xref: /freebsd/sys/compat/linux/linux_misc.c (revision 5945da0bc9ac42f531b1079a246eb8ce4f0d63db)
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 	exit1(td, args->error_code, 0);
1549 		/* NOTREACHED */
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 	default:
1839 		linux_msg(td, "unsupported prctl option %d", args->option);
1840 		error = EINVAL;
1841 		break;
1842 	}
1843 
1844 	return (error);
1845 }
1846 
1847 int
1848 linux_sched_setparam(struct thread *td,
1849     struct linux_sched_setparam_args *uap)
1850 {
1851 	struct sched_param sched_param;
1852 	struct thread *tdt;
1853 	int error, policy;
1854 
1855 	error = copyin(uap->param, &sched_param, sizeof(sched_param));
1856 	if (error)
1857 		return (error);
1858 
1859 	tdt = linux_tdfind(td, uap->pid, -1);
1860 	if (tdt == NULL)
1861 		return (ESRCH);
1862 
1863 	if (linux_map_sched_prio) {
1864 		error = kern_sched_getscheduler(td, tdt, &policy);
1865 		if (error)
1866 			goto out;
1867 
1868 		switch (policy) {
1869 		case SCHED_OTHER:
1870 			if (sched_param.sched_priority != 0) {
1871 				error = EINVAL;
1872 				goto out;
1873 			}
1874 			sched_param.sched_priority =
1875 			    PRI_MAX_TIMESHARE - PRI_MIN_TIMESHARE;
1876 			break;
1877 		case SCHED_FIFO:
1878 		case SCHED_RR:
1879 			if (sched_param.sched_priority < 1 ||
1880 			    sched_param.sched_priority >= LINUX_MAX_RT_PRIO) {
1881 				error = EINVAL;
1882 				goto out;
1883 			}
1884 			/*
1885 			 * Map [1, LINUX_MAX_RT_PRIO - 1] to
1886 			 * [0, RTP_PRIO_MAX - RTP_PRIO_MIN] (rounding down).
1887 			 */
1888 			sched_param.sched_priority =
1889 			    (sched_param.sched_priority - 1) *
1890 			    (RTP_PRIO_MAX - RTP_PRIO_MIN + 1) /
1891 			    (LINUX_MAX_RT_PRIO - 1);
1892 			break;
1893 		}
1894 	}
1895 
1896 	error = kern_sched_setparam(td, tdt, &sched_param);
1897 out:	PROC_UNLOCK(tdt->td_proc);
1898 	return (error);
1899 }
1900 
1901 int
1902 linux_sched_getparam(struct thread *td,
1903     struct linux_sched_getparam_args *uap)
1904 {
1905 	struct sched_param sched_param;
1906 	struct thread *tdt;
1907 	int error, policy;
1908 
1909 	tdt = linux_tdfind(td, uap->pid, -1);
1910 	if (tdt == NULL)
1911 		return (ESRCH);
1912 
1913 	error = kern_sched_getparam(td, tdt, &sched_param);
1914 	if (error) {
1915 		PROC_UNLOCK(tdt->td_proc);
1916 		return (error);
1917 	}
1918 
1919 	if (linux_map_sched_prio) {
1920 		error = kern_sched_getscheduler(td, tdt, &policy);
1921 		PROC_UNLOCK(tdt->td_proc);
1922 		if (error)
1923 			return (error);
1924 
1925 		switch (policy) {
1926 		case SCHED_OTHER:
1927 			sched_param.sched_priority = 0;
1928 			break;
1929 		case SCHED_FIFO:
1930 		case SCHED_RR:
1931 			/*
1932 			 * Map [0, RTP_PRIO_MAX - RTP_PRIO_MIN] to
1933 			 * [1, LINUX_MAX_RT_PRIO - 1] (rounding up).
1934 			 */
1935 			sched_param.sched_priority =
1936 			    (sched_param.sched_priority *
1937 			    (LINUX_MAX_RT_PRIO - 1) +
1938 			    (RTP_PRIO_MAX - RTP_PRIO_MIN - 1)) /
1939 			    (RTP_PRIO_MAX - RTP_PRIO_MIN) + 1;
1940 			break;
1941 		}
1942 	} else
1943 		PROC_UNLOCK(tdt->td_proc);
1944 
1945 	error = copyout(&sched_param, uap->param, sizeof(sched_param));
1946 	return (error);
1947 }
1948 
1949 /*
1950  * Get affinity of a process.
1951  */
1952 int
1953 linux_sched_getaffinity(struct thread *td,
1954     struct linux_sched_getaffinity_args *args)
1955 {
1956 	struct thread *tdt;
1957 	cpuset_t *mask;
1958 	size_t size;
1959 	int error;
1960 	id_t tid;
1961 
1962 	tdt = linux_tdfind(td, args->pid, -1);
1963 	if (tdt == NULL)
1964 		return (ESRCH);
1965 	tid = tdt->td_tid;
1966 	PROC_UNLOCK(tdt->td_proc);
1967 
1968 	mask = malloc(sizeof(cpuset_t), M_LINUX, M_WAITOK | M_ZERO);
1969 	size = min(args->len, sizeof(cpuset_t));
1970 	error = kern_cpuset_getaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
1971 	    tid, size, mask);
1972 	if (error == ERANGE)
1973 		error = EINVAL;
1974  	if (error == 0)
1975 		error = copyout(mask, args->user_mask_ptr, size);
1976 	if (error == 0)
1977 		td->td_retval[0] = size;
1978 	free(mask, M_LINUX);
1979 	return (error);
1980 }
1981 
1982 /*
1983  *  Set affinity of a process.
1984  */
1985 int
1986 linux_sched_setaffinity(struct thread *td,
1987     struct linux_sched_setaffinity_args *args)
1988 {
1989 	struct thread *tdt;
1990 	cpuset_t *mask;
1991 	int cpu, error;
1992 	size_t len;
1993 	id_t tid;
1994 
1995 	tdt = linux_tdfind(td, args->pid, -1);
1996 	if (tdt == NULL)
1997 		return (ESRCH);
1998 	tid = tdt->td_tid;
1999 	PROC_UNLOCK(tdt->td_proc);
2000 
2001 	len = min(args->len, sizeof(cpuset_t));
2002 	mask = malloc(sizeof(cpuset_t), M_TEMP, M_WAITOK | M_ZERO);
2003 	error = copyin(args->user_mask_ptr, mask, len);
2004 	if (error != 0)
2005 		goto out;
2006 	/* Linux ignore high bits */
2007 	CPU_FOREACH_ISSET(cpu, mask)
2008 		if (cpu > mp_maxid)
2009 			CPU_CLR(cpu, mask);
2010 
2011 	error = kern_cpuset_setaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2012 	    tid, mask);
2013 	if (error == EDEADLK)
2014 		error = EINVAL;
2015 out:
2016 	free(mask, M_TEMP);
2017 	return (error);
2018 }
2019 
2020 struct linux_rlimit64 {
2021 	uint64_t	rlim_cur;
2022 	uint64_t	rlim_max;
2023 };
2024 
2025 int
2026 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
2027 {
2028 	struct rlimit rlim, nrlim;
2029 	struct linux_rlimit64 lrlim;
2030 	struct proc *p;
2031 	u_int which;
2032 	int flags;
2033 	int error;
2034 
2035 	if (args->new == NULL && args->old != NULL) {
2036 		if (linux_get_dummy_limit(td, args->resource, &rlim)) {
2037 			lrlim.rlim_cur = rlim.rlim_cur;
2038 			lrlim.rlim_max = rlim.rlim_max;
2039 			return (copyout(&lrlim, args->old, sizeof(lrlim)));
2040 		}
2041 	}
2042 
2043 	if (args->resource >= LINUX_RLIM_NLIMITS)
2044 		return (EINVAL);
2045 
2046 	which = linux_to_bsd_resource[args->resource];
2047 	if (which == -1)
2048 		return (EINVAL);
2049 
2050 	if (args->new != NULL) {
2051 		/*
2052 		 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
2053 		 * rlim is unsigned 64-bit. FreeBSD treats negative limits
2054 		 * as INFINITY so we do not need a conversion even.
2055 		 */
2056 		error = copyin(args->new, &nrlim, sizeof(nrlim));
2057 		if (error != 0)
2058 			return (error);
2059 	}
2060 
2061 	flags = PGET_HOLD | PGET_NOTWEXIT;
2062 	if (args->new != NULL)
2063 		flags |= PGET_CANDEBUG;
2064 	else
2065 		flags |= PGET_CANSEE;
2066 	if (args->pid == 0) {
2067 		p = td->td_proc;
2068 		PHOLD(p);
2069 	} else {
2070 		error = pget(args->pid, flags, &p);
2071 		if (error != 0)
2072 			return (error);
2073 	}
2074 	if (args->old != NULL) {
2075 		PROC_LOCK(p);
2076 		lim_rlimit_proc(p, which, &rlim);
2077 		PROC_UNLOCK(p);
2078 		if (rlim.rlim_cur == RLIM_INFINITY)
2079 			lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2080 		else
2081 			lrlim.rlim_cur = rlim.rlim_cur;
2082 		if (rlim.rlim_max == RLIM_INFINITY)
2083 			lrlim.rlim_max = LINUX_RLIM_INFINITY;
2084 		else
2085 			lrlim.rlim_max = rlim.rlim_max;
2086 		error = copyout(&lrlim, args->old, sizeof(lrlim));
2087 		if (error != 0)
2088 			goto out;
2089 	}
2090 
2091 	if (args->new != NULL)
2092 		error = kern_proc_setrlimit(td, p, which, &nrlim);
2093 
2094  out:
2095 	PRELE(p);
2096 	return (error);
2097 }
2098 
2099 int
2100 linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2101 {
2102 	struct timespec ts, *tsp;
2103 	int error;
2104 
2105 	if (args->tsp != NULL) {
2106 		error = linux_get_timespec(&ts, args->tsp);
2107 		if (error != 0)
2108 			return (error);
2109 		tsp = &ts;
2110 	} else
2111 		tsp = NULL;
2112 
2113 	error = linux_common_pselect6(td, args->nfds, args->readfds,
2114 	    args->writefds, args->exceptfds, tsp, args->sig);
2115 
2116 	if (args->tsp != NULL)
2117 		linux_put_timespec(&ts, args->tsp);
2118 	return (error);
2119 }
2120 
2121 static int
2122 linux_common_pselect6(struct thread *td, l_int nfds, l_fd_set *readfds,
2123     l_fd_set *writefds, l_fd_set *exceptfds, struct timespec *tsp,
2124     l_uintptr_t *sig)
2125 {
2126 	struct timeval utv, tv0, tv1, *tvp;
2127 	struct l_pselect6arg lpse6;
2128 	sigset_t *ssp;
2129 	sigset_t ss;
2130 	int error;
2131 
2132 	ssp = NULL;
2133 	if (sig != NULL) {
2134 		error = copyin(sig, &lpse6, sizeof(lpse6));
2135 		if (error != 0)
2136 			return (error);
2137 		error = linux_copyin_sigset(td, PTRIN(lpse6.ss),
2138 		    lpse6.ss_len, &ss, &ssp);
2139 		if (error != 0)
2140 		    return (error);
2141 	} else
2142 		ssp = NULL;
2143 
2144 	/*
2145 	 * Currently glibc changes nanosecond number to microsecond.
2146 	 * This mean losing precision but for now it is hardly seen.
2147 	 */
2148 	if (tsp != NULL) {
2149 		TIMESPEC_TO_TIMEVAL(&utv, tsp);
2150 		if (itimerfix(&utv))
2151 			return (EINVAL);
2152 
2153 		microtime(&tv0);
2154 		tvp = &utv;
2155 	} else
2156 		tvp = NULL;
2157 
2158 	error = kern_pselect(td, nfds, readfds, writefds,
2159 	    exceptfds, tvp, ssp, LINUX_NFDBITS);
2160 
2161 	if (tsp != NULL) {
2162 		/*
2163 		 * Compute how much time was left of the timeout,
2164 		 * by subtracting the current time and the time
2165 		 * before we started the call, and subtracting
2166 		 * that result from the user-supplied value.
2167 		 */
2168 		microtime(&tv1);
2169 		timevalsub(&tv1, &tv0);
2170 		timevalsub(&utv, &tv1);
2171 		if (utv.tv_sec < 0)
2172 			timevalclear(&utv);
2173 		TIMEVAL_TO_TIMESPEC(&utv, tsp);
2174 	}
2175 	return (error);
2176 }
2177 
2178 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
2179 int
2180 linux_pselect6_time64(struct thread *td,
2181     struct linux_pselect6_time64_args *args)
2182 {
2183 	struct timespec ts, *tsp;
2184 	int error;
2185 
2186 	if (args->tsp != NULL) {
2187 		error = linux_get_timespec64(&ts, args->tsp);
2188 		if (error != 0)
2189 			return (error);
2190 		tsp = &ts;
2191 	} else
2192 		tsp = NULL;
2193 
2194 	error = linux_common_pselect6(td, args->nfds, args->readfds,
2195 	    args->writefds, args->exceptfds, tsp, args->sig);
2196 
2197 	if (args->tsp != NULL)
2198 		linux_put_timespec64(&ts, args->tsp);
2199 	return (error);
2200 }
2201 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
2202 
2203 int
2204 linux_ppoll(struct thread *td, struct linux_ppoll_args *args)
2205 {
2206 	struct timespec uts, *tsp;
2207 	int error;
2208 
2209 	if (args->tsp != NULL) {
2210 		error = linux_get_timespec(&uts, args->tsp);
2211 		if (error != 0)
2212 			return (error);
2213 		tsp = &uts;
2214 	} else
2215 		tsp = NULL;
2216 
2217 	error = linux_common_ppoll(td, args->fds, args->nfds, tsp,
2218 	    args->sset, args->ssize);
2219 	if (error == 0 && args->tsp != NULL)
2220 		error = linux_put_timespec(&uts, args->tsp);
2221 	return (error);
2222 }
2223 
2224 static int
2225 linux_common_ppoll(struct thread *td, struct pollfd *fds, uint32_t nfds,
2226     struct timespec *tsp, l_sigset_t *sset, l_size_t ssize)
2227 {
2228 	struct timespec ts0, ts1;
2229 	struct pollfd stackfds[32];
2230 	struct pollfd *kfds;
2231  	sigset_t *ssp;
2232  	sigset_t ss;
2233  	int error;
2234 
2235 	if (kern_poll_maxfds(nfds))
2236 		return (EINVAL);
2237 	if (sset != NULL) {
2238 		error = linux_copyin_sigset(td, sset, ssize, &ss, &ssp);
2239 		if (error != 0)
2240 		    return (error);
2241 	} else
2242 		ssp = NULL;
2243 	if (tsp != NULL)
2244 		nanotime(&ts0);
2245 
2246 	if (nfds > nitems(stackfds))
2247 		kfds = mallocarray(nfds, sizeof(*kfds), M_TEMP, M_WAITOK);
2248 	else
2249 		kfds = stackfds;
2250 	error = linux_pollin(td, kfds, fds, nfds);
2251 	if (error != 0)
2252 		goto out;
2253 
2254 	error = kern_poll_kfds(td, kfds, nfds, tsp, ssp);
2255 	if (error == 0)
2256 		error = linux_pollout(td, kfds, fds, nfds);
2257 
2258 	if (error == 0 && tsp != NULL) {
2259 		if (td->td_retval[0]) {
2260 			nanotime(&ts1);
2261 			timespecsub(&ts1, &ts0, &ts1);
2262 			timespecsub(tsp, &ts1, tsp);
2263 			if (tsp->tv_sec < 0)
2264 				timespecclear(tsp);
2265 		} else
2266 			timespecclear(tsp);
2267 	}
2268 
2269 out:
2270 	if (nfds > nitems(stackfds))
2271 		free(kfds, M_TEMP);
2272 	return (error);
2273 }
2274 
2275 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
2276 int
2277 linux_ppoll_time64(struct thread *td, struct linux_ppoll_time64_args *args)
2278 {
2279 	struct timespec uts, *tsp;
2280 	int error;
2281 
2282 	if (args->tsp != NULL) {
2283 		error = linux_get_timespec64(&uts, args->tsp);
2284 		if (error != 0)
2285 			return (error);
2286 		tsp = &uts;
2287 	} else
2288  		tsp = NULL;
2289 	error = linux_common_ppoll(td, args->fds, args->nfds, tsp,
2290 	    args->sset, args->ssize);
2291 	if (error == 0 && args->tsp != NULL)
2292 		error = linux_put_timespec64(&uts, args->tsp);
2293 	return (error);
2294 }
2295 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
2296 
2297 static int
2298 linux_pollin(struct thread *td, struct pollfd *fds, struct pollfd *ufds, u_int nfd)
2299 {
2300 	int error;
2301 	u_int i;
2302 
2303 	error = copyin(ufds, fds, nfd * sizeof(*fds));
2304 	if (error != 0)
2305 		return (error);
2306 
2307 	for (i = 0; i < nfd; i++) {
2308 		if (fds->events != 0)
2309 			linux_to_bsd_poll_events(td, fds->fd,
2310 			    fds->events, &fds->events);
2311 		fds++;
2312 	}
2313 	return (0);
2314 }
2315 
2316 static int
2317 linux_pollout(struct thread *td, struct pollfd *fds, struct pollfd *ufds, u_int nfd)
2318 {
2319 	int error = 0;
2320 	u_int i, n = 0;
2321 
2322 	for (i = 0; i < nfd; i++) {
2323 		if (fds->revents != 0) {
2324 			bsd_to_linux_poll_events(fds->revents,
2325 			    &fds->revents);
2326 			n++;
2327 		}
2328 		error = copyout(&fds->revents, &ufds->revents,
2329 		    sizeof(ufds->revents));
2330 		if (error)
2331 			return (error);
2332 		fds++;
2333 		ufds++;
2334 	}
2335 	td->td_retval[0] = n;
2336 	return (0);
2337 }
2338 
2339 static int
2340 linux_sched_rr_get_interval_common(struct thread *td, pid_t pid,
2341     struct timespec *ts)
2342 {
2343 	struct thread *tdt;
2344 	int error;
2345 
2346 	/*
2347 	 * According to man in case the invalid pid specified
2348 	 * EINVAL should be returned.
2349 	 */
2350 	if (pid < 0)
2351 		return (EINVAL);
2352 
2353 	tdt = linux_tdfind(td, pid, -1);
2354 	if (tdt == NULL)
2355 		return (ESRCH);
2356 
2357 	error = kern_sched_rr_get_interval_td(td, tdt, ts);
2358 	PROC_UNLOCK(tdt->td_proc);
2359 	return (error);
2360 }
2361 
2362 int
2363 linux_sched_rr_get_interval(struct thread *td,
2364     struct linux_sched_rr_get_interval_args *uap)
2365 {
2366 	struct timespec ts;
2367 	int error;
2368 
2369 	error = linux_sched_rr_get_interval_common(td, uap->pid, &ts);
2370 	if (error != 0)
2371 		return (error);
2372 	return (linux_put_timespec(&ts, uap->interval));
2373 }
2374 
2375 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
2376 int
2377 linux_sched_rr_get_interval_time64(struct thread *td,
2378     struct linux_sched_rr_get_interval_time64_args *uap)
2379 {
2380 	struct timespec ts;
2381 	int error;
2382 
2383 	error = linux_sched_rr_get_interval_common(td, uap->pid, &ts);
2384 	if (error != 0)
2385 		return (error);
2386 	return (linux_put_timespec64(&ts, uap->interval));
2387 }
2388 #endif
2389 
2390 /*
2391  * In case when the Linux thread is the initial thread in
2392  * the thread group thread id is equal to the process id.
2393  * Glibc depends on this magic (assert in pthread_getattr_np.c).
2394  */
2395 struct thread *
2396 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid)
2397 {
2398 	struct linux_emuldata *em;
2399 	struct thread *tdt;
2400 	struct proc *p;
2401 
2402 	tdt = NULL;
2403 	if (tid == 0 || tid == td->td_tid) {
2404 		if (pid != -1 && td->td_proc->p_pid != pid)
2405 			return (NULL);
2406 		PROC_LOCK(td->td_proc);
2407 		return (td);
2408 	} else if (tid > PID_MAX)
2409 		return (tdfind(tid, pid));
2410 
2411 	/*
2412 	 * Initial thread where the tid equal to the pid.
2413 	 */
2414 	p = pfind(tid);
2415 	if (p != NULL) {
2416 		if (SV_PROC_ABI(p) != SV_ABI_LINUX ||
2417 		    (pid != -1 && tid != pid)) {
2418 			/*
2419 			 * p is not a Linuxulator process.
2420 			 */
2421 			PROC_UNLOCK(p);
2422 			return (NULL);
2423 		}
2424 		FOREACH_THREAD_IN_PROC(p, tdt) {
2425 			em = em_find(tdt);
2426 			if (tid == em->em_tid)
2427 				return (tdt);
2428 		}
2429 		PROC_UNLOCK(p);
2430 	}
2431 	return (NULL);
2432 }
2433 
2434 void
2435 linux_to_bsd_waitopts(int options, int *bsdopts)
2436 {
2437 
2438 	if (options & LINUX_WNOHANG)
2439 		*bsdopts |= WNOHANG;
2440 	if (options & LINUX_WUNTRACED)
2441 		*bsdopts |= WUNTRACED;
2442 	if (options & LINUX_WEXITED)
2443 		*bsdopts |= WEXITED;
2444 	if (options & LINUX_WCONTINUED)
2445 		*bsdopts |= WCONTINUED;
2446 	if (options & LINUX_WNOWAIT)
2447 		*bsdopts |= WNOWAIT;
2448 
2449 	if (options & __WCLONE)
2450 		*bsdopts |= WLINUXCLONE;
2451 }
2452 
2453 int
2454 linux_getrandom(struct thread *td, struct linux_getrandom_args *args)
2455 {
2456 	struct uio uio;
2457 	struct iovec iov;
2458 	int error;
2459 
2460 	if (args->flags & ~(LINUX_GRND_NONBLOCK|LINUX_GRND_RANDOM))
2461 		return (EINVAL);
2462 	if (args->count > INT_MAX)
2463 		args->count = INT_MAX;
2464 
2465 	iov.iov_base = args->buf;
2466 	iov.iov_len = args->count;
2467 
2468 	uio.uio_iov = &iov;
2469 	uio.uio_iovcnt = 1;
2470 	uio.uio_resid = iov.iov_len;
2471 	uio.uio_segflg = UIO_USERSPACE;
2472 	uio.uio_rw = UIO_READ;
2473 	uio.uio_td = td;
2474 
2475 	error = read_random_uio(&uio, args->flags & LINUX_GRND_NONBLOCK);
2476 	if (error == 0)
2477 		td->td_retval[0] = args->count - uio.uio_resid;
2478 	return (error);
2479 }
2480 
2481 int
2482 linux_mincore(struct thread *td, struct linux_mincore_args *args)
2483 {
2484 
2485 	/* Needs to be page-aligned */
2486 	if (args->start & PAGE_MASK)
2487 		return (EINVAL);
2488 	return (kern_mincore(td, args->start, args->len, args->vec));
2489 }
2490 
2491 #define	SYSLOG_TAG	"<6>"
2492 
2493 int
2494 linux_syslog(struct thread *td, struct linux_syslog_args *args)
2495 {
2496 	char buf[128], *src, *dst;
2497 	u_int seq;
2498 	int buflen, error;
2499 
2500 	if (args->type != LINUX_SYSLOG_ACTION_READ_ALL) {
2501 		linux_msg(td, "syslog unsupported type 0x%x", args->type);
2502 		return (EINVAL);
2503 	}
2504 
2505 	if (args->len < 6) {
2506 		td->td_retval[0] = 0;
2507 		return (0);
2508 	}
2509 
2510 	error = priv_check(td, PRIV_MSGBUF);
2511 	if (error)
2512 		return (error);
2513 
2514 	mtx_lock(&msgbuf_lock);
2515 	msgbuf_peekbytes(msgbufp, NULL, 0, &seq);
2516 	mtx_unlock(&msgbuf_lock);
2517 
2518 	dst = args->buf;
2519 	error = copyout(&SYSLOG_TAG, dst, sizeof(SYSLOG_TAG));
2520 	/* The -1 is to skip the trailing '\0'. */
2521 	dst += sizeof(SYSLOG_TAG) - 1;
2522 
2523 	while (error == 0) {
2524 		mtx_lock(&msgbuf_lock);
2525 		buflen = msgbuf_peekbytes(msgbufp, buf, sizeof(buf), &seq);
2526 		mtx_unlock(&msgbuf_lock);
2527 
2528 		if (buflen == 0)
2529 			break;
2530 
2531 		for (src = buf; src < buf + buflen && error == 0; src++) {
2532 			if (*src == '\0')
2533 				continue;
2534 
2535 			if (dst >= args->buf + args->len)
2536 				goto out;
2537 
2538 			error = copyout(src, dst, 1);
2539 			dst++;
2540 
2541 			if (*src == '\n' && *(src + 1) != '<' &&
2542 			    dst + sizeof(SYSLOG_TAG) < args->buf + args->len) {
2543 				error = copyout(&SYSLOG_TAG,
2544 				    dst, sizeof(SYSLOG_TAG));
2545 				dst += sizeof(SYSLOG_TAG) - 1;
2546 			}
2547 		}
2548 	}
2549 out:
2550 	td->td_retval[0] = dst - args->buf;
2551 	return (error);
2552 }
2553 
2554 int
2555 linux_getcpu(struct thread *td, struct linux_getcpu_args *args)
2556 {
2557 	int cpu, error, node;
2558 
2559 	cpu = td->td_oncpu; /* Make sure it doesn't change during copyout(9) */
2560 	error = 0;
2561 	node = cpuid_to_pcpu[cpu]->pc_domain;
2562 
2563 	if (args->cpu != NULL)
2564 		error = copyout(&cpu, args->cpu, sizeof(l_int));
2565 	if (args->node != NULL)
2566 		error = copyout(&node, args->node, sizeof(l_int));
2567 	return (error);
2568 }
2569 
2570 #if defined(__i386__) || defined(__amd64__)
2571 int
2572 linux_poll(struct thread *td, struct linux_poll_args *args)
2573 {
2574 	struct timespec ts, *tsp;
2575 
2576 	if (args->timeout != INFTIM) {
2577 		if (args->timeout < 0)
2578 			return (EINVAL);
2579 		ts.tv_sec = args->timeout / 1000;
2580 		ts.tv_nsec = (args->timeout % 1000) * 1000000;
2581 		tsp = &ts;
2582 	} else
2583 		tsp = NULL;
2584 
2585 	return (linux_common_ppoll(td, args->fds, args->nfds,
2586 	    tsp, NULL, 0));
2587 }
2588 #endif /* __i386__ || __amd64__ */
2589 
2590 int
2591 linux_seccomp(struct thread *td, struct linux_seccomp_args *args)
2592 {
2593 
2594 	switch (args->op) {
2595 	case LINUX_SECCOMP_GET_ACTION_AVAIL:
2596 		return (EOPNOTSUPP);
2597 	default:
2598 		/*
2599 		 * Ignore unknown operations, just like Linux kernel built
2600 		 * without CONFIG_SECCOMP.
2601 		 */
2602 		return (EINVAL);
2603 	}
2604 }
2605 
2606 /*
2607  * Custom version of exec_copyin_args(), to copy out argument and environment
2608  * strings from the old process address space into the temporary string buffer.
2609  * Based on freebsd32_exec_copyin_args.
2610  */
2611 static int
2612 linux_exec_copyin_args(struct image_args *args, const char *fname,
2613     l_uintptr_t *argv, l_uintptr_t *envv)
2614 {
2615 	char *argp, *envp;
2616 	l_uintptr_t *ptr, arg;
2617 	int error;
2618 
2619 	bzero(args, sizeof(*args));
2620 	if (argv == NULL)
2621 		return (EFAULT);
2622 
2623 	/*
2624 	 * Allocate demand-paged memory for the file name, argument, and
2625 	 * environment strings.
2626 	 */
2627 	error = exec_alloc_args(args);
2628 	if (error != 0)
2629 		return (error);
2630 
2631 	/*
2632 	 * Copy the file name.
2633 	 */
2634 	error = exec_args_add_fname(args, fname, UIO_USERSPACE);
2635 	if (error != 0)
2636 		goto err_exit;
2637 
2638 	/*
2639 	 * extract arguments first
2640 	 */
2641 	ptr = argv;
2642 	for (;;) {
2643 		error = copyin(ptr++, &arg, sizeof(arg));
2644 		if (error)
2645 			goto err_exit;
2646 		if (arg == 0)
2647 			break;
2648 		argp = PTRIN(arg);
2649 		error = exec_args_add_arg(args, argp, UIO_USERSPACE);
2650 		if (error != 0)
2651 			goto err_exit;
2652 	}
2653 
2654 	/*
2655 	 * This comment is from Linux do_execveat_common:
2656 	 * When argv is empty, add an empty string ("") as argv[0] to
2657 	 * ensure confused userspace programs that start processing
2658 	 * from argv[1] won't end up walking envp.
2659 	 */
2660 	if (args->argc == 0 &&
2661 	    (error = exec_args_add_arg(args, "", UIO_SYSSPACE) != 0))
2662 		goto err_exit;
2663 
2664 	/*
2665 	 * extract environment strings
2666 	 */
2667 	if (envv) {
2668 		ptr = envv;
2669 		for (;;) {
2670 			error = copyin(ptr++, &arg, sizeof(arg));
2671 			if (error)
2672 				goto err_exit;
2673 			if (arg == 0)
2674 				break;
2675 			envp = PTRIN(arg);
2676 			error = exec_args_add_env(args, envp, UIO_USERSPACE);
2677 			if (error != 0)
2678 				goto err_exit;
2679 		}
2680 	}
2681 
2682 	return (0);
2683 
2684 err_exit:
2685 	exec_free_args(args);
2686 	return (error);
2687 }
2688 
2689 int
2690 linux_execve(struct thread *td, struct linux_execve_args *args)
2691 {
2692 	struct image_args eargs;
2693 	int error;
2694 
2695 	LINUX_CTR(execve);
2696 
2697 	error = linux_exec_copyin_args(&eargs, args->path, args->argp,
2698 	    args->envp);
2699 	if (error == 0)
2700 		error = linux_common_execve(td, &eargs);
2701 	AUDIT_SYSCALL_EXIT(error == EJUSTRETURN ? 0 : error, td);
2702 	return (error);
2703 }
2704 
2705 static void
2706 linux_up_rtprio_if(struct thread *td1, struct rtprio *rtp)
2707 {
2708 	struct rtprio rtp2;
2709 
2710 	pri_to_rtp(td1, &rtp2);
2711 	if (rtp2.type <  rtp->type ||
2712 	    (rtp2.type == rtp->type &&
2713 	    rtp2.prio < rtp->prio)) {
2714 		rtp->type = rtp2.type;
2715 		rtp->prio = rtp2.prio;
2716 	}
2717 }
2718 
2719 #define	LINUX_PRIO_DIVIDER	RTP_PRIO_MAX / LINUX_IOPRIO_MAX
2720 
2721 static int
2722 linux_rtprio2ioprio(struct rtprio *rtp)
2723 {
2724 	int ioprio, prio;
2725 
2726 	switch (rtp->type) {
2727 	case RTP_PRIO_IDLE:
2728 		prio = RTP_PRIO_MIN;
2729 		ioprio = LINUX_IOPRIO_PRIO(LINUX_IOPRIO_CLASS_IDLE, prio);
2730 		break;
2731 	case RTP_PRIO_NORMAL:
2732 		prio = rtp->prio / LINUX_PRIO_DIVIDER;
2733 		ioprio = LINUX_IOPRIO_PRIO(LINUX_IOPRIO_CLASS_BE, prio);
2734 		break;
2735 	case RTP_PRIO_REALTIME:
2736 		prio = rtp->prio / LINUX_PRIO_DIVIDER;
2737 		ioprio = LINUX_IOPRIO_PRIO(LINUX_IOPRIO_CLASS_RT, prio);
2738 		break;
2739 	default:
2740 		prio = RTP_PRIO_MIN;
2741 		ioprio = LINUX_IOPRIO_PRIO(LINUX_IOPRIO_CLASS_NONE, prio);
2742 		break;
2743 	}
2744 	return (ioprio);
2745 }
2746 
2747 static int
2748 linux_ioprio2rtprio(int ioprio, struct rtprio *rtp)
2749 {
2750 
2751 	switch (LINUX_IOPRIO_PRIO_CLASS(ioprio)) {
2752 	case LINUX_IOPRIO_CLASS_IDLE:
2753 		rtp->prio = RTP_PRIO_MIN;
2754 		rtp->type = RTP_PRIO_IDLE;
2755 		break;
2756 	case LINUX_IOPRIO_CLASS_BE:
2757 		rtp->prio = LINUX_IOPRIO_PRIO_DATA(ioprio) * LINUX_PRIO_DIVIDER;
2758 		rtp->type = RTP_PRIO_NORMAL;
2759 		break;
2760 	case LINUX_IOPRIO_CLASS_RT:
2761 		rtp->prio = LINUX_IOPRIO_PRIO_DATA(ioprio) * LINUX_PRIO_DIVIDER;
2762 		rtp->type = RTP_PRIO_REALTIME;
2763 		break;
2764 	default:
2765 		return (EINVAL);
2766 	}
2767 	return (0);
2768 }
2769 #undef LINUX_PRIO_DIVIDER
2770 
2771 int
2772 linux_ioprio_get(struct thread *td, struct linux_ioprio_get_args *args)
2773 {
2774 	struct thread *td1;
2775 	struct rtprio rtp;
2776 	struct pgrp *pg;
2777 	struct proc *p;
2778 	int error, found;
2779 
2780 	p = NULL;
2781 	td1 = NULL;
2782 	error = 0;
2783 	found = 0;
2784 	rtp.type = RTP_PRIO_IDLE;
2785 	rtp.prio = RTP_PRIO_MAX;
2786 	switch (args->which) {
2787 	case LINUX_IOPRIO_WHO_PROCESS:
2788 		if (args->who == 0) {
2789 			td1 = td;
2790 			p = td1->td_proc;
2791 			PROC_LOCK(p);
2792 		} else if (args->who > PID_MAX) {
2793 			td1 = linux_tdfind(td, args->who, -1);
2794 			if (td1 != NULL)
2795 				p = td1->td_proc;
2796 		} else
2797 			p = pfind(args->who);
2798 		if (p == NULL)
2799 			return (ESRCH);
2800 		if ((error = p_cansee(td, p))) {
2801 			PROC_UNLOCK(p);
2802 			break;
2803 		}
2804 		if (td1 != NULL) {
2805 			pri_to_rtp(td1, &rtp);
2806 		} else {
2807 			FOREACH_THREAD_IN_PROC(p, td1) {
2808 				linux_up_rtprio_if(td1, &rtp);
2809 			}
2810 		}
2811 		found++;
2812 		PROC_UNLOCK(p);
2813 		break;
2814 	case LINUX_IOPRIO_WHO_PGRP:
2815 		sx_slock(&proctree_lock);
2816 		if (args->who == 0) {
2817 			pg = td->td_proc->p_pgrp;
2818 			PGRP_LOCK(pg);
2819 		} else {
2820 			pg = pgfind(args->who);
2821 			if (pg == NULL) {
2822 				sx_sunlock(&proctree_lock);
2823 				error = ESRCH;
2824 				break;
2825 			}
2826 		}
2827 		sx_sunlock(&proctree_lock);
2828 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
2829 			PROC_LOCK(p);
2830 			if (p->p_state == PRS_NORMAL &&
2831 			    p_cansee(td, p) == 0) {
2832 				FOREACH_THREAD_IN_PROC(p, td1) {
2833 					linux_up_rtprio_if(td1, &rtp);
2834 					found++;
2835 				}
2836 			}
2837 			PROC_UNLOCK(p);
2838 		}
2839 		PGRP_UNLOCK(pg);
2840 		break;
2841 	case LINUX_IOPRIO_WHO_USER:
2842 		if (args->who == 0)
2843 			args->who = td->td_ucred->cr_uid;
2844 		sx_slock(&allproc_lock);
2845 		FOREACH_PROC_IN_SYSTEM(p) {
2846 			PROC_LOCK(p);
2847 			if (p->p_state == PRS_NORMAL &&
2848 			    p->p_ucred->cr_uid == args->who &&
2849 			    p_cansee(td, p) == 0) {
2850 				FOREACH_THREAD_IN_PROC(p, td1) {
2851 					linux_up_rtprio_if(td1, &rtp);
2852 					found++;
2853 				}
2854 			}
2855 			PROC_UNLOCK(p);
2856 		}
2857 		sx_sunlock(&allproc_lock);
2858 		break;
2859 	default:
2860 		error = EINVAL;
2861 		break;
2862 	}
2863 	if (error == 0) {
2864 		if (found != 0)
2865 			td->td_retval[0] = linux_rtprio2ioprio(&rtp);
2866 		else
2867 			error = ESRCH;
2868 	}
2869 	return (error);
2870 }
2871 
2872 int
2873 linux_ioprio_set(struct thread *td, struct linux_ioprio_set_args *args)
2874 {
2875 	struct thread *td1;
2876 	struct rtprio rtp;
2877 	struct pgrp *pg;
2878 	struct proc *p;
2879 	int error;
2880 
2881 	if ((error = linux_ioprio2rtprio(args->ioprio, &rtp)) != 0)
2882 		return (error);
2883 	/* Attempts to set high priorities (REALTIME) require su privileges. */
2884 	if (RTP_PRIO_BASE(rtp.type) == RTP_PRIO_REALTIME &&
2885 	    (error = priv_check(td, PRIV_SCHED_RTPRIO)) != 0)
2886 		return (error);
2887 
2888 	p = NULL;
2889 	td1 = NULL;
2890 	switch (args->which) {
2891 	case LINUX_IOPRIO_WHO_PROCESS:
2892 		if (args->who == 0) {
2893 			td1 = td;
2894 			p = td1->td_proc;
2895 			PROC_LOCK(p);
2896 		} else if (args->who > PID_MAX) {
2897 			td1 = linux_tdfind(td, args->who, -1);
2898 			if (td1 != NULL)
2899 				p = td1->td_proc;
2900 		} else
2901 			p = pfind(args->who);
2902 		if (p == NULL)
2903 			return (ESRCH);
2904 		if ((error = p_cansched(td, p))) {
2905 			PROC_UNLOCK(p);
2906 			break;
2907 		}
2908 		if (td1 != NULL) {
2909 			error = rtp_to_pri(&rtp, td1);
2910 		} else {
2911 			FOREACH_THREAD_IN_PROC(p, td1) {
2912 				if ((error = rtp_to_pri(&rtp, td1)) != 0)
2913 					break;
2914 			}
2915 		}
2916 		PROC_UNLOCK(p);
2917 		break;
2918 	case LINUX_IOPRIO_WHO_PGRP:
2919 		sx_slock(&proctree_lock);
2920 		if (args->who == 0) {
2921 			pg = td->td_proc->p_pgrp;
2922 			PGRP_LOCK(pg);
2923 		} else {
2924 			pg = pgfind(args->who);
2925 			if (pg == NULL) {
2926 				sx_sunlock(&proctree_lock);
2927 				error = ESRCH;
2928 				break;
2929 			}
2930 		}
2931 		sx_sunlock(&proctree_lock);
2932 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
2933 			PROC_LOCK(p);
2934 			if (p->p_state == PRS_NORMAL &&
2935 			    p_cansched(td, p) == 0) {
2936 				FOREACH_THREAD_IN_PROC(p, td1) {
2937 					if ((error = rtp_to_pri(&rtp, td1)) != 0)
2938 						break;
2939 				}
2940 			}
2941 			PROC_UNLOCK(p);
2942 			if (error != 0)
2943 				break;
2944 		}
2945 		PGRP_UNLOCK(pg);
2946 		break;
2947 	case LINUX_IOPRIO_WHO_USER:
2948 		if (args->who == 0)
2949 			args->who = td->td_ucred->cr_uid;
2950 		sx_slock(&allproc_lock);
2951 		FOREACH_PROC_IN_SYSTEM(p) {
2952 			PROC_LOCK(p);
2953 			if (p->p_state == PRS_NORMAL &&
2954 			    p->p_ucred->cr_uid == args->who &&
2955 			    p_cansched(td, p) == 0) {
2956 				FOREACH_THREAD_IN_PROC(p, td1) {
2957 					if ((error = rtp_to_pri(&rtp, td1)) != 0)
2958 						break;
2959 				}
2960 			}
2961 			PROC_UNLOCK(p);
2962 			if (error != 0)
2963 				break;
2964 		}
2965 		sx_sunlock(&allproc_lock);
2966 		break;
2967 	default:
2968 		error = EINVAL;
2969 		break;
2970 	}
2971 	return (error);
2972 }
2973 
2974 /* The only flag is O_NONBLOCK */
2975 #define B2L_MQ_FLAGS(bflags)	((bflags) != 0 ? LINUX_O_NONBLOCK : 0)
2976 #define L2B_MQ_FLAGS(lflags)	((lflags) != 0 ? O_NONBLOCK : 0)
2977 
2978 int
2979 linux_mq_open(struct thread *td, struct linux_mq_open_args *args)
2980 {
2981 	struct mq_attr attr;
2982 	int error, flags;
2983 
2984 	flags = linux_common_openflags(args->oflag);
2985 	if ((flags & O_ACCMODE) == O_ACCMODE || (flags & O_EXEC) != 0)
2986 		return (EINVAL);
2987 	flags = FFLAGS(flags);
2988 	if ((flags & O_CREAT) != 0 && args->attr != NULL) {
2989 		error = copyin(args->attr, &attr, sizeof(attr));
2990 		if (error != 0)
2991 			return (error);
2992 		attr.mq_flags = L2B_MQ_FLAGS(attr.mq_flags);
2993 	}
2994 
2995 	return (kern_kmq_open(td, args->name, flags, args->mode,
2996 	    args->attr != NULL ? &attr : NULL));
2997 }
2998 
2999 int
3000 linux_mq_unlink(struct thread *td, struct linux_mq_unlink_args *args)
3001 {
3002 	struct kmq_unlink_args bsd_args = {
3003 		.path = PTRIN(args->name)
3004 	};
3005 
3006 	return (sys_kmq_unlink(td, &bsd_args));
3007 }
3008 
3009 int
3010 linux_mq_timedsend(struct thread *td, struct linux_mq_timedsend_args *args)
3011 {
3012 	struct timespec ts, *abs_timeout;
3013 	int error;
3014 
3015 	if (args->abs_timeout == NULL)
3016 		abs_timeout = NULL;
3017 	else {
3018 		error = linux_get_timespec(&ts, args->abs_timeout);
3019 		if (error != 0)
3020 			return (error);
3021 		abs_timeout = &ts;
3022 	}
3023 
3024 	return (kern_kmq_timedsend(td, args->mqd, PTRIN(args->msg_ptr),
3025 		args->msg_len, args->msg_prio, abs_timeout));
3026 }
3027 
3028 int
3029 linux_mq_timedreceive(struct thread *td, struct linux_mq_timedreceive_args *args)
3030 {
3031 	struct timespec ts, *abs_timeout;
3032 	int error;
3033 
3034 	if (args->abs_timeout == NULL)
3035 		abs_timeout = NULL;
3036 	else {
3037 		error = linux_get_timespec(&ts, args->abs_timeout);
3038 		if (error != 0)
3039 			return (error);
3040 		abs_timeout = &ts;
3041 	}
3042 
3043 	return (kern_kmq_timedreceive(td, args->mqd, PTRIN(args->msg_ptr),
3044 		args->msg_len, args->msg_prio, abs_timeout));
3045 }
3046 
3047 int
3048 linux_mq_notify(struct thread *td, struct linux_mq_notify_args *args)
3049 {
3050 	struct sigevent ev, *evp;
3051 	struct l_sigevent l_ev;
3052 	int error;
3053 
3054 	if (args->sevp == NULL)
3055 		evp = NULL;
3056 	else {
3057 		error = copyin(args->sevp, &l_ev, sizeof(l_ev));
3058 		if (error != 0)
3059 			return (error);
3060 		error = linux_convert_l_sigevent(&l_ev, &ev);
3061 		if (error != 0)
3062 			return (error);
3063 		evp = &ev;
3064 	}
3065 
3066 	return (kern_kmq_notify(td, args->mqd, evp));
3067 }
3068 
3069 int
3070 linux_mq_getsetattr(struct thread *td, struct linux_mq_getsetattr_args *args)
3071 {
3072 	struct mq_attr attr, oattr;
3073 	int error;
3074 
3075 	if (args->attr != NULL) {
3076 		error = copyin(args->attr, &attr, sizeof(attr));
3077 		if (error != 0)
3078 			return (error);
3079 		attr.mq_flags = L2B_MQ_FLAGS(attr.mq_flags);
3080 	}
3081 
3082 	error = kern_kmq_setattr(td, args->mqd, args->attr != NULL ? &attr : NULL,
3083 	    &oattr);
3084 	if (error == 0 && args->oattr != NULL) {
3085 		oattr.mq_flags = B2L_MQ_FLAGS(oattr.mq_flags);
3086 		bzero(oattr.__reserved, sizeof(oattr.__reserved));
3087 		error = copyout(&oattr, args->oattr, sizeof(oattr));
3088 	}
3089 
3090 	return (error);
3091 }
3092 
3093 int
3094 linux_kcmp(struct thread *td, struct linux_kcmp_args *args)
3095 {
3096 	int type;
3097 
3098 	switch (args->type) {
3099 	case LINUX_KCMP_FILE:
3100 		type = KCMP_FILE;
3101 		break;
3102 	case LINUX_KCMP_FILES:
3103 		type = KCMP_FILES;
3104 		break;
3105 	case LINUX_KCMP_SIGHAND:
3106 		type = KCMP_SIGHAND;
3107 		break;
3108 	case LINUX_KCMP_VM:
3109 		type = KCMP_VM;
3110 		break;
3111 	default:
3112 		return (EINVAL);
3113 	}
3114 
3115 	return (kern_kcmp(td, args->pid1, args->pid2, type, args->idx1,
3116 	    args->idx));
3117 }
3118 
3119 int
3120 linux_membarrier(struct thread *td, struct linux_membarrier_args *args)
3121 {
3122 	static const struct {
3123 		int linux_cmd;
3124 		int freebsd_cmd;
3125 	} cmds[] = {
3126 		{ LINUX_MEMBARRIER_CMD_QUERY,
3127 		    MEMBARRIER_CMD_QUERY },
3128 		{ LINUX_MEMBARRIER_CMD_GLOBAL,
3129 		    MEMBARRIER_CMD_GLOBAL },
3130 		{ LINUX_MEMBARRIER_CMD_GLOBAL_EXPEDITED,
3131 		    MEMBARRIER_CMD_GLOBAL_EXPEDITED },
3132 		{ LINUX_MEMBARRIER_CMD_REGISTER_GLOBAL_EXPEDITED,
3133 		    MEMBARRIER_CMD_REGISTER_GLOBAL_EXPEDITED },
3134 		{ LINUX_MEMBARRIER_CMD_PRIVATE_EXPEDITED,
3135 		    MEMBARRIER_CMD_PRIVATE_EXPEDITED },
3136 		{ LINUX_MEMBARRIER_CMD_REGISTER_PRIVATE_EXPEDITED,
3137 		    MEMBARRIER_CMD_REGISTER_PRIVATE_EXPEDITED },
3138 		{ LINUX_MEMBARRIER_CMD_PRIVATE_EXPEDITED_SYNC_CORE,
3139 		    MEMBARRIER_CMD_PRIVATE_EXPEDITED_SYNC_CORE },
3140 		{ LINUX_MEMBARRIER_CMD_REGISTER_PRIVATE_EXPEDITED_SYNC_CORE,
3141 		    MEMBARRIER_CMD_REGISTER_PRIVATE_EXPEDITED_SYNC_CORE },
3142 		{ LINUX_MEMBARRIER_CMD_PRIVATE_EXPEDITED_RSEQ,
3143 		    MEMBARRIER_CMD_PRIVATE_EXPEDITED_RSEQ },
3144 		{ LINUX_MEMBARRIER_CMD_REGISTER_PRIVATE_EXPEDITED_RSEQ,
3145 		    MEMBARRIER_CMD_REGISTER_PRIVATE_EXPEDITED_RSEQ },
3146 		{ LINUX_MEMBARRIER_CMD_GET_REGISTRATIONS,
3147 		    MEMBARRIER_CMD_GET_REGISTRATIONS },
3148 	};
3149 	int cmd, error, flags, i, mask;
3150 
3151 	cmd = -1;
3152 	for (i = 0; i < nitems(cmds); i++) {
3153 		if (args->cmd == cmds[i].linux_cmd) {
3154 			cmd = cmds[i].freebsd_cmd;
3155 			break;
3156 		}
3157 	}
3158 
3159 	if (cmd == -1 || (args->flags & ~LINUX_MEMBARRIER_CMD_FLAG_CPU) != 0)
3160 		return (EINVAL);
3161 
3162 	flags = 0;
3163 	if ((args->flags & LINUX_MEMBARRIER_CMD_FLAG_CPU) != 0)
3164 		flags |= MEMBARRIER_CMD_FLAG_CPU;
3165 
3166 	error = kern_membarrier(td, cmd, flags, args->cpu_id);
3167 	if (error != 0)
3168 		return (error);
3169 
3170 	if (args->cmd == LINUX_MEMBARRIER_CMD_QUERY ||
3171 	    args->cmd == LINUX_MEMBARRIER_CMD_GET_REGISTRATIONS) {
3172 		mask = td->td_retval[0];
3173 		td->td_retval[0] = 0;
3174 		for (i = 0; i < nitems(cmds); i++)
3175 			if ((mask & cmds[i].freebsd_cmd) != 0)
3176 				td->td_retval[0] |= cmds[i].linux_cmd;
3177 	}
3178 
3179 	return (0);
3180 }
3181 
3182 MODULE_DEPEND(linux, mqueuefs, 1, 1, 1);
3183