xref: /freebsd/sys/compat/linux/linux_misc.c (revision ff0ba87247820afbdfdc1b307c803f7923d0e4d3)
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_util.h>
92 #include <compat/linux/linux_sysproto.h>
93 #include <compat/linux/linux_emul.h>
94 #include <compat/linux/linux_misc.h>
95 
96 /* DTrace init */
97 LIN_SDT_PROVIDER_DECLARE(LINUX_DTRACE);
98 
99 /* Linuxulator-global DTrace probes */
100 LIN_SDT_PROBE_DECLARE(locks, emul_lock, locked);
101 LIN_SDT_PROBE_DECLARE(locks, emul_lock, unlock);
102 LIN_SDT_PROBE_DECLARE(locks, emul_shared_rlock, locked);
103 LIN_SDT_PROBE_DECLARE(locks, emul_shared_rlock, unlock);
104 LIN_SDT_PROBE_DECLARE(locks, emul_shared_wlock, locked);
105 LIN_SDT_PROBE_DECLARE(locks, emul_shared_wlock, unlock);
106 
107 int stclohz;				/* Statistics clock frequency */
108 
109 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
110 	RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
111 	RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
112 	RLIMIT_MEMLOCK, RLIMIT_AS
113 };
114 
115 struct l_sysinfo {
116 	l_long		uptime;		/* Seconds since boot */
117 	l_ulong		loads[3];	/* 1, 5, and 15 minute load averages */
118 #define LINUX_SYSINFO_LOADS_SCALE 65536
119 	l_ulong		totalram;	/* Total usable main memory size */
120 	l_ulong		freeram;	/* Available memory size */
121 	l_ulong		sharedram;	/* Amount of shared memory */
122 	l_ulong		bufferram;	/* Memory used by buffers */
123 	l_ulong		totalswap;	/* Total swap space size */
124 	l_ulong		freeswap;	/* swap space still available */
125 	l_ushort	procs;		/* Number of current processes */
126 	l_ushort	pads;
127 	l_ulong		totalbig;
128 	l_ulong		freebig;
129 	l_uint		mem_unit;
130 	char		_f[20-2*sizeof(l_long)-sizeof(l_int)];	/* padding */
131 };
132 int
133 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
134 {
135 	struct l_sysinfo sysinfo;
136 	vm_object_t object;
137 	int i, j;
138 	struct timespec ts;
139 
140 	getnanouptime(&ts);
141 	if (ts.tv_nsec != 0)
142 		ts.tv_sec++;
143 	sysinfo.uptime = ts.tv_sec;
144 
145 	/* Use the information from the mib to get our load averages */
146 	for (i = 0; i < 3; i++)
147 		sysinfo.loads[i] = averunnable.ldavg[i] *
148 		    LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
149 
150 	sysinfo.totalram = physmem * PAGE_SIZE;
151 	sysinfo.freeram = sysinfo.totalram - vm_cnt.v_wire_count * PAGE_SIZE;
152 
153 	sysinfo.sharedram = 0;
154 	mtx_lock(&vm_object_list_mtx);
155 	TAILQ_FOREACH(object, &vm_object_list, object_list)
156 		if (object->shadow_count > 1)
157 			sysinfo.sharedram += object->resident_page_count;
158 	mtx_unlock(&vm_object_list_mtx);
159 
160 	sysinfo.sharedram *= PAGE_SIZE;
161 	sysinfo.bufferram = 0;
162 
163 	swap_pager_status(&i, &j);
164 	sysinfo.totalswap = i * PAGE_SIZE;
165 	sysinfo.freeswap = (i - j) * PAGE_SIZE;
166 
167 	sysinfo.procs = nprocs;
168 
169 	/* The following are only present in newer Linux kernels. */
170 	sysinfo.totalbig = 0;
171 	sysinfo.freebig = 0;
172 	sysinfo.mem_unit = 1;
173 
174 	return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
175 }
176 
177 int
178 linux_alarm(struct thread *td, struct linux_alarm_args *args)
179 {
180 	struct itimerval it, old_it;
181 	u_int secs;
182 	int error;
183 
184 #ifdef DEBUG
185 	if (ldebug(alarm))
186 		printf(ARGS(alarm, "%u"), args->secs);
187 #endif
188 
189 	secs = args->secs;
190 
191 	if (secs > INT_MAX)
192 		secs = INT_MAX;
193 
194 	it.it_value.tv_sec = (long) secs;
195 	it.it_value.tv_usec = 0;
196 	it.it_interval.tv_sec = 0;
197 	it.it_interval.tv_usec = 0;
198 	error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
199 	if (error)
200 		return (error);
201 	if (timevalisset(&old_it.it_value)) {
202 		if (old_it.it_value.tv_usec != 0)
203 			old_it.it_value.tv_sec++;
204 		td->td_retval[0] = old_it.it_value.tv_sec;
205 	}
206 	return (0);
207 }
208 
209 int
210 linux_brk(struct thread *td, struct linux_brk_args *args)
211 {
212 	struct vmspace *vm = td->td_proc->p_vmspace;
213 	vm_offset_t new, old;
214 	struct obreak_args /* {
215 		char * nsize;
216 	} */ tmp;
217 
218 #ifdef DEBUG
219 	if (ldebug(brk))
220 		printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
221 #endif
222 	old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize);
223 	new = (vm_offset_t)args->dsend;
224 	tmp.nsize = (char *)new;
225 	if (((caddr_t)new > vm->vm_daddr) && !sys_obreak(td, &tmp))
226 		td->td_retval[0] = (long)new;
227 	else
228 		td->td_retval[0] = (long)old;
229 
230 	return (0);
231 }
232 
233 #if defined(__i386__)
234 /* XXX: what about amd64/linux32? */
235 
236 int
237 linux_uselib(struct thread *td, struct linux_uselib_args *args)
238 {
239 	struct nameidata ni;
240 	struct vnode *vp;
241 	struct exec *a_out;
242 	struct vattr attr;
243 	vm_offset_t vmaddr;
244 	unsigned long file_offset;
245 	unsigned long bss_size;
246 	char *library;
247 	ssize_t aresid;
248 	int error, locked, writecount;
249 
250 	LCONVPATHEXIST(td, args->library, &library);
251 
252 #ifdef DEBUG
253 	if (ldebug(uselib))
254 		printf(ARGS(uselib, "%s"), library);
255 #endif
256 
257 	a_out = NULL;
258 	locked = 0;
259 	vp = NULL;
260 
261 	NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
262 	    UIO_SYSSPACE, library, td);
263 	error = namei(&ni);
264 	LFREEPATH(library);
265 	if (error)
266 		goto cleanup;
267 
268 	vp = ni.ni_vp;
269 	NDFREE(&ni, NDF_ONLY_PNBUF);
270 
271 	/*
272 	 * From here on down, we have a locked vnode that must be unlocked.
273 	 * XXX: The code below largely duplicates exec_check_permissions().
274 	 */
275 	locked = 1;
276 
277 	/* Writable? */
278 	error = VOP_GET_WRITECOUNT(vp, &writecount);
279 	if (error != 0)
280 		goto cleanup;
281 	if (writecount != 0) {
282 		error = ETXTBSY;
283 		goto cleanup;
284 	}
285 
286 	/* Executable? */
287 	error = VOP_GETATTR(vp, &attr, td->td_ucred);
288 	if (error)
289 		goto cleanup;
290 
291 	if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
292 	    ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
293 		/* EACCESS is what exec(2) returns. */
294 		error = ENOEXEC;
295 		goto cleanup;
296 	}
297 
298 	/* Sensible size? */
299 	if (attr.va_size == 0) {
300 		error = ENOEXEC;
301 		goto cleanup;
302 	}
303 
304 	/* Can we access it? */
305 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
306 	if (error)
307 		goto cleanup;
308 
309 	/*
310 	 * XXX: This should use vn_open() so that it is properly authorized,
311 	 * and to reduce code redundancy all over the place here.
312 	 * XXX: Not really, it duplicates far more of exec_check_permissions()
313 	 * than vn_open().
314 	 */
315 #ifdef MAC
316 	error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
317 	if (error)
318 		goto cleanup;
319 #endif
320 	error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
321 	if (error)
322 		goto cleanup;
323 
324 	/* Pull in executable header into exec_map */
325 	error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
326 	    VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
327 	if (error)
328 		goto cleanup;
329 
330 	/* Is it a Linux binary ? */
331 	if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
332 		error = ENOEXEC;
333 		goto cleanup;
334 	}
335 
336 	/*
337 	 * While we are here, we should REALLY do some more checks
338 	 */
339 
340 	/* Set file/virtual offset based on a.out variant. */
341 	switch ((int)(a_out->a_magic & 0xffff)) {
342 	case 0413:			/* ZMAGIC */
343 		file_offset = 1024;
344 		break;
345 	case 0314:			/* QMAGIC */
346 		file_offset = 0;
347 		break;
348 	default:
349 		error = ENOEXEC;
350 		goto cleanup;
351 	}
352 
353 	bss_size = round_page(a_out->a_bss);
354 
355 	/* Check various fields in header for validity/bounds. */
356 	if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
357 		error = ENOEXEC;
358 		goto cleanup;
359 	}
360 
361 	/* text + data can't exceed file size */
362 	if (a_out->a_data + a_out->a_text > attr.va_size) {
363 		error = EFAULT;
364 		goto cleanup;
365 	}
366 
367 	/*
368 	 * text/data/bss must not exceed limits
369 	 * XXX - this is not complete. it should check current usage PLUS
370 	 * the resources needed by this library.
371 	 */
372 	PROC_LOCK(td->td_proc);
373 	if (a_out->a_text > maxtsiz ||
374 	    a_out->a_data + bss_size > lim_cur(td->td_proc, RLIMIT_DATA) ||
375 	    racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
376 	    bss_size) != 0) {
377 		PROC_UNLOCK(td->td_proc);
378 		error = ENOMEM;
379 		goto cleanup;
380 	}
381 	PROC_UNLOCK(td->td_proc);
382 
383 	/*
384 	 * Prevent more writers.
385 	 * XXX: Note that if any of the VM operations fail below we don't
386 	 * clear this flag.
387 	 */
388 	VOP_SET_TEXT(vp);
389 
390 	/*
391 	 * Lock no longer needed
392 	 */
393 	locked = 0;
394 	VOP_UNLOCK(vp, 0);
395 
396 	/*
397 	 * Check if file_offset page aligned. Currently we cannot handle
398 	 * misalinged file offsets, and so we read in the entire image
399 	 * (what a waste).
400 	 */
401 	if (file_offset & PAGE_MASK) {
402 #ifdef DEBUG
403 		printf("uselib: Non page aligned binary %lu\n", file_offset);
404 #endif
405 		/* Map text+data read/write/execute */
406 
407 		/* a_entry is the load address and is page aligned */
408 		vmaddr = trunc_page(a_out->a_entry);
409 
410 		/* get anon user mapping, read+write+execute */
411 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
412 		    &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
413 		    VM_PROT_ALL, VM_PROT_ALL, 0);
414 		if (error)
415 			goto cleanup;
416 
417 		error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
418 		    a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
419 		    td->td_ucred, NOCRED, &aresid, td);
420 		if (error != 0)
421 			goto cleanup;
422 		if (aresid != 0) {
423 			error = ENOEXEC;
424 			goto cleanup;
425 		}
426 	} else {
427 #ifdef DEBUG
428 		printf("uselib: Page aligned binary %lu\n", file_offset);
429 #endif
430 		/*
431 		 * for QMAGIC, a_entry is 20 bytes beyond the load address
432 		 * to skip the executable header
433 		 */
434 		vmaddr = trunc_page(a_out->a_entry);
435 
436 		/*
437 		 * Map it all into the process's space as a single
438 		 * copy-on-write "data" segment.
439 		 */
440 		error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr,
441 		    a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
442 		    MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
443 		if (error)
444 			goto cleanup;
445 	}
446 #ifdef DEBUG
447 	printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
448 	    ((long *)vmaddr)[1]);
449 #endif
450 	if (bss_size != 0) {
451 		/* Calculate BSS start address */
452 		vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
453 		    a_out->a_data;
454 
455 		/* allocate some 'anon' space */
456 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
457 		    &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
458 		    VM_PROT_ALL, 0);
459 		if (error)
460 			goto cleanup;
461 	}
462 
463 cleanup:
464 	/* Unlock vnode if needed */
465 	if (locked)
466 		VOP_UNLOCK(vp, 0);
467 
468 	/* Release the temporary mapping. */
469 	if (a_out)
470 		kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
471 
472 	return (error);
473 }
474 
475 #endif	/* __i386__ */
476 
477 int
478 linux_select(struct thread *td, struct linux_select_args *args)
479 {
480 	l_timeval ltv;
481 	struct timeval tv0, tv1, utv, *tvp;
482 	int error;
483 
484 #ifdef DEBUG
485 	if (ldebug(select))
486 		printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
487 		    (void *)args->readfds, (void *)args->writefds,
488 		    (void *)args->exceptfds, (void *)args->timeout);
489 #endif
490 
491 	/*
492 	 * Store current time for computation of the amount of
493 	 * time left.
494 	 */
495 	if (args->timeout) {
496 		if ((error = copyin(args->timeout, &ltv, sizeof(ltv))))
497 			goto select_out;
498 		utv.tv_sec = ltv.tv_sec;
499 		utv.tv_usec = ltv.tv_usec;
500 #ifdef DEBUG
501 		if (ldebug(select))
502 			printf(LMSG("incoming timeout (%jd/%ld)"),
503 			    (intmax_t)utv.tv_sec, utv.tv_usec);
504 #endif
505 
506 		if (itimerfix(&utv)) {
507 			/*
508 			 * The timeval was invalid.  Convert it to something
509 			 * valid that will act as it does under Linux.
510 			 */
511 			utv.tv_sec += utv.tv_usec / 1000000;
512 			utv.tv_usec %= 1000000;
513 			if (utv.tv_usec < 0) {
514 				utv.tv_sec -= 1;
515 				utv.tv_usec += 1000000;
516 			}
517 			if (utv.tv_sec < 0)
518 				timevalclear(&utv);
519 		}
520 		microtime(&tv0);
521 		tvp = &utv;
522 	} else
523 		tvp = NULL;
524 
525 	error = kern_select(td, args->nfds, args->readfds, args->writefds,
526 	    args->exceptfds, tvp, sizeof(l_int) * 8);
527 
528 #ifdef DEBUG
529 	if (ldebug(select))
530 		printf(LMSG("real select returns %d"), error);
531 #endif
532 	if (error)
533 		goto select_out;
534 
535 	if (args->timeout) {
536 		if (td->td_retval[0]) {
537 			/*
538 			 * Compute how much time was left of the timeout,
539 			 * by subtracting the current time and the time
540 			 * before we started the call, and subtracting
541 			 * that result from the user-supplied value.
542 			 */
543 			microtime(&tv1);
544 			timevalsub(&tv1, &tv0);
545 			timevalsub(&utv, &tv1);
546 			if (utv.tv_sec < 0)
547 				timevalclear(&utv);
548 		} else
549 			timevalclear(&utv);
550 #ifdef DEBUG
551 		if (ldebug(select))
552 			printf(LMSG("outgoing timeout (%jd/%ld)"),
553 			    (intmax_t)utv.tv_sec, utv.tv_usec);
554 #endif
555 		ltv.tv_sec = utv.tv_sec;
556 		ltv.tv_usec = utv.tv_usec;
557 		if ((error = copyout(&ltv, args->timeout, sizeof(ltv))))
558 			goto select_out;
559 	}
560 
561 select_out:
562 #ifdef DEBUG
563 	if (ldebug(select))
564 		printf(LMSG("select_out -> %d"), error);
565 #endif
566 	return (error);
567 }
568 
569 int
570 linux_mremap(struct thread *td, struct linux_mremap_args *args)
571 {
572 	struct munmap_args /* {
573 		void *addr;
574 		size_t len;
575 	} */ bsd_args;
576 	int error = 0;
577 
578 #ifdef DEBUG
579 	if (ldebug(mremap))
580 		printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
581 		    (void *)(uintptr_t)args->addr,
582 		    (unsigned long)args->old_len,
583 		    (unsigned long)args->new_len,
584 		    (unsigned long)args->flags);
585 #endif
586 
587 	if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
588 		td->td_retval[0] = 0;
589 		return (EINVAL);
590 	}
591 
592 	/*
593 	 * Check for the page alignment.
594 	 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
595 	 */
596 	if (args->addr & PAGE_MASK) {
597 		td->td_retval[0] = 0;
598 		return (EINVAL);
599 	}
600 
601 	args->new_len = round_page(args->new_len);
602 	args->old_len = round_page(args->old_len);
603 
604 	if (args->new_len > args->old_len) {
605 		td->td_retval[0] = 0;
606 		return (ENOMEM);
607 	}
608 
609 	if (args->new_len < args->old_len) {
610 		bsd_args.addr =
611 		    (caddr_t)((uintptr_t)args->addr + args->new_len);
612 		bsd_args.len = args->old_len - args->new_len;
613 		error = sys_munmap(td, &bsd_args);
614 	}
615 
616 	td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
617 	return (error);
618 }
619 
620 #define LINUX_MS_ASYNC       0x0001
621 #define LINUX_MS_INVALIDATE  0x0002
622 #define LINUX_MS_SYNC        0x0004
623 
624 int
625 linux_msync(struct thread *td, struct linux_msync_args *args)
626 {
627 	struct msync_args bsd_args;
628 
629 	bsd_args.addr = (caddr_t)(uintptr_t)args->addr;
630 	bsd_args.len = (uintptr_t)args->len;
631 	bsd_args.flags = args->fl & ~LINUX_MS_SYNC;
632 
633 	return (sys_msync(td, &bsd_args));
634 }
635 
636 int
637 linux_time(struct thread *td, struct linux_time_args *args)
638 {
639 	struct timeval tv;
640 	l_time_t tm;
641 	int error;
642 
643 #ifdef DEBUG
644 	if (ldebug(time))
645 		printf(ARGS(time, "*"));
646 #endif
647 
648 	microtime(&tv);
649 	tm = tv.tv_sec;
650 	if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
651 		return (error);
652 	td->td_retval[0] = tm;
653 	return (0);
654 }
655 
656 struct l_times_argv {
657 	l_clock_t	tms_utime;
658 	l_clock_t	tms_stime;
659 	l_clock_t	tms_cutime;
660 	l_clock_t	tms_cstime;
661 };
662 
663 
664 /*
665  * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
666  * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
667  * auxiliary vector entry.
668  */
669 #define	CLK_TCK		100
670 
671 #define	CONVOTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
672 #define	CONVNTCK(r)	(r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
673 
674 #define	CONVTCK(r)	(linux_kernver(td) >= LINUX_KERNVER_2004000 ?		\
675 			    CONVNTCK(r) : CONVOTCK(r))
676 
677 int
678 linux_times(struct thread *td, struct linux_times_args *args)
679 {
680 	struct timeval tv, utime, stime, cutime, cstime;
681 	struct l_times_argv tms;
682 	struct proc *p;
683 	int error;
684 
685 #ifdef DEBUG
686 	if (ldebug(times))
687 		printf(ARGS(times, "*"));
688 #endif
689 
690 	if (args->buf != NULL) {
691 		p = td->td_proc;
692 		PROC_LOCK(p);
693 		PROC_SLOCK(p);
694 		calcru(p, &utime, &stime);
695 		PROC_SUNLOCK(p);
696 		calccru(p, &cutime, &cstime);
697 		PROC_UNLOCK(p);
698 
699 		tms.tms_utime = CONVTCK(utime);
700 		tms.tms_stime = CONVTCK(stime);
701 
702 		tms.tms_cutime = CONVTCK(cutime);
703 		tms.tms_cstime = CONVTCK(cstime);
704 
705 		if ((error = copyout(&tms, args->buf, sizeof(tms))))
706 			return (error);
707 	}
708 
709 	microuptime(&tv);
710 	td->td_retval[0] = (int)CONVTCK(tv);
711 	return (0);
712 }
713 
714 int
715 linux_newuname(struct thread *td, struct linux_newuname_args *args)
716 {
717 	struct l_new_utsname utsname;
718 	char osname[LINUX_MAX_UTSNAME];
719 	char osrelease[LINUX_MAX_UTSNAME];
720 	char *p;
721 
722 #ifdef DEBUG
723 	if (ldebug(newuname))
724 		printf(ARGS(newuname, "*"));
725 #endif
726 
727 	linux_get_osname(td, osname);
728 	linux_get_osrelease(td, osrelease);
729 
730 	bzero(&utsname, sizeof(utsname));
731 	strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
732 	getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
733 	getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
734 	strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
735 	strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
736 	for (p = utsname.version; *p != '\0'; ++p)
737 		if (*p == '\n') {
738 			*p = '\0';
739 			break;
740 		}
741 	strlcpy(utsname.machine, linux_platform, LINUX_MAX_UTSNAME);
742 
743 	return (copyout(&utsname, args->buf, sizeof(utsname)));
744 }
745 
746 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
747 struct l_utimbuf {
748 	l_time_t l_actime;
749 	l_time_t l_modtime;
750 };
751 
752 int
753 linux_utime(struct thread *td, struct linux_utime_args *args)
754 {
755 	struct timeval tv[2], *tvp;
756 	struct l_utimbuf lut;
757 	char *fname;
758 	int error;
759 
760 	LCONVPATHEXIST(td, args->fname, &fname);
761 
762 #ifdef DEBUG
763 	if (ldebug(utime))
764 		printf(ARGS(utime, "%s, *"), fname);
765 #endif
766 
767 	if (args->times) {
768 		if ((error = copyin(args->times, &lut, sizeof lut))) {
769 			LFREEPATH(fname);
770 			return (error);
771 		}
772 		tv[0].tv_sec = lut.l_actime;
773 		tv[0].tv_usec = 0;
774 		tv[1].tv_sec = lut.l_modtime;
775 		tv[1].tv_usec = 0;
776 		tvp = tv;
777 	} else
778 		tvp = NULL;
779 
780 	error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, tvp,
781 	    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_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE,
814 	    tvp, UIO_SYSSPACE);
815 	LFREEPATH(fname);
816 	return (error);
817 }
818 
819 int
820 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
821 {
822 	l_timeval ltv[2];
823 	struct timeval tv[2], *tvp = NULL;
824 	char *fname;
825 	int error, dfd;
826 
827 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
828 	LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
829 
830 #ifdef DEBUG
831 	if (ldebug(futimesat))
832 		printf(ARGS(futimesat, "%s, *"), fname);
833 #endif
834 
835 	if (args->utimes != NULL) {
836 		if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
837 			LFREEPATH(fname);
838 			return (error);
839 		}
840 		tv[0].tv_sec = ltv[0].tv_sec;
841 		tv[0].tv_usec = ltv[0].tv_usec;
842 		tv[1].tv_sec = ltv[1].tv_sec;
843 		tv[1].tv_usec = ltv[1].tv_usec;
844 		tvp = tv;
845 	}
846 
847 	error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
848 	LFREEPATH(fname);
849 	return (error);
850 }
851 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
852 
853 int
854 linux_common_wait(struct thread *td, int pid, int *status,
855     int options, struct rusage *ru)
856 {
857 	int error, tmpstat;
858 
859 	error = kern_wait(td, pid, &tmpstat, options, ru);
860 	if (error)
861 		return (error);
862 
863 	if (status) {
864 		tmpstat &= 0xffff;
865 		if (WIFSIGNALED(tmpstat))
866 			tmpstat = (tmpstat & 0xffffff80) |
867 			    BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat));
868 		else if (WIFSTOPPED(tmpstat))
869 			tmpstat = (tmpstat & 0xffff00ff) |
870 			    (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8);
871 		error = copyout(&tmpstat, status, sizeof(int));
872 	}
873 
874 	return (error);
875 }
876 
877 int
878 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
879 {
880 	int options;
881 
882 #ifdef DEBUG
883 	if (ldebug(waitpid))
884 		printf(ARGS(waitpid, "%d, %p, %d"),
885 		    args->pid, (void *)args->status, args->options);
886 #endif
887 	/*
888 	 * this is necessary because the test in kern_wait doesn't work
889 	 * because we mess with the options here
890 	 */
891 	if (args->options & ~(WUNTRACED | WNOHANG | WCONTINUED | __WCLONE))
892 		return (EINVAL);
893 
894 	options = (args->options & (WNOHANG | WUNTRACED));
895 	/* WLINUXCLONE should be equal to __WCLONE, but we make sure */
896 	if (args->options & __WCLONE)
897 		options |= WLINUXCLONE;
898 
899 	return (linux_common_wait(td, args->pid, args->status, options, NULL));
900 }
901 
902 
903 int
904 linux_mknod(struct thread *td, struct linux_mknod_args *args)
905 {
906 	char *path;
907 	int error;
908 
909 	LCONVPATHCREAT(td, args->path, &path);
910 
911 #ifdef DEBUG
912 	if (ldebug(mknod))
913 		printf(ARGS(mknod, "%s, %d, %d"), path, args->mode, args->dev);
914 #endif
915 
916 	switch (args->mode & S_IFMT) {
917 	case S_IFIFO:
918 	case S_IFSOCK:
919 		error = kern_mkfifoat(td, AT_FDCWD, path, UIO_SYSSPACE,
920 		    args->mode);
921 		break;
922 
923 	case S_IFCHR:
924 	case S_IFBLK:
925 		error = kern_mknodat(td, AT_FDCWD, path, UIO_SYSSPACE,
926 		    args->mode, args->dev);
927 		break;
928 
929 	case S_IFDIR:
930 		error = EPERM;
931 		break;
932 
933 	case 0:
934 		args->mode |= S_IFREG;
935 		/* FALLTHROUGH */
936 	case S_IFREG:
937 		error = kern_openat(td, AT_FDCWD, path, UIO_SYSSPACE,
938 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
939 		if (error == 0)
940 			kern_close(td, td->td_retval[0]);
941 		break;
942 
943 	default:
944 		error = EINVAL;
945 		break;
946 	}
947 	LFREEPATH(path);
948 	return (error);
949 }
950 
951 int
952 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
953 {
954 	char *path;
955 	int error, dfd;
956 
957 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
958 	LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
959 
960 #ifdef DEBUG
961 	if (ldebug(mknodat))
962 		printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev);
963 #endif
964 
965 	switch (args->mode & S_IFMT) {
966 	case S_IFIFO:
967 	case S_IFSOCK:
968 		error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
969 		break;
970 
971 	case S_IFCHR:
972 	case S_IFBLK:
973 		error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
974 		    args->dev);
975 		break;
976 
977 	case S_IFDIR:
978 		error = EPERM;
979 		break;
980 
981 	case 0:
982 		args->mode |= S_IFREG;
983 		/* FALLTHROUGH */
984 	case S_IFREG:
985 		error = kern_openat(td, dfd, path, UIO_SYSSPACE,
986 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
987 		if (error == 0)
988 			kern_close(td, td->td_retval[0]);
989 		break;
990 
991 	default:
992 		error = EINVAL;
993 		break;
994 	}
995 	LFREEPATH(path);
996 	return (error);
997 }
998 
999 /*
1000  * UGH! This is just about the dumbest idea I've ever heard!!
1001  */
1002 int
1003 linux_personality(struct thread *td, struct linux_personality_args *args)
1004 {
1005 #ifdef DEBUG
1006 	if (ldebug(personality))
1007 		printf(ARGS(personality, "%lu"), (unsigned long)args->per);
1008 #endif
1009 	if (args->per != 0)
1010 		return (EINVAL);
1011 
1012 	/* Yes Jim, it's still a Linux... */
1013 	td->td_retval[0] = 0;
1014 	return (0);
1015 }
1016 
1017 struct l_itimerval {
1018 	l_timeval it_interval;
1019 	l_timeval it_value;
1020 };
1021 
1022 #define	B2L_ITIMERVAL(bip, lip) 					\
1023 	(bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec;		\
1024 	(bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec;	\
1025 	(bip)->it_value.tv_sec = (lip)->it_value.tv_sec;		\
1026 	(bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1027 
1028 int
1029 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1030 {
1031 	int error;
1032 	struct l_itimerval ls;
1033 	struct itimerval aitv, oitv;
1034 
1035 #ifdef DEBUG
1036 	if (ldebug(setitimer))
1037 		printf(ARGS(setitimer, "%p, %p"),
1038 		    (void *)uap->itv, (void *)uap->oitv);
1039 #endif
1040 
1041 	if (uap->itv == NULL) {
1042 		uap->itv = uap->oitv;
1043 		return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1044 	}
1045 
1046 	error = copyin(uap->itv, &ls, sizeof(ls));
1047 	if (error != 0)
1048 		return (error);
1049 	B2L_ITIMERVAL(&aitv, &ls);
1050 #ifdef DEBUG
1051 	if (ldebug(setitimer)) {
1052 		printf("setitimer: value: sec: %jd, usec: %ld\n",
1053 		    (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1054 		printf("setitimer: interval: sec: %jd, usec: %ld\n",
1055 		    (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1056 	}
1057 #endif
1058 	error = kern_setitimer(td, uap->which, &aitv, &oitv);
1059 	if (error != 0 || uap->oitv == NULL)
1060 		return (error);
1061 	B2L_ITIMERVAL(&ls, &oitv);
1062 
1063 	return (copyout(&ls, uap->oitv, sizeof(ls)));
1064 }
1065 
1066 int
1067 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1068 {
1069 	int error;
1070 	struct l_itimerval ls;
1071 	struct itimerval aitv;
1072 
1073 #ifdef DEBUG
1074 	if (ldebug(getitimer))
1075 		printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1076 #endif
1077 	error = kern_getitimer(td, uap->which, &aitv);
1078 	if (error != 0)
1079 		return (error);
1080 	B2L_ITIMERVAL(&ls, &aitv);
1081 	return (copyout(&ls, uap->itv, sizeof(ls)));
1082 }
1083 
1084 int
1085 linux_nice(struct thread *td, struct linux_nice_args *args)
1086 {
1087 	struct setpriority_args bsd_args;
1088 
1089 	bsd_args.which = PRIO_PROCESS;
1090 	bsd_args.who = 0;		/* current process */
1091 	bsd_args.prio = args->inc;
1092 	return (sys_setpriority(td, &bsd_args));
1093 }
1094 
1095 int
1096 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1097 {
1098 	struct ucred *newcred, *oldcred;
1099 	l_gid_t *linux_gidset;
1100 	gid_t *bsd_gidset;
1101 	int ngrp, error;
1102 	struct proc *p;
1103 
1104 	ngrp = args->gidsetsize;
1105 	if (ngrp < 0 || ngrp >= ngroups_max + 1)
1106 		return (EINVAL);
1107 	linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_TEMP, M_WAITOK);
1108 	error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1109 	if (error)
1110 		goto out;
1111 	newcred = crget();
1112 	p = td->td_proc;
1113 	PROC_LOCK(p);
1114 	oldcred = crcopysafe(p, newcred);
1115 
1116 	/*
1117 	 * cr_groups[0] holds egid. Setting the whole set from
1118 	 * the supplied set will cause egid to be changed too.
1119 	 * Keep cr_groups[0] unchanged to prevent that.
1120 	 */
1121 
1122 	if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) {
1123 		PROC_UNLOCK(p);
1124 		crfree(newcred);
1125 		goto out;
1126 	}
1127 
1128 	if (ngrp > 0) {
1129 		newcred->cr_ngroups = ngrp + 1;
1130 
1131 		bsd_gidset = newcred->cr_groups;
1132 		ngrp--;
1133 		while (ngrp >= 0) {
1134 			bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1135 			ngrp--;
1136 		}
1137 	} else
1138 		newcred->cr_ngroups = 1;
1139 
1140 	setsugid(p);
1141 	p->p_ucred = newcred;
1142 	PROC_UNLOCK(p);
1143 	crfree(oldcred);
1144 	error = 0;
1145 out:
1146 	free(linux_gidset, M_TEMP);
1147 	return (error);
1148 }
1149 
1150 int
1151 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1152 {
1153 	struct ucred *cred;
1154 	l_gid_t *linux_gidset;
1155 	gid_t *bsd_gidset;
1156 	int bsd_gidsetsz, ngrp, error;
1157 
1158 	cred = td->td_ucred;
1159 	bsd_gidset = cred->cr_groups;
1160 	bsd_gidsetsz = cred->cr_ngroups - 1;
1161 
1162 	/*
1163 	 * cr_groups[0] holds egid. Returning the whole set
1164 	 * here will cause a duplicate. Exclude cr_groups[0]
1165 	 * to prevent that.
1166 	 */
1167 
1168 	if ((ngrp = args->gidsetsize) == 0) {
1169 		td->td_retval[0] = bsd_gidsetsz;
1170 		return (0);
1171 	}
1172 
1173 	if (ngrp < bsd_gidsetsz)
1174 		return (EINVAL);
1175 
1176 	ngrp = 0;
1177 	linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1178 	    M_TEMP, M_WAITOK);
1179 	while (ngrp < bsd_gidsetsz) {
1180 		linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1181 		ngrp++;
1182 	}
1183 
1184 	error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1185 	free(linux_gidset, M_TEMP);
1186 	if (error)
1187 		return (error);
1188 
1189 	td->td_retval[0] = ngrp;
1190 	return (0);
1191 }
1192 
1193 int
1194 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1195 {
1196 	struct rlimit bsd_rlim;
1197 	struct l_rlimit rlim;
1198 	u_int which;
1199 	int error;
1200 
1201 #ifdef DEBUG
1202 	if (ldebug(setrlimit))
1203 		printf(ARGS(setrlimit, "%d, %p"),
1204 		    args->resource, (void *)args->rlim);
1205 #endif
1206 
1207 	if (args->resource >= LINUX_RLIM_NLIMITS)
1208 		return (EINVAL);
1209 
1210 	which = linux_to_bsd_resource[args->resource];
1211 	if (which == -1)
1212 		return (EINVAL);
1213 
1214 	error = copyin(args->rlim, &rlim, sizeof(rlim));
1215 	if (error)
1216 		return (error);
1217 
1218 	bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1219 	bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1220 	return (kern_setrlimit(td, which, &bsd_rlim));
1221 }
1222 
1223 int
1224 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1225 {
1226 	struct l_rlimit rlim;
1227 	struct proc *p = td->td_proc;
1228 	struct rlimit bsd_rlim;
1229 	u_int which;
1230 
1231 #ifdef DEBUG
1232 	if (ldebug(old_getrlimit))
1233 		printf(ARGS(old_getrlimit, "%d, %p"),
1234 		    args->resource, (void *)args->rlim);
1235 #endif
1236 
1237 	if (args->resource >= LINUX_RLIM_NLIMITS)
1238 		return (EINVAL);
1239 
1240 	which = linux_to_bsd_resource[args->resource];
1241 	if (which == -1)
1242 		return (EINVAL);
1243 
1244 	PROC_LOCK(p);
1245 	lim_rlimit(p, which, &bsd_rlim);
1246 	PROC_UNLOCK(p);
1247 
1248 #ifdef COMPAT_LINUX32
1249 	rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1250 	if (rlim.rlim_cur == UINT_MAX)
1251 		rlim.rlim_cur = INT_MAX;
1252 	rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1253 	if (rlim.rlim_max == UINT_MAX)
1254 		rlim.rlim_max = INT_MAX;
1255 #else
1256 	rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1257 	if (rlim.rlim_cur == ULONG_MAX)
1258 		rlim.rlim_cur = LONG_MAX;
1259 	rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1260 	if (rlim.rlim_max == ULONG_MAX)
1261 		rlim.rlim_max = LONG_MAX;
1262 #endif
1263 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1264 }
1265 
1266 int
1267 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1268 {
1269 	struct l_rlimit rlim;
1270 	struct proc *p = td->td_proc;
1271 	struct rlimit bsd_rlim;
1272 	u_int which;
1273 
1274 #ifdef DEBUG
1275 	if (ldebug(getrlimit))
1276 		printf(ARGS(getrlimit, "%d, %p"),
1277 		    args->resource, (void *)args->rlim);
1278 #endif
1279 
1280 	if (args->resource >= LINUX_RLIM_NLIMITS)
1281 		return (EINVAL);
1282 
1283 	which = linux_to_bsd_resource[args->resource];
1284 	if (which == -1)
1285 		return (EINVAL);
1286 
1287 	PROC_LOCK(p);
1288 	lim_rlimit(p, which, &bsd_rlim);
1289 	PROC_UNLOCK(p);
1290 
1291 	rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1292 	rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1293 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1294 }
1295 
1296 int
1297 linux_sched_setscheduler(struct thread *td,
1298     struct linux_sched_setscheduler_args *args)
1299 {
1300 	struct sched_setscheduler_args bsd;
1301 
1302 #ifdef DEBUG
1303 	if (ldebug(sched_setscheduler))
1304 		printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1305 		    args->pid, args->policy, (const void *)args->param);
1306 #endif
1307 
1308 	switch (args->policy) {
1309 	case LINUX_SCHED_OTHER:
1310 		bsd.policy = SCHED_OTHER;
1311 		break;
1312 	case LINUX_SCHED_FIFO:
1313 		bsd.policy = SCHED_FIFO;
1314 		break;
1315 	case LINUX_SCHED_RR:
1316 		bsd.policy = SCHED_RR;
1317 		break;
1318 	default:
1319 		return (EINVAL);
1320 	}
1321 
1322 	bsd.pid = args->pid;
1323 	bsd.param = (struct sched_param *)args->param;
1324 	return (sys_sched_setscheduler(td, &bsd));
1325 }
1326 
1327 int
1328 linux_sched_getscheduler(struct thread *td,
1329     struct linux_sched_getscheduler_args *args)
1330 {
1331 	struct sched_getscheduler_args bsd;
1332 	int error;
1333 
1334 #ifdef DEBUG
1335 	if (ldebug(sched_getscheduler))
1336 		printf(ARGS(sched_getscheduler, "%d"), args->pid);
1337 #endif
1338 
1339 	bsd.pid = args->pid;
1340 	error = sys_sched_getscheduler(td, &bsd);
1341 
1342 	switch (td->td_retval[0]) {
1343 	case SCHED_OTHER:
1344 		td->td_retval[0] = LINUX_SCHED_OTHER;
1345 		break;
1346 	case SCHED_FIFO:
1347 		td->td_retval[0] = LINUX_SCHED_FIFO;
1348 		break;
1349 	case SCHED_RR:
1350 		td->td_retval[0] = LINUX_SCHED_RR;
1351 		break;
1352 	}
1353 
1354 	return (error);
1355 }
1356 
1357 int
1358 linux_sched_get_priority_max(struct thread *td,
1359     struct linux_sched_get_priority_max_args *args)
1360 {
1361 	struct sched_get_priority_max_args bsd;
1362 
1363 #ifdef DEBUG
1364 	if (ldebug(sched_get_priority_max))
1365 		printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1366 #endif
1367 
1368 	switch (args->policy) {
1369 	case LINUX_SCHED_OTHER:
1370 		bsd.policy = SCHED_OTHER;
1371 		break;
1372 	case LINUX_SCHED_FIFO:
1373 		bsd.policy = SCHED_FIFO;
1374 		break;
1375 	case LINUX_SCHED_RR:
1376 		bsd.policy = SCHED_RR;
1377 		break;
1378 	default:
1379 		return (EINVAL);
1380 	}
1381 	return (sys_sched_get_priority_max(td, &bsd));
1382 }
1383 
1384 int
1385 linux_sched_get_priority_min(struct thread *td,
1386     struct linux_sched_get_priority_min_args *args)
1387 {
1388 	struct sched_get_priority_min_args bsd;
1389 
1390 #ifdef DEBUG
1391 	if (ldebug(sched_get_priority_min))
1392 		printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1393 #endif
1394 
1395 	switch (args->policy) {
1396 	case LINUX_SCHED_OTHER:
1397 		bsd.policy = SCHED_OTHER;
1398 		break;
1399 	case LINUX_SCHED_FIFO:
1400 		bsd.policy = SCHED_FIFO;
1401 		break;
1402 	case LINUX_SCHED_RR:
1403 		bsd.policy = SCHED_RR;
1404 		break;
1405 	default:
1406 		return (EINVAL);
1407 	}
1408 	return (sys_sched_get_priority_min(td, &bsd));
1409 }
1410 
1411 #define REBOOT_CAD_ON	0x89abcdef
1412 #define REBOOT_CAD_OFF	0
1413 #define REBOOT_HALT	0xcdef0123
1414 #define REBOOT_RESTART	0x01234567
1415 #define REBOOT_RESTART2	0xA1B2C3D4
1416 #define REBOOT_POWEROFF	0x4321FEDC
1417 #define REBOOT_MAGIC1	0xfee1dead
1418 #define REBOOT_MAGIC2	0x28121969
1419 #define REBOOT_MAGIC2A	0x05121996
1420 #define REBOOT_MAGIC2B	0x16041998
1421 
1422 int
1423 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1424 {
1425 	struct reboot_args bsd_args;
1426 
1427 #ifdef DEBUG
1428 	if (ldebug(reboot))
1429 		printf(ARGS(reboot, "0x%x"), args->cmd);
1430 #endif
1431 
1432 	if (args->magic1 != REBOOT_MAGIC1)
1433 		return (EINVAL);
1434 
1435 	switch (args->magic2) {
1436 	case REBOOT_MAGIC2:
1437 	case REBOOT_MAGIC2A:
1438 	case REBOOT_MAGIC2B:
1439 		break;
1440 	default:
1441 		return (EINVAL);
1442 	}
1443 
1444 	switch (args->cmd) {
1445 	case REBOOT_CAD_ON:
1446 	case REBOOT_CAD_OFF:
1447 		return (priv_check(td, PRIV_REBOOT));
1448 	case REBOOT_HALT:
1449 		bsd_args.opt = RB_HALT;
1450 		break;
1451 	case REBOOT_RESTART:
1452 	case REBOOT_RESTART2:
1453 		bsd_args.opt = 0;
1454 		break;
1455 	case REBOOT_POWEROFF:
1456 		bsd_args.opt = RB_POWEROFF;
1457 		break;
1458 	default:
1459 		return (EINVAL);
1460 	}
1461 	return (sys_reboot(td, &bsd_args));
1462 }
1463 
1464 
1465 /*
1466  * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1467  * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
1468  * are assumed to be preserved. The following lightweight syscalls fixes
1469  * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
1470  *
1471  * linux_getpid() - MP SAFE
1472  * linux_getgid() - MP SAFE
1473  * linux_getuid() - MP SAFE
1474  */
1475 
1476 int
1477 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1478 {
1479 	struct linux_emuldata *em;
1480 
1481 #ifdef DEBUG
1482 	if (ldebug(getpid))
1483 		printf(ARGS(getpid, ""));
1484 #endif
1485 
1486 	if (linux_use26(td)) {
1487 		em = em_find(td->td_proc, EMUL_DONTLOCK);
1488 		KASSERT(em != NULL, ("getpid: emuldata not found.\n"));
1489 		td->td_retval[0] = em->shared->group_pid;
1490 	} else {
1491 		td->td_retval[0] = td->td_proc->p_pid;
1492 	}
1493 
1494 	return (0);
1495 }
1496 
1497 int
1498 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1499 {
1500 
1501 #ifdef DEBUG
1502 	if (ldebug(gettid))
1503 		printf(ARGS(gettid, ""));
1504 #endif
1505 
1506 	td->td_retval[0] = td->td_proc->p_pid;
1507 	return (0);
1508 }
1509 
1510 
1511 int
1512 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1513 {
1514 	struct linux_emuldata *em;
1515 	struct proc *p, *pp;
1516 
1517 #ifdef DEBUG
1518 	if (ldebug(getppid))
1519 		printf(ARGS(getppid, ""));
1520 #endif
1521 
1522 	if (!linux_use26(td)) {
1523 		PROC_LOCK(td->td_proc);
1524 		td->td_retval[0] = td->td_proc->p_pptr->p_pid;
1525 		PROC_UNLOCK(td->td_proc);
1526 		return (0);
1527 	}
1528 
1529 	em = em_find(td->td_proc, EMUL_DONTLOCK);
1530 
1531 	KASSERT(em != NULL, ("getppid: process emuldata not found.\n"));
1532 
1533 	/* find the group leader */
1534 	p = pfind(em->shared->group_pid);
1535 
1536 	if (p == NULL) {
1537 #ifdef DEBUG
1538 	   	printf(LMSG("parent process not found.\n"));
1539 #endif
1540 		return (0);
1541 	}
1542 
1543 	pp = p->p_pptr;		/* switch to parent */
1544 	PROC_LOCK(pp);
1545 	PROC_UNLOCK(p);
1546 
1547 	/* if its also linux process */
1548 	if (pp->p_sysent == &elf_linux_sysvec) {
1549 		em = em_find(pp, EMUL_DONTLOCK);
1550 		KASSERT(em != NULL, ("getppid: parent emuldata not found.\n"));
1551 
1552 		td->td_retval[0] = em->shared->group_pid;
1553 	} else
1554 		td->td_retval[0] = pp->p_pid;
1555 
1556 	PROC_UNLOCK(pp);
1557 
1558 	return (0);
1559 }
1560 
1561 int
1562 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1563 {
1564 
1565 #ifdef DEBUG
1566 	if (ldebug(getgid))
1567 		printf(ARGS(getgid, ""));
1568 #endif
1569 
1570 	td->td_retval[0] = td->td_ucred->cr_rgid;
1571 	return (0);
1572 }
1573 
1574 int
1575 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1576 {
1577 
1578 #ifdef DEBUG
1579 	if (ldebug(getuid))
1580 		printf(ARGS(getuid, ""));
1581 #endif
1582 
1583 	td->td_retval[0] = td->td_ucred->cr_ruid;
1584 	return (0);
1585 }
1586 
1587 
1588 int
1589 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1590 {
1591 	struct getsid_args bsd;
1592 
1593 #ifdef DEBUG
1594 	if (ldebug(getsid))
1595 		printf(ARGS(getsid, "%i"), args->pid);
1596 #endif
1597 
1598 	bsd.pid = args->pid;
1599 	return (sys_getsid(td, &bsd));
1600 }
1601 
1602 int
1603 linux_nosys(struct thread *td, struct nosys_args *ignore)
1604 {
1605 
1606 	return (ENOSYS);
1607 }
1608 
1609 int
1610 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1611 {
1612 	struct getpriority_args bsd_args;
1613 	int error;
1614 
1615 #ifdef DEBUG
1616 	if (ldebug(getpriority))
1617 		printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
1618 #endif
1619 
1620 	bsd_args.which = args->which;
1621 	bsd_args.who = args->who;
1622 	error = sys_getpriority(td, &bsd_args);
1623 	td->td_retval[0] = 20 - td->td_retval[0];
1624 	return (error);
1625 }
1626 
1627 int
1628 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1629 {
1630 	int name[2];
1631 
1632 #ifdef DEBUG
1633 	if (ldebug(sethostname))
1634 		printf(ARGS(sethostname, "*, %i"), args->len);
1635 #endif
1636 
1637 	name[0] = CTL_KERN;
1638 	name[1] = KERN_HOSTNAME;
1639 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1640 	    args->len, 0, 0));
1641 }
1642 
1643 int
1644 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1645 {
1646 	int name[2];
1647 
1648 #ifdef DEBUG
1649 	if (ldebug(setdomainname))
1650 		printf(ARGS(setdomainname, "*, %i"), args->len);
1651 #endif
1652 
1653 	name[0] = CTL_KERN;
1654 	name[1] = KERN_NISDOMAINNAME;
1655 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1656 	    args->len, 0, 0));
1657 }
1658 
1659 int
1660 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1661 {
1662 	struct linux_emuldata *em;
1663 
1664 #ifdef DEBUG
1665 	if (ldebug(exit_group))
1666 		printf(ARGS(exit_group, "%i"), args->error_code);
1667 #endif
1668 
1669 	em = em_find(td->td_proc, EMUL_DONTLOCK);
1670 	if (em->shared->refs > 1) {
1671 		EMUL_SHARED_WLOCK(&emul_shared_lock);
1672 		em->shared->flags |= EMUL_SHARED_HASXSTAT;
1673 		em->shared->xstat = W_EXITCODE(args->error_code, 0);
1674 		EMUL_SHARED_WUNLOCK(&emul_shared_lock);
1675 		if (linux_use26(td))
1676 			linux_kill_threads(td, SIGKILL);
1677 	}
1678 
1679 	/*
1680 	 * XXX: we should send a signal to the parent if
1681 	 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1682 	 * as it doesnt occur often.
1683 	 */
1684 	exit1(td, W_EXITCODE(args->error_code, 0));
1685 
1686 	return (0);
1687 }
1688 
1689 #define _LINUX_CAPABILITY_VERSION  0x19980330
1690 
1691 struct l_user_cap_header {
1692 	l_int	version;
1693 	l_int	pid;
1694 };
1695 
1696 struct l_user_cap_data {
1697 	l_int	effective;
1698 	l_int	permitted;
1699 	l_int	inheritable;
1700 };
1701 
1702 int
1703 linux_capget(struct thread *td, struct linux_capget_args *args)
1704 {
1705 	struct l_user_cap_header luch;
1706 	struct l_user_cap_data lucd;
1707 	int error;
1708 
1709 	if (args->hdrp == NULL)
1710 		return (EFAULT);
1711 
1712 	error = copyin(args->hdrp, &luch, sizeof(luch));
1713 	if (error != 0)
1714 		return (error);
1715 
1716 	if (luch.version != _LINUX_CAPABILITY_VERSION) {
1717 		luch.version = _LINUX_CAPABILITY_VERSION;
1718 		error = copyout(&luch, args->hdrp, sizeof(luch));
1719 		if (error)
1720 			return (error);
1721 		return (EINVAL);
1722 	}
1723 
1724 	if (luch.pid)
1725 		return (EPERM);
1726 
1727 	if (args->datap) {
1728 		/*
1729 		 * The current implementation doesn't support setting
1730 		 * a capability (it's essentially a stub) so indicate
1731 		 * that no capabilities are currently set or available
1732 		 * to request.
1733 		 */
1734 		bzero (&lucd, sizeof(lucd));
1735 		error = copyout(&lucd, args->datap, sizeof(lucd));
1736 	}
1737 
1738 	return (error);
1739 }
1740 
1741 int
1742 linux_capset(struct thread *td, struct linux_capset_args *args)
1743 {
1744 	struct l_user_cap_header luch;
1745 	struct l_user_cap_data lucd;
1746 	int error;
1747 
1748 	if (args->hdrp == NULL || args->datap == NULL)
1749 		return (EFAULT);
1750 
1751 	error = copyin(args->hdrp, &luch, sizeof(luch));
1752 	if (error != 0)
1753 		return (error);
1754 
1755 	if (luch.version != _LINUX_CAPABILITY_VERSION) {
1756 		luch.version = _LINUX_CAPABILITY_VERSION;
1757 		error = copyout(&luch, args->hdrp, sizeof(luch));
1758 		if (error)
1759 			return (error);
1760 		return (EINVAL);
1761 	}
1762 
1763 	if (luch.pid)
1764 		return (EPERM);
1765 
1766 	error = copyin(args->datap, &lucd, sizeof(lucd));
1767 	if (error != 0)
1768 		return (error);
1769 
1770 	/* We currently don't support setting any capabilities. */
1771 	if (lucd.effective || lucd.permitted || lucd.inheritable) {
1772 		linux_msg(td,
1773 			  "capset effective=0x%x, permitted=0x%x, "
1774 			  "inheritable=0x%x is not implemented",
1775 			  (int)lucd.effective, (int)lucd.permitted,
1776 			  (int)lucd.inheritable);
1777 		return (EPERM);
1778 	}
1779 
1780 	return (0);
1781 }
1782 
1783 int
1784 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1785 {
1786 	int error = 0, max_size;
1787 	struct proc *p = td->td_proc;
1788 	char comm[LINUX_MAX_COMM_LEN];
1789 	struct linux_emuldata *em;
1790 	int pdeath_signal;
1791 
1792 #ifdef DEBUG
1793 	if (ldebug(prctl))
1794 		printf(ARGS(prctl, "%d, %d, %d, %d, %d"), args->option,
1795 		    args->arg2, args->arg3, args->arg4, args->arg5);
1796 #endif
1797 
1798 	switch (args->option) {
1799 	case LINUX_PR_SET_PDEATHSIG:
1800 		if (!LINUX_SIG_VALID(args->arg2))
1801 			return (EINVAL);
1802 		em = em_find(p, EMUL_DOLOCK);
1803 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1804 		em->pdeath_signal = args->arg2;
1805 		EMUL_UNLOCK(&emul_lock);
1806 		break;
1807 	case LINUX_PR_GET_PDEATHSIG:
1808 		em = em_find(p, EMUL_DOLOCK);
1809 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1810 		pdeath_signal = em->pdeath_signal;
1811 		EMUL_UNLOCK(&emul_lock);
1812 		error = copyout(&pdeath_signal,
1813 		    (void *)(register_t)args->arg2,
1814 		    sizeof(pdeath_signal));
1815 		break;
1816 	case LINUX_PR_GET_KEEPCAPS:
1817 		/*
1818 		 * Indicate that we always clear the effective and
1819 		 * permitted capability sets when the user id becomes
1820 		 * non-zero (actually the capability sets are simply
1821 		 * always zero in the current implementation).
1822 		 */
1823 		td->td_retval[0] = 0;
1824 		break;
1825 	case LINUX_PR_SET_KEEPCAPS:
1826 		/*
1827 		 * Ignore requests to keep the effective and permitted
1828 		 * capability sets when the user id becomes non-zero.
1829 		 */
1830 		break;
1831 	case LINUX_PR_SET_NAME:
1832 		/*
1833 		 * To be on the safe side we need to make sure to not
1834 		 * overflow the size a linux program expects. We already
1835 		 * do this here in the copyin, so that we don't need to
1836 		 * check on copyout.
1837 		 */
1838 		max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1839 		error = copyinstr((void *)(register_t)args->arg2, comm,
1840 		    max_size, NULL);
1841 
1842 		/* Linux silently truncates the name if it is too long. */
1843 		if (error == ENAMETOOLONG) {
1844 			/*
1845 			 * XXX: copyinstr() isn't documented to populate the
1846 			 * array completely, so do a copyin() to be on the
1847 			 * safe side. This should be changed in case
1848 			 * copyinstr() is changed to guarantee this.
1849 			 */
1850 			error = copyin((void *)(register_t)args->arg2, comm,
1851 			    max_size - 1);
1852 			comm[max_size - 1] = '\0';
1853 		}
1854 		if (error)
1855 			return (error);
1856 
1857 		PROC_LOCK(p);
1858 		strlcpy(p->p_comm, comm, sizeof(p->p_comm));
1859 		PROC_UNLOCK(p);
1860 		break;
1861 	case LINUX_PR_GET_NAME:
1862 		PROC_LOCK(p);
1863 		strlcpy(comm, p->p_comm, sizeof(comm));
1864 		PROC_UNLOCK(p);
1865 		error = copyout(comm, (void *)(register_t)args->arg2,
1866 		    strlen(comm) + 1);
1867 		break;
1868 	default:
1869 		error = EINVAL;
1870 		break;
1871 	}
1872 
1873 	return (error);
1874 }
1875 
1876 /*
1877  * Get affinity of a process.
1878  */
1879 int
1880 linux_sched_getaffinity(struct thread *td,
1881     struct linux_sched_getaffinity_args *args)
1882 {
1883 	int error;
1884 	struct cpuset_getaffinity_args cga;
1885 
1886 #ifdef DEBUG
1887 	if (ldebug(sched_getaffinity))
1888 		printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid,
1889 		    args->len);
1890 #endif
1891 	if (args->len < sizeof(cpuset_t))
1892 		return (EINVAL);
1893 
1894 	cga.level = CPU_LEVEL_WHICH;
1895 	cga.which = CPU_WHICH_PID;
1896 	cga.id = args->pid;
1897 	cga.cpusetsize = sizeof(cpuset_t);
1898 	cga.mask = (cpuset_t *) args->user_mask_ptr;
1899 
1900 	if ((error = sys_cpuset_getaffinity(td, &cga)) == 0)
1901 		td->td_retval[0] = sizeof(cpuset_t);
1902 
1903 	return (error);
1904 }
1905 
1906 /*
1907  *  Set affinity of a process.
1908  */
1909 int
1910 linux_sched_setaffinity(struct thread *td,
1911     struct linux_sched_setaffinity_args *args)
1912 {
1913 	struct cpuset_setaffinity_args csa;
1914 
1915 #ifdef DEBUG
1916 	if (ldebug(sched_setaffinity))
1917 		printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid,
1918 		    args->len);
1919 #endif
1920 	if (args->len < sizeof(cpuset_t))
1921 		return (EINVAL);
1922 
1923 	csa.level = CPU_LEVEL_WHICH;
1924 	csa.which = CPU_WHICH_PID;
1925 	csa.id = args->pid;
1926 	csa.cpusetsize = sizeof(cpuset_t);
1927 	csa.mask = (cpuset_t *) args->user_mask_ptr;
1928 
1929 	return (sys_cpuset_setaffinity(td, &csa));
1930 }
1931