xref: /freebsd/sys/compat/linux/linux_misc.c (revision 830940567b49bb0c08dfaed40418999e76616909)
1 /*-
2  * Copyright (c) 2002 Doug Rabson
3  * Copyright (c) 1994-1995 S�ren Schmidt
4  * All rights reserved.
5  *
6  * Redistribution and use in source and binary forms, with or without
7  * modification, are permitted provided that the following conditions
8  * are met:
9  * 1. Redistributions of source code must retain the above copyright
10  *    notice, this list of conditions and the following disclaimer
11  *    in this position and unchanged.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. The name of the author may not be used to endorse or promote products
16  *    derived from this software without specific prior written permission
17  *
18  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
19  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
20  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
21  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
22  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
23  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
24  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
25  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
26  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
27  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
28  */
29 
30 #include <sys/cdefs.h>
31 __FBSDID("$FreeBSD$");
32 
33 #include "opt_compat.h"
34 
35 #include <sys/param.h>
36 #include <sys/blist.h>
37 #include <sys/fcntl.h>
38 #if defined(__i386__)
39 #include <sys/imgact_aout.h>
40 #endif
41 #include <sys/jail.h>
42 #include <sys/kernel.h>
43 #include <sys/limits.h>
44 #include <sys/lock.h>
45 #include <sys/malloc.h>
46 #include <sys/mman.h>
47 #include <sys/mount.h>
48 #include <sys/mutex.h>
49 #include <sys/namei.h>
50 #include <sys/priv.h>
51 #include <sys/proc.h>
52 #include <sys/reboot.h>
53 #include <sys/resourcevar.h>
54 #include <sys/sched.h>
55 #include <sys/signalvar.h>
56 #include <sys/stat.h>
57 #include <sys/syscallsubr.h>
58 #include <sys/sysctl.h>
59 #include <sys/sysproto.h>
60 #include <sys/systm.h>
61 #include <sys/time.h>
62 #include <sys/vmmeter.h>
63 #include <sys/vnode.h>
64 #include <sys/wait.h>
65 #include <sys/cpuset.h>
66 
67 #include <security/mac/mac_framework.h>
68 
69 #include <vm/vm.h>
70 #include <vm/pmap.h>
71 #include <vm/vm_kern.h>
72 #include <vm/vm_map.h>
73 #include <vm/vm_extern.h>
74 #include <vm/vm_object.h>
75 #include <vm/swap_pager.h>
76 
77 #ifdef COMPAT_LINUX32
78 #include <machine/../linux32/linux.h>
79 #include <machine/../linux32/linux32_proto.h>
80 #else
81 #include <machine/../linux/linux.h>
82 #include <machine/../linux/linux_proto.h>
83 #endif
84 
85 #include <compat/linux/linux_file.h>
86 #include <compat/linux/linux_mib.h>
87 #include <compat/linux/linux_signal.h>
88 #include <compat/linux/linux_util.h>
89 #include <compat/linux/linux_sysproto.h>
90 #include <compat/linux/linux_emul.h>
91 #include <compat/linux/linux_misc.h>
92 
93 int stclohz;				/* Statistics clock frequency */
94 
95 #define BSD_TO_LINUX_SIGNAL(sig)	\
96 	(((sig) <= LINUX_SIGTBLSZ) ? bsd_to_linux_signal[_SIG_IDX(sig)] : sig)
97 
98 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
99 	RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
100 	RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
101 	RLIMIT_MEMLOCK, RLIMIT_AS
102 };
103 
104 struct l_sysinfo {
105 	l_long		uptime;		/* Seconds since boot */
106 	l_ulong		loads[3];	/* 1, 5, and 15 minute load averages */
107 #define LINUX_SYSINFO_LOADS_SCALE 65536
108 	l_ulong		totalram;	/* Total usable main memory size */
109 	l_ulong		freeram;	/* Available memory size */
110 	l_ulong		sharedram;	/* Amount of shared memory */
111 	l_ulong		bufferram;	/* Memory used by buffers */
112 	l_ulong		totalswap;	/* Total swap space size */
113 	l_ulong		freeswap;	/* swap space still available */
114 	l_ushort	procs;		/* Number of current processes */
115 	l_ushort	pads;
116 	l_ulong		totalbig;
117 	l_ulong		freebig;
118 	l_uint		mem_unit;
119 	char		_f[20-2*sizeof(l_long)-sizeof(l_int)];	/* padding */
120 };
121 int
122 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
123 {
124 	struct l_sysinfo sysinfo;
125 	vm_object_t object;
126 	int i, j;
127 	struct timespec ts;
128 
129 	getnanouptime(&ts);
130 	if (ts.tv_nsec != 0)
131 		ts.tv_sec++;
132 	sysinfo.uptime = ts.tv_sec;
133 
134 	/* Use the information from the mib to get our load averages */
135 	for (i = 0; i < 3; i++)
136 		sysinfo.loads[i] = averunnable.ldavg[i] *
137 		    LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
138 
139 	sysinfo.totalram = physmem * PAGE_SIZE;
140 	sysinfo.freeram = sysinfo.totalram - cnt.v_wire_count * PAGE_SIZE;
141 
142 	sysinfo.sharedram = 0;
143 	mtx_lock(&vm_object_list_mtx);
144 	TAILQ_FOREACH(object, &vm_object_list, object_list)
145 		if (object->shadow_count > 1)
146 			sysinfo.sharedram += object->resident_page_count;
147 	mtx_unlock(&vm_object_list_mtx);
148 
149 	sysinfo.sharedram *= PAGE_SIZE;
150 	sysinfo.bufferram = 0;
151 
152 	swap_pager_status(&i, &j);
153 	sysinfo.totalswap = i * PAGE_SIZE;
154 	sysinfo.freeswap = (i - j) * PAGE_SIZE;
155 
156 	sysinfo.procs = nprocs;
157 
158 	/* The following are only present in newer Linux kernels. */
159 	sysinfo.totalbig = 0;
160 	sysinfo.freebig = 0;
161 	sysinfo.mem_unit = 1;
162 
163 	return copyout(&sysinfo, args->info, sizeof(sysinfo));
164 }
165 
166 int
167 linux_alarm(struct thread *td, struct linux_alarm_args *args)
168 {
169 	struct itimerval it, old_it;
170 	u_int secs;
171 	int error;
172 
173 #ifdef DEBUG
174 	if (ldebug(alarm))
175 		printf(ARGS(alarm, "%u"), args->secs);
176 #endif
177 
178 	secs = args->secs;
179 
180 	if (secs > INT_MAX)
181 		secs = INT_MAX;
182 
183 	it.it_value.tv_sec = (long) secs;
184 	it.it_value.tv_usec = 0;
185 	it.it_interval.tv_sec = 0;
186 	it.it_interval.tv_usec = 0;
187 	error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
188 	if (error)
189 		return (error);
190 	if (timevalisset(&old_it.it_value)) {
191 		if (old_it.it_value.tv_usec != 0)
192 			old_it.it_value.tv_sec++;
193 		td->td_retval[0] = old_it.it_value.tv_sec;
194 	}
195 	return (0);
196 }
197 
198 int
199 linux_brk(struct thread *td, struct linux_brk_args *args)
200 {
201 	struct vmspace *vm = td->td_proc->p_vmspace;
202 	vm_offset_t new, old;
203 	struct obreak_args /* {
204 		char * nsize;
205 	} */ tmp;
206 
207 #ifdef DEBUG
208 	if (ldebug(brk))
209 		printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
210 #endif
211 	old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize);
212 	new = (vm_offset_t)args->dsend;
213 	tmp.nsize = (char *)new;
214 	if (((caddr_t)new > vm->vm_daddr) && !obreak(td, &tmp))
215 		td->td_retval[0] = (long)new;
216 	else
217 		td->td_retval[0] = (long)old;
218 
219 	return 0;
220 }
221 
222 #if defined(__i386__)
223 /* XXX: what about amd64/linux32? */
224 
225 int
226 linux_uselib(struct thread *td, struct linux_uselib_args *args)
227 {
228 	struct nameidata ni;
229 	struct vnode *vp;
230 	struct exec *a_out;
231 	struct vattr attr;
232 	vm_offset_t vmaddr;
233 	unsigned long file_offset;
234 	vm_offset_t buffer;
235 	unsigned long bss_size;
236 	char *library;
237 	int error;
238 	int locked, vfslocked;
239 
240 	LCONVPATHEXIST(td, args->library, &library);
241 
242 #ifdef DEBUG
243 	if (ldebug(uselib))
244 		printf(ARGS(uselib, "%s"), library);
245 #endif
246 
247 	a_out = NULL;
248 	vfslocked = 0;
249 	locked = 0;
250 	vp = NULL;
251 
252 	NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
253 	    UIO_SYSSPACE, library, td);
254 	error = namei(&ni);
255 	LFREEPATH(library);
256 	if (error)
257 		goto cleanup;
258 
259 	vp = ni.ni_vp;
260 	vfslocked = NDHASGIANT(&ni);
261 	NDFREE(&ni, NDF_ONLY_PNBUF);
262 
263 	/*
264 	 * From here on down, we have a locked vnode that must be unlocked.
265 	 * XXX: The code below largely duplicates exec_check_permissions().
266 	 */
267 	locked = 1;
268 
269 	/* Writable? */
270 	if (vp->v_writecount) {
271 		error = ETXTBSY;
272 		goto cleanup;
273 	}
274 
275 	/* Executable? */
276 	error = VOP_GETATTR(vp, &attr, td->td_ucred);
277 	if (error)
278 		goto cleanup;
279 
280 	if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
281 	    ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
282 		/* EACCESS is what exec(2) returns. */
283 		error = ENOEXEC;
284 		goto cleanup;
285 	}
286 
287 	/* Sensible size? */
288 	if (attr.va_size == 0) {
289 		error = ENOEXEC;
290 		goto cleanup;
291 	}
292 
293 	/* Can we access it? */
294 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
295 	if (error)
296 		goto cleanup;
297 
298 	/*
299 	 * XXX: This should use vn_open() so that it is properly authorized,
300 	 * and to reduce code redundancy all over the place here.
301 	 * XXX: Not really, it duplicates far more of exec_check_permissions()
302 	 * than vn_open().
303 	 */
304 #ifdef MAC
305 	error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
306 	if (error)
307 		goto cleanup;
308 #endif
309 	error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
310 	if (error)
311 		goto cleanup;
312 
313 	/* Pull in executable header into kernel_map */
314 	error = vm_mmap(kernel_map, (vm_offset_t *)&a_out, PAGE_SIZE,
315 	    VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
316 	if (error)
317 		goto cleanup;
318 
319 	/* Is it a Linux binary ? */
320 	if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
321 		error = ENOEXEC;
322 		goto cleanup;
323 	}
324 
325 	/*
326 	 * While we are here, we should REALLY do some more checks
327 	 */
328 
329 	/* Set file/virtual offset based on a.out variant. */
330 	switch ((int)(a_out->a_magic & 0xffff)) {
331 	case 0413:			/* ZMAGIC */
332 		file_offset = 1024;
333 		break;
334 	case 0314:			/* QMAGIC */
335 		file_offset = 0;
336 		break;
337 	default:
338 		error = ENOEXEC;
339 		goto cleanup;
340 	}
341 
342 	bss_size = round_page(a_out->a_bss);
343 
344 	/* Check various fields in header for validity/bounds. */
345 	if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
346 		error = ENOEXEC;
347 		goto cleanup;
348 	}
349 
350 	/* text + data can't exceed file size */
351 	if (a_out->a_data + a_out->a_text > attr.va_size) {
352 		error = EFAULT;
353 		goto cleanup;
354 	}
355 
356 	/*
357 	 * text/data/bss must not exceed limits
358 	 * XXX - this is not complete. it should check current usage PLUS
359 	 * the resources needed by this library.
360 	 */
361 	PROC_LOCK(td->td_proc);
362 	if (a_out->a_text > maxtsiz ||
363 	    a_out->a_data + bss_size > lim_cur(td->td_proc, RLIMIT_DATA)) {
364 		PROC_UNLOCK(td->td_proc);
365 		error = ENOMEM;
366 		goto cleanup;
367 	}
368 	PROC_UNLOCK(td->td_proc);
369 
370 	/*
371 	 * Prevent more writers.
372 	 * XXX: Note that if any of the VM operations fail below we don't
373 	 * clear this flag.
374 	 */
375 	vp->v_vflag |= VV_TEXT;
376 
377 	/*
378 	 * Lock no longer needed
379 	 */
380 	locked = 0;
381 	VOP_UNLOCK(vp, 0);
382 	VFS_UNLOCK_GIANT(vfslocked);
383 
384 	/*
385 	 * Check if file_offset page aligned. Currently we cannot handle
386 	 * misalinged file offsets, and so we read in the entire image
387 	 * (what a waste).
388 	 */
389 	if (file_offset & PAGE_MASK) {
390 #ifdef DEBUG
391 		printf("uselib: Non page aligned binary %lu\n", file_offset);
392 #endif
393 		/* Map text+data read/write/execute */
394 
395 		/* a_entry is the load address and is page aligned */
396 		vmaddr = trunc_page(a_out->a_entry);
397 
398 		/* get anon user mapping, read+write+execute */
399 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
400 		    &vmaddr, a_out->a_text + a_out->a_data, FALSE, VM_PROT_ALL,
401 		    VM_PROT_ALL, 0);
402 		if (error)
403 			goto cleanup;
404 
405 		/* map file into kernel_map */
406 		error = vm_mmap(kernel_map, &buffer,
407 		    round_page(a_out->a_text + a_out->a_data + file_offset),
408 		    VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp,
409 		    trunc_page(file_offset));
410 		if (error)
411 			goto cleanup;
412 
413 		/* copy from kernel VM space to user space */
414 		error = copyout(PTRIN(buffer + file_offset),
415 		    (void *)vmaddr, a_out->a_text + a_out->a_data);
416 
417 		/* release temporary kernel space */
418 		vm_map_remove(kernel_map, buffer, buffer +
419 		    round_page(a_out->a_text + a_out->a_data + file_offset));
420 
421 		if (error)
422 			goto cleanup;
423 	} else {
424 #ifdef DEBUG
425 		printf("uselib: Page aligned binary %lu\n", file_offset);
426 #endif
427 		/*
428 		 * for QMAGIC, a_entry is 20 bytes beyond the load address
429 		 * to skip the executable header
430 		 */
431 		vmaddr = trunc_page(a_out->a_entry);
432 
433 		/*
434 		 * Map it all into the process's space as a single
435 		 * copy-on-write "data" segment.
436 		 */
437 		error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr,
438 		    a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
439 		    MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
440 		if (error)
441 			goto cleanup;
442 	}
443 #ifdef DEBUG
444 	printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
445 	    ((long *)vmaddr)[1]);
446 #endif
447 	if (bss_size != 0) {
448 		/* Calculate BSS start address */
449 		vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
450 		    a_out->a_data;
451 
452 		/* allocate some 'anon' space */
453 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
454 		    &vmaddr, bss_size, FALSE, VM_PROT_ALL, VM_PROT_ALL, 0);
455 		if (error)
456 			goto cleanup;
457 	}
458 
459 cleanup:
460 	/* Unlock vnode if needed */
461 	if (locked) {
462 		VOP_UNLOCK(vp, 0);
463 		VFS_UNLOCK_GIANT(vfslocked);
464 	}
465 
466 	/* Release the kernel mapping. */
467 	if (a_out)
468 		vm_map_remove(kernel_map, (vm_offset_t)a_out,
469 		    (vm_offset_t)a_out + PAGE_SIZE);
470 
471 	return error;
472 }
473 
474 #endif	/* __i386__ */
475 
476 int
477 linux_select(struct thread *td, struct linux_select_args *args)
478 {
479 	l_timeval ltv;
480 	struct timeval tv0, tv1, utv, *tvp;
481 	int error;
482 
483 #ifdef DEBUG
484 	if (ldebug(select))
485 		printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
486 		    (void *)args->readfds, (void *)args->writefds,
487 		    (void *)args->exceptfds, (void *)args->timeout);
488 #endif
489 
490 	/*
491 	 * Store current time for computation of the amount of
492 	 * time left.
493 	 */
494 	if (args->timeout) {
495 		if ((error = copyin(args->timeout, &ltv, sizeof(ltv))))
496 			goto select_out;
497 		utv.tv_sec = ltv.tv_sec;
498 		utv.tv_usec = ltv.tv_usec;
499 #ifdef DEBUG
500 		if (ldebug(select))
501 			printf(LMSG("incoming timeout (%jd/%ld)"),
502 			    (intmax_t)utv.tv_sec, utv.tv_usec);
503 #endif
504 
505 		if (itimerfix(&utv)) {
506 			/*
507 			 * The timeval was invalid.  Convert it to something
508 			 * valid that will act as it does under Linux.
509 			 */
510 			utv.tv_sec += utv.tv_usec / 1000000;
511 			utv.tv_usec %= 1000000;
512 			if (utv.tv_usec < 0) {
513 				utv.tv_sec -= 1;
514 				utv.tv_usec += 1000000;
515 			}
516 			if (utv.tv_sec < 0)
517 				timevalclear(&utv);
518 		}
519 		microtime(&tv0);
520 		tvp = &utv;
521 	} else
522 		tvp = NULL;
523 
524 	error = kern_select(td, args->nfds, args->readfds, args->writefds,
525 	    args->exceptfds, tvp);
526 
527 #ifdef DEBUG
528 	if (ldebug(select))
529 		printf(LMSG("real select returns %d"), error);
530 #endif
531 	if (error)
532 		goto select_out;
533 
534 	if (args->timeout) {
535 		if (td->td_retval[0]) {
536 			/*
537 			 * Compute how much time was left of the timeout,
538 			 * by subtracting the current time and the time
539 			 * before we started the call, and subtracting
540 			 * that result from the user-supplied value.
541 			 */
542 			microtime(&tv1);
543 			timevalsub(&tv1, &tv0);
544 			timevalsub(&utv, &tv1);
545 			if (utv.tv_sec < 0)
546 				timevalclear(&utv);
547 		} else
548 			timevalclear(&utv);
549 #ifdef DEBUG
550 		if (ldebug(select))
551 			printf(LMSG("outgoing timeout (%jd/%ld)"),
552 			    (intmax_t)utv.tv_sec, utv.tv_usec);
553 #endif
554 		ltv.tv_sec = utv.tv_sec;
555 		ltv.tv_usec = utv.tv_usec;
556 		if ((error = copyout(&ltv, args->timeout, sizeof(ltv))))
557 			goto select_out;
558 	}
559 
560 select_out:
561 #ifdef DEBUG
562 	if (ldebug(select))
563 		printf(LMSG("select_out -> %d"), error);
564 #endif
565 	return error;
566 }
567 
568 int
569 linux_mremap(struct thread *td, struct linux_mremap_args *args)
570 {
571 	struct munmap_args /* {
572 		void *addr;
573 		size_t len;
574 	} */ bsd_args;
575 	int error = 0;
576 
577 #ifdef DEBUG
578 	if (ldebug(mremap))
579 		printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
580 		    (void *)(uintptr_t)args->addr,
581 		    (unsigned long)args->old_len,
582 		    (unsigned long)args->new_len,
583 		    (unsigned long)args->flags);
584 #endif
585 
586 	if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
587 		td->td_retval[0] = 0;
588 		return (EINVAL);
589 	}
590 
591 	/*
592 	 * Check for the page alignment.
593 	 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
594 	 */
595 	if (args->addr & PAGE_MASK) {
596 		td->td_retval[0] = 0;
597 		return (EINVAL);
598 	}
599 
600 	args->new_len = round_page(args->new_len);
601 	args->old_len = round_page(args->old_len);
602 
603 	if (args->new_len > args->old_len) {
604 		td->td_retval[0] = 0;
605 		return ENOMEM;
606 	}
607 
608 	if (args->new_len < args->old_len) {
609 		bsd_args.addr =
610 		    (caddr_t)((uintptr_t)args->addr + args->new_len);
611 		bsd_args.len = args->old_len - args->new_len;
612 		error = munmap(td, &bsd_args);
613 	}
614 
615 	td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
616 	return error;
617 }
618 
619 #define LINUX_MS_ASYNC       0x0001
620 #define LINUX_MS_INVALIDATE  0x0002
621 #define LINUX_MS_SYNC        0x0004
622 
623 int
624 linux_msync(struct thread *td, struct linux_msync_args *args)
625 {
626 	struct msync_args bsd_args;
627 
628 	bsd_args.addr = (caddr_t)(uintptr_t)args->addr;
629 	bsd_args.len = (uintptr_t)args->len;
630 	bsd_args.flags = args->fl & ~LINUX_MS_SYNC;
631 
632 	return msync(td, &bsd_args);
633 }
634 
635 int
636 linux_time(struct thread *td, struct linux_time_args *args)
637 {
638 	struct timeval tv;
639 	l_time_t tm;
640 	int error;
641 
642 #ifdef DEBUG
643 	if (ldebug(time))
644 		printf(ARGS(time, "*"));
645 #endif
646 
647 	microtime(&tv);
648 	tm = tv.tv_sec;
649 	if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
650 		return error;
651 	td->td_retval[0] = tm;
652 	return 0;
653 }
654 
655 struct l_times_argv {
656 	l_clock_t	tms_utime;
657 	l_clock_t	tms_stime;
658 	l_clock_t	tms_cutime;
659 	l_clock_t	tms_cstime;
660 };
661 
662 
663 /*
664  * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
665  * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
666  * auxiliary vector entry.
667  */
668 #define	CLK_TCK		100
669 
670 #define	CONVOTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
671 #define	CONVNTCK(r)	(r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
672 
673 #define	CONVTCK(r)	(linux_kernver(td) >= LINUX_KERNVER_2004000 ?		\
674 			    CONVNTCK(r) : CONVOTCK(r))
675 
676 int
677 linux_times(struct thread *td, struct linux_times_args *args)
678 {
679 	struct timeval tv, utime, stime, cutime, cstime;
680 	struct l_times_argv tms;
681 	struct proc *p;
682 	int error;
683 
684 #ifdef DEBUG
685 	if (ldebug(times))
686 		printf(ARGS(times, "*"));
687 #endif
688 
689 	if (args->buf != NULL) {
690 		p = td->td_proc;
691 		PROC_LOCK(p);
692 		PROC_SLOCK(p);
693 		calcru(p, &utime, &stime);
694 		PROC_SUNLOCK(p);
695 		calccru(p, &cutime, &cstime);
696 		PROC_UNLOCK(p);
697 
698 		tms.tms_utime = CONVTCK(utime);
699 		tms.tms_stime = CONVTCK(stime);
700 
701 		tms.tms_cutime = CONVTCK(cutime);
702 		tms.tms_cstime = CONVTCK(cstime);
703 
704 		if ((error = copyout(&tms, args->buf, sizeof(tms))))
705 			return error;
706 	}
707 
708 	microuptime(&tv);
709 	td->td_retval[0] = (int)CONVTCK(tv);
710 	return 0;
711 }
712 
713 int
714 linux_newuname(struct thread *td, struct linux_newuname_args *args)
715 {
716 	struct l_new_utsname utsname;
717 	char osname[LINUX_MAX_UTSNAME];
718 	char osrelease[LINUX_MAX_UTSNAME];
719 	char *p;
720 
721 #ifdef DEBUG
722 	if (ldebug(newuname))
723 		printf(ARGS(newuname, "*"));
724 #endif
725 
726 	linux_get_osname(td, osname);
727 	linux_get_osrelease(td, osrelease);
728 
729 	bzero(&utsname, sizeof(utsname));
730 	strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
731 	getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
732 	getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
733 	strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
734 	strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
735 	for (p = utsname.version; *p != '\0'; ++p)
736 		if (*p == '\n') {
737 			*p = '\0';
738 			break;
739 		}
740 	strlcpy(utsname.machine, linux_platform, LINUX_MAX_UTSNAME);
741 
742 	return (copyout(&utsname, args->buf, sizeof(utsname)));
743 }
744 
745 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
746 struct l_utimbuf {
747 	l_time_t l_actime;
748 	l_time_t l_modtime;
749 };
750 
751 int
752 linux_utime(struct thread *td, struct linux_utime_args *args)
753 {
754 	struct timeval tv[2], *tvp;
755 	struct l_utimbuf lut;
756 	char *fname;
757 	int error;
758 
759 	LCONVPATHEXIST(td, args->fname, &fname);
760 
761 #ifdef DEBUG
762 	if (ldebug(utime))
763 		printf(ARGS(utime, "%s, *"), fname);
764 #endif
765 
766 	if (args->times) {
767 		if ((error = copyin(args->times, &lut, sizeof lut))) {
768 			LFREEPATH(fname);
769 			return error;
770 		}
771 		tv[0].tv_sec = lut.l_actime;
772 		tv[0].tv_usec = 0;
773 		tv[1].tv_sec = lut.l_modtime;
774 		tv[1].tv_usec = 0;
775 		tvp = tv;
776 	} else
777 		tvp = NULL;
778 
779 	error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
780 	LFREEPATH(fname);
781 	return (error);
782 }
783 
784 int
785 linux_utimes(struct thread *td, struct linux_utimes_args *args)
786 {
787 	l_timeval ltv[2];
788 	struct timeval tv[2], *tvp = NULL;
789 	char *fname;
790 	int error;
791 
792 	LCONVPATHEXIST(td, args->fname, &fname);
793 
794 #ifdef DEBUG
795 	if (ldebug(utimes))
796 		printf(ARGS(utimes, "%s, *"), fname);
797 #endif
798 
799 	if (args->tptr != NULL) {
800 		if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
801 			LFREEPATH(fname);
802 			return (error);
803 		}
804 		tv[0].tv_sec = ltv[0].tv_sec;
805 		tv[0].tv_usec = ltv[0].tv_usec;
806 		tv[1].tv_sec = ltv[1].tv_sec;
807 		tv[1].tv_usec = ltv[1].tv_usec;
808 		tvp = tv;
809 	}
810 
811 	error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
812 	LFREEPATH(fname);
813 	return (error);
814 }
815 
816 int
817 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
818 {
819 	l_timeval ltv[2];
820 	struct timeval tv[2], *tvp = NULL;
821 	char *fname;
822 	int error, dfd;
823 
824 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
825 	LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
826 
827 #ifdef DEBUG
828 	if (ldebug(futimesat))
829 		printf(ARGS(futimesat, "%s, *"), fname);
830 #endif
831 
832 	if (args->utimes != NULL) {
833 		if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
834 			LFREEPATH(fname);
835 			return (error);
836 		}
837 		tv[0].tv_sec = ltv[0].tv_sec;
838 		tv[0].tv_usec = ltv[0].tv_usec;
839 		tv[1].tv_sec = ltv[1].tv_sec;
840 		tv[1].tv_usec = ltv[1].tv_usec;
841 		tvp = tv;
842 	}
843 
844 	error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
845 	LFREEPATH(fname);
846 	return (error);
847 }
848 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
849 
850 #define __WCLONE 0x80000000
851 
852 int
853 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
854 {
855 	int error, options, tmpstat;
856 
857 #ifdef DEBUG
858 	if (ldebug(waitpid))
859 		printf(ARGS(waitpid, "%d, %p, %d"),
860 		    args->pid, (void *)args->status, args->options);
861 #endif
862 	/*
863 	 * this is necessary because the test in kern_wait doesn't work
864 	 * because we mess with the options here
865 	 */
866 	if (args->options & ~(WUNTRACED | WNOHANG | WCONTINUED | __WCLONE))
867 		return (EINVAL);
868 
869 	options = (args->options & (WNOHANG | WUNTRACED));
870 	/* WLINUXCLONE should be equal to __WCLONE, but we make sure */
871 	if (args->options & __WCLONE)
872 		options |= WLINUXCLONE;
873 
874 	error = kern_wait(td, args->pid, &tmpstat, options, NULL);
875 	if (error)
876 		return error;
877 
878 	if (args->status) {
879 		tmpstat &= 0xffff;
880 		if (WIFSIGNALED(tmpstat))
881 			tmpstat = (tmpstat & 0xffffff80) |
882 			    BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat));
883 		else if (WIFSTOPPED(tmpstat))
884 			tmpstat = (tmpstat & 0xffff00ff) |
885 			    (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8);
886 		return copyout(&tmpstat, args->status, sizeof(int));
887 	}
888 
889 	return 0;
890 }
891 
892 int
893 linux_wait4(struct thread *td, struct linux_wait4_args *args)
894 {
895 	int error, options, tmpstat;
896 	struct rusage ru, *rup;
897 	struct proc *p;
898 
899 #ifdef DEBUG
900 	if (ldebug(wait4))
901 		printf(ARGS(wait4, "%d, %p, %d, %p"),
902 		    args->pid, (void *)args->status, args->options,
903 		    (void *)args->rusage);
904 #endif
905 
906 	options = (args->options & (WNOHANG | WUNTRACED));
907 	/* WLINUXCLONE should be equal to __WCLONE, but we make sure */
908 	if (args->options & __WCLONE)
909 		options |= WLINUXCLONE;
910 
911 	if (args->rusage != NULL)
912 		rup = &ru;
913 	else
914 		rup = NULL;
915 	error = kern_wait(td, args->pid, &tmpstat, options, rup);
916 	if (error)
917 		return error;
918 
919 	p = td->td_proc;
920 	PROC_LOCK(p);
921 	sigqueue_delete(&p->p_sigqueue, SIGCHLD);
922 	PROC_UNLOCK(p);
923 
924 	if (args->status) {
925 		tmpstat &= 0xffff;
926 		if (WIFSIGNALED(tmpstat))
927 			tmpstat = (tmpstat & 0xffffff80) |
928 			    BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat));
929 		else if (WIFSTOPPED(tmpstat))
930 			tmpstat = (tmpstat & 0xffff00ff) |
931 			    (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8);
932 		error = copyout(&tmpstat, args->status, sizeof(int));
933 	}
934 	if (args->rusage != NULL && error == 0)
935 		error = copyout(&ru, args->rusage, sizeof(ru));
936 
937 	return (error);
938 }
939 
940 int
941 linux_mknod(struct thread *td, struct linux_mknod_args *args)
942 {
943 	char *path;
944 	int error;
945 
946 	LCONVPATHCREAT(td, args->path, &path);
947 
948 #ifdef DEBUG
949 	if (ldebug(mknod))
950 		printf(ARGS(mknod, "%s, %d, %d"), path, args->mode, args->dev);
951 #endif
952 
953 	switch (args->mode & S_IFMT) {
954 	case S_IFIFO:
955 	case S_IFSOCK:
956 		error = kern_mkfifo(td, path, UIO_SYSSPACE, args->mode);
957 		break;
958 
959 	case S_IFCHR:
960 	case S_IFBLK:
961 		error = kern_mknod(td, path, UIO_SYSSPACE, args->mode,
962 		    args->dev);
963 		break;
964 
965 	case S_IFDIR:
966 		error = EPERM;
967 		break;
968 
969 	case 0:
970 		args->mode |= S_IFREG;
971 		/* FALLTHROUGH */
972 	case S_IFREG:
973 		error = kern_open(td, path, UIO_SYSSPACE,
974 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
975 		if (error == 0)
976 			kern_close(td, td->td_retval[0]);
977 		break;
978 
979 	default:
980 		error = EINVAL;
981 		break;
982 	}
983 	LFREEPATH(path);
984 	return (error);
985 }
986 
987 int
988 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
989 {
990 	char *path;
991 	int error, dfd;
992 
993 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
994 	LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
995 
996 #ifdef DEBUG
997 	if (ldebug(mknodat))
998 		printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev);
999 #endif
1000 
1001 	switch (args->mode & S_IFMT) {
1002 	case S_IFIFO:
1003 	case S_IFSOCK:
1004 		error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
1005 		break;
1006 
1007 	case S_IFCHR:
1008 	case S_IFBLK:
1009 		error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
1010 		    args->dev);
1011 		break;
1012 
1013 	case S_IFDIR:
1014 		error = EPERM;
1015 		break;
1016 
1017 	case 0:
1018 		args->mode |= S_IFREG;
1019 		/* FALLTHROUGH */
1020 	case S_IFREG:
1021 		error = kern_openat(td, dfd, path, UIO_SYSSPACE,
1022 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1023 		if (error == 0)
1024 			kern_close(td, td->td_retval[0]);
1025 		break;
1026 
1027 	default:
1028 		error = EINVAL;
1029 		break;
1030 	}
1031 	LFREEPATH(path);
1032 	return (error);
1033 }
1034 
1035 /*
1036  * UGH! This is just about the dumbest idea I've ever heard!!
1037  */
1038 int
1039 linux_personality(struct thread *td, struct linux_personality_args *args)
1040 {
1041 #ifdef DEBUG
1042 	if (ldebug(personality))
1043 		printf(ARGS(personality, "%lu"), (unsigned long)args->per);
1044 #endif
1045 	if (args->per != 0)
1046 		return EINVAL;
1047 
1048 	/* Yes Jim, it's still a Linux... */
1049 	td->td_retval[0] = 0;
1050 	return 0;
1051 }
1052 
1053 struct l_itimerval {
1054 	l_timeval it_interval;
1055 	l_timeval it_value;
1056 };
1057 
1058 #define	B2L_ITIMERVAL(bip, lip) 					\
1059 	(bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec;		\
1060 	(bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec;	\
1061 	(bip)->it_value.tv_sec = (lip)->it_value.tv_sec;		\
1062 	(bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1063 
1064 int
1065 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1066 {
1067 	int error;
1068 	struct l_itimerval ls;
1069 	struct itimerval aitv, oitv;
1070 
1071 #ifdef DEBUG
1072 	if (ldebug(setitimer))
1073 		printf(ARGS(setitimer, "%p, %p"),
1074 		    (void *)uap->itv, (void *)uap->oitv);
1075 #endif
1076 
1077 	if (uap->itv == NULL) {
1078 		uap->itv = uap->oitv;
1079 		return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1080 	}
1081 
1082 	error = copyin(uap->itv, &ls, sizeof(ls));
1083 	if (error != 0)
1084 		return (error);
1085 	B2L_ITIMERVAL(&aitv, &ls);
1086 #ifdef DEBUG
1087 	if (ldebug(setitimer)) {
1088 		printf("setitimer: value: sec: %jd, usec: %ld\n",
1089 		    (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1090 		printf("setitimer: interval: sec: %jd, usec: %ld\n",
1091 		    (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1092 	}
1093 #endif
1094 	error = kern_setitimer(td, uap->which, &aitv, &oitv);
1095 	if (error != 0 || uap->oitv == NULL)
1096 		return (error);
1097 	B2L_ITIMERVAL(&ls, &oitv);
1098 
1099 	return (copyout(&ls, uap->oitv, sizeof(ls)));
1100 }
1101 
1102 int
1103 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1104 {
1105 	int error;
1106 	struct l_itimerval ls;
1107 	struct itimerval aitv;
1108 
1109 #ifdef DEBUG
1110 	if (ldebug(getitimer))
1111 		printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1112 #endif
1113 	error = kern_getitimer(td, uap->which, &aitv);
1114 	if (error != 0)
1115 		return (error);
1116 	B2L_ITIMERVAL(&ls, &aitv);
1117 	return (copyout(&ls, uap->itv, sizeof(ls)));
1118 }
1119 
1120 int
1121 linux_nice(struct thread *td, struct linux_nice_args *args)
1122 {
1123 	struct setpriority_args bsd_args;
1124 
1125 	bsd_args.which = PRIO_PROCESS;
1126 	bsd_args.who = 0;		/* current process */
1127 	bsd_args.prio = args->inc;
1128 	return setpriority(td, &bsd_args);
1129 }
1130 
1131 int
1132 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1133 {
1134 	struct ucred *newcred, *oldcred;
1135 	l_gid_t *linux_gidset;
1136 	gid_t *bsd_gidset;
1137 	int ngrp, error;
1138 	struct proc *p;
1139 
1140 	ngrp = args->gidsetsize;
1141 	if (ngrp < 0 || ngrp >= NGROUPS)
1142 		return (EINVAL);
1143 	linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_TEMP, M_WAITOK);
1144 	error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1145 	if (error)
1146 		goto out;
1147 	newcred = crget();
1148 	p = td->td_proc;
1149 	PROC_LOCK(p);
1150 	oldcred = crcopysafe(p, newcred);
1151 
1152 	/*
1153 	 * cr_groups[0] holds egid. Setting the whole set from
1154 	 * the supplied set will cause egid to be changed too.
1155 	 * Keep cr_groups[0] unchanged to prevent that.
1156 	 */
1157 
1158 	if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) {
1159 		PROC_UNLOCK(p);
1160 		crfree(newcred);
1161 		goto out;
1162 	}
1163 
1164 	if (ngrp > 0) {
1165 		newcred->cr_ngroups = ngrp + 1;
1166 
1167 		bsd_gidset = newcred->cr_groups;
1168 		ngrp--;
1169 		while (ngrp >= 0) {
1170 			bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1171 			ngrp--;
1172 		}
1173 	} else
1174 		newcred->cr_ngroups = 1;
1175 
1176 	setsugid(p);
1177 	p->p_ucred = newcred;
1178 	PROC_UNLOCK(p);
1179 	crfree(oldcred);
1180 	error = 0;
1181 out:
1182 	free(linux_gidset, M_TEMP);
1183 	return (error);
1184 }
1185 
1186 int
1187 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1188 {
1189 	struct ucred *cred;
1190 	l_gid_t *linux_gidset;
1191 	gid_t *bsd_gidset;
1192 	int bsd_gidsetsz, ngrp, error;
1193 
1194 	cred = td->td_ucred;
1195 	bsd_gidset = cred->cr_groups;
1196 	bsd_gidsetsz = cred->cr_ngroups - 1;
1197 
1198 	/*
1199 	 * cr_groups[0] holds egid. Returning the whole set
1200 	 * here will cause a duplicate. Exclude cr_groups[0]
1201 	 * to prevent that.
1202 	 */
1203 
1204 	if ((ngrp = args->gidsetsize) == 0) {
1205 		td->td_retval[0] = bsd_gidsetsz;
1206 		return (0);
1207 	}
1208 
1209 	if (ngrp < bsd_gidsetsz)
1210 		return (EINVAL);
1211 
1212 	ngrp = 0;
1213 	linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1214 	    M_TEMP, M_WAITOK);
1215 	while (ngrp < bsd_gidsetsz) {
1216 		linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1217 		ngrp++;
1218 	}
1219 
1220 	error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1221 	free(linux_gidset, M_TEMP);
1222 	if (error)
1223 		return (error);
1224 
1225 	td->td_retval[0] = ngrp;
1226 	return (0);
1227 }
1228 
1229 int
1230 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1231 {
1232 	struct rlimit bsd_rlim;
1233 	struct l_rlimit rlim;
1234 	u_int which;
1235 	int error;
1236 
1237 #ifdef DEBUG
1238 	if (ldebug(setrlimit))
1239 		printf(ARGS(setrlimit, "%d, %p"),
1240 		    args->resource, (void *)args->rlim);
1241 #endif
1242 
1243 	if (args->resource >= LINUX_RLIM_NLIMITS)
1244 		return (EINVAL);
1245 
1246 	which = linux_to_bsd_resource[args->resource];
1247 	if (which == -1)
1248 		return (EINVAL);
1249 
1250 	error = copyin(args->rlim, &rlim, sizeof(rlim));
1251 	if (error)
1252 		return (error);
1253 
1254 	bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1255 	bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1256 	return (kern_setrlimit(td, which, &bsd_rlim));
1257 }
1258 
1259 int
1260 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1261 {
1262 	struct l_rlimit rlim;
1263 	struct proc *p = td->td_proc;
1264 	struct rlimit bsd_rlim;
1265 	u_int which;
1266 
1267 #ifdef DEBUG
1268 	if (ldebug(old_getrlimit))
1269 		printf(ARGS(old_getrlimit, "%d, %p"),
1270 		    args->resource, (void *)args->rlim);
1271 #endif
1272 
1273 	if (args->resource >= LINUX_RLIM_NLIMITS)
1274 		return (EINVAL);
1275 
1276 	which = linux_to_bsd_resource[args->resource];
1277 	if (which == -1)
1278 		return (EINVAL);
1279 
1280 	PROC_LOCK(p);
1281 	lim_rlimit(p, which, &bsd_rlim);
1282 	PROC_UNLOCK(p);
1283 
1284 #ifdef COMPAT_LINUX32
1285 	rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1286 	if (rlim.rlim_cur == UINT_MAX)
1287 		rlim.rlim_cur = INT_MAX;
1288 	rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1289 	if (rlim.rlim_max == UINT_MAX)
1290 		rlim.rlim_max = INT_MAX;
1291 #else
1292 	rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1293 	if (rlim.rlim_cur == ULONG_MAX)
1294 		rlim.rlim_cur = LONG_MAX;
1295 	rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1296 	if (rlim.rlim_max == ULONG_MAX)
1297 		rlim.rlim_max = LONG_MAX;
1298 #endif
1299 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1300 }
1301 
1302 int
1303 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1304 {
1305 	struct l_rlimit rlim;
1306 	struct proc *p = td->td_proc;
1307 	struct rlimit bsd_rlim;
1308 	u_int which;
1309 
1310 #ifdef DEBUG
1311 	if (ldebug(getrlimit))
1312 		printf(ARGS(getrlimit, "%d, %p"),
1313 		    args->resource, (void *)args->rlim);
1314 #endif
1315 
1316 	if (args->resource >= LINUX_RLIM_NLIMITS)
1317 		return (EINVAL);
1318 
1319 	which = linux_to_bsd_resource[args->resource];
1320 	if (which == -1)
1321 		return (EINVAL);
1322 
1323 	PROC_LOCK(p);
1324 	lim_rlimit(p, which, &bsd_rlim);
1325 	PROC_UNLOCK(p);
1326 
1327 	rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1328 	rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1329 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1330 }
1331 
1332 int
1333 linux_sched_setscheduler(struct thread *td,
1334     struct linux_sched_setscheduler_args *args)
1335 {
1336 	struct sched_setscheduler_args bsd;
1337 
1338 #ifdef DEBUG
1339 	if (ldebug(sched_setscheduler))
1340 		printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1341 		    args->pid, args->policy, (const void *)args->param);
1342 #endif
1343 
1344 	switch (args->policy) {
1345 	case LINUX_SCHED_OTHER:
1346 		bsd.policy = SCHED_OTHER;
1347 		break;
1348 	case LINUX_SCHED_FIFO:
1349 		bsd.policy = SCHED_FIFO;
1350 		break;
1351 	case LINUX_SCHED_RR:
1352 		bsd.policy = SCHED_RR;
1353 		break;
1354 	default:
1355 		return EINVAL;
1356 	}
1357 
1358 	bsd.pid = args->pid;
1359 	bsd.param = (struct sched_param *)args->param;
1360 	return sched_setscheduler(td, &bsd);
1361 }
1362 
1363 int
1364 linux_sched_getscheduler(struct thread *td,
1365     struct linux_sched_getscheduler_args *args)
1366 {
1367 	struct sched_getscheduler_args bsd;
1368 	int error;
1369 
1370 #ifdef DEBUG
1371 	if (ldebug(sched_getscheduler))
1372 		printf(ARGS(sched_getscheduler, "%d"), args->pid);
1373 #endif
1374 
1375 	bsd.pid = args->pid;
1376 	error = sched_getscheduler(td, &bsd);
1377 
1378 	switch (td->td_retval[0]) {
1379 	case SCHED_OTHER:
1380 		td->td_retval[0] = LINUX_SCHED_OTHER;
1381 		break;
1382 	case SCHED_FIFO:
1383 		td->td_retval[0] = LINUX_SCHED_FIFO;
1384 		break;
1385 	case SCHED_RR:
1386 		td->td_retval[0] = LINUX_SCHED_RR;
1387 		break;
1388 	}
1389 
1390 	return error;
1391 }
1392 
1393 int
1394 linux_sched_get_priority_max(struct thread *td,
1395     struct linux_sched_get_priority_max_args *args)
1396 {
1397 	struct sched_get_priority_max_args bsd;
1398 
1399 #ifdef DEBUG
1400 	if (ldebug(sched_get_priority_max))
1401 		printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1402 #endif
1403 
1404 	switch (args->policy) {
1405 	case LINUX_SCHED_OTHER:
1406 		bsd.policy = SCHED_OTHER;
1407 		break;
1408 	case LINUX_SCHED_FIFO:
1409 		bsd.policy = SCHED_FIFO;
1410 		break;
1411 	case LINUX_SCHED_RR:
1412 		bsd.policy = SCHED_RR;
1413 		break;
1414 	default:
1415 		return EINVAL;
1416 	}
1417 	return sched_get_priority_max(td, &bsd);
1418 }
1419 
1420 int
1421 linux_sched_get_priority_min(struct thread *td,
1422     struct linux_sched_get_priority_min_args *args)
1423 {
1424 	struct sched_get_priority_min_args bsd;
1425 
1426 #ifdef DEBUG
1427 	if (ldebug(sched_get_priority_min))
1428 		printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1429 #endif
1430 
1431 	switch (args->policy) {
1432 	case LINUX_SCHED_OTHER:
1433 		bsd.policy = SCHED_OTHER;
1434 		break;
1435 	case LINUX_SCHED_FIFO:
1436 		bsd.policy = SCHED_FIFO;
1437 		break;
1438 	case LINUX_SCHED_RR:
1439 		bsd.policy = SCHED_RR;
1440 		break;
1441 	default:
1442 		return EINVAL;
1443 	}
1444 	return sched_get_priority_min(td, &bsd);
1445 }
1446 
1447 #define REBOOT_CAD_ON	0x89abcdef
1448 #define REBOOT_CAD_OFF	0
1449 #define REBOOT_HALT	0xcdef0123
1450 #define REBOOT_RESTART	0x01234567
1451 #define REBOOT_RESTART2	0xA1B2C3D4
1452 #define REBOOT_POWEROFF	0x4321FEDC
1453 #define REBOOT_MAGIC1	0xfee1dead
1454 #define REBOOT_MAGIC2	0x28121969
1455 #define REBOOT_MAGIC2A	0x05121996
1456 #define REBOOT_MAGIC2B	0x16041998
1457 
1458 int
1459 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1460 {
1461 	struct reboot_args bsd_args;
1462 
1463 #ifdef DEBUG
1464 	if (ldebug(reboot))
1465 		printf(ARGS(reboot, "0x%x"), args->cmd);
1466 #endif
1467 
1468 	if (args->magic1 != REBOOT_MAGIC1)
1469 		return EINVAL;
1470 
1471 	switch (args->magic2) {
1472 	case REBOOT_MAGIC2:
1473 	case REBOOT_MAGIC2A:
1474 	case REBOOT_MAGIC2B:
1475 		break;
1476 	default:
1477 		return EINVAL;
1478 	}
1479 
1480 	switch (args->cmd) {
1481 	case REBOOT_CAD_ON:
1482 	case REBOOT_CAD_OFF:
1483 		return (priv_check(td, PRIV_REBOOT));
1484 	case REBOOT_HALT:
1485 		bsd_args.opt = RB_HALT;
1486 		break;
1487 	case REBOOT_RESTART:
1488 	case REBOOT_RESTART2:
1489 		bsd_args.opt = 0;
1490 		break;
1491 	case REBOOT_POWEROFF:
1492 		bsd_args.opt = RB_POWEROFF;
1493 		break;
1494 	default:
1495 		return EINVAL;
1496 	}
1497 	return reboot(td, &bsd_args);
1498 }
1499 
1500 
1501 /*
1502  * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1503  * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
1504  * are assumed to be preserved. The following lightweight syscalls fixes
1505  * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
1506  *
1507  * linux_getpid() - MP SAFE
1508  * linux_getgid() - MP SAFE
1509  * linux_getuid() - MP SAFE
1510  */
1511 
1512 int
1513 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1514 {
1515 	struct linux_emuldata *em;
1516 
1517 #ifdef DEBUG
1518 	if (ldebug(getpid))
1519 		printf(ARGS(getpid, ""));
1520 #endif
1521 
1522 	if (linux_use26(td)) {
1523 		em = em_find(td->td_proc, EMUL_DONTLOCK);
1524 		KASSERT(em != NULL, ("getpid: emuldata not found.\n"));
1525 		td->td_retval[0] = em->shared->group_pid;
1526 	} else {
1527 		td->td_retval[0] = td->td_proc->p_pid;
1528 	}
1529 
1530 	return (0);
1531 }
1532 
1533 int
1534 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1535 {
1536 
1537 #ifdef DEBUG
1538 	if (ldebug(gettid))
1539 		printf(ARGS(gettid, ""));
1540 #endif
1541 
1542 	td->td_retval[0] = td->td_proc->p_pid;
1543 	return (0);
1544 }
1545 
1546 
1547 int
1548 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1549 {
1550 	struct linux_emuldata *em;
1551 	struct proc *p, *pp;
1552 
1553 #ifdef DEBUG
1554 	if (ldebug(getppid))
1555 		printf(ARGS(getppid, ""));
1556 #endif
1557 
1558 	if (!linux_use26(td)) {
1559 		PROC_LOCK(td->td_proc);
1560 		td->td_retval[0] = td->td_proc->p_pptr->p_pid;
1561 		PROC_UNLOCK(td->td_proc);
1562 		return (0);
1563 	}
1564 
1565 	em = em_find(td->td_proc, EMUL_DONTLOCK);
1566 
1567 	KASSERT(em != NULL, ("getppid: process emuldata not found.\n"));
1568 
1569 	/* find the group leader */
1570 	p = pfind(em->shared->group_pid);
1571 
1572 	if (p == NULL) {
1573 #ifdef DEBUG
1574 	   	printf(LMSG("parent process not found.\n"));
1575 #endif
1576 		return (0);
1577 	}
1578 
1579 	pp = p->p_pptr;		/* switch to parent */
1580 	PROC_LOCK(pp);
1581 	PROC_UNLOCK(p);
1582 
1583 	/* if its also linux process */
1584 	if (pp->p_sysent == &elf_linux_sysvec) {
1585 		em = em_find(pp, EMUL_DONTLOCK);
1586 		KASSERT(em != NULL, ("getppid: parent emuldata not found.\n"));
1587 
1588 		td->td_retval[0] = em->shared->group_pid;
1589 	} else
1590 		td->td_retval[0] = pp->p_pid;
1591 
1592 	PROC_UNLOCK(pp);
1593 
1594 	return (0);
1595 }
1596 
1597 int
1598 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1599 {
1600 
1601 #ifdef DEBUG
1602 	if (ldebug(getgid))
1603 		printf(ARGS(getgid, ""));
1604 #endif
1605 
1606 	td->td_retval[0] = td->td_ucred->cr_rgid;
1607 	return (0);
1608 }
1609 
1610 int
1611 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1612 {
1613 
1614 #ifdef DEBUG
1615 	if (ldebug(getuid))
1616 		printf(ARGS(getuid, ""));
1617 #endif
1618 
1619 	td->td_retval[0] = td->td_ucred->cr_ruid;
1620 	return (0);
1621 }
1622 
1623 
1624 int
1625 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1626 {
1627 	struct getsid_args bsd;
1628 
1629 #ifdef DEBUG
1630 	if (ldebug(getsid))
1631 		printf(ARGS(getsid, "%i"), args->pid);
1632 #endif
1633 
1634 	bsd.pid = args->pid;
1635 	return getsid(td, &bsd);
1636 }
1637 
1638 int
1639 linux_nosys(struct thread *td, struct nosys_args *ignore)
1640 {
1641 
1642 	return (ENOSYS);
1643 }
1644 
1645 int
1646 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1647 {
1648 	struct getpriority_args bsd_args;
1649 	int error;
1650 
1651 #ifdef DEBUG
1652 	if (ldebug(getpriority))
1653 		printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
1654 #endif
1655 
1656 	bsd_args.which = args->which;
1657 	bsd_args.who = args->who;
1658 	error = getpriority(td, &bsd_args);
1659 	td->td_retval[0] = 20 - td->td_retval[0];
1660 	return error;
1661 }
1662 
1663 int
1664 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1665 {
1666 	int name[2];
1667 
1668 #ifdef DEBUG
1669 	if (ldebug(sethostname))
1670 		printf(ARGS(sethostname, "*, %i"), args->len);
1671 #endif
1672 
1673 	name[0] = CTL_KERN;
1674 	name[1] = KERN_HOSTNAME;
1675 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1676 	    args->len, 0, 0));
1677 }
1678 
1679 int
1680 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1681 {
1682 	int name[2];
1683 
1684 #ifdef DEBUG
1685 	if (ldebug(setdomainname))
1686 		printf(ARGS(setdomainname, "*, %i"), args->len);
1687 #endif
1688 
1689 	name[0] = CTL_KERN;
1690 	name[1] = KERN_NISDOMAINNAME;
1691 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1692 	    args->len, 0, 0));
1693 }
1694 
1695 int
1696 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1697 {
1698 	struct linux_emuldata *em, *td_em, *tmp_em;
1699 	struct proc *sp;
1700 
1701 #ifdef DEBUG
1702 	if (ldebug(exit_group))
1703 		printf(ARGS(exit_group, "%i"), args->error_code);
1704 #endif
1705 
1706 	if (linux_use26(td)) {
1707 		td_em = em_find(td->td_proc, EMUL_DONTLOCK);
1708 
1709 		KASSERT(td_em != NULL, ("exit_group: emuldata not found.\n"));
1710 
1711 		EMUL_SHARED_RLOCK(&emul_shared_lock);
1712 		LIST_FOREACH_SAFE(em, &td_em->shared->threads, threads, tmp_em) {
1713 			if (em->pid == td_em->pid)
1714 				continue;
1715 
1716 			sp = pfind(em->pid);
1717 			psignal(sp, SIGKILL);
1718 			PROC_UNLOCK(sp);
1719 #ifdef DEBUG
1720 			printf(LMSG("linux_sys_exit_group: kill PID %d\n"), em->pid);
1721 #endif
1722 		}
1723 
1724 		EMUL_SHARED_RUNLOCK(&emul_shared_lock);
1725 	}
1726 	/*
1727 	 * XXX: we should send a signal to the parent if
1728 	 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1729 	 * as it doesnt occur often.
1730 	 */
1731 	exit1(td, W_EXITCODE(args->error_code, 0));
1732 
1733 	return (0);
1734 }
1735 
1736 int
1737 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1738 {
1739 	int error = 0, max_size;
1740 	struct proc *p = td->td_proc;
1741 	char comm[LINUX_MAX_COMM_LEN];
1742 	struct linux_emuldata *em;
1743 	int pdeath_signal;
1744 
1745 #ifdef DEBUG
1746 	if (ldebug(prctl))
1747 		printf(ARGS(prctl, "%d, %d, %d, %d, %d"), args->option,
1748 		    args->arg2, args->arg3, args->arg4, args->arg5);
1749 #endif
1750 
1751 	switch (args->option) {
1752 	case LINUX_PR_SET_PDEATHSIG:
1753 		if (!LINUX_SIG_VALID(args->arg2))
1754 			return (EINVAL);
1755 		em = em_find(p, EMUL_DOLOCK);
1756 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1757 		em->pdeath_signal = args->arg2;
1758 		EMUL_UNLOCK(&emul_lock);
1759 		break;
1760 	case LINUX_PR_GET_PDEATHSIG:
1761 		em = em_find(p, EMUL_DOLOCK);
1762 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1763 		pdeath_signal = em->pdeath_signal;
1764 		EMUL_UNLOCK(&emul_lock);
1765 		error = copyout(&pdeath_signal,
1766 		    (void *)(register_t)args->arg2,
1767 		    sizeof(pdeath_signal));
1768 		break;
1769 	case LINUX_PR_SET_NAME:
1770 		/*
1771 		 * To be on the safe side we need to make sure to not
1772 		 * overflow the size a linux program expects. We already
1773 		 * do this here in the copyin, so that we don't need to
1774 		 * check on copyout.
1775 		 */
1776 		max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1777 		error = copyinstr((void *)(register_t)args->arg2, comm,
1778 		    max_size, NULL);
1779 
1780 		/* Linux silently truncates the name if it is too long. */
1781 		if (error == ENAMETOOLONG) {
1782 			/*
1783 			 * XXX: copyinstr() isn't documented to populate the
1784 			 * array completely, so do a copyin() to be on the
1785 			 * safe side. This should be changed in case
1786 			 * copyinstr() is changed to guarantee this.
1787 			 */
1788 			error = copyin((void *)(register_t)args->arg2, comm,
1789 			    max_size - 1);
1790 			comm[max_size - 1] = '\0';
1791 		}
1792 		if (error)
1793 			return (error);
1794 
1795 		PROC_LOCK(p);
1796 		strlcpy(p->p_comm, comm, sizeof(p->p_comm));
1797 		PROC_UNLOCK(p);
1798 		break;
1799 	case LINUX_PR_GET_NAME:
1800 		PROC_LOCK(p);
1801 		strlcpy(comm, p->p_comm, sizeof(comm));
1802 		PROC_UNLOCK(p);
1803 		error = copyout(comm, (void *)(register_t)args->arg2,
1804 		    strlen(comm) + 1);
1805 		break;
1806 	default:
1807 		error = EINVAL;
1808 		break;
1809 	}
1810 
1811 	return (error);
1812 }
1813 
1814 /*
1815  * Get affinity of a process.
1816  */
1817 int
1818 linux_sched_getaffinity(struct thread *td,
1819     struct linux_sched_getaffinity_args *args)
1820 {
1821 	int error;
1822 	struct cpuset_getaffinity_args cga;
1823 
1824 #ifdef DEBUG
1825 	if (ldebug(sched_getaffinity))
1826 		printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid,
1827 		    args->len);
1828 #endif
1829 	if (args->len < sizeof(cpuset_t))
1830 		return (EINVAL);
1831 
1832 	cga.level = CPU_LEVEL_WHICH;
1833 	cga.which = CPU_WHICH_PID;
1834 	cga.id = args->pid;
1835 	cga.cpusetsize = sizeof(cpuset_t);
1836 	cga.mask = (cpuset_t *) args->user_mask_ptr;
1837 
1838 	if ((error = cpuset_getaffinity(td, &cga)) == 0)
1839 		td->td_retval[0] = sizeof(cpuset_t);
1840 
1841 	return (error);
1842 }
1843 
1844 /*
1845  *  Set affinity of a process.
1846  */
1847 int
1848 linux_sched_setaffinity(struct thread *td,
1849     struct linux_sched_setaffinity_args *args)
1850 {
1851 	struct cpuset_setaffinity_args csa;
1852 
1853 #ifdef DEBUG
1854 	if (ldebug(sched_setaffinity))
1855 		printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid,
1856 		    args->len);
1857 #endif
1858 	if (args->len < sizeof(cpuset_t))
1859 		return (EINVAL);
1860 
1861 	csa.level = CPU_LEVEL_WHICH;
1862 	csa.which = CPU_WHICH_PID;
1863 	csa.id = args->pid;
1864 	csa.cpusetsize = sizeof(cpuset_t);
1865 	csa.mask = (cpuset_t *) args->user_mask_ptr;
1866 
1867 	return (cpuset_setaffinity(td, &csa));
1868 }
1869