xref: /freebsd/sys/compat/linux/linux_misc.c (revision 3a1bf59d195ced99c0f69774969d7090d21f6097)
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_pkey_mprotect(struct thread *td, struct linux_pkey_mprotect_args *uap)
364 {
365 
366 	return (linux_pkey_mprotect_common(td, uap->start, uap->len,
367 	    uap->prot, uap->pkey));
368 }
369 
370 int
371 linux_pkey_alloc(struct thread *td, struct linux_pkey_alloc_args *uap)
372 {
373 
374 	return (linux_pkey_alloc_common(td, uap->flags, uap->init_val));
375 }
376 
377 int
378 linux_pkey_free(struct thread *td, struct linux_pkey_free_args *uap)
379 {
380 
381 	return (linux_pkey_free_common(td, uap->pkey));
382 }
383 
384 int
385 linux_madvise(struct thread *td, struct linux_madvise_args *uap)
386 {
387 
388 	return (linux_madvise_common(td, PTROUT(uap->addr), uap->len,
389 	    uap->behav));
390 }
391 
392 int
393 linux_mmap2(struct thread *td, struct linux_mmap2_args *uap)
394 {
395 #if defined(LINUX_ARCHWANT_MMAP2PGOFF)
396 	/*
397 	 * For architectures with sizeof (off_t) < sizeof (loff_t) mmap is
398 	 * implemented with mmap2 syscall and the offset is represented in
399 	 * multiples of page size.
400 	 */
401 	return (linux_mmap_common(td, PTROUT(uap->addr), uap->len, uap->prot,
402 	    uap->flags, uap->fd, (uint64_t)(uint32_t)uap->pgoff * PAGE_SIZE));
403 #else
404 	return (linux_mmap_common(td, PTROUT(uap->addr), uap->len, uap->prot,
405 	    uap->flags, uap->fd, uap->pgoff));
406 #endif
407 }
408 
409 #ifdef LINUX_LEGACY_SYSCALLS
410 int
411 linux_time(struct thread *td, struct linux_time_args *args)
412 {
413 	struct timeval tv;
414 	l_time_t tm;
415 	int error;
416 
417 	microtime(&tv);
418 	tm = tv.tv_sec;
419 	if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
420 		return (error);
421 	td->td_retval[0] = tm;
422 	return (0);
423 }
424 #endif
425 
426 struct l_times_argv {
427 	l_clock_t	tms_utime;
428 	l_clock_t	tms_stime;
429 	l_clock_t	tms_cutime;
430 	l_clock_t	tms_cstime;
431 };
432 
433 /*
434  * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
435  * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
436  * auxiliary vector entry.
437  */
438 #define	CLK_TCK		100
439 
440 #define	CONVOTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
441 #define	CONVNTCK(r)	(r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
442 
443 #define	CONVTCK(r)	(linux_kernver(td) >= LINUX_KERNVER(2,4,0) ?	\
444 			    CONVNTCK(r) : CONVOTCK(r))
445 
446 int
447 linux_times(struct thread *td, struct linux_times_args *args)
448 {
449 	struct timeval tv, utime, stime, cutime, cstime;
450 	struct l_times_argv tms;
451 	struct proc *p;
452 	int error;
453 
454 	if (args->buf != NULL) {
455 		p = td->td_proc;
456 		PROC_LOCK(p);
457 		PROC_STATLOCK(p);
458 		calcru(p, &utime, &stime);
459 		PROC_STATUNLOCK(p);
460 		calccru(p, &cutime, &cstime);
461 		PROC_UNLOCK(p);
462 
463 		tms.tms_utime = CONVTCK(utime);
464 		tms.tms_stime = CONVTCK(stime);
465 
466 		tms.tms_cutime = CONVTCK(cutime);
467 		tms.tms_cstime = CONVTCK(cstime);
468 
469 		if ((error = copyout(&tms, args->buf, sizeof(tms))))
470 			return (error);
471 	}
472 
473 	microuptime(&tv);
474 	td->td_retval[0] = (int)CONVTCK(tv);
475 	return (0);
476 }
477 
478 int
479 linux_newuname(struct thread *td, struct linux_newuname_args *args)
480 {
481 	struct l_new_utsname utsname;
482 	char osname[LINUX_MAX_UTSNAME];
483 	char osrelease[LINUX_MAX_UTSNAME];
484 	char *p;
485 
486 	linux_get_osname(td, osname);
487 	linux_get_osrelease(td, osrelease);
488 
489 	bzero(&utsname, sizeof(utsname));
490 	strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
491 	getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
492 	getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
493 	strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
494 	strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
495 	for (p = utsname.version; *p != '\0'; ++p)
496 		if (*p == '\n') {
497 			*p = '\0';
498 			break;
499 		}
500 #if defined(__amd64__)
501 	/*
502 	 * On amd64, Linux uname(2) needs to return "x86_64"
503 	 * for both 64-bit and 32-bit applications.  On 32-bit,
504 	 * the string returned by getauxval(AT_PLATFORM) needs
505 	 * to remain "i686", though.
506 	 */
507 #if defined(COMPAT_LINUX32)
508 	if (linux32_emulate_i386)
509 		strlcpy(utsname.machine, "i686", LINUX_MAX_UTSNAME);
510 	else
511 #endif
512 	strlcpy(utsname.machine, "x86_64", LINUX_MAX_UTSNAME);
513 #elif defined(__aarch64__)
514 	strlcpy(utsname.machine, "aarch64", LINUX_MAX_UTSNAME);
515 #elif defined(__i386__)
516 	strlcpy(utsname.machine, "i686", LINUX_MAX_UTSNAME);
517 #endif
518 
519 	return (copyout(&utsname, args->buf, sizeof(utsname)));
520 }
521 
522 struct l_utimbuf {
523 	l_time_t l_actime;
524 	l_time_t l_modtime;
525 };
526 
527 #ifdef LINUX_LEGACY_SYSCALLS
528 int
529 linux_utime(struct thread *td, struct linux_utime_args *args)
530 {
531 	struct timeval tv[2], *tvp;
532 	struct l_utimbuf lut;
533 	int error;
534 
535 	if (args->times) {
536 		if ((error = copyin(args->times, &lut, sizeof lut)) != 0)
537 			return (error);
538 		tv[0].tv_sec = lut.l_actime;
539 		tv[0].tv_usec = 0;
540 		tv[1].tv_sec = lut.l_modtime;
541 		tv[1].tv_usec = 0;
542 		tvp = tv;
543 	} else
544 		tvp = NULL;
545 
546 	return (kern_utimesat(td, AT_FDCWD, args->fname, UIO_USERSPACE,
547 	    tvp, UIO_SYSSPACE));
548 }
549 #endif
550 
551 #ifdef LINUX_LEGACY_SYSCALLS
552 int
553 linux_utimes(struct thread *td, struct linux_utimes_args *args)
554 {
555 	l_timeval ltv[2];
556 	struct timeval tv[2], *tvp = NULL;
557 	int error;
558 
559 	if (args->tptr != NULL) {
560 		if ((error = copyin(args->tptr, ltv, sizeof ltv)) != 0)
561 			return (error);
562 		tv[0].tv_sec = ltv[0].tv_sec;
563 		tv[0].tv_usec = ltv[0].tv_usec;
564 		tv[1].tv_sec = ltv[1].tv_sec;
565 		tv[1].tv_usec = ltv[1].tv_usec;
566 		tvp = tv;
567 	}
568 
569 	return (kern_utimesat(td, AT_FDCWD, args->fname, UIO_USERSPACE,
570 	    tvp, UIO_SYSSPACE));
571 }
572 #endif
573 
574 static int
575 linux_utimensat_lts_to_ts(struct l_timespec *l_times, struct timespec *times)
576 {
577 
578 	if (l_times->tv_nsec != LINUX_UTIME_OMIT &&
579 	    l_times->tv_nsec != LINUX_UTIME_NOW &&
580 	    (l_times->tv_nsec < 0 || l_times->tv_nsec > 999999999))
581 		return (EINVAL);
582 
583 	times->tv_sec = l_times->tv_sec;
584 	switch (l_times->tv_nsec)
585 	{
586 	case LINUX_UTIME_OMIT:
587 		times->tv_nsec = UTIME_OMIT;
588 		break;
589 	case LINUX_UTIME_NOW:
590 		times->tv_nsec = UTIME_NOW;
591 		break;
592 	default:
593 		times->tv_nsec = l_times->tv_nsec;
594 	}
595 
596 	return (0);
597 }
598 
599 static int
600 linux_common_utimensat(struct thread *td, int ldfd, const char *pathname,
601     struct timespec *timesp, int lflags)
602 {
603 	int dfd, flags = 0;
604 
605 	dfd = (ldfd == LINUX_AT_FDCWD) ? AT_FDCWD : ldfd;
606 
607 	if (lflags & ~(LINUX_AT_SYMLINK_NOFOLLOW | LINUX_AT_EMPTY_PATH))
608 		return (EINVAL);
609 
610 	if (timesp != NULL) {
611 		/* This breaks POSIX, but is what the Linux kernel does
612 		 * _on purpose_ (documented in the man page for utimensat(2)),
613 		 * so we must follow that behaviour. */
614 		if (timesp[0].tv_nsec == UTIME_OMIT &&
615 		    timesp[1].tv_nsec == UTIME_OMIT)
616 			return (0);
617 	}
618 
619 	if (lflags & LINUX_AT_SYMLINK_NOFOLLOW)
620 		flags |= AT_SYMLINK_NOFOLLOW;
621 	if (lflags & LINUX_AT_EMPTY_PATH)
622 		flags |= AT_EMPTY_PATH;
623 
624 	if (pathname != NULL)
625 		return (kern_utimensat(td, dfd, pathname,
626 		    UIO_USERSPACE, timesp, UIO_SYSSPACE, flags));
627 
628 	if (lflags != 0)
629 		return (EINVAL);
630 
631 	return (kern_futimens(td, dfd, timesp, UIO_SYSSPACE));
632 }
633 
634 int
635 linux_utimensat(struct thread *td, struct linux_utimensat_args *args)
636 {
637 	struct l_timespec l_times[2];
638 	struct timespec times[2], *timesp;
639 	int error;
640 
641 	if (args->times != NULL) {
642 		error = copyin(args->times, l_times, sizeof(l_times));
643 		if (error != 0)
644 			return (error);
645 
646 		error = linux_utimensat_lts_to_ts(&l_times[0], &times[0]);
647 		if (error != 0)
648 			return (error);
649 		error = linux_utimensat_lts_to_ts(&l_times[1], &times[1]);
650 		if (error != 0)
651 			return (error);
652 		timesp = times;
653 	} else
654 		timesp = NULL;
655 
656 	return (linux_common_utimensat(td, args->dfd, args->pathname,
657 	    timesp, args->flags));
658 }
659 
660 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
661 static int
662 linux_utimensat_lts64_to_ts(struct l_timespec64 *l_times, struct timespec *times)
663 {
664 
665 	/* Zero out the padding in compat mode. */
666 	l_times->tv_nsec &= 0xFFFFFFFFUL;
667 
668 	if (l_times->tv_nsec != LINUX_UTIME_OMIT &&
669 	    l_times->tv_nsec != LINUX_UTIME_NOW &&
670 	    (l_times->tv_nsec < 0 || l_times->tv_nsec > 999999999))
671 		return (EINVAL);
672 
673 	times->tv_sec = l_times->tv_sec;
674 	switch (l_times->tv_nsec)
675 	{
676 	case LINUX_UTIME_OMIT:
677 		times->tv_nsec = UTIME_OMIT;
678 		break;
679 	case LINUX_UTIME_NOW:
680 		times->tv_nsec = UTIME_NOW;
681 		break;
682 	default:
683 		times->tv_nsec = l_times->tv_nsec;
684 	}
685 
686 	return (0);
687 }
688 
689 int
690 linux_utimensat_time64(struct thread *td, struct linux_utimensat_time64_args *args)
691 {
692 	struct l_timespec64 l_times[2];
693 	struct timespec times[2], *timesp;
694 	int error;
695 
696 	if (args->times64 != NULL) {
697 		error = copyin(args->times64, l_times, sizeof(l_times));
698 		if (error != 0)
699 			return (error);
700 
701 		error = linux_utimensat_lts64_to_ts(&l_times[0], &times[0]);
702 		if (error != 0)
703 			return (error);
704 		error = linux_utimensat_lts64_to_ts(&l_times[1], &times[1]);
705 		if (error != 0)
706 			return (error);
707 		timesp = times;
708 	} else
709 		timesp = NULL;
710 
711 	return (linux_common_utimensat(td, args->dfd, args->pathname,
712 	    timesp, args->flags));
713 }
714 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
715 
716 #ifdef LINUX_LEGACY_SYSCALLS
717 int
718 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
719 {
720 	l_timeval ltv[2];
721 	struct timeval tv[2], *tvp = NULL;
722 	int error, dfd;
723 
724 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
725 
726 	if (args->utimes != NULL) {
727 		if ((error = copyin(args->utimes, ltv, sizeof ltv)) != 0)
728 			return (error);
729 		tv[0].tv_sec = ltv[0].tv_sec;
730 		tv[0].tv_usec = ltv[0].tv_usec;
731 		tv[1].tv_sec = ltv[1].tv_sec;
732 		tv[1].tv_usec = ltv[1].tv_usec;
733 		tvp = tv;
734 	}
735 
736 	return (kern_utimesat(td, dfd, args->filename, UIO_USERSPACE,
737 	    tvp, UIO_SYSSPACE));
738 }
739 #endif
740 
741 static int
742 linux_common_wait(struct thread *td, idtype_t idtype, int id, int *statusp,
743     int options, void *rup, l_siginfo_t *infop)
744 {
745 	l_siginfo_t lsi;
746 	siginfo_t siginfo;
747 	struct __wrusage wru;
748 	int error, status, tmpstat, sig;
749 
750 	error = kern_wait6(td, idtype, id, &status, options,
751 	    rup != NULL ? &wru : NULL, &siginfo);
752 
753 	if (error == 0 && statusp) {
754 		tmpstat = status & 0xffff;
755 		if (WIFSIGNALED(tmpstat)) {
756 			tmpstat = (tmpstat & 0xffffff80) |
757 			    bsd_to_linux_signal(WTERMSIG(tmpstat));
758 		} else if (WIFSTOPPED(tmpstat)) {
759 			tmpstat = (tmpstat & 0xffff00ff) |
760 			    (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8);
761 #if defined(__aarch64__) || (defined(__amd64__) && !defined(COMPAT_LINUX32))
762 			if (WSTOPSIG(status) == SIGTRAP) {
763 				tmpstat = linux_ptrace_status(td,
764 				    siginfo.si_pid, tmpstat);
765 			}
766 #endif
767 		} else if (WIFCONTINUED(tmpstat)) {
768 			tmpstat = 0xffff;
769 		}
770 		error = copyout(&tmpstat, statusp, sizeof(int));
771 	}
772 	if (error == 0 && rup != NULL)
773 		error = linux_copyout_rusage(&wru.wru_self, rup);
774 	if (error == 0 && infop != NULL && td->td_retval[0] != 0) {
775 		sig = bsd_to_linux_signal(siginfo.si_signo);
776 		memset(&lsi, 0, sizeof(lsi));
777 		siginfo_to_lsiginfo(&siginfo, &lsi, sig);
778 		error = copyout(&lsi, infop, sizeof(lsi));
779 	}
780 
781 	return (error);
782 }
783 
784 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
785 int
786 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
787 {
788 	struct linux_wait4_args wait4_args = {
789 		.pid = args->pid,
790 		.status = args->status,
791 		.options = args->options,
792 		.rusage = NULL,
793 	};
794 
795 	return (linux_wait4(td, &wait4_args));
796 }
797 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
798 
799 int
800 linux_wait4(struct thread *td, struct linux_wait4_args *args)
801 {
802 	struct proc *p;
803 	int options, id, idtype;
804 
805 	if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG |
806 	    LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
807 		return (EINVAL);
808 
809 	/* -INT_MIN is not defined. */
810 	if (args->pid == INT_MIN)
811 		return (ESRCH);
812 
813 	options = 0;
814 	linux_to_bsd_waitopts(args->options, &options);
815 
816 	/*
817 	 * For backward compatibility we implicitly add flags WEXITED
818 	 * and WTRAPPED here.
819 	 */
820 	options |= WEXITED | WTRAPPED;
821 
822 	if (args->pid == WAIT_ANY) {
823 		idtype = P_ALL;
824 		id = 0;
825 	} else if (args->pid < 0) {
826 		idtype = P_PGID;
827 		id = (id_t)-args->pid;
828 	} else if (args->pid == 0) {
829 		idtype = P_PGID;
830 		p = td->td_proc;
831 		PROC_LOCK(p);
832 		id = p->p_pgid;
833 		PROC_UNLOCK(p);
834 	} else {
835 		idtype = P_PID;
836 		id = (id_t)args->pid;
837 	}
838 
839 	return (linux_common_wait(td, idtype, id, args->status, options,
840 	    args->rusage, NULL));
841 }
842 
843 int
844 linux_waitid(struct thread *td, struct linux_waitid_args *args)
845 {
846 	idtype_t idtype;
847 	int error, options;
848 	struct proc *p;
849 	pid_t id;
850 
851 	if (args->options & ~(LINUX_WNOHANG | LINUX_WNOWAIT | LINUX_WEXITED |
852 	    LINUX_WSTOPPED | LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
853 		return (EINVAL);
854 
855 	options = 0;
856 	linux_to_bsd_waitopts(args->options, &options);
857 
858 	id = args->id;
859 	switch (args->idtype) {
860 	case LINUX_P_ALL:
861 		idtype = P_ALL;
862 		break;
863 	case LINUX_P_PID:
864 		if (args->id <= 0)
865 			return (EINVAL);
866 		idtype = P_PID;
867 		break;
868 	case LINUX_P_PGID:
869 		if (linux_kernver(td) >= LINUX_KERNVER(5,4,0) && args->id == 0) {
870 			p = td->td_proc;
871 			PROC_LOCK(p);
872 			id = p->p_pgid;
873 			PROC_UNLOCK(p);
874 		} else if (args->id <= 0)
875 			return (EINVAL);
876 		idtype = P_PGID;
877 		break;
878 	case LINUX_P_PIDFD:
879 		LINUX_RATELIMIT_MSG("unsupported waitid P_PIDFD idtype");
880 		return (ENOSYS);
881 	default:
882 		return (EINVAL);
883 	}
884 
885 	error = linux_common_wait(td, idtype, id, NULL, options,
886 	    args->rusage, args->info);
887 	td->td_retval[0] = 0;
888 
889 	return (error);
890 }
891 
892 #ifdef LINUX_LEGACY_SYSCALLS
893 int
894 linux_mknod(struct thread *td, struct linux_mknod_args *args)
895 {
896 	int error;
897 
898 	switch (args->mode & S_IFMT) {
899 	case S_IFIFO:
900 	case S_IFSOCK:
901 		error = kern_mkfifoat(td, AT_FDCWD, args->path, UIO_USERSPACE,
902 		    args->mode);
903 		break;
904 
905 	case S_IFCHR:
906 	case S_IFBLK:
907 		error = kern_mknodat(td, AT_FDCWD, args->path, UIO_USERSPACE,
908 		    args->mode, linux_decode_dev(args->dev));
909 		break;
910 
911 	case S_IFDIR:
912 		error = EPERM;
913 		break;
914 
915 	case 0:
916 		args->mode |= S_IFREG;
917 		/* FALLTHROUGH */
918 	case S_IFREG:
919 		error = kern_openat(td, AT_FDCWD, args->path, UIO_USERSPACE,
920 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
921 		if (error == 0)
922 			kern_close(td, td->td_retval[0]);
923 		break;
924 
925 	default:
926 		error = EINVAL;
927 		break;
928 	}
929 	return (error);
930 }
931 #endif
932 
933 int
934 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
935 {
936 	int error, dfd;
937 
938 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
939 
940 	switch (args->mode & S_IFMT) {
941 	case S_IFIFO:
942 	case S_IFSOCK:
943 		error = kern_mkfifoat(td, dfd, args->filename, UIO_USERSPACE,
944 		    args->mode);
945 		break;
946 
947 	case S_IFCHR:
948 	case S_IFBLK:
949 		error = kern_mknodat(td, dfd, args->filename, UIO_USERSPACE,
950 		    args->mode, linux_decode_dev(args->dev));
951 		break;
952 
953 	case S_IFDIR:
954 		error = EPERM;
955 		break;
956 
957 	case 0:
958 		args->mode |= S_IFREG;
959 		/* FALLTHROUGH */
960 	case S_IFREG:
961 		error = kern_openat(td, dfd, args->filename, UIO_USERSPACE,
962 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
963 		if (error == 0)
964 			kern_close(td, td->td_retval[0]);
965 		break;
966 
967 	default:
968 		error = EINVAL;
969 		break;
970 	}
971 	return (error);
972 }
973 
974 /*
975  * UGH! This is just about the dumbest idea I've ever heard!!
976  */
977 int
978 linux_personality(struct thread *td, struct linux_personality_args *args)
979 {
980 	struct linux_pemuldata *pem;
981 	struct proc *p = td->td_proc;
982 	uint32_t old;
983 
984 	PROC_LOCK(p);
985 	pem = pem_find(p);
986 	old = pem->persona;
987 	if (args->per != 0xffffffff)
988 		pem->persona = args->per;
989 	PROC_UNLOCK(p);
990 
991 	td->td_retval[0] = old;
992 	return (0);
993 }
994 
995 struct l_itimerval {
996 	l_timeval it_interval;
997 	l_timeval it_value;
998 };
999 
1000 #define	B2L_ITIMERVAL(bip, lip)						\
1001 	(bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec;		\
1002 	(bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec;	\
1003 	(bip)->it_value.tv_sec = (lip)->it_value.tv_sec;		\
1004 	(bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1005 
1006 int
1007 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1008 {
1009 	int error;
1010 	struct l_itimerval ls;
1011 	struct itimerval aitv, oitv;
1012 
1013 	if (uap->itv == NULL) {
1014 		uap->itv = uap->oitv;
1015 		return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1016 	}
1017 
1018 	error = copyin(uap->itv, &ls, sizeof(ls));
1019 	if (error != 0)
1020 		return (error);
1021 	B2L_ITIMERVAL(&aitv, &ls);
1022 	error = kern_setitimer(td, uap->which, &aitv, &oitv);
1023 	if (error != 0 || uap->oitv == NULL)
1024 		return (error);
1025 	B2L_ITIMERVAL(&ls, &oitv);
1026 
1027 	return (copyout(&ls, uap->oitv, sizeof(ls)));
1028 }
1029 
1030 int
1031 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1032 {
1033 	int error;
1034 	struct l_itimerval ls;
1035 	struct itimerval aitv;
1036 
1037 	error = kern_getitimer(td, uap->which, &aitv);
1038 	if (error != 0)
1039 		return (error);
1040 	B2L_ITIMERVAL(&ls, &aitv);
1041 	return (copyout(&ls, uap->itv, sizeof(ls)));
1042 }
1043 
1044 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1045 int
1046 linux_nice(struct thread *td, struct linux_nice_args *args)
1047 {
1048 
1049 	return (kern_setpriority(td, PRIO_PROCESS, 0, args->inc));
1050 }
1051 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1052 
1053 int
1054 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1055 {
1056 	const int ngrp = args->gidsetsize;
1057 	struct ucred *newcred, *oldcred;
1058 	l_gid_t *linux_gidset;
1059 	int error;
1060 	struct proc *p;
1061 
1062 	if (ngrp < 0 || ngrp > ngroups_max)
1063 		return (EINVAL);
1064 	linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK);
1065 	error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1066 	if (error)
1067 		goto out;
1068 
1069 	newcred = crget();
1070 	crextend(newcred, ngrp);
1071 	p = td->td_proc;
1072 	PROC_LOCK(p);
1073 	oldcred = crcopysafe(p, newcred);
1074 
1075 	if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS)) != 0) {
1076 		PROC_UNLOCK(p);
1077 		crfree(newcred);
1078 		goto out;
1079 	}
1080 
1081 	newcred->cr_ngroups = ngrp;
1082 	for (int i = 0; i < ngrp; i++)
1083 		newcred->cr_groups[i] = linux_gidset[i];
1084 	newcred->cr_flags |= CRED_FLAG_GROUPSET;
1085 
1086 	setsugid(p);
1087 	proc_set_cred(p, newcred);
1088 	PROC_UNLOCK(p);
1089 	crfree(oldcred);
1090 	error = 0;
1091 out:
1092 	free(linux_gidset, M_LINUX);
1093 	return (error);
1094 }
1095 
1096 int
1097 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1098 {
1099 	const struct ucred *const cred = td->td_ucred;
1100 	l_gid_t *linux_gidset;
1101 	int ngrp, error;
1102 
1103 	ngrp = args->gidsetsize;
1104 
1105 	if (ngrp == 0) {
1106 		td->td_retval[0] = cred->cr_ngroups;
1107 		return (0);
1108 	}
1109 	if (ngrp < cred->cr_ngroups)
1110 		return (EINVAL);
1111 
1112 	ngrp = cred->cr_ngroups;
1113 
1114 	linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK);
1115 	for (int i = 0; i < ngrp; ++i)
1116 		linux_gidset[i] = cred->cr_groups[i];
1117 
1118 	error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1119 	free(linux_gidset, M_LINUX);
1120 
1121 	if (error != 0)
1122 		return (error);
1123 
1124 	td->td_retval[0] = ngrp;
1125 	return (0);
1126 }
1127 
1128 static bool
1129 linux_get_dummy_limit(struct thread *td, l_uint resource, struct rlimit *rlim)
1130 {
1131 	ssize_t size;
1132 	int res, error;
1133 
1134 	if (linux_dummy_rlimits == 0)
1135 		return (false);
1136 
1137 	switch (resource) {
1138 	case LINUX_RLIMIT_LOCKS:
1139 	case LINUX_RLIMIT_RTTIME:
1140 		rlim->rlim_cur = LINUX_RLIM_INFINITY;
1141 		rlim->rlim_max = LINUX_RLIM_INFINITY;
1142 		return (true);
1143 	case LINUX_RLIMIT_NICE:
1144 	case LINUX_RLIMIT_RTPRIO:
1145 		rlim->rlim_cur = 0;
1146 		rlim->rlim_max = 0;
1147 		return (true);
1148 	case LINUX_RLIMIT_SIGPENDING:
1149 		error = kernel_sysctlbyname(td,
1150 		    "kern.sigqueue.max_pending_per_proc",
1151 		    &res, &size, 0, 0, 0, 0);
1152 		if (error != 0)
1153 			return (false);
1154 		rlim->rlim_cur = res;
1155 		rlim->rlim_max = res;
1156 		return (true);
1157 	case LINUX_RLIMIT_MSGQUEUE:
1158 		error = kernel_sysctlbyname(td,
1159 		    "kern.ipc.msgmnb", &res, &size, 0, 0, 0, 0);
1160 		if (error != 0)
1161 			return (false);
1162 		rlim->rlim_cur = res;
1163 		rlim->rlim_max = res;
1164 		return (true);
1165 	default:
1166 		return (false);
1167 	}
1168 }
1169 
1170 int
1171 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1172 {
1173 	struct rlimit bsd_rlim;
1174 	struct l_rlimit rlim;
1175 	u_int which;
1176 	int error;
1177 
1178 	if (args->resource >= LINUX_RLIM_NLIMITS)
1179 		return (EINVAL);
1180 
1181 	which = linux_to_bsd_resource[args->resource];
1182 	if (which == -1)
1183 		return (EINVAL);
1184 
1185 	error = copyin(args->rlim, &rlim, sizeof(rlim));
1186 	if (error)
1187 		return (error);
1188 
1189 	bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1190 	bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1191 	return (kern_setrlimit(td, which, &bsd_rlim));
1192 }
1193 
1194 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1195 int
1196 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1197 {
1198 	struct l_rlimit rlim;
1199 	struct rlimit bsd_rlim;
1200 	u_int which;
1201 
1202 	if (linux_get_dummy_limit(td, args->resource, &bsd_rlim)) {
1203 		rlim.rlim_cur = bsd_rlim.rlim_cur;
1204 		rlim.rlim_max = bsd_rlim.rlim_max;
1205 		return (copyout(&rlim, args->rlim, sizeof(rlim)));
1206 	}
1207 
1208 	if (args->resource >= LINUX_RLIM_NLIMITS)
1209 		return (EINVAL);
1210 
1211 	which = linux_to_bsd_resource[args->resource];
1212 	if (which == -1)
1213 		return (EINVAL);
1214 
1215 	lim_rlimit(td, which, &bsd_rlim);
1216 
1217 #ifdef COMPAT_LINUX32
1218 	rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1219 	if (rlim.rlim_cur == UINT_MAX)
1220 		rlim.rlim_cur = INT_MAX;
1221 	rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1222 	if (rlim.rlim_max == UINT_MAX)
1223 		rlim.rlim_max = INT_MAX;
1224 #else
1225 	rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1226 	if (rlim.rlim_cur == ULONG_MAX)
1227 		rlim.rlim_cur = LONG_MAX;
1228 	rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1229 	if (rlim.rlim_max == ULONG_MAX)
1230 		rlim.rlim_max = LONG_MAX;
1231 #endif
1232 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1233 }
1234 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1235 
1236 int
1237 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1238 {
1239 	struct l_rlimit rlim;
1240 	struct rlimit bsd_rlim;
1241 	u_int which;
1242 
1243 	if (linux_get_dummy_limit(td, args->resource, &bsd_rlim)) {
1244 		rlim.rlim_cur = bsd_rlim.rlim_cur;
1245 		rlim.rlim_max = bsd_rlim.rlim_max;
1246 		return (copyout(&rlim, args->rlim, sizeof(rlim)));
1247 	}
1248 
1249 	if (args->resource >= LINUX_RLIM_NLIMITS)
1250 		return (EINVAL);
1251 
1252 	which = linux_to_bsd_resource[args->resource];
1253 	if (which == -1)
1254 		return (EINVAL);
1255 
1256 	lim_rlimit(td, which, &bsd_rlim);
1257 
1258 	rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1259 	rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1260 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1261 }
1262 
1263 int
1264 linux_sched_setscheduler(struct thread *td,
1265     struct linux_sched_setscheduler_args *args)
1266 {
1267 	struct sched_param sched_param;
1268 	struct thread *tdt;
1269 	int error, policy;
1270 
1271 	switch (args->policy) {
1272 	case LINUX_SCHED_OTHER:
1273 		policy = SCHED_OTHER;
1274 		break;
1275 	case LINUX_SCHED_FIFO:
1276 		policy = SCHED_FIFO;
1277 		break;
1278 	case LINUX_SCHED_RR:
1279 		policy = SCHED_RR;
1280 		break;
1281 	default:
1282 		return (EINVAL);
1283 	}
1284 
1285 	error = copyin(args->param, &sched_param, sizeof(sched_param));
1286 	if (error)
1287 		return (error);
1288 
1289 	if (linux_map_sched_prio) {
1290 		switch (policy) {
1291 		case SCHED_OTHER:
1292 			if (sched_param.sched_priority != 0)
1293 				return (EINVAL);
1294 
1295 			sched_param.sched_priority =
1296 			    PRI_MAX_TIMESHARE - PRI_MIN_TIMESHARE;
1297 			break;
1298 		case SCHED_FIFO:
1299 		case SCHED_RR:
1300 			if (sched_param.sched_priority < 1 ||
1301 			    sched_param.sched_priority >= LINUX_MAX_RT_PRIO)
1302 				return (EINVAL);
1303 
1304 			/*
1305 			 * Map [1, LINUX_MAX_RT_PRIO - 1] to
1306 			 * [0, RTP_PRIO_MAX - RTP_PRIO_MIN] (rounding down).
1307 			 */
1308 			sched_param.sched_priority =
1309 			    (sched_param.sched_priority - 1) *
1310 			    (RTP_PRIO_MAX - RTP_PRIO_MIN + 1) /
1311 			    (LINUX_MAX_RT_PRIO - 1);
1312 			break;
1313 		}
1314 	}
1315 
1316 	tdt = linux_tdfind(td, args->pid, -1);
1317 	if (tdt == NULL)
1318 		return (ESRCH);
1319 
1320 	error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1321 	PROC_UNLOCK(tdt->td_proc);
1322 	return (error);
1323 }
1324 
1325 int
1326 linux_sched_getscheduler(struct thread *td,
1327     struct linux_sched_getscheduler_args *args)
1328 {
1329 	struct thread *tdt;
1330 	int error, policy;
1331 
1332 	tdt = linux_tdfind(td, args->pid, -1);
1333 	if (tdt == NULL)
1334 		return (ESRCH);
1335 
1336 	error = kern_sched_getscheduler(td, tdt, &policy);
1337 	PROC_UNLOCK(tdt->td_proc);
1338 
1339 	switch (policy) {
1340 	case SCHED_OTHER:
1341 		td->td_retval[0] = LINUX_SCHED_OTHER;
1342 		break;
1343 	case SCHED_FIFO:
1344 		td->td_retval[0] = LINUX_SCHED_FIFO;
1345 		break;
1346 	case SCHED_RR:
1347 		td->td_retval[0] = LINUX_SCHED_RR;
1348 		break;
1349 	}
1350 	return (error);
1351 }
1352 
1353 int
1354 linux_sched_get_priority_max(struct thread *td,
1355     struct linux_sched_get_priority_max_args *args)
1356 {
1357 	struct sched_get_priority_max_args bsd;
1358 
1359 	if (linux_map_sched_prio) {
1360 		switch (args->policy) {
1361 		case LINUX_SCHED_OTHER:
1362 			td->td_retval[0] = 0;
1363 			return (0);
1364 		case LINUX_SCHED_FIFO:
1365 		case LINUX_SCHED_RR:
1366 			td->td_retval[0] = LINUX_MAX_RT_PRIO - 1;
1367 			return (0);
1368 		default:
1369 			return (EINVAL);
1370 		}
1371 	}
1372 
1373 	switch (args->policy) {
1374 	case LINUX_SCHED_OTHER:
1375 		bsd.policy = SCHED_OTHER;
1376 		break;
1377 	case LINUX_SCHED_FIFO:
1378 		bsd.policy = SCHED_FIFO;
1379 		break;
1380 	case LINUX_SCHED_RR:
1381 		bsd.policy = SCHED_RR;
1382 		break;
1383 	default:
1384 		return (EINVAL);
1385 	}
1386 	return (sys_sched_get_priority_max(td, &bsd));
1387 }
1388 
1389 int
1390 linux_sched_get_priority_min(struct thread *td,
1391     struct linux_sched_get_priority_min_args *args)
1392 {
1393 	struct sched_get_priority_min_args bsd;
1394 
1395 	if (linux_map_sched_prio) {
1396 		switch (args->policy) {
1397 		case LINUX_SCHED_OTHER:
1398 			td->td_retval[0] = 0;
1399 			return (0);
1400 		case LINUX_SCHED_FIFO:
1401 		case LINUX_SCHED_RR:
1402 			td->td_retval[0] = 1;
1403 			return (0);
1404 		default:
1405 			return (EINVAL);
1406 		}
1407 	}
1408 
1409 	switch (args->policy) {
1410 	case LINUX_SCHED_OTHER:
1411 		bsd.policy = SCHED_OTHER;
1412 		break;
1413 	case LINUX_SCHED_FIFO:
1414 		bsd.policy = SCHED_FIFO;
1415 		break;
1416 	case LINUX_SCHED_RR:
1417 		bsd.policy = SCHED_RR;
1418 		break;
1419 	default:
1420 		return (EINVAL);
1421 	}
1422 	return (sys_sched_get_priority_min(td, &bsd));
1423 }
1424 
1425 #define REBOOT_CAD_ON	0x89abcdef
1426 #define REBOOT_CAD_OFF	0
1427 #define REBOOT_HALT	0xcdef0123
1428 #define REBOOT_RESTART	0x01234567
1429 #define REBOOT_RESTART2	0xA1B2C3D4
1430 #define REBOOT_POWEROFF	0x4321FEDC
1431 #define REBOOT_MAGIC1	0xfee1dead
1432 #define REBOOT_MAGIC2	0x28121969
1433 #define REBOOT_MAGIC2A	0x05121996
1434 #define REBOOT_MAGIC2B	0x16041998
1435 
1436 int
1437 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1438 {
1439 	struct reboot_args bsd_args;
1440 
1441 	if (args->magic1 != REBOOT_MAGIC1)
1442 		return (EINVAL);
1443 
1444 	switch (args->magic2) {
1445 	case REBOOT_MAGIC2:
1446 	case REBOOT_MAGIC2A:
1447 	case REBOOT_MAGIC2B:
1448 		break;
1449 	default:
1450 		return (EINVAL);
1451 	}
1452 
1453 	switch (args->cmd) {
1454 	case REBOOT_CAD_ON:
1455 	case REBOOT_CAD_OFF:
1456 		return (priv_check(td, PRIV_REBOOT));
1457 	case REBOOT_HALT:
1458 		bsd_args.opt = RB_HALT;
1459 		break;
1460 	case REBOOT_RESTART:
1461 	case REBOOT_RESTART2:
1462 		bsd_args.opt = 0;
1463 		break;
1464 	case REBOOT_POWEROFF:
1465 		bsd_args.opt = RB_POWEROFF;
1466 		break;
1467 	default:
1468 		return (EINVAL);
1469 	}
1470 	return (sys_reboot(td, &bsd_args));
1471 }
1472 
1473 int
1474 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1475 {
1476 
1477 	td->td_retval[0] = td->td_proc->p_pid;
1478 
1479 	return (0);
1480 }
1481 
1482 int
1483 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1484 {
1485 	struct linux_emuldata *em;
1486 
1487 	em = em_find(td);
1488 	KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1489 
1490 	td->td_retval[0] = em->em_tid;
1491 
1492 	return (0);
1493 }
1494 
1495 int
1496 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1497 {
1498 
1499 	td->td_retval[0] = kern_getppid(td);
1500 	return (0);
1501 }
1502 
1503 int
1504 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1505 {
1506 
1507 	td->td_retval[0] = td->td_ucred->cr_rgid;
1508 	return (0);
1509 }
1510 
1511 int
1512 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1513 {
1514 
1515 	td->td_retval[0] = td->td_ucred->cr_ruid;
1516 	return (0);
1517 }
1518 
1519 int
1520 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1521 {
1522 
1523 	return (kern_getsid(td, args->pid));
1524 }
1525 
1526 int
1527 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1528 {
1529 	int error;
1530 
1531 	error = kern_getpriority(td, args->which, args->who);
1532 	td->td_retval[0] = 20 - td->td_retval[0];
1533 	return (error);
1534 }
1535 
1536 int
1537 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1538 {
1539 	int name[2];
1540 
1541 	name[0] = CTL_KERN;
1542 	name[1] = KERN_HOSTNAME;
1543 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1544 	    args->len, 0, 0));
1545 }
1546 
1547 int
1548 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1549 {
1550 	int name[2];
1551 
1552 	name[0] = CTL_KERN;
1553 	name[1] = KERN_NISDOMAINNAME;
1554 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1555 	    args->len, 0, 0));
1556 }
1557 
1558 int
1559 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1560 {
1561 
1562 	LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1563 	    args->error_code);
1564 
1565 	/*
1566 	 * XXX: we should send a signal to the parent if
1567 	 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1568 	 * as it doesnt occur often.
1569 	 */
1570 	kern_exit(td, args->error_code, 0);
1571 	return (0);
1572 }
1573 
1574 #define _LINUX_CAPABILITY_VERSION_1  0x19980330
1575 #define _LINUX_CAPABILITY_VERSION_2  0x20071026
1576 #define _LINUX_CAPABILITY_VERSION_3  0x20080522
1577 
1578 struct l_user_cap_header {
1579 	l_int	version;
1580 	l_int	pid;
1581 };
1582 
1583 struct l_user_cap_data {
1584 	l_int	effective;
1585 	l_int	permitted;
1586 	l_int	inheritable;
1587 };
1588 
1589 int
1590 linux_capget(struct thread *td, struct linux_capget_args *uap)
1591 {
1592 	struct l_user_cap_header luch;
1593 	struct l_user_cap_data lucd[2];
1594 	int error, u32s;
1595 
1596 	if (uap->hdrp == NULL)
1597 		return (EFAULT);
1598 
1599 	error = copyin(uap->hdrp, &luch, sizeof(luch));
1600 	if (error != 0)
1601 		return (error);
1602 
1603 	switch (luch.version) {
1604 	case _LINUX_CAPABILITY_VERSION_1:
1605 		u32s = 1;
1606 		break;
1607 	case _LINUX_CAPABILITY_VERSION_2:
1608 	case _LINUX_CAPABILITY_VERSION_3:
1609 		u32s = 2;
1610 		break;
1611 	default:
1612 		luch.version = _LINUX_CAPABILITY_VERSION_1;
1613 		error = copyout(&luch, uap->hdrp, sizeof(luch));
1614 		if (error)
1615 			return (error);
1616 		return (EINVAL);
1617 	}
1618 
1619 	if (luch.pid)
1620 		return (EPERM);
1621 
1622 	if (uap->datap) {
1623 		/*
1624 		 * The current implementation doesn't support setting
1625 		 * a capability (it's essentially a stub) so indicate
1626 		 * that no capabilities are currently set or available
1627 		 * to request.
1628 		 */
1629 		memset(&lucd, 0, u32s * sizeof(lucd[0]));
1630 		error = copyout(&lucd, uap->datap, u32s * sizeof(lucd[0]));
1631 	}
1632 
1633 	return (error);
1634 }
1635 
1636 int
1637 linux_capset(struct thread *td, struct linux_capset_args *uap)
1638 {
1639 	struct l_user_cap_header luch;
1640 	struct l_user_cap_data lucd[2];
1641 	int error, i, u32s;
1642 
1643 	if (uap->hdrp == NULL || uap->datap == NULL)
1644 		return (EFAULT);
1645 
1646 	error = copyin(uap->hdrp, &luch, sizeof(luch));
1647 	if (error != 0)
1648 		return (error);
1649 
1650 	switch (luch.version) {
1651 	case _LINUX_CAPABILITY_VERSION_1:
1652 		u32s = 1;
1653 		break;
1654 	case _LINUX_CAPABILITY_VERSION_2:
1655 	case _LINUX_CAPABILITY_VERSION_3:
1656 		u32s = 2;
1657 		break;
1658 	default:
1659 		luch.version = _LINUX_CAPABILITY_VERSION_1;
1660 		error = copyout(&luch, uap->hdrp, sizeof(luch));
1661 		if (error)
1662 			return (error);
1663 		return (EINVAL);
1664 	}
1665 
1666 	if (luch.pid)
1667 		return (EPERM);
1668 
1669 	error = copyin(uap->datap, &lucd, u32s * sizeof(lucd[0]));
1670 	if (error != 0)
1671 		return (error);
1672 
1673 	/* We currently don't support setting any capabilities. */
1674 	for (i = 0; i < u32s; i++) {
1675 		if (lucd[i].effective || lucd[i].permitted ||
1676 		    lucd[i].inheritable) {
1677 			linux_msg(td,
1678 			    "capset[%d] effective=0x%x, permitted=0x%x, "
1679 			    "inheritable=0x%x is not implemented", i,
1680 			    (int)lucd[i].effective, (int)lucd[i].permitted,
1681 			    (int)lucd[i].inheritable);
1682 			return (EPERM);
1683 		}
1684 	}
1685 
1686 	return (0);
1687 }
1688 
1689 int
1690 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1691 {
1692 	int error = 0, max_size, arg;
1693 	struct proc *p = td->td_proc;
1694 	char comm[LINUX_MAX_COMM_LEN];
1695 	int pdeath_signal, trace_state;
1696 
1697 	switch (args->option) {
1698 	case LINUX_PR_SET_PDEATHSIG:
1699 		if (!LINUX_SIG_VALID(args->arg2))
1700 			return (EINVAL);
1701 		pdeath_signal = linux_to_bsd_signal(args->arg2);
1702 		return (kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_CTL,
1703 		    &pdeath_signal));
1704 	case LINUX_PR_GET_PDEATHSIG:
1705 		error = kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_STATUS,
1706 		    &pdeath_signal);
1707 		if (error != 0)
1708 			return (error);
1709 		pdeath_signal = bsd_to_linux_signal(pdeath_signal);
1710 		return (copyout(&pdeath_signal,
1711 		    (void *)(register_t)args->arg2,
1712 		    sizeof(pdeath_signal)));
1713 	/*
1714 	 * In Linux, this flag controls if set[gu]id processes can coredump.
1715 	 * There are additional semantics imposed on processes that cannot
1716 	 * coredump:
1717 	 * - Such processes can not be ptraced.
1718 	 * - There are some semantics around ownership of process-related files
1719 	 *   in the /proc namespace.
1720 	 *
1721 	 * In FreeBSD, we can (and by default, do) disable setuid coredump
1722 	 * system-wide with 'sugid_coredump.'  We control tracability on a
1723 	 * per-process basis with the procctl PROC_TRACE (=> P2_NOTRACE flag).
1724 	 * By happy coincidence, P2_NOTRACE also prevents coredumping.  So the
1725 	 * procctl is roughly analogous to Linux's DUMPABLE.
1726 	 *
1727 	 * So, proxy these knobs to the corresponding PROC_TRACE setting.
1728 	 */
1729 	case LINUX_PR_GET_DUMPABLE:
1730 		error = kern_procctl(td, P_PID, p->p_pid, PROC_TRACE_STATUS,
1731 		    &trace_state);
1732 		if (error != 0)
1733 			return (error);
1734 		td->td_retval[0] = (trace_state != -1);
1735 		return (0);
1736 	case LINUX_PR_SET_DUMPABLE:
1737 		/*
1738 		 * It is only valid for userspace to set one of these two
1739 		 * flags, and only one at a time.
1740 		 */
1741 		switch (args->arg2) {
1742 		case LINUX_SUID_DUMP_DISABLE:
1743 			trace_state = PROC_TRACE_CTL_DISABLE_EXEC;
1744 			break;
1745 		case LINUX_SUID_DUMP_USER:
1746 			trace_state = PROC_TRACE_CTL_ENABLE;
1747 			break;
1748 		default:
1749 			return (EINVAL);
1750 		}
1751 		return (kern_procctl(td, P_PID, p->p_pid, PROC_TRACE_CTL,
1752 		    &trace_state));
1753 	case LINUX_PR_GET_KEEPCAPS:
1754 		/*
1755 		 * Indicate that we always clear the effective and
1756 		 * permitted capability sets when the user id becomes
1757 		 * non-zero (actually the capability sets are simply
1758 		 * always zero in the current implementation).
1759 		 */
1760 		td->td_retval[0] = 0;
1761 		break;
1762 	case LINUX_PR_SET_KEEPCAPS:
1763 		/*
1764 		 * Ignore requests to keep the effective and permitted
1765 		 * capability sets when the user id becomes non-zero.
1766 		 */
1767 		break;
1768 	case LINUX_PR_SET_NAME:
1769 		/*
1770 		 * To be on the safe side we need to make sure to not
1771 		 * overflow the size a Linux program expects. We already
1772 		 * do this here in the copyin, so that we don't need to
1773 		 * check on copyout.
1774 		 */
1775 		max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1776 		error = copyinstr((void *)(register_t)args->arg2, comm,
1777 		    max_size, NULL);
1778 
1779 		/* Linux silently truncates the name if it is too long. */
1780 		if (error == ENAMETOOLONG) {
1781 			/*
1782 			 * XXX: copyinstr() isn't documented to populate the
1783 			 * array completely, so do a copyin() to be on the
1784 			 * safe side. This should be changed in case
1785 			 * copyinstr() is changed to guarantee this.
1786 			 */
1787 			error = copyin((void *)(register_t)args->arg2, comm,
1788 			    max_size - 1);
1789 			comm[max_size - 1] = '\0';
1790 		}
1791 		if (error)
1792 			return (error);
1793 
1794 		PROC_LOCK(p);
1795 		strlcpy(p->p_comm, comm, sizeof(p->p_comm));
1796 		PROC_UNLOCK(p);
1797 		break;
1798 	case LINUX_PR_GET_NAME:
1799 		PROC_LOCK(p);
1800 		strlcpy(comm, p->p_comm, sizeof(comm));
1801 		PROC_UNLOCK(p);
1802 		error = copyout(comm, (void *)(register_t)args->arg2,
1803 		    strlen(comm) + 1);
1804 		break;
1805 	case LINUX_PR_GET_SECCOMP:
1806 	case LINUX_PR_SET_SECCOMP:
1807 		/*
1808 		 * Same as returned by Linux without CONFIG_SECCOMP enabled.
1809 		 */
1810 		error = EINVAL;
1811 		break;
1812 	case LINUX_PR_CAPBSET_READ:
1813 #if 0
1814 		/*
1815 		 * This makes too much noise with Ubuntu Focal.
1816 		 */
1817 		linux_msg(td, "unsupported prctl PR_CAPBSET_READ %d",
1818 		    (int)args->arg2);
1819 #endif
1820 		error = EINVAL;
1821 		break;
1822 	case LINUX_PR_SET_CHILD_SUBREAPER:
1823 		if (args->arg2 == 0) {
1824 			return (kern_procctl(td, P_PID, 0, PROC_REAP_RELEASE,
1825 			    NULL));
1826 		}
1827 
1828 		return (kern_procctl(td, P_PID, 0, PROC_REAP_ACQUIRE,
1829 		    NULL));
1830 	case LINUX_PR_GET_CHILD_SUBREAPER: {
1831 		struct procctl_reaper_status rs;
1832 		l_int val;
1833 
1834 		error = kern_procctl(td, P_PID, 0, PROC_REAP_STATUS, &rs);
1835 		if (error != 0)
1836 			return (error);
1837 		val = rs.rs_reaper == p->p_pid ? 1 : 0;
1838 		error = copyout(&val, (void *)(register_t)args->arg2,
1839 		    sizeof(val));
1840 		break;
1841 	}
1842 	case LINUX_PR_SET_NO_NEW_PRIVS:
1843 		arg = args->arg2 == 1 ?
1844 		    PROC_NO_NEW_PRIVS_ENABLE : PROC_NO_NEW_PRIVS_DISABLE;
1845 		error = kern_procctl(td, P_PID, p->p_pid,
1846 		    PROC_NO_NEW_PRIVS_CTL, &arg);
1847 		break;
1848 	case LINUX_PR_GET_NO_NEW_PRIVS:
1849 		error = kern_procctl(td, P_PID, p->p_pid,
1850 		    PROC_NO_NEW_PRIVS_STATUS, &arg);
1851 		if (error != 0)
1852 			return (error);
1853 		/* Linux returns the value as the syscall return */
1854 		td->td_retval[0] = arg == PROC_NO_NEW_PRIVS_ENABLE ? 1 : 0;
1855 		break;
1856 	case LINUX_PR_SET_PTRACER:
1857 		linux_msg(td, "unsupported prctl PR_SET_PTRACER");
1858 		error = EINVAL;
1859 		break;
1860 	case LINUX_PR_SET_VMA:
1861 		if (args->arg2 != LINUX_PR_SET_VMA_ANON_NAME) {
1862 			linux_msg(td, "unsupported prctl PR_SET_VMA attr %ju",
1863 			    (uintmax_t)args->arg2);
1864 			error = EINVAL;
1865 		}
1866 		break;
1867 	default:
1868 		linux_msg(td, "unsupported prctl option %d", args->option);
1869 		error = EINVAL;
1870 		break;
1871 	}
1872 
1873 	return (error);
1874 }
1875 
1876 int
1877 linux_sched_setparam(struct thread *td,
1878     struct linux_sched_setparam_args *uap)
1879 {
1880 	struct sched_param sched_param;
1881 	struct thread *tdt;
1882 	int error, policy;
1883 
1884 	error = copyin(uap->param, &sched_param, sizeof(sched_param));
1885 	if (error)
1886 		return (error);
1887 
1888 	tdt = linux_tdfind(td, uap->pid, -1);
1889 	if (tdt == NULL)
1890 		return (ESRCH);
1891 
1892 	if (linux_map_sched_prio) {
1893 		error = kern_sched_getscheduler(td, tdt, &policy);
1894 		if (error)
1895 			goto out;
1896 
1897 		switch (policy) {
1898 		case SCHED_OTHER:
1899 			if (sched_param.sched_priority != 0) {
1900 				error = EINVAL;
1901 				goto out;
1902 			}
1903 			sched_param.sched_priority =
1904 			    PRI_MAX_TIMESHARE - PRI_MIN_TIMESHARE;
1905 			break;
1906 		case SCHED_FIFO:
1907 		case SCHED_RR:
1908 			if (sched_param.sched_priority < 1 ||
1909 			    sched_param.sched_priority >= LINUX_MAX_RT_PRIO) {
1910 				error = EINVAL;
1911 				goto out;
1912 			}
1913 			/*
1914 			 * Map [1, LINUX_MAX_RT_PRIO - 1] to
1915 			 * [0, RTP_PRIO_MAX - RTP_PRIO_MIN] (rounding down).
1916 			 */
1917 			sched_param.sched_priority =
1918 			    (sched_param.sched_priority - 1) *
1919 			    (RTP_PRIO_MAX - RTP_PRIO_MIN + 1) /
1920 			    (LINUX_MAX_RT_PRIO - 1);
1921 			break;
1922 		}
1923 	}
1924 
1925 	error = kern_sched_setparam(td, tdt, &sched_param);
1926 out:	PROC_UNLOCK(tdt->td_proc);
1927 	return (error);
1928 }
1929 
1930 int
1931 linux_sched_getparam(struct thread *td,
1932     struct linux_sched_getparam_args *uap)
1933 {
1934 	struct sched_param sched_param;
1935 	struct thread *tdt;
1936 	int error, policy;
1937 
1938 	tdt = linux_tdfind(td, uap->pid, -1);
1939 	if (tdt == NULL)
1940 		return (ESRCH);
1941 
1942 	error = kern_sched_getparam(td, tdt, &sched_param);
1943 	if (error) {
1944 		PROC_UNLOCK(tdt->td_proc);
1945 		return (error);
1946 	}
1947 
1948 	if (linux_map_sched_prio) {
1949 		error = kern_sched_getscheduler(td, tdt, &policy);
1950 		PROC_UNLOCK(tdt->td_proc);
1951 		if (error)
1952 			return (error);
1953 
1954 		switch (policy) {
1955 		case SCHED_OTHER:
1956 			sched_param.sched_priority = 0;
1957 			break;
1958 		case SCHED_FIFO:
1959 		case SCHED_RR:
1960 			/*
1961 			 * Map [0, RTP_PRIO_MAX - RTP_PRIO_MIN] to
1962 			 * [1, LINUX_MAX_RT_PRIO - 1] (rounding up).
1963 			 */
1964 			sched_param.sched_priority =
1965 			    (sched_param.sched_priority *
1966 			    (LINUX_MAX_RT_PRIO - 1) +
1967 			    (RTP_PRIO_MAX - RTP_PRIO_MIN - 1)) /
1968 			    (RTP_PRIO_MAX - RTP_PRIO_MIN) + 1;
1969 			break;
1970 		}
1971 	} else
1972 		PROC_UNLOCK(tdt->td_proc);
1973 
1974 	error = copyout(&sched_param, uap->param, sizeof(sched_param));
1975 	return (error);
1976 }
1977 
1978 /*
1979  * Get affinity of a process.
1980  */
1981 int
1982 linux_sched_getaffinity(struct thread *td,
1983     struct linux_sched_getaffinity_args *args)
1984 {
1985 	struct thread *tdt;
1986 	cpuset_t *mask;
1987 	size_t size;
1988 	int error;
1989 	id_t tid;
1990 
1991 	tdt = linux_tdfind(td, args->pid, -1);
1992 	if (tdt == NULL)
1993 		return (ESRCH);
1994 	tid = tdt->td_tid;
1995 	PROC_UNLOCK(tdt->td_proc);
1996 
1997 	mask = malloc(sizeof(cpuset_t), M_LINUX, M_WAITOK | M_ZERO);
1998 	size = min(args->len, sizeof(cpuset_t));
1999 	error = kern_cpuset_getaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2000 	    tid, size, mask);
2001 	if (error == ERANGE)
2002 		error = EINVAL;
2003  	if (error == 0)
2004 		error = copyout(mask, args->user_mask_ptr, size);
2005 	if (error == 0)
2006 		td->td_retval[0] = size;
2007 	free(mask, M_LINUX);
2008 	return (error);
2009 }
2010 
2011 /*
2012  *  Set affinity of a process.
2013  */
2014 int
2015 linux_sched_setaffinity(struct thread *td,
2016     struct linux_sched_setaffinity_args *args)
2017 {
2018 	struct thread *tdt;
2019 	cpuset_t *mask;
2020 	int cpu, error;
2021 	size_t len;
2022 	id_t tid;
2023 
2024 	tdt = linux_tdfind(td, args->pid, -1);
2025 	if (tdt == NULL)
2026 		return (ESRCH);
2027 	tid = tdt->td_tid;
2028 	PROC_UNLOCK(tdt->td_proc);
2029 
2030 	len = min(args->len, sizeof(cpuset_t));
2031 	mask = malloc(sizeof(cpuset_t), M_TEMP, M_WAITOK | M_ZERO);
2032 	error = copyin(args->user_mask_ptr, mask, len);
2033 	if (error != 0)
2034 		goto out;
2035 	/* Linux ignore high bits */
2036 	CPU_FOREACH_ISSET(cpu, mask)
2037 		if (cpu > mp_maxid)
2038 			CPU_CLR(cpu, mask);
2039 
2040 	error = kern_cpuset_setaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2041 	    tid, mask);
2042 	if (error == EDEADLK)
2043 		error = EINVAL;
2044 out:
2045 	free(mask, M_TEMP);
2046 	return (error);
2047 }
2048 
2049 struct linux_rlimit64 {
2050 	uint64_t	rlim_cur;
2051 	uint64_t	rlim_max;
2052 };
2053 
2054 int
2055 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
2056 {
2057 	struct rlimit rlim, nrlim;
2058 	struct linux_rlimit64 lrlim;
2059 	struct proc *p;
2060 	u_int which;
2061 	int flags;
2062 	int error;
2063 	bool exec_blocked;
2064 
2065 	if (args->new == NULL && args->old != NULL) {
2066 		if (linux_get_dummy_limit(td, args->resource, &rlim)) {
2067 			lrlim.rlim_cur = rlim.rlim_cur;
2068 			lrlim.rlim_max = rlim.rlim_max;
2069 			return (copyout(&lrlim, args->old, sizeof(lrlim)));
2070 		}
2071 	}
2072 
2073 	if (args->resource >= LINUX_RLIM_NLIMITS)
2074 		return (EINVAL);
2075 
2076 	which = linux_to_bsd_resource[args->resource];
2077 	if (which == -1)
2078 		return (EINVAL);
2079 
2080 	if (args->new != NULL) {
2081 		/*
2082 		 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
2083 		 * rlim is unsigned 64-bit. FreeBSD treats negative limits
2084 		 * as INFINITY so we do not need a conversion even.
2085 		 */
2086 		error = copyin(args->new, &nrlim, sizeof(nrlim));
2087 		if (error != 0)
2088 			return (error);
2089 	}
2090 
2091 	exec_blocked = false;
2092 	flags = PGET_HOLD | PGET_NOTWEXIT;
2093 	if (args->new != NULL)
2094 		flags |= PGET_CANDEBUG;
2095 	else
2096 		flags |= PGET_CANSEE;
2097 	if (args->pid == 0) {
2098 		p = td->td_proc;
2099 		PHOLD(p);
2100 	} else {
2101 		error = pget(args->pid, flags, &p);
2102 		if (error != 0)
2103 			return (error);
2104 		exec_blocked = true;
2105 		PROC_LOCK(p);
2106 		execve_block_wait(td, p);
2107 		error = args->new != NULL ? p_candebug(td, p) :
2108 		    p_cansee(td, p);
2109 		PROC_UNLOCK(p);
2110 		if (error != 0)
2111 			goto out;
2112 	}
2113 	if (args->old != NULL) {
2114 		PROC_LOCK(p);
2115 		lim_rlimit_proc(p, which, &rlim);
2116 		PROC_UNLOCK(p);
2117 		if (rlim.rlim_cur == RLIM_INFINITY)
2118 			lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2119 		else
2120 			lrlim.rlim_cur = rlim.rlim_cur;
2121 		if (rlim.rlim_max == RLIM_INFINITY)
2122 			lrlim.rlim_max = LINUX_RLIM_INFINITY;
2123 		else
2124 			lrlim.rlim_max = rlim.rlim_max;
2125 		error = copyout(&lrlim, args->old, sizeof(lrlim));
2126 		if (error != 0)
2127 			goto out;
2128 	}
2129 
2130 	if (args->new != NULL)
2131 		error = kern_proc_setrlimit(td, p, which, &nrlim);
2132 
2133  out:
2134 	if (exec_blocked) {
2135 		PROC_LOCK(p);
2136 		execve_unblock(td, p);
2137 		PROC_UNLOCK(p);
2138 	}
2139 	PRELE(p);
2140 	return (error);
2141 }
2142 
2143 int
2144 linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2145 {
2146 	struct timespec ts, *tsp;
2147 	int error;
2148 
2149 	if (args->tsp != NULL) {
2150 		error = linux_get_timespec(&ts, args->tsp);
2151 		if (error != 0)
2152 			return (error);
2153 		tsp = &ts;
2154 	} else
2155 		tsp = NULL;
2156 
2157 	error = linux_common_pselect6(td, args->nfds, args->readfds,
2158 	    args->writefds, args->exceptfds, tsp, args->sig);
2159 
2160 	if (args->tsp != NULL)
2161 		linux_put_timespec(&ts, args->tsp);
2162 	return (error);
2163 }
2164 
2165 static int
2166 linux_common_pselect6(struct thread *td, l_int nfds, l_fd_set *readfds,
2167     l_fd_set *writefds, l_fd_set *exceptfds, struct timespec *tsp,
2168     l_uintptr_t *sig)
2169 {
2170 	struct timeval utv, tv0, tv1, *tvp;
2171 	struct l_pselect6arg lpse6;
2172 	sigset_t *ssp;
2173 	sigset_t ss;
2174 	int error;
2175 
2176 	ssp = NULL;
2177 	if (sig != NULL) {
2178 		error = copyin(sig, &lpse6, sizeof(lpse6));
2179 		if (error != 0)
2180 			return (error);
2181 		error = linux_copyin_sigset(td, PTRIN(lpse6.ss),
2182 		    lpse6.ss_len, &ss, &ssp);
2183 		if (error != 0)
2184 		    return (error);
2185 	} else
2186 		ssp = NULL;
2187 
2188 	/*
2189 	 * Currently glibc changes nanosecond number to microsecond.
2190 	 * This mean losing precision but for now it is hardly seen.
2191 	 */
2192 	if (tsp != NULL) {
2193 		TIMESPEC_TO_TIMEVAL(&utv, tsp);
2194 		if (itimerfix(&utv))
2195 			return (EINVAL);
2196 
2197 		microtime(&tv0);
2198 		tvp = &utv;
2199 	} else
2200 		tvp = NULL;
2201 
2202 	error = kern_pselect(td, nfds, readfds, writefds,
2203 	    exceptfds, tvp, ssp, LINUX_NFDBITS);
2204 
2205 	if (tsp != NULL) {
2206 		/*
2207 		 * Compute how much time was left of the timeout,
2208 		 * by subtracting the current time and the time
2209 		 * before we started the call, and subtracting
2210 		 * that result from the user-supplied value.
2211 		 */
2212 		microtime(&tv1);
2213 		timevalsub(&tv1, &tv0);
2214 		timevalsub(&utv, &tv1);
2215 		if (utv.tv_sec < 0)
2216 			timevalclear(&utv);
2217 		TIMEVAL_TO_TIMESPEC(&utv, tsp);
2218 	}
2219 	return (error);
2220 }
2221 
2222 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
2223 int
2224 linux_pselect6_time64(struct thread *td,
2225     struct linux_pselect6_time64_args *args)
2226 {
2227 	struct timespec ts, *tsp;
2228 	int error;
2229 
2230 	if (args->tsp != NULL) {
2231 		error = linux_get_timespec64(&ts, args->tsp);
2232 		if (error != 0)
2233 			return (error);
2234 		tsp = &ts;
2235 	} else
2236 		tsp = NULL;
2237 
2238 	error = linux_common_pselect6(td, args->nfds, args->readfds,
2239 	    args->writefds, args->exceptfds, tsp, args->sig);
2240 
2241 	if (args->tsp != NULL)
2242 		linux_put_timespec64(&ts, args->tsp);
2243 	return (error);
2244 }
2245 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
2246 
2247 int
2248 linux_ppoll(struct thread *td, struct linux_ppoll_args *args)
2249 {
2250 	struct timespec uts, *tsp;
2251 	int error;
2252 
2253 	if (args->tsp != NULL) {
2254 		error = linux_get_timespec(&uts, args->tsp);
2255 		if (error != 0)
2256 			return (error);
2257 		tsp = &uts;
2258 	} else
2259 		tsp = NULL;
2260 
2261 	error = linux_common_ppoll(td, args->fds, args->nfds, tsp,
2262 	    args->sset, args->ssize);
2263 	if (error == 0 && args->tsp != NULL)
2264 		error = linux_put_timespec(&uts, args->tsp);
2265 	return (error);
2266 }
2267 
2268 static int
2269 linux_common_ppoll(struct thread *td, struct pollfd *fds, uint32_t nfds,
2270     struct timespec *tsp, l_sigset_t *sset, l_size_t ssize)
2271 {
2272 	struct timespec ts0, ts1;
2273 	struct pollfd stackfds[32];
2274 	struct pollfd *kfds;
2275  	sigset_t *ssp;
2276  	sigset_t ss;
2277  	int error;
2278 
2279 	if (kern_poll_maxfds(nfds))
2280 		return (EINVAL);
2281 	if (sset != NULL) {
2282 		error = linux_copyin_sigset(td, sset, ssize, &ss, &ssp);
2283 		if (error != 0)
2284 		    return (error);
2285 	} else
2286 		ssp = NULL;
2287 	if (tsp != NULL)
2288 		nanotime(&ts0);
2289 
2290 	if (nfds > nitems(stackfds))
2291 		kfds = mallocarray(nfds, sizeof(*kfds), M_TEMP, M_WAITOK);
2292 	else
2293 		kfds = stackfds;
2294 	error = linux_pollin(td, kfds, fds, nfds);
2295 	if (error != 0)
2296 		goto out;
2297 
2298 	error = kern_poll_kfds(td, kfds, nfds, tsp, ssp);
2299 	if (error == 0)
2300 		error = linux_pollout(td, kfds, fds, nfds);
2301 
2302 	if (error == 0 && tsp != NULL) {
2303 		if (td->td_retval[0]) {
2304 			nanotime(&ts1);
2305 			timespecsub(&ts1, &ts0, &ts1);
2306 			timespecsub(tsp, &ts1, tsp);
2307 			if (tsp->tv_sec < 0)
2308 				timespecclear(tsp);
2309 		} else
2310 			timespecclear(tsp);
2311 	}
2312 
2313 out:
2314 	if (nfds > nitems(stackfds))
2315 		free(kfds, M_TEMP);
2316 	return (error);
2317 }
2318 
2319 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
2320 int
2321 linux_ppoll_time64(struct thread *td, struct linux_ppoll_time64_args *args)
2322 {
2323 	struct timespec uts, *tsp;
2324 	int error;
2325 
2326 	if (args->tsp != NULL) {
2327 		error = linux_get_timespec64(&uts, args->tsp);
2328 		if (error != 0)
2329 			return (error);
2330 		tsp = &uts;
2331 	} else
2332  		tsp = NULL;
2333 	error = linux_common_ppoll(td, args->fds, args->nfds, tsp,
2334 	    args->sset, args->ssize);
2335 	if (error == 0 && args->tsp != NULL)
2336 		error = linux_put_timespec64(&uts, args->tsp);
2337 	return (error);
2338 }
2339 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
2340 
2341 static int
2342 linux_pollin(struct thread *td, struct pollfd *fds, struct pollfd *ufds, u_int nfd)
2343 {
2344 	int error;
2345 	u_int i;
2346 
2347 	error = copyin(ufds, fds, nfd * sizeof(*fds));
2348 	if (error != 0)
2349 		return (error);
2350 
2351 	for (i = 0; i < nfd; i++) {
2352 		if (fds->events != 0)
2353 			linux_to_bsd_poll_events(td, fds->fd,
2354 			    fds->events, &fds->events);
2355 		fds++;
2356 	}
2357 	return (0);
2358 }
2359 
2360 static int
2361 linux_pollout(struct thread *td, struct pollfd *fds, struct pollfd *ufds, u_int nfd)
2362 {
2363 	int error = 0;
2364 	u_int i, n = 0;
2365 
2366 	for (i = 0; i < nfd; i++) {
2367 		if (fds->revents != 0) {
2368 			bsd_to_linux_poll_events(fds->revents,
2369 			    &fds->revents);
2370 			n++;
2371 		}
2372 		error = copyout(&fds->revents, &ufds->revents,
2373 		    sizeof(ufds->revents));
2374 		if (error)
2375 			return (error);
2376 		fds++;
2377 		ufds++;
2378 	}
2379 	td->td_retval[0] = n;
2380 	return (0);
2381 }
2382 
2383 static int
2384 linux_sched_rr_get_interval_common(struct thread *td, pid_t pid,
2385     struct timespec *ts)
2386 {
2387 	struct thread *tdt;
2388 	int error;
2389 
2390 	/*
2391 	 * According to man in case the invalid pid specified
2392 	 * EINVAL should be returned.
2393 	 */
2394 	if (pid < 0)
2395 		return (EINVAL);
2396 
2397 	tdt = linux_tdfind(td, pid, -1);
2398 	if (tdt == NULL)
2399 		return (ESRCH);
2400 
2401 	error = kern_sched_rr_get_interval_td(td, tdt, ts);
2402 	PROC_UNLOCK(tdt->td_proc);
2403 	return (error);
2404 }
2405 
2406 int
2407 linux_sched_rr_get_interval(struct thread *td,
2408     struct linux_sched_rr_get_interval_args *uap)
2409 {
2410 	struct timespec ts;
2411 	int error;
2412 
2413 	error = linux_sched_rr_get_interval_common(td, uap->pid, &ts);
2414 	if (error != 0)
2415 		return (error);
2416 	return (linux_put_timespec(&ts, uap->interval));
2417 }
2418 
2419 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
2420 int
2421 linux_sched_rr_get_interval_time64(struct thread *td,
2422     struct linux_sched_rr_get_interval_time64_args *uap)
2423 {
2424 	struct timespec ts;
2425 	int error;
2426 
2427 	error = linux_sched_rr_get_interval_common(td, uap->pid, &ts);
2428 	if (error != 0)
2429 		return (error);
2430 	return (linux_put_timespec64(&ts, uap->interval));
2431 }
2432 #endif
2433 
2434 /*
2435  * In case when the Linux thread is the initial thread in
2436  * the thread group thread id is equal to the process id.
2437  * Glibc depends on this magic (assert in pthread_getattr_np.c).
2438  */
2439 struct thread *
2440 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid)
2441 {
2442 	struct linux_emuldata *em;
2443 	struct thread *tdt;
2444 	struct proc *p;
2445 
2446 	tdt = NULL;
2447 	if (tid == 0 || tid == td->td_tid) {
2448 		if (pid != -1 && td->td_proc->p_pid != pid)
2449 			return (NULL);
2450 		PROC_LOCK(td->td_proc);
2451 		return (td);
2452 	} else if (tid > PID_MAX)
2453 		return (tdfind(tid, pid));
2454 
2455 	/*
2456 	 * Initial thread where the tid equal to the pid.
2457 	 */
2458 	p = pfind(tid);
2459 	if (p != NULL) {
2460 		if (SV_PROC_ABI(p) != SV_ABI_LINUX ||
2461 		    (pid != -1 && tid != pid)) {
2462 			/*
2463 			 * p is not a Linuxulator process.
2464 			 */
2465 			PROC_UNLOCK(p);
2466 			return (NULL);
2467 		}
2468 		FOREACH_THREAD_IN_PROC(p, tdt) {
2469 			em = em_find(tdt);
2470 			if (tid == em->em_tid)
2471 				return (tdt);
2472 		}
2473 		PROC_UNLOCK(p);
2474 	}
2475 	return (NULL);
2476 }
2477 
2478 void
2479 linux_to_bsd_waitopts(int options, int *bsdopts)
2480 {
2481 
2482 	if (options & LINUX_WNOHANG)
2483 		*bsdopts |= WNOHANG;
2484 	if (options & LINUX_WUNTRACED)
2485 		*bsdopts |= WUNTRACED;
2486 	if (options & LINUX_WEXITED)
2487 		*bsdopts |= WEXITED;
2488 	if (options & LINUX_WCONTINUED)
2489 		*bsdopts |= WCONTINUED;
2490 	if (options & LINUX_WNOWAIT)
2491 		*bsdopts |= WNOWAIT;
2492 
2493 	if (options & __WCLONE)
2494 		*bsdopts |= WLINUXCLONE;
2495 }
2496 
2497 int
2498 linux_getrandom(struct thread *td, struct linux_getrandom_args *args)
2499 {
2500 	struct uio uio;
2501 	struct iovec iov;
2502 	int error;
2503 
2504 	if (args->flags & ~(LINUX_GRND_NONBLOCK|LINUX_GRND_RANDOM))
2505 		return (EINVAL);
2506 	if (args->count > INT_MAX)
2507 		args->count = INT_MAX;
2508 
2509 	iov.iov_base = args->buf;
2510 	iov.iov_len = args->count;
2511 
2512 	uio.uio_iov = &iov;
2513 	uio.uio_iovcnt = 1;
2514 	uio.uio_resid = iov.iov_len;
2515 	uio.uio_segflg = UIO_USERSPACE;
2516 	uio.uio_rw = UIO_READ;
2517 	uio.uio_td = td;
2518 
2519 	error = read_random_uio(&uio, args->flags & LINUX_GRND_NONBLOCK);
2520 	if (error == 0)
2521 		td->td_retval[0] = args->count - uio.uio_resid;
2522 	return (error);
2523 }
2524 
2525 int
2526 linux_mincore(struct thread *td, struct linux_mincore_args *args)
2527 {
2528 
2529 	/* Needs to be page-aligned */
2530 	if (args->start & PAGE_MASK)
2531 		return (EINVAL);
2532 	return (kern_mincore(td, args->start, args->len, args->vec));
2533 }
2534 
2535 #define	SYSLOG_TAG	"<6>"
2536 
2537 int
2538 linux_syslog(struct thread *td, struct linux_syslog_args *args)
2539 {
2540 	char buf[128], *src, *dst;
2541 	u_int seq;
2542 	int buflen, error;
2543 
2544 	if (args->type != LINUX_SYSLOG_ACTION_READ_ALL) {
2545 		linux_msg(td, "syslog unsupported type 0x%x", args->type);
2546 		return (EINVAL);
2547 	}
2548 
2549 	if (args->len < 6) {
2550 		td->td_retval[0] = 0;
2551 		return (0);
2552 	}
2553 
2554 	error = priv_check(td, PRIV_MSGBUF);
2555 	if (error)
2556 		return (error);
2557 
2558 	mtx_lock(&msgbuf_lock);
2559 	msgbuf_peekbytes(msgbufp, NULL, 0, &seq);
2560 	mtx_unlock(&msgbuf_lock);
2561 
2562 	dst = args->buf;
2563 	error = copyout(&SYSLOG_TAG, dst, sizeof(SYSLOG_TAG));
2564 	/* The -1 is to skip the trailing '\0'. */
2565 	dst += sizeof(SYSLOG_TAG) - 1;
2566 
2567 	while (error == 0) {
2568 		mtx_lock(&msgbuf_lock);
2569 		buflen = msgbuf_peekbytes(msgbufp, buf, sizeof(buf), &seq);
2570 		mtx_unlock(&msgbuf_lock);
2571 
2572 		if (buflen == 0)
2573 			break;
2574 
2575 		for (src = buf; src < buf + buflen && error == 0; src++) {
2576 			if (*src == '\0')
2577 				continue;
2578 
2579 			if (dst >= args->buf + args->len)
2580 				goto out;
2581 
2582 			error = copyout(src, dst, 1);
2583 			dst++;
2584 
2585 			if (*src == '\n' && *(src + 1) != '<' &&
2586 			    dst + sizeof(SYSLOG_TAG) < args->buf + args->len) {
2587 				error = copyout(&SYSLOG_TAG,
2588 				    dst, sizeof(SYSLOG_TAG));
2589 				dst += sizeof(SYSLOG_TAG) - 1;
2590 			}
2591 		}
2592 	}
2593 out:
2594 	td->td_retval[0] = dst - args->buf;
2595 	return (error);
2596 }
2597 
2598 int
2599 linux_getcpu(struct thread *td, struct linux_getcpu_args *args)
2600 {
2601 	int cpu, error, node;
2602 
2603 	cpu = td->td_oncpu; /* Make sure it doesn't change during copyout(9) */
2604 	error = 0;
2605 	node = cpuid_to_pcpu[cpu]->pc_domain;
2606 
2607 	if (args->cpu != NULL)
2608 		error = copyout(&cpu, args->cpu, sizeof(l_int));
2609 	if (args->node != NULL)
2610 		error = copyout(&node, args->node, sizeof(l_int));
2611 	return (error);
2612 }
2613 
2614 #if defined(__i386__) || defined(__amd64__)
2615 int
2616 linux_poll(struct thread *td, struct linux_poll_args *args)
2617 {
2618 	struct timespec ts, *tsp;
2619 
2620 	if (args->timeout != INFTIM) {
2621 		if (args->timeout < 0)
2622 			return (EINVAL);
2623 		ts.tv_sec = args->timeout / 1000;
2624 		ts.tv_nsec = (args->timeout % 1000) * 1000000;
2625 		tsp = &ts;
2626 	} else
2627 		tsp = NULL;
2628 
2629 	return (linux_common_ppoll(td, args->fds, args->nfds,
2630 	    tsp, NULL, 0));
2631 }
2632 #endif /* __i386__ || __amd64__ */
2633 
2634 int
2635 linux_seccomp(struct thread *td, struct linux_seccomp_args *args)
2636 {
2637 
2638 	switch (args->op) {
2639 	case LINUX_SECCOMP_GET_ACTION_AVAIL:
2640 		return (EOPNOTSUPP);
2641 	default:
2642 		/*
2643 		 * Ignore unknown operations, just like Linux kernel built
2644 		 * without CONFIG_SECCOMP.
2645 		 */
2646 		return (EINVAL);
2647 	}
2648 }
2649 
2650 /*
2651  * Custom version of exec_copyin_args(), to copy out argument and environment
2652  * strings from the old process address space into the temporary string buffer.
2653  * Based on freebsd32_exec_copyin_args.
2654  */
2655 static int
2656 linux_exec_copyin_args(struct image_args *args, const char *fname,
2657     l_uintptr_t *argv, l_uintptr_t *envv)
2658 {
2659 	char *argp, *envp;
2660 	l_uintptr_t *ptr, arg;
2661 	int error;
2662 
2663 	bzero(args, sizeof(*args));
2664 	if (argv == NULL)
2665 		return (EFAULT);
2666 
2667 	/*
2668 	 * Allocate demand-paged memory for the file name, argument, and
2669 	 * environment strings.
2670 	 */
2671 	error = exec_alloc_args(args);
2672 	if (error != 0)
2673 		return (error);
2674 
2675 	/*
2676 	 * Copy the file name.
2677 	 */
2678 	error = exec_args_add_fname(args, fname, UIO_USERSPACE);
2679 	if (error != 0)
2680 		goto err_exit;
2681 
2682 	/*
2683 	 * extract arguments first
2684 	 */
2685 	ptr = argv;
2686 	for (;;) {
2687 		error = copyin(ptr++, &arg, sizeof(arg));
2688 		if (error)
2689 			goto err_exit;
2690 		if (arg == 0)
2691 			break;
2692 		argp = PTRIN(arg);
2693 		error = exec_args_add_arg(args, argp, UIO_USERSPACE);
2694 		if (error != 0)
2695 			goto err_exit;
2696 	}
2697 
2698 	/*
2699 	 * This comment is from Linux do_execveat_common:
2700 	 * When argv is empty, add an empty string ("") as argv[0] to
2701 	 * ensure confused userspace programs that start processing
2702 	 * from argv[1] won't end up walking envp.
2703 	 */
2704 	if (args->argc == 0 &&
2705 	    (error = exec_args_add_arg(args, "", UIO_SYSSPACE) != 0))
2706 		goto err_exit;
2707 
2708 	/*
2709 	 * extract environment strings
2710 	 */
2711 	if (envv) {
2712 		ptr = envv;
2713 		for (;;) {
2714 			error = copyin(ptr++, &arg, sizeof(arg));
2715 			if (error)
2716 				goto err_exit;
2717 			if (arg == 0)
2718 				break;
2719 			envp = PTRIN(arg);
2720 			error = exec_args_add_env(args, envp, UIO_USERSPACE);
2721 			if (error != 0)
2722 				goto err_exit;
2723 		}
2724 	}
2725 
2726 	return (0);
2727 
2728 err_exit:
2729 	exec_free_args(args);
2730 	return (error);
2731 }
2732 
2733 int
2734 linux_execve(struct thread *td, struct linux_execve_args *args)
2735 {
2736 	struct image_args eargs;
2737 	int error;
2738 
2739 	LINUX_CTR(execve);
2740 
2741 	error = linux_exec_copyin_args(&eargs, args->path, args->argp,
2742 	    args->envp);
2743 	if (error == 0)
2744 		error = linux_common_execve(td, &eargs);
2745 	AUDIT_SYSCALL_EXIT(error == EJUSTRETURN ? 0 : error, td);
2746 	return (error);
2747 }
2748 
2749 static void
2750 linux_up_rtprio_if(struct thread *td1, struct rtprio *rtp)
2751 {
2752 	struct rtprio rtp2;
2753 
2754 	pri_to_rtp(td1, &rtp2);
2755 	if (rtp2.type <  rtp->type ||
2756 	    (rtp2.type == rtp->type &&
2757 	    rtp2.prio < rtp->prio)) {
2758 		rtp->type = rtp2.type;
2759 		rtp->prio = rtp2.prio;
2760 	}
2761 }
2762 
2763 #define	LINUX_PRIO_DIVIDER	RTP_PRIO_MAX / LINUX_IOPRIO_MAX
2764 
2765 static int
2766 linux_rtprio2ioprio(struct rtprio *rtp)
2767 {
2768 	int ioprio, prio;
2769 
2770 	switch (rtp->type) {
2771 	case RTP_PRIO_IDLE:
2772 		prio = RTP_PRIO_MIN;
2773 		ioprio = LINUX_IOPRIO_PRIO(LINUX_IOPRIO_CLASS_IDLE, prio);
2774 		break;
2775 	case RTP_PRIO_NORMAL:
2776 		prio = rtp->prio / LINUX_PRIO_DIVIDER;
2777 		ioprio = LINUX_IOPRIO_PRIO(LINUX_IOPRIO_CLASS_BE, prio);
2778 		break;
2779 	case RTP_PRIO_REALTIME:
2780 		prio = rtp->prio / LINUX_PRIO_DIVIDER;
2781 		ioprio = LINUX_IOPRIO_PRIO(LINUX_IOPRIO_CLASS_RT, prio);
2782 		break;
2783 	default:
2784 		prio = RTP_PRIO_MIN;
2785 		ioprio = LINUX_IOPRIO_PRIO(LINUX_IOPRIO_CLASS_NONE, prio);
2786 		break;
2787 	}
2788 	return (ioprio);
2789 }
2790 
2791 static int
2792 linux_ioprio2rtprio(int ioprio, struct rtprio *rtp)
2793 {
2794 
2795 	switch (LINUX_IOPRIO_PRIO_CLASS(ioprio)) {
2796 	case LINUX_IOPRIO_CLASS_IDLE:
2797 		rtp->prio = RTP_PRIO_MIN;
2798 		rtp->type = RTP_PRIO_IDLE;
2799 		break;
2800 	case LINUX_IOPRIO_CLASS_BE:
2801 		rtp->prio = LINUX_IOPRIO_PRIO_DATA(ioprio) * LINUX_PRIO_DIVIDER;
2802 		rtp->type = RTP_PRIO_NORMAL;
2803 		break;
2804 	case LINUX_IOPRIO_CLASS_RT:
2805 		rtp->prio = LINUX_IOPRIO_PRIO_DATA(ioprio) * LINUX_PRIO_DIVIDER;
2806 		rtp->type = RTP_PRIO_REALTIME;
2807 		break;
2808 	default:
2809 		return (EINVAL);
2810 	}
2811 	return (0);
2812 }
2813 #undef LINUX_PRIO_DIVIDER
2814 
2815 int
2816 linux_ioprio_get(struct thread *td, struct linux_ioprio_get_args *args)
2817 {
2818 	struct thread *td1;
2819 	struct rtprio rtp;
2820 	struct pgrp *pg;
2821 	struct proc *p;
2822 	int error, found;
2823 
2824 	p = NULL;
2825 	td1 = NULL;
2826 	error = 0;
2827 	found = 0;
2828 	rtp.type = RTP_PRIO_IDLE;
2829 	rtp.prio = RTP_PRIO_MAX;
2830 	switch (args->which) {
2831 	case LINUX_IOPRIO_WHO_PROCESS:
2832 		if (args->who == 0) {
2833 			td1 = td;
2834 			p = td1->td_proc;
2835 			PROC_LOCK(p);
2836 		} else if (args->who > PID_MAX) {
2837 			td1 = linux_tdfind(td, args->who, -1);
2838 			if (td1 != NULL)
2839 				p = td1->td_proc;
2840 		} else
2841 			p = pfind(args->who);
2842 		if (p == NULL)
2843 			return (ESRCH);
2844 		if ((error = p_cansee(td, p))) {
2845 			PROC_UNLOCK(p);
2846 			break;
2847 		}
2848 		if (td1 != NULL) {
2849 			pri_to_rtp(td1, &rtp);
2850 		} else {
2851 			FOREACH_THREAD_IN_PROC(p, td1) {
2852 				linux_up_rtprio_if(td1, &rtp);
2853 			}
2854 		}
2855 		found++;
2856 		PROC_UNLOCK(p);
2857 		break;
2858 	case LINUX_IOPRIO_WHO_PGRP:
2859 		sx_slock(&proctree_lock);
2860 		if (args->who == 0) {
2861 			pg = td->td_proc->p_pgrp;
2862 			PGRP_LOCK(pg);
2863 		} else {
2864 			pg = pgfind(args->who);
2865 			if (pg == NULL) {
2866 				sx_sunlock(&proctree_lock);
2867 				error = ESRCH;
2868 				break;
2869 			}
2870 		}
2871 		sx_sunlock(&proctree_lock);
2872 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
2873 			PROC_LOCK(p);
2874 			if (p->p_state == PRS_NORMAL &&
2875 			    p_cansee(td, p) == 0) {
2876 				FOREACH_THREAD_IN_PROC(p, td1) {
2877 					linux_up_rtprio_if(td1, &rtp);
2878 					found++;
2879 				}
2880 			}
2881 			PROC_UNLOCK(p);
2882 		}
2883 		PGRP_UNLOCK(pg);
2884 		break;
2885 	case LINUX_IOPRIO_WHO_USER:
2886 		if (args->who == 0)
2887 			args->who = td->td_ucred->cr_uid;
2888 		sx_slock(&allproc_lock);
2889 		FOREACH_PROC_IN_SYSTEM(p) {
2890 			PROC_LOCK(p);
2891 			if (p->p_state == PRS_NORMAL &&
2892 			    p->p_ucred->cr_uid == args->who &&
2893 			    p_cansee(td, p) == 0) {
2894 				FOREACH_THREAD_IN_PROC(p, td1) {
2895 					linux_up_rtprio_if(td1, &rtp);
2896 					found++;
2897 				}
2898 			}
2899 			PROC_UNLOCK(p);
2900 		}
2901 		sx_sunlock(&allproc_lock);
2902 		break;
2903 	default:
2904 		error = EINVAL;
2905 		break;
2906 	}
2907 	if (error == 0) {
2908 		if (found != 0)
2909 			td->td_retval[0] = linux_rtprio2ioprio(&rtp);
2910 		else
2911 			error = ESRCH;
2912 	}
2913 	return (error);
2914 }
2915 
2916 int
2917 linux_ioprio_set(struct thread *td, struct linux_ioprio_set_args *args)
2918 {
2919 	struct thread *td1;
2920 	struct rtprio rtp;
2921 	struct pgrp *pg;
2922 	struct proc *p;
2923 	int error;
2924 
2925 	if ((error = linux_ioprio2rtprio(args->ioprio, &rtp)) != 0)
2926 		return (error);
2927 	/* Attempts to set high priorities (REALTIME) require su privileges. */
2928 	if (RTP_PRIO_BASE(rtp.type) == RTP_PRIO_REALTIME &&
2929 	    (error = priv_check(td, PRIV_SCHED_RTPRIO)) != 0)
2930 		return (error);
2931 
2932 	p = NULL;
2933 	td1 = NULL;
2934 	switch (args->which) {
2935 	case LINUX_IOPRIO_WHO_PROCESS:
2936 		if (args->who == 0) {
2937 			td1 = td;
2938 			p = td1->td_proc;
2939 			PROC_LOCK(p);
2940 		} else if (args->who > PID_MAX) {
2941 			td1 = linux_tdfind(td, args->who, -1);
2942 			if (td1 != NULL)
2943 				p = td1->td_proc;
2944 		} else
2945 			p = pfind(args->who);
2946 		if (p == NULL)
2947 			return (ESRCH);
2948 		if ((error = p_cansched(td, p))) {
2949 			PROC_UNLOCK(p);
2950 			break;
2951 		}
2952 		if (td1 != NULL) {
2953 			error = rtp_to_pri(&rtp, td1);
2954 		} else {
2955 			FOREACH_THREAD_IN_PROC(p, td1) {
2956 				if ((error = rtp_to_pri(&rtp, td1)) != 0)
2957 					break;
2958 			}
2959 		}
2960 		PROC_UNLOCK(p);
2961 		break;
2962 	case LINUX_IOPRIO_WHO_PGRP:
2963 		sx_slock(&proctree_lock);
2964 		if (args->who == 0) {
2965 			pg = td->td_proc->p_pgrp;
2966 			PGRP_LOCK(pg);
2967 		} else {
2968 			pg = pgfind(args->who);
2969 			if (pg == NULL) {
2970 				sx_sunlock(&proctree_lock);
2971 				error = ESRCH;
2972 				break;
2973 			}
2974 		}
2975 		sx_sunlock(&proctree_lock);
2976 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
2977 			PROC_LOCK(p);
2978 			if (p->p_state == PRS_NORMAL &&
2979 			    p_cansched(td, p) == 0) {
2980 				FOREACH_THREAD_IN_PROC(p, td1) {
2981 					if ((error = rtp_to_pri(&rtp, td1)) != 0)
2982 						break;
2983 				}
2984 			}
2985 			PROC_UNLOCK(p);
2986 			if (error != 0)
2987 				break;
2988 		}
2989 		PGRP_UNLOCK(pg);
2990 		break;
2991 	case LINUX_IOPRIO_WHO_USER:
2992 		if (args->who == 0)
2993 			args->who = td->td_ucred->cr_uid;
2994 		sx_slock(&allproc_lock);
2995 		FOREACH_PROC_IN_SYSTEM(p) {
2996 			PROC_LOCK(p);
2997 			if (p->p_state == PRS_NORMAL &&
2998 			    p->p_ucred->cr_uid == args->who &&
2999 			    p_cansched(td, p) == 0) {
3000 				FOREACH_THREAD_IN_PROC(p, td1) {
3001 					if ((error = rtp_to_pri(&rtp, td1)) != 0)
3002 						break;
3003 				}
3004 			}
3005 			PROC_UNLOCK(p);
3006 			if (error != 0)
3007 				break;
3008 		}
3009 		sx_sunlock(&allproc_lock);
3010 		break;
3011 	default:
3012 		error = EINVAL;
3013 		break;
3014 	}
3015 	return (error);
3016 }
3017 
3018 /* The only flag is O_NONBLOCK */
3019 #define B2L_MQ_FLAGS(bflags)	((bflags) != 0 ? LINUX_O_NONBLOCK : 0)
3020 #define L2B_MQ_FLAGS(lflags)	((lflags) != 0 ? O_NONBLOCK : 0)
3021 
3022 int
3023 linux_mq_open(struct thread *td, struct linux_mq_open_args *args)
3024 {
3025 	struct mq_attr attr;
3026 	int error, flags;
3027 
3028 	flags = linux_common_openflags(args->oflag);
3029 	if ((flags & O_ACCMODE) == O_ACCMODE || (flags & O_EXEC) != 0)
3030 		return (EINVAL);
3031 	flags = FFLAGS(flags);
3032 	if ((flags & O_CREAT) != 0 && args->attr != NULL) {
3033 		error = copyin(args->attr, &attr, sizeof(attr));
3034 		if (error != 0)
3035 			return (error);
3036 		attr.mq_flags = L2B_MQ_FLAGS(attr.mq_flags);
3037 	}
3038 
3039 	return (kern_kmq_open(td, args->name, flags, args->mode,
3040 	    args->attr != NULL ? &attr : NULL));
3041 }
3042 
3043 int
3044 linux_mq_unlink(struct thread *td, struct linux_mq_unlink_args *args)
3045 {
3046 	struct kmq_unlink_args bsd_args = {
3047 		.path = PTRIN(args->name)
3048 	};
3049 
3050 	return (sys_kmq_unlink(td, &bsd_args));
3051 }
3052 
3053 int
3054 linux_mq_timedsend(struct thread *td, struct linux_mq_timedsend_args *args)
3055 {
3056 	struct timespec ts, *abs_timeout;
3057 	int error;
3058 
3059 	if (args->abs_timeout == NULL)
3060 		abs_timeout = NULL;
3061 	else {
3062 		error = linux_get_timespec(&ts, args->abs_timeout);
3063 		if (error != 0)
3064 			return (error);
3065 		abs_timeout = &ts;
3066 	}
3067 
3068 	return (kern_kmq_timedsend(td, args->mqd, PTRIN(args->msg_ptr),
3069 		args->msg_len, args->msg_prio, abs_timeout));
3070 }
3071 
3072 int
3073 linux_mq_timedreceive(struct thread *td, struct linux_mq_timedreceive_args *args)
3074 {
3075 	struct timespec ts, *abs_timeout;
3076 	int error;
3077 
3078 	if (args->abs_timeout == NULL)
3079 		abs_timeout = NULL;
3080 	else {
3081 		error = linux_get_timespec(&ts, args->abs_timeout);
3082 		if (error != 0)
3083 			return (error);
3084 		abs_timeout = &ts;
3085 	}
3086 
3087 	return (kern_kmq_timedreceive(td, args->mqd, PTRIN(args->msg_ptr),
3088 		args->msg_len, args->msg_prio, abs_timeout));
3089 }
3090 
3091 int
3092 linux_mq_notify(struct thread *td, struct linux_mq_notify_args *args)
3093 {
3094 	struct sigevent ev, *evp;
3095 	struct l_sigevent l_ev;
3096 	int error;
3097 
3098 	if (args->sevp == NULL)
3099 		evp = NULL;
3100 	else {
3101 		error = copyin(args->sevp, &l_ev, sizeof(l_ev));
3102 		if (error != 0)
3103 			return (error);
3104 		error = linux_convert_l_sigevent(&l_ev, &ev);
3105 		if (error != 0)
3106 			return (error);
3107 		evp = &ev;
3108 	}
3109 
3110 	return (kern_kmq_notify(td, args->mqd, evp));
3111 }
3112 
3113 int
3114 linux_mq_getsetattr(struct thread *td, struct linux_mq_getsetattr_args *args)
3115 {
3116 	struct mq_attr attr, oattr;
3117 	int error;
3118 
3119 	if (args->attr != NULL) {
3120 		error = copyin(args->attr, &attr, sizeof(attr));
3121 		if (error != 0)
3122 			return (error);
3123 		attr.mq_flags = L2B_MQ_FLAGS(attr.mq_flags);
3124 	}
3125 
3126 	error = kern_kmq_setattr(td, args->mqd, args->attr != NULL ? &attr : NULL,
3127 	    &oattr);
3128 	if (error == 0 && args->oattr != NULL) {
3129 		oattr.mq_flags = B2L_MQ_FLAGS(oattr.mq_flags);
3130 		bzero(oattr.__reserved, sizeof(oattr.__reserved));
3131 		error = copyout(&oattr, args->oattr, sizeof(oattr));
3132 	}
3133 
3134 	return (error);
3135 }
3136 
3137 int
3138 linux_kcmp(struct thread *td, struct linux_kcmp_args *args)
3139 {
3140 	int type;
3141 
3142 	switch (args->type) {
3143 	case LINUX_KCMP_FILE:
3144 		type = KCMP_FILE;
3145 		break;
3146 	case LINUX_KCMP_FILES:
3147 		type = KCMP_FILES;
3148 		break;
3149 	case LINUX_KCMP_SIGHAND:
3150 		type = KCMP_SIGHAND;
3151 		break;
3152 	case LINUX_KCMP_VM:
3153 		type = KCMP_VM;
3154 		break;
3155 	default:
3156 		return (EINVAL);
3157 	}
3158 
3159 	return (kern_kcmp(td, args->pid1, args->pid2, type, args->idx1,
3160 	    args->idx));
3161 }
3162 
3163 int
3164 linux_membarrier(struct thread *td, struct linux_membarrier_args *args)
3165 {
3166 	static const struct {
3167 		int linux_cmd;
3168 		int freebsd_cmd;
3169 	} cmds[] = {
3170 		{ LINUX_MEMBARRIER_CMD_QUERY,
3171 		    MEMBARRIER_CMD_QUERY },
3172 		{ LINUX_MEMBARRIER_CMD_GLOBAL,
3173 		    MEMBARRIER_CMD_GLOBAL },
3174 		{ LINUX_MEMBARRIER_CMD_GLOBAL_EXPEDITED,
3175 		    MEMBARRIER_CMD_GLOBAL_EXPEDITED },
3176 		{ LINUX_MEMBARRIER_CMD_REGISTER_GLOBAL_EXPEDITED,
3177 		    MEMBARRIER_CMD_REGISTER_GLOBAL_EXPEDITED },
3178 		{ LINUX_MEMBARRIER_CMD_PRIVATE_EXPEDITED,
3179 		    MEMBARRIER_CMD_PRIVATE_EXPEDITED },
3180 		{ LINUX_MEMBARRIER_CMD_REGISTER_PRIVATE_EXPEDITED,
3181 		    MEMBARRIER_CMD_REGISTER_PRIVATE_EXPEDITED },
3182 		{ LINUX_MEMBARRIER_CMD_PRIVATE_EXPEDITED_SYNC_CORE,
3183 		    MEMBARRIER_CMD_PRIVATE_EXPEDITED_SYNC_CORE },
3184 		{ LINUX_MEMBARRIER_CMD_REGISTER_PRIVATE_EXPEDITED_SYNC_CORE,
3185 		    MEMBARRIER_CMD_REGISTER_PRIVATE_EXPEDITED_SYNC_CORE },
3186 		{ LINUX_MEMBARRIER_CMD_PRIVATE_EXPEDITED_RSEQ,
3187 		    MEMBARRIER_CMD_PRIVATE_EXPEDITED_RSEQ },
3188 		{ LINUX_MEMBARRIER_CMD_REGISTER_PRIVATE_EXPEDITED_RSEQ,
3189 		    MEMBARRIER_CMD_REGISTER_PRIVATE_EXPEDITED_RSEQ },
3190 		{ LINUX_MEMBARRIER_CMD_GET_REGISTRATIONS,
3191 		    MEMBARRIER_CMD_GET_REGISTRATIONS },
3192 	};
3193 	int cmd, error, flags, i, mask;
3194 
3195 	cmd = -1;
3196 	for (i = 0; i < nitems(cmds); i++) {
3197 		if (args->cmd == cmds[i].linux_cmd) {
3198 			cmd = cmds[i].freebsd_cmd;
3199 			break;
3200 		}
3201 	}
3202 
3203 	if (cmd == -1 || (args->flags & ~LINUX_MEMBARRIER_CMD_FLAG_CPU) != 0)
3204 		return (EINVAL);
3205 
3206 	flags = 0;
3207 	if ((args->flags & LINUX_MEMBARRIER_CMD_FLAG_CPU) != 0)
3208 		flags |= MEMBARRIER_CMD_FLAG_CPU;
3209 
3210 	error = kern_membarrier(td, cmd, flags, args->cpu_id);
3211 	if (error != 0)
3212 		return (error);
3213 
3214 	if (args->cmd == LINUX_MEMBARRIER_CMD_QUERY ||
3215 	    args->cmd == LINUX_MEMBARRIER_CMD_GET_REGISTRATIONS) {
3216 		mask = td->td_retval[0];
3217 		td->td_retval[0] = 0;
3218 		for (i = 0; i < nitems(cmds); i++)
3219 			if ((mask & cmds[i].freebsd_cmd) != 0)
3220 				td->td_retval[0] |= cmds[i].linux_cmd;
3221 	}
3222 
3223 	return (0);
3224 }
3225 
3226 /*
3227  * setfsuid() & setfsgid() exist to decouple the Linux filesystem credentials
3228  * from the effective credentials, avoiding signal exposure during privilege
3229  * transitions. The signal permission model that motivated this was revised in
3230  * Linux 2.0, making these syscalls obsolete for new applications.
3231  *
3232  * As there's no FreeBSD equivalent, implement both syscalls as no-ops that
3233  * return the current effective UID/GID as the previous filesystem UID/GID.
3234  * Linux returns the previous filesystem UID/GID for these syscalls, with no
3235  * error indication.
3236  */
3237 
3238 int
3239 linux_setfsuid(struct thread *td, struct linux_setfsuid_args *args)
3240 {
3241 	td->td_retval[0] = td->td_ucred->cr_uid;
3242 	return (0);
3243 }
3244 
3245 int
3246 linux_setfsgid(struct thread *td, struct linux_setfsgid_args *args)
3247 {
3248 	td->td_retval[0] = td->td_ucred->cr_gid;
3249 	return (0);
3250 }
3251 
3252 MODULE_DEPEND(linux, mqueuefs, 1, 1, 1);
3253