xref: /illumos-gate/usr/src/uts/common/fs/smbclnt/smbfs/smbfs_vfsops.c (revision d8a7fe16f62711cdc5c4267da8b34ff24a6b668c)
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 2009 Sun Microsystems, Inc.  All rights reserved.
37  * Use is subject to license terms.
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 #ifdef NOT_YET
91 static char *force_dio_cancel[] = { MNTOPT_NOFORCEDIRECTIO, NULL };
92 static char *noforce_dio_cancel[] = { MNTOPT_FORCEDIRECTIO, NULL };
93 static char *largefiles_cancel[] = { MNTOPT_NOLARGEFILES, NULL };
94 static char *nolargefiles_cancel[] = { MNTOPT_LARGEFILES, NULL };
95 #endif
96 static char *xattr_cancel[] = { MNTOPT_NOXATTR, NULL };
97 static char *noxattr_cancel[] = { MNTOPT_XATTR, NULL };
98 
99 static mntopt_t mntopts[] = {
100 /*
101  *	option name		cancel option	default arg	flags
102  *		ufs arg flag
103  */
104 	{ MNTOPT_INTR,		intr_cancel,	NULL,	MO_DEFAULT, 0 },
105 	{ MNTOPT_NOINTR,	nointr_cancel,	NULL,	0,	0 },
106 #ifdef NOT_YET
107 	{ MNTOPT_FORCEDIRECTIO,	force_dio_cancel, NULL, 0,	0 },
108 	{ MNTOPT_NOFORCEDIRECTIO, noforce_dio_cancel, NULL, 0, 0 },
109 	{ MNTOPT_LARGEFILES,	largefiles_cancel, NULL, MO_DEFAULT, 0 },
110 	{ MNTOPT_NOLARGEFILES,	nolargefiles_cancel, NULL, 0,	0 },
111 #endif
112 	{ MNTOPT_XATTR,		xattr_cancel,	NULL,	MO_DEFAULT, 0 },
113 	{ MNTOPT_NOXATTR,	noxattr_cancel, NULL,	0,	0 }
114 };
115 
116 static mntopts_t smbfs_mntopts = {
117 	sizeof (mntopts) / sizeof (mntopt_t),
118 	mntopts
119 };
120 
121 static const char fs_type_name[FSTYPSZ] = "smbfs";
122 
123 static vfsdef_t vfw = {
124 	VFSDEF_VERSION,
125 	(char *)fs_type_name,
126 	smbfsinit,		/* init routine */
127 	VSW_HASPROTO|VSW_NOTZONESAFE,	/* flags */
128 	&smbfs_mntopts			/* mount options table prototype */
129 };
130 
131 static struct modlfs modlfs = {
132 	&mod_fsops,
133 	"SMBFS filesystem",
134 	&vfw
135 };
136 
137 static struct modlinkage modlinkage = {
138 	MODREV_1, (void *)&modlfs, NULL
139 };
140 
141 /*
142  * Mutex to protect the following variables:
143  *	  smbfs_major
144  *	  smbfs_minor
145  */
146 extern	kmutex_t	smbfs_minor_lock;
147 extern	int		smbfs_major;
148 extern	int		smbfs_minor;
149 
150 /*
151  * Prevent unloads while we have mounts
152  */
153 uint32_t	smbfs_mountcount;
154 
155 /*
156  * smbfs vfs operations.
157  */
158 static int	smbfs_mount(vfs_t *, vnode_t *, struct mounta *, cred_t *);
159 static int	smbfs_unmount(vfs_t *, int, cred_t *);
160 static int	smbfs_root(vfs_t *, vnode_t **);
161 static int	smbfs_statvfs(vfs_t *, statvfs64_t *);
162 static int	smbfs_sync(vfs_t *, short, cred_t *);
163 static void	smbfs_freevfs(vfs_t *);
164 
165 /*
166  * Module loading
167  */
168 
169 /*
170  * This routine is invoked automatically when the kernel module
171  * containing this routine is loaded.  This allows module specific
172  * initialization to be done when the module is loaded.
173  */
174 int
175 _init(void)
176 {
177 	int		error;
178 
179 	/*
180 	 * Check compiled-in version of "nsmb"
181 	 * that we're linked with.  (paranoid)
182 	 */
183 	if (nsmb_version != NSMB_VERSION) {
184 		cmn_err(CE_WARN, "_init: nsmb version mismatch");
185 		return (ENOTTY);
186 	}
187 
188 	smbfs_mountcount = 0;
189 
190 	/*
191 	 * NFS calls these two in _clntinit
192 	 * Easier to follow this way.
193 	 */
194 	if ((error = smbfs_subrinit()) != 0) {
195 		cmn_err(CE_WARN, "_init: smbfs_subrinit failed");
196 		return (error);
197 	}
198 
199 	if ((error = smbfs_vfsinit()) != 0) {
200 		cmn_err(CE_WARN, "_init: smbfs_vfsinit failed");
201 		smbfs_subrfini();
202 		return (error);
203 	}
204 
205 	if ((error = smbfs_clntinit()) != 0) {
206 		cmn_err(CE_WARN, "_init: smbfs_clntinit failed");
207 		smbfs_vfsfini();
208 		smbfs_subrfini();
209 		return (error);
210 	}
211 
212 	error = mod_install((struct modlinkage *)&modlinkage);
213 	return (error);
214 }
215 
216 /*
217  * Free kernel module resources that were allocated in _init
218  * and remove the linkage information into the kernel
219  */
220 int
221 _fini(void)
222 {
223 	int	error;
224 
225 	/*
226 	 * If a forcedly unmounted instance is still hanging around,
227 	 * we cannot allow the module to be unloaded because that would
228 	 * cause panics once the VFS framework decides it's time to call
229 	 * into VFS_FREEVFS().
230 	 */
231 	if (smbfs_mountcount)
232 		return (EBUSY);
233 
234 	error = mod_remove(&modlinkage);
235 	if (error)
236 		return (error);
237 
238 	/*
239 	 * Free the allocated smbnodes, etc.
240 	 */
241 	smbfs_clntfini();
242 
243 	/* NFS calls these two in _clntfini */
244 	smbfs_vfsfini();
245 	smbfs_subrfini();
246 
247 	/*
248 	 * Free the ops vectors
249 	 */
250 	smbfsfini();
251 	return (0);
252 }
253 
254 /*
255  * Return information about the module
256  */
257 int
258 _info(struct modinfo *modinfop)
259 {
260 	return (mod_info((struct modlinkage *)&modlinkage, modinfop));
261 }
262 
263 /*
264  * Initialize the vfs structure
265  */
266 
267 int smbfsfstyp;
268 vfsops_t *smbfs_vfsops = NULL;
269 
270 static const fs_operation_def_t smbfs_vfsops_template[] = {
271 	{ VFSNAME_MOUNT, { .vfs_mount = smbfs_mount } },
272 	{ VFSNAME_UNMOUNT, { .vfs_unmount = smbfs_unmount } },
273 	{ VFSNAME_ROOT,	{ .vfs_root = smbfs_root } },
274 	{ VFSNAME_STATVFS, { .vfs_statvfs = smbfs_statvfs } },
275 	{ VFSNAME_SYNC,	{ .vfs_sync = smbfs_sync } },
276 	{ VFSNAME_VGET,	{ .error = fs_nosys } },
277 	{ VFSNAME_MOUNTROOT, { .error = fs_nosys } },
278 	{ VFSNAME_FREEVFS, { .vfs_freevfs = smbfs_freevfs } },
279 	{ NULL, NULL }
280 };
281 
282 int
283 smbfsinit(int fstyp, char *name)
284 {
285 	int		error;
286 
287 	error = vfs_setfsops(fstyp, smbfs_vfsops_template, &smbfs_vfsops);
288 	if (error != 0) {
289 		zcmn_err(GLOBAL_ZONEID, CE_WARN,
290 		    "smbfsinit: bad vfs ops template");
291 		return (error);
292 	}
293 
294 	error = vn_make_ops(name, smbfs_vnodeops_template, &smbfs_vnodeops);
295 	if (error != 0) {
296 		(void) vfs_freevfsops_by_type(fstyp);
297 		zcmn_err(GLOBAL_ZONEID, CE_WARN,
298 		    "smbfsinit: bad vnode ops template");
299 		return (error);
300 	}
301 
302 	smbfsfstyp = fstyp;
303 
304 	return (0);
305 }
306 
307 void
308 smbfsfini()
309 {
310 	if (smbfs_vfsops) {
311 		(void) vfs_freevfsops_by_type(smbfsfstyp);
312 		smbfs_vfsops = NULL;
313 	}
314 	if (smbfs_vnodeops) {
315 		vn_freevnodeops(smbfs_vnodeops);
316 		smbfs_vnodeops = NULL;
317 	}
318 }
319 
320 void
321 smbfs_free_smi(smbmntinfo_t *smi)
322 {
323 	if (smi == NULL)
324 		return;
325 
326 	if (smi->smi_zone != NULL)
327 		zone_rele(smi->smi_zone);
328 
329 	if (smi->smi_share != NULL)
330 		smb_share_rele(smi->smi_share);
331 
332 	avl_destroy(&smi->smi_hash_avl);
333 	rw_destroy(&smi->smi_hash_lk);
334 	cv_destroy(&smi->smi_statvfs_cv);
335 	mutex_destroy(&smi->smi_lock);
336 
337 	kmem_free(smi, sizeof (smbmntinfo_t));
338 }
339 
340 /*
341  * smbfs mount vfsop
342  * Set up mount info record and attach it to vfs struct.
343  */
344 static int
345 smbfs_mount(vfs_t *vfsp, vnode_t *mvp, struct mounta *uap, cred_t *cr)
346 {
347 	char		*data = uap->dataptr;
348 	int		error;
349 	smbnode_t 	*rtnp = NULL;	/* root of this fs */
350 	smbmntinfo_t 	*smi = NULL;
351 	dev_t 		smbfs_dev;
352 	int 		version;
353 	int 		devfd;
354 	zone_t		*zone = curproc->p_zone;
355 	zone_t		*mntzone = NULL;
356 	smb_share_t 	*ssp = NULL;
357 	smb_cred_t 	scred;
358 	int		flags, sec;
359 
360 	STRUCT_DECL(smbfs_args, args);		/* smbfs mount arguments */
361 
362 	if ((error = secpolicy_fs_mount(cr, mvp, vfsp)) != 0)
363 		return (error);
364 
365 	if (mvp->v_type != VDIR)
366 		return (ENOTDIR);
367 
368 	/*
369 	 * get arguments
370 	 *
371 	 * uap->datalen might be different from sizeof (args)
372 	 * in a compatible situation.
373 	 */
374 	STRUCT_INIT(args, get_udatamodel());
375 	bzero(STRUCT_BUF(args), SIZEOF_STRUCT(smbfs_args, DATAMODEL_NATIVE));
376 	if (copyin(data, STRUCT_BUF(args), MIN(uap->datalen,
377 	    SIZEOF_STRUCT(smbfs_args, DATAMODEL_NATIVE))))
378 		return (EFAULT);
379 
380 	/*
381 	 * Check mount program version
382 	 */
383 	version = STRUCT_FGET(args, version);
384 	if (version != SMBFS_VERSION) {
385 		cmn_err(CE_WARN, "mount version mismatch:"
386 		    " kernel=%d, mount=%d\n",
387 		    SMBFS_VERSION, version);
388 		return (EINVAL);
389 	}
390 
391 	/*
392 	 * Deal with re-mount requests.
393 	 */
394 	if (uap->flags & MS_REMOUNT) {
395 		cmn_err(CE_WARN, "MS_REMOUNT not implemented");
396 		return (ENOTSUP);
397 	}
398 
399 	/*
400 	 * Check for busy
401 	 */
402 	mutex_enter(&mvp->v_lock);
403 	if (!(uap->flags & MS_OVERLAY) &&
404 	    (mvp->v_count != 1 || (mvp->v_flag & VROOT))) {
405 		mutex_exit(&mvp->v_lock);
406 		return (EBUSY);
407 	}
408 	mutex_exit(&mvp->v_lock);
409 
410 	/*
411 	 * Get the "share" from the netsmb driver (ssp).
412 	 * It is returned with a "ref" (hold) for us.
413 	 * Release this hold: at errout below, or in
414 	 * smbfs_freevfs().
415 	 */
416 	devfd = STRUCT_FGET(args, devfd);
417 	error = smb_dev2share(devfd, &ssp);
418 	if (error) {
419 		cmn_err(CE_WARN, "invalid device handle %d (%d)\n",
420 		    devfd, error);
421 		return (error);
422 	}
423 
424 	/*
425 	 * Use "goto errout" from here on.
426 	 * See: ssp, smi, rtnp, mntzone
427 	 */
428 
429 	/*
430 	 * Determine the zone we're being mounted into.
431 	 */
432 	zone_hold(mntzone = zone);		/* start with this assumption */
433 	if (getzoneid() == GLOBAL_ZONEID) {
434 		zone_rele(mntzone);
435 		mntzone = zone_find_by_path(refstr_value(vfsp->vfs_mntpt));
436 		ASSERT(mntzone != NULL);
437 		if (mntzone != zone) {
438 			error = EBUSY;
439 			goto errout;
440 		}
441 	}
442 
443 	/*
444 	 * Stop the mount from going any further if the zone is going away.
445 	 */
446 	if (zone_status_get(mntzone) >= ZONE_IS_SHUTTING_DOWN) {
447 		error = EBUSY;
448 		goto errout;
449 	}
450 
451 	/*
452 	 * On a Trusted Extensions client, we may have to force read-only
453 	 * for read-down mounts.
454 	 */
455 	if (is_system_labeled()) {
456 		void *addr;
457 		int ipvers = 0;
458 		struct smb_vc *vcp;
459 
460 		vcp = SSTOVC(ssp);
461 		addr = smb_vc_getipaddr(vcp, &ipvers);
462 		error = smbfs_mount_label_policy(vfsp, addr, ipvers, cr);
463 
464 		if (error > 0)
465 			goto errout;
466 
467 		if (error == -1) {
468 			/* change mount to read-only to prevent write-down */
469 			vfs_setmntopt(vfsp, MNTOPT_RO, NULL, 0);
470 		}
471 	}
472 
473 	/* Prevent unload. */
474 	atomic_inc_32(&smbfs_mountcount);
475 
476 	/*
477 	 * Create a mount record and link it to the vfs struct.
478 	 * No more possiblities for errors from here on.
479 	 * Tear-down of this stuff is in smbfs_free_smi()
480 	 *
481 	 * Compare with NFS: nfsrootvp()
482 	 */
483 	smi = kmem_zalloc(sizeof (*smi), KM_SLEEP);
484 
485 	mutex_init(&smi->smi_lock, NULL, MUTEX_DEFAULT, NULL);
486 	cv_init(&smi->smi_statvfs_cv, NULL, CV_DEFAULT, NULL);
487 
488 	rw_init(&smi->smi_hash_lk, NULL, RW_DEFAULT, NULL);
489 	smbfs_init_hash_avl(&smi->smi_hash_avl);
490 
491 	smi->smi_share = ssp;
492 	ssp = NULL;
493 	smi->smi_zone = 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 
513 	/*
514 	 * Get the mount options that come in as smbfs_args,
515 	 * starting with args.flags (SMBFS_MF_xxx)
516 	 */
517 	flags = STRUCT_FGET(args, flags);
518 	smi->smi_uid 	= STRUCT_FGET(args, uid);
519 	smi->smi_gid 	= STRUCT_FGET(args, gid);
520 	smi->smi_fmode	= STRUCT_FGET(args, file_mode) & 0777;
521 	smi->smi_dmode	= STRUCT_FGET(args, dir_mode) & 0777;
522 
523 	/*
524 	 * Hande the SMBFS_MF_xxx flags.
525 	 */
526 	if (flags & SMBFS_MF_NOAC)
527 		smi->smi_flags |= SMI_NOAC;
528 	if (flags & SMBFS_MF_ACREGMIN) {
529 		sec = STRUCT_FGET(args, acregmin);
530 		if (sec < 0 || sec > SMBFS_ACMINMAX)
531 			sec = SMBFS_ACMINMAX;
532 		smi->smi_acregmin = SEC2HR(sec);
533 	}
534 	if (flags & SMBFS_MF_ACREGMAX) {
535 		sec = STRUCT_FGET(args, acregmax);
536 		if (sec < 0 || sec > SMBFS_ACMAXMAX)
537 			sec = SMBFS_ACMAXMAX;
538 		smi->smi_acregmax = SEC2HR(sec);
539 	}
540 	if (flags & SMBFS_MF_ACDIRMIN) {
541 		sec = STRUCT_FGET(args, acdirmin);
542 		if (sec < 0 || sec > SMBFS_ACMINMAX)
543 			sec = SMBFS_ACMINMAX;
544 		smi->smi_acdirmin = SEC2HR(sec);
545 	}
546 	if (flags & SMBFS_MF_ACDIRMAX) {
547 		sec = STRUCT_FGET(args, acdirmax);
548 		if (sec < 0 || sec > SMBFS_ACMAXMAX)
549 			sec = SMBFS_ACMAXMAX;
550 		smi->smi_acdirmax = SEC2HR(sec);
551 	}
552 
553 	/*
554 	 * Get attributes of the remote file system,
555 	 * i.e. ACL support, named streams, etc.
556 	 */
557 	smb_credinit(&scred, cr);
558 	error = smbfs_smb_qfsattr(smi->smi_share, &smi->smi_fsa, &scred);
559 	smb_credrele(&scred);
560 	if (error) {
561 		SMBVDEBUG("smbfs_smb_qfsattr error %d\n", error);
562 	}
563 
564 	/*
565 	 * We enable XATTR by default (via smbfs_mntopts)
566 	 * but if the share does not support named streams,
567 	 * force the NOXATTR option (also clears XATTR).
568 	 * Caller will set or clear VFS_XATTR after this.
569 	 */
570 	if ((smi->smi_fsattr & FILE_NAMED_STREAMS) == 0)
571 		vfs_setmntopt(vfsp, MNTOPT_NOXATTR, NULL, 0);
572 
573 	/*
574 	 * Assign a unique device id to the mount
575 	 */
576 	mutex_enter(&smbfs_minor_lock);
577 	do {
578 		smbfs_minor = (smbfs_minor + 1) & MAXMIN32;
579 		smbfs_dev = makedevice(smbfs_major, smbfs_minor);
580 	} while (vfs_devismounted(smbfs_dev));
581 	mutex_exit(&smbfs_minor_lock);
582 
583 	vfsp->vfs_dev	= smbfs_dev;
584 	vfs_make_fsid(&vfsp->vfs_fsid, smbfs_dev, smbfsfstyp);
585 	vfsp->vfs_data	= (caddr_t)smi;
586 	vfsp->vfs_fstype = smbfsfstyp;
587 	vfsp->vfs_bsize = MAXBSIZE;
588 	vfsp->vfs_bcount = 0;
589 
590 	smi->smi_vfsp	= vfsp;
591 	smbfs_zonelist_add(smi);	/* undo in smbfs_freevfs */
592 
593 	/*
594 	 * Create the root vnode, which we need in unmount
595 	 * for the call to smbfs_check_table(), etc.
596 	 * Release this hold in smbfs_unmount.
597 	 */
598 	rtnp = smbfs_node_findcreate(smi, "\\", 1, NULL, 0, 0,
599 	    &smbfs_fattr0);
600 	ASSERT(rtnp != NULL);
601 	rtnp->r_vnode->v_type = VDIR;
602 	rtnp->r_vnode->v_flag |= VROOT;
603 	smi->smi_root = rtnp;
604 
605 	/*
606 	 * NFS does other stuff here too:
607 	 *   async worker threads
608 	 *   init kstats
609 	 *
610 	 * End of code from NFS nfsrootvp()
611 	 */
612 	return (0);
613 
614 errout:
615 	vfsp->vfs_data = NULL;
616 	if (smi != NULL)
617 		smbfs_free_smi(smi);
618 
619 	if (mntzone != NULL)
620 		zone_rele(mntzone);
621 
622 	if (ssp != NULL)
623 		smb_share_rele(ssp);
624 
625 	return (error);
626 }
627 
628 /*
629  * vfs operations
630  */
631 static int
632 smbfs_unmount(vfs_t *vfsp, int flag, cred_t *cr)
633 {
634 	smbmntinfo_t	*smi;
635 	smbnode_t	*rtnp;
636 
637 	smi = VFTOSMI(vfsp);
638 
639 	if (secpolicy_fs_unmount(cr, vfsp) != 0)
640 		return (EPERM);
641 
642 	if ((flag & MS_FORCE) == 0) {
643 		smbfs_rflush(vfsp, cr);
644 
645 		/*
646 		 * If there are any active vnodes on this file system,
647 		 * (other than the root vnode) then the file system is
648 		 * busy and can't be umounted.
649 		 */
650 		if (smbfs_check_table(vfsp, smi->smi_root))
651 			return (EBUSY);
652 
653 		/*
654 		 * We normally hold a ref to the root vnode, so
655 		 * check for references beyond the one we expect:
656 		 *   smbmntinfo_t -> smi_root
657 		 * Note that NFS does not hold the root vnode.
658 		 */
659 		if (smi->smi_root &&
660 		    smi->smi_root->r_vnode->v_count > 1)
661 			return (EBUSY);
662 	}
663 
664 	/*
665 	 * common code for both forced and non-forced
666 	 *
667 	 * Setting VFS_UNMOUNTED prevents new operations.
668 	 * Operations already underway may continue,
669 	 * but not for long.
670 	 */
671 	vfsp->vfs_flag |= VFS_UNMOUNTED;
672 
673 	/*
674 	 * Shutdown any outstanding I/O requests on this share,
675 	 * and force a tree disconnect.  The share object will
676 	 * continue to hang around until smb_share_rele().
677 	 * This should also cause most active nodes to be
678 	 * released as their operations fail with EIO.
679 	 */
680 	smb_share_kill(smi->smi_share);
681 
682 	/*
683 	 * If we hold the root VP (and we normally do)
684 	 * then it's safe to release it now.
685 	 */
686 	if (smi->smi_root) {
687 		rtnp = smi->smi_root;
688 		smi->smi_root = NULL;
689 		VN_RELE(rtnp->r_vnode);	/* release root vnode */
690 	}
691 
692 	/*
693 	 * Remove all nodes from the node hash tables.
694 	 * This (indirectly) calls: smbfs_addfree, smbinactive,
695 	 * which will try to flush dirty pages, etc. so
696 	 * don't destroy the underlying share just yet.
697 	 *
698 	 * Also, with a forced unmount, some nodes may
699 	 * remain active, and those will get cleaned up
700 	 * after their last vn_rele.
701 	 */
702 	smbfs_destroy_table(vfsp);
703 
704 	/*
705 	 * Delete our kstats...
706 	 *
707 	 * Doing it here, rather than waiting until
708 	 * smbfs_freevfs so these are not visible
709 	 * after the unmount.
710 	 */
711 	if (smi->smi_io_kstats) {
712 		kstat_delete(smi->smi_io_kstats);
713 		smi->smi_io_kstats = NULL;
714 	}
715 	if (smi->smi_ro_kstats) {
716 		kstat_delete(smi->smi_ro_kstats);
717 		smi->smi_ro_kstats = NULL;
718 	}
719 
720 	/*
721 	 * The rest happens in smbfs_freevfs()
722 	 */
723 	return (0);
724 }
725 
726 
727 /*
728  * find root of smbfs
729  */
730 static int
731 smbfs_root(vfs_t *vfsp, vnode_t **vpp)
732 {
733 	smbmntinfo_t	*smi;
734 	vnode_t		*vp;
735 
736 	smi = VFTOSMI(vfsp);
737 
738 	if (curproc->p_zone != smi->smi_zone)
739 		return (EPERM);
740 
741 	if (smi->smi_flags & SMI_DEAD || vfsp->vfs_flag & VFS_UNMOUNTED)
742 		return (EIO);
743 
744 	/*
745 	 * The root vp is created in mount and held
746 	 * until unmount, so this is paranoia.
747 	 */
748 	if (smi->smi_root == NULL)
749 		return (EIO);
750 
751 	/* Just take a reference and return it. */
752 	vp = SMBTOV(smi->smi_root);
753 	VN_HOLD(vp);
754 	*vpp = vp;
755 
756 	return (0);
757 }
758 
759 /*
760  * Get file system statistics.
761  */
762 static int
763 smbfs_statvfs(vfs_t *vfsp, statvfs64_t *sbp)
764 {
765 	int		error;
766 	smbmntinfo_t	*smi = VFTOSMI(vfsp);
767 	smb_share_t	*ssp = smi->smi_share;
768 	statvfs64_t	stvfs;
769 	hrtime_t now;
770 	smb_cred_t	scred;
771 
772 	if (curproc->p_zone != smi->smi_zone)
773 		return (EPERM);
774 
775 	if (smi->smi_flags & SMI_DEAD || vfsp->vfs_flag & VFS_UNMOUNTED)
776 		return (EIO);
777 
778 	mutex_enter(&smi->smi_lock);
779 
780 	/*
781 	 * Use cached result if still valid.
782 	 */
783 recheck:
784 	now = gethrtime();
785 	if (now < smi->smi_statfstime) {
786 		error = 0;
787 		goto cache_hit;
788 	}
789 
790 	/*
791 	 * FS attributes are stale, so someone
792 	 * needs to do an OTW call to get them.
793 	 * Serialize here so only one thread
794 	 * does the OTW call.
795 	 */
796 	if (smi->smi_status & SM_STATUS_STATFS_BUSY) {
797 		smi->smi_status |= SM_STATUS_STATFS_WANT;
798 		if (!cv_wait_sig(&smi->smi_statvfs_cv, &smi->smi_lock)) {
799 			mutex_exit(&smi->smi_lock);
800 			return (EINTR);
801 		}
802 		/* Hope status is valid now. */
803 		goto recheck;
804 	}
805 	smi->smi_status |= SM_STATUS_STATFS_BUSY;
806 	mutex_exit(&smi->smi_lock);
807 
808 	/*
809 	 * Do the OTW call.  Note: lock NOT held.
810 	 */
811 	smb_credinit(&scred, NULL);
812 	bzero(&stvfs, sizeof (stvfs));
813 	error = smbfs_smb_statfs(ssp, &stvfs, &scred);
814 	smb_credrele(&scred);
815 	if (error) {
816 		SMBVDEBUG("statfs error=%d\n", error);
817 	} else {
818 
819 		/*
820 		 * Set a few things the OTW call didn't get.
821 		 */
822 		stvfs.f_frsize = stvfs.f_bsize;
823 		stvfs.f_favail = stvfs.f_ffree;
824 		stvfs.f_fsid = (unsigned long)vfsp->vfs_fsid.val[0];
825 		bcopy(fs_type_name, stvfs.f_basetype, FSTYPSZ);
826 		stvfs.f_flag	= vf_to_stf(vfsp->vfs_flag);
827 		stvfs.f_namemax	= smi->smi_fsa.fsa_maxname;
828 
829 		/*
830 		 * Save the result, update lifetime
831 		 */
832 		now = gethrtime();
833 		smi->smi_statfstime = now +
834 		    (SM_MAX_STATFSTIME * (hrtime_t)NANOSEC);
835 		smi->smi_statvfsbuf = stvfs; /* struct assign! */
836 	}
837 
838 	mutex_enter(&smi->smi_lock);
839 	if (smi->smi_status & SM_STATUS_STATFS_WANT)
840 		cv_broadcast(&smi->smi_statvfs_cv);
841 	smi->smi_status &= ~(SM_STATUS_STATFS_BUSY | SM_STATUS_STATFS_WANT);
842 
843 	/*
844 	 * Copy the statvfs data to caller's buf.
845 	 * Note: struct assignment
846 	 */
847 cache_hit:
848 	if (error == 0)
849 		*sbp = smi->smi_statvfsbuf;
850 	mutex_exit(&smi->smi_lock);
851 	return (error);
852 }
853 
854 static kmutex_t smbfs_syncbusy;
855 
856 /*
857  * Flush dirty smbfs files for file system vfsp.
858  * If vfsp == NULL, all smbfs files are flushed.
859  */
860 /*ARGSUSED*/
861 static int
862 smbfs_sync(vfs_t *vfsp, short flag, cred_t *cr)
863 {
864 	/*
865 	 * Cross-zone calls are OK here, since this translates to a
866 	 * VOP_PUTPAGE(B_ASYNC), which gets picked up by the right zone.
867 	 */
868 	if (!(flag & SYNC_ATTR) && mutex_tryenter(&smbfs_syncbusy) != 0) {
869 		smbfs_rflush(vfsp, cr);
870 		mutex_exit(&smbfs_syncbusy);
871 	}
872 
873 	return (0);
874 }
875 
876 /*
877  * Initialization routine for VFS routines.  Should only be called once
878  */
879 int
880 smbfs_vfsinit(void)
881 {
882 	mutex_init(&smbfs_syncbusy, NULL, MUTEX_DEFAULT, NULL);
883 	return (0);
884 }
885 
886 /*
887  * Shutdown routine for VFS routines.  Should only be called once
888  */
889 void
890 smbfs_vfsfini(void)
891 {
892 	mutex_destroy(&smbfs_syncbusy);
893 }
894 
895 void
896 smbfs_freevfs(vfs_t *vfsp)
897 {
898 	smbmntinfo_t    *smi;
899 
900 	/* free up the resources */
901 	smi = VFTOSMI(vfsp);
902 
903 	/*
904 	 * By this time we should have already deleted the
905 	 * smi kstats in the unmount code.  If they are still around
906 	 * something is wrong
907 	 */
908 	ASSERT(smi->smi_io_kstats == NULL);
909 
910 	smbfs_zonelist_remove(smi);
911 
912 	smbfs_free_smi(smi);
913 
914 	/*
915 	 * Allow _fini() to succeed now, if so desired.
916 	 */
917 	atomic_dec_32(&smbfs_mountcount);
918 }
919 
920 /*
921  * smbfs_mount_label_policy:
922  *	Determine whether the mount is allowed according to MAC check,
923  *	by comparing (where appropriate) label of the remote server
924  *	against the label of the zone being mounted into.
925  *
926  *	Returns:
927  *		 0 :	access allowed
928  *		-1 :	read-only access allowed (i.e., read-down)
929  *		>0 :	error code, such as EACCES
930  *
931  * NB:
932  * NFS supports Cipso labels by parsing the vfs_resource
933  * to see what the Solaris server global zone has shared.
934  * We can't support that for CIFS since resource names
935  * contain share names, not paths.
936  */
937 static int
938 smbfs_mount_label_policy(vfs_t *vfsp, void *ipaddr, int addr_type, cred_t *cr)
939 {
940 	bslabel_t	*server_sl, *mntlabel;
941 	zone_t		*mntzone = NULL;
942 	ts_label_t	*zlabel;
943 	tsol_tpc_t	*tp;
944 	ts_label_t	*tsl = NULL;
945 	int		retv;
946 
947 	/*
948 	 * Get the zone's label.  Each zone on a labeled system has a label.
949 	 */
950 	mntzone = zone_find_by_any_path(refstr_value(vfsp->vfs_mntpt), B_FALSE);
951 	zlabel = mntzone->zone_slabel;
952 	ASSERT(zlabel != NULL);
953 	label_hold(zlabel);
954 
955 	retv = EACCES;				/* assume the worst */
956 
957 	/*
958 	 * Next, get the assigned label of the remote server.
959 	 */
960 	tp = find_tpc(ipaddr, addr_type, B_FALSE);
961 	if (tp == NULL)
962 		goto out;			/* error getting host entry */
963 
964 	if (tp->tpc_tp.tp_doi != zlabel->tsl_doi)
965 		goto rel_tpc;			/* invalid domain */
966 	if ((tp->tpc_tp.host_type != UNLABELED))
967 		goto rel_tpc;			/* invalid hosttype */
968 
969 	server_sl = &tp->tpc_tp.tp_def_label;
970 	mntlabel = label2bslabel(zlabel);
971 
972 	/*
973 	 * Now compare labels to complete the MAC check.  If the labels
974 	 * are equal or if the requestor is in the global zone and has
975 	 * NET_MAC_AWARE, then allow read-write access.   (Except for
976 	 * mounts into the global zone itself; restrict these to
977 	 * read-only.)
978 	 *
979 	 * If the requestor is in some other zone, but his label
980 	 * dominates the server, then allow read-down.
981 	 *
982 	 * Otherwise, access is denied.
983 	 */
984 	if (blequal(mntlabel, server_sl) ||
985 	    (crgetzoneid(cr) == GLOBAL_ZONEID &&
986 	    getpflags(NET_MAC_AWARE, cr) != 0)) {
987 		if ((mntzone == global_zone) ||
988 		    !blequal(mntlabel, server_sl))
989 			retv = -1;		/* read-only */
990 		else
991 			retv = 0;		/* access OK */
992 	} else if (bldominates(mntlabel, server_sl)) {
993 		retv = -1;			/* read-only */
994 	} else {
995 		retv = EACCES;
996 	}
997 
998 	if (tsl != NULL)
999 		label_rele(tsl);
1000 
1001 rel_tpc:
1002 	/*LINTED*/
1003 	TPC_RELE(tp);
1004 out:
1005 	if (mntzone)
1006 		zone_rele(mntzone);
1007 	label_rele(zlabel);
1008 	return (retv);
1009 }
1010