xref: /freebsd-14.2/sys/compat/linux/linux_misc.c (revision bc273677)
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_util.h>
92 #include <compat/linux/linux_sysproto.h>
93 #include <compat/linux/linux_emul.h>
94 #include <compat/linux/linux_misc.h>
95 
96 /**
97  * Special DTrace provider for the linuxulator.
98  *
99  * In this file we define the provider for the entire linuxulator. All
100  * modules (= files of the linuxulator) use it.
101  *
102  * We define a different name depending on the emulated bitsize, see
103  * ../../<ARCH>/linux{,32}/linux.h, e.g.:
104  *      native bitsize          = linuxulator
105  *      amd64, 32bit emulation  = linuxulator32
106  */
107 LIN_SDT_PROVIDER_DEFINE(LINUX_DTRACE);
108 
109 int stclohz;				/* Statistics clock frequency */
110 
111 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
112 	RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
113 	RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
114 	RLIMIT_MEMLOCK, RLIMIT_AS
115 };
116 
117 struct l_sysinfo {
118 	l_long		uptime;		/* Seconds since boot */
119 	l_ulong		loads[3];	/* 1, 5, and 15 minute load averages */
120 #define LINUX_SYSINFO_LOADS_SCALE 65536
121 	l_ulong		totalram;	/* Total usable main memory size */
122 	l_ulong		freeram;	/* Available memory size */
123 	l_ulong		sharedram;	/* Amount of shared memory */
124 	l_ulong		bufferram;	/* Memory used by buffers */
125 	l_ulong		totalswap;	/* Total swap space size */
126 	l_ulong		freeswap;	/* swap space still available */
127 	l_ushort	procs;		/* Number of current processes */
128 	l_ushort	pads;
129 	l_ulong		totalbig;
130 	l_ulong		freebig;
131 	l_uint		mem_unit;
132 	char		_f[20-2*sizeof(l_long)-sizeof(l_int)];	/* padding */
133 };
134 
135 struct l_pselect6arg {
136 	l_uintptr_t	ss;
137 	l_size_t	ss_len;
138 };
139 
140 int
141 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
142 {
143 	struct l_sysinfo sysinfo;
144 	vm_object_t object;
145 	int i, j;
146 	struct timespec ts;
147 
148 	getnanouptime(&ts);
149 	if (ts.tv_nsec != 0)
150 		ts.tv_sec++;
151 	sysinfo.uptime = ts.tv_sec;
152 
153 	/* Use the information from the mib to get our load averages */
154 	for (i = 0; i < 3; i++)
155 		sysinfo.loads[i] = averunnable.ldavg[i] *
156 		    LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
157 
158 	sysinfo.totalram = physmem * PAGE_SIZE;
159 	sysinfo.freeram = sysinfo.totalram - vm_cnt.v_wire_count * PAGE_SIZE;
160 
161 	sysinfo.sharedram = 0;
162 	mtx_lock(&vm_object_list_mtx);
163 	TAILQ_FOREACH(object, &vm_object_list, object_list)
164 		if (object->shadow_count > 1)
165 			sysinfo.sharedram += object->resident_page_count;
166 	mtx_unlock(&vm_object_list_mtx);
167 
168 	sysinfo.sharedram *= PAGE_SIZE;
169 	sysinfo.bufferram = 0;
170 
171 	swap_pager_status(&i, &j);
172 	sysinfo.totalswap = i * PAGE_SIZE;
173 	sysinfo.freeswap = (i - j) * PAGE_SIZE;
174 
175 	sysinfo.procs = nprocs;
176 
177 	/* The following are only present in newer Linux kernels. */
178 	sysinfo.totalbig = 0;
179 	sysinfo.freebig = 0;
180 	sysinfo.mem_unit = 1;
181 
182 	return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
183 }
184 
185 int
186 linux_alarm(struct thread *td, struct linux_alarm_args *args)
187 {
188 	struct itimerval it, old_it;
189 	u_int secs;
190 	int error;
191 
192 #ifdef DEBUG
193 	if (ldebug(alarm))
194 		printf(ARGS(alarm, "%u"), args->secs);
195 #endif
196 
197 	secs = args->secs;
198 
199 	if (secs > INT_MAX)
200 		secs = INT_MAX;
201 
202 	it.it_value.tv_sec = (long) secs;
203 	it.it_value.tv_usec = 0;
204 	it.it_interval.tv_sec = 0;
205 	it.it_interval.tv_usec = 0;
206 	error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
207 	if (error)
208 		return (error);
209 	if (timevalisset(&old_it.it_value)) {
210 		if (old_it.it_value.tv_usec != 0)
211 			old_it.it_value.tv_sec++;
212 		td->td_retval[0] = old_it.it_value.tv_sec;
213 	}
214 	return (0);
215 }
216 
217 int
218 linux_brk(struct thread *td, struct linux_brk_args *args)
219 {
220 	struct vmspace *vm = td->td_proc->p_vmspace;
221 	vm_offset_t new, old;
222 	struct obreak_args /* {
223 		char * nsize;
224 	} */ tmp;
225 
226 #ifdef DEBUG
227 	if (ldebug(brk))
228 		printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
229 #endif
230 	old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize);
231 	new = (vm_offset_t)args->dsend;
232 	tmp.nsize = (char *)new;
233 	if (((caddr_t)new > vm->vm_daddr) && !sys_obreak(td, &tmp))
234 		td->td_retval[0] = (long)new;
235 	else
236 		td->td_retval[0] = (long)old;
237 
238 	return (0);
239 }
240 
241 #if defined(__i386__)
242 /* XXX: what about amd64/linux32? */
243 
244 int
245 linux_uselib(struct thread *td, struct linux_uselib_args *args)
246 {
247 	struct nameidata ni;
248 	struct vnode *vp;
249 	struct exec *a_out;
250 	struct vattr attr;
251 	vm_offset_t vmaddr;
252 	unsigned long file_offset;
253 	unsigned long bss_size;
254 	char *library;
255 	ssize_t aresid;
256 	int error, locked, writecount;
257 
258 	LCONVPATHEXIST(td, args->library, &library);
259 
260 #ifdef DEBUG
261 	if (ldebug(uselib))
262 		printf(ARGS(uselib, "%s"), library);
263 #endif
264 
265 	a_out = NULL;
266 	locked = 0;
267 	vp = NULL;
268 
269 	NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
270 	    UIO_SYSSPACE, library, td);
271 	error = namei(&ni);
272 	LFREEPATH(library);
273 	if (error)
274 		goto cleanup;
275 
276 	vp = ni.ni_vp;
277 	NDFREE(&ni, NDF_ONLY_PNBUF);
278 
279 	/*
280 	 * From here on down, we have a locked vnode that must be unlocked.
281 	 * XXX: The code below largely duplicates exec_check_permissions().
282 	 */
283 	locked = 1;
284 
285 	/* Writable? */
286 	error = VOP_GET_WRITECOUNT(vp, &writecount);
287 	if (error != 0)
288 		goto cleanup;
289 	if (writecount != 0) {
290 		error = ETXTBSY;
291 		goto cleanup;
292 	}
293 
294 	/* Executable? */
295 	error = VOP_GETATTR(vp, &attr, td->td_ucred);
296 	if (error)
297 		goto cleanup;
298 
299 	if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
300 	    ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
301 		/* EACCESS is what exec(2) returns. */
302 		error = ENOEXEC;
303 		goto cleanup;
304 	}
305 
306 	/* Sensible size? */
307 	if (attr.va_size == 0) {
308 		error = ENOEXEC;
309 		goto cleanup;
310 	}
311 
312 	/* Can we access it? */
313 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
314 	if (error)
315 		goto cleanup;
316 
317 	/*
318 	 * XXX: This should use vn_open() so that it is properly authorized,
319 	 * and to reduce code redundancy all over the place here.
320 	 * XXX: Not really, it duplicates far more of exec_check_permissions()
321 	 * than vn_open().
322 	 */
323 #ifdef MAC
324 	error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
325 	if (error)
326 		goto cleanup;
327 #endif
328 	error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
329 	if (error)
330 		goto cleanup;
331 
332 	/* Pull in executable header into exec_map */
333 	error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
334 	    VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
335 	if (error)
336 		goto cleanup;
337 
338 	/* Is it a Linux binary ? */
339 	if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
340 		error = ENOEXEC;
341 		goto cleanup;
342 	}
343 
344 	/*
345 	 * While we are here, we should REALLY do some more checks
346 	 */
347 
348 	/* Set file/virtual offset based on a.out variant. */
349 	switch ((int)(a_out->a_magic & 0xffff)) {
350 	case 0413:			/* ZMAGIC */
351 		file_offset = 1024;
352 		break;
353 	case 0314:			/* QMAGIC */
354 		file_offset = 0;
355 		break;
356 	default:
357 		error = ENOEXEC;
358 		goto cleanup;
359 	}
360 
361 	bss_size = round_page(a_out->a_bss);
362 
363 	/* Check various fields in header for validity/bounds. */
364 	if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
365 		error = ENOEXEC;
366 		goto cleanup;
367 	}
368 
369 	/* text + data can't exceed file size */
370 	if (a_out->a_data + a_out->a_text > attr.va_size) {
371 		error = EFAULT;
372 		goto cleanup;
373 	}
374 
375 	/*
376 	 * text/data/bss must not exceed limits
377 	 * XXX - this is not complete. it should check current usage PLUS
378 	 * the resources needed by this library.
379 	 */
380 	PROC_LOCK(td->td_proc);
381 	if (a_out->a_text > maxtsiz ||
382 	    a_out->a_data + bss_size > lim_cur(td->td_proc, RLIMIT_DATA) ||
383 	    racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
384 	    bss_size) != 0) {
385 		PROC_UNLOCK(td->td_proc);
386 		error = ENOMEM;
387 		goto cleanup;
388 	}
389 	PROC_UNLOCK(td->td_proc);
390 
391 	/*
392 	 * Prevent more writers.
393 	 * XXX: Note that if any of the VM operations fail below we don't
394 	 * clear this flag.
395 	 */
396 	VOP_SET_TEXT(vp);
397 
398 	/*
399 	 * Lock no longer needed
400 	 */
401 	locked = 0;
402 	VOP_UNLOCK(vp, 0);
403 
404 	/*
405 	 * Check if file_offset page aligned. Currently we cannot handle
406 	 * misalinged file offsets, and so we read in the entire image
407 	 * (what a waste).
408 	 */
409 	if (file_offset & PAGE_MASK) {
410 #ifdef DEBUG
411 		printf("uselib: Non page aligned binary %lu\n", file_offset);
412 #endif
413 		/* Map text+data read/write/execute */
414 
415 		/* a_entry is the load address and is page aligned */
416 		vmaddr = trunc_page(a_out->a_entry);
417 
418 		/* get anon user mapping, read+write+execute */
419 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
420 		    &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
421 		    VM_PROT_ALL, VM_PROT_ALL, 0);
422 		if (error)
423 			goto cleanup;
424 
425 		error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
426 		    a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
427 		    td->td_ucred, NOCRED, &aresid, td);
428 		if (error != 0)
429 			goto cleanup;
430 		if (aresid != 0) {
431 			error = ENOEXEC;
432 			goto cleanup;
433 		}
434 	} else {
435 #ifdef DEBUG
436 		printf("uselib: Page aligned binary %lu\n", file_offset);
437 #endif
438 		/*
439 		 * for QMAGIC, a_entry is 20 bytes beyond the load address
440 		 * to skip the executable header
441 		 */
442 		vmaddr = trunc_page(a_out->a_entry);
443 
444 		/*
445 		 * Map it all into the process's space as a single
446 		 * copy-on-write "data" segment.
447 		 */
448 		error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr,
449 		    a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
450 		    MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
451 		if (error)
452 			goto cleanup;
453 	}
454 #ifdef DEBUG
455 	printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
456 	    ((long *)vmaddr)[1]);
457 #endif
458 	if (bss_size != 0) {
459 		/* Calculate BSS start address */
460 		vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
461 		    a_out->a_data;
462 
463 		/* allocate some 'anon' space */
464 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
465 		    &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
466 		    VM_PROT_ALL, 0);
467 		if (error)
468 			goto cleanup;
469 	}
470 
471 cleanup:
472 	/* Unlock vnode if needed */
473 	if (locked)
474 		VOP_UNLOCK(vp, 0);
475 
476 	/* Release the temporary mapping. */
477 	if (a_out)
478 		kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
479 
480 	return (error);
481 }
482 
483 #endif	/* __i386__ */
484 
485 int
486 linux_select(struct thread *td, struct linux_select_args *args)
487 {
488 	l_timeval ltv;
489 	struct timeval tv0, tv1, utv, *tvp;
490 	int error;
491 
492 #ifdef DEBUG
493 	if (ldebug(select))
494 		printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
495 		    (void *)args->readfds, (void *)args->writefds,
496 		    (void *)args->exceptfds, (void *)args->timeout);
497 #endif
498 
499 	/*
500 	 * Store current time for computation of the amount of
501 	 * time left.
502 	 */
503 	if (args->timeout) {
504 		if ((error = copyin(args->timeout, &ltv, sizeof(ltv))))
505 			goto select_out;
506 		utv.tv_sec = ltv.tv_sec;
507 		utv.tv_usec = ltv.tv_usec;
508 #ifdef DEBUG
509 		if (ldebug(select))
510 			printf(LMSG("incoming timeout (%jd/%ld)"),
511 			    (intmax_t)utv.tv_sec, utv.tv_usec);
512 #endif
513 
514 		if (itimerfix(&utv)) {
515 			/*
516 			 * The timeval was invalid.  Convert it to something
517 			 * valid that will act as it does under Linux.
518 			 */
519 			utv.tv_sec += utv.tv_usec / 1000000;
520 			utv.tv_usec %= 1000000;
521 			if (utv.tv_usec < 0) {
522 				utv.tv_sec -= 1;
523 				utv.tv_usec += 1000000;
524 			}
525 			if (utv.tv_sec < 0)
526 				timevalclear(&utv);
527 		}
528 		microtime(&tv0);
529 		tvp = &utv;
530 	} else
531 		tvp = NULL;
532 
533 	error = kern_select(td, args->nfds, args->readfds, args->writefds,
534 	    args->exceptfds, tvp, sizeof(l_int) * 8);
535 
536 #ifdef DEBUG
537 	if (ldebug(select))
538 		printf(LMSG("real select returns %d"), error);
539 #endif
540 	if (error)
541 		goto select_out;
542 
543 	if (args->timeout) {
544 		if (td->td_retval[0]) {
545 			/*
546 			 * Compute how much time was left of the timeout,
547 			 * by subtracting the current time and the time
548 			 * before we started the call, and subtracting
549 			 * that result from the user-supplied value.
550 			 */
551 			microtime(&tv1);
552 			timevalsub(&tv1, &tv0);
553 			timevalsub(&utv, &tv1);
554 			if (utv.tv_sec < 0)
555 				timevalclear(&utv);
556 		} else
557 			timevalclear(&utv);
558 #ifdef DEBUG
559 		if (ldebug(select))
560 			printf(LMSG("outgoing timeout (%jd/%ld)"),
561 			    (intmax_t)utv.tv_sec, utv.tv_usec);
562 #endif
563 		ltv.tv_sec = utv.tv_sec;
564 		ltv.tv_usec = utv.tv_usec;
565 		if ((error = copyout(&ltv, args->timeout, sizeof(ltv))))
566 			goto select_out;
567 	}
568 
569 select_out:
570 #ifdef DEBUG
571 	if (ldebug(select))
572 		printf(LMSG("select_out -> %d"), error);
573 #endif
574 	return (error);
575 }
576 
577 int
578 linux_mremap(struct thread *td, struct linux_mremap_args *args)
579 {
580 	struct munmap_args /* {
581 		void *addr;
582 		size_t len;
583 	} */ bsd_args;
584 	int error = 0;
585 
586 #ifdef DEBUG
587 	if (ldebug(mremap))
588 		printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
589 		    (void *)(uintptr_t)args->addr,
590 		    (unsigned long)args->old_len,
591 		    (unsigned long)args->new_len,
592 		    (unsigned long)args->flags);
593 #endif
594 
595 	if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
596 		td->td_retval[0] = 0;
597 		return (EINVAL);
598 	}
599 
600 	/*
601 	 * Check for the page alignment.
602 	 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
603 	 */
604 	if (args->addr & PAGE_MASK) {
605 		td->td_retval[0] = 0;
606 		return (EINVAL);
607 	}
608 
609 	args->new_len = round_page(args->new_len);
610 	args->old_len = round_page(args->old_len);
611 
612 	if (args->new_len > args->old_len) {
613 		td->td_retval[0] = 0;
614 		return (ENOMEM);
615 	}
616 
617 	if (args->new_len < args->old_len) {
618 		bsd_args.addr =
619 		    (caddr_t)((uintptr_t)args->addr + args->new_len);
620 		bsd_args.len = args->old_len - args->new_len;
621 		error = sys_munmap(td, &bsd_args);
622 	}
623 
624 	td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
625 	return (error);
626 }
627 
628 #define LINUX_MS_ASYNC       0x0001
629 #define LINUX_MS_INVALIDATE  0x0002
630 #define LINUX_MS_SYNC        0x0004
631 
632 int
633 linux_msync(struct thread *td, struct linux_msync_args *args)
634 {
635 	struct msync_args bsd_args;
636 
637 	bsd_args.addr = (caddr_t)(uintptr_t)args->addr;
638 	bsd_args.len = (uintptr_t)args->len;
639 	bsd_args.flags = args->fl & ~LINUX_MS_SYNC;
640 
641 	return (sys_msync(td, &bsd_args));
642 }
643 
644 int
645 linux_time(struct thread *td, struct linux_time_args *args)
646 {
647 	struct timeval tv;
648 	l_time_t tm;
649 	int error;
650 
651 #ifdef DEBUG
652 	if (ldebug(time))
653 		printf(ARGS(time, "*"));
654 #endif
655 
656 	microtime(&tv);
657 	tm = tv.tv_sec;
658 	if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
659 		return (error);
660 	td->td_retval[0] = tm;
661 	return (0);
662 }
663 
664 struct l_times_argv {
665 	l_clock_t	tms_utime;
666 	l_clock_t	tms_stime;
667 	l_clock_t	tms_cutime;
668 	l_clock_t	tms_cstime;
669 };
670 
671 
672 /*
673  * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
674  * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
675  * auxiliary vector entry.
676  */
677 #define	CLK_TCK		100
678 
679 #define	CONVOTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
680 #define	CONVNTCK(r)	(r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
681 
682 #define	CONVTCK(r)	(linux_kernver(td) >= LINUX_KERNVER_2004000 ?		\
683 			    CONVNTCK(r) : CONVOTCK(r))
684 
685 int
686 linux_times(struct thread *td, struct linux_times_args *args)
687 {
688 	struct timeval tv, utime, stime, cutime, cstime;
689 	struct l_times_argv tms;
690 	struct proc *p;
691 	int error;
692 
693 #ifdef DEBUG
694 	if (ldebug(times))
695 		printf(ARGS(times, "*"));
696 #endif
697 
698 	if (args->buf != NULL) {
699 		p = td->td_proc;
700 		PROC_LOCK(p);
701 		PROC_STATLOCK(p);
702 		calcru(p, &utime, &stime);
703 		PROC_STATUNLOCK(p);
704 		calccru(p, &cutime, &cstime);
705 		PROC_UNLOCK(p);
706 
707 		tms.tms_utime = CONVTCK(utime);
708 		tms.tms_stime = CONVTCK(stime);
709 
710 		tms.tms_cutime = CONVTCK(cutime);
711 		tms.tms_cstime = CONVTCK(cstime);
712 
713 		if ((error = copyout(&tms, args->buf, sizeof(tms))))
714 			return (error);
715 	}
716 
717 	microuptime(&tv);
718 	td->td_retval[0] = (int)CONVTCK(tv);
719 	return (0);
720 }
721 
722 int
723 linux_newuname(struct thread *td, struct linux_newuname_args *args)
724 {
725 	struct l_new_utsname utsname;
726 	char osname[LINUX_MAX_UTSNAME];
727 	char osrelease[LINUX_MAX_UTSNAME];
728 	char *p;
729 
730 #ifdef DEBUG
731 	if (ldebug(newuname))
732 		printf(ARGS(newuname, "*"));
733 #endif
734 
735 	linux_get_osname(td, osname);
736 	linux_get_osrelease(td, osrelease);
737 
738 	bzero(&utsname, sizeof(utsname));
739 	strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
740 	getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
741 	getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
742 	strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
743 	strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
744 	for (p = utsname.version; *p != '\0'; ++p)
745 		if (*p == '\n') {
746 			*p = '\0';
747 			break;
748 		}
749 	strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME);
750 
751 	return (copyout(&utsname, args->buf, sizeof(utsname)));
752 }
753 
754 struct l_utimbuf {
755 	l_time_t l_actime;
756 	l_time_t l_modtime;
757 };
758 
759 int
760 linux_utime(struct thread *td, struct linux_utime_args *args)
761 {
762 	struct timeval tv[2], *tvp;
763 	struct l_utimbuf lut;
764 	char *fname;
765 	int error;
766 
767 	LCONVPATHEXIST(td, args->fname, &fname);
768 
769 #ifdef DEBUG
770 	if (ldebug(utime))
771 		printf(ARGS(utime, "%s, *"), fname);
772 #endif
773 
774 	if (args->times) {
775 		if ((error = copyin(args->times, &lut, sizeof lut))) {
776 			LFREEPATH(fname);
777 			return (error);
778 		}
779 		tv[0].tv_sec = lut.l_actime;
780 		tv[0].tv_usec = 0;
781 		tv[1].tv_sec = lut.l_modtime;
782 		tv[1].tv_usec = 0;
783 		tvp = tv;
784 	} else
785 		tvp = NULL;
786 
787 	error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, tvp,
788 	    UIO_SYSSPACE);
789 	LFREEPATH(fname);
790 	return (error);
791 }
792 
793 int
794 linux_utimes(struct thread *td, struct linux_utimes_args *args)
795 {
796 	l_timeval ltv[2];
797 	struct timeval tv[2], *tvp = NULL;
798 	char *fname;
799 	int error;
800 
801 	LCONVPATHEXIST(td, args->fname, &fname);
802 
803 #ifdef DEBUG
804 	if (ldebug(utimes))
805 		printf(ARGS(utimes, "%s, *"), fname);
806 #endif
807 
808 	if (args->tptr != NULL) {
809 		if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
810 			LFREEPATH(fname);
811 			return (error);
812 		}
813 		tv[0].tv_sec = ltv[0].tv_sec;
814 		tv[0].tv_usec = ltv[0].tv_usec;
815 		tv[1].tv_sec = ltv[1].tv_sec;
816 		tv[1].tv_usec = ltv[1].tv_usec;
817 		tvp = tv;
818 	}
819 
820 	error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE,
821 	    tvp, UIO_SYSSPACE);
822 	LFREEPATH(fname);
823 	return (error);
824 }
825 
826 int
827 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
828 {
829 	l_timeval ltv[2];
830 	struct timeval tv[2], *tvp = NULL;
831 	char *fname;
832 	int error, dfd;
833 
834 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
835 	LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
836 
837 #ifdef DEBUG
838 	if (ldebug(futimesat))
839 		printf(ARGS(futimesat, "%s, *"), fname);
840 #endif
841 
842 	if (args->utimes != NULL) {
843 		if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
844 			LFREEPATH(fname);
845 			return (error);
846 		}
847 		tv[0].tv_sec = ltv[0].tv_sec;
848 		tv[0].tv_usec = ltv[0].tv_usec;
849 		tv[1].tv_sec = ltv[1].tv_sec;
850 		tv[1].tv_usec = ltv[1].tv_usec;
851 		tvp = tv;
852 	}
853 
854 	error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
855 	LFREEPATH(fname);
856 	return (error);
857 }
858 
859 int
860 linux_common_wait(struct thread *td, int pid, int *status,
861     int options, struct rusage *ru)
862 {
863 	int error, tmpstat;
864 
865 	error = kern_wait(td, pid, &tmpstat, options, ru);
866 	if (error)
867 		return (error);
868 
869 	if (status) {
870 		tmpstat &= 0xffff;
871 		if (WIFSIGNALED(tmpstat))
872 			tmpstat = (tmpstat & 0xffffff80) |
873 			    BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat));
874 		else if (WIFSTOPPED(tmpstat))
875 			tmpstat = (tmpstat & 0xffff00ff) |
876 			    (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8);
877 		error = copyout(&tmpstat, status, sizeof(int));
878 	}
879 
880 	return (error);
881 }
882 
883 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
884 int
885 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
886 {
887 	int options;
888 
889 #ifdef DEBUG
890 	if (ldebug(waitpid))
891 		printf(ARGS(waitpid, "%d, %p, %d"),
892 		    args->pid, (void *)args->status, args->options);
893 #endif
894 	/*
895 	 * this is necessary because the test in kern_wait doesn't work
896 	 * because we mess with the options here
897 	 */
898 	if (args->options & ~(WUNTRACED | WNOHANG | WCONTINUED | __WCLONE))
899 		return (EINVAL);
900 
901 	options = (args->options & (WNOHANG | WUNTRACED));
902 	/* WLINUXCLONE should be equal to __WCLONE, but we make sure */
903 	if (args->options & __WCLONE)
904 		options |= WLINUXCLONE;
905 
906 	return (linux_common_wait(td, args->pid, args->status, options, NULL));
907 }
908 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
909 
910 int
911 linux_wait4(struct thread *td, struct linux_wait4_args *args)
912 {
913 	int error, options;
914 	struct rusage ru, *rup;
915 
916 #ifdef DEBUG
917 	if (ldebug(wait4))
918 		printf(ARGS(wait4, "%d, %p, %d, %p"),
919 		    args->pid, (void *)args->status, args->options,
920 		    (void *)args->rusage);
921 #endif
922 
923 	options = (args->options & (WNOHANG | WUNTRACED));
924 	/* WLINUXCLONE should be equal to __WCLONE, but we make sure */
925 	if (args->options & __WCLONE)
926 		options |= WLINUXCLONE;
927 
928 	if (args->rusage != NULL)
929 		rup = &ru;
930 	else
931 		rup = NULL;
932 	error = linux_common_wait(td, args->pid, args->status, options, rup);
933 	if (error != 0)
934 		return (error);
935 	if (args->rusage != NULL)
936 		error = linux_copyout_rusage(&ru, args->rusage);
937 	return (error);
938 }
939 
940 int
941 linux_waitid(struct thread *td, struct linux_waitid_args *args)
942 {
943 	int status, options, sig;
944 	struct __wrusage wru;
945 	siginfo_t siginfo;
946 	l_siginfo_t lsi;
947 	idtype_t idtype;
948 	struct proc *p;
949 	int error;
950 
951 	options = 0;
952 	linux_to_bsd_waitopts(args->options, &options);
953 
954 	if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED))
955 		return (EINVAL);
956 	if (!(options & (WEXITED | WUNTRACED | WCONTINUED)))
957 		return (EINVAL);
958 
959 	switch (args->idtype) {
960 	case LINUX_P_ALL:
961 		idtype = P_ALL;
962 		break;
963 	case LINUX_P_PID:
964 		if (args->id <= 0)
965 			return (EINVAL);
966 		idtype = P_PID;
967 		break;
968 	case LINUX_P_PGID:
969 		if (args->id <= 0)
970 			return (EINVAL);
971 		idtype = P_PGID;
972 		break;
973 	default:
974 		return (EINVAL);
975 	}
976 
977 	error = kern_wait6(td, idtype, args->id, &status, options,
978 	    &wru, &siginfo);
979 	if (error != 0)
980 		return (error);
981 	if (args->rusage != NULL) {
982 		error = linux_copyout_rusage(&wru.wru_children,
983 		    args->rusage);
984 		if (error != 0)
985 			return (error);
986 	}
987 	if (args->info != NULL) {
988 		p = td->td_proc;
989 		if (td->td_retval[0] == 0)
990 			bzero(&lsi, sizeof(lsi));
991 		else {
992 			sig = BSD_TO_LINUX_SIGNAL(siginfo.si_signo);
993 			siginfo_to_lsiginfo(&siginfo, &lsi, sig);
994 		}
995 		error = copyout(&lsi, args->info, sizeof(lsi));
996 	}
997 	td->td_retval[0] = 0;
998 
999 	return (error);
1000 }
1001 
1002 int
1003 linux_mknod(struct thread *td, struct linux_mknod_args *args)
1004 {
1005 	char *path;
1006 	int error;
1007 
1008 	LCONVPATHCREAT(td, args->path, &path);
1009 
1010 #ifdef DEBUG
1011 	if (ldebug(mknod))
1012 		printf(ARGS(mknod, "%s, %d, %ju"), path, args->mode,
1013 		    (uintmax_t)args->dev);
1014 #endif
1015 
1016 	switch (args->mode & S_IFMT) {
1017 	case S_IFIFO:
1018 	case S_IFSOCK:
1019 		error = kern_mkfifoat(td, AT_FDCWD, path, UIO_SYSSPACE,
1020 		    args->mode);
1021 		break;
1022 
1023 	case S_IFCHR:
1024 	case S_IFBLK:
1025 		error = kern_mknodat(td, AT_FDCWD, path, UIO_SYSSPACE,
1026 		    args->mode, args->dev);
1027 		break;
1028 
1029 	case S_IFDIR:
1030 		error = EPERM;
1031 		break;
1032 
1033 	case 0:
1034 		args->mode |= S_IFREG;
1035 		/* FALLTHROUGH */
1036 	case S_IFREG:
1037 		error = kern_openat(td, AT_FDCWD, path, UIO_SYSSPACE,
1038 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1039 		if (error == 0)
1040 			kern_close(td, td->td_retval[0]);
1041 		break;
1042 
1043 	default:
1044 		error = EINVAL;
1045 		break;
1046 	}
1047 	LFREEPATH(path);
1048 	return (error);
1049 }
1050 
1051 int
1052 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
1053 {
1054 	char *path;
1055 	int error, dfd;
1056 
1057 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
1058 	LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
1059 
1060 #ifdef DEBUG
1061 	if (ldebug(mknodat))
1062 		printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev);
1063 #endif
1064 
1065 	switch (args->mode & S_IFMT) {
1066 	case S_IFIFO:
1067 	case S_IFSOCK:
1068 		error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
1069 		break;
1070 
1071 	case S_IFCHR:
1072 	case S_IFBLK:
1073 		error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
1074 		    args->dev);
1075 		break;
1076 
1077 	case S_IFDIR:
1078 		error = EPERM;
1079 		break;
1080 
1081 	case 0:
1082 		args->mode |= S_IFREG;
1083 		/* FALLTHROUGH */
1084 	case S_IFREG:
1085 		error = kern_openat(td, dfd, path, UIO_SYSSPACE,
1086 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1087 		if (error == 0)
1088 			kern_close(td, td->td_retval[0]);
1089 		break;
1090 
1091 	default:
1092 		error = EINVAL;
1093 		break;
1094 	}
1095 	LFREEPATH(path);
1096 	return (error);
1097 }
1098 
1099 /*
1100  * UGH! This is just about the dumbest idea I've ever heard!!
1101  */
1102 int
1103 linux_personality(struct thread *td, struct linux_personality_args *args)
1104 {
1105 #ifdef DEBUG
1106 	if (ldebug(personality))
1107 		printf(ARGS(personality, "%lu"), (unsigned long)args->per);
1108 #endif
1109 	if (args->per != 0)
1110 		return (EINVAL);
1111 
1112 	/* Yes Jim, it's still a Linux... */
1113 	td->td_retval[0] = 0;
1114 	return (0);
1115 }
1116 
1117 struct l_itimerval {
1118 	l_timeval it_interval;
1119 	l_timeval it_value;
1120 };
1121 
1122 #define	B2L_ITIMERVAL(bip, lip) 					\
1123 	(bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec;		\
1124 	(bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec;	\
1125 	(bip)->it_value.tv_sec = (lip)->it_value.tv_sec;		\
1126 	(bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1127 
1128 int
1129 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1130 {
1131 	int error;
1132 	struct l_itimerval ls;
1133 	struct itimerval aitv, oitv;
1134 
1135 #ifdef DEBUG
1136 	if (ldebug(setitimer))
1137 		printf(ARGS(setitimer, "%p, %p"),
1138 		    (void *)uap->itv, (void *)uap->oitv);
1139 #endif
1140 
1141 	if (uap->itv == NULL) {
1142 		uap->itv = uap->oitv;
1143 		return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1144 	}
1145 
1146 	error = copyin(uap->itv, &ls, sizeof(ls));
1147 	if (error != 0)
1148 		return (error);
1149 	B2L_ITIMERVAL(&aitv, &ls);
1150 #ifdef DEBUG
1151 	if (ldebug(setitimer)) {
1152 		printf("setitimer: value: sec: %jd, usec: %ld\n",
1153 		    (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1154 		printf("setitimer: interval: sec: %jd, usec: %ld\n",
1155 		    (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1156 	}
1157 #endif
1158 	error = kern_setitimer(td, uap->which, &aitv, &oitv);
1159 	if (error != 0 || uap->oitv == NULL)
1160 		return (error);
1161 	B2L_ITIMERVAL(&ls, &oitv);
1162 
1163 	return (copyout(&ls, uap->oitv, sizeof(ls)));
1164 }
1165 
1166 int
1167 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1168 {
1169 	int error;
1170 	struct l_itimerval ls;
1171 	struct itimerval aitv;
1172 
1173 #ifdef DEBUG
1174 	if (ldebug(getitimer))
1175 		printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1176 #endif
1177 	error = kern_getitimer(td, uap->which, &aitv);
1178 	if (error != 0)
1179 		return (error);
1180 	B2L_ITIMERVAL(&ls, &aitv);
1181 	return (copyout(&ls, uap->itv, sizeof(ls)));
1182 }
1183 
1184 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1185 int
1186 linux_nice(struct thread *td, struct linux_nice_args *args)
1187 {
1188 	struct setpriority_args bsd_args;
1189 
1190 	bsd_args.which = PRIO_PROCESS;
1191 	bsd_args.who = 0;		/* current process */
1192 	bsd_args.prio = args->inc;
1193 	return (sys_setpriority(td, &bsd_args));
1194 }
1195 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1196 
1197 int
1198 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1199 {
1200 	struct ucred *newcred, *oldcred;
1201 	l_gid_t *linux_gidset;
1202 	gid_t *bsd_gidset;
1203 	int ngrp, error;
1204 	struct proc *p;
1205 
1206 	ngrp = args->gidsetsize;
1207 	if (ngrp < 0 || ngrp >= ngroups_max + 1)
1208 		return (EINVAL);
1209 	linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_TEMP, M_WAITOK);
1210 	error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1211 	if (error)
1212 		goto out;
1213 	newcred = crget();
1214 	p = td->td_proc;
1215 	PROC_LOCK(p);
1216 	oldcred = crcopysafe(p, newcred);
1217 
1218 	/*
1219 	 * cr_groups[0] holds egid. Setting the whole set from
1220 	 * the supplied set will cause egid to be changed too.
1221 	 * Keep cr_groups[0] unchanged to prevent that.
1222 	 */
1223 
1224 	if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) {
1225 		PROC_UNLOCK(p);
1226 		crfree(newcred);
1227 		goto out;
1228 	}
1229 
1230 	if (ngrp > 0) {
1231 		newcred->cr_ngroups = ngrp + 1;
1232 
1233 		bsd_gidset = newcred->cr_groups;
1234 		ngrp--;
1235 		while (ngrp >= 0) {
1236 			bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1237 			ngrp--;
1238 		}
1239 	} else
1240 		newcred->cr_ngroups = 1;
1241 
1242 	setsugid(p);
1243 	proc_set_cred(p, newcred);
1244 	PROC_UNLOCK(p);
1245 	crfree(oldcred);
1246 	error = 0;
1247 out:
1248 	free(linux_gidset, M_TEMP);
1249 	return (error);
1250 }
1251 
1252 int
1253 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1254 {
1255 	struct ucred *cred;
1256 	l_gid_t *linux_gidset;
1257 	gid_t *bsd_gidset;
1258 	int bsd_gidsetsz, ngrp, error;
1259 
1260 	cred = td->td_ucred;
1261 	bsd_gidset = cred->cr_groups;
1262 	bsd_gidsetsz = cred->cr_ngroups - 1;
1263 
1264 	/*
1265 	 * cr_groups[0] holds egid. Returning the whole set
1266 	 * here will cause a duplicate. Exclude cr_groups[0]
1267 	 * to prevent that.
1268 	 */
1269 
1270 	if ((ngrp = args->gidsetsize) == 0) {
1271 		td->td_retval[0] = bsd_gidsetsz;
1272 		return (0);
1273 	}
1274 
1275 	if (ngrp < bsd_gidsetsz)
1276 		return (EINVAL);
1277 
1278 	ngrp = 0;
1279 	linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1280 	    M_TEMP, M_WAITOK);
1281 	while (ngrp < bsd_gidsetsz) {
1282 		linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1283 		ngrp++;
1284 	}
1285 
1286 	error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1287 	free(linux_gidset, M_TEMP);
1288 	if (error)
1289 		return (error);
1290 
1291 	td->td_retval[0] = ngrp;
1292 	return (0);
1293 }
1294 
1295 int
1296 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1297 {
1298 	struct rlimit bsd_rlim;
1299 	struct l_rlimit rlim;
1300 	u_int which;
1301 	int error;
1302 
1303 #ifdef DEBUG
1304 	if (ldebug(setrlimit))
1305 		printf(ARGS(setrlimit, "%d, %p"),
1306 		    args->resource, (void *)args->rlim);
1307 #endif
1308 
1309 	if (args->resource >= LINUX_RLIM_NLIMITS)
1310 		return (EINVAL);
1311 
1312 	which = linux_to_bsd_resource[args->resource];
1313 	if (which == -1)
1314 		return (EINVAL);
1315 
1316 	error = copyin(args->rlim, &rlim, sizeof(rlim));
1317 	if (error)
1318 		return (error);
1319 
1320 	bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1321 	bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1322 	return (kern_setrlimit(td, which, &bsd_rlim));
1323 }
1324 
1325 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1326 int
1327 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1328 {
1329 	struct l_rlimit rlim;
1330 	struct proc *p = td->td_proc;
1331 	struct rlimit bsd_rlim;
1332 	u_int which;
1333 
1334 #ifdef DEBUG
1335 	if (ldebug(old_getrlimit))
1336 		printf(ARGS(old_getrlimit, "%d, %p"),
1337 		    args->resource, (void *)args->rlim);
1338 #endif
1339 
1340 	if (args->resource >= LINUX_RLIM_NLIMITS)
1341 		return (EINVAL);
1342 
1343 	which = linux_to_bsd_resource[args->resource];
1344 	if (which == -1)
1345 		return (EINVAL);
1346 
1347 	PROC_LOCK(p);
1348 	lim_rlimit(p, which, &bsd_rlim);
1349 	PROC_UNLOCK(p);
1350 
1351 #ifdef COMPAT_LINUX32
1352 	rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1353 	if (rlim.rlim_cur == UINT_MAX)
1354 		rlim.rlim_cur = INT_MAX;
1355 	rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1356 	if (rlim.rlim_max == UINT_MAX)
1357 		rlim.rlim_max = INT_MAX;
1358 #else
1359 	rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1360 	if (rlim.rlim_cur == ULONG_MAX)
1361 		rlim.rlim_cur = LONG_MAX;
1362 	rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1363 	if (rlim.rlim_max == ULONG_MAX)
1364 		rlim.rlim_max = LONG_MAX;
1365 #endif
1366 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1367 }
1368 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1369 
1370 int
1371 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1372 {
1373 	struct l_rlimit rlim;
1374 	struct proc *p = td->td_proc;
1375 	struct rlimit bsd_rlim;
1376 	u_int which;
1377 
1378 #ifdef DEBUG
1379 	if (ldebug(getrlimit))
1380 		printf(ARGS(getrlimit, "%d, %p"),
1381 		    args->resource, (void *)args->rlim);
1382 #endif
1383 
1384 	if (args->resource >= LINUX_RLIM_NLIMITS)
1385 		return (EINVAL);
1386 
1387 	which = linux_to_bsd_resource[args->resource];
1388 	if (which == -1)
1389 		return (EINVAL);
1390 
1391 	PROC_LOCK(p);
1392 	lim_rlimit(p, which, &bsd_rlim);
1393 	PROC_UNLOCK(p);
1394 
1395 	rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1396 	rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1397 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1398 }
1399 
1400 int
1401 linux_sched_setscheduler(struct thread *td,
1402     struct linux_sched_setscheduler_args *args)
1403 {
1404 	struct sched_param sched_param;
1405 	struct thread *tdt;
1406 	int error, policy;
1407 
1408 #ifdef DEBUG
1409 	if (ldebug(sched_setscheduler))
1410 		printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1411 		    args->pid, args->policy, (const void *)args->param);
1412 #endif
1413 
1414 	switch (args->policy) {
1415 	case LINUX_SCHED_OTHER:
1416 		policy = SCHED_OTHER;
1417 		break;
1418 	case LINUX_SCHED_FIFO:
1419 		policy = SCHED_FIFO;
1420 		break;
1421 	case LINUX_SCHED_RR:
1422 		policy = SCHED_RR;
1423 		break;
1424 	default:
1425 		return (EINVAL);
1426 	}
1427 
1428 	error = copyin(args->param, &sched_param, sizeof(sched_param));
1429 	if (error)
1430 		return (error);
1431 
1432 	tdt = linux_tdfind(td, args->pid, -1);
1433 	if (tdt == NULL)
1434 		return (ESRCH);
1435 
1436 	error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1437 	PROC_UNLOCK(tdt->td_proc);
1438 	return (error);
1439 }
1440 
1441 int
1442 linux_sched_getscheduler(struct thread *td,
1443     struct linux_sched_getscheduler_args *args)
1444 {
1445 	struct thread *tdt;
1446 	int error, policy;
1447 
1448 #ifdef DEBUG
1449 	if (ldebug(sched_getscheduler))
1450 		printf(ARGS(sched_getscheduler, "%d"), args->pid);
1451 #endif
1452 
1453 	tdt = linux_tdfind(td, args->pid, -1);
1454 	if (tdt == NULL)
1455 		return (ESRCH);
1456 
1457 	error = kern_sched_getscheduler(td, tdt, &policy);
1458 	PROC_UNLOCK(tdt->td_proc);
1459 
1460 	switch (policy) {
1461 	case SCHED_OTHER:
1462 		td->td_retval[0] = LINUX_SCHED_OTHER;
1463 		break;
1464 	case SCHED_FIFO:
1465 		td->td_retval[0] = LINUX_SCHED_FIFO;
1466 		break;
1467 	case SCHED_RR:
1468 		td->td_retval[0] = LINUX_SCHED_RR;
1469 		break;
1470 	}
1471 	return (error);
1472 }
1473 
1474 int
1475 linux_sched_get_priority_max(struct thread *td,
1476     struct linux_sched_get_priority_max_args *args)
1477 {
1478 	struct sched_get_priority_max_args bsd;
1479 
1480 #ifdef DEBUG
1481 	if (ldebug(sched_get_priority_max))
1482 		printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1483 #endif
1484 
1485 	switch (args->policy) {
1486 	case LINUX_SCHED_OTHER:
1487 		bsd.policy = SCHED_OTHER;
1488 		break;
1489 	case LINUX_SCHED_FIFO:
1490 		bsd.policy = SCHED_FIFO;
1491 		break;
1492 	case LINUX_SCHED_RR:
1493 		bsd.policy = SCHED_RR;
1494 		break;
1495 	default:
1496 		return (EINVAL);
1497 	}
1498 	return (sys_sched_get_priority_max(td, &bsd));
1499 }
1500 
1501 int
1502 linux_sched_get_priority_min(struct thread *td,
1503     struct linux_sched_get_priority_min_args *args)
1504 {
1505 	struct sched_get_priority_min_args bsd;
1506 
1507 #ifdef DEBUG
1508 	if (ldebug(sched_get_priority_min))
1509 		printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1510 #endif
1511 
1512 	switch (args->policy) {
1513 	case LINUX_SCHED_OTHER:
1514 		bsd.policy = SCHED_OTHER;
1515 		break;
1516 	case LINUX_SCHED_FIFO:
1517 		bsd.policy = SCHED_FIFO;
1518 		break;
1519 	case LINUX_SCHED_RR:
1520 		bsd.policy = SCHED_RR;
1521 		break;
1522 	default:
1523 		return (EINVAL);
1524 	}
1525 	return (sys_sched_get_priority_min(td, &bsd));
1526 }
1527 
1528 #define REBOOT_CAD_ON	0x89abcdef
1529 #define REBOOT_CAD_OFF	0
1530 #define REBOOT_HALT	0xcdef0123
1531 #define REBOOT_RESTART	0x01234567
1532 #define REBOOT_RESTART2	0xA1B2C3D4
1533 #define REBOOT_POWEROFF	0x4321FEDC
1534 #define REBOOT_MAGIC1	0xfee1dead
1535 #define REBOOT_MAGIC2	0x28121969
1536 #define REBOOT_MAGIC2A	0x05121996
1537 #define REBOOT_MAGIC2B	0x16041998
1538 
1539 int
1540 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1541 {
1542 	struct reboot_args bsd_args;
1543 
1544 #ifdef DEBUG
1545 	if (ldebug(reboot))
1546 		printf(ARGS(reboot, "0x%x"), args->cmd);
1547 #endif
1548 
1549 	if (args->magic1 != REBOOT_MAGIC1)
1550 		return (EINVAL);
1551 
1552 	switch (args->magic2) {
1553 	case REBOOT_MAGIC2:
1554 	case REBOOT_MAGIC2A:
1555 	case REBOOT_MAGIC2B:
1556 		break;
1557 	default:
1558 		return (EINVAL);
1559 	}
1560 
1561 	switch (args->cmd) {
1562 	case REBOOT_CAD_ON:
1563 	case REBOOT_CAD_OFF:
1564 		return (priv_check(td, PRIV_REBOOT));
1565 	case REBOOT_HALT:
1566 		bsd_args.opt = RB_HALT;
1567 		break;
1568 	case REBOOT_RESTART:
1569 	case REBOOT_RESTART2:
1570 		bsd_args.opt = 0;
1571 		break;
1572 	case REBOOT_POWEROFF:
1573 		bsd_args.opt = RB_POWEROFF;
1574 		break;
1575 	default:
1576 		return (EINVAL);
1577 	}
1578 	return (sys_reboot(td, &bsd_args));
1579 }
1580 
1581 
1582 /*
1583  * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1584  * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
1585  * are assumed to be preserved. The following lightweight syscalls fixes
1586  * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
1587  *
1588  * linux_getpid() - MP SAFE
1589  * linux_getgid() - MP SAFE
1590  * linux_getuid() - MP SAFE
1591  */
1592 
1593 int
1594 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1595 {
1596 
1597 #ifdef DEBUG
1598 	if (ldebug(getpid))
1599 		printf(ARGS(getpid, ""));
1600 #endif
1601 	td->td_retval[0] = td->td_proc->p_pid;
1602 
1603 	return (0);
1604 }
1605 
1606 int
1607 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1608 {
1609 	struct linux_emuldata *em;
1610 
1611 #ifdef DEBUG
1612 	if (ldebug(gettid))
1613 		printf(ARGS(gettid, ""));
1614 #endif
1615 
1616 	em = em_find(td);
1617 	KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1618 
1619 	td->td_retval[0] = em->em_tid;
1620 
1621 	return (0);
1622 }
1623 
1624 
1625 int
1626 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1627 {
1628 
1629 #ifdef DEBUG
1630 	if (ldebug(getppid))
1631 		printf(ARGS(getppid, ""));
1632 #endif
1633 
1634 	PROC_LOCK(td->td_proc);
1635 	td->td_retval[0] = td->td_proc->p_pptr->p_pid;
1636 	PROC_UNLOCK(td->td_proc);
1637 	return (0);
1638 }
1639 
1640 int
1641 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1642 {
1643 
1644 #ifdef DEBUG
1645 	if (ldebug(getgid))
1646 		printf(ARGS(getgid, ""));
1647 #endif
1648 
1649 	td->td_retval[0] = td->td_ucred->cr_rgid;
1650 	return (0);
1651 }
1652 
1653 int
1654 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1655 {
1656 
1657 #ifdef DEBUG
1658 	if (ldebug(getuid))
1659 		printf(ARGS(getuid, ""));
1660 #endif
1661 
1662 	td->td_retval[0] = td->td_ucred->cr_ruid;
1663 	return (0);
1664 }
1665 
1666 
1667 int
1668 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1669 {
1670 	struct getsid_args bsd;
1671 
1672 #ifdef DEBUG
1673 	if (ldebug(getsid))
1674 		printf(ARGS(getsid, "%i"), args->pid);
1675 #endif
1676 
1677 	bsd.pid = args->pid;
1678 	return (sys_getsid(td, &bsd));
1679 }
1680 
1681 int
1682 linux_nosys(struct thread *td, struct nosys_args *ignore)
1683 {
1684 
1685 	return (ENOSYS);
1686 }
1687 
1688 int
1689 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1690 {
1691 	struct getpriority_args bsd_args;
1692 	int error;
1693 
1694 #ifdef DEBUG
1695 	if (ldebug(getpriority))
1696 		printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
1697 #endif
1698 
1699 	bsd_args.which = args->which;
1700 	bsd_args.who = args->who;
1701 	error = sys_getpriority(td, &bsd_args);
1702 	td->td_retval[0] = 20 - td->td_retval[0];
1703 	return (error);
1704 }
1705 
1706 int
1707 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1708 {
1709 	int name[2];
1710 
1711 #ifdef DEBUG
1712 	if (ldebug(sethostname))
1713 		printf(ARGS(sethostname, "*, %i"), args->len);
1714 #endif
1715 
1716 	name[0] = CTL_KERN;
1717 	name[1] = KERN_HOSTNAME;
1718 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1719 	    args->len, 0, 0));
1720 }
1721 
1722 int
1723 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1724 {
1725 	int name[2];
1726 
1727 #ifdef DEBUG
1728 	if (ldebug(setdomainname))
1729 		printf(ARGS(setdomainname, "*, %i"), args->len);
1730 #endif
1731 
1732 	name[0] = CTL_KERN;
1733 	name[1] = KERN_NISDOMAINNAME;
1734 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1735 	    args->len, 0, 0));
1736 }
1737 
1738 int
1739 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1740 {
1741 
1742 #ifdef DEBUG
1743 	if (ldebug(exit_group))
1744 		printf(ARGS(exit_group, "%i"), args->error_code);
1745 #endif
1746 
1747 	LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1748 	    args->error_code);
1749 
1750 	/*
1751 	 * XXX: we should send a signal to the parent if
1752 	 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1753 	 * as it doesnt occur often.
1754 	 */
1755 	exit1(td, W_EXITCODE(args->error_code, 0));
1756 		/* NOTREACHED */
1757 }
1758 
1759 #define _LINUX_CAPABILITY_VERSION  0x19980330
1760 
1761 struct l_user_cap_header {
1762 	l_int	version;
1763 	l_int	pid;
1764 };
1765 
1766 struct l_user_cap_data {
1767 	l_int	effective;
1768 	l_int	permitted;
1769 	l_int	inheritable;
1770 };
1771 
1772 int
1773 linux_capget(struct thread *td, struct linux_capget_args *args)
1774 {
1775 	struct l_user_cap_header luch;
1776 	struct l_user_cap_data lucd;
1777 	int error;
1778 
1779 	if (args->hdrp == NULL)
1780 		return (EFAULT);
1781 
1782 	error = copyin(args->hdrp, &luch, sizeof(luch));
1783 	if (error != 0)
1784 		return (error);
1785 
1786 	if (luch.version != _LINUX_CAPABILITY_VERSION) {
1787 		luch.version = _LINUX_CAPABILITY_VERSION;
1788 		error = copyout(&luch, args->hdrp, sizeof(luch));
1789 		if (error)
1790 			return (error);
1791 		return (EINVAL);
1792 	}
1793 
1794 	if (luch.pid)
1795 		return (EPERM);
1796 
1797 	if (args->datap) {
1798 		/*
1799 		 * The current implementation doesn't support setting
1800 		 * a capability (it's essentially a stub) so indicate
1801 		 * that no capabilities are currently set or available
1802 		 * to request.
1803 		 */
1804 		bzero (&lucd, sizeof(lucd));
1805 		error = copyout(&lucd, args->datap, sizeof(lucd));
1806 	}
1807 
1808 	return (error);
1809 }
1810 
1811 int
1812 linux_capset(struct thread *td, struct linux_capset_args *args)
1813 {
1814 	struct l_user_cap_header luch;
1815 	struct l_user_cap_data lucd;
1816 	int error;
1817 
1818 	if (args->hdrp == NULL || args->datap == NULL)
1819 		return (EFAULT);
1820 
1821 	error = copyin(args->hdrp, &luch, sizeof(luch));
1822 	if (error != 0)
1823 		return (error);
1824 
1825 	if (luch.version != _LINUX_CAPABILITY_VERSION) {
1826 		luch.version = _LINUX_CAPABILITY_VERSION;
1827 		error = copyout(&luch, args->hdrp, sizeof(luch));
1828 		if (error)
1829 			return (error);
1830 		return (EINVAL);
1831 	}
1832 
1833 	if (luch.pid)
1834 		return (EPERM);
1835 
1836 	error = copyin(args->datap, &lucd, sizeof(lucd));
1837 	if (error != 0)
1838 		return (error);
1839 
1840 	/* We currently don't support setting any capabilities. */
1841 	if (lucd.effective || lucd.permitted || lucd.inheritable) {
1842 		linux_msg(td,
1843 			  "capset effective=0x%x, permitted=0x%x, "
1844 			  "inheritable=0x%x is not implemented",
1845 			  (int)lucd.effective, (int)lucd.permitted,
1846 			  (int)lucd.inheritable);
1847 		return (EPERM);
1848 	}
1849 
1850 	return (0);
1851 }
1852 
1853 int
1854 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1855 {
1856 	int error = 0, max_size;
1857 	struct proc *p = td->td_proc;
1858 	char comm[LINUX_MAX_COMM_LEN];
1859 	struct linux_emuldata *em;
1860 	int pdeath_signal;
1861 
1862 #ifdef DEBUG
1863 	if (ldebug(prctl))
1864 		printf(ARGS(prctl, "%d, %ju, %ju, %ju, %ju"), args->option,
1865 		    (uintmax_t)args->arg2, (uintmax_t)args->arg3,
1866 		    (uintmax_t)args->arg4, (uintmax_t)args->arg5);
1867 #endif
1868 
1869 	switch (args->option) {
1870 	case LINUX_PR_SET_PDEATHSIG:
1871 		if (!LINUX_SIG_VALID(args->arg2))
1872 			return (EINVAL);
1873 		em = em_find(td);
1874 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1875 		em->pdeath_signal = args->arg2;
1876 		break;
1877 	case LINUX_PR_GET_PDEATHSIG:
1878 		em = em_find(td);
1879 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1880 		pdeath_signal = em->pdeath_signal;
1881 		error = copyout(&pdeath_signal,
1882 		    (void *)(register_t)args->arg2,
1883 		    sizeof(pdeath_signal));
1884 		break;
1885 	case LINUX_PR_GET_KEEPCAPS:
1886 		/*
1887 		 * Indicate that we always clear the effective and
1888 		 * permitted capability sets when the user id becomes
1889 		 * non-zero (actually the capability sets are simply
1890 		 * always zero in the current implementation).
1891 		 */
1892 		td->td_retval[0] = 0;
1893 		break;
1894 	case LINUX_PR_SET_KEEPCAPS:
1895 		/*
1896 		 * Ignore requests to keep the effective and permitted
1897 		 * capability sets when the user id becomes non-zero.
1898 		 */
1899 		break;
1900 	case LINUX_PR_SET_NAME:
1901 		/*
1902 		 * To be on the safe side we need to make sure to not
1903 		 * overflow the size a linux program expects. We already
1904 		 * do this here in the copyin, so that we don't need to
1905 		 * check on copyout.
1906 		 */
1907 		max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1908 		error = copyinstr((void *)(register_t)args->arg2, comm,
1909 		    max_size, NULL);
1910 
1911 		/* Linux silently truncates the name if it is too long. */
1912 		if (error == ENAMETOOLONG) {
1913 			/*
1914 			 * XXX: copyinstr() isn't documented to populate the
1915 			 * array completely, so do a copyin() to be on the
1916 			 * safe side. This should be changed in case
1917 			 * copyinstr() is changed to guarantee this.
1918 			 */
1919 			error = copyin((void *)(register_t)args->arg2, comm,
1920 			    max_size - 1);
1921 			comm[max_size - 1] = '\0';
1922 		}
1923 		if (error)
1924 			return (error);
1925 
1926 		PROC_LOCK(p);
1927 		strlcpy(p->p_comm, comm, sizeof(p->p_comm));
1928 		PROC_UNLOCK(p);
1929 		break;
1930 	case LINUX_PR_GET_NAME:
1931 		PROC_LOCK(p);
1932 		strlcpy(comm, p->p_comm, sizeof(comm));
1933 		PROC_UNLOCK(p);
1934 		error = copyout(comm, (void *)(register_t)args->arg2,
1935 		    strlen(comm) + 1);
1936 		break;
1937 	default:
1938 		error = EINVAL;
1939 		break;
1940 	}
1941 
1942 	return (error);
1943 }
1944 
1945 int
1946 linux_sched_setparam(struct thread *td,
1947     struct linux_sched_setparam_args *uap)
1948 {
1949 	struct sched_param sched_param;
1950 	struct thread *tdt;
1951 	int error;
1952 
1953 #ifdef DEBUG
1954 	if (ldebug(sched_setparam))
1955 		printf(ARGS(sched_setparam, "%d, *"), uap->pid);
1956 #endif
1957 
1958 	error = copyin(uap->param, &sched_param, sizeof(sched_param));
1959 	if (error)
1960 		return (error);
1961 
1962 	tdt = linux_tdfind(td, uap->pid, -1);
1963 	if (tdt == NULL)
1964 		return (ESRCH);
1965 
1966 	error = kern_sched_setparam(td, tdt, &sched_param);
1967 	PROC_UNLOCK(tdt->td_proc);
1968 	return (error);
1969 }
1970 
1971 int
1972 linux_sched_getparam(struct thread *td,
1973     struct linux_sched_getparam_args *uap)
1974 {
1975 	struct sched_param sched_param;
1976 	struct thread *tdt;
1977 	int error;
1978 
1979 #ifdef DEBUG
1980 	if (ldebug(sched_getparam))
1981 		printf(ARGS(sched_getparam, "%d, *"), uap->pid);
1982 #endif
1983 
1984 	tdt = linux_tdfind(td, uap->pid, -1);
1985 	if (tdt == NULL)
1986 		return (ESRCH);
1987 
1988 	error = kern_sched_getparam(td, tdt, &sched_param);
1989 	PROC_UNLOCK(tdt->td_proc);
1990 	if (error == 0)
1991 		error = copyout(&sched_param, uap->param,
1992 		    sizeof(sched_param));
1993 	return (error);
1994 }
1995 
1996 /*
1997  * Get affinity of a process.
1998  */
1999 int
2000 linux_sched_getaffinity(struct thread *td,
2001     struct linux_sched_getaffinity_args *args)
2002 {
2003 	int error;
2004 	struct thread *tdt;
2005 	struct cpuset_getaffinity_args cga;
2006 
2007 #ifdef DEBUG
2008 	if (ldebug(sched_getaffinity))
2009 		printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid,
2010 		    args->len);
2011 #endif
2012 	if (args->len < sizeof(cpuset_t))
2013 		return (EINVAL);
2014 
2015 	tdt = linux_tdfind(td, args->pid, -1);
2016 	if (tdt == NULL)
2017 		return (ESRCH);
2018 
2019 	PROC_UNLOCK(tdt->td_proc);
2020 	cga.level = CPU_LEVEL_WHICH;
2021 	cga.which = CPU_WHICH_TID;
2022 	cga.id = tdt->td_tid;
2023 	cga.cpusetsize = sizeof(cpuset_t);
2024 	cga.mask = (cpuset_t *) args->user_mask_ptr;
2025 
2026 	if ((error = sys_cpuset_getaffinity(td, &cga)) == 0)
2027 		td->td_retval[0] = sizeof(cpuset_t);
2028 
2029 	return (error);
2030 }
2031 
2032 /*
2033  *  Set affinity of a process.
2034  */
2035 int
2036 linux_sched_setaffinity(struct thread *td,
2037     struct linux_sched_setaffinity_args *args)
2038 {
2039 	struct cpuset_setaffinity_args csa;
2040 	struct thread *tdt;
2041 
2042 #ifdef DEBUG
2043 	if (ldebug(sched_setaffinity))
2044 		printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid,
2045 		    args->len);
2046 #endif
2047 	if (args->len < sizeof(cpuset_t))
2048 		return (EINVAL);
2049 
2050 	tdt = linux_tdfind(td, args->pid, -1);
2051 	if (tdt == NULL)
2052 		return (ESRCH);
2053 
2054 	PROC_UNLOCK(tdt->td_proc);
2055 	csa.level = CPU_LEVEL_WHICH;
2056 	csa.which = CPU_WHICH_TID;
2057 	csa.id = tdt->td_tid;
2058 	csa.cpusetsize = sizeof(cpuset_t);
2059 	csa.mask = (cpuset_t *) args->user_mask_ptr;
2060 
2061 	return (sys_cpuset_setaffinity(td, &csa));
2062 }
2063 
2064 struct linux_rlimit64 {
2065 	uint64_t	rlim_cur;
2066 	uint64_t	rlim_max;
2067 };
2068 
2069 int
2070 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
2071 {
2072 	struct rlimit rlim, nrlim;
2073 	struct linux_rlimit64 lrlim;
2074 	struct proc *p;
2075 	u_int which;
2076 	int flags;
2077 	int error;
2078 
2079 #ifdef DEBUG
2080 	if (ldebug(prlimit64))
2081 		printf(ARGS(prlimit64, "%d, %d, %p, %p"), args->pid,
2082 		    args->resource, (void *)args->new, (void *)args->old);
2083 #endif
2084 
2085 	if (args->resource >= LINUX_RLIM_NLIMITS)
2086 		return (EINVAL);
2087 
2088 	which = linux_to_bsd_resource[args->resource];
2089 	if (which == -1)
2090 		return (EINVAL);
2091 
2092 	if (args->new != NULL) {
2093 		/*
2094 		 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
2095 		 * rlim is unsigned 64-bit. FreeBSD treats negative limits
2096 		 * as INFINITY so we do not need a conversion even.
2097 		 */
2098 		error = copyin(args->new, &nrlim, sizeof(nrlim));
2099 		if (error != 0)
2100 			return (error);
2101 	}
2102 
2103 	flags = PGET_HOLD | PGET_NOTWEXIT;
2104 	if (args->new != NULL)
2105 		flags |= PGET_CANDEBUG;
2106 	else
2107 		flags |= PGET_CANSEE;
2108 	error = pget(args->pid, flags, &p);
2109 	if (error != 0)
2110 		return (error);
2111 
2112 	if (args->old != NULL) {
2113 		PROC_LOCK(p);
2114 		lim_rlimit(p, which, &rlim);
2115 		PROC_UNLOCK(p);
2116 		if (rlim.rlim_cur == RLIM_INFINITY)
2117 			lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2118 		else
2119 			lrlim.rlim_cur = rlim.rlim_cur;
2120 		if (rlim.rlim_max == RLIM_INFINITY)
2121 			lrlim.rlim_max = LINUX_RLIM_INFINITY;
2122 		else
2123 			lrlim.rlim_max = rlim.rlim_max;
2124 		error = copyout(&lrlim, args->old, sizeof(lrlim));
2125 		if (error != 0)
2126 			goto out;
2127 	}
2128 
2129 	if (args->new != NULL)
2130 		error = kern_proc_setrlimit(td, p, which, &nrlim);
2131 
2132  out:
2133 	PRELE(p);
2134 	return (error);
2135 }
2136 
2137 int
2138 linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2139 {
2140 	struct timeval utv, tv0, tv1, *tvp;
2141 	struct l_pselect6arg lpse6;
2142 	struct l_timespec lts;
2143 	struct timespec uts;
2144 	l_sigset_t l_ss;
2145 	sigset_t *ssp;
2146 	sigset_t ss;
2147 	int error;
2148 
2149 	ssp = NULL;
2150 	if (args->sig != NULL) {
2151 		error = copyin(args->sig, &lpse6, sizeof(lpse6));
2152 		if (error != 0)
2153 			return (error);
2154 		if (lpse6.ss_len != sizeof(l_ss))
2155 			return (EINVAL);
2156 		if (lpse6.ss != 0) {
2157 			error = copyin(PTRIN(lpse6.ss), &l_ss,
2158 			    sizeof(l_ss));
2159 			if (error != 0)
2160 				return (error);
2161 			linux_to_bsd_sigset(&l_ss, &ss);
2162 			ssp = &ss;
2163 		}
2164 	}
2165 
2166 	/*
2167 	 * Currently glibc changes nanosecond number to microsecond.
2168 	 * This mean losing precision but for now it is hardly seen.
2169 	 */
2170 	if (args->tsp != NULL) {
2171 		error = copyin(args->tsp, &lts, sizeof(lts));
2172 		if (error != 0)
2173 			return (error);
2174 		uts.tv_sec = lts.tv_sec;
2175 		uts.tv_nsec = lts.tv_nsec;
2176 
2177 		TIMESPEC_TO_TIMEVAL(&utv, &uts);
2178 		if (itimerfix(&utv))
2179 			return (EINVAL);
2180 
2181 		microtime(&tv0);
2182 		tvp = &utv;
2183 	} else
2184 		tvp = NULL;
2185 
2186 	error = kern_pselect(td, args->nfds, args->readfds, args->writefds,
2187 	    args->exceptfds, tvp, ssp, sizeof(l_int) * 8);
2188 
2189 	if (error == 0 && args->tsp != NULL) {
2190 		if (td->td_retval[0] != 0) {
2191 			/*
2192 			 * Compute how much time was left of the timeout,
2193 			 * by subtracting the current time and the time
2194 			 * before we started the call, and subtracting
2195 			 * that result from the user-supplied value.
2196 			 */
2197 
2198 			microtime(&tv1);
2199 			timevalsub(&tv1, &tv0);
2200 			timevalsub(&utv, &tv1);
2201 			if (utv.tv_sec < 0)
2202 				timevalclear(&utv);
2203 		} else
2204 			timevalclear(&utv);
2205 
2206 		TIMEVAL_TO_TIMESPEC(&utv, &uts);
2207 		lts.tv_sec = uts.tv_sec;
2208 		lts.tv_nsec = uts.tv_nsec;
2209 		error = copyout(&lts, args->tsp, sizeof(lts));
2210 	}
2211 
2212 	return (error);
2213 }
2214 
2215 #if defined(DEBUG) || defined(KTR)
2216 /* XXX: can be removed when every ldebug(...) and KTR stuff are removed. */
2217 
2218 u_char linux_debug_map[howmany(LINUX_SYS_MAXSYSCALL, sizeof(u_char))];
2219 
2220 static int
2221 linux_debug(int syscall, int toggle, int global)
2222 {
2223 
2224 	if (global) {
2225 		char c = toggle ? 0 : 0xff;
2226 
2227 		memset(linux_debug_map, c, sizeof(linux_debug_map));
2228 		return (0);
2229 	}
2230 	if (syscall < 0 || syscall >= LINUX_SYS_MAXSYSCALL)
2231 		return (EINVAL);
2232 	if (toggle)
2233 		clrbit(linux_debug_map, syscall);
2234 	else
2235 		setbit(linux_debug_map, syscall);
2236 	return (0);
2237 }
2238 
2239 /*
2240  * Usage: sysctl linux.debug=<syscall_nr>.<0/1>
2241  *
2242  *    E.g.: sysctl linux.debug=21.0
2243  *
2244  * As a special case, syscall "all" will apply to all syscalls globally.
2245  */
2246 #define LINUX_MAX_DEBUGSTR	16
2247 int
2248 linux_sysctl_debug(SYSCTL_HANDLER_ARGS)
2249 {
2250 	char value[LINUX_MAX_DEBUGSTR], *p;
2251 	int error, sysc, toggle;
2252 	int global = 0;
2253 
2254 	value[0] = '\0';
2255 	error = sysctl_handle_string(oidp, value, LINUX_MAX_DEBUGSTR, req);
2256 	if (error || req->newptr == NULL)
2257 		return (error);
2258 	for (p = value; *p != '\0' && *p != '.'; p++);
2259 	if (*p == '\0')
2260 		return (EINVAL);
2261 	*p++ = '\0';
2262 	sysc = strtol(value, NULL, 0);
2263 	toggle = strtol(p, NULL, 0);
2264 	if (strcmp(value, "all") == 0)
2265 		global = 1;
2266 	error = linux_debug(sysc, toggle, global);
2267 	return (error);
2268 }
2269 
2270 #endif /* DEBUG || KTR */
2271 
2272 int
2273 linux_sched_rr_get_interval(struct thread *td,
2274     struct linux_sched_rr_get_interval_args *uap)
2275 {
2276 	struct timespec ts;
2277 	struct l_timespec lts;
2278 	struct thread *tdt;
2279 	int error;
2280 
2281 	/*
2282 	 * According to man in case the invalid pid specified
2283 	 * EINVAL should be returned.
2284 	 */
2285 	if (uap->pid < 0)
2286 		return (EINVAL);
2287 
2288 	tdt = linux_tdfind(td, uap->pid, -1);
2289 	if (tdt == NULL)
2290 		return (ESRCH);
2291 
2292 	error = kern_sched_rr_get_interval_td(td, tdt, &ts);
2293 	PROC_UNLOCK(tdt->td_proc);
2294 	if (error != 0)
2295 		return (error);
2296 	lts.tv_sec = ts.tv_sec;
2297 	lts.tv_nsec = ts.tv_nsec;
2298 	return (copyout(&lts, uap->interval, sizeof(lts)));
2299 }
2300 
2301 /*
2302  * In case when the Linux thread is the initial thread in
2303  * the thread group thread id is equal to the process id.
2304  * Glibc depends on this magic (assert in pthread_getattr_np.c).
2305  */
2306 struct thread *
2307 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid)
2308 {
2309 	struct linux_emuldata *em;
2310 	struct thread *tdt;
2311 	struct proc *p;
2312 
2313 	tdt = NULL;
2314 	if (tid == 0 || tid == td->td_tid) {
2315 		tdt = td;
2316 		PROC_LOCK(tdt->td_proc);
2317 	} else if (tid > PID_MAX)
2318 		tdt = tdfind(tid, pid);
2319 	else {
2320 		/*
2321 		 * Initial thread where the tid equal to the pid.
2322 		 */
2323 		p = pfind(tid);
2324 		if (p != NULL) {
2325 			if (SV_PROC_ABI(p) != SV_ABI_LINUX) {
2326 				/*
2327 				 * p is not a Linuxulator process.
2328 				 */
2329 				PROC_UNLOCK(p);
2330 				return (NULL);
2331 			}
2332 			FOREACH_THREAD_IN_PROC(p, tdt) {
2333 				em = em_find(tdt);
2334 				if (tid == em->em_tid)
2335 					return (tdt);
2336 			}
2337 			PROC_UNLOCK(p);
2338 		}
2339 		return (NULL);
2340 	}
2341 
2342 	return (tdt);
2343 }
2344 
2345 void
2346 linux_to_bsd_waitopts(int options, int *bsdopts)
2347 {
2348 
2349 	if (options & LINUX_WNOHANG)
2350 		*bsdopts |= WNOHANG;
2351 	if (options & LINUX_WUNTRACED)
2352 		*bsdopts |= WUNTRACED;
2353 	if (options & LINUX_WEXITED)
2354 		*bsdopts |= WEXITED;
2355 	if (options & LINUX_WCONTINUED)
2356 		*bsdopts |= WCONTINUED;
2357 	if (options & LINUX_WNOWAIT)
2358 		*bsdopts |= WNOWAIT;
2359 
2360 	if (options & __WCLONE)
2361 		*bsdopts |= WLINUXCLONE;
2362 }
2363