xref: /freebsd/sys/compat/linux/linux_misc.c (revision 404825a72b6fdaada592ae93c5e39f0536a96d5a)
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 
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 #include <compat/linux/linux_sysproto.h>
78 #include <compat/linux/linux_emul.h>
79 #include <compat/linux/linux_misc.h>
80 
81 #ifdef COMPAT_LINUX32
82 #include <machine/../linux32/linux.h>
83 #include <machine/../linux32/linux32_proto.h>
84 #else
85 #include <machine/../linux/linux.h>
86 #include <machine/../linux/linux_proto.h>
87 #endif
88 
89 #include <compat/linux/linux_mib.h>
90 #include <compat/linux/linux_signal.h>
91 #include <compat/linux/linux_util.h>
92 
93 #ifdef __i386__
94 #include <machine/cputypes.h>
95 #endif
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 	int error;
173 
174 #ifdef DEBUG
175 	if (ldebug(alarm))
176 		printf(ARGS(alarm, "%u"), args->secs);
177 #endif
178 
179 	if (args->secs > 100000000)
180 		return (EINVAL);
181 
182 	it.it_value.tv_sec = (long)args->secs;
183 	it.it_value.tv_usec = 0;
184 	it.it_interval.tv_sec = 0;
185 	it.it_interval.tv_usec = 0;
186 	error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
187 	if (error)
188 		return (error);
189 	if (timevalisset(&old_it.it_value)) {
190 		if (old_it.it_value.tv_usec != 0)
191 			old_it.it_value.tv_sec++;
192 		td->td_retval[0] = old_it.it_value.tv_sec;
193 	}
194 	return (0);
195 }
196 
197 int
198 linux_brk(struct thread *td, struct linux_brk_args *args)
199 {
200 	struct vmspace *vm = td->td_proc->p_vmspace;
201 	vm_offset_t new, old;
202 	struct obreak_args /* {
203 		char * nsize;
204 	} */ tmp;
205 
206 #ifdef DEBUG
207 	if (ldebug(brk))
208 		printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
209 #endif
210 	old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize);
211 	new = (vm_offset_t)args->dsend;
212 	tmp.nsize = (char *)new;
213 	if (((caddr_t)new > vm->vm_daddr) && !obreak(td, &tmp))
214 		td->td_retval[0] = (long)new;
215 	else
216 		td->td_retval[0] = (long)old;
217 
218 	return 0;
219 }
220 
221 #if defined(__i386__)
222 /* XXX: what about amd64/linux32? */
223 
224 int
225 linux_uselib(struct thread *td, struct linux_uselib_args *args)
226 {
227 	struct nameidata ni;
228 	struct vnode *vp;
229 	struct exec *a_out;
230 	struct vattr attr;
231 	vm_offset_t vmaddr;
232 	unsigned long file_offset;
233 	vm_offset_t buffer;
234 	unsigned long bss_size;
235 	char *library;
236 	int error;
237 	int locked, vfslocked;
238 
239 	LCONVPATHEXIST(td, args->library, &library);
240 
241 #ifdef DEBUG
242 	if (ldebug(uselib))
243 		printf(ARGS(uselib, "%s"), library);
244 #endif
245 
246 	a_out = NULL;
247 	vfslocked = 0;
248 	locked = 0;
249 	vp = NULL;
250 
251 	NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
252 	    UIO_SYSSPACE, library, td);
253 	error = namei(&ni);
254 	LFREEPATH(library);
255 	if (error)
256 		goto cleanup;
257 
258 	vp = ni.ni_vp;
259 	vfslocked = NDHASGIANT(&ni);
260 	NDFREE(&ni, NDF_ONLY_PNBUF);
261 
262 	/*
263 	 * From here on down, we have a locked vnode that must be unlocked.
264 	 * XXX: The code below largely duplicates exec_check_permissions().
265 	 */
266 	locked = 1;
267 
268 	/* Writable? */
269 	if (vp->v_writecount) {
270 		error = ETXTBSY;
271 		goto cleanup;
272 	}
273 
274 	/* Executable? */
275 	error = VOP_GETATTR(vp, &attr, td->td_ucred, td);
276 	if (error)
277 		goto cleanup;
278 
279 	if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
280 	    ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
281 		/* EACCESS is what exec(2) returns. */
282 		error = ENOEXEC;
283 		goto cleanup;
284 	}
285 
286 	/* Sensible size? */
287 	if (attr.va_size == 0) {
288 		error = ENOEXEC;
289 		goto cleanup;
290 	}
291 
292 	/* Can we access it? */
293 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
294 	if (error)
295 		goto cleanup;
296 
297 	/*
298 	 * XXX: This should use vn_open() so that it is properly authorized,
299 	 * and to reduce code redundancy all over the place here.
300 	 * XXX: Not really, it duplicates far more of exec_check_permissions()
301 	 * than vn_open().
302 	 */
303 #ifdef MAC
304 	error = mac_vnode_check_open(td->td_ucred, vp, FREAD);
305 	if (error)
306 		goto cleanup;
307 #endif
308 	error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
309 	if (error)
310 		goto cleanup;
311 
312 	/* Pull in executable header into kernel_map */
313 	error = vm_mmap(kernel_map, (vm_offset_t *)&a_out, PAGE_SIZE,
314 	    VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
315 	if (error)
316 		goto cleanup;
317 
318 	/* Is it a Linux binary ? */
319 	if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
320 		error = ENOEXEC;
321 		goto cleanup;
322 	}
323 
324 	/*
325 	 * While we are here, we should REALLY do some more checks
326 	 */
327 
328 	/* Set file/virtual offset based on a.out variant. */
329 	switch ((int)(a_out->a_magic & 0xffff)) {
330 	case 0413:			/* ZMAGIC */
331 		file_offset = 1024;
332 		break;
333 	case 0314:			/* QMAGIC */
334 		file_offset = 0;
335 		break;
336 	default:
337 		error = ENOEXEC;
338 		goto cleanup;
339 	}
340 
341 	bss_size = round_page(a_out->a_bss);
342 
343 	/* Check various fields in header for validity/bounds. */
344 	if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
345 		error = ENOEXEC;
346 		goto cleanup;
347 	}
348 
349 	/* text + data can't exceed file size */
350 	if (a_out->a_data + a_out->a_text > attr.va_size) {
351 		error = EFAULT;
352 		goto cleanup;
353 	}
354 
355 	/*
356 	 * text/data/bss must not exceed limits
357 	 * XXX - this is not complete. it should check current usage PLUS
358 	 * the resources needed by this library.
359 	 */
360 	PROC_LOCK(td->td_proc);
361 	if (a_out->a_text > maxtsiz ||
362 	    a_out->a_data + bss_size > lim_cur(td->td_proc, RLIMIT_DATA)) {
363 		PROC_UNLOCK(td->td_proc);
364 		error = ENOMEM;
365 		goto cleanup;
366 	}
367 	PROC_UNLOCK(td->td_proc);
368 
369 	/*
370 	 * Prevent more writers.
371 	 * XXX: Note that if any of the VM operations fail below we don't
372 	 * clear this flag.
373 	 */
374 	vp->v_vflag |= VV_TEXT;
375 
376 	/*
377 	 * Lock no longer needed
378 	 */
379 	locked = 0;
380 	VOP_UNLOCK(vp, 0);
381 	VFS_UNLOCK_GIANT(vfslocked);
382 
383 	/*
384 	 * Check if file_offset page aligned. Currently we cannot handle
385 	 * misalinged file offsets, and so we read in the entire image
386 	 * (what a waste).
387 	 */
388 	if (file_offset & PAGE_MASK) {
389 #ifdef DEBUG
390 		printf("uselib: Non page aligned binary %lu\n", file_offset);
391 #endif
392 		/* Map text+data read/write/execute */
393 
394 		/* a_entry is the load address and is page aligned */
395 		vmaddr = trunc_page(a_out->a_entry);
396 
397 		/* get anon user mapping, read+write+execute */
398 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
399 		    &vmaddr, a_out->a_text + a_out->a_data, FALSE, VM_PROT_ALL,
400 		    VM_PROT_ALL, 0);
401 		if (error)
402 			goto cleanup;
403 
404 		/* map file into kernel_map */
405 		error = vm_mmap(kernel_map, &buffer,
406 		    round_page(a_out->a_text + a_out->a_data + file_offset),
407 		    VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp,
408 		    trunc_page(file_offset));
409 		if (error)
410 			goto cleanup;
411 
412 		/* copy from kernel VM space to user space */
413 		error = copyout(PTRIN(buffer + file_offset),
414 		    (void *)vmaddr, a_out->a_text + a_out->a_data);
415 
416 		/* release temporary kernel space */
417 		vm_map_remove(kernel_map, buffer, buffer +
418 		    round_page(a_out->a_text + a_out->a_data + file_offset));
419 
420 		if (error)
421 			goto cleanup;
422 	} else {
423 #ifdef DEBUG
424 		printf("uselib: Page aligned binary %lu\n", file_offset);
425 #endif
426 		/*
427 		 * for QMAGIC, a_entry is 20 bytes beyond the load address
428 		 * to skip the executable header
429 		 */
430 		vmaddr = trunc_page(a_out->a_entry);
431 
432 		/*
433 		 * Map it all into the process's space as a single
434 		 * copy-on-write "data" segment.
435 		 */
436 		error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr,
437 		    a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
438 		    MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
439 		if (error)
440 			goto cleanup;
441 	}
442 #ifdef DEBUG
443 	printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
444 	    ((long *)vmaddr)[1]);
445 #endif
446 	if (bss_size != 0) {
447 		/* Calculate BSS start address */
448 		vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
449 		    a_out->a_data;
450 
451 		/* allocate some 'anon' space */
452 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
453 		    &vmaddr, bss_size, FALSE, VM_PROT_ALL, VM_PROT_ALL, 0);
454 		if (error)
455 			goto cleanup;
456 	}
457 
458 cleanup:
459 	/* Unlock vnode if needed */
460 	if (locked) {
461 		VOP_UNLOCK(vp, 0);
462 		VFS_UNLOCK_GIANT(vfslocked);
463 	}
464 
465 	/* Release the kernel mapping. */
466 	if (a_out)
467 		vm_map_remove(kernel_map, (vm_offset_t)a_out,
468 		    (vm_offset_t)a_out + PAGE_SIZE);
469 
470 	return error;
471 }
472 
473 #endif	/* __i386__ */
474 
475 int
476 linux_select(struct thread *td, struct linux_select_args *args)
477 {
478 	l_timeval ltv;
479 	struct timeval tv0, tv1, utv, *tvp;
480 	int error;
481 
482 #ifdef DEBUG
483 	if (ldebug(select))
484 		printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
485 		    (void *)args->readfds, (void *)args->writefds,
486 		    (void *)args->exceptfds, (void *)args->timeout);
487 #endif
488 
489 	/*
490 	 * Store current time for computation of the amount of
491 	 * time left.
492 	 */
493 	if (args->timeout) {
494 		if ((error = copyin(args->timeout, &ltv, sizeof(ltv))))
495 			goto select_out;
496 		utv.tv_sec = ltv.tv_sec;
497 		utv.tv_usec = ltv.tv_usec;
498 #ifdef DEBUG
499 		if (ldebug(select))
500 			printf(LMSG("incoming timeout (%jd/%ld)"),
501 			    (intmax_t)utv.tv_sec, utv.tv_usec);
502 #endif
503 
504 		if (itimerfix(&utv)) {
505 			/*
506 			 * The timeval was invalid.  Convert it to something
507 			 * valid that will act as it does under Linux.
508 			 */
509 			utv.tv_sec += utv.tv_usec / 1000000;
510 			utv.tv_usec %= 1000000;
511 			if (utv.tv_usec < 0) {
512 				utv.tv_sec -= 1;
513 				utv.tv_usec += 1000000;
514 			}
515 			if (utv.tv_sec < 0)
516 				timevalclear(&utv);
517 		}
518 		microtime(&tv0);
519 		tvp = &utv;
520 	} else
521 		tvp = NULL;
522 
523 	error = kern_select(td, args->nfds, args->readfds, args->writefds,
524 	    args->exceptfds, tvp);
525 
526 #ifdef DEBUG
527 	if (ldebug(select))
528 		printf(LMSG("real select returns %d"), error);
529 #endif
530 	if (error) {
531 		/*
532 		 * See fs/select.c in the Linux kernel.  Without this,
533 		 * Maelstrom doesn't work.
534 		 */
535 		if (error == ERESTART)
536 			error = EINTR;
537 		goto select_out;
538 	}
539 
540 	if (args->timeout) {
541 		if (td->td_retval[0]) {
542 			/*
543 			 * Compute how much time was left of the timeout,
544 			 * by subtracting the current time and the time
545 			 * before we started the call, and subtracting
546 			 * that result from the user-supplied value.
547 			 */
548 			microtime(&tv1);
549 			timevalsub(&tv1, &tv0);
550 			timevalsub(&utv, &tv1);
551 			if (utv.tv_sec < 0)
552 				timevalclear(&utv);
553 		} else
554 			timevalclear(&utv);
555 #ifdef DEBUG
556 		if (ldebug(select))
557 			printf(LMSG("outgoing timeout (%jd/%ld)"),
558 			    (intmax_t)utv.tv_sec, utv.tv_usec);
559 #endif
560 		ltv.tv_sec = utv.tv_sec;
561 		ltv.tv_usec = utv.tv_usec;
562 		if ((error = copyout(&ltv, args->timeout, sizeof(ltv))))
563 			goto select_out;
564 	}
565 
566 select_out:
567 #ifdef DEBUG
568 	if (ldebug(select))
569 		printf(LMSG("select_out -> %d"), error);
570 #endif
571 	return error;
572 }
573 
574 int
575 linux_mremap(struct thread *td, struct linux_mremap_args *args)
576 {
577 	struct munmap_args /* {
578 		void *addr;
579 		size_t len;
580 	} */ bsd_args;
581 	int error = 0;
582 
583 #ifdef DEBUG
584 	if (ldebug(mremap))
585 		printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
586 		    (void *)(uintptr_t)args->addr,
587 		    (unsigned long)args->old_len,
588 		    (unsigned long)args->new_len,
589 		    (unsigned long)args->flags);
590 #endif
591 
592 	if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
593 		td->td_retval[0] = 0;
594 		return (EINVAL);
595 	}
596 
597 	/*
598 	 * Check for the page alignment.
599 	 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
600 	 */
601 	if (args->addr & PAGE_MASK) {
602 		td->td_retval[0] = 0;
603 		return (EINVAL);
604 	}
605 
606 	args->new_len = round_page(args->new_len);
607 	args->old_len = round_page(args->old_len);
608 
609 	if (args->new_len > args->old_len) {
610 		td->td_retval[0] = 0;
611 		return ENOMEM;
612 	}
613 
614 	if (args->new_len < args->old_len) {
615 		bsd_args.addr =
616 		    (caddr_t)((uintptr_t)args->addr + args->new_len);
617 		bsd_args.len = args->old_len - args->new_len;
618 		error = munmap(td, &bsd_args);
619 	}
620 
621 	td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
622 	return error;
623 }
624 
625 #define LINUX_MS_ASYNC       0x0001
626 #define LINUX_MS_INVALIDATE  0x0002
627 #define LINUX_MS_SYNC        0x0004
628 
629 int
630 linux_msync(struct thread *td, struct linux_msync_args *args)
631 {
632 	struct msync_args bsd_args;
633 
634 	bsd_args.addr = (caddr_t)(uintptr_t)args->addr;
635 	bsd_args.len = (uintptr_t)args->len;
636 	bsd_args.flags = args->fl & ~LINUX_MS_SYNC;
637 
638 	return msync(td, &bsd_args);
639 }
640 
641 int
642 linux_time(struct thread *td, struct linux_time_args *args)
643 {
644 	struct timeval tv;
645 	l_time_t tm;
646 	int error;
647 
648 #ifdef DEBUG
649 	if (ldebug(time))
650 		printf(ARGS(time, "*"));
651 #endif
652 
653 	microtime(&tv);
654 	tm = tv.tv_sec;
655 	if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
656 		return error;
657 	td->td_retval[0] = tm;
658 	return 0;
659 }
660 
661 struct l_times_argv {
662 	l_long	tms_utime;
663 	l_long	tms_stime;
664 	l_long	tms_cutime;
665 	l_long	tms_cstime;
666 };
667 
668 #define CLK_TCK 100			/* Linux uses 100 */
669 
670 #define CONVTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
671 
672 int
673 linux_times(struct thread *td, struct linux_times_args *args)
674 {
675 	struct timeval tv, utime, stime, cutime, cstime;
676 	struct l_times_argv tms;
677 	struct proc *p;
678 	int error;
679 
680 #ifdef DEBUG
681 	if (ldebug(times))
682 		printf(ARGS(times, "*"));
683 #endif
684 
685 	if (args->buf != NULL) {
686 		p = td->td_proc;
687 		PROC_LOCK(p);
688 		PROC_SLOCK(p);
689 		calcru(p, &utime, &stime);
690 		PROC_SUNLOCK(p);
691 		calccru(p, &cutime, &cstime);
692 		PROC_UNLOCK(p);
693 
694 		tms.tms_utime = CONVTCK(utime);
695 		tms.tms_stime = CONVTCK(stime);
696 
697 		tms.tms_cutime = CONVTCK(cutime);
698 		tms.tms_cstime = CONVTCK(cstime);
699 
700 		if ((error = copyout(&tms, args->buf, sizeof(tms))))
701 			return error;
702 	}
703 
704 	microuptime(&tv);
705 	td->td_retval[0] = (int)CONVTCK(tv);
706 	return 0;
707 }
708 
709 int
710 linux_newuname(struct thread *td, struct linux_newuname_args *args)
711 {
712 	struct l_new_utsname utsname;
713 	char osname[LINUX_MAX_UTSNAME];
714 	char osrelease[LINUX_MAX_UTSNAME];
715 	char *p;
716 
717 #ifdef DEBUG
718 	if (ldebug(newuname))
719 		printf(ARGS(newuname, "*"));
720 #endif
721 
722 	linux_get_osname(td, osname);
723 	linux_get_osrelease(td, osrelease);
724 
725 	bzero(&utsname, sizeof(utsname));
726 	strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
727 	getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
728 	strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
729 	strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
730 	for (p = utsname.version; *p != '\0'; ++p)
731 		if (*p == '\n') {
732 			*p = '\0';
733 			break;
734 		}
735 #ifdef __i386__
736 	{
737 		const char *class;
738 
739 		switch (cpu_class) {
740 		case CPUCLASS_686:
741 			class = "i686";
742 			break;
743 		case CPUCLASS_586:
744 			class = "i586";
745 			break;
746 		case CPUCLASS_486:
747 			class = "i486";
748 			break;
749 		default:
750 			class = "i386";
751 		}
752 		strlcpy(utsname.machine, class, LINUX_MAX_UTSNAME);
753 	}
754 #elif defined(__amd64__)	/* XXX: Linux can change 'personality'. */
755 #ifdef COMPAT_LINUX32
756 	strlcpy(utsname.machine, "i686", LINUX_MAX_UTSNAME);
757 #else
758 	strlcpy(utsname.machine, "x86_64", LINUX_MAX_UTSNAME);
759 #endif /* COMPAT_LINUX32 */
760 #else /* something other than i386 or amd64 - assume we and Linux agree */
761 	strlcpy(utsname.machine, machine, LINUX_MAX_UTSNAME);
762 #endif /* __i386__ */
763 	strlcpy(utsname.domainname, domainname, LINUX_MAX_UTSNAME);
764 
765 	return (copyout(&utsname, args->buf, sizeof(utsname)));
766 }
767 
768 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
769 struct l_utimbuf {
770 	l_time_t l_actime;
771 	l_time_t l_modtime;
772 };
773 
774 int
775 linux_utime(struct thread *td, struct linux_utime_args *args)
776 {
777 	struct timeval tv[2], *tvp;
778 	struct l_utimbuf lut;
779 	char *fname;
780 	int error;
781 
782 	LCONVPATHEXIST(td, args->fname, &fname);
783 
784 #ifdef DEBUG
785 	if (ldebug(utime))
786 		printf(ARGS(utime, "%s, *"), fname);
787 #endif
788 
789 	if (args->times) {
790 		if ((error = copyin(args->times, &lut, sizeof lut))) {
791 			LFREEPATH(fname);
792 			return error;
793 		}
794 		tv[0].tv_sec = lut.l_actime;
795 		tv[0].tv_usec = 0;
796 		tv[1].tv_sec = lut.l_modtime;
797 		tv[1].tv_usec = 0;
798 		tvp = tv;
799 	} else
800 		tvp = NULL;
801 
802 	error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
803 	LFREEPATH(fname);
804 	return (error);
805 }
806 
807 int
808 linux_utimes(struct thread *td, struct linux_utimes_args *args)
809 {
810 	l_timeval ltv[2];
811 	struct timeval tv[2], *tvp = NULL;
812 	char *fname;
813 	int error;
814 
815 	LCONVPATHEXIST(td, args->fname, &fname);
816 
817 #ifdef DEBUG
818 	if (ldebug(utimes))
819 		printf(ARGS(utimes, "%s, *"), fname);
820 #endif
821 
822 	if (args->tptr != NULL) {
823 		if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
824 			LFREEPATH(fname);
825 			return (error);
826 		}
827 		tv[0].tv_sec = ltv[0].tv_sec;
828 		tv[0].tv_usec = ltv[0].tv_usec;
829 		tv[1].tv_sec = ltv[1].tv_sec;
830 		tv[1].tv_usec = ltv[1].tv_usec;
831 		tvp = tv;
832 	}
833 
834 	error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
835 	LFREEPATH(fname);
836 	return (error);
837 }
838 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
839 
840 #define __WCLONE 0x80000000
841 
842 int
843 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
844 {
845 	int error, options, tmpstat;
846 
847 #ifdef DEBUG
848 	if (ldebug(waitpid))
849 		printf(ARGS(waitpid, "%d, %p, %d"),
850 		    args->pid, (void *)args->status, args->options);
851 #endif
852 	/*
853 	 * this is necessary because the test in kern_wait doesn't work
854 	 * because we mess with the options here
855 	 */
856 	if (args->options & ~(WUNTRACED | WNOHANG | WCONTINUED | __WCLONE))
857 		return (EINVAL);
858 
859 	options = (args->options & (WNOHANG | WUNTRACED));
860 	/* WLINUXCLONE should be equal to __WCLONE, but we make sure */
861 	if (args->options & __WCLONE)
862 		options |= WLINUXCLONE;
863 
864 	error = kern_wait(td, args->pid, &tmpstat, options, NULL);
865 	if (error)
866 		return error;
867 
868 	if (args->status) {
869 		tmpstat &= 0xffff;
870 		if (WIFSIGNALED(tmpstat))
871 			tmpstat = (tmpstat & 0xffffff80) |
872 			    BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat));
873 		else if (WIFSTOPPED(tmpstat))
874 			tmpstat = (tmpstat & 0xffff00ff) |
875 			    (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8);
876 		return copyout(&tmpstat, args->status, sizeof(int));
877 	}
878 
879 	return 0;
880 }
881 
882 int
883 linux_wait4(struct thread *td, struct linux_wait4_args *args)
884 {
885 	int error, options, tmpstat;
886 	struct rusage ru, *rup;
887 	struct proc *p;
888 
889 #ifdef DEBUG
890 	if (ldebug(wait4))
891 		printf(ARGS(wait4, "%d, %p, %d, %p"),
892 		    args->pid, (void *)args->status, args->options,
893 		    (void *)args->rusage);
894 #endif
895 
896 	options = (args->options & (WNOHANG | WUNTRACED));
897 	/* WLINUXCLONE should be equal to __WCLONE, but we make sure */
898 	if (args->options & __WCLONE)
899 		options |= WLINUXCLONE;
900 
901 	if (args->rusage != NULL)
902 		rup = &ru;
903 	else
904 		rup = NULL;
905 	error = kern_wait(td, args->pid, &tmpstat, options, rup);
906 	if (error)
907 		return error;
908 
909 	p = td->td_proc;
910 	PROC_LOCK(p);
911 	sigqueue_delete(&p->p_sigqueue, SIGCHLD);
912 	PROC_UNLOCK(p);
913 
914 	if (args->status) {
915 		tmpstat &= 0xffff;
916 		if (WIFSIGNALED(tmpstat))
917 			tmpstat = (tmpstat & 0xffffff80) |
918 			    BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat));
919 		else if (WIFSTOPPED(tmpstat))
920 			tmpstat = (tmpstat & 0xffff00ff) |
921 			    (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8);
922 		error = copyout(&tmpstat, args->status, sizeof(int));
923 	}
924 	if (args->rusage != NULL && error == 0)
925 		error = copyout(&ru, args->rusage, sizeof(ru));
926 
927 	return (error);
928 }
929 
930 int
931 linux_mknod(struct thread *td, struct linux_mknod_args *args)
932 {
933 	char *path;
934 	int error;
935 
936 	LCONVPATHCREAT(td, args->path, &path);
937 
938 #ifdef DEBUG
939 	if (ldebug(mknod))
940 		printf(ARGS(mknod, "%s, %d, %d"), path, args->mode, args->dev);
941 #endif
942 
943 	switch (args->mode & S_IFMT) {
944 	case S_IFIFO:
945 	case S_IFSOCK:
946 		error = kern_mkfifo(td, path, UIO_SYSSPACE, args->mode);
947 		break;
948 
949 	case S_IFCHR:
950 	case S_IFBLK:
951 		error = kern_mknod(td, path, UIO_SYSSPACE, args->mode,
952 		    args->dev);
953 		break;
954 
955 	case S_IFDIR:
956 		error = EPERM;
957 		break;
958 
959 	case 0:
960 		args->mode |= S_IFREG;
961 		/* FALLTHROUGH */
962 	case S_IFREG:
963 		error = kern_open(td, path, UIO_SYSSPACE,
964 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
965 		break;
966 
967 	default:
968 		error = EINVAL;
969 		break;
970 	}
971 	LFREEPATH(path);
972 	return (error);
973 }
974 
975 /*
976  * UGH! This is just about the dumbest idea I've ever heard!!
977  */
978 int
979 linux_personality(struct thread *td, struct linux_personality_args *args)
980 {
981 #ifdef DEBUG
982 	if (ldebug(personality))
983 		printf(ARGS(personality, "%lu"), (unsigned long)args->per);
984 #endif
985 	if (args->per != 0)
986 		return EINVAL;
987 
988 	/* Yes Jim, it's still a Linux... */
989 	td->td_retval[0] = 0;
990 	return 0;
991 }
992 
993 struct l_itimerval {
994 	l_timeval it_interval;
995 	l_timeval it_value;
996 };
997 
998 #define	B2L_ITIMERVAL(bip, lip) 					\
999 	(bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec;		\
1000 	(bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec;	\
1001 	(bip)->it_value.tv_sec = (lip)->it_value.tv_sec;		\
1002 	(bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1003 
1004 int
1005 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1006 {
1007 	int error;
1008 	struct l_itimerval ls;
1009 	struct itimerval aitv, oitv;
1010 
1011 #ifdef DEBUG
1012 	if (ldebug(setitimer))
1013 		printf(ARGS(setitimer, "%p, %p"),
1014 		    (void *)uap->itv, (void *)uap->oitv);
1015 #endif
1016 
1017 	if (uap->itv == NULL) {
1018 		uap->itv = uap->oitv;
1019 		return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1020 	}
1021 
1022 	error = copyin(uap->itv, &ls, sizeof(ls));
1023 	if (error != 0)
1024 		return (error);
1025 	B2L_ITIMERVAL(&aitv, &ls);
1026 #ifdef DEBUG
1027 	if (ldebug(setitimer)) {
1028 		printf("setitimer: value: sec: %jd, usec: %ld\n",
1029 		    (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1030 		printf("setitimer: interval: sec: %jd, usec: %ld\n",
1031 		    (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1032 	}
1033 #endif
1034 	error = kern_setitimer(td, uap->which, &aitv, &oitv);
1035 	if (error != 0 || uap->oitv == NULL)
1036 		return (error);
1037 	B2L_ITIMERVAL(&ls, &oitv);
1038 
1039 	return (copyout(&ls, uap->oitv, sizeof(ls)));
1040 }
1041 
1042 int
1043 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1044 {
1045 	int error;
1046 	struct l_itimerval ls;
1047 	struct itimerval aitv;
1048 
1049 #ifdef DEBUG
1050 	if (ldebug(getitimer))
1051 		printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1052 #endif
1053 	error = kern_getitimer(td, uap->which, &aitv);
1054 	if (error != 0)
1055 		return (error);
1056 	B2L_ITIMERVAL(&ls, &aitv);
1057 	return (copyout(&ls, uap->itv, sizeof(ls)));
1058 }
1059 
1060 int
1061 linux_nice(struct thread *td, struct linux_nice_args *args)
1062 {
1063 	struct setpriority_args bsd_args;
1064 
1065 	bsd_args.which = PRIO_PROCESS;
1066 	bsd_args.who = 0;		/* current process */
1067 	bsd_args.prio = args->inc;
1068 	return setpriority(td, &bsd_args);
1069 }
1070 
1071 int
1072 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1073 {
1074 	struct ucred *newcred, *oldcred;
1075 	l_gid_t linux_gidset[NGROUPS];
1076 	gid_t *bsd_gidset;
1077 	int ngrp, error;
1078 	struct proc *p;
1079 
1080 	ngrp = args->gidsetsize;
1081 	if (ngrp < 0 || ngrp >= NGROUPS)
1082 		return (EINVAL);
1083 	error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1084 	if (error)
1085 		return (error);
1086 	newcred = crget();
1087 	p = td->td_proc;
1088 	PROC_LOCK(p);
1089 	oldcred = p->p_ucred;
1090 
1091 	/*
1092 	 * cr_groups[0] holds egid. Setting the whole set from
1093 	 * the supplied set will cause egid to be changed too.
1094 	 * Keep cr_groups[0] unchanged to prevent that.
1095 	 */
1096 
1097 	if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) {
1098 		PROC_UNLOCK(p);
1099 		crfree(newcred);
1100 		return (error);
1101 	}
1102 
1103 	crcopy(newcred, oldcred);
1104 	if (ngrp > 0) {
1105 		newcred->cr_ngroups = ngrp + 1;
1106 
1107 		bsd_gidset = newcred->cr_groups;
1108 		ngrp--;
1109 		while (ngrp >= 0) {
1110 			bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1111 			ngrp--;
1112 		}
1113 	} else
1114 		newcred->cr_ngroups = 1;
1115 
1116 	setsugid(p);
1117 	p->p_ucred = newcred;
1118 	PROC_UNLOCK(p);
1119 	crfree(oldcred);
1120 	return (0);
1121 }
1122 
1123 int
1124 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1125 {
1126 	struct ucred *cred;
1127 	l_gid_t linux_gidset[NGROUPS];
1128 	gid_t *bsd_gidset;
1129 	int bsd_gidsetsz, ngrp, error;
1130 
1131 	cred = td->td_ucred;
1132 	bsd_gidset = cred->cr_groups;
1133 	bsd_gidsetsz = cred->cr_ngroups - 1;
1134 
1135 	/*
1136 	 * cr_groups[0] holds egid. Returning the whole set
1137 	 * here will cause a duplicate. Exclude cr_groups[0]
1138 	 * to prevent that.
1139 	 */
1140 
1141 	if ((ngrp = args->gidsetsize) == 0) {
1142 		td->td_retval[0] = bsd_gidsetsz;
1143 		return (0);
1144 	}
1145 
1146 	if (ngrp < bsd_gidsetsz)
1147 		return (EINVAL);
1148 
1149 	ngrp = 0;
1150 	while (ngrp < bsd_gidsetsz) {
1151 		linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1152 		ngrp++;
1153 	}
1154 
1155 	if ((error = copyout(linux_gidset, args->grouplist,
1156 	    ngrp * sizeof(l_gid_t))))
1157 		return (error);
1158 
1159 	td->td_retval[0] = ngrp;
1160 	return (0);
1161 }
1162 
1163 int
1164 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1165 {
1166 	struct rlimit bsd_rlim;
1167 	struct l_rlimit rlim;
1168 	u_int which;
1169 	int error;
1170 
1171 #ifdef DEBUG
1172 	if (ldebug(setrlimit))
1173 		printf(ARGS(setrlimit, "%d, %p"),
1174 		    args->resource, (void *)args->rlim);
1175 #endif
1176 
1177 	if (args->resource >= LINUX_RLIM_NLIMITS)
1178 		return (EINVAL);
1179 
1180 	which = linux_to_bsd_resource[args->resource];
1181 	if (which == -1)
1182 		return (EINVAL);
1183 
1184 	error = copyin(args->rlim, &rlim, sizeof(rlim));
1185 	if (error)
1186 		return (error);
1187 
1188 	bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1189 	bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1190 	return (kern_setrlimit(td, which, &bsd_rlim));
1191 }
1192 
1193 int
1194 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1195 {
1196 	struct l_rlimit rlim;
1197 	struct proc *p = td->td_proc;
1198 	struct rlimit bsd_rlim;
1199 	u_int which;
1200 
1201 #ifdef DEBUG
1202 	if (ldebug(old_getrlimit))
1203 		printf(ARGS(old_getrlimit, "%d, %p"),
1204 		    args->resource, (void *)args->rlim);
1205 #endif
1206 
1207 	if (args->resource >= LINUX_RLIM_NLIMITS)
1208 		return (EINVAL);
1209 
1210 	which = linux_to_bsd_resource[args->resource];
1211 	if (which == -1)
1212 		return (EINVAL);
1213 
1214 	PROC_LOCK(p);
1215 	lim_rlimit(p, which, &bsd_rlim);
1216 	PROC_UNLOCK(p);
1217 
1218 #ifdef COMPAT_LINUX32
1219 	rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1220 	if (rlim.rlim_cur == UINT_MAX)
1221 		rlim.rlim_cur = INT_MAX;
1222 	rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1223 	if (rlim.rlim_max == UINT_MAX)
1224 		rlim.rlim_max = INT_MAX;
1225 #else
1226 	rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1227 	if (rlim.rlim_cur == ULONG_MAX)
1228 		rlim.rlim_cur = LONG_MAX;
1229 	rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1230 	if (rlim.rlim_max == ULONG_MAX)
1231 		rlim.rlim_max = LONG_MAX;
1232 #endif
1233 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1234 }
1235 
1236 int
1237 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1238 {
1239 	struct l_rlimit rlim;
1240 	struct proc *p = td->td_proc;
1241 	struct rlimit bsd_rlim;
1242 	u_int which;
1243 
1244 #ifdef DEBUG
1245 	if (ldebug(getrlimit))
1246 		printf(ARGS(getrlimit, "%d, %p"),
1247 		    args->resource, (void *)args->rlim);
1248 #endif
1249 
1250 	if (args->resource >= LINUX_RLIM_NLIMITS)
1251 		return (EINVAL);
1252 
1253 	which = linux_to_bsd_resource[args->resource];
1254 	if (which == -1)
1255 		return (EINVAL);
1256 
1257 	PROC_LOCK(p);
1258 	lim_rlimit(p, which, &bsd_rlim);
1259 	PROC_UNLOCK(p);
1260 
1261 	rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1262 	rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1263 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1264 }
1265 
1266 int
1267 linux_sched_setscheduler(struct thread *td,
1268     struct linux_sched_setscheduler_args *args)
1269 {
1270 	struct sched_setscheduler_args bsd;
1271 
1272 #ifdef DEBUG
1273 	if (ldebug(sched_setscheduler))
1274 		printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1275 		    args->pid, args->policy, (const void *)args->param);
1276 #endif
1277 
1278 	switch (args->policy) {
1279 	case LINUX_SCHED_OTHER:
1280 		bsd.policy = SCHED_OTHER;
1281 		break;
1282 	case LINUX_SCHED_FIFO:
1283 		bsd.policy = SCHED_FIFO;
1284 		break;
1285 	case LINUX_SCHED_RR:
1286 		bsd.policy = SCHED_RR;
1287 		break;
1288 	default:
1289 		return EINVAL;
1290 	}
1291 
1292 	bsd.pid = args->pid;
1293 	bsd.param = (struct sched_param *)args->param;
1294 	return sched_setscheduler(td, &bsd);
1295 }
1296 
1297 int
1298 linux_sched_getscheduler(struct thread *td,
1299     struct linux_sched_getscheduler_args *args)
1300 {
1301 	struct sched_getscheduler_args bsd;
1302 	int error;
1303 
1304 #ifdef DEBUG
1305 	if (ldebug(sched_getscheduler))
1306 		printf(ARGS(sched_getscheduler, "%d"), args->pid);
1307 #endif
1308 
1309 	bsd.pid = args->pid;
1310 	error = sched_getscheduler(td, &bsd);
1311 
1312 	switch (td->td_retval[0]) {
1313 	case SCHED_OTHER:
1314 		td->td_retval[0] = LINUX_SCHED_OTHER;
1315 		break;
1316 	case SCHED_FIFO:
1317 		td->td_retval[0] = LINUX_SCHED_FIFO;
1318 		break;
1319 	case SCHED_RR:
1320 		td->td_retval[0] = LINUX_SCHED_RR;
1321 		break;
1322 	}
1323 
1324 	return error;
1325 }
1326 
1327 int
1328 linux_sched_get_priority_max(struct thread *td,
1329     struct linux_sched_get_priority_max_args *args)
1330 {
1331 	struct sched_get_priority_max_args bsd;
1332 
1333 #ifdef DEBUG
1334 	if (ldebug(sched_get_priority_max))
1335 		printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1336 #endif
1337 
1338 	switch (args->policy) {
1339 	case LINUX_SCHED_OTHER:
1340 		bsd.policy = SCHED_OTHER;
1341 		break;
1342 	case LINUX_SCHED_FIFO:
1343 		bsd.policy = SCHED_FIFO;
1344 		break;
1345 	case LINUX_SCHED_RR:
1346 		bsd.policy = SCHED_RR;
1347 		break;
1348 	default:
1349 		return EINVAL;
1350 	}
1351 	return sched_get_priority_max(td, &bsd);
1352 }
1353 
1354 int
1355 linux_sched_get_priority_min(struct thread *td,
1356     struct linux_sched_get_priority_min_args *args)
1357 {
1358 	struct sched_get_priority_min_args bsd;
1359 
1360 #ifdef DEBUG
1361 	if (ldebug(sched_get_priority_min))
1362 		printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1363 #endif
1364 
1365 	switch (args->policy) {
1366 	case LINUX_SCHED_OTHER:
1367 		bsd.policy = SCHED_OTHER;
1368 		break;
1369 	case LINUX_SCHED_FIFO:
1370 		bsd.policy = SCHED_FIFO;
1371 		break;
1372 	case LINUX_SCHED_RR:
1373 		bsd.policy = SCHED_RR;
1374 		break;
1375 	default:
1376 		return EINVAL;
1377 	}
1378 	return sched_get_priority_min(td, &bsd);
1379 }
1380 
1381 #define REBOOT_CAD_ON	0x89abcdef
1382 #define REBOOT_CAD_OFF	0
1383 #define REBOOT_HALT	0xcdef0123
1384 #define REBOOT_RESTART	0x01234567
1385 #define REBOOT_RESTART2	0xA1B2C3D4
1386 #define REBOOT_POWEROFF	0x4321FEDC
1387 #define REBOOT_MAGIC1	0xfee1dead
1388 #define REBOOT_MAGIC2	0x28121969
1389 #define REBOOT_MAGIC2A	0x05121996
1390 #define REBOOT_MAGIC2B	0x16041998
1391 
1392 int
1393 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1394 {
1395 	struct reboot_args bsd_args;
1396 
1397 #ifdef DEBUG
1398 	if (ldebug(reboot))
1399 		printf(ARGS(reboot, "0x%x"), args->cmd);
1400 #endif
1401 
1402 	if (args->magic1 != REBOOT_MAGIC1)
1403 		return EINVAL;
1404 
1405 	switch (args->magic2) {
1406 	case REBOOT_MAGIC2:
1407 	case REBOOT_MAGIC2A:
1408 	case REBOOT_MAGIC2B:
1409 		break;
1410 	default:
1411 		return EINVAL;
1412 	}
1413 
1414 	switch (args->cmd) {
1415 	case REBOOT_CAD_ON:
1416 	case REBOOT_CAD_OFF:
1417 		return (priv_check(td, PRIV_REBOOT));
1418 	case REBOOT_HALT:
1419 		bsd_args.opt = RB_HALT;
1420 		break;
1421 	case REBOOT_RESTART:
1422 	case REBOOT_RESTART2:
1423 		bsd_args.opt = 0;
1424 		break;
1425 	case REBOOT_POWEROFF:
1426 		bsd_args.opt = RB_POWEROFF;
1427 		break;
1428 	default:
1429 		return EINVAL;
1430 	}
1431 	return reboot(td, &bsd_args);
1432 }
1433 
1434 
1435 /*
1436  * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1437  * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
1438  * are assumed to be preserved. The following lightweight syscalls fixes
1439  * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
1440  *
1441  * linux_getpid() - MP SAFE
1442  * linux_getgid() - MP SAFE
1443  * linux_getuid() - MP SAFE
1444  */
1445 
1446 int
1447 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1448 {
1449 	struct linux_emuldata *em;
1450 
1451 #ifdef DEBUG
1452 	if (ldebug(getpid))
1453 		printf(ARGS(getpid, ""));
1454 #endif
1455 
1456 	if (linux_use26(td)) {
1457 		em = em_find(td->td_proc, EMUL_DONTLOCK);
1458 		KASSERT(em != NULL, ("getpid: emuldata not found.\n"));
1459 		td->td_retval[0] = em->shared->group_pid;
1460 	} else {
1461 		td->td_retval[0] = td->td_proc->p_pid;
1462 	}
1463 
1464 	return (0);
1465 }
1466 
1467 int
1468 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1469 {
1470 
1471 #ifdef DEBUG
1472 	if (ldebug(gettid))
1473 		printf(ARGS(gettid, ""));
1474 #endif
1475 
1476 	td->td_retval[0] = td->td_proc->p_pid;
1477 	return (0);
1478 }
1479 
1480 
1481 int
1482 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1483 {
1484 	struct linux_emuldata *em;
1485 	struct proc *p, *pp;
1486 
1487 #ifdef DEBUG
1488 	if (ldebug(getppid))
1489 		printf(ARGS(getppid, ""));
1490 #endif
1491 
1492 	if (!linux_use26(td)) {
1493 		PROC_LOCK(td->td_proc);
1494 		td->td_retval[0] = td->td_proc->p_pptr->p_pid;
1495 		PROC_UNLOCK(td->td_proc);
1496 		return (0);
1497 	}
1498 
1499 	em = em_find(td->td_proc, EMUL_DONTLOCK);
1500 
1501 	KASSERT(em != NULL, ("getppid: process emuldata not found.\n"));
1502 
1503 	/* find the group leader */
1504 	p = pfind(em->shared->group_pid);
1505 
1506 	if (p == NULL) {
1507 #ifdef DEBUG
1508 	   	printf(LMSG("parent process not found.\n"));
1509 #endif
1510 		return (0);
1511 	}
1512 
1513 	pp = p->p_pptr;		/* switch to parent */
1514 	PROC_LOCK(pp);
1515 	PROC_UNLOCK(p);
1516 
1517 	/* if its also linux process */
1518 	if (pp->p_sysent == &elf_linux_sysvec) {
1519 		em = em_find(pp, EMUL_DONTLOCK);
1520 		KASSERT(em != NULL, ("getppid: parent emuldata not found.\n"));
1521 
1522 		td->td_retval[0] = em->shared->group_pid;
1523 	} else
1524 		td->td_retval[0] = pp->p_pid;
1525 
1526 	PROC_UNLOCK(pp);
1527 
1528 	return (0);
1529 }
1530 
1531 int
1532 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1533 {
1534 
1535 #ifdef DEBUG
1536 	if (ldebug(getgid))
1537 		printf(ARGS(getgid, ""));
1538 #endif
1539 
1540 	td->td_retval[0] = td->td_ucred->cr_rgid;
1541 	return (0);
1542 }
1543 
1544 int
1545 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1546 {
1547 
1548 #ifdef DEBUG
1549 	if (ldebug(getuid))
1550 		printf(ARGS(getuid, ""));
1551 #endif
1552 
1553 	td->td_retval[0] = td->td_ucred->cr_ruid;
1554 	return (0);
1555 }
1556 
1557 
1558 int
1559 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1560 {
1561 	struct getsid_args bsd;
1562 
1563 #ifdef DEBUG
1564 	if (ldebug(getsid))
1565 		printf(ARGS(getsid, "%i"), args->pid);
1566 #endif
1567 
1568 	bsd.pid = args->pid;
1569 	return getsid(td, &bsd);
1570 }
1571 
1572 int
1573 linux_nosys(struct thread *td, struct nosys_args *ignore)
1574 {
1575 
1576 	return (ENOSYS);
1577 }
1578 
1579 int
1580 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1581 {
1582 	struct getpriority_args bsd_args;
1583 	int error;
1584 
1585 #ifdef DEBUG
1586 	if (ldebug(getpriority))
1587 		printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
1588 #endif
1589 
1590 	bsd_args.which = args->which;
1591 	bsd_args.who = args->who;
1592 	error = getpriority(td, &bsd_args);
1593 	td->td_retval[0] = 20 - td->td_retval[0];
1594 	return error;
1595 }
1596 
1597 int
1598 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1599 {
1600 	int name[2];
1601 
1602 #ifdef DEBUG
1603 	if (ldebug(sethostname))
1604 		printf(ARGS(sethostname, "*, %i"), args->len);
1605 #endif
1606 
1607 	name[0] = CTL_KERN;
1608 	name[1] = KERN_HOSTNAME;
1609 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1610 	    args->len, 0, 0));
1611 }
1612 
1613 int
1614 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1615 {
1616 	struct linux_emuldata *em, *td_em, *tmp_em;
1617 	struct proc *sp;
1618 
1619 #ifdef DEBUG
1620 	if (ldebug(exit_group))
1621 		printf(ARGS(exit_group, "%i"), args->error_code);
1622 #endif
1623 
1624 	if (linux_use26(td)) {
1625 		td_em = em_find(td->td_proc, EMUL_DONTLOCK);
1626 
1627 		KASSERT(td_em != NULL, ("exit_group: emuldata not found.\n"));
1628 
1629 		EMUL_SHARED_RLOCK(&emul_shared_lock);
1630 		LIST_FOREACH_SAFE(em, &td_em->shared->threads, threads, tmp_em) {
1631 			if (em->pid == td_em->pid)
1632 				continue;
1633 
1634 			sp = pfind(em->pid);
1635 			psignal(sp, SIGKILL);
1636 			PROC_UNLOCK(sp);
1637 #ifdef DEBUG
1638 			printf(LMSG("linux_sys_exit_group: kill PID %d\n"), em->pid);
1639 #endif
1640 		}
1641 
1642 		EMUL_SHARED_RUNLOCK(&emul_shared_lock);
1643 	}
1644 	/*
1645 	 * XXX: we should send a signal to the parent if
1646 	 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1647 	 * as it doesnt occur often.
1648 	 */
1649 	exit1(td, W_EXITCODE(args->error_code, 0));
1650 
1651 	return (0);
1652 }
1653 
1654 int
1655 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1656 {
1657 	int error = 0, max_size;
1658 	struct proc *p = td->td_proc;
1659 	char comm[LINUX_MAX_COMM_LEN];
1660 	struct linux_emuldata *em;
1661 	int pdeath_signal;
1662 
1663 #ifdef DEBUG
1664 	if (ldebug(prctl))
1665 		printf(ARGS(prctl, "%d, %d, %d, %d, %d"), args->option,
1666 		    args->arg2, args->arg3, args->arg4, args->arg5);
1667 #endif
1668 
1669 	switch (args->option) {
1670 	case LINUX_PR_SET_PDEATHSIG:
1671 		if (!LINUX_SIG_VALID(args->arg2))
1672 			return (EINVAL);
1673 		em = em_find(p, EMUL_DOLOCK);
1674 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1675 		em->pdeath_signal = args->arg2;
1676 		EMUL_UNLOCK(&emul_lock);
1677 		break;
1678 	case LINUX_PR_GET_PDEATHSIG:
1679 		em = em_find(p, EMUL_DOLOCK);
1680 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1681 		pdeath_signal = em->pdeath_signal;
1682 		EMUL_UNLOCK(&emul_lock);
1683 		error = copyout(&pdeath_signal,
1684 		    (void *)(register_t)args->arg2,
1685 		    sizeof(pdeath_signal));
1686 		break;
1687 	case LINUX_PR_SET_NAME:
1688 		/*
1689 		 * To be on the safe side we need to make sure to not
1690 		 * overflow the size a linux program expects. We already
1691 		 * do this here in the copyin, so that we don't need to
1692 		 * check on copyout.
1693 		 */
1694 		max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1695 		error = copyinstr((void *)(register_t)args->arg2, comm,
1696 		    max_size, NULL);
1697 
1698 		/* Linux silently truncates the name if it is too long. */
1699 		if (error == ENAMETOOLONG) {
1700 			/*
1701 			 * XXX: copyinstr() isn't documented to populate the
1702 			 * array completely, so do a copyin() to be on the
1703 			 * safe side. This should be changed in case
1704 			 * copyinstr() is changed to guarantee this.
1705 			 */
1706 			error = copyin((void *)(register_t)args->arg2, comm,
1707 			    max_size - 1);
1708 			comm[max_size - 1] = '\0';
1709 		}
1710 		if (error)
1711 			return (error);
1712 
1713 		PROC_LOCK(p);
1714 		strlcpy(p->p_comm, comm, sizeof(p->p_comm));
1715 		PROC_UNLOCK(p);
1716 		break;
1717 	case LINUX_PR_GET_NAME:
1718 		PROC_LOCK(p);
1719 		strlcpy(comm, p->p_comm, sizeof(comm));
1720 		PROC_UNLOCK(p);
1721 		error = copyout(comm, (void *)(register_t)args->arg2,
1722 		    strlen(comm) + 1);
1723 		break;
1724 	default:
1725 		error = EINVAL;
1726 		break;
1727 	}
1728 
1729 	return (error);
1730 }
1731 
1732 /*
1733  * XXX: fake one.. waiting for real implementation of affinity mask.
1734  */
1735 int
1736 linux_sched_getaffinity(struct thread *td,
1737     struct linux_sched_getaffinity_args *args)
1738 {
1739 	int error;
1740 	cpumask_t i = ~0;
1741 
1742 	if (args->len < sizeof(cpumask_t))
1743 		return (EINVAL);
1744 
1745 	error = copyout(&i, args->user_mask_ptr, sizeof(cpumask_t));
1746 	if (error)
1747 		return (EFAULT);
1748 
1749 	td->td_retval[0] = sizeof(cpumask_t);
1750 	return (0);
1751 }
1752