xref: /illumos-gate/usr/src/uts/common/os/policy.c (revision 9a016c63ca347047a236dff12f0da83aac8981d1)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 /*
22  * Copyright 2006 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 
26 #pragma ident	"%Z%%M%	%I%	%E% SMI"
27 
28 #include <sys/types.h>
29 #include <sys/sysmacros.h>
30 #include <sys/param.h>
31 #include <sys/systm.h>
32 #include <sys/cred_impl.h>
33 #include <sys/vnode.h>
34 #include <sys/vfs.h>
35 #include <sys/stat.h>
36 #include <sys/errno.h>
37 #include <sys/kmem.h>
38 #include <sys/user.h>
39 #include <sys/proc.h>
40 #include <sys/acct.h>
41 #include <sys/ipc_impl.h>
42 #include <sys/cmn_err.h>
43 #include <sys/debug.h>
44 #include <sys/policy.h>
45 #include <sys/kobj.h>
46 #include <sys/msg.h>
47 #include <sys/devpolicy.h>
48 #include <c2/audit.h>
49 #include <sys/varargs.h>
50 #include <sys/modctl.h>
51 #include <sys/disp.h>
52 #include <sys/zone.h>
53 #include <inet/optcom.h>
54 #include <sys/sdt.h>
55 #include <sys/vfs.h>
56 #include <sys/mntent.h>
57 #include <sys/contract_impl.h>
58 
59 /*
60  * There are two possible layers of privilege routines and two possible
61  * levels of secpolicy.  Plus one other we may not be interested in, so
62  * we may need as many as 6 but no more.
63  */
64 #define	MAXPRIVSTACK		6
65 
66 int priv_debug = 0;
67 
68 /*
69  * This file contains the majority of the policy routines.
70  * Since the policy routines are defined by function and not
71  * by privilege, there is quite a bit of duplication of
72  * functions.
73  *
74  * The secpolicy functions must not make asssumptions about
75  * locks held or not held as any lock can be held while they're
76  * being called.
77  *
78  * Credentials are read-only so no special precautions need to
79  * be taken while locking them.
80  *
81  * When a new policy check needs to be added to the system the
82  * following procedure should be followed:
83  *
84  *		Pick an appropriate secpolicy_*() function
85  *			-> done if one exists.
86  *		Create a new secpolicy function, preferably with
87  *		a descriptive name using the standard template.
88  *		Pick an appropriate privilege for the policy.
89  *		If no appropraite privilege exists, define new one
90  *		(this should be done with extreme care; in most cases
91  *		little is gained by adding another privilege)
92  *
93  * WHY ROOT IS STILL SPECIAL.
94  *
95  * In a number of the policy functions, there are still explicit
96  * checks for uid 0.  The rationale behind these is that many root
97  * owned files/objects hold configuration information which can give full
98  * privileges to the user once written to.  To prevent escalation
99  * of privilege by allowing just a single privilege to modify root owned
100  * objects, we've added these root specific checks where we considered
101  * them necessary: modifying root owned files, changing uids to 0, etc.
102  *
103  * PRIVILEGE ESCALATION AND ZONES.
104  *
105  * A number of operations potentially allow the caller to achieve
106  * privileges beyond the ones normally required to perform the operation.
107  * For example, if allowed to create a setuid 0 executable, a process can
108  * gain privileges beyond PRIV_FILE_SETID.  Zones, however, place
109  * restrictions on the ability to gain privileges beyond those available
110  * within the zone through file and process manipulation.  Hence, such
111  * operations require that the caller have an effective set that includes
112  * all privileges available within the current zone, or all privileges
113  * if executing in the global zone.
114  *
115  * This is indicated in the priv_policy* policy checking functions
116  * through a combination of parameters.  The "priv" parameter indicates
117  * the privilege that is required, and the "allzone" parameter indicates
118  * whether or not all privileges in the zone are required.  In addition,
119  * priv can be set to PRIV_ALL to indicate that all privileges are
120  * required (regardless of zone).  There are three scenarios of interest:
121  * (1) operation requires a specific privilege
122  * (2) operation requires a specific privilege, and requires all
123  *     privileges available within the zone (or all privileges if in
124  *     the global zone)
125  * (3) operation requires all privileges, regardless of zone
126  *
127  * For (1), priv should be set to the specific privilege, and allzone
128  * should be set to B_FALSE.
129  * For (2), priv should be set to the specific privilege, and allzone
130  * should be set to B_TRUE.
131  * For (3), priv should be set to PRIV_ALL, and allzone should be set
132  * to B_FALSE.
133  *
134  */
135 
136 /*
137  * The privileges are checked against the Effective set for
138  * ordinary processes and checked against the Limit set
139  * for euid 0 processes that haven't manipulated their privilege
140  * sets.
141  */
142 #define	HAS_ALLPRIVS(cr)	priv_isfullset(&CR_OEPRIV(cr))
143 #define	ZONEPRIVS(cr)		((cr)->cr_zone->zone_privset)
144 #define	HAS_ALLZONEPRIVS(cr)	priv_issubset(ZONEPRIVS(cr), &CR_OEPRIV(cr))
145 #define	HAS_PRIVILEGE(cr, pr)	((pr) == PRIV_ALL ? \
146 					HAS_ALLPRIVS(cr) : \
147 					PRIV_ISASSERT(&CR_OEPRIV(cr), pr))
148 
149 /*
150  * Policy checking functions
151  *
152  * In future, these will migrate to several files when policy
153  * becomes more or less pluggable.
154  *
155  * For now, there's only one policy and this is it.
156  */
157 
158 /*
159  * Generic policy calls
160  *
161  * The "bottom" functions of policy control
162  */
163 
164 static char *
165 mprintf(const char *fmt, ...)
166 {
167 	va_list args;
168 	char *buf;
169 	size_t len;
170 
171 	va_start(args, fmt);
172 	len = vsnprintf(NULL, 0, fmt, args) + 1;
173 	va_end(args);
174 
175 	buf = kmem_alloc(len, KM_NOSLEEP);
176 
177 	if (buf == NULL)
178 		return (NULL);
179 
180 	va_start(args, fmt);
181 	(void) vsnprintf(buf, len, fmt, args);
182 	va_end(args);
183 
184 	return (buf);
185 }
186 
187 /*
188  * priv_policy_errmsg()
189  *
190  * Generate an error message if privilege debugging is enabled system wide
191  * or for this particular process.
192  */
193 
194 #define	FMTHDR	"%s[%d]: missing privilege \"%s\" (euid = %d, syscall = %d)"
195 #define	FMTMSG	" for \"%s\""
196 #define	FMTFUN	" needed at %s+0x%lx"
197 
198 /* The maximum size privilege format: the concatenation of the above */
199 #define	FMTMAX	FMTHDR FMTMSG FMTFUN "\n"
200 
201 static void
202 priv_policy_errmsg(const cred_t *cr, int priv, const char *msg)
203 {
204 	struct proc *me;
205 	pc_t stack[MAXPRIVSTACK];
206 	int depth;
207 	int i;
208 	char *sym;
209 	ulong_t off;
210 	const char *pname;
211 
212 	char *cmd;
213 	char fmt[sizeof (FMTMAX)];
214 
215 	if ((me = curproc) == &p0)
216 		return;
217 
218 	/* Privileges must be defined  */
219 	ASSERT(priv == PRIV_ALL || priv == PRIV_MULTIPLE ||
220 	    priv == PRIV_ALLZONE || priv == PRIV_GLOBAL ||
221 	    priv_getbynum(priv) != NULL);
222 
223 	if (priv == PRIV_ALLZONE && INGLOBALZONE(me))
224 		priv = PRIV_ALL;
225 
226 	if (curthread->t_pre_sys)
227 		ttolwp(curthread)->lwp_badpriv = (short)priv;
228 
229 	if (priv_debug == 0 && (CR_FLAGS(cr) & PRIV_DEBUG) == 0)
230 		return;
231 
232 	(void) strcpy(fmt, FMTHDR);
233 
234 	if (me->p_user.u_comm[0])
235 		cmd = &me->p_user.u_comm[0];
236 	else
237 		cmd = "priv_policy";
238 
239 	if (msg != NULL && *msg != '\0') {
240 		(void) strcat(fmt, FMTMSG);
241 	} else {
242 		(void) strcat(fmt, "%s");
243 		msg = "";
244 	}
245 
246 	sym = NULL;
247 
248 	depth = getpcstack(stack, MAXPRIVSTACK);
249 
250 	/*
251 	 * Try to find the first interesting function on the stack.
252 	 * priv_policy* that's us, so completely uninteresting.
253 	 * suser(), drv_priv(), secpolicy_* are also called from
254 	 * too many locations to convey useful information.
255 	 */
256 	for (i = 0; i < depth; i++) {
257 		sym = kobj_getsymname((uintptr_t)stack[i], &off);
258 		if (sym != NULL &&
259 		    strstr(sym, "hasprocperm") == 0 &&
260 		    strcmp("suser", sym) != 0 &&
261 		    strcmp("ipcaccess", sym) != 0 &&
262 		    strcmp("drv_priv", sym) != 0 &&
263 		    strncmp("secpolicy_", sym, 10) != 0 &&
264 		    strncmp("priv_policy", sym, 11) != 0)
265 			break;
266 	}
267 
268 	if (sym != NULL)
269 		(void) strcat(fmt, FMTFUN);
270 
271 	(void) strcat(fmt, "\n");
272 
273 	switch (priv) {
274 	case PRIV_ALL:
275 		pname = "ALL";
276 		break;
277 	case PRIV_MULTIPLE:
278 		pname = "MULTIPLE";
279 		break;
280 	case PRIV_ALLZONE:
281 		pname = "ZONE";
282 		break;
283 	case PRIV_GLOBAL:
284 		pname = "GLOBAL";
285 		break;
286 	default:
287 		pname = priv_getbynum(priv);
288 		break;
289 	}
290 
291 	if (CR_FLAGS(cr) & PRIV_DEBUG) {
292 		/* Remember last message, just like lwp_badpriv. */
293 		if (curthread->t_pdmsg != NULL) {
294 			kmem_free(curthread->t_pdmsg,
295 			    strlen(curthread->t_pdmsg) + 1);
296 		}
297 
298 		curthread->t_pdmsg = mprintf(fmt, cmd, me->p_pid, pname,
299 			    cr->cr_uid, curthread->t_sysnum, msg, sym, off);
300 
301 		curthread->t_post_sys = 1;
302 	} else {
303 		cmn_err(CE_NOTE, fmt, cmd, me->p_pid, pname, cr->cr_uid,
304 		    curthread->t_sysnum, msg, sym, off);
305 	}
306 }
307 
308 /*
309  * Audit failure, log error message.
310  */
311 static void
312 priv_policy_err(const cred_t *cr, int priv, boolean_t allzone, const char *msg)
313 {
314 
315 #ifdef C2_AUDIT
316 	if (audit_active)
317 		audit_priv(priv, allzone ? ZONEPRIVS(cr) : NULL, 0);
318 #endif
319 	DTRACE_PROBE2(priv__err, int, priv, boolean_t, allzone);
320 
321 	if (priv_debug || (CR_FLAGS(cr) & PRIV_DEBUG) ||
322 	    curthread->t_pre_sys) {
323 		if (allzone && !HAS_ALLZONEPRIVS(cr)) {
324 			priv_policy_errmsg(cr, PRIV_ALLZONE, msg);
325 		} else {
326 			ASSERT(!HAS_PRIVILEGE(cr, priv));
327 			priv_policy_errmsg(cr, priv, msg);
328 		}
329 	}
330 }
331 
332 /*
333  * priv_policy()
334  * return 0 or error.
335  * See block comment above for a description of "priv" and "allzone" usage.
336  */
337 int
338 priv_policy(const cred_t *cr, int priv, boolean_t allzone, int err,
339     const char *msg)
340 {
341 	if (HAS_PRIVILEGE(cr, priv) && (!allzone || HAS_ALLZONEPRIVS(cr))) {
342 		if ((allzone || priv == PRIV_ALL ||
343 		    !PRIV_ISASSERT(priv_basic, priv)) &&
344 		    !servicing_interrupt()) {
345 			u.u_acflag |= ASU;		/* Needed for SVVS */
346 #ifdef C2_AUDIT
347 			if (audit_active)
348 				audit_priv(priv,
349 				    allzone ? ZONEPRIVS(cr) : NULL, 1);
350 #endif
351 		}
352 		err = 0;
353 		DTRACE_PROBE2(priv__ok, int, priv, boolean_t, allzone);
354 	} else if (!servicing_interrupt()) {
355 		/* Failure audited in this procedure */
356 		priv_policy_err(cr, priv, allzone, msg);
357 	}
358 
359 	return (err);
360 }
361 
362 /*
363  * Return B_TRUE for sufficient privileges, B_FALSE for insufficient privileges.
364  */
365 boolean_t
366 priv_policy_choice(const cred_t *cr, int priv, boolean_t allzone)
367 {
368 	boolean_t res = HAS_PRIVILEGE(cr, priv) &&
369 	    (!allzone || HAS_ALLZONEPRIVS(cr));
370 
371 #ifdef C2_AUDIT
372 	/* Audit success only */
373 	if (res && audit_active &&
374 	    (allzone || priv == PRIV_ALL || !PRIV_ISASSERT(priv_basic, priv)) &&
375 	    !servicing_interrupt()) {
376 		audit_priv(priv, allzone ? ZONEPRIVS(cr) : NULL, 1);
377 	}
378 #endif
379 	if (res) {
380 		DTRACE_PROBE2(priv__ok, int, priv, boolean_t, allzone);
381 	} else {
382 		DTRACE_PROBE2(priv__err, int, priv, boolean_t, allzone);
383 	}
384 	return (res);
385 }
386 
387 /*
388  * Non-auditing variant of priv_policy_choice().
389  */
390 boolean_t
391 priv_policy_only(const cred_t *cr, int priv, boolean_t allzone)
392 {
393 	boolean_t res = HAS_PRIVILEGE(cr, priv) &&
394 	    (!allzone || HAS_ALLZONEPRIVS(cr));
395 
396 	if (res) {
397 		DTRACE_PROBE2(priv__ok, int, priv, boolean_t, allzone);
398 	} else {
399 		DTRACE_PROBE2(priv__err, int, priv, boolean_t, allzone);
400 	}
401 	return (res);
402 }
403 
404 /*
405  * Check whether all privileges in the required set are present.
406  */
407 static int
408 secpolicy_require_set(const cred_t *cr, const priv_set_t *req, const char *msg)
409 {
410 	int priv;
411 	int pfound = -1;
412 	priv_set_t pset;
413 
414 	if (req == PRIV_FULLSET ? HAS_ALLPRIVS(cr) : priv_issubset(req,
415 							    &CR_OEPRIV(cr))) {
416 		return (0);
417 	}
418 
419 	if (req == PRIV_FULLSET || priv_isfullset(req)) {
420 		priv_policy_err(cr, PRIV_ALL, B_FALSE, msg);
421 		return (EACCES);
422 	}
423 
424 	pset = CR_OEPRIV(cr);		/* present privileges */
425 	priv_inverse(&pset);		/* all non present privileges */
426 	priv_intersect(req, &pset);	/* the actual missing privs */
427 
428 #ifdef C2_AUDIT
429 	if (audit_active)
430 		audit_priv(PRIV_NONE, &pset, 0);
431 #endif
432 	/*
433 	 * Privilege debugging; special case "one privilege in set".
434 	 */
435 	if (priv_debug || (CR_FLAGS(cr) & PRIV_DEBUG) || curthread->t_pre_sys) {
436 		for (priv = 0; priv < nprivs; priv++) {
437 			if (priv_ismember(&pset, priv)) {
438 				if (pfound != -1) {
439 					/* Multiple missing privs */
440 					priv_policy_errmsg(cr, PRIV_MULTIPLE,
441 								    msg);
442 					return (EACCES);
443 				}
444 				pfound = priv;
445 			}
446 		}
447 		ASSERT(pfound != -1);
448 		/* Just the one missing privilege */
449 		priv_policy_errmsg(cr, pfound, msg);
450 	}
451 
452 	return (EACCES);
453 }
454 
455 /*
456  * Called when an operation requires that the caller be in the
457  * global zone, regardless of privilege.
458  */
459 static int
460 priv_policy_global(const cred_t *cr)
461 {
462 	if (crgetzoneid(cr) == GLOBAL_ZONEID)
463 		return (0);	/* success */
464 
465 	if (priv_debug || (CR_FLAGS(cr) & PRIV_DEBUG) ||
466 	    curthread->t_pre_sys) {
467 		priv_policy_errmsg(cr, PRIV_GLOBAL, NULL);
468 	}
469 	return (EPERM);
470 }
471 
472 /*
473  * Changing process priority
474  */
475 int
476 secpolicy_setpriority(const cred_t *cr)
477 {
478 	return (PRIV_POLICY(cr, PRIV_PROC_PRIOCNTL, B_FALSE, EPERM, NULL));
479 }
480 
481 /*
482  * Binding to a privileged port, port must be specified in host byte
483  * order.
484  */
485 int
486 secpolicy_net_privaddr(const cred_t *cr, in_port_t port)
487 {
488 	/*
489 	 * NFS ports, these are extra privileged ports, allow bind
490 	 * only if the SYS_NFS privilege is present.
491 	 */
492 	if (port == 2049 || port == 4045)
493 		return (PRIV_POLICY(cr, PRIV_SYS_NFS, B_FALSE, EACCES,
494 		    "NFS port"));
495 	else
496 		return (PRIV_POLICY(cr, PRIV_NET_PRIVADDR, B_FALSE, EACCES,
497 		    NULL));
498 }
499 
500 /*
501  * Binding to a multilevel port on a trusted (labeled) system.
502  */
503 int
504 secpolicy_net_bindmlp(const cred_t *cr)
505 {
506 	return (PRIV_POLICY(cr, PRIV_NET_BINDMLP, B_FALSE, EACCES,
507 	    NULL));
508 }
509 
510 /*
511  * Allow a communication between a zone and an unlabeled host when their
512  * labels don't match.
513  */
514 int
515 secpolicy_net_mac_aware(const cred_t *cr)
516 {
517 	return (PRIV_POLICY(cr, PRIV_NET_MAC_AWARE, B_FALSE, EACCES,
518 	    NULL));
519 }
520 
521 /*
522  * Common routine which determines whether a given credential can
523  * act on a given mount.
524  * When called through mount, the parameter needoptcheck is a pointer
525  * to a boolean variable which will be set to either true or false,
526  * depending on whether the mount policy should change the mount options.
527  * In all other cases, needoptcheck should be a NULL pointer.
528  */
529 static int
530 secpolicy_fs_common(cred_t *cr, vnode_t *mvp, const vfs_t *vfsp,
531     boolean_t *needoptcheck)
532 {
533 	boolean_t allzone = B_FALSE;
534 	boolean_t mounting = needoptcheck != NULL;
535 
536 	/*
537 	 * Short circuit the following cases:
538 	 *	vfsp == NULL or mvp == NULL (pure privilege check)
539 	 *	have all privileges - no further checks required
540 	 *	and no mount options need to be set.
541 	 */
542 	if (vfsp == NULL || mvp == NULL || HAS_ALLPRIVS(cr)) {
543 		if (mounting)
544 			*needoptcheck = B_FALSE;
545 
546 		return (PRIV_POLICY(cr, PRIV_SYS_MOUNT, allzone, EPERM, NULL));
547 	}
548 
549 	/*
550 	 * When operating on an existing mount (either we're not mounting
551 	 * or we're doing a remount and VFS_REMOUNT will be set), zones
552 	 * can operate only on mounts established by the zone itself.
553 	 */
554 	if (!mounting || (vfsp->vfs_flag & VFS_REMOUNT) != 0) {
555 		zoneid_t zoneid = crgetzoneid(cr);
556 
557 		if (zoneid != GLOBAL_ZONEID &&
558 		    vfsp->vfs_zone->zone_id != zoneid) {
559 			return (EPERM);
560 		}
561 	}
562 
563 	if (mounting)
564 		*needoptcheck = B_TRUE;
565 
566 	/*
567 	 * Overlay mounts may hide important stuff; if you can't write to a
568 	 * mount point but would be able to mount on top of it, you can
569 	 * escalate your privileges.
570 	 * So we go about asking the same questions namefs does when it
571 	 * decides whether you can mount over a file or not but with the
572 	 * added restriction that you can only mount on top of a regular
573 	 * file or directory.
574 	 * If we have all the zone's privileges, we skip all other checks,
575 	 * or else we may actually get in trouble inside the automounter.
576 	 */
577 	if ((mvp->v_flag & VROOT) != 0 ||
578 	    (mvp->v_type != VDIR && mvp->v_type != VREG) ||
579 	    HAS_ALLZONEPRIVS(cr)) {
580 		allzone = B_TRUE;
581 	} else {
582 		vattr_t va;
583 		int err;
584 
585 		va.va_mask = AT_UID|AT_MODE;
586 		err = VOP_GETATTR(mvp, &va, 0, cr);
587 		if (err != 0)
588 			return (err);
589 
590 		if ((err = secpolicy_vnode_owner(cr, va.va_uid)) != 0)
591 			return (err);
592 
593 		if ((va.va_mode & VWRITE) == 0 &&
594 		    secpolicy_vnode_access(cr, mvp, va.va_uid, VWRITE) != 0) {
595 			return (EACCES);
596 		}
597 	}
598 	return (PRIV_POLICY(cr, PRIV_SYS_MOUNT, allzone, EPERM, NULL));
599 }
600 
601 extern vnode_t *rootvp;
602 extern vfs_t *rootvfs;
603 
604 int
605 secpolicy_fs_mount(cred_t *cr, vnode_t *mvp, struct vfs *vfsp)
606 {
607 	boolean_t needoptchk;
608 	int error;
609 
610 	/*
611 	 * If it's a remount, get the underlying mount point,
612 	 * except for the root where we use the rootvp.
613 	 */
614 	if ((vfsp->vfs_flag & VFS_REMOUNT) != 0) {
615 		if (vfsp == rootvfs)
616 			mvp = rootvp;
617 		else
618 			mvp = vfsp->vfs_vnodecovered;
619 	}
620 
621 	error = secpolicy_fs_common(cr, mvp, vfsp, &needoptchk);
622 
623 	if (error == 0 && needoptchk) {
624 		boolean_t amsuper = HAS_ALLZONEPRIVS(cr);
625 
626 		/*
627 		 * Third check; if we don't have either "nosuid" or
628 		 * both "nosetuid" and "nodevices", then we add
629 		 * "nosuid"; this depends on how the current
630 		 * implementation works (it first checks nosuid).  In a
631 		 * zone, a user with all zone privileges can mount with
632 		 * "setuid" but never with "devices".
633 		 */
634 		if (!vfs_optionisset(vfsp, MNTOPT_NOSUID, NULL) &&
635 		    (!vfs_optionisset(vfsp, MNTOPT_NODEVICES, NULL) ||
636 		    !vfs_optionisset(vfsp, MNTOPT_NOSETUID, NULL))) {
637 			if (crgetzoneid(cr) == GLOBAL_ZONEID || !amsuper)
638 				vfs_setmntopt(vfsp, MNTOPT_NOSUID, NULL, 0);
639 			else
640 				vfs_setmntopt(vfsp, MNTOPT_NODEVICES, NULL, 0);
641 		}
642 		/*
643 		 * If we're not the local super user, we set the "restrict"
644 		 * option to indicate to automountd that this mount should
645 		 * be handled with care.
646 		 */
647 		if (!amsuper)
648 			vfs_setmntopt(vfsp, MNTOPT_RESTRICT, NULL, 0);
649 
650 	}
651 	return (error);
652 }
653 
654 /*
655  * Does the policy computations for "ownership" of a mount;
656  * here ownership is defined as the ability to "mount"
657  * the filesystem originally.  The rootvfs doesn't cover any
658  * vnodes; we attribute its ownership to the rootvp.
659  */
660 static int
661 secpolicy_fs_owner(cred_t *cr, const struct vfs *vfsp)
662 {
663 	vnode_t *mvp;
664 
665 	if (vfsp == NULL)
666 		mvp = NULL;
667 	else if (vfsp == rootvfs)
668 		mvp = rootvp;
669 	else
670 		mvp = vfsp->vfs_vnodecovered;
671 
672 	return (secpolicy_fs_common(cr, mvp, vfsp, NULL));
673 }
674 
675 int
676 secpolicy_fs_unmount(cred_t *cr, struct vfs *vfsp)
677 {
678 	return (secpolicy_fs_owner(cr, vfsp));
679 }
680 
681 /*
682  * Quotas are a resource, but if one has the ability to mount a filesystem, he
683  * should be able to modify quotas on it.
684  */
685 int
686 secpolicy_fs_quota(const cred_t *cr, const vfs_t *vfsp)
687 {
688 	return (secpolicy_fs_owner((cred_t *)cr, vfsp));
689 }
690 
691 /*
692  * Exceeding minfree: also a per-mount resource constraint.
693  */
694 int
695 secpolicy_fs_minfree(const cred_t *cr, const vfs_t *vfsp)
696 {
697 	return (secpolicy_fs_owner((cred_t *)cr, vfsp));
698 }
699 
700 int
701 secpolicy_fs_config(const cred_t *cr, const vfs_t *vfsp)
702 {
703 	return (secpolicy_fs_owner((cred_t *)cr, vfsp));
704 }
705 
706 /* ARGSUSED */
707 int
708 secpolicy_fs_linkdir(const cred_t *cr, const vfs_t *vfsp)
709 {
710 	return (PRIV_POLICY(cr, PRIV_SYS_LINKDIR, B_FALSE, EPERM, NULL));
711 }
712 
713 /*
714  * Name:        secpolicy_vnode_access()
715  *
716  * Parameters:  Process credential
717  *		vnode
718  *		uid of owner of vnode
719  *		permission bits not granted to the caller when examining
720  *		file mode bits (i.e., when a process wants to open a
721  *		mode 444 file for VREAD|VWRITE, this function should be
722  *		called only with a VWRITE argument).
723  *
724  * Normal:      Verifies that cred has the appropriate privileges to
725  *              override the mode bits that were denied.
726  *
727  * Override:    file_dac_execute - if VEXEC bit was denied and vnode is
728  *                      not a directory.
729  *              file_dac_read - if VREAD bit was denied.
730  *              file_dac_search - if VEXEC bit was denied and vnode is
731  *                      a directory.
732  *              file_dac_write - if VWRITE bit was denied.
733  *
734  *		Root owned files are special cased to protect system
735  *		configuration files and such.
736  *
737  * Output:      EACCES - if privilege check fails.
738  */
739 
740 /* ARGSUSED */
741 int
742 secpolicy_vnode_access(const cred_t *cr, vnode_t *vp, uid_t owner, mode_t mode)
743 {
744 	if ((mode & VREAD) &&
745 	    PRIV_POLICY(cr, PRIV_FILE_DAC_READ, B_FALSE, EACCES, NULL) != 0)
746 		return (EACCES);
747 
748 	if (mode & VWRITE) {
749 		boolean_t allzone;
750 
751 		if (owner == 0 && cr->cr_uid != 0)
752 			allzone = B_TRUE;
753 		else
754 			allzone = B_FALSE;
755 		if (PRIV_POLICY(cr, PRIV_FILE_DAC_WRITE, allzone, EACCES, NULL)
756 		    != 0)
757 			return (EACCES);
758 	}
759 
760 	if (mode & VEXEC) {
761 		/*
762 		 * Directories use file_dac_search to override the execute bit.
763 		 */
764 		vtype_t vtype = vp->v_type;
765 
766 		if (vtype == VDIR)
767 			return (PRIV_POLICY(cr, PRIV_FILE_DAC_SEARCH, B_FALSE,
768 			    EACCES, NULL));
769 		else
770 			return (PRIV_POLICY(cr, PRIV_FILE_DAC_EXECUTE, B_FALSE,
771 			    EACCES, NULL));
772 	}
773 	return (0);
774 }
775 
776 /*
777  * Name:	secpolicy_vnode_setid_modify()
778  *
779  * Normal:	verify that subject can set the file setid flags.
780  *
781  * Output:	EPERM - if not privileged.
782  */
783 
784 static int
785 secpolicy_vnode_setid_modify(const cred_t *cr, uid_t owner)
786 {
787 	/* If changing to suid root, must have all zone privs */
788 	boolean_t allzone = B_TRUE;
789 
790 	if (owner != 0) {
791 		if (owner == cr->cr_uid)
792 			return (0);
793 		allzone = B_FALSE;
794 	}
795 	return (PRIV_POLICY(cr, PRIV_FILE_SETID, allzone, EPERM, NULL));
796 }
797 
798 /*
799  * Are we allowed to retain the set-uid/set-gid bits when
800  * changing ownership or when writing to a file?
801  * "issuid" should be true when set-uid; only in that case
802  * root ownership is checked (setgid is assumed).
803  */
804 int
805 secpolicy_vnode_setid_retain(const cred_t *cred, boolean_t issuidroot)
806 {
807 	if (issuidroot && !HAS_ALLZONEPRIVS(cred))
808 		return (EPERM);
809 
810 	return (!PRIV_POLICY_CHOICE(cred, PRIV_FILE_SETID, B_FALSE));
811 }
812 
813 /*
814  * Name:	secpolicy_vnode_setids_setgids()
815  *
816  * Normal:	verify that subject can set the file setgid flag.
817  *
818  * Output:	EPERM - if not privileged
819  */
820 
821 int
822 secpolicy_vnode_setids_setgids(const cred_t *cred, gid_t gid)
823 {
824 	if (!groupmember(gid, cred))
825 		return (PRIV_POLICY(cred, PRIV_FILE_SETID, B_FALSE, EPERM,
826 		    NULL));
827 	return (0);
828 }
829 
830 /*
831  * Create a file with a group different than any of the groups allowed:
832  * the group of the directory the file is created in, the effective
833  * group or any of the supplementary groups.
834  */
835 int
836 secpolicy_vnode_create_gid(const cred_t *cred)
837 {
838 	if (HAS_PRIVILEGE(cred, PRIV_FILE_CHOWN))
839 		return (PRIV_POLICY(cred, PRIV_FILE_CHOWN, B_FALSE, EPERM,
840 		    NULL));
841 	else
842 		return (PRIV_POLICY(cred, PRIV_FILE_CHOWN_SELF, B_FALSE, EPERM,
843 		    NULL));
844 }
845 
846 /*
847  * Name:	secpolicy_vnode_utime_modify()
848  *
849  * Normal:	verify that subject can modify the utime on a file.
850  *
851  * Output:	EPERM - if access denied.
852  */
853 
854 static int
855 secpolicy_vnode_utime_modify(const cred_t *cred)
856 {
857 	return (PRIV_POLICY(cred, PRIV_FILE_OWNER, B_FALSE, EPERM,
858 	    "modify file times"));
859 }
860 
861 
862 /*
863  * Name:	secpolicy_vnode_setdac()
864  *
865  * Normal:	verify that subject can modify the mode of a file.
866  *		allzone privilege needed when modifying root owned object.
867  *
868  * Output:	EPERM - if access denied.
869  */
870 
871 int
872 secpolicy_vnode_setdac(const cred_t *cred, uid_t owner)
873 {
874 	if (owner == cred->cr_uid)
875 		return (0);
876 
877 	return (PRIV_POLICY(cred, PRIV_FILE_OWNER, owner == 0, EPERM, NULL));
878 }
879 /*
880  * Name:	secpolicy_vnode_stky_modify()
881  *
882  * Normal:	verify that subject can make a file a "sticky".
883  *
884  * Output:	EPERM - if access denied.
885  */
886 
887 int
888 secpolicy_vnode_stky_modify(const cred_t *cred)
889 {
890 	return (PRIV_POLICY(cred, PRIV_SYS_CONFIG, B_FALSE, EPERM,
891 	    "set file sticky"));
892 }
893 
894 /*
895  * Policy determines whether we can remove an entry from a directory,
896  * regardless of permission bits.
897  */
898 int
899 secpolicy_vnode_remove(const cred_t *cr)
900 {
901 	return (PRIV_POLICY(cr, PRIV_FILE_OWNER, B_FALSE, EACCES,
902 	    "sticky directory"));
903 }
904 
905 int
906 secpolicy_vnode_owner(const cred_t *cr, uid_t owner)
907 {
908 	boolean_t allzone = (owner == 0);
909 
910 	if (owner == cr->cr_uid)
911 		return (0);
912 
913 	return (PRIV_POLICY(cr, PRIV_FILE_OWNER, allzone, EPERM, NULL));
914 }
915 
916 void
917 secpolicy_setid_clear(vattr_t *vap, cred_t *cr)
918 {
919 	if ((vap->va_mode & (S_ISUID | S_ISGID)) != 0 &&
920 	    secpolicy_vnode_setid_retain(cr,
921 	    (vap->va_mode & S_ISUID) != 0 &&
922 	    (vap->va_mask & AT_UID) != 0 && vap->va_uid == 0) != 0) {
923 		vap->va_mask |= AT_MODE;
924 		vap->va_mode &= ~(S_ISUID|S_ISGID);
925 	}
926 }
927 
928 /*
929  * This function checks the policy decisions surrounding the
930  * vop setattr call.
931  *
932  * It should be called after sufficient locks have been established
933  * on the underlying data structures.  No concurrent modifications
934  * should be allowed.
935  *
936  * The caller must pass in unlocked version of its vaccess function
937  * this is required because vop_access function should lock the
938  * node for reading.  A three argument function should be defined
939  * which accepts the following argument:
940  * 	A pointer to the internal "node" type (inode *)
941  *	vnode access bits (VREAD|VWRITE|VEXEC)
942  *	a pointer to the credential
943  *
944  * This function makes the following policy decisions:
945  *
946  *		- change permissions
947  *			- permission to change file mode if not owner
948  *			- permission to add sticky bit to non-directory
949  *			- permission to add set-gid bit
950  *
951  * The ovap argument should include AT_MODE|AT_UID|AT_GID.
952  *
953  * If the vap argument does not include AT_MODE, the mode will be copied from
954  * ovap.  In certain situations set-uid/set-gid bits need to be removed;
955  * this is done by marking vap->va_mask to include AT_MODE and va_mode
956  * is updated to the newly computed mode.
957  */
958 
959 int
960 secpolicy_vnode_setattr(cred_t *cr, struct vnode *vp, struct vattr *vap,
961 	const struct vattr *ovap, int flags,
962 	int unlocked_access(void *, int, cred_t *),
963 	void *node)
964 {
965 	int mask = vap->va_mask;
966 	int error = 0;
967 
968 	if (mask & AT_SIZE) {
969 		if (vp->v_type == VDIR) {
970 			error = EISDIR;
971 			goto out;
972 		}
973 		error = unlocked_access(node, VWRITE, cr);
974 		if (error)
975 			goto out;
976 	}
977 	if (mask & AT_MODE) {
978 		/*
979 		 * If not the owner of the file then check privilege
980 		 * for two things: the privilege to set the mode at all
981 		 * and, if we're setting setuid, we also need permissions
982 		 * to add the set-uid bit, if we're not the owner.
983 		 * In the specific case of creating a set-uid root
984 		 * file, we need even more permissions.
985 		 */
986 		if ((error = secpolicy_vnode_setdac(cr, ovap->va_uid)) != 0)
987 			goto out;
988 
989 		if ((vap->va_mode & S_ISUID) != 0 &&
990 		    (error = secpolicy_vnode_setid_modify(cr,
991 							ovap->va_uid)) != 0) {
992 			goto out;
993 		}
994 
995 		/*
996 		 * Check privilege if attempting to set the
997 		 * sticky bit on a non-directory.
998 		 */
999 		if (vp->v_type != VDIR && (vap->va_mode & S_ISVTX) != 0 &&
1000 		    secpolicy_vnode_stky_modify(cr) != 0) {
1001 			vap->va_mode &= ~S_ISVTX;
1002 		}
1003 
1004 		/*
1005 		 * Check for privilege if attempting to set the
1006 		 * group-id bit.
1007 		 */
1008 		if ((vap->va_mode & S_ISGID) != 0 &&
1009 		    secpolicy_vnode_setids_setgids(cr, ovap->va_gid) != 0) {
1010 			vap->va_mode &= ~S_ISGID;
1011 		}
1012 
1013 	} else
1014 		vap->va_mode = ovap->va_mode;
1015 
1016 	if (mask & (AT_UID|AT_GID)) {
1017 		boolean_t checkpriv = B_FALSE;
1018 		int priv;
1019 		boolean_t allzone = B_FALSE;
1020 
1021 		/*
1022 		 * Chowning files.
1023 		 *
1024 		 * If you are the file owner:
1025 		 *	chown to other uid		FILE_CHOWN_SELF
1026 		 *	chown to gid (non-member) 	FILE_CHOWN_SELF
1027 		 *	chown to gid (member) 		<none>
1028 		 *
1029 		 * Instead of PRIV_FILE_CHOWN_SELF, FILE_CHOWN is also
1030 		 * acceptable but the first one is reported when debugging.
1031 		 *
1032 		 * If you are not the file owner:
1033 		 *	chown from root			PRIV_FILE_CHOWN + zone
1034 		 *	chown from other to any		PRIV_FILE_CHOWN
1035 		 *
1036 		 */
1037 		if (cr->cr_uid != ovap->va_uid) {
1038 			checkpriv = B_TRUE;
1039 			allzone = (ovap->va_uid == 0);
1040 			priv = PRIV_FILE_CHOWN;
1041 		} else {
1042 			if (((mask & AT_UID) && vap->va_uid != ovap->va_uid) ||
1043 			    ((mask & AT_GID) && vap->va_gid != ovap->va_gid &&
1044 			    !groupmember(vap->va_gid, cr))) {
1045 				checkpriv = B_TRUE;
1046 				priv = HAS_PRIVILEGE(cr, PRIV_FILE_CHOWN) ?
1047 				    PRIV_FILE_CHOWN : PRIV_FILE_CHOWN_SELF;
1048 			}
1049 		}
1050 		/*
1051 		 * If necessary, check privilege to see if update can be done.
1052 		 */
1053 		if (checkpriv &&
1054 		    (error = PRIV_POLICY(cr, priv, allzone, EPERM, NULL))
1055 		    != 0) {
1056 			goto out;
1057 		}
1058 
1059 		/*
1060 		 * If the file has either the set UID or set GID bits
1061 		 * set and the caller can set the bits, then leave them.
1062 		 */
1063 		secpolicy_setid_clear(vap, cr);
1064 	}
1065 	if (mask & (AT_ATIME|AT_MTIME)) {
1066 		/*
1067 		 * If not the file owner and not otherwise privileged,
1068 		 * always return an error when setting the
1069 		 * time other than the current (ATTR_UTIME flag set).
1070 		 * If setting the current time (ATTR_UTIME not set) then
1071 		 * unlocked_access will check permissions according to policy.
1072 		 */
1073 		if (cr->cr_uid != ovap->va_uid) {
1074 			if (flags & ATTR_UTIME)
1075 				error = secpolicy_vnode_utime_modify(cr);
1076 			else {
1077 				error = unlocked_access(node, VWRITE, cr);
1078 				if (error == EACCES &&
1079 				    secpolicy_vnode_utime_modify(cr) == 0)
1080 					error = 0;
1081 			}
1082 			if (error)
1083 				goto out;
1084 		}
1085 	}
1086 out:
1087 	return (error);
1088 }
1089 
1090 /*
1091  * Name:	secpolicy_pcfs_modify_bootpartition()
1092  *
1093  * Normal:	verify that subject can modify a pcfs boot partition.
1094  *
1095  * Output:	EACCES - if privilege check failed.
1096  */
1097 /*ARGSUSED*/
1098 int
1099 secpolicy_pcfs_modify_bootpartition(const cred_t *cred)
1100 {
1101 	return (PRIV_POLICY(cred, PRIV_ALL, B_FALSE, EACCES,
1102 	    "modify pcfs boot partition"));
1103 }
1104 
1105 /*
1106  * System V IPC routines
1107  */
1108 int
1109 secpolicy_ipc_owner(const cred_t *cr, const struct kipc_perm *ip)
1110 {
1111 	if (crgetzoneid(cr) != ip->ipc_zoneid ||
1112 	    (cr->cr_uid != ip->ipc_uid && cr->cr_uid != ip->ipc_cuid)) {
1113 		boolean_t allzone = B_FALSE;
1114 		if (ip->ipc_uid == 0 || ip->ipc_cuid == 0)
1115 			allzone = B_TRUE;
1116 		return (PRIV_POLICY(cr, PRIV_IPC_OWNER, allzone, EPERM, NULL));
1117 	}
1118 	return (0);
1119 }
1120 
1121 int
1122 secpolicy_ipc_config(const cred_t *cr)
1123 {
1124 	return (PRIV_POLICY(cr, PRIV_SYS_IPC_CONFIG, B_FALSE, EPERM, NULL));
1125 }
1126 
1127 int
1128 secpolicy_ipc_access(const cred_t *cr, const struct kipc_perm *ip, mode_t mode)
1129 {
1130 
1131 	boolean_t allzone = B_FALSE;
1132 
1133 	ASSERT((mode & (MSG_R|MSG_W)) != 0);
1134 
1135 	if ((mode & MSG_R) &&
1136 	    PRIV_POLICY(cr, PRIV_IPC_DAC_READ, allzone, EACCES, NULL) != 0)
1137 		return (EACCES);
1138 
1139 	if (mode & MSG_W) {
1140 		if (cr->cr_uid != 0 && (ip->ipc_uid == 0 || ip->ipc_cuid == 0))
1141 			allzone = B_TRUE;
1142 
1143 		return (PRIV_POLICY(cr, PRIV_IPC_DAC_WRITE, allzone, EACCES,
1144 		    NULL));
1145 	}
1146 	return (0);
1147 }
1148 
1149 int
1150 secpolicy_rsm_access(const cred_t *cr, uid_t owner, mode_t mode)
1151 {
1152 	boolean_t allzone = B_FALSE;
1153 
1154 	ASSERT((mode & (MSG_R|MSG_W)) != 0);
1155 
1156 	if ((mode & MSG_R) &&
1157 	    PRIV_POLICY(cr, PRIV_IPC_DAC_READ, allzone, EACCES, NULL) != 0)
1158 		return (EACCES);
1159 
1160 	if (mode & MSG_W) {
1161 		if (cr->cr_uid != 0 && owner == 0)
1162 			allzone = B_TRUE;
1163 
1164 		return (PRIV_POLICY(cr, PRIV_IPC_DAC_WRITE, allzone, EACCES,
1165 		    NULL));
1166 	}
1167 	return (0);
1168 }
1169 
1170 /*
1171  * Audit configuration.
1172  */
1173 int
1174 secpolicy_audit_config(const cred_t *cr)
1175 {
1176 	return (PRIV_POLICY(cr, PRIV_SYS_AUDIT, B_FALSE, EPERM, NULL));
1177 }
1178 
1179 /*
1180  * Audit record generation.
1181  */
1182 int
1183 secpolicy_audit_modify(const cred_t *cr)
1184 {
1185 	return (PRIV_POLICY(cr, PRIV_PROC_AUDIT, B_FALSE, EPERM, NULL));
1186 }
1187 
1188 /*
1189  * Get audit attributes.
1190  * Either PRIV_SYS_AUDIT or PRIV_PROC_AUDIT required; report the
1191  * "Least" of the two privileges on error.
1192  */
1193 int
1194 secpolicy_audit_getattr(const cred_t *cr)
1195 {
1196 	if (!PRIV_POLICY_ONLY(cr, PRIV_SYS_AUDIT, B_FALSE)) {
1197 		return (PRIV_POLICY(cr, PRIV_PROC_AUDIT, B_FALSE, EPERM,
1198 		    NULL));
1199 	} else {
1200 		return (PRIV_POLICY(cr, PRIV_SYS_AUDIT, B_FALSE, EPERM, NULL));
1201 	}
1202 }
1203 
1204 
1205 /*
1206  * Locking physical memory
1207  */
1208 int
1209 secpolicy_lock_memory(const cred_t *cr)
1210 {
1211 	return (PRIV_POLICY(cr, PRIV_PROC_LOCK_MEMORY, B_FALSE, EPERM, NULL));
1212 }
1213 
1214 /*
1215  * Accounting (both acct(2) and exacct).
1216  */
1217 int
1218 secpolicy_acct(const cred_t *cr)
1219 {
1220 	return (PRIV_POLICY(cr, PRIV_SYS_ACCT, B_FALSE, EPERM, NULL));
1221 }
1222 
1223 /*
1224  * Is this process privileged to change its uids at will?
1225  * Uid 0 is still considered "special" and having the SETID
1226  * privilege is not sufficient to get uid 0.
1227  * Files are owned by root, so the privilege would give
1228  * full access and euid 0 is still effective.
1229  *
1230  * If you have the privilege and euid 0 only then do you
1231  * get the powers of root wrt uid 0.
1232  *
1233  * For gid manipulations, this is should be called with an
1234  * uid of -1.
1235  *
1236  */
1237 int
1238 secpolicy_allow_setid(const cred_t *cr, uid_t newuid, boolean_t checkonly)
1239 {
1240 	boolean_t allzone = B_FALSE;
1241 
1242 	if (newuid == 0 && cr->cr_uid != 0 && cr->cr_suid != 0 &&
1243 	    cr->cr_ruid != 0) {
1244 		allzone = B_TRUE;
1245 	}
1246 
1247 	return (checkonly ? !PRIV_POLICY_ONLY(cr, PRIV_PROC_SETID, allzone) :
1248 	    PRIV_POLICY(cr, PRIV_PROC_SETID, allzone, EPERM, NULL));
1249 }
1250 
1251 
1252 /*
1253  * Acting on a different process: if the mode is for writing,
1254  * the restrictions are more severe.  This is called after
1255  * we've verified that the uids do not match.
1256  */
1257 int
1258 secpolicy_proc_owner(const cred_t *scr, const cred_t *tcr, int mode)
1259 {
1260 	boolean_t allzone = B_FALSE;
1261 
1262 	if ((mode & VWRITE) && scr->cr_uid != 0 &&
1263 	    (tcr->cr_uid == 0 || tcr->cr_ruid == 0 || tcr->cr_suid == 0))
1264 		allzone = B_TRUE;
1265 
1266 	return (PRIV_POLICY(scr, PRIV_PROC_OWNER, allzone, EPERM, NULL));
1267 }
1268 
1269 int
1270 secpolicy_proc_access(const cred_t *scr)
1271 {
1272 	return (PRIV_POLICY(scr, PRIV_PROC_OWNER, B_FALSE, EACCES, NULL));
1273 }
1274 
1275 int
1276 secpolicy_proc_excl_open(const cred_t *scr)
1277 {
1278 	return (PRIV_POLICY(scr, PRIV_PROC_OWNER, B_FALSE, EBUSY, NULL));
1279 }
1280 
1281 int
1282 secpolicy_proc_zone(const cred_t *scr)
1283 {
1284 	return (PRIV_POLICY(scr, PRIV_PROC_ZONE, B_FALSE, EPERM, NULL));
1285 }
1286 
1287 /*
1288  * Destroying the system
1289  */
1290 
1291 int
1292 secpolicy_kmdb(const cred_t *scr)
1293 {
1294 	return (PRIV_POLICY(scr, PRIV_ALL, B_FALSE, EPERM, NULL));
1295 }
1296 
1297 int
1298 secpolicy_error_inject(const cred_t *scr)
1299 {
1300 	return (PRIV_POLICY(scr, PRIV_ALL, B_FALSE, EPERM, NULL));
1301 }
1302 
1303 /*
1304  * Processor sets, cpu configuration, resource pools.
1305  */
1306 int
1307 secpolicy_pset(const cred_t *cr)
1308 {
1309 	return (PRIV_POLICY(cr, PRIV_SYS_RES_CONFIG, B_FALSE, EPERM, NULL));
1310 }
1311 
1312 int
1313 secpolicy_ponline(const cred_t *cr)
1314 {
1315 	return (PRIV_POLICY(cr, PRIV_SYS_RES_CONFIG, B_FALSE, EPERM, NULL));
1316 }
1317 
1318 int
1319 secpolicy_pool(const cred_t *cr)
1320 {
1321 	return (PRIV_POLICY(cr, PRIV_SYS_RES_CONFIG, B_FALSE, EPERM, NULL));
1322 }
1323 
1324 int
1325 secpolicy_blacklist(const cred_t *cr)
1326 {
1327 	return (PRIV_POLICY(cr, PRIV_SYS_RES_CONFIG, B_FALSE, EPERM, NULL));
1328 }
1329 
1330 /*
1331  * Catch all system configuration.
1332  */
1333 int
1334 secpolicy_sys_config(const cred_t *cr, boolean_t checkonly)
1335 {
1336 	if (checkonly) {
1337 		return (PRIV_POLICY_ONLY(cr, PRIV_SYS_CONFIG, B_FALSE) ? 0 :
1338 		    EPERM);
1339 	} else {
1340 		return (PRIV_POLICY(cr, PRIV_SYS_CONFIG, B_FALSE, EPERM, NULL));
1341 	}
1342 }
1343 
1344 /*
1345  * Zone administration (halt, reboot, etc.) from within zone.
1346  */
1347 int
1348 secpolicy_zone_admin(const cred_t *cr, boolean_t checkonly)
1349 {
1350 	if (checkonly) {
1351 		return (PRIV_POLICY_ONLY(cr, PRIV_SYS_ADMIN, B_FALSE) ? 0 :
1352 		    EPERM);
1353 	} else {
1354 		return (PRIV_POLICY(cr, PRIV_SYS_ADMIN, B_FALSE, EPERM,
1355 		    NULL));
1356 	}
1357 }
1358 
1359 /*
1360  * Zone configuration (create, halt, enter).
1361  */
1362 int
1363 secpolicy_zone_config(const cred_t *cr)
1364 {
1365 	/*
1366 	 * Require all privileges to avoid possibility of privilege
1367 	 * escalation.
1368 	 */
1369 	return (secpolicy_require_set(cr, PRIV_FULLSET, NULL));
1370 }
1371 
1372 /*
1373  * Various other system configuration calls
1374  */
1375 int
1376 secpolicy_coreadm(const cred_t *cr)
1377 {
1378 	return (PRIV_POLICY(cr, PRIV_SYS_ADMIN, B_FALSE, EPERM, NULL));
1379 }
1380 
1381 int
1382 secpolicy_systeminfo(const cred_t *cr)
1383 {
1384 	return (PRIV_POLICY(cr, PRIV_SYS_ADMIN, B_FALSE, EPERM, NULL));
1385 }
1386 
1387 int
1388 secpolicy_dispadm(const cred_t *cr)
1389 {
1390 	return (PRIV_POLICY(cr, PRIV_SYS_CONFIG, B_FALSE, EPERM, NULL));
1391 }
1392 
1393 int
1394 secpolicy_settime(const cred_t *cr)
1395 {
1396 	return (PRIV_POLICY(cr, PRIV_SYS_TIME, B_FALSE, EPERM, NULL));
1397 }
1398 
1399 /*
1400  * For realtime users: high resolution clock.
1401  */
1402 int
1403 secpolicy_clock_highres(const cred_t *cr)
1404 {
1405 	return (PRIV_POLICY(cr, PRIV_PROC_CLOCK_HIGHRES, B_FALSE, EPERM,
1406 	    NULL));
1407 }
1408 
1409 /*
1410  * drv_priv() is documented as callable from interrupt context, not that
1411  * anyone ever does, but still.  No debugging or auditing can be done when
1412  * it is called from interrupt context.
1413  * returns 0 on succes, EPERM on failure.
1414  */
1415 int
1416 drv_priv(cred_t *cr)
1417 {
1418 	return (PRIV_POLICY(cr, PRIV_SYS_DEVICES, B_FALSE, EPERM, NULL));
1419 }
1420 
1421 int
1422 secpolicy_sys_devices(const cred_t *cr)
1423 {
1424 	return (PRIV_POLICY(cr, PRIV_SYS_DEVICES, B_FALSE, EPERM, NULL));
1425 }
1426 
1427 int
1428 secpolicy_excl_open(const cred_t *cr)
1429 {
1430 	return (PRIV_POLICY(cr, PRIV_SYS_DEVICES, B_FALSE, EBUSY, NULL));
1431 }
1432 
1433 int
1434 secpolicy_rctlsys(const cred_t *cr, boolean_t is_zone_rctl)
1435 {
1436 	/* zone.* rctls can only be set from the global zone */
1437 	if (is_zone_rctl && priv_policy_global(cr) != 0)
1438 		return (EPERM);
1439 	return (PRIV_POLICY(cr, PRIV_SYS_RESOURCE, B_FALSE, EPERM, NULL));
1440 }
1441 
1442 int
1443 secpolicy_resource(const cred_t *cr)
1444 {
1445 	return (PRIV_POLICY(cr, PRIV_SYS_RESOURCE, B_FALSE, EPERM, NULL));
1446 }
1447 
1448 /*
1449  * Processes with a real uid of 0 escape any form of accounting, much
1450  * like before.
1451  */
1452 int
1453 secpolicy_newproc(const cred_t *cr)
1454 {
1455 	if (cr->cr_ruid == 0)
1456 		return (0);
1457 
1458 	return (PRIV_POLICY(cr, PRIV_SYS_RESOURCE, B_FALSE, EPERM, NULL));
1459 }
1460 
1461 /*
1462  * Networking
1463  */
1464 int
1465 secpolicy_net_rawaccess(const cred_t *cr)
1466 {
1467 	return (PRIV_POLICY(cr, PRIV_NET_RAWACCESS, B_FALSE, EACCES, NULL));
1468 }
1469 
1470 /*
1471  * Need this privilege for accessing the ICMP device
1472  */
1473 int
1474 secpolicy_net_icmpaccess(const cred_t *cr)
1475 {
1476 	return (PRIV_POLICY(cr, PRIV_NET_ICMPACCESS, B_FALSE, EACCES, NULL));
1477 }
1478 
1479 /*
1480  * There are a few rare cases where the kernel generates ioctls() from
1481  * interrupt context with a credential of kcred rather than NULL.
1482  * In those cases, we take the safe and cheap test.
1483  */
1484 int
1485 secpolicy_net_config(const cred_t *cr, boolean_t checkonly)
1486 {
1487 	if (checkonly) {
1488 		return (PRIV_POLICY_ONLY(cr, PRIV_SYS_NET_CONFIG, B_FALSE) ?
1489 		    0 : EPERM);
1490 	} else {
1491 		return (PRIV_POLICY(cr, PRIV_SYS_NET_CONFIG, B_FALSE, EPERM,
1492 		    NULL));
1493 	}
1494 }
1495 
1496 
1497 /*
1498  * Map network pseudo privileges to actual privileges.
1499  * So we don't need to recompile IP when we change the privileges.
1500  */
1501 int
1502 secpolicy_net(const cred_t *cr, int netpriv, boolean_t checkonly)
1503 {
1504 	int priv = PRIV_ALL;
1505 
1506 	switch (netpriv) {
1507 	case OP_CONFIG:
1508 		priv = PRIV_SYS_NET_CONFIG;
1509 		break;
1510 	case OP_RAW:
1511 		priv = PRIV_NET_RAWACCESS;
1512 		break;
1513 	case OP_PRIVPORT:
1514 		priv = PRIV_NET_PRIVADDR;
1515 		break;
1516 	}
1517 	ASSERT(priv != PRIV_ALL);
1518 	if (checkonly)
1519 		return (PRIV_POLICY_ONLY(cr, priv, B_FALSE) ? 0 : EPERM);
1520 	else
1521 		return (PRIV_POLICY(cr, priv, B_FALSE, EPERM, NULL));
1522 }
1523 
1524 /*
1525  * Checks for operations that are either client-only or are used by
1526  * both clients and servers.
1527  */
1528 int
1529 secpolicy_nfs(const cred_t *cr)
1530 {
1531 	return (PRIV_POLICY(cr, PRIV_SYS_NFS, B_FALSE, EPERM, NULL));
1532 }
1533 
1534 /*
1535  * Special case for opening rpcmod: have NFS privileges or network
1536  * config privileges.
1537  */
1538 int
1539 secpolicy_rpcmod_open(const cred_t *cr)
1540 {
1541 	if (PRIV_POLICY_ONLY(cr, PRIV_SYS_NFS, B_FALSE))
1542 		return (secpolicy_nfs(cr));
1543 	else
1544 		return (secpolicy_net_config(cr, NULL));
1545 }
1546 
1547 int
1548 secpolicy_chroot(const cred_t *cr)
1549 {
1550 	return (PRIV_POLICY(cr, PRIV_PROC_CHROOT, B_FALSE, EPERM, NULL));
1551 }
1552 
1553 int
1554 secpolicy_tasksys(const cred_t *cr)
1555 {
1556 	return (PRIV_POLICY(cr, PRIV_PROC_TASKID, B_FALSE, EPERM, NULL));
1557 }
1558 
1559 /*
1560  * Basic privilege checks.
1561  */
1562 int
1563 secpolicy_basic_exec(const cred_t *cr)
1564 {
1565 	return (PRIV_POLICY(cr, PRIV_PROC_EXEC, B_FALSE, EPERM, NULL));
1566 }
1567 
1568 int
1569 secpolicy_basic_fork(const cred_t *cr)
1570 {
1571 	return (PRIV_POLICY(cr, PRIV_PROC_FORK, B_FALSE, EPERM, NULL));
1572 }
1573 
1574 int
1575 secpolicy_basic_proc(const cred_t *cr)
1576 {
1577 	return (PRIV_POLICY(cr, PRIV_PROC_SESSION, B_FALSE, EPERM, NULL));
1578 }
1579 
1580 /*
1581  * Slightly complicated because we don't want to trigger the policy too
1582  * often.  First we shortcircuit access to "self" (tp == sp) or if
1583  * we don't have the privilege but if we have permission
1584  * just return (0) and we don't flag the privilege as needed.
1585  * Else, we test for the privilege because we either have it or need it.
1586  */
1587 int
1588 secpolicy_basic_procinfo(const cred_t *cr, proc_t *tp, proc_t *sp)
1589 {
1590 	if (tp == sp ||
1591 	    !HAS_PRIVILEGE(cr, PRIV_PROC_INFO) && prochasprocperm(tp, sp, cr)) {
1592 		return (0);
1593 	} else {
1594 		return (PRIV_POLICY(cr, PRIV_PROC_INFO, B_FALSE, EPERM, NULL));
1595 	}
1596 }
1597 
1598 int
1599 secpolicy_basic_link(const cred_t *cr)
1600 {
1601 	return (PRIV_POLICY(cr, PRIV_FILE_LINK_ANY, B_FALSE, EPERM, NULL));
1602 }
1603 
1604 /*
1605  * Additional device protection.
1606  *
1607  * Traditionally, a device has specific permissions on the node in
1608  * the filesystem which govern which devices can be opened by what
1609  * processes.  In certain cases, it is desirable to add extra
1610  * restrictions, as writing to certain devices is identical to
1611  * having a complete run of the system.
1612  *
1613  * This mechanism is called the device policy.
1614  *
1615  * When a device is opened, its policy entry is looked up in the
1616  * policy cache and checked.
1617  */
1618 int
1619 secpolicy_spec_open(const cred_t *cr, struct vnode *vp, int oflag)
1620 {
1621 	devplcy_t *plcy;
1622 	int err;
1623 	struct snode *csp = VTOS(common_specvp(vp));
1624 
1625 	mutex_enter(&csp->s_lock);
1626 
1627 	if (csp->s_plcy == NULL || csp->s_plcy->dp_gen != devplcy_gen) {
1628 		plcy = devpolicy_find(vp);
1629 		if (csp->s_plcy)
1630 			dpfree(csp->s_plcy);
1631 		csp->s_plcy = plcy;
1632 		ASSERT(plcy != NULL);
1633 	} else
1634 		plcy = csp->s_plcy;
1635 
1636 	if (plcy == nullpolicy) {
1637 		mutex_exit(&csp->s_lock);
1638 		return (0);
1639 	}
1640 
1641 	dphold(plcy);
1642 
1643 	mutex_exit(&csp->s_lock);
1644 
1645 	err = secpolicy_require_set(cr,
1646 	    (oflag & FWRITE) ? &plcy->dp_wrp : &plcy->dp_rdp, "devpolicy");
1647 	dpfree(plcy);
1648 
1649 	return (err);
1650 }
1651 
1652 int
1653 secpolicy_modctl(const cred_t *cr, int cmd)
1654 {
1655 	switch (cmd) {
1656 	case MODINFO:
1657 	case MODGETPATH:
1658 	case MODGETPATHLEN:
1659 	case MODGETFBNAME:
1660 	case MODGETNAME:
1661 	case MODGETDEVPOLICY:
1662 	case MODGETDEVPOLICYBYNAME:
1663 	case MODGETMAJBIND:
1664 		/* Unprivileged */
1665 		return (0);
1666 	case MODLOAD:
1667 	case MODSETDEVPOLICY:
1668 		return (secpolicy_require_set(cr, PRIV_FULLSET, NULL));
1669 	default:
1670 		return (secpolicy_sys_config(cr, B_FALSE));
1671 	}
1672 }
1673 
1674 int
1675 secpolicy_console(const cred_t *cr)
1676 {
1677 	return (PRIV_POLICY(cr, PRIV_SYS_DEVICES, B_FALSE, EPERM, NULL));
1678 }
1679 
1680 int
1681 secpolicy_power_mgmt(const cred_t *cr)
1682 {
1683 	return (PRIV_POLICY(cr, PRIV_SYS_DEVICES, B_FALSE, EPERM, NULL));
1684 }
1685 
1686 /*
1687  * Simulate terminal input; another escalation of privileges avenue.
1688  */
1689 
1690 int
1691 secpolicy_sti(const cred_t *cr)
1692 {
1693 	return (secpolicy_require_set(cr, PRIV_FULLSET, NULL));
1694 }
1695 
1696 boolean_t
1697 secpolicy_net_reply_equal(const cred_t *cr)
1698 {
1699 	return (PRIV_POLICY(cr, PRIV_SYS_CONFIG, B_FALSE, EPERM, NULL));
1700 }
1701 
1702 int
1703 secpolicy_swapctl(const cred_t *cr)
1704 {
1705 	return (PRIV_POLICY(cr, PRIV_SYS_CONFIG, B_FALSE, EPERM, NULL));
1706 }
1707 
1708 int
1709 secpolicy_cpc_cpu(const cred_t *cr)
1710 {
1711 	return (PRIV_POLICY(cr, PRIV_CPC_CPU, B_FALSE, EACCES, NULL));
1712 }
1713 
1714 /*
1715  * secpolicy_contract_observer
1716  *
1717  * Determine if the subject may observe a specific contract's events.
1718  */
1719 int
1720 secpolicy_contract_observer(const cred_t *cr, struct contract *ct)
1721 {
1722 	if (contract_owned(ct, cr, B_FALSE))
1723 		return (0);
1724 	return (PRIV_POLICY(cr, PRIV_CONTRACT_OBSERVER, B_FALSE, EPERM, NULL));
1725 }
1726 
1727 /*
1728  * secpolicy_contract_observer_choice
1729  *
1730  * Determine if the subject may observe any contract's events.  Just
1731  * tests privilege and audits on success.
1732  */
1733 boolean_t
1734 secpolicy_contract_observer_choice(const cred_t *cr)
1735 {
1736 	return (PRIV_POLICY_CHOICE(cr, PRIV_CONTRACT_OBSERVER, B_FALSE));
1737 }
1738 
1739 /*
1740  * secpolicy_contract_event
1741  *
1742  * Determine if the subject may request critical contract events or
1743  * reliable contract event delivery.
1744  */
1745 int
1746 secpolicy_contract_event(const cred_t *cr)
1747 {
1748 	return (PRIV_POLICY(cr, PRIV_CONTRACT_EVENT, B_FALSE, EPERM, NULL));
1749 }
1750 
1751 /*
1752  * secpolicy_contract_event_choice
1753  *
1754  * Determine if the subject may retain contract events in its critical
1755  * set when a change in other terms would normally require a change in
1756  * the critical set.  Just tests privilege and audits on success.
1757  */
1758 boolean_t
1759 secpolicy_contract_event_choice(const cred_t *cr)
1760 {
1761 	return (PRIV_POLICY_CHOICE(cr, PRIV_CONTRACT_EVENT, B_FALSE));
1762 }
1763 
1764 /*
1765  * secpolicy_gart_access
1766  *
1767  * Determine if the subject has sufficient priveleges to make ioctls to agpgart
1768  * device.
1769  */
1770 int
1771 secpolicy_gart_access(const cred_t *cr)
1772 {
1773 	return (PRIV_POLICY(cr, PRIV_GRAPHICS_ACCESS, B_FALSE, EPERM, NULL));
1774 }
1775 
1776 /*
1777  * secpolicy_gart_map
1778  *
1779  * Determine if the subject has sufficient priveleges to map aperture range
1780  * through agpgart driver.
1781  */
1782 int
1783 secpolicy_gart_map(const cred_t *cr)
1784 {
1785 	if (PRIV_POLICY_ONLY(cr, PRIV_GRAPHICS_ACCESS, B_FALSE)) {
1786 		return (PRIV_POLICY(cr, PRIV_GRAPHICS_ACCESS, B_FALSE, EPERM,
1787 		    NULL));
1788 	} else {
1789 		return (PRIV_POLICY(cr, PRIV_GRAPHICS_MAP, B_FALSE, EPERM,
1790 		    NULL));
1791 	}
1792 }
1793 
1794 /*
1795  * secpolicy_zinject
1796  *
1797  * Determine if the subject can inject faults in the ZFS fault injection
1798  * framework.  Requires all privileges.
1799  */
1800 int
1801 secpolicy_zinject(const cred_t *cr)
1802 {
1803 	return (secpolicy_require_set(cr, PRIV_FULLSET, NULL));
1804 }
1805 
1806 /*
1807  * secpolicy_zfs
1808  *
1809  * Determine if the subject has permission to manipulate ZFS datasets
1810  * (not pools).  Equivalent to the SYS_MOUNT privilege.
1811  */
1812 int
1813 secpolicy_zfs(const cred_t *cr)
1814 {
1815 	return (PRIV_POLICY(cr, PRIV_SYS_MOUNT, B_FALSE, EPERM, NULL));
1816 }
1817