xref: /freebsd/sys/compat/linux/linux_misc.c (revision 82d4dc0621c92e3c05a86013eec35afbdec057a5)
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/poll.h>
54 #include <sys/priv.h>
55 #include <sys/proc.h>
56 #include <sys/procctl.h>
57 #include <sys/reboot.h>
58 #include <sys/racct.h>
59 #include <sys/random.h>
60 #include <sys/resourcevar.h>
61 #include <sys/sched.h>
62 #include <sys/sdt.h>
63 #include <sys/signalvar.h>
64 #include <sys/stat.h>
65 #include <sys/syscallsubr.h>
66 #include <sys/sysctl.h>
67 #include <sys/sysproto.h>
68 #include <sys/systm.h>
69 #include <sys/time.h>
70 #include <sys/vmmeter.h>
71 #include <sys/vnode.h>
72 #include <sys/wait.h>
73 #include <sys/cpuset.h>
74 #include <sys/uio.h>
75 
76 #include <security/mac/mac_framework.h>
77 
78 #include <vm/vm.h>
79 #include <vm/pmap.h>
80 #include <vm/vm_kern.h>
81 #include <vm/vm_map.h>
82 #include <vm/vm_extern.h>
83 #include <vm/swap_pager.h>
84 
85 #ifdef COMPAT_LINUX32
86 #include <machine/../linux32/linux.h>
87 #include <machine/../linux32/linux32_proto.h>
88 #else
89 #include <machine/../linux/linux.h>
90 #include <machine/../linux/linux_proto.h>
91 #endif
92 
93 #include <compat/linux/linux_common.h>
94 #include <compat/linux/linux_dtrace.h>
95 #include <compat/linux/linux_file.h>
96 #include <compat/linux/linux_mib.h>
97 #include <compat/linux/linux_signal.h>
98 #include <compat/linux/linux_timer.h>
99 #include <compat/linux/linux_util.h>
100 #include <compat/linux/linux_sysproto.h>
101 #include <compat/linux/linux_emul.h>
102 #include <compat/linux/linux_misc.h>
103 
104 int stclohz;				/* Statistics clock frequency */
105 
106 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
107 	RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
108 	RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
109 	RLIMIT_MEMLOCK, RLIMIT_AS
110 };
111 
112 struct l_sysinfo {
113 	l_long		uptime;		/* Seconds since boot */
114 	l_ulong		loads[3];	/* 1, 5, and 15 minute load averages */
115 #define LINUX_SYSINFO_LOADS_SCALE 65536
116 	l_ulong		totalram;	/* Total usable main memory size */
117 	l_ulong		freeram;	/* Available memory size */
118 	l_ulong		sharedram;	/* Amount of shared memory */
119 	l_ulong		bufferram;	/* Memory used by buffers */
120 	l_ulong		totalswap;	/* Total swap space size */
121 	l_ulong		freeswap;	/* swap space still available */
122 	l_ushort	procs;		/* Number of current processes */
123 	l_ushort	pads;
124 	l_ulong		totalhigh;
125 	l_ulong		freehigh;
126 	l_uint		mem_unit;
127 	char		_f[20-2*sizeof(l_long)-sizeof(l_int)];	/* padding */
128 };
129 
130 struct l_pselect6arg {
131 	l_uintptr_t	ss;
132 	l_size_t	ss_len;
133 };
134 
135 static int	linux_utimensat_lts_to_ts(struct l_timespec *,
136 			struct timespec *);
137 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
138 static int	linux_utimensat_lts64_to_ts(struct l_timespec64 *,
139 			struct timespec *);
140 #endif
141 static int	linux_common_utimensat(struct thread *, int,
142 			const char *, struct timespec *, int);
143 static int	linux_common_pselect6(struct thread *, l_int,
144 			l_fd_set *, l_fd_set *, l_fd_set *,
145 			struct timespec *, l_uintptr_t *);
146 static int	linux_common_ppoll(struct thread *, struct pollfd *,
147 			uint32_t, struct timespec *, l_sigset_t *,
148 			l_size_t);
149 static int	linux_pollin(struct thread *, struct pollfd *,
150 			struct pollfd *, u_int);
151 static int	linux_pollout(struct thread *, struct pollfd *,
152 			struct pollfd *, u_int);
153 
154 int
155 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
156 {
157 	struct l_sysinfo sysinfo;
158 	int i, j;
159 	struct timespec ts;
160 
161 	bzero(&sysinfo, sizeof(sysinfo));
162 	getnanouptime(&ts);
163 	if (ts.tv_nsec != 0)
164 		ts.tv_sec++;
165 	sysinfo.uptime = ts.tv_sec;
166 
167 	/* Use the information from the mib to get our load averages */
168 	for (i = 0; i < 3; i++)
169 		sysinfo.loads[i] = averunnable.ldavg[i] *
170 		    LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
171 
172 	sysinfo.totalram = physmem * PAGE_SIZE;
173 	sysinfo.freeram = (u_long)vm_free_count() * PAGE_SIZE;
174 
175 	/*
176 	 * sharedram counts pages allocated to named, swap-backed objects such
177 	 * as shared memory segments and tmpfs files.  There is no cheap way to
178 	 * compute this, so just leave the field unpopulated.  Linux itself only
179 	 * started setting this field in the 3.x timeframe.
180 	 */
181 	sysinfo.sharedram = 0;
182 	sysinfo.bufferram = 0;
183 
184 	swap_pager_status(&i, &j);
185 	sysinfo.totalswap = i * PAGE_SIZE;
186 	sysinfo.freeswap = (i - j) * PAGE_SIZE;
187 
188 	sysinfo.procs = nprocs;
189 
190 	/*
191 	 * Platforms supported by the emulation layer do not have a notion of
192 	 * high memory.
193 	 */
194 	sysinfo.totalhigh = 0;
195 	sysinfo.freehigh = 0;
196 
197 	sysinfo.mem_unit = 1;
198 
199 	return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
200 }
201 
202 #ifdef LINUX_LEGACY_SYSCALLS
203 int
204 linux_alarm(struct thread *td, struct linux_alarm_args *args)
205 {
206 	struct itimerval it, old_it;
207 	u_int secs;
208 	int error __diagused;
209 
210 	secs = args->secs;
211 	/*
212 	 * Linux alarm() is always successful. Limit secs to INT32_MAX / 2
213 	 * to match kern_setitimer()'s limit to avoid error from it.
214 	 *
215 	 * XXX. Linux limit secs to INT_MAX on 32 and does not limit on 64-bit
216 	 * platforms.
217 	 */
218 	if (secs > INT32_MAX / 2)
219 		secs = INT32_MAX / 2;
220 
221 	it.it_value.tv_sec = secs;
222 	it.it_value.tv_usec = 0;
223 	timevalclear(&it.it_interval);
224 	error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
225 	KASSERT(error == 0, ("kern_setitimer returns %d", error));
226 
227 	if ((old_it.it_value.tv_sec == 0 && old_it.it_value.tv_usec > 0) ||
228 	    old_it.it_value.tv_usec >= 500000)
229 		old_it.it_value.tv_sec++;
230 	td->td_retval[0] = old_it.it_value.tv_sec;
231 	return (0);
232 }
233 #endif
234 
235 int
236 linux_brk(struct thread *td, struct linux_brk_args *args)
237 {
238 	struct vmspace *vm = td->td_proc->p_vmspace;
239 	uintptr_t new, old;
240 
241 	old = (uintptr_t)vm->vm_daddr + ctob(vm->vm_dsize);
242 	new = (uintptr_t)args->dsend;
243 	if ((caddr_t)new > vm->vm_daddr && !kern_break(td, &new))
244 		td->td_retval[0] = (register_t)new;
245 	else
246 		td->td_retval[0] = (register_t)old;
247 
248 	return (0);
249 }
250 
251 #if defined(__i386__)
252 /* XXX: what about amd64/linux32? */
253 
254 int
255 linux_uselib(struct thread *td, struct linux_uselib_args *args)
256 {
257 	struct nameidata ni;
258 	struct vnode *vp;
259 	struct exec *a_out;
260 	vm_map_t map;
261 	vm_map_entry_t entry;
262 	struct vattr attr;
263 	vm_offset_t vmaddr;
264 	unsigned long file_offset;
265 	unsigned long bss_size;
266 	char *library;
267 	ssize_t aresid;
268 	int error;
269 	bool locked, opened, textset;
270 
271 	a_out = NULL;
272 	vp = NULL;
273 	locked = false;
274 	textset = false;
275 	opened = false;
276 
277 	if (!LUSECONVPATH(td)) {
278 		NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
279 		    UIO_USERSPACE, args->library);
280 		error = namei(&ni);
281 	} else {
282 		LCONVPATHEXIST(args->library, &library);
283 		NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
284 		    UIO_SYSSPACE, library);
285 		error = namei(&ni);
286 		LFREEPATH(library);
287 	}
288 	if (error)
289 		goto cleanup;
290 
291 	vp = ni.ni_vp;
292 	NDFREE_PNBUF(&ni);
293 
294 	/*
295 	 * From here on down, we have a locked vnode that must be unlocked.
296 	 * XXX: The code below largely duplicates exec_check_permissions().
297 	 */
298 	locked = true;
299 
300 	/* Executable? */
301 	error = VOP_GETATTR(vp, &attr, td->td_ucred);
302 	if (error)
303 		goto cleanup;
304 
305 	if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
306 	    ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
307 		/* EACCESS is what exec(2) returns. */
308 		error = ENOEXEC;
309 		goto cleanup;
310 	}
311 
312 	/* Sensible size? */
313 	if (attr.va_size == 0) {
314 		error = ENOEXEC;
315 		goto cleanup;
316 	}
317 
318 	/* Can we access it? */
319 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
320 	if (error)
321 		goto cleanup;
322 
323 	/*
324 	 * XXX: This should use vn_open() so that it is properly authorized,
325 	 * and to reduce code redundancy all over the place here.
326 	 * XXX: Not really, it duplicates far more of exec_check_permissions()
327 	 * than vn_open().
328 	 */
329 #ifdef MAC
330 	error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
331 	if (error)
332 		goto cleanup;
333 #endif
334 	error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
335 	if (error)
336 		goto cleanup;
337 	opened = true;
338 
339 	/* Pull in executable header into exec_map */
340 	error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
341 	    VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
342 	if (error)
343 		goto cleanup;
344 
345 	/* Is it a Linux binary ? */
346 	if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
347 		error = ENOEXEC;
348 		goto cleanup;
349 	}
350 
351 	/*
352 	 * While we are here, we should REALLY do some more checks
353 	 */
354 
355 	/* Set file/virtual offset based on a.out variant. */
356 	switch ((int)(a_out->a_magic & 0xffff)) {
357 	case 0413:			/* ZMAGIC */
358 		file_offset = 1024;
359 		break;
360 	case 0314:			/* QMAGIC */
361 		file_offset = 0;
362 		break;
363 	default:
364 		error = ENOEXEC;
365 		goto cleanup;
366 	}
367 
368 	bss_size = round_page(a_out->a_bss);
369 
370 	/* Check various fields in header for validity/bounds. */
371 	if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
372 		error = ENOEXEC;
373 		goto cleanup;
374 	}
375 
376 	/* text + data can't exceed file size */
377 	if (a_out->a_data + a_out->a_text > attr.va_size) {
378 		error = EFAULT;
379 		goto cleanup;
380 	}
381 
382 	/*
383 	 * text/data/bss must not exceed limits
384 	 * XXX - this is not complete. it should check current usage PLUS
385 	 * the resources needed by this library.
386 	 */
387 	PROC_LOCK(td->td_proc);
388 	if (a_out->a_text > maxtsiz ||
389 	    a_out->a_data + bss_size > lim_cur_proc(td->td_proc, RLIMIT_DATA) ||
390 	    racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
391 	    bss_size) != 0) {
392 		PROC_UNLOCK(td->td_proc);
393 		error = ENOMEM;
394 		goto cleanup;
395 	}
396 	PROC_UNLOCK(td->td_proc);
397 
398 	/*
399 	 * Prevent more writers.
400 	 */
401 	error = VOP_SET_TEXT(vp);
402 	if (error != 0)
403 		goto cleanup;
404 	textset = true;
405 
406 	/*
407 	 * Lock no longer needed
408 	 */
409 	locked = false;
410 	VOP_UNLOCK(vp);
411 
412 	/*
413 	 * Check if file_offset page aligned. Currently we cannot handle
414 	 * misalinged file offsets, and so we read in the entire image
415 	 * (what a waste).
416 	 */
417 	if (file_offset & PAGE_MASK) {
418 		/* Map text+data read/write/execute */
419 
420 		/* a_entry is the load address and is page aligned */
421 		vmaddr = trunc_page(a_out->a_entry);
422 
423 		/* get anon user mapping, read+write+execute */
424 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
425 		    &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
426 		    VM_PROT_ALL, VM_PROT_ALL, 0);
427 		if (error)
428 			goto cleanup;
429 
430 		error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
431 		    a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
432 		    td->td_ucred, NOCRED, &aresid, td);
433 		if (error != 0)
434 			goto cleanup;
435 		if (aresid != 0) {
436 			error = ENOEXEC;
437 			goto cleanup;
438 		}
439 	} else {
440 		/*
441 		 * for QMAGIC, a_entry is 20 bytes beyond the load address
442 		 * to skip the executable header
443 		 */
444 		vmaddr = trunc_page(a_out->a_entry);
445 
446 		/*
447 		 * Map it all into the process's space as a single
448 		 * copy-on-write "data" segment.
449 		 */
450 		map = &td->td_proc->p_vmspace->vm_map;
451 		error = vm_mmap(map, &vmaddr,
452 		    a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
453 		    MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
454 		if (error)
455 			goto cleanup;
456 		vm_map_lock(map);
457 		if (!vm_map_lookup_entry(map, vmaddr, &entry)) {
458 			vm_map_unlock(map);
459 			error = EDOOFUS;
460 			goto cleanup;
461 		}
462 		entry->eflags |= MAP_ENTRY_VN_EXEC;
463 		vm_map_unlock(map);
464 		textset = false;
465 	}
466 
467 	if (bss_size != 0) {
468 		/* Calculate BSS start address */
469 		vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
470 		    a_out->a_data;
471 
472 		/* allocate some 'anon' space */
473 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
474 		    &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
475 		    VM_PROT_ALL, 0);
476 		if (error)
477 			goto cleanup;
478 	}
479 
480 cleanup:
481 	if (opened) {
482 		if (locked)
483 			VOP_UNLOCK(vp);
484 		locked = false;
485 		VOP_CLOSE(vp, FREAD, td->td_ucred, td);
486 	}
487 	if (textset) {
488 		if (!locked) {
489 			locked = true;
490 			VOP_LOCK(vp, LK_SHARED | LK_RETRY);
491 		}
492 		VOP_UNSET_TEXT_CHECKED(vp);
493 	}
494 	if (locked)
495 		VOP_UNLOCK(vp);
496 
497 	/* Release the temporary mapping. */
498 	if (a_out)
499 		kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
500 
501 	return (error);
502 }
503 
504 #endif	/* __i386__ */
505 
506 #ifdef LINUX_LEGACY_SYSCALLS
507 int
508 linux_select(struct thread *td, struct linux_select_args *args)
509 {
510 	l_timeval ltv;
511 	struct timeval tv0, tv1, utv, *tvp;
512 	int error;
513 
514 	/*
515 	 * Store current time for computation of the amount of
516 	 * time left.
517 	 */
518 	if (args->timeout) {
519 		if ((error = copyin(args->timeout, &ltv, sizeof(ltv))))
520 			goto select_out;
521 		utv.tv_sec = ltv.tv_sec;
522 		utv.tv_usec = ltv.tv_usec;
523 
524 		if (itimerfix(&utv)) {
525 			/*
526 			 * The timeval was invalid.  Convert it to something
527 			 * valid that will act as it does under Linux.
528 			 */
529 			utv.tv_sec += utv.tv_usec / 1000000;
530 			utv.tv_usec %= 1000000;
531 			if (utv.tv_usec < 0) {
532 				utv.tv_sec -= 1;
533 				utv.tv_usec += 1000000;
534 			}
535 			if (utv.tv_sec < 0)
536 				timevalclear(&utv);
537 		}
538 		microtime(&tv0);
539 		tvp = &utv;
540 	} else
541 		tvp = NULL;
542 
543 	error = kern_select(td, args->nfds, args->readfds, args->writefds,
544 	    args->exceptfds, tvp, LINUX_NFDBITS);
545 	if (error)
546 		goto select_out;
547 
548 	if (args->timeout) {
549 		if (td->td_retval[0]) {
550 			/*
551 			 * Compute how much time was left of the timeout,
552 			 * by subtracting the current time and the time
553 			 * before we started the call, and subtracting
554 			 * that result from the user-supplied value.
555 			 */
556 			microtime(&tv1);
557 			timevalsub(&tv1, &tv0);
558 			timevalsub(&utv, &tv1);
559 			if (utv.tv_sec < 0)
560 				timevalclear(&utv);
561 		} else
562 			timevalclear(&utv);
563 		ltv.tv_sec = utv.tv_sec;
564 		ltv.tv_usec = utv.tv_usec;
565 		if ((error = copyout(&ltv, args->timeout, sizeof(ltv))))
566 			goto select_out;
567 	}
568 
569 select_out:
570 	return (error);
571 }
572 #endif
573 
574 int
575 linux_mremap(struct thread *td, struct linux_mremap_args *args)
576 {
577 	uintptr_t addr;
578 	size_t len;
579 	int error = 0;
580 
581 	if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
582 		td->td_retval[0] = 0;
583 		return (EINVAL);
584 	}
585 
586 	/*
587 	 * Check for the page alignment.
588 	 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
589 	 */
590 	if (args->addr & PAGE_MASK) {
591 		td->td_retval[0] = 0;
592 		return (EINVAL);
593 	}
594 
595 	args->new_len = round_page(args->new_len);
596 	args->old_len = round_page(args->old_len);
597 
598 	if (args->new_len > args->old_len) {
599 		td->td_retval[0] = 0;
600 		return (ENOMEM);
601 	}
602 
603 	if (args->new_len < args->old_len) {
604 		addr = args->addr + args->new_len;
605 		len = args->old_len - args->new_len;
606 		error = kern_munmap(td, addr, len);
607 	}
608 
609 	td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
610 	return (error);
611 }
612 
613 #define LINUX_MS_ASYNC       0x0001
614 #define LINUX_MS_INVALIDATE  0x0002
615 #define LINUX_MS_SYNC        0x0004
616 
617 int
618 linux_msync(struct thread *td, struct linux_msync_args *args)
619 {
620 
621 	return (kern_msync(td, args->addr, args->len,
622 	    args->fl & ~LINUX_MS_SYNC));
623 }
624 
625 #ifdef LINUX_LEGACY_SYSCALLS
626 int
627 linux_time(struct thread *td, struct linux_time_args *args)
628 {
629 	struct timeval tv;
630 	l_time_t tm;
631 	int error;
632 
633 	microtime(&tv);
634 	tm = tv.tv_sec;
635 	if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
636 		return (error);
637 	td->td_retval[0] = tm;
638 	return (0);
639 }
640 #endif
641 
642 struct l_times_argv {
643 	l_clock_t	tms_utime;
644 	l_clock_t	tms_stime;
645 	l_clock_t	tms_cutime;
646 	l_clock_t	tms_cstime;
647 };
648 
649 /*
650  * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
651  * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
652  * auxiliary vector entry.
653  */
654 #define	CLK_TCK		100
655 
656 #define	CONVOTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
657 #define	CONVNTCK(r)	(r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
658 
659 #define	CONVTCK(r)	(linux_kernver(td) >= LINUX_KERNVER_2004000 ?		\
660 			    CONVNTCK(r) : CONVOTCK(r))
661 
662 int
663 linux_times(struct thread *td, struct linux_times_args *args)
664 {
665 	struct timeval tv, utime, stime, cutime, cstime;
666 	struct l_times_argv tms;
667 	struct proc *p;
668 	int error;
669 
670 	if (args->buf != NULL) {
671 		p = td->td_proc;
672 		PROC_LOCK(p);
673 		PROC_STATLOCK(p);
674 		calcru(p, &utime, &stime);
675 		PROC_STATUNLOCK(p);
676 		calccru(p, &cutime, &cstime);
677 		PROC_UNLOCK(p);
678 
679 		tms.tms_utime = CONVTCK(utime);
680 		tms.tms_stime = CONVTCK(stime);
681 
682 		tms.tms_cutime = CONVTCK(cutime);
683 		tms.tms_cstime = CONVTCK(cstime);
684 
685 		if ((error = copyout(&tms, args->buf, sizeof(tms))))
686 			return (error);
687 	}
688 
689 	microuptime(&tv);
690 	td->td_retval[0] = (int)CONVTCK(tv);
691 	return (0);
692 }
693 
694 int
695 linux_newuname(struct thread *td, struct linux_newuname_args *args)
696 {
697 	struct l_new_utsname utsname;
698 	char osname[LINUX_MAX_UTSNAME];
699 	char osrelease[LINUX_MAX_UTSNAME];
700 	char *p;
701 
702 	linux_get_osname(td, osname);
703 	linux_get_osrelease(td, osrelease);
704 
705 	bzero(&utsname, sizeof(utsname));
706 	strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
707 	getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
708 	getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
709 	strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
710 	strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
711 	for (p = utsname.version; *p != '\0'; ++p)
712 		if (*p == '\n') {
713 			*p = '\0';
714 			break;
715 		}
716 #if defined(__amd64__)
717 	/*
718 	 * On amd64, Linux uname(2) needs to return "x86_64"
719 	 * for both 64-bit and 32-bit applications.  On 32-bit,
720 	 * the string returned by getauxval(AT_PLATFORM) needs
721 	 * to remain "i686", though.
722 	 */
723 #if defined(COMPAT_LINUX32)
724 	if (linux32_emulate_i386)
725 		strlcpy(utsname.machine, "i686", LINUX_MAX_UTSNAME);
726 	else
727 #endif
728 	strlcpy(utsname.machine, "x86_64", LINUX_MAX_UTSNAME);
729 #elif defined(__aarch64__)
730 	strlcpy(utsname.machine, "aarch64", LINUX_MAX_UTSNAME);
731 #elif defined(__i386__)
732 	strlcpy(utsname.machine, "i686", LINUX_MAX_UTSNAME);
733 #endif
734 
735 	return (copyout(&utsname, args->buf, sizeof(utsname)));
736 }
737 
738 struct l_utimbuf {
739 	l_time_t l_actime;
740 	l_time_t l_modtime;
741 };
742 
743 #ifdef LINUX_LEGACY_SYSCALLS
744 int
745 linux_utime(struct thread *td, struct linux_utime_args *args)
746 {
747 	struct timeval tv[2], *tvp;
748 	struct l_utimbuf lut;
749 	char *fname;
750 	int error;
751 
752 	if (args->times) {
753 		if ((error = copyin(args->times, &lut, sizeof lut)) != 0)
754 			return (error);
755 		tv[0].tv_sec = lut.l_actime;
756 		tv[0].tv_usec = 0;
757 		tv[1].tv_sec = lut.l_modtime;
758 		tv[1].tv_usec = 0;
759 		tvp = tv;
760 	} else
761 		tvp = NULL;
762 
763 	if (!LUSECONVPATH(td)) {
764 		error = kern_utimesat(td, AT_FDCWD, args->fname, UIO_USERSPACE,
765 		    tvp, UIO_SYSSPACE);
766 	} else {
767 		LCONVPATHEXIST(args->fname, &fname);
768 		error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, tvp,
769 		    UIO_SYSSPACE);
770 		LFREEPATH(fname);
771 	}
772 	return (error);
773 }
774 #endif
775 
776 #ifdef LINUX_LEGACY_SYSCALLS
777 int
778 linux_utimes(struct thread *td, struct linux_utimes_args *args)
779 {
780 	l_timeval ltv[2];
781 	struct timeval tv[2], *tvp = NULL;
782 	char *fname;
783 	int error;
784 
785 	if (args->tptr != NULL) {
786 		if ((error = copyin(args->tptr, ltv, sizeof ltv)) != 0)
787 			return (error);
788 		tv[0].tv_sec = ltv[0].tv_sec;
789 		tv[0].tv_usec = ltv[0].tv_usec;
790 		tv[1].tv_sec = ltv[1].tv_sec;
791 		tv[1].tv_usec = ltv[1].tv_usec;
792 		tvp = tv;
793 	}
794 
795 	if (!LUSECONVPATH(td)) {
796 		error = kern_utimesat(td, AT_FDCWD, args->fname, UIO_USERSPACE,
797 		    tvp, UIO_SYSSPACE);
798 	} else {
799 		LCONVPATHEXIST(args->fname, &fname);
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_lts_to_ts(struct l_timespec *l_times, struct timespec *times)
810 {
811 
812 	if (l_times->tv_nsec != LINUX_UTIME_OMIT &&
813 	    l_times->tv_nsec != LINUX_UTIME_NOW &&
814 	    (l_times->tv_nsec < 0 || l_times->tv_nsec > 999999999))
815 		return (EINVAL);
816 
817 	times->tv_sec = l_times->tv_sec;
818 	switch (l_times->tv_nsec)
819 	{
820 	case LINUX_UTIME_OMIT:
821 		times->tv_nsec = UTIME_OMIT;
822 		break;
823 	case LINUX_UTIME_NOW:
824 		times->tv_nsec = UTIME_NOW;
825 		break;
826 	default:
827 		times->tv_nsec = l_times->tv_nsec;
828 	}
829 
830 	return (0);
831 }
832 
833 static int
834 linux_common_utimensat(struct thread *td, int ldfd, const char *pathname,
835     struct timespec *timesp, int lflags)
836 {
837 	char *path = NULL;
838 	int error, dfd, flags = 0;
839 
840 	dfd = (ldfd == LINUX_AT_FDCWD) ? AT_FDCWD : ldfd;
841 
842 	if (lflags & ~(LINUX_AT_SYMLINK_NOFOLLOW | LINUX_AT_EMPTY_PATH))
843 		return (EINVAL);
844 
845 	if (timesp != NULL) {
846 		/* This breaks POSIX, but is what the Linux kernel does
847 		 * _on purpose_ (documented in the man page for utimensat(2)),
848 		 * so we must follow that behaviour. */
849 		if (timesp[0].tv_nsec == UTIME_OMIT &&
850 		    timesp[1].tv_nsec == UTIME_OMIT)
851 			return (0);
852 	}
853 
854 	if (lflags & LINUX_AT_SYMLINK_NOFOLLOW)
855 		flags |= AT_SYMLINK_NOFOLLOW;
856 	if (lflags & LINUX_AT_EMPTY_PATH)
857 		flags |= AT_EMPTY_PATH;
858 
859 	if (!LUSECONVPATH(td)) {
860 		if (pathname != NULL) {
861 			return (kern_utimensat(td, dfd, pathname,
862 			    UIO_USERSPACE, timesp, UIO_SYSSPACE, flags));
863 		}
864 	}
865 
866 	if (pathname != NULL)
867 		LCONVPATHEXIST_AT(pathname, &path, dfd);
868 	else if (lflags != 0)
869 		return (EINVAL);
870 
871 	if (path == NULL)
872 		error = kern_futimens(td, dfd, timesp, UIO_SYSSPACE);
873 	else {
874 		error = kern_utimensat(td, dfd, path, UIO_SYSSPACE, timesp,
875 			UIO_SYSSPACE, flags);
876 		LFREEPATH(path);
877 	}
878 
879 	return (error);
880 }
881 
882 int
883 linux_utimensat(struct thread *td, struct linux_utimensat_args *args)
884 {
885 	struct l_timespec l_times[2];
886 	struct timespec times[2], *timesp;
887 	int error;
888 
889 	if (args->times != NULL) {
890 		error = copyin(args->times, l_times, sizeof(l_times));
891 		if (error != 0)
892 			return (error);
893 
894 		error = linux_utimensat_lts_to_ts(&l_times[0], &times[0]);
895 		if (error != 0)
896 			return (error);
897 		error = linux_utimensat_lts_to_ts(&l_times[1], &times[1]);
898 		if (error != 0)
899 			return (error);
900 		timesp = times;
901 	} else
902 		timesp = NULL;
903 
904 	return (linux_common_utimensat(td, args->dfd, args->pathname,
905 	    timesp, args->flags));
906 }
907 
908 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
909 static int
910 linux_utimensat_lts64_to_ts(struct l_timespec64 *l_times, struct timespec *times)
911 {
912 
913 	/* Zero out the padding in compat mode. */
914 	l_times->tv_nsec &= 0xFFFFFFFFUL;
915 
916 	if (l_times->tv_nsec != LINUX_UTIME_OMIT &&
917 	    l_times->tv_nsec != LINUX_UTIME_NOW &&
918 	    (l_times->tv_nsec < 0 || l_times->tv_nsec > 999999999))
919 		return (EINVAL);
920 
921 	times->tv_sec = l_times->tv_sec;
922 	switch (l_times->tv_nsec)
923 	{
924 	case LINUX_UTIME_OMIT:
925 		times->tv_nsec = UTIME_OMIT;
926 		break;
927 	case LINUX_UTIME_NOW:
928 		times->tv_nsec = UTIME_NOW;
929 		break;
930 	default:
931 		times->tv_nsec = l_times->tv_nsec;
932 	}
933 
934 	return (0);
935 }
936 
937 int
938 linux_utimensat_time64(struct thread *td, struct linux_utimensat_time64_args *args)
939 {
940 	struct l_timespec64 l_times[2];
941 	struct timespec times[2], *timesp;
942 	int error;
943 
944 	if (args->times64 != NULL) {
945 		error = copyin(args->times64, l_times, sizeof(l_times));
946 		if (error != 0)
947 			return (error);
948 
949 		error = linux_utimensat_lts64_to_ts(&l_times[0], &times[0]);
950 		if (error != 0)
951 			return (error);
952 		error = linux_utimensat_lts64_to_ts(&l_times[1], &times[1]);
953 		if (error != 0)
954 			return (error);
955 		timesp = times;
956 	} else
957 		timesp = NULL;
958 
959 	return (linux_common_utimensat(td, args->dfd, args->pathname,
960 	    timesp, args->flags));
961 }
962 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
963 
964 #ifdef LINUX_LEGACY_SYSCALLS
965 int
966 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
967 {
968 	l_timeval ltv[2];
969 	struct timeval tv[2], *tvp = NULL;
970 	char *fname;
971 	int error, dfd;
972 
973 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
974 
975 	if (args->utimes != NULL) {
976 		if ((error = copyin(args->utimes, ltv, sizeof ltv)) != 0)
977 			return (error);
978 		tv[0].tv_sec = ltv[0].tv_sec;
979 		tv[0].tv_usec = ltv[0].tv_usec;
980 		tv[1].tv_sec = ltv[1].tv_sec;
981 		tv[1].tv_usec = ltv[1].tv_usec;
982 		tvp = tv;
983 	}
984 
985 	if (!LUSECONVPATH(td)) {
986 		error = kern_utimesat(td, dfd, args->filename, UIO_USERSPACE,
987 		    tvp, UIO_SYSSPACE);
988 	} else {
989 		LCONVPATHEXIST_AT(args->filename, &fname, dfd);
990 		error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE,
991 		    tvp, UIO_SYSSPACE);
992 		LFREEPATH(fname);
993 	}
994 	return (error);
995 }
996 #endif
997 
998 static int
999 linux_common_wait(struct thread *td, idtype_t idtype, int id, int *statusp,
1000     int options, void *rup, l_siginfo_t *infop)
1001 {
1002 	l_siginfo_t lsi;
1003 	siginfo_t siginfo;
1004 	struct __wrusage wru;
1005 	int error, status, tmpstat, sig;
1006 
1007 	error = kern_wait6(td, idtype, id, &status, options,
1008 	    rup != NULL ? &wru : NULL, &siginfo);
1009 
1010 	if (error == 0 && statusp) {
1011 		tmpstat = status & 0xffff;
1012 		if (WIFSIGNALED(tmpstat)) {
1013 			tmpstat = (tmpstat & 0xffffff80) |
1014 			    bsd_to_linux_signal(WTERMSIG(tmpstat));
1015 		} else if (WIFSTOPPED(tmpstat)) {
1016 			tmpstat = (tmpstat & 0xffff00ff) |
1017 			    (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8);
1018 #if defined(__aarch64__) || (defined(__amd64__) && !defined(COMPAT_LINUX32))
1019 			if (WSTOPSIG(status) == SIGTRAP) {
1020 				tmpstat = linux_ptrace_status(td,
1021 				    siginfo.si_pid, tmpstat);
1022 			}
1023 #endif
1024 		} else if (WIFCONTINUED(tmpstat)) {
1025 			tmpstat = 0xffff;
1026 		}
1027 		error = copyout(&tmpstat, statusp, sizeof(int));
1028 	}
1029 	if (error == 0 && rup != NULL)
1030 		error = linux_copyout_rusage(&wru.wru_self, rup);
1031 	if (error == 0 && infop != NULL && td->td_retval[0] != 0) {
1032 		sig = bsd_to_linux_signal(siginfo.si_signo);
1033 		siginfo_to_lsiginfo(&siginfo, &lsi, sig);
1034 		error = copyout(&lsi, infop, sizeof(lsi));
1035 	}
1036 
1037 	return (error);
1038 }
1039 
1040 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1041 int
1042 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
1043 {
1044 	struct linux_wait4_args wait4_args;
1045 
1046 	wait4_args.pid = args->pid;
1047 	wait4_args.status = args->status;
1048 	wait4_args.options = args->options;
1049 	wait4_args.rusage = NULL;
1050 
1051 	return (linux_wait4(td, &wait4_args));
1052 }
1053 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1054 
1055 int
1056 linux_wait4(struct thread *td, struct linux_wait4_args *args)
1057 {
1058 	struct proc *p;
1059 	int options, id, idtype;
1060 
1061 	if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG |
1062 	    LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
1063 		return (EINVAL);
1064 
1065 	/* -INT_MIN is not defined. */
1066 	if (args->pid == INT_MIN)
1067 		return (ESRCH);
1068 
1069 	options = 0;
1070 	linux_to_bsd_waitopts(args->options, &options);
1071 
1072 	/*
1073 	 * For backward compatibility we implicitly add flags WEXITED
1074 	 * and WTRAPPED here.
1075 	 */
1076 	options |= WEXITED | WTRAPPED;
1077 
1078 	/*
1079 	 * As FreeBSD does not have __WALL option bit analogue explicitly set all
1080 	 * possible option bits to emulate Linux __WALL wait option bit. The same
1081 	 * for waitid system call.
1082 	 */
1083 	if ((args->options & __WALL) != 0)
1084 		options |= WUNTRACED | WCONTINUED | WLINUXCLONE;
1085 
1086 	if (args->pid == WAIT_ANY) {
1087 		idtype = P_ALL;
1088 		id = 0;
1089 	} else if (args->pid < 0) {
1090 		idtype = P_PGID;
1091 		id = (id_t)-args->pid;
1092 	} else if (args->pid == 0) {
1093 		idtype = P_PGID;
1094 		p = td->td_proc;
1095 		PROC_LOCK(p);
1096 		id = p->p_pgid;
1097 		PROC_UNLOCK(p);
1098 	} else {
1099 		idtype = P_PID;
1100 		id = (id_t)args->pid;
1101 	}
1102 
1103 	return (linux_common_wait(td, idtype, id, args->status, options,
1104 	    args->rusage, NULL));
1105 }
1106 
1107 int
1108 linux_waitid(struct thread *td, struct linux_waitid_args *args)
1109 {
1110 	idtype_t idtype;
1111 	int error, options;
1112 	struct proc *p;
1113 	pid_t id;
1114 
1115 	if (args->options & ~(LINUX_WNOHANG | LINUX_WNOWAIT | LINUX_WEXITED |
1116 	    LINUX_WSTOPPED | LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
1117 		return (EINVAL);
1118 
1119 	options = 0;
1120 	linux_to_bsd_waitopts(args->options, &options);
1121 	if ((args->options & __WALL) != 0)
1122 		options |= WEXITED | WTRAPPED | WUNTRACED |
1123 		    WCONTINUED | WLINUXCLONE;
1124 
1125 	id = args->id;
1126 	switch (args->idtype) {
1127 	case LINUX_P_ALL:
1128 		idtype = P_ALL;
1129 		break;
1130 	case LINUX_P_PID:
1131 		if (args->id <= 0)
1132 			return (EINVAL);
1133 		idtype = P_PID;
1134 		break;
1135 	case LINUX_P_PGID:
1136 		if (linux_use54(td) && args->id == 0) {
1137 			p = td->td_proc;
1138 			PROC_LOCK(p);
1139 			id = p->p_pgid;
1140 			PROC_UNLOCK(p);
1141 		} else if (args->id <= 0)
1142 			return (EINVAL);
1143 		idtype = P_PGID;
1144 		break;
1145 	case LINUX_P_PIDFD:
1146 		LINUX_RATELIMIT_MSG("unsupported waitid P_PIDFD idtype");
1147 		return (ENOSYS);
1148 	default:
1149 		return (EINVAL);
1150 	}
1151 
1152 	error = linux_common_wait(td, idtype, id, NULL, options,
1153 	    args->rusage, args->info);
1154 	td->td_retval[0] = 0;
1155 
1156 	return (error);
1157 }
1158 
1159 #ifdef LINUX_LEGACY_SYSCALLS
1160 int
1161 linux_mknod(struct thread *td, struct linux_mknod_args *args)
1162 {
1163 	char *path;
1164 	int error;
1165 	enum uio_seg seg;
1166 	bool convpath;
1167 
1168 	convpath = LUSECONVPATH(td);
1169 	if (!convpath) {
1170 		path = args->path;
1171 		seg = UIO_USERSPACE;
1172 	} else {
1173 		LCONVPATHCREAT(args->path, &path);
1174 		seg = UIO_SYSSPACE;
1175 	}
1176 
1177 	switch (args->mode & S_IFMT) {
1178 	case S_IFIFO:
1179 	case S_IFSOCK:
1180 		error = kern_mkfifoat(td, AT_FDCWD, path, seg,
1181 		    args->mode);
1182 		break;
1183 
1184 	case S_IFCHR:
1185 	case S_IFBLK:
1186 		error = kern_mknodat(td, AT_FDCWD, path, seg,
1187 		    args->mode, args->dev);
1188 		break;
1189 
1190 	case S_IFDIR:
1191 		error = EPERM;
1192 		break;
1193 
1194 	case 0:
1195 		args->mode |= S_IFREG;
1196 		/* FALLTHROUGH */
1197 	case S_IFREG:
1198 		error = kern_openat(td, AT_FDCWD, path, seg,
1199 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1200 		if (error == 0)
1201 			kern_close(td, td->td_retval[0]);
1202 		break;
1203 
1204 	default:
1205 		error = EINVAL;
1206 		break;
1207 	}
1208 	if (convpath)
1209 		LFREEPATH(path);
1210 	return (error);
1211 }
1212 #endif
1213 
1214 int
1215 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
1216 {
1217 	char *path;
1218 	int error, dfd;
1219 	enum uio_seg seg;
1220 	bool convpath;
1221 
1222 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
1223 
1224 	convpath = LUSECONVPATH(td);
1225 	if (!convpath) {
1226 		path = __DECONST(char *, args->filename);
1227 		seg = UIO_USERSPACE;
1228 	} else {
1229 		LCONVPATHCREAT_AT(args->filename, &path, dfd);
1230 		seg = UIO_SYSSPACE;
1231 	}
1232 
1233 	switch (args->mode & S_IFMT) {
1234 	case S_IFIFO:
1235 	case S_IFSOCK:
1236 		error = kern_mkfifoat(td, dfd, path, seg, args->mode);
1237 		break;
1238 
1239 	case S_IFCHR:
1240 	case S_IFBLK:
1241 		error = kern_mknodat(td, dfd, path, seg, args->mode,
1242 		    args->dev);
1243 		break;
1244 
1245 	case S_IFDIR:
1246 		error = EPERM;
1247 		break;
1248 
1249 	case 0:
1250 		args->mode |= S_IFREG;
1251 		/* FALLTHROUGH */
1252 	case S_IFREG:
1253 		error = kern_openat(td, dfd, path, seg,
1254 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1255 		if (error == 0)
1256 			kern_close(td, td->td_retval[0]);
1257 		break;
1258 
1259 	default:
1260 		error = EINVAL;
1261 		break;
1262 	}
1263 	if (convpath)
1264 		LFREEPATH(path);
1265 	return (error);
1266 }
1267 
1268 /*
1269  * UGH! This is just about the dumbest idea I've ever heard!!
1270  */
1271 int
1272 linux_personality(struct thread *td, struct linux_personality_args *args)
1273 {
1274 	struct linux_pemuldata *pem;
1275 	struct proc *p = td->td_proc;
1276 	uint32_t old;
1277 
1278 	PROC_LOCK(p);
1279 	pem = pem_find(p);
1280 	old = pem->persona;
1281 	if (args->per != 0xffffffff)
1282 		pem->persona = args->per;
1283 	PROC_UNLOCK(p);
1284 
1285 	td->td_retval[0] = old;
1286 	return (0);
1287 }
1288 
1289 struct l_itimerval {
1290 	l_timeval it_interval;
1291 	l_timeval it_value;
1292 };
1293 
1294 #define	B2L_ITIMERVAL(bip, lip)						\
1295 	(bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec;		\
1296 	(bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec;	\
1297 	(bip)->it_value.tv_sec = (lip)->it_value.tv_sec;		\
1298 	(bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1299 
1300 int
1301 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1302 {
1303 	int error;
1304 	struct l_itimerval ls;
1305 	struct itimerval aitv, oitv;
1306 
1307 	if (uap->itv == NULL) {
1308 		uap->itv = uap->oitv;
1309 		return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1310 	}
1311 
1312 	error = copyin(uap->itv, &ls, sizeof(ls));
1313 	if (error != 0)
1314 		return (error);
1315 	B2L_ITIMERVAL(&aitv, &ls);
1316 	error = kern_setitimer(td, uap->which, &aitv, &oitv);
1317 	if (error != 0 || uap->oitv == NULL)
1318 		return (error);
1319 	B2L_ITIMERVAL(&ls, &oitv);
1320 
1321 	return (copyout(&ls, uap->oitv, sizeof(ls)));
1322 }
1323 
1324 int
1325 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1326 {
1327 	int error;
1328 	struct l_itimerval ls;
1329 	struct itimerval aitv;
1330 
1331 	error = kern_getitimer(td, uap->which, &aitv);
1332 	if (error != 0)
1333 		return (error);
1334 	B2L_ITIMERVAL(&ls, &aitv);
1335 	return (copyout(&ls, uap->itv, sizeof(ls)));
1336 }
1337 
1338 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1339 int
1340 linux_nice(struct thread *td, struct linux_nice_args *args)
1341 {
1342 
1343 	return (kern_setpriority(td, PRIO_PROCESS, 0, args->inc));
1344 }
1345 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1346 
1347 int
1348 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1349 {
1350 	struct ucred *newcred, *oldcred;
1351 	l_gid_t *linux_gidset;
1352 	gid_t *bsd_gidset;
1353 	int ngrp, error;
1354 	struct proc *p;
1355 
1356 	ngrp = args->gidsetsize;
1357 	if (ngrp < 0 || ngrp >= ngroups_max + 1)
1358 		return (EINVAL);
1359 	linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK);
1360 	error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1361 	if (error)
1362 		goto out;
1363 	newcred = crget();
1364 	crextend(newcred, ngrp + 1);
1365 	p = td->td_proc;
1366 	PROC_LOCK(p);
1367 	oldcred = p->p_ucred;
1368 	crcopy(newcred, oldcred);
1369 
1370 	/*
1371 	 * cr_groups[0] holds egid. Setting the whole set from
1372 	 * the supplied set will cause egid to be changed too.
1373 	 * Keep cr_groups[0] unchanged to prevent that.
1374 	 */
1375 
1376 	if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS)) != 0) {
1377 		PROC_UNLOCK(p);
1378 		crfree(newcred);
1379 		goto out;
1380 	}
1381 
1382 	if (ngrp > 0) {
1383 		newcred->cr_ngroups = ngrp + 1;
1384 
1385 		bsd_gidset = newcred->cr_groups;
1386 		ngrp--;
1387 		while (ngrp >= 0) {
1388 			bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1389 			ngrp--;
1390 		}
1391 	} else
1392 		newcred->cr_ngroups = 1;
1393 
1394 	setsugid(p);
1395 	proc_set_cred(p, newcred);
1396 	PROC_UNLOCK(p);
1397 	crfree(oldcred);
1398 	error = 0;
1399 out:
1400 	free(linux_gidset, M_LINUX);
1401 	return (error);
1402 }
1403 
1404 int
1405 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1406 {
1407 	struct ucred *cred;
1408 	l_gid_t *linux_gidset;
1409 	gid_t *bsd_gidset;
1410 	int bsd_gidsetsz, ngrp, error;
1411 
1412 	cred = td->td_ucred;
1413 	bsd_gidset = cred->cr_groups;
1414 	bsd_gidsetsz = cred->cr_ngroups - 1;
1415 
1416 	/*
1417 	 * cr_groups[0] holds egid. Returning the whole set
1418 	 * here will cause a duplicate. Exclude cr_groups[0]
1419 	 * to prevent that.
1420 	 */
1421 
1422 	if ((ngrp = args->gidsetsize) == 0) {
1423 		td->td_retval[0] = bsd_gidsetsz;
1424 		return (0);
1425 	}
1426 
1427 	if (ngrp < bsd_gidsetsz)
1428 		return (EINVAL);
1429 
1430 	ngrp = 0;
1431 	linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1432 	    M_LINUX, M_WAITOK);
1433 	while (ngrp < bsd_gidsetsz) {
1434 		linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1435 		ngrp++;
1436 	}
1437 
1438 	error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1439 	free(linux_gidset, M_LINUX);
1440 	if (error)
1441 		return (error);
1442 
1443 	td->td_retval[0] = ngrp;
1444 	return (0);
1445 }
1446 
1447 static bool
1448 linux_get_dummy_limit(l_uint resource, struct rlimit *rlim)
1449 {
1450 
1451 	if (linux_dummy_rlimits == 0)
1452 		return (false);
1453 
1454 	switch (resource) {
1455 	case LINUX_RLIMIT_LOCKS:
1456 	case LINUX_RLIMIT_SIGPENDING:
1457 	case LINUX_RLIMIT_MSGQUEUE:
1458 	case LINUX_RLIMIT_RTTIME:
1459 		rlim->rlim_cur = LINUX_RLIM_INFINITY;
1460 		rlim->rlim_max = LINUX_RLIM_INFINITY;
1461 		return (true);
1462 	case LINUX_RLIMIT_NICE:
1463 	case LINUX_RLIMIT_RTPRIO:
1464 		rlim->rlim_cur = 0;
1465 		rlim->rlim_max = 0;
1466 		return (true);
1467 	default:
1468 		return (false);
1469 	}
1470 }
1471 
1472 int
1473 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1474 {
1475 	struct rlimit bsd_rlim;
1476 	struct l_rlimit rlim;
1477 	u_int which;
1478 	int error;
1479 
1480 	if (args->resource >= LINUX_RLIM_NLIMITS)
1481 		return (EINVAL);
1482 
1483 	which = linux_to_bsd_resource[args->resource];
1484 	if (which == -1)
1485 		return (EINVAL);
1486 
1487 	error = copyin(args->rlim, &rlim, sizeof(rlim));
1488 	if (error)
1489 		return (error);
1490 
1491 	bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1492 	bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1493 	return (kern_setrlimit(td, which, &bsd_rlim));
1494 }
1495 
1496 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1497 int
1498 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1499 {
1500 	struct l_rlimit rlim;
1501 	struct rlimit bsd_rlim;
1502 	u_int which;
1503 
1504 	if (linux_get_dummy_limit(args->resource, &bsd_rlim)) {
1505 		rlim.rlim_cur = bsd_rlim.rlim_cur;
1506 		rlim.rlim_max = bsd_rlim.rlim_max;
1507 		return (copyout(&rlim, args->rlim, sizeof(rlim)));
1508 	}
1509 
1510 	if (args->resource >= LINUX_RLIM_NLIMITS)
1511 		return (EINVAL);
1512 
1513 	which = linux_to_bsd_resource[args->resource];
1514 	if (which == -1)
1515 		return (EINVAL);
1516 
1517 	lim_rlimit(td, which, &bsd_rlim);
1518 
1519 #ifdef COMPAT_LINUX32
1520 	rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1521 	if (rlim.rlim_cur == UINT_MAX)
1522 		rlim.rlim_cur = INT_MAX;
1523 	rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1524 	if (rlim.rlim_max == UINT_MAX)
1525 		rlim.rlim_max = INT_MAX;
1526 #else
1527 	rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1528 	if (rlim.rlim_cur == ULONG_MAX)
1529 		rlim.rlim_cur = LONG_MAX;
1530 	rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1531 	if (rlim.rlim_max == ULONG_MAX)
1532 		rlim.rlim_max = LONG_MAX;
1533 #endif
1534 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1535 }
1536 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1537 
1538 int
1539 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1540 {
1541 	struct l_rlimit rlim;
1542 	struct rlimit bsd_rlim;
1543 	u_int which;
1544 
1545 	if (linux_get_dummy_limit(args->resource, &bsd_rlim)) {
1546 		rlim.rlim_cur = bsd_rlim.rlim_cur;
1547 		rlim.rlim_max = bsd_rlim.rlim_max;
1548 		return (copyout(&rlim, args->rlim, sizeof(rlim)));
1549 	}
1550 
1551 	if (args->resource >= LINUX_RLIM_NLIMITS)
1552 		return (EINVAL);
1553 
1554 	which = linux_to_bsd_resource[args->resource];
1555 	if (which == -1)
1556 		return (EINVAL);
1557 
1558 	lim_rlimit(td, which, &bsd_rlim);
1559 
1560 	rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1561 	rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1562 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1563 }
1564 
1565 int
1566 linux_sched_setscheduler(struct thread *td,
1567     struct linux_sched_setscheduler_args *args)
1568 {
1569 	struct sched_param sched_param;
1570 	struct thread *tdt;
1571 	int error, policy;
1572 
1573 	switch (args->policy) {
1574 	case LINUX_SCHED_OTHER:
1575 		policy = SCHED_OTHER;
1576 		break;
1577 	case LINUX_SCHED_FIFO:
1578 		policy = SCHED_FIFO;
1579 		break;
1580 	case LINUX_SCHED_RR:
1581 		policy = SCHED_RR;
1582 		break;
1583 	default:
1584 		return (EINVAL);
1585 	}
1586 
1587 	error = copyin(args->param, &sched_param, sizeof(sched_param));
1588 	if (error)
1589 		return (error);
1590 
1591 	if (linux_map_sched_prio) {
1592 		switch (policy) {
1593 		case SCHED_OTHER:
1594 			if (sched_param.sched_priority != 0)
1595 				return (EINVAL);
1596 
1597 			sched_param.sched_priority =
1598 			    PRI_MAX_TIMESHARE - PRI_MIN_TIMESHARE;
1599 			break;
1600 		case SCHED_FIFO:
1601 		case SCHED_RR:
1602 			if (sched_param.sched_priority < 1 ||
1603 			    sched_param.sched_priority >= LINUX_MAX_RT_PRIO)
1604 				return (EINVAL);
1605 
1606 			/*
1607 			 * Map [1, LINUX_MAX_RT_PRIO - 1] to
1608 			 * [0, RTP_PRIO_MAX - RTP_PRIO_MIN] (rounding down).
1609 			 */
1610 			sched_param.sched_priority =
1611 			    (sched_param.sched_priority - 1) *
1612 			    (RTP_PRIO_MAX - RTP_PRIO_MIN + 1) /
1613 			    (LINUX_MAX_RT_PRIO - 1);
1614 			break;
1615 		}
1616 	}
1617 
1618 	tdt = linux_tdfind(td, args->pid, -1);
1619 	if (tdt == NULL)
1620 		return (ESRCH);
1621 
1622 	error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1623 	PROC_UNLOCK(tdt->td_proc);
1624 	return (error);
1625 }
1626 
1627 int
1628 linux_sched_getscheduler(struct thread *td,
1629     struct linux_sched_getscheduler_args *args)
1630 {
1631 	struct thread *tdt;
1632 	int error, policy;
1633 
1634 	tdt = linux_tdfind(td, args->pid, -1);
1635 	if (tdt == NULL)
1636 		return (ESRCH);
1637 
1638 	error = kern_sched_getscheduler(td, tdt, &policy);
1639 	PROC_UNLOCK(tdt->td_proc);
1640 
1641 	switch (policy) {
1642 	case SCHED_OTHER:
1643 		td->td_retval[0] = LINUX_SCHED_OTHER;
1644 		break;
1645 	case SCHED_FIFO:
1646 		td->td_retval[0] = LINUX_SCHED_FIFO;
1647 		break;
1648 	case SCHED_RR:
1649 		td->td_retval[0] = LINUX_SCHED_RR;
1650 		break;
1651 	}
1652 	return (error);
1653 }
1654 
1655 int
1656 linux_sched_get_priority_max(struct thread *td,
1657     struct linux_sched_get_priority_max_args *args)
1658 {
1659 	struct sched_get_priority_max_args bsd;
1660 
1661 	if (linux_map_sched_prio) {
1662 		switch (args->policy) {
1663 		case LINUX_SCHED_OTHER:
1664 			td->td_retval[0] = 0;
1665 			return (0);
1666 		case LINUX_SCHED_FIFO:
1667 		case LINUX_SCHED_RR:
1668 			td->td_retval[0] = LINUX_MAX_RT_PRIO - 1;
1669 			return (0);
1670 		default:
1671 			return (EINVAL);
1672 		}
1673 	}
1674 
1675 	switch (args->policy) {
1676 	case LINUX_SCHED_OTHER:
1677 		bsd.policy = SCHED_OTHER;
1678 		break;
1679 	case LINUX_SCHED_FIFO:
1680 		bsd.policy = SCHED_FIFO;
1681 		break;
1682 	case LINUX_SCHED_RR:
1683 		bsd.policy = SCHED_RR;
1684 		break;
1685 	default:
1686 		return (EINVAL);
1687 	}
1688 	return (sys_sched_get_priority_max(td, &bsd));
1689 }
1690 
1691 int
1692 linux_sched_get_priority_min(struct thread *td,
1693     struct linux_sched_get_priority_min_args *args)
1694 {
1695 	struct sched_get_priority_min_args bsd;
1696 
1697 	if (linux_map_sched_prio) {
1698 		switch (args->policy) {
1699 		case LINUX_SCHED_OTHER:
1700 			td->td_retval[0] = 0;
1701 			return (0);
1702 		case LINUX_SCHED_FIFO:
1703 		case LINUX_SCHED_RR:
1704 			td->td_retval[0] = 1;
1705 			return (0);
1706 		default:
1707 			return (EINVAL);
1708 		}
1709 	}
1710 
1711 	switch (args->policy) {
1712 	case LINUX_SCHED_OTHER:
1713 		bsd.policy = SCHED_OTHER;
1714 		break;
1715 	case LINUX_SCHED_FIFO:
1716 		bsd.policy = SCHED_FIFO;
1717 		break;
1718 	case LINUX_SCHED_RR:
1719 		bsd.policy = SCHED_RR;
1720 		break;
1721 	default:
1722 		return (EINVAL);
1723 	}
1724 	return (sys_sched_get_priority_min(td, &bsd));
1725 }
1726 
1727 #define REBOOT_CAD_ON	0x89abcdef
1728 #define REBOOT_CAD_OFF	0
1729 #define REBOOT_HALT	0xcdef0123
1730 #define REBOOT_RESTART	0x01234567
1731 #define REBOOT_RESTART2	0xA1B2C3D4
1732 #define REBOOT_POWEROFF	0x4321FEDC
1733 #define REBOOT_MAGIC1	0xfee1dead
1734 #define REBOOT_MAGIC2	0x28121969
1735 #define REBOOT_MAGIC2A	0x05121996
1736 #define REBOOT_MAGIC2B	0x16041998
1737 
1738 int
1739 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1740 {
1741 	struct reboot_args bsd_args;
1742 
1743 	if (args->magic1 != REBOOT_MAGIC1)
1744 		return (EINVAL);
1745 
1746 	switch (args->magic2) {
1747 	case REBOOT_MAGIC2:
1748 	case REBOOT_MAGIC2A:
1749 	case REBOOT_MAGIC2B:
1750 		break;
1751 	default:
1752 		return (EINVAL);
1753 	}
1754 
1755 	switch (args->cmd) {
1756 	case REBOOT_CAD_ON:
1757 	case REBOOT_CAD_OFF:
1758 		return (priv_check(td, PRIV_REBOOT));
1759 	case REBOOT_HALT:
1760 		bsd_args.opt = RB_HALT;
1761 		break;
1762 	case REBOOT_RESTART:
1763 	case REBOOT_RESTART2:
1764 		bsd_args.opt = 0;
1765 		break;
1766 	case REBOOT_POWEROFF:
1767 		bsd_args.opt = RB_POWEROFF;
1768 		break;
1769 	default:
1770 		return (EINVAL);
1771 	}
1772 	return (sys_reboot(td, &bsd_args));
1773 }
1774 
1775 int
1776 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1777 {
1778 
1779 	td->td_retval[0] = td->td_proc->p_pid;
1780 
1781 	return (0);
1782 }
1783 
1784 int
1785 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1786 {
1787 	struct linux_emuldata *em;
1788 
1789 	em = em_find(td);
1790 	KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1791 
1792 	td->td_retval[0] = em->em_tid;
1793 
1794 	return (0);
1795 }
1796 
1797 int
1798 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1799 {
1800 
1801 	td->td_retval[0] = kern_getppid(td);
1802 	return (0);
1803 }
1804 
1805 int
1806 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1807 {
1808 
1809 	td->td_retval[0] = td->td_ucred->cr_rgid;
1810 	return (0);
1811 }
1812 
1813 int
1814 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1815 {
1816 
1817 	td->td_retval[0] = td->td_ucred->cr_ruid;
1818 	return (0);
1819 }
1820 
1821 int
1822 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1823 {
1824 
1825 	return (kern_getsid(td, args->pid));
1826 }
1827 
1828 int
1829 linux_nosys(struct thread *td, struct nosys_args *ignore)
1830 {
1831 
1832 	return (ENOSYS);
1833 }
1834 
1835 int
1836 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1837 {
1838 	int error;
1839 
1840 	error = kern_getpriority(td, args->which, args->who);
1841 	td->td_retval[0] = 20 - td->td_retval[0];
1842 	return (error);
1843 }
1844 
1845 int
1846 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1847 {
1848 	int name[2];
1849 
1850 	name[0] = CTL_KERN;
1851 	name[1] = KERN_HOSTNAME;
1852 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1853 	    args->len, 0, 0));
1854 }
1855 
1856 int
1857 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1858 {
1859 	int name[2];
1860 
1861 	name[0] = CTL_KERN;
1862 	name[1] = KERN_NISDOMAINNAME;
1863 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1864 	    args->len, 0, 0));
1865 }
1866 
1867 int
1868 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1869 {
1870 
1871 	LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1872 	    args->error_code);
1873 
1874 	/*
1875 	 * XXX: we should send a signal to the parent if
1876 	 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1877 	 * as it doesnt occur often.
1878 	 */
1879 	exit1(td, args->error_code, 0);
1880 		/* NOTREACHED */
1881 }
1882 
1883 #define _LINUX_CAPABILITY_VERSION_1  0x19980330
1884 #define _LINUX_CAPABILITY_VERSION_2  0x20071026
1885 #define _LINUX_CAPABILITY_VERSION_3  0x20080522
1886 
1887 struct l_user_cap_header {
1888 	l_int	version;
1889 	l_int	pid;
1890 };
1891 
1892 struct l_user_cap_data {
1893 	l_int	effective;
1894 	l_int	permitted;
1895 	l_int	inheritable;
1896 };
1897 
1898 int
1899 linux_capget(struct thread *td, struct linux_capget_args *uap)
1900 {
1901 	struct l_user_cap_header luch;
1902 	struct l_user_cap_data lucd[2];
1903 	int error, u32s;
1904 
1905 	if (uap->hdrp == NULL)
1906 		return (EFAULT);
1907 
1908 	error = copyin(uap->hdrp, &luch, sizeof(luch));
1909 	if (error != 0)
1910 		return (error);
1911 
1912 	switch (luch.version) {
1913 	case _LINUX_CAPABILITY_VERSION_1:
1914 		u32s = 1;
1915 		break;
1916 	case _LINUX_CAPABILITY_VERSION_2:
1917 	case _LINUX_CAPABILITY_VERSION_3:
1918 		u32s = 2;
1919 		break;
1920 	default:
1921 		luch.version = _LINUX_CAPABILITY_VERSION_1;
1922 		error = copyout(&luch, uap->hdrp, sizeof(luch));
1923 		if (error)
1924 			return (error);
1925 		return (EINVAL);
1926 	}
1927 
1928 	if (luch.pid)
1929 		return (EPERM);
1930 
1931 	if (uap->datap) {
1932 		/*
1933 		 * The current implementation doesn't support setting
1934 		 * a capability (it's essentially a stub) so indicate
1935 		 * that no capabilities are currently set or available
1936 		 * to request.
1937 		 */
1938 		memset(&lucd, 0, u32s * sizeof(lucd[0]));
1939 		error = copyout(&lucd, uap->datap, u32s * sizeof(lucd[0]));
1940 	}
1941 
1942 	return (error);
1943 }
1944 
1945 int
1946 linux_capset(struct thread *td, struct linux_capset_args *uap)
1947 {
1948 	struct l_user_cap_header luch;
1949 	struct l_user_cap_data lucd[2];
1950 	int error, i, u32s;
1951 
1952 	if (uap->hdrp == NULL || uap->datap == NULL)
1953 		return (EFAULT);
1954 
1955 	error = copyin(uap->hdrp, &luch, sizeof(luch));
1956 	if (error != 0)
1957 		return (error);
1958 
1959 	switch (luch.version) {
1960 	case _LINUX_CAPABILITY_VERSION_1:
1961 		u32s = 1;
1962 		break;
1963 	case _LINUX_CAPABILITY_VERSION_2:
1964 	case _LINUX_CAPABILITY_VERSION_3:
1965 		u32s = 2;
1966 		break;
1967 	default:
1968 		luch.version = _LINUX_CAPABILITY_VERSION_1;
1969 		error = copyout(&luch, uap->hdrp, sizeof(luch));
1970 		if (error)
1971 			return (error);
1972 		return (EINVAL);
1973 	}
1974 
1975 	if (luch.pid)
1976 		return (EPERM);
1977 
1978 	error = copyin(uap->datap, &lucd, u32s * sizeof(lucd[0]));
1979 	if (error != 0)
1980 		return (error);
1981 
1982 	/* We currently don't support setting any capabilities. */
1983 	for (i = 0; i < u32s; i++) {
1984 		if (lucd[i].effective || lucd[i].permitted ||
1985 		    lucd[i].inheritable) {
1986 			linux_msg(td,
1987 			    "capset[%d] effective=0x%x, permitted=0x%x, "
1988 			    "inheritable=0x%x is not implemented", i,
1989 			    (int)lucd[i].effective, (int)lucd[i].permitted,
1990 			    (int)lucd[i].inheritable);
1991 			return (EPERM);
1992 		}
1993 	}
1994 
1995 	return (0);
1996 }
1997 
1998 int
1999 linux_prctl(struct thread *td, struct linux_prctl_args *args)
2000 {
2001 	int error = 0, max_size, arg;
2002 	struct proc *p = td->td_proc;
2003 	char comm[LINUX_MAX_COMM_LEN];
2004 	int pdeath_signal, trace_state;
2005 
2006 	switch (args->option) {
2007 	case LINUX_PR_SET_PDEATHSIG:
2008 		if (!LINUX_SIG_VALID(args->arg2))
2009 			return (EINVAL);
2010 		pdeath_signal = linux_to_bsd_signal(args->arg2);
2011 		return (kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_CTL,
2012 		    &pdeath_signal));
2013 	case LINUX_PR_GET_PDEATHSIG:
2014 		error = kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_STATUS,
2015 		    &pdeath_signal);
2016 		if (error != 0)
2017 			return (error);
2018 		pdeath_signal = bsd_to_linux_signal(pdeath_signal);
2019 		return (copyout(&pdeath_signal,
2020 		    (void *)(register_t)args->arg2,
2021 		    sizeof(pdeath_signal)));
2022 	/*
2023 	 * In Linux, this flag controls if set[gu]id processes can coredump.
2024 	 * There are additional semantics imposed on processes that cannot
2025 	 * coredump:
2026 	 * - Such processes can not be ptraced.
2027 	 * - There are some semantics around ownership of process-related files
2028 	 *   in the /proc namespace.
2029 	 *
2030 	 * In FreeBSD, we can (and by default, do) disable setuid coredump
2031 	 * system-wide with 'sugid_coredump.'  We control tracability on a
2032 	 * per-process basis with the procctl PROC_TRACE (=> P2_NOTRACE flag).
2033 	 * By happy coincidence, P2_NOTRACE also prevents coredumping.  So the
2034 	 * procctl is roughly analogous to Linux's DUMPABLE.
2035 	 *
2036 	 * So, proxy these knobs to the corresponding PROC_TRACE setting.
2037 	 */
2038 	case LINUX_PR_GET_DUMPABLE:
2039 		error = kern_procctl(td, P_PID, p->p_pid, PROC_TRACE_STATUS,
2040 		    &trace_state);
2041 		if (error != 0)
2042 			return (error);
2043 		td->td_retval[0] = (trace_state != -1);
2044 		return (0);
2045 	case LINUX_PR_SET_DUMPABLE:
2046 		/*
2047 		 * It is only valid for userspace to set one of these two
2048 		 * flags, and only one at a time.
2049 		 */
2050 		switch (args->arg2) {
2051 		case LINUX_SUID_DUMP_DISABLE:
2052 			trace_state = PROC_TRACE_CTL_DISABLE_EXEC;
2053 			break;
2054 		case LINUX_SUID_DUMP_USER:
2055 			trace_state = PROC_TRACE_CTL_ENABLE;
2056 			break;
2057 		default:
2058 			return (EINVAL);
2059 		}
2060 		return (kern_procctl(td, P_PID, p->p_pid, PROC_TRACE_CTL,
2061 		    &trace_state));
2062 	case LINUX_PR_GET_KEEPCAPS:
2063 		/*
2064 		 * Indicate that we always clear the effective and
2065 		 * permitted capability sets when the user id becomes
2066 		 * non-zero (actually the capability sets are simply
2067 		 * always zero in the current implementation).
2068 		 */
2069 		td->td_retval[0] = 0;
2070 		break;
2071 	case LINUX_PR_SET_KEEPCAPS:
2072 		/*
2073 		 * Ignore requests to keep the effective and permitted
2074 		 * capability sets when the user id becomes non-zero.
2075 		 */
2076 		break;
2077 	case LINUX_PR_SET_NAME:
2078 		/*
2079 		 * To be on the safe side we need to make sure to not
2080 		 * overflow the size a Linux program expects. We already
2081 		 * do this here in the copyin, so that we don't need to
2082 		 * check on copyout.
2083 		 */
2084 		max_size = MIN(sizeof(comm), sizeof(p->p_comm));
2085 		error = copyinstr((void *)(register_t)args->arg2, comm,
2086 		    max_size, NULL);
2087 
2088 		/* Linux silently truncates the name if it is too long. */
2089 		if (error == ENAMETOOLONG) {
2090 			/*
2091 			 * XXX: copyinstr() isn't documented to populate the
2092 			 * array completely, so do a copyin() to be on the
2093 			 * safe side. This should be changed in case
2094 			 * copyinstr() is changed to guarantee this.
2095 			 */
2096 			error = copyin((void *)(register_t)args->arg2, comm,
2097 			    max_size - 1);
2098 			comm[max_size - 1] = '\0';
2099 		}
2100 		if (error)
2101 			return (error);
2102 
2103 		PROC_LOCK(p);
2104 		strlcpy(p->p_comm, comm, sizeof(p->p_comm));
2105 		PROC_UNLOCK(p);
2106 		break;
2107 	case LINUX_PR_GET_NAME:
2108 		PROC_LOCK(p);
2109 		strlcpy(comm, p->p_comm, sizeof(comm));
2110 		PROC_UNLOCK(p);
2111 		error = copyout(comm, (void *)(register_t)args->arg2,
2112 		    strlen(comm) + 1);
2113 		break;
2114 	case LINUX_PR_GET_SECCOMP:
2115 	case LINUX_PR_SET_SECCOMP:
2116 		/*
2117 		 * Same as returned by Linux without CONFIG_SECCOMP enabled.
2118 		 */
2119 		error = EINVAL;
2120 		break;
2121 	case LINUX_PR_CAPBSET_READ:
2122 #if 0
2123 		/*
2124 		 * This makes too much noise with Ubuntu Focal.
2125 		 */
2126 		linux_msg(td, "unsupported prctl PR_CAPBSET_READ %d",
2127 		    (int)args->arg2);
2128 #endif
2129 		error = EINVAL;
2130 		break;
2131 	case LINUX_PR_SET_NO_NEW_PRIVS:
2132 		arg = args->arg2 == 1 ?
2133 		    PROC_NO_NEW_PRIVS_ENABLE : PROC_NO_NEW_PRIVS_DISABLE;
2134 		error = kern_procctl(td, P_PID, p->p_pid,
2135 		    PROC_NO_NEW_PRIVS_CTL, &arg);
2136 		break;
2137 	case LINUX_PR_SET_PTRACER:
2138 		linux_msg(td, "unsupported prctl PR_SET_PTRACER");
2139 		error = EINVAL;
2140 		break;
2141 	default:
2142 		linux_msg(td, "unsupported prctl option %d", args->option);
2143 		error = EINVAL;
2144 		break;
2145 	}
2146 
2147 	return (error);
2148 }
2149 
2150 int
2151 linux_sched_setparam(struct thread *td,
2152     struct linux_sched_setparam_args *uap)
2153 {
2154 	struct sched_param sched_param;
2155 	struct thread *tdt;
2156 	int error, policy;
2157 
2158 	error = copyin(uap->param, &sched_param, sizeof(sched_param));
2159 	if (error)
2160 		return (error);
2161 
2162 	tdt = linux_tdfind(td, uap->pid, -1);
2163 	if (tdt == NULL)
2164 		return (ESRCH);
2165 
2166 	if (linux_map_sched_prio) {
2167 		error = kern_sched_getscheduler(td, tdt, &policy);
2168 		if (error)
2169 			goto out;
2170 
2171 		switch (policy) {
2172 		case SCHED_OTHER:
2173 			if (sched_param.sched_priority != 0) {
2174 				error = EINVAL;
2175 				goto out;
2176 			}
2177 			sched_param.sched_priority =
2178 			    PRI_MAX_TIMESHARE - PRI_MIN_TIMESHARE;
2179 			break;
2180 		case SCHED_FIFO:
2181 		case SCHED_RR:
2182 			if (sched_param.sched_priority < 1 ||
2183 			    sched_param.sched_priority >= LINUX_MAX_RT_PRIO) {
2184 				error = EINVAL;
2185 				goto out;
2186 			}
2187 			/*
2188 			 * Map [1, LINUX_MAX_RT_PRIO - 1] to
2189 			 * [0, RTP_PRIO_MAX - RTP_PRIO_MIN] (rounding down).
2190 			 */
2191 			sched_param.sched_priority =
2192 			    (sched_param.sched_priority - 1) *
2193 			    (RTP_PRIO_MAX - RTP_PRIO_MIN + 1) /
2194 			    (LINUX_MAX_RT_PRIO - 1);
2195 			break;
2196 		}
2197 	}
2198 
2199 	error = kern_sched_setparam(td, tdt, &sched_param);
2200 out:	PROC_UNLOCK(tdt->td_proc);
2201 	return (error);
2202 }
2203 
2204 int
2205 linux_sched_getparam(struct thread *td,
2206     struct linux_sched_getparam_args *uap)
2207 {
2208 	struct sched_param sched_param;
2209 	struct thread *tdt;
2210 	int error, policy;
2211 
2212 	tdt = linux_tdfind(td, uap->pid, -1);
2213 	if (tdt == NULL)
2214 		return (ESRCH);
2215 
2216 	error = kern_sched_getparam(td, tdt, &sched_param);
2217 	if (error) {
2218 		PROC_UNLOCK(tdt->td_proc);
2219 		return (error);
2220 	}
2221 
2222 	if (linux_map_sched_prio) {
2223 		error = kern_sched_getscheduler(td, tdt, &policy);
2224 		PROC_UNLOCK(tdt->td_proc);
2225 		if (error)
2226 			return (error);
2227 
2228 		switch (policy) {
2229 		case SCHED_OTHER:
2230 			sched_param.sched_priority = 0;
2231 			break;
2232 		case SCHED_FIFO:
2233 		case SCHED_RR:
2234 			/*
2235 			 * Map [0, RTP_PRIO_MAX - RTP_PRIO_MIN] to
2236 			 * [1, LINUX_MAX_RT_PRIO - 1] (rounding up).
2237 			 */
2238 			sched_param.sched_priority =
2239 			    (sched_param.sched_priority *
2240 			    (LINUX_MAX_RT_PRIO - 1) +
2241 			    (RTP_PRIO_MAX - RTP_PRIO_MIN - 1)) /
2242 			    (RTP_PRIO_MAX - RTP_PRIO_MIN) + 1;
2243 			break;
2244 		}
2245 	} else
2246 		PROC_UNLOCK(tdt->td_proc);
2247 
2248 	error = copyout(&sched_param, uap->param, sizeof(sched_param));
2249 	return (error);
2250 }
2251 
2252 /*
2253  * Get affinity of a process.
2254  */
2255 int
2256 linux_sched_getaffinity(struct thread *td,
2257     struct linux_sched_getaffinity_args *args)
2258 {
2259 	int error;
2260 	struct thread *tdt;
2261 
2262 	if (args->len < sizeof(cpuset_t))
2263 		return (EINVAL);
2264 
2265 	tdt = linux_tdfind(td, args->pid, -1);
2266 	if (tdt == NULL)
2267 		return (ESRCH);
2268 
2269 	PROC_UNLOCK(tdt->td_proc);
2270 
2271 	error = kern_cpuset_getaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2272 	    tdt->td_tid, sizeof(cpuset_t), (cpuset_t *)args->user_mask_ptr);
2273 	if (error == 0)
2274 		td->td_retval[0] = sizeof(cpuset_t);
2275 
2276 	return (error);
2277 }
2278 
2279 /*
2280  *  Set affinity of a process.
2281  */
2282 int
2283 linux_sched_setaffinity(struct thread *td,
2284     struct linux_sched_setaffinity_args *args)
2285 {
2286 	struct thread *tdt;
2287 
2288 	if (args->len < sizeof(cpuset_t))
2289 		return (EINVAL);
2290 
2291 	tdt = linux_tdfind(td, args->pid, -1);
2292 	if (tdt == NULL)
2293 		return (ESRCH);
2294 
2295 	PROC_UNLOCK(tdt->td_proc);
2296 
2297 	return (kern_cpuset_setaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2298 	    tdt->td_tid, sizeof(cpuset_t), (cpuset_t *) args->user_mask_ptr));
2299 }
2300 
2301 struct linux_rlimit64 {
2302 	uint64_t	rlim_cur;
2303 	uint64_t	rlim_max;
2304 };
2305 
2306 int
2307 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
2308 {
2309 	struct rlimit rlim, nrlim;
2310 	struct linux_rlimit64 lrlim;
2311 	struct proc *p;
2312 	u_int which;
2313 	int flags;
2314 	int error;
2315 
2316 	if (args->new == NULL && args->old != NULL) {
2317 		if (linux_get_dummy_limit(args->resource, &rlim)) {
2318 			lrlim.rlim_cur = rlim.rlim_cur;
2319 			lrlim.rlim_max = rlim.rlim_max;
2320 			return (copyout(&lrlim, args->old, sizeof(lrlim)));
2321 		}
2322 	}
2323 
2324 	if (args->resource >= LINUX_RLIM_NLIMITS)
2325 		return (EINVAL);
2326 
2327 	which = linux_to_bsd_resource[args->resource];
2328 	if (which == -1)
2329 		return (EINVAL);
2330 
2331 	if (args->new != NULL) {
2332 		/*
2333 		 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
2334 		 * rlim is unsigned 64-bit. FreeBSD treats negative limits
2335 		 * as INFINITY so we do not need a conversion even.
2336 		 */
2337 		error = copyin(args->new, &nrlim, sizeof(nrlim));
2338 		if (error != 0)
2339 			return (error);
2340 	}
2341 
2342 	flags = PGET_HOLD | PGET_NOTWEXIT;
2343 	if (args->new != NULL)
2344 		flags |= PGET_CANDEBUG;
2345 	else
2346 		flags |= PGET_CANSEE;
2347 	if (args->pid == 0) {
2348 		p = td->td_proc;
2349 		PHOLD(p);
2350 	} else {
2351 		error = pget(args->pid, flags, &p);
2352 		if (error != 0)
2353 			return (error);
2354 	}
2355 	if (args->old != NULL) {
2356 		PROC_LOCK(p);
2357 		lim_rlimit_proc(p, which, &rlim);
2358 		PROC_UNLOCK(p);
2359 		if (rlim.rlim_cur == RLIM_INFINITY)
2360 			lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2361 		else
2362 			lrlim.rlim_cur = rlim.rlim_cur;
2363 		if (rlim.rlim_max == RLIM_INFINITY)
2364 			lrlim.rlim_max = LINUX_RLIM_INFINITY;
2365 		else
2366 			lrlim.rlim_max = rlim.rlim_max;
2367 		error = copyout(&lrlim, args->old, sizeof(lrlim));
2368 		if (error != 0)
2369 			goto out;
2370 	}
2371 
2372 	if (args->new != NULL)
2373 		error = kern_proc_setrlimit(td, p, which, &nrlim);
2374 
2375  out:
2376 	PRELE(p);
2377 	return (error);
2378 }
2379 
2380 int
2381 linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2382 {
2383 	struct timespec ts, *tsp;
2384 	int error;
2385 
2386 	if (args->tsp != NULL) {
2387 		error = linux_get_timespec(&ts, args->tsp);
2388 		if (error != 0)
2389 			return (error);
2390 		tsp = &ts;
2391 	} else
2392 		tsp = NULL;
2393 
2394 	error = linux_common_pselect6(td, args->nfds, args->readfds,
2395 	    args->writefds, args->exceptfds, tsp, args->sig);
2396 
2397 	if (args->tsp != NULL)
2398 		linux_put_timespec(&ts, args->tsp);
2399 	return (error);
2400 }
2401 
2402 static int
2403 linux_common_pselect6(struct thread *td, l_int nfds, l_fd_set *readfds,
2404     l_fd_set *writefds, l_fd_set *exceptfds, struct timespec *tsp,
2405     l_uintptr_t *sig)
2406 {
2407 	struct timeval utv, tv0, tv1, *tvp;
2408 	struct l_pselect6arg lpse6;
2409 	l_sigset_t l_ss;
2410 	sigset_t *ssp;
2411 	sigset_t ss;
2412 	int error;
2413 
2414 	ssp = NULL;
2415 	if (sig != NULL) {
2416 		error = copyin(sig, &lpse6, sizeof(lpse6));
2417 		if (error != 0)
2418 			return (error);
2419 		if (lpse6.ss_len != sizeof(l_ss))
2420 			return (EINVAL);
2421 		if (lpse6.ss != 0) {
2422 			error = copyin(PTRIN(lpse6.ss), &l_ss,
2423 			    sizeof(l_ss));
2424 			if (error != 0)
2425 				return (error);
2426 			linux_to_bsd_sigset(&l_ss, &ss);
2427 			ssp = &ss;
2428 		}
2429 	} else
2430 		ssp = NULL;
2431 
2432 	/*
2433 	 * Currently glibc changes nanosecond number to microsecond.
2434 	 * This mean losing precision but for now it is hardly seen.
2435 	 */
2436 	if (tsp != NULL) {
2437 		TIMESPEC_TO_TIMEVAL(&utv, tsp);
2438 		if (itimerfix(&utv))
2439 			return (EINVAL);
2440 
2441 		microtime(&tv0);
2442 		tvp = &utv;
2443 	} else
2444 		tvp = NULL;
2445 
2446 	error = kern_pselect(td, nfds, readfds, writefds,
2447 	    exceptfds, tvp, ssp, LINUX_NFDBITS);
2448 
2449 	if (tsp != NULL) {
2450 		/*
2451 		 * Compute how much time was left of the timeout,
2452 		 * by subtracting the current time and the time
2453 		 * before we started the call, and subtracting
2454 		 * that result from the user-supplied value.
2455 		 */
2456 		microtime(&tv1);
2457 		timevalsub(&tv1, &tv0);
2458 		timevalsub(&utv, &tv1);
2459 		if (utv.tv_sec < 0)
2460 			timevalclear(&utv);
2461 		TIMEVAL_TO_TIMESPEC(&utv, tsp);
2462 	}
2463 	return (error);
2464 }
2465 
2466 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
2467 int
2468 linux_pselect6_time64(struct thread *td,
2469     struct linux_pselect6_time64_args *args)
2470 {
2471 	struct timespec ts, *tsp;
2472 	int error;
2473 
2474 	if (args->tsp != NULL) {
2475 		error = linux_get_timespec64(&ts, args->tsp);
2476 		if (error != 0)
2477 			return (error);
2478 		tsp = &ts;
2479 	} else
2480 		tsp = NULL;
2481 
2482 	error = linux_common_pselect6(td, args->nfds, args->readfds,
2483 	    args->writefds, args->exceptfds, tsp, args->sig);
2484 
2485 	if (args->tsp != NULL)
2486 		linux_put_timespec64(&ts, args->tsp);
2487 	return (error);
2488 }
2489 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
2490 
2491 int
2492 linux_ppoll(struct thread *td, struct linux_ppoll_args *args)
2493 {
2494 	struct timespec uts, *tsp;
2495 	int error;
2496 
2497 	if (args->tsp != NULL) {
2498 		error = linux_get_timespec(&uts, args->tsp);
2499 		if (error != 0)
2500 			return (error);
2501 		tsp = &uts;
2502 	} else
2503 		tsp = NULL;
2504 
2505 	error = linux_common_ppoll(td, args->fds, args->nfds, tsp,
2506 	    args->sset, args->ssize);
2507 	if (error == 0 && args->tsp != NULL)
2508 		error = linux_put_timespec(&uts, args->tsp);
2509 	return (error);
2510 }
2511 
2512 static int
2513 linux_common_ppoll(struct thread *td, struct pollfd *fds, uint32_t nfds,
2514     struct timespec *tsp, l_sigset_t *sset, l_size_t ssize)
2515 {
2516 	struct timespec ts0, ts1;
2517 	struct pollfd stackfds[32];
2518 	struct pollfd *kfds;
2519  	l_sigset_t l_ss;
2520  	sigset_t *ssp;
2521  	sigset_t ss;
2522  	int error;
2523 
2524 	if (kern_poll_maxfds(nfds))
2525 		return (EINVAL);
2526 	if (sset != NULL) {
2527 		if (ssize != sizeof(l_ss))
2528 			return (EINVAL);
2529 		error = copyin(sset, &l_ss, sizeof(l_ss));
2530 		if (error)
2531 			return (error);
2532 		linux_to_bsd_sigset(&l_ss, &ss);
2533 		ssp = &ss;
2534 	} else
2535 		ssp = NULL;
2536 	if (tsp != NULL)
2537 		nanotime(&ts0);
2538 
2539 	if (nfds > nitems(stackfds))
2540 		kfds = mallocarray(nfds, sizeof(*kfds), M_TEMP, M_WAITOK);
2541 	else
2542 		kfds = stackfds;
2543 	error = linux_pollin(td, kfds, fds, nfds);
2544 	if (error != 0)
2545 		goto out;
2546 
2547 	error = kern_poll_kfds(td, kfds, nfds, tsp, ssp);
2548 	if (error == 0)
2549 		error = linux_pollout(td, kfds, fds, nfds);
2550 
2551 	if (error == 0 && tsp != NULL) {
2552 		if (td->td_retval[0]) {
2553 			nanotime(&ts1);
2554 			timespecsub(&ts1, &ts0, &ts1);
2555 			timespecsub(tsp, &ts1, tsp);
2556 			if (tsp->tv_sec < 0)
2557 				timespecclear(tsp);
2558 		} else
2559 			timespecclear(tsp);
2560 	}
2561 
2562 out:
2563 	if (nfds > nitems(stackfds))
2564 		free(kfds, M_TEMP);
2565 	return (error);
2566 }
2567 
2568 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
2569 int
2570 linux_ppoll_time64(struct thread *td, struct linux_ppoll_time64_args *args)
2571 {
2572 	struct timespec uts, *tsp;
2573 	int error;
2574 
2575 	if (args->tsp != NULL) {
2576 		error = linux_get_timespec64(&uts, args->tsp);
2577 		if (error != 0)
2578 			return (error);
2579 		tsp = &uts;
2580 	} else
2581  		tsp = NULL;
2582 	error = linux_common_ppoll(td, args->fds, args->nfds, tsp,
2583 	    args->sset, args->ssize);
2584 	if (error == 0 && args->tsp != NULL)
2585 		error = linux_put_timespec64(&uts, args->tsp);
2586 	return (error);
2587 }
2588 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
2589 
2590 static int
2591 linux_pollin(struct thread *td, struct pollfd *fds, struct pollfd *ufds, u_int nfd)
2592 {
2593 	int error;
2594 	u_int i;
2595 
2596 	error = copyin(ufds, fds, nfd * sizeof(*fds));
2597 	if (error != 0)
2598 		return (error);
2599 
2600 	for (i = 0; i < nfd; i++) {
2601 		if (fds->events != 0)
2602 			linux_to_bsd_poll_events(td, fds->fd,
2603 			    fds->events, &fds->events);
2604 		fds++;
2605 	}
2606 	return (0);
2607 }
2608 
2609 static int
2610 linux_pollout(struct thread *td, struct pollfd *fds, struct pollfd *ufds, u_int nfd)
2611 {
2612 	int error = 0;
2613 	u_int i, n = 0;
2614 
2615 	for (i = 0; i < nfd; i++) {
2616 		if (fds->revents != 0) {
2617 			bsd_to_linux_poll_events(fds->revents,
2618 			    &fds->revents);
2619 			n++;
2620 		}
2621 		error = copyout(&fds->revents, &ufds->revents,
2622 		    sizeof(ufds->revents));
2623 		if (error)
2624 			return (error);
2625 		fds++;
2626 		ufds++;
2627 	}
2628 	td->td_retval[0] = n;
2629 	return (0);
2630 }
2631 
2632 static int
2633 linux_sched_rr_get_interval_common(struct thread *td, pid_t pid,
2634     struct timespec *ts)
2635 {
2636 	struct thread *tdt;
2637 	int error;
2638 
2639 	/*
2640 	 * According to man in case the invalid pid specified
2641 	 * EINVAL should be returned.
2642 	 */
2643 	if (pid < 0)
2644 		return (EINVAL);
2645 
2646 	tdt = linux_tdfind(td, pid, -1);
2647 	if (tdt == NULL)
2648 		return (ESRCH);
2649 
2650 	error = kern_sched_rr_get_interval_td(td, tdt, ts);
2651 	PROC_UNLOCK(tdt->td_proc);
2652 	return (error);
2653 }
2654 
2655 int
2656 linux_sched_rr_get_interval(struct thread *td,
2657     struct linux_sched_rr_get_interval_args *uap)
2658 {
2659 	struct timespec ts;
2660 	int error;
2661 
2662 	error = linux_sched_rr_get_interval_common(td, uap->pid, &ts);
2663 	if (error != 0)
2664 		return (error);
2665 	return (linux_put_timespec(&ts, uap->interval));
2666 }
2667 
2668 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
2669 int
2670 linux_sched_rr_get_interval_time64(struct thread *td,
2671     struct linux_sched_rr_get_interval_time64_args *uap)
2672 {
2673 	struct timespec ts;
2674 	int error;
2675 
2676 	error = linux_sched_rr_get_interval_common(td, uap->pid, &ts);
2677 	if (error != 0)
2678 		return (error);
2679 	return (linux_put_timespec64(&ts, uap->interval));
2680 }
2681 #endif
2682 
2683 /*
2684  * In case when the Linux thread is the initial thread in
2685  * the thread group thread id is equal to the process id.
2686  * Glibc depends on this magic (assert in pthread_getattr_np.c).
2687  */
2688 struct thread *
2689 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid)
2690 {
2691 	struct linux_emuldata *em;
2692 	struct thread *tdt;
2693 	struct proc *p;
2694 
2695 	tdt = NULL;
2696 	if (tid == 0 || tid == td->td_tid) {
2697 		if (pid != -1 && td->td_proc->p_pid != pid)
2698 			return (NULL);
2699 		PROC_LOCK(td->td_proc);
2700 		return (td);
2701 	} else if (tid > PID_MAX)
2702 		return (tdfind(tid, pid));
2703 
2704 	/*
2705 	 * Initial thread where the tid equal to the pid.
2706 	 */
2707 	p = pfind(tid);
2708 	if (p != NULL) {
2709 		if (SV_PROC_ABI(p) != SV_ABI_LINUX ||
2710 		    (pid != -1 && tid != pid)) {
2711 			/*
2712 			 * p is not a Linuxulator process.
2713 			 */
2714 			PROC_UNLOCK(p);
2715 			return (NULL);
2716 		}
2717 		FOREACH_THREAD_IN_PROC(p, tdt) {
2718 			em = em_find(tdt);
2719 			if (tid == em->em_tid)
2720 				return (tdt);
2721 		}
2722 		PROC_UNLOCK(p);
2723 	}
2724 	return (NULL);
2725 }
2726 
2727 void
2728 linux_to_bsd_waitopts(int options, int *bsdopts)
2729 {
2730 
2731 	if (options & LINUX_WNOHANG)
2732 		*bsdopts |= WNOHANG;
2733 	if (options & LINUX_WUNTRACED)
2734 		*bsdopts |= WUNTRACED;
2735 	if (options & LINUX_WEXITED)
2736 		*bsdopts |= WEXITED;
2737 	if (options & LINUX_WCONTINUED)
2738 		*bsdopts |= WCONTINUED;
2739 	if (options & LINUX_WNOWAIT)
2740 		*bsdopts |= WNOWAIT;
2741 
2742 	if (options & __WCLONE)
2743 		*bsdopts |= WLINUXCLONE;
2744 }
2745 
2746 int
2747 linux_getrandom(struct thread *td, struct linux_getrandom_args *args)
2748 {
2749 	struct uio uio;
2750 	struct iovec iov;
2751 	int error;
2752 
2753 	if (args->flags & ~(LINUX_GRND_NONBLOCK|LINUX_GRND_RANDOM))
2754 		return (EINVAL);
2755 	if (args->count > INT_MAX)
2756 		args->count = INT_MAX;
2757 
2758 	iov.iov_base = args->buf;
2759 	iov.iov_len = args->count;
2760 
2761 	uio.uio_iov = &iov;
2762 	uio.uio_iovcnt = 1;
2763 	uio.uio_resid = iov.iov_len;
2764 	uio.uio_segflg = UIO_USERSPACE;
2765 	uio.uio_rw = UIO_READ;
2766 	uio.uio_td = td;
2767 
2768 	error = read_random_uio(&uio, args->flags & LINUX_GRND_NONBLOCK);
2769 	if (error == 0)
2770 		td->td_retval[0] = args->count - uio.uio_resid;
2771 	return (error);
2772 }
2773 
2774 int
2775 linux_mincore(struct thread *td, struct linux_mincore_args *args)
2776 {
2777 
2778 	/* Needs to be page-aligned */
2779 	if (args->start & PAGE_MASK)
2780 		return (EINVAL);
2781 	return (kern_mincore(td, args->start, args->len, args->vec));
2782 }
2783 
2784 #define	SYSLOG_TAG	"<6>"
2785 
2786 int
2787 linux_syslog(struct thread *td, struct linux_syslog_args *args)
2788 {
2789 	char buf[128], *src, *dst;
2790 	u_int seq;
2791 	int buflen, error;
2792 
2793 	if (args->type != LINUX_SYSLOG_ACTION_READ_ALL) {
2794 		linux_msg(td, "syslog unsupported type 0x%x", args->type);
2795 		return (EINVAL);
2796 	}
2797 
2798 	if (args->len < 6) {
2799 		td->td_retval[0] = 0;
2800 		return (0);
2801 	}
2802 
2803 	error = priv_check(td, PRIV_MSGBUF);
2804 	if (error)
2805 		return (error);
2806 
2807 	mtx_lock(&msgbuf_lock);
2808 	msgbuf_peekbytes(msgbufp, NULL, 0, &seq);
2809 	mtx_unlock(&msgbuf_lock);
2810 
2811 	dst = args->buf;
2812 	error = copyout(&SYSLOG_TAG, dst, sizeof(SYSLOG_TAG));
2813 	/* The -1 is to skip the trailing '\0'. */
2814 	dst += sizeof(SYSLOG_TAG) - 1;
2815 
2816 	while (error == 0) {
2817 		mtx_lock(&msgbuf_lock);
2818 		buflen = msgbuf_peekbytes(msgbufp, buf, sizeof(buf), &seq);
2819 		mtx_unlock(&msgbuf_lock);
2820 
2821 		if (buflen == 0)
2822 			break;
2823 
2824 		for (src = buf; src < buf + buflen && error == 0; src++) {
2825 			if (*src == '\0')
2826 				continue;
2827 
2828 			if (dst >= args->buf + args->len)
2829 				goto out;
2830 
2831 			error = copyout(src, dst, 1);
2832 			dst++;
2833 
2834 			if (*src == '\n' && *(src + 1) != '<' &&
2835 			    dst + sizeof(SYSLOG_TAG) < args->buf + args->len) {
2836 				error = copyout(&SYSLOG_TAG,
2837 				    dst, sizeof(SYSLOG_TAG));
2838 				dst += sizeof(SYSLOG_TAG) - 1;
2839 			}
2840 		}
2841 	}
2842 out:
2843 	td->td_retval[0] = dst - args->buf;
2844 	return (error);
2845 }
2846 
2847 int
2848 linux_getcpu(struct thread *td, struct linux_getcpu_args *args)
2849 {
2850 	int cpu, error, node;
2851 
2852 	cpu = td->td_oncpu; /* Make sure it doesn't change during copyout(9) */
2853 	error = 0;
2854 	node = cpuid_to_pcpu[cpu]->pc_domain;
2855 
2856 	if (args->cpu != NULL)
2857 		error = copyout(&cpu, args->cpu, sizeof(l_int));
2858 	if (args->node != NULL)
2859 		error = copyout(&node, args->node, sizeof(l_int));
2860 	return (error);
2861 }
2862 
2863 #if defined(__i386__) || defined(__amd64__)
2864 int
2865 linux_poll(struct thread *td, struct linux_poll_args *args)
2866 {
2867 	struct timespec ts, *tsp;
2868 
2869 	if (args->timeout != INFTIM) {
2870 		if (args->timeout < 0)
2871 			return (EINVAL);
2872 		ts.tv_sec = args->timeout / 1000;
2873 		ts.tv_nsec = (args->timeout % 1000) * 1000000;
2874 		tsp = &ts;
2875 	} else
2876 		tsp = NULL;
2877 
2878 	return (linux_common_ppoll(td, args->fds, args->nfds,
2879 	    tsp, NULL, 0));
2880 }
2881 #endif /* __i386__ || __amd64__ */
2882 
2883 int
2884 linux_seccomp(struct thread *td, struct linux_seccomp_args *args)
2885 {
2886 
2887 	switch (args->op) {
2888 	case LINUX_SECCOMP_GET_ACTION_AVAIL:
2889 		return (EOPNOTSUPP);
2890 	default:
2891 		/*
2892 		 * Ignore unknown operations, just like Linux kernel built
2893 		 * without CONFIG_SECCOMP.
2894 		 */
2895 		return (EINVAL);
2896 	}
2897 }
2898