xref: /freebsd/sys/compat/linux/linux_misc.c (revision a063878a50beb4e1af5cca0c91cfe62966e129dc)
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 #include "opt_mac.h"
35 
36 #include <sys/param.h>
37 #include <sys/blist.h>
38 #include <sys/fcntl.h>
39 #if defined(__i386__)
40 #include <sys/imgact_aout.h>
41 #endif
42 #include <sys/jail.h>
43 #include <sys/kernel.h>
44 #include <sys/limits.h>
45 #include <sys/lock.h>
46 #include <sys/malloc.h>
47 #include <sys/mman.h>
48 #include <sys/mount.h>
49 #include <sys/mutex.h>
50 #include <sys/namei.h>
51 #include <sys/priv.h>
52 #include <sys/proc.h>
53 #include <sys/reboot.h>
54 #include <sys/resourcevar.h>
55 #include <sys/sched.h>
56 #include <sys/signalvar.h>
57 #include <sys/stat.h>
58 #include <sys/syscallsubr.h>
59 #include <sys/sysctl.h>
60 #include <sys/sysproto.h>
61 #include <sys/systm.h>
62 #include <sys/time.h>
63 #include <sys/vmmeter.h>
64 #include <sys/vnode.h>
65 #include <sys/wait.h>
66 #include <sys/cpuset.h>
67 #include <sys/vimage.h>
68 
69 #include <security/mac/mac_framework.h>
70 
71 #include <vm/vm.h>
72 #include <vm/pmap.h>
73 #include <vm/vm_kern.h>
74 #include <vm/vm_map.h>
75 #include <vm/vm_extern.h>
76 #include <vm/vm_object.h>
77 #include <vm/swap_pager.h>
78 
79 #ifdef COMPAT_LINUX32
80 #include <machine/../linux32/linux.h>
81 #include <machine/../linux32/linux32_proto.h>
82 #else
83 #include <machine/../linux/linux.h>
84 #include <machine/../linux/linux_proto.h>
85 #endif
86 
87 #include <compat/linux/linux_file.h>
88 #include <compat/linux/linux_mib.h>
89 #include <compat/linux/linux_signal.h>
90 #include <compat/linux/linux_util.h>
91 #include <compat/linux/linux_sysproto.h>
92 #include <compat/linux/linux_emul.h>
93 #include <compat/linux/linux_misc.h>
94 
95 int stclohz;				/* Statistics clock frequency */
96 
97 #define BSD_TO_LINUX_SIGNAL(sig)	\
98 	(((sig) <= LINUX_SIGTBLSZ) ? bsd_to_linux_signal[_SIG_IDX(sig)] : sig)
99 
100 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
101 	RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
102 	RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
103 	RLIMIT_MEMLOCK, RLIMIT_AS
104 };
105 
106 struct l_sysinfo {
107 	l_long		uptime;		/* Seconds since boot */
108 	l_ulong		loads[3];	/* 1, 5, and 15 minute load averages */
109 #define LINUX_SYSINFO_LOADS_SCALE 65536
110 	l_ulong		totalram;	/* Total usable main memory size */
111 	l_ulong		freeram;	/* Available memory size */
112 	l_ulong		sharedram;	/* Amount of shared memory */
113 	l_ulong		bufferram;	/* Memory used by buffers */
114 	l_ulong		totalswap;	/* Total swap space size */
115 	l_ulong		freeswap;	/* swap space still available */
116 	l_ushort	procs;		/* Number of current processes */
117 	l_ushort	pads;
118 	l_ulong		totalbig;
119 	l_ulong		freebig;
120 	l_uint		mem_unit;
121 	char		_f[20-2*sizeof(l_long)-sizeof(l_int)];	/* padding */
122 };
123 int
124 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
125 {
126 	struct l_sysinfo sysinfo;
127 	vm_object_t object;
128 	int i, j;
129 	struct timespec ts;
130 
131 	getnanouptime(&ts);
132 	if (ts.tv_nsec != 0)
133 		ts.tv_sec++;
134 	sysinfo.uptime = ts.tv_sec;
135 
136 	/* Use the information from the mib to get our load averages */
137 	for (i = 0; i < 3; i++)
138 		sysinfo.loads[i] = averunnable.ldavg[i] *
139 		    LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
140 
141 	sysinfo.totalram = physmem * PAGE_SIZE;
142 	sysinfo.freeram = sysinfo.totalram - cnt.v_wire_count * PAGE_SIZE;
143 
144 	sysinfo.sharedram = 0;
145 	mtx_lock(&vm_object_list_mtx);
146 	TAILQ_FOREACH(object, &vm_object_list, object_list)
147 		if (object->shadow_count > 1)
148 			sysinfo.sharedram += object->resident_page_count;
149 	mtx_unlock(&vm_object_list_mtx);
150 
151 	sysinfo.sharedram *= PAGE_SIZE;
152 	sysinfo.bufferram = 0;
153 
154 	swap_pager_status(&i, &j);
155 	sysinfo.totalswap = i * PAGE_SIZE;
156 	sysinfo.freeswap = (i - j) * PAGE_SIZE;
157 
158 	sysinfo.procs = nprocs;
159 
160 	/* The following are only present in newer Linux kernels. */
161 	sysinfo.totalbig = 0;
162 	sysinfo.freebig = 0;
163 	sysinfo.mem_unit = 1;
164 
165 	return copyout(&sysinfo, args->info, sizeof(sysinfo));
166 }
167 
168 int
169 linux_alarm(struct thread *td, struct linux_alarm_args *args)
170 {
171 	struct itimerval it, old_it;
172 	u_int secs;
173 	int error;
174 
175 #ifdef DEBUG
176 	if (ldebug(alarm))
177 		printf(ARGS(alarm, "%u"), args->secs);
178 #endif
179 
180 	secs = args->secs;
181 
182 	if (secs > INT_MAX)
183 		secs = INT_MAX;
184 
185 	it.it_value.tv_sec = (long) secs;
186 	it.it_value.tv_usec = 0;
187 	it.it_interval.tv_sec = 0;
188 	it.it_interval.tv_usec = 0;
189 	error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
190 	if (error)
191 		return (error);
192 	if (timevalisset(&old_it.it_value)) {
193 		if (old_it.it_value.tv_usec != 0)
194 			old_it.it_value.tv_sec++;
195 		td->td_retval[0] = old_it.it_value.tv_sec;
196 	}
197 	return (0);
198 }
199 
200 int
201 linux_brk(struct thread *td, struct linux_brk_args *args)
202 {
203 	struct vmspace *vm = td->td_proc->p_vmspace;
204 	vm_offset_t new, old;
205 	struct obreak_args /* {
206 		char * nsize;
207 	} */ tmp;
208 
209 #ifdef DEBUG
210 	if (ldebug(brk))
211 		printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
212 #endif
213 	old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize);
214 	new = (vm_offset_t)args->dsend;
215 	tmp.nsize = (char *)new;
216 	if (((caddr_t)new > vm->vm_daddr) && !obreak(td, &tmp))
217 		td->td_retval[0] = (long)new;
218 	else
219 		td->td_retval[0] = (long)old;
220 
221 	return 0;
222 }
223 
224 #if defined(__i386__)
225 /* XXX: what about amd64/linux32? */
226 
227 int
228 linux_uselib(struct thread *td, struct linux_uselib_args *args)
229 {
230 	struct nameidata ni;
231 	struct vnode *vp;
232 	struct exec *a_out;
233 	struct vattr attr;
234 	vm_offset_t vmaddr;
235 	unsigned long file_offset;
236 	vm_offset_t buffer;
237 	unsigned long bss_size;
238 	char *library;
239 	int error;
240 	int locked, vfslocked;
241 
242 	LCONVPATHEXIST(td, args->library, &library);
243 
244 #ifdef DEBUG
245 	if (ldebug(uselib))
246 		printf(ARGS(uselib, "%s"), library);
247 #endif
248 
249 	a_out = NULL;
250 	vfslocked = 0;
251 	locked = 0;
252 	vp = NULL;
253 
254 	NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
255 	    UIO_SYSSPACE, library, td);
256 	error = namei(&ni);
257 	LFREEPATH(library);
258 	if (error)
259 		goto cleanup;
260 
261 	vp = ni.ni_vp;
262 	vfslocked = NDHASGIANT(&ni);
263 	NDFREE(&ni, NDF_ONLY_PNBUF);
264 
265 	/*
266 	 * From here on down, we have a locked vnode that must be unlocked.
267 	 * XXX: The code below largely duplicates exec_check_permissions().
268 	 */
269 	locked = 1;
270 
271 	/* Writable? */
272 	if (vp->v_writecount) {
273 		error = ETXTBSY;
274 		goto cleanup;
275 	}
276 
277 	/* Executable? */
278 	error = VOP_GETATTR(vp, &attr, td->td_ucred);
279 	if (error)
280 		goto cleanup;
281 
282 	if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
283 	    ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
284 		/* EACCESS is what exec(2) returns. */
285 		error = ENOEXEC;
286 		goto cleanup;
287 	}
288 
289 	/* Sensible size? */
290 	if (attr.va_size == 0) {
291 		error = ENOEXEC;
292 		goto cleanup;
293 	}
294 
295 	/* Can we access it? */
296 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
297 	if (error)
298 		goto cleanup;
299 
300 	/*
301 	 * XXX: This should use vn_open() so that it is properly authorized,
302 	 * and to reduce code redundancy all over the place here.
303 	 * XXX: Not really, it duplicates far more of exec_check_permissions()
304 	 * than vn_open().
305 	 */
306 #ifdef MAC
307 	error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
308 	if (error)
309 		goto cleanup;
310 #endif
311 	error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
312 	if (error)
313 		goto cleanup;
314 
315 	/* Pull in executable header into kernel_map */
316 	error = vm_mmap(kernel_map, (vm_offset_t *)&a_out, PAGE_SIZE,
317 	    VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
318 	if (error)
319 		goto cleanup;
320 
321 	/* Is it a Linux binary ? */
322 	if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
323 		error = ENOEXEC;
324 		goto cleanup;
325 	}
326 
327 	/*
328 	 * While we are here, we should REALLY do some more checks
329 	 */
330 
331 	/* Set file/virtual offset based on a.out variant. */
332 	switch ((int)(a_out->a_magic & 0xffff)) {
333 	case 0413:			/* ZMAGIC */
334 		file_offset = 1024;
335 		break;
336 	case 0314:			/* QMAGIC */
337 		file_offset = 0;
338 		break;
339 	default:
340 		error = ENOEXEC;
341 		goto cleanup;
342 	}
343 
344 	bss_size = round_page(a_out->a_bss);
345 
346 	/* Check various fields in header for validity/bounds. */
347 	if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
348 		error = ENOEXEC;
349 		goto cleanup;
350 	}
351 
352 	/* text + data can't exceed file size */
353 	if (a_out->a_data + a_out->a_text > attr.va_size) {
354 		error = EFAULT;
355 		goto cleanup;
356 	}
357 
358 	/*
359 	 * text/data/bss must not exceed limits
360 	 * XXX - this is not complete. it should check current usage PLUS
361 	 * the resources needed by this library.
362 	 */
363 	PROC_LOCK(td->td_proc);
364 	if (a_out->a_text > maxtsiz ||
365 	    a_out->a_data + bss_size > lim_cur(td->td_proc, RLIMIT_DATA)) {
366 		PROC_UNLOCK(td->td_proc);
367 		error = ENOMEM;
368 		goto cleanup;
369 	}
370 	PROC_UNLOCK(td->td_proc);
371 
372 	/*
373 	 * Prevent more writers.
374 	 * XXX: Note that if any of the VM operations fail below we don't
375 	 * clear this flag.
376 	 */
377 	vp->v_vflag |= VV_TEXT;
378 
379 	/*
380 	 * Lock no longer needed
381 	 */
382 	locked = 0;
383 	VOP_UNLOCK(vp, 0);
384 	VFS_UNLOCK_GIANT(vfslocked);
385 
386 	/*
387 	 * Check if file_offset page aligned. Currently we cannot handle
388 	 * misalinged file offsets, and so we read in the entire image
389 	 * (what a waste).
390 	 */
391 	if (file_offset & PAGE_MASK) {
392 #ifdef DEBUG
393 		printf("uselib: Non page aligned binary %lu\n", file_offset);
394 #endif
395 		/* Map text+data read/write/execute */
396 
397 		/* a_entry is the load address and is page aligned */
398 		vmaddr = trunc_page(a_out->a_entry);
399 
400 		/* get anon user mapping, read+write+execute */
401 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
402 		    &vmaddr, a_out->a_text + a_out->a_data, FALSE, VM_PROT_ALL,
403 		    VM_PROT_ALL, 0);
404 		if (error)
405 			goto cleanup;
406 
407 		/* map file into kernel_map */
408 		error = vm_mmap(kernel_map, &buffer,
409 		    round_page(a_out->a_text + a_out->a_data + file_offset),
410 		    VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp,
411 		    trunc_page(file_offset));
412 		if (error)
413 			goto cleanup;
414 
415 		/* copy from kernel VM space to user space */
416 		error = copyout(PTRIN(buffer + file_offset),
417 		    (void *)vmaddr, a_out->a_text + a_out->a_data);
418 
419 		/* release temporary kernel space */
420 		vm_map_remove(kernel_map, buffer, buffer +
421 		    round_page(a_out->a_text + a_out->a_data + file_offset));
422 
423 		if (error)
424 			goto cleanup;
425 	} else {
426 #ifdef DEBUG
427 		printf("uselib: Page aligned binary %lu\n", file_offset);
428 #endif
429 		/*
430 		 * for QMAGIC, a_entry is 20 bytes beyond the load address
431 		 * to skip the executable header
432 		 */
433 		vmaddr = trunc_page(a_out->a_entry);
434 
435 		/*
436 		 * Map it all into the process's space as a single
437 		 * copy-on-write "data" segment.
438 		 */
439 		error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr,
440 		    a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
441 		    MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
442 		if (error)
443 			goto cleanup;
444 	}
445 #ifdef DEBUG
446 	printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
447 	    ((long *)vmaddr)[1]);
448 #endif
449 	if (bss_size != 0) {
450 		/* Calculate BSS start address */
451 		vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
452 		    a_out->a_data;
453 
454 		/* allocate some 'anon' space */
455 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
456 		    &vmaddr, bss_size, FALSE, VM_PROT_ALL, VM_PROT_ALL, 0);
457 		if (error)
458 			goto cleanup;
459 	}
460 
461 cleanup:
462 	/* Unlock vnode if needed */
463 	if (locked) {
464 		VOP_UNLOCK(vp, 0);
465 		VFS_UNLOCK_GIANT(vfslocked);
466 	}
467 
468 	/* Release the kernel mapping. */
469 	if (a_out)
470 		vm_map_remove(kernel_map, (vm_offset_t)a_out,
471 		    (vm_offset_t)a_out + PAGE_SIZE);
472 
473 	return error;
474 }
475 
476 #endif	/* __i386__ */
477 
478 int
479 linux_select(struct thread *td, struct linux_select_args *args)
480 {
481 	l_timeval ltv;
482 	struct timeval tv0, tv1, utv, *tvp;
483 	int error;
484 
485 #ifdef DEBUG
486 	if (ldebug(select))
487 		printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
488 		    (void *)args->readfds, (void *)args->writefds,
489 		    (void *)args->exceptfds, (void *)args->timeout);
490 #endif
491 
492 	/*
493 	 * Store current time for computation of the amount of
494 	 * time left.
495 	 */
496 	if (args->timeout) {
497 		if ((error = copyin(args->timeout, &ltv, sizeof(ltv))))
498 			goto select_out;
499 		utv.tv_sec = ltv.tv_sec;
500 		utv.tv_usec = ltv.tv_usec;
501 #ifdef DEBUG
502 		if (ldebug(select))
503 			printf(LMSG("incoming timeout (%jd/%ld)"),
504 			    (intmax_t)utv.tv_sec, utv.tv_usec);
505 #endif
506 
507 		if (itimerfix(&utv)) {
508 			/*
509 			 * The timeval was invalid.  Convert it to something
510 			 * valid that will act as it does under Linux.
511 			 */
512 			utv.tv_sec += utv.tv_usec / 1000000;
513 			utv.tv_usec %= 1000000;
514 			if (utv.tv_usec < 0) {
515 				utv.tv_sec -= 1;
516 				utv.tv_usec += 1000000;
517 			}
518 			if (utv.tv_sec < 0)
519 				timevalclear(&utv);
520 		}
521 		microtime(&tv0);
522 		tvp = &utv;
523 	} else
524 		tvp = NULL;
525 
526 	error = kern_select(td, args->nfds, args->readfds, args->writefds,
527 	    args->exceptfds, tvp);
528 
529 #ifdef DEBUG
530 	if (ldebug(select))
531 		printf(LMSG("real select returns %d"), error);
532 #endif
533 	if (error)
534 		goto select_out;
535 
536 	if (args->timeout) {
537 		if (td->td_retval[0]) {
538 			/*
539 			 * Compute how much time was left of the timeout,
540 			 * by subtracting the current time and the time
541 			 * before we started the call, and subtracting
542 			 * that result from the user-supplied value.
543 			 */
544 			microtime(&tv1);
545 			timevalsub(&tv1, &tv0);
546 			timevalsub(&utv, &tv1);
547 			if (utv.tv_sec < 0)
548 				timevalclear(&utv);
549 		} else
550 			timevalclear(&utv);
551 #ifdef DEBUG
552 		if (ldebug(select))
553 			printf(LMSG("outgoing timeout (%jd/%ld)"),
554 			    (intmax_t)utv.tv_sec, utv.tv_usec);
555 #endif
556 		ltv.tv_sec = utv.tv_sec;
557 		ltv.tv_usec = utv.tv_usec;
558 		if ((error = copyout(&ltv, args->timeout, sizeof(ltv))))
559 			goto select_out;
560 	}
561 
562 select_out:
563 #ifdef DEBUG
564 	if (ldebug(select))
565 		printf(LMSG("select_out -> %d"), error);
566 #endif
567 	return error;
568 }
569 
570 int
571 linux_mremap(struct thread *td, struct linux_mremap_args *args)
572 {
573 	struct munmap_args /* {
574 		void *addr;
575 		size_t len;
576 	} */ bsd_args;
577 	int error = 0;
578 
579 #ifdef DEBUG
580 	if (ldebug(mremap))
581 		printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
582 		    (void *)(uintptr_t)args->addr,
583 		    (unsigned long)args->old_len,
584 		    (unsigned long)args->new_len,
585 		    (unsigned long)args->flags);
586 #endif
587 
588 	if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
589 		td->td_retval[0] = 0;
590 		return (EINVAL);
591 	}
592 
593 	/*
594 	 * Check for the page alignment.
595 	 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
596 	 */
597 	if (args->addr & PAGE_MASK) {
598 		td->td_retval[0] = 0;
599 		return (EINVAL);
600 	}
601 
602 	args->new_len = round_page(args->new_len);
603 	args->old_len = round_page(args->old_len);
604 
605 	if (args->new_len > args->old_len) {
606 		td->td_retval[0] = 0;
607 		return ENOMEM;
608 	}
609 
610 	if (args->new_len < args->old_len) {
611 		bsd_args.addr =
612 		    (caddr_t)((uintptr_t)args->addr + args->new_len);
613 		bsd_args.len = args->old_len - args->new_len;
614 		error = munmap(td, &bsd_args);
615 	}
616 
617 	td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
618 	return error;
619 }
620 
621 #define LINUX_MS_ASYNC       0x0001
622 #define LINUX_MS_INVALIDATE  0x0002
623 #define LINUX_MS_SYNC        0x0004
624 
625 int
626 linux_msync(struct thread *td, struct linux_msync_args *args)
627 {
628 	struct msync_args bsd_args;
629 
630 	bsd_args.addr = (caddr_t)(uintptr_t)args->addr;
631 	bsd_args.len = (uintptr_t)args->len;
632 	bsd_args.flags = args->fl & ~LINUX_MS_SYNC;
633 
634 	return msync(td, &bsd_args);
635 }
636 
637 int
638 linux_time(struct thread *td, struct linux_time_args *args)
639 {
640 	struct timeval tv;
641 	l_time_t tm;
642 	int error;
643 
644 #ifdef DEBUG
645 	if (ldebug(time))
646 		printf(ARGS(time, "*"));
647 #endif
648 
649 	microtime(&tv);
650 	tm = tv.tv_sec;
651 	if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
652 		return error;
653 	td->td_retval[0] = tm;
654 	return 0;
655 }
656 
657 struct l_times_argv {
658 	l_clock_t	tms_utime;
659 	l_clock_t	tms_stime;
660 	l_clock_t	tms_cutime;
661 	l_clock_t	tms_cstime;
662 };
663 
664 
665 /*
666  * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
667  * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
668  * auxiliary vector entry.
669  */
670 #define	CLK_TCK		100
671 
672 #define	CONVOTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
673 #define	CONVNTCK(r)	(r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
674 
675 #define	CONVTCK(r)	(linux_kernver(td) >= LINUX_KERNVER_2004000 ?		\
676 			    CONVNTCK(r) : CONVOTCK(r))
677 
678 int
679 linux_times(struct thread *td, struct linux_times_args *args)
680 {
681 	struct timeval tv, utime, stime, cutime, cstime;
682 	struct l_times_argv tms;
683 	struct proc *p;
684 	int error;
685 
686 #ifdef DEBUG
687 	if (ldebug(times))
688 		printf(ARGS(times, "*"));
689 #endif
690 
691 	if (args->buf != NULL) {
692 		p = td->td_proc;
693 		PROC_LOCK(p);
694 		PROC_SLOCK(p);
695 		calcru(p, &utime, &stime);
696 		PROC_SUNLOCK(p);
697 		calccru(p, &cutime, &cstime);
698 		PROC_UNLOCK(p);
699 
700 		tms.tms_utime = CONVTCK(utime);
701 		tms.tms_stime = CONVTCK(stime);
702 
703 		tms.tms_cutime = CONVTCK(cutime);
704 		tms.tms_cstime = CONVTCK(cstime);
705 
706 		if ((error = copyout(&tms, args->buf, sizeof(tms))))
707 			return error;
708 	}
709 
710 	microuptime(&tv);
711 	td->td_retval[0] = (int)CONVTCK(tv);
712 	return 0;
713 }
714 
715 int
716 linux_newuname(struct thread *td, struct linux_newuname_args *args)
717 {
718 	INIT_VPROCG(TD_TO_VPROCG(td));
719 	struct l_new_utsname utsname;
720 	char osname[LINUX_MAX_UTSNAME];
721 	char osrelease[LINUX_MAX_UTSNAME];
722 	char *p;
723 
724 #ifdef DEBUG
725 	if (ldebug(newuname))
726 		printf(ARGS(newuname, "*"));
727 #endif
728 
729 	linux_get_osname(td, osname);
730 	linux_get_osrelease(td, osrelease);
731 
732 	bzero(&utsname, sizeof(utsname));
733 	strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
734 	getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
735 	strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
736 	strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
737 	for (p = utsname.version; *p != '\0'; ++p)
738 		if (*p == '\n') {
739 			*p = '\0';
740 			break;
741 		}
742 	strlcpy(utsname.machine, linux_platform, LINUX_MAX_UTSNAME);
743 
744 	mtx_lock(&hostname_mtx);
745 	strlcpy(utsname.domainname, V_domainname, LINUX_MAX_UTSNAME);
746 	mtx_unlock(&hostname_mtx);
747 
748 	return (copyout(&utsname, args->buf, sizeof(utsname)));
749 }
750 
751 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
752 struct l_utimbuf {
753 	l_time_t l_actime;
754 	l_time_t l_modtime;
755 };
756 
757 int
758 linux_utime(struct thread *td, struct linux_utime_args *args)
759 {
760 	struct timeval tv[2], *tvp;
761 	struct l_utimbuf lut;
762 	char *fname;
763 	int error;
764 
765 	LCONVPATHEXIST(td, args->fname, &fname);
766 
767 #ifdef DEBUG
768 	if (ldebug(utime))
769 		printf(ARGS(utime, "%s, *"), fname);
770 #endif
771 
772 	if (args->times) {
773 		if ((error = copyin(args->times, &lut, sizeof lut))) {
774 			LFREEPATH(fname);
775 			return error;
776 		}
777 		tv[0].tv_sec = lut.l_actime;
778 		tv[0].tv_usec = 0;
779 		tv[1].tv_sec = lut.l_modtime;
780 		tv[1].tv_usec = 0;
781 		tvp = tv;
782 	} else
783 		tvp = NULL;
784 
785 	error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
786 	LFREEPATH(fname);
787 	return (error);
788 }
789 
790 int
791 linux_utimes(struct thread *td, struct linux_utimes_args *args)
792 {
793 	l_timeval ltv[2];
794 	struct timeval tv[2], *tvp = NULL;
795 	char *fname;
796 	int error;
797 
798 	LCONVPATHEXIST(td, args->fname, &fname);
799 
800 #ifdef DEBUG
801 	if (ldebug(utimes))
802 		printf(ARGS(utimes, "%s, *"), fname);
803 #endif
804 
805 	if (args->tptr != NULL) {
806 		if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
807 			LFREEPATH(fname);
808 			return (error);
809 		}
810 		tv[0].tv_sec = ltv[0].tv_sec;
811 		tv[0].tv_usec = ltv[0].tv_usec;
812 		tv[1].tv_sec = ltv[1].tv_sec;
813 		tv[1].tv_usec = ltv[1].tv_usec;
814 		tvp = tv;
815 	}
816 
817 	error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
818 	LFREEPATH(fname);
819 	return (error);
820 }
821 
822 int
823 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
824 {
825 	l_timeval ltv[2];
826 	struct timeval tv[2], *tvp = NULL;
827 	char *fname;
828 	int error, dfd;
829 
830 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
831 	LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
832 
833 #ifdef DEBUG
834 	if (ldebug(futimesat))
835 		printf(ARGS(futimesat, "%s, *"), fname);
836 #endif
837 
838 	if (args->utimes != NULL) {
839 		if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
840 			LFREEPATH(fname);
841 			return (error);
842 		}
843 		tv[0].tv_sec = ltv[0].tv_sec;
844 		tv[0].tv_usec = ltv[0].tv_usec;
845 		tv[1].tv_sec = ltv[1].tv_sec;
846 		tv[1].tv_usec = ltv[1].tv_usec;
847 		tvp = tv;
848 	}
849 
850 	error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
851 	LFREEPATH(fname);
852 	return (error);
853 }
854 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
855 
856 #define __WCLONE 0x80000000
857 
858 int
859 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
860 {
861 	int error, options, tmpstat;
862 
863 #ifdef DEBUG
864 	if (ldebug(waitpid))
865 		printf(ARGS(waitpid, "%d, %p, %d"),
866 		    args->pid, (void *)args->status, args->options);
867 #endif
868 	/*
869 	 * this is necessary because the test in kern_wait doesn't work
870 	 * because we mess with the options here
871 	 */
872 	if (args->options & ~(WUNTRACED | WNOHANG | WCONTINUED | __WCLONE))
873 		return (EINVAL);
874 
875 	options = (args->options & (WNOHANG | WUNTRACED));
876 	/* WLINUXCLONE should be equal to __WCLONE, but we make sure */
877 	if (args->options & __WCLONE)
878 		options |= WLINUXCLONE;
879 
880 	error = kern_wait(td, args->pid, &tmpstat, options, NULL);
881 	if (error)
882 		return error;
883 
884 	if (args->status) {
885 		tmpstat &= 0xffff;
886 		if (WIFSIGNALED(tmpstat))
887 			tmpstat = (tmpstat & 0xffffff80) |
888 			    BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat));
889 		else if (WIFSTOPPED(tmpstat))
890 			tmpstat = (tmpstat & 0xffff00ff) |
891 			    (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8);
892 		return copyout(&tmpstat, args->status, sizeof(int));
893 	}
894 
895 	return 0;
896 }
897 
898 int
899 linux_wait4(struct thread *td, struct linux_wait4_args *args)
900 {
901 	int error, options, tmpstat;
902 	struct rusage ru, *rup;
903 	struct proc *p;
904 
905 #ifdef DEBUG
906 	if (ldebug(wait4))
907 		printf(ARGS(wait4, "%d, %p, %d, %p"),
908 		    args->pid, (void *)args->status, args->options,
909 		    (void *)args->rusage);
910 #endif
911 
912 	options = (args->options & (WNOHANG | WUNTRACED));
913 	/* WLINUXCLONE should be equal to __WCLONE, but we make sure */
914 	if (args->options & __WCLONE)
915 		options |= WLINUXCLONE;
916 
917 	if (args->rusage != NULL)
918 		rup = &ru;
919 	else
920 		rup = NULL;
921 	error = kern_wait(td, args->pid, &tmpstat, options, rup);
922 	if (error)
923 		return error;
924 
925 	p = td->td_proc;
926 	PROC_LOCK(p);
927 	sigqueue_delete(&p->p_sigqueue, SIGCHLD);
928 	PROC_UNLOCK(p);
929 
930 	if (args->status) {
931 		tmpstat &= 0xffff;
932 		if (WIFSIGNALED(tmpstat))
933 			tmpstat = (tmpstat & 0xffffff80) |
934 			    BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat));
935 		else if (WIFSTOPPED(tmpstat))
936 			tmpstat = (tmpstat & 0xffff00ff) |
937 			    (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8);
938 		error = copyout(&tmpstat, args->status, sizeof(int));
939 	}
940 	if (args->rusage != NULL && error == 0)
941 		error = copyout(&ru, args->rusage, sizeof(ru));
942 
943 	return (error);
944 }
945 
946 int
947 linux_mknod(struct thread *td, struct linux_mknod_args *args)
948 {
949 	char *path;
950 	int error;
951 
952 	LCONVPATHCREAT(td, args->path, &path);
953 
954 #ifdef DEBUG
955 	if (ldebug(mknod))
956 		printf(ARGS(mknod, "%s, %d, %d"), path, args->mode, args->dev);
957 #endif
958 
959 	switch (args->mode & S_IFMT) {
960 	case S_IFIFO:
961 	case S_IFSOCK:
962 		error = kern_mkfifo(td, path, UIO_SYSSPACE, args->mode);
963 		break;
964 
965 	case S_IFCHR:
966 	case S_IFBLK:
967 		error = kern_mknod(td, path, UIO_SYSSPACE, args->mode,
968 		    args->dev);
969 		break;
970 
971 	case S_IFDIR:
972 		error = EPERM;
973 		break;
974 
975 	case 0:
976 		args->mode |= S_IFREG;
977 		/* FALLTHROUGH */
978 	case S_IFREG:
979 		error = kern_open(td, path, UIO_SYSSPACE,
980 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
981 		if (error == 0)
982 			kern_close(td, td->td_retval[0]);
983 		break;
984 
985 	default:
986 		error = EINVAL;
987 		break;
988 	}
989 	LFREEPATH(path);
990 	return (error);
991 }
992 
993 int
994 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
995 {
996 	char *path;
997 	int error, dfd;
998 
999 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
1000 	LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
1001 
1002 #ifdef DEBUG
1003 	if (ldebug(mknodat))
1004 		printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev);
1005 #endif
1006 
1007 	switch (args->mode & S_IFMT) {
1008 	case S_IFIFO:
1009 	case S_IFSOCK:
1010 		error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
1011 		break;
1012 
1013 	case S_IFCHR:
1014 	case S_IFBLK:
1015 		error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
1016 		    args->dev);
1017 		break;
1018 
1019 	case S_IFDIR:
1020 		error = EPERM;
1021 		break;
1022 
1023 	case 0:
1024 		args->mode |= S_IFREG;
1025 		/* FALLTHROUGH */
1026 	case S_IFREG:
1027 		error = kern_openat(td, dfd, path, UIO_SYSSPACE,
1028 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1029 		if (error == 0)
1030 			kern_close(td, td->td_retval[0]);
1031 		break;
1032 
1033 	default:
1034 		error = EINVAL;
1035 		break;
1036 	}
1037 	LFREEPATH(path);
1038 	return (error);
1039 }
1040 
1041 /*
1042  * UGH! This is just about the dumbest idea I've ever heard!!
1043  */
1044 int
1045 linux_personality(struct thread *td, struct linux_personality_args *args)
1046 {
1047 #ifdef DEBUG
1048 	if (ldebug(personality))
1049 		printf(ARGS(personality, "%lu"), (unsigned long)args->per);
1050 #endif
1051 	if (args->per != 0)
1052 		return EINVAL;
1053 
1054 	/* Yes Jim, it's still a Linux... */
1055 	td->td_retval[0] = 0;
1056 	return 0;
1057 }
1058 
1059 struct l_itimerval {
1060 	l_timeval it_interval;
1061 	l_timeval it_value;
1062 };
1063 
1064 #define	B2L_ITIMERVAL(bip, lip) 					\
1065 	(bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec;		\
1066 	(bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec;	\
1067 	(bip)->it_value.tv_sec = (lip)->it_value.tv_sec;		\
1068 	(bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1069 
1070 int
1071 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1072 {
1073 	int error;
1074 	struct l_itimerval ls;
1075 	struct itimerval aitv, oitv;
1076 
1077 #ifdef DEBUG
1078 	if (ldebug(setitimer))
1079 		printf(ARGS(setitimer, "%p, %p"),
1080 		    (void *)uap->itv, (void *)uap->oitv);
1081 #endif
1082 
1083 	if (uap->itv == NULL) {
1084 		uap->itv = uap->oitv;
1085 		return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1086 	}
1087 
1088 	error = copyin(uap->itv, &ls, sizeof(ls));
1089 	if (error != 0)
1090 		return (error);
1091 	B2L_ITIMERVAL(&aitv, &ls);
1092 #ifdef DEBUG
1093 	if (ldebug(setitimer)) {
1094 		printf("setitimer: value: sec: %jd, usec: %ld\n",
1095 		    (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1096 		printf("setitimer: interval: sec: %jd, usec: %ld\n",
1097 		    (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1098 	}
1099 #endif
1100 	error = kern_setitimer(td, uap->which, &aitv, &oitv);
1101 	if (error != 0 || uap->oitv == NULL)
1102 		return (error);
1103 	B2L_ITIMERVAL(&ls, &oitv);
1104 
1105 	return (copyout(&ls, uap->oitv, sizeof(ls)));
1106 }
1107 
1108 int
1109 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1110 {
1111 	int error;
1112 	struct l_itimerval ls;
1113 	struct itimerval aitv;
1114 
1115 #ifdef DEBUG
1116 	if (ldebug(getitimer))
1117 		printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1118 #endif
1119 	error = kern_getitimer(td, uap->which, &aitv);
1120 	if (error != 0)
1121 		return (error);
1122 	B2L_ITIMERVAL(&ls, &aitv);
1123 	return (copyout(&ls, uap->itv, sizeof(ls)));
1124 }
1125 
1126 int
1127 linux_nice(struct thread *td, struct linux_nice_args *args)
1128 {
1129 	struct setpriority_args bsd_args;
1130 
1131 	bsd_args.which = PRIO_PROCESS;
1132 	bsd_args.who = 0;		/* current process */
1133 	bsd_args.prio = args->inc;
1134 	return setpriority(td, &bsd_args);
1135 }
1136 
1137 int
1138 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1139 {
1140 	struct ucred *newcred, *oldcred;
1141 	l_gid_t linux_gidset[NGROUPS];
1142 	gid_t *bsd_gidset;
1143 	int ngrp, error;
1144 	struct proc *p;
1145 
1146 	ngrp = args->gidsetsize;
1147 	if (ngrp < 0 || ngrp >= NGROUPS)
1148 		return (EINVAL);
1149 	error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1150 	if (error)
1151 		return (error);
1152 	newcred = crget();
1153 	p = td->td_proc;
1154 	PROC_LOCK(p);
1155 	oldcred = p->p_ucred;
1156 
1157 	/*
1158 	 * cr_groups[0] holds egid. Setting the whole set from
1159 	 * the supplied set will cause egid to be changed too.
1160 	 * Keep cr_groups[0] unchanged to prevent that.
1161 	 */
1162 
1163 	if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) {
1164 		PROC_UNLOCK(p);
1165 		crfree(newcred);
1166 		return (error);
1167 	}
1168 
1169 	crcopy(newcred, oldcred);
1170 	if (ngrp > 0) {
1171 		newcred->cr_ngroups = ngrp + 1;
1172 
1173 		bsd_gidset = newcred->cr_groups;
1174 		ngrp--;
1175 		while (ngrp >= 0) {
1176 			bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1177 			ngrp--;
1178 		}
1179 	} else
1180 		newcred->cr_ngroups = 1;
1181 
1182 	setsugid(p);
1183 	p->p_ucred = newcred;
1184 	PROC_UNLOCK(p);
1185 	crfree(oldcred);
1186 	return (0);
1187 }
1188 
1189 int
1190 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1191 {
1192 	struct ucred *cred;
1193 	l_gid_t linux_gidset[NGROUPS];
1194 	gid_t *bsd_gidset;
1195 	int bsd_gidsetsz, ngrp, error;
1196 
1197 	cred = td->td_ucred;
1198 	bsd_gidset = cred->cr_groups;
1199 	bsd_gidsetsz = cred->cr_ngroups - 1;
1200 
1201 	/*
1202 	 * cr_groups[0] holds egid. Returning the whole set
1203 	 * here will cause a duplicate. Exclude cr_groups[0]
1204 	 * to prevent that.
1205 	 */
1206 
1207 	if ((ngrp = args->gidsetsize) == 0) {
1208 		td->td_retval[0] = bsd_gidsetsz;
1209 		return (0);
1210 	}
1211 
1212 	if (ngrp < bsd_gidsetsz)
1213 		return (EINVAL);
1214 
1215 	ngrp = 0;
1216 	while (ngrp < bsd_gidsetsz) {
1217 		linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1218 		ngrp++;
1219 	}
1220 
1221 	if ((error = copyout(linux_gidset, args->grouplist,
1222 	    ngrp * sizeof(l_gid_t))))
1223 		return (error);
1224 
1225 	td->td_retval[0] = ngrp;
1226 	return (0);
1227 }
1228 
1229 int
1230 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1231 {
1232 	struct rlimit bsd_rlim;
1233 	struct l_rlimit rlim;
1234 	u_int which;
1235 	int error;
1236 
1237 #ifdef DEBUG
1238 	if (ldebug(setrlimit))
1239 		printf(ARGS(setrlimit, "%d, %p"),
1240 		    args->resource, (void *)args->rlim);
1241 #endif
1242 
1243 	if (args->resource >= LINUX_RLIM_NLIMITS)
1244 		return (EINVAL);
1245 
1246 	which = linux_to_bsd_resource[args->resource];
1247 	if (which == -1)
1248 		return (EINVAL);
1249 
1250 	error = copyin(args->rlim, &rlim, sizeof(rlim));
1251 	if (error)
1252 		return (error);
1253 
1254 	bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1255 	bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1256 	return (kern_setrlimit(td, which, &bsd_rlim));
1257 }
1258 
1259 int
1260 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1261 {
1262 	struct l_rlimit rlim;
1263 	struct proc *p = td->td_proc;
1264 	struct rlimit bsd_rlim;
1265 	u_int which;
1266 
1267 #ifdef DEBUG
1268 	if (ldebug(old_getrlimit))
1269 		printf(ARGS(old_getrlimit, "%d, %p"),
1270 		    args->resource, (void *)args->rlim);
1271 #endif
1272 
1273 	if (args->resource >= LINUX_RLIM_NLIMITS)
1274 		return (EINVAL);
1275 
1276 	which = linux_to_bsd_resource[args->resource];
1277 	if (which == -1)
1278 		return (EINVAL);
1279 
1280 	PROC_LOCK(p);
1281 	lim_rlimit(p, which, &bsd_rlim);
1282 	PROC_UNLOCK(p);
1283 
1284 #ifdef COMPAT_LINUX32
1285 	rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1286 	if (rlim.rlim_cur == UINT_MAX)
1287 		rlim.rlim_cur = INT_MAX;
1288 	rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1289 	if (rlim.rlim_max == UINT_MAX)
1290 		rlim.rlim_max = INT_MAX;
1291 #else
1292 	rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1293 	if (rlim.rlim_cur == ULONG_MAX)
1294 		rlim.rlim_cur = LONG_MAX;
1295 	rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1296 	if (rlim.rlim_max == ULONG_MAX)
1297 		rlim.rlim_max = LONG_MAX;
1298 #endif
1299 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1300 }
1301 
1302 int
1303 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1304 {
1305 	struct l_rlimit rlim;
1306 	struct proc *p = td->td_proc;
1307 	struct rlimit bsd_rlim;
1308 	u_int which;
1309 
1310 #ifdef DEBUG
1311 	if (ldebug(getrlimit))
1312 		printf(ARGS(getrlimit, "%d, %p"),
1313 		    args->resource, (void *)args->rlim);
1314 #endif
1315 
1316 	if (args->resource >= LINUX_RLIM_NLIMITS)
1317 		return (EINVAL);
1318 
1319 	which = linux_to_bsd_resource[args->resource];
1320 	if (which == -1)
1321 		return (EINVAL);
1322 
1323 	PROC_LOCK(p);
1324 	lim_rlimit(p, which, &bsd_rlim);
1325 	PROC_UNLOCK(p);
1326 
1327 	rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1328 	rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1329 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1330 }
1331 
1332 int
1333 linux_sched_setscheduler(struct thread *td,
1334     struct linux_sched_setscheduler_args *args)
1335 {
1336 	struct sched_setscheduler_args bsd;
1337 
1338 #ifdef DEBUG
1339 	if (ldebug(sched_setscheduler))
1340 		printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1341 		    args->pid, args->policy, (const void *)args->param);
1342 #endif
1343 
1344 	switch (args->policy) {
1345 	case LINUX_SCHED_OTHER:
1346 		bsd.policy = SCHED_OTHER;
1347 		break;
1348 	case LINUX_SCHED_FIFO:
1349 		bsd.policy = SCHED_FIFO;
1350 		break;
1351 	case LINUX_SCHED_RR:
1352 		bsd.policy = SCHED_RR;
1353 		break;
1354 	default:
1355 		return EINVAL;
1356 	}
1357 
1358 	bsd.pid = args->pid;
1359 	bsd.param = (struct sched_param *)args->param;
1360 	return sched_setscheduler(td, &bsd);
1361 }
1362 
1363 int
1364 linux_sched_getscheduler(struct thread *td,
1365     struct linux_sched_getscheduler_args *args)
1366 {
1367 	struct sched_getscheduler_args bsd;
1368 	int error;
1369 
1370 #ifdef DEBUG
1371 	if (ldebug(sched_getscheduler))
1372 		printf(ARGS(sched_getscheduler, "%d"), args->pid);
1373 #endif
1374 
1375 	bsd.pid = args->pid;
1376 	error = sched_getscheduler(td, &bsd);
1377 
1378 	switch (td->td_retval[0]) {
1379 	case SCHED_OTHER:
1380 		td->td_retval[0] = LINUX_SCHED_OTHER;
1381 		break;
1382 	case SCHED_FIFO:
1383 		td->td_retval[0] = LINUX_SCHED_FIFO;
1384 		break;
1385 	case SCHED_RR:
1386 		td->td_retval[0] = LINUX_SCHED_RR;
1387 		break;
1388 	}
1389 
1390 	return error;
1391 }
1392 
1393 int
1394 linux_sched_get_priority_max(struct thread *td,
1395     struct linux_sched_get_priority_max_args *args)
1396 {
1397 	struct sched_get_priority_max_args bsd;
1398 
1399 #ifdef DEBUG
1400 	if (ldebug(sched_get_priority_max))
1401 		printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1402 #endif
1403 
1404 	switch (args->policy) {
1405 	case LINUX_SCHED_OTHER:
1406 		bsd.policy = SCHED_OTHER;
1407 		break;
1408 	case LINUX_SCHED_FIFO:
1409 		bsd.policy = SCHED_FIFO;
1410 		break;
1411 	case LINUX_SCHED_RR:
1412 		bsd.policy = SCHED_RR;
1413 		break;
1414 	default:
1415 		return EINVAL;
1416 	}
1417 	return sched_get_priority_max(td, &bsd);
1418 }
1419 
1420 int
1421 linux_sched_get_priority_min(struct thread *td,
1422     struct linux_sched_get_priority_min_args *args)
1423 {
1424 	struct sched_get_priority_min_args bsd;
1425 
1426 #ifdef DEBUG
1427 	if (ldebug(sched_get_priority_min))
1428 		printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1429 #endif
1430 
1431 	switch (args->policy) {
1432 	case LINUX_SCHED_OTHER:
1433 		bsd.policy = SCHED_OTHER;
1434 		break;
1435 	case LINUX_SCHED_FIFO:
1436 		bsd.policy = SCHED_FIFO;
1437 		break;
1438 	case LINUX_SCHED_RR:
1439 		bsd.policy = SCHED_RR;
1440 		break;
1441 	default:
1442 		return EINVAL;
1443 	}
1444 	return sched_get_priority_min(td, &bsd);
1445 }
1446 
1447 #define REBOOT_CAD_ON	0x89abcdef
1448 #define REBOOT_CAD_OFF	0
1449 #define REBOOT_HALT	0xcdef0123
1450 #define REBOOT_RESTART	0x01234567
1451 #define REBOOT_RESTART2	0xA1B2C3D4
1452 #define REBOOT_POWEROFF	0x4321FEDC
1453 #define REBOOT_MAGIC1	0xfee1dead
1454 #define REBOOT_MAGIC2	0x28121969
1455 #define REBOOT_MAGIC2A	0x05121996
1456 #define REBOOT_MAGIC2B	0x16041998
1457 
1458 int
1459 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1460 {
1461 	struct reboot_args bsd_args;
1462 
1463 #ifdef DEBUG
1464 	if (ldebug(reboot))
1465 		printf(ARGS(reboot, "0x%x"), args->cmd);
1466 #endif
1467 
1468 	if (args->magic1 != REBOOT_MAGIC1)
1469 		return EINVAL;
1470 
1471 	switch (args->magic2) {
1472 	case REBOOT_MAGIC2:
1473 	case REBOOT_MAGIC2A:
1474 	case REBOOT_MAGIC2B:
1475 		break;
1476 	default:
1477 		return EINVAL;
1478 	}
1479 
1480 	switch (args->cmd) {
1481 	case REBOOT_CAD_ON:
1482 	case REBOOT_CAD_OFF:
1483 		return (priv_check(td, PRIV_REBOOT));
1484 	case REBOOT_HALT:
1485 		bsd_args.opt = RB_HALT;
1486 		break;
1487 	case REBOOT_RESTART:
1488 	case REBOOT_RESTART2:
1489 		bsd_args.opt = 0;
1490 		break;
1491 	case REBOOT_POWEROFF:
1492 		bsd_args.opt = RB_POWEROFF;
1493 		break;
1494 	default:
1495 		return EINVAL;
1496 	}
1497 	return reboot(td, &bsd_args);
1498 }
1499 
1500 
1501 /*
1502  * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1503  * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
1504  * are assumed to be preserved. The following lightweight syscalls fixes
1505  * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
1506  *
1507  * linux_getpid() - MP SAFE
1508  * linux_getgid() - MP SAFE
1509  * linux_getuid() - MP SAFE
1510  */
1511 
1512 int
1513 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1514 {
1515 	struct linux_emuldata *em;
1516 
1517 #ifdef DEBUG
1518 	if (ldebug(getpid))
1519 		printf(ARGS(getpid, ""));
1520 #endif
1521 
1522 	if (linux_use26(td)) {
1523 		em = em_find(td->td_proc, EMUL_DONTLOCK);
1524 		KASSERT(em != NULL, ("getpid: emuldata not found.\n"));
1525 		td->td_retval[0] = em->shared->group_pid;
1526 	} else {
1527 		td->td_retval[0] = td->td_proc->p_pid;
1528 	}
1529 
1530 	return (0);
1531 }
1532 
1533 int
1534 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1535 {
1536 
1537 #ifdef DEBUG
1538 	if (ldebug(gettid))
1539 		printf(ARGS(gettid, ""));
1540 #endif
1541 
1542 	td->td_retval[0] = td->td_proc->p_pid;
1543 	return (0);
1544 }
1545 
1546 
1547 int
1548 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1549 {
1550 	struct linux_emuldata *em;
1551 	struct proc *p, *pp;
1552 
1553 #ifdef DEBUG
1554 	if (ldebug(getppid))
1555 		printf(ARGS(getppid, ""));
1556 #endif
1557 
1558 	if (!linux_use26(td)) {
1559 		PROC_LOCK(td->td_proc);
1560 		td->td_retval[0] = td->td_proc->p_pptr->p_pid;
1561 		PROC_UNLOCK(td->td_proc);
1562 		return (0);
1563 	}
1564 
1565 	em = em_find(td->td_proc, EMUL_DONTLOCK);
1566 
1567 	KASSERT(em != NULL, ("getppid: process emuldata not found.\n"));
1568 
1569 	/* find the group leader */
1570 	p = pfind(em->shared->group_pid);
1571 
1572 	if (p == NULL) {
1573 #ifdef DEBUG
1574 	   	printf(LMSG("parent process not found.\n"));
1575 #endif
1576 		return (0);
1577 	}
1578 
1579 	pp = p->p_pptr;		/* switch to parent */
1580 	PROC_LOCK(pp);
1581 	PROC_UNLOCK(p);
1582 
1583 	/* if its also linux process */
1584 	if (pp->p_sysent == &elf_linux_sysvec) {
1585 		em = em_find(pp, EMUL_DONTLOCK);
1586 		KASSERT(em != NULL, ("getppid: parent emuldata not found.\n"));
1587 
1588 		td->td_retval[0] = em->shared->group_pid;
1589 	} else
1590 		td->td_retval[0] = pp->p_pid;
1591 
1592 	PROC_UNLOCK(pp);
1593 
1594 	return (0);
1595 }
1596 
1597 int
1598 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1599 {
1600 
1601 #ifdef DEBUG
1602 	if (ldebug(getgid))
1603 		printf(ARGS(getgid, ""));
1604 #endif
1605 
1606 	td->td_retval[0] = td->td_ucred->cr_rgid;
1607 	return (0);
1608 }
1609 
1610 int
1611 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1612 {
1613 
1614 #ifdef DEBUG
1615 	if (ldebug(getuid))
1616 		printf(ARGS(getuid, ""));
1617 #endif
1618 
1619 	td->td_retval[0] = td->td_ucred->cr_ruid;
1620 	return (0);
1621 }
1622 
1623 
1624 int
1625 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1626 {
1627 	struct getsid_args bsd;
1628 
1629 #ifdef DEBUG
1630 	if (ldebug(getsid))
1631 		printf(ARGS(getsid, "%i"), args->pid);
1632 #endif
1633 
1634 	bsd.pid = args->pid;
1635 	return getsid(td, &bsd);
1636 }
1637 
1638 int
1639 linux_nosys(struct thread *td, struct nosys_args *ignore)
1640 {
1641 
1642 	return (ENOSYS);
1643 }
1644 
1645 int
1646 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1647 {
1648 	struct getpriority_args bsd_args;
1649 	int error;
1650 
1651 #ifdef DEBUG
1652 	if (ldebug(getpriority))
1653 		printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
1654 #endif
1655 
1656 	bsd_args.which = args->which;
1657 	bsd_args.who = args->who;
1658 	error = getpriority(td, &bsd_args);
1659 	td->td_retval[0] = 20 - td->td_retval[0];
1660 	return error;
1661 }
1662 
1663 int
1664 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1665 {
1666 	int name[2];
1667 
1668 #ifdef DEBUG
1669 	if (ldebug(sethostname))
1670 		printf(ARGS(sethostname, "*, %i"), args->len);
1671 #endif
1672 
1673 	name[0] = CTL_KERN;
1674 	name[1] = KERN_HOSTNAME;
1675 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1676 	    args->len, 0, 0));
1677 }
1678 
1679 int
1680 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1681 {
1682 	int name[2];
1683 
1684 #ifdef DEBUG
1685 	if (ldebug(setdomainname))
1686 		printf(ARGS(setdomainname, "*, %i"), args->len);
1687 #endif
1688 
1689 	name[0] = CTL_KERN;
1690 	name[1] = KERN_NISDOMAINNAME;
1691 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1692 	    args->len, 0, 0));
1693 }
1694 
1695 int
1696 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1697 {
1698 	struct linux_emuldata *em, *td_em, *tmp_em;
1699 	struct proc *sp;
1700 
1701 #ifdef DEBUG
1702 	if (ldebug(exit_group))
1703 		printf(ARGS(exit_group, "%i"), args->error_code);
1704 #endif
1705 
1706 	if (linux_use26(td)) {
1707 		td_em = em_find(td->td_proc, EMUL_DONTLOCK);
1708 
1709 		KASSERT(td_em != NULL, ("exit_group: emuldata not found.\n"));
1710 
1711 		EMUL_SHARED_RLOCK(&emul_shared_lock);
1712 		LIST_FOREACH_SAFE(em, &td_em->shared->threads, threads, tmp_em) {
1713 			if (em->pid == td_em->pid)
1714 				continue;
1715 
1716 			sp = pfind(em->pid);
1717 			psignal(sp, SIGKILL);
1718 			PROC_UNLOCK(sp);
1719 #ifdef DEBUG
1720 			printf(LMSG("linux_sys_exit_group: kill PID %d\n"), em->pid);
1721 #endif
1722 		}
1723 
1724 		EMUL_SHARED_RUNLOCK(&emul_shared_lock);
1725 	}
1726 	/*
1727 	 * XXX: we should send a signal to the parent if
1728 	 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1729 	 * as it doesnt occur often.
1730 	 */
1731 	exit1(td, W_EXITCODE(args->error_code, 0));
1732 
1733 	return (0);
1734 }
1735 
1736 int
1737 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1738 {
1739 	int error = 0, max_size;
1740 	struct proc *p = td->td_proc;
1741 	char comm[LINUX_MAX_COMM_LEN];
1742 	struct linux_emuldata *em;
1743 	int pdeath_signal;
1744 
1745 #ifdef DEBUG
1746 	if (ldebug(prctl))
1747 		printf(ARGS(prctl, "%d, %d, %d, %d, %d"), args->option,
1748 		    args->arg2, args->arg3, args->arg4, args->arg5);
1749 #endif
1750 
1751 	switch (args->option) {
1752 	case LINUX_PR_SET_PDEATHSIG:
1753 		if (!LINUX_SIG_VALID(args->arg2))
1754 			return (EINVAL);
1755 		em = em_find(p, EMUL_DOLOCK);
1756 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1757 		em->pdeath_signal = args->arg2;
1758 		EMUL_UNLOCK(&emul_lock);
1759 		break;
1760 	case LINUX_PR_GET_PDEATHSIG:
1761 		em = em_find(p, EMUL_DOLOCK);
1762 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1763 		pdeath_signal = em->pdeath_signal;
1764 		EMUL_UNLOCK(&emul_lock);
1765 		error = copyout(&pdeath_signal,
1766 		    (void *)(register_t)args->arg2,
1767 		    sizeof(pdeath_signal));
1768 		break;
1769 	case LINUX_PR_SET_NAME:
1770 		/*
1771 		 * To be on the safe side we need to make sure to not
1772 		 * overflow the size a linux program expects. We already
1773 		 * do this here in the copyin, so that we don't need to
1774 		 * check on copyout.
1775 		 */
1776 		max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1777 		error = copyinstr((void *)(register_t)args->arg2, comm,
1778 		    max_size, NULL);
1779 
1780 		/* Linux silently truncates the name if it is too long. */
1781 		if (error == ENAMETOOLONG) {
1782 			/*
1783 			 * XXX: copyinstr() isn't documented to populate the
1784 			 * array completely, so do a copyin() to be on the
1785 			 * safe side. This should be changed in case
1786 			 * copyinstr() is changed to guarantee this.
1787 			 */
1788 			error = copyin((void *)(register_t)args->arg2, comm,
1789 			    max_size - 1);
1790 			comm[max_size - 1] = '\0';
1791 		}
1792 		if (error)
1793 			return (error);
1794 
1795 		PROC_LOCK(p);
1796 		strlcpy(p->p_comm, comm, sizeof(p->p_comm));
1797 		PROC_UNLOCK(p);
1798 		break;
1799 	case LINUX_PR_GET_NAME:
1800 		PROC_LOCK(p);
1801 		strlcpy(comm, p->p_comm, sizeof(comm));
1802 		PROC_UNLOCK(p);
1803 		error = copyout(comm, (void *)(register_t)args->arg2,
1804 		    strlen(comm) + 1);
1805 		break;
1806 	default:
1807 		error = EINVAL;
1808 		break;
1809 	}
1810 
1811 	return (error);
1812 }
1813 
1814 /*
1815  * Get affinity of a process.
1816  */
1817 int
1818 linux_sched_getaffinity(struct thread *td,
1819     struct linux_sched_getaffinity_args *args)
1820 {
1821 	int error;
1822 	struct cpuset_getaffinity_args cga;
1823 
1824 #ifdef DEBUG
1825 	if (ldebug(sched_getaffinity))
1826 		printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid,
1827 		    args->len);
1828 #endif
1829 	if (args->len < sizeof(cpuset_t))
1830 		return (EINVAL);
1831 
1832 	cga.level = CPU_LEVEL_WHICH;
1833 	cga.which = CPU_WHICH_PID;
1834 	cga.id = args->pid;
1835 	cga.cpusetsize = sizeof(cpuset_t);
1836 	cga.mask = (cpuset_t *) args->user_mask_ptr;
1837 
1838 	if ((error = cpuset_getaffinity(td, &cga)) == 0)
1839 		td->td_retval[0] = sizeof(cpuset_t);
1840 
1841 	return (error);
1842 }
1843 
1844 /*
1845  *  Set affinity of a process.
1846  */
1847 int
1848 linux_sched_setaffinity(struct thread *td,
1849     struct linux_sched_setaffinity_args *args)
1850 {
1851 	struct cpuset_setaffinity_args csa;
1852 
1853 #ifdef DEBUG
1854 	if (ldebug(sched_setaffinity))
1855 		printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid,
1856 		    args->len);
1857 #endif
1858 	if (args->len < sizeof(cpuset_t))
1859 		return (EINVAL);
1860 
1861 	csa.level = CPU_LEVEL_WHICH;
1862 	csa.which = CPU_WHICH_PID;
1863 	csa.id = args->pid;
1864 	csa.cpusetsize = sizeof(cpuset_t);
1865 	csa.mask = (cpuset_t *) args->user_mask_ptr;
1866 
1867 	return (cpuset_setaffinity(td, &csa));
1868 }
1869