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