xref: /freebsd/sys/compat/linux/linux_misc.c (revision 895f86f15fbf6540071feb9328c3c50ed1f027b8)
1 /*-
2  * Copyright (c) 2002 Doug Rabson
3  * Copyright (c) 1994-1995 Søren Schmidt
4  * All rights reserved.
5  *
6  * Redistribution and use in source and binary forms, with or without
7  * modification, are permitted provided that the following conditions
8  * are met:
9  * 1. Redistributions of source code must retain the above copyright
10  *    notice, this list of conditions and the following disclaimer
11  *    in this position and unchanged.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. The name of the author may not be used to endorse or promote products
16  *    derived from this software without specific prior written permission
17  *
18  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
19  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
20  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
21  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
22  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
23  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
24  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
25  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
26  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
27  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
28  */
29 
30 #include <sys/cdefs.h>
31 __FBSDID("$FreeBSD$");
32 
33 #include "opt_compat.h"
34 
35 #include <sys/param.h>
36 #include <sys/blist.h>
37 #include <sys/fcntl.h>
38 #if defined(__i386__)
39 #include <sys/imgact_aout.h>
40 #endif
41 #include <sys/jail.h>
42 #include <sys/kernel.h>
43 #include <sys/limits.h>
44 #include <sys/lock.h>
45 #include <sys/malloc.h>
46 #include <sys/mman.h>
47 #include <sys/mount.h>
48 #include <sys/mutex.h>
49 #include <sys/namei.h>
50 #include <sys/priv.h>
51 #include <sys/proc.h>
52 #include <sys/reboot.h>
53 #include <sys/racct.h>
54 #include <sys/resourcevar.h>
55 #include <sys/sched.h>
56 #include <sys/sdt.h>
57 #include <sys/signalvar.h>
58 #include <sys/stat.h>
59 #include <sys/syscallsubr.h>
60 #include <sys/sysctl.h>
61 #include <sys/sysproto.h>
62 #include <sys/systm.h>
63 #include <sys/time.h>
64 #include <sys/vmmeter.h>
65 #include <sys/vnode.h>
66 #include <sys/wait.h>
67 #include <sys/cpuset.h>
68 
69 #include <security/mac/mac_framework.h>
70 
71 #include <vm/vm.h>
72 #include <vm/pmap.h>
73 #include <vm/vm_kern.h>
74 #include <vm/vm_map.h>
75 #include <vm/vm_extern.h>
76 #include <vm/vm_object.h>
77 #include <vm/swap_pager.h>
78 
79 #ifdef COMPAT_LINUX32
80 #include <machine/../linux32/linux.h>
81 #include <machine/../linux32/linux32_proto.h>
82 #else
83 #include <machine/../linux/linux.h>
84 #include <machine/../linux/linux_proto.h>
85 #endif
86 
87 #include <compat/linux/linux_dtrace.h>
88 #include <compat/linux/linux_file.h>
89 #include <compat/linux/linux_mib.h>
90 #include <compat/linux/linux_signal.h>
91 #include <compat/linux/linux_timer.h>
92 #include <compat/linux/linux_util.h>
93 #include <compat/linux/linux_sysproto.h>
94 #include <compat/linux/linux_emul.h>
95 #include <compat/linux/linux_misc.h>
96 
97 /**
98  * Special DTrace provider for the linuxulator.
99  *
100  * In this file we define the provider for the entire linuxulator. All
101  * modules (= files of the linuxulator) use it.
102  *
103  * We define a different name depending on the emulated bitsize, see
104  * ../../<ARCH>/linux{,32}/linux.h, e.g.:
105  *      native bitsize          = linuxulator
106  *      amd64, 32bit emulation  = linuxulator32
107  */
108 LIN_SDT_PROVIDER_DEFINE(LINUX_DTRACE);
109 
110 int stclohz;				/* Statistics clock frequency */
111 
112 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
113 	RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
114 	RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
115 	RLIMIT_MEMLOCK, RLIMIT_AS
116 };
117 
118 struct l_sysinfo {
119 	l_long		uptime;		/* Seconds since boot */
120 	l_ulong		loads[3];	/* 1, 5, and 15 minute load averages */
121 #define LINUX_SYSINFO_LOADS_SCALE 65536
122 	l_ulong		totalram;	/* Total usable main memory size */
123 	l_ulong		freeram;	/* Available memory size */
124 	l_ulong		sharedram;	/* Amount of shared memory */
125 	l_ulong		bufferram;	/* Memory used by buffers */
126 	l_ulong		totalswap;	/* Total swap space size */
127 	l_ulong		freeswap;	/* swap space still available */
128 	l_ushort	procs;		/* Number of current processes */
129 	l_ushort	pads;
130 	l_ulong		totalbig;
131 	l_ulong		freebig;
132 	l_uint		mem_unit;
133 	char		_f[20-2*sizeof(l_long)-sizeof(l_int)];	/* padding */
134 };
135 
136 struct l_pselect6arg {
137 	l_uintptr_t	ss;
138 	l_size_t	ss_len;
139 };
140 
141 static int	linux_utimensat_nsec_valid(l_long);
142 
143 
144 int
145 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
146 {
147 	struct l_sysinfo sysinfo;
148 	vm_object_t object;
149 	int i, j;
150 	struct timespec ts;
151 
152 	getnanouptime(&ts);
153 	if (ts.tv_nsec != 0)
154 		ts.tv_sec++;
155 	sysinfo.uptime = ts.tv_sec;
156 
157 	/* Use the information from the mib to get our load averages */
158 	for (i = 0; i < 3; i++)
159 		sysinfo.loads[i] = averunnable.ldavg[i] *
160 		    LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
161 
162 	sysinfo.totalram = physmem * PAGE_SIZE;
163 	sysinfo.freeram = sysinfo.totalram - vm_cnt.v_wire_count * PAGE_SIZE;
164 
165 	sysinfo.sharedram = 0;
166 	mtx_lock(&vm_object_list_mtx);
167 	TAILQ_FOREACH(object, &vm_object_list, object_list)
168 		if (object->shadow_count > 1)
169 			sysinfo.sharedram += object->resident_page_count;
170 	mtx_unlock(&vm_object_list_mtx);
171 
172 	sysinfo.sharedram *= PAGE_SIZE;
173 	sysinfo.bufferram = 0;
174 
175 	swap_pager_status(&i, &j);
176 	sysinfo.totalswap = i * PAGE_SIZE;
177 	sysinfo.freeswap = (i - j) * PAGE_SIZE;
178 
179 	sysinfo.procs = nprocs;
180 
181 	/* The following are only present in newer Linux kernels. */
182 	sysinfo.totalbig = 0;
183 	sysinfo.freebig = 0;
184 	sysinfo.mem_unit = 1;
185 
186 	return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
187 }
188 
189 int
190 linux_alarm(struct thread *td, struct linux_alarm_args *args)
191 {
192 	struct itimerval it, old_it;
193 	u_int secs;
194 	int error;
195 
196 #ifdef DEBUG
197 	if (ldebug(alarm))
198 		printf(ARGS(alarm, "%u"), args->secs);
199 #endif
200 
201 	secs = args->secs;
202 
203 	if (secs > INT_MAX)
204 		secs = INT_MAX;
205 
206 	it.it_value.tv_sec = (long) secs;
207 	it.it_value.tv_usec = 0;
208 	it.it_interval.tv_sec = 0;
209 	it.it_interval.tv_usec = 0;
210 	error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
211 	if (error)
212 		return (error);
213 	if (timevalisset(&old_it.it_value)) {
214 		if (old_it.it_value.tv_usec != 0)
215 			old_it.it_value.tv_sec++;
216 		td->td_retval[0] = old_it.it_value.tv_sec;
217 	}
218 	return (0);
219 }
220 
221 int
222 linux_brk(struct thread *td, struct linux_brk_args *args)
223 {
224 	struct vmspace *vm = td->td_proc->p_vmspace;
225 	vm_offset_t new, old;
226 	struct obreak_args /* {
227 		char * nsize;
228 	} */ tmp;
229 
230 #ifdef DEBUG
231 	if (ldebug(brk))
232 		printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
233 #endif
234 	old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize);
235 	new = (vm_offset_t)args->dsend;
236 	tmp.nsize = (char *)new;
237 	if (((caddr_t)new > vm->vm_daddr) && !sys_obreak(td, &tmp))
238 		td->td_retval[0] = (long)new;
239 	else
240 		td->td_retval[0] = (long)old;
241 
242 	return (0);
243 }
244 
245 #if defined(__i386__)
246 /* XXX: what about amd64/linux32? */
247 
248 int
249 linux_uselib(struct thread *td, struct linux_uselib_args *args)
250 {
251 	struct nameidata ni;
252 	struct vnode *vp;
253 	struct exec *a_out;
254 	struct vattr attr;
255 	vm_offset_t vmaddr;
256 	unsigned long file_offset;
257 	unsigned long bss_size;
258 	char *library;
259 	ssize_t aresid;
260 	int error, locked, writecount;
261 
262 	LCONVPATHEXIST(td, args->library, &library);
263 
264 #ifdef DEBUG
265 	if (ldebug(uselib))
266 		printf(ARGS(uselib, "%s"), library);
267 #endif
268 
269 	a_out = NULL;
270 	locked = 0;
271 	vp = NULL;
272 
273 	NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
274 	    UIO_SYSSPACE, library, td);
275 	error = namei(&ni);
276 	LFREEPATH(library);
277 	if (error)
278 		goto cleanup;
279 
280 	vp = ni.ni_vp;
281 	NDFREE(&ni, NDF_ONLY_PNBUF);
282 
283 	/*
284 	 * From here on down, we have a locked vnode that must be unlocked.
285 	 * XXX: The code below largely duplicates exec_check_permissions().
286 	 */
287 	locked = 1;
288 
289 	/* Writable? */
290 	error = VOP_GET_WRITECOUNT(vp, &writecount);
291 	if (error != 0)
292 		goto cleanup;
293 	if (writecount != 0) {
294 		error = ETXTBSY;
295 		goto cleanup;
296 	}
297 
298 	/* Executable? */
299 	error = VOP_GETATTR(vp, &attr, td->td_ucred);
300 	if (error)
301 		goto cleanup;
302 
303 	if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
304 	    ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
305 		/* EACCESS is what exec(2) returns. */
306 		error = ENOEXEC;
307 		goto cleanup;
308 	}
309 
310 	/* Sensible size? */
311 	if (attr.va_size == 0) {
312 		error = ENOEXEC;
313 		goto cleanup;
314 	}
315 
316 	/* Can we access it? */
317 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
318 	if (error)
319 		goto cleanup;
320 
321 	/*
322 	 * XXX: This should use vn_open() so that it is properly authorized,
323 	 * and to reduce code redundancy all over the place here.
324 	 * XXX: Not really, it duplicates far more of exec_check_permissions()
325 	 * than vn_open().
326 	 */
327 #ifdef MAC
328 	error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
329 	if (error)
330 		goto cleanup;
331 #endif
332 	error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
333 	if (error)
334 		goto cleanup;
335 
336 	/* Pull in executable header into exec_map */
337 	error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
338 	    VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
339 	if (error)
340 		goto cleanup;
341 
342 	/* Is it a Linux binary ? */
343 	if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
344 		error = ENOEXEC;
345 		goto cleanup;
346 	}
347 
348 	/*
349 	 * While we are here, we should REALLY do some more checks
350 	 */
351 
352 	/* Set file/virtual offset based on a.out variant. */
353 	switch ((int)(a_out->a_magic & 0xffff)) {
354 	case 0413:			/* ZMAGIC */
355 		file_offset = 1024;
356 		break;
357 	case 0314:			/* QMAGIC */
358 		file_offset = 0;
359 		break;
360 	default:
361 		error = ENOEXEC;
362 		goto cleanup;
363 	}
364 
365 	bss_size = round_page(a_out->a_bss);
366 
367 	/* Check various fields in header for validity/bounds. */
368 	if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
369 		error = ENOEXEC;
370 		goto cleanup;
371 	}
372 
373 	/* text + data can't exceed file size */
374 	if (a_out->a_data + a_out->a_text > attr.va_size) {
375 		error = EFAULT;
376 		goto cleanup;
377 	}
378 
379 	/*
380 	 * text/data/bss must not exceed limits
381 	 * XXX - this is not complete. it should check current usage PLUS
382 	 * the resources needed by this library.
383 	 */
384 	PROC_LOCK(td->td_proc);
385 	if (a_out->a_text > maxtsiz ||
386 	    a_out->a_data + bss_size > lim_cur_proc(td->td_proc, RLIMIT_DATA) ||
387 	    racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
388 	    bss_size) != 0) {
389 		PROC_UNLOCK(td->td_proc);
390 		error = ENOMEM;
391 		goto cleanup;
392 	}
393 	PROC_UNLOCK(td->td_proc);
394 
395 	/*
396 	 * Prevent more writers.
397 	 * XXX: Note that if any of the VM operations fail below we don't
398 	 * clear this flag.
399 	 */
400 	VOP_SET_TEXT(vp);
401 
402 	/*
403 	 * Lock no longer needed
404 	 */
405 	locked = 0;
406 	VOP_UNLOCK(vp, 0);
407 
408 	/*
409 	 * Check if file_offset page aligned. Currently we cannot handle
410 	 * misalinged file offsets, and so we read in the entire image
411 	 * (what a waste).
412 	 */
413 	if (file_offset & PAGE_MASK) {
414 #ifdef DEBUG
415 		printf("uselib: Non page aligned binary %lu\n", file_offset);
416 #endif
417 		/* Map text+data read/write/execute */
418 
419 		/* a_entry is the load address and is page aligned */
420 		vmaddr = trunc_page(a_out->a_entry);
421 
422 		/* get anon user mapping, read+write+execute */
423 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
424 		    &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
425 		    VM_PROT_ALL, VM_PROT_ALL, 0);
426 		if (error)
427 			goto cleanup;
428 
429 		error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
430 		    a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
431 		    td->td_ucred, NOCRED, &aresid, td);
432 		if (error != 0)
433 			goto cleanup;
434 		if (aresid != 0) {
435 			error = ENOEXEC;
436 			goto cleanup;
437 		}
438 	} else {
439 #ifdef DEBUG
440 		printf("uselib: Page aligned binary %lu\n", file_offset);
441 #endif
442 		/*
443 		 * for QMAGIC, a_entry is 20 bytes beyond the load address
444 		 * to skip the executable header
445 		 */
446 		vmaddr = trunc_page(a_out->a_entry);
447 
448 		/*
449 		 * Map it all into the process's space as a single
450 		 * copy-on-write "data" segment.
451 		 */
452 		error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr,
453 		    a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
454 		    MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
455 		if (error)
456 			goto cleanup;
457 	}
458 #ifdef DEBUG
459 	printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
460 	    ((long *)vmaddr)[1]);
461 #endif
462 	if (bss_size != 0) {
463 		/* Calculate BSS start address */
464 		vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
465 		    a_out->a_data;
466 
467 		/* allocate some 'anon' space */
468 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
469 		    &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
470 		    VM_PROT_ALL, 0);
471 		if (error)
472 			goto cleanup;
473 	}
474 
475 cleanup:
476 	/* Unlock vnode if needed */
477 	if (locked)
478 		VOP_UNLOCK(vp, 0);
479 
480 	/* Release the temporary mapping. */
481 	if (a_out)
482 		kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
483 
484 	return (error);
485 }
486 
487 #endif	/* __i386__ */
488 
489 int
490 linux_select(struct thread *td, struct linux_select_args *args)
491 {
492 	l_timeval ltv;
493 	struct timeval tv0, tv1, utv, *tvp;
494 	int error;
495 
496 #ifdef DEBUG
497 	if (ldebug(select))
498 		printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
499 		    (void *)args->readfds, (void *)args->writefds,
500 		    (void *)args->exceptfds, (void *)args->timeout);
501 #endif
502 
503 	/*
504 	 * Store current time for computation of the amount of
505 	 * time left.
506 	 */
507 	if (args->timeout) {
508 		if ((error = copyin(args->timeout, &ltv, sizeof(ltv))))
509 			goto select_out;
510 		utv.tv_sec = ltv.tv_sec;
511 		utv.tv_usec = ltv.tv_usec;
512 #ifdef DEBUG
513 		if (ldebug(select))
514 			printf(LMSG("incoming timeout (%jd/%ld)"),
515 			    (intmax_t)utv.tv_sec, utv.tv_usec);
516 #endif
517 
518 		if (itimerfix(&utv)) {
519 			/*
520 			 * The timeval was invalid.  Convert it to something
521 			 * valid that will act as it does under Linux.
522 			 */
523 			utv.tv_sec += utv.tv_usec / 1000000;
524 			utv.tv_usec %= 1000000;
525 			if (utv.tv_usec < 0) {
526 				utv.tv_sec -= 1;
527 				utv.tv_usec += 1000000;
528 			}
529 			if (utv.tv_sec < 0)
530 				timevalclear(&utv);
531 		}
532 		microtime(&tv0);
533 		tvp = &utv;
534 	} else
535 		tvp = NULL;
536 
537 	error = kern_select(td, args->nfds, args->readfds, args->writefds,
538 	    args->exceptfds, tvp, LINUX_NFDBITS);
539 
540 #ifdef DEBUG
541 	if (ldebug(select))
542 		printf(LMSG("real select returns %d"), error);
543 #endif
544 	if (error)
545 		goto select_out;
546 
547 	if (args->timeout) {
548 		if (td->td_retval[0]) {
549 			/*
550 			 * Compute how much time was left of the timeout,
551 			 * by subtracting the current time and the time
552 			 * before we started the call, and subtracting
553 			 * that result from the user-supplied value.
554 			 */
555 			microtime(&tv1);
556 			timevalsub(&tv1, &tv0);
557 			timevalsub(&utv, &tv1);
558 			if (utv.tv_sec < 0)
559 				timevalclear(&utv);
560 		} else
561 			timevalclear(&utv);
562 #ifdef DEBUG
563 		if (ldebug(select))
564 			printf(LMSG("outgoing timeout (%jd/%ld)"),
565 			    (intmax_t)utv.tv_sec, utv.tv_usec);
566 #endif
567 		ltv.tv_sec = utv.tv_sec;
568 		ltv.tv_usec = utv.tv_usec;
569 		if ((error = copyout(&ltv, args->timeout, sizeof(ltv))))
570 			goto select_out;
571 	}
572 
573 select_out:
574 #ifdef DEBUG
575 	if (ldebug(select))
576 		printf(LMSG("select_out -> %d"), error);
577 #endif
578 	return (error);
579 }
580 
581 int
582 linux_mremap(struct thread *td, struct linux_mremap_args *args)
583 {
584 	struct munmap_args /* {
585 		void *addr;
586 		size_t len;
587 	} */ bsd_args;
588 	int error = 0;
589 
590 #ifdef DEBUG
591 	if (ldebug(mremap))
592 		printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
593 		    (void *)(uintptr_t)args->addr,
594 		    (unsigned long)args->old_len,
595 		    (unsigned long)args->new_len,
596 		    (unsigned long)args->flags);
597 #endif
598 
599 	if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
600 		td->td_retval[0] = 0;
601 		return (EINVAL);
602 	}
603 
604 	/*
605 	 * Check for the page alignment.
606 	 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
607 	 */
608 	if (args->addr & PAGE_MASK) {
609 		td->td_retval[0] = 0;
610 		return (EINVAL);
611 	}
612 
613 	args->new_len = round_page(args->new_len);
614 	args->old_len = round_page(args->old_len);
615 
616 	if (args->new_len > args->old_len) {
617 		td->td_retval[0] = 0;
618 		return (ENOMEM);
619 	}
620 
621 	if (args->new_len < args->old_len) {
622 		bsd_args.addr =
623 		    (caddr_t)((uintptr_t)args->addr + args->new_len);
624 		bsd_args.len = args->old_len - args->new_len;
625 		error = sys_munmap(td, &bsd_args);
626 	}
627 
628 	td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
629 	return (error);
630 }
631 
632 #define LINUX_MS_ASYNC       0x0001
633 #define LINUX_MS_INVALIDATE  0x0002
634 #define LINUX_MS_SYNC        0x0004
635 
636 int
637 linux_msync(struct thread *td, struct linux_msync_args *args)
638 {
639 	struct msync_args bsd_args;
640 
641 	bsd_args.addr = (caddr_t)(uintptr_t)args->addr;
642 	bsd_args.len = (uintptr_t)args->len;
643 	bsd_args.flags = args->fl & ~LINUX_MS_SYNC;
644 
645 	return (sys_msync(td, &bsd_args));
646 }
647 
648 int
649 linux_time(struct thread *td, struct linux_time_args *args)
650 {
651 	struct timeval tv;
652 	l_time_t tm;
653 	int error;
654 
655 #ifdef DEBUG
656 	if (ldebug(time))
657 		printf(ARGS(time, "*"));
658 #endif
659 
660 	microtime(&tv);
661 	tm = tv.tv_sec;
662 	if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
663 		return (error);
664 	td->td_retval[0] = tm;
665 	return (0);
666 }
667 
668 struct l_times_argv {
669 	l_clock_t	tms_utime;
670 	l_clock_t	tms_stime;
671 	l_clock_t	tms_cutime;
672 	l_clock_t	tms_cstime;
673 };
674 
675 
676 /*
677  * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
678  * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
679  * auxiliary vector entry.
680  */
681 #define	CLK_TCK		100
682 
683 #define	CONVOTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
684 #define	CONVNTCK(r)	(r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
685 
686 #define	CONVTCK(r)	(linux_kernver(td) >= LINUX_KERNVER_2004000 ?		\
687 			    CONVNTCK(r) : CONVOTCK(r))
688 
689 int
690 linux_times(struct thread *td, struct linux_times_args *args)
691 {
692 	struct timeval tv, utime, stime, cutime, cstime;
693 	struct l_times_argv tms;
694 	struct proc *p;
695 	int error;
696 
697 #ifdef DEBUG
698 	if (ldebug(times))
699 		printf(ARGS(times, "*"));
700 #endif
701 
702 	if (args->buf != NULL) {
703 		p = td->td_proc;
704 		PROC_LOCK(p);
705 		PROC_STATLOCK(p);
706 		calcru(p, &utime, &stime);
707 		PROC_STATUNLOCK(p);
708 		calccru(p, &cutime, &cstime);
709 		PROC_UNLOCK(p);
710 
711 		tms.tms_utime = CONVTCK(utime);
712 		tms.tms_stime = CONVTCK(stime);
713 
714 		tms.tms_cutime = CONVTCK(cutime);
715 		tms.tms_cstime = CONVTCK(cstime);
716 
717 		if ((error = copyout(&tms, args->buf, sizeof(tms))))
718 			return (error);
719 	}
720 
721 	microuptime(&tv);
722 	td->td_retval[0] = (int)CONVTCK(tv);
723 	return (0);
724 }
725 
726 int
727 linux_newuname(struct thread *td, struct linux_newuname_args *args)
728 {
729 	struct l_new_utsname utsname;
730 	char osname[LINUX_MAX_UTSNAME];
731 	char osrelease[LINUX_MAX_UTSNAME];
732 	char *p;
733 
734 #ifdef DEBUG
735 	if (ldebug(newuname))
736 		printf(ARGS(newuname, "*"));
737 #endif
738 
739 	linux_get_osname(td, osname);
740 	linux_get_osrelease(td, osrelease);
741 
742 	bzero(&utsname, sizeof(utsname));
743 	strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
744 	getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
745 	getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
746 	strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
747 	strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
748 	for (p = utsname.version; *p != '\0'; ++p)
749 		if (*p == '\n') {
750 			*p = '\0';
751 			break;
752 		}
753 	strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME);
754 
755 	return (copyout(&utsname, args->buf, sizeof(utsname)));
756 }
757 
758 struct l_utimbuf {
759 	l_time_t l_actime;
760 	l_time_t l_modtime;
761 };
762 
763 int
764 linux_utime(struct thread *td, struct linux_utime_args *args)
765 {
766 	struct timeval tv[2], *tvp;
767 	struct l_utimbuf lut;
768 	char *fname;
769 	int error;
770 
771 	LCONVPATHEXIST(td, args->fname, &fname);
772 
773 #ifdef DEBUG
774 	if (ldebug(utime))
775 		printf(ARGS(utime, "%s, *"), fname);
776 #endif
777 
778 	if (args->times) {
779 		if ((error = copyin(args->times, &lut, sizeof lut))) {
780 			LFREEPATH(fname);
781 			return (error);
782 		}
783 		tv[0].tv_sec = lut.l_actime;
784 		tv[0].tv_usec = 0;
785 		tv[1].tv_sec = lut.l_modtime;
786 		tv[1].tv_usec = 0;
787 		tvp = tv;
788 	} else
789 		tvp = NULL;
790 
791 	error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, tvp,
792 	    UIO_SYSSPACE);
793 	LFREEPATH(fname);
794 	return (error);
795 }
796 
797 int
798 linux_utimes(struct thread *td, struct linux_utimes_args *args)
799 {
800 	l_timeval ltv[2];
801 	struct timeval tv[2], *tvp = NULL;
802 	char *fname;
803 	int error;
804 
805 	LCONVPATHEXIST(td, args->fname, &fname);
806 
807 #ifdef DEBUG
808 	if (ldebug(utimes))
809 		printf(ARGS(utimes, "%s, *"), fname);
810 #endif
811 
812 	if (args->tptr != NULL) {
813 		if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
814 			LFREEPATH(fname);
815 			return (error);
816 		}
817 		tv[0].tv_sec = ltv[0].tv_sec;
818 		tv[0].tv_usec = ltv[0].tv_usec;
819 		tv[1].tv_sec = ltv[1].tv_sec;
820 		tv[1].tv_usec = ltv[1].tv_usec;
821 		tvp = tv;
822 	}
823 
824 	error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE,
825 	    tvp, UIO_SYSSPACE);
826 	LFREEPATH(fname);
827 	return (error);
828 }
829 
830 static int
831 linux_utimensat_nsec_valid(l_long nsec)
832 {
833 
834 	if (nsec == LINUX_UTIME_OMIT || nsec == LINUX_UTIME_NOW)
835 		return (0);
836 	if (nsec >= 0 && nsec <= 999999999)
837 		return (0);
838 	return (1);
839 }
840 
841 int
842 linux_utimensat(struct thread *td, struct linux_utimensat_args *args)
843 {
844 	struct l_timespec l_times[2];
845 	struct timespec times[2], *timesp = NULL;
846 	char *path = NULL;
847 	int error, dfd, flags = 0;
848 
849 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
850 
851 #ifdef DEBUG
852 	if (ldebug(utimensat))
853 		printf(ARGS(utimensat, "%d, *"), dfd);
854 #endif
855 
856 	if (args->flags & ~LINUX_AT_SYMLINK_NOFOLLOW)
857 		return (EINVAL);
858 
859 	if (args->times != NULL) {
860 		error = copyin(args->times, l_times, sizeof(l_times));
861 		if (error != 0)
862 			return (error);
863 
864 		if (linux_utimensat_nsec_valid(l_times[0].tv_nsec) != 0 ||
865 		    linux_utimensat_nsec_valid(l_times[1].tv_nsec) != 0)
866 			return (EINVAL);
867 
868 		times[0].tv_sec = l_times[0].tv_sec;
869 		switch (l_times[0].tv_nsec)
870 		{
871 		case LINUX_UTIME_OMIT:
872 			times[0].tv_nsec = UTIME_OMIT;
873 			break;
874 		case LINUX_UTIME_NOW:
875 			times[0].tv_nsec = UTIME_NOW;
876 			break;
877 		default:
878 			times[0].tv_nsec = l_times[0].tv_nsec;
879 		}
880 
881 		times[1].tv_sec = l_times[1].tv_sec;
882 		switch (l_times[1].tv_nsec)
883 		{
884 		case LINUX_UTIME_OMIT:
885 			times[1].tv_nsec = UTIME_OMIT;
886 			break;
887 		case LINUX_UTIME_NOW:
888 			times[1].tv_nsec = UTIME_NOW;
889 			break;
890 		default:
891 			times[1].tv_nsec = l_times[1].tv_nsec;
892 			break;
893 		}
894 		timesp = times;
895 	}
896 
897 	if (times[0].tv_nsec == UTIME_OMIT && times[1].tv_nsec == UTIME_OMIT)
898 		/* This breaks POSIX, but is what the Linux kernel does
899 		 * _on purpose_ (documented in the man page for utimensat(2)),
900 		 * so we must follow that behaviour. */
901 		return (0);
902 
903 	if (args->pathname != NULL)
904 		LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
905 	else if (args->flags != 0)
906 		return (EINVAL);
907 
908 	if (args->flags & LINUX_AT_SYMLINK_NOFOLLOW)
909 		flags |= AT_SYMLINK_NOFOLLOW;
910 
911 	if (path == NULL)
912 		error = kern_futimens(td, dfd, timesp, UIO_SYSSPACE);
913 	else {
914 		error = kern_utimensat(td, dfd, path, UIO_SYSSPACE, timesp,
915 	    		UIO_SYSSPACE, flags);
916 		LFREEPATH(path);
917 	}
918 
919 	return (error);
920 }
921 
922 int
923 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
924 {
925 	l_timeval ltv[2];
926 	struct timeval tv[2], *tvp = NULL;
927 	char *fname;
928 	int error, dfd;
929 
930 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
931 	LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
932 
933 #ifdef DEBUG
934 	if (ldebug(futimesat))
935 		printf(ARGS(futimesat, "%s, *"), fname);
936 #endif
937 
938 	if (args->utimes != NULL) {
939 		if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
940 			LFREEPATH(fname);
941 			return (error);
942 		}
943 		tv[0].tv_sec = ltv[0].tv_sec;
944 		tv[0].tv_usec = ltv[0].tv_usec;
945 		tv[1].tv_sec = ltv[1].tv_sec;
946 		tv[1].tv_usec = ltv[1].tv_usec;
947 		tvp = tv;
948 	}
949 
950 	error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
951 	LFREEPATH(fname);
952 	return (error);
953 }
954 
955 int
956 linux_common_wait(struct thread *td, int pid, int *status,
957     int options, struct rusage *ru)
958 {
959 	int error, tmpstat;
960 
961 	error = kern_wait(td, pid, &tmpstat, options, ru);
962 	if (error)
963 		return (error);
964 
965 	if (status) {
966 		tmpstat &= 0xffff;
967 		if (WIFSIGNALED(tmpstat))
968 			tmpstat = (tmpstat & 0xffffff80) |
969 			    bsd_to_linux_signal(WTERMSIG(tmpstat));
970 		else if (WIFSTOPPED(tmpstat))
971 			tmpstat = (tmpstat & 0xffff00ff) |
972 			    (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8);
973 		else if (WIFCONTINUED(tmpstat))
974 			tmpstat = 0xffff;
975 		error = copyout(&tmpstat, status, sizeof(int));
976 	}
977 
978 	return (error);
979 }
980 
981 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
982 int
983 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
984 {
985 	struct linux_wait4_args wait4_args;
986 
987 #ifdef DEBUG
988 	if (ldebug(waitpid))
989 		printf(ARGS(waitpid, "%d, %p, %d"),
990 		    args->pid, (void *)args->status, args->options);
991 #endif
992 
993 	wait4_args.pid = args->pid;
994 	wait4_args.status = args->status;
995 	wait4_args.options = args->options;
996 	wait4_args.rusage = NULL;
997 
998 	return (linux_wait4(td, &wait4_args));
999 }
1000 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1001 
1002 int
1003 linux_wait4(struct thread *td, struct linux_wait4_args *args)
1004 {
1005 	int error, options;
1006 	struct rusage ru, *rup;
1007 
1008 #ifdef DEBUG
1009 	if (ldebug(wait4))
1010 		printf(ARGS(wait4, "%d, %p, %d, %p"),
1011 		    args->pid, (void *)args->status, args->options,
1012 		    (void *)args->rusage);
1013 #endif
1014 	if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG |
1015 	    LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
1016 		return (EINVAL);
1017 
1018 	options = WEXITED;
1019 	linux_to_bsd_waitopts(args->options, &options);
1020 
1021 	if (args->rusage != NULL)
1022 		rup = &ru;
1023 	else
1024 		rup = NULL;
1025 	error = linux_common_wait(td, args->pid, args->status, options, rup);
1026 	if (error != 0)
1027 		return (error);
1028 	if (args->rusage != NULL)
1029 		error = linux_copyout_rusage(&ru, args->rusage);
1030 	return (error);
1031 }
1032 
1033 int
1034 linux_waitid(struct thread *td, struct linux_waitid_args *args)
1035 {
1036 	int status, options, sig;
1037 	struct __wrusage wru;
1038 	siginfo_t siginfo;
1039 	l_siginfo_t lsi;
1040 	idtype_t idtype;
1041 	struct proc *p;
1042 	int error;
1043 
1044 	options = 0;
1045 	linux_to_bsd_waitopts(args->options, &options);
1046 
1047 	if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED))
1048 		return (EINVAL);
1049 	if (!(options & (WEXITED | WUNTRACED | WCONTINUED)))
1050 		return (EINVAL);
1051 
1052 	switch (args->idtype) {
1053 	case LINUX_P_ALL:
1054 		idtype = P_ALL;
1055 		break;
1056 	case LINUX_P_PID:
1057 		if (args->id <= 0)
1058 			return (EINVAL);
1059 		idtype = P_PID;
1060 		break;
1061 	case LINUX_P_PGID:
1062 		if (args->id <= 0)
1063 			return (EINVAL);
1064 		idtype = P_PGID;
1065 		break;
1066 	default:
1067 		return (EINVAL);
1068 	}
1069 
1070 	error = kern_wait6(td, idtype, args->id, &status, options,
1071 	    &wru, &siginfo);
1072 	if (error != 0)
1073 		return (error);
1074 	if (args->rusage != NULL) {
1075 		error = linux_copyout_rusage(&wru.wru_children,
1076 		    args->rusage);
1077 		if (error != 0)
1078 			return (error);
1079 	}
1080 	if (args->info != NULL) {
1081 		p = td->td_proc;
1082 		if (td->td_retval[0] == 0)
1083 			bzero(&lsi, sizeof(lsi));
1084 		else {
1085 			sig = bsd_to_linux_signal(siginfo.si_signo);
1086 			siginfo_to_lsiginfo(&siginfo, &lsi, sig);
1087 		}
1088 		error = copyout(&lsi, args->info, sizeof(lsi));
1089 	}
1090 	td->td_retval[0] = 0;
1091 
1092 	return (error);
1093 }
1094 
1095 int
1096 linux_mknod(struct thread *td, struct linux_mknod_args *args)
1097 {
1098 	char *path;
1099 	int error;
1100 
1101 	LCONVPATHCREAT(td, args->path, &path);
1102 
1103 #ifdef DEBUG
1104 	if (ldebug(mknod))
1105 		printf(ARGS(mknod, "%s, %d, %ju"), path, args->mode,
1106 		    (uintmax_t)args->dev);
1107 #endif
1108 
1109 	switch (args->mode & S_IFMT) {
1110 	case S_IFIFO:
1111 	case S_IFSOCK:
1112 		error = kern_mkfifoat(td, AT_FDCWD, path, UIO_SYSSPACE,
1113 		    args->mode);
1114 		break;
1115 
1116 	case S_IFCHR:
1117 	case S_IFBLK:
1118 		error = kern_mknodat(td, AT_FDCWD, path, UIO_SYSSPACE,
1119 		    args->mode, args->dev);
1120 		break;
1121 
1122 	case S_IFDIR:
1123 		error = EPERM;
1124 		break;
1125 
1126 	case 0:
1127 		args->mode |= S_IFREG;
1128 		/* FALLTHROUGH */
1129 	case S_IFREG:
1130 		error = kern_openat(td, AT_FDCWD, path, UIO_SYSSPACE,
1131 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1132 		if (error == 0)
1133 			kern_close(td, td->td_retval[0]);
1134 		break;
1135 
1136 	default:
1137 		error = EINVAL;
1138 		break;
1139 	}
1140 	LFREEPATH(path);
1141 	return (error);
1142 }
1143 
1144 int
1145 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
1146 {
1147 	char *path;
1148 	int error, dfd;
1149 
1150 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
1151 	LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
1152 
1153 #ifdef DEBUG
1154 	if (ldebug(mknodat))
1155 		printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev);
1156 #endif
1157 
1158 	switch (args->mode & S_IFMT) {
1159 	case S_IFIFO:
1160 	case S_IFSOCK:
1161 		error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
1162 		break;
1163 
1164 	case S_IFCHR:
1165 	case S_IFBLK:
1166 		error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
1167 		    args->dev);
1168 		break;
1169 
1170 	case S_IFDIR:
1171 		error = EPERM;
1172 		break;
1173 
1174 	case 0:
1175 		args->mode |= S_IFREG;
1176 		/* FALLTHROUGH */
1177 	case S_IFREG:
1178 		error = kern_openat(td, dfd, path, UIO_SYSSPACE,
1179 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1180 		if (error == 0)
1181 			kern_close(td, td->td_retval[0]);
1182 		break;
1183 
1184 	default:
1185 		error = EINVAL;
1186 		break;
1187 	}
1188 	LFREEPATH(path);
1189 	return (error);
1190 }
1191 
1192 /*
1193  * UGH! This is just about the dumbest idea I've ever heard!!
1194  */
1195 int
1196 linux_personality(struct thread *td, struct linux_personality_args *args)
1197 {
1198 #ifdef DEBUG
1199 	if (ldebug(personality))
1200 		printf(ARGS(personality, "%lu"), (unsigned long)args->per);
1201 #endif
1202 	if (args->per != 0)
1203 		return (EINVAL);
1204 
1205 	/* Yes Jim, it's still a Linux... */
1206 	td->td_retval[0] = 0;
1207 	return (0);
1208 }
1209 
1210 struct l_itimerval {
1211 	l_timeval it_interval;
1212 	l_timeval it_value;
1213 };
1214 
1215 #define	B2L_ITIMERVAL(bip, lip) 					\
1216 	(bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec;		\
1217 	(bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec;	\
1218 	(bip)->it_value.tv_sec = (lip)->it_value.tv_sec;		\
1219 	(bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1220 
1221 int
1222 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1223 {
1224 	int error;
1225 	struct l_itimerval ls;
1226 	struct itimerval aitv, oitv;
1227 
1228 #ifdef DEBUG
1229 	if (ldebug(setitimer))
1230 		printf(ARGS(setitimer, "%p, %p"),
1231 		    (void *)uap->itv, (void *)uap->oitv);
1232 #endif
1233 
1234 	if (uap->itv == NULL) {
1235 		uap->itv = uap->oitv;
1236 		return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1237 	}
1238 
1239 	error = copyin(uap->itv, &ls, sizeof(ls));
1240 	if (error != 0)
1241 		return (error);
1242 	B2L_ITIMERVAL(&aitv, &ls);
1243 #ifdef DEBUG
1244 	if (ldebug(setitimer)) {
1245 		printf("setitimer: value: sec: %jd, usec: %ld\n",
1246 		    (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1247 		printf("setitimer: interval: sec: %jd, usec: %ld\n",
1248 		    (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1249 	}
1250 #endif
1251 	error = kern_setitimer(td, uap->which, &aitv, &oitv);
1252 	if (error != 0 || uap->oitv == NULL)
1253 		return (error);
1254 	B2L_ITIMERVAL(&ls, &oitv);
1255 
1256 	return (copyout(&ls, uap->oitv, sizeof(ls)));
1257 }
1258 
1259 int
1260 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1261 {
1262 	int error;
1263 	struct l_itimerval ls;
1264 	struct itimerval aitv;
1265 
1266 #ifdef DEBUG
1267 	if (ldebug(getitimer))
1268 		printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1269 #endif
1270 	error = kern_getitimer(td, uap->which, &aitv);
1271 	if (error != 0)
1272 		return (error);
1273 	B2L_ITIMERVAL(&ls, &aitv);
1274 	return (copyout(&ls, uap->itv, sizeof(ls)));
1275 }
1276 
1277 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1278 int
1279 linux_nice(struct thread *td, struct linux_nice_args *args)
1280 {
1281 	struct setpriority_args bsd_args;
1282 
1283 	bsd_args.which = PRIO_PROCESS;
1284 	bsd_args.who = 0;		/* current process */
1285 	bsd_args.prio = args->inc;
1286 	return (sys_setpriority(td, &bsd_args));
1287 }
1288 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1289 
1290 int
1291 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1292 {
1293 	struct ucred *newcred, *oldcred;
1294 	l_gid_t *linux_gidset;
1295 	gid_t *bsd_gidset;
1296 	int ngrp, error;
1297 	struct proc *p;
1298 
1299 	ngrp = args->gidsetsize;
1300 	if (ngrp < 0 || ngrp >= ngroups_max + 1)
1301 		return (EINVAL);
1302 	linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK);
1303 	error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1304 	if (error)
1305 		goto out;
1306 	newcred = crget();
1307 	p = td->td_proc;
1308 	PROC_LOCK(p);
1309 	oldcred = crcopysafe(p, newcred);
1310 
1311 	/*
1312 	 * cr_groups[0] holds egid. Setting the whole set from
1313 	 * the supplied set will cause egid to be changed too.
1314 	 * Keep cr_groups[0] unchanged to prevent that.
1315 	 */
1316 
1317 	if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) {
1318 		PROC_UNLOCK(p);
1319 		crfree(newcred);
1320 		goto out;
1321 	}
1322 
1323 	if (ngrp > 0) {
1324 		newcred->cr_ngroups = ngrp + 1;
1325 
1326 		bsd_gidset = newcred->cr_groups;
1327 		ngrp--;
1328 		while (ngrp >= 0) {
1329 			bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1330 			ngrp--;
1331 		}
1332 	} else
1333 		newcred->cr_ngroups = 1;
1334 
1335 	setsugid(p);
1336 	proc_set_cred(p, newcred);
1337 	PROC_UNLOCK(p);
1338 	crfree(oldcred);
1339 	error = 0;
1340 out:
1341 	free(linux_gidset, M_LINUX);
1342 	return (error);
1343 }
1344 
1345 int
1346 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1347 {
1348 	struct ucred *cred;
1349 	l_gid_t *linux_gidset;
1350 	gid_t *bsd_gidset;
1351 	int bsd_gidsetsz, ngrp, error;
1352 
1353 	cred = td->td_ucred;
1354 	bsd_gidset = cred->cr_groups;
1355 	bsd_gidsetsz = cred->cr_ngroups - 1;
1356 
1357 	/*
1358 	 * cr_groups[0] holds egid. Returning the whole set
1359 	 * here will cause a duplicate. Exclude cr_groups[0]
1360 	 * to prevent that.
1361 	 */
1362 
1363 	if ((ngrp = args->gidsetsize) == 0) {
1364 		td->td_retval[0] = bsd_gidsetsz;
1365 		return (0);
1366 	}
1367 
1368 	if (ngrp < bsd_gidsetsz)
1369 		return (EINVAL);
1370 
1371 	ngrp = 0;
1372 	linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1373 	    M_LINUX, M_WAITOK);
1374 	while (ngrp < bsd_gidsetsz) {
1375 		linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1376 		ngrp++;
1377 	}
1378 
1379 	error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1380 	free(linux_gidset, M_LINUX);
1381 	if (error)
1382 		return (error);
1383 
1384 	td->td_retval[0] = ngrp;
1385 	return (0);
1386 }
1387 
1388 int
1389 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1390 {
1391 	struct rlimit bsd_rlim;
1392 	struct l_rlimit rlim;
1393 	u_int which;
1394 	int error;
1395 
1396 #ifdef DEBUG
1397 	if (ldebug(setrlimit))
1398 		printf(ARGS(setrlimit, "%d, %p"),
1399 		    args->resource, (void *)args->rlim);
1400 #endif
1401 
1402 	if (args->resource >= LINUX_RLIM_NLIMITS)
1403 		return (EINVAL);
1404 
1405 	which = linux_to_bsd_resource[args->resource];
1406 	if (which == -1)
1407 		return (EINVAL);
1408 
1409 	error = copyin(args->rlim, &rlim, sizeof(rlim));
1410 	if (error)
1411 		return (error);
1412 
1413 	bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1414 	bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1415 	return (kern_setrlimit(td, which, &bsd_rlim));
1416 }
1417 
1418 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1419 int
1420 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1421 {
1422 	struct l_rlimit rlim;
1423 	struct rlimit bsd_rlim;
1424 	u_int which;
1425 
1426 #ifdef DEBUG
1427 	if (ldebug(old_getrlimit))
1428 		printf(ARGS(old_getrlimit, "%d, %p"),
1429 		    args->resource, (void *)args->rlim);
1430 #endif
1431 
1432 	if (args->resource >= LINUX_RLIM_NLIMITS)
1433 		return (EINVAL);
1434 
1435 	which = linux_to_bsd_resource[args->resource];
1436 	if (which == -1)
1437 		return (EINVAL);
1438 
1439 	lim_rlimit(td, which, &bsd_rlim);
1440 
1441 #ifdef COMPAT_LINUX32
1442 	rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1443 	if (rlim.rlim_cur == UINT_MAX)
1444 		rlim.rlim_cur = INT_MAX;
1445 	rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1446 	if (rlim.rlim_max == UINT_MAX)
1447 		rlim.rlim_max = INT_MAX;
1448 #else
1449 	rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1450 	if (rlim.rlim_cur == ULONG_MAX)
1451 		rlim.rlim_cur = LONG_MAX;
1452 	rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1453 	if (rlim.rlim_max == ULONG_MAX)
1454 		rlim.rlim_max = LONG_MAX;
1455 #endif
1456 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1457 }
1458 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1459 
1460 int
1461 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1462 {
1463 	struct l_rlimit rlim;
1464 	struct rlimit bsd_rlim;
1465 	u_int which;
1466 
1467 #ifdef DEBUG
1468 	if (ldebug(getrlimit))
1469 		printf(ARGS(getrlimit, "%d, %p"),
1470 		    args->resource, (void *)args->rlim);
1471 #endif
1472 
1473 	if (args->resource >= LINUX_RLIM_NLIMITS)
1474 		return (EINVAL);
1475 
1476 	which = linux_to_bsd_resource[args->resource];
1477 	if (which == -1)
1478 		return (EINVAL);
1479 
1480 	lim_rlimit(td, which, &bsd_rlim);
1481 
1482 	rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1483 	rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1484 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1485 }
1486 
1487 int
1488 linux_sched_setscheduler(struct thread *td,
1489     struct linux_sched_setscheduler_args *args)
1490 {
1491 	struct sched_param sched_param;
1492 	struct thread *tdt;
1493 	int error, policy;
1494 
1495 #ifdef DEBUG
1496 	if (ldebug(sched_setscheduler))
1497 		printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1498 		    args->pid, args->policy, (const void *)args->param);
1499 #endif
1500 
1501 	switch (args->policy) {
1502 	case LINUX_SCHED_OTHER:
1503 		policy = SCHED_OTHER;
1504 		break;
1505 	case LINUX_SCHED_FIFO:
1506 		policy = SCHED_FIFO;
1507 		break;
1508 	case LINUX_SCHED_RR:
1509 		policy = SCHED_RR;
1510 		break;
1511 	default:
1512 		return (EINVAL);
1513 	}
1514 
1515 	error = copyin(args->param, &sched_param, sizeof(sched_param));
1516 	if (error)
1517 		return (error);
1518 
1519 	tdt = linux_tdfind(td, args->pid, -1);
1520 	if (tdt == NULL)
1521 		return (ESRCH);
1522 
1523 	error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1524 	PROC_UNLOCK(tdt->td_proc);
1525 	return (error);
1526 }
1527 
1528 int
1529 linux_sched_getscheduler(struct thread *td,
1530     struct linux_sched_getscheduler_args *args)
1531 {
1532 	struct thread *tdt;
1533 	int error, policy;
1534 
1535 #ifdef DEBUG
1536 	if (ldebug(sched_getscheduler))
1537 		printf(ARGS(sched_getscheduler, "%d"), args->pid);
1538 #endif
1539 
1540 	tdt = linux_tdfind(td, args->pid, -1);
1541 	if (tdt == NULL)
1542 		return (ESRCH);
1543 
1544 	error = kern_sched_getscheduler(td, tdt, &policy);
1545 	PROC_UNLOCK(tdt->td_proc);
1546 
1547 	switch (policy) {
1548 	case SCHED_OTHER:
1549 		td->td_retval[0] = LINUX_SCHED_OTHER;
1550 		break;
1551 	case SCHED_FIFO:
1552 		td->td_retval[0] = LINUX_SCHED_FIFO;
1553 		break;
1554 	case SCHED_RR:
1555 		td->td_retval[0] = LINUX_SCHED_RR;
1556 		break;
1557 	}
1558 	return (error);
1559 }
1560 
1561 int
1562 linux_sched_get_priority_max(struct thread *td,
1563     struct linux_sched_get_priority_max_args *args)
1564 {
1565 	struct sched_get_priority_max_args bsd;
1566 
1567 #ifdef DEBUG
1568 	if (ldebug(sched_get_priority_max))
1569 		printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1570 #endif
1571 
1572 	switch (args->policy) {
1573 	case LINUX_SCHED_OTHER:
1574 		bsd.policy = SCHED_OTHER;
1575 		break;
1576 	case LINUX_SCHED_FIFO:
1577 		bsd.policy = SCHED_FIFO;
1578 		break;
1579 	case LINUX_SCHED_RR:
1580 		bsd.policy = SCHED_RR;
1581 		break;
1582 	default:
1583 		return (EINVAL);
1584 	}
1585 	return (sys_sched_get_priority_max(td, &bsd));
1586 }
1587 
1588 int
1589 linux_sched_get_priority_min(struct thread *td,
1590     struct linux_sched_get_priority_min_args *args)
1591 {
1592 	struct sched_get_priority_min_args bsd;
1593 
1594 #ifdef DEBUG
1595 	if (ldebug(sched_get_priority_min))
1596 		printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1597 #endif
1598 
1599 	switch (args->policy) {
1600 	case LINUX_SCHED_OTHER:
1601 		bsd.policy = SCHED_OTHER;
1602 		break;
1603 	case LINUX_SCHED_FIFO:
1604 		bsd.policy = SCHED_FIFO;
1605 		break;
1606 	case LINUX_SCHED_RR:
1607 		bsd.policy = SCHED_RR;
1608 		break;
1609 	default:
1610 		return (EINVAL);
1611 	}
1612 	return (sys_sched_get_priority_min(td, &bsd));
1613 }
1614 
1615 #define REBOOT_CAD_ON	0x89abcdef
1616 #define REBOOT_CAD_OFF	0
1617 #define REBOOT_HALT	0xcdef0123
1618 #define REBOOT_RESTART	0x01234567
1619 #define REBOOT_RESTART2	0xA1B2C3D4
1620 #define REBOOT_POWEROFF	0x4321FEDC
1621 #define REBOOT_MAGIC1	0xfee1dead
1622 #define REBOOT_MAGIC2	0x28121969
1623 #define REBOOT_MAGIC2A	0x05121996
1624 #define REBOOT_MAGIC2B	0x16041998
1625 
1626 int
1627 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1628 {
1629 	struct reboot_args bsd_args;
1630 
1631 #ifdef DEBUG
1632 	if (ldebug(reboot))
1633 		printf(ARGS(reboot, "0x%x"), args->cmd);
1634 #endif
1635 
1636 	if (args->magic1 != REBOOT_MAGIC1)
1637 		return (EINVAL);
1638 
1639 	switch (args->magic2) {
1640 	case REBOOT_MAGIC2:
1641 	case REBOOT_MAGIC2A:
1642 	case REBOOT_MAGIC2B:
1643 		break;
1644 	default:
1645 		return (EINVAL);
1646 	}
1647 
1648 	switch (args->cmd) {
1649 	case REBOOT_CAD_ON:
1650 	case REBOOT_CAD_OFF:
1651 		return (priv_check(td, PRIV_REBOOT));
1652 	case REBOOT_HALT:
1653 		bsd_args.opt = RB_HALT;
1654 		break;
1655 	case REBOOT_RESTART:
1656 	case REBOOT_RESTART2:
1657 		bsd_args.opt = 0;
1658 		break;
1659 	case REBOOT_POWEROFF:
1660 		bsd_args.opt = RB_POWEROFF;
1661 		break;
1662 	default:
1663 		return (EINVAL);
1664 	}
1665 	return (sys_reboot(td, &bsd_args));
1666 }
1667 
1668 
1669 /*
1670  * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1671  * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
1672  * are assumed to be preserved. The following lightweight syscalls fixes
1673  * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
1674  *
1675  * linux_getpid() - MP SAFE
1676  * linux_getgid() - MP SAFE
1677  * linux_getuid() - MP SAFE
1678  */
1679 
1680 int
1681 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1682 {
1683 
1684 #ifdef DEBUG
1685 	if (ldebug(getpid))
1686 		printf(ARGS(getpid, ""));
1687 #endif
1688 	td->td_retval[0] = td->td_proc->p_pid;
1689 
1690 	return (0);
1691 }
1692 
1693 int
1694 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1695 {
1696 	struct linux_emuldata *em;
1697 
1698 #ifdef DEBUG
1699 	if (ldebug(gettid))
1700 		printf(ARGS(gettid, ""));
1701 #endif
1702 
1703 	em = em_find(td);
1704 	KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1705 
1706 	td->td_retval[0] = em->em_tid;
1707 
1708 	return (0);
1709 }
1710 
1711 
1712 int
1713 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1714 {
1715 
1716 #ifdef DEBUG
1717 	if (ldebug(getppid))
1718 		printf(ARGS(getppid, ""));
1719 #endif
1720 
1721 	PROC_LOCK(td->td_proc);
1722 	td->td_retval[0] = td->td_proc->p_pptr->p_pid;
1723 	PROC_UNLOCK(td->td_proc);
1724 	return (0);
1725 }
1726 
1727 int
1728 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1729 {
1730 
1731 #ifdef DEBUG
1732 	if (ldebug(getgid))
1733 		printf(ARGS(getgid, ""));
1734 #endif
1735 
1736 	td->td_retval[0] = td->td_ucred->cr_rgid;
1737 	return (0);
1738 }
1739 
1740 int
1741 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1742 {
1743 
1744 #ifdef DEBUG
1745 	if (ldebug(getuid))
1746 		printf(ARGS(getuid, ""));
1747 #endif
1748 
1749 	td->td_retval[0] = td->td_ucred->cr_ruid;
1750 	return (0);
1751 }
1752 
1753 
1754 int
1755 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1756 {
1757 	struct getsid_args bsd;
1758 
1759 #ifdef DEBUG
1760 	if (ldebug(getsid))
1761 		printf(ARGS(getsid, "%i"), args->pid);
1762 #endif
1763 
1764 	bsd.pid = args->pid;
1765 	return (sys_getsid(td, &bsd));
1766 }
1767 
1768 int
1769 linux_nosys(struct thread *td, struct nosys_args *ignore)
1770 {
1771 
1772 	return (ENOSYS);
1773 }
1774 
1775 int
1776 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1777 {
1778 	struct getpriority_args bsd_args;
1779 	int error;
1780 
1781 #ifdef DEBUG
1782 	if (ldebug(getpriority))
1783 		printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
1784 #endif
1785 
1786 	bsd_args.which = args->which;
1787 	bsd_args.who = args->who;
1788 	error = sys_getpriority(td, &bsd_args);
1789 	td->td_retval[0] = 20 - td->td_retval[0];
1790 	return (error);
1791 }
1792 
1793 int
1794 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1795 {
1796 	int name[2];
1797 
1798 #ifdef DEBUG
1799 	if (ldebug(sethostname))
1800 		printf(ARGS(sethostname, "*, %i"), args->len);
1801 #endif
1802 
1803 	name[0] = CTL_KERN;
1804 	name[1] = KERN_HOSTNAME;
1805 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1806 	    args->len, 0, 0));
1807 }
1808 
1809 int
1810 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1811 {
1812 	int name[2];
1813 
1814 #ifdef DEBUG
1815 	if (ldebug(setdomainname))
1816 		printf(ARGS(setdomainname, "*, %i"), args->len);
1817 #endif
1818 
1819 	name[0] = CTL_KERN;
1820 	name[1] = KERN_NISDOMAINNAME;
1821 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1822 	    args->len, 0, 0));
1823 }
1824 
1825 int
1826 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1827 {
1828 
1829 #ifdef DEBUG
1830 	if (ldebug(exit_group))
1831 		printf(ARGS(exit_group, "%i"), args->error_code);
1832 #endif
1833 
1834 	LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1835 	    args->error_code);
1836 
1837 	/*
1838 	 * XXX: we should send a signal to the parent if
1839 	 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1840 	 * as it doesnt occur often.
1841 	 */
1842 	exit1(td, args->error_code, 0);
1843 		/* NOTREACHED */
1844 }
1845 
1846 #define _LINUX_CAPABILITY_VERSION  0x19980330
1847 
1848 struct l_user_cap_header {
1849 	l_int	version;
1850 	l_int	pid;
1851 };
1852 
1853 struct l_user_cap_data {
1854 	l_int	effective;
1855 	l_int	permitted;
1856 	l_int	inheritable;
1857 };
1858 
1859 int
1860 linux_capget(struct thread *td, struct linux_capget_args *args)
1861 {
1862 	struct l_user_cap_header luch;
1863 	struct l_user_cap_data lucd;
1864 	int error;
1865 
1866 	if (args->hdrp == NULL)
1867 		return (EFAULT);
1868 
1869 	error = copyin(args->hdrp, &luch, sizeof(luch));
1870 	if (error != 0)
1871 		return (error);
1872 
1873 	if (luch.version != _LINUX_CAPABILITY_VERSION) {
1874 		luch.version = _LINUX_CAPABILITY_VERSION;
1875 		error = copyout(&luch, args->hdrp, sizeof(luch));
1876 		if (error)
1877 			return (error);
1878 		return (EINVAL);
1879 	}
1880 
1881 	if (luch.pid)
1882 		return (EPERM);
1883 
1884 	if (args->datap) {
1885 		/*
1886 		 * The current implementation doesn't support setting
1887 		 * a capability (it's essentially a stub) so indicate
1888 		 * that no capabilities are currently set or available
1889 		 * to request.
1890 		 */
1891 		bzero (&lucd, sizeof(lucd));
1892 		error = copyout(&lucd, args->datap, sizeof(lucd));
1893 	}
1894 
1895 	return (error);
1896 }
1897 
1898 int
1899 linux_capset(struct thread *td, struct linux_capset_args *args)
1900 {
1901 	struct l_user_cap_header luch;
1902 	struct l_user_cap_data lucd;
1903 	int error;
1904 
1905 	if (args->hdrp == NULL || args->datap == NULL)
1906 		return (EFAULT);
1907 
1908 	error = copyin(args->hdrp, &luch, sizeof(luch));
1909 	if (error != 0)
1910 		return (error);
1911 
1912 	if (luch.version != _LINUX_CAPABILITY_VERSION) {
1913 		luch.version = _LINUX_CAPABILITY_VERSION;
1914 		error = copyout(&luch, args->hdrp, sizeof(luch));
1915 		if (error)
1916 			return (error);
1917 		return (EINVAL);
1918 	}
1919 
1920 	if (luch.pid)
1921 		return (EPERM);
1922 
1923 	error = copyin(args->datap, &lucd, sizeof(lucd));
1924 	if (error != 0)
1925 		return (error);
1926 
1927 	/* We currently don't support setting any capabilities. */
1928 	if (lucd.effective || lucd.permitted || lucd.inheritable) {
1929 		linux_msg(td,
1930 			  "capset effective=0x%x, permitted=0x%x, "
1931 			  "inheritable=0x%x is not implemented",
1932 			  (int)lucd.effective, (int)lucd.permitted,
1933 			  (int)lucd.inheritable);
1934 		return (EPERM);
1935 	}
1936 
1937 	return (0);
1938 }
1939 
1940 int
1941 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1942 {
1943 	int error = 0, max_size;
1944 	struct proc *p = td->td_proc;
1945 	char comm[LINUX_MAX_COMM_LEN];
1946 	struct linux_emuldata *em;
1947 	int pdeath_signal;
1948 
1949 #ifdef DEBUG
1950 	if (ldebug(prctl))
1951 		printf(ARGS(prctl, "%d, %ju, %ju, %ju, %ju"), args->option,
1952 		    (uintmax_t)args->arg2, (uintmax_t)args->arg3,
1953 		    (uintmax_t)args->arg4, (uintmax_t)args->arg5);
1954 #endif
1955 
1956 	switch (args->option) {
1957 	case LINUX_PR_SET_PDEATHSIG:
1958 		if (!LINUX_SIG_VALID(args->arg2))
1959 			return (EINVAL);
1960 		em = em_find(td);
1961 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1962 		em->pdeath_signal = args->arg2;
1963 		break;
1964 	case LINUX_PR_GET_PDEATHSIG:
1965 		em = em_find(td);
1966 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1967 		pdeath_signal = em->pdeath_signal;
1968 		error = copyout(&pdeath_signal,
1969 		    (void *)(register_t)args->arg2,
1970 		    sizeof(pdeath_signal));
1971 		break;
1972 	case LINUX_PR_GET_KEEPCAPS:
1973 		/*
1974 		 * Indicate that we always clear the effective and
1975 		 * permitted capability sets when the user id becomes
1976 		 * non-zero (actually the capability sets are simply
1977 		 * always zero in the current implementation).
1978 		 */
1979 		td->td_retval[0] = 0;
1980 		break;
1981 	case LINUX_PR_SET_KEEPCAPS:
1982 		/*
1983 		 * Ignore requests to keep the effective and permitted
1984 		 * capability sets when the user id becomes non-zero.
1985 		 */
1986 		break;
1987 	case LINUX_PR_SET_NAME:
1988 		/*
1989 		 * To be on the safe side we need to make sure to not
1990 		 * overflow the size a linux program expects. We already
1991 		 * do this here in the copyin, so that we don't need to
1992 		 * check on copyout.
1993 		 */
1994 		max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1995 		error = copyinstr((void *)(register_t)args->arg2, comm,
1996 		    max_size, NULL);
1997 
1998 		/* Linux silently truncates the name if it is too long. */
1999 		if (error == ENAMETOOLONG) {
2000 			/*
2001 			 * XXX: copyinstr() isn't documented to populate the
2002 			 * array completely, so do a copyin() to be on the
2003 			 * safe side. This should be changed in case
2004 			 * copyinstr() is changed to guarantee this.
2005 			 */
2006 			error = copyin((void *)(register_t)args->arg2, comm,
2007 			    max_size - 1);
2008 			comm[max_size - 1] = '\0';
2009 		}
2010 		if (error)
2011 			return (error);
2012 
2013 		PROC_LOCK(p);
2014 		strlcpy(p->p_comm, comm, sizeof(p->p_comm));
2015 		PROC_UNLOCK(p);
2016 		break;
2017 	case LINUX_PR_GET_NAME:
2018 		PROC_LOCK(p);
2019 		strlcpy(comm, p->p_comm, sizeof(comm));
2020 		PROC_UNLOCK(p);
2021 		error = copyout(comm, (void *)(register_t)args->arg2,
2022 		    strlen(comm) + 1);
2023 		break;
2024 	default:
2025 		error = EINVAL;
2026 		break;
2027 	}
2028 
2029 	return (error);
2030 }
2031 
2032 int
2033 linux_sched_setparam(struct thread *td,
2034     struct linux_sched_setparam_args *uap)
2035 {
2036 	struct sched_param sched_param;
2037 	struct thread *tdt;
2038 	int error;
2039 
2040 #ifdef DEBUG
2041 	if (ldebug(sched_setparam))
2042 		printf(ARGS(sched_setparam, "%d, *"), uap->pid);
2043 #endif
2044 
2045 	error = copyin(uap->param, &sched_param, sizeof(sched_param));
2046 	if (error)
2047 		return (error);
2048 
2049 	tdt = linux_tdfind(td, uap->pid, -1);
2050 	if (tdt == NULL)
2051 		return (ESRCH);
2052 
2053 	error = kern_sched_setparam(td, tdt, &sched_param);
2054 	PROC_UNLOCK(tdt->td_proc);
2055 	return (error);
2056 }
2057 
2058 int
2059 linux_sched_getparam(struct thread *td,
2060     struct linux_sched_getparam_args *uap)
2061 {
2062 	struct sched_param sched_param;
2063 	struct thread *tdt;
2064 	int error;
2065 
2066 #ifdef DEBUG
2067 	if (ldebug(sched_getparam))
2068 		printf(ARGS(sched_getparam, "%d, *"), uap->pid);
2069 #endif
2070 
2071 	tdt = linux_tdfind(td, uap->pid, -1);
2072 	if (tdt == NULL)
2073 		return (ESRCH);
2074 
2075 	error = kern_sched_getparam(td, tdt, &sched_param);
2076 	PROC_UNLOCK(tdt->td_proc);
2077 	if (error == 0)
2078 		error = copyout(&sched_param, uap->param,
2079 		    sizeof(sched_param));
2080 	return (error);
2081 }
2082 
2083 /*
2084  * Get affinity of a process.
2085  */
2086 int
2087 linux_sched_getaffinity(struct thread *td,
2088     struct linux_sched_getaffinity_args *args)
2089 {
2090 	int error;
2091 	struct thread *tdt;
2092 	struct cpuset_getaffinity_args cga;
2093 
2094 #ifdef DEBUG
2095 	if (ldebug(sched_getaffinity))
2096 		printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid,
2097 		    args->len);
2098 #endif
2099 	if (args->len < sizeof(cpuset_t))
2100 		return (EINVAL);
2101 
2102 	tdt = linux_tdfind(td, args->pid, -1);
2103 	if (tdt == NULL)
2104 		return (ESRCH);
2105 
2106 	PROC_UNLOCK(tdt->td_proc);
2107 	cga.level = CPU_LEVEL_WHICH;
2108 	cga.which = CPU_WHICH_TID;
2109 	cga.id = tdt->td_tid;
2110 	cga.cpusetsize = sizeof(cpuset_t);
2111 	cga.mask = (cpuset_t *) args->user_mask_ptr;
2112 
2113 	if ((error = sys_cpuset_getaffinity(td, &cga)) == 0)
2114 		td->td_retval[0] = sizeof(cpuset_t);
2115 
2116 	return (error);
2117 }
2118 
2119 /*
2120  *  Set affinity of a process.
2121  */
2122 int
2123 linux_sched_setaffinity(struct thread *td,
2124     struct linux_sched_setaffinity_args *args)
2125 {
2126 	struct cpuset_setaffinity_args csa;
2127 	struct thread *tdt;
2128 
2129 #ifdef DEBUG
2130 	if (ldebug(sched_setaffinity))
2131 		printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid,
2132 		    args->len);
2133 #endif
2134 	if (args->len < sizeof(cpuset_t))
2135 		return (EINVAL);
2136 
2137 	tdt = linux_tdfind(td, args->pid, -1);
2138 	if (tdt == NULL)
2139 		return (ESRCH);
2140 
2141 	PROC_UNLOCK(tdt->td_proc);
2142 	csa.level = CPU_LEVEL_WHICH;
2143 	csa.which = CPU_WHICH_TID;
2144 	csa.id = tdt->td_tid;
2145 	csa.cpusetsize = sizeof(cpuset_t);
2146 	csa.mask = (cpuset_t *) args->user_mask_ptr;
2147 
2148 	return (sys_cpuset_setaffinity(td, &csa));
2149 }
2150 
2151 struct linux_rlimit64 {
2152 	uint64_t	rlim_cur;
2153 	uint64_t	rlim_max;
2154 };
2155 
2156 int
2157 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
2158 {
2159 	struct rlimit rlim, nrlim;
2160 	struct linux_rlimit64 lrlim;
2161 	struct proc *p;
2162 	u_int which;
2163 	int flags;
2164 	int error;
2165 
2166 #ifdef DEBUG
2167 	if (ldebug(prlimit64))
2168 		printf(ARGS(prlimit64, "%d, %d, %p, %p"), args->pid,
2169 		    args->resource, (void *)args->new, (void *)args->old);
2170 #endif
2171 
2172 	if (args->resource >= LINUX_RLIM_NLIMITS)
2173 		return (EINVAL);
2174 
2175 	which = linux_to_bsd_resource[args->resource];
2176 	if (which == -1)
2177 		return (EINVAL);
2178 
2179 	if (args->new != NULL) {
2180 		/*
2181 		 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
2182 		 * rlim is unsigned 64-bit. FreeBSD treats negative limits
2183 		 * as INFINITY so we do not need a conversion even.
2184 		 */
2185 		error = copyin(args->new, &nrlim, sizeof(nrlim));
2186 		if (error != 0)
2187 			return (error);
2188 	}
2189 
2190 	flags = PGET_HOLD | PGET_NOTWEXIT;
2191 	if (args->new != NULL)
2192 		flags |= PGET_CANDEBUG;
2193 	else
2194 		flags |= PGET_CANSEE;
2195 	error = pget(args->pid, flags, &p);
2196 	if (error != 0)
2197 		return (error);
2198 
2199 	if (args->old != NULL) {
2200 		PROC_LOCK(p);
2201 		lim_rlimit_proc(p, which, &rlim);
2202 		PROC_UNLOCK(p);
2203 		if (rlim.rlim_cur == RLIM_INFINITY)
2204 			lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2205 		else
2206 			lrlim.rlim_cur = rlim.rlim_cur;
2207 		if (rlim.rlim_max == RLIM_INFINITY)
2208 			lrlim.rlim_max = LINUX_RLIM_INFINITY;
2209 		else
2210 			lrlim.rlim_max = rlim.rlim_max;
2211 		error = copyout(&lrlim, args->old, sizeof(lrlim));
2212 		if (error != 0)
2213 			goto out;
2214 	}
2215 
2216 	if (args->new != NULL)
2217 		error = kern_proc_setrlimit(td, p, which, &nrlim);
2218 
2219  out:
2220 	PRELE(p);
2221 	return (error);
2222 }
2223 
2224 int
2225 linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2226 {
2227 	struct timeval utv, tv0, tv1, *tvp;
2228 	struct l_pselect6arg lpse6;
2229 	struct l_timespec lts;
2230 	struct timespec uts;
2231 	l_sigset_t l_ss;
2232 	sigset_t *ssp;
2233 	sigset_t ss;
2234 	int error;
2235 
2236 	ssp = NULL;
2237 	if (args->sig != NULL) {
2238 		error = copyin(args->sig, &lpse6, sizeof(lpse6));
2239 		if (error != 0)
2240 			return (error);
2241 		if (lpse6.ss_len != sizeof(l_ss))
2242 			return (EINVAL);
2243 		if (lpse6.ss != 0) {
2244 			error = copyin(PTRIN(lpse6.ss), &l_ss,
2245 			    sizeof(l_ss));
2246 			if (error != 0)
2247 				return (error);
2248 			linux_to_bsd_sigset(&l_ss, &ss);
2249 			ssp = &ss;
2250 		}
2251 	}
2252 
2253 	/*
2254 	 * Currently glibc changes nanosecond number to microsecond.
2255 	 * This mean losing precision but for now it is hardly seen.
2256 	 */
2257 	if (args->tsp != NULL) {
2258 		error = copyin(args->tsp, &lts, sizeof(lts));
2259 		if (error != 0)
2260 			return (error);
2261 		error = linux_to_native_timespec(&uts, &lts);
2262 		if (error != 0)
2263 			return (error);
2264 
2265 		TIMESPEC_TO_TIMEVAL(&utv, &uts);
2266 		if (itimerfix(&utv))
2267 			return (EINVAL);
2268 
2269 		microtime(&tv0);
2270 		tvp = &utv;
2271 	} else
2272 		tvp = NULL;
2273 
2274 	error = kern_pselect(td, args->nfds, args->readfds, args->writefds,
2275 	    args->exceptfds, tvp, ssp, LINUX_NFDBITS);
2276 
2277 	if (error == 0 && args->tsp != NULL) {
2278 		if (td->td_retval[0] != 0) {
2279 			/*
2280 			 * Compute how much time was left of the timeout,
2281 			 * by subtracting the current time and the time
2282 			 * before we started the call, and subtracting
2283 			 * that result from the user-supplied value.
2284 			 */
2285 
2286 			microtime(&tv1);
2287 			timevalsub(&tv1, &tv0);
2288 			timevalsub(&utv, &tv1);
2289 			if (utv.tv_sec < 0)
2290 				timevalclear(&utv);
2291 		} else
2292 			timevalclear(&utv);
2293 
2294 		TIMEVAL_TO_TIMESPEC(&utv, &uts);
2295 
2296 		native_to_linux_timespec(&lts, &uts);
2297 		error = copyout(&lts, args->tsp, sizeof(lts));
2298 	}
2299 
2300 	return (error);
2301 }
2302 
2303 int
2304 linux_ppoll(struct thread *td, struct linux_ppoll_args *args)
2305 {
2306 	struct timespec ts0, ts1;
2307 	struct l_timespec lts;
2308 	struct timespec uts, *tsp;
2309 	l_sigset_t l_ss;
2310 	sigset_t *ssp;
2311 	sigset_t ss;
2312 	int error;
2313 
2314 	if (args->sset != NULL) {
2315 		if (args->ssize != sizeof(l_ss))
2316 			return (EINVAL);
2317 		error = copyin(args->sset, &l_ss, sizeof(l_ss));
2318 		if (error)
2319 			return (error);
2320 		linux_to_bsd_sigset(&l_ss, &ss);
2321 		ssp = &ss;
2322 	} else
2323 		ssp = NULL;
2324 	if (args->tsp != NULL) {
2325 		error = copyin(args->tsp, &lts, sizeof(lts));
2326 		if (error)
2327 			return (error);
2328 		error = linux_to_native_timespec(&uts, &lts);
2329 		if (error != 0)
2330 			return (error);
2331 
2332 		nanotime(&ts0);
2333 		tsp = &uts;
2334 	} else
2335 		tsp = NULL;
2336 
2337 	error = kern_poll(td, args->fds, args->nfds, tsp, ssp);
2338 
2339 	if (error == 0 && args->tsp != NULL) {
2340 		if (td->td_retval[0]) {
2341 			nanotime(&ts1);
2342 			timespecsub(&ts1, &ts0);
2343 			timespecsub(&uts, &ts1);
2344 			if (uts.tv_sec < 0)
2345 				timespecclear(&uts);
2346 		} else
2347 			timespecclear(&uts);
2348 
2349 		native_to_linux_timespec(&lts, &uts);
2350 		error = copyout(&lts, args->tsp, sizeof(lts));
2351 	}
2352 
2353 	return (error);
2354 }
2355 
2356 #if defined(DEBUG) || defined(KTR)
2357 /* XXX: can be removed when every ldebug(...) and KTR stuff are removed. */
2358 
2359 #ifdef COMPAT_LINUX32
2360 #define	L_MAXSYSCALL	LINUX32_SYS_MAXSYSCALL
2361 #else
2362 #define	L_MAXSYSCALL	LINUX_SYS_MAXSYSCALL
2363 #endif
2364 
2365 u_char linux_debug_map[howmany(L_MAXSYSCALL, sizeof(u_char))];
2366 
2367 static int
2368 linux_debug(int syscall, int toggle, int global)
2369 {
2370 
2371 	if (global) {
2372 		char c = toggle ? 0 : 0xff;
2373 
2374 		memset(linux_debug_map, c, sizeof(linux_debug_map));
2375 		return (0);
2376 	}
2377 	if (syscall < 0 || syscall >= L_MAXSYSCALL)
2378 		return (EINVAL);
2379 	if (toggle)
2380 		clrbit(linux_debug_map, syscall);
2381 	else
2382 		setbit(linux_debug_map, syscall);
2383 	return (0);
2384 }
2385 #undef L_MAXSYSCALL
2386 
2387 /*
2388  * Usage: sysctl linux.debug=<syscall_nr>.<0/1>
2389  *
2390  *    E.g.: sysctl linux.debug=21.0
2391  *
2392  * As a special case, syscall "all" will apply to all syscalls globally.
2393  */
2394 #define LINUX_MAX_DEBUGSTR	16
2395 int
2396 linux_sysctl_debug(SYSCTL_HANDLER_ARGS)
2397 {
2398 	char value[LINUX_MAX_DEBUGSTR], *p;
2399 	int error, sysc, toggle;
2400 	int global = 0;
2401 
2402 	value[0] = '\0';
2403 	error = sysctl_handle_string(oidp, value, LINUX_MAX_DEBUGSTR, req);
2404 	if (error || req->newptr == NULL)
2405 		return (error);
2406 	for (p = value; *p != '\0' && *p != '.'; p++);
2407 	if (*p == '\0')
2408 		return (EINVAL);
2409 	*p++ = '\0';
2410 	sysc = strtol(value, NULL, 0);
2411 	toggle = strtol(p, NULL, 0);
2412 	if (strcmp(value, "all") == 0)
2413 		global = 1;
2414 	error = linux_debug(sysc, toggle, global);
2415 	return (error);
2416 }
2417 
2418 #endif /* DEBUG || KTR */
2419 
2420 int
2421 linux_sched_rr_get_interval(struct thread *td,
2422     struct linux_sched_rr_get_interval_args *uap)
2423 {
2424 	struct timespec ts;
2425 	struct l_timespec lts;
2426 	struct thread *tdt;
2427 	int error;
2428 
2429 	/*
2430 	 * According to man in case the invalid pid specified
2431 	 * EINVAL should be returned.
2432 	 */
2433 	if (uap->pid < 0)
2434 		return (EINVAL);
2435 
2436 	tdt = linux_tdfind(td, uap->pid, -1);
2437 	if (tdt == NULL)
2438 		return (ESRCH);
2439 
2440 	error = kern_sched_rr_get_interval_td(td, tdt, &ts);
2441 	PROC_UNLOCK(tdt->td_proc);
2442 	if (error != 0)
2443 		return (error);
2444 	native_to_linux_timespec(&lts, &ts);
2445 	return (copyout(&lts, uap->interval, sizeof(lts)));
2446 }
2447 
2448 /*
2449  * In case when the Linux thread is the initial thread in
2450  * the thread group thread id is equal to the process id.
2451  * Glibc depends on this magic (assert in pthread_getattr_np.c).
2452  */
2453 struct thread *
2454 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid)
2455 {
2456 	struct linux_emuldata *em;
2457 	struct thread *tdt;
2458 	struct proc *p;
2459 
2460 	tdt = NULL;
2461 	if (tid == 0 || tid == td->td_tid) {
2462 		tdt = td;
2463 		PROC_LOCK(tdt->td_proc);
2464 	} else if (tid > PID_MAX)
2465 		tdt = tdfind(tid, pid);
2466 	else {
2467 		/*
2468 		 * Initial thread where the tid equal to the pid.
2469 		 */
2470 		p = pfind(tid);
2471 		if (p != NULL) {
2472 			if (SV_PROC_ABI(p) != SV_ABI_LINUX) {
2473 				/*
2474 				 * p is not a Linuxulator process.
2475 				 */
2476 				PROC_UNLOCK(p);
2477 				return (NULL);
2478 			}
2479 			FOREACH_THREAD_IN_PROC(p, tdt) {
2480 				em = em_find(tdt);
2481 				if (tid == em->em_tid)
2482 					return (tdt);
2483 			}
2484 			PROC_UNLOCK(p);
2485 		}
2486 		return (NULL);
2487 	}
2488 
2489 	return (tdt);
2490 }
2491 
2492 void
2493 linux_to_bsd_waitopts(int options, int *bsdopts)
2494 {
2495 
2496 	if (options & LINUX_WNOHANG)
2497 		*bsdopts |= WNOHANG;
2498 	if (options & LINUX_WUNTRACED)
2499 		*bsdopts |= WUNTRACED;
2500 	if (options & LINUX_WEXITED)
2501 		*bsdopts |= WEXITED;
2502 	if (options & LINUX_WCONTINUED)
2503 		*bsdopts |= WCONTINUED;
2504 	if (options & LINUX_WNOWAIT)
2505 		*bsdopts |= WNOWAIT;
2506 
2507 	if (options & __WCLONE)
2508 		*bsdopts |= WLINUXCLONE;
2509 }
2510