xref: /freebsd/sys/compat/linux/linux_misc.c (revision 7d536dc855c85c15bf45f033d108a61b1f3cecc3)
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 	getnanouptime(&ts);
153 	if (ts.tv_nsec != 0)
154 		ts.tv_sec++;
155 	sysinfo.uptime = ts.tv_sec;
156 
157 	/* Use the information from the mib to get our load averages */
158 	for (i = 0; i < 3; i++)
159 		sysinfo.loads[i] = averunnable.ldavg[i] *
160 		    LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
161 
162 	sysinfo.totalram = physmem * PAGE_SIZE;
163 	sysinfo.freeram = sysinfo.totalram - vm_cnt.v_wire_count * PAGE_SIZE;
164 
165 	sysinfo.sharedram = 0;
166 	mtx_lock(&vm_object_list_mtx);
167 	TAILQ_FOREACH(object, &vm_object_list, object_list)
168 		if (object->shadow_count > 1)
169 			sysinfo.sharedram += object->resident_page_count;
170 	mtx_unlock(&vm_object_list_mtx);
171 
172 	sysinfo.sharedram *= PAGE_SIZE;
173 	sysinfo.bufferram = 0;
174 
175 	swap_pager_status(&i, &j);
176 	sysinfo.totalswap = i * PAGE_SIZE;
177 	sysinfo.freeswap = (i - j) * PAGE_SIZE;
178 
179 	sysinfo.procs = nprocs;
180 
181 	/* The following are only present in newer Linux kernels. */
182 	sysinfo.totalbig = 0;
183 	sysinfo.freebig = 0;
184 	sysinfo.mem_unit = 1;
185 
186 	return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
187 }
188 
189 int
190 linux_alarm(struct thread *td, struct linux_alarm_args *args)
191 {
192 	struct itimerval it, old_it;
193 	u_int secs;
194 
195 #ifdef DEBUG
196 	if (ldebug(alarm))
197 		printf(ARGS(alarm, "%u"), args->secs);
198 #endif
199 
200 	secs = args->secs;
201 
202 	if (secs > INT_MAX)
203 		secs = INT_MAX;
204 
205 	it.it_value.tv_sec = (long) secs;
206 	it.it_value.tv_usec = 0;
207 	it.it_interval.tv_sec = 0;
208 	it.it_interval.tv_usec = 0;
209 	/*
210 	 * According to POSIX and Linux implementation
211 	 * the alarm() system call is always successfull.
212 	 * Ignore errors and return 0 as a Linux does.
213 	 */
214 	kern_setitimer(td, ITIMER_REAL, &it, &old_it);
215 	if (timevalisset(&old_it.it_value)) {
216 		if (old_it.it_value.tv_usec != 0)
217 			old_it.it_value.tv_sec++;
218 		td->td_retval[0] = old_it.it_value.tv_sec;
219 	}
220 	return (0);
221 }
222 
223 int
224 linux_brk(struct thread *td, struct linux_brk_args *args)
225 {
226 	struct vmspace *vm = td->td_proc->p_vmspace;
227 	vm_offset_t new, old;
228 	struct obreak_args /* {
229 		char * nsize;
230 	} */ tmp;
231 
232 #ifdef DEBUG
233 	if (ldebug(brk))
234 		printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
235 #endif
236 	old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize);
237 	new = (vm_offset_t)args->dsend;
238 	tmp.nsize = (char *)new;
239 	if (((caddr_t)new > vm->vm_daddr) && !sys_obreak(td, &tmp))
240 		td->td_retval[0] = (long)new;
241 	else
242 		td->td_retval[0] = (long)old;
243 
244 	return (0);
245 }
246 
247 #if defined(__i386__)
248 /* XXX: what about amd64/linux32? */
249 
250 int
251 linux_uselib(struct thread *td, struct linux_uselib_args *args)
252 {
253 	struct nameidata ni;
254 	struct vnode *vp;
255 	struct exec *a_out;
256 	struct vattr attr;
257 	vm_offset_t vmaddr;
258 	unsigned long file_offset;
259 	unsigned long bss_size;
260 	char *library;
261 	ssize_t aresid;
262 	int error, locked, writecount;
263 
264 	LCONVPATHEXIST(td, args->library, &library);
265 
266 #ifdef DEBUG
267 	if (ldebug(uselib))
268 		printf(ARGS(uselib, "%s"), library);
269 #endif
270 
271 	a_out = NULL;
272 	locked = 0;
273 	vp = NULL;
274 
275 	NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
276 	    UIO_SYSSPACE, library, td);
277 	error = namei(&ni);
278 	LFREEPATH(library);
279 	if (error)
280 		goto cleanup;
281 
282 	vp = ni.ni_vp;
283 	NDFREE(&ni, NDF_ONLY_PNBUF);
284 
285 	/*
286 	 * From here on down, we have a locked vnode that must be unlocked.
287 	 * XXX: The code below largely duplicates exec_check_permissions().
288 	 */
289 	locked = 1;
290 
291 	/* Writable? */
292 	error = VOP_GET_WRITECOUNT(vp, &writecount);
293 	if (error != 0)
294 		goto cleanup;
295 	if (writecount != 0) {
296 		error = ETXTBSY;
297 		goto cleanup;
298 	}
299 
300 	/* Executable? */
301 	error = VOP_GETATTR(vp, &attr, td->td_ucred);
302 	if (error)
303 		goto cleanup;
304 
305 	if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
306 	    ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
307 		/* EACCESS is what exec(2) returns. */
308 		error = ENOEXEC;
309 		goto cleanup;
310 	}
311 
312 	/* Sensible size? */
313 	if (attr.va_size == 0) {
314 		error = ENOEXEC;
315 		goto cleanup;
316 	}
317 
318 	/* Can we access it? */
319 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
320 	if (error)
321 		goto cleanup;
322 
323 	/*
324 	 * XXX: This should use vn_open() so that it is properly authorized,
325 	 * and to reduce code redundancy all over the place here.
326 	 * XXX: Not really, it duplicates far more of exec_check_permissions()
327 	 * than vn_open().
328 	 */
329 #ifdef MAC
330 	error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
331 	if (error)
332 		goto cleanup;
333 #endif
334 	error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
335 	if (error)
336 		goto cleanup;
337 
338 	/* Pull in executable header into exec_map */
339 	error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
340 	    VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
341 	if (error)
342 		goto cleanup;
343 
344 	/* Is it a Linux binary ? */
345 	if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
346 		error = ENOEXEC;
347 		goto cleanup;
348 	}
349 
350 	/*
351 	 * While we are here, we should REALLY do some more checks
352 	 */
353 
354 	/* Set file/virtual offset based on a.out variant. */
355 	switch ((int)(a_out->a_magic & 0xffff)) {
356 	case 0413:			/* ZMAGIC */
357 		file_offset = 1024;
358 		break;
359 	case 0314:			/* QMAGIC */
360 		file_offset = 0;
361 		break;
362 	default:
363 		error = ENOEXEC;
364 		goto cleanup;
365 	}
366 
367 	bss_size = round_page(a_out->a_bss);
368 
369 	/* Check various fields in header for validity/bounds. */
370 	if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
371 		error = ENOEXEC;
372 		goto cleanup;
373 	}
374 
375 	/* text + data can't exceed file size */
376 	if (a_out->a_data + a_out->a_text > attr.va_size) {
377 		error = EFAULT;
378 		goto cleanup;
379 	}
380 
381 	/*
382 	 * text/data/bss must not exceed limits
383 	 * XXX - this is not complete. it should check current usage PLUS
384 	 * the resources needed by this library.
385 	 */
386 	PROC_LOCK(td->td_proc);
387 	if (a_out->a_text > maxtsiz ||
388 	    a_out->a_data + bss_size > lim_cur_proc(td->td_proc, RLIMIT_DATA) ||
389 	    racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
390 	    bss_size) != 0) {
391 		PROC_UNLOCK(td->td_proc);
392 		error = ENOMEM;
393 		goto cleanup;
394 	}
395 	PROC_UNLOCK(td->td_proc);
396 
397 	/*
398 	 * Prevent more writers.
399 	 * XXX: Note that if any of the VM operations fail below we don't
400 	 * clear this flag.
401 	 */
402 	VOP_SET_TEXT(vp);
403 
404 	/*
405 	 * Lock no longer needed
406 	 */
407 	locked = 0;
408 	VOP_UNLOCK(vp, 0);
409 
410 	/*
411 	 * Check if file_offset page aligned. Currently we cannot handle
412 	 * misalinged file offsets, and so we read in the entire image
413 	 * (what a waste).
414 	 */
415 	if (file_offset & PAGE_MASK) {
416 #ifdef DEBUG
417 		printf("uselib: Non page aligned binary %lu\n", file_offset);
418 #endif
419 		/* Map text+data read/write/execute */
420 
421 		/* a_entry is the load address and is page aligned */
422 		vmaddr = trunc_page(a_out->a_entry);
423 
424 		/* get anon user mapping, read+write+execute */
425 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
426 		    &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
427 		    VM_PROT_ALL, VM_PROT_ALL, 0);
428 		if (error)
429 			goto cleanup;
430 
431 		error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
432 		    a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
433 		    td->td_ucred, NOCRED, &aresid, td);
434 		if (error != 0)
435 			goto cleanup;
436 		if (aresid != 0) {
437 			error = ENOEXEC;
438 			goto cleanup;
439 		}
440 	} else {
441 #ifdef DEBUG
442 		printf("uselib: Page aligned binary %lu\n", file_offset);
443 #endif
444 		/*
445 		 * for QMAGIC, a_entry is 20 bytes beyond the load address
446 		 * to skip the executable header
447 		 */
448 		vmaddr = trunc_page(a_out->a_entry);
449 
450 		/*
451 		 * Map it all into the process's space as a single
452 		 * copy-on-write "data" segment.
453 		 */
454 		error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr,
455 		    a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
456 		    MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
457 		if (error)
458 			goto cleanup;
459 	}
460 #ifdef DEBUG
461 	printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
462 	    ((long *)vmaddr)[1]);
463 #endif
464 	if (bss_size != 0) {
465 		/* Calculate BSS start address */
466 		vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
467 		    a_out->a_data;
468 
469 		/* allocate some 'anon' space */
470 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
471 		    &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
472 		    VM_PROT_ALL, 0);
473 		if (error)
474 			goto cleanup;
475 	}
476 
477 cleanup:
478 	/* Unlock vnode if needed */
479 	if (locked)
480 		VOP_UNLOCK(vp, 0);
481 
482 	/* Release the temporary mapping. */
483 	if (a_out)
484 		kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
485 
486 	return (error);
487 }
488 
489 #endif	/* __i386__ */
490 
491 int
492 linux_select(struct thread *td, struct linux_select_args *args)
493 {
494 	l_timeval ltv;
495 	struct timeval tv0, tv1, utv, *tvp;
496 	int error;
497 
498 #ifdef DEBUG
499 	if (ldebug(select))
500 		printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
501 		    (void *)args->readfds, (void *)args->writefds,
502 		    (void *)args->exceptfds, (void *)args->timeout);
503 #endif
504 
505 	/*
506 	 * Store current time for computation of the amount of
507 	 * time left.
508 	 */
509 	if (args->timeout) {
510 		if ((error = copyin(args->timeout, &ltv, sizeof(ltv))))
511 			goto select_out;
512 		utv.tv_sec = ltv.tv_sec;
513 		utv.tv_usec = ltv.tv_usec;
514 #ifdef DEBUG
515 		if (ldebug(select))
516 			printf(LMSG("incoming timeout (%jd/%ld)"),
517 			    (intmax_t)utv.tv_sec, utv.tv_usec);
518 #endif
519 
520 		if (itimerfix(&utv)) {
521 			/*
522 			 * The timeval was invalid.  Convert it to something
523 			 * valid that will act as it does under Linux.
524 			 */
525 			utv.tv_sec += utv.tv_usec / 1000000;
526 			utv.tv_usec %= 1000000;
527 			if (utv.tv_usec < 0) {
528 				utv.tv_sec -= 1;
529 				utv.tv_usec += 1000000;
530 			}
531 			if (utv.tv_sec < 0)
532 				timevalclear(&utv);
533 		}
534 		microtime(&tv0);
535 		tvp = &utv;
536 	} else
537 		tvp = NULL;
538 
539 	error = kern_select(td, args->nfds, args->readfds, args->writefds,
540 	    args->exceptfds, tvp, LINUX_NFDBITS);
541 
542 #ifdef DEBUG
543 	if (ldebug(select))
544 		printf(LMSG("real select returns %d"), error);
545 #endif
546 	if (error)
547 		goto select_out;
548 
549 	if (args->timeout) {
550 		if (td->td_retval[0]) {
551 			/*
552 			 * Compute how much time was left of the timeout,
553 			 * by subtracting the current time and the time
554 			 * before we started the call, and subtracting
555 			 * that result from the user-supplied value.
556 			 */
557 			microtime(&tv1);
558 			timevalsub(&tv1, &tv0);
559 			timevalsub(&utv, &tv1);
560 			if (utv.tv_sec < 0)
561 				timevalclear(&utv);
562 		} else
563 			timevalclear(&utv);
564 #ifdef DEBUG
565 		if (ldebug(select))
566 			printf(LMSG("outgoing timeout (%jd/%ld)"),
567 			    (intmax_t)utv.tv_sec, utv.tv_usec);
568 #endif
569 		ltv.tv_sec = utv.tv_sec;
570 		ltv.tv_usec = utv.tv_usec;
571 		if ((error = copyout(&ltv, args->timeout, sizeof(ltv))))
572 			goto select_out;
573 	}
574 
575 select_out:
576 #ifdef DEBUG
577 	if (ldebug(select))
578 		printf(LMSG("select_out -> %d"), error);
579 #endif
580 	return (error);
581 }
582 
583 int
584 linux_mremap(struct thread *td, struct linux_mremap_args *args)
585 {
586 	struct munmap_args /* {
587 		void *addr;
588 		size_t len;
589 	} */ bsd_args;
590 	int error = 0;
591 
592 #ifdef DEBUG
593 	if (ldebug(mremap))
594 		printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
595 		    (void *)(uintptr_t)args->addr,
596 		    (unsigned long)args->old_len,
597 		    (unsigned long)args->new_len,
598 		    (unsigned long)args->flags);
599 #endif
600 
601 	if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
602 		td->td_retval[0] = 0;
603 		return (EINVAL);
604 	}
605 
606 	/*
607 	 * Check for the page alignment.
608 	 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
609 	 */
610 	if (args->addr & PAGE_MASK) {
611 		td->td_retval[0] = 0;
612 		return (EINVAL);
613 	}
614 
615 	args->new_len = round_page(args->new_len);
616 	args->old_len = round_page(args->old_len);
617 
618 	if (args->new_len > args->old_len) {
619 		td->td_retval[0] = 0;
620 		return (ENOMEM);
621 	}
622 
623 	if (args->new_len < args->old_len) {
624 		bsd_args.addr =
625 		    (caddr_t)((uintptr_t)args->addr + args->new_len);
626 		bsd_args.len = args->old_len - args->new_len;
627 		error = sys_munmap(td, &bsd_args);
628 	}
629 
630 	td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
631 	return (error);
632 }
633 
634 #define LINUX_MS_ASYNC       0x0001
635 #define LINUX_MS_INVALIDATE  0x0002
636 #define LINUX_MS_SYNC        0x0004
637 
638 int
639 linux_msync(struct thread *td, struct linux_msync_args *args)
640 {
641 	struct msync_args bsd_args;
642 
643 	bsd_args.addr = (caddr_t)(uintptr_t)args->addr;
644 	bsd_args.len = (uintptr_t)args->len;
645 	bsd_args.flags = args->fl & ~LINUX_MS_SYNC;
646 
647 	return (sys_msync(td, &bsd_args));
648 }
649 
650 int
651 linux_time(struct thread *td, struct linux_time_args *args)
652 {
653 	struct timeval tv;
654 	l_time_t tm;
655 	int error;
656 
657 #ifdef DEBUG
658 	if (ldebug(time))
659 		printf(ARGS(time, "*"));
660 #endif
661 
662 	microtime(&tv);
663 	tm = tv.tv_sec;
664 	if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
665 		return (error);
666 	td->td_retval[0] = tm;
667 	return (0);
668 }
669 
670 struct l_times_argv {
671 	l_clock_t	tms_utime;
672 	l_clock_t	tms_stime;
673 	l_clock_t	tms_cutime;
674 	l_clock_t	tms_cstime;
675 };
676 
677 
678 /*
679  * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
680  * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
681  * auxiliary vector entry.
682  */
683 #define	CLK_TCK		100
684 
685 #define	CONVOTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
686 #define	CONVNTCK(r)	(r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
687 
688 #define	CONVTCK(r)	(linux_kernver(td) >= LINUX_KERNVER_2004000 ?		\
689 			    CONVNTCK(r) : CONVOTCK(r))
690 
691 int
692 linux_times(struct thread *td, struct linux_times_args *args)
693 {
694 	struct timeval tv, utime, stime, cutime, cstime;
695 	struct l_times_argv tms;
696 	struct proc *p;
697 	int error;
698 
699 #ifdef DEBUG
700 	if (ldebug(times))
701 		printf(ARGS(times, "*"));
702 #endif
703 
704 	if (args->buf != NULL) {
705 		p = td->td_proc;
706 		PROC_LOCK(p);
707 		PROC_STATLOCK(p);
708 		calcru(p, &utime, &stime);
709 		PROC_STATUNLOCK(p);
710 		calccru(p, &cutime, &cstime);
711 		PROC_UNLOCK(p);
712 
713 		tms.tms_utime = CONVTCK(utime);
714 		tms.tms_stime = CONVTCK(stime);
715 
716 		tms.tms_cutime = CONVTCK(cutime);
717 		tms.tms_cstime = CONVTCK(cstime);
718 
719 		if ((error = copyout(&tms, args->buf, sizeof(tms))))
720 			return (error);
721 	}
722 
723 	microuptime(&tv);
724 	td->td_retval[0] = (int)CONVTCK(tv);
725 	return (0);
726 }
727 
728 int
729 linux_newuname(struct thread *td, struct linux_newuname_args *args)
730 {
731 	struct l_new_utsname utsname;
732 	char osname[LINUX_MAX_UTSNAME];
733 	char osrelease[LINUX_MAX_UTSNAME];
734 	char *p;
735 
736 #ifdef DEBUG
737 	if (ldebug(newuname))
738 		printf(ARGS(newuname, "*"));
739 #endif
740 
741 	linux_get_osname(td, osname);
742 	linux_get_osrelease(td, osrelease);
743 
744 	bzero(&utsname, sizeof(utsname));
745 	strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
746 	getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
747 	getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
748 	strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
749 	strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
750 	for (p = utsname.version; *p != '\0'; ++p)
751 		if (*p == '\n') {
752 			*p = '\0';
753 			break;
754 		}
755 	strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME);
756 
757 	return (copyout(&utsname, args->buf, sizeof(utsname)));
758 }
759 
760 struct l_utimbuf {
761 	l_time_t l_actime;
762 	l_time_t l_modtime;
763 };
764 
765 int
766 linux_utime(struct thread *td, struct linux_utime_args *args)
767 {
768 	struct timeval tv[2], *tvp;
769 	struct l_utimbuf lut;
770 	char *fname;
771 	int error;
772 
773 	LCONVPATHEXIST(td, args->fname, &fname);
774 
775 #ifdef DEBUG
776 	if (ldebug(utime))
777 		printf(ARGS(utime, "%s, *"), fname);
778 #endif
779 
780 	if (args->times) {
781 		if ((error = copyin(args->times, &lut, sizeof lut))) {
782 			LFREEPATH(fname);
783 			return (error);
784 		}
785 		tv[0].tv_sec = lut.l_actime;
786 		tv[0].tv_usec = 0;
787 		tv[1].tv_sec = lut.l_modtime;
788 		tv[1].tv_usec = 0;
789 		tvp = tv;
790 	} else
791 		tvp = NULL;
792 
793 	error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, tvp,
794 	    UIO_SYSSPACE);
795 	LFREEPATH(fname);
796 	return (error);
797 }
798 
799 int
800 linux_utimes(struct thread *td, struct linux_utimes_args *args)
801 {
802 	l_timeval ltv[2];
803 	struct timeval tv[2], *tvp = NULL;
804 	char *fname;
805 	int error;
806 
807 	LCONVPATHEXIST(td, args->fname, &fname);
808 
809 #ifdef DEBUG
810 	if (ldebug(utimes))
811 		printf(ARGS(utimes, "%s, *"), fname);
812 #endif
813 
814 	if (args->tptr != NULL) {
815 		if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
816 			LFREEPATH(fname);
817 			return (error);
818 		}
819 		tv[0].tv_sec = ltv[0].tv_sec;
820 		tv[0].tv_usec = ltv[0].tv_usec;
821 		tv[1].tv_sec = ltv[1].tv_sec;
822 		tv[1].tv_usec = ltv[1].tv_usec;
823 		tvp = tv;
824 	}
825 
826 	error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE,
827 	    tvp, UIO_SYSSPACE);
828 	LFREEPATH(fname);
829 	return (error);
830 }
831 
832 static int
833 linux_utimensat_nsec_valid(l_long nsec)
834 {
835 
836 	if (nsec == LINUX_UTIME_OMIT || nsec == LINUX_UTIME_NOW)
837 		return (0);
838 	if (nsec >= 0 && nsec <= 999999999)
839 		return (0);
840 	return (1);
841 }
842 
843 int
844 linux_utimensat(struct thread *td, struct linux_utimensat_args *args)
845 {
846 	struct l_timespec l_times[2];
847 	struct timespec times[2], *timesp = NULL;
848 	char *path = NULL;
849 	int error, dfd, flags = 0;
850 
851 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
852 
853 #ifdef DEBUG
854 	if (ldebug(utimensat))
855 		printf(ARGS(utimensat, "%d, *"), dfd);
856 #endif
857 
858 	if (args->flags & ~LINUX_AT_SYMLINK_NOFOLLOW)
859 		return (EINVAL);
860 
861 	if (args->times != NULL) {
862 		error = copyin(args->times, l_times, sizeof(l_times));
863 		if (error != 0)
864 			return (error);
865 
866 		if (linux_utimensat_nsec_valid(l_times[0].tv_nsec) != 0 ||
867 		    linux_utimensat_nsec_valid(l_times[1].tv_nsec) != 0)
868 			return (EINVAL);
869 
870 		times[0].tv_sec = l_times[0].tv_sec;
871 		switch (l_times[0].tv_nsec)
872 		{
873 		case LINUX_UTIME_OMIT:
874 			times[0].tv_nsec = UTIME_OMIT;
875 			break;
876 		case LINUX_UTIME_NOW:
877 			times[0].tv_nsec = UTIME_NOW;
878 			break;
879 		default:
880 			times[0].tv_nsec = l_times[0].tv_nsec;
881 		}
882 
883 		times[1].tv_sec = l_times[1].tv_sec;
884 		switch (l_times[1].tv_nsec)
885 		{
886 		case LINUX_UTIME_OMIT:
887 			times[1].tv_nsec = UTIME_OMIT;
888 			break;
889 		case LINUX_UTIME_NOW:
890 			times[1].tv_nsec = UTIME_NOW;
891 			break;
892 		default:
893 			times[1].tv_nsec = l_times[1].tv_nsec;
894 			break;
895 		}
896 		timesp = times;
897 	}
898 
899 	if (times[0].tv_nsec == UTIME_OMIT && times[1].tv_nsec == UTIME_OMIT)
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 		return (0);
904 
905 	if (args->pathname != NULL)
906 		LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
907 	else if (args->flags != 0)
908 		return (EINVAL);
909 
910 	if (args->flags & LINUX_AT_SYMLINK_NOFOLLOW)
911 		flags |= AT_SYMLINK_NOFOLLOW;
912 
913 	if (path == NULL)
914 		error = kern_futimens(td, dfd, timesp, UIO_SYSSPACE);
915 	else {
916 		error = kern_utimensat(td, dfd, path, UIO_SYSSPACE, timesp,
917 	    		UIO_SYSSPACE, flags);
918 		LFREEPATH(path);
919 	}
920 
921 	return (error);
922 }
923 
924 int
925 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
926 {
927 	l_timeval ltv[2];
928 	struct timeval tv[2], *tvp = NULL;
929 	char *fname;
930 	int error, dfd;
931 
932 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
933 	LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
934 
935 #ifdef DEBUG
936 	if (ldebug(futimesat))
937 		printf(ARGS(futimesat, "%s, *"), fname);
938 #endif
939 
940 	if (args->utimes != NULL) {
941 		if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
942 			LFREEPATH(fname);
943 			return (error);
944 		}
945 		tv[0].tv_sec = ltv[0].tv_sec;
946 		tv[0].tv_usec = ltv[0].tv_usec;
947 		tv[1].tv_sec = ltv[1].tv_sec;
948 		tv[1].tv_usec = ltv[1].tv_usec;
949 		tvp = tv;
950 	}
951 
952 	error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
953 	LFREEPATH(fname);
954 	return (error);
955 }
956 
957 int
958 linux_common_wait(struct thread *td, int pid, int *status,
959     int options, struct rusage *ru)
960 {
961 	int error, tmpstat;
962 
963 	error = kern_wait(td, pid, &tmpstat, options, ru);
964 	if (error)
965 		return (error);
966 
967 	if (status) {
968 		tmpstat &= 0xffff;
969 		if (WIFSIGNALED(tmpstat))
970 			tmpstat = (tmpstat & 0xffffff80) |
971 			    bsd_to_linux_signal(WTERMSIG(tmpstat));
972 		else if (WIFSTOPPED(tmpstat))
973 			tmpstat = (tmpstat & 0xffff00ff) |
974 			    (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8);
975 		else if (WIFCONTINUED(tmpstat))
976 			tmpstat = 0xffff;
977 		error = copyout(&tmpstat, status, sizeof(int));
978 	}
979 
980 	return (error);
981 }
982 
983 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
984 int
985 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
986 {
987 	struct linux_wait4_args wait4_args;
988 
989 #ifdef DEBUG
990 	if (ldebug(waitpid))
991 		printf(ARGS(waitpid, "%d, %p, %d"),
992 		    args->pid, (void *)args->status, args->options);
993 #endif
994 
995 	wait4_args.pid = args->pid;
996 	wait4_args.status = args->status;
997 	wait4_args.options = args->options;
998 	wait4_args.rusage = NULL;
999 
1000 	return (linux_wait4(td, &wait4_args));
1001 }
1002 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1003 
1004 int
1005 linux_wait4(struct thread *td, struct linux_wait4_args *args)
1006 {
1007 	int error, options;
1008 	struct rusage ru, *rup;
1009 
1010 #ifdef DEBUG
1011 	if (ldebug(wait4))
1012 		printf(ARGS(wait4, "%d, %p, %d, %p"),
1013 		    args->pid, (void *)args->status, args->options,
1014 		    (void *)args->rusage);
1015 #endif
1016 	if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG |
1017 	    LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
1018 		return (EINVAL);
1019 
1020 	options = WEXITED;
1021 	linux_to_bsd_waitopts(args->options, &options);
1022 
1023 	if (args->rusage != NULL)
1024 		rup = &ru;
1025 	else
1026 		rup = NULL;
1027 	error = linux_common_wait(td, args->pid, args->status, options, rup);
1028 	if (error != 0)
1029 		return (error);
1030 	if (args->rusage != NULL)
1031 		error = linux_copyout_rusage(&ru, args->rusage);
1032 	return (error);
1033 }
1034 
1035 int
1036 linux_waitid(struct thread *td, struct linux_waitid_args *args)
1037 {
1038 	int status, options, sig;
1039 	struct __wrusage wru;
1040 	siginfo_t siginfo;
1041 	l_siginfo_t lsi;
1042 	idtype_t idtype;
1043 	struct proc *p;
1044 	int error;
1045 
1046 	options = 0;
1047 	linux_to_bsd_waitopts(args->options, &options);
1048 
1049 	if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED))
1050 		return (EINVAL);
1051 	if (!(options & (WEXITED | WUNTRACED | WCONTINUED)))
1052 		return (EINVAL);
1053 
1054 	switch (args->idtype) {
1055 	case LINUX_P_ALL:
1056 		idtype = P_ALL;
1057 		break;
1058 	case LINUX_P_PID:
1059 		if (args->id <= 0)
1060 			return (EINVAL);
1061 		idtype = P_PID;
1062 		break;
1063 	case LINUX_P_PGID:
1064 		if (args->id <= 0)
1065 			return (EINVAL);
1066 		idtype = P_PGID;
1067 		break;
1068 	default:
1069 		return (EINVAL);
1070 	}
1071 
1072 	error = kern_wait6(td, idtype, args->id, &status, options,
1073 	    &wru, &siginfo);
1074 	if (error != 0)
1075 		return (error);
1076 	if (args->rusage != NULL) {
1077 		error = linux_copyout_rusage(&wru.wru_children,
1078 		    args->rusage);
1079 		if (error != 0)
1080 			return (error);
1081 	}
1082 	if (args->info != NULL) {
1083 		p = td->td_proc;
1084 		if (td->td_retval[0] == 0)
1085 			bzero(&lsi, sizeof(lsi));
1086 		else {
1087 			sig = bsd_to_linux_signal(siginfo.si_signo);
1088 			siginfo_to_lsiginfo(&siginfo, &lsi, sig);
1089 		}
1090 		error = copyout(&lsi, args->info, sizeof(lsi));
1091 	}
1092 	td->td_retval[0] = 0;
1093 
1094 	return (error);
1095 }
1096 
1097 int
1098 linux_mknod(struct thread *td, struct linux_mknod_args *args)
1099 {
1100 	char *path;
1101 	int error;
1102 
1103 	LCONVPATHCREAT(td, args->path, &path);
1104 
1105 #ifdef DEBUG
1106 	if (ldebug(mknod))
1107 		printf(ARGS(mknod, "%s, %d, %ju"), path, args->mode,
1108 		    (uintmax_t)args->dev);
1109 #endif
1110 
1111 	switch (args->mode & S_IFMT) {
1112 	case S_IFIFO:
1113 	case S_IFSOCK:
1114 		error = kern_mkfifoat(td, AT_FDCWD, path, UIO_SYSSPACE,
1115 		    args->mode);
1116 		break;
1117 
1118 	case S_IFCHR:
1119 	case S_IFBLK:
1120 		error = kern_mknodat(td, AT_FDCWD, path, UIO_SYSSPACE,
1121 		    args->mode, args->dev);
1122 		break;
1123 
1124 	case S_IFDIR:
1125 		error = EPERM;
1126 		break;
1127 
1128 	case 0:
1129 		args->mode |= S_IFREG;
1130 		/* FALLTHROUGH */
1131 	case S_IFREG:
1132 		error = kern_openat(td, AT_FDCWD, path, UIO_SYSSPACE,
1133 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1134 		if (error == 0)
1135 			kern_close(td, td->td_retval[0]);
1136 		break;
1137 
1138 	default:
1139 		error = EINVAL;
1140 		break;
1141 	}
1142 	LFREEPATH(path);
1143 	return (error);
1144 }
1145 
1146 int
1147 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
1148 {
1149 	char *path;
1150 	int error, dfd;
1151 
1152 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
1153 	LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
1154 
1155 #ifdef DEBUG
1156 	if (ldebug(mknodat))
1157 		printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev);
1158 #endif
1159 
1160 	switch (args->mode & S_IFMT) {
1161 	case S_IFIFO:
1162 	case S_IFSOCK:
1163 		error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
1164 		break;
1165 
1166 	case S_IFCHR:
1167 	case S_IFBLK:
1168 		error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
1169 		    args->dev);
1170 		break;
1171 
1172 	case S_IFDIR:
1173 		error = EPERM;
1174 		break;
1175 
1176 	case 0:
1177 		args->mode |= S_IFREG;
1178 		/* FALLTHROUGH */
1179 	case S_IFREG:
1180 		error = kern_openat(td, dfd, path, UIO_SYSSPACE,
1181 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1182 		if (error == 0)
1183 			kern_close(td, td->td_retval[0]);
1184 		break;
1185 
1186 	default:
1187 		error = EINVAL;
1188 		break;
1189 	}
1190 	LFREEPATH(path);
1191 	return (error);
1192 }
1193 
1194 /*
1195  * UGH! This is just about the dumbest idea I've ever heard!!
1196  */
1197 int
1198 linux_personality(struct thread *td, struct linux_personality_args *args)
1199 {
1200 #ifdef DEBUG
1201 	if (ldebug(personality))
1202 		printf(ARGS(personality, "%lu"), (unsigned long)args->per);
1203 #endif
1204 	if (args->per != 0)
1205 		return (EINVAL);
1206 
1207 	/* Yes Jim, it's still a Linux... */
1208 	td->td_retval[0] = 0;
1209 	return (0);
1210 }
1211 
1212 struct l_itimerval {
1213 	l_timeval it_interval;
1214 	l_timeval it_value;
1215 };
1216 
1217 #define	B2L_ITIMERVAL(bip, lip) 					\
1218 	(bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec;		\
1219 	(bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec;	\
1220 	(bip)->it_value.tv_sec = (lip)->it_value.tv_sec;		\
1221 	(bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1222 
1223 int
1224 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1225 {
1226 	int error;
1227 	struct l_itimerval ls;
1228 	struct itimerval aitv, oitv;
1229 
1230 #ifdef DEBUG
1231 	if (ldebug(setitimer))
1232 		printf(ARGS(setitimer, "%p, %p"),
1233 		    (void *)uap->itv, (void *)uap->oitv);
1234 #endif
1235 
1236 	if (uap->itv == NULL) {
1237 		uap->itv = uap->oitv;
1238 		return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1239 	}
1240 
1241 	error = copyin(uap->itv, &ls, sizeof(ls));
1242 	if (error != 0)
1243 		return (error);
1244 	B2L_ITIMERVAL(&aitv, &ls);
1245 #ifdef DEBUG
1246 	if (ldebug(setitimer)) {
1247 		printf("setitimer: value: sec: %jd, usec: %ld\n",
1248 		    (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1249 		printf("setitimer: interval: sec: %jd, usec: %ld\n",
1250 		    (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1251 	}
1252 #endif
1253 	error = kern_setitimer(td, uap->which, &aitv, &oitv);
1254 	if (error != 0 || uap->oitv == NULL)
1255 		return (error);
1256 	B2L_ITIMERVAL(&ls, &oitv);
1257 
1258 	return (copyout(&ls, uap->oitv, sizeof(ls)));
1259 }
1260 
1261 int
1262 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1263 {
1264 	int error;
1265 	struct l_itimerval ls;
1266 	struct itimerval aitv;
1267 
1268 #ifdef DEBUG
1269 	if (ldebug(getitimer))
1270 		printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1271 #endif
1272 	error = kern_getitimer(td, uap->which, &aitv);
1273 	if (error != 0)
1274 		return (error);
1275 	B2L_ITIMERVAL(&ls, &aitv);
1276 	return (copyout(&ls, uap->itv, sizeof(ls)));
1277 }
1278 
1279 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1280 int
1281 linux_nice(struct thread *td, struct linux_nice_args *args)
1282 {
1283 	struct setpriority_args bsd_args;
1284 
1285 	bsd_args.which = PRIO_PROCESS;
1286 	bsd_args.who = 0;		/* current process */
1287 	bsd_args.prio = args->inc;
1288 	return (sys_setpriority(td, &bsd_args));
1289 }
1290 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1291 
1292 int
1293 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1294 {
1295 	struct ucred *newcred, *oldcred;
1296 	l_gid_t *linux_gidset;
1297 	gid_t *bsd_gidset;
1298 	int ngrp, error;
1299 	struct proc *p;
1300 
1301 	ngrp = args->gidsetsize;
1302 	if (ngrp < 0 || ngrp >= ngroups_max + 1)
1303 		return (EINVAL);
1304 	linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK);
1305 	error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1306 	if (error)
1307 		goto out;
1308 	newcred = crget();
1309 	crextend(newcred, ngrp + 1);
1310 	p = td->td_proc;
1311 	PROC_LOCK(p);
1312 	oldcred = p->p_ucred;
1313 	crcopy(newcred, oldcred);
1314 
1315 	/*
1316 	 * cr_groups[0] holds egid. Setting the whole set from
1317 	 * the supplied set will cause egid to be changed too.
1318 	 * Keep cr_groups[0] unchanged to prevent that.
1319 	 */
1320 
1321 	if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) {
1322 		PROC_UNLOCK(p);
1323 		crfree(newcred);
1324 		goto out;
1325 	}
1326 
1327 	if (ngrp > 0) {
1328 		newcred->cr_ngroups = ngrp + 1;
1329 
1330 		bsd_gidset = newcred->cr_groups;
1331 		ngrp--;
1332 		while (ngrp >= 0) {
1333 			bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1334 			ngrp--;
1335 		}
1336 	} else
1337 		newcred->cr_ngroups = 1;
1338 
1339 	setsugid(p);
1340 	proc_set_cred(p, newcred);
1341 	PROC_UNLOCK(p);
1342 	crfree(oldcred);
1343 	error = 0;
1344 out:
1345 	free(linux_gidset, M_LINUX);
1346 	return (error);
1347 }
1348 
1349 int
1350 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1351 {
1352 	struct ucred *cred;
1353 	l_gid_t *linux_gidset;
1354 	gid_t *bsd_gidset;
1355 	int bsd_gidsetsz, ngrp, error;
1356 
1357 	cred = td->td_ucred;
1358 	bsd_gidset = cred->cr_groups;
1359 	bsd_gidsetsz = cred->cr_ngroups - 1;
1360 
1361 	/*
1362 	 * cr_groups[0] holds egid. Returning the whole set
1363 	 * here will cause a duplicate. Exclude cr_groups[0]
1364 	 * to prevent that.
1365 	 */
1366 
1367 	if ((ngrp = args->gidsetsize) == 0) {
1368 		td->td_retval[0] = bsd_gidsetsz;
1369 		return (0);
1370 	}
1371 
1372 	if (ngrp < bsd_gidsetsz)
1373 		return (EINVAL);
1374 
1375 	ngrp = 0;
1376 	linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1377 	    M_LINUX, M_WAITOK);
1378 	while (ngrp < bsd_gidsetsz) {
1379 		linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1380 		ngrp++;
1381 	}
1382 
1383 	error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1384 	free(linux_gidset, M_LINUX);
1385 	if (error)
1386 		return (error);
1387 
1388 	td->td_retval[0] = ngrp;
1389 	return (0);
1390 }
1391 
1392 int
1393 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1394 {
1395 	struct rlimit bsd_rlim;
1396 	struct l_rlimit rlim;
1397 	u_int which;
1398 	int error;
1399 
1400 #ifdef DEBUG
1401 	if (ldebug(setrlimit))
1402 		printf(ARGS(setrlimit, "%d, %p"),
1403 		    args->resource, (void *)args->rlim);
1404 #endif
1405 
1406 	if (args->resource >= LINUX_RLIM_NLIMITS)
1407 		return (EINVAL);
1408 
1409 	which = linux_to_bsd_resource[args->resource];
1410 	if (which == -1)
1411 		return (EINVAL);
1412 
1413 	error = copyin(args->rlim, &rlim, sizeof(rlim));
1414 	if (error)
1415 		return (error);
1416 
1417 	bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1418 	bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1419 	return (kern_setrlimit(td, which, &bsd_rlim));
1420 }
1421 
1422 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1423 int
1424 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1425 {
1426 	struct l_rlimit rlim;
1427 	struct rlimit bsd_rlim;
1428 	u_int which;
1429 
1430 #ifdef DEBUG
1431 	if (ldebug(old_getrlimit))
1432 		printf(ARGS(old_getrlimit, "%d, %p"),
1433 		    args->resource, (void *)args->rlim);
1434 #endif
1435 
1436 	if (args->resource >= LINUX_RLIM_NLIMITS)
1437 		return (EINVAL);
1438 
1439 	which = linux_to_bsd_resource[args->resource];
1440 	if (which == -1)
1441 		return (EINVAL);
1442 
1443 	lim_rlimit(td, which, &bsd_rlim);
1444 
1445 #ifdef COMPAT_LINUX32
1446 	rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1447 	if (rlim.rlim_cur == UINT_MAX)
1448 		rlim.rlim_cur = INT_MAX;
1449 	rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1450 	if (rlim.rlim_max == UINT_MAX)
1451 		rlim.rlim_max = INT_MAX;
1452 #else
1453 	rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1454 	if (rlim.rlim_cur == ULONG_MAX)
1455 		rlim.rlim_cur = LONG_MAX;
1456 	rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1457 	if (rlim.rlim_max == ULONG_MAX)
1458 		rlim.rlim_max = LONG_MAX;
1459 #endif
1460 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1461 }
1462 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1463 
1464 int
1465 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1466 {
1467 	struct l_rlimit rlim;
1468 	struct rlimit bsd_rlim;
1469 	u_int which;
1470 
1471 #ifdef DEBUG
1472 	if (ldebug(getrlimit))
1473 		printf(ARGS(getrlimit, "%d, %p"),
1474 		    args->resource, (void *)args->rlim);
1475 #endif
1476 
1477 	if (args->resource >= LINUX_RLIM_NLIMITS)
1478 		return (EINVAL);
1479 
1480 	which = linux_to_bsd_resource[args->resource];
1481 	if (which == -1)
1482 		return (EINVAL);
1483 
1484 	lim_rlimit(td, which, &bsd_rlim);
1485 
1486 	rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1487 	rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1488 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1489 }
1490 
1491 int
1492 linux_sched_setscheduler(struct thread *td,
1493     struct linux_sched_setscheduler_args *args)
1494 {
1495 	struct sched_param sched_param;
1496 	struct thread *tdt;
1497 	int error, policy;
1498 
1499 #ifdef DEBUG
1500 	if (ldebug(sched_setscheduler))
1501 		printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1502 		    args->pid, args->policy, (const void *)args->param);
1503 #endif
1504 
1505 	switch (args->policy) {
1506 	case LINUX_SCHED_OTHER:
1507 		policy = SCHED_OTHER;
1508 		break;
1509 	case LINUX_SCHED_FIFO:
1510 		policy = SCHED_FIFO;
1511 		break;
1512 	case LINUX_SCHED_RR:
1513 		policy = SCHED_RR;
1514 		break;
1515 	default:
1516 		return (EINVAL);
1517 	}
1518 
1519 	error = copyin(args->param, &sched_param, sizeof(sched_param));
1520 	if (error)
1521 		return (error);
1522 
1523 	tdt = linux_tdfind(td, args->pid, -1);
1524 	if (tdt == NULL)
1525 		return (ESRCH);
1526 
1527 	error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1528 	PROC_UNLOCK(tdt->td_proc);
1529 	return (error);
1530 }
1531 
1532 int
1533 linux_sched_getscheduler(struct thread *td,
1534     struct linux_sched_getscheduler_args *args)
1535 {
1536 	struct thread *tdt;
1537 	int error, policy;
1538 
1539 #ifdef DEBUG
1540 	if (ldebug(sched_getscheduler))
1541 		printf(ARGS(sched_getscheduler, "%d"), args->pid);
1542 #endif
1543 
1544 	tdt = linux_tdfind(td, args->pid, -1);
1545 	if (tdt == NULL)
1546 		return (ESRCH);
1547 
1548 	error = kern_sched_getscheduler(td, tdt, &policy);
1549 	PROC_UNLOCK(tdt->td_proc);
1550 
1551 	switch (policy) {
1552 	case SCHED_OTHER:
1553 		td->td_retval[0] = LINUX_SCHED_OTHER;
1554 		break;
1555 	case SCHED_FIFO:
1556 		td->td_retval[0] = LINUX_SCHED_FIFO;
1557 		break;
1558 	case SCHED_RR:
1559 		td->td_retval[0] = LINUX_SCHED_RR;
1560 		break;
1561 	}
1562 	return (error);
1563 }
1564 
1565 int
1566 linux_sched_get_priority_max(struct thread *td,
1567     struct linux_sched_get_priority_max_args *args)
1568 {
1569 	struct sched_get_priority_max_args bsd;
1570 
1571 #ifdef DEBUG
1572 	if (ldebug(sched_get_priority_max))
1573 		printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1574 #endif
1575 
1576 	switch (args->policy) {
1577 	case LINUX_SCHED_OTHER:
1578 		bsd.policy = SCHED_OTHER;
1579 		break;
1580 	case LINUX_SCHED_FIFO:
1581 		bsd.policy = SCHED_FIFO;
1582 		break;
1583 	case LINUX_SCHED_RR:
1584 		bsd.policy = SCHED_RR;
1585 		break;
1586 	default:
1587 		return (EINVAL);
1588 	}
1589 	return (sys_sched_get_priority_max(td, &bsd));
1590 }
1591 
1592 int
1593 linux_sched_get_priority_min(struct thread *td,
1594     struct linux_sched_get_priority_min_args *args)
1595 {
1596 	struct sched_get_priority_min_args bsd;
1597 
1598 #ifdef DEBUG
1599 	if (ldebug(sched_get_priority_min))
1600 		printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1601 #endif
1602 
1603 	switch (args->policy) {
1604 	case LINUX_SCHED_OTHER:
1605 		bsd.policy = SCHED_OTHER;
1606 		break;
1607 	case LINUX_SCHED_FIFO:
1608 		bsd.policy = SCHED_FIFO;
1609 		break;
1610 	case LINUX_SCHED_RR:
1611 		bsd.policy = SCHED_RR;
1612 		break;
1613 	default:
1614 		return (EINVAL);
1615 	}
1616 	return (sys_sched_get_priority_min(td, &bsd));
1617 }
1618 
1619 #define REBOOT_CAD_ON	0x89abcdef
1620 #define REBOOT_CAD_OFF	0
1621 #define REBOOT_HALT	0xcdef0123
1622 #define REBOOT_RESTART	0x01234567
1623 #define REBOOT_RESTART2	0xA1B2C3D4
1624 #define REBOOT_POWEROFF	0x4321FEDC
1625 #define REBOOT_MAGIC1	0xfee1dead
1626 #define REBOOT_MAGIC2	0x28121969
1627 #define REBOOT_MAGIC2A	0x05121996
1628 #define REBOOT_MAGIC2B	0x16041998
1629 
1630 int
1631 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1632 {
1633 	struct reboot_args bsd_args;
1634 
1635 #ifdef DEBUG
1636 	if (ldebug(reboot))
1637 		printf(ARGS(reboot, "0x%x"), args->cmd);
1638 #endif
1639 
1640 	if (args->magic1 != REBOOT_MAGIC1)
1641 		return (EINVAL);
1642 
1643 	switch (args->magic2) {
1644 	case REBOOT_MAGIC2:
1645 	case REBOOT_MAGIC2A:
1646 	case REBOOT_MAGIC2B:
1647 		break;
1648 	default:
1649 		return (EINVAL);
1650 	}
1651 
1652 	switch (args->cmd) {
1653 	case REBOOT_CAD_ON:
1654 	case REBOOT_CAD_OFF:
1655 		return (priv_check(td, PRIV_REBOOT));
1656 	case REBOOT_HALT:
1657 		bsd_args.opt = RB_HALT;
1658 		break;
1659 	case REBOOT_RESTART:
1660 	case REBOOT_RESTART2:
1661 		bsd_args.opt = 0;
1662 		break;
1663 	case REBOOT_POWEROFF:
1664 		bsd_args.opt = RB_POWEROFF;
1665 		break;
1666 	default:
1667 		return (EINVAL);
1668 	}
1669 	return (sys_reboot(td, &bsd_args));
1670 }
1671 
1672 
1673 /*
1674  * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1675  * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
1676  * are assumed to be preserved. The following lightweight syscalls fixes
1677  * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
1678  *
1679  * linux_getpid() - MP SAFE
1680  * linux_getgid() - MP SAFE
1681  * linux_getuid() - MP SAFE
1682  */
1683 
1684 int
1685 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1686 {
1687 
1688 #ifdef DEBUG
1689 	if (ldebug(getpid))
1690 		printf(ARGS(getpid, ""));
1691 #endif
1692 	td->td_retval[0] = td->td_proc->p_pid;
1693 
1694 	return (0);
1695 }
1696 
1697 int
1698 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1699 {
1700 	struct linux_emuldata *em;
1701 
1702 #ifdef DEBUG
1703 	if (ldebug(gettid))
1704 		printf(ARGS(gettid, ""));
1705 #endif
1706 
1707 	em = em_find(td);
1708 	KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1709 
1710 	td->td_retval[0] = em->em_tid;
1711 
1712 	return (0);
1713 }
1714 
1715 
1716 int
1717 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1718 {
1719 
1720 #ifdef DEBUG
1721 	if (ldebug(getppid))
1722 		printf(ARGS(getppid, ""));
1723 #endif
1724 
1725 	PROC_LOCK(td->td_proc);
1726 	td->td_retval[0] = td->td_proc->p_pptr->p_pid;
1727 	PROC_UNLOCK(td->td_proc);
1728 	return (0);
1729 }
1730 
1731 int
1732 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1733 {
1734 
1735 #ifdef DEBUG
1736 	if (ldebug(getgid))
1737 		printf(ARGS(getgid, ""));
1738 #endif
1739 
1740 	td->td_retval[0] = td->td_ucred->cr_rgid;
1741 	return (0);
1742 }
1743 
1744 int
1745 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1746 {
1747 
1748 #ifdef DEBUG
1749 	if (ldebug(getuid))
1750 		printf(ARGS(getuid, ""));
1751 #endif
1752 
1753 	td->td_retval[0] = td->td_ucred->cr_ruid;
1754 	return (0);
1755 }
1756 
1757 
1758 int
1759 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1760 {
1761 	struct getsid_args bsd;
1762 
1763 #ifdef DEBUG
1764 	if (ldebug(getsid))
1765 		printf(ARGS(getsid, "%i"), args->pid);
1766 #endif
1767 
1768 	bsd.pid = args->pid;
1769 	return (sys_getsid(td, &bsd));
1770 }
1771 
1772 int
1773 linux_nosys(struct thread *td, struct nosys_args *ignore)
1774 {
1775 
1776 	return (ENOSYS);
1777 }
1778 
1779 int
1780 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1781 {
1782 	struct getpriority_args bsd_args;
1783 	int error;
1784 
1785 #ifdef DEBUG
1786 	if (ldebug(getpriority))
1787 		printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
1788 #endif
1789 
1790 	bsd_args.which = args->which;
1791 	bsd_args.who = args->who;
1792 	error = sys_getpriority(td, &bsd_args);
1793 	td->td_retval[0] = 20 - td->td_retval[0];
1794 	return (error);
1795 }
1796 
1797 int
1798 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1799 {
1800 	int name[2];
1801 
1802 #ifdef DEBUG
1803 	if (ldebug(sethostname))
1804 		printf(ARGS(sethostname, "*, %i"), args->len);
1805 #endif
1806 
1807 	name[0] = CTL_KERN;
1808 	name[1] = KERN_HOSTNAME;
1809 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1810 	    args->len, 0, 0));
1811 }
1812 
1813 int
1814 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1815 {
1816 	int name[2];
1817 
1818 #ifdef DEBUG
1819 	if (ldebug(setdomainname))
1820 		printf(ARGS(setdomainname, "*, %i"), args->len);
1821 #endif
1822 
1823 	name[0] = CTL_KERN;
1824 	name[1] = KERN_NISDOMAINNAME;
1825 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1826 	    args->len, 0, 0));
1827 }
1828 
1829 int
1830 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1831 {
1832 
1833 #ifdef DEBUG
1834 	if (ldebug(exit_group))
1835 		printf(ARGS(exit_group, "%i"), args->error_code);
1836 #endif
1837 
1838 	LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1839 	    args->error_code);
1840 
1841 	/*
1842 	 * XXX: we should send a signal to the parent if
1843 	 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1844 	 * as it doesnt occur often.
1845 	 */
1846 	exit1(td, args->error_code, 0);
1847 		/* NOTREACHED */
1848 }
1849 
1850 #define _LINUX_CAPABILITY_VERSION  0x19980330
1851 
1852 struct l_user_cap_header {
1853 	l_int	version;
1854 	l_int	pid;
1855 };
1856 
1857 struct l_user_cap_data {
1858 	l_int	effective;
1859 	l_int	permitted;
1860 	l_int	inheritable;
1861 };
1862 
1863 int
1864 linux_capget(struct thread *td, struct linux_capget_args *args)
1865 {
1866 	struct l_user_cap_header luch;
1867 	struct l_user_cap_data lucd;
1868 	int error;
1869 
1870 	if (args->hdrp == NULL)
1871 		return (EFAULT);
1872 
1873 	error = copyin(args->hdrp, &luch, sizeof(luch));
1874 	if (error != 0)
1875 		return (error);
1876 
1877 	if (luch.version != _LINUX_CAPABILITY_VERSION) {
1878 		luch.version = _LINUX_CAPABILITY_VERSION;
1879 		error = copyout(&luch, args->hdrp, sizeof(luch));
1880 		if (error)
1881 			return (error);
1882 		return (EINVAL);
1883 	}
1884 
1885 	if (luch.pid)
1886 		return (EPERM);
1887 
1888 	if (args->datap) {
1889 		/*
1890 		 * The current implementation doesn't support setting
1891 		 * a capability (it's essentially a stub) so indicate
1892 		 * that no capabilities are currently set or available
1893 		 * to request.
1894 		 */
1895 		bzero (&lucd, sizeof(lucd));
1896 		error = copyout(&lucd, args->datap, sizeof(lucd));
1897 	}
1898 
1899 	return (error);
1900 }
1901 
1902 int
1903 linux_capset(struct thread *td, struct linux_capset_args *args)
1904 {
1905 	struct l_user_cap_header luch;
1906 	struct l_user_cap_data lucd;
1907 	int error;
1908 
1909 	if (args->hdrp == NULL || args->datap == NULL)
1910 		return (EFAULT);
1911 
1912 	error = copyin(args->hdrp, &luch, sizeof(luch));
1913 	if (error != 0)
1914 		return (error);
1915 
1916 	if (luch.version != _LINUX_CAPABILITY_VERSION) {
1917 		luch.version = _LINUX_CAPABILITY_VERSION;
1918 		error = copyout(&luch, args->hdrp, sizeof(luch));
1919 		if (error)
1920 			return (error);
1921 		return (EINVAL);
1922 	}
1923 
1924 	if (luch.pid)
1925 		return (EPERM);
1926 
1927 	error = copyin(args->datap, &lucd, sizeof(lucd));
1928 	if (error != 0)
1929 		return (error);
1930 
1931 	/* We currently don't support setting any capabilities. */
1932 	if (lucd.effective || lucd.permitted || lucd.inheritable) {
1933 		linux_msg(td,
1934 			  "capset effective=0x%x, permitted=0x%x, "
1935 			  "inheritable=0x%x is not implemented",
1936 			  (int)lucd.effective, (int)lucd.permitted,
1937 			  (int)lucd.inheritable);
1938 		return (EPERM);
1939 	}
1940 
1941 	return (0);
1942 }
1943 
1944 int
1945 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1946 {
1947 	int error = 0, max_size;
1948 	struct proc *p = td->td_proc;
1949 	char comm[LINUX_MAX_COMM_LEN];
1950 	struct linux_emuldata *em;
1951 	int pdeath_signal;
1952 
1953 #ifdef DEBUG
1954 	if (ldebug(prctl))
1955 		printf(ARGS(prctl, "%d, %ju, %ju, %ju, %ju"), args->option,
1956 		    (uintmax_t)args->arg2, (uintmax_t)args->arg3,
1957 		    (uintmax_t)args->arg4, (uintmax_t)args->arg5);
1958 #endif
1959 
1960 	switch (args->option) {
1961 	case LINUX_PR_SET_PDEATHSIG:
1962 		if (!LINUX_SIG_VALID(args->arg2))
1963 			return (EINVAL);
1964 		em = em_find(td);
1965 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1966 		em->pdeath_signal = args->arg2;
1967 		break;
1968 	case LINUX_PR_GET_PDEATHSIG:
1969 		em = em_find(td);
1970 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1971 		pdeath_signal = em->pdeath_signal;
1972 		error = copyout(&pdeath_signal,
1973 		    (void *)(register_t)args->arg2,
1974 		    sizeof(pdeath_signal));
1975 		break;
1976 	case LINUX_PR_GET_KEEPCAPS:
1977 		/*
1978 		 * Indicate that we always clear the effective and
1979 		 * permitted capability sets when the user id becomes
1980 		 * non-zero (actually the capability sets are simply
1981 		 * always zero in the current implementation).
1982 		 */
1983 		td->td_retval[0] = 0;
1984 		break;
1985 	case LINUX_PR_SET_KEEPCAPS:
1986 		/*
1987 		 * Ignore requests to keep the effective and permitted
1988 		 * capability sets when the user id becomes non-zero.
1989 		 */
1990 		break;
1991 	case LINUX_PR_SET_NAME:
1992 		/*
1993 		 * To be on the safe side we need to make sure to not
1994 		 * overflow the size a linux program expects. We already
1995 		 * do this here in the copyin, so that we don't need to
1996 		 * check on copyout.
1997 		 */
1998 		max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1999 		error = copyinstr((void *)(register_t)args->arg2, comm,
2000 		    max_size, NULL);
2001 
2002 		/* Linux silently truncates the name if it is too long. */
2003 		if (error == ENAMETOOLONG) {
2004 			/*
2005 			 * XXX: copyinstr() isn't documented to populate the
2006 			 * array completely, so do a copyin() to be on the
2007 			 * safe side. This should be changed in case
2008 			 * copyinstr() is changed to guarantee this.
2009 			 */
2010 			error = copyin((void *)(register_t)args->arg2, comm,
2011 			    max_size - 1);
2012 			comm[max_size - 1] = '\0';
2013 		}
2014 		if (error)
2015 			return (error);
2016 
2017 		PROC_LOCK(p);
2018 		strlcpy(p->p_comm, comm, sizeof(p->p_comm));
2019 		PROC_UNLOCK(p);
2020 		break;
2021 	case LINUX_PR_GET_NAME:
2022 		PROC_LOCK(p);
2023 		strlcpy(comm, p->p_comm, sizeof(comm));
2024 		PROC_UNLOCK(p);
2025 		error = copyout(comm, (void *)(register_t)args->arg2,
2026 		    strlen(comm) + 1);
2027 		break;
2028 	default:
2029 		error = EINVAL;
2030 		break;
2031 	}
2032 
2033 	return (error);
2034 }
2035 
2036 int
2037 linux_sched_setparam(struct thread *td,
2038     struct linux_sched_setparam_args *uap)
2039 {
2040 	struct sched_param sched_param;
2041 	struct thread *tdt;
2042 	int error;
2043 
2044 #ifdef DEBUG
2045 	if (ldebug(sched_setparam))
2046 		printf(ARGS(sched_setparam, "%d, *"), uap->pid);
2047 #endif
2048 
2049 	error = copyin(uap->param, &sched_param, sizeof(sched_param));
2050 	if (error)
2051 		return (error);
2052 
2053 	tdt = linux_tdfind(td, uap->pid, -1);
2054 	if (tdt == NULL)
2055 		return (ESRCH);
2056 
2057 	error = kern_sched_setparam(td, tdt, &sched_param);
2058 	PROC_UNLOCK(tdt->td_proc);
2059 	return (error);
2060 }
2061 
2062 int
2063 linux_sched_getparam(struct thread *td,
2064     struct linux_sched_getparam_args *uap)
2065 {
2066 	struct sched_param sched_param;
2067 	struct thread *tdt;
2068 	int error;
2069 
2070 #ifdef DEBUG
2071 	if (ldebug(sched_getparam))
2072 		printf(ARGS(sched_getparam, "%d, *"), uap->pid);
2073 #endif
2074 
2075 	tdt = linux_tdfind(td, uap->pid, -1);
2076 	if (tdt == NULL)
2077 		return (ESRCH);
2078 
2079 	error = kern_sched_getparam(td, tdt, &sched_param);
2080 	PROC_UNLOCK(tdt->td_proc);
2081 	if (error == 0)
2082 		error = copyout(&sched_param, uap->param,
2083 		    sizeof(sched_param));
2084 	return (error);
2085 }
2086 
2087 /*
2088  * Get affinity of a process.
2089  */
2090 int
2091 linux_sched_getaffinity(struct thread *td,
2092     struct linux_sched_getaffinity_args *args)
2093 {
2094 	int error;
2095 	struct thread *tdt;
2096 	struct cpuset_getaffinity_args cga;
2097 
2098 #ifdef DEBUG
2099 	if (ldebug(sched_getaffinity))
2100 		printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid,
2101 		    args->len);
2102 #endif
2103 	if (args->len < sizeof(cpuset_t))
2104 		return (EINVAL);
2105 
2106 	tdt = linux_tdfind(td, args->pid, -1);
2107 	if (tdt == NULL)
2108 		return (ESRCH);
2109 
2110 	PROC_UNLOCK(tdt->td_proc);
2111 	cga.level = CPU_LEVEL_WHICH;
2112 	cga.which = CPU_WHICH_TID;
2113 	cga.id = tdt->td_tid;
2114 	cga.cpusetsize = sizeof(cpuset_t);
2115 	cga.mask = (cpuset_t *) args->user_mask_ptr;
2116 
2117 	if ((error = sys_cpuset_getaffinity(td, &cga)) == 0)
2118 		td->td_retval[0] = sizeof(cpuset_t);
2119 
2120 	return (error);
2121 }
2122 
2123 /*
2124  *  Set affinity of a process.
2125  */
2126 int
2127 linux_sched_setaffinity(struct thread *td,
2128     struct linux_sched_setaffinity_args *args)
2129 {
2130 	struct cpuset_setaffinity_args csa;
2131 	struct thread *tdt;
2132 
2133 #ifdef DEBUG
2134 	if (ldebug(sched_setaffinity))
2135 		printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid,
2136 		    args->len);
2137 #endif
2138 	if (args->len < sizeof(cpuset_t))
2139 		return (EINVAL);
2140 
2141 	tdt = linux_tdfind(td, args->pid, -1);
2142 	if (tdt == NULL)
2143 		return (ESRCH);
2144 
2145 	PROC_UNLOCK(tdt->td_proc);
2146 	csa.level = CPU_LEVEL_WHICH;
2147 	csa.which = CPU_WHICH_TID;
2148 	csa.id = tdt->td_tid;
2149 	csa.cpusetsize = sizeof(cpuset_t);
2150 	csa.mask = (cpuset_t *) args->user_mask_ptr;
2151 
2152 	return (sys_cpuset_setaffinity(td, &csa));
2153 }
2154 
2155 struct linux_rlimit64 {
2156 	uint64_t	rlim_cur;
2157 	uint64_t	rlim_max;
2158 };
2159 
2160 int
2161 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
2162 {
2163 	struct rlimit rlim, nrlim;
2164 	struct linux_rlimit64 lrlim;
2165 	struct proc *p;
2166 	u_int which;
2167 	int flags;
2168 	int error;
2169 
2170 #ifdef DEBUG
2171 	if (ldebug(prlimit64))
2172 		printf(ARGS(prlimit64, "%d, %d, %p, %p"), args->pid,
2173 		    args->resource, (void *)args->new, (void *)args->old);
2174 #endif
2175 
2176 	if (args->resource >= LINUX_RLIM_NLIMITS)
2177 		return (EINVAL);
2178 
2179 	which = linux_to_bsd_resource[args->resource];
2180 	if (which == -1)
2181 		return (EINVAL);
2182 
2183 	if (args->new != NULL) {
2184 		/*
2185 		 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
2186 		 * rlim is unsigned 64-bit. FreeBSD treats negative limits
2187 		 * as INFINITY so we do not need a conversion even.
2188 		 */
2189 		error = copyin(args->new, &nrlim, sizeof(nrlim));
2190 		if (error != 0)
2191 			return (error);
2192 	}
2193 
2194 	flags = PGET_HOLD | PGET_NOTWEXIT;
2195 	if (args->new != NULL)
2196 		flags |= PGET_CANDEBUG;
2197 	else
2198 		flags |= PGET_CANSEE;
2199 	error = pget(args->pid, flags, &p);
2200 	if (error != 0)
2201 		return (error);
2202 
2203 	if (args->old != NULL) {
2204 		PROC_LOCK(p);
2205 		lim_rlimit_proc(p, which, &rlim);
2206 		PROC_UNLOCK(p);
2207 		if (rlim.rlim_cur == RLIM_INFINITY)
2208 			lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2209 		else
2210 			lrlim.rlim_cur = rlim.rlim_cur;
2211 		if (rlim.rlim_max == RLIM_INFINITY)
2212 			lrlim.rlim_max = LINUX_RLIM_INFINITY;
2213 		else
2214 			lrlim.rlim_max = rlim.rlim_max;
2215 		error = copyout(&lrlim, args->old, sizeof(lrlim));
2216 		if (error != 0)
2217 			goto out;
2218 	}
2219 
2220 	if (args->new != NULL)
2221 		error = kern_proc_setrlimit(td, p, which, &nrlim);
2222 
2223  out:
2224 	PRELE(p);
2225 	return (error);
2226 }
2227 
2228 int
2229 linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2230 {
2231 	struct timeval utv, tv0, tv1, *tvp;
2232 	struct l_pselect6arg lpse6;
2233 	struct l_timespec lts;
2234 	struct timespec uts;
2235 	l_sigset_t l_ss;
2236 	sigset_t *ssp;
2237 	sigset_t ss;
2238 	int error;
2239 
2240 	ssp = NULL;
2241 	if (args->sig != NULL) {
2242 		error = copyin(args->sig, &lpse6, sizeof(lpse6));
2243 		if (error != 0)
2244 			return (error);
2245 		if (lpse6.ss_len != sizeof(l_ss))
2246 			return (EINVAL);
2247 		if (lpse6.ss != 0) {
2248 			error = copyin(PTRIN(lpse6.ss), &l_ss,
2249 			    sizeof(l_ss));
2250 			if (error != 0)
2251 				return (error);
2252 			linux_to_bsd_sigset(&l_ss, &ss);
2253 			ssp = &ss;
2254 		}
2255 	}
2256 
2257 	/*
2258 	 * Currently glibc changes nanosecond number to microsecond.
2259 	 * This mean losing precision but for now it is hardly seen.
2260 	 */
2261 	if (args->tsp != NULL) {
2262 		error = copyin(args->tsp, &lts, sizeof(lts));
2263 		if (error != 0)
2264 			return (error);
2265 		error = linux_to_native_timespec(&uts, &lts);
2266 		if (error != 0)
2267 			return (error);
2268 
2269 		TIMESPEC_TO_TIMEVAL(&utv, &uts);
2270 		if (itimerfix(&utv))
2271 			return (EINVAL);
2272 
2273 		microtime(&tv0);
2274 		tvp = &utv;
2275 	} else
2276 		tvp = NULL;
2277 
2278 	error = kern_pselect(td, args->nfds, args->readfds, args->writefds,
2279 	    args->exceptfds, tvp, ssp, LINUX_NFDBITS);
2280 
2281 	if (error == 0 && args->tsp != NULL) {
2282 		if (td->td_retval[0] != 0) {
2283 			/*
2284 			 * Compute how much time was left of the timeout,
2285 			 * by subtracting the current time and the time
2286 			 * before we started the call, and subtracting
2287 			 * that result from the user-supplied value.
2288 			 */
2289 
2290 			microtime(&tv1);
2291 			timevalsub(&tv1, &tv0);
2292 			timevalsub(&utv, &tv1);
2293 			if (utv.tv_sec < 0)
2294 				timevalclear(&utv);
2295 		} else
2296 			timevalclear(&utv);
2297 
2298 		TIMEVAL_TO_TIMESPEC(&utv, &uts);
2299 
2300 		native_to_linux_timespec(&lts, &uts);
2301 		error = copyout(&lts, args->tsp, sizeof(lts));
2302 	}
2303 
2304 	return (error);
2305 }
2306 
2307 int
2308 linux_ppoll(struct thread *td, struct linux_ppoll_args *args)
2309 {
2310 	struct timespec ts0, ts1;
2311 	struct l_timespec lts;
2312 	struct timespec uts, *tsp;
2313 	l_sigset_t l_ss;
2314 	sigset_t *ssp;
2315 	sigset_t ss;
2316 	int error;
2317 
2318 	if (args->sset != NULL) {
2319 		if (args->ssize != sizeof(l_ss))
2320 			return (EINVAL);
2321 		error = copyin(args->sset, &l_ss, sizeof(l_ss));
2322 		if (error)
2323 			return (error);
2324 		linux_to_bsd_sigset(&l_ss, &ss);
2325 		ssp = &ss;
2326 	} else
2327 		ssp = NULL;
2328 	if (args->tsp != NULL) {
2329 		error = copyin(args->tsp, &lts, sizeof(lts));
2330 		if (error)
2331 			return (error);
2332 		error = linux_to_native_timespec(&uts, &lts);
2333 		if (error != 0)
2334 			return (error);
2335 
2336 		nanotime(&ts0);
2337 		tsp = &uts;
2338 	} else
2339 		tsp = NULL;
2340 
2341 	error = kern_poll(td, args->fds, args->nfds, tsp, ssp);
2342 
2343 	if (error == 0 && args->tsp != NULL) {
2344 		if (td->td_retval[0]) {
2345 			nanotime(&ts1);
2346 			timespecsub(&ts1, &ts0);
2347 			timespecsub(&uts, &ts1);
2348 			if (uts.tv_sec < 0)
2349 				timespecclear(&uts);
2350 		} else
2351 			timespecclear(&uts);
2352 
2353 		native_to_linux_timespec(&lts, &uts);
2354 		error = copyout(&lts, args->tsp, sizeof(lts));
2355 	}
2356 
2357 	return (error);
2358 }
2359 
2360 #if defined(DEBUG) || defined(KTR)
2361 /* XXX: can be removed when every ldebug(...) and KTR stuff are removed. */
2362 
2363 #ifdef COMPAT_LINUX32
2364 #define	L_MAXSYSCALL	LINUX32_SYS_MAXSYSCALL
2365 #else
2366 #define	L_MAXSYSCALL	LINUX_SYS_MAXSYSCALL
2367 #endif
2368 
2369 u_char linux_debug_map[howmany(L_MAXSYSCALL, sizeof(u_char))];
2370 
2371 static int
2372 linux_debug(int syscall, int toggle, int global)
2373 {
2374 
2375 	if (global) {
2376 		char c = toggle ? 0 : 0xff;
2377 
2378 		memset(linux_debug_map, c, sizeof(linux_debug_map));
2379 		return (0);
2380 	}
2381 	if (syscall < 0 || syscall >= L_MAXSYSCALL)
2382 		return (EINVAL);
2383 	if (toggle)
2384 		clrbit(linux_debug_map, syscall);
2385 	else
2386 		setbit(linux_debug_map, syscall);
2387 	return (0);
2388 }
2389 #undef L_MAXSYSCALL
2390 
2391 /*
2392  * Usage: sysctl linux.debug=<syscall_nr>.<0/1>
2393  *
2394  *    E.g.: sysctl linux.debug=21.0
2395  *
2396  * As a special case, syscall "all" will apply to all syscalls globally.
2397  */
2398 #define LINUX_MAX_DEBUGSTR	16
2399 int
2400 linux_sysctl_debug(SYSCTL_HANDLER_ARGS)
2401 {
2402 	char value[LINUX_MAX_DEBUGSTR], *p;
2403 	int error, sysc, toggle;
2404 	int global = 0;
2405 
2406 	value[0] = '\0';
2407 	error = sysctl_handle_string(oidp, value, LINUX_MAX_DEBUGSTR, req);
2408 	if (error || req->newptr == NULL)
2409 		return (error);
2410 	for (p = value; *p != '\0' && *p != '.'; p++);
2411 	if (*p == '\0')
2412 		return (EINVAL);
2413 	*p++ = '\0';
2414 	sysc = strtol(value, NULL, 0);
2415 	toggle = strtol(p, NULL, 0);
2416 	if (strcmp(value, "all") == 0)
2417 		global = 1;
2418 	error = linux_debug(sysc, toggle, global);
2419 	return (error);
2420 }
2421 
2422 #endif /* DEBUG || KTR */
2423 
2424 int
2425 linux_sched_rr_get_interval(struct thread *td,
2426     struct linux_sched_rr_get_interval_args *uap)
2427 {
2428 	struct timespec ts;
2429 	struct l_timespec lts;
2430 	struct thread *tdt;
2431 	int error;
2432 
2433 	/*
2434 	 * According to man in case the invalid pid specified
2435 	 * EINVAL should be returned.
2436 	 */
2437 	if (uap->pid < 0)
2438 		return (EINVAL);
2439 
2440 	tdt = linux_tdfind(td, uap->pid, -1);
2441 	if (tdt == NULL)
2442 		return (ESRCH);
2443 
2444 	error = kern_sched_rr_get_interval_td(td, tdt, &ts);
2445 	PROC_UNLOCK(tdt->td_proc);
2446 	if (error != 0)
2447 		return (error);
2448 	native_to_linux_timespec(&lts, &ts);
2449 	return (copyout(&lts, uap->interval, sizeof(lts)));
2450 }
2451 
2452 /*
2453  * In case when the Linux thread is the initial thread in
2454  * the thread group thread id is equal to the process id.
2455  * Glibc depends on this magic (assert in pthread_getattr_np.c).
2456  */
2457 struct thread *
2458 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid)
2459 {
2460 	struct linux_emuldata *em;
2461 	struct thread *tdt;
2462 	struct proc *p;
2463 
2464 	tdt = NULL;
2465 	if (tid == 0 || tid == td->td_tid) {
2466 		tdt = td;
2467 		PROC_LOCK(tdt->td_proc);
2468 	} else if (tid > PID_MAX)
2469 		tdt = tdfind(tid, pid);
2470 	else {
2471 		/*
2472 		 * Initial thread where the tid equal to the pid.
2473 		 */
2474 		p = pfind(tid);
2475 		if (p != NULL) {
2476 			if (SV_PROC_ABI(p) != SV_ABI_LINUX) {
2477 				/*
2478 				 * p is not a Linuxulator process.
2479 				 */
2480 				PROC_UNLOCK(p);
2481 				return (NULL);
2482 			}
2483 			FOREACH_THREAD_IN_PROC(p, tdt) {
2484 				em = em_find(tdt);
2485 				if (tid == em->em_tid)
2486 					return (tdt);
2487 			}
2488 			PROC_UNLOCK(p);
2489 		}
2490 		return (NULL);
2491 	}
2492 
2493 	return (tdt);
2494 }
2495 
2496 void
2497 linux_to_bsd_waitopts(int options, int *bsdopts)
2498 {
2499 
2500 	if (options & LINUX_WNOHANG)
2501 		*bsdopts |= WNOHANG;
2502 	if (options & LINUX_WUNTRACED)
2503 		*bsdopts |= WUNTRACED;
2504 	if (options & LINUX_WEXITED)
2505 		*bsdopts |= WEXITED;
2506 	if (options & LINUX_WCONTINUED)
2507 		*bsdopts |= WCONTINUED;
2508 	if (options & LINUX_WNOWAIT)
2509 		*bsdopts |= WNOWAIT;
2510 
2511 	if (options & __WCLONE)
2512 		*bsdopts |= WLINUXCLONE;
2513 }
2514