xref: /freebsd/sys/compat/linux/linux_misc.c (revision 193d9e768ba63fcfb187cfd17f461f7d41345048)
1 /*-
2  * Copyright (c) 2002 Doug Rabson
3  * Copyright (c) 1994-1995 Søren Schmidt
4  * All rights reserved.
5  *
6  * Redistribution and use in source and binary forms, with or without
7  * modification, are permitted provided that the following conditions
8  * are met:
9  * 1. Redistributions of source code must retain the above copyright
10  *    notice, this list of conditions and the following disclaimer
11  *    in this position and unchanged.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. The name of the author may not be used to endorse or promote products
16  *    derived from this software without specific prior written permission
17  *
18  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
19  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
20  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
21  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
22  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
23  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
24  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
25  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
26  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
27  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
28  */
29 
30 #include <sys/cdefs.h>
31 __FBSDID("$FreeBSD$");
32 
33 #include "opt_compat.h"
34 
35 #include <sys/param.h>
36 #include <sys/blist.h>
37 #include <sys/fcntl.h>
38 #if defined(__i386__)
39 #include <sys/imgact_aout.h>
40 #endif
41 #include <sys/jail.h>
42 #include <sys/kernel.h>
43 #include <sys/limits.h>
44 #include <sys/lock.h>
45 #include <sys/malloc.h>
46 #include <sys/mman.h>
47 #include <sys/mount.h>
48 #include <sys/mutex.h>
49 #include <sys/namei.h>
50 #include <sys/priv.h>
51 #include <sys/proc.h>
52 #include <sys/reboot.h>
53 #include <sys/racct.h>
54 #include <sys/resourcevar.h>
55 #include <sys/sched.h>
56 #include <sys/sdt.h>
57 #include <sys/signalvar.h>
58 #include <sys/stat.h>
59 #include <sys/syscallsubr.h>
60 #include <sys/sysctl.h>
61 #include <sys/sysproto.h>
62 #include <sys/systm.h>
63 #include <sys/time.h>
64 #include <sys/vmmeter.h>
65 #include <sys/vnode.h>
66 #include <sys/wait.h>
67 #include <sys/cpuset.h>
68 
69 #include <security/mac/mac_framework.h>
70 
71 #include <vm/vm.h>
72 #include <vm/pmap.h>
73 #include <vm/vm_kern.h>
74 #include <vm/vm_map.h>
75 #include <vm/vm_extern.h>
76 #include <vm/vm_object.h>
77 #include <vm/swap_pager.h>
78 
79 #ifdef COMPAT_LINUX32
80 #include <machine/../linux32/linux.h>
81 #include <machine/../linux32/linux32_proto.h>
82 #else
83 #include <machine/../linux/linux.h>
84 #include <machine/../linux/linux_proto.h>
85 #endif
86 
87 #include <compat/linux/linux_dtrace.h>
88 #include <compat/linux/linux_file.h>
89 #include <compat/linux/linux_mib.h>
90 #include <compat/linux/linux_signal.h>
91 #include <compat/linux/linux_timer.h>
92 #include <compat/linux/linux_util.h>
93 #include <compat/linux/linux_sysproto.h>
94 #include <compat/linux/linux_emul.h>
95 #include <compat/linux/linux_misc.h>
96 
97 /**
98  * Special DTrace provider for the linuxulator.
99  *
100  * In this file we define the provider for the entire linuxulator. All
101  * modules (= files of the linuxulator) use it.
102  *
103  * We define a different name depending on the emulated bitsize, see
104  * ../../<ARCH>/linux{,32}/linux.h, e.g.:
105  *      native bitsize          = linuxulator
106  *      amd64, 32bit emulation  = linuxulator32
107  */
108 LIN_SDT_PROVIDER_DEFINE(LINUX_DTRACE);
109 
110 int stclohz;				/* Statistics clock frequency */
111 
112 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
113 	RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
114 	RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
115 	RLIMIT_MEMLOCK, RLIMIT_AS
116 };
117 
118 struct l_sysinfo {
119 	l_long		uptime;		/* Seconds since boot */
120 	l_ulong		loads[3];	/* 1, 5, and 15 minute load averages */
121 #define LINUX_SYSINFO_LOADS_SCALE 65536
122 	l_ulong		totalram;	/* Total usable main memory size */
123 	l_ulong		freeram;	/* Available memory size */
124 	l_ulong		sharedram;	/* Amount of shared memory */
125 	l_ulong		bufferram;	/* Memory used by buffers */
126 	l_ulong		totalswap;	/* Total swap space size */
127 	l_ulong		freeswap;	/* swap space still available */
128 	l_ushort	procs;		/* Number of current processes */
129 	l_ushort	pads;
130 	l_ulong		totalbig;
131 	l_ulong		freebig;
132 	l_uint		mem_unit;
133 	char		_f[20-2*sizeof(l_long)-sizeof(l_int)];	/* padding */
134 };
135 
136 struct l_pselect6arg {
137 	l_uintptr_t	ss;
138 	l_size_t	ss_len;
139 };
140 
141 static int	linux_utimensat_nsec_valid(l_long);
142 
143 
144 int
145 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
146 {
147 	struct l_sysinfo sysinfo;
148 	vm_object_t object;
149 	int i, j;
150 	struct timespec ts;
151 
152 	bzero(&sysinfo, sizeof(sysinfo));
153 	getnanouptime(&ts);
154 	if (ts.tv_nsec != 0)
155 		ts.tv_sec++;
156 	sysinfo.uptime = ts.tv_sec;
157 
158 	/* Use the information from the mib to get our load averages */
159 	for (i = 0; i < 3; i++)
160 		sysinfo.loads[i] = averunnable.ldavg[i] *
161 		    LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
162 
163 	sysinfo.totalram = physmem * PAGE_SIZE;
164 	sysinfo.freeram = sysinfo.totalram - vm_cnt.v_wire_count * PAGE_SIZE;
165 
166 	sysinfo.sharedram = 0;
167 	mtx_lock(&vm_object_list_mtx);
168 	TAILQ_FOREACH(object, &vm_object_list, object_list)
169 		if (object->shadow_count > 1)
170 			sysinfo.sharedram += object->resident_page_count;
171 	mtx_unlock(&vm_object_list_mtx);
172 
173 	sysinfo.sharedram *= PAGE_SIZE;
174 	sysinfo.bufferram = 0;
175 
176 	swap_pager_status(&i, &j);
177 	sysinfo.totalswap = i * PAGE_SIZE;
178 	sysinfo.freeswap = (i - j) * PAGE_SIZE;
179 
180 	sysinfo.procs = nprocs;
181 
182 	/* The following are only present in newer Linux kernels. */
183 	sysinfo.totalbig = 0;
184 	sysinfo.freebig = 0;
185 	sysinfo.mem_unit = 1;
186 
187 	return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
188 }
189 
190 int
191 linux_alarm(struct thread *td, struct linux_alarm_args *args)
192 {
193 	struct itimerval it, old_it;
194 	u_int secs;
195 	int error;
196 
197 #ifdef DEBUG
198 	if (ldebug(alarm))
199 		printf(ARGS(alarm, "%u"), args->secs);
200 #endif
201 	secs = args->secs;
202 	/*
203 	 * Linux alarm() is always successful. Limit secs to INT32_MAX / 2
204 	 * to match kern_setitimer()'s limit to avoid error from it.
205 	 *
206 	 * XXX. Linux limit secs to INT_MAX on 32 and does not limit on 64-bit
207 	 * platforms.
208 	 */
209 	if (secs > INT32_MAX / 2)
210 		secs = INT32_MAX / 2;
211 
212 	it.it_value.tv_sec = secs;
213 	it.it_value.tv_usec = 0;
214 	timevalclear(&it.it_interval);
215 	error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
216 	KASSERT(error == 0, ("kern_setitimer returns %d", error));
217 
218 	if ((old_it.it_value.tv_sec == 0 && old_it.it_value.tv_usec > 0) ||
219 	    old_it.it_value.tv_usec >= 500000)
220 		old_it.it_value.tv_sec++;
221 	td->td_retval[0] = old_it.it_value.tv_sec;
222 	return (0);
223 }
224 
225 int
226 linux_brk(struct thread *td, struct linux_brk_args *args)
227 {
228 	struct vmspace *vm = td->td_proc->p_vmspace;
229 	vm_offset_t new, old;
230 	struct obreak_args /* {
231 		char * nsize;
232 	} */ tmp;
233 
234 #ifdef DEBUG
235 	if (ldebug(brk))
236 		printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
237 #endif
238 	old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize);
239 	new = (vm_offset_t)args->dsend;
240 	tmp.nsize = (char *)new;
241 	if (((caddr_t)new > vm->vm_daddr) && !sys_obreak(td, &tmp))
242 		td->td_retval[0] = (long)new;
243 	else
244 		td->td_retval[0] = (long)old;
245 
246 	return (0);
247 }
248 
249 #if defined(__i386__)
250 /* XXX: what about amd64/linux32? */
251 
252 int
253 linux_uselib(struct thread *td, struct linux_uselib_args *args)
254 {
255 	struct nameidata ni;
256 	struct vnode *vp;
257 	struct exec *a_out;
258 	struct vattr attr;
259 	vm_offset_t vmaddr;
260 	unsigned long file_offset;
261 	unsigned long bss_size;
262 	char *library;
263 	ssize_t aresid;
264 	int error, locked, writecount;
265 
266 	LCONVPATHEXIST(td, args->library, &library);
267 
268 #ifdef DEBUG
269 	if (ldebug(uselib))
270 		printf(ARGS(uselib, "%s"), library);
271 #endif
272 
273 	a_out = NULL;
274 	locked = 0;
275 	vp = NULL;
276 
277 	NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
278 	    UIO_SYSSPACE, library, td);
279 	error = namei(&ni);
280 	LFREEPATH(library);
281 	if (error)
282 		goto cleanup;
283 
284 	vp = ni.ni_vp;
285 	NDFREE(&ni, NDF_ONLY_PNBUF);
286 
287 	/*
288 	 * From here on down, we have a locked vnode that must be unlocked.
289 	 * XXX: The code below largely duplicates exec_check_permissions().
290 	 */
291 	locked = 1;
292 
293 	/* Writable? */
294 	error = VOP_GET_WRITECOUNT(vp, &writecount);
295 	if (error != 0)
296 		goto cleanup;
297 	if (writecount != 0) {
298 		error = ETXTBSY;
299 		goto cleanup;
300 	}
301 
302 	/* Executable? */
303 	error = VOP_GETATTR(vp, &attr, td->td_ucred);
304 	if (error)
305 		goto cleanup;
306 
307 	if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
308 	    ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
309 		/* EACCESS is what exec(2) returns. */
310 		error = ENOEXEC;
311 		goto cleanup;
312 	}
313 
314 	/* Sensible size? */
315 	if (attr.va_size == 0) {
316 		error = ENOEXEC;
317 		goto cleanup;
318 	}
319 
320 	/* Can we access it? */
321 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
322 	if (error)
323 		goto cleanup;
324 
325 	/*
326 	 * XXX: This should use vn_open() so that it is properly authorized,
327 	 * and to reduce code redundancy all over the place here.
328 	 * XXX: Not really, it duplicates far more of exec_check_permissions()
329 	 * than vn_open().
330 	 */
331 #ifdef MAC
332 	error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
333 	if (error)
334 		goto cleanup;
335 #endif
336 	error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
337 	if (error)
338 		goto cleanup;
339 
340 	/* Pull in executable header into exec_map */
341 	error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
342 	    VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
343 	if (error)
344 		goto cleanup;
345 
346 	/* Is it a Linux binary ? */
347 	if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
348 		error = ENOEXEC;
349 		goto cleanup;
350 	}
351 
352 	/*
353 	 * While we are here, we should REALLY do some more checks
354 	 */
355 
356 	/* Set file/virtual offset based on a.out variant. */
357 	switch ((int)(a_out->a_magic & 0xffff)) {
358 	case 0413:			/* ZMAGIC */
359 		file_offset = 1024;
360 		break;
361 	case 0314:			/* QMAGIC */
362 		file_offset = 0;
363 		break;
364 	default:
365 		error = ENOEXEC;
366 		goto cleanup;
367 	}
368 
369 	bss_size = round_page(a_out->a_bss);
370 
371 	/* Check various fields in header for validity/bounds. */
372 	if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
373 		error = ENOEXEC;
374 		goto cleanup;
375 	}
376 
377 	/* text + data can't exceed file size */
378 	if (a_out->a_data + a_out->a_text > attr.va_size) {
379 		error = EFAULT;
380 		goto cleanup;
381 	}
382 
383 	/*
384 	 * text/data/bss must not exceed limits
385 	 * XXX - this is not complete. it should check current usage PLUS
386 	 * the resources needed by this library.
387 	 */
388 	PROC_LOCK(td->td_proc);
389 	if (a_out->a_text > maxtsiz ||
390 	    a_out->a_data + bss_size > lim_cur_proc(td->td_proc, RLIMIT_DATA) ||
391 	    racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
392 	    bss_size) != 0) {
393 		PROC_UNLOCK(td->td_proc);
394 		error = ENOMEM;
395 		goto cleanup;
396 	}
397 	PROC_UNLOCK(td->td_proc);
398 
399 	/*
400 	 * Prevent more writers.
401 	 * XXX: Note that if any of the VM operations fail below we don't
402 	 * clear this flag.
403 	 */
404 	VOP_SET_TEXT(vp);
405 
406 	/*
407 	 * Lock no longer needed
408 	 */
409 	locked = 0;
410 	VOP_UNLOCK(vp, 0);
411 
412 	/*
413 	 * Check if file_offset page aligned. Currently we cannot handle
414 	 * misalinged file offsets, and so we read in the entire image
415 	 * (what a waste).
416 	 */
417 	if (file_offset & PAGE_MASK) {
418 #ifdef DEBUG
419 		printf("uselib: Non page aligned binary %lu\n", file_offset);
420 #endif
421 		/* Map text+data read/write/execute */
422 
423 		/* a_entry is the load address and is page aligned */
424 		vmaddr = trunc_page(a_out->a_entry);
425 
426 		/* get anon user mapping, read+write+execute */
427 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
428 		    &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
429 		    VM_PROT_ALL, VM_PROT_ALL, 0);
430 		if (error)
431 			goto cleanup;
432 
433 		error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
434 		    a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
435 		    td->td_ucred, NOCRED, &aresid, td);
436 		if (error != 0)
437 			goto cleanup;
438 		if (aresid != 0) {
439 			error = ENOEXEC;
440 			goto cleanup;
441 		}
442 	} else {
443 #ifdef DEBUG
444 		printf("uselib: Page aligned binary %lu\n", file_offset);
445 #endif
446 		/*
447 		 * for QMAGIC, a_entry is 20 bytes beyond the load address
448 		 * to skip the executable header
449 		 */
450 		vmaddr = trunc_page(a_out->a_entry);
451 
452 		/*
453 		 * Map it all into the process's space as a single
454 		 * copy-on-write "data" segment.
455 		 */
456 		error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr,
457 		    a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
458 		    MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
459 		if (error)
460 			goto cleanup;
461 	}
462 #ifdef DEBUG
463 	printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
464 	    ((long *)vmaddr)[1]);
465 #endif
466 	if (bss_size != 0) {
467 		/* Calculate BSS start address */
468 		vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
469 		    a_out->a_data;
470 
471 		/* allocate some 'anon' space */
472 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
473 		    &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
474 		    VM_PROT_ALL, 0);
475 		if (error)
476 			goto cleanup;
477 	}
478 
479 cleanup:
480 	/* Unlock vnode if needed */
481 	if (locked)
482 		VOP_UNLOCK(vp, 0);
483 
484 	/* Release the temporary mapping. */
485 	if (a_out)
486 		kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
487 
488 	return (error);
489 }
490 
491 #endif	/* __i386__ */
492 
493 int
494 linux_select(struct thread *td, struct linux_select_args *args)
495 {
496 	l_timeval ltv;
497 	struct timeval tv0, tv1, utv, *tvp;
498 	int error;
499 
500 #ifdef DEBUG
501 	if (ldebug(select))
502 		printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
503 		    (void *)args->readfds, (void *)args->writefds,
504 		    (void *)args->exceptfds, (void *)args->timeout);
505 #endif
506 
507 	/*
508 	 * Store current time for computation of the amount of
509 	 * time left.
510 	 */
511 	if (args->timeout) {
512 		if ((error = copyin(args->timeout, &ltv, sizeof(ltv))))
513 			goto select_out;
514 		utv.tv_sec = ltv.tv_sec;
515 		utv.tv_usec = ltv.tv_usec;
516 #ifdef DEBUG
517 		if (ldebug(select))
518 			printf(LMSG("incoming timeout (%jd/%ld)"),
519 			    (intmax_t)utv.tv_sec, utv.tv_usec);
520 #endif
521 
522 		if (itimerfix(&utv)) {
523 			/*
524 			 * The timeval was invalid.  Convert it to something
525 			 * valid that will act as it does under Linux.
526 			 */
527 			utv.tv_sec += utv.tv_usec / 1000000;
528 			utv.tv_usec %= 1000000;
529 			if (utv.tv_usec < 0) {
530 				utv.tv_sec -= 1;
531 				utv.tv_usec += 1000000;
532 			}
533 			if (utv.tv_sec < 0)
534 				timevalclear(&utv);
535 		}
536 		microtime(&tv0);
537 		tvp = &utv;
538 	} else
539 		tvp = NULL;
540 
541 	error = kern_select(td, args->nfds, args->readfds, args->writefds,
542 	    args->exceptfds, tvp, LINUX_NFDBITS);
543 
544 #ifdef DEBUG
545 	if (ldebug(select))
546 		printf(LMSG("real select returns %d"), error);
547 #endif
548 	if (error)
549 		goto select_out;
550 
551 	if (args->timeout) {
552 		if (td->td_retval[0]) {
553 			/*
554 			 * Compute how much time was left of the timeout,
555 			 * by subtracting the current time and the time
556 			 * before we started the call, and subtracting
557 			 * that result from the user-supplied value.
558 			 */
559 			microtime(&tv1);
560 			timevalsub(&tv1, &tv0);
561 			timevalsub(&utv, &tv1);
562 			if (utv.tv_sec < 0)
563 				timevalclear(&utv);
564 		} else
565 			timevalclear(&utv);
566 #ifdef DEBUG
567 		if (ldebug(select))
568 			printf(LMSG("outgoing timeout (%jd/%ld)"),
569 			    (intmax_t)utv.tv_sec, utv.tv_usec);
570 #endif
571 		ltv.tv_sec = utv.tv_sec;
572 		ltv.tv_usec = utv.tv_usec;
573 		if ((error = copyout(&ltv, args->timeout, sizeof(ltv))))
574 			goto select_out;
575 	}
576 
577 select_out:
578 #ifdef DEBUG
579 	if (ldebug(select))
580 		printf(LMSG("select_out -> %d"), error);
581 #endif
582 	return (error);
583 }
584 
585 int
586 linux_mremap(struct thread *td, struct linux_mremap_args *args)
587 {
588 	uintptr_t addr;
589 	size_t len;
590 	int error = 0;
591 
592 #ifdef DEBUG
593 	if (ldebug(mremap))
594 		printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
595 		    (void *)(uintptr_t)args->addr,
596 		    (unsigned long)args->old_len,
597 		    (unsigned long)args->new_len,
598 		    (unsigned long)args->flags);
599 #endif
600 
601 	if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
602 		td->td_retval[0] = 0;
603 		return (EINVAL);
604 	}
605 
606 	/*
607 	 * Check for the page alignment.
608 	 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
609 	 */
610 	if (args->addr & PAGE_MASK) {
611 		td->td_retval[0] = 0;
612 		return (EINVAL);
613 	}
614 
615 	args->new_len = round_page(args->new_len);
616 	args->old_len = round_page(args->old_len);
617 
618 	if (args->new_len > args->old_len) {
619 		td->td_retval[0] = 0;
620 		return (ENOMEM);
621 	}
622 
623 	if (args->new_len < args->old_len) {
624 		addr = args->addr + args->new_len;
625 		len = args->old_len - args->new_len;
626 		error = kern_munmap(td, addr, len);
627 	}
628 
629 	td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
630 	return (error);
631 }
632 
633 #define LINUX_MS_ASYNC       0x0001
634 #define LINUX_MS_INVALIDATE  0x0002
635 #define LINUX_MS_SYNC        0x0004
636 
637 int
638 linux_msync(struct thread *td, struct linux_msync_args *args)
639 {
640 
641 	return (kern_msync(td, args->addr, args->len,
642 	    args->fl & ~LINUX_MS_SYNC));
643 }
644 
645 int
646 linux_time(struct thread *td, struct linux_time_args *args)
647 {
648 	struct timeval tv;
649 	l_time_t tm;
650 	int error;
651 
652 #ifdef DEBUG
653 	if (ldebug(time))
654 		printf(ARGS(time, "*"));
655 #endif
656 
657 	microtime(&tv);
658 	tm = tv.tv_sec;
659 	if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
660 		return (error);
661 	td->td_retval[0] = tm;
662 	return (0);
663 }
664 
665 struct l_times_argv {
666 	l_clock_t	tms_utime;
667 	l_clock_t	tms_stime;
668 	l_clock_t	tms_cutime;
669 	l_clock_t	tms_cstime;
670 };
671 
672 
673 /*
674  * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
675  * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
676  * auxiliary vector entry.
677  */
678 #define	CLK_TCK		100
679 
680 #define	CONVOTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
681 #define	CONVNTCK(r)	(r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
682 
683 #define	CONVTCK(r)	(linux_kernver(td) >= LINUX_KERNVER_2004000 ?		\
684 			    CONVNTCK(r) : CONVOTCK(r))
685 
686 int
687 linux_times(struct thread *td, struct linux_times_args *args)
688 {
689 	struct timeval tv, utime, stime, cutime, cstime;
690 	struct l_times_argv tms;
691 	struct proc *p;
692 	int error;
693 
694 #ifdef DEBUG
695 	if (ldebug(times))
696 		printf(ARGS(times, "*"));
697 #endif
698 
699 	if (args->buf != NULL) {
700 		p = td->td_proc;
701 		PROC_LOCK(p);
702 		PROC_STATLOCK(p);
703 		calcru(p, &utime, &stime);
704 		PROC_STATUNLOCK(p);
705 		calccru(p, &cutime, &cstime);
706 		PROC_UNLOCK(p);
707 
708 		tms.tms_utime = CONVTCK(utime);
709 		tms.tms_stime = CONVTCK(stime);
710 
711 		tms.tms_cutime = CONVTCK(cutime);
712 		tms.tms_cstime = CONVTCK(cstime);
713 
714 		if ((error = copyout(&tms, args->buf, sizeof(tms))))
715 			return (error);
716 	}
717 
718 	microuptime(&tv);
719 	td->td_retval[0] = (int)CONVTCK(tv);
720 	return (0);
721 }
722 
723 int
724 linux_newuname(struct thread *td, struct linux_newuname_args *args)
725 {
726 	struct l_new_utsname utsname;
727 	char osname[LINUX_MAX_UTSNAME];
728 	char osrelease[LINUX_MAX_UTSNAME];
729 	char *p;
730 
731 #ifdef DEBUG
732 	if (ldebug(newuname))
733 		printf(ARGS(newuname, "*"));
734 #endif
735 
736 	linux_get_osname(td, osname);
737 	linux_get_osrelease(td, osrelease);
738 
739 	bzero(&utsname, sizeof(utsname));
740 	strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
741 	getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
742 	getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
743 	strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
744 	strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
745 	for (p = utsname.version; *p != '\0'; ++p)
746 		if (*p == '\n') {
747 			*p = '\0';
748 			break;
749 		}
750 	strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME);
751 
752 	return (copyout(&utsname, args->buf, sizeof(utsname)));
753 }
754 
755 struct l_utimbuf {
756 	l_time_t l_actime;
757 	l_time_t l_modtime;
758 };
759 
760 int
761 linux_utime(struct thread *td, struct linux_utime_args *args)
762 {
763 	struct timeval tv[2], *tvp;
764 	struct l_utimbuf lut;
765 	char *fname;
766 	int error;
767 
768 	LCONVPATHEXIST(td, args->fname, &fname);
769 
770 #ifdef DEBUG
771 	if (ldebug(utime))
772 		printf(ARGS(utime, "%s, *"), fname);
773 #endif
774 
775 	if (args->times) {
776 		if ((error = copyin(args->times, &lut, sizeof lut))) {
777 			LFREEPATH(fname);
778 			return (error);
779 		}
780 		tv[0].tv_sec = lut.l_actime;
781 		tv[0].tv_usec = 0;
782 		tv[1].tv_sec = lut.l_modtime;
783 		tv[1].tv_usec = 0;
784 		tvp = tv;
785 	} else
786 		tvp = NULL;
787 
788 	error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, tvp,
789 	    UIO_SYSSPACE);
790 	LFREEPATH(fname);
791 	return (error);
792 }
793 
794 int
795 linux_utimes(struct thread *td, struct linux_utimes_args *args)
796 {
797 	l_timeval ltv[2];
798 	struct timeval tv[2], *tvp = NULL;
799 	char *fname;
800 	int error;
801 
802 	LCONVPATHEXIST(td, args->fname, &fname);
803 
804 #ifdef DEBUG
805 	if (ldebug(utimes))
806 		printf(ARGS(utimes, "%s, *"), fname);
807 #endif
808 
809 	if (args->tptr != NULL) {
810 		if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
811 			LFREEPATH(fname);
812 			return (error);
813 		}
814 		tv[0].tv_sec = ltv[0].tv_sec;
815 		tv[0].tv_usec = ltv[0].tv_usec;
816 		tv[1].tv_sec = ltv[1].tv_sec;
817 		tv[1].tv_usec = ltv[1].tv_usec;
818 		tvp = tv;
819 	}
820 
821 	error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE,
822 	    tvp, UIO_SYSSPACE);
823 	LFREEPATH(fname);
824 	return (error);
825 }
826 
827 static int
828 linux_utimensat_nsec_valid(l_long nsec)
829 {
830 
831 	if (nsec == LINUX_UTIME_OMIT || nsec == LINUX_UTIME_NOW)
832 		return (0);
833 	if (nsec >= 0 && nsec <= 999999999)
834 		return (0);
835 	return (1);
836 }
837 
838 int
839 linux_utimensat(struct thread *td, struct linux_utimensat_args *args)
840 {
841 	struct l_timespec l_times[2];
842 	struct timespec times[2], *timesp = NULL;
843 	char *path = NULL;
844 	int error, dfd, flags = 0;
845 
846 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
847 
848 #ifdef DEBUG
849 	if (ldebug(utimensat))
850 		printf(ARGS(utimensat, "%d, *"), dfd);
851 #endif
852 
853 	if (args->flags & ~LINUX_AT_SYMLINK_NOFOLLOW)
854 		return (EINVAL);
855 
856 	if (args->times != NULL) {
857 		error = copyin(args->times, l_times, sizeof(l_times));
858 		if (error != 0)
859 			return (error);
860 
861 		if (linux_utimensat_nsec_valid(l_times[0].tv_nsec) != 0 ||
862 		    linux_utimensat_nsec_valid(l_times[1].tv_nsec) != 0)
863 			return (EINVAL);
864 
865 		times[0].tv_sec = l_times[0].tv_sec;
866 		switch (l_times[0].tv_nsec)
867 		{
868 		case LINUX_UTIME_OMIT:
869 			times[0].tv_nsec = UTIME_OMIT;
870 			break;
871 		case LINUX_UTIME_NOW:
872 			times[0].tv_nsec = UTIME_NOW;
873 			break;
874 		default:
875 			times[0].tv_nsec = l_times[0].tv_nsec;
876 		}
877 
878 		times[1].tv_sec = l_times[1].tv_sec;
879 		switch (l_times[1].tv_nsec)
880 		{
881 		case LINUX_UTIME_OMIT:
882 			times[1].tv_nsec = UTIME_OMIT;
883 			break;
884 		case LINUX_UTIME_NOW:
885 			times[1].tv_nsec = UTIME_NOW;
886 			break;
887 		default:
888 			times[1].tv_nsec = l_times[1].tv_nsec;
889 			break;
890 		}
891 		timesp = times;
892 
893 		/* This breaks POSIX, but is what the Linux kernel does
894 		 * _on purpose_ (documented in the man page for utimensat(2)),
895 		 * so we must follow that behaviour. */
896 		if (times[0].tv_nsec == UTIME_OMIT &&
897 		    times[1].tv_nsec == UTIME_OMIT)
898 			return (0);
899 	}
900 
901 	if (args->pathname != NULL)
902 		LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
903 	else if (args->flags != 0)
904 		return (EINVAL);
905 
906 	if (args->flags & LINUX_AT_SYMLINK_NOFOLLOW)
907 		flags |= AT_SYMLINK_NOFOLLOW;
908 
909 	if (path == NULL)
910 		error = kern_futimens(td, dfd, timesp, UIO_SYSSPACE);
911 	else {
912 		error = kern_utimensat(td, dfd, path, UIO_SYSSPACE, timesp,
913 	    		UIO_SYSSPACE, flags);
914 		LFREEPATH(path);
915 	}
916 
917 	return (error);
918 }
919 
920 int
921 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
922 {
923 	l_timeval ltv[2];
924 	struct timeval tv[2], *tvp = NULL;
925 	char *fname;
926 	int error, dfd;
927 
928 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
929 	LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
930 
931 #ifdef DEBUG
932 	if (ldebug(futimesat))
933 		printf(ARGS(futimesat, "%s, *"), fname);
934 #endif
935 
936 	if (args->utimes != NULL) {
937 		if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
938 			LFREEPATH(fname);
939 			return (error);
940 		}
941 		tv[0].tv_sec = ltv[0].tv_sec;
942 		tv[0].tv_usec = ltv[0].tv_usec;
943 		tv[1].tv_sec = ltv[1].tv_sec;
944 		tv[1].tv_usec = ltv[1].tv_usec;
945 		tvp = tv;
946 	}
947 
948 	error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
949 	LFREEPATH(fname);
950 	return (error);
951 }
952 
953 int
954 linux_common_wait(struct thread *td, int pid, int *status,
955     int options, struct rusage *ru)
956 {
957 	int error, tmpstat;
958 
959 	error = kern_wait(td, pid, &tmpstat, options, ru);
960 	if (error)
961 		return (error);
962 
963 	if (status) {
964 		tmpstat &= 0xffff;
965 		if (WIFSIGNALED(tmpstat))
966 			tmpstat = (tmpstat & 0xffffff80) |
967 			    bsd_to_linux_signal(WTERMSIG(tmpstat));
968 		else if (WIFSTOPPED(tmpstat))
969 			tmpstat = (tmpstat & 0xffff00ff) |
970 			    (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8);
971 		else if (WIFCONTINUED(tmpstat))
972 			tmpstat = 0xffff;
973 		error = copyout(&tmpstat, status, sizeof(int));
974 	}
975 
976 	return (error);
977 }
978 
979 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
980 int
981 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
982 {
983 	struct linux_wait4_args wait4_args;
984 
985 #ifdef DEBUG
986 	if (ldebug(waitpid))
987 		printf(ARGS(waitpid, "%d, %p, %d"),
988 		    args->pid, (void *)args->status, args->options);
989 #endif
990 
991 	wait4_args.pid = args->pid;
992 	wait4_args.status = args->status;
993 	wait4_args.options = args->options;
994 	wait4_args.rusage = NULL;
995 
996 	return (linux_wait4(td, &wait4_args));
997 }
998 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
999 
1000 int
1001 linux_wait4(struct thread *td, struct linux_wait4_args *args)
1002 {
1003 	int error, options;
1004 	struct rusage ru, *rup;
1005 
1006 #ifdef DEBUG
1007 	if (ldebug(wait4))
1008 		printf(ARGS(wait4, "%d, %p, %d, %p"),
1009 		    args->pid, (void *)args->status, args->options,
1010 		    (void *)args->rusage);
1011 #endif
1012 	if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG |
1013 	    LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
1014 		return (EINVAL);
1015 
1016 	options = WEXITED;
1017 	linux_to_bsd_waitopts(args->options, &options);
1018 
1019 	if (args->rusage != NULL)
1020 		rup = &ru;
1021 	else
1022 		rup = NULL;
1023 	error = linux_common_wait(td, args->pid, args->status, options, rup);
1024 	if (error != 0)
1025 		return (error);
1026 	if (args->rusage != NULL)
1027 		error = linux_copyout_rusage(&ru, args->rusage);
1028 	return (error);
1029 }
1030 
1031 int
1032 linux_waitid(struct thread *td, struct linux_waitid_args *args)
1033 {
1034 	int status, options, sig;
1035 	struct __wrusage wru;
1036 	siginfo_t siginfo;
1037 	l_siginfo_t lsi;
1038 	idtype_t idtype;
1039 	struct proc *p;
1040 	int error;
1041 
1042 	options = 0;
1043 	linux_to_bsd_waitopts(args->options, &options);
1044 
1045 	if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED))
1046 		return (EINVAL);
1047 	if (!(options & (WEXITED | WUNTRACED | WCONTINUED)))
1048 		return (EINVAL);
1049 
1050 	switch (args->idtype) {
1051 	case LINUX_P_ALL:
1052 		idtype = P_ALL;
1053 		break;
1054 	case LINUX_P_PID:
1055 		if (args->id <= 0)
1056 			return (EINVAL);
1057 		idtype = P_PID;
1058 		break;
1059 	case LINUX_P_PGID:
1060 		if (args->id <= 0)
1061 			return (EINVAL);
1062 		idtype = P_PGID;
1063 		break;
1064 	default:
1065 		return (EINVAL);
1066 	}
1067 
1068 	error = kern_wait6(td, idtype, args->id, &status, options,
1069 	    &wru, &siginfo);
1070 	if (error != 0)
1071 		return (error);
1072 	if (args->rusage != NULL) {
1073 		error = linux_copyout_rusage(&wru.wru_children,
1074 		    args->rusage);
1075 		if (error != 0)
1076 			return (error);
1077 	}
1078 	if (args->info != NULL) {
1079 		p = td->td_proc;
1080 		if (td->td_retval[0] == 0)
1081 			bzero(&lsi, sizeof(lsi));
1082 		else {
1083 			sig = bsd_to_linux_signal(siginfo.si_signo);
1084 			siginfo_to_lsiginfo(&siginfo, &lsi, sig);
1085 		}
1086 		error = copyout(&lsi, args->info, sizeof(lsi));
1087 	}
1088 	td->td_retval[0] = 0;
1089 
1090 	return (error);
1091 }
1092 
1093 int
1094 linux_mknod(struct thread *td, struct linux_mknod_args *args)
1095 {
1096 	char *path;
1097 	int error;
1098 
1099 	LCONVPATHCREAT(td, args->path, &path);
1100 
1101 #ifdef DEBUG
1102 	if (ldebug(mknod))
1103 		printf(ARGS(mknod, "%s, %d, %ju"), path, args->mode,
1104 		    (uintmax_t)args->dev);
1105 #endif
1106 
1107 	switch (args->mode & S_IFMT) {
1108 	case S_IFIFO:
1109 	case S_IFSOCK:
1110 		error = kern_mkfifoat(td, AT_FDCWD, path, UIO_SYSSPACE,
1111 		    args->mode);
1112 		break;
1113 
1114 	case S_IFCHR:
1115 	case S_IFBLK:
1116 		error = kern_mknodat(td, AT_FDCWD, path, UIO_SYSSPACE,
1117 		    args->mode, args->dev);
1118 		break;
1119 
1120 	case S_IFDIR:
1121 		error = EPERM;
1122 		break;
1123 
1124 	case 0:
1125 		args->mode |= S_IFREG;
1126 		/* FALLTHROUGH */
1127 	case S_IFREG:
1128 		error = kern_openat(td, AT_FDCWD, path, UIO_SYSSPACE,
1129 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1130 		if (error == 0)
1131 			kern_close(td, td->td_retval[0]);
1132 		break;
1133 
1134 	default:
1135 		error = EINVAL;
1136 		break;
1137 	}
1138 	LFREEPATH(path);
1139 	return (error);
1140 }
1141 
1142 int
1143 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
1144 {
1145 	char *path;
1146 	int error, dfd;
1147 
1148 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
1149 	LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
1150 
1151 #ifdef DEBUG
1152 	if (ldebug(mknodat))
1153 		printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev);
1154 #endif
1155 
1156 	switch (args->mode & S_IFMT) {
1157 	case S_IFIFO:
1158 	case S_IFSOCK:
1159 		error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
1160 		break;
1161 
1162 	case S_IFCHR:
1163 	case S_IFBLK:
1164 		error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
1165 		    args->dev);
1166 		break;
1167 
1168 	case S_IFDIR:
1169 		error = EPERM;
1170 		break;
1171 
1172 	case 0:
1173 		args->mode |= S_IFREG;
1174 		/* FALLTHROUGH */
1175 	case S_IFREG:
1176 		error = kern_openat(td, dfd, path, UIO_SYSSPACE,
1177 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1178 		if (error == 0)
1179 			kern_close(td, td->td_retval[0]);
1180 		break;
1181 
1182 	default:
1183 		error = EINVAL;
1184 		break;
1185 	}
1186 	LFREEPATH(path);
1187 	return (error);
1188 }
1189 
1190 /*
1191  * UGH! This is just about the dumbest idea I've ever heard!!
1192  */
1193 int
1194 linux_personality(struct thread *td, struct linux_personality_args *args)
1195 {
1196 	struct linux_pemuldata *pem;
1197 	struct proc *p = td->td_proc;
1198 	uint32_t old;
1199 
1200 #ifdef DEBUG
1201 	if (ldebug(personality))
1202 		printf(ARGS(personality, "%u"), args->per);
1203 #endif
1204 
1205 	PROC_LOCK(p);
1206 	pem = pem_find(p);
1207 	old = pem->persona;
1208 	if (args->per != 0xffffffff)
1209 		pem->persona = args->per;
1210 	PROC_UNLOCK(p);
1211 
1212 	td->td_retval[0] = old;
1213 	return (0);
1214 }
1215 
1216 struct l_itimerval {
1217 	l_timeval it_interval;
1218 	l_timeval it_value;
1219 };
1220 
1221 #define	B2L_ITIMERVAL(bip, lip) 					\
1222 	(bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec;		\
1223 	(bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec;	\
1224 	(bip)->it_value.tv_sec = (lip)->it_value.tv_sec;		\
1225 	(bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1226 
1227 int
1228 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1229 {
1230 	int error;
1231 	struct l_itimerval ls;
1232 	struct itimerval aitv, oitv;
1233 
1234 #ifdef DEBUG
1235 	if (ldebug(setitimer))
1236 		printf(ARGS(setitimer, "%p, %p"),
1237 		    (void *)uap->itv, (void *)uap->oitv);
1238 #endif
1239 
1240 	if (uap->itv == NULL) {
1241 		uap->itv = uap->oitv;
1242 		return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1243 	}
1244 
1245 	error = copyin(uap->itv, &ls, sizeof(ls));
1246 	if (error != 0)
1247 		return (error);
1248 	B2L_ITIMERVAL(&aitv, &ls);
1249 #ifdef DEBUG
1250 	if (ldebug(setitimer)) {
1251 		printf("setitimer: value: sec: %jd, usec: %ld\n",
1252 		    (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1253 		printf("setitimer: interval: sec: %jd, usec: %ld\n",
1254 		    (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1255 	}
1256 #endif
1257 	error = kern_setitimer(td, uap->which, &aitv, &oitv);
1258 	if (error != 0 || uap->oitv == NULL)
1259 		return (error);
1260 	B2L_ITIMERVAL(&ls, &oitv);
1261 
1262 	return (copyout(&ls, uap->oitv, sizeof(ls)));
1263 }
1264 
1265 int
1266 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1267 {
1268 	int error;
1269 	struct l_itimerval ls;
1270 	struct itimerval aitv;
1271 
1272 #ifdef DEBUG
1273 	if (ldebug(getitimer))
1274 		printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1275 #endif
1276 	error = kern_getitimer(td, uap->which, &aitv);
1277 	if (error != 0)
1278 		return (error);
1279 	B2L_ITIMERVAL(&ls, &aitv);
1280 	return (copyout(&ls, uap->itv, sizeof(ls)));
1281 }
1282 
1283 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1284 int
1285 linux_nice(struct thread *td, struct linux_nice_args *args)
1286 {
1287 	struct setpriority_args bsd_args;
1288 
1289 	bsd_args.which = PRIO_PROCESS;
1290 	bsd_args.who = 0;		/* current process */
1291 	bsd_args.prio = args->inc;
1292 	return (sys_setpriority(td, &bsd_args));
1293 }
1294 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1295 
1296 int
1297 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1298 {
1299 	struct ucred *newcred, *oldcred;
1300 	l_gid_t *linux_gidset;
1301 	gid_t *bsd_gidset;
1302 	int ngrp, error;
1303 	struct proc *p;
1304 
1305 	ngrp = args->gidsetsize;
1306 	if (ngrp < 0 || ngrp >= ngroups_max + 1)
1307 		return (EINVAL);
1308 	linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK);
1309 	error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1310 	if (error)
1311 		goto out;
1312 	newcred = crget();
1313 	crextend(newcred, ngrp + 1);
1314 	p = td->td_proc;
1315 	PROC_LOCK(p);
1316 	oldcred = p->p_ucred;
1317 	crcopy(newcred, oldcred);
1318 
1319 	/*
1320 	 * cr_groups[0] holds egid. Setting the whole set from
1321 	 * the supplied set will cause egid to be changed too.
1322 	 * Keep cr_groups[0] unchanged to prevent that.
1323 	 */
1324 
1325 	if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) {
1326 		PROC_UNLOCK(p);
1327 		crfree(newcred);
1328 		goto out;
1329 	}
1330 
1331 	if (ngrp > 0) {
1332 		newcred->cr_ngroups = ngrp + 1;
1333 
1334 		bsd_gidset = newcred->cr_groups;
1335 		ngrp--;
1336 		while (ngrp >= 0) {
1337 			bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1338 			ngrp--;
1339 		}
1340 	} else
1341 		newcred->cr_ngroups = 1;
1342 
1343 	setsugid(p);
1344 	proc_set_cred(p, newcred);
1345 	PROC_UNLOCK(p);
1346 	crfree(oldcred);
1347 	error = 0;
1348 out:
1349 	free(linux_gidset, M_LINUX);
1350 	return (error);
1351 }
1352 
1353 int
1354 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1355 {
1356 	struct ucred *cred;
1357 	l_gid_t *linux_gidset;
1358 	gid_t *bsd_gidset;
1359 	int bsd_gidsetsz, ngrp, error;
1360 
1361 	cred = td->td_ucred;
1362 	bsd_gidset = cred->cr_groups;
1363 	bsd_gidsetsz = cred->cr_ngroups - 1;
1364 
1365 	/*
1366 	 * cr_groups[0] holds egid. Returning the whole set
1367 	 * here will cause a duplicate. Exclude cr_groups[0]
1368 	 * to prevent that.
1369 	 */
1370 
1371 	if ((ngrp = args->gidsetsize) == 0) {
1372 		td->td_retval[0] = bsd_gidsetsz;
1373 		return (0);
1374 	}
1375 
1376 	if (ngrp < bsd_gidsetsz)
1377 		return (EINVAL);
1378 
1379 	ngrp = 0;
1380 	linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1381 	    M_LINUX, M_WAITOK);
1382 	while (ngrp < bsd_gidsetsz) {
1383 		linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1384 		ngrp++;
1385 	}
1386 
1387 	error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1388 	free(linux_gidset, M_LINUX);
1389 	if (error)
1390 		return (error);
1391 
1392 	td->td_retval[0] = ngrp;
1393 	return (0);
1394 }
1395 
1396 int
1397 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1398 {
1399 	struct rlimit bsd_rlim;
1400 	struct l_rlimit rlim;
1401 	u_int which;
1402 	int error;
1403 
1404 #ifdef DEBUG
1405 	if (ldebug(setrlimit))
1406 		printf(ARGS(setrlimit, "%d, %p"),
1407 		    args->resource, (void *)args->rlim);
1408 #endif
1409 
1410 	if (args->resource >= LINUX_RLIM_NLIMITS)
1411 		return (EINVAL);
1412 
1413 	which = linux_to_bsd_resource[args->resource];
1414 	if (which == -1)
1415 		return (EINVAL);
1416 
1417 	error = copyin(args->rlim, &rlim, sizeof(rlim));
1418 	if (error)
1419 		return (error);
1420 
1421 	bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1422 	bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1423 	return (kern_setrlimit(td, which, &bsd_rlim));
1424 }
1425 
1426 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1427 int
1428 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1429 {
1430 	struct l_rlimit rlim;
1431 	struct rlimit bsd_rlim;
1432 	u_int which;
1433 
1434 #ifdef DEBUG
1435 	if (ldebug(old_getrlimit))
1436 		printf(ARGS(old_getrlimit, "%d, %p"),
1437 		    args->resource, (void *)args->rlim);
1438 #endif
1439 
1440 	if (args->resource >= LINUX_RLIM_NLIMITS)
1441 		return (EINVAL);
1442 
1443 	which = linux_to_bsd_resource[args->resource];
1444 	if (which == -1)
1445 		return (EINVAL);
1446 
1447 	lim_rlimit(td, which, &bsd_rlim);
1448 
1449 #ifdef COMPAT_LINUX32
1450 	rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1451 	if (rlim.rlim_cur == UINT_MAX)
1452 		rlim.rlim_cur = INT_MAX;
1453 	rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1454 	if (rlim.rlim_max == UINT_MAX)
1455 		rlim.rlim_max = INT_MAX;
1456 #else
1457 	rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1458 	if (rlim.rlim_cur == ULONG_MAX)
1459 		rlim.rlim_cur = LONG_MAX;
1460 	rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1461 	if (rlim.rlim_max == ULONG_MAX)
1462 		rlim.rlim_max = LONG_MAX;
1463 #endif
1464 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1465 }
1466 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1467 
1468 int
1469 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1470 {
1471 	struct l_rlimit rlim;
1472 	struct rlimit bsd_rlim;
1473 	u_int which;
1474 
1475 #ifdef DEBUG
1476 	if (ldebug(getrlimit))
1477 		printf(ARGS(getrlimit, "%d, %p"),
1478 		    args->resource, (void *)args->rlim);
1479 #endif
1480 
1481 	if (args->resource >= LINUX_RLIM_NLIMITS)
1482 		return (EINVAL);
1483 
1484 	which = linux_to_bsd_resource[args->resource];
1485 	if (which == -1)
1486 		return (EINVAL);
1487 
1488 	lim_rlimit(td, which, &bsd_rlim);
1489 
1490 	rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1491 	rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1492 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1493 }
1494 
1495 int
1496 linux_sched_setscheduler(struct thread *td,
1497     struct linux_sched_setscheduler_args *args)
1498 {
1499 	struct sched_param sched_param;
1500 	struct thread *tdt;
1501 	int error, policy;
1502 
1503 #ifdef DEBUG
1504 	if (ldebug(sched_setscheduler))
1505 		printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1506 		    args->pid, args->policy, (const void *)args->param);
1507 #endif
1508 
1509 	switch (args->policy) {
1510 	case LINUX_SCHED_OTHER:
1511 		policy = SCHED_OTHER;
1512 		break;
1513 	case LINUX_SCHED_FIFO:
1514 		policy = SCHED_FIFO;
1515 		break;
1516 	case LINUX_SCHED_RR:
1517 		policy = SCHED_RR;
1518 		break;
1519 	default:
1520 		return (EINVAL);
1521 	}
1522 
1523 	error = copyin(args->param, &sched_param, sizeof(sched_param));
1524 	if (error)
1525 		return (error);
1526 
1527 	tdt = linux_tdfind(td, args->pid, -1);
1528 	if (tdt == NULL)
1529 		return (ESRCH);
1530 
1531 	error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1532 	PROC_UNLOCK(tdt->td_proc);
1533 	return (error);
1534 }
1535 
1536 int
1537 linux_sched_getscheduler(struct thread *td,
1538     struct linux_sched_getscheduler_args *args)
1539 {
1540 	struct thread *tdt;
1541 	int error, policy;
1542 
1543 #ifdef DEBUG
1544 	if (ldebug(sched_getscheduler))
1545 		printf(ARGS(sched_getscheduler, "%d"), args->pid);
1546 #endif
1547 
1548 	tdt = linux_tdfind(td, args->pid, -1);
1549 	if (tdt == NULL)
1550 		return (ESRCH);
1551 
1552 	error = kern_sched_getscheduler(td, tdt, &policy);
1553 	PROC_UNLOCK(tdt->td_proc);
1554 
1555 	switch (policy) {
1556 	case SCHED_OTHER:
1557 		td->td_retval[0] = LINUX_SCHED_OTHER;
1558 		break;
1559 	case SCHED_FIFO:
1560 		td->td_retval[0] = LINUX_SCHED_FIFO;
1561 		break;
1562 	case SCHED_RR:
1563 		td->td_retval[0] = LINUX_SCHED_RR;
1564 		break;
1565 	}
1566 	return (error);
1567 }
1568 
1569 int
1570 linux_sched_get_priority_max(struct thread *td,
1571     struct linux_sched_get_priority_max_args *args)
1572 {
1573 	struct sched_get_priority_max_args bsd;
1574 
1575 #ifdef DEBUG
1576 	if (ldebug(sched_get_priority_max))
1577 		printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1578 #endif
1579 
1580 	switch (args->policy) {
1581 	case LINUX_SCHED_OTHER:
1582 		bsd.policy = SCHED_OTHER;
1583 		break;
1584 	case LINUX_SCHED_FIFO:
1585 		bsd.policy = SCHED_FIFO;
1586 		break;
1587 	case LINUX_SCHED_RR:
1588 		bsd.policy = SCHED_RR;
1589 		break;
1590 	default:
1591 		return (EINVAL);
1592 	}
1593 	return (sys_sched_get_priority_max(td, &bsd));
1594 }
1595 
1596 int
1597 linux_sched_get_priority_min(struct thread *td,
1598     struct linux_sched_get_priority_min_args *args)
1599 {
1600 	struct sched_get_priority_min_args bsd;
1601 
1602 #ifdef DEBUG
1603 	if (ldebug(sched_get_priority_min))
1604 		printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1605 #endif
1606 
1607 	switch (args->policy) {
1608 	case LINUX_SCHED_OTHER:
1609 		bsd.policy = SCHED_OTHER;
1610 		break;
1611 	case LINUX_SCHED_FIFO:
1612 		bsd.policy = SCHED_FIFO;
1613 		break;
1614 	case LINUX_SCHED_RR:
1615 		bsd.policy = SCHED_RR;
1616 		break;
1617 	default:
1618 		return (EINVAL);
1619 	}
1620 	return (sys_sched_get_priority_min(td, &bsd));
1621 }
1622 
1623 #define REBOOT_CAD_ON	0x89abcdef
1624 #define REBOOT_CAD_OFF	0
1625 #define REBOOT_HALT	0xcdef0123
1626 #define REBOOT_RESTART	0x01234567
1627 #define REBOOT_RESTART2	0xA1B2C3D4
1628 #define REBOOT_POWEROFF	0x4321FEDC
1629 #define REBOOT_MAGIC1	0xfee1dead
1630 #define REBOOT_MAGIC2	0x28121969
1631 #define REBOOT_MAGIC2A	0x05121996
1632 #define REBOOT_MAGIC2B	0x16041998
1633 
1634 int
1635 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1636 {
1637 	struct reboot_args bsd_args;
1638 
1639 #ifdef DEBUG
1640 	if (ldebug(reboot))
1641 		printf(ARGS(reboot, "0x%x"), args->cmd);
1642 #endif
1643 
1644 	if (args->magic1 != REBOOT_MAGIC1)
1645 		return (EINVAL);
1646 
1647 	switch (args->magic2) {
1648 	case REBOOT_MAGIC2:
1649 	case REBOOT_MAGIC2A:
1650 	case REBOOT_MAGIC2B:
1651 		break;
1652 	default:
1653 		return (EINVAL);
1654 	}
1655 
1656 	switch (args->cmd) {
1657 	case REBOOT_CAD_ON:
1658 	case REBOOT_CAD_OFF:
1659 		return (priv_check(td, PRIV_REBOOT));
1660 	case REBOOT_HALT:
1661 		bsd_args.opt = RB_HALT;
1662 		break;
1663 	case REBOOT_RESTART:
1664 	case REBOOT_RESTART2:
1665 		bsd_args.opt = 0;
1666 		break;
1667 	case REBOOT_POWEROFF:
1668 		bsd_args.opt = RB_POWEROFF;
1669 		break;
1670 	default:
1671 		return (EINVAL);
1672 	}
1673 	return (sys_reboot(td, &bsd_args));
1674 }
1675 
1676 
1677 /*
1678  * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1679  * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
1680  * are assumed to be preserved. The following lightweight syscalls fixes
1681  * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
1682  *
1683  * linux_getpid() - MP SAFE
1684  * linux_getgid() - MP SAFE
1685  * linux_getuid() - MP SAFE
1686  */
1687 
1688 int
1689 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1690 {
1691 
1692 #ifdef DEBUG
1693 	if (ldebug(getpid))
1694 		printf(ARGS(getpid, ""));
1695 #endif
1696 	td->td_retval[0] = td->td_proc->p_pid;
1697 
1698 	return (0);
1699 }
1700 
1701 int
1702 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1703 {
1704 	struct linux_emuldata *em;
1705 
1706 #ifdef DEBUG
1707 	if (ldebug(gettid))
1708 		printf(ARGS(gettid, ""));
1709 #endif
1710 
1711 	em = em_find(td);
1712 	KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1713 
1714 	td->td_retval[0] = em->em_tid;
1715 
1716 	return (0);
1717 }
1718 
1719 
1720 int
1721 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1722 {
1723 
1724 #ifdef DEBUG
1725 	if (ldebug(getppid))
1726 		printf(ARGS(getppid, ""));
1727 #endif
1728 
1729 	td->td_retval[0] = kern_getppid(td);
1730 	return (0);
1731 }
1732 
1733 int
1734 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1735 {
1736 
1737 #ifdef DEBUG
1738 	if (ldebug(getgid))
1739 		printf(ARGS(getgid, ""));
1740 #endif
1741 
1742 	td->td_retval[0] = td->td_ucred->cr_rgid;
1743 	return (0);
1744 }
1745 
1746 int
1747 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1748 {
1749 
1750 #ifdef DEBUG
1751 	if (ldebug(getuid))
1752 		printf(ARGS(getuid, ""));
1753 #endif
1754 
1755 	td->td_retval[0] = td->td_ucred->cr_ruid;
1756 	return (0);
1757 }
1758 
1759 
1760 int
1761 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1762 {
1763 	struct getsid_args bsd;
1764 
1765 #ifdef DEBUG
1766 	if (ldebug(getsid))
1767 		printf(ARGS(getsid, "%i"), args->pid);
1768 #endif
1769 
1770 	bsd.pid = args->pid;
1771 	return (sys_getsid(td, &bsd));
1772 }
1773 
1774 int
1775 linux_nosys(struct thread *td, struct nosys_args *ignore)
1776 {
1777 
1778 	return (ENOSYS);
1779 }
1780 
1781 int
1782 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1783 {
1784 	struct getpriority_args bsd_args;
1785 	int error;
1786 
1787 #ifdef DEBUG
1788 	if (ldebug(getpriority))
1789 		printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
1790 #endif
1791 
1792 	bsd_args.which = args->which;
1793 	bsd_args.who = args->who;
1794 	error = sys_getpriority(td, &bsd_args);
1795 	td->td_retval[0] = 20 - td->td_retval[0];
1796 	return (error);
1797 }
1798 
1799 int
1800 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1801 {
1802 	int name[2];
1803 
1804 #ifdef DEBUG
1805 	if (ldebug(sethostname))
1806 		printf(ARGS(sethostname, "*, %i"), args->len);
1807 #endif
1808 
1809 	name[0] = CTL_KERN;
1810 	name[1] = KERN_HOSTNAME;
1811 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1812 	    args->len, 0, 0));
1813 }
1814 
1815 int
1816 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1817 {
1818 	int name[2];
1819 
1820 #ifdef DEBUG
1821 	if (ldebug(setdomainname))
1822 		printf(ARGS(setdomainname, "*, %i"), args->len);
1823 #endif
1824 
1825 	name[0] = CTL_KERN;
1826 	name[1] = KERN_NISDOMAINNAME;
1827 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1828 	    args->len, 0, 0));
1829 }
1830 
1831 int
1832 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1833 {
1834 
1835 #ifdef DEBUG
1836 	if (ldebug(exit_group))
1837 		printf(ARGS(exit_group, "%i"), args->error_code);
1838 #endif
1839 
1840 	LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1841 	    args->error_code);
1842 
1843 	/*
1844 	 * XXX: we should send a signal to the parent if
1845 	 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1846 	 * as it doesnt occur often.
1847 	 */
1848 	exit1(td, args->error_code, 0);
1849 		/* NOTREACHED */
1850 }
1851 
1852 #define _LINUX_CAPABILITY_VERSION  0x19980330
1853 
1854 struct l_user_cap_header {
1855 	l_int	version;
1856 	l_int	pid;
1857 };
1858 
1859 struct l_user_cap_data {
1860 	l_int	effective;
1861 	l_int	permitted;
1862 	l_int	inheritable;
1863 };
1864 
1865 int
1866 linux_capget(struct thread *td, struct linux_capget_args *args)
1867 {
1868 	struct l_user_cap_header luch;
1869 	struct l_user_cap_data lucd;
1870 	int error;
1871 
1872 	if (args->hdrp == NULL)
1873 		return (EFAULT);
1874 
1875 	error = copyin(args->hdrp, &luch, sizeof(luch));
1876 	if (error != 0)
1877 		return (error);
1878 
1879 	if (luch.version != _LINUX_CAPABILITY_VERSION) {
1880 		luch.version = _LINUX_CAPABILITY_VERSION;
1881 		error = copyout(&luch, args->hdrp, sizeof(luch));
1882 		if (error)
1883 			return (error);
1884 		return (EINVAL);
1885 	}
1886 
1887 	if (luch.pid)
1888 		return (EPERM);
1889 
1890 	if (args->datap) {
1891 		/*
1892 		 * The current implementation doesn't support setting
1893 		 * a capability (it's essentially a stub) so indicate
1894 		 * that no capabilities are currently set or available
1895 		 * to request.
1896 		 */
1897 		bzero (&lucd, sizeof(lucd));
1898 		error = copyout(&lucd, args->datap, sizeof(lucd));
1899 	}
1900 
1901 	return (error);
1902 }
1903 
1904 int
1905 linux_capset(struct thread *td, struct linux_capset_args *args)
1906 {
1907 	struct l_user_cap_header luch;
1908 	struct l_user_cap_data lucd;
1909 	int error;
1910 
1911 	if (args->hdrp == NULL || args->datap == NULL)
1912 		return (EFAULT);
1913 
1914 	error = copyin(args->hdrp, &luch, sizeof(luch));
1915 	if (error != 0)
1916 		return (error);
1917 
1918 	if (luch.version != _LINUX_CAPABILITY_VERSION) {
1919 		luch.version = _LINUX_CAPABILITY_VERSION;
1920 		error = copyout(&luch, args->hdrp, sizeof(luch));
1921 		if (error)
1922 			return (error);
1923 		return (EINVAL);
1924 	}
1925 
1926 	if (luch.pid)
1927 		return (EPERM);
1928 
1929 	error = copyin(args->datap, &lucd, sizeof(lucd));
1930 	if (error != 0)
1931 		return (error);
1932 
1933 	/* We currently don't support setting any capabilities. */
1934 	if (lucd.effective || lucd.permitted || lucd.inheritable) {
1935 		linux_msg(td,
1936 			  "capset effective=0x%x, permitted=0x%x, "
1937 			  "inheritable=0x%x is not implemented",
1938 			  (int)lucd.effective, (int)lucd.permitted,
1939 			  (int)lucd.inheritable);
1940 		return (EPERM);
1941 	}
1942 
1943 	return (0);
1944 }
1945 
1946 int
1947 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1948 {
1949 	int error = 0, max_size;
1950 	struct proc *p = td->td_proc;
1951 	char comm[LINUX_MAX_COMM_LEN];
1952 	struct linux_emuldata *em;
1953 	int pdeath_signal;
1954 
1955 #ifdef DEBUG
1956 	if (ldebug(prctl))
1957 		printf(ARGS(prctl, "%d, %ju, %ju, %ju, %ju"), args->option,
1958 		    (uintmax_t)args->arg2, (uintmax_t)args->arg3,
1959 		    (uintmax_t)args->arg4, (uintmax_t)args->arg5);
1960 #endif
1961 
1962 	switch (args->option) {
1963 	case LINUX_PR_SET_PDEATHSIG:
1964 		if (!LINUX_SIG_VALID(args->arg2))
1965 			return (EINVAL);
1966 		em = em_find(td);
1967 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1968 		em->pdeath_signal = args->arg2;
1969 		break;
1970 	case LINUX_PR_GET_PDEATHSIG:
1971 		em = em_find(td);
1972 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1973 		pdeath_signal = em->pdeath_signal;
1974 		error = copyout(&pdeath_signal,
1975 		    (void *)(register_t)args->arg2,
1976 		    sizeof(pdeath_signal));
1977 		break;
1978 	case LINUX_PR_GET_KEEPCAPS:
1979 		/*
1980 		 * Indicate that we always clear the effective and
1981 		 * permitted capability sets when the user id becomes
1982 		 * non-zero (actually the capability sets are simply
1983 		 * always zero in the current implementation).
1984 		 */
1985 		td->td_retval[0] = 0;
1986 		break;
1987 	case LINUX_PR_SET_KEEPCAPS:
1988 		/*
1989 		 * Ignore requests to keep the effective and permitted
1990 		 * capability sets when the user id becomes non-zero.
1991 		 */
1992 		break;
1993 	case LINUX_PR_SET_NAME:
1994 		/*
1995 		 * To be on the safe side we need to make sure to not
1996 		 * overflow the size a linux program expects. We already
1997 		 * do this here in the copyin, so that we don't need to
1998 		 * check on copyout.
1999 		 */
2000 		max_size = MIN(sizeof(comm), sizeof(p->p_comm));
2001 		error = copyinstr((void *)(register_t)args->arg2, comm,
2002 		    max_size, NULL);
2003 
2004 		/* Linux silently truncates the name if it is too long. */
2005 		if (error == ENAMETOOLONG) {
2006 			/*
2007 			 * XXX: copyinstr() isn't documented to populate the
2008 			 * array completely, so do a copyin() to be on the
2009 			 * safe side. This should be changed in case
2010 			 * copyinstr() is changed to guarantee this.
2011 			 */
2012 			error = copyin((void *)(register_t)args->arg2, comm,
2013 			    max_size - 1);
2014 			comm[max_size - 1] = '\0';
2015 		}
2016 		if (error)
2017 			return (error);
2018 
2019 		PROC_LOCK(p);
2020 		strlcpy(p->p_comm, comm, sizeof(p->p_comm));
2021 		PROC_UNLOCK(p);
2022 		break;
2023 	case LINUX_PR_GET_NAME:
2024 		PROC_LOCK(p);
2025 		strlcpy(comm, p->p_comm, sizeof(comm));
2026 		PROC_UNLOCK(p);
2027 		error = copyout(comm, (void *)(register_t)args->arg2,
2028 		    strlen(comm) + 1);
2029 		break;
2030 	default:
2031 		error = EINVAL;
2032 		break;
2033 	}
2034 
2035 	return (error);
2036 }
2037 
2038 int
2039 linux_sched_setparam(struct thread *td,
2040     struct linux_sched_setparam_args *uap)
2041 {
2042 	struct sched_param sched_param;
2043 	struct thread *tdt;
2044 	int error;
2045 
2046 #ifdef DEBUG
2047 	if (ldebug(sched_setparam))
2048 		printf(ARGS(sched_setparam, "%d, *"), uap->pid);
2049 #endif
2050 
2051 	error = copyin(uap->param, &sched_param, sizeof(sched_param));
2052 	if (error)
2053 		return (error);
2054 
2055 	tdt = linux_tdfind(td, uap->pid, -1);
2056 	if (tdt == NULL)
2057 		return (ESRCH);
2058 
2059 	error = kern_sched_setparam(td, tdt, &sched_param);
2060 	PROC_UNLOCK(tdt->td_proc);
2061 	return (error);
2062 }
2063 
2064 int
2065 linux_sched_getparam(struct thread *td,
2066     struct linux_sched_getparam_args *uap)
2067 {
2068 	struct sched_param sched_param;
2069 	struct thread *tdt;
2070 	int error;
2071 
2072 #ifdef DEBUG
2073 	if (ldebug(sched_getparam))
2074 		printf(ARGS(sched_getparam, "%d, *"), uap->pid);
2075 #endif
2076 
2077 	tdt = linux_tdfind(td, uap->pid, -1);
2078 	if (tdt == NULL)
2079 		return (ESRCH);
2080 
2081 	error = kern_sched_getparam(td, tdt, &sched_param);
2082 	PROC_UNLOCK(tdt->td_proc);
2083 	if (error == 0)
2084 		error = copyout(&sched_param, uap->param,
2085 		    sizeof(sched_param));
2086 	return (error);
2087 }
2088 
2089 /*
2090  * Get affinity of a process.
2091  */
2092 int
2093 linux_sched_getaffinity(struct thread *td,
2094     struct linux_sched_getaffinity_args *args)
2095 {
2096 	int error;
2097 	struct thread *tdt;
2098 
2099 #ifdef DEBUG
2100 	if (ldebug(sched_getaffinity))
2101 		printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid,
2102 		    args->len);
2103 #endif
2104 	if (args->len < sizeof(cpuset_t))
2105 		return (EINVAL);
2106 
2107 	tdt = linux_tdfind(td, args->pid, -1);
2108 	if (tdt == NULL)
2109 		return (ESRCH);
2110 
2111 	PROC_UNLOCK(tdt->td_proc);
2112 
2113 	error = kern_cpuset_getaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2114 	    tdt->td_tid, sizeof(cpuset_t), (cpuset_t *)args->user_mask_ptr);
2115 	if (error == 0)
2116 		td->td_retval[0] = sizeof(cpuset_t);
2117 
2118 	return (error);
2119 }
2120 
2121 /*
2122  *  Set affinity of a process.
2123  */
2124 int
2125 linux_sched_setaffinity(struct thread *td,
2126     struct linux_sched_setaffinity_args *args)
2127 {
2128 	struct thread *tdt;
2129 
2130 #ifdef DEBUG
2131 	if (ldebug(sched_setaffinity))
2132 		printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid,
2133 		    args->len);
2134 #endif
2135 	if (args->len < sizeof(cpuset_t))
2136 		return (EINVAL);
2137 
2138 	tdt = linux_tdfind(td, args->pid, -1);
2139 	if (tdt == NULL)
2140 		return (ESRCH);
2141 
2142 	PROC_UNLOCK(tdt->td_proc);
2143 
2144 	return (kern_cpuset_setaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2145 	    tdt->td_tid, sizeof(cpuset_t), (cpuset_t *) args->user_mask_ptr));
2146 }
2147 
2148 struct linux_rlimit64 {
2149 	uint64_t	rlim_cur;
2150 	uint64_t	rlim_max;
2151 };
2152 
2153 int
2154 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
2155 {
2156 	struct rlimit rlim, nrlim;
2157 	struct linux_rlimit64 lrlim;
2158 	struct proc *p;
2159 	u_int which;
2160 	int flags;
2161 	int error;
2162 
2163 #ifdef DEBUG
2164 	if (ldebug(prlimit64))
2165 		printf(ARGS(prlimit64, "%d, %d, %p, %p"), args->pid,
2166 		    args->resource, (void *)args->new, (void *)args->old);
2167 #endif
2168 
2169 	if (args->resource >= LINUX_RLIM_NLIMITS)
2170 		return (EINVAL);
2171 
2172 	which = linux_to_bsd_resource[args->resource];
2173 	if (which == -1)
2174 		return (EINVAL);
2175 
2176 	if (args->new != NULL) {
2177 		/*
2178 		 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
2179 		 * rlim is unsigned 64-bit. FreeBSD treats negative limits
2180 		 * as INFINITY so we do not need a conversion even.
2181 		 */
2182 		error = copyin(args->new, &nrlim, sizeof(nrlim));
2183 		if (error != 0)
2184 			return (error);
2185 	}
2186 
2187 	flags = PGET_HOLD | PGET_NOTWEXIT;
2188 	if (args->new != NULL)
2189 		flags |= PGET_CANDEBUG;
2190 	else
2191 		flags |= PGET_CANSEE;
2192 	error = pget(args->pid, flags, &p);
2193 	if (error != 0)
2194 		return (error);
2195 
2196 	if (args->old != NULL) {
2197 		PROC_LOCK(p);
2198 		lim_rlimit_proc(p, which, &rlim);
2199 		PROC_UNLOCK(p);
2200 		if (rlim.rlim_cur == RLIM_INFINITY)
2201 			lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2202 		else
2203 			lrlim.rlim_cur = rlim.rlim_cur;
2204 		if (rlim.rlim_max == RLIM_INFINITY)
2205 			lrlim.rlim_max = LINUX_RLIM_INFINITY;
2206 		else
2207 			lrlim.rlim_max = rlim.rlim_max;
2208 		error = copyout(&lrlim, args->old, sizeof(lrlim));
2209 		if (error != 0)
2210 			goto out;
2211 	}
2212 
2213 	if (args->new != NULL)
2214 		error = kern_proc_setrlimit(td, p, which, &nrlim);
2215 
2216  out:
2217 	PRELE(p);
2218 	return (error);
2219 }
2220 
2221 int
2222 linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2223 {
2224 	struct timeval utv, tv0, tv1, *tvp;
2225 	struct l_pselect6arg lpse6;
2226 	struct l_timespec lts;
2227 	struct timespec uts;
2228 	l_sigset_t l_ss;
2229 	sigset_t *ssp;
2230 	sigset_t ss;
2231 	int error;
2232 
2233 	ssp = NULL;
2234 	if (args->sig != NULL) {
2235 		error = copyin(args->sig, &lpse6, sizeof(lpse6));
2236 		if (error != 0)
2237 			return (error);
2238 		if (lpse6.ss_len != sizeof(l_ss))
2239 			return (EINVAL);
2240 		if (lpse6.ss != 0) {
2241 			error = copyin(PTRIN(lpse6.ss), &l_ss,
2242 			    sizeof(l_ss));
2243 			if (error != 0)
2244 				return (error);
2245 			linux_to_bsd_sigset(&l_ss, &ss);
2246 			ssp = &ss;
2247 		}
2248 	}
2249 
2250 	/*
2251 	 * Currently glibc changes nanosecond number to microsecond.
2252 	 * This mean losing precision but for now it is hardly seen.
2253 	 */
2254 	if (args->tsp != NULL) {
2255 		error = copyin(args->tsp, &lts, sizeof(lts));
2256 		if (error != 0)
2257 			return (error);
2258 		error = linux_to_native_timespec(&uts, &lts);
2259 		if (error != 0)
2260 			return (error);
2261 
2262 		TIMESPEC_TO_TIMEVAL(&utv, &uts);
2263 		if (itimerfix(&utv))
2264 			return (EINVAL);
2265 
2266 		microtime(&tv0);
2267 		tvp = &utv;
2268 	} else
2269 		tvp = NULL;
2270 
2271 	error = kern_pselect(td, args->nfds, args->readfds, args->writefds,
2272 	    args->exceptfds, tvp, ssp, LINUX_NFDBITS);
2273 
2274 	if (error == 0 && args->tsp != NULL) {
2275 		if (td->td_retval[0] != 0) {
2276 			/*
2277 			 * Compute how much time was left of the timeout,
2278 			 * by subtracting the current time and the time
2279 			 * before we started the call, and subtracting
2280 			 * that result from the user-supplied value.
2281 			 */
2282 
2283 			microtime(&tv1);
2284 			timevalsub(&tv1, &tv0);
2285 			timevalsub(&utv, &tv1);
2286 			if (utv.tv_sec < 0)
2287 				timevalclear(&utv);
2288 		} else
2289 			timevalclear(&utv);
2290 
2291 		TIMEVAL_TO_TIMESPEC(&utv, &uts);
2292 
2293 		error = native_to_linux_timespec(&lts, &uts);
2294 		if (error == 0)
2295 			error = copyout(&lts, args->tsp, sizeof(lts));
2296 	}
2297 
2298 	return (error);
2299 }
2300 
2301 int
2302 linux_ppoll(struct thread *td, struct linux_ppoll_args *args)
2303 {
2304 	struct timespec ts0, ts1;
2305 	struct l_timespec lts;
2306 	struct timespec uts, *tsp;
2307 	l_sigset_t l_ss;
2308 	sigset_t *ssp;
2309 	sigset_t ss;
2310 	int error;
2311 
2312 	if (args->sset != NULL) {
2313 		if (args->ssize != sizeof(l_ss))
2314 			return (EINVAL);
2315 		error = copyin(args->sset, &l_ss, sizeof(l_ss));
2316 		if (error)
2317 			return (error);
2318 		linux_to_bsd_sigset(&l_ss, &ss);
2319 		ssp = &ss;
2320 	} else
2321 		ssp = NULL;
2322 	if (args->tsp != NULL) {
2323 		error = copyin(args->tsp, &lts, sizeof(lts));
2324 		if (error)
2325 			return (error);
2326 		error = linux_to_native_timespec(&uts, &lts);
2327 		if (error != 0)
2328 			return (error);
2329 
2330 		nanotime(&ts0);
2331 		tsp = &uts;
2332 	} else
2333 		tsp = NULL;
2334 
2335 	error = kern_poll(td, args->fds, args->nfds, tsp, ssp);
2336 
2337 	if (error == 0 && args->tsp != NULL) {
2338 		if (td->td_retval[0]) {
2339 			nanotime(&ts1);
2340 			timespecsub(&ts1, &ts0);
2341 			timespecsub(&uts, &ts1);
2342 			if (uts.tv_sec < 0)
2343 				timespecclear(&uts);
2344 		} else
2345 			timespecclear(&uts);
2346 
2347 		error = native_to_linux_timespec(&lts, &uts);
2348 		if (error == 0)
2349 			error = copyout(&lts, args->tsp, sizeof(lts));
2350 	}
2351 
2352 	return (error);
2353 }
2354 
2355 #if defined(DEBUG) || defined(KTR)
2356 /* XXX: can be removed when every ldebug(...) and KTR stuff are removed. */
2357 
2358 #ifdef COMPAT_LINUX32
2359 #define	L_MAXSYSCALL	LINUX32_SYS_MAXSYSCALL
2360 #else
2361 #define	L_MAXSYSCALL	LINUX_SYS_MAXSYSCALL
2362 #endif
2363 
2364 u_char linux_debug_map[howmany(L_MAXSYSCALL, sizeof(u_char))];
2365 
2366 static int
2367 linux_debug(int syscall, int toggle, int global)
2368 {
2369 
2370 	if (global) {
2371 		char c = toggle ? 0 : 0xff;
2372 
2373 		memset(linux_debug_map, c, sizeof(linux_debug_map));
2374 		return (0);
2375 	}
2376 	if (syscall < 0 || syscall >= L_MAXSYSCALL)
2377 		return (EINVAL);
2378 	if (toggle)
2379 		clrbit(linux_debug_map, syscall);
2380 	else
2381 		setbit(linux_debug_map, syscall);
2382 	return (0);
2383 }
2384 #undef L_MAXSYSCALL
2385 
2386 /*
2387  * Usage: sysctl linux.debug=<syscall_nr>.<0/1>
2388  *
2389  *    E.g.: sysctl linux.debug=21.0
2390  *
2391  * As a special case, syscall "all" will apply to all syscalls globally.
2392  */
2393 #define LINUX_MAX_DEBUGSTR	16
2394 int
2395 linux_sysctl_debug(SYSCTL_HANDLER_ARGS)
2396 {
2397 	char value[LINUX_MAX_DEBUGSTR], *p;
2398 	int error, sysc, toggle;
2399 	int global = 0;
2400 
2401 	value[0] = '\0';
2402 	error = sysctl_handle_string(oidp, value, LINUX_MAX_DEBUGSTR, req);
2403 	if (error || req->newptr == NULL)
2404 		return (error);
2405 	for (p = value; *p != '\0' && *p != '.'; p++);
2406 	if (*p == '\0')
2407 		return (EINVAL);
2408 	*p++ = '\0';
2409 	sysc = strtol(value, NULL, 0);
2410 	toggle = strtol(p, NULL, 0);
2411 	if (strcmp(value, "all") == 0)
2412 		global = 1;
2413 	error = linux_debug(sysc, toggle, global);
2414 	return (error);
2415 }
2416 
2417 #endif /* DEBUG || KTR */
2418 
2419 int
2420 linux_sched_rr_get_interval(struct thread *td,
2421     struct linux_sched_rr_get_interval_args *uap)
2422 {
2423 	struct timespec ts;
2424 	struct l_timespec lts;
2425 	struct thread *tdt;
2426 	int error;
2427 
2428 	/*
2429 	 * According to man in case the invalid pid specified
2430 	 * EINVAL should be returned.
2431 	 */
2432 	if (uap->pid < 0)
2433 		return (EINVAL);
2434 
2435 	tdt = linux_tdfind(td, uap->pid, -1);
2436 	if (tdt == NULL)
2437 		return (ESRCH);
2438 
2439 	error = kern_sched_rr_get_interval_td(td, tdt, &ts);
2440 	PROC_UNLOCK(tdt->td_proc);
2441 	if (error != 0)
2442 		return (error);
2443 	error = native_to_linux_timespec(&lts, &ts);
2444 	if (error != 0)
2445 		return (error);
2446 	return (copyout(&lts, uap->interval, sizeof(lts)));
2447 }
2448 
2449 /*
2450  * In case when the Linux thread is the initial thread in
2451  * the thread group thread id is equal to the process id.
2452  * Glibc depends on this magic (assert in pthread_getattr_np.c).
2453  */
2454 struct thread *
2455 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid)
2456 {
2457 	struct linux_emuldata *em;
2458 	struct thread *tdt;
2459 	struct proc *p;
2460 
2461 	tdt = NULL;
2462 	if (tid == 0 || tid == td->td_tid) {
2463 		tdt = td;
2464 		PROC_LOCK(tdt->td_proc);
2465 	} else if (tid > PID_MAX)
2466 		tdt = tdfind(tid, pid);
2467 	else {
2468 		/*
2469 		 * Initial thread where the tid equal to the pid.
2470 		 */
2471 		p = pfind(tid);
2472 		if (p != NULL) {
2473 			if (SV_PROC_ABI(p) != SV_ABI_LINUX) {
2474 				/*
2475 				 * p is not a Linuxulator process.
2476 				 */
2477 				PROC_UNLOCK(p);
2478 				return (NULL);
2479 			}
2480 			FOREACH_THREAD_IN_PROC(p, tdt) {
2481 				em = em_find(tdt);
2482 				if (tid == em->em_tid)
2483 					return (tdt);
2484 			}
2485 			PROC_UNLOCK(p);
2486 		}
2487 		return (NULL);
2488 	}
2489 
2490 	return (tdt);
2491 }
2492 
2493 void
2494 linux_to_bsd_waitopts(int options, int *bsdopts)
2495 {
2496 
2497 	if (options & LINUX_WNOHANG)
2498 		*bsdopts |= WNOHANG;
2499 	if (options & LINUX_WUNTRACED)
2500 		*bsdopts |= WUNTRACED;
2501 	if (options & LINUX_WEXITED)
2502 		*bsdopts |= WEXITED;
2503 	if (options & LINUX_WCONTINUED)
2504 		*bsdopts |= WCONTINUED;
2505 	if (options & LINUX_WNOWAIT)
2506 		*bsdopts |= WNOWAIT;
2507 
2508 	if (options & __WCLONE)
2509 		*bsdopts |= WLINUXCLONE;
2510 }
2511