xref: /freebsd/sys/compat/linux/linux_misc.c (revision 22cf89c938886d14f5796fc49f9f020c23ea8eaf)
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_sysproto.h>
83 #include <compat/linux/linux_emul.h>
84 #include <compat/linux/linux_misc.h>
85 
86 int stclohz;				/* Statistics clock frequency */
87 
88 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
89 	RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
90 	RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
91 	RLIMIT_MEMLOCK, RLIMIT_AS
92 };
93 
94 struct l_sysinfo {
95 	l_long		uptime;		/* Seconds since boot */
96 	l_ulong		loads[3];	/* 1, 5, and 15 minute load averages */
97 #define LINUX_SYSINFO_LOADS_SCALE 65536
98 	l_ulong		totalram;	/* Total usable main memory size */
99 	l_ulong		freeram;	/* Available memory size */
100 	l_ulong		sharedram;	/* Amount of shared memory */
101 	l_ulong		bufferram;	/* Memory used by buffers */
102 	l_ulong		totalswap;	/* Total swap space size */
103 	l_ulong		freeswap;	/* swap space still available */
104 	l_ushort	procs;		/* Number of current processes */
105 	l_ushort	pads;
106 	l_ulong		totalhigh;
107 	l_ulong		freehigh;
108 	l_uint		mem_unit;
109 	char		_f[20-2*sizeof(l_long)-sizeof(l_int)];	/* padding */
110 };
111 
112 struct l_pselect6arg {
113 	l_uintptr_t	ss;
114 	l_size_t	ss_len;
115 };
116 
117 static int	linux_utimensat_lts_to_ts(struct l_timespec *,
118 			struct timespec *);
119 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
120 static int	linux_utimensat_lts64_to_ts(struct l_timespec64 *,
121 			struct timespec *);
122 #endif
123 static int	linux_common_utimensat(struct thread *, int,
124 			const char *, struct timespec *, int);
125 static int	linux_common_pselect6(struct thread *, l_int,
126 			l_fd_set *, l_fd_set *, l_fd_set *,
127 			struct timespec *, l_uintptr_t *);
128 static int	linux_common_ppoll(struct thread *, struct pollfd *,
129 			uint32_t, struct timespec *, l_sigset_t *,
130 			l_size_t);
131 static int	linux_pollin(struct thread *, struct pollfd *,
132 			struct pollfd *, u_int);
133 static int	linux_pollout(struct thread *, struct pollfd *,
134 			struct pollfd *, u_int);
135 
136 int
137 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
138 {
139 	struct l_sysinfo sysinfo;
140 	int i, j;
141 	struct timespec ts;
142 
143 	bzero(&sysinfo, sizeof(sysinfo));
144 	getnanouptime(&ts);
145 	if (ts.tv_nsec != 0)
146 		ts.tv_sec++;
147 	sysinfo.uptime = ts.tv_sec;
148 
149 	/* Use the information from the mib to get our load averages */
150 	for (i = 0; i < 3; i++)
151 		sysinfo.loads[i] = averunnable.ldavg[i] *
152 		    LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
153 
154 	sysinfo.totalram = physmem * PAGE_SIZE;
155 	sysinfo.freeram = (u_long)vm_free_count() * PAGE_SIZE;
156 
157 	/*
158 	 * sharedram counts pages allocated to named, swap-backed objects such
159 	 * as shared memory segments and tmpfs files.  There is no cheap way to
160 	 * compute this, so just leave the field unpopulated.  Linux itself only
161 	 * started setting this field in the 3.x timeframe.
162 	 */
163 	sysinfo.sharedram = 0;
164 	sysinfo.bufferram = 0;
165 
166 	swap_pager_status(&i, &j);
167 	sysinfo.totalswap = i * PAGE_SIZE;
168 	sysinfo.freeswap = (i - j) * PAGE_SIZE;
169 
170 	sysinfo.procs = nprocs;
171 
172 	/*
173 	 * Platforms supported by the emulation layer do not have a notion of
174 	 * high memory.
175 	 */
176 	sysinfo.totalhigh = 0;
177 	sysinfo.freehigh = 0;
178 
179 	sysinfo.mem_unit = 1;
180 
181 	return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
182 }
183 
184 #ifdef LINUX_LEGACY_SYSCALLS
185 int
186 linux_alarm(struct thread *td, struct linux_alarm_args *args)
187 {
188 	struct itimerval it, old_it;
189 	u_int secs;
190 	int error __diagused;
191 
192 	secs = args->secs;
193 	/*
194 	 * Linux alarm() is always successful. Limit secs to INT32_MAX / 2
195 	 * to match kern_setitimer()'s limit to avoid error from it.
196 	 *
197 	 * XXX. Linux limit secs to INT_MAX on 32 and does not limit on 64-bit
198 	 * platforms.
199 	 */
200 	if (secs > INT32_MAX / 2)
201 		secs = INT32_MAX / 2;
202 
203 	it.it_value.tv_sec = secs;
204 	it.it_value.tv_usec = 0;
205 	timevalclear(&it.it_interval);
206 	error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
207 	KASSERT(error == 0, ("kern_setitimer returns %d", error));
208 
209 	if ((old_it.it_value.tv_sec == 0 && old_it.it_value.tv_usec > 0) ||
210 	    old_it.it_value.tv_usec >= 500000)
211 		old_it.it_value.tv_sec++;
212 	td->td_retval[0] = old_it.it_value.tv_sec;
213 	return (0);
214 }
215 #endif
216 
217 int
218 linux_brk(struct thread *td, struct linux_brk_args *args)
219 {
220 	struct vmspace *vm = td->td_proc->p_vmspace;
221 	uintptr_t new, old;
222 
223 	old = (uintptr_t)vm->vm_daddr + ctob(vm->vm_dsize);
224 	new = (uintptr_t)args->dsend;
225 	if ((caddr_t)new > vm->vm_daddr && !kern_break(td, &new))
226 		td->td_retval[0] = (register_t)new;
227 	else
228 		td->td_retval[0] = (register_t)old;
229 
230 	return (0);
231 }
232 
233 #ifdef LINUX_LEGACY_SYSCALLS
234 int
235 linux_select(struct thread *td, struct linux_select_args *args)
236 {
237 	l_timeval ltv;
238 	struct timeval tv0, tv1, utv, *tvp;
239 	int error;
240 
241 	/*
242 	 * Store current time for computation of the amount of
243 	 * time left.
244 	 */
245 	if (args->timeout) {
246 		if ((error = copyin(args->timeout, &ltv, sizeof(ltv))))
247 			goto select_out;
248 		utv.tv_sec = ltv.tv_sec;
249 		utv.tv_usec = ltv.tv_usec;
250 
251 		if (itimerfix(&utv)) {
252 			/*
253 			 * The timeval was invalid.  Convert it to something
254 			 * valid that will act as it does under Linux.
255 			 */
256 			utv.tv_sec += utv.tv_usec / 1000000;
257 			utv.tv_usec %= 1000000;
258 			if (utv.tv_usec < 0) {
259 				utv.tv_sec -= 1;
260 				utv.tv_usec += 1000000;
261 			}
262 			if (utv.tv_sec < 0)
263 				timevalclear(&utv);
264 		}
265 		microtime(&tv0);
266 		tvp = &utv;
267 	} else
268 		tvp = NULL;
269 
270 	error = kern_select(td, args->nfds, args->readfds, args->writefds,
271 	    args->exceptfds, tvp, LINUX_NFDBITS);
272 	if (error)
273 		goto select_out;
274 
275 	if (args->timeout) {
276 		if (td->td_retval[0]) {
277 			/*
278 			 * Compute how much time was left of the timeout,
279 			 * by subtracting the current time and the time
280 			 * before we started the call, and subtracting
281 			 * that result from the user-supplied value.
282 			 */
283 			microtime(&tv1);
284 			timevalsub(&tv1, &tv0);
285 			timevalsub(&utv, &tv1);
286 			if (utv.tv_sec < 0)
287 				timevalclear(&utv);
288 		} else
289 			timevalclear(&utv);
290 		ltv.tv_sec = utv.tv_sec;
291 		ltv.tv_usec = utv.tv_usec;
292 		if ((error = copyout(&ltv, args->timeout, sizeof(ltv))))
293 			goto select_out;
294 	}
295 
296 select_out:
297 	return (error);
298 }
299 #endif
300 
301 int
302 linux_mremap(struct thread *td, struct linux_mremap_args *args)
303 {
304 	uintptr_t addr;
305 	size_t len;
306 	int error = 0;
307 
308 	if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
309 		td->td_retval[0] = 0;
310 		return (EINVAL);
311 	}
312 
313 	/*
314 	 * Check for the page alignment.
315 	 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
316 	 */
317 	if (args->addr & PAGE_MASK) {
318 		td->td_retval[0] = 0;
319 		return (EINVAL);
320 	}
321 
322 	args->new_len = round_page(args->new_len);
323 	args->old_len = round_page(args->old_len);
324 
325 	if (args->new_len > args->old_len) {
326 		td->td_retval[0] = 0;
327 		return (ENOMEM);
328 	}
329 
330 	if (args->new_len < args->old_len) {
331 		addr = args->addr + args->new_len;
332 		len = args->old_len - args->new_len;
333 		error = kern_munmap(td, addr, len);
334 	}
335 
336 	td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
337 	return (error);
338 }
339 
340 #define LINUX_MS_ASYNC       0x0001
341 #define LINUX_MS_INVALIDATE  0x0002
342 #define LINUX_MS_SYNC        0x0004
343 
344 int
345 linux_msync(struct thread *td, struct linux_msync_args *args)
346 {
347 
348 	return (kern_msync(td, args->addr, args->len,
349 	    args->fl & ~LINUX_MS_SYNC));
350 }
351 
352 int
353 linux_mprotect(struct thread *td, struct linux_mprotect_args *uap)
354 {
355 
356 	return (linux_mprotect_common(td, PTROUT(uap->addr), uap->len,
357 	    uap->prot));
358 }
359 
360 int
361 linux_madvise(struct thread *td, struct linux_madvise_args *uap)
362 {
363 
364 	return (linux_madvise_common(td, PTROUT(uap->addr), uap->len,
365 	    uap->behav));
366 }
367 
368 int
369 linux_mmap2(struct thread *td, struct linux_mmap2_args *uap)
370 {
371 #if defined(LINUX_ARCHWANT_MMAP2PGOFF)
372 	/*
373 	 * For architectures with sizeof (off_t) < sizeof (loff_t) mmap is
374 	 * implemented with mmap2 syscall and the offset is represented in
375 	 * multiples of page size.
376 	 */
377 	return (linux_mmap_common(td, PTROUT(uap->addr), uap->len, uap->prot,
378 	    uap->flags, uap->fd, (uint64_t)(uint32_t)uap->pgoff * PAGE_SIZE));
379 #else
380 	return (linux_mmap_common(td, PTROUT(uap->addr), uap->len, uap->prot,
381 	    uap->flags, uap->fd, uap->pgoff));
382 #endif
383 }
384 
385 #ifdef LINUX_LEGACY_SYSCALLS
386 int
387 linux_time(struct thread *td, struct linux_time_args *args)
388 {
389 	struct timeval tv;
390 	l_time_t tm;
391 	int error;
392 
393 	microtime(&tv);
394 	tm = tv.tv_sec;
395 	if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
396 		return (error);
397 	td->td_retval[0] = tm;
398 	return (0);
399 }
400 #endif
401 
402 struct l_times_argv {
403 	l_clock_t	tms_utime;
404 	l_clock_t	tms_stime;
405 	l_clock_t	tms_cutime;
406 	l_clock_t	tms_cstime;
407 };
408 
409 /*
410  * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
411  * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
412  * auxiliary vector entry.
413  */
414 #define	CLK_TCK		100
415 
416 #define	CONVOTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
417 #define	CONVNTCK(r)	(r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
418 
419 #define	CONVTCK(r)	(linux_kernver(td) >= LINUX_KERNVER(2,4,0) ?	\
420 			    CONVNTCK(r) : CONVOTCK(r))
421 
422 int
423 linux_times(struct thread *td, struct linux_times_args *args)
424 {
425 	struct timeval tv, utime, stime, cutime, cstime;
426 	struct l_times_argv tms;
427 	struct proc *p;
428 	int error;
429 
430 	if (args->buf != NULL) {
431 		p = td->td_proc;
432 		PROC_LOCK(p);
433 		PROC_STATLOCK(p);
434 		calcru(p, &utime, &stime);
435 		PROC_STATUNLOCK(p);
436 		calccru(p, &cutime, &cstime);
437 		PROC_UNLOCK(p);
438 
439 		tms.tms_utime = CONVTCK(utime);
440 		tms.tms_stime = CONVTCK(stime);
441 
442 		tms.tms_cutime = CONVTCK(cutime);
443 		tms.tms_cstime = CONVTCK(cstime);
444 
445 		if ((error = copyout(&tms, args->buf, sizeof(tms))))
446 			return (error);
447 	}
448 
449 	microuptime(&tv);
450 	td->td_retval[0] = (int)CONVTCK(tv);
451 	return (0);
452 }
453 
454 int
455 linux_newuname(struct thread *td, struct linux_newuname_args *args)
456 {
457 	struct l_new_utsname utsname;
458 	char osname[LINUX_MAX_UTSNAME];
459 	char osrelease[LINUX_MAX_UTSNAME];
460 	char *p;
461 
462 	linux_get_osname(td, osname);
463 	linux_get_osrelease(td, osrelease);
464 
465 	bzero(&utsname, sizeof(utsname));
466 	strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
467 	getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
468 	getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
469 	strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
470 	strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
471 	for (p = utsname.version; *p != '\0'; ++p)
472 		if (*p == '\n') {
473 			*p = '\0';
474 			break;
475 		}
476 #if defined(__amd64__)
477 	/*
478 	 * On amd64, Linux uname(2) needs to return "x86_64"
479 	 * for both 64-bit and 32-bit applications.  On 32-bit,
480 	 * the string returned by getauxval(AT_PLATFORM) needs
481 	 * to remain "i686", though.
482 	 */
483 #if defined(COMPAT_LINUX32)
484 	if (linux32_emulate_i386)
485 		strlcpy(utsname.machine, "i686", LINUX_MAX_UTSNAME);
486 	else
487 #endif
488 	strlcpy(utsname.machine, "x86_64", LINUX_MAX_UTSNAME);
489 #elif defined(__aarch64__)
490 	strlcpy(utsname.machine, "aarch64", LINUX_MAX_UTSNAME);
491 #elif defined(__i386__)
492 	strlcpy(utsname.machine, "i686", LINUX_MAX_UTSNAME);
493 #endif
494 
495 	return (copyout(&utsname, args->buf, sizeof(utsname)));
496 }
497 
498 struct l_utimbuf {
499 	l_time_t l_actime;
500 	l_time_t l_modtime;
501 };
502 
503 #ifdef LINUX_LEGACY_SYSCALLS
504 int
505 linux_utime(struct thread *td, struct linux_utime_args *args)
506 {
507 	struct timeval tv[2], *tvp;
508 	struct l_utimbuf lut;
509 	int error;
510 
511 	if (args->times) {
512 		if ((error = copyin(args->times, &lut, sizeof lut)) != 0)
513 			return (error);
514 		tv[0].tv_sec = lut.l_actime;
515 		tv[0].tv_usec = 0;
516 		tv[1].tv_sec = lut.l_modtime;
517 		tv[1].tv_usec = 0;
518 		tvp = tv;
519 	} else
520 		tvp = NULL;
521 
522 	return (kern_utimesat(td, AT_FDCWD, args->fname, UIO_USERSPACE,
523 	    tvp, UIO_SYSSPACE));
524 }
525 #endif
526 
527 #ifdef LINUX_LEGACY_SYSCALLS
528 int
529 linux_utimes(struct thread *td, struct linux_utimes_args *args)
530 {
531 	l_timeval ltv[2];
532 	struct timeval tv[2], *tvp = NULL;
533 	int error;
534 
535 	if (args->tptr != NULL) {
536 		if ((error = copyin(args->tptr, ltv, sizeof ltv)) != 0)
537 			return (error);
538 		tv[0].tv_sec = ltv[0].tv_sec;
539 		tv[0].tv_usec = ltv[0].tv_usec;
540 		tv[1].tv_sec = ltv[1].tv_sec;
541 		tv[1].tv_usec = ltv[1].tv_usec;
542 		tvp = tv;
543 	}
544 
545 	return (kern_utimesat(td, AT_FDCWD, args->fname, UIO_USERSPACE,
546 	    tvp, UIO_SYSSPACE));
547 }
548 #endif
549 
550 static int
551 linux_utimensat_lts_to_ts(struct l_timespec *l_times, struct timespec *times)
552 {
553 
554 	if (l_times->tv_nsec != LINUX_UTIME_OMIT &&
555 	    l_times->tv_nsec != LINUX_UTIME_NOW &&
556 	    (l_times->tv_nsec < 0 || l_times->tv_nsec > 999999999))
557 		return (EINVAL);
558 
559 	times->tv_sec = l_times->tv_sec;
560 	switch (l_times->tv_nsec)
561 	{
562 	case LINUX_UTIME_OMIT:
563 		times->tv_nsec = UTIME_OMIT;
564 		break;
565 	case LINUX_UTIME_NOW:
566 		times->tv_nsec = UTIME_NOW;
567 		break;
568 	default:
569 		times->tv_nsec = l_times->tv_nsec;
570 	}
571 
572 	return (0);
573 }
574 
575 static int
576 linux_common_utimensat(struct thread *td, int ldfd, const char *pathname,
577     struct timespec *timesp, int lflags)
578 {
579 	int dfd, flags = 0;
580 
581 	dfd = (ldfd == LINUX_AT_FDCWD) ? AT_FDCWD : ldfd;
582 
583 	if (lflags & ~(LINUX_AT_SYMLINK_NOFOLLOW | LINUX_AT_EMPTY_PATH))
584 		return (EINVAL);
585 
586 	if (timesp != NULL) {
587 		/* This breaks POSIX, but is what the Linux kernel does
588 		 * _on purpose_ (documented in the man page for utimensat(2)),
589 		 * so we must follow that behaviour. */
590 		if (timesp[0].tv_nsec == UTIME_OMIT &&
591 		    timesp[1].tv_nsec == UTIME_OMIT)
592 			return (0);
593 	}
594 
595 	if (lflags & LINUX_AT_SYMLINK_NOFOLLOW)
596 		flags |= AT_SYMLINK_NOFOLLOW;
597 	if (lflags & LINUX_AT_EMPTY_PATH)
598 		flags |= AT_EMPTY_PATH;
599 
600 	if (pathname != NULL)
601 		return (kern_utimensat(td, dfd, pathname,
602 		    UIO_USERSPACE, timesp, UIO_SYSSPACE, flags));
603 
604 	if (lflags != 0)
605 		return (EINVAL);
606 
607 	return (kern_futimens(td, dfd, timesp, UIO_SYSSPACE));
608 }
609 
610 int
611 linux_utimensat(struct thread *td, struct linux_utimensat_args *args)
612 {
613 	struct l_timespec l_times[2];
614 	struct timespec times[2], *timesp;
615 	int error;
616 
617 	if (args->times != NULL) {
618 		error = copyin(args->times, l_times, sizeof(l_times));
619 		if (error != 0)
620 			return (error);
621 
622 		error = linux_utimensat_lts_to_ts(&l_times[0], &times[0]);
623 		if (error != 0)
624 			return (error);
625 		error = linux_utimensat_lts_to_ts(&l_times[1], &times[1]);
626 		if (error != 0)
627 			return (error);
628 		timesp = times;
629 	} else
630 		timesp = NULL;
631 
632 	return (linux_common_utimensat(td, args->dfd, args->pathname,
633 	    timesp, args->flags));
634 }
635 
636 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
637 static int
638 linux_utimensat_lts64_to_ts(struct l_timespec64 *l_times, struct timespec *times)
639 {
640 
641 	/* Zero out the padding in compat mode. */
642 	l_times->tv_nsec &= 0xFFFFFFFFUL;
643 
644 	if (l_times->tv_nsec != LINUX_UTIME_OMIT &&
645 	    l_times->tv_nsec != LINUX_UTIME_NOW &&
646 	    (l_times->tv_nsec < 0 || l_times->tv_nsec > 999999999))
647 		return (EINVAL);
648 
649 	times->tv_sec = l_times->tv_sec;
650 	switch (l_times->tv_nsec)
651 	{
652 	case LINUX_UTIME_OMIT:
653 		times->tv_nsec = UTIME_OMIT;
654 		break;
655 	case LINUX_UTIME_NOW:
656 		times->tv_nsec = UTIME_NOW;
657 		break;
658 	default:
659 		times->tv_nsec = l_times->tv_nsec;
660 	}
661 
662 	return (0);
663 }
664 
665 int
666 linux_utimensat_time64(struct thread *td, struct linux_utimensat_time64_args *args)
667 {
668 	struct l_timespec64 l_times[2];
669 	struct timespec times[2], *timesp;
670 	int error;
671 
672 	if (args->times64 != NULL) {
673 		error = copyin(args->times64, l_times, sizeof(l_times));
674 		if (error != 0)
675 			return (error);
676 
677 		error = linux_utimensat_lts64_to_ts(&l_times[0], &times[0]);
678 		if (error != 0)
679 			return (error);
680 		error = linux_utimensat_lts64_to_ts(&l_times[1], &times[1]);
681 		if (error != 0)
682 			return (error);
683 		timesp = times;
684 	} else
685 		timesp = NULL;
686 
687 	return (linux_common_utimensat(td, args->dfd, args->pathname,
688 	    timesp, args->flags));
689 }
690 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
691 
692 #ifdef LINUX_LEGACY_SYSCALLS
693 int
694 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
695 {
696 	l_timeval ltv[2];
697 	struct timeval tv[2], *tvp = NULL;
698 	int error, dfd;
699 
700 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
701 
702 	if (args->utimes != NULL) {
703 		if ((error = copyin(args->utimes, ltv, sizeof ltv)) != 0)
704 			return (error);
705 		tv[0].tv_sec = ltv[0].tv_sec;
706 		tv[0].tv_usec = ltv[0].tv_usec;
707 		tv[1].tv_sec = ltv[1].tv_sec;
708 		tv[1].tv_usec = ltv[1].tv_usec;
709 		tvp = tv;
710 	}
711 
712 	return (kern_utimesat(td, dfd, args->filename, UIO_USERSPACE,
713 	    tvp, UIO_SYSSPACE));
714 }
715 #endif
716 
717 static int
718 linux_common_wait(struct thread *td, idtype_t idtype, int id, int *statusp,
719     int options, void *rup, l_siginfo_t *infop)
720 {
721 	l_siginfo_t lsi;
722 	siginfo_t siginfo;
723 	struct __wrusage wru;
724 	int error, status, tmpstat, sig;
725 
726 	error = kern_wait6(td, idtype, id, &status, options,
727 	    rup != NULL ? &wru : NULL, &siginfo);
728 
729 	if (error == 0 && statusp) {
730 		tmpstat = status & 0xffff;
731 		if (WIFSIGNALED(tmpstat)) {
732 			tmpstat = (tmpstat & 0xffffff80) |
733 			    bsd_to_linux_signal(WTERMSIG(tmpstat));
734 		} else if (WIFSTOPPED(tmpstat)) {
735 			tmpstat = (tmpstat & 0xffff00ff) |
736 			    (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8);
737 #if defined(__aarch64__) || (defined(__amd64__) && !defined(COMPAT_LINUX32))
738 			if (WSTOPSIG(status) == SIGTRAP) {
739 				tmpstat = linux_ptrace_status(td,
740 				    siginfo.si_pid, tmpstat);
741 			}
742 #endif
743 		} else if (WIFCONTINUED(tmpstat)) {
744 			tmpstat = 0xffff;
745 		}
746 		error = copyout(&tmpstat, statusp, sizeof(int));
747 	}
748 	if (error == 0 && rup != NULL)
749 		error = linux_copyout_rusage(&wru.wru_self, rup);
750 	if (error == 0 && infop != NULL && td->td_retval[0] != 0) {
751 		sig = bsd_to_linux_signal(siginfo.si_signo);
752 		siginfo_to_lsiginfo(&siginfo, &lsi, sig);
753 		error = copyout(&lsi, infop, sizeof(lsi));
754 	}
755 
756 	return (error);
757 }
758 
759 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
760 int
761 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
762 {
763 	struct linux_wait4_args wait4_args = {
764 		.pid = args->pid,
765 		.status = args->status,
766 		.options = args->options,
767 		.rusage = NULL,
768 	};
769 
770 	return (linux_wait4(td, &wait4_args));
771 }
772 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
773 
774 int
775 linux_wait4(struct thread *td, struct linux_wait4_args *args)
776 {
777 	struct proc *p;
778 	int options, id, idtype;
779 
780 	if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG |
781 	    LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
782 		return (EINVAL);
783 
784 	/* -INT_MIN is not defined. */
785 	if (args->pid == INT_MIN)
786 		return (ESRCH);
787 
788 	options = 0;
789 	linux_to_bsd_waitopts(args->options, &options);
790 
791 	/*
792 	 * For backward compatibility we implicitly add flags WEXITED
793 	 * and WTRAPPED here.
794 	 */
795 	options |= WEXITED | WTRAPPED;
796 
797 	if (args->pid == WAIT_ANY) {
798 		idtype = P_ALL;
799 		id = 0;
800 	} else if (args->pid < 0) {
801 		idtype = P_PGID;
802 		id = (id_t)-args->pid;
803 	} else if (args->pid == 0) {
804 		idtype = P_PGID;
805 		p = td->td_proc;
806 		PROC_LOCK(p);
807 		id = p->p_pgid;
808 		PROC_UNLOCK(p);
809 	} else {
810 		idtype = P_PID;
811 		id = (id_t)args->pid;
812 	}
813 
814 	return (linux_common_wait(td, idtype, id, args->status, options,
815 	    args->rusage, NULL));
816 }
817 
818 int
819 linux_waitid(struct thread *td, struct linux_waitid_args *args)
820 {
821 	idtype_t idtype;
822 	int error, options;
823 	struct proc *p;
824 	pid_t id;
825 
826 	if (args->options & ~(LINUX_WNOHANG | LINUX_WNOWAIT | LINUX_WEXITED |
827 	    LINUX_WSTOPPED | LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
828 		return (EINVAL);
829 
830 	options = 0;
831 	linux_to_bsd_waitopts(args->options, &options);
832 
833 	id = args->id;
834 	switch (args->idtype) {
835 	case LINUX_P_ALL:
836 		idtype = P_ALL;
837 		break;
838 	case LINUX_P_PID:
839 		if (args->id <= 0)
840 			return (EINVAL);
841 		idtype = P_PID;
842 		break;
843 	case LINUX_P_PGID:
844 		if (linux_kernver(td) >= LINUX_KERNVER(5,4,0) && args->id == 0) {
845 			p = td->td_proc;
846 			PROC_LOCK(p);
847 			id = p->p_pgid;
848 			PROC_UNLOCK(p);
849 		} else if (args->id <= 0)
850 			return (EINVAL);
851 		idtype = P_PGID;
852 		break;
853 	case LINUX_P_PIDFD:
854 		LINUX_RATELIMIT_MSG("unsupported waitid P_PIDFD idtype");
855 		return (ENOSYS);
856 	default:
857 		return (EINVAL);
858 	}
859 
860 	error = linux_common_wait(td, idtype, id, NULL, options,
861 	    args->rusage, args->info);
862 	td->td_retval[0] = 0;
863 
864 	return (error);
865 }
866 
867 #ifdef LINUX_LEGACY_SYSCALLS
868 int
869 linux_mknod(struct thread *td, struct linux_mknod_args *args)
870 {
871 	int error;
872 
873 	switch (args->mode & S_IFMT) {
874 	case S_IFIFO:
875 	case S_IFSOCK:
876 		error = kern_mkfifoat(td, AT_FDCWD, args->path, UIO_USERSPACE,
877 		    args->mode);
878 		break;
879 
880 	case S_IFCHR:
881 	case S_IFBLK:
882 		error = kern_mknodat(td, AT_FDCWD, args->path, UIO_USERSPACE,
883 		    args->mode, linux_decode_dev(args->dev));
884 		break;
885 
886 	case S_IFDIR:
887 		error = EPERM;
888 		break;
889 
890 	case 0:
891 		args->mode |= S_IFREG;
892 		/* FALLTHROUGH */
893 	case S_IFREG:
894 		error = kern_openat(td, AT_FDCWD, args->path, UIO_USERSPACE,
895 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
896 		if (error == 0)
897 			kern_close(td, td->td_retval[0]);
898 		break;
899 
900 	default:
901 		error = EINVAL;
902 		break;
903 	}
904 	return (error);
905 }
906 #endif
907 
908 int
909 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
910 {
911 	int error, dfd;
912 
913 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
914 
915 	switch (args->mode & S_IFMT) {
916 	case S_IFIFO:
917 	case S_IFSOCK:
918 		error = kern_mkfifoat(td, dfd, args->filename, UIO_USERSPACE,
919 		    args->mode);
920 		break;
921 
922 	case S_IFCHR:
923 	case S_IFBLK:
924 		error = kern_mknodat(td, dfd, args->filename, UIO_USERSPACE,
925 		    args->mode, linux_decode_dev(args->dev));
926 		break;
927 
928 	case S_IFDIR:
929 		error = EPERM;
930 		break;
931 
932 	case 0:
933 		args->mode |= S_IFREG;
934 		/* FALLTHROUGH */
935 	case S_IFREG:
936 		error = kern_openat(td, dfd, args->filename, UIO_USERSPACE,
937 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
938 		if (error == 0)
939 			kern_close(td, td->td_retval[0]);
940 		break;
941 
942 	default:
943 		error = EINVAL;
944 		break;
945 	}
946 	return (error);
947 }
948 
949 /*
950  * UGH! This is just about the dumbest idea I've ever heard!!
951  */
952 int
953 linux_personality(struct thread *td, struct linux_personality_args *args)
954 {
955 	struct linux_pemuldata *pem;
956 	struct proc *p = td->td_proc;
957 	uint32_t old;
958 
959 	PROC_LOCK(p);
960 	pem = pem_find(p);
961 	old = pem->persona;
962 	if (args->per != 0xffffffff)
963 		pem->persona = args->per;
964 	PROC_UNLOCK(p);
965 
966 	td->td_retval[0] = old;
967 	return (0);
968 }
969 
970 struct l_itimerval {
971 	l_timeval it_interval;
972 	l_timeval it_value;
973 };
974 
975 #define	B2L_ITIMERVAL(bip, lip)						\
976 	(bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec;		\
977 	(bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec;	\
978 	(bip)->it_value.tv_sec = (lip)->it_value.tv_sec;		\
979 	(bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
980 
981 int
982 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
983 {
984 	int error;
985 	struct l_itimerval ls;
986 	struct itimerval aitv, oitv;
987 
988 	if (uap->itv == NULL) {
989 		uap->itv = uap->oitv;
990 		return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
991 	}
992 
993 	error = copyin(uap->itv, &ls, sizeof(ls));
994 	if (error != 0)
995 		return (error);
996 	B2L_ITIMERVAL(&aitv, &ls);
997 	error = kern_setitimer(td, uap->which, &aitv, &oitv);
998 	if (error != 0 || uap->oitv == NULL)
999 		return (error);
1000 	B2L_ITIMERVAL(&ls, &oitv);
1001 
1002 	return (copyout(&ls, uap->oitv, sizeof(ls)));
1003 }
1004 
1005 int
1006 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1007 {
1008 	int error;
1009 	struct l_itimerval ls;
1010 	struct itimerval aitv;
1011 
1012 	error = kern_getitimer(td, uap->which, &aitv);
1013 	if (error != 0)
1014 		return (error);
1015 	B2L_ITIMERVAL(&ls, &aitv);
1016 	return (copyout(&ls, uap->itv, sizeof(ls)));
1017 }
1018 
1019 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1020 int
1021 linux_nice(struct thread *td, struct linux_nice_args *args)
1022 {
1023 
1024 	return (kern_setpriority(td, PRIO_PROCESS, 0, args->inc));
1025 }
1026 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1027 
1028 int
1029 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1030 {
1031 	struct ucred *newcred, *oldcred;
1032 	l_gid_t *linux_gidset;
1033 	gid_t *bsd_gidset;
1034 	int ngrp, error;
1035 	struct proc *p;
1036 
1037 	ngrp = args->gidsetsize;
1038 	if (ngrp < 0 || ngrp >= ngroups_max + 1)
1039 		return (EINVAL);
1040 	linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK);
1041 	error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1042 	if (error)
1043 		goto out;
1044 	newcred = crget();
1045 	crextend(newcred, ngrp + 1);
1046 	p = td->td_proc;
1047 	PROC_LOCK(p);
1048 	oldcred = p->p_ucred;
1049 	crcopy(newcred, oldcred);
1050 
1051 	/*
1052 	 * cr_groups[0] holds egid. Setting the whole set from
1053 	 * the supplied set will cause egid to be changed too.
1054 	 * Keep cr_groups[0] unchanged to prevent that.
1055 	 */
1056 
1057 	if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS)) != 0) {
1058 		PROC_UNLOCK(p);
1059 		crfree(newcred);
1060 		goto out;
1061 	}
1062 
1063 	if (ngrp > 0) {
1064 		newcred->cr_ngroups = ngrp + 1;
1065 
1066 		bsd_gidset = newcred->cr_groups;
1067 		ngrp--;
1068 		while (ngrp >= 0) {
1069 			bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1070 			ngrp--;
1071 		}
1072 	} else
1073 		newcred->cr_ngroups = 1;
1074 
1075 	setsugid(p);
1076 	proc_set_cred(p, newcred);
1077 	PROC_UNLOCK(p);
1078 	crfree(oldcred);
1079 	error = 0;
1080 out:
1081 	free(linux_gidset, M_LINUX);
1082 	return (error);
1083 }
1084 
1085 int
1086 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1087 {
1088 	struct ucred *cred;
1089 	l_gid_t *linux_gidset;
1090 	gid_t *bsd_gidset;
1091 	int bsd_gidsetsz, ngrp, error;
1092 
1093 	cred = td->td_ucred;
1094 	bsd_gidset = cred->cr_groups;
1095 	bsd_gidsetsz = cred->cr_ngroups - 1;
1096 
1097 	/*
1098 	 * cr_groups[0] holds egid. Returning the whole set
1099 	 * here will cause a duplicate. Exclude cr_groups[0]
1100 	 * to prevent that.
1101 	 */
1102 
1103 	if ((ngrp = args->gidsetsize) == 0) {
1104 		td->td_retval[0] = bsd_gidsetsz;
1105 		return (0);
1106 	}
1107 
1108 	if (ngrp < bsd_gidsetsz)
1109 		return (EINVAL);
1110 
1111 	ngrp = 0;
1112 	linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1113 	    M_LINUX, M_WAITOK);
1114 	while (ngrp < bsd_gidsetsz) {
1115 		linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1116 		ngrp++;
1117 	}
1118 
1119 	error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1120 	free(linux_gidset, M_LINUX);
1121 	if (error)
1122 		return (error);
1123 
1124 	td->td_retval[0] = ngrp;
1125 	return (0);
1126 }
1127 
1128 static bool
1129 linux_get_dummy_limit(l_uint resource, struct rlimit *rlim)
1130 {
1131 
1132 	if (linux_dummy_rlimits == 0)
1133 		return (false);
1134 
1135 	switch (resource) {
1136 	case LINUX_RLIMIT_LOCKS:
1137 	case LINUX_RLIMIT_SIGPENDING:
1138 	case LINUX_RLIMIT_MSGQUEUE:
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 	default:
1149 		return (false);
1150 	}
1151 }
1152 
1153 int
1154 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1155 {
1156 	struct rlimit bsd_rlim;
1157 	struct l_rlimit rlim;
1158 	u_int which;
1159 	int error;
1160 
1161 	if (args->resource >= LINUX_RLIM_NLIMITS)
1162 		return (EINVAL);
1163 
1164 	which = linux_to_bsd_resource[args->resource];
1165 	if (which == -1)
1166 		return (EINVAL);
1167 
1168 	error = copyin(args->rlim, &rlim, sizeof(rlim));
1169 	if (error)
1170 		return (error);
1171 
1172 	bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1173 	bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1174 	return (kern_setrlimit(td, which, &bsd_rlim));
1175 }
1176 
1177 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1178 int
1179 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1180 {
1181 	struct l_rlimit rlim;
1182 	struct rlimit bsd_rlim;
1183 	u_int which;
1184 
1185 	if (linux_get_dummy_limit(args->resource, &bsd_rlim)) {
1186 		rlim.rlim_cur = bsd_rlim.rlim_cur;
1187 		rlim.rlim_max = bsd_rlim.rlim_max;
1188 		return (copyout(&rlim, args->rlim, sizeof(rlim)));
1189 	}
1190 
1191 	if (args->resource >= LINUX_RLIM_NLIMITS)
1192 		return (EINVAL);
1193 
1194 	which = linux_to_bsd_resource[args->resource];
1195 	if (which == -1)
1196 		return (EINVAL);
1197 
1198 	lim_rlimit(td, which, &bsd_rlim);
1199 
1200 #ifdef COMPAT_LINUX32
1201 	rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1202 	if (rlim.rlim_cur == UINT_MAX)
1203 		rlim.rlim_cur = INT_MAX;
1204 	rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1205 	if (rlim.rlim_max == UINT_MAX)
1206 		rlim.rlim_max = INT_MAX;
1207 #else
1208 	rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1209 	if (rlim.rlim_cur == ULONG_MAX)
1210 		rlim.rlim_cur = LONG_MAX;
1211 	rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1212 	if (rlim.rlim_max == ULONG_MAX)
1213 		rlim.rlim_max = LONG_MAX;
1214 #endif
1215 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1216 }
1217 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1218 
1219 int
1220 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1221 {
1222 	struct l_rlimit rlim;
1223 	struct rlimit bsd_rlim;
1224 	u_int which;
1225 
1226 	if (linux_get_dummy_limit(args->resource, &bsd_rlim)) {
1227 		rlim.rlim_cur = bsd_rlim.rlim_cur;
1228 		rlim.rlim_max = bsd_rlim.rlim_max;
1229 		return (copyout(&rlim, args->rlim, sizeof(rlim)));
1230 	}
1231 
1232 	if (args->resource >= LINUX_RLIM_NLIMITS)
1233 		return (EINVAL);
1234 
1235 	which = linux_to_bsd_resource[args->resource];
1236 	if (which == -1)
1237 		return (EINVAL);
1238 
1239 	lim_rlimit(td, which, &bsd_rlim);
1240 
1241 	rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1242 	rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1243 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1244 }
1245 
1246 int
1247 linux_sched_setscheduler(struct thread *td,
1248     struct linux_sched_setscheduler_args *args)
1249 {
1250 	struct sched_param sched_param;
1251 	struct thread *tdt;
1252 	int error, policy;
1253 
1254 	switch (args->policy) {
1255 	case LINUX_SCHED_OTHER:
1256 		policy = SCHED_OTHER;
1257 		break;
1258 	case LINUX_SCHED_FIFO:
1259 		policy = SCHED_FIFO;
1260 		break;
1261 	case LINUX_SCHED_RR:
1262 		policy = SCHED_RR;
1263 		break;
1264 	default:
1265 		return (EINVAL);
1266 	}
1267 
1268 	error = copyin(args->param, &sched_param, sizeof(sched_param));
1269 	if (error)
1270 		return (error);
1271 
1272 	if (linux_map_sched_prio) {
1273 		switch (policy) {
1274 		case SCHED_OTHER:
1275 			if (sched_param.sched_priority != 0)
1276 				return (EINVAL);
1277 
1278 			sched_param.sched_priority =
1279 			    PRI_MAX_TIMESHARE - PRI_MIN_TIMESHARE;
1280 			break;
1281 		case SCHED_FIFO:
1282 		case SCHED_RR:
1283 			if (sched_param.sched_priority < 1 ||
1284 			    sched_param.sched_priority >= LINUX_MAX_RT_PRIO)
1285 				return (EINVAL);
1286 
1287 			/*
1288 			 * Map [1, LINUX_MAX_RT_PRIO - 1] to
1289 			 * [0, RTP_PRIO_MAX - RTP_PRIO_MIN] (rounding down).
1290 			 */
1291 			sched_param.sched_priority =
1292 			    (sched_param.sched_priority - 1) *
1293 			    (RTP_PRIO_MAX - RTP_PRIO_MIN + 1) /
1294 			    (LINUX_MAX_RT_PRIO - 1);
1295 			break;
1296 		}
1297 	}
1298 
1299 	tdt = linux_tdfind(td, args->pid, -1);
1300 	if (tdt == NULL)
1301 		return (ESRCH);
1302 
1303 	error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1304 	PROC_UNLOCK(tdt->td_proc);
1305 	return (error);
1306 }
1307 
1308 int
1309 linux_sched_getscheduler(struct thread *td,
1310     struct linux_sched_getscheduler_args *args)
1311 {
1312 	struct thread *tdt;
1313 	int error, policy;
1314 
1315 	tdt = linux_tdfind(td, args->pid, -1);
1316 	if (tdt == NULL)
1317 		return (ESRCH);
1318 
1319 	error = kern_sched_getscheduler(td, tdt, &policy);
1320 	PROC_UNLOCK(tdt->td_proc);
1321 
1322 	switch (policy) {
1323 	case SCHED_OTHER:
1324 		td->td_retval[0] = LINUX_SCHED_OTHER;
1325 		break;
1326 	case SCHED_FIFO:
1327 		td->td_retval[0] = LINUX_SCHED_FIFO;
1328 		break;
1329 	case SCHED_RR:
1330 		td->td_retval[0] = LINUX_SCHED_RR;
1331 		break;
1332 	}
1333 	return (error);
1334 }
1335 
1336 int
1337 linux_sched_get_priority_max(struct thread *td,
1338     struct linux_sched_get_priority_max_args *args)
1339 {
1340 	struct sched_get_priority_max_args bsd;
1341 
1342 	if (linux_map_sched_prio) {
1343 		switch (args->policy) {
1344 		case LINUX_SCHED_OTHER:
1345 			td->td_retval[0] = 0;
1346 			return (0);
1347 		case LINUX_SCHED_FIFO:
1348 		case LINUX_SCHED_RR:
1349 			td->td_retval[0] = LINUX_MAX_RT_PRIO - 1;
1350 			return (0);
1351 		default:
1352 			return (EINVAL);
1353 		}
1354 	}
1355 
1356 	switch (args->policy) {
1357 	case LINUX_SCHED_OTHER:
1358 		bsd.policy = SCHED_OTHER;
1359 		break;
1360 	case LINUX_SCHED_FIFO:
1361 		bsd.policy = SCHED_FIFO;
1362 		break;
1363 	case LINUX_SCHED_RR:
1364 		bsd.policy = SCHED_RR;
1365 		break;
1366 	default:
1367 		return (EINVAL);
1368 	}
1369 	return (sys_sched_get_priority_max(td, &bsd));
1370 }
1371 
1372 int
1373 linux_sched_get_priority_min(struct thread *td,
1374     struct linux_sched_get_priority_min_args *args)
1375 {
1376 	struct sched_get_priority_min_args bsd;
1377 
1378 	if (linux_map_sched_prio) {
1379 		switch (args->policy) {
1380 		case LINUX_SCHED_OTHER:
1381 			td->td_retval[0] = 0;
1382 			return (0);
1383 		case LINUX_SCHED_FIFO:
1384 		case LINUX_SCHED_RR:
1385 			td->td_retval[0] = 1;
1386 			return (0);
1387 		default:
1388 			return (EINVAL);
1389 		}
1390 	}
1391 
1392 	switch (args->policy) {
1393 	case LINUX_SCHED_OTHER:
1394 		bsd.policy = SCHED_OTHER;
1395 		break;
1396 	case LINUX_SCHED_FIFO:
1397 		bsd.policy = SCHED_FIFO;
1398 		break;
1399 	case LINUX_SCHED_RR:
1400 		bsd.policy = SCHED_RR;
1401 		break;
1402 	default:
1403 		return (EINVAL);
1404 	}
1405 	return (sys_sched_get_priority_min(td, &bsd));
1406 }
1407 
1408 #define REBOOT_CAD_ON	0x89abcdef
1409 #define REBOOT_CAD_OFF	0
1410 #define REBOOT_HALT	0xcdef0123
1411 #define REBOOT_RESTART	0x01234567
1412 #define REBOOT_RESTART2	0xA1B2C3D4
1413 #define REBOOT_POWEROFF	0x4321FEDC
1414 #define REBOOT_MAGIC1	0xfee1dead
1415 #define REBOOT_MAGIC2	0x28121969
1416 #define REBOOT_MAGIC2A	0x05121996
1417 #define REBOOT_MAGIC2B	0x16041998
1418 
1419 int
1420 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1421 {
1422 	struct reboot_args bsd_args;
1423 
1424 	if (args->magic1 != REBOOT_MAGIC1)
1425 		return (EINVAL);
1426 
1427 	switch (args->magic2) {
1428 	case REBOOT_MAGIC2:
1429 	case REBOOT_MAGIC2A:
1430 	case REBOOT_MAGIC2B:
1431 		break;
1432 	default:
1433 		return (EINVAL);
1434 	}
1435 
1436 	switch (args->cmd) {
1437 	case REBOOT_CAD_ON:
1438 	case REBOOT_CAD_OFF:
1439 		return (priv_check(td, PRIV_REBOOT));
1440 	case REBOOT_HALT:
1441 		bsd_args.opt = RB_HALT;
1442 		break;
1443 	case REBOOT_RESTART:
1444 	case REBOOT_RESTART2:
1445 		bsd_args.opt = 0;
1446 		break;
1447 	case REBOOT_POWEROFF:
1448 		bsd_args.opt = RB_POWEROFF;
1449 		break;
1450 	default:
1451 		return (EINVAL);
1452 	}
1453 	return (sys_reboot(td, &bsd_args));
1454 }
1455 
1456 int
1457 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1458 {
1459 
1460 	td->td_retval[0] = td->td_proc->p_pid;
1461 
1462 	return (0);
1463 }
1464 
1465 int
1466 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1467 {
1468 	struct linux_emuldata *em;
1469 
1470 	em = em_find(td);
1471 	KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1472 
1473 	td->td_retval[0] = em->em_tid;
1474 
1475 	return (0);
1476 }
1477 
1478 int
1479 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1480 {
1481 
1482 	td->td_retval[0] = kern_getppid(td);
1483 	return (0);
1484 }
1485 
1486 int
1487 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1488 {
1489 
1490 	td->td_retval[0] = td->td_ucred->cr_rgid;
1491 	return (0);
1492 }
1493 
1494 int
1495 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1496 {
1497 
1498 	td->td_retval[0] = td->td_ucred->cr_ruid;
1499 	return (0);
1500 }
1501 
1502 int
1503 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1504 {
1505 
1506 	return (kern_getsid(td, args->pid));
1507 }
1508 
1509 int
1510 linux_nosys(struct thread *td, struct nosys_args *ignore)
1511 {
1512 
1513 	return (ENOSYS);
1514 }
1515 
1516 int
1517 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1518 {
1519 	int error;
1520 
1521 	error = kern_getpriority(td, args->which, args->who);
1522 	td->td_retval[0] = 20 - td->td_retval[0];
1523 	return (error);
1524 }
1525 
1526 int
1527 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1528 {
1529 	int name[2];
1530 
1531 	name[0] = CTL_KERN;
1532 	name[1] = KERN_HOSTNAME;
1533 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1534 	    args->len, 0, 0));
1535 }
1536 
1537 int
1538 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1539 {
1540 	int name[2];
1541 
1542 	name[0] = CTL_KERN;
1543 	name[1] = KERN_NISDOMAINNAME;
1544 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1545 	    args->len, 0, 0));
1546 }
1547 
1548 int
1549 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1550 {
1551 
1552 	LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1553 	    args->error_code);
1554 
1555 	/*
1556 	 * XXX: we should send a signal to the parent if
1557 	 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1558 	 * as it doesnt occur often.
1559 	 */
1560 	exit1(td, args->error_code, 0);
1561 		/* NOTREACHED */
1562 }
1563 
1564 #define _LINUX_CAPABILITY_VERSION_1  0x19980330
1565 #define _LINUX_CAPABILITY_VERSION_2  0x20071026
1566 #define _LINUX_CAPABILITY_VERSION_3  0x20080522
1567 
1568 struct l_user_cap_header {
1569 	l_int	version;
1570 	l_int	pid;
1571 };
1572 
1573 struct l_user_cap_data {
1574 	l_int	effective;
1575 	l_int	permitted;
1576 	l_int	inheritable;
1577 };
1578 
1579 int
1580 linux_capget(struct thread *td, struct linux_capget_args *uap)
1581 {
1582 	struct l_user_cap_header luch;
1583 	struct l_user_cap_data lucd[2];
1584 	int error, u32s;
1585 
1586 	if (uap->hdrp == NULL)
1587 		return (EFAULT);
1588 
1589 	error = copyin(uap->hdrp, &luch, sizeof(luch));
1590 	if (error != 0)
1591 		return (error);
1592 
1593 	switch (luch.version) {
1594 	case _LINUX_CAPABILITY_VERSION_1:
1595 		u32s = 1;
1596 		break;
1597 	case _LINUX_CAPABILITY_VERSION_2:
1598 	case _LINUX_CAPABILITY_VERSION_3:
1599 		u32s = 2;
1600 		break;
1601 	default:
1602 		luch.version = _LINUX_CAPABILITY_VERSION_1;
1603 		error = copyout(&luch, uap->hdrp, sizeof(luch));
1604 		if (error)
1605 			return (error);
1606 		return (EINVAL);
1607 	}
1608 
1609 	if (luch.pid)
1610 		return (EPERM);
1611 
1612 	if (uap->datap) {
1613 		/*
1614 		 * The current implementation doesn't support setting
1615 		 * a capability (it's essentially a stub) so indicate
1616 		 * that no capabilities are currently set or available
1617 		 * to request.
1618 		 */
1619 		memset(&lucd, 0, u32s * sizeof(lucd[0]));
1620 		error = copyout(&lucd, uap->datap, u32s * sizeof(lucd[0]));
1621 	}
1622 
1623 	return (error);
1624 }
1625 
1626 int
1627 linux_capset(struct thread *td, struct linux_capset_args *uap)
1628 {
1629 	struct l_user_cap_header luch;
1630 	struct l_user_cap_data lucd[2];
1631 	int error, i, u32s;
1632 
1633 	if (uap->hdrp == NULL || uap->datap == NULL)
1634 		return (EFAULT);
1635 
1636 	error = copyin(uap->hdrp, &luch, sizeof(luch));
1637 	if (error != 0)
1638 		return (error);
1639 
1640 	switch (luch.version) {
1641 	case _LINUX_CAPABILITY_VERSION_1:
1642 		u32s = 1;
1643 		break;
1644 	case _LINUX_CAPABILITY_VERSION_2:
1645 	case _LINUX_CAPABILITY_VERSION_3:
1646 		u32s = 2;
1647 		break;
1648 	default:
1649 		luch.version = _LINUX_CAPABILITY_VERSION_1;
1650 		error = copyout(&luch, uap->hdrp, sizeof(luch));
1651 		if (error)
1652 			return (error);
1653 		return (EINVAL);
1654 	}
1655 
1656 	if (luch.pid)
1657 		return (EPERM);
1658 
1659 	error = copyin(uap->datap, &lucd, u32s * sizeof(lucd[0]));
1660 	if (error != 0)
1661 		return (error);
1662 
1663 	/* We currently don't support setting any capabilities. */
1664 	for (i = 0; i < u32s; i++) {
1665 		if (lucd[i].effective || lucd[i].permitted ||
1666 		    lucd[i].inheritable) {
1667 			linux_msg(td,
1668 			    "capset[%d] effective=0x%x, permitted=0x%x, "
1669 			    "inheritable=0x%x is not implemented", i,
1670 			    (int)lucd[i].effective, (int)lucd[i].permitted,
1671 			    (int)lucd[i].inheritable);
1672 			return (EPERM);
1673 		}
1674 	}
1675 
1676 	return (0);
1677 }
1678 
1679 int
1680 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1681 {
1682 	int error = 0, max_size, arg;
1683 	struct proc *p = td->td_proc;
1684 	char comm[LINUX_MAX_COMM_LEN];
1685 	int pdeath_signal, trace_state;
1686 
1687 	switch (args->option) {
1688 	case LINUX_PR_SET_PDEATHSIG:
1689 		if (!LINUX_SIG_VALID(args->arg2))
1690 			return (EINVAL);
1691 		pdeath_signal = linux_to_bsd_signal(args->arg2);
1692 		return (kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_CTL,
1693 		    &pdeath_signal));
1694 	case LINUX_PR_GET_PDEATHSIG:
1695 		error = kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_STATUS,
1696 		    &pdeath_signal);
1697 		if (error != 0)
1698 			return (error);
1699 		pdeath_signal = bsd_to_linux_signal(pdeath_signal);
1700 		return (copyout(&pdeath_signal,
1701 		    (void *)(register_t)args->arg2,
1702 		    sizeof(pdeath_signal)));
1703 	/*
1704 	 * In Linux, this flag controls if set[gu]id processes can coredump.
1705 	 * There are additional semantics imposed on processes that cannot
1706 	 * coredump:
1707 	 * - Such processes can not be ptraced.
1708 	 * - There are some semantics around ownership of process-related files
1709 	 *   in the /proc namespace.
1710 	 *
1711 	 * In FreeBSD, we can (and by default, do) disable setuid coredump
1712 	 * system-wide with 'sugid_coredump.'  We control tracability on a
1713 	 * per-process basis with the procctl PROC_TRACE (=> P2_NOTRACE flag).
1714 	 * By happy coincidence, P2_NOTRACE also prevents coredumping.  So the
1715 	 * procctl is roughly analogous to Linux's DUMPABLE.
1716 	 *
1717 	 * So, proxy these knobs to the corresponding PROC_TRACE setting.
1718 	 */
1719 	case LINUX_PR_GET_DUMPABLE:
1720 		error = kern_procctl(td, P_PID, p->p_pid, PROC_TRACE_STATUS,
1721 		    &trace_state);
1722 		if (error != 0)
1723 			return (error);
1724 		td->td_retval[0] = (trace_state != -1);
1725 		return (0);
1726 	case LINUX_PR_SET_DUMPABLE:
1727 		/*
1728 		 * It is only valid for userspace to set one of these two
1729 		 * flags, and only one at a time.
1730 		 */
1731 		switch (args->arg2) {
1732 		case LINUX_SUID_DUMP_DISABLE:
1733 			trace_state = PROC_TRACE_CTL_DISABLE_EXEC;
1734 			break;
1735 		case LINUX_SUID_DUMP_USER:
1736 			trace_state = PROC_TRACE_CTL_ENABLE;
1737 			break;
1738 		default:
1739 			return (EINVAL);
1740 		}
1741 		return (kern_procctl(td, P_PID, p->p_pid, PROC_TRACE_CTL,
1742 		    &trace_state));
1743 	case LINUX_PR_GET_KEEPCAPS:
1744 		/*
1745 		 * Indicate that we always clear the effective and
1746 		 * permitted capability sets when the user id becomes
1747 		 * non-zero (actually the capability sets are simply
1748 		 * always zero in the current implementation).
1749 		 */
1750 		td->td_retval[0] = 0;
1751 		break;
1752 	case LINUX_PR_SET_KEEPCAPS:
1753 		/*
1754 		 * Ignore requests to keep the effective and permitted
1755 		 * capability sets when the user id becomes non-zero.
1756 		 */
1757 		break;
1758 	case LINUX_PR_SET_NAME:
1759 		/*
1760 		 * To be on the safe side we need to make sure to not
1761 		 * overflow the size a Linux program expects. We already
1762 		 * do this here in the copyin, so that we don't need to
1763 		 * check on copyout.
1764 		 */
1765 		max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1766 		error = copyinstr((void *)(register_t)args->arg2, comm,
1767 		    max_size, NULL);
1768 
1769 		/* Linux silently truncates the name if it is too long. */
1770 		if (error == ENAMETOOLONG) {
1771 			/*
1772 			 * XXX: copyinstr() isn't documented to populate the
1773 			 * array completely, so do a copyin() to be on the
1774 			 * safe side. This should be changed in case
1775 			 * copyinstr() is changed to guarantee this.
1776 			 */
1777 			error = copyin((void *)(register_t)args->arg2, comm,
1778 			    max_size - 1);
1779 			comm[max_size - 1] = '\0';
1780 		}
1781 		if (error)
1782 			return (error);
1783 
1784 		PROC_LOCK(p);
1785 		strlcpy(p->p_comm, comm, sizeof(p->p_comm));
1786 		PROC_UNLOCK(p);
1787 		break;
1788 	case LINUX_PR_GET_NAME:
1789 		PROC_LOCK(p);
1790 		strlcpy(comm, p->p_comm, sizeof(comm));
1791 		PROC_UNLOCK(p);
1792 		error = copyout(comm, (void *)(register_t)args->arg2,
1793 		    strlen(comm) + 1);
1794 		break;
1795 	case LINUX_PR_GET_SECCOMP:
1796 	case LINUX_PR_SET_SECCOMP:
1797 		/*
1798 		 * Same as returned by Linux without CONFIG_SECCOMP enabled.
1799 		 */
1800 		error = EINVAL;
1801 		break;
1802 	case LINUX_PR_CAPBSET_READ:
1803 #if 0
1804 		/*
1805 		 * This makes too much noise with Ubuntu Focal.
1806 		 */
1807 		linux_msg(td, "unsupported prctl PR_CAPBSET_READ %d",
1808 		    (int)args->arg2);
1809 #endif
1810 		error = EINVAL;
1811 		break;
1812 	case LINUX_PR_SET_NO_NEW_PRIVS:
1813 		arg = args->arg2 == 1 ?
1814 		    PROC_NO_NEW_PRIVS_ENABLE : PROC_NO_NEW_PRIVS_DISABLE;
1815 		error = kern_procctl(td, P_PID, p->p_pid,
1816 		    PROC_NO_NEW_PRIVS_CTL, &arg);
1817 		break;
1818 	case LINUX_PR_SET_PTRACER:
1819 		linux_msg(td, "unsupported prctl PR_SET_PTRACER");
1820 		error = EINVAL;
1821 		break;
1822 	default:
1823 		linux_msg(td, "unsupported prctl option %d", args->option);
1824 		error = EINVAL;
1825 		break;
1826 	}
1827 
1828 	return (error);
1829 }
1830 
1831 int
1832 linux_sched_setparam(struct thread *td,
1833     struct linux_sched_setparam_args *uap)
1834 {
1835 	struct sched_param sched_param;
1836 	struct thread *tdt;
1837 	int error, policy;
1838 
1839 	error = copyin(uap->param, &sched_param, sizeof(sched_param));
1840 	if (error)
1841 		return (error);
1842 
1843 	tdt = linux_tdfind(td, uap->pid, -1);
1844 	if (tdt == NULL)
1845 		return (ESRCH);
1846 
1847 	if (linux_map_sched_prio) {
1848 		error = kern_sched_getscheduler(td, tdt, &policy);
1849 		if (error)
1850 			goto out;
1851 
1852 		switch (policy) {
1853 		case SCHED_OTHER:
1854 			if (sched_param.sched_priority != 0) {
1855 				error = EINVAL;
1856 				goto out;
1857 			}
1858 			sched_param.sched_priority =
1859 			    PRI_MAX_TIMESHARE - PRI_MIN_TIMESHARE;
1860 			break;
1861 		case SCHED_FIFO:
1862 		case SCHED_RR:
1863 			if (sched_param.sched_priority < 1 ||
1864 			    sched_param.sched_priority >= LINUX_MAX_RT_PRIO) {
1865 				error = EINVAL;
1866 				goto out;
1867 			}
1868 			/*
1869 			 * Map [1, LINUX_MAX_RT_PRIO - 1] to
1870 			 * [0, RTP_PRIO_MAX - RTP_PRIO_MIN] (rounding down).
1871 			 */
1872 			sched_param.sched_priority =
1873 			    (sched_param.sched_priority - 1) *
1874 			    (RTP_PRIO_MAX - RTP_PRIO_MIN + 1) /
1875 			    (LINUX_MAX_RT_PRIO - 1);
1876 			break;
1877 		}
1878 	}
1879 
1880 	error = kern_sched_setparam(td, tdt, &sched_param);
1881 out:	PROC_UNLOCK(tdt->td_proc);
1882 	return (error);
1883 }
1884 
1885 int
1886 linux_sched_getparam(struct thread *td,
1887     struct linux_sched_getparam_args *uap)
1888 {
1889 	struct sched_param sched_param;
1890 	struct thread *tdt;
1891 	int error, policy;
1892 
1893 	tdt = linux_tdfind(td, uap->pid, -1);
1894 	if (tdt == NULL)
1895 		return (ESRCH);
1896 
1897 	error = kern_sched_getparam(td, tdt, &sched_param);
1898 	if (error) {
1899 		PROC_UNLOCK(tdt->td_proc);
1900 		return (error);
1901 	}
1902 
1903 	if (linux_map_sched_prio) {
1904 		error = kern_sched_getscheduler(td, tdt, &policy);
1905 		PROC_UNLOCK(tdt->td_proc);
1906 		if (error)
1907 			return (error);
1908 
1909 		switch (policy) {
1910 		case SCHED_OTHER:
1911 			sched_param.sched_priority = 0;
1912 			break;
1913 		case SCHED_FIFO:
1914 		case SCHED_RR:
1915 			/*
1916 			 * Map [0, RTP_PRIO_MAX - RTP_PRIO_MIN] to
1917 			 * [1, LINUX_MAX_RT_PRIO - 1] (rounding up).
1918 			 */
1919 			sched_param.sched_priority =
1920 			    (sched_param.sched_priority *
1921 			    (LINUX_MAX_RT_PRIO - 1) +
1922 			    (RTP_PRIO_MAX - RTP_PRIO_MIN - 1)) /
1923 			    (RTP_PRIO_MAX - RTP_PRIO_MIN) + 1;
1924 			break;
1925 		}
1926 	} else
1927 		PROC_UNLOCK(tdt->td_proc);
1928 
1929 	error = copyout(&sched_param, uap->param, sizeof(sched_param));
1930 	return (error);
1931 }
1932 
1933 /*
1934  * Get affinity of a process.
1935  */
1936 int
1937 linux_sched_getaffinity(struct thread *td,
1938     struct linux_sched_getaffinity_args *args)
1939 {
1940 	struct thread *tdt;
1941 	cpuset_t *mask;
1942 	size_t size;
1943 	int error;
1944 	id_t tid;
1945 
1946 	tdt = linux_tdfind(td, args->pid, -1);
1947 	if (tdt == NULL)
1948 		return (ESRCH);
1949 	tid = tdt->td_tid;
1950 	PROC_UNLOCK(tdt->td_proc);
1951 
1952 	mask = malloc(sizeof(cpuset_t), M_LINUX, M_WAITOK | M_ZERO);
1953 	size = min(args->len, sizeof(cpuset_t));
1954 	error = kern_cpuset_getaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
1955 	    tid, size, mask);
1956 	if (error == ERANGE)
1957 		error = EINVAL;
1958  	if (error == 0)
1959 		error = copyout(mask, args->user_mask_ptr, size);
1960 	if (error == 0)
1961 		td->td_retval[0] = size;
1962 	free(mask, M_LINUX);
1963 	return (error);
1964 }
1965 
1966 /*
1967  *  Set affinity of a process.
1968  */
1969 int
1970 linux_sched_setaffinity(struct thread *td,
1971     struct linux_sched_setaffinity_args *args)
1972 {
1973 	struct thread *tdt;
1974 	cpuset_t *mask;
1975 	int cpu, error;
1976 	size_t len;
1977 	id_t tid;
1978 
1979 	tdt = linux_tdfind(td, args->pid, -1);
1980 	if (tdt == NULL)
1981 		return (ESRCH);
1982 	tid = tdt->td_tid;
1983 	PROC_UNLOCK(tdt->td_proc);
1984 
1985 	len = min(args->len, sizeof(cpuset_t));
1986 	mask = malloc(sizeof(cpuset_t), M_TEMP, M_WAITOK | M_ZERO);;
1987 	error = copyin(args->user_mask_ptr, mask, len);
1988 	if (error != 0)
1989 		goto out;
1990 	/* Linux ignore high bits */
1991 	CPU_FOREACH_ISSET(cpu, mask)
1992 		if (cpu > mp_maxid)
1993 			CPU_CLR(cpu, mask);
1994 
1995 	error = kern_cpuset_setaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
1996 	    tid, mask);
1997 	if (error == EDEADLK)
1998 		error = EINVAL;
1999 out:
2000 	free(mask, M_TEMP);
2001 	return (error);
2002 }
2003 
2004 struct linux_rlimit64 {
2005 	uint64_t	rlim_cur;
2006 	uint64_t	rlim_max;
2007 };
2008 
2009 int
2010 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
2011 {
2012 	struct rlimit rlim, nrlim;
2013 	struct linux_rlimit64 lrlim;
2014 	struct proc *p;
2015 	u_int which;
2016 	int flags;
2017 	int error;
2018 
2019 	if (args->new == NULL && args->old != NULL) {
2020 		if (linux_get_dummy_limit(args->resource, &rlim)) {
2021 			lrlim.rlim_cur = rlim.rlim_cur;
2022 			lrlim.rlim_max = rlim.rlim_max;
2023 			return (copyout(&lrlim, args->old, sizeof(lrlim)));
2024 		}
2025 	}
2026 
2027 	if (args->resource >= LINUX_RLIM_NLIMITS)
2028 		return (EINVAL);
2029 
2030 	which = linux_to_bsd_resource[args->resource];
2031 	if (which == -1)
2032 		return (EINVAL);
2033 
2034 	if (args->new != NULL) {
2035 		/*
2036 		 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
2037 		 * rlim is unsigned 64-bit. FreeBSD treats negative limits
2038 		 * as INFINITY so we do not need a conversion even.
2039 		 */
2040 		error = copyin(args->new, &nrlim, sizeof(nrlim));
2041 		if (error != 0)
2042 			return (error);
2043 	}
2044 
2045 	flags = PGET_HOLD | PGET_NOTWEXIT;
2046 	if (args->new != NULL)
2047 		flags |= PGET_CANDEBUG;
2048 	else
2049 		flags |= PGET_CANSEE;
2050 	if (args->pid == 0) {
2051 		p = td->td_proc;
2052 		PHOLD(p);
2053 	} else {
2054 		error = pget(args->pid, flags, &p);
2055 		if (error != 0)
2056 			return (error);
2057 	}
2058 	if (args->old != NULL) {
2059 		PROC_LOCK(p);
2060 		lim_rlimit_proc(p, which, &rlim);
2061 		PROC_UNLOCK(p);
2062 		if (rlim.rlim_cur == RLIM_INFINITY)
2063 			lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2064 		else
2065 			lrlim.rlim_cur = rlim.rlim_cur;
2066 		if (rlim.rlim_max == RLIM_INFINITY)
2067 			lrlim.rlim_max = LINUX_RLIM_INFINITY;
2068 		else
2069 			lrlim.rlim_max = rlim.rlim_max;
2070 		error = copyout(&lrlim, args->old, sizeof(lrlim));
2071 		if (error != 0)
2072 			goto out;
2073 	}
2074 
2075 	if (args->new != NULL)
2076 		error = kern_proc_setrlimit(td, p, which, &nrlim);
2077 
2078  out:
2079 	PRELE(p);
2080 	return (error);
2081 }
2082 
2083 int
2084 linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2085 {
2086 	struct timespec ts, *tsp;
2087 	int error;
2088 
2089 	if (args->tsp != NULL) {
2090 		error = linux_get_timespec(&ts, args->tsp);
2091 		if (error != 0)
2092 			return (error);
2093 		tsp = &ts;
2094 	} else
2095 		tsp = NULL;
2096 
2097 	error = linux_common_pselect6(td, args->nfds, args->readfds,
2098 	    args->writefds, args->exceptfds, tsp, args->sig);
2099 
2100 	if (args->tsp != NULL)
2101 		linux_put_timespec(&ts, args->tsp);
2102 	return (error);
2103 }
2104 
2105 static int
2106 linux_common_pselect6(struct thread *td, l_int nfds, l_fd_set *readfds,
2107     l_fd_set *writefds, l_fd_set *exceptfds, struct timespec *tsp,
2108     l_uintptr_t *sig)
2109 {
2110 	struct timeval utv, tv0, tv1, *tvp;
2111 	struct l_pselect6arg lpse6;
2112 	sigset_t *ssp;
2113 	sigset_t ss;
2114 	int error;
2115 
2116 	ssp = NULL;
2117 	if (sig != NULL) {
2118 		error = copyin(sig, &lpse6, sizeof(lpse6));
2119 		if (error != 0)
2120 			return (error);
2121 		error = linux_copyin_sigset(td, PTRIN(lpse6.ss),
2122 		    lpse6.ss_len, &ss, &ssp);
2123 		if (error != 0)
2124 		    return (error);
2125 	} else
2126 		ssp = NULL;
2127 
2128 	/*
2129 	 * Currently glibc changes nanosecond number to microsecond.
2130 	 * This mean losing precision but for now it is hardly seen.
2131 	 */
2132 	if (tsp != NULL) {
2133 		TIMESPEC_TO_TIMEVAL(&utv, tsp);
2134 		if (itimerfix(&utv))
2135 			return (EINVAL);
2136 
2137 		microtime(&tv0);
2138 		tvp = &utv;
2139 	} else
2140 		tvp = NULL;
2141 
2142 	error = kern_pselect(td, nfds, readfds, writefds,
2143 	    exceptfds, tvp, ssp, LINUX_NFDBITS);
2144 
2145 	if (tsp != NULL) {
2146 		/*
2147 		 * Compute how much time was left of the timeout,
2148 		 * by subtracting the current time and the time
2149 		 * before we started the call, and subtracting
2150 		 * that result from the user-supplied value.
2151 		 */
2152 		microtime(&tv1);
2153 		timevalsub(&tv1, &tv0);
2154 		timevalsub(&utv, &tv1);
2155 		if (utv.tv_sec < 0)
2156 			timevalclear(&utv);
2157 		TIMEVAL_TO_TIMESPEC(&utv, tsp);
2158 	}
2159 	return (error);
2160 }
2161 
2162 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
2163 int
2164 linux_pselect6_time64(struct thread *td,
2165     struct linux_pselect6_time64_args *args)
2166 {
2167 	struct timespec ts, *tsp;
2168 	int error;
2169 
2170 	if (args->tsp != NULL) {
2171 		error = linux_get_timespec64(&ts, args->tsp);
2172 		if (error != 0)
2173 			return (error);
2174 		tsp = &ts;
2175 	} else
2176 		tsp = NULL;
2177 
2178 	error = linux_common_pselect6(td, args->nfds, args->readfds,
2179 	    args->writefds, args->exceptfds, tsp, args->sig);
2180 
2181 	if (args->tsp != NULL)
2182 		linux_put_timespec64(&ts, args->tsp);
2183 	return (error);
2184 }
2185 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
2186 
2187 int
2188 linux_ppoll(struct thread *td, struct linux_ppoll_args *args)
2189 {
2190 	struct timespec uts, *tsp;
2191 	int error;
2192 
2193 	if (args->tsp != NULL) {
2194 		error = linux_get_timespec(&uts, args->tsp);
2195 		if (error != 0)
2196 			return (error);
2197 		tsp = &uts;
2198 	} else
2199 		tsp = NULL;
2200 
2201 	error = linux_common_ppoll(td, args->fds, args->nfds, tsp,
2202 	    args->sset, args->ssize);
2203 	if (error == 0 && args->tsp != NULL)
2204 		error = linux_put_timespec(&uts, args->tsp);
2205 	return (error);
2206 }
2207 
2208 static int
2209 linux_common_ppoll(struct thread *td, struct pollfd *fds, uint32_t nfds,
2210     struct timespec *tsp, l_sigset_t *sset, l_size_t ssize)
2211 {
2212 	struct timespec ts0, ts1;
2213 	struct pollfd stackfds[32];
2214 	struct pollfd *kfds;
2215  	sigset_t *ssp;
2216  	sigset_t ss;
2217  	int error;
2218 
2219 	if (kern_poll_maxfds(nfds))
2220 		return (EINVAL);
2221 	if (sset != NULL) {
2222 		error = linux_copyin_sigset(td, sset, ssize, &ss, &ssp);
2223 		if (error != 0)
2224 		    return (error);
2225 	} else
2226 		ssp = NULL;
2227 	if (tsp != NULL)
2228 		nanotime(&ts0);
2229 
2230 	if (nfds > nitems(stackfds))
2231 		kfds = mallocarray(nfds, sizeof(*kfds), M_TEMP, M_WAITOK);
2232 	else
2233 		kfds = stackfds;
2234 	error = linux_pollin(td, kfds, fds, nfds);
2235 	if (error != 0)
2236 		goto out;
2237 
2238 	error = kern_poll_kfds(td, kfds, nfds, tsp, ssp);
2239 	if (error == 0)
2240 		error = linux_pollout(td, kfds, fds, nfds);
2241 
2242 	if (error == 0 && tsp != NULL) {
2243 		if (td->td_retval[0]) {
2244 			nanotime(&ts1);
2245 			timespecsub(&ts1, &ts0, &ts1);
2246 			timespecsub(tsp, &ts1, tsp);
2247 			if (tsp->tv_sec < 0)
2248 				timespecclear(tsp);
2249 		} else
2250 			timespecclear(tsp);
2251 	}
2252 
2253 out:
2254 	if (nfds > nitems(stackfds))
2255 		free(kfds, M_TEMP);
2256 	return (error);
2257 }
2258 
2259 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
2260 int
2261 linux_ppoll_time64(struct thread *td, struct linux_ppoll_time64_args *args)
2262 {
2263 	struct timespec uts, *tsp;
2264 	int error;
2265 
2266 	if (args->tsp != NULL) {
2267 		error = linux_get_timespec64(&uts, args->tsp);
2268 		if (error != 0)
2269 			return (error);
2270 		tsp = &uts;
2271 	} else
2272  		tsp = NULL;
2273 	error = linux_common_ppoll(td, args->fds, args->nfds, tsp,
2274 	    args->sset, args->ssize);
2275 	if (error == 0 && args->tsp != NULL)
2276 		error = linux_put_timespec64(&uts, args->tsp);
2277 	return (error);
2278 }
2279 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
2280 
2281 static int
2282 linux_pollin(struct thread *td, struct pollfd *fds, struct pollfd *ufds, u_int nfd)
2283 {
2284 	int error;
2285 	u_int i;
2286 
2287 	error = copyin(ufds, fds, nfd * sizeof(*fds));
2288 	if (error != 0)
2289 		return (error);
2290 
2291 	for (i = 0; i < nfd; i++) {
2292 		if (fds->events != 0)
2293 			linux_to_bsd_poll_events(td, fds->fd,
2294 			    fds->events, &fds->events);
2295 		fds++;
2296 	}
2297 	return (0);
2298 }
2299 
2300 static int
2301 linux_pollout(struct thread *td, struct pollfd *fds, struct pollfd *ufds, u_int nfd)
2302 {
2303 	int error = 0;
2304 	u_int i, n = 0;
2305 
2306 	for (i = 0; i < nfd; i++) {
2307 		if (fds->revents != 0) {
2308 			bsd_to_linux_poll_events(fds->revents,
2309 			    &fds->revents);
2310 			n++;
2311 		}
2312 		error = copyout(&fds->revents, &ufds->revents,
2313 		    sizeof(ufds->revents));
2314 		if (error)
2315 			return (error);
2316 		fds++;
2317 		ufds++;
2318 	}
2319 	td->td_retval[0] = n;
2320 	return (0);
2321 }
2322 
2323 static int
2324 linux_sched_rr_get_interval_common(struct thread *td, pid_t pid,
2325     struct timespec *ts)
2326 {
2327 	struct thread *tdt;
2328 	int error;
2329 
2330 	/*
2331 	 * According to man in case the invalid pid specified
2332 	 * EINVAL should be returned.
2333 	 */
2334 	if (pid < 0)
2335 		return (EINVAL);
2336 
2337 	tdt = linux_tdfind(td, pid, -1);
2338 	if (tdt == NULL)
2339 		return (ESRCH);
2340 
2341 	error = kern_sched_rr_get_interval_td(td, tdt, ts);
2342 	PROC_UNLOCK(tdt->td_proc);
2343 	return (error);
2344 }
2345 
2346 int
2347 linux_sched_rr_get_interval(struct thread *td,
2348     struct linux_sched_rr_get_interval_args *uap)
2349 {
2350 	struct timespec ts;
2351 	int error;
2352 
2353 	error = linux_sched_rr_get_interval_common(td, uap->pid, &ts);
2354 	if (error != 0)
2355 		return (error);
2356 	return (linux_put_timespec(&ts, uap->interval));
2357 }
2358 
2359 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
2360 int
2361 linux_sched_rr_get_interval_time64(struct thread *td,
2362     struct linux_sched_rr_get_interval_time64_args *uap)
2363 {
2364 	struct timespec ts;
2365 	int error;
2366 
2367 	error = linux_sched_rr_get_interval_common(td, uap->pid, &ts);
2368 	if (error != 0)
2369 		return (error);
2370 	return (linux_put_timespec64(&ts, uap->interval));
2371 }
2372 #endif
2373 
2374 /*
2375  * In case when the Linux thread is the initial thread in
2376  * the thread group thread id is equal to the process id.
2377  * Glibc depends on this magic (assert in pthread_getattr_np.c).
2378  */
2379 struct thread *
2380 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid)
2381 {
2382 	struct linux_emuldata *em;
2383 	struct thread *tdt;
2384 	struct proc *p;
2385 
2386 	tdt = NULL;
2387 	if (tid == 0 || tid == td->td_tid) {
2388 		if (pid != -1 && td->td_proc->p_pid != pid)
2389 			return (NULL);
2390 		PROC_LOCK(td->td_proc);
2391 		return (td);
2392 	} else if (tid > PID_MAX)
2393 		return (tdfind(tid, pid));
2394 
2395 	/*
2396 	 * Initial thread where the tid equal to the pid.
2397 	 */
2398 	p = pfind(tid);
2399 	if (p != NULL) {
2400 		if (SV_PROC_ABI(p) != SV_ABI_LINUX ||
2401 		    (pid != -1 && tid != pid)) {
2402 			/*
2403 			 * p is not a Linuxulator process.
2404 			 */
2405 			PROC_UNLOCK(p);
2406 			return (NULL);
2407 		}
2408 		FOREACH_THREAD_IN_PROC(p, tdt) {
2409 			em = em_find(tdt);
2410 			if (tid == em->em_tid)
2411 				return (tdt);
2412 		}
2413 		PROC_UNLOCK(p);
2414 	}
2415 	return (NULL);
2416 }
2417 
2418 void
2419 linux_to_bsd_waitopts(int options, int *bsdopts)
2420 {
2421 
2422 	if (options & LINUX_WNOHANG)
2423 		*bsdopts |= WNOHANG;
2424 	if (options & LINUX_WUNTRACED)
2425 		*bsdopts |= WUNTRACED;
2426 	if (options & LINUX_WEXITED)
2427 		*bsdopts |= WEXITED;
2428 	if (options & LINUX_WCONTINUED)
2429 		*bsdopts |= WCONTINUED;
2430 	if (options & LINUX_WNOWAIT)
2431 		*bsdopts |= WNOWAIT;
2432 
2433 	if (options & __WCLONE)
2434 		*bsdopts |= WLINUXCLONE;
2435 }
2436 
2437 int
2438 linux_getrandom(struct thread *td, struct linux_getrandom_args *args)
2439 {
2440 	struct uio uio;
2441 	struct iovec iov;
2442 	int error;
2443 
2444 	if (args->flags & ~(LINUX_GRND_NONBLOCK|LINUX_GRND_RANDOM))
2445 		return (EINVAL);
2446 	if (args->count > INT_MAX)
2447 		args->count = INT_MAX;
2448 
2449 	iov.iov_base = args->buf;
2450 	iov.iov_len = args->count;
2451 
2452 	uio.uio_iov = &iov;
2453 	uio.uio_iovcnt = 1;
2454 	uio.uio_resid = iov.iov_len;
2455 	uio.uio_segflg = UIO_USERSPACE;
2456 	uio.uio_rw = UIO_READ;
2457 	uio.uio_td = td;
2458 
2459 	error = read_random_uio(&uio, args->flags & LINUX_GRND_NONBLOCK);
2460 	if (error == 0)
2461 		td->td_retval[0] = args->count - uio.uio_resid;
2462 	return (error);
2463 }
2464 
2465 int
2466 linux_mincore(struct thread *td, struct linux_mincore_args *args)
2467 {
2468 
2469 	/* Needs to be page-aligned */
2470 	if (args->start & PAGE_MASK)
2471 		return (EINVAL);
2472 	return (kern_mincore(td, args->start, args->len, args->vec));
2473 }
2474 
2475 #define	SYSLOG_TAG	"<6>"
2476 
2477 int
2478 linux_syslog(struct thread *td, struct linux_syslog_args *args)
2479 {
2480 	char buf[128], *src, *dst;
2481 	u_int seq;
2482 	int buflen, error;
2483 
2484 	if (args->type != LINUX_SYSLOG_ACTION_READ_ALL) {
2485 		linux_msg(td, "syslog unsupported type 0x%x", args->type);
2486 		return (EINVAL);
2487 	}
2488 
2489 	if (args->len < 6) {
2490 		td->td_retval[0] = 0;
2491 		return (0);
2492 	}
2493 
2494 	error = priv_check(td, PRIV_MSGBUF);
2495 	if (error)
2496 		return (error);
2497 
2498 	mtx_lock(&msgbuf_lock);
2499 	msgbuf_peekbytes(msgbufp, NULL, 0, &seq);
2500 	mtx_unlock(&msgbuf_lock);
2501 
2502 	dst = args->buf;
2503 	error = copyout(&SYSLOG_TAG, dst, sizeof(SYSLOG_TAG));
2504 	/* The -1 is to skip the trailing '\0'. */
2505 	dst += sizeof(SYSLOG_TAG) - 1;
2506 
2507 	while (error == 0) {
2508 		mtx_lock(&msgbuf_lock);
2509 		buflen = msgbuf_peekbytes(msgbufp, buf, sizeof(buf), &seq);
2510 		mtx_unlock(&msgbuf_lock);
2511 
2512 		if (buflen == 0)
2513 			break;
2514 
2515 		for (src = buf; src < buf + buflen && error == 0; src++) {
2516 			if (*src == '\0')
2517 				continue;
2518 
2519 			if (dst >= args->buf + args->len)
2520 				goto out;
2521 
2522 			error = copyout(src, dst, 1);
2523 			dst++;
2524 
2525 			if (*src == '\n' && *(src + 1) != '<' &&
2526 			    dst + sizeof(SYSLOG_TAG) < args->buf + args->len) {
2527 				error = copyout(&SYSLOG_TAG,
2528 				    dst, sizeof(SYSLOG_TAG));
2529 				dst += sizeof(SYSLOG_TAG) - 1;
2530 			}
2531 		}
2532 	}
2533 out:
2534 	td->td_retval[0] = dst - args->buf;
2535 	return (error);
2536 }
2537 
2538 int
2539 linux_getcpu(struct thread *td, struct linux_getcpu_args *args)
2540 {
2541 	int cpu, error, node;
2542 
2543 	cpu = td->td_oncpu; /* Make sure it doesn't change during copyout(9) */
2544 	error = 0;
2545 	node = cpuid_to_pcpu[cpu]->pc_domain;
2546 
2547 	if (args->cpu != NULL)
2548 		error = copyout(&cpu, args->cpu, sizeof(l_int));
2549 	if (args->node != NULL)
2550 		error = copyout(&node, args->node, sizeof(l_int));
2551 	return (error);
2552 }
2553 
2554 #if defined(__i386__) || defined(__amd64__)
2555 int
2556 linux_poll(struct thread *td, struct linux_poll_args *args)
2557 {
2558 	struct timespec ts, *tsp;
2559 
2560 	if (args->timeout != INFTIM) {
2561 		if (args->timeout < 0)
2562 			return (EINVAL);
2563 		ts.tv_sec = args->timeout / 1000;
2564 		ts.tv_nsec = (args->timeout % 1000) * 1000000;
2565 		tsp = &ts;
2566 	} else
2567 		tsp = NULL;
2568 
2569 	return (linux_common_ppoll(td, args->fds, args->nfds,
2570 	    tsp, NULL, 0));
2571 }
2572 #endif /* __i386__ || __amd64__ */
2573 
2574 int
2575 linux_seccomp(struct thread *td, struct linux_seccomp_args *args)
2576 {
2577 
2578 	switch (args->op) {
2579 	case LINUX_SECCOMP_GET_ACTION_AVAIL:
2580 		return (EOPNOTSUPP);
2581 	default:
2582 		/*
2583 		 * Ignore unknown operations, just like Linux kernel built
2584 		 * without CONFIG_SECCOMP.
2585 		 */
2586 		return (EINVAL);
2587 	}
2588 }
2589 
2590 /*
2591  * Custom version of exec_copyin_args(), to copy out argument and environment
2592  * strings from the old process address space into the temporary string buffer.
2593  * Based on freebsd32_exec_copyin_args.
2594  */
2595 static int
2596 linux_exec_copyin_args(struct image_args *args, const char *fname,
2597     enum uio_seg segflg, l_uintptr_t *argv, l_uintptr_t *envv)
2598 {
2599 	char *argp, *envp;
2600 	l_uintptr_t *ptr, arg;
2601 	int error;
2602 
2603 	bzero(args, sizeof(*args));
2604 	if (argv == NULL)
2605 		return (EFAULT);
2606 
2607 	/*
2608 	 * Allocate demand-paged memory for the file name, argument, and
2609 	 * environment strings.
2610 	 */
2611 	error = exec_alloc_args(args);
2612 	if (error != 0)
2613 		return (error);
2614 
2615 	/*
2616 	 * Copy the file name.
2617 	 */
2618 	error = exec_args_add_fname(args, fname, segflg);
2619 	if (error != 0)
2620 		goto err_exit;
2621 
2622 	/*
2623 	 * extract arguments first
2624 	 */
2625 	ptr = argv;
2626 	for (;;) {
2627 		error = copyin(ptr++, &arg, sizeof(arg));
2628 		if (error)
2629 			goto err_exit;
2630 		if (arg == 0)
2631 			break;
2632 		argp = PTRIN(arg);
2633 		error = exec_args_add_arg(args, argp, UIO_USERSPACE);
2634 		if (error != 0)
2635 			goto err_exit;
2636 	}
2637 
2638 	/*
2639 	 * This comment is from Linux do_execveat_common:
2640 	 * When argv is empty, add an empty string ("") as argv[0] to
2641 	 * ensure confused userspace programs that start processing
2642 	 * from argv[1] won't end up walking envp.
2643 	 */
2644 	if (args->argc == 0 &&
2645 	    (error = exec_args_add_arg(args, "", UIO_SYSSPACE) != 0))
2646 		goto err_exit;
2647 
2648 	/*
2649 	 * extract environment strings
2650 	 */
2651 	if (envv) {
2652 		ptr = envv;
2653 		for (;;) {
2654 			error = copyin(ptr++, &arg, sizeof(arg));
2655 			if (error)
2656 				goto err_exit;
2657 			if (arg == 0)
2658 				break;
2659 			envp = PTRIN(arg);
2660 			error = exec_args_add_env(args, envp, UIO_USERSPACE);
2661 			if (error != 0)
2662 				goto err_exit;
2663 		}
2664 	}
2665 
2666 	return (0);
2667 
2668 err_exit:
2669 	exec_free_args(args);
2670 	return (error);
2671 }
2672 
2673 int
2674 linux_execve(struct thread *td, struct linux_execve_args *args)
2675 {
2676 	struct image_args eargs;
2677 	int error;
2678 
2679 	LINUX_CTR(execve);
2680 
2681 	error = linux_exec_copyin_args(&eargs, args->path, UIO_USERSPACE,
2682 	    args->argp, args->envp);
2683 	if (error == 0)
2684 		error = linux_common_execve(td, &eargs);
2685 	AUDIT_SYSCALL_EXIT(error == EJUSTRETURN ? 0 : error, td);
2686 	return (error);
2687 }
2688 
2689 static void
2690 linux_up_rtprio_if(struct thread *td1, struct rtprio *rtp)
2691 {
2692 	struct rtprio rtp2;
2693 
2694 	pri_to_rtp(td1, &rtp2);
2695 	if (rtp2.type <  rtp->type ||
2696 	    (rtp2.type == rtp->type &&
2697 	    rtp2.prio < rtp->prio)) {
2698 		rtp->type = rtp2.type;
2699 		rtp->prio = rtp2.prio;
2700 	}
2701 }
2702 
2703 #define	LINUX_PRIO_DIVIDER	RTP_PRIO_MAX / LINUX_IOPRIO_MAX
2704 
2705 static int
2706 linux_rtprio2ioprio(struct rtprio *rtp)
2707 {
2708 	int ioprio, prio;
2709 
2710 	switch (rtp->type) {
2711 	case RTP_PRIO_IDLE:
2712 		prio = RTP_PRIO_MIN;
2713 		ioprio = LINUX_IOPRIO_PRIO(LINUX_IOPRIO_CLASS_IDLE, prio);
2714 		break;
2715 	case RTP_PRIO_NORMAL:
2716 		prio = rtp->prio / LINUX_PRIO_DIVIDER;
2717 		ioprio = LINUX_IOPRIO_PRIO(LINUX_IOPRIO_CLASS_BE, prio);
2718 		break;
2719 	case RTP_PRIO_REALTIME:
2720 		prio = rtp->prio / LINUX_PRIO_DIVIDER;
2721 		ioprio = LINUX_IOPRIO_PRIO(LINUX_IOPRIO_CLASS_RT, prio);
2722 		break;
2723 	default:
2724 		prio = RTP_PRIO_MIN;
2725 		ioprio = LINUX_IOPRIO_PRIO(LINUX_IOPRIO_CLASS_NONE, prio);
2726 		break;
2727 	}
2728 	return (ioprio);
2729 }
2730 
2731 static int
2732 linux_ioprio2rtprio(int ioprio, struct rtprio *rtp)
2733 {
2734 
2735 	switch (LINUX_IOPRIO_PRIO_CLASS(ioprio)) {
2736 	case LINUX_IOPRIO_CLASS_IDLE:
2737 		rtp->prio = RTP_PRIO_MIN;
2738 		rtp->type = RTP_PRIO_IDLE;
2739 		break;
2740 	case LINUX_IOPRIO_CLASS_BE:
2741 		rtp->prio = LINUX_IOPRIO_PRIO_DATA(ioprio) * LINUX_PRIO_DIVIDER;
2742 		rtp->type = RTP_PRIO_NORMAL;
2743 		break;
2744 	case LINUX_IOPRIO_CLASS_RT:
2745 		rtp->prio = LINUX_IOPRIO_PRIO_DATA(ioprio) * LINUX_PRIO_DIVIDER;
2746 		rtp->type = RTP_PRIO_REALTIME;
2747 		break;
2748 	default:
2749 		return (EINVAL);
2750 	}
2751 	return (0);
2752 }
2753 #undef LINUX_PRIO_DIVIDER
2754 
2755 int
2756 linux_ioprio_get(struct thread *td, struct linux_ioprio_get_args *args)
2757 {
2758 	struct thread *td1;
2759 	struct rtprio rtp;
2760 	struct pgrp *pg;
2761 	struct proc *p;
2762 	int error, found;
2763 
2764 	p = NULL;
2765 	td1 = NULL;
2766 	error = 0;
2767 	found = 0;
2768 	rtp.type = RTP_PRIO_IDLE;
2769 	rtp.prio = RTP_PRIO_MAX;
2770 	switch (args->which) {
2771 	case LINUX_IOPRIO_WHO_PROCESS:
2772 		if (args->who == 0) {
2773 			td1 = td;
2774 			p = td1->td_proc;
2775 			PROC_LOCK(p);
2776 		} else if (args->who > PID_MAX) {
2777 			td1 = linux_tdfind(td, args->who, -1);
2778 			if (td1 != NULL)
2779 				p = td1->td_proc;
2780 		} else
2781 			p = pfind(args->who);
2782 		if (p == NULL)
2783 			return (ESRCH);
2784 		if ((error = p_cansee(td, p))) {
2785 			PROC_UNLOCK(p);
2786 			break;
2787 		}
2788 		if (td1 != NULL) {
2789 			pri_to_rtp(td1, &rtp);
2790 		} else {
2791 			FOREACH_THREAD_IN_PROC(p, td1) {
2792 				linux_up_rtprio_if(td1, &rtp);
2793 			}
2794 		}
2795 		found++;
2796 		PROC_UNLOCK(p);
2797 		break;
2798 	case LINUX_IOPRIO_WHO_PGRP:
2799 		sx_slock(&proctree_lock);
2800 		if (args->who == 0) {
2801 			pg = td->td_proc->p_pgrp;
2802 			PGRP_LOCK(pg);
2803 		} else {
2804 			pg = pgfind(args->who);
2805 			if (pg == NULL) {
2806 				sx_sunlock(&proctree_lock);
2807 				error = ESRCH;
2808 				break;
2809 			}
2810 		}
2811 		sx_sunlock(&proctree_lock);
2812 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
2813 			PROC_LOCK(p);
2814 			if (p->p_state == PRS_NORMAL &&
2815 			    p_cansee(td, p) == 0) {
2816 				FOREACH_THREAD_IN_PROC(p, td1) {
2817 					linux_up_rtprio_if(td1, &rtp);
2818 					found++;
2819 				}
2820 			}
2821 			PROC_UNLOCK(p);
2822 		}
2823 		PGRP_UNLOCK(pg);
2824 		break;
2825 	case LINUX_IOPRIO_WHO_USER:
2826 		if (args->who == 0)
2827 			args->who = td->td_ucred->cr_uid;
2828 		sx_slock(&allproc_lock);
2829 		FOREACH_PROC_IN_SYSTEM(p) {
2830 			PROC_LOCK(p);
2831 			if (p->p_state == PRS_NORMAL &&
2832 			    p->p_ucred->cr_uid == args->who &&
2833 			    p_cansee(td, p) == 0) {
2834 				FOREACH_THREAD_IN_PROC(p, td1) {
2835 					linux_up_rtprio_if(td1, &rtp);
2836 					found++;
2837 				}
2838 			}
2839 			PROC_UNLOCK(p);
2840 		}
2841 		sx_sunlock(&allproc_lock);
2842 		break;
2843 	default:
2844 		error = EINVAL;
2845 		break;
2846 	}
2847 	if (error == 0) {
2848 		if (found != 0)
2849 			td->td_retval[0] = linux_rtprio2ioprio(&rtp);
2850 		else
2851 			error = ESRCH;
2852 	}
2853 	return (error);
2854 }
2855 
2856 int
2857 linux_ioprio_set(struct thread *td, struct linux_ioprio_set_args *args)
2858 {
2859 	struct thread *td1;
2860 	struct rtprio rtp;
2861 	struct pgrp *pg;
2862 	struct proc *p;
2863 	int error;
2864 
2865 	if ((error = linux_ioprio2rtprio(args->ioprio, &rtp)) != 0)
2866 		return (error);
2867 	/* Attempts to set high priorities (REALTIME) require su privileges. */
2868 	if (RTP_PRIO_BASE(rtp.type) == RTP_PRIO_REALTIME &&
2869 	    (error = priv_check(td, PRIV_SCHED_RTPRIO)) != 0)
2870 		return (error);
2871 
2872 	p = NULL;
2873 	td1 = NULL;
2874 	switch (args->which) {
2875 	case LINUX_IOPRIO_WHO_PROCESS:
2876 		if (args->who == 0) {
2877 			td1 = td;
2878 			p = td1->td_proc;
2879 			PROC_LOCK(p);
2880 		} else if (args->who > PID_MAX) {
2881 			td1 = linux_tdfind(td, args->who, -1);
2882 			if (td1 != NULL)
2883 				p = td1->td_proc;
2884 		} else
2885 			p = pfind(args->who);
2886 		if (p == NULL)
2887 			return (ESRCH);
2888 		if ((error = p_cansched(td, p))) {
2889 			PROC_UNLOCK(p);
2890 			break;
2891 		}
2892 		if (td1 != NULL) {
2893 			error = rtp_to_pri(&rtp, td1);
2894 		} else {
2895 			FOREACH_THREAD_IN_PROC(p, td1) {
2896 				if ((error = rtp_to_pri(&rtp, td1)) != 0)
2897 					break;
2898 			}
2899 		}
2900 		PROC_UNLOCK(p);
2901 		break;
2902 	case LINUX_IOPRIO_WHO_PGRP:
2903 		sx_slock(&proctree_lock);
2904 		if (args->who == 0) {
2905 			pg = td->td_proc->p_pgrp;
2906 			PGRP_LOCK(pg);
2907 		} else {
2908 			pg = pgfind(args->who);
2909 			if (pg == NULL) {
2910 				sx_sunlock(&proctree_lock);
2911 				error = ESRCH;
2912 				break;
2913 			}
2914 		}
2915 		sx_sunlock(&proctree_lock);
2916 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
2917 			PROC_LOCK(p);
2918 			if (p->p_state == PRS_NORMAL &&
2919 			    p_cansched(td, p) == 0) {
2920 				FOREACH_THREAD_IN_PROC(p, td1) {
2921 					if ((error = rtp_to_pri(&rtp, td1)) != 0)
2922 						break;
2923 				}
2924 			}
2925 			PROC_UNLOCK(p);
2926 			if (error != 0)
2927 				break;
2928 		}
2929 		PGRP_UNLOCK(pg);
2930 		break;
2931 	case LINUX_IOPRIO_WHO_USER:
2932 		if (args->who == 0)
2933 			args->who = td->td_ucred->cr_uid;
2934 		sx_slock(&allproc_lock);
2935 		FOREACH_PROC_IN_SYSTEM(p) {
2936 			PROC_LOCK(p);
2937 			if (p->p_state == PRS_NORMAL &&
2938 			    p->p_ucred->cr_uid == args->who &&
2939 			    p_cansched(td, p) == 0) {
2940 				FOREACH_THREAD_IN_PROC(p, td1) {
2941 					if ((error = rtp_to_pri(&rtp, td1)) != 0)
2942 						break;
2943 				}
2944 			}
2945 			PROC_UNLOCK(p);
2946 			if (error != 0)
2947 				break;
2948 		}
2949 		sx_sunlock(&allproc_lock);
2950 		break;
2951 	default:
2952 		error = EINVAL;
2953 		break;
2954 	}
2955 	return (error);
2956 }
2957