xref: /freebsd/sys/compat/linux/linux_misc.c (revision 09d325677d53a12c79a43664ff29871e92247629)
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 #include "opt_kdtrace.h"
35 
36 #include <sys/param.h>
37 #include <sys/blist.h>
38 #include <sys/fcntl.h>
39 #if defined(__i386__)
40 #include <sys/imgact_aout.h>
41 #endif
42 #include <sys/jail.h>
43 #include <sys/kernel.h>
44 #include <sys/limits.h>
45 #include <sys/lock.h>
46 #include <sys/malloc.h>
47 #include <sys/mman.h>
48 #include <sys/mount.h>
49 #include <sys/mutex.h>
50 #include <sys/namei.h>
51 #include <sys/priv.h>
52 #include <sys/proc.h>
53 #include <sys/reboot.h>
54 #include <sys/racct.h>
55 #include <sys/resourcevar.h>
56 #include <sys/sched.h>
57 #include <sys/sdt.h>
58 #include <sys/signalvar.h>
59 #include <sys/stat.h>
60 #include <sys/syscallsubr.h>
61 #include <sys/sysctl.h>
62 #include <sys/sysproto.h>
63 #include <sys/systm.h>
64 #include <sys/time.h>
65 #include <sys/vmmeter.h>
66 #include <sys/vnode.h>
67 #include <sys/wait.h>
68 #include <sys/cpuset.h>
69 
70 #include <security/mac/mac_framework.h>
71 
72 #include <vm/vm.h>
73 #include <vm/pmap.h>
74 #include <vm/vm_kern.h>
75 #include <vm/vm_map.h>
76 #include <vm/vm_extern.h>
77 #include <vm/vm_object.h>
78 #include <vm/swap_pager.h>
79 
80 #ifdef COMPAT_LINUX32
81 #include <machine/../linux32/linux.h>
82 #include <machine/../linux32/linux32_proto.h>
83 #else
84 #include <machine/../linux/linux.h>
85 #include <machine/../linux/linux_proto.h>
86 #endif
87 
88 #include <compat/linux/linux_dtrace.h>
89 #include <compat/linux/linux_file.h>
90 #include <compat/linux/linux_mib.h>
91 #include <compat/linux/linux_signal.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 /* DTrace init */
98 LIN_SDT_PROVIDER_DECLARE(LINUX_DTRACE);
99 
100 /* Linuxulator-global DTrace probes */
101 LIN_SDT_PROBE_DECLARE(locks, emul_lock, locked);
102 LIN_SDT_PROBE_DECLARE(locks, emul_lock, unlock);
103 LIN_SDT_PROBE_DECLARE(locks, emul_shared_rlock, locked);
104 LIN_SDT_PROBE_DECLARE(locks, emul_shared_rlock, unlock);
105 LIN_SDT_PROBE_DECLARE(locks, emul_shared_wlock, locked);
106 LIN_SDT_PROBE_DECLARE(locks, emul_shared_wlock, unlock);
107 
108 int stclohz;				/* Statistics clock frequency */
109 
110 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
111 	RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
112 	RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
113 	RLIMIT_MEMLOCK, RLIMIT_AS
114 };
115 
116 struct l_sysinfo {
117 	l_long		uptime;		/* Seconds since boot */
118 	l_ulong		loads[3];	/* 1, 5, and 15 minute load averages */
119 #define LINUX_SYSINFO_LOADS_SCALE 65536
120 	l_ulong		totalram;	/* Total usable main memory size */
121 	l_ulong		freeram;	/* Available memory size */
122 	l_ulong		sharedram;	/* Amount of shared memory */
123 	l_ulong		bufferram;	/* Memory used by buffers */
124 	l_ulong		totalswap;	/* Total swap space size */
125 	l_ulong		freeswap;	/* swap space still available */
126 	l_ushort	procs;		/* Number of current processes */
127 	l_ushort	pads;
128 	l_ulong		totalbig;
129 	l_ulong		freebig;
130 	l_uint		mem_unit;
131 	char		_f[20-2*sizeof(l_long)-sizeof(l_int)];	/* padding */
132 };
133 int
134 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
135 {
136 	struct l_sysinfo sysinfo;
137 	vm_object_t object;
138 	int i, j;
139 	struct timespec ts;
140 
141 	getnanouptime(&ts);
142 	if (ts.tv_nsec != 0)
143 		ts.tv_sec++;
144 	sysinfo.uptime = ts.tv_sec;
145 
146 	/* Use the information from the mib to get our load averages */
147 	for (i = 0; i < 3; i++)
148 		sysinfo.loads[i] = averunnable.ldavg[i] *
149 		    LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
150 
151 	sysinfo.totalram = physmem * PAGE_SIZE;
152 	sysinfo.freeram = sysinfo.totalram - cnt.v_wire_count * PAGE_SIZE;
153 
154 	sysinfo.sharedram = 0;
155 	mtx_lock(&vm_object_list_mtx);
156 	TAILQ_FOREACH(object, &vm_object_list, object_list)
157 		if (object->shadow_count > 1)
158 			sysinfo.sharedram += object->resident_page_count;
159 	mtx_unlock(&vm_object_list_mtx);
160 
161 	sysinfo.sharedram *= PAGE_SIZE;
162 	sysinfo.bufferram = 0;
163 
164 	swap_pager_status(&i, &j);
165 	sysinfo.totalswap = i * PAGE_SIZE;
166 	sysinfo.freeswap = (i - j) * PAGE_SIZE;
167 
168 	sysinfo.procs = nprocs;
169 
170 	/* The following are only present in newer Linux kernels. */
171 	sysinfo.totalbig = 0;
172 	sysinfo.freebig = 0;
173 	sysinfo.mem_unit = 1;
174 
175 	return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
176 }
177 
178 int
179 linux_alarm(struct thread *td, struct linux_alarm_args *args)
180 {
181 	struct itimerval it, old_it;
182 	u_int secs;
183 	int error;
184 
185 #ifdef DEBUG
186 	if (ldebug(alarm))
187 		printf(ARGS(alarm, "%u"), args->secs);
188 #endif
189 
190 	secs = args->secs;
191 
192 	if (secs > INT_MAX)
193 		secs = INT_MAX;
194 
195 	it.it_value.tv_sec = (long) secs;
196 	it.it_value.tv_usec = 0;
197 	it.it_interval.tv_sec = 0;
198 	it.it_interval.tv_usec = 0;
199 	error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
200 	if (error)
201 		return (error);
202 	if (timevalisset(&old_it.it_value)) {
203 		if (old_it.it_value.tv_usec != 0)
204 			old_it.it_value.tv_sec++;
205 		td->td_retval[0] = old_it.it_value.tv_sec;
206 	}
207 	return (0);
208 }
209 
210 int
211 linux_brk(struct thread *td, struct linux_brk_args *args)
212 {
213 	struct vmspace *vm = td->td_proc->p_vmspace;
214 	vm_offset_t new, old;
215 	struct obreak_args /* {
216 		char * nsize;
217 	} */ tmp;
218 
219 #ifdef DEBUG
220 	if (ldebug(brk))
221 		printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
222 #endif
223 	old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize);
224 	new = (vm_offset_t)args->dsend;
225 	tmp.nsize = (char *)new;
226 	if (((caddr_t)new > vm->vm_daddr) && !sys_obreak(td, &tmp))
227 		td->td_retval[0] = (long)new;
228 	else
229 		td->td_retval[0] = (long)old;
230 
231 	return (0);
232 }
233 
234 #if defined(__i386__)
235 /* XXX: what about amd64/linux32? */
236 
237 int
238 linux_uselib(struct thread *td, struct linux_uselib_args *args)
239 {
240 	struct nameidata ni;
241 	struct vnode *vp;
242 	struct exec *a_out;
243 	struct vattr attr;
244 	vm_offset_t vmaddr;
245 	unsigned long file_offset;
246 	unsigned long bss_size;
247 	char *library;
248 	ssize_t aresid;
249 	int error, locked, writecount;
250 
251 	LCONVPATHEXIST(td, args->library, &library);
252 
253 #ifdef DEBUG
254 	if (ldebug(uselib))
255 		printf(ARGS(uselib, "%s"), library);
256 #endif
257 
258 	a_out = NULL;
259 	locked = 0;
260 	vp = NULL;
261 
262 	NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
263 	    UIO_SYSSPACE, library, td);
264 	error = namei(&ni);
265 	LFREEPATH(library);
266 	if (error)
267 		goto cleanup;
268 
269 	vp = ni.ni_vp;
270 	NDFREE(&ni, NDF_ONLY_PNBUF);
271 
272 	/*
273 	 * From here on down, we have a locked vnode that must be unlocked.
274 	 * XXX: The code below largely duplicates exec_check_permissions().
275 	 */
276 	locked = 1;
277 
278 	/* Writable? */
279 	error = VOP_GET_WRITECOUNT(vp, &writecount);
280 	if (error != 0)
281 		goto cleanup;
282 	if (writecount != 0) {
283 		error = ETXTBSY;
284 		goto cleanup;
285 	}
286 
287 	/* Executable? */
288 	error = VOP_GETATTR(vp, &attr, td->td_ucred);
289 	if (error)
290 		goto cleanup;
291 
292 	if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
293 	    ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
294 		/* EACCESS is what exec(2) returns. */
295 		error = ENOEXEC;
296 		goto cleanup;
297 	}
298 
299 	/* Sensible size? */
300 	if (attr.va_size == 0) {
301 		error = ENOEXEC;
302 		goto cleanup;
303 	}
304 
305 	/* Can we access it? */
306 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
307 	if (error)
308 		goto cleanup;
309 
310 	/*
311 	 * XXX: This should use vn_open() so that it is properly authorized,
312 	 * and to reduce code redundancy all over the place here.
313 	 * XXX: Not really, it duplicates far more of exec_check_permissions()
314 	 * than vn_open().
315 	 */
316 #ifdef MAC
317 	error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
318 	if (error)
319 		goto cleanup;
320 #endif
321 	error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
322 	if (error)
323 		goto cleanup;
324 
325 	/* Pull in executable header into exec_map */
326 	error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
327 	    VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
328 	if (error)
329 		goto cleanup;
330 
331 	/* Is it a Linux binary ? */
332 	if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
333 		error = ENOEXEC;
334 		goto cleanup;
335 	}
336 
337 	/*
338 	 * While we are here, we should REALLY do some more checks
339 	 */
340 
341 	/* Set file/virtual offset based on a.out variant. */
342 	switch ((int)(a_out->a_magic & 0xffff)) {
343 	case 0413:			/* ZMAGIC */
344 		file_offset = 1024;
345 		break;
346 	case 0314:			/* QMAGIC */
347 		file_offset = 0;
348 		break;
349 	default:
350 		error = ENOEXEC;
351 		goto cleanup;
352 	}
353 
354 	bss_size = round_page(a_out->a_bss);
355 
356 	/* Check various fields in header for validity/bounds. */
357 	if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
358 		error = ENOEXEC;
359 		goto cleanup;
360 	}
361 
362 	/* text + data can't exceed file size */
363 	if (a_out->a_data + a_out->a_text > attr.va_size) {
364 		error = EFAULT;
365 		goto cleanup;
366 	}
367 
368 	/*
369 	 * text/data/bss must not exceed limits
370 	 * XXX - this is not complete. it should check current usage PLUS
371 	 * the resources needed by this library.
372 	 */
373 	PROC_LOCK(td->td_proc);
374 	if (a_out->a_text > maxtsiz ||
375 	    a_out->a_data + bss_size > lim_cur(td->td_proc, RLIMIT_DATA) ||
376 	    racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
377 	    bss_size) != 0) {
378 		PROC_UNLOCK(td->td_proc);
379 		error = ENOMEM;
380 		goto cleanup;
381 	}
382 	PROC_UNLOCK(td->td_proc);
383 
384 	/*
385 	 * Prevent more writers.
386 	 * XXX: Note that if any of the VM operations fail below we don't
387 	 * clear this flag.
388 	 */
389 	VOP_SET_TEXT(vp);
390 
391 	/*
392 	 * Lock no longer needed
393 	 */
394 	locked = 0;
395 	VOP_UNLOCK(vp, 0);
396 
397 	/*
398 	 * Check if file_offset page aligned. Currently we cannot handle
399 	 * misalinged file offsets, and so we read in the entire image
400 	 * (what a waste).
401 	 */
402 	if (file_offset & PAGE_MASK) {
403 #ifdef DEBUG
404 		printf("uselib: Non page aligned binary %lu\n", file_offset);
405 #endif
406 		/* Map text+data read/write/execute */
407 
408 		/* a_entry is the load address and is page aligned */
409 		vmaddr = trunc_page(a_out->a_entry);
410 
411 		/* get anon user mapping, read+write+execute */
412 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
413 		    &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
414 		    VM_PROT_ALL, VM_PROT_ALL, 0);
415 		if (error)
416 			goto cleanup;
417 
418 		error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
419 		    a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
420 		    td->td_ucred, NOCRED, &aresid, td);
421 		if (error != 0)
422 			goto cleanup;
423 		if (aresid != 0) {
424 			error = ENOEXEC;
425 			goto cleanup;
426 		}
427 	} else {
428 #ifdef DEBUG
429 		printf("uselib: Page aligned binary %lu\n", file_offset);
430 #endif
431 		/*
432 		 * for QMAGIC, a_entry is 20 bytes beyond the load address
433 		 * to skip the executable header
434 		 */
435 		vmaddr = trunc_page(a_out->a_entry);
436 
437 		/*
438 		 * Map it all into the process's space as a single
439 		 * copy-on-write "data" segment.
440 		 */
441 		error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr,
442 		    a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
443 		    MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
444 		if (error)
445 			goto cleanup;
446 	}
447 #ifdef DEBUG
448 	printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
449 	    ((long *)vmaddr)[1]);
450 #endif
451 	if (bss_size != 0) {
452 		/* Calculate BSS start address */
453 		vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
454 		    a_out->a_data;
455 
456 		/* allocate some 'anon' space */
457 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
458 		    &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
459 		    VM_PROT_ALL, 0);
460 		if (error)
461 			goto cleanup;
462 	}
463 
464 cleanup:
465 	/* Unlock vnode if needed */
466 	if (locked)
467 		VOP_UNLOCK(vp, 0);
468 
469 	/* Release the temporary mapping. */
470 	if (a_out)
471 		kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
472 
473 	return (error);
474 }
475 
476 #endif	/* __i386__ */
477 
478 int
479 linux_select(struct thread *td, struct linux_select_args *args)
480 {
481 	l_timeval ltv;
482 	struct timeval tv0, tv1, utv, *tvp;
483 	int error;
484 
485 #ifdef DEBUG
486 	if (ldebug(select))
487 		printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
488 		    (void *)args->readfds, (void *)args->writefds,
489 		    (void *)args->exceptfds, (void *)args->timeout);
490 #endif
491 
492 	/*
493 	 * Store current time for computation of the amount of
494 	 * time left.
495 	 */
496 	if (args->timeout) {
497 		if ((error = copyin(args->timeout, &ltv, sizeof(ltv))))
498 			goto select_out;
499 		utv.tv_sec = ltv.tv_sec;
500 		utv.tv_usec = ltv.tv_usec;
501 #ifdef DEBUG
502 		if (ldebug(select))
503 			printf(LMSG("incoming timeout (%jd/%ld)"),
504 			    (intmax_t)utv.tv_sec, utv.tv_usec);
505 #endif
506 
507 		if (itimerfix(&utv)) {
508 			/*
509 			 * The timeval was invalid.  Convert it to something
510 			 * valid that will act as it does under Linux.
511 			 */
512 			utv.tv_sec += utv.tv_usec / 1000000;
513 			utv.tv_usec %= 1000000;
514 			if (utv.tv_usec < 0) {
515 				utv.tv_sec -= 1;
516 				utv.tv_usec += 1000000;
517 			}
518 			if (utv.tv_sec < 0)
519 				timevalclear(&utv);
520 		}
521 		microtime(&tv0);
522 		tvp = &utv;
523 	} else
524 		tvp = NULL;
525 
526 	error = kern_select(td, args->nfds, args->readfds, args->writefds,
527 	    args->exceptfds, tvp, sizeof(l_int) * 8);
528 
529 #ifdef DEBUG
530 	if (ldebug(select))
531 		printf(LMSG("real select returns %d"), error);
532 #endif
533 	if (error)
534 		goto select_out;
535 
536 	if (args->timeout) {
537 		if (td->td_retval[0]) {
538 			/*
539 			 * Compute how much time was left of the timeout,
540 			 * by subtracting the current time and the time
541 			 * before we started the call, and subtracting
542 			 * that result from the user-supplied value.
543 			 */
544 			microtime(&tv1);
545 			timevalsub(&tv1, &tv0);
546 			timevalsub(&utv, &tv1);
547 			if (utv.tv_sec < 0)
548 				timevalclear(&utv);
549 		} else
550 			timevalclear(&utv);
551 #ifdef DEBUG
552 		if (ldebug(select))
553 			printf(LMSG("outgoing timeout (%jd/%ld)"),
554 			    (intmax_t)utv.tv_sec, utv.tv_usec);
555 #endif
556 		ltv.tv_sec = utv.tv_sec;
557 		ltv.tv_usec = utv.tv_usec;
558 		if ((error = copyout(&ltv, args->timeout, sizeof(ltv))))
559 			goto select_out;
560 	}
561 
562 select_out:
563 #ifdef DEBUG
564 	if (ldebug(select))
565 		printf(LMSG("select_out -> %d"), error);
566 #endif
567 	return (error);
568 }
569 
570 int
571 linux_mremap(struct thread *td, struct linux_mremap_args *args)
572 {
573 	struct munmap_args /* {
574 		void *addr;
575 		size_t len;
576 	} */ bsd_args;
577 	int error = 0;
578 
579 #ifdef DEBUG
580 	if (ldebug(mremap))
581 		printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
582 		    (void *)(uintptr_t)args->addr,
583 		    (unsigned long)args->old_len,
584 		    (unsigned long)args->new_len,
585 		    (unsigned long)args->flags);
586 #endif
587 
588 	if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
589 		td->td_retval[0] = 0;
590 		return (EINVAL);
591 	}
592 
593 	/*
594 	 * Check for the page alignment.
595 	 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
596 	 */
597 	if (args->addr & PAGE_MASK) {
598 		td->td_retval[0] = 0;
599 		return (EINVAL);
600 	}
601 
602 	args->new_len = round_page(args->new_len);
603 	args->old_len = round_page(args->old_len);
604 
605 	if (args->new_len > args->old_len) {
606 		td->td_retval[0] = 0;
607 		return (ENOMEM);
608 	}
609 
610 	if (args->new_len < args->old_len) {
611 		bsd_args.addr =
612 		    (caddr_t)((uintptr_t)args->addr + args->new_len);
613 		bsd_args.len = args->old_len - args->new_len;
614 		error = sys_munmap(td, &bsd_args);
615 	}
616 
617 	td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
618 	return (error);
619 }
620 
621 #define LINUX_MS_ASYNC       0x0001
622 #define LINUX_MS_INVALIDATE  0x0002
623 #define LINUX_MS_SYNC        0x0004
624 
625 int
626 linux_msync(struct thread *td, struct linux_msync_args *args)
627 {
628 	struct msync_args bsd_args;
629 
630 	bsd_args.addr = (caddr_t)(uintptr_t)args->addr;
631 	bsd_args.len = (uintptr_t)args->len;
632 	bsd_args.flags = args->fl & ~LINUX_MS_SYNC;
633 
634 	return (sys_msync(td, &bsd_args));
635 }
636 
637 int
638 linux_time(struct thread *td, struct linux_time_args *args)
639 {
640 	struct timeval tv;
641 	l_time_t tm;
642 	int error;
643 
644 #ifdef DEBUG
645 	if (ldebug(time))
646 		printf(ARGS(time, "*"));
647 #endif
648 
649 	microtime(&tv);
650 	tm = tv.tv_sec;
651 	if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
652 		return (error);
653 	td->td_retval[0] = tm;
654 	return (0);
655 }
656 
657 struct l_times_argv {
658 	l_clock_t	tms_utime;
659 	l_clock_t	tms_stime;
660 	l_clock_t	tms_cutime;
661 	l_clock_t	tms_cstime;
662 };
663 
664 
665 /*
666  * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
667  * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
668  * auxiliary vector entry.
669  */
670 #define	CLK_TCK		100
671 
672 #define	CONVOTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
673 #define	CONVNTCK(r)	(r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
674 
675 #define	CONVTCK(r)	(linux_kernver(td) >= LINUX_KERNVER_2004000 ?		\
676 			    CONVNTCK(r) : CONVOTCK(r))
677 
678 int
679 linux_times(struct thread *td, struct linux_times_args *args)
680 {
681 	struct timeval tv, utime, stime, cutime, cstime;
682 	struct l_times_argv tms;
683 	struct proc *p;
684 	int error;
685 
686 #ifdef DEBUG
687 	if (ldebug(times))
688 		printf(ARGS(times, "*"));
689 #endif
690 
691 	if (args->buf != NULL) {
692 		p = td->td_proc;
693 		PROC_LOCK(p);
694 		PROC_SLOCK(p);
695 		calcru(p, &utime, &stime);
696 		PROC_SUNLOCK(p);
697 		calccru(p, &cutime, &cstime);
698 		PROC_UNLOCK(p);
699 
700 		tms.tms_utime = CONVTCK(utime);
701 		tms.tms_stime = CONVTCK(stime);
702 
703 		tms.tms_cutime = CONVTCK(cutime);
704 		tms.tms_cstime = CONVTCK(cstime);
705 
706 		if ((error = copyout(&tms, args->buf, sizeof(tms))))
707 			return (error);
708 	}
709 
710 	microuptime(&tv);
711 	td->td_retval[0] = (int)CONVTCK(tv);
712 	return (0);
713 }
714 
715 int
716 linux_newuname(struct thread *td, struct linux_newuname_args *args)
717 {
718 	struct l_new_utsname utsname;
719 	char osname[LINUX_MAX_UTSNAME];
720 	char osrelease[LINUX_MAX_UTSNAME];
721 	char *p;
722 
723 #ifdef DEBUG
724 	if (ldebug(newuname))
725 		printf(ARGS(newuname, "*"));
726 #endif
727 
728 	linux_get_osname(td, osname);
729 	linux_get_osrelease(td, osrelease);
730 
731 	bzero(&utsname, sizeof(utsname));
732 	strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
733 	getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
734 	getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
735 	strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
736 	strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
737 	for (p = utsname.version; *p != '\0'; ++p)
738 		if (*p == '\n') {
739 			*p = '\0';
740 			break;
741 		}
742 	strlcpy(utsname.machine, linux_platform, LINUX_MAX_UTSNAME);
743 
744 	return (copyout(&utsname, args->buf, sizeof(utsname)));
745 }
746 
747 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
748 struct l_utimbuf {
749 	l_time_t l_actime;
750 	l_time_t l_modtime;
751 };
752 
753 int
754 linux_utime(struct thread *td, struct linux_utime_args *args)
755 {
756 	struct timeval tv[2], *tvp;
757 	struct l_utimbuf lut;
758 	char *fname;
759 	int error;
760 
761 	LCONVPATHEXIST(td, args->fname, &fname);
762 
763 #ifdef DEBUG
764 	if (ldebug(utime))
765 		printf(ARGS(utime, "%s, *"), fname);
766 #endif
767 
768 	if (args->times) {
769 		if ((error = copyin(args->times, &lut, sizeof lut))) {
770 			LFREEPATH(fname);
771 			return (error);
772 		}
773 		tv[0].tv_sec = lut.l_actime;
774 		tv[0].tv_usec = 0;
775 		tv[1].tv_sec = lut.l_modtime;
776 		tv[1].tv_usec = 0;
777 		tvp = tv;
778 	} else
779 		tvp = NULL;
780 
781 	error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
782 	LFREEPATH(fname);
783 	return (error);
784 }
785 
786 int
787 linux_utimes(struct thread *td, struct linux_utimes_args *args)
788 {
789 	l_timeval ltv[2];
790 	struct timeval tv[2], *tvp = NULL;
791 	char *fname;
792 	int error;
793 
794 	LCONVPATHEXIST(td, args->fname, &fname);
795 
796 #ifdef DEBUG
797 	if (ldebug(utimes))
798 		printf(ARGS(utimes, "%s, *"), fname);
799 #endif
800 
801 	if (args->tptr != NULL) {
802 		if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
803 			LFREEPATH(fname);
804 			return (error);
805 		}
806 		tv[0].tv_sec = ltv[0].tv_sec;
807 		tv[0].tv_usec = ltv[0].tv_usec;
808 		tv[1].tv_sec = ltv[1].tv_sec;
809 		tv[1].tv_usec = ltv[1].tv_usec;
810 		tvp = tv;
811 	}
812 
813 	error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
814 	LFREEPATH(fname);
815 	return (error);
816 }
817 
818 int
819 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
820 {
821 	l_timeval ltv[2];
822 	struct timeval tv[2], *tvp = NULL;
823 	char *fname;
824 	int error, dfd;
825 
826 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
827 	LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
828 
829 #ifdef DEBUG
830 	if (ldebug(futimesat))
831 		printf(ARGS(futimesat, "%s, *"), fname);
832 #endif
833 
834 	if (args->utimes != NULL) {
835 		if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
836 			LFREEPATH(fname);
837 			return (error);
838 		}
839 		tv[0].tv_sec = ltv[0].tv_sec;
840 		tv[0].tv_usec = ltv[0].tv_usec;
841 		tv[1].tv_sec = ltv[1].tv_sec;
842 		tv[1].tv_usec = ltv[1].tv_usec;
843 		tvp = tv;
844 	}
845 
846 	error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
847 	LFREEPATH(fname);
848 	return (error);
849 }
850 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
851 
852 int
853 linux_common_wait(struct thread *td, int pid, int *status,
854     int options, struct rusage *ru)
855 {
856 	int error, tmpstat;
857 
858 	error = kern_wait(td, pid, &tmpstat, options, ru);
859 	if (error)
860 		return (error);
861 
862 	if (status) {
863 		tmpstat &= 0xffff;
864 		if (WIFSIGNALED(tmpstat))
865 			tmpstat = (tmpstat & 0xffffff80) |
866 			    BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat));
867 		else if (WIFSTOPPED(tmpstat))
868 			tmpstat = (tmpstat & 0xffff00ff) |
869 			    (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8);
870 		error = copyout(&tmpstat, status, sizeof(int));
871 	}
872 
873 	return (error);
874 }
875 
876 int
877 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
878 {
879 	int options;
880 
881 #ifdef DEBUG
882 	if (ldebug(waitpid))
883 		printf(ARGS(waitpid, "%d, %p, %d"),
884 		    args->pid, (void *)args->status, args->options);
885 #endif
886 	/*
887 	 * this is necessary because the test in kern_wait doesn't work
888 	 * because we mess with the options here
889 	 */
890 	if (args->options & ~(WUNTRACED | WNOHANG | WCONTINUED | __WCLONE))
891 		return (EINVAL);
892 
893 	options = (args->options & (WNOHANG | WUNTRACED));
894 	/* WLINUXCLONE should be equal to __WCLONE, but we make sure */
895 	if (args->options & __WCLONE)
896 		options |= WLINUXCLONE;
897 
898 	return (linux_common_wait(td, args->pid, args->status, options, NULL));
899 }
900 
901 
902 int
903 linux_mknod(struct thread *td, struct linux_mknod_args *args)
904 {
905 	char *path;
906 	int error;
907 
908 	LCONVPATHCREAT(td, args->path, &path);
909 
910 #ifdef DEBUG
911 	if (ldebug(mknod))
912 		printf(ARGS(mknod, "%s, %d, %d"), path, args->mode, args->dev);
913 #endif
914 
915 	switch (args->mode & S_IFMT) {
916 	case S_IFIFO:
917 	case S_IFSOCK:
918 		error = kern_mkfifo(td, path, UIO_SYSSPACE, args->mode);
919 		break;
920 
921 	case S_IFCHR:
922 	case S_IFBLK:
923 		error = kern_mknod(td, path, UIO_SYSSPACE, args->mode,
924 		    args->dev);
925 		break;
926 
927 	case S_IFDIR:
928 		error = EPERM;
929 		break;
930 
931 	case 0:
932 		args->mode |= S_IFREG;
933 		/* FALLTHROUGH */
934 	case S_IFREG:
935 		error = kern_open(td, path, UIO_SYSSPACE,
936 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
937 		if (error == 0)
938 			kern_close(td, td->td_retval[0]);
939 		break;
940 
941 	default:
942 		error = EINVAL;
943 		break;
944 	}
945 	LFREEPATH(path);
946 	return (error);
947 }
948 
949 int
950 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
951 {
952 	char *path;
953 	int error, dfd;
954 
955 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
956 	LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
957 
958 #ifdef DEBUG
959 	if (ldebug(mknodat))
960 		printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev);
961 #endif
962 
963 	switch (args->mode & S_IFMT) {
964 	case S_IFIFO:
965 	case S_IFSOCK:
966 		error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
967 		break;
968 
969 	case S_IFCHR:
970 	case S_IFBLK:
971 		error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
972 		    args->dev);
973 		break;
974 
975 	case S_IFDIR:
976 		error = EPERM;
977 		break;
978 
979 	case 0:
980 		args->mode |= S_IFREG;
981 		/* FALLTHROUGH */
982 	case S_IFREG:
983 		error = kern_openat(td, dfd, path, UIO_SYSSPACE,
984 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
985 		if (error == 0)
986 			kern_close(td, td->td_retval[0]);
987 		break;
988 
989 	default:
990 		error = EINVAL;
991 		break;
992 	}
993 	LFREEPATH(path);
994 	return (error);
995 }
996 
997 /*
998  * UGH! This is just about the dumbest idea I've ever heard!!
999  */
1000 int
1001 linux_personality(struct thread *td, struct linux_personality_args *args)
1002 {
1003 #ifdef DEBUG
1004 	if (ldebug(personality))
1005 		printf(ARGS(personality, "%lu"), (unsigned long)args->per);
1006 #endif
1007 	if (args->per != 0)
1008 		return (EINVAL);
1009 
1010 	/* Yes Jim, it's still a Linux... */
1011 	td->td_retval[0] = 0;
1012 	return (0);
1013 }
1014 
1015 struct l_itimerval {
1016 	l_timeval it_interval;
1017 	l_timeval it_value;
1018 };
1019 
1020 #define	B2L_ITIMERVAL(bip, lip) 					\
1021 	(bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec;		\
1022 	(bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec;	\
1023 	(bip)->it_value.tv_sec = (lip)->it_value.tv_sec;		\
1024 	(bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1025 
1026 int
1027 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1028 {
1029 	int error;
1030 	struct l_itimerval ls;
1031 	struct itimerval aitv, oitv;
1032 
1033 #ifdef DEBUG
1034 	if (ldebug(setitimer))
1035 		printf(ARGS(setitimer, "%p, %p"),
1036 		    (void *)uap->itv, (void *)uap->oitv);
1037 #endif
1038 
1039 	if (uap->itv == NULL) {
1040 		uap->itv = uap->oitv;
1041 		return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1042 	}
1043 
1044 	error = copyin(uap->itv, &ls, sizeof(ls));
1045 	if (error != 0)
1046 		return (error);
1047 	B2L_ITIMERVAL(&aitv, &ls);
1048 #ifdef DEBUG
1049 	if (ldebug(setitimer)) {
1050 		printf("setitimer: value: sec: %jd, usec: %ld\n",
1051 		    (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1052 		printf("setitimer: interval: sec: %jd, usec: %ld\n",
1053 		    (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1054 	}
1055 #endif
1056 	error = kern_setitimer(td, uap->which, &aitv, &oitv);
1057 	if (error != 0 || uap->oitv == NULL)
1058 		return (error);
1059 	B2L_ITIMERVAL(&ls, &oitv);
1060 
1061 	return (copyout(&ls, uap->oitv, sizeof(ls)));
1062 }
1063 
1064 int
1065 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1066 {
1067 	int error;
1068 	struct l_itimerval ls;
1069 	struct itimerval aitv;
1070 
1071 #ifdef DEBUG
1072 	if (ldebug(getitimer))
1073 		printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1074 #endif
1075 	error = kern_getitimer(td, uap->which, &aitv);
1076 	if (error != 0)
1077 		return (error);
1078 	B2L_ITIMERVAL(&ls, &aitv);
1079 	return (copyout(&ls, uap->itv, sizeof(ls)));
1080 }
1081 
1082 int
1083 linux_nice(struct thread *td, struct linux_nice_args *args)
1084 {
1085 	struct setpriority_args bsd_args;
1086 
1087 	bsd_args.which = PRIO_PROCESS;
1088 	bsd_args.who = 0;		/* current process */
1089 	bsd_args.prio = args->inc;
1090 	return (sys_setpriority(td, &bsd_args));
1091 }
1092 
1093 int
1094 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1095 {
1096 	struct ucred *newcred, *oldcred;
1097 	l_gid_t *linux_gidset;
1098 	gid_t *bsd_gidset;
1099 	int ngrp, error;
1100 	struct proc *p;
1101 
1102 	ngrp = args->gidsetsize;
1103 	if (ngrp < 0 || ngrp >= ngroups_max + 1)
1104 		return (EINVAL);
1105 	linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_TEMP, M_WAITOK);
1106 	error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1107 	if (error)
1108 		goto out;
1109 	newcred = crget();
1110 	p = td->td_proc;
1111 	PROC_LOCK(p);
1112 	oldcred = crcopysafe(p, newcred);
1113 
1114 	/*
1115 	 * cr_groups[0] holds egid. Setting the whole set from
1116 	 * the supplied set will cause egid to be changed too.
1117 	 * Keep cr_groups[0] unchanged to prevent that.
1118 	 */
1119 
1120 	if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) {
1121 		PROC_UNLOCK(p);
1122 		crfree(newcred);
1123 		goto out;
1124 	}
1125 
1126 	if (ngrp > 0) {
1127 		newcred->cr_ngroups = ngrp + 1;
1128 
1129 		bsd_gidset = newcred->cr_groups;
1130 		ngrp--;
1131 		while (ngrp >= 0) {
1132 			bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1133 			ngrp--;
1134 		}
1135 	} else
1136 		newcred->cr_ngroups = 1;
1137 
1138 	setsugid(p);
1139 	p->p_ucred = newcred;
1140 	PROC_UNLOCK(p);
1141 	crfree(oldcred);
1142 	error = 0;
1143 out:
1144 	free(linux_gidset, M_TEMP);
1145 	return (error);
1146 }
1147 
1148 int
1149 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1150 {
1151 	struct ucred *cred;
1152 	l_gid_t *linux_gidset;
1153 	gid_t *bsd_gidset;
1154 	int bsd_gidsetsz, ngrp, error;
1155 
1156 	cred = td->td_ucred;
1157 	bsd_gidset = cred->cr_groups;
1158 	bsd_gidsetsz = cred->cr_ngroups - 1;
1159 
1160 	/*
1161 	 * cr_groups[0] holds egid. Returning the whole set
1162 	 * here will cause a duplicate. Exclude cr_groups[0]
1163 	 * to prevent that.
1164 	 */
1165 
1166 	if ((ngrp = args->gidsetsize) == 0) {
1167 		td->td_retval[0] = bsd_gidsetsz;
1168 		return (0);
1169 	}
1170 
1171 	if (ngrp < bsd_gidsetsz)
1172 		return (EINVAL);
1173 
1174 	ngrp = 0;
1175 	linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1176 	    M_TEMP, M_WAITOK);
1177 	while (ngrp < bsd_gidsetsz) {
1178 		linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1179 		ngrp++;
1180 	}
1181 
1182 	error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1183 	free(linux_gidset, M_TEMP);
1184 	if (error)
1185 		return (error);
1186 
1187 	td->td_retval[0] = ngrp;
1188 	return (0);
1189 }
1190 
1191 int
1192 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1193 {
1194 	struct rlimit bsd_rlim;
1195 	struct l_rlimit rlim;
1196 	u_int which;
1197 	int error;
1198 
1199 #ifdef DEBUG
1200 	if (ldebug(setrlimit))
1201 		printf(ARGS(setrlimit, "%d, %p"),
1202 		    args->resource, (void *)args->rlim);
1203 #endif
1204 
1205 	if (args->resource >= LINUX_RLIM_NLIMITS)
1206 		return (EINVAL);
1207 
1208 	which = linux_to_bsd_resource[args->resource];
1209 	if (which == -1)
1210 		return (EINVAL);
1211 
1212 	error = copyin(args->rlim, &rlim, sizeof(rlim));
1213 	if (error)
1214 		return (error);
1215 
1216 	bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1217 	bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1218 	return (kern_setrlimit(td, which, &bsd_rlim));
1219 }
1220 
1221 int
1222 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1223 {
1224 	struct l_rlimit rlim;
1225 	struct proc *p = td->td_proc;
1226 	struct rlimit bsd_rlim;
1227 	u_int which;
1228 
1229 #ifdef DEBUG
1230 	if (ldebug(old_getrlimit))
1231 		printf(ARGS(old_getrlimit, "%d, %p"),
1232 		    args->resource, (void *)args->rlim);
1233 #endif
1234 
1235 	if (args->resource >= LINUX_RLIM_NLIMITS)
1236 		return (EINVAL);
1237 
1238 	which = linux_to_bsd_resource[args->resource];
1239 	if (which == -1)
1240 		return (EINVAL);
1241 
1242 	PROC_LOCK(p);
1243 	lim_rlimit(p, which, &bsd_rlim);
1244 	PROC_UNLOCK(p);
1245 
1246 #ifdef COMPAT_LINUX32
1247 	rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1248 	if (rlim.rlim_cur == UINT_MAX)
1249 		rlim.rlim_cur = INT_MAX;
1250 	rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1251 	if (rlim.rlim_max == UINT_MAX)
1252 		rlim.rlim_max = INT_MAX;
1253 #else
1254 	rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1255 	if (rlim.rlim_cur == ULONG_MAX)
1256 		rlim.rlim_cur = LONG_MAX;
1257 	rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1258 	if (rlim.rlim_max == ULONG_MAX)
1259 		rlim.rlim_max = LONG_MAX;
1260 #endif
1261 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1262 }
1263 
1264 int
1265 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1266 {
1267 	struct l_rlimit rlim;
1268 	struct proc *p = td->td_proc;
1269 	struct rlimit bsd_rlim;
1270 	u_int which;
1271 
1272 #ifdef DEBUG
1273 	if (ldebug(getrlimit))
1274 		printf(ARGS(getrlimit, "%d, %p"),
1275 		    args->resource, (void *)args->rlim);
1276 #endif
1277 
1278 	if (args->resource >= LINUX_RLIM_NLIMITS)
1279 		return (EINVAL);
1280 
1281 	which = linux_to_bsd_resource[args->resource];
1282 	if (which == -1)
1283 		return (EINVAL);
1284 
1285 	PROC_LOCK(p);
1286 	lim_rlimit(p, which, &bsd_rlim);
1287 	PROC_UNLOCK(p);
1288 
1289 	rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1290 	rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1291 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1292 }
1293 
1294 int
1295 linux_sched_setscheduler(struct thread *td,
1296     struct linux_sched_setscheduler_args *args)
1297 {
1298 	struct sched_setscheduler_args bsd;
1299 
1300 #ifdef DEBUG
1301 	if (ldebug(sched_setscheduler))
1302 		printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1303 		    args->pid, args->policy, (const void *)args->param);
1304 #endif
1305 
1306 	switch (args->policy) {
1307 	case LINUX_SCHED_OTHER:
1308 		bsd.policy = SCHED_OTHER;
1309 		break;
1310 	case LINUX_SCHED_FIFO:
1311 		bsd.policy = SCHED_FIFO;
1312 		break;
1313 	case LINUX_SCHED_RR:
1314 		bsd.policy = SCHED_RR;
1315 		break;
1316 	default:
1317 		return (EINVAL);
1318 	}
1319 
1320 	bsd.pid = args->pid;
1321 	bsd.param = (struct sched_param *)args->param;
1322 	return (sys_sched_setscheduler(td, &bsd));
1323 }
1324 
1325 int
1326 linux_sched_getscheduler(struct thread *td,
1327     struct linux_sched_getscheduler_args *args)
1328 {
1329 	struct sched_getscheduler_args bsd;
1330 	int error;
1331 
1332 #ifdef DEBUG
1333 	if (ldebug(sched_getscheduler))
1334 		printf(ARGS(sched_getscheduler, "%d"), args->pid);
1335 #endif
1336 
1337 	bsd.pid = args->pid;
1338 	error = sys_sched_getscheduler(td, &bsd);
1339 
1340 	switch (td->td_retval[0]) {
1341 	case SCHED_OTHER:
1342 		td->td_retval[0] = LINUX_SCHED_OTHER;
1343 		break;
1344 	case SCHED_FIFO:
1345 		td->td_retval[0] = LINUX_SCHED_FIFO;
1346 		break;
1347 	case SCHED_RR:
1348 		td->td_retval[0] = LINUX_SCHED_RR;
1349 		break;
1350 	}
1351 
1352 	return (error);
1353 }
1354 
1355 int
1356 linux_sched_get_priority_max(struct thread *td,
1357     struct linux_sched_get_priority_max_args *args)
1358 {
1359 	struct sched_get_priority_max_args bsd;
1360 
1361 #ifdef DEBUG
1362 	if (ldebug(sched_get_priority_max))
1363 		printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1364 #endif
1365 
1366 	switch (args->policy) {
1367 	case LINUX_SCHED_OTHER:
1368 		bsd.policy = SCHED_OTHER;
1369 		break;
1370 	case LINUX_SCHED_FIFO:
1371 		bsd.policy = SCHED_FIFO;
1372 		break;
1373 	case LINUX_SCHED_RR:
1374 		bsd.policy = SCHED_RR;
1375 		break;
1376 	default:
1377 		return (EINVAL);
1378 	}
1379 	return (sys_sched_get_priority_max(td, &bsd));
1380 }
1381 
1382 int
1383 linux_sched_get_priority_min(struct thread *td,
1384     struct linux_sched_get_priority_min_args *args)
1385 {
1386 	struct sched_get_priority_min_args bsd;
1387 
1388 #ifdef DEBUG
1389 	if (ldebug(sched_get_priority_min))
1390 		printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1391 #endif
1392 
1393 	switch (args->policy) {
1394 	case LINUX_SCHED_OTHER:
1395 		bsd.policy = SCHED_OTHER;
1396 		break;
1397 	case LINUX_SCHED_FIFO:
1398 		bsd.policy = SCHED_FIFO;
1399 		break;
1400 	case LINUX_SCHED_RR:
1401 		bsd.policy = SCHED_RR;
1402 		break;
1403 	default:
1404 		return (EINVAL);
1405 	}
1406 	return (sys_sched_get_priority_min(td, &bsd));
1407 }
1408 
1409 #define REBOOT_CAD_ON	0x89abcdef
1410 #define REBOOT_CAD_OFF	0
1411 #define REBOOT_HALT	0xcdef0123
1412 #define REBOOT_RESTART	0x01234567
1413 #define REBOOT_RESTART2	0xA1B2C3D4
1414 #define REBOOT_POWEROFF	0x4321FEDC
1415 #define REBOOT_MAGIC1	0xfee1dead
1416 #define REBOOT_MAGIC2	0x28121969
1417 #define REBOOT_MAGIC2A	0x05121996
1418 #define REBOOT_MAGIC2B	0x16041998
1419 
1420 int
1421 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1422 {
1423 	struct reboot_args bsd_args;
1424 
1425 #ifdef DEBUG
1426 	if (ldebug(reboot))
1427 		printf(ARGS(reboot, "0x%x"), args->cmd);
1428 #endif
1429 
1430 	if (args->magic1 != REBOOT_MAGIC1)
1431 		return (EINVAL);
1432 
1433 	switch (args->magic2) {
1434 	case REBOOT_MAGIC2:
1435 	case REBOOT_MAGIC2A:
1436 	case REBOOT_MAGIC2B:
1437 		break;
1438 	default:
1439 		return (EINVAL);
1440 	}
1441 
1442 	switch (args->cmd) {
1443 	case REBOOT_CAD_ON:
1444 	case REBOOT_CAD_OFF:
1445 		return (priv_check(td, PRIV_REBOOT));
1446 	case REBOOT_HALT:
1447 		bsd_args.opt = RB_HALT;
1448 		break;
1449 	case REBOOT_RESTART:
1450 	case REBOOT_RESTART2:
1451 		bsd_args.opt = 0;
1452 		break;
1453 	case REBOOT_POWEROFF:
1454 		bsd_args.opt = RB_POWEROFF;
1455 		break;
1456 	default:
1457 		return (EINVAL);
1458 	}
1459 	return (sys_reboot(td, &bsd_args));
1460 }
1461 
1462 
1463 /*
1464  * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1465  * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
1466  * are assumed to be preserved. The following lightweight syscalls fixes
1467  * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
1468  *
1469  * linux_getpid() - MP SAFE
1470  * linux_getgid() - MP SAFE
1471  * linux_getuid() - MP SAFE
1472  */
1473 
1474 int
1475 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1476 {
1477 	struct linux_emuldata *em;
1478 
1479 #ifdef DEBUG
1480 	if (ldebug(getpid))
1481 		printf(ARGS(getpid, ""));
1482 #endif
1483 
1484 	if (linux_use26(td)) {
1485 		em = em_find(td->td_proc, EMUL_DONTLOCK);
1486 		KASSERT(em != NULL, ("getpid: emuldata not found.\n"));
1487 		td->td_retval[0] = em->shared->group_pid;
1488 	} else {
1489 		td->td_retval[0] = td->td_proc->p_pid;
1490 	}
1491 
1492 	return (0);
1493 }
1494 
1495 int
1496 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1497 {
1498 
1499 #ifdef DEBUG
1500 	if (ldebug(gettid))
1501 		printf(ARGS(gettid, ""));
1502 #endif
1503 
1504 	td->td_retval[0] = td->td_proc->p_pid;
1505 	return (0);
1506 }
1507 
1508 
1509 int
1510 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1511 {
1512 	struct linux_emuldata *em;
1513 	struct proc *p, *pp;
1514 
1515 #ifdef DEBUG
1516 	if (ldebug(getppid))
1517 		printf(ARGS(getppid, ""));
1518 #endif
1519 
1520 	if (!linux_use26(td)) {
1521 		PROC_LOCK(td->td_proc);
1522 		td->td_retval[0] = td->td_proc->p_pptr->p_pid;
1523 		PROC_UNLOCK(td->td_proc);
1524 		return (0);
1525 	}
1526 
1527 	em = em_find(td->td_proc, EMUL_DONTLOCK);
1528 
1529 	KASSERT(em != NULL, ("getppid: process emuldata not found.\n"));
1530 
1531 	/* find the group leader */
1532 	p = pfind(em->shared->group_pid);
1533 
1534 	if (p == NULL) {
1535 #ifdef DEBUG
1536 	   	printf(LMSG("parent process not found.\n"));
1537 #endif
1538 		return (0);
1539 	}
1540 
1541 	pp = p->p_pptr;		/* switch to parent */
1542 	PROC_LOCK(pp);
1543 	PROC_UNLOCK(p);
1544 
1545 	/* if its also linux process */
1546 	if (pp->p_sysent == &elf_linux_sysvec) {
1547 		em = em_find(pp, EMUL_DONTLOCK);
1548 		KASSERT(em != NULL, ("getppid: parent emuldata not found.\n"));
1549 
1550 		td->td_retval[0] = em->shared->group_pid;
1551 	} else
1552 		td->td_retval[0] = pp->p_pid;
1553 
1554 	PROC_UNLOCK(pp);
1555 
1556 	return (0);
1557 }
1558 
1559 int
1560 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1561 {
1562 
1563 #ifdef DEBUG
1564 	if (ldebug(getgid))
1565 		printf(ARGS(getgid, ""));
1566 #endif
1567 
1568 	td->td_retval[0] = td->td_ucred->cr_rgid;
1569 	return (0);
1570 }
1571 
1572 int
1573 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1574 {
1575 
1576 #ifdef DEBUG
1577 	if (ldebug(getuid))
1578 		printf(ARGS(getuid, ""));
1579 #endif
1580 
1581 	td->td_retval[0] = td->td_ucred->cr_ruid;
1582 	return (0);
1583 }
1584 
1585 
1586 int
1587 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1588 {
1589 	struct getsid_args bsd;
1590 
1591 #ifdef DEBUG
1592 	if (ldebug(getsid))
1593 		printf(ARGS(getsid, "%i"), args->pid);
1594 #endif
1595 
1596 	bsd.pid = args->pid;
1597 	return (sys_getsid(td, &bsd));
1598 }
1599 
1600 int
1601 linux_nosys(struct thread *td, struct nosys_args *ignore)
1602 {
1603 
1604 	return (ENOSYS);
1605 }
1606 
1607 int
1608 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1609 {
1610 	struct getpriority_args bsd_args;
1611 	int error;
1612 
1613 #ifdef DEBUG
1614 	if (ldebug(getpriority))
1615 		printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
1616 #endif
1617 
1618 	bsd_args.which = args->which;
1619 	bsd_args.who = args->who;
1620 	error = sys_getpriority(td, &bsd_args);
1621 	td->td_retval[0] = 20 - td->td_retval[0];
1622 	return (error);
1623 }
1624 
1625 int
1626 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1627 {
1628 	int name[2];
1629 
1630 #ifdef DEBUG
1631 	if (ldebug(sethostname))
1632 		printf(ARGS(sethostname, "*, %i"), args->len);
1633 #endif
1634 
1635 	name[0] = CTL_KERN;
1636 	name[1] = KERN_HOSTNAME;
1637 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1638 	    args->len, 0, 0));
1639 }
1640 
1641 int
1642 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1643 {
1644 	int name[2];
1645 
1646 #ifdef DEBUG
1647 	if (ldebug(setdomainname))
1648 		printf(ARGS(setdomainname, "*, %i"), args->len);
1649 #endif
1650 
1651 	name[0] = CTL_KERN;
1652 	name[1] = KERN_NISDOMAINNAME;
1653 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1654 	    args->len, 0, 0));
1655 }
1656 
1657 int
1658 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1659 {
1660 	struct linux_emuldata *em;
1661 
1662 #ifdef DEBUG
1663 	if (ldebug(exit_group))
1664 		printf(ARGS(exit_group, "%i"), args->error_code);
1665 #endif
1666 
1667 	em = em_find(td->td_proc, EMUL_DONTLOCK);
1668 	if (em->shared->refs > 1) {
1669 		EMUL_SHARED_WLOCK(&emul_shared_lock);
1670 		em->shared->flags |= EMUL_SHARED_HASXSTAT;
1671 		em->shared->xstat = W_EXITCODE(args->error_code, 0);
1672 		EMUL_SHARED_WUNLOCK(&emul_shared_lock);
1673 		if (linux_use26(td))
1674 			linux_kill_threads(td, SIGKILL);
1675 	}
1676 
1677 	/*
1678 	 * XXX: we should send a signal to the parent if
1679 	 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1680 	 * as it doesnt occur often.
1681 	 */
1682 	exit1(td, W_EXITCODE(args->error_code, 0));
1683 
1684 	return (0);
1685 }
1686 
1687 #define _LINUX_CAPABILITY_VERSION  0x19980330
1688 
1689 struct l_user_cap_header {
1690 	l_int	version;
1691 	l_int	pid;
1692 };
1693 
1694 struct l_user_cap_data {
1695 	l_int	effective;
1696 	l_int	permitted;
1697 	l_int	inheritable;
1698 };
1699 
1700 int
1701 linux_capget(struct thread *td, struct linux_capget_args *args)
1702 {
1703 	struct l_user_cap_header luch;
1704 	struct l_user_cap_data lucd;
1705 	int error;
1706 
1707 	if (args->hdrp == NULL)
1708 		return (EFAULT);
1709 
1710 	error = copyin(args->hdrp, &luch, sizeof(luch));
1711 	if (error != 0)
1712 		return (error);
1713 
1714 	if (luch.version != _LINUX_CAPABILITY_VERSION) {
1715 		luch.version = _LINUX_CAPABILITY_VERSION;
1716 		error = copyout(&luch, args->hdrp, sizeof(luch));
1717 		if (error)
1718 			return (error);
1719 		return (EINVAL);
1720 	}
1721 
1722 	if (luch.pid)
1723 		return (EPERM);
1724 
1725 	if (args->datap) {
1726 		/*
1727 		 * The current implementation doesn't support setting
1728 		 * a capability (it's essentially a stub) so indicate
1729 		 * that no capabilities are currently set or available
1730 		 * to request.
1731 		 */
1732 		bzero (&lucd, sizeof(lucd));
1733 		error = copyout(&lucd, args->datap, sizeof(lucd));
1734 	}
1735 
1736 	return (error);
1737 }
1738 
1739 int
1740 linux_capset(struct thread *td, struct linux_capset_args *args)
1741 {
1742 	struct l_user_cap_header luch;
1743 	struct l_user_cap_data lucd;
1744 	int error;
1745 
1746 	if (args->hdrp == NULL || args->datap == NULL)
1747 		return (EFAULT);
1748 
1749 	error = copyin(args->hdrp, &luch, sizeof(luch));
1750 	if (error != 0)
1751 		return (error);
1752 
1753 	if (luch.version != _LINUX_CAPABILITY_VERSION) {
1754 		luch.version = _LINUX_CAPABILITY_VERSION;
1755 		error = copyout(&luch, args->hdrp, sizeof(luch));
1756 		if (error)
1757 			return (error);
1758 		return (EINVAL);
1759 	}
1760 
1761 	if (luch.pid)
1762 		return (EPERM);
1763 
1764 	error = copyin(args->datap, &lucd, sizeof(lucd));
1765 	if (error != 0)
1766 		return (error);
1767 
1768 	/* We currently don't support setting any capabilities. */
1769 	if (lucd.effective || lucd.permitted || lucd.inheritable) {
1770 		linux_msg(td,
1771 			  "capset effective=0x%x, permitted=0x%x, "
1772 			  "inheritable=0x%x is not implemented",
1773 			  (int)lucd.effective, (int)lucd.permitted,
1774 			  (int)lucd.inheritable);
1775 		return (EPERM);
1776 	}
1777 
1778 	return (0);
1779 }
1780 
1781 int
1782 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1783 {
1784 	int error = 0, max_size;
1785 	struct proc *p = td->td_proc;
1786 	char comm[LINUX_MAX_COMM_LEN];
1787 	struct linux_emuldata *em;
1788 	int pdeath_signal;
1789 
1790 #ifdef DEBUG
1791 	if (ldebug(prctl))
1792 		printf(ARGS(prctl, "%d, %d, %d, %d, %d"), args->option,
1793 		    args->arg2, args->arg3, args->arg4, args->arg5);
1794 #endif
1795 
1796 	switch (args->option) {
1797 	case LINUX_PR_SET_PDEATHSIG:
1798 		if (!LINUX_SIG_VALID(args->arg2))
1799 			return (EINVAL);
1800 		em = em_find(p, EMUL_DOLOCK);
1801 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1802 		em->pdeath_signal = args->arg2;
1803 		EMUL_UNLOCK(&emul_lock);
1804 		break;
1805 	case LINUX_PR_GET_PDEATHSIG:
1806 		em = em_find(p, EMUL_DOLOCK);
1807 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1808 		pdeath_signal = em->pdeath_signal;
1809 		EMUL_UNLOCK(&emul_lock);
1810 		error = copyout(&pdeath_signal,
1811 		    (void *)(register_t)args->arg2,
1812 		    sizeof(pdeath_signal));
1813 		break;
1814 	case LINUX_PR_GET_KEEPCAPS:
1815 		/*
1816 		 * Indicate that we always clear the effective and
1817 		 * permitted capability sets when the user id becomes
1818 		 * non-zero (actually the capability sets are simply
1819 		 * always zero in the current implementation).
1820 		 */
1821 		td->td_retval[0] = 0;
1822 		break;
1823 	case LINUX_PR_SET_KEEPCAPS:
1824 		/*
1825 		 * Ignore requests to keep the effective and permitted
1826 		 * capability sets when the user id becomes non-zero.
1827 		 */
1828 		break;
1829 	case LINUX_PR_SET_NAME:
1830 		/*
1831 		 * To be on the safe side we need to make sure to not
1832 		 * overflow the size a linux program expects. We already
1833 		 * do this here in the copyin, so that we don't need to
1834 		 * check on copyout.
1835 		 */
1836 		max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1837 		error = copyinstr((void *)(register_t)args->arg2, comm,
1838 		    max_size, NULL);
1839 
1840 		/* Linux silently truncates the name if it is too long. */
1841 		if (error == ENAMETOOLONG) {
1842 			/*
1843 			 * XXX: copyinstr() isn't documented to populate the
1844 			 * array completely, so do a copyin() to be on the
1845 			 * safe side. This should be changed in case
1846 			 * copyinstr() is changed to guarantee this.
1847 			 */
1848 			error = copyin((void *)(register_t)args->arg2, comm,
1849 			    max_size - 1);
1850 			comm[max_size - 1] = '\0';
1851 		}
1852 		if (error)
1853 			return (error);
1854 
1855 		PROC_LOCK(p);
1856 		strlcpy(p->p_comm, comm, sizeof(p->p_comm));
1857 		PROC_UNLOCK(p);
1858 		break;
1859 	case LINUX_PR_GET_NAME:
1860 		PROC_LOCK(p);
1861 		strlcpy(comm, p->p_comm, sizeof(comm));
1862 		PROC_UNLOCK(p);
1863 		error = copyout(comm, (void *)(register_t)args->arg2,
1864 		    strlen(comm) + 1);
1865 		break;
1866 	default:
1867 		error = EINVAL;
1868 		break;
1869 	}
1870 
1871 	return (error);
1872 }
1873 
1874 /*
1875  * Get affinity of a process.
1876  */
1877 int
1878 linux_sched_getaffinity(struct thread *td,
1879     struct linux_sched_getaffinity_args *args)
1880 {
1881 	int error;
1882 	struct cpuset_getaffinity_args cga;
1883 
1884 #ifdef DEBUG
1885 	if (ldebug(sched_getaffinity))
1886 		printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid,
1887 		    args->len);
1888 #endif
1889 	if (args->len < sizeof(cpuset_t))
1890 		return (EINVAL);
1891 
1892 	cga.level = CPU_LEVEL_WHICH;
1893 	cga.which = CPU_WHICH_PID;
1894 	cga.id = args->pid;
1895 	cga.cpusetsize = sizeof(cpuset_t);
1896 	cga.mask = (cpuset_t *) args->user_mask_ptr;
1897 
1898 	if ((error = sys_cpuset_getaffinity(td, &cga)) == 0)
1899 		td->td_retval[0] = sizeof(cpuset_t);
1900 
1901 	return (error);
1902 }
1903 
1904 /*
1905  *  Set affinity of a process.
1906  */
1907 int
1908 linux_sched_setaffinity(struct thread *td,
1909     struct linux_sched_setaffinity_args *args)
1910 {
1911 	struct cpuset_setaffinity_args csa;
1912 
1913 #ifdef DEBUG
1914 	if (ldebug(sched_setaffinity))
1915 		printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid,
1916 		    args->len);
1917 #endif
1918 	if (args->len < sizeof(cpuset_t))
1919 		return (EINVAL);
1920 
1921 	csa.level = CPU_LEVEL_WHICH;
1922 	csa.which = CPU_WHICH_PID;
1923 	csa.id = args->pid;
1924 	csa.cpusetsize = sizeof(cpuset_t);
1925 	csa.mask = (cpuset_t *) args->user_mask_ptr;
1926 
1927 	return (sys_cpuset_setaffinity(td, &csa));
1928 }
1929