xref: /freebsd/sys/compat/linux/linux_misc.c (revision 8ef24a0d4b28fe230e20637f56869cc4148cd2ca)
1 /*-
2  * Copyright (c) 2002 Doug Rabson
3  * Copyright (c) 1994-1995 Søren Schmidt
4  * All rights reserved.
5  *
6  * Redistribution and use in source and binary forms, with or without
7  * modification, are permitted provided that the following conditions
8  * are met:
9  * 1. Redistributions of source code must retain the above copyright
10  *    notice, this list of conditions and the following disclaimer
11  *    in this position and unchanged.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. The name of the author may not be used to endorse or promote products
16  *    derived from this software without specific prior written permission
17  *
18  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
19  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
20  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
21  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
22  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
23  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
24  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
25  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
26  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
27  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
28  */
29 
30 #include <sys/cdefs.h>
31 __FBSDID("$FreeBSD$");
32 
33 #include "opt_compat.h"
34 
35 #include <sys/param.h>
36 #include <sys/blist.h>
37 #include <sys/fcntl.h>
38 #if defined(__i386__)
39 #include <sys/imgact_aout.h>
40 #endif
41 #include <sys/jail.h>
42 #include <sys/kernel.h>
43 #include <sys/limits.h>
44 #include <sys/lock.h>
45 #include <sys/malloc.h>
46 #include <sys/mman.h>
47 #include <sys/mount.h>
48 #include <sys/mutex.h>
49 #include <sys/namei.h>
50 #include <sys/priv.h>
51 #include <sys/proc.h>
52 #include <sys/reboot.h>
53 #include <sys/racct.h>
54 #include <sys/resourcevar.h>
55 #include <sys/sched.h>
56 #include <sys/sdt.h>
57 #include <sys/signalvar.h>
58 #include <sys/stat.h>
59 #include <sys/syscallsubr.h>
60 #include <sys/sysctl.h>
61 #include <sys/sysproto.h>
62 #include <sys/systm.h>
63 #include <sys/time.h>
64 #include <sys/vmmeter.h>
65 #include <sys/vnode.h>
66 #include <sys/wait.h>
67 #include <sys/cpuset.h>
68 
69 #include <security/mac/mac_framework.h>
70 
71 #include <vm/vm.h>
72 #include <vm/pmap.h>
73 #include <vm/vm_kern.h>
74 #include <vm/vm_map.h>
75 #include <vm/vm_extern.h>
76 #include <vm/vm_object.h>
77 #include <vm/swap_pager.h>
78 
79 #ifdef COMPAT_LINUX32
80 #include <machine/../linux32/linux.h>
81 #include <machine/../linux32/linux32_proto.h>
82 #else
83 #include <machine/../linux/linux.h>
84 #include <machine/../linux/linux_proto.h>
85 #endif
86 
87 #include <compat/linux/linux_dtrace.h>
88 #include <compat/linux/linux_file.h>
89 #include <compat/linux/linux_mib.h>
90 #include <compat/linux/linux_signal.h>
91 #include <compat/linux/linux_timer.h>
92 #include <compat/linux/linux_util.h>
93 #include <compat/linux/linux_sysproto.h>
94 #include <compat/linux/linux_emul.h>
95 #include <compat/linux/linux_misc.h>
96 
97 /**
98  * Special DTrace provider for the linuxulator.
99  *
100  * In this file we define the provider for the entire linuxulator. All
101  * modules (= files of the linuxulator) use it.
102  *
103  * We define a different name depending on the emulated bitsize, see
104  * ../../<ARCH>/linux{,32}/linux.h, e.g.:
105  *      native bitsize          = linuxulator
106  *      amd64, 32bit emulation  = linuxulator32
107  */
108 LIN_SDT_PROVIDER_DEFINE(LINUX_DTRACE);
109 
110 int stclohz;				/* Statistics clock frequency */
111 
112 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
113 	RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
114 	RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
115 	RLIMIT_MEMLOCK, RLIMIT_AS
116 };
117 
118 struct l_sysinfo {
119 	l_long		uptime;		/* Seconds since boot */
120 	l_ulong		loads[3];	/* 1, 5, and 15 minute load averages */
121 #define LINUX_SYSINFO_LOADS_SCALE 65536
122 	l_ulong		totalram;	/* Total usable main memory size */
123 	l_ulong		freeram;	/* Available memory size */
124 	l_ulong		sharedram;	/* Amount of shared memory */
125 	l_ulong		bufferram;	/* Memory used by buffers */
126 	l_ulong		totalswap;	/* Total swap space size */
127 	l_ulong		freeswap;	/* swap space still available */
128 	l_ushort	procs;		/* Number of current processes */
129 	l_ushort	pads;
130 	l_ulong		totalbig;
131 	l_ulong		freebig;
132 	l_uint		mem_unit;
133 	char		_f[20-2*sizeof(l_long)-sizeof(l_int)];	/* padding */
134 };
135 
136 struct l_pselect6arg {
137 	l_uintptr_t	ss;
138 	l_size_t	ss_len;
139 };
140 
141 static int	linux_utimensat_nsec_valid(l_long);
142 
143 
144 int
145 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
146 {
147 	struct l_sysinfo sysinfo;
148 	vm_object_t object;
149 	int i, j;
150 	struct timespec ts;
151 
152 	getnanouptime(&ts);
153 	if (ts.tv_nsec != 0)
154 		ts.tv_sec++;
155 	sysinfo.uptime = ts.tv_sec;
156 
157 	/* Use the information from the mib to get our load averages */
158 	for (i = 0; i < 3; i++)
159 		sysinfo.loads[i] = averunnable.ldavg[i] *
160 		    LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
161 
162 	sysinfo.totalram = physmem * PAGE_SIZE;
163 	sysinfo.freeram = sysinfo.totalram - vm_cnt.v_wire_count * PAGE_SIZE;
164 
165 	sysinfo.sharedram = 0;
166 	mtx_lock(&vm_object_list_mtx);
167 	TAILQ_FOREACH(object, &vm_object_list, object_list)
168 		if (object->shadow_count > 1)
169 			sysinfo.sharedram += object->resident_page_count;
170 	mtx_unlock(&vm_object_list_mtx);
171 
172 	sysinfo.sharedram *= PAGE_SIZE;
173 	sysinfo.bufferram = 0;
174 
175 	swap_pager_status(&i, &j);
176 	sysinfo.totalswap = i * PAGE_SIZE;
177 	sysinfo.freeswap = (i - j) * PAGE_SIZE;
178 
179 	sysinfo.procs = nprocs;
180 
181 	/* The following are only present in newer Linux kernels. */
182 	sysinfo.totalbig = 0;
183 	sysinfo.freebig = 0;
184 	sysinfo.mem_unit = 1;
185 
186 	return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
187 }
188 
189 int
190 linux_alarm(struct thread *td, struct linux_alarm_args *args)
191 {
192 	struct itimerval it, old_it;
193 	u_int secs;
194 	int error;
195 
196 #ifdef DEBUG
197 	if (ldebug(alarm))
198 		printf(ARGS(alarm, "%u"), args->secs);
199 #endif
200 	secs = args->secs;
201 	/*
202 	 * Linux alarm() is always successful. Limit secs to INT32_MAX / 2
203 	 * to match kern_setitimer()'s limit to avoid error from it.
204 	 *
205 	 * XXX. Linux limit secs to INT_MAX on 32 and does not limit on 64-bit
206 	 * platforms.
207 	 */
208 	if (secs > INT32_MAX / 2)
209 		secs = INT32_MAX / 2;
210 
211 	it.it_value.tv_sec = secs;
212 	it.it_value.tv_usec = 0;
213 	timevalclear(&it.it_interval);
214 	error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
215 	KASSERT(error == 0, ("kern_setitimer returns %d", error));
216 
217 	if ((old_it.it_value.tv_sec == 0 && old_it.it_value.tv_usec > 0) ||
218 	    old_it.it_value.tv_usec >= 500000)
219 		old_it.it_value.tv_sec++;
220 	td->td_retval[0] = old_it.it_value.tv_sec;
221 	return (0);
222 }
223 
224 int
225 linux_brk(struct thread *td, struct linux_brk_args *args)
226 {
227 	struct vmspace *vm = td->td_proc->p_vmspace;
228 	vm_offset_t new, old;
229 	struct obreak_args /* {
230 		char * nsize;
231 	} */ tmp;
232 
233 #ifdef DEBUG
234 	if (ldebug(brk))
235 		printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
236 #endif
237 	old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize);
238 	new = (vm_offset_t)args->dsend;
239 	tmp.nsize = (char *)new;
240 	if (((caddr_t)new > vm->vm_daddr) && !sys_obreak(td, &tmp))
241 		td->td_retval[0] = (long)new;
242 	else
243 		td->td_retval[0] = (long)old;
244 
245 	return (0);
246 }
247 
248 #if defined(__i386__)
249 /* XXX: what about amd64/linux32? */
250 
251 int
252 linux_uselib(struct thread *td, struct linux_uselib_args *args)
253 {
254 	struct nameidata ni;
255 	struct vnode *vp;
256 	struct exec *a_out;
257 	struct vattr attr;
258 	vm_offset_t vmaddr;
259 	unsigned long file_offset;
260 	unsigned long bss_size;
261 	char *library;
262 	ssize_t aresid;
263 	int error, locked, writecount;
264 
265 	LCONVPATHEXIST(td, args->library, &library);
266 
267 #ifdef DEBUG
268 	if (ldebug(uselib))
269 		printf(ARGS(uselib, "%s"), library);
270 #endif
271 
272 	a_out = NULL;
273 	locked = 0;
274 	vp = NULL;
275 
276 	NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
277 	    UIO_SYSSPACE, library, td);
278 	error = namei(&ni);
279 	LFREEPATH(library);
280 	if (error)
281 		goto cleanup;
282 
283 	vp = ni.ni_vp;
284 	NDFREE(&ni, NDF_ONLY_PNBUF);
285 
286 	/*
287 	 * From here on down, we have a locked vnode that must be unlocked.
288 	 * XXX: The code below largely duplicates exec_check_permissions().
289 	 */
290 	locked = 1;
291 
292 	/* Writable? */
293 	error = VOP_GET_WRITECOUNT(vp, &writecount);
294 	if (error != 0)
295 		goto cleanup;
296 	if (writecount != 0) {
297 		error = ETXTBSY;
298 		goto cleanup;
299 	}
300 
301 	/* Executable? */
302 	error = VOP_GETATTR(vp, &attr, td->td_ucred);
303 	if (error)
304 		goto cleanup;
305 
306 	if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
307 	    ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
308 		/* EACCESS is what exec(2) returns. */
309 		error = ENOEXEC;
310 		goto cleanup;
311 	}
312 
313 	/* Sensible size? */
314 	if (attr.va_size == 0) {
315 		error = ENOEXEC;
316 		goto cleanup;
317 	}
318 
319 	/* Can we access it? */
320 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
321 	if (error)
322 		goto cleanup;
323 
324 	/*
325 	 * XXX: This should use vn_open() so that it is properly authorized,
326 	 * and to reduce code redundancy all over the place here.
327 	 * XXX: Not really, it duplicates far more of exec_check_permissions()
328 	 * than vn_open().
329 	 */
330 #ifdef MAC
331 	error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
332 	if (error)
333 		goto cleanup;
334 #endif
335 	error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
336 	if (error)
337 		goto cleanup;
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 	 * XXX: Note that if any of the VM operations fail below we don't
401 	 * clear this flag.
402 	 */
403 	VOP_SET_TEXT(vp);
404 
405 	/*
406 	 * Lock no longer needed
407 	 */
408 	locked = 0;
409 	VOP_UNLOCK(vp, 0);
410 
411 	/*
412 	 * Check if file_offset page aligned. Currently we cannot handle
413 	 * misalinged file offsets, and so we read in the entire image
414 	 * (what a waste).
415 	 */
416 	if (file_offset & PAGE_MASK) {
417 #ifdef DEBUG
418 		printf("uselib: Non page aligned binary %lu\n", file_offset);
419 #endif
420 		/* Map text+data read/write/execute */
421 
422 		/* a_entry is the load address and is page aligned */
423 		vmaddr = trunc_page(a_out->a_entry);
424 
425 		/* get anon user mapping, read+write+execute */
426 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
427 		    &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
428 		    VM_PROT_ALL, VM_PROT_ALL, 0);
429 		if (error)
430 			goto cleanup;
431 
432 		error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
433 		    a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
434 		    td->td_ucred, NOCRED, &aresid, td);
435 		if (error != 0)
436 			goto cleanup;
437 		if (aresid != 0) {
438 			error = ENOEXEC;
439 			goto cleanup;
440 		}
441 	} else {
442 #ifdef DEBUG
443 		printf("uselib: Page aligned binary %lu\n", file_offset);
444 #endif
445 		/*
446 		 * for QMAGIC, a_entry is 20 bytes beyond the load address
447 		 * to skip the executable header
448 		 */
449 		vmaddr = trunc_page(a_out->a_entry);
450 
451 		/*
452 		 * Map it all into the process's space as a single
453 		 * copy-on-write "data" segment.
454 		 */
455 		error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr,
456 		    a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
457 		    MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
458 		if (error)
459 			goto cleanup;
460 	}
461 #ifdef DEBUG
462 	printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
463 	    ((long *)vmaddr)[1]);
464 #endif
465 	if (bss_size != 0) {
466 		/* Calculate BSS start address */
467 		vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
468 		    a_out->a_data;
469 
470 		/* allocate some 'anon' space */
471 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
472 		    &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
473 		    VM_PROT_ALL, 0);
474 		if (error)
475 			goto cleanup;
476 	}
477 
478 cleanup:
479 	/* Unlock vnode if needed */
480 	if (locked)
481 		VOP_UNLOCK(vp, 0);
482 
483 	/* Release the temporary mapping. */
484 	if (a_out)
485 		kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
486 
487 	return (error);
488 }
489 
490 #endif	/* __i386__ */
491 
492 int
493 linux_select(struct thread *td, struct linux_select_args *args)
494 {
495 	l_timeval ltv;
496 	struct timeval tv0, tv1, utv, *tvp;
497 	int error;
498 
499 #ifdef DEBUG
500 	if (ldebug(select))
501 		printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
502 		    (void *)args->readfds, (void *)args->writefds,
503 		    (void *)args->exceptfds, (void *)args->timeout);
504 #endif
505 
506 	/*
507 	 * Store current time for computation of the amount of
508 	 * time left.
509 	 */
510 	if (args->timeout) {
511 		if ((error = copyin(args->timeout, &ltv, sizeof(ltv))))
512 			goto select_out;
513 		utv.tv_sec = ltv.tv_sec;
514 		utv.tv_usec = ltv.tv_usec;
515 #ifdef DEBUG
516 		if (ldebug(select))
517 			printf(LMSG("incoming timeout (%jd/%ld)"),
518 			    (intmax_t)utv.tv_sec, utv.tv_usec);
519 #endif
520 
521 		if (itimerfix(&utv)) {
522 			/*
523 			 * The timeval was invalid.  Convert it to something
524 			 * valid that will act as it does under Linux.
525 			 */
526 			utv.tv_sec += utv.tv_usec / 1000000;
527 			utv.tv_usec %= 1000000;
528 			if (utv.tv_usec < 0) {
529 				utv.tv_sec -= 1;
530 				utv.tv_usec += 1000000;
531 			}
532 			if (utv.tv_sec < 0)
533 				timevalclear(&utv);
534 		}
535 		microtime(&tv0);
536 		tvp = &utv;
537 	} else
538 		tvp = NULL;
539 
540 	error = kern_select(td, args->nfds, args->readfds, args->writefds,
541 	    args->exceptfds, tvp, LINUX_NFDBITS);
542 
543 #ifdef DEBUG
544 	if (ldebug(select))
545 		printf(LMSG("real select returns %d"), error);
546 #endif
547 	if (error)
548 		goto select_out;
549 
550 	if (args->timeout) {
551 		if (td->td_retval[0]) {
552 			/*
553 			 * Compute how much time was left of the timeout,
554 			 * by subtracting the current time and the time
555 			 * before we started the call, and subtracting
556 			 * that result from the user-supplied value.
557 			 */
558 			microtime(&tv1);
559 			timevalsub(&tv1, &tv0);
560 			timevalsub(&utv, &tv1);
561 			if (utv.tv_sec < 0)
562 				timevalclear(&utv);
563 		} else
564 			timevalclear(&utv);
565 #ifdef DEBUG
566 		if (ldebug(select))
567 			printf(LMSG("outgoing timeout (%jd/%ld)"),
568 			    (intmax_t)utv.tv_sec, utv.tv_usec);
569 #endif
570 		ltv.tv_sec = utv.tv_sec;
571 		ltv.tv_usec = utv.tv_usec;
572 		if ((error = copyout(&ltv, args->timeout, sizeof(ltv))))
573 			goto select_out;
574 	}
575 
576 select_out:
577 #ifdef DEBUG
578 	if (ldebug(select))
579 		printf(LMSG("select_out -> %d"), error);
580 #endif
581 	return (error);
582 }
583 
584 int
585 linux_mremap(struct thread *td, struct linux_mremap_args *args)
586 {
587 	struct munmap_args /* {
588 		void *addr;
589 		size_t len;
590 	} */ bsd_args;
591 	int error = 0;
592 
593 #ifdef DEBUG
594 	if (ldebug(mremap))
595 		printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
596 		    (void *)(uintptr_t)args->addr,
597 		    (unsigned long)args->old_len,
598 		    (unsigned long)args->new_len,
599 		    (unsigned long)args->flags);
600 #endif
601 
602 	if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
603 		td->td_retval[0] = 0;
604 		return (EINVAL);
605 	}
606 
607 	/*
608 	 * Check for the page alignment.
609 	 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
610 	 */
611 	if (args->addr & PAGE_MASK) {
612 		td->td_retval[0] = 0;
613 		return (EINVAL);
614 	}
615 
616 	args->new_len = round_page(args->new_len);
617 	args->old_len = round_page(args->old_len);
618 
619 	if (args->new_len > args->old_len) {
620 		td->td_retval[0] = 0;
621 		return (ENOMEM);
622 	}
623 
624 	if (args->new_len < args->old_len) {
625 		bsd_args.addr =
626 		    (caddr_t)((uintptr_t)args->addr + args->new_len);
627 		bsd_args.len = args->old_len - args->new_len;
628 		error = sys_munmap(td, &bsd_args);
629 	}
630 
631 	td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
632 	return (error);
633 }
634 
635 #define LINUX_MS_ASYNC       0x0001
636 #define LINUX_MS_INVALIDATE  0x0002
637 #define LINUX_MS_SYNC        0x0004
638 
639 int
640 linux_msync(struct thread *td, struct linux_msync_args *args)
641 {
642 	struct msync_args bsd_args;
643 
644 	bsd_args.addr = (caddr_t)(uintptr_t)args->addr;
645 	bsd_args.len = (uintptr_t)args->len;
646 	bsd_args.flags = args->fl & ~LINUX_MS_SYNC;
647 
648 	return (sys_msync(td, &bsd_args));
649 }
650 
651 int
652 linux_time(struct thread *td, struct linux_time_args *args)
653 {
654 	struct timeval tv;
655 	l_time_t tm;
656 	int error;
657 
658 #ifdef DEBUG
659 	if (ldebug(time))
660 		printf(ARGS(time, "*"));
661 #endif
662 
663 	microtime(&tv);
664 	tm = tv.tv_sec;
665 	if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
666 		return (error);
667 	td->td_retval[0] = tm;
668 	return (0);
669 }
670 
671 struct l_times_argv {
672 	l_clock_t	tms_utime;
673 	l_clock_t	tms_stime;
674 	l_clock_t	tms_cutime;
675 	l_clock_t	tms_cstime;
676 };
677 
678 
679 /*
680  * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
681  * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
682  * auxiliary vector entry.
683  */
684 #define	CLK_TCK		100
685 
686 #define	CONVOTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
687 #define	CONVNTCK(r)	(r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
688 
689 #define	CONVTCK(r)	(linux_kernver(td) >= LINUX_KERNVER_2004000 ?		\
690 			    CONVNTCK(r) : CONVOTCK(r))
691 
692 int
693 linux_times(struct thread *td, struct linux_times_args *args)
694 {
695 	struct timeval tv, utime, stime, cutime, cstime;
696 	struct l_times_argv tms;
697 	struct proc *p;
698 	int error;
699 
700 #ifdef DEBUG
701 	if (ldebug(times))
702 		printf(ARGS(times, "*"));
703 #endif
704 
705 	if (args->buf != NULL) {
706 		p = td->td_proc;
707 		PROC_LOCK(p);
708 		PROC_STATLOCK(p);
709 		calcru(p, &utime, &stime);
710 		PROC_STATUNLOCK(p);
711 		calccru(p, &cutime, &cstime);
712 		PROC_UNLOCK(p);
713 
714 		tms.tms_utime = CONVTCK(utime);
715 		tms.tms_stime = CONVTCK(stime);
716 
717 		tms.tms_cutime = CONVTCK(cutime);
718 		tms.tms_cstime = CONVTCK(cstime);
719 
720 		if ((error = copyout(&tms, args->buf, sizeof(tms))))
721 			return (error);
722 	}
723 
724 	microuptime(&tv);
725 	td->td_retval[0] = (int)CONVTCK(tv);
726 	return (0);
727 }
728 
729 int
730 linux_newuname(struct thread *td, struct linux_newuname_args *args)
731 {
732 	struct l_new_utsname utsname;
733 	char osname[LINUX_MAX_UTSNAME];
734 	char osrelease[LINUX_MAX_UTSNAME];
735 	char *p;
736 
737 #ifdef DEBUG
738 	if (ldebug(newuname))
739 		printf(ARGS(newuname, "*"));
740 #endif
741 
742 	linux_get_osname(td, osname);
743 	linux_get_osrelease(td, osrelease);
744 
745 	bzero(&utsname, sizeof(utsname));
746 	strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
747 	getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
748 	getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
749 	strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
750 	strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
751 	for (p = utsname.version; *p != '\0'; ++p)
752 		if (*p == '\n') {
753 			*p = '\0';
754 			break;
755 		}
756 	strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME);
757 
758 	return (copyout(&utsname, args->buf, sizeof(utsname)));
759 }
760 
761 struct l_utimbuf {
762 	l_time_t l_actime;
763 	l_time_t l_modtime;
764 };
765 
766 int
767 linux_utime(struct thread *td, struct linux_utime_args *args)
768 {
769 	struct timeval tv[2], *tvp;
770 	struct l_utimbuf lut;
771 	char *fname;
772 	int error;
773 
774 	LCONVPATHEXIST(td, args->fname, &fname);
775 
776 #ifdef DEBUG
777 	if (ldebug(utime))
778 		printf(ARGS(utime, "%s, *"), fname);
779 #endif
780 
781 	if (args->times) {
782 		if ((error = copyin(args->times, &lut, sizeof lut))) {
783 			LFREEPATH(fname);
784 			return (error);
785 		}
786 		tv[0].tv_sec = lut.l_actime;
787 		tv[0].tv_usec = 0;
788 		tv[1].tv_sec = lut.l_modtime;
789 		tv[1].tv_usec = 0;
790 		tvp = tv;
791 	} else
792 		tvp = NULL;
793 
794 	error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, tvp,
795 	    UIO_SYSSPACE);
796 	LFREEPATH(fname);
797 	return (error);
798 }
799 
800 int
801 linux_utimes(struct thread *td, struct linux_utimes_args *args)
802 {
803 	l_timeval ltv[2];
804 	struct timeval tv[2], *tvp = NULL;
805 	char *fname;
806 	int error;
807 
808 	LCONVPATHEXIST(td, args->fname, &fname);
809 
810 #ifdef DEBUG
811 	if (ldebug(utimes))
812 		printf(ARGS(utimes, "%s, *"), fname);
813 #endif
814 
815 	if (args->tptr != NULL) {
816 		if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
817 			LFREEPATH(fname);
818 			return (error);
819 		}
820 		tv[0].tv_sec = ltv[0].tv_sec;
821 		tv[0].tv_usec = ltv[0].tv_usec;
822 		tv[1].tv_sec = ltv[1].tv_sec;
823 		tv[1].tv_usec = ltv[1].tv_usec;
824 		tvp = tv;
825 	}
826 
827 	error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE,
828 	    tvp, UIO_SYSSPACE);
829 	LFREEPATH(fname);
830 	return (error);
831 }
832 
833 static int
834 linux_utimensat_nsec_valid(l_long nsec)
835 {
836 
837 	if (nsec == LINUX_UTIME_OMIT || nsec == LINUX_UTIME_NOW)
838 		return (0);
839 	if (nsec >= 0 && nsec <= 999999999)
840 		return (0);
841 	return (1);
842 }
843 
844 int
845 linux_utimensat(struct thread *td, struct linux_utimensat_args *args)
846 {
847 	struct l_timespec l_times[2];
848 	struct timespec times[2], *timesp = NULL;
849 	char *path = NULL;
850 	int error, dfd, flags = 0;
851 
852 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
853 
854 #ifdef DEBUG
855 	if (ldebug(utimensat))
856 		printf(ARGS(utimensat, "%d, *"), dfd);
857 #endif
858 
859 	if (args->flags & ~LINUX_AT_SYMLINK_NOFOLLOW)
860 		return (EINVAL);
861 
862 	if (args->times != NULL) {
863 		error = copyin(args->times, l_times, sizeof(l_times));
864 		if (error != 0)
865 			return (error);
866 
867 		if (linux_utimensat_nsec_valid(l_times[0].tv_nsec) != 0 ||
868 		    linux_utimensat_nsec_valid(l_times[1].tv_nsec) != 0)
869 			return (EINVAL);
870 
871 		times[0].tv_sec = l_times[0].tv_sec;
872 		switch (l_times[0].tv_nsec)
873 		{
874 		case LINUX_UTIME_OMIT:
875 			times[0].tv_nsec = UTIME_OMIT;
876 			break;
877 		case LINUX_UTIME_NOW:
878 			times[0].tv_nsec = UTIME_NOW;
879 			break;
880 		default:
881 			times[0].tv_nsec = l_times[0].tv_nsec;
882 		}
883 
884 		times[1].tv_sec = l_times[1].tv_sec;
885 		switch (l_times[1].tv_nsec)
886 		{
887 		case LINUX_UTIME_OMIT:
888 			times[1].tv_nsec = UTIME_OMIT;
889 			break;
890 		case LINUX_UTIME_NOW:
891 			times[1].tv_nsec = UTIME_NOW;
892 			break;
893 		default:
894 			times[1].tv_nsec = l_times[1].tv_nsec;
895 			break;
896 		}
897 		timesp = times;
898 
899 		/* This breaks POSIX, but is what the Linux kernel does
900 		 * _on purpose_ (documented in the man page for utimensat(2)),
901 		 * so we must follow that behaviour. */
902 		if (times[0].tv_nsec == UTIME_OMIT &&
903 		    times[1].tv_nsec == UTIME_OMIT)
904 			return (0);
905 	}
906 
907 	if (args->pathname != NULL)
908 		LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
909 	else if (args->flags != 0)
910 		return (EINVAL);
911 
912 	if (args->flags & LINUX_AT_SYMLINK_NOFOLLOW)
913 		flags |= AT_SYMLINK_NOFOLLOW;
914 
915 	if (path == NULL)
916 		error = kern_futimens(td, dfd, timesp, UIO_SYSSPACE);
917 	else {
918 		error = kern_utimensat(td, dfd, path, UIO_SYSSPACE, timesp,
919 	    		UIO_SYSSPACE, flags);
920 		LFREEPATH(path);
921 	}
922 
923 	return (error);
924 }
925 
926 int
927 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
928 {
929 	l_timeval ltv[2];
930 	struct timeval tv[2], *tvp = NULL;
931 	char *fname;
932 	int error, dfd;
933 
934 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
935 	LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
936 
937 #ifdef DEBUG
938 	if (ldebug(futimesat))
939 		printf(ARGS(futimesat, "%s, *"), fname);
940 #endif
941 
942 	if (args->utimes != NULL) {
943 		if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
944 			LFREEPATH(fname);
945 			return (error);
946 		}
947 		tv[0].tv_sec = ltv[0].tv_sec;
948 		tv[0].tv_usec = ltv[0].tv_usec;
949 		tv[1].tv_sec = ltv[1].tv_sec;
950 		tv[1].tv_usec = ltv[1].tv_usec;
951 		tvp = tv;
952 	}
953 
954 	error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
955 	LFREEPATH(fname);
956 	return (error);
957 }
958 
959 int
960 linux_common_wait(struct thread *td, int pid, int *status,
961     int options, struct rusage *ru)
962 {
963 	int error, tmpstat;
964 
965 	error = kern_wait(td, pid, &tmpstat, options, ru);
966 	if (error)
967 		return (error);
968 
969 	if (status) {
970 		tmpstat &= 0xffff;
971 		if (WIFSIGNALED(tmpstat))
972 			tmpstat = (tmpstat & 0xffffff80) |
973 			    bsd_to_linux_signal(WTERMSIG(tmpstat));
974 		else if (WIFSTOPPED(tmpstat))
975 			tmpstat = (tmpstat & 0xffff00ff) |
976 			    (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8);
977 		else if (WIFCONTINUED(tmpstat))
978 			tmpstat = 0xffff;
979 		error = copyout(&tmpstat, status, sizeof(int));
980 	}
981 
982 	return (error);
983 }
984 
985 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
986 int
987 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
988 {
989 	struct linux_wait4_args wait4_args;
990 
991 #ifdef DEBUG
992 	if (ldebug(waitpid))
993 		printf(ARGS(waitpid, "%d, %p, %d"),
994 		    args->pid, (void *)args->status, args->options);
995 #endif
996 
997 	wait4_args.pid = args->pid;
998 	wait4_args.status = args->status;
999 	wait4_args.options = args->options;
1000 	wait4_args.rusage = NULL;
1001 
1002 	return (linux_wait4(td, &wait4_args));
1003 }
1004 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1005 
1006 int
1007 linux_wait4(struct thread *td, struct linux_wait4_args *args)
1008 {
1009 	int error, options;
1010 	struct rusage ru, *rup;
1011 
1012 #ifdef DEBUG
1013 	if (ldebug(wait4))
1014 		printf(ARGS(wait4, "%d, %p, %d, %p"),
1015 		    args->pid, (void *)args->status, args->options,
1016 		    (void *)args->rusage);
1017 #endif
1018 	if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG |
1019 	    LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
1020 		return (EINVAL);
1021 
1022 	options = WEXITED;
1023 	linux_to_bsd_waitopts(args->options, &options);
1024 
1025 	if (args->rusage != NULL)
1026 		rup = &ru;
1027 	else
1028 		rup = NULL;
1029 	error = linux_common_wait(td, args->pid, args->status, options, rup);
1030 	if (error != 0)
1031 		return (error);
1032 	if (args->rusage != NULL)
1033 		error = linux_copyout_rusage(&ru, args->rusage);
1034 	return (error);
1035 }
1036 
1037 int
1038 linux_waitid(struct thread *td, struct linux_waitid_args *args)
1039 {
1040 	int status, options, sig;
1041 	struct __wrusage wru;
1042 	siginfo_t siginfo;
1043 	l_siginfo_t lsi;
1044 	idtype_t idtype;
1045 	struct proc *p;
1046 	int error;
1047 
1048 	options = 0;
1049 	linux_to_bsd_waitopts(args->options, &options);
1050 
1051 	if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED))
1052 		return (EINVAL);
1053 	if (!(options & (WEXITED | WUNTRACED | WCONTINUED)))
1054 		return (EINVAL);
1055 
1056 	switch (args->idtype) {
1057 	case LINUX_P_ALL:
1058 		idtype = P_ALL;
1059 		break;
1060 	case LINUX_P_PID:
1061 		if (args->id <= 0)
1062 			return (EINVAL);
1063 		idtype = P_PID;
1064 		break;
1065 	case LINUX_P_PGID:
1066 		if (args->id <= 0)
1067 			return (EINVAL);
1068 		idtype = P_PGID;
1069 		break;
1070 	default:
1071 		return (EINVAL);
1072 	}
1073 
1074 	error = kern_wait6(td, idtype, args->id, &status, options,
1075 	    &wru, &siginfo);
1076 	if (error != 0)
1077 		return (error);
1078 	if (args->rusage != NULL) {
1079 		error = linux_copyout_rusage(&wru.wru_children,
1080 		    args->rusage);
1081 		if (error != 0)
1082 			return (error);
1083 	}
1084 	if (args->info != NULL) {
1085 		p = td->td_proc;
1086 		if (td->td_retval[0] == 0)
1087 			bzero(&lsi, sizeof(lsi));
1088 		else {
1089 			sig = bsd_to_linux_signal(siginfo.si_signo);
1090 			siginfo_to_lsiginfo(&siginfo, &lsi, sig);
1091 		}
1092 		error = copyout(&lsi, args->info, sizeof(lsi));
1093 	}
1094 	td->td_retval[0] = 0;
1095 
1096 	return (error);
1097 }
1098 
1099 int
1100 linux_mknod(struct thread *td, struct linux_mknod_args *args)
1101 {
1102 	char *path;
1103 	int error;
1104 
1105 	LCONVPATHCREAT(td, args->path, &path);
1106 
1107 #ifdef DEBUG
1108 	if (ldebug(mknod))
1109 		printf(ARGS(mknod, "%s, %d, %ju"), path, args->mode,
1110 		    (uintmax_t)args->dev);
1111 #endif
1112 
1113 	switch (args->mode & S_IFMT) {
1114 	case S_IFIFO:
1115 	case S_IFSOCK:
1116 		error = kern_mkfifoat(td, AT_FDCWD, path, UIO_SYSSPACE,
1117 		    args->mode);
1118 		break;
1119 
1120 	case S_IFCHR:
1121 	case S_IFBLK:
1122 		error = kern_mknodat(td, AT_FDCWD, path, UIO_SYSSPACE,
1123 		    args->mode, args->dev);
1124 		break;
1125 
1126 	case S_IFDIR:
1127 		error = EPERM;
1128 		break;
1129 
1130 	case 0:
1131 		args->mode |= S_IFREG;
1132 		/* FALLTHROUGH */
1133 	case S_IFREG:
1134 		error = kern_openat(td, AT_FDCWD, path, UIO_SYSSPACE,
1135 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1136 		if (error == 0)
1137 			kern_close(td, td->td_retval[0]);
1138 		break;
1139 
1140 	default:
1141 		error = EINVAL;
1142 		break;
1143 	}
1144 	LFREEPATH(path);
1145 	return (error);
1146 }
1147 
1148 int
1149 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
1150 {
1151 	char *path;
1152 	int error, dfd;
1153 
1154 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
1155 	LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
1156 
1157 #ifdef DEBUG
1158 	if (ldebug(mknodat))
1159 		printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev);
1160 #endif
1161 
1162 	switch (args->mode & S_IFMT) {
1163 	case S_IFIFO:
1164 	case S_IFSOCK:
1165 		error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
1166 		break;
1167 
1168 	case S_IFCHR:
1169 	case S_IFBLK:
1170 		error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
1171 		    args->dev);
1172 		break;
1173 
1174 	case S_IFDIR:
1175 		error = EPERM;
1176 		break;
1177 
1178 	case 0:
1179 		args->mode |= S_IFREG;
1180 		/* FALLTHROUGH */
1181 	case S_IFREG:
1182 		error = kern_openat(td, dfd, path, UIO_SYSSPACE,
1183 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1184 		if (error == 0)
1185 			kern_close(td, td->td_retval[0]);
1186 		break;
1187 
1188 	default:
1189 		error = EINVAL;
1190 		break;
1191 	}
1192 	LFREEPATH(path);
1193 	return (error);
1194 }
1195 
1196 /*
1197  * UGH! This is just about the dumbest idea I've ever heard!!
1198  */
1199 int
1200 linux_personality(struct thread *td, struct linux_personality_args *args)
1201 {
1202 #ifdef DEBUG
1203 	if (ldebug(personality))
1204 		printf(ARGS(personality, "%lu"), (unsigned long)args->per);
1205 #endif
1206 	if (args->per != 0)
1207 		return (EINVAL);
1208 
1209 	/* Yes Jim, it's still a Linux... */
1210 	td->td_retval[0] = 0;
1211 	return (0);
1212 }
1213 
1214 struct l_itimerval {
1215 	l_timeval it_interval;
1216 	l_timeval it_value;
1217 };
1218 
1219 #define	B2L_ITIMERVAL(bip, lip) 					\
1220 	(bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec;		\
1221 	(bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec;	\
1222 	(bip)->it_value.tv_sec = (lip)->it_value.tv_sec;		\
1223 	(bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1224 
1225 int
1226 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1227 {
1228 	int error;
1229 	struct l_itimerval ls;
1230 	struct itimerval aitv, oitv;
1231 
1232 #ifdef DEBUG
1233 	if (ldebug(setitimer))
1234 		printf(ARGS(setitimer, "%p, %p"),
1235 		    (void *)uap->itv, (void *)uap->oitv);
1236 #endif
1237 
1238 	if (uap->itv == NULL) {
1239 		uap->itv = uap->oitv;
1240 		return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1241 	}
1242 
1243 	error = copyin(uap->itv, &ls, sizeof(ls));
1244 	if (error != 0)
1245 		return (error);
1246 	B2L_ITIMERVAL(&aitv, &ls);
1247 #ifdef DEBUG
1248 	if (ldebug(setitimer)) {
1249 		printf("setitimer: value: sec: %jd, usec: %ld\n",
1250 		    (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1251 		printf("setitimer: interval: sec: %jd, usec: %ld\n",
1252 		    (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1253 	}
1254 #endif
1255 	error = kern_setitimer(td, uap->which, &aitv, &oitv);
1256 	if (error != 0 || uap->oitv == NULL)
1257 		return (error);
1258 	B2L_ITIMERVAL(&ls, &oitv);
1259 
1260 	return (copyout(&ls, uap->oitv, sizeof(ls)));
1261 }
1262 
1263 int
1264 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1265 {
1266 	int error;
1267 	struct l_itimerval ls;
1268 	struct itimerval aitv;
1269 
1270 #ifdef DEBUG
1271 	if (ldebug(getitimer))
1272 		printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1273 #endif
1274 	error = kern_getitimer(td, uap->which, &aitv);
1275 	if (error != 0)
1276 		return (error);
1277 	B2L_ITIMERVAL(&ls, &aitv);
1278 	return (copyout(&ls, uap->itv, sizeof(ls)));
1279 }
1280 
1281 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1282 int
1283 linux_nice(struct thread *td, struct linux_nice_args *args)
1284 {
1285 	struct setpriority_args bsd_args;
1286 
1287 	bsd_args.which = PRIO_PROCESS;
1288 	bsd_args.who = 0;		/* current process */
1289 	bsd_args.prio = args->inc;
1290 	return (sys_setpriority(td, &bsd_args));
1291 }
1292 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1293 
1294 int
1295 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1296 {
1297 	struct ucred *newcred, *oldcred;
1298 	l_gid_t *linux_gidset;
1299 	gid_t *bsd_gidset;
1300 	int ngrp, error;
1301 	struct proc *p;
1302 
1303 	ngrp = args->gidsetsize;
1304 	if (ngrp < 0 || ngrp >= ngroups_max + 1)
1305 		return (EINVAL);
1306 	linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK);
1307 	error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1308 	if (error)
1309 		goto out;
1310 	newcred = crget();
1311 	crextend(newcred, ngrp + 1);
1312 	p = td->td_proc;
1313 	PROC_LOCK(p);
1314 	oldcred = p->p_ucred;
1315 	crcopy(newcred, oldcred);
1316 
1317 	/*
1318 	 * cr_groups[0] holds egid. Setting the whole set from
1319 	 * the supplied set will cause egid to be changed too.
1320 	 * Keep cr_groups[0] unchanged to prevent that.
1321 	 */
1322 
1323 	if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) {
1324 		PROC_UNLOCK(p);
1325 		crfree(newcred);
1326 		goto out;
1327 	}
1328 
1329 	if (ngrp > 0) {
1330 		newcred->cr_ngroups = ngrp + 1;
1331 
1332 		bsd_gidset = newcred->cr_groups;
1333 		ngrp--;
1334 		while (ngrp >= 0) {
1335 			bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1336 			ngrp--;
1337 		}
1338 	} else
1339 		newcred->cr_ngroups = 1;
1340 
1341 	setsugid(p);
1342 	proc_set_cred(p, newcred);
1343 	PROC_UNLOCK(p);
1344 	crfree(oldcred);
1345 	error = 0;
1346 out:
1347 	free(linux_gidset, M_LINUX);
1348 	return (error);
1349 }
1350 
1351 int
1352 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1353 {
1354 	struct ucred *cred;
1355 	l_gid_t *linux_gidset;
1356 	gid_t *bsd_gidset;
1357 	int bsd_gidsetsz, ngrp, error;
1358 
1359 	cred = td->td_ucred;
1360 	bsd_gidset = cred->cr_groups;
1361 	bsd_gidsetsz = cred->cr_ngroups - 1;
1362 
1363 	/*
1364 	 * cr_groups[0] holds egid. Returning the whole set
1365 	 * here will cause a duplicate. Exclude cr_groups[0]
1366 	 * to prevent that.
1367 	 */
1368 
1369 	if ((ngrp = args->gidsetsize) == 0) {
1370 		td->td_retval[0] = bsd_gidsetsz;
1371 		return (0);
1372 	}
1373 
1374 	if (ngrp < bsd_gidsetsz)
1375 		return (EINVAL);
1376 
1377 	ngrp = 0;
1378 	linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1379 	    M_LINUX, M_WAITOK);
1380 	while (ngrp < bsd_gidsetsz) {
1381 		linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1382 		ngrp++;
1383 	}
1384 
1385 	error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1386 	free(linux_gidset, M_LINUX);
1387 	if (error)
1388 		return (error);
1389 
1390 	td->td_retval[0] = ngrp;
1391 	return (0);
1392 }
1393 
1394 int
1395 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1396 {
1397 	struct rlimit bsd_rlim;
1398 	struct l_rlimit rlim;
1399 	u_int which;
1400 	int error;
1401 
1402 #ifdef DEBUG
1403 	if (ldebug(setrlimit))
1404 		printf(ARGS(setrlimit, "%d, %p"),
1405 		    args->resource, (void *)args->rlim);
1406 #endif
1407 
1408 	if (args->resource >= LINUX_RLIM_NLIMITS)
1409 		return (EINVAL);
1410 
1411 	which = linux_to_bsd_resource[args->resource];
1412 	if (which == -1)
1413 		return (EINVAL);
1414 
1415 	error = copyin(args->rlim, &rlim, sizeof(rlim));
1416 	if (error)
1417 		return (error);
1418 
1419 	bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1420 	bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1421 	return (kern_setrlimit(td, which, &bsd_rlim));
1422 }
1423 
1424 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1425 int
1426 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1427 {
1428 	struct l_rlimit rlim;
1429 	struct rlimit bsd_rlim;
1430 	u_int which;
1431 
1432 #ifdef DEBUG
1433 	if (ldebug(old_getrlimit))
1434 		printf(ARGS(old_getrlimit, "%d, %p"),
1435 		    args->resource, (void *)args->rlim);
1436 #endif
1437 
1438 	if (args->resource >= LINUX_RLIM_NLIMITS)
1439 		return (EINVAL);
1440 
1441 	which = linux_to_bsd_resource[args->resource];
1442 	if (which == -1)
1443 		return (EINVAL);
1444 
1445 	lim_rlimit(td, which, &bsd_rlim);
1446 
1447 #ifdef COMPAT_LINUX32
1448 	rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1449 	if (rlim.rlim_cur == UINT_MAX)
1450 		rlim.rlim_cur = INT_MAX;
1451 	rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1452 	if (rlim.rlim_max == UINT_MAX)
1453 		rlim.rlim_max = INT_MAX;
1454 #else
1455 	rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1456 	if (rlim.rlim_cur == ULONG_MAX)
1457 		rlim.rlim_cur = LONG_MAX;
1458 	rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1459 	if (rlim.rlim_max == ULONG_MAX)
1460 		rlim.rlim_max = LONG_MAX;
1461 #endif
1462 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1463 }
1464 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1465 
1466 int
1467 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1468 {
1469 	struct l_rlimit rlim;
1470 	struct rlimit bsd_rlim;
1471 	u_int which;
1472 
1473 #ifdef DEBUG
1474 	if (ldebug(getrlimit))
1475 		printf(ARGS(getrlimit, "%d, %p"),
1476 		    args->resource, (void *)args->rlim);
1477 #endif
1478 
1479 	if (args->resource >= LINUX_RLIM_NLIMITS)
1480 		return (EINVAL);
1481 
1482 	which = linux_to_bsd_resource[args->resource];
1483 	if (which == -1)
1484 		return (EINVAL);
1485 
1486 	lim_rlimit(td, which, &bsd_rlim);
1487 
1488 	rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1489 	rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1490 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1491 }
1492 
1493 int
1494 linux_sched_setscheduler(struct thread *td,
1495     struct linux_sched_setscheduler_args *args)
1496 {
1497 	struct sched_param sched_param;
1498 	struct thread *tdt;
1499 	int error, policy;
1500 
1501 #ifdef DEBUG
1502 	if (ldebug(sched_setscheduler))
1503 		printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1504 		    args->pid, args->policy, (const void *)args->param);
1505 #endif
1506 
1507 	switch (args->policy) {
1508 	case LINUX_SCHED_OTHER:
1509 		policy = SCHED_OTHER;
1510 		break;
1511 	case LINUX_SCHED_FIFO:
1512 		policy = SCHED_FIFO;
1513 		break;
1514 	case LINUX_SCHED_RR:
1515 		policy = SCHED_RR;
1516 		break;
1517 	default:
1518 		return (EINVAL);
1519 	}
1520 
1521 	error = copyin(args->param, &sched_param, sizeof(sched_param));
1522 	if (error)
1523 		return (error);
1524 
1525 	tdt = linux_tdfind(td, args->pid, -1);
1526 	if (tdt == NULL)
1527 		return (ESRCH);
1528 
1529 	error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1530 	PROC_UNLOCK(tdt->td_proc);
1531 	return (error);
1532 }
1533 
1534 int
1535 linux_sched_getscheduler(struct thread *td,
1536     struct linux_sched_getscheduler_args *args)
1537 {
1538 	struct thread *tdt;
1539 	int error, policy;
1540 
1541 #ifdef DEBUG
1542 	if (ldebug(sched_getscheduler))
1543 		printf(ARGS(sched_getscheduler, "%d"), args->pid);
1544 #endif
1545 
1546 	tdt = linux_tdfind(td, args->pid, -1);
1547 	if (tdt == NULL)
1548 		return (ESRCH);
1549 
1550 	error = kern_sched_getscheduler(td, tdt, &policy);
1551 	PROC_UNLOCK(tdt->td_proc);
1552 
1553 	switch (policy) {
1554 	case SCHED_OTHER:
1555 		td->td_retval[0] = LINUX_SCHED_OTHER;
1556 		break;
1557 	case SCHED_FIFO:
1558 		td->td_retval[0] = LINUX_SCHED_FIFO;
1559 		break;
1560 	case SCHED_RR:
1561 		td->td_retval[0] = LINUX_SCHED_RR;
1562 		break;
1563 	}
1564 	return (error);
1565 }
1566 
1567 int
1568 linux_sched_get_priority_max(struct thread *td,
1569     struct linux_sched_get_priority_max_args *args)
1570 {
1571 	struct sched_get_priority_max_args bsd;
1572 
1573 #ifdef DEBUG
1574 	if (ldebug(sched_get_priority_max))
1575 		printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1576 #endif
1577 
1578 	switch (args->policy) {
1579 	case LINUX_SCHED_OTHER:
1580 		bsd.policy = SCHED_OTHER;
1581 		break;
1582 	case LINUX_SCHED_FIFO:
1583 		bsd.policy = SCHED_FIFO;
1584 		break;
1585 	case LINUX_SCHED_RR:
1586 		bsd.policy = SCHED_RR;
1587 		break;
1588 	default:
1589 		return (EINVAL);
1590 	}
1591 	return (sys_sched_get_priority_max(td, &bsd));
1592 }
1593 
1594 int
1595 linux_sched_get_priority_min(struct thread *td,
1596     struct linux_sched_get_priority_min_args *args)
1597 {
1598 	struct sched_get_priority_min_args bsd;
1599 
1600 #ifdef DEBUG
1601 	if (ldebug(sched_get_priority_min))
1602 		printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1603 #endif
1604 
1605 	switch (args->policy) {
1606 	case LINUX_SCHED_OTHER:
1607 		bsd.policy = SCHED_OTHER;
1608 		break;
1609 	case LINUX_SCHED_FIFO:
1610 		bsd.policy = SCHED_FIFO;
1611 		break;
1612 	case LINUX_SCHED_RR:
1613 		bsd.policy = SCHED_RR;
1614 		break;
1615 	default:
1616 		return (EINVAL);
1617 	}
1618 	return (sys_sched_get_priority_min(td, &bsd));
1619 }
1620 
1621 #define REBOOT_CAD_ON	0x89abcdef
1622 #define REBOOT_CAD_OFF	0
1623 #define REBOOT_HALT	0xcdef0123
1624 #define REBOOT_RESTART	0x01234567
1625 #define REBOOT_RESTART2	0xA1B2C3D4
1626 #define REBOOT_POWEROFF	0x4321FEDC
1627 #define REBOOT_MAGIC1	0xfee1dead
1628 #define REBOOT_MAGIC2	0x28121969
1629 #define REBOOT_MAGIC2A	0x05121996
1630 #define REBOOT_MAGIC2B	0x16041998
1631 
1632 int
1633 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1634 {
1635 	struct reboot_args bsd_args;
1636 
1637 #ifdef DEBUG
1638 	if (ldebug(reboot))
1639 		printf(ARGS(reboot, "0x%x"), args->cmd);
1640 #endif
1641 
1642 	if (args->magic1 != REBOOT_MAGIC1)
1643 		return (EINVAL);
1644 
1645 	switch (args->magic2) {
1646 	case REBOOT_MAGIC2:
1647 	case REBOOT_MAGIC2A:
1648 	case REBOOT_MAGIC2B:
1649 		break;
1650 	default:
1651 		return (EINVAL);
1652 	}
1653 
1654 	switch (args->cmd) {
1655 	case REBOOT_CAD_ON:
1656 	case REBOOT_CAD_OFF:
1657 		return (priv_check(td, PRIV_REBOOT));
1658 	case REBOOT_HALT:
1659 		bsd_args.opt = RB_HALT;
1660 		break;
1661 	case REBOOT_RESTART:
1662 	case REBOOT_RESTART2:
1663 		bsd_args.opt = 0;
1664 		break;
1665 	case REBOOT_POWEROFF:
1666 		bsd_args.opt = RB_POWEROFF;
1667 		break;
1668 	default:
1669 		return (EINVAL);
1670 	}
1671 	return (sys_reboot(td, &bsd_args));
1672 }
1673 
1674 
1675 /*
1676  * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1677  * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
1678  * are assumed to be preserved. The following lightweight syscalls fixes
1679  * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
1680  *
1681  * linux_getpid() - MP SAFE
1682  * linux_getgid() - MP SAFE
1683  * linux_getuid() - MP SAFE
1684  */
1685 
1686 int
1687 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1688 {
1689 
1690 #ifdef DEBUG
1691 	if (ldebug(getpid))
1692 		printf(ARGS(getpid, ""));
1693 #endif
1694 	td->td_retval[0] = td->td_proc->p_pid;
1695 
1696 	return (0);
1697 }
1698 
1699 int
1700 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1701 {
1702 	struct linux_emuldata *em;
1703 
1704 #ifdef DEBUG
1705 	if (ldebug(gettid))
1706 		printf(ARGS(gettid, ""));
1707 #endif
1708 
1709 	em = em_find(td);
1710 	KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1711 
1712 	td->td_retval[0] = em->em_tid;
1713 
1714 	return (0);
1715 }
1716 
1717 
1718 int
1719 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1720 {
1721 
1722 #ifdef DEBUG
1723 	if (ldebug(getppid))
1724 		printf(ARGS(getppid, ""));
1725 #endif
1726 
1727 	PROC_LOCK(td->td_proc);
1728 	td->td_retval[0] = td->td_proc->p_pptr->p_pid;
1729 	PROC_UNLOCK(td->td_proc);
1730 	return (0);
1731 }
1732 
1733 int
1734 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1735 {
1736 
1737 #ifdef DEBUG
1738 	if (ldebug(getgid))
1739 		printf(ARGS(getgid, ""));
1740 #endif
1741 
1742 	td->td_retval[0] = td->td_ucred->cr_rgid;
1743 	return (0);
1744 }
1745 
1746 int
1747 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1748 {
1749 
1750 #ifdef DEBUG
1751 	if (ldebug(getuid))
1752 		printf(ARGS(getuid, ""));
1753 #endif
1754 
1755 	td->td_retval[0] = td->td_ucred->cr_ruid;
1756 	return (0);
1757 }
1758 
1759 
1760 int
1761 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1762 {
1763 	struct getsid_args bsd;
1764 
1765 #ifdef DEBUG
1766 	if (ldebug(getsid))
1767 		printf(ARGS(getsid, "%i"), args->pid);
1768 #endif
1769 
1770 	bsd.pid = args->pid;
1771 	return (sys_getsid(td, &bsd));
1772 }
1773 
1774 int
1775 linux_nosys(struct thread *td, struct nosys_args *ignore)
1776 {
1777 
1778 	return (ENOSYS);
1779 }
1780 
1781 int
1782 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1783 {
1784 	struct getpriority_args bsd_args;
1785 	int error;
1786 
1787 #ifdef DEBUG
1788 	if (ldebug(getpriority))
1789 		printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
1790 #endif
1791 
1792 	bsd_args.which = args->which;
1793 	bsd_args.who = args->who;
1794 	error = sys_getpriority(td, &bsd_args);
1795 	td->td_retval[0] = 20 - td->td_retval[0];
1796 	return (error);
1797 }
1798 
1799 int
1800 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1801 {
1802 	int name[2];
1803 
1804 #ifdef DEBUG
1805 	if (ldebug(sethostname))
1806 		printf(ARGS(sethostname, "*, %i"), args->len);
1807 #endif
1808 
1809 	name[0] = CTL_KERN;
1810 	name[1] = KERN_HOSTNAME;
1811 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1812 	    args->len, 0, 0));
1813 }
1814 
1815 int
1816 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1817 {
1818 	int name[2];
1819 
1820 #ifdef DEBUG
1821 	if (ldebug(setdomainname))
1822 		printf(ARGS(setdomainname, "*, %i"), args->len);
1823 #endif
1824 
1825 	name[0] = CTL_KERN;
1826 	name[1] = KERN_NISDOMAINNAME;
1827 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1828 	    args->len, 0, 0));
1829 }
1830 
1831 int
1832 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1833 {
1834 
1835 #ifdef DEBUG
1836 	if (ldebug(exit_group))
1837 		printf(ARGS(exit_group, "%i"), args->error_code);
1838 #endif
1839 
1840 	LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1841 	    args->error_code);
1842 
1843 	/*
1844 	 * XXX: we should send a signal to the parent if
1845 	 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1846 	 * as it doesnt occur often.
1847 	 */
1848 	exit1(td, args->error_code, 0);
1849 		/* NOTREACHED */
1850 }
1851 
1852 #define _LINUX_CAPABILITY_VERSION  0x19980330
1853 
1854 struct l_user_cap_header {
1855 	l_int	version;
1856 	l_int	pid;
1857 };
1858 
1859 struct l_user_cap_data {
1860 	l_int	effective;
1861 	l_int	permitted;
1862 	l_int	inheritable;
1863 };
1864 
1865 int
1866 linux_capget(struct thread *td, struct linux_capget_args *args)
1867 {
1868 	struct l_user_cap_header luch;
1869 	struct l_user_cap_data lucd;
1870 	int error;
1871 
1872 	if (args->hdrp == NULL)
1873 		return (EFAULT);
1874 
1875 	error = copyin(args->hdrp, &luch, sizeof(luch));
1876 	if (error != 0)
1877 		return (error);
1878 
1879 	if (luch.version != _LINUX_CAPABILITY_VERSION) {
1880 		luch.version = _LINUX_CAPABILITY_VERSION;
1881 		error = copyout(&luch, args->hdrp, sizeof(luch));
1882 		if (error)
1883 			return (error);
1884 		return (EINVAL);
1885 	}
1886 
1887 	if (luch.pid)
1888 		return (EPERM);
1889 
1890 	if (args->datap) {
1891 		/*
1892 		 * The current implementation doesn't support setting
1893 		 * a capability (it's essentially a stub) so indicate
1894 		 * that no capabilities are currently set or available
1895 		 * to request.
1896 		 */
1897 		bzero (&lucd, sizeof(lucd));
1898 		error = copyout(&lucd, args->datap, sizeof(lucd));
1899 	}
1900 
1901 	return (error);
1902 }
1903 
1904 int
1905 linux_capset(struct thread *td, struct linux_capset_args *args)
1906 {
1907 	struct l_user_cap_header luch;
1908 	struct l_user_cap_data lucd;
1909 	int error;
1910 
1911 	if (args->hdrp == NULL || args->datap == NULL)
1912 		return (EFAULT);
1913 
1914 	error = copyin(args->hdrp, &luch, sizeof(luch));
1915 	if (error != 0)
1916 		return (error);
1917 
1918 	if (luch.version != _LINUX_CAPABILITY_VERSION) {
1919 		luch.version = _LINUX_CAPABILITY_VERSION;
1920 		error = copyout(&luch, args->hdrp, sizeof(luch));
1921 		if (error)
1922 			return (error);
1923 		return (EINVAL);
1924 	}
1925 
1926 	if (luch.pid)
1927 		return (EPERM);
1928 
1929 	error = copyin(args->datap, &lucd, sizeof(lucd));
1930 	if (error != 0)
1931 		return (error);
1932 
1933 	/* We currently don't support setting any capabilities. */
1934 	if (lucd.effective || lucd.permitted || lucd.inheritable) {
1935 		linux_msg(td,
1936 			  "capset effective=0x%x, permitted=0x%x, "
1937 			  "inheritable=0x%x is not implemented",
1938 			  (int)lucd.effective, (int)lucd.permitted,
1939 			  (int)lucd.inheritable);
1940 		return (EPERM);
1941 	}
1942 
1943 	return (0);
1944 }
1945 
1946 int
1947 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1948 {
1949 	int error = 0, max_size;
1950 	struct proc *p = td->td_proc;
1951 	char comm[LINUX_MAX_COMM_LEN];
1952 	struct linux_emuldata *em;
1953 	int pdeath_signal;
1954 
1955 #ifdef DEBUG
1956 	if (ldebug(prctl))
1957 		printf(ARGS(prctl, "%d, %ju, %ju, %ju, %ju"), args->option,
1958 		    (uintmax_t)args->arg2, (uintmax_t)args->arg3,
1959 		    (uintmax_t)args->arg4, (uintmax_t)args->arg5);
1960 #endif
1961 
1962 	switch (args->option) {
1963 	case LINUX_PR_SET_PDEATHSIG:
1964 		if (!LINUX_SIG_VALID(args->arg2))
1965 			return (EINVAL);
1966 		em = em_find(td);
1967 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1968 		em->pdeath_signal = args->arg2;
1969 		break;
1970 	case LINUX_PR_GET_PDEATHSIG:
1971 		em = em_find(td);
1972 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1973 		pdeath_signal = em->pdeath_signal;
1974 		error = copyout(&pdeath_signal,
1975 		    (void *)(register_t)args->arg2,
1976 		    sizeof(pdeath_signal));
1977 		break;
1978 	case LINUX_PR_GET_KEEPCAPS:
1979 		/*
1980 		 * Indicate that we always clear the effective and
1981 		 * permitted capability sets when the user id becomes
1982 		 * non-zero (actually the capability sets are simply
1983 		 * always zero in the current implementation).
1984 		 */
1985 		td->td_retval[0] = 0;
1986 		break;
1987 	case LINUX_PR_SET_KEEPCAPS:
1988 		/*
1989 		 * Ignore requests to keep the effective and permitted
1990 		 * capability sets when the user id becomes non-zero.
1991 		 */
1992 		break;
1993 	case LINUX_PR_SET_NAME:
1994 		/*
1995 		 * To be on the safe side we need to make sure to not
1996 		 * overflow the size a linux program expects. We already
1997 		 * do this here in the copyin, so that we don't need to
1998 		 * check on copyout.
1999 		 */
2000 		max_size = MIN(sizeof(comm), sizeof(p->p_comm));
2001 		error = copyinstr((void *)(register_t)args->arg2, comm,
2002 		    max_size, NULL);
2003 
2004 		/* Linux silently truncates the name if it is too long. */
2005 		if (error == ENAMETOOLONG) {
2006 			/*
2007 			 * XXX: copyinstr() isn't documented to populate the
2008 			 * array completely, so do a copyin() to be on the
2009 			 * safe side. This should be changed in case
2010 			 * copyinstr() is changed to guarantee this.
2011 			 */
2012 			error = copyin((void *)(register_t)args->arg2, comm,
2013 			    max_size - 1);
2014 			comm[max_size - 1] = '\0';
2015 		}
2016 		if (error)
2017 			return (error);
2018 
2019 		PROC_LOCK(p);
2020 		strlcpy(p->p_comm, comm, sizeof(p->p_comm));
2021 		PROC_UNLOCK(p);
2022 		break;
2023 	case LINUX_PR_GET_NAME:
2024 		PROC_LOCK(p);
2025 		strlcpy(comm, p->p_comm, sizeof(comm));
2026 		PROC_UNLOCK(p);
2027 		error = copyout(comm, (void *)(register_t)args->arg2,
2028 		    strlen(comm) + 1);
2029 		break;
2030 	default:
2031 		error = EINVAL;
2032 		break;
2033 	}
2034 
2035 	return (error);
2036 }
2037 
2038 int
2039 linux_sched_setparam(struct thread *td,
2040     struct linux_sched_setparam_args *uap)
2041 {
2042 	struct sched_param sched_param;
2043 	struct thread *tdt;
2044 	int error;
2045 
2046 #ifdef DEBUG
2047 	if (ldebug(sched_setparam))
2048 		printf(ARGS(sched_setparam, "%d, *"), uap->pid);
2049 #endif
2050 
2051 	error = copyin(uap->param, &sched_param, sizeof(sched_param));
2052 	if (error)
2053 		return (error);
2054 
2055 	tdt = linux_tdfind(td, uap->pid, -1);
2056 	if (tdt == NULL)
2057 		return (ESRCH);
2058 
2059 	error = kern_sched_setparam(td, tdt, &sched_param);
2060 	PROC_UNLOCK(tdt->td_proc);
2061 	return (error);
2062 }
2063 
2064 int
2065 linux_sched_getparam(struct thread *td,
2066     struct linux_sched_getparam_args *uap)
2067 {
2068 	struct sched_param sched_param;
2069 	struct thread *tdt;
2070 	int error;
2071 
2072 #ifdef DEBUG
2073 	if (ldebug(sched_getparam))
2074 		printf(ARGS(sched_getparam, "%d, *"), uap->pid);
2075 #endif
2076 
2077 	tdt = linux_tdfind(td, uap->pid, -1);
2078 	if (tdt == NULL)
2079 		return (ESRCH);
2080 
2081 	error = kern_sched_getparam(td, tdt, &sched_param);
2082 	PROC_UNLOCK(tdt->td_proc);
2083 	if (error == 0)
2084 		error = copyout(&sched_param, uap->param,
2085 		    sizeof(sched_param));
2086 	return (error);
2087 }
2088 
2089 /*
2090  * Get affinity of a process.
2091  */
2092 int
2093 linux_sched_getaffinity(struct thread *td,
2094     struct linux_sched_getaffinity_args *args)
2095 {
2096 	int error;
2097 	struct thread *tdt;
2098 	struct cpuset_getaffinity_args cga;
2099 
2100 #ifdef DEBUG
2101 	if (ldebug(sched_getaffinity))
2102 		printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid,
2103 		    args->len);
2104 #endif
2105 	if (args->len < sizeof(cpuset_t))
2106 		return (EINVAL);
2107 
2108 	tdt = linux_tdfind(td, args->pid, -1);
2109 	if (tdt == NULL)
2110 		return (ESRCH);
2111 
2112 	PROC_UNLOCK(tdt->td_proc);
2113 	cga.level = CPU_LEVEL_WHICH;
2114 	cga.which = CPU_WHICH_TID;
2115 	cga.id = tdt->td_tid;
2116 	cga.cpusetsize = sizeof(cpuset_t);
2117 	cga.mask = (cpuset_t *) args->user_mask_ptr;
2118 
2119 	if ((error = sys_cpuset_getaffinity(td, &cga)) == 0)
2120 		td->td_retval[0] = sizeof(cpuset_t);
2121 
2122 	return (error);
2123 }
2124 
2125 /*
2126  *  Set affinity of a process.
2127  */
2128 int
2129 linux_sched_setaffinity(struct thread *td,
2130     struct linux_sched_setaffinity_args *args)
2131 {
2132 	struct cpuset_setaffinity_args csa;
2133 	struct thread *tdt;
2134 
2135 #ifdef DEBUG
2136 	if (ldebug(sched_setaffinity))
2137 		printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid,
2138 		    args->len);
2139 #endif
2140 	if (args->len < sizeof(cpuset_t))
2141 		return (EINVAL);
2142 
2143 	tdt = linux_tdfind(td, args->pid, -1);
2144 	if (tdt == NULL)
2145 		return (ESRCH);
2146 
2147 	PROC_UNLOCK(tdt->td_proc);
2148 	csa.level = CPU_LEVEL_WHICH;
2149 	csa.which = CPU_WHICH_TID;
2150 	csa.id = tdt->td_tid;
2151 	csa.cpusetsize = sizeof(cpuset_t);
2152 	csa.mask = (cpuset_t *) args->user_mask_ptr;
2153 
2154 	return (sys_cpuset_setaffinity(td, &csa));
2155 }
2156 
2157 struct linux_rlimit64 {
2158 	uint64_t	rlim_cur;
2159 	uint64_t	rlim_max;
2160 };
2161 
2162 int
2163 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
2164 {
2165 	struct rlimit rlim, nrlim;
2166 	struct linux_rlimit64 lrlim;
2167 	struct proc *p;
2168 	u_int which;
2169 	int flags;
2170 	int error;
2171 
2172 #ifdef DEBUG
2173 	if (ldebug(prlimit64))
2174 		printf(ARGS(prlimit64, "%d, %d, %p, %p"), args->pid,
2175 		    args->resource, (void *)args->new, (void *)args->old);
2176 #endif
2177 
2178 	if (args->resource >= LINUX_RLIM_NLIMITS)
2179 		return (EINVAL);
2180 
2181 	which = linux_to_bsd_resource[args->resource];
2182 	if (which == -1)
2183 		return (EINVAL);
2184 
2185 	if (args->new != NULL) {
2186 		/*
2187 		 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
2188 		 * rlim is unsigned 64-bit. FreeBSD treats negative limits
2189 		 * as INFINITY so we do not need a conversion even.
2190 		 */
2191 		error = copyin(args->new, &nrlim, sizeof(nrlim));
2192 		if (error != 0)
2193 			return (error);
2194 	}
2195 
2196 	flags = PGET_HOLD | PGET_NOTWEXIT;
2197 	if (args->new != NULL)
2198 		flags |= PGET_CANDEBUG;
2199 	else
2200 		flags |= PGET_CANSEE;
2201 	error = pget(args->pid, flags, &p);
2202 	if (error != 0)
2203 		return (error);
2204 
2205 	if (args->old != NULL) {
2206 		PROC_LOCK(p);
2207 		lim_rlimit_proc(p, which, &rlim);
2208 		PROC_UNLOCK(p);
2209 		if (rlim.rlim_cur == RLIM_INFINITY)
2210 			lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2211 		else
2212 			lrlim.rlim_cur = rlim.rlim_cur;
2213 		if (rlim.rlim_max == RLIM_INFINITY)
2214 			lrlim.rlim_max = LINUX_RLIM_INFINITY;
2215 		else
2216 			lrlim.rlim_max = rlim.rlim_max;
2217 		error = copyout(&lrlim, args->old, sizeof(lrlim));
2218 		if (error != 0)
2219 			goto out;
2220 	}
2221 
2222 	if (args->new != NULL)
2223 		error = kern_proc_setrlimit(td, p, which, &nrlim);
2224 
2225  out:
2226 	PRELE(p);
2227 	return (error);
2228 }
2229 
2230 int
2231 linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2232 {
2233 	struct timeval utv, tv0, tv1, *tvp;
2234 	struct l_pselect6arg lpse6;
2235 	struct l_timespec lts;
2236 	struct timespec uts;
2237 	l_sigset_t l_ss;
2238 	sigset_t *ssp;
2239 	sigset_t ss;
2240 	int error;
2241 
2242 	ssp = NULL;
2243 	if (args->sig != NULL) {
2244 		error = copyin(args->sig, &lpse6, sizeof(lpse6));
2245 		if (error != 0)
2246 			return (error);
2247 		if (lpse6.ss_len != sizeof(l_ss))
2248 			return (EINVAL);
2249 		if (lpse6.ss != 0) {
2250 			error = copyin(PTRIN(lpse6.ss), &l_ss,
2251 			    sizeof(l_ss));
2252 			if (error != 0)
2253 				return (error);
2254 			linux_to_bsd_sigset(&l_ss, &ss);
2255 			ssp = &ss;
2256 		}
2257 	}
2258 
2259 	/*
2260 	 * Currently glibc changes nanosecond number to microsecond.
2261 	 * This mean losing precision but for now it is hardly seen.
2262 	 */
2263 	if (args->tsp != NULL) {
2264 		error = copyin(args->tsp, &lts, sizeof(lts));
2265 		if (error != 0)
2266 			return (error);
2267 		error = linux_to_native_timespec(&uts, &lts);
2268 		if (error != 0)
2269 			return (error);
2270 
2271 		TIMESPEC_TO_TIMEVAL(&utv, &uts);
2272 		if (itimerfix(&utv))
2273 			return (EINVAL);
2274 
2275 		microtime(&tv0);
2276 		tvp = &utv;
2277 	} else
2278 		tvp = NULL;
2279 
2280 	error = kern_pselect(td, args->nfds, args->readfds, args->writefds,
2281 	    args->exceptfds, tvp, ssp, LINUX_NFDBITS);
2282 
2283 	if (error == 0 && args->tsp != NULL) {
2284 		if (td->td_retval[0] != 0) {
2285 			/*
2286 			 * Compute how much time was left of the timeout,
2287 			 * by subtracting the current time and the time
2288 			 * before we started the call, and subtracting
2289 			 * that result from the user-supplied value.
2290 			 */
2291 
2292 			microtime(&tv1);
2293 			timevalsub(&tv1, &tv0);
2294 			timevalsub(&utv, &tv1);
2295 			if (utv.tv_sec < 0)
2296 				timevalclear(&utv);
2297 		} else
2298 			timevalclear(&utv);
2299 
2300 		TIMEVAL_TO_TIMESPEC(&utv, &uts);
2301 
2302 		native_to_linux_timespec(&lts, &uts);
2303 		error = copyout(&lts, args->tsp, sizeof(lts));
2304 	}
2305 
2306 	return (error);
2307 }
2308 
2309 int
2310 linux_ppoll(struct thread *td, struct linux_ppoll_args *args)
2311 {
2312 	struct timespec ts0, ts1;
2313 	struct l_timespec lts;
2314 	struct timespec uts, *tsp;
2315 	l_sigset_t l_ss;
2316 	sigset_t *ssp;
2317 	sigset_t ss;
2318 	int error;
2319 
2320 	if (args->sset != NULL) {
2321 		if (args->ssize != sizeof(l_ss))
2322 			return (EINVAL);
2323 		error = copyin(args->sset, &l_ss, sizeof(l_ss));
2324 		if (error)
2325 			return (error);
2326 		linux_to_bsd_sigset(&l_ss, &ss);
2327 		ssp = &ss;
2328 	} else
2329 		ssp = NULL;
2330 	if (args->tsp != NULL) {
2331 		error = copyin(args->tsp, &lts, sizeof(lts));
2332 		if (error)
2333 			return (error);
2334 		error = linux_to_native_timespec(&uts, &lts);
2335 		if (error != 0)
2336 			return (error);
2337 
2338 		nanotime(&ts0);
2339 		tsp = &uts;
2340 	} else
2341 		tsp = NULL;
2342 
2343 	error = kern_poll(td, args->fds, args->nfds, tsp, ssp);
2344 
2345 	if (error == 0 && args->tsp != NULL) {
2346 		if (td->td_retval[0]) {
2347 			nanotime(&ts1);
2348 			timespecsub(&ts1, &ts0);
2349 			timespecsub(&uts, &ts1);
2350 			if (uts.tv_sec < 0)
2351 				timespecclear(&uts);
2352 		} else
2353 			timespecclear(&uts);
2354 
2355 		native_to_linux_timespec(&lts, &uts);
2356 		error = copyout(&lts, args->tsp, sizeof(lts));
2357 	}
2358 
2359 	return (error);
2360 }
2361 
2362 #if defined(DEBUG) || defined(KTR)
2363 /* XXX: can be removed when every ldebug(...) and KTR stuff are removed. */
2364 
2365 #ifdef COMPAT_LINUX32
2366 #define	L_MAXSYSCALL	LINUX32_SYS_MAXSYSCALL
2367 #else
2368 #define	L_MAXSYSCALL	LINUX_SYS_MAXSYSCALL
2369 #endif
2370 
2371 u_char linux_debug_map[howmany(L_MAXSYSCALL, sizeof(u_char))];
2372 
2373 static int
2374 linux_debug(int syscall, int toggle, int global)
2375 {
2376 
2377 	if (global) {
2378 		char c = toggle ? 0 : 0xff;
2379 
2380 		memset(linux_debug_map, c, sizeof(linux_debug_map));
2381 		return (0);
2382 	}
2383 	if (syscall < 0 || syscall >= L_MAXSYSCALL)
2384 		return (EINVAL);
2385 	if (toggle)
2386 		clrbit(linux_debug_map, syscall);
2387 	else
2388 		setbit(linux_debug_map, syscall);
2389 	return (0);
2390 }
2391 #undef L_MAXSYSCALL
2392 
2393 /*
2394  * Usage: sysctl linux.debug=<syscall_nr>.<0/1>
2395  *
2396  *    E.g.: sysctl linux.debug=21.0
2397  *
2398  * As a special case, syscall "all" will apply to all syscalls globally.
2399  */
2400 #define LINUX_MAX_DEBUGSTR	16
2401 int
2402 linux_sysctl_debug(SYSCTL_HANDLER_ARGS)
2403 {
2404 	char value[LINUX_MAX_DEBUGSTR], *p;
2405 	int error, sysc, toggle;
2406 	int global = 0;
2407 
2408 	value[0] = '\0';
2409 	error = sysctl_handle_string(oidp, value, LINUX_MAX_DEBUGSTR, req);
2410 	if (error || req->newptr == NULL)
2411 		return (error);
2412 	for (p = value; *p != '\0' && *p != '.'; p++);
2413 	if (*p == '\0')
2414 		return (EINVAL);
2415 	*p++ = '\0';
2416 	sysc = strtol(value, NULL, 0);
2417 	toggle = strtol(p, NULL, 0);
2418 	if (strcmp(value, "all") == 0)
2419 		global = 1;
2420 	error = linux_debug(sysc, toggle, global);
2421 	return (error);
2422 }
2423 
2424 #endif /* DEBUG || KTR */
2425 
2426 int
2427 linux_sched_rr_get_interval(struct thread *td,
2428     struct linux_sched_rr_get_interval_args *uap)
2429 {
2430 	struct timespec ts;
2431 	struct l_timespec lts;
2432 	struct thread *tdt;
2433 	int error;
2434 
2435 	/*
2436 	 * According to man in case the invalid pid specified
2437 	 * EINVAL should be returned.
2438 	 */
2439 	if (uap->pid < 0)
2440 		return (EINVAL);
2441 
2442 	tdt = linux_tdfind(td, uap->pid, -1);
2443 	if (tdt == NULL)
2444 		return (ESRCH);
2445 
2446 	error = kern_sched_rr_get_interval_td(td, tdt, &ts);
2447 	PROC_UNLOCK(tdt->td_proc);
2448 	if (error != 0)
2449 		return (error);
2450 	native_to_linux_timespec(&lts, &ts);
2451 	return (copyout(&lts, uap->interval, sizeof(lts)));
2452 }
2453 
2454 /*
2455  * In case when the Linux thread is the initial thread in
2456  * the thread group thread id is equal to the process id.
2457  * Glibc depends on this magic (assert in pthread_getattr_np.c).
2458  */
2459 struct thread *
2460 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid)
2461 {
2462 	struct linux_emuldata *em;
2463 	struct thread *tdt;
2464 	struct proc *p;
2465 
2466 	tdt = NULL;
2467 	if (tid == 0 || tid == td->td_tid) {
2468 		tdt = td;
2469 		PROC_LOCK(tdt->td_proc);
2470 	} else if (tid > PID_MAX)
2471 		tdt = tdfind(tid, pid);
2472 	else {
2473 		/*
2474 		 * Initial thread where the tid equal to the pid.
2475 		 */
2476 		p = pfind(tid);
2477 		if (p != NULL) {
2478 			if (SV_PROC_ABI(p) != SV_ABI_LINUX) {
2479 				/*
2480 				 * p is not a Linuxulator process.
2481 				 */
2482 				PROC_UNLOCK(p);
2483 				return (NULL);
2484 			}
2485 			FOREACH_THREAD_IN_PROC(p, tdt) {
2486 				em = em_find(tdt);
2487 				if (tid == em->em_tid)
2488 					return (tdt);
2489 			}
2490 			PROC_UNLOCK(p);
2491 		}
2492 		return (NULL);
2493 	}
2494 
2495 	return (tdt);
2496 }
2497 
2498 void
2499 linux_to_bsd_waitopts(int options, int *bsdopts)
2500 {
2501 
2502 	if (options & LINUX_WNOHANG)
2503 		*bsdopts |= WNOHANG;
2504 	if (options & LINUX_WUNTRACED)
2505 		*bsdopts |= WUNTRACED;
2506 	if (options & LINUX_WEXITED)
2507 		*bsdopts |= WEXITED;
2508 	if (options & LINUX_WCONTINUED)
2509 		*bsdopts |= WCONTINUED;
2510 	if (options & LINUX_WNOWAIT)
2511 		*bsdopts |= WNOWAIT;
2512 
2513 	if (options & __WCLONE)
2514 		*bsdopts |= WLINUXCLONE;
2515 }
2516