xref: /openbsd-src/sys/kern/kern_pledge.c (revision 1ad61ae0a79a724d2d3ec69e69c8e1d1ff6b53a0)
1 /*	$OpenBSD: kern_pledge.c,v 1.309 2023/09/29 12:47:34 claudio 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 #include <sys/mount.h>
22 #include <sys/proc.h>
23 #include <sys/mutex.h>
24 #include <sys/fcntl.h>
25 #include <sys/file.h>
26 #include <sys/namei.h>
27 #include <sys/socketvar.h>
28 #include <sys/vnode.h>
29 #include <sys/mman.h>
30 #include <sys/sysctl.h>
31 #include <sys/syslog.h>
32 #include <sys/ktrace.h>
33 #include <sys/acct.h>
34 #include <sys/swap.h>
35 
36 #include <sys/ioctl.h>
37 #include <sys/termios.h>
38 #include <sys/tty.h>
39 #include <sys/device.h>
40 #include <sys/disklabel.h>
41 #include <sys/dkio.h>
42 #include <sys/mtio.h>
43 #include <sys/audioio.h>
44 #include <sys/videoio.h>
45 #include <net/bpf.h>
46 #include <net/route.h>
47 #include <net/if.h>
48 #include <net/if_var.h>
49 #include <netinet/in.h>
50 #include <netinet6/in6_var.h>
51 #include <netinet6/nd6.h>
52 #include <netinet/tcp.h>
53 #include <net/pfvar.h>
54 
55 #include <sys/conf.h>
56 #include <sys/specdev.h>
57 #include <sys/signal.h>
58 #include <sys/signalvar.h>
59 #include <sys/syscall.h>
60 #include <sys/syscallargs.h>
61 #include <sys/systm.h>
62 
63 #include <dev/biovar.h>
64 
65 #define PLEDGENAMES
66 #include <sys/pledge.h>
67 
68 #include "audio.h"
69 #include "bpfilter.h"
70 #include "pf.h"
71 #include "video.h"
72 #include "pty.h"
73 
74 #if defined(__amd64__)
75 #include "vmm.h"
76 #if NVMM > 0
77 #include <machine/conf.h>
78 #endif
79 #endif
80 
81 #include "drm.h"
82 
83 uint64_t pledgereq_flags(const char *req);
84 int	 parsepledges(struct proc *p, const char *kname,
85 	    const char *promises, u_int64_t *fp);
86 int	 canonpath(const char *input, char *buf, size_t bufsize);
87 void	 unveil_destroy(struct process *ps);
88 
89 /* #define DEBUG_PLEDGE */
90 #ifdef DEBUG_PLEDGE
91 int debug_pledge = 1;
92 #define DPRINTF(x...)    do { if (debug_pledge) printf(x); } while (0)
93 #define DNPRINTF(n,x...) do { if (debug_pledge >= (n)) printf(x); } while (0)
94 #else
95 #define DPRINTF(x...)
96 #define DNPRINTF(n,x...)
97 #endif
98 
99 /*
100  * Ordered in blocks starting with least risky and most required.
101  */
102 const uint64_t pledge_syscalls[SYS_MAXSYSCALL] = {
103 	/*
104 	 * Minimum required
105 	 */
106 	[SYS_exit] = PLEDGE_ALWAYS,
107 	[SYS_kbind] = PLEDGE_ALWAYS,
108 	[SYS_msyscall] = PLEDGE_ALWAYS,
109 	[SYS___get_tcb] = PLEDGE_ALWAYS,
110 	[SYS___set_tcb] = PLEDGE_ALWAYS,
111 	[SYS_pledge] = PLEDGE_ALWAYS,
112 	[SYS_sendsyslog] = PLEDGE_ALWAYS,	/* stack protector reporting */
113 	[SYS_thrkill] = PLEDGE_ALWAYS,		/* raise, abort, stack pro */
114 	[SYS_utrace] = PLEDGE_ALWAYS,		/* ltrace(1) from ld.so */
115 	[SYS_pinsyscall] = PLEDGE_ALWAYS,
116 
117 	/* "getting" information about self is considered safe */
118 	[SYS_getuid] = PLEDGE_STDIO,
119 	[SYS_geteuid] = PLEDGE_STDIO,
120 	[SYS_getresuid] = PLEDGE_STDIO,
121 	[SYS_getgid] = PLEDGE_STDIO,
122 	[SYS_getegid] = PLEDGE_STDIO,
123 	[SYS_getresgid] = PLEDGE_STDIO,
124 	[SYS_getgroups] = PLEDGE_STDIO,
125 	[SYS_getlogin_r] = PLEDGE_STDIO,
126 	[SYS_getpgrp] = PLEDGE_STDIO,
127 	[SYS_getpgid] = PLEDGE_STDIO,
128 	[SYS_getppid] = PLEDGE_STDIO,
129 	[SYS_getsid] = PLEDGE_STDIO,
130 	[SYS_getthrid] = PLEDGE_STDIO,
131 	[SYS_getrlimit] = PLEDGE_STDIO,
132 	[SYS_getrtable] = PLEDGE_STDIO,
133 	[SYS_gettimeofday] = PLEDGE_STDIO,
134 	[SYS_getdtablecount] = PLEDGE_STDIO,
135 	[SYS_getrusage] = PLEDGE_STDIO,
136 	[SYS_issetugid] = PLEDGE_STDIO,
137 	[SYS_clock_getres] = PLEDGE_STDIO,
138 	[SYS_clock_gettime] = PLEDGE_STDIO,
139 	[SYS_getpid] = PLEDGE_STDIO,
140 
141 	/*
142 	 * Almost exclusively read-only, Very narrow subset.
143 	 * Use of "route", "inet", "dns", "ps", or "vminfo"
144 	 * expands access.
145 	 */
146 	[SYS_sysctl] = PLEDGE_STDIO,
147 
148 	/* For moncontrol(3).  Only allowed to disable profiling. */
149 	[SYS_profil] = PLEDGE_STDIO,
150 
151 	/* Support for malloc(3) family of operations */
152 	[SYS_getentropy] = PLEDGE_STDIO,
153 	[SYS_madvise] = PLEDGE_STDIO,
154 	[SYS_minherit] = PLEDGE_STDIO,
155 	[SYS_mmap] = PLEDGE_STDIO,
156 	[SYS_mprotect] = PLEDGE_STDIO,
157 	[SYS_mimmutable] = PLEDGE_STDIO,
158 	[SYS_mquery] = PLEDGE_STDIO,
159 	[SYS_munmap] = PLEDGE_STDIO,
160 	[SYS_msync] = PLEDGE_STDIO,
161 	[SYS_break] = PLEDGE_STDIO,
162 
163 	[SYS_umask] = PLEDGE_STDIO,
164 
165 	/* read/write operations */
166 	[SYS_read] = PLEDGE_STDIO,
167 	[SYS_readv] = PLEDGE_STDIO,
168 	[SYS_pread] = PLEDGE_STDIO,
169 	[SYS_preadv] = PLEDGE_STDIO,
170 	[SYS_write] = PLEDGE_STDIO,
171 	[SYS_writev] = PLEDGE_STDIO,
172 	[SYS_pwrite] = PLEDGE_STDIO,
173 	[SYS_pwritev] = PLEDGE_STDIO,
174 	[SYS_recvmsg] = PLEDGE_STDIO,
175 	[SYS_recvmmsg] = PLEDGE_STDIO,
176 	[SYS_recvfrom] = PLEDGE_STDIO,
177 	[SYS_ftruncate] = PLEDGE_STDIO,
178 	[SYS_lseek] = PLEDGE_STDIO,
179 	[SYS_fpathconf] = PLEDGE_STDIO,
180 
181 	/*
182 	 * Address selection required a network pledge ("inet",
183 	 * "unix", "dns".
184 	 */
185 	[SYS_sendto] = PLEDGE_STDIO,
186 
187 	/*
188 	 * Address specification required a network pledge ("inet",
189 	 * "unix", "dns".  SCM_RIGHTS requires "sendfd" or "recvfd".
190 	 */
191 	[SYS_sendmsg] = PLEDGE_STDIO,
192 	[SYS_sendmmsg] = PLEDGE_STDIO,
193 
194 	/* Common signal operations */
195 	[SYS_nanosleep] = PLEDGE_STDIO,
196 	[SYS_sigaltstack] = PLEDGE_STDIO,
197 	[SYS_sigprocmask] = PLEDGE_STDIO,
198 	[SYS_sigsuspend] = PLEDGE_STDIO,
199 	[SYS_sigaction] = PLEDGE_STDIO,
200 	[SYS_sigreturn] = PLEDGE_STDIO,
201 	[SYS_sigpending] = PLEDGE_STDIO,
202 	[SYS_getitimer] = PLEDGE_STDIO,
203 	[SYS_setitimer] = PLEDGE_STDIO,
204 
205 	/*
206 	 * To support event driven programming.
207 	 */
208 	[SYS_poll] = PLEDGE_STDIO,
209 	[SYS_ppoll] = PLEDGE_STDIO,
210 	[SYS_kevent] = PLEDGE_STDIO,
211 	[SYS_kqueue] = PLEDGE_STDIO,
212 	[SYS_kqueue1] = PLEDGE_STDIO,
213 	[SYS_select] = PLEDGE_STDIO,
214 	[SYS_pselect] = PLEDGE_STDIO,
215 
216 	[SYS_fstat] = PLEDGE_STDIO,
217 	[SYS_fsync] = PLEDGE_STDIO,
218 
219 	[SYS_setsockopt] = PLEDGE_STDIO,	/* narrow whitelist */
220 	[SYS_getsockopt] = PLEDGE_STDIO,	/* narrow whitelist */
221 
222 	/* F_SETOWN requires PLEDGE_PROC */
223 	[SYS_fcntl] = PLEDGE_STDIO,
224 
225 	[SYS_close] = PLEDGE_STDIO,
226 	[SYS_dup] = PLEDGE_STDIO,
227 	[SYS_dup2] = PLEDGE_STDIO,
228 	[SYS_dup3] = PLEDGE_STDIO,
229 	[SYS_closefrom] = PLEDGE_STDIO,
230 	[SYS_shutdown] = PLEDGE_STDIO,
231 	[SYS_fchdir] = PLEDGE_STDIO,	/* XXX consider tightening */
232 
233 	[SYS_pipe] = PLEDGE_STDIO,
234 	[SYS_pipe2] = PLEDGE_STDIO,
235 	[SYS_socketpair] = PLEDGE_STDIO,
236 
237 	[SYS_wait4] = PLEDGE_STDIO,
238 	[SYS_waitid] = PLEDGE_STDIO,
239 
240 	/*
241 	 * Can kill self with "stdio".  Killing another pid
242 	 * requires "proc"
243 	 */
244 	[SYS_kill] = PLEDGE_STDIO,
245 
246 	/*
247 	 * FIONREAD/FIONBIO for "stdio"
248 	 * Other ioctl are selectively allowed based upon other pledges.
249 	 */
250 	[SYS_ioctl] = PLEDGE_STDIO,
251 
252 	/*
253 	 * Path access/creation calls encounter many extensive
254 	 * checks done during pledge_namei()
255 	 */
256 	[SYS_open] = PLEDGE_STDIO,
257 	[SYS_stat] = PLEDGE_STDIO,
258 	[SYS_access] = PLEDGE_STDIO,
259 	[SYS_readlink] = PLEDGE_STDIO,
260 	[SYS___realpath] = PLEDGE_STDIO,
261 
262 	[SYS_adjtime] = PLEDGE_STDIO,   /* setting requires "settime" */
263 	[SYS_adjfreq] = PLEDGE_SETTIME,
264 	[SYS_settimeofday] = PLEDGE_SETTIME,
265 
266 	/*
267 	 * Needed by threaded programs
268 	 * XXX should we have a new "threads"?
269 	 */
270 	[SYS___tfork] = PLEDGE_STDIO,
271 	[SYS_sched_yield] = PLEDGE_STDIO,
272 	[SYS_futex] = PLEDGE_STDIO,
273 	[SYS___thrsleep] = PLEDGE_STDIO,
274 	[SYS___thrwakeup] = PLEDGE_STDIO,
275 	[SYS___threxit] = PLEDGE_STDIO,
276 	[SYS___thrsigdivert] = PLEDGE_STDIO,
277 	[SYS_getthrname] = PLEDGE_STDIO,
278 	[SYS_setthrname] = PLEDGE_STDIO,
279 
280 	[SYS_fork] = PLEDGE_PROC,
281 	[SYS_vfork] = PLEDGE_PROC,
282 	[SYS_setpgid] = PLEDGE_PROC,
283 	[SYS_setsid] = PLEDGE_PROC,
284 
285 	[SYS_setrlimit] = PLEDGE_PROC | PLEDGE_ID,
286 	[SYS_getpriority] = PLEDGE_PROC | PLEDGE_ID,
287 
288 	[SYS_setpriority] = PLEDGE_PROC | PLEDGE_ID,
289 
290 	[SYS_setuid] = PLEDGE_ID,
291 	[SYS_seteuid] = PLEDGE_ID,
292 	[SYS_setreuid] = PLEDGE_ID,
293 	[SYS_setresuid] = PLEDGE_ID,
294 	[SYS_setgid] = PLEDGE_ID,
295 	[SYS_setegid] = PLEDGE_ID,
296 	[SYS_setregid] = PLEDGE_ID,
297 	[SYS_setresgid] = PLEDGE_ID,
298 	[SYS_setgroups] = PLEDGE_ID,
299 	[SYS_setlogin] = PLEDGE_ID,
300 	[SYS_setrtable] = PLEDGE_ID,
301 
302 	[SYS_unveil] = PLEDGE_UNVEIL,
303 
304 	[SYS_execve] = PLEDGE_EXEC,
305 
306 	[SYS_chdir] = PLEDGE_RPATH,
307 	[SYS_openat] = PLEDGE_RPATH | PLEDGE_WPATH,
308 	[SYS_fstatat] = PLEDGE_RPATH | PLEDGE_WPATH,
309 	[SYS_faccessat] = PLEDGE_RPATH | PLEDGE_WPATH,
310 	[SYS_readlinkat] = PLEDGE_RPATH | PLEDGE_WPATH,
311 	[SYS_lstat] = PLEDGE_RPATH | PLEDGE_WPATH | PLEDGE_TMPPATH,
312 	[SYS_truncate] = PLEDGE_WPATH,
313 	[SYS_rename] = PLEDGE_RPATH | PLEDGE_CPATH,
314 	[SYS_rmdir] = PLEDGE_CPATH,
315 	[SYS_renameat] = PLEDGE_CPATH,
316 	[SYS_link] = PLEDGE_CPATH,
317 	[SYS_linkat] = PLEDGE_CPATH,
318 	[SYS_symlink] = PLEDGE_CPATH,
319 	[SYS_symlinkat] = PLEDGE_CPATH,
320 	[SYS_unlink] = PLEDGE_CPATH | PLEDGE_TMPPATH,
321 	[SYS_unlinkat] = PLEDGE_CPATH,
322 	[SYS_mkdir] = PLEDGE_CPATH,
323 	[SYS_mkdirat] = PLEDGE_CPATH,
324 
325 	[SYS_mkfifo] = PLEDGE_DPATH,
326 	[SYS_mkfifoat] = PLEDGE_DPATH,
327 	[SYS_mknod] = PLEDGE_DPATH,
328 	[SYS_mknodat] = PLEDGE_DPATH,
329 
330 	[SYS_revoke] = PLEDGE_TTY,	/* also requires PLEDGE_RPATH */
331 
332 	/*
333 	 * Classify as RPATH|WPATH, because of path information leakage.
334 	 * WPATH due to unknown use of mk*temp(3) on non-/tmp paths..
335 	 */
336 	[SYS___getcwd] = PLEDGE_RPATH | PLEDGE_WPATH,
337 
338 	/* Classify as RPATH, because these leak path information */
339 	[SYS_getdents] = PLEDGE_RPATH,
340 	[SYS_getfsstat] = PLEDGE_RPATH,
341 	[SYS_statfs] = PLEDGE_RPATH,
342 	[SYS_fstatfs] = PLEDGE_RPATH,
343 	[SYS_pathconf] = PLEDGE_RPATH,
344 
345 	[SYS_utimes] = PLEDGE_FATTR,
346 	[SYS_futimes] = PLEDGE_FATTR,
347 	[SYS_utimensat] = PLEDGE_FATTR,
348 	[SYS_futimens] = PLEDGE_FATTR,
349 	[SYS_chmod] = PLEDGE_FATTR,
350 	[SYS_fchmod] = PLEDGE_FATTR,
351 	[SYS_fchmodat] = PLEDGE_FATTR,
352 	[SYS_chflags] = PLEDGE_FATTR,
353 	[SYS_chflagsat] = PLEDGE_FATTR,
354 	[SYS_fchflags] = PLEDGE_FATTR,
355 
356 	[SYS_chown] = PLEDGE_CHOWN,
357 	[SYS_fchownat] = PLEDGE_CHOWN,
358 	[SYS_lchown] = PLEDGE_CHOWN,
359 	[SYS_fchown] = PLEDGE_CHOWN,
360 
361 	[SYS_socket] = PLEDGE_INET | PLEDGE_UNIX | PLEDGE_DNS,
362 	[SYS_connect] = PLEDGE_INET | PLEDGE_UNIX | PLEDGE_DNS,
363 	[SYS_bind] = PLEDGE_INET | PLEDGE_UNIX | PLEDGE_DNS,
364 	[SYS_getsockname] = PLEDGE_INET | PLEDGE_UNIX | PLEDGE_DNS,
365 
366 	[SYS_listen] = PLEDGE_INET | PLEDGE_UNIX,
367 	[SYS_accept4] = PLEDGE_INET | PLEDGE_UNIX,
368 	[SYS_accept] = PLEDGE_INET | PLEDGE_UNIX,
369 	[SYS_getpeername] = PLEDGE_INET | PLEDGE_UNIX,
370 
371 	[SYS_flock] = PLEDGE_FLOCK,
372 
373 	[SYS_ypconnect] = PLEDGE_GETPW,
374 
375 	[SYS_swapctl] = PLEDGE_VMINFO,
376 };
377 
378 static const struct {
379 	char *name;
380 	uint64_t flags;
381 } pledgereq[] = {
382 	{ "audio",		PLEDGE_AUDIO },
383 	{ "bpf",		PLEDGE_BPF },
384 	{ "chown",		PLEDGE_CHOWN | PLEDGE_CHOWNUID },
385 	{ "cpath",		PLEDGE_CPATH },
386 	{ "disklabel",		PLEDGE_DISKLABEL },
387 	{ "dns",		PLEDGE_DNS },
388 	{ "dpath",		PLEDGE_DPATH },
389 	{ "drm",		PLEDGE_DRM },
390 	{ "error",		PLEDGE_ERROR },
391 	{ "exec",		PLEDGE_EXEC },
392 	{ "fattr",		PLEDGE_FATTR | PLEDGE_CHOWN },
393 	{ "flock",		PLEDGE_FLOCK },
394 	{ "getpw",		PLEDGE_GETPW },
395 	{ "id",			PLEDGE_ID },
396 	{ "inet",		PLEDGE_INET },
397 	{ "mcast",		PLEDGE_MCAST },
398 	{ "pf",			PLEDGE_PF },
399 	{ "proc",		PLEDGE_PROC },
400 	{ "prot_exec",		PLEDGE_PROTEXEC },
401 	{ "ps",			PLEDGE_PS },
402 	{ "recvfd",		PLEDGE_RECVFD },
403 	{ "route",		PLEDGE_ROUTE },
404 	{ "rpath",		PLEDGE_RPATH },
405 	{ "sendfd",		PLEDGE_SENDFD },
406 	{ "settime",		PLEDGE_SETTIME },
407 	{ "stdio",		PLEDGE_STDIO },
408 	{ "tape",		PLEDGE_TAPE },
409 	{ "tmppath",		PLEDGE_TMPPATH },
410 	{ "tty",		PLEDGE_TTY },
411 	{ "unix",		PLEDGE_UNIX },
412 	{ "unveil",		PLEDGE_UNVEIL },
413 	{ "video",		PLEDGE_VIDEO },
414 	{ "vminfo",		PLEDGE_VMINFO },
415 	{ "vmm",		PLEDGE_VMM },
416 	{ "wpath",		PLEDGE_WPATH },
417 	{ "wroute",		PLEDGE_WROUTE },
418 };
419 
420 int
421 parsepledges(struct proc *p, const char *kname, const char *promises, u_int64_t *fp)
422 {
423 	size_t rbuflen;
424 	char *rbuf, *rp, *pn;
425 	u_int64_t flags = 0, f;
426 	int error;
427 
428 	rbuf = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
429 	error = copyinstr(promises, rbuf, MAXPATHLEN, &rbuflen);
430 	if (error) {
431 		free(rbuf, M_TEMP, MAXPATHLEN);
432 		return (error);
433 	}
434 #ifdef KTRACE
435 	if (KTRPOINT(p, KTR_STRUCT))
436 		ktrstruct(p, kname, rbuf, rbuflen-1);
437 #endif
438 
439 	for (rp = rbuf; rp && *rp; rp = pn) {
440 		pn = strchr(rp, ' ');	/* find terminator */
441 		if (pn) {
442 			while (*pn == ' ')
443 				*pn++ = '\0';
444 		}
445 		if ((f = pledgereq_flags(rp)) == 0) {
446 			free(rbuf, M_TEMP, MAXPATHLEN);
447 			return (EINVAL);
448 		}
449 		flags |= f;
450 	}
451 	free(rbuf, M_TEMP, MAXPATHLEN);
452 	*fp = flags;
453 	return 0;
454 }
455 
456 int
457 sys_pledge(struct proc *p, void *v, register_t *retval)
458 {
459 	struct sys_pledge_args /* {
460 		syscallarg(const char *)promises;
461 		syscallarg(const char *)execpromises;
462 	} */	*uap = v;
463 	struct process *pr = p->p_p;
464 	uint64_t promises, execpromises;
465 	int error;
466 	int unveil_cleanup = 0;
467 
468 	/* Check for any error in user input */
469 	if (SCARG(uap, promises)) {
470 		error = parsepledges(p, "pledgereq",
471 		    SCARG(uap, promises), &promises);
472 		if (error)
473 			return (error);
474 	}
475 	if (SCARG(uap, execpromises)) {
476 		error = parsepledges(p, "pledgeexecreq",
477 		    SCARG(uap, execpromises), &execpromises);
478 		if (error)
479 			return (error);
480 	}
481 
482 	mtx_enter(&pr->ps_mtx);
483 
484 	/* Check for any error wrt current promises */
485 	if (SCARG(uap, promises)) {
486 		/* In "error" mode, ignore promise increase requests,
487 		 * but accept promise decrease requests */
488 		if (ISSET(pr->ps_flags, PS_PLEDGE) &&
489 		    (pr->ps_pledge & PLEDGE_ERROR))
490 			promises &= (pr->ps_pledge & PLEDGE_USERSET);
491 
492 		/* Only permit reductions */
493 		if (ISSET(pr->ps_flags, PS_PLEDGE) &&
494 		    (((promises | pr->ps_pledge) != pr->ps_pledge))) {
495 			mtx_leave(&pr->ps_mtx);
496 			return (EPERM);
497 		}
498 	}
499 	if (SCARG(uap, execpromises)) {
500 		/* Only permit reductions */
501 		if (ISSET(pr->ps_flags, PS_EXECPLEDGE) &&
502 		    (((execpromises | pr->ps_execpledge) != pr->ps_execpledge))) {
503 			mtx_leave(&pr->ps_mtx);
504 			return (EPERM);
505 		}
506 	}
507 
508 	/* Set up promises */
509 	if (SCARG(uap, promises)) {
510 		pr->ps_pledge = promises;
511 		atomic_setbits_int(&pr->ps_flags, PS_PLEDGE);
512 
513 		if ((pr->ps_pledge & (PLEDGE_RPATH | PLEDGE_WPATH |
514 		    PLEDGE_CPATH | PLEDGE_DPATH | PLEDGE_TMPPATH | PLEDGE_EXEC |
515 		    PLEDGE_UNIX | PLEDGE_UNVEIL)) == 0)
516 			unveil_cleanup = 1;
517 	}
518 	if (SCARG(uap, execpromises)) {
519 		pr->ps_execpledge = execpromises;
520 		atomic_setbits_int(&pr->ps_flags, PS_EXECPLEDGE);
521 	}
522 
523 	mtx_leave(&pr->ps_mtx);
524 
525 	if (unveil_cleanup) {
526 		/*
527 		 * Kill off unveil and drop unveil vnode refs if we no
528 		 * longer are holding any path-accessing pledge
529 		 */
530 		KERNEL_LOCK();
531 		unveil_destroy(pr);
532 		KERNEL_UNLOCK();
533 	}
534 
535 	return (0);
536 }
537 
538 int
539 pledge_syscall(struct proc *p, int code, uint64_t *tval)
540 {
541 	p->p_pledge_syscall = code;
542 	*tval = 0;
543 
544 	if (code < 0 || code > SYS_MAXSYSCALL - 1)
545 		return (EINVAL);
546 
547 	if (pledge_syscalls[code] == PLEDGE_ALWAYS)
548 		return (0);
549 
550 	if (p->p_p->ps_pledge & pledge_syscalls[code])
551 		return (0);
552 
553 	*tval = pledge_syscalls[code];
554 	return (EPERM);
555 }
556 
557 int
558 pledge_fail(struct proc *p, int error, uint64_t code)
559 {
560 	const char *codes = "";
561 	int i;
562 
563 	/* Print first matching pledge */
564 	for (i = 0; code && pledgenames[i].bits != 0; i++)
565 		if (pledgenames[i].bits & code) {
566 			codes = pledgenames[i].name;
567 			break;
568 		}
569 #ifdef KTRACE
570 	if (KTRPOINT(p, KTR_PLEDGE))
571 		ktrpledge(p, error, code, p->p_pledge_syscall);
572 #endif
573 	if (p->p_p->ps_pledge & PLEDGE_ERROR)
574 		return (ENOSYS);
575 
576 	KERNEL_LOCK();
577 	log(LOG_ERR, "%s[%d]: pledge \"%s\", syscall %d\n",
578 	    p->p_p->ps_comm, p->p_p->ps_pid, codes, p->p_pledge_syscall);
579 	p->p_p->ps_acflag |= APLEDGE;
580 
581 	/* Try to stop threads immediately, because this process is suspect */
582 	if (P_HASSIBLING(p))
583 		single_thread_set(p, SINGLE_UNWIND | SINGLE_DEEP);
584 
585 	/* Send uncatchable SIGABRT for coredump */
586 	sigabort(p);
587 
588 	p->p_p->ps_pledge = 0;		/* Disable all PLEDGE_ flags */
589 	KERNEL_UNLOCK();
590 	return (error);
591 }
592 
593 /*
594  * Need to make it more obvious that one cannot get through here
595  * without the right flags set
596  */
597 int
598 pledge_namei(struct proc *p, struct nameidata *ni, char *origpath)
599 {
600 	char path[PATH_MAX];
601 	uint64_t pledge;
602 	int error;
603 
604 	if ((p->p_p->ps_flags & PS_PLEDGE) == 0 ||
605 	    (p->p_p->ps_flags & PS_COREDUMP))
606 		return (0);
607 	pledge = READ_ONCE(p->p_p->ps_pledge);
608 
609 	if (ni->ni_pledge == 0)
610 		panic("pledge_namei: ni_pledge");
611 
612 	/*
613 	 * We set the BYPASSUNVEIL flag to skip unveil checks
614 	 * as necessary
615 	 */
616 
617 	/* Doing a permitted execve() */
618 	if ((ni->ni_pledge & PLEDGE_EXEC) &&
619 	    (pledge & PLEDGE_EXEC))
620 		return (0);
621 
622 	error = canonpath(origpath, path, sizeof(path));
623 	if (error)
624 		return (error);
625 
626 	/* Detect what looks like a mkstemp(3) family operation */
627 	if ((pledge & PLEDGE_TMPPATH) &&
628 	    (p->p_pledge_syscall == SYS_open) &&
629 	    (ni->ni_pledge & PLEDGE_CPATH) &&
630 	    strncmp(path, "/tmp/", sizeof("/tmp/") - 1) == 0) {
631 		ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
632 		return (0);
633 	}
634 
635 	/* Allow unlinking of a mkstemp(3) file...
636 	 * Good opportunity for strict checks here.
637 	 */
638 	if ((pledge & PLEDGE_TMPPATH) &&
639 	    (p->p_pledge_syscall == SYS_unlink) &&
640 	    strncmp(path, "/tmp/", sizeof("/tmp/") - 1) == 0) {
641 		ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
642 		return (0);
643 	}
644 
645 	/* Whitelisted paths */
646 	switch (p->p_pledge_syscall) {
647 	case SYS_access:
648 		/* tzset() needs this. */
649 		if (ni->ni_pledge == PLEDGE_RPATH &&
650 		    strcmp(path, "/etc/localtime") == 0) {
651 			ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
652 			return (0);
653 		}
654 		break;
655 	case SYS_open:
656 		/* daemon(3) or other such functions */
657 		if ((ni->ni_pledge & ~(PLEDGE_RPATH | PLEDGE_WPATH)) == 0 &&
658 		    strcmp(path, "/dev/null") == 0) {
659 			ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
660 			return (0);
661 		}
662 
663 		/* readpassphrase(3), getpass(3) */
664 		if ((pledge & PLEDGE_TTY) &&
665 		    (ni->ni_pledge & ~(PLEDGE_RPATH | PLEDGE_WPATH)) == 0 &&
666 		    strcmp(path, "/dev/tty") == 0) {
667 			ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
668 			return (0);
669 		}
670 
671 		/* getpw* and friends need a few files */
672 		if ((ni->ni_pledge == PLEDGE_RPATH) &&
673 		    (pledge & PLEDGE_GETPW)) {
674 			if (strcmp(path, "/etc/spwd.db") == 0)
675 				return (EPERM); /* don't call pledge_fail */
676 			if (strcmp(path, "/etc/pwd.db") == 0) {
677 				ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
678 				return (0);
679 			}
680 			if (strcmp(path, "/etc/group") == 0) {
681 				ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
682 				return (0);
683 			}
684 			if (strcmp(path, "/etc/netid") == 0) {
685 				ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
686 				return (0);
687 			}
688 		}
689 
690 		/* DNS needs /etc/{resolv.conf,hosts,services,protocols}. */
691 		if ((ni->ni_pledge == PLEDGE_RPATH) &&
692 		    (pledge & PLEDGE_DNS)) {
693 			if (strcmp(path, "/etc/resolv.conf") == 0) {
694 				ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
695 				return (0);
696 			}
697 			if (strcmp(path, "/etc/hosts") == 0) {
698 				ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
699 				return (0);
700 			}
701 			if (strcmp(path, "/etc/services") == 0) {
702 				ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
703 				return (0);
704 			}
705 			if (strcmp(path, "/etc/protocols") == 0) {
706 				ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
707 				return (0);
708 			}
709 		}
710 
711 		/* tzset() needs these. */
712 		if ((ni->ni_pledge == PLEDGE_RPATH) &&
713 		    strncmp(path, "/usr/share/zoneinfo/",
714 		    sizeof("/usr/share/zoneinfo/") - 1) == 0)  {
715 			ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
716 			return (0);
717 		}
718 		if ((ni->ni_pledge == PLEDGE_RPATH) &&
719 		    strcmp(path, "/etc/localtime") == 0) {
720 			ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
721 			return (0);
722 		}
723 
724 		break;
725 	case SYS_stat:
726 		/* DNS needs /etc/{resolv.conf,hosts}. */
727 		if ((ni->ni_pledge == PLEDGE_RPATH) &&
728 		    (pledge & PLEDGE_DNS)) {
729 			if (strcmp(path, "/etc/resolv.conf") == 0) {
730 				ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
731 				return (0);
732 			}
733 			if (strcmp(path, "/etc/hosts") == 0) {
734 				ni->ni_cnd.cn_flags |= BYPASSUNVEIL;
735 				return (0);
736 			}
737 		}
738 		break;
739 	}
740 
741 	/*
742 	 * Ensure each flag of ni_pledge has counterpart allowing it in
743 	 * ps_pledge.
744 	 */
745 	if (ni->ni_pledge & ~pledge)
746 		return (pledge_fail(p, EPERM, (ni->ni_pledge & ~pledge)));
747 
748 	/* continue, and check unveil if present */
749 	return (0);
750 }
751 
752 /*
753  * Only allow reception of safe file descriptors.
754  */
755 int
756 pledge_recvfd(struct proc *p, struct file *fp)
757 {
758 	struct vnode *vp;
759 
760 	if ((p->p_p->ps_flags & PS_PLEDGE) == 0)
761 		return (0);
762 	if ((p->p_p->ps_pledge & PLEDGE_RECVFD) == 0)
763 		return pledge_fail(p, EPERM, PLEDGE_RECVFD);
764 
765 	switch (fp->f_type) {
766 	case DTYPE_SOCKET:
767 	case DTYPE_PIPE:
768 	case DTYPE_DMABUF:
769 	case DTYPE_SYNC:
770 		return (0);
771 	case DTYPE_VNODE:
772 		vp = fp->f_data;
773 
774 		if (vp->v_type != VDIR)
775 			return (0);
776 	}
777 	return pledge_fail(p, EINVAL, PLEDGE_RECVFD);
778 }
779 
780 /*
781  * Only allow sending of safe file descriptors.
782  */
783 int
784 pledge_sendfd(struct proc *p, struct file *fp)
785 {
786 	struct vnode *vp;
787 
788 	if ((p->p_p->ps_flags & PS_PLEDGE) == 0)
789 		return (0);
790 	if ((p->p_p->ps_pledge & PLEDGE_SENDFD) == 0)
791 		return pledge_fail(p, EPERM, PLEDGE_SENDFD);
792 
793 	switch (fp->f_type) {
794 	case DTYPE_SOCKET:
795 	case DTYPE_PIPE:
796 	case DTYPE_DMABUF:
797 	case DTYPE_SYNC:
798 		return (0);
799 	case DTYPE_VNODE:
800 		vp = fp->f_data;
801 
802 		if (vp->v_type != VDIR)
803 			return (0);
804 		break;
805 	}
806 	return pledge_fail(p, EINVAL, PLEDGE_SENDFD);
807 }
808 
809 int
810 pledge_sysctl(struct proc *p, int miblen, int *mib, void *new)
811 {
812 	char	buf[80];
813 	uint64_t pledge;
814 	int	i;
815 
816 	if ((p->p_p->ps_flags & PS_PLEDGE) == 0)
817 		return (0);
818 	pledge = READ_ONCE(p->p_p->ps_pledge);
819 
820 	if (new)
821 		return pledge_fail(p, EFAULT, 0);
822 
823 	/* routing table observation */
824 	if ((pledge & PLEDGE_ROUTE)) {
825 		if ((miblen == 6 || miblen == 7) &&
826 		    mib[0] == CTL_NET && mib[1] == PF_ROUTE &&
827 		    mib[2] == 0 &&
828 		    mib[4] == NET_RT_DUMP)
829 			return (0);
830 
831 		if (miblen == 6 &&
832 		    mib[0] == CTL_NET && mib[1] == PF_ROUTE &&
833 		    mib[2] == 0 &&
834 		    (mib[3] == 0 || mib[3] == AF_INET6 || mib[3] == AF_INET) &&
835 		    (mib[4] == NET_RT_TABLE || mib[4] == NET_RT_SOURCE))
836 			return (0);
837 
838 		if (miblen == 7 &&		/* exposes MACs */
839 		    mib[0] == CTL_NET && mib[1] == PF_ROUTE &&
840 		    mib[2] == 0 &&
841 		    (mib[3] == 0 || mib[3] == AF_INET6 || mib[3] == AF_INET) &&
842 		    mib[4] == NET_RT_FLAGS && mib[5] == RTF_LLINFO)
843 			return (0);
844 	}
845 
846 	if ((pledge & PLEDGE_WROUTE)) {
847 		if (miblen == 4 &&
848 		    mib[0] == CTL_NET && mib[1] == PF_INET6 &&
849 		    mib[2] == IPPROTO_IPV6 && mib[3] == IPV6CTL_SOIIKEY)
850 			return (0);
851 	}
852 
853 	if (pledge & (PLEDGE_PS | PLEDGE_VMINFO)) {
854 		if (miblen == 2 &&		/* kern.fscale */
855 		    mib[0] == CTL_KERN && mib[1] == KERN_FSCALE)
856 			return (0);
857 		if (miblen == 2 &&		/* kern.boottime */
858 		    mib[0] == CTL_KERN && mib[1] == KERN_BOOTTIME)
859 			return (0);
860 		if (miblen == 2 &&		/* kern.consdev */
861 		    mib[0] == CTL_KERN && mib[1] == KERN_CONSDEV)
862 			return (0);
863 		if (miblen == 2 &&			/* kern.cptime */
864 		    mib[0] == CTL_KERN && mib[1] == KERN_CPTIME)
865 			return (0);
866 		if (miblen == 3 &&			/* kern.cptime2 */
867 		    mib[0] == CTL_KERN && mib[1] == KERN_CPTIME2)
868 			return (0);
869 		if (miblen == 3 &&			/* kern.cpustats */
870 		    mib[0] == CTL_KERN && mib[1] == KERN_CPUSTATS)
871 			return (0);
872 	}
873 
874 	if ((pledge & PLEDGE_PS)) {
875 		if (miblen == 4 &&		/* kern.procargs.* */
876 		    mib[0] == CTL_KERN && mib[1] == KERN_PROC_ARGS &&
877 		    (mib[3] == KERN_PROC_ARGV || mib[3] == KERN_PROC_ENV))
878 			return (0);
879 		if (miblen == 6 &&		/* kern.proc.* */
880 		    mib[0] == CTL_KERN && mib[1] == KERN_PROC)
881 			return (0);
882 		if (miblen == 3 &&		/* kern.proc_cwd.* */
883 		    mib[0] == CTL_KERN && mib[1] == KERN_PROC_CWD)
884 			return (0);
885 		if (miblen == 2 &&		/* kern.ccpu */
886 		    mib[0] == CTL_KERN && mib[1] == KERN_CCPU)
887 			return (0);
888 		if (miblen == 2 &&		/* vm.maxslp */
889 		    mib[0] == CTL_VM && mib[1] == VM_MAXSLP)
890 			return (0);
891 	}
892 
893 	if ((pledge & PLEDGE_VMINFO)) {
894 		if (miblen == 2 &&		/* vm.uvmexp */
895 		    mib[0] == CTL_VM && mib[1] == VM_UVMEXP)
896 			return (0);
897 		if (miblen == 3 &&		/* vfs.generic.bcachestat */
898 		    mib[0] == CTL_VFS && mib[1] == VFS_GENERIC &&
899 		    mib[2] == VFS_BCACHESTAT)
900 			return (0);
901 		if (miblen == 3 &&		/* for sysconf(3) */
902 		    mib[0] == CTL_NET && mib[1] == PF_INET6)
903 			return (0);
904 	}
905 
906 	if ((pledge & (PLEDGE_INET | PLEDGE_UNIX))) {
907 		if (miblen == 2 &&		/* kern.somaxconn */
908 		    mib[0] == CTL_KERN && mib[1] == KERN_SOMAXCONN)
909 			return (0);
910 	}
911 
912 	if ((pledge & (PLEDGE_ROUTE | PLEDGE_INET | PLEDGE_DNS))) {
913 		if (miblen == 6 &&		/* getifaddrs() */
914 		    mib[0] == CTL_NET && mib[1] == PF_ROUTE &&
915 		    mib[2] == 0 &&
916 		    (mib[3] == 0 || mib[3] == AF_INET6 || mib[3] == AF_INET) &&
917 		    mib[4] == NET_RT_IFLIST)
918 			return (0);
919 	}
920 
921 	if ((pledge & PLEDGE_DISKLABEL)) {
922 		if (miblen == 2 &&		/* kern.rawpartition */
923 		    mib[0] == CTL_KERN &&
924 		    mib[1] == KERN_RAWPARTITION)
925 			return (0);
926 		if (miblen == 2 &&		/* kern.maxpartitions */
927 		    mib[0] == CTL_KERN &&
928 		    mib[1] == KERN_MAXPARTITIONS)
929 			return (0);
930 #ifdef CPU_CHR2BLK
931 		if (miblen == 3 &&		/* machdep.chr2blk */
932 		    mib[0] == CTL_MACHDEP &&
933 		    mib[1] == CPU_CHR2BLK)
934 			return (0);
935 #endif /* CPU_CHR2BLK */
936 	}
937 
938 	if (miblen >= 3 &&			/* ntpd(8) to read sensors */
939 	    mib[0] == CTL_HW && mib[1] == HW_SENSORS)
940 		return (0);
941 
942 	if (miblen == 6 &&		/* if_nameindex() */
943 	    mib[0] == CTL_NET && mib[1] == PF_ROUTE &&
944 	    mib[2] == 0 && mib[3] == 0 && mib[4] == NET_RT_IFNAMES)
945 		return (0);
946 
947 	if (miblen == 2) {
948 		switch (mib[0]) {
949 		case CTL_KERN:
950 			switch (mib[1]) {
951 			case KERN_DOMAINNAME:	/* getdomainname() */
952 			case KERN_HOSTNAME:	/* gethostname() */
953 			case KERN_OSTYPE:	/* uname() */
954 			case KERN_OSRELEASE:	/* uname() */
955 			case KERN_OSVERSION:	/* uname() */
956 			case KERN_VERSION:	/* uname() */
957 			case KERN_CLOCKRATE:	/* kern.clockrate */
958 			case KERN_ARGMAX:	/* kern.argmax */
959 			case KERN_NGROUPS:	/* kern.ngroups */
960 			case KERN_SYSVSHM:	/* kern.sysvshm */
961 			case KERN_POSIX1:	/* kern.posix1version */
962 			case KERN_AUTOCONF_SERIAL:	/* kern.autoconf_serial */
963 				return (0);
964 			}
965 			break;
966 		case CTL_HW:
967 			switch (mib[1]) {
968 			case HW_MACHINE: 	/* uname() */
969 			case HW_PAGESIZE: 	/* getpagesize() */
970 			case HW_PHYSMEM64:	/* hw.physmem */
971 			case HW_NCPU:		/* hw.ncpu */
972 			case HW_NCPUONLINE:	/* hw.ncpuonline */
973 			case HW_USERMEM64:	/* hw.usermem */
974 				return (0);
975 			}
976 			break;
977 		case CTL_VM:
978 			switch (mib[1]) {
979 			case VM_PSSTRINGS:	/* setproctitle() */
980 			case VM_LOADAVG:	/* vm.loadavg / getloadavg(3) */
981 			case VM_MALLOC_CONF:	/* vm.malloc_conf */
982 				return (0);
983 			}
984 			break;
985 		default:
986 			break;
987 		}
988 	}
989 
990 #ifdef CPU_SSE
991 	if (miblen == 2 &&		/* i386 libm tests for SSE */
992 	    mib[0] == CTL_MACHDEP && mib[1] == CPU_SSE)
993 		return (0);
994 #endif /* CPU_SSE */
995 
996 #ifdef CPU_ID_AA64ISAR0
997 	if (miblen == 2 &&		/* arm64 libcrypto inspects CPU features */
998 	    mib[0] == CTL_MACHDEP && mib[1] == CPU_ID_AA64ISAR0)
999 		return (0);
1000 #endif /* CPU_ID_AA64ISAR0 */
1001 
1002 	snprintf(buf, sizeof(buf), "%s(%d): pledge sysctl %d:",
1003 	    p->p_p->ps_comm, p->p_p->ps_pid, miblen);
1004 	for (i = 0; i < miblen; i++) {
1005 		char *p = buf + strlen(buf);
1006 		snprintf(p, sizeof(buf) - (p - buf), " %d", mib[i]);
1007 	}
1008 	log(LOG_ERR, "%s\n", buf);
1009 
1010 	return pledge_fail(p, EINVAL, 0);
1011 }
1012 
1013 int
1014 pledge_chown(struct proc *p, uid_t uid, gid_t gid)
1015 {
1016 	if ((p->p_p->ps_flags & PS_PLEDGE) == 0)
1017 		return (0);
1018 
1019 	if (p->p_p->ps_pledge & PLEDGE_CHOWNUID)
1020 		return (0);
1021 
1022 	if (uid != -1 && uid != p->p_ucred->cr_uid)
1023 		return (EPERM);
1024 	if (gid != -1 && !groupmember(gid, p->p_ucred))
1025 		return (EPERM);
1026 	return (0);
1027 }
1028 
1029 int
1030 pledge_adjtime(struct proc *p, const void *v)
1031 {
1032 	const struct timeval *delta = v;
1033 
1034 	if ((p->p_p->ps_flags & PS_PLEDGE) == 0)
1035 		return (0);
1036 
1037 	if ((p->p_p->ps_pledge & PLEDGE_SETTIME))
1038 		return (0);
1039 	if (delta)
1040 		return (EPERM);
1041 	return (0);
1042 }
1043 
1044 int
1045 pledge_sendit(struct proc *p, const void *to)
1046 {
1047 	if ((p->p_p->ps_flags & PS_PLEDGE) == 0)
1048 		return (0);
1049 
1050 	if ((p->p_p->ps_pledge & (PLEDGE_INET | PLEDGE_UNIX | PLEDGE_DNS)))
1051 		return (0);		/* may use address */
1052 	if (to == NULL)
1053 		return (0);		/* behaves just like write */
1054 	return pledge_fail(p, EPERM, PLEDGE_INET);
1055 }
1056 
1057 int
1058 pledge_ioctl(struct proc *p, long com, struct file *fp)
1059 {
1060 	struct vnode *vp = NULL;
1061 	int error = EPERM;
1062 	uint64_t pledge;
1063 
1064 	if ((p->p_p->ps_flags & PS_PLEDGE) == 0)
1065 		return (0);
1066 	pledge = READ_ONCE(p->p_p->ps_pledge);
1067 
1068 	/*
1069 	 * The ioctl's which are always allowed.
1070 	 */
1071 	switch (com) {
1072 	case FIONREAD:
1073 	case FIONBIO:
1074 	case FIOCLEX:
1075 	case FIONCLEX:
1076 		return (0);
1077 	}
1078 
1079 	/* fp != NULL was already checked */
1080 	if (fp->f_type == DTYPE_VNODE) {
1081 		vp = fp->f_data;
1082 		if (vp->v_type == VBAD)
1083 			return (ENOTTY);
1084 	}
1085 
1086 	if ((pledge & PLEDGE_INET)) {
1087 		switch (com) {
1088 		case SIOCATMARK:
1089 		case SIOCGIFGROUP:
1090 			if (fp->f_type == DTYPE_SOCKET)
1091 				return (0);
1092 			break;
1093 		}
1094 	}
1095 
1096 #if NBPFILTER > 0
1097 	if ((pledge & PLEDGE_BPF)) {
1098 		switch (com) {
1099 		case BIOCGSTATS:	/* bpf: tcpdump privsep on ^C */
1100 			if (fp->f_type == DTYPE_VNODE &&
1101 			    fp->f_ops->fo_ioctl == vn_ioctl &&
1102 			    vp->v_type == VCHR &&
1103 			    cdevsw[major(vp->v_rdev)].d_open == bpfopen)
1104 				return (0);
1105 			break;
1106 		}
1107 	}
1108 #endif /* NBPFILTER > 0 */
1109 
1110 	if ((pledge & PLEDGE_TAPE)) {
1111 		switch (com) {
1112 		case MTIOCGET:
1113 		case MTIOCTOP:
1114 			/* for pax(1) and such, checking tapes... */
1115 			if (fp->f_type == DTYPE_VNODE &&
1116 			    vp->v_type == VCHR) {
1117 				if (vp->v_flag & VISTTY)
1118 					return (ENOTTY);
1119 				else
1120 					return (0);
1121 			}
1122 			break;
1123 		}
1124 	}
1125 
1126 #if NDRM > 0
1127 	if ((pledge & PLEDGE_DRM)) {
1128 		if ((fp->f_type == DTYPE_VNODE) &&
1129 		    (vp->v_type == VCHR) &&
1130 		    (cdevsw[major(vp->v_rdev)].d_open == drmopen)) {
1131 			error = pledge_ioctl_drm(p, com, vp->v_rdev);
1132 			if (error == 0)
1133 				return 0;
1134 		}
1135 	}
1136 #endif /* NDRM > 0 */
1137 
1138 #if NAUDIO > 0
1139 	if ((pledge & PLEDGE_AUDIO)) {
1140 		switch (com) {
1141 		case AUDIO_GETPOS:
1142 		case AUDIO_GETPAR:
1143 		case AUDIO_SETPAR:
1144 		case AUDIO_START:
1145 		case AUDIO_STOP:
1146 		case AUDIO_MIXER_DEVINFO:
1147 		case AUDIO_MIXER_READ:
1148 		case AUDIO_MIXER_WRITE:
1149 			if (fp->f_type == DTYPE_VNODE &&
1150 			    vp->v_type == VCHR &&
1151 			    cdevsw[major(vp->v_rdev)].d_open == audioopen)
1152 				return (0);
1153 		}
1154 	}
1155 #endif /* NAUDIO > 0 */
1156 
1157 	if ((pledge & PLEDGE_DISKLABEL)) {
1158 		switch (com) {
1159 		case DIOCGDINFO:
1160 		case DIOCGPDINFO:
1161 		case DIOCRLDINFO:
1162 		case DIOCWDINFO:
1163 		case BIOCDISK:
1164 		case BIOCINQ:
1165 		case BIOCINSTALLBOOT:
1166 		case BIOCVOL:
1167 			if (fp->f_type == DTYPE_VNODE &&
1168 			    ((vp->v_type == VCHR &&
1169 			    cdevsw[major(vp->v_rdev)].d_type == D_DISK) ||
1170 			    (vp->v_type == VBLK &&
1171 			    bdevsw[major(vp->v_rdev)].d_type == D_DISK)))
1172 				return (0);
1173 			break;
1174 		case DIOCMAP:
1175 			if (fp->f_type == DTYPE_VNODE &&
1176 			    vp->v_type == VCHR &&
1177 			    cdevsw[major(vp->v_rdev)].d_ioctl == diskmapioctl)
1178 				return (0);
1179 			break;
1180 		}
1181 	}
1182 
1183 #if NVIDEO > 0
1184 	if ((pledge & PLEDGE_VIDEO)) {
1185 		switch (com) {
1186 		case VIDIOC_QUERYCAP:
1187 		case VIDIOC_TRY_FMT:
1188 		case VIDIOC_ENUM_FMT:
1189 		case VIDIOC_S_FMT:
1190 		case VIDIOC_QUERYCTRL:
1191 		case VIDIOC_G_CTRL:
1192 		case VIDIOC_S_CTRL:
1193 		case VIDIOC_G_PARM:
1194 		case VIDIOC_S_PARM:
1195 		case VIDIOC_REQBUFS:
1196 		case VIDIOC_QBUF:
1197 		case VIDIOC_DQBUF:
1198 		case VIDIOC_QUERYBUF:
1199 		case VIDIOC_STREAMON:
1200 		case VIDIOC_STREAMOFF:
1201 		case VIDIOC_ENUM_FRAMESIZES:
1202 		case VIDIOC_ENUM_FRAMEINTERVALS:
1203 		case VIDIOC_DQEVENT:
1204 		case VIDIOC_ENCODER_CMD:
1205 		case VIDIOC_EXPBUF:
1206 		case VIDIOC_G_CROP:
1207 		case VIDIOC_G_EXT_CTRLS:
1208 		case VIDIOC_G_FMT:
1209 		case VIDIOC_G_SELECTION:
1210 		case VIDIOC_QUERYMENU:
1211 		case VIDIOC_SUBSCRIBE_EVENT:
1212 		case VIDIOC_S_EXT_CTRLS:
1213 		case VIDIOC_S_SELECTION:
1214 		case VIDIOC_TRY_DECODER_CMD:
1215 		case VIDIOC_TRY_ENCODER_CMD:
1216 			if (fp->f_type == DTYPE_VNODE &&
1217 			    vp->v_type == VCHR &&
1218 			    cdevsw[major(vp->v_rdev)].d_open == videoopen)
1219 				return (0);
1220 			break;
1221 		}
1222 	}
1223 #endif
1224 
1225 #if NPF > 0
1226 	if ((pledge & PLEDGE_PF)) {
1227 		switch (com) {
1228 		case DIOCADDRULE:
1229 		case DIOCGETSTATUS:
1230 		case DIOCNATLOOK:
1231 		case DIOCRADDTABLES:
1232 		case DIOCRCLRADDRS:
1233 		case DIOCRCLRTABLES:
1234 		case DIOCRCLRTSTATS:
1235 		case DIOCRGETTSTATS:
1236 		case DIOCRSETADDRS:
1237 		case DIOCXBEGIN:
1238 		case DIOCXCOMMIT:
1239 		case DIOCKILLSRCNODES:
1240 			if ((fp->f_type == DTYPE_VNODE) &&
1241 			    (vp->v_type == VCHR) &&
1242 			    (cdevsw[major(vp->v_rdev)].d_open == pfopen))
1243 				return (0);
1244 			break;
1245 		}
1246 	}
1247 #endif
1248 
1249 	if ((pledge & PLEDGE_TTY)) {
1250 		switch (com) {
1251 #if NPTY > 0
1252 		case PTMGET:
1253 			if ((pledge & PLEDGE_RPATH) == 0)
1254 				break;
1255 			if ((pledge & PLEDGE_WPATH) == 0)
1256 				break;
1257 			if (fp->f_type != DTYPE_VNODE || vp->v_type != VCHR)
1258 				break;
1259 			if (cdevsw[major(vp->v_rdev)].d_open != ptmopen)
1260 				break;
1261 			return (0);
1262 		case TIOCUCNTL:		/* vmd */
1263 			if ((pledge & PLEDGE_RPATH) == 0)
1264 				break;
1265 			if ((pledge & PLEDGE_WPATH) == 0)
1266 				break;
1267 			if (cdevsw[major(vp->v_rdev)].d_open != ptcopen)
1268 				break;
1269 			return (0);
1270 #endif /* NPTY > 0 */
1271 		case TIOCSPGRP:
1272 			if ((pledge & PLEDGE_PROC) == 0)
1273 				break;
1274 			/* FALLTHROUGH */
1275 		case TIOCFLUSH:		/* getty, telnet */
1276 		case TIOCSTART:		/* emacs, etc */
1277 		case TIOCGPGRP:
1278 		case TIOCGETA:
1279 		case TIOCGWINSZ:	/* ENOTTY return for non-tty */
1280 		case TIOCSTAT:		/* csh */
1281 			if (fp->f_type == DTYPE_VNODE && (vp->v_flag & VISTTY))
1282 				return (0);
1283 			return (ENOTTY);
1284 		case TIOCSWINSZ:
1285 		case TIOCEXT:		/* mail, libedit .. */
1286 		case TIOCCBRK:		/* cu */
1287 		case TIOCSBRK:		/* cu */
1288 		case TIOCCDTR:		/* cu */
1289 		case TIOCSDTR:		/* cu */
1290 		case TIOCEXCL:		/* cu */
1291 		case TIOCSETA:		/* cu, ... */
1292 		case TIOCSETAW:		/* cu, ... */
1293 		case TIOCSETAF:		/* tcsetattr TCSAFLUSH, script */
1294 		case TIOCSCTTY:		/* forkpty(3), login_tty(3), ... */
1295 			if (fp->f_type == DTYPE_VNODE && (vp->v_flag & VISTTY))
1296 				return (0);
1297 			break;
1298 		}
1299 	}
1300 
1301 	if ((pledge & PLEDGE_ROUTE)) {
1302 		switch (com) {
1303 		case SIOCGIFADDR:
1304 		case SIOCGIFAFLAG_IN6:
1305 		case SIOCGIFALIFETIME_IN6:
1306 		case SIOCGIFDESCR:
1307 		case SIOCGIFFLAGS:
1308 		case SIOCGIFMETRIC:
1309 		case SIOCGIFGMEMB:
1310 		case SIOCGIFRDOMAIN:
1311 		case SIOCGIFDSTADDR_IN6:
1312 		case SIOCGIFNETMASK_IN6:
1313 		case SIOCGIFXFLAGS:
1314 		case SIOCGNBRINFO_IN6:
1315 		case SIOCGIFINFO_IN6:
1316 		case SIOCGIFMEDIA:
1317 			if (fp->f_type == DTYPE_SOCKET)
1318 				return (0);
1319 			break;
1320 		}
1321 	}
1322 
1323 	if ((pledge & PLEDGE_WROUTE)) {
1324 		switch (com) {
1325 		case SIOCAIFADDR:
1326 		case SIOCDIFADDR:
1327 		case SIOCAIFADDR_IN6:
1328 		case SIOCDIFADDR_IN6:
1329 			if (fp->f_type == DTYPE_SOCKET)
1330 				return (0);
1331 			break;
1332 		case SIOCSIFMTU:
1333 			if (fp->f_type == DTYPE_SOCKET)
1334 				return (0);
1335 			break;
1336 		}
1337 	}
1338 
1339 #if NVMM > 0
1340 	if ((pledge & PLEDGE_VMM)) {
1341 		if ((fp->f_type == DTYPE_VNODE) &&
1342 		    (vp->v_type == VCHR) &&
1343 		    (cdevsw[major(vp->v_rdev)].d_open == vmmopen)) {
1344 			error = pledge_ioctl_vmm(p, com);
1345 			if (error == 0)
1346 				return 0;
1347 		}
1348 	}
1349 #endif
1350 
1351 	return pledge_fail(p, error, PLEDGE_TTY);
1352 }
1353 
1354 int
1355 pledge_sockopt(struct proc *p, int set, int level, int optname)
1356 {
1357 	uint64_t pledge;
1358 
1359 	if ((p->p_p->ps_flags & PS_PLEDGE) == 0)
1360 		return (0);
1361 	pledge = READ_ONCE(p->p_p->ps_pledge);
1362 
1363 	/* Always allow these, which are too common to reject */
1364 	switch (level) {
1365 	case SOL_SOCKET:
1366 		switch (optname) {
1367 		case SO_RCVBUF:
1368 		case SO_ERROR:
1369 			return (0);
1370 		}
1371 		break;
1372 	case IPPROTO_TCP:
1373 		switch (optname) {
1374 		case TCP_NODELAY:
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)) == 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 	if ((pledge & (PLEDGE_INET|PLEDGE_UNIX)) == 0)
1415 		return pledge_fail(p, EPERM, PLEDGE_INET);
1416 	switch (level) {
1417 	case SOL_SOCKET:
1418 		switch (optname) {
1419 		case SO_RTABLE:
1420 			return pledge_fail(p, EINVAL, PLEDGE_WROUTE);
1421 		}
1422 		return (0);
1423 	}
1424 
1425 	if ((pledge & PLEDGE_INET) == 0)
1426 		return pledge_fail(p, EPERM, PLEDGE_INET);
1427 	switch (level) {
1428 	case IPPROTO_TCP:
1429 		switch (optname) {
1430 		case TCP_MD5SIG:
1431 		case TCP_SACK_ENABLE:
1432 		case TCP_MAXSEG:
1433 		case TCP_NOPUSH:
1434 		case TCP_INFO:
1435 			return (0);
1436 		}
1437 		break;
1438 	case IPPROTO_IP:
1439 		switch (optname) {
1440 		case IP_OPTIONS:
1441 			if (!set)
1442 				return (0);
1443 			break;
1444 		case IP_TOS:
1445 		case IP_TTL:
1446 		case IP_MINTTL:
1447 		case IP_IPDEFTTL:
1448 		case IP_PORTRANGE:
1449 		case IP_RECVDSTADDR:
1450 		case IP_RECVDSTPORT:
1451 			return (0);
1452 		case IP_MULTICAST_IF:
1453 		case IP_MULTICAST_TTL:
1454 		case IP_MULTICAST_LOOP:
1455 		case IP_ADD_MEMBERSHIP:
1456 		case IP_DROP_MEMBERSHIP:
1457 			if (pledge & PLEDGE_MCAST)
1458 				return (0);
1459 			break;
1460 		}
1461 		break;
1462 	case IPPROTO_ICMP:
1463 		break;
1464 	case IPPROTO_IPV6:
1465 		switch (optname) {
1466 		case IPV6_TCLASS:
1467 		case IPV6_UNICAST_HOPS:
1468 		case IPV6_MINHOPCOUNT:
1469 		case IPV6_RECVHOPLIMIT:
1470 		case IPV6_PORTRANGE:
1471 		case IPV6_RECVPKTINFO:
1472 		case IPV6_RECVDSTPORT:
1473 		case IPV6_V6ONLY:
1474 			return (0);
1475 		case IPV6_MULTICAST_IF:
1476 		case IPV6_MULTICAST_HOPS:
1477 		case IPV6_MULTICAST_LOOP:
1478 		case IPV6_JOIN_GROUP:
1479 		case IPV6_LEAVE_GROUP:
1480 			if (pledge & PLEDGE_MCAST)
1481 				return (0);
1482 			break;
1483 		}
1484 		break;
1485 	case IPPROTO_ICMPV6:
1486 		break;
1487 	}
1488 	return pledge_fail(p, EPERM, PLEDGE_INET);
1489 }
1490 
1491 int
1492 pledge_socket(struct proc *p, int domain, unsigned int state)
1493 {
1494 	uint64_t pledge;
1495 
1496 	if (!ISSET(p->p_p->ps_flags, PS_PLEDGE))
1497 		return 0;
1498 	pledge = READ_ONCE(p->p_p->ps_pledge);
1499 
1500 	if (ISSET(state, SS_DNS)) {
1501 		if (ISSET(pledge, PLEDGE_DNS))
1502 			return 0;
1503 		return pledge_fail(p, EPERM, PLEDGE_DNS);
1504 	}
1505 
1506 	switch (domain) {
1507 	case -1:		/* accept on any domain */
1508 		return (0);
1509 	case AF_INET:
1510 	case AF_INET6:
1511 		if (ISSET(pledge, PLEDGE_INET))
1512 			return 0;
1513 		return pledge_fail(p, EPERM, PLEDGE_INET);
1514 
1515 	case AF_UNIX:
1516 		if (ISSET(pledge, PLEDGE_UNIX))
1517 			return 0;
1518 		return pledge_fail(p, EPERM, PLEDGE_UNIX);
1519 	}
1520 
1521 	return pledge_fail(p, EINVAL, PLEDGE_INET);
1522 }
1523 
1524 int
1525 pledge_flock(struct proc *p)
1526 {
1527 	if ((p->p_p->ps_flags & PS_PLEDGE) == 0)
1528 		return (0);
1529 
1530 	if ((p->p_p->ps_pledge & PLEDGE_FLOCK))
1531 		return (0);
1532 	return (pledge_fail(p, EPERM, PLEDGE_FLOCK));
1533 }
1534 
1535 int
1536 pledge_swapctl(struct proc *p, int cmd)
1537 {
1538 	if ((p->p_p->ps_flags & PS_PLEDGE) == 0)
1539 		return (0);
1540 
1541 	if (p->p_p->ps_pledge & PLEDGE_VMINFO) {
1542 		switch (cmd) {
1543 		case SWAP_NSWAP:
1544 		case SWAP_STATS:
1545 			return (0);
1546 		}
1547 	}
1548 
1549 	return pledge_fail(p, EPERM, PLEDGE_VMINFO);
1550 }
1551 
1552 /* bsearch over pledgereq. return flags value if found, 0 else */
1553 uint64_t
1554 pledgereq_flags(const char *req_name)
1555 {
1556 	int base = 0, cmp, i, lim;
1557 
1558 	for (lim = nitems(pledgereq); lim != 0; lim >>= 1) {
1559 		i = base + (lim >> 1);
1560 		cmp = strcmp(req_name, pledgereq[i].name);
1561 		if (cmp == 0)
1562 			return (pledgereq[i].flags);
1563 		if (cmp > 0) { /* not found before, move right */
1564 			base = i + 1;
1565 			lim--;
1566 		} /* else move left */
1567 	}
1568 	return (0);
1569 }
1570 
1571 int
1572 pledge_fcntl(struct proc *p, int cmd)
1573 {
1574 	if ((p->p_p->ps_flags & PS_PLEDGE) == 0)
1575 		return (0);
1576 	if ((p->p_p->ps_pledge & PLEDGE_PROC) == 0 && cmd == F_SETOWN)
1577 		return pledge_fail(p, EPERM, PLEDGE_PROC);
1578 	return (0);
1579 }
1580 
1581 int
1582 pledge_kill(struct proc *p, pid_t pid)
1583 {
1584 	if ((p->p_p->ps_flags & PS_PLEDGE) == 0)
1585 		return 0;
1586 	if (p->p_p->ps_pledge & PLEDGE_PROC)
1587 		return 0;
1588 	if (pid == 0 || pid == p->p_p->ps_pid)
1589 		return 0;
1590 	return pledge_fail(p, EPERM, PLEDGE_PROC);
1591 }
1592 
1593 int
1594 pledge_profil(struct proc *p, u_int scale)
1595 {
1596 	if ((p->p_p->ps_flags & PS_PLEDGE) == 0)
1597 		return 0;
1598 	if (scale != 0)
1599 		return pledge_fail(p, EPERM, PLEDGE_STDIO);
1600 	return 0;
1601 }
1602 
1603 int
1604 pledge_protexec(struct proc *p, int prot)
1605 {
1606 	if ((p->p_p->ps_flags & PS_PLEDGE) == 0)
1607 		return 0;
1608 	/* Before kbind(2) call, ld.so and crt may create EXEC mappings */
1609 	if (p->p_p->ps_kbind_addr == 0 && p->p_p->ps_kbind_cookie == 0)
1610 		return 0;
1611 	if (!(p->p_p->ps_pledge & PLEDGE_PROTEXEC) && (prot & PROT_EXEC))
1612 		return pledge_fail(p, EPERM, PLEDGE_PROTEXEC);
1613 	return 0;
1614 }
1615 
1616 int
1617 canonpath(const char *input, char *buf, size_t bufsize)
1618 {
1619 	const char *p;
1620 	char *q;
1621 
1622 	/* can't canon relative paths, don't bother */
1623 	if (input[0] != '/') {
1624 		if (strlcpy(buf, input, bufsize) >= bufsize)
1625 			return ENAMETOOLONG;
1626 		return 0;
1627 	}
1628 
1629 	p = input;
1630 	q = buf;
1631 	while (*p && (q - buf < bufsize)) {
1632 		if (p[0] == '/' && (p[1] == '/' || p[1] == '\0')) {
1633 			p += 1;
1634 
1635 		} else if (p[0] == '/' && p[1] == '.' &&
1636 		    (p[2] == '/' || p[2] == '\0')) {
1637 			p += 2;
1638 
1639 		} else if (p[0] == '/' && p[1] == '.' && p[2] == '.' &&
1640 		    (p[3] == '/' || p[3] == '\0')) {
1641 			p += 3;
1642 			if (q != buf)	/* "/../" at start of buf */
1643 				while (*--q != '/')
1644 					continue;
1645 
1646 		} else {
1647 			*q++ = *p++;
1648 		}
1649 	}
1650 	if ((*p == '\0') && (q - buf < bufsize)) {
1651 		*q = 0;
1652 		return 0;
1653 	} else
1654 		return ENAMETOOLONG;
1655 }
1656