xref: /freebsd/sys/compat/linux/linux_misc.c (revision f30188c4680a85126e793de157b851bf5ee47529)
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/mqueue.h>
40 #include <sys/mutex.h>
41 #include <sys/poll.h>
42 #include <sys/priv.h>
43 #include <sys/proc.h>
44 #include <sys/procctl.h>
45 #include <sys/reboot.h>
46 #include <sys/random.h>
47 #include <sys/resourcevar.h>
48 #include <sys/rtprio.h>
49 #include <sys/sched.h>
50 #include <sys/smp.h>
51 #include <sys/stat.h>
52 #include <sys/syscallsubr.h>
53 #include <sys/sysctl.h>
54 #include <sys/sysent.h>
55 #include <sys/sysproto.h>
56 #include <sys/time.h>
57 #include <sys/vmmeter.h>
58 #include <sys/vnode.h>
59 
60 #include <security/audit/audit.h>
61 #include <security/mac/mac_framework.h>
62 
63 #include <vm/pmap.h>
64 #include <vm/vm_map.h>
65 #include <vm/swap_pager.h>
66 
67 #ifdef COMPAT_LINUX32
68 #include <machine/../linux32/linux.h>
69 #include <machine/../linux32/linux32_proto.h>
70 #else
71 #include <machine/../linux/linux.h>
72 #include <machine/../linux/linux_proto.h>
73 #endif
74 
75 #include <compat/linux/linux_common.h>
76 #include <compat/linux/linux_dtrace.h>
77 #include <compat/linux/linux_file.h>
78 #include <compat/linux/linux_mib.h>
79 #include <compat/linux/linux_mmap.h>
80 #include <compat/linux/linux_signal.h>
81 #include <compat/linux/linux_time.h>
82 #include <compat/linux/linux_util.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(struct thread *td, l_uint resource, struct rlimit *rlim)
1130 {
1131 	ssize_t size;
1132 	int res, error;
1133 
1134 	if (linux_dummy_rlimits == 0)
1135 		return (false);
1136 
1137 	switch (resource) {
1138 	case LINUX_RLIMIT_LOCKS:
1139 	case LINUX_RLIMIT_RTTIME:
1140 		rlim->rlim_cur = LINUX_RLIM_INFINITY;
1141 		rlim->rlim_max = LINUX_RLIM_INFINITY;
1142 		return (true);
1143 	case LINUX_RLIMIT_NICE:
1144 	case LINUX_RLIMIT_RTPRIO:
1145 		rlim->rlim_cur = 0;
1146 		rlim->rlim_max = 0;
1147 		return (true);
1148 	case LINUX_RLIMIT_SIGPENDING:
1149 		error = kernel_sysctlbyname(td,
1150 		    "kern.sigqueue.max_pending_per_proc",
1151 		    &res, &size, 0, 0, 0, 0);
1152 		if (error != 0)
1153 			return (false);
1154 		rlim->rlim_cur = res;
1155 		rlim->rlim_max = res;
1156 		return (true);
1157 	case LINUX_RLIMIT_MSGQUEUE:
1158 		error = kernel_sysctlbyname(td,
1159 		    "kern.ipc.msgmnb", &res, &size, 0, 0, 0, 0);
1160 		if (error != 0)
1161 			return (false);
1162 		rlim->rlim_cur = res;
1163 		rlim->rlim_max = res;
1164 		return (true);
1165 	default:
1166 		return (false);
1167 	}
1168 }
1169 
1170 int
1171 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1172 {
1173 	struct rlimit bsd_rlim;
1174 	struct l_rlimit rlim;
1175 	u_int which;
1176 	int error;
1177 
1178 	if (args->resource >= LINUX_RLIM_NLIMITS)
1179 		return (EINVAL);
1180 
1181 	which = linux_to_bsd_resource[args->resource];
1182 	if (which == -1)
1183 		return (EINVAL);
1184 
1185 	error = copyin(args->rlim, &rlim, sizeof(rlim));
1186 	if (error)
1187 		return (error);
1188 
1189 	bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1190 	bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1191 	return (kern_setrlimit(td, which, &bsd_rlim));
1192 }
1193 
1194 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1195 int
1196 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1197 {
1198 	struct l_rlimit rlim;
1199 	struct rlimit bsd_rlim;
1200 	u_int which;
1201 
1202 	if (linux_get_dummy_limit(td, args->resource, &bsd_rlim)) {
1203 		rlim.rlim_cur = bsd_rlim.rlim_cur;
1204 		rlim.rlim_max = bsd_rlim.rlim_max;
1205 		return (copyout(&rlim, args->rlim, sizeof(rlim)));
1206 	}
1207 
1208 	if (args->resource >= LINUX_RLIM_NLIMITS)
1209 		return (EINVAL);
1210 
1211 	which = linux_to_bsd_resource[args->resource];
1212 	if (which == -1)
1213 		return (EINVAL);
1214 
1215 	lim_rlimit(td, which, &bsd_rlim);
1216 
1217 #ifdef COMPAT_LINUX32
1218 	rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1219 	if (rlim.rlim_cur == UINT_MAX)
1220 		rlim.rlim_cur = INT_MAX;
1221 	rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1222 	if (rlim.rlim_max == UINT_MAX)
1223 		rlim.rlim_max = INT_MAX;
1224 #else
1225 	rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1226 	if (rlim.rlim_cur == ULONG_MAX)
1227 		rlim.rlim_cur = LONG_MAX;
1228 	rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1229 	if (rlim.rlim_max == ULONG_MAX)
1230 		rlim.rlim_max = LONG_MAX;
1231 #endif
1232 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1233 }
1234 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1235 
1236 int
1237 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1238 {
1239 	struct l_rlimit rlim;
1240 	struct rlimit bsd_rlim;
1241 	u_int which;
1242 
1243 	if (linux_get_dummy_limit(td, args->resource, &bsd_rlim)) {
1244 		rlim.rlim_cur = bsd_rlim.rlim_cur;
1245 		rlim.rlim_max = bsd_rlim.rlim_max;
1246 		return (copyout(&rlim, args->rlim, sizeof(rlim)));
1247 	}
1248 
1249 	if (args->resource >= LINUX_RLIM_NLIMITS)
1250 		return (EINVAL);
1251 
1252 	which = linux_to_bsd_resource[args->resource];
1253 	if (which == -1)
1254 		return (EINVAL);
1255 
1256 	lim_rlimit(td, which, &bsd_rlim);
1257 
1258 	rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1259 	rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1260 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1261 }
1262 
1263 int
1264 linux_sched_setscheduler(struct thread *td,
1265     struct linux_sched_setscheduler_args *args)
1266 {
1267 	struct sched_param sched_param;
1268 	struct thread *tdt;
1269 	int error, policy;
1270 
1271 	switch (args->policy) {
1272 	case LINUX_SCHED_OTHER:
1273 		policy = SCHED_OTHER;
1274 		break;
1275 	case LINUX_SCHED_FIFO:
1276 		policy = SCHED_FIFO;
1277 		break;
1278 	case LINUX_SCHED_RR:
1279 		policy = SCHED_RR;
1280 		break;
1281 	default:
1282 		return (EINVAL);
1283 	}
1284 
1285 	error = copyin(args->param, &sched_param, sizeof(sched_param));
1286 	if (error)
1287 		return (error);
1288 
1289 	if (linux_map_sched_prio) {
1290 		switch (policy) {
1291 		case SCHED_OTHER:
1292 			if (sched_param.sched_priority != 0)
1293 				return (EINVAL);
1294 
1295 			sched_param.sched_priority =
1296 			    PRI_MAX_TIMESHARE - PRI_MIN_TIMESHARE;
1297 			break;
1298 		case SCHED_FIFO:
1299 		case SCHED_RR:
1300 			if (sched_param.sched_priority < 1 ||
1301 			    sched_param.sched_priority >= LINUX_MAX_RT_PRIO)
1302 				return (EINVAL);
1303 
1304 			/*
1305 			 * Map [1, LINUX_MAX_RT_PRIO - 1] to
1306 			 * [0, RTP_PRIO_MAX - RTP_PRIO_MIN] (rounding down).
1307 			 */
1308 			sched_param.sched_priority =
1309 			    (sched_param.sched_priority - 1) *
1310 			    (RTP_PRIO_MAX - RTP_PRIO_MIN + 1) /
1311 			    (LINUX_MAX_RT_PRIO - 1);
1312 			break;
1313 		}
1314 	}
1315 
1316 	tdt = linux_tdfind(td, args->pid, -1);
1317 	if (tdt == NULL)
1318 		return (ESRCH);
1319 
1320 	error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1321 	PROC_UNLOCK(tdt->td_proc);
1322 	return (error);
1323 }
1324 
1325 int
1326 linux_sched_getscheduler(struct thread *td,
1327     struct linux_sched_getscheduler_args *args)
1328 {
1329 	struct thread *tdt;
1330 	int error, policy;
1331 
1332 	tdt = linux_tdfind(td, args->pid, -1);
1333 	if (tdt == NULL)
1334 		return (ESRCH);
1335 
1336 	error = kern_sched_getscheduler(td, tdt, &policy);
1337 	PROC_UNLOCK(tdt->td_proc);
1338 
1339 	switch (policy) {
1340 	case SCHED_OTHER:
1341 		td->td_retval[0] = LINUX_SCHED_OTHER;
1342 		break;
1343 	case SCHED_FIFO:
1344 		td->td_retval[0] = LINUX_SCHED_FIFO;
1345 		break;
1346 	case SCHED_RR:
1347 		td->td_retval[0] = LINUX_SCHED_RR;
1348 		break;
1349 	}
1350 	return (error);
1351 }
1352 
1353 int
1354 linux_sched_get_priority_max(struct thread *td,
1355     struct linux_sched_get_priority_max_args *args)
1356 {
1357 	struct sched_get_priority_max_args bsd;
1358 
1359 	if (linux_map_sched_prio) {
1360 		switch (args->policy) {
1361 		case LINUX_SCHED_OTHER:
1362 			td->td_retval[0] = 0;
1363 			return (0);
1364 		case LINUX_SCHED_FIFO:
1365 		case LINUX_SCHED_RR:
1366 			td->td_retval[0] = LINUX_MAX_RT_PRIO - 1;
1367 			return (0);
1368 		default:
1369 			return (EINVAL);
1370 		}
1371 	}
1372 
1373 	switch (args->policy) {
1374 	case LINUX_SCHED_OTHER:
1375 		bsd.policy = SCHED_OTHER;
1376 		break;
1377 	case LINUX_SCHED_FIFO:
1378 		bsd.policy = SCHED_FIFO;
1379 		break;
1380 	case LINUX_SCHED_RR:
1381 		bsd.policy = SCHED_RR;
1382 		break;
1383 	default:
1384 		return (EINVAL);
1385 	}
1386 	return (sys_sched_get_priority_max(td, &bsd));
1387 }
1388 
1389 int
1390 linux_sched_get_priority_min(struct thread *td,
1391     struct linux_sched_get_priority_min_args *args)
1392 {
1393 	struct sched_get_priority_min_args bsd;
1394 
1395 	if (linux_map_sched_prio) {
1396 		switch (args->policy) {
1397 		case LINUX_SCHED_OTHER:
1398 			td->td_retval[0] = 0;
1399 			return (0);
1400 		case LINUX_SCHED_FIFO:
1401 		case LINUX_SCHED_RR:
1402 			td->td_retval[0] = 1;
1403 			return (0);
1404 		default:
1405 			return (EINVAL);
1406 		}
1407 	}
1408 
1409 	switch (args->policy) {
1410 	case LINUX_SCHED_OTHER:
1411 		bsd.policy = SCHED_OTHER;
1412 		break;
1413 	case LINUX_SCHED_FIFO:
1414 		bsd.policy = SCHED_FIFO;
1415 		break;
1416 	case LINUX_SCHED_RR:
1417 		bsd.policy = SCHED_RR;
1418 		break;
1419 	default:
1420 		return (EINVAL);
1421 	}
1422 	return (sys_sched_get_priority_min(td, &bsd));
1423 }
1424 
1425 #define REBOOT_CAD_ON	0x89abcdef
1426 #define REBOOT_CAD_OFF	0
1427 #define REBOOT_HALT	0xcdef0123
1428 #define REBOOT_RESTART	0x01234567
1429 #define REBOOT_RESTART2	0xA1B2C3D4
1430 #define REBOOT_POWEROFF	0x4321FEDC
1431 #define REBOOT_MAGIC1	0xfee1dead
1432 #define REBOOT_MAGIC2	0x28121969
1433 #define REBOOT_MAGIC2A	0x05121996
1434 #define REBOOT_MAGIC2B	0x16041998
1435 
1436 int
1437 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1438 {
1439 	struct reboot_args bsd_args;
1440 
1441 	if (args->magic1 != REBOOT_MAGIC1)
1442 		return (EINVAL);
1443 
1444 	switch (args->magic2) {
1445 	case REBOOT_MAGIC2:
1446 	case REBOOT_MAGIC2A:
1447 	case REBOOT_MAGIC2B:
1448 		break;
1449 	default:
1450 		return (EINVAL);
1451 	}
1452 
1453 	switch (args->cmd) {
1454 	case REBOOT_CAD_ON:
1455 	case REBOOT_CAD_OFF:
1456 		return (priv_check(td, PRIV_REBOOT));
1457 	case REBOOT_HALT:
1458 		bsd_args.opt = RB_HALT;
1459 		break;
1460 	case REBOOT_RESTART:
1461 	case REBOOT_RESTART2:
1462 		bsd_args.opt = 0;
1463 		break;
1464 	case REBOOT_POWEROFF:
1465 		bsd_args.opt = RB_POWEROFF;
1466 		break;
1467 	default:
1468 		return (EINVAL);
1469 	}
1470 	return (sys_reboot(td, &bsd_args));
1471 }
1472 
1473 int
1474 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1475 {
1476 
1477 	td->td_retval[0] = td->td_proc->p_pid;
1478 
1479 	return (0);
1480 }
1481 
1482 int
1483 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1484 {
1485 	struct linux_emuldata *em;
1486 
1487 	em = em_find(td);
1488 	KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1489 
1490 	td->td_retval[0] = em->em_tid;
1491 
1492 	return (0);
1493 }
1494 
1495 int
1496 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1497 {
1498 
1499 	td->td_retval[0] = kern_getppid(td);
1500 	return (0);
1501 }
1502 
1503 int
1504 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1505 {
1506 
1507 	td->td_retval[0] = td->td_ucred->cr_rgid;
1508 	return (0);
1509 }
1510 
1511 int
1512 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1513 {
1514 
1515 	td->td_retval[0] = td->td_ucred->cr_ruid;
1516 	return (0);
1517 }
1518 
1519 int
1520 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1521 {
1522 
1523 	return (kern_getsid(td, args->pid));
1524 }
1525 
1526 int
1527 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1528 {
1529 	int error;
1530 
1531 	error = kern_getpriority(td, args->which, args->who);
1532 	td->td_retval[0] = 20 - td->td_retval[0];
1533 	return (error);
1534 }
1535 
1536 int
1537 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1538 {
1539 	int name[2];
1540 
1541 	name[0] = CTL_KERN;
1542 	name[1] = KERN_HOSTNAME;
1543 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1544 	    args->len, 0, 0));
1545 }
1546 
1547 int
1548 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1549 {
1550 	int name[2];
1551 
1552 	name[0] = CTL_KERN;
1553 	name[1] = KERN_NISDOMAINNAME;
1554 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1555 	    args->len, 0, 0));
1556 }
1557 
1558 int
1559 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1560 {
1561 
1562 	LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1563 	    args->error_code);
1564 
1565 	/*
1566 	 * XXX: we should send a signal to the parent if
1567 	 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1568 	 * as it doesnt occur often.
1569 	 */
1570 	exit1(td, args->error_code, 0);
1571 		/* NOTREACHED */
1572 }
1573 
1574 #define _LINUX_CAPABILITY_VERSION_1  0x19980330
1575 #define _LINUX_CAPABILITY_VERSION_2  0x20071026
1576 #define _LINUX_CAPABILITY_VERSION_3  0x20080522
1577 
1578 struct l_user_cap_header {
1579 	l_int	version;
1580 	l_int	pid;
1581 };
1582 
1583 struct l_user_cap_data {
1584 	l_int	effective;
1585 	l_int	permitted;
1586 	l_int	inheritable;
1587 };
1588 
1589 int
1590 linux_capget(struct thread *td, struct linux_capget_args *uap)
1591 {
1592 	struct l_user_cap_header luch;
1593 	struct l_user_cap_data lucd[2];
1594 	int error, u32s;
1595 
1596 	if (uap->hdrp == NULL)
1597 		return (EFAULT);
1598 
1599 	error = copyin(uap->hdrp, &luch, sizeof(luch));
1600 	if (error != 0)
1601 		return (error);
1602 
1603 	switch (luch.version) {
1604 	case _LINUX_CAPABILITY_VERSION_1:
1605 		u32s = 1;
1606 		break;
1607 	case _LINUX_CAPABILITY_VERSION_2:
1608 	case _LINUX_CAPABILITY_VERSION_3:
1609 		u32s = 2;
1610 		break;
1611 	default:
1612 		luch.version = _LINUX_CAPABILITY_VERSION_1;
1613 		error = copyout(&luch, uap->hdrp, sizeof(luch));
1614 		if (error)
1615 			return (error);
1616 		return (EINVAL);
1617 	}
1618 
1619 	if (luch.pid)
1620 		return (EPERM);
1621 
1622 	if (uap->datap) {
1623 		/*
1624 		 * The current implementation doesn't support setting
1625 		 * a capability (it's essentially a stub) so indicate
1626 		 * that no capabilities are currently set or available
1627 		 * to request.
1628 		 */
1629 		memset(&lucd, 0, u32s * sizeof(lucd[0]));
1630 		error = copyout(&lucd, uap->datap, u32s * sizeof(lucd[0]));
1631 	}
1632 
1633 	return (error);
1634 }
1635 
1636 int
1637 linux_capset(struct thread *td, struct linux_capset_args *uap)
1638 {
1639 	struct l_user_cap_header luch;
1640 	struct l_user_cap_data lucd[2];
1641 	int error, i, u32s;
1642 
1643 	if (uap->hdrp == NULL || uap->datap == NULL)
1644 		return (EFAULT);
1645 
1646 	error = copyin(uap->hdrp, &luch, sizeof(luch));
1647 	if (error != 0)
1648 		return (error);
1649 
1650 	switch (luch.version) {
1651 	case _LINUX_CAPABILITY_VERSION_1:
1652 		u32s = 1;
1653 		break;
1654 	case _LINUX_CAPABILITY_VERSION_2:
1655 	case _LINUX_CAPABILITY_VERSION_3:
1656 		u32s = 2;
1657 		break;
1658 	default:
1659 		luch.version = _LINUX_CAPABILITY_VERSION_1;
1660 		error = copyout(&luch, uap->hdrp, sizeof(luch));
1661 		if (error)
1662 			return (error);
1663 		return (EINVAL);
1664 	}
1665 
1666 	if (luch.pid)
1667 		return (EPERM);
1668 
1669 	error = copyin(uap->datap, &lucd, u32s * sizeof(lucd[0]));
1670 	if (error != 0)
1671 		return (error);
1672 
1673 	/* We currently don't support setting any capabilities. */
1674 	for (i = 0; i < u32s; i++) {
1675 		if (lucd[i].effective || lucd[i].permitted ||
1676 		    lucd[i].inheritable) {
1677 			linux_msg(td,
1678 			    "capset[%d] effective=0x%x, permitted=0x%x, "
1679 			    "inheritable=0x%x is not implemented", i,
1680 			    (int)lucd[i].effective, (int)lucd[i].permitted,
1681 			    (int)lucd[i].inheritable);
1682 			return (EPERM);
1683 		}
1684 	}
1685 
1686 	return (0);
1687 }
1688 
1689 int
1690 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1691 {
1692 	int error = 0, max_size, arg;
1693 	struct proc *p = td->td_proc;
1694 	char comm[LINUX_MAX_COMM_LEN];
1695 	int pdeath_signal, trace_state;
1696 
1697 	switch (args->option) {
1698 	case LINUX_PR_SET_PDEATHSIG:
1699 		if (!LINUX_SIG_VALID(args->arg2))
1700 			return (EINVAL);
1701 		pdeath_signal = linux_to_bsd_signal(args->arg2);
1702 		return (kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_CTL,
1703 		    &pdeath_signal));
1704 	case LINUX_PR_GET_PDEATHSIG:
1705 		error = kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_STATUS,
1706 		    &pdeath_signal);
1707 		if (error != 0)
1708 			return (error);
1709 		pdeath_signal = bsd_to_linux_signal(pdeath_signal);
1710 		return (copyout(&pdeath_signal,
1711 		    (void *)(register_t)args->arg2,
1712 		    sizeof(pdeath_signal)));
1713 	/*
1714 	 * In Linux, this flag controls if set[gu]id processes can coredump.
1715 	 * There are additional semantics imposed on processes that cannot
1716 	 * coredump:
1717 	 * - Such processes can not be ptraced.
1718 	 * - There are some semantics around ownership of process-related files
1719 	 *   in the /proc namespace.
1720 	 *
1721 	 * In FreeBSD, we can (and by default, do) disable setuid coredump
1722 	 * system-wide with 'sugid_coredump.'  We control tracability on a
1723 	 * per-process basis with the procctl PROC_TRACE (=> P2_NOTRACE flag).
1724 	 * By happy coincidence, P2_NOTRACE also prevents coredumping.  So the
1725 	 * procctl is roughly analogous to Linux's DUMPABLE.
1726 	 *
1727 	 * So, proxy these knobs to the corresponding PROC_TRACE setting.
1728 	 */
1729 	case LINUX_PR_GET_DUMPABLE:
1730 		error = kern_procctl(td, P_PID, p->p_pid, PROC_TRACE_STATUS,
1731 		    &trace_state);
1732 		if (error != 0)
1733 			return (error);
1734 		td->td_retval[0] = (trace_state != -1);
1735 		return (0);
1736 	case LINUX_PR_SET_DUMPABLE:
1737 		/*
1738 		 * It is only valid for userspace to set one of these two
1739 		 * flags, and only one at a time.
1740 		 */
1741 		switch (args->arg2) {
1742 		case LINUX_SUID_DUMP_DISABLE:
1743 			trace_state = PROC_TRACE_CTL_DISABLE_EXEC;
1744 			break;
1745 		case LINUX_SUID_DUMP_USER:
1746 			trace_state = PROC_TRACE_CTL_ENABLE;
1747 			break;
1748 		default:
1749 			return (EINVAL);
1750 		}
1751 		return (kern_procctl(td, P_PID, p->p_pid, PROC_TRACE_CTL,
1752 		    &trace_state));
1753 	case LINUX_PR_GET_KEEPCAPS:
1754 		/*
1755 		 * Indicate that we always clear the effective and
1756 		 * permitted capability sets when the user id becomes
1757 		 * non-zero (actually the capability sets are simply
1758 		 * always zero in the current implementation).
1759 		 */
1760 		td->td_retval[0] = 0;
1761 		break;
1762 	case LINUX_PR_SET_KEEPCAPS:
1763 		/*
1764 		 * Ignore requests to keep the effective and permitted
1765 		 * capability sets when the user id becomes non-zero.
1766 		 */
1767 		break;
1768 	case LINUX_PR_SET_NAME:
1769 		/*
1770 		 * To be on the safe side we need to make sure to not
1771 		 * overflow the size a Linux program expects. We already
1772 		 * do this here in the copyin, so that we don't need to
1773 		 * check on copyout.
1774 		 */
1775 		max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1776 		error = copyinstr((void *)(register_t)args->arg2, comm,
1777 		    max_size, NULL);
1778 
1779 		/* Linux silently truncates the name if it is too long. */
1780 		if (error == ENAMETOOLONG) {
1781 			/*
1782 			 * XXX: copyinstr() isn't documented to populate the
1783 			 * array completely, so do a copyin() to be on the
1784 			 * safe side. This should be changed in case
1785 			 * copyinstr() is changed to guarantee this.
1786 			 */
1787 			error = copyin((void *)(register_t)args->arg2, comm,
1788 			    max_size - 1);
1789 			comm[max_size - 1] = '\0';
1790 		}
1791 		if (error)
1792 			return (error);
1793 
1794 		PROC_LOCK(p);
1795 		strlcpy(p->p_comm, comm, sizeof(p->p_comm));
1796 		PROC_UNLOCK(p);
1797 		break;
1798 	case LINUX_PR_GET_NAME:
1799 		PROC_LOCK(p);
1800 		strlcpy(comm, p->p_comm, sizeof(comm));
1801 		PROC_UNLOCK(p);
1802 		error = copyout(comm, (void *)(register_t)args->arg2,
1803 		    strlen(comm) + 1);
1804 		break;
1805 	case LINUX_PR_GET_SECCOMP:
1806 	case LINUX_PR_SET_SECCOMP:
1807 		/*
1808 		 * Same as returned by Linux without CONFIG_SECCOMP enabled.
1809 		 */
1810 		error = EINVAL;
1811 		break;
1812 	case LINUX_PR_CAPBSET_READ:
1813 #if 0
1814 		/*
1815 		 * This makes too much noise with Ubuntu Focal.
1816 		 */
1817 		linux_msg(td, "unsupported prctl PR_CAPBSET_READ %d",
1818 		    (int)args->arg2);
1819 #endif
1820 		error = EINVAL;
1821 		break;
1822 	case LINUX_PR_SET_CHILD_SUBREAPER:
1823 		if (args->arg2 == 0) {
1824 			return (kern_procctl(td, P_PID, 0, PROC_REAP_RELEASE,
1825 			    NULL));
1826 		}
1827 
1828 		return (kern_procctl(td, P_PID, 0, PROC_REAP_ACQUIRE,
1829 		    NULL));
1830 	case LINUX_PR_SET_NO_NEW_PRIVS:
1831 		arg = args->arg2 == 1 ?
1832 		    PROC_NO_NEW_PRIVS_ENABLE : PROC_NO_NEW_PRIVS_DISABLE;
1833 		error = kern_procctl(td, P_PID, p->p_pid,
1834 		    PROC_NO_NEW_PRIVS_CTL, &arg);
1835 		break;
1836 	case LINUX_PR_SET_PTRACER:
1837 		linux_msg(td, "unsupported prctl PR_SET_PTRACER");
1838 		error = EINVAL;
1839 		break;
1840 	default:
1841 		linux_msg(td, "unsupported prctl option %d", args->option);
1842 		error = EINVAL;
1843 		break;
1844 	}
1845 
1846 	return (error);
1847 }
1848 
1849 int
1850 linux_sched_setparam(struct thread *td,
1851     struct linux_sched_setparam_args *uap)
1852 {
1853 	struct sched_param sched_param;
1854 	struct thread *tdt;
1855 	int error, policy;
1856 
1857 	error = copyin(uap->param, &sched_param, sizeof(sched_param));
1858 	if (error)
1859 		return (error);
1860 
1861 	tdt = linux_tdfind(td, uap->pid, -1);
1862 	if (tdt == NULL)
1863 		return (ESRCH);
1864 
1865 	if (linux_map_sched_prio) {
1866 		error = kern_sched_getscheduler(td, tdt, &policy);
1867 		if (error)
1868 			goto out;
1869 
1870 		switch (policy) {
1871 		case SCHED_OTHER:
1872 			if (sched_param.sched_priority != 0) {
1873 				error = EINVAL;
1874 				goto out;
1875 			}
1876 			sched_param.sched_priority =
1877 			    PRI_MAX_TIMESHARE - PRI_MIN_TIMESHARE;
1878 			break;
1879 		case SCHED_FIFO:
1880 		case SCHED_RR:
1881 			if (sched_param.sched_priority < 1 ||
1882 			    sched_param.sched_priority >= LINUX_MAX_RT_PRIO) {
1883 				error = EINVAL;
1884 				goto out;
1885 			}
1886 			/*
1887 			 * Map [1, LINUX_MAX_RT_PRIO - 1] to
1888 			 * [0, RTP_PRIO_MAX - RTP_PRIO_MIN] (rounding down).
1889 			 */
1890 			sched_param.sched_priority =
1891 			    (sched_param.sched_priority - 1) *
1892 			    (RTP_PRIO_MAX - RTP_PRIO_MIN + 1) /
1893 			    (LINUX_MAX_RT_PRIO - 1);
1894 			break;
1895 		}
1896 	}
1897 
1898 	error = kern_sched_setparam(td, tdt, &sched_param);
1899 out:	PROC_UNLOCK(tdt->td_proc);
1900 	return (error);
1901 }
1902 
1903 int
1904 linux_sched_getparam(struct thread *td,
1905     struct linux_sched_getparam_args *uap)
1906 {
1907 	struct sched_param sched_param;
1908 	struct thread *tdt;
1909 	int error, policy;
1910 
1911 	tdt = linux_tdfind(td, uap->pid, -1);
1912 	if (tdt == NULL)
1913 		return (ESRCH);
1914 
1915 	error = kern_sched_getparam(td, tdt, &sched_param);
1916 	if (error) {
1917 		PROC_UNLOCK(tdt->td_proc);
1918 		return (error);
1919 	}
1920 
1921 	if (linux_map_sched_prio) {
1922 		error = kern_sched_getscheduler(td, tdt, &policy);
1923 		PROC_UNLOCK(tdt->td_proc);
1924 		if (error)
1925 			return (error);
1926 
1927 		switch (policy) {
1928 		case SCHED_OTHER:
1929 			sched_param.sched_priority = 0;
1930 			break;
1931 		case SCHED_FIFO:
1932 		case SCHED_RR:
1933 			/*
1934 			 * Map [0, RTP_PRIO_MAX - RTP_PRIO_MIN] to
1935 			 * [1, LINUX_MAX_RT_PRIO - 1] (rounding up).
1936 			 */
1937 			sched_param.sched_priority =
1938 			    (sched_param.sched_priority *
1939 			    (LINUX_MAX_RT_PRIO - 1) +
1940 			    (RTP_PRIO_MAX - RTP_PRIO_MIN - 1)) /
1941 			    (RTP_PRIO_MAX - RTP_PRIO_MIN) + 1;
1942 			break;
1943 		}
1944 	} else
1945 		PROC_UNLOCK(tdt->td_proc);
1946 
1947 	error = copyout(&sched_param, uap->param, sizeof(sched_param));
1948 	return (error);
1949 }
1950 
1951 /*
1952  * Get affinity of a process.
1953  */
1954 int
1955 linux_sched_getaffinity(struct thread *td,
1956     struct linux_sched_getaffinity_args *args)
1957 {
1958 	struct thread *tdt;
1959 	cpuset_t *mask;
1960 	size_t size;
1961 	int error;
1962 	id_t tid;
1963 
1964 	tdt = linux_tdfind(td, args->pid, -1);
1965 	if (tdt == NULL)
1966 		return (ESRCH);
1967 	tid = tdt->td_tid;
1968 	PROC_UNLOCK(tdt->td_proc);
1969 
1970 	mask = malloc(sizeof(cpuset_t), M_LINUX, M_WAITOK | M_ZERO);
1971 	size = min(args->len, sizeof(cpuset_t));
1972 	error = kern_cpuset_getaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
1973 	    tid, size, mask);
1974 	if (error == ERANGE)
1975 		error = EINVAL;
1976  	if (error == 0)
1977 		error = copyout(mask, args->user_mask_ptr, size);
1978 	if (error == 0)
1979 		td->td_retval[0] = size;
1980 	free(mask, M_LINUX);
1981 	return (error);
1982 }
1983 
1984 /*
1985  *  Set affinity of a process.
1986  */
1987 int
1988 linux_sched_setaffinity(struct thread *td,
1989     struct linux_sched_setaffinity_args *args)
1990 {
1991 	struct thread *tdt;
1992 	cpuset_t *mask;
1993 	int cpu, error;
1994 	size_t len;
1995 	id_t tid;
1996 
1997 	tdt = linux_tdfind(td, args->pid, -1);
1998 	if (tdt == NULL)
1999 		return (ESRCH);
2000 	tid = tdt->td_tid;
2001 	PROC_UNLOCK(tdt->td_proc);
2002 
2003 	len = min(args->len, sizeof(cpuset_t));
2004 	mask = malloc(sizeof(cpuset_t), M_TEMP, M_WAITOK | M_ZERO);
2005 	error = copyin(args->user_mask_ptr, mask, len);
2006 	if (error != 0)
2007 		goto out;
2008 	/* Linux ignore high bits */
2009 	CPU_FOREACH_ISSET(cpu, mask)
2010 		if (cpu > mp_maxid)
2011 			CPU_CLR(cpu, mask);
2012 
2013 	error = kern_cpuset_setaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2014 	    tid, mask);
2015 	if (error == EDEADLK)
2016 		error = EINVAL;
2017 out:
2018 	free(mask, M_TEMP);
2019 	return (error);
2020 }
2021 
2022 struct linux_rlimit64 {
2023 	uint64_t	rlim_cur;
2024 	uint64_t	rlim_max;
2025 };
2026 
2027 int
2028 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
2029 {
2030 	struct rlimit rlim, nrlim;
2031 	struct linux_rlimit64 lrlim;
2032 	struct proc *p;
2033 	u_int which;
2034 	int flags;
2035 	int error;
2036 
2037 	if (args->new == NULL && args->old != NULL) {
2038 		if (linux_get_dummy_limit(td, args->resource, &rlim)) {
2039 			lrlim.rlim_cur = rlim.rlim_cur;
2040 			lrlim.rlim_max = rlim.rlim_max;
2041 			return (copyout(&lrlim, args->old, sizeof(lrlim)));
2042 		}
2043 	}
2044 
2045 	if (args->resource >= LINUX_RLIM_NLIMITS)
2046 		return (EINVAL);
2047 
2048 	which = linux_to_bsd_resource[args->resource];
2049 	if (which == -1)
2050 		return (EINVAL);
2051 
2052 	if (args->new != NULL) {
2053 		/*
2054 		 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
2055 		 * rlim is unsigned 64-bit. FreeBSD treats negative limits
2056 		 * as INFINITY so we do not need a conversion even.
2057 		 */
2058 		error = copyin(args->new, &nrlim, sizeof(nrlim));
2059 		if (error != 0)
2060 			return (error);
2061 	}
2062 
2063 	flags = PGET_HOLD | PGET_NOTWEXIT;
2064 	if (args->new != NULL)
2065 		flags |= PGET_CANDEBUG;
2066 	else
2067 		flags |= PGET_CANSEE;
2068 	if (args->pid == 0) {
2069 		p = td->td_proc;
2070 		PHOLD(p);
2071 	} else {
2072 		error = pget(args->pid, flags, &p);
2073 		if (error != 0)
2074 			return (error);
2075 	}
2076 	if (args->old != NULL) {
2077 		PROC_LOCK(p);
2078 		lim_rlimit_proc(p, which, &rlim);
2079 		PROC_UNLOCK(p);
2080 		if (rlim.rlim_cur == RLIM_INFINITY)
2081 			lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2082 		else
2083 			lrlim.rlim_cur = rlim.rlim_cur;
2084 		if (rlim.rlim_max == RLIM_INFINITY)
2085 			lrlim.rlim_max = LINUX_RLIM_INFINITY;
2086 		else
2087 			lrlim.rlim_max = rlim.rlim_max;
2088 		error = copyout(&lrlim, args->old, sizeof(lrlim));
2089 		if (error != 0)
2090 			goto out;
2091 	}
2092 
2093 	if (args->new != NULL)
2094 		error = kern_proc_setrlimit(td, p, which, &nrlim);
2095 
2096  out:
2097 	PRELE(p);
2098 	return (error);
2099 }
2100 
2101 int
2102 linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2103 {
2104 	struct timespec ts, *tsp;
2105 	int error;
2106 
2107 	if (args->tsp != NULL) {
2108 		error = linux_get_timespec(&ts, args->tsp);
2109 		if (error != 0)
2110 			return (error);
2111 		tsp = &ts;
2112 	} else
2113 		tsp = NULL;
2114 
2115 	error = linux_common_pselect6(td, args->nfds, args->readfds,
2116 	    args->writefds, args->exceptfds, tsp, args->sig);
2117 
2118 	if (args->tsp != NULL)
2119 		linux_put_timespec(&ts, args->tsp);
2120 	return (error);
2121 }
2122 
2123 static int
2124 linux_common_pselect6(struct thread *td, l_int nfds, l_fd_set *readfds,
2125     l_fd_set *writefds, l_fd_set *exceptfds, struct timespec *tsp,
2126     l_uintptr_t *sig)
2127 {
2128 	struct timeval utv, tv0, tv1, *tvp;
2129 	struct l_pselect6arg lpse6;
2130 	sigset_t *ssp;
2131 	sigset_t ss;
2132 	int error;
2133 
2134 	ssp = NULL;
2135 	if (sig != NULL) {
2136 		error = copyin(sig, &lpse6, sizeof(lpse6));
2137 		if (error != 0)
2138 			return (error);
2139 		error = linux_copyin_sigset(td, PTRIN(lpse6.ss),
2140 		    lpse6.ss_len, &ss, &ssp);
2141 		if (error != 0)
2142 		    return (error);
2143 	} else
2144 		ssp = NULL;
2145 
2146 	/*
2147 	 * Currently glibc changes nanosecond number to microsecond.
2148 	 * This mean losing precision but for now it is hardly seen.
2149 	 */
2150 	if (tsp != NULL) {
2151 		TIMESPEC_TO_TIMEVAL(&utv, tsp);
2152 		if (itimerfix(&utv))
2153 			return (EINVAL);
2154 
2155 		microtime(&tv0);
2156 		tvp = &utv;
2157 	} else
2158 		tvp = NULL;
2159 
2160 	error = kern_pselect(td, nfds, readfds, writefds,
2161 	    exceptfds, tvp, ssp, LINUX_NFDBITS);
2162 
2163 	if (tsp != NULL) {
2164 		/*
2165 		 * Compute how much time was left of the timeout,
2166 		 * by subtracting the current time and the time
2167 		 * before we started the call, and subtracting
2168 		 * that result from the user-supplied value.
2169 		 */
2170 		microtime(&tv1);
2171 		timevalsub(&tv1, &tv0);
2172 		timevalsub(&utv, &tv1);
2173 		if (utv.tv_sec < 0)
2174 			timevalclear(&utv);
2175 		TIMEVAL_TO_TIMESPEC(&utv, tsp);
2176 	}
2177 	return (error);
2178 }
2179 
2180 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
2181 int
2182 linux_pselect6_time64(struct thread *td,
2183     struct linux_pselect6_time64_args *args)
2184 {
2185 	struct timespec ts, *tsp;
2186 	int error;
2187 
2188 	if (args->tsp != NULL) {
2189 		error = linux_get_timespec64(&ts, args->tsp);
2190 		if (error != 0)
2191 			return (error);
2192 		tsp = &ts;
2193 	} else
2194 		tsp = NULL;
2195 
2196 	error = linux_common_pselect6(td, args->nfds, args->readfds,
2197 	    args->writefds, args->exceptfds, tsp, args->sig);
2198 
2199 	if (args->tsp != NULL)
2200 		linux_put_timespec64(&ts, args->tsp);
2201 	return (error);
2202 }
2203 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
2204 
2205 int
2206 linux_ppoll(struct thread *td, struct linux_ppoll_args *args)
2207 {
2208 	struct timespec uts, *tsp;
2209 	int error;
2210 
2211 	if (args->tsp != NULL) {
2212 		error = linux_get_timespec(&uts, args->tsp);
2213 		if (error != 0)
2214 			return (error);
2215 		tsp = &uts;
2216 	} else
2217 		tsp = NULL;
2218 
2219 	error = linux_common_ppoll(td, args->fds, args->nfds, tsp,
2220 	    args->sset, args->ssize);
2221 	if (error == 0 && args->tsp != NULL)
2222 		error = linux_put_timespec(&uts, args->tsp);
2223 	return (error);
2224 }
2225 
2226 static int
2227 linux_common_ppoll(struct thread *td, struct pollfd *fds, uint32_t nfds,
2228     struct timespec *tsp, l_sigset_t *sset, l_size_t ssize)
2229 {
2230 	struct timespec ts0, ts1;
2231 	struct pollfd stackfds[32];
2232 	struct pollfd *kfds;
2233  	sigset_t *ssp;
2234  	sigset_t ss;
2235  	int error;
2236 
2237 	if (kern_poll_maxfds(nfds))
2238 		return (EINVAL);
2239 	if (sset != NULL) {
2240 		error = linux_copyin_sigset(td, sset, ssize, &ss, &ssp);
2241 		if (error != 0)
2242 		    return (error);
2243 	} else
2244 		ssp = NULL;
2245 	if (tsp != NULL)
2246 		nanotime(&ts0);
2247 
2248 	if (nfds > nitems(stackfds))
2249 		kfds = mallocarray(nfds, sizeof(*kfds), M_TEMP, M_WAITOK);
2250 	else
2251 		kfds = stackfds;
2252 	error = linux_pollin(td, kfds, fds, nfds);
2253 	if (error != 0)
2254 		goto out;
2255 
2256 	error = kern_poll_kfds(td, kfds, nfds, tsp, ssp);
2257 	if (error == 0)
2258 		error = linux_pollout(td, kfds, fds, nfds);
2259 
2260 	if (error == 0 && tsp != NULL) {
2261 		if (td->td_retval[0]) {
2262 			nanotime(&ts1);
2263 			timespecsub(&ts1, &ts0, &ts1);
2264 			timespecsub(tsp, &ts1, tsp);
2265 			if (tsp->tv_sec < 0)
2266 				timespecclear(tsp);
2267 		} else
2268 			timespecclear(tsp);
2269 	}
2270 
2271 out:
2272 	if (nfds > nitems(stackfds))
2273 		free(kfds, M_TEMP);
2274 	return (error);
2275 }
2276 
2277 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
2278 int
2279 linux_ppoll_time64(struct thread *td, struct linux_ppoll_time64_args *args)
2280 {
2281 	struct timespec uts, *tsp;
2282 	int error;
2283 
2284 	if (args->tsp != NULL) {
2285 		error = linux_get_timespec64(&uts, args->tsp);
2286 		if (error != 0)
2287 			return (error);
2288 		tsp = &uts;
2289 	} else
2290  		tsp = NULL;
2291 	error = linux_common_ppoll(td, args->fds, args->nfds, tsp,
2292 	    args->sset, args->ssize);
2293 	if (error == 0 && args->tsp != NULL)
2294 		error = linux_put_timespec64(&uts, args->tsp);
2295 	return (error);
2296 }
2297 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
2298 
2299 static int
2300 linux_pollin(struct thread *td, struct pollfd *fds, struct pollfd *ufds, u_int nfd)
2301 {
2302 	int error;
2303 	u_int i;
2304 
2305 	error = copyin(ufds, fds, nfd * sizeof(*fds));
2306 	if (error != 0)
2307 		return (error);
2308 
2309 	for (i = 0; i < nfd; i++) {
2310 		if (fds->events != 0)
2311 			linux_to_bsd_poll_events(td, fds->fd,
2312 			    fds->events, &fds->events);
2313 		fds++;
2314 	}
2315 	return (0);
2316 }
2317 
2318 static int
2319 linux_pollout(struct thread *td, struct pollfd *fds, struct pollfd *ufds, u_int nfd)
2320 {
2321 	int error = 0;
2322 	u_int i, n = 0;
2323 
2324 	for (i = 0; i < nfd; i++) {
2325 		if (fds->revents != 0) {
2326 			bsd_to_linux_poll_events(fds->revents,
2327 			    &fds->revents);
2328 			n++;
2329 		}
2330 		error = copyout(&fds->revents, &ufds->revents,
2331 		    sizeof(ufds->revents));
2332 		if (error)
2333 			return (error);
2334 		fds++;
2335 		ufds++;
2336 	}
2337 	td->td_retval[0] = n;
2338 	return (0);
2339 }
2340 
2341 static int
2342 linux_sched_rr_get_interval_common(struct thread *td, pid_t pid,
2343     struct timespec *ts)
2344 {
2345 	struct thread *tdt;
2346 	int error;
2347 
2348 	/*
2349 	 * According to man in case the invalid pid specified
2350 	 * EINVAL should be returned.
2351 	 */
2352 	if (pid < 0)
2353 		return (EINVAL);
2354 
2355 	tdt = linux_tdfind(td, pid, -1);
2356 	if (tdt == NULL)
2357 		return (ESRCH);
2358 
2359 	error = kern_sched_rr_get_interval_td(td, tdt, ts);
2360 	PROC_UNLOCK(tdt->td_proc);
2361 	return (error);
2362 }
2363 
2364 int
2365 linux_sched_rr_get_interval(struct thread *td,
2366     struct linux_sched_rr_get_interval_args *uap)
2367 {
2368 	struct timespec ts;
2369 	int error;
2370 
2371 	error = linux_sched_rr_get_interval_common(td, uap->pid, &ts);
2372 	if (error != 0)
2373 		return (error);
2374 	return (linux_put_timespec(&ts, uap->interval));
2375 }
2376 
2377 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
2378 int
2379 linux_sched_rr_get_interval_time64(struct thread *td,
2380     struct linux_sched_rr_get_interval_time64_args *uap)
2381 {
2382 	struct timespec ts;
2383 	int error;
2384 
2385 	error = linux_sched_rr_get_interval_common(td, uap->pid, &ts);
2386 	if (error != 0)
2387 		return (error);
2388 	return (linux_put_timespec64(&ts, uap->interval));
2389 }
2390 #endif
2391 
2392 /*
2393  * In case when the Linux thread is the initial thread in
2394  * the thread group thread id is equal to the process id.
2395  * Glibc depends on this magic (assert in pthread_getattr_np.c).
2396  */
2397 struct thread *
2398 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid)
2399 {
2400 	struct linux_emuldata *em;
2401 	struct thread *tdt;
2402 	struct proc *p;
2403 
2404 	tdt = NULL;
2405 	if (tid == 0 || tid == td->td_tid) {
2406 		if (pid != -1 && td->td_proc->p_pid != pid)
2407 			return (NULL);
2408 		PROC_LOCK(td->td_proc);
2409 		return (td);
2410 	} else if (tid > PID_MAX)
2411 		return (tdfind(tid, pid));
2412 
2413 	/*
2414 	 * Initial thread where the tid equal to the pid.
2415 	 */
2416 	p = pfind(tid);
2417 	if (p != NULL) {
2418 		if (SV_PROC_ABI(p) != SV_ABI_LINUX ||
2419 		    (pid != -1 && tid != pid)) {
2420 			/*
2421 			 * p is not a Linuxulator process.
2422 			 */
2423 			PROC_UNLOCK(p);
2424 			return (NULL);
2425 		}
2426 		FOREACH_THREAD_IN_PROC(p, tdt) {
2427 			em = em_find(tdt);
2428 			if (tid == em->em_tid)
2429 				return (tdt);
2430 		}
2431 		PROC_UNLOCK(p);
2432 	}
2433 	return (NULL);
2434 }
2435 
2436 void
2437 linux_to_bsd_waitopts(int options, int *bsdopts)
2438 {
2439 
2440 	if (options & LINUX_WNOHANG)
2441 		*bsdopts |= WNOHANG;
2442 	if (options & LINUX_WUNTRACED)
2443 		*bsdopts |= WUNTRACED;
2444 	if (options & LINUX_WEXITED)
2445 		*bsdopts |= WEXITED;
2446 	if (options & LINUX_WCONTINUED)
2447 		*bsdopts |= WCONTINUED;
2448 	if (options & LINUX_WNOWAIT)
2449 		*bsdopts |= WNOWAIT;
2450 
2451 	if (options & __WCLONE)
2452 		*bsdopts |= WLINUXCLONE;
2453 }
2454 
2455 int
2456 linux_getrandom(struct thread *td, struct linux_getrandom_args *args)
2457 {
2458 	struct uio uio;
2459 	struct iovec iov;
2460 	int error;
2461 
2462 	if (args->flags & ~(LINUX_GRND_NONBLOCK|LINUX_GRND_RANDOM))
2463 		return (EINVAL);
2464 	if (args->count > INT_MAX)
2465 		args->count = INT_MAX;
2466 
2467 	iov.iov_base = args->buf;
2468 	iov.iov_len = args->count;
2469 
2470 	uio.uio_iov = &iov;
2471 	uio.uio_iovcnt = 1;
2472 	uio.uio_resid = iov.iov_len;
2473 	uio.uio_segflg = UIO_USERSPACE;
2474 	uio.uio_rw = UIO_READ;
2475 	uio.uio_td = td;
2476 
2477 	error = read_random_uio(&uio, args->flags & LINUX_GRND_NONBLOCK);
2478 	if (error == 0)
2479 		td->td_retval[0] = args->count - uio.uio_resid;
2480 	return (error);
2481 }
2482 
2483 int
2484 linux_mincore(struct thread *td, struct linux_mincore_args *args)
2485 {
2486 
2487 	/* Needs to be page-aligned */
2488 	if (args->start & PAGE_MASK)
2489 		return (EINVAL);
2490 	return (kern_mincore(td, args->start, args->len, args->vec));
2491 }
2492 
2493 #define	SYSLOG_TAG	"<6>"
2494 
2495 int
2496 linux_syslog(struct thread *td, struct linux_syslog_args *args)
2497 {
2498 	char buf[128], *src, *dst;
2499 	u_int seq;
2500 	int buflen, error;
2501 
2502 	if (args->type != LINUX_SYSLOG_ACTION_READ_ALL) {
2503 		linux_msg(td, "syslog unsupported type 0x%x", args->type);
2504 		return (EINVAL);
2505 	}
2506 
2507 	if (args->len < 6) {
2508 		td->td_retval[0] = 0;
2509 		return (0);
2510 	}
2511 
2512 	error = priv_check(td, PRIV_MSGBUF);
2513 	if (error)
2514 		return (error);
2515 
2516 	mtx_lock(&msgbuf_lock);
2517 	msgbuf_peekbytes(msgbufp, NULL, 0, &seq);
2518 	mtx_unlock(&msgbuf_lock);
2519 
2520 	dst = args->buf;
2521 	error = copyout(&SYSLOG_TAG, dst, sizeof(SYSLOG_TAG));
2522 	/* The -1 is to skip the trailing '\0'. */
2523 	dst += sizeof(SYSLOG_TAG) - 1;
2524 
2525 	while (error == 0) {
2526 		mtx_lock(&msgbuf_lock);
2527 		buflen = msgbuf_peekbytes(msgbufp, buf, sizeof(buf), &seq);
2528 		mtx_unlock(&msgbuf_lock);
2529 
2530 		if (buflen == 0)
2531 			break;
2532 
2533 		for (src = buf; src < buf + buflen && error == 0; src++) {
2534 			if (*src == '\0')
2535 				continue;
2536 
2537 			if (dst >= args->buf + args->len)
2538 				goto out;
2539 
2540 			error = copyout(src, dst, 1);
2541 			dst++;
2542 
2543 			if (*src == '\n' && *(src + 1) != '<' &&
2544 			    dst + sizeof(SYSLOG_TAG) < args->buf + args->len) {
2545 				error = copyout(&SYSLOG_TAG,
2546 				    dst, sizeof(SYSLOG_TAG));
2547 				dst += sizeof(SYSLOG_TAG) - 1;
2548 			}
2549 		}
2550 	}
2551 out:
2552 	td->td_retval[0] = dst - args->buf;
2553 	return (error);
2554 }
2555 
2556 int
2557 linux_getcpu(struct thread *td, struct linux_getcpu_args *args)
2558 {
2559 	int cpu, error, node;
2560 
2561 	cpu = td->td_oncpu; /* Make sure it doesn't change during copyout(9) */
2562 	error = 0;
2563 	node = cpuid_to_pcpu[cpu]->pc_domain;
2564 
2565 	if (args->cpu != NULL)
2566 		error = copyout(&cpu, args->cpu, sizeof(l_int));
2567 	if (args->node != NULL)
2568 		error = copyout(&node, args->node, sizeof(l_int));
2569 	return (error);
2570 }
2571 
2572 #if defined(__i386__) || defined(__amd64__)
2573 int
2574 linux_poll(struct thread *td, struct linux_poll_args *args)
2575 {
2576 	struct timespec ts, *tsp;
2577 
2578 	if (args->timeout != INFTIM) {
2579 		if (args->timeout < 0)
2580 			return (EINVAL);
2581 		ts.tv_sec = args->timeout / 1000;
2582 		ts.tv_nsec = (args->timeout % 1000) * 1000000;
2583 		tsp = &ts;
2584 	} else
2585 		tsp = NULL;
2586 
2587 	return (linux_common_ppoll(td, args->fds, args->nfds,
2588 	    tsp, NULL, 0));
2589 }
2590 #endif /* __i386__ || __amd64__ */
2591 
2592 int
2593 linux_seccomp(struct thread *td, struct linux_seccomp_args *args)
2594 {
2595 
2596 	switch (args->op) {
2597 	case LINUX_SECCOMP_GET_ACTION_AVAIL:
2598 		return (EOPNOTSUPP);
2599 	default:
2600 		/*
2601 		 * Ignore unknown operations, just like Linux kernel built
2602 		 * without CONFIG_SECCOMP.
2603 		 */
2604 		return (EINVAL);
2605 	}
2606 }
2607 
2608 /*
2609  * Custom version of exec_copyin_args(), to copy out argument and environment
2610  * strings from the old process address space into the temporary string buffer.
2611  * Based on freebsd32_exec_copyin_args.
2612  */
2613 static int
2614 linux_exec_copyin_args(struct image_args *args, const char *fname,
2615     enum uio_seg segflg, l_uintptr_t *argv, l_uintptr_t *envv)
2616 {
2617 	char *argp, *envp;
2618 	l_uintptr_t *ptr, arg;
2619 	int error;
2620 
2621 	bzero(args, sizeof(*args));
2622 	if (argv == NULL)
2623 		return (EFAULT);
2624 
2625 	/*
2626 	 * Allocate demand-paged memory for the file name, argument, and
2627 	 * environment strings.
2628 	 */
2629 	error = exec_alloc_args(args);
2630 	if (error != 0)
2631 		return (error);
2632 
2633 	/*
2634 	 * Copy the file name.
2635 	 */
2636 	error = exec_args_add_fname(args, fname, segflg);
2637 	if (error != 0)
2638 		goto err_exit;
2639 
2640 	/*
2641 	 * extract arguments first
2642 	 */
2643 	ptr = argv;
2644 	for (;;) {
2645 		error = copyin(ptr++, &arg, sizeof(arg));
2646 		if (error)
2647 			goto err_exit;
2648 		if (arg == 0)
2649 			break;
2650 		argp = PTRIN(arg);
2651 		error = exec_args_add_arg(args, argp, UIO_USERSPACE);
2652 		if (error != 0)
2653 			goto err_exit;
2654 	}
2655 
2656 	/*
2657 	 * This comment is from Linux do_execveat_common:
2658 	 * When argv is empty, add an empty string ("") as argv[0] to
2659 	 * ensure confused userspace programs that start processing
2660 	 * from argv[1] won't end up walking envp.
2661 	 */
2662 	if (args->argc == 0 &&
2663 	    (error = exec_args_add_arg(args, "", UIO_SYSSPACE) != 0))
2664 		goto err_exit;
2665 
2666 	/*
2667 	 * extract environment strings
2668 	 */
2669 	if (envv) {
2670 		ptr = envv;
2671 		for (;;) {
2672 			error = copyin(ptr++, &arg, sizeof(arg));
2673 			if (error)
2674 				goto err_exit;
2675 			if (arg == 0)
2676 				break;
2677 			envp = PTRIN(arg);
2678 			error = exec_args_add_env(args, envp, UIO_USERSPACE);
2679 			if (error != 0)
2680 				goto err_exit;
2681 		}
2682 	}
2683 
2684 	return (0);
2685 
2686 err_exit:
2687 	exec_free_args(args);
2688 	return (error);
2689 }
2690 
2691 int
2692 linux_execve(struct thread *td, struct linux_execve_args *args)
2693 {
2694 	struct image_args eargs;
2695 	int error;
2696 
2697 	LINUX_CTR(execve);
2698 
2699 	error = linux_exec_copyin_args(&eargs, args->path, UIO_USERSPACE,
2700 	    args->argp, args->envp);
2701 	if (error == 0)
2702 		error = linux_common_execve(td, &eargs);
2703 	AUDIT_SYSCALL_EXIT(error == EJUSTRETURN ? 0 : error, td);
2704 	return (error);
2705 }
2706 
2707 static void
2708 linux_up_rtprio_if(struct thread *td1, struct rtprio *rtp)
2709 {
2710 	struct rtprio rtp2;
2711 
2712 	pri_to_rtp(td1, &rtp2);
2713 	if (rtp2.type <  rtp->type ||
2714 	    (rtp2.type == rtp->type &&
2715 	    rtp2.prio < rtp->prio)) {
2716 		rtp->type = rtp2.type;
2717 		rtp->prio = rtp2.prio;
2718 	}
2719 }
2720 
2721 #define	LINUX_PRIO_DIVIDER	RTP_PRIO_MAX / LINUX_IOPRIO_MAX
2722 
2723 static int
2724 linux_rtprio2ioprio(struct rtprio *rtp)
2725 {
2726 	int ioprio, prio;
2727 
2728 	switch (rtp->type) {
2729 	case RTP_PRIO_IDLE:
2730 		prio = RTP_PRIO_MIN;
2731 		ioprio = LINUX_IOPRIO_PRIO(LINUX_IOPRIO_CLASS_IDLE, prio);
2732 		break;
2733 	case RTP_PRIO_NORMAL:
2734 		prio = rtp->prio / LINUX_PRIO_DIVIDER;
2735 		ioprio = LINUX_IOPRIO_PRIO(LINUX_IOPRIO_CLASS_BE, prio);
2736 		break;
2737 	case RTP_PRIO_REALTIME:
2738 		prio = rtp->prio / LINUX_PRIO_DIVIDER;
2739 		ioprio = LINUX_IOPRIO_PRIO(LINUX_IOPRIO_CLASS_RT, prio);
2740 		break;
2741 	default:
2742 		prio = RTP_PRIO_MIN;
2743 		ioprio = LINUX_IOPRIO_PRIO(LINUX_IOPRIO_CLASS_NONE, prio);
2744 		break;
2745 	}
2746 	return (ioprio);
2747 }
2748 
2749 static int
2750 linux_ioprio2rtprio(int ioprio, struct rtprio *rtp)
2751 {
2752 
2753 	switch (LINUX_IOPRIO_PRIO_CLASS(ioprio)) {
2754 	case LINUX_IOPRIO_CLASS_IDLE:
2755 		rtp->prio = RTP_PRIO_MIN;
2756 		rtp->type = RTP_PRIO_IDLE;
2757 		break;
2758 	case LINUX_IOPRIO_CLASS_BE:
2759 		rtp->prio = LINUX_IOPRIO_PRIO_DATA(ioprio) * LINUX_PRIO_DIVIDER;
2760 		rtp->type = RTP_PRIO_NORMAL;
2761 		break;
2762 	case LINUX_IOPRIO_CLASS_RT:
2763 		rtp->prio = LINUX_IOPRIO_PRIO_DATA(ioprio) * LINUX_PRIO_DIVIDER;
2764 		rtp->type = RTP_PRIO_REALTIME;
2765 		break;
2766 	default:
2767 		return (EINVAL);
2768 	}
2769 	return (0);
2770 }
2771 #undef LINUX_PRIO_DIVIDER
2772 
2773 int
2774 linux_ioprio_get(struct thread *td, struct linux_ioprio_get_args *args)
2775 {
2776 	struct thread *td1;
2777 	struct rtprio rtp;
2778 	struct pgrp *pg;
2779 	struct proc *p;
2780 	int error, found;
2781 
2782 	p = NULL;
2783 	td1 = NULL;
2784 	error = 0;
2785 	found = 0;
2786 	rtp.type = RTP_PRIO_IDLE;
2787 	rtp.prio = RTP_PRIO_MAX;
2788 	switch (args->which) {
2789 	case LINUX_IOPRIO_WHO_PROCESS:
2790 		if (args->who == 0) {
2791 			td1 = td;
2792 			p = td1->td_proc;
2793 			PROC_LOCK(p);
2794 		} else if (args->who > PID_MAX) {
2795 			td1 = linux_tdfind(td, args->who, -1);
2796 			if (td1 != NULL)
2797 				p = td1->td_proc;
2798 		} else
2799 			p = pfind(args->who);
2800 		if (p == NULL)
2801 			return (ESRCH);
2802 		if ((error = p_cansee(td, p))) {
2803 			PROC_UNLOCK(p);
2804 			break;
2805 		}
2806 		if (td1 != NULL) {
2807 			pri_to_rtp(td1, &rtp);
2808 		} else {
2809 			FOREACH_THREAD_IN_PROC(p, td1) {
2810 				linux_up_rtprio_if(td1, &rtp);
2811 			}
2812 		}
2813 		found++;
2814 		PROC_UNLOCK(p);
2815 		break;
2816 	case LINUX_IOPRIO_WHO_PGRP:
2817 		sx_slock(&proctree_lock);
2818 		if (args->who == 0) {
2819 			pg = td->td_proc->p_pgrp;
2820 			PGRP_LOCK(pg);
2821 		} else {
2822 			pg = pgfind(args->who);
2823 			if (pg == NULL) {
2824 				sx_sunlock(&proctree_lock);
2825 				error = ESRCH;
2826 				break;
2827 			}
2828 		}
2829 		sx_sunlock(&proctree_lock);
2830 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
2831 			PROC_LOCK(p);
2832 			if (p->p_state == PRS_NORMAL &&
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 		PGRP_UNLOCK(pg);
2842 		break;
2843 	case LINUX_IOPRIO_WHO_USER:
2844 		if (args->who == 0)
2845 			args->who = td->td_ucred->cr_uid;
2846 		sx_slock(&allproc_lock);
2847 		FOREACH_PROC_IN_SYSTEM(p) {
2848 			PROC_LOCK(p);
2849 			if (p->p_state == PRS_NORMAL &&
2850 			    p->p_ucred->cr_uid == args->who &&
2851 			    p_cansee(td, p) == 0) {
2852 				FOREACH_THREAD_IN_PROC(p, td1) {
2853 					linux_up_rtprio_if(td1, &rtp);
2854 					found++;
2855 				}
2856 			}
2857 			PROC_UNLOCK(p);
2858 		}
2859 		sx_sunlock(&allproc_lock);
2860 		break;
2861 	default:
2862 		error = EINVAL;
2863 		break;
2864 	}
2865 	if (error == 0) {
2866 		if (found != 0)
2867 			td->td_retval[0] = linux_rtprio2ioprio(&rtp);
2868 		else
2869 			error = ESRCH;
2870 	}
2871 	return (error);
2872 }
2873 
2874 int
2875 linux_ioprio_set(struct thread *td, struct linux_ioprio_set_args *args)
2876 {
2877 	struct thread *td1;
2878 	struct rtprio rtp;
2879 	struct pgrp *pg;
2880 	struct proc *p;
2881 	int error;
2882 
2883 	if ((error = linux_ioprio2rtprio(args->ioprio, &rtp)) != 0)
2884 		return (error);
2885 	/* Attempts to set high priorities (REALTIME) require su privileges. */
2886 	if (RTP_PRIO_BASE(rtp.type) == RTP_PRIO_REALTIME &&
2887 	    (error = priv_check(td, PRIV_SCHED_RTPRIO)) != 0)
2888 		return (error);
2889 
2890 	p = NULL;
2891 	td1 = NULL;
2892 	switch (args->which) {
2893 	case LINUX_IOPRIO_WHO_PROCESS:
2894 		if (args->who == 0) {
2895 			td1 = td;
2896 			p = td1->td_proc;
2897 			PROC_LOCK(p);
2898 		} else if (args->who > PID_MAX) {
2899 			td1 = linux_tdfind(td, args->who, -1);
2900 			if (td1 != NULL)
2901 				p = td1->td_proc;
2902 		} else
2903 			p = pfind(args->who);
2904 		if (p == NULL)
2905 			return (ESRCH);
2906 		if ((error = p_cansched(td, p))) {
2907 			PROC_UNLOCK(p);
2908 			break;
2909 		}
2910 		if (td1 != NULL) {
2911 			error = rtp_to_pri(&rtp, td1);
2912 		} else {
2913 			FOREACH_THREAD_IN_PROC(p, td1) {
2914 				if ((error = rtp_to_pri(&rtp, td1)) != 0)
2915 					break;
2916 			}
2917 		}
2918 		PROC_UNLOCK(p);
2919 		break;
2920 	case LINUX_IOPRIO_WHO_PGRP:
2921 		sx_slock(&proctree_lock);
2922 		if (args->who == 0) {
2923 			pg = td->td_proc->p_pgrp;
2924 			PGRP_LOCK(pg);
2925 		} else {
2926 			pg = pgfind(args->who);
2927 			if (pg == NULL) {
2928 				sx_sunlock(&proctree_lock);
2929 				error = ESRCH;
2930 				break;
2931 			}
2932 		}
2933 		sx_sunlock(&proctree_lock);
2934 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
2935 			PROC_LOCK(p);
2936 			if (p->p_state == PRS_NORMAL &&
2937 			    p_cansched(td, p) == 0) {
2938 				FOREACH_THREAD_IN_PROC(p, td1) {
2939 					if ((error = rtp_to_pri(&rtp, td1)) != 0)
2940 						break;
2941 				}
2942 			}
2943 			PROC_UNLOCK(p);
2944 			if (error != 0)
2945 				break;
2946 		}
2947 		PGRP_UNLOCK(pg);
2948 		break;
2949 	case LINUX_IOPRIO_WHO_USER:
2950 		if (args->who == 0)
2951 			args->who = td->td_ucred->cr_uid;
2952 		sx_slock(&allproc_lock);
2953 		FOREACH_PROC_IN_SYSTEM(p) {
2954 			PROC_LOCK(p);
2955 			if (p->p_state == PRS_NORMAL &&
2956 			    p->p_ucred->cr_uid == args->who &&
2957 			    p_cansched(td, p) == 0) {
2958 				FOREACH_THREAD_IN_PROC(p, td1) {
2959 					if ((error = rtp_to_pri(&rtp, td1)) != 0)
2960 						break;
2961 				}
2962 			}
2963 			PROC_UNLOCK(p);
2964 			if (error != 0)
2965 				break;
2966 		}
2967 		sx_sunlock(&allproc_lock);
2968 		break;
2969 	default:
2970 		error = EINVAL;
2971 		break;
2972 	}
2973 	return (error);
2974 }
2975 
2976 /* The only flag is O_NONBLOCK */
2977 #define B2L_MQ_FLAGS(bflags)	((bflags) != 0 ? LINUX_O_NONBLOCK : 0)
2978 #define L2B_MQ_FLAGS(lflags)	((lflags) != 0 ? O_NONBLOCK : 0)
2979 
2980 int
2981 linux_mq_open(struct thread *td, struct linux_mq_open_args *args)
2982 {
2983 	struct mq_attr attr;
2984 	int error, flags;
2985 
2986 	flags = linux_common_openflags(args->oflag);
2987 	if ((flags & O_ACCMODE) == O_ACCMODE || (flags & O_EXEC) != 0)
2988 		return (EINVAL);
2989 	flags = FFLAGS(flags);
2990 	if ((flags & O_CREAT) != 0 && args->attr != NULL) {
2991 		error = copyin(args->attr, &attr, sizeof(attr));
2992 		if (error != 0)
2993 			return (error);
2994 		attr.mq_flags = L2B_MQ_FLAGS(attr.mq_flags);
2995 	}
2996 
2997 	return (kern_kmq_open(td, args->name, flags, args->mode,
2998 	    args->attr != NULL ? &attr : NULL));
2999 }
3000 
3001 int
3002 linux_mq_unlink(struct thread *td, struct linux_mq_unlink_args *args)
3003 {
3004 	struct kmq_unlink_args bsd_args = {
3005 		.path = PTRIN(args->name)
3006 	};
3007 
3008 	return (sys_kmq_unlink(td, &bsd_args));
3009 }
3010 
3011 int
3012 linux_mq_timedsend(struct thread *td, struct linux_mq_timedsend_args *args)
3013 {
3014 	struct timespec ts, *abs_timeout;
3015 	int error;
3016 
3017 	if (args->abs_timeout == NULL)
3018 		abs_timeout = NULL;
3019 	else {
3020 		error = linux_get_timespec(&ts, args->abs_timeout);
3021 		if (error != 0)
3022 			return (error);
3023 		abs_timeout = &ts;
3024 	}
3025 
3026 	return (kern_kmq_timedsend(td, args->mqd, PTRIN(args->msg_ptr),
3027 		args->msg_len, args->msg_prio, abs_timeout));
3028 }
3029 
3030 int
3031 linux_mq_timedreceive(struct thread *td, struct linux_mq_timedreceive_args *args)
3032 {
3033 	struct timespec ts, *abs_timeout;
3034 	int error;
3035 
3036 	if (args->abs_timeout == NULL)
3037 		abs_timeout = NULL;
3038 	else {
3039 		error = linux_get_timespec(&ts, args->abs_timeout);
3040 		if (error != 0)
3041 			return (error);
3042 		abs_timeout = &ts;
3043 	}
3044 
3045 	return (kern_kmq_timedreceive(td, args->mqd, PTRIN(args->msg_ptr),
3046 		args->msg_len, args->msg_prio, abs_timeout));
3047 }
3048 
3049 int
3050 linux_mq_notify(struct thread *td, struct linux_mq_notify_args *args)
3051 {
3052 	struct sigevent ev, *evp;
3053 	struct l_sigevent l_ev;
3054 	int error;
3055 
3056 	if (args->sevp == NULL)
3057 		evp = NULL;
3058 	else {
3059 		error = copyin(args->sevp, &l_ev, sizeof(l_ev));
3060 		if (error != 0)
3061 			return (error);
3062 		error = linux_convert_l_sigevent(&l_ev, &ev);
3063 		if (error != 0)
3064 			return (error);
3065 		evp = &ev;
3066 	}
3067 
3068 	return (kern_kmq_notify(td, args->mqd, evp));
3069 }
3070 
3071 int
3072 linux_mq_getsetattr(struct thread *td, struct linux_mq_getsetattr_args *args)
3073 {
3074 	struct mq_attr attr, oattr;
3075 	int error;
3076 
3077 	if (args->attr != NULL) {
3078 		error = copyin(args->attr, &attr, sizeof(attr));
3079 		if (error != 0)
3080 			return (error);
3081 		attr.mq_flags = L2B_MQ_FLAGS(attr.mq_flags);
3082 	}
3083 
3084 	error = kern_kmq_setattr(td, args->mqd, args->attr != NULL ? &attr : NULL,
3085 	    &oattr);
3086 	if (error == 0 && args->oattr != NULL) {
3087 		oattr.mq_flags = B2L_MQ_FLAGS(oattr.mq_flags);
3088 		bzero(oattr.__reserved, sizeof(oattr.__reserved));
3089 		error = copyout(&oattr, args->oattr, sizeof(oattr));
3090 	}
3091 
3092 	return (error);
3093 }
3094 
3095 MODULE_DEPEND(linux, mqueuefs, 1, 1, 1);
3096