xref: /freebsd/sys/compat/linux/linux_misc.c (revision 6871d4882591c9a8fcab24d084c93f0a2972e1af)
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/reboot.h>
55 #include <sys/racct.h>
56 #include <sys/random.h>
57 #include <sys/resourcevar.h>
58 #include <sys/sched.h>
59 #include <sys/sdt.h>
60 #include <sys/signalvar.h>
61 #include <sys/stat.h>
62 #include <sys/syscallsubr.h>
63 #include <sys/sysctl.h>
64 #include <sys/sysproto.h>
65 #include <sys/systm.h>
66 #include <sys/time.h>
67 #include <sys/vmmeter.h>
68 #include <sys/vnode.h>
69 #include <sys/wait.h>
70 #include <sys/cpuset.h>
71 #include <sys/uio.h>
72 
73 #include <security/mac/mac_framework.h>
74 
75 #include <vm/vm.h>
76 #include <vm/pmap.h>
77 #include <vm/vm_kern.h>
78 #include <vm/vm_map.h>
79 #include <vm/vm_extern.h>
80 #include <vm/vm_object.h>
81 #include <vm/swap_pager.h>
82 
83 #ifdef COMPAT_LINUX32
84 #include <machine/../linux32/linux.h>
85 #include <machine/../linux32/linux32_proto.h>
86 #else
87 #include <machine/../linux/linux.h>
88 #include <machine/../linux/linux_proto.h>
89 #endif
90 
91 #include <compat/linux/linux_dtrace.h>
92 #include <compat/linux/linux_file.h>
93 #include <compat/linux/linux_mib.h>
94 #include <compat/linux/linux_signal.h>
95 #include <compat/linux/linux_timer.h>
96 #include <compat/linux/linux_util.h>
97 #include <compat/linux/linux_sysproto.h>
98 #include <compat/linux/linux_emul.h>
99 #include <compat/linux/linux_misc.h>
100 
101 /**
102  * Special DTrace provider for the linuxulator.
103  *
104  * In this file we define the provider for the entire linuxulator. All
105  * modules (= files of the linuxulator) use it.
106  *
107  * We define a different name depending on the emulated bitsize, see
108  * ../../<ARCH>/linux{,32}/linux.h, e.g.:
109  *      native bitsize          = linuxulator
110  *      amd64, 32bit emulation  = linuxulator32
111  */
112 LIN_SDT_PROVIDER_DEFINE(LINUX_DTRACE);
113 
114 int stclohz;				/* Statistics clock frequency */
115 
116 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
117 	RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
118 	RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
119 	RLIMIT_MEMLOCK, RLIMIT_AS
120 };
121 
122 struct l_sysinfo {
123 	l_long		uptime;		/* Seconds since boot */
124 	l_ulong		loads[3];	/* 1, 5, and 15 minute load averages */
125 #define LINUX_SYSINFO_LOADS_SCALE 65536
126 	l_ulong		totalram;	/* Total usable main memory size */
127 	l_ulong		freeram;	/* Available memory size */
128 	l_ulong		sharedram;	/* Amount of shared memory */
129 	l_ulong		bufferram;	/* Memory used by buffers */
130 	l_ulong		totalswap;	/* Total swap space size */
131 	l_ulong		freeswap;	/* swap space still available */
132 	l_ushort	procs;		/* Number of current processes */
133 	l_ushort	pads;
134 	l_ulong		totalbig;
135 	l_ulong		freebig;
136 	l_uint		mem_unit;
137 	char		_f[20-2*sizeof(l_long)-sizeof(l_int)];	/* padding */
138 };
139 
140 struct l_pselect6arg {
141 	l_uintptr_t	ss;
142 	l_size_t	ss_len;
143 };
144 
145 static int	linux_utimensat_nsec_valid(l_long);
146 
147 
148 int
149 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
150 {
151 	struct l_sysinfo sysinfo;
152 	vm_object_t object;
153 	int i, j;
154 	struct timespec ts;
155 
156 	bzero(&sysinfo, sizeof(sysinfo));
157 	getnanouptime(&ts);
158 	if (ts.tv_nsec != 0)
159 		ts.tv_sec++;
160 	sysinfo.uptime = ts.tv_sec;
161 
162 	/* Use the information from the mib to get our load averages */
163 	for (i = 0; i < 3; i++)
164 		sysinfo.loads[i] = averunnable.ldavg[i] *
165 		    LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
166 
167 	sysinfo.totalram = physmem * PAGE_SIZE;
168 	sysinfo.freeram = sysinfo.totalram - vm_wire_count() * PAGE_SIZE;
169 
170 	sysinfo.sharedram = 0;
171 	mtx_lock(&vm_object_list_mtx);
172 	TAILQ_FOREACH(object, &vm_object_list, object_list)
173 		if (object->shadow_count > 1)
174 			sysinfo.sharedram += object->resident_page_count;
175 	mtx_unlock(&vm_object_list_mtx);
176 
177 	sysinfo.sharedram *= PAGE_SIZE;
178 	sysinfo.bufferram = 0;
179 
180 	swap_pager_status(&i, &j);
181 	sysinfo.totalswap = i * PAGE_SIZE;
182 	sysinfo.freeswap = (i - j) * PAGE_SIZE;
183 
184 	sysinfo.procs = nprocs;
185 
186 	/* The following are only present in newer Linux kernels. */
187 	sysinfo.totalbig = 0;
188 	sysinfo.freebig = 0;
189 	sysinfo.mem_unit = 1;
190 
191 	return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
192 }
193 
194 #ifdef LINUX_LEGACY_SYSCALLS
195 int
196 linux_alarm(struct thread *td, struct linux_alarm_args *args)
197 {
198 	struct itimerval it, old_it;
199 	u_int secs;
200 	int error;
201 
202 #ifdef DEBUG
203 	if (ldebug(alarm))
204 		printf(ARGS(alarm, "%u"), args->secs);
205 #endif
206 	secs = args->secs;
207 	/*
208 	 * Linux alarm() is always successful. Limit secs to INT32_MAX / 2
209 	 * to match kern_setitimer()'s limit to avoid error from it.
210 	 *
211 	 * XXX. Linux limit secs to INT_MAX on 32 and does not limit on 64-bit
212 	 * platforms.
213 	 */
214 	if (secs > INT32_MAX / 2)
215 		secs = INT32_MAX / 2;
216 
217 	it.it_value.tv_sec = secs;
218 	it.it_value.tv_usec = 0;
219 	timevalclear(&it.it_interval);
220 	error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
221 	KASSERT(error == 0, ("kern_setitimer returns %d", error));
222 
223 	if ((old_it.it_value.tv_sec == 0 && old_it.it_value.tv_usec > 0) ||
224 	    old_it.it_value.tv_usec >= 500000)
225 		old_it.it_value.tv_sec++;
226 	td->td_retval[0] = old_it.it_value.tv_sec;
227 	return (0);
228 }
229 #endif
230 
231 int
232 linux_brk(struct thread *td, struct linux_brk_args *args)
233 {
234 	struct vmspace *vm = td->td_proc->p_vmspace;
235 	uintptr_t new, old;
236 
237 #ifdef DEBUG
238 	if (ldebug(brk))
239 		printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
240 #endif
241 	old = (uintptr_t)vm->vm_daddr + ctob(vm->vm_dsize);
242 	new = (uintptr_t)args->dsend;
243 	if ((caddr_t)new > vm->vm_daddr && !kern_break(td, &new))
244 		td->td_retval[0] = (register_t)new;
245 	else
246 		td->td_retval[0] = (register_t)old;
247 
248 	return (0);
249 }
250 
251 #if defined(__i386__)
252 /* XXX: what about amd64/linux32? */
253 
254 int
255 linux_uselib(struct thread *td, struct linux_uselib_args *args)
256 {
257 	struct nameidata ni;
258 	struct vnode *vp;
259 	struct exec *a_out;
260 	struct vattr attr;
261 	vm_offset_t vmaddr;
262 	unsigned long file_offset;
263 	unsigned long bss_size;
264 	char *library;
265 	ssize_t aresid;
266 	int error, locked, writecount;
267 
268 	LCONVPATHEXIST(td, args->library, &library);
269 
270 #ifdef DEBUG
271 	if (ldebug(uselib))
272 		printf(ARGS(uselib, "%s"), library);
273 #endif
274 
275 	a_out = NULL;
276 	locked = 0;
277 	vp = NULL;
278 
279 	NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
280 	    UIO_SYSSPACE, library, td);
281 	error = namei(&ni);
282 	LFREEPATH(library);
283 	if (error)
284 		goto cleanup;
285 
286 	vp = ni.ni_vp;
287 	NDFREE(&ni, NDF_ONLY_PNBUF);
288 
289 	/*
290 	 * From here on down, we have a locked vnode that must be unlocked.
291 	 * XXX: The code below largely duplicates exec_check_permissions().
292 	 */
293 	locked = 1;
294 
295 	/* Writable? */
296 	error = VOP_GET_WRITECOUNT(vp, &writecount);
297 	if (error != 0)
298 		goto cleanup;
299 	if (writecount != 0) {
300 		error = ETXTBSY;
301 		goto cleanup;
302 	}
303 
304 	/* Executable? */
305 	error = VOP_GETATTR(vp, &attr, td->td_ucred);
306 	if (error)
307 		goto cleanup;
308 
309 	if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
310 	    ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
311 		/* EACCESS is what exec(2) returns. */
312 		error = ENOEXEC;
313 		goto cleanup;
314 	}
315 
316 	/* Sensible size? */
317 	if (attr.va_size == 0) {
318 		error = ENOEXEC;
319 		goto cleanup;
320 	}
321 
322 	/* Can we access it? */
323 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
324 	if (error)
325 		goto cleanup;
326 
327 	/*
328 	 * XXX: This should use vn_open() so that it is properly authorized,
329 	 * and to reduce code redundancy all over the place here.
330 	 * XXX: Not really, it duplicates far more of exec_check_permissions()
331 	 * than vn_open().
332 	 */
333 #ifdef MAC
334 	error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
335 	if (error)
336 		goto cleanup;
337 #endif
338 	error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
339 	if (error)
340 		goto cleanup;
341 
342 	/* Pull in executable header into exec_map */
343 	error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
344 	    VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
345 	if (error)
346 		goto cleanup;
347 
348 	/* Is it a Linux binary ? */
349 	if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
350 		error = ENOEXEC;
351 		goto cleanup;
352 	}
353 
354 	/*
355 	 * While we are here, we should REALLY do some more checks
356 	 */
357 
358 	/* Set file/virtual offset based on a.out variant. */
359 	switch ((int)(a_out->a_magic & 0xffff)) {
360 	case 0413:			/* ZMAGIC */
361 		file_offset = 1024;
362 		break;
363 	case 0314:			/* QMAGIC */
364 		file_offset = 0;
365 		break;
366 	default:
367 		error = ENOEXEC;
368 		goto cleanup;
369 	}
370 
371 	bss_size = round_page(a_out->a_bss);
372 
373 	/* Check various fields in header for validity/bounds. */
374 	if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
375 		error = ENOEXEC;
376 		goto cleanup;
377 	}
378 
379 	/* text + data can't exceed file size */
380 	if (a_out->a_data + a_out->a_text > attr.va_size) {
381 		error = EFAULT;
382 		goto cleanup;
383 	}
384 
385 	/*
386 	 * text/data/bss must not exceed limits
387 	 * XXX - this is not complete. it should check current usage PLUS
388 	 * the resources needed by this library.
389 	 */
390 	PROC_LOCK(td->td_proc);
391 	if (a_out->a_text > maxtsiz ||
392 	    a_out->a_data + bss_size > lim_cur_proc(td->td_proc, RLIMIT_DATA) ||
393 	    racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
394 	    bss_size) != 0) {
395 		PROC_UNLOCK(td->td_proc);
396 		error = ENOMEM;
397 		goto cleanup;
398 	}
399 	PROC_UNLOCK(td->td_proc);
400 
401 	/*
402 	 * Prevent more writers.
403 	 * XXX: Note that if any of the VM operations fail below we don't
404 	 * clear this flag.
405 	 */
406 	VOP_SET_TEXT(vp);
407 
408 	/*
409 	 * Lock no longer needed
410 	 */
411 	locked = 0;
412 	VOP_UNLOCK(vp, 0);
413 
414 	/*
415 	 * Check if file_offset page aligned. Currently we cannot handle
416 	 * misalinged file offsets, and so we read in the entire image
417 	 * (what a waste).
418 	 */
419 	if (file_offset & PAGE_MASK) {
420 #ifdef DEBUG
421 		printf("uselib: Non page aligned binary %lu\n", file_offset);
422 #endif
423 		/* Map text+data read/write/execute */
424 
425 		/* a_entry is the load address and is page aligned */
426 		vmaddr = trunc_page(a_out->a_entry);
427 
428 		/* get anon user mapping, read+write+execute */
429 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
430 		    &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
431 		    VM_PROT_ALL, VM_PROT_ALL, 0);
432 		if (error)
433 			goto cleanup;
434 
435 		error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
436 		    a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
437 		    td->td_ucred, NOCRED, &aresid, td);
438 		if (error != 0)
439 			goto cleanup;
440 		if (aresid != 0) {
441 			error = ENOEXEC;
442 			goto cleanup;
443 		}
444 	} else {
445 #ifdef DEBUG
446 		printf("uselib: Page aligned binary %lu\n", file_offset);
447 #endif
448 		/*
449 		 * for QMAGIC, a_entry is 20 bytes beyond the load address
450 		 * to skip the executable header
451 		 */
452 		vmaddr = trunc_page(a_out->a_entry);
453 
454 		/*
455 		 * Map it all into the process's space as a single
456 		 * copy-on-write "data" segment.
457 		 */
458 		error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr,
459 		    a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
460 		    MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
461 		if (error)
462 			goto cleanup;
463 	}
464 #ifdef DEBUG
465 	printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
466 	    ((long *)vmaddr)[1]);
467 #endif
468 	if (bss_size != 0) {
469 		/* Calculate BSS start address */
470 		vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
471 		    a_out->a_data;
472 
473 		/* allocate some 'anon' space */
474 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
475 		    &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
476 		    VM_PROT_ALL, 0);
477 		if (error)
478 			goto cleanup;
479 	}
480 
481 cleanup:
482 	/* Unlock vnode if needed */
483 	if (locked)
484 		VOP_UNLOCK(vp, 0);
485 
486 	/* Release the temporary mapping. */
487 	if (a_out)
488 		kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
489 
490 	return (error);
491 }
492 
493 #endif	/* __i386__ */
494 
495 #ifdef LINUX_LEGACY_SYSCALLS
496 int
497 linux_select(struct thread *td, struct linux_select_args *args)
498 {
499 	l_timeval ltv;
500 	struct timeval tv0, tv1, utv, *tvp;
501 	int error;
502 
503 #ifdef DEBUG
504 	if (ldebug(select))
505 		printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
506 		    (void *)args->readfds, (void *)args->writefds,
507 		    (void *)args->exceptfds, (void *)args->timeout);
508 #endif
509 
510 	/*
511 	 * Store current time for computation of the amount of
512 	 * time left.
513 	 */
514 	if (args->timeout) {
515 		if ((error = copyin(args->timeout, &ltv, sizeof(ltv))))
516 			goto select_out;
517 		utv.tv_sec = ltv.tv_sec;
518 		utv.tv_usec = ltv.tv_usec;
519 #ifdef DEBUG
520 		if (ldebug(select))
521 			printf(LMSG("incoming timeout (%jd/%ld)"),
522 			    (intmax_t)utv.tv_sec, utv.tv_usec);
523 #endif
524 
525 		if (itimerfix(&utv)) {
526 			/*
527 			 * The timeval was invalid.  Convert it to something
528 			 * valid that will act as it does under Linux.
529 			 */
530 			utv.tv_sec += utv.tv_usec / 1000000;
531 			utv.tv_usec %= 1000000;
532 			if (utv.tv_usec < 0) {
533 				utv.tv_sec -= 1;
534 				utv.tv_usec += 1000000;
535 			}
536 			if (utv.tv_sec < 0)
537 				timevalclear(&utv);
538 		}
539 		microtime(&tv0);
540 		tvp = &utv;
541 	} else
542 		tvp = NULL;
543 
544 	error = kern_select(td, args->nfds, args->readfds, args->writefds,
545 	    args->exceptfds, tvp, LINUX_NFDBITS);
546 
547 #ifdef DEBUG
548 	if (ldebug(select))
549 		printf(LMSG("real select returns %d"), error);
550 #endif
551 	if (error)
552 		goto select_out;
553 
554 	if (args->timeout) {
555 		if (td->td_retval[0]) {
556 			/*
557 			 * Compute how much time was left of the timeout,
558 			 * by subtracting the current time and the time
559 			 * before we started the call, and subtracting
560 			 * that result from the user-supplied value.
561 			 */
562 			microtime(&tv1);
563 			timevalsub(&tv1, &tv0);
564 			timevalsub(&utv, &tv1);
565 			if (utv.tv_sec < 0)
566 				timevalclear(&utv);
567 		} else
568 			timevalclear(&utv);
569 #ifdef DEBUG
570 		if (ldebug(select))
571 			printf(LMSG("outgoing timeout (%jd/%ld)"),
572 			    (intmax_t)utv.tv_sec, utv.tv_usec);
573 #endif
574 		ltv.tv_sec = utv.tv_sec;
575 		ltv.tv_usec = utv.tv_usec;
576 		if ((error = copyout(&ltv, args->timeout, sizeof(ltv))))
577 			goto select_out;
578 	}
579 
580 select_out:
581 #ifdef DEBUG
582 	if (ldebug(select))
583 		printf(LMSG("select_out -> %d"), error);
584 #endif
585 	return (error);
586 }
587 #endif
588 
589 int
590 linux_mremap(struct thread *td, struct linux_mremap_args *args)
591 {
592 	uintptr_t addr;
593 	size_t len;
594 	int error = 0;
595 
596 #ifdef DEBUG
597 	if (ldebug(mremap))
598 		printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
599 		    (void *)(uintptr_t)args->addr,
600 		    (unsigned long)args->old_len,
601 		    (unsigned long)args->new_len,
602 		    (unsigned long)args->flags);
603 #endif
604 
605 	if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
606 		td->td_retval[0] = 0;
607 		return (EINVAL);
608 	}
609 
610 	/*
611 	 * Check for the page alignment.
612 	 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
613 	 */
614 	if (args->addr & PAGE_MASK) {
615 		td->td_retval[0] = 0;
616 		return (EINVAL);
617 	}
618 
619 	args->new_len = round_page(args->new_len);
620 	args->old_len = round_page(args->old_len);
621 
622 	if (args->new_len > args->old_len) {
623 		td->td_retval[0] = 0;
624 		return (ENOMEM);
625 	}
626 
627 	if (args->new_len < args->old_len) {
628 		addr = args->addr + args->new_len;
629 		len = args->old_len - args->new_len;
630 		error = kern_munmap(td, addr, len);
631 	}
632 
633 	td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
634 	return (error);
635 }
636 
637 #define LINUX_MS_ASYNC       0x0001
638 #define LINUX_MS_INVALIDATE  0x0002
639 #define LINUX_MS_SYNC        0x0004
640 
641 int
642 linux_msync(struct thread *td, struct linux_msync_args *args)
643 {
644 
645 	return (kern_msync(td, args->addr, args->len,
646 	    args->fl & ~LINUX_MS_SYNC));
647 }
648 
649 #ifdef LINUX_LEGACY_SYSCALLS
650 int
651 linux_time(struct thread *td, struct linux_time_args *args)
652 {
653 	struct timeval tv;
654 	l_time_t tm;
655 	int error;
656 
657 #ifdef DEBUG
658 	if (ldebug(time))
659 		printf(ARGS(time, "*"));
660 #endif
661 
662 	microtime(&tv);
663 	tm = tv.tv_sec;
664 	if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
665 		return (error);
666 	td->td_retval[0] = tm;
667 	return (0);
668 }
669 #endif
670 
671 struct l_times_argv {
672 	l_clock_t	tms_utime;
673 	l_clock_t	tms_stime;
674 	l_clock_t	tms_cutime;
675 	l_clock_t	tms_cstime;
676 };
677 
678 
679 /*
680  * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
681  * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
682  * auxiliary vector entry.
683  */
684 #define	CLK_TCK		100
685 
686 #define	CONVOTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
687 #define	CONVNTCK(r)	(r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
688 
689 #define	CONVTCK(r)	(linux_kernver(td) >= LINUX_KERNVER_2004000 ?		\
690 			    CONVNTCK(r) : CONVOTCK(r))
691 
692 int
693 linux_times(struct thread *td, struct linux_times_args *args)
694 {
695 	struct timeval tv, utime, stime, cutime, cstime;
696 	struct l_times_argv tms;
697 	struct proc *p;
698 	int error;
699 
700 #ifdef DEBUG
701 	if (ldebug(times))
702 		printf(ARGS(times, "*"));
703 #endif
704 
705 	if (args->buf != NULL) {
706 		p = td->td_proc;
707 		PROC_LOCK(p);
708 		PROC_STATLOCK(p);
709 		calcru(p, &utime, &stime);
710 		PROC_STATUNLOCK(p);
711 		calccru(p, &cutime, &cstime);
712 		PROC_UNLOCK(p);
713 
714 		tms.tms_utime = CONVTCK(utime);
715 		tms.tms_stime = CONVTCK(stime);
716 
717 		tms.tms_cutime = CONVTCK(cutime);
718 		tms.tms_cstime = CONVTCK(cstime);
719 
720 		if ((error = copyout(&tms, args->buf, sizeof(tms))))
721 			return (error);
722 	}
723 
724 	microuptime(&tv);
725 	td->td_retval[0] = (int)CONVTCK(tv);
726 	return (0);
727 }
728 
729 int
730 linux_newuname(struct thread *td, struct linux_newuname_args *args)
731 {
732 	struct l_new_utsname utsname;
733 	char osname[LINUX_MAX_UTSNAME];
734 	char osrelease[LINUX_MAX_UTSNAME];
735 	char *p;
736 
737 #ifdef DEBUG
738 	if (ldebug(newuname))
739 		printf(ARGS(newuname, "*"));
740 #endif
741 
742 	linux_get_osname(td, osname);
743 	linux_get_osrelease(td, osrelease);
744 
745 	bzero(&utsname, sizeof(utsname));
746 	strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
747 	getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
748 	getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
749 	strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
750 	strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
751 	for (p = utsname.version; *p != '\0'; ++p)
752 		if (*p == '\n') {
753 			*p = '\0';
754 			break;
755 		}
756 	strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME);
757 
758 	return (copyout(&utsname, args->buf, sizeof(utsname)));
759 }
760 
761 struct l_utimbuf {
762 	l_time_t l_actime;
763 	l_time_t l_modtime;
764 };
765 
766 #ifdef LINUX_LEGACY_SYSCALLS
767 int
768 linux_utime(struct thread *td, struct linux_utime_args *args)
769 {
770 	struct timeval tv[2], *tvp;
771 	struct l_utimbuf lut;
772 	char *fname;
773 	int error;
774 
775 	LCONVPATHEXIST(td, args->fname, &fname);
776 
777 #ifdef DEBUG
778 	if (ldebug(utime))
779 		printf(ARGS(utime, "%s, *"), fname);
780 #endif
781 
782 	if (args->times) {
783 		if ((error = copyin(args->times, &lut, sizeof lut))) {
784 			LFREEPATH(fname);
785 			return (error);
786 		}
787 		tv[0].tv_sec = lut.l_actime;
788 		tv[0].tv_usec = 0;
789 		tv[1].tv_sec = lut.l_modtime;
790 		tv[1].tv_usec = 0;
791 		tvp = tv;
792 	} else
793 		tvp = NULL;
794 
795 	error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, tvp,
796 	    UIO_SYSSPACE);
797 	LFREEPATH(fname);
798 	return (error);
799 }
800 #endif
801 
802 #ifdef LINUX_LEGACY_SYSCALLS
803 int
804 linux_utimes(struct thread *td, struct linux_utimes_args *args)
805 {
806 	l_timeval ltv[2];
807 	struct timeval tv[2], *tvp = NULL;
808 	char *fname;
809 	int error;
810 
811 	LCONVPATHEXIST(td, args->fname, &fname);
812 
813 #ifdef DEBUG
814 	if (ldebug(utimes))
815 		printf(ARGS(utimes, "%s, *"), fname);
816 #endif
817 
818 	if (args->tptr != NULL) {
819 		if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
820 			LFREEPATH(fname);
821 			return (error);
822 		}
823 		tv[0].tv_sec = ltv[0].tv_sec;
824 		tv[0].tv_usec = ltv[0].tv_usec;
825 		tv[1].tv_sec = ltv[1].tv_sec;
826 		tv[1].tv_usec = ltv[1].tv_usec;
827 		tvp = tv;
828 	}
829 
830 	error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE,
831 	    tvp, UIO_SYSSPACE);
832 	LFREEPATH(fname);
833 	return (error);
834 }
835 #endif
836 
837 static int
838 linux_utimensat_nsec_valid(l_long nsec)
839 {
840 
841 	if (nsec == LINUX_UTIME_OMIT || nsec == LINUX_UTIME_NOW)
842 		return (0);
843 	if (nsec >= 0 && nsec <= 999999999)
844 		return (0);
845 	return (1);
846 }
847 
848 int
849 linux_utimensat(struct thread *td, struct linux_utimensat_args *args)
850 {
851 	struct l_timespec l_times[2];
852 	struct timespec times[2], *timesp = NULL;
853 	char *path = NULL;
854 	int error, dfd, flags = 0;
855 
856 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
857 
858 #ifdef DEBUG
859 	if (ldebug(utimensat))
860 		printf(ARGS(utimensat, "%d, *"), dfd);
861 #endif
862 
863 	if (args->flags & ~LINUX_AT_SYMLINK_NOFOLLOW)
864 		return (EINVAL);
865 
866 	if (args->times != NULL) {
867 		error = copyin(args->times, l_times, sizeof(l_times));
868 		if (error != 0)
869 			return (error);
870 
871 		if (linux_utimensat_nsec_valid(l_times[0].tv_nsec) != 0 ||
872 		    linux_utimensat_nsec_valid(l_times[1].tv_nsec) != 0)
873 			return (EINVAL);
874 
875 		times[0].tv_sec = l_times[0].tv_sec;
876 		switch (l_times[0].tv_nsec)
877 		{
878 		case LINUX_UTIME_OMIT:
879 			times[0].tv_nsec = UTIME_OMIT;
880 			break;
881 		case LINUX_UTIME_NOW:
882 			times[0].tv_nsec = UTIME_NOW;
883 			break;
884 		default:
885 			times[0].tv_nsec = l_times[0].tv_nsec;
886 		}
887 
888 		times[1].tv_sec = l_times[1].tv_sec;
889 		switch (l_times[1].tv_nsec)
890 		{
891 		case LINUX_UTIME_OMIT:
892 			times[1].tv_nsec = UTIME_OMIT;
893 			break;
894 		case LINUX_UTIME_NOW:
895 			times[1].tv_nsec = UTIME_NOW;
896 			break;
897 		default:
898 			times[1].tv_nsec = l_times[1].tv_nsec;
899 			break;
900 		}
901 		timesp = times;
902 
903 		/* This breaks POSIX, but is what the Linux kernel does
904 		 * _on purpose_ (documented in the man page for utimensat(2)),
905 		 * so we must follow that behaviour. */
906 		if (times[0].tv_nsec == UTIME_OMIT &&
907 		    times[1].tv_nsec == UTIME_OMIT)
908 			return (0);
909 	}
910 
911 	if (args->pathname != NULL)
912 		LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
913 	else if (args->flags != 0)
914 		return (EINVAL);
915 
916 	if (args->flags & LINUX_AT_SYMLINK_NOFOLLOW)
917 		flags |= AT_SYMLINK_NOFOLLOW;
918 
919 	if (path == NULL)
920 		error = kern_futimens(td, dfd, timesp, UIO_SYSSPACE);
921 	else {
922 		error = kern_utimensat(td, dfd, path, UIO_SYSSPACE, timesp,
923 			UIO_SYSSPACE, flags);
924 		LFREEPATH(path);
925 	}
926 
927 	return (error);
928 }
929 
930 #ifdef LINUX_LEGACY_SYSCALLS
931 int
932 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
933 {
934 	l_timeval ltv[2];
935 	struct timeval tv[2], *tvp = NULL;
936 	char *fname;
937 	int error, dfd;
938 
939 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
940 	LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
941 
942 #ifdef DEBUG
943 	if (ldebug(futimesat))
944 		printf(ARGS(futimesat, "%s, *"), fname);
945 #endif
946 
947 	if (args->utimes != NULL) {
948 		if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
949 			LFREEPATH(fname);
950 			return (error);
951 		}
952 		tv[0].tv_sec = ltv[0].tv_sec;
953 		tv[0].tv_usec = ltv[0].tv_usec;
954 		tv[1].tv_sec = ltv[1].tv_sec;
955 		tv[1].tv_usec = ltv[1].tv_usec;
956 		tvp = tv;
957 	}
958 
959 	error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
960 	LFREEPATH(fname);
961 	return (error);
962 }
963 #endif
964 
965 int
966 linux_common_wait(struct thread *td, int pid, int *status,
967     int options, struct rusage *ru)
968 {
969 	int error, tmpstat;
970 
971 	error = kern_wait(td, pid, &tmpstat, options, ru);
972 	if (error)
973 		return (error);
974 
975 	if (status) {
976 		tmpstat &= 0xffff;
977 		if (WIFSIGNALED(tmpstat))
978 			tmpstat = (tmpstat & 0xffffff80) |
979 			    bsd_to_linux_signal(WTERMSIG(tmpstat));
980 		else if (WIFSTOPPED(tmpstat))
981 			tmpstat = (tmpstat & 0xffff00ff) |
982 			    (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8);
983 		else if (WIFCONTINUED(tmpstat))
984 			tmpstat = 0xffff;
985 		error = copyout(&tmpstat, status, sizeof(int));
986 	}
987 
988 	return (error);
989 }
990 
991 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
992 int
993 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
994 {
995 	struct linux_wait4_args wait4_args;
996 
997 #ifdef DEBUG
998 	if (ldebug(waitpid))
999 		printf(ARGS(waitpid, "%d, %p, %d"),
1000 		    args->pid, (void *)args->status, args->options);
1001 #endif
1002 
1003 	wait4_args.pid = args->pid;
1004 	wait4_args.status = args->status;
1005 	wait4_args.options = args->options;
1006 	wait4_args.rusage = NULL;
1007 
1008 	return (linux_wait4(td, &wait4_args));
1009 }
1010 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1011 
1012 int
1013 linux_wait4(struct thread *td, struct linux_wait4_args *args)
1014 {
1015 	int error, options;
1016 	struct rusage ru, *rup;
1017 
1018 #ifdef DEBUG
1019 	if (ldebug(wait4))
1020 		printf(ARGS(wait4, "%d, %p, %d, %p"),
1021 		    args->pid, (void *)args->status, args->options,
1022 		    (void *)args->rusage);
1023 #endif
1024 	if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG |
1025 	    LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
1026 		return (EINVAL);
1027 
1028 	options = WEXITED;
1029 	linux_to_bsd_waitopts(args->options, &options);
1030 
1031 	if (args->rusage != NULL)
1032 		rup = &ru;
1033 	else
1034 		rup = NULL;
1035 	error = linux_common_wait(td, args->pid, args->status, options, rup);
1036 	if (error != 0)
1037 		return (error);
1038 	if (args->rusage != NULL)
1039 		error = linux_copyout_rusage(&ru, args->rusage);
1040 	return (error);
1041 }
1042 
1043 int
1044 linux_waitid(struct thread *td, struct linux_waitid_args *args)
1045 {
1046 	int status, options, sig;
1047 	struct __wrusage wru;
1048 	siginfo_t siginfo;
1049 	l_siginfo_t lsi;
1050 	idtype_t idtype;
1051 	struct proc *p;
1052 	int error;
1053 
1054 	options = 0;
1055 	linux_to_bsd_waitopts(args->options, &options);
1056 
1057 	if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED))
1058 		return (EINVAL);
1059 	if (!(options & (WEXITED | WUNTRACED | WCONTINUED)))
1060 		return (EINVAL);
1061 
1062 	switch (args->idtype) {
1063 	case LINUX_P_ALL:
1064 		idtype = P_ALL;
1065 		break;
1066 	case LINUX_P_PID:
1067 		if (args->id <= 0)
1068 			return (EINVAL);
1069 		idtype = P_PID;
1070 		break;
1071 	case LINUX_P_PGID:
1072 		if (args->id <= 0)
1073 			return (EINVAL);
1074 		idtype = P_PGID;
1075 		break;
1076 	default:
1077 		return (EINVAL);
1078 	}
1079 
1080 	error = kern_wait6(td, idtype, args->id, &status, options,
1081 	    &wru, &siginfo);
1082 	if (error != 0)
1083 		return (error);
1084 	if (args->rusage != NULL) {
1085 		error = linux_copyout_rusage(&wru.wru_children,
1086 		    args->rusage);
1087 		if (error != 0)
1088 			return (error);
1089 	}
1090 	if (args->info != NULL) {
1091 		p = td->td_proc;
1092 		if (td->td_retval[0] == 0)
1093 			bzero(&lsi, sizeof(lsi));
1094 		else {
1095 			sig = bsd_to_linux_signal(siginfo.si_signo);
1096 			siginfo_to_lsiginfo(&siginfo, &lsi, sig);
1097 		}
1098 		error = copyout(&lsi, args->info, sizeof(lsi));
1099 	}
1100 	td->td_retval[0] = 0;
1101 
1102 	return (error);
1103 }
1104 
1105 #ifdef LINUX_LEGACY_SYSCALLS
1106 int
1107 linux_mknod(struct thread *td, struct linux_mknod_args *args)
1108 {
1109 	char *path;
1110 	int error;
1111 
1112 	LCONVPATHCREAT(td, args->path, &path);
1113 
1114 #ifdef DEBUG
1115 	if (ldebug(mknod))
1116 		printf(ARGS(mknod, "%s, %d, %ju"), path, args->mode,
1117 		    (uintmax_t)args->dev);
1118 #endif
1119 
1120 	switch (args->mode & S_IFMT) {
1121 	case S_IFIFO:
1122 	case S_IFSOCK:
1123 		error = kern_mkfifoat(td, AT_FDCWD, path, UIO_SYSSPACE,
1124 		    args->mode);
1125 		break;
1126 
1127 	case S_IFCHR:
1128 	case S_IFBLK:
1129 		error = kern_mknodat(td, AT_FDCWD, path, UIO_SYSSPACE,
1130 		    args->mode, args->dev);
1131 		break;
1132 
1133 	case S_IFDIR:
1134 		error = EPERM;
1135 		break;
1136 
1137 	case 0:
1138 		args->mode |= S_IFREG;
1139 		/* FALLTHROUGH */
1140 	case S_IFREG:
1141 		error = kern_openat(td, AT_FDCWD, path, UIO_SYSSPACE,
1142 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1143 		if (error == 0)
1144 			kern_close(td, td->td_retval[0]);
1145 		break;
1146 
1147 	default:
1148 		error = EINVAL;
1149 		break;
1150 	}
1151 	LFREEPATH(path);
1152 	return (error);
1153 }
1154 #endif
1155 
1156 int
1157 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
1158 {
1159 	char *path;
1160 	int error, dfd;
1161 
1162 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
1163 	LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
1164 
1165 #ifdef DEBUG
1166 	if (ldebug(mknodat))
1167 		printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev);
1168 #endif
1169 
1170 	switch (args->mode & S_IFMT) {
1171 	case S_IFIFO:
1172 	case S_IFSOCK:
1173 		error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
1174 		break;
1175 
1176 	case S_IFCHR:
1177 	case S_IFBLK:
1178 		error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
1179 		    args->dev);
1180 		break;
1181 
1182 	case S_IFDIR:
1183 		error = EPERM;
1184 		break;
1185 
1186 	case 0:
1187 		args->mode |= S_IFREG;
1188 		/* FALLTHROUGH */
1189 	case S_IFREG:
1190 		error = kern_openat(td, dfd, path, UIO_SYSSPACE,
1191 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1192 		if (error == 0)
1193 			kern_close(td, td->td_retval[0]);
1194 		break;
1195 
1196 	default:
1197 		error = EINVAL;
1198 		break;
1199 	}
1200 	LFREEPATH(path);
1201 	return (error);
1202 }
1203 
1204 /*
1205  * UGH! This is just about the dumbest idea I've ever heard!!
1206  */
1207 int
1208 linux_personality(struct thread *td, struct linux_personality_args *args)
1209 {
1210 	struct linux_pemuldata *pem;
1211 	struct proc *p = td->td_proc;
1212 	uint32_t old;
1213 
1214 #ifdef DEBUG
1215 	if (ldebug(personality))
1216 		printf(ARGS(personality, "%u"), args->per);
1217 #endif
1218 
1219 	PROC_LOCK(p);
1220 	pem = pem_find(p);
1221 	old = pem->persona;
1222 	if (args->per != 0xffffffff)
1223 		pem->persona = args->per;
1224 	PROC_UNLOCK(p);
1225 
1226 	td->td_retval[0] = old;
1227 	return (0);
1228 }
1229 
1230 struct l_itimerval {
1231 	l_timeval it_interval;
1232 	l_timeval it_value;
1233 };
1234 
1235 #define	B2L_ITIMERVAL(bip, lip)						\
1236 	(bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec;		\
1237 	(bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec;	\
1238 	(bip)->it_value.tv_sec = (lip)->it_value.tv_sec;		\
1239 	(bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1240 
1241 int
1242 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1243 {
1244 	int error;
1245 	struct l_itimerval ls;
1246 	struct itimerval aitv, oitv;
1247 
1248 #ifdef DEBUG
1249 	if (ldebug(setitimer))
1250 		printf(ARGS(setitimer, "%p, %p"),
1251 		    (void *)uap->itv, (void *)uap->oitv);
1252 #endif
1253 
1254 	if (uap->itv == NULL) {
1255 		uap->itv = uap->oitv;
1256 		return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1257 	}
1258 
1259 	error = copyin(uap->itv, &ls, sizeof(ls));
1260 	if (error != 0)
1261 		return (error);
1262 	B2L_ITIMERVAL(&aitv, &ls);
1263 #ifdef DEBUG
1264 	if (ldebug(setitimer)) {
1265 		printf("setitimer: value: sec: %jd, usec: %ld\n",
1266 		    (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1267 		printf("setitimer: interval: sec: %jd, usec: %ld\n",
1268 		    (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1269 	}
1270 #endif
1271 	error = kern_setitimer(td, uap->which, &aitv, &oitv);
1272 	if (error != 0 || uap->oitv == NULL)
1273 		return (error);
1274 	B2L_ITIMERVAL(&ls, &oitv);
1275 
1276 	return (copyout(&ls, uap->oitv, sizeof(ls)));
1277 }
1278 
1279 int
1280 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1281 {
1282 	int error;
1283 	struct l_itimerval ls;
1284 	struct itimerval aitv;
1285 
1286 #ifdef DEBUG
1287 	if (ldebug(getitimer))
1288 		printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1289 #endif
1290 	error = kern_getitimer(td, uap->which, &aitv);
1291 	if (error != 0)
1292 		return (error);
1293 	B2L_ITIMERVAL(&ls, &aitv);
1294 	return (copyout(&ls, uap->itv, sizeof(ls)));
1295 }
1296 
1297 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1298 int
1299 linux_nice(struct thread *td, struct linux_nice_args *args)
1300 {
1301 	struct setpriority_args bsd_args;
1302 
1303 	bsd_args.which = PRIO_PROCESS;
1304 	bsd_args.who = 0;		/* current process */
1305 	bsd_args.prio = args->inc;
1306 	return (sys_setpriority(td, &bsd_args));
1307 }
1308 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1309 
1310 int
1311 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1312 {
1313 	struct ucred *newcred, *oldcred;
1314 	l_gid_t *linux_gidset;
1315 	gid_t *bsd_gidset;
1316 	int ngrp, error;
1317 	struct proc *p;
1318 
1319 	ngrp = args->gidsetsize;
1320 	if (ngrp < 0 || ngrp >= ngroups_max + 1)
1321 		return (EINVAL);
1322 	linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK);
1323 	error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1324 	if (error)
1325 		goto out;
1326 	newcred = crget();
1327 	crextend(newcred, ngrp + 1);
1328 	p = td->td_proc;
1329 	PROC_LOCK(p);
1330 	oldcred = p->p_ucred;
1331 	crcopy(newcred, oldcred);
1332 
1333 	/*
1334 	 * cr_groups[0] holds egid. Setting the whole set from
1335 	 * the supplied set will cause egid to be changed too.
1336 	 * Keep cr_groups[0] unchanged to prevent that.
1337 	 */
1338 
1339 	if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS)) != 0) {
1340 		PROC_UNLOCK(p);
1341 		crfree(newcred);
1342 		goto out;
1343 	}
1344 
1345 	if (ngrp > 0) {
1346 		newcred->cr_ngroups = ngrp + 1;
1347 
1348 		bsd_gidset = newcred->cr_groups;
1349 		ngrp--;
1350 		while (ngrp >= 0) {
1351 			bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1352 			ngrp--;
1353 		}
1354 	} else
1355 		newcred->cr_ngroups = 1;
1356 
1357 	setsugid(p);
1358 	proc_set_cred(p, newcred);
1359 	PROC_UNLOCK(p);
1360 	crfree(oldcred);
1361 	error = 0;
1362 out:
1363 	free(linux_gidset, M_LINUX);
1364 	return (error);
1365 }
1366 
1367 int
1368 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1369 {
1370 	struct ucred *cred;
1371 	l_gid_t *linux_gidset;
1372 	gid_t *bsd_gidset;
1373 	int bsd_gidsetsz, ngrp, error;
1374 
1375 	cred = td->td_ucred;
1376 	bsd_gidset = cred->cr_groups;
1377 	bsd_gidsetsz = cred->cr_ngroups - 1;
1378 
1379 	/*
1380 	 * cr_groups[0] holds egid. Returning the whole set
1381 	 * here will cause a duplicate. Exclude cr_groups[0]
1382 	 * to prevent that.
1383 	 */
1384 
1385 	if ((ngrp = args->gidsetsize) == 0) {
1386 		td->td_retval[0] = bsd_gidsetsz;
1387 		return (0);
1388 	}
1389 
1390 	if (ngrp < bsd_gidsetsz)
1391 		return (EINVAL);
1392 
1393 	ngrp = 0;
1394 	linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1395 	    M_LINUX, M_WAITOK);
1396 	while (ngrp < bsd_gidsetsz) {
1397 		linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1398 		ngrp++;
1399 	}
1400 
1401 	error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1402 	free(linux_gidset, M_LINUX);
1403 	if (error)
1404 		return (error);
1405 
1406 	td->td_retval[0] = ngrp;
1407 	return (0);
1408 }
1409 
1410 int
1411 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1412 {
1413 	struct rlimit bsd_rlim;
1414 	struct l_rlimit rlim;
1415 	u_int which;
1416 	int error;
1417 
1418 #ifdef DEBUG
1419 	if (ldebug(setrlimit))
1420 		printf(ARGS(setrlimit, "%d, %p"),
1421 		    args->resource, (void *)args->rlim);
1422 #endif
1423 
1424 	if (args->resource >= LINUX_RLIM_NLIMITS)
1425 		return (EINVAL);
1426 
1427 	which = linux_to_bsd_resource[args->resource];
1428 	if (which == -1)
1429 		return (EINVAL);
1430 
1431 	error = copyin(args->rlim, &rlim, sizeof(rlim));
1432 	if (error)
1433 		return (error);
1434 
1435 	bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1436 	bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1437 	return (kern_setrlimit(td, which, &bsd_rlim));
1438 }
1439 
1440 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1441 int
1442 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1443 {
1444 	struct l_rlimit rlim;
1445 	struct rlimit bsd_rlim;
1446 	u_int which;
1447 
1448 #ifdef DEBUG
1449 	if (ldebug(old_getrlimit))
1450 		printf(ARGS(old_getrlimit, "%d, %p"),
1451 		    args->resource, (void *)args->rlim);
1452 #endif
1453 
1454 	if (args->resource >= LINUX_RLIM_NLIMITS)
1455 		return (EINVAL);
1456 
1457 	which = linux_to_bsd_resource[args->resource];
1458 	if (which == -1)
1459 		return (EINVAL);
1460 
1461 	lim_rlimit(td, which, &bsd_rlim);
1462 
1463 #ifdef COMPAT_LINUX32
1464 	rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1465 	if (rlim.rlim_cur == UINT_MAX)
1466 		rlim.rlim_cur = INT_MAX;
1467 	rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1468 	if (rlim.rlim_max == UINT_MAX)
1469 		rlim.rlim_max = INT_MAX;
1470 #else
1471 	rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1472 	if (rlim.rlim_cur == ULONG_MAX)
1473 		rlim.rlim_cur = LONG_MAX;
1474 	rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1475 	if (rlim.rlim_max == ULONG_MAX)
1476 		rlim.rlim_max = LONG_MAX;
1477 #endif
1478 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1479 }
1480 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1481 
1482 int
1483 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1484 {
1485 	struct l_rlimit rlim;
1486 	struct rlimit bsd_rlim;
1487 	u_int which;
1488 
1489 #ifdef DEBUG
1490 	if (ldebug(getrlimit))
1491 		printf(ARGS(getrlimit, "%d, %p"),
1492 		    args->resource, (void *)args->rlim);
1493 #endif
1494 
1495 	if (args->resource >= LINUX_RLIM_NLIMITS)
1496 		return (EINVAL);
1497 
1498 	which = linux_to_bsd_resource[args->resource];
1499 	if (which == -1)
1500 		return (EINVAL);
1501 
1502 	lim_rlimit(td, which, &bsd_rlim);
1503 
1504 	rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1505 	rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1506 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1507 }
1508 
1509 int
1510 linux_sched_setscheduler(struct thread *td,
1511     struct linux_sched_setscheduler_args *args)
1512 {
1513 	struct sched_param sched_param;
1514 	struct thread *tdt;
1515 	int error, policy;
1516 
1517 #ifdef DEBUG
1518 	if (ldebug(sched_setscheduler))
1519 		printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1520 		    args->pid, args->policy, (const void *)args->param);
1521 #endif
1522 
1523 	switch (args->policy) {
1524 	case LINUX_SCHED_OTHER:
1525 		policy = SCHED_OTHER;
1526 		break;
1527 	case LINUX_SCHED_FIFO:
1528 		policy = SCHED_FIFO;
1529 		break;
1530 	case LINUX_SCHED_RR:
1531 		policy = SCHED_RR;
1532 		break;
1533 	default:
1534 		return (EINVAL);
1535 	}
1536 
1537 	error = copyin(args->param, &sched_param, sizeof(sched_param));
1538 	if (error)
1539 		return (error);
1540 
1541 	tdt = linux_tdfind(td, args->pid, -1);
1542 	if (tdt == NULL)
1543 		return (ESRCH);
1544 
1545 	error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1546 	PROC_UNLOCK(tdt->td_proc);
1547 	return (error);
1548 }
1549 
1550 int
1551 linux_sched_getscheduler(struct thread *td,
1552     struct linux_sched_getscheduler_args *args)
1553 {
1554 	struct thread *tdt;
1555 	int error, policy;
1556 
1557 #ifdef DEBUG
1558 	if (ldebug(sched_getscheduler))
1559 		printf(ARGS(sched_getscheduler, "%d"), args->pid);
1560 #endif
1561 
1562 	tdt = linux_tdfind(td, args->pid, -1);
1563 	if (tdt == NULL)
1564 		return (ESRCH);
1565 
1566 	error = kern_sched_getscheduler(td, tdt, &policy);
1567 	PROC_UNLOCK(tdt->td_proc);
1568 
1569 	switch (policy) {
1570 	case SCHED_OTHER:
1571 		td->td_retval[0] = LINUX_SCHED_OTHER;
1572 		break;
1573 	case SCHED_FIFO:
1574 		td->td_retval[0] = LINUX_SCHED_FIFO;
1575 		break;
1576 	case SCHED_RR:
1577 		td->td_retval[0] = LINUX_SCHED_RR;
1578 		break;
1579 	}
1580 	return (error);
1581 }
1582 
1583 int
1584 linux_sched_get_priority_max(struct thread *td,
1585     struct linux_sched_get_priority_max_args *args)
1586 {
1587 	struct sched_get_priority_max_args bsd;
1588 
1589 #ifdef DEBUG
1590 	if (ldebug(sched_get_priority_max))
1591 		printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1592 #endif
1593 
1594 	switch (args->policy) {
1595 	case LINUX_SCHED_OTHER:
1596 		bsd.policy = SCHED_OTHER;
1597 		break;
1598 	case LINUX_SCHED_FIFO:
1599 		bsd.policy = SCHED_FIFO;
1600 		break;
1601 	case LINUX_SCHED_RR:
1602 		bsd.policy = SCHED_RR;
1603 		break;
1604 	default:
1605 		return (EINVAL);
1606 	}
1607 	return (sys_sched_get_priority_max(td, &bsd));
1608 }
1609 
1610 int
1611 linux_sched_get_priority_min(struct thread *td,
1612     struct linux_sched_get_priority_min_args *args)
1613 {
1614 	struct sched_get_priority_min_args bsd;
1615 
1616 #ifdef DEBUG
1617 	if (ldebug(sched_get_priority_min))
1618 		printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1619 #endif
1620 
1621 	switch (args->policy) {
1622 	case LINUX_SCHED_OTHER:
1623 		bsd.policy = SCHED_OTHER;
1624 		break;
1625 	case LINUX_SCHED_FIFO:
1626 		bsd.policy = SCHED_FIFO;
1627 		break;
1628 	case LINUX_SCHED_RR:
1629 		bsd.policy = SCHED_RR;
1630 		break;
1631 	default:
1632 		return (EINVAL);
1633 	}
1634 	return (sys_sched_get_priority_min(td, &bsd));
1635 }
1636 
1637 #define REBOOT_CAD_ON	0x89abcdef
1638 #define REBOOT_CAD_OFF	0
1639 #define REBOOT_HALT	0xcdef0123
1640 #define REBOOT_RESTART	0x01234567
1641 #define REBOOT_RESTART2	0xA1B2C3D4
1642 #define REBOOT_POWEROFF	0x4321FEDC
1643 #define REBOOT_MAGIC1	0xfee1dead
1644 #define REBOOT_MAGIC2	0x28121969
1645 #define REBOOT_MAGIC2A	0x05121996
1646 #define REBOOT_MAGIC2B	0x16041998
1647 
1648 int
1649 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1650 {
1651 	struct reboot_args bsd_args;
1652 
1653 #ifdef DEBUG
1654 	if (ldebug(reboot))
1655 		printf(ARGS(reboot, "0x%x"), args->cmd);
1656 #endif
1657 
1658 	if (args->magic1 != REBOOT_MAGIC1)
1659 		return (EINVAL);
1660 
1661 	switch (args->magic2) {
1662 	case REBOOT_MAGIC2:
1663 	case REBOOT_MAGIC2A:
1664 	case REBOOT_MAGIC2B:
1665 		break;
1666 	default:
1667 		return (EINVAL);
1668 	}
1669 
1670 	switch (args->cmd) {
1671 	case REBOOT_CAD_ON:
1672 	case REBOOT_CAD_OFF:
1673 		return (priv_check(td, PRIV_REBOOT));
1674 	case REBOOT_HALT:
1675 		bsd_args.opt = RB_HALT;
1676 		break;
1677 	case REBOOT_RESTART:
1678 	case REBOOT_RESTART2:
1679 		bsd_args.opt = 0;
1680 		break;
1681 	case REBOOT_POWEROFF:
1682 		bsd_args.opt = RB_POWEROFF;
1683 		break;
1684 	default:
1685 		return (EINVAL);
1686 	}
1687 	return (sys_reboot(td, &bsd_args));
1688 }
1689 
1690 
1691 /*
1692  * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1693  * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
1694  * are assumed to be preserved. The following lightweight syscalls fixes
1695  * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
1696  *
1697  * linux_getpid() - MP SAFE
1698  * linux_getgid() - MP SAFE
1699  * linux_getuid() - MP SAFE
1700  */
1701 
1702 int
1703 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1704 {
1705 
1706 #ifdef DEBUG
1707 	if (ldebug(getpid))
1708 		printf(ARGS(getpid, ""));
1709 #endif
1710 	td->td_retval[0] = td->td_proc->p_pid;
1711 
1712 	return (0);
1713 }
1714 
1715 int
1716 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1717 {
1718 	struct linux_emuldata *em;
1719 
1720 #ifdef DEBUG
1721 	if (ldebug(gettid))
1722 		printf(ARGS(gettid, ""));
1723 #endif
1724 
1725 	em = em_find(td);
1726 	KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1727 
1728 	td->td_retval[0] = em->em_tid;
1729 
1730 	return (0);
1731 }
1732 
1733 
1734 int
1735 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1736 {
1737 
1738 #ifdef DEBUG
1739 	if (ldebug(getppid))
1740 		printf(ARGS(getppid, ""));
1741 #endif
1742 
1743 	td->td_retval[0] = kern_getppid(td);
1744 	return (0);
1745 }
1746 
1747 int
1748 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1749 {
1750 
1751 #ifdef DEBUG
1752 	if (ldebug(getgid))
1753 		printf(ARGS(getgid, ""));
1754 #endif
1755 
1756 	td->td_retval[0] = td->td_ucred->cr_rgid;
1757 	return (0);
1758 }
1759 
1760 int
1761 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1762 {
1763 
1764 #ifdef DEBUG
1765 	if (ldebug(getuid))
1766 		printf(ARGS(getuid, ""));
1767 #endif
1768 
1769 	td->td_retval[0] = td->td_ucred->cr_ruid;
1770 	return (0);
1771 }
1772 
1773 
1774 int
1775 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1776 {
1777 	struct getsid_args bsd;
1778 
1779 #ifdef DEBUG
1780 	if (ldebug(getsid))
1781 		printf(ARGS(getsid, "%i"), args->pid);
1782 #endif
1783 
1784 	bsd.pid = args->pid;
1785 	return (sys_getsid(td, &bsd));
1786 }
1787 
1788 int
1789 linux_nosys(struct thread *td, struct nosys_args *ignore)
1790 {
1791 
1792 	return (ENOSYS);
1793 }
1794 
1795 int
1796 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1797 {
1798 	struct getpriority_args bsd_args;
1799 	int error;
1800 
1801 #ifdef DEBUG
1802 	if (ldebug(getpriority))
1803 		printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
1804 #endif
1805 
1806 	bsd_args.which = args->which;
1807 	bsd_args.who = args->who;
1808 	error = sys_getpriority(td, &bsd_args);
1809 	td->td_retval[0] = 20 - td->td_retval[0];
1810 	return (error);
1811 }
1812 
1813 int
1814 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1815 {
1816 	int name[2];
1817 
1818 #ifdef DEBUG
1819 	if (ldebug(sethostname))
1820 		printf(ARGS(sethostname, "*, %i"), args->len);
1821 #endif
1822 
1823 	name[0] = CTL_KERN;
1824 	name[1] = KERN_HOSTNAME;
1825 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1826 	    args->len, 0, 0));
1827 }
1828 
1829 int
1830 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1831 {
1832 	int name[2];
1833 
1834 #ifdef DEBUG
1835 	if (ldebug(setdomainname))
1836 		printf(ARGS(setdomainname, "*, %i"), args->len);
1837 #endif
1838 
1839 	name[0] = CTL_KERN;
1840 	name[1] = KERN_NISDOMAINNAME;
1841 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1842 	    args->len, 0, 0));
1843 }
1844 
1845 int
1846 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1847 {
1848 
1849 #ifdef DEBUG
1850 	if (ldebug(exit_group))
1851 		printf(ARGS(exit_group, "%i"), args->error_code);
1852 #endif
1853 
1854 	LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1855 	    args->error_code);
1856 
1857 	/*
1858 	 * XXX: we should send a signal to the parent if
1859 	 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1860 	 * as it doesnt occur often.
1861 	 */
1862 	exit1(td, args->error_code, 0);
1863 		/* NOTREACHED */
1864 }
1865 
1866 #define _LINUX_CAPABILITY_VERSION_1  0x19980330
1867 #define _LINUX_CAPABILITY_VERSION_2  0x20071026
1868 #define _LINUX_CAPABILITY_VERSION_3  0x20080522
1869 
1870 struct l_user_cap_header {
1871 	l_int	version;
1872 	l_int	pid;
1873 };
1874 
1875 struct l_user_cap_data {
1876 	l_int	effective;
1877 	l_int	permitted;
1878 	l_int	inheritable;
1879 };
1880 
1881 int
1882 linux_capget(struct thread *td, struct linux_capget_args *uap)
1883 {
1884 	struct l_user_cap_header luch;
1885 	struct l_user_cap_data lucd[2];
1886 	int error, u32s;
1887 
1888 	if (uap->hdrp == NULL)
1889 		return (EFAULT);
1890 
1891 	error = copyin(uap->hdrp, &luch, sizeof(luch));
1892 	if (error != 0)
1893 		return (error);
1894 
1895 	switch (luch.version) {
1896 	case _LINUX_CAPABILITY_VERSION_1:
1897 		u32s = 1;
1898 		break;
1899 	case _LINUX_CAPABILITY_VERSION_2:
1900 	case _LINUX_CAPABILITY_VERSION_3:
1901 		u32s = 2;
1902 		break;
1903 	default:
1904 #ifdef DEBUG
1905 		if (ldebug(capget))
1906 			printf(LMSG("invalid capget capability version 0x%x"),
1907 			    luch.version);
1908 #endif
1909 		luch.version = _LINUX_CAPABILITY_VERSION_1;
1910 		error = copyout(&luch, uap->hdrp, sizeof(luch));
1911 		if (error)
1912 			return (error);
1913 		return (EINVAL);
1914 	}
1915 
1916 	if (luch.pid)
1917 		return (EPERM);
1918 
1919 	if (uap->datap) {
1920 		/*
1921 		 * The current implementation doesn't support setting
1922 		 * a capability (it's essentially a stub) so indicate
1923 		 * that no capabilities are currently set or available
1924 		 * to request.
1925 		 */
1926 		memset(&lucd, 0, u32s * sizeof(lucd[0]));
1927 		error = copyout(&lucd, uap->datap, u32s * sizeof(lucd[0]));
1928 	}
1929 
1930 	return (error);
1931 }
1932 
1933 int
1934 linux_capset(struct thread *td, struct linux_capset_args *uap)
1935 {
1936 	struct l_user_cap_header luch;
1937 	struct l_user_cap_data lucd[2];
1938 	int error, i, u32s;
1939 
1940 	if (uap->hdrp == NULL || uap->datap == NULL)
1941 		return (EFAULT);
1942 
1943 	error = copyin(uap->hdrp, &luch, sizeof(luch));
1944 	if (error != 0)
1945 		return (error);
1946 
1947 	switch (luch.version) {
1948 	case _LINUX_CAPABILITY_VERSION_1:
1949 		u32s = 1;
1950 		break;
1951 	case _LINUX_CAPABILITY_VERSION_2:
1952 	case _LINUX_CAPABILITY_VERSION_3:
1953 		u32s = 2;
1954 		break;
1955 	default:
1956 #ifdef DEBUG
1957 		if (ldebug(capset))
1958 			printf(LMSG("invalid capset capability version 0x%x"),
1959 			    luch.version);
1960 #endif
1961 		luch.version = _LINUX_CAPABILITY_VERSION_1;
1962 		error = copyout(&luch, uap->hdrp, sizeof(luch));
1963 		if (error)
1964 			return (error);
1965 		return (EINVAL);
1966 	}
1967 
1968 	if (luch.pid)
1969 		return (EPERM);
1970 
1971 	error = copyin(uap->datap, &lucd, u32s * sizeof(lucd[0]));
1972 	if (error != 0)
1973 		return (error);
1974 
1975 	/* We currently don't support setting any capabilities. */
1976 	for (i = 0; i < u32s; i++) {
1977 		if (lucd[i].effective || lucd[i].permitted ||
1978 		    lucd[i].inheritable) {
1979 			linux_msg(td,
1980 			    "capset[%d] effective=0x%x, permitted=0x%x, "
1981 			    "inheritable=0x%x is not implemented", i,
1982 			    (int)lucd[i].effective, (int)lucd[i].permitted,
1983 			    (int)lucd[i].inheritable);
1984 			return (EPERM);
1985 		}
1986 	}
1987 
1988 	return (0);
1989 }
1990 
1991 int
1992 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1993 {
1994 	int error = 0, max_size;
1995 	struct proc *p = td->td_proc;
1996 	char comm[LINUX_MAX_COMM_LEN];
1997 	struct linux_emuldata *em;
1998 	int pdeath_signal;
1999 
2000 #ifdef DEBUG
2001 	if (ldebug(prctl))
2002 		printf(ARGS(prctl, "%d, %ju, %ju, %ju, %ju"), args->option,
2003 		    (uintmax_t)args->arg2, (uintmax_t)args->arg3,
2004 		    (uintmax_t)args->arg4, (uintmax_t)args->arg5);
2005 #endif
2006 
2007 	switch (args->option) {
2008 	case LINUX_PR_SET_PDEATHSIG:
2009 		if (!LINUX_SIG_VALID(args->arg2))
2010 			return (EINVAL);
2011 		em = em_find(td);
2012 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
2013 		em->pdeath_signal = args->arg2;
2014 		break;
2015 	case LINUX_PR_GET_PDEATHSIG:
2016 		em = em_find(td);
2017 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
2018 		pdeath_signal = em->pdeath_signal;
2019 		error = copyout(&pdeath_signal,
2020 		    (void *)(register_t)args->arg2,
2021 		    sizeof(pdeath_signal));
2022 		break;
2023 	case LINUX_PR_GET_KEEPCAPS:
2024 		/*
2025 		 * Indicate that we always clear the effective and
2026 		 * permitted capability sets when the user id becomes
2027 		 * non-zero (actually the capability sets are simply
2028 		 * always zero in the current implementation).
2029 		 */
2030 		td->td_retval[0] = 0;
2031 		break;
2032 	case LINUX_PR_SET_KEEPCAPS:
2033 		/*
2034 		 * Ignore requests to keep the effective and permitted
2035 		 * capability sets when the user id becomes non-zero.
2036 		 */
2037 		break;
2038 	case LINUX_PR_SET_NAME:
2039 		/*
2040 		 * To be on the safe side we need to make sure to not
2041 		 * overflow the size a Linux program expects. We already
2042 		 * do this here in the copyin, so that we don't need to
2043 		 * check on copyout.
2044 		 */
2045 		max_size = MIN(sizeof(comm), sizeof(p->p_comm));
2046 		error = copyinstr((void *)(register_t)args->arg2, comm,
2047 		    max_size, NULL);
2048 
2049 		/* Linux silently truncates the name if it is too long. */
2050 		if (error == ENAMETOOLONG) {
2051 			/*
2052 			 * XXX: copyinstr() isn't documented to populate the
2053 			 * array completely, so do a copyin() to be on the
2054 			 * safe side. This should be changed in case
2055 			 * copyinstr() is changed to guarantee this.
2056 			 */
2057 			error = copyin((void *)(register_t)args->arg2, comm,
2058 			    max_size - 1);
2059 			comm[max_size - 1] = '\0';
2060 		}
2061 		if (error)
2062 			return (error);
2063 
2064 		PROC_LOCK(p);
2065 		strlcpy(p->p_comm, comm, sizeof(p->p_comm));
2066 		PROC_UNLOCK(p);
2067 		break;
2068 	case LINUX_PR_GET_NAME:
2069 		PROC_LOCK(p);
2070 		strlcpy(comm, p->p_comm, sizeof(comm));
2071 		PROC_UNLOCK(p);
2072 		error = copyout(comm, (void *)(register_t)args->arg2,
2073 		    strlen(comm) + 1);
2074 		break;
2075 	default:
2076 		error = EINVAL;
2077 		break;
2078 	}
2079 
2080 	return (error);
2081 }
2082 
2083 int
2084 linux_sched_setparam(struct thread *td,
2085     struct linux_sched_setparam_args *uap)
2086 {
2087 	struct sched_param sched_param;
2088 	struct thread *tdt;
2089 	int error;
2090 
2091 #ifdef DEBUG
2092 	if (ldebug(sched_setparam))
2093 		printf(ARGS(sched_setparam, "%d, *"), uap->pid);
2094 #endif
2095 
2096 	error = copyin(uap->param, &sched_param, sizeof(sched_param));
2097 	if (error)
2098 		return (error);
2099 
2100 	tdt = linux_tdfind(td, uap->pid, -1);
2101 	if (tdt == NULL)
2102 		return (ESRCH);
2103 
2104 	error = kern_sched_setparam(td, tdt, &sched_param);
2105 	PROC_UNLOCK(tdt->td_proc);
2106 	return (error);
2107 }
2108 
2109 int
2110 linux_sched_getparam(struct thread *td,
2111     struct linux_sched_getparam_args *uap)
2112 {
2113 	struct sched_param sched_param;
2114 	struct thread *tdt;
2115 	int error;
2116 
2117 #ifdef DEBUG
2118 	if (ldebug(sched_getparam))
2119 		printf(ARGS(sched_getparam, "%d, *"), uap->pid);
2120 #endif
2121 
2122 	tdt = linux_tdfind(td, uap->pid, -1);
2123 	if (tdt == NULL)
2124 		return (ESRCH);
2125 
2126 	error = kern_sched_getparam(td, tdt, &sched_param);
2127 	PROC_UNLOCK(tdt->td_proc);
2128 	if (error == 0)
2129 		error = copyout(&sched_param, uap->param,
2130 		    sizeof(sched_param));
2131 	return (error);
2132 }
2133 
2134 /*
2135  * Get affinity of a process.
2136  */
2137 int
2138 linux_sched_getaffinity(struct thread *td,
2139     struct linux_sched_getaffinity_args *args)
2140 {
2141 	int error;
2142 	struct thread *tdt;
2143 
2144 #ifdef DEBUG
2145 	if (ldebug(sched_getaffinity))
2146 		printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid,
2147 		    args->len);
2148 #endif
2149 	if (args->len < sizeof(cpuset_t))
2150 		return (EINVAL);
2151 
2152 	tdt = linux_tdfind(td, args->pid, -1);
2153 	if (tdt == NULL)
2154 		return (ESRCH);
2155 
2156 	PROC_UNLOCK(tdt->td_proc);
2157 
2158 	error = kern_cpuset_getaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2159 	    tdt->td_tid, sizeof(cpuset_t), (cpuset_t *)args->user_mask_ptr);
2160 	if (error == 0)
2161 		td->td_retval[0] = sizeof(cpuset_t);
2162 
2163 	return (error);
2164 }
2165 
2166 /*
2167  *  Set affinity of a process.
2168  */
2169 int
2170 linux_sched_setaffinity(struct thread *td,
2171     struct linux_sched_setaffinity_args *args)
2172 {
2173 	struct thread *tdt;
2174 
2175 #ifdef DEBUG
2176 	if (ldebug(sched_setaffinity))
2177 		printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid,
2178 		    args->len);
2179 #endif
2180 	if (args->len < sizeof(cpuset_t))
2181 		return (EINVAL);
2182 
2183 	tdt = linux_tdfind(td, args->pid, -1);
2184 	if (tdt == NULL)
2185 		return (ESRCH);
2186 
2187 	PROC_UNLOCK(tdt->td_proc);
2188 
2189 	return (kern_cpuset_setaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2190 	    tdt->td_tid, sizeof(cpuset_t), (cpuset_t *) args->user_mask_ptr));
2191 }
2192 
2193 struct linux_rlimit64 {
2194 	uint64_t	rlim_cur;
2195 	uint64_t	rlim_max;
2196 };
2197 
2198 int
2199 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
2200 {
2201 	struct rlimit rlim, nrlim;
2202 	struct linux_rlimit64 lrlim;
2203 	struct proc *p;
2204 	u_int which;
2205 	int flags;
2206 	int error;
2207 
2208 #ifdef DEBUG
2209 	if (ldebug(prlimit64))
2210 		printf(ARGS(prlimit64, "%d, %d, %p, %p"), args->pid,
2211 		    args->resource, (void *)args->new, (void *)args->old);
2212 #endif
2213 
2214 	if (args->resource >= LINUX_RLIM_NLIMITS)
2215 		return (EINVAL);
2216 
2217 	which = linux_to_bsd_resource[args->resource];
2218 	if (which == -1)
2219 		return (EINVAL);
2220 
2221 	if (args->new != NULL) {
2222 		/*
2223 		 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
2224 		 * rlim is unsigned 64-bit. FreeBSD treats negative limits
2225 		 * as INFINITY so we do not need a conversion even.
2226 		 */
2227 		error = copyin(args->new, &nrlim, sizeof(nrlim));
2228 		if (error != 0)
2229 			return (error);
2230 	}
2231 
2232 	flags = PGET_HOLD | PGET_NOTWEXIT;
2233 	if (args->new != NULL)
2234 		flags |= PGET_CANDEBUG;
2235 	else
2236 		flags |= PGET_CANSEE;
2237 	error = pget(args->pid, flags, &p);
2238 	if (error != 0)
2239 		return (error);
2240 
2241 	if (args->old != NULL) {
2242 		PROC_LOCK(p);
2243 		lim_rlimit_proc(p, which, &rlim);
2244 		PROC_UNLOCK(p);
2245 		if (rlim.rlim_cur == RLIM_INFINITY)
2246 			lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2247 		else
2248 			lrlim.rlim_cur = rlim.rlim_cur;
2249 		if (rlim.rlim_max == RLIM_INFINITY)
2250 			lrlim.rlim_max = LINUX_RLIM_INFINITY;
2251 		else
2252 			lrlim.rlim_max = rlim.rlim_max;
2253 		error = copyout(&lrlim, args->old, sizeof(lrlim));
2254 		if (error != 0)
2255 			goto out;
2256 	}
2257 
2258 	if (args->new != NULL)
2259 		error = kern_proc_setrlimit(td, p, which, &nrlim);
2260 
2261  out:
2262 	PRELE(p);
2263 	return (error);
2264 }
2265 
2266 int
2267 linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2268 {
2269 	struct timeval utv, tv0, tv1, *tvp;
2270 	struct l_pselect6arg lpse6;
2271 	struct l_timespec lts;
2272 	struct timespec uts;
2273 	l_sigset_t l_ss;
2274 	sigset_t *ssp;
2275 	sigset_t ss;
2276 	int error;
2277 
2278 	ssp = NULL;
2279 	if (args->sig != NULL) {
2280 		error = copyin(args->sig, &lpse6, sizeof(lpse6));
2281 		if (error != 0)
2282 			return (error);
2283 		if (lpse6.ss_len != sizeof(l_ss))
2284 			return (EINVAL);
2285 		if (lpse6.ss != 0) {
2286 			error = copyin(PTRIN(lpse6.ss), &l_ss,
2287 			    sizeof(l_ss));
2288 			if (error != 0)
2289 				return (error);
2290 			linux_to_bsd_sigset(&l_ss, &ss);
2291 			ssp = &ss;
2292 		}
2293 	}
2294 
2295 	/*
2296 	 * Currently glibc changes nanosecond number to microsecond.
2297 	 * This mean losing precision but for now it is hardly seen.
2298 	 */
2299 	if (args->tsp != NULL) {
2300 		error = copyin(args->tsp, &lts, sizeof(lts));
2301 		if (error != 0)
2302 			return (error);
2303 		error = linux_to_native_timespec(&uts, &lts);
2304 		if (error != 0)
2305 			return (error);
2306 
2307 		TIMESPEC_TO_TIMEVAL(&utv, &uts);
2308 		if (itimerfix(&utv))
2309 			return (EINVAL);
2310 
2311 		microtime(&tv0);
2312 		tvp = &utv;
2313 	} else
2314 		tvp = NULL;
2315 
2316 	error = kern_pselect(td, args->nfds, args->readfds, args->writefds,
2317 	    args->exceptfds, tvp, ssp, LINUX_NFDBITS);
2318 
2319 	if (error == 0 && args->tsp != NULL) {
2320 		if (td->td_retval[0] != 0) {
2321 			/*
2322 			 * Compute how much time was left of the timeout,
2323 			 * by subtracting the current time and the time
2324 			 * before we started the call, and subtracting
2325 			 * that result from the user-supplied value.
2326 			 */
2327 
2328 			microtime(&tv1);
2329 			timevalsub(&tv1, &tv0);
2330 			timevalsub(&utv, &tv1);
2331 			if (utv.tv_sec < 0)
2332 				timevalclear(&utv);
2333 		} else
2334 			timevalclear(&utv);
2335 
2336 		TIMEVAL_TO_TIMESPEC(&utv, &uts);
2337 
2338 		error = native_to_linux_timespec(&lts, &uts);
2339 		if (error == 0)
2340 			error = copyout(&lts, args->tsp, sizeof(lts));
2341 	}
2342 
2343 	return (error);
2344 }
2345 
2346 int
2347 linux_ppoll(struct thread *td, struct linux_ppoll_args *args)
2348 {
2349 	struct timespec ts0, ts1;
2350 	struct l_timespec lts;
2351 	struct timespec uts, *tsp;
2352 	l_sigset_t l_ss;
2353 	sigset_t *ssp;
2354 	sigset_t ss;
2355 	int error;
2356 
2357 	if (args->sset != NULL) {
2358 		if (args->ssize != sizeof(l_ss))
2359 			return (EINVAL);
2360 		error = copyin(args->sset, &l_ss, sizeof(l_ss));
2361 		if (error)
2362 			return (error);
2363 		linux_to_bsd_sigset(&l_ss, &ss);
2364 		ssp = &ss;
2365 	} else
2366 		ssp = NULL;
2367 	if (args->tsp != NULL) {
2368 		error = copyin(args->tsp, &lts, sizeof(lts));
2369 		if (error)
2370 			return (error);
2371 		error = linux_to_native_timespec(&uts, &lts);
2372 		if (error != 0)
2373 			return (error);
2374 
2375 		nanotime(&ts0);
2376 		tsp = &uts;
2377 	} else
2378 		tsp = NULL;
2379 
2380 	error = kern_poll(td, args->fds, args->nfds, tsp, ssp);
2381 
2382 	if (error == 0 && args->tsp != NULL) {
2383 		if (td->td_retval[0]) {
2384 			nanotime(&ts1);
2385 			timespecsub(&ts1, &ts0, &ts1);
2386 			timespecsub(&uts, &ts1, &uts);
2387 			if (uts.tv_sec < 0)
2388 				timespecclear(&uts);
2389 		} else
2390 			timespecclear(&uts);
2391 
2392 		error = native_to_linux_timespec(&lts, &uts);
2393 		if (error == 0)
2394 			error = copyout(&lts, args->tsp, sizeof(lts));
2395 	}
2396 
2397 	return (error);
2398 }
2399 
2400 #if defined(DEBUG) || defined(KTR)
2401 /* XXX: can be removed when every ldebug(...) and KTR stuff are removed. */
2402 
2403 #ifdef COMPAT_LINUX32
2404 #define	L_MAXSYSCALL	LINUX32_SYS_MAXSYSCALL
2405 #else
2406 #define	L_MAXSYSCALL	LINUX_SYS_MAXSYSCALL
2407 #endif
2408 
2409 u_char linux_debug_map[howmany(L_MAXSYSCALL, sizeof(u_char))];
2410 
2411 static int
2412 linux_debug(int syscall, int toggle, int global)
2413 {
2414 
2415 	if (global) {
2416 		char c = toggle ? 0 : 0xff;
2417 
2418 		memset(linux_debug_map, c, sizeof(linux_debug_map));
2419 		return (0);
2420 	}
2421 	if (syscall < 0 || syscall >= L_MAXSYSCALL)
2422 		return (EINVAL);
2423 	if (toggle)
2424 		clrbit(linux_debug_map, syscall);
2425 	else
2426 		setbit(linux_debug_map, syscall);
2427 	return (0);
2428 }
2429 #undef L_MAXSYSCALL
2430 
2431 /*
2432  * Usage: sysctl linux.debug=<syscall_nr>.<0/1>
2433  *
2434  *    E.g.: sysctl linux.debug=21.0
2435  *
2436  * As a special case, syscall "all" will apply to all syscalls globally.
2437  */
2438 #define LINUX_MAX_DEBUGSTR	16
2439 int
2440 linux_sysctl_debug(SYSCTL_HANDLER_ARGS)
2441 {
2442 	char value[LINUX_MAX_DEBUGSTR], *p;
2443 	int error, sysc, toggle;
2444 	int global = 0;
2445 
2446 	value[0] = '\0';
2447 	error = sysctl_handle_string(oidp, value, LINUX_MAX_DEBUGSTR, req);
2448 	if (error || req->newptr == NULL)
2449 		return (error);
2450 	for (p = value; *p != '\0' && *p != '.'; p++);
2451 	if (*p == '\0')
2452 		return (EINVAL);
2453 	*p++ = '\0';
2454 	sysc = strtol(value, NULL, 0);
2455 	toggle = strtol(p, NULL, 0);
2456 	if (strcmp(value, "all") == 0)
2457 		global = 1;
2458 	error = linux_debug(sysc, toggle, global);
2459 	return (error);
2460 }
2461 
2462 #endif /* DEBUG || KTR */
2463 
2464 int
2465 linux_sched_rr_get_interval(struct thread *td,
2466     struct linux_sched_rr_get_interval_args *uap)
2467 {
2468 	struct timespec ts;
2469 	struct l_timespec lts;
2470 	struct thread *tdt;
2471 	int error;
2472 
2473 	/*
2474 	 * According to man in case the invalid pid specified
2475 	 * EINVAL should be returned.
2476 	 */
2477 	if (uap->pid < 0)
2478 		return (EINVAL);
2479 
2480 	tdt = linux_tdfind(td, uap->pid, -1);
2481 	if (tdt == NULL)
2482 		return (ESRCH);
2483 
2484 	error = kern_sched_rr_get_interval_td(td, tdt, &ts);
2485 	PROC_UNLOCK(tdt->td_proc);
2486 	if (error != 0)
2487 		return (error);
2488 	error = native_to_linux_timespec(&lts, &ts);
2489 	if (error != 0)
2490 		return (error);
2491 	return (copyout(&lts, uap->interval, sizeof(lts)));
2492 }
2493 
2494 /*
2495  * In case when the Linux thread is the initial thread in
2496  * the thread group thread id is equal to the process id.
2497  * Glibc depends on this magic (assert in pthread_getattr_np.c).
2498  */
2499 struct thread *
2500 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid)
2501 {
2502 	struct linux_emuldata *em;
2503 	struct thread *tdt;
2504 	struct proc *p;
2505 
2506 	tdt = NULL;
2507 	if (tid == 0 || tid == td->td_tid) {
2508 		tdt = td;
2509 		PROC_LOCK(tdt->td_proc);
2510 	} else if (tid > PID_MAX)
2511 		tdt = tdfind(tid, pid);
2512 	else {
2513 		/*
2514 		 * Initial thread where the tid equal to the pid.
2515 		 */
2516 		p = pfind(tid);
2517 		if (p != NULL) {
2518 			if (SV_PROC_ABI(p) != SV_ABI_LINUX) {
2519 				/*
2520 				 * p is not a Linuxulator process.
2521 				 */
2522 				PROC_UNLOCK(p);
2523 				return (NULL);
2524 			}
2525 			FOREACH_THREAD_IN_PROC(p, tdt) {
2526 				em = em_find(tdt);
2527 				if (tid == em->em_tid)
2528 					return (tdt);
2529 			}
2530 			PROC_UNLOCK(p);
2531 		}
2532 		return (NULL);
2533 	}
2534 
2535 	return (tdt);
2536 }
2537 
2538 void
2539 linux_to_bsd_waitopts(int options, int *bsdopts)
2540 {
2541 
2542 	if (options & LINUX_WNOHANG)
2543 		*bsdopts |= WNOHANG;
2544 	if (options & LINUX_WUNTRACED)
2545 		*bsdopts |= WUNTRACED;
2546 	if (options & LINUX_WEXITED)
2547 		*bsdopts |= WEXITED;
2548 	if (options & LINUX_WCONTINUED)
2549 		*bsdopts |= WCONTINUED;
2550 	if (options & LINUX_WNOWAIT)
2551 		*bsdopts |= WNOWAIT;
2552 
2553 	if (options & __WCLONE)
2554 		*bsdopts |= WLINUXCLONE;
2555 }
2556 
2557 int
2558 linux_getrandom(struct thread *td, struct linux_getrandom_args *args)
2559 {
2560 	struct uio uio;
2561 	struct iovec iov;
2562 	int error;
2563 
2564 	if (args->flags & ~(LINUX_GRND_NONBLOCK|LINUX_GRND_RANDOM))
2565 		return (EINVAL);
2566 	if (args->count > INT_MAX)
2567 		args->count = INT_MAX;
2568 
2569 	iov.iov_base = args->buf;
2570 	iov.iov_len = args->count;
2571 
2572 	uio.uio_iov = &iov;
2573 	uio.uio_iovcnt = 1;
2574 	uio.uio_resid = iov.iov_len;
2575 	uio.uio_segflg = UIO_USERSPACE;
2576 	uio.uio_rw = UIO_READ;
2577 	uio.uio_td = td;
2578 
2579 	error = read_random_uio(&uio, args->flags & LINUX_GRND_NONBLOCK);
2580 	if (error == 0)
2581 		td->td_retval[0] = args->count - uio.uio_resid;
2582 	return (error);
2583 }
2584 
2585 int
2586 linux_mincore(struct thread *td, struct linux_mincore_args *args)
2587 {
2588 
2589 	/* Needs to be page-aligned */
2590 	if (args->start & PAGE_MASK)
2591 		return (EINVAL);
2592 	return (kern_mincore(td, args->start, args->len, args->vec));
2593 }
2594