xref: /freebsd/sys/compat/linux/linux_misc.c (revision e41c28e67fac9cd22a85b160c5c9d0477ed03600)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 2002 Doug Rabson
5  * Copyright (c) 1994-1995 Søren Schmidt
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer
13  *    in this position and unchanged.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. The name of the author may not be used to endorse or promote products
18  *    derived from this software without specific prior written permission
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
21  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
22  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
23  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
24  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
25  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
26  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
27  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
28  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
29  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
30  */
31 
32 #include <sys/param.h>
33 #include <sys/fcntl.h>
34 #include <sys/jail.h>
35 #include <sys/imgact.h>
36 #include <sys/limits.h>
37 #include <sys/lock.h>
38 #include <sys/membarrier.h>
39 #include <sys/msgbuf.h>
40 #include <sys/mqueue.h>
41 #include <sys/mutex.h>
42 #include <sys/poll.h>
43 #include <sys/priv.h>
44 #include <sys/proc.h>
45 #include <sys/procctl.h>
46 #include <sys/reboot.h>
47 #include <sys/random.h>
48 #include <sys/resourcevar.h>
49 #include <sys/rtprio.h>
50 #include <sys/sched.h>
51 #include <sys/smp.h>
52 #include <sys/stat.h>
53 #include <sys/syscallsubr.h>
54 #include <sys/sysctl.h>
55 #include <sys/sysent.h>
56 #include <sys/sysproto.h>
57 #include <sys/time.h>
58 #include <sys/unistd.h>
59 #include <sys/vmmeter.h>
60 #include <sys/vnode.h>
61 
62 #include <security/audit/audit.h>
63 #include <security/mac/mac_framework.h>
64 
65 #include <vm/pmap.h>
66 #include <vm/vm_map.h>
67 #include <vm/swap_pager.h>
68 
69 #ifdef COMPAT_LINUX32
70 #include <machine/../linux32/linux.h>
71 #include <machine/../linux32/linux32_proto.h>
72 #else
73 #include <machine/../linux/linux.h>
74 #include <machine/../linux/linux_proto.h>
75 #endif
76 
77 #include <compat/linux/linux_common.h>
78 #include <compat/linux/linux_dtrace.h>
79 #include <compat/linux/linux_file.h>
80 #include <compat/linux/linux_mib.h>
81 #include <compat/linux/linux_mmap.h>
82 #include <compat/linux/linux_signal.h>
83 #include <compat/linux/linux_time.h>
84 #include <compat/linux/linux_util.h>
85 #include <compat/linux/linux_emul.h>
86 #include <compat/linux/linux_misc.h>
87 
88 int stclohz;				/* Statistics clock frequency */
89 
90 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
91 	RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
92 	RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
93 	RLIMIT_MEMLOCK, RLIMIT_AS
94 };
95 
96 struct l_sysinfo {
97 	l_long		uptime;		/* Seconds since boot */
98 	l_ulong		loads[3];	/* 1, 5, and 15 minute load averages */
99 #define LINUX_SYSINFO_LOADS_SCALE 65536
100 	l_ulong		totalram;	/* Total usable main memory size */
101 	l_ulong		freeram;	/* Available memory size */
102 	l_ulong		sharedram;	/* Amount of shared memory */
103 	l_ulong		bufferram;	/* Memory used by buffers */
104 	l_ulong		totalswap;	/* Total swap space size */
105 	l_ulong		freeswap;	/* swap space still available */
106 	l_ushort	procs;		/* Number of current processes */
107 	l_ushort	pads;
108 	l_ulong		totalhigh;
109 	l_ulong		freehigh;
110 	l_uint		mem_unit;
111 	char		_f[20-2*sizeof(l_long)-sizeof(l_int)];	/* padding */
112 };
113 
114 struct l_pselect6arg {
115 	l_uintptr_t	ss;
116 	l_size_t	ss_len;
117 };
118 
119 static int	linux_utimensat_lts_to_ts(struct l_timespec *,
120 			struct timespec *);
121 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
122 static int	linux_utimensat_lts64_to_ts(struct l_timespec64 *,
123 			struct timespec *);
124 #endif
125 static int	linux_common_utimensat(struct thread *, int,
126 			const char *, struct timespec *, int);
127 static int	linux_common_pselect6(struct thread *, l_int,
128 			l_fd_set *, l_fd_set *, l_fd_set *,
129 			struct timespec *, l_uintptr_t *);
130 static int	linux_common_ppoll(struct thread *, struct pollfd *,
131 			uint32_t, struct timespec *, l_sigset_t *,
132 			l_size_t);
133 static int	linux_pollin(struct thread *, struct pollfd *,
134 			struct pollfd *, u_int);
135 static int	linux_pollout(struct thread *, struct pollfd *,
136 			struct pollfd *, u_int);
137 
138 int
139 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
140 {
141 	struct l_sysinfo sysinfo;
142 	int i, j;
143 	struct timespec ts;
144 
145 	bzero(&sysinfo, sizeof(sysinfo));
146 	getnanouptime(&ts);
147 	if (ts.tv_nsec != 0)
148 		ts.tv_sec++;
149 	sysinfo.uptime = ts.tv_sec;
150 
151 	/* Use the information from the mib to get our load averages */
152 	for (i = 0; i < 3; i++)
153 		sysinfo.loads[i] = averunnable.ldavg[i] *
154 		    LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
155 
156 	sysinfo.totalram = physmem * PAGE_SIZE;
157 	sysinfo.freeram = (u_long)vm_free_count() * PAGE_SIZE;
158 
159 	/*
160 	 * sharedram counts pages allocated to named, swap-backed objects such
161 	 * as shared memory segments and tmpfs files.  There is no cheap way to
162 	 * compute this, so just leave the field unpopulated.  Linux itself only
163 	 * started setting this field in the 3.x timeframe.
164 	 */
165 	sysinfo.sharedram = 0;
166 	sysinfo.bufferram = 0;
167 
168 	swap_pager_status(&i, &j);
169 	sysinfo.totalswap = i * PAGE_SIZE;
170 	sysinfo.freeswap = (i - j) * PAGE_SIZE;
171 
172 	sysinfo.procs = nprocs;
173 
174 	/*
175 	 * Platforms supported by the emulation layer do not have a notion of
176 	 * high memory.
177 	 */
178 	sysinfo.totalhigh = 0;
179 	sysinfo.freehigh = 0;
180 
181 	sysinfo.mem_unit = 1;
182 
183 	return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
184 }
185 
186 #ifdef LINUX_LEGACY_SYSCALLS
187 int
188 linux_alarm(struct thread *td, struct linux_alarm_args *args)
189 {
190 	struct itimerval it, old_it;
191 	u_int secs;
192 	int error __diagused;
193 
194 	secs = args->secs;
195 	/*
196 	 * Linux alarm() is always successful. Limit secs to INT32_MAX / 2
197 	 * to match kern_setitimer()'s limit to avoid error from it.
198 	 *
199 	 * XXX. Linux limit secs to INT_MAX on 32 and does not limit on 64-bit
200 	 * platforms.
201 	 */
202 	if (secs > INT32_MAX / 2)
203 		secs = INT32_MAX / 2;
204 
205 	it.it_value.tv_sec = secs;
206 	it.it_value.tv_usec = 0;
207 	timevalclear(&it.it_interval);
208 	error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
209 	KASSERT(error == 0, ("kern_setitimer returns %d", error));
210 
211 	if ((old_it.it_value.tv_sec == 0 && old_it.it_value.tv_usec > 0) ||
212 	    old_it.it_value.tv_usec >= 500000)
213 		old_it.it_value.tv_sec++;
214 	td->td_retval[0] = old_it.it_value.tv_sec;
215 	return (0);
216 }
217 #endif
218 
219 int
220 linux_brk(struct thread *td, struct linux_brk_args *args)
221 {
222 	struct vmspace *vm = td->td_proc->p_vmspace;
223 	uintptr_t new, old;
224 
225 	old = (uintptr_t)vm->vm_daddr + ctob(vm->vm_dsize);
226 	new = (uintptr_t)args->dsend;
227 	if ((caddr_t)new > vm->vm_daddr && !kern_break(td, &new))
228 		td->td_retval[0] = (register_t)new;
229 	else
230 		td->td_retval[0] = (register_t)old;
231 
232 	return (0);
233 }
234 
235 #ifdef LINUX_LEGACY_SYSCALLS
236 int
237 linux_select(struct thread *td, struct linux_select_args *args)
238 {
239 	l_timeval ltv;
240 	struct timeval tv0, tv1, utv, *tvp;
241 	int error;
242 
243 	/*
244 	 * Store current time for computation of the amount of
245 	 * time left.
246 	 */
247 	if (args->timeout) {
248 		if ((error = copyin(args->timeout, &ltv, sizeof(ltv))))
249 			goto select_out;
250 		utv.tv_sec = ltv.tv_sec;
251 		utv.tv_usec = ltv.tv_usec;
252 
253 		if (itimerfix(&utv)) {
254 			/*
255 			 * The timeval was invalid.  Convert it to something
256 			 * valid that will act as it does under Linux.
257 			 */
258 			utv.tv_sec += utv.tv_usec / 1000000;
259 			utv.tv_usec %= 1000000;
260 			if (utv.tv_usec < 0) {
261 				utv.tv_sec -= 1;
262 				utv.tv_usec += 1000000;
263 			}
264 			if (utv.tv_sec < 0)
265 				timevalclear(&utv);
266 		}
267 		microtime(&tv0);
268 		tvp = &utv;
269 	} else
270 		tvp = NULL;
271 
272 	error = kern_select(td, args->nfds, args->readfds, args->writefds,
273 	    args->exceptfds, tvp, LINUX_NFDBITS);
274 	if (error)
275 		goto select_out;
276 
277 	if (args->timeout) {
278 		if (td->td_retval[0]) {
279 			/*
280 			 * Compute how much time was left of the timeout,
281 			 * by subtracting the current time and the time
282 			 * before we started the call, and subtracting
283 			 * that result from the user-supplied value.
284 			 */
285 			microtime(&tv1);
286 			timevalsub(&tv1, &tv0);
287 			timevalsub(&utv, &tv1);
288 			if (utv.tv_sec < 0)
289 				timevalclear(&utv);
290 		} else
291 			timevalclear(&utv);
292 		ltv.tv_sec = utv.tv_sec;
293 		ltv.tv_usec = utv.tv_usec;
294 		if ((error = copyout(&ltv, args->timeout, sizeof(ltv))))
295 			goto select_out;
296 	}
297 
298 select_out:
299 	return (error);
300 }
301 #endif
302 
303 int
304 linux_mremap(struct thread *td, struct linux_mremap_args *args)
305 {
306 	uintptr_t addr;
307 	size_t len;
308 	int error = 0;
309 
310 	if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
311 		td->td_retval[0] = 0;
312 		return (EINVAL);
313 	}
314 
315 	/*
316 	 * Check for the page alignment.
317 	 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
318 	 */
319 	if (args->addr & PAGE_MASK) {
320 		td->td_retval[0] = 0;
321 		return (EINVAL);
322 	}
323 
324 	args->new_len = round_page(args->new_len);
325 	args->old_len = round_page(args->old_len);
326 
327 	if (args->new_len > args->old_len) {
328 		td->td_retval[0] = 0;
329 		return (ENOMEM);
330 	}
331 
332 	if (args->new_len < args->old_len) {
333 		addr = args->addr + args->new_len;
334 		len = args->old_len - args->new_len;
335 		error = kern_munmap(td, addr, len);
336 	}
337 
338 	td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
339 	return (error);
340 }
341 
342 #define LINUX_MS_ASYNC       0x0001
343 #define LINUX_MS_INVALIDATE  0x0002
344 #define LINUX_MS_SYNC        0x0004
345 
346 int
347 linux_msync(struct thread *td, struct linux_msync_args *args)
348 {
349 
350 	return (kern_msync(td, args->addr, args->len,
351 	    args->fl & ~LINUX_MS_SYNC));
352 }
353 
354 int
355 linux_mprotect(struct thread *td, struct linux_mprotect_args *uap)
356 {
357 
358 	return (linux_mprotect_common(td, PTROUT(uap->addr), uap->len,
359 	    uap->prot));
360 }
361 
362 int
363 linux_madvise(struct thread *td, struct linux_madvise_args *uap)
364 {
365 
366 	return (linux_madvise_common(td, PTROUT(uap->addr), uap->len,
367 	    uap->behav));
368 }
369 
370 int
371 linux_mmap2(struct thread *td, struct linux_mmap2_args *uap)
372 {
373 #if defined(LINUX_ARCHWANT_MMAP2PGOFF)
374 	/*
375 	 * For architectures with sizeof (off_t) < sizeof (loff_t) mmap is
376 	 * implemented with mmap2 syscall and the offset is represented in
377 	 * multiples of page size.
378 	 */
379 	return (linux_mmap_common(td, PTROUT(uap->addr), uap->len, uap->prot,
380 	    uap->flags, uap->fd, (uint64_t)(uint32_t)uap->pgoff * PAGE_SIZE));
381 #else
382 	return (linux_mmap_common(td, PTROUT(uap->addr), uap->len, uap->prot,
383 	    uap->flags, uap->fd, uap->pgoff));
384 #endif
385 }
386 
387 #ifdef LINUX_LEGACY_SYSCALLS
388 int
389 linux_time(struct thread *td, struct linux_time_args *args)
390 {
391 	struct timeval tv;
392 	l_time_t tm;
393 	int error;
394 
395 	microtime(&tv);
396 	tm = tv.tv_sec;
397 	if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
398 		return (error);
399 	td->td_retval[0] = tm;
400 	return (0);
401 }
402 #endif
403 
404 struct l_times_argv {
405 	l_clock_t	tms_utime;
406 	l_clock_t	tms_stime;
407 	l_clock_t	tms_cutime;
408 	l_clock_t	tms_cstime;
409 };
410 
411 /*
412  * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
413  * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
414  * auxiliary vector entry.
415  */
416 #define	CLK_TCK		100
417 
418 #define	CONVOTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
419 #define	CONVNTCK(r)	(r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
420 
421 #define	CONVTCK(r)	(linux_kernver(td) >= LINUX_KERNVER(2,4,0) ?	\
422 			    CONVNTCK(r) : CONVOTCK(r))
423 
424 int
425 linux_times(struct thread *td, struct linux_times_args *args)
426 {
427 	struct timeval tv, utime, stime, cutime, cstime;
428 	struct l_times_argv tms;
429 	struct proc *p;
430 	int error;
431 
432 	if (args->buf != NULL) {
433 		p = td->td_proc;
434 		PROC_LOCK(p);
435 		PROC_STATLOCK(p);
436 		calcru(p, &utime, &stime);
437 		PROC_STATUNLOCK(p);
438 		calccru(p, &cutime, &cstime);
439 		PROC_UNLOCK(p);
440 
441 		tms.tms_utime = CONVTCK(utime);
442 		tms.tms_stime = CONVTCK(stime);
443 
444 		tms.tms_cutime = CONVTCK(cutime);
445 		tms.tms_cstime = CONVTCK(cstime);
446 
447 		if ((error = copyout(&tms, args->buf, sizeof(tms))))
448 			return (error);
449 	}
450 
451 	microuptime(&tv);
452 	td->td_retval[0] = (int)CONVTCK(tv);
453 	return (0);
454 }
455 
456 int
457 linux_newuname(struct thread *td, struct linux_newuname_args *args)
458 {
459 	struct l_new_utsname utsname;
460 	char osname[LINUX_MAX_UTSNAME];
461 	char osrelease[LINUX_MAX_UTSNAME];
462 	char *p;
463 
464 	linux_get_osname(td, osname);
465 	linux_get_osrelease(td, osrelease);
466 
467 	bzero(&utsname, sizeof(utsname));
468 	strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
469 	getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
470 	getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
471 	strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
472 	strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
473 	for (p = utsname.version; *p != '\0'; ++p)
474 		if (*p == '\n') {
475 			*p = '\0';
476 			break;
477 		}
478 #if defined(__amd64__)
479 	/*
480 	 * On amd64, Linux uname(2) needs to return "x86_64"
481 	 * for both 64-bit and 32-bit applications.  On 32-bit,
482 	 * the string returned by getauxval(AT_PLATFORM) needs
483 	 * to remain "i686", though.
484 	 */
485 #if defined(COMPAT_LINUX32)
486 	if (linux32_emulate_i386)
487 		strlcpy(utsname.machine, "i686", LINUX_MAX_UTSNAME);
488 	else
489 #endif
490 	strlcpy(utsname.machine, "x86_64", LINUX_MAX_UTSNAME);
491 #elif defined(__aarch64__)
492 	strlcpy(utsname.machine, "aarch64", LINUX_MAX_UTSNAME);
493 #elif defined(__i386__)
494 	strlcpy(utsname.machine, "i686", LINUX_MAX_UTSNAME);
495 #endif
496 
497 	return (copyout(&utsname, args->buf, sizeof(utsname)));
498 }
499 
500 struct l_utimbuf {
501 	l_time_t l_actime;
502 	l_time_t l_modtime;
503 };
504 
505 #ifdef LINUX_LEGACY_SYSCALLS
506 int
507 linux_utime(struct thread *td, struct linux_utime_args *args)
508 {
509 	struct timeval tv[2], *tvp;
510 	struct l_utimbuf lut;
511 	int error;
512 
513 	if (args->times) {
514 		if ((error = copyin(args->times, &lut, sizeof lut)) != 0)
515 			return (error);
516 		tv[0].tv_sec = lut.l_actime;
517 		tv[0].tv_usec = 0;
518 		tv[1].tv_sec = lut.l_modtime;
519 		tv[1].tv_usec = 0;
520 		tvp = tv;
521 	} else
522 		tvp = NULL;
523 
524 	return (kern_utimesat(td, AT_FDCWD, args->fname, UIO_USERSPACE,
525 	    tvp, UIO_SYSSPACE));
526 }
527 #endif
528 
529 #ifdef LINUX_LEGACY_SYSCALLS
530 int
531 linux_utimes(struct thread *td, struct linux_utimes_args *args)
532 {
533 	l_timeval ltv[2];
534 	struct timeval tv[2], *tvp = NULL;
535 	int error;
536 
537 	if (args->tptr != NULL) {
538 		if ((error = copyin(args->tptr, ltv, sizeof ltv)) != 0)
539 			return (error);
540 		tv[0].tv_sec = ltv[0].tv_sec;
541 		tv[0].tv_usec = ltv[0].tv_usec;
542 		tv[1].tv_sec = ltv[1].tv_sec;
543 		tv[1].tv_usec = ltv[1].tv_usec;
544 		tvp = tv;
545 	}
546 
547 	return (kern_utimesat(td, AT_FDCWD, args->fname, UIO_USERSPACE,
548 	    tvp, UIO_SYSSPACE));
549 }
550 #endif
551 
552 static int
553 linux_utimensat_lts_to_ts(struct l_timespec *l_times, struct timespec *times)
554 {
555 
556 	if (l_times->tv_nsec != LINUX_UTIME_OMIT &&
557 	    l_times->tv_nsec != LINUX_UTIME_NOW &&
558 	    (l_times->tv_nsec < 0 || l_times->tv_nsec > 999999999))
559 		return (EINVAL);
560 
561 	times->tv_sec = l_times->tv_sec;
562 	switch (l_times->tv_nsec)
563 	{
564 	case LINUX_UTIME_OMIT:
565 		times->tv_nsec = UTIME_OMIT;
566 		break;
567 	case LINUX_UTIME_NOW:
568 		times->tv_nsec = UTIME_NOW;
569 		break;
570 	default:
571 		times->tv_nsec = l_times->tv_nsec;
572 	}
573 
574 	return (0);
575 }
576 
577 static int
578 linux_common_utimensat(struct thread *td, int ldfd, const char *pathname,
579     struct timespec *timesp, int lflags)
580 {
581 	int dfd, flags = 0;
582 
583 	dfd = (ldfd == LINUX_AT_FDCWD) ? AT_FDCWD : ldfd;
584 
585 	if (lflags & ~(LINUX_AT_SYMLINK_NOFOLLOW | LINUX_AT_EMPTY_PATH))
586 		return (EINVAL);
587 
588 	if (timesp != NULL) {
589 		/* This breaks POSIX, but is what the Linux kernel does
590 		 * _on purpose_ (documented in the man page for utimensat(2)),
591 		 * so we must follow that behaviour. */
592 		if (timesp[0].tv_nsec == UTIME_OMIT &&
593 		    timesp[1].tv_nsec == UTIME_OMIT)
594 			return (0);
595 	}
596 
597 	if (lflags & LINUX_AT_SYMLINK_NOFOLLOW)
598 		flags |= AT_SYMLINK_NOFOLLOW;
599 	if (lflags & LINUX_AT_EMPTY_PATH)
600 		flags |= AT_EMPTY_PATH;
601 
602 	if (pathname != NULL)
603 		return (kern_utimensat(td, dfd, pathname,
604 		    UIO_USERSPACE, timesp, UIO_SYSSPACE, flags));
605 
606 	if (lflags != 0)
607 		return (EINVAL);
608 
609 	return (kern_futimens(td, dfd, timesp, UIO_SYSSPACE));
610 }
611 
612 int
613 linux_utimensat(struct thread *td, struct linux_utimensat_args *args)
614 {
615 	struct l_timespec l_times[2];
616 	struct timespec times[2], *timesp;
617 	int error;
618 
619 	if (args->times != NULL) {
620 		error = copyin(args->times, l_times, sizeof(l_times));
621 		if (error != 0)
622 			return (error);
623 
624 		error = linux_utimensat_lts_to_ts(&l_times[0], &times[0]);
625 		if (error != 0)
626 			return (error);
627 		error = linux_utimensat_lts_to_ts(&l_times[1], &times[1]);
628 		if (error != 0)
629 			return (error);
630 		timesp = times;
631 	} else
632 		timesp = NULL;
633 
634 	return (linux_common_utimensat(td, args->dfd, args->pathname,
635 	    timesp, args->flags));
636 }
637 
638 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
639 static int
640 linux_utimensat_lts64_to_ts(struct l_timespec64 *l_times, struct timespec *times)
641 {
642 
643 	/* Zero out the padding in compat mode. */
644 	l_times->tv_nsec &= 0xFFFFFFFFUL;
645 
646 	if (l_times->tv_nsec != LINUX_UTIME_OMIT &&
647 	    l_times->tv_nsec != LINUX_UTIME_NOW &&
648 	    (l_times->tv_nsec < 0 || l_times->tv_nsec > 999999999))
649 		return (EINVAL);
650 
651 	times->tv_sec = l_times->tv_sec;
652 	switch (l_times->tv_nsec)
653 	{
654 	case LINUX_UTIME_OMIT:
655 		times->tv_nsec = UTIME_OMIT;
656 		break;
657 	case LINUX_UTIME_NOW:
658 		times->tv_nsec = UTIME_NOW;
659 		break;
660 	default:
661 		times->tv_nsec = l_times->tv_nsec;
662 	}
663 
664 	return (0);
665 }
666 
667 int
668 linux_utimensat_time64(struct thread *td, struct linux_utimensat_time64_args *args)
669 {
670 	struct l_timespec64 l_times[2];
671 	struct timespec times[2], *timesp;
672 	int error;
673 
674 	if (args->times64 != NULL) {
675 		error = copyin(args->times64, l_times, sizeof(l_times));
676 		if (error != 0)
677 			return (error);
678 
679 		error = linux_utimensat_lts64_to_ts(&l_times[0], &times[0]);
680 		if (error != 0)
681 			return (error);
682 		error = linux_utimensat_lts64_to_ts(&l_times[1], &times[1]);
683 		if (error != 0)
684 			return (error);
685 		timesp = times;
686 	} else
687 		timesp = NULL;
688 
689 	return (linux_common_utimensat(td, args->dfd, args->pathname,
690 	    timesp, args->flags));
691 }
692 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
693 
694 #ifdef LINUX_LEGACY_SYSCALLS
695 int
696 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
697 {
698 	l_timeval ltv[2];
699 	struct timeval tv[2], *tvp = NULL;
700 	int error, dfd;
701 
702 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
703 
704 	if (args->utimes != NULL) {
705 		if ((error = copyin(args->utimes, ltv, sizeof ltv)) != 0)
706 			return (error);
707 		tv[0].tv_sec = ltv[0].tv_sec;
708 		tv[0].tv_usec = ltv[0].tv_usec;
709 		tv[1].tv_sec = ltv[1].tv_sec;
710 		tv[1].tv_usec = ltv[1].tv_usec;
711 		tvp = tv;
712 	}
713 
714 	return (kern_utimesat(td, dfd, args->filename, UIO_USERSPACE,
715 	    tvp, UIO_SYSSPACE));
716 }
717 #endif
718 
719 static int
720 linux_common_wait(struct thread *td, idtype_t idtype, int id, int *statusp,
721     int options, void *rup, l_siginfo_t *infop)
722 {
723 	l_siginfo_t lsi;
724 	siginfo_t siginfo;
725 	struct __wrusage wru;
726 	int error, status, tmpstat, sig;
727 
728 	error = kern_wait6(td, idtype, id, &status, options,
729 	    rup != NULL ? &wru : NULL, &siginfo);
730 
731 	if (error == 0 && statusp) {
732 		tmpstat = status & 0xffff;
733 		if (WIFSIGNALED(tmpstat)) {
734 			tmpstat = (tmpstat & 0xffffff80) |
735 			    bsd_to_linux_signal(WTERMSIG(tmpstat));
736 		} else if (WIFSTOPPED(tmpstat)) {
737 			tmpstat = (tmpstat & 0xffff00ff) |
738 			    (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8);
739 #if defined(__aarch64__) || (defined(__amd64__) && !defined(COMPAT_LINUX32))
740 			if (WSTOPSIG(status) == SIGTRAP) {
741 				tmpstat = linux_ptrace_status(td,
742 				    siginfo.si_pid, tmpstat);
743 			}
744 #endif
745 		} else if (WIFCONTINUED(tmpstat)) {
746 			tmpstat = 0xffff;
747 		}
748 		error = copyout(&tmpstat, statusp, sizeof(int));
749 	}
750 	if (error == 0 && rup != NULL)
751 		error = linux_copyout_rusage(&wru.wru_self, rup);
752 	if (error == 0 && infop != NULL && td->td_retval[0] != 0) {
753 		sig = bsd_to_linux_signal(siginfo.si_signo);
754 		memset(&lsi, 0, sizeof(lsi));
755 		siginfo_to_lsiginfo(&siginfo, &lsi, sig);
756 		error = copyout(&lsi, infop, sizeof(lsi));
757 	}
758 
759 	return (error);
760 }
761 
762 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
763 int
764 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
765 {
766 	struct linux_wait4_args wait4_args = {
767 		.pid = args->pid,
768 		.status = args->status,
769 		.options = args->options,
770 		.rusage = NULL,
771 	};
772 
773 	return (linux_wait4(td, &wait4_args));
774 }
775 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
776 
777 int
778 linux_wait4(struct thread *td, struct linux_wait4_args *args)
779 {
780 	struct proc *p;
781 	int options, id, idtype;
782 
783 	if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG |
784 	    LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
785 		return (EINVAL);
786 
787 	/* -INT_MIN is not defined. */
788 	if (args->pid == INT_MIN)
789 		return (ESRCH);
790 
791 	options = 0;
792 	linux_to_bsd_waitopts(args->options, &options);
793 
794 	/*
795 	 * For backward compatibility we implicitly add flags WEXITED
796 	 * and WTRAPPED here.
797 	 */
798 	options |= WEXITED | WTRAPPED;
799 
800 	if (args->pid == WAIT_ANY) {
801 		idtype = P_ALL;
802 		id = 0;
803 	} else if (args->pid < 0) {
804 		idtype = P_PGID;
805 		id = (id_t)-args->pid;
806 	} else if (args->pid == 0) {
807 		idtype = P_PGID;
808 		p = td->td_proc;
809 		PROC_LOCK(p);
810 		id = p->p_pgid;
811 		PROC_UNLOCK(p);
812 	} else {
813 		idtype = P_PID;
814 		id = (id_t)args->pid;
815 	}
816 
817 	return (linux_common_wait(td, idtype, id, args->status, options,
818 	    args->rusage, NULL));
819 }
820 
821 int
822 linux_waitid(struct thread *td, struct linux_waitid_args *args)
823 {
824 	idtype_t idtype;
825 	int error, options;
826 	struct proc *p;
827 	pid_t id;
828 
829 	if (args->options & ~(LINUX_WNOHANG | LINUX_WNOWAIT | LINUX_WEXITED |
830 	    LINUX_WSTOPPED | LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
831 		return (EINVAL);
832 
833 	options = 0;
834 	linux_to_bsd_waitopts(args->options, &options);
835 
836 	id = args->id;
837 	switch (args->idtype) {
838 	case LINUX_P_ALL:
839 		idtype = P_ALL;
840 		break;
841 	case LINUX_P_PID:
842 		if (args->id <= 0)
843 			return (EINVAL);
844 		idtype = P_PID;
845 		break;
846 	case LINUX_P_PGID:
847 		if (linux_kernver(td) >= LINUX_KERNVER(5,4,0) && args->id == 0) {
848 			p = td->td_proc;
849 			PROC_LOCK(p);
850 			id = p->p_pgid;
851 			PROC_UNLOCK(p);
852 		} else if (args->id <= 0)
853 			return (EINVAL);
854 		idtype = P_PGID;
855 		break;
856 	case LINUX_P_PIDFD:
857 		LINUX_RATELIMIT_MSG("unsupported waitid P_PIDFD idtype");
858 		return (ENOSYS);
859 	default:
860 		return (EINVAL);
861 	}
862 
863 	error = linux_common_wait(td, idtype, id, NULL, options,
864 	    args->rusage, args->info);
865 	td->td_retval[0] = 0;
866 
867 	return (error);
868 }
869 
870 #ifdef LINUX_LEGACY_SYSCALLS
871 int
872 linux_mknod(struct thread *td, struct linux_mknod_args *args)
873 {
874 	int error;
875 
876 	switch (args->mode & S_IFMT) {
877 	case S_IFIFO:
878 	case S_IFSOCK:
879 		error = kern_mkfifoat(td, AT_FDCWD, args->path, UIO_USERSPACE,
880 		    args->mode);
881 		break;
882 
883 	case S_IFCHR:
884 	case S_IFBLK:
885 		error = kern_mknodat(td, AT_FDCWD, args->path, UIO_USERSPACE,
886 		    args->mode, linux_decode_dev(args->dev));
887 		break;
888 
889 	case S_IFDIR:
890 		error = EPERM;
891 		break;
892 
893 	case 0:
894 		args->mode |= S_IFREG;
895 		/* FALLTHROUGH */
896 	case S_IFREG:
897 		error = kern_openat(td, AT_FDCWD, args->path, UIO_USERSPACE,
898 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
899 		if (error == 0)
900 			kern_close(td, td->td_retval[0]);
901 		break;
902 
903 	default:
904 		error = EINVAL;
905 		break;
906 	}
907 	return (error);
908 }
909 #endif
910 
911 int
912 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
913 {
914 	int error, dfd;
915 
916 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
917 
918 	switch (args->mode & S_IFMT) {
919 	case S_IFIFO:
920 	case S_IFSOCK:
921 		error = kern_mkfifoat(td, dfd, args->filename, UIO_USERSPACE,
922 		    args->mode);
923 		break;
924 
925 	case S_IFCHR:
926 	case S_IFBLK:
927 		error = kern_mknodat(td, dfd, args->filename, UIO_USERSPACE,
928 		    args->mode, linux_decode_dev(args->dev));
929 		break;
930 
931 	case S_IFDIR:
932 		error = EPERM;
933 		break;
934 
935 	case 0:
936 		args->mode |= S_IFREG;
937 		/* FALLTHROUGH */
938 	case S_IFREG:
939 		error = kern_openat(td, dfd, args->filename, UIO_USERSPACE,
940 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
941 		if (error == 0)
942 			kern_close(td, td->td_retval[0]);
943 		break;
944 
945 	default:
946 		error = EINVAL;
947 		break;
948 	}
949 	return (error);
950 }
951 
952 /*
953  * UGH! This is just about the dumbest idea I've ever heard!!
954  */
955 int
956 linux_personality(struct thread *td, struct linux_personality_args *args)
957 {
958 	struct linux_pemuldata *pem;
959 	struct proc *p = td->td_proc;
960 	uint32_t old;
961 
962 	PROC_LOCK(p);
963 	pem = pem_find(p);
964 	old = pem->persona;
965 	if (args->per != 0xffffffff)
966 		pem->persona = args->per;
967 	PROC_UNLOCK(p);
968 
969 	td->td_retval[0] = old;
970 	return (0);
971 }
972 
973 struct l_itimerval {
974 	l_timeval it_interval;
975 	l_timeval it_value;
976 };
977 
978 #define	B2L_ITIMERVAL(bip, lip)						\
979 	(bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec;		\
980 	(bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec;	\
981 	(bip)->it_value.tv_sec = (lip)->it_value.tv_sec;		\
982 	(bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
983 
984 int
985 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
986 {
987 	int error;
988 	struct l_itimerval ls;
989 	struct itimerval aitv, oitv;
990 
991 	if (uap->itv == NULL) {
992 		uap->itv = uap->oitv;
993 		return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
994 	}
995 
996 	error = copyin(uap->itv, &ls, sizeof(ls));
997 	if (error != 0)
998 		return (error);
999 	B2L_ITIMERVAL(&aitv, &ls);
1000 	error = kern_setitimer(td, uap->which, &aitv, &oitv);
1001 	if (error != 0 || uap->oitv == NULL)
1002 		return (error);
1003 	B2L_ITIMERVAL(&ls, &oitv);
1004 
1005 	return (copyout(&ls, uap->oitv, sizeof(ls)));
1006 }
1007 
1008 int
1009 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1010 {
1011 	int error;
1012 	struct l_itimerval ls;
1013 	struct itimerval aitv;
1014 
1015 	error = kern_getitimer(td, uap->which, &aitv);
1016 	if (error != 0)
1017 		return (error);
1018 	B2L_ITIMERVAL(&ls, &aitv);
1019 	return (copyout(&ls, uap->itv, sizeof(ls)));
1020 }
1021 
1022 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1023 int
1024 linux_nice(struct thread *td, struct linux_nice_args *args)
1025 {
1026 
1027 	return (kern_setpriority(td, PRIO_PROCESS, 0, args->inc));
1028 }
1029 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1030 
1031 int
1032 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1033 {
1034 	const int ngrp = args->gidsetsize;
1035 	struct ucred *newcred, *oldcred;
1036 	l_gid_t *linux_gidset;
1037 	int error;
1038 	struct proc *p;
1039 
1040 	if (ngrp < 0 || ngrp > ngroups_max)
1041 		return (EINVAL);
1042 	linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK);
1043 	error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1044 	if (error)
1045 		goto out;
1046 
1047 	newcred = crget();
1048 	crextend(newcred, ngrp);
1049 	p = td->td_proc;
1050 	PROC_LOCK(p);
1051 	oldcred = crcopysafe(p, newcred);
1052 
1053 	if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS)) != 0) {
1054 		PROC_UNLOCK(p);
1055 		crfree(newcred);
1056 		goto out;
1057 	}
1058 
1059 	newcred->cr_ngroups = ngrp;
1060 	for (int i = 0; i < ngrp; i++)
1061 		newcred->cr_groups[i] = linux_gidset[i];
1062 	newcred->cr_flags |= CRED_FLAG_GROUPSET;
1063 
1064 	setsugid(p);
1065 	proc_set_cred(p, newcred);
1066 	PROC_UNLOCK(p);
1067 	crfree(oldcred);
1068 	error = 0;
1069 out:
1070 	free(linux_gidset, M_LINUX);
1071 	return (error);
1072 }
1073 
1074 int
1075 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1076 {
1077 	const struct ucred *const cred = td->td_ucred;
1078 	l_gid_t *linux_gidset;
1079 	int ngrp, error;
1080 
1081 	ngrp = args->gidsetsize;
1082 
1083 	if (ngrp == 0) {
1084 		td->td_retval[0] = cred->cr_ngroups;
1085 		return (0);
1086 	}
1087 	if (ngrp < cred->cr_ngroups)
1088 		return (EINVAL);
1089 
1090 	ngrp = cred->cr_ngroups;
1091 
1092 	linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK);
1093 	for (int i = 0; i < ngrp; ++i)
1094 		linux_gidset[i] = cred->cr_groups[i];
1095 
1096 	error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1097 	free(linux_gidset, M_LINUX);
1098 
1099 	if (error != 0)
1100 		return (error);
1101 
1102 	td->td_retval[0] = ngrp;
1103 	return (0);
1104 }
1105 
1106 static bool
1107 linux_get_dummy_limit(struct thread *td, l_uint resource, struct rlimit *rlim)
1108 {
1109 	ssize_t size;
1110 	int res, error;
1111 
1112 	if (linux_dummy_rlimits == 0)
1113 		return (false);
1114 
1115 	switch (resource) {
1116 	case LINUX_RLIMIT_LOCKS:
1117 	case LINUX_RLIMIT_RTTIME:
1118 		rlim->rlim_cur = LINUX_RLIM_INFINITY;
1119 		rlim->rlim_max = LINUX_RLIM_INFINITY;
1120 		return (true);
1121 	case LINUX_RLIMIT_NICE:
1122 	case LINUX_RLIMIT_RTPRIO:
1123 		rlim->rlim_cur = 0;
1124 		rlim->rlim_max = 0;
1125 		return (true);
1126 	case LINUX_RLIMIT_SIGPENDING:
1127 		error = kernel_sysctlbyname(td,
1128 		    "kern.sigqueue.max_pending_per_proc",
1129 		    &res, &size, 0, 0, 0, 0);
1130 		if (error != 0)
1131 			return (false);
1132 		rlim->rlim_cur = res;
1133 		rlim->rlim_max = res;
1134 		return (true);
1135 	case LINUX_RLIMIT_MSGQUEUE:
1136 		error = kernel_sysctlbyname(td,
1137 		    "kern.ipc.msgmnb", &res, &size, 0, 0, 0, 0);
1138 		if (error != 0)
1139 			return (false);
1140 		rlim->rlim_cur = res;
1141 		rlim->rlim_max = res;
1142 		return (true);
1143 	default:
1144 		return (false);
1145 	}
1146 }
1147 
1148 int
1149 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1150 {
1151 	struct rlimit bsd_rlim;
1152 	struct l_rlimit rlim;
1153 	u_int which;
1154 	int error;
1155 
1156 	if (args->resource >= LINUX_RLIM_NLIMITS)
1157 		return (EINVAL);
1158 
1159 	which = linux_to_bsd_resource[args->resource];
1160 	if (which == -1)
1161 		return (EINVAL);
1162 
1163 	error = copyin(args->rlim, &rlim, sizeof(rlim));
1164 	if (error)
1165 		return (error);
1166 
1167 	bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1168 	bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1169 	return (kern_setrlimit(td, which, &bsd_rlim));
1170 }
1171 
1172 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1173 int
1174 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1175 {
1176 	struct l_rlimit rlim;
1177 	struct rlimit bsd_rlim;
1178 	u_int which;
1179 
1180 	if (linux_get_dummy_limit(td, args->resource, &bsd_rlim)) {
1181 		rlim.rlim_cur = bsd_rlim.rlim_cur;
1182 		rlim.rlim_max = bsd_rlim.rlim_max;
1183 		return (copyout(&rlim, args->rlim, sizeof(rlim)));
1184 	}
1185 
1186 	if (args->resource >= LINUX_RLIM_NLIMITS)
1187 		return (EINVAL);
1188 
1189 	which = linux_to_bsd_resource[args->resource];
1190 	if (which == -1)
1191 		return (EINVAL);
1192 
1193 	lim_rlimit(td, which, &bsd_rlim);
1194 
1195 #ifdef COMPAT_LINUX32
1196 	rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1197 	if (rlim.rlim_cur == UINT_MAX)
1198 		rlim.rlim_cur = INT_MAX;
1199 	rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1200 	if (rlim.rlim_max == UINT_MAX)
1201 		rlim.rlim_max = INT_MAX;
1202 #else
1203 	rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1204 	if (rlim.rlim_cur == ULONG_MAX)
1205 		rlim.rlim_cur = LONG_MAX;
1206 	rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1207 	if (rlim.rlim_max == ULONG_MAX)
1208 		rlim.rlim_max = LONG_MAX;
1209 #endif
1210 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1211 }
1212 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1213 
1214 int
1215 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1216 {
1217 	struct l_rlimit rlim;
1218 	struct rlimit bsd_rlim;
1219 	u_int which;
1220 
1221 	if (linux_get_dummy_limit(td, args->resource, &bsd_rlim)) {
1222 		rlim.rlim_cur = bsd_rlim.rlim_cur;
1223 		rlim.rlim_max = bsd_rlim.rlim_max;
1224 		return (copyout(&rlim, args->rlim, sizeof(rlim)));
1225 	}
1226 
1227 	if (args->resource >= LINUX_RLIM_NLIMITS)
1228 		return (EINVAL);
1229 
1230 	which = linux_to_bsd_resource[args->resource];
1231 	if (which == -1)
1232 		return (EINVAL);
1233 
1234 	lim_rlimit(td, which, &bsd_rlim);
1235 
1236 	rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1237 	rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1238 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1239 }
1240 
1241 int
1242 linux_sched_setscheduler(struct thread *td,
1243     struct linux_sched_setscheduler_args *args)
1244 {
1245 	struct sched_param sched_param;
1246 	struct thread *tdt;
1247 	int error, policy;
1248 
1249 	switch (args->policy) {
1250 	case LINUX_SCHED_OTHER:
1251 		policy = SCHED_OTHER;
1252 		break;
1253 	case LINUX_SCHED_FIFO:
1254 		policy = SCHED_FIFO;
1255 		break;
1256 	case LINUX_SCHED_RR:
1257 		policy = SCHED_RR;
1258 		break;
1259 	default:
1260 		return (EINVAL);
1261 	}
1262 
1263 	error = copyin(args->param, &sched_param, sizeof(sched_param));
1264 	if (error)
1265 		return (error);
1266 
1267 	if (linux_map_sched_prio) {
1268 		switch (policy) {
1269 		case SCHED_OTHER:
1270 			if (sched_param.sched_priority != 0)
1271 				return (EINVAL);
1272 
1273 			sched_param.sched_priority =
1274 			    PRI_MAX_TIMESHARE - PRI_MIN_TIMESHARE;
1275 			break;
1276 		case SCHED_FIFO:
1277 		case SCHED_RR:
1278 			if (sched_param.sched_priority < 1 ||
1279 			    sched_param.sched_priority >= LINUX_MAX_RT_PRIO)
1280 				return (EINVAL);
1281 
1282 			/*
1283 			 * Map [1, LINUX_MAX_RT_PRIO - 1] to
1284 			 * [0, RTP_PRIO_MAX - RTP_PRIO_MIN] (rounding down).
1285 			 */
1286 			sched_param.sched_priority =
1287 			    (sched_param.sched_priority - 1) *
1288 			    (RTP_PRIO_MAX - RTP_PRIO_MIN + 1) /
1289 			    (LINUX_MAX_RT_PRIO - 1);
1290 			break;
1291 		}
1292 	}
1293 
1294 	tdt = linux_tdfind(td, args->pid, -1);
1295 	if (tdt == NULL)
1296 		return (ESRCH);
1297 
1298 	error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1299 	PROC_UNLOCK(tdt->td_proc);
1300 	return (error);
1301 }
1302 
1303 int
1304 linux_sched_getscheduler(struct thread *td,
1305     struct linux_sched_getscheduler_args *args)
1306 {
1307 	struct thread *tdt;
1308 	int error, policy;
1309 
1310 	tdt = linux_tdfind(td, args->pid, -1);
1311 	if (tdt == NULL)
1312 		return (ESRCH);
1313 
1314 	error = kern_sched_getscheduler(td, tdt, &policy);
1315 	PROC_UNLOCK(tdt->td_proc);
1316 
1317 	switch (policy) {
1318 	case SCHED_OTHER:
1319 		td->td_retval[0] = LINUX_SCHED_OTHER;
1320 		break;
1321 	case SCHED_FIFO:
1322 		td->td_retval[0] = LINUX_SCHED_FIFO;
1323 		break;
1324 	case SCHED_RR:
1325 		td->td_retval[0] = LINUX_SCHED_RR;
1326 		break;
1327 	}
1328 	return (error);
1329 }
1330 
1331 int
1332 linux_sched_get_priority_max(struct thread *td,
1333     struct linux_sched_get_priority_max_args *args)
1334 {
1335 	struct sched_get_priority_max_args bsd;
1336 
1337 	if (linux_map_sched_prio) {
1338 		switch (args->policy) {
1339 		case LINUX_SCHED_OTHER:
1340 			td->td_retval[0] = 0;
1341 			return (0);
1342 		case LINUX_SCHED_FIFO:
1343 		case LINUX_SCHED_RR:
1344 			td->td_retval[0] = LINUX_MAX_RT_PRIO - 1;
1345 			return (0);
1346 		default:
1347 			return (EINVAL);
1348 		}
1349 	}
1350 
1351 	switch (args->policy) {
1352 	case LINUX_SCHED_OTHER:
1353 		bsd.policy = SCHED_OTHER;
1354 		break;
1355 	case LINUX_SCHED_FIFO:
1356 		bsd.policy = SCHED_FIFO;
1357 		break;
1358 	case LINUX_SCHED_RR:
1359 		bsd.policy = SCHED_RR;
1360 		break;
1361 	default:
1362 		return (EINVAL);
1363 	}
1364 	return (sys_sched_get_priority_max(td, &bsd));
1365 }
1366 
1367 int
1368 linux_sched_get_priority_min(struct thread *td,
1369     struct linux_sched_get_priority_min_args *args)
1370 {
1371 	struct sched_get_priority_min_args bsd;
1372 
1373 	if (linux_map_sched_prio) {
1374 		switch (args->policy) {
1375 		case LINUX_SCHED_OTHER:
1376 			td->td_retval[0] = 0;
1377 			return (0);
1378 		case LINUX_SCHED_FIFO:
1379 		case LINUX_SCHED_RR:
1380 			td->td_retval[0] = 1;
1381 			return (0);
1382 		default:
1383 			return (EINVAL);
1384 		}
1385 	}
1386 
1387 	switch (args->policy) {
1388 	case LINUX_SCHED_OTHER:
1389 		bsd.policy = SCHED_OTHER;
1390 		break;
1391 	case LINUX_SCHED_FIFO:
1392 		bsd.policy = SCHED_FIFO;
1393 		break;
1394 	case LINUX_SCHED_RR:
1395 		bsd.policy = SCHED_RR;
1396 		break;
1397 	default:
1398 		return (EINVAL);
1399 	}
1400 	return (sys_sched_get_priority_min(td, &bsd));
1401 }
1402 
1403 #define REBOOT_CAD_ON	0x89abcdef
1404 #define REBOOT_CAD_OFF	0
1405 #define REBOOT_HALT	0xcdef0123
1406 #define REBOOT_RESTART	0x01234567
1407 #define REBOOT_RESTART2	0xA1B2C3D4
1408 #define REBOOT_POWEROFF	0x4321FEDC
1409 #define REBOOT_MAGIC1	0xfee1dead
1410 #define REBOOT_MAGIC2	0x28121969
1411 #define REBOOT_MAGIC2A	0x05121996
1412 #define REBOOT_MAGIC2B	0x16041998
1413 
1414 int
1415 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1416 {
1417 	struct reboot_args bsd_args;
1418 
1419 	if (args->magic1 != REBOOT_MAGIC1)
1420 		return (EINVAL);
1421 
1422 	switch (args->magic2) {
1423 	case REBOOT_MAGIC2:
1424 	case REBOOT_MAGIC2A:
1425 	case REBOOT_MAGIC2B:
1426 		break;
1427 	default:
1428 		return (EINVAL);
1429 	}
1430 
1431 	switch (args->cmd) {
1432 	case REBOOT_CAD_ON:
1433 	case REBOOT_CAD_OFF:
1434 		return (priv_check(td, PRIV_REBOOT));
1435 	case REBOOT_HALT:
1436 		bsd_args.opt = RB_HALT;
1437 		break;
1438 	case REBOOT_RESTART:
1439 	case REBOOT_RESTART2:
1440 		bsd_args.opt = 0;
1441 		break;
1442 	case REBOOT_POWEROFF:
1443 		bsd_args.opt = RB_POWEROFF;
1444 		break;
1445 	default:
1446 		return (EINVAL);
1447 	}
1448 	return (sys_reboot(td, &bsd_args));
1449 }
1450 
1451 int
1452 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1453 {
1454 
1455 	td->td_retval[0] = td->td_proc->p_pid;
1456 
1457 	return (0);
1458 }
1459 
1460 int
1461 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1462 {
1463 	struct linux_emuldata *em;
1464 
1465 	em = em_find(td);
1466 	KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1467 
1468 	td->td_retval[0] = em->em_tid;
1469 
1470 	return (0);
1471 }
1472 
1473 int
1474 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1475 {
1476 
1477 	td->td_retval[0] = kern_getppid(td);
1478 	return (0);
1479 }
1480 
1481 int
1482 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1483 {
1484 
1485 	td->td_retval[0] = td->td_ucred->cr_rgid;
1486 	return (0);
1487 }
1488 
1489 int
1490 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1491 {
1492 
1493 	td->td_retval[0] = td->td_ucred->cr_ruid;
1494 	return (0);
1495 }
1496 
1497 int
1498 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1499 {
1500 
1501 	return (kern_getsid(td, args->pid));
1502 }
1503 
1504 int
1505 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1506 {
1507 	int error;
1508 
1509 	error = kern_getpriority(td, args->which, args->who);
1510 	td->td_retval[0] = 20 - td->td_retval[0];
1511 	return (error);
1512 }
1513 
1514 int
1515 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1516 {
1517 	int name[2];
1518 
1519 	name[0] = CTL_KERN;
1520 	name[1] = KERN_HOSTNAME;
1521 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1522 	    args->len, 0, 0));
1523 }
1524 
1525 int
1526 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1527 {
1528 	int name[2];
1529 
1530 	name[0] = CTL_KERN;
1531 	name[1] = KERN_NISDOMAINNAME;
1532 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1533 	    args->len, 0, 0));
1534 }
1535 
1536 int
1537 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1538 {
1539 
1540 	LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1541 	    args->error_code);
1542 
1543 	/*
1544 	 * XXX: we should send a signal to the parent if
1545 	 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1546 	 * as it doesnt occur often.
1547 	 */
1548 	kern_exit(td, args->error_code, 0);
1549 	return (0);
1550 }
1551 
1552 #define _LINUX_CAPABILITY_VERSION_1  0x19980330
1553 #define _LINUX_CAPABILITY_VERSION_2  0x20071026
1554 #define _LINUX_CAPABILITY_VERSION_3  0x20080522
1555 
1556 struct l_user_cap_header {
1557 	l_int	version;
1558 	l_int	pid;
1559 };
1560 
1561 struct l_user_cap_data {
1562 	l_int	effective;
1563 	l_int	permitted;
1564 	l_int	inheritable;
1565 };
1566 
1567 int
1568 linux_capget(struct thread *td, struct linux_capget_args *uap)
1569 {
1570 	struct l_user_cap_header luch;
1571 	struct l_user_cap_data lucd[2];
1572 	int error, u32s;
1573 
1574 	if (uap->hdrp == NULL)
1575 		return (EFAULT);
1576 
1577 	error = copyin(uap->hdrp, &luch, sizeof(luch));
1578 	if (error != 0)
1579 		return (error);
1580 
1581 	switch (luch.version) {
1582 	case _LINUX_CAPABILITY_VERSION_1:
1583 		u32s = 1;
1584 		break;
1585 	case _LINUX_CAPABILITY_VERSION_2:
1586 	case _LINUX_CAPABILITY_VERSION_3:
1587 		u32s = 2;
1588 		break;
1589 	default:
1590 		luch.version = _LINUX_CAPABILITY_VERSION_1;
1591 		error = copyout(&luch, uap->hdrp, sizeof(luch));
1592 		if (error)
1593 			return (error);
1594 		return (EINVAL);
1595 	}
1596 
1597 	if (luch.pid)
1598 		return (EPERM);
1599 
1600 	if (uap->datap) {
1601 		/*
1602 		 * The current implementation doesn't support setting
1603 		 * a capability (it's essentially a stub) so indicate
1604 		 * that no capabilities are currently set or available
1605 		 * to request.
1606 		 */
1607 		memset(&lucd, 0, u32s * sizeof(lucd[0]));
1608 		error = copyout(&lucd, uap->datap, u32s * sizeof(lucd[0]));
1609 	}
1610 
1611 	return (error);
1612 }
1613 
1614 int
1615 linux_capset(struct thread *td, struct linux_capset_args *uap)
1616 {
1617 	struct l_user_cap_header luch;
1618 	struct l_user_cap_data lucd[2];
1619 	int error, i, u32s;
1620 
1621 	if (uap->hdrp == NULL || uap->datap == NULL)
1622 		return (EFAULT);
1623 
1624 	error = copyin(uap->hdrp, &luch, sizeof(luch));
1625 	if (error != 0)
1626 		return (error);
1627 
1628 	switch (luch.version) {
1629 	case _LINUX_CAPABILITY_VERSION_1:
1630 		u32s = 1;
1631 		break;
1632 	case _LINUX_CAPABILITY_VERSION_2:
1633 	case _LINUX_CAPABILITY_VERSION_3:
1634 		u32s = 2;
1635 		break;
1636 	default:
1637 		luch.version = _LINUX_CAPABILITY_VERSION_1;
1638 		error = copyout(&luch, uap->hdrp, sizeof(luch));
1639 		if (error)
1640 			return (error);
1641 		return (EINVAL);
1642 	}
1643 
1644 	if (luch.pid)
1645 		return (EPERM);
1646 
1647 	error = copyin(uap->datap, &lucd, u32s * sizeof(lucd[0]));
1648 	if (error != 0)
1649 		return (error);
1650 
1651 	/* We currently don't support setting any capabilities. */
1652 	for (i = 0; i < u32s; i++) {
1653 		if (lucd[i].effective || lucd[i].permitted ||
1654 		    lucd[i].inheritable) {
1655 			linux_msg(td,
1656 			    "capset[%d] effective=0x%x, permitted=0x%x, "
1657 			    "inheritable=0x%x is not implemented", i,
1658 			    (int)lucd[i].effective, (int)lucd[i].permitted,
1659 			    (int)lucd[i].inheritable);
1660 			return (EPERM);
1661 		}
1662 	}
1663 
1664 	return (0);
1665 }
1666 
1667 int
1668 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1669 {
1670 	int error = 0, max_size, arg;
1671 	struct proc *p = td->td_proc;
1672 	char comm[LINUX_MAX_COMM_LEN];
1673 	int pdeath_signal, trace_state;
1674 
1675 	switch (args->option) {
1676 	case LINUX_PR_SET_PDEATHSIG:
1677 		if (!LINUX_SIG_VALID(args->arg2))
1678 			return (EINVAL);
1679 		pdeath_signal = linux_to_bsd_signal(args->arg2);
1680 		return (kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_CTL,
1681 		    &pdeath_signal));
1682 	case LINUX_PR_GET_PDEATHSIG:
1683 		error = kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_STATUS,
1684 		    &pdeath_signal);
1685 		if (error != 0)
1686 			return (error);
1687 		pdeath_signal = bsd_to_linux_signal(pdeath_signal);
1688 		return (copyout(&pdeath_signal,
1689 		    (void *)(register_t)args->arg2,
1690 		    sizeof(pdeath_signal)));
1691 	/*
1692 	 * In Linux, this flag controls if set[gu]id processes can coredump.
1693 	 * There are additional semantics imposed on processes that cannot
1694 	 * coredump:
1695 	 * - Such processes can not be ptraced.
1696 	 * - There are some semantics around ownership of process-related files
1697 	 *   in the /proc namespace.
1698 	 *
1699 	 * In FreeBSD, we can (and by default, do) disable setuid coredump
1700 	 * system-wide with 'sugid_coredump.'  We control tracability on a
1701 	 * per-process basis with the procctl PROC_TRACE (=> P2_NOTRACE flag).
1702 	 * By happy coincidence, P2_NOTRACE also prevents coredumping.  So the
1703 	 * procctl is roughly analogous to Linux's DUMPABLE.
1704 	 *
1705 	 * So, proxy these knobs to the corresponding PROC_TRACE setting.
1706 	 */
1707 	case LINUX_PR_GET_DUMPABLE:
1708 		error = kern_procctl(td, P_PID, p->p_pid, PROC_TRACE_STATUS,
1709 		    &trace_state);
1710 		if (error != 0)
1711 			return (error);
1712 		td->td_retval[0] = (trace_state != -1);
1713 		return (0);
1714 	case LINUX_PR_SET_DUMPABLE:
1715 		/*
1716 		 * It is only valid for userspace to set one of these two
1717 		 * flags, and only one at a time.
1718 		 */
1719 		switch (args->arg2) {
1720 		case LINUX_SUID_DUMP_DISABLE:
1721 			trace_state = PROC_TRACE_CTL_DISABLE_EXEC;
1722 			break;
1723 		case LINUX_SUID_DUMP_USER:
1724 			trace_state = PROC_TRACE_CTL_ENABLE;
1725 			break;
1726 		default:
1727 			return (EINVAL);
1728 		}
1729 		return (kern_procctl(td, P_PID, p->p_pid, PROC_TRACE_CTL,
1730 		    &trace_state));
1731 	case LINUX_PR_GET_KEEPCAPS:
1732 		/*
1733 		 * Indicate that we always clear the effective and
1734 		 * permitted capability sets when the user id becomes
1735 		 * non-zero (actually the capability sets are simply
1736 		 * always zero in the current implementation).
1737 		 */
1738 		td->td_retval[0] = 0;
1739 		break;
1740 	case LINUX_PR_SET_KEEPCAPS:
1741 		/*
1742 		 * Ignore requests to keep the effective and permitted
1743 		 * capability sets when the user id becomes non-zero.
1744 		 */
1745 		break;
1746 	case LINUX_PR_SET_NAME:
1747 		/*
1748 		 * To be on the safe side we need to make sure to not
1749 		 * overflow the size a Linux program expects. We already
1750 		 * do this here in the copyin, so that we don't need to
1751 		 * check on copyout.
1752 		 */
1753 		max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1754 		error = copyinstr((void *)(register_t)args->arg2, comm,
1755 		    max_size, NULL);
1756 
1757 		/* Linux silently truncates the name if it is too long. */
1758 		if (error == ENAMETOOLONG) {
1759 			/*
1760 			 * XXX: copyinstr() isn't documented to populate the
1761 			 * array completely, so do a copyin() to be on the
1762 			 * safe side. This should be changed in case
1763 			 * copyinstr() is changed to guarantee this.
1764 			 */
1765 			error = copyin((void *)(register_t)args->arg2, comm,
1766 			    max_size - 1);
1767 			comm[max_size - 1] = '\0';
1768 		}
1769 		if (error)
1770 			return (error);
1771 
1772 		PROC_LOCK(p);
1773 		strlcpy(p->p_comm, comm, sizeof(p->p_comm));
1774 		PROC_UNLOCK(p);
1775 		break;
1776 	case LINUX_PR_GET_NAME:
1777 		PROC_LOCK(p);
1778 		strlcpy(comm, p->p_comm, sizeof(comm));
1779 		PROC_UNLOCK(p);
1780 		error = copyout(comm, (void *)(register_t)args->arg2,
1781 		    strlen(comm) + 1);
1782 		break;
1783 	case LINUX_PR_GET_SECCOMP:
1784 	case LINUX_PR_SET_SECCOMP:
1785 		/*
1786 		 * Same as returned by Linux without CONFIG_SECCOMP enabled.
1787 		 */
1788 		error = EINVAL;
1789 		break;
1790 	case LINUX_PR_CAPBSET_READ:
1791 #if 0
1792 		/*
1793 		 * This makes too much noise with Ubuntu Focal.
1794 		 */
1795 		linux_msg(td, "unsupported prctl PR_CAPBSET_READ %d",
1796 		    (int)args->arg2);
1797 #endif
1798 		error = EINVAL;
1799 		break;
1800 	case LINUX_PR_SET_CHILD_SUBREAPER:
1801 		if (args->arg2 == 0) {
1802 			return (kern_procctl(td, P_PID, 0, PROC_REAP_RELEASE,
1803 			    NULL));
1804 		}
1805 
1806 		return (kern_procctl(td, P_PID, 0, PROC_REAP_ACQUIRE,
1807 		    NULL));
1808 	case LINUX_PR_GET_CHILD_SUBREAPER: {
1809 		struct procctl_reaper_status rs;
1810 		l_int val;
1811 
1812 		error = kern_procctl(td, P_PID, 0, PROC_REAP_STATUS, &rs);
1813 		if (error != 0)
1814 			return (error);
1815 		val = rs.rs_reaper == p->p_pid ? 1 : 0;
1816 		error = copyout(&val, (void *)(register_t)args->arg2,
1817 		    sizeof(val));
1818 		break;
1819 	}
1820 	case LINUX_PR_SET_NO_NEW_PRIVS:
1821 		arg = args->arg2 == 1 ?
1822 		    PROC_NO_NEW_PRIVS_ENABLE : PROC_NO_NEW_PRIVS_DISABLE;
1823 		error = kern_procctl(td, P_PID, p->p_pid,
1824 		    PROC_NO_NEW_PRIVS_CTL, &arg);
1825 		break;
1826 	case LINUX_PR_GET_NO_NEW_PRIVS:
1827 		error = kern_procctl(td, P_PID, p->p_pid,
1828 		    PROC_NO_NEW_PRIVS_STATUS, &arg);
1829 		if (error != 0)
1830 			return (error);
1831 		/* Linux returns the value as the syscall return */
1832 		td->td_retval[0] = arg == PROC_NO_NEW_PRIVS_ENABLE ? 1 : 0;
1833 		break;
1834 	case LINUX_PR_SET_PTRACER:
1835 		linux_msg(td, "unsupported prctl PR_SET_PTRACER");
1836 		error = EINVAL;
1837 		break;
1838 	case LINUX_PR_SET_VMA:
1839 		if (args->arg2 != LINUX_PR_SET_VMA_ANON_NAME) {
1840 			linux_msg(td, "unsupported prctl PR_SET_VMA attr %ju",
1841 			    (uintmax_t)args->arg2);
1842 			error = EINVAL;
1843 		}
1844 		break;
1845 	default:
1846 		linux_msg(td, "unsupported prctl option %d", args->option);
1847 		error = EINVAL;
1848 		break;
1849 	}
1850 
1851 	return (error);
1852 }
1853 
1854 int
1855 linux_sched_setparam(struct thread *td,
1856     struct linux_sched_setparam_args *uap)
1857 {
1858 	struct sched_param sched_param;
1859 	struct thread *tdt;
1860 	int error, policy;
1861 
1862 	error = copyin(uap->param, &sched_param, sizeof(sched_param));
1863 	if (error)
1864 		return (error);
1865 
1866 	tdt = linux_tdfind(td, uap->pid, -1);
1867 	if (tdt == NULL)
1868 		return (ESRCH);
1869 
1870 	if (linux_map_sched_prio) {
1871 		error = kern_sched_getscheduler(td, tdt, &policy);
1872 		if (error)
1873 			goto out;
1874 
1875 		switch (policy) {
1876 		case SCHED_OTHER:
1877 			if (sched_param.sched_priority != 0) {
1878 				error = EINVAL;
1879 				goto out;
1880 			}
1881 			sched_param.sched_priority =
1882 			    PRI_MAX_TIMESHARE - PRI_MIN_TIMESHARE;
1883 			break;
1884 		case SCHED_FIFO:
1885 		case SCHED_RR:
1886 			if (sched_param.sched_priority < 1 ||
1887 			    sched_param.sched_priority >= LINUX_MAX_RT_PRIO) {
1888 				error = EINVAL;
1889 				goto out;
1890 			}
1891 			/*
1892 			 * Map [1, LINUX_MAX_RT_PRIO - 1] to
1893 			 * [0, RTP_PRIO_MAX - RTP_PRIO_MIN] (rounding down).
1894 			 */
1895 			sched_param.sched_priority =
1896 			    (sched_param.sched_priority - 1) *
1897 			    (RTP_PRIO_MAX - RTP_PRIO_MIN + 1) /
1898 			    (LINUX_MAX_RT_PRIO - 1);
1899 			break;
1900 		}
1901 	}
1902 
1903 	error = kern_sched_setparam(td, tdt, &sched_param);
1904 out:	PROC_UNLOCK(tdt->td_proc);
1905 	return (error);
1906 }
1907 
1908 int
1909 linux_sched_getparam(struct thread *td,
1910     struct linux_sched_getparam_args *uap)
1911 {
1912 	struct sched_param sched_param;
1913 	struct thread *tdt;
1914 	int error, policy;
1915 
1916 	tdt = linux_tdfind(td, uap->pid, -1);
1917 	if (tdt == NULL)
1918 		return (ESRCH);
1919 
1920 	error = kern_sched_getparam(td, tdt, &sched_param);
1921 	if (error) {
1922 		PROC_UNLOCK(tdt->td_proc);
1923 		return (error);
1924 	}
1925 
1926 	if (linux_map_sched_prio) {
1927 		error = kern_sched_getscheduler(td, tdt, &policy);
1928 		PROC_UNLOCK(tdt->td_proc);
1929 		if (error)
1930 			return (error);
1931 
1932 		switch (policy) {
1933 		case SCHED_OTHER:
1934 			sched_param.sched_priority = 0;
1935 			break;
1936 		case SCHED_FIFO:
1937 		case SCHED_RR:
1938 			/*
1939 			 * Map [0, RTP_PRIO_MAX - RTP_PRIO_MIN] to
1940 			 * [1, LINUX_MAX_RT_PRIO - 1] (rounding up).
1941 			 */
1942 			sched_param.sched_priority =
1943 			    (sched_param.sched_priority *
1944 			    (LINUX_MAX_RT_PRIO - 1) +
1945 			    (RTP_PRIO_MAX - RTP_PRIO_MIN - 1)) /
1946 			    (RTP_PRIO_MAX - RTP_PRIO_MIN) + 1;
1947 			break;
1948 		}
1949 	} else
1950 		PROC_UNLOCK(tdt->td_proc);
1951 
1952 	error = copyout(&sched_param, uap->param, sizeof(sched_param));
1953 	return (error);
1954 }
1955 
1956 /*
1957  * Get affinity of a process.
1958  */
1959 int
1960 linux_sched_getaffinity(struct thread *td,
1961     struct linux_sched_getaffinity_args *args)
1962 {
1963 	struct thread *tdt;
1964 	cpuset_t *mask;
1965 	size_t size;
1966 	int error;
1967 	id_t tid;
1968 
1969 	tdt = linux_tdfind(td, args->pid, -1);
1970 	if (tdt == NULL)
1971 		return (ESRCH);
1972 	tid = tdt->td_tid;
1973 	PROC_UNLOCK(tdt->td_proc);
1974 
1975 	mask = malloc(sizeof(cpuset_t), M_LINUX, M_WAITOK | M_ZERO);
1976 	size = min(args->len, sizeof(cpuset_t));
1977 	error = kern_cpuset_getaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
1978 	    tid, size, mask);
1979 	if (error == ERANGE)
1980 		error = EINVAL;
1981  	if (error == 0)
1982 		error = copyout(mask, args->user_mask_ptr, size);
1983 	if (error == 0)
1984 		td->td_retval[0] = size;
1985 	free(mask, M_LINUX);
1986 	return (error);
1987 }
1988 
1989 /*
1990  *  Set affinity of a process.
1991  */
1992 int
1993 linux_sched_setaffinity(struct thread *td,
1994     struct linux_sched_setaffinity_args *args)
1995 {
1996 	struct thread *tdt;
1997 	cpuset_t *mask;
1998 	int cpu, error;
1999 	size_t len;
2000 	id_t tid;
2001 
2002 	tdt = linux_tdfind(td, args->pid, -1);
2003 	if (tdt == NULL)
2004 		return (ESRCH);
2005 	tid = tdt->td_tid;
2006 	PROC_UNLOCK(tdt->td_proc);
2007 
2008 	len = min(args->len, sizeof(cpuset_t));
2009 	mask = malloc(sizeof(cpuset_t), M_TEMP, M_WAITOK | M_ZERO);
2010 	error = copyin(args->user_mask_ptr, mask, len);
2011 	if (error != 0)
2012 		goto out;
2013 	/* Linux ignore high bits */
2014 	CPU_FOREACH_ISSET(cpu, mask)
2015 		if (cpu > mp_maxid)
2016 			CPU_CLR(cpu, mask);
2017 
2018 	error = kern_cpuset_setaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2019 	    tid, mask);
2020 	if (error == EDEADLK)
2021 		error = EINVAL;
2022 out:
2023 	free(mask, M_TEMP);
2024 	return (error);
2025 }
2026 
2027 struct linux_rlimit64 {
2028 	uint64_t	rlim_cur;
2029 	uint64_t	rlim_max;
2030 };
2031 
2032 int
2033 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
2034 {
2035 	struct rlimit rlim, nrlim;
2036 	struct linux_rlimit64 lrlim;
2037 	struct proc *p;
2038 	u_int which;
2039 	int flags;
2040 	int error;
2041 	bool exec_blocked;
2042 
2043 	if (args->new == NULL && args->old != NULL) {
2044 		if (linux_get_dummy_limit(td, args->resource, &rlim)) {
2045 			lrlim.rlim_cur = rlim.rlim_cur;
2046 			lrlim.rlim_max = rlim.rlim_max;
2047 			return (copyout(&lrlim, args->old, sizeof(lrlim)));
2048 		}
2049 	}
2050 
2051 	if (args->resource >= LINUX_RLIM_NLIMITS)
2052 		return (EINVAL);
2053 
2054 	which = linux_to_bsd_resource[args->resource];
2055 	if (which == -1)
2056 		return (EINVAL);
2057 
2058 	if (args->new != NULL) {
2059 		/*
2060 		 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
2061 		 * rlim is unsigned 64-bit. FreeBSD treats negative limits
2062 		 * as INFINITY so we do not need a conversion even.
2063 		 */
2064 		error = copyin(args->new, &nrlim, sizeof(nrlim));
2065 		if (error != 0)
2066 			return (error);
2067 	}
2068 
2069 	exec_blocked = false;
2070 	flags = PGET_HOLD | PGET_NOTWEXIT;
2071 	if (args->new != NULL)
2072 		flags |= PGET_CANDEBUG;
2073 	else
2074 		flags |= PGET_CANSEE;
2075 	if (args->pid == 0) {
2076 		p = td->td_proc;
2077 		PHOLD(p);
2078 	} else {
2079 		error = pget(args->pid, flags, &p);
2080 		if (error != 0)
2081 			return (error);
2082 		exec_blocked = true;
2083 		PROC_LOCK(p);
2084 		execve_block_wait(td, p);
2085 		error = args->new != NULL ? p_candebug(td, p) :
2086 		    p_cansee(td, p);
2087 		PROC_UNLOCK(p);
2088 		if (error != 0)
2089 			goto out;
2090 	}
2091 	if (args->old != NULL) {
2092 		PROC_LOCK(p);
2093 		lim_rlimit_proc(p, which, &rlim);
2094 		PROC_UNLOCK(p);
2095 		if (rlim.rlim_cur == RLIM_INFINITY)
2096 			lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2097 		else
2098 			lrlim.rlim_cur = rlim.rlim_cur;
2099 		if (rlim.rlim_max == RLIM_INFINITY)
2100 			lrlim.rlim_max = LINUX_RLIM_INFINITY;
2101 		else
2102 			lrlim.rlim_max = rlim.rlim_max;
2103 		error = copyout(&lrlim, args->old, sizeof(lrlim));
2104 		if (error != 0)
2105 			goto out;
2106 	}
2107 
2108 	if (args->new != NULL)
2109 		error = kern_proc_setrlimit(td, p, which, &nrlim);
2110 
2111  out:
2112 	if (exec_blocked) {
2113 		PROC_LOCK(p);
2114 		execve_unblock(td, p);
2115 		PROC_UNLOCK(p);
2116 	}
2117 	PRELE(p);
2118 	return (error);
2119 }
2120 
2121 int
2122 linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2123 {
2124 	struct timespec ts, *tsp;
2125 	int error;
2126 
2127 	if (args->tsp != NULL) {
2128 		error = linux_get_timespec(&ts, args->tsp);
2129 		if (error != 0)
2130 			return (error);
2131 		tsp = &ts;
2132 	} else
2133 		tsp = NULL;
2134 
2135 	error = linux_common_pselect6(td, args->nfds, args->readfds,
2136 	    args->writefds, args->exceptfds, tsp, args->sig);
2137 
2138 	if (args->tsp != NULL)
2139 		linux_put_timespec(&ts, args->tsp);
2140 	return (error);
2141 }
2142 
2143 static int
2144 linux_common_pselect6(struct thread *td, l_int nfds, l_fd_set *readfds,
2145     l_fd_set *writefds, l_fd_set *exceptfds, struct timespec *tsp,
2146     l_uintptr_t *sig)
2147 {
2148 	struct timeval utv, tv0, tv1, *tvp;
2149 	struct l_pselect6arg lpse6;
2150 	sigset_t *ssp;
2151 	sigset_t ss;
2152 	int error;
2153 
2154 	ssp = NULL;
2155 	if (sig != NULL) {
2156 		error = copyin(sig, &lpse6, sizeof(lpse6));
2157 		if (error != 0)
2158 			return (error);
2159 		error = linux_copyin_sigset(td, PTRIN(lpse6.ss),
2160 		    lpse6.ss_len, &ss, &ssp);
2161 		if (error != 0)
2162 		    return (error);
2163 	} else
2164 		ssp = NULL;
2165 
2166 	/*
2167 	 * Currently glibc changes nanosecond number to microsecond.
2168 	 * This mean losing precision but for now it is hardly seen.
2169 	 */
2170 	if (tsp != NULL) {
2171 		TIMESPEC_TO_TIMEVAL(&utv, tsp);
2172 		if (itimerfix(&utv))
2173 			return (EINVAL);
2174 
2175 		microtime(&tv0);
2176 		tvp = &utv;
2177 	} else
2178 		tvp = NULL;
2179 
2180 	error = kern_pselect(td, nfds, readfds, writefds,
2181 	    exceptfds, tvp, ssp, LINUX_NFDBITS);
2182 
2183 	if (tsp != NULL) {
2184 		/*
2185 		 * Compute how much time was left of the timeout,
2186 		 * by subtracting the current time and the time
2187 		 * before we started the call, and subtracting
2188 		 * that result from the user-supplied value.
2189 		 */
2190 		microtime(&tv1);
2191 		timevalsub(&tv1, &tv0);
2192 		timevalsub(&utv, &tv1);
2193 		if (utv.tv_sec < 0)
2194 			timevalclear(&utv);
2195 		TIMEVAL_TO_TIMESPEC(&utv, tsp);
2196 	}
2197 	return (error);
2198 }
2199 
2200 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
2201 int
2202 linux_pselect6_time64(struct thread *td,
2203     struct linux_pselect6_time64_args *args)
2204 {
2205 	struct timespec ts, *tsp;
2206 	int error;
2207 
2208 	if (args->tsp != NULL) {
2209 		error = linux_get_timespec64(&ts, args->tsp);
2210 		if (error != 0)
2211 			return (error);
2212 		tsp = &ts;
2213 	} else
2214 		tsp = NULL;
2215 
2216 	error = linux_common_pselect6(td, args->nfds, args->readfds,
2217 	    args->writefds, args->exceptfds, tsp, args->sig);
2218 
2219 	if (args->tsp != NULL)
2220 		linux_put_timespec64(&ts, args->tsp);
2221 	return (error);
2222 }
2223 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
2224 
2225 int
2226 linux_ppoll(struct thread *td, struct linux_ppoll_args *args)
2227 {
2228 	struct timespec uts, *tsp;
2229 	int error;
2230 
2231 	if (args->tsp != NULL) {
2232 		error = linux_get_timespec(&uts, args->tsp);
2233 		if (error != 0)
2234 			return (error);
2235 		tsp = &uts;
2236 	} else
2237 		tsp = NULL;
2238 
2239 	error = linux_common_ppoll(td, args->fds, args->nfds, tsp,
2240 	    args->sset, args->ssize);
2241 	if (error == 0 && args->tsp != NULL)
2242 		error = linux_put_timespec(&uts, args->tsp);
2243 	return (error);
2244 }
2245 
2246 static int
2247 linux_common_ppoll(struct thread *td, struct pollfd *fds, uint32_t nfds,
2248     struct timespec *tsp, l_sigset_t *sset, l_size_t ssize)
2249 {
2250 	struct timespec ts0, ts1;
2251 	struct pollfd stackfds[32];
2252 	struct pollfd *kfds;
2253  	sigset_t *ssp;
2254  	sigset_t ss;
2255  	int error;
2256 
2257 	if (kern_poll_maxfds(nfds))
2258 		return (EINVAL);
2259 	if (sset != NULL) {
2260 		error = linux_copyin_sigset(td, sset, ssize, &ss, &ssp);
2261 		if (error != 0)
2262 		    return (error);
2263 	} else
2264 		ssp = NULL;
2265 	if (tsp != NULL)
2266 		nanotime(&ts0);
2267 
2268 	if (nfds > nitems(stackfds))
2269 		kfds = mallocarray(nfds, sizeof(*kfds), M_TEMP, M_WAITOK);
2270 	else
2271 		kfds = stackfds;
2272 	error = linux_pollin(td, kfds, fds, nfds);
2273 	if (error != 0)
2274 		goto out;
2275 
2276 	error = kern_poll_kfds(td, kfds, nfds, tsp, ssp);
2277 	if (error == 0)
2278 		error = linux_pollout(td, kfds, fds, nfds);
2279 
2280 	if (error == 0 && tsp != NULL) {
2281 		if (td->td_retval[0]) {
2282 			nanotime(&ts1);
2283 			timespecsub(&ts1, &ts0, &ts1);
2284 			timespecsub(tsp, &ts1, tsp);
2285 			if (tsp->tv_sec < 0)
2286 				timespecclear(tsp);
2287 		} else
2288 			timespecclear(tsp);
2289 	}
2290 
2291 out:
2292 	if (nfds > nitems(stackfds))
2293 		free(kfds, M_TEMP);
2294 	return (error);
2295 }
2296 
2297 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
2298 int
2299 linux_ppoll_time64(struct thread *td, struct linux_ppoll_time64_args *args)
2300 {
2301 	struct timespec uts, *tsp;
2302 	int error;
2303 
2304 	if (args->tsp != NULL) {
2305 		error = linux_get_timespec64(&uts, args->tsp);
2306 		if (error != 0)
2307 			return (error);
2308 		tsp = &uts;
2309 	} else
2310  		tsp = NULL;
2311 	error = linux_common_ppoll(td, args->fds, args->nfds, tsp,
2312 	    args->sset, args->ssize);
2313 	if (error == 0 && args->tsp != NULL)
2314 		error = linux_put_timespec64(&uts, args->tsp);
2315 	return (error);
2316 }
2317 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
2318 
2319 static int
2320 linux_pollin(struct thread *td, struct pollfd *fds, struct pollfd *ufds, u_int nfd)
2321 {
2322 	int error;
2323 	u_int i;
2324 
2325 	error = copyin(ufds, fds, nfd * sizeof(*fds));
2326 	if (error != 0)
2327 		return (error);
2328 
2329 	for (i = 0; i < nfd; i++) {
2330 		if (fds->events != 0)
2331 			linux_to_bsd_poll_events(td, fds->fd,
2332 			    fds->events, &fds->events);
2333 		fds++;
2334 	}
2335 	return (0);
2336 }
2337 
2338 static int
2339 linux_pollout(struct thread *td, struct pollfd *fds, struct pollfd *ufds, u_int nfd)
2340 {
2341 	int error = 0;
2342 	u_int i, n = 0;
2343 
2344 	for (i = 0; i < nfd; i++) {
2345 		if (fds->revents != 0) {
2346 			bsd_to_linux_poll_events(fds->revents,
2347 			    &fds->revents);
2348 			n++;
2349 		}
2350 		error = copyout(&fds->revents, &ufds->revents,
2351 		    sizeof(ufds->revents));
2352 		if (error)
2353 			return (error);
2354 		fds++;
2355 		ufds++;
2356 	}
2357 	td->td_retval[0] = n;
2358 	return (0);
2359 }
2360 
2361 static int
2362 linux_sched_rr_get_interval_common(struct thread *td, pid_t pid,
2363     struct timespec *ts)
2364 {
2365 	struct thread *tdt;
2366 	int error;
2367 
2368 	/*
2369 	 * According to man in case the invalid pid specified
2370 	 * EINVAL should be returned.
2371 	 */
2372 	if (pid < 0)
2373 		return (EINVAL);
2374 
2375 	tdt = linux_tdfind(td, pid, -1);
2376 	if (tdt == NULL)
2377 		return (ESRCH);
2378 
2379 	error = kern_sched_rr_get_interval_td(td, tdt, ts);
2380 	PROC_UNLOCK(tdt->td_proc);
2381 	return (error);
2382 }
2383 
2384 int
2385 linux_sched_rr_get_interval(struct thread *td,
2386     struct linux_sched_rr_get_interval_args *uap)
2387 {
2388 	struct timespec ts;
2389 	int error;
2390 
2391 	error = linux_sched_rr_get_interval_common(td, uap->pid, &ts);
2392 	if (error != 0)
2393 		return (error);
2394 	return (linux_put_timespec(&ts, uap->interval));
2395 }
2396 
2397 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
2398 int
2399 linux_sched_rr_get_interval_time64(struct thread *td,
2400     struct linux_sched_rr_get_interval_time64_args *uap)
2401 {
2402 	struct timespec ts;
2403 	int error;
2404 
2405 	error = linux_sched_rr_get_interval_common(td, uap->pid, &ts);
2406 	if (error != 0)
2407 		return (error);
2408 	return (linux_put_timespec64(&ts, uap->interval));
2409 }
2410 #endif
2411 
2412 /*
2413  * In case when the Linux thread is the initial thread in
2414  * the thread group thread id is equal to the process id.
2415  * Glibc depends on this magic (assert in pthread_getattr_np.c).
2416  */
2417 struct thread *
2418 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid)
2419 {
2420 	struct linux_emuldata *em;
2421 	struct thread *tdt;
2422 	struct proc *p;
2423 
2424 	tdt = NULL;
2425 	if (tid == 0 || tid == td->td_tid) {
2426 		if (pid != -1 && td->td_proc->p_pid != pid)
2427 			return (NULL);
2428 		PROC_LOCK(td->td_proc);
2429 		return (td);
2430 	} else if (tid > PID_MAX)
2431 		return (tdfind(tid, pid));
2432 
2433 	/*
2434 	 * Initial thread where the tid equal to the pid.
2435 	 */
2436 	p = pfind(tid);
2437 	if (p != NULL) {
2438 		if (SV_PROC_ABI(p) != SV_ABI_LINUX ||
2439 		    (pid != -1 && tid != pid)) {
2440 			/*
2441 			 * p is not a Linuxulator process.
2442 			 */
2443 			PROC_UNLOCK(p);
2444 			return (NULL);
2445 		}
2446 		FOREACH_THREAD_IN_PROC(p, tdt) {
2447 			em = em_find(tdt);
2448 			if (tid == em->em_tid)
2449 				return (tdt);
2450 		}
2451 		PROC_UNLOCK(p);
2452 	}
2453 	return (NULL);
2454 }
2455 
2456 void
2457 linux_to_bsd_waitopts(int options, int *bsdopts)
2458 {
2459 
2460 	if (options & LINUX_WNOHANG)
2461 		*bsdopts |= WNOHANG;
2462 	if (options & LINUX_WUNTRACED)
2463 		*bsdopts |= WUNTRACED;
2464 	if (options & LINUX_WEXITED)
2465 		*bsdopts |= WEXITED;
2466 	if (options & LINUX_WCONTINUED)
2467 		*bsdopts |= WCONTINUED;
2468 	if (options & LINUX_WNOWAIT)
2469 		*bsdopts |= WNOWAIT;
2470 
2471 	if (options & __WCLONE)
2472 		*bsdopts |= WLINUXCLONE;
2473 }
2474 
2475 int
2476 linux_getrandom(struct thread *td, struct linux_getrandom_args *args)
2477 {
2478 	struct uio uio;
2479 	struct iovec iov;
2480 	int error;
2481 
2482 	if (args->flags & ~(LINUX_GRND_NONBLOCK|LINUX_GRND_RANDOM))
2483 		return (EINVAL);
2484 	if (args->count > INT_MAX)
2485 		args->count = INT_MAX;
2486 
2487 	iov.iov_base = args->buf;
2488 	iov.iov_len = args->count;
2489 
2490 	uio.uio_iov = &iov;
2491 	uio.uio_iovcnt = 1;
2492 	uio.uio_resid = iov.iov_len;
2493 	uio.uio_segflg = UIO_USERSPACE;
2494 	uio.uio_rw = UIO_READ;
2495 	uio.uio_td = td;
2496 
2497 	error = read_random_uio(&uio, args->flags & LINUX_GRND_NONBLOCK);
2498 	if (error == 0)
2499 		td->td_retval[0] = args->count - uio.uio_resid;
2500 	return (error);
2501 }
2502 
2503 int
2504 linux_mincore(struct thread *td, struct linux_mincore_args *args)
2505 {
2506 
2507 	/* Needs to be page-aligned */
2508 	if (args->start & PAGE_MASK)
2509 		return (EINVAL);
2510 	return (kern_mincore(td, args->start, args->len, args->vec));
2511 }
2512 
2513 #define	SYSLOG_TAG	"<6>"
2514 
2515 int
2516 linux_syslog(struct thread *td, struct linux_syslog_args *args)
2517 {
2518 	char buf[128], *src, *dst;
2519 	u_int seq;
2520 	int buflen, error;
2521 
2522 	if (args->type != LINUX_SYSLOG_ACTION_READ_ALL) {
2523 		linux_msg(td, "syslog unsupported type 0x%x", args->type);
2524 		return (EINVAL);
2525 	}
2526 
2527 	if (args->len < 6) {
2528 		td->td_retval[0] = 0;
2529 		return (0);
2530 	}
2531 
2532 	error = priv_check(td, PRIV_MSGBUF);
2533 	if (error)
2534 		return (error);
2535 
2536 	mtx_lock(&msgbuf_lock);
2537 	msgbuf_peekbytes(msgbufp, NULL, 0, &seq);
2538 	mtx_unlock(&msgbuf_lock);
2539 
2540 	dst = args->buf;
2541 	error = copyout(&SYSLOG_TAG, dst, sizeof(SYSLOG_TAG));
2542 	/* The -1 is to skip the trailing '\0'. */
2543 	dst += sizeof(SYSLOG_TAG) - 1;
2544 
2545 	while (error == 0) {
2546 		mtx_lock(&msgbuf_lock);
2547 		buflen = msgbuf_peekbytes(msgbufp, buf, sizeof(buf), &seq);
2548 		mtx_unlock(&msgbuf_lock);
2549 
2550 		if (buflen == 0)
2551 			break;
2552 
2553 		for (src = buf; src < buf + buflen && error == 0; src++) {
2554 			if (*src == '\0')
2555 				continue;
2556 
2557 			if (dst >= args->buf + args->len)
2558 				goto out;
2559 
2560 			error = copyout(src, dst, 1);
2561 			dst++;
2562 
2563 			if (*src == '\n' && *(src + 1) != '<' &&
2564 			    dst + sizeof(SYSLOG_TAG) < args->buf + args->len) {
2565 				error = copyout(&SYSLOG_TAG,
2566 				    dst, sizeof(SYSLOG_TAG));
2567 				dst += sizeof(SYSLOG_TAG) - 1;
2568 			}
2569 		}
2570 	}
2571 out:
2572 	td->td_retval[0] = dst - args->buf;
2573 	return (error);
2574 }
2575 
2576 int
2577 linux_getcpu(struct thread *td, struct linux_getcpu_args *args)
2578 {
2579 	int cpu, error, node;
2580 
2581 	cpu = td->td_oncpu; /* Make sure it doesn't change during copyout(9) */
2582 	error = 0;
2583 	node = cpuid_to_pcpu[cpu]->pc_domain;
2584 
2585 	if (args->cpu != NULL)
2586 		error = copyout(&cpu, args->cpu, sizeof(l_int));
2587 	if (args->node != NULL)
2588 		error = copyout(&node, args->node, sizeof(l_int));
2589 	return (error);
2590 }
2591 
2592 #if defined(__i386__) || defined(__amd64__)
2593 int
2594 linux_poll(struct thread *td, struct linux_poll_args *args)
2595 {
2596 	struct timespec ts, *tsp;
2597 
2598 	if (args->timeout != INFTIM) {
2599 		if (args->timeout < 0)
2600 			return (EINVAL);
2601 		ts.tv_sec = args->timeout / 1000;
2602 		ts.tv_nsec = (args->timeout % 1000) * 1000000;
2603 		tsp = &ts;
2604 	} else
2605 		tsp = NULL;
2606 
2607 	return (linux_common_ppoll(td, args->fds, args->nfds,
2608 	    tsp, NULL, 0));
2609 }
2610 #endif /* __i386__ || __amd64__ */
2611 
2612 int
2613 linux_seccomp(struct thread *td, struct linux_seccomp_args *args)
2614 {
2615 
2616 	switch (args->op) {
2617 	case LINUX_SECCOMP_GET_ACTION_AVAIL:
2618 		return (EOPNOTSUPP);
2619 	default:
2620 		/*
2621 		 * Ignore unknown operations, just like Linux kernel built
2622 		 * without CONFIG_SECCOMP.
2623 		 */
2624 		return (EINVAL);
2625 	}
2626 }
2627 
2628 /*
2629  * Custom version of exec_copyin_args(), to copy out argument and environment
2630  * strings from the old process address space into the temporary string buffer.
2631  * Based on freebsd32_exec_copyin_args.
2632  */
2633 static int
2634 linux_exec_copyin_args(struct image_args *args, const char *fname,
2635     l_uintptr_t *argv, l_uintptr_t *envv)
2636 {
2637 	char *argp, *envp;
2638 	l_uintptr_t *ptr, arg;
2639 	int error;
2640 
2641 	bzero(args, sizeof(*args));
2642 	if (argv == NULL)
2643 		return (EFAULT);
2644 
2645 	/*
2646 	 * Allocate demand-paged memory for the file name, argument, and
2647 	 * environment strings.
2648 	 */
2649 	error = exec_alloc_args(args);
2650 	if (error != 0)
2651 		return (error);
2652 
2653 	/*
2654 	 * Copy the file name.
2655 	 */
2656 	error = exec_args_add_fname(args, fname, UIO_USERSPACE);
2657 	if (error != 0)
2658 		goto err_exit;
2659 
2660 	/*
2661 	 * extract arguments first
2662 	 */
2663 	ptr = argv;
2664 	for (;;) {
2665 		error = copyin(ptr++, &arg, sizeof(arg));
2666 		if (error)
2667 			goto err_exit;
2668 		if (arg == 0)
2669 			break;
2670 		argp = PTRIN(arg);
2671 		error = exec_args_add_arg(args, argp, UIO_USERSPACE);
2672 		if (error != 0)
2673 			goto err_exit;
2674 	}
2675 
2676 	/*
2677 	 * This comment is from Linux do_execveat_common:
2678 	 * When argv is empty, add an empty string ("") as argv[0] to
2679 	 * ensure confused userspace programs that start processing
2680 	 * from argv[1] won't end up walking envp.
2681 	 */
2682 	if (args->argc == 0 &&
2683 	    (error = exec_args_add_arg(args, "", UIO_SYSSPACE) != 0))
2684 		goto err_exit;
2685 
2686 	/*
2687 	 * extract environment strings
2688 	 */
2689 	if (envv) {
2690 		ptr = envv;
2691 		for (;;) {
2692 			error = copyin(ptr++, &arg, sizeof(arg));
2693 			if (error)
2694 				goto err_exit;
2695 			if (arg == 0)
2696 				break;
2697 			envp = PTRIN(arg);
2698 			error = exec_args_add_env(args, envp, UIO_USERSPACE);
2699 			if (error != 0)
2700 				goto err_exit;
2701 		}
2702 	}
2703 
2704 	return (0);
2705 
2706 err_exit:
2707 	exec_free_args(args);
2708 	return (error);
2709 }
2710 
2711 int
2712 linux_execve(struct thread *td, struct linux_execve_args *args)
2713 {
2714 	struct image_args eargs;
2715 	int error;
2716 
2717 	LINUX_CTR(execve);
2718 
2719 	error = linux_exec_copyin_args(&eargs, args->path, args->argp,
2720 	    args->envp);
2721 	if (error == 0)
2722 		error = linux_common_execve(td, &eargs);
2723 	AUDIT_SYSCALL_EXIT(error == EJUSTRETURN ? 0 : error, td);
2724 	return (error);
2725 }
2726 
2727 static void
2728 linux_up_rtprio_if(struct thread *td1, struct rtprio *rtp)
2729 {
2730 	struct rtprio rtp2;
2731 
2732 	pri_to_rtp(td1, &rtp2);
2733 	if (rtp2.type <  rtp->type ||
2734 	    (rtp2.type == rtp->type &&
2735 	    rtp2.prio < rtp->prio)) {
2736 		rtp->type = rtp2.type;
2737 		rtp->prio = rtp2.prio;
2738 	}
2739 }
2740 
2741 #define	LINUX_PRIO_DIVIDER	RTP_PRIO_MAX / LINUX_IOPRIO_MAX
2742 
2743 static int
2744 linux_rtprio2ioprio(struct rtprio *rtp)
2745 {
2746 	int ioprio, prio;
2747 
2748 	switch (rtp->type) {
2749 	case RTP_PRIO_IDLE:
2750 		prio = RTP_PRIO_MIN;
2751 		ioprio = LINUX_IOPRIO_PRIO(LINUX_IOPRIO_CLASS_IDLE, prio);
2752 		break;
2753 	case RTP_PRIO_NORMAL:
2754 		prio = rtp->prio / LINUX_PRIO_DIVIDER;
2755 		ioprio = LINUX_IOPRIO_PRIO(LINUX_IOPRIO_CLASS_BE, prio);
2756 		break;
2757 	case RTP_PRIO_REALTIME:
2758 		prio = rtp->prio / LINUX_PRIO_DIVIDER;
2759 		ioprio = LINUX_IOPRIO_PRIO(LINUX_IOPRIO_CLASS_RT, prio);
2760 		break;
2761 	default:
2762 		prio = RTP_PRIO_MIN;
2763 		ioprio = LINUX_IOPRIO_PRIO(LINUX_IOPRIO_CLASS_NONE, prio);
2764 		break;
2765 	}
2766 	return (ioprio);
2767 }
2768 
2769 static int
2770 linux_ioprio2rtprio(int ioprio, struct rtprio *rtp)
2771 {
2772 
2773 	switch (LINUX_IOPRIO_PRIO_CLASS(ioprio)) {
2774 	case LINUX_IOPRIO_CLASS_IDLE:
2775 		rtp->prio = RTP_PRIO_MIN;
2776 		rtp->type = RTP_PRIO_IDLE;
2777 		break;
2778 	case LINUX_IOPRIO_CLASS_BE:
2779 		rtp->prio = LINUX_IOPRIO_PRIO_DATA(ioprio) * LINUX_PRIO_DIVIDER;
2780 		rtp->type = RTP_PRIO_NORMAL;
2781 		break;
2782 	case LINUX_IOPRIO_CLASS_RT:
2783 		rtp->prio = LINUX_IOPRIO_PRIO_DATA(ioprio) * LINUX_PRIO_DIVIDER;
2784 		rtp->type = RTP_PRIO_REALTIME;
2785 		break;
2786 	default:
2787 		return (EINVAL);
2788 	}
2789 	return (0);
2790 }
2791 #undef LINUX_PRIO_DIVIDER
2792 
2793 int
2794 linux_ioprio_get(struct thread *td, struct linux_ioprio_get_args *args)
2795 {
2796 	struct thread *td1;
2797 	struct rtprio rtp;
2798 	struct pgrp *pg;
2799 	struct proc *p;
2800 	int error, found;
2801 
2802 	p = NULL;
2803 	td1 = NULL;
2804 	error = 0;
2805 	found = 0;
2806 	rtp.type = RTP_PRIO_IDLE;
2807 	rtp.prio = RTP_PRIO_MAX;
2808 	switch (args->which) {
2809 	case LINUX_IOPRIO_WHO_PROCESS:
2810 		if (args->who == 0) {
2811 			td1 = td;
2812 			p = td1->td_proc;
2813 			PROC_LOCK(p);
2814 		} else if (args->who > PID_MAX) {
2815 			td1 = linux_tdfind(td, args->who, -1);
2816 			if (td1 != NULL)
2817 				p = td1->td_proc;
2818 		} else
2819 			p = pfind(args->who);
2820 		if (p == NULL)
2821 			return (ESRCH);
2822 		if ((error = p_cansee(td, p))) {
2823 			PROC_UNLOCK(p);
2824 			break;
2825 		}
2826 		if (td1 != NULL) {
2827 			pri_to_rtp(td1, &rtp);
2828 		} else {
2829 			FOREACH_THREAD_IN_PROC(p, td1) {
2830 				linux_up_rtprio_if(td1, &rtp);
2831 			}
2832 		}
2833 		found++;
2834 		PROC_UNLOCK(p);
2835 		break;
2836 	case LINUX_IOPRIO_WHO_PGRP:
2837 		sx_slock(&proctree_lock);
2838 		if (args->who == 0) {
2839 			pg = td->td_proc->p_pgrp;
2840 			PGRP_LOCK(pg);
2841 		} else {
2842 			pg = pgfind(args->who);
2843 			if (pg == NULL) {
2844 				sx_sunlock(&proctree_lock);
2845 				error = ESRCH;
2846 				break;
2847 			}
2848 		}
2849 		sx_sunlock(&proctree_lock);
2850 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
2851 			PROC_LOCK(p);
2852 			if (p->p_state == PRS_NORMAL &&
2853 			    p_cansee(td, p) == 0) {
2854 				FOREACH_THREAD_IN_PROC(p, td1) {
2855 					linux_up_rtprio_if(td1, &rtp);
2856 					found++;
2857 				}
2858 			}
2859 			PROC_UNLOCK(p);
2860 		}
2861 		PGRP_UNLOCK(pg);
2862 		break;
2863 	case LINUX_IOPRIO_WHO_USER:
2864 		if (args->who == 0)
2865 			args->who = td->td_ucred->cr_uid;
2866 		sx_slock(&allproc_lock);
2867 		FOREACH_PROC_IN_SYSTEM(p) {
2868 			PROC_LOCK(p);
2869 			if (p->p_state == PRS_NORMAL &&
2870 			    p->p_ucred->cr_uid == args->who &&
2871 			    p_cansee(td, p) == 0) {
2872 				FOREACH_THREAD_IN_PROC(p, td1) {
2873 					linux_up_rtprio_if(td1, &rtp);
2874 					found++;
2875 				}
2876 			}
2877 			PROC_UNLOCK(p);
2878 		}
2879 		sx_sunlock(&allproc_lock);
2880 		break;
2881 	default:
2882 		error = EINVAL;
2883 		break;
2884 	}
2885 	if (error == 0) {
2886 		if (found != 0)
2887 			td->td_retval[0] = linux_rtprio2ioprio(&rtp);
2888 		else
2889 			error = ESRCH;
2890 	}
2891 	return (error);
2892 }
2893 
2894 int
2895 linux_ioprio_set(struct thread *td, struct linux_ioprio_set_args *args)
2896 {
2897 	struct thread *td1;
2898 	struct rtprio rtp;
2899 	struct pgrp *pg;
2900 	struct proc *p;
2901 	int error;
2902 
2903 	if ((error = linux_ioprio2rtprio(args->ioprio, &rtp)) != 0)
2904 		return (error);
2905 	/* Attempts to set high priorities (REALTIME) require su privileges. */
2906 	if (RTP_PRIO_BASE(rtp.type) == RTP_PRIO_REALTIME &&
2907 	    (error = priv_check(td, PRIV_SCHED_RTPRIO)) != 0)
2908 		return (error);
2909 
2910 	p = NULL;
2911 	td1 = NULL;
2912 	switch (args->which) {
2913 	case LINUX_IOPRIO_WHO_PROCESS:
2914 		if (args->who == 0) {
2915 			td1 = td;
2916 			p = td1->td_proc;
2917 			PROC_LOCK(p);
2918 		} else if (args->who > PID_MAX) {
2919 			td1 = linux_tdfind(td, args->who, -1);
2920 			if (td1 != NULL)
2921 				p = td1->td_proc;
2922 		} else
2923 			p = pfind(args->who);
2924 		if (p == NULL)
2925 			return (ESRCH);
2926 		if ((error = p_cansched(td, p))) {
2927 			PROC_UNLOCK(p);
2928 			break;
2929 		}
2930 		if (td1 != NULL) {
2931 			error = rtp_to_pri(&rtp, td1);
2932 		} else {
2933 			FOREACH_THREAD_IN_PROC(p, td1) {
2934 				if ((error = rtp_to_pri(&rtp, td1)) != 0)
2935 					break;
2936 			}
2937 		}
2938 		PROC_UNLOCK(p);
2939 		break;
2940 	case LINUX_IOPRIO_WHO_PGRP:
2941 		sx_slock(&proctree_lock);
2942 		if (args->who == 0) {
2943 			pg = td->td_proc->p_pgrp;
2944 			PGRP_LOCK(pg);
2945 		} else {
2946 			pg = pgfind(args->who);
2947 			if (pg == NULL) {
2948 				sx_sunlock(&proctree_lock);
2949 				error = ESRCH;
2950 				break;
2951 			}
2952 		}
2953 		sx_sunlock(&proctree_lock);
2954 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
2955 			PROC_LOCK(p);
2956 			if (p->p_state == PRS_NORMAL &&
2957 			    p_cansched(td, p) == 0) {
2958 				FOREACH_THREAD_IN_PROC(p, td1) {
2959 					if ((error = rtp_to_pri(&rtp, td1)) != 0)
2960 						break;
2961 				}
2962 			}
2963 			PROC_UNLOCK(p);
2964 			if (error != 0)
2965 				break;
2966 		}
2967 		PGRP_UNLOCK(pg);
2968 		break;
2969 	case LINUX_IOPRIO_WHO_USER:
2970 		if (args->who == 0)
2971 			args->who = td->td_ucred->cr_uid;
2972 		sx_slock(&allproc_lock);
2973 		FOREACH_PROC_IN_SYSTEM(p) {
2974 			PROC_LOCK(p);
2975 			if (p->p_state == PRS_NORMAL &&
2976 			    p->p_ucred->cr_uid == args->who &&
2977 			    p_cansched(td, p) == 0) {
2978 				FOREACH_THREAD_IN_PROC(p, td1) {
2979 					if ((error = rtp_to_pri(&rtp, td1)) != 0)
2980 						break;
2981 				}
2982 			}
2983 			PROC_UNLOCK(p);
2984 			if (error != 0)
2985 				break;
2986 		}
2987 		sx_sunlock(&allproc_lock);
2988 		break;
2989 	default:
2990 		error = EINVAL;
2991 		break;
2992 	}
2993 	return (error);
2994 }
2995 
2996 /* The only flag is O_NONBLOCK */
2997 #define B2L_MQ_FLAGS(bflags)	((bflags) != 0 ? LINUX_O_NONBLOCK : 0)
2998 #define L2B_MQ_FLAGS(lflags)	((lflags) != 0 ? O_NONBLOCK : 0)
2999 
3000 int
3001 linux_mq_open(struct thread *td, struct linux_mq_open_args *args)
3002 {
3003 	struct mq_attr attr;
3004 	int error, flags;
3005 
3006 	flags = linux_common_openflags(args->oflag);
3007 	if ((flags & O_ACCMODE) == O_ACCMODE || (flags & O_EXEC) != 0)
3008 		return (EINVAL);
3009 	flags = FFLAGS(flags);
3010 	if ((flags & O_CREAT) != 0 && args->attr != NULL) {
3011 		error = copyin(args->attr, &attr, sizeof(attr));
3012 		if (error != 0)
3013 			return (error);
3014 		attr.mq_flags = L2B_MQ_FLAGS(attr.mq_flags);
3015 	}
3016 
3017 	return (kern_kmq_open(td, args->name, flags, args->mode,
3018 	    args->attr != NULL ? &attr : NULL));
3019 }
3020 
3021 int
3022 linux_mq_unlink(struct thread *td, struct linux_mq_unlink_args *args)
3023 {
3024 	struct kmq_unlink_args bsd_args = {
3025 		.path = PTRIN(args->name)
3026 	};
3027 
3028 	return (sys_kmq_unlink(td, &bsd_args));
3029 }
3030 
3031 int
3032 linux_mq_timedsend(struct thread *td, struct linux_mq_timedsend_args *args)
3033 {
3034 	struct timespec ts, *abs_timeout;
3035 	int error;
3036 
3037 	if (args->abs_timeout == NULL)
3038 		abs_timeout = NULL;
3039 	else {
3040 		error = linux_get_timespec(&ts, args->abs_timeout);
3041 		if (error != 0)
3042 			return (error);
3043 		abs_timeout = &ts;
3044 	}
3045 
3046 	return (kern_kmq_timedsend(td, args->mqd, PTRIN(args->msg_ptr),
3047 		args->msg_len, args->msg_prio, abs_timeout));
3048 }
3049 
3050 int
3051 linux_mq_timedreceive(struct thread *td, struct linux_mq_timedreceive_args *args)
3052 {
3053 	struct timespec ts, *abs_timeout;
3054 	int error;
3055 
3056 	if (args->abs_timeout == NULL)
3057 		abs_timeout = NULL;
3058 	else {
3059 		error = linux_get_timespec(&ts, args->abs_timeout);
3060 		if (error != 0)
3061 			return (error);
3062 		abs_timeout = &ts;
3063 	}
3064 
3065 	return (kern_kmq_timedreceive(td, args->mqd, PTRIN(args->msg_ptr),
3066 		args->msg_len, args->msg_prio, abs_timeout));
3067 }
3068 
3069 int
3070 linux_mq_notify(struct thread *td, struct linux_mq_notify_args *args)
3071 {
3072 	struct sigevent ev, *evp;
3073 	struct l_sigevent l_ev;
3074 	int error;
3075 
3076 	if (args->sevp == NULL)
3077 		evp = NULL;
3078 	else {
3079 		error = copyin(args->sevp, &l_ev, sizeof(l_ev));
3080 		if (error != 0)
3081 			return (error);
3082 		error = linux_convert_l_sigevent(&l_ev, &ev);
3083 		if (error != 0)
3084 			return (error);
3085 		evp = &ev;
3086 	}
3087 
3088 	return (kern_kmq_notify(td, args->mqd, evp));
3089 }
3090 
3091 int
3092 linux_mq_getsetattr(struct thread *td, struct linux_mq_getsetattr_args *args)
3093 {
3094 	struct mq_attr attr, oattr;
3095 	int error;
3096 
3097 	if (args->attr != NULL) {
3098 		error = copyin(args->attr, &attr, sizeof(attr));
3099 		if (error != 0)
3100 			return (error);
3101 		attr.mq_flags = L2B_MQ_FLAGS(attr.mq_flags);
3102 	}
3103 
3104 	error = kern_kmq_setattr(td, args->mqd, args->attr != NULL ? &attr : NULL,
3105 	    &oattr);
3106 	if (error == 0 && args->oattr != NULL) {
3107 		oattr.mq_flags = B2L_MQ_FLAGS(oattr.mq_flags);
3108 		bzero(oattr.__reserved, sizeof(oattr.__reserved));
3109 		error = copyout(&oattr, args->oattr, sizeof(oattr));
3110 	}
3111 
3112 	return (error);
3113 }
3114 
3115 int
3116 linux_kcmp(struct thread *td, struct linux_kcmp_args *args)
3117 {
3118 	int type;
3119 
3120 	switch (args->type) {
3121 	case LINUX_KCMP_FILE:
3122 		type = KCMP_FILE;
3123 		break;
3124 	case LINUX_KCMP_FILES:
3125 		type = KCMP_FILES;
3126 		break;
3127 	case LINUX_KCMP_SIGHAND:
3128 		type = KCMP_SIGHAND;
3129 		break;
3130 	case LINUX_KCMP_VM:
3131 		type = KCMP_VM;
3132 		break;
3133 	default:
3134 		return (EINVAL);
3135 	}
3136 
3137 	return (kern_kcmp(td, args->pid1, args->pid2, type, args->idx1,
3138 	    args->idx));
3139 }
3140 
3141 int
3142 linux_membarrier(struct thread *td, struct linux_membarrier_args *args)
3143 {
3144 	static const struct {
3145 		int linux_cmd;
3146 		int freebsd_cmd;
3147 	} cmds[] = {
3148 		{ LINUX_MEMBARRIER_CMD_QUERY,
3149 		    MEMBARRIER_CMD_QUERY },
3150 		{ LINUX_MEMBARRIER_CMD_GLOBAL,
3151 		    MEMBARRIER_CMD_GLOBAL },
3152 		{ LINUX_MEMBARRIER_CMD_GLOBAL_EXPEDITED,
3153 		    MEMBARRIER_CMD_GLOBAL_EXPEDITED },
3154 		{ LINUX_MEMBARRIER_CMD_REGISTER_GLOBAL_EXPEDITED,
3155 		    MEMBARRIER_CMD_REGISTER_GLOBAL_EXPEDITED },
3156 		{ LINUX_MEMBARRIER_CMD_PRIVATE_EXPEDITED,
3157 		    MEMBARRIER_CMD_PRIVATE_EXPEDITED },
3158 		{ LINUX_MEMBARRIER_CMD_REGISTER_PRIVATE_EXPEDITED,
3159 		    MEMBARRIER_CMD_REGISTER_PRIVATE_EXPEDITED },
3160 		{ LINUX_MEMBARRIER_CMD_PRIVATE_EXPEDITED_SYNC_CORE,
3161 		    MEMBARRIER_CMD_PRIVATE_EXPEDITED_SYNC_CORE },
3162 		{ LINUX_MEMBARRIER_CMD_REGISTER_PRIVATE_EXPEDITED_SYNC_CORE,
3163 		    MEMBARRIER_CMD_REGISTER_PRIVATE_EXPEDITED_SYNC_CORE },
3164 		{ LINUX_MEMBARRIER_CMD_PRIVATE_EXPEDITED_RSEQ,
3165 		    MEMBARRIER_CMD_PRIVATE_EXPEDITED_RSEQ },
3166 		{ LINUX_MEMBARRIER_CMD_REGISTER_PRIVATE_EXPEDITED_RSEQ,
3167 		    MEMBARRIER_CMD_REGISTER_PRIVATE_EXPEDITED_RSEQ },
3168 		{ LINUX_MEMBARRIER_CMD_GET_REGISTRATIONS,
3169 		    MEMBARRIER_CMD_GET_REGISTRATIONS },
3170 	};
3171 	int cmd, error, flags, i, mask;
3172 
3173 	cmd = -1;
3174 	for (i = 0; i < nitems(cmds); i++) {
3175 		if (args->cmd == cmds[i].linux_cmd) {
3176 			cmd = cmds[i].freebsd_cmd;
3177 			break;
3178 		}
3179 	}
3180 
3181 	if (cmd == -1 || (args->flags & ~LINUX_MEMBARRIER_CMD_FLAG_CPU) != 0)
3182 		return (EINVAL);
3183 
3184 	flags = 0;
3185 	if ((args->flags & LINUX_MEMBARRIER_CMD_FLAG_CPU) != 0)
3186 		flags |= MEMBARRIER_CMD_FLAG_CPU;
3187 
3188 	error = kern_membarrier(td, cmd, flags, args->cpu_id);
3189 	if (error != 0)
3190 		return (error);
3191 
3192 	if (args->cmd == LINUX_MEMBARRIER_CMD_QUERY ||
3193 	    args->cmd == LINUX_MEMBARRIER_CMD_GET_REGISTRATIONS) {
3194 		mask = td->td_retval[0];
3195 		td->td_retval[0] = 0;
3196 		for (i = 0; i < nitems(cmds); i++)
3197 			if ((mask & cmds[i].freebsd_cmd) != 0)
3198 				td->td_retval[0] |= cmds[i].linux_cmd;
3199 	}
3200 
3201 	return (0);
3202 }
3203 
3204 /*
3205  * setfsuid() & setfsgid() exist to decouple the Linux filesystem credentials
3206  * from the effective credentials, avoiding signal exposure during privilege
3207  * transitions. The signal permission model that motivated this was revised in
3208  * Linux 2.0, making these syscalls obsolete for new applications.
3209  *
3210  * As there's no FreeBSD equivalent, implement both syscalls as no-ops that
3211  * return the current effective UID/GID as the previous filesystem UID/GID.
3212  * Linux returns the previous filesystem UID/GID for these syscalls, with no
3213  * error indication.
3214  */
3215 
3216 int
3217 linux_setfsuid(struct thread *td, struct linux_setfsuid_args *args)
3218 {
3219 	td->td_retval[0] = td->td_ucred->cr_uid;
3220 	return (0);
3221 }
3222 
3223 int
3224 linux_setfsgid(struct thread *td, struct linux_setfsgid_args *args)
3225 {
3226 	td->td_retval[0] = td->td_ucred->cr_gid;
3227 	return (0);
3228 }
3229 
3230 MODULE_DEPEND(linux, mqueuefs, 1, 1, 1);
3231