xref: /openbsd-src/sys/kern/kern_pledge.c (revision c1a45aed656e7d5627c30c92421893a76f370ccb)
1 /*	$OpenBSD: kern_pledge.c,v 1.281 2022/03/25 17:40:59 robert Exp $	*/
2 
3 /*
4  * Copyright (c) 2015 Nicholas Marriott <nicm@openbsd.org>
5  * Copyright (c) 2015 Theo de Raadt <deraadt@openbsd.org>
6  *
7  * Permission to use, copy, modify, and distribute this software for any
8  * purpose with or without fee is hereby granted, provided that the above
9  * copyright notice and this permission notice appear in all copies.
10  *
11  * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
12  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
13  * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
14  * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
15  * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
16  * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
17  * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
18  */
19 
20 #include <sys/param.h>
21 
22 #include <sys/mount.h>
23 #include <sys/proc.h>
24 #include <sys/fcntl.h>
25 #include <sys/file.h>
26 #include <sys/filedesc.h>
27 #include <sys/namei.h>
28 #include <sys/pool.h>
29 #include <sys/socketvar.h>
30 #include <sys/vnode.h>
31 #include <sys/mbuf.h>
32 #include <sys/mman.h>
33 #include <sys/sysctl.h>
34 #include <sys/syslog.h>
35 #include <sys/ktrace.h>
36 #include <sys/acct.h>
37 
38 #include <sys/ioctl.h>
39 #include <sys/termios.h>
40 #include <sys/tty.h>
41 #include <sys/device.h>
42 #include <sys/disklabel.h>
43 #include <sys/dkio.h>
44 #include <sys/mtio.h>
45 #include <sys/audioio.h>
46 #include <sys/videoio.h>
47 #include <net/bpf.h>
48 #include <net/route.h>
49 #include <net/if.h>
50 #include <net/if_var.h>
51 #include <netinet/in.h>
52 #include <netinet6/in6_var.h>
53 #include <netinet6/nd6.h>
54 #include <netinet/tcp.h>
55 #include <net/pfvar.h>
56 
57 #include <sys/conf.h>
58 #include <sys/specdev.h>
59 #include <sys/signal.h>
60 #include <sys/signalvar.h>
61 #include <sys/syscall.h>
62 #include <sys/syscallargs.h>
63 #include <sys/systm.h>
64 
65 #include <dev/biovar.h>
66 
67 #define PLEDGENAMES
68 #include <sys/pledge.h>
69 
70 #include "audio.h"
71 #include "bpfilter.h"
72 #include "pf.h"
73 #include "video.h"
74 #include "pty.h"
75 
76 #if defined(__amd64__)
77 #include "vmm.h"
78 #if NVMM > 0
79 #include <machine/conf.h>
80 #endif
81 #endif
82 
83 #include "drm.h"
84 
85 uint64_t pledgereq_flags(const char *req);
86 int	 parsepledges(struct proc *p, const char *kname,
87 	    const char *promises, u_int64_t *fp);
88 int	 canonpath(const char *input, char *buf, size_t bufsize);
89 void	 unveil_destroy(struct process *ps);
90 
91 /* #define DEBUG_PLEDGE */
92 #ifdef DEBUG_PLEDGE
93 int debug_pledge = 1;
94 #define DPRINTF(x...)    do { if (debug_pledge) printf(x); } while (0)
95 #define DNPRINTF(n,x...) do { if (debug_pledge >= (n)) printf(x); } while (0)
96 #else
97 #define DPRINTF(x...)
98 #define DNPRINTF(n,x...)
99 #endif
100 
101 /*
102  * Ordered in blocks starting with least risky and most required.
103  */
104 const uint64_t pledge_syscalls[SYS_MAXSYSCALL] = {
105 	/*
106 	 * Minimum required
107 	 */
108 	[SYS_exit] = PLEDGE_ALWAYS,
109 	[SYS_kbind] = PLEDGE_ALWAYS,
110 	[SYS_msyscall] = PLEDGE_ALWAYS,
111 	[SYS___get_tcb] = PLEDGE_ALWAYS,
112 	[SYS___set_tcb] = PLEDGE_ALWAYS,
113 	[SYS_pledge] = PLEDGE_ALWAYS,
114 	[SYS_sendsyslog] = PLEDGE_ALWAYS,	/* stack protector reporting */
115 	[SYS_thrkill] = PLEDGE_ALWAYS,		/* raise, abort, stack pro */
116 	[SYS_utrace] = PLEDGE_ALWAYS,		/* ltrace(1) from ld.so */
117 
118 	/* "getting" information about self is considered safe */
119 	[SYS_getuid] = PLEDGE_STDIO,
120 	[SYS_geteuid] = PLEDGE_STDIO,
121 	[SYS_getresuid] = PLEDGE_STDIO,
122 	[SYS_getgid] = PLEDGE_STDIO,
123 	[SYS_getegid] = PLEDGE_STDIO,
124 	[SYS_getresgid] = PLEDGE_STDIO,
125 	[SYS_getgroups] = PLEDGE_STDIO,
126 	[SYS_getlogin_r] = PLEDGE_STDIO,
127 	[SYS_getpgrp] = PLEDGE_STDIO,
128 	[SYS_getpgid] = PLEDGE_STDIO,
129 	[SYS_getppid] = PLEDGE_STDIO,
130 	[SYS_getsid] = PLEDGE_STDIO,
131 	[SYS_getthrid] = PLEDGE_STDIO,
132 	[SYS_getrlimit] = PLEDGE_STDIO,
133 	[SYS_getrtable] = PLEDGE_STDIO,
134 	[SYS_gettimeofday] = PLEDGE_STDIO,
135 	[SYS_getdtablecount] = PLEDGE_STDIO,
136 	[SYS_getrusage] = PLEDGE_STDIO,
137 	[SYS_issetugid] = PLEDGE_STDIO,
138 	[SYS_clock_getres] = PLEDGE_STDIO,
139 	[SYS_clock_gettime] = PLEDGE_STDIO,
140 	[SYS_getpid] = PLEDGE_STDIO,
141 
142 	/*
143 	 * Almost exclusively read-only, Very narrow subset.
144 	 * Use of "route", "inet", "dns", "ps", or "vminfo"
145 	 * expands access.
146 	 */
147 	[SYS_sysctl] = PLEDGE_STDIO,
148 
149 	/* Support for malloc(3) family of operations */
150 	[SYS_getentropy] = PLEDGE_STDIO,
151 	[SYS_madvise] = PLEDGE_STDIO,
152 	[SYS_minherit] = PLEDGE_STDIO,
153 	[SYS_mmap] = PLEDGE_STDIO,
154 	[SYS_mprotect] = PLEDGE_STDIO,
155 	[SYS_mquery] = PLEDGE_STDIO,
156 	[SYS_munmap] = PLEDGE_STDIO,
157 	[SYS_msync] = PLEDGE_STDIO,
158 	[SYS_break] = PLEDGE_STDIO,
159 
160 	[SYS_umask] = PLEDGE_STDIO,
161 
162 	/* read/write operations */
163 	[SYS_read] = PLEDGE_STDIO,
164 	[SYS_readv] = PLEDGE_STDIO,
165 	[SYS_pread] = PLEDGE_STDIO,
166 	[SYS_preadv] = PLEDGE_STDIO,
167 	[SYS_write] = PLEDGE_STDIO,
168 	[SYS_writev] = PLEDGE_STDIO,
169 	[SYS_pwrite] = PLEDGE_STDIO,
170 	[SYS_pwritev] = PLEDGE_STDIO,
171 	[SYS_recvmsg] = PLEDGE_STDIO,
172 	[SYS_recvfrom] = PLEDGE_STDIO | PLEDGE_YPACTIVE,
173 	[SYS_ftruncate] = PLEDGE_STDIO,
174 	[SYS_lseek] = PLEDGE_STDIO,
175 	[SYS_fpathconf] = PLEDGE_STDIO,
176 
177 #if 1
178 	[SYS_pad_mquery] = PLEDGE_STDIO,
179 	[SYS_pad_mmap] = PLEDGE_STDIO,
180 	[SYS_pad_pread] = PLEDGE_STDIO,
181 	[SYS_pad_preadv] = PLEDGE_STDIO,
182 	[SYS_pad_pwrite] = PLEDGE_STDIO,
183 	[SYS_pad_pwritev] = PLEDGE_STDIO,
184 	[SYS_pad_ftruncate] = PLEDGE_STDIO,
185 	[SYS_pad_lseek] = PLEDGE_STDIO,
186 	[SYS_pad_truncate] = PLEDGE_WPATH,
187 #endif
188 
189 	/*
190 	 * Address selection required a network pledge ("inet",
191 	 * "unix", "dns".
192 	 */
193 	[SYS_sendto] = PLEDGE_STDIO | PLEDGE_YPACTIVE,
194 
195 	/*
196 	 * Address specification required a network pledge ("inet",
197 	 * "unix", "dns".  SCM_RIGHTS requires "sendfd" or "recvfd".
198 	 */
199 	[SYS_sendmsg] = PLEDGE_STDIO,
200 
201 	/* Common signal operations */
202 	[SYS_nanosleep] = PLEDGE_STDIO,
203 	[SYS_sigaltstack] = PLEDGE_STDIO,
204 	[SYS_sigprocmask] = PLEDGE_STDIO,
205 	[SYS_sigsuspend] = PLEDGE_STDIO,
206 	[SYS_sigaction] = PLEDGE_STDIO,
207 	[SYS_sigreturn] = PLEDGE_STDIO,
208 	[SYS_sigpending] = PLEDGE_STDIO,
209 	[SYS_getitimer] = PLEDGE_STDIO,
210 	[SYS_setitimer] = PLEDGE_STDIO,
211 
212 	/*
213 	 * To support event driven programming.
214 	 */
215 	[SYS_poll] = PLEDGE_STDIO,
216 	[SYS_ppoll] = PLEDGE_STDIO,
217 	[SYS_kevent] = PLEDGE_STDIO,
218 	[SYS_kqueue] = PLEDGE_STDIO,
219 	[SYS_select] = PLEDGE_STDIO,
220 	[SYS_pselect] = PLEDGE_STDIO,
221 
222 	[SYS_fstat] = PLEDGE_STDIO,
223 	[SYS_fsync] = PLEDGE_STDIO,
224 
225 	[SYS_setsockopt] = PLEDGE_STDIO,	/* narrow whitelist */
226 	[SYS_getsockopt] = PLEDGE_STDIO,	/* narrow whitelist */
227 
228 	/* F_SETOWN requires PLEDGE_PROC */
229 	[SYS_fcntl] = PLEDGE_STDIO,
230 
231 	[SYS_close] = PLEDGE_STDIO,
232 	[SYS_dup] = PLEDGE_STDIO,
233 	[SYS_dup2] = PLEDGE_STDIO,
234 	[SYS_dup3] = PLEDGE_STDIO,
235 	[SYS_closefrom] = PLEDGE_STDIO,
236 	[SYS_shutdown] = PLEDGE_STDIO,
237 	[SYS_fchdir] = PLEDGE_STDIO,	/* XXX consider tightening */
238 
239 	[SYS_pipe] = PLEDGE_STDIO,
240 	[SYS_pipe2] = PLEDGE_STDIO,
241 	[SYS_socketpair] = PLEDGE_STDIO,
242 
243 	[SYS_wait4] = PLEDGE_STDIO,
244 
245 	/*
246 	 * Can kill self with "stdio".  Killing another pid
247 	 * requires "proc"
248 	 */
249 	[SYS_kill] = PLEDGE_STDIO,
250 
251 	/*
252 	 * FIONREAD/FIONBIO for "stdio"
253 	 * Other ioctl are selectively allowed based upon other pledges.
254 	 */
255 	[SYS_ioctl] = PLEDGE_STDIO,
256 
257 	/*
258 	 * Path access/creation calls encounter many extensive
259 	 * checks done during pledge_namei()
260 	 */
261 	[SYS_open] = PLEDGE_STDIO,
262 	[SYS_stat] = PLEDGE_STDIO,
263 	[SYS_access] = PLEDGE_STDIO,
264 	[SYS_readlink] = PLEDGE_STDIO,
265 	[SYS___realpath] = PLEDGE_STDIO,
266 
267 	[SYS_adjtime] = PLEDGE_STDIO,   /* setting requires "settime" */
268 	[SYS_adjfreq] = PLEDGE_SETTIME,
269 	[SYS_settimeofday] = PLEDGE_SETTIME,
270 
271 	/*
272 	 * Needed by threaded programs
273 	 * XXX should we have a new "threads"?
274 	 */
275 	[SYS___tfork] = PLEDGE_STDIO,
276 	[SYS_sched_yield] = PLEDGE_STDIO,
277 	[SYS_futex] = PLEDGE_STDIO,
278 	[SYS___thrsleep] = PLEDGE_STDIO,
279 	[SYS___thrwakeup] = PLEDGE_STDIO,
280 	[SYS___threxit] = PLEDGE_STDIO,
281 	[SYS___thrsigdivert] = PLEDGE_STDIO,
282 
283 	[SYS_fork] = PLEDGE_PROC,
284 	[SYS_vfork] = PLEDGE_PROC,
285 	[SYS_setpgid] = PLEDGE_PROC,
286 	[SYS_setsid] = PLEDGE_PROC,
287 
288 	[SYS_setrlimit] = PLEDGE_PROC | PLEDGE_ID,
289 	[SYS_getpriority] = PLEDGE_PROC | PLEDGE_ID,
290 
291 	[SYS_setpriority] = PLEDGE_PROC | PLEDGE_ID,
292 
293 	[SYS_setuid] = PLEDGE_ID,
294 	[SYS_seteuid] = PLEDGE_ID,
295 	[SYS_setreuid] = PLEDGE_ID,
296 	[SYS_setresuid] = PLEDGE_ID,
297 	[SYS_setgid] = PLEDGE_ID,
298 	[SYS_setegid] = PLEDGE_ID,
299 	[SYS_setregid] = PLEDGE_ID,
300 	[SYS_setresgid] = PLEDGE_ID,
301 	[SYS_setgroups] = PLEDGE_ID,
302 	[SYS_setlogin] = PLEDGE_ID,
303 	[SYS_setrtable] = PLEDGE_ID,
304 
305 	[SYS_unveil] = PLEDGE_UNVEIL,
306 
307 	[SYS_execve] = PLEDGE_EXEC,
308 
309 	[SYS_chdir] = PLEDGE_RPATH,
310 	[SYS_openat] = PLEDGE_RPATH | PLEDGE_WPATH,
311 	[SYS_fstatat] = PLEDGE_RPATH | PLEDGE_WPATH,
312 	[SYS_faccessat] = PLEDGE_RPATH | PLEDGE_WPATH,
313 	[SYS_readlinkat] = PLEDGE_RPATH | PLEDGE_WPATH,
314 	[SYS_lstat] = PLEDGE_RPATH | PLEDGE_WPATH | PLEDGE_TMPPATH,
315 	[SYS_truncate] = PLEDGE_WPATH,
316 	[SYS_rename] = PLEDGE_RPATH | PLEDGE_CPATH,
317 	[SYS_rmdir] = PLEDGE_CPATH,
318 	[SYS_renameat] = PLEDGE_CPATH,
319 	[SYS_link] = PLEDGE_CPATH,
320 	[SYS_linkat] = PLEDGE_CPATH,
321 	[SYS_symlink] = PLEDGE_CPATH,
322 	[SYS_symlinkat] = PLEDGE_CPATH,
323 	[SYS_unlink] = PLEDGE_CPATH | PLEDGE_TMPPATH,
324 	[SYS_unlinkat] = PLEDGE_CPATH,
325 	[SYS_mkdir] = PLEDGE_CPATH,
326 	[SYS_mkdirat] = PLEDGE_CPATH,
327 
328 	[SYS_mkfifo] = PLEDGE_DPATH,
329 	[SYS_mkfifoat] = PLEDGE_DPATH,
330 	[SYS_mknod] = PLEDGE_DPATH,
331 	[SYS_mknodat] = PLEDGE_DPATH,
332 
333 	[SYS_revoke] = PLEDGE_TTY,	/* also requires PLEDGE_RPATH */
334 
335 	/*
336 	 * Classify as RPATH|WPATH, because of path information leakage.
337 	 * WPATH due to unknown use of mk*temp(3) on non-/tmp paths..
338 	 */
339 	[SYS___getcwd] = PLEDGE_RPATH | PLEDGE_WPATH,
340 
341 	/* Classify as RPATH, because these leak path information */
342 	[SYS_getdents] = PLEDGE_RPATH,
343 	[SYS_getfsstat] = PLEDGE_RPATH,
344 	[SYS_statfs] = PLEDGE_RPATH,
345 	[SYS_fstatfs] = PLEDGE_RPATH,
346 	[SYS_pathconf] = PLEDGE_RPATH,
347 
348 	[SYS_utimes] = PLEDGE_FATTR,
349 	[SYS_futimes] = PLEDGE_FATTR,
350 	[SYS_utimensat] = PLEDGE_FATTR,
351 	[SYS_futimens] = PLEDGE_FATTR,
352 	[SYS_chmod] = PLEDGE_FATTR,
353 	[SYS_fchmod] = PLEDGE_FATTR,
354 	[SYS_fchmodat] = PLEDGE_FATTR,
355 	[SYS_chflags] = PLEDGE_FATTR,
356 	[SYS_chflagsat] = PLEDGE_FATTR,
357 	[SYS_fchflags] = PLEDGE_FATTR,
358 
359 	[SYS_chown] = PLEDGE_CHOWN,
360 	[SYS_fchownat] = PLEDGE_CHOWN,
361 	[SYS_lchown] = PLEDGE_CHOWN,
362 	[SYS_fchown] = PLEDGE_CHOWN,
363 
364 	[SYS_socket] = PLEDGE_INET | PLEDGE_UNIX | PLEDGE_DNS | PLEDGE_YPACTIVE,
365 	[SYS_connect] = PLEDGE_INET | PLEDGE_UNIX | PLEDGE_DNS | PLEDGE_YPACTIVE,
366 	[SYS_bind] = PLEDGE_INET | PLEDGE_UNIX | PLEDGE_DNS | PLEDGE_YPACTIVE,
367 	[SYS_getsockname] = PLEDGE_INET | PLEDGE_UNIX | PLEDGE_DNS | PLEDGE_YPACTIVE,
368 
369 	[SYS_listen] = PLEDGE_INET | PLEDGE_UNIX,
370 	[SYS_accept4] = PLEDGE_INET | PLEDGE_UNIX,
371 	[SYS_accept] = PLEDGE_INET | PLEDGE_UNIX,
372 	[SYS_getpeername] = PLEDGE_INET | PLEDGE_UNIX,
373 
374 	[SYS_flock] = PLEDGE_FLOCK | PLEDGE_YPACTIVE,
375 
376 	[SYS_swapctl] = PLEDGE_VMINFO,	/* XXX should limit to "get" operations */
377 };
378 
379 static const struct {
380 	char *name;
381 	uint64_t flags;
382 } pledgereq[] = {
383 	{ "audio",		PLEDGE_AUDIO },
384 	{ "bpf",		PLEDGE_BPF },
385 	{ "chown",		PLEDGE_CHOWN | PLEDGE_CHOWNUID },
386 	{ "cpath",		PLEDGE_CPATH },
387 	{ "disklabel",		PLEDGE_DISKLABEL },
388 	{ "dns",		PLEDGE_DNS },
389 	{ "dpath",		PLEDGE_DPATH },
390 	{ "drm",		PLEDGE_DRM },
391 	{ "error",		PLEDGE_ERROR },
392 	{ "exec",		PLEDGE_EXEC },
393 	{ "fattr",		PLEDGE_FATTR | PLEDGE_CHOWN },
394 	{ "flock",		PLEDGE_FLOCK },
395 	{ "getpw",		PLEDGE_GETPW },
396 	{ "id",			PLEDGE_ID },
397 	{ "inet",		PLEDGE_INET },
398 	{ "mcast",		PLEDGE_MCAST },
399 	{ "pf",			PLEDGE_PF },
400 	{ "proc",		PLEDGE_PROC },
401 	{ "prot_exec",		PLEDGE_PROTEXEC },
402 	{ "ps",			PLEDGE_PS },
403 	{ "recvfd",		PLEDGE_RECVFD },
404 	{ "route",		PLEDGE_ROUTE },
405 	{ "rpath",		PLEDGE_RPATH },
406 	{ "sendfd",		PLEDGE_SENDFD },
407 	{ "settime",		PLEDGE_SETTIME },
408 	{ "stdio",		PLEDGE_STDIO },
409 	{ "tape",		PLEDGE_TAPE },
410 	{ "tmppath",		PLEDGE_TMPPATH },
411 	{ "tty",		PLEDGE_TTY },
412 	{ "unix",		PLEDGE_UNIX },
413 	{ "unveil",		PLEDGE_UNVEIL },
414 	{ "video",		PLEDGE_VIDEO },
415 	{ "vminfo",		PLEDGE_VMINFO },
416 	{ "vmm",		PLEDGE_VMM },
417 	{ "wpath",		PLEDGE_WPATH },
418 	{ "wroute",		PLEDGE_WROUTE },
419 };
420 
421 int
422 parsepledges(struct proc *p, const char *kname, const char *promises, u_int64_t *fp)
423 {
424 	size_t rbuflen;
425 	char *rbuf, *rp, *pn;
426 	u_int64_t flags = 0, f;
427 	int error;
428 
429 	rbuf = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
430 	error = copyinstr(promises, rbuf, MAXPATHLEN,
431 	    &rbuflen);
432 	if (error) {
433 		free(rbuf, M_TEMP, MAXPATHLEN);
434 		return (error);
435 	}
436 #ifdef KTRACE
437 	if (KTRPOINT(p, KTR_STRUCT))
438 		ktrstruct(p, kname, rbuf, rbuflen-1);
439 #endif
440 
441 	for (rp = rbuf; rp && *rp; rp = pn) {
442 		pn = strchr(rp, ' ');	/* find terminator */
443 		if (pn) {
444 			while (*pn == ' ')
445 				*pn++ = '\0';
446 		}
447 		if ((f = pledgereq_flags(rp)) == 0) {
448 			free(rbuf, M_TEMP, MAXPATHLEN);
449 			return (EINVAL);
450 		}
451 		flags |= f;
452 	}
453 	free(rbuf, M_TEMP, MAXPATHLEN);
454 	*fp = flags;
455 	return 0;
456 }
457 
458 int
459 sys_pledge(struct proc *p, void *v, register_t *retval)
460 {
461 	struct sys_pledge_args /* {
462 		syscallarg(const char *)promises;
463 		syscallarg(const char *)execpromises;
464 	} */	*uap = v;
465 	struct process *pr = p->p_p;
466 	uint64_t promises, execpromises;
467 	int error;
468 
469 	if (SCARG(uap, promises)) {
470 		error = parsepledges(p, "pledgereq",
471 		    SCARG(uap, promises), &promises);
472 		if (error)
473 			return (error);
474 
475 		/* In "error" mode, ignore promise increase requests,
476 		 * but accept promise decrease requests */
477 		if (ISSET(pr->ps_flags, PS_PLEDGE) &&
478 		    (pr->ps_pledge & PLEDGE_ERROR))
479 			promises &= (pr->ps_pledge & PLEDGE_USERSET);
480 
481 		/* Only permit reductions */
482 		if (ISSET(pr->ps_flags, PS_PLEDGE) &&
483 		    (((promises | pr->ps_pledge) != pr->ps_pledge)))
484 			return (EPERM);
485 	}
486 	if (SCARG(uap, execpromises)) {
487 		error = parsepledges(p, "pledgeexecreq",
488 		    SCARG(uap, execpromises), &execpromises);
489 		if (error)
490 			return (error);
491 
492 		/* Only permit reductions */
493 		if (ISSET(pr->ps_flags, PS_EXECPLEDGE) &&
494 		    (((execpromises | pr->ps_execpledge) != pr->ps_execpledge)))
495 			return (EPERM);
496 	}
497 
498 	if (SCARG(uap, promises)) {
499 		pr->ps_pledge = promises;
500 		atomic_setbits_int(&pr->ps_flags, PS_PLEDGE);
501 		/*
502 		 * Kill off unveil and drop unveil vnode refs if we no
503 		 * longer are holding any path-accessing pledge
504 		 */
505 		if ((pr->ps_pledge & (PLEDGE_RPATH | PLEDGE_WPATH |
506 		    PLEDGE_CPATH | PLEDGE_DPATH | PLEDGE_TMPPATH | PLEDGE_EXEC |
507 		    PLEDGE_UNIX | PLEDGE_UNVEIL)) == 0)
508 			unveil_destroy(pr);
509 	}
510 	if (SCARG(uap, execpromises)) {
511 		pr->ps_execpledge = execpromises;
512 		atomic_setbits_int(&pr->ps_flags, PS_EXECPLEDGE);
513 	}
514 	return (0);
515 }
516 
517 int
518 pledge_syscall(struct proc *p, int code, uint64_t *tval)
519 {
520 	p->p_pledge_syscall = code;
521 	*tval = 0;
522 
523 	if (code < 0 || code > SYS_MAXSYSCALL - 1)
524 		return (EINVAL);
525 
526 	if (pledge_syscalls[code] == PLEDGE_ALWAYS)
527 		return (0);
528 
529 	if (p->p_p->ps_pledge & pledge_syscalls[code])
530 		return (0);
531 
532 	*tval = pledge_syscalls[code];
533 	return (EPERM);
534 }
535 
536 int
537 pledge_fail(struct proc *p, int error, uint64_t code)
538 {
539 	const char *codes = "";
540 	int i;
541 
542 	/* Print first matching pledge */
543 	for (i = 0; code && pledgenames[i].bits != 0; i++)
544 		if (pledgenames[i].bits & code) {
545 			codes = pledgenames[i].name;
546 			break;
547 		}
548 #ifdef KTRACE
549 	if (KTRPOINT(p, KTR_PLEDGE))
550 		ktrpledge(p, error, code, p->p_pledge_syscall);
551 #endif
552 	if (p->p_p->ps_pledge & PLEDGE_ERROR)
553 		return (ENOSYS);
554 
555 	KERNEL_LOCK();
556 	log(LOG_ERR, "%s[%d]: pledge \"%s\", syscall %d\n",
557 	    p->p_p->ps_comm, p->p_p->ps_pid, codes, p->p_pledge_syscall);
558 	p->p_p->ps_acflag |= APLEDGE;
559 
560 	/* Stop threads immediately, because this process is suspect */
561 	if (P_HASSIBLING(p))
562 		single_thread_set(p, SINGLE_SUSPEND, 1);
563 
564 	/* Send uncatchable SIGABRT for coredump */
565 	sigabort(p);
566 
567 	p->p_p->ps_pledge = 0;		/* Disable all PLEDGE_ flags */
568 	KERNEL_UNLOCK();
569 	return (error);
570 }
571 
572 /*
573  * Need to make it more obvious that one cannot get through here
574  * without the right flags set
575  */
576 int
577 pledge_namei(struct proc *p, struct nameidata *ni, char *origpath)
578 {
579 	char path[PATH_MAX];
580 	uint64_t pledge;
581 	int error;
582 
583 	if ((p->p_p->ps_flags & PS_PLEDGE) == 0 ||
584 	    (p->p_p->ps_flags & PS_COREDUMP))
585 		return (0);
586 	pledge = p->p_p->ps_pledge;
587 
588 	if (ni->ni_pledge == 0)
589 		panic("pledge_namei: ni_pledge");
590 
591 	/*
592 	 * We set the BYPASSUNVEIL flag to skip unveil checks
593 	 * as necessary
594 	 */
595 
596 	/* Doing a permitted execve() */
597 	if ((ni->ni_pledge & PLEDGE_EXEC) &&
598 	    (pledge & PLEDGE_EXEC))
599 		return (0);
600 
601 	error = canonpath(origpath, path, sizeof(path));
602 	if (error)
603 		return (error);
604 
605 	/* Detect what looks like a mkstemp(3) family operation */
606 	if ((pledge & PLEDGE_TMPPATH) &&
607 	    (p->p_pledge_syscall == SYS_open) &&
608 	    (ni->ni_pledge & PLEDGE_CPATH) &&
609 	    strncmp(path, "/tmp/", sizeof("/tmp/") - 1) == 0) {
610 		ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
611 		return (0);
612 	}
613 
614 	/* Allow unlinking of a mkstemp(3) file...
615 	 * Good opportunity for strict checks here.
616 	 */
617 	if ((pledge & PLEDGE_TMPPATH) &&
618 	    (p->p_pledge_syscall == SYS_unlink) &&
619 	    strncmp(path, "/tmp/", sizeof("/tmp/") - 1) == 0) {
620 		ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
621 		return (0);
622 	}
623 
624 	/* Whitelisted paths */
625 	switch (p->p_pledge_syscall) {
626 	case SYS_access:
627 		/* tzset() needs this. */
628 		if (ni->ni_pledge == PLEDGE_RPATH &&
629 		    strcmp(path, "/etc/localtime") == 0) {
630 			ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
631 			return (0);
632 		}
633 
634 		/* when avoiding YP mode, getpw* functions touch this */
635 		if (ni->ni_pledge == PLEDGE_RPATH &&
636 		    strcmp(path, "/var/run/ypbind.lock") == 0) {
637 			if (pledge & PLEDGE_GETPW) {
638 				ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
639 				return (0);
640 			} else
641 				return (pledge_fail(p, error, PLEDGE_GETPW));
642 		}
643 		break;
644 	case SYS_open:
645 		/* daemon(3) or other such functions */
646 		if ((ni->ni_pledge & ~(PLEDGE_RPATH | PLEDGE_WPATH)) == 0 &&
647 		    strcmp(path, "/dev/null") == 0) {
648 			ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
649 			return (0);
650 		}
651 
652 		/* readpassphrase(3), getpass(3) */
653 		if ((pledge & PLEDGE_TTY) &&
654 		    (ni->ni_pledge & ~(PLEDGE_RPATH | PLEDGE_WPATH)) == 0 &&
655 		    strcmp(path, "/dev/tty") == 0) {
656 			ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
657 			return (0);
658 		}
659 
660 		/* getpw* and friends need a few files */
661 		if ((ni->ni_pledge == PLEDGE_RPATH) &&
662 		    (pledge & PLEDGE_GETPW)) {
663 			if (strcmp(path, "/etc/spwd.db") == 0)
664 				return (EPERM); /* don't call pledge_fail */
665 			if (strcmp(path, "/etc/pwd.db") == 0) {
666 				ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
667 				return (0);
668 			}
669 			if (strcmp(path, "/etc/group") == 0) {
670 				ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
671 				return (0);
672 			}
673 			if (strcmp(path, "/etc/netid") == 0) {
674 				ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
675 				return (0);
676 			}
677 		}
678 
679 		/* DNS needs /etc/{resolv.conf,hosts,services,protocols}. */
680 		if ((ni->ni_pledge == PLEDGE_RPATH) &&
681 		    (pledge & PLEDGE_DNS)) {
682 			if (strcmp(path, "/etc/resolv.conf") == 0) {
683 				ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
684 				return (0);
685 			}
686 			if (strcmp(path, "/etc/hosts") == 0) {
687 				ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
688 				return (0);
689 			}
690 			if (strcmp(path, "/etc/services") == 0) {
691 				ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
692 				return (0);
693 			}
694 			if (strcmp(path, "/etc/protocols") == 0) {
695 				ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
696 				return (0);
697 			}
698 		}
699 
700 		if ((ni->ni_pledge == PLEDGE_RPATH) &&
701 		    (pledge & PLEDGE_GETPW)) {
702 			if (strcmp(path, "/var/run/ypbind.lock") == 0) {
703 				/*
704 				 * XXX
705 				 * The current hack for YP support in "getpw"
706 				 * is to enable some "inet" features until
707 				 * next pledge call.  Setting a bit in ps_pledge
708 				 * is not safe with respect to multiple threads,
709 				 * a very different approach is needed.
710 				 */
711 				p->p_p->ps_pledge |= PLEDGE_YPACTIVE;
712 				ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
713 				return (0);
714 			}
715 			if (strncmp(path, "/var/yp/binding/",
716 			    sizeof("/var/yp/binding/") - 1) == 0) {
717 				ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
718 				return (0);
719 			}
720 		}
721 
722 		/* tzset() needs these. */
723 		if ((ni->ni_pledge == PLEDGE_RPATH) &&
724 		    strncmp(path, "/usr/share/zoneinfo/",
725 		    sizeof("/usr/share/zoneinfo/") - 1) == 0)  {
726 			ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
727 			return (0);
728 		}
729 		if ((ni->ni_pledge == PLEDGE_RPATH) &&
730 		    strcmp(path, "/etc/localtime") == 0) {
731 			ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
732 			return (0);
733 		}
734 
735 		break;
736 	case SYS_stat:
737 		/* DNS needs /etc/{resolv.conf,hosts}. */
738 		if ((ni->ni_pledge == PLEDGE_RPATH) &&
739 		    (pledge & PLEDGE_DNS)) {
740 			if (strcmp(path, "/etc/resolv.conf") == 0) {
741 				ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
742 				return (0);
743 			}
744 			if (strcmp(path, "/etc/hosts") == 0) {
745 				ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
746 				return (0);
747 			}
748 		}
749 		break;
750 	}
751 
752 	/*
753 	 * Ensure each flag of ni_pledge has counterpart allowing it in
754 	 * ps_pledge.
755 	 */
756 	if (ni->ni_pledge & ~pledge)
757 		return (pledge_fail(p, EPERM, (ni->ni_pledge & ~pledge)));
758 
759 	/* continue, and check unveil if present */
760 	return (0);
761 }
762 
763 /*
764  * Only allow reception of safe file descriptors.
765  */
766 int
767 pledge_recvfd(struct proc *p, struct file *fp)
768 {
769 	struct vnode *vp;
770 
771 	if ((p->p_p->ps_flags & PS_PLEDGE) == 0)
772 		return (0);
773 	if ((p->p_p->ps_pledge & PLEDGE_RECVFD) == 0)
774 		return pledge_fail(p, EPERM, PLEDGE_RECVFD);
775 
776 	switch (fp->f_type) {
777 	case DTYPE_SOCKET:
778 	case DTYPE_PIPE:
779 	case DTYPE_DMABUF:
780 	case DTYPE_SYNC:
781 		return (0);
782 	case DTYPE_VNODE:
783 		vp = fp->f_data;
784 
785 		if (vp->v_type != VDIR)
786 			return (0);
787 	}
788 	return pledge_fail(p, EINVAL, PLEDGE_RECVFD);
789 }
790 
791 /*
792  * Only allow sending of safe file descriptors.
793  */
794 int
795 pledge_sendfd(struct proc *p, struct file *fp)
796 {
797 	struct vnode *vp;
798 
799 	if ((p->p_p->ps_flags & PS_PLEDGE) == 0)
800 		return (0);
801 	if ((p->p_p->ps_pledge & PLEDGE_SENDFD) == 0)
802 		return pledge_fail(p, EPERM, PLEDGE_SENDFD);
803 
804 	switch (fp->f_type) {
805 	case DTYPE_SOCKET:
806 	case DTYPE_PIPE:
807 	case DTYPE_DMABUF:
808 	case DTYPE_SYNC:
809 		return (0);
810 	case DTYPE_VNODE:
811 		vp = fp->f_data;
812 
813 		if (vp->v_type != VDIR)
814 			return (0);
815 		break;
816 	}
817 	return pledge_fail(p, EINVAL, PLEDGE_SENDFD);
818 }
819 
820 int
821 pledge_sysctl(struct proc *p, int miblen, int *mib, void *new)
822 {
823 	char	buf[80];
824 	uint64_t pledge;
825 	int	i;
826 
827 	if ((p->p_p->ps_flags & PS_PLEDGE) == 0)
828 		return (0);
829 	pledge = p->p_p->ps_pledge;
830 
831 	if (new)
832 		return pledge_fail(p, EFAULT, 0);
833 
834 	/* routing table observation */
835 	if ((pledge & PLEDGE_ROUTE)) {
836 		if ((miblen == 6 || miblen == 7) &&
837 		    mib[0] == CTL_NET && mib[1] == PF_ROUTE &&
838 		    mib[2] == 0 &&
839 		    mib[4] == NET_RT_DUMP)
840 			return (0);
841 
842 		if (miblen == 6 &&
843 		    mib[0] == CTL_NET && mib[1] == PF_ROUTE &&
844 		    mib[2] == 0 &&
845 		    (mib[3] == 0 || mib[3] == AF_INET6 || mib[3] == AF_INET) &&
846 		    (mib[4] == NET_RT_TABLE || mib[4] == NET_RT_SOURCE))
847 			return (0);
848 
849 		if (miblen == 7 &&		/* exposes MACs */
850 		    mib[0] == CTL_NET && mib[1] == PF_ROUTE &&
851 		    mib[2] == 0 &&
852 		    (mib[3] == 0 || mib[3] == AF_INET6 || mib[3] == AF_INET) &&
853 		    mib[4] == NET_RT_FLAGS && mib[5] == RTF_LLINFO)
854 			return (0);
855 	}
856 
857 	if ((pledge & PLEDGE_WROUTE)) {
858 		if (miblen == 4 &&
859 		    mib[0] == CTL_NET && mib[1] == PF_INET6 &&
860 		    mib[2] == IPPROTO_IPV6 && mib[3] == IPV6CTL_SOIIKEY)
861 			return (0);
862 	}
863 
864 	if (pledge & (PLEDGE_PS | PLEDGE_VMINFO)) {
865 		if (miblen == 2 &&		/* kern.fscale */
866 		    mib[0] == CTL_KERN && mib[1] == KERN_FSCALE)
867 			return (0);
868 		if (miblen == 2 &&		/* kern.boottime */
869 		    mib[0] == CTL_KERN && mib[1] == KERN_BOOTTIME)
870 			return (0);
871 		if (miblen == 2 &&		/* kern.consdev */
872 		    mib[0] == CTL_KERN && mib[1] == KERN_CONSDEV)
873 			return (0);
874 		if (miblen == 2 &&			/* kern.cptime */
875 		    mib[0] == CTL_KERN && mib[1] == KERN_CPTIME)
876 			return (0);
877 		if (miblen == 3 &&			/* kern.cptime2 */
878 		    mib[0] == CTL_KERN && mib[1] == KERN_CPTIME2)
879 			return (0);
880 		if (miblen == 3 &&			/* kern.cpustats */
881 		    mib[0] == CTL_KERN && mib[1] == KERN_CPUSTATS)
882 			return (0);
883 	}
884 
885 	if ((pledge & PLEDGE_PS)) {
886 		if (miblen == 4 &&		/* kern.procargs.* */
887 		    mib[0] == CTL_KERN && mib[1] == KERN_PROC_ARGS &&
888 		    (mib[3] == KERN_PROC_ARGV || mib[3] == KERN_PROC_ENV))
889 			return (0);
890 		if (miblen == 6 &&		/* kern.proc.* */
891 		    mib[0] == CTL_KERN && mib[1] == KERN_PROC)
892 			return (0);
893 		if (miblen == 3 &&		/* kern.proc_cwd.* */
894 		    mib[0] == CTL_KERN && mib[1] == KERN_PROC_CWD)
895 			return (0);
896 		if (miblen == 2 &&		/* kern.ccpu */
897 		    mib[0] == CTL_KERN && mib[1] == KERN_CCPU)
898 			return (0);
899 		if (miblen == 2 &&		/* vm.maxslp */
900 		    mib[0] == CTL_VM && mib[1] == VM_MAXSLP)
901 			return (0);
902 	}
903 
904 	if ((pledge & PLEDGE_VMINFO)) {
905 		if (miblen == 2 &&		/* vm.uvmexp */
906 		    mib[0] == CTL_VM && mib[1] == VM_UVMEXP)
907 			return (0);
908 		if (miblen == 3 &&		/* vfs.generic.bcachestat */
909 		    mib[0] == CTL_VFS && mib[1] == VFS_GENERIC &&
910 		    mib[2] == VFS_BCACHESTAT)
911 			return (0);
912 	}
913 
914 	if ((pledge & (PLEDGE_INET | PLEDGE_UNIX))) {
915 		if (miblen == 2 &&		/* kern.somaxconn */
916 		    mib[0] == CTL_KERN && mib[1] == KERN_SOMAXCONN)
917 			return (0);
918 	}
919 
920 	if ((pledge & (PLEDGE_ROUTE | PLEDGE_INET | PLEDGE_DNS))) {
921 		if (miblen == 6 &&		/* getifaddrs() */
922 		    mib[0] == CTL_NET && mib[1] == PF_ROUTE &&
923 		    mib[2] == 0 &&
924 		    (mib[3] == 0 || mib[3] == AF_INET6 || mib[3] == AF_INET) &&
925 		    mib[4] == NET_RT_IFLIST)
926 			return (0);
927 	}
928 
929 	if ((pledge & PLEDGE_DISKLABEL)) {
930 		if (miblen == 2 &&		/* kern.rawpartition */
931 		    mib[0] == CTL_KERN &&
932 		    mib[1] == KERN_RAWPARTITION)
933 			return (0);
934 		if (miblen == 2 &&		/* kern.maxpartitions */
935 		    mib[0] == CTL_KERN &&
936 		    mib[1] == KERN_MAXPARTITIONS)
937 			return (0);
938 #ifdef CPU_CHR2BLK
939 		if (miblen == 3 &&		/* machdep.chr2blk */
940 		    mib[0] == CTL_MACHDEP &&
941 		    mib[1] == CPU_CHR2BLK)
942 			return (0);
943 #endif /* CPU_CHR2BLK */
944 	}
945 
946 	if (miblen >= 3 &&			/* ntpd(8) to read sensors */
947 	    mib[0] == CTL_HW && mib[1] == HW_SENSORS)
948 		return (0);
949 
950 	if (miblen == 6 &&		/* if_nameindex() */
951 	    mib[0] == CTL_NET && mib[1] == PF_ROUTE &&
952 	    mib[2] == 0 && mib[3] == 0 && mib[4] == NET_RT_IFNAMES)
953 		return (0);
954 
955 	if (miblen == 2) {
956 		switch (mib[0]) {
957 		case CTL_KERN:
958 			switch (mib[1]) {
959 			case KERN_DOMAINNAME:	/* getdomainname() */
960 			case KERN_HOSTNAME:	/* gethostname() */
961 			case KERN_OSTYPE:	/* uname() */
962 			case KERN_OSRELEASE:	/* uname() */
963 			case KERN_OSVERSION:	/* uname() */
964 			case KERN_VERSION:	/* uname() */
965 			case KERN_CLOCKRATE:	/* kern.clockrate */
966 			case KERN_ARGMAX:	/* kern.argmax */
967 			case KERN_NGROUPS:	/* kern.ngroups */
968 			case KERN_SYSVSHM:	/* kern.sysvshm */
969 			case KERN_POSIX1:	/* kern.posix1version */
970 				return (0);
971 			}
972 			break;
973 		case CTL_HW:
974 			switch (mib[1]) {
975 			case HW_MACHINE: 	/* uname() */
976 			case HW_PAGESIZE: 	/* getpagesize() */
977 			case HW_PHYSMEM64:	/* hw.physmem */
978 			case HW_NCPU:		/* hw.ncpu */
979 			case HW_NCPUONLINE:	/* hw.ncpuonline */
980 				return (0);
981 			}
982 			break;
983 		case CTL_VM:
984 			switch (mib[1]) {
985 			case VM_PSSTRINGS:	/* setproctitle() */
986 			case VM_LOADAVG:	/* vm.loadavg / getloadavg(3) */
987 			case VM_MALLOC_CONF:	/* vm.malloc_conf */
988 				return (0);
989 			}
990 			break;
991 		default:
992 			break;
993 		}
994 	}
995 
996 #ifdef CPU_SSE
997 	if (miblen == 2 &&		/* i386 libm tests for SSE */
998 	    mib[0] == CTL_MACHDEP && mib[1] == CPU_SSE)
999 		return (0);
1000 #endif /* CPU_SSE */
1001 
1002 #ifdef CPU_ID_AA64ISAR0
1003 	if (miblen == 2 &&		/* arm64 libcrypto inspects CPU features */
1004 	    mib[0] == CTL_MACHDEP && mib[1] == CPU_ID_AA64ISAR0)
1005 		return (0);
1006 #endif /* CPU_ID_AA64ISAR0 */
1007 
1008 	snprintf(buf, sizeof(buf), "%s(%d): pledge sysctl %d:",
1009 	    p->p_p->ps_comm, p->p_p->ps_pid, miblen);
1010 	for (i = 0; i < miblen; i++) {
1011 		char *p = buf + strlen(buf);
1012 		snprintf(p, sizeof(buf) - (p - buf), " %d", mib[i]);
1013 	}
1014 	log(LOG_ERR, "%s\n", buf);
1015 
1016 	return pledge_fail(p, EINVAL, 0);
1017 }
1018 
1019 int
1020 pledge_chown(struct proc *p, uid_t uid, gid_t gid)
1021 {
1022 	if ((p->p_p->ps_flags & PS_PLEDGE) == 0)
1023 		return (0);
1024 
1025 	if (p->p_p->ps_pledge & PLEDGE_CHOWNUID)
1026 		return (0);
1027 
1028 	if (uid != -1 && uid != p->p_ucred->cr_uid)
1029 		return (EPERM);
1030 	if (gid != -1 && !groupmember(gid, p->p_ucred))
1031 		return (EPERM);
1032 	return (0);
1033 }
1034 
1035 int
1036 pledge_adjtime(struct proc *p, const void *v)
1037 {
1038 	const struct timeval *delta = v;
1039 
1040 	if ((p->p_p->ps_flags & PS_PLEDGE) == 0)
1041 		return (0);
1042 
1043 	if ((p->p_p->ps_pledge & PLEDGE_SETTIME))
1044 		return (0);
1045 	if (delta)
1046 		return (EPERM);
1047 	return (0);
1048 }
1049 
1050 int
1051 pledge_sendit(struct proc *p, const void *to)
1052 {
1053 	if ((p->p_p->ps_flags & PS_PLEDGE) == 0)
1054 		return (0);
1055 
1056 	if ((p->p_p->ps_pledge & (PLEDGE_INET | PLEDGE_UNIX | PLEDGE_DNS | PLEDGE_YPACTIVE)))
1057 		return (0);		/* may use address */
1058 	if (to == NULL)
1059 		return (0);		/* behaves just like write */
1060 	return pledge_fail(p, EPERM, PLEDGE_INET);
1061 }
1062 
1063 int
1064 pledge_ioctl(struct proc *p, long com, struct file *fp)
1065 {
1066 	struct vnode *vp = NULL;
1067 	int error = EPERM;
1068 	uint64_t pledge;
1069 
1070 	if ((p->p_p->ps_flags & PS_PLEDGE) == 0)
1071 		return (0);
1072 	pledge = p->p_p->ps_pledge;
1073 
1074 	/*
1075 	 * The ioctl's which are always allowed.
1076 	 */
1077 	switch (com) {
1078 	case FIONREAD:
1079 	case FIONBIO:
1080 	case FIOCLEX:
1081 	case FIONCLEX:
1082 		return (0);
1083 	}
1084 
1085 	/* fp != NULL was already checked */
1086 	if (fp->f_type == DTYPE_VNODE) {
1087 		vp = fp->f_data;
1088 		if (vp->v_type == VBAD)
1089 			return (ENOTTY);
1090 	}
1091 
1092 	if ((pledge & PLEDGE_INET)) {
1093 		switch (com) {
1094 		case SIOCATMARK:
1095 		case SIOCGIFGROUP:
1096 			if (fp->f_type == DTYPE_SOCKET)
1097 				return (0);
1098 			break;
1099 		}
1100 	}
1101 
1102 #if NBPFILTER > 0
1103 	if ((pledge & PLEDGE_BPF)) {
1104 		switch (com) {
1105 		case BIOCGSTATS:	/* bpf: tcpdump privsep on ^C */
1106 			if (fp->f_type == DTYPE_VNODE &&
1107 			    fp->f_ops->fo_ioctl == vn_ioctl &&
1108 			    vp->v_type == VCHR &&
1109 			    cdevsw[major(vp->v_rdev)].d_open == bpfopen)
1110 				return (0);
1111 			break;
1112 		}
1113 	}
1114 #endif /* NBPFILTER > 0 */
1115 
1116 	if ((pledge & PLEDGE_TAPE)) {
1117 		switch (com) {
1118 		case MTIOCGET:
1119 		case MTIOCTOP:
1120 			/* for pax(1) and such, checking tapes... */
1121 			if (fp->f_type == DTYPE_VNODE &&
1122 			    vp->v_type == VCHR) {
1123 				if (vp->v_flag & VISTTY)
1124 					return (ENOTTY);
1125 				else
1126 					return (0);
1127 			}
1128 			break;
1129 		}
1130 	}
1131 
1132 #if NDRM > 0
1133 	if ((pledge & PLEDGE_DRM)) {
1134 		if ((fp->f_type == DTYPE_VNODE) &&
1135 		    (vp->v_type == VCHR) &&
1136 		    (cdevsw[major(vp->v_rdev)].d_open == drmopen)) {
1137 			error = pledge_ioctl_drm(p, com, vp->v_rdev);
1138 			if (error == 0)
1139 				return 0;
1140 		}
1141 	}
1142 #endif /* NDRM > 0 */
1143 
1144 #if NAUDIO > 0
1145 	if ((pledge & PLEDGE_AUDIO)) {
1146 		switch (com) {
1147 		case AUDIO_GETPOS:
1148 		case AUDIO_GETPAR:
1149 		case AUDIO_SETPAR:
1150 		case AUDIO_START:
1151 		case AUDIO_STOP:
1152 		case AUDIO_MIXER_DEVINFO:
1153 		case AUDIO_MIXER_READ:
1154 		case AUDIO_MIXER_WRITE:
1155 			if (fp->f_type == DTYPE_VNODE &&
1156 			    vp->v_type == VCHR &&
1157 			    cdevsw[major(vp->v_rdev)].d_open == audioopen)
1158 				return (0);
1159 		}
1160 	}
1161 #endif /* NAUDIO > 0 */
1162 
1163 	if ((pledge & PLEDGE_DISKLABEL)) {
1164 		switch (com) {
1165 		case DIOCGDINFO:
1166 		case DIOCGPDINFO:
1167 		case DIOCRLDINFO:
1168 		case DIOCWDINFO:
1169 		case BIOCDISK:
1170 		case BIOCINQ:
1171 		case BIOCINSTALLBOOT:
1172 		case BIOCVOL:
1173 			if (fp->f_type == DTYPE_VNODE &&
1174 			    ((vp->v_type == VCHR &&
1175 			    cdevsw[major(vp->v_rdev)].d_type == D_DISK) ||
1176 			    (vp->v_type == VBLK &&
1177 			    bdevsw[major(vp->v_rdev)].d_type == D_DISK)))
1178 				return (0);
1179 			break;
1180 		case DIOCMAP:
1181 			if (fp->f_type == DTYPE_VNODE &&
1182 			    vp->v_type == VCHR &&
1183 			    cdevsw[major(vp->v_rdev)].d_ioctl == diskmapioctl)
1184 				return (0);
1185 			break;
1186 		}
1187 	}
1188 
1189 #if NVIDEO > 0
1190 	if ((pledge & PLEDGE_VIDEO)) {
1191 		switch (com) {
1192 		case VIDIOC_QUERYCAP:
1193 		case VIDIOC_TRY_FMT:
1194 		case VIDIOC_ENUM_FMT:
1195 		case VIDIOC_S_FMT:
1196 		case VIDIOC_QUERYCTRL:
1197 		case VIDIOC_G_CTRL:
1198 		case VIDIOC_S_CTRL:
1199 		case VIDIOC_G_PARM:
1200 		case VIDIOC_S_PARM:
1201 		case VIDIOC_REQBUFS:
1202 		case VIDIOC_QBUF:
1203 		case VIDIOC_DQBUF:
1204 		case VIDIOC_QUERYBUF:
1205 		case VIDIOC_STREAMON:
1206 		case VIDIOC_STREAMOFF:
1207 		case VIDIOC_ENUM_FRAMESIZES:
1208 		case VIDIOC_ENUM_FRAMEINTERVALS:
1209 		case VIDIOC_DQEVENT:
1210 		case VIDIOC_ENCODER_CMD:
1211 		case VIDIOC_EXPBUF:
1212 		case VIDIOC_G_CROP:
1213 		case VIDIOC_G_EXT_CTRLS:
1214 		case VIDIOC_G_FMT:
1215 		case VIDIOC_G_SELECTION:
1216 		case VIDIOC_QUERYMENU:
1217 		case VIDIOC_SUBSCRIBE_EVENT:
1218 		case VIDIOC_S_EXT_CTRLS:
1219 		case VIDIOC_S_SELECTION:
1220 		case VIDIOC_TRY_DECODER_CMD:
1221 		case VIDIOC_TRY_ENCODER_CMD:
1222 			if (fp->f_type == DTYPE_VNODE &&
1223 			    vp->v_type == VCHR &&
1224 			    cdevsw[major(vp->v_rdev)].d_open == videoopen)
1225 				return (0);
1226 			break;
1227 		}
1228 	}
1229 #endif
1230 
1231 #if NPF > 0
1232 	if ((pledge & PLEDGE_PF)) {
1233 		switch (com) {
1234 		case DIOCADDRULE:
1235 		case DIOCGETSTATUS:
1236 		case DIOCNATLOOK:
1237 		case DIOCRADDTABLES:
1238 		case DIOCRCLRADDRS:
1239 		case DIOCRCLRTABLES:
1240 		case DIOCRCLRTSTATS:
1241 		case DIOCRGETTSTATS:
1242 		case DIOCRSETADDRS:
1243 		case DIOCXBEGIN:
1244 		case DIOCXCOMMIT:
1245 		case DIOCKILLSRCNODES:
1246 			if ((fp->f_type == DTYPE_VNODE) &&
1247 			    (vp->v_type == VCHR) &&
1248 			    (cdevsw[major(vp->v_rdev)].d_open == pfopen))
1249 				return (0);
1250 			break;
1251 		}
1252 	}
1253 #endif
1254 
1255 	if ((pledge & PLEDGE_TTY)) {
1256 		switch (com) {
1257 #if NPTY > 0
1258 		case PTMGET:
1259 			if ((pledge & PLEDGE_RPATH) == 0)
1260 				break;
1261 			if ((pledge & PLEDGE_WPATH) == 0)
1262 				break;
1263 			if (fp->f_type != DTYPE_VNODE || vp->v_type != VCHR)
1264 				break;
1265 			if (cdevsw[major(vp->v_rdev)].d_open != ptmopen)
1266 				break;
1267 			return (0);
1268 		case TIOCUCNTL:		/* vmd */
1269 			if ((pledge & PLEDGE_RPATH) == 0)
1270 				break;
1271 			if ((pledge & PLEDGE_WPATH) == 0)
1272 				break;
1273 			if (cdevsw[major(vp->v_rdev)].d_open != ptcopen)
1274 				break;
1275 			return (0);
1276 #endif /* NPTY > 0 */
1277 		case TIOCSPGRP:
1278 			if ((pledge & PLEDGE_PROC) == 0)
1279 				break;
1280 			/* FALLTHROUGH */
1281 		case TIOCFLUSH:		/* getty, telnet */
1282 		case TIOCSTART:		/* emacs, etc */
1283 		case TIOCGPGRP:
1284 		case TIOCGETA:
1285 		case TIOCGWINSZ:	/* ENOTTY return for non-tty */
1286 		case TIOCSTAT:		/* csh */
1287 			if (fp->f_type == DTYPE_VNODE && (vp->v_flag & VISTTY))
1288 				return (0);
1289 			return (ENOTTY);
1290 		case TIOCSWINSZ:
1291 		case TIOCEXT:		/* mail, libedit .. */
1292 		case TIOCCBRK:		/* cu */
1293 		case TIOCSBRK:		/* cu */
1294 		case TIOCCDTR:		/* cu */
1295 		case TIOCSDTR:		/* cu */
1296 		case TIOCEXCL:		/* cu */
1297 		case TIOCSETA:		/* cu, ... */
1298 		case TIOCSETAW:		/* cu, ... */
1299 		case TIOCSETAF:		/* tcsetattr TCSAFLUSH, script */
1300 		case TIOCSCTTY:		/* forkpty(3), login_tty(3), ... */
1301 			if (fp->f_type == DTYPE_VNODE && (vp->v_flag & VISTTY))
1302 				return (0);
1303 			break;
1304 		}
1305 	}
1306 
1307 	if ((pledge & PLEDGE_ROUTE)) {
1308 		switch (com) {
1309 		case SIOCGIFADDR:
1310 		case SIOCGIFAFLAG_IN6:
1311 		case SIOCGIFALIFETIME_IN6:
1312 		case SIOCGIFDESCR:
1313 		case SIOCGIFFLAGS:
1314 		case SIOCGIFMETRIC:
1315 		case SIOCGIFGMEMB:
1316 		case SIOCGIFRDOMAIN:
1317 		case SIOCGIFDSTADDR_IN6:
1318 		case SIOCGIFNETMASK_IN6:
1319 		case SIOCGIFXFLAGS:
1320 		case SIOCGNBRINFO_IN6:
1321 		case SIOCGIFINFO_IN6:
1322 		case SIOCGIFMEDIA:
1323 			if (fp->f_type == DTYPE_SOCKET)
1324 				return (0);
1325 			break;
1326 		}
1327 	}
1328 
1329 	if ((pledge & PLEDGE_WROUTE)) {
1330 		switch (com) {
1331 		case SIOCAIFADDR:
1332 		case SIOCDIFADDR:
1333 		case SIOCAIFADDR_IN6:
1334 		case SIOCDIFADDR_IN6:
1335 			if (fp->f_type == DTYPE_SOCKET)
1336 				return (0);
1337 			break;
1338 		case SIOCSIFMTU:
1339 			if (fp->f_type == DTYPE_SOCKET)
1340 				return (0);
1341 			break;
1342 		}
1343 	}
1344 
1345 #if NVMM > 0
1346 	if ((pledge & PLEDGE_VMM)) {
1347 		if ((fp->f_type == DTYPE_VNODE) &&
1348 		    (vp->v_type == VCHR) &&
1349 		    (cdevsw[major(vp->v_rdev)].d_open == vmmopen)) {
1350 			error = pledge_ioctl_vmm(p, com);
1351 			if (error == 0)
1352 				return 0;
1353 		}
1354 	}
1355 #endif
1356 
1357 	return pledge_fail(p, error, PLEDGE_TTY);
1358 }
1359 
1360 int
1361 pledge_sockopt(struct proc *p, int set, int level, int optname)
1362 {
1363 	uint64_t pledge;
1364 
1365 	if ((p->p_p->ps_flags & PS_PLEDGE) == 0)
1366 		return (0);
1367 	pledge = p->p_p->ps_pledge;
1368 
1369 	/* Always allow these, which are too common to reject */
1370 	switch (level) {
1371 	case SOL_SOCKET:
1372 		switch (optname) {
1373 		case SO_RCVBUF:
1374 		case SO_ERROR:
1375 			return 0;
1376 		}
1377 		break;
1378 	}
1379 
1380 	if ((pledge & PLEDGE_WROUTE)) {
1381 		switch (level) {
1382 		case SOL_SOCKET:
1383 			switch (optname) {
1384 			case SO_RTABLE:
1385 				return (0);
1386 			}
1387 		}
1388 	}
1389 
1390 	if ((pledge & (PLEDGE_INET|PLEDGE_UNIX|PLEDGE_DNS|PLEDGE_YPACTIVE)) == 0)
1391 		return pledge_fail(p, EPERM, PLEDGE_INET);
1392 	/* In use by some service libraries */
1393 	switch (level) {
1394 	case SOL_SOCKET:
1395 		switch (optname) {
1396 		case SO_TIMESTAMP:
1397 			return 0;
1398 		}
1399 		break;
1400 	}
1401 
1402 	/* DNS resolver may do these requests */
1403 	if ((pledge & PLEDGE_DNS)) {
1404 		switch (level) {
1405 		case IPPROTO_IPV6:
1406 			switch (optname) {
1407 			case IPV6_RECVPKTINFO:
1408 			case IPV6_USE_MIN_MTU:
1409 				return (0);
1410 			}
1411 		}
1412 	}
1413 
1414 	/* YP may do these requests */
1415 	if (pledge & PLEDGE_YPACTIVE) {
1416 		switch (level) {
1417 		case IPPROTO_IP:
1418 			switch (optname) {
1419 			case IP_PORTRANGE:
1420 				return (0);
1421 			}
1422 			break;
1423 
1424 		case IPPROTO_IPV6:
1425 			switch (optname) {
1426 			case IPV6_PORTRANGE:
1427 				return (0);
1428 			}
1429 			break;
1430 		}
1431 	}
1432 
1433 	if ((pledge & (PLEDGE_INET|PLEDGE_UNIX)) == 0)
1434 		return pledge_fail(p, EPERM, PLEDGE_INET);
1435 	switch (level) {
1436 	case SOL_SOCKET:
1437 		switch (optname) {
1438 		case SO_RTABLE:
1439 			return pledge_fail(p, EINVAL, PLEDGE_WROUTE);
1440 		}
1441 		return (0);
1442 	}
1443 
1444 	if ((pledge & PLEDGE_INET) == 0)
1445 		return pledge_fail(p, EPERM, PLEDGE_INET);
1446 	switch (level) {
1447 	case IPPROTO_TCP:
1448 		switch (optname) {
1449 		case TCP_NODELAY:
1450 		case TCP_MD5SIG:
1451 		case TCP_SACK_ENABLE:
1452 		case TCP_MAXSEG:
1453 		case TCP_NOPUSH:
1454 			return (0);
1455 		}
1456 		break;
1457 	case IPPROTO_IP:
1458 		switch (optname) {
1459 		case IP_OPTIONS:
1460 			if (!set)
1461 				return (0);
1462 			break;
1463 		case IP_TOS:
1464 		case IP_TTL:
1465 		case IP_MINTTL:
1466 		case IP_IPDEFTTL:
1467 		case IP_PORTRANGE:
1468 		case IP_RECVDSTADDR:
1469 		case IP_RECVDSTPORT:
1470 			return (0);
1471 		case IP_MULTICAST_IF:
1472 		case IP_MULTICAST_TTL:
1473 		case IP_MULTICAST_LOOP:
1474 		case IP_ADD_MEMBERSHIP:
1475 		case IP_DROP_MEMBERSHIP:
1476 			if (pledge & PLEDGE_MCAST)
1477 				return (0);
1478 			break;
1479 		}
1480 		break;
1481 	case IPPROTO_ICMP:
1482 		break;
1483 	case IPPROTO_IPV6:
1484 		switch (optname) {
1485 		case IPV6_TCLASS:
1486 		case IPV6_UNICAST_HOPS:
1487 		case IPV6_MINHOPCOUNT:
1488 		case IPV6_RECVHOPLIMIT:
1489 		case IPV6_PORTRANGE:
1490 		case IPV6_RECVPKTINFO:
1491 		case IPV6_RECVDSTPORT:
1492 		case IPV6_V6ONLY:
1493 			return (0);
1494 		case IPV6_MULTICAST_IF:
1495 		case IPV6_MULTICAST_HOPS:
1496 		case IPV6_MULTICAST_LOOP:
1497 		case IPV6_JOIN_GROUP:
1498 		case IPV6_LEAVE_GROUP:
1499 			if (pledge & PLEDGE_MCAST)
1500 				return (0);
1501 			break;
1502 		}
1503 		break;
1504 	case IPPROTO_ICMPV6:
1505 		break;
1506 	}
1507 	return pledge_fail(p, EPERM, PLEDGE_INET);
1508 }
1509 
1510 int
1511 pledge_socket(struct proc *p, int domain, unsigned int state)
1512 {
1513 	uint64_t pledge;
1514 
1515 	if (!ISSET(p->p_p->ps_flags, PS_PLEDGE))
1516 		return 0;
1517 	pledge = p->p_p->ps_pledge;
1518 
1519 	if (ISSET(state, SS_DNS)) {
1520 		if (ISSET(pledge, PLEDGE_DNS))
1521 			return 0;
1522 		return pledge_fail(p, EPERM, PLEDGE_DNS);
1523 	}
1524 
1525 	switch (domain) {
1526 	case -1:		/* accept on any domain */
1527 		return (0);
1528 	case AF_INET:
1529 	case AF_INET6:
1530 		if (ISSET(pledge, PLEDGE_INET) ||
1531 		    ISSET(pledge, PLEDGE_YPACTIVE))
1532 			return 0;
1533 		return pledge_fail(p, EPERM, PLEDGE_INET);
1534 
1535 	case AF_UNIX:
1536 		if (ISSET(pledge, PLEDGE_UNIX))
1537 			return 0;
1538 		return pledge_fail(p, EPERM, PLEDGE_UNIX);
1539 	}
1540 
1541 	return pledge_fail(p, EINVAL, PLEDGE_INET);
1542 }
1543 
1544 int
1545 pledge_flock(struct proc *p)
1546 {
1547 	if ((p->p_p->ps_flags & PS_PLEDGE) == 0)
1548 		return (0);
1549 
1550 	if ((p->p_p->ps_pledge & PLEDGE_FLOCK))
1551 		return (0);
1552 	return (pledge_fail(p, EPERM, PLEDGE_FLOCK));
1553 }
1554 
1555 int
1556 pledge_swapctl(struct proc *p)
1557 {
1558 	if ((p->p_p->ps_flags & PS_PLEDGE) == 0)
1559 		return (0);
1560 	return (EPERM);
1561 }
1562 
1563 /* bsearch over pledgereq. return flags value if found, 0 else */
1564 uint64_t
1565 pledgereq_flags(const char *req_name)
1566 {
1567 	int base = 0, cmp, i, lim;
1568 
1569 	for (lim = nitems(pledgereq); lim != 0; lim >>= 1) {
1570 		i = base + (lim >> 1);
1571 		cmp = strcmp(req_name, pledgereq[i].name);
1572 		if (cmp == 0)
1573 			return (pledgereq[i].flags);
1574 		if (cmp > 0) { /* not found before, move right */
1575 			base = i + 1;
1576 			lim--;
1577 		} /* else move left */
1578 	}
1579 	return (0);
1580 }
1581 
1582 int
1583 pledge_fcntl(struct proc *p, int cmd)
1584 {
1585 	if ((p->p_p->ps_flags & PS_PLEDGE) == 0)
1586 		return (0);
1587 	if ((p->p_p->ps_pledge & PLEDGE_PROC) == 0 && cmd == F_SETOWN)
1588 		return pledge_fail(p, EPERM, PLEDGE_PROC);
1589 	return (0);
1590 }
1591 
1592 int
1593 pledge_kill(struct proc *p, pid_t pid)
1594 {
1595 	if ((p->p_p->ps_flags & PS_PLEDGE) == 0)
1596 		return 0;
1597 	if (p->p_p->ps_pledge & PLEDGE_PROC)
1598 		return 0;
1599 	if (pid == 0 || pid == p->p_p->ps_pid)
1600 		return 0;
1601 	return pledge_fail(p, EPERM, PLEDGE_PROC);
1602 }
1603 
1604 int
1605 pledge_protexec(struct proc *p, int prot)
1606 {
1607 	if ((p->p_p->ps_flags & PS_PLEDGE) == 0)
1608 		return 0;
1609 	/* Before kbind(2) call, ld.so and crt may create EXEC mappings */
1610 	if (p->p_p->ps_kbind_addr == 0 && p->p_p->ps_kbind_cookie == 0)
1611 		return 0;
1612 	if (!(p->p_p->ps_pledge & PLEDGE_PROTEXEC) && (prot & PROT_EXEC))
1613 		return pledge_fail(p, EPERM, PLEDGE_PROTEXEC);
1614 	return 0;
1615 }
1616 
1617 int
1618 canonpath(const char *input, char *buf, size_t bufsize)
1619 {
1620 	const char *p;
1621 	char *q;
1622 
1623 	/* can't canon relative paths, don't bother */
1624 	if (input[0] != '/') {
1625 		if (strlcpy(buf, input, bufsize) >= bufsize)
1626 			return ENAMETOOLONG;
1627 		return 0;
1628 	}
1629 
1630 	p = input;
1631 	q = buf;
1632 	while (*p && (q - buf < bufsize)) {
1633 		if (p[0] == '/' && (p[1] == '/' || p[1] == '\0')) {
1634 			p += 1;
1635 
1636 		} else if (p[0] == '/' && p[1] == '.' &&
1637 		    (p[2] == '/' || p[2] == '\0')) {
1638 			p += 2;
1639 
1640 		} else if (p[0] == '/' && p[1] == '.' && p[2] == '.' &&
1641 		    (p[3] == '/' || p[3] == '\0')) {
1642 			p += 3;
1643 			if (q != buf)	/* "/../" at start of buf */
1644 				while (*--q != '/')
1645 					continue;
1646 
1647 		} else {
1648 			*q++ = *p++;
1649 		}
1650 	}
1651 	if ((*p == '\0') && (q - buf < bufsize)) {
1652 		*q = 0;
1653 		return 0;
1654 	} else
1655 		return ENAMETOOLONG;
1656 }
1657