
688 
689 	if (error)
690 		goto cleanup;
691 
692 	DTRACE_PROBE4(findvol,
693 	    struct hsfs *, fsp,
694 	    struct hs_volume *, &fsp->hsfs_vol,
695 	    struct hs_volume *, svp,
696 	    struct hs_volume *, jvp);
697 
698 	/*
699 	 * Generate a file system ID from the CD-ROM,
700 	 * and check it for uniqueness.
701 	 *
702 	 * What we are aiming for is some chance of integrity
703 	 * across disk change.  That is, if a client has an fhandle,
704 	 * it will be valid as long as the same disk is mounted.
705 	 */
706 	fsid = compute_cdrom_id(fsp, devvp);
707 
708 	mutex_enter(&hs_mounttab_lock);
709 
710 	if (fsid == 0 || fsid == -1) {
711 		uniqtime(&tv);
712 		fsid = tv.tv_sec;
713 	} else	/* make sure that the fsid is unique */
714 		for (tsp = hs_mounttab; tsp != NULL; tsp = tsp->hsfs_next) {
715 			if (fsid == tsp->hsfs_vfs->vfs_fsid.val[0]) {
716 				uniqtime(&tv);
717 				fsid = tv.tv_sec;
718 				break;
719 			}
720 		}
721 
722 	fsp->hsfs_next = hs_mounttab;
723 	hs_mounttab = fsp;
724 
725 	fsp->hsfs_devvp = devvp;
726 	fsp->hsfs_vfs = vfsp;
727 	fsp->hsfs_fsmnt = kmem_alloc(pathbufsz, KM_SLEEP);
728 	(void) strlcpy(fsp->hsfs_fsmnt, path, pathbufsz);
729 
730 	mutex_init(&fsp->hsfs_free_lock, NULL, MUTEX_DEFAULT, NULL);
731 	rw_init(&fsp->hsfs_hash_lock, NULL, RW_DEFAULT, NULL);
732 
733 	vfsp->vfs_data = (caddr_t)fsp;
734 	vfsp->vfs_dev = dev;
735 	vfsp->vfs_fstype = hsfsfstype;
736 	vfsp->vfs_bsize = fsp->hsfs_vol.lbn_size; /* %% */
737 	vfsp->vfs_fsid.val[0] = fsid;
738 	vfsp->vfs_fsid.val[1] =  hsfsfstype;
739 
740 	if (!hs_getrootvp(vfsp, fsp, pathbufsz)) {
741 		DTRACE_PROBE1(rootvp__failed, struct hsfs *, fsp);
742 		error = EINVAL;
743 		goto cleanup;
744 	}
745 	DTRACE_PROBE1(rootvp, struct hsfs *, fsp);
746 
747 	/*
748 	 * Attempt to discover a RR extension.
749 	 */
750 	if (use_rrip) {
751 		hp = VTOH(fsp->hsfs_rootvp);
752 		hs_check_root_dirent(fsp->hsfs_rootvp, &(hp->hs_dirent));
753 	}
754 
755 	has_rrip = IS_RRIP_IMPLEMENTED(fsp);
756 	has_vers2 = (svp->lbn_size != 0);
757 	has_joliet = (jvp->lbn_size != 0);
758 
759 	DTRACE_PROBE4(voltype__suggested, struct hsfs *, fsp,
760 	    int, use_rrip, int, use_vers2, int, use_joliet);
761 
762 	DTRACE_PROBE4(voltype__actual, struct hsfs *, fsp,
763 	    int, has_rrip, int, has_vers2, int, has_joliet);
764 
765 	DTRACE_PROBE4(findvol,
766 	    struct hsfs *, fsp,
767 	    struct hs_volume *, &fsp->hsfs_vol,
768 	    struct hs_volume *, svp,
769 	    struct hs_volume *, jvp);
770 
771 	force_rrip_off = !use_rrip ||
772 	    (vfs_optionisset(vfsp, HOPT_JOLIET, NULL) && has_joliet) ||
773 	    (vfs_optionisset(vfsp, HOPT_VERS2, NULL) && has_vers2);
774 
775 	force_vers2_off = !use_vers2 ||
776 	    (vfs_optionisset(vfsp, HOPT_JOLIET, NULL) && has_joliet);
777 
778 	force_joliet_off = !use_joliet;
779 
780 	DTRACE_PROBE4(voltype__force_off, struct hsfs *, fsp,
781 	    int, force_rrip_off, int, force_vers2_off, int, force_joliet_off);
782 
783 	/*
784 	 * At the moment, we have references of all three possible
785 	 * extensions (RR, ISO9660:1999/v2 and Joliet) if present.
786 	 *
787 	 * The "active" volume descriptor is RRIP (or ISO9660:1988).
788 	 * We now switch to the user-requested one.
789 	 */
790 	redo_rootvp = 0;
791 
792 	if (force_rrip_off || !has_rrip) {
793 		if (has_vers2 && !force_vers2_off) {
794 			VN_RELE(fsp->hsfs_rootvp);
795 			bcopy(svp, &fsp->hsfs_vol, sizeof (struct hs_volume));
796 			fsp->hsfs_vol_type = HS_VOL_TYPE_ISO_V2;
797 			vfsp->vfs_bsize = fsp->hsfs_vol.lbn_size;
798 			redo_rootvp = 1;
799 			has_joliet = 0;
800 		} else if (has_joliet && !force_joliet_off) {
801 			VN_RELE(fsp->hsfs_rootvp);
802 			bcopy(jvp, &fsp->hsfs_vol, sizeof (struct hs_volume));
803 			fsp->hsfs_vol_type = HS_VOL_TYPE_JOLIET;
804 			vfsp->vfs_bsize = fsp->hsfs_vol.lbn_size;
805 			redo_rootvp = 1;
806 			has_vers2 = 0;
807 		}
808 	}
809 
810 	if (redo_rootvp) {
811 		/*
812 		 * Make sure not to use Rock Ridge.
813 		 */
814 		UNSET_IMPL_BIT(fsp, RRIP_BIT);
815 		UNSET_SUSP_BIT(fsp);
816 		has_rrip = 0;
817 
818 		if (!hs_getrootvp(vfsp, fsp, pathbufsz)) {
819 			DTRACE_PROBE1(rootvp__failed, struct hsfs *, fsp);
820 			error = EINVAL;
821 			goto cleanup;
822 		}
823 		DTRACE_PROBE1(rootvp, struct hsfs *, fsp);
824 	}
825 	if (IS_RRIP_IMPLEMENTED(fsp)) {
826 		has_vers2 = 0;
827 		has_joliet = 0;
828 	}
829 	if (force_vers2_off)
830 		has_vers2 = 0;
831 	if (force_joliet_off)
832 		has_joliet = 0;
833 	DTRACE_PROBE4(voltype__taken, struct hsfs *, fsp,
834 	    int, has_rrip, int, has_vers2, int, has_joliet);
835 
836 	/*
837 	 * mark root node as VROOT
838 	 */
839 	fsp->hsfs_rootvp->v_flag |= VROOT;
840 
841 	/* Here we take care of some special case stuff for mountroot */
842 	if (isroot) {
843 		fsp->hsfs_rootvp->v_rdev = devvp->v_rdev;
844 		rootvp = fsp->hsfs_rootvp;
845 	}
846 
847 	if (IS_RRIP_IMPLEMENTED(fsp)) {
848 		/*
849 		 * if RRIP, don't copy NOMAPLCASE or NOTRAILDOT to hsfs_flags
850 		 */
851 		mount_flags &= ~(HSFSMNT_NOMAPLCASE | HSFSMNT_NOTRAILDOT);
852 
853 		fsp->hsfs_namemax = RRIP_FILE_NAMELEN;
854 		fsp->hsfs_namelen = RRIP_FILE_NAMELEN;
855 
856 		ASSERT(vfs_optionisset(vfsp, HOPT_RR, NULL));
857 		vfs_clearmntopt(vfsp, HOPT_VERS2);
858 		vfs_clearmntopt(vfsp, HOPT_JOLIET);
859 
860 	} else switch (fsp->hsfs_vol_type) {
861 
862 	case HS_VOL_TYPE_HS:
863 	case HS_VOL_TYPE_ISO:
864 	default:
865 		/*
866 		 * if iso v1, don't allow trailing spaces in iso file names
867 		 */
868 		mount_flags |= HSFSMNT_NOTRAILSPACE;
869 		fsp->hsfs_namemax = ISO_NAMELEN_V2_MAX;
870 		fsp->hsfs_namelen = ISO_FILE_NAMELEN;
871 		vfs_clearmntopt(vfsp, HOPT_RR);
872 		vfs_clearmntopt(vfsp, HOPT_VERS2);
873 		vfs_clearmntopt(vfsp, HOPT_JOLIET);
874 		break;
875 
876 	case HS_VOL_TYPE_ISO_V2:
877 		/*
878 		 * if iso v2, don't copy NOTRAILDOT to hsfs_flags
879 		 */
880 		mount_flags &= ~HSFSMNT_NOTRAILDOT;
881 		mount_flags |= HSFSMNT_NOMAPLCASE | HSFSMNT_NOVERSION;
882 		fsp->hsfs_namemax = ISO_NAMELEN_V2_MAX;
883 		fsp->hsfs_namelen = ISO_NAMELEN_V2;
884 		vfs_setmntopt(vfsp, HOPT_VERS2, NULL, 0);
885 		vfs_clearmntopt(vfsp, HOPT_RR);
886 		vfs_clearmntopt(vfsp, HOPT_JOLIET);
887 		break;
888 
889 	case HS_VOL_TYPE_JOLIET:
890 		/*
891 		 * if Joliet, don't copy NOMAPLCASE or NOTRAILDOT to hsfs_flags
892 		 */
893 		mount_flags &= ~(HSFSMNT_NOMAPLCASE | HSFSMNT_NOTRAILDOT);
894 		mount_flags |= HSFSMNT_NOMAPLCASE;
895 		if (mount_flags & HSFSMNT_JOLIETLONG)
896 			fsp->hsfs_namemax = JOLIET_NAMELEN_MAX*3; /* UTF-8 */
897 		else
898 			fsp->hsfs_namemax = MAXNAMELEN-1;
899 		fsp->hsfs_namelen = JOLIET_NAMELEN*2;
900 		vfs_setmntopt(vfsp, HOPT_JOLIET, NULL, 0);
901 		vfs_clearmntopt(vfsp, HOPT_RR);
902 		vfs_clearmntopt(vfsp, HOPT_VERS2);
903 		break;
904 	}
905 
906 	/*
907 	 * Add the HSFSMNT_INODE pseudo mount flag to the current mount flags.
908 	 */
909 	fsp->hsfs_flags = mount_flags | (fsp->hsfs_flags & HSFSMNT_INODE);
910 
911 	/*
912 	 * Setup I/O Scheduling structures
913 	 */
914 	if (do_schedio) {
915 		fsp->hqueue = kmem_alloc(sizeof (struct hsfs_queue), KM_SLEEP);
916 		hsched_init(fsp, fsid, &modlinkage);
917 	}
918 
919 	/*
920 	 * Setup kstats
921 	 */
922 	hsfs_init_kstats(fsp, fsid);
923 
924 	DTRACE_PROBE1(mount__done, struct hsfs *, fsp);
925 
926 	/*
927 	 * set the magic word
928 	 */
929 	fsp->hsfs_magic = HSFS_MAGIC;
930 	mutex_exit(&hs_mounttab_lock);
931 
932 	kmem_free(svp, sizeof (*svp));
933 	kmem_free(jvp, sizeof (*jvp));
934 
935 	return (0);
936 
937 cleanup:
938 	(void) VOP_CLOSE(devvp, FREAD, 1, (offset_t)0, cr, NULL);
939 	VN_RELE(devvp);
940 	if (fsp)
941 		kmem_free(fsp, sizeof (*fsp));
942 	if (svp)
943 		kmem_free(svp, sizeof (*svp));
944 	if (jvp)
945 		kmem_free(jvp, sizeof (*jvp));
946 	return (error);
947 }
948 
949 /*
950  * Get the rootvp associated with fsp->hsfs_vol
951  */
952 static int
953 hs_getrootvp(
954 	struct vfs	*vfsp,
955 	struct hsfs	*fsp,
956 	size_t		pathsize)
957 {
958 	struct hsnode	*hp;
959 
960 	ASSERT(pathsize == strlen(fsp->hsfs_fsmnt) + 1);
961 
962 	/*
963 	 * If the root directory does not appear to be
964 	 * valid, use what it points to as "." instead.
965 	 * Some Defense Mapping Agency disks are non-conformant
966 	 * in this way.
967 	 */
968 	if (!hsfs_valid_dir(&fsp->hsfs_vol.root_dir)) {
969 		hs_log_bogus_disk_warning(fsp, HSFS_ERR_BAD_ROOT_DIR, 0);
970 		if (hs_remakenode(fsp->hsfs_vol.root_dir.ext_lbn,
971 		    (uint_t)0, vfsp, &fsp->hsfs_rootvp)) {
972 			hs_mounttab = hs_mounttab->hsfs_next;
973 			mutex_destroy(&fsp->hsfs_free_lock);
974 			rw_destroy(&fsp->hsfs_hash_lock);
975 			kmem_free(fsp->hsfs_fsmnt, pathsize);
976 			mutex_exit(&hs_mounttab_lock);
977 			return (0);
978 		}
979 	} else {
980 		fsp->hsfs_rootvp = hs_makenode(&fsp->hsfs_vol.root_dir,
981 		    fsp->hsfs_vol.root_dir.ext_lbn, 0, vfsp);
982 	}
983 
984 	/* XXX - ignore the path table for now */
985 	fsp->hsfs_ptbl = NULL;
986 	hp = VTOH(fsp->hsfs_rootvp);
987 	hp->hs_ptbl_idx = NULL;
988 
989 	return (1);
990 }
991 
992 /*
993  * hs_findhsvol()
994  *
995  * Locate the Standard File Structure Volume Descriptor and
996  * parse it into an hs_volume structure.
997  *
998  * XXX - May someday want to look for Coded Character Set FSVD, too.
999  */
1000 static int
1001 hs_findhsvol(struct hsfs *fsp, struct vnode *vp, struct hs_volume *hvp)
1002 {
1003 	struct buf *secbp;
1004 	int i;
1005 	int n;
1006 	uchar_t *volp;
1007 	int error;
1008 	uint_t secno;
1009 
1010 	secno = hs_findvoldesc(vp->v_rdev, HS_VOLDESC_SEC);
1011 	secbp = bread(vp->v_rdev, secno * 4, HS_SECTOR_SIZE);
1012 	error = geterror(secbp);
1013 
1014 	if (error != 0) {
1015 		cmn_err(CE_NOTE, "hs_findhsvol: bread: error=(%d)", error);
1016 		brelse(secbp);
1017 		return (error);
1018 	}
1019 
1020 	volp = (uchar_t *)secbp->b_un.b_addr;
1021 
1022 	/*
1023 	 * To avoid that we read the whole medium in case that someone prepares
1024 	 * a malicious "fs image", we read at most 32 blocks.
1025 	 */
1026 	for (n = 0; n < 32 &&
1027 	    HSV_DESC_TYPE(volp) != VD_EOV; n++) {
1028 		for (i = 0; i < HSV_ID_STRLEN; i++)
1029 			if (HSV_STD_ID(volp)[i] != HSV_ID_STRING[i])
1030 				goto cantfind;
1031 		if (HSV_STD_VER(volp) != HSV_ID_VER)
1032 			goto cantfind;
1033 		switch (HSV_DESC_TYPE(volp)) {
1034 		case VD_SFS:
1035 			/* Standard File Structure */
1036 			fsp->hsfs_vol_type = HS_VOL_TYPE_HS;
1037 			error = hs_parsehsvol(fsp, volp, hvp);
1038 			brelse(secbp);
1039 			return (error);
1040 
1041 		case VD_CCFS:
1042 			/* Coded Character File Structure */
1043 		case VD_BOOT:
1044 		case VD_UNSPEC:
1045 		case VD_EOV:
1046 			break;
1047 		}
1048 		brelse(secbp);
1049 		++secno;
1050 		secbp = bread(vp->v_rdev, secno * 4, HS_SECTOR_SIZE);
1051 
1052 		error = geterror(secbp);
1053 
1054 		if (error != 0) {
1055 			cmn_err(CE_NOTE, "hs_findhsvol: bread: error=(%d)",
1056 			    error);
1057 			brelse(secbp);
1058 			return (error);
1059 		}
1060 
1061 		volp = (uchar_t *)secbp->b_un.b_addr;
1062 	}
1063 cantfind:
1064 	brelse(secbp);
1065 	return (EINVAL);
1066 }
1067 
1068 /*
1069  * hs_parsehsvol
1070  *
1071  * Parse the Standard File Structure Volume Descriptor into
1072  * an hs_volume structure.  We can't just bcopy it into the
1073  * structure because of byte-ordering problems.
1074  *
1075  */
1076 static int
1077 hs_parsehsvol(struct hsfs *fsp, uchar_t *volp, struct hs_volume *hvp)
1078 {
1079 	hvp->vol_size = HSV_VOL_SIZE(volp);
1080 	hvp->lbn_size = HSV_BLK_SIZE(volp);
1081 	if (hvp->lbn_size == 0) {
1082 		cmn_err(CE_NOTE, "hs_parsehsvol: logical block size in the "
1083 		    "SFSVD is zero");
1084 		return (EINVAL);
1085 	}
1086 	hvp->lbn_shift = ffs((long)hvp->lbn_size) - 1;
1087 	hvp->lbn_secshift =
1088 	    ffs((long)howmany(HS_SECTOR_SIZE, (int)hvp->lbn_size)) - 1;
1089 	hvp->lbn_maxoffset = hvp->lbn_size - 1;
1090 	hs_parse_longdate(HSV_cre_date(volp), &hvp->cre_date);
1091 	hs_parse_longdate(HSV_mod_date(volp), &hvp->mod_date);
1092 	hvp->file_struct_ver = HSV_FILE_STRUCT_VER(volp);
1093 	hvp->ptbl_len = HSV_PTBL_SIZE(volp);
1094 	hvp->vol_set_size = (ushort_t)HSV_SET_SIZE(volp);
1095 	hvp->vol_set_seq = (ushort_t)HSV_SET_SEQ(volp);
1096 #if defined(_LITTLE_ENDIAN)
1097 	hvp->ptbl_lbn = HSV_PTBL_MAN_LS(volp);
1098 #else
1099 	hvp->ptbl_lbn = HSV_PTBL_MAN_MS(volp);
1100 #endif
1101 	hs_copylabel(hvp, HSV_VOL_ID(volp), 0);
1102 
1103 	/*
1104 	 * Make sure that lbn_size is a power of two and otherwise valid.
1105 	 */
1106 	if (hvp->lbn_size & ~(1 << hvp->lbn_shift)) {
1107 		cmn_err(CE_NOTE,
1108 		    "hsfs: %d-byte logical block size not supported",
1109 		    hvp->lbn_size);
1110 		return (EINVAL);
1111 	}
1112 	return (hs_parsedir(fsp, HSV_ROOT_DIR(volp), &hvp->root_dir,
1113 	    (char *)NULL, (int *)NULL, HDE_ROOT_DIR_REC_SIZE));
1114 }
1115 
1116 /*
1117  * hs_findisovol()
1118  *
1119  * Locate the Primary Volume Descriptor
1120  * parse it into an hs_volume structure.
1121  *
1122  * XXX - Partition not yet done
1123  *
1124  * Except for fsp->hsfs_vol_type, no fsp member may be modified.
1125  * fsp->hsfs_vol is modified indirectly via the *hvp argument.
1126  */
1127 static int
1128 hs_findisovol(struct hsfs *fsp, struct vnode *vp,
1129     struct hs_volume *hvp,
1130     struct hs_volume *svp,
1131     struct hs_volume *jvp)
1132 {
1133 	struct buf *secbp;
1134 	int i;
1135 	int n;
1136 	uchar_t *volp;
1137 	int error;
1138 	uint_t secno;
1139 	int foundpvd = 0;
1140 	int foundsvd = 0;
1141 	int foundjvd = 0;
1142 	int pvd_sum = 0;
1143 
1144 	secno = hs_findvoldesc(vp->v_rdev, ISO_VOLDESC_SEC);
1145 	secbp = bread(vp->v_rdev, secno * 4, ISO_SECTOR_SIZE);
1146 	error = geterror(secbp);
1147 
1148 	if (error != 0) {
1149 		cmn_err(CE_NOTE, "hs_findisovol: bread: error=(%d)", error);
1150 		brelse(secbp);
1151 		return (error);
1152 	}
1153 
1154 	volp = (uchar_t *)secbp->b_un.b_addr;
1155 
1156 	/*
1157 	 * To avoid that we read the whole medium in case that someone prepares
1158 	 * a malicious "fs image", we read at most 32 blocks.
1159 	 */
1160 	for (n = 0; n < 32 &&
1161 	    (enum iso_voldesc_type) ISO_DESC_TYPE(volp) != ISO_VD_EOV; n++) {
1162 		for (i = 0; i < ISO_ID_STRLEN; i++)
1163 			if (ISO_STD_ID(volp)[i] != ISO_ID_STRING[i])
1164 				goto cantfind;
1165 		switch (ISO_DESC_TYPE(volp)) {
1166 		case ISO_VD_PVD:
1167 			/* Standard File Structure */
1168 			if (ISO_STD_VER(volp) != ISO_ID_VER)
1169 				goto cantfind;
1170 			if (foundpvd != 1) {
1171 				fsp->hsfs_vol_type = HS_VOL_TYPE_ISO;
1172 				if (error = hs_parseisovol(fsp, volp, hvp)) {
1173 					brelse(secbp);
1174 					return (error);
1175 				}
1176 				foundpvd = 1;
1177 				for (i = 0; i < ISO_SECTOR_SIZE; i++)
1178 					pvd_sum += volp[i];
1179 			}
1180 			break;
1181 		case ISO_VD_SVD:
1182 			/* Supplementary Volume Descriptor */
1183 			if (ISO_STD_VER(volp) == ISO_ID_VER2 &&
1184 			    foundsvd != 1) {
1185 				fsp->hsfs_vol_type = HS_VOL_TYPE_ISO;
1186 				if (error = hs_parseisovol(fsp, volp, svp)) {
1187 					brelse(secbp);
1188 					return (error);
1189 				}
1190 				foundsvd = 1;
1191 			}
1192 			if (hs_joliet_level(volp) >= 1 && foundjvd != 1) {
1193 				fsp->hsfs_vol_type = HS_VOL_TYPE_ISO;
1194 				if (error = hs_parseisovol(fsp, volp, jvp)) {
1195 					brelse(secbp);
1196 					return (error);
1197 				}
1198 				foundjvd = 1;
1199 			}
1200 			break;
1201 		case ISO_VD_BOOT:
1202 			break;
1203 		case ISO_VD_VPD:
1204 			/* currently cannot handle partition */
1205 			break;
1206 		case VD_EOV:
1207 			break;
1208 		}
1209 		brelse(secbp);
1210 		++secno;
1211 		secbp = bread(vp->v_rdev, secno * 4, HS_SECTOR_SIZE);
1212 		error = geterror(secbp);
1213 
1214 		if (error != 0) {
1215 			cmn_err(CE_NOTE, "hs_findisovol: bread: error=(%d)",
1216 			    error);
1217 			brelse(secbp);
1218 			return (error);
1219 		}
1220 
1221 		volp = (uchar_t *)secbp->b_un.b_addr;
1222 	}
1223 	for (n = 0; n < 16; n++) {
1224 		brelse(secbp);
1225 		++secno;
1226 		secbp = bread(vp->v_rdev, secno * 4, HS_SECTOR_SIZE);
1227 		error = geterror(secbp);
1228 
1229 		if (error != 0) {
1230 			cmn_err(CE_NOTE, "hs_findisovol: bread: error=(%d)",
1231 			    error);
1232 			brelse(secbp);
1233 			return (error);
1234 		}
1235 
1236 		/*
1237 		 * Check for the signature from mkisofs that grants that
1238 		 * the current filesystem allows to use the extent lbn as
1239 		 * inode number even in pure ISO9660 mode.
1240 		 */
1241 		volp = (uchar_t *)secbp->b_un.b_addr;
1242 		if (strncmp((char *)volp, "MKI ", 4) == 0) {
1243 			int	sum;
1244 
1245 			sum  = volp[2045];
1246 			sum *= 256;
1247 			sum += volp[2046];
1248 			sum *= 256;
1249 			sum += volp[2047];
1250 			if (sum == pvd_sum)
1251 				fsp->hsfs_flags |= HSFSMNT_INODE;
1252 			break;
1253 		}
1254 	}
1255 	if (foundpvd) {
1256 		brelse(secbp);
1257 		return (0);
1258 	}
1259 cantfind:
1260 	brelse(secbp);
1261 	return (EINVAL);
1262 }
1263 
1264 /*
1265  * Return 0 if no Joliet is found
1266  * else return Joliet Level 1..3
1267  */
1268 static int
1269 hs_joliet_level(uchar_t *volp)
1270 {
1271 	if (ISO_std_ver(volp)[0] == ISO_ID_VER &&
1272 	    ISO_svd_esc(volp)[0] == '%' &&
1273 	    ISO_svd_esc(volp)[1] == '/') {
1274 
1275 		switch (ISO_svd_esc(volp)[2]) {
1276 
1277 		case '@':
1278 			return (1);
1279 		case 'C':
1280 			return (2);
1281 		case 'E':
1282 			return (3);
1283 		}
1284 	}
1285 	return (0);
1286 }
1287 
1288 /*
1289  * hs_parseisovol
1290  *
1291  * Parse the Primary Volume Descriptor into an hs_volume structure.
1292  *
1293  */
1294 static int
1295 hs_parseisovol(struct hsfs *fsp, uchar_t *volp, struct hs_volume *hvp)
1296 {
1297 	hvp->vol_size = ISO_VOL_SIZE(volp);
1298 	hvp->lbn_size = ISO_BLK_SIZE(volp);
1299 	if (hvp->lbn_size == 0) {
1300 		cmn_err(CE_NOTE, "hs_parseisovol: logical block size in the "
1301 		    "PVD is zero");
1302 		return (EINVAL);
1303 	}
1304 	hvp->lbn_shift = ffs((long)hvp->lbn_size) - 1;
1305 	hvp->lbn_secshift =
1306 	    ffs((long)howmany(ISO_SECTOR_SIZE, (int)hvp->lbn_size)) - 1;
1307 	hvp->lbn_maxoffset = hvp->lbn_size - 1;
1308 	hs_parse_longdate(ISO_cre_date(volp), &hvp->cre_date);
1309 	hs_parse_longdate(ISO_mod_date(volp), &hvp->mod_date);
1310 	hvp->file_struct_ver = ISO_FILE_STRUCT_VER(volp);
1311 	hvp->ptbl_len = ISO_PTBL_SIZE(volp);
1312 	hvp->vol_set_size = (ushort_t)ISO_SET_SIZE(volp);
1313 	hvp->vol_set_seq = (ushort_t)ISO_SET_SEQ(volp);
1314 #if defined(_LITTLE_ENDIAN)
1315 	hvp->ptbl_lbn = ISO_PTBL_MAN_LS(volp);
1316 #else
1317 	hvp->ptbl_lbn = ISO_PTBL_MAN_MS(volp);
1318 #endif
1319 	hs_copylabel(hvp, ISO_VOL_ID(volp), hs_joliet_level(volp) >= 1);
1320 
1321 	/*
1322 	 * Make sure that lbn_size is a power of two and otherwise valid.
1323 	 */
1324 	if (hvp->lbn_size & ~(1 << hvp->lbn_shift)) {
1325 		cmn_err(CE_NOTE,
1326 		    "hsfs: %d-byte logical block size not supported",
1327 		    hvp->lbn_size);
1328 		return (EINVAL);
1329 	}
1330 	return (hs_parsedir(fsp, ISO_ROOT_DIR(volp), &hvp->root_dir,
1331 	    (char *)NULL, (int *)NULL, IDE_ROOT_DIR_REC_SIZE));
1332 }
1333 
1334 /*
1335  * Common code for mount and umount.
1336  * Check that the user's argument is a reasonable
1337  * thing on which to mount, and return the device number if so.
1338  */
1339 static int
1340 hs_getmdev(struct vfs *vfsp, char *fspec, int flags, dev_t *pdev, mode_t *mode,
1341     cred_t *cr)
1342 {
1343 	int error;
1344 	struct vnode *svp = NULL;
1345 	struct vnode *lvp = NULL;
1346 	struct vnode *bvp;
1347 	struct vattr vap;
1348 	dev_t dev;
1349 	enum uio_seg fromspace = (flags & MS_SYSSPACE) ?
1350 	    UIO_SYSSPACE : UIO_USERSPACE;
1351 
1352 	/*
1353 	 * Look up the device/file to be mounted.
1354 	 */
1355 	error = lookupname(fspec, fromspace, FOLLOW, NULLVPP, &svp);
1356 	if (error) {
1357 		if (error == ENOENT)
1358 			error = ENODEV;
1359 		goto out;
1360 	}
1361 
1362 	error = vfs_get_lofi(vfsp, &lvp);
1363 
1364 	if (error > 0) {
1365 		if (error == ENOENT)
1366 			error = ENODEV;
1367 		goto out;
1368 	} else if (error == 0) {
1369 		bvp = lvp;
1370 	} else {
1371 		bvp = svp;
1372 
1373 		if (bvp->v_type != VBLK) {
1374 			error = ENOTBLK;
1375 			goto out;
1376 		}
1377 
1378 		if ((error = secpolicy_spec_open(cr, bvp, FREAD)) != 0)
1379 			goto out;
1380 	}
1381 
1382 	/*
1383 	 * Can we read from the device/file ?
1384 	 */
1385 	if ((error = VOP_ACCESS(svp, VREAD, 0, cr, NULL)) != 0)
1386 		goto out;
1387 
1388 	vap.va_mask = AT_MODE;		/* get protection mode */
1389 	(void) VOP_GETATTR(bvp, &vap, 0, CRED(), NULL);
1390 	*mode = vap.va_mode;
1391 
1392 	dev = *pdev = bvp->v_rdev;
1393 
1394 	error = EBUSY;
1395 
1396 	/*
1397 	 * Ensure that this device isn't already mounted,
1398 	 * unless this is a REMOUNT request or we are told to suppress
1399 	 * mount checks.
1400 	 */
1401 	if ((flags & MS_NOCHECK) == 0) {
1402 		if (vfs_devmounting(dev, vfsp))
1403 			goto out;
1404 		if (vfs_devismounted(dev) && !(flags & MS_REMOUNT))
1405 			goto out;
1406 	}
1407 
1408 	if (getmajor(*pdev) >= devcnt) {
1409 		error = ENXIO;
1410 		goto out;
1411 	}
1412 
1413 	error = 0;
1414 out:
1415 	if (svp != NULL)
1416 		VN_RELE(svp);
1417 	if (lvp != NULL)
1418 		VN_RELE(lvp);
1419 	return (error);
1420 }
1421 
1422 static void
1423 hs_copylabel(struct hs_volume *hvp, unsigned char *label, int isjoliet)
1424 {
1425 	char	lbuf[64];	/* hs_joliet_cp() creates 48 bytes at most */
1426 
1427 	if (isjoliet) {
1428 		/*
1429 		 * hs_joliet_cp() will output 16..48 bytes.
1430 		 * We need to clear 'lbuf' to avoid junk chars past byte 15.
1431 		 */
1432 		bzero(lbuf, sizeof (lbuf));
1433 		(void) hs_joliet_cp((char *)label, lbuf, 32);
1434 		label = (unsigned char *)lbuf;
1435 	}
1436 	/* cdrom volid is at most 32 bytes */
1437 	bcopy(label, hvp->vol_id, 32);
1438 	hvp->vol_id[31] = NULL;
1439 }
1440 
1441 /*
1442  * Mount root file system.
1443  * "why" is ROOT_INIT on initial call, ROOT_REMOUNT if called to
1444  * remount the root file system, and ROOT_UNMOUNT if called to
1445  * unmount the root (e.g., as part of a system shutdown).
1446  *
1447  * XXX - this may be partially machine-dependent; it, along with the VFS_SWAPVP
1448  * operation, goes along with auto-configuration.  A mechanism should be
1449  * provided by which machine-INdependent code in the kernel can say "get me the
1450  * right root file system" and "get me the right initial swap area", and have
1451  * that done in what may well be a machine-dependent fashion.
1452  * Unfortunately, it is also file-system-type dependent (NFS gets it via
1453  * bootparams calls, UFS gets it from various and sundry machine-dependent
1454  * mechanisms, as SPECFS does for swap).
1455  */
1456 static int
1457 hsfs_mountroot(struct vfs *vfsp, enum whymountroot why)
1458 {
1459 	int error;
1460 	struct hsfs *fsp;
1461 	struct hs_volume *fvolp;
1462 	static int hsfsrootdone = 0;
1463 	dev_t rootdev;
1464 	mode_t mode = 0;
1465 
1466 	if (why == ROOT_INIT) {
1467 		if (hsfsrootdone++)
1468 			return (EBUSY);
1469 		rootdev = getrootdev();
1470 		if (rootdev == (dev_t)NODEV)
1471 			return (ENODEV);
1472 		vfsp->vfs_dev = rootdev;
1473 		vfsp->vfs_flag |= VFS_RDONLY;
1474 	} else if (why == ROOT_REMOUNT) {
1475 		cmn_err(CE_NOTE, "hsfs_mountroot: ROOT_REMOUNT");
1476 		return (0);
1477 	} else if (why == ROOT_UNMOUNT) {
1478 		return (0);
1479 	}
1480 	error = vfs_lock(vfsp);
1481 	if (error) {
1482 		cmn_err(CE_NOTE, "hsfs_mountroot: couldn't get vfs_lock");
1483 		return (error);
1484 	}
1485 
1486 	error = hs_mountfs(vfsp, rootdev, "/", mode, 1, CRED(), 1);
1487 	/*
1488 	 * XXX - assumes root device is not indirect, because we don't set
1489 	 * rootvp.  Is rootvp used for anything?  If so, make another arg
1490 	 * to mountfs.
1491 	 */
1492 	if (error) {
1493 		vfs_unlock(vfsp);
1494 		if (rootvp) {
1495 			VN_RELE(rootvp);
1496 			rootvp = (struct vnode *)0;
1497 		}
1498 		return (error);
1499 	}
1500 	if (why == ROOT_INIT)
1501 		vfs_add((struct vnode *)0, vfsp,
1502 		    (vfsp->vfs_flag & VFS_RDONLY) ? MS_RDONLY : 0);
1503 	vfs_unlock(vfsp);
1504 	fsp = VFS_TO_HSFS(vfsp);
1505 	fvolp = &fsp->hsfs_vol;
1506 #ifdef HSFS_CLKSET
1507 	if (fvolp->cre_date.tv_sec == 0) {
1508 		cmn_err(CE_NOTE, "hsfs_mountroot: cre_date.tv_sec == 0");
1509 		if (fvolp->mod_date.tv_sec == 0) {
1510 			cmn_err(CE_NOTE,
1511 			    "hsfs_mountroot: mod_date.tv_sec == 0");
1512 			cmn_err(CE_NOTE, "hsfs_mountroot: clkset(-1L)");
1513 			clkset(-1L);
1514 		} else {
1515 			clkset(fvolp->mod_date.tv_sec);
1516 		}
1517 	} else {
1518 		clkset(fvolp->mod_date.tv_sec);
1519 	}
1520 #else	/* HSFS_CLKSET */
1521 	clkset(-1L);
1522 #endif	/* HSFS_CLKSET */
1523 	return (0);
1524 }
1525 
1526 /*
1527  * hs_findvoldesc()
1528  *
1529  * Return the sector where the volume descriptor lives.  This is
1530  * a fixed value for "normal" cd-rom's, but can change for
1531  * multisession cd's.
1532  *
1533  * desc_sec is the same for high-sierra and iso 9660 formats, why
1534  * there are two different #defines used in the code for this is
1535  * beyond me.  These are standards, cast in concrete, right?
1536  * To be general, however, this function supports passing in different
1537  * values.
1538  */
1539 static int
1540 hs_findvoldesc(dev_t rdev, int desc_sec)
1541 {
1542 	int secno;
1543 	int error;
1544 	int rval;	/* ignored */
1545 
1546 #ifdef CDROMREADOFFSET
1547 	/*
1548 	 * Issue the Read Offset ioctl directly to the
1549 	 * device. Ignore any errors and set starting
1550 	 * secno to the default, otherwise add the
1551 	 * VOLDESC sector number to the offset.
1552 	 */
1553 	error = cdev_ioctl(rdev, CDROMREADOFFSET, (intptr_t)&secno,
1554 	    FNATIVE|FKIOCTL|FREAD, CRED(), &rval);
1555 	if (error) {
1556 		secno = desc_sec;
1557 	} else {
1558 		secno += desc_sec;
1559 	}
1560 #else
1561 	secno = desc_sec;
1562 #endif
1563 
1564 	return (secno);
1565 }
1566