xref: /freebsd/sys/compat/linux/linux_misc.c (revision 3e7865bcfcf79abd482da6144348218137ccbaaa)
1 /*-
2  * Copyright (c) 2002 Doug Rabson
3  * Copyright (c) 1994-1995 S�ren Schmidt
4  * All rights reserved.
5  *
6  * Redistribution and use in source and binary forms, with or without
7  * modification, are permitted provided that the following conditions
8  * are met:
9  * 1. Redistributions of source code must retain the above copyright
10  *    notice, this list of conditions and the following disclaimer
11  *    in this position and unchanged.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. The name of the author may not be used to endorse or promote products
16  *    derived from this software without specific prior written permission
17  *
18  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
19  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
20  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
21  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
22  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
23  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
24  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
25  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
26  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
27  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
28  */
29 
30 #include <sys/cdefs.h>
31 __FBSDID("$FreeBSD$");
32 
33 #include "opt_compat.h"
34 
35 #include <sys/param.h>
36 #include <sys/blist.h>
37 #include <sys/fcntl.h>
38 #if defined(__i386__)
39 #include <sys/imgact_aout.h>
40 #endif
41 #include <sys/jail.h>
42 #include <sys/kernel.h>
43 #include <sys/limits.h>
44 #include <sys/lock.h>
45 #include <sys/malloc.h>
46 #include <sys/mman.h>
47 #include <sys/mount.h>
48 #include <sys/mutex.h>
49 #include <sys/namei.h>
50 #include <sys/priv.h>
51 #include <sys/proc.h>
52 #include <sys/reboot.h>
53 #include <sys/resourcevar.h>
54 #include <sys/sched.h>
55 #include <sys/signalvar.h>
56 #include <sys/stat.h>
57 #include <sys/syscallsubr.h>
58 #include <sys/sysctl.h>
59 #include <sys/sysproto.h>
60 #include <sys/systm.h>
61 #include <sys/time.h>
62 #include <sys/vmmeter.h>
63 #include <sys/vnode.h>
64 #include <sys/wait.h>
65 #include <sys/cpuset.h>
66 
67 #include <security/mac/mac_framework.h>
68 
69 #include <vm/vm.h>
70 #include <vm/pmap.h>
71 #include <vm/vm_kern.h>
72 #include <vm/vm_map.h>
73 #include <vm/vm_extern.h>
74 #include <vm/vm_object.h>
75 #include <vm/swap_pager.h>
76 
77 #ifdef COMPAT_LINUX32
78 #include <machine/../linux32/linux.h>
79 #include <machine/../linux32/linux32_proto.h>
80 #else
81 #include <machine/../linux/linux.h>
82 #include <machine/../linux/linux_proto.h>
83 #endif
84 
85 #include <compat/linux/linux_file.h>
86 #include <compat/linux/linux_mib.h>
87 #include <compat/linux/linux_signal.h>
88 #include <compat/linux/linux_util.h>
89 #include <compat/linux/linux_sysproto.h>
90 #include <compat/linux/linux_emul.h>
91 #include <compat/linux/linux_misc.h>
92 
93 int stclohz;				/* Statistics clock frequency */
94 
95 #define BSD_TO_LINUX_SIGNAL(sig)	\
96 	(((sig) <= LINUX_SIGTBLSZ) ? bsd_to_linux_signal[_SIG_IDX(sig)] : sig)
97 
98 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
99 	RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
100 	RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
101 	RLIMIT_MEMLOCK, RLIMIT_AS
102 };
103 
104 struct l_sysinfo {
105 	l_long		uptime;		/* Seconds since boot */
106 	l_ulong		loads[3];	/* 1, 5, and 15 minute load averages */
107 #define LINUX_SYSINFO_LOADS_SCALE 65536
108 	l_ulong		totalram;	/* Total usable main memory size */
109 	l_ulong		freeram;	/* Available memory size */
110 	l_ulong		sharedram;	/* Amount of shared memory */
111 	l_ulong		bufferram;	/* Memory used by buffers */
112 	l_ulong		totalswap;	/* Total swap space size */
113 	l_ulong		freeswap;	/* swap space still available */
114 	l_ushort	procs;		/* Number of current processes */
115 	l_ushort	pads;
116 	l_ulong		totalbig;
117 	l_ulong		freebig;
118 	l_uint		mem_unit;
119 	char		_f[20-2*sizeof(l_long)-sizeof(l_int)];	/* padding */
120 };
121 int
122 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
123 {
124 	struct l_sysinfo sysinfo;
125 	vm_object_t object;
126 	int i, j;
127 	struct timespec ts;
128 
129 	getnanouptime(&ts);
130 	if (ts.tv_nsec != 0)
131 		ts.tv_sec++;
132 	sysinfo.uptime = ts.tv_sec;
133 
134 	/* Use the information from the mib to get our load averages */
135 	for (i = 0; i < 3; i++)
136 		sysinfo.loads[i] = averunnable.ldavg[i] *
137 		    LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
138 
139 	sysinfo.totalram = physmem * PAGE_SIZE;
140 	sysinfo.freeram = sysinfo.totalram - cnt.v_wire_count * PAGE_SIZE;
141 
142 	sysinfo.sharedram = 0;
143 	mtx_lock(&vm_object_list_mtx);
144 	TAILQ_FOREACH(object, &vm_object_list, object_list)
145 		if (object->shadow_count > 1)
146 			sysinfo.sharedram += object->resident_page_count;
147 	mtx_unlock(&vm_object_list_mtx);
148 
149 	sysinfo.sharedram *= PAGE_SIZE;
150 	sysinfo.bufferram = 0;
151 
152 	swap_pager_status(&i, &j);
153 	sysinfo.totalswap = i * PAGE_SIZE;
154 	sysinfo.freeswap = (i - j) * PAGE_SIZE;
155 
156 	sysinfo.procs = nprocs;
157 
158 	/* The following are only present in newer Linux kernels. */
159 	sysinfo.totalbig = 0;
160 	sysinfo.freebig = 0;
161 	sysinfo.mem_unit = 1;
162 
163 	return copyout(&sysinfo, args->info, sizeof(sysinfo));
164 }
165 
166 int
167 linux_alarm(struct thread *td, struct linux_alarm_args *args)
168 {
169 	struct itimerval it, old_it;
170 	u_int secs;
171 	int error;
172 
173 #ifdef DEBUG
174 	if (ldebug(alarm))
175 		printf(ARGS(alarm, "%u"), args->secs);
176 #endif
177 
178 	secs = args->secs;
179 
180 	if (secs > INT_MAX)
181 		secs = INT_MAX;
182 
183 	it.it_value.tv_sec = (long) secs;
184 	it.it_value.tv_usec = 0;
185 	it.it_interval.tv_sec = 0;
186 	it.it_interval.tv_usec = 0;
187 	error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
188 	if (error)
189 		return (error);
190 	if (timevalisset(&old_it.it_value)) {
191 		if (old_it.it_value.tv_usec != 0)
192 			old_it.it_value.tv_sec++;
193 		td->td_retval[0] = old_it.it_value.tv_sec;
194 	}
195 	return (0);
196 }
197 
198 int
199 linux_brk(struct thread *td, struct linux_brk_args *args)
200 {
201 	struct vmspace *vm = td->td_proc->p_vmspace;
202 	vm_offset_t new, old;
203 	struct obreak_args /* {
204 		char * nsize;
205 	} */ tmp;
206 
207 #ifdef DEBUG
208 	if (ldebug(brk))
209 		printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
210 #endif
211 	old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize);
212 	new = (vm_offset_t)args->dsend;
213 	tmp.nsize = (char *)new;
214 	if (((caddr_t)new > vm->vm_daddr) && !obreak(td, &tmp))
215 		td->td_retval[0] = (long)new;
216 	else
217 		td->td_retval[0] = (long)old;
218 
219 	return 0;
220 }
221 
222 #if defined(__i386__)
223 /* XXX: what about amd64/linux32? */
224 
225 int
226 linux_uselib(struct thread *td, struct linux_uselib_args *args)
227 {
228 	struct nameidata ni;
229 	struct vnode *vp;
230 	struct exec *a_out;
231 	struct vattr attr;
232 	vm_offset_t vmaddr;
233 	unsigned long file_offset;
234 	vm_offset_t buffer;
235 	unsigned long bss_size;
236 	char *library;
237 	int error;
238 	int locked, vfslocked;
239 
240 	LCONVPATHEXIST(td, args->library, &library);
241 
242 #ifdef DEBUG
243 	if (ldebug(uselib))
244 		printf(ARGS(uselib, "%s"), library);
245 #endif
246 
247 	a_out = NULL;
248 	vfslocked = 0;
249 	locked = 0;
250 	vp = NULL;
251 
252 	NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
253 	    UIO_SYSSPACE, library, td);
254 	error = namei(&ni);
255 	LFREEPATH(library);
256 	if (error)
257 		goto cleanup;
258 
259 	vp = ni.ni_vp;
260 	vfslocked = NDHASGIANT(&ni);
261 	NDFREE(&ni, NDF_ONLY_PNBUF);
262 
263 	/*
264 	 * From here on down, we have a locked vnode that must be unlocked.
265 	 * XXX: The code below largely duplicates exec_check_permissions().
266 	 */
267 	locked = 1;
268 
269 	/* Writable? */
270 	if (vp->v_writecount) {
271 		error = ETXTBSY;
272 		goto cleanup;
273 	}
274 
275 	/* Executable? */
276 	error = VOP_GETATTR(vp, &attr, td->td_ucred);
277 	if (error)
278 		goto cleanup;
279 
280 	if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
281 	    ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
282 		/* EACCESS is what exec(2) returns. */
283 		error = ENOEXEC;
284 		goto cleanup;
285 	}
286 
287 	/* Sensible size? */
288 	if (attr.va_size == 0) {
289 		error = ENOEXEC;
290 		goto cleanup;
291 	}
292 
293 	/* Can we access it? */
294 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
295 	if (error)
296 		goto cleanup;
297 
298 	/*
299 	 * XXX: This should use vn_open() so that it is properly authorized,
300 	 * and to reduce code redundancy all over the place here.
301 	 * XXX: Not really, it duplicates far more of exec_check_permissions()
302 	 * than vn_open().
303 	 */
304 #ifdef MAC
305 	error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
306 	if (error)
307 		goto cleanup;
308 #endif
309 	error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
310 	if (error)
311 		goto cleanup;
312 
313 	/* Pull in executable header into kernel_map */
314 	error = vm_mmap(kernel_map, (vm_offset_t *)&a_out, PAGE_SIZE,
315 	    VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
316 	if (error)
317 		goto cleanup;
318 
319 	/* Is it a Linux binary ? */
320 	if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
321 		error = ENOEXEC;
322 		goto cleanup;
323 	}
324 
325 	/*
326 	 * While we are here, we should REALLY do some more checks
327 	 */
328 
329 	/* Set file/virtual offset based on a.out variant. */
330 	switch ((int)(a_out->a_magic & 0xffff)) {
331 	case 0413:			/* ZMAGIC */
332 		file_offset = 1024;
333 		break;
334 	case 0314:			/* QMAGIC */
335 		file_offset = 0;
336 		break;
337 	default:
338 		error = ENOEXEC;
339 		goto cleanup;
340 	}
341 
342 	bss_size = round_page(a_out->a_bss);
343 
344 	/* Check various fields in header for validity/bounds. */
345 	if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
346 		error = ENOEXEC;
347 		goto cleanup;
348 	}
349 
350 	/* text + data can't exceed file size */
351 	if (a_out->a_data + a_out->a_text > attr.va_size) {
352 		error = EFAULT;
353 		goto cleanup;
354 	}
355 
356 	/*
357 	 * text/data/bss must not exceed limits
358 	 * XXX - this is not complete. it should check current usage PLUS
359 	 * the resources needed by this library.
360 	 */
361 	PROC_LOCK(td->td_proc);
362 	if (a_out->a_text > maxtsiz ||
363 	    a_out->a_data + bss_size > lim_cur(td->td_proc, RLIMIT_DATA)) {
364 		PROC_UNLOCK(td->td_proc);
365 		error = ENOMEM;
366 		goto cleanup;
367 	}
368 	PROC_UNLOCK(td->td_proc);
369 
370 	/*
371 	 * Prevent more writers.
372 	 * XXX: Note that if any of the VM operations fail below we don't
373 	 * clear this flag.
374 	 */
375 	vp->v_vflag |= VV_TEXT;
376 
377 	/*
378 	 * Lock no longer needed
379 	 */
380 	locked = 0;
381 	VOP_UNLOCK(vp, 0);
382 	VFS_UNLOCK_GIANT(vfslocked);
383 
384 	/*
385 	 * Check if file_offset page aligned. Currently we cannot handle
386 	 * misalinged file offsets, and so we read in the entire image
387 	 * (what a waste).
388 	 */
389 	if (file_offset & PAGE_MASK) {
390 #ifdef DEBUG
391 		printf("uselib: Non page aligned binary %lu\n", file_offset);
392 #endif
393 		/* Map text+data read/write/execute */
394 
395 		/* a_entry is the load address and is page aligned */
396 		vmaddr = trunc_page(a_out->a_entry);
397 
398 		/* get anon user mapping, read+write+execute */
399 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
400 		    &vmaddr, a_out->a_text + a_out->a_data, FALSE, VM_PROT_ALL,
401 		    VM_PROT_ALL, 0);
402 		if (error)
403 			goto cleanup;
404 
405 		/* map file into kernel_map */
406 		error = vm_mmap(kernel_map, &buffer,
407 		    round_page(a_out->a_text + a_out->a_data + file_offset),
408 		    VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp,
409 		    trunc_page(file_offset));
410 		if (error)
411 			goto cleanup;
412 
413 		/* copy from kernel VM space to user space */
414 		error = copyout(PTRIN(buffer + file_offset),
415 		    (void *)vmaddr, a_out->a_text + a_out->a_data);
416 
417 		/* release temporary kernel space */
418 		vm_map_remove(kernel_map, buffer, buffer +
419 		    round_page(a_out->a_text + a_out->a_data + file_offset));
420 
421 		if (error)
422 			goto cleanup;
423 	} else {
424 #ifdef DEBUG
425 		printf("uselib: Page aligned binary %lu\n", file_offset);
426 #endif
427 		/*
428 		 * for QMAGIC, a_entry is 20 bytes beyond the load address
429 		 * to skip the executable header
430 		 */
431 		vmaddr = trunc_page(a_out->a_entry);
432 
433 		/*
434 		 * Map it all into the process's space as a single
435 		 * copy-on-write "data" segment.
436 		 */
437 		error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr,
438 		    a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
439 		    MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
440 		if (error)
441 			goto cleanup;
442 	}
443 #ifdef DEBUG
444 	printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
445 	    ((long *)vmaddr)[1]);
446 #endif
447 	if (bss_size != 0) {
448 		/* Calculate BSS start address */
449 		vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
450 		    a_out->a_data;
451 
452 		/* allocate some 'anon' space */
453 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
454 		    &vmaddr, bss_size, FALSE, VM_PROT_ALL, VM_PROT_ALL, 0);
455 		if (error)
456 			goto cleanup;
457 	}
458 
459 cleanup:
460 	/* Unlock vnode if needed */
461 	if (locked) {
462 		VOP_UNLOCK(vp, 0);
463 		VFS_UNLOCK_GIANT(vfslocked);
464 	}
465 
466 	/* Release the kernel mapping. */
467 	if (a_out)
468 		vm_map_remove(kernel_map, (vm_offset_t)a_out,
469 		    (vm_offset_t)a_out + PAGE_SIZE);
470 
471 	return error;
472 }
473 
474 #endif	/* __i386__ */
475 
476 int
477 linux_select(struct thread *td, struct linux_select_args *args)
478 {
479 	l_timeval ltv;
480 	struct timeval tv0, tv1, utv, *tvp;
481 	int error;
482 
483 #ifdef DEBUG
484 	if (ldebug(select))
485 		printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
486 		    (void *)args->readfds, (void *)args->writefds,
487 		    (void *)args->exceptfds, (void *)args->timeout);
488 #endif
489 
490 	/*
491 	 * Store current time for computation of the amount of
492 	 * time left.
493 	 */
494 	if (args->timeout) {
495 		if ((error = copyin(args->timeout, &ltv, sizeof(ltv))))
496 			goto select_out;
497 		utv.tv_sec = ltv.tv_sec;
498 		utv.tv_usec = ltv.tv_usec;
499 #ifdef DEBUG
500 		if (ldebug(select))
501 			printf(LMSG("incoming timeout (%jd/%ld)"),
502 			    (intmax_t)utv.tv_sec, utv.tv_usec);
503 #endif
504 
505 		if (itimerfix(&utv)) {
506 			/*
507 			 * The timeval was invalid.  Convert it to something
508 			 * valid that will act as it does under Linux.
509 			 */
510 			utv.tv_sec += utv.tv_usec / 1000000;
511 			utv.tv_usec %= 1000000;
512 			if (utv.tv_usec < 0) {
513 				utv.tv_sec -= 1;
514 				utv.tv_usec += 1000000;
515 			}
516 			if (utv.tv_sec < 0)
517 				timevalclear(&utv);
518 		}
519 		microtime(&tv0);
520 		tvp = &utv;
521 	} else
522 		tvp = NULL;
523 
524 	error = kern_select(td, args->nfds, args->readfds, args->writefds,
525 	    args->exceptfds, tvp);
526 
527 #ifdef DEBUG
528 	if (ldebug(select))
529 		printf(LMSG("real select returns %d"), error);
530 #endif
531 	if (error)
532 		goto select_out;
533 
534 	if (args->timeout) {
535 		if (td->td_retval[0]) {
536 			/*
537 			 * Compute how much time was left of the timeout,
538 			 * by subtracting the current time and the time
539 			 * before we started the call, and subtracting
540 			 * that result from the user-supplied value.
541 			 */
542 			microtime(&tv1);
543 			timevalsub(&tv1, &tv0);
544 			timevalsub(&utv, &tv1);
545 			if (utv.tv_sec < 0)
546 				timevalclear(&utv);
547 		} else
548 			timevalclear(&utv);
549 #ifdef DEBUG
550 		if (ldebug(select))
551 			printf(LMSG("outgoing timeout (%jd/%ld)"),
552 			    (intmax_t)utv.tv_sec, utv.tv_usec);
553 #endif
554 		ltv.tv_sec = utv.tv_sec;
555 		ltv.tv_usec = utv.tv_usec;
556 		if ((error = copyout(&ltv, args->timeout, sizeof(ltv))))
557 			goto select_out;
558 	}
559 
560 select_out:
561 #ifdef DEBUG
562 	if (ldebug(select))
563 		printf(LMSG("select_out -> %d"), error);
564 #endif
565 	return error;
566 }
567 
568 int
569 linux_mremap(struct thread *td, struct linux_mremap_args *args)
570 {
571 	struct munmap_args /* {
572 		void *addr;
573 		size_t len;
574 	} */ bsd_args;
575 	int error = 0;
576 
577 #ifdef DEBUG
578 	if (ldebug(mremap))
579 		printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
580 		    (void *)(uintptr_t)args->addr,
581 		    (unsigned long)args->old_len,
582 		    (unsigned long)args->new_len,
583 		    (unsigned long)args->flags);
584 #endif
585 
586 	if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
587 		td->td_retval[0] = 0;
588 		return (EINVAL);
589 	}
590 
591 	/*
592 	 * Check for the page alignment.
593 	 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
594 	 */
595 	if (args->addr & PAGE_MASK) {
596 		td->td_retval[0] = 0;
597 		return (EINVAL);
598 	}
599 
600 	args->new_len = round_page(args->new_len);
601 	args->old_len = round_page(args->old_len);
602 
603 	if (args->new_len > args->old_len) {
604 		td->td_retval[0] = 0;
605 		return ENOMEM;
606 	}
607 
608 	if (args->new_len < args->old_len) {
609 		bsd_args.addr =
610 		    (caddr_t)((uintptr_t)args->addr + args->new_len);
611 		bsd_args.len = args->old_len - args->new_len;
612 		error = munmap(td, &bsd_args);
613 	}
614 
615 	td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
616 	return error;
617 }
618 
619 #define LINUX_MS_ASYNC       0x0001
620 #define LINUX_MS_INVALIDATE  0x0002
621 #define LINUX_MS_SYNC        0x0004
622 
623 int
624 linux_msync(struct thread *td, struct linux_msync_args *args)
625 {
626 	struct msync_args bsd_args;
627 
628 	bsd_args.addr = (caddr_t)(uintptr_t)args->addr;
629 	bsd_args.len = (uintptr_t)args->len;
630 	bsd_args.flags = args->fl & ~LINUX_MS_SYNC;
631 
632 	return msync(td, &bsd_args);
633 }
634 
635 int
636 linux_time(struct thread *td, struct linux_time_args *args)
637 {
638 	struct timeval tv;
639 	l_time_t tm;
640 	int error;
641 
642 #ifdef DEBUG
643 	if (ldebug(time))
644 		printf(ARGS(time, "*"));
645 #endif
646 
647 	microtime(&tv);
648 	tm = tv.tv_sec;
649 	if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
650 		return error;
651 	td->td_retval[0] = tm;
652 	return 0;
653 }
654 
655 struct l_times_argv {
656 	l_clock_t	tms_utime;
657 	l_clock_t	tms_stime;
658 	l_clock_t	tms_cutime;
659 	l_clock_t	tms_cstime;
660 };
661 
662 
663 /*
664  * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
665  * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
666  * auxiliary vector entry.
667  */
668 #define	CLK_TCK		100
669 
670 #define	CONVOTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
671 #define	CONVNTCK(r)	(r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
672 
673 #define	CONVTCK(r)	(linux_kernver(td) >= LINUX_KERNVER_2004000 ?		\
674 			    CONVNTCK(r) : CONVOTCK(r))
675 
676 int
677 linux_times(struct thread *td, struct linux_times_args *args)
678 {
679 	struct timeval tv, utime, stime, cutime, cstime;
680 	struct l_times_argv tms;
681 	struct proc *p;
682 	int error;
683 
684 #ifdef DEBUG
685 	if (ldebug(times))
686 		printf(ARGS(times, "*"));
687 #endif
688 
689 	if (args->buf != NULL) {
690 		p = td->td_proc;
691 		PROC_LOCK(p);
692 		PROC_SLOCK(p);
693 		calcru(p, &utime, &stime);
694 		PROC_SUNLOCK(p);
695 		calccru(p, &cutime, &cstime);
696 		PROC_UNLOCK(p);
697 
698 		tms.tms_utime = CONVTCK(utime);
699 		tms.tms_stime = CONVTCK(stime);
700 
701 		tms.tms_cutime = CONVTCK(cutime);
702 		tms.tms_cstime = CONVTCK(cstime);
703 
704 		if ((error = copyout(&tms, args->buf, sizeof(tms))))
705 			return error;
706 	}
707 
708 	microuptime(&tv);
709 	td->td_retval[0] = (int)CONVTCK(tv);
710 	return 0;
711 }
712 
713 int
714 linux_newuname(struct thread *td, struct linux_newuname_args *args)
715 {
716 	struct l_new_utsname utsname;
717 	char osname[LINUX_MAX_UTSNAME];
718 	char osrelease[LINUX_MAX_UTSNAME];
719 	struct prison *pr;
720 	char *p;
721 
722 #ifdef DEBUG
723 	if (ldebug(newuname))
724 		printf(ARGS(newuname, "*"));
725 #endif
726 
727 	linux_get_osname(td, osname);
728 	linux_get_osrelease(td, osrelease);
729 
730 	bzero(&utsname, sizeof(utsname));
731 	strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
732 	getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
733 	strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
734 	strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
735 	for (p = utsname.version; *p != '\0'; ++p)
736 		if (*p == '\n') {
737 			*p = '\0';
738 			break;
739 		}
740 	strlcpy(utsname.machine, linux_platform, LINUX_MAX_UTSNAME);
741 
742 	pr = td->td_ucred->cr_prison;
743 	mtx_lock(&pr->pr_mtx);
744 	strlcpy(utsname.domainname, pr->pr_domain, LINUX_MAX_UTSNAME);
745 	mtx_unlock(&pr->pr_mtx);
746 
747 	return (copyout(&utsname, args->buf, sizeof(utsname)));
748 }
749 
750 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
751 struct l_utimbuf {
752 	l_time_t l_actime;
753 	l_time_t l_modtime;
754 };
755 
756 int
757 linux_utime(struct thread *td, struct linux_utime_args *args)
758 {
759 	struct timeval tv[2], *tvp;
760 	struct l_utimbuf lut;
761 	char *fname;
762 	int error;
763 
764 	LCONVPATHEXIST(td, args->fname, &fname);
765 
766 #ifdef DEBUG
767 	if (ldebug(utime))
768 		printf(ARGS(utime, "%s, *"), fname);
769 #endif
770 
771 	if (args->times) {
772 		if ((error = copyin(args->times, &lut, sizeof lut))) {
773 			LFREEPATH(fname);
774 			return error;
775 		}
776 		tv[0].tv_sec = lut.l_actime;
777 		tv[0].tv_usec = 0;
778 		tv[1].tv_sec = lut.l_modtime;
779 		tv[1].tv_usec = 0;
780 		tvp = tv;
781 	} else
782 		tvp = NULL;
783 
784 	error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
785 	LFREEPATH(fname);
786 	return (error);
787 }
788 
789 int
790 linux_utimes(struct thread *td, struct linux_utimes_args *args)
791 {
792 	l_timeval ltv[2];
793 	struct timeval tv[2], *tvp = NULL;
794 	char *fname;
795 	int error;
796 
797 	LCONVPATHEXIST(td, args->fname, &fname);
798 
799 #ifdef DEBUG
800 	if (ldebug(utimes))
801 		printf(ARGS(utimes, "%s, *"), fname);
802 #endif
803 
804 	if (args->tptr != NULL) {
805 		if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
806 			LFREEPATH(fname);
807 			return (error);
808 		}
809 		tv[0].tv_sec = ltv[0].tv_sec;
810 		tv[0].tv_usec = ltv[0].tv_usec;
811 		tv[1].tv_sec = ltv[1].tv_sec;
812 		tv[1].tv_usec = ltv[1].tv_usec;
813 		tvp = tv;
814 	}
815 
816 	error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
817 	LFREEPATH(fname);
818 	return (error);
819 }
820 
821 int
822 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
823 {
824 	l_timeval ltv[2];
825 	struct timeval tv[2], *tvp = NULL;
826 	char *fname;
827 	int error, dfd;
828 
829 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
830 	LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
831 
832 #ifdef DEBUG
833 	if (ldebug(futimesat))
834 		printf(ARGS(futimesat, "%s, *"), fname);
835 #endif
836 
837 	if (args->utimes != NULL) {
838 		if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
839 			LFREEPATH(fname);
840 			return (error);
841 		}
842 		tv[0].tv_sec = ltv[0].tv_sec;
843 		tv[0].tv_usec = ltv[0].tv_usec;
844 		tv[1].tv_sec = ltv[1].tv_sec;
845 		tv[1].tv_usec = ltv[1].tv_usec;
846 		tvp = tv;
847 	}
848 
849 	error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
850 	LFREEPATH(fname);
851 	return (error);
852 }
853 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
854 
855 #define __WCLONE 0x80000000
856 
857 int
858 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
859 {
860 	int error, options, tmpstat;
861 
862 #ifdef DEBUG
863 	if (ldebug(waitpid))
864 		printf(ARGS(waitpid, "%d, %p, %d"),
865 		    args->pid, (void *)args->status, args->options);
866 #endif
867 	/*
868 	 * this is necessary because the test in kern_wait doesn't work
869 	 * because we mess with the options here
870 	 */
871 	if (args->options & ~(WUNTRACED | WNOHANG | WCONTINUED | __WCLONE))
872 		return (EINVAL);
873 
874 	options = (args->options & (WNOHANG | WUNTRACED));
875 	/* WLINUXCLONE should be equal to __WCLONE, but we make sure */
876 	if (args->options & __WCLONE)
877 		options |= WLINUXCLONE;
878 
879 	error = kern_wait(td, args->pid, &tmpstat, options, NULL);
880 	if (error)
881 		return error;
882 
883 	if (args->status) {
884 		tmpstat &= 0xffff;
885 		if (WIFSIGNALED(tmpstat))
886 			tmpstat = (tmpstat & 0xffffff80) |
887 			    BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat));
888 		else if (WIFSTOPPED(tmpstat))
889 			tmpstat = (tmpstat & 0xffff00ff) |
890 			    (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8);
891 		return copyout(&tmpstat, args->status, sizeof(int));
892 	}
893 
894 	return 0;
895 }
896 
897 int
898 linux_wait4(struct thread *td, struct linux_wait4_args *args)
899 {
900 	int error, options, tmpstat;
901 	struct rusage ru, *rup;
902 	struct proc *p;
903 
904 #ifdef DEBUG
905 	if (ldebug(wait4))
906 		printf(ARGS(wait4, "%d, %p, %d, %p"),
907 		    args->pid, (void *)args->status, args->options,
908 		    (void *)args->rusage);
909 #endif
910 
911 	options = (args->options & (WNOHANG | WUNTRACED));
912 	/* WLINUXCLONE should be equal to __WCLONE, but we make sure */
913 	if (args->options & __WCLONE)
914 		options |= WLINUXCLONE;
915 
916 	if (args->rusage != NULL)
917 		rup = &ru;
918 	else
919 		rup = NULL;
920 	error = kern_wait(td, args->pid, &tmpstat, options, rup);
921 	if (error)
922 		return error;
923 
924 	p = td->td_proc;
925 	PROC_LOCK(p);
926 	sigqueue_delete(&p->p_sigqueue, SIGCHLD);
927 	PROC_UNLOCK(p);
928 
929 	if (args->status) {
930 		tmpstat &= 0xffff;
931 		if (WIFSIGNALED(tmpstat))
932 			tmpstat = (tmpstat & 0xffffff80) |
933 			    BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat));
934 		else if (WIFSTOPPED(tmpstat))
935 			tmpstat = (tmpstat & 0xffff00ff) |
936 			    (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8);
937 		error = copyout(&tmpstat, args->status, sizeof(int));
938 	}
939 	if (args->rusage != NULL && error == 0)
940 		error = copyout(&ru, args->rusage, sizeof(ru));
941 
942 	return (error);
943 }
944 
945 int
946 linux_mknod(struct thread *td, struct linux_mknod_args *args)
947 {
948 	char *path;
949 	int error;
950 
951 	LCONVPATHCREAT(td, args->path, &path);
952 
953 #ifdef DEBUG
954 	if (ldebug(mknod))
955 		printf(ARGS(mknod, "%s, %d, %d"), path, args->mode, args->dev);
956 #endif
957 
958 	switch (args->mode & S_IFMT) {
959 	case S_IFIFO:
960 	case S_IFSOCK:
961 		error = kern_mkfifo(td, path, UIO_SYSSPACE, args->mode);
962 		break;
963 
964 	case S_IFCHR:
965 	case S_IFBLK:
966 		error = kern_mknod(td, path, UIO_SYSSPACE, args->mode,
967 		    args->dev);
968 		break;
969 
970 	case S_IFDIR:
971 		error = EPERM;
972 		break;
973 
974 	case 0:
975 		args->mode |= S_IFREG;
976 		/* FALLTHROUGH */
977 	case S_IFREG:
978 		error = kern_open(td, path, UIO_SYSSPACE,
979 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
980 		if (error == 0)
981 			kern_close(td, td->td_retval[0]);
982 		break;
983 
984 	default:
985 		error = EINVAL;
986 		break;
987 	}
988 	LFREEPATH(path);
989 	return (error);
990 }
991 
992 int
993 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
994 {
995 	char *path;
996 	int error, dfd;
997 
998 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
999 	LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
1000 
1001 #ifdef DEBUG
1002 	if (ldebug(mknodat))
1003 		printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev);
1004 #endif
1005 
1006 	switch (args->mode & S_IFMT) {
1007 	case S_IFIFO:
1008 	case S_IFSOCK:
1009 		error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
1010 		break;
1011 
1012 	case S_IFCHR:
1013 	case S_IFBLK:
1014 		error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
1015 		    args->dev);
1016 		break;
1017 
1018 	case S_IFDIR:
1019 		error = EPERM;
1020 		break;
1021 
1022 	case 0:
1023 		args->mode |= S_IFREG;
1024 		/* FALLTHROUGH */
1025 	case S_IFREG:
1026 		error = kern_openat(td, dfd, path, UIO_SYSSPACE,
1027 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1028 		if (error == 0)
1029 			kern_close(td, td->td_retval[0]);
1030 		break;
1031 
1032 	default:
1033 		error = EINVAL;
1034 		break;
1035 	}
1036 	LFREEPATH(path);
1037 	return (error);
1038 }
1039 
1040 /*
1041  * UGH! This is just about the dumbest idea I've ever heard!!
1042  */
1043 int
1044 linux_personality(struct thread *td, struct linux_personality_args *args)
1045 {
1046 #ifdef DEBUG
1047 	if (ldebug(personality))
1048 		printf(ARGS(personality, "%lu"), (unsigned long)args->per);
1049 #endif
1050 	if (args->per != 0)
1051 		return EINVAL;
1052 
1053 	/* Yes Jim, it's still a Linux... */
1054 	td->td_retval[0] = 0;
1055 	return 0;
1056 }
1057 
1058 struct l_itimerval {
1059 	l_timeval it_interval;
1060 	l_timeval it_value;
1061 };
1062 
1063 #define	B2L_ITIMERVAL(bip, lip) 					\
1064 	(bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec;		\
1065 	(bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec;	\
1066 	(bip)->it_value.tv_sec = (lip)->it_value.tv_sec;		\
1067 	(bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1068 
1069 int
1070 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1071 {
1072 	int error;
1073 	struct l_itimerval ls;
1074 	struct itimerval aitv, oitv;
1075 
1076 #ifdef DEBUG
1077 	if (ldebug(setitimer))
1078 		printf(ARGS(setitimer, "%p, %p"),
1079 		    (void *)uap->itv, (void *)uap->oitv);
1080 #endif
1081 
1082 	if (uap->itv == NULL) {
1083 		uap->itv = uap->oitv;
1084 		return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1085 	}
1086 
1087 	error = copyin(uap->itv, &ls, sizeof(ls));
1088 	if (error != 0)
1089 		return (error);
1090 	B2L_ITIMERVAL(&aitv, &ls);
1091 #ifdef DEBUG
1092 	if (ldebug(setitimer)) {
1093 		printf("setitimer: value: sec: %jd, usec: %ld\n",
1094 		    (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1095 		printf("setitimer: interval: sec: %jd, usec: %ld\n",
1096 		    (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1097 	}
1098 #endif
1099 	error = kern_setitimer(td, uap->which, &aitv, &oitv);
1100 	if (error != 0 || uap->oitv == NULL)
1101 		return (error);
1102 	B2L_ITIMERVAL(&ls, &oitv);
1103 
1104 	return (copyout(&ls, uap->oitv, sizeof(ls)));
1105 }
1106 
1107 int
1108 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1109 {
1110 	int error;
1111 	struct l_itimerval ls;
1112 	struct itimerval aitv;
1113 
1114 #ifdef DEBUG
1115 	if (ldebug(getitimer))
1116 		printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1117 #endif
1118 	error = kern_getitimer(td, uap->which, &aitv);
1119 	if (error != 0)
1120 		return (error);
1121 	B2L_ITIMERVAL(&ls, &aitv);
1122 	return (copyout(&ls, uap->itv, sizeof(ls)));
1123 }
1124 
1125 int
1126 linux_nice(struct thread *td, struct linux_nice_args *args)
1127 {
1128 	struct setpriority_args bsd_args;
1129 
1130 	bsd_args.which = PRIO_PROCESS;
1131 	bsd_args.who = 0;		/* current process */
1132 	bsd_args.prio = args->inc;
1133 	return setpriority(td, &bsd_args);
1134 }
1135 
1136 int
1137 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1138 {
1139 	struct ucred *newcred, *oldcred;
1140 	l_gid_t linux_gidset[NGROUPS];
1141 	gid_t *bsd_gidset;
1142 	int ngrp, error;
1143 	struct proc *p;
1144 
1145 	ngrp = args->gidsetsize;
1146 	if (ngrp < 0 || ngrp >= NGROUPS)
1147 		return (EINVAL);
1148 	error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1149 	if (error)
1150 		return (error);
1151 	newcred = crget();
1152 	p = td->td_proc;
1153 	PROC_LOCK(p);
1154 	oldcred = p->p_ucred;
1155 
1156 	/*
1157 	 * cr_groups[0] holds egid. Setting the whole set from
1158 	 * the supplied set will cause egid to be changed too.
1159 	 * Keep cr_groups[0] unchanged to prevent that.
1160 	 */
1161 
1162 	if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) {
1163 		PROC_UNLOCK(p);
1164 		crfree(newcred);
1165 		return (error);
1166 	}
1167 
1168 	crcopy(newcred, oldcred);
1169 	if (ngrp > 0) {
1170 		newcred->cr_ngroups = ngrp + 1;
1171 
1172 		bsd_gidset = newcred->cr_groups;
1173 		ngrp--;
1174 		while (ngrp >= 0) {
1175 			bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1176 			ngrp--;
1177 		}
1178 	} else
1179 		newcred->cr_ngroups = 1;
1180 
1181 	setsugid(p);
1182 	p->p_ucred = newcred;
1183 	PROC_UNLOCK(p);
1184 	crfree(oldcred);
1185 	return (0);
1186 }
1187 
1188 int
1189 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1190 {
1191 	struct ucred *cred;
1192 	l_gid_t linux_gidset[NGROUPS];
1193 	gid_t *bsd_gidset;
1194 	int bsd_gidsetsz, ngrp, error;
1195 
1196 	cred = td->td_ucred;
1197 	bsd_gidset = cred->cr_groups;
1198 	bsd_gidsetsz = cred->cr_ngroups - 1;
1199 
1200 	/*
1201 	 * cr_groups[0] holds egid. Returning the whole set
1202 	 * here will cause a duplicate. Exclude cr_groups[0]
1203 	 * to prevent that.
1204 	 */
1205 
1206 	if ((ngrp = args->gidsetsize) == 0) {
1207 		td->td_retval[0] = bsd_gidsetsz;
1208 		return (0);
1209 	}
1210 
1211 	if (ngrp < bsd_gidsetsz)
1212 		return (EINVAL);
1213 
1214 	ngrp = 0;
1215 	while (ngrp < bsd_gidsetsz) {
1216 		linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1217 		ngrp++;
1218 	}
1219 
1220 	if ((error = copyout(linux_gidset, args->grouplist,
1221 	    ngrp * sizeof(l_gid_t))))
1222 		return (error);
1223 
1224 	td->td_retval[0] = ngrp;
1225 	return (0);
1226 }
1227 
1228 int
1229 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1230 {
1231 	struct rlimit bsd_rlim;
1232 	struct l_rlimit rlim;
1233 	u_int which;
1234 	int error;
1235 
1236 #ifdef DEBUG
1237 	if (ldebug(setrlimit))
1238 		printf(ARGS(setrlimit, "%d, %p"),
1239 		    args->resource, (void *)args->rlim);
1240 #endif
1241 
1242 	if (args->resource >= LINUX_RLIM_NLIMITS)
1243 		return (EINVAL);
1244 
1245 	which = linux_to_bsd_resource[args->resource];
1246 	if (which == -1)
1247 		return (EINVAL);
1248 
1249 	error = copyin(args->rlim, &rlim, sizeof(rlim));
1250 	if (error)
1251 		return (error);
1252 
1253 	bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1254 	bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1255 	return (kern_setrlimit(td, which, &bsd_rlim));
1256 }
1257 
1258 int
1259 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1260 {
1261 	struct l_rlimit rlim;
1262 	struct proc *p = td->td_proc;
1263 	struct rlimit bsd_rlim;
1264 	u_int which;
1265 
1266 #ifdef DEBUG
1267 	if (ldebug(old_getrlimit))
1268 		printf(ARGS(old_getrlimit, "%d, %p"),
1269 		    args->resource, (void *)args->rlim);
1270 #endif
1271 
1272 	if (args->resource >= LINUX_RLIM_NLIMITS)
1273 		return (EINVAL);
1274 
1275 	which = linux_to_bsd_resource[args->resource];
1276 	if (which == -1)
1277 		return (EINVAL);
1278 
1279 	PROC_LOCK(p);
1280 	lim_rlimit(p, which, &bsd_rlim);
1281 	PROC_UNLOCK(p);
1282 
1283 #ifdef COMPAT_LINUX32
1284 	rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1285 	if (rlim.rlim_cur == UINT_MAX)
1286 		rlim.rlim_cur = INT_MAX;
1287 	rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1288 	if (rlim.rlim_max == UINT_MAX)
1289 		rlim.rlim_max = INT_MAX;
1290 #else
1291 	rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1292 	if (rlim.rlim_cur == ULONG_MAX)
1293 		rlim.rlim_cur = LONG_MAX;
1294 	rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1295 	if (rlim.rlim_max == ULONG_MAX)
1296 		rlim.rlim_max = LONG_MAX;
1297 #endif
1298 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1299 }
1300 
1301 int
1302 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1303 {
1304 	struct l_rlimit rlim;
1305 	struct proc *p = td->td_proc;
1306 	struct rlimit bsd_rlim;
1307 	u_int which;
1308 
1309 #ifdef DEBUG
1310 	if (ldebug(getrlimit))
1311 		printf(ARGS(getrlimit, "%d, %p"),
1312 		    args->resource, (void *)args->rlim);
1313 #endif
1314 
1315 	if (args->resource >= LINUX_RLIM_NLIMITS)
1316 		return (EINVAL);
1317 
1318 	which = linux_to_bsd_resource[args->resource];
1319 	if (which == -1)
1320 		return (EINVAL);
1321 
1322 	PROC_LOCK(p);
1323 	lim_rlimit(p, which, &bsd_rlim);
1324 	PROC_UNLOCK(p);
1325 
1326 	rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1327 	rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1328 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1329 }
1330 
1331 int
1332 linux_sched_setscheduler(struct thread *td,
1333     struct linux_sched_setscheduler_args *args)
1334 {
1335 	struct sched_setscheduler_args bsd;
1336 
1337 #ifdef DEBUG
1338 	if (ldebug(sched_setscheduler))
1339 		printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1340 		    args->pid, args->policy, (const void *)args->param);
1341 #endif
1342 
1343 	switch (args->policy) {
1344 	case LINUX_SCHED_OTHER:
1345 		bsd.policy = SCHED_OTHER;
1346 		break;
1347 	case LINUX_SCHED_FIFO:
1348 		bsd.policy = SCHED_FIFO;
1349 		break;
1350 	case LINUX_SCHED_RR:
1351 		bsd.policy = SCHED_RR;
1352 		break;
1353 	default:
1354 		return EINVAL;
1355 	}
1356 
1357 	bsd.pid = args->pid;
1358 	bsd.param = (struct sched_param *)args->param;
1359 	return sched_setscheduler(td, &bsd);
1360 }
1361 
1362 int
1363 linux_sched_getscheduler(struct thread *td,
1364     struct linux_sched_getscheduler_args *args)
1365 {
1366 	struct sched_getscheduler_args bsd;
1367 	int error;
1368 
1369 #ifdef DEBUG
1370 	if (ldebug(sched_getscheduler))
1371 		printf(ARGS(sched_getscheduler, "%d"), args->pid);
1372 #endif
1373 
1374 	bsd.pid = args->pid;
1375 	error = sched_getscheduler(td, &bsd);
1376 
1377 	switch (td->td_retval[0]) {
1378 	case SCHED_OTHER:
1379 		td->td_retval[0] = LINUX_SCHED_OTHER;
1380 		break;
1381 	case SCHED_FIFO:
1382 		td->td_retval[0] = LINUX_SCHED_FIFO;
1383 		break;
1384 	case SCHED_RR:
1385 		td->td_retval[0] = LINUX_SCHED_RR;
1386 		break;
1387 	}
1388 
1389 	return error;
1390 }
1391 
1392 int
1393 linux_sched_get_priority_max(struct thread *td,
1394     struct linux_sched_get_priority_max_args *args)
1395 {
1396 	struct sched_get_priority_max_args bsd;
1397 
1398 #ifdef DEBUG
1399 	if (ldebug(sched_get_priority_max))
1400 		printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1401 #endif
1402 
1403 	switch (args->policy) {
1404 	case LINUX_SCHED_OTHER:
1405 		bsd.policy = SCHED_OTHER;
1406 		break;
1407 	case LINUX_SCHED_FIFO:
1408 		bsd.policy = SCHED_FIFO;
1409 		break;
1410 	case LINUX_SCHED_RR:
1411 		bsd.policy = SCHED_RR;
1412 		break;
1413 	default:
1414 		return EINVAL;
1415 	}
1416 	return sched_get_priority_max(td, &bsd);
1417 }
1418 
1419 int
1420 linux_sched_get_priority_min(struct thread *td,
1421     struct linux_sched_get_priority_min_args *args)
1422 {
1423 	struct sched_get_priority_min_args bsd;
1424 
1425 #ifdef DEBUG
1426 	if (ldebug(sched_get_priority_min))
1427 		printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1428 #endif
1429 
1430 	switch (args->policy) {
1431 	case LINUX_SCHED_OTHER:
1432 		bsd.policy = SCHED_OTHER;
1433 		break;
1434 	case LINUX_SCHED_FIFO:
1435 		bsd.policy = SCHED_FIFO;
1436 		break;
1437 	case LINUX_SCHED_RR:
1438 		bsd.policy = SCHED_RR;
1439 		break;
1440 	default:
1441 		return EINVAL;
1442 	}
1443 	return sched_get_priority_min(td, &bsd);
1444 }
1445 
1446 #define REBOOT_CAD_ON	0x89abcdef
1447 #define REBOOT_CAD_OFF	0
1448 #define REBOOT_HALT	0xcdef0123
1449 #define REBOOT_RESTART	0x01234567
1450 #define REBOOT_RESTART2	0xA1B2C3D4
1451 #define REBOOT_POWEROFF	0x4321FEDC
1452 #define REBOOT_MAGIC1	0xfee1dead
1453 #define REBOOT_MAGIC2	0x28121969
1454 #define REBOOT_MAGIC2A	0x05121996
1455 #define REBOOT_MAGIC2B	0x16041998
1456 
1457 int
1458 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1459 {
1460 	struct reboot_args bsd_args;
1461 
1462 #ifdef DEBUG
1463 	if (ldebug(reboot))
1464 		printf(ARGS(reboot, "0x%x"), args->cmd);
1465 #endif
1466 
1467 	if (args->magic1 != REBOOT_MAGIC1)
1468 		return EINVAL;
1469 
1470 	switch (args->magic2) {
1471 	case REBOOT_MAGIC2:
1472 	case REBOOT_MAGIC2A:
1473 	case REBOOT_MAGIC2B:
1474 		break;
1475 	default:
1476 		return EINVAL;
1477 	}
1478 
1479 	switch (args->cmd) {
1480 	case REBOOT_CAD_ON:
1481 	case REBOOT_CAD_OFF:
1482 		return (priv_check(td, PRIV_REBOOT));
1483 	case REBOOT_HALT:
1484 		bsd_args.opt = RB_HALT;
1485 		break;
1486 	case REBOOT_RESTART:
1487 	case REBOOT_RESTART2:
1488 		bsd_args.opt = 0;
1489 		break;
1490 	case REBOOT_POWEROFF:
1491 		bsd_args.opt = RB_POWEROFF;
1492 		break;
1493 	default:
1494 		return EINVAL;
1495 	}
1496 	return reboot(td, &bsd_args);
1497 }
1498 
1499 
1500 /*
1501  * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1502  * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
1503  * are assumed to be preserved. The following lightweight syscalls fixes
1504  * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
1505  *
1506  * linux_getpid() - MP SAFE
1507  * linux_getgid() - MP SAFE
1508  * linux_getuid() - MP SAFE
1509  */
1510 
1511 int
1512 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1513 {
1514 	struct linux_emuldata *em;
1515 
1516 #ifdef DEBUG
1517 	if (ldebug(getpid))
1518 		printf(ARGS(getpid, ""));
1519 #endif
1520 
1521 	if (linux_use26(td)) {
1522 		em = em_find(td->td_proc, EMUL_DONTLOCK);
1523 		KASSERT(em != NULL, ("getpid: emuldata not found.\n"));
1524 		td->td_retval[0] = em->shared->group_pid;
1525 	} else {
1526 		td->td_retval[0] = td->td_proc->p_pid;
1527 	}
1528 
1529 	return (0);
1530 }
1531 
1532 int
1533 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1534 {
1535 
1536 #ifdef DEBUG
1537 	if (ldebug(gettid))
1538 		printf(ARGS(gettid, ""));
1539 #endif
1540 
1541 	td->td_retval[0] = td->td_proc->p_pid;
1542 	return (0);
1543 }
1544 
1545 
1546 int
1547 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1548 {
1549 	struct linux_emuldata *em;
1550 	struct proc *p, *pp;
1551 
1552 #ifdef DEBUG
1553 	if (ldebug(getppid))
1554 		printf(ARGS(getppid, ""));
1555 #endif
1556 
1557 	if (!linux_use26(td)) {
1558 		PROC_LOCK(td->td_proc);
1559 		td->td_retval[0] = td->td_proc->p_pptr->p_pid;
1560 		PROC_UNLOCK(td->td_proc);
1561 		return (0);
1562 	}
1563 
1564 	em = em_find(td->td_proc, EMUL_DONTLOCK);
1565 
1566 	KASSERT(em != NULL, ("getppid: process emuldata not found.\n"));
1567 
1568 	/* find the group leader */
1569 	p = pfind(em->shared->group_pid);
1570 
1571 	if (p == NULL) {
1572 #ifdef DEBUG
1573 	   	printf(LMSG("parent process not found.\n"));
1574 #endif
1575 		return (0);
1576 	}
1577 
1578 	pp = p->p_pptr;		/* switch to parent */
1579 	PROC_LOCK(pp);
1580 	PROC_UNLOCK(p);
1581 
1582 	/* if its also linux process */
1583 	if (pp->p_sysent == &elf_linux_sysvec) {
1584 		em = em_find(pp, EMUL_DONTLOCK);
1585 		KASSERT(em != NULL, ("getppid: parent emuldata not found.\n"));
1586 
1587 		td->td_retval[0] = em->shared->group_pid;
1588 	} else
1589 		td->td_retval[0] = pp->p_pid;
1590 
1591 	PROC_UNLOCK(pp);
1592 
1593 	return (0);
1594 }
1595 
1596 int
1597 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1598 {
1599 
1600 #ifdef DEBUG
1601 	if (ldebug(getgid))
1602 		printf(ARGS(getgid, ""));
1603 #endif
1604 
1605 	td->td_retval[0] = td->td_ucred->cr_rgid;
1606 	return (0);
1607 }
1608 
1609 int
1610 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1611 {
1612 
1613 #ifdef DEBUG
1614 	if (ldebug(getuid))
1615 		printf(ARGS(getuid, ""));
1616 #endif
1617 
1618 	td->td_retval[0] = td->td_ucred->cr_ruid;
1619 	return (0);
1620 }
1621 
1622 
1623 int
1624 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1625 {
1626 	struct getsid_args bsd;
1627 
1628 #ifdef DEBUG
1629 	if (ldebug(getsid))
1630 		printf(ARGS(getsid, "%i"), args->pid);
1631 #endif
1632 
1633 	bsd.pid = args->pid;
1634 	return getsid(td, &bsd);
1635 }
1636 
1637 int
1638 linux_nosys(struct thread *td, struct nosys_args *ignore)
1639 {
1640 
1641 	return (ENOSYS);
1642 }
1643 
1644 int
1645 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1646 {
1647 	struct getpriority_args bsd_args;
1648 	int error;
1649 
1650 #ifdef DEBUG
1651 	if (ldebug(getpriority))
1652 		printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
1653 #endif
1654 
1655 	bsd_args.which = args->which;
1656 	bsd_args.who = args->who;
1657 	error = getpriority(td, &bsd_args);
1658 	td->td_retval[0] = 20 - td->td_retval[0];
1659 	return error;
1660 }
1661 
1662 int
1663 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1664 {
1665 	int name[2];
1666 
1667 #ifdef DEBUG
1668 	if (ldebug(sethostname))
1669 		printf(ARGS(sethostname, "*, %i"), args->len);
1670 #endif
1671 
1672 	name[0] = CTL_KERN;
1673 	name[1] = KERN_HOSTNAME;
1674 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1675 	    args->len, 0, 0));
1676 }
1677 
1678 int
1679 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1680 {
1681 	int name[2];
1682 
1683 #ifdef DEBUG
1684 	if (ldebug(setdomainname))
1685 		printf(ARGS(setdomainname, "*, %i"), args->len);
1686 #endif
1687 
1688 	name[0] = CTL_KERN;
1689 	name[1] = KERN_NISDOMAINNAME;
1690 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1691 	    args->len, 0, 0));
1692 }
1693 
1694 int
1695 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1696 {
1697 	struct linux_emuldata *em, *td_em, *tmp_em;
1698 	struct proc *sp;
1699 
1700 #ifdef DEBUG
1701 	if (ldebug(exit_group))
1702 		printf(ARGS(exit_group, "%i"), args->error_code);
1703 #endif
1704 
1705 	if (linux_use26(td)) {
1706 		td_em = em_find(td->td_proc, EMUL_DONTLOCK);
1707 
1708 		KASSERT(td_em != NULL, ("exit_group: emuldata not found.\n"));
1709 
1710 		EMUL_SHARED_RLOCK(&emul_shared_lock);
1711 		LIST_FOREACH_SAFE(em, &td_em->shared->threads, threads, tmp_em) {
1712 			if (em->pid == td_em->pid)
1713 				continue;
1714 
1715 			sp = pfind(em->pid);
1716 			psignal(sp, SIGKILL);
1717 			PROC_UNLOCK(sp);
1718 #ifdef DEBUG
1719 			printf(LMSG("linux_sys_exit_group: kill PID %d\n"), em->pid);
1720 #endif
1721 		}
1722 
1723 		EMUL_SHARED_RUNLOCK(&emul_shared_lock);
1724 	}
1725 	/*
1726 	 * XXX: we should send a signal to the parent if
1727 	 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1728 	 * as it doesnt occur often.
1729 	 */
1730 	exit1(td, W_EXITCODE(args->error_code, 0));
1731 
1732 	return (0);
1733 }
1734 
1735 int
1736 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1737 {
1738 	int error = 0, max_size;
1739 	struct proc *p = td->td_proc;
1740 	char comm[LINUX_MAX_COMM_LEN];
1741 	struct linux_emuldata *em;
1742 	int pdeath_signal;
1743 
1744 #ifdef DEBUG
1745 	if (ldebug(prctl))
1746 		printf(ARGS(prctl, "%d, %d, %d, %d, %d"), args->option,
1747 		    args->arg2, args->arg3, args->arg4, args->arg5);
1748 #endif
1749 
1750 	switch (args->option) {
1751 	case LINUX_PR_SET_PDEATHSIG:
1752 		if (!LINUX_SIG_VALID(args->arg2))
1753 			return (EINVAL);
1754 		em = em_find(p, EMUL_DOLOCK);
1755 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1756 		em->pdeath_signal = args->arg2;
1757 		EMUL_UNLOCK(&emul_lock);
1758 		break;
1759 	case LINUX_PR_GET_PDEATHSIG:
1760 		em = em_find(p, EMUL_DOLOCK);
1761 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1762 		pdeath_signal = em->pdeath_signal;
1763 		EMUL_UNLOCK(&emul_lock);
1764 		error = copyout(&pdeath_signal,
1765 		    (void *)(register_t)args->arg2,
1766 		    sizeof(pdeath_signal));
1767 		break;
1768 	case LINUX_PR_SET_NAME:
1769 		/*
1770 		 * To be on the safe side we need to make sure to not
1771 		 * overflow the size a linux program expects. We already
1772 		 * do this here in the copyin, so that we don't need to
1773 		 * check on copyout.
1774 		 */
1775 		max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1776 		error = copyinstr((void *)(register_t)args->arg2, comm,
1777 		    max_size, NULL);
1778 
1779 		/* Linux silently truncates the name if it is too long. */
1780 		if (error == ENAMETOOLONG) {
1781 			/*
1782 			 * XXX: copyinstr() isn't documented to populate the
1783 			 * array completely, so do a copyin() to be on the
1784 			 * safe side. This should be changed in case
1785 			 * copyinstr() is changed to guarantee this.
1786 			 */
1787 			error = copyin((void *)(register_t)args->arg2, comm,
1788 			    max_size - 1);
1789 			comm[max_size - 1] = '\0';
1790 		}
1791 		if (error)
1792 			return (error);
1793 
1794 		PROC_LOCK(p);
1795 		strlcpy(p->p_comm, comm, sizeof(p->p_comm));
1796 		PROC_UNLOCK(p);
1797 		break;
1798 	case LINUX_PR_GET_NAME:
1799 		PROC_LOCK(p);
1800 		strlcpy(comm, p->p_comm, sizeof(comm));
1801 		PROC_UNLOCK(p);
1802 		error = copyout(comm, (void *)(register_t)args->arg2,
1803 		    strlen(comm) + 1);
1804 		break;
1805 	default:
1806 		error = EINVAL;
1807 		break;
1808 	}
1809 
1810 	return (error);
1811 }
1812 
1813 /*
1814  * Get affinity of a process.
1815  */
1816 int
1817 linux_sched_getaffinity(struct thread *td,
1818     struct linux_sched_getaffinity_args *args)
1819 {
1820 	int error;
1821 	struct cpuset_getaffinity_args cga;
1822 
1823 #ifdef DEBUG
1824 	if (ldebug(sched_getaffinity))
1825 		printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid,
1826 		    args->len);
1827 #endif
1828 	if (args->len < sizeof(cpuset_t))
1829 		return (EINVAL);
1830 
1831 	cga.level = CPU_LEVEL_WHICH;
1832 	cga.which = CPU_WHICH_PID;
1833 	cga.id = args->pid;
1834 	cga.cpusetsize = sizeof(cpuset_t);
1835 	cga.mask = (cpuset_t *) args->user_mask_ptr;
1836 
1837 	if ((error = cpuset_getaffinity(td, &cga)) == 0)
1838 		td->td_retval[0] = sizeof(cpuset_t);
1839 
1840 	return (error);
1841 }
1842 
1843 /*
1844  *  Set affinity of a process.
1845  */
1846 int
1847 linux_sched_setaffinity(struct thread *td,
1848     struct linux_sched_setaffinity_args *args)
1849 {
1850 	struct cpuset_setaffinity_args csa;
1851 
1852 #ifdef DEBUG
1853 	if (ldebug(sched_setaffinity))
1854 		printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid,
1855 		    args->len);
1856 #endif
1857 	if (args->len < sizeof(cpuset_t))
1858 		return (EINVAL);
1859 
1860 	csa.level = CPU_LEVEL_WHICH;
1861 	csa.which = CPU_WHICH_PID;
1862 	csa.id = args->pid;
1863 	csa.cpusetsize = sizeof(cpuset_t);
1864 	csa.mask = (cpuset_t *) args->user_mask_ptr;
1865 
1866 	return (cpuset_setaffinity(td, &csa));
1867 }
1868