xref: /freebsd/sys/compat/linux/linux_misc.c (revision 5ca8e32633c4ffbbcd6762e5888b6a4ba0708c6c)
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/msgbuf.h>
39 #include <sys/mutex.h>
40 #include <sys/poll.h>
41 #include <sys/priv.h>
42 #include <sys/proc.h>
43 #include <sys/procctl.h>
44 #include <sys/reboot.h>
45 #include <sys/random.h>
46 #include <sys/resourcevar.h>
47 #include <sys/rtprio.h>
48 #include <sys/sched.h>
49 #include <sys/smp.h>
50 #include <sys/stat.h>
51 #include <sys/syscallsubr.h>
52 #include <sys/sysctl.h>
53 #include <sys/sysent.h>
54 #include <sys/sysproto.h>
55 #include <sys/time.h>
56 #include <sys/vmmeter.h>
57 #include <sys/vnode.h>
58 
59 #include <security/audit/audit.h>
60 #include <security/mac/mac_framework.h>
61 
62 #include <vm/pmap.h>
63 #include <vm/vm_map.h>
64 #include <vm/swap_pager.h>
65 
66 #ifdef COMPAT_LINUX32
67 #include <machine/../linux32/linux.h>
68 #include <machine/../linux32/linux32_proto.h>
69 #else
70 #include <machine/../linux/linux.h>
71 #include <machine/../linux/linux_proto.h>
72 #endif
73 
74 #include <compat/linux/linux_common.h>
75 #include <compat/linux/linux_dtrace.h>
76 #include <compat/linux/linux_file.h>
77 #include <compat/linux/linux_mib.h>
78 #include <compat/linux/linux_mmap.h>
79 #include <compat/linux/linux_signal.h>
80 #include <compat/linux/linux_time.h>
81 #include <compat/linux/linux_util.h>
82 #include <compat/linux/linux_emul.h>
83 #include <compat/linux/linux_misc.h>
84 
85 int stclohz;				/* Statistics clock frequency */
86 
87 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
88 	RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
89 	RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
90 	RLIMIT_MEMLOCK, RLIMIT_AS
91 };
92 
93 struct l_sysinfo {
94 	l_long		uptime;		/* Seconds since boot */
95 	l_ulong		loads[3];	/* 1, 5, and 15 minute load averages */
96 #define LINUX_SYSINFO_LOADS_SCALE 65536
97 	l_ulong		totalram;	/* Total usable main memory size */
98 	l_ulong		freeram;	/* Available memory size */
99 	l_ulong		sharedram;	/* Amount of shared memory */
100 	l_ulong		bufferram;	/* Memory used by buffers */
101 	l_ulong		totalswap;	/* Total swap space size */
102 	l_ulong		freeswap;	/* swap space still available */
103 	l_ushort	procs;		/* Number of current processes */
104 	l_ushort	pads;
105 	l_ulong		totalhigh;
106 	l_ulong		freehigh;
107 	l_uint		mem_unit;
108 	char		_f[20-2*sizeof(l_long)-sizeof(l_int)];	/* padding */
109 };
110 
111 struct l_pselect6arg {
112 	l_uintptr_t	ss;
113 	l_size_t	ss_len;
114 };
115 
116 static int	linux_utimensat_lts_to_ts(struct l_timespec *,
117 			struct timespec *);
118 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
119 static int	linux_utimensat_lts64_to_ts(struct l_timespec64 *,
120 			struct timespec *);
121 #endif
122 static int	linux_common_utimensat(struct thread *, int,
123 			const char *, struct timespec *, int);
124 static int	linux_common_pselect6(struct thread *, l_int,
125 			l_fd_set *, l_fd_set *, l_fd_set *,
126 			struct timespec *, l_uintptr_t *);
127 static int	linux_common_ppoll(struct thread *, struct pollfd *,
128 			uint32_t, struct timespec *, l_sigset_t *,
129 			l_size_t);
130 static int	linux_pollin(struct thread *, struct pollfd *,
131 			struct pollfd *, u_int);
132 static int	linux_pollout(struct thread *, struct pollfd *,
133 			struct pollfd *, u_int);
134 
135 int
136 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
137 {
138 	struct l_sysinfo sysinfo;
139 	int i, j;
140 	struct timespec ts;
141 
142 	bzero(&sysinfo, sizeof(sysinfo));
143 	getnanouptime(&ts);
144 	if (ts.tv_nsec != 0)
145 		ts.tv_sec++;
146 	sysinfo.uptime = ts.tv_sec;
147 
148 	/* Use the information from the mib to get our load averages */
149 	for (i = 0; i < 3; i++)
150 		sysinfo.loads[i] = averunnable.ldavg[i] *
151 		    LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
152 
153 	sysinfo.totalram = physmem * PAGE_SIZE;
154 	sysinfo.freeram = (u_long)vm_free_count() * PAGE_SIZE;
155 
156 	/*
157 	 * sharedram counts pages allocated to named, swap-backed objects such
158 	 * as shared memory segments and tmpfs files.  There is no cheap way to
159 	 * compute this, so just leave the field unpopulated.  Linux itself only
160 	 * started setting this field in the 3.x timeframe.
161 	 */
162 	sysinfo.sharedram = 0;
163 	sysinfo.bufferram = 0;
164 
165 	swap_pager_status(&i, &j);
166 	sysinfo.totalswap = i * PAGE_SIZE;
167 	sysinfo.freeswap = (i - j) * PAGE_SIZE;
168 
169 	sysinfo.procs = nprocs;
170 
171 	/*
172 	 * Platforms supported by the emulation layer do not have a notion of
173 	 * high memory.
174 	 */
175 	sysinfo.totalhigh = 0;
176 	sysinfo.freehigh = 0;
177 
178 	sysinfo.mem_unit = 1;
179 
180 	return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
181 }
182 
183 #ifdef LINUX_LEGACY_SYSCALLS
184 int
185 linux_alarm(struct thread *td, struct linux_alarm_args *args)
186 {
187 	struct itimerval it, old_it;
188 	u_int secs;
189 	int error __diagused;
190 
191 	secs = args->secs;
192 	/*
193 	 * Linux alarm() is always successful. Limit secs to INT32_MAX / 2
194 	 * to match kern_setitimer()'s limit to avoid error from it.
195 	 *
196 	 * XXX. Linux limit secs to INT_MAX on 32 and does not limit on 64-bit
197 	 * platforms.
198 	 */
199 	if (secs > INT32_MAX / 2)
200 		secs = INT32_MAX / 2;
201 
202 	it.it_value.tv_sec = secs;
203 	it.it_value.tv_usec = 0;
204 	timevalclear(&it.it_interval);
205 	error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
206 	KASSERT(error == 0, ("kern_setitimer returns %d", error));
207 
208 	if ((old_it.it_value.tv_sec == 0 && old_it.it_value.tv_usec > 0) ||
209 	    old_it.it_value.tv_usec >= 500000)
210 		old_it.it_value.tv_sec++;
211 	td->td_retval[0] = old_it.it_value.tv_sec;
212 	return (0);
213 }
214 #endif
215 
216 int
217 linux_brk(struct thread *td, struct linux_brk_args *args)
218 {
219 	struct vmspace *vm = td->td_proc->p_vmspace;
220 	uintptr_t new, old;
221 
222 	old = (uintptr_t)vm->vm_daddr + ctob(vm->vm_dsize);
223 	new = (uintptr_t)args->dsend;
224 	if ((caddr_t)new > vm->vm_daddr && !kern_break(td, &new))
225 		td->td_retval[0] = (register_t)new;
226 	else
227 		td->td_retval[0] = (register_t)old;
228 
229 	return (0);
230 }
231 
232 #ifdef LINUX_LEGACY_SYSCALLS
233 int
234 linux_select(struct thread *td, struct linux_select_args *args)
235 {
236 	l_timeval ltv;
237 	struct timeval tv0, tv1, utv, *tvp;
238 	int error;
239 
240 	/*
241 	 * Store current time for computation of the amount of
242 	 * time left.
243 	 */
244 	if (args->timeout) {
245 		if ((error = copyin(args->timeout, &ltv, sizeof(ltv))))
246 			goto select_out;
247 		utv.tv_sec = ltv.tv_sec;
248 		utv.tv_usec = ltv.tv_usec;
249 
250 		if (itimerfix(&utv)) {
251 			/*
252 			 * The timeval was invalid.  Convert it to something
253 			 * valid that will act as it does under Linux.
254 			 */
255 			utv.tv_sec += utv.tv_usec / 1000000;
256 			utv.tv_usec %= 1000000;
257 			if (utv.tv_usec < 0) {
258 				utv.tv_sec -= 1;
259 				utv.tv_usec += 1000000;
260 			}
261 			if (utv.tv_sec < 0)
262 				timevalclear(&utv);
263 		}
264 		microtime(&tv0);
265 		tvp = &utv;
266 	} else
267 		tvp = NULL;
268 
269 	error = kern_select(td, args->nfds, args->readfds, args->writefds,
270 	    args->exceptfds, tvp, LINUX_NFDBITS);
271 	if (error)
272 		goto select_out;
273 
274 	if (args->timeout) {
275 		if (td->td_retval[0]) {
276 			/*
277 			 * Compute how much time was left of the timeout,
278 			 * by subtracting the current time and the time
279 			 * before we started the call, and subtracting
280 			 * that result from the user-supplied value.
281 			 */
282 			microtime(&tv1);
283 			timevalsub(&tv1, &tv0);
284 			timevalsub(&utv, &tv1);
285 			if (utv.tv_sec < 0)
286 				timevalclear(&utv);
287 		} else
288 			timevalclear(&utv);
289 		ltv.tv_sec = utv.tv_sec;
290 		ltv.tv_usec = utv.tv_usec;
291 		if ((error = copyout(&ltv, args->timeout, sizeof(ltv))))
292 			goto select_out;
293 	}
294 
295 select_out:
296 	return (error);
297 }
298 #endif
299 
300 int
301 linux_mremap(struct thread *td, struct linux_mremap_args *args)
302 {
303 	uintptr_t addr;
304 	size_t len;
305 	int error = 0;
306 
307 	if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
308 		td->td_retval[0] = 0;
309 		return (EINVAL);
310 	}
311 
312 	/*
313 	 * Check for the page alignment.
314 	 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
315 	 */
316 	if (args->addr & PAGE_MASK) {
317 		td->td_retval[0] = 0;
318 		return (EINVAL);
319 	}
320 
321 	args->new_len = round_page(args->new_len);
322 	args->old_len = round_page(args->old_len);
323 
324 	if (args->new_len > args->old_len) {
325 		td->td_retval[0] = 0;
326 		return (ENOMEM);
327 	}
328 
329 	if (args->new_len < args->old_len) {
330 		addr = args->addr + args->new_len;
331 		len = args->old_len - args->new_len;
332 		error = kern_munmap(td, addr, len);
333 	}
334 
335 	td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
336 	return (error);
337 }
338 
339 #define LINUX_MS_ASYNC       0x0001
340 #define LINUX_MS_INVALIDATE  0x0002
341 #define LINUX_MS_SYNC        0x0004
342 
343 int
344 linux_msync(struct thread *td, struct linux_msync_args *args)
345 {
346 
347 	return (kern_msync(td, args->addr, args->len,
348 	    args->fl & ~LINUX_MS_SYNC));
349 }
350 
351 int
352 linux_mprotect(struct thread *td, struct linux_mprotect_args *uap)
353 {
354 
355 	return (linux_mprotect_common(td, PTROUT(uap->addr), uap->len,
356 	    uap->prot));
357 }
358 
359 int
360 linux_madvise(struct thread *td, struct linux_madvise_args *uap)
361 {
362 
363 	return (linux_madvise_common(td, PTROUT(uap->addr), uap->len,
364 	    uap->behav));
365 }
366 
367 int
368 linux_mmap2(struct thread *td, struct linux_mmap2_args *uap)
369 {
370 #if defined(LINUX_ARCHWANT_MMAP2PGOFF)
371 	/*
372 	 * For architectures with sizeof (off_t) < sizeof (loff_t) mmap is
373 	 * implemented with mmap2 syscall and the offset is represented in
374 	 * multiples of page size.
375 	 */
376 	return (linux_mmap_common(td, PTROUT(uap->addr), uap->len, uap->prot,
377 	    uap->flags, uap->fd, (uint64_t)(uint32_t)uap->pgoff * PAGE_SIZE));
378 #else
379 	return (linux_mmap_common(td, PTROUT(uap->addr), uap->len, uap->prot,
380 	    uap->flags, uap->fd, uap->pgoff));
381 #endif
382 }
383 
384 #ifdef LINUX_LEGACY_SYSCALLS
385 int
386 linux_time(struct thread *td, struct linux_time_args *args)
387 {
388 	struct timeval tv;
389 	l_time_t tm;
390 	int error;
391 
392 	microtime(&tv);
393 	tm = tv.tv_sec;
394 	if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
395 		return (error);
396 	td->td_retval[0] = tm;
397 	return (0);
398 }
399 #endif
400 
401 struct l_times_argv {
402 	l_clock_t	tms_utime;
403 	l_clock_t	tms_stime;
404 	l_clock_t	tms_cutime;
405 	l_clock_t	tms_cstime;
406 };
407 
408 /*
409  * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
410  * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
411  * auxiliary vector entry.
412  */
413 #define	CLK_TCK		100
414 
415 #define	CONVOTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
416 #define	CONVNTCK(r)	(r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
417 
418 #define	CONVTCK(r)	(linux_kernver(td) >= LINUX_KERNVER(2,4,0) ?	\
419 			    CONVNTCK(r) : CONVOTCK(r))
420 
421 int
422 linux_times(struct thread *td, struct linux_times_args *args)
423 {
424 	struct timeval tv, utime, stime, cutime, cstime;
425 	struct l_times_argv tms;
426 	struct proc *p;
427 	int error;
428 
429 	if (args->buf != NULL) {
430 		p = td->td_proc;
431 		PROC_LOCK(p);
432 		PROC_STATLOCK(p);
433 		calcru(p, &utime, &stime);
434 		PROC_STATUNLOCK(p);
435 		calccru(p, &cutime, &cstime);
436 		PROC_UNLOCK(p);
437 
438 		tms.tms_utime = CONVTCK(utime);
439 		tms.tms_stime = CONVTCK(stime);
440 
441 		tms.tms_cutime = CONVTCK(cutime);
442 		tms.tms_cstime = CONVTCK(cstime);
443 
444 		if ((error = copyout(&tms, args->buf, sizeof(tms))))
445 			return (error);
446 	}
447 
448 	microuptime(&tv);
449 	td->td_retval[0] = (int)CONVTCK(tv);
450 	return (0);
451 }
452 
453 int
454 linux_newuname(struct thread *td, struct linux_newuname_args *args)
455 {
456 	struct l_new_utsname utsname;
457 	char osname[LINUX_MAX_UTSNAME];
458 	char osrelease[LINUX_MAX_UTSNAME];
459 	char *p;
460 
461 	linux_get_osname(td, osname);
462 	linux_get_osrelease(td, osrelease);
463 
464 	bzero(&utsname, sizeof(utsname));
465 	strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
466 	getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
467 	getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
468 	strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
469 	strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
470 	for (p = utsname.version; *p != '\0'; ++p)
471 		if (*p == '\n') {
472 			*p = '\0';
473 			break;
474 		}
475 #if defined(__amd64__)
476 	/*
477 	 * On amd64, Linux uname(2) needs to return "x86_64"
478 	 * for both 64-bit and 32-bit applications.  On 32-bit,
479 	 * the string returned by getauxval(AT_PLATFORM) needs
480 	 * to remain "i686", though.
481 	 */
482 #if defined(COMPAT_LINUX32)
483 	if (linux32_emulate_i386)
484 		strlcpy(utsname.machine, "i686", LINUX_MAX_UTSNAME);
485 	else
486 #endif
487 	strlcpy(utsname.machine, "x86_64", LINUX_MAX_UTSNAME);
488 #elif defined(__aarch64__)
489 	strlcpy(utsname.machine, "aarch64", LINUX_MAX_UTSNAME);
490 #elif defined(__i386__)
491 	strlcpy(utsname.machine, "i686", LINUX_MAX_UTSNAME);
492 #endif
493 
494 	return (copyout(&utsname, args->buf, sizeof(utsname)));
495 }
496 
497 struct l_utimbuf {
498 	l_time_t l_actime;
499 	l_time_t l_modtime;
500 };
501 
502 #ifdef LINUX_LEGACY_SYSCALLS
503 int
504 linux_utime(struct thread *td, struct linux_utime_args *args)
505 {
506 	struct timeval tv[2], *tvp;
507 	struct l_utimbuf lut;
508 	int error;
509 
510 	if (args->times) {
511 		if ((error = copyin(args->times, &lut, sizeof lut)) != 0)
512 			return (error);
513 		tv[0].tv_sec = lut.l_actime;
514 		tv[0].tv_usec = 0;
515 		tv[1].tv_sec = lut.l_modtime;
516 		tv[1].tv_usec = 0;
517 		tvp = tv;
518 	} else
519 		tvp = NULL;
520 
521 	return (kern_utimesat(td, AT_FDCWD, args->fname, UIO_USERSPACE,
522 	    tvp, UIO_SYSSPACE));
523 }
524 #endif
525 
526 #ifdef LINUX_LEGACY_SYSCALLS
527 int
528 linux_utimes(struct thread *td, struct linux_utimes_args *args)
529 {
530 	l_timeval ltv[2];
531 	struct timeval tv[2], *tvp = NULL;
532 	int error;
533 
534 	if (args->tptr != NULL) {
535 		if ((error = copyin(args->tptr, ltv, sizeof ltv)) != 0)
536 			return (error);
537 		tv[0].tv_sec = ltv[0].tv_sec;
538 		tv[0].tv_usec = ltv[0].tv_usec;
539 		tv[1].tv_sec = ltv[1].tv_sec;
540 		tv[1].tv_usec = ltv[1].tv_usec;
541 		tvp = tv;
542 	}
543 
544 	return (kern_utimesat(td, AT_FDCWD, args->fname, UIO_USERSPACE,
545 	    tvp, UIO_SYSSPACE));
546 }
547 #endif
548 
549 static int
550 linux_utimensat_lts_to_ts(struct l_timespec *l_times, struct timespec *times)
551 {
552 
553 	if (l_times->tv_nsec != LINUX_UTIME_OMIT &&
554 	    l_times->tv_nsec != LINUX_UTIME_NOW &&
555 	    (l_times->tv_nsec < 0 || l_times->tv_nsec > 999999999))
556 		return (EINVAL);
557 
558 	times->tv_sec = l_times->tv_sec;
559 	switch (l_times->tv_nsec)
560 	{
561 	case LINUX_UTIME_OMIT:
562 		times->tv_nsec = UTIME_OMIT;
563 		break;
564 	case LINUX_UTIME_NOW:
565 		times->tv_nsec = UTIME_NOW;
566 		break;
567 	default:
568 		times->tv_nsec = l_times->tv_nsec;
569 	}
570 
571 	return (0);
572 }
573 
574 static int
575 linux_common_utimensat(struct thread *td, int ldfd, const char *pathname,
576     struct timespec *timesp, int lflags)
577 {
578 	int dfd, flags = 0;
579 
580 	dfd = (ldfd == LINUX_AT_FDCWD) ? AT_FDCWD : ldfd;
581 
582 	if (lflags & ~(LINUX_AT_SYMLINK_NOFOLLOW | LINUX_AT_EMPTY_PATH))
583 		return (EINVAL);
584 
585 	if (timesp != NULL) {
586 		/* This breaks POSIX, but is what the Linux kernel does
587 		 * _on purpose_ (documented in the man page for utimensat(2)),
588 		 * so we must follow that behaviour. */
589 		if (timesp[0].tv_nsec == UTIME_OMIT &&
590 		    timesp[1].tv_nsec == UTIME_OMIT)
591 			return (0);
592 	}
593 
594 	if (lflags & LINUX_AT_SYMLINK_NOFOLLOW)
595 		flags |= AT_SYMLINK_NOFOLLOW;
596 	if (lflags & LINUX_AT_EMPTY_PATH)
597 		flags |= AT_EMPTY_PATH;
598 
599 	if (pathname != NULL)
600 		return (kern_utimensat(td, dfd, pathname,
601 		    UIO_USERSPACE, timesp, UIO_SYSSPACE, flags));
602 
603 	if (lflags != 0)
604 		return (EINVAL);
605 
606 	return (kern_futimens(td, dfd, timesp, UIO_SYSSPACE));
607 }
608 
609 int
610 linux_utimensat(struct thread *td, struct linux_utimensat_args *args)
611 {
612 	struct l_timespec l_times[2];
613 	struct timespec times[2], *timesp;
614 	int error;
615 
616 	if (args->times != NULL) {
617 		error = copyin(args->times, l_times, sizeof(l_times));
618 		if (error != 0)
619 			return (error);
620 
621 		error = linux_utimensat_lts_to_ts(&l_times[0], &times[0]);
622 		if (error != 0)
623 			return (error);
624 		error = linux_utimensat_lts_to_ts(&l_times[1], &times[1]);
625 		if (error != 0)
626 			return (error);
627 		timesp = times;
628 	} else
629 		timesp = NULL;
630 
631 	return (linux_common_utimensat(td, args->dfd, args->pathname,
632 	    timesp, args->flags));
633 }
634 
635 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
636 static int
637 linux_utimensat_lts64_to_ts(struct l_timespec64 *l_times, struct timespec *times)
638 {
639 
640 	/* Zero out the padding in compat mode. */
641 	l_times->tv_nsec &= 0xFFFFFFFFUL;
642 
643 	if (l_times->tv_nsec != LINUX_UTIME_OMIT &&
644 	    l_times->tv_nsec != LINUX_UTIME_NOW &&
645 	    (l_times->tv_nsec < 0 || l_times->tv_nsec > 999999999))
646 		return (EINVAL);
647 
648 	times->tv_sec = l_times->tv_sec;
649 	switch (l_times->tv_nsec)
650 	{
651 	case LINUX_UTIME_OMIT:
652 		times->tv_nsec = UTIME_OMIT;
653 		break;
654 	case LINUX_UTIME_NOW:
655 		times->tv_nsec = UTIME_NOW;
656 		break;
657 	default:
658 		times->tv_nsec = l_times->tv_nsec;
659 	}
660 
661 	return (0);
662 }
663 
664 int
665 linux_utimensat_time64(struct thread *td, struct linux_utimensat_time64_args *args)
666 {
667 	struct l_timespec64 l_times[2];
668 	struct timespec times[2], *timesp;
669 	int error;
670 
671 	if (args->times64 != NULL) {
672 		error = copyin(args->times64, l_times, sizeof(l_times));
673 		if (error != 0)
674 			return (error);
675 
676 		error = linux_utimensat_lts64_to_ts(&l_times[0], &times[0]);
677 		if (error != 0)
678 			return (error);
679 		error = linux_utimensat_lts64_to_ts(&l_times[1], &times[1]);
680 		if (error != 0)
681 			return (error);
682 		timesp = times;
683 	} else
684 		timesp = NULL;
685 
686 	return (linux_common_utimensat(td, args->dfd, args->pathname,
687 	    timesp, args->flags));
688 }
689 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
690 
691 #ifdef LINUX_LEGACY_SYSCALLS
692 int
693 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
694 {
695 	l_timeval ltv[2];
696 	struct timeval tv[2], *tvp = NULL;
697 	int error, dfd;
698 
699 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
700 
701 	if (args->utimes != NULL) {
702 		if ((error = copyin(args->utimes, ltv, sizeof ltv)) != 0)
703 			return (error);
704 		tv[0].tv_sec = ltv[0].tv_sec;
705 		tv[0].tv_usec = ltv[0].tv_usec;
706 		tv[1].tv_sec = ltv[1].tv_sec;
707 		tv[1].tv_usec = ltv[1].tv_usec;
708 		tvp = tv;
709 	}
710 
711 	return (kern_utimesat(td, dfd, args->filename, UIO_USERSPACE,
712 	    tvp, UIO_SYSSPACE));
713 }
714 #endif
715 
716 static int
717 linux_common_wait(struct thread *td, idtype_t idtype, int id, int *statusp,
718     int options, void *rup, l_siginfo_t *infop)
719 {
720 	l_siginfo_t lsi;
721 	siginfo_t siginfo;
722 	struct __wrusage wru;
723 	int error, status, tmpstat, sig;
724 
725 	error = kern_wait6(td, idtype, id, &status, options,
726 	    rup != NULL ? &wru : NULL, &siginfo);
727 
728 	if (error == 0 && statusp) {
729 		tmpstat = status & 0xffff;
730 		if (WIFSIGNALED(tmpstat)) {
731 			tmpstat = (tmpstat & 0xffffff80) |
732 			    bsd_to_linux_signal(WTERMSIG(tmpstat));
733 		} else if (WIFSTOPPED(tmpstat)) {
734 			tmpstat = (tmpstat & 0xffff00ff) |
735 			    (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8);
736 #if defined(__aarch64__) || (defined(__amd64__) && !defined(COMPAT_LINUX32))
737 			if (WSTOPSIG(status) == SIGTRAP) {
738 				tmpstat = linux_ptrace_status(td,
739 				    siginfo.si_pid, tmpstat);
740 			}
741 #endif
742 		} else if (WIFCONTINUED(tmpstat)) {
743 			tmpstat = 0xffff;
744 		}
745 		error = copyout(&tmpstat, statusp, sizeof(int));
746 	}
747 	if (error == 0 && rup != NULL)
748 		error = linux_copyout_rusage(&wru.wru_self, rup);
749 	if (error == 0 && infop != NULL && td->td_retval[0] != 0) {
750 		sig = bsd_to_linux_signal(siginfo.si_signo);
751 		siginfo_to_lsiginfo(&siginfo, &lsi, sig);
752 		error = copyout(&lsi, infop, sizeof(lsi));
753 	}
754 
755 	return (error);
756 }
757 
758 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
759 int
760 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
761 {
762 	struct linux_wait4_args wait4_args = {
763 		.pid = args->pid,
764 		.status = args->status,
765 		.options = args->options,
766 		.rusage = NULL,
767 	};
768 
769 	return (linux_wait4(td, &wait4_args));
770 }
771 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
772 
773 int
774 linux_wait4(struct thread *td, struct linux_wait4_args *args)
775 {
776 	struct proc *p;
777 	int options, id, idtype;
778 
779 	if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG |
780 	    LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
781 		return (EINVAL);
782 
783 	/* -INT_MIN is not defined. */
784 	if (args->pid == INT_MIN)
785 		return (ESRCH);
786 
787 	options = 0;
788 	linux_to_bsd_waitopts(args->options, &options);
789 
790 	/*
791 	 * For backward compatibility we implicitly add flags WEXITED
792 	 * and WTRAPPED here.
793 	 */
794 	options |= WEXITED | WTRAPPED;
795 
796 	if (args->pid == WAIT_ANY) {
797 		idtype = P_ALL;
798 		id = 0;
799 	} else if (args->pid < 0) {
800 		idtype = P_PGID;
801 		id = (id_t)-args->pid;
802 	} else if (args->pid == 0) {
803 		idtype = P_PGID;
804 		p = td->td_proc;
805 		PROC_LOCK(p);
806 		id = p->p_pgid;
807 		PROC_UNLOCK(p);
808 	} else {
809 		idtype = P_PID;
810 		id = (id_t)args->pid;
811 	}
812 
813 	return (linux_common_wait(td, idtype, id, args->status, options,
814 	    args->rusage, NULL));
815 }
816 
817 int
818 linux_waitid(struct thread *td, struct linux_waitid_args *args)
819 {
820 	idtype_t idtype;
821 	int error, options;
822 	struct proc *p;
823 	pid_t id;
824 
825 	if (args->options & ~(LINUX_WNOHANG | LINUX_WNOWAIT | LINUX_WEXITED |
826 	    LINUX_WSTOPPED | LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
827 		return (EINVAL);
828 
829 	options = 0;
830 	linux_to_bsd_waitopts(args->options, &options);
831 
832 	id = args->id;
833 	switch (args->idtype) {
834 	case LINUX_P_ALL:
835 		idtype = P_ALL;
836 		break;
837 	case LINUX_P_PID:
838 		if (args->id <= 0)
839 			return (EINVAL);
840 		idtype = P_PID;
841 		break;
842 	case LINUX_P_PGID:
843 		if (linux_kernver(td) >= LINUX_KERNVER(5,4,0) && args->id == 0) {
844 			p = td->td_proc;
845 			PROC_LOCK(p);
846 			id = p->p_pgid;
847 			PROC_UNLOCK(p);
848 		} else if (args->id <= 0)
849 			return (EINVAL);
850 		idtype = P_PGID;
851 		break;
852 	case LINUX_P_PIDFD:
853 		LINUX_RATELIMIT_MSG("unsupported waitid P_PIDFD idtype");
854 		return (ENOSYS);
855 	default:
856 		return (EINVAL);
857 	}
858 
859 	error = linux_common_wait(td, idtype, id, NULL, options,
860 	    args->rusage, args->info);
861 	td->td_retval[0] = 0;
862 
863 	return (error);
864 }
865 
866 #ifdef LINUX_LEGACY_SYSCALLS
867 int
868 linux_mknod(struct thread *td, struct linux_mknod_args *args)
869 {
870 	int error;
871 
872 	switch (args->mode & S_IFMT) {
873 	case S_IFIFO:
874 	case S_IFSOCK:
875 		error = kern_mkfifoat(td, AT_FDCWD, args->path, UIO_USERSPACE,
876 		    args->mode);
877 		break;
878 
879 	case S_IFCHR:
880 	case S_IFBLK:
881 		error = kern_mknodat(td, AT_FDCWD, args->path, UIO_USERSPACE,
882 		    args->mode, linux_decode_dev(args->dev));
883 		break;
884 
885 	case S_IFDIR:
886 		error = EPERM;
887 		break;
888 
889 	case 0:
890 		args->mode |= S_IFREG;
891 		/* FALLTHROUGH */
892 	case S_IFREG:
893 		error = kern_openat(td, AT_FDCWD, args->path, UIO_USERSPACE,
894 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
895 		if (error == 0)
896 			kern_close(td, td->td_retval[0]);
897 		break;
898 
899 	default:
900 		error = EINVAL;
901 		break;
902 	}
903 	return (error);
904 }
905 #endif
906 
907 int
908 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
909 {
910 	int error, dfd;
911 
912 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
913 
914 	switch (args->mode & S_IFMT) {
915 	case S_IFIFO:
916 	case S_IFSOCK:
917 		error = kern_mkfifoat(td, dfd, args->filename, UIO_USERSPACE,
918 		    args->mode);
919 		break;
920 
921 	case S_IFCHR:
922 	case S_IFBLK:
923 		error = kern_mknodat(td, dfd, args->filename, UIO_USERSPACE,
924 		    args->mode, linux_decode_dev(args->dev));
925 		break;
926 
927 	case S_IFDIR:
928 		error = EPERM;
929 		break;
930 
931 	case 0:
932 		args->mode |= S_IFREG;
933 		/* FALLTHROUGH */
934 	case S_IFREG:
935 		error = kern_openat(td, dfd, args->filename, UIO_USERSPACE,
936 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
937 		if (error == 0)
938 			kern_close(td, td->td_retval[0]);
939 		break;
940 
941 	default:
942 		error = EINVAL;
943 		break;
944 	}
945 	return (error);
946 }
947 
948 /*
949  * UGH! This is just about the dumbest idea I've ever heard!!
950  */
951 int
952 linux_personality(struct thread *td, struct linux_personality_args *args)
953 {
954 	struct linux_pemuldata *pem;
955 	struct proc *p = td->td_proc;
956 	uint32_t old;
957 
958 	PROC_LOCK(p);
959 	pem = pem_find(p);
960 	old = pem->persona;
961 	if (args->per != 0xffffffff)
962 		pem->persona = args->per;
963 	PROC_UNLOCK(p);
964 
965 	td->td_retval[0] = old;
966 	return (0);
967 }
968 
969 struct l_itimerval {
970 	l_timeval it_interval;
971 	l_timeval it_value;
972 };
973 
974 #define	B2L_ITIMERVAL(bip, lip)						\
975 	(bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec;		\
976 	(bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec;	\
977 	(bip)->it_value.tv_sec = (lip)->it_value.tv_sec;		\
978 	(bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
979 
980 int
981 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
982 {
983 	int error;
984 	struct l_itimerval ls;
985 	struct itimerval aitv, oitv;
986 
987 	if (uap->itv == NULL) {
988 		uap->itv = uap->oitv;
989 		return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
990 	}
991 
992 	error = copyin(uap->itv, &ls, sizeof(ls));
993 	if (error != 0)
994 		return (error);
995 	B2L_ITIMERVAL(&aitv, &ls);
996 	error = kern_setitimer(td, uap->which, &aitv, &oitv);
997 	if (error != 0 || uap->oitv == NULL)
998 		return (error);
999 	B2L_ITIMERVAL(&ls, &oitv);
1000 
1001 	return (copyout(&ls, uap->oitv, sizeof(ls)));
1002 }
1003 
1004 int
1005 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1006 {
1007 	int error;
1008 	struct l_itimerval ls;
1009 	struct itimerval aitv;
1010 
1011 	error = kern_getitimer(td, uap->which, &aitv);
1012 	if (error != 0)
1013 		return (error);
1014 	B2L_ITIMERVAL(&ls, &aitv);
1015 	return (copyout(&ls, uap->itv, sizeof(ls)));
1016 }
1017 
1018 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1019 int
1020 linux_nice(struct thread *td, struct linux_nice_args *args)
1021 {
1022 
1023 	return (kern_setpriority(td, PRIO_PROCESS, 0, args->inc));
1024 }
1025 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1026 
1027 int
1028 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1029 {
1030 	struct ucred *newcred, *oldcred;
1031 	l_gid_t *linux_gidset;
1032 	gid_t *bsd_gidset;
1033 	int ngrp, error;
1034 	struct proc *p;
1035 
1036 	ngrp = args->gidsetsize;
1037 	if (ngrp < 0 || ngrp >= ngroups_max + 1)
1038 		return (EINVAL);
1039 	linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK);
1040 	error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1041 	if (error)
1042 		goto out;
1043 	newcred = crget();
1044 	crextend(newcred, ngrp + 1);
1045 	p = td->td_proc;
1046 	PROC_LOCK(p);
1047 	oldcred = p->p_ucred;
1048 	crcopy(newcred, oldcred);
1049 
1050 	/*
1051 	 * cr_groups[0] holds egid. Setting the whole set from
1052 	 * the supplied set will cause egid to be changed too.
1053 	 * Keep cr_groups[0] unchanged to prevent that.
1054 	 */
1055 
1056 	if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS)) != 0) {
1057 		PROC_UNLOCK(p);
1058 		crfree(newcred);
1059 		goto out;
1060 	}
1061 
1062 	if (ngrp > 0) {
1063 		newcred->cr_ngroups = ngrp + 1;
1064 
1065 		bsd_gidset = newcred->cr_groups;
1066 		ngrp--;
1067 		while (ngrp >= 0) {
1068 			bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1069 			ngrp--;
1070 		}
1071 	} else
1072 		newcred->cr_ngroups = 1;
1073 
1074 	setsugid(p);
1075 	proc_set_cred(p, newcred);
1076 	PROC_UNLOCK(p);
1077 	crfree(oldcred);
1078 	error = 0;
1079 out:
1080 	free(linux_gidset, M_LINUX);
1081 	return (error);
1082 }
1083 
1084 int
1085 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1086 {
1087 	struct ucred *cred;
1088 	l_gid_t *linux_gidset;
1089 	gid_t *bsd_gidset;
1090 	int bsd_gidsetsz, ngrp, error;
1091 
1092 	cred = td->td_ucred;
1093 	bsd_gidset = cred->cr_groups;
1094 	bsd_gidsetsz = cred->cr_ngroups - 1;
1095 
1096 	/*
1097 	 * cr_groups[0] holds egid. Returning the whole set
1098 	 * here will cause a duplicate. Exclude cr_groups[0]
1099 	 * to prevent that.
1100 	 */
1101 
1102 	if ((ngrp = args->gidsetsize) == 0) {
1103 		td->td_retval[0] = bsd_gidsetsz;
1104 		return (0);
1105 	}
1106 
1107 	if (ngrp < bsd_gidsetsz)
1108 		return (EINVAL);
1109 
1110 	ngrp = 0;
1111 	linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1112 	    M_LINUX, M_WAITOK);
1113 	while (ngrp < bsd_gidsetsz) {
1114 		linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1115 		ngrp++;
1116 	}
1117 
1118 	error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1119 	free(linux_gidset, M_LINUX);
1120 	if (error)
1121 		return (error);
1122 
1123 	td->td_retval[0] = ngrp;
1124 	return (0);
1125 }
1126 
1127 static bool
1128 linux_get_dummy_limit(l_uint resource, struct rlimit *rlim)
1129 {
1130 
1131 	if (linux_dummy_rlimits == 0)
1132 		return (false);
1133 
1134 	switch (resource) {
1135 	case LINUX_RLIMIT_LOCKS:
1136 	case LINUX_RLIMIT_SIGPENDING:
1137 	case LINUX_RLIMIT_MSGQUEUE:
1138 	case LINUX_RLIMIT_RTTIME:
1139 		rlim->rlim_cur = LINUX_RLIM_INFINITY;
1140 		rlim->rlim_max = LINUX_RLIM_INFINITY;
1141 		return (true);
1142 	case LINUX_RLIMIT_NICE:
1143 	case LINUX_RLIMIT_RTPRIO:
1144 		rlim->rlim_cur = 0;
1145 		rlim->rlim_max = 0;
1146 		return (true);
1147 	default:
1148 		return (false);
1149 	}
1150 }
1151 
1152 int
1153 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1154 {
1155 	struct rlimit bsd_rlim;
1156 	struct l_rlimit rlim;
1157 	u_int which;
1158 	int error;
1159 
1160 	if (args->resource >= LINUX_RLIM_NLIMITS)
1161 		return (EINVAL);
1162 
1163 	which = linux_to_bsd_resource[args->resource];
1164 	if (which == -1)
1165 		return (EINVAL);
1166 
1167 	error = copyin(args->rlim, &rlim, sizeof(rlim));
1168 	if (error)
1169 		return (error);
1170 
1171 	bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1172 	bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1173 	return (kern_setrlimit(td, which, &bsd_rlim));
1174 }
1175 
1176 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1177 int
1178 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1179 {
1180 	struct l_rlimit rlim;
1181 	struct rlimit bsd_rlim;
1182 	u_int which;
1183 
1184 	if (linux_get_dummy_limit(args->resource, &bsd_rlim)) {
1185 		rlim.rlim_cur = bsd_rlim.rlim_cur;
1186 		rlim.rlim_max = bsd_rlim.rlim_max;
1187 		return (copyout(&rlim, args->rlim, sizeof(rlim)));
1188 	}
1189 
1190 	if (args->resource >= LINUX_RLIM_NLIMITS)
1191 		return (EINVAL);
1192 
1193 	which = linux_to_bsd_resource[args->resource];
1194 	if (which == -1)
1195 		return (EINVAL);
1196 
1197 	lim_rlimit(td, which, &bsd_rlim);
1198 
1199 #ifdef COMPAT_LINUX32
1200 	rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1201 	if (rlim.rlim_cur == UINT_MAX)
1202 		rlim.rlim_cur = INT_MAX;
1203 	rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1204 	if (rlim.rlim_max == UINT_MAX)
1205 		rlim.rlim_max = INT_MAX;
1206 #else
1207 	rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1208 	if (rlim.rlim_cur == ULONG_MAX)
1209 		rlim.rlim_cur = LONG_MAX;
1210 	rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1211 	if (rlim.rlim_max == ULONG_MAX)
1212 		rlim.rlim_max = LONG_MAX;
1213 #endif
1214 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1215 }
1216 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1217 
1218 int
1219 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1220 {
1221 	struct l_rlimit rlim;
1222 	struct rlimit bsd_rlim;
1223 	u_int which;
1224 
1225 	if (linux_get_dummy_limit(args->resource, &bsd_rlim)) {
1226 		rlim.rlim_cur = bsd_rlim.rlim_cur;
1227 		rlim.rlim_max = bsd_rlim.rlim_max;
1228 		return (copyout(&rlim, args->rlim, sizeof(rlim)));
1229 	}
1230 
1231 	if (args->resource >= LINUX_RLIM_NLIMITS)
1232 		return (EINVAL);
1233 
1234 	which = linux_to_bsd_resource[args->resource];
1235 	if (which == -1)
1236 		return (EINVAL);
1237 
1238 	lim_rlimit(td, which, &bsd_rlim);
1239 
1240 	rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1241 	rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1242 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1243 }
1244 
1245 int
1246 linux_sched_setscheduler(struct thread *td,
1247     struct linux_sched_setscheduler_args *args)
1248 {
1249 	struct sched_param sched_param;
1250 	struct thread *tdt;
1251 	int error, policy;
1252 
1253 	switch (args->policy) {
1254 	case LINUX_SCHED_OTHER:
1255 		policy = SCHED_OTHER;
1256 		break;
1257 	case LINUX_SCHED_FIFO:
1258 		policy = SCHED_FIFO;
1259 		break;
1260 	case LINUX_SCHED_RR:
1261 		policy = SCHED_RR;
1262 		break;
1263 	default:
1264 		return (EINVAL);
1265 	}
1266 
1267 	error = copyin(args->param, &sched_param, sizeof(sched_param));
1268 	if (error)
1269 		return (error);
1270 
1271 	if (linux_map_sched_prio) {
1272 		switch (policy) {
1273 		case SCHED_OTHER:
1274 			if (sched_param.sched_priority != 0)
1275 				return (EINVAL);
1276 
1277 			sched_param.sched_priority =
1278 			    PRI_MAX_TIMESHARE - PRI_MIN_TIMESHARE;
1279 			break;
1280 		case SCHED_FIFO:
1281 		case SCHED_RR:
1282 			if (sched_param.sched_priority < 1 ||
1283 			    sched_param.sched_priority >= LINUX_MAX_RT_PRIO)
1284 				return (EINVAL);
1285 
1286 			/*
1287 			 * Map [1, LINUX_MAX_RT_PRIO - 1] to
1288 			 * [0, RTP_PRIO_MAX - RTP_PRIO_MIN] (rounding down).
1289 			 */
1290 			sched_param.sched_priority =
1291 			    (sched_param.sched_priority - 1) *
1292 			    (RTP_PRIO_MAX - RTP_PRIO_MIN + 1) /
1293 			    (LINUX_MAX_RT_PRIO - 1);
1294 			break;
1295 		}
1296 	}
1297 
1298 	tdt = linux_tdfind(td, args->pid, -1);
1299 	if (tdt == NULL)
1300 		return (ESRCH);
1301 
1302 	error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1303 	PROC_UNLOCK(tdt->td_proc);
1304 	return (error);
1305 }
1306 
1307 int
1308 linux_sched_getscheduler(struct thread *td,
1309     struct linux_sched_getscheduler_args *args)
1310 {
1311 	struct thread *tdt;
1312 	int error, policy;
1313 
1314 	tdt = linux_tdfind(td, args->pid, -1);
1315 	if (tdt == NULL)
1316 		return (ESRCH);
1317 
1318 	error = kern_sched_getscheduler(td, tdt, &policy);
1319 	PROC_UNLOCK(tdt->td_proc);
1320 
1321 	switch (policy) {
1322 	case SCHED_OTHER:
1323 		td->td_retval[0] = LINUX_SCHED_OTHER;
1324 		break;
1325 	case SCHED_FIFO:
1326 		td->td_retval[0] = LINUX_SCHED_FIFO;
1327 		break;
1328 	case SCHED_RR:
1329 		td->td_retval[0] = LINUX_SCHED_RR;
1330 		break;
1331 	}
1332 	return (error);
1333 }
1334 
1335 int
1336 linux_sched_get_priority_max(struct thread *td,
1337     struct linux_sched_get_priority_max_args *args)
1338 {
1339 	struct sched_get_priority_max_args bsd;
1340 
1341 	if (linux_map_sched_prio) {
1342 		switch (args->policy) {
1343 		case LINUX_SCHED_OTHER:
1344 			td->td_retval[0] = 0;
1345 			return (0);
1346 		case LINUX_SCHED_FIFO:
1347 		case LINUX_SCHED_RR:
1348 			td->td_retval[0] = LINUX_MAX_RT_PRIO - 1;
1349 			return (0);
1350 		default:
1351 			return (EINVAL);
1352 		}
1353 	}
1354 
1355 	switch (args->policy) {
1356 	case LINUX_SCHED_OTHER:
1357 		bsd.policy = SCHED_OTHER;
1358 		break;
1359 	case LINUX_SCHED_FIFO:
1360 		bsd.policy = SCHED_FIFO;
1361 		break;
1362 	case LINUX_SCHED_RR:
1363 		bsd.policy = SCHED_RR;
1364 		break;
1365 	default:
1366 		return (EINVAL);
1367 	}
1368 	return (sys_sched_get_priority_max(td, &bsd));
1369 }
1370 
1371 int
1372 linux_sched_get_priority_min(struct thread *td,
1373     struct linux_sched_get_priority_min_args *args)
1374 {
1375 	struct sched_get_priority_min_args bsd;
1376 
1377 	if (linux_map_sched_prio) {
1378 		switch (args->policy) {
1379 		case LINUX_SCHED_OTHER:
1380 			td->td_retval[0] = 0;
1381 			return (0);
1382 		case LINUX_SCHED_FIFO:
1383 		case LINUX_SCHED_RR:
1384 			td->td_retval[0] = 1;
1385 			return (0);
1386 		default:
1387 			return (EINVAL);
1388 		}
1389 	}
1390 
1391 	switch (args->policy) {
1392 	case LINUX_SCHED_OTHER:
1393 		bsd.policy = SCHED_OTHER;
1394 		break;
1395 	case LINUX_SCHED_FIFO:
1396 		bsd.policy = SCHED_FIFO;
1397 		break;
1398 	case LINUX_SCHED_RR:
1399 		bsd.policy = SCHED_RR;
1400 		break;
1401 	default:
1402 		return (EINVAL);
1403 	}
1404 	return (sys_sched_get_priority_min(td, &bsd));
1405 }
1406 
1407 #define REBOOT_CAD_ON	0x89abcdef
1408 #define REBOOT_CAD_OFF	0
1409 #define REBOOT_HALT	0xcdef0123
1410 #define REBOOT_RESTART	0x01234567
1411 #define REBOOT_RESTART2	0xA1B2C3D4
1412 #define REBOOT_POWEROFF	0x4321FEDC
1413 #define REBOOT_MAGIC1	0xfee1dead
1414 #define REBOOT_MAGIC2	0x28121969
1415 #define REBOOT_MAGIC2A	0x05121996
1416 #define REBOOT_MAGIC2B	0x16041998
1417 
1418 int
1419 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1420 {
1421 	struct reboot_args bsd_args;
1422 
1423 	if (args->magic1 != REBOOT_MAGIC1)
1424 		return (EINVAL);
1425 
1426 	switch (args->magic2) {
1427 	case REBOOT_MAGIC2:
1428 	case REBOOT_MAGIC2A:
1429 	case REBOOT_MAGIC2B:
1430 		break;
1431 	default:
1432 		return (EINVAL);
1433 	}
1434 
1435 	switch (args->cmd) {
1436 	case REBOOT_CAD_ON:
1437 	case REBOOT_CAD_OFF:
1438 		return (priv_check(td, PRIV_REBOOT));
1439 	case REBOOT_HALT:
1440 		bsd_args.opt = RB_HALT;
1441 		break;
1442 	case REBOOT_RESTART:
1443 	case REBOOT_RESTART2:
1444 		bsd_args.opt = 0;
1445 		break;
1446 	case REBOOT_POWEROFF:
1447 		bsd_args.opt = RB_POWEROFF;
1448 		break;
1449 	default:
1450 		return (EINVAL);
1451 	}
1452 	return (sys_reboot(td, &bsd_args));
1453 }
1454 
1455 int
1456 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1457 {
1458 
1459 	td->td_retval[0] = td->td_proc->p_pid;
1460 
1461 	return (0);
1462 }
1463 
1464 int
1465 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1466 {
1467 	struct linux_emuldata *em;
1468 
1469 	em = em_find(td);
1470 	KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1471 
1472 	td->td_retval[0] = em->em_tid;
1473 
1474 	return (0);
1475 }
1476 
1477 int
1478 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1479 {
1480 
1481 	td->td_retval[0] = kern_getppid(td);
1482 	return (0);
1483 }
1484 
1485 int
1486 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1487 {
1488 
1489 	td->td_retval[0] = td->td_ucred->cr_rgid;
1490 	return (0);
1491 }
1492 
1493 int
1494 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1495 {
1496 
1497 	td->td_retval[0] = td->td_ucred->cr_ruid;
1498 	return (0);
1499 }
1500 
1501 int
1502 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1503 {
1504 
1505 	return (kern_getsid(td, args->pid));
1506 }
1507 
1508 int
1509 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1510 {
1511 	int error;
1512 
1513 	error = kern_getpriority(td, args->which, args->who);
1514 	td->td_retval[0] = 20 - td->td_retval[0];
1515 	return (error);
1516 }
1517 
1518 int
1519 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1520 {
1521 	int name[2];
1522 
1523 	name[0] = CTL_KERN;
1524 	name[1] = KERN_HOSTNAME;
1525 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1526 	    args->len, 0, 0));
1527 }
1528 
1529 int
1530 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1531 {
1532 	int name[2];
1533 
1534 	name[0] = CTL_KERN;
1535 	name[1] = KERN_NISDOMAINNAME;
1536 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1537 	    args->len, 0, 0));
1538 }
1539 
1540 int
1541 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1542 {
1543 
1544 	LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1545 	    args->error_code);
1546 
1547 	/*
1548 	 * XXX: we should send a signal to the parent if
1549 	 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1550 	 * as it doesnt occur often.
1551 	 */
1552 	exit1(td, args->error_code, 0);
1553 		/* NOTREACHED */
1554 }
1555 
1556 #define _LINUX_CAPABILITY_VERSION_1  0x19980330
1557 #define _LINUX_CAPABILITY_VERSION_2  0x20071026
1558 #define _LINUX_CAPABILITY_VERSION_3  0x20080522
1559 
1560 struct l_user_cap_header {
1561 	l_int	version;
1562 	l_int	pid;
1563 };
1564 
1565 struct l_user_cap_data {
1566 	l_int	effective;
1567 	l_int	permitted;
1568 	l_int	inheritable;
1569 };
1570 
1571 int
1572 linux_capget(struct thread *td, struct linux_capget_args *uap)
1573 {
1574 	struct l_user_cap_header luch;
1575 	struct l_user_cap_data lucd[2];
1576 	int error, u32s;
1577 
1578 	if (uap->hdrp == NULL)
1579 		return (EFAULT);
1580 
1581 	error = copyin(uap->hdrp, &luch, sizeof(luch));
1582 	if (error != 0)
1583 		return (error);
1584 
1585 	switch (luch.version) {
1586 	case _LINUX_CAPABILITY_VERSION_1:
1587 		u32s = 1;
1588 		break;
1589 	case _LINUX_CAPABILITY_VERSION_2:
1590 	case _LINUX_CAPABILITY_VERSION_3:
1591 		u32s = 2;
1592 		break;
1593 	default:
1594 		luch.version = _LINUX_CAPABILITY_VERSION_1;
1595 		error = copyout(&luch, uap->hdrp, sizeof(luch));
1596 		if (error)
1597 			return (error);
1598 		return (EINVAL);
1599 	}
1600 
1601 	if (luch.pid)
1602 		return (EPERM);
1603 
1604 	if (uap->datap) {
1605 		/*
1606 		 * The current implementation doesn't support setting
1607 		 * a capability (it's essentially a stub) so indicate
1608 		 * that no capabilities are currently set or available
1609 		 * to request.
1610 		 */
1611 		memset(&lucd, 0, u32s * sizeof(lucd[0]));
1612 		error = copyout(&lucd, uap->datap, u32s * sizeof(lucd[0]));
1613 	}
1614 
1615 	return (error);
1616 }
1617 
1618 int
1619 linux_capset(struct thread *td, struct linux_capset_args *uap)
1620 {
1621 	struct l_user_cap_header luch;
1622 	struct l_user_cap_data lucd[2];
1623 	int error, i, u32s;
1624 
1625 	if (uap->hdrp == NULL || uap->datap == NULL)
1626 		return (EFAULT);
1627 
1628 	error = copyin(uap->hdrp, &luch, sizeof(luch));
1629 	if (error != 0)
1630 		return (error);
1631 
1632 	switch (luch.version) {
1633 	case _LINUX_CAPABILITY_VERSION_1:
1634 		u32s = 1;
1635 		break;
1636 	case _LINUX_CAPABILITY_VERSION_2:
1637 	case _LINUX_CAPABILITY_VERSION_3:
1638 		u32s = 2;
1639 		break;
1640 	default:
1641 		luch.version = _LINUX_CAPABILITY_VERSION_1;
1642 		error = copyout(&luch, uap->hdrp, sizeof(luch));
1643 		if (error)
1644 			return (error);
1645 		return (EINVAL);
1646 	}
1647 
1648 	if (luch.pid)
1649 		return (EPERM);
1650 
1651 	error = copyin(uap->datap, &lucd, u32s * sizeof(lucd[0]));
1652 	if (error != 0)
1653 		return (error);
1654 
1655 	/* We currently don't support setting any capabilities. */
1656 	for (i = 0; i < u32s; i++) {
1657 		if (lucd[i].effective || lucd[i].permitted ||
1658 		    lucd[i].inheritable) {
1659 			linux_msg(td,
1660 			    "capset[%d] effective=0x%x, permitted=0x%x, "
1661 			    "inheritable=0x%x is not implemented", i,
1662 			    (int)lucd[i].effective, (int)lucd[i].permitted,
1663 			    (int)lucd[i].inheritable);
1664 			return (EPERM);
1665 		}
1666 	}
1667 
1668 	return (0);
1669 }
1670 
1671 int
1672 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1673 {
1674 	int error = 0, max_size, arg;
1675 	struct proc *p = td->td_proc;
1676 	char comm[LINUX_MAX_COMM_LEN];
1677 	int pdeath_signal, trace_state;
1678 
1679 	switch (args->option) {
1680 	case LINUX_PR_SET_PDEATHSIG:
1681 		if (!LINUX_SIG_VALID(args->arg2))
1682 			return (EINVAL);
1683 		pdeath_signal = linux_to_bsd_signal(args->arg2);
1684 		return (kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_CTL,
1685 		    &pdeath_signal));
1686 	case LINUX_PR_GET_PDEATHSIG:
1687 		error = kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_STATUS,
1688 		    &pdeath_signal);
1689 		if (error != 0)
1690 			return (error);
1691 		pdeath_signal = bsd_to_linux_signal(pdeath_signal);
1692 		return (copyout(&pdeath_signal,
1693 		    (void *)(register_t)args->arg2,
1694 		    sizeof(pdeath_signal)));
1695 	/*
1696 	 * In Linux, this flag controls if set[gu]id processes can coredump.
1697 	 * There are additional semantics imposed on processes that cannot
1698 	 * coredump:
1699 	 * - Such processes can not be ptraced.
1700 	 * - There are some semantics around ownership of process-related files
1701 	 *   in the /proc namespace.
1702 	 *
1703 	 * In FreeBSD, we can (and by default, do) disable setuid coredump
1704 	 * system-wide with 'sugid_coredump.'  We control tracability on a
1705 	 * per-process basis with the procctl PROC_TRACE (=> P2_NOTRACE flag).
1706 	 * By happy coincidence, P2_NOTRACE also prevents coredumping.  So the
1707 	 * procctl is roughly analogous to Linux's DUMPABLE.
1708 	 *
1709 	 * So, proxy these knobs to the corresponding PROC_TRACE setting.
1710 	 */
1711 	case LINUX_PR_GET_DUMPABLE:
1712 		error = kern_procctl(td, P_PID, p->p_pid, PROC_TRACE_STATUS,
1713 		    &trace_state);
1714 		if (error != 0)
1715 			return (error);
1716 		td->td_retval[0] = (trace_state != -1);
1717 		return (0);
1718 	case LINUX_PR_SET_DUMPABLE:
1719 		/*
1720 		 * It is only valid for userspace to set one of these two
1721 		 * flags, and only one at a time.
1722 		 */
1723 		switch (args->arg2) {
1724 		case LINUX_SUID_DUMP_DISABLE:
1725 			trace_state = PROC_TRACE_CTL_DISABLE_EXEC;
1726 			break;
1727 		case LINUX_SUID_DUMP_USER:
1728 			trace_state = PROC_TRACE_CTL_ENABLE;
1729 			break;
1730 		default:
1731 			return (EINVAL);
1732 		}
1733 		return (kern_procctl(td, P_PID, p->p_pid, PROC_TRACE_CTL,
1734 		    &trace_state));
1735 	case LINUX_PR_GET_KEEPCAPS:
1736 		/*
1737 		 * Indicate that we always clear the effective and
1738 		 * permitted capability sets when the user id becomes
1739 		 * non-zero (actually the capability sets are simply
1740 		 * always zero in the current implementation).
1741 		 */
1742 		td->td_retval[0] = 0;
1743 		break;
1744 	case LINUX_PR_SET_KEEPCAPS:
1745 		/*
1746 		 * Ignore requests to keep the effective and permitted
1747 		 * capability sets when the user id becomes non-zero.
1748 		 */
1749 		break;
1750 	case LINUX_PR_SET_NAME:
1751 		/*
1752 		 * To be on the safe side we need to make sure to not
1753 		 * overflow the size a Linux program expects. We already
1754 		 * do this here in the copyin, so that we don't need to
1755 		 * check on copyout.
1756 		 */
1757 		max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1758 		error = copyinstr((void *)(register_t)args->arg2, comm,
1759 		    max_size, NULL);
1760 
1761 		/* Linux silently truncates the name if it is too long. */
1762 		if (error == ENAMETOOLONG) {
1763 			/*
1764 			 * XXX: copyinstr() isn't documented to populate the
1765 			 * array completely, so do a copyin() to be on the
1766 			 * safe side. This should be changed in case
1767 			 * copyinstr() is changed to guarantee this.
1768 			 */
1769 			error = copyin((void *)(register_t)args->arg2, comm,
1770 			    max_size - 1);
1771 			comm[max_size - 1] = '\0';
1772 		}
1773 		if (error)
1774 			return (error);
1775 
1776 		PROC_LOCK(p);
1777 		strlcpy(p->p_comm, comm, sizeof(p->p_comm));
1778 		PROC_UNLOCK(p);
1779 		break;
1780 	case LINUX_PR_GET_NAME:
1781 		PROC_LOCK(p);
1782 		strlcpy(comm, p->p_comm, sizeof(comm));
1783 		PROC_UNLOCK(p);
1784 		error = copyout(comm, (void *)(register_t)args->arg2,
1785 		    strlen(comm) + 1);
1786 		break;
1787 	case LINUX_PR_GET_SECCOMP:
1788 	case LINUX_PR_SET_SECCOMP:
1789 		/*
1790 		 * Same as returned by Linux without CONFIG_SECCOMP enabled.
1791 		 */
1792 		error = EINVAL;
1793 		break;
1794 	case LINUX_PR_CAPBSET_READ:
1795 #if 0
1796 		/*
1797 		 * This makes too much noise with Ubuntu Focal.
1798 		 */
1799 		linux_msg(td, "unsupported prctl PR_CAPBSET_READ %d",
1800 		    (int)args->arg2);
1801 #endif
1802 		error = EINVAL;
1803 		break;
1804 	case LINUX_PR_SET_NO_NEW_PRIVS:
1805 		arg = args->arg2 == 1 ?
1806 		    PROC_NO_NEW_PRIVS_ENABLE : PROC_NO_NEW_PRIVS_DISABLE;
1807 		error = kern_procctl(td, P_PID, p->p_pid,
1808 		    PROC_NO_NEW_PRIVS_CTL, &arg);
1809 		break;
1810 	case LINUX_PR_SET_PTRACER:
1811 		linux_msg(td, "unsupported prctl PR_SET_PTRACER");
1812 		error = EINVAL;
1813 		break;
1814 	default:
1815 		linux_msg(td, "unsupported prctl option %d", args->option);
1816 		error = EINVAL;
1817 		break;
1818 	}
1819 
1820 	return (error);
1821 }
1822 
1823 int
1824 linux_sched_setparam(struct thread *td,
1825     struct linux_sched_setparam_args *uap)
1826 {
1827 	struct sched_param sched_param;
1828 	struct thread *tdt;
1829 	int error, policy;
1830 
1831 	error = copyin(uap->param, &sched_param, sizeof(sched_param));
1832 	if (error)
1833 		return (error);
1834 
1835 	tdt = linux_tdfind(td, uap->pid, -1);
1836 	if (tdt == NULL)
1837 		return (ESRCH);
1838 
1839 	if (linux_map_sched_prio) {
1840 		error = kern_sched_getscheduler(td, tdt, &policy);
1841 		if (error)
1842 			goto out;
1843 
1844 		switch (policy) {
1845 		case SCHED_OTHER:
1846 			if (sched_param.sched_priority != 0) {
1847 				error = EINVAL;
1848 				goto out;
1849 			}
1850 			sched_param.sched_priority =
1851 			    PRI_MAX_TIMESHARE - PRI_MIN_TIMESHARE;
1852 			break;
1853 		case SCHED_FIFO:
1854 		case SCHED_RR:
1855 			if (sched_param.sched_priority < 1 ||
1856 			    sched_param.sched_priority >= LINUX_MAX_RT_PRIO) {
1857 				error = EINVAL;
1858 				goto out;
1859 			}
1860 			/*
1861 			 * Map [1, LINUX_MAX_RT_PRIO - 1] to
1862 			 * [0, RTP_PRIO_MAX - RTP_PRIO_MIN] (rounding down).
1863 			 */
1864 			sched_param.sched_priority =
1865 			    (sched_param.sched_priority - 1) *
1866 			    (RTP_PRIO_MAX - RTP_PRIO_MIN + 1) /
1867 			    (LINUX_MAX_RT_PRIO - 1);
1868 			break;
1869 		}
1870 	}
1871 
1872 	error = kern_sched_setparam(td, tdt, &sched_param);
1873 out:	PROC_UNLOCK(tdt->td_proc);
1874 	return (error);
1875 }
1876 
1877 int
1878 linux_sched_getparam(struct thread *td,
1879     struct linux_sched_getparam_args *uap)
1880 {
1881 	struct sched_param sched_param;
1882 	struct thread *tdt;
1883 	int error, policy;
1884 
1885 	tdt = linux_tdfind(td, uap->pid, -1);
1886 	if (tdt == NULL)
1887 		return (ESRCH);
1888 
1889 	error = kern_sched_getparam(td, tdt, &sched_param);
1890 	if (error) {
1891 		PROC_UNLOCK(tdt->td_proc);
1892 		return (error);
1893 	}
1894 
1895 	if (linux_map_sched_prio) {
1896 		error = kern_sched_getscheduler(td, tdt, &policy);
1897 		PROC_UNLOCK(tdt->td_proc);
1898 		if (error)
1899 			return (error);
1900 
1901 		switch (policy) {
1902 		case SCHED_OTHER:
1903 			sched_param.sched_priority = 0;
1904 			break;
1905 		case SCHED_FIFO:
1906 		case SCHED_RR:
1907 			/*
1908 			 * Map [0, RTP_PRIO_MAX - RTP_PRIO_MIN] to
1909 			 * [1, LINUX_MAX_RT_PRIO - 1] (rounding up).
1910 			 */
1911 			sched_param.sched_priority =
1912 			    (sched_param.sched_priority *
1913 			    (LINUX_MAX_RT_PRIO - 1) +
1914 			    (RTP_PRIO_MAX - RTP_PRIO_MIN - 1)) /
1915 			    (RTP_PRIO_MAX - RTP_PRIO_MIN) + 1;
1916 			break;
1917 		}
1918 	} else
1919 		PROC_UNLOCK(tdt->td_proc);
1920 
1921 	error = copyout(&sched_param, uap->param, sizeof(sched_param));
1922 	return (error);
1923 }
1924 
1925 /*
1926  * Get affinity of a process.
1927  */
1928 int
1929 linux_sched_getaffinity(struct thread *td,
1930     struct linux_sched_getaffinity_args *args)
1931 {
1932 	struct thread *tdt;
1933 	cpuset_t *mask;
1934 	size_t size;
1935 	int error;
1936 	id_t tid;
1937 
1938 	tdt = linux_tdfind(td, args->pid, -1);
1939 	if (tdt == NULL)
1940 		return (ESRCH);
1941 	tid = tdt->td_tid;
1942 	PROC_UNLOCK(tdt->td_proc);
1943 
1944 	mask = malloc(sizeof(cpuset_t), M_LINUX, M_WAITOK | M_ZERO);
1945 	size = min(args->len, sizeof(cpuset_t));
1946 	error = kern_cpuset_getaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
1947 	    tid, size, mask);
1948 	if (error == ERANGE)
1949 		error = EINVAL;
1950  	if (error == 0)
1951 		error = copyout(mask, args->user_mask_ptr, size);
1952 	if (error == 0)
1953 		td->td_retval[0] = size;
1954 	free(mask, M_LINUX);
1955 	return (error);
1956 }
1957 
1958 /*
1959  *  Set affinity of a process.
1960  */
1961 int
1962 linux_sched_setaffinity(struct thread *td,
1963     struct linux_sched_setaffinity_args *args)
1964 {
1965 	struct thread *tdt;
1966 	cpuset_t *mask;
1967 	int cpu, error;
1968 	size_t len;
1969 	id_t tid;
1970 
1971 	tdt = linux_tdfind(td, args->pid, -1);
1972 	if (tdt == NULL)
1973 		return (ESRCH);
1974 	tid = tdt->td_tid;
1975 	PROC_UNLOCK(tdt->td_proc);
1976 
1977 	len = min(args->len, sizeof(cpuset_t));
1978 	mask = malloc(sizeof(cpuset_t), M_TEMP, M_WAITOK | M_ZERO);
1979 	error = copyin(args->user_mask_ptr, mask, len);
1980 	if (error != 0)
1981 		goto out;
1982 	/* Linux ignore high bits */
1983 	CPU_FOREACH_ISSET(cpu, mask)
1984 		if (cpu > mp_maxid)
1985 			CPU_CLR(cpu, mask);
1986 
1987 	error = kern_cpuset_setaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
1988 	    tid, mask);
1989 	if (error == EDEADLK)
1990 		error = EINVAL;
1991 out:
1992 	free(mask, M_TEMP);
1993 	return (error);
1994 }
1995 
1996 struct linux_rlimit64 {
1997 	uint64_t	rlim_cur;
1998 	uint64_t	rlim_max;
1999 };
2000 
2001 int
2002 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
2003 {
2004 	struct rlimit rlim, nrlim;
2005 	struct linux_rlimit64 lrlim;
2006 	struct proc *p;
2007 	u_int which;
2008 	int flags;
2009 	int error;
2010 
2011 	if (args->new == NULL && args->old != NULL) {
2012 		if (linux_get_dummy_limit(args->resource, &rlim)) {
2013 			lrlim.rlim_cur = rlim.rlim_cur;
2014 			lrlim.rlim_max = rlim.rlim_max;
2015 			return (copyout(&lrlim, args->old, sizeof(lrlim)));
2016 		}
2017 	}
2018 
2019 	if (args->resource >= LINUX_RLIM_NLIMITS)
2020 		return (EINVAL);
2021 
2022 	which = linux_to_bsd_resource[args->resource];
2023 	if (which == -1)
2024 		return (EINVAL);
2025 
2026 	if (args->new != NULL) {
2027 		/*
2028 		 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
2029 		 * rlim is unsigned 64-bit. FreeBSD treats negative limits
2030 		 * as INFINITY so we do not need a conversion even.
2031 		 */
2032 		error = copyin(args->new, &nrlim, sizeof(nrlim));
2033 		if (error != 0)
2034 			return (error);
2035 	}
2036 
2037 	flags = PGET_HOLD | PGET_NOTWEXIT;
2038 	if (args->new != NULL)
2039 		flags |= PGET_CANDEBUG;
2040 	else
2041 		flags |= PGET_CANSEE;
2042 	if (args->pid == 0) {
2043 		p = td->td_proc;
2044 		PHOLD(p);
2045 	} else {
2046 		error = pget(args->pid, flags, &p);
2047 		if (error != 0)
2048 			return (error);
2049 	}
2050 	if (args->old != NULL) {
2051 		PROC_LOCK(p);
2052 		lim_rlimit_proc(p, which, &rlim);
2053 		PROC_UNLOCK(p);
2054 		if (rlim.rlim_cur == RLIM_INFINITY)
2055 			lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2056 		else
2057 			lrlim.rlim_cur = rlim.rlim_cur;
2058 		if (rlim.rlim_max == RLIM_INFINITY)
2059 			lrlim.rlim_max = LINUX_RLIM_INFINITY;
2060 		else
2061 			lrlim.rlim_max = rlim.rlim_max;
2062 		error = copyout(&lrlim, args->old, sizeof(lrlim));
2063 		if (error != 0)
2064 			goto out;
2065 	}
2066 
2067 	if (args->new != NULL)
2068 		error = kern_proc_setrlimit(td, p, which, &nrlim);
2069 
2070  out:
2071 	PRELE(p);
2072 	return (error);
2073 }
2074 
2075 int
2076 linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2077 {
2078 	struct timespec ts, *tsp;
2079 	int error;
2080 
2081 	if (args->tsp != NULL) {
2082 		error = linux_get_timespec(&ts, args->tsp);
2083 		if (error != 0)
2084 			return (error);
2085 		tsp = &ts;
2086 	} else
2087 		tsp = NULL;
2088 
2089 	error = linux_common_pselect6(td, args->nfds, args->readfds,
2090 	    args->writefds, args->exceptfds, tsp, args->sig);
2091 
2092 	if (args->tsp != NULL)
2093 		linux_put_timespec(&ts, args->tsp);
2094 	return (error);
2095 }
2096 
2097 static int
2098 linux_common_pselect6(struct thread *td, l_int nfds, l_fd_set *readfds,
2099     l_fd_set *writefds, l_fd_set *exceptfds, struct timespec *tsp,
2100     l_uintptr_t *sig)
2101 {
2102 	struct timeval utv, tv0, tv1, *tvp;
2103 	struct l_pselect6arg lpse6;
2104 	sigset_t *ssp;
2105 	sigset_t ss;
2106 	int error;
2107 
2108 	ssp = NULL;
2109 	if (sig != NULL) {
2110 		error = copyin(sig, &lpse6, sizeof(lpse6));
2111 		if (error != 0)
2112 			return (error);
2113 		error = linux_copyin_sigset(td, PTRIN(lpse6.ss),
2114 		    lpse6.ss_len, &ss, &ssp);
2115 		if (error != 0)
2116 		    return (error);
2117 	} else
2118 		ssp = NULL;
2119 
2120 	/*
2121 	 * Currently glibc changes nanosecond number to microsecond.
2122 	 * This mean losing precision but for now it is hardly seen.
2123 	 */
2124 	if (tsp != NULL) {
2125 		TIMESPEC_TO_TIMEVAL(&utv, tsp);
2126 		if (itimerfix(&utv))
2127 			return (EINVAL);
2128 
2129 		microtime(&tv0);
2130 		tvp = &utv;
2131 	} else
2132 		tvp = NULL;
2133 
2134 	error = kern_pselect(td, nfds, readfds, writefds,
2135 	    exceptfds, tvp, ssp, LINUX_NFDBITS);
2136 
2137 	if (tsp != NULL) {
2138 		/*
2139 		 * Compute how much time was left of the timeout,
2140 		 * by subtracting the current time and the time
2141 		 * before we started the call, and subtracting
2142 		 * that result from the user-supplied value.
2143 		 */
2144 		microtime(&tv1);
2145 		timevalsub(&tv1, &tv0);
2146 		timevalsub(&utv, &tv1);
2147 		if (utv.tv_sec < 0)
2148 			timevalclear(&utv);
2149 		TIMEVAL_TO_TIMESPEC(&utv, tsp);
2150 	}
2151 	return (error);
2152 }
2153 
2154 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
2155 int
2156 linux_pselect6_time64(struct thread *td,
2157     struct linux_pselect6_time64_args *args)
2158 {
2159 	struct timespec ts, *tsp;
2160 	int error;
2161 
2162 	if (args->tsp != NULL) {
2163 		error = linux_get_timespec64(&ts, args->tsp);
2164 		if (error != 0)
2165 			return (error);
2166 		tsp = &ts;
2167 	} else
2168 		tsp = NULL;
2169 
2170 	error = linux_common_pselect6(td, args->nfds, args->readfds,
2171 	    args->writefds, args->exceptfds, tsp, args->sig);
2172 
2173 	if (args->tsp != NULL)
2174 		linux_put_timespec64(&ts, args->tsp);
2175 	return (error);
2176 }
2177 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
2178 
2179 int
2180 linux_ppoll(struct thread *td, struct linux_ppoll_args *args)
2181 {
2182 	struct timespec uts, *tsp;
2183 	int error;
2184 
2185 	if (args->tsp != NULL) {
2186 		error = linux_get_timespec(&uts, args->tsp);
2187 		if (error != 0)
2188 			return (error);
2189 		tsp = &uts;
2190 	} else
2191 		tsp = NULL;
2192 
2193 	error = linux_common_ppoll(td, args->fds, args->nfds, tsp,
2194 	    args->sset, args->ssize);
2195 	if (error == 0 && args->tsp != NULL)
2196 		error = linux_put_timespec(&uts, args->tsp);
2197 	return (error);
2198 }
2199 
2200 static int
2201 linux_common_ppoll(struct thread *td, struct pollfd *fds, uint32_t nfds,
2202     struct timespec *tsp, l_sigset_t *sset, l_size_t ssize)
2203 {
2204 	struct timespec ts0, ts1;
2205 	struct pollfd stackfds[32];
2206 	struct pollfd *kfds;
2207  	sigset_t *ssp;
2208  	sigset_t ss;
2209  	int error;
2210 
2211 	if (kern_poll_maxfds(nfds))
2212 		return (EINVAL);
2213 	if (sset != NULL) {
2214 		error = linux_copyin_sigset(td, sset, ssize, &ss, &ssp);
2215 		if (error != 0)
2216 		    return (error);
2217 	} else
2218 		ssp = NULL;
2219 	if (tsp != NULL)
2220 		nanotime(&ts0);
2221 
2222 	if (nfds > nitems(stackfds))
2223 		kfds = mallocarray(nfds, sizeof(*kfds), M_TEMP, M_WAITOK);
2224 	else
2225 		kfds = stackfds;
2226 	error = linux_pollin(td, kfds, fds, nfds);
2227 	if (error != 0)
2228 		goto out;
2229 
2230 	error = kern_poll_kfds(td, kfds, nfds, tsp, ssp);
2231 	if (error == 0)
2232 		error = linux_pollout(td, kfds, fds, nfds);
2233 
2234 	if (error == 0 && tsp != NULL) {
2235 		if (td->td_retval[0]) {
2236 			nanotime(&ts1);
2237 			timespecsub(&ts1, &ts0, &ts1);
2238 			timespecsub(tsp, &ts1, tsp);
2239 			if (tsp->tv_sec < 0)
2240 				timespecclear(tsp);
2241 		} else
2242 			timespecclear(tsp);
2243 	}
2244 
2245 out:
2246 	if (nfds > nitems(stackfds))
2247 		free(kfds, M_TEMP);
2248 	return (error);
2249 }
2250 
2251 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
2252 int
2253 linux_ppoll_time64(struct thread *td, struct linux_ppoll_time64_args *args)
2254 {
2255 	struct timespec uts, *tsp;
2256 	int error;
2257 
2258 	if (args->tsp != NULL) {
2259 		error = linux_get_timespec64(&uts, args->tsp);
2260 		if (error != 0)
2261 			return (error);
2262 		tsp = &uts;
2263 	} else
2264  		tsp = NULL;
2265 	error = linux_common_ppoll(td, args->fds, args->nfds, tsp,
2266 	    args->sset, args->ssize);
2267 	if (error == 0 && args->tsp != NULL)
2268 		error = linux_put_timespec64(&uts, args->tsp);
2269 	return (error);
2270 }
2271 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
2272 
2273 static int
2274 linux_pollin(struct thread *td, struct pollfd *fds, struct pollfd *ufds, u_int nfd)
2275 {
2276 	int error;
2277 	u_int i;
2278 
2279 	error = copyin(ufds, fds, nfd * sizeof(*fds));
2280 	if (error != 0)
2281 		return (error);
2282 
2283 	for (i = 0; i < nfd; i++) {
2284 		if (fds->events != 0)
2285 			linux_to_bsd_poll_events(td, fds->fd,
2286 			    fds->events, &fds->events);
2287 		fds++;
2288 	}
2289 	return (0);
2290 }
2291 
2292 static int
2293 linux_pollout(struct thread *td, struct pollfd *fds, struct pollfd *ufds, u_int nfd)
2294 {
2295 	int error = 0;
2296 	u_int i, n = 0;
2297 
2298 	for (i = 0; i < nfd; i++) {
2299 		if (fds->revents != 0) {
2300 			bsd_to_linux_poll_events(fds->revents,
2301 			    &fds->revents);
2302 			n++;
2303 		}
2304 		error = copyout(&fds->revents, &ufds->revents,
2305 		    sizeof(ufds->revents));
2306 		if (error)
2307 			return (error);
2308 		fds++;
2309 		ufds++;
2310 	}
2311 	td->td_retval[0] = n;
2312 	return (0);
2313 }
2314 
2315 static int
2316 linux_sched_rr_get_interval_common(struct thread *td, pid_t pid,
2317     struct timespec *ts)
2318 {
2319 	struct thread *tdt;
2320 	int error;
2321 
2322 	/*
2323 	 * According to man in case the invalid pid specified
2324 	 * EINVAL should be returned.
2325 	 */
2326 	if (pid < 0)
2327 		return (EINVAL);
2328 
2329 	tdt = linux_tdfind(td, pid, -1);
2330 	if (tdt == NULL)
2331 		return (ESRCH);
2332 
2333 	error = kern_sched_rr_get_interval_td(td, tdt, ts);
2334 	PROC_UNLOCK(tdt->td_proc);
2335 	return (error);
2336 }
2337 
2338 int
2339 linux_sched_rr_get_interval(struct thread *td,
2340     struct linux_sched_rr_get_interval_args *uap)
2341 {
2342 	struct timespec ts;
2343 	int error;
2344 
2345 	error = linux_sched_rr_get_interval_common(td, uap->pid, &ts);
2346 	if (error != 0)
2347 		return (error);
2348 	return (linux_put_timespec(&ts, uap->interval));
2349 }
2350 
2351 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
2352 int
2353 linux_sched_rr_get_interval_time64(struct thread *td,
2354     struct linux_sched_rr_get_interval_time64_args *uap)
2355 {
2356 	struct timespec ts;
2357 	int error;
2358 
2359 	error = linux_sched_rr_get_interval_common(td, uap->pid, &ts);
2360 	if (error != 0)
2361 		return (error);
2362 	return (linux_put_timespec64(&ts, uap->interval));
2363 }
2364 #endif
2365 
2366 /*
2367  * In case when the Linux thread is the initial thread in
2368  * the thread group thread id is equal to the process id.
2369  * Glibc depends on this magic (assert in pthread_getattr_np.c).
2370  */
2371 struct thread *
2372 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid)
2373 {
2374 	struct linux_emuldata *em;
2375 	struct thread *tdt;
2376 	struct proc *p;
2377 
2378 	tdt = NULL;
2379 	if (tid == 0 || tid == td->td_tid) {
2380 		if (pid != -1 && td->td_proc->p_pid != pid)
2381 			return (NULL);
2382 		PROC_LOCK(td->td_proc);
2383 		return (td);
2384 	} else if (tid > PID_MAX)
2385 		return (tdfind(tid, pid));
2386 
2387 	/*
2388 	 * Initial thread where the tid equal to the pid.
2389 	 */
2390 	p = pfind(tid);
2391 	if (p != NULL) {
2392 		if (SV_PROC_ABI(p) != SV_ABI_LINUX ||
2393 		    (pid != -1 && tid != pid)) {
2394 			/*
2395 			 * p is not a Linuxulator process.
2396 			 */
2397 			PROC_UNLOCK(p);
2398 			return (NULL);
2399 		}
2400 		FOREACH_THREAD_IN_PROC(p, tdt) {
2401 			em = em_find(tdt);
2402 			if (tid == em->em_tid)
2403 				return (tdt);
2404 		}
2405 		PROC_UNLOCK(p);
2406 	}
2407 	return (NULL);
2408 }
2409 
2410 void
2411 linux_to_bsd_waitopts(int options, int *bsdopts)
2412 {
2413 
2414 	if (options & LINUX_WNOHANG)
2415 		*bsdopts |= WNOHANG;
2416 	if (options & LINUX_WUNTRACED)
2417 		*bsdopts |= WUNTRACED;
2418 	if (options & LINUX_WEXITED)
2419 		*bsdopts |= WEXITED;
2420 	if (options & LINUX_WCONTINUED)
2421 		*bsdopts |= WCONTINUED;
2422 	if (options & LINUX_WNOWAIT)
2423 		*bsdopts |= WNOWAIT;
2424 
2425 	if (options & __WCLONE)
2426 		*bsdopts |= WLINUXCLONE;
2427 }
2428 
2429 int
2430 linux_getrandom(struct thread *td, struct linux_getrandom_args *args)
2431 {
2432 	struct uio uio;
2433 	struct iovec iov;
2434 	int error;
2435 
2436 	if (args->flags & ~(LINUX_GRND_NONBLOCK|LINUX_GRND_RANDOM))
2437 		return (EINVAL);
2438 	if (args->count > INT_MAX)
2439 		args->count = INT_MAX;
2440 
2441 	iov.iov_base = args->buf;
2442 	iov.iov_len = args->count;
2443 
2444 	uio.uio_iov = &iov;
2445 	uio.uio_iovcnt = 1;
2446 	uio.uio_resid = iov.iov_len;
2447 	uio.uio_segflg = UIO_USERSPACE;
2448 	uio.uio_rw = UIO_READ;
2449 	uio.uio_td = td;
2450 
2451 	error = read_random_uio(&uio, args->flags & LINUX_GRND_NONBLOCK);
2452 	if (error == 0)
2453 		td->td_retval[0] = args->count - uio.uio_resid;
2454 	return (error);
2455 }
2456 
2457 int
2458 linux_mincore(struct thread *td, struct linux_mincore_args *args)
2459 {
2460 
2461 	/* Needs to be page-aligned */
2462 	if (args->start & PAGE_MASK)
2463 		return (EINVAL);
2464 	return (kern_mincore(td, args->start, args->len, args->vec));
2465 }
2466 
2467 #define	SYSLOG_TAG	"<6>"
2468 
2469 int
2470 linux_syslog(struct thread *td, struct linux_syslog_args *args)
2471 {
2472 	char buf[128], *src, *dst;
2473 	u_int seq;
2474 	int buflen, error;
2475 
2476 	if (args->type != LINUX_SYSLOG_ACTION_READ_ALL) {
2477 		linux_msg(td, "syslog unsupported type 0x%x", args->type);
2478 		return (EINVAL);
2479 	}
2480 
2481 	if (args->len < 6) {
2482 		td->td_retval[0] = 0;
2483 		return (0);
2484 	}
2485 
2486 	error = priv_check(td, PRIV_MSGBUF);
2487 	if (error)
2488 		return (error);
2489 
2490 	mtx_lock(&msgbuf_lock);
2491 	msgbuf_peekbytes(msgbufp, NULL, 0, &seq);
2492 	mtx_unlock(&msgbuf_lock);
2493 
2494 	dst = args->buf;
2495 	error = copyout(&SYSLOG_TAG, dst, sizeof(SYSLOG_TAG));
2496 	/* The -1 is to skip the trailing '\0'. */
2497 	dst += sizeof(SYSLOG_TAG) - 1;
2498 
2499 	while (error == 0) {
2500 		mtx_lock(&msgbuf_lock);
2501 		buflen = msgbuf_peekbytes(msgbufp, buf, sizeof(buf), &seq);
2502 		mtx_unlock(&msgbuf_lock);
2503 
2504 		if (buflen == 0)
2505 			break;
2506 
2507 		for (src = buf; src < buf + buflen && error == 0; src++) {
2508 			if (*src == '\0')
2509 				continue;
2510 
2511 			if (dst >= args->buf + args->len)
2512 				goto out;
2513 
2514 			error = copyout(src, dst, 1);
2515 			dst++;
2516 
2517 			if (*src == '\n' && *(src + 1) != '<' &&
2518 			    dst + sizeof(SYSLOG_TAG) < args->buf + args->len) {
2519 				error = copyout(&SYSLOG_TAG,
2520 				    dst, sizeof(SYSLOG_TAG));
2521 				dst += sizeof(SYSLOG_TAG) - 1;
2522 			}
2523 		}
2524 	}
2525 out:
2526 	td->td_retval[0] = dst - args->buf;
2527 	return (error);
2528 }
2529 
2530 int
2531 linux_getcpu(struct thread *td, struct linux_getcpu_args *args)
2532 {
2533 	int cpu, error, node;
2534 
2535 	cpu = td->td_oncpu; /* Make sure it doesn't change during copyout(9) */
2536 	error = 0;
2537 	node = cpuid_to_pcpu[cpu]->pc_domain;
2538 
2539 	if (args->cpu != NULL)
2540 		error = copyout(&cpu, args->cpu, sizeof(l_int));
2541 	if (args->node != NULL)
2542 		error = copyout(&node, args->node, sizeof(l_int));
2543 	return (error);
2544 }
2545 
2546 #if defined(__i386__) || defined(__amd64__)
2547 int
2548 linux_poll(struct thread *td, struct linux_poll_args *args)
2549 {
2550 	struct timespec ts, *tsp;
2551 
2552 	if (args->timeout != INFTIM) {
2553 		if (args->timeout < 0)
2554 			return (EINVAL);
2555 		ts.tv_sec = args->timeout / 1000;
2556 		ts.tv_nsec = (args->timeout % 1000) * 1000000;
2557 		tsp = &ts;
2558 	} else
2559 		tsp = NULL;
2560 
2561 	return (linux_common_ppoll(td, args->fds, args->nfds,
2562 	    tsp, NULL, 0));
2563 }
2564 #endif /* __i386__ || __amd64__ */
2565 
2566 int
2567 linux_seccomp(struct thread *td, struct linux_seccomp_args *args)
2568 {
2569 
2570 	switch (args->op) {
2571 	case LINUX_SECCOMP_GET_ACTION_AVAIL:
2572 		return (EOPNOTSUPP);
2573 	default:
2574 		/*
2575 		 * Ignore unknown operations, just like Linux kernel built
2576 		 * without CONFIG_SECCOMP.
2577 		 */
2578 		return (EINVAL);
2579 	}
2580 }
2581 
2582 /*
2583  * Custom version of exec_copyin_args(), to copy out argument and environment
2584  * strings from the old process address space into the temporary string buffer.
2585  * Based on freebsd32_exec_copyin_args.
2586  */
2587 static int
2588 linux_exec_copyin_args(struct image_args *args, const char *fname,
2589     enum uio_seg segflg, l_uintptr_t *argv, l_uintptr_t *envv)
2590 {
2591 	char *argp, *envp;
2592 	l_uintptr_t *ptr, arg;
2593 	int error;
2594 
2595 	bzero(args, sizeof(*args));
2596 	if (argv == NULL)
2597 		return (EFAULT);
2598 
2599 	/*
2600 	 * Allocate demand-paged memory for the file name, argument, and
2601 	 * environment strings.
2602 	 */
2603 	error = exec_alloc_args(args);
2604 	if (error != 0)
2605 		return (error);
2606 
2607 	/*
2608 	 * Copy the file name.
2609 	 */
2610 	error = exec_args_add_fname(args, fname, segflg);
2611 	if (error != 0)
2612 		goto err_exit;
2613 
2614 	/*
2615 	 * extract arguments first
2616 	 */
2617 	ptr = argv;
2618 	for (;;) {
2619 		error = copyin(ptr++, &arg, sizeof(arg));
2620 		if (error)
2621 			goto err_exit;
2622 		if (arg == 0)
2623 			break;
2624 		argp = PTRIN(arg);
2625 		error = exec_args_add_arg(args, argp, UIO_USERSPACE);
2626 		if (error != 0)
2627 			goto err_exit;
2628 	}
2629 
2630 	/*
2631 	 * This comment is from Linux do_execveat_common:
2632 	 * When argv is empty, add an empty string ("") as argv[0] to
2633 	 * ensure confused userspace programs that start processing
2634 	 * from argv[1] won't end up walking envp.
2635 	 */
2636 	if (args->argc == 0 &&
2637 	    (error = exec_args_add_arg(args, "", UIO_SYSSPACE) != 0))
2638 		goto err_exit;
2639 
2640 	/*
2641 	 * extract environment strings
2642 	 */
2643 	if (envv) {
2644 		ptr = envv;
2645 		for (;;) {
2646 			error = copyin(ptr++, &arg, sizeof(arg));
2647 			if (error)
2648 				goto err_exit;
2649 			if (arg == 0)
2650 				break;
2651 			envp = PTRIN(arg);
2652 			error = exec_args_add_env(args, envp, UIO_USERSPACE);
2653 			if (error != 0)
2654 				goto err_exit;
2655 		}
2656 	}
2657 
2658 	return (0);
2659 
2660 err_exit:
2661 	exec_free_args(args);
2662 	return (error);
2663 }
2664 
2665 int
2666 linux_execve(struct thread *td, struct linux_execve_args *args)
2667 {
2668 	struct image_args eargs;
2669 	int error;
2670 
2671 	LINUX_CTR(execve);
2672 
2673 	error = linux_exec_copyin_args(&eargs, args->path, UIO_USERSPACE,
2674 	    args->argp, args->envp);
2675 	if (error == 0)
2676 		error = linux_common_execve(td, &eargs);
2677 	AUDIT_SYSCALL_EXIT(error == EJUSTRETURN ? 0 : error, td);
2678 	return (error);
2679 }
2680 
2681 static void
2682 linux_up_rtprio_if(struct thread *td1, struct rtprio *rtp)
2683 {
2684 	struct rtprio rtp2;
2685 
2686 	pri_to_rtp(td1, &rtp2);
2687 	if (rtp2.type <  rtp->type ||
2688 	    (rtp2.type == rtp->type &&
2689 	    rtp2.prio < rtp->prio)) {
2690 		rtp->type = rtp2.type;
2691 		rtp->prio = rtp2.prio;
2692 	}
2693 }
2694 
2695 #define	LINUX_PRIO_DIVIDER	RTP_PRIO_MAX / LINUX_IOPRIO_MAX
2696 
2697 static int
2698 linux_rtprio2ioprio(struct rtprio *rtp)
2699 {
2700 	int ioprio, prio;
2701 
2702 	switch (rtp->type) {
2703 	case RTP_PRIO_IDLE:
2704 		prio = RTP_PRIO_MIN;
2705 		ioprio = LINUX_IOPRIO_PRIO(LINUX_IOPRIO_CLASS_IDLE, prio);
2706 		break;
2707 	case RTP_PRIO_NORMAL:
2708 		prio = rtp->prio / LINUX_PRIO_DIVIDER;
2709 		ioprio = LINUX_IOPRIO_PRIO(LINUX_IOPRIO_CLASS_BE, prio);
2710 		break;
2711 	case RTP_PRIO_REALTIME:
2712 		prio = rtp->prio / LINUX_PRIO_DIVIDER;
2713 		ioprio = LINUX_IOPRIO_PRIO(LINUX_IOPRIO_CLASS_RT, prio);
2714 		break;
2715 	default:
2716 		prio = RTP_PRIO_MIN;
2717 		ioprio = LINUX_IOPRIO_PRIO(LINUX_IOPRIO_CLASS_NONE, prio);
2718 		break;
2719 	}
2720 	return (ioprio);
2721 }
2722 
2723 static int
2724 linux_ioprio2rtprio(int ioprio, struct rtprio *rtp)
2725 {
2726 
2727 	switch (LINUX_IOPRIO_PRIO_CLASS(ioprio)) {
2728 	case LINUX_IOPRIO_CLASS_IDLE:
2729 		rtp->prio = RTP_PRIO_MIN;
2730 		rtp->type = RTP_PRIO_IDLE;
2731 		break;
2732 	case LINUX_IOPRIO_CLASS_BE:
2733 		rtp->prio = LINUX_IOPRIO_PRIO_DATA(ioprio) * LINUX_PRIO_DIVIDER;
2734 		rtp->type = RTP_PRIO_NORMAL;
2735 		break;
2736 	case LINUX_IOPRIO_CLASS_RT:
2737 		rtp->prio = LINUX_IOPRIO_PRIO_DATA(ioprio) * LINUX_PRIO_DIVIDER;
2738 		rtp->type = RTP_PRIO_REALTIME;
2739 		break;
2740 	default:
2741 		return (EINVAL);
2742 	}
2743 	return (0);
2744 }
2745 #undef LINUX_PRIO_DIVIDER
2746 
2747 int
2748 linux_ioprio_get(struct thread *td, struct linux_ioprio_get_args *args)
2749 {
2750 	struct thread *td1;
2751 	struct rtprio rtp;
2752 	struct pgrp *pg;
2753 	struct proc *p;
2754 	int error, found;
2755 
2756 	p = NULL;
2757 	td1 = NULL;
2758 	error = 0;
2759 	found = 0;
2760 	rtp.type = RTP_PRIO_IDLE;
2761 	rtp.prio = RTP_PRIO_MAX;
2762 	switch (args->which) {
2763 	case LINUX_IOPRIO_WHO_PROCESS:
2764 		if (args->who == 0) {
2765 			td1 = td;
2766 			p = td1->td_proc;
2767 			PROC_LOCK(p);
2768 		} else if (args->who > PID_MAX) {
2769 			td1 = linux_tdfind(td, args->who, -1);
2770 			if (td1 != NULL)
2771 				p = td1->td_proc;
2772 		} else
2773 			p = pfind(args->who);
2774 		if (p == NULL)
2775 			return (ESRCH);
2776 		if ((error = p_cansee(td, p))) {
2777 			PROC_UNLOCK(p);
2778 			break;
2779 		}
2780 		if (td1 != NULL) {
2781 			pri_to_rtp(td1, &rtp);
2782 		} else {
2783 			FOREACH_THREAD_IN_PROC(p, td1) {
2784 				linux_up_rtprio_if(td1, &rtp);
2785 			}
2786 		}
2787 		found++;
2788 		PROC_UNLOCK(p);
2789 		break;
2790 	case LINUX_IOPRIO_WHO_PGRP:
2791 		sx_slock(&proctree_lock);
2792 		if (args->who == 0) {
2793 			pg = td->td_proc->p_pgrp;
2794 			PGRP_LOCK(pg);
2795 		} else {
2796 			pg = pgfind(args->who);
2797 			if (pg == NULL) {
2798 				sx_sunlock(&proctree_lock);
2799 				error = ESRCH;
2800 				break;
2801 			}
2802 		}
2803 		sx_sunlock(&proctree_lock);
2804 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
2805 			PROC_LOCK(p);
2806 			if (p->p_state == PRS_NORMAL &&
2807 			    p_cansee(td, p) == 0) {
2808 				FOREACH_THREAD_IN_PROC(p, td1) {
2809 					linux_up_rtprio_if(td1, &rtp);
2810 					found++;
2811 				}
2812 			}
2813 			PROC_UNLOCK(p);
2814 		}
2815 		PGRP_UNLOCK(pg);
2816 		break;
2817 	case LINUX_IOPRIO_WHO_USER:
2818 		if (args->who == 0)
2819 			args->who = td->td_ucred->cr_uid;
2820 		sx_slock(&allproc_lock);
2821 		FOREACH_PROC_IN_SYSTEM(p) {
2822 			PROC_LOCK(p);
2823 			if (p->p_state == PRS_NORMAL &&
2824 			    p->p_ucred->cr_uid == args->who &&
2825 			    p_cansee(td, p) == 0) {
2826 				FOREACH_THREAD_IN_PROC(p, td1) {
2827 					linux_up_rtprio_if(td1, &rtp);
2828 					found++;
2829 				}
2830 			}
2831 			PROC_UNLOCK(p);
2832 		}
2833 		sx_sunlock(&allproc_lock);
2834 		break;
2835 	default:
2836 		error = EINVAL;
2837 		break;
2838 	}
2839 	if (error == 0) {
2840 		if (found != 0)
2841 			td->td_retval[0] = linux_rtprio2ioprio(&rtp);
2842 		else
2843 			error = ESRCH;
2844 	}
2845 	return (error);
2846 }
2847 
2848 int
2849 linux_ioprio_set(struct thread *td, struct linux_ioprio_set_args *args)
2850 {
2851 	struct thread *td1;
2852 	struct rtprio rtp;
2853 	struct pgrp *pg;
2854 	struct proc *p;
2855 	int error;
2856 
2857 	if ((error = linux_ioprio2rtprio(args->ioprio, &rtp)) != 0)
2858 		return (error);
2859 	/* Attempts to set high priorities (REALTIME) require su privileges. */
2860 	if (RTP_PRIO_BASE(rtp.type) == RTP_PRIO_REALTIME &&
2861 	    (error = priv_check(td, PRIV_SCHED_RTPRIO)) != 0)
2862 		return (error);
2863 
2864 	p = NULL;
2865 	td1 = NULL;
2866 	switch (args->which) {
2867 	case LINUX_IOPRIO_WHO_PROCESS:
2868 		if (args->who == 0) {
2869 			td1 = td;
2870 			p = td1->td_proc;
2871 			PROC_LOCK(p);
2872 		} else if (args->who > PID_MAX) {
2873 			td1 = linux_tdfind(td, args->who, -1);
2874 			if (td1 != NULL)
2875 				p = td1->td_proc;
2876 		} else
2877 			p = pfind(args->who);
2878 		if (p == NULL)
2879 			return (ESRCH);
2880 		if ((error = p_cansched(td, p))) {
2881 			PROC_UNLOCK(p);
2882 			break;
2883 		}
2884 		if (td1 != NULL) {
2885 			error = rtp_to_pri(&rtp, td1);
2886 		} else {
2887 			FOREACH_THREAD_IN_PROC(p, td1) {
2888 				if ((error = rtp_to_pri(&rtp, td1)) != 0)
2889 					break;
2890 			}
2891 		}
2892 		PROC_UNLOCK(p);
2893 		break;
2894 	case LINUX_IOPRIO_WHO_PGRP:
2895 		sx_slock(&proctree_lock);
2896 		if (args->who == 0) {
2897 			pg = td->td_proc->p_pgrp;
2898 			PGRP_LOCK(pg);
2899 		} else {
2900 			pg = pgfind(args->who);
2901 			if (pg == NULL) {
2902 				sx_sunlock(&proctree_lock);
2903 				error = ESRCH;
2904 				break;
2905 			}
2906 		}
2907 		sx_sunlock(&proctree_lock);
2908 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
2909 			PROC_LOCK(p);
2910 			if (p->p_state == PRS_NORMAL &&
2911 			    p_cansched(td, p) == 0) {
2912 				FOREACH_THREAD_IN_PROC(p, td1) {
2913 					if ((error = rtp_to_pri(&rtp, td1)) != 0)
2914 						break;
2915 				}
2916 			}
2917 			PROC_UNLOCK(p);
2918 			if (error != 0)
2919 				break;
2920 		}
2921 		PGRP_UNLOCK(pg);
2922 		break;
2923 	case LINUX_IOPRIO_WHO_USER:
2924 		if (args->who == 0)
2925 			args->who = td->td_ucred->cr_uid;
2926 		sx_slock(&allproc_lock);
2927 		FOREACH_PROC_IN_SYSTEM(p) {
2928 			PROC_LOCK(p);
2929 			if (p->p_state == PRS_NORMAL &&
2930 			    p->p_ucred->cr_uid == args->who &&
2931 			    p_cansched(td, p) == 0) {
2932 				FOREACH_THREAD_IN_PROC(p, td1) {
2933 					if ((error = rtp_to_pri(&rtp, td1)) != 0)
2934 						break;
2935 				}
2936 			}
2937 			PROC_UNLOCK(p);
2938 			if (error != 0)
2939 				break;
2940 		}
2941 		sx_sunlock(&allproc_lock);
2942 		break;
2943 	default:
2944 		error = EINVAL;
2945 		break;
2946 	}
2947 	return (error);
2948 }
2949