xref: /illumos-gate/usr/src/uts/common/fs/smbclnt/smbfs/smbfs_vfsops.c (revision 67a4bb8f9ad4c49e9aa9e21e2114a7c093c3a73a)
1 /*
2  * Copyright (c) 2000-2001, Boris Popov
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *    This product includes software developed by Boris Popov.
16  * 4. Neither the name of the author nor the names of any co-contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  *
32  * $Id: smbfs_vfsops.c,v 1.73.64.1 2005/05/27 02:35:28 lindak Exp $
33  */
34 
35 /*
36  * Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved.
37  * Copyright 2013, Joyent, Inc. All rights reserved.
38  */
39 
40 #include <sys/systm.h>
41 #include <sys/cred.h>
42 #include <sys/time.h>
43 #include <sys/vfs.h>
44 #include <sys/vnode.h>
45 #include <fs/fs_subr.h>
46 #include <sys/sysmacros.h>
47 #include <sys/kmem.h>
48 #include <sys/mkdev.h>
49 #include <sys/mount.h>
50 #include <sys/statvfs.h>
51 #include <sys/errno.h>
52 #include <sys/debug.h>
53 #include <sys/cmn_err.h>
54 #include <sys/modctl.h>
55 #include <sys/policy.h>
56 #include <sys/atomic.h>
57 #include <sys/zone.h>
58 #include <sys/vfs_opreg.h>
59 #include <sys/mntent.h>
60 #include <sys/priv.h>
61 #include <sys/tsol/label.h>
62 #include <sys/tsol/tndb.h>
63 #include <inet/ip.h>
64 
65 #include <netsmb/smb_osdep.h>
66 #include <netsmb/smb.h>
67 #include <netsmb/smb_conn.h>
68 #include <netsmb/smb_subr.h>
69 #include <netsmb/smb_dev.h>
70 
71 #include <smbfs/smbfs.h>
72 #include <smbfs/smbfs_node.h>
73 #include <smbfs/smbfs_subr.h>
74 
75 /*
76  * Local functions definitions.
77  */
78 int		smbfsinit(int fstyp, char *name);
79 void		smbfsfini();
80 static int	smbfs_mount_label_policy(vfs_t *, void *, int, cred_t *);
81 
82 /*
83  * SMBFS Mount options table for MS_OPTIONSTR
84  * Note: These are not all the options.
85  * Some options come in via MS_DATA.
86  * Others are generic (see vfs.c)
87  */
88 static char *intr_cancel[] = { MNTOPT_NOINTR, NULL };
89 static char *nointr_cancel[] = { MNTOPT_INTR, NULL };
90 static char *acl_cancel[] = { MNTOPT_NOACL, NULL };
91 static char *noacl_cancel[] = { MNTOPT_ACL, NULL };
92 static char *xattr_cancel[] = { MNTOPT_NOXATTR, NULL };
93 static char *noxattr_cancel[] = { MNTOPT_XATTR, NULL };
94 
95 static mntopt_t mntopts[] = {
96 /*
97  *	option name		cancel option	default arg	flags
98  *		ufs arg flag
99  */
100 	{ MNTOPT_INTR,		intr_cancel,	NULL,	MO_DEFAULT, 0 },
101 	{ MNTOPT_NOINTR,	nointr_cancel,	NULL,	0,	0 },
102 	{ MNTOPT_ACL,		acl_cancel,	NULL,	MO_DEFAULT, 0 },
103 	{ MNTOPT_NOACL,		noacl_cancel,	NULL,	0,	0 },
104 	{ MNTOPT_XATTR,		xattr_cancel,	NULL,	MO_DEFAULT, 0 },
105 	{ MNTOPT_NOXATTR,	noxattr_cancel, NULL,	0,	0 }
106 };
107 
108 static mntopts_t smbfs_mntopts = {
109 	sizeof (mntopts) / sizeof (mntopt_t),
110 	mntopts
111 };
112 
113 static const char fs_type_name[FSTYPSZ] = "smbfs";
114 
115 static vfsdef_t vfw = {
116 	VFSDEF_VERSION,
117 	(char *)fs_type_name,
118 	smbfsinit,		/* init routine */
119 	VSW_HASPROTO|VSW_NOTZONESAFE,	/* flags */
120 	&smbfs_mntopts			/* mount options table prototype */
121 };
122 
123 static struct modlfs modlfs = {
124 	&mod_fsops,
125 	"SMBFS filesystem",
126 	&vfw
127 };
128 
129 static struct modlinkage modlinkage = {
130 	MODREV_1, (void *)&modlfs, NULL
131 };
132 
133 /*
134  * Mutex to protect the following variables:
135  *	  smbfs_major
136  *	  smbfs_minor
137  */
138 extern	kmutex_t	smbfs_minor_lock;
139 extern	int		smbfs_major;
140 extern	int		smbfs_minor;
141 
142 /*
143  * Prevent unloads while we have mounts
144  */
145 uint32_t	smbfs_mountcount;
146 
147 /*
148  * smbfs vfs operations.
149  */
150 static int	smbfs_mount(vfs_t *, vnode_t *, struct mounta *, cred_t *);
151 static int	smbfs_unmount(vfs_t *, int, cred_t *);
152 static int	smbfs_root(vfs_t *, vnode_t **);
153 static int	smbfs_statvfs(vfs_t *, statvfs64_t *);
154 static int	smbfs_sync(vfs_t *, short, cred_t *);
155 static void	smbfs_freevfs(vfs_t *);
156 
157 /*
158  * Module loading
159  */
160 
161 /*
162  * This routine is invoked automatically when the kernel module
163  * containing this routine is loaded.  This allows module specific
164  * initialization to be done when the module is loaded.
165  */
166 int
167 _init(void)
168 {
169 	int		error;
170 
171 	/*
172 	 * Check compiled-in version of "nsmb"
173 	 * that we're linked with.  (paranoid)
174 	 */
175 	if (nsmb_version != NSMB_VERSION) {
176 		cmn_err(CE_WARN, "_init: nsmb version mismatch");
177 		return (ENOTTY);
178 	}
179 
180 	smbfs_mountcount = 0;
181 
182 	/*
183 	 * NFS calls these two in _clntinit
184 	 * Easier to follow this way.
185 	 */
186 	if ((error = smbfs_subrinit()) != 0) {
187 		cmn_err(CE_WARN, "_init: smbfs_subrinit failed");
188 		return (error);
189 	}
190 
191 	if ((error = smbfs_vfsinit()) != 0) {
192 		cmn_err(CE_WARN, "_init: smbfs_vfsinit failed");
193 		smbfs_subrfini();
194 		return (error);
195 	}
196 
197 	if ((error = smbfs_clntinit()) != 0) {
198 		cmn_err(CE_WARN, "_init: smbfs_clntinit failed");
199 		smbfs_vfsfini();
200 		smbfs_subrfini();
201 		return (error);
202 	}
203 
204 	error = mod_install((struct modlinkage *)&modlinkage);
205 	return (error);
206 }
207 
208 /*
209  * Free kernel module resources that were allocated in _init
210  * and remove the linkage information into the kernel
211  */
212 int
213 _fini(void)
214 {
215 	int	error;
216 
217 	/*
218 	 * If a forcedly unmounted instance is still hanging around,
219 	 * we cannot allow the module to be unloaded because that would
220 	 * cause panics once the VFS framework decides it's time to call
221 	 * into VFS_FREEVFS().
222 	 */
223 	if (smbfs_mountcount)
224 		return (EBUSY);
225 
226 	error = mod_remove(&modlinkage);
227 	if (error)
228 		return (error);
229 
230 	/*
231 	 * Free the allocated smbnodes, etc.
232 	 */
233 	smbfs_clntfini();
234 
235 	/* NFS calls these two in _clntfini */
236 	smbfs_vfsfini();
237 	smbfs_subrfini();
238 
239 	/*
240 	 * Free the ops vectors
241 	 */
242 	smbfsfini();
243 	return (0);
244 }
245 
246 /*
247  * Return information about the module
248  */
249 int
250 _info(struct modinfo *modinfop)
251 {
252 	return (mod_info((struct modlinkage *)&modlinkage, modinfop));
253 }
254 
255 /*
256  * Initialize the vfs structure
257  */
258 
259 int smbfsfstyp;
260 vfsops_t *smbfs_vfsops = NULL;
261 
262 static const fs_operation_def_t smbfs_vfsops_template[] = {
263 	{ VFSNAME_MOUNT, { .vfs_mount = smbfs_mount } },
264 	{ VFSNAME_UNMOUNT, { .vfs_unmount = smbfs_unmount } },
265 	{ VFSNAME_ROOT,	{ .vfs_root = smbfs_root } },
266 	{ VFSNAME_STATVFS, { .vfs_statvfs = smbfs_statvfs } },
267 	{ VFSNAME_SYNC,	{ .vfs_sync = smbfs_sync } },
268 	{ VFSNAME_VGET,	{ .error = fs_nosys } },
269 	{ VFSNAME_MOUNTROOT, { .error = fs_nosys } },
270 	{ VFSNAME_FREEVFS, { .vfs_freevfs = smbfs_freevfs } },
271 	{ NULL, NULL }
272 };
273 
274 int
275 smbfsinit(int fstyp, char *name)
276 {
277 	int		error;
278 
279 	error = vfs_setfsops(fstyp, smbfs_vfsops_template, &smbfs_vfsops);
280 	if (error != 0) {
281 		zcmn_err(GLOBAL_ZONEID, CE_WARN,
282 		    "smbfsinit: bad vfs ops template");
283 		return (error);
284 	}
285 
286 	error = vn_make_ops(name, smbfs_vnodeops_template, &smbfs_vnodeops);
287 	if (error != 0) {
288 		(void) vfs_freevfsops_by_type(fstyp);
289 		zcmn_err(GLOBAL_ZONEID, CE_WARN,
290 		    "smbfsinit: bad vnode ops template");
291 		return (error);
292 	}
293 
294 	smbfsfstyp = fstyp;
295 
296 	return (0);
297 }
298 
299 void
300 smbfsfini()
301 {
302 	if (smbfs_vfsops) {
303 		(void) vfs_freevfsops_by_type(smbfsfstyp);
304 		smbfs_vfsops = NULL;
305 	}
306 	if (smbfs_vnodeops) {
307 		vn_freevnodeops(smbfs_vnodeops);
308 		smbfs_vnodeops = NULL;
309 	}
310 }
311 
312 void
313 smbfs_free_smi(smbmntinfo_t *smi)
314 {
315 	if (smi == NULL)
316 		return;
317 
318 	if (smi->smi_zone_ref.zref_zone != NULL)
319 		zone_rele_ref(&smi->smi_zone_ref, ZONE_REF_SMBFS);
320 
321 	if (smi->smi_share != NULL)
322 		smb_share_rele(smi->smi_share);
323 
324 	avl_destroy(&smi->smi_hash_avl);
325 	rw_destroy(&smi->smi_hash_lk);
326 	cv_destroy(&smi->smi_statvfs_cv);
327 	mutex_destroy(&smi->smi_lock);
328 
329 	kmem_free(smi, sizeof (smbmntinfo_t));
330 }
331 
332 /*
333  * smbfs mount vfsop
334  * Set up mount info record and attach it to vfs struct.
335  */
336 static int
337 smbfs_mount(vfs_t *vfsp, vnode_t *mvp, struct mounta *uap, cred_t *cr)
338 {
339 	char		*data = uap->dataptr;
340 	int		error;
341 	smbnode_t 	*rtnp = NULL;	/* root of this fs */
342 	smbmntinfo_t 	*smi = NULL;
343 	dev_t 		smbfs_dev;
344 	int 		version;
345 	int 		devfd;
346 	zone_t		*zone = curproc->p_zone;
347 	zone_t		*mntzone = NULL;
348 	smb_share_t 	*ssp = NULL;
349 	smb_cred_t 	scred;
350 	int		flags, sec;
351 
352 	STRUCT_DECL(smbfs_args, args);		/* smbfs mount arguments */
353 
354 	if ((error = secpolicy_fs_mount(cr, mvp, vfsp)) != 0)
355 		return (error);
356 
357 	if (mvp->v_type != VDIR)
358 		return (ENOTDIR);
359 
360 	/*
361 	 * get arguments
362 	 *
363 	 * uap->datalen might be different from sizeof (args)
364 	 * in a compatible situation.
365 	 */
366 	STRUCT_INIT(args, get_udatamodel());
367 	bzero(STRUCT_BUF(args), SIZEOF_STRUCT(smbfs_args, DATAMODEL_NATIVE));
368 	if (copyin(data, STRUCT_BUF(args), MIN(uap->datalen,
369 	    SIZEOF_STRUCT(smbfs_args, DATAMODEL_NATIVE))))
370 		return (EFAULT);
371 
372 	/*
373 	 * Check mount program version
374 	 */
375 	version = STRUCT_FGET(args, version);
376 	if (version != SMBFS_VERSION) {
377 		cmn_err(CE_WARN, "mount version mismatch:"
378 		    " kernel=%d, mount=%d\n",
379 		    SMBFS_VERSION, version);
380 		return (EINVAL);
381 	}
382 
383 	/*
384 	 * Deal with re-mount requests.
385 	 */
386 	if (uap->flags & MS_REMOUNT) {
387 		cmn_err(CE_WARN, "MS_REMOUNT not implemented");
388 		return (ENOTSUP);
389 	}
390 
391 	/*
392 	 * Check for busy
393 	 */
394 	mutex_enter(&mvp->v_lock);
395 	if (!(uap->flags & MS_OVERLAY) &&
396 	    (mvp->v_count != 1 || (mvp->v_flag & VROOT))) {
397 		mutex_exit(&mvp->v_lock);
398 		return (EBUSY);
399 	}
400 	mutex_exit(&mvp->v_lock);
401 
402 	/*
403 	 * Get the "share" from the netsmb driver (ssp).
404 	 * It is returned with a "ref" (hold) for us.
405 	 * Release this hold: at errout below, or in
406 	 * smbfs_freevfs().
407 	 */
408 	devfd = STRUCT_FGET(args, devfd);
409 	error = smb_dev2share(devfd, &ssp);
410 	if (error) {
411 		cmn_err(CE_WARN, "invalid device handle %d (%d)\n",
412 		    devfd, error);
413 		return (error);
414 	}
415 
416 	/*
417 	 * Use "goto errout" from here on.
418 	 * See: ssp, smi, rtnp, mntzone
419 	 */
420 
421 	/*
422 	 * Determine the zone we're being mounted into.
423 	 */
424 	zone_hold(mntzone = zone);		/* start with this assumption */
425 	if (getzoneid() == GLOBAL_ZONEID) {
426 		zone_rele(mntzone);
427 		mntzone = zone_find_by_path(refstr_value(vfsp->vfs_mntpt));
428 		ASSERT(mntzone != NULL);
429 		if (mntzone != zone) {
430 			error = EBUSY;
431 			goto errout;
432 		}
433 	}
434 
435 	/*
436 	 * Stop the mount from going any further if the zone is going away.
437 	 */
438 	if (zone_status_get(mntzone) >= ZONE_IS_SHUTTING_DOWN) {
439 		error = EBUSY;
440 		goto errout;
441 	}
442 
443 	/*
444 	 * On a Trusted Extensions client, we may have to force read-only
445 	 * for read-down mounts.
446 	 */
447 	if (is_system_labeled()) {
448 		void *addr;
449 		int ipvers = 0;
450 		struct smb_vc *vcp;
451 
452 		vcp = SSTOVC(ssp);
453 		addr = smb_vc_getipaddr(vcp, &ipvers);
454 		error = smbfs_mount_label_policy(vfsp, addr, ipvers, cr);
455 
456 		if (error > 0)
457 			goto errout;
458 
459 		if (error == -1) {
460 			/* change mount to read-only to prevent write-down */
461 			vfs_setmntopt(vfsp, MNTOPT_RO, NULL, 0);
462 		}
463 	}
464 
465 	/* Prevent unload. */
466 	atomic_inc_32(&smbfs_mountcount);
467 
468 	/*
469 	 * Create a mount record and link it to the vfs struct.
470 	 * No more possiblities for errors from here on.
471 	 * Tear-down of this stuff is in smbfs_free_smi()
472 	 *
473 	 * Compare with NFS: nfsrootvp()
474 	 */
475 	smi = kmem_zalloc(sizeof (*smi), KM_SLEEP);
476 
477 	mutex_init(&smi->smi_lock, NULL, MUTEX_DEFAULT, NULL);
478 	cv_init(&smi->smi_statvfs_cv, NULL, CV_DEFAULT, NULL);
479 
480 	rw_init(&smi->smi_hash_lk, NULL, RW_DEFAULT, NULL);
481 	smbfs_init_hash_avl(&smi->smi_hash_avl);
482 
483 	smi->smi_share = ssp;
484 	ssp = NULL;
485 
486 	/*
487 	 * Convert the anonymous zone hold acquired via zone_hold() above
488 	 * into a zone reference.
489 	 */
490 	zone_init_ref(&smi->smi_zone_ref);
491 	zone_hold_ref(mntzone, &smi->smi_zone_ref, ZONE_REF_SMBFS);
492 	zone_rele(mntzone);
493 	mntzone = NULL;
494 
495 	/*
496 	 * Initialize option defaults
497 	 */
498 	smi->smi_flags	= SMI_LLOCK;
499 	smi->smi_acregmin = SEC2HR(SMBFS_ACREGMIN);
500 	smi->smi_acregmax = SEC2HR(SMBFS_ACREGMAX);
501 	smi->smi_acdirmin = SEC2HR(SMBFS_ACDIRMIN);
502 	smi->smi_acdirmax = SEC2HR(SMBFS_ACDIRMAX);
503 
504 	/*
505 	 * All "generic" mount options have already been
506 	 * handled in vfs.c:domount() - see mntopts stuff.
507 	 * Query generic options using vfs_optionisset().
508 	 */
509 	if (vfs_optionisset(vfsp, MNTOPT_INTR, NULL))
510 		smi->smi_flags |= SMI_INT;
511 	if (vfs_optionisset(vfsp, MNTOPT_ACL, NULL))
512 		smi->smi_flags |= SMI_ACL;
513 
514 	/*
515 	 * Get the mount options that come in as smbfs_args,
516 	 * starting with args.flags (SMBFS_MF_xxx)
517 	 */
518 	flags = STRUCT_FGET(args, flags);
519 	smi->smi_uid 	= STRUCT_FGET(args, uid);
520 	smi->smi_gid 	= STRUCT_FGET(args, gid);
521 	smi->smi_fmode	= STRUCT_FGET(args, file_mode) & 0777;
522 	smi->smi_dmode	= STRUCT_FGET(args, dir_mode) & 0777;
523 
524 	/*
525 	 * Hande the SMBFS_MF_xxx flags.
526 	 */
527 	if (flags & SMBFS_MF_NOAC)
528 		smi->smi_flags |= SMI_NOAC;
529 	if (flags & SMBFS_MF_ACREGMIN) {
530 		sec = STRUCT_FGET(args, acregmin);
531 		if (sec < 0 || sec > SMBFS_ACMINMAX)
532 			sec = SMBFS_ACMINMAX;
533 		smi->smi_acregmin = SEC2HR(sec);
534 	}
535 	if (flags & SMBFS_MF_ACREGMAX) {
536 		sec = STRUCT_FGET(args, acregmax);
537 		if (sec < 0 || sec > SMBFS_ACMAXMAX)
538 			sec = SMBFS_ACMAXMAX;
539 		smi->smi_acregmax = SEC2HR(sec);
540 	}
541 	if (flags & SMBFS_MF_ACDIRMIN) {
542 		sec = STRUCT_FGET(args, acdirmin);
543 		if (sec < 0 || sec > SMBFS_ACMINMAX)
544 			sec = SMBFS_ACMINMAX;
545 		smi->smi_acdirmin = SEC2HR(sec);
546 	}
547 	if (flags & SMBFS_MF_ACDIRMAX) {
548 		sec = STRUCT_FGET(args, acdirmax);
549 		if (sec < 0 || sec > SMBFS_ACMAXMAX)
550 			sec = SMBFS_ACMAXMAX;
551 		smi->smi_acdirmax = SEC2HR(sec);
552 	}
553 
554 	/*
555 	 * Get attributes of the remote file system,
556 	 * i.e. ACL support, named streams, etc.
557 	 */
558 	smb_credinit(&scred, cr);
559 	error = smbfs_smb_qfsattr(smi->smi_share, &smi->smi_fsa, &scred);
560 	smb_credrele(&scred);
561 	if (error) {
562 		SMBVDEBUG("smbfs_smb_qfsattr error %d\n", error);
563 	}
564 
565 	/*
566 	 * We enable XATTR by default (via smbfs_mntopts)
567 	 * but if the share does not support named streams,
568 	 * force the NOXATTR option (also clears XATTR).
569 	 * Caller will set or clear VFS_XATTR after this.
570 	 */
571 	if ((smi->smi_fsattr & FILE_NAMED_STREAMS) == 0)
572 		vfs_setmntopt(vfsp, MNTOPT_NOXATTR, NULL, 0);
573 
574 	/*
575 	 * Ditto ACLs (disable if not supported on this share)
576 	 */
577 	if ((smi->smi_fsattr & FILE_PERSISTENT_ACLS) == 0) {
578 		vfs_setmntopt(vfsp, MNTOPT_NOACL, NULL, 0);
579 		smi->smi_flags &= ~SMI_ACL;
580 	}
581 
582 	/*
583 	 * Assign a unique device id to the mount
584 	 */
585 	mutex_enter(&smbfs_minor_lock);
586 	do {
587 		smbfs_minor = (smbfs_minor + 1) & MAXMIN32;
588 		smbfs_dev = makedevice(smbfs_major, smbfs_minor);
589 	} while (vfs_devismounted(smbfs_dev));
590 	mutex_exit(&smbfs_minor_lock);
591 
592 	vfsp->vfs_dev	= smbfs_dev;
593 	vfs_make_fsid(&vfsp->vfs_fsid, smbfs_dev, smbfsfstyp);
594 	vfsp->vfs_data	= (caddr_t)smi;
595 	vfsp->vfs_fstype = smbfsfstyp;
596 	vfsp->vfs_bsize = MAXBSIZE;
597 	vfsp->vfs_bcount = 0;
598 
599 	smi->smi_vfsp	= vfsp;
600 	smbfs_zonelist_add(smi);	/* undo in smbfs_freevfs */
601 
602 	/*
603 	 * Create the root vnode, which we need in unmount
604 	 * for the call to smbfs_check_table(), etc.
605 	 * Release this hold in smbfs_unmount.
606 	 */
607 	rtnp = smbfs_node_findcreate(smi, "\\", 1, NULL, 0, 0,
608 	    &smbfs_fattr0);
609 	ASSERT(rtnp != NULL);
610 	rtnp->r_vnode->v_type = VDIR;
611 	rtnp->r_vnode->v_flag |= VROOT;
612 	smi->smi_root = rtnp;
613 
614 	/*
615 	 * NFS does other stuff here too:
616 	 *   async worker threads
617 	 *   init kstats
618 	 *
619 	 * End of code from NFS nfsrootvp()
620 	 */
621 	return (0);
622 
623 errout:
624 	vfsp->vfs_data = NULL;
625 	if (smi != NULL)
626 		smbfs_free_smi(smi);
627 
628 	if (mntzone != NULL)
629 		zone_rele(mntzone);
630 
631 	if (ssp != NULL)
632 		smb_share_rele(ssp);
633 
634 	return (error);
635 }
636 
637 /*
638  * vfs operations
639  */
640 static int
641 smbfs_unmount(vfs_t *vfsp, int flag, cred_t *cr)
642 {
643 	smbmntinfo_t	*smi;
644 	smbnode_t	*rtnp;
645 
646 	smi = VFTOSMI(vfsp);
647 
648 	if (secpolicy_fs_unmount(cr, vfsp) != 0)
649 		return (EPERM);
650 
651 	if ((flag & MS_FORCE) == 0) {
652 		smbfs_rflush(vfsp, cr);
653 
654 		/*
655 		 * If there are any active vnodes on this file system,
656 		 * (other than the root vnode) then the file system is
657 		 * busy and can't be umounted.
658 		 */
659 		if (smbfs_check_table(vfsp, smi->smi_root))
660 			return (EBUSY);
661 
662 		/*
663 		 * We normally hold a ref to the root vnode, so
664 		 * check for references beyond the one we expect:
665 		 *   smbmntinfo_t -> smi_root
666 		 * Note that NFS does not hold the root vnode.
667 		 */
668 		if (smi->smi_root &&
669 		    smi->smi_root->r_vnode->v_count > 1)
670 			return (EBUSY);
671 	}
672 
673 	/*
674 	 * common code for both forced and non-forced
675 	 *
676 	 * Setting VFS_UNMOUNTED prevents new operations.
677 	 * Operations already underway may continue,
678 	 * but not for long.
679 	 */
680 	vfsp->vfs_flag |= VFS_UNMOUNTED;
681 
682 	/*
683 	 * Shutdown any outstanding I/O requests on this share,
684 	 * and force a tree disconnect.  The share object will
685 	 * continue to hang around until smb_share_rele().
686 	 * This should also cause most active nodes to be
687 	 * released as their operations fail with EIO.
688 	 */
689 	smb_share_kill(smi->smi_share);
690 
691 	/*
692 	 * If we hold the root VP (and we normally do)
693 	 * then it's safe to release it now.
694 	 */
695 	if (smi->smi_root) {
696 		rtnp = smi->smi_root;
697 		smi->smi_root = NULL;
698 		VN_RELE(rtnp->r_vnode);	/* release root vnode */
699 	}
700 
701 	/*
702 	 * Remove all nodes from the node hash tables.
703 	 * This (indirectly) calls: smbfs_addfree, smbinactive,
704 	 * which will try to flush dirty pages, etc. so
705 	 * don't destroy the underlying share just yet.
706 	 *
707 	 * Also, with a forced unmount, some nodes may
708 	 * remain active, and those will get cleaned up
709 	 * after their last vn_rele.
710 	 */
711 	smbfs_destroy_table(vfsp);
712 
713 	/*
714 	 * Delete our kstats...
715 	 *
716 	 * Doing it here, rather than waiting until
717 	 * smbfs_freevfs so these are not visible
718 	 * after the unmount.
719 	 */
720 	if (smi->smi_io_kstats) {
721 		kstat_delete(smi->smi_io_kstats);
722 		smi->smi_io_kstats = NULL;
723 	}
724 	if (smi->smi_ro_kstats) {
725 		kstat_delete(smi->smi_ro_kstats);
726 		smi->smi_ro_kstats = NULL;
727 	}
728 
729 	/*
730 	 * The rest happens in smbfs_freevfs()
731 	 */
732 	return (0);
733 }
734 
735 
736 /*
737  * find root of smbfs
738  */
739 static int
740 smbfs_root(vfs_t *vfsp, vnode_t **vpp)
741 {
742 	smbmntinfo_t	*smi;
743 	vnode_t		*vp;
744 
745 	smi = VFTOSMI(vfsp);
746 
747 	if (curproc->p_zone != smi->smi_zone_ref.zref_zone)
748 		return (EPERM);
749 
750 	if (smi->smi_flags & SMI_DEAD || vfsp->vfs_flag & VFS_UNMOUNTED)
751 		return (EIO);
752 
753 	/*
754 	 * The root vp is created in mount and held
755 	 * until unmount, so this is paranoia.
756 	 */
757 	if (smi->smi_root == NULL)
758 		return (EIO);
759 
760 	/* Just take a reference and return it. */
761 	vp = SMBTOV(smi->smi_root);
762 	VN_HOLD(vp);
763 	*vpp = vp;
764 
765 	return (0);
766 }
767 
768 /*
769  * Get file system statistics.
770  */
771 static int
772 smbfs_statvfs(vfs_t *vfsp, statvfs64_t *sbp)
773 {
774 	int		error;
775 	smbmntinfo_t	*smi = VFTOSMI(vfsp);
776 	smb_share_t	*ssp = smi->smi_share;
777 	statvfs64_t	stvfs;
778 	hrtime_t now;
779 	smb_cred_t	scred;
780 
781 	if (curproc->p_zone != smi->smi_zone_ref.zref_zone)
782 		return (EPERM);
783 
784 	if (smi->smi_flags & SMI_DEAD || vfsp->vfs_flag & VFS_UNMOUNTED)
785 		return (EIO);
786 
787 	mutex_enter(&smi->smi_lock);
788 
789 	/*
790 	 * Use cached result if still valid.
791 	 */
792 recheck:
793 	now = gethrtime();
794 	if (now < smi->smi_statfstime) {
795 		error = 0;
796 		goto cache_hit;
797 	}
798 
799 	/*
800 	 * FS attributes are stale, so someone
801 	 * needs to do an OTW call to get them.
802 	 * Serialize here so only one thread
803 	 * does the OTW call.
804 	 */
805 	if (smi->smi_status & SM_STATUS_STATFS_BUSY) {
806 		smi->smi_status |= SM_STATUS_STATFS_WANT;
807 		if (!cv_wait_sig(&smi->smi_statvfs_cv, &smi->smi_lock)) {
808 			mutex_exit(&smi->smi_lock);
809 			return (EINTR);
810 		}
811 		/* Hope status is valid now. */
812 		goto recheck;
813 	}
814 	smi->smi_status |= SM_STATUS_STATFS_BUSY;
815 	mutex_exit(&smi->smi_lock);
816 
817 	/*
818 	 * Do the OTW call.  Note: lock NOT held.
819 	 */
820 	smb_credinit(&scred, NULL);
821 	bzero(&stvfs, sizeof (stvfs));
822 	error = smbfs_smb_statfs(ssp, &stvfs, &scred);
823 	smb_credrele(&scred);
824 	if (error) {
825 		SMBVDEBUG("statfs error=%d\n", error);
826 	} else {
827 
828 		/*
829 		 * Set a few things the OTW call didn't get.
830 		 */
831 		stvfs.f_frsize = stvfs.f_bsize;
832 		stvfs.f_favail = stvfs.f_ffree;
833 		stvfs.f_fsid = (unsigned long)vfsp->vfs_fsid.val[0];
834 		bcopy(fs_type_name, stvfs.f_basetype, FSTYPSZ);
835 		stvfs.f_flag	= vf_to_stf(vfsp->vfs_flag);
836 		stvfs.f_namemax	= smi->smi_fsa.fsa_maxname;
837 
838 		/*
839 		 * Save the result, update lifetime
840 		 */
841 		now = gethrtime();
842 		smi->smi_statfstime = now +
843 		    (SM_MAX_STATFSTIME * (hrtime_t)NANOSEC);
844 		smi->smi_statvfsbuf = stvfs; /* struct assign! */
845 	}
846 
847 	mutex_enter(&smi->smi_lock);
848 	if (smi->smi_status & SM_STATUS_STATFS_WANT)
849 		cv_broadcast(&smi->smi_statvfs_cv);
850 	smi->smi_status &= ~(SM_STATUS_STATFS_BUSY | SM_STATUS_STATFS_WANT);
851 
852 	/*
853 	 * Copy the statvfs data to caller's buf.
854 	 * Note: struct assignment
855 	 */
856 cache_hit:
857 	if (error == 0)
858 		*sbp = smi->smi_statvfsbuf;
859 	mutex_exit(&smi->smi_lock);
860 	return (error);
861 }
862 
863 static kmutex_t smbfs_syncbusy;
864 
865 /*
866  * Flush dirty smbfs files for file system vfsp.
867  * If vfsp == NULL, all smbfs files are flushed.
868  */
869 /*ARGSUSED*/
870 static int
871 smbfs_sync(vfs_t *vfsp, short flag, cred_t *cr)
872 {
873 	/*
874 	 * Cross-zone calls are OK here, since this translates to a
875 	 * VOP_PUTPAGE(B_ASYNC), which gets picked up by the right zone.
876 	 */
877 	if (!(flag & SYNC_ATTR) && mutex_tryenter(&smbfs_syncbusy) != 0) {
878 		smbfs_rflush(vfsp, cr);
879 		mutex_exit(&smbfs_syncbusy);
880 	}
881 
882 	return (0);
883 }
884 
885 /*
886  * Initialization routine for VFS routines.  Should only be called once
887  */
888 int
889 smbfs_vfsinit(void)
890 {
891 	mutex_init(&smbfs_syncbusy, NULL, MUTEX_DEFAULT, NULL);
892 	return (0);
893 }
894 
895 /*
896  * Shutdown routine for VFS routines.  Should only be called once
897  */
898 void
899 smbfs_vfsfini(void)
900 {
901 	mutex_destroy(&smbfs_syncbusy);
902 }
903 
904 void
905 smbfs_freevfs(vfs_t *vfsp)
906 {
907 	smbmntinfo_t    *smi;
908 
909 	/* free up the resources */
910 	smi = VFTOSMI(vfsp);
911 
912 	/*
913 	 * By this time we should have already deleted the
914 	 * smi kstats in the unmount code.  If they are still around
915 	 * something is wrong
916 	 */
917 	ASSERT(smi->smi_io_kstats == NULL);
918 
919 	smbfs_zonelist_remove(smi);
920 
921 	smbfs_free_smi(smi);
922 
923 	/*
924 	 * Allow _fini() to succeed now, if so desired.
925 	 */
926 	atomic_dec_32(&smbfs_mountcount);
927 }
928 
929 /*
930  * smbfs_mount_label_policy:
931  *	Determine whether the mount is allowed according to MAC check,
932  *	by comparing (where appropriate) label of the remote server
933  *	against the label of the zone being mounted into.
934  *
935  *	Returns:
936  *		 0 :	access allowed
937  *		-1 :	read-only access allowed (i.e., read-down)
938  *		>0 :	error code, such as EACCES
939  *
940  * NB:
941  * NFS supports Cipso labels by parsing the vfs_resource
942  * to see what the Solaris server global zone has shared.
943  * We can't support that for CIFS since resource names
944  * contain share names, not paths.
945  */
946 static int
947 smbfs_mount_label_policy(vfs_t *vfsp, void *ipaddr, int addr_type, cred_t *cr)
948 {
949 	bslabel_t	*server_sl, *mntlabel;
950 	zone_t		*mntzone = NULL;
951 	ts_label_t	*zlabel;
952 	tsol_tpc_t	*tp;
953 	ts_label_t	*tsl = NULL;
954 	int		retv;
955 
956 	/*
957 	 * Get the zone's label.  Each zone on a labeled system has a label.
958 	 */
959 	mntzone = zone_find_by_any_path(refstr_value(vfsp->vfs_mntpt), B_FALSE);
960 	zlabel = mntzone->zone_slabel;
961 	ASSERT(zlabel != NULL);
962 	label_hold(zlabel);
963 
964 	retv = EACCES;				/* assume the worst */
965 
966 	/*
967 	 * Next, get the assigned label of the remote server.
968 	 */
969 	tp = find_tpc(ipaddr, addr_type, B_FALSE);
970 	if (tp == NULL)
971 		goto out;			/* error getting host entry */
972 
973 	if (tp->tpc_tp.tp_doi != zlabel->tsl_doi)
974 		goto rel_tpc;			/* invalid domain */
975 	if ((tp->tpc_tp.host_type != UNLABELED))
976 		goto rel_tpc;			/* invalid hosttype */
977 
978 	server_sl = &tp->tpc_tp.tp_def_label;
979 	mntlabel = label2bslabel(zlabel);
980 
981 	/*
982 	 * Now compare labels to complete the MAC check.  If the labels
983 	 * are equal or if the requestor is in the global zone and has
984 	 * NET_MAC_AWARE, then allow read-write access.   (Except for
985 	 * mounts into the global zone itself; restrict these to
986 	 * read-only.)
987 	 *
988 	 * If the requestor is in some other zone, but his label
989 	 * dominates the server, then allow read-down.
990 	 *
991 	 * Otherwise, access is denied.
992 	 */
993 	if (blequal(mntlabel, server_sl) ||
994 	    (crgetzoneid(cr) == GLOBAL_ZONEID &&
995 	    getpflags(NET_MAC_AWARE, cr) != 0)) {
996 		if ((mntzone == global_zone) ||
997 		    !blequal(mntlabel, server_sl))
998 			retv = -1;		/* read-only */
999 		else
1000 			retv = 0;		/* access OK */
1001 	} else if (bldominates(mntlabel, server_sl)) {
1002 		retv = -1;			/* read-only */
1003 	} else {
1004 		retv = EACCES;
1005 	}
1006 
1007 	if (tsl != NULL)
1008 		label_rele(tsl);
1009 
1010 rel_tpc:
1011 	/*LINTED*/
1012 	TPC_RELE(tp);
1013 out:
1014 	if (mntzone)
1015 		zone_rele(mntzone);
1016 	label_rele(zlabel);
1017 	return (retv);
1018 }
1019