xref: /freebsd/sys/compat/linux/linux_misc.c (revision aa0a1e58f0189b0fde359a8bda032887e72057fa)
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 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
96 	RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
97 	RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
98 	RLIMIT_MEMLOCK, RLIMIT_AS
99 };
100 
101 struct l_sysinfo {
102 	l_long		uptime;		/* Seconds since boot */
103 	l_ulong		loads[3];	/* 1, 5, and 15 minute load averages */
104 #define LINUX_SYSINFO_LOADS_SCALE 65536
105 	l_ulong		totalram;	/* Total usable main memory size */
106 	l_ulong		freeram;	/* Available memory size */
107 	l_ulong		sharedram;	/* Amount of shared memory */
108 	l_ulong		bufferram;	/* Memory used by buffers */
109 	l_ulong		totalswap;	/* Total swap space size */
110 	l_ulong		freeswap;	/* swap space still available */
111 	l_ushort	procs;		/* Number of current processes */
112 	l_ushort	pads;
113 	l_ulong		totalbig;
114 	l_ulong		freebig;
115 	l_uint		mem_unit;
116 	char		_f[20-2*sizeof(l_long)-sizeof(l_int)];	/* padding */
117 };
118 int
119 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
120 {
121 	struct l_sysinfo sysinfo;
122 	vm_object_t object;
123 	int i, j;
124 	struct timespec ts;
125 
126 	getnanouptime(&ts);
127 	if (ts.tv_nsec != 0)
128 		ts.tv_sec++;
129 	sysinfo.uptime = ts.tv_sec;
130 
131 	/* Use the information from the mib to get our load averages */
132 	for (i = 0; i < 3; i++)
133 		sysinfo.loads[i] = averunnable.ldavg[i] *
134 		    LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
135 
136 	sysinfo.totalram = physmem * PAGE_SIZE;
137 	sysinfo.freeram = sysinfo.totalram - cnt.v_wire_count * PAGE_SIZE;
138 
139 	sysinfo.sharedram = 0;
140 	mtx_lock(&vm_object_list_mtx);
141 	TAILQ_FOREACH(object, &vm_object_list, object_list)
142 		if (object->shadow_count > 1)
143 			sysinfo.sharedram += object->resident_page_count;
144 	mtx_unlock(&vm_object_list_mtx);
145 
146 	sysinfo.sharedram *= PAGE_SIZE;
147 	sysinfo.bufferram = 0;
148 
149 	swap_pager_status(&i, &j);
150 	sysinfo.totalswap = i * PAGE_SIZE;
151 	sysinfo.freeswap = (i - j) * PAGE_SIZE;
152 
153 	sysinfo.procs = nprocs;
154 
155 	/* The following are only present in newer Linux kernels. */
156 	sysinfo.totalbig = 0;
157 	sysinfo.freebig = 0;
158 	sysinfo.mem_unit = 1;
159 
160 	return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
161 }
162 
163 int
164 linux_alarm(struct thread *td, struct linux_alarm_args *args)
165 {
166 	struct itimerval it, old_it;
167 	u_int secs;
168 	int error;
169 
170 #ifdef DEBUG
171 	if (ldebug(alarm))
172 		printf(ARGS(alarm, "%u"), args->secs);
173 #endif
174 
175 	secs = args->secs;
176 
177 	if (secs > INT_MAX)
178 		secs = INT_MAX;
179 
180 	it.it_value.tv_sec = (long) secs;
181 	it.it_value.tv_usec = 0;
182 	it.it_interval.tv_sec = 0;
183 	it.it_interval.tv_usec = 0;
184 	error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
185 	if (error)
186 		return (error);
187 	if (timevalisset(&old_it.it_value)) {
188 		if (old_it.it_value.tv_usec != 0)
189 			old_it.it_value.tv_sec++;
190 		td->td_retval[0] = old_it.it_value.tv_sec;
191 	}
192 	return (0);
193 }
194 
195 int
196 linux_brk(struct thread *td, struct linux_brk_args *args)
197 {
198 	struct vmspace *vm = td->td_proc->p_vmspace;
199 	vm_offset_t new, old;
200 	struct obreak_args /* {
201 		char * nsize;
202 	} */ tmp;
203 
204 #ifdef DEBUG
205 	if (ldebug(brk))
206 		printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
207 #endif
208 	old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize);
209 	new = (vm_offset_t)args->dsend;
210 	tmp.nsize = (char *)new;
211 	if (((caddr_t)new > vm->vm_daddr) && !obreak(td, &tmp))
212 		td->td_retval[0] = (long)new;
213 	else
214 		td->td_retval[0] = (long)old;
215 
216 	return (0);
217 }
218 
219 #if defined(__i386__)
220 /* XXX: what about amd64/linux32? */
221 
222 int
223 linux_uselib(struct thread *td, struct linux_uselib_args *args)
224 {
225 	struct nameidata ni;
226 	struct vnode *vp;
227 	struct exec *a_out;
228 	struct vattr attr;
229 	vm_offset_t vmaddr;
230 	unsigned long file_offset;
231 	vm_offset_t buffer;
232 	unsigned long bss_size;
233 	char *library;
234 	int error;
235 	int locked, vfslocked;
236 
237 	LCONVPATHEXIST(td, args->library, &library);
238 
239 #ifdef DEBUG
240 	if (ldebug(uselib))
241 		printf(ARGS(uselib, "%s"), library);
242 #endif
243 
244 	a_out = NULL;
245 	vfslocked = 0;
246 	locked = 0;
247 	vp = NULL;
248 
249 	NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
250 	    UIO_SYSSPACE, library, td);
251 	error = namei(&ni);
252 	LFREEPATH(library);
253 	if (error)
254 		goto cleanup;
255 
256 	vp = ni.ni_vp;
257 	vfslocked = NDHASGIANT(&ni);
258 	NDFREE(&ni, NDF_ONLY_PNBUF);
259 
260 	/*
261 	 * From here on down, we have a locked vnode that must be unlocked.
262 	 * XXX: The code below largely duplicates exec_check_permissions().
263 	 */
264 	locked = 1;
265 
266 	/* Writable? */
267 	if (vp->v_writecount) {
268 		error = ETXTBSY;
269 		goto cleanup;
270 	}
271 
272 	/* Executable? */
273 	error = VOP_GETATTR(vp, &attr, td->td_ucred);
274 	if (error)
275 		goto cleanup;
276 
277 	if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
278 	    ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
279 		/* EACCESS is what exec(2) returns. */
280 		error = ENOEXEC;
281 		goto cleanup;
282 	}
283 
284 	/* Sensible size? */
285 	if (attr.va_size == 0) {
286 		error = ENOEXEC;
287 		goto cleanup;
288 	}
289 
290 	/* Can we access it? */
291 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
292 	if (error)
293 		goto cleanup;
294 
295 	/*
296 	 * XXX: This should use vn_open() so that it is properly authorized,
297 	 * and to reduce code redundancy all over the place here.
298 	 * XXX: Not really, it duplicates far more of exec_check_permissions()
299 	 * than vn_open().
300 	 */
301 #ifdef MAC
302 	error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
303 	if (error)
304 		goto cleanup;
305 #endif
306 	error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
307 	if (error)
308 		goto cleanup;
309 
310 	/* Pull in executable header into kernel_map */
311 	error = vm_mmap(kernel_map, (vm_offset_t *)&a_out, PAGE_SIZE,
312 	    VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
313 	if (error)
314 		goto cleanup;
315 
316 	/* Is it a Linux binary ? */
317 	if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
318 		error = ENOEXEC;
319 		goto cleanup;
320 	}
321 
322 	/*
323 	 * While we are here, we should REALLY do some more checks
324 	 */
325 
326 	/* Set file/virtual offset based on a.out variant. */
327 	switch ((int)(a_out->a_magic & 0xffff)) {
328 	case 0413:			/* ZMAGIC */
329 		file_offset = 1024;
330 		break;
331 	case 0314:			/* QMAGIC */
332 		file_offset = 0;
333 		break;
334 	default:
335 		error = ENOEXEC;
336 		goto cleanup;
337 	}
338 
339 	bss_size = round_page(a_out->a_bss);
340 
341 	/* Check various fields in header for validity/bounds. */
342 	if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
343 		error = ENOEXEC;
344 		goto cleanup;
345 	}
346 
347 	/* text + data can't exceed file size */
348 	if (a_out->a_data + a_out->a_text > attr.va_size) {
349 		error = EFAULT;
350 		goto cleanup;
351 	}
352 
353 	/*
354 	 * text/data/bss must not exceed limits
355 	 * XXX - this is not complete. it should check current usage PLUS
356 	 * the resources needed by this library.
357 	 */
358 	PROC_LOCK(td->td_proc);
359 	if (a_out->a_text > maxtsiz ||
360 	    a_out->a_data + bss_size > lim_cur(td->td_proc, RLIMIT_DATA)) {
361 		PROC_UNLOCK(td->td_proc);
362 		error = ENOMEM;
363 		goto cleanup;
364 	}
365 	PROC_UNLOCK(td->td_proc);
366 
367 	/*
368 	 * Prevent more writers.
369 	 * XXX: Note that if any of the VM operations fail below we don't
370 	 * clear this flag.
371 	 */
372 	vp->v_vflag |= VV_TEXT;
373 
374 	/*
375 	 * Lock no longer needed
376 	 */
377 	locked = 0;
378 	VOP_UNLOCK(vp, 0);
379 	VFS_UNLOCK_GIANT(vfslocked);
380 
381 	/*
382 	 * Check if file_offset page aligned. Currently we cannot handle
383 	 * misalinged file offsets, and so we read in the entire image
384 	 * (what a waste).
385 	 */
386 	if (file_offset & PAGE_MASK) {
387 #ifdef DEBUG
388 		printf("uselib: Non page aligned binary %lu\n", file_offset);
389 #endif
390 		/* Map text+data read/write/execute */
391 
392 		/* a_entry is the load address and is page aligned */
393 		vmaddr = trunc_page(a_out->a_entry);
394 
395 		/* get anon user mapping, read+write+execute */
396 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
397 		    &vmaddr, a_out->a_text + a_out->a_data, FALSE, VM_PROT_ALL,
398 		    VM_PROT_ALL, 0);
399 		if (error)
400 			goto cleanup;
401 
402 		/* map file into kernel_map */
403 		error = vm_mmap(kernel_map, &buffer,
404 		    round_page(a_out->a_text + a_out->a_data + file_offset),
405 		    VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp,
406 		    trunc_page(file_offset));
407 		if (error)
408 			goto cleanup;
409 
410 		/* copy from kernel VM space to user space */
411 		error = copyout(PTRIN(buffer + file_offset),
412 		    (void *)vmaddr, a_out->a_text + a_out->a_data);
413 
414 		/* release temporary kernel space */
415 		vm_map_remove(kernel_map, buffer, buffer +
416 		    round_page(a_out->a_text + a_out->a_data + file_offset));
417 
418 		if (error)
419 			goto cleanup;
420 	} else {
421 #ifdef DEBUG
422 		printf("uselib: Page aligned binary %lu\n", file_offset);
423 #endif
424 		/*
425 		 * for QMAGIC, a_entry is 20 bytes beyond the load address
426 		 * to skip the executable header
427 		 */
428 		vmaddr = trunc_page(a_out->a_entry);
429 
430 		/*
431 		 * Map it all into the process's space as a single
432 		 * copy-on-write "data" segment.
433 		 */
434 		error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr,
435 		    a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
436 		    MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
437 		if (error)
438 			goto cleanup;
439 	}
440 #ifdef DEBUG
441 	printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
442 	    ((long *)vmaddr)[1]);
443 #endif
444 	if (bss_size != 0) {
445 		/* Calculate BSS start address */
446 		vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
447 		    a_out->a_data;
448 
449 		/* allocate some 'anon' space */
450 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
451 		    &vmaddr, bss_size, FALSE, VM_PROT_ALL, VM_PROT_ALL, 0);
452 		if (error)
453 			goto cleanup;
454 	}
455 
456 cleanup:
457 	/* Unlock vnode if needed */
458 	if (locked) {
459 		VOP_UNLOCK(vp, 0);
460 		VFS_UNLOCK_GIANT(vfslocked);
461 	}
462 
463 	/* Release the kernel mapping. */
464 	if (a_out)
465 		vm_map_remove(kernel_map, (vm_offset_t)a_out,
466 		    (vm_offset_t)a_out + PAGE_SIZE);
467 
468 	return (error);
469 }
470 
471 #endif	/* __i386__ */
472 
473 int
474 linux_select(struct thread *td, struct linux_select_args *args)
475 {
476 	l_timeval ltv;
477 	struct timeval tv0, tv1, utv, *tvp;
478 	int error;
479 
480 #ifdef DEBUG
481 	if (ldebug(select))
482 		printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
483 		    (void *)args->readfds, (void *)args->writefds,
484 		    (void *)args->exceptfds, (void *)args->timeout);
485 #endif
486 
487 	/*
488 	 * Store current time for computation of the amount of
489 	 * time left.
490 	 */
491 	if (args->timeout) {
492 		if ((error = copyin(args->timeout, &ltv, sizeof(ltv))))
493 			goto select_out;
494 		utv.tv_sec = ltv.tv_sec;
495 		utv.tv_usec = ltv.tv_usec;
496 #ifdef DEBUG
497 		if (ldebug(select))
498 			printf(LMSG("incoming timeout (%jd/%ld)"),
499 			    (intmax_t)utv.tv_sec, utv.tv_usec);
500 #endif
501 
502 		if (itimerfix(&utv)) {
503 			/*
504 			 * The timeval was invalid.  Convert it to something
505 			 * valid that will act as it does under Linux.
506 			 */
507 			utv.tv_sec += utv.tv_usec / 1000000;
508 			utv.tv_usec %= 1000000;
509 			if (utv.tv_usec < 0) {
510 				utv.tv_sec -= 1;
511 				utv.tv_usec += 1000000;
512 			}
513 			if (utv.tv_sec < 0)
514 				timevalclear(&utv);
515 		}
516 		microtime(&tv0);
517 		tvp = &utv;
518 	} else
519 		tvp = NULL;
520 
521 	error = kern_select(td, args->nfds, args->readfds, args->writefds,
522 	    args->exceptfds, tvp, sizeof(l_int) * 8);
523 
524 #ifdef DEBUG
525 	if (ldebug(select))
526 		printf(LMSG("real select returns %d"), error);
527 #endif
528 	if (error)
529 		goto select_out;
530 
531 	if (args->timeout) {
532 		if (td->td_retval[0]) {
533 			/*
534 			 * Compute how much time was left of the timeout,
535 			 * by subtracting the current time and the time
536 			 * before we started the call, and subtracting
537 			 * that result from the user-supplied value.
538 			 */
539 			microtime(&tv1);
540 			timevalsub(&tv1, &tv0);
541 			timevalsub(&utv, &tv1);
542 			if (utv.tv_sec < 0)
543 				timevalclear(&utv);
544 		} else
545 			timevalclear(&utv);
546 #ifdef DEBUG
547 		if (ldebug(select))
548 			printf(LMSG("outgoing timeout (%jd/%ld)"),
549 			    (intmax_t)utv.tv_sec, utv.tv_usec);
550 #endif
551 		ltv.tv_sec = utv.tv_sec;
552 		ltv.tv_usec = utv.tv_usec;
553 		if ((error = copyout(&ltv, args->timeout, sizeof(ltv))))
554 			goto select_out;
555 	}
556 
557 select_out:
558 #ifdef DEBUG
559 	if (ldebug(select))
560 		printf(LMSG("select_out -> %d"), error);
561 #endif
562 	return (error);
563 }
564 
565 int
566 linux_mremap(struct thread *td, struct linux_mremap_args *args)
567 {
568 	struct munmap_args /* {
569 		void *addr;
570 		size_t len;
571 	} */ bsd_args;
572 	int error = 0;
573 
574 #ifdef DEBUG
575 	if (ldebug(mremap))
576 		printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
577 		    (void *)(uintptr_t)args->addr,
578 		    (unsigned long)args->old_len,
579 		    (unsigned long)args->new_len,
580 		    (unsigned long)args->flags);
581 #endif
582 
583 	if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
584 		td->td_retval[0] = 0;
585 		return (EINVAL);
586 	}
587 
588 	/*
589 	 * Check for the page alignment.
590 	 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
591 	 */
592 	if (args->addr & PAGE_MASK) {
593 		td->td_retval[0] = 0;
594 		return (EINVAL);
595 	}
596 
597 	args->new_len = round_page(args->new_len);
598 	args->old_len = round_page(args->old_len);
599 
600 	if (args->new_len > args->old_len) {
601 		td->td_retval[0] = 0;
602 		return (ENOMEM);
603 	}
604 
605 	if (args->new_len < args->old_len) {
606 		bsd_args.addr =
607 		    (caddr_t)((uintptr_t)args->addr + args->new_len);
608 		bsd_args.len = args->old_len - args->new_len;
609 		error = munmap(td, &bsd_args);
610 	}
611 
612 	td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
613 	return (error);
614 }
615 
616 #define LINUX_MS_ASYNC       0x0001
617 #define LINUX_MS_INVALIDATE  0x0002
618 #define LINUX_MS_SYNC        0x0004
619 
620 int
621 linux_msync(struct thread *td, struct linux_msync_args *args)
622 {
623 	struct msync_args bsd_args;
624 
625 	bsd_args.addr = (caddr_t)(uintptr_t)args->addr;
626 	bsd_args.len = (uintptr_t)args->len;
627 	bsd_args.flags = args->fl & ~LINUX_MS_SYNC;
628 
629 	return (msync(td, &bsd_args));
630 }
631 
632 int
633 linux_time(struct thread *td, struct linux_time_args *args)
634 {
635 	struct timeval tv;
636 	l_time_t tm;
637 	int error;
638 
639 #ifdef DEBUG
640 	if (ldebug(time))
641 		printf(ARGS(time, "*"));
642 #endif
643 
644 	microtime(&tv);
645 	tm = tv.tv_sec;
646 	if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
647 		return (error);
648 	td->td_retval[0] = tm;
649 	return (0);
650 }
651 
652 struct l_times_argv {
653 	l_clock_t	tms_utime;
654 	l_clock_t	tms_stime;
655 	l_clock_t	tms_cutime;
656 	l_clock_t	tms_cstime;
657 };
658 
659 
660 /*
661  * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
662  * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
663  * auxiliary vector entry.
664  */
665 #define	CLK_TCK		100
666 
667 #define	CONVOTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
668 #define	CONVNTCK(r)	(r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
669 
670 #define	CONVTCK(r)	(linux_kernver(td) >= LINUX_KERNVER_2004000 ?		\
671 			    CONVNTCK(r) : CONVOTCK(r))
672 
673 int
674 linux_times(struct thread *td, struct linux_times_args *args)
675 {
676 	struct timeval tv, utime, stime, cutime, cstime;
677 	struct l_times_argv tms;
678 	struct proc *p;
679 	int error;
680 
681 #ifdef DEBUG
682 	if (ldebug(times))
683 		printf(ARGS(times, "*"));
684 #endif
685 
686 	if (args->buf != NULL) {
687 		p = td->td_proc;
688 		PROC_LOCK(p);
689 		PROC_SLOCK(p);
690 		calcru(p, &utime, &stime);
691 		PROC_SUNLOCK(p);
692 		calccru(p, &cutime, &cstime);
693 		PROC_UNLOCK(p);
694 
695 		tms.tms_utime = CONVTCK(utime);
696 		tms.tms_stime = CONVTCK(stime);
697 
698 		tms.tms_cutime = CONVTCK(cutime);
699 		tms.tms_cstime = CONVTCK(cstime);
700 
701 		if ((error = copyout(&tms, args->buf, sizeof(tms))))
702 			return (error);
703 	}
704 
705 	microuptime(&tv);
706 	td->td_retval[0] = (int)CONVTCK(tv);
707 	return (0);
708 }
709 
710 int
711 linux_newuname(struct thread *td, struct linux_newuname_args *args)
712 {
713 	struct l_new_utsname utsname;
714 	char osname[LINUX_MAX_UTSNAME];
715 	char osrelease[LINUX_MAX_UTSNAME];
716 	char *p;
717 
718 #ifdef DEBUG
719 	if (ldebug(newuname))
720 		printf(ARGS(newuname, "*"));
721 #endif
722 
723 	linux_get_osname(td, osname);
724 	linux_get_osrelease(td, osrelease);
725 
726 	bzero(&utsname, sizeof(utsname));
727 	strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
728 	getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
729 	getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
730 	strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
731 	strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
732 	for (p = utsname.version; *p != '\0'; ++p)
733 		if (*p == '\n') {
734 			*p = '\0';
735 			break;
736 		}
737 	strlcpy(utsname.machine, linux_platform, LINUX_MAX_UTSNAME);
738 
739 	return (copyout(&utsname, args->buf, sizeof(utsname)));
740 }
741 
742 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
743 struct l_utimbuf {
744 	l_time_t l_actime;
745 	l_time_t l_modtime;
746 };
747 
748 int
749 linux_utime(struct thread *td, struct linux_utime_args *args)
750 {
751 	struct timeval tv[2], *tvp;
752 	struct l_utimbuf lut;
753 	char *fname;
754 	int error;
755 
756 	LCONVPATHEXIST(td, args->fname, &fname);
757 
758 #ifdef DEBUG
759 	if (ldebug(utime))
760 		printf(ARGS(utime, "%s, *"), fname);
761 #endif
762 
763 	if (args->times) {
764 		if ((error = copyin(args->times, &lut, sizeof lut))) {
765 			LFREEPATH(fname);
766 			return (error);
767 		}
768 		tv[0].tv_sec = lut.l_actime;
769 		tv[0].tv_usec = 0;
770 		tv[1].tv_sec = lut.l_modtime;
771 		tv[1].tv_usec = 0;
772 		tvp = tv;
773 	} else
774 		tvp = NULL;
775 
776 	error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
777 	LFREEPATH(fname);
778 	return (error);
779 }
780 
781 int
782 linux_utimes(struct thread *td, struct linux_utimes_args *args)
783 {
784 	l_timeval ltv[2];
785 	struct timeval tv[2], *tvp = NULL;
786 	char *fname;
787 	int error;
788 
789 	LCONVPATHEXIST(td, args->fname, &fname);
790 
791 #ifdef DEBUG
792 	if (ldebug(utimes))
793 		printf(ARGS(utimes, "%s, *"), fname);
794 #endif
795 
796 	if (args->tptr != NULL) {
797 		if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
798 			LFREEPATH(fname);
799 			return (error);
800 		}
801 		tv[0].tv_sec = ltv[0].tv_sec;
802 		tv[0].tv_usec = ltv[0].tv_usec;
803 		tv[1].tv_sec = ltv[1].tv_sec;
804 		tv[1].tv_usec = ltv[1].tv_usec;
805 		tvp = tv;
806 	}
807 
808 	error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
809 	LFREEPATH(fname);
810 	return (error);
811 }
812 
813 int
814 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
815 {
816 	l_timeval ltv[2];
817 	struct timeval tv[2], *tvp = NULL;
818 	char *fname;
819 	int error, dfd;
820 
821 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
822 	LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
823 
824 #ifdef DEBUG
825 	if (ldebug(futimesat))
826 		printf(ARGS(futimesat, "%s, *"), fname);
827 #endif
828 
829 	if (args->utimes != NULL) {
830 		if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
831 			LFREEPATH(fname);
832 			return (error);
833 		}
834 		tv[0].tv_sec = ltv[0].tv_sec;
835 		tv[0].tv_usec = ltv[0].tv_usec;
836 		tv[1].tv_sec = ltv[1].tv_sec;
837 		tv[1].tv_usec = ltv[1].tv_usec;
838 		tvp = tv;
839 	}
840 
841 	error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
842 	LFREEPATH(fname);
843 	return (error);
844 }
845 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
846 
847 int
848 linux_common_wait(struct thread *td, int pid, int *status,
849     int options, struct rusage *ru)
850 {
851 	int error, tmpstat;
852 
853 	error = kern_wait(td, pid, &tmpstat, options, ru);
854 	if (error)
855 		return (error);
856 
857 	if (status) {
858 		tmpstat &= 0xffff;
859 		if (WIFSIGNALED(tmpstat))
860 			tmpstat = (tmpstat & 0xffffff80) |
861 			    BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat));
862 		else if (WIFSTOPPED(tmpstat))
863 			tmpstat = (tmpstat & 0xffff00ff) |
864 			    (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8);
865 		error = copyout(&tmpstat, status, sizeof(int));
866 	}
867 
868 	return (error);
869 }
870 
871 int
872 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
873 {
874 	int options;
875 
876 #ifdef DEBUG
877 	if (ldebug(waitpid))
878 		printf(ARGS(waitpid, "%d, %p, %d"),
879 		    args->pid, (void *)args->status, args->options);
880 #endif
881 	/*
882 	 * this is necessary because the test in kern_wait doesn't work
883 	 * because we mess with the options here
884 	 */
885 	if (args->options & ~(WUNTRACED | WNOHANG | WCONTINUED | __WCLONE))
886 		return (EINVAL);
887 
888 	options = (args->options & (WNOHANG | WUNTRACED));
889 	/* WLINUXCLONE should be equal to __WCLONE, but we make sure */
890 	if (args->options & __WCLONE)
891 		options |= WLINUXCLONE;
892 
893 	return (linux_common_wait(td, args->pid, args->status, options, NULL));
894 }
895 
896 
897 int
898 linux_mknod(struct thread *td, struct linux_mknod_args *args)
899 {
900 	char *path;
901 	int error;
902 
903 	LCONVPATHCREAT(td, args->path, &path);
904 
905 #ifdef DEBUG
906 	if (ldebug(mknod))
907 		printf(ARGS(mknod, "%s, %d, %d"), path, args->mode, args->dev);
908 #endif
909 
910 	switch (args->mode & S_IFMT) {
911 	case S_IFIFO:
912 	case S_IFSOCK:
913 		error = kern_mkfifo(td, path, UIO_SYSSPACE, args->mode);
914 		break;
915 
916 	case S_IFCHR:
917 	case S_IFBLK:
918 		error = kern_mknod(td, path, UIO_SYSSPACE, args->mode,
919 		    args->dev);
920 		break;
921 
922 	case S_IFDIR:
923 		error = EPERM;
924 		break;
925 
926 	case 0:
927 		args->mode |= S_IFREG;
928 		/* FALLTHROUGH */
929 	case S_IFREG:
930 		error = kern_open(td, path, UIO_SYSSPACE,
931 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
932 		if (error == 0)
933 			kern_close(td, td->td_retval[0]);
934 		break;
935 
936 	default:
937 		error = EINVAL;
938 		break;
939 	}
940 	LFREEPATH(path);
941 	return (error);
942 }
943 
944 int
945 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
946 {
947 	char *path;
948 	int error, dfd;
949 
950 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
951 	LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
952 
953 #ifdef DEBUG
954 	if (ldebug(mknodat))
955 		printf(ARGS(mknodat, "%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_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
962 		break;
963 
964 	case S_IFCHR:
965 	case S_IFBLK:
966 		error = kern_mknodat(td, dfd, 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_openat(td, dfd, 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 /*
993  * UGH! This is just about the dumbest idea I've ever heard!!
994  */
995 int
996 linux_personality(struct thread *td, struct linux_personality_args *args)
997 {
998 #ifdef DEBUG
999 	if (ldebug(personality))
1000 		printf(ARGS(personality, "%lu"), (unsigned long)args->per);
1001 #endif
1002 	if (args->per != 0)
1003 		return (EINVAL);
1004 
1005 	/* Yes Jim, it's still a Linux... */
1006 	td->td_retval[0] = 0;
1007 	return (0);
1008 }
1009 
1010 struct l_itimerval {
1011 	l_timeval it_interval;
1012 	l_timeval it_value;
1013 };
1014 
1015 #define	B2L_ITIMERVAL(bip, lip) 					\
1016 	(bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec;		\
1017 	(bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec;	\
1018 	(bip)->it_value.tv_sec = (lip)->it_value.tv_sec;		\
1019 	(bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1020 
1021 int
1022 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1023 {
1024 	int error;
1025 	struct l_itimerval ls;
1026 	struct itimerval aitv, oitv;
1027 
1028 #ifdef DEBUG
1029 	if (ldebug(setitimer))
1030 		printf(ARGS(setitimer, "%p, %p"),
1031 		    (void *)uap->itv, (void *)uap->oitv);
1032 #endif
1033 
1034 	if (uap->itv == NULL) {
1035 		uap->itv = uap->oitv;
1036 		return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1037 	}
1038 
1039 	error = copyin(uap->itv, &ls, sizeof(ls));
1040 	if (error != 0)
1041 		return (error);
1042 	B2L_ITIMERVAL(&aitv, &ls);
1043 #ifdef DEBUG
1044 	if (ldebug(setitimer)) {
1045 		printf("setitimer: value: sec: %jd, usec: %ld\n",
1046 		    (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1047 		printf("setitimer: interval: sec: %jd, usec: %ld\n",
1048 		    (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1049 	}
1050 #endif
1051 	error = kern_setitimer(td, uap->which, &aitv, &oitv);
1052 	if (error != 0 || uap->oitv == NULL)
1053 		return (error);
1054 	B2L_ITIMERVAL(&ls, &oitv);
1055 
1056 	return (copyout(&ls, uap->oitv, sizeof(ls)));
1057 }
1058 
1059 int
1060 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1061 {
1062 	int error;
1063 	struct l_itimerval ls;
1064 	struct itimerval aitv;
1065 
1066 #ifdef DEBUG
1067 	if (ldebug(getitimer))
1068 		printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1069 #endif
1070 	error = kern_getitimer(td, uap->which, &aitv);
1071 	if (error != 0)
1072 		return (error);
1073 	B2L_ITIMERVAL(&ls, &aitv);
1074 	return (copyout(&ls, uap->itv, sizeof(ls)));
1075 }
1076 
1077 int
1078 linux_nice(struct thread *td, struct linux_nice_args *args)
1079 {
1080 	struct setpriority_args bsd_args;
1081 
1082 	bsd_args.which = PRIO_PROCESS;
1083 	bsd_args.who = 0;		/* current process */
1084 	bsd_args.prio = args->inc;
1085 	return (setpriority(td, &bsd_args));
1086 }
1087 
1088 int
1089 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1090 {
1091 	struct ucred *newcred, *oldcred;
1092 	l_gid_t *linux_gidset;
1093 	gid_t *bsd_gidset;
1094 	int ngrp, error;
1095 	struct proc *p;
1096 
1097 	ngrp = args->gidsetsize;
1098 	if (ngrp < 0 || ngrp >= ngroups_max + 1)
1099 		return (EINVAL);
1100 	linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_TEMP, M_WAITOK);
1101 	error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1102 	if (error)
1103 		goto out;
1104 	newcred = crget();
1105 	p = td->td_proc;
1106 	PROC_LOCK(p);
1107 	oldcred = crcopysafe(p, newcred);
1108 
1109 	/*
1110 	 * cr_groups[0] holds egid. Setting the whole set from
1111 	 * the supplied set will cause egid to be changed too.
1112 	 * Keep cr_groups[0] unchanged to prevent that.
1113 	 */
1114 
1115 	if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) {
1116 		PROC_UNLOCK(p);
1117 		crfree(newcred);
1118 		goto out;
1119 	}
1120 
1121 	if (ngrp > 0) {
1122 		newcred->cr_ngroups = ngrp + 1;
1123 
1124 		bsd_gidset = newcred->cr_groups;
1125 		ngrp--;
1126 		while (ngrp >= 0) {
1127 			bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1128 			ngrp--;
1129 		}
1130 	} else
1131 		newcred->cr_ngroups = 1;
1132 
1133 	setsugid(p);
1134 	p->p_ucred = newcred;
1135 	PROC_UNLOCK(p);
1136 	crfree(oldcred);
1137 	error = 0;
1138 out:
1139 	free(linux_gidset, M_TEMP);
1140 	return (error);
1141 }
1142 
1143 int
1144 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1145 {
1146 	struct ucred *cred;
1147 	l_gid_t *linux_gidset;
1148 	gid_t *bsd_gidset;
1149 	int bsd_gidsetsz, ngrp, error;
1150 
1151 	cred = td->td_ucred;
1152 	bsd_gidset = cred->cr_groups;
1153 	bsd_gidsetsz = cred->cr_ngroups - 1;
1154 
1155 	/*
1156 	 * cr_groups[0] holds egid. Returning the whole set
1157 	 * here will cause a duplicate. Exclude cr_groups[0]
1158 	 * to prevent that.
1159 	 */
1160 
1161 	if ((ngrp = args->gidsetsize) == 0) {
1162 		td->td_retval[0] = bsd_gidsetsz;
1163 		return (0);
1164 	}
1165 
1166 	if (ngrp < bsd_gidsetsz)
1167 		return (EINVAL);
1168 
1169 	ngrp = 0;
1170 	linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1171 	    M_TEMP, M_WAITOK);
1172 	while (ngrp < bsd_gidsetsz) {
1173 		linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1174 		ngrp++;
1175 	}
1176 
1177 	error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1178 	free(linux_gidset, M_TEMP);
1179 	if (error)
1180 		return (error);
1181 
1182 	td->td_retval[0] = ngrp;
1183 	return (0);
1184 }
1185 
1186 int
1187 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1188 {
1189 	struct rlimit bsd_rlim;
1190 	struct l_rlimit rlim;
1191 	u_int which;
1192 	int error;
1193 
1194 #ifdef DEBUG
1195 	if (ldebug(setrlimit))
1196 		printf(ARGS(setrlimit, "%d, %p"),
1197 		    args->resource, (void *)args->rlim);
1198 #endif
1199 
1200 	if (args->resource >= LINUX_RLIM_NLIMITS)
1201 		return (EINVAL);
1202 
1203 	which = linux_to_bsd_resource[args->resource];
1204 	if (which == -1)
1205 		return (EINVAL);
1206 
1207 	error = copyin(args->rlim, &rlim, sizeof(rlim));
1208 	if (error)
1209 		return (error);
1210 
1211 	bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1212 	bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1213 	return (kern_setrlimit(td, which, &bsd_rlim));
1214 }
1215 
1216 int
1217 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1218 {
1219 	struct l_rlimit rlim;
1220 	struct proc *p = td->td_proc;
1221 	struct rlimit bsd_rlim;
1222 	u_int which;
1223 
1224 #ifdef DEBUG
1225 	if (ldebug(old_getrlimit))
1226 		printf(ARGS(old_getrlimit, "%d, %p"),
1227 		    args->resource, (void *)args->rlim);
1228 #endif
1229 
1230 	if (args->resource >= LINUX_RLIM_NLIMITS)
1231 		return (EINVAL);
1232 
1233 	which = linux_to_bsd_resource[args->resource];
1234 	if (which == -1)
1235 		return (EINVAL);
1236 
1237 	PROC_LOCK(p);
1238 	lim_rlimit(p, which, &bsd_rlim);
1239 	PROC_UNLOCK(p);
1240 
1241 #ifdef COMPAT_LINUX32
1242 	rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1243 	if (rlim.rlim_cur == UINT_MAX)
1244 		rlim.rlim_cur = INT_MAX;
1245 	rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1246 	if (rlim.rlim_max == UINT_MAX)
1247 		rlim.rlim_max = INT_MAX;
1248 #else
1249 	rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1250 	if (rlim.rlim_cur == ULONG_MAX)
1251 		rlim.rlim_cur = LONG_MAX;
1252 	rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1253 	if (rlim.rlim_max == ULONG_MAX)
1254 		rlim.rlim_max = LONG_MAX;
1255 #endif
1256 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1257 }
1258 
1259 int
1260 linux_getrlimit(struct thread *td, struct linux_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(getrlimit))
1269 		printf(ARGS(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 	rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1285 	rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1286 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1287 }
1288 
1289 int
1290 linux_sched_setscheduler(struct thread *td,
1291     struct linux_sched_setscheduler_args *args)
1292 {
1293 	struct sched_setscheduler_args bsd;
1294 
1295 #ifdef DEBUG
1296 	if (ldebug(sched_setscheduler))
1297 		printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1298 		    args->pid, args->policy, (const void *)args->param);
1299 #endif
1300 
1301 	switch (args->policy) {
1302 	case LINUX_SCHED_OTHER:
1303 		bsd.policy = SCHED_OTHER;
1304 		break;
1305 	case LINUX_SCHED_FIFO:
1306 		bsd.policy = SCHED_FIFO;
1307 		break;
1308 	case LINUX_SCHED_RR:
1309 		bsd.policy = SCHED_RR;
1310 		break;
1311 	default:
1312 		return (EINVAL);
1313 	}
1314 
1315 	bsd.pid = args->pid;
1316 	bsd.param = (struct sched_param *)args->param;
1317 	return (sched_setscheduler(td, &bsd));
1318 }
1319 
1320 int
1321 linux_sched_getscheduler(struct thread *td,
1322     struct linux_sched_getscheduler_args *args)
1323 {
1324 	struct sched_getscheduler_args bsd;
1325 	int error;
1326 
1327 #ifdef DEBUG
1328 	if (ldebug(sched_getscheduler))
1329 		printf(ARGS(sched_getscheduler, "%d"), args->pid);
1330 #endif
1331 
1332 	bsd.pid = args->pid;
1333 	error = sched_getscheduler(td, &bsd);
1334 
1335 	switch (td->td_retval[0]) {
1336 	case SCHED_OTHER:
1337 		td->td_retval[0] = LINUX_SCHED_OTHER;
1338 		break;
1339 	case SCHED_FIFO:
1340 		td->td_retval[0] = LINUX_SCHED_FIFO;
1341 		break;
1342 	case SCHED_RR:
1343 		td->td_retval[0] = LINUX_SCHED_RR;
1344 		break;
1345 	}
1346 
1347 	return (error);
1348 }
1349 
1350 int
1351 linux_sched_get_priority_max(struct thread *td,
1352     struct linux_sched_get_priority_max_args *args)
1353 {
1354 	struct sched_get_priority_max_args bsd;
1355 
1356 #ifdef DEBUG
1357 	if (ldebug(sched_get_priority_max))
1358 		printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1359 #endif
1360 
1361 	switch (args->policy) {
1362 	case LINUX_SCHED_OTHER:
1363 		bsd.policy = SCHED_OTHER;
1364 		break;
1365 	case LINUX_SCHED_FIFO:
1366 		bsd.policy = SCHED_FIFO;
1367 		break;
1368 	case LINUX_SCHED_RR:
1369 		bsd.policy = SCHED_RR;
1370 		break;
1371 	default:
1372 		return (EINVAL);
1373 	}
1374 	return (sched_get_priority_max(td, &bsd));
1375 }
1376 
1377 int
1378 linux_sched_get_priority_min(struct thread *td,
1379     struct linux_sched_get_priority_min_args *args)
1380 {
1381 	struct sched_get_priority_min_args bsd;
1382 
1383 #ifdef DEBUG
1384 	if (ldebug(sched_get_priority_min))
1385 		printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1386 #endif
1387 
1388 	switch (args->policy) {
1389 	case LINUX_SCHED_OTHER:
1390 		bsd.policy = SCHED_OTHER;
1391 		break;
1392 	case LINUX_SCHED_FIFO:
1393 		bsd.policy = SCHED_FIFO;
1394 		break;
1395 	case LINUX_SCHED_RR:
1396 		bsd.policy = SCHED_RR;
1397 		break;
1398 	default:
1399 		return (EINVAL);
1400 	}
1401 	return (sched_get_priority_min(td, &bsd));
1402 }
1403 
1404 #define REBOOT_CAD_ON	0x89abcdef
1405 #define REBOOT_CAD_OFF	0
1406 #define REBOOT_HALT	0xcdef0123
1407 #define REBOOT_RESTART	0x01234567
1408 #define REBOOT_RESTART2	0xA1B2C3D4
1409 #define REBOOT_POWEROFF	0x4321FEDC
1410 #define REBOOT_MAGIC1	0xfee1dead
1411 #define REBOOT_MAGIC2	0x28121969
1412 #define REBOOT_MAGIC2A	0x05121996
1413 #define REBOOT_MAGIC2B	0x16041998
1414 
1415 int
1416 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1417 {
1418 	struct reboot_args bsd_args;
1419 
1420 #ifdef DEBUG
1421 	if (ldebug(reboot))
1422 		printf(ARGS(reboot, "0x%x"), args->cmd);
1423 #endif
1424 
1425 	if (args->magic1 != REBOOT_MAGIC1)
1426 		return (EINVAL);
1427 
1428 	switch (args->magic2) {
1429 	case REBOOT_MAGIC2:
1430 	case REBOOT_MAGIC2A:
1431 	case REBOOT_MAGIC2B:
1432 		break;
1433 	default:
1434 		return (EINVAL);
1435 	}
1436 
1437 	switch (args->cmd) {
1438 	case REBOOT_CAD_ON:
1439 	case REBOOT_CAD_OFF:
1440 		return (priv_check(td, PRIV_REBOOT));
1441 	case REBOOT_HALT:
1442 		bsd_args.opt = RB_HALT;
1443 		break;
1444 	case REBOOT_RESTART:
1445 	case REBOOT_RESTART2:
1446 		bsd_args.opt = 0;
1447 		break;
1448 	case REBOOT_POWEROFF:
1449 		bsd_args.opt = RB_POWEROFF;
1450 		break;
1451 	default:
1452 		return (EINVAL);
1453 	}
1454 	return (reboot(td, &bsd_args));
1455 }
1456 
1457 
1458 /*
1459  * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1460  * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
1461  * are assumed to be preserved. The following lightweight syscalls fixes
1462  * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
1463  *
1464  * linux_getpid() - MP SAFE
1465  * linux_getgid() - MP SAFE
1466  * linux_getuid() - MP SAFE
1467  */
1468 
1469 int
1470 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1471 {
1472 	struct linux_emuldata *em;
1473 
1474 #ifdef DEBUG
1475 	if (ldebug(getpid))
1476 		printf(ARGS(getpid, ""));
1477 #endif
1478 
1479 	if (linux_use26(td)) {
1480 		em = em_find(td->td_proc, EMUL_DONTLOCK);
1481 		KASSERT(em != NULL, ("getpid: emuldata not found.\n"));
1482 		td->td_retval[0] = em->shared->group_pid;
1483 	} else {
1484 		td->td_retval[0] = td->td_proc->p_pid;
1485 	}
1486 
1487 	return (0);
1488 }
1489 
1490 int
1491 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1492 {
1493 
1494 #ifdef DEBUG
1495 	if (ldebug(gettid))
1496 		printf(ARGS(gettid, ""));
1497 #endif
1498 
1499 	td->td_retval[0] = td->td_proc->p_pid;
1500 	return (0);
1501 }
1502 
1503 
1504 int
1505 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1506 {
1507 	struct linux_emuldata *em;
1508 	struct proc *p, *pp;
1509 
1510 #ifdef DEBUG
1511 	if (ldebug(getppid))
1512 		printf(ARGS(getppid, ""));
1513 #endif
1514 
1515 	if (!linux_use26(td)) {
1516 		PROC_LOCK(td->td_proc);
1517 		td->td_retval[0] = td->td_proc->p_pptr->p_pid;
1518 		PROC_UNLOCK(td->td_proc);
1519 		return (0);
1520 	}
1521 
1522 	em = em_find(td->td_proc, EMUL_DONTLOCK);
1523 
1524 	KASSERT(em != NULL, ("getppid: process emuldata not found.\n"));
1525 
1526 	/* find the group leader */
1527 	p = pfind(em->shared->group_pid);
1528 
1529 	if (p == NULL) {
1530 #ifdef DEBUG
1531 	   	printf(LMSG("parent process not found.\n"));
1532 #endif
1533 		return (0);
1534 	}
1535 
1536 	pp = p->p_pptr;		/* switch to parent */
1537 	PROC_LOCK(pp);
1538 	PROC_UNLOCK(p);
1539 
1540 	/* if its also linux process */
1541 	if (pp->p_sysent == &elf_linux_sysvec) {
1542 		em = em_find(pp, EMUL_DONTLOCK);
1543 		KASSERT(em != NULL, ("getppid: parent emuldata not found.\n"));
1544 
1545 		td->td_retval[0] = em->shared->group_pid;
1546 	} else
1547 		td->td_retval[0] = pp->p_pid;
1548 
1549 	PROC_UNLOCK(pp);
1550 
1551 	return (0);
1552 }
1553 
1554 int
1555 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1556 {
1557 
1558 #ifdef DEBUG
1559 	if (ldebug(getgid))
1560 		printf(ARGS(getgid, ""));
1561 #endif
1562 
1563 	td->td_retval[0] = td->td_ucred->cr_rgid;
1564 	return (0);
1565 }
1566 
1567 int
1568 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1569 {
1570 
1571 #ifdef DEBUG
1572 	if (ldebug(getuid))
1573 		printf(ARGS(getuid, ""));
1574 #endif
1575 
1576 	td->td_retval[0] = td->td_ucred->cr_ruid;
1577 	return (0);
1578 }
1579 
1580 
1581 int
1582 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1583 {
1584 	struct getsid_args bsd;
1585 
1586 #ifdef DEBUG
1587 	if (ldebug(getsid))
1588 		printf(ARGS(getsid, "%i"), args->pid);
1589 #endif
1590 
1591 	bsd.pid = args->pid;
1592 	return (getsid(td, &bsd));
1593 }
1594 
1595 int
1596 linux_nosys(struct thread *td, struct nosys_args *ignore)
1597 {
1598 
1599 	return (ENOSYS);
1600 }
1601 
1602 int
1603 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1604 {
1605 	struct getpriority_args bsd_args;
1606 	int error;
1607 
1608 #ifdef DEBUG
1609 	if (ldebug(getpriority))
1610 		printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
1611 #endif
1612 
1613 	bsd_args.which = args->which;
1614 	bsd_args.who = args->who;
1615 	error = getpriority(td, &bsd_args);
1616 	td->td_retval[0] = 20 - td->td_retval[0];
1617 	return (error);
1618 }
1619 
1620 int
1621 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1622 {
1623 	int name[2];
1624 
1625 #ifdef DEBUG
1626 	if (ldebug(sethostname))
1627 		printf(ARGS(sethostname, "*, %i"), args->len);
1628 #endif
1629 
1630 	name[0] = CTL_KERN;
1631 	name[1] = KERN_HOSTNAME;
1632 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1633 	    args->len, 0, 0));
1634 }
1635 
1636 int
1637 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1638 {
1639 	int name[2];
1640 
1641 #ifdef DEBUG
1642 	if (ldebug(setdomainname))
1643 		printf(ARGS(setdomainname, "*, %i"), args->len);
1644 #endif
1645 
1646 	name[0] = CTL_KERN;
1647 	name[1] = KERN_NISDOMAINNAME;
1648 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1649 	    args->len, 0, 0));
1650 }
1651 
1652 int
1653 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1654 {
1655 	struct linux_emuldata *em;
1656 
1657 #ifdef DEBUG
1658 	if (ldebug(exit_group))
1659 		printf(ARGS(exit_group, "%i"), args->error_code);
1660 #endif
1661 
1662 	em = em_find(td->td_proc, EMUL_DONTLOCK);
1663 	if (em->shared->refs > 1) {
1664 		EMUL_SHARED_WLOCK(&emul_shared_lock);
1665 		em->shared->flags |= EMUL_SHARED_HASXSTAT;
1666 		em->shared->xstat = W_EXITCODE(args->error_code, 0);
1667 		EMUL_SHARED_WUNLOCK(&emul_shared_lock);
1668 		if (linux_use26(td))
1669 			linux_kill_threads(td, SIGKILL);
1670 	}
1671 
1672 	/*
1673 	 * XXX: we should send a signal to the parent if
1674 	 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1675 	 * as it doesnt occur often.
1676 	 */
1677 	exit1(td, W_EXITCODE(args->error_code, 0));
1678 
1679 	return (0);
1680 }
1681 
1682 int
1683 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1684 {
1685 	int error = 0, max_size;
1686 	struct proc *p = td->td_proc;
1687 	char comm[LINUX_MAX_COMM_LEN];
1688 	struct linux_emuldata *em;
1689 	int pdeath_signal;
1690 
1691 #ifdef DEBUG
1692 	if (ldebug(prctl))
1693 		printf(ARGS(prctl, "%d, %d, %d, %d, %d"), args->option,
1694 		    args->arg2, args->arg3, args->arg4, args->arg5);
1695 #endif
1696 
1697 	switch (args->option) {
1698 	case LINUX_PR_SET_PDEATHSIG:
1699 		if (!LINUX_SIG_VALID(args->arg2))
1700 			return (EINVAL);
1701 		em = em_find(p, EMUL_DOLOCK);
1702 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1703 		em->pdeath_signal = args->arg2;
1704 		EMUL_UNLOCK(&emul_lock);
1705 		break;
1706 	case LINUX_PR_GET_PDEATHSIG:
1707 		em = em_find(p, EMUL_DOLOCK);
1708 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1709 		pdeath_signal = em->pdeath_signal;
1710 		EMUL_UNLOCK(&emul_lock);
1711 		error = copyout(&pdeath_signal,
1712 		    (void *)(register_t)args->arg2,
1713 		    sizeof(pdeath_signal));
1714 		break;
1715 	case LINUX_PR_SET_NAME:
1716 		/*
1717 		 * To be on the safe side we need to make sure to not
1718 		 * overflow the size a linux program expects. We already
1719 		 * do this here in the copyin, so that we don't need to
1720 		 * check on copyout.
1721 		 */
1722 		max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1723 		error = copyinstr((void *)(register_t)args->arg2, comm,
1724 		    max_size, NULL);
1725 
1726 		/* Linux silently truncates the name if it is too long. */
1727 		if (error == ENAMETOOLONG) {
1728 			/*
1729 			 * XXX: copyinstr() isn't documented to populate the
1730 			 * array completely, so do a copyin() to be on the
1731 			 * safe side. This should be changed in case
1732 			 * copyinstr() is changed to guarantee this.
1733 			 */
1734 			error = copyin((void *)(register_t)args->arg2, comm,
1735 			    max_size - 1);
1736 			comm[max_size - 1] = '\0';
1737 		}
1738 		if (error)
1739 			return (error);
1740 
1741 		PROC_LOCK(p);
1742 		strlcpy(p->p_comm, comm, sizeof(p->p_comm));
1743 		PROC_UNLOCK(p);
1744 		break;
1745 	case LINUX_PR_GET_NAME:
1746 		PROC_LOCK(p);
1747 		strlcpy(comm, p->p_comm, sizeof(comm));
1748 		PROC_UNLOCK(p);
1749 		error = copyout(comm, (void *)(register_t)args->arg2,
1750 		    strlen(comm) + 1);
1751 		break;
1752 	default:
1753 		error = EINVAL;
1754 		break;
1755 	}
1756 
1757 	return (error);
1758 }
1759 
1760 /*
1761  * Get affinity of a process.
1762  */
1763 int
1764 linux_sched_getaffinity(struct thread *td,
1765     struct linux_sched_getaffinity_args *args)
1766 {
1767 	int error;
1768 	struct cpuset_getaffinity_args cga;
1769 
1770 #ifdef DEBUG
1771 	if (ldebug(sched_getaffinity))
1772 		printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid,
1773 		    args->len);
1774 #endif
1775 	if (args->len < sizeof(cpuset_t))
1776 		return (EINVAL);
1777 
1778 	cga.level = CPU_LEVEL_WHICH;
1779 	cga.which = CPU_WHICH_PID;
1780 	cga.id = args->pid;
1781 	cga.cpusetsize = sizeof(cpuset_t);
1782 	cga.mask = (cpuset_t *) args->user_mask_ptr;
1783 
1784 	if ((error = cpuset_getaffinity(td, &cga)) == 0)
1785 		td->td_retval[0] = sizeof(cpuset_t);
1786 
1787 	return (error);
1788 }
1789 
1790 /*
1791  *  Set affinity of a process.
1792  */
1793 int
1794 linux_sched_setaffinity(struct thread *td,
1795     struct linux_sched_setaffinity_args *args)
1796 {
1797 	struct cpuset_setaffinity_args csa;
1798 
1799 #ifdef DEBUG
1800 	if (ldebug(sched_setaffinity))
1801 		printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid,
1802 		    args->len);
1803 #endif
1804 	if (args->len < sizeof(cpuset_t))
1805 		return (EINVAL);
1806 
1807 	csa.level = CPU_LEVEL_WHICH;
1808 	csa.which = CPU_WHICH_PID;
1809 	csa.id = args->pid;
1810 	csa.cpusetsize = sizeof(cpuset_t);
1811 	csa.mask = (cpuset_t *) args->user_mask_ptr;
1812 
1813 	return (cpuset_setaffinity(td, &csa));
1814 }
1815