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  * Copyright (c) 2016 by Delphix. All rights reserved.
39  */
40 
41 #include <sys/systm.h>
42 #include <sys/cred.h>
43 #include <sys/time.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 static char *acl_cancel[] = { MNTOPT_NOACL, NULL };
92 static char *noacl_cancel[] = { MNTOPT_ACL, NULL };
93 static char *xattr_cancel[] = { MNTOPT_NOXATTR, NULL };
94 static char *noxattr_cancel[] = { MNTOPT_XATTR, NULL };
95 
96 static mntopt_t mntopts[] = {
97 /*
98  *	option name		cancel option	default arg	flags
99  *		ufs arg flag
100  */
101 	{ MNTOPT_INTR,		intr_cancel,	NULL,	MO_DEFAULT, 0 },
102 	{ MNTOPT_NOINTR,	nointr_cancel,	NULL,	0,	0 },
103 	{ MNTOPT_ACL,		acl_cancel,	NULL,	MO_DEFAULT, 0 },
104 	{ MNTOPT_NOACL,		noacl_cancel,	NULL,	0,	0 },
105 	{ MNTOPT_XATTR,		xattr_cancel,	NULL,	MO_DEFAULT, 0 },
106 	{ MNTOPT_NOXATTR,	noxattr_cancel, NULL,	0,	0 }
107 };
108 
109 static mntopts_t smbfs_mntopts = {
110 	sizeof (mntopts) / sizeof (mntopt_t),
111 	mntopts
112 };
113 
114 static const char fs_type_name[FSTYPSZ] = "smbfs";
115 
116 static vfsdef_t vfw = {
117 	VFSDEF_VERSION,
118 	(char *)fs_type_name,
119 	smbfsinit,		/* init routine */
120 	VSW_HASPROTO|VSW_NOTZONESAFE,	/* flags */
121 	&smbfs_mntopts			/* mount options table prototype */
122 };
123 
124 static struct modlfs modlfs = {
125 	&mod_fsops,
126 	"SMBFS filesystem",
127 	&vfw
128 };
129 
130 static struct modlinkage modlinkage = {
131 	MODREV_1, (void *)&modlfs, NULL
132 };
133 
134 /*
135  * Mutex to protect the following variables:
136  *	  smbfs_major
137  *	  smbfs_minor
138  */
139 extern	kmutex_t	smbfs_minor_lock;
140 extern	int		smbfs_major;
141 extern	int		smbfs_minor;
142 
143 /*
144  * Prevent unloads while we have mounts
145  */
146 uint32_t	smbfs_mountcount;
147 
148 /*
149  * smbfs vfs operations.
150  */
151 static int	smbfs_mount(vfs_t *, vnode_t *, struct mounta *, cred_t *);
152 static int	smbfs_unmount(vfs_t *, int, cred_t *);
153 static int	smbfs_root(vfs_t *, vnode_t **);
154 static int	smbfs_statvfs(vfs_t *, statvfs64_t *);
155 static int	smbfs_sync(vfs_t *, short, cred_t *);
156 static void	smbfs_freevfs(vfs_t *);
157 
158 /*
159  * Module loading
160  */
161 
162 /*
163  * This routine is invoked automatically when the kernel module
164  * containing this routine is loaded.  This allows module specific
165  * initialization to be done when the module is loaded.
166  */
167 int
168 _init(void)
169 {
170 	int		error;
171 
172 	/*
173 	 * Check compiled-in version of "nsmb"
174 	 * that we're linked with.  (paranoid)
175 	 */
176 	if (nsmb_version != NSMB_VERSION) {
177 		cmn_err(CE_WARN, "_init: nsmb version mismatch");
178 		return (ENOTTY);
179 	}
180 
181 	smbfs_mountcount = 0;
182 
183 	/*
184 	 * NFS calls these two in _clntinit
185 	 * Easier to follow this way.
186 	 */
187 	if ((error = smbfs_subrinit()) != 0) {
188 		cmn_err(CE_WARN, "_init: smbfs_subrinit failed");
189 		return (error);
190 	}
191 
192 	if ((error = smbfs_vfsinit()) != 0) {
193 		cmn_err(CE_WARN, "_init: smbfs_vfsinit failed");
194 		smbfs_subrfini();
195 		return (error);
196 	}
197 
198 	if ((error = smbfs_clntinit()) != 0) {
199 		cmn_err(CE_WARN, "_init: smbfs_clntinit failed");
200 		smbfs_vfsfini();
201 		smbfs_subrfini();
202 		return (error);
203 	}
204 
205 	error = mod_install((struct modlinkage *)&modlinkage);
206 	return (error);
207 }
208 
209 /*
210  * Free kernel module resources that were allocated in _init
211  * and remove the linkage information into the kernel
212  */
213 int
214 _fini(void)
215 {
216 	int	error;
217 
218 	/*
219 	 * If a forcedly unmounted instance is still hanging around,
220 	 * we cannot allow the module to be unloaded because that would
221 	 * cause panics once the VFS framework decides it's time to call
222 	 * into VFS_FREEVFS().
223 	 */
224 	if (smbfs_mountcount)
225 		return (EBUSY);
226 
227 	error = mod_remove(&modlinkage);
228 	if (error)
229 		return (error);
230 
231 	/*
232 	 * Free the allocated smbnodes, etc.
233 	 */
234 	smbfs_clntfini();
235 
236 	/* NFS calls these two in _clntfini */
237 	smbfs_vfsfini();
238 	smbfs_subrfini();
239 
240 	/*
241 	 * Free the ops vectors
242 	 */
243 	smbfsfini();
244 	return (0);
245 }
246 
247 /*
248  * Return information about the module
249  */
250 int
251 _info(struct modinfo *modinfop)
252 {
253 	return (mod_info((struct modlinkage *)&modlinkage, modinfop));
254 }
255 
256 /*
257  * Initialize the vfs structure
258  */
259 
260 int smbfsfstyp;
261 vfsops_t *smbfs_vfsops = NULL;
262 
263 static const fs_operation_def_t smbfs_vfsops_template[] = {
264 	{ VFSNAME_MOUNT, { .vfs_mount = smbfs_mount } },
265 	{ VFSNAME_UNMOUNT, { .vfs_unmount = smbfs_unmount } },
266 	{ VFSNAME_ROOT,	{ .vfs_root = smbfs_root } },
267 	{ VFSNAME_STATVFS, { .vfs_statvfs = smbfs_statvfs } },
268 	{ VFSNAME_SYNC,	{ .vfs_sync = smbfs_sync } },
269 	{ VFSNAME_VGET,	{ .error = fs_nosys } },
270 	{ VFSNAME_MOUNTROOT, { .error = fs_nosys } },
271 	{ VFSNAME_FREEVFS, { .vfs_freevfs = smbfs_freevfs } },
272 	{ NULL, NULL }
273 };
274 
275 int
276 smbfsinit(int fstyp, char *name)
277 {
278 	int		error;
279 
280 	error = vfs_setfsops(fstyp, smbfs_vfsops_template, &smbfs_vfsops);
281 	if (error != 0) {
282 		zcmn_err(GLOBAL_ZONEID, CE_WARN,
283 		    "smbfsinit: bad vfs ops template");
284 		return (error);
285 	}
286 
287 	error = vn_make_ops(name, smbfs_vnodeops_template, &smbfs_vnodeops);
288 	if (error != 0) {
289 		(void) vfs_freevfsops_by_type(fstyp);
290 		zcmn_err(GLOBAL_ZONEID, CE_WARN,
291 		    "smbfsinit: bad vnode ops template");
292 		return (error);
293 	}
294 
295 	smbfsfstyp = fstyp;
296 
297 	return (0);
298 }
299 
300 void
301 smbfsfini()
302 {
303 	if (smbfs_vfsops) {
304 		(void) vfs_freevfsops_by_type(smbfsfstyp);
305 		smbfs_vfsops = NULL;
306 	}
307 	if (smbfs_vnodeops) {
308 		vn_freevnodeops(smbfs_vnodeops);
309 		smbfs_vnodeops = NULL;
310 	}
311 }
312 
313 void
314 smbfs_free_smi(smbmntinfo_t *smi)
315 {
316 	if (smi == NULL)
317 		return;
318 
319 	if (smi->smi_zone_ref.zref_zone != NULL)
320 		zone_rele_ref(&smi->smi_zone_ref, ZONE_REF_SMBFS);
321 
322 	if (smi->smi_share != NULL)
323 		smb_share_rele(smi->smi_share);
324 
325 	avl_destroy(&smi->smi_hash_avl);
326 	rw_destroy(&smi->smi_hash_lk);
327 	cv_destroy(&smi->smi_statvfs_cv);
328 	mutex_destroy(&smi->smi_lock);
329 
330 	kmem_free(smi, sizeof (smbmntinfo_t));
331 }
332 
333 /*
334  * smbfs mount vfsop
335  * Set up mount info record and attach it to vfs struct.
336  */
337 static int
338 smbfs_mount(vfs_t *vfsp, vnode_t *mvp, struct mounta *uap, cred_t *cr)
339 {
340 	char		*data = uap->dataptr;
341 	int		error;
342 	smbnode_t 	*rtnp = NULL;	/* root of this fs */
343 	smbmntinfo_t 	*smi = NULL;
344 	dev_t 		smbfs_dev;
345 	int 		version;
346 	int 		devfd;
347 	zone_t		*zone = curproc->p_zone;
348 	zone_t		*mntzone = NULL;
349 	smb_share_t 	*ssp = NULL;
350 	smb_cred_t 	scred;
351 	int		flags, sec;
352 
353 	STRUCT_DECL(smbfs_args, args);		/* smbfs mount arguments */
354 
355 	if ((error = secpolicy_fs_mount(cr, mvp, vfsp)) != 0)
356 		return (error);
357 
358 	if (mvp->v_type != VDIR)
359 		return (ENOTDIR);
360 
361 	/*
362 	 * get arguments
363 	 *
364 	 * uap->datalen might be different from sizeof (args)
365 	 * in a compatible situation.
366 	 */
367 	STRUCT_INIT(args, get_udatamodel());
368 	bzero(STRUCT_BUF(args), SIZEOF_STRUCT(smbfs_args, DATAMODEL_NATIVE));
369 	if (copyin(data, STRUCT_BUF(args), MIN(uap->datalen,
370 	    SIZEOF_STRUCT(smbfs_args, DATAMODEL_NATIVE))))
371 		return (EFAULT);
372 
373 	/*
374 	 * Check mount program version
375 	 */
376 	version = STRUCT_FGET(args, version);
377 	if (version != SMBFS_VERSION) {
378 		cmn_err(CE_WARN, "mount version mismatch:"
379 		    " kernel=%d, mount=%d\n",
380 		    SMBFS_VERSION, version);
381 		return (EINVAL);
382 	}
383 
384 	/*
385 	 * Deal with re-mount requests.
386 	 */
387 	if (uap->flags & MS_REMOUNT) {
388 		cmn_err(CE_WARN, "MS_REMOUNT not implemented");
389 		return (ENOTSUP);
390 	}
391 
392 	/*
393 	 * Check for busy
394 	 */
395 	mutex_enter(&mvp->v_lock);
396 	if (!(uap->flags & MS_OVERLAY) &&
397 	    (mvp->v_count != 1 || (mvp->v_flag & VROOT))) {
398 		mutex_exit(&mvp->v_lock);
399 		return (EBUSY);
400 	}
401 	mutex_exit(&mvp->v_lock);
402 
403 	/*
404 	 * Get the "share" from the netsmb driver (ssp).
405 	 * It is returned with a "ref" (hold) for us.
406 	 * Release this hold: at errout below, or in
407 	 * smbfs_freevfs().
408 	 */
409 	devfd = STRUCT_FGET(args, devfd);
410 	error = smb_dev2share(devfd, &ssp);
411 	if (error) {
412 		cmn_err(CE_WARN, "invalid device handle %d (%d)\n",
413 		    devfd, error);
414 		return (error);
415 	}
416 
417 	/*
418 	 * Use "goto errout" from here on.
419 	 * See: ssp, smi, rtnp, mntzone
420 	 */
421 
422 	/*
423 	 * Determine the zone we're being mounted into.
424 	 */
425 	zone_hold(mntzone = zone);		/* start with this assumption */
426 	if (getzoneid() == GLOBAL_ZONEID) {
427 		zone_rele(mntzone);
428 		mntzone = zone_find_by_path(refstr_value(vfsp->vfs_mntpt));
429 		ASSERT(mntzone != NULL);
430 		if (mntzone != zone) {
431 			error = EBUSY;
432 			goto errout;
433 		}
434 	}
435 
436 	/*
437 	 * Stop the mount from going any further if the zone is going away.
438 	 */
439 	if (zone_status_get(mntzone) >= ZONE_IS_SHUTTING_DOWN) {
440 		error = EBUSY;
441 		goto errout;
442 	}
443 
444 	/*
445 	 * On a Trusted Extensions client, we may have to force read-only
446 	 * for read-down mounts.
447 	 */
448 	if (is_system_labeled()) {
449 		void *addr;
450 		int ipvers = 0;
451 		struct smb_vc *vcp;
452 
453 		vcp = SSTOVC(ssp);
454 		addr = smb_vc_getipaddr(vcp, &ipvers);
455 		error = smbfs_mount_label_policy(vfsp, addr, ipvers, cr);
456 
457 		if (error > 0)
458 			goto errout;
459 
460 		if (error == -1) {
461 			/* change mount to read-only to prevent write-down */
462 			vfs_setmntopt(vfsp, MNTOPT_RO, NULL, 0);
463 		}
464 	}
465 
466 	/* Prevent unload. */
467 	atomic_inc_32(&smbfs_mountcount);
468 
469 	/*
470 	 * Create a mount record and link it to the vfs struct.
471 	 * No more possiblities for errors from here on.
472 	 * Tear-down of this stuff is in smbfs_free_smi()
473 	 *
474 	 * Compare with NFS: nfsrootvp()
475 	 */
476 	smi = kmem_zalloc(sizeof (*smi), KM_SLEEP);
477 
478 	mutex_init(&smi->smi_lock, NULL, MUTEX_DEFAULT, NULL);
479 	cv_init(&smi->smi_statvfs_cv, NULL, CV_DEFAULT, NULL);
480 
481 	rw_init(&smi->smi_hash_lk, NULL, RW_DEFAULT, NULL);
482 	smbfs_init_hash_avl(&smi->smi_hash_avl);
483 
484 	smi->smi_share = ssp;
485 	ssp = NULL;
486 
487 	/*
488 	 * Convert the anonymous zone hold acquired via zone_hold() above
489 	 * into a zone reference.
490 	 */
491 	zone_init_ref(&smi->smi_zone_ref);
492 	zone_hold_ref(mntzone, &smi->smi_zone_ref, ZONE_REF_SMBFS);
493 	zone_rele(mntzone);
494 	mntzone = NULL;
495 
496 	/*
497 	 * Initialize option defaults
498 	 */
499 	smi->smi_flags	= SMI_LLOCK;
500 	smi->smi_acregmin = SEC2HR(SMBFS_ACREGMIN);
501 	smi->smi_acregmax = SEC2HR(SMBFS_ACREGMAX);
502 	smi->smi_acdirmin = SEC2HR(SMBFS_ACDIRMIN);
503 	smi->smi_acdirmax = SEC2HR(SMBFS_ACDIRMAX);
504 
505 	/*
506 	 * All "generic" mount options have already been
507 	 * handled in vfs.c:domount() - see mntopts stuff.
508 	 * Query generic options using vfs_optionisset().
509 	 */
510 	if (vfs_optionisset(vfsp, MNTOPT_INTR, NULL))
511 		smi->smi_flags |= SMI_INT;
512 	if (vfs_optionisset(vfsp, MNTOPT_ACL, NULL))
513 		smi->smi_flags |= SMI_ACL;
514 
515 	/*
516 	 * Get the mount options that come in as smbfs_args,
517 	 * starting with args.flags (SMBFS_MF_xxx)
518 	 */
519 	flags = STRUCT_FGET(args, flags);
520 	smi->smi_uid 	= STRUCT_FGET(args, uid);
521 	smi->smi_gid 	= STRUCT_FGET(args, gid);
522 	smi->smi_fmode	= STRUCT_FGET(args, file_mode) & 0777;
523 	smi->smi_dmode	= STRUCT_FGET(args, dir_mode) & 0777;
524 
525 	/*
526 	 * Hande the SMBFS_MF_xxx flags.
527 	 */
528 	if (flags & SMBFS_MF_NOAC)
529 		smi->smi_flags |= SMI_NOAC;
530 	if (flags & SMBFS_MF_ACREGMIN) {
531 		sec = STRUCT_FGET(args, acregmin);
532 		if (sec < 0 || sec > SMBFS_ACMINMAX)
533 			sec = SMBFS_ACMINMAX;
534 		smi->smi_acregmin = SEC2HR(sec);
535 	}
536 	if (flags & SMBFS_MF_ACREGMAX) {
537 		sec = STRUCT_FGET(args, acregmax);
538 		if (sec < 0 || sec > SMBFS_ACMAXMAX)
539 			sec = SMBFS_ACMAXMAX;
540 		smi->smi_acregmax = SEC2HR(sec);
541 	}
542 	if (flags & SMBFS_MF_ACDIRMIN) {
543 		sec = STRUCT_FGET(args, acdirmin);
544 		if (sec < 0 || sec > SMBFS_ACMINMAX)
545 			sec = SMBFS_ACMINMAX;
546 		smi->smi_acdirmin = SEC2HR(sec);
547 	}
548 	if (flags & SMBFS_MF_ACDIRMAX) {
549 		sec = STRUCT_FGET(args, acdirmax);
550 		if (sec < 0 || sec > SMBFS_ACMAXMAX)
551 			sec = SMBFS_ACMAXMAX;
552 		smi->smi_acdirmax = SEC2HR(sec);
553 	}
554 
555 	/*
556 	 * Get attributes of the remote file system,
557 	 * i.e. ACL support, named streams, etc.
558 	 */
559 	smb_credinit(&scred, cr);
560 	error = smbfs_smb_qfsattr(smi->smi_share, &smi->smi_fsa, &scred);
561 	smb_credrele(&scred);
562 	if (error) {
563 		SMBVDEBUG("smbfs_smb_qfsattr error %d\n", error);
564 	}
565 
566 	/*
567 	 * We enable XATTR by default (via smbfs_mntopts)
568 	 * but if the share does not support named streams,
569 	 * force the NOXATTR option (also clears XATTR).
570 	 * Caller will set or clear VFS_XATTR after this.
571 	 */
572 	if ((smi->smi_fsattr & FILE_NAMED_STREAMS) == 0)
573 		vfs_setmntopt(vfsp, MNTOPT_NOXATTR, NULL, 0);
574 
575 	/*
576 	 * Ditto ACLs (disable if not supported on this share)
577 	 */
578 	if ((smi->smi_fsattr & FILE_PERSISTENT_ACLS) == 0) {
579 		vfs_setmntopt(vfsp, MNTOPT_NOACL, NULL, 0);
580 		smi->smi_flags &= ~SMI_ACL;
581 	}
582 
583 	/*
584 	 * Assign a unique device id to the mount
585 	 */
586 	mutex_enter(&smbfs_minor_lock);
587 	do {
588 		smbfs_minor = (smbfs_minor + 1) & MAXMIN32;
589 		smbfs_dev = makedevice(smbfs_major, smbfs_minor);
590 	} while (vfs_devismounted(smbfs_dev));
591 	mutex_exit(&smbfs_minor_lock);
592 
593 	vfsp->vfs_dev	= smbfs_dev;
594 	vfs_make_fsid(&vfsp->vfs_fsid, smbfs_dev, smbfsfstyp);
595 	vfsp->vfs_data	= (caddr_t)smi;
596 	vfsp->vfs_fstype = smbfsfstyp;
597 	vfsp->vfs_bsize = MAXBSIZE;
598 	vfsp->vfs_bcount = 0;
599 
600 	smi->smi_vfsp	= vfsp;
601 	smbfs_zonelist_add(smi);	/* undo in smbfs_freevfs */
602 
603 	/* PSARC 2007/227 VFS Feature Registration */
604 	vfs_set_feature(vfsp, VFSFT_XVATTR);
605 	vfs_set_feature(vfsp, VFSFT_SYSATTR_VIEWS);
606 
607 	/*
608 	 * Create the root vnode, which we need in unmount
609 	 * for the call to smbfs_check_table(), etc.
610 	 * Release this hold in smbfs_unmount.
611 	 */
612 	rtnp = smbfs_node_findcreate(smi, "\\", 1, NULL, 0, 0,
613 	    &smbfs_fattr0);
614 	ASSERT(rtnp != NULL);
615 	rtnp->r_vnode->v_type = VDIR;
616 	rtnp->r_vnode->v_flag |= VROOT;
617 	smi->smi_root = rtnp;
618 
619 	/*
620 	 * NFS does other stuff here too:
621 	 *   async worker threads
622 	 *   init kstats
623 	 *
624 	 * End of code from NFS nfsrootvp()
625 	 */
626 	return (0);
627 
628 errout:
629 	vfsp->vfs_data = NULL;
630 	if (smi != NULL)
631 		smbfs_free_smi(smi);
632 
633 	if (mntzone != NULL)
634 		zone_rele(mntzone);
635 
636 	if (ssp != NULL)
637 		smb_share_rele(ssp);
638 
639 	return (error);
640 }
641 
642 /*
643  * vfs operations
644  */
645 static int
646 smbfs_unmount(vfs_t *vfsp, int flag, cred_t *cr)
647 {
648 	smbmntinfo_t	*smi;
649 	smbnode_t	*rtnp;
650 
651 	smi = VFTOSMI(vfsp);
652 
653 	if (secpolicy_fs_unmount(cr, vfsp) != 0)
654 		return (EPERM);
655 
656 	if ((flag & MS_FORCE) == 0) {
657 		smbfs_rflush(vfsp, cr);
658 
659 		/*
660 		 * If there are any active vnodes on this file system,
661 		 * (other than the root vnode) then the file system is
662 		 * busy and can't be umounted.
663 		 */
664 		if (smbfs_check_table(vfsp, smi->smi_root))
665 			return (EBUSY);
666 
667 		/*
668 		 * We normally hold a ref to the root vnode, so
669 		 * check for references beyond the one we expect:
670 		 *   smbmntinfo_t -> smi_root
671 		 * Note that NFS does not hold the root vnode.
672 		 */
673 		if (smi->smi_root &&
674 		    smi->smi_root->r_vnode->v_count > 1)
675 			return (EBUSY);
676 	}
677 
678 	/*
679 	 * common code for both forced and non-forced
680 	 *
681 	 * Setting VFS_UNMOUNTED prevents new operations.
682 	 * Operations already underway may continue,
683 	 * but not for long.
684 	 */
685 	vfsp->vfs_flag |= VFS_UNMOUNTED;
686 
687 	/*
688 	 * Shutdown any outstanding I/O requests on this share,
689 	 * and force a tree disconnect.  The share object will
690 	 * continue to hang around until smb_share_rele().
691 	 * This should also cause most active nodes to be
692 	 * released as their operations fail with EIO.
693 	 */
694 	smb_share_kill(smi->smi_share);
695 
696 	/*
697 	 * If we hold the root VP (and we normally do)
698 	 * then it's safe to release it now.
699 	 */
700 	if (smi->smi_root) {
701 		rtnp = smi->smi_root;
702 		smi->smi_root = NULL;
703 		VN_RELE(rtnp->r_vnode);	/* release root vnode */
704 	}
705 
706 	/*
707 	 * Remove all nodes from the node hash tables.
708 	 * This (indirectly) calls: smbfs_addfree, smbinactive,
709 	 * which will try to flush dirty pages, etc. so
710 	 * don't destroy the underlying share just yet.
711 	 *
712 	 * Also, with a forced unmount, some nodes may
713 	 * remain active, and those will get cleaned up
714 	 * after their last vn_rele.
715 	 */
716 	smbfs_destroy_table(vfsp);
717 
718 	/*
719 	 * Delete our kstats...
720 	 *
721 	 * Doing it here, rather than waiting until
722 	 * smbfs_freevfs so these are not visible
723 	 * after the unmount.
724 	 */
725 	if (smi->smi_io_kstats) {
726 		kstat_delete(smi->smi_io_kstats);
727 		smi->smi_io_kstats = NULL;
728 	}
729 	if (smi->smi_ro_kstats) {
730 		kstat_delete(smi->smi_ro_kstats);
731 		smi->smi_ro_kstats = NULL;
732 	}
733 
734 	/*
735 	 * The rest happens in smbfs_freevfs()
736 	 */
737 	return (0);
738 }
739 
740 
741 /*
742  * find root of smbfs
743  */
744 static int
745 smbfs_root(vfs_t *vfsp, vnode_t **vpp)
746 {
747 	smbmntinfo_t	*smi;
748 	vnode_t		*vp;
749 
750 	smi = VFTOSMI(vfsp);
751 
752 	if (curproc->p_zone != smi->smi_zone_ref.zref_zone)
753 		return (EPERM);
754 
755 	if (smi->smi_flags & SMI_DEAD || vfsp->vfs_flag & VFS_UNMOUNTED)
756 		return (EIO);
757 
758 	/*
759 	 * The root vp is created in mount and held
760 	 * until unmount, so this is paranoia.
761 	 */
762 	if (smi->smi_root == NULL)
763 		return (EIO);
764 
765 	/* Just take a reference and return it. */
766 	vp = SMBTOV(smi->smi_root);
767 	VN_HOLD(vp);
768 	*vpp = vp;
769 
770 	return (0);
771 }
772 
773 /*
774  * Get file system statistics.
775  */
776 static int
777 smbfs_statvfs(vfs_t *vfsp, statvfs64_t *sbp)
778 {
779 	int		error;
780 	smbmntinfo_t	*smi = VFTOSMI(vfsp);
781 	smb_share_t	*ssp = smi->smi_share;
782 	statvfs64_t	stvfs;
783 	hrtime_t now;
784 	smb_cred_t	scred;
785 
786 	if (curproc->p_zone != smi->smi_zone_ref.zref_zone)
787 		return (EPERM);
788 
789 	if (smi->smi_flags & SMI_DEAD || vfsp->vfs_flag & VFS_UNMOUNTED)
790 		return (EIO);
791 
792 	mutex_enter(&smi->smi_lock);
793 
794 	/*
795 	 * Use cached result if still valid.
796 	 */
797 recheck:
798 	now = gethrtime();
799 	if (now < smi->smi_statfstime) {
800 		error = 0;
801 		goto cache_hit;
802 	}
803 
804 	/*
805 	 * FS attributes are stale, so someone
806 	 * needs to do an OTW call to get them.
807 	 * Serialize here so only one thread
808 	 * does the OTW call.
809 	 */
810 	if (smi->smi_status & SM_STATUS_STATFS_BUSY) {
811 		smi->smi_status |= SM_STATUS_STATFS_WANT;
812 		if (!cv_wait_sig(&smi->smi_statvfs_cv, &smi->smi_lock)) {
813 			mutex_exit(&smi->smi_lock);
814 			return (EINTR);
815 		}
816 		/* Hope status is valid now. */
817 		goto recheck;
818 	}
819 	smi->smi_status |= SM_STATUS_STATFS_BUSY;
820 	mutex_exit(&smi->smi_lock);
821 
822 	/*
823 	 * Do the OTW call.  Note: lock NOT held.
824 	 */
825 	smb_credinit(&scred, NULL);
826 	bzero(&stvfs, sizeof (stvfs));
827 	error = smbfs_smb_statfs(ssp, &stvfs, &scred);
828 	smb_credrele(&scred);
829 	if (error) {
830 		SMBVDEBUG("statfs error=%d\n", error);
831 	} else {
832 
833 		/*
834 		 * Set a few things the OTW call didn't get.
835 		 */
836 		stvfs.f_frsize = stvfs.f_bsize;
837 		stvfs.f_favail = stvfs.f_ffree;
838 		stvfs.f_fsid = (unsigned long)vfsp->vfs_fsid.val[0];
839 		bcopy(fs_type_name, stvfs.f_basetype, FSTYPSZ);
840 		stvfs.f_flag	= vf_to_stf(vfsp->vfs_flag);
841 		stvfs.f_namemax	= smi->smi_fsa.fsa_maxname;
842 
843 		/*
844 		 * Save the result, update lifetime
845 		 */
846 		now = gethrtime();
847 		smi->smi_statfstime = now +
848 		    (SM_MAX_STATFSTIME * (hrtime_t)NANOSEC);
849 		smi->smi_statvfsbuf = stvfs; /* struct assign! */
850 	}
851 
852 	mutex_enter(&smi->smi_lock);
853 	if (smi->smi_status & SM_STATUS_STATFS_WANT)
854 		cv_broadcast(&smi->smi_statvfs_cv);
855 	smi->smi_status &= ~(SM_STATUS_STATFS_BUSY | SM_STATUS_STATFS_WANT);
856 
857 	/*
858 	 * Copy the statvfs data to caller's buf.
859 	 * Note: struct assignment
860 	 */
861 cache_hit:
862 	if (error == 0)
863 		*sbp = smi->smi_statvfsbuf;
864 	mutex_exit(&smi->smi_lock);
865 	return (error);
866 }
867 
868 static kmutex_t smbfs_syncbusy;
869 
870 /*
871  * Flush dirty smbfs files for file system vfsp.
872  * If vfsp == NULL, all smbfs files are flushed.
873  */
874 /*ARGSUSED*/
875 static int
876 smbfs_sync(vfs_t *vfsp, short flag, cred_t *cr)
877 {
878 	/*
879 	 * Cross-zone calls are OK here, since this translates to a
880 	 * VOP_PUTPAGE(B_ASYNC), which gets picked up by the right zone.
881 	 */
882 	if (!(flag & SYNC_ATTR) && mutex_tryenter(&smbfs_syncbusy) != 0) {
883 		smbfs_rflush(vfsp, cr);
884 		mutex_exit(&smbfs_syncbusy);
885 	}
886 
887 	return (0);
888 }
889 
890 /*
891  * Initialization routine for VFS routines.  Should only be called once
892  */
893 int
894 smbfs_vfsinit(void)
895 {
896 	mutex_init(&smbfs_syncbusy, NULL, MUTEX_DEFAULT, NULL);
897 	return (0);
898 }
899 
900 /*
901  * Shutdown routine for VFS routines.  Should only be called once
902  */
903 void
904 smbfs_vfsfini(void)
905 {
906 	mutex_destroy(&smbfs_syncbusy);
907 }
908 
909 void
910 smbfs_freevfs(vfs_t *vfsp)
911 {
912 	smbmntinfo_t    *smi;
913 
914 	/* free up the resources */
915 	smi = VFTOSMI(vfsp);
916 
917 	/*
918 	 * By this time we should have already deleted the
919 	 * smi kstats in the unmount code.  If they are still around
920 	 * something is wrong
921 	 */
922 	ASSERT(smi->smi_io_kstats == NULL);
923 
924 	smbfs_zonelist_remove(smi);
925 
926 	smbfs_free_smi(smi);
927 
928 	/*
929 	 * Allow _fini() to succeed now, if so desired.
930 	 */
931 	atomic_dec_32(&smbfs_mountcount);
932 }
933 
934 /*
935  * smbfs_mount_label_policy:
936  *	Determine whether the mount is allowed according to MAC check,
937  *	by comparing (where appropriate) label of the remote server
938  *	against the label of the zone being mounted into.
939  *
940  *	Returns:
941  *		 0 :	access allowed
942  *		-1 :	read-only access allowed (i.e., read-down)
943  *		>0 :	error code, such as EACCES
944  *
945  * NB:
946  * NFS supports Cipso labels by parsing the vfs_resource
947  * to see what the Solaris server global zone has shared.
948  * We can't support that for CIFS since resource names
949  * contain share names, not paths.
950  */
951 static int
952 smbfs_mount_label_policy(vfs_t *vfsp, void *ipaddr, int addr_type, cred_t *cr)
953 {
954 	bslabel_t	*server_sl, *mntlabel;
955 	zone_t		*mntzone = NULL;
956 	ts_label_t	*zlabel;
957 	tsol_tpc_t	*tp;
958 	ts_label_t	*tsl = NULL;
959 	int		retv;
960 
961 	/*
962 	 * Get the zone's label.  Each zone on a labeled system has a label.
963 	 */
964 	mntzone = zone_find_by_any_path(refstr_value(vfsp->vfs_mntpt), B_FALSE);
965 	zlabel = mntzone->zone_slabel;
966 	ASSERT(zlabel != NULL);
967 	label_hold(zlabel);
968 
969 	retv = EACCES;				/* assume the worst */
970 
971 	/*
972 	 * Next, get the assigned label of the remote server.
973 	 */
974 	tp = find_tpc(ipaddr, addr_type, B_FALSE);
975 	if (tp == NULL)
976 		goto out;			/* error getting host entry */
977 
978 	if (tp->tpc_tp.tp_doi != zlabel->tsl_doi)
979 		goto rel_tpc;			/* invalid domain */
980 	if ((tp->tpc_tp.host_type != UNLABELED))
981 		goto rel_tpc;			/* invalid hosttype */
982 
983 	server_sl = &tp->tpc_tp.tp_def_label;
984 	mntlabel = label2bslabel(zlabel);
985 
986 	/*
987 	 * Now compare labels to complete the MAC check.  If the labels
988 	 * are equal or if the requestor is in the global zone and has
989 	 * NET_MAC_AWARE, then allow read-write access.   (Except for
990 	 * mounts into the global zone itself; restrict these to
991 	 * read-only.)
992 	 *
993 	 * If the requestor is in some other zone, but their label
994 	 * dominates the server, then allow read-down.
995 	 *
996 	 * Otherwise, access is denied.
997 	 */
998 	if (blequal(mntlabel, server_sl) ||
999 	    (crgetzoneid(cr) == GLOBAL_ZONEID &&
1000 	    getpflags(NET_MAC_AWARE, cr) != 0)) {
1001 		if ((mntzone == global_zone) ||
1002 		    !blequal(mntlabel, server_sl))
1003 			retv = -1;		/* read-only */
1004 		else
1005 			retv = 0;		/* access OK */
1006 	} else if (bldominates(mntlabel, server_sl)) {
1007 		retv = -1;			/* read-only */
1008 	} else {
1009 		retv = EACCES;
1010 	}
1011 
1012 	if (tsl != NULL)
1013 		label_rele(tsl);
1014 
1015 rel_tpc:
1016 	/*LINTED*/
1017 	TPC_RELE(tp);
1018 out:
1019 	if (mntzone)
1020 		zone_rele(mntzone);
1021 	label_rele(zlabel);
1022 	return (retv);
1023 }
1024