xref: /freebsd/sys/compat/linux/linux_misc.c (revision ec513841b3ac59ce581ffb5eb859926555f6834b)
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/signalvar.h>
57 #include <sys/stat.h>
58 #include <sys/syscallsubr.h>
59 #include <sys/sysctl.h>
60 #include <sys/sysproto.h>
61 #include <sys/systm.h>
62 #include <sys/time.h>
63 #include <sys/vmmeter.h>
64 #include <sys/vnode.h>
65 #include <sys/wait.h>
66 #include <sys/cpuset.h>
67 
68 #include <security/mac/mac_framework.h>
69 
70 #include <vm/vm.h>
71 #include <vm/pmap.h>
72 #include <vm/vm_kern.h>
73 #include <vm/vm_map.h>
74 #include <vm/vm_extern.h>
75 #include <vm/vm_object.h>
76 #include <vm/swap_pager.h>
77 
78 #ifdef COMPAT_LINUX32
79 #include <machine/../linux32/linux.h>
80 #include <machine/../linux32/linux32_proto.h>
81 #else
82 #include <machine/../linux/linux.h>
83 #include <machine/../linux/linux_proto.h>
84 #endif
85 
86 #include <compat/linux/linux_file.h>
87 #include <compat/linux/linux_mib.h>
88 #include <compat/linux/linux_signal.h>
89 #include <compat/linux/linux_util.h>
90 #include <compat/linux/linux_sysproto.h>
91 #include <compat/linux/linux_emul.h>
92 #include <compat/linux/linux_misc.h>
93 
94 int stclohz;				/* Statistics clock frequency */
95 
96 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
97 	RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
98 	RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
99 	RLIMIT_MEMLOCK, RLIMIT_AS
100 };
101 
102 struct l_sysinfo {
103 	l_long		uptime;		/* Seconds since boot */
104 	l_ulong		loads[3];	/* 1, 5, and 15 minute load averages */
105 #define LINUX_SYSINFO_LOADS_SCALE 65536
106 	l_ulong		totalram;	/* Total usable main memory size */
107 	l_ulong		freeram;	/* Available memory size */
108 	l_ulong		sharedram;	/* Amount of shared memory */
109 	l_ulong		bufferram;	/* Memory used by buffers */
110 	l_ulong		totalswap;	/* Total swap space size */
111 	l_ulong		freeswap;	/* swap space still available */
112 	l_ushort	procs;		/* Number of current processes */
113 	l_ushort	pads;
114 	l_ulong		totalbig;
115 	l_ulong		freebig;
116 	l_uint		mem_unit;
117 	char		_f[20-2*sizeof(l_long)-sizeof(l_int)];	/* padding */
118 };
119 
120 struct l_pselect6arg {
121 	l_uintptr_t	ss;
122 	l_size_t	ss_len;
123 };
124 
125 int
126 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
127 {
128 	struct l_sysinfo sysinfo;
129 	vm_object_t object;
130 	int i, j;
131 	struct timespec ts;
132 
133 	getnanouptime(&ts);
134 	if (ts.tv_nsec != 0)
135 		ts.tv_sec++;
136 	sysinfo.uptime = ts.tv_sec;
137 
138 	/* Use the information from the mib to get our load averages */
139 	for (i = 0; i < 3; i++)
140 		sysinfo.loads[i] = averunnable.ldavg[i] *
141 		    LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
142 
143 	sysinfo.totalram = physmem * PAGE_SIZE;
144 	sysinfo.freeram = sysinfo.totalram - vm_cnt.v_wire_count * PAGE_SIZE;
145 
146 	sysinfo.sharedram = 0;
147 	mtx_lock(&vm_object_list_mtx);
148 	TAILQ_FOREACH(object, &vm_object_list, object_list)
149 		if (object->shadow_count > 1)
150 			sysinfo.sharedram += object->resident_page_count;
151 	mtx_unlock(&vm_object_list_mtx);
152 
153 	sysinfo.sharedram *= PAGE_SIZE;
154 	sysinfo.bufferram = 0;
155 
156 	swap_pager_status(&i, &j);
157 	sysinfo.totalswap = i * PAGE_SIZE;
158 	sysinfo.freeswap = (i - j) * PAGE_SIZE;
159 
160 	sysinfo.procs = nprocs;
161 
162 	/* The following are only present in newer Linux kernels. */
163 	sysinfo.totalbig = 0;
164 	sysinfo.freebig = 0;
165 	sysinfo.mem_unit = 1;
166 
167 	return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
168 }
169 
170 int
171 linux_alarm(struct thread *td, struct linux_alarm_args *args)
172 {
173 	struct itimerval it, old_it;
174 	u_int secs;
175 	int error;
176 
177 #ifdef DEBUG
178 	if (ldebug(alarm))
179 		printf(ARGS(alarm, "%u"), args->secs);
180 #endif
181 
182 	secs = args->secs;
183 
184 	if (secs > INT_MAX)
185 		secs = INT_MAX;
186 
187 	it.it_value.tv_sec = (long) secs;
188 	it.it_value.tv_usec = 0;
189 	it.it_interval.tv_sec = 0;
190 	it.it_interval.tv_usec = 0;
191 	error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
192 	if (error)
193 		return (error);
194 	if (timevalisset(&old_it.it_value)) {
195 		if (old_it.it_value.tv_usec != 0)
196 			old_it.it_value.tv_sec++;
197 		td->td_retval[0] = old_it.it_value.tv_sec;
198 	}
199 	return (0);
200 }
201 
202 int
203 linux_brk(struct thread *td, struct linux_brk_args *args)
204 {
205 	struct vmspace *vm = td->td_proc->p_vmspace;
206 	vm_offset_t new, old;
207 	struct obreak_args /* {
208 		char * nsize;
209 	} */ tmp;
210 
211 #ifdef DEBUG
212 	if (ldebug(brk))
213 		printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
214 #endif
215 	old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize);
216 	new = (vm_offset_t)args->dsend;
217 	tmp.nsize = (char *)new;
218 	if (((caddr_t)new > vm->vm_daddr) && !sys_obreak(td, &tmp))
219 		td->td_retval[0] = (long)new;
220 	else
221 		td->td_retval[0] = (long)old;
222 
223 	return (0);
224 }
225 
226 #if defined(__i386__)
227 /* XXX: what about amd64/linux32? */
228 
229 int
230 linux_uselib(struct thread *td, struct linux_uselib_args *args)
231 {
232 	struct nameidata ni;
233 	struct vnode *vp;
234 	struct exec *a_out;
235 	struct vattr attr;
236 	vm_offset_t vmaddr;
237 	unsigned long file_offset;
238 	unsigned long bss_size;
239 	char *library;
240 	ssize_t aresid;
241 	int error, locked, writecount;
242 
243 	LCONVPATHEXIST(td, args->library, &library);
244 
245 #ifdef DEBUG
246 	if (ldebug(uselib))
247 		printf(ARGS(uselib, "%s"), library);
248 #endif
249 
250 	a_out = NULL;
251 	locked = 0;
252 	vp = NULL;
253 
254 	NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
255 	    UIO_SYSSPACE, library, td);
256 	error = namei(&ni);
257 	LFREEPATH(library);
258 	if (error)
259 		goto cleanup;
260 
261 	vp = ni.ni_vp;
262 	NDFREE(&ni, NDF_ONLY_PNBUF);
263 
264 	/*
265 	 * From here on down, we have a locked vnode that must be unlocked.
266 	 * XXX: The code below largely duplicates exec_check_permissions().
267 	 */
268 	locked = 1;
269 
270 	/* Writable? */
271 	error = VOP_GET_WRITECOUNT(vp, &writecount);
272 	if (error != 0)
273 		goto cleanup;
274 	if (writecount != 0) {
275 		error = ETXTBSY;
276 		goto cleanup;
277 	}
278 
279 	/* Executable? */
280 	error = VOP_GETATTR(vp, &attr, td->td_ucred);
281 	if (error)
282 		goto cleanup;
283 
284 	if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
285 	    ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
286 		/* EACCESS is what exec(2) returns. */
287 		error = ENOEXEC;
288 		goto cleanup;
289 	}
290 
291 	/* Sensible size? */
292 	if (attr.va_size == 0) {
293 		error = ENOEXEC;
294 		goto cleanup;
295 	}
296 
297 	/* Can we access it? */
298 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
299 	if (error)
300 		goto cleanup;
301 
302 	/*
303 	 * XXX: This should use vn_open() so that it is properly authorized,
304 	 * and to reduce code redundancy all over the place here.
305 	 * XXX: Not really, it duplicates far more of exec_check_permissions()
306 	 * than vn_open().
307 	 */
308 #ifdef MAC
309 	error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
310 	if (error)
311 		goto cleanup;
312 #endif
313 	error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
314 	if (error)
315 		goto cleanup;
316 
317 	/* Pull in executable header into exec_map */
318 	error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
319 	    VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
320 	if (error)
321 		goto cleanup;
322 
323 	/* Is it a Linux binary ? */
324 	if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
325 		error = ENOEXEC;
326 		goto cleanup;
327 	}
328 
329 	/*
330 	 * While we are here, we should REALLY do some more checks
331 	 */
332 
333 	/* Set file/virtual offset based on a.out variant. */
334 	switch ((int)(a_out->a_magic & 0xffff)) {
335 	case 0413:			/* ZMAGIC */
336 		file_offset = 1024;
337 		break;
338 	case 0314:			/* QMAGIC */
339 		file_offset = 0;
340 		break;
341 	default:
342 		error = ENOEXEC;
343 		goto cleanup;
344 	}
345 
346 	bss_size = round_page(a_out->a_bss);
347 
348 	/* Check various fields in header for validity/bounds. */
349 	if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
350 		error = ENOEXEC;
351 		goto cleanup;
352 	}
353 
354 	/* text + data can't exceed file size */
355 	if (a_out->a_data + a_out->a_text > attr.va_size) {
356 		error = EFAULT;
357 		goto cleanup;
358 	}
359 
360 	/*
361 	 * text/data/bss must not exceed limits
362 	 * XXX - this is not complete. it should check current usage PLUS
363 	 * the resources needed by this library.
364 	 */
365 	PROC_LOCK(td->td_proc);
366 	if (a_out->a_text > maxtsiz ||
367 	    a_out->a_data + bss_size > lim_cur(td->td_proc, RLIMIT_DATA) ||
368 	    racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
369 	    bss_size) != 0) {
370 		PROC_UNLOCK(td->td_proc);
371 		error = ENOMEM;
372 		goto cleanup;
373 	}
374 	PROC_UNLOCK(td->td_proc);
375 
376 	/*
377 	 * Prevent more writers.
378 	 * XXX: Note that if any of the VM operations fail below we don't
379 	 * clear this flag.
380 	 */
381 	VOP_SET_TEXT(vp);
382 
383 	/*
384 	 * Lock no longer needed
385 	 */
386 	locked = 0;
387 	VOP_UNLOCK(vp, 0);
388 
389 	/*
390 	 * Check if file_offset page aligned. Currently we cannot handle
391 	 * misalinged file offsets, and so we read in the entire image
392 	 * (what a waste).
393 	 */
394 	if (file_offset & PAGE_MASK) {
395 #ifdef DEBUG
396 		printf("uselib: Non page aligned binary %lu\n", file_offset);
397 #endif
398 		/* Map text+data read/write/execute */
399 
400 		/* a_entry is the load address and is page aligned */
401 		vmaddr = trunc_page(a_out->a_entry);
402 
403 		/* get anon user mapping, read+write+execute */
404 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
405 		    &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
406 		    VM_PROT_ALL, VM_PROT_ALL, 0);
407 		if (error)
408 			goto cleanup;
409 
410 		error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
411 		    a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
412 		    td->td_ucred, NOCRED, &aresid, td);
413 		if (error != 0)
414 			goto cleanup;
415 		if (aresid != 0) {
416 			error = ENOEXEC;
417 			goto cleanup;
418 		}
419 	} else {
420 #ifdef DEBUG
421 		printf("uselib: Page aligned binary %lu\n", file_offset);
422 #endif
423 		/*
424 		 * for QMAGIC, a_entry is 20 bytes beyond the load address
425 		 * to skip the executable header
426 		 */
427 		vmaddr = trunc_page(a_out->a_entry);
428 
429 		/*
430 		 * Map it all into the process's space as a single
431 		 * copy-on-write "data" segment.
432 		 */
433 		error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr,
434 		    a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
435 		    MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
436 		if (error)
437 			goto cleanup;
438 	}
439 #ifdef DEBUG
440 	printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
441 	    ((long *)vmaddr)[1]);
442 #endif
443 	if (bss_size != 0) {
444 		/* Calculate BSS start address */
445 		vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
446 		    a_out->a_data;
447 
448 		/* allocate some 'anon' space */
449 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
450 		    &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
451 		    VM_PROT_ALL, 0);
452 		if (error)
453 			goto cleanup;
454 	}
455 
456 cleanup:
457 	/* Unlock vnode if needed */
458 	if (locked)
459 		VOP_UNLOCK(vp, 0);
460 
461 	/* Release the temporary mapping. */
462 	if (a_out)
463 		kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
464 
465 	return (error);
466 }
467 
468 #endif	/* __i386__ */
469 
470 int
471 linux_select(struct thread *td, struct linux_select_args *args)
472 {
473 	l_timeval ltv;
474 	struct timeval tv0, tv1, utv, *tvp;
475 	int error;
476 
477 #ifdef DEBUG
478 	if (ldebug(select))
479 		printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
480 		    (void *)args->readfds, (void *)args->writefds,
481 		    (void *)args->exceptfds, (void *)args->timeout);
482 #endif
483 
484 	/*
485 	 * Store current time for computation of the amount of
486 	 * time left.
487 	 */
488 	if (args->timeout) {
489 		if ((error = copyin(args->timeout, &ltv, sizeof(ltv))))
490 			goto select_out;
491 		utv.tv_sec = ltv.tv_sec;
492 		utv.tv_usec = ltv.tv_usec;
493 #ifdef DEBUG
494 		if (ldebug(select))
495 			printf(LMSG("incoming timeout (%jd/%ld)"),
496 			    (intmax_t)utv.tv_sec, utv.tv_usec);
497 #endif
498 
499 		if (itimerfix(&utv)) {
500 			/*
501 			 * The timeval was invalid.  Convert it to something
502 			 * valid that will act as it does under Linux.
503 			 */
504 			utv.tv_sec += utv.tv_usec / 1000000;
505 			utv.tv_usec %= 1000000;
506 			if (utv.tv_usec < 0) {
507 				utv.tv_sec -= 1;
508 				utv.tv_usec += 1000000;
509 			}
510 			if (utv.tv_sec < 0)
511 				timevalclear(&utv);
512 		}
513 		microtime(&tv0);
514 		tvp = &utv;
515 	} else
516 		tvp = NULL;
517 
518 	error = kern_select(td, args->nfds, args->readfds, args->writefds,
519 	    args->exceptfds, tvp, sizeof(l_int) * 8);
520 
521 #ifdef DEBUG
522 	if (ldebug(select))
523 		printf(LMSG("real select returns %d"), error);
524 #endif
525 	if (error)
526 		goto select_out;
527 
528 	if (args->timeout) {
529 		if (td->td_retval[0]) {
530 			/*
531 			 * Compute how much time was left of the timeout,
532 			 * by subtracting the current time and the time
533 			 * before we started the call, and subtracting
534 			 * that result from the user-supplied value.
535 			 */
536 			microtime(&tv1);
537 			timevalsub(&tv1, &tv0);
538 			timevalsub(&utv, &tv1);
539 			if (utv.tv_sec < 0)
540 				timevalclear(&utv);
541 		} else
542 			timevalclear(&utv);
543 #ifdef DEBUG
544 		if (ldebug(select))
545 			printf(LMSG("outgoing timeout (%jd/%ld)"),
546 			    (intmax_t)utv.tv_sec, utv.tv_usec);
547 #endif
548 		ltv.tv_sec = utv.tv_sec;
549 		ltv.tv_usec = utv.tv_usec;
550 		if ((error = copyout(&ltv, args->timeout, sizeof(ltv))))
551 			goto select_out;
552 	}
553 
554 select_out:
555 #ifdef DEBUG
556 	if (ldebug(select))
557 		printf(LMSG("select_out -> %d"), error);
558 #endif
559 	return (error);
560 }
561 
562 int
563 linux_mremap(struct thread *td, struct linux_mremap_args *args)
564 {
565 	struct munmap_args /* {
566 		void *addr;
567 		size_t len;
568 	} */ bsd_args;
569 	int error = 0;
570 
571 #ifdef DEBUG
572 	if (ldebug(mremap))
573 		printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
574 		    (void *)(uintptr_t)args->addr,
575 		    (unsigned long)args->old_len,
576 		    (unsigned long)args->new_len,
577 		    (unsigned long)args->flags);
578 #endif
579 
580 	if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
581 		td->td_retval[0] = 0;
582 		return (EINVAL);
583 	}
584 
585 	/*
586 	 * Check for the page alignment.
587 	 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
588 	 */
589 	if (args->addr & PAGE_MASK) {
590 		td->td_retval[0] = 0;
591 		return (EINVAL);
592 	}
593 
594 	args->new_len = round_page(args->new_len);
595 	args->old_len = round_page(args->old_len);
596 
597 	if (args->new_len > args->old_len) {
598 		td->td_retval[0] = 0;
599 		return (ENOMEM);
600 	}
601 
602 	if (args->new_len < args->old_len) {
603 		bsd_args.addr =
604 		    (caddr_t)((uintptr_t)args->addr + args->new_len);
605 		bsd_args.len = args->old_len - args->new_len;
606 		error = sys_munmap(td, &bsd_args);
607 	}
608 
609 	td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
610 	return (error);
611 }
612 
613 #define LINUX_MS_ASYNC       0x0001
614 #define LINUX_MS_INVALIDATE  0x0002
615 #define LINUX_MS_SYNC        0x0004
616 
617 int
618 linux_msync(struct thread *td, struct linux_msync_args *args)
619 {
620 	struct msync_args bsd_args;
621 
622 	bsd_args.addr = (caddr_t)(uintptr_t)args->addr;
623 	bsd_args.len = (uintptr_t)args->len;
624 	bsd_args.flags = args->fl & ~LINUX_MS_SYNC;
625 
626 	return (sys_msync(td, &bsd_args));
627 }
628 
629 int
630 linux_time(struct thread *td, struct linux_time_args *args)
631 {
632 	struct timeval tv;
633 	l_time_t tm;
634 	int error;
635 
636 #ifdef DEBUG
637 	if (ldebug(time))
638 		printf(ARGS(time, "*"));
639 #endif
640 
641 	microtime(&tv);
642 	tm = tv.tv_sec;
643 	if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
644 		return (error);
645 	td->td_retval[0] = tm;
646 	return (0);
647 }
648 
649 struct l_times_argv {
650 	l_clock_t	tms_utime;
651 	l_clock_t	tms_stime;
652 	l_clock_t	tms_cutime;
653 	l_clock_t	tms_cstime;
654 };
655 
656 
657 /*
658  * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
659  * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
660  * auxiliary vector entry.
661  */
662 #define	CLK_TCK		100
663 
664 #define	CONVOTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
665 #define	CONVNTCK(r)	(r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
666 
667 #define	CONVTCK(r)	(linux_kernver(td) >= LINUX_KERNVER_2004000 ?		\
668 			    CONVNTCK(r) : CONVOTCK(r))
669 
670 int
671 linux_times(struct thread *td, struct linux_times_args *args)
672 {
673 	struct timeval tv, utime, stime, cutime, cstime;
674 	struct l_times_argv tms;
675 	struct proc *p;
676 	int error;
677 
678 #ifdef DEBUG
679 	if (ldebug(times))
680 		printf(ARGS(times, "*"));
681 #endif
682 
683 	if (args->buf != NULL) {
684 		p = td->td_proc;
685 		PROC_LOCK(p);
686 		PROC_STATLOCK(p);
687 		calcru(p, &utime, &stime);
688 		PROC_STATUNLOCK(p);
689 		calccru(p, &cutime, &cstime);
690 		PROC_UNLOCK(p);
691 
692 		tms.tms_utime = CONVTCK(utime);
693 		tms.tms_stime = CONVTCK(stime);
694 
695 		tms.tms_cutime = CONVTCK(cutime);
696 		tms.tms_cstime = CONVTCK(cstime);
697 
698 		if ((error = copyout(&tms, args->buf, sizeof(tms))))
699 			return (error);
700 	}
701 
702 	microuptime(&tv);
703 	td->td_retval[0] = (int)CONVTCK(tv);
704 	return (0);
705 }
706 
707 int
708 linux_newuname(struct thread *td, struct linux_newuname_args *args)
709 {
710 	struct l_new_utsname utsname;
711 	char osname[LINUX_MAX_UTSNAME];
712 	char osrelease[LINUX_MAX_UTSNAME];
713 	char *p;
714 
715 #ifdef DEBUG
716 	if (ldebug(newuname))
717 		printf(ARGS(newuname, "*"));
718 #endif
719 
720 	linux_get_osname(td, osname);
721 	linux_get_osrelease(td, osrelease);
722 
723 	bzero(&utsname, sizeof(utsname));
724 	strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
725 	getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
726 	getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
727 	strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
728 	strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
729 	for (p = utsname.version; *p != '\0'; ++p)
730 		if (*p == '\n') {
731 			*p = '\0';
732 			break;
733 		}
734 	strlcpy(utsname.machine, linux_platform, LINUX_MAX_UTSNAME);
735 
736 	return (copyout(&utsname, args->buf, sizeof(utsname)));
737 }
738 
739 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
740 struct l_utimbuf {
741 	l_time_t l_actime;
742 	l_time_t l_modtime;
743 };
744 
745 int
746 linux_utime(struct thread *td, struct linux_utime_args *args)
747 {
748 	struct timeval tv[2], *tvp;
749 	struct l_utimbuf lut;
750 	char *fname;
751 	int error;
752 
753 	LCONVPATHEXIST(td, args->fname, &fname);
754 
755 #ifdef DEBUG
756 	if (ldebug(utime))
757 		printf(ARGS(utime, "%s, *"), fname);
758 #endif
759 
760 	if (args->times) {
761 		if ((error = copyin(args->times, &lut, sizeof lut))) {
762 			LFREEPATH(fname);
763 			return (error);
764 		}
765 		tv[0].tv_sec = lut.l_actime;
766 		tv[0].tv_usec = 0;
767 		tv[1].tv_sec = lut.l_modtime;
768 		tv[1].tv_usec = 0;
769 		tvp = tv;
770 	} else
771 		tvp = NULL;
772 
773 	error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, tvp,
774 	    UIO_SYSSPACE);
775 	LFREEPATH(fname);
776 	return (error);
777 }
778 
779 int
780 linux_utimes(struct thread *td, struct linux_utimes_args *args)
781 {
782 	l_timeval ltv[2];
783 	struct timeval tv[2], *tvp = NULL;
784 	char *fname;
785 	int error;
786 
787 	LCONVPATHEXIST(td, args->fname, &fname);
788 
789 #ifdef DEBUG
790 	if (ldebug(utimes))
791 		printf(ARGS(utimes, "%s, *"), fname);
792 #endif
793 
794 	if (args->tptr != NULL) {
795 		if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
796 			LFREEPATH(fname);
797 			return (error);
798 		}
799 		tv[0].tv_sec = ltv[0].tv_sec;
800 		tv[0].tv_usec = ltv[0].tv_usec;
801 		tv[1].tv_sec = ltv[1].tv_sec;
802 		tv[1].tv_usec = ltv[1].tv_usec;
803 		tvp = tv;
804 	}
805 
806 	error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE,
807 	    tvp, UIO_SYSSPACE);
808 	LFREEPATH(fname);
809 	return (error);
810 }
811 
812 int
813 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
814 {
815 	l_timeval ltv[2];
816 	struct timeval tv[2], *tvp = NULL;
817 	char *fname;
818 	int error, dfd;
819 
820 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
821 	LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
822 
823 #ifdef DEBUG
824 	if (ldebug(futimesat))
825 		printf(ARGS(futimesat, "%s, *"), fname);
826 #endif
827 
828 	if (args->utimes != NULL) {
829 		if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
830 			LFREEPATH(fname);
831 			return (error);
832 		}
833 		tv[0].tv_sec = ltv[0].tv_sec;
834 		tv[0].tv_usec = ltv[0].tv_usec;
835 		tv[1].tv_sec = ltv[1].tv_sec;
836 		tv[1].tv_usec = ltv[1].tv_usec;
837 		tvp = tv;
838 	}
839 
840 	error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
841 	LFREEPATH(fname);
842 	return (error);
843 }
844 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
845 
846 int
847 linux_common_wait(struct thread *td, int pid, int *status,
848     int options, struct rusage *ru)
849 {
850 	int error, tmpstat;
851 
852 	error = kern_wait(td, pid, &tmpstat, options, ru);
853 	if (error)
854 		return (error);
855 
856 	if (status) {
857 		tmpstat &= 0xffff;
858 		if (WIFSIGNALED(tmpstat))
859 			tmpstat = (tmpstat & 0xffffff80) |
860 			    BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat));
861 		else if (WIFSTOPPED(tmpstat))
862 			tmpstat = (tmpstat & 0xffff00ff) |
863 			    (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8);
864 		error = copyout(&tmpstat, status, sizeof(int));
865 	}
866 
867 	return (error);
868 }
869 
870 int
871 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
872 {
873 	int options;
874 
875 #ifdef DEBUG
876 	if (ldebug(waitpid))
877 		printf(ARGS(waitpid, "%d, %p, %d"),
878 		    args->pid, (void *)args->status, args->options);
879 #endif
880 	/*
881 	 * this is necessary because the test in kern_wait doesn't work
882 	 * because we mess with the options here
883 	 */
884 	if (args->options & ~(WUNTRACED | WNOHANG | WCONTINUED | __WCLONE))
885 		return (EINVAL);
886 
887 	options = (args->options & (WNOHANG | WUNTRACED));
888 	/* WLINUXCLONE should be equal to __WCLONE, but we make sure */
889 	if (args->options & __WCLONE)
890 		options |= WLINUXCLONE;
891 
892 	return (linux_common_wait(td, args->pid, args->status, options, NULL));
893 }
894 
895 int
896 linux_wait4(struct thread *td, struct linux_wait4_args *args)
897 {
898 	int error, options;
899 	struct rusage ru, *rup;
900 
901 #ifdef DEBUG
902 	if (ldebug(wait4))
903 		printf(ARGS(wait4, "%d, %p, %d, %p"),
904 		    args->pid, (void *)args->status, args->options,
905 		    (void *)args->rusage);
906 #endif
907 
908 	options = (args->options & (WNOHANG | WUNTRACED));
909 	/* WLINUXCLONE should be equal to __WCLONE, but we make sure */
910 	if (args->options & __WCLONE)
911 		options |= WLINUXCLONE;
912 
913 	if (args->rusage != NULL)
914 		rup = &ru;
915 	else
916 		rup = NULL;
917 	error = linux_common_wait(td, args->pid, args->status, options, rup);
918 	if (error != 0)
919 		return (error);
920 	if (args->rusage != NULL)
921 		error = linux_copyout_rusage(&ru, args->rusage);
922 	return (error);
923 }
924 
925 int
926 linux_waitid(struct thread *td, struct linux_waitid_args *args)
927 {
928 	int status, options, sig;
929 	struct __wrusage wru;
930 	siginfo_t siginfo;
931 	l_siginfo_t lsi;
932 	idtype_t idtype;
933 	struct proc *p;
934 	int error;
935 
936 	options = 0;
937 	linux_to_bsd_waitopts(args->options, &options);
938 
939 	if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED))
940 		return (EINVAL);
941 	if (!(options & (WEXITED | WUNTRACED | WCONTINUED)))
942 		return (EINVAL);
943 
944 	switch (args->idtype) {
945 	case LINUX_P_ALL:
946 		idtype = P_ALL;
947 		break;
948 	case LINUX_P_PID:
949 		if (args->id <= 0)
950 			return (EINVAL);
951 		idtype = P_PID;
952 		break;
953 	case LINUX_P_PGID:
954 		if (args->id <= 0)
955 			return (EINVAL);
956 		idtype = P_PGID;
957 		break;
958 	default:
959 		return (EINVAL);
960 	}
961 
962 	error = kern_wait6(td, idtype, args->id, &status, options,
963 	    &wru, &siginfo);
964 	if (error != 0)
965 		return (error);
966 	if (args->rusage != NULL) {
967 		error = linux_copyout_rusage(&wru.wru_children,
968 		    args->rusage);
969 		if (error != 0)
970 			return (error);
971 	}
972 	if (args->info != NULL) {
973 		p = td->td_proc;
974 		if (td->td_retval[0] == 0)
975 			bzero(&lsi, sizeof(lsi));
976 		else {
977 			sig = BSD_TO_LINUX_SIGNAL(siginfo.si_signo);
978 			siginfo_to_lsiginfo(&siginfo, &lsi, sig);
979 		}
980 		error = copyout(&lsi, args->info, sizeof(lsi));
981 	}
982 	td->td_retval[0] = 0;
983 
984 	return (error);
985 }
986 
987 int
988 linux_mknod(struct thread *td, struct linux_mknod_args *args)
989 {
990 	char *path;
991 	int error;
992 
993 	LCONVPATHCREAT(td, args->path, &path);
994 
995 #ifdef DEBUG
996 	if (ldebug(mknod))
997 		printf(ARGS(mknod, "%s, %d, %d"), path, args->mode, args->dev);
998 #endif
999 
1000 	switch (args->mode & S_IFMT) {
1001 	case S_IFIFO:
1002 	case S_IFSOCK:
1003 		error = kern_mkfifoat(td, AT_FDCWD, path, UIO_SYSSPACE,
1004 		    args->mode);
1005 		break;
1006 
1007 	case S_IFCHR:
1008 	case S_IFBLK:
1009 		error = kern_mknodat(td, AT_FDCWD, path, UIO_SYSSPACE,
1010 		    args->mode, args->dev);
1011 		break;
1012 
1013 	case S_IFDIR:
1014 		error = EPERM;
1015 		break;
1016 
1017 	case 0:
1018 		args->mode |= S_IFREG;
1019 		/* FALLTHROUGH */
1020 	case S_IFREG:
1021 		error = kern_openat(td, AT_FDCWD, path, UIO_SYSSPACE,
1022 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1023 		if (error == 0)
1024 			kern_close(td, td->td_retval[0]);
1025 		break;
1026 
1027 	default:
1028 		error = EINVAL;
1029 		break;
1030 	}
1031 	LFREEPATH(path);
1032 	return (error);
1033 }
1034 
1035 int
1036 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
1037 {
1038 	char *path;
1039 	int error, dfd;
1040 
1041 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
1042 	LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
1043 
1044 #ifdef DEBUG
1045 	if (ldebug(mknodat))
1046 		printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev);
1047 #endif
1048 
1049 	switch (args->mode & S_IFMT) {
1050 	case S_IFIFO:
1051 	case S_IFSOCK:
1052 		error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
1053 		break;
1054 
1055 	case S_IFCHR:
1056 	case S_IFBLK:
1057 		error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
1058 		    args->dev);
1059 		break;
1060 
1061 	case S_IFDIR:
1062 		error = EPERM;
1063 		break;
1064 
1065 	case 0:
1066 		args->mode |= S_IFREG;
1067 		/* FALLTHROUGH */
1068 	case S_IFREG:
1069 		error = kern_openat(td, dfd, path, UIO_SYSSPACE,
1070 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1071 		if (error == 0)
1072 			kern_close(td, td->td_retval[0]);
1073 		break;
1074 
1075 	default:
1076 		error = EINVAL;
1077 		break;
1078 	}
1079 	LFREEPATH(path);
1080 	return (error);
1081 }
1082 
1083 /*
1084  * UGH! This is just about the dumbest idea I've ever heard!!
1085  */
1086 int
1087 linux_personality(struct thread *td, struct linux_personality_args *args)
1088 {
1089 #ifdef DEBUG
1090 	if (ldebug(personality))
1091 		printf(ARGS(personality, "%lu"), (unsigned long)args->per);
1092 #endif
1093 	if (args->per != 0)
1094 		return (EINVAL);
1095 
1096 	/* Yes Jim, it's still a Linux... */
1097 	td->td_retval[0] = 0;
1098 	return (0);
1099 }
1100 
1101 struct l_itimerval {
1102 	l_timeval it_interval;
1103 	l_timeval it_value;
1104 };
1105 
1106 #define	B2L_ITIMERVAL(bip, lip) 					\
1107 	(bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec;		\
1108 	(bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec;	\
1109 	(bip)->it_value.tv_sec = (lip)->it_value.tv_sec;		\
1110 	(bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1111 
1112 int
1113 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1114 {
1115 	int error;
1116 	struct l_itimerval ls;
1117 	struct itimerval aitv, oitv;
1118 
1119 #ifdef DEBUG
1120 	if (ldebug(setitimer))
1121 		printf(ARGS(setitimer, "%p, %p"),
1122 		    (void *)uap->itv, (void *)uap->oitv);
1123 #endif
1124 
1125 	if (uap->itv == NULL) {
1126 		uap->itv = uap->oitv;
1127 		return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1128 	}
1129 
1130 	error = copyin(uap->itv, &ls, sizeof(ls));
1131 	if (error != 0)
1132 		return (error);
1133 	B2L_ITIMERVAL(&aitv, &ls);
1134 #ifdef DEBUG
1135 	if (ldebug(setitimer)) {
1136 		printf("setitimer: value: sec: %jd, usec: %ld\n",
1137 		    (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1138 		printf("setitimer: interval: sec: %jd, usec: %ld\n",
1139 		    (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1140 	}
1141 #endif
1142 	error = kern_setitimer(td, uap->which, &aitv, &oitv);
1143 	if (error != 0 || uap->oitv == NULL)
1144 		return (error);
1145 	B2L_ITIMERVAL(&ls, &oitv);
1146 
1147 	return (copyout(&ls, uap->oitv, sizeof(ls)));
1148 }
1149 
1150 int
1151 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1152 {
1153 	int error;
1154 	struct l_itimerval ls;
1155 	struct itimerval aitv;
1156 
1157 #ifdef DEBUG
1158 	if (ldebug(getitimer))
1159 		printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1160 #endif
1161 	error = kern_getitimer(td, uap->which, &aitv);
1162 	if (error != 0)
1163 		return (error);
1164 	B2L_ITIMERVAL(&ls, &aitv);
1165 	return (copyout(&ls, uap->itv, sizeof(ls)));
1166 }
1167 
1168 int
1169 linux_nice(struct thread *td, struct linux_nice_args *args)
1170 {
1171 	struct setpriority_args bsd_args;
1172 
1173 	bsd_args.which = PRIO_PROCESS;
1174 	bsd_args.who = 0;		/* current process */
1175 	bsd_args.prio = args->inc;
1176 	return (sys_setpriority(td, &bsd_args));
1177 }
1178 
1179 int
1180 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1181 {
1182 	struct ucred *newcred, *oldcred;
1183 	l_gid_t *linux_gidset;
1184 	gid_t *bsd_gidset;
1185 	int ngrp, error;
1186 	struct proc *p;
1187 
1188 	ngrp = args->gidsetsize;
1189 	if (ngrp < 0 || ngrp >= ngroups_max + 1)
1190 		return (EINVAL);
1191 	linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_TEMP, M_WAITOK);
1192 	error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1193 	if (error)
1194 		goto out;
1195 	newcred = crget();
1196 	p = td->td_proc;
1197 	PROC_LOCK(p);
1198 	oldcred = crcopysafe(p, newcred);
1199 
1200 	/*
1201 	 * cr_groups[0] holds egid. Setting the whole set from
1202 	 * the supplied set will cause egid to be changed too.
1203 	 * Keep cr_groups[0] unchanged to prevent that.
1204 	 */
1205 
1206 	if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) {
1207 		PROC_UNLOCK(p);
1208 		crfree(newcred);
1209 		goto out;
1210 	}
1211 
1212 	if (ngrp > 0) {
1213 		newcred->cr_ngroups = ngrp + 1;
1214 
1215 		bsd_gidset = newcred->cr_groups;
1216 		ngrp--;
1217 		while (ngrp >= 0) {
1218 			bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1219 			ngrp--;
1220 		}
1221 	} else
1222 		newcred->cr_ngroups = 1;
1223 
1224 	setsugid(p);
1225 	proc_set_cred(p, newcred);
1226 	PROC_UNLOCK(p);
1227 	crfree(oldcred);
1228 	error = 0;
1229 out:
1230 	free(linux_gidset, M_TEMP);
1231 	return (error);
1232 }
1233 
1234 int
1235 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1236 {
1237 	struct ucred *cred;
1238 	l_gid_t *linux_gidset;
1239 	gid_t *bsd_gidset;
1240 	int bsd_gidsetsz, ngrp, error;
1241 
1242 	cred = td->td_ucred;
1243 	bsd_gidset = cred->cr_groups;
1244 	bsd_gidsetsz = cred->cr_ngroups - 1;
1245 
1246 	/*
1247 	 * cr_groups[0] holds egid. Returning the whole set
1248 	 * here will cause a duplicate. Exclude cr_groups[0]
1249 	 * to prevent that.
1250 	 */
1251 
1252 	if ((ngrp = args->gidsetsize) == 0) {
1253 		td->td_retval[0] = bsd_gidsetsz;
1254 		return (0);
1255 	}
1256 
1257 	if (ngrp < bsd_gidsetsz)
1258 		return (EINVAL);
1259 
1260 	ngrp = 0;
1261 	linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1262 	    M_TEMP, M_WAITOK);
1263 	while (ngrp < bsd_gidsetsz) {
1264 		linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1265 		ngrp++;
1266 	}
1267 
1268 	error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1269 	free(linux_gidset, M_TEMP);
1270 	if (error)
1271 		return (error);
1272 
1273 	td->td_retval[0] = ngrp;
1274 	return (0);
1275 }
1276 
1277 int
1278 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1279 {
1280 	struct rlimit bsd_rlim;
1281 	struct l_rlimit rlim;
1282 	u_int which;
1283 	int error;
1284 
1285 #ifdef DEBUG
1286 	if (ldebug(setrlimit))
1287 		printf(ARGS(setrlimit, "%d, %p"),
1288 		    args->resource, (void *)args->rlim);
1289 #endif
1290 
1291 	if (args->resource >= LINUX_RLIM_NLIMITS)
1292 		return (EINVAL);
1293 
1294 	which = linux_to_bsd_resource[args->resource];
1295 	if (which == -1)
1296 		return (EINVAL);
1297 
1298 	error = copyin(args->rlim, &rlim, sizeof(rlim));
1299 	if (error)
1300 		return (error);
1301 
1302 	bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1303 	bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1304 	return (kern_setrlimit(td, which, &bsd_rlim));
1305 }
1306 
1307 int
1308 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1309 {
1310 	struct l_rlimit rlim;
1311 	struct proc *p = td->td_proc;
1312 	struct rlimit bsd_rlim;
1313 	u_int which;
1314 
1315 #ifdef DEBUG
1316 	if (ldebug(old_getrlimit))
1317 		printf(ARGS(old_getrlimit, "%d, %p"),
1318 		    args->resource, (void *)args->rlim);
1319 #endif
1320 
1321 	if (args->resource >= LINUX_RLIM_NLIMITS)
1322 		return (EINVAL);
1323 
1324 	which = linux_to_bsd_resource[args->resource];
1325 	if (which == -1)
1326 		return (EINVAL);
1327 
1328 	PROC_LOCK(p);
1329 	lim_rlimit(p, which, &bsd_rlim);
1330 	PROC_UNLOCK(p);
1331 
1332 #ifdef COMPAT_LINUX32
1333 	rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1334 	if (rlim.rlim_cur == UINT_MAX)
1335 		rlim.rlim_cur = INT_MAX;
1336 	rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1337 	if (rlim.rlim_max == UINT_MAX)
1338 		rlim.rlim_max = INT_MAX;
1339 #else
1340 	rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1341 	if (rlim.rlim_cur == ULONG_MAX)
1342 		rlim.rlim_cur = LONG_MAX;
1343 	rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1344 	if (rlim.rlim_max == ULONG_MAX)
1345 		rlim.rlim_max = LONG_MAX;
1346 #endif
1347 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1348 }
1349 
1350 int
1351 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1352 {
1353 	struct l_rlimit rlim;
1354 	struct proc *p = td->td_proc;
1355 	struct rlimit bsd_rlim;
1356 	u_int which;
1357 
1358 #ifdef DEBUG
1359 	if (ldebug(getrlimit))
1360 		printf(ARGS(getrlimit, "%d, %p"),
1361 		    args->resource, (void *)args->rlim);
1362 #endif
1363 
1364 	if (args->resource >= LINUX_RLIM_NLIMITS)
1365 		return (EINVAL);
1366 
1367 	which = linux_to_bsd_resource[args->resource];
1368 	if (which == -1)
1369 		return (EINVAL);
1370 
1371 	PROC_LOCK(p);
1372 	lim_rlimit(p, which, &bsd_rlim);
1373 	PROC_UNLOCK(p);
1374 
1375 	rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1376 	rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1377 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1378 }
1379 
1380 int
1381 linux_sched_setscheduler(struct thread *td,
1382     struct linux_sched_setscheduler_args *args)
1383 {
1384 	struct sched_param sched_param;
1385 	struct thread *tdt;
1386 	int error, policy;
1387 
1388 #ifdef DEBUG
1389 	if (ldebug(sched_setscheduler))
1390 		printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1391 		    args->pid, args->policy, (const void *)args->param);
1392 #endif
1393 
1394 	switch (args->policy) {
1395 	case LINUX_SCHED_OTHER:
1396 		policy = SCHED_OTHER;
1397 		break;
1398 	case LINUX_SCHED_FIFO:
1399 		policy = SCHED_FIFO;
1400 		break;
1401 	case LINUX_SCHED_RR:
1402 		policy = SCHED_RR;
1403 		break;
1404 	default:
1405 		return (EINVAL);
1406 	}
1407 
1408 	error = copyin(args->param, &sched_param, sizeof(sched_param));
1409 	if (error)
1410 		return (error);
1411 
1412 	tdt = linux_tdfind(td, args->pid, -1);
1413 	if (tdt == NULL)
1414 		return (ESRCH);
1415 
1416 	error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1417 	PROC_UNLOCK(tdt->td_proc);
1418 	return (error);
1419 }
1420 
1421 int
1422 linux_sched_getscheduler(struct thread *td,
1423     struct linux_sched_getscheduler_args *args)
1424 {
1425 	struct thread *tdt;
1426 	int error, policy;
1427 
1428 #ifdef DEBUG
1429 	if (ldebug(sched_getscheduler))
1430 		printf(ARGS(sched_getscheduler, "%d"), args->pid);
1431 #endif
1432 
1433 	tdt = linux_tdfind(td, args->pid, -1);
1434 	if (tdt == NULL)
1435 		return (ESRCH);
1436 
1437 	error = kern_sched_getscheduler(td, tdt, &policy);
1438 	PROC_UNLOCK(tdt->td_proc);
1439 
1440 	switch (policy) {
1441 	case SCHED_OTHER:
1442 		td->td_retval[0] = LINUX_SCHED_OTHER;
1443 		break;
1444 	case SCHED_FIFO:
1445 		td->td_retval[0] = LINUX_SCHED_FIFO;
1446 		break;
1447 	case SCHED_RR:
1448 		td->td_retval[0] = LINUX_SCHED_RR;
1449 		break;
1450 	}
1451 	return (error);
1452 }
1453 
1454 int
1455 linux_sched_get_priority_max(struct thread *td,
1456     struct linux_sched_get_priority_max_args *args)
1457 {
1458 	struct sched_get_priority_max_args bsd;
1459 
1460 #ifdef DEBUG
1461 	if (ldebug(sched_get_priority_max))
1462 		printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1463 #endif
1464 
1465 	switch (args->policy) {
1466 	case LINUX_SCHED_OTHER:
1467 		bsd.policy = SCHED_OTHER;
1468 		break;
1469 	case LINUX_SCHED_FIFO:
1470 		bsd.policy = SCHED_FIFO;
1471 		break;
1472 	case LINUX_SCHED_RR:
1473 		bsd.policy = SCHED_RR;
1474 		break;
1475 	default:
1476 		return (EINVAL);
1477 	}
1478 	return (sys_sched_get_priority_max(td, &bsd));
1479 }
1480 
1481 int
1482 linux_sched_get_priority_min(struct thread *td,
1483     struct linux_sched_get_priority_min_args *args)
1484 {
1485 	struct sched_get_priority_min_args bsd;
1486 
1487 #ifdef DEBUG
1488 	if (ldebug(sched_get_priority_min))
1489 		printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1490 #endif
1491 
1492 	switch (args->policy) {
1493 	case LINUX_SCHED_OTHER:
1494 		bsd.policy = SCHED_OTHER;
1495 		break;
1496 	case LINUX_SCHED_FIFO:
1497 		bsd.policy = SCHED_FIFO;
1498 		break;
1499 	case LINUX_SCHED_RR:
1500 		bsd.policy = SCHED_RR;
1501 		break;
1502 	default:
1503 		return (EINVAL);
1504 	}
1505 	return (sys_sched_get_priority_min(td, &bsd));
1506 }
1507 
1508 #define REBOOT_CAD_ON	0x89abcdef
1509 #define REBOOT_CAD_OFF	0
1510 #define REBOOT_HALT	0xcdef0123
1511 #define REBOOT_RESTART	0x01234567
1512 #define REBOOT_RESTART2	0xA1B2C3D4
1513 #define REBOOT_POWEROFF	0x4321FEDC
1514 #define REBOOT_MAGIC1	0xfee1dead
1515 #define REBOOT_MAGIC2	0x28121969
1516 #define REBOOT_MAGIC2A	0x05121996
1517 #define REBOOT_MAGIC2B	0x16041998
1518 
1519 int
1520 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1521 {
1522 	struct reboot_args bsd_args;
1523 
1524 #ifdef DEBUG
1525 	if (ldebug(reboot))
1526 		printf(ARGS(reboot, "0x%x"), args->cmd);
1527 #endif
1528 
1529 	if (args->magic1 != REBOOT_MAGIC1)
1530 		return (EINVAL);
1531 
1532 	switch (args->magic2) {
1533 	case REBOOT_MAGIC2:
1534 	case REBOOT_MAGIC2A:
1535 	case REBOOT_MAGIC2B:
1536 		break;
1537 	default:
1538 		return (EINVAL);
1539 	}
1540 
1541 	switch (args->cmd) {
1542 	case REBOOT_CAD_ON:
1543 	case REBOOT_CAD_OFF:
1544 		return (priv_check(td, PRIV_REBOOT));
1545 	case REBOOT_HALT:
1546 		bsd_args.opt = RB_HALT;
1547 		break;
1548 	case REBOOT_RESTART:
1549 	case REBOOT_RESTART2:
1550 		bsd_args.opt = 0;
1551 		break;
1552 	case REBOOT_POWEROFF:
1553 		bsd_args.opt = RB_POWEROFF;
1554 		break;
1555 	default:
1556 		return (EINVAL);
1557 	}
1558 	return (sys_reboot(td, &bsd_args));
1559 }
1560 
1561 
1562 /*
1563  * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1564  * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
1565  * are assumed to be preserved. The following lightweight syscalls fixes
1566  * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
1567  *
1568  * linux_getpid() - MP SAFE
1569  * linux_getgid() - MP SAFE
1570  * linux_getuid() - MP SAFE
1571  */
1572 
1573 int
1574 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1575 {
1576 
1577 #ifdef DEBUG
1578 	if (ldebug(getpid))
1579 		printf(ARGS(getpid, ""));
1580 #endif
1581 	td->td_retval[0] = td->td_proc->p_pid;
1582 
1583 	return (0);
1584 }
1585 
1586 int
1587 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1588 {
1589 	struct linux_emuldata *em;
1590 
1591 #ifdef DEBUG
1592 	if (ldebug(gettid))
1593 		printf(ARGS(gettid, ""));
1594 #endif
1595 
1596 	em = em_find(td);
1597 	KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1598 
1599 	td->td_retval[0] = em->em_tid;
1600 
1601 	return (0);
1602 }
1603 
1604 
1605 int
1606 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1607 {
1608 
1609 #ifdef DEBUG
1610 	if (ldebug(getppid))
1611 		printf(ARGS(getppid, ""));
1612 #endif
1613 
1614 	PROC_LOCK(td->td_proc);
1615 	td->td_retval[0] = td->td_proc->p_pptr->p_pid;
1616 	PROC_UNLOCK(td->td_proc);
1617 	return (0);
1618 }
1619 
1620 int
1621 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1622 {
1623 
1624 #ifdef DEBUG
1625 	if (ldebug(getgid))
1626 		printf(ARGS(getgid, ""));
1627 #endif
1628 
1629 	td->td_retval[0] = td->td_ucred->cr_rgid;
1630 	return (0);
1631 }
1632 
1633 int
1634 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1635 {
1636 
1637 #ifdef DEBUG
1638 	if (ldebug(getuid))
1639 		printf(ARGS(getuid, ""));
1640 #endif
1641 
1642 	td->td_retval[0] = td->td_ucred->cr_ruid;
1643 	return (0);
1644 }
1645 
1646 
1647 int
1648 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1649 {
1650 	struct getsid_args bsd;
1651 
1652 #ifdef DEBUG
1653 	if (ldebug(getsid))
1654 		printf(ARGS(getsid, "%i"), args->pid);
1655 #endif
1656 
1657 	bsd.pid = args->pid;
1658 	return (sys_getsid(td, &bsd));
1659 }
1660 
1661 int
1662 linux_nosys(struct thread *td, struct nosys_args *ignore)
1663 {
1664 
1665 	return (ENOSYS);
1666 }
1667 
1668 int
1669 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1670 {
1671 	struct getpriority_args bsd_args;
1672 	int error;
1673 
1674 #ifdef DEBUG
1675 	if (ldebug(getpriority))
1676 		printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
1677 #endif
1678 
1679 	bsd_args.which = args->which;
1680 	bsd_args.who = args->who;
1681 	error = sys_getpriority(td, &bsd_args);
1682 	td->td_retval[0] = 20 - td->td_retval[0];
1683 	return (error);
1684 }
1685 
1686 int
1687 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1688 {
1689 	int name[2];
1690 
1691 #ifdef DEBUG
1692 	if (ldebug(sethostname))
1693 		printf(ARGS(sethostname, "*, %i"), args->len);
1694 #endif
1695 
1696 	name[0] = CTL_KERN;
1697 	name[1] = KERN_HOSTNAME;
1698 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1699 	    args->len, 0, 0));
1700 }
1701 
1702 int
1703 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1704 {
1705 	int name[2];
1706 
1707 #ifdef DEBUG
1708 	if (ldebug(setdomainname))
1709 		printf(ARGS(setdomainname, "*, %i"), args->len);
1710 #endif
1711 
1712 	name[0] = CTL_KERN;
1713 	name[1] = KERN_NISDOMAINNAME;
1714 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1715 	    args->len, 0, 0));
1716 }
1717 
1718 int
1719 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1720 {
1721 
1722 #ifdef DEBUG
1723 	if (ldebug(exit_group))
1724 		printf(ARGS(exit_group, "%i"), args->error_code);
1725 #endif
1726 
1727 	LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1728 	    args->error_code);
1729 
1730 	/*
1731 	 * XXX: we should send a signal to the parent if
1732 	 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1733 	 * as it doesnt occur often.
1734 	 */
1735 	exit1(td, W_EXITCODE(args->error_code, 0));
1736 		/* NOTREACHED */
1737 }
1738 
1739 #define _LINUX_CAPABILITY_VERSION  0x19980330
1740 
1741 struct l_user_cap_header {
1742 	l_int	version;
1743 	l_int	pid;
1744 };
1745 
1746 struct l_user_cap_data {
1747 	l_int	effective;
1748 	l_int	permitted;
1749 	l_int	inheritable;
1750 };
1751 
1752 int
1753 linux_capget(struct thread *td, struct linux_capget_args *args)
1754 {
1755 	struct l_user_cap_header luch;
1756 	struct l_user_cap_data lucd;
1757 	int error;
1758 
1759 	if (args->hdrp == NULL)
1760 		return (EFAULT);
1761 
1762 	error = copyin(args->hdrp, &luch, sizeof(luch));
1763 	if (error != 0)
1764 		return (error);
1765 
1766 	if (luch.version != _LINUX_CAPABILITY_VERSION) {
1767 		luch.version = _LINUX_CAPABILITY_VERSION;
1768 		error = copyout(&luch, args->hdrp, sizeof(luch));
1769 		if (error)
1770 			return (error);
1771 		return (EINVAL);
1772 	}
1773 
1774 	if (luch.pid)
1775 		return (EPERM);
1776 
1777 	if (args->datap) {
1778 		/*
1779 		 * The current implementation doesn't support setting
1780 		 * a capability (it's essentially a stub) so indicate
1781 		 * that no capabilities are currently set or available
1782 		 * to request.
1783 		 */
1784 		bzero (&lucd, sizeof(lucd));
1785 		error = copyout(&lucd, args->datap, sizeof(lucd));
1786 	}
1787 
1788 	return (error);
1789 }
1790 
1791 int
1792 linux_capset(struct thread *td, struct linux_capset_args *args)
1793 {
1794 	struct l_user_cap_header luch;
1795 	struct l_user_cap_data lucd;
1796 	int error;
1797 
1798 	if (args->hdrp == NULL || args->datap == NULL)
1799 		return (EFAULT);
1800 
1801 	error = copyin(args->hdrp, &luch, sizeof(luch));
1802 	if (error != 0)
1803 		return (error);
1804 
1805 	if (luch.version != _LINUX_CAPABILITY_VERSION) {
1806 		luch.version = _LINUX_CAPABILITY_VERSION;
1807 		error = copyout(&luch, args->hdrp, sizeof(luch));
1808 		if (error)
1809 			return (error);
1810 		return (EINVAL);
1811 	}
1812 
1813 	if (luch.pid)
1814 		return (EPERM);
1815 
1816 	error = copyin(args->datap, &lucd, sizeof(lucd));
1817 	if (error != 0)
1818 		return (error);
1819 
1820 	/* We currently don't support setting any capabilities. */
1821 	if (lucd.effective || lucd.permitted || lucd.inheritable) {
1822 		linux_msg(td,
1823 			  "capset effective=0x%x, permitted=0x%x, "
1824 			  "inheritable=0x%x is not implemented",
1825 			  (int)lucd.effective, (int)lucd.permitted,
1826 			  (int)lucd.inheritable);
1827 		return (EPERM);
1828 	}
1829 
1830 	return (0);
1831 }
1832 
1833 int
1834 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1835 {
1836 	int error = 0, max_size;
1837 	struct proc *p = td->td_proc;
1838 	char comm[LINUX_MAX_COMM_LEN];
1839 	struct linux_emuldata *em;
1840 	int pdeath_signal;
1841 
1842 #ifdef DEBUG
1843 	if (ldebug(prctl))
1844 		printf(ARGS(prctl, "%d, %d, %d, %d, %d"), args->option,
1845 		    args->arg2, args->arg3, args->arg4, args->arg5);
1846 #endif
1847 
1848 	switch (args->option) {
1849 	case LINUX_PR_SET_PDEATHSIG:
1850 		if (!LINUX_SIG_VALID(args->arg2))
1851 			return (EINVAL);
1852 		em = em_find(td);
1853 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1854 		em->pdeath_signal = args->arg2;
1855 		break;
1856 	case LINUX_PR_GET_PDEATHSIG:
1857 		em = em_find(td);
1858 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1859 		pdeath_signal = em->pdeath_signal;
1860 		error = copyout(&pdeath_signal,
1861 		    (void *)(register_t)args->arg2,
1862 		    sizeof(pdeath_signal));
1863 		break;
1864 	case LINUX_PR_GET_KEEPCAPS:
1865 		/*
1866 		 * Indicate that we always clear the effective and
1867 		 * permitted capability sets when the user id becomes
1868 		 * non-zero (actually the capability sets are simply
1869 		 * always zero in the current implementation).
1870 		 */
1871 		td->td_retval[0] = 0;
1872 		break;
1873 	case LINUX_PR_SET_KEEPCAPS:
1874 		/*
1875 		 * Ignore requests to keep the effective and permitted
1876 		 * capability sets when the user id becomes non-zero.
1877 		 */
1878 		break;
1879 	case LINUX_PR_SET_NAME:
1880 		/*
1881 		 * To be on the safe side we need to make sure to not
1882 		 * overflow the size a linux program expects. We already
1883 		 * do this here in the copyin, so that we don't need to
1884 		 * check on copyout.
1885 		 */
1886 		max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1887 		error = copyinstr((void *)(register_t)args->arg2, comm,
1888 		    max_size, NULL);
1889 
1890 		/* Linux silently truncates the name if it is too long. */
1891 		if (error == ENAMETOOLONG) {
1892 			/*
1893 			 * XXX: copyinstr() isn't documented to populate the
1894 			 * array completely, so do a copyin() to be on the
1895 			 * safe side. This should be changed in case
1896 			 * copyinstr() is changed to guarantee this.
1897 			 */
1898 			error = copyin((void *)(register_t)args->arg2, comm,
1899 			    max_size - 1);
1900 			comm[max_size - 1] = '\0';
1901 		}
1902 		if (error)
1903 			return (error);
1904 
1905 		PROC_LOCK(p);
1906 		strlcpy(p->p_comm, comm, sizeof(p->p_comm));
1907 		PROC_UNLOCK(p);
1908 		break;
1909 	case LINUX_PR_GET_NAME:
1910 		PROC_LOCK(p);
1911 		strlcpy(comm, p->p_comm, sizeof(comm));
1912 		PROC_UNLOCK(p);
1913 		error = copyout(comm, (void *)(register_t)args->arg2,
1914 		    strlen(comm) + 1);
1915 		break;
1916 	default:
1917 		error = EINVAL;
1918 		break;
1919 	}
1920 
1921 	return (error);
1922 }
1923 
1924 int
1925 linux_sched_setparam(struct thread *td,
1926     struct linux_sched_setparam_args *uap)
1927 {
1928 	struct sched_param sched_param;
1929 	struct thread *tdt;
1930 	int error;
1931 
1932 #ifdef DEBUG
1933 	if (ldebug(sched_setparam))
1934 		printf(ARGS(sched_setparam, "%d, *"), uap->pid);
1935 #endif
1936 
1937 	error = copyin(uap->param, &sched_param, sizeof(sched_param));
1938 	if (error)
1939 		return (error);
1940 
1941 	tdt = linux_tdfind(td, uap->pid, -1);
1942 	if (tdt == NULL)
1943 		return (ESRCH);
1944 
1945 	error = kern_sched_setparam(td, tdt, &sched_param);
1946 	PROC_UNLOCK(tdt->td_proc);
1947 	return (error);
1948 }
1949 
1950 int
1951 linux_sched_getparam(struct thread *td,
1952     struct linux_sched_getparam_args *uap)
1953 {
1954 	struct sched_param sched_param;
1955 	struct thread *tdt;
1956 	int error;
1957 
1958 #ifdef DEBUG
1959 	if (ldebug(sched_getparam))
1960 		printf(ARGS(sched_getparam, "%d, *"), uap->pid);
1961 #endif
1962 
1963 	tdt = linux_tdfind(td, uap->pid, -1);
1964 	if (tdt == NULL)
1965 		return (ESRCH);
1966 
1967 	error = kern_sched_getparam(td, tdt, &sched_param);
1968 	PROC_UNLOCK(tdt->td_proc);
1969 	if (error == 0)
1970 		error = copyout(&sched_param, uap->param,
1971 		    sizeof(sched_param));
1972 	return (error);
1973 }
1974 
1975 /*
1976  * Get affinity of a process.
1977  */
1978 int
1979 linux_sched_getaffinity(struct thread *td,
1980     struct linux_sched_getaffinity_args *args)
1981 {
1982 	int error;
1983 	struct thread *tdt;
1984 	struct cpuset_getaffinity_args cga;
1985 
1986 #ifdef DEBUG
1987 	if (ldebug(sched_getaffinity))
1988 		printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid,
1989 		    args->len);
1990 #endif
1991 	if (args->len < sizeof(cpuset_t))
1992 		return (EINVAL);
1993 
1994 	tdt = linux_tdfind(td, args->pid, -1);
1995 	if (tdt == NULL)
1996 		return (ESRCH);
1997 
1998 	PROC_UNLOCK(tdt->td_proc);
1999 	cga.level = CPU_LEVEL_WHICH;
2000 	cga.which = CPU_WHICH_TID;
2001 	cga.id = tdt->td_tid;
2002 	cga.cpusetsize = sizeof(cpuset_t);
2003 	cga.mask = (cpuset_t *) args->user_mask_ptr;
2004 
2005 	if ((error = sys_cpuset_getaffinity(td, &cga)) == 0)
2006 		td->td_retval[0] = sizeof(cpuset_t);
2007 
2008 	return (error);
2009 }
2010 
2011 /*
2012  *  Set affinity of a process.
2013  */
2014 int
2015 linux_sched_setaffinity(struct thread *td,
2016     struct linux_sched_setaffinity_args *args)
2017 {
2018 	struct cpuset_setaffinity_args csa;
2019 	struct thread *tdt;
2020 
2021 #ifdef DEBUG
2022 	if (ldebug(sched_setaffinity))
2023 		printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid,
2024 		    args->len);
2025 #endif
2026 	if (args->len < sizeof(cpuset_t))
2027 		return (EINVAL);
2028 
2029 	tdt = linux_tdfind(td, args->pid, -1);
2030 	if (tdt == NULL)
2031 		return (ESRCH);
2032 
2033 	PROC_UNLOCK(tdt->td_proc);
2034 	csa.level = CPU_LEVEL_WHICH;
2035 	csa.which = CPU_WHICH_TID;
2036 	csa.id = tdt->td_tid;
2037 	csa.cpusetsize = sizeof(cpuset_t);
2038 	csa.mask = (cpuset_t *) args->user_mask_ptr;
2039 
2040 	return (sys_cpuset_setaffinity(td, &csa));
2041 }
2042 
2043 struct linux_rlimit64 {
2044 	uint64_t	rlim_cur;
2045 	uint64_t	rlim_max;
2046 };
2047 
2048 int
2049 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
2050 {
2051 	struct rlimit rlim, nrlim;
2052 	struct linux_rlimit64 lrlim;
2053 	struct proc *p;
2054 	u_int which;
2055 	int flags;
2056 	int error;
2057 
2058 #ifdef DEBUG
2059 	if (ldebug(prlimit64))
2060 		printf(ARGS(prlimit64, "%d, %d, %p, %p"), args->pid,
2061 		    args->resource, (void *)args->new, (void *)args->old);
2062 #endif
2063 
2064 	if (args->resource >= LINUX_RLIM_NLIMITS)
2065 		return (EINVAL);
2066 
2067 	which = linux_to_bsd_resource[args->resource];
2068 	if (which == -1)
2069 		return (EINVAL);
2070 
2071 	if (args->new != NULL) {
2072 		/*
2073 		 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
2074 		 * rlim is unsigned 64-bit. FreeBSD treats negative limits
2075 		 * as INFINITY so we do not need a conversion even.
2076 		 */
2077 		error = copyin(args->new, &nrlim, sizeof(nrlim));
2078 		if (error != 0)
2079 			return (error);
2080 	}
2081 
2082 	flags = PGET_HOLD | PGET_NOTWEXIT;
2083 	if (args->new != NULL)
2084 		flags |= PGET_CANDEBUG;
2085 	else
2086 		flags |= PGET_CANSEE;
2087 	error = pget(args->pid, flags, &p);
2088 	if (error != 0)
2089 		return (error);
2090 
2091 	if (args->old != NULL) {
2092 		PROC_LOCK(p);
2093 		lim_rlimit(p, which, &rlim);
2094 		PROC_UNLOCK(p);
2095 		if (rlim.rlim_cur == RLIM_INFINITY)
2096 			lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2097 		else
2098 			lrlim.rlim_cur = rlim.rlim_cur;
2099 		if (rlim.rlim_max == RLIM_INFINITY)
2100 			lrlim.rlim_max = LINUX_RLIM_INFINITY;
2101 		else
2102 			lrlim.rlim_max = rlim.rlim_max;
2103 		error = copyout(&lrlim, args->old, sizeof(lrlim));
2104 		if (error != 0)
2105 			goto out;
2106 	}
2107 
2108 	if (args->new != NULL)
2109 		error = kern_proc_setrlimit(td, p, which, &nrlim);
2110 
2111  out:
2112 	PRELE(p);
2113 	return (error);
2114 }
2115 
2116 int
2117 linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2118 {
2119 	struct timeval utv, tv0, tv1, *tvp;
2120 	struct l_pselect6arg lpse6;
2121 	struct l_timespec lts;
2122 	struct timespec uts;
2123 	l_sigset_t l_ss;
2124 	sigset_t *ssp;
2125 	sigset_t ss;
2126 	int error;
2127 
2128 	ssp = NULL;
2129 	if (args->sig != NULL) {
2130 		error = copyin(args->sig, &lpse6, sizeof(lpse6));
2131 		if (error != 0)
2132 			return (error);
2133 		if (lpse6.ss_len != sizeof(l_ss))
2134 			return (EINVAL);
2135 		if (lpse6.ss != 0) {
2136 			error = copyin(PTRIN(lpse6.ss), &l_ss,
2137 			    sizeof(l_ss));
2138 			if (error != 0)
2139 				return (error);
2140 			linux_to_bsd_sigset(&l_ss, &ss);
2141 			ssp = &ss;
2142 		}
2143 	}
2144 
2145 	/*
2146 	 * Currently glibc changes nanosecond number to microsecond.
2147 	 * This mean losing precision but for now it is hardly seen.
2148 	 */
2149 	if (args->tsp != NULL) {
2150 		error = copyin(args->tsp, &lts, sizeof(lts));
2151 		if (error != 0)
2152 			return (error);
2153 		uts.tv_sec = lts.tv_sec;
2154 		uts.tv_nsec = lts.tv_nsec;
2155 
2156 		TIMESPEC_TO_TIMEVAL(&utv, &uts);
2157 		if (itimerfix(&utv))
2158 			return (EINVAL);
2159 
2160 		microtime(&tv0);
2161 		tvp = &utv;
2162 	} else
2163 		tvp = NULL;
2164 
2165 	error = kern_pselect(td, args->nfds, args->readfds, args->writefds,
2166 	    args->exceptfds, tvp, ssp, sizeof(l_int) * 8);
2167 
2168 	if (error == 0 && args->tsp != NULL) {
2169 		if (td->td_retval[0] != 0) {
2170 			/*
2171 			 * Compute how much time was left of the timeout,
2172 			 * by subtracting the current time and the time
2173 			 * before we started the call, and subtracting
2174 			 * that result from the user-supplied value.
2175 			 */
2176 
2177 			microtime(&tv1);
2178 			timevalsub(&tv1, &tv0);
2179 			timevalsub(&utv, &tv1);
2180 			if (utv.tv_sec < 0)
2181 				timevalclear(&utv);
2182 		} else
2183 			timevalclear(&utv);
2184 
2185 		TIMEVAL_TO_TIMESPEC(&utv, &uts);
2186 		lts.tv_sec = uts.tv_sec;
2187 		lts.tv_nsec = uts.tv_nsec;
2188 		error = copyout(&lts, args->tsp, sizeof(lts));
2189 	}
2190 
2191 	return (error);
2192 }
2193 
2194 int
2195 linux_sched_rr_get_interval(struct thread *td,
2196     struct linux_sched_rr_get_interval_args *uap)
2197 {
2198 	struct timespec ts;
2199 	struct l_timespec lts;
2200 	struct thread *tdt;
2201 	int error;
2202 
2203 	/*
2204 	 * According to man in case the invalid pid specified
2205 	 * EINVAL should be returned.
2206 	 */
2207 	if (uap->pid < 0)
2208 		return (EINVAL);
2209 
2210 	tdt = linux_tdfind(td, uap->pid, -1);
2211 	if (tdt == NULL)
2212 		return (ESRCH);
2213 
2214 	error = kern_sched_rr_get_interval_td(td, tdt, &ts);
2215 	PROC_UNLOCK(tdt->td_proc);
2216 	if (error != 0)
2217 		return (error);
2218 	lts.tv_sec = ts.tv_sec;
2219 	lts.tv_nsec = ts.tv_nsec;
2220 	return (copyout(&lts, uap->interval, sizeof(lts)));
2221 }
2222 
2223 /*
2224  * In case when the Linux thread is the initial thread in
2225  * the thread group thread id is equal to the process id.
2226  * Glibc depends on this magic (assert in pthread_getattr_np.c).
2227  */
2228 struct thread *
2229 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid)
2230 {
2231 	struct linux_emuldata *em;
2232 	struct thread *tdt;
2233 	struct proc *p;
2234 
2235 	tdt = NULL;
2236 	if (tid == 0 || tid == td->td_tid) {
2237 		tdt = td;
2238 		PROC_LOCK(tdt->td_proc);
2239 	} else if (tid > PID_MAX)
2240 		tdt = tdfind(tid, pid);
2241 	else {
2242 		/*
2243 		 * Initial thread where the tid equal to the pid.
2244 		 */
2245 		p = pfind(tid);
2246 		if (p != NULL) {
2247 			if (SV_PROC_ABI(p) != SV_ABI_LINUX) {
2248 				/*
2249 				 * p is not a Linuxulator process.
2250 				 */
2251 				PROC_UNLOCK(p);
2252 				return (NULL);
2253 			}
2254 			FOREACH_THREAD_IN_PROC(p, tdt) {
2255 				em = em_find(tdt);
2256 				if (tid == em->em_tid)
2257 					return (tdt);
2258 			}
2259 			PROC_UNLOCK(p);
2260 		}
2261 		return (NULL);
2262 	}
2263 
2264 	return (tdt);
2265 }
2266 
2267 void
2268 linux_to_bsd_waitopts(int options, int *bsdopts)
2269 {
2270 
2271 	if (options & LINUX_WNOHANG)
2272 		*bsdopts |= WNOHANG;
2273 	if (options & LINUX_WUNTRACED)
2274 		*bsdopts |= WUNTRACED;
2275 	if (options & LINUX_WEXITED)
2276 		*bsdopts |= WEXITED;
2277 	if (options & LINUX_WCONTINUED)
2278 		*bsdopts |= WCONTINUED;
2279 	if (options & LINUX_WNOWAIT)
2280 		*bsdopts |= WNOWAIT;
2281 
2282 	if (options & __WCLONE)
2283 		*bsdopts |= WLINUXCLONE;
2284 }
2285