xref: /titanic_41/usr/src/uts/common/os/policy.c (revision 8b0ef7ed9d02af20a3d5106d84a4fc6aa0d7dd9f)
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 (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
23  */
24 
25 #include <sys/types.h>
26 #include <sys/sysmacros.h>
27 #include <sys/param.h>
28 #include <sys/systm.h>
29 #include <sys/cred_impl.h>
30 #include <sys/vnode.h>
31 #include <sys/vfs.h>
32 #include <sys/stat.h>
33 #include <sys/errno.h>
34 #include <sys/kmem.h>
35 #include <sys/user.h>
36 #include <sys/proc.h>
37 #include <sys/acct.h>
38 #include <sys/ipc_impl.h>
39 #include <sys/cmn_err.h>
40 #include <sys/debug.h>
41 #include <sys/policy.h>
42 #include <sys/kobj.h>
43 #include <sys/msg.h>
44 #include <sys/devpolicy.h>
45 #include <c2/audit.h>
46 #include <sys/varargs.h>
47 #include <sys/klpd.h>
48 #include <sys/modctl.h>
49 #include <sys/disp.h>
50 #include <sys/zone.h>
51 #include <inet/optcom.h>
52 #include <sys/sdt.h>
53 #include <sys/vfs.h>
54 #include <sys/mntent.h>
55 #include <sys/contract_impl.h>
56 #include <sys/dld_ioc.h>
57 
58 /*
59  * There are two possible layers of privilege routines and two possible
60  * levels of secpolicy.  Plus one other we may not be interested in, so
61  * we may need as many as 6 but no more.
62  */
63 #define	MAXPRIVSTACK		6
64 
65 int priv_debug = 0;
66 int priv_basic_test = -1;
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 assumptions 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 #define	FAST_BASIC_CHECK(cr, priv)	\
150 	if (PRIV_ISASSERT(&CR_OEPRIV(cr), priv)) { \
151 		DTRACE_PROBE2(priv__ok, int, priv, boolean_t, B_FALSE); \
152 		return (0); \
153 	}
154 
155 /*
156  * Policy checking functions.
157  *
158  * All of the system's policy should be implemented here.
159  */
160 
161 /*
162  * Private functions which take an additional va_list argument to
163  * implement an object specific policy override.
164  */
165 static int priv_policy_ap(const cred_t *, int, boolean_t, int,
166     const char *, va_list);
167 static int priv_policy_va(const cred_t *, int, boolean_t, int,
168     const char *, ...);
169 
170 /*
171  * Generic policy calls
172  *
173  * The "bottom" functions of policy control
174  */
175 static char *
176 mprintf(const char *fmt, ...)
177 {
178 	va_list args;
179 	char *buf;
180 	size_t len;
181 
182 	va_start(args, fmt);
183 	len = vsnprintf(NULL, 0, fmt, args) + 1;
184 	va_end(args);
185 
186 	buf = kmem_alloc(len, KM_NOSLEEP);
187 
188 	if (buf == NULL)
189 		return (NULL);
190 
191 	va_start(args, fmt);
192 	(void) vsnprintf(buf, len, fmt, args);
193 	va_end(args);
194 
195 	return (buf);
196 }
197 
198 /*
199  * priv_policy_errmsg()
200  *
201  * Generate an error message if privilege debugging is enabled system wide
202  * or for this particular process.
203  */
204 
205 #define	FMTHDR	"%s[%d]: missing privilege \"%s\" (euid = %d, syscall = %d)"
206 #define	FMTMSG	" for \"%s\""
207 #define	FMTFUN	" needed at %s+0x%lx"
208 
209 /* The maximum size privilege format: the concatenation of the above */
210 #define	FMTMAX	FMTHDR FMTMSG FMTFUN "\n"
211 
212 static void
213 priv_policy_errmsg(const cred_t *cr, int priv, const char *msg)
214 {
215 	struct proc *me;
216 	pc_t stack[MAXPRIVSTACK];
217 	int depth;
218 	int i;
219 	char *sym;
220 	ulong_t off;
221 	const char *pname;
222 
223 	char *cmd;
224 	char fmt[sizeof (FMTMAX)];
225 
226 	if ((me = curproc) == &p0)
227 		return;
228 
229 	/* Privileges must be defined  */
230 	ASSERT(priv == PRIV_ALL || priv == PRIV_MULTIPLE ||
231 	    priv == PRIV_ALLZONE || priv == PRIV_GLOBAL ||
232 	    priv_getbynum(priv) != NULL);
233 
234 	if (priv == PRIV_ALLZONE && INGLOBALZONE(me))
235 		priv = PRIV_ALL;
236 
237 	if (curthread->t_pre_sys)
238 		ttolwp(curthread)->lwp_badpriv = (short)priv;
239 
240 	if (priv_debug == 0 && (CR_FLAGS(cr) & PRIV_DEBUG) == 0)
241 		return;
242 
243 	(void) strcpy(fmt, FMTHDR);
244 
245 	if (me->p_user.u_comm[0])
246 		cmd = &me->p_user.u_comm[0];
247 	else
248 		cmd = "priv_policy";
249 
250 	if (msg != NULL && *msg != '\0') {
251 		(void) strcat(fmt, FMTMSG);
252 	} else {
253 		(void) strcat(fmt, "%s");
254 		msg = "";
255 	}
256 
257 	sym = NULL;
258 
259 	depth = getpcstack(stack, MAXPRIVSTACK);
260 
261 	/*
262 	 * Try to find the first interesting function on the stack.
263 	 * priv_policy* that's us, so completely uninteresting.
264 	 * suser(), drv_priv(), secpolicy_* are also called from
265 	 * too many locations to convey useful information.
266 	 */
267 	for (i = 0; i < depth; i++) {
268 		sym = kobj_getsymname((uintptr_t)stack[i], &off);
269 		if (sym != NULL &&
270 		    strstr(sym, "hasprocperm") == 0 &&
271 		    strcmp("suser", sym) != 0 &&
272 		    strcmp("ipcaccess", sym) != 0 &&
273 		    strcmp("drv_priv", sym) != 0 &&
274 		    strncmp("secpolicy_", sym, 10) != 0 &&
275 		    strncmp("priv_policy", sym, 11) != 0)
276 			break;
277 	}
278 
279 	if (sym != NULL)
280 		(void) strcat(fmt, FMTFUN);
281 
282 	(void) strcat(fmt, "\n");
283 
284 	switch (priv) {
285 	case PRIV_ALL:
286 		pname = "ALL";
287 		break;
288 	case PRIV_MULTIPLE:
289 		pname = "MULTIPLE";
290 		break;
291 	case PRIV_ALLZONE:
292 		pname = "ZONE";
293 		break;
294 	case PRIV_GLOBAL:
295 		pname = "GLOBAL";
296 		break;
297 	default:
298 		pname = priv_getbynum(priv);
299 		break;
300 	}
301 
302 	if (CR_FLAGS(cr) & PRIV_DEBUG) {
303 		/* Remember last message, just like lwp_badpriv. */
304 		if (curthread->t_pdmsg != NULL) {
305 			kmem_free(curthread->t_pdmsg,
306 			    strlen(curthread->t_pdmsg) + 1);
307 		}
308 
309 		curthread->t_pdmsg = mprintf(fmt, cmd, me->p_pid, pname,
310 		    cr->cr_uid, curthread->t_sysnum, msg, sym, off);
311 
312 		curthread->t_post_sys = 1;
313 	}
314 	if (priv_debug) {
315 		cmn_err(CE_NOTE, fmt, cmd, me->p_pid, pname, cr->cr_uid,
316 		    curthread->t_sysnum, msg, sym, off);
317 	}
318 }
319 
320 /*
321  * Override the policy, if appropriate.  Return 0 if the external
322  * policy engine approves.
323  */
324 static int
325 priv_policy_override(const cred_t *cr, int priv, boolean_t allzone, va_list ap)
326 {
327 	priv_set_t set;
328 	int ret;
329 
330 	if (!(CR_FLAGS(cr) & PRIV_XPOLICY))
331 		return (-1);
332 
333 	if (priv == PRIV_ALL) {
334 		priv_fillset(&set);
335 	} else if (allzone) {
336 		set = *ZONEPRIVS(cr);
337 	} else {
338 		priv_emptyset(&set);
339 		priv_addset(&set, priv);
340 	}
341 	ret = klpd_call(cr, &set, ap);
342 	return (ret);
343 }
344 
345 static int
346 priv_policy_override_set(const cred_t *cr, const priv_set_t *req, va_list ap)
347 {
348 	if (CR_FLAGS(cr) & PRIV_PFEXEC)
349 		return (check_user_privs(cr, req));
350 	if (CR_FLAGS(cr) & PRIV_XPOLICY) {
351 		return (klpd_call(cr, req, ap));
352 	}
353 	return (-1);
354 }
355 
356 static int
357 priv_policy_override_set_va(const cred_t *cr, const priv_set_t *req, ...)
358 {
359 	va_list ap;
360 	int ret;
361 
362 	va_start(ap, req);
363 	ret = priv_policy_override_set(cr, req, ap);
364 	va_end(ap);
365 	return (ret);
366 }
367 
368 /*
369  * Audit failure, log error message.
370  */
371 static void
372 priv_policy_err(const cred_t *cr, int priv, boolean_t allzone, const char *msg)
373 {
374 
375 	if (AU_AUDITING())
376 		audit_priv(priv, allzone ? ZONEPRIVS(cr) : NULL, 0);
377 	DTRACE_PROBE2(priv__err, int, priv, boolean_t, allzone);
378 
379 	if (priv_debug || (CR_FLAGS(cr) & PRIV_DEBUG) ||
380 	    curthread->t_pre_sys) {
381 		if (allzone && !HAS_ALLZONEPRIVS(cr)) {
382 			priv_policy_errmsg(cr, PRIV_ALLZONE, msg);
383 		} else {
384 			ASSERT(!HAS_PRIVILEGE(cr, priv));
385 			priv_policy_errmsg(cr, priv, msg);
386 		}
387 	}
388 }
389 
390 /*
391  * priv_policy_ap()
392  * return 0 or error.
393  * See block comment above for a description of "priv" and "allzone" usage.
394  */
395 static int
396 priv_policy_ap(const cred_t *cr, int priv, boolean_t allzone, int err,
397     const char *msg, va_list ap)
398 {
399 	if ((HAS_PRIVILEGE(cr, priv) && (!allzone || HAS_ALLZONEPRIVS(cr))) ||
400 	    (!servicing_interrupt() &&
401 	    priv_policy_override(cr, priv, allzone, ap) == 0)) {
402 		if ((allzone || priv == PRIV_ALL ||
403 		    !PRIV_ISASSERT(priv_basic, priv)) &&
404 		    !servicing_interrupt()) {
405 			PTOU(curproc)->u_acflag |= ASU; /* Needed for SVVS */
406 			if (AU_AUDITING())
407 				audit_priv(priv,
408 				    allzone ? ZONEPRIVS(cr) : NULL, 1);
409 		}
410 		err = 0;
411 		DTRACE_PROBE2(priv__ok, int, priv, boolean_t, allzone);
412 	} else if (!servicing_interrupt()) {
413 		/* Failure audited in this procedure */
414 		priv_policy_err(cr, priv, allzone, msg);
415 	}
416 	return (err);
417 }
418 
419 int
420 priv_policy_va(const cred_t *cr, int priv, boolean_t allzone, int err,
421     const char *msg, ...)
422 {
423 	int ret;
424 	va_list ap;
425 
426 	va_start(ap, msg);
427 	ret = priv_policy_ap(cr, priv, allzone, err, msg, ap);
428 	va_end(ap);
429 
430 	return (ret);
431 }
432 
433 int
434 priv_policy(const cred_t *cr, int priv, boolean_t allzone, int err,
435     const char *msg)
436 {
437 	return (priv_policy_va(cr, priv, allzone, err, msg, KLPDARG_NONE));
438 }
439 
440 /*
441  * Return B_TRUE for sufficient privileges, B_FALSE for insufficient privileges.
442  */
443 boolean_t
444 priv_policy_choice(const cred_t *cr, int priv, boolean_t allzone)
445 {
446 	boolean_t res = HAS_PRIVILEGE(cr, priv) &&
447 	    (!allzone || HAS_ALLZONEPRIVS(cr));
448 
449 	/* Audit success only */
450 	if (res && AU_AUDITING() &&
451 	    (allzone || priv == PRIV_ALL || !PRIV_ISASSERT(priv_basic, priv)) &&
452 	    !servicing_interrupt()) {
453 		audit_priv(priv, allzone ? ZONEPRIVS(cr) : NULL, 1);
454 	}
455 	if (res) {
456 		DTRACE_PROBE2(priv__ok, int, priv, boolean_t, allzone);
457 	} else {
458 		DTRACE_PROBE2(priv__err, int, priv, boolean_t, allzone);
459 	}
460 	return (res);
461 }
462 
463 /*
464  * Non-auditing variant of priv_policy_choice().
465  */
466 boolean_t
467 priv_policy_only(const cred_t *cr, int priv, boolean_t allzone)
468 {
469 	boolean_t res = HAS_PRIVILEGE(cr, priv) &&
470 	    (!allzone || HAS_ALLZONEPRIVS(cr));
471 
472 	if (res) {
473 		DTRACE_PROBE2(priv__ok, int, priv, boolean_t, allzone);
474 	} else {
475 		DTRACE_PROBE2(priv__err, int, priv, boolean_t, allzone);
476 	}
477 	return (res);
478 }
479 
480 /*
481  * Check whether all privileges in the required set are present.
482  */
483 static int
484 secpolicy_require_set(const cred_t *cr, const priv_set_t *req,
485     const char *msg, ...)
486 {
487 	int priv;
488 	int pfound = -1;
489 	priv_set_t pset;
490 	va_list ap;
491 	int ret;
492 
493 	if (req == PRIV_FULLSET ? HAS_ALLPRIVS(cr) : priv_issubset(req,
494 	    &CR_OEPRIV(cr))) {
495 		return (0);
496 	}
497 
498 	va_start(ap, msg);
499 	ret = priv_policy_override_set(cr, req, ap);
500 	va_end(ap);
501 	if (ret == 0)
502 		return (0);
503 
504 	if (req == PRIV_FULLSET || priv_isfullset(req)) {
505 		priv_policy_err(cr, PRIV_ALL, B_FALSE, msg);
506 		return (EACCES);
507 	}
508 
509 	pset = CR_OEPRIV(cr);		/* present privileges */
510 	priv_inverse(&pset);		/* all non present privileges */
511 	priv_intersect(req, &pset);	/* the actual missing privs */
512 
513 	if (AU_AUDITING())
514 		audit_priv(PRIV_NONE, &pset, 0);
515 	/*
516 	 * Privilege debugging; special case "one privilege in set".
517 	 */
518 	if (priv_debug || (CR_FLAGS(cr) & PRIV_DEBUG) || curthread->t_pre_sys) {
519 		for (priv = 0; priv < nprivs; priv++) {
520 			if (priv_ismember(&pset, priv)) {
521 				if (pfound != -1) {
522 					/* Multiple missing privs */
523 					priv_policy_errmsg(cr, PRIV_MULTIPLE,
524 					    msg);
525 					return (EACCES);
526 				}
527 				pfound = priv;
528 			}
529 		}
530 		ASSERT(pfound != -1);
531 		/* Just the one missing privilege */
532 		priv_policy_errmsg(cr, pfound, msg);
533 	}
534 
535 	return (EACCES);
536 }
537 
538 /*
539  * Called when an operation requires that the caller be in the
540  * global zone, regardless of privilege.
541  */
542 static int
543 priv_policy_global(const cred_t *cr)
544 {
545 	if (crgetzoneid(cr) == GLOBAL_ZONEID)
546 		return (0);	/* success */
547 
548 	if (priv_debug || (CR_FLAGS(cr) & PRIV_DEBUG) ||
549 	    curthread->t_pre_sys) {
550 		priv_policy_errmsg(cr, PRIV_GLOBAL, NULL);
551 	}
552 	return (EPERM);
553 }
554 
555 /*
556  * Changing process priority
557  */
558 int
559 secpolicy_setpriority(const cred_t *cr)
560 {
561 	return (PRIV_POLICY(cr, PRIV_PROC_PRIOCNTL, B_FALSE, EPERM, NULL));
562 }
563 
564 /*
565  * Binding to a privileged port, port must be specified in host byte
566  * order.
567  * When adding a new privilege which allows binding to currently privileged
568  * ports, then you MUST also allow processes with PRIV_NET_PRIVADDR bind
569  * to these ports because of backward compatibility.
570  */
571 int
572 secpolicy_net_privaddr(const cred_t *cr, in_port_t port, int proto)
573 {
574 	char *reason;
575 	int priv;
576 
577 	switch (port) {
578 	case 137:
579 	case 138:
580 	case 139:
581 	case 445:
582 		/*
583 		 * NBT and SMB ports, these are normal privileged ports,
584 		 * allow bind only if the SYS_SMB or NET_PRIVADDR privilege
585 		 * is present.
586 		 * Try both, if neither is present return an error for
587 		 * priv SYS_SMB.
588 		 */
589 		if (PRIV_POLICY_ONLY(cr, PRIV_NET_PRIVADDR, B_FALSE))
590 			priv = PRIV_NET_PRIVADDR;
591 		else
592 			priv = PRIV_SYS_SMB;
593 		reason = "NBT or SMB port";
594 		break;
595 
596 	case 2049:
597 	case 4045:
598 		/*
599 		 * NFS ports, these are extra privileged ports, allow bind
600 		 * only if the SYS_NFS privilege is present.
601 		 */
602 		priv = PRIV_SYS_NFS;
603 		reason = "NFS port";
604 		break;
605 
606 	default:
607 		priv = PRIV_NET_PRIVADDR;
608 		reason = NULL;
609 		break;
610 
611 	}
612 
613 	return (priv_policy_va(cr, priv, B_FALSE, EACCES, reason,
614 	    KLPDARG_PORT, (int)proto, (int)port, KLPDARG_NOMORE));
615 }
616 
617 /*
618  * Binding to a multilevel port on a trusted (labeled) system.
619  */
620 int
621 secpolicy_net_bindmlp(const cred_t *cr)
622 {
623 	return (PRIV_POLICY(cr, PRIV_NET_BINDMLP, B_FALSE, EACCES, NULL));
624 }
625 
626 /*
627  * Allow a communication between a zone and an unlabeled host when their
628  * labels don't match.
629  */
630 int
631 secpolicy_net_mac_aware(const cred_t *cr)
632 {
633 	return (PRIV_POLICY(cr, PRIV_NET_MAC_AWARE, B_FALSE, EACCES, NULL));
634 }
635 
636 /*
637  * Allow a privileged process to transmit traffic without explicit labels
638  */
639 int
640 secpolicy_net_mac_implicit(const cred_t *cr)
641 {
642 	return (PRIV_POLICY(cr, PRIV_NET_MAC_IMPLICIT, B_FALSE, EACCES, NULL));
643 }
644 
645 /*
646  * Common routine which determines whether a given credential can
647  * act on a given mount.
648  * When called through mount, the parameter needoptcheck is a pointer
649  * to a boolean variable which will be set to either true or false,
650  * depending on whether the mount policy should change the mount options.
651  * In all other cases, needoptcheck should be a NULL pointer.
652  */
653 static int
654 secpolicy_fs_common(cred_t *cr, vnode_t *mvp, const vfs_t *vfsp,
655     boolean_t *needoptcheck)
656 {
657 	boolean_t allzone = B_FALSE;
658 	boolean_t mounting = needoptcheck != NULL;
659 
660 	/*
661 	 * Short circuit the following cases:
662 	 *	vfsp == NULL or mvp == NULL (pure privilege check)
663 	 *	have all privileges - no further checks required
664 	 *	and no mount options need to be set.
665 	 */
666 	if (vfsp == NULL || mvp == NULL || HAS_ALLPRIVS(cr)) {
667 		if (mounting)
668 			*needoptcheck = B_FALSE;
669 
670 		return (priv_policy_va(cr, PRIV_SYS_MOUNT, allzone, EPERM,
671 		    NULL, KLPDARG_VNODE, mvp, (char *)NULL, KLPDARG_NOMORE));
672 	}
673 
674 	/*
675 	 * When operating on an existing mount (either we're not mounting
676 	 * or we're doing a remount and VFS_REMOUNT will be set), zones
677 	 * can operate only on mounts established by the zone itself.
678 	 */
679 	if (!mounting || (vfsp->vfs_flag & VFS_REMOUNT) != 0) {
680 		zoneid_t zoneid = crgetzoneid(cr);
681 
682 		if (zoneid != GLOBAL_ZONEID &&
683 		    vfsp->vfs_zone->zone_id != zoneid) {
684 			return (EPERM);
685 		}
686 	}
687 
688 	if (mounting)
689 		*needoptcheck = B_TRUE;
690 
691 	/*
692 	 * Overlay mounts may hide important stuff; if you can't write to a
693 	 * mount point but would be able to mount on top of it, you can
694 	 * escalate your privileges.
695 	 * So we go about asking the same questions namefs does when it
696 	 * decides whether you can mount over a file or not but with the
697 	 * added restriction that you can only mount on top of a regular
698 	 * file or directory.
699 	 * If we have all the zone's privileges, we skip all other checks,
700 	 * or else we may actually get in trouble inside the automounter.
701 	 */
702 	if ((mvp->v_flag & VROOT) != 0 ||
703 	    (mvp->v_type != VDIR && mvp->v_type != VREG) ||
704 	    HAS_ALLZONEPRIVS(cr)) {
705 		allzone = B_TRUE;
706 	} else {
707 		vattr_t va;
708 		int err;
709 
710 		va.va_mask = AT_UID|AT_MODE;
711 		err = VOP_GETATTR(mvp, &va, 0, cr, NULL);
712 		if (err != 0)
713 			return (err);
714 
715 		if ((err = secpolicy_vnode_owner(cr, va.va_uid)) != 0)
716 			return (err);
717 
718 		if (secpolicy_vnode_access2(cr, mvp, va.va_uid, va.va_mode,
719 		    VWRITE) != 0) {
720 			return (EACCES);
721 		}
722 	}
723 	return (priv_policy_va(cr, PRIV_SYS_MOUNT, allzone, EPERM,
724 	    NULL, KLPDARG_VNODE, mvp, (char *)NULL, KLPDARG_NOMORE));
725 }
726 
727 void
728 secpolicy_fs_mount_clearopts(cred_t *cr, struct vfs *vfsp)
729 {
730 	boolean_t amsuper = HAS_ALLZONEPRIVS(cr);
731 
732 	/*
733 	 * check; if we don't have either "nosuid" or
734 	 * both "nosetuid" and "nodevices", then we add
735 	 * "nosuid"; this depends on how the current
736 	 * implementation works (it first checks nosuid).  In a
737 	 * zone, a user with all zone privileges can mount with
738 	 * "setuid" but never with "devices".
739 	 */
740 	if (!vfs_optionisset(vfsp, MNTOPT_NOSUID, NULL) &&
741 	    (!vfs_optionisset(vfsp, MNTOPT_NODEVICES, NULL) ||
742 	    !vfs_optionisset(vfsp, MNTOPT_NOSETUID, NULL))) {
743 		if (crgetzoneid(cr) == GLOBAL_ZONEID || !amsuper)
744 			vfs_setmntopt(vfsp, MNTOPT_NOSUID, NULL, 0);
745 		else
746 			vfs_setmntopt(vfsp, MNTOPT_NODEVICES, NULL, 0);
747 	}
748 	/*
749 	 * If we're not the local super user, we set the "restrict"
750 	 * option to indicate to automountd that this mount should
751 	 * be handled with care.
752 	 */
753 	if (!amsuper)
754 		vfs_setmntopt(vfsp, MNTOPT_RESTRICT, NULL, 0);
755 
756 }
757 
758 extern vnode_t *rootvp;
759 extern vfs_t *rootvfs;
760 
761 int
762 secpolicy_fs_mount(cred_t *cr, vnode_t *mvp, struct vfs *vfsp)
763 {
764 	boolean_t needoptchk;
765 	int error;
766 
767 	/*
768 	 * If it's a remount, get the underlying mount point,
769 	 * except for the root where we use the rootvp.
770 	 */
771 	if ((vfsp->vfs_flag & VFS_REMOUNT) != 0) {
772 		if (vfsp == rootvfs)
773 			mvp = rootvp;
774 		else
775 			mvp = vfsp->vfs_vnodecovered;
776 	}
777 
778 	error = secpolicy_fs_common(cr, mvp, vfsp, &needoptchk);
779 
780 	if (error == 0 && needoptchk) {
781 		secpolicy_fs_mount_clearopts(cr, vfsp);
782 	}
783 
784 	return (error);
785 }
786 
787 /*
788  * Does the policy computations for "ownership" of a mount;
789  * here ownership is defined as the ability to "mount"
790  * the filesystem originally.  The rootvfs doesn't cover any
791  * vnodes; we attribute its ownership to the rootvp.
792  */
793 static int
794 secpolicy_fs_owner(cred_t *cr, const struct vfs *vfsp)
795 {
796 	vnode_t *mvp;
797 
798 	if (vfsp == NULL)
799 		mvp = NULL;
800 	else if (vfsp == rootvfs)
801 		mvp = rootvp;
802 	else
803 		mvp = vfsp->vfs_vnodecovered;
804 
805 	return (secpolicy_fs_common(cr, mvp, vfsp, NULL));
806 }
807 
808 int
809 secpolicy_fs_unmount(cred_t *cr, struct vfs *vfsp)
810 {
811 	return (secpolicy_fs_owner(cr, vfsp));
812 }
813 
814 /*
815  * Quotas are a resource, but if one has the ability to mount a filesystem, he
816  * should be able to modify quotas on it.
817  */
818 int
819 secpolicy_fs_quota(const cred_t *cr, const vfs_t *vfsp)
820 {
821 	return (secpolicy_fs_owner((cred_t *)cr, vfsp));
822 }
823 
824 /*
825  * Exceeding minfree: also a per-mount resource constraint.
826  */
827 int
828 secpolicy_fs_minfree(const cred_t *cr, const vfs_t *vfsp)
829 {
830 	return (secpolicy_fs_owner((cred_t *)cr, vfsp));
831 }
832 
833 int
834 secpolicy_fs_config(const cred_t *cr, const vfs_t *vfsp)
835 {
836 	return (secpolicy_fs_owner((cred_t *)cr, vfsp));
837 }
838 
839 /* ARGSUSED */
840 int
841 secpolicy_fs_linkdir(const cred_t *cr, const vfs_t *vfsp)
842 {
843 	return (PRIV_POLICY(cr, PRIV_SYS_LINKDIR, B_FALSE, EPERM, NULL));
844 }
845 
846 /*
847  * Name:        secpolicy_vnode_access()
848  *
849  * Parameters:  Process credential
850  *		vnode
851  *		uid of owner of vnode
852  *		permission bits not granted to the caller when examining
853  *		file mode bits (i.e., when a process wants to open a
854  *		mode 444 file for VREAD|VWRITE, this function should be
855  *		called only with a VWRITE argument).
856  *
857  * Normal:      Verifies that cred has the appropriate privileges to
858  *              override the mode bits that were denied.
859  *
860  * Override:    file_dac_execute - if VEXEC bit was denied and vnode is
861  *                      not a directory.
862  *              file_dac_read - if VREAD bit was denied.
863  *              file_dac_search - if VEXEC bit was denied and vnode is
864  *                      a directory.
865  *              file_dac_write - if VWRITE bit was denied.
866  *
867  *		Root owned files are special cased to protect system
868  *		configuration files and such.
869  *
870  * Output:      EACCES - if privilege check fails.
871  */
872 
873 int
874 secpolicy_vnode_access(const cred_t *cr, vnode_t *vp, uid_t owner, mode_t mode)
875 {
876 	if ((mode & VREAD) && priv_policy_va(cr, PRIV_FILE_DAC_READ, B_FALSE,
877 	    EACCES, NULL, KLPDARG_VNODE, vp, (char *)NULL,
878 	    KLPDARG_NOMORE) != 0) {
879 		return (EACCES);
880 	}
881 
882 	if (mode & VWRITE) {
883 		boolean_t allzone;
884 
885 		if (owner == 0 && cr->cr_uid != 0)
886 			allzone = B_TRUE;
887 		else
888 			allzone = B_FALSE;
889 		if (priv_policy_va(cr, PRIV_FILE_DAC_WRITE, allzone, EACCES,
890 		    NULL, KLPDARG_VNODE, vp, (char *)NULL,
891 		    KLPDARG_NOMORE) != 0) {
892 			return (EACCES);
893 		}
894 	}
895 
896 	if (mode & VEXEC) {
897 		/*
898 		 * Directories use file_dac_search to override the execute bit.
899 		 */
900 		int p = vp->v_type == VDIR ? PRIV_FILE_DAC_SEARCH :
901 		    PRIV_FILE_DAC_EXECUTE;
902 
903 		return (priv_policy_va(cr, p, B_FALSE, EACCES, NULL,
904 		    KLPDARG_VNODE, vp, (char *)NULL, KLPDARG_NOMORE));
905 	}
906 	return (0);
907 }
908 
909 /*
910  * Like secpolicy_vnode_access() but we get the actual wanted mode and the
911  * current mode of the file, not the missing bits.
912  */
913 int
914 secpolicy_vnode_access2(const cred_t *cr, vnode_t *vp, uid_t owner,
915     mode_t curmode, mode_t wantmode)
916 {
917 	mode_t mode;
918 
919 	/* Inline the basic privileges tests. */
920 	if ((wantmode & VREAD) &&
921 	    !PRIV_ISASSERT(&CR_OEPRIV(cr), PRIV_FILE_READ) &&
922 	    priv_policy_va(cr, PRIV_FILE_READ, B_FALSE, EACCES, NULL,
923 	    KLPDARG_VNODE, vp, (char *)NULL, KLPDARG_NOMORE) != 0) {
924 		return (EACCES);
925 	}
926 
927 	if ((wantmode & VWRITE) &&
928 	    !PRIV_ISASSERT(&CR_OEPRIV(cr), PRIV_FILE_WRITE) &&
929 	    priv_policy_va(cr, PRIV_FILE_WRITE, B_FALSE, EACCES, NULL,
930 	    KLPDARG_VNODE, vp, (char *)NULL, KLPDARG_NOMORE) != 0) {
931 		return (EACCES);
932 	}
933 
934 	mode = ~curmode & wantmode;
935 
936 	if (mode == 0)
937 		return (0);
938 
939 	if ((mode & VREAD) && priv_policy_va(cr, PRIV_FILE_DAC_READ, B_FALSE,
940 	    EACCES, NULL, KLPDARG_VNODE, vp, (char *)NULL,
941 	    KLPDARG_NOMORE) != 0) {
942 		return (EACCES);
943 	}
944 
945 	if (mode & VWRITE) {
946 		boolean_t allzone;
947 
948 		if (owner == 0 && cr->cr_uid != 0)
949 			allzone = B_TRUE;
950 		else
951 			allzone = B_FALSE;
952 		if (priv_policy_va(cr, PRIV_FILE_DAC_WRITE, allzone, EACCES,
953 		    NULL, KLPDARG_VNODE, vp, (char *)NULL,
954 		    KLPDARG_NOMORE) != 0) {
955 			return (EACCES);
956 		}
957 	}
958 
959 	if (mode & VEXEC) {
960 		/*
961 		 * Directories use file_dac_search to override the execute bit.
962 		 */
963 		int p = vp->v_type == VDIR ? PRIV_FILE_DAC_SEARCH :
964 		    PRIV_FILE_DAC_EXECUTE;
965 
966 		return (priv_policy_va(cr, p, B_FALSE, EACCES, NULL,
967 		    KLPDARG_VNODE, vp, (char *)NULL, KLPDARG_NOMORE));
968 	}
969 	return (0);
970 }
971 
972 /*
973  * This is a special routine for ZFS; it is used to determine whether
974  * any of the privileges in effect allow any form of access to the
975  * file.  There's no reason to audit this or any reason to record
976  * this.  More work is needed to do the "KPLD" stuff.
977  */
978 int
979 secpolicy_vnode_any_access(const cred_t *cr, vnode_t *vp, uid_t owner)
980 {
981 	static int privs[] = {
982 	    PRIV_FILE_OWNER,
983 	    PRIV_FILE_CHOWN,
984 	    PRIV_FILE_DAC_READ,
985 	    PRIV_FILE_DAC_WRITE,
986 	    PRIV_FILE_DAC_EXECUTE,
987 	    PRIV_FILE_DAC_SEARCH,
988 	};
989 	int i;
990 
991 	/* Same as secpolicy_vnode_setdac */
992 	if (owner == cr->cr_uid)
993 		return (0);
994 
995 	for (i = 0; i < sizeof (privs)/sizeof (int); i++) {
996 		boolean_t allzone = B_FALSE;
997 		int priv;
998 
999 		switch (priv = privs[i]) {
1000 		case PRIV_FILE_DAC_EXECUTE:
1001 			if (vp->v_type == VDIR)
1002 				continue;
1003 			break;
1004 		case PRIV_FILE_DAC_SEARCH:
1005 			if (vp->v_type != VDIR)
1006 				continue;
1007 			break;
1008 		case PRIV_FILE_DAC_WRITE:
1009 		case PRIV_FILE_OWNER:
1010 		case PRIV_FILE_CHOWN:
1011 			/* We know here that if owner == 0, that cr_uid != 0 */
1012 			allzone = owner == 0;
1013 			break;
1014 		}
1015 		if (PRIV_POLICY_CHOICE(cr, priv, allzone))
1016 			return (0);
1017 	}
1018 	return (EPERM);
1019 }
1020 
1021 /*
1022  * Name:	secpolicy_vnode_setid_modify()
1023  *
1024  * Normal:	verify that subject can set the file setid flags.
1025  *
1026  * Output:	EPERM - if not privileged.
1027  */
1028 
1029 static int
1030 secpolicy_vnode_setid_modify(const cred_t *cr, uid_t owner)
1031 {
1032 	/* If changing to suid root, must have all zone privs */
1033 	boolean_t allzone = B_TRUE;
1034 
1035 	if (owner != 0) {
1036 		if (owner == cr->cr_uid)
1037 			return (0);
1038 		allzone = B_FALSE;
1039 	}
1040 	return (PRIV_POLICY(cr, PRIV_FILE_SETID, allzone, EPERM, NULL));
1041 }
1042 
1043 /*
1044  * Are we allowed to retain the set-uid/set-gid bits when
1045  * changing ownership or when writing to a file?
1046  * "issuid" should be true when set-uid; only in that case
1047  * root ownership is checked (setgid is assumed).
1048  */
1049 int
1050 secpolicy_vnode_setid_retain(const cred_t *cred, boolean_t issuidroot)
1051 {
1052 	if (issuidroot && !HAS_ALLZONEPRIVS(cred))
1053 		return (EPERM);
1054 
1055 	return (!PRIV_POLICY_CHOICE(cred, PRIV_FILE_SETID, B_FALSE));
1056 }
1057 
1058 /*
1059  * Name:	secpolicy_vnode_setids_setgids()
1060  *
1061  * Normal:	verify that subject can set the file setgid flag.
1062  *
1063  * Output:	EPERM - if not privileged
1064  */
1065 
1066 int
1067 secpolicy_vnode_setids_setgids(const cred_t *cred, gid_t gid)
1068 {
1069 	if (!groupmember(gid, cred))
1070 		return (PRIV_POLICY(cred, PRIV_FILE_SETID, B_FALSE, EPERM,
1071 		    NULL));
1072 	return (0);
1073 }
1074 
1075 /*
1076  * Name:	secpolicy_vnode_chown
1077  *
1078  * Normal:	Determine if subject can chown owner of a file.
1079  *
1080  * Output:	EPERM - if access denied
1081  */
1082 
1083 int
1084 secpolicy_vnode_chown(const cred_t *cred, uid_t owner)
1085 {
1086 	boolean_t is_owner = (owner == crgetuid(cred));
1087 	boolean_t allzone = B_FALSE;
1088 	int priv;
1089 
1090 	if (!is_owner) {
1091 		allzone = (owner == 0);
1092 		priv = PRIV_FILE_CHOWN;
1093 	} else {
1094 		priv = HAS_PRIVILEGE(cred, PRIV_FILE_CHOWN) ?
1095 		    PRIV_FILE_CHOWN : PRIV_FILE_CHOWN_SELF;
1096 	}
1097 
1098 	return (PRIV_POLICY(cred, priv, allzone, EPERM, NULL));
1099 }
1100 
1101 /*
1102  * Name:	secpolicy_vnode_create_gid
1103  *
1104  * Normal:	Determine if subject can change group ownership of a file.
1105  *
1106  * Output:	EPERM - if access denied
1107  */
1108 int
1109 secpolicy_vnode_create_gid(const cred_t *cred)
1110 {
1111 	if (HAS_PRIVILEGE(cred, PRIV_FILE_CHOWN))
1112 		return (PRIV_POLICY(cred, PRIV_FILE_CHOWN, B_FALSE, EPERM,
1113 		    NULL));
1114 	else
1115 		return (PRIV_POLICY(cred, PRIV_FILE_CHOWN_SELF, B_FALSE, EPERM,
1116 		    NULL));
1117 }
1118 
1119 /*
1120  * Name:	secpolicy_vnode_utime_modify()
1121  *
1122  * Normal:	verify that subject can modify the utime on a file.
1123  *
1124  * Output:	EPERM - if access denied.
1125  */
1126 
1127 static int
1128 secpolicy_vnode_utime_modify(const cred_t *cred)
1129 {
1130 	return (PRIV_POLICY(cred, PRIV_FILE_OWNER, B_FALSE, EPERM,
1131 	    "modify file times"));
1132 }
1133 
1134 
1135 /*
1136  * Name:	secpolicy_vnode_setdac()
1137  *
1138  * Normal:	verify that subject can modify the mode of a file.
1139  *		allzone privilege needed when modifying root owned object.
1140  *
1141  * Output:	EPERM - if access denied.
1142  */
1143 
1144 int
1145 secpolicy_vnode_setdac(const cred_t *cred, uid_t owner)
1146 {
1147 	if (owner == cred->cr_uid)
1148 		return (0);
1149 
1150 	return (PRIV_POLICY(cred, PRIV_FILE_OWNER, owner == 0, EPERM, NULL));
1151 }
1152 /*
1153  * Name:	secpolicy_vnode_stky_modify()
1154  *
1155  * Normal:	verify that subject can make a file a "sticky".
1156  *
1157  * Output:	EPERM - if access denied.
1158  */
1159 
1160 int
1161 secpolicy_vnode_stky_modify(const cred_t *cred)
1162 {
1163 	return (PRIV_POLICY(cred, PRIV_SYS_CONFIG, B_FALSE, EPERM,
1164 	    "set file sticky"));
1165 }
1166 
1167 /*
1168  * Policy determines whether we can remove an entry from a directory,
1169  * regardless of permission bits.
1170  */
1171 int
1172 secpolicy_vnode_remove(const cred_t *cr)
1173 {
1174 	return (PRIV_POLICY(cr, PRIV_FILE_OWNER, B_FALSE, EACCES,
1175 	    "sticky directory"));
1176 }
1177 
1178 int
1179 secpolicy_vnode_owner(const cred_t *cr, uid_t owner)
1180 {
1181 	boolean_t allzone = (owner == 0);
1182 
1183 	if (owner == cr->cr_uid)
1184 		return (0);
1185 
1186 	return (PRIV_POLICY(cr, PRIV_FILE_OWNER, allzone, EPERM, NULL));
1187 }
1188 
1189 void
1190 secpolicy_setid_clear(vattr_t *vap, cred_t *cr)
1191 {
1192 	if ((vap->va_mode & (S_ISUID | S_ISGID)) != 0 &&
1193 	    secpolicy_vnode_setid_retain(cr,
1194 	    (vap->va_mode & S_ISUID) != 0 &&
1195 	    (vap->va_mask & AT_UID) != 0 && vap->va_uid == 0) != 0) {
1196 		vap->va_mask |= AT_MODE;
1197 		vap->va_mode &= ~(S_ISUID|S_ISGID);
1198 	}
1199 }
1200 
1201 int
1202 secpolicy_setid_setsticky_clear(vnode_t *vp, vattr_t *vap, const vattr_t *ovap,
1203     cred_t *cr)
1204 {
1205 	int error;
1206 
1207 	if ((vap->va_mode & S_ISUID) != 0 &&
1208 	    (error = secpolicy_vnode_setid_modify(cr,
1209 	    ovap->va_uid)) != 0) {
1210 		return (error);
1211 	}
1212 
1213 	/*
1214 	 * Check privilege if attempting to set the
1215 	 * sticky bit on a non-directory.
1216 	 */
1217 	if (vp->v_type != VDIR && (vap->va_mode & S_ISVTX) != 0 &&
1218 	    secpolicy_vnode_stky_modify(cr) != 0) {
1219 		vap->va_mode &= ~S_ISVTX;
1220 	}
1221 
1222 	/*
1223 	 * Check for privilege if attempting to set the
1224 	 * group-id bit.
1225 	 */
1226 	if ((vap->va_mode & S_ISGID) != 0 &&
1227 	    secpolicy_vnode_setids_setgids(cr, ovap->va_gid) != 0) {
1228 		vap->va_mode &= ~S_ISGID;
1229 	}
1230 
1231 	return (0);
1232 }
1233 
1234 #define	ATTR_FLAG_PRIV(attr, value, cr)	\
1235 	PRIV_POLICY(cr, value ? PRIV_FILE_FLAG_SET : PRIV_ALL, \
1236 	B_FALSE, EPERM, NULL)
1237 
1238 /*
1239  * Check privileges for setting xvattr attributes
1240  */
1241 int
1242 secpolicy_xvattr(xvattr_t *xvap, uid_t owner, cred_t *cr, vtype_t vtype)
1243 {
1244 	xoptattr_t *xoap;
1245 	int error = 0;
1246 
1247 	if ((xoap = xva_getxoptattr(xvap)) == NULL)
1248 		return (EINVAL);
1249 
1250 	/*
1251 	 * First process the DOS bits
1252 	 */
1253 	if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE) ||
1254 	    XVA_ISSET_REQ(xvap, XAT_HIDDEN) ||
1255 	    XVA_ISSET_REQ(xvap, XAT_READONLY) ||
1256 	    XVA_ISSET_REQ(xvap, XAT_SYSTEM) ||
1257 	    XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
1258 		if ((error = secpolicy_vnode_owner(cr, owner)) != 0)
1259 			return (error);
1260 	}
1261 
1262 	/*
1263 	 * Now handle special attributes
1264 	 */
1265 
1266 	if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE))
1267 		error = ATTR_FLAG_PRIV(XAT_IMMUTABLE,
1268 		    xoap->xoa_immutable, cr);
1269 	if (error == 0 && XVA_ISSET_REQ(xvap, XAT_NOUNLINK))
1270 		error = ATTR_FLAG_PRIV(XAT_NOUNLINK,
1271 		    xoap->xoa_nounlink, cr);
1272 	if (error == 0 && XVA_ISSET_REQ(xvap, XAT_APPENDONLY))
1273 		error = ATTR_FLAG_PRIV(XAT_APPENDONLY,
1274 		    xoap->xoa_appendonly, cr);
1275 	if (error == 0 && XVA_ISSET_REQ(xvap, XAT_NODUMP))
1276 		error = ATTR_FLAG_PRIV(XAT_NODUMP,
1277 		    xoap->xoa_nodump, cr);
1278 	if (error == 0 && XVA_ISSET_REQ(xvap, XAT_OPAQUE))
1279 		error = EPERM;
1280 	if (error == 0 && XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
1281 		error = ATTR_FLAG_PRIV(XAT_AV_QUARANTINED,
1282 		    xoap->xoa_av_quarantined, cr);
1283 		if (error == 0 && vtype != VREG && xoap->xoa_av_quarantined)
1284 			error = EINVAL;
1285 	}
1286 	if (error == 0 && XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED))
1287 		error = ATTR_FLAG_PRIV(XAT_AV_MODIFIED,
1288 		    xoap->xoa_av_modified, cr);
1289 	if (error == 0 && XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
1290 		error = ATTR_FLAG_PRIV(XAT_AV_SCANSTAMP,
1291 		    xoap->xoa_av_scanstamp, cr);
1292 		if (error == 0 && vtype != VREG)
1293 			error = EINVAL;
1294 	}
1295 	return (error);
1296 }
1297 
1298 /*
1299  * This function checks the policy decisions surrounding the
1300  * vop setattr call.
1301  *
1302  * It should be called after sufficient locks have been established
1303  * on the underlying data structures.  No concurrent modifications
1304  * should be allowed.
1305  *
1306  * The caller must pass in unlocked version of its vaccess function
1307  * this is required because vop_access function should lock the
1308  * node for reading.  A three argument function should be defined
1309  * which accepts the following argument:
1310  * 	A pointer to the internal "node" type (inode *)
1311  *	vnode access bits (VREAD|VWRITE|VEXEC)
1312  *	a pointer to the credential
1313  *
1314  * This function makes the following policy decisions:
1315  *
1316  *		- change permissions
1317  *			- permission to change file mode if not owner
1318  *			- permission to add sticky bit to non-directory
1319  *			- permission to add set-gid bit
1320  *
1321  * The ovap argument should include AT_MODE|AT_UID|AT_GID.
1322  *
1323  * If the vap argument does not include AT_MODE, the mode will be copied from
1324  * ovap.  In certain situations set-uid/set-gid bits need to be removed;
1325  * this is done by marking vap->va_mask to include AT_MODE and va_mode
1326  * is updated to the newly computed mode.
1327  */
1328 
1329 int
1330 secpolicy_vnode_setattr(cred_t *cr, struct vnode *vp, struct vattr *vap,
1331 	const struct vattr *ovap, int flags,
1332 	int unlocked_access(void *, int, cred_t *),
1333 	void *node)
1334 {
1335 	int mask = vap->va_mask;
1336 	int error = 0;
1337 	boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
1338 
1339 	if (mask & AT_SIZE) {
1340 		if (vp->v_type == VDIR) {
1341 			error = EISDIR;
1342 			goto out;
1343 		}
1344 
1345 		/*
1346 		 * If ATTR_NOACLCHECK is set in the flags, then we don't
1347 		 * perform the secondary unlocked_access() call since the
1348 		 * ACL (if any) is being checked there.
1349 		 */
1350 		if (skipaclchk == B_FALSE) {
1351 			error = unlocked_access(node, VWRITE, cr);
1352 			if (error)
1353 				goto out;
1354 		}
1355 	}
1356 	if (mask & AT_MODE) {
1357 		/*
1358 		 * If not the owner of the file then check privilege
1359 		 * for two things: the privilege to set the mode at all
1360 		 * and, if we're setting setuid, we also need permissions
1361 		 * to add the set-uid bit, if we're not the owner.
1362 		 * In the specific case of creating a set-uid root
1363 		 * file, we need even more permissions.
1364 		 */
1365 		if ((error = secpolicy_vnode_setdac(cr, ovap->va_uid)) != 0)
1366 			goto out;
1367 
1368 		if ((error = secpolicy_setid_setsticky_clear(vp, vap,
1369 		    ovap, cr)) != 0)
1370 			goto out;
1371 	} else
1372 		vap->va_mode = ovap->va_mode;
1373 
1374 	if (mask & (AT_UID|AT_GID)) {
1375 		boolean_t checkpriv = B_FALSE;
1376 
1377 		/*
1378 		 * Chowning files.
1379 		 *
1380 		 * If you are the file owner:
1381 		 *	chown to other uid		FILE_CHOWN_SELF
1382 		 *	chown to gid (non-member) 	FILE_CHOWN_SELF
1383 		 *	chown to gid (member) 		<none>
1384 		 *
1385 		 * Instead of PRIV_FILE_CHOWN_SELF, FILE_CHOWN is also
1386 		 * acceptable but the first one is reported when debugging.
1387 		 *
1388 		 * If you are not the file owner:
1389 		 *	chown from root			PRIV_FILE_CHOWN + zone
1390 		 *	chown from other to any		PRIV_FILE_CHOWN
1391 		 *
1392 		 */
1393 		if (cr->cr_uid != ovap->va_uid) {
1394 			checkpriv = B_TRUE;
1395 		} else {
1396 			if (((mask & AT_UID) && vap->va_uid != ovap->va_uid) ||
1397 			    ((mask & AT_GID) && vap->va_gid != ovap->va_gid &&
1398 			    !groupmember(vap->va_gid, cr))) {
1399 				checkpriv = B_TRUE;
1400 			}
1401 		}
1402 		/*
1403 		 * If necessary, check privilege to see if update can be done.
1404 		 */
1405 		if (checkpriv &&
1406 		    (error = secpolicy_vnode_chown(cr, ovap->va_uid)) != 0) {
1407 			goto out;
1408 		}
1409 
1410 		/*
1411 		 * If the file has either the set UID or set GID bits
1412 		 * set and the caller can set the bits, then leave them.
1413 		 */
1414 		secpolicy_setid_clear(vap, cr);
1415 	}
1416 	if (mask & (AT_ATIME|AT_MTIME)) {
1417 		/*
1418 		 * If not the file owner and not otherwise privileged,
1419 		 * always return an error when setting the
1420 		 * time other than the current (ATTR_UTIME flag set).
1421 		 * If setting the current time (ATTR_UTIME not set) then
1422 		 * unlocked_access will check permissions according to policy.
1423 		 */
1424 		if (cr->cr_uid != ovap->va_uid) {
1425 			if (flags & ATTR_UTIME)
1426 				error = secpolicy_vnode_utime_modify(cr);
1427 			else if (skipaclchk == B_FALSE) {
1428 				error = unlocked_access(node, VWRITE, cr);
1429 				if (error == EACCES &&
1430 				    secpolicy_vnode_utime_modify(cr) == 0)
1431 					error = 0;
1432 			}
1433 			if (error)
1434 				goto out;
1435 		}
1436 	}
1437 
1438 	/*
1439 	 * Check for optional attributes here by checking the following:
1440 	 */
1441 	if (mask & AT_XVATTR)
1442 		error = secpolicy_xvattr((xvattr_t *)vap, ovap->va_uid, cr,
1443 		    vp->v_type);
1444 out:
1445 	return (error);
1446 }
1447 
1448 /*
1449  * Name:	secpolicy_pcfs_modify_bootpartition()
1450  *
1451  * Normal:	verify that subject can modify a pcfs boot partition.
1452  *
1453  * Output:	EACCES - if privilege check failed.
1454  */
1455 /*ARGSUSED*/
1456 int
1457 secpolicy_pcfs_modify_bootpartition(const cred_t *cred)
1458 {
1459 	return (PRIV_POLICY(cred, PRIV_ALL, B_FALSE, EACCES,
1460 	    "modify pcfs boot partition"));
1461 }
1462 
1463 /*
1464  * System V IPC routines
1465  */
1466 int
1467 secpolicy_ipc_owner(const cred_t *cr, const struct kipc_perm *ip)
1468 {
1469 	if (crgetzoneid(cr) != ip->ipc_zoneid ||
1470 	    (cr->cr_uid != ip->ipc_uid && cr->cr_uid != ip->ipc_cuid)) {
1471 		boolean_t allzone = B_FALSE;
1472 		if (ip->ipc_uid == 0 || ip->ipc_cuid == 0)
1473 			allzone = B_TRUE;
1474 		return (PRIV_POLICY(cr, PRIV_IPC_OWNER, allzone, EPERM, NULL));
1475 	}
1476 	return (0);
1477 }
1478 
1479 int
1480 secpolicy_ipc_config(const cred_t *cr)
1481 {
1482 	return (PRIV_POLICY(cr, PRIV_SYS_IPC_CONFIG, B_FALSE, EPERM, NULL));
1483 }
1484 
1485 int
1486 secpolicy_ipc_access(const cred_t *cr, const struct kipc_perm *ip, mode_t mode)
1487 {
1488 
1489 	boolean_t allzone = B_FALSE;
1490 
1491 	ASSERT((mode & (MSG_R|MSG_W)) != 0);
1492 
1493 	if ((mode & MSG_R) &&
1494 	    PRIV_POLICY(cr, PRIV_IPC_DAC_READ, allzone, EACCES, NULL) != 0)
1495 		return (EACCES);
1496 
1497 	if (mode & MSG_W) {
1498 		if (cr->cr_uid != 0 && (ip->ipc_uid == 0 || ip->ipc_cuid == 0))
1499 			allzone = B_TRUE;
1500 
1501 		return (PRIV_POLICY(cr, PRIV_IPC_DAC_WRITE, allzone, EACCES,
1502 		    NULL));
1503 	}
1504 	return (0);
1505 }
1506 
1507 int
1508 secpolicy_rsm_access(const cred_t *cr, uid_t owner, mode_t mode)
1509 {
1510 	boolean_t allzone = B_FALSE;
1511 
1512 	ASSERT((mode & (MSG_R|MSG_W)) != 0);
1513 
1514 	if ((mode & MSG_R) &&
1515 	    PRIV_POLICY(cr, PRIV_IPC_DAC_READ, allzone, EACCES, NULL) != 0)
1516 		return (EACCES);
1517 
1518 	if (mode & MSG_W) {
1519 		if (cr->cr_uid != 0 && owner == 0)
1520 			allzone = B_TRUE;
1521 
1522 		return (PRIV_POLICY(cr, PRIV_IPC_DAC_WRITE, allzone, EACCES,
1523 		    NULL));
1524 	}
1525 	return (0);
1526 }
1527 
1528 /*
1529  * Audit configuration.
1530  */
1531 int
1532 secpolicy_audit_config(const cred_t *cr)
1533 {
1534 	return (PRIV_POLICY(cr, PRIV_SYS_AUDIT, B_FALSE, EPERM, NULL));
1535 }
1536 
1537 /*
1538  * Audit record generation.
1539  */
1540 int
1541 secpolicy_audit_modify(const cred_t *cr)
1542 {
1543 	return (PRIV_POLICY(cr, PRIV_PROC_AUDIT, B_FALSE, EPERM, NULL));
1544 }
1545 
1546 /*
1547  * Get audit attributes.
1548  * Either PRIV_SYS_AUDIT or PRIV_PROC_AUDIT required; report the
1549  * "Least" of the two privileges on error.
1550  */
1551 int
1552 secpolicy_audit_getattr(const cred_t *cr, boolean_t checkonly)
1553 {
1554 	int priv;
1555 
1556 	if (PRIV_POLICY_ONLY(cr, PRIV_SYS_AUDIT, B_FALSE))
1557 		priv = PRIV_SYS_AUDIT;
1558 	else
1559 		priv = PRIV_PROC_AUDIT;
1560 
1561 	if (checkonly)
1562 		return (!PRIV_POLICY_ONLY(cr, priv, B_FALSE));
1563 	else
1564 		return (PRIV_POLICY(cr, priv, B_FALSE, EPERM, NULL));
1565 }
1566 
1567 
1568 /*
1569  * Locking physical memory
1570  */
1571 int
1572 secpolicy_lock_memory(const cred_t *cr)
1573 {
1574 	return (PRIV_POLICY(cr, PRIV_PROC_LOCK_MEMORY, B_FALSE, EPERM, NULL));
1575 }
1576 
1577 /*
1578  * Accounting (both acct(2) and exacct).
1579  */
1580 int
1581 secpolicy_acct(const cred_t *cr)
1582 {
1583 	return (PRIV_POLICY(cr, PRIV_SYS_ACCT, B_FALSE, EPERM, NULL));
1584 }
1585 
1586 /*
1587  * Is this process privileged to change its uids at will?
1588  * Uid 0 is still considered "special" and having the SETID
1589  * privilege is not sufficient to get uid 0.
1590  * Files are owned by root, so the privilege would give
1591  * full access and euid 0 is still effective.
1592  *
1593  * If you have the privilege and euid 0 only then do you
1594  * get the powers of root wrt uid 0.
1595  *
1596  * For gid manipulations, this is should be called with an
1597  * uid of -1.
1598  *
1599  */
1600 int
1601 secpolicy_allow_setid(const cred_t *cr, uid_t newuid, boolean_t checkonly)
1602 {
1603 	boolean_t allzone = B_FALSE;
1604 
1605 	if (newuid == 0 && cr->cr_uid != 0 && cr->cr_suid != 0 &&
1606 	    cr->cr_ruid != 0) {
1607 		allzone = B_TRUE;
1608 	}
1609 
1610 	return (checkonly ? !PRIV_POLICY_ONLY(cr, PRIV_PROC_SETID, allzone) :
1611 	    PRIV_POLICY(cr, PRIV_PROC_SETID, allzone, EPERM, NULL));
1612 }
1613 
1614 
1615 /*
1616  * Acting on a different process: if the mode is for writing,
1617  * the restrictions are more severe.  This is called after
1618  * we've verified that the uids do not match.
1619  */
1620 int
1621 secpolicy_proc_owner(const cred_t *scr, const cred_t *tcr, int mode)
1622 {
1623 	boolean_t allzone = B_FALSE;
1624 
1625 	if ((mode & VWRITE) && scr->cr_uid != 0 &&
1626 	    (tcr->cr_uid == 0 || tcr->cr_ruid == 0 || tcr->cr_suid == 0))
1627 		allzone = B_TRUE;
1628 
1629 	return (PRIV_POLICY(scr, PRIV_PROC_OWNER, allzone, EPERM, NULL));
1630 }
1631 
1632 int
1633 secpolicy_proc_access(const cred_t *scr)
1634 {
1635 	return (PRIV_POLICY(scr, PRIV_PROC_OWNER, B_FALSE, EACCES, NULL));
1636 }
1637 
1638 int
1639 secpolicy_proc_excl_open(const cred_t *scr)
1640 {
1641 	return (PRIV_POLICY(scr, PRIV_PROC_OWNER, B_FALSE, EBUSY, NULL));
1642 }
1643 
1644 int
1645 secpolicy_proc_zone(const cred_t *scr)
1646 {
1647 	return (PRIV_POLICY(scr, PRIV_PROC_ZONE, B_FALSE, EPERM, NULL));
1648 }
1649 
1650 /*
1651  * Destroying the system
1652  */
1653 
1654 int
1655 secpolicy_kmdb(const cred_t *scr)
1656 {
1657 	return (PRIV_POLICY(scr, PRIV_ALL, B_FALSE, EPERM, NULL));
1658 }
1659 
1660 int
1661 secpolicy_error_inject(const cred_t *scr)
1662 {
1663 	return (PRIV_POLICY(scr, PRIV_ALL, B_FALSE, EPERM, NULL));
1664 }
1665 
1666 /*
1667  * Processor sets, cpu configuration, resource pools.
1668  */
1669 int
1670 secpolicy_pset(const cred_t *cr)
1671 {
1672 	return (PRIV_POLICY(cr, PRIV_SYS_RES_CONFIG, B_FALSE, EPERM, NULL));
1673 }
1674 
1675 /*
1676  * Processor set binding.
1677  */
1678 int
1679 secpolicy_pbind(const cred_t *cr)
1680 {
1681 	if (PRIV_POLICY_ONLY(cr, PRIV_SYS_RES_CONFIG, B_FALSE))
1682 		return (secpolicy_pset(cr));
1683 	return (PRIV_POLICY(cr, PRIV_SYS_RES_BIND, B_FALSE, EPERM, NULL));
1684 }
1685 
1686 int
1687 secpolicy_ponline(const cred_t *cr)
1688 {
1689 	return (PRIV_POLICY(cr, PRIV_SYS_RES_CONFIG, B_FALSE, EPERM, NULL));
1690 }
1691 
1692 int
1693 secpolicy_pool(const cred_t *cr)
1694 {
1695 	return (PRIV_POLICY(cr, PRIV_SYS_RES_CONFIG, B_FALSE, EPERM, NULL));
1696 }
1697 
1698 int
1699 secpolicy_blacklist(const cred_t *cr)
1700 {
1701 	return (PRIV_POLICY(cr, PRIV_SYS_RES_CONFIG, B_FALSE, EPERM, NULL));
1702 }
1703 
1704 /*
1705  * Catch all system configuration.
1706  */
1707 int
1708 secpolicy_sys_config(const cred_t *cr, boolean_t checkonly)
1709 {
1710 	if (checkonly) {
1711 		return (PRIV_POLICY_ONLY(cr, PRIV_SYS_CONFIG, B_FALSE) ? 0 :
1712 		    EPERM);
1713 	} else {
1714 		return (PRIV_POLICY(cr, PRIV_SYS_CONFIG, B_FALSE, EPERM, NULL));
1715 	}
1716 }
1717 
1718 /*
1719  * Zone administration (halt, reboot, etc.) from within zone.
1720  */
1721 int
1722 secpolicy_zone_admin(const cred_t *cr, boolean_t checkonly)
1723 {
1724 	if (checkonly) {
1725 		return (PRIV_POLICY_ONLY(cr, PRIV_SYS_ADMIN, B_FALSE) ? 0 :
1726 		    EPERM);
1727 	} else {
1728 		return (PRIV_POLICY(cr, PRIV_SYS_ADMIN, B_FALSE, EPERM,
1729 		    NULL));
1730 	}
1731 }
1732 
1733 /*
1734  * Zone configuration (create, halt, enter).
1735  */
1736 int
1737 secpolicy_zone_config(const cred_t *cr)
1738 {
1739 	/*
1740 	 * Require all privileges to avoid possibility of privilege
1741 	 * escalation.
1742 	 */
1743 	return (secpolicy_require_set(cr, PRIV_FULLSET, NULL, KLPDARG_NONE));
1744 }
1745 
1746 /*
1747  * Various other system configuration calls
1748  */
1749 int
1750 secpolicy_coreadm(const cred_t *cr)
1751 {
1752 	return (PRIV_POLICY(cr, PRIV_SYS_ADMIN, B_FALSE, EPERM, NULL));
1753 }
1754 
1755 int
1756 secpolicy_systeminfo(const cred_t *cr)
1757 {
1758 	return (PRIV_POLICY(cr, PRIV_SYS_ADMIN, B_FALSE, EPERM, NULL));
1759 }
1760 
1761 int
1762 secpolicy_dispadm(const cred_t *cr)
1763 {
1764 	return (PRIV_POLICY(cr, PRIV_SYS_CONFIG, B_FALSE, EPERM, NULL));
1765 }
1766 
1767 int
1768 secpolicy_settime(const cred_t *cr)
1769 {
1770 	return (PRIV_POLICY(cr, PRIV_SYS_TIME, B_FALSE, EPERM, NULL));
1771 }
1772 
1773 /*
1774  * For realtime users: high resolution clock.
1775  */
1776 int
1777 secpolicy_clock_highres(const cred_t *cr)
1778 {
1779 	return (PRIV_POLICY(cr, PRIV_PROC_CLOCK_HIGHRES, B_FALSE, EPERM,
1780 	    NULL));
1781 }
1782 
1783 /*
1784  * drv_priv() is documented as callable from interrupt context, not that
1785  * anyone ever does, but still.  No debugging or auditing can be done when
1786  * it is called from interrupt context.
1787  * returns 0 on succes, EPERM on failure.
1788  */
1789 int
1790 drv_priv(cred_t *cr)
1791 {
1792 	return (PRIV_POLICY(cr, PRIV_SYS_DEVICES, B_FALSE, EPERM, NULL));
1793 }
1794 
1795 int
1796 secpolicy_sys_devices(const cred_t *cr)
1797 {
1798 	return (PRIV_POLICY(cr, PRIV_SYS_DEVICES, B_FALSE, EPERM, NULL));
1799 }
1800 
1801 int
1802 secpolicy_excl_open(const cred_t *cr)
1803 {
1804 	return (PRIV_POLICY(cr, PRIV_SYS_DEVICES, B_FALSE, EBUSY, NULL));
1805 }
1806 
1807 int
1808 secpolicy_rctlsys(const cred_t *cr, boolean_t is_zone_rctl)
1809 {
1810 	/* zone.* rctls can only be set from the global zone */
1811 	if (is_zone_rctl && priv_policy_global(cr) != 0)
1812 		return (EPERM);
1813 	return (PRIV_POLICY(cr, PRIV_SYS_RESOURCE, B_FALSE, EPERM, NULL));
1814 }
1815 
1816 int
1817 secpolicy_resource(const cred_t *cr)
1818 {
1819 	return (PRIV_POLICY(cr, PRIV_SYS_RESOURCE, B_FALSE, EPERM, NULL));
1820 }
1821 
1822 int
1823 secpolicy_resource_anon_mem(const cred_t *cr)
1824 {
1825 	return (PRIV_POLICY_ONLY(cr, PRIV_SYS_RESOURCE, B_FALSE));
1826 }
1827 
1828 /*
1829  * Processes with a real uid of 0 escape any form of accounting, much
1830  * like before.
1831  */
1832 int
1833 secpolicy_newproc(const cred_t *cr)
1834 {
1835 	if (cr->cr_ruid == 0)
1836 		return (0);
1837 
1838 	return (PRIV_POLICY(cr, PRIV_SYS_RESOURCE, B_FALSE, EPERM, NULL));
1839 }
1840 
1841 /*
1842  * Networking
1843  */
1844 int
1845 secpolicy_net_rawaccess(const cred_t *cr)
1846 {
1847 	return (PRIV_POLICY(cr, PRIV_NET_RAWACCESS, B_FALSE, EACCES, NULL));
1848 }
1849 
1850 int
1851 secpolicy_net_observability(const cred_t *cr)
1852 {
1853 	return (PRIV_POLICY(cr, PRIV_NET_OBSERVABILITY, B_FALSE, EACCES, NULL));
1854 }
1855 
1856 /*
1857  * Need this privilege for accessing the ICMP device
1858  */
1859 int
1860 secpolicy_net_icmpaccess(const cred_t *cr)
1861 {
1862 	return (PRIV_POLICY(cr, PRIV_NET_ICMPACCESS, B_FALSE, EACCES, NULL));
1863 }
1864 
1865 /*
1866  * There are a few rare cases where the kernel generates ioctls() from
1867  * interrupt context with a credential of kcred rather than NULL.
1868  * In those cases, we take the safe and cheap test.
1869  */
1870 int
1871 secpolicy_net_config(const cred_t *cr, boolean_t checkonly)
1872 {
1873 	if (checkonly) {
1874 		return (PRIV_POLICY_ONLY(cr, PRIV_SYS_NET_CONFIG, B_FALSE) ?
1875 		    0 : EPERM);
1876 	} else {
1877 		return (PRIV_POLICY(cr, PRIV_SYS_NET_CONFIG, B_FALSE, EPERM,
1878 		    NULL));
1879 	}
1880 }
1881 
1882 
1883 /*
1884  * PRIV_SYS_NET_CONFIG is a superset of PRIV_SYS_IP_CONFIG.
1885  *
1886  * There are a few rare cases where the kernel generates ioctls() from
1887  * interrupt context with a credential of kcred rather than NULL.
1888  * In those cases, we take the safe and cheap test.
1889  */
1890 int
1891 secpolicy_ip_config(const cred_t *cr, boolean_t checkonly)
1892 {
1893 	if (PRIV_POLICY_ONLY(cr, PRIV_SYS_NET_CONFIG, B_FALSE))
1894 		return (secpolicy_net_config(cr, checkonly));
1895 
1896 	if (checkonly) {
1897 		return (PRIV_POLICY_ONLY(cr, PRIV_SYS_IP_CONFIG, B_FALSE) ?
1898 		    0 : EPERM);
1899 	} else {
1900 		return (PRIV_POLICY(cr, PRIV_SYS_IP_CONFIG, B_FALSE, EPERM,
1901 		    NULL));
1902 	}
1903 }
1904 
1905 /*
1906  * PRIV_SYS_NET_CONFIG is a superset of PRIV_SYS_DL_CONFIG.
1907  */
1908 int
1909 secpolicy_dl_config(const cred_t *cr)
1910 {
1911 	if (PRIV_POLICY_ONLY(cr, PRIV_SYS_NET_CONFIG, B_FALSE))
1912 		return (secpolicy_net_config(cr, B_FALSE));
1913 	return (PRIV_POLICY(cr, PRIV_SYS_DL_CONFIG, B_FALSE, EPERM, NULL));
1914 }
1915 
1916 /*
1917  * PRIV_SYS_DL_CONFIG is a superset of PRIV_SYS_IPTUN_CONFIG.
1918  */
1919 int
1920 secpolicy_iptun_config(const cred_t *cr)
1921 {
1922 	if (PRIV_POLICY_ONLY(cr, PRIV_SYS_NET_CONFIG, B_FALSE))
1923 		return (secpolicy_net_config(cr, B_FALSE));
1924 	if (PRIV_POLICY_ONLY(cr, PRIV_SYS_DL_CONFIG, B_FALSE))
1925 		return (secpolicy_dl_config(cr));
1926 	return (PRIV_POLICY(cr, PRIV_SYS_IPTUN_CONFIG, B_FALSE, EPERM, NULL));
1927 }
1928 
1929 /*
1930  * Map IP pseudo privileges to actual privileges.
1931  * So we don't need to recompile IP when we change the privileges.
1932  */
1933 int
1934 secpolicy_ip(const cred_t *cr, int netpriv, boolean_t checkonly)
1935 {
1936 	int priv = PRIV_ALL;
1937 
1938 	switch (netpriv) {
1939 	case OP_CONFIG:
1940 		priv = PRIV_SYS_IP_CONFIG;
1941 		break;
1942 	case OP_RAW:
1943 		priv = PRIV_NET_RAWACCESS;
1944 		break;
1945 	case OP_PRIVPORT:
1946 		priv = PRIV_NET_PRIVADDR;
1947 		break;
1948 	}
1949 	ASSERT(priv != PRIV_ALL);
1950 	if (checkonly)
1951 		return (PRIV_POLICY_ONLY(cr, priv, B_FALSE) ? 0 : EPERM);
1952 	else
1953 		return (PRIV_POLICY(cr, priv, B_FALSE, EPERM, NULL));
1954 }
1955 
1956 /*
1957  * Map network pseudo privileges to actual privileges.
1958  * So we don't need to recompile IP when we change the privileges.
1959  */
1960 int
1961 secpolicy_net(const cred_t *cr, int netpriv, boolean_t checkonly)
1962 {
1963 	int priv = PRIV_ALL;
1964 
1965 	switch (netpriv) {
1966 	case OP_CONFIG:
1967 		priv = PRIV_SYS_NET_CONFIG;
1968 		break;
1969 	case OP_RAW:
1970 		priv = PRIV_NET_RAWACCESS;
1971 		break;
1972 	case OP_PRIVPORT:
1973 		priv = PRIV_NET_PRIVADDR;
1974 		break;
1975 	}
1976 	ASSERT(priv != PRIV_ALL);
1977 	if (checkonly)
1978 		return (PRIV_POLICY_ONLY(cr, priv, B_FALSE) ? 0 : EPERM);
1979 	else
1980 		return (PRIV_POLICY(cr, priv, B_FALSE, EPERM, NULL));
1981 }
1982 
1983 /*
1984  * Checks for operations that are either client-only or are used by
1985  * both clients and servers.
1986  */
1987 int
1988 secpolicy_nfs(const cred_t *cr)
1989 {
1990 	return (PRIV_POLICY(cr, PRIV_SYS_NFS, B_FALSE, EPERM, NULL));
1991 }
1992 
1993 /*
1994  * Special case for opening rpcmod: have NFS privileges or network
1995  * config privileges.
1996  */
1997 int
1998 secpolicy_rpcmod_open(const cred_t *cr)
1999 {
2000 	if (PRIV_POLICY_ONLY(cr, PRIV_SYS_NFS, B_FALSE))
2001 		return (secpolicy_nfs(cr));
2002 	else
2003 		return (secpolicy_net_config(cr, NULL));
2004 }
2005 
2006 int
2007 secpolicy_chroot(const cred_t *cr)
2008 {
2009 	return (PRIV_POLICY(cr, PRIV_PROC_CHROOT, B_FALSE, EPERM, NULL));
2010 }
2011 
2012 int
2013 secpolicy_tasksys(const cred_t *cr)
2014 {
2015 	return (PRIV_POLICY(cr, PRIV_PROC_TASKID, B_FALSE, EPERM, NULL));
2016 }
2017 
2018 int
2019 secpolicy_pfexec_register(const cred_t *cr)
2020 {
2021 	return (PRIV_POLICY(cr, PRIV_SYS_ADMIN, B_TRUE, EPERM, NULL));
2022 }
2023 
2024 /*
2025  * Basic privilege checks.
2026  */
2027 int
2028 secpolicy_basic_exec(const cred_t *cr, vnode_t *vp)
2029 {
2030 	FAST_BASIC_CHECK(cr, PRIV_PROC_EXEC);
2031 
2032 	return (priv_policy_va(cr, PRIV_PROC_EXEC, B_FALSE, EPERM, NULL,
2033 	    KLPDARG_VNODE, vp, (char *)NULL, KLPDARG_NOMORE));
2034 }
2035 
2036 int
2037 secpolicy_basic_fork(const cred_t *cr)
2038 {
2039 	FAST_BASIC_CHECK(cr, PRIV_PROC_FORK);
2040 
2041 	return (PRIV_POLICY(cr, PRIV_PROC_FORK, B_FALSE, EPERM, NULL));
2042 }
2043 
2044 int
2045 secpolicy_basic_proc(const cred_t *cr)
2046 {
2047 	FAST_BASIC_CHECK(cr, PRIV_PROC_SESSION);
2048 
2049 	return (PRIV_POLICY(cr, PRIV_PROC_SESSION, B_FALSE, EPERM, NULL));
2050 }
2051 
2052 /*
2053  * Slightly complicated because we don't want to trigger the policy too
2054  * often.  First we shortcircuit access to "self" (tp == sp) or if
2055  * we don't have the privilege but if we have permission
2056  * just return (0) and we don't flag the privilege as needed.
2057  * Else, we test for the privilege because we either have it or need it.
2058  */
2059 int
2060 secpolicy_basic_procinfo(const cred_t *cr, proc_t *tp, proc_t *sp)
2061 {
2062 	if (tp == sp ||
2063 	    !HAS_PRIVILEGE(cr, PRIV_PROC_INFO) && prochasprocperm(tp, sp, cr)) {
2064 		return (0);
2065 	} else {
2066 		return (PRIV_POLICY(cr, PRIV_PROC_INFO, B_FALSE, EPERM, NULL));
2067 	}
2068 }
2069 
2070 int
2071 secpolicy_basic_link(const cred_t *cr)
2072 {
2073 	FAST_BASIC_CHECK(cr, PRIV_FILE_LINK_ANY);
2074 
2075 	return (PRIV_POLICY(cr, PRIV_FILE_LINK_ANY, B_FALSE, EPERM, NULL));
2076 }
2077 
2078 int
2079 secpolicy_basic_net_access(const cred_t *cr)
2080 {
2081 	FAST_BASIC_CHECK(cr, PRIV_NET_ACCESS);
2082 
2083 	return (PRIV_POLICY(cr, PRIV_NET_ACCESS, B_FALSE, EACCES, NULL));
2084 }
2085 
2086 /* ARGSUSED */
2087 int
2088 secpolicy_basic_file_read(const cred_t *cr, vnode_t *vp, const char *pn)
2089 {
2090 	FAST_BASIC_CHECK(cr, PRIV_FILE_READ);
2091 
2092 	return (priv_policy_va(cr, PRIV_FILE_READ, B_FALSE, EACCES, NULL,
2093 	    KLPDARG_VNODE, vp, (char *)pn, KLPDARG_NOMORE));
2094 }
2095 
2096 /* ARGSUSED */
2097 int
2098 secpolicy_basic_file_write(const cred_t *cr, vnode_t *vp, const char *pn)
2099 {
2100 	FAST_BASIC_CHECK(cr, PRIV_FILE_WRITE);
2101 
2102 	return (priv_policy_va(cr, PRIV_FILE_WRITE, B_FALSE, EACCES, NULL,
2103 	    KLPDARG_VNODE, vp, (char *)pn, KLPDARG_NOMORE));
2104 }
2105 
2106 /*
2107  * Additional device protection.
2108  *
2109  * Traditionally, a device has specific permissions on the node in
2110  * the filesystem which govern which devices can be opened by what
2111  * processes.  In certain cases, it is desirable to add extra
2112  * restrictions, as writing to certain devices is identical to
2113  * having a complete run of the system.
2114  *
2115  * This mechanism is called the device policy.
2116  *
2117  * When a device is opened, its policy entry is looked up in the
2118  * policy cache and checked.
2119  */
2120 int
2121 secpolicy_spec_open(const cred_t *cr, struct vnode *vp, int oflag)
2122 {
2123 	devplcy_t *plcy;
2124 	int err;
2125 	struct snode *csp = VTOS(common_specvp(vp));
2126 	priv_set_t pset;
2127 
2128 	mutex_enter(&csp->s_lock);
2129 
2130 	if (csp->s_plcy == NULL || csp->s_plcy->dp_gen != devplcy_gen) {
2131 		plcy = devpolicy_find(vp);
2132 		if (csp->s_plcy)
2133 			dpfree(csp->s_plcy);
2134 		csp->s_plcy = plcy;
2135 		ASSERT(plcy != NULL);
2136 	} else
2137 		plcy = csp->s_plcy;
2138 
2139 	if (plcy == nullpolicy) {
2140 		mutex_exit(&csp->s_lock);
2141 		return (0);
2142 	}
2143 
2144 	dphold(plcy);
2145 
2146 	mutex_exit(&csp->s_lock);
2147 
2148 	if (oflag & FWRITE)
2149 		pset = plcy->dp_wrp;
2150 	else
2151 		pset = plcy->dp_rdp;
2152 	/*
2153 	 * Special case:
2154 	 * PRIV_SYS_NET_CONFIG is a superset of PRIV_SYS_IP_CONFIG.
2155 	 * If PRIV_SYS_NET_CONFIG is present and PRIV_SYS_IP_CONFIG is
2156 	 * required, replace PRIV_SYS_IP_CONFIG with PRIV_SYS_NET_CONFIG
2157 	 * in the required privilege set before doing the check.
2158 	 */
2159 	if (priv_ismember(&pset, PRIV_SYS_IP_CONFIG) &&
2160 	    priv_ismember(&CR_OEPRIV(cr), PRIV_SYS_NET_CONFIG) &&
2161 	    !priv_ismember(&CR_OEPRIV(cr), PRIV_SYS_IP_CONFIG)) {
2162 		priv_delset(&pset, PRIV_SYS_IP_CONFIG);
2163 		priv_addset(&pset, PRIV_SYS_NET_CONFIG);
2164 	}
2165 
2166 	err = secpolicy_require_set(cr, &pset, "devpolicy", KLPDARG_NONE);
2167 	dpfree(plcy);
2168 
2169 	return (err);
2170 }
2171 
2172 int
2173 secpolicy_modctl(const cred_t *cr, int cmd)
2174 {
2175 	switch (cmd) {
2176 	case MODINFO:
2177 	case MODGETMAJBIND:
2178 	case MODGETPATH:
2179 	case MODGETPATHLEN:
2180 	case MODGETNAME:
2181 	case MODGETFBNAME:
2182 	case MODGETDEVPOLICY:
2183 	case MODGETDEVPOLICYBYNAME:
2184 	case MODDEVT2INSTANCE:
2185 	case MODSIZEOF_DEVID:
2186 	case MODGETDEVID:
2187 	case MODSIZEOF_MINORNAME:
2188 	case MODGETMINORNAME:
2189 	case MODGETDEVFSPATH_LEN:
2190 	case MODGETDEVFSPATH:
2191 	case MODGETDEVFSPATH_MI_LEN:
2192 	case MODGETDEVFSPATH_MI:
2193 		/* Unprivileged */
2194 		return (0);
2195 	case MODLOAD:
2196 	case MODSETDEVPOLICY:
2197 		return (secpolicy_require_set(cr, PRIV_FULLSET, NULL,
2198 		    KLPDARG_NONE));
2199 	default:
2200 		return (secpolicy_sys_config(cr, B_FALSE));
2201 	}
2202 }
2203 
2204 int
2205 secpolicy_console(const cred_t *cr)
2206 {
2207 	return (PRIV_POLICY(cr, PRIV_SYS_DEVICES, B_FALSE, EPERM, NULL));
2208 }
2209 
2210 int
2211 secpolicy_power_mgmt(const cred_t *cr)
2212 {
2213 	return (PRIV_POLICY(cr, PRIV_SYS_DEVICES, B_FALSE, EPERM, NULL));
2214 }
2215 
2216 /*
2217  * Simulate terminal input; another escalation of privileges avenue.
2218  */
2219 
2220 int
2221 secpolicy_sti(const cred_t *cr)
2222 {
2223 	return (secpolicy_require_set(cr, PRIV_FULLSET, NULL, KLPDARG_NONE));
2224 }
2225 
2226 boolean_t
2227 secpolicy_net_reply_equal(const cred_t *cr)
2228 {
2229 	return (PRIV_POLICY(cr, PRIV_SYS_CONFIG, B_FALSE, EPERM, NULL));
2230 }
2231 
2232 int
2233 secpolicy_swapctl(const cred_t *cr)
2234 {
2235 	return (PRIV_POLICY(cr, PRIV_SYS_CONFIG, B_FALSE, EPERM, NULL));
2236 }
2237 
2238 int
2239 secpolicy_cpc_cpu(const cred_t *cr)
2240 {
2241 	return (PRIV_POLICY(cr, PRIV_CPC_CPU, B_FALSE, EACCES, NULL));
2242 }
2243 
2244 /*
2245  * secpolicy_contract_identity
2246  *
2247  * Determine if the subject may set the process contract FMRI value
2248  */
2249 int
2250 secpolicy_contract_identity(const cred_t *cr)
2251 {
2252 	return (PRIV_POLICY(cr, PRIV_CONTRACT_IDENTITY, B_FALSE, EPERM, NULL));
2253 }
2254 
2255 /*
2256  * secpolicy_contract_observer
2257  *
2258  * Determine if the subject may observe a specific contract's events.
2259  */
2260 int
2261 secpolicy_contract_observer(const cred_t *cr, struct contract *ct)
2262 {
2263 	if (contract_owned(ct, cr, B_FALSE))
2264 		return (0);
2265 	return (PRIV_POLICY(cr, PRIV_CONTRACT_OBSERVER, B_FALSE, EPERM, NULL));
2266 }
2267 
2268 /*
2269  * secpolicy_contract_observer_choice
2270  *
2271  * Determine if the subject may observe any contract's events.  Just
2272  * tests privilege and audits on success.
2273  */
2274 boolean_t
2275 secpolicy_contract_observer_choice(const cred_t *cr)
2276 {
2277 	return (PRIV_POLICY_CHOICE(cr, PRIV_CONTRACT_OBSERVER, B_FALSE));
2278 }
2279 
2280 /*
2281  * secpolicy_contract_event
2282  *
2283  * Determine if the subject may request critical contract events or
2284  * reliable contract event delivery.
2285  */
2286 int
2287 secpolicy_contract_event(const cred_t *cr)
2288 {
2289 	return (PRIV_POLICY(cr, PRIV_CONTRACT_EVENT, B_FALSE, EPERM, NULL));
2290 }
2291 
2292 /*
2293  * secpolicy_contract_event_choice
2294  *
2295  * Determine if the subject may retain contract events in its critical
2296  * set when a change in other terms would normally require a change in
2297  * the critical set.  Just tests privilege and audits on success.
2298  */
2299 boolean_t
2300 secpolicy_contract_event_choice(const cred_t *cr)
2301 {
2302 	return (PRIV_POLICY_CHOICE(cr, PRIV_CONTRACT_EVENT, B_FALSE));
2303 }
2304 
2305 /*
2306  * secpolicy_gart_access
2307  *
2308  * Determine if the subject has sufficient priveleges to make ioctls to agpgart
2309  * device.
2310  */
2311 int
2312 secpolicy_gart_access(const cred_t *cr)
2313 {
2314 	return (PRIV_POLICY(cr, PRIV_GRAPHICS_ACCESS, B_FALSE, EPERM, NULL));
2315 }
2316 
2317 /*
2318  * secpolicy_gart_map
2319  *
2320  * Determine if the subject has sufficient priveleges to map aperture range
2321  * through agpgart driver.
2322  */
2323 int
2324 secpolicy_gart_map(const cred_t *cr)
2325 {
2326 	if (PRIV_POLICY_ONLY(cr, PRIV_GRAPHICS_ACCESS, B_FALSE)) {
2327 		return (PRIV_POLICY(cr, PRIV_GRAPHICS_ACCESS, B_FALSE, EPERM,
2328 		    NULL));
2329 	} else {
2330 		return (PRIV_POLICY(cr, PRIV_GRAPHICS_MAP, B_FALSE, EPERM,
2331 		    NULL));
2332 	}
2333 }
2334 
2335 /*
2336  * secpolicy_zinject
2337  *
2338  * Determine if the subject can inject faults in the ZFS fault injection
2339  * framework.  Requires all privileges.
2340  */
2341 int
2342 secpolicy_zinject(const cred_t *cr)
2343 {
2344 	return (secpolicy_require_set(cr, PRIV_FULLSET, NULL, KLPDARG_NONE));
2345 }
2346 
2347 /*
2348  * secpolicy_zfs
2349  *
2350  * Determine if the subject has permission to manipulate ZFS datasets
2351  * (not pools).  Equivalent to the SYS_MOUNT privilege.
2352  */
2353 int
2354 secpolicy_zfs(const cred_t *cr)
2355 {
2356 	return (PRIV_POLICY(cr, PRIV_SYS_MOUNT, B_FALSE, EPERM, NULL));
2357 }
2358 
2359 /*
2360  * secpolicy_idmap
2361  *
2362  * Determine if the calling process has permissions to register an SID
2363  * mapping daemon and allocate ephemeral IDs.
2364  */
2365 int
2366 secpolicy_idmap(const cred_t *cr)
2367 {
2368 	return (PRIV_POLICY(cr, PRIV_FILE_SETID, B_TRUE, EPERM, NULL));
2369 }
2370 
2371 /*
2372  * secpolicy_ucode_update
2373  *
2374  * Determine if the subject has sufficient privilege to update microcode.
2375  */
2376 int
2377 secpolicy_ucode_update(const cred_t *scr)
2378 {
2379 	return (PRIV_POLICY(scr, PRIV_ALL, B_FALSE, EPERM, NULL));
2380 }
2381 
2382 /*
2383  * secpolicy_sadopen
2384  *
2385  * Determine if the subject has sufficient privilege to access /dev/sad/admin.
2386  * /dev/sad/admin appear in global zone and exclusive-IP zones only.
2387  * In global zone, sys_config is required.
2388  * In exclusive-IP zones, sys_ip_config is required.
2389  * Note that sys_config is prohibited in non-global zones.
2390  */
2391 int
2392 secpolicy_sadopen(const cred_t *credp)
2393 {
2394 	priv_set_t pset;
2395 
2396 	priv_emptyset(&pset);
2397 
2398 	if (crgetzoneid(credp) == GLOBAL_ZONEID)
2399 		priv_addset(&pset, PRIV_SYS_CONFIG);
2400 	else
2401 		priv_addset(&pset, PRIV_SYS_IP_CONFIG);
2402 
2403 	return (secpolicy_require_set(credp, &pset, "devpolicy", KLPDARG_NONE));
2404 }
2405 
2406 
2407 /*
2408  * Add privileges to a particular privilege set; this is called when the
2409  * current sets of privileges are not sufficient.  I.e., we should always
2410  * call the policy override functions from here.
2411  * What we are allowed to have is in the Observed Permitted set; so
2412  * we compute the difference between that and the newset.
2413  */
2414 int
2415 secpolicy_require_privs(const cred_t *cr, const priv_set_t *nset)
2416 {
2417 	priv_set_t rqd;
2418 
2419 	rqd = CR_OPPRIV(cr);
2420 
2421 	priv_inverse(&rqd);
2422 	priv_intersect(nset, &rqd);
2423 
2424 	return (secpolicy_require_set(cr, &rqd, NULL, KLPDARG_NONE));
2425 }
2426 
2427 /*
2428  * secpolicy_smb
2429  *
2430  * Determine if the cred_t has PRIV_SYS_SMB privilege, indicating
2431  * that it has permission to access the smbsrv kernel driver.
2432  * PRIV_POLICY checks the privilege and audits the check.
2433  *
2434  * Returns:
2435  * 0       Driver access is allowed.
2436  * EPERM   Driver access is NOT permitted.
2437  */
2438 int
2439 secpolicy_smb(const cred_t *cr)
2440 {
2441 	return (PRIV_POLICY(cr, PRIV_SYS_SMB, B_FALSE, EPERM, NULL));
2442 }
2443 
2444 /*
2445  * secpolicy_vscan
2446  *
2447  * Determine if cred_t has the necessary privileges to access a file
2448  * for virus scanning and update its extended system attributes.
2449  * PRIV_FILE_DAC_SEARCH, PRIV_FILE_DAC_READ - file access
2450  * PRIV_FILE_FLAG_SET - set extended system attributes
2451  *
2452  * PRIV_POLICY checks the privilege and audits the check.
2453  *
2454  * Returns:
2455  * 0      file access for virus scanning allowed.
2456  * EPERM  file access for virus scanning is NOT permitted.
2457  */
2458 int
2459 secpolicy_vscan(const cred_t *cr)
2460 {
2461 	if ((PRIV_POLICY(cr, PRIV_FILE_DAC_SEARCH, B_FALSE, EPERM, NULL)) ||
2462 	    (PRIV_POLICY(cr, PRIV_FILE_DAC_READ, B_FALSE, EPERM, NULL)) ||
2463 	    (PRIV_POLICY(cr, PRIV_FILE_FLAG_SET, B_FALSE, EPERM, NULL))) {
2464 		return (EPERM);
2465 	}
2466 
2467 	return (0);
2468 }
2469 
2470 /*
2471  * secpolicy_smbfs_login
2472  *
2473  * Determines if the caller can add and delete the smbfs login
2474  * password in the the nsmb kernel module for the CIFS client.
2475  *
2476  * Returns:
2477  * 0       access is allowed.
2478  * EPERM   access is NOT allowed.
2479  */
2480 int
2481 secpolicy_smbfs_login(const cred_t *cr, uid_t uid)
2482 {
2483 	uid_t cruid = crgetruid(cr);
2484 
2485 	if (cruid == uid)
2486 		return (0);
2487 	return (PRIV_POLICY(cr, PRIV_PROC_OWNER, B_FALSE,
2488 	    EPERM, NULL));
2489 }
2490 
2491 /*
2492  * secpolicy_xvm_control
2493  *
2494  * Determines if a caller can control the xVM hypervisor and/or running
2495  * domains (x86 specific).
2496  *
2497  * Returns:
2498  * 0       access is allowed.
2499  * EPERM   access is NOT allowed.
2500  */
2501 int
2502 secpolicy_xvm_control(const cred_t *cr)
2503 {
2504 	if (PRIV_POLICY(cr, PRIV_XVM_CONTROL, B_FALSE, EPERM, NULL))
2505 		return (EPERM);
2506 	return (0);
2507 }
2508 
2509 /*
2510  * secpolicy_ppp_config
2511  *
2512  * Determine if the subject has sufficient privileges to configure PPP and
2513  * PPP-related devices.
2514  */
2515 int
2516 secpolicy_ppp_config(const cred_t *cr)
2517 {
2518 	if (PRIV_POLICY_ONLY(cr, PRIV_SYS_NET_CONFIG, B_FALSE))
2519 		return (secpolicy_net_config(cr, B_FALSE));
2520 	return (PRIV_POLICY(cr, PRIV_SYS_PPP_CONFIG, B_FALSE, EPERM, NULL));
2521 }
2522