xref: /freebsd/sys/compat/linux/linux_misc.c (revision 25fb30bd9abc492359ad1f66901a06cb8cd08370)
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/cdefs.h>
33 __FBSDID("$FreeBSD$");
34 
35 #include "opt_compat.h"
36 
37 #include <sys/param.h>
38 #include <sys/blist.h>
39 #include <sys/fcntl.h>
40 #if defined(__i386__)
41 #include <sys/imgact_aout.h>
42 #endif
43 #include <sys/jail.h>
44 #include <sys/kernel.h>
45 #include <sys/limits.h>
46 #include <sys/lock.h>
47 #include <sys/malloc.h>
48 #include <sys/mman.h>
49 #include <sys/mount.h>
50 #include <sys/msgbuf.h>
51 #include <sys/mutex.h>
52 #include <sys/namei.h>
53 #include <sys/priv.h>
54 #include <sys/proc.h>
55 #include <sys/procctl.h>
56 #include <sys/reboot.h>
57 #include <sys/racct.h>
58 #include <sys/random.h>
59 #include <sys/resourcevar.h>
60 #include <sys/sched.h>
61 #include <sys/sdt.h>
62 #include <sys/signalvar.h>
63 #include <sys/stat.h>
64 #include <sys/syscallsubr.h>
65 #include <sys/sysctl.h>
66 #include <sys/sysproto.h>
67 #include <sys/systm.h>
68 #include <sys/time.h>
69 #include <sys/vmmeter.h>
70 #include <sys/vnode.h>
71 #include <sys/wait.h>
72 #include <sys/cpuset.h>
73 #include <sys/uio.h>
74 
75 #include <security/mac/mac_framework.h>
76 
77 #include <vm/vm.h>
78 #include <vm/pmap.h>
79 #include <vm/vm_kern.h>
80 #include <vm/vm_map.h>
81 #include <vm/vm_extern.h>
82 #include <vm/swap_pager.h>
83 
84 #ifdef COMPAT_LINUX32
85 #include <machine/../linux32/linux.h>
86 #include <machine/../linux32/linux32_proto.h>
87 #else
88 #include <machine/../linux/linux.h>
89 #include <machine/../linux/linux_proto.h>
90 #endif
91 
92 #include <compat/linux/linux_dtrace.h>
93 #include <compat/linux/linux_file.h>
94 #include <compat/linux/linux_mib.h>
95 #include <compat/linux/linux_signal.h>
96 #include <compat/linux/linux_timer.h>
97 #include <compat/linux/linux_util.h>
98 #include <compat/linux/linux_sysproto.h>
99 #include <compat/linux/linux_emul.h>
100 #include <compat/linux/linux_misc.h>
101 
102 /**
103  * Special DTrace provider for the linuxulator.
104  *
105  * In this file we define the provider for the entire linuxulator. All
106  * modules (= files of the linuxulator) use it.
107  *
108  * We define a different name depending on the emulated bitsize, see
109  * ../../<ARCH>/linux{,32}/linux.h, e.g.:
110  *      native bitsize          = linuxulator
111  *      amd64, 32bit emulation  = linuxulator32
112  */
113 LIN_SDT_PROVIDER_DEFINE(LINUX_DTRACE);
114 
115 int stclohz;				/* Statistics clock frequency */
116 
117 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
118 	RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
119 	RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
120 	RLIMIT_MEMLOCK, RLIMIT_AS
121 };
122 
123 struct l_sysinfo {
124 	l_long		uptime;		/* Seconds since boot */
125 	l_ulong		loads[3];	/* 1, 5, and 15 minute load averages */
126 #define LINUX_SYSINFO_LOADS_SCALE 65536
127 	l_ulong		totalram;	/* Total usable main memory size */
128 	l_ulong		freeram;	/* Available memory size */
129 	l_ulong		sharedram;	/* Amount of shared memory */
130 	l_ulong		bufferram;	/* Memory used by buffers */
131 	l_ulong		totalswap;	/* Total swap space size */
132 	l_ulong		freeswap;	/* swap space still available */
133 	l_ushort	procs;		/* Number of current processes */
134 	l_ushort	pads;
135 	l_ulong		totalhigh;
136 	l_ulong		freehigh;
137 	l_uint		mem_unit;
138 	char		_f[20-2*sizeof(l_long)-sizeof(l_int)];	/* padding */
139 };
140 
141 struct l_pselect6arg {
142 	l_uintptr_t	ss;
143 	l_size_t	ss_len;
144 };
145 
146 static int	linux_utimensat_nsec_valid(l_long);
147 
148 int
149 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
150 {
151 	struct l_sysinfo sysinfo;
152 	int i, j;
153 	struct timespec ts;
154 
155 	bzero(&sysinfo, sizeof(sysinfo));
156 	getnanouptime(&ts);
157 	if (ts.tv_nsec != 0)
158 		ts.tv_sec++;
159 	sysinfo.uptime = ts.tv_sec;
160 
161 	/* Use the information from the mib to get our load averages */
162 	for (i = 0; i < 3; i++)
163 		sysinfo.loads[i] = averunnable.ldavg[i] *
164 		    LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
165 
166 	sysinfo.totalram = physmem * PAGE_SIZE;
167 	sysinfo.freeram = (u_long)vm_free_count() * PAGE_SIZE;
168 
169 	/*
170 	 * sharedram counts pages allocated to named, swap-backed objects such
171 	 * as shared memory segments and tmpfs files.  There is no cheap way to
172 	 * compute this, so just leave the field unpopulated.  Linux itself only
173 	 * started setting this field in the 3.x timeframe.
174 	 */
175 	sysinfo.sharedram = 0;
176 	sysinfo.bufferram = 0;
177 
178 	swap_pager_status(&i, &j);
179 	sysinfo.totalswap = i * PAGE_SIZE;
180 	sysinfo.freeswap = (i - j) * PAGE_SIZE;
181 
182 	sysinfo.procs = nprocs;
183 
184 	/*
185 	 * Platforms supported by the emulation layer do not have a notion of
186 	 * high memory.
187 	 */
188 	sysinfo.totalhigh = 0;
189 	sysinfo.freehigh = 0;
190 
191 	sysinfo.mem_unit = 1;
192 
193 	return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
194 }
195 
196 #ifdef LINUX_LEGACY_SYSCALLS
197 int
198 linux_alarm(struct thread *td, struct linux_alarm_args *args)
199 {
200 	struct itimerval it, old_it;
201 	u_int secs;
202 	int error;
203 
204 	secs = args->secs;
205 	/*
206 	 * Linux alarm() is always successful. Limit secs to INT32_MAX / 2
207 	 * to match kern_setitimer()'s limit to avoid error from it.
208 	 *
209 	 * XXX. Linux limit secs to INT_MAX on 32 and does not limit on 64-bit
210 	 * platforms.
211 	 */
212 	if (secs > INT32_MAX / 2)
213 		secs = INT32_MAX / 2;
214 
215 	it.it_value.tv_sec = secs;
216 	it.it_value.tv_usec = 0;
217 	timevalclear(&it.it_interval);
218 	error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
219 	KASSERT(error == 0, ("kern_setitimer returns %d", error));
220 
221 	if ((old_it.it_value.tv_sec == 0 && old_it.it_value.tv_usec > 0) ||
222 	    old_it.it_value.tv_usec >= 500000)
223 		old_it.it_value.tv_sec++;
224 	td->td_retval[0] = old_it.it_value.tv_sec;
225 	return (0);
226 }
227 #endif
228 
229 int
230 linux_brk(struct thread *td, struct linux_brk_args *args)
231 {
232 	struct vmspace *vm = td->td_proc->p_vmspace;
233 	uintptr_t new, old;
234 
235 	old = (uintptr_t)vm->vm_daddr + ctob(vm->vm_dsize);
236 	new = (uintptr_t)args->dsend;
237 	if ((caddr_t)new > vm->vm_daddr && !kern_break(td, &new))
238 		td->td_retval[0] = (register_t)new;
239 	else
240 		td->td_retval[0] = (register_t)old;
241 
242 	return (0);
243 }
244 
245 #if defined(__i386__)
246 /* XXX: what about amd64/linux32? */
247 
248 int
249 linux_uselib(struct thread *td, struct linux_uselib_args *args)
250 {
251 	struct nameidata ni;
252 	struct vnode *vp;
253 	struct exec *a_out;
254 	vm_map_t map;
255 	vm_map_entry_t entry;
256 	struct vattr attr;
257 	vm_offset_t vmaddr;
258 	unsigned long file_offset;
259 	unsigned long bss_size;
260 	char *library;
261 	ssize_t aresid;
262 	int error;
263 	bool locked, opened, textset;
264 
265 	a_out = NULL;
266 	vp = NULL;
267 	locked = false;
268 	textset = false;
269 	opened = false;
270 
271 	if (!LUSECONVPATH(td)) {
272 		NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
273 		    UIO_USERSPACE, args->library, td);
274 		error = namei(&ni);
275 	} else {
276 		LCONVPATHEXIST(td, args->library, &library);
277 		NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
278 		    UIO_SYSSPACE, library, td);
279 		error = namei(&ni);
280 		LFREEPATH(library);
281 	}
282 	if (error)
283 		goto cleanup;
284 
285 	vp = ni.ni_vp;
286 	NDFREE(&ni, NDF_ONLY_PNBUF);
287 
288 	/*
289 	 * From here on down, we have a locked vnode that must be unlocked.
290 	 * XXX: The code below largely duplicates exec_check_permissions().
291 	 */
292 	locked = true;
293 
294 	/* Executable? */
295 	error = VOP_GETATTR(vp, &attr, td->td_ucred);
296 	if (error)
297 		goto cleanup;
298 
299 	if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
300 	    ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
301 		/* EACCESS is what exec(2) returns. */
302 		error = ENOEXEC;
303 		goto cleanup;
304 	}
305 
306 	/* Sensible size? */
307 	if (attr.va_size == 0) {
308 		error = ENOEXEC;
309 		goto cleanup;
310 	}
311 
312 	/* Can we access it? */
313 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
314 	if (error)
315 		goto cleanup;
316 
317 	/*
318 	 * XXX: This should use vn_open() so that it is properly authorized,
319 	 * and to reduce code redundancy all over the place here.
320 	 * XXX: Not really, it duplicates far more of exec_check_permissions()
321 	 * than vn_open().
322 	 */
323 #ifdef MAC
324 	error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
325 	if (error)
326 		goto cleanup;
327 #endif
328 	error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
329 	if (error)
330 		goto cleanup;
331 	opened = true;
332 
333 	/* Pull in executable header into exec_map */
334 	error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
335 	    VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
336 	if (error)
337 		goto cleanup;
338 
339 	/* Is it a Linux binary ? */
340 	if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
341 		error = ENOEXEC;
342 		goto cleanup;
343 	}
344 
345 	/*
346 	 * While we are here, we should REALLY do some more checks
347 	 */
348 
349 	/* Set file/virtual offset based on a.out variant. */
350 	switch ((int)(a_out->a_magic & 0xffff)) {
351 	case 0413:			/* ZMAGIC */
352 		file_offset = 1024;
353 		break;
354 	case 0314:			/* QMAGIC */
355 		file_offset = 0;
356 		break;
357 	default:
358 		error = ENOEXEC;
359 		goto cleanup;
360 	}
361 
362 	bss_size = round_page(a_out->a_bss);
363 
364 	/* Check various fields in header for validity/bounds. */
365 	if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
366 		error = ENOEXEC;
367 		goto cleanup;
368 	}
369 
370 	/* text + data can't exceed file size */
371 	if (a_out->a_data + a_out->a_text > attr.va_size) {
372 		error = EFAULT;
373 		goto cleanup;
374 	}
375 
376 	/*
377 	 * text/data/bss must not exceed limits
378 	 * XXX - this is not complete. it should check current usage PLUS
379 	 * the resources needed by this library.
380 	 */
381 	PROC_LOCK(td->td_proc);
382 	if (a_out->a_text > maxtsiz ||
383 	    a_out->a_data + bss_size > lim_cur_proc(td->td_proc, RLIMIT_DATA) ||
384 	    racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
385 	    bss_size) != 0) {
386 		PROC_UNLOCK(td->td_proc);
387 		error = ENOMEM;
388 		goto cleanup;
389 	}
390 	PROC_UNLOCK(td->td_proc);
391 
392 	/*
393 	 * Prevent more writers.
394 	 */
395 	error = VOP_SET_TEXT(vp);
396 	if (error != 0)
397 		goto cleanup;
398 	textset = true;
399 
400 	/*
401 	 * Lock no longer needed
402 	 */
403 	locked = false;
404 	VOP_UNLOCK(vp);
405 
406 	/*
407 	 * Check if file_offset page aligned. Currently we cannot handle
408 	 * misalinged file offsets, and so we read in the entire image
409 	 * (what a waste).
410 	 */
411 	if (file_offset & PAGE_MASK) {
412 		/* Map text+data read/write/execute */
413 
414 		/* a_entry is the load address and is page aligned */
415 		vmaddr = trunc_page(a_out->a_entry);
416 
417 		/* get anon user mapping, read+write+execute */
418 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
419 		    &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
420 		    VM_PROT_ALL, VM_PROT_ALL, 0);
421 		if (error)
422 			goto cleanup;
423 
424 		error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
425 		    a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
426 		    td->td_ucred, NOCRED, &aresid, td);
427 		if (error != 0)
428 			goto cleanup;
429 		if (aresid != 0) {
430 			error = ENOEXEC;
431 			goto cleanup;
432 		}
433 	} else {
434 		/*
435 		 * for QMAGIC, a_entry is 20 bytes beyond the load address
436 		 * to skip the executable header
437 		 */
438 		vmaddr = trunc_page(a_out->a_entry);
439 
440 		/*
441 		 * Map it all into the process's space as a single
442 		 * copy-on-write "data" segment.
443 		 */
444 		map = &td->td_proc->p_vmspace->vm_map;
445 		error = vm_mmap(map, &vmaddr,
446 		    a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
447 		    MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
448 		if (error)
449 			goto cleanup;
450 		vm_map_lock(map);
451 		if (!vm_map_lookup_entry(map, vmaddr, &entry)) {
452 			vm_map_unlock(map);
453 			error = EDOOFUS;
454 			goto cleanup;
455 		}
456 		entry->eflags |= MAP_ENTRY_VN_EXEC;
457 		vm_map_unlock(map);
458 		textset = false;
459 	}
460 
461 	if (bss_size != 0) {
462 		/* Calculate BSS start address */
463 		vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
464 		    a_out->a_data;
465 
466 		/* allocate some 'anon' space */
467 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
468 		    &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
469 		    VM_PROT_ALL, 0);
470 		if (error)
471 			goto cleanup;
472 	}
473 
474 cleanup:
475 	if (opened) {
476 		if (locked)
477 			VOP_UNLOCK(vp);
478 		locked = false;
479 		VOP_CLOSE(vp, FREAD, td->td_ucred, td);
480 	}
481 	if (textset) {
482 		if (!locked) {
483 			locked = true;
484 			VOP_LOCK(vp, LK_SHARED | LK_RETRY);
485 		}
486 		VOP_UNSET_TEXT_CHECKED(vp);
487 	}
488 	if (locked)
489 		VOP_UNLOCK(vp);
490 
491 	/* Release the temporary mapping. */
492 	if (a_out)
493 		kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
494 
495 	return (error);
496 }
497 
498 #endif	/* __i386__ */
499 
500 #ifdef LINUX_LEGACY_SYSCALLS
501 int
502 linux_select(struct thread *td, struct linux_select_args *args)
503 {
504 	l_timeval ltv;
505 	struct timeval tv0, tv1, utv, *tvp;
506 	int error;
507 
508 	/*
509 	 * Store current time for computation of the amount of
510 	 * time left.
511 	 */
512 	if (args->timeout) {
513 		if ((error = copyin(args->timeout, &ltv, sizeof(ltv))))
514 			goto select_out;
515 		utv.tv_sec = ltv.tv_sec;
516 		utv.tv_usec = ltv.tv_usec;
517 
518 		if (itimerfix(&utv)) {
519 			/*
520 			 * The timeval was invalid.  Convert it to something
521 			 * valid that will act as it does under Linux.
522 			 */
523 			utv.tv_sec += utv.tv_usec / 1000000;
524 			utv.tv_usec %= 1000000;
525 			if (utv.tv_usec < 0) {
526 				utv.tv_sec -= 1;
527 				utv.tv_usec += 1000000;
528 			}
529 			if (utv.tv_sec < 0)
530 				timevalclear(&utv);
531 		}
532 		microtime(&tv0);
533 		tvp = &utv;
534 	} else
535 		tvp = NULL;
536 
537 	error = kern_select(td, args->nfds, args->readfds, args->writefds,
538 	    args->exceptfds, tvp, LINUX_NFDBITS);
539 	if (error)
540 		goto select_out;
541 
542 	if (args->timeout) {
543 		if (td->td_retval[0]) {
544 			/*
545 			 * Compute how much time was left of the timeout,
546 			 * by subtracting the current time and the time
547 			 * before we started the call, and subtracting
548 			 * that result from the user-supplied value.
549 			 */
550 			microtime(&tv1);
551 			timevalsub(&tv1, &tv0);
552 			timevalsub(&utv, &tv1);
553 			if (utv.tv_sec < 0)
554 				timevalclear(&utv);
555 		} else
556 			timevalclear(&utv);
557 		ltv.tv_sec = utv.tv_sec;
558 		ltv.tv_usec = utv.tv_usec;
559 		if ((error = copyout(&ltv, args->timeout, sizeof(ltv))))
560 			goto select_out;
561 	}
562 
563 select_out:
564 	return (error);
565 }
566 #endif
567 
568 int
569 linux_mremap(struct thread *td, struct linux_mremap_args *args)
570 {
571 	uintptr_t addr;
572 	size_t len;
573 	int error = 0;
574 
575 	if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
576 		td->td_retval[0] = 0;
577 		return (EINVAL);
578 	}
579 
580 	/*
581 	 * Check for the page alignment.
582 	 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
583 	 */
584 	if (args->addr & PAGE_MASK) {
585 		td->td_retval[0] = 0;
586 		return (EINVAL);
587 	}
588 
589 	args->new_len = round_page(args->new_len);
590 	args->old_len = round_page(args->old_len);
591 
592 	if (args->new_len > args->old_len) {
593 		td->td_retval[0] = 0;
594 		return (ENOMEM);
595 	}
596 
597 	if (args->new_len < args->old_len) {
598 		addr = args->addr + args->new_len;
599 		len = args->old_len - args->new_len;
600 		error = kern_munmap(td, addr, len);
601 	}
602 
603 	td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
604 	return (error);
605 }
606 
607 #define LINUX_MS_ASYNC       0x0001
608 #define LINUX_MS_INVALIDATE  0x0002
609 #define LINUX_MS_SYNC        0x0004
610 
611 int
612 linux_msync(struct thread *td, struct linux_msync_args *args)
613 {
614 
615 	return (kern_msync(td, args->addr, args->len,
616 	    args->fl & ~LINUX_MS_SYNC));
617 }
618 
619 #ifdef LINUX_LEGACY_SYSCALLS
620 int
621 linux_time(struct thread *td, struct linux_time_args *args)
622 {
623 	struct timeval tv;
624 	l_time_t tm;
625 	int error;
626 
627 	microtime(&tv);
628 	tm = tv.tv_sec;
629 	if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
630 		return (error);
631 	td->td_retval[0] = tm;
632 	return (0);
633 }
634 #endif
635 
636 struct l_times_argv {
637 	l_clock_t	tms_utime;
638 	l_clock_t	tms_stime;
639 	l_clock_t	tms_cutime;
640 	l_clock_t	tms_cstime;
641 };
642 
643 /*
644  * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
645  * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
646  * auxiliary vector entry.
647  */
648 #define	CLK_TCK		100
649 
650 #define	CONVOTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
651 #define	CONVNTCK(r)	(r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
652 
653 #define	CONVTCK(r)	(linux_kernver(td) >= LINUX_KERNVER_2004000 ?		\
654 			    CONVNTCK(r) : CONVOTCK(r))
655 
656 int
657 linux_times(struct thread *td, struct linux_times_args *args)
658 {
659 	struct timeval tv, utime, stime, cutime, cstime;
660 	struct l_times_argv tms;
661 	struct proc *p;
662 	int error;
663 
664 	if (args->buf != NULL) {
665 		p = td->td_proc;
666 		PROC_LOCK(p);
667 		PROC_STATLOCK(p);
668 		calcru(p, &utime, &stime);
669 		PROC_STATUNLOCK(p);
670 		calccru(p, &cutime, &cstime);
671 		PROC_UNLOCK(p);
672 
673 		tms.tms_utime = CONVTCK(utime);
674 		tms.tms_stime = CONVTCK(stime);
675 
676 		tms.tms_cutime = CONVTCK(cutime);
677 		tms.tms_cstime = CONVTCK(cstime);
678 
679 		if ((error = copyout(&tms, args->buf, sizeof(tms))))
680 			return (error);
681 	}
682 
683 	microuptime(&tv);
684 	td->td_retval[0] = (int)CONVTCK(tv);
685 	return (0);
686 }
687 
688 int
689 linux_newuname(struct thread *td, struct linux_newuname_args *args)
690 {
691 	struct l_new_utsname utsname;
692 	char osname[LINUX_MAX_UTSNAME];
693 	char osrelease[LINUX_MAX_UTSNAME];
694 	char *p;
695 
696 	linux_get_osname(td, osname);
697 	linux_get_osrelease(td, osrelease);
698 
699 	bzero(&utsname, sizeof(utsname));
700 	strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
701 	getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
702 	getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
703 	strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
704 	strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
705 	for (p = utsname.version; *p != '\0'; ++p)
706 		if (*p == '\n') {
707 			*p = '\0';
708 			break;
709 		}
710 #if defined(__amd64__)
711 	/*
712 	 * On amd64, Linux uname(2) needs to return "x86_64"
713 	 * for both 64-bit and 32-bit applications.  On 32-bit,
714 	 * the string returned by getauxval(AT_PLATFORM) needs
715 	 * to remain "i686", though.
716 	 */
717 	strlcpy(utsname.machine, "x86_64", LINUX_MAX_UTSNAME);
718 #else
719 	strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME);
720 #endif
721 
722 	return (copyout(&utsname, args->buf, sizeof(utsname)));
723 }
724 
725 struct l_utimbuf {
726 	l_time_t l_actime;
727 	l_time_t l_modtime;
728 };
729 
730 #ifdef LINUX_LEGACY_SYSCALLS
731 int
732 linux_utime(struct thread *td, struct linux_utime_args *args)
733 {
734 	struct timeval tv[2], *tvp;
735 	struct l_utimbuf lut;
736 	char *fname;
737 	int error;
738 	bool convpath;
739 
740 	convpath = LUSECONVPATH(td);
741 	if (convpath)
742 		LCONVPATHEXIST(td, args->fname, &fname);
743 
744 	if (args->times) {
745 		if ((error = copyin(args->times, &lut, sizeof lut))) {
746 			if (convpath)
747 				LFREEPATH(fname);
748 			return (error);
749 		}
750 		tv[0].tv_sec = lut.l_actime;
751 		tv[0].tv_usec = 0;
752 		tv[1].tv_sec = lut.l_modtime;
753 		tv[1].tv_usec = 0;
754 		tvp = tv;
755 	} else
756 		tvp = NULL;
757 
758 	if (!convpath) {
759 		error = kern_utimesat(td, AT_FDCWD, args->fname, UIO_USERSPACE,
760 		    tvp, UIO_SYSSPACE);
761 	} else {
762 		error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, tvp,
763 		    UIO_SYSSPACE);
764 		LFREEPATH(fname);
765 	}
766 	return (error);
767 }
768 #endif
769 
770 #ifdef LINUX_LEGACY_SYSCALLS
771 int
772 linux_utimes(struct thread *td, struct linux_utimes_args *args)
773 {
774 	l_timeval ltv[2];
775 	struct timeval tv[2], *tvp = NULL;
776 	char *fname;
777 	int error;
778 	bool convpath;
779 
780 	convpath = LUSECONVPATH(td);
781 	if (convpath)
782 		LCONVPATHEXIST(td, args->fname, &fname);
783 
784 	if (args->tptr != NULL) {
785 		if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
786 			LFREEPATH(fname);
787 			return (error);
788 		}
789 		tv[0].tv_sec = ltv[0].tv_sec;
790 		tv[0].tv_usec = ltv[0].tv_usec;
791 		tv[1].tv_sec = ltv[1].tv_sec;
792 		tv[1].tv_usec = ltv[1].tv_usec;
793 		tvp = tv;
794 	}
795 
796 	if (!convpath) {
797 		error = kern_utimesat(td, AT_FDCWD, args->fname, UIO_USERSPACE,
798 		    tvp, UIO_SYSSPACE);
799 	} else {
800 		error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE,
801 		    tvp, UIO_SYSSPACE);
802 		LFREEPATH(fname);
803 	}
804 	return (error);
805 }
806 #endif
807 
808 static int
809 linux_utimensat_nsec_valid(l_long nsec)
810 {
811 
812 	if (nsec == LINUX_UTIME_OMIT || nsec == LINUX_UTIME_NOW)
813 		return (0);
814 	if (nsec >= 0 && nsec <= 999999999)
815 		return (0);
816 	return (1);
817 }
818 
819 int
820 linux_utimensat(struct thread *td, struct linux_utimensat_args *args)
821 {
822 	struct l_timespec l_times[2];
823 	struct timespec times[2], *timesp = NULL;
824 	char *path = NULL;
825 	int error, dfd, flags = 0;
826 
827 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
828 
829 	if (args->flags & ~LINUX_AT_SYMLINK_NOFOLLOW)
830 		return (EINVAL);
831 
832 	if (args->times != NULL) {
833 		error = copyin(args->times, l_times, sizeof(l_times));
834 		if (error != 0)
835 			return (error);
836 
837 		if (linux_utimensat_nsec_valid(l_times[0].tv_nsec) != 0 ||
838 		    linux_utimensat_nsec_valid(l_times[1].tv_nsec) != 0)
839 			return (EINVAL);
840 
841 		times[0].tv_sec = l_times[0].tv_sec;
842 		switch (l_times[0].tv_nsec)
843 		{
844 		case LINUX_UTIME_OMIT:
845 			times[0].tv_nsec = UTIME_OMIT;
846 			break;
847 		case LINUX_UTIME_NOW:
848 			times[0].tv_nsec = UTIME_NOW;
849 			break;
850 		default:
851 			times[0].tv_nsec = l_times[0].tv_nsec;
852 		}
853 
854 		times[1].tv_sec = l_times[1].tv_sec;
855 		switch (l_times[1].tv_nsec)
856 		{
857 		case LINUX_UTIME_OMIT:
858 			times[1].tv_nsec = UTIME_OMIT;
859 			break;
860 		case LINUX_UTIME_NOW:
861 			times[1].tv_nsec = UTIME_NOW;
862 			break;
863 		default:
864 			times[1].tv_nsec = l_times[1].tv_nsec;
865 			break;
866 		}
867 		timesp = times;
868 
869 		/* This breaks POSIX, but is what the Linux kernel does
870 		 * _on purpose_ (documented in the man page for utimensat(2)),
871 		 * so we must follow that behaviour. */
872 		if (times[0].tv_nsec == UTIME_OMIT &&
873 		    times[1].tv_nsec == UTIME_OMIT)
874 			return (0);
875 	}
876 
877 	if (args->pathname != NULL)
878 		LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
879 	else if (args->flags != 0)
880 		return (EINVAL);
881 
882 	if (args->flags & LINUX_AT_SYMLINK_NOFOLLOW)
883 		flags |= AT_SYMLINK_NOFOLLOW;
884 
885 	if (path == NULL)
886 		error = kern_futimens(td, dfd, timesp, UIO_SYSSPACE);
887 	else {
888 		error = kern_utimensat(td, dfd, path, UIO_SYSSPACE, timesp,
889 			UIO_SYSSPACE, flags);
890 		LFREEPATH(path);
891 	}
892 
893 	return (error);
894 }
895 
896 #ifdef LINUX_LEGACY_SYSCALLS
897 int
898 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
899 {
900 	l_timeval ltv[2];
901 	struct timeval tv[2], *tvp = NULL;
902 	char *fname;
903 	int error, dfd;
904 	bool convpath;
905 
906 	convpath = LUSECONVPATH(td);
907 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
908 	if (convpath)
909 		LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
910 
911 	if (args->utimes != NULL) {
912 		if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
913 			if (convpath)
914 				LFREEPATH(fname);
915 			return (error);
916 		}
917 		tv[0].tv_sec = ltv[0].tv_sec;
918 		tv[0].tv_usec = ltv[0].tv_usec;
919 		tv[1].tv_sec = ltv[1].tv_sec;
920 		tv[1].tv_usec = ltv[1].tv_usec;
921 		tvp = tv;
922 	}
923 
924 	if (!convpath) {
925 		error = kern_utimesat(td, dfd, args->filename, UIO_USERSPACE,
926 		    tvp, UIO_SYSSPACE);
927 	} else {
928 		error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
929 		LFREEPATH(fname);
930 	}
931 	return (error);
932 }
933 #endif
934 
935 static int
936 linux_common_wait(struct thread *td, int pid, int *statusp,
937     int options, struct __wrusage *wrup)
938 {
939 	siginfo_t siginfo;
940 	idtype_t idtype;
941 	id_t id;
942 	int error, status, tmpstat;
943 
944 	if (pid == WAIT_ANY) {
945 		idtype = P_ALL;
946 		id = 0;
947 	} else if (pid < 0) {
948 		idtype = P_PGID;
949 		id = (id_t)-pid;
950 	} else {
951 		idtype = P_PID;
952 		id = (id_t)pid;
953 	}
954 
955 	/*
956 	 * For backward compatibility we implicitly add flags WEXITED
957 	 * and WTRAPPED here.
958 	 */
959 	options |= WEXITED | WTRAPPED;
960 	error = kern_wait6(td, idtype, id, &status, options, wrup, &siginfo);
961 	if (error)
962 		return (error);
963 
964 	if (statusp) {
965 		tmpstat = status & 0xffff;
966 		if (WIFSIGNALED(tmpstat)) {
967 			tmpstat = (tmpstat & 0xffffff80) |
968 			    bsd_to_linux_signal(WTERMSIG(tmpstat));
969 		} else if (WIFSTOPPED(tmpstat)) {
970 			tmpstat = (tmpstat & 0xffff00ff) |
971 			    (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8);
972 #if defined(__amd64__) && !defined(COMPAT_LINUX32)
973 			if (WSTOPSIG(status) == SIGTRAP) {
974 				tmpstat = linux_ptrace_status(td,
975 				    siginfo.si_pid, tmpstat);
976 			}
977 #endif
978 		} else if (WIFCONTINUED(tmpstat)) {
979 			tmpstat = 0xffff;
980 		}
981 		error = copyout(&tmpstat, statusp, sizeof(int));
982 	}
983 
984 	return (error);
985 }
986 
987 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
988 int
989 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
990 {
991 	struct linux_wait4_args wait4_args;
992 
993 	wait4_args.pid = args->pid;
994 	wait4_args.status = args->status;
995 	wait4_args.options = args->options;
996 	wait4_args.rusage = NULL;
997 
998 	return (linux_wait4(td, &wait4_args));
999 }
1000 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1001 
1002 int
1003 linux_wait4(struct thread *td, struct linux_wait4_args *args)
1004 {
1005 	int error, options;
1006 	struct __wrusage wru, *wrup;
1007 
1008 	if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG |
1009 	    LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
1010 		return (EINVAL);
1011 
1012 	options = WEXITED;
1013 	linux_to_bsd_waitopts(args->options, &options);
1014 
1015 	if (args->rusage != NULL)
1016 		wrup = &wru;
1017 	else
1018 		wrup = NULL;
1019 	error = linux_common_wait(td, args->pid, args->status, options, wrup);
1020 	if (error != 0)
1021 		return (error);
1022 	if (args->rusage != NULL)
1023 		error = linux_copyout_rusage(&wru.wru_self, args->rusage);
1024 	return (error);
1025 }
1026 
1027 int
1028 linux_waitid(struct thread *td, struct linux_waitid_args *args)
1029 {
1030 	int status, options, sig;
1031 	struct __wrusage wru;
1032 	siginfo_t siginfo;
1033 	l_siginfo_t lsi;
1034 	idtype_t idtype;
1035 	struct proc *p;
1036 	int error;
1037 
1038 	options = 0;
1039 	linux_to_bsd_waitopts(args->options, &options);
1040 
1041 	if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED))
1042 		return (EINVAL);
1043 	if (!(options & (WEXITED | WUNTRACED | WCONTINUED)))
1044 		return (EINVAL);
1045 
1046 	switch (args->idtype) {
1047 	case LINUX_P_ALL:
1048 		idtype = P_ALL;
1049 		break;
1050 	case LINUX_P_PID:
1051 		if (args->id <= 0)
1052 			return (EINVAL);
1053 		idtype = P_PID;
1054 		break;
1055 	case LINUX_P_PGID:
1056 		if (args->id <= 0)
1057 			return (EINVAL);
1058 		idtype = P_PGID;
1059 		break;
1060 	default:
1061 		return (EINVAL);
1062 	}
1063 
1064 	error = kern_wait6(td, idtype, args->id, &status, options,
1065 	    &wru, &siginfo);
1066 	if (error != 0)
1067 		return (error);
1068 	if (args->rusage != NULL) {
1069 		error = linux_copyout_rusage(&wru.wru_children,
1070 		    args->rusage);
1071 		if (error != 0)
1072 			return (error);
1073 	}
1074 	if (args->info != NULL) {
1075 		p = td->td_proc;
1076 		bzero(&lsi, sizeof(lsi));
1077 		if (td->td_retval[0] != 0) {
1078 			sig = bsd_to_linux_signal(siginfo.si_signo);
1079 			siginfo_to_lsiginfo(&siginfo, &lsi, sig);
1080 		}
1081 		error = copyout(&lsi, args->info, sizeof(lsi));
1082 	}
1083 	td->td_retval[0] = 0;
1084 
1085 	return (error);
1086 }
1087 
1088 #ifdef LINUX_LEGACY_SYSCALLS
1089 int
1090 linux_mknod(struct thread *td, struct linux_mknod_args *args)
1091 {
1092 	char *path;
1093 	int error;
1094 	enum uio_seg seg;
1095 	bool convpath;
1096 
1097 	convpath = LUSECONVPATH(td);
1098 	if (!convpath) {
1099 		path = args->path;
1100 		seg = UIO_USERSPACE;
1101 	} else {
1102 		LCONVPATHCREAT(td, args->path, &path);
1103 		seg = UIO_SYSSPACE;
1104 	}
1105 
1106 	switch (args->mode & S_IFMT) {
1107 	case S_IFIFO:
1108 	case S_IFSOCK:
1109 		error = kern_mkfifoat(td, AT_FDCWD, path, seg,
1110 		    args->mode);
1111 		break;
1112 
1113 	case S_IFCHR:
1114 	case S_IFBLK:
1115 		error = kern_mknodat(td, AT_FDCWD, path, seg,
1116 		    args->mode, args->dev);
1117 		break;
1118 
1119 	case S_IFDIR:
1120 		error = EPERM;
1121 		break;
1122 
1123 	case 0:
1124 		args->mode |= S_IFREG;
1125 		/* FALLTHROUGH */
1126 	case S_IFREG:
1127 		error = kern_openat(td, AT_FDCWD, path, seg,
1128 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1129 		if (error == 0)
1130 			kern_close(td, td->td_retval[0]);
1131 		break;
1132 
1133 	default:
1134 		error = EINVAL;
1135 		break;
1136 	}
1137 	if (convpath)
1138 		LFREEPATH(path);
1139 	return (error);
1140 }
1141 #endif
1142 
1143 int
1144 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
1145 {
1146 	char *path;
1147 	int error, dfd;
1148 	enum uio_seg seg;
1149 	bool convpath;
1150 
1151 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
1152 
1153 	convpath = LUSECONVPATH(td);
1154 	if (!convpath) {
1155 		path = __DECONST(char *, args->filename);
1156 		seg = UIO_USERSPACE;
1157 	} else {
1158 		LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
1159 		seg = UIO_SYSSPACE;
1160 	}
1161 
1162 	switch (args->mode & S_IFMT) {
1163 	case S_IFIFO:
1164 	case S_IFSOCK:
1165 		error = kern_mkfifoat(td, dfd, path, seg, args->mode);
1166 		break;
1167 
1168 	case S_IFCHR:
1169 	case S_IFBLK:
1170 		error = kern_mknodat(td, dfd, path, seg, args->mode,
1171 		    args->dev);
1172 		break;
1173 
1174 	case S_IFDIR:
1175 		error = EPERM;
1176 		break;
1177 
1178 	case 0:
1179 		args->mode |= S_IFREG;
1180 		/* FALLTHROUGH */
1181 	case S_IFREG:
1182 		error = kern_openat(td, dfd, path, seg,
1183 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1184 		if (error == 0)
1185 			kern_close(td, td->td_retval[0]);
1186 		break;
1187 
1188 	default:
1189 		error = EINVAL;
1190 		break;
1191 	}
1192 	if (convpath)
1193 		LFREEPATH(path);
1194 	return (error);
1195 }
1196 
1197 /*
1198  * UGH! This is just about the dumbest idea I've ever heard!!
1199  */
1200 int
1201 linux_personality(struct thread *td, struct linux_personality_args *args)
1202 {
1203 	struct linux_pemuldata *pem;
1204 	struct proc *p = td->td_proc;
1205 	uint32_t old;
1206 
1207 	PROC_LOCK(p);
1208 	pem = pem_find(p);
1209 	old = pem->persona;
1210 	if (args->per != 0xffffffff)
1211 		pem->persona = args->per;
1212 	PROC_UNLOCK(p);
1213 
1214 	td->td_retval[0] = old;
1215 	return (0);
1216 }
1217 
1218 struct l_itimerval {
1219 	l_timeval it_interval;
1220 	l_timeval it_value;
1221 };
1222 
1223 #define	B2L_ITIMERVAL(bip, lip)						\
1224 	(bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec;		\
1225 	(bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec;	\
1226 	(bip)->it_value.tv_sec = (lip)->it_value.tv_sec;		\
1227 	(bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1228 
1229 int
1230 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1231 {
1232 	int error;
1233 	struct l_itimerval ls;
1234 	struct itimerval aitv, oitv;
1235 
1236 	if (uap->itv == NULL) {
1237 		uap->itv = uap->oitv;
1238 		return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1239 	}
1240 
1241 	error = copyin(uap->itv, &ls, sizeof(ls));
1242 	if (error != 0)
1243 		return (error);
1244 	B2L_ITIMERVAL(&aitv, &ls);
1245 	error = kern_setitimer(td, uap->which, &aitv, &oitv);
1246 	if (error != 0 || uap->oitv == NULL)
1247 		return (error);
1248 	B2L_ITIMERVAL(&ls, &oitv);
1249 
1250 	return (copyout(&ls, uap->oitv, sizeof(ls)));
1251 }
1252 
1253 int
1254 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1255 {
1256 	int error;
1257 	struct l_itimerval ls;
1258 	struct itimerval aitv;
1259 
1260 	error = kern_getitimer(td, uap->which, &aitv);
1261 	if (error != 0)
1262 		return (error);
1263 	B2L_ITIMERVAL(&ls, &aitv);
1264 	return (copyout(&ls, uap->itv, sizeof(ls)));
1265 }
1266 
1267 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1268 int
1269 linux_nice(struct thread *td, struct linux_nice_args *args)
1270 {
1271 
1272 	return (kern_setpriority(td, PRIO_PROCESS, 0, args->inc));
1273 }
1274 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1275 
1276 int
1277 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1278 {
1279 	struct ucred *newcred, *oldcred;
1280 	l_gid_t *linux_gidset;
1281 	gid_t *bsd_gidset;
1282 	int ngrp, error;
1283 	struct proc *p;
1284 
1285 	ngrp = args->gidsetsize;
1286 	if (ngrp < 0 || ngrp >= ngroups_max + 1)
1287 		return (EINVAL);
1288 	linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK);
1289 	error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1290 	if (error)
1291 		goto out;
1292 	newcred = crget();
1293 	crextend(newcred, ngrp + 1);
1294 	p = td->td_proc;
1295 	PROC_LOCK(p);
1296 	oldcred = p->p_ucred;
1297 	crcopy(newcred, oldcred);
1298 
1299 	/*
1300 	 * cr_groups[0] holds egid. Setting the whole set from
1301 	 * the supplied set will cause egid to be changed too.
1302 	 * Keep cr_groups[0] unchanged to prevent that.
1303 	 */
1304 
1305 	if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS)) != 0) {
1306 		PROC_UNLOCK(p);
1307 		crfree(newcred);
1308 		goto out;
1309 	}
1310 
1311 	if (ngrp > 0) {
1312 		newcred->cr_ngroups = ngrp + 1;
1313 
1314 		bsd_gidset = newcred->cr_groups;
1315 		ngrp--;
1316 		while (ngrp >= 0) {
1317 			bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1318 			ngrp--;
1319 		}
1320 	} else
1321 		newcred->cr_ngroups = 1;
1322 
1323 	setsugid(p);
1324 	proc_set_cred(p, newcred);
1325 	PROC_UNLOCK(p);
1326 	crfree(oldcred);
1327 	error = 0;
1328 out:
1329 	free(linux_gidset, M_LINUX);
1330 	return (error);
1331 }
1332 
1333 int
1334 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1335 {
1336 	struct ucred *cred;
1337 	l_gid_t *linux_gidset;
1338 	gid_t *bsd_gidset;
1339 	int bsd_gidsetsz, ngrp, error;
1340 
1341 	cred = td->td_ucred;
1342 	bsd_gidset = cred->cr_groups;
1343 	bsd_gidsetsz = cred->cr_ngroups - 1;
1344 
1345 	/*
1346 	 * cr_groups[0] holds egid. Returning the whole set
1347 	 * here will cause a duplicate. Exclude cr_groups[0]
1348 	 * to prevent that.
1349 	 */
1350 
1351 	if ((ngrp = args->gidsetsize) == 0) {
1352 		td->td_retval[0] = bsd_gidsetsz;
1353 		return (0);
1354 	}
1355 
1356 	if (ngrp < bsd_gidsetsz)
1357 		return (EINVAL);
1358 
1359 	ngrp = 0;
1360 	linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1361 	    M_LINUX, M_WAITOK);
1362 	while (ngrp < bsd_gidsetsz) {
1363 		linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1364 		ngrp++;
1365 	}
1366 
1367 	error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1368 	free(linux_gidset, M_LINUX);
1369 	if (error)
1370 		return (error);
1371 
1372 	td->td_retval[0] = ngrp;
1373 	return (0);
1374 }
1375 
1376 static bool
1377 linux_get_dummy_limit(l_uint resource, struct rlimit *rlim)
1378 {
1379 
1380 	if (linux_dummy_rlimits == 0)
1381 		return (false);
1382 
1383 	switch (resource) {
1384 	case LINUX_RLIMIT_LOCKS:
1385 	case LINUX_RLIMIT_SIGPENDING:
1386 	case LINUX_RLIMIT_MSGQUEUE:
1387 	case LINUX_RLIMIT_RTTIME:
1388 		rlim->rlim_cur = LINUX_RLIM_INFINITY;
1389 		rlim->rlim_max = LINUX_RLIM_INFINITY;
1390 		return (true);
1391 	case LINUX_RLIMIT_NICE:
1392 	case LINUX_RLIMIT_RTPRIO:
1393 		rlim->rlim_cur = 0;
1394 		rlim->rlim_max = 0;
1395 		return (true);
1396 	default:
1397 		return (false);
1398 	}
1399 }
1400 
1401 int
1402 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1403 {
1404 	struct rlimit bsd_rlim;
1405 	struct l_rlimit rlim;
1406 	u_int which;
1407 	int error;
1408 
1409 	if (args->resource >= LINUX_RLIM_NLIMITS)
1410 		return (EINVAL);
1411 
1412 	which = linux_to_bsd_resource[args->resource];
1413 	if (which == -1)
1414 		return (EINVAL);
1415 
1416 	error = copyin(args->rlim, &rlim, sizeof(rlim));
1417 	if (error)
1418 		return (error);
1419 
1420 	bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1421 	bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1422 	return (kern_setrlimit(td, which, &bsd_rlim));
1423 }
1424 
1425 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1426 int
1427 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1428 {
1429 	struct l_rlimit rlim;
1430 	struct rlimit bsd_rlim;
1431 	u_int which;
1432 
1433 	if (linux_get_dummy_limit(args->resource, &bsd_rlim)) {
1434 		rlim.rlim_cur = bsd_rlim.rlim_cur;
1435 		rlim.rlim_max = bsd_rlim.rlim_max;
1436 		return (copyout(&rlim, args->rlim, sizeof(rlim)));
1437 	}
1438 
1439 	if (args->resource >= LINUX_RLIM_NLIMITS)
1440 		return (EINVAL);
1441 
1442 	which = linux_to_bsd_resource[args->resource];
1443 	if (which == -1)
1444 		return (EINVAL);
1445 
1446 	lim_rlimit(td, which, &bsd_rlim);
1447 
1448 #ifdef COMPAT_LINUX32
1449 	rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1450 	if (rlim.rlim_cur == UINT_MAX)
1451 		rlim.rlim_cur = INT_MAX;
1452 	rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1453 	if (rlim.rlim_max == UINT_MAX)
1454 		rlim.rlim_max = INT_MAX;
1455 #else
1456 	rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1457 	if (rlim.rlim_cur == ULONG_MAX)
1458 		rlim.rlim_cur = LONG_MAX;
1459 	rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1460 	if (rlim.rlim_max == ULONG_MAX)
1461 		rlim.rlim_max = LONG_MAX;
1462 #endif
1463 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1464 }
1465 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1466 
1467 int
1468 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1469 {
1470 	struct l_rlimit rlim;
1471 	struct rlimit bsd_rlim;
1472 	u_int which;
1473 
1474 	if (linux_get_dummy_limit(args->resource, &bsd_rlim)) {
1475 		rlim.rlim_cur = bsd_rlim.rlim_cur;
1476 		rlim.rlim_max = bsd_rlim.rlim_max;
1477 		return (copyout(&rlim, args->rlim, sizeof(rlim)));
1478 	}
1479 
1480 	if (args->resource >= LINUX_RLIM_NLIMITS)
1481 		return (EINVAL);
1482 
1483 	which = linux_to_bsd_resource[args->resource];
1484 	if (which == -1)
1485 		return (EINVAL);
1486 
1487 	lim_rlimit(td, which, &bsd_rlim);
1488 
1489 	rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1490 	rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1491 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1492 }
1493 
1494 int
1495 linux_sched_setscheduler(struct thread *td,
1496     struct linux_sched_setscheduler_args *args)
1497 {
1498 	struct sched_param sched_param;
1499 	struct thread *tdt;
1500 	int error, policy;
1501 
1502 	switch (args->policy) {
1503 	case LINUX_SCHED_OTHER:
1504 		policy = SCHED_OTHER;
1505 		break;
1506 	case LINUX_SCHED_FIFO:
1507 		policy = SCHED_FIFO;
1508 		break;
1509 	case LINUX_SCHED_RR:
1510 		policy = SCHED_RR;
1511 		break;
1512 	default:
1513 		return (EINVAL);
1514 	}
1515 
1516 	error = copyin(args->param, &sched_param, sizeof(sched_param));
1517 	if (error)
1518 		return (error);
1519 
1520 	if (linux_map_sched_prio) {
1521 		switch (policy) {
1522 		case SCHED_OTHER:
1523 			if (sched_param.sched_priority != 0)
1524 				return (EINVAL);
1525 
1526 			sched_param.sched_priority =
1527 			    PRI_MAX_TIMESHARE - PRI_MIN_TIMESHARE;
1528 			break;
1529 		case SCHED_FIFO:
1530 		case SCHED_RR:
1531 			if (sched_param.sched_priority < 1 ||
1532 			    sched_param.sched_priority >= LINUX_MAX_RT_PRIO)
1533 				return (EINVAL);
1534 
1535 			/*
1536 			 * Map [1, LINUX_MAX_RT_PRIO - 1] to
1537 			 * [0, RTP_PRIO_MAX - RTP_PRIO_MIN] (rounding down).
1538 			 */
1539 			sched_param.sched_priority =
1540 			    (sched_param.sched_priority - 1) *
1541 			    (RTP_PRIO_MAX - RTP_PRIO_MIN + 1) /
1542 			    (LINUX_MAX_RT_PRIO - 1);
1543 			break;
1544 		}
1545 	}
1546 
1547 	tdt = linux_tdfind(td, args->pid, -1);
1548 	if (tdt == NULL)
1549 		return (ESRCH);
1550 
1551 	error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1552 	PROC_UNLOCK(tdt->td_proc);
1553 	return (error);
1554 }
1555 
1556 int
1557 linux_sched_getscheduler(struct thread *td,
1558     struct linux_sched_getscheduler_args *args)
1559 {
1560 	struct thread *tdt;
1561 	int error, policy;
1562 
1563 	tdt = linux_tdfind(td, args->pid, -1);
1564 	if (tdt == NULL)
1565 		return (ESRCH);
1566 
1567 	error = kern_sched_getscheduler(td, tdt, &policy);
1568 	PROC_UNLOCK(tdt->td_proc);
1569 
1570 	switch (policy) {
1571 	case SCHED_OTHER:
1572 		td->td_retval[0] = LINUX_SCHED_OTHER;
1573 		break;
1574 	case SCHED_FIFO:
1575 		td->td_retval[0] = LINUX_SCHED_FIFO;
1576 		break;
1577 	case SCHED_RR:
1578 		td->td_retval[0] = LINUX_SCHED_RR;
1579 		break;
1580 	}
1581 	return (error);
1582 }
1583 
1584 int
1585 linux_sched_get_priority_max(struct thread *td,
1586     struct linux_sched_get_priority_max_args *args)
1587 {
1588 	struct sched_get_priority_max_args bsd;
1589 
1590 	if (linux_map_sched_prio) {
1591 		switch (args->policy) {
1592 		case LINUX_SCHED_OTHER:
1593 			td->td_retval[0] = 0;
1594 			return (0);
1595 		case LINUX_SCHED_FIFO:
1596 		case LINUX_SCHED_RR:
1597 			td->td_retval[0] = LINUX_MAX_RT_PRIO - 1;
1598 			return (0);
1599 		default:
1600 			return (EINVAL);
1601 		}
1602 	}
1603 
1604 	switch (args->policy) {
1605 	case LINUX_SCHED_OTHER:
1606 		bsd.policy = SCHED_OTHER;
1607 		break;
1608 	case LINUX_SCHED_FIFO:
1609 		bsd.policy = SCHED_FIFO;
1610 		break;
1611 	case LINUX_SCHED_RR:
1612 		bsd.policy = SCHED_RR;
1613 		break;
1614 	default:
1615 		return (EINVAL);
1616 	}
1617 	return (sys_sched_get_priority_max(td, &bsd));
1618 }
1619 
1620 int
1621 linux_sched_get_priority_min(struct thread *td,
1622     struct linux_sched_get_priority_min_args *args)
1623 {
1624 	struct sched_get_priority_min_args bsd;
1625 
1626 	if (linux_map_sched_prio) {
1627 		switch (args->policy) {
1628 		case LINUX_SCHED_OTHER:
1629 			td->td_retval[0] = 0;
1630 			return (0);
1631 		case LINUX_SCHED_FIFO:
1632 		case LINUX_SCHED_RR:
1633 			td->td_retval[0] = 1;
1634 			return (0);
1635 		default:
1636 			return (EINVAL);
1637 		}
1638 	}
1639 
1640 	switch (args->policy) {
1641 	case LINUX_SCHED_OTHER:
1642 		bsd.policy = SCHED_OTHER;
1643 		break;
1644 	case LINUX_SCHED_FIFO:
1645 		bsd.policy = SCHED_FIFO;
1646 		break;
1647 	case LINUX_SCHED_RR:
1648 		bsd.policy = SCHED_RR;
1649 		break;
1650 	default:
1651 		return (EINVAL);
1652 	}
1653 	return (sys_sched_get_priority_min(td, &bsd));
1654 }
1655 
1656 #define REBOOT_CAD_ON	0x89abcdef
1657 #define REBOOT_CAD_OFF	0
1658 #define REBOOT_HALT	0xcdef0123
1659 #define REBOOT_RESTART	0x01234567
1660 #define REBOOT_RESTART2	0xA1B2C3D4
1661 #define REBOOT_POWEROFF	0x4321FEDC
1662 #define REBOOT_MAGIC1	0xfee1dead
1663 #define REBOOT_MAGIC2	0x28121969
1664 #define REBOOT_MAGIC2A	0x05121996
1665 #define REBOOT_MAGIC2B	0x16041998
1666 
1667 int
1668 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1669 {
1670 	struct reboot_args bsd_args;
1671 
1672 	if (args->magic1 != REBOOT_MAGIC1)
1673 		return (EINVAL);
1674 
1675 	switch (args->magic2) {
1676 	case REBOOT_MAGIC2:
1677 	case REBOOT_MAGIC2A:
1678 	case REBOOT_MAGIC2B:
1679 		break;
1680 	default:
1681 		return (EINVAL);
1682 	}
1683 
1684 	switch (args->cmd) {
1685 	case REBOOT_CAD_ON:
1686 	case REBOOT_CAD_OFF:
1687 		return (priv_check(td, PRIV_REBOOT));
1688 	case REBOOT_HALT:
1689 		bsd_args.opt = RB_HALT;
1690 		break;
1691 	case REBOOT_RESTART:
1692 	case REBOOT_RESTART2:
1693 		bsd_args.opt = 0;
1694 		break;
1695 	case REBOOT_POWEROFF:
1696 		bsd_args.opt = RB_POWEROFF;
1697 		break;
1698 	default:
1699 		return (EINVAL);
1700 	}
1701 	return (sys_reboot(td, &bsd_args));
1702 }
1703 
1704 int
1705 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1706 {
1707 
1708 	td->td_retval[0] = td->td_proc->p_pid;
1709 
1710 	return (0);
1711 }
1712 
1713 int
1714 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1715 {
1716 	struct linux_emuldata *em;
1717 
1718 	em = em_find(td);
1719 	KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1720 
1721 	td->td_retval[0] = em->em_tid;
1722 
1723 	return (0);
1724 }
1725 
1726 int
1727 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1728 {
1729 
1730 	td->td_retval[0] = kern_getppid(td);
1731 	return (0);
1732 }
1733 
1734 int
1735 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1736 {
1737 
1738 	td->td_retval[0] = td->td_ucred->cr_rgid;
1739 	return (0);
1740 }
1741 
1742 int
1743 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1744 {
1745 
1746 	td->td_retval[0] = td->td_ucred->cr_ruid;
1747 	return (0);
1748 }
1749 
1750 int
1751 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1752 {
1753 
1754 	return (kern_getsid(td, args->pid));
1755 }
1756 
1757 int
1758 linux_nosys(struct thread *td, struct nosys_args *ignore)
1759 {
1760 
1761 	return (ENOSYS);
1762 }
1763 
1764 int
1765 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1766 {
1767 	int error;
1768 
1769 	error = kern_getpriority(td, args->which, args->who);
1770 	td->td_retval[0] = 20 - td->td_retval[0];
1771 	return (error);
1772 }
1773 
1774 int
1775 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1776 {
1777 	int name[2];
1778 
1779 	name[0] = CTL_KERN;
1780 	name[1] = KERN_HOSTNAME;
1781 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1782 	    args->len, 0, 0));
1783 }
1784 
1785 int
1786 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1787 {
1788 	int name[2];
1789 
1790 	name[0] = CTL_KERN;
1791 	name[1] = KERN_NISDOMAINNAME;
1792 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1793 	    args->len, 0, 0));
1794 }
1795 
1796 int
1797 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1798 {
1799 
1800 	LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1801 	    args->error_code);
1802 
1803 	/*
1804 	 * XXX: we should send a signal to the parent if
1805 	 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1806 	 * as it doesnt occur often.
1807 	 */
1808 	exit1(td, args->error_code, 0);
1809 		/* NOTREACHED */
1810 }
1811 
1812 #define _LINUX_CAPABILITY_VERSION_1  0x19980330
1813 #define _LINUX_CAPABILITY_VERSION_2  0x20071026
1814 #define _LINUX_CAPABILITY_VERSION_3  0x20080522
1815 
1816 struct l_user_cap_header {
1817 	l_int	version;
1818 	l_int	pid;
1819 };
1820 
1821 struct l_user_cap_data {
1822 	l_int	effective;
1823 	l_int	permitted;
1824 	l_int	inheritable;
1825 };
1826 
1827 int
1828 linux_capget(struct thread *td, struct linux_capget_args *uap)
1829 {
1830 	struct l_user_cap_header luch;
1831 	struct l_user_cap_data lucd[2];
1832 	int error, u32s;
1833 
1834 	if (uap->hdrp == NULL)
1835 		return (EFAULT);
1836 
1837 	error = copyin(uap->hdrp, &luch, sizeof(luch));
1838 	if (error != 0)
1839 		return (error);
1840 
1841 	switch (luch.version) {
1842 	case _LINUX_CAPABILITY_VERSION_1:
1843 		u32s = 1;
1844 		break;
1845 	case _LINUX_CAPABILITY_VERSION_2:
1846 	case _LINUX_CAPABILITY_VERSION_3:
1847 		u32s = 2;
1848 		break;
1849 	default:
1850 		luch.version = _LINUX_CAPABILITY_VERSION_1;
1851 		error = copyout(&luch, uap->hdrp, sizeof(luch));
1852 		if (error)
1853 			return (error);
1854 		return (EINVAL);
1855 	}
1856 
1857 	if (luch.pid)
1858 		return (EPERM);
1859 
1860 	if (uap->datap) {
1861 		/*
1862 		 * The current implementation doesn't support setting
1863 		 * a capability (it's essentially a stub) so indicate
1864 		 * that no capabilities are currently set or available
1865 		 * to request.
1866 		 */
1867 		memset(&lucd, 0, u32s * sizeof(lucd[0]));
1868 		error = copyout(&lucd, uap->datap, u32s * sizeof(lucd[0]));
1869 	}
1870 
1871 	return (error);
1872 }
1873 
1874 int
1875 linux_capset(struct thread *td, struct linux_capset_args *uap)
1876 {
1877 	struct l_user_cap_header luch;
1878 	struct l_user_cap_data lucd[2];
1879 	int error, i, u32s;
1880 
1881 	if (uap->hdrp == NULL || uap->datap == NULL)
1882 		return (EFAULT);
1883 
1884 	error = copyin(uap->hdrp, &luch, sizeof(luch));
1885 	if (error != 0)
1886 		return (error);
1887 
1888 	switch (luch.version) {
1889 	case _LINUX_CAPABILITY_VERSION_1:
1890 		u32s = 1;
1891 		break;
1892 	case _LINUX_CAPABILITY_VERSION_2:
1893 	case _LINUX_CAPABILITY_VERSION_3:
1894 		u32s = 2;
1895 		break;
1896 	default:
1897 		luch.version = _LINUX_CAPABILITY_VERSION_1;
1898 		error = copyout(&luch, uap->hdrp, sizeof(luch));
1899 		if (error)
1900 			return (error);
1901 		return (EINVAL);
1902 	}
1903 
1904 	if (luch.pid)
1905 		return (EPERM);
1906 
1907 	error = copyin(uap->datap, &lucd, u32s * sizeof(lucd[0]));
1908 	if (error != 0)
1909 		return (error);
1910 
1911 	/* We currently don't support setting any capabilities. */
1912 	for (i = 0; i < u32s; i++) {
1913 		if (lucd[i].effective || lucd[i].permitted ||
1914 		    lucd[i].inheritable) {
1915 			linux_msg(td,
1916 			    "capset[%d] effective=0x%x, permitted=0x%x, "
1917 			    "inheritable=0x%x is not implemented", i,
1918 			    (int)lucd[i].effective, (int)lucd[i].permitted,
1919 			    (int)lucd[i].inheritable);
1920 			return (EPERM);
1921 		}
1922 	}
1923 
1924 	return (0);
1925 }
1926 
1927 int
1928 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1929 {
1930 	int error = 0, max_size;
1931 	struct proc *p = td->td_proc;
1932 	char comm[LINUX_MAX_COMM_LEN];
1933 	int pdeath_signal;
1934 
1935 	switch (args->option) {
1936 	case LINUX_PR_SET_PDEATHSIG:
1937 		if (!LINUX_SIG_VALID(args->arg2))
1938 			return (EINVAL);
1939 		pdeath_signal = linux_to_bsd_signal(args->arg2);
1940 		return (kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_CTL,
1941 		    &pdeath_signal));
1942 	case LINUX_PR_GET_PDEATHSIG:
1943 		error = kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_STATUS,
1944 		    &pdeath_signal);
1945 		if (error != 0)
1946 			return (error);
1947 		pdeath_signal = bsd_to_linux_signal(pdeath_signal);
1948 		return (copyout(&pdeath_signal,
1949 		    (void *)(register_t)args->arg2,
1950 		    sizeof(pdeath_signal)));
1951 		break;
1952 	case LINUX_PR_GET_KEEPCAPS:
1953 		/*
1954 		 * Indicate that we always clear the effective and
1955 		 * permitted capability sets when the user id becomes
1956 		 * non-zero (actually the capability sets are simply
1957 		 * always zero in the current implementation).
1958 		 */
1959 		td->td_retval[0] = 0;
1960 		break;
1961 	case LINUX_PR_SET_KEEPCAPS:
1962 		/*
1963 		 * Ignore requests to keep the effective and permitted
1964 		 * capability sets when the user id becomes non-zero.
1965 		 */
1966 		break;
1967 	case LINUX_PR_SET_NAME:
1968 		/*
1969 		 * To be on the safe side we need to make sure to not
1970 		 * overflow the size a Linux program expects. We already
1971 		 * do this here in the copyin, so that we don't need to
1972 		 * check on copyout.
1973 		 */
1974 		max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1975 		error = copyinstr((void *)(register_t)args->arg2, comm,
1976 		    max_size, NULL);
1977 
1978 		/* Linux silently truncates the name if it is too long. */
1979 		if (error == ENAMETOOLONG) {
1980 			/*
1981 			 * XXX: copyinstr() isn't documented to populate the
1982 			 * array completely, so do a copyin() to be on the
1983 			 * safe side. This should be changed in case
1984 			 * copyinstr() is changed to guarantee this.
1985 			 */
1986 			error = copyin((void *)(register_t)args->arg2, comm,
1987 			    max_size - 1);
1988 			comm[max_size - 1] = '\0';
1989 		}
1990 		if (error)
1991 			return (error);
1992 
1993 		PROC_LOCK(p);
1994 		strlcpy(p->p_comm, comm, sizeof(p->p_comm));
1995 		PROC_UNLOCK(p);
1996 		break;
1997 	case LINUX_PR_GET_NAME:
1998 		PROC_LOCK(p);
1999 		strlcpy(comm, p->p_comm, sizeof(comm));
2000 		PROC_UNLOCK(p);
2001 		error = copyout(comm, (void *)(register_t)args->arg2,
2002 		    strlen(comm) + 1);
2003 		break;
2004 	case LINUX_PR_GET_SECCOMP:
2005 	case LINUX_PR_SET_SECCOMP:
2006 		/*
2007 		 * Same as returned by Linux without CONFIG_SECCOMP enabled.
2008 		 */
2009 		error = EINVAL;
2010 		break;
2011 	default:
2012 		linux_msg(td, "unsupported prctl option %d", args->option);
2013 		error = EINVAL;
2014 		break;
2015 	}
2016 
2017 	return (error);
2018 }
2019 
2020 int
2021 linux_sched_setparam(struct thread *td,
2022     struct linux_sched_setparam_args *uap)
2023 {
2024 	struct sched_param sched_param;
2025 	struct thread *tdt;
2026 	int error, policy;
2027 
2028 	error = copyin(uap->param, &sched_param, sizeof(sched_param));
2029 	if (error)
2030 		return (error);
2031 
2032 	tdt = linux_tdfind(td, uap->pid, -1);
2033 	if (tdt == NULL)
2034 		return (ESRCH);
2035 
2036 	if (linux_map_sched_prio) {
2037 		error = kern_sched_getscheduler(td, tdt, &policy);
2038 		if (error)
2039 			goto out;
2040 
2041 		switch (policy) {
2042 		case SCHED_OTHER:
2043 			if (sched_param.sched_priority != 0) {
2044 				error = EINVAL;
2045 				goto out;
2046 			}
2047 			sched_param.sched_priority =
2048 			    PRI_MAX_TIMESHARE - PRI_MIN_TIMESHARE;
2049 			break;
2050 		case SCHED_FIFO:
2051 		case SCHED_RR:
2052 			if (sched_param.sched_priority < 1 ||
2053 			    sched_param.sched_priority >= LINUX_MAX_RT_PRIO) {
2054 				error = EINVAL;
2055 				goto out;
2056 			}
2057 			/*
2058 			 * Map [1, LINUX_MAX_RT_PRIO - 1] to
2059 			 * [0, RTP_PRIO_MAX - RTP_PRIO_MIN] (rounding down).
2060 			 */
2061 			sched_param.sched_priority =
2062 			    (sched_param.sched_priority - 1) *
2063 			    (RTP_PRIO_MAX - RTP_PRIO_MIN + 1) /
2064 			    (LINUX_MAX_RT_PRIO - 1);
2065 			break;
2066 		}
2067 	}
2068 
2069 	error = kern_sched_setparam(td, tdt, &sched_param);
2070 out:	PROC_UNLOCK(tdt->td_proc);
2071 	return (error);
2072 }
2073 
2074 int
2075 linux_sched_getparam(struct thread *td,
2076     struct linux_sched_getparam_args *uap)
2077 {
2078 	struct sched_param sched_param;
2079 	struct thread *tdt;
2080 	int error, policy;
2081 
2082 	tdt = linux_tdfind(td, uap->pid, -1);
2083 	if (tdt == NULL)
2084 		return (ESRCH);
2085 
2086 	error = kern_sched_getparam(td, tdt, &sched_param);
2087 	if (error) {
2088 		PROC_UNLOCK(tdt->td_proc);
2089 		return (error);
2090 	}
2091 
2092 	if (linux_map_sched_prio) {
2093 		error = kern_sched_getscheduler(td, tdt, &policy);
2094 		PROC_UNLOCK(tdt->td_proc);
2095 		if (error)
2096 			return (error);
2097 
2098 		switch (policy) {
2099 		case SCHED_OTHER:
2100 			sched_param.sched_priority = 0;
2101 			break;
2102 		case SCHED_FIFO:
2103 		case SCHED_RR:
2104 			/*
2105 			 * Map [0, RTP_PRIO_MAX - RTP_PRIO_MIN] to
2106 			 * [1, LINUX_MAX_RT_PRIO - 1] (rounding up).
2107 			 */
2108 			sched_param.sched_priority =
2109 			    (sched_param.sched_priority *
2110 			    (LINUX_MAX_RT_PRIO - 1) +
2111 			    (RTP_PRIO_MAX - RTP_PRIO_MIN - 1)) /
2112 			    (RTP_PRIO_MAX - RTP_PRIO_MIN) + 1;
2113 			break;
2114 		}
2115 	} else
2116 		PROC_UNLOCK(tdt->td_proc);
2117 
2118 	error = copyout(&sched_param, uap->param, sizeof(sched_param));
2119 	return (error);
2120 }
2121 
2122 /*
2123  * Get affinity of a process.
2124  */
2125 int
2126 linux_sched_getaffinity(struct thread *td,
2127     struct linux_sched_getaffinity_args *args)
2128 {
2129 	int error;
2130 	struct thread *tdt;
2131 
2132 	if (args->len < sizeof(cpuset_t))
2133 		return (EINVAL);
2134 
2135 	tdt = linux_tdfind(td, args->pid, -1);
2136 	if (tdt == NULL)
2137 		return (ESRCH);
2138 
2139 	PROC_UNLOCK(tdt->td_proc);
2140 
2141 	error = kern_cpuset_getaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2142 	    tdt->td_tid, sizeof(cpuset_t), (cpuset_t *)args->user_mask_ptr);
2143 	if (error == 0)
2144 		td->td_retval[0] = sizeof(cpuset_t);
2145 
2146 	return (error);
2147 }
2148 
2149 /*
2150  *  Set affinity of a process.
2151  */
2152 int
2153 linux_sched_setaffinity(struct thread *td,
2154     struct linux_sched_setaffinity_args *args)
2155 {
2156 	struct thread *tdt;
2157 
2158 	if (args->len < sizeof(cpuset_t))
2159 		return (EINVAL);
2160 
2161 	tdt = linux_tdfind(td, args->pid, -1);
2162 	if (tdt == NULL)
2163 		return (ESRCH);
2164 
2165 	PROC_UNLOCK(tdt->td_proc);
2166 
2167 	return (kern_cpuset_setaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2168 	    tdt->td_tid, sizeof(cpuset_t), (cpuset_t *) args->user_mask_ptr));
2169 }
2170 
2171 struct linux_rlimit64 {
2172 	uint64_t	rlim_cur;
2173 	uint64_t	rlim_max;
2174 };
2175 
2176 int
2177 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
2178 {
2179 	struct rlimit rlim, nrlim;
2180 	struct linux_rlimit64 lrlim;
2181 	struct proc *p;
2182 	u_int which;
2183 	int flags;
2184 	int error;
2185 
2186 	if (args->new == NULL && args->old != NULL) {
2187 		if (linux_get_dummy_limit(args->resource, &rlim)) {
2188 			lrlim.rlim_cur = rlim.rlim_cur;
2189 			lrlim.rlim_max = rlim.rlim_max;
2190 			return (copyout(&lrlim, args->old, sizeof(lrlim)));
2191 		}
2192 	}
2193 
2194 	if (args->resource >= LINUX_RLIM_NLIMITS)
2195 		return (EINVAL);
2196 
2197 	which = linux_to_bsd_resource[args->resource];
2198 	if (which == -1)
2199 		return (EINVAL);
2200 
2201 	if (args->new != NULL) {
2202 		/*
2203 		 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
2204 		 * rlim is unsigned 64-bit. FreeBSD treats negative limits
2205 		 * as INFINITY so we do not need a conversion even.
2206 		 */
2207 		error = copyin(args->new, &nrlim, sizeof(nrlim));
2208 		if (error != 0)
2209 			return (error);
2210 	}
2211 
2212 	flags = PGET_HOLD | PGET_NOTWEXIT;
2213 	if (args->new != NULL)
2214 		flags |= PGET_CANDEBUG;
2215 	else
2216 		flags |= PGET_CANSEE;
2217 	if (args->pid == 0) {
2218 		p = td->td_proc;
2219 		PHOLD(p);
2220 	} else {
2221 		error = pget(args->pid, flags, &p);
2222 		if (error != 0)
2223 			return (error);
2224 	}
2225 	if (args->old != NULL) {
2226 		PROC_LOCK(p);
2227 		lim_rlimit_proc(p, which, &rlim);
2228 		PROC_UNLOCK(p);
2229 		if (rlim.rlim_cur == RLIM_INFINITY)
2230 			lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2231 		else
2232 			lrlim.rlim_cur = rlim.rlim_cur;
2233 		if (rlim.rlim_max == RLIM_INFINITY)
2234 			lrlim.rlim_max = LINUX_RLIM_INFINITY;
2235 		else
2236 			lrlim.rlim_max = rlim.rlim_max;
2237 		error = copyout(&lrlim, args->old, sizeof(lrlim));
2238 		if (error != 0)
2239 			goto out;
2240 	}
2241 
2242 	if (args->new != NULL)
2243 		error = kern_proc_setrlimit(td, p, which, &nrlim);
2244 
2245  out:
2246 	PRELE(p);
2247 	return (error);
2248 }
2249 
2250 int
2251 linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2252 {
2253 	struct timeval utv, tv0, tv1, *tvp;
2254 	struct l_pselect6arg lpse6;
2255 	struct l_timespec lts;
2256 	struct timespec uts;
2257 	l_sigset_t l_ss;
2258 	sigset_t *ssp;
2259 	sigset_t ss;
2260 	int error;
2261 
2262 	ssp = NULL;
2263 	if (args->sig != NULL) {
2264 		error = copyin(args->sig, &lpse6, sizeof(lpse6));
2265 		if (error != 0)
2266 			return (error);
2267 		if (lpse6.ss_len != sizeof(l_ss))
2268 			return (EINVAL);
2269 		if (lpse6.ss != 0) {
2270 			error = copyin(PTRIN(lpse6.ss), &l_ss,
2271 			    sizeof(l_ss));
2272 			if (error != 0)
2273 				return (error);
2274 			linux_to_bsd_sigset(&l_ss, &ss);
2275 			ssp = &ss;
2276 		}
2277 	}
2278 
2279 	/*
2280 	 * Currently glibc changes nanosecond number to microsecond.
2281 	 * This mean losing precision but for now it is hardly seen.
2282 	 */
2283 	if (args->tsp != NULL) {
2284 		error = copyin(args->tsp, &lts, sizeof(lts));
2285 		if (error != 0)
2286 			return (error);
2287 		error = linux_to_native_timespec(&uts, &lts);
2288 		if (error != 0)
2289 			return (error);
2290 
2291 		TIMESPEC_TO_TIMEVAL(&utv, &uts);
2292 		if (itimerfix(&utv))
2293 			return (EINVAL);
2294 
2295 		microtime(&tv0);
2296 		tvp = &utv;
2297 	} else
2298 		tvp = NULL;
2299 
2300 	error = kern_pselect(td, args->nfds, args->readfds, args->writefds,
2301 	    args->exceptfds, tvp, ssp, LINUX_NFDBITS);
2302 
2303 	if (error == 0 && args->tsp != NULL) {
2304 		if (td->td_retval[0] != 0) {
2305 			/*
2306 			 * Compute how much time was left of the timeout,
2307 			 * by subtracting the current time and the time
2308 			 * before we started the call, and subtracting
2309 			 * that result from the user-supplied value.
2310 			 */
2311 
2312 			microtime(&tv1);
2313 			timevalsub(&tv1, &tv0);
2314 			timevalsub(&utv, &tv1);
2315 			if (utv.tv_sec < 0)
2316 				timevalclear(&utv);
2317 		} else
2318 			timevalclear(&utv);
2319 
2320 		TIMEVAL_TO_TIMESPEC(&utv, &uts);
2321 
2322 		error = native_to_linux_timespec(&lts, &uts);
2323 		if (error == 0)
2324 			error = copyout(&lts, args->tsp, sizeof(lts));
2325 	}
2326 
2327 	return (error);
2328 }
2329 
2330 int
2331 linux_ppoll(struct thread *td, struct linux_ppoll_args *args)
2332 {
2333 	struct timespec ts0, ts1;
2334 	struct l_timespec lts;
2335 	struct timespec uts, *tsp;
2336 	l_sigset_t l_ss;
2337 	sigset_t *ssp;
2338 	sigset_t ss;
2339 	int error;
2340 
2341 	if (args->sset != NULL) {
2342 		if (args->ssize != sizeof(l_ss))
2343 			return (EINVAL);
2344 		error = copyin(args->sset, &l_ss, sizeof(l_ss));
2345 		if (error)
2346 			return (error);
2347 		linux_to_bsd_sigset(&l_ss, &ss);
2348 		ssp = &ss;
2349 	} else
2350 		ssp = NULL;
2351 	if (args->tsp != NULL) {
2352 		error = copyin(args->tsp, &lts, sizeof(lts));
2353 		if (error)
2354 			return (error);
2355 		error = linux_to_native_timespec(&uts, &lts);
2356 		if (error != 0)
2357 			return (error);
2358 
2359 		nanotime(&ts0);
2360 		tsp = &uts;
2361 	} else
2362 		tsp = NULL;
2363 
2364 	error = kern_poll(td, args->fds, args->nfds, tsp, ssp);
2365 
2366 	if (error == 0 && args->tsp != NULL) {
2367 		if (td->td_retval[0]) {
2368 			nanotime(&ts1);
2369 			timespecsub(&ts1, &ts0, &ts1);
2370 			timespecsub(&uts, &ts1, &uts);
2371 			if (uts.tv_sec < 0)
2372 				timespecclear(&uts);
2373 		} else
2374 			timespecclear(&uts);
2375 
2376 		error = native_to_linux_timespec(&lts, &uts);
2377 		if (error == 0)
2378 			error = copyout(&lts, args->tsp, sizeof(lts));
2379 	}
2380 
2381 	return (error);
2382 }
2383 
2384 int
2385 linux_sched_rr_get_interval(struct thread *td,
2386     struct linux_sched_rr_get_interval_args *uap)
2387 {
2388 	struct timespec ts;
2389 	struct l_timespec lts;
2390 	struct thread *tdt;
2391 	int error;
2392 
2393 	/*
2394 	 * According to man in case the invalid pid specified
2395 	 * EINVAL should be returned.
2396 	 */
2397 	if (uap->pid < 0)
2398 		return (EINVAL);
2399 
2400 	tdt = linux_tdfind(td, uap->pid, -1);
2401 	if (tdt == NULL)
2402 		return (ESRCH);
2403 
2404 	error = kern_sched_rr_get_interval_td(td, tdt, &ts);
2405 	PROC_UNLOCK(tdt->td_proc);
2406 	if (error != 0)
2407 		return (error);
2408 	error = native_to_linux_timespec(&lts, &ts);
2409 	if (error != 0)
2410 		return (error);
2411 	return (copyout(&lts, uap->interval, sizeof(lts)));
2412 }
2413 
2414 /*
2415  * In case when the Linux thread is the initial thread in
2416  * the thread group thread id is equal to the process id.
2417  * Glibc depends on this magic (assert in pthread_getattr_np.c).
2418  */
2419 struct thread *
2420 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid)
2421 {
2422 	struct linux_emuldata *em;
2423 	struct thread *tdt;
2424 	struct proc *p;
2425 
2426 	tdt = NULL;
2427 	if (tid == 0 || tid == td->td_tid) {
2428 		tdt = td;
2429 		PROC_LOCK(tdt->td_proc);
2430 	} else if (tid > PID_MAX)
2431 		tdt = tdfind(tid, pid);
2432 	else {
2433 		/*
2434 		 * Initial thread where the tid equal to the pid.
2435 		 */
2436 		p = pfind(tid);
2437 		if (p != NULL) {
2438 			if (SV_PROC_ABI(p) != SV_ABI_LINUX) {
2439 				/*
2440 				 * p is not a Linuxulator process.
2441 				 */
2442 				PROC_UNLOCK(p);
2443 				return (NULL);
2444 			}
2445 			FOREACH_THREAD_IN_PROC(p, tdt) {
2446 				em = em_find(tdt);
2447 				if (tid == em->em_tid)
2448 					return (tdt);
2449 			}
2450 			PROC_UNLOCK(p);
2451 		}
2452 		return (NULL);
2453 	}
2454 
2455 	return (tdt);
2456 }
2457 
2458 void
2459 linux_to_bsd_waitopts(int options, int *bsdopts)
2460 {
2461 
2462 	if (options & LINUX_WNOHANG)
2463 		*bsdopts |= WNOHANG;
2464 	if (options & LINUX_WUNTRACED)
2465 		*bsdopts |= WUNTRACED;
2466 	if (options & LINUX_WEXITED)
2467 		*bsdopts |= WEXITED;
2468 	if (options & LINUX_WCONTINUED)
2469 		*bsdopts |= WCONTINUED;
2470 	if (options & LINUX_WNOWAIT)
2471 		*bsdopts |= WNOWAIT;
2472 
2473 	if (options & __WCLONE)
2474 		*bsdopts |= WLINUXCLONE;
2475 }
2476 
2477 int
2478 linux_getrandom(struct thread *td, struct linux_getrandom_args *args)
2479 {
2480 	struct uio uio;
2481 	struct iovec iov;
2482 	int error;
2483 
2484 	if (args->flags & ~(LINUX_GRND_NONBLOCK|LINUX_GRND_RANDOM))
2485 		return (EINVAL);
2486 	if (args->count > INT_MAX)
2487 		args->count = INT_MAX;
2488 
2489 	iov.iov_base = args->buf;
2490 	iov.iov_len = args->count;
2491 
2492 	uio.uio_iov = &iov;
2493 	uio.uio_iovcnt = 1;
2494 	uio.uio_resid = iov.iov_len;
2495 	uio.uio_segflg = UIO_USERSPACE;
2496 	uio.uio_rw = UIO_READ;
2497 	uio.uio_td = td;
2498 
2499 	error = read_random_uio(&uio, args->flags & LINUX_GRND_NONBLOCK);
2500 	if (error == 0)
2501 		td->td_retval[0] = args->count - uio.uio_resid;
2502 	return (error);
2503 }
2504 
2505 int
2506 linux_mincore(struct thread *td, struct linux_mincore_args *args)
2507 {
2508 
2509 	/* Needs to be page-aligned */
2510 	if (args->start & PAGE_MASK)
2511 		return (EINVAL);
2512 	return (kern_mincore(td, args->start, args->len, args->vec));
2513 }
2514 
2515 #define	SYSLOG_TAG	"<6>"
2516 
2517 int
2518 linux_syslog(struct thread *td, struct linux_syslog_args *args)
2519 {
2520 	char buf[128], *src, *dst;
2521 	u_int seq;
2522 	int buflen, error;
2523 
2524 	if (args->type != LINUX_SYSLOG_ACTION_READ_ALL) {
2525 		linux_msg(td, "syslog unsupported type 0x%x", args->type);
2526 		return (EINVAL);
2527 	}
2528 
2529 	if (args->len < 6) {
2530 		td->td_retval[0] = 0;
2531 		return (0);
2532 	}
2533 
2534 	error = priv_check(td, PRIV_MSGBUF);
2535 	if (error)
2536 		return (error);
2537 
2538 	mtx_lock(&msgbuf_lock);
2539 	msgbuf_peekbytes(msgbufp, NULL, 0, &seq);
2540 	mtx_unlock(&msgbuf_lock);
2541 
2542 	dst = args->buf;
2543 	error = copyout(&SYSLOG_TAG, dst, sizeof(SYSLOG_TAG));
2544 	/* The -1 is to skip the trailing '\0'. */
2545 	dst += sizeof(SYSLOG_TAG) - 1;
2546 
2547 	while (error == 0) {
2548 		mtx_lock(&msgbuf_lock);
2549 		buflen = msgbuf_peekbytes(msgbufp, buf, sizeof(buf), &seq);
2550 		mtx_unlock(&msgbuf_lock);
2551 
2552 		if (buflen == 0)
2553 			break;
2554 
2555 		for (src = buf; src < buf + buflen && error == 0; src++) {
2556 			if (*src == '\0')
2557 				continue;
2558 
2559 			if (dst >= args->buf + args->len)
2560 				goto out;
2561 
2562 			error = copyout(src, dst, 1);
2563 			dst++;
2564 
2565 			if (*src == '\n' && *(src + 1) != '<' &&
2566 			    dst + sizeof(SYSLOG_TAG) < args->buf + args->len) {
2567 				error = copyout(&SYSLOG_TAG,
2568 				    dst, sizeof(SYSLOG_TAG));
2569 				dst += sizeof(SYSLOG_TAG) - 1;
2570 			}
2571 		}
2572 	}
2573 out:
2574 	td->td_retval[0] = dst - args->buf;
2575 	return (error);
2576 }
2577 
2578 int
2579 linux_getcpu(struct thread *td, struct linux_getcpu_args *args)
2580 {
2581 	int cpu, error, node;
2582 
2583 	cpu = td->td_oncpu; /* Make sure it doesn't change during copyout(9) */
2584 	error = 0;
2585 	node = cpuid_to_pcpu[cpu]->pc_domain;
2586 
2587 	if (args->cpu != NULL)
2588 		error = copyout(&cpu, args->cpu, sizeof(l_int));
2589 	if (args->node != NULL)
2590 		error = copyout(&node, args->node, sizeof(l_int));
2591 	return (error);
2592 }
2593