1/*-
2 * Copyright (c) 1994
3 *	The Regents of the University of California.  All rights reserved.
4 *
5 * This code is derived from software contributed to Berkeley
6 * by Pace Willisson (pace@blitz.com).  The Rock Ridge Extension
7 * Support code is derived from software contributed to Berkeley
8 * by Atsushi Murai (amurai@spec.co.jp).
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 *    notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 *    notice, this list of conditions and the following disclaimer in the
17 *    documentation and/or other materials provided with the distribution.
18 * 4. Neither the name of the University nor the names of its contributors
19 *    may be used to endorse or promote products derived from this software
20 *    without specific prior written permission.
21 *
22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * SUCH DAMAGE.
33 *
34 *	@(#)cd9660_vfsops.c	8.18 (Berkeley) 5/22/95
35 */
36
37#include <sys/cdefs.h>
38__FBSDID("$FreeBSD$");
39
40#include <sys/param.h>
41#include <sys/systm.h>
42#include <sys/namei.h>
43#include <sys/priv.h>
44#include <sys/proc.h>
45#include <sys/kernel.h>
46#include <sys/vnode.h>
47#include <sys/mount.h>
48#include <sys/bio.h>
49#include <sys/buf.h>
50#include <sys/cdio.h>
51#include <sys/conf.h>
52#include <sys/fcntl.h>
53#include <sys/malloc.h>
54#include <sys/stat.h>
55#include <sys/syslog.h>
56#include <sys/iconv.h>
57
58#include <fs/cd9660/iso.h>
59#include <fs/cd9660/iso_rrip.h>
60#include <fs/cd9660/cd9660_node.h>
61#include <fs/cd9660/cd9660_mount.h>
62
63#include <geom/geom.h>
64#include <geom/geom_vfs.h>
65
66MALLOC_DEFINE(M_ISOFSMNT, "isofs_mount", "ISOFS mount structure");
67MALLOC_DEFINE(M_ISOFSNODE, "isofs_node", "ISOFS vnode private part");
68
69struct iconv_functions *cd9660_iconv = NULL;
70
71static vfs_mount_t	cd9660_mount;
72static vfs_cmount_t	cd9660_cmount;
73static vfs_unmount_t	cd9660_unmount;
74static vfs_root_t	cd9660_root;
75static vfs_statfs_t	cd9660_statfs;
76static vfs_vget_t	cd9660_vget;
77static vfs_fhtovp_t	cd9660_fhtovp;
78
79static struct vfsops cd9660_vfsops = {
80	.vfs_fhtovp =		cd9660_fhtovp,
81	.vfs_mount =		cd9660_mount,
82	.vfs_cmount =		cd9660_cmount,
83	.vfs_root =		cd9660_root,
84	.vfs_statfs =		cd9660_statfs,
85	.vfs_unmount =		cd9660_unmount,
86	.vfs_vget =		cd9660_vget,
87};
88VFS_SET(cd9660_vfsops, cd9660, VFCF_READONLY);
89MODULE_VERSION(cd9660, 1);
90
91static int iso_mountfs(struct vnode *devvp, struct mount *mp);
92
93/*
94 * VFS Operations.
95 */
96
97static int
98cd9660_cmount(struct mntarg *ma, void *data, uint64_t flags)
99{
100	struct iso_args args;
101	struct export_args exp;
102	int error;
103
104	error = copyin(data, &args, sizeof args);
105	if (error)
106		return (error);
107	vfs_oexport_conv(&args.export, &exp);
108
109	ma = mount_argsu(ma, "from", args.fspec, MAXPATHLEN);
110	ma = mount_arg(ma, "export", &exp, sizeof(exp));
111	ma = mount_argsu(ma, "cs_disk", args.cs_disk, 64);
112	ma = mount_argsu(ma, "cs_local", args.cs_local, 64);
113	ma = mount_argf(ma, "ssector", "%u", args.ssector);
114	ma = mount_argb(ma, !(args.flags & ISOFSMNT_NORRIP), "norrip");
115	ma = mount_argb(ma, args.flags & ISOFSMNT_GENS, "nogens");
116	ma = mount_argb(ma, args.flags & ISOFSMNT_EXTATT, "noextatt");
117	ma = mount_argb(ma, !(args.flags & ISOFSMNT_NOJOLIET), "nojoliet");
118	ma = mount_argb(ma,
119	    args.flags & ISOFSMNT_BROKENJOLIET, "nobrokenjoliet");
120	ma = mount_argb(ma, args.flags & ISOFSMNT_KICONV, "nokiconv");
121
122	error = kernel_mount(ma, flags);
123
124	return (error);
125}
126
127static int
128cd9660_mount(struct mount *mp)
129{
130	struct vnode *devvp;
131	struct thread *td;
132	char *fspec;
133	int error;
134	accmode_t accmode;
135	struct nameidata ndp;
136	struct iso_mnt *imp = NULL;
137
138	td = curthread;
139
140	/*
141	 * Unconditionally mount as read-only.
142	 */
143	MNT_ILOCK(mp);
144	mp->mnt_flag |= MNT_RDONLY;
145	MNT_IUNLOCK(mp);
146
147	fspec = vfs_getopts(mp->mnt_optnew, "from", &error);
148	if (error)
149		return (error);
150
151	imp = VFSTOISOFS(mp);
152
153	if (mp->mnt_flag & MNT_UPDATE) {
154		if (vfs_flagopt(mp->mnt_optnew, "export", NULL, 0))
155			return (0);
156	}
157	/*
158	 * Not an update, or updating the name: look up the name
159	 * and verify that it refers to a sensible block device.
160	 */
161	NDINIT(&ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, fspec, td);
162	if ((error = namei(&ndp)))
163		return (error);
164	NDFREE(&ndp, NDF_ONLY_PNBUF);
165	devvp = ndp.ni_vp;
166
167	if (!vn_isdisk(devvp, &error)) {
168		vput(devvp);
169		return (error);
170	}
171
172	/*
173	 * Verify that user has necessary permissions on the device,
174	 * or has superuser abilities
175	 */
176	accmode = VREAD;
177	error = VOP_ACCESS(devvp, accmode, td->td_ucred, td);
178	if (error)
179		error = priv_check(td, PRIV_VFS_MOUNT_PERM);
180	if (error) {
181		vput(devvp);
182		return (error);
183	}
184
185	if ((mp->mnt_flag & MNT_UPDATE) == 0) {
186		error = iso_mountfs(devvp, mp);
187		if (error)
188			vrele(devvp);
189	} else {
190		if (devvp != imp->im_devvp)
191			error = EINVAL;	/* needs translation */
192		vput(devvp);
193	}
194	if (error)
195		return (error);
196	vfs_mountedfrom(mp, fspec);
197	return (0);
198}
199
200/*
201 * Common code for mount and mountroot
202 */
203static int
204iso_mountfs(devvp, mp)
205	struct vnode *devvp;
206	struct mount *mp;
207{
208	struct iso_mnt *isomp = NULL;
209	struct buf *bp = NULL;
210	struct buf *pribp = NULL, *supbp = NULL;
211	struct cdev *dev;
212	int error = EINVAL;
213	int high_sierra = 0;
214	int iso_bsize;
215	int iso_blknum;
216	int joliet_level;
217	struct iso_volume_descriptor *vdp = NULL;
218	struct iso_primary_descriptor *pri = NULL;
219	struct iso_sierra_primary_descriptor *pri_sierra = NULL;
220	struct iso_supplementary_descriptor *sup = NULL;
221	struct iso_directory_record *rootp;
222	int logical_block_size, ssector;
223	struct g_consumer *cp;
224	struct bufobj *bo;
225	char *cs_local, *cs_disk;
226
227	dev = devvp->v_rdev;
228	dev_ref(dev);
229	DROP_GIANT();
230	g_topology_lock();
231	error = g_vfs_open(devvp, &cp, "cd9660", 0);
232	g_topology_unlock();
233	PICKUP_GIANT();
234	VOP_UNLOCK(devvp, 0);
235	if (error)
236		goto out;
237	if (devvp->v_rdev->si_iosize_max != 0)
238		mp->mnt_iosize_max = devvp->v_rdev->si_iosize_max;
239	if (mp->mnt_iosize_max > MAXPHYS)
240		mp->mnt_iosize_max = MAXPHYS;
241
242	bo = &devvp->v_bufobj;
243
244	/* This is the "logical sector size".  The standard says this
245	 * should be 2048 or the physical sector size on the device,
246	 * whichever is greater.
247	 */
248	if ((ISO_DEFAULT_BLOCK_SIZE % cp->provider->sectorsize) != 0) {
249		error = EINVAL;
250		goto out;
251	}
252
253	iso_bsize = cp->provider->sectorsize;
254
255	joliet_level = 0;
256	if (1 != vfs_scanopt(mp->mnt_optnew, "ssector", "%d", &ssector))
257		ssector = 0;
258	for (iso_blknum = 16 + ssector;
259	     iso_blknum < 100 + ssector;
260	     iso_blknum++) {
261		if ((error = bread(devvp, iso_blknum * btodb(ISO_DEFAULT_BLOCK_SIZE),
262				  iso_bsize, NOCRED, &bp)) != 0)
263			goto out;
264
265		vdp = (struct iso_volume_descriptor *)bp->b_data;
266		if (bcmp (vdp->id, ISO_STANDARD_ID, sizeof vdp->id) != 0) {
267			if (bcmp (vdp->id_sierra, ISO_SIERRA_ID,
268				  sizeof vdp->id_sierra) != 0) {
269				error = EINVAL;
270				goto out;
271			} else
272				high_sierra = 1;
273		}
274		switch (isonum_711 (high_sierra? vdp->type_sierra: vdp->type)){
275		case ISO_VD_PRIMARY:
276			if (pribp == NULL) {
277				pribp = bp;
278				bp = NULL;
279				pri = (struct iso_primary_descriptor *)vdp;
280				pri_sierra =
281				  (struct iso_sierra_primary_descriptor *)vdp;
282			}
283			break;
284
285		case ISO_VD_SUPPLEMENTARY:
286			if (supbp == NULL) {
287				supbp = bp;
288				bp = NULL;
289				sup = (struct iso_supplementary_descriptor *)vdp;
290
291				if (!vfs_flagopt(mp->mnt_optnew, "nojoliet", NULL, 0)) {
292					if (bcmp(sup->escape, "%/@", 3) == 0)
293						joliet_level = 1;
294					if (bcmp(sup->escape, "%/C", 3) == 0)
295						joliet_level = 2;
296					if (bcmp(sup->escape, "%/E", 3) == 0)
297						joliet_level = 3;
298
299					if ((isonum_711 (sup->flags) & 1) &&
300					    !vfs_flagopt(mp->mnt_optnew, "brokenjoliet", NULL, 0))
301						joliet_level = 0;
302				}
303			}
304			break;
305
306		case ISO_VD_END:
307			goto vd_end;
308
309		default:
310			break;
311		}
312		if (bp) {
313			brelse(bp);
314			bp = NULL;
315		}
316	}
317 vd_end:
318	if (bp) {
319		brelse(bp);
320		bp = NULL;
321	}
322
323	if (pri == NULL) {
324		error = EINVAL;
325		goto out;
326	}
327
328	logical_block_size =
329		isonum_723 (high_sierra?
330			    pri_sierra->logical_block_size:
331			    pri->logical_block_size);
332
333	if (logical_block_size < DEV_BSIZE || logical_block_size > MAXBSIZE
334	    || (logical_block_size & (logical_block_size - 1)) != 0) {
335		error = EINVAL;
336		goto out;
337	}
338
339	rootp = (struct iso_directory_record *)
340		(high_sierra?
341		 pri_sierra->root_directory_record:
342		 pri->root_directory_record);
343
344	isomp = malloc(sizeof *isomp, M_ISOFSMNT, M_WAITOK | M_ZERO);
345	isomp->im_cp = cp;
346	isomp->im_bo = bo;
347	isomp->logical_block_size = logical_block_size;
348	isomp->volume_space_size =
349		isonum_733 (high_sierra?
350			    pri_sierra->volume_space_size:
351			    pri->volume_space_size);
352	isomp->joliet_level = 0;
353	/*
354	 * Since an ISO9660 multi-session CD can also access previous
355	 * sessions, we have to include them into the space consider-
356	 * ations.  This doesn't yield a very accurate number since
357	 * parts of the old sessions might be inaccessible now, but we
358	 * can't do much better.  This is also important for the NFS
359	 * filehandle validation.
360	 */
361	isomp->volume_space_size += ssector;
362	bcopy (rootp, isomp->root, sizeof isomp->root);
363	isomp->root_extent = isonum_733 (rootp->extent);
364	isomp->root_size = isonum_733 (rootp->size);
365
366	isomp->im_bmask = logical_block_size - 1;
367	isomp->im_bshift = ffs(logical_block_size) - 1;
368
369	pribp->b_flags |= B_AGE;
370	brelse(pribp);
371	pribp = NULL;
372	rootp = NULL;
373	pri = NULL;
374	pri_sierra = NULL;
375
376	mp->mnt_data = isomp;
377	mp->mnt_stat.f_fsid.val[0] = dev2udev(dev);
378	mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
379	mp->mnt_maxsymlinklen = 0;
380	MNT_ILOCK(mp);
381	mp->mnt_flag |= MNT_LOCAL;
382	mp->mnt_kern_flag |= MNTK_LOOKUP_SHARED | MNTK_EXTENDED_SHARED;
383	MNT_IUNLOCK(mp);
384	isomp->im_mountp = mp;
385	isomp->im_dev = dev;
386	isomp->im_devvp = devvp;
387
388	vfs_flagopt(mp->mnt_optnew, "norrip", &isomp->im_flags, ISOFSMNT_NORRIP);
389	vfs_flagopt(mp->mnt_optnew, "gens", &isomp->im_flags, ISOFSMNT_GENS);
390	vfs_flagopt(mp->mnt_optnew, "extatt", &isomp->im_flags, ISOFSMNT_EXTATT);
391	vfs_flagopt(mp->mnt_optnew, "nojoliet", &isomp->im_flags, ISOFSMNT_NOJOLIET);
392	vfs_flagopt(mp->mnt_optnew, "kiconv", &isomp->im_flags, ISOFSMNT_KICONV);
393
394	/* Check the Rock Ridge Extension support */
395	if (!(isomp->im_flags & ISOFSMNT_NORRIP)) {
396		if ((error = bread(isomp->im_devvp, (isomp->root_extent +
397		    isonum_711(((struct iso_directory_record *)isomp->root)->
398		    ext_attr_length)) << (isomp->im_bshift - DEV_BSHIFT),
399		    isomp->logical_block_size, NOCRED, &bp)) != 0)
400			goto out;
401
402		rootp = (struct iso_directory_record *)bp->b_data;
403
404		if ((isomp->rr_skip = cd9660_rrip_offset(rootp,isomp)) < 0) {
405		    isomp->im_flags |= ISOFSMNT_NORRIP;
406		} else {
407		    isomp->im_flags &= ~ISOFSMNT_GENS;
408		}
409
410		/*
411		 * The contents are valid,
412		 * but they will get reread as part of another vnode, so...
413		 */
414		bp->b_flags |= B_AGE;
415		brelse(bp);
416		bp = NULL;
417		rootp = NULL;
418	}
419
420	if (isomp->im_flags & ISOFSMNT_KICONV && cd9660_iconv) {
421		cs_local = vfs_getopts(mp->mnt_optnew, "cs_local", &error);
422		if (error)
423			goto out;
424		cs_disk = vfs_getopts(mp->mnt_optnew, "cs_disk", &error);
425		if (error)
426			goto out;
427		cd9660_iconv->open(cs_local, cs_disk, &isomp->im_d2l);
428		cd9660_iconv->open(cs_disk, cs_local, &isomp->im_l2d);
429	} else {
430		isomp->im_d2l = NULL;
431		isomp->im_l2d = NULL;
432	}
433
434	if (high_sierra) {
435		/* this effectively ignores all the mount flags */
436		if (bootverbose)
437			log(LOG_INFO, "cd9660: High Sierra Format\n");
438		isomp->iso_ftype = ISO_FTYPE_HIGH_SIERRA;
439	} else
440		switch (isomp->im_flags&(ISOFSMNT_NORRIP|ISOFSMNT_GENS)) {
441		  default:
442			  isomp->iso_ftype = ISO_FTYPE_DEFAULT;
443			  break;
444		  case ISOFSMNT_GENS|ISOFSMNT_NORRIP:
445			  isomp->iso_ftype = ISO_FTYPE_9660;
446			  break;
447		  case 0:
448			  if (bootverbose)
449			  	  log(LOG_INFO, "cd9660: RockRidge Extension\n");
450			  isomp->iso_ftype = ISO_FTYPE_RRIP;
451			  break;
452		}
453
454	/* Decide whether to use the Joliet descriptor */
455
456	if (isomp->iso_ftype != ISO_FTYPE_RRIP && joliet_level) {
457		if (bootverbose)
458			log(LOG_INFO, "cd9660: Joliet Extension (Level %d)\n",
459			    joliet_level);
460		rootp = (struct iso_directory_record *)
461			sup->root_directory_record;
462		bcopy (rootp, isomp->root, sizeof isomp->root);
463		isomp->root_extent = isonum_733 (rootp->extent);
464		isomp->root_size = isonum_733 (rootp->size);
465		isomp->joliet_level = joliet_level;
466		supbp->b_flags |= B_AGE;
467	}
468
469	if (supbp) {
470		brelse(supbp);
471		supbp = NULL;
472		sup = NULL;
473	}
474
475	return 0;
476out:
477	if (bp)
478		brelse(bp);
479	if (pribp)
480		brelse(pribp);
481	if (supbp)
482		brelse(supbp);
483	if (cp != NULL) {
484		DROP_GIANT();
485		g_topology_lock();
486		g_vfs_close(cp);
487		g_topology_unlock();
488		PICKUP_GIANT();
489	}
490	if (isomp) {
491		free(isomp, M_ISOFSMNT);
492		mp->mnt_data = NULL;
493	}
494	dev_rel(dev);
495	return error;
496}
497
498/*
499 * unmount system call
500 */
501static int
502cd9660_unmount(mp, mntflags)
503	struct mount *mp;
504	int mntflags;
505{
506	struct iso_mnt *isomp;
507	int error, flags = 0;
508
509	if (mntflags & MNT_FORCE)
510		flags |= FORCECLOSE;
511	if ((error = vflush(mp, 0, flags, curthread)))
512		return (error);
513
514	isomp = VFSTOISOFS(mp);
515
516	if (isomp->im_flags & ISOFSMNT_KICONV && cd9660_iconv) {
517		if (isomp->im_d2l)
518			cd9660_iconv->close(isomp->im_d2l);
519		if (isomp->im_l2d)
520			cd9660_iconv->close(isomp->im_l2d);
521	}
522	DROP_GIANT();
523	g_topology_lock();
524	g_vfs_close(isomp->im_cp);
525	g_topology_unlock();
526	PICKUP_GIANT();
527	vrele(isomp->im_devvp);
528	dev_rel(isomp->im_dev);
529	free(isomp, M_ISOFSMNT);
530	mp->mnt_data = NULL;
531	MNT_ILOCK(mp);
532	mp->mnt_flag &= ~MNT_LOCAL;
533	MNT_IUNLOCK(mp);
534	return (error);
535}
536
537/*
538 * Return root of a filesystem
539 */
540static int
541cd9660_root(mp, flags, vpp)
542	struct mount *mp;
543	int flags;
544	struct vnode **vpp;
545{
546	struct iso_mnt *imp = VFSTOISOFS(mp);
547	struct iso_directory_record *dp =
548	    (struct iso_directory_record *)imp->root;
549	ino_t ino = isodirino(dp, imp);
550
551	/*
552	 * With RRIP we must use the `.' entry of the root directory.
553	 * Simply tell vget, that it's a relocated directory.
554	 */
555	return (cd9660_vget_internal(mp, ino, flags, vpp,
556	    imp->iso_ftype == ISO_FTYPE_RRIP, dp));
557}
558
559/*
560 * Get filesystem statistics.
561 */
562static int
563cd9660_statfs(mp, sbp)
564	struct mount *mp;
565	struct statfs *sbp;
566{
567	struct iso_mnt *isomp;
568
569	isomp = VFSTOISOFS(mp);
570
571	sbp->f_bsize = isomp->logical_block_size;
572	sbp->f_iosize = sbp->f_bsize;	/* XXX */
573	sbp->f_blocks = isomp->volume_space_size;
574	sbp->f_bfree = 0; /* total free blocks */
575	sbp->f_bavail = 0; /* blocks free for non superuser */
576	sbp->f_files =	0; /* total files */
577	sbp->f_ffree = 0; /* free file nodes */
578	return 0;
579}
580
581/*
582 * File handle to vnode
583 *
584 * Have to be really careful about stale file handles:
585 * - check that the inode number is in range
586 * - call iget() to get the locked inode
587 * - check for an unallocated inode (i_mode == 0)
588 * - check that the generation number matches
589 */
590
591/* ARGSUSED */
592static int
593cd9660_fhtovp(mp, fhp, flags, vpp)
594	struct mount *mp;
595	struct fid *fhp;
596	int flags;
597	struct vnode **vpp;
598{
599	struct ifid ifh;
600	struct iso_node *ip;
601	struct vnode *nvp;
602	int error;
603
604	memcpy(&ifh, fhp, sizeof(ifh));
605
606#ifdef	ISOFS_DBG
607	printf("fhtovp: ino %d, start %ld\n",
608	    ifh.ifid_ino, ifh.ifid_start);
609#endif
610
611	if ((error = VFS_VGET(mp, ifh.ifid_ino, LK_EXCLUSIVE, &nvp)) != 0) {
612		*vpp = NULLVP;
613		return (error);
614	}
615	ip = VTOI(nvp);
616	if (ip->inode.iso_mode == 0) {
617		vput(nvp);
618		*vpp = NULLVP;
619		return (ESTALE);
620	}
621	*vpp = nvp;
622	vnode_create_vobject(*vpp, ip->i_size, curthread);
623	return (0);
624}
625
626static int
627cd9660_vget(mp, ino, flags, vpp)
628	struct mount *mp;
629	ino_t ino;
630	int flags;
631	struct vnode **vpp;
632{
633
634	/*
635	 * XXXX
636	 * It would be nice if we didn't always set the `relocated' flag
637	 * and force the extra read, but I don't want to think about fixing
638	 * that right now.
639	 */
640	return (cd9660_vget_internal(mp, ino, flags, vpp,
641#if 0
642	    VFSTOISOFS(mp)->iso_ftype == ISO_FTYPE_RRIP,
643#else
644	    0,
645#endif
646	    (struct iso_directory_record *)0));
647}
648
649int
650cd9660_vget_internal(mp, ino, flags, vpp, relocated, isodir)
651	struct mount *mp;
652	ino_t ino;
653	int flags;
654	struct vnode **vpp;
655	int relocated;
656	struct iso_directory_record *isodir;
657{
658	struct iso_mnt *imp;
659	struct iso_node *ip;
660	struct buf *bp;
661	struct vnode *vp;
662	struct cdev *dev;
663	int error;
664	struct thread *td;
665
666	td = curthread;
667	error = vfs_hash_get(mp, ino, flags, td, vpp, NULL, NULL);
668	if (error || *vpp != NULL)
669		return (error);
670
671	/*
672	 * We must promote to an exclusive lock for vnode creation.  This
673	 * can happen if lookup is passed LOCKSHARED.
674 	 */
675	if ((flags & LK_TYPE_MASK) == LK_SHARED) {
676		flags &= ~LK_TYPE_MASK;
677		flags |= LK_EXCLUSIVE;
678	}
679
680	/*
681	 * We do not lock vnode creation as it is believed to be too
682	 * expensive for such rare case as simultaneous creation of vnode
683	 * for same ino by different processes. We just allow them to race
684	 * and check later to decide who wins. Let the race begin!
685	 */
686
687	imp = VFSTOISOFS(mp);
688	dev = imp->im_dev;
689
690	/* Allocate a new vnode/iso_node. */
691	if ((error = getnewvnode("isofs", mp, &cd9660_vnodeops, &vp)) != 0) {
692		*vpp = NULLVP;
693		return (error);
694	}
695	ip = malloc(sizeof(struct iso_node), M_ISOFSNODE,
696	    M_WAITOK | M_ZERO);
697	vp->v_data = ip;
698	ip->i_vnode = vp;
699	ip->i_number = ino;
700
701	lockmgr(vp->v_vnlock, LK_EXCLUSIVE, NULL);
702	error = insmntque(vp, mp);
703	if (error != 0) {
704		free(ip, M_ISOFSNODE);
705		*vpp = NULLVP;
706		return (error);
707	}
708	error = vfs_hash_insert(vp, ino, flags, td, vpp, NULL, NULL);
709	if (error || *vpp != NULL)
710		return (error);
711
712	if (isodir == 0) {
713		int lbn, off;
714
715		lbn = lblkno(imp, ino);
716		if (lbn >= imp->volume_space_size) {
717			vput(vp);
718			printf("fhtovp: lbn exceed volume space %d\n", lbn);
719			return (ESTALE);
720		}
721
722		off = blkoff(imp, ino);
723		if (off + ISO_DIRECTORY_RECORD_SIZE > imp->logical_block_size) {
724			vput(vp);
725			printf("fhtovp: crosses block boundary %d\n",
726			       off + ISO_DIRECTORY_RECORD_SIZE);
727			return (ESTALE);
728		}
729
730		error = bread(imp->im_devvp,
731			      lbn << (imp->im_bshift - DEV_BSHIFT),
732			      imp->logical_block_size, NOCRED, &bp);
733		if (error) {
734			vput(vp);
735			brelse(bp);
736			printf("fhtovp: bread error %d\n",error);
737			return (error);
738		}
739		isodir = (struct iso_directory_record *)(bp->b_data + off);
740
741		if (off + isonum_711(isodir->length) >
742		    imp->logical_block_size) {
743			vput(vp);
744			if (bp != 0)
745				brelse(bp);
746			printf("fhtovp: directory crosses block boundary %d[off=%d/len=%d]\n",
747			       off +isonum_711(isodir->length), off,
748			       isonum_711(isodir->length));
749			return (ESTALE);
750		}
751
752#if 0
753		if (isonum_733(isodir->extent) +
754		    isonum_711(isodir->ext_attr_length) != ifhp->ifid_start) {
755			if (bp != 0)
756				brelse(bp);
757			printf("fhtovp: file start miss %d vs %d\n",
758			       isonum_733(isodir->extent) + isonum_711(isodir->ext_attr_length),
759			       ifhp->ifid_start);
760			return (ESTALE);
761		}
762#endif
763	} else
764		bp = 0;
765
766	ip->i_mnt = imp;
767
768	if (relocated) {
769		/*
770		 * On relocated directories we must
771		 * read the `.' entry out of a dir.
772		 */
773		ip->iso_start = ino >> imp->im_bshift;
774		if (bp != 0)
775			brelse(bp);
776		if ((error = cd9660_blkatoff(vp, (off_t)0, NULL, &bp)) != 0) {
777			vput(vp);
778			return (error);
779		}
780		isodir = (struct iso_directory_record *)bp->b_data;
781	}
782
783	ip->iso_extent = isonum_733(isodir->extent);
784	ip->i_size = isonum_733(isodir->size);
785	ip->iso_start = isonum_711(isodir->ext_attr_length) + ip->iso_extent;
786
787	/*
788	 * Setup time stamp, attribute
789	 */
790	vp->v_type = VNON;
791	switch (imp->iso_ftype) {
792	default:	/* ISO_FTYPE_9660 */
793	    {
794		struct buf *bp2;
795		int off;
796		if ((imp->im_flags & ISOFSMNT_EXTATT)
797		    && (off = isonum_711(isodir->ext_attr_length)))
798			cd9660_blkatoff(vp, (off_t)-(off << imp->im_bshift), NULL,
799				     &bp2);
800		else
801			bp2 = NULL;
802		cd9660_defattr(isodir, ip, bp2, ISO_FTYPE_9660);
803		cd9660_deftstamp(isodir, ip, bp2, ISO_FTYPE_9660);
804		if (bp2)
805			brelse(bp2);
806		break;
807	    }
808	case ISO_FTYPE_RRIP:
809		cd9660_rrip_analyze(isodir, ip, imp);
810		break;
811	}
812
813	if (bp != 0)
814		brelse(bp);
815
816	/*
817	 * Initialize the associated vnode
818	 */
819	switch (vp->v_type = IFTOVT(ip->inode.iso_mode)) {
820	case VFIFO:
821		vp->v_op = &cd9660_fifoops;
822		break;
823	default:
824		VN_LOCK_ASHARE(vp);
825		break;
826	}
827
828	if (ip->iso_extent == imp->root_extent)
829		vp->v_vflag |= VV_ROOT;
830
831	/*
832	 * XXX need generation number?
833	 */
834
835	*vpp = vp;
836	return (0);
837}
838