xref: /freebsd/sys/compat/linux/linux_misc.c (revision daceb336172a6b0572de864b97e70b28451ca636)
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 #ifdef DEBUG
204 	if (ldebug(alarm))
205 		printf(ARGS(alarm, "%u"), args->secs);
206 #endif
207 	secs = args->secs;
208 	/*
209 	 * Linux alarm() is always successful. Limit secs to INT32_MAX / 2
210 	 * to match kern_setitimer()'s limit to avoid error from it.
211 	 *
212 	 * XXX. Linux limit secs to INT_MAX on 32 and does not limit on 64-bit
213 	 * platforms.
214 	 */
215 	if (secs > INT32_MAX / 2)
216 		secs = INT32_MAX / 2;
217 
218 	it.it_value.tv_sec = secs;
219 	it.it_value.tv_usec = 0;
220 	timevalclear(&it.it_interval);
221 	error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
222 	KASSERT(error == 0, ("kern_setitimer returns %d", error));
223 
224 	if ((old_it.it_value.tv_sec == 0 && old_it.it_value.tv_usec > 0) ||
225 	    old_it.it_value.tv_usec >= 500000)
226 		old_it.it_value.tv_sec++;
227 	td->td_retval[0] = old_it.it_value.tv_sec;
228 	return (0);
229 }
230 #endif
231 
232 int
233 linux_brk(struct thread *td, struct linux_brk_args *args)
234 {
235 	struct vmspace *vm = td->td_proc->p_vmspace;
236 	uintptr_t new, old;
237 
238 #ifdef DEBUG
239 	if (ldebug(brk))
240 		printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
241 #endif
242 	old = (uintptr_t)vm->vm_daddr + ctob(vm->vm_dsize);
243 	new = (uintptr_t)args->dsend;
244 	if ((caddr_t)new > vm->vm_daddr && !kern_break(td, &new))
245 		td->td_retval[0] = (register_t)new;
246 	else
247 		td->td_retval[0] = (register_t)old;
248 
249 	return (0);
250 }
251 
252 #if defined(__i386__)
253 /* XXX: what about amd64/linux32? */
254 
255 int
256 linux_uselib(struct thread *td, struct linux_uselib_args *args)
257 {
258 	struct nameidata ni;
259 	struct vnode *vp;
260 	struct exec *a_out;
261 	vm_map_t map;
262 	vm_map_entry_t entry;
263 	struct vattr attr;
264 	vm_offset_t vmaddr;
265 	unsigned long file_offset;
266 	unsigned long bss_size;
267 	char *library;
268 	ssize_t aresid;
269 	int error;
270 	bool locked, opened, textset;
271 
272 	LCONVPATHEXIST(td, args->library, &library);
273 
274 #ifdef DEBUG
275 	if (ldebug(uselib))
276 		printf(ARGS(uselib, "%s"), library);
277 #endif
278 
279 	a_out = NULL;
280 	vp = NULL;
281 	locked = false;
282 	textset = false;
283 	opened = false;
284 
285 	NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
286 	    UIO_SYSSPACE, library, td);
287 	error = namei(&ni);
288 	LFREEPATH(library);
289 	if (error)
290 		goto cleanup;
291 
292 	vp = ni.ni_vp;
293 	NDFREE(&ni, NDF_ONLY_PNBUF);
294 
295 	/*
296 	 * From here on down, we have a locked vnode that must be unlocked.
297 	 * XXX: The code below largely duplicates exec_check_permissions().
298 	 */
299 	locked = true;
300 
301 	/* Executable? */
302 	error = VOP_GETATTR(vp, &attr, td->td_ucred);
303 	if (error)
304 		goto cleanup;
305 
306 	if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
307 	    ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
308 		/* EACCESS is what exec(2) returns. */
309 		error = ENOEXEC;
310 		goto cleanup;
311 	}
312 
313 	/* Sensible size? */
314 	if (attr.va_size == 0) {
315 		error = ENOEXEC;
316 		goto cleanup;
317 	}
318 
319 	/* Can we access it? */
320 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
321 	if (error)
322 		goto cleanup;
323 
324 	/*
325 	 * XXX: This should use vn_open() so that it is properly authorized,
326 	 * and to reduce code redundancy all over the place here.
327 	 * XXX: Not really, it duplicates far more of exec_check_permissions()
328 	 * than vn_open().
329 	 */
330 #ifdef MAC
331 	error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
332 	if (error)
333 		goto cleanup;
334 #endif
335 	error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
336 	if (error)
337 		goto cleanup;
338 	opened = true;
339 
340 	/* Pull in executable header into exec_map */
341 	error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
342 	    VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
343 	if (error)
344 		goto cleanup;
345 
346 	/* Is it a Linux binary ? */
347 	if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
348 		error = ENOEXEC;
349 		goto cleanup;
350 	}
351 
352 	/*
353 	 * While we are here, we should REALLY do some more checks
354 	 */
355 
356 	/* Set file/virtual offset based on a.out variant. */
357 	switch ((int)(a_out->a_magic & 0xffff)) {
358 	case 0413:			/* ZMAGIC */
359 		file_offset = 1024;
360 		break;
361 	case 0314:			/* QMAGIC */
362 		file_offset = 0;
363 		break;
364 	default:
365 		error = ENOEXEC;
366 		goto cleanup;
367 	}
368 
369 	bss_size = round_page(a_out->a_bss);
370 
371 	/* Check various fields in header for validity/bounds. */
372 	if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
373 		error = ENOEXEC;
374 		goto cleanup;
375 	}
376 
377 	/* text + data can't exceed file size */
378 	if (a_out->a_data + a_out->a_text > attr.va_size) {
379 		error = EFAULT;
380 		goto cleanup;
381 	}
382 
383 	/*
384 	 * text/data/bss must not exceed limits
385 	 * XXX - this is not complete. it should check current usage PLUS
386 	 * the resources needed by this library.
387 	 */
388 	PROC_LOCK(td->td_proc);
389 	if (a_out->a_text > maxtsiz ||
390 	    a_out->a_data + bss_size > lim_cur_proc(td->td_proc, RLIMIT_DATA) ||
391 	    racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
392 	    bss_size) != 0) {
393 		PROC_UNLOCK(td->td_proc);
394 		error = ENOMEM;
395 		goto cleanup;
396 	}
397 	PROC_UNLOCK(td->td_proc);
398 
399 	/*
400 	 * Prevent more writers.
401 	 */
402 	error = VOP_SET_TEXT(vp);
403 	if (error != 0)
404 		goto cleanup;
405 	textset = true;
406 
407 	/*
408 	 * Lock no longer needed
409 	 */
410 	locked = false;
411 	VOP_UNLOCK(vp, 0);
412 
413 	/*
414 	 * Check if file_offset page aligned. Currently we cannot handle
415 	 * misalinged file offsets, and so we read in the entire image
416 	 * (what a waste).
417 	 */
418 	if (file_offset & PAGE_MASK) {
419 #ifdef DEBUG
420 		printf("uselib: Non page aligned binary %lu\n", file_offset);
421 #endif
422 		/* Map text+data read/write/execute */
423 
424 		/* a_entry is the load address and is page aligned */
425 		vmaddr = trunc_page(a_out->a_entry);
426 
427 		/* get anon user mapping, read+write+execute */
428 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
429 		    &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
430 		    VM_PROT_ALL, VM_PROT_ALL, 0);
431 		if (error)
432 			goto cleanup;
433 
434 		error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
435 		    a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
436 		    td->td_ucred, NOCRED, &aresid, td);
437 		if (error != 0)
438 			goto cleanup;
439 		if (aresid != 0) {
440 			error = ENOEXEC;
441 			goto cleanup;
442 		}
443 	} else {
444 #ifdef DEBUG
445 		printf("uselib: Page aligned binary %lu\n", file_offset);
446 #endif
447 		/*
448 		 * for QMAGIC, a_entry is 20 bytes beyond the load address
449 		 * to skip the executable header
450 		 */
451 		vmaddr = trunc_page(a_out->a_entry);
452 
453 		/*
454 		 * Map it all into the process's space as a single
455 		 * copy-on-write "data" segment.
456 		 */
457 		map = &td->td_proc->p_vmspace->vm_map;
458 		error = vm_mmap(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 		vm_map_lock(map);
464 		if (!vm_map_lookup_entry(map, vmaddr, &entry)) {
465 			vm_map_unlock(map);
466 			error = EDOOFUS;
467 			goto cleanup;
468 		}
469 		entry->eflags |= MAP_ENTRY_VN_EXEC;
470 		vm_map_unlock(map);
471 		textset = false;
472 	}
473 #ifdef DEBUG
474 	printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
475 	    ((long *)vmaddr)[1]);
476 #endif
477 	if (bss_size != 0) {
478 		/* Calculate BSS start address */
479 		vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
480 		    a_out->a_data;
481 
482 		/* allocate some 'anon' space */
483 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
484 		    &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
485 		    VM_PROT_ALL, 0);
486 		if (error)
487 			goto cleanup;
488 	}
489 
490 cleanup:
491 	if (opened) {
492 		if (locked)
493 			VOP_UNLOCK(vp, 0);
494 		locked = false;
495 		VOP_CLOSE(vp, FREAD, td->td_ucred, td);
496 	}
497 	if (textset)
498 		VOP_UNSET_TEXT_CHECKED(vp);
499 	if (locked)
500 		VOP_UNLOCK(vp, 0);
501 
502 	/* Release the temporary mapping. */
503 	if (a_out)
504 		kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
505 
506 	return (error);
507 }
508 
509 #endif	/* __i386__ */
510 
511 #ifdef LINUX_LEGACY_SYSCALLS
512 int
513 linux_select(struct thread *td, struct linux_select_args *args)
514 {
515 	l_timeval ltv;
516 	struct timeval tv0, tv1, utv, *tvp;
517 	int error;
518 
519 #ifdef DEBUG
520 	if (ldebug(select))
521 		printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
522 		    (void *)args->readfds, (void *)args->writefds,
523 		    (void *)args->exceptfds, (void *)args->timeout);
524 #endif
525 
526 	/*
527 	 * Store current time for computation of the amount of
528 	 * time left.
529 	 */
530 	if (args->timeout) {
531 		if ((error = copyin(args->timeout, &ltv, sizeof(ltv))))
532 			goto select_out;
533 		utv.tv_sec = ltv.tv_sec;
534 		utv.tv_usec = ltv.tv_usec;
535 #ifdef DEBUG
536 		if (ldebug(select))
537 			printf(LMSG("incoming timeout (%jd/%ld)"),
538 			    (intmax_t)utv.tv_sec, utv.tv_usec);
539 #endif
540 
541 		if (itimerfix(&utv)) {
542 			/*
543 			 * The timeval was invalid.  Convert it to something
544 			 * valid that will act as it does under Linux.
545 			 */
546 			utv.tv_sec += utv.tv_usec / 1000000;
547 			utv.tv_usec %= 1000000;
548 			if (utv.tv_usec < 0) {
549 				utv.tv_sec -= 1;
550 				utv.tv_usec += 1000000;
551 			}
552 			if (utv.tv_sec < 0)
553 				timevalclear(&utv);
554 		}
555 		microtime(&tv0);
556 		tvp = &utv;
557 	} else
558 		tvp = NULL;
559 
560 	error = kern_select(td, args->nfds, args->readfds, args->writefds,
561 	    args->exceptfds, tvp, LINUX_NFDBITS);
562 
563 #ifdef DEBUG
564 	if (ldebug(select))
565 		printf(LMSG("real select returns %d"), error);
566 #endif
567 	if (error)
568 		goto select_out;
569 
570 	if (args->timeout) {
571 		if (td->td_retval[0]) {
572 			/*
573 			 * Compute how much time was left of the timeout,
574 			 * by subtracting the current time and the time
575 			 * before we started the call, and subtracting
576 			 * that result from the user-supplied value.
577 			 */
578 			microtime(&tv1);
579 			timevalsub(&tv1, &tv0);
580 			timevalsub(&utv, &tv1);
581 			if (utv.tv_sec < 0)
582 				timevalclear(&utv);
583 		} else
584 			timevalclear(&utv);
585 #ifdef DEBUG
586 		if (ldebug(select))
587 			printf(LMSG("outgoing timeout (%jd/%ld)"),
588 			    (intmax_t)utv.tv_sec, utv.tv_usec);
589 #endif
590 		ltv.tv_sec = utv.tv_sec;
591 		ltv.tv_usec = utv.tv_usec;
592 		if ((error = copyout(&ltv, args->timeout, sizeof(ltv))))
593 			goto select_out;
594 	}
595 
596 select_out:
597 #ifdef DEBUG
598 	if (ldebug(select))
599 		printf(LMSG("select_out -> %d"), error);
600 #endif
601 	return (error);
602 }
603 #endif
604 
605 int
606 linux_mremap(struct thread *td, struct linux_mremap_args *args)
607 {
608 	uintptr_t addr;
609 	size_t len;
610 	int error = 0;
611 
612 #ifdef DEBUG
613 	if (ldebug(mremap))
614 		printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
615 		    (void *)(uintptr_t)args->addr,
616 		    (unsigned long)args->old_len,
617 		    (unsigned long)args->new_len,
618 		    (unsigned long)args->flags);
619 #endif
620 
621 	if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
622 		td->td_retval[0] = 0;
623 		return (EINVAL);
624 	}
625 
626 	/*
627 	 * Check for the page alignment.
628 	 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
629 	 */
630 	if (args->addr & PAGE_MASK) {
631 		td->td_retval[0] = 0;
632 		return (EINVAL);
633 	}
634 
635 	args->new_len = round_page(args->new_len);
636 	args->old_len = round_page(args->old_len);
637 
638 	if (args->new_len > args->old_len) {
639 		td->td_retval[0] = 0;
640 		return (ENOMEM);
641 	}
642 
643 	if (args->new_len < args->old_len) {
644 		addr = args->addr + args->new_len;
645 		len = args->old_len - args->new_len;
646 		error = kern_munmap(td, addr, len);
647 	}
648 
649 	td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
650 	return (error);
651 }
652 
653 #define LINUX_MS_ASYNC       0x0001
654 #define LINUX_MS_INVALIDATE  0x0002
655 #define LINUX_MS_SYNC        0x0004
656 
657 int
658 linux_msync(struct thread *td, struct linux_msync_args *args)
659 {
660 
661 	return (kern_msync(td, args->addr, args->len,
662 	    args->fl & ~LINUX_MS_SYNC));
663 }
664 
665 #ifdef LINUX_LEGACY_SYSCALLS
666 int
667 linux_time(struct thread *td, struct linux_time_args *args)
668 {
669 	struct timeval tv;
670 	l_time_t tm;
671 	int error;
672 
673 #ifdef DEBUG
674 	if (ldebug(time))
675 		printf(ARGS(time, "*"));
676 #endif
677 
678 	microtime(&tv);
679 	tm = tv.tv_sec;
680 	if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
681 		return (error);
682 	td->td_retval[0] = tm;
683 	return (0);
684 }
685 #endif
686 
687 struct l_times_argv {
688 	l_clock_t	tms_utime;
689 	l_clock_t	tms_stime;
690 	l_clock_t	tms_cutime;
691 	l_clock_t	tms_cstime;
692 };
693 
694 
695 /*
696  * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
697  * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
698  * auxiliary vector entry.
699  */
700 #define	CLK_TCK		100
701 
702 #define	CONVOTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
703 #define	CONVNTCK(r)	(r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
704 
705 #define	CONVTCK(r)	(linux_kernver(td) >= LINUX_KERNVER_2004000 ?		\
706 			    CONVNTCK(r) : CONVOTCK(r))
707 
708 int
709 linux_times(struct thread *td, struct linux_times_args *args)
710 {
711 	struct timeval tv, utime, stime, cutime, cstime;
712 	struct l_times_argv tms;
713 	struct proc *p;
714 	int error;
715 
716 #ifdef DEBUG
717 	if (ldebug(times))
718 		printf(ARGS(times, "*"));
719 #endif
720 
721 	if (args->buf != NULL) {
722 		p = td->td_proc;
723 		PROC_LOCK(p);
724 		PROC_STATLOCK(p);
725 		calcru(p, &utime, &stime);
726 		PROC_STATUNLOCK(p);
727 		calccru(p, &cutime, &cstime);
728 		PROC_UNLOCK(p);
729 
730 		tms.tms_utime = CONVTCK(utime);
731 		tms.tms_stime = CONVTCK(stime);
732 
733 		tms.tms_cutime = CONVTCK(cutime);
734 		tms.tms_cstime = CONVTCK(cstime);
735 
736 		if ((error = copyout(&tms, args->buf, sizeof(tms))))
737 			return (error);
738 	}
739 
740 	microuptime(&tv);
741 	td->td_retval[0] = (int)CONVTCK(tv);
742 	return (0);
743 }
744 
745 int
746 linux_newuname(struct thread *td, struct linux_newuname_args *args)
747 {
748 	struct l_new_utsname utsname;
749 	char osname[LINUX_MAX_UTSNAME];
750 	char osrelease[LINUX_MAX_UTSNAME];
751 	char *p;
752 
753 #ifdef DEBUG
754 	if (ldebug(newuname))
755 		printf(ARGS(newuname, "*"));
756 #endif
757 
758 	linux_get_osname(td, osname);
759 	linux_get_osrelease(td, osrelease);
760 
761 	bzero(&utsname, sizeof(utsname));
762 	strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
763 	getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
764 	getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
765 	strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
766 	strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
767 	for (p = utsname.version; *p != '\0'; ++p)
768 		if (*p == '\n') {
769 			*p = '\0';
770 			break;
771 		}
772 	strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME);
773 
774 	return (copyout(&utsname, args->buf, sizeof(utsname)));
775 }
776 
777 struct l_utimbuf {
778 	l_time_t l_actime;
779 	l_time_t l_modtime;
780 };
781 
782 #ifdef LINUX_LEGACY_SYSCALLS
783 int
784 linux_utime(struct thread *td, struct linux_utime_args *args)
785 {
786 	struct timeval tv[2], *tvp;
787 	struct l_utimbuf lut;
788 	char *fname;
789 	int error;
790 
791 	LCONVPATHEXIST(td, args->fname, &fname);
792 
793 #ifdef DEBUG
794 	if (ldebug(utime))
795 		printf(ARGS(utime, "%s, *"), fname);
796 #endif
797 
798 	if (args->times) {
799 		if ((error = copyin(args->times, &lut, sizeof lut))) {
800 			LFREEPATH(fname);
801 			return (error);
802 		}
803 		tv[0].tv_sec = lut.l_actime;
804 		tv[0].tv_usec = 0;
805 		tv[1].tv_sec = lut.l_modtime;
806 		tv[1].tv_usec = 0;
807 		tvp = tv;
808 	} else
809 		tvp = NULL;
810 
811 	error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, tvp,
812 	    UIO_SYSSPACE);
813 	LFREEPATH(fname);
814 	return (error);
815 }
816 #endif
817 
818 #ifdef LINUX_LEGACY_SYSCALLS
819 int
820 linux_utimes(struct thread *td, struct linux_utimes_args *args)
821 {
822 	l_timeval ltv[2];
823 	struct timeval tv[2], *tvp = NULL;
824 	char *fname;
825 	int error;
826 
827 	LCONVPATHEXIST(td, args->fname, &fname);
828 
829 #ifdef DEBUG
830 	if (ldebug(utimes))
831 		printf(ARGS(utimes, "%s, *"), fname);
832 #endif
833 
834 	if (args->tptr != NULL) {
835 		if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
836 			LFREEPATH(fname);
837 			return (error);
838 		}
839 		tv[0].tv_sec = ltv[0].tv_sec;
840 		tv[0].tv_usec = ltv[0].tv_usec;
841 		tv[1].tv_sec = ltv[1].tv_sec;
842 		tv[1].tv_usec = ltv[1].tv_usec;
843 		tvp = tv;
844 	}
845 
846 	error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE,
847 	    tvp, UIO_SYSSPACE);
848 	LFREEPATH(fname);
849 	return (error);
850 }
851 #endif
852 
853 static int
854 linux_utimensat_nsec_valid(l_long nsec)
855 {
856 
857 	if (nsec == LINUX_UTIME_OMIT || nsec == LINUX_UTIME_NOW)
858 		return (0);
859 	if (nsec >= 0 && nsec <= 999999999)
860 		return (0);
861 	return (1);
862 }
863 
864 int
865 linux_utimensat(struct thread *td, struct linux_utimensat_args *args)
866 {
867 	struct l_timespec l_times[2];
868 	struct timespec times[2], *timesp = NULL;
869 	char *path = NULL;
870 	int error, dfd, flags = 0;
871 
872 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
873 
874 #ifdef DEBUG
875 	if (ldebug(utimensat))
876 		printf(ARGS(utimensat, "%d, *"), dfd);
877 #endif
878 
879 	if (args->flags & ~LINUX_AT_SYMLINK_NOFOLLOW)
880 		return (EINVAL);
881 
882 	if (args->times != NULL) {
883 		error = copyin(args->times, l_times, sizeof(l_times));
884 		if (error != 0)
885 			return (error);
886 
887 		if (linux_utimensat_nsec_valid(l_times[0].tv_nsec) != 0 ||
888 		    linux_utimensat_nsec_valid(l_times[1].tv_nsec) != 0)
889 			return (EINVAL);
890 
891 		times[0].tv_sec = l_times[0].tv_sec;
892 		switch (l_times[0].tv_nsec)
893 		{
894 		case LINUX_UTIME_OMIT:
895 			times[0].tv_nsec = UTIME_OMIT;
896 			break;
897 		case LINUX_UTIME_NOW:
898 			times[0].tv_nsec = UTIME_NOW;
899 			break;
900 		default:
901 			times[0].tv_nsec = l_times[0].tv_nsec;
902 		}
903 
904 		times[1].tv_sec = l_times[1].tv_sec;
905 		switch (l_times[1].tv_nsec)
906 		{
907 		case LINUX_UTIME_OMIT:
908 			times[1].tv_nsec = UTIME_OMIT;
909 			break;
910 		case LINUX_UTIME_NOW:
911 			times[1].tv_nsec = UTIME_NOW;
912 			break;
913 		default:
914 			times[1].tv_nsec = l_times[1].tv_nsec;
915 			break;
916 		}
917 		timesp = times;
918 
919 		/* This breaks POSIX, but is what the Linux kernel does
920 		 * _on purpose_ (documented in the man page for utimensat(2)),
921 		 * so we must follow that behaviour. */
922 		if (times[0].tv_nsec == UTIME_OMIT &&
923 		    times[1].tv_nsec == UTIME_OMIT)
924 			return (0);
925 	}
926 
927 	if (args->pathname != NULL)
928 		LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
929 	else if (args->flags != 0)
930 		return (EINVAL);
931 
932 	if (args->flags & LINUX_AT_SYMLINK_NOFOLLOW)
933 		flags |= AT_SYMLINK_NOFOLLOW;
934 
935 	if (path == NULL)
936 		error = kern_futimens(td, dfd, timesp, UIO_SYSSPACE);
937 	else {
938 		error = kern_utimensat(td, dfd, path, UIO_SYSSPACE, timesp,
939 			UIO_SYSSPACE, flags);
940 		LFREEPATH(path);
941 	}
942 
943 	return (error);
944 }
945 
946 #ifdef LINUX_LEGACY_SYSCALLS
947 int
948 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
949 {
950 	l_timeval ltv[2];
951 	struct timeval tv[2], *tvp = NULL;
952 	char *fname;
953 	int error, dfd;
954 
955 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
956 	LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
957 
958 #ifdef DEBUG
959 	if (ldebug(futimesat))
960 		printf(ARGS(futimesat, "%s, *"), fname);
961 #endif
962 
963 	if (args->utimes != NULL) {
964 		if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
965 			LFREEPATH(fname);
966 			return (error);
967 		}
968 		tv[0].tv_sec = ltv[0].tv_sec;
969 		tv[0].tv_usec = ltv[0].tv_usec;
970 		tv[1].tv_sec = ltv[1].tv_sec;
971 		tv[1].tv_usec = ltv[1].tv_usec;
972 		tvp = tv;
973 	}
974 
975 	error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
976 	LFREEPATH(fname);
977 	return (error);
978 }
979 #endif
980 
981 int
982 linux_common_wait(struct thread *td, int pid, int *status,
983     int options, struct rusage *ru)
984 {
985 	int error, tmpstat;
986 
987 	error = kern_wait(td, pid, &tmpstat, options, ru);
988 	if (error)
989 		return (error);
990 
991 	if (status) {
992 		tmpstat &= 0xffff;
993 		if (WIFSIGNALED(tmpstat))
994 			tmpstat = (tmpstat & 0xffffff80) |
995 			    bsd_to_linux_signal(WTERMSIG(tmpstat));
996 		else if (WIFSTOPPED(tmpstat))
997 			tmpstat = (tmpstat & 0xffff00ff) |
998 			    (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8);
999 		else if (WIFCONTINUED(tmpstat))
1000 			tmpstat = 0xffff;
1001 		error = copyout(&tmpstat, status, sizeof(int));
1002 	}
1003 
1004 	return (error);
1005 }
1006 
1007 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1008 int
1009 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
1010 {
1011 	struct linux_wait4_args wait4_args;
1012 
1013 #ifdef DEBUG
1014 	if (ldebug(waitpid))
1015 		printf(ARGS(waitpid, "%d, %p, %d"),
1016 		    args->pid, (void *)args->status, args->options);
1017 #endif
1018 
1019 	wait4_args.pid = args->pid;
1020 	wait4_args.status = args->status;
1021 	wait4_args.options = args->options;
1022 	wait4_args.rusage = NULL;
1023 
1024 	return (linux_wait4(td, &wait4_args));
1025 }
1026 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1027 
1028 int
1029 linux_wait4(struct thread *td, struct linux_wait4_args *args)
1030 {
1031 	int error, options;
1032 	struct rusage ru, *rup;
1033 
1034 #ifdef DEBUG
1035 	if (ldebug(wait4))
1036 		printf(ARGS(wait4, "%d, %p, %d, %p"),
1037 		    args->pid, (void *)args->status, args->options,
1038 		    (void *)args->rusage);
1039 #endif
1040 	if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG |
1041 	    LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
1042 		return (EINVAL);
1043 
1044 	options = WEXITED;
1045 	linux_to_bsd_waitopts(args->options, &options);
1046 
1047 	if (args->rusage != NULL)
1048 		rup = &ru;
1049 	else
1050 		rup = NULL;
1051 	error = linux_common_wait(td, args->pid, args->status, options, rup);
1052 	if (error != 0)
1053 		return (error);
1054 	if (args->rusage != NULL)
1055 		error = linux_copyout_rusage(&ru, args->rusage);
1056 	return (error);
1057 }
1058 
1059 int
1060 linux_waitid(struct thread *td, struct linux_waitid_args *args)
1061 {
1062 	int status, options, sig;
1063 	struct __wrusage wru;
1064 	siginfo_t siginfo;
1065 	l_siginfo_t lsi;
1066 	idtype_t idtype;
1067 	struct proc *p;
1068 	int error;
1069 
1070 	options = 0;
1071 	linux_to_bsd_waitopts(args->options, &options);
1072 
1073 	if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED))
1074 		return (EINVAL);
1075 	if (!(options & (WEXITED | WUNTRACED | WCONTINUED)))
1076 		return (EINVAL);
1077 
1078 	switch (args->idtype) {
1079 	case LINUX_P_ALL:
1080 		idtype = P_ALL;
1081 		break;
1082 	case LINUX_P_PID:
1083 		if (args->id <= 0)
1084 			return (EINVAL);
1085 		idtype = P_PID;
1086 		break;
1087 	case LINUX_P_PGID:
1088 		if (args->id <= 0)
1089 			return (EINVAL);
1090 		idtype = P_PGID;
1091 		break;
1092 	default:
1093 		return (EINVAL);
1094 	}
1095 
1096 	error = kern_wait6(td, idtype, args->id, &status, options,
1097 	    &wru, &siginfo);
1098 	if (error != 0)
1099 		return (error);
1100 	if (args->rusage != NULL) {
1101 		error = linux_copyout_rusage(&wru.wru_children,
1102 		    args->rusage);
1103 		if (error != 0)
1104 			return (error);
1105 	}
1106 	if (args->info != NULL) {
1107 		p = td->td_proc;
1108 		bzero(&lsi, sizeof(lsi));
1109 		if (td->td_retval[0] != 0) {
1110 			sig = bsd_to_linux_signal(siginfo.si_signo);
1111 			siginfo_to_lsiginfo(&siginfo, &lsi, sig);
1112 		}
1113 		error = copyout(&lsi, args->info, sizeof(lsi));
1114 	}
1115 	td->td_retval[0] = 0;
1116 
1117 	return (error);
1118 }
1119 
1120 #ifdef LINUX_LEGACY_SYSCALLS
1121 int
1122 linux_mknod(struct thread *td, struct linux_mknod_args *args)
1123 {
1124 	char *path;
1125 	int error;
1126 
1127 	LCONVPATHCREAT(td, args->path, &path);
1128 
1129 #ifdef DEBUG
1130 	if (ldebug(mknod))
1131 		printf(ARGS(mknod, "%s, %d, %ju"), path, args->mode,
1132 		    (uintmax_t)args->dev);
1133 #endif
1134 
1135 	switch (args->mode & S_IFMT) {
1136 	case S_IFIFO:
1137 	case S_IFSOCK:
1138 		error = kern_mkfifoat(td, AT_FDCWD, path, UIO_SYSSPACE,
1139 		    args->mode);
1140 		break;
1141 
1142 	case S_IFCHR:
1143 	case S_IFBLK:
1144 		error = kern_mknodat(td, AT_FDCWD, path, UIO_SYSSPACE,
1145 		    args->mode, args->dev);
1146 		break;
1147 
1148 	case S_IFDIR:
1149 		error = EPERM;
1150 		break;
1151 
1152 	case 0:
1153 		args->mode |= S_IFREG;
1154 		/* FALLTHROUGH */
1155 	case S_IFREG:
1156 		error = kern_openat(td, AT_FDCWD, path, UIO_SYSSPACE,
1157 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1158 		if (error == 0)
1159 			kern_close(td, td->td_retval[0]);
1160 		break;
1161 
1162 	default:
1163 		error = EINVAL;
1164 		break;
1165 	}
1166 	LFREEPATH(path);
1167 	return (error);
1168 }
1169 #endif
1170 
1171 int
1172 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
1173 {
1174 	char *path;
1175 	int error, dfd;
1176 
1177 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
1178 	LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
1179 
1180 #ifdef DEBUG
1181 	if (ldebug(mknodat))
1182 		printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev);
1183 #endif
1184 
1185 	switch (args->mode & S_IFMT) {
1186 	case S_IFIFO:
1187 	case S_IFSOCK:
1188 		error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
1189 		break;
1190 
1191 	case S_IFCHR:
1192 	case S_IFBLK:
1193 		error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
1194 		    args->dev);
1195 		break;
1196 
1197 	case S_IFDIR:
1198 		error = EPERM;
1199 		break;
1200 
1201 	case 0:
1202 		args->mode |= S_IFREG;
1203 		/* FALLTHROUGH */
1204 	case S_IFREG:
1205 		error = kern_openat(td, dfd, path, UIO_SYSSPACE,
1206 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1207 		if (error == 0)
1208 			kern_close(td, td->td_retval[0]);
1209 		break;
1210 
1211 	default:
1212 		error = EINVAL;
1213 		break;
1214 	}
1215 	LFREEPATH(path);
1216 	return (error);
1217 }
1218 
1219 /*
1220  * UGH! This is just about the dumbest idea I've ever heard!!
1221  */
1222 int
1223 linux_personality(struct thread *td, struct linux_personality_args *args)
1224 {
1225 	struct linux_pemuldata *pem;
1226 	struct proc *p = td->td_proc;
1227 	uint32_t old;
1228 
1229 #ifdef DEBUG
1230 	if (ldebug(personality))
1231 		printf(ARGS(personality, "%u"), args->per);
1232 #endif
1233 
1234 	PROC_LOCK(p);
1235 	pem = pem_find(p);
1236 	old = pem->persona;
1237 	if (args->per != 0xffffffff)
1238 		pem->persona = args->per;
1239 	PROC_UNLOCK(p);
1240 
1241 	td->td_retval[0] = old;
1242 	return (0);
1243 }
1244 
1245 struct l_itimerval {
1246 	l_timeval it_interval;
1247 	l_timeval it_value;
1248 };
1249 
1250 #define	B2L_ITIMERVAL(bip, lip)						\
1251 	(bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec;		\
1252 	(bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec;	\
1253 	(bip)->it_value.tv_sec = (lip)->it_value.tv_sec;		\
1254 	(bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1255 
1256 int
1257 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1258 {
1259 	int error;
1260 	struct l_itimerval ls;
1261 	struct itimerval aitv, oitv;
1262 
1263 #ifdef DEBUG
1264 	if (ldebug(setitimer))
1265 		printf(ARGS(setitimer, "%p, %p"),
1266 		    (void *)uap->itv, (void *)uap->oitv);
1267 #endif
1268 
1269 	if (uap->itv == NULL) {
1270 		uap->itv = uap->oitv;
1271 		return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1272 	}
1273 
1274 	error = copyin(uap->itv, &ls, sizeof(ls));
1275 	if (error != 0)
1276 		return (error);
1277 	B2L_ITIMERVAL(&aitv, &ls);
1278 #ifdef DEBUG
1279 	if (ldebug(setitimer)) {
1280 		printf("setitimer: value: sec: %jd, usec: %ld\n",
1281 		    (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1282 		printf("setitimer: interval: sec: %jd, usec: %ld\n",
1283 		    (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1284 	}
1285 #endif
1286 	error = kern_setitimer(td, uap->which, &aitv, &oitv);
1287 	if (error != 0 || uap->oitv == NULL)
1288 		return (error);
1289 	B2L_ITIMERVAL(&ls, &oitv);
1290 
1291 	return (copyout(&ls, uap->oitv, sizeof(ls)));
1292 }
1293 
1294 int
1295 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1296 {
1297 	int error;
1298 	struct l_itimerval ls;
1299 	struct itimerval aitv;
1300 
1301 #ifdef DEBUG
1302 	if (ldebug(getitimer))
1303 		printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1304 #endif
1305 	error = kern_getitimer(td, uap->which, &aitv);
1306 	if (error != 0)
1307 		return (error);
1308 	B2L_ITIMERVAL(&ls, &aitv);
1309 	return (copyout(&ls, uap->itv, sizeof(ls)));
1310 }
1311 
1312 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1313 int
1314 linux_nice(struct thread *td, struct linux_nice_args *args)
1315 {
1316 	struct setpriority_args bsd_args;
1317 
1318 	bsd_args.which = PRIO_PROCESS;
1319 	bsd_args.who = 0;		/* current process */
1320 	bsd_args.prio = args->inc;
1321 	return (sys_setpriority(td, &bsd_args));
1322 }
1323 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1324 
1325 int
1326 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1327 {
1328 	struct ucred *newcred, *oldcred;
1329 	l_gid_t *linux_gidset;
1330 	gid_t *bsd_gidset;
1331 	int ngrp, error;
1332 	struct proc *p;
1333 
1334 	ngrp = args->gidsetsize;
1335 	if (ngrp < 0 || ngrp >= ngroups_max + 1)
1336 		return (EINVAL);
1337 	linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK);
1338 	error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1339 	if (error)
1340 		goto out;
1341 	newcred = crget();
1342 	crextend(newcred, ngrp + 1);
1343 	p = td->td_proc;
1344 	PROC_LOCK(p);
1345 	oldcred = p->p_ucred;
1346 	crcopy(newcred, oldcred);
1347 
1348 	/*
1349 	 * cr_groups[0] holds egid. Setting the whole set from
1350 	 * the supplied set will cause egid to be changed too.
1351 	 * Keep cr_groups[0] unchanged to prevent that.
1352 	 */
1353 
1354 	if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS)) != 0) {
1355 		PROC_UNLOCK(p);
1356 		crfree(newcred);
1357 		goto out;
1358 	}
1359 
1360 	if (ngrp > 0) {
1361 		newcred->cr_ngroups = ngrp + 1;
1362 
1363 		bsd_gidset = newcred->cr_groups;
1364 		ngrp--;
1365 		while (ngrp >= 0) {
1366 			bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1367 			ngrp--;
1368 		}
1369 	} else
1370 		newcred->cr_ngroups = 1;
1371 
1372 	setsugid(p);
1373 	proc_set_cred(p, newcred);
1374 	PROC_UNLOCK(p);
1375 	crfree(oldcred);
1376 	error = 0;
1377 out:
1378 	free(linux_gidset, M_LINUX);
1379 	return (error);
1380 }
1381 
1382 int
1383 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1384 {
1385 	struct ucred *cred;
1386 	l_gid_t *linux_gidset;
1387 	gid_t *bsd_gidset;
1388 	int bsd_gidsetsz, ngrp, error;
1389 
1390 	cred = td->td_ucred;
1391 	bsd_gidset = cred->cr_groups;
1392 	bsd_gidsetsz = cred->cr_ngroups - 1;
1393 
1394 	/*
1395 	 * cr_groups[0] holds egid. Returning the whole set
1396 	 * here will cause a duplicate. Exclude cr_groups[0]
1397 	 * to prevent that.
1398 	 */
1399 
1400 	if ((ngrp = args->gidsetsize) == 0) {
1401 		td->td_retval[0] = bsd_gidsetsz;
1402 		return (0);
1403 	}
1404 
1405 	if (ngrp < bsd_gidsetsz)
1406 		return (EINVAL);
1407 
1408 	ngrp = 0;
1409 	linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1410 	    M_LINUX, M_WAITOK);
1411 	while (ngrp < bsd_gidsetsz) {
1412 		linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1413 		ngrp++;
1414 	}
1415 
1416 	error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1417 	free(linux_gidset, M_LINUX);
1418 	if (error)
1419 		return (error);
1420 
1421 	td->td_retval[0] = ngrp;
1422 	return (0);
1423 }
1424 
1425 int
1426 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1427 {
1428 	struct rlimit bsd_rlim;
1429 	struct l_rlimit rlim;
1430 	u_int which;
1431 	int error;
1432 
1433 #ifdef DEBUG
1434 	if (ldebug(setrlimit))
1435 		printf(ARGS(setrlimit, "%d, %p"),
1436 		    args->resource, (void *)args->rlim);
1437 #endif
1438 
1439 	if (args->resource >= LINUX_RLIM_NLIMITS)
1440 		return (EINVAL);
1441 
1442 	which = linux_to_bsd_resource[args->resource];
1443 	if (which == -1)
1444 		return (EINVAL);
1445 
1446 	error = copyin(args->rlim, &rlim, sizeof(rlim));
1447 	if (error)
1448 		return (error);
1449 
1450 	bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1451 	bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1452 	return (kern_setrlimit(td, which, &bsd_rlim));
1453 }
1454 
1455 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1456 int
1457 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1458 {
1459 	struct l_rlimit rlim;
1460 	struct rlimit bsd_rlim;
1461 	u_int which;
1462 
1463 #ifdef DEBUG
1464 	if (ldebug(old_getrlimit))
1465 		printf(ARGS(old_getrlimit, "%d, %p"),
1466 		    args->resource, (void *)args->rlim);
1467 #endif
1468 
1469 	if (args->resource >= LINUX_RLIM_NLIMITS)
1470 		return (EINVAL);
1471 
1472 	which = linux_to_bsd_resource[args->resource];
1473 	if (which == -1)
1474 		return (EINVAL);
1475 
1476 	lim_rlimit(td, which, &bsd_rlim);
1477 
1478 #ifdef COMPAT_LINUX32
1479 	rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1480 	if (rlim.rlim_cur == UINT_MAX)
1481 		rlim.rlim_cur = INT_MAX;
1482 	rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1483 	if (rlim.rlim_max == UINT_MAX)
1484 		rlim.rlim_max = INT_MAX;
1485 #else
1486 	rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1487 	if (rlim.rlim_cur == ULONG_MAX)
1488 		rlim.rlim_cur = LONG_MAX;
1489 	rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1490 	if (rlim.rlim_max == ULONG_MAX)
1491 		rlim.rlim_max = LONG_MAX;
1492 #endif
1493 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1494 }
1495 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1496 
1497 int
1498 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1499 {
1500 	struct l_rlimit rlim;
1501 	struct rlimit bsd_rlim;
1502 	u_int which;
1503 
1504 #ifdef DEBUG
1505 	if (ldebug(getrlimit))
1506 		printf(ARGS(getrlimit, "%d, %p"),
1507 		    args->resource, (void *)args->rlim);
1508 #endif
1509 
1510 	if (args->resource >= LINUX_RLIM_NLIMITS)
1511 		return (EINVAL);
1512 
1513 	which = linux_to_bsd_resource[args->resource];
1514 	if (which == -1)
1515 		return (EINVAL);
1516 
1517 	lim_rlimit(td, which, &bsd_rlim);
1518 
1519 	rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1520 	rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1521 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1522 }
1523 
1524 int
1525 linux_sched_setscheduler(struct thread *td,
1526     struct linux_sched_setscheduler_args *args)
1527 {
1528 	struct sched_param sched_param;
1529 	struct thread *tdt;
1530 	int error, policy;
1531 
1532 #ifdef DEBUG
1533 	if (ldebug(sched_setscheduler))
1534 		printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1535 		    args->pid, args->policy, (const void *)args->param);
1536 #endif
1537 
1538 	switch (args->policy) {
1539 	case LINUX_SCHED_OTHER:
1540 		policy = SCHED_OTHER;
1541 		break;
1542 	case LINUX_SCHED_FIFO:
1543 		policy = SCHED_FIFO;
1544 		break;
1545 	case LINUX_SCHED_RR:
1546 		policy = SCHED_RR;
1547 		break;
1548 	default:
1549 		return (EINVAL);
1550 	}
1551 
1552 	error = copyin(args->param, &sched_param, sizeof(sched_param));
1553 	if (error)
1554 		return (error);
1555 
1556 	tdt = linux_tdfind(td, args->pid, -1);
1557 	if (tdt == NULL)
1558 		return (ESRCH);
1559 
1560 	error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1561 	PROC_UNLOCK(tdt->td_proc);
1562 	return (error);
1563 }
1564 
1565 int
1566 linux_sched_getscheduler(struct thread *td,
1567     struct linux_sched_getscheduler_args *args)
1568 {
1569 	struct thread *tdt;
1570 	int error, policy;
1571 
1572 #ifdef DEBUG
1573 	if (ldebug(sched_getscheduler))
1574 		printf(ARGS(sched_getscheduler, "%d"), args->pid);
1575 #endif
1576 
1577 	tdt = linux_tdfind(td, args->pid, -1);
1578 	if (tdt == NULL)
1579 		return (ESRCH);
1580 
1581 	error = kern_sched_getscheduler(td, tdt, &policy);
1582 	PROC_UNLOCK(tdt->td_proc);
1583 
1584 	switch (policy) {
1585 	case SCHED_OTHER:
1586 		td->td_retval[0] = LINUX_SCHED_OTHER;
1587 		break;
1588 	case SCHED_FIFO:
1589 		td->td_retval[0] = LINUX_SCHED_FIFO;
1590 		break;
1591 	case SCHED_RR:
1592 		td->td_retval[0] = LINUX_SCHED_RR;
1593 		break;
1594 	}
1595 	return (error);
1596 }
1597 
1598 int
1599 linux_sched_get_priority_max(struct thread *td,
1600     struct linux_sched_get_priority_max_args *args)
1601 {
1602 	struct sched_get_priority_max_args bsd;
1603 
1604 #ifdef DEBUG
1605 	if (ldebug(sched_get_priority_max))
1606 		printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1607 #endif
1608 
1609 	switch (args->policy) {
1610 	case LINUX_SCHED_OTHER:
1611 		bsd.policy = SCHED_OTHER;
1612 		break;
1613 	case LINUX_SCHED_FIFO:
1614 		bsd.policy = SCHED_FIFO;
1615 		break;
1616 	case LINUX_SCHED_RR:
1617 		bsd.policy = SCHED_RR;
1618 		break;
1619 	default:
1620 		return (EINVAL);
1621 	}
1622 	return (sys_sched_get_priority_max(td, &bsd));
1623 }
1624 
1625 int
1626 linux_sched_get_priority_min(struct thread *td,
1627     struct linux_sched_get_priority_min_args *args)
1628 {
1629 	struct sched_get_priority_min_args bsd;
1630 
1631 #ifdef DEBUG
1632 	if (ldebug(sched_get_priority_min))
1633 		printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1634 #endif
1635 
1636 	switch (args->policy) {
1637 	case LINUX_SCHED_OTHER:
1638 		bsd.policy = SCHED_OTHER;
1639 		break;
1640 	case LINUX_SCHED_FIFO:
1641 		bsd.policy = SCHED_FIFO;
1642 		break;
1643 	case LINUX_SCHED_RR:
1644 		bsd.policy = SCHED_RR;
1645 		break;
1646 	default:
1647 		return (EINVAL);
1648 	}
1649 	return (sys_sched_get_priority_min(td, &bsd));
1650 }
1651 
1652 #define REBOOT_CAD_ON	0x89abcdef
1653 #define REBOOT_CAD_OFF	0
1654 #define REBOOT_HALT	0xcdef0123
1655 #define REBOOT_RESTART	0x01234567
1656 #define REBOOT_RESTART2	0xA1B2C3D4
1657 #define REBOOT_POWEROFF	0x4321FEDC
1658 #define REBOOT_MAGIC1	0xfee1dead
1659 #define REBOOT_MAGIC2	0x28121969
1660 #define REBOOT_MAGIC2A	0x05121996
1661 #define REBOOT_MAGIC2B	0x16041998
1662 
1663 int
1664 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1665 {
1666 	struct reboot_args bsd_args;
1667 
1668 #ifdef DEBUG
1669 	if (ldebug(reboot))
1670 		printf(ARGS(reboot, "0x%x"), args->cmd);
1671 #endif
1672 
1673 	if (args->magic1 != REBOOT_MAGIC1)
1674 		return (EINVAL);
1675 
1676 	switch (args->magic2) {
1677 	case REBOOT_MAGIC2:
1678 	case REBOOT_MAGIC2A:
1679 	case REBOOT_MAGIC2B:
1680 		break;
1681 	default:
1682 		return (EINVAL);
1683 	}
1684 
1685 	switch (args->cmd) {
1686 	case REBOOT_CAD_ON:
1687 	case REBOOT_CAD_OFF:
1688 		return (priv_check(td, PRIV_REBOOT));
1689 	case REBOOT_HALT:
1690 		bsd_args.opt = RB_HALT;
1691 		break;
1692 	case REBOOT_RESTART:
1693 	case REBOOT_RESTART2:
1694 		bsd_args.opt = 0;
1695 		break;
1696 	case REBOOT_POWEROFF:
1697 		bsd_args.opt = RB_POWEROFF;
1698 		break;
1699 	default:
1700 		return (EINVAL);
1701 	}
1702 	return (sys_reboot(td, &bsd_args));
1703 }
1704 
1705 
1706 /*
1707  * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1708  * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
1709  * are assumed to be preserved. The following lightweight syscalls fixes
1710  * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
1711  *
1712  * linux_getpid() - MP SAFE
1713  * linux_getgid() - MP SAFE
1714  * linux_getuid() - MP SAFE
1715  */
1716 
1717 int
1718 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1719 {
1720 
1721 #ifdef DEBUG
1722 	if (ldebug(getpid))
1723 		printf(ARGS(getpid, ""));
1724 #endif
1725 	td->td_retval[0] = td->td_proc->p_pid;
1726 
1727 	return (0);
1728 }
1729 
1730 int
1731 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1732 {
1733 	struct linux_emuldata *em;
1734 
1735 #ifdef DEBUG
1736 	if (ldebug(gettid))
1737 		printf(ARGS(gettid, ""));
1738 #endif
1739 
1740 	em = em_find(td);
1741 	KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1742 
1743 	td->td_retval[0] = em->em_tid;
1744 
1745 	return (0);
1746 }
1747 
1748 
1749 int
1750 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1751 {
1752 
1753 #ifdef DEBUG
1754 	if (ldebug(getppid))
1755 		printf(ARGS(getppid, ""));
1756 #endif
1757 
1758 	td->td_retval[0] = kern_getppid(td);
1759 	return (0);
1760 }
1761 
1762 int
1763 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1764 {
1765 
1766 #ifdef DEBUG
1767 	if (ldebug(getgid))
1768 		printf(ARGS(getgid, ""));
1769 #endif
1770 
1771 	td->td_retval[0] = td->td_ucred->cr_rgid;
1772 	return (0);
1773 }
1774 
1775 int
1776 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1777 {
1778 
1779 #ifdef DEBUG
1780 	if (ldebug(getuid))
1781 		printf(ARGS(getuid, ""));
1782 #endif
1783 
1784 	td->td_retval[0] = td->td_ucred->cr_ruid;
1785 	return (0);
1786 }
1787 
1788 
1789 int
1790 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1791 {
1792 	struct getsid_args bsd;
1793 
1794 #ifdef DEBUG
1795 	if (ldebug(getsid))
1796 		printf(ARGS(getsid, "%i"), args->pid);
1797 #endif
1798 
1799 	bsd.pid = args->pid;
1800 	return (sys_getsid(td, &bsd));
1801 }
1802 
1803 int
1804 linux_nosys(struct thread *td, struct nosys_args *ignore)
1805 {
1806 
1807 	return (ENOSYS);
1808 }
1809 
1810 int
1811 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1812 {
1813 	struct getpriority_args bsd_args;
1814 	int error;
1815 
1816 #ifdef DEBUG
1817 	if (ldebug(getpriority))
1818 		printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
1819 #endif
1820 
1821 	bsd_args.which = args->which;
1822 	bsd_args.who = args->who;
1823 	error = sys_getpriority(td, &bsd_args);
1824 	td->td_retval[0] = 20 - td->td_retval[0];
1825 	return (error);
1826 }
1827 
1828 int
1829 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1830 {
1831 	int name[2];
1832 
1833 #ifdef DEBUG
1834 	if (ldebug(sethostname))
1835 		printf(ARGS(sethostname, "*, %i"), args->len);
1836 #endif
1837 
1838 	name[0] = CTL_KERN;
1839 	name[1] = KERN_HOSTNAME;
1840 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1841 	    args->len, 0, 0));
1842 }
1843 
1844 int
1845 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1846 {
1847 	int name[2];
1848 
1849 #ifdef DEBUG
1850 	if (ldebug(setdomainname))
1851 		printf(ARGS(setdomainname, "*, %i"), args->len);
1852 #endif
1853 
1854 	name[0] = CTL_KERN;
1855 	name[1] = KERN_NISDOMAINNAME;
1856 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1857 	    args->len, 0, 0));
1858 }
1859 
1860 int
1861 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1862 {
1863 
1864 #ifdef DEBUG
1865 	if (ldebug(exit_group))
1866 		printf(ARGS(exit_group, "%i"), args->error_code);
1867 #endif
1868 
1869 	LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1870 	    args->error_code);
1871 
1872 	/*
1873 	 * XXX: we should send a signal to the parent if
1874 	 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1875 	 * as it doesnt occur often.
1876 	 */
1877 	exit1(td, args->error_code, 0);
1878 		/* NOTREACHED */
1879 }
1880 
1881 #define _LINUX_CAPABILITY_VERSION_1  0x19980330
1882 #define _LINUX_CAPABILITY_VERSION_2  0x20071026
1883 #define _LINUX_CAPABILITY_VERSION_3  0x20080522
1884 
1885 struct l_user_cap_header {
1886 	l_int	version;
1887 	l_int	pid;
1888 };
1889 
1890 struct l_user_cap_data {
1891 	l_int	effective;
1892 	l_int	permitted;
1893 	l_int	inheritable;
1894 };
1895 
1896 int
1897 linux_capget(struct thread *td, struct linux_capget_args *uap)
1898 {
1899 	struct l_user_cap_header luch;
1900 	struct l_user_cap_data lucd[2];
1901 	int error, u32s;
1902 
1903 	if (uap->hdrp == NULL)
1904 		return (EFAULT);
1905 
1906 	error = copyin(uap->hdrp, &luch, sizeof(luch));
1907 	if (error != 0)
1908 		return (error);
1909 
1910 	switch (luch.version) {
1911 	case _LINUX_CAPABILITY_VERSION_1:
1912 		u32s = 1;
1913 		break;
1914 	case _LINUX_CAPABILITY_VERSION_2:
1915 	case _LINUX_CAPABILITY_VERSION_3:
1916 		u32s = 2;
1917 		break;
1918 	default:
1919 #ifdef DEBUG
1920 		if (ldebug(capget))
1921 			printf(LMSG("invalid capget capability version 0x%x"),
1922 			    luch.version);
1923 #endif
1924 		luch.version = _LINUX_CAPABILITY_VERSION_1;
1925 		error = copyout(&luch, uap->hdrp, sizeof(luch));
1926 		if (error)
1927 			return (error);
1928 		return (EINVAL);
1929 	}
1930 
1931 	if (luch.pid)
1932 		return (EPERM);
1933 
1934 	if (uap->datap) {
1935 		/*
1936 		 * The current implementation doesn't support setting
1937 		 * a capability (it's essentially a stub) so indicate
1938 		 * that no capabilities are currently set or available
1939 		 * to request.
1940 		 */
1941 		memset(&lucd, 0, u32s * sizeof(lucd[0]));
1942 		error = copyout(&lucd, uap->datap, u32s * sizeof(lucd[0]));
1943 	}
1944 
1945 	return (error);
1946 }
1947 
1948 int
1949 linux_capset(struct thread *td, struct linux_capset_args *uap)
1950 {
1951 	struct l_user_cap_header luch;
1952 	struct l_user_cap_data lucd[2];
1953 	int error, i, u32s;
1954 
1955 	if (uap->hdrp == NULL || uap->datap == NULL)
1956 		return (EFAULT);
1957 
1958 	error = copyin(uap->hdrp, &luch, sizeof(luch));
1959 	if (error != 0)
1960 		return (error);
1961 
1962 	switch (luch.version) {
1963 	case _LINUX_CAPABILITY_VERSION_1:
1964 		u32s = 1;
1965 		break;
1966 	case _LINUX_CAPABILITY_VERSION_2:
1967 	case _LINUX_CAPABILITY_VERSION_3:
1968 		u32s = 2;
1969 		break;
1970 	default:
1971 #ifdef DEBUG
1972 		if (ldebug(capset))
1973 			printf(LMSG("invalid capset capability version 0x%x"),
1974 			    luch.version);
1975 #endif
1976 		luch.version = _LINUX_CAPABILITY_VERSION_1;
1977 		error = copyout(&luch, uap->hdrp, sizeof(luch));
1978 		if (error)
1979 			return (error);
1980 		return (EINVAL);
1981 	}
1982 
1983 	if (luch.pid)
1984 		return (EPERM);
1985 
1986 	error = copyin(uap->datap, &lucd, u32s * sizeof(lucd[0]));
1987 	if (error != 0)
1988 		return (error);
1989 
1990 	/* We currently don't support setting any capabilities. */
1991 	for (i = 0; i < u32s; i++) {
1992 		if (lucd[i].effective || lucd[i].permitted ||
1993 		    lucd[i].inheritable) {
1994 			linux_msg(td,
1995 			    "capset[%d] effective=0x%x, permitted=0x%x, "
1996 			    "inheritable=0x%x is not implemented", i,
1997 			    (int)lucd[i].effective, (int)lucd[i].permitted,
1998 			    (int)lucd[i].inheritable);
1999 			return (EPERM);
2000 		}
2001 	}
2002 
2003 	return (0);
2004 }
2005 
2006 int
2007 linux_prctl(struct thread *td, struct linux_prctl_args *args)
2008 {
2009 	int error = 0, max_size;
2010 	struct proc *p = td->td_proc;
2011 	char comm[LINUX_MAX_COMM_LEN];
2012 	int pdeath_signal;
2013 
2014 #ifdef DEBUG
2015 	if (ldebug(prctl))
2016 		printf(ARGS(prctl, "%d, %ju, %ju, %ju, %ju"), args->option,
2017 		    (uintmax_t)args->arg2, (uintmax_t)args->arg3,
2018 		    (uintmax_t)args->arg4, (uintmax_t)args->arg5);
2019 #endif
2020 
2021 	switch (args->option) {
2022 	case LINUX_PR_SET_PDEATHSIG:
2023 		if (!LINUX_SIG_VALID(args->arg2))
2024 			return (EINVAL);
2025 		pdeath_signal = linux_to_bsd_signal(args->arg2);
2026 		return (kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_CTL,
2027 		    &pdeath_signal));
2028 	case LINUX_PR_GET_PDEATHSIG:
2029 		error = kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_STATUS,
2030 		    &pdeath_signal);
2031 		if (error != 0)
2032 			return (error);
2033 		pdeath_signal = bsd_to_linux_signal(pdeath_signal);
2034 		return (copyout(&pdeath_signal,
2035 		    (void *)(register_t)args->arg2,
2036 		    sizeof(pdeath_signal)));
2037 		break;
2038 	case LINUX_PR_GET_KEEPCAPS:
2039 		/*
2040 		 * Indicate that we always clear the effective and
2041 		 * permitted capability sets when the user id becomes
2042 		 * non-zero (actually the capability sets are simply
2043 		 * always zero in the current implementation).
2044 		 */
2045 		td->td_retval[0] = 0;
2046 		break;
2047 	case LINUX_PR_SET_KEEPCAPS:
2048 		/*
2049 		 * Ignore requests to keep the effective and permitted
2050 		 * capability sets when the user id becomes non-zero.
2051 		 */
2052 		break;
2053 	case LINUX_PR_SET_NAME:
2054 		/*
2055 		 * To be on the safe side we need to make sure to not
2056 		 * overflow the size a Linux program expects. We already
2057 		 * do this here in the copyin, so that we don't need to
2058 		 * check on copyout.
2059 		 */
2060 		max_size = MIN(sizeof(comm), sizeof(p->p_comm));
2061 		error = copyinstr((void *)(register_t)args->arg2, comm,
2062 		    max_size, NULL);
2063 
2064 		/* Linux silently truncates the name if it is too long. */
2065 		if (error == ENAMETOOLONG) {
2066 			/*
2067 			 * XXX: copyinstr() isn't documented to populate the
2068 			 * array completely, so do a copyin() to be on the
2069 			 * safe side. This should be changed in case
2070 			 * copyinstr() is changed to guarantee this.
2071 			 */
2072 			error = copyin((void *)(register_t)args->arg2, comm,
2073 			    max_size - 1);
2074 			comm[max_size - 1] = '\0';
2075 		}
2076 		if (error)
2077 			return (error);
2078 
2079 		PROC_LOCK(p);
2080 		strlcpy(p->p_comm, comm, sizeof(p->p_comm));
2081 		PROC_UNLOCK(p);
2082 		break;
2083 	case LINUX_PR_GET_NAME:
2084 		PROC_LOCK(p);
2085 		strlcpy(comm, p->p_comm, sizeof(comm));
2086 		PROC_UNLOCK(p);
2087 		error = copyout(comm, (void *)(register_t)args->arg2,
2088 		    strlen(comm) + 1);
2089 		break;
2090 	default:
2091 		error = EINVAL;
2092 		break;
2093 	}
2094 
2095 	return (error);
2096 }
2097 
2098 int
2099 linux_sched_setparam(struct thread *td,
2100     struct linux_sched_setparam_args *uap)
2101 {
2102 	struct sched_param sched_param;
2103 	struct thread *tdt;
2104 	int error;
2105 
2106 #ifdef DEBUG
2107 	if (ldebug(sched_setparam))
2108 		printf(ARGS(sched_setparam, "%d, *"), uap->pid);
2109 #endif
2110 
2111 	error = copyin(uap->param, &sched_param, sizeof(sched_param));
2112 	if (error)
2113 		return (error);
2114 
2115 	tdt = linux_tdfind(td, uap->pid, -1);
2116 	if (tdt == NULL)
2117 		return (ESRCH);
2118 
2119 	error = kern_sched_setparam(td, tdt, &sched_param);
2120 	PROC_UNLOCK(tdt->td_proc);
2121 	return (error);
2122 }
2123 
2124 int
2125 linux_sched_getparam(struct thread *td,
2126     struct linux_sched_getparam_args *uap)
2127 {
2128 	struct sched_param sched_param;
2129 	struct thread *tdt;
2130 	int error;
2131 
2132 #ifdef DEBUG
2133 	if (ldebug(sched_getparam))
2134 		printf(ARGS(sched_getparam, "%d, *"), uap->pid);
2135 #endif
2136 
2137 	tdt = linux_tdfind(td, uap->pid, -1);
2138 	if (tdt == NULL)
2139 		return (ESRCH);
2140 
2141 	error = kern_sched_getparam(td, tdt, &sched_param);
2142 	PROC_UNLOCK(tdt->td_proc);
2143 	if (error == 0)
2144 		error = copyout(&sched_param, uap->param,
2145 		    sizeof(sched_param));
2146 	return (error);
2147 }
2148 
2149 /*
2150  * Get affinity of a process.
2151  */
2152 int
2153 linux_sched_getaffinity(struct thread *td,
2154     struct linux_sched_getaffinity_args *args)
2155 {
2156 	int error;
2157 	struct thread *tdt;
2158 
2159 #ifdef DEBUG
2160 	if (ldebug(sched_getaffinity))
2161 		printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid,
2162 		    args->len);
2163 #endif
2164 	if (args->len < sizeof(cpuset_t))
2165 		return (EINVAL);
2166 
2167 	tdt = linux_tdfind(td, args->pid, -1);
2168 	if (tdt == NULL)
2169 		return (ESRCH);
2170 
2171 	PROC_UNLOCK(tdt->td_proc);
2172 
2173 	error = kern_cpuset_getaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2174 	    tdt->td_tid, sizeof(cpuset_t), (cpuset_t *)args->user_mask_ptr);
2175 	if (error == 0)
2176 		td->td_retval[0] = sizeof(cpuset_t);
2177 
2178 	return (error);
2179 }
2180 
2181 /*
2182  *  Set affinity of a process.
2183  */
2184 int
2185 linux_sched_setaffinity(struct thread *td,
2186     struct linux_sched_setaffinity_args *args)
2187 {
2188 	struct thread *tdt;
2189 
2190 #ifdef DEBUG
2191 	if (ldebug(sched_setaffinity))
2192 		printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid,
2193 		    args->len);
2194 #endif
2195 	if (args->len < sizeof(cpuset_t))
2196 		return (EINVAL);
2197 
2198 	tdt = linux_tdfind(td, args->pid, -1);
2199 	if (tdt == NULL)
2200 		return (ESRCH);
2201 
2202 	PROC_UNLOCK(tdt->td_proc);
2203 
2204 	return (kern_cpuset_setaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2205 	    tdt->td_tid, sizeof(cpuset_t), (cpuset_t *) args->user_mask_ptr));
2206 }
2207 
2208 struct linux_rlimit64 {
2209 	uint64_t	rlim_cur;
2210 	uint64_t	rlim_max;
2211 };
2212 
2213 int
2214 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
2215 {
2216 	struct rlimit rlim, nrlim;
2217 	struct linux_rlimit64 lrlim;
2218 	struct proc *p;
2219 	u_int which;
2220 	int flags;
2221 	int error;
2222 
2223 #ifdef DEBUG
2224 	if (ldebug(prlimit64))
2225 		printf(ARGS(prlimit64, "%d, %d, %p, %p"), args->pid,
2226 		    args->resource, (void *)args->new, (void *)args->old);
2227 #endif
2228 
2229 	if (args->resource >= LINUX_RLIM_NLIMITS)
2230 		return (EINVAL);
2231 
2232 	which = linux_to_bsd_resource[args->resource];
2233 	if (which == -1)
2234 		return (EINVAL);
2235 
2236 	if (args->new != NULL) {
2237 		/*
2238 		 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
2239 		 * rlim is unsigned 64-bit. FreeBSD treats negative limits
2240 		 * as INFINITY so we do not need a conversion even.
2241 		 */
2242 		error = copyin(args->new, &nrlim, sizeof(nrlim));
2243 		if (error != 0)
2244 			return (error);
2245 	}
2246 
2247 	flags = PGET_HOLD | PGET_NOTWEXIT;
2248 	if (args->new != NULL)
2249 		flags |= PGET_CANDEBUG;
2250 	else
2251 		flags |= PGET_CANSEE;
2252 	error = pget(args->pid, flags, &p);
2253 	if (error != 0)
2254 		return (error);
2255 
2256 	if (args->old != NULL) {
2257 		PROC_LOCK(p);
2258 		lim_rlimit_proc(p, which, &rlim);
2259 		PROC_UNLOCK(p);
2260 		if (rlim.rlim_cur == RLIM_INFINITY)
2261 			lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2262 		else
2263 			lrlim.rlim_cur = rlim.rlim_cur;
2264 		if (rlim.rlim_max == RLIM_INFINITY)
2265 			lrlim.rlim_max = LINUX_RLIM_INFINITY;
2266 		else
2267 			lrlim.rlim_max = rlim.rlim_max;
2268 		error = copyout(&lrlim, args->old, sizeof(lrlim));
2269 		if (error != 0)
2270 			goto out;
2271 	}
2272 
2273 	if (args->new != NULL)
2274 		error = kern_proc_setrlimit(td, p, which, &nrlim);
2275 
2276  out:
2277 	PRELE(p);
2278 	return (error);
2279 }
2280 
2281 int
2282 linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2283 {
2284 	struct timeval utv, tv0, tv1, *tvp;
2285 	struct l_pselect6arg lpse6;
2286 	struct l_timespec lts;
2287 	struct timespec uts;
2288 	l_sigset_t l_ss;
2289 	sigset_t *ssp;
2290 	sigset_t ss;
2291 	int error;
2292 
2293 	ssp = NULL;
2294 	if (args->sig != NULL) {
2295 		error = copyin(args->sig, &lpse6, sizeof(lpse6));
2296 		if (error != 0)
2297 			return (error);
2298 		if (lpse6.ss_len != sizeof(l_ss))
2299 			return (EINVAL);
2300 		if (lpse6.ss != 0) {
2301 			error = copyin(PTRIN(lpse6.ss), &l_ss,
2302 			    sizeof(l_ss));
2303 			if (error != 0)
2304 				return (error);
2305 			linux_to_bsd_sigset(&l_ss, &ss);
2306 			ssp = &ss;
2307 		}
2308 	}
2309 
2310 	/*
2311 	 * Currently glibc changes nanosecond number to microsecond.
2312 	 * This mean losing precision but for now it is hardly seen.
2313 	 */
2314 	if (args->tsp != NULL) {
2315 		error = copyin(args->tsp, &lts, sizeof(lts));
2316 		if (error != 0)
2317 			return (error);
2318 		error = linux_to_native_timespec(&uts, &lts);
2319 		if (error != 0)
2320 			return (error);
2321 
2322 		TIMESPEC_TO_TIMEVAL(&utv, &uts);
2323 		if (itimerfix(&utv))
2324 			return (EINVAL);
2325 
2326 		microtime(&tv0);
2327 		tvp = &utv;
2328 	} else
2329 		tvp = NULL;
2330 
2331 	error = kern_pselect(td, args->nfds, args->readfds, args->writefds,
2332 	    args->exceptfds, tvp, ssp, LINUX_NFDBITS);
2333 
2334 	if (error == 0 && args->tsp != NULL) {
2335 		if (td->td_retval[0] != 0) {
2336 			/*
2337 			 * Compute how much time was left of the timeout,
2338 			 * by subtracting the current time and the time
2339 			 * before we started the call, and subtracting
2340 			 * that result from the user-supplied value.
2341 			 */
2342 
2343 			microtime(&tv1);
2344 			timevalsub(&tv1, &tv0);
2345 			timevalsub(&utv, &tv1);
2346 			if (utv.tv_sec < 0)
2347 				timevalclear(&utv);
2348 		} else
2349 			timevalclear(&utv);
2350 
2351 		TIMEVAL_TO_TIMESPEC(&utv, &uts);
2352 
2353 		error = native_to_linux_timespec(&lts, &uts);
2354 		if (error == 0)
2355 			error = copyout(&lts, args->tsp, sizeof(lts));
2356 	}
2357 
2358 	return (error);
2359 }
2360 
2361 int
2362 linux_ppoll(struct thread *td, struct linux_ppoll_args *args)
2363 {
2364 	struct timespec ts0, ts1;
2365 	struct l_timespec lts;
2366 	struct timespec uts, *tsp;
2367 	l_sigset_t l_ss;
2368 	sigset_t *ssp;
2369 	sigset_t ss;
2370 	int error;
2371 
2372 	if (args->sset != NULL) {
2373 		if (args->ssize != sizeof(l_ss))
2374 			return (EINVAL);
2375 		error = copyin(args->sset, &l_ss, sizeof(l_ss));
2376 		if (error)
2377 			return (error);
2378 		linux_to_bsd_sigset(&l_ss, &ss);
2379 		ssp = &ss;
2380 	} else
2381 		ssp = NULL;
2382 	if (args->tsp != NULL) {
2383 		error = copyin(args->tsp, &lts, sizeof(lts));
2384 		if (error)
2385 			return (error);
2386 		error = linux_to_native_timespec(&uts, &lts);
2387 		if (error != 0)
2388 			return (error);
2389 
2390 		nanotime(&ts0);
2391 		tsp = &uts;
2392 	} else
2393 		tsp = NULL;
2394 
2395 	error = kern_poll(td, args->fds, args->nfds, tsp, ssp);
2396 
2397 	if (error == 0 && args->tsp != NULL) {
2398 		if (td->td_retval[0]) {
2399 			nanotime(&ts1);
2400 			timespecsub(&ts1, &ts0, &ts1);
2401 			timespecsub(&uts, &ts1, &uts);
2402 			if (uts.tv_sec < 0)
2403 				timespecclear(&uts);
2404 		} else
2405 			timespecclear(&uts);
2406 
2407 		error = native_to_linux_timespec(&lts, &uts);
2408 		if (error == 0)
2409 			error = copyout(&lts, args->tsp, sizeof(lts));
2410 	}
2411 
2412 	return (error);
2413 }
2414 
2415 #if defined(DEBUG) || defined(KTR)
2416 /* XXX: can be removed when every ldebug(...) and KTR stuff are removed. */
2417 
2418 #ifdef COMPAT_LINUX32
2419 #define	L_MAXSYSCALL	LINUX32_SYS_MAXSYSCALL
2420 #else
2421 #define	L_MAXSYSCALL	LINUX_SYS_MAXSYSCALL
2422 #endif
2423 
2424 u_char linux_debug_map[howmany(L_MAXSYSCALL, sizeof(u_char))];
2425 
2426 static int
2427 linux_debug(int syscall, int toggle, int global)
2428 {
2429 
2430 	if (global) {
2431 		char c = toggle ? 0 : 0xff;
2432 
2433 		memset(linux_debug_map, c, sizeof(linux_debug_map));
2434 		return (0);
2435 	}
2436 	if (syscall < 0 || syscall >= L_MAXSYSCALL)
2437 		return (EINVAL);
2438 	if (toggle)
2439 		clrbit(linux_debug_map, syscall);
2440 	else
2441 		setbit(linux_debug_map, syscall);
2442 	return (0);
2443 }
2444 #undef L_MAXSYSCALL
2445 
2446 /*
2447  * Usage: sysctl linux.debug=<syscall_nr>.<0/1>
2448  *
2449  *    E.g.: sysctl linux.debug=21.0
2450  *
2451  * As a special case, syscall "all" will apply to all syscalls globally.
2452  */
2453 #define LINUX_MAX_DEBUGSTR	16
2454 int
2455 linux_sysctl_debug(SYSCTL_HANDLER_ARGS)
2456 {
2457 	char value[LINUX_MAX_DEBUGSTR], *p;
2458 	int error, sysc, toggle;
2459 	int global = 0;
2460 
2461 	value[0] = '\0';
2462 	error = sysctl_handle_string(oidp, value, LINUX_MAX_DEBUGSTR, req);
2463 	if (error || req->newptr == NULL)
2464 		return (error);
2465 	for (p = value; *p != '\0' && *p != '.'; p++);
2466 	if (*p == '\0')
2467 		return (EINVAL);
2468 	*p++ = '\0';
2469 	sysc = strtol(value, NULL, 0);
2470 	toggle = strtol(p, NULL, 0);
2471 	if (strcmp(value, "all") == 0)
2472 		global = 1;
2473 	error = linux_debug(sysc, toggle, global);
2474 	return (error);
2475 }
2476 
2477 #endif /* DEBUG || KTR */
2478 
2479 int
2480 linux_sched_rr_get_interval(struct thread *td,
2481     struct linux_sched_rr_get_interval_args *uap)
2482 {
2483 	struct timespec ts;
2484 	struct l_timespec lts;
2485 	struct thread *tdt;
2486 	int error;
2487 
2488 	/*
2489 	 * According to man in case the invalid pid specified
2490 	 * EINVAL should be returned.
2491 	 */
2492 	if (uap->pid < 0)
2493 		return (EINVAL);
2494 
2495 	tdt = linux_tdfind(td, uap->pid, -1);
2496 	if (tdt == NULL)
2497 		return (ESRCH);
2498 
2499 	error = kern_sched_rr_get_interval_td(td, tdt, &ts);
2500 	PROC_UNLOCK(tdt->td_proc);
2501 	if (error != 0)
2502 		return (error);
2503 	error = native_to_linux_timespec(&lts, &ts);
2504 	if (error != 0)
2505 		return (error);
2506 	return (copyout(&lts, uap->interval, sizeof(lts)));
2507 }
2508 
2509 /*
2510  * In case when the Linux thread is the initial thread in
2511  * the thread group thread id is equal to the process id.
2512  * Glibc depends on this magic (assert in pthread_getattr_np.c).
2513  */
2514 struct thread *
2515 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid)
2516 {
2517 	struct linux_emuldata *em;
2518 	struct thread *tdt;
2519 	struct proc *p;
2520 
2521 	tdt = NULL;
2522 	if (tid == 0 || tid == td->td_tid) {
2523 		tdt = td;
2524 		PROC_LOCK(tdt->td_proc);
2525 	} else if (tid > PID_MAX)
2526 		tdt = tdfind(tid, pid);
2527 	else {
2528 		/*
2529 		 * Initial thread where the tid equal to the pid.
2530 		 */
2531 		p = pfind(tid);
2532 		if (p != NULL) {
2533 			if (SV_PROC_ABI(p) != SV_ABI_LINUX) {
2534 				/*
2535 				 * p is not a Linuxulator process.
2536 				 */
2537 				PROC_UNLOCK(p);
2538 				return (NULL);
2539 			}
2540 			FOREACH_THREAD_IN_PROC(p, tdt) {
2541 				em = em_find(tdt);
2542 				if (tid == em->em_tid)
2543 					return (tdt);
2544 			}
2545 			PROC_UNLOCK(p);
2546 		}
2547 		return (NULL);
2548 	}
2549 
2550 	return (tdt);
2551 }
2552 
2553 void
2554 linux_to_bsd_waitopts(int options, int *bsdopts)
2555 {
2556 
2557 	if (options & LINUX_WNOHANG)
2558 		*bsdopts |= WNOHANG;
2559 	if (options & LINUX_WUNTRACED)
2560 		*bsdopts |= WUNTRACED;
2561 	if (options & LINUX_WEXITED)
2562 		*bsdopts |= WEXITED;
2563 	if (options & LINUX_WCONTINUED)
2564 		*bsdopts |= WCONTINUED;
2565 	if (options & LINUX_WNOWAIT)
2566 		*bsdopts |= WNOWAIT;
2567 
2568 	if (options & __WCLONE)
2569 		*bsdopts |= WLINUXCLONE;
2570 }
2571 
2572 int
2573 linux_getrandom(struct thread *td, struct linux_getrandom_args *args)
2574 {
2575 	struct uio uio;
2576 	struct iovec iov;
2577 	int error;
2578 
2579 	if (args->flags & ~(LINUX_GRND_NONBLOCK|LINUX_GRND_RANDOM))
2580 		return (EINVAL);
2581 	if (args->count > INT_MAX)
2582 		args->count = INT_MAX;
2583 
2584 	iov.iov_base = args->buf;
2585 	iov.iov_len = args->count;
2586 
2587 	uio.uio_iov = &iov;
2588 	uio.uio_iovcnt = 1;
2589 	uio.uio_resid = iov.iov_len;
2590 	uio.uio_segflg = UIO_USERSPACE;
2591 	uio.uio_rw = UIO_READ;
2592 	uio.uio_td = td;
2593 
2594 	error = read_random_uio(&uio, args->flags & LINUX_GRND_NONBLOCK);
2595 	if (error == 0)
2596 		td->td_retval[0] = args->count - uio.uio_resid;
2597 	return (error);
2598 }
2599 
2600 int
2601 linux_mincore(struct thread *td, struct linux_mincore_args *args)
2602 {
2603 
2604 	/* Needs to be page-aligned */
2605 	if (args->start & PAGE_MASK)
2606 		return (EINVAL);
2607 	return (kern_mincore(td, args->start, args->len, args->vec));
2608 }
2609