xref: /freebsd/sys/compat/linux/linux_misc.c (revision b0d29bc47dba79f6f38e67eabadfb4b32ffd9390)
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/vm_object.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_dtrace.h>
94 #include <compat/linux/linux_file.h>
95 #include <compat/linux/linux_mib.h>
96 #include <compat/linux/linux_signal.h>
97 #include <compat/linux/linux_timer.h>
98 #include <compat/linux/linux_util.h>
99 #include <compat/linux/linux_sysproto.h>
100 #include <compat/linux/linux_emul.h>
101 #include <compat/linux/linux_misc.h>
102 
103 /**
104  * Special DTrace provider for the linuxulator.
105  *
106  * In this file we define the provider for the entire linuxulator. All
107  * modules (= files of the linuxulator) use it.
108  *
109  * We define a different name depending on the emulated bitsize, see
110  * ../../<ARCH>/linux{,32}/linux.h, e.g.:
111  *      native bitsize          = linuxulator
112  *      amd64, 32bit emulation  = linuxulator32
113  */
114 LIN_SDT_PROVIDER_DEFINE(LINUX_DTRACE);
115 
116 int stclohz;				/* Statistics clock frequency */
117 
118 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
119 	RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
120 	RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
121 	RLIMIT_MEMLOCK, RLIMIT_AS
122 };
123 
124 struct l_sysinfo {
125 	l_long		uptime;		/* Seconds since boot */
126 	l_ulong		loads[3];	/* 1, 5, and 15 minute load averages */
127 #define LINUX_SYSINFO_LOADS_SCALE 65536
128 	l_ulong		totalram;	/* Total usable main memory size */
129 	l_ulong		freeram;	/* Available memory size */
130 	l_ulong		sharedram;	/* Amount of shared memory */
131 	l_ulong		bufferram;	/* Memory used by buffers */
132 	l_ulong		totalswap;	/* Total swap space size */
133 	l_ulong		freeswap;	/* swap space still available */
134 	l_ushort	procs;		/* Number of current processes */
135 	l_ushort	pads;
136 	l_ulong		totalbig;
137 	l_ulong		freebig;
138 	l_uint		mem_unit;
139 	char		_f[20-2*sizeof(l_long)-sizeof(l_int)];	/* padding */
140 };
141 
142 struct l_pselect6arg {
143 	l_uintptr_t	ss;
144 	l_size_t	ss_len;
145 };
146 
147 static int	linux_utimensat_nsec_valid(l_long);
148 
149 
150 int
151 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
152 {
153 	struct l_sysinfo sysinfo;
154 	vm_object_t object;
155 	int i, j;
156 	struct timespec ts;
157 
158 	bzero(&sysinfo, sizeof(sysinfo));
159 	getnanouptime(&ts);
160 	if (ts.tv_nsec != 0)
161 		ts.tv_sec++;
162 	sysinfo.uptime = ts.tv_sec;
163 
164 	/* Use the information from the mib to get our load averages */
165 	for (i = 0; i < 3; i++)
166 		sysinfo.loads[i] = averunnable.ldavg[i] *
167 		    LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
168 
169 	sysinfo.totalram = physmem * PAGE_SIZE;
170 	sysinfo.freeram = sysinfo.totalram - vm_wire_count() * PAGE_SIZE;
171 
172 	sysinfo.sharedram = 0;
173 	mtx_lock(&vm_object_list_mtx);
174 	TAILQ_FOREACH(object, &vm_object_list, object_list)
175 		if (object->shadow_count > 1)
176 			sysinfo.sharedram += object->resident_page_count;
177 	mtx_unlock(&vm_object_list_mtx);
178 
179 	sysinfo.sharedram *= PAGE_SIZE;
180 	sysinfo.bufferram = 0;
181 
182 	swap_pager_status(&i, &j);
183 	sysinfo.totalswap = i * PAGE_SIZE;
184 	sysinfo.freeswap = (i - j) * PAGE_SIZE;
185 
186 	sysinfo.procs = nprocs;
187 
188 	/* The following are only present in newer Linux kernels. */
189 	sysinfo.totalbig = 0;
190 	sysinfo.freebig = 0;
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 	LCONVPATHEXIST(td, args->library, &library);
266 
267 	a_out = NULL;
268 	vp = NULL;
269 	locked = false;
270 	textset = false;
271 	opened = false;
272 
273 	NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
274 	    UIO_SYSSPACE, library, td);
275 	error = namei(&ni);
276 	LFREEPATH(library);
277 	if (error)
278 		goto cleanup;
279 
280 	vp = ni.ni_vp;
281 	NDFREE(&ni, NDF_ONLY_PNBUF);
282 
283 	/*
284 	 * From here on down, we have a locked vnode that must be unlocked.
285 	 * XXX: The code below largely duplicates exec_check_permissions().
286 	 */
287 	locked = true;
288 
289 	/* Executable? */
290 	error = VOP_GETATTR(vp, &attr, td->td_ucred);
291 	if (error)
292 		goto cleanup;
293 
294 	if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
295 	    ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
296 		/* EACCESS is what exec(2) returns. */
297 		error = ENOEXEC;
298 		goto cleanup;
299 	}
300 
301 	/* Sensible size? */
302 	if (attr.va_size == 0) {
303 		error = ENOEXEC;
304 		goto cleanup;
305 	}
306 
307 	/* Can we access it? */
308 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
309 	if (error)
310 		goto cleanup;
311 
312 	/*
313 	 * XXX: This should use vn_open() so that it is properly authorized,
314 	 * and to reduce code redundancy all over the place here.
315 	 * XXX: Not really, it duplicates far more of exec_check_permissions()
316 	 * than vn_open().
317 	 */
318 #ifdef MAC
319 	error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
320 	if (error)
321 		goto cleanup;
322 #endif
323 	error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
324 	if (error)
325 		goto cleanup;
326 	opened = true;
327 
328 	/* Pull in executable header into exec_map */
329 	error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
330 	    VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
331 	if (error)
332 		goto cleanup;
333 
334 	/* Is it a Linux binary ? */
335 	if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
336 		error = ENOEXEC;
337 		goto cleanup;
338 	}
339 
340 	/*
341 	 * While we are here, we should REALLY do some more checks
342 	 */
343 
344 	/* Set file/virtual offset based on a.out variant. */
345 	switch ((int)(a_out->a_magic & 0xffff)) {
346 	case 0413:			/* ZMAGIC */
347 		file_offset = 1024;
348 		break;
349 	case 0314:			/* QMAGIC */
350 		file_offset = 0;
351 		break;
352 	default:
353 		error = ENOEXEC;
354 		goto cleanup;
355 	}
356 
357 	bss_size = round_page(a_out->a_bss);
358 
359 	/* Check various fields in header for validity/bounds. */
360 	if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
361 		error = ENOEXEC;
362 		goto cleanup;
363 	}
364 
365 	/* text + data can't exceed file size */
366 	if (a_out->a_data + a_out->a_text > attr.va_size) {
367 		error = EFAULT;
368 		goto cleanup;
369 	}
370 
371 	/*
372 	 * text/data/bss must not exceed limits
373 	 * XXX - this is not complete. it should check current usage PLUS
374 	 * the resources needed by this library.
375 	 */
376 	PROC_LOCK(td->td_proc);
377 	if (a_out->a_text > maxtsiz ||
378 	    a_out->a_data + bss_size > lim_cur_proc(td->td_proc, RLIMIT_DATA) ||
379 	    racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
380 	    bss_size) != 0) {
381 		PROC_UNLOCK(td->td_proc);
382 		error = ENOMEM;
383 		goto cleanup;
384 	}
385 	PROC_UNLOCK(td->td_proc);
386 
387 	/*
388 	 * Prevent more writers.
389 	 */
390 	error = VOP_SET_TEXT(vp);
391 	if (error != 0)
392 		goto cleanup;
393 	textset = true;
394 
395 	/*
396 	 * Lock no longer needed
397 	 */
398 	locked = false;
399 	VOP_UNLOCK(vp);
400 
401 	/*
402 	 * Check if file_offset page aligned. Currently we cannot handle
403 	 * misalinged file offsets, and so we read in the entire image
404 	 * (what a waste).
405 	 */
406 	if (file_offset & PAGE_MASK) {
407 		/* Map text+data read/write/execute */
408 
409 		/* a_entry is the load address and is page aligned */
410 		vmaddr = trunc_page(a_out->a_entry);
411 
412 		/* get anon user mapping, read+write+execute */
413 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
414 		    &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
415 		    VM_PROT_ALL, VM_PROT_ALL, 0);
416 		if (error)
417 			goto cleanup;
418 
419 		error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
420 		    a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
421 		    td->td_ucred, NOCRED, &aresid, td);
422 		if (error != 0)
423 			goto cleanup;
424 		if (aresid != 0) {
425 			error = ENOEXEC;
426 			goto cleanup;
427 		}
428 	} else {
429 		/*
430 		 * for QMAGIC, a_entry is 20 bytes beyond the load address
431 		 * to skip the executable header
432 		 */
433 		vmaddr = trunc_page(a_out->a_entry);
434 
435 		/*
436 		 * Map it all into the process's space as a single
437 		 * copy-on-write "data" segment.
438 		 */
439 		map = &td->td_proc->p_vmspace->vm_map;
440 		error = vm_mmap(map, &vmaddr,
441 		    a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
442 		    MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
443 		if (error)
444 			goto cleanup;
445 		vm_map_lock(map);
446 		if (!vm_map_lookup_entry(map, vmaddr, &entry)) {
447 			vm_map_unlock(map);
448 			error = EDOOFUS;
449 			goto cleanup;
450 		}
451 		entry->eflags |= MAP_ENTRY_VN_EXEC;
452 		vm_map_unlock(map);
453 		textset = false;
454 	}
455 
456 	if (bss_size != 0) {
457 		/* Calculate BSS start address */
458 		vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
459 		    a_out->a_data;
460 
461 		/* allocate some 'anon' space */
462 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
463 		    &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
464 		    VM_PROT_ALL, 0);
465 		if (error)
466 			goto cleanup;
467 	}
468 
469 cleanup:
470 	if (opened) {
471 		if (locked)
472 			VOP_UNLOCK(vp);
473 		locked = false;
474 		VOP_CLOSE(vp, FREAD, td->td_ucred, td);
475 	}
476 	if (textset) {
477 		if (!locked) {
478 			locked = true;
479 			VOP_LOCK(vp, LK_SHARED | LK_RETRY);
480 		}
481 		VOP_UNSET_TEXT_CHECKED(vp);
482 	}
483 	if (locked)
484 		VOP_UNLOCK(vp);
485 
486 	/* Release the temporary mapping. */
487 	if (a_out)
488 		kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
489 
490 	return (error);
491 }
492 
493 #endif	/* __i386__ */
494 
495 #ifdef LINUX_LEGACY_SYSCALLS
496 int
497 linux_select(struct thread *td, struct linux_select_args *args)
498 {
499 	l_timeval ltv;
500 	struct timeval tv0, tv1, utv, *tvp;
501 	int error;
502 
503 	/*
504 	 * Store current time for computation of the amount of
505 	 * time left.
506 	 */
507 	if (args->timeout) {
508 		if ((error = copyin(args->timeout, &ltv, sizeof(ltv))))
509 			goto select_out;
510 		utv.tv_sec = ltv.tv_sec;
511 		utv.tv_usec = ltv.tv_usec;
512 
513 		if (itimerfix(&utv)) {
514 			/*
515 			 * The timeval was invalid.  Convert it to something
516 			 * valid that will act as it does under Linux.
517 			 */
518 			utv.tv_sec += utv.tv_usec / 1000000;
519 			utv.tv_usec %= 1000000;
520 			if (utv.tv_usec < 0) {
521 				utv.tv_sec -= 1;
522 				utv.tv_usec += 1000000;
523 			}
524 			if (utv.tv_sec < 0)
525 				timevalclear(&utv);
526 		}
527 		microtime(&tv0);
528 		tvp = &utv;
529 	} else
530 		tvp = NULL;
531 
532 	error = kern_select(td, args->nfds, args->readfds, args->writefds,
533 	    args->exceptfds, tvp, LINUX_NFDBITS);
534 	if (error)
535 		goto select_out;
536 
537 	if (args->timeout) {
538 		if (td->td_retval[0]) {
539 			/*
540 			 * Compute how much time was left of the timeout,
541 			 * by subtracting the current time and the time
542 			 * before we started the call, and subtracting
543 			 * that result from the user-supplied value.
544 			 */
545 			microtime(&tv1);
546 			timevalsub(&tv1, &tv0);
547 			timevalsub(&utv, &tv1);
548 			if (utv.tv_sec < 0)
549 				timevalclear(&utv);
550 		} else
551 			timevalclear(&utv);
552 		ltv.tv_sec = utv.tv_sec;
553 		ltv.tv_usec = utv.tv_usec;
554 		if ((error = copyout(&ltv, args->timeout, sizeof(ltv))))
555 			goto select_out;
556 	}
557 
558 select_out:
559 	return (error);
560 }
561 #endif
562 
563 int
564 linux_mremap(struct thread *td, struct linux_mremap_args *args)
565 {
566 	uintptr_t addr;
567 	size_t len;
568 	int error = 0;
569 
570 	if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
571 		td->td_retval[0] = 0;
572 		return (EINVAL);
573 	}
574 
575 	/*
576 	 * Check for the page alignment.
577 	 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
578 	 */
579 	if (args->addr & PAGE_MASK) {
580 		td->td_retval[0] = 0;
581 		return (EINVAL);
582 	}
583 
584 	args->new_len = round_page(args->new_len);
585 	args->old_len = round_page(args->old_len);
586 
587 	if (args->new_len > args->old_len) {
588 		td->td_retval[0] = 0;
589 		return (ENOMEM);
590 	}
591 
592 	if (args->new_len < args->old_len) {
593 		addr = args->addr + args->new_len;
594 		len = args->old_len - args->new_len;
595 		error = kern_munmap(td, addr, len);
596 	}
597 
598 	td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
599 	return (error);
600 }
601 
602 #define LINUX_MS_ASYNC       0x0001
603 #define LINUX_MS_INVALIDATE  0x0002
604 #define LINUX_MS_SYNC        0x0004
605 
606 int
607 linux_msync(struct thread *td, struct linux_msync_args *args)
608 {
609 
610 	return (kern_msync(td, args->addr, args->len,
611 	    args->fl & ~LINUX_MS_SYNC));
612 }
613 
614 #ifdef LINUX_LEGACY_SYSCALLS
615 int
616 linux_time(struct thread *td, struct linux_time_args *args)
617 {
618 	struct timeval tv;
619 	l_time_t tm;
620 	int error;
621 
622 	microtime(&tv);
623 	tm = tv.tv_sec;
624 	if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
625 		return (error);
626 	td->td_retval[0] = tm;
627 	return (0);
628 }
629 #endif
630 
631 struct l_times_argv {
632 	l_clock_t	tms_utime;
633 	l_clock_t	tms_stime;
634 	l_clock_t	tms_cutime;
635 	l_clock_t	tms_cstime;
636 };
637 
638 
639 /*
640  * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
641  * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
642  * auxiliary vector entry.
643  */
644 #define	CLK_TCK		100
645 
646 #define	CONVOTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
647 #define	CONVNTCK(r)	(r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
648 
649 #define	CONVTCK(r)	(linux_kernver(td) >= LINUX_KERNVER_2004000 ?		\
650 			    CONVNTCK(r) : CONVOTCK(r))
651 
652 int
653 linux_times(struct thread *td, struct linux_times_args *args)
654 {
655 	struct timeval tv, utime, stime, cutime, cstime;
656 	struct l_times_argv tms;
657 	struct proc *p;
658 	int error;
659 
660 	if (args->buf != NULL) {
661 		p = td->td_proc;
662 		PROC_LOCK(p);
663 		PROC_STATLOCK(p);
664 		calcru(p, &utime, &stime);
665 		PROC_STATUNLOCK(p);
666 		calccru(p, &cutime, &cstime);
667 		PROC_UNLOCK(p);
668 
669 		tms.tms_utime = CONVTCK(utime);
670 		tms.tms_stime = CONVTCK(stime);
671 
672 		tms.tms_cutime = CONVTCK(cutime);
673 		tms.tms_cstime = CONVTCK(cstime);
674 
675 		if ((error = copyout(&tms, args->buf, sizeof(tms))))
676 			return (error);
677 	}
678 
679 	microuptime(&tv);
680 	td->td_retval[0] = (int)CONVTCK(tv);
681 	return (0);
682 }
683 
684 int
685 linux_newuname(struct thread *td, struct linux_newuname_args *args)
686 {
687 	struct l_new_utsname utsname;
688 	char osname[LINUX_MAX_UTSNAME];
689 	char osrelease[LINUX_MAX_UTSNAME];
690 	char *p;
691 
692 	linux_get_osname(td, osname);
693 	linux_get_osrelease(td, osrelease);
694 
695 	bzero(&utsname, sizeof(utsname));
696 	strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
697 	getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
698 	getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
699 	strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
700 	strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
701 	for (p = utsname.version; *p != '\0'; ++p)
702 		if (*p == '\n') {
703 			*p = '\0';
704 			break;
705 		}
706 	strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME);
707 
708 	return (copyout(&utsname, args->buf, sizeof(utsname)));
709 }
710 
711 struct l_utimbuf {
712 	l_time_t l_actime;
713 	l_time_t l_modtime;
714 };
715 
716 #ifdef LINUX_LEGACY_SYSCALLS
717 int
718 linux_utime(struct thread *td, struct linux_utime_args *args)
719 {
720 	struct timeval tv[2], *tvp;
721 	struct l_utimbuf lut;
722 	char *fname;
723 	int error;
724 
725 	LCONVPATHEXIST(td, args->fname, &fname);
726 
727 	if (args->times) {
728 		if ((error = copyin(args->times, &lut, sizeof lut))) {
729 			LFREEPATH(fname);
730 			return (error);
731 		}
732 		tv[0].tv_sec = lut.l_actime;
733 		tv[0].tv_usec = 0;
734 		tv[1].tv_sec = lut.l_modtime;
735 		tv[1].tv_usec = 0;
736 		tvp = tv;
737 	} else
738 		tvp = NULL;
739 
740 	error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, tvp,
741 	    UIO_SYSSPACE);
742 	LFREEPATH(fname);
743 	return (error);
744 }
745 #endif
746 
747 #ifdef LINUX_LEGACY_SYSCALLS
748 int
749 linux_utimes(struct thread *td, struct linux_utimes_args *args)
750 {
751 	l_timeval ltv[2];
752 	struct timeval tv[2], *tvp = NULL;
753 	char *fname;
754 	int error;
755 
756 	LCONVPATHEXIST(td, args->fname, &fname);
757 
758 	if (args->tptr != NULL) {
759 		if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
760 			LFREEPATH(fname);
761 			return (error);
762 		}
763 		tv[0].tv_sec = ltv[0].tv_sec;
764 		tv[0].tv_usec = ltv[0].tv_usec;
765 		tv[1].tv_sec = ltv[1].tv_sec;
766 		tv[1].tv_usec = ltv[1].tv_usec;
767 		tvp = tv;
768 	}
769 
770 	error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE,
771 	    tvp, UIO_SYSSPACE);
772 	LFREEPATH(fname);
773 	return (error);
774 }
775 #endif
776 
777 static int
778 linux_utimensat_nsec_valid(l_long nsec)
779 {
780 
781 	if (nsec == LINUX_UTIME_OMIT || nsec == LINUX_UTIME_NOW)
782 		return (0);
783 	if (nsec >= 0 && nsec <= 999999999)
784 		return (0);
785 	return (1);
786 }
787 
788 int
789 linux_utimensat(struct thread *td, struct linux_utimensat_args *args)
790 {
791 	struct l_timespec l_times[2];
792 	struct timespec times[2], *timesp = NULL;
793 	char *path = NULL;
794 	int error, dfd, flags = 0;
795 
796 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
797 
798 	if (args->flags & ~LINUX_AT_SYMLINK_NOFOLLOW)
799 		return (EINVAL);
800 
801 	if (args->times != NULL) {
802 		error = copyin(args->times, l_times, sizeof(l_times));
803 		if (error != 0)
804 			return (error);
805 
806 		if (linux_utimensat_nsec_valid(l_times[0].tv_nsec) != 0 ||
807 		    linux_utimensat_nsec_valid(l_times[1].tv_nsec) != 0)
808 			return (EINVAL);
809 
810 		times[0].tv_sec = l_times[0].tv_sec;
811 		switch (l_times[0].tv_nsec)
812 		{
813 		case LINUX_UTIME_OMIT:
814 			times[0].tv_nsec = UTIME_OMIT;
815 			break;
816 		case LINUX_UTIME_NOW:
817 			times[0].tv_nsec = UTIME_NOW;
818 			break;
819 		default:
820 			times[0].tv_nsec = l_times[0].tv_nsec;
821 		}
822 
823 		times[1].tv_sec = l_times[1].tv_sec;
824 		switch (l_times[1].tv_nsec)
825 		{
826 		case LINUX_UTIME_OMIT:
827 			times[1].tv_nsec = UTIME_OMIT;
828 			break;
829 		case LINUX_UTIME_NOW:
830 			times[1].tv_nsec = UTIME_NOW;
831 			break;
832 		default:
833 			times[1].tv_nsec = l_times[1].tv_nsec;
834 			break;
835 		}
836 		timesp = times;
837 
838 		/* This breaks POSIX, but is what the Linux kernel does
839 		 * _on purpose_ (documented in the man page for utimensat(2)),
840 		 * so we must follow that behaviour. */
841 		if (times[0].tv_nsec == UTIME_OMIT &&
842 		    times[1].tv_nsec == UTIME_OMIT)
843 			return (0);
844 	}
845 
846 	if (args->pathname != NULL)
847 		LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
848 	else if (args->flags != 0)
849 		return (EINVAL);
850 
851 	if (args->flags & LINUX_AT_SYMLINK_NOFOLLOW)
852 		flags |= AT_SYMLINK_NOFOLLOW;
853 
854 	if (path == NULL)
855 		error = kern_futimens(td, dfd, timesp, UIO_SYSSPACE);
856 	else {
857 		error = kern_utimensat(td, dfd, path, UIO_SYSSPACE, timesp,
858 			UIO_SYSSPACE, flags);
859 		LFREEPATH(path);
860 	}
861 
862 	return (error);
863 }
864 
865 #ifdef LINUX_LEGACY_SYSCALLS
866 int
867 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
868 {
869 	l_timeval ltv[2];
870 	struct timeval tv[2], *tvp = NULL;
871 	char *fname;
872 	int error, dfd;
873 
874 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
875 	LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
876 
877 	if (args->utimes != NULL) {
878 		if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
879 			LFREEPATH(fname);
880 			return (error);
881 		}
882 		tv[0].tv_sec = ltv[0].tv_sec;
883 		tv[0].tv_usec = ltv[0].tv_usec;
884 		tv[1].tv_sec = ltv[1].tv_sec;
885 		tv[1].tv_usec = ltv[1].tv_usec;
886 		tvp = tv;
887 	}
888 
889 	error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
890 	LFREEPATH(fname);
891 	return (error);
892 }
893 #endif
894 
895 static int
896 linux_common_wait(struct thread *td, int pid, int *statusp,
897     int options, struct __wrusage *wrup)
898 {
899 	siginfo_t siginfo;
900 	idtype_t idtype;
901 	id_t id;
902 	int error, status, tmpstat;
903 
904 	if (pid == WAIT_ANY) {
905 		idtype = P_ALL;
906 		id = 0;
907 	} else if (pid < 0) {
908 		idtype = P_PGID;
909 		id = (id_t)-pid;
910 	} else {
911 		idtype = P_PID;
912 		id = (id_t)pid;
913 	}
914 
915 	/*
916 	 * For backward compatibility we implicitly add flags WEXITED
917 	 * and WTRAPPED here.
918 	 */
919 	options |= WEXITED | WTRAPPED;
920 	error = kern_wait6(td, idtype, id, &status, options, wrup, &siginfo);
921 	if (error)
922 		return (error);
923 
924 	if (statusp) {
925 		tmpstat = status & 0xffff;
926 		if (WIFSIGNALED(tmpstat)) {
927 			tmpstat = (tmpstat & 0xffffff80) |
928 			    bsd_to_linux_signal(WTERMSIG(tmpstat));
929 		} else if (WIFSTOPPED(tmpstat)) {
930 			tmpstat = (tmpstat & 0xffff00ff) |
931 			    (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8);
932 #if defined(__amd64__) && !defined(COMPAT_LINUX32)
933 			if (WSTOPSIG(status) == SIGTRAP) {
934 				tmpstat = linux_ptrace_status(td,
935 				    siginfo.si_pid, tmpstat);
936 			}
937 #endif
938 		} else if (WIFCONTINUED(tmpstat)) {
939 			tmpstat = 0xffff;
940 		}
941 		error = copyout(&tmpstat, statusp, sizeof(int));
942 	}
943 
944 	return (error);
945 }
946 
947 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
948 int
949 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
950 {
951 	struct linux_wait4_args wait4_args;
952 
953 	wait4_args.pid = args->pid;
954 	wait4_args.status = args->status;
955 	wait4_args.options = args->options;
956 	wait4_args.rusage = NULL;
957 
958 	return (linux_wait4(td, &wait4_args));
959 }
960 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
961 
962 int
963 linux_wait4(struct thread *td, struct linux_wait4_args *args)
964 {
965 	int error, options;
966 	struct __wrusage wru, *wrup;
967 
968 	if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG |
969 	    LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
970 		return (EINVAL);
971 
972 	options = WEXITED;
973 	linux_to_bsd_waitopts(args->options, &options);
974 
975 	if (args->rusage != NULL)
976 		wrup = &wru;
977 	else
978 		wrup = NULL;
979 	error = linux_common_wait(td, args->pid, args->status, options, wrup);
980 	if (error != 0)
981 		return (error);
982 	if (args->rusage != NULL)
983 		error = linux_copyout_rusage(&wru.wru_self, args->rusage);
984 	return (error);
985 }
986 
987 int
988 linux_waitid(struct thread *td, struct linux_waitid_args *args)
989 {
990 	int status, options, sig;
991 	struct __wrusage wru;
992 	siginfo_t siginfo;
993 	l_siginfo_t lsi;
994 	idtype_t idtype;
995 	struct proc *p;
996 	int error;
997 
998 	options = 0;
999 	linux_to_bsd_waitopts(args->options, &options);
1000 
1001 	if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED))
1002 		return (EINVAL);
1003 	if (!(options & (WEXITED | WUNTRACED | WCONTINUED)))
1004 		return (EINVAL);
1005 
1006 	switch (args->idtype) {
1007 	case LINUX_P_ALL:
1008 		idtype = P_ALL;
1009 		break;
1010 	case LINUX_P_PID:
1011 		if (args->id <= 0)
1012 			return (EINVAL);
1013 		idtype = P_PID;
1014 		break;
1015 	case LINUX_P_PGID:
1016 		if (args->id <= 0)
1017 			return (EINVAL);
1018 		idtype = P_PGID;
1019 		break;
1020 	default:
1021 		return (EINVAL);
1022 	}
1023 
1024 	error = kern_wait6(td, idtype, args->id, &status, options,
1025 	    &wru, &siginfo);
1026 	if (error != 0)
1027 		return (error);
1028 	if (args->rusage != NULL) {
1029 		error = linux_copyout_rusage(&wru.wru_children,
1030 		    args->rusage);
1031 		if (error != 0)
1032 			return (error);
1033 	}
1034 	if (args->info != NULL) {
1035 		p = td->td_proc;
1036 		bzero(&lsi, sizeof(lsi));
1037 		if (td->td_retval[0] != 0) {
1038 			sig = bsd_to_linux_signal(siginfo.si_signo);
1039 			siginfo_to_lsiginfo(&siginfo, &lsi, sig);
1040 		}
1041 		error = copyout(&lsi, args->info, sizeof(lsi));
1042 	}
1043 	td->td_retval[0] = 0;
1044 
1045 	return (error);
1046 }
1047 
1048 #ifdef LINUX_LEGACY_SYSCALLS
1049 int
1050 linux_mknod(struct thread *td, struct linux_mknod_args *args)
1051 {
1052 	char *path;
1053 	int error;
1054 
1055 	LCONVPATHCREAT(td, args->path, &path);
1056 
1057 	switch (args->mode & S_IFMT) {
1058 	case S_IFIFO:
1059 	case S_IFSOCK:
1060 		error = kern_mkfifoat(td, AT_FDCWD, path, UIO_SYSSPACE,
1061 		    args->mode);
1062 		break;
1063 
1064 	case S_IFCHR:
1065 	case S_IFBLK:
1066 		error = kern_mknodat(td, AT_FDCWD, path, UIO_SYSSPACE,
1067 		    args->mode, args->dev);
1068 		break;
1069 
1070 	case S_IFDIR:
1071 		error = EPERM;
1072 		break;
1073 
1074 	case 0:
1075 		args->mode |= S_IFREG;
1076 		/* FALLTHROUGH */
1077 	case S_IFREG:
1078 		error = kern_openat(td, AT_FDCWD, path, UIO_SYSSPACE,
1079 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1080 		if (error == 0)
1081 			kern_close(td, td->td_retval[0]);
1082 		break;
1083 
1084 	default:
1085 		error = EINVAL;
1086 		break;
1087 	}
1088 	LFREEPATH(path);
1089 	return (error);
1090 }
1091 #endif
1092 
1093 int
1094 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
1095 {
1096 	char *path;
1097 	int error, dfd;
1098 
1099 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
1100 	LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
1101 
1102 	switch (args->mode & S_IFMT) {
1103 	case S_IFIFO:
1104 	case S_IFSOCK:
1105 		error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
1106 		break;
1107 
1108 	case S_IFCHR:
1109 	case S_IFBLK:
1110 		error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
1111 		    args->dev);
1112 		break;
1113 
1114 	case S_IFDIR:
1115 		error = EPERM;
1116 		break;
1117 
1118 	case 0:
1119 		args->mode |= S_IFREG;
1120 		/* FALLTHROUGH */
1121 	case S_IFREG:
1122 		error = kern_openat(td, dfd, path, UIO_SYSSPACE,
1123 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1124 		if (error == 0)
1125 			kern_close(td, td->td_retval[0]);
1126 		break;
1127 
1128 	default:
1129 		error = EINVAL;
1130 		break;
1131 	}
1132 	LFREEPATH(path);
1133 	return (error);
1134 }
1135 
1136 /*
1137  * UGH! This is just about the dumbest idea I've ever heard!!
1138  */
1139 int
1140 linux_personality(struct thread *td, struct linux_personality_args *args)
1141 {
1142 	struct linux_pemuldata *pem;
1143 	struct proc *p = td->td_proc;
1144 	uint32_t old;
1145 
1146 	PROC_LOCK(p);
1147 	pem = pem_find(p);
1148 	old = pem->persona;
1149 	if (args->per != 0xffffffff)
1150 		pem->persona = args->per;
1151 	PROC_UNLOCK(p);
1152 
1153 	td->td_retval[0] = old;
1154 	return (0);
1155 }
1156 
1157 struct l_itimerval {
1158 	l_timeval it_interval;
1159 	l_timeval it_value;
1160 };
1161 
1162 #define	B2L_ITIMERVAL(bip, lip)						\
1163 	(bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec;		\
1164 	(bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec;	\
1165 	(bip)->it_value.tv_sec = (lip)->it_value.tv_sec;		\
1166 	(bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1167 
1168 int
1169 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1170 {
1171 	int error;
1172 	struct l_itimerval ls;
1173 	struct itimerval aitv, oitv;
1174 
1175 	if (uap->itv == NULL) {
1176 		uap->itv = uap->oitv;
1177 		return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1178 	}
1179 
1180 	error = copyin(uap->itv, &ls, sizeof(ls));
1181 	if (error != 0)
1182 		return (error);
1183 	B2L_ITIMERVAL(&aitv, &ls);
1184 	error = kern_setitimer(td, uap->which, &aitv, &oitv);
1185 	if (error != 0 || uap->oitv == NULL)
1186 		return (error);
1187 	B2L_ITIMERVAL(&ls, &oitv);
1188 
1189 	return (copyout(&ls, uap->oitv, sizeof(ls)));
1190 }
1191 
1192 int
1193 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1194 {
1195 	int error;
1196 	struct l_itimerval ls;
1197 	struct itimerval aitv;
1198 
1199 	error = kern_getitimer(td, uap->which, &aitv);
1200 	if (error != 0)
1201 		return (error);
1202 	B2L_ITIMERVAL(&ls, &aitv);
1203 	return (copyout(&ls, uap->itv, sizeof(ls)));
1204 }
1205 
1206 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1207 int
1208 linux_nice(struct thread *td, struct linux_nice_args *args)
1209 {
1210 
1211 	return (kern_setpriority(td, PRIO_PROCESS, 0, args->inc));
1212 }
1213 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1214 
1215 int
1216 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1217 {
1218 	struct ucred *newcred, *oldcred;
1219 	l_gid_t *linux_gidset;
1220 	gid_t *bsd_gidset;
1221 	int ngrp, error;
1222 	struct proc *p;
1223 
1224 	ngrp = args->gidsetsize;
1225 	if (ngrp < 0 || ngrp >= ngroups_max + 1)
1226 		return (EINVAL);
1227 	linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK);
1228 	error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1229 	if (error)
1230 		goto out;
1231 	newcred = crget();
1232 	crextend(newcred, ngrp + 1);
1233 	p = td->td_proc;
1234 	PROC_LOCK(p);
1235 	oldcred = p->p_ucred;
1236 	crcopy(newcred, oldcred);
1237 
1238 	/*
1239 	 * cr_groups[0] holds egid. Setting the whole set from
1240 	 * the supplied set will cause egid to be changed too.
1241 	 * Keep cr_groups[0] unchanged to prevent that.
1242 	 */
1243 
1244 	if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS)) != 0) {
1245 		PROC_UNLOCK(p);
1246 		crfree(newcred);
1247 		goto out;
1248 	}
1249 
1250 	if (ngrp > 0) {
1251 		newcred->cr_ngroups = ngrp + 1;
1252 
1253 		bsd_gidset = newcred->cr_groups;
1254 		ngrp--;
1255 		while (ngrp >= 0) {
1256 			bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1257 			ngrp--;
1258 		}
1259 	} else
1260 		newcred->cr_ngroups = 1;
1261 
1262 	setsugid(p);
1263 	proc_set_cred(p, newcred);
1264 	PROC_UNLOCK(p);
1265 	crfree(oldcred);
1266 	error = 0;
1267 out:
1268 	free(linux_gidset, M_LINUX);
1269 	return (error);
1270 }
1271 
1272 int
1273 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1274 {
1275 	struct ucred *cred;
1276 	l_gid_t *linux_gidset;
1277 	gid_t *bsd_gidset;
1278 	int bsd_gidsetsz, ngrp, error;
1279 
1280 	cred = td->td_ucred;
1281 	bsd_gidset = cred->cr_groups;
1282 	bsd_gidsetsz = cred->cr_ngroups - 1;
1283 
1284 	/*
1285 	 * cr_groups[0] holds egid. Returning the whole set
1286 	 * here will cause a duplicate. Exclude cr_groups[0]
1287 	 * to prevent that.
1288 	 */
1289 
1290 	if ((ngrp = args->gidsetsize) == 0) {
1291 		td->td_retval[0] = bsd_gidsetsz;
1292 		return (0);
1293 	}
1294 
1295 	if (ngrp < bsd_gidsetsz)
1296 		return (EINVAL);
1297 
1298 	ngrp = 0;
1299 	linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1300 	    M_LINUX, M_WAITOK);
1301 	while (ngrp < bsd_gidsetsz) {
1302 		linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1303 		ngrp++;
1304 	}
1305 
1306 	error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1307 	free(linux_gidset, M_LINUX);
1308 	if (error)
1309 		return (error);
1310 
1311 	td->td_retval[0] = ngrp;
1312 	return (0);
1313 }
1314 
1315 int
1316 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1317 {
1318 	struct rlimit bsd_rlim;
1319 	struct l_rlimit rlim;
1320 	u_int which;
1321 	int error;
1322 
1323 	if (args->resource >= LINUX_RLIM_NLIMITS)
1324 		return (EINVAL);
1325 
1326 	which = linux_to_bsd_resource[args->resource];
1327 	if (which == -1)
1328 		return (EINVAL);
1329 
1330 	error = copyin(args->rlim, &rlim, sizeof(rlim));
1331 	if (error)
1332 		return (error);
1333 
1334 	bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1335 	bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1336 	return (kern_setrlimit(td, which, &bsd_rlim));
1337 }
1338 
1339 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1340 int
1341 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1342 {
1343 	struct l_rlimit rlim;
1344 	struct rlimit bsd_rlim;
1345 	u_int which;
1346 
1347 	if (args->resource >= LINUX_RLIM_NLIMITS)
1348 		return (EINVAL);
1349 
1350 	which = linux_to_bsd_resource[args->resource];
1351 	if (which == -1)
1352 		return (EINVAL);
1353 
1354 	lim_rlimit(td, which, &bsd_rlim);
1355 
1356 #ifdef COMPAT_LINUX32
1357 	rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1358 	if (rlim.rlim_cur == UINT_MAX)
1359 		rlim.rlim_cur = INT_MAX;
1360 	rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1361 	if (rlim.rlim_max == UINT_MAX)
1362 		rlim.rlim_max = INT_MAX;
1363 #else
1364 	rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1365 	if (rlim.rlim_cur == ULONG_MAX)
1366 		rlim.rlim_cur = LONG_MAX;
1367 	rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1368 	if (rlim.rlim_max == ULONG_MAX)
1369 		rlim.rlim_max = LONG_MAX;
1370 #endif
1371 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1372 }
1373 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1374 
1375 int
1376 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1377 {
1378 	struct l_rlimit rlim;
1379 	struct rlimit bsd_rlim;
1380 	u_int which;
1381 
1382 	if (args->resource >= LINUX_RLIM_NLIMITS)
1383 		return (EINVAL);
1384 
1385 	which = linux_to_bsd_resource[args->resource];
1386 	if (which == -1)
1387 		return (EINVAL);
1388 
1389 	lim_rlimit(td, which, &bsd_rlim);
1390 
1391 	rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1392 	rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1393 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1394 }
1395 
1396 int
1397 linux_sched_setscheduler(struct thread *td,
1398     struct linux_sched_setscheduler_args *args)
1399 {
1400 	struct sched_param sched_param;
1401 	struct thread *tdt;
1402 	int error, policy;
1403 
1404 	switch (args->policy) {
1405 	case LINUX_SCHED_OTHER:
1406 		policy = SCHED_OTHER;
1407 		break;
1408 	case LINUX_SCHED_FIFO:
1409 		policy = SCHED_FIFO;
1410 		break;
1411 	case LINUX_SCHED_RR:
1412 		policy = SCHED_RR;
1413 		break;
1414 	default:
1415 		return (EINVAL);
1416 	}
1417 
1418 	error = copyin(args->param, &sched_param, sizeof(sched_param));
1419 	if (error)
1420 		return (error);
1421 
1422 	if (linux_map_sched_prio) {
1423 		switch (policy) {
1424 		case SCHED_OTHER:
1425 			if (sched_param.sched_priority != 0)
1426 				return (EINVAL);
1427 
1428 			sched_param.sched_priority =
1429 			    PRI_MAX_TIMESHARE - PRI_MIN_TIMESHARE;
1430 			break;
1431 		case SCHED_FIFO:
1432 		case SCHED_RR:
1433 			if (sched_param.sched_priority < 1 ||
1434 			    sched_param.sched_priority >= LINUX_MAX_RT_PRIO)
1435 				return (EINVAL);
1436 
1437 			/*
1438 			 * Map [1, LINUX_MAX_RT_PRIO - 1] to
1439 			 * [0, RTP_PRIO_MAX - RTP_PRIO_MIN] (rounding down).
1440 			 */
1441 			sched_param.sched_priority =
1442 			    (sched_param.sched_priority - 1) *
1443 			    (RTP_PRIO_MAX - RTP_PRIO_MIN + 1) /
1444 			    (LINUX_MAX_RT_PRIO - 1);
1445 			break;
1446 		}
1447 	}
1448 
1449 	tdt = linux_tdfind(td, args->pid, -1);
1450 	if (tdt == NULL)
1451 		return (ESRCH);
1452 
1453 	error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1454 	PROC_UNLOCK(tdt->td_proc);
1455 	return (error);
1456 }
1457 
1458 int
1459 linux_sched_getscheduler(struct thread *td,
1460     struct linux_sched_getscheduler_args *args)
1461 {
1462 	struct thread *tdt;
1463 	int error, policy;
1464 
1465 	tdt = linux_tdfind(td, args->pid, -1);
1466 	if (tdt == NULL)
1467 		return (ESRCH);
1468 
1469 	error = kern_sched_getscheduler(td, tdt, &policy);
1470 	PROC_UNLOCK(tdt->td_proc);
1471 
1472 	switch (policy) {
1473 	case SCHED_OTHER:
1474 		td->td_retval[0] = LINUX_SCHED_OTHER;
1475 		break;
1476 	case SCHED_FIFO:
1477 		td->td_retval[0] = LINUX_SCHED_FIFO;
1478 		break;
1479 	case SCHED_RR:
1480 		td->td_retval[0] = LINUX_SCHED_RR;
1481 		break;
1482 	}
1483 	return (error);
1484 }
1485 
1486 int
1487 linux_sched_get_priority_max(struct thread *td,
1488     struct linux_sched_get_priority_max_args *args)
1489 {
1490 	struct sched_get_priority_max_args bsd;
1491 
1492 	if (linux_map_sched_prio) {
1493 		switch (args->policy) {
1494 		case LINUX_SCHED_OTHER:
1495 			td->td_retval[0] = 0;
1496 			return (0);
1497 		case LINUX_SCHED_FIFO:
1498 		case LINUX_SCHED_RR:
1499 			td->td_retval[0] = LINUX_MAX_RT_PRIO - 1;
1500 			return (0);
1501 		default:
1502 			return (EINVAL);
1503 		}
1504 	}
1505 
1506 	switch (args->policy) {
1507 	case LINUX_SCHED_OTHER:
1508 		bsd.policy = SCHED_OTHER;
1509 		break;
1510 	case LINUX_SCHED_FIFO:
1511 		bsd.policy = SCHED_FIFO;
1512 		break;
1513 	case LINUX_SCHED_RR:
1514 		bsd.policy = SCHED_RR;
1515 		break;
1516 	default:
1517 		return (EINVAL);
1518 	}
1519 	return (sys_sched_get_priority_max(td, &bsd));
1520 }
1521 
1522 int
1523 linux_sched_get_priority_min(struct thread *td,
1524     struct linux_sched_get_priority_min_args *args)
1525 {
1526 	struct sched_get_priority_min_args bsd;
1527 
1528 	if (linux_map_sched_prio) {
1529 		switch (args->policy) {
1530 		case LINUX_SCHED_OTHER:
1531 			td->td_retval[0] = 0;
1532 			return (0);
1533 		case LINUX_SCHED_FIFO:
1534 		case LINUX_SCHED_RR:
1535 			td->td_retval[0] = 1;
1536 			return (0);
1537 		default:
1538 			return (EINVAL);
1539 		}
1540 	}
1541 
1542 	switch (args->policy) {
1543 	case LINUX_SCHED_OTHER:
1544 		bsd.policy = SCHED_OTHER;
1545 		break;
1546 	case LINUX_SCHED_FIFO:
1547 		bsd.policy = SCHED_FIFO;
1548 		break;
1549 	case LINUX_SCHED_RR:
1550 		bsd.policy = SCHED_RR;
1551 		break;
1552 	default:
1553 		return (EINVAL);
1554 	}
1555 	return (sys_sched_get_priority_min(td, &bsd));
1556 }
1557 
1558 #define REBOOT_CAD_ON	0x89abcdef
1559 #define REBOOT_CAD_OFF	0
1560 #define REBOOT_HALT	0xcdef0123
1561 #define REBOOT_RESTART	0x01234567
1562 #define REBOOT_RESTART2	0xA1B2C3D4
1563 #define REBOOT_POWEROFF	0x4321FEDC
1564 #define REBOOT_MAGIC1	0xfee1dead
1565 #define REBOOT_MAGIC2	0x28121969
1566 #define REBOOT_MAGIC2A	0x05121996
1567 #define REBOOT_MAGIC2B	0x16041998
1568 
1569 int
1570 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1571 {
1572 	struct reboot_args bsd_args;
1573 
1574 	if (args->magic1 != REBOOT_MAGIC1)
1575 		return (EINVAL);
1576 
1577 	switch (args->magic2) {
1578 	case REBOOT_MAGIC2:
1579 	case REBOOT_MAGIC2A:
1580 	case REBOOT_MAGIC2B:
1581 		break;
1582 	default:
1583 		return (EINVAL);
1584 	}
1585 
1586 	switch (args->cmd) {
1587 	case REBOOT_CAD_ON:
1588 	case REBOOT_CAD_OFF:
1589 		return (priv_check(td, PRIV_REBOOT));
1590 	case REBOOT_HALT:
1591 		bsd_args.opt = RB_HALT;
1592 		break;
1593 	case REBOOT_RESTART:
1594 	case REBOOT_RESTART2:
1595 		bsd_args.opt = 0;
1596 		break;
1597 	case REBOOT_POWEROFF:
1598 		bsd_args.opt = RB_POWEROFF;
1599 		break;
1600 	default:
1601 		return (EINVAL);
1602 	}
1603 	return (sys_reboot(td, &bsd_args));
1604 }
1605 
1606 
1607 int
1608 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1609 {
1610 
1611 	td->td_retval[0] = td->td_proc->p_pid;
1612 
1613 	return (0);
1614 }
1615 
1616 int
1617 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1618 {
1619 	struct linux_emuldata *em;
1620 
1621 	em = em_find(td);
1622 	KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1623 
1624 	td->td_retval[0] = em->em_tid;
1625 
1626 	return (0);
1627 }
1628 
1629 
1630 int
1631 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1632 {
1633 
1634 	td->td_retval[0] = kern_getppid(td);
1635 	return (0);
1636 }
1637 
1638 int
1639 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1640 {
1641 
1642 	td->td_retval[0] = td->td_ucred->cr_rgid;
1643 	return (0);
1644 }
1645 
1646 int
1647 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1648 {
1649 
1650 	td->td_retval[0] = td->td_ucred->cr_ruid;
1651 	return (0);
1652 }
1653 
1654 int
1655 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1656 {
1657 
1658 	return (kern_getsid(td, args->pid));
1659 }
1660 
1661 int
1662 linux_nosys(struct thread *td, struct nosys_args *ignore)
1663 {
1664 
1665 	return (ENOSYS);
1666 }
1667 
1668 int
1669 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1670 {
1671 	int error;
1672 
1673 	error = kern_getpriority(td, args->which, args->who);
1674 	td->td_retval[0] = 20 - td->td_retval[0];
1675 	return (error);
1676 }
1677 
1678 int
1679 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1680 {
1681 	int name[2];
1682 
1683 	name[0] = CTL_KERN;
1684 	name[1] = KERN_HOSTNAME;
1685 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1686 	    args->len, 0, 0));
1687 }
1688 
1689 int
1690 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1691 {
1692 	int name[2];
1693 
1694 	name[0] = CTL_KERN;
1695 	name[1] = KERN_NISDOMAINNAME;
1696 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1697 	    args->len, 0, 0));
1698 }
1699 
1700 int
1701 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1702 {
1703 
1704 	LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1705 	    args->error_code);
1706 
1707 	/*
1708 	 * XXX: we should send a signal to the parent if
1709 	 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1710 	 * as it doesnt occur often.
1711 	 */
1712 	exit1(td, args->error_code, 0);
1713 		/* NOTREACHED */
1714 }
1715 
1716 #define _LINUX_CAPABILITY_VERSION_1  0x19980330
1717 #define _LINUX_CAPABILITY_VERSION_2  0x20071026
1718 #define _LINUX_CAPABILITY_VERSION_3  0x20080522
1719 
1720 struct l_user_cap_header {
1721 	l_int	version;
1722 	l_int	pid;
1723 };
1724 
1725 struct l_user_cap_data {
1726 	l_int	effective;
1727 	l_int	permitted;
1728 	l_int	inheritable;
1729 };
1730 
1731 int
1732 linux_capget(struct thread *td, struct linux_capget_args *uap)
1733 {
1734 	struct l_user_cap_header luch;
1735 	struct l_user_cap_data lucd[2];
1736 	int error, u32s;
1737 
1738 	if (uap->hdrp == NULL)
1739 		return (EFAULT);
1740 
1741 	error = copyin(uap->hdrp, &luch, sizeof(luch));
1742 	if (error != 0)
1743 		return (error);
1744 
1745 	switch (luch.version) {
1746 	case _LINUX_CAPABILITY_VERSION_1:
1747 		u32s = 1;
1748 		break;
1749 	case _LINUX_CAPABILITY_VERSION_2:
1750 	case _LINUX_CAPABILITY_VERSION_3:
1751 		u32s = 2;
1752 		break;
1753 	default:
1754 		luch.version = _LINUX_CAPABILITY_VERSION_1;
1755 		error = copyout(&luch, uap->hdrp, sizeof(luch));
1756 		if (error)
1757 			return (error);
1758 		return (EINVAL);
1759 	}
1760 
1761 	if (luch.pid)
1762 		return (EPERM);
1763 
1764 	if (uap->datap) {
1765 		/*
1766 		 * The current implementation doesn't support setting
1767 		 * a capability (it's essentially a stub) so indicate
1768 		 * that no capabilities are currently set or available
1769 		 * to request.
1770 		 */
1771 		memset(&lucd, 0, u32s * sizeof(lucd[0]));
1772 		error = copyout(&lucd, uap->datap, u32s * sizeof(lucd[0]));
1773 	}
1774 
1775 	return (error);
1776 }
1777 
1778 int
1779 linux_capset(struct thread *td, struct linux_capset_args *uap)
1780 {
1781 	struct l_user_cap_header luch;
1782 	struct l_user_cap_data lucd[2];
1783 	int error, i, u32s;
1784 
1785 	if (uap->hdrp == NULL || uap->datap == NULL)
1786 		return (EFAULT);
1787 
1788 	error = copyin(uap->hdrp, &luch, sizeof(luch));
1789 	if (error != 0)
1790 		return (error);
1791 
1792 	switch (luch.version) {
1793 	case _LINUX_CAPABILITY_VERSION_1:
1794 		u32s = 1;
1795 		break;
1796 	case _LINUX_CAPABILITY_VERSION_2:
1797 	case _LINUX_CAPABILITY_VERSION_3:
1798 		u32s = 2;
1799 		break;
1800 	default:
1801 		luch.version = _LINUX_CAPABILITY_VERSION_1;
1802 		error = copyout(&luch, uap->hdrp, sizeof(luch));
1803 		if (error)
1804 			return (error);
1805 		return (EINVAL);
1806 	}
1807 
1808 	if (luch.pid)
1809 		return (EPERM);
1810 
1811 	error = copyin(uap->datap, &lucd, u32s * sizeof(lucd[0]));
1812 	if (error != 0)
1813 		return (error);
1814 
1815 	/* We currently don't support setting any capabilities. */
1816 	for (i = 0; i < u32s; i++) {
1817 		if (lucd[i].effective || lucd[i].permitted ||
1818 		    lucd[i].inheritable) {
1819 			linux_msg(td,
1820 			    "capset[%d] effective=0x%x, permitted=0x%x, "
1821 			    "inheritable=0x%x is not implemented", i,
1822 			    (int)lucd[i].effective, (int)lucd[i].permitted,
1823 			    (int)lucd[i].inheritable);
1824 			return (EPERM);
1825 		}
1826 	}
1827 
1828 	return (0);
1829 }
1830 
1831 int
1832 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1833 {
1834 	int error = 0, max_size;
1835 	struct proc *p = td->td_proc;
1836 	char comm[LINUX_MAX_COMM_LEN];
1837 	int pdeath_signal;
1838 
1839 	switch (args->option) {
1840 	case LINUX_PR_SET_PDEATHSIG:
1841 		if (!LINUX_SIG_VALID(args->arg2))
1842 			return (EINVAL);
1843 		pdeath_signal = linux_to_bsd_signal(args->arg2);
1844 		return (kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_CTL,
1845 		    &pdeath_signal));
1846 	case LINUX_PR_GET_PDEATHSIG:
1847 		error = kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_STATUS,
1848 		    &pdeath_signal);
1849 		if (error != 0)
1850 			return (error);
1851 		pdeath_signal = bsd_to_linux_signal(pdeath_signal);
1852 		return (copyout(&pdeath_signal,
1853 		    (void *)(register_t)args->arg2,
1854 		    sizeof(pdeath_signal)));
1855 		break;
1856 	case LINUX_PR_GET_KEEPCAPS:
1857 		/*
1858 		 * Indicate that we always clear the effective and
1859 		 * permitted capability sets when the user id becomes
1860 		 * non-zero (actually the capability sets are simply
1861 		 * always zero in the current implementation).
1862 		 */
1863 		td->td_retval[0] = 0;
1864 		break;
1865 	case LINUX_PR_SET_KEEPCAPS:
1866 		/*
1867 		 * Ignore requests to keep the effective and permitted
1868 		 * capability sets when the user id becomes non-zero.
1869 		 */
1870 		break;
1871 	case LINUX_PR_SET_NAME:
1872 		/*
1873 		 * To be on the safe side we need to make sure to not
1874 		 * overflow the size a Linux program expects. We already
1875 		 * do this here in the copyin, so that we don't need to
1876 		 * check on copyout.
1877 		 */
1878 		max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1879 		error = copyinstr((void *)(register_t)args->arg2, comm,
1880 		    max_size, NULL);
1881 
1882 		/* Linux silently truncates the name if it is too long. */
1883 		if (error == ENAMETOOLONG) {
1884 			/*
1885 			 * XXX: copyinstr() isn't documented to populate the
1886 			 * array completely, so do a copyin() to be on the
1887 			 * safe side. This should be changed in case
1888 			 * copyinstr() is changed to guarantee this.
1889 			 */
1890 			error = copyin((void *)(register_t)args->arg2, comm,
1891 			    max_size - 1);
1892 			comm[max_size - 1] = '\0';
1893 		}
1894 		if (error)
1895 			return (error);
1896 
1897 		PROC_LOCK(p);
1898 		strlcpy(p->p_comm, comm, sizeof(p->p_comm));
1899 		PROC_UNLOCK(p);
1900 		break;
1901 	case LINUX_PR_GET_NAME:
1902 		PROC_LOCK(p);
1903 		strlcpy(comm, p->p_comm, sizeof(comm));
1904 		PROC_UNLOCK(p);
1905 		error = copyout(comm, (void *)(register_t)args->arg2,
1906 		    strlen(comm) + 1);
1907 		break;
1908 	default:
1909 		error = EINVAL;
1910 		break;
1911 	}
1912 
1913 	return (error);
1914 }
1915 
1916 int
1917 linux_sched_setparam(struct thread *td,
1918     struct linux_sched_setparam_args *uap)
1919 {
1920 	struct sched_param sched_param;
1921 	struct thread *tdt;
1922 	int error, policy;
1923 
1924 	error = copyin(uap->param, &sched_param, sizeof(sched_param));
1925 	if (error)
1926 		return (error);
1927 
1928 	tdt = linux_tdfind(td, uap->pid, -1);
1929 	if (tdt == NULL)
1930 		return (ESRCH);
1931 
1932 	if (linux_map_sched_prio) {
1933 		error = kern_sched_getscheduler(td, tdt, &policy);
1934 		if (error)
1935 			goto out;
1936 
1937 		switch (policy) {
1938 		case SCHED_OTHER:
1939 			if (sched_param.sched_priority != 0) {
1940 				error = EINVAL;
1941 				goto out;
1942 			}
1943 			sched_param.sched_priority =
1944 			    PRI_MAX_TIMESHARE - PRI_MIN_TIMESHARE;
1945 			break;
1946 		case SCHED_FIFO:
1947 		case SCHED_RR:
1948 			if (sched_param.sched_priority < 1 ||
1949 			    sched_param.sched_priority >= LINUX_MAX_RT_PRIO) {
1950 				error = EINVAL;
1951 				goto out;
1952 			}
1953 			/*
1954 			 * Map [1, LINUX_MAX_RT_PRIO - 1] to
1955 			 * [0, RTP_PRIO_MAX - RTP_PRIO_MIN] (rounding down).
1956 			 */
1957 			sched_param.sched_priority =
1958 			    (sched_param.sched_priority - 1) *
1959 			    (RTP_PRIO_MAX - RTP_PRIO_MIN + 1) /
1960 			    (LINUX_MAX_RT_PRIO - 1);
1961 			break;
1962 		}
1963 	}
1964 
1965 	error = kern_sched_setparam(td, tdt, &sched_param);
1966 out:	PROC_UNLOCK(tdt->td_proc);
1967 	return (error);
1968 }
1969 
1970 int
1971 linux_sched_getparam(struct thread *td,
1972     struct linux_sched_getparam_args *uap)
1973 {
1974 	struct sched_param sched_param;
1975 	struct thread *tdt;
1976 	int error, policy;
1977 
1978 	tdt = linux_tdfind(td, uap->pid, -1);
1979 	if (tdt == NULL)
1980 		return (ESRCH);
1981 
1982 	error = kern_sched_getparam(td, tdt, &sched_param);
1983 	if (error) {
1984 		PROC_UNLOCK(tdt->td_proc);
1985 		return (error);
1986 	}
1987 
1988 	if (linux_map_sched_prio) {
1989 		error = kern_sched_getscheduler(td, tdt, &policy);
1990 		PROC_UNLOCK(tdt->td_proc);
1991 		if (error)
1992 			return (error);
1993 
1994 		switch (policy) {
1995 		case SCHED_OTHER:
1996 			sched_param.sched_priority = 0;
1997 			break;
1998 		case SCHED_FIFO:
1999 		case SCHED_RR:
2000 			/*
2001 			 * Map [0, RTP_PRIO_MAX - RTP_PRIO_MIN] to
2002 			 * [1, LINUX_MAX_RT_PRIO - 1] (rounding up).
2003 			 */
2004 			sched_param.sched_priority =
2005 			    (sched_param.sched_priority *
2006 			    (LINUX_MAX_RT_PRIO - 1) +
2007 			    (RTP_PRIO_MAX - RTP_PRIO_MIN - 1)) /
2008 			    (RTP_PRIO_MAX - RTP_PRIO_MIN) + 1;
2009 			break;
2010 		}
2011 	} else
2012 		PROC_UNLOCK(tdt->td_proc);
2013 
2014 	error = copyout(&sched_param, uap->param, sizeof(sched_param));
2015 	return (error);
2016 }
2017 
2018 /*
2019  * Get affinity of a process.
2020  */
2021 int
2022 linux_sched_getaffinity(struct thread *td,
2023     struct linux_sched_getaffinity_args *args)
2024 {
2025 	int error;
2026 	struct thread *tdt;
2027 
2028 	if (args->len < sizeof(cpuset_t))
2029 		return (EINVAL);
2030 
2031 	tdt = linux_tdfind(td, args->pid, -1);
2032 	if (tdt == NULL)
2033 		return (ESRCH);
2034 
2035 	PROC_UNLOCK(tdt->td_proc);
2036 
2037 	error = kern_cpuset_getaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2038 	    tdt->td_tid, sizeof(cpuset_t), (cpuset_t *)args->user_mask_ptr);
2039 	if (error == 0)
2040 		td->td_retval[0] = sizeof(cpuset_t);
2041 
2042 	return (error);
2043 }
2044 
2045 /*
2046  *  Set affinity of a process.
2047  */
2048 int
2049 linux_sched_setaffinity(struct thread *td,
2050     struct linux_sched_setaffinity_args *args)
2051 {
2052 	struct thread *tdt;
2053 
2054 	if (args->len < sizeof(cpuset_t))
2055 		return (EINVAL);
2056 
2057 	tdt = linux_tdfind(td, args->pid, -1);
2058 	if (tdt == NULL)
2059 		return (ESRCH);
2060 
2061 	PROC_UNLOCK(tdt->td_proc);
2062 
2063 	return (kern_cpuset_setaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2064 	    tdt->td_tid, sizeof(cpuset_t), (cpuset_t *) args->user_mask_ptr));
2065 }
2066 
2067 struct linux_rlimit64 {
2068 	uint64_t	rlim_cur;
2069 	uint64_t	rlim_max;
2070 };
2071 
2072 int
2073 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
2074 {
2075 	struct rlimit rlim, nrlim;
2076 	struct linux_rlimit64 lrlim;
2077 	struct proc *p;
2078 	u_int which;
2079 	int flags;
2080 	int error;
2081 
2082 	if (args->resource >= LINUX_RLIM_NLIMITS)
2083 		return (EINVAL);
2084 
2085 	which = linux_to_bsd_resource[args->resource];
2086 	if (which == -1)
2087 		return (EINVAL);
2088 
2089 	if (args->new != NULL) {
2090 		/*
2091 		 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
2092 		 * rlim is unsigned 64-bit. FreeBSD treats negative limits
2093 		 * as INFINITY so we do not need a conversion even.
2094 		 */
2095 		error = copyin(args->new, &nrlim, sizeof(nrlim));
2096 		if (error != 0)
2097 			return (error);
2098 	}
2099 
2100 	flags = PGET_HOLD | PGET_NOTWEXIT;
2101 	if (args->new != NULL)
2102 		flags |= PGET_CANDEBUG;
2103 	else
2104 		flags |= PGET_CANSEE;
2105 	if (args->pid == 0) {
2106 		p = td->td_proc;
2107 		PHOLD(p);
2108 	} else {
2109 		error = pget(args->pid, flags, &p);
2110 		if (error != 0)
2111 			return (error);
2112 	}
2113 	if (args->old != NULL) {
2114 		PROC_LOCK(p);
2115 		lim_rlimit_proc(p, which, &rlim);
2116 		PROC_UNLOCK(p);
2117 		if (rlim.rlim_cur == RLIM_INFINITY)
2118 			lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2119 		else
2120 			lrlim.rlim_cur = rlim.rlim_cur;
2121 		if (rlim.rlim_max == RLIM_INFINITY)
2122 			lrlim.rlim_max = LINUX_RLIM_INFINITY;
2123 		else
2124 			lrlim.rlim_max = rlim.rlim_max;
2125 		error = copyout(&lrlim, args->old, sizeof(lrlim));
2126 		if (error != 0)
2127 			goto out;
2128 	}
2129 
2130 	if (args->new != NULL)
2131 		error = kern_proc_setrlimit(td, p, which, &nrlim);
2132 
2133  out:
2134 	PRELE(p);
2135 	return (error);
2136 }
2137 
2138 int
2139 linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2140 {
2141 	struct timeval utv, tv0, tv1, *tvp;
2142 	struct l_pselect6arg lpse6;
2143 	struct l_timespec lts;
2144 	struct timespec uts;
2145 	l_sigset_t l_ss;
2146 	sigset_t *ssp;
2147 	sigset_t ss;
2148 	int error;
2149 
2150 	ssp = NULL;
2151 	if (args->sig != NULL) {
2152 		error = copyin(args->sig, &lpse6, sizeof(lpse6));
2153 		if (error != 0)
2154 			return (error);
2155 		if (lpse6.ss_len != sizeof(l_ss))
2156 			return (EINVAL);
2157 		if (lpse6.ss != 0) {
2158 			error = copyin(PTRIN(lpse6.ss), &l_ss,
2159 			    sizeof(l_ss));
2160 			if (error != 0)
2161 				return (error);
2162 			linux_to_bsd_sigset(&l_ss, &ss);
2163 			ssp = &ss;
2164 		}
2165 	}
2166 
2167 	/*
2168 	 * Currently glibc changes nanosecond number to microsecond.
2169 	 * This mean losing precision but for now it is hardly seen.
2170 	 */
2171 	if (args->tsp != NULL) {
2172 		error = copyin(args->tsp, &lts, sizeof(lts));
2173 		if (error != 0)
2174 			return (error);
2175 		error = linux_to_native_timespec(&uts, &lts);
2176 		if (error != 0)
2177 			return (error);
2178 
2179 		TIMESPEC_TO_TIMEVAL(&utv, &uts);
2180 		if (itimerfix(&utv))
2181 			return (EINVAL);
2182 
2183 		microtime(&tv0);
2184 		tvp = &utv;
2185 	} else
2186 		tvp = NULL;
2187 
2188 	error = kern_pselect(td, args->nfds, args->readfds, args->writefds,
2189 	    args->exceptfds, tvp, ssp, LINUX_NFDBITS);
2190 
2191 	if (error == 0 && args->tsp != NULL) {
2192 		if (td->td_retval[0] != 0) {
2193 			/*
2194 			 * Compute how much time was left of the timeout,
2195 			 * by subtracting the current time and the time
2196 			 * before we started the call, and subtracting
2197 			 * that result from the user-supplied value.
2198 			 */
2199 
2200 			microtime(&tv1);
2201 			timevalsub(&tv1, &tv0);
2202 			timevalsub(&utv, &tv1);
2203 			if (utv.tv_sec < 0)
2204 				timevalclear(&utv);
2205 		} else
2206 			timevalclear(&utv);
2207 
2208 		TIMEVAL_TO_TIMESPEC(&utv, &uts);
2209 
2210 		error = native_to_linux_timespec(&lts, &uts);
2211 		if (error == 0)
2212 			error = copyout(&lts, args->tsp, sizeof(lts));
2213 	}
2214 
2215 	return (error);
2216 }
2217 
2218 int
2219 linux_ppoll(struct thread *td, struct linux_ppoll_args *args)
2220 {
2221 	struct timespec ts0, ts1;
2222 	struct l_timespec lts;
2223 	struct timespec uts, *tsp;
2224 	l_sigset_t l_ss;
2225 	sigset_t *ssp;
2226 	sigset_t ss;
2227 	int error;
2228 
2229 	if (args->sset != NULL) {
2230 		if (args->ssize != sizeof(l_ss))
2231 			return (EINVAL);
2232 		error = copyin(args->sset, &l_ss, sizeof(l_ss));
2233 		if (error)
2234 			return (error);
2235 		linux_to_bsd_sigset(&l_ss, &ss);
2236 		ssp = &ss;
2237 	} else
2238 		ssp = NULL;
2239 	if (args->tsp != NULL) {
2240 		error = copyin(args->tsp, &lts, sizeof(lts));
2241 		if (error)
2242 			return (error);
2243 		error = linux_to_native_timespec(&uts, &lts);
2244 		if (error != 0)
2245 			return (error);
2246 
2247 		nanotime(&ts0);
2248 		tsp = &uts;
2249 	} else
2250 		tsp = NULL;
2251 
2252 	error = kern_poll(td, args->fds, args->nfds, tsp, ssp);
2253 
2254 	if (error == 0 && args->tsp != NULL) {
2255 		if (td->td_retval[0]) {
2256 			nanotime(&ts1);
2257 			timespecsub(&ts1, &ts0, &ts1);
2258 			timespecsub(&uts, &ts1, &uts);
2259 			if (uts.tv_sec < 0)
2260 				timespecclear(&uts);
2261 		} else
2262 			timespecclear(&uts);
2263 
2264 		error = native_to_linux_timespec(&lts, &uts);
2265 		if (error == 0)
2266 			error = copyout(&lts, args->tsp, sizeof(lts));
2267 	}
2268 
2269 	return (error);
2270 }
2271 
2272 int
2273 linux_sched_rr_get_interval(struct thread *td,
2274     struct linux_sched_rr_get_interval_args *uap)
2275 {
2276 	struct timespec ts;
2277 	struct l_timespec lts;
2278 	struct thread *tdt;
2279 	int error;
2280 
2281 	/*
2282 	 * According to man in case the invalid pid specified
2283 	 * EINVAL should be returned.
2284 	 */
2285 	if (uap->pid < 0)
2286 		return (EINVAL);
2287 
2288 	tdt = linux_tdfind(td, uap->pid, -1);
2289 	if (tdt == NULL)
2290 		return (ESRCH);
2291 
2292 	error = kern_sched_rr_get_interval_td(td, tdt, &ts);
2293 	PROC_UNLOCK(tdt->td_proc);
2294 	if (error != 0)
2295 		return (error);
2296 	error = native_to_linux_timespec(&lts, &ts);
2297 	if (error != 0)
2298 		return (error);
2299 	return (copyout(&lts, uap->interval, sizeof(lts)));
2300 }
2301 
2302 /*
2303  * In case when the Linux thread is the initial thread in
2304  * the thread group thread id is equal to the process id.
2305  * Glibc depends on this magic (assert in pthread_getattr_np.c).
2306  */
2307 struct thread *
2308 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid)
2309 {
2310 	struct linux_emuldata *em;
2311 	struct thread *tdt;
2312 	struct proc *p;
2313 
2314 	tdt = NULL;
2315 	if (tid == 0 || tid == td->td_tid) {
2316 		tdt = td;
2317 		PROC_LOCK(tdt->td_proc);
2318 	} else if (tid > PID_MAX)
2319 		tdt = tdfind(tid, pid);
2320 	else {
2321 		/*
2322 		 * Initial thread where the tid equal to the pid.
2323 		 */
2324 		p = pfind(tid);
2325 		if (p != NULL) {
2326 			if (SV_PROC_ABI(p) != SV_ABI_LINUX) {
2327 				/*
2328 				 * p is not a Linuxulator process.
2329 				 */
2330 				PROC_UNLOCK(p);
2331 				return (NULL);
2332 			}
2333 			FOREACH_THREAD_IN_PROC(p, tdt) {
2334 				em = em_find(tdt);
2335 				if (tid == em->em_tid)
2336 					return (tdt);
2337 			}
2338 			PROC_UNLOCK(p);
2339 		}
2340 		return (NULL);
2341 	}
2342 
2343 	return (tdt);
2344 }
2345 
2346 void
2347 linux_to_bsd_waitopts(int options, int *bsdopts)
2348 {
2349 
2350 	if (options & LINUX_WNOHANG)
2351 		*bsdopts |= WNOHANG;
2352 	if (options & LINUX_WUNTRACED)
2353 		*bsdopts |= WUNTRACED;
2354 	if (options & LINUX_WEXITED)
2355 		*bsdopts |= WEXITED;
2356 	if (options & LINUX_WCONTINUED)
2357 		*bsdopts |= WCONTINUED;
2358 	if (options & LINUX_WNOWAIT)
2359 		*bsdopts |= WNOWAIT;
2360 
2361 	if (options & __WCLONE)
2362 		*bsdopts |= WLINUXCLONE;
2363 }
2364 
2365 int
2366 linux_getrandom(struct thread *td, struct linux_getrandom_args *args)
2367 {
2368 	struct uio uio;
2369 	struct iovec iov;
2370 	int error;
2371 
2372 	if (args->flags & ~(LINUX_GRND_NONBLOCK|LINUX_GRND_RANDOM))
2373 		return (EINVAL);
2374 	if (args->count > INT_MAX)
2375 		args->count = INT_MAX;
2376 
2377 	iov.iov_base = args->buf;
2378 	iov.iov_len = args->count;
2379 
2380 	uio.uio_iov = &iov;
2381 	uio.uio_iovcnt = 1;
2382 	uio.uio_resid = iov.iov_len;
2383 	uio.uio_segflg = UIO_USERSPACE;
2384 	uio.uio_rw = UIO_READ;
2385 	uio.uio_td = td;
2386 
2387 	error = read_random_uio(&uio, args->flags & LINUX_GRND_NONBLOCK);
2388 	if (error == 0)
2389 		td->td_retval[0] = args->count - uio.uio_resid;
2390 	return (error);
2391 }
2392 
2393 int
2394 linux_mincore(struct thread *td, struct linux_mincore_args *args)
2395 {
2396 
2397 	/* Needs to be page-aligned */
2398 	if (args->start & PAGE_MASK)
2399 		return (EINVAL);
2400 	return (kern_mincore(td, args->start, args->len, args->vec));
2401 }
2402 
2403 #define	SYSLOG_TAG	"<6>"
2404 
2405 int
2406 linux_syslog(struct thread *td, struct linux_syslog_args *args)
2407 {
2408 	char buf[128], *src, *dst;
2409 	u_int seq;
2410 	int buflen, error;
2411 
2412 	if (args->type != LINUX_SYSLOG_ACTION_READ_ALL) {
2413 		linux_msg(td, "syslog unsupported type 0x%x", args->type);
2414 		return (EINVAL);
2415 	}
2416 
2417 	if (args->len < 6) {
2418 		td->td_retval[0] = 0;
2419 		return (0);
2420 	}
2421 
2422 	error = priv_check(td, PRIV_MSGBUF);
2423 	if (error)
2424 		return (error);
2425 
2426 	mtx_lock(&msgbuf_lock);
2427 	msgbuf_peekbytes(msgbufp, NULL, 0, &seq);
2428 	mtx_unlock(&msgbuf_lock);
2429 
2430 	dst = args->buf;
2431 	error = copyout(&SYSLOG_TAG, dst, sizeof(SYSLOG_TAG));
2432 	/* The -1 is to skip the trailing '\0'. */
2433 	dst += sizeof(SYSLOG_TAG) - 1;
2434 
2435 	while (error == 0) {
2436 		mtx_lock(&msgbuf_lock);
2437 		buflen = msgbuf_peekbytes(msgbufp, buf, sizeof(buf), &seq);
2438 		mtx_unlock(&msgbuf_lock);
2439 
2440 		if (buflen == 0)
2441 			break;
2442 
2443 		for (src = buf; src < buf + buflen && error == 0; src++) {
2444 			if (*src == '\0')
2445 				continue;
2446 
2447 			if (dst >= args->buf + args->len)
2448 				goto out;
2449 
2450 			error = copyout(src, dst, 1);
2451 			dst++;
2452 
2453 			if (*src == '\n' && *(src + 1) != '<' &&
2454 			    dst + sizeof(SYSLOG_TAG) < args->buf + args->len) {
2455 				error = copyout(&SYSLOG_TAG,
2456 				    dst, sizeof(SYSLOG_TAG));
2457 				dst += sizeof(SYSLOG_TAG) - 1;
2458 			}
2459 		}
2460 	}
2461 out:
2462 	td->td_retval[0] = dst - args->buf;
2463 	return (error);
2464 }
2465 
2466 int
2467 linux_getcpu(struct thread *td, struct linux_getcpu_args *args)
2468 {
2469 	int cpu, error, node;
2470 
2471 	cpu = td->td_oncpu; /* Make sure it doesn't change during copyout(9) */
2472 	error = 0;
2473 	node = cpuid_to_pcpu[cpu]->pc_domain;
2474 
2475 	if (args->cpu != NULL)
2476 		error = copyout(&cpu, args->cpu, sizeof(l_int));
2477 	if (args->node != NULL)
2478 		error = copyout(&node, args->node, sizeof(l_int));
2479 	return (error);
2480 }
2481