xref: /freebsd/sys/compat/linux/linux_misc.c (revision 34e05ebe7281cdb7f4378339ab183e33780d801f)
1 /*-
2  * Copyright (c) 2002 Doug Rabson
3  * Copyright (c) 1994-1995 Søren 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/racct.h>
54 #include <sys/resourcevar.h>
55 #include <sys/sched.h>
56 #include <sys/sdt.h>
57 #include <sys/signalvar.h>
58 #include <sys/stat.h>
59 #include <sys/syscallsubr.h>
60 #include <sys/sysctl.h>
61 #include <sys/sysproto.h>
62 #include <sys/systm.h>
63 #include <sys/time.h>
64 #include <sys/vmmeter.h>
65 #include <sys/vnode.h>
66 #include <sys/wait.h>
67 #include <sys/cpuset.h>
68 
69 #include <security/mac/mac_framework.h>
70 
71 #include <vm/vm.h>
72 #include <vm/pmap.h>
73 #include <vm/vm_kern.h>
74 #include <vm/vm_map.h>
75 #include <vm/vm_extern.h>
76 #include <vm/vm_object.h>
77 #include <vm/swap_pager.h>
78 
79 #ifdef COMPAT_LINUX32
80 #include <machine/../linux32/linux.h>
81 #include <machine/../linux32/linux32_proto.h>
82 #else
83 #include <machine/../linux/linux.h>
84 #include <machine/../linux/linux_proto.h>
85 #endif
86 
87 #include <compat/linux/linux_dtrace.h>
88 #include <compat/linux/linux_file.h>
89 #include <compat/linux/linux_mib.h>
90 #include <compat/linux/linux_signal.h>
91 #include <compat/linux/linux_timer.h>
92 #include <compat/linux/linux_util.h>
93 #include <compat/linux/linux_sysproto.h>
94 #include <compat/linux/linux_emul.h>
95 #include <compat/linux/linux_misc.h>
96 
97 /**
98  * Special DTrace provider for the linuxulator.
99  *
100  * In this file we define the provider for the entire linuxulator. All
101  * modules (= files of the linuxulator) use it.
102  *
103  * We define a different name depending on the emulated bitsize, see
104  * ../../<ARCH>/linux{,32}/linux.h, e.g.:
105  *      native bitsize          = linuxulator
106  *      amd64, 32bit emulation  = linuxulator32
107  */
108 LIN_SDT_PROVIDER_DEFINE(LINUX_DTRACE);
109 
110 int stclohz;				/* Statistics clock frequency */
111 
112 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
113 	RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
114 	RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
115 	RLIMIT_MEMLOCK, RLIMIT_AS
116 };
117 
118 struct l_sysinfo {
119 	l_long		uptime;		/* Seconds since boot */
120 	l_ulong		loads[3];	/* 1, 5, and 15 minute load averages */
121 #define LINUX_SYSINFO_LOADS_SCALE 65536
122 	l_ulong		totalram;	/* Total usable main memory size */
123 	l_ulong		freeram;	/* Available memory size */
124 	l_ulong		sharedram;	/* Amount of shared memory */
125 	l_ulong		bufferram;	/* Memory used by buffers */
126 	l_ulong		totalswap;	/* Total swap space size */
127 	l_ulong		freeswap;	/* swap space still available */
128 	l_ushort	procs;		/* Number of current processes */
129 	l_ushort	pads;
130 	l_ulong		totalbig;
131 	l_ulong		freebig;
132 	l_uint		mem_unit;
133 	char		_f[20-2*sizeof(l_long)-sizeof(l_int)];	/* padding */
134 };
135 
136 struct l_pselect6arg {
137 	l_uintptr_t	ss;
138 	l_size_t	ss_len;
139 };
140 
141 static int	linux_utimensat_nsec_valid(l_long);
142 
143 
144 int
145 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
146 {
147 	struct l_sysinfo sysinfo;
148 	vm_object_t object;
149 	int i, j;
150 	struct timespec ts;
151 
152 	bzero(&sysinfo, sizeof(sysinfo));
153 	getnanouptime(&ts);
154 	if (ts.tv_nsec != 0)
155 		ts.tv_sec++;
156 	sysinfo.uptime = ts.tv_sec;
157 
158 	/* Use the information from the mib to get our load averages */
159 	for (i = 0; i < 3; i++)
160 		sysinfo.loads[i] = averunnable.ldavg[i] *
161 		    LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
162 
163 	sysinfo.totalram = physmem * PAGE_SIZE;
164 	sysinfo.freeram = sysinfo.totalram - vm_cnt.v_wire_count * PAGE_SIZE;
165 
166 	sysinfo.sharedram = 0;
167 	mtx_lock(&vm_object_list_mtx);
168 	TAILQ_FOREACH(object, &vm_object_list, object_list)
169 		if (object->shadow_count > 1)
170 			sysinfo.sharedram += object->resident_page_count;
171 	mtx_unlock(&vm_object_list_mtx);
172 
173 	sysinfo.sharedram *= PAGE_SIZE;
174 	sysinfo.bufferram = 0;
175 
176 	swap_pager_status(&i, &j);
177 	sysinfo.totalswap = i * PAGE_SIZE;
178 	sysinfo.freeswap = (i - j) * PAGE_SIZE;
179 
180 	sysinfo.procs = nprocs;
181 
182 	/* The following are only present in newer Linux kernels. */
183 	sysinfo.totalbig = 0;
184 	sysinfo.freebig = 0;
185 	sysinfo.mem_unit = 1;
186 
187 	return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
188 }
189 
190 int
191 linux_alarm(struct thread *td, struct linux_alarm_args *args)
192 {
193 	struct itimerval it, old_it;
194 	u_int secs;
195 	int error;
196 
197 #ifdef DEBUG
198 	if (ldebug(alarm))
199 		printf(ARGS(alarm, "%u"), args->secs);
200 #endif
201 	secs = args->secs;
202 	/*
203 	 * Linux alarm() is always successful. Limit secs to INT32_MAX / 2
204 	 * to match kern_setitimer()'s limit to avoid error from it.
205 	 *
206 	 * XXX. Linux limit secs to INT_MAX on 32 and does not limit on 64-bit
207 	 * platforms.
208 	 */
209 	if (secs > INT32_MAX / 2)
210 		secs = INT32_MAX / 2;
211 
212 	it.it_value.tv_sec = secs;
213 	it.it_value.tv_usec = 0;
214 	timevalclear(&it.it_interval);
215 	error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
216 	KASSERT(error == 0, ("kern_setitimer returns %d", error));
217 
218 	if ((old_it.it_value.tv_sec == 0 && old_it.it_value.tv_usec > 0) ||
219 	    old_it.it_value.tv_usec >= 500000)
220 		old_it.it_value.tv_sec++;
221 	td->td_retval[0] = old_it.it_value.tv_sec;
222 	return (0);
223 }
224 
225 int
226 linux_brk(struct thread *td, struct linux_brk_args *args)
227 {
228 	struct vmspace *vm = td->td_proc->p_vmspace;
229 	vm_offset_t new, old;
230 	struct obreak_args /* {
231 		char * nsize;
232 	} */ tmp;
233 
234 #ifdef DEBUG
235 	if (ldebug(brk))
236 		printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
237 #endif
238 	old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize);
239 	new = (vm_offset_t)args->dsend;
240 	tmp.nsize = (char *)new;
241 	if (((caddr_t)new > vm->vm_daddr) && !sys_obreak(td, &tmp))
242 		td->td_retval[0] = (long)new;
243 	else
244 		td->td_retval[0] = (long)old;
245 
246 	return (0);
247 }
248 
249 #if defined(__i386__)
250 /* XXX: what about amd64/linux32? */
251 
252 int
253 linux_uselib(struct thread *td, struct linux_uselib_args *args)
254 {
255 	struct nameidata ni;
256 	struct vnode *vp;
257 	struct exec *a_out;
258 	struct vattr attr;
259 	vm_offset_t vmaddr;
260 	unsigned long file_offset;
261 	unsigned long bss_size;
262 	char *library;
263 	ssize_t aresid;
264 	int error, locked, writecount;
265 
266 	LCONVPATHEXIST(td, args->library, &library);
267 
268 #ifdef DEBUG
269 	if (ldebug(uselib))
270 		printf(ARGS(uselib, "%s"), library);
271 #endif
272 
273 	a_out = NULL;
274 	locked = 0;
275 	vp = NULL;
276 
277 	NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
278 	    UIO_SYSSPACE, library, td);
279 	error = namei(&ni);
280 	LFREEPATH(library);
281 	if (error)
282 		goto cleanup;
283 
284 	vp = ni.ni_vp;
285 	NDFREE(&ni, NDF_ONLY_PNBUF);
286 
287 	/*
288 	 * From here on down, we have a locked vnode that must be unlocked.
289 	 * XXX: The code below largely duplicates exec_check_permissions().
290 	 */
291 	locked = 1;
292 
293 	/* Writable? */
294 	error = VOP_GET_WRITECOUNT(vp, &writecount);
295 	if (error != 0)
296 		goto cleanup;
297 	if (writecount != 0) {
298 		error = ETXTBSY;
299 		goto cleanup;
300 	}
301 
302 	/* Executable? */
303 	error = VOP_GETATTR(vp, &attr, td->td_ucred);
304 	if (error)
305 		goto cleanup;
306 
307 	if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
308 	    ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
309 		/* EACCESS is what exec(2) returns. */
310 		error = ENOEXEC;
311 		goto cleanup;
312 	}
313 
314 	/* Sensible size? */
315 	if (attr.va_size == 0) {
316 		error = ENOEXEC;
317 		goto cleanup;
318 	}
319 
320 	/* Can we access it? */
321 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
322 	if (error)
323 		goto cleanup;
324 
325 	/*
326 	 * XXX: This should use vn_open() so that it is properly authorized,
327 	 * and to reduce code redundancy all over the place here.
328 	 * XXX: Not really, it duplicates far more of exec_check_permissions()
329 	 * than vn_open().
330 	 */
331 #ifdef MAC
332 	error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
333 	if (error)
334 		goto cleanup;
335 #endif
336 	error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
337 	if (error)
338 		goto cleanup;
339 
340 	/* Pull in executable header into exec_map */
341 	error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
342 	    VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
343 	if (error)
344 		goto cleanup;
345 
346 	/* Is it a Linux binary ? */
347 	if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
348 		error = ENOEXEC;
349 		goto cleanup;
350 	}
351 
352 	/*
353 	 * While we are here, we should REALLY do some more checks
354 	 */
355 
356 	/* Set file/virtual offset based on a.out variant. */
357 	switch ((int)(a_out->a_magic & 0xffff)) {
358 	case 0413:			/* ZMAGIC */
359 		file_offset = 1024;
360 		break;
361 	case 0314:			/* QMAGIC */
362 		file_offset = 0;
363 		break;
364 	default:
365 		error = ENOEXEC;
366 		goto cleanup;
367 	}
368 
369 	bss_size = round_page(a_out->a_bss);
370 
371 	/* Check various fields in header for validity/bounds. */
372 	if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
373 		error = ENOEXEC;
374 		goto cleanup;
375 	}
376 
377 	/* text + data can't exceed file size */
378 	if (a_out->a_data + a_out->a_text > attr.va_size) {
379 		error = EFAULT;
380 		goto cleanup;
381 	}
382 
383 	/*
384 	 * text/data/bss must not exceed limits
385 	 * XXX - this is not complete. it should check current usage PLUS
386 	 * the resources needed by this library.
387 	 */
388 	PROC_LOCK(td->td_proc);
389 	if (a_out->a_text > maxtsiz ||
390 	    a_out->a_data + bss_size > lim_cur_proc(td->td_proc, RLIMIT_DATA) ||
391 	    racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
392 	    bss_size) != 0) {
393 		PROC_UNLOCK(td->td_proc);
394 		error = ENOMEM;
395 		goto cleanup;
396 	}
397 	PROC_UNLOCK(td->td_proc);
398 
399 	/*
400 	 * Prevent more writers.
401 	 * XXX: Note that if any of the VM operations fail below we don't
402 	 * clear this flag.
403 	 */
404 	VOP_SET_TEXT(vp);
405 
406 	/*
407 	 * Lock no longer needed
408 	 */
409 	locked = 0;
410 	VOP_UNLOCK(vp, 0);
411 
412 	/*
413 	 * Check if file_offset page aligned. Currently we cannot handle
414 	 * misalinged file offsets, and so we read in the entire image
415 	 * (what a waste).
416 	 */
417 	if (file_offset & PAGE_MASK) {
418 #ifdef DEBUG
419 		printf("uselib: Non page aligned binary %lu\n", file_offset);
420 #endif
421 		/* Map text+data read/write/execute */
422 
423 		/* a_entry is the load address and is page aligned */
424 		vmaddr = trunc_page(a_out->a_entry);
425 
426 		/* get anon user mapping, read+write+execute */
427 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
428 		    &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
429 		    VM_PROT_ALL, VM_PROT_ALL, 0);
430 		if (error)
431 			goto cleanup;
432 
433 		error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
434 		    a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
435 		    td->td_ucred, NOCRED, &aresid, td);
436 		if (error != 0)
437 			goto cleanup;
438 		if (aresid != 0) {
439 			error = ENOEXEC;
440 			goto cleanup;
441 		}
442 	} else {
443 #ifdef DEBUG
444 		printf("uselib: Page aligned binary %lu\n", file_offset);
445 #endif
446 		/*
447 		 * for QMAGIC, a_entry is 20 bytes beyond the load address
448 		 * to skip the executable header
449 		 */
450 		vmaddr = trunc_page(a_out->a_entry);
451 
452 		/*
453 		 * Map it all into the process's space as a single
454 		 * copy-on-write "data" segment.
455 		 */
456 		error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr,
457 		    a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
458 		    MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
459 		if (error)
460 			goto cleanup;
461 	}
462 #ifdef DEBUG
463 	printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
464 	    ((long *)vmaddr)[1]);
465 #endif
466 	if (bss_size != 0) {
467 		/* Calculate BSS start address */
468 		vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
469 		    a_out->a_data;
470 
471 		/* allocate some 'anon' space */
472 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
473 		    &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
474 		    VM_PROT_ALL, 0);
475 		if (error)
476 			goto cleanup;
477 	}
478 
479 cleanup:
480 	/* Unlock vnode if needed */
481 	if (locked)
482 		VOP_UNLOCK(vp, 0);
483 
484 	/* Release the temporary mapping. */
485 	if (a_out)
486 		kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
487 
488 	return (error);
489 }
490 
491 #endif	/* __i386__ */
492 
493 int
494 linux_select(struct thread *td, struct linux_select_args *args)
495 {
496 	l_timeval ltv;
497 	struct timeval tv0, tv1, utv, *tvp;
498 	int error;
499 
500 #ifdef DEBUG
501 	if (ldebug(select))
502 		printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
503 		    (void *)args->readfds, (void *)args->writefds,
504 		    (void *)args->exceptfds, (void *)args->timeout);
505 #endif
506 
507 	/*
508 	 * Store current time for computation of the amount of
509 	 * time left.
510 	 */
511 	if (args->timeout) {
512 		if ((error = copyin(args->timeout, &ltv, sizeof(ltv))))
513 			goto select_out;
514 		utv.tv_sec = ltv.tv_sec;
515 		utv.tv_usec = ltv.tv_usec;
516 #ifdef DEBUG
517 		if (ldebug(select))
518 			printf(LMSG("incoming timeout (%jd/%ld)"),
519 			    (intmax_t)utv.tv_sec, utv.tv_usec);
520 #endif
521 
522 		if (itimerfix(&utv)) {
523 			/*
524 			 * The timeval was invalid.  Convert it to something
525 			 * valid that will act as it does under Linux.
526 			 */
527 			utv.tv_sec += utv.tv_usec / 1000000;
528 			utv.tv_usec %= 1000000;
529 			if (utv.tv_usec < 0) {
530 				utv.tv_sec -= 1;
531 				utv.tv_usec += 1000000;
532 			}
533 			if (utv.tv_sec < 0)
534 				timevalclear(&utv);
535 		}
536 		microtime(&tv0);
537 		tvp = &utv;
538 	} else
539 		tvp = NULL;
540 
541 	error = kern_select(td, args->nfds, args->readfds, args->writefds,
542 	    args->exceptfds, tvp, LINUX_NFDBITS);
543 
544 #ifdef DEBUG
545 	if (ldebug(select))
546 		printf(LMSG("real select returns %d"), error);
547 #endif
548 	if (error)
549 		goto select_out;
550 
551 	if (args->timeout) {
552 		if (td->td_retval[0]) {
553 			/*
554 			 * Compute how much time was left of the timeout,
555 			 * by subtracting the current time and the time
556 			 * before we started the call, and subtracting
557 			 * that result from the user-supplied value.
558 			 */
559 			microtime(&tv1);
560 			timevalsub(&tv1, &tv0);
561 			timevalsub(&utv, &tv1);
562 			if (utv.tv_sec < 0)
563 				timevalclear(&utv);
564 		} else
565 			timevalclear(&utv);
566 #ifdef DEBUG
567 		if (ldebug(select))
568 			printf(LMSG("outgoing timeout (%jd/%ld)"),
569 			    (intmax_t)utv.tv_sec, utv.tv_usec);
570 #endif
571 		ltv.tv_sec = utv.tv_sec;
572 		ltv.tv_usec = utv.tv_usec;
573 		if ((error = copyout(&ltv, args->timeout, sizeof(ltv))))
574 			goto select_out;
575 	}
576 
577 select_out:
578 #ifdef DEBUG
579 	if (ldebug(select))
580 		printf(LMSG("select_out -> %d"), error);
581 #endif
582 	return (error);
583 }
584 
585 int
586 linux_mremap(struct thread *td, struct linux_mremap_args *args)
587 {
588 	struct munmap_args /* {
589 		void *addr;
590 		size_t len;
591 	} */ bsd_args;
592 	int error = 0;
593 
594 #ifdef DEBUG
595 	if (ldebug(mremap))
596 		printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
597 		    (void *)(uintptr_t)args->addr,
598 		    (unsigned long)args->old_len,
599 		    (unsigned long)args->new_len,
600 		    (unsigned long)args->flags);
601 #endif
602 
603 	if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
604 		td->td_retval[0] = 0;
605 		return (EINVAL);
606 	}
607 
608 	/*
609 	 * Check for the page alignment.
610 	 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
611 	 */
612 	if (args->addr & PAGE_MASK) {
613 		td->td_retval[0] = 0;
614 		return (EINVAL);
615 	}
616 
617 	args->new_len = round_page(args->new_len);
618 	args->old_len = round_page(args->old_len);
619 
620 	if (args->new_len > args->old_len) {
621 		td->td_retval[0] = 0;
622 		return (ENOMEM);
623 	}
624 
625 	if (args->new_len < args->old_len) {
626 		bsd_args.addr =
627 		    (caddr_t)((uintptr_t)args->addr + args->new_len);
628 		bsd_args.len = args->old_len - args->new_len;
629 		error = sys_munmap(td, &bsd_args);
630 	}
631 
632 	td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
633 	return (error);
634 }
635 
636 #define LINUX_MS_ASYNC       0x0001
637 #define LINUX_MS_INVALIDATE  0x0002
638 #define LINUX_MS_SYNC        0x0004
639 
640 int
641 linux_msync(struct thread *td, struct linux_msync_args *args)
642 {
643 	struct msync_args bsd_args;
644 
645 	bsd_args.addr = (caddr_t)(uintptr_t)args->addr;
646 	bsd_args.len = (uintptr_t)args->len;
647 	bsd_args.flags = args->fl & ~LINUX_MS_SYNC;
648 
649 	return (sys_msync(td, &bsd_args));
650 }
651 
652 int
653 linux_time(struct thread *td, struct linux_time_args *args)
654 {
655 	struct timeval tv;
656 	l_time_t tm;
657 	int error;
658 
659 #ifdef DEBUG
660 	if (ldebug(time))
661 		printf(ARGS(time, "*"));
662 #endif
663 
664 	microtime(&tv);
665 	tm = tv.tv_sec;
666 	if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
667 		return (error);
668 	td->td_retval[0] = tm;
669 	return (0);
670 }
671 
672 struct l_times_argv {
673 	l_clock_t	tms_utime;
674 	l_clock_t	tms_stime;
675 	l_clock_t	tms_cutime;
676 	l_clock_t	tms_cstime;
677 };
678 
679 
680 /*
681  * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
682  * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
683  * auxiliary vector entry.
684  */
685 #define	CLK_TCK		100
686 
687 #define	CONVOTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
688 #define	CONVNTCK(r)	(r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
689 
690 #define	CONVTCK(r)	(linux_kernver(td) >= LINUX_KERNVER_2004000 ?		\
691 			    CONVNTCK(r) : CONVOTCK(r))
692 
693 int
694 linux_times(struct thread *td, struct linux_times_args *args)
695 {
696 	struct timeval tv, utime, stime, cutime, cstime;
697 	struct l_times_argv tms;
698 	struct proc *p;
699 	int error;
700 
701 #ifdef DEBUG
702 	if (ldebug(times))
703 		printf(ARGS(times, "*"));
704 #endif
705 
706 	if (args->buf != NULL) {
707 		p = td->td_proc;
708 		PROC_LOCK(p);
709 		PROC_STATLOCK(p);
710 		calcru(p, &utime, &stime);
711 		PROC_STATUNLOCK(p);
712 		calccru(p, &cutime, &cstime);
713 		PROC_UNLOCK(p);
714 
715 		tms.tms_utime = CONVTCK(utime);
716 		tms.tms_stime = CONVTCK(stime);
717 
718 		tms.tms_cutime = CONVTCK(cutime);
719 		tms.tms_cstime = CONVTCK(cstime);
720 
721 		if ((error = copyout(&tms, args->buf, sizeof(tms))))
722 			return (error);
723 	}
724 
725 	microuptime(&tv);
726 	td->td_retval[0] = (int)CONVTCK(tv);
727 	return (0);
728 }
729 
730 int
731 linux_newuname(struct thread *td, struct linux_newuname_args *args)
732 {
733 	struct l_new_utsname utsname;
734 	char osname[LINUX_MAX_UTSNAME];
735 	char osrelease[LINUX_MAX_UTSNAME];
736 	char *p;
737 
738 #ifdef DEBUG
739 	if (ldebug(newuname))
740 		printf(ARGS(newuname, "*"));
741 #endif
742 
743 	linux_get_osname(td, osname);
744 	linux_get_osrelease(td, osrelease);
745 
746 	bzero(&utsname, sizeof(utsname));
747 	strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
748 	getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
749 	getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
750 	strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
751 	strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
752 	for (p = utsname.version; *p != '\0'; ++p)
753 		if (*p == '\n') {
754 			*p = '\0';
755 			break;
756 		}
757 	strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME);
758 
759 	return (copyout(&utsname, args->buf, sizeof(utsname)));
760 }
761 
762 struct l_utimbuf {
763 	l_time_t l_actime;
764 	l_time_t l_modtime;
765 };
766 
767 int
768 linux_utime(struct thread *td, struct linux_utime_args *args)
769 {
770 	struct timeval tv[2], *tvp;
771 	struct l_utimbuf lut;
772 	char *fname;
773 	int error;
774 
775 	LCONVPATHEXIST(td, args->fname, &fname);
776 
777 #ifdef DEBUG
778 	if (ldebug(utime))
779 		printf(ARGS(utime, "%s, *"), fname);
780 #endif
781 
782 	if (args->times) {
783 		if ((error = copyin(args->times, &lut, sizeof lut))) {
784 			LFREEPATH(fname);
785 			return (error);
786 		}
787 		tv[0].tv_sec = lut.l_actime;
788 		tv[0].tv_usec = 0;
789 		tv[1].tv_sec = lut.l_modtime;
790 		tv[1].tv_usec = 0;
791 		tvp = tv;
792 	} else
793 		tvp = NULL;
794 
795 	error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, tvp,
796 	    UIO_SYSSPACE);
797 	LFREEPATH(fname);
798 	return (error);
799 }
800 
801 int
802 linux_utimes(struct thread *td, struct linux_utimes_args *args)
803 {
804 	l_timeval ltv[2];
805 	struct timeval tv[2], *tvp = NULL;
806 	char *fname;
807 	int error;
808 
809 	LCONVPATHEXIST(td, args->fname, &fname);
810 
811 #ifdef DEBUG
812 	if (ldebug(utimes))
813 		printf(ARGS(utimes, "%s, *"), fname);
814 #endif
815 
816 	if (args->tptr != NULL) {
817 		if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
818 			LFREEPATH(fname);
819 			return (error);
820 		}
821 		tv[0].tv_sec = ltv[0].tv_sec;
822 		tv[0].tv_usec = ltv[0].tv_usec;
823 		tv[1].tv_sec = ltv[1].tv_sec;
824 		tv[1].tv_usec = ltv[1].tv_usec;
825 		tvp = tv;
826 	}
827 
828 	error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE,
829 	    tvp, UIO_SYSSPACE);
830 	LFREEPATH(fname);
831 	return (error);
832 }
833 
834 static int
835 linux_utimensat_nsec_valid(l_long nsec)
836 {
837 
838 	if (nsec == LINUX_UTIME_OMIT || nsec == LINUX_UTIME_NOW)
839 		return (0);
840 	if (nsec >= 0 && nsec <= 999999999)
841 		return (0);
842 	return (1);
843 }
844 
845 int
846 linux_utimensat(struct thread *td, struct linux_utimensat_args *args)
847 {
848 	struct l_timespec l_times[2];
849 	struct timespec times[2], *timesp = NULL;
850 	char *path = NULL;
851 	int error, dfd, flags = 0;
852 
853 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
854 
855 #ifdef DEBUG
856 	if (ldebug(utimensat))
857 		printf(ARGS(utimensat, "%d, *"), dfd);
858 #endif
859 
860 	if (args->flags & ~LINUX_AT_SYMLINK_NOFOLLOW)
861 		return (EINVAL);
862 
863 	if (args->times != NULL) {
864 		error = copyin(args->times, l_times, sizeof(l_times));
865 		if (error != 0)
866 			return (error);
867 
868 		if (linux_utimensat_nsec_valid(l_times[0].tv_nsec) != 0 ||
869 		    linux_utimensat_nsec_valid(l_times[1].tv_nsec) != 0)
870 			return (EINVAL);
871 
872 		times[0].tv_sec = l_times[0].tv_sec;
873 		switch (l_times[0].tv_nsec)
874 		{
875 		case LINUX_UTIME_OMIT:
876 			times[0].tv_nsec = UTIME_OMIT;
877 			break;
878 		case LINUX_UTIME_NOW:
879 			times[0].tv_nsec = UTIME_NOW;
880 			break;
881 		default:
882 			times[0].tv_nsec = l_times[0].tv_nsec;
883 		}
884 
885 		times[1].tv_sec = l_times[1].tv_sec;
886 		switch (l_times[1].tv_nsec)
887 		{
888 		case LINUX_UTIME_OMIT:
889 			times[1].tv_nsec = UTIME_OMIT;
890 			break;
891 		case LINUX_UTIME_NOW:
892 			times[1].tv_nsec = UTIME_NOW;
893 			break;
894 		default:
895 			times[1].tv_nsec = l_times[1].tv_nsec;
896 			break;
897 		}
898 		timesp = times;
899 
900 		/* This breaks POSIX, but is what the Linux kernel does
901 		 * _on purpose_ (documented in the man page for utimensat(2)),
902 		 * so we must follow that behaviour. */
903 		if (times[0].tv_nsec == UTIME_OMIT &&
904 		    times[1].tv_nsec == UTIME_OMIT)
905 			return (0);
906 	}
907 
908 	if (args->pathname != NULL)
909 		LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
910 	else if (args->flags != 0)
911 		return (EINVAL);
912 
913 	if (args->flags & LINUX_AT_SYMLINK_NOFOLLOW)
914 		flags |= AT_SYMLINK_NOFOLLOW;
915 
916 	if (path == NULL)
917 		error = kern_futimens(td, dfd, timesp, UIO_SYSSPACE);
918 	else {
919 		error = kern_utimensat(td, dfd, path, UIO_SYSSPACE, timesp,
920 	    		UIO_SYSSPACE, flags);
921 		LFREEPATH(path);
922 	}
923 
924 	return (error);
925 }
926 
927 int
928 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
929 {
930 	l_timeval ltv[2];
931 	struct timeval tv[2], *tvp = NULL;
932 	char *fname;
933 	int error, dfd;
934 
935 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
936 	LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
937 
938 #ifdef DEBUG
939 	if (ldebug(futimesat))
940 		printf(ARGS(futimesat, "%s, *"), fname);
941 #endif
942 
943 	if (args->utimes != NULL) {
944 		if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
945 			LFREEPATH(fname);
946 			return (error);
947 		}
948 		tv[0].tv_sec = ltv[0].tv_sec;
949 		tv[0].tv_usec = ltv[0].tv_usec;
950 		tv[1].tv_sec = ltv[1].tv_sec;
951 		tv[1].tv_usec = ltv[1].tv_usec;
952 		tvp = tv;
953 	}
954 
955 	error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
956 	LFREEPATH(fname);
957 	return (error);
958 }
959 
960 int
961 linux_common_wait(struct thread *td, int pid, int *status,
962     int options, struct rusage *ru)
963 {
964 	int error, tmpstat;
965 
966 	error = kern_wait(td, pid, &tmpstat, options, ru);
967 	if (error)
968 		return (error);
969 
970 	if (status) {
971 		tmpstat &= 0xffff;
972 		if (WIFSIGNALED(tmpstat))
973 			tmpstat = (tmpstat & 0xffffff80) |
974 			    bsd_to_linux_signal(WTERMSIG(tmpstat));
975 		else if (WIFSTOPPED(tmpstat))
976 			tmpstat = (tmpstat & 0xffff00ff) |
977 			    (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8);
978 		else if (WIFCONTINUED(tmpstat))
979 			tmpstat = 0xffff;
980 		error = copyout(&tmpstat, status, sizeof(int));
981 	}
982 
983 	return (error);
984 }
985 
986 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
987 int
988 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
989 {
990 	struct linux_wait4_args wait4_args;
991 
992 #ifdef DEBUG
993 	if (ldebug(waitpid))
994 		printf(ARGS(waitpid, "%d, %p, %d"),
995 		    args->pid, (void *)args->status, args->options);
996 #endif
997 
998 	wait4_args.pid = args->pid;
999 	wait4_args.status = args->status;
1000 	wait4_args.options = args->options;
1001 	wait4_args.rusage = NULL;
1002 
1003 	return (linux_wait4(td, &wait4_args));
1004 }
1005 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1006 
1007 int
1008 linux_wait4(struct thread *td, struct linux_wait4_args *args)
1009 {
1010 	int error, options;
1011 	struct rusage ru, *rup;
1012 
1013 #ifdef DEBUG
1014 	if (ldebug(wait4))
1015 		printf(ARGS(wait4, "%d, %p, %d, %p"),
1016 		    args->pid, (void *)args->status, args->options,
1017 		    (void *)args->rusage);
1018 #endif
1019 	if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG |
1020 	    LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
1021 		return (EINVAL);
1022 
1023 	options = WEXITED;
1024 	linux_to_bsd_waitopts(args->options, &options);
1025 
1026 	if (args->rusage != NULL)
1027 		rup = &ru;
1028 	else
1029 		rup = NULL;
1030 	error = linux_common_wait(td, args->pid, args->status, options, rup);
1031 	if (error != 0)
1032 		return (error);
1033 	if (args->rusage != NULL)
1034 		error = linux_copyout_rusage(&ru, args->rusage);
1035 	return (error);
1036 }
1037 
1038 int
1039 linux_waitid(struct thread *td, struct linux_waitid_args *args)
1040 {
1041 	int status, options, sig;
1042 	struct __wrusage wru;
1043 	siginfo_t siginfo;
1044 	l_siginfo_t lsi;
1045 	idtype_t idtype;
1046 	struct proc *p;
1047 	int error;
1048 
1049 	options = 0;
1050 	linux_to_bsd_waitopts(args->options, &options);
1051 
1052 	if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED))
1053 		return (EINVAL);
1054 	if (!(options & (WEXITED | WUNTRACED | WCONTINUED)))
1055 		return (EINVAL);
1056 
1057 	switch (args->idtype) {
1058 	case LINUX_P_ALL:
1059 		idtype = P_ALL;
1060 		break;
1061 	case LINUX_P_PID:
1062 		if (args->id <= 0)
1063 			return (EINVAL);
1064 		idtype = P_PID;
1065 		break;
1066 	case LINUX_P_PGID:
1067 		if (args->id <= 0)
1068 			return (EINVAL);
1069 		idtype = P_PGID;
1070 		break;
1071 	default:
1072 		return (EINVAL);
1073 	}
1074 
1075 	error = kern_wait6(td, idtype, args->id, &status, options,
1076 	    &wru, &siginfo);
1077 	if (error != 0)
1078 		return (error);
1079 	if (args->rusage != NULL) {
1080 		error = linux_copyout_rusage(&wru.wru_children,
1081 		    args->rusage);
1082 		if (error != 0)
1083 			return (error);
1084 	}
1085 	if (args->info != NULL) {
1086 		p = td->td_proc;
1087 		if (td->td_retval[0] == 0)
1088 			bzero(&lsi, sizeof(lsi));
1089 		else {
1090 			sig = bsd_to_linux_signal(siginfo.si_signo);
1091 			siginfo_to_lsiginfo(&siginfo, &lsi, sig);
1092 		}
1093 		error = copyout(&lsi, args->info, sizeof(lsi));
1094 	}
1095 	td->td_retval[0] = 0;
1096 
1097 	return (error);
1098 }
1099 
1100 int
1101 linux_mknod(struct thread *td, struct linux_mknod_args *args)
1102 {
1103 	char *path;
1104 	int error;
1105 
1106 	LCONVPATHCREAT(td, args->path, &path);
1107 
1108 #ifdef DEBUG
1109 	if (ldebug(mknod))
1110 		printf(ARGS(mknod, "%s, %d, %ju"), path, args->mode,
1111 		    (uintmax_t)args->dev);
1112 #endif
1113 
1114 	switch (args->mode & S_IFMT) {
1115 	case S_IFIFO:
1116 	case S_IFSOCK:
1117 		error = kern_mkfifoat(td, AT_FDCWD, path, UIO_SYSSPACE,
1118 		    args->mode);
1119 		break;
1120 
1121 	case S_IFCHR:
1122 	case S_IFBLK:
1123 		error = kern_mknodat(td, AT_FDCWD, path, UIO_SYSSPACE,
1124 		    args->mode, args->dev);
1125 		break;
1126 
1127 	case S_IFDIR:
1128 		error = EPERM;
1129 		break;
1130 
1131 	case 0:
1132 		args->mode |= S_IFREG;
1133 		/* FALLTHROUGH */
1134 	case S_IFREG:
1135 		error = kern_openat(td, AT_FDCWD, path, UIO_SYSSPACE,
1136 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1137 		if (error == 0)
1138 			kern_close(td, td->td_retval[0]);
1139 		break;
1140 
1141 	default:
1142 		error = EINVAL;
1143 		break;
1144 	}
1145 	LFREEPATH(path);
1146 	return (error);
1147 }
1148 
1149 int
1150 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
1151 {
1152 	char *path;
1153 	int error, dfd;
1154 
1155 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
1156 	LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
1157 
1158 #ifdef DEBUG
1159 	if (ldebug(mknodat))
1160 		printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev);
1161 #endif
1162 
1163 	switch (args->mode & S_IFMT) {
1164 	case S_IFIFO:
1165 	case S_IFSOCK:
1166 		error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
1167 		break;
1168 
1169 	case S_IFCHR:
1170 	case S_IFBLK:
1171 		error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
1172 		    args->dev);
1173 		break;
1174 
1175 	case S_IFDIR:
1176 		error = EPERM;
1177 		break;
1178 
1179 	case 0:
1180 		args->mode |= S_IFREG;
1181 		/* FALLTHROUGH */
1182 	case S_IFREG:
1183 		error = kern_openat(td, dfd, path, UIO_SYSSPACE,
1184 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1185 		if (error == 0)
1186 			kern_close(td, td->td_retval[0]);
1187 		break;
1188 
1189 	default:
1190 		error = EINVAL;
1191 		break;
1192 	}
1193 	LFREEPATH(path);
1194 	return (error);
1195 }
1196 
1197 /*
1198  * UGH! This is just about the dumbest idea I've ever heard!!
1199  */
1200 int
1201 linux_personality(struct thread *td, struct linux_personality_args *args)
1202 {
1203 #ifdef DEBUG
1204 	if (ldebug(personality))
1205 		printf(ARGS(personality, "%lu"), (unsigned long)args->per);
1206 #endif
1207 	if (args->per != 0)
1208 		return (EINVAL);
1209 
1210 	/* Yes Jim, it's still a Linux... */
1211 	td->td_retval[0] = 0;
1212 	return (0);
1213 }
1214 
1215 struct l_itimerval {
1216 	l_timeval it_interval;
1217 	l_timeval it_value;
1218 };
1219 
1220 #define	B2L_ITIMERVAL(bip, lip) 					\
1221 	(bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec;		\
1222 	(bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec;	\
1223 	(bip)->it_value.tv_sec = (lip)->it_value.tv_sec;		\
1224 	(bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1225 
1226 int
1227 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1228 {
1229 	int error;
1230 	struct l_itimerval ls;
1231 	struct itimerval aitv, oitv;
1232 
1233 #ifdef DEBUG
1234 	if (ldebug(setitimer))
1235 		printf(ARGS(setitimer, "%p, %p"),
1236 		    (void *)uap->itv, (void *)uap->oitv);
1237 #endif
1238 
1239 	if (uap->itv == NULL) {
1240 		uap->itv = uap->oitv;
1241 		return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1242 	}
1243 
1244 	error = copyin(uap->itv, &ls, sizeof(ls));
1245 	if (error != 0)
1246 		return (error);
1247 	B2L_ITIMERVAL(&aitv, &ls);
1248 #ifdef DEBUG
1249 	if (ldebug(setitimer)) {
1250 		printf("setitimer: value: sec: %jd, usec: %ld\n",
1251 		    (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1252 		printf("setitimer: interval: sec: %jd, usec: %ld\n",
1253 		    (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1254 	}
1255 #endif
1256 	error = kern_setitimer(td, uap->which, &aitv, &oitv);
1257 	if (error != 0 || uap->oitv == NULL)
1258 		return (error);
1259 	B2L_ITIMERVAL(&ls, &oitv);
1260 
1261 	return (copyout(&ls, uap->oitv, sizeof(ls)));
1262 }
1263 
1264 int
1265 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1266 {
1267 	int error;
1268 	struct l_itimerval ls;
1269 	struct itimerval aitv;
1270 
1271 #ifdef DEBUG
1272 	if (ldebug(getitimer))
1273 		printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1274 #endif
1275 	error = kern_getitimer(td, uap->which, &aitv);
1276 	if (error != 0)
1277 		return (error);
1278 	B2L_ITIMERVAL(&ls, &aitv);
1279 	return (copyout(&ls, uap->itv, sizeof(ls)));
1280 }
1281 
1282 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1283 int
1284 linux_nice(struct thread *td, struct linux_nice_args *args)
1285 {
1286 	struct setpriority_args bsd_args;
1287 
1288 	bsd_args.which = PRIO_PROCESS;
1289 	bsd_args.who = 0;		/* current process */
1290 	bsd_args.prio = args->inc;
1291 	return (sys_setpriority(td, &bsd_args));
1292 }
1293 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1294 
1295 int
1296 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1297 {
1298 	struct ucred *newcred, *oldcred;
1299 	l_gid_t *linux_gidset;
1300 	gid_t *bsd_gidset;
1301 	int ngrp, error;
1302 	struct proc *p;
1303 
1304 	ngrp = args->gidsetsize;
1305 	if (ngrp < 0 || ngrp >= ngroups_max + 1)
1306 		return (EINVAL);
1307 	linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK);
1308 	error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1309 	if (error)
1310 		goto out;
1311 	newcred = crget();
1312 	crextend(newcred, ngrp + 1);
1313 	p = td->td_proc;
1314 	PROC_LOCK(p);
1315 	oldcred = p->p_ucred;
1316 	crcopy(newcred, oldcred);
1317 
1318 	/*
1319 	 * cr_groups[0] holds egid. Setting the whole set from
1320 	 * the supplied set will cause egid to be changed too.
1321 	 * Keep cr_groups[0] unchanged to prevent that.
1322 	 */
1323 
1324 	if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) {
1325 		PROC_UNLOCK(p);
1326 		crfree(newcred);
1327 		goto out;
1328 	}
1329 
1330 	if (ngrp > 0) {
1331 		newcred->cr_ngroups = ngrp + 1;
1332 
1333 		bsd_gidset = newcred->cr_groups;
1334 		ngrp--;
1335 		while (ngrp >= 0) {
1336 			bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1337 			ngrp--;
1338 		}
1339 	} else
1340 		newcred->cr_ngroups = 1;
1341 
1342 	setsugid(p);
1343 	proc_set_cred(p, newcred);
1344 	PROC_UNLOCK(p);
1345 	crfree(oldcred);
1346 	error = 0;
1347 out:
1348 	free(linux_gidset, M_LINUX);
1349 	return (error);
1350 }
1351 
1352 int
1353 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1354 {
1355 	struct ucred *cred;
1356 	l_gid_t *linux_gidset;
1357 	gid_t *bsd_gidset;
1358 	int bsd_gidsetsz, ngrp, error;
1359 
1360 	cred = td->td_ucred;
1361 	bsd_gidset = cred->cr_groups;
1362 	bsd_gidsetsz = cred->cr_ngroups - 1;
1363 
1364 	/*
1365 	 * cr_groups[0] holds egid. Returning the whole set
1366 	 * here will cause a duplicate. Exclude cr_groups[0]
1367 	 * to prevent that.
1368 	 */
1369 
1370 	if ((ngrp = args->gidsetsize) == 0) {
1371 		td->td_retval[0] = bsd_gidsetsz;
1372 		return (0);
1373 	}
1374 
1375 	if (ngrp < bsd_gidsetsz)
1376 		return (EINVAL);
1377 
1378 	ngrp = 0;
1379 	linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1380 	    M_LINUX, M_WAITOK);
1381 	while (ngrp < bsd_gidsetsz) {
1382 		linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1383 		ngrp++;
1384 	}
1385 
1386 	error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1387 	free(linux_gidset, M_LINUX);
1388 	if (error)
1389 		return (error);
1390 
1391 	td->td_retval[0] = ngrp;
1392 	return (0);
1393 }
1394 
1395 int
1396 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1397 {
1398 	struct rlimit bsd_rlim;
1399 	struct l_rlimit rlim;
1400 	u_int which;
1401 	int error;
1402 
1403 #ifdef DEBUG
1404 	if (ldebug(setrlimit))
1405 		printf(ARGS(setrlimit, "%d, %p"),
1406 		    args->resource, (void *)args->rlim);
1407 #endif
1408 
1409 	if (args->resource >= LINUX_RLIM_NLIMITS)
1410 		return (EINVAL);
1411 
1412 	which = linux_to_bsd_resource[args->resource];
1413 	if (which == -1)
1414 		return (EINVAL);
1415 
1416 	error = copyin(args->rlim, &rlim, sizeof(rlim));
1417 	if (error)
1418 		return (error);
1419 
1420 	bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1421 	bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1422 	return (kern_setrlimit(td, which, &bsd_rlim));
1423 }
1424 
1425 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1426 int
1427 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1428 {
1429 	struct l_rlimit rlim;
1430 	struct rlimit bsd_rlim;
1431 	u_int which;
1432 
1433 #ifdef DEBUG
1434 	if (ldebug(old_getrlimit))
1435 		printf(ARGS(old_getrlimit, "%d, %p"),
1436 		    args->resource, (void *)args->rlim);
1437 #endif
1438 
1439 	if (args->resource >= LINUX_RLIM_NLIMITS)
1440 		return (EINVAL);
1441 
1442 	which = linux_to_bsd_resource[args->resource];
1443 	if (which == -1)
1444 		return (EINVAL);
1445 
1446 	lim_rlimit(td, which, &bsd_rlim);
1447 
1448 #ifdef COMPAT_LINUX32
1449 	rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1450 	if (rlim.rlim_cur == UINT_MAX)
1451 		rlim.rlim_cur = INT_MAX;
1452 	rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1453 	if (rlim.rlim_max == UINT_MAX)
1454 		rlim.rlim_max = INT_MAX;
1455 #else
1456 	rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1457 	if (rlim.rlim_cur == ULONG_MAX)
1458 		rlim.rlim_cur = LONG_MAX;
1459 	rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1460 	if (rlim.rlim_max == ULONG_MAX)
1461 		rlim.rlim_max = LONG_MAX;
1462 #endif
1463 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1464 }
1465 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1466 
1467 int
1468 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1469 {
1470 	struct l_rlimit rlim;
1471 	struct rlimit bsd_rlim;
1472 	u_int which;
1473 
1474 #ifdef DEBUG
1475 	if (ldebug(getrlimit))
1476 		printf(ARGS(getrlimit, "%d, %p"),
1477 		    args->resource, (void *)args->rlim);
1478 #endif
1479 
1480 	if (args->resource >= LINUX_RLIM_NLIMITS)
1481 		return (EINVAL);
1482 
1483 	which = linux_to_bsd_resource[args->resource];
1484 	if (which == -1)
1485 		return (EINVAL);
1486 
1487 	lim_rlimit(td, which, &bsd_rlim);
1488 
1489 	rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1490 	rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1491 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1492 }
1493 
1494 int
1495 linux_sched_setscheduler(struct thread *td,
1496     struct linux_sched_setscheduler_args *args)
1497 {
1498 	struct sched_param sched_param;
1499 	struct thread *tdt;
1500 	int error, policy;
1501 
1502 #ifdef DEBUG
1503 	if (ldebug(sched_setscheduler))
1504 		printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1505 		    args->pid, args->policy, (const void *)args->param);
1506 #endif
1507 
1508 	switch (args->policy) {
1509 	case LINUX_SCHED_OTHER:
1510 		policy = SCHED_OTHER;
1511 		break;
1512 	case LINUX_SCHED_FIFO:
1513 		policy = SCHED_FIFO;
1514 		break;
1515 	case LINUX_SCHED_RR:
1516 		policy = SCHED_RR;
1517 		break;
1518 	default:
1519 		return (EINVAL);
1520 	}
1521 
1522 	error = copyin(args->param, &sched_param, sizeof(sched_param));
1523 	if (error)
1524 		return (error);
1525 
1526 	tdt = linux_tdfind(td, args->pid, -1);
1527 	if (tdt == NULL)
1528 		return (ESRCH);
1529 
1530 	error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1531 	PROC_UNLOCK(tdt->td_proc);
1532 	return (error);
1533 }
1534 
1535 int
1536 linux_sched_getscheduler(struct thread *td,
1537     struct linux_sched_getscheduler_args *args)
1538 {
1539 	struct thread *tdt;
1540 	int error, policy;
1541 
1542 #ifdef DEBUG
1543 	if (ldebug(sched_getscheduler))
1544 		printf(ARGS(sched_getscheduler, "%d"), args->pid);
1545 #endif
1546 
1547 	tdt = linux_tdfind(td, args->pid, -1);
1548 	if (tdt == NULL)
1549 		return (ESRCH);
1550 
1551 	error = kern_sched_getscheduler(td, tdt, &policy);
1552 	PROC_UNLOCK(tdt->td_proc);
1553 
1554 	switch (policy) {
1555 	case SCHED_OTHER:
1556 		td->td_retval[0] = LINUX_SCHED_OTHER;
1557 		break;
1558 	case SCHED_FIFO:
1559 		td->td_retval[0] = LINUX_SCHED_FIFO;
1560 		break;
1561 	case SCHED_RR:
1562 		td->td_retval[0] = LINUX_SCHED_RR;
1563 		break;
1564 	}
1565 	return (error);
1566 }
1567 
1568 int
1569 linux_sched_get_priority_max(struct thread *td,
1570     struct linux_sched_get_priority_max_args *args)
1571 {
1572 	struct sched_get_priority_max_args bsd;
1573 
1574 #ifdef DEBUG
1575 	if (ldebug(sched_get_priority_max))
1576 		printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1577 #endif
1578 
1579 	switch (args->policy) {
1580 	case LINUX_SCHED_OTHER:
1581 		bsd.policy = SCHED_OTHER;
1582 		break;
1583 	case LINUX_SCHED_FIFO:
1584 		bsd.policy = SCHED_FIFO;
1585 		break;
1586 	case LINUX_SCHED_RR:
1587 		bsd.policy = SCHED_RR;
1588 		break;
1589 	default:
1590 		return (EINVAL);
1591 	}
1592 	return (sys_sched_get_priority_max(td, &bsd));
1593 }
1594 
1595 int
1596 linux_sched_get_priority_min(struct thread *td,
1597     struct linux_sched_get_priority_min_args *args)
1598 {
1599 	struct sched_get_priority_min_args bsd;
1600 
1601 #ifdef DEBUG
1602 	if (ldebug(sched_get_priority_min))
1603 		printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1604 #endif
1605 
1606 	switch (args->policy) {
1607 	case LINUX_SCHED_OTHER:
1608 		bsd.policy = SCHED_OTHER;
1609 		break;
1610 	case LINUX_SCHED_FIFO:
1611 		bsd.policy = SCHED_FIFO;
1612 		break;
1613 	case LINUX_SCHED_RR:
1614 		bsd.policy = SCHED_RR;
1615 		break;
1616 	default:
1617 		return (EINVAL);
1618 	}
1619 	return (sys_sched_get_priority_min(td, &bsd));
1620 }
1621 
1622 #define REBOOT_CAD_ON	0x89abcdef
1623 #define REBOOT_CAD_OFF	0
1624 #define REBOOT_HALT	0xcdef0123
1625 #define REBOOT_RESTART	0x01234567
1626 #define REBOOT_RESTART2	0xA1B2C3D4
1627 #define REBOOT_POWEROFF	0x4321FEDC
1628 #define REBOOT_MAGIC1	0xfee1dead
1629 #define REBOOT_MAGIC2	0x28121969
1630 #define REBOOT_MAGIC2A	0x05121996
1631 #define REBOOT_MAGIC2B	0x16041998
1632 
1633 int
1634 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1635 {
1636 	struct reboot_args bsd_args;
1637 
1638 #ifdef DEBUG
1639 	if (ldebug(reboot))
1640 		printf(ARGS(reboot, "0x%x"), args->cmd);
1641 #endif
1642 
1643 	if (args->magic1 != REBOOT_MAGIC1)
1644 		return (EINVAL);
1645 
1646 	switch (args->magic2) {
1647 	case REBOOT_MAGIC2:
1648 	case REBOOT_MAGIC2A:
1649 	case REBOOT_MAGIC2B:
1650 		break;
1651 	default:
1652 		return (EINVAL);
1653 	}
1654 
1655 	switch (args->cmd) {
1656 	case REBOOT_CAD_ON:
1657 	case REBOOT_CAD_OFF:
1658 		return (priv_check(td, PRIV_REBOOT));
1659 	case REBOOT_HALT:
1660 		bsd_args.opt = RB_HALT;
1661 		break;
1662 	case REBOOT_RESTART:
1663 	case REBOOT_RESTART2:
1664 		bsd_args.opt = 0;
1665 		break;
1666 	case REBOOT_POWEROFF:
1667 		bsd_args.opt = RB_POWEROFF;
1668 		break;
1669 	default:
1670 		return (EINVAL);
1671 	}
1672 	return (sys_reboot(td, &bsd_args));
1673 }
1674 
1675 
1676 /*
1677  * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1678  * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
1679  * are assumed to be preserved. The following lightweight syscalls fixes
1680  * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
1681  *
1682  * linux_getpid() - MP SAFE
1683  * linux_getgid() - MP SAFE
1684  * linux_getuid() - MP SAFE
1685  */
1686 
1687 int
1688 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1689 {
1690 
1691 #ifdef DEBUG
1692 	if (ldebug(getpid))
1693 		printf(ARGS(getpid, ""));
1694 #endif
1695 	td->td_retval[0] = td->td_proc->p_pid;
1696 
1697 	return (0);
1698 }
1699 
1700 int
1701 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1702 {
1703 	struct linux_emuldata *em;
1704 
1705 #ifdef DEBUG
1706 	if (ldebug(gettid))
1707 		printf(ARGS(gettid, ""));
1708 #endif
1709 
1710 	em = em_find(td);
1711 	KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1712 
1713 	td->td_retval[0] = em->em_tid;
1714 
1715 	return (0);
1716 }
1717 
1718 
1719 int
1720 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1721 {
1722 
1723 #ifdef DEBUG
1724 	if (ldebug(getppid))
1725 		printf(ARGS(getppid, ""));
1726 #endif
1727 
1728 	PROC_LOCK(td->td_proc);
1729 	td->td_retval[0] = td->td_proc->p_pptr->p_pid;
1730 	PROC_UNLOCK(td->td_proc);
1731 	return (0);
1732 }
1733 
1734 int
1735 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1736 {
1737 
1738 #ifdef DEBUG
1739 	if (ldebug(getgid))
1740 		printf(ARGS(getgid, ""));
1741 #endif
1742 
1743 	td->td_retval[0] = td->td_ucred->cr_rgid;
1744 	return (0);
1745 }
1746 
1747 int
1748 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1749 {
1750 
1751 #ifdef DEBUG
1752 	if (ldebug(getuid))
1753 		printf(ARGS(getuid, ""));
1754 #endif
1755 
1756 	td->td_retval[0] = td->td_ucred->cr_ruid;
1757 	return (0);
1758 }
1759 
1760 
1761 int
1762 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1763 {
1764 	struct getsid_args bsd;
1765 
1766 #ifdef DEBUG
1767 	if (ldebug(getsid))
1768 		printf(ARGS(getsid, "%i"), args->pid);
1769 #endif
1770 
1771 	bsd.pid = args->pid;
1772 	return (sys_getsid(td, &bsd));
1773 }
1774 
1775 int
1776 linux_nosys(struct thread *td, struct nosys_args *ignore)
1777 {
1778 
1779 	return (ENOSYS);
1780 }
1781 
1782 int
1783 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1784 {
1785 	struct getpriority_args bsd_args;
1786 	int error;
1787 
1788 #ifdef DEBUG
1789 	if (ldebug(getpriority))
1790 		printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
1791 #endif
1792 
1793 	bsd_args.which = args->which;
1794 	bsd_args.who = args->who;
1795 	error = sys_getpriority(td, &bsd_args);
1796 	td->td_retval[0] = 20 - td->td_retval[0];
1797 	return (error);
1798 }
1799 
1800 int
1801 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1802 {
1803 	int name[2];
1804 
1805 #ifdef DEBUG
1806 	if (ldebug(sethostname))
1807 		printf(ARGS(sethostname, "*, %i"), args->len);
1808 #endif
1809 
1810 	name[0] = CTL_KERN;
1811 	name[1] = KERN_HOSTNAME;
1812 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1813 	    args->len, 0, 0));
1814 }
1815 
1816 int
1817 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1818 {
1819 	int name[2];
1820 
1821 #ifdef DEBUG
1822 	if (ldebug(setdomainname))
1823 		printf(ARGS(setdomainname, "*, %i"), args->len);
1824 #endif
1825 
1826 	name[0] = CTL_KERN;
1827 	name[1] = KERN_NISDOMAINNAME;
1828 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1829 	    args->len, 0, 0));
1830 }
1831 
1832 int
1833 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1834 {
1835 
1836 #ifdef DEBUG
1837 	if (ldebug(exit_group))
1838 		printf(ARGS(exit_group, "%i"), args->error_code);
1839 #endif
1840 
1841 	LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1842 	    args->error_code);
1843 
1844 	/*
1845 	 * XXX: we should send a signal to the parent if
1846 	 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1847 	 * as it doesnt occur often.
1848 	 */
1849 	exit1(td, args->error_code, 0);
1850 		/* NOTREACHED */
1851 }
1852 
1853 #define _LINUX_CAPABILITY_VERSION  0x19980330
1854 
1855 struct l_user_cap_header {
1856 	l_int	version;
1857 	l_int	pid;
1858 };
1859 
1860 struct l_user_cap_data {
1861 	l_int	effective;
1862 	l_int	permitted;
1863 	l_int	inheritable;
1864 };
1865 
1866 int
1867 linux_capget(struct thread *td, struct linux_capget_args *args)
1868 {
1869 	struct l_user_cap_header luch;
1870 	struct l_user_cap_data lucd;
1871 	int error;
1872 
1873 	if (args->hdrp == NULL)
1874 		return (EFAULT);
1875 
1876 	error = copyin(args->hdrp, &luch, sizeof(luch));
1877 	if (error != 0)
1878 		return (error);
1879 
1880 	if (luch.version != _LINUX_CAPABILITY_VERSION) {
1881 		luch.version = _LINUX_CAPABILITY_VERSION;
1882 		error = copyout(&luch, args->hdrp, sizeof(luch));
1883 		if (error)
1884 			return (error);
1885 		return (EINVAL);
1886 	}
1887 
1888 	if (luch.pid)
1889 		return (EPERM);
1890 
1891 	if (args->datap) {
1892 		/*
1893 		 * The current implementation doesn't support setting
1894 		 * a capability (it's essentially a stub) so indicate
1895 		 * that no capabilities are currently set or available
1896 		 * to request.
1897 		 */
1898 		bzero (&lucd, sizeof(lucd));
1899 		error = copyout(&lucd, args->datap, sizeof(lucd));
1900 	}
1901 
1902 	return (error);
1903 }
1904 
1905 int
1906 linux_capset(struct thread *td, struct linux_capset_args *args)
1907 {
1908 	struct l_user_cap_header luch;
1909 	struct l_user_cap_data lucd;
1910 	int error;
1911 
1912 	if (args->hdrp == NULL || args->datap == NULL)
1913 		return (EFAULT);
1914 
1915 	error = copyin(args->hdrp, &luch, sizeof(luch));
1916 	if (error != 0)
1917 		return (error);
1918 
1919 	if (luch.version != _LINUX_CAPABILITY_VERSION) {
1920 		luch.version = _LINUX_CAPABILITY_VERSION;
1921 		error = copyout(&luch, args->hdrp, sizeof(luch));
1922 		if (error)
1923 			return (error);
1924 		return (EINVAL);
1925 	}
1926 
1927 	if (luch.pid)
1928 		return (EPERM);
1929 
1930 	error = copyin(args->datap, &lucd, sizeof(lucd));
1931 	if (error != 0)
1932 		return (error);
1933 
1934 	/* We currently don't support setting any capabilities. */
1935 	if (lucd.effective || lucd.permitted || lucd.inheritable) {
1936 		linux_msg(td,
1937 			  "capset effective=0x%x, permitted=0x%x, "
1938 			  "inheritable=0x%x is not implemented",
1939 			  (int)lucd.effective, (int)lucd.permitted,
1940 			  (int)lucd.inheritable);
1941 		return (EPERM);
1942 	}
1943 
1944 	return (0);
1945 }
1946 
1947 int
1948 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1949 {
1950 	int error = 0, max_size;
1951 	struct proc *p = td->td_proc;
1952 	char comm[LINUX_MAX_COMM_LEN];
1953 	struct linux_emuldata *em;
1954 	int pdeath_signal;
1955 
1956 #ifdef DEBUG
1957 	if (ldebug(prctl))
1958 		printf(ARGS(prctl, "%d, %ju, %ju, %ju, %ju"), args->option,
1959 		    (uintmax_t)args->arg2, (uintmax_t)args->arg3,
1960 		    (uintmax_t)args->arg4, (uintmax_t)args->arg5);
1961 #endif
1962 
1963 	switch (args->option) {
1964 	case LINUX_PR_SET_PDEATHSIG:
1965 		if (!LINUX_SIG_VALID(args->arg2))
1966 			return (EINVAL);
1967 		em = em_find(td);
1968 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1969 		em->pdeath_signal = args->arg2;
1970 		break;
1971 	case LINUX_PR_GET_PDEATHSIG:
1972 		em = em_find(td);
1973 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1974 		pdeath_signal = em->pdeath_signal;
1975 		error = copyout(&pdeath_signal,
1976 		    (void *)(register_t)args->arg2,
1977 		    sizeof(pdeath_signal));
1978 		break;
1979 	case LINUX_PR_GET_KEEPCAPS:
1980 		/*
1981 		 * Indicate that we always clear the effective and
1982 		 * permitted capability sets when the user id becomes
1983 		 * non-zero (actually the capability sets are simply
1984 		 * always zero in the current implementation).
1985 		 */
1986 		td->td_retval[0] = 0;
1987 		break;
1988 	case LINUX_PR_SET_KEEPCAPS:
1989 		/*
1990 		 * Ignore requests to keep the effective and permitted
1991 		 * capability sets when the user id becomes non-zero.
1992 		 */
1993 		break;
1994 	case LINUX_PR_SET_NAME:
1995 		/*
1996 		 * To be on the safe side we need to make sure to not
1997 		 * overflow the size a linux program expects. We already
1998 		 * do this here in the copyin, so that we don't need to
1999 		 * check on copyout.
2000 		 */
2001 		max_size = MIN(sizeof(comm), sizeof(p->p_comm));
2002 		error = copyinstr((void *)(register_t)args->arg2, comm,
2003 		    max_size, NULL);
2004 
2005 		/* Linux silently truncates the name if it is too long. */
2006 		if (error == ENAMETOOLONG) {
2007 			/*
2008 			 * XXX: copyinstr() isn't documented to populate the
2009 			 * array completely, so do a copyin() to be on the
2010 			 * safe side. This should be changed in case
2011 			 * copyinstr() is changed to guarantee this.
2012 			 */
2013 			error = copyin((void *)(register_t)args->arg2, comm,
2014 			    max_size - 1);
2015 			comm[max_size - 1] = '\0';
2016 		}
2017 		if (error)
2018 			return (error);
2019 
2020 		PROC_LOCK(p);
2021 		strlcpy(p->p_comm, comm, sizeof(p->p_comm));
2022 		PROC_UNLOCK(p);
2023 		break;
2024 	case LINUX_PR_GET_NAME:
2025 		PROC_LOCK(p);
2026 		strlcpy(comm, p->p_comm, sizeof(comm));
2027 		PROC_UNLOCK(p);
2028 		error = copyout(comm, (void *)(register_t)args->arg2,
2029 		    strlen(comm) + 1);
2030 		break;
2031 	default:
2032 		error = EINVAL;
2033 		break;
2034 	}
2035 
2036 	return (error);
2037 }
2038 
2039 int
2040 linux_sched_setparam(struct thread *td,
2041     struct linux_sched_setparam_args *uap)
2042 {
2043 	struct sched_param sched_param;
2044 	struct thread *tdt;
2045 	int error;
2046 
2047 #ifdef DEBUG
2048 	if (ldebug(sched_setparam))
2049 		printf(ARGS(sched_setparam, "%d, *"), uap->pid);
2050 #endif
2051 
2052 	error = copyin(uap->param, &sched_param, sizeof(sched_param));
2053 	if (error)
2054 		return (error);
2055 
2056 	tdt = linux_tdfind(td, uap->pid, -1);
2057 	if (tdt == NULL)
2058 		return (ESRCH);
2059 
2060 	error = kern_sched_setparam(td, tdt, &sched_param);
2061 	PROC_UNLOCK(tdt->td_proc);
2062 	return (error);
2063 }
2064 
2065 int
2066 linux_sched_getparam(struct thread *td,
2067     struct linux_sched_getparam_args *uap)
2068 {
2069 	struct sched_param sched_param;
2070 	struct thread *tdt;
2071 	int error;
2072 
2073 #ifdef DEBUG
2074 	if (ldebug(sched_getparam))
2075 		printf(ARGS(sched_getparam, "%d, *"), uap->pid);
2076 #endif
2077 
2078 	tdt = linux_tdfind(td, uap->pid, -1);
2079 	if (tdt == NULL)
2080 		return (ESRCH);
2081 
2082 	error = kern_sched_getparam(td, tdt, &sched_param);
2083 	PROC_UNLOCK(tdt->td_proc);
2084 	if (error == 0)
2085 		error = copyout(&sched_param, uap->param,
2086 		    sizeof(sched_param));
2087 	return (error);
2088 }
2089 
2090 /*
2091  * Get affinity of a process.
2092  */
2093 int
2094 linux_sched_getaffinity(struct thread *td,
2095     struct linux_sched_getaffinity_args *args)
2096 {
2097 	int error;
2098 	struct thread *tdt;
2099 	struct cpuset_getaffinity_args cga;
2100 
2101 #ifdef DEBUG
2102 	if (ldebug(sched_getaffinity))
2103 		printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid,
2104 		    args->len);
2105 #endif
2106 	if (args->len < sizeof(cpuset_t))
2107 		return (EINVAL);
2108 
2109 	tdt = linux_tdfind(td, args->pid, -1);
2110 	if (tdt == NULL)
2111 		return (ESRCH);
2112 
2113 	PROC_UNLOCK(tdt->td_proc);
2114 	cga.level = CPU_LEVEL_WHICH;
2115 	cga.which = CPU_WHICH_TID;
2116 	cga.id = tdt->td_tid;
2117 	cga.cpusetsize = sizeof(cpuset_t);
2118 	cga.mask = (cpuset_t *) args->user_mask_ptr;
2119 
2120 	if ((error = sys_cpuset_getaffinity(td, &cga)) == 0)
2121 		td->td_retval[0] = sizeof(cpuset_t);
2122 
2123 	return (error);
2124 }
2125 
2126 /*
2127  *  Set affinity of a process.
2128  */
2129 int
2130 linux_sched_setaffinity(struct thread *td,
2131     struct linux_sched_setaffinity_args *args)
2132 {
2133 	struct cpuset_setaffinity_args csa;
2134 	struct thread *tdt;
2135 
2136 #ifdef DEBUG
2137 	if (ldebug(sched_setaffinity))
2138 		printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid,
2139 		    args->len);
2140 #endif
2141 	if (args->len < sizeof(cpuset_t))
2142 		return (EINVAL);
2143 
2144 	tdt = linux_tdfind(td, args->pid, -1);
2145 	if (tdt == NULL)
2146 		return (ESRCH);
2147 
2148 	PROC_UNLOCK(tdt->td_proc);
2149 	csa.level = CPU_LEVEL_WHICH;
2150 	csa.which = CPU_WHICH_TID;
2151 	csa.id = tdt->td_tid;
2152 	csa.cpusetsize = sizeof(cpuset_t);
2153 	csa.mask = (cpuset_t *) args->user_mask_ptr;
2154 
2155 	return (sys_cpuset_setaffinity(td, &csa));
2156 }
2157 
2158 struct linux_rlimit64 {
2159 	uint64_t	rlim_cur;
2160 	uint64_t	rlim_max;
2161 };
2162 
2163 int
2164 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
2165 {
2166 	struct rlimit rlim, nrlim;
2167 	struct linux_rlimit64 lrlim;
2168 	struct proc *p;
2169 	u_int which;
2170 	int flags;
2171 	int error;
2172 
2173 #ifdef DEBUG
2174 	if (ldebug(prlimit64))
2175 		printf(ARGS(prlimit64, "%d, %d, %p, %p"), args->pid,
2176 		    args->resource, (void *)args->new, (void *)args->old);
2177 #endif
2178 
2179 	if (args->resource >= LINUX_RLIM_NLIMITS)
2180 		return (EINVAL);
2181 
2182 	which = linux_to_bsd_resource[args->resource];
2183 	if (which == -1)
2184 		return (EINVAL);
2185 
2186 	if (args->new != NULL) {
2187 		/*
2188 		 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
2189 		 * rlim is unsigned 64-bit. FreeBSD treats negative limits
2190 		 * as INFINITY so we do not need a conversion even.
2191 		 */
2192 		error = copyin(args->new, &nrlim, sizeof(nrlim));
2193 		if (error != 0)
2194 			return (error);
2195 	}
2196 
2197 	flags = PGET_HOLD | PGET_NOTWEXIT;
2198 	if (args->new != NULL)
2199 		flags |= PGET_CANDEBUG;
2200 	else
2201 		flags |= PGET_CANSEE;
2202 	error = pget(args->pid, flags, &p);
2203 	if (error != 0)
2204 		return (error);
2205 
2206 	if (args->old != NULL) {
2207 		PROC_LOCK(p);
2208 		lim_rlimit_proc(p, which, &rlim);
2209 		PROC_UNLOCK(p);
2210 		if (rlim.rlim_cur == RLIM_INFINITY)
2211 			lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2212 		else
2213 			lrlim.rlim_cur = rlim.rlim_cur;
2214 		if (rlim.rlim_max == RLIM_INFINITY)
2215 			lrlim.rlim_max = LINUX_RLIM_INFINITY;
2216 		else
2217 			lrlim.rlim_max = rlim.rlim_max;
2218 		error = copyout(&lrlim, args->old, sizeof(lrlim));
2219 		if (error != 0)
2220 			goto out;
2221 	}
2222 
2223 	if (args->new != NULL)
2224 		error = kern_proc_setrlimit(td, p, which, &nrlim);
2225 
2226  out:
2227 	PRELE(p);
2228 	return (error);
2229 }
2230 
2231 int
2232 linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2233 {
2234 	struct timeval utv, tv0, tv1, *tvp;
2235 	struct l_pselect6arg lpse6;
2236 	struct l_timespec lts;
2237 	struct timespec uts;
2238 	l_sigset_t l_ss;
2239 	sigset_t *ssp;
2240 	sigset_t ss;
2241 	int error;
2242 
2243 	ssp = NULL;
2244 	if (args->sig != NULL) {
2245 		error = copyin(args->sig, &lpse6, sizeof(lpse6));
2246 		if (error != 0)
2247 			return (error);
2248 		if (lpse6.ss_len != sizeof(l_ss))
2249 			return (EINVAL);
2250 		if (lpse6.ss != 0) {
2251 			error = copyin(PTRIN(lpse6.ss), &l_ss,
2252 			    sizeof(l_ss));
2253 			if (error != 0)
2254 				return (error);
2255 			linux_to_bsd_sigset(&l_ss, &ss);
2256 			ssp = &ss;
2257 		}
2258 	}
2259 
2260 	/*
2261 	 * Currently glibc changes nanosecond number to microsecond.
2262 	 * This mean losing precision but for now it is hardly seen.
2263 	 */
2264 	if (args->tsp != NULL) {
2265 		error = copyin(args->tsp, &lts, sizeof(lts));
2266 		if (error != 0)
2267 			return (error);
2268 		error = linux_to_native_timespec(&uts, &lts);
2269 		if (error != 0)
2270 			return (error);
2271 
2272 		TIMESPEC_TO_TIMEVAL(&utv, &uts);
2273 		if (itimerfix(&utv))
2274 			return (EINVAL);
2275 
2276 		microtime(&tv0);
2277 		tvp = &utv;
2278 	} else
2279 		tvp = NULL;
2280 
2281 	error = kern_pselect(td, args->nfds, args->readfds, args->writefds,
2282 	    args->exceptfds, tvp, ssp, LINUX_NFDBITS);
2283 
2284 	if (error == 0 && args->tsp != NULL) {
2285 		if (td->td_retval[0] != 0) {
2286 			/*
2287 			 * Compute how much time was left of the timeout,
2288 			 * by subtracting the current time and the time
2289 			 * before we started the call, and subtracting
2290 			 * that result from the user-supplied value.
2291 			 */
2292 
2293 			microtime(&tv1);
2294 			timevalsub(&tv1, &tv0);
2295 			timevalsub(&utv, &tv1);
2296 			if (utv.tv_sec < 0)
2297 				timevalclear(&utv);
2298 		} else
2299 			timevalclear(&utv);
2300 
2301 		TIMEVAL_TO_TIMESPEC(&utv, &uts);
2302 
2303 		native_to_linux_timespec(&lts, &uts);
2304 		error = copyout(&lts, args->tsp, sizeof(lts));
2305 	}
2306 
2307 	return (error);
2308 }
2309 
2310 int
2311 linux_ppoll(struct thread *td, struct linux_ppoll_args *args)
2312 {
2313 	struct timespec ts0, ts1;
2314 	struct l_timespec lts;
2315 	struct timespec uts, *tsp;
2316 	l_sigset_t l_ss;
2317 	sigset_t *ssp;
2318 	sigset_t ss;
2319 	int error;
2320 
2321 	if (args->sset != NULL) {
2322 		if (args->ssize != sizeof(l_ss))
2323 			return (EINVAL);
2324 		error = copyin(args->sset, &l_ss, sizeof(l_ss));
2325 		if (error)
2326 			return (error);
2327 		linux_to_bsd_sigset(&l_ss, &ss);
2328 		ssp = &ss;
2329 	} else
2330 		ssp = NULL;
2331 	if (args->tsp != NULL) {
2332 		error = copyin(args->tsp, &lts, sizeof(lts));
2333 		if (error)
2334 			return (error);
2335 		error = linux_to_native_timespec(&uts, &lts);
2336 		if (error != 0)
2337 			return (error);
2338 
2339 		nanotime(&ts0);
2340 		tsp = &uts;
2341 	} else
2342 		tsp = NULL;
2343 
2344 	error = kern_poll(td, args->fds, args->nfds, tsp, ssp);
2345 
2346 	if (error == 0 && args->tsp != NULL) {
2347 		if (td->td_retval[0]) {
2348 			nanotime(&ts1);
2349 			timespecsub(&ts1, &ts0);
2350 			timespecsub(&uts, &ts1);
2351 			if (uts.tv_sec < 0)
2352 				timespecclear(&uts);
2353 		} else
2354 			timespecclear(&uts);
2355 
2356 		native_to_linux_timespec(&lts, &uts);
2357 		error = copyout(&lts, args->tsp, sizeof(lts));
2358 	}
2359 
2360 	return (error);
2361 }
2362 
2363 #if defined(DEBUG) || defined(KTR)
2364 /* XXX: can be removed when every ldebug(...) and KTR stuff are removed. */
2365 
2366 #ifdef COMPAT_LINUX32
2367 #define	L_MAXSYSCALL	LINUX32_SYS_MAXSYSCALL
2368 #else
2369 #define	L_MAXSYSCALL	LINUX_SYS_MAXSYSCALL
2370 #endif
2371 
2372 u_char linux_debug_map[howmany(L_MAXSYSCALL, sizeof(u_char))];
2373 
2374 static int
2375 linux_debug(int syscall, int toggle, int global)
2376 {
2377 
2378 	if (global) {
2379 		char c = toggle ? 0 : 0xff;
2380 
2381 		memset(linux_debug_map, c, sizeof(linux_debug_map));
2382 		return (0);
2383 	}
2384 	if (syscall < 0 || syscall >= L_MAXSYSCALL)
2385 		return (EINVAL);
2386 	if (toggle)
2387 		clrbit(linux_debug_map, syscall);
2388 	else
2389 		setbit(linux_debug_map, syscall);
2390 	return (0);
2391 }
2392 #undef L_MAXSYSCALL
2393 
2394 /*
2395  * Usage: sysctl linux.debug=<syscall_nr>.<0/1>
2396  *
2397  *    E.g.: sysctl linux.debug=21.0
2398  *
2399  * As a special case, syscall "all" will apply to all syscalls globally.
2400  */
2401 #define LINUX_MAX_DEBUGSTR	16
2402 int
2403 linux_sysctl_debug(SYSCTL_HANDLER_ARGS)
2404 {
2405 	char value[LINUX_MAX_DEBUGSTR], *p;
2406 	int error, sysc, toggle;
2407 	int global = 0;
2408 
2409 	value[0] = '\0';
2410 	error = sysctl_handle_string(oidp, value, LINUX_MAX_DEBUGSTR, req);
2411 	if (error || req->newptr == NULL)
2412 		return (error);
2413 	for (p = value; *p != '\0' && *p != '.'; p++);
2414 	if (*p == '\0')
2415 		return (EINVAL);
2416 	*p++ = '\0';
2417 	sysc = strtol(value, NULL, 0);
2418 	toggle = strtol(p, NULL, 0);
2419 	if (strcmp(value, "all") == 0)
2420 		global = 1;
2421 	error = linux_debug(sysc, toggle, global);
2422 	return (error);
2423 }
2424 
2425 #endif /* DEBUG || KTR */
2426 
2427 int
2428 linux_sched_rr_get_interval(struct thread *td,
2429     struct linux_sched_rr_get_interval_args *uap)
2430 {
2431 	struct timespec ts;
2432 	struct l_timespec lts;
2433 	struct thread *tdt;
2434 	int error;
2435 
2436 	/*
2437 	 * According to man in case the invalid pid specified
2438 	 * EINVAL should be returned.
2439 	 */
2440 	if (uap->pid < 0)
2441 		return (EINVAL);
2442 
2443 	tdt = linux_tdfind(td, uap->pid, -1);
2444 	if (tdt == NULL)
2445 		return (ESRCH);
2446 
2447 	error = kern_sched_rr_get_interval_td(td, tdt, &ts);
2448 	PROC_UNLOCK(tdt->td_proc);
2449 	if (error != 0)
2450 		return (error);
2451 	native_to_linux_timespec(&lts, &ts);
2452 	return (copyout(&lts, uap->interval, sizeof(lts)));
2453 }
2454 
2455 /*
2456  * In case when the Linux thread is the initial thread in
2457  * the thread group thread id is equal to the process id.
2458  * Glibc depends on this magic (assert in pthread_getattr_np.c).
2459  */
2460 struct thread *
2461 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid)
2462 {
2463 	struct linux_emuldata *em;
2464 	struct thread *tdt;
2465 	struct proc *p;
2466 
2467 	tdt = NULL;
2468 	if (tid == 0 || tid == td->td_tid) {
2469 		tdt = td;
2470 		PROC_LOCK(tdt->td_proc);
2471 	} else if (tid > PID_MAX)
2472 		tdt = tdfind(tid, pid);
2473 	else {
2474 		/*
2475 		 * Initial thread where the tid equal to the pid.
2476 		 */
2477 		p = pfind(tid);
2478 		if (p != NULL) {
2479 			if (SV_PROC_ABI(p) != SV_ABI_LINUX) {
2480 				/*
2481 				 * p is not a Linuxulator process.
2482 				 */
2483 				PROC_UNLOCK(p);
2484 				return (NULL);
2485 			}
2486 			FOREACH_THREAD_IN_PROC(p, tdt) {
2487 				em = em_find(tdt);
2488 				if (tid == em->em_tid)
2489 					return (tdt);
2490 			}
2491 			PROC_UNLOCK(p);
2492 		}
2493 		return (NULL);
2494 	}
2495 
2496 	return (tdt);
2497 }
2498 
2499 void
2500 linux_to_bsd_waitopts(int options, int *bsdopts)
2501 {
2502 
2503 	if (options & LINUX_WNOHANG)
2504 		*bsdopts |= WNOHANG;
2505 	if (options & LINUX_WUNTRACED)
2506 		*bsdopts |= WUNTRACED;
2507 	if (options & LINUX_WEXITED)
2508 		*bsdopts |= WEXITED;
2509 	if (options & LINUX_WCONTINUED)
2510 		*bsdopts |= WCONTINUED;
2511 	if (options & LINUX_WNOWAIT)
2512 		*bsdopts |= WNOWAIT;
2513 
2514 	if (options & __WCLONE)
2515 		*bsdopts |= WLINUXCLONE;
2516 }
2517