xref: /freebsd/sys/compat/linux/linux_misc.c (revision b685454a1107a512e105ceb09b5ff04dadc04d1a)
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/mutex.h>
51 #include <sys/namei.h>
52 #include <sys/priv.h>
53 #include <sys/proc.h>
54 #include <sys/procctl.h>
55 #include <sys/reboot.h>
56 #include <sys/racct.h>
57 #include <sys/random.h>
58 #include <sys/resourcevar.h>
59 #include <sys/sched.h>
60 #include <sys/sdt.h>
61 #include <sys/signalvar.h>
62 #include <sys/stat.h>
63 #include <sys/syscallsubr.h>
64 #include <sys/sysctl.h>
65 #include <sys/sysproto.h>
66 #include <sys/systm.h>
67 #include <sys/time.h>
68 #include <sys/vmmeter.h>
69 #include <sys/vnode.h>
70 #include <sys/wait.h>
71 #include <sys/cpuset.h>
72 #include <sys/uio.h>
73 
74 #include <security/mac/mac_framework.h>
75 
76 #include <vm/vm.h>
77 #include <vm/pmap.h>
78 #include <vm/vm_kern.h>
79 #include <vm/vm_map.h>
80 #include <vm/vm_extern.h>
81 #include <vm/vm_object.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		totalbig;
136 	l_ulong		freebig;
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 
149 int
150 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
151 {
152 	struct l_sysinfo sysinfo;
153 	vm_object_t object;
154 	int i, j;
155 	struct timespec ts;
156 
157 	bzero(&sysinfo, sizeof(sysinfo));
158 	getnanouptime(&ts);
159 	if (ts.tv_nsec != 0)
160 		ts.tv_sec++;
161 	sysinfo.uptime = ts.tv_sec;
162 
163 	/* Use the information from the mib to get our load averages */
164 	for (i = 0; i < 3; i++)
165 		sysinfo.loads[i] = averunnable.ldavg[i] *
166 		    LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
167 
168 	sysinfo.totalram = physmem * PAGE_SIZE;
169 	sysinfo.freeram = sysinfo.totalram - vm_wire_count() * PAGE_SIZE;
170 
171 	sysinfo.sharedram = 0;
172 	mtx_lock(&vm_object_list_mtx);
173 	TAILQ_FOREACH(object, &vm_object_list, object_list)
174 		if (object->shadow_count > 1)
175 			sysinfo.sharedram += object->resident_page_count;
176 	mtx_unlock(&vm_object_list_mtx);
177 
178 	sysinfo.sharedram *= PAGE_SIZE;
179 	sysinfo.bufferram = 0;
180 
181 	swap_pager_status(&i, &j);
182 	sysinfo.totalswap = i * PAGE_SIZE;
183 	sysinfo.freeswap = (i - j) * PAGE_SIZE;
184 
185 	sysinfo.procs = nprocs;
186 
187 	/* The following are only present in newer Linux kernels. */
188 	sysinfo.totalbig = 0;
189 	sysinfo.freebig = 0;
190 	sysinfo.mem_unit = 1;
191 
192 	return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
193 }
194 
195 #ifdef LINUX_LEGACY_SYSCALLS
196 int
197 linux_alarm(struct thread *td, struct linux_alarm_args *args)
198 {
199 	struct itimerval it, old_it;
200 	u_int secs;
201 	int error;
202 
203 	secs = args->secs;
204 	/*
205 	 * Linux alarm() is always successful. Limit secs to INT32_MAX / 2
206 	 * to match kern_setitimer()'s limit to avoid error from it.
207 	 *
208 	 * XXX. Linux limit secs to INT_MAX on 32 and does not limit on 64-bit
209 	 * platforms.
210 	 */
211 	if (secs > INT32_MAX / 2)
212 		secs = INT32_MAX / 2;
213 
214 	it.it_value.tv_sec = secs;
215 	it.it_value.tv_usec = 0;
216 	timevalclear(&it.it_interval);
217 	error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
218 	KASSERT(error == 0, ("kern_setitimer returns %d", error));
219 
220 	if ((old_it.it_value.tv_sec == 0 && old_it.it_value.tv_usec > 0) ||
221 	    old_it.it_value.tv_usec >= 500000)
222 		old_it.it_value.tv_sec++;
223 	td->td_retval[0] = old_it.it_value.tv_sec;
224 	return (0);
225 }
226 #endif
227 
228 int
229 linux_brk(struct thread *td, struct linux_brk_args *args)
230 {
231 	struct vmspace *vm = td->td_proc->p_vmspace;
232 	uintptr_t new, old;
233 
234 	old = (uintptr_t)vm->vm_daddr + ctob(vm->vm_dsize);
235 	new = (uintptr_t)args->dsend;
236 	if ((caddr_t)new > vm->vm_daddr && !kern_break(td, &new))
237 		td->td_retval[0] = (register_t)new;
238 	else
239 		td->td_retval[0] = (register_t)old;
240 
241 	return (0);
242 }
243 
244 #if defined(__i386__)
245 /* XXX: what about amd64/linux32? */
246 
247 int
248 linux_uselib(struct thread *td, struct linux_uselib_args *args)
249 {
250 	struct nameidata ni;
251 	struct vnode *vp;
252 	struct exec *a_out;
253 	vm_map_t map;
254 	vm_map_entry_t entry;
255 	struct vattr attr;
256 	vm_offset_t vmaddr;
257 	unsigned long file_offset;
258 	unsigned long bss_size;
259 	char *library;
260 	ssize_t aresid;
261 	int error;
262 	bool locked, opened, textset;
263 
264 	LCONVPATHEXIST(td, args->library, &library);
265 
266 	a_out = NULL;
267 	vp = NULL;
268 	locked = false;
269 	textset = false;
270 	opened = false;
271 
272 	NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
273 	    UIO_SYSSPACE, library, td);
274 	error = namei(&ni);
275 	LFREEPATH(library);
276 	if (error)
277 		goto cleanup;
278 
279 	vp = ni.ni_vp;
280 	NDFREE(&ni, NDF_ONLY_PNBUF);
281 
282 	/*
283 	 * From here on down, we have a locked vnode that must be unlocked.
284 	 * XXX: The code below largely duplicates exec_check_permissions().
285 	 */
286 	locked = true;
287 
288 	/* Executable? */
289 	error = VOP_GETATTR(vp, &attr, td->td_ucred);
290 	if (error)
291 		goto cleanup;
292 
293 	if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
294 	    ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
295 		/* EACCESS is what exec(2) returns. */
296 		error = ENOEXEC;
297 		goto cleanup;
298 	}
299 
300 	/* Sensible size? */
301 	if (attr.va_size == 0) {
302 		error = ENOEXEC;
303 		goto cleanup;
304 	}
305 
306 	/* Can we access it? */
307 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
308 	if (error)
309 		goto cleanup;
310 
311 	/*
312 	 * XXX: This should use vn_open() so that it is properly authorized,
313 	 * and to reduce code redundancy all over the place here.
314 	 * XXX: Not really, it duplicates far more of exec_check_permissions()
315 	 * than vn_open().
316 	 */
317 #ifdef MAC
318 	error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
319 	if (error)
320 		goto cleanup;
321 #endif
322 	error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
323 	if (error)
324 		goto cleanup;
325 	opened = true;
326 
327 	/* Pull in executable header into exec_map */
328 	error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
329 	    VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
330 	if (error)
331 		goto cleanup;
332 
333 	/* Is it a Linux binary ? */
334 	if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
335 		error = ENOEXEC;
336 		goto cleanup;
337 	}
338 
339 	/*
340 	 * While we are here, we should REALLY do some more checks
341 	 */
342 
343 	/* Set file/virtual offset based on a.out variant. */
344 	switch ((int)(a_out->a_magic & 0xffff)) {
345 	case 0413:			/* ZMAGIC */
346 		file_offset = 1024;
347 		break;
348 	case 0314:			/* QMAGIC */
349 		file_offset = 0;
350 		break;
351 	default:
352 		error = ENOEXEC;
353 		goto cleanup;
354 	}
355 
356 	bss_size = round_page(a_out->a_bss);
357 
358 	/* Check various fields in header for validity/bounds. */
359 	if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
360 		error = ENOEXEC;
361 		goto cleanup;
362 	}
363 
364 	/* text + data can't exceed file size */
365 	if (a_out->a_data + a_out->a_text > attr.va_size) {
366 		error = EFAULT;
367 		goto cleanup;
368 	}
369 
370 	/*
371 	 * text/data/bss must not exceed limits
372 	 * XXX - this is not complete. it should check current usage PLUS
373 	 * the resources needed by this library.
374 	 */
375 	PROC_LOCK(td->td_proc);
376 	if (a_out->a_text > maxtsiz ||
377 	    a_out->a_data + bss_size > lim_cur_proc(td->td_proc, RLIMIT_DATA) ||
378 	    racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
379 	    bss_size) != 0) {
380 		PROC_UNLOCK(td->td_proc);
381 		error = ENOMEM;
382 		goto cleanup;
383 	}
384 	PROC_UNLOCK(td->td_proc);
385 
386 	/*
387 	 * Prevent more writers.
388 	 */
389 	error = VOP_SET_TEXT(vp);
390 	if (error != 0)
391 		goto cleanup;
392 	textset = true;
393 
394 	/*
395 	 * Lock no longer needed
396 	 */
397 	locked = false;
398 	VOP_UNLOCK(vp, 0);
399 
400 	/*
401 	 * Check if file_offset page aligned. Currently we cannot handle
402 	 * misalinged file offsets, and so we read in the entire image
403 	 * (what a waste).
404 	 */
405 	if (file_offset & PAGE_MASK) {
406 		/* Map text+data read/write/execute */
407 
408 		/* a_entry is the load address and is page aligned */
409 		vmaddr = trunc_page(a_out->a_entry);
410 
411 		/* get anon user mapping, read+write+execute */
412 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
413 		    &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
414 		    VM_PROT_ALL, VM_PROT_ALL, 0);
415 		if (error)
416 			goto cleanup;
417 
418 		error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
419 		    a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
420 		    td->td_ucred, NOCRED, &aresid, td);
421 		if (error != 0)
422 			goto cleanup;
423 		if (aresid != 0) {
424 			error = ENOEXEC;
425 			goto cleanup;
426 		}
427 	} else {
428 		/*
429 		 * for QMAGIC, a_entry is 20 bytes beyond the load address
430 		 * to skip the executable header
431 		 */
432 		vmaddr = trunc_page(a_out->a_entry);
433 
434 		/*
435 		 * Map it all into the process's space as a single
436 		 * copy-on-write "data" segment.
437 		 */
438 		map = &td->td_proc->p_vmspace->vm_map;
439 		error = vm_mmap(map, &vmaddr,
440 		    a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
441 		    MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
442 		if (error)
443 			goto cleanup;
444 		vm_map_lock(map);
445 		if (!vm_map_lookup_entry(map, vmaddr, &entry)) {
446 			vm_map_unlock(map);
447 			error = EDOOFUS;
448 			goto cleanup;
449 		}
450 		entry->eflags |= MAP_ENTRY_VN_EXEC;
451 		vm_map_unlock(map);
452 		textset = false;
453 	}
454 
455 	if (bss_size != 0) {
456 		/* Calculate BSS start address */
457 		vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
458 		    a_out->a_data;
459 
460 		/* allocate some 'anon' space */
461 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
462 		    &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
463 		    VM_PROT_ALL, 0);
464 		if (error)
465 			goto cleanup;
466 	}
467 
468 cleanup:
469 	if (opened) {
470 		if (locked)
471 			VOP_UNLOCK(vp, 0);
472 		locked = false;
473 		VOP_CLOSE(vp, FREAD, td->td_ucred, td);
474 	}
475 	if (textset)
476 		VOP_UNSET_TEXT_CHECKED(vp);
477 	if (locked)
478 		VOP_UNLOCK(vp, 0);
479 
480 	/* Release the temporary mapping. */
481 	if (a_out)
482 		kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
483 
484 	return (error);
485 }
486 
487 #endif	/* __i386__ */
488 
489 #ifdef LINUX_LEGACY_SYSCALLS
490 int
491 linux_select(struct thread *td, struct linux_select_args *args)
492 {
493 	l_timeval ltv;
494 	struct timeval tv0, tv1, utv, *tvp;
495 	int error;
496 
497 	/*
498 	 * Store current time for computation of the amount of
499 	 * time left.
500 	 */
501 	if (args->timeout) {
502 		if ((error = copyin(args->timeout, &ltv, sizeof(ltv))))
503 			goto select_out;
504 		utv.tv_sec = ltv.tv_sec;
505 		utv.tv_usec = ltv.tv_usec;
506 
507 		if (itimerfix(&utv)) {
508 			/*
509 			 * The timeval was invalid.  Convert it to something
510 			 * valid that will act as it does under Linux.
511 			 */
512 			utv.tv_sec += utv.tv_usec / 1000000;
513 			utv.tv_usec %= 1000000;
514 			if (utv.tv_usec < 0) {
515 				utv.tv_sec -= 1;
516 				utv.tv_usec += 1000000;
517 			}
518 			if (utv.tv_sec < 0)
519 				timevalclear(&utv);
520 		}
521 		microtime(&tv0);
522 		tvp = &utv;
523 	} else
524 		tvp = NULL;
525 
526 	error = kern_select(td, args->nfds, args->readfds, args->writefds,
527 	    args->exceptfds, tvp, LINUX_NFDBITS);
528 	if (error)
529 		goto select_out;
530 
531 	if (args->timeout) {
532 		if (td->td_retval[0]) {
533 			/*
534 			 * Compute how much time was left of the timeout,
535 			 * by subtracting the current time and the time
536 			 * before we started the call, and subtracting
537 			 * that result from the user-supplied value.
538 			 */
539 			microtime(&tv1);
540 			timevalsub(&tv1, &tv0);
541 			timevalsub(&utv, &tv1);
542 			if (utv.tv_sec < 0)
543 				timevalclear(&utv);
544 		} else
545 			timevalclear(&utv);
546 		ltv.tv_sec = utv.tv_sec;
547 		ltv.tv_usec = utv.tv_usec;
548 		if ((error = copyout(&ltv, args->timeout, sizeof(ltv))))
549 			goto select_out;
550 	}
551 
552 select_out:
553 	return (error);
554 }
555 #endif
556 
557 int
558 linux_mremap(struct thread *td, struct linux_mremap_args *args)
559 {
560 	uintptr_t addr;
561 	size_t len;
562 	int error = 0;
563 
564 	if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
565 		td->td_retval[0] = 0;
566 		return (EINVAL);
567 	}
568 
569 	/*
570 	 * Check for the page alignment.
571 	 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
572 	 */
573 	if (args->addr & PAGE_MASK) {
574 		td->td_retval[0] = 0;
575 		return (EINVAL);
576 	}
577 
578 	args->new_len = round_page(args->new_len);
579 	args->old_len = round_page(args->old_len);
580 
581 	if (args->new_len > args->old_len) {
582 		td->td_retval[0] = 0;
583 		return (ENOMEM);
584 	}
585 
586 	if (args->new_len < args->old_len) {
587 		addr = args->addr + args->new_len;
588 		len = args->old_len - args->new_len;
589 		error = kern_munmap(td, addr, len);
590 	}
591 
592 	td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
593 	return (error);
594 }
595 
596 #define LINUX_MS_ASYNC       0x0001
597 #define LINUX_MS_INVALIDATE  0x0002
598 #define LINUX_MS_SYNC        0x0004
599 
600 int
601 linux_msync(struct thread *td, struct linux_msync_args *args)
602 {
603 
604 	return (kern_msync(td, args->addr, args->len,
605 	    args->fl & ~LINUX_MS_SYNC));
606 }
607 
608 #ifdef LINUX_LEGACY_SYSCALLS
609 int
610 linux_time(struct thread *td, struct linux_time_args *args)
611 {
612 	struct timeval tv;
613 	l_time_t tm;
614 	int error;
615 
616 	microtime(&tv);
617 	tm = tv.tv_sec;
618 	if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
619 		return (error);
620 	td->td_retval[0] = tm;
621 	return (0);
622 }
623 #endif
624 
625 struct l_times_argv {
626 	l_clock_t	tms_utime;
627 	l_clock_t	tms_stime;
628 	l_clock_t	tms_cutime;
629 	l_clock_t	tms_cstime;
630 };
631 
632 
633 /*
634  * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
635  * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
636  * auxiliary vector entry.
637  */
638 #define	CLK_TCK		100
639 
640 #define	CONVOTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
641 #define	CONVNTCK(r)	(r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
642 
643 #define	CONVTCK(r)	(linux_kernver(td) >= LINUX_KERNVER_2004000 ?		\
644 			    CONVNTCK(r) : CONVOTCK(r))
645 
646 int
647 linux_times(struct thread *td, struct linux_times_args *args)
648 {
649 	struct timeval tv, utime, stime, cutime, cstime;
650 	struct l_times_argv tms;
651 	struct proc *p;
652 	int error;
653 
654 	if (args->buf != NULL) {
655 		p = td->td_proc;
656 		PROC_LOCK(p);
657 		PROC_STATLOCK(p);
658 		calcru(p, &utime, &stime);
659 		PROC_STATUNLOCK(p);
660 		calccru(p, &cutime, &cstime);
661 		PROC_UNLOCK(p);
662 
663 		tms.tms_utime = CONVTCK(utime);
664 		tms.tms_stime = CONVTCK(stime);
665 
666 		tms.tms_cutime = CONVTCK(cutime);
667 		tms.tms_cstime = CONVTCK(cstime);
668 
669 		if ((error = copyout(&tms, args->buf, sizeof(tms))))
670 			return (error);
671 	}
672 
673 	microuptime(&tv);
674 	td->td_retval[0] = (int)CONVTCK(tv);
675 	return (0);
676 }
677 
678 int
679 linux_newuname(struct thread *td, struct linux_newuname_args *args)
680 {
681 	struct l_new_utsname utsname;
682 	char osname[LINUX_MAX_UTSNAME];
683 	char osrelease[LINUX_MAX_UTSNAME];
684 	char *p;
685 
686 	linux_get_osname(td, osname);
687 	linux_get_osrelease(td, osrelease);
688 
689 	bzero(&utsname, sizeof(utsname));
690 	strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
691 	getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
692 	getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
693 	strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
694 	strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
695 	for (p = utsname.version; *p != '\0'; ++p)
696 		if (*p == '\n') {
697 			*p = '\0';
698 			break;
699 		}
700 	strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME);
701 
702 	return (copyout(&utsname, args->buf, sizeof(utsname)));
703 }
704 
705 struct l_utimbuf {
706 	l_time_t l_actime;
707 	l_time_t l_modtime;
708 };
709 
710 #ifdef LINUX_LEGACY_SYSCALLS
711 int
712 linux_utime(struct thread *td, struct linux_utime_args *args)
713 {
714 	struct timeval tv[2], *tvp;
715 	struct l_utimbuf lut;
716 	char *fname;
717 	int error;
718 
719 	LCONVPATHEXIST(td, args->fname, &fname);
720 
721 	if (args->times) {
722 		if ((error = copyin(args->times, &lut, sizeof lut))) {
723 			LFREEPATH(fname);
724 			return (error);
725 		}
726 		tv[0].tv_sec = lut.l_actime;
727 		tv[0].tv_usec = 0;
728 		tv[1].tv_sec = lut.l_modtime;
729 		tv[1].tv_usec = 0;
730 		tvp = tv;
731 	} else
732 		tvp = NULL;
733 
734 	error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, tvp,
735 	    UIO_SYSSPACE);
736 	LFREEPATH(fname);
737 	return (error);
738 }
739 #endif
740 
741 #ifdef LINUX_LEGACY_SYSCALLS
742 int
743 linux_utimes(struct thread *td, struct linux_utimes_args *args)
744 {
745 	l_timeval ltv[2];
746 	struct timeval tv[2], *tvp = NULL;
747 	char *fname;
748 	int error;
749 
750 	LCONVPATHEXIST(td, args->fname, &fname);
751 
752 	if (args->tptr != NULL) {
753 		if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
754 			LFREEPATH(fname);
755 			return (error);
756 		}
757 		tv[0].tv_sec = ltv[0].tv_sec;
758 		tv[0].tv_usec = ltv[0].tv_usec;
759 		tv[1].tv_sec = ltv[1].tv_sec;
760 		tv[1].tv_usec = ltv[1].tv_usec;
761 		tvp = tv;
762 	}
763 
764 	error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE,
765 	    tvp, UIO_SYSSPACE);
766 	LFREEPATH(fname);
767 	return (error);
768 }
769 #endif
770 
771 static int
772 linux_utimensat_nsec_valid(l_long nsec)
773 {
774 
775 	if (nsec == LINUX_UTIME_OMIT || nsec == LINUX_UTIME_NOW)
776 		return (0);
777 	if (nsec >= 0 && nsec <= 999999999)
778 		return (0);
779 	return (1);
780 }
781 
782 int
783 linux_utimensat(struct thread *td, struct linux_utimensat_args *args)
784 {
785 	struct l_timespec l_times[2];
786 	struct timespec times[2], *timesp = NULL;
787 	char *path = NULL;
788 	int error, dfd, flags = 0;
789 
790 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
791 
792 	if (args->flags & ~LINUX_AT_SYMLINK_NOFOLLOW)
793 		return (EINVAL);
794 
795 	if (args->times != NULL) {
796 		error = copyin(args->times, l_times, sizeof(l_times));
797 		if (error != 0)
798 			return (error);
799 
800 		if (linux_utimensat_nsec_valid(l_times[0].tv_nsec) != 0 ||
801 		    linux_utimensat_nsec_valid(l_times[1].tv_nsec) != 0)
802 			return (EINVAL);
803 
804 		times[0].tv_sec = l_times[0].tv_sec;
805 		switch (l_times[0].tv_nsec)
806 		{
807 		case LINUX_UTIME_OMIT:
808 			times[0].tv_nsec = UTIME_OMIT;
809 			break;
810 		case LINUX_UTIME_NOW:
811 			times[0].tv_nsec = UTIME_NOW;
812 			break;
813 		default:
814 			times[0].tv_nsec = l_times[0].tv_nsec;
815 		}
816 
817 		times[1].tv_sec = l_times[1].tv_sec;
818 		switch (l_times[1].tv_nsec)
819 		{
820 		case LINUX_UTIME_OMIT:
821 			times[1].tv_nsec = UTIME_OMIT;
822 			break;
823 		case LINUX_UTIME_NOW:
824 			times[1].tv_nsec = UTIME_NOW;
825 			break;
826 		default:
827 			times[1].tv_nsec = l_times[1].tv_nsec;
828 			break;
829 		}
830 		timesp = times;
831 
832 		/* This breaks POSIX, but is what the Linux kernel does
833 		 * _on purpose_ (documented in the man page for utimensat(2)),
834 		 * so we must follow that behaviour. */
835 		if (times[0].tv_nsec == UTIME_OMIT &&
836 		    times[1].tv_nsec == UTIME_OMIT)
837 			return (0);
838 	}
839 
840 	if (args->pathname != NULL)
841 		LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
842 	else if (args->flags != 0)
843 		return (EINVAL);
844 
845 	if (args->flags & LINUX_AT_SYMLINK_NOFOLLOW)
846 		flags |= AT_SYMLINK_NOFOLLOW;
847 
848 	if (path == NULL)
849 		error = kern_futimens(td, dfd, timesp, UIO_SYSSPACE);
850 	else {
851 		error = kern_utimensat(td, dfd, path, UIO_SYSSPACE, timesp,
852 			UIO_SYSSPACE, flags);
853 		LFREEPATH(path);
854 	}
855 
856 	return (error);
857 }
858 
859 #ifdef LINUX_LEGACY_SYSCALLS
860 int
861 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
862 {
863 	l_timeval ltv[2];
864 	struct timeval tv[2], *tvp = NULL;
865 	char *fname;
866 	int error, dfd;
867 
868 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
869 	LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
870 
871 	if (args->utimes != NULL) {
872 		if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
873 			LFREEPATH(fname);
874 			return (error);
875 		}
876 		tv[0].tv_sec = ltv[0].tv_sec;
877 		tv[0].tv_usec = ltv[0].tv_usec;
878 		tv[1].tv_sec = ltv[1].tv_sec;
879 		tv[1].tv_usec = ltv[1].tv_usec;
880 		tvp = tv;
881 	}
882 
883 	error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
884 	LFREEPATH(fname);
885 	return (error);
886 }
887 #endif
888 
889 int
890 linux_common_wait(struct thread *td, int pid, int *status,
891     int options, struct rusage *ru)
892 {
893 	int error, tmpstat;
894 
895 	error = kern_wait(td, pid, &tmpstat, options, ru);
896 	if (error)
897 		return (error);
898 
899 	if (status) {
900 		tmpstat &= 0xffff;
901 		if (WIFSIGNALED(tmpstat))
902 			tmpstat = (tmpstat & 0xffffff80) |
903 			    bsd_to_linux_signal(WTERMSIG(tmpstat));
904 		else if (WIFSTOPPED(tmpstat))
905 			tmpstat = (tmpstat & 0xffff00ff) |
906 			    (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8);
907 		else if (WIFCONTINUED(tmpstat))
908 			tmpstat = 0xffff;
909 		error = copyout(&tmpstat, status, sizeof(int));
910 	}
911 
912 	return (error);
913 }
914 
915 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
916 int
917 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
918 {
919 	struct linux_wait4_args wait4_args;
920 
921 	wait4_args.pid = args->pid;
922 	wait4_args.status = args->status;
923 	wait4_args.options = args->options;
924 	wait4_args.rusage = NULL;
925 
926 	return (linux_wait4(td, &wait4_args));
927 }
928 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
929 
930 int
931 linux_wait4(struct thread *td, struct linux_wait4_args *args)
932 {
933 	int error, options;
934 	struct rusage ru, *rup;
935 
936 	if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG |
937 	    LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
938 		return (EINVAL);
939 
940 	options = WEXITED;
941 	linux_to_bsd_waitopts(args->options, &options);
942 
943 	if (args->rusage != NULL)
944 		rup = &ru;
945 	else
946 		rup = NULL;
947 	error = linux_common_wait(td, args->pid, args->status, options, rup);
948 	if (error != 0)
949 		return (error);
950 	if (args->rusage != NULL)
951 		error = linux_copyout_rusage(&ru, args->rusage);
952 	return (error);
953 }
954 
955 int
956 linux_waitid(struct thread *td, struct linux_waitid_args *args)
957 {
958 	int status, options, sig;
959 	struct __wrusage wru;
960 	siginfo_t siginfo;
961 	l_siginfo_t lsi;
962 	idtype_t idtype;
963 	struct proc *p;
964 	int error;
965 
966 	options = 0;
967 	linux_to_bsd_waitopts(args->options, &options);
968 
969 	if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED))
970 		return (EINVAL);
971 	if (!(options & (WEXITED | WUNTRACED | WCONTINUED)))
972 		return (EINVAL);
973 
974 	switch (args->idtype) {
975 	case LINUX_P_ALL:
976 		idtype = P_ALL;
977 		break;
978 	case LINUX_P_PID:
979 		if (args->id <= 0)
980 			return (EINVAL);
981 		idtype = P_PID;
982 		break;
983 	case LINUX_P_PGID:
984 		if (args->id <= 0)
985 			return (EINVAL);
986 		idtype = P_PGID;
987 		break;
988 	default:
989 		return (EINVAL);
990 	}
991 
992 	error = kern_wait6(td, idtype, args->id, &status, options,
993 	    &wru, &siginfo);
994 	if (error != 0)
995 		return (error);
996 	if (args->rusage != NULL) {
997 		error = linux_copyout_rusage(&wru.wru_children,
998 		    args->rusage);
999 		if (error != 0)
1000 			return (error);
1001 	}
1002 	if (args->info != NULL) {
1003 		p = td->td_proc;
1004 		bzero(&lsi, sizeof(lsi));
1005 		if (td->td_retval[0] != 0) {
1006 			sig = bsd_to_linux_signal(siginfo.si_signo);
1007 			siginfo_to_lsiginfo(&siginfo, &lsi, sig);
1008 		}
1009 		error = copyout(&lsi, args->info, sizeof(lsi));
1010 	}
1011 	td->td_retval[0] = 0;
1012 
1013 	return (error);
1014 }
1015 
1016 #ifdef LINUX_LEGACY_SYSCALLS
1017 int
1018 linux_mknod(struct thread *td, struct linux_mknod_args *args)
1019 {
1020 	char *path;
1021 	int error;
1022 
1023 	LCONVPATHCREAT(td, args->path, &path);
1024 
1025 	switch (args->mode & S_IFMT) {
1026 	case S_IFIFO:
1027 	case S_IFSOCK:
1028 		error = kern_mkfifoat(td, AT_FDCWD, path, UIO_SYSSPACE,
1029 		    args->mode);
1030 		break;
1031 
1032 	case S_IFCHR:
1033 	case S_IFBLK:
1034 		error = kern_mknodat(td, AT_FDCWD, path, UIO_SYSSPACE,
1035 		    args->mode, args->dev);
1036 		break;
1037 
1038 	case S_IFDIR:
1039 		error = EPERM;
1040 		break;
1041 
1042 	case 0:
1043 		args->mode |= S_IFREG;
1044 		/* FALLTHROUGH */
1045 	case S_IFREG:
1046 		error = kern_openat(td, AT_FDCWD, path, UIO_SYSSPACE,
1047 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1048 		if (error == 0)
1049 			kern_close(td, td->td_retval[0]);
1050 		break;
1051 
1052 	default:
1053 		error = EINVAL;
1054 		break;
1055 	}
1056 	LFREEPATH(path);
1057 	return (error);
1058 }
1059 #endif
1060 
1061 int
1062 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
1063 {
1064 	char *path;
1065 	int error, dfd;
1066 
1067 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
1068 	LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
1069 
1070 	switch (args->mode & S_IFMT) {
1071 	case S_IFIFO:
1072 	case S_IFSOCK:
1073 		error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
1074 		break;
1075 
1076 	case S_IFCHR:
1077 	case S_IFBLK:
1078 		error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
1079 		    args->dev);
1080 		break;
1081 
1082 	case S_IFDIR:
1083 		error = EPERM;
1084 		break;
1085 
1086 	case 0:
1087 		args->mode |= S_IFREG;
1088 		/* FALLTHROUGH */
1089 	case S_IFREG:
1090 		error = kern_openat(td, dfd, path, UIO_SYSSPACE,
1091 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1092 		if (error == 0)
1093 			kern_close(td, td->td_retval[0]);
1094 		break;
1095 
1096 	default:
1097 		error = EINVAL;
1098 		break;
1099 	}
1100 	LFREEPATH(path);
1101 	return (error);
1102 }
1103 
1104 /*
1105  * UGH! This is just about the dumbest idea I've ever heard!!
1106  */
1107 int
1108 linux_personality(struct thread *td, struct linux_personality_args *args)
1109 {
1110 	struct linux_pemuldata *pem;
1111 	struct proc *p = td->td_proc;
1112 	uint32_t old;
1113 
1114 	PROC_LOCK(p);
1115 	pem = pem_find(p);
1116 	old = pem->persona;
1117 	if (args->per != 0xffffffff)
1118 		pem->persona = args->per;
1119 	PROC_UNLOCK(p);
1120 
1121 	td->td_retval[0] = old;
1122 	return (0);
1123 }
1124 
1125 struct l_itimerval {
1126 	l_timeval it_interval;
1127 	l_timeval it_value;
1128 };
1129 
1130 #define	B2L_ITIMERVAL(bip, lip)						\
1131 	(bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec;		\
1132 	(bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec;	\
1133 	(bip)->it_value.tv_sec = (lip)->it_value.tv_sec;		\
1134 	(bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1135 
1136 int
1137 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1138 {
1139 	int error;
1140 	struct l_itimerval ls;
1141 	struct itimerval aitv, oitv;
1142 
1143 	if (uap->itv == NULL) {
1144 		uap->itv = uap->oitv;
1145 		return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1146 	}
1147 
1148 	error = copyin(uap->itv, &ls, sizeof(ls));
1149 	if (error != 0)
1150 		return (error);
1151 	B2L_ITIMERVAL(&aitv, &ls);
1152 	error = kern_setitimer(td, uap->which, &aitv, &oitv);
1153 	if (error != 0 || uap->oitv == NULL)
1154 		return (error);
1155 	B2L_ITIMERVAL(&ls, &oitv);
1156 
1157 	return (copyout(&ls, uap->oitv, sizeof(ls)));
1158 }
1159 
1160 int
1161 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1162 {
1163 	int error;
1164 	struct l_itimerval ls;
1165 	struct itimerval aitv;
1166 
1167 	error = kern_getitimer(td, uap->which, &aitv);
1168 	if (error != 0)
1169 		return (error);
1170 	B2L_ITIMERVAL(&ls, &aitv);
1171 	return (copyout(&ls, uap->itv, sizeof(ls)));
1172 }
1173 
1174 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1175 int
1176 linux_nice(struct thread *td, struct linux_nice_args *args)
1177 {
1178 	struct setpriority_args bsd_args;
1179 
1180 	bsd_args.which = PRIO_PROCESS;
1181 	bsd_args.who = 0;		/* current process */
1182 	bsd_args.prio = args->inc;
1183 	return (sys_setpriority(td, &bsd_args));
1184 }
1185 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1186 
1187 int
1188 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1189 {
1190 	struct ucred *newcred, *oldcred;
1191 	l_gid_t *linux_gidset;
1192 	gid_t *bsd_gidset;
1193 	int ngrp, error;
1194 	struct proc *p;
1195 
1196 	ngrp = args->gidsetsize;
1197 	if (ngrp < 0 || ngrp >= ngroups_max + 1)
1198 		return (EINVAL);
1199 	linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK);
1200 	error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1201 	if (error)
1202 		goto out;
1203 	newcred = crget();
1204 	crextend(newcred, ngrp + 1);
1205 	p = td->td_proc;
1206 	PROC_LOCK(p);
1207 	oldcred = p->p_ucred;
1208 	crcopy(newcred, oldcred);
1209 
1210 	/*
1211 	 * cr_groups[0] holds egid. Setting the whole set from
1212 	 * the supplied set will cause egid to be changed too.
1213 	 * Keep cr_groups[0] unchanged to prevent that.
1214 	 */
1215 
1216 	if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS)) != 0) {
1217 		PROC_UNLOCK(p);
1218 		crfree(newcred);
1219 		goto out;
1220 	}
1221 
1222 	if (ngrp > 0) {
1223 		newcred->cr_ngroups = ngrp + 1;
1224 
1225 		bsd_gidset = newcred->cr_groups;
1226 		ngrp--;
1227 		while (ngrp >= 0) {
1228 			bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1229 			ngrp--;
1230 		}
1231 	} else
1232 		newcred->cr_ngroups = 1;
1233 
1234 	setsugid(p);
1235 	proc_set_cred(p, newcred);
1236 	PROC_UNLOCK(p);
1237 	crfree(oldcred);
1238 	error = 0;
1239 out:
1240 	free(linux_gidset, M_LINUX);
1241 	return (error);
1242 }
1243 
1244 int
1245 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1246 {
1247 	struct ucred *cred;
1248 	l_gid_t *linux_gidset;
1249 	gid_t *bsd_gidset;
1250 	int bsd_gidsetsz, ngrp, error;
1251 
1252 	cred = td->td_ucred;
1253 	bsd_gidset = cred->cr_groups;
1254 	bsd_gidsetsz = cred->cr_ngroups - 1;
1255 
1256 	/*
1257 	 * cr_groups[0] holds egid. Returning the whole set
1258 	 * here will cause a duplicate. Exclude cr_groups[0]
1259 	 * to prevent that.
1260 	 */
1261 
1262 	if ((ngrp = args->gidsetsize) == 0) {
1263 		td->td_retval[0] = bsd_gidsetsz;
1264 		return (0);
1265 	}
1266 
1267 	if (ngrp < bsd_gidsetsz)
1268 		return (EINVAL);
1269 
1270 	ngrp = 0;
1271 	linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1272 	    M_LINUX, M_WAITOK);
1273 	while (ngrp < bsd_gidsetsz) {
1274 		linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1275 		ngrp++;
1276 	}
1277 
1278 	error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1279 	free(linux_gidset, M_LINUX);
1280 	if (error)
1281 		return (error);
1282 
1283 	td->td_retval[0] = ngrp;
1284 	return (0);
1285 }
1286 
1287 int
1288 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1289 {
1290 	struct rlimit bsd_rlim;
1291 	struct l_rlimit rlim;
1292 	u_int which;
1293 	int error;
1294 
1295 	if (args->resource >= LINUX_RLIM_NLIMITS)
1296 		return (EINVAL);
1297 
1298 	which = linux_to_bsd_resource[args->resource];
1299 	if (which == -1)
1300 		return (EINVAL);
1301 
1302 	error = copyin(args->rlim, &rlim, sizeof(rlim));
1303 	if (error)
1304 		return (error);
1305 
1306 	bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1307 	bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1308 	return (kern_setrlimit(td, which, &bsd_rlim));
1309 }
1310 
1311 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1312 int
1313 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1314 {
1315 	struct l_rlimit rlim;
1316 	struct rlimit bsd_rlim;
1317 	u_int which;
1318 
1319 	if (args->resource >= LINUX_RLIM_NLIMITS)
1320 		return (EINVAL);
1321 
1322 	which = linux_to_bsd_resource[args->resource];
1323 	if (which == -1)
1324 		return (EINVAL);
1325 
1326 	lim_rlimit(td, which, &bsd_rlim);
1327 
1328 #ifdef COMPAT_LINUX32
1329 	rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1330 	if (rlim.rlim_cur == UINT_MAX)
1331 		rlim.rlim_cur = INT_MAX;
1332 	rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1333 	if (rlim.rlim_max == UINT_MAX)
1334 		rlim.rlim_max = INT_MAX;
1335 #else
1336 	rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1337 	if (rlim.rlim_cur == ULONG_MAX)
1338 		rlim.rlim_cur = LONG_MAX;
1339 	rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1340 	if (rlim.rlim_max == ULONG_MAX)
1341 		rlim.rlim_max = LONG_MAX;
1342 #endif
1343 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1344 }
1345 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1346 
1347 int
1348 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1349 {
1350 	struct l_rlimit rlim;
1351 	struct rlimit bsd_rlim;
1352 	u_int which;
1353 
1354 	if (args->resource >= LINUX_RLIM_NLIMITS)
1355 		return (EINVAL);
1356 
1357 	which = linux_to_bsd_resource[args->resource];
1358 	if (which == -1)
1359 		return (EINVAL);
1360 
1361 	lim_rlimit(td, which, &bsd_rlim);
1362 
1363 	rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1364 	rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1365 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1366 }
1367 
1368 int
1369 linux_sched_setscheduler(struct thread *td,
1370     struct linux_sched_setscheduler_args *args)
1371 {
1372 	struct sched_param sched_param;
1373 	struct thread *tdt;
1374 	int error, policy;
1375 
1376 	switch (args->policy) {
1377 	case LINUX_SCHED_OTHER:
1378 		policy = SCHED_OTHER;
1379 		break;
1380 	case LINUX_SCHED_FIFO:
1381 		policy = SCHED_FIFO;
1382 		break;
1383 	case LINUX_SCHED_RR:
1384 		policy = SCHED_RR;
1385 		break;
1386 	default:
1387 		return (EINVAL);
1388 	}
1389 
1390 	error = copyin(args->param, &sched_param, sizeof(sched_param));
1391 	if (error)
1392 		return (error);
1393 
1394 	tdt = linux_tdfind(td, args->pid, -1);
1395 	if (tdt == NULL)
1396 		return (ESRCH);
1397 
1398 	error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1399 	PROC_UNLOCK(tdt->td_proc);
1400 	return (error);
1401 }
1402 
1403 int
1404 linux_sched_getscheduler(struct thread *td,
1405     struct linux_sched_getscheduler_args *args)
1406 {
1407 	struct thread *tdt;
1408 	int error, policy;
1409 
1410 	tdt = linux_tdfind(td, args->pid, -1);
1411 	if (tdt == NULL)
1412 		return (ESRCH);
1413 
1414 	error = kern_sched_getscheduler(td, tdt, &policy);
1415 	PROC_UNLOCK(tdt->td_proc);
1416 
1417 	switch (policy) {
1418 	case SCHED_OTHER:
1419 		td->td_retval[0] = LINUX_SCHED_OTHER;
1420 		break;
1421 	case SCHED_FIFO:
1422 		td->td_retval[0] = LINUX_SCHED_FIFO;
1423 		break;
1424 	case SCHED_RR:
1425 		td->td_retval[0] = LINUX_SCHED_RR;
1426 		break;
1427 	}
1428 	return (error);
1429 }
1430 
1431 int
1432 linux_sched_get_priority_max(struct thread *td,
1433     struct linux_sched_get_priority_max_args *args)
1434 {
1435 	struct sched_get_priority_max_args bsd;
1436 
1437 	switch (args->policy) {
1438 	case LINUX_SCHED_OTHER:
1439 		bsd.policy = SCHED_OTHER;
1440 		break;
1441 	case LINUX_SCHED_FIFO:
1442 		bsd.policy = SCHED_FIFO;
1443 		break;
1444 	case LINUX_SCHED_RR:
1445 		bsd.policy = SCHED_RR;
1446 		break;
1447 	default:
1448 		return (EINVAL);
1449 	}
1450 	return (sys_sched_get_priority_max(td, &bsd));
1451 }
1452 
1453 int
1454 linux_sched_get_priority_min(struct thread *td,
1455     struct linux_sched_get_priority_min_args *args)
1456 {
1457 	struct sched_get_priority_min_args bsd;
1458 
1459 	switch (args->policy) {
1460 	case LINUX_SCHED_OTHER:
1461 		bsd.policy = SCHED_OTHER;
1462 		break;
1463 	case LINUX_SCHED_FIFO:
1464 		bsd.policy = SCHED_FIFO;
1465 		break;
1466 	case LINUX_SCHED_RR:
1467 		bsd.policy = SCHED_RR;
1468 		break;
1469 	default:
1470 		return (EINVAL);
1471 	}
1472 	return (sys_sched_get_priority_min(td, &bsd));
1473 }
1474 
1475 #define REBOOT_CAD_ON	0x89abcdef
1476 #define REBOOT_CAD_OFF	0
1477 #define REBOOT_HALT	0xcdef0123
1478 #define REBOOT_RESTART	0x01234567
1479 #define REBOOT_RESTART2	0xA1B2C3D4
1480 #define REBOOT_POWEROFF	0x4321FEDC
1481 #define REBOOT_MAGIC1	0xfee1dead
1482 #define REBOOT_MAGIC2	0x28121969
1483 #define REBOOT_MAGIC2A	0x05121996
1484 #define REBOOT_MAGIC2B	0x16041998
1485 
1486 int
1487 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1488 {
1489 	struct reboot_args bsd_args;
1490 
1491 	if (args->magic1 != REBOOT_MAGIC1)
1492 		return (EINVAL);
1493 
1494 	switch (args->magic2) {
1495 	case REBOOT_MAGIC2:
1496 	case REBOOT_MAGIC2A:
1497 	case REBOOT_MAGIC2B:
1498 		break;
1499 	default:
1500 		return (EINVAL);
1501 	}
1502 
1503 	switch (args->cmd) {
1504 	case REBOOT_CAD_ON:
1505 	case REBOOT_CAD_OFF:
1506 		return (priv_check(td, PRIV_REBOOT));
1507 	case REBOOT_HALT:
1508 		bsd_args.opt = RB_HALT;
1509 		break;
1510 	case REBOOT_RESTART:
1511 	case REBOOT_RESTART2:
1512 		bsd_args.opt = 0;
1513 		break;
1514 	case REBOOT_POWEROFF:
1515 		bsd_args.opt = RB_POWEROFF;
1516 		break;
1517 	default:
1518 		return (EINVAL);
1519 	}
1520 	return (sys_reboot(td, &bsd_args));
1521 }
1522 
1523 
1524 /*
1525  * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1526  * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
1527  * are assumed to be preserved. The following lightweight syscalls fixes
1528  * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
1529  *
1530  * linux_getpid() - MP SAFE
1531  * linux_getgid() - MP SAFE
1532  * linux_getuid() - MP SAFE
1533  */
1534 
1535 int
1536 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1537 {
1538 
1539 	td->td_retval[0] = td->td_proc->p_pid;
1540 
1541 	return (0);
1542 }
1543 
1544 int
1545 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1546 {
1547 	struct linux_emuldata *em;
1548 
1549 	em = em_find(td);
1550 	KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1551 
1552 	td->td_retval[0] = em->em_tid;
1553 
1554 	return (0);
1555 }
1556 
1557 
1558 int
1559 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1560 {
1561 
1562 	td->td_retval[0] = kern_getppid(td);
1563 	return (0);
1564 }
1565 
1566 int
1567 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1568 {
1569 
1570 	td->td_retval[0] = td->td_ucred->cr_rgid;
1571 	return (0);
1572 }
1573 
1574 int
1575 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1576 {
1577 
1578 	td->td_retval[0] = td->td_ucred->cr_ruid;
1579 	return (0);
1580 }
1581 
1582 
1583 int
1584 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1585 {
1586 	struct getsid_args bsd;
1587 
1588 	bsd.pid = args->pid;
1589 	return (sys_getsid(td, &bsd));
1590 }
1591 
1592 int
1593 linux_nosys(struct thread *td, struct nosys_args *ignore)
1594 {
1595 
1596 	return (ENOSYS);
1597 }
1598 
1599 int
1600 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1601 {
1602 	struct getpriority_args bsd_args;
1603 	int error;
1604 
1605 	bsd_args.which = args->which;
1606 	bsd_args.who = args->who;
1607 	error = sys_getpriority(td, &bsd_args);
1608 	td->td_retval[0] = 20 - td->td_retval[0];
1609 	return (error);
1610 }
1611 
1612 int
1613 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1614 {
1615 	int name[2];
1616 
1617 	name[0] = CTL_KERN;
1618 	name[1] = KERN_HOSTNAME;
1619 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1620 	    args->len, 0, 0));
1621 }
1622 
1623 int
1624 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1625 {
1626 	int name[2];
1627 
1628 	name[0] = CTL_KERN;
1629 	name[1] = KERN_NISDOMAINNAME;
1630 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1631 	    args->len, 0, 0));
1632 }
1633 
1634 int
1635 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1636 {
1637 
1638 	LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1639 	    args->error_code);
1640 
1641 	/*
1642 	 * XXX: we should send a signal to the parent if
1643 	 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1644 	 * as it doesnt occur often.
1645 	 */
1646 	exit1(td, args->error_code, 0);
1647 		/* NOTREACHED */
1648 }
1649 
1650 #define _LINUX_CAPABILITY_VERSION_1  0x19980330
1651 #define _LINUX_CAPABILITY_VERSION_2  0x20071026
1652 #define _LINUX_CAPABILITY_VERSION_3  0x20080522
1653 
1654 struct l_user_cap_header {
1655 	l_int	version;
1656 	l_int	pid;
1657 };
1658 
1659 struct l_user_cap_data {
1660 	l_int	effective;
1661 	l_int	permitted;
1662 	l_int	inheritable;
1663 };
1664 
1665 int
1666 linux_capget(struct thread *td, struct linux_capget_args *uap)
1667 {
1668 	struct l_user_cap_header luch;
1669 	struct l_user_cap_data lucd[2];
1670 	int error, u32s;
1671 
1672 	if (uap->hdrp == NULL)
1673 		return (EFAULT);
1674 
1675 	error = copyin(uap->hdrp, &luch, sizeof(luch));
1676 	if (error != 0)
1677 		return (error);
1678 
1679 	switch (luch.version) {
1680 	case _LINUX_CAPABILITY_VERSION_1:
1681 		u32s = 1;
1682 		break;
1683 	case _LINUX_CAPABILITY_VERSION_2:
1684 	case _LINUX_CAPABILITY_VERSION_3:
1685 		u32s = 2;
1686 		break;
1687 	default:
1688 		luch.version = _LINUX_CAPABILITY_VERSION_1;
1689 		error = copyout(&luch, uap->hdrp, sizeof(luch));
1690 		if (error)
1691 			return (error);
1692 		return (EINVAL);
1693 	}
1694 
1695 	if (luch.pid)
1696 		return (EPERM);
1697 
1698 	if (uap->datap) {
1699 		/*
1700 		 * The current implementation doesn't support setting
1701 		 * a capability (it's essentially a stub) so indicate
1702 		 * that no capabilities are currently set or available
1703 		 * to request.
1704 		 */
1705 		memset(&lucd, 0, u32s * sizeof(lucd[0]));
1706 		error = copyout(&lucd, uap->datap, u32s * sizeof(lucd[0]));
1707 	}
1708 
1709 	return (error);
1710 }
1711 
1712 int
1713 linux_capset(struct thread *td, struct linux_capset_args *uap)
1714 {
1715 	struct l_user_cap_header luch;
1716 	struct l_user_cap_data lucd[2];
1717 	int error, i, u32s;
1718 
1719 	if (uap->hdrp == NULL || uap->datap == NULL)
1720 		return (EFAULT);
1721 
1722 	error = copyin(uap->hdrp, &luch, sizeof(luch));
1723 	if (error != 0)
1724 		return (error);
1725 
1726 	switch (luch.version) {
1727 	case _LINUX_CAPABILITY_VERSION_1:
1728 		u32s = 1;
1729 		break;
1730 	case _LINUX_CAPABILITY_VERSION_2:
1731 	case _LINUX_CAPABILITY_VERSION_3:
1732 		u32s = 2;
1733 		break;
1734 	default:
1735 		luch.version = _LINUX_CAPABILITY_VERSION_1;
1736 		error = copyout(&luch, uap->hdrp, sizeof(luch));
1737 		if (error)
1738 			return (error);
1739 		return (EINVAL);
1740 	}
1741 
1742 	if (luch.pid)
1743 		return (EPERM);
1744 
1745 	error = copyin(uap->datap, &lucd, u32s * sizeof(lucd[0]));
1746 	if (error != 0)
1747 		return (error);
1748 
1749 	/* We currently don't support setting any capabilities. */
1750 	for (i = 0; i < u32s; i++) {
1751 		if (lucd[i].effective || lucd[i].permitted ||
1752 		    lucd[i].inheritable) {
1753 			linux_msg(td,
1754 			    "capset[%d] effective=0x%x, permitted=0x%x, "
1755 			    "inheritable=0x%x is not implemented", i,
1756 			    (int)lucd[i].effective, (int)lucd[i].permitted,
1757 			    (int)lucd[i].inheritable);
1758 			return (EPERM);
1759 		}
1760 	}
1761 
1762 	return (0);
1763 }
1764 
1765 int
1766 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1767 {
1768 	int error = 0, max_size;
1769 	struct proc *p = td->td_proc;
1770 	char comm[LINUX_MAX_COMM_LEN];
1771 	int pdeath_signal;
1772 
1773 	switch (args->option) {
1774 	case LINUX_PR_SET_PDEATHSIG:
1775 		if (!LINUX_SIG_VALID(args->arg2))
1776 			return (EINVAL);
1777 		pdeath_signal = linux_to_bsd_signal(args->arg2);
1778 		return (kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_CTL,
1779 		    &pdeath_signal));
1780 	case LINUX_PR_GET_PDEATHSIG:
1781 		error = kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_STATUS,
1782 		    &pdeath_signal);
1783 		if (error != 0)
1784 			return (error);
1785 		pdeath_signal = bsd_to_linux_signal(pdeath_signal);
1786 		return (copyout(&pdeath_signal,
1787 		    (void *)(register_t)args->arg2,
1788 		    sizeof(pdeath_signal)));
1789 		break;
1790 	case LINUX_PR_GET_KEEPCAPS:
1791 		/*
1792 		 * Indicate that we always clear the effective and
1793 		 * permitted capability sets when the user id becomes
1794 		 * non-zero (actually the capability sets are simply
1795 		 * always zero in the current implementation).
1796 		 */
1797 		td->td_retval[0] = 0;
1798 		break;
1799 	case LINUX_PR_SET_KEEPCAPS:
1800 		/*
1801 		 * Ignore requests to keep the effective and permitted
1802 		 * capability sets when the user id becomes non-zero.
1803 		 */
1804 		break;
1805 	case LINUX_PR_SET_NAME:
1806 		/*
1807 		 * To be on the safe side we need to make sure to not
1808 		 * overflow the size a Linux program expects. We already
1809 		 * do this here in the copyin, so that we don't need to
1810 		 * check on copyout.
1811 		 */
1812 		max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1813 		error = copyinstr((void *)(register_t)args->arg2, comm,
1814 		    max_size, NULL);
1815 
1816 		/* Linux silently truncates the name if it is too long. */
1817 		if (error == ENAMETOOLONG) {
1818 			/*
1819 			 * XXX: copyinstr() isn't documented to populate the
1820 			 * array completely, so do a copyin() to be on the
1821 			 * safe side. This should be changed in case
1822 			 * copyinstr() is changed to guarantee this.
1823 			 */
1824 			error = copyin((void *)(register_t)args->arg2, comm,
1825 			    max_size - 1);
1826 			comm[max_size - 1] = '\0';
1827 		}
1828 		if (error)
1829 			return (error);
1830 
1831 		PROC_LOCK(p);
1832 		strlcpy(p->p_comm, comm, sizeof(p->p_comm));
1833 		PROC_UNLOCK(p);
1834 		break;
1835 	case LINUX_PR_GET_NAME:
1836 		PROC_LOCK(p);
1837 		strlcpy(comm, p->p_comm, sizeof(comm));
1838 		PROC_UNLOCK(p);
1839 		error = copyout(comm, (void *)(register_t)args->arg2,
1840 		    strlen(comm) + 1);
1841 		break;
1842 	default:
1843 		error = EINVAL;
1844 		break;
1845 	}
1846 
1847 	return (error);
1848 }
1849 
1850 int
1851 linux_sched_setparam(struct thread *td,
1852     struct linux_sched_setparam_args *uap)
1853 {
1854 	struct sched_param sched_param;
1855 	struct thread *tdt;
1856 	int error;
1857 
1858 	error = copyin(uap->param, &sched_param, sizeof(sched_param));
1859 	if (error)
1860 		return (error);
1861 
1862 	tdt = linux_tdfind(td, uap->pid, -1);
1863 	if (tdt == NULL)
1864 		return (ESRCH);
1865 
1866 	error = kern_sched_setparam(td, tdt, &sched_param);
1867 	PROC_UNLOCK(tdt->td_proc);
1868 	return (error);
1869 }
1870 
1871 int
1872 linux_sched_getparam(struct thread *td,
1873     struct linux_sched_getparam_args *uap)
1874 {
1875 	struct sched_param sched_param;
1876 	struct thread *tdt;
1877 	int error;
1878 
1879 	tdt = linux_tdfind(td, uap->pid, -1);
1880 	if (tdt == NULL)
1881 		return (ESRCH);
1882 
1883 	error = kern_sched_getparam(td, tdt, &sched_param);
1884 	PROC_UNLOCK(tdt->td_proc);
1885 	if (error == 0)
1886 		error = copyout(&sched_param, uap->param,
1887 		    sizeof(sched_param));
1888 	return (error);
1889 }
1890 
1891 /*
1892  * Get affinity of a process.
1893  */
1894 int
1895 linux_sched_getaffinity(struct thread *td,
1896     struct linux_sched_getaffinity_args *args)
1897 {
1898 	int error;
1899 	struct thread *tdt;
1900 
1901 	if (args->len < sizeof(cpuset_t))
1902 		return (EINVAL);
1903 
1904 	tdt = linux_tdfind(td, args->pid, -1);
1905 	if (tdt == NULL)
1906 		return (ESRCH);
1907 
1908 	PROC_UNLOCK(tdt->td_proc);
1909 
1910 	error = kern_cpuset_getaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
1911 	    tdt->td_tid, sizeof(cpuset_t), (cpuset_t *)args->user_mask_ptr);
1912 	if (error == 0)
1913 		td->td_retval[0] = sizeof(cpuset_t);
1914 
1915 	return (error);
1916 }
1917 
1918 /*
1919  *  Set affinity of a process.
1920  */
1921 int
1922 linux_sched_setaffinity(struct thread *td,
1923     struct linux_sched_setaffinity_args *args)
1924 {
1925 	struct thread *tdt;
1926 
1927 	if (args->len < sizeof(cpuset_t))
1928 		return (EINVAL);
1929 
1930 	tdt = linux_tdfind(td, args->pid, -1);
1931 	if (tdt == NULL)
1932 		return (ESRCH);
1933 
1934 	PROC_UNLOCK(tdt->td_proc);
1935 
1936 	return (kern_cpuset_setaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
1937 	    tdt->td_tid, sizeof(cpuset_t), (cpuset_t *) args->user_mask_ptr));
1938 }
1939 
1940 struct linux_rlimit64 {
1941 	uint64_t	rlim_cur;
1942 	uint64_t	rlim_max;
1943 };
1944 
1945 int
1946 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
1947 {
1948 	struct rlimit rlim, nrlim;
1949 	struct linux_rlimit64 lrlim;
1950 	struct proc *p;
1951 	u_int which;
1952 	int flags;
1953 	int error;
1954 
1955 	if (args->resource >= LINUX_RLIM_NLIMITS)
1956 		return (EINVAL);
1957 
1958 	which = linux_to_bsd_resource[args->resource];
1959 	if (which == -1)
1960 		return (EINVAL);
1961 
1962 	if (args->new != NULL) {
1963 		/*
1964 		 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
1965 		 * rlim is unsigned 64-bit. FreeBSD treats negative limits
1966 		 * as INFINITY so we do not need a conversion even.
1967 		 */
1968 		error = copyin(args->new, &nrlim, sizeof(nrlim));
1969 		if (error != 0)
1970 			return (error);
1971 	}
1972 
1973 	flags = PGET_HOLD | PGET_NOTWEXIT;
1974 	if (args->new != NULL)
1975 		flags |= PGET_CANDEBUG;
1976 	else
1977 		flags |= PGET_CANSEE;
1978 	error = pget(args->pid, flags, &p);
1979 	if (error != 0)
1980 		return (error);
1981 
1982 	if (args->old != NULL) {
1983 		PROC_LOCK(p);
1984 		lim_rlimit_proc(p, which, &rlim);
1985 		PROC_UNLOCK(p);
1986 		if (rlim.rlim_cur == RLIM_INFINITY)
1987 			lrlim.rlim_cur = LINUX_RLIM_INFINITY;
1988 		else
1989 			lrlim.rlim_cur = rlim.rlim_cur;
1990 		if (rlim.rlim_max == RLIM_INFINITY)
1991 			lrlim.rlim_max = LINUX_RLIM_INFINITY;
1992 		else
1993 			lrlim.rlim_max = rlim.rlim_max;
1994 		error = copyout(&lrlim, args->old, sizeof(lrlim));
1995 		if (error != 0)
1996 			goto out;
1997 	}
1998 
1999 	if (args->new != NULL)
2000 		error = kern_proc_setrlimit(td, p, which, &nrlim);
2001 
2002  out:
2003 	PRELE(p);
2004 	return (error);
2005 }
2006 
2007 int
2008 linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2009 {
2010 	struct timeval utv, tv0, tv1, *tvp;
2011 	struct l_pselect6arg lpse6;
2012 	struct l_timespec lts;
2013 	struct timespec uts;
2014 	l_sigset_t l_ss;
2015 	sigset_t *ssp;
2016 	sigset_t ss;
2017 	int error;
2018 
2019 	ssp = NULL;
2020 	if (args->sig != NULL) {
2021 		error = copyin(args->sig, &lpse6, sizeof(lpse6));
2022 		if (error != 0)
2023 			return (error);
2024 		if (lpse6.ss_len != sizeof(l_ss))
2025 			return (EINVAL);
2026 		if (lpse6.ss != 0) {
2027 			error = copyin(PTRIN(lpse6.ss), &l_ss,
2028 			    sizeof(l_ss));
2029 			if (error != 0)
2030 				return (error);
2031 			linux_to_bsd_sigset(&l_ss, &ss);
2032 			ssp = &ss;
2033 		}
2034 	}
2035 
2036 	/*
2037 	 * Currently glibc changes nanosecond number to microsecond.
2038 	 * This mean losing precision but for now it is hardly seen.
2039 	 */
2040 	if (args->tsp != NULL) {
2041 		error = copyin(args->tsp, &lts, sizeof(lts));
2042 		if (error != 0)
2043 			return (error);
2044 		error = linux_to_native_timespec(&uts, &lts);
2045 		if (error != 0)
2046 			return (error);
2047 
2048 		TIMESPEC_TO_TIMEVAL(&utv, &uts);
2049 		if (itimerfix(&utv))
2050 			return (EINVAL);
2051 
2052 		microtime(&tv0);
2053 		tvp = &utv;
2054 	} else
2055 		tvp = NULL;
2056 
2057 	error = kern_pselect(td, args->nfds, args->readfds, args->writefds,
2058 	    args->exceptfds, tvp, ssp, LINUX_NFDBITS);
2059 
2060 	if (error == 0 && args->tsp != NULL) {
2061 		if (td->td_retval[0] != 0) {
2062 			/*
2063 			 * Compute how much time was left of the timeout,
2064 			 * by subtracting the current time and the time
2065 			 * before we started the call, and subtracting
2066 			 * that result from the user-supplied value.
2067 			 */
2068 
2069 			microtime(&tv1);
2070 			timevalsub(&tv1, &tv0);
2071 			timevalsub(&utv, &tv1);
2072 			if (utv.tv_sec < 0)
2073 				timevalclear(&utv);
2074 		} else
2075 			timevalclear(&utv);
2076 
2077 		TIMEVAL_TO_TIMESPEC(&utv, &uts);
2078 
2079 		error = native_to_linux_timespec(&lts, &uts);
2080 		if (error == 0)
2081 			error = copyout(&lts, args->tsp, sizeof(lts));
2082 	}
2083 
2084 	return (error);
2085 }
2086 
2087 int
2088 linux_ppoll(struct thread *td, struct linux_ppoll_args *args)
2089 {
2090 	struct timespec ts0, ts1;
2091 	struct l_timespec lts;
2092 	struct timespec uts, *tsp;
2093 	l_sigset_t l_ss;
2094 	sigset_t *ssp;
2095 	sigset_t ss;
2096 	int error;
2097 
2098 	if (args->sset != NULL) {
2099 		if (args->ssize != sizeof(l_ss))
2100 			return (EINVAL);
2101 		error = copyin(args->sset, &l_ss, sizeof(l_ss));
2102 		if (error)
2103 			return (error);
2104 		linux_to_bsd_sigset(&l_ss, &ss);
2105 		ssp = &ss;
2106 	} else
2107 		ssp = NULL;
2108 	if (args->tsp != NULL) {
2109 		error = copyin(args->tsp, &lts, sizeof(lts));
2110 		if (error)
2111 			return (error);
2112 		error = linux_to_native_timespec(&uts, &lts);
2113 		if (error != 0)
2114 			return (error);
2115 
2116 		nanotime(&ts0);
2117 		tsp = &uts;
2118 	} else
2119 		tsp = NULL;
2120 
2121 	error = kern_poll(td, args->fds, args->nfds, tsp, ssp);
2122 
2123 	if (error == 0 && args->tsp != NULL) {
2124 		if (td->td_retval[0]) {
2125 			nanotime(&ts1);
2126 			timespecsub(&ts1, &ts0, &ts1);
2127 			timespecsub(&uts, &ts1, &uts);
2128 			if (uts.tv_sec < 0)
2129 				timespecclear(&uts);
2130 		} else
2131 			timespecclear(&uts);
2132 
2133 		error = native_to_linux_timespec(&lts, &uts);
2134 		if (error == 0)
2135 			error = copyout(&lts, args->tsp, sizeof(lts));
2136 	}
2137 
2138 	return (error);
2139 }
2140 
2141 int
2142 linux_sched_rr_get_interval(struct thread *td,
2143     struct linux_sched_rr_get_interval_args *uap)
2144 {
2145 	struct timespec ts;
2146 	struct l_timespec lts;
2147 	struct thread *tdt;
2148 	int error;
2149 
2150 	/*
2151 	 * According to man in case the invalid pid specified
2152 	 * EINVAL should be returned.
2153 	 */
2154 	if (uap->pid < 0)
2155 		return (EINVAL);
2156 
2157 	tdt = linux_tdfind(td, uap->pid, -1);
2158 	if (tdt == NULL)
2159 		return (ESRCH);
2160 
2161 	error = kern_sched_rr_get_interval_td(td, tdt, &ts);
2162 	PROC_UNLOCK(tdt->td_proc);
2163 	if (error != 0)
2164 		return (error);
2165 	error = native_to_linux_timespec(&lts, &ts);
2166 	if (error != 0)
2167 		return (error);
2168 	return (copyout(&lts, uap->interval, sizeof(lts)));
2169 }
2170 
2171 /*
2172  * In case when the Linux thread is the initial thread in
2173  * the thread group thread id is equal to the process id.
2174  * Glibc depends on this magic (assert in pthread_getattr_np.c).
2175  */
2176 struct thread *
2177 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid)
2178 {
2179 	struct linux_emuldata *em;
2180 	struct thread *tdt;
2181 	struct proc *p;
2182 
2183 	tdt = NULL;
2184 	if (tid == 0 || tid == td->td_tid) {
2185 		tdt = td;
2186 		PROC_LOCK(tdt->td_proc);
2187 	} else if (tid > PID_MAX)
2188 		tdt = tdfind(tid, pid);
2189 	else {
2190 		/*
2191 		 * Initial thread where the tid equal to the pid.
2192 		 */
2193 		p = pfind(tid);
2194 		if (p != NULL) {
2195 			if (SV_PROC_ABI(p) != SV_ABI_LINUX) {
2196 				/*
2197 				 * p is not a Linuxulator process.
2198 				 */
2199 				PROC_UNLOCK(p);
2200 				return (NULL);
2201 			}
2202 			FOREACH_THREAD_IN_PROC(p, tdt) {
2203 				em = em_find(tdt);
2204 				if (tid == em->em_tid)
2205 					return (tdt);
2206 			}
2207 			PROC_UNLOCK(p);
2208 		}
2209 		return (NULL);
2210 	}
2211 
2212 	return (tdt);
2213 }
2214 
2215 void
2216 linux_to_bsd_waitopts(int options, int *bsdopts)
2217 {
2218 
2219 	if (options & LINUX_WNOHANG)
2220 		*bsdopts |= WNOHANG;
2221 	if (options & LINUX_WUNTRACED)
2222 		*bsdopts |= WUNTRACED;
2223 	if (options & LINUX_WEXITED)
2224 		*bsdopts |= WEXITED;
2225 	if (options & LINUX_WCONTINUED)
2226 		*bsdopts |= WCONTINUED;
2227 	if (options & LINUX_WNOWAIT)
2228 		*bsdopts |= WNOWAIT;
2229 
2230 	if (options & __WCLONE)
2231 		*bsdopts |= WLINUXCLONE;
2232 }
2233 
2234 int
2235 linux_getrandom(struct thread *td, struct linux_getrandom_args *args)
2236 {
2237 	struct uio uio;
2238 	struct iovec iov;
2239 	int error;
2240 
2241 	if (args->flags & ~(LINUX_GRND_NONBLOCK|LINUX_GRND_RANDOM))
2242 		return (EINVAL);
2243 	if (args->count > INT_MAX)
2244 		args->count = INT_MAX;
2245 
2246 	iov.iov_base = args->buf;
2247 	iov.iov_len = args->count;
2248 
2249 	uio.uio_iov = &iov;
2250 	uio.uio_iovcnt = 1;
2251 	uio.uio_resid = iov.iov_len;
2252 	uio.uio_segflg = UIO_USERSPACE;
2253 	uio.uio_rw = UIO_READ;
2254 	uio.uio_td = td;
2255 
2256 	error = read_random_uio(&uio, args->flags & LINUX_GRND_NONBLOCK);
2257 	if (error == 0)
2258 		td->td_retval[0] = args->count - uio.uio_resid;
2259 	return (error);
2260 }
2261 
2262 int
2263 linux_mincore(struct thread *td, struct linux_mincore_args *args)
2264 {
2265 
2266 	/* Needs to be page-aligned */
2267 	if (args->start & PAGE_MASK)
2268 		return (EINVAL);
2269 	return (kern_mincore(td, args->start, args->len, args->vec));
2270 }
2271