xref: /freebsd/sys/compat/linux/linux_misc.c (revision bce7ee9d412b6410e6d799c4a417617cbb148e09)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 2002 Doug Rabson
5  * Copyright (c) 1994-1995 Søren Schmidt
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer
13  *    in this position and unchanged.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. The name of the author may not be used to endorse or promote products
18  *    derived from this software without specific prior written permission
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
21  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
22  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
23  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
24  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
25  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
26  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
27  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
28  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
29  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
30  */
31 
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
34 
35 #include "opt_compat.h"
36 
37 #include <sys/param.h>
38 #include <sys/blist.h>
39 #include <sys/fcntl.h>
40 #if defined(__i386__)
41 #include <sys/imgact_aout.h>
42 #endif
43 #include <sys/jail.h>
44 #include <sys/kernel.h>
45 #include <sys/limits.h>
46 #include <sys/lock.h>
47 #include <sys/malloc.h>
48 #include <sys/mman.h>
49 #include <sys/mount.h>
50 #include <sys/msgbuf.h>
51 #include <sys/mutex.h>
52 #include <sys/namei.h>
53 #include <sys/priv.h>
54 #include <sys/proc.h>
55 #include <sys/procctl.h>
56 #include <sys/reboot.h>
57 #include <sys/racct.h>
58 #include <sys/random.h>
59 #include <sys/resourcevar.h>
60 #include <sys/sched.h>
61 #include <sys/sdt.h>
62 #include <sys/signalvar.h>
63 #include <sys/stat.h>
64 #include <sys/syscallsubr.h>
65 #include <sys/sysctl.h>
66 #include <sys/sysproto.h>
67 #include <sys/systm.h>
68 #include <sys/time.h>
69 #include <sys/vmmeter.h>
70 #include <sys/vnode.h>
71 #include <sys/wait.h>
72 #include <sys/cpuset.h>
73 #include <sys/uio.h>
74 
75 #include <security/mac/mac_framework.h>
76 
77 #include <vm/vm.h>
78 #include <vm/pmap.h>
79 #include <vm/vm_kern.h>
80 #include <vm/vm_map.h>
81 #include <vm/vm_extern.h>
82 #include <vm/swap_pager.h>
83 
84 #ifdef COMPAT_LINUX32
85 #include <machine/../linux32/linux.h>
86 #include <machine/../linux32/linux32_proto.h>
87 #else
88 #include <machine/../linux/linux.h>
89 #include <machine/../linux/linux_proto.h>
90 #endif
91 
92 #include <compat/linux/linux_dtrace.h>
93 #include <compat/linux/linux_file.h>
94 #include <compat/linux/linux_mib.h>
95 #include <compat/linux/linux_signal.h>
96 #include <compat/linux/linux_timer.h>
97 #include <compat/linux/linux_util.h>
98 #include <compat/linux/linux_sysproto.h>
99 #include <compat/linux/linux_emul.h>
100 #include <compat/linux/linux_misc.h>
101 
102 /**
103  * Special DTrace provider for the linuxulator.
104  *
105  * In this file we define the provider for the entire linuxulator. All
106  * modules (= files of the linuxulator) use it.
107  *
108  * We define a different name depending on the emulated bitsize, see
109  * ../../<ARCH>/linux{,32}/linux.h, e.g.:
110  *      native bitsize          = linuxulator
111  *      amd64, 32bit emulation  = linuxulator32
112  */
113 LIN_SDT_PROVIDER_DEFINE(LINUX_DTRACE);
114 
115 int stclohz;				/* Statistics clock frequency */
116 
117 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
118 	RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
119 	RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
120 	RLIMIT_MEMLOCK, RLIMIT_AS
121 };
122 
123 struct l_sysinfo {
124 	l_long		uptime;		/* Seconds since boot */
125 	l_ulong		loads[3];	/* 1, 5, and 15 minute load averages */
126 #define LINUX_SYSINFO_LOADS_SCALE 65536
127 	l_ulong		totalram;	/* Total usable main memory size */
128 	l_ulong		freeram;	/* Available memory size */
129 	l_ulong		sharedram;	/* Amount of shared memory */
130 	l_ulong		bufferram;	/* Memory used by buffers */
131 	l_ulong		totalswap;	/* Total swap space size */
132 	l_ulong		freeswap;	/* swap space still available */
133 	l_ushort	procs;		/* Number of current processes */
134 	l_ushort	pads;
135 	l_ulong		totalhigh;
136 	l_ulong		freehigh;
137 	l_uint		mem_unit;
138 	char		_f[20-2*sizeof(l_long)-sizeof(l_int)];	/* padding */
139 };
140 
141 struct l_pselect6arg {
142 	l_uintptr_t	ss;
143 	l_size_t	ss_len;
144 };
145 
146 static int	linux_utimensat_nsec_valid(l_long);
147 
148 int
149 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
150 {
151 	struct l_sysinfo sysinfo;
152 	int i, j;
153 	struct timespec ts;
154 
155 	bzero(&sysinfo, sizeof(sysinfo));
156 	getnanouptime(&ts);
157 	if (ts.tv_nsec != 0)
158 		ts.tv_sec++;
159 	sysinfo.uptime = ts.tv_sec;
160 
161 	/* Use the information from the mib to get our load averages */
162 	for (i = 0; i < 3; i++)
163 		sysinfo.loads[i] = averunnable.ldavg[i] *
164 		    LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
165 
166 	sysinfo.totalram = physmem * PAGE_SIZE;
167 	sysinfo.freeram = (u_long)vm_free_count() * PAGE_SIZE;
168 
169 	/*
170 	 * sharedram counts pages allocated to named, swap-backed objects such
171 	 * as shared memory segments and tmpfs files.  There is no cheap way to
172 	 * compute this, so just leave the field unpopulated.  Linux itself only
173 	 * started setting this field in the 3.x timeframe.
174 	 */
175 	sysinfo.sharedram = 0;
176 	sysinfo.bufferram = 0;
177 
178 	swap_pager_status(&i, &j);
179 	sysinfo.totalswap = i * PAGE_SIZE;
180 	sysinfo.freeswap = (i - j) * PAGE_SIZE;
181 
182 	sysinfo.procs = nprocs;
183 
184 	/*
185 	 * Platforms supported by the emulation layer do not have a notion of
186 	 * high memory.
187 	 */
188 	sysinfo.totalhigh = 0;
189 	sysinfo.freehigh = 0;
190 
191 	sysinfo.mem_unit = 1;
192 
193 	return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
194 }
195 
196 #ifdef LINUX_LEGACY_SYSCALLS
197 int
198 linux_alarm(struct thread *td, struct linux_alarm_args *args)
199 {
200 	struct itimerval it, old_it;
201 	u_int secs;
202 	int error;
203 
204 	secs = args->secs;
205 	/*
206 	 * Linux alarm() is always successful. Limit secs to INT32_MAX / 2
207 	 * to match kern_setitimer()'s limit to avoid error from it.
208 	 *
209 	 * XXX. Linux limit secs to INT_MAX on 32 and does not limit on 64-bit
210 	 * platforms.
211 	 */
212 	if (secs > INT32_MAX / 2)
213 		secs = INT32_MAX / 2;
214 
215 	it.it_value.tv_sec = secs;
216 	it.it_value.tv_usec = 0;
217 	timevalclear(&it.it_interval);
218 	error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
219 	KASSERT(error == 0, ("kern_setitimer returns %d", error));
220 
221 	if ((old_it.it_value.tv_sec == 0 && old_it.it_value.tv_usec > 0) ||
222 	    old_it.it_value.tv_usec >= 500000)
223 		old_it.it_value.tv_sec++;
224 	td->td_retval[0] = old_it.it_value.tv_sec;
225 	return (0);
226 }
227 #endif
228 
229 int
230 linux_brk(struct thread *td, struct linux_brk_args *args)
231 {
232 	struct vmspace *vm = td->td_proc->p_vmspace;
233 	uintptr_t new, old;
234 
235 	old = (uintptr_t)vm->vm_daddr + ctob(vm->vm_dsize);
236 	new = (uintptr_t)args->dsend;
237 	if ((caddr_t)new > vm->vm_daddr && !kern_break(td, &new))
238 		td->td_retval[0] = (register_t)new;
239 	else
240 		td->td_retval[0] = (register_t)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 	vm_map_t map;
255 	vm_map_entry_t entry;
256 	struct vattr attr;
257 	vm_offset_t vmaddr;
258 	unsigned long file_offset;
259 	unsigned long bss_size;
260 	char *library;
261 	ssize_t aresid;
262 	int error;
263 	bool locked, opened, textset;
264 
265 	a_out = NULL;
266 	vp = NULL;
267 	locked = false;
268 	textset = false;
269 	opened = false;
270 
271 	if (!LUSECONVPATH(td)) {
272 		NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
273 		    UIO_USERSPACE, args->library, td);
274 		error = namei(&ni);
275 	} else {
276 		LCONVPATHEXIST(td, args->library, &library);
277 		NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
278 		    UIO_SYSSPACE, library, td);
279 		error = namei(&ni);
280 		LFREEPATH(library);
281 	}
282 	if (error)
283 		goto cleanup;
284 
285 	vp = ni.ni_vp;
286 	NDFREE(&ni, NDF_ONLY_PNBUF);
287 
288 	/*
289 	 * From here on down, we have a locked vnode that must be unlocked.
290 	 * XXX: The code below largely duplicates exec_check_permissions().
291 	 */
292 	locked = true;
293 
294 	/* Executable? */
295 	error = VOP_GETATTR(vp, &attr, td->td_ucred);
296 	if (error)
297 		goto cleanup;
298 
299 	if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
300 	    ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
301 		/* EACCESS is what exec(2) returns. */
302 		error = ENOEXEC;
303 		goto cleanup;
304 	}
305 
306 	/* Sensible size? */
307 	if (attr.va_size == 0) {
308 		error = ENOEXEC;
309 		goto cleanup;
310 	}
311 
312 	/* Can we access it? */
313 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
314 	if (error)
315 		goto cleanup;
316 
317 	/*
318 	 * XXX: This should use vn_open() so that it is properly authorized,
319 	 * and to reduce code redundancy all over the place here.
320 	 * XXX: Not really, it duplicates far more of exec_check_permissions()
321 	 * than vn_open().
322 	 */
323 #ifdef MAC
324 	error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
325 	if (error)
326 		goto cleanup;
327 #endif
328 	error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
329 	if (error)
330 		goto cleanup;
331 	opened = true;
332 
333 	/* Pull in executable header into exec_map */
334 	error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
335 	    VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
336 	if (error)
337 		goto cleanup;
338 
339 	/* Is it a Linux binary ? */
340 	if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
341 		error = ENOEXEC;
342 		goto cleanup;
343 	}
344 
345 	/*
346 	 * While we are here, we should REALLY do some more checks
347 	 */
348 
349 	/* Set file/virtual offset based on a.out variant. */
350 	switch ((int)(a_out->a_magic & 0xffff)) {
351 	case 0413:			/* ZMAGIC */
352 		file_offset = 1024;
353 		break;
354 	case 0314:			/* QMAGIC */
355 		file_offset = 0;
356 		break;
357 	default:
358 		error = ENOEXEC;
359 		goto cleanup;
360 	}
361 
362 	bss_size = round_page(a_out->a_bss);
363 
364 	/* Check various fields in header for validity/bounds. */
365 	if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
366 		error = ENOEXEC;
367 		goto cleanup;
368 	}
369 
370 	/* text + data can't exceed file size */
371 	if (a_out->a_data + a_out->a_text > attr.va_size) {
372 		error = EFAULT;
373 		goto cleanup;
374 	}
375 
376 	/*
377 	 * text/data/bss must not exceed limits
378 	 * XXX - this is not complete. it should check current usage PLUS
379 	 * the resources needed by this library.
380 	 */
381 	PROC_LOCK(td->td_proc);
382 	if (a_out->a_text > maxtsiz ||
383 	    a_out->a_data + bss_size > lim_cur_proc(td->td_proc, RLIMIT_DATA) ||
384 	    racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
385 	    bss_size) != 0) {
386 		PROC_UNLOCK(td->td_proc);
387 		error = ENOMEM;
388 		goto cleanup;
389 	}
390 	PROC_UNLOCK(td->td_proc);
391 
392 	/*
393 	 * Prevent more writers.
394 	 */
395 	error = VOP_SET_TEXT(vp);
396 	if (error != 0)
397 		goto cleanup;
398 	textset = true;
399 
400 	/*
401 	 * Lock no longer needed
402 	 */
403 	locked = false;
404 	VOP_UNLOCK(vp);
405 
406 	/*
407 	 * Check if file_offset page aligned. Currently we cannot handle
408 	 * misalinged file offsets, and so we read in the entire image
409 	 * (what a waste).
410 	 */
411 	if (file_offset & PAGE_MASK) {
412 		/* Map text+data read/write/execute */
413 
414 		/* a_entry is the load address and is page aligned */
415 		vmaddr = trunc_page(a_out->a_entry);
416 
417 		/* get anon user mapping, read+write+execute */
418 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
419 		    &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
420 		    VM_PROT_ALL, VM_PROT_ALL, 0);
421 		if (error)
422 			goto cleanup;
423 
424 		error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
425 		    a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
426 		    td->td_ucred, NOCRED, &aresid, td);
427 		if (error != 0)
428 			goto cleanup;
429 		if (aresid != 0) {
430 			error = ENOEXEC;
431 			goto cleanup;
432 		}
433 	} else {
434 		/*
435 		 * for QMAGIC, a_entry is 20 bytes beyond the load address
436 		 * to skip the executable header
437 		 */
438 		vmaddr = trunc_page(a_out->a_entry);
439 
440 		/*
441 		 * Map it all into the process's space as a single
442 		 * copy-on-write "data" segment.
443 		 */
444 		map = &td->td_proc->p_vmspace->vm_map;
445 		error = vm_mmap(map, &vmaddr,
446 		    a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
447 		    MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
448 		if (error)
449 			goto cleanup;
450 		vm_map_lock(map);
451 		if (!vm_map_lookup_entry(map, vmaddr, &entry)) {
452 			vm_map_unlock(map);
453 			error = EDOOFUS;
454 			goto cleanup;
455 		}
456 		entry->eflags |= MAP_ENTRY_VN_EXEC;
457 		vm_map_unlock(map);
458 		textset = false;
459 	}
460 
461 	if (bss_size != 0) {
462 		/* Calculate BSS start address */
463 		vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
464 		    a_out->a_data;
465 
466 		/* allocate some 'anon' space */
467 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
468 		    &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
469 		    VM_PROT_ALL, 0);
470 		if (error)
471 			goto cleanup;
472 	}
473 
474 cleanup:
475 	if (opened) {
476 		if (locked)
477 			VOP_UNLOCK(vp);
478 		locked = false;
479 		VOP_CLOSE(vp, FREAD, td->td_ucred, td);
480 	}
481 	if (textset) {
482 		if (!locked) {
483 			locked = true;
484 			VOP_LOCK(vp, LK_SHARED | LK_RETRY);
485 		}
486 		VOP_UNSET_TEXT_CHECKED(vp);
487 	}
488 	if (locked)
489 		VOP_UNLOCK(vp);
490 
491 	/* Release the temporary mapping. */
492 	if (a_out)
493 		kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
494 
495 	return (error);
496 }
497 
498 #endif	/* __i386__ */
499 
500 #ifdef LINUX_LEGACY_SYSCALLS
501 int
502 linux_select(struct thread *td, struct linux_select_args *args)
503 {
504 	l_timeval ltv;
505 	struct timeval tv0, tv1, utv, *tvp;
506 	int error;
507 
508 	/*
509 	 * Store current time for computation of the amount of
510 	 * time left.
511 	 */
512 	if (args->timeout) {
513 		if ((error = copyin(args->timeout, &ltv, sizeof(ltv))))
514 			goto select_out;
515 		utv.tv_sec = ltv.tv_sec;
516 		utv.tv_usec = ltv.tv_usec;
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 	if (error)
540 		goto select_out;
541 
542 	if (args->timeout) {
543 		if (td->td_retval[0]) {
544 			/*
545 			 * Compute how much time was left of the timeout,
546 			 * by subtracting the current time and the time
547 			 * before we started the call, and subtracting
548 			 * that result from the user-supplied value.
549 			 */
550 			microtime(&tv1);
551 			timevalsub(&tv1, &tv0);
552 			timevalsub(&utv, &tv1);
553 			if (utv.tv_sec < 0)
554 				timevalclear(&utv);
555 		} else
556 			timevalclear(&utv);
557 		ltv.tv_sec = utv.tv_sec;
558 		ltv.tv_usec = utv.tv_usec;
559 		if ((error = copyout(&ltv, args->timeout, sizeof(ltv))))
560 			goto select_out;
561 	}
562 
563 select_out:
564 	return (error);
565 }
566 #endif
567 
568 int
569 linux_mremap(struct thread *td, struct linux_mremap_args *args)
570 {
571 	uintptr_t addr;
572 	size_t len;
573 	int error = 0;
574 
575 	if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
576 		td->td_retval[0] = 0;
577 		return (EINVAL);
578 	}
579 
580 	/*
581 	 * Check for the page alignment.
582 	 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
583 	 */
584 	if (args->addr & PAGE_MASK) {
585 		td->td_retval[0] = 0;
586 		return (EINVAL);
587 	}
588 
589 	args->new_len = round_page(args->new_len);
590 	args->old_len = round_page(args->old_len);
591 
592 	if (args->new_len > args->old_len) {
593 		td->td_retval[0] = 0;
594 		return (ENOMEM);
595 	}
596 
597 	if (args->new_len < args->old_len) {
598 		addr = args->addr + args->new_len;
599 		len = args->old_len - args->new_len;
600 		error = kern_munmap(td, addr, len);
601 	}
602 
603 	td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
604 	return (error);
605 }
606 
607 #define LINUX_MS_ASYNC       0x0001
608 #define LINUX_MS_INVALIDATE  0x0002
609 #define LINUX_MS_SYNC        0x0004
610 
611 int
612 linux_msync(struct thread *td, struct linux_msync_args *args)
613 {
614 
615 	return (kern_msync(td, args->addr, args->len,
616 	    args->fl & ~LINUX_MS_SYNC));
617 }
618 
619 #ifdef LINUX_LEGACY_SYSCALLS
620 int
621 linux_time(struct thread *td, struct linux_time_args *args)
622 {
623 	struct timeval tv;
624 	l_time_t tm;
625 	int error;
626 
627 	microtime(&tv);
628 	tm = tv.tv_sec;
629 	if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
630 		return (error);
631 	td->td_retval[0] = tm;
632 	return (0);
633 }
634 #endif
635 
636 struct l_times_argv {
637 	l_clock_t	tms_utime;
638 	l_clock_t	tms_stime;
639 	l_clock_t	tms_cutime;
640 	l_clock_t	tms_cstime;
641 };
642 
643 /*
644  * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
645  * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
646  * auxiliary vector entry.
647  */
648 #define	CLK_TCK		100
649 
650 #define	CONVOTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
651 #define	CONVNTCK(r)	(r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
652 
653 #define	CONVTCK(r)	(linux_kernver(td) >= LINUX_KERNVER_2004000 ?		\
654 			    CONVNTCK(r) : CONVOTCK(r))
655 
656 int
657 linux_times(struct thread *td, struct linux_times_args *args)
658 {
659 	struct timeval tv, utime, stime, cutime, cstime;
660 	struct l_times_argv tms;
661 	struct proc *p;
662 	int error;
663 
664 	if (args->buf != NULL) {
665 		p = td->td_proc;
666 		PROC_LOCK(p);
667 		PROC_STATLOCK(p);
668 		calcru(p, &utime, &stime);
669 		PROC_STATUNLOCK(p);
670 		calccru(p, &cutime, &cstime);
671 		PROC_UNLOCK(p);
672 
673 		tms.tms_utime = CONVTCK(utime);
674 		tms.tms_stime = CONVTCK(stime);
675 
676 		tms.tms_cutime = CONVTCK(cutime);
677 		tms.tms_cstime = CONVTCK(cstime);
678 
679 		if ((error = copyout(&tms, args->buf, sizeof(tms))))
680 			return (error);
681 	}
682 
683 	microuptime(&tv);
684 	td->td_retval[0] = (int)CONVTCK(tv);
685 	return (0);
686 }
687 
688 int
689 linux_newuname(struct thread *td, struct linux_newuname_args *args)
690 {
691 	struct l_new_utsname utsname;
692 	char osname[LINUX_MAX_UTSNAME];
693 	char osrelease[LINUX_MAX_UTSNAME];
694 	char *p;
695 
696 	linux_get_osname(td, osname);
697 	linux_get_osrelease(td, osrelease);
698 
699 	bzero(&utsname, sizeof(utsname));
700 	strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
701 	getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
702 	getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
703 	strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
704 	strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
705 	for (p = utsname.version; *p != '\0'; ++p)
706 		if (*p == '\n') {
707 			*p = '\0';
708 			break;
709 		}
710 #if defined(__amd64__)
711 	/*
712 	 * On amd64, Linux uname(2) needs to return "x86_64"
713 	 * for both 64-bit and 32-bit applications.  On 32-bit,
714 	 * the string returned by getauxval(AT_PLATFORM) needs
715 	 * to remain "i686", though.
716 	 */
717 	strlcpy(utsname.machine, "x86_64", LINUX_MAX_UTSNAME);
718 #else
719 	strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME);
720 #endif
721 
722 	return (copyout(&utsname, args->buf, sizeof(utsname)));
723 }
724 
725 struct l_utimbuf {
726 	l_time_t l_actime;
727 	l_time_t l_modtime;
728 };
729 
730 #ifdef LINUX_LEGACY_SYSCALLS
731 int
732 linux_utime(struct thread *td, struct linux_utime_args *args)
733 {
734 	struct timeval tv[2], *tvp;
735 	struct l_utimbuf lut;
736 	char *fname;
737 	int error;
738 	bool convpath;
739 
740 	convpath = LUSECONVPATH(td);
741 	if (convpath)
742 		LCONVPATHEXIST(td, args->fname, &fname);
743 
744 	if (args->times) {
745 		if ((error = copyin(args->times, &lut, sizeof lut))) {
746 			if (convpath)
747 				LFREEPATH(fname);
748 			return (error);
749 		}
750 		tv[0].tv_sec = lut.l_actime;
751 		tv[0].tv_usec = 0;
752 		tv[1].tv_sec = lut.l_modtime;
753 		tv[1].tv_usec = 0;
754 		tvp = tv;
755 	} else
756 		tvp = NULL;
757 
758 	if (!convpath) {
759 		error = kern_utimesat(td, AT_FDCWD, args->fname, UIO_USERSPACE,
760 		    tvp, UIO_SYSSPACE);
761 	} else {
762 		error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, tvp,
763 		    UIO_SYSSPACE);
764 		LFREEPATH(fname);
765 	}
766 	return (error);
767 }
768 #endif
769 
770 #ifdef LINUX_LEGACY_SYSCALLS
771 int
772 linux_utimes(struct thread *td, struct linux_utimes_args *args)
773 {
774 	l_timeval ltv[2];
775 	struct timeval tv[2], *tvp = NULL;
776 	char *fname;
777 	int error;
778 	bool convpath;
779 
780 	convpath = LUSECONVPATH(td);
781 	if (convpath)
782 		LCONVPATHEXIST(td, args->fname, &fname);
783 
784 	if (args->tptr != NULL) {
785 		if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
786 			LFREEPATH(fname);
787 			return (error);
788 		}
789 		tv[0].tv_sec = ltv[0].tv_sec;
790 		tv[0].tv_usec = ltv[0].tv_usec;
791 		tv[1].tv_sec = ltv[1].tv_sec;
792 		tv[1].tv_usec = ltv[1].tv_usec;
793 		tvp = tv;
794 	}
795 
796 	if (!convpath) {
797 		error = kern_utimesat(td, AT_FDCWD, args->fname, UIO_USERSPACE,
798 		    tvp, UIO_SYSSPACE);
799 	} else {
800 		error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE,
801 		    tvp, UIO_SYSSPACE);
802 		LFREEPATH(fname);
803 	}
804 	return (error);
805 }
806 #endif
807 
808 static int
809 linux_utimensat_nsec_valid(l_long nsec)
810 {
811 
812 	if (nsec == LINUX_UTIME_OMIT || nsec == LINUX_UTIME_NOW)
813 		return (0);
814 	if (nsec >= 0 && nsec <= 999999999)
815 		return (0);
816 	return (1);
817 }
818 
819 int
820 linux_utimensat(struct thread *td, struct linux_utimensat_args *args)
821 {
822 	struct l_timespec l_times[2];
823 	struct timespec times[2], *timesp = NULL;
824 	char *path = NULL;
825 	int error, dfd, flags = 0;
826 
827 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
828 
829 	if (args->flags & ~LINUX_AT_SYMLINK_NOFOLLOW)
830 		return (EINVAL);
831 
832 	if (args->times != NULL) {
833 		error = copyin(args->times, l_times, sizeof(l_times));
834 		if (error != 0)
835 			return (error);
836 
837 		if (linux_utimensat_nsec_valid(l_times[0].tv_nsec) != 0 ||
838 		    linux_utimensat_nsec_valid(l_times[1].tv_nsec) != 0)
839 			return (EINVAL);
840 
841 		times[0].tv_sec = l_times[0].tv_sec;
842 		switch (l_times[0].tv_nsec)
843 		{
844 		case LINUX_UTIME_OMIT:
845 			times[0].tv_nsec = UTIME_OMIT;
846 			break;
847 		case LINUX_UTIME_NOW:
848 			times[0].tv_nsec = UTIME_NOW;
849 			break;
850 		default:
851 			times[0].tv_nsec = l_times[0].tv_nsec;
852 		}
853 
854 		times[1].tv_sec = l_times[1].tv_sec;
855 		switch (l_times[1].tv_nsec)
856 		{
857 		case LINUX_UTIME_OMIT:
858 			times[1].tv_nsec = UTIME_OMIT;
859 			break;
860 		case LINUX_UTIME_NOW:
861 			times[1].tv_nsec = UTIME_NOW;
862 			break;
863 		default:
864 			times[1].tv_nsec = l_times[1].tv_nsec;
865 			break;
866 		}
867 		timesp = times;
868 
869 		/* This breaks POSIX, but is what the Linux kernel does
870 		 * _on purpose_ (documented in the man page for utimensat(2)),
871 		 * so we must follow that behaviour. */
872 		if (times[0].tv_nsec == UTIME_OMIT &&
873 		    times[1].tv_nsec == UTIME_OMIT)
874 			return (0);
875 	}
876 
877 	if (!LUSECONVPATH(td)) {
878 		if (args->pathname != NULL) {
879 			return (kern_utimensat(td, dfd, args->pathname,
880 			    UIO_USERSPACE, timesp, UIO_SYSSPACE, flags));
881 		}
882 	}
883 
884 	if (args->pathname != NULL)
885 		LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
886 	else if (args->flags != 0)
887 		return (EINVAL);
888 
889 	if (args->flags & LINUX_AT_SYMLINK_NOFOLLOW)
890 		flags |= AT_SYMLINK_NOFOLLOW;
891 
892 	if (path == NULL)
893 		error = kern_futimens(td, dfd, timesp, UIO_SYSSPACE);
894 	else {
895 		error = kern_utimensat(td, dfd, path, UIO_SYSSPACE, timesp,
896 			UIO_SYSSPACE, flags);
897 		LFREEPATH(path);
898 	}
899 
900 	return (error);
901 }
902 
903 #ifdef LINUX_LEGACY_SYSCALLS
904 int
905 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
906 {
907 	l_timeval ltv[2];
908 	struct timeval tv[2], *tvp = NULL;
909 	char *fname;
910 	int error, dfd;
911 	bool convpath;
912 
913 	convpath = LUSECONVPATH(td);
914 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
915 	if (convpath)
916 		LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
917 
918 	if (args->utimes != NULL) {
919 		if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
920 			if (convpath)
921 				LFREEPATH(fname);
922 			return (error);
923 		}
924 		tv[0].tv_sec = ltv[0].tv_sec;
925 		tv[0].tv_usec = ltv[0].tv_usec;
926 		tv[1].tv_sec = ltv[1].tv_sec;
927 		tv[1].tv_usec = ltv[1].tv_usec;
928 		tvp = tv;
929 	}
930 
931 	if (!convpath) {
932 		error = kern_utimesat(td, dfd, args->filename, UIO_USERSPACE,
933 		    tvp, UIO_SYSSPACE);
934 	} else {
935 		error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
936 		LFREEPATH(fname);
937 	}
938 	return (error);
939 }
940 #endif
941 
942 static int
943 linux_common_wait(struct thread *td, int pid, int *statusp,
944     int options, struct __wrusage *wrup)
945 {
946 	siginfo_t siginfo;
947 	idtype_t idtype;
948 	id_t id;
949 	int error, status, tmpstat;
950 
951 	if (pid == WAIT_ANY) {
952 		idtype = P_ALL;
953 		id = 0;
954 	} else if (pid < 0) {
955 		idtype = P_PGID;
956 		id = (id_t)-pid;
957 	} else {
958 		idtype = P_PID;
959 		id = (id_t)pid;
960 	}
961 
962 	/*
963 	 * For backward compatibility we implicitly add flags WEXITED
964 	 * and WTRAPPED here.
965 	 */
966 	options |= WEXITED | WTRAPPED;
967 	error = kern_wait6(td, idtype, id, &status, options, wrup, &siginfo);
968 	if (error)
969 		return (error);
970 
971 	if (statusp) {
972 		tmpstat = status & 0xffff;
973 		if (WIFSIGNALED(tmpstat)) {
974 			tmpstat = (tmpstat & 0xffffff80) |
975 			    bsd_to_linux_signal(WTERMSIG(tmpstat));
976 		} else if (WIFSTOPPED(tmpstat)) {
977 			tmpstat = (tmpstat & 0xffff00ff) |
978 			    (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8);
979 #if defined(__amd64__) && !defined(COMPAT_LINUX32)
980 			if (WSTOPSIG(status) == SIGTRAP) {
981 				tmpstat = linux_ptrace_status(td,
982 				    siginfo.si_pid, tmpstat);
983 			}
984 #endif
985 		} else if (WIFCONTINUED(tmpstat)) {
986 			tmpstat = 0xffff;
987 		}
988 		error = copyout(&tmpstat, statusp, sizeof(int));
989 	}
990 
991 	return (error);
992 }
993 
994 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
995 int
996 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
997 {
998 	struct linux_wait4_args wait4_args;
999 
1000 	wait4_args.pid = args->pid;
1001 	wait4_args.status = args->status;
1002 	wait4_args.options = args->options;
1003 	wait4_args.rusage = NULL;
1004 
1005 	return (linux_wait4(td, &wait4_args));
1006 }
1007 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1008 
1009 int
1010 linux_wait4(struct thread *td, struct linux_wait4_args *args)
1011 {
1012 	int error, options;
1013 	struct __wrusage wru, *wrup;
1014 
1015 	if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG |
1016 	    LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
1017 		return (EINVAL);
1018 
1019 	options = WEXITED;
1020 	linux_to_bsd_waitopts(args->options, &options);
1021 
1022 	if (args->rusage != NULL)
1023 		wrup = &wru;
1024 	else
1025 		wrup = NULL;
1026 	error = linux_common_wait(td, args->pid, args->status, options, wrup);
1027 	if (error != 0)
1028 		return (error);
1029 	if (args->rusage != NULL)
1030 		error = linux_copyout_rusage(&wru.wru_self, args->rusage);
1031 	return (error);
1032 }
1033 
1034 int
1035 linux_waitid(struct thread *td, struct linux_waitid_args *args)
1036 {
1037 	int status, options, sig;
1038 	struct __wrusage wru;
1039 	siginfo_t siginfo;
1040 	l_siginfo_t lsi;
1041 	idtype_t idtype;
1042 	struct proc *p;
1043 	int error;
1044 
1045 	options = 0;
1046 	linux_to_bsd_waitopts(args->options, &options);
1047 
1048 	if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED))
1049 		return (EINVAL);
1050 	if (!(options & (WEXITED | WUNTRACED | WCONTINUED)))
1051 		return (EINVAL);
1052 
1053 	switch (args->idtype) {
1054 	case LINUX_P_ALL:
1055 		idtype = P_ALL;
1056 		break;
1057 	case LINUX_P_PID:
1058 		if (args->id <= 0)
1059 			return (EINVAL);
1060 		idtype = P_PID;
1061 		break;
1062 	case LINUX_P_PGID:
1063 		if (args->id <= 0)
1064 			return (EINVAL);
1065 		idtype = P_PGID;
1066 		break;
1067 	default:
1068 		return (EINVAL);
1069 	}
1070 
1071 	error = kern_wait6(td, idtype, args->id, &status, options,
1072 	    &wru, &siginfo);
1073 	if (error != 0)
1074 		return (error);
1075 	if (args->rusage != NULL) {
1076 		error = linux_copyout_rusage(&wru.wru_children,
1077 		    args->rusage);
1078 		if (error != 0)
1079 			return (error);
1080 	}
1081 	if (args->info != NULL) {
1082 		p = td->td_proc;
1083 		bzero(&lsi, sizeof(lsi));
1084 		if (td->td_retval[0] != 0) {
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 #ifdef LINUX_LEGACY_SYSCALLS
1096 int
1097 linux_mknod(struct thread *td, struct linux_mknod_args *args)
1098 {
1099 	char *path;
1100 	int error;
1101 	enum uio_seg seg;
1102 	bool convpath;
1103 
1104 	convpath = LUSECONVPATH(td);
1105 	if (!convpath) {
1106 		path = args->path;
1107 		seg = UIO_USERSPACE;
1108 	} else {
1109 		LCONVPATHCREAT(td, args->path, &path);
1110 		seg = UIO_SYSSPACE;
1111 	}
1112 
1113 	switch (args->mode & S_IFMT) {
1114 	case S_IFIFO:
1115 	case S_IFSOCK:
1116 		error = kern_mkfifoat(td, AT_FDCWD, path, seg,
1117 		    args->mode);
1118 		break;
1119 
1120 	case S_IFCHR:
1121 	case S_IFBLK:
1122 		error = kern_mknodat(td, AT_FDCWD, path, seg,
1123 		    args->mode, args->dev);
1124 		break;
1125 
1126 	case S_IFDIR:
1127 		error = EPERM;
1128 		break;
1129 
1130 	case 0:
1131 		args->mode |= S_IFREG;
1132 		/* FALLTHROUGH */
1133 	case S_IFREG:
1134 		error = kern_openat(td, AT_FDCWD, path, seg,
1135 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1136 		if (error == 0)
1137 			kern_close(td, td->td_retval[0]);
1138 		break;
1139 
1140 	default:
1141 		error = EINVAL;
1142 		break;
1143 	}
1144 	if (convpath)
1145 		LFREEPATH(path);
1146 	return (error);
1147 }
1148 #endif
1149 
1150 int
1151 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
1152 {
1153 	char *path;
1154 	int error, dfd;
1155 	enum uio_seg seg;
1156 	bool convpath;
1157 
1158 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
1159 
1160 	convpath = LUSECONVPATH(td);
1161 	if (!convpath) {
1162 		path = __DECONST(char *, args->filename);
1163 		seg = UIO_USERSPACE;
1164 	} else {
1165 		LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
1166 		seg = UIO_SYSSPACE;
1167 	}
1168 
1169 	switch (args->mode & S_IFMT) {
1170 	case S_IFIFO:
1171 	case S_IFSOCK:
1172 		error = kern_mkfifoat(td, dfd, path, seg, args->mode);
1173 		break;
1174 
1175 	case S_IFCHR:
1176 	case S_IFBLK:
1177 		error = kern_mknodat(td, dfd, path, seg, args->mode,
1178 		    args->dev);
1179 		break;
1180 
1181 	case S_IFDIR:
1182 		error = EPERM;
1183 		break;
1184 
1185 	case 0:
1186 		args->mode |= S_IFREG;
1187 		/* FALLTHROUGH */
1188 	case S_IFREG:
1189 		error = kern_openat(td, dfd, path, seg,
1190 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1191 		if (error == 0)
1192 			kern_close(td, td->td_retval[0]);
1193 		break;
1194 
1195 	default:
1196 		error = EINVAL;
1197 		break;
1198 	}
1199 	if (convpath)
1200 		LFREEPATH(path);
1201 	return (error);
1202 }
1203 
1204 /*
1205  * UGH! This is just about the dumbest idea I've ever heard!!
1206  */
1207 int
1208 linux_personality(struct thread *td, struct linux_personality_args *args)
1209 {
1210 	struct linux_pemuldata *pem;
1211 	struct proc *p = td->td_proc;
1212 	uint32_t old;
1213 
1214 	PROC_LOCK(p);
1215 	pem = pem_find(p);
1216 	old = pem->persona;
1217 	if (args->per != 0xffffffff)
1218 		pem->persona = args->per;
1219 	PROC_UNLOCK(p);
1220 
1221 	td->td_retval[0] = old;
1222 	return (0);
1223 }
1224 
1225 struct l_itimerval {
1226 	l_timeval it_interval;
1227 	l_timeval it_value;
1228 };
1229 
1230 #define	B2L_ITIMERVAL(bip, lip)						\
1231 	(bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec;		\
1232 	(bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec;	\
1233 	(bip)->it_value.tv_sec = (lip)->it_value.tv_sec;		\
1234 	(bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1235 
1236 int
1237 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1238 {
1239 	int error;
1240 	struct l_itimerval ls;
1241 	struct itimerval aitv, oitv;
1242 
1243 	if (uap->itv == NULL) {
1244 		uap->itv = uap->oitv;
1245 		return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1246 	}
1247 
1248 	error = copyin(uap->itv, &ls, sizeof(ls));
1249 	if (error != 0)
1250 		return (error);
1251 	B2L_ITIMERVAL(&aitv, &ls);
1252 	error = kern_setitimer(td, uap->which, &aitv, &oitv);
1253 	if (error != 0 || uap->oitv == NULL)
1254 		return (error);
1255 	B2L_ITIMERVAL(&ls, &oitv);
1256 
1257 	return (copyout(&ls, uap->oitv, sizeof(ls)));
1258 }
1259 
1260 int
1261 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1262 {
1263 	int error;
1264 	struct l_itimerval ls;
1265 	struct itimerval aitv;
1266 
1267 	error = kern_getitimer(td, uap->which, &aitv);
1268 	if (error != 0)
1269 		return (error);
1270 	B2L_ITIMERVAL(&ls, &aitv);
1271 	return (copyout(&ls, uap->itv, sizeof(ls)));
1272 }
1273 
1274 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1275 int
1276 linux_nice(struct thread *td, struct linux_nice_args *args)
1277 {
1278 
1279 	return (kern_setpriority(td, PRIO_PROCESS, 0, args->inc));
1280 }
1281 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1282 
1283 int
1284 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1285 {
1286 	struct ucred *newcred, *oldcred;
1287 	l_gid_t *linux_gidset;
1288 	gid_t *bsd_gidset;
1289 	int ngrp, error;
1290 	struct proc *p;
1291 
1292 	ngrp = args->gidsetsize;
1293 	if (ngrp < 0 || ngrp >= ngroups_max + 1)
1294 		return (EINVAL);
1295 	linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK);
1296 	error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1297 	if (error)
1298 		goto out;
1299 	newcred = crget();
1300 	crextend(newcred, ngrp + 1);
1301 	p = td->td_proc;
1302 	PROC_LOCK(p);
1303 	oldcred = p->p_ucred;
1304 	crcopy(newcred, oldcred);
1305 
1306 	/*
1307 	 * cr_groups[0] holds egid. Setting the whole set from
1308 	 * the supplied set will cause egid to be changed too.
1309 	 * Keep cr_groups[0] unchanged to prevent that.
1310 	 */
1311 
1312 	if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS)) != 0) {
1313 		PROC_UNLOCK(p);
1314 		crfree(newcred);
1315 		goto out;
1316 	}
1317 
1318 	if (ngrp > 0) {
1319 		newcred->cr_ngroups = ngrp + 1;
1320 
1321 		bsd_gidset = newcred->cr_groups;
1322 		ngrp--;
1323 		while (ngrp >= 0) {
1324 			bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1325 			ngrp--;
1326 		}
1327 	} else
1328 		newcred->cr_ngroups = 1;
1329 
1330 	setsugid(p);
1331 	proc_set_cred(p, newcred);
1332 	PROC_UNLOCK(p);
1333 	crfree(oldcred);
1334 	error = 0;
1335 out:
1336 	free(linux_gidset, M_LINUX);
1337 	return (error);
1338 }
1339 
1340 int
1341 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1342 {
1343 	struct ucred *cred;
1344 	l_gid_t *linux_gidset;
1345 	gid_t *bsd_gidset;
1346 	int bsd_gidsetsz, ngrp, error;
1347 
1348 	cred = td->td_ucred;
1349 	bsd_gidset = cred->cr_groups;
1350 	bsd_gidsetsz = cred->cr_ngroups - 1;
1351 
1352 	/*
1353 	 * cr_groups[0] holds egid. Returning the whole set
1354 	 * here will cause a duplicate. Exclude cr_groups[0]
1355 	 * to prevent that.
1356 	 */
1357 
1358 	if ((ngrp = args->gidsetsize) == 0) {
1359 		td->td_retval[0] = bsd_gidsetsz;
1360 		return (0);
1361 	}
1362 
1363 	if (ngrp < bsd_gidsetsz)
1364 		return (EINVAL);
1365 
1366 	ngrp = 0;
1367 	linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1368 	    M_LINUX, M_WAITOK);
1369 	while (ngrp < bsd_gidsetsz) {
1370 		linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1371 		ngrp++;
1372 	}
1373 
1374 	error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1375 	free(linux_gidset, M_LINUX);
1376 	if (error)
1377 		return (error);
1378 
1379 	td->td_retval[0] = ngrp;
1380 	return (0);
1381 }
1382 
1383 static bool
1384 linux_get_dummy_limit(l_uint resource, struct rlimit *rlim)
1385 {
1386 
1387 	if (linux_dummy_rlimits == 0)
1388 		return (false);
1389 
1390 	switch (resource) {
1391 	case LINUX_RLIMIT_LOCKS:
1392 	case LINUX_RLIMIT_SIGPENDING:
1393 	case LINUX_RLIMIT_MSGQUEUE:
1394 	case LINUX_RLIMIT_RTTIME:
1395 		rlim->rlim_cur = LINUX_RLIM_INFINITY;
1396 		rlim->rlim_max = LINUX_RLIM_INFINITY;
1397 		return (true);
1398 	case LINUX_RLIMIT_NICE:
1399 	case LINUX_RLIMIT_RTPRIO:
1400 		rlim->rlim_cur = 0;
1401 		rlim->rlim_max = 0;
1402 		return (true);
1403 	default:
1404 		return (false);
1405 	}
1406 }
1407 
1408 int
1409 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1410 {
1411 	struct rlimit bsd_rlim;
1412 	struct l_rlimit rlim;
1413 	u_int which;
1414 	int error;
1415 
1416 	if (args->resource >= LINUX_RLIM_NLIMITS)
1417 		return (EINVAL);
1418 
1419 	which = linux_to_bsd_resource[args->resource];
1420 	if (which == -1)
1421 		return (EINVAL);
1422 
1423 	error = copyin(args->rlim, &rlim, sizeof(rlim));
1424 	if (error)
1425 		return (error);
1426 
1427 	bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1428 	bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1429 	return (kern_setrlimit(td, which, &bsd_rlim));
1430 }
1431 
1432 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1433 int
1434 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1435 {
1436 	struct l_rlimit rlim;
1437 	struct rlimit bsd_rlim;
1438 	u_int which;
1439 
1440 	if (linux_get_dummy_limit(args->resource, &bsd_rlim)) {
1441 		rlim.rlim_cur = bsd_rlim.rlim_cur;
1442 		rlim.rlim_max = bsd_rlim.rlim_max;
1443 		return (copyout(&rlim, args->rlim, sizeof(rlim)));
1444 	}
1445 
1446 	if (args->resource >= LINUX_RLIM_NLIMITS)
1447 		return (EINVAL);
1448 
1449 	which = linux_to_bsd_resource[args->resource];
1450 	if (which == -1)
1451 		return (EINVAL);
1452 
1453 	lim_rlimit(td, which, &bsd_rlim);
1454 
1455 #ifdef COMPAT_LINUX32
1456 	rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1457 	if (rlim.rlim_cur == UINT_MAX)
1458 		rlim.rlim_cur = INT_MAX;
1459 	rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1460 	if (rlim.rlim_max == UINT_MAX)
1461 		rlim.rlim_max = INT_MAX;
1462 #else
1463 	rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1464 	if (rlim.rlim_cur == ULONG_MAX)
1465 		rlim.rlim_cur = LONG_MAX;
1466 	rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1467 	if (rlim.rlim_max == ULONG_MAX)
1468 		rlim.rlim_max = LONG_MAX;
1469 #endif
1470 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1471 }
1472 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1473 
1474 int
1475 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1476 {
1477 	struct l_rlimit rlim;
1478 	struct rlimit bsd_rlim;
1479 	u_int which;
1480 
1481 	if (linux_get_dummy_limit(args->resource, &bsd_rlim)) {
1482 		rlim.rlim_cur = bsd_rlim.rlim_cur;
1483 		rlim.rlim_max = bsd_rlim.rlim_max;
1484 		return (copyout(&rlim, args->rlim, sizeof(rlim)));
1485 	}
1486 
1487 	if (args->resource >= LINUX_RLIM_NLIMITS)
1488 		return (EINVAL);
1489 
1490 	which = linux_to_bsd_resource[args->resource];
1491 	if (which == -1)
1492 		return (EINVAL);
1493 
1494 	lim_rlimit(td, which, &bsd_rlim);
1495 
1496 	rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1497 	rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1498 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1499 }
1500 
1501 int
1502 linux_sched_setscheduler(struct thread *td,
1503     struct linux_sched_setscheduler_args *args)
1504 {
1505 	struct sched_param sched_param;
1506 	struct thread *tdt;
1507 	int error, policy;
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 	if (linux_map_sched_prio) {
1528 		switch (policy) {
1529 		case SCHED_OTHER:
1530 			if (sched_param.sched_priority != 0)
1531 				return (EINVAL);
1532 
1533 			sched_param.sched_priority =
1534 			    PRI_MAX_TIMESHARE - PRI_MIN_TIMESHARE;
1535 			break;
1536 		case SCHED_FIFO:
1537 		case SCHED_RR:
1538 			if (sched_param.sched_priority < 1 ||
1539 			    sched_param.sched_priority >= LINUX_MAX_RT_PRIO)
1540 				return (EINVAL);
1541 
1542 			/*
1543 			 * Map [1, LINUX_MAX_RT_PRIO - 1] to
1544 			 * [0, RTP_PRIO_MAX - RTP_PRIO_MIN] (rounding down).
1545 			 */
1546 			sched_param.sched_priority =
1547 			    (sched_param.sched_priority - 1) *
1548 			    (RTP_PRIO_MAX - RTP_PRIO_MIN + 1) /
1549 			    (LINUX_MAX_RT_PRIO - 1);
1550 			break;
1551 		}
1552 	}
1553 
1554 	tdt = linux_tdfind(td, args->pid, -1);
1555 	if (tdt == NULL)
1556 		return (ESRCH);
1557 
1558 	error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1559 	PROC_UNLOCK(tdt->td_proc);
1560 	return (error);
1561 }
1562 
1563 int
1564 linux_sched_getscheduler(struct thread *td,
1565     struct linux_sched_getscheduler_args *args)
1566 {
1567 	struct thread *tdt;
1568 	int error, policy;
1569 
1570 	tdt = linux_tdfind(td, args->pid, -1);
1571 	if (tdt == NULL)
1572 		return (ESRCH);
1573 
1574 	error = kern_sched_getscheduler(td, tdt, &policy);
1575 	PROC_UNLOCK(tdt->td_proc);
1576 
1577 	switch (policy) {
1578 	case SCHED_OTHER:
1579 		td->td_retval[0] = LINUX_SCHED_OTHER;
1580 		break;
1581 	case SCHED_FIFO:
1582 		td->td_retval[0] = LINUX_SCHED_FIFO;
1583 		break;
1584 	case SCHED_RR:
1585 		td->td_retval[0] = LINUX_SCHED_RR;
1586 		break;
1587 	}
1588 	return (error);
1589 }
1590 
1591 int
1592 linux_sched_get_priority_max(struct thread *td,
1593     struct linux_sched_get_priority_max_args *args)
1594 {
1595 	struct sched_get_priority_max_args bsd;
1596 
1597 	if (linux_map_sched_prio) {
1598 		switch (args->policy) {
1599 		case LINUX_SCHED_OTHER:
1600 			td->td_retval[0] = 0;
1601 			return (0);
1602 		case LINUX_SCHED_FIFO:
1603 		case LINUX_SCHED_RR:
1604 			td->td_retval[0] = LINUX_MAX_RT_PRIO - 1;
1605 			return (0);
1606 		default:
1607 			return (EINVAL);
1608 		}
1609 	}
1610 
1611 	switch (args->policy) {
1612 	case LINUX_SCHED_OTHER:
1613 		bsd.policy = SCHED_OTHER;
1614 		break;
1615 	case LINUX_SCHED_FIFO:
1616 		bsd.policy = SCHED_FIFO;
1617 		break;
1618 	case LINUX_SCHED_RR:
1619 		bsd.policy = SCHED_RR;
1620 		break;
1621 	default:
1622 		return (EINVAL);
1623 	}
1624 	return (sys_sched_get_priority_max(td, &bsd));
1625 }
1626 
1627 int
1628 linux_sched_get_priority_min(struct thread *td,
1629     struct linux_sched_get_priority_min_args *args)
1630 {
1631 	struct sched_get_priority_min_args bsd;
1632 
1633 	if (linux_map_sched_prio) {
1634 		switch (args->policy) {
1635 		case LINUX_SCHED_OTHER:
1636 			td->td_retval[0] = 0;
1637 			return (0);
1638 		case LINUX_SCHED_FIFO:
1639 		case LINUX_SCHED_RR:
1640 			td->td_retval[0] = 1;
1641 			return (0);
1642 		default:
1643 			return (EINVAL);
1644 		}
1645 	}
1646 
1647 	switch (args->policy) {
1648 	case LINUX_SCHED_OTHER:
1649 		bsd.policy = SCHED_OTHER;
1650 		break;
1651 	case LINUX_SCHED_FIFO:
1652 		bsd.policy = SCHED_FIFO;
1653 		break;
1654 	case LINUX_SCHED_RR:
1655 		bsd.policy = SCHED_RR;
1656 		break;
1657 	default:
1658 		return (EINVAL);
1659 	}
1660 	return (sys_sched_get_priority_min(td, &bsd));
1661 }
1662 
1663 #define REBOOT_CAD_ON	0x89abcdef
1664 #define REBOOT_CAD_OFF	0
1665 #define REBOOT_HALT	0xcdef0123
1666 #define REBOOT_RESTART	0x01234567
1667 #define REBOOT_RESTART2	0xA1B2C3D4
1668 #define REBOOT_POWEROFF	0x4321FEDC
1669 #define REBOOT_MAGIC1	0xfee1dead
1670 #define REBOOT_MAGIC2	0x28121969
1671 #define REBOOT_MAGIC2A	0x05121996
1672 #define REBOOT_MAGIC2B	0x16041998
1673 
1674 int
1675 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1676 {
1677 	struct reboot_args bsd_args;
1678 
1679 	if (args->magic1 != REBOOT_MAGIC1)
1680 		return (EINVAL);
1681 
1682 	switch (args->magic2) {
1683 	case REBOOT_MAGIC2:
1684 	case REBOOT_MAGIC2A:
1685 	case REBOOT_MAGIC2B:
1686 		break;
1687 	default:
1688 		return (EINVAL);
1689 	}
1690 
1691 	switch (args->cmd) {
1692 	case REBOOT_CAD_ON:
1693 	case REBOOT_CAD_OFF:
1694 		return (priv_check(td, PRIV_REBOOT));
1695 	case REBOOT_HALT:
1696 		bsd_args.opt = RB_HALT;
1697 		break;
1698 	case REBOOT_RESTART:
1699 	case REBOOT_RESTART2:
1700 		bsd_args.opt = 0;
1701 		break;
1702 	case REBOOT_POWEROFF:
1703 		bsd_args.opt = RB_POWEROFF;
1704 		break;
1705 	default:
1706 		return (EINVAL);
1707 	}
1708 	return (sys_reboot(td, &bsd_args));
1709 }
1710 
1711 int
1712 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1713 {
1714 
1715 	td->td_retval[0] = td->td_proc->p_pid;
1716 
1717 	return (0);
1718 }
1719 
1720 int
1721 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1722 {
1723 	struct linux_emuldata *em;
1724 
1725 	em = em_find(td);
1726 	KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1727 
1728 	td->td_retval[0] = em->em_tid;
1729 
1730 	return (0);
1731 }
1732 
1733 int
1734 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1735 {
1736 
1737 	td->td_retval[0] = kern_getppid(td);
1738 	return (0);
1739 }
1740 
1741 int
1742 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1743 {
1744 
1745 	td->td_retval[0] = td->td_ucred->cr_rgid;
1746 	return (0);
1747 }
1748 
1749 int
1750 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1751 {
1752 
1753 	td->td_retval[0] = td->td_ucred->cr_ruid;
1754 	return (0);
1755 }
1756 
1757 int
1758 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1759 {
1760 
1761 	return (kern_getsid(td, args->pid));
1762 }
1763 
1764 int
1765 linux_nosys(struct thread *td, struct nosys_args *ignore)
1766 {
1767 
1768 	return (ENOSYS);
1769 }
1770 
1771 int
1772 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1773 {
1774 	int error;
1775 
1776 	error = kern_getpriority(td, args->which, args->who);
1777 	td->td_retval[0] = 20 - td->td_retval[0];
1778 	return (error);
1779 }
1780 
1781 int
1782 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1783 {
1784 	int name[2];
1785 
1786 	name[0] = CTL_KERN;
1787 	name[1] = KERN_HOSTNAME;
1788 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1789 	    args->len, 0, 0));
1790 }
1791 
1792 int
1793 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1794 {
1795 	int name[2];
1796 
1797 	name[0] = CTL_KERN;
1798 	name[1] = KERN_NISDOMAINNAME;
1799 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1800 	    args->len, 0, 0));
1801 }
1802 
1803 int
1804 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1805 {
1806 
1807 	LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1808 	    args->error_code);
1809 
1810 	/*
1811 	 * XXX: we should send a signal to the parent if
1812 	 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1813 	 * as it doesnt occur often.
1814 	 */
1815 	exit1(td, args->error_code, 0);
1816 		/* NOTREACHED */
1817 }
1818 
1819 #define _LINUX_CAPABILITY_VERSION_1  0x19980330
1820 #define _LINUX_CAPABILITY_VERSION_2  0x20071026
1821 #define _LINUX_CAPABILITY_VERSION_3  0x20080522
1822 
1823 struct l_user_cap_header {
1824 	l_int	version;
1825 	l_int	pid;
1826 };
1827 
1828 struct l_user_cap_data {
1829 	l_int	effective;
1830 	l_int	permitted;
1831 	l_int	inheritable;
1832 };
1833 
1834 int
1835 linux_capget(struct thread *td, struct linux_capget_args *uap)
1836 {
1837 	struct l_user_cap_header luch;
1838 	struct l_user_cap_data lucd[2];
1839 	int error, u32s;
1840 
1841 	if (uap->hdrp == NULL)
1842 		return (EFAULT);
1843 
1844 	error = copyin(uap->hdrp, &luch, sizeof(luch));
1845 	if (error != 0)
1846 		return (error);
1847 
1848 	switch (luch.version) {
1849 	case _LINUX_CAPABILITY_VERSION_1:
1850 		u32s = 1;
1851 		break;
1852 	case _LINUX_CAPABILITY_VERSION_2:
1853 	case _LINUX_CAPABILITY_VERSION_3:
1854 		u32s = 2;
1855 		break;
1856 	default:
1857 		luch.version = _LINUX_CAPABILITY_VERSION_1;
1858 		error = copyout(&luch, uap->hdrp, sizeof(luch));
1859 		if (error)
1860 			return (error);
1861 		return (EINVAL);
1862 	}
1863 
1864 	if (luch.pid)
1865 		return (EPERM);
1866 
1867 	if (uap->datap) {
1868 		/*
1869 		 * The current implementation doesn't support setting
1870 		 * a capability (it's essentially a stub) so indicate
1871 		 * that no capabilities are currently set or available
1872 		 * to request.
1873 		 */
1874 		memset(&lucd, 0, u32s * sizeof(lucd[0]));
1875 		error = copyout(&lucd, uap->datap, u32s * sizeof(lucd[0]));
1876 	}
1877 
1878 	return (error);
1879 }
1880 
1881 int
1882 linux_capset(struct thread *td, struct linux_capset_args *uap)
1883 {
1884 	struct l_user_cap_header luch;
1885 	struct l_user_cap_data lucd[2];
1886 	int error, i, u32s;
1887 
1888 	if (uap->hdrp == NULL || uap->datap == NULL)
1889 		return (EFAULT);
1890 
1891 	error = copyin(uap->hdrp, &luch, sizeof(luch));
1892 	if (error != 0)
1893 		return (error);
1894 
1895 	switch (luch.version) {
1896 	case _LINUX_CAPABILITY_VERSION_1:
1897 		u32s = 1;
1898 		break;
1899 	case _LINUX_CAPABILITY_VERSION_2:
1900 	case _LINUX_CAPABILITY_VERSION_3:
1901 		u32s = 2;
1902 		break;
1903 	default:
1904 		luch.version = _LINUX_CAPABILITY_VERSION_1;
1905 		error = copyout(&luch, uap->hdrp, sizeof(luch));
1906 		if (error)
1907 			return (error);
1908 		return (EINVAL);
1909 	}
1910 
1911 	if (luch.pid)
1912 		return (EPERM);
1913 
1914 	error = copyin(uap->datap, &lucd, u32s * sizeof(lucd[0]));
1915 	if (error != 0)
1916 		return (error);
1917 
1918 	/* We currently don't support setting any capabilities. */
1919 	for (i = 0; i < u32s; i++) {
1920 		if (lucd[i].effective || lucd[i].permitted ||
1921 		    lucd[i].inheritable) {
1922 			linux_msg(td,
1923 			    "capset[%d] effective=0x%x, permitted=0x%x, "
1924 			    "inheritable=0x%x is not implemented", i,
1925 			    (int)lucd[i].effective, (int)lucd[i].permitted,
1926 			    (int)lucd[i].inheritable);
1927 			return (EPERM);
1928 		}
1929 	}
1930 
1931 	return (0);
1932 }
1933 
1934 int
1935 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1936 {
1937 	int error = 0, max_size;
1938 	struct proc *p = td->td_proc;
1939 	char comm[LINUX_MAX_COMM_LEN];
1940 	int pdeath_signal;
1941 
1942 	switch (args->option) {
1943 	case LINUX_PR_SET_PDEATHSIG:
1944 		if (!LINUX_SIG_VALID(args->arg2))
1945 			return (EINVAL);
1946 		pdeath_signal = linux_to_bsd_signal(args->arg2);
1947 		return (kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_CTL,
1948 		    &pdeath_signal));
1949 	case LINUX_PR_GET_PDEATHSIG:
1950 		error = kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_STATUS,
1951 		    &pdeath_signal);
1952 		if (error != 0)
1953 			return (error);
1954 		pdeath_signal = bsd_to_linux_signal(pdeath_signal);
1955 		return (copyout(&pdeath_signal,
1956 		    (void *)(register_t)args->arg2,
1957 		    sizeof(pdeath_signal)));
1958 		break;
1959 	case LINUX_PR_SET_DUMPABLE:
1960 		linux_msg(td, "unsupported prctl PR_SET_DUMPABLE");
1961 		error = EINVAL;
1962 		break;
1963 	case LINUX_PR_GET_KEEPCAPS:
1964 		/*
1965 		 * Indicate that we always clear the effective and
1966 		 * permitted capability sets when the user id becomes
1967 		 * non-zero (actually the capability sets are simply
1968 		 * always zero in the current implementation).
1969 		 */
1970 		td->td_retval[0] = 0;
1971 		break;
1972 	case LINUX_PR_SET_KEEPCAPS:
1973 		/*
1974 		 * Ignore requests to keep the effective and permitted
1975 		 * capability sets when the user id becomes non-zero.
1976 		 */
1977 		break;
1978 	case LINUX_PR_SET_NAME:
1979 		/*
1980 		 * To be on the safe side we need to make sure to not
1981 		 * overflow the size a Linux program expects. We already
1982 		 * do this here in the copyin, so that we don't need to
1983 		 * check on copyout.
1984 		 */
1985 		max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1986 		error = copyinstr((void *)(register_t)args->arg2, comm,
1987 		    max_size, NULL);
1988 
1989 		/* Linux silently truncates the name if it is too long. */
1990 		if (error == ENAMETOOLONG) {
1991 			/*
1992 			 * XXX: copyinstr() isn't documented to populate the
1993 			 * array completely, so do a copyin() to be on the
1994 			 * safe side. This should be changed in case
1995 			 * copyinstr() is changed to guarantee this.
1996 			 */
1997 			error = copyin((void *)(register_t)args->arg2, comm,
1998 			    max_size - 1);
1999 			comm[max_size - 1] = '\0';
2000 		}
2001 		if (error)
2002 			return (error);
2003 
2004 		PROC_LOCK(p);
2005 		strlcpy(p->p_comm, comm, sizeof(p->p_comm));
2006 		PROC_UNLOCK(p);
2007 		break;
2008 	case LINUX_PR_GET_NAME:
2009 		PROC_LOCK(p);
2010 		strlcpy(comm, p->p_comm, sizeof(comm));
2011 		PROC_UNLOCK(p);
2012 		error = copyout(comm, (void *)(register_t)args->arg2,
2013 		    strlen(comm) + 1);
2014 		break;
2015 	case LINUX_PR_GET_SECCOMP:
2016 	case LINUX_PR_SET_SECCOMP:
2017 		/*
2018 		 * Same as returned by Linux without CONFIG_SECCOMP enabled.
2019 		 */
2020 		error = EINVAL;
2021 		break;
2022 	case LINUX_PR_SET_NO_NEW_PRIVS:
2023 		linux_msg(td, "unsupported prctl PR_SET_NO_NEW_PRIVS");
2024 		error = EINVAL;
2025 		break;
2026 	case LINUX_PR_SET_PTRACER:
2027 		linux_msg(td, "unsupported prctl PR_SET_PTRACER");
2028 		error = EINVAL;
2029 		break;
2030 	default:
2031 		linux_msg(td, "unsupported prctl option %d", args->option);
2032 		error = EINVAL;
2033 		break;
2034 	}
2035 
2036 	return (error);
2037 }
2038 
2039 int
2040 linux_sched_setparam(struct thread *td,
2041     struct linux_sched_setparam_args *uap)
2042 {
2043 	struct sched_param sched_param;
2044 	struct thread *tdt;
2045 	int error, policy;
2046 
2047 	error = copyin(uap->param, &sched_param, sizeof(sched_param));
2048 	if (error)
2049 		return (error);
2050 
2051 	tdt = linux_tdfind(td, uap->pid, -1);
2052 	if (tdt == NULL)
2053 		return (ESRCH);
2054 
2055 	if (linux_map_sched_prio) {
2056 		error = kern_sched_getscheduler(td, tdt, &policy);
2057 		if (error)
2058 			goto out;
2059 
2060 		switch (policy) {
2061 		case SCHED_OTHER:
2062 			if (sched_param.sched_priority != 0) {
2063 				error = EINVAL;
2064 				goto out;
2065 			}
2066 			sched_param.sched_priority =
2067 			    PRI_MAX_TIMESHARE - PRI_MIN_TIMESHARE;
2068 			break;
2069 		case SCHED_FIFO:
2070 		case SCHED_RR:
2071 			if (sched_param.sched_priority < 1 ||
2072 			    sched_param.sched_priority >= LINUX_MAX_RT_PRIO) {
2073 				error = EINVAL;
2074 				goto out;
2075 			}
2076 			/*
2077 			 * Map [1, LINUX_MAX_RT_PRIO - 1] to
2078 			 * [0, RTP_PRIO_MAX - RTP_PRIO_MIN] (rounding down).
2079 			 */
2080 			sched_param.sched_priority =
2081 			    (sched_param.sched_priority - 1) *
2082 			    (RTP_PRIO_MAX - RTP_PRIO_MIN + 1) /
2083 			    (LINUX_MAX_RT_PRIO - 1);
2084 			break;
2085 		}
2086 	}
2087 
2088 	error = kern_sched_setparam(td, tdt, &sched_param);
2089 out:	PROC_UNLOCK(tdt->td_proc);
2090 	return (error);
2091 }
2092 
2093 int
2094 linux_sched_getparam(struct thread *td,
2095     struct linux_sched_getparam_args *uap)
2096 {
2097 	struct sched_param sched_param;
2098 	struct thread *tdt;
2099 	int error, policy;
2100 
2101 	tdt = linux_tdfind(td, uap->pid, -1);
2102 	if (tdt == NULL)
2103 		return (ESRCH);
2104 
2105 	error = kern_sched_getparam(td, tdt, &sched_param);
2106 	if (error) {
2107 		PROC_UNLOCK(tdt->td_proc);
2108 		return (error);
2109 	}
2110 
2111 	if (linux_map_sched_prio) {
2112 		error = kern_sched_getscheduler(td, tdt, &policy);
2113 		PROC_UNLOCK(tdt->td_proc);
2114 		if (error)
2115 			return (error);
2116 
2117 		switch (policy) {
2118 		case SCHED_OTHER:
2119 			sched_param.sched_priority = 0;
2120 			break;
2121 		case SCHED_FIFO:
2122 		case SCHED_RR:
2123 			/*
2124 			 * Map [0, RTP_PRIO_MAX - RTP_PRIO_MIN] to
2125 			 * [1, LINUX_MAX_RT_PRIO - 1] (rounding up).
2126 			 */
2127 			sched_param.sched_priority =
2128 			    (sched_param.sched_priority *
2129 			    (LINUX_MAX_RT_PRIO - 1) +
2130 			    (RTP_PRIO_MAX - RTP_PRIO_MIN - 1)) /
2131 			    (RTP_PRIO_MAX - RTP_PRIO_MIN) + 1;
2132 			break;
2133 		}
2134 	} else
2135 		PROC_UNLOCK(tdt->td_proc);
2136 
2137 	error = copyout(&sched_param, uap->param, sizeof(sched_param));
2138 	return (error);
2139 }
2140 
2141 /*
2142  * Get affinity of a process.
2143  */
2144 int
2145 linux_sched_getaffinity(struct thread *td,
2146     struct linux_sched_getaffinity_args *args)
2147 {
2148 	int error;
2149 	struct thread *tdt;
2150 
2151 	if (args->len < sizeof(cpuset_t))
2152 		return (EINVAL);
2153 
2154 	tdt = linux_tdfind(td, args->pid, -1);
2155 	if (tdt == NULL)
2156 		return (ESRCH);
2157 
2158 	PROC_UNLOCK(tdt->td_proc);
2159 
2160 	error = kern_cpuset_getaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2161 	    tdt->td_tid, sizeof(cpuset_t), (cpuset_t *)args->user_mask_ptr);
2162 	if (error == 0)
2163 		td->td_retval[0] = sizeof(cpuset_t);
2164 
2165 	return (error);
2166 }
2167 
2168 /*
2169  *  Set affinity of a process.
2170  */
2171 int
2172 linux_sched_setaffinity(struct thread *td,
2173     struct linux_sched_setaffinity_args *args)
2174 {
2175 	struct thread *tdt;
2176 
2177 	if (args->len < sizeof(cpuset_t))
2178 		return (EINVAL);
2179 
2180 	tdt = linux_tdfind(td, args->pid, -1);
2181 	if (tdt == NULL)
2182 		return (ESRCH);
2183 
2184 	PROC_UNLOCK(tdt->td_proc);
2185 
2186 	return (kern_cpuset_setaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2187 	    tdt->td_tid, sizeof(cpuset_t), (cpuset_t *) args->user_mask_ptr));
2188 }
2189 
2190 struct linux_rlimit64 {
2191 	uint64_t	rlim_cur;
2192 	uint64_t	rlim_max;
2193 };
2194 
2195 int
2196 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
2197 {
2198 	struct rlimit rlim, nrlim;
2199 	struct linux_rlimit64 lrlim;
2200 	struct proc *p;
2201 	u_int which;
2202 	int flags;
2203 	int error;
2204 
2205 	if (args->new == NULL && args->old != NULL) {
2206 		if (linux_get_dummy_limit(args->resource, &rlim)) {
2207 			lrlim.rlim_cur = rlim.rlim_cur;
2208 			lrlim.rlim_max = rlim.rlim_max;
2209 			return (copyout(&lrlim, args->old, sizeof(lrlim)));
2210 		}
2211 	}
2212 
2213 	if (args->resource >= LINUX_RLIM_NLIMITS)
2214 		return (EINVAL);
2215 
2216 	which = linux_to_bsd_resource[args->resource];
2217 	if (which == -1)
2218 		return (EINVAL);
2219 
2220 	if (args->new != NULL) {
2221 		/*
2222 		 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
2223 		 * rlim is unsigned 64-bit. FreeBSD treats negative limits
2224 		 * as INFINITY so we do not need a conversion even.
2225 		 */
2226 		error = copyin(args->new, &nrlim, sizeof(nrlim));
2227 		if (error != 0)
2228 			return (error);
2229 	}
2230 
2231 	flags = PGET_HOLD | PGET_NOTWEXIT;
2232 	if (args->new != NULL)
2233 		flags |= PGET_CANDEBUG;
2234 	else
2235 		flags |= PGET_CANSEE;
2236 	if (args->pid == 0) {
2237 		p = td->td_proc;
2238 		PHOLD(p);
2239 	} else {
2240 		error = pget(args->pid, flags, &p);
2241 		if (error != 0)
2242 			return (error);
2243 	}
2244 	if (args->old != NULL) {
2245 		PROC_LOCK(p);
2246 		lim_rlimit_proc(p, which, &rlim);
2247 		PROC_UNLOCK(p);
2248 		if (rlim.rlim_cur == RLIM_INFINITY)
2249 			lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2250 		else
2251 			lrlim.rlim_cur = rlim.rlim_cur;
2252 		if (rlim.rlim_max == RLIM_INFINITY)
2253 			lrlim.rlim_max = LINUX_RLIM_INFINITY;
2254 		else
2255 			lrlim.rlim_max = rlim.rlim_max;
2256 		error = copyout(&lrlim, args->old, sizeof(lrlim));
2257 		if (error != 0)
2258 			goto out;
2259 	}
2260 
2261 	if (args->new != NULL)
2262 		error = kern_proc_setrlimit(td, p, which, &nrlim);
2263 
2264  out:
2265 	PRELE(p);
2266 	return (error);
2267 }
2268 
2269 int
2270 linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2271 {
2272 	struct timeval utv, tv0, tv1, *tvp;
2273 	struct l_pselect6arg lpse6;
2274 	struct l_timespec lts;
2275 	struct timespec uts;
2276 	l_sigset_t l_ss;
2277 	sigset_t *ssp;
2278 	sigset_t ss;
2279 	int error;
2280 
2281 	ssp = NULL;
2282 	if (args->sig != NULL) {
2283 		error = copyin(args->sig, &lpse6, sizeof(lpse6));
2284 		if (error != 0)
2285 			return (error);
2286 		if (lpse6.ss_len != sizeof(l_ss))
2287 			return (EINVAL);
2288 		if (lpse6.ss != 0) {
2289 			error = copyin(PTRIN(lpse6.ss), &l_ss,
2290 			    sizeof(l_ss));
2291 			if (error != 0)
2292 				return (error);
2293 			linux_to_bsd_sigset(&l_ss, &ss);
2294 			ssp = &ss;
2295 		}
2296 	}
2297 
2298 	/*
2299 	 * Currently glibc changes nanosecond number to microsecond.
2300 	 * This mean losing precision but for now it is hardly seen.
2301 	 */
2302 	if (args->tsp != NULL) {
2303 		error = copyin(args->tsp, &lts, sizeof(lts));
2304 		if (error != 0)
2305 			return (error);
2306 		error = linux_to_native_timespec(&uts, &lts);
2307 		if (error != 0)
2308 			return (error);
2309 
2310 		TIMESPEC_TO_TIMEVAL(&utv, &uts);
2311 		if (itimerfix(&utv))
2312 			return (EINVAL);
2313 
2314 		microtime(&tv0);
2315 		tvp = &utv;
2316 	} else
2317 		tvp = NULL;
2318 
2319 	error = kern_pselect(td, args->nfds, args->readfds, args->writefds,
2320 	    args->exceptfds, tvp, ssp, LINUX_NFDBITS);
2321 
2322 	if (error == 0 && args->tsp != NULL) {
2323 		if (td->td_retval[0] != 0) {
2324 			/*
2325 			 * Compute how much time was left of the timeout,
2326 			 * by subtracting the current time and the time
2327 			 * before we started the call, and subtracting
2328 			 * that result from the user-supplied value.
2329 			 */
2330 
2331 			microtime(&tv1);
2332 			timevalsub(&tv1, &tv0);
2333 			timevalsub(&utv, &tv1);
2334 			if (utv.tv_sec < 0)
2335 				timevalclear(&utv);
2336 		} else
2337 			timevalclear(&utv);
2338 
2339 		TIMEVAL_TO_TIMESPEC(&utv, &uts);
2340 
2341 		error = native_to_linux_timespec(&lts, &uts);
2342 		if (error == 0)
2343 			error = copyout(&lts, args->tsp, sizeof(lts));
2344 	}
2345 
2346 	return (error);
2347 }
2348 
2349 int
2350 linux_ppoll(struct thread *td, struct linux_ppoll_args *args)
2351 {
2352 	struct timespec ts0, ts1;
2353 	struct l_timespec lts;
2354 	struct timespec uts, *tsp;
2355 	l_sigset_t l_ss;
2356 	sigset_t *ssp;
2357 	sigset_t ss;
2358 	int error;
2359 
2360 	if (args->sset != NULL) {
2361 		if (args->ssize != sizeof(l_ss))
2362 			return (EINVAL);
2363 		error = copyin(args->sset, &l_ss, sizeof(l_ss));
2364 		if (error)
2365 			return (error);
2366 		linux_to_bsd_sigset(&l_ss, &ss);
2367 		ssp = &ss;
2368 	} else
2369 		ssp = NULL;
2370 	if (args->tsp != NULL) {
2371 		error = copyin(args->tsp, &lts, sizeof(lts));
2372 		if (error)
2373 			return (error);
2374 		error = linux_to_native_timespec(&uts, &lts);
2375 		if (error != 0)
2376 			return (error);
2377 
2378 		nanotime(&ts0);
2379 		tsp = &uts;
2380 	} else
2381 		tsp = NULL;
2382 
2383 	error = kern_poll(td, args->fds, args->nfds, tsp, ssp);
2384 
2385 	if (error == 0 && args->tsp != NULL) {
2386 		if (td->td_retval[0]) {
2387 			nanotime(&ts1);
2388 			timespecsub(&ts1, &ts0, &ts1);
2389 			timespecsub(&uts, &ts1, &uts);
2390 			if (uts.tv_sec < 0)
2391 				timespecclear(&uts);
2392 		} else
2393 			timespecclear(&uts);
2394 
2395 		error = native_to_linux_timespec(&lts, &uts);
2396 		if (error == 0)
2397 			error = copyout(&lts, args->tsp, sizeof(lts));
2398 	}
2399 
2400 	return (error);
2401 }
2402 
2403 int
2404 linux_sched_rr_get_interval(struct thread *td,
2405     struct linux_sched_rr_get_interval_args *uap)
2406 {
2407 	struct timespec ts;
2408 	struct l_timespec lts;
2409 	struct thread *tdt;
2410 	int error;
2411 
2412 	/*
2413 	 * According to man in case the invalid pid specified
2414 	 * EINVAL should be returned.
2415 	 */
2416 	if (uap->pid < 0)
2417 		return (EINVAL);
2418 
2419 	tdt = linux_tdfind(td, uap->pid, -1);
2420 	if (tdt == NULL)
2421 		return (ESRCH);
2422 
2423 	error = kern_sched_rr_get_interval_td(td, tdt, &ts);
2424 	PROC_UNLOCK(tdt->td_proc);
2425 	if (error != 0)
2426 		return (error);
2427 	error = native_to_linux_timespec(&lts, &ts);
2428 	if (error != 0)
2429 		return (error);
2430 	return (copyout(&lts, uap->interval, sizeof(lts)));
2431 }
2432 
2433 /*
2434  * In case when the Linux thread is the initial thread in
2435  * the thread group thread id is equal to the process id.
2436  * Glibc depends on this magic (assert in pthread_getattr_np.c).
2437  */
2438 struct thread *
2439 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid)
2440 {
2441 	struct linux_emuldata *em;
2442 	struct thread *tdt;
2443 	struct proc *p;
2444 
2445 	tdt = NULL;
2446 	if (tid == 0 || tid == td->td_tid) {
2447 		tdt = td;
2448 		PROC_LOCK(tdt->td_proc);
2449 	} else if (tid > PID_MAX)
2450 		tdt = tdfind(tid, pid);
2451 	else {
2452 		/*
2453 		 * Initial thread where the tid equal to the pid.
2454 		 */
2455 		p = pfind(tid);
2456 		if (p != NULL) {
2457 			if (SV_PROC_ABI(p) != SV_ABI_LINUX) {
2458 				/*
2459 				 * p is not a Linuxulator process.
2460 				 */
2461 				PROC_UNLOCK(p);
2462 				return (NULL);
2463 			}
2464 			FOREACH_THREAD_IN_PROC(p, tdt) {
2465 				em = em_find(tdt);
2466 				if (tid == em->em_tid)
2467 					return (tdt);
2468 			}
2469 			PROC_UNLOCK(p);
2470 		}
2471 		return (NULL);
2472 	}
2473 
2474 	return (tdt);
2475 }
2476 
2477 void
2478 linux_to_bsd_waitopts(int options, int *bsdopts)
2479 {
2480 
2481 	if (options & LINUX_WNOHANG)
2482 		*bsdopts |= WNOHANG;
2483 	if (options & LINUX_WUNTRACED)
2484 		*bsdopts |= WUNTRACED;
2485 	if (options & LINUX_WEXITED)
2486 		*bsdopts |= WEXITED;
2487 	if (options & LINUX_WCONTINUED)
2488 		*bsdopts |= WCONTINUED;
2489 	if (options & LINUX_WNOWAIT)
2490 		*bsdopts |= WNOWAIT;
2491 
2492 	if (options & __WCLONE)
2493 		*bsdopts |= WLINUXCLONE;
2494 }
2495 
2496 int
2497 linux_getrandom(struct thread *td, struct linux_getrandom_args *args)
2498 {
2499 	struct uio uio;
2500 	struct iovec iov;
2501 	int error;
2502 
2503 	if (args->flags & ~(LINUX_GRND_NONBLOCK|LINUX_GRND_RANDOM))
2504 		return (EINVAL);
2505 	if (args->count > INT_MAX)
2506 		args->count = INT_MAX;
2507 
2508 	iov.iov_base = args->buf;
2509 	iov.iov_len = args->count;
2510 
2511 	uio.uio_iov = &iov;
2512 	uio.uio_iovcnt = 1;
2513 	uio.uio_resid = iov.iov_len;
2514 	uio.uio_segflg = UIO_USERSPACE;
2515 	uio.uio_rw = UIO_READ;
2516 	uio.uio_td = td;
2517 
2518 	error = read_random_uio(&uio, args->flags & LINUX_GRND_NONBLOCK);
2519 	if (error == 0)
2520 		td->td_retval[0] = args->count - uio.uio_resid;
2521 	return (error);
2522 }
2523 
2524 int
2525 linux_mincore(struct thread *td, struct linux_mincore_args *args)
2526 {
2527 
2528 	/* Needs to be page-aligned */
2529 	if (args->start & PAGE_MASK)
2530 		return (EINVAL);
2531 	return (kern_mincore(td, args->start, args->len, args->vec));
2532 }
2533 
2534 #define	SYSLOG_TAG	"<6>"
2535 
2536 int
2537 linux_syslog(struct thread *td, struct linux_syslog_args *args)
2538 {
2539 	char buf[128], *src, *dst;
2540 	u_int seq;
2541 	int buflen, error;
2542 
2543 	if (args->type != LINUX_SYSLOG_ACTION_READ_ALL) {
2544 		linux_msg(td, "syslog unsupported type 0x%x", args->type);
2545 		return (EINVAL);
2546 	}
2547 
2548 	if (args->len < 6) {
2549 		td->td_retval[0] = 0;
2550 		return (0);
2551 	}
2552 
2553 	error = priv_check(td, PRIV_MSGBUF);
2554 	if (error)
2555 		return (error);
2556 
2557 	mtx_lock(&msgbuf_lock);
2558 	msgbuf_peekbytes(msgbufp, NULL, 0, &seq);
2559 	mtx_unlock(&msgbuf_lock);
2560 
2561 	dst = args->buf;
2562 	error = copyout(&SYSLOG_TAG, dst, sizeof(SYSLOG_TAG));
2563 	/* The -1 is to skip the trailing '\0'. */
2564 	dst += sizeof(SYSLOG_TAG) - 1;
2565 
2566 	while (error == 0) {
2567 		mtx_lock(&msgbuf_lock);
2568 		buflen = msgbuf_peekbytes(msgbufp, buf, sizeof(buf), &seq);
2569 		mtx_unlock(&msgbuf_lock);
2570 
2571 		if (buflen == 0)
2572 			break;
2573 
2574 		for (src = buf; src < buf + buflen && error == 0; src++) {
2575 			if (*src == '\0')
2576 				continue;
2577 
2578 			if (dst >= args->buf + args->len)
2579 				goto out;
2580 
2581 			error = copyout(src, dst, 1);
2582 			dst++;
2583 
2584 			if (*src == '\n' && *(src + 1) != '<' &&
2585 			    dst + sizeof(SYSLOG_TAG) < args->buf + args->len) {
2586 				error = copyout(&SYSLOG_TAG,
2587 				    dst, sizeof(SYSLOG_TAG));
2588 				dst += sizeof(SYSLOG_TAG) - 1;
2589 			}
2590 		}
2591 	}
2592 out:
2593 	td->td_retval[0] = dst - args->buf;
2594 	return (error);
2595 }
2596 
2597 int
2598 linux_getcpu(struct thread *td, struct linux_getcpu_args *args)
2599 {
2600 	int cpu, error, node;
2601 
2602 	cpu = td->td_oncpu; /* Make sure it doesn't change during copyout(9) */
2603 	error = 0;
2604 	node = cpuid_to_pcpu[cpu]->pc_domain;
2605 
2606 	if (args->cpu != NULL)
2607 		error = copyout(&cpu, args->cpu, sizeof(l_int));
2608 	if (args->node != NULL)
2609 		error = copyout(&node, args->node, sizeof(l_int));
2610 	return (error);
2611 }
2612