xref: /freebsd/sys/compat/linux/linux_misc.c (revision 6486b015fc84e96725fef22b0e3363351399ae83)
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/signalvar.h>
57 #include <sys/stat.h>
58 #include <sys/syscallsubr.h>
59 #include <sys/sysctl.h>
60 #include <sys/sysproto.h>
61 #include <sys/systm.h>
62 #include <sys/time.h>
63 #include <sys/vmmeter.h>
64 #include <sys/vnode.h>
65 #include <sys/wait.h>
66 #include <sys/cpuset.h>
67 
68 #include <security/mac/mac_framework.h>
69 
70 #include <vm/vm.h>
71 #include <vm/pmap.h>
72 #include <vm/vm_kern.h>
73 #include <vm/vm_map.h>
74 #include <vm/vm_extern.h>
75 #include <vm/vm_object.h>
76 #include <vm/swap_pager.h>
77 
78 #ifdef COMPAT_LINUX32
79 #include <machine/../linux32/linux.h>
80 #include <machine/../linux32/linux32_proto.h>
81 #else
82 #include <machine/../linux/linux.h>
83 #include <machine/../linux/linux_proto.h>
84 #endif
85 
86 #include <compat/linux/linux_file.h>
87 #include <compat/linux/linux_mib.h>
88 #include <compat/linux/linux_signal.h>
89 #include <compat/linux/linux_util.h>
90 #include <compat/linux/linux_sysproto.h>
91 #include <compat/linux/linux_emul.h>
92 #include <compat/linux/linux_misc.h>
93 
94 int stclohz;				/* Statistics clock frequency */
95 
96 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
97 	RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
98 	RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
99 	RLIMIT_MEMLOCK, RLIMIT_AS
100 };
101 
102 struct l_sysinfo {
103 	l_long		uptime;		/* Seconds since boot */
104 	l_ulong		loads[3];	/* 1, 5, and 15 minute load averages */
105 #define LINUX_SYSINFO_LOADS_SCALE 65536
106 	l_ulong		totalram;	/* Total usable main memory size */
107 	l_ulong		freeram;	/* Available memory size */
108 	l_ulong		sharedram;	/* Amount of shared memory */
109 	l_ulong		bufferram;	/* Memory used by buffers */
110 	l_ulong		totalswap;	/* Total swap space size */
111 	l_ulong		freeswap;	/* swap space still available */
112 	l_ushort	procs;		/* Number of current processes */
113 	l_ushort	pads;
114 	l_ulong		totalbig;
115 	l_ulong		freebig;
116 	l_uint		mem_unit;
117 	char		_f[20-2*sizeof(l_long)-sizeof(l_int)];	/* padding */
118 };
119 int
120 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
121 {
122 	struct l_sysinfo sysinfo;
123 	vm_object_t object;
124 	int i, j;
125 	struct timespec ts;
126 
127 	getnanouptime(&ts);
128 	if (ts.tv_nsec != 0)
129 		ts.tv_sec++;
130 	sysinfo.uptime = ts.tv_sec;
131 
132 	/* Use the information from the mib to get our load averages */
133 	for (i = 0; i < 3; i++)
134 		sysinfo.loads[i] = averunnable.ldavg[i] *
135 		    LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
136 
137 	sysinfo.totalram = physmem * PAGE_SIZE;
138 	sysinfo.freeram = sysinfo.totalram - cnt.v_wire_count * PAGE_SIZE;
139 
140 	sysinfo.sharedram = 0;
141 	mtx_lock(&vm_object_list_mtx);
142 	TAILQ_FOREACH(object, &vm_object_list, object_list)
143 		if (object->shadow_count > 1)
144 			sysinfo.sharedram += object->resident_page_count;
145 	mtx_unlock(&vm_object_list_mtx);
146 
147 	sysinfo.sharedram *= PAGE_SIZE;
148 	sysinfo.bufferram = 0;
149 
150 	swap_pager_status(&i, &j);
151 	sysinfo.totalswap = i * PAGE_SIZE;
152 	sysinfo.freeswap = (i - j) * PAGE_SIZE;
153 
154 	sysinfo.procs = nprocs;
155 
156 	/* The following are only present in newer Linux kernels. */
157 	sysinfo.totalbig = 0;
158 	sysinfo.freebig = 0;
159 	sysinfo.mem_unit = 1;
160 
161 	return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
162 }
163 
164 int
165 linux_alarm(struct thread *td, struct linux_alarm_args *args)
166 {
167 	struct itimerval it, old_it;
168 	u_int secs;
169 	int error;
170 
171 #ifdef DEBUG
172 	if (ldebug(alarm))
173 		printf(ARGS(alarm, "%u"), args->secs);
174 #endif
175 
176 	secs = args->secs;
177 
178 	if (secs > INT_MAX)
179 		secs = INT_MAX;
180 
181 	it.it_value.tv_sec = (long) secs;
182 	it.it_value.tv_usec = 0;
183 	it.it_interval.tv_sec = 0;
184 	it.it_interval.tv_usec = 0;
185 	error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
186 	if (error)
187 		return (error);
188 	if (timevalisset(&old_it.it_value)) {
189 		if (old_it.it_value.tv_usec != 0)
190 			old_it.it_value.tv_sec++;
191 		td->td_retval[0] = old_it.it_value.tv_sec;
192 	}
193 	return (0);
194 }
195 
196 int
197 linux_brk(struct thread *td, struct linux_brk_args *args)
198 {
199 	struct vmspace *vm = td->td_proc->p_vmspace;
200 	vm_offset_t new, old;
201 	struct obreak_args /* {
202 		char * nsize;
203 	} */ tmp;
204 
205 #ifdef DEBUG
206 	if (ldebug(brk))
207 		printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
208 #endif
209 	old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize);
210 	new = (vm_offset_t)args->dsend;
211 	tmp.nsize = (char *)new;
212 	if (((caddr_t)new > vm->vm_daddr) && !sys_obreak(td, &tmp))
213 		td->td_retval[0] = (long)new;
214 	else
215 		td->td_retval[0] = (long)old;
216 
217 	return (0);
218 }
219 
220 #if defined(__i386__)
221 /* XXX: what about amd64/linux32? */
222 
223 int
224 linux_uselib(struct thread *td, struct linux_uselib_args *args)
225 {
226 	struct nameidata ni;
227 	struct vnode *vp;
228 	struct exec *a_out;
229 	struct vattr attr;
230 	vm_offset_t vmaddr;
231 	unsigned long file_offset;
232 	unsigned long bss_size;
233 	char *library;
234 	ssize_t aresid;
235 	int error;
236 	int locked, vfslocked;
237 
238 	LCONVPATHEXIST(td, args->library, &library);
239 
240 #ifdef DEBUG
241 	if (ldebug(uselib))
242 		printf(ARGS(uselib, "%s"), library);
243 #endif
244 
245 	a_out = NULL;
246 	vfslocked = 0;
247 	locked = 0;
248 	vp = NULL;
249 
250 	NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
251 	    UIO_SYSSPACE, library, td);
252 	error = namei(&ni);
253 	LFREEPATH(library);
254 	if (error)
255 		goto cleanup;
256 
257 	vp = ni.ni_vp;
258 	vfslocked = NDHASGIANT(&ni);
259 	NDFREE(&ni, NDF_ONLY_PNBUF);
260 
261 	/*
262 	 * From here on down, we have a locked vnode that must be unlocked.
263 	 * XXX: The code below largely duplicates exec_check_permissions().
264 	 */
265 	locked = 1;
266 
267 	/* Writable? */
268 	if (vp->v_writecount) {
269 		error = ETXTBSY;
270 		goto cleanup;
271 	}
272 
273 	/* Executable? */
274 	error = VOP_GETATTR(vp, &attr, td->td_ucred);
275 	if (error)
276 		goto cleanup;
277 
278 	if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
279 	    ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
280 		/* EACCESS is what exec(2) returns. */
281 		error = ENOEXEC;
282 		goto cleanup;
283 	}
284 
285 	/* Sensible size? */
286 	if (attr.va_size == 0) {
287 		error = ENOEXEC;
288 		goto cleanup;
289 	}
290 
291 	/* Can we access it? */
292 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
293 	if (error)
294 		goto cleanup;
295 
296 	/*
297 	 * XXX: This should use vn_open() so that it is properly authorized,
298 	 * and to reduce code redundancy all over the place here.
299 	 * XXX: Not really, it duplicates far more of exec_check_permissions()
300 	 * than vn_open().
301 	 */
302 #ifdef MAC
303 	error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
304 	if (error)
305 		goto cleanup;
306 #endif
307 	error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
308 	if (error)
309 		goto cleanup;
310 
311 	/* Pull in executable header into exec_map */
312 	error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
313 	    VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
314 	if (error)
315 		goto cleanup;
316 
317 	/* Is it a Linux binary ? */
318 	if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
319 		error = ENOEXEC;
320 		goto cleanup;
321 	}
322 
323 	/*
324 	 * While we are here, we should REALLY do some more checks
325 	 */
326 
327 	/* Set file/virtual offset based on a.out variant. */
328 	switch ((int)(a_out->a_magic & 0xffff)) {
329 	case 0413:			/* ZMAGIC */
330 		file_offset = 1024;
331 		break;
332 	case 0314:			/* QMAGIC */
333 		file_offset = 0;
334 		break;
335 	default:
336 		error = ENOEXEC;
337 		goto cleanup;
338 	}
339 
340 	bss_size = round_page(a_out->a_bss);
341 
342 	/* Check various fields in header for validity/bounds. */
343 	if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
344 		error = ENOEXEC;
345 		goto cleanup;
346 	}
347 
348 	/* text + data can't exceed file size */
349 	if (a_out->a_data + a_out->a_text > attr.va_size) {
350 		error = EFAULT;
351 		goto cleanup;
352 	}
353 
354 	/*
355 	 * text/data/bss must not exceed limits
356 	 * XXX - this is not complete. it should check current usage PLUS
357 	 * the resources needed by this library.
358 	 */
359 	PROC_LOCK(td->td_proc);
360 	if (a_out->a_text > maxtsiz ||
361 	    a_out->a_data + bss_size > lim_cur(td->td_proc, RLIMIT_DATA) ||
362 	    racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
363 	    bss_size) != 0) {
364 		PROC_UNLOCK(td->td_proc);
365 		error = ENOMEM;
366 		goto cleanup;
367 	}
368 	PROC_UNLOCK(td->td_proc);
369 
370 	/*
371 	 * Prevent more writers.
372 	 * XXX: Note that if any of the VM operations fail below we don't
373 	 * clear this flag.
374 	 */
375 	vp->v_vflag |= VV_TEXT;
376 
377 	/*
378 	 * Lock no longer needed
379 	 */
380 	locked = 0;
381 	VOP_UNLOCK(vp, 0);
382 	VFS_UNLOCK_GIANT(vfslocked);
383 
384 	/*
385 	 * Check if file_offset page aligned. Currently we cannot handle
386 	 * misalinged file offsets, and so we read in the entire image
387 	 * (what a waste).
388 	 */
389 	if (file_offset & PAGE_MASK) {
390 #ifdef DEBUG
391 		printf("uselib: Non page aligned binary %lu\n", file_offset);
392 #endif
393 		/* Map text+data read/write/execute */
394 
395 		/* a_entry is the load address and is page aligned */
396 		vmaddr = trunc_page(a_out->a_entry);
397 
398 		/* get anon user mapping, read+write+execute */
399 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
400 		    &vmaddr, a_out->a_text + a_out->a_data, FALSE, VM_PROT_ALL,
401 		    VM_PROT_ALL, 0);
402 		if (error)
403 			goto cleanup;
404 
405 		error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
406 		    a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
407 		    td->td_ucred, NOCRED, &aresid, td);
408 		if (error != 0)
409 			goto cleanup;
410 		if (aresid != 0) {
411 			error = ENOEXEC;
412 			goto cleanup;
413 		}
414 	} else {
415 #ifdef DEBUG
416 		printf("uselib: Page aligned binary %lu\n", file_offset);
417 #endif
418 		/*
419 		 * for QMAGIC, a_entry is 20 bytes beyond the load address
420 		 * to skip the executable header
421 		 */
422 		vmaddr = trunc_page(a_out->a_entry);
423 
424 		/*
425 		 * Map it all into the process's space as a single
426 		 * copy-on-write "data" segment.
427 		 */
428 		error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr,
429 		    a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
430 		    MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
431 		if (error)
432 			goto cleanup;
433 	}
434 #ifdef DEBUG
435 	printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
436 	    ((long *)vmaddr)[1]);
437 #endif
438 	if (bss_size != 0) {
439 		/* Calculate BSS start address */
440 		vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
441 		    a_out->a_data;
442 
443 		/* allocate some 'anon' space */
444 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
445 		    &vmaddr, bss_size, FALSE, VM_PROT_ALL, VM_PROT_ALL, 0);
446 		if (error)
447 			goto cleanup;
448 	}
449 
450 cleanup:
451 	/* Unlock vnode if needed */
452 	if (locked) {
453 		VOP_UNLOCK(vp, 0);
454 		VFS_UNLOCK_GIANT(vfslocked);
455 	}
456 
457 	/* Release the temporary mapping. */
458 	if (a_out)
459 		kmem_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
460 
461 	return (error);
462 }
463 
464 #endif	/* __i386__ */
465 
466 int
467 linux_select(struct thread *td, struct linux_select_args *args)
468 {
469 	l_timeval ltv;
470 	struct timeval tv0, tv1, utv, *tvp;
471 	int error;
472 
473 #ifdef DEBUG
474 	if (ldebug(select))
475 		printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
476 		    (void *)args->readfds, (void *)args->writefds,
477 		    (void *)args->exceptfds, (void *)args->timeout);
478 #endif
479 
480 	/*
481 	 * Store current time for computation of the amount of
482 	 * time left.
483 	 */
484 	if (args->timeout) {
485 		if ((error = copyin(args->timeout, &ltv, sizeof(ltv))))
486 			goto select_out;
487 		utv.tv_sec = ltv.tv_sec;
488 		utv.tv_usec = ltv.tv_usec;
489 #ifdef DEBUG
490 		if (ldebug(select))
491 			printf(LMSG("incoming timeout (%jd/%ld)"),
492 			    (intmax_t)utv.tv_sec, utv.tv_usec);
493 #endif
494 
495 		if (itimerfix(&utv)) {
496 			/*
497 			 * The timeval was invalid.  Convert it to something
498 			 * valid that will act as it does under Linux.
499 			 */
500 			utv.tv_sec += utv.tv_usec / 1000000;
501 			utv.tv_usec %= 1000000;
502 			if (utv.tv_usec < 0) {
503 				utv.tv_sec -= 1;
504 				utv.tv_usec += 1000000;
505 			}
506 			if (utv.tv_sec < 0)
507 				timevalclear(&utv);
508 		}
509 		microtime(&tv0);
510 		tvp = &utv;
511 	} else
512 		tvp = NULL;
513 
514 	error = kern_select(td, args->nfds, args->readfds, args->writefds,
515 	    args->exceptfds, tvp, sizeof(l_int) * 8);
516 
517 #ifdef DEBUG
518 	if (ldebug(select))
519 		printf(LMSG("real select returns %d"), error);
520 #endif
521 	if (error)
522 		goto select_out;
523 
524 	if (args->timeout) {
525 		if (td->td_retval[0]) {
526 			/*
527 			 * Compute how much time was left of the timeout,
528 			 * by subtracting the current time and the time
529 			 * before we started the call, and subtracting
530 			 * that result from the user-supplied value.
531 			 */
532 			microtime(&tv1);
533 			timevalsub(&tv1, &tv0);
534 			timevalsub(&utv, &tv1);
535 			if (utv.tv_sec < 0)
536 				timevalclear(&utv);
537 		} else
538 			timevalclear(&utv);
539 #ifdef DEBUG
540 		if (ldebug(select))
541 			printf(LMSG("outgoing timeout (%jd/%ld)"),
542 			    (intmax_t)utv.tv_sec, utv.tv_usec);
543 #endif
544 		ltv.tv_sec = utv.tv_sec;
545 		ltv.tv_usec = utv.tv_usec;
546 		if ((error = copyout(&ltv, args->timeout, sizeof(ltv))))
547 			goto select_out;
548 	}
549 
550 select_out:
551 #ifdef DEBUG
552 	if (ldebug(select))
553 		printf(LMSG("select_out -> %d"), error);
554 #endif
555 	return (error);
556 }
557 
558 int
559 linux_mremap(struct thread *td, struct linux_mremap_args *args)
560 {
561 	struct munmap_args /* {
562 		void *addr;
563 		size_t len;
564 	} */ bsd_args;
565 	int error = 0;
566 
567 #ifdef DEBUG
568 	if (ldebug(mremap))
569 		printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
570 		    (void *)(uintptr_t)args->addr,
571 		    (unsigned long)args->old_len,
572 		    (unsigned long)args->new_len,
573 		    (unsigned long)args->flags);
574 #endif
575 
576 	if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
577 		td->td_retval[0] = 0;
578 		return (EINVAL);
579 	}
580 
581 	/*
582 	 * Check for the page alignment.
583 	 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
584 	 */
585 	if (args->addr & PAGE_MASK) {
586 		td->td_retval[0] = 0;
587 		return (EINVAL);
588 	}
589 
590 	args->new_len = round_page(args->new_len);
591 	args->old_len = round_page(args->old_len);
592 
593 	if (args->new_len > args->old_len) {
594 		td->td_retval[0] = 0;
595 		return (ENOMEM);
596 	}
597 
598 	if (args->new_len < args->old_len) {
599 		bsd_args.addr =
600 		    (caddr_t)((uintptr_t)args->addr + args->new_len);
601 		bsd_args.len = args->old_len - args->new_len;
602 		error = sys_munmap(td, &bsd_args);
603 	}
604 
605 	td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
606 	return (error);
607 }
608 
609 #define LINUX_MS_ASYNC       0x0001
610 #define LINUX_MS_INVALIDATE  0x0002
611 #define LINUX_MS_SYNC        0x0004
612 
613 int
614 linux_msync(struct thread *td, struct linux_msync_args *args)
615 {
616 	struct msync_args bsd_args;
617 
618 	bsd_args.addr = (caddr_t)(uintptr_t)args->addr;
619 	bsd_args.len = (uintptr_t)args->len;
620 	bsd_args.flags = args->fl & ~LINUX_MS_SYNC;
621 
622 	return (sys_msync(td, &bsd_args));
623 }
624 
625 int
626 linux_time(struct thread *td, struct linux_time_args *args)
627 {
628 	struct timeval tv;
629 	l_time_t tm;
630 	int error;
631 
632 #ifdef DEBUG
633 	if (ldebug(time))
634 		printf(ARGS(time, "*"));
635 #endif
636 
637 	microtime(&tv);
638 	tm = tv.tv_sec;
639 	if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
640 		return (error);
641 	td->td_retval[0] = tm;
642 	return (0);
643 }
644 
645 struct l_times_argv {
646 	l_clock_t	tms_utime;
647 	l_clock_t	tms_stime;
648 	l_clock_t	tms_cutime;
649 	l_clock_t	tms_cstime;
650 };
651 
652 
653 /*
654  * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
655  * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
656  * auxiliary vector entry.
657  */
658 #define	CLK_TCK		100
659 
660 #define	CONVOTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
661 #define	CONVNTCK(r)	(r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
662 
663 #define	CONVTCK(r)	(linux_kernver(td) >= LINUX_KERNVER_2004000 ?		\
664 			    CONVNTCK(r) : CONVOTCK(r))
665 
666 int
667 linux_times(struct thread *td, struct linux_times_args *args)
668 {
669 	struct timeval tv, utime, stime, cutime, cstime;
670 	struct l_times_argv tms;
671 	struct proc *p;
672 	int error;
673 
674 #ifdef DEBUG
675 	if (ldebug(times))
676 		printf(ARGS(times, "*"));
677 #endif
678 
679 	if (args->buf != NULL) {
680 		p = td->td_proc;
681 		PROC_LOCK(p);
682 		PROC_SLOCK(p);
683 		calcru(p, &utime, &stime);
684 		PROC_SUNLOCK(p);
685 		calccru(p, &cutime, &cstime);
686 		PROC_UNLOCK(p);
687 
688 		tms.tms_utime = CONVTCK(utime);
689 		tms.tms_stime = CONVTCK(stime);
690 
691 		tms.tms_cutime = CONVTCK(cutime);
692 		tms.tms_cstime = CONVTCK(cstime);
693 
694 		if ((error = copyout(&tms, args->buf, sizeof(tms))))
695 			return (error);
696 	}
697 
698 	microuptime(&tv);
699 	td->td_retval[0] = (int)CONVTCK(tv);
700 	return (0);
701 }
702 
703 int
704 linux_newuname(struct thread *td, struct linux_newuname_args *args)
705 {
706 	struct l_new_utsname utsname;
707 	char osname[LINUX_MAX_UTSNAME];
708 	char osrelease[LINUX_MAX_UTSNAME];
709 	char *p;
710 
711 #ifdef DEBUG
712 	if (ldebug(newuname))
713 		printf(ARGS(newuname, "*"));
714 #endif
715 
716 	linux_get_osname(td, osname);
717 	linux_get_osrelease(td, osrelease);
718 
719 	bzero(&utsname, sizeof(utsname));
720 	strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
721 	getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
722 	getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
723 	strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
724 	strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
725 	for (p = utsname.version; *p != '\0'; ++p)
726 		if (*p == '\n') {
727 			*p = '\0';
728 			break;
729 		}
730 	strlcpy(utsname.machine, linux_platform, LINUX_MAX_UTSNAME);
731 
732 	return (copyout(&utsname, args->buf, sizeof(utsname)));
733 }
734 
735 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
736 struct l_utimbuf {
737 	l_time_t l_actime;
738 	l_time_t l_modtime;
739 };
740 
741 int
742 linux_utime(struct thread *td, struct linux_utime_args *args)
743 {
744 	struct timeval tv[2], *tvp;
745 	struct l_utimbuf lut;
746 	char *fname;
747 	int error;
748 
749 	LCONVPATHEXIST(td, args->fname, &fname);
750 
751 #ifdef DEBUG
752 	if (ldebug(utime))
753 		printf(ARGS(utime, "%s, *"), fname);
754 #endif
755 
756 	if (args->times) {
757 		if ((error = copyin(args->times, &lut, sizeof lut))) {
758 			LFREEPATH(fname);
759 			return (error);
760 		}
761 		tv[0].tv_sec = lut.l_actime;
762 		tv[0].tv_usec = 0;
763 		tv[1].tv_sec = lut.l_modtime;
764 		tv[1].tv_usec = 0;
765 		tvp = tv;
766 	} else
767 		tvp = NULL;
768 
769 	error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
770 	LFREEPATH(fname);
771 	return (error);
772 }
773 
774 int
775 linux_utimes(struct thread *td, struct linux_utimes_args *args)
776 {
777 	l_timeval ltv[2];
778 	struct timeval tv[2], *tvp = NULL;
779 	char *fname;
780 	int error;
781 
782 	LCONVPATHEXIST(td, args->fname, &fname);
783 
784 #ifdef DEBUG
785 	if (ldebug(utimes))
786 		printf(ARGS(utimes, "%s, *"), fname);
787 #endif
788 
789 	if (args->tptr != NULL) {
790 		if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
791 			LFREEPATH(fname);
792 			return (error);
793 		}
794 		tv[0].tv_sec = ltv[0].tv_sec;
795 		tv[0].tv_usec = ltv[0].tv_usec;
796 		tv[1].tv_sec = ltv[1].tv_sec;
797 		tv[1].tv_usec = ltv[1].tv_usec;
798 		tvp = tv;
799 	}
800 
801 	error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
802 	LFREEPATH(fname);
803 	return (error);
804 }
805 
806 int
807 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
808 {
809 	l_timeval ltv[2];
810 	struct timeval tv[2], *tvp = NULL;
811 	char *fname;
812 	int error, dfd;
813 
814 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
815 	LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
816 
817 #ifdef DEBUG
818 	if (ldebug(futimesat))
819 		printf(ARGS(futimesat, "%s, *"), fname);
820 #endif
821 
822 	if (args->utimes != NULL) {
823 		if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
824 			LFREEPATH(fname);
825 			return (error);
826 		}
827 		tv[0].tv_sec = ltv[0].tv_sec;
828 		tv[0].tv_usec = ltv[0].tv_usec;
829 		tv[1].tv_sec = ltv[1].tv_sec;
830 		tv[1].tv_usec = ltv[1].tv_usec;
831 		tvp = tv;
832 	}
833 
834 	error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
835 	LFREEPATH(fname);
836 	return (error);
837 }
838 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
839 
840 int
841 linux_common_wait(struct thread *td, int pid, int *status,
842     int options, struct rusage *ru)
843 {
844 	int error, tmpstat;
845 
846 	error = kern_wait(td, pid, &tmpstat, options, ru);
847 	if (error)
848 		return (error);
849 
850 	if (status) {
851 		tmpstat &= 0xffff;
852 		if (WIFSIGNALED(tmpstat))
853 			tmpstat = (tmpstat & 0xffffff80) |
854 			    BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat));
855 		else if (WIFSTOPPED(tmpstat))
856 			tmpstat = (tmpstat & 0xffff00ff) |
857 			    (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8);
858 		error = copyout(&tmpstat, status, sizeof(int));
859 	}
860 
861 	return (error);
862 }
863 
864 int
865 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
866 {
867 	int options;
868 
869 #ifdef DEBUG
870 	if (ldebug(waitpid))
871 		printf(ARGS(waitpid, "%d, %p, %d"),
872 		    args->pid, (void *)args->status, args->options);
873 #endif
874 	/*
875 	 * this is necessary because the test in kern_wait doesn't work
876 	 * because we mess with the options here
877 	 */
878 	if (args->options & ~(WUNTRACED | WNOHANG | WCONTINUED | __WCLONE))
879 		return (EINVAL);
880 
881 	options = (args->options & (WNOHANG | WUNTRACED));
882 	/* WLINUXCLONE should be equal to __WCLONE, but we make sure */
883 	if (args->options & __WCLONE)
884 		options |= WLINUXCLONE;
885 
886 	return (linux_common_wait(td, args->pid, args->status, options, NULL));
887 }
888 
889 
890 int
891 linux_mknod(struct thread *td, struct linux_mknod_args *args)
892 {
893 	char *path;
894 	int error;
895 
896 	LCONVPATHCREAT(td, args->path, &path);
897 
898 #ifdef DEBUG
899 	if (ldebug(mknod))
900 		printf(ARGS(mknod, "%s, %d, %d"), path, args->mode, args->dev);
901 #endif
902 
903 	switch (args->mode & S_IFMT) {
904 	case S_IFIFO:
905 	case S_IFSOCK:
906 		error = kern_mkfifo(td, path, UIO_SYSSPACE, args->mode);
907 		break;
908 
909 	case S_IFCHR:
910 	case S_IFBLK:
911 		error = kern_mknod(td, path, UIO_SYSSPACE, args->mode,
912 		    args->dev);
913 		break;
914 
915 	case S_IFDIR:
916 		error = EPERM;
917 		break;
918 
919 	case 0:
920 		args->mode |= S_IFREG;
921 		/* FALLTHROUGH */
922 	case S_IFREG:
923 		error = kern_open(td, path, UIO_SYSSPACE,
924 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
925 		if (error == 0)
926 			kern_close(td, td->td_retval[0]);
927 		break;
928 
929 	default:
930 		error = EINVAL;
931 		break;
932 	}
933 	LFREEPATH(path);
934 	return (error);
935 }
936 
937 int
938 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
939 {
940 	char *path;
941 	int error, dfd;
942 
943 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
944 	LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
945 
946 #ifdef DEBUG
947 	if (ldebug(mknodat))
948 		printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev);
949 #endif
950 
951 	switch (args->mode & S_IFMT) {
952 	case S_IFIFO:
953 	case S_IFSOCK:
954 		error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
955 		break;
956 
957 	case S_IFCHR:
958 	case S_IFBLK:
959 		error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
960 		    args->dev);
961 		break;
962 
963 	case S_IFDIR:
964 		error = EPERM;
965 		break;
966 
967 	case 0:
968 		args->mode |= S_IFREG;
969 		/* FALLTHROUGH */
970 	case S_IFREG:
971 		error = kern_openat(td, dfd, path, UIO_SYSSPACE,
972 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
973 		if (error == 0)
974 			kern_close(td, td->td_retval[0]);
975 		break;
976 
977 	default:
978 		error = EINVAL;
979 		break;
980 	}
981 	LFREEPATH(path);
982 	return (error);
983 }
984 
985 /*
986  * UGH! This is just about the dumbest idea I've ever heard!!
987  */
988 int
989 linux_personality(struct thread *td, struct linux_personality_args *args)
990 {
991 #ifdef DEBUG
992 	if (ldebug(personality))
993 		printf(ARGS(personality, "%lu"), (unsigned long)args->per);
994 #endif
995 	if (args->per != 0)
996 		return (EINVAL);
997 
998 	/* Yes Jim, it's still a Linux... */
999 	td->td_retval[0] = 0;
1000 	return (0);
1001 }
1002 
1003 struct l_itimerval {
1004 	l_timeval it_interval;
1005 	l_timeval it_value;
1006 };
1007 
1008 #define	B2L_ITIMERVAL(bip, lip) 					\
1009 	(bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec;		\
1010 	(bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec;	\
1011 	(bip)->it_value.tv_sec = (lip)->it_value.tv_sec;		\
1012 	(bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1013 
1014 int
1015 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1016 {
1017 	int error;
1018 	struct l_itimerval ls;
1019 	struct itimerval aitv, oitv;
1020 
1021 #ifdef DEBUG
1022 	if (ldebug(setitimer))
1023 		printf(ARGS(setitimer, "%p, %p"),
1024 		    (void *)uap->itv, (void *)uap->oitv);
1025 #endif
1026 
1027 	if (uap->itv == NULL) {
1028 		uap->itv = uap->oitv;
1029 		return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1030 	}
1031 
1032 	error = copyin(uap->itv, &ls, sizeof(ls));
1033 	if (error != 0)
1034 		return (error);
1035 	B2L_ITIMERVAL(&aitv, &ls);
1036 #ifdef DEBUG
1037 	if (ldebug(setitimer)) {
1038 		printf("setitimer: value: sec: %jd, usec: %ld\n",
1039 		    (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1040 		printf("setitimer: interval: sec: %jd, usec: %ld\n",
1041 		    (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1042 	}
1043 #endif
1044 	error = kern_setitimer(td, uap->which, &aitv, &oitv);
1045 	if (error != 0 || uap->oitv == NULL)
1046 		return (error);
1047 	B2L_ITIMERVAL(&ls, &oitv);
1048 
1049 	return (copyout(&ls, uap->oitv, sizeof(ls)));
1050 }
1051 
1052 int
1053 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1054 {
1055 	int error;
1056 	struct l_itimerval ls;
1057 	struct itimerval aitv;
1058 
1059 #ifdef DEBUG
1060 	if (ldebug(getitimer))
1061 		printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1062 #endif
1063 	error = kern_getitimer(td, uap->which, &aitv);
1064 	if (error != 0)
1065 		return (error);
1066 	B2L_ITIMERVAL(&ls, &aitv);
1067 	return (copyout(&ls, uap->itv, sizeof(ls)));
1068 }
1069 
1070 int
1071 linux_nice(struct thread *td, struct linux_nice_args *args)
1072 {
1073 	struct setpriority_args bsd_args;
1074 
1075 	bsd_args.which = PRIO_PROCESS;
1076 	bsd_args.who = 0;		/* current process */
1077 	bsd_args.prio = args->inc;
1078 	return (sys_setpriority(td, &bsd_args));
1079 }
1080 
1081 int
1082 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1083 {
1084 	struct ucred *newcred, *oldcred;
1085 	l_gid_t *linux_gidset;
1086 	gid_t *bsd_gidset;
1087 	int ngrp, error;
1088 	struct proc *p;
1089 
1090 	ngrp = args->gidsetsize;
1091 	if (ngrp < 0 || ngrp >= ngroups_max + 1)
1092 		return (EINVAL);
1093 	linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_TEMP, M_WAITOK);
1094 	error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1095 	if (error)
1096 		goto out;
1097 	newcred = crget();
1098 	p = td->td_proc;
1099 	PROC_LOCK(p);
1100 	oldcred = crcopysafe(p, newcred);
1101 
1102 	/*
1103 	 * cr_groups[0] holds egid. Setting the whole set from
1104 	 * the supplied set will cause egid to be changed too.
1105 	 * Keep cr_groups[0] unchanged to prevent that.
1106 	 */
1107 
1108 	if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) {
1109 		PROC_UNLOCK(p);
1110 		crfree(newcred);
1111 		goto out;
1112 	}
1113 
1114 	if (ngrp > 0) {
1115 		newcred->cr_ngroups = ngrp + 1;
1116 
1117 		bsd_gidset = newcred->cr_groups;
1118 		ngrp--;
1119 		while (ngrp >= 0) {
1120 			bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1121 			ngrp--;
1122 		}
1123 	} else
1124 		newcred->cr_ngroups = 1;
1125 
1126 	setsugid(p);
1127 	p->p_ucred = newcred;
1128 	PROC_UNLOCK(p);
1129 	crfree(oldcred);
1130 	error = 0;
1131 out:
1132 	free(linux_gidset, M_TEMP);
1133 	return (error);
1134 }
1135 
1136 int
1137 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1138 {
1139 	struct ucred *cred;
1140 	l_gid_t *linux_gidset;
1141 	gid_t *bsd_gidset;
1142 	int bsd_gidsetsz, ngrp, error;
1143 
1144 	cred = td->td_ucred;
1145 	bsd_gidset = cred->cr_groups;
1146 	bsd_gidsetsz = cred->cr_ngroups - 1;
1147 
1148 	/*
1149 	 * cr_groups[0] holds egid. Returning the whole set
1150 	 * here will cause a duplicate. Exclude cr_groups[0]
1151 	 * to prevent that.
1152 	 */
1153 
1154 	if ((ngrp = args->gidsetsize) == 0) {
1155 		td->td_retval[0] = bsd_gidsetsz;
1156 		return (0);
1157 	}
1158 
1159 	if (ngrp < bsd_gidsetsz)
1160 		return (EINVAL);
1161 
1162 	ngrp = 0;
1163 	linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1164 	    M_TEMP, M_WAITOK);
1165 	while (ngrp < bsd_gidsetsz) {
1166 		linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1167 		ngrp++;
1168 	}
1169 
1170 	error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1171 	free(linux_gidset, M_TEMP);
1172 	if (error)
1173 		return (error);
1174 
1175 	td->td_retval[0] = ngrp;
1176 	return (0);
1177 }
1178 
1179 int
1180 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1181 {
1182 	struct rlimit bsd_rlim;
1183 	struct l_rlimit rlim;
1184 	u_int which;
1185 	int error;
1186 
1187 #ifdef DEBUG
1188 	if (ldebug(setrlimit))
1189 		printf(ARGS(setrlimit, "%d, %p"),
1190 		    args->resource, (void *)args->rlim);
1191 #endif
1192 
1193 	if (args->resource >= LINUX_RLIM_NLIMITS)
1194 		return (EINVAL);
1195 
1196 	which = linux_to_bsd_resource[args->resource];
1197 	if (which == -1)
1198 		return (EINVAL);
1199 
1200 	error = copyin(args->rlim, &rlim, sizeof(rlim));
1201 	if (error)
1202 		return (error);
1203 
1204 	bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1205 	bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1206 	return (kern_setrlimit(td, which, &bsd_rlim));
1207 }
1208 
1209 int
1210 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1211 {
1212 	struct l_rlimit rlim;
1213 	struct proc *p = td->td_proc;
1214 	struct rlimit bsd_rlim;
1215 	u_int which;
1216 
1217 #ifdef DEBUG
1218 	if (ldebug(old_getrlimit))
1219 		printf(ARGS(old_getrlimit, "%d, %p"),
1220 		    args->resource, (void *)args->rlim);
1221 #endif
1222 
1223 	if (args->resource >= LINUX_RLIM_NLIMITS)
1224 		return (EINVAL);
1225 
1226 	which = linux_to_bsd_resource[args->resource];
1227 	if (which == -1)
1228 		return (EINVAL);
1229 
1230 	PROC_LOCK(p);
1231 	lim_rlimit(p, which, &bsd_rlim);
1232 	PROC_UNLOCK(p);
1233 
1234 #ifdef COMPAT_LINUX32
1235 	rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1236 	if (rlim.rlim_cur == UINT_MAX)
1237 		rlim.rlim_cur = INT_MAX;
1238 	rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1239 	if (rlim.rlim_max == UINT_MAX)
1240 		rlim.rlim_max = INT_MAX;
1241 #else
1242 	rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1243 	if (rlim.rlim_cur == ULONG_MAX)
1244 		rlim.rlim_cur = LONG_MAX;
1245 	rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1246 	if (rlim.rlim_max == ULONG_MAX)
1247 		rlim.rlim_max = LONG_MAX;
1248 #endif
1249 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1250 }
1251 
1252 int
1253 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1254 {
1255 	struct l_rlimit rlim;
1256 	struct proc *p = td->td_proc;
1257 	struct rlimit bsd_rlim;
1258 	u_int which;
1259 
1260 #ifdef DEBUG
1261 	if (ldebug(getrlimit))
1262 		printf(ARGS(getrlimit, "%d, %p"),
1263 		    args->resource, (void *)args->rlim);
1264 #endif
1265 
1266 	if (args->resource >= LINUX_RLIM_NLIMITS)
1267 		return (EINVAL);
1268 
1269 	which = linux_to_bsd_resource[args->resource];
1270 	if (which == -1)
1271 		return (EINVAL);
1272 
1273 	PROC_LOCK(p);
1274 	lim_rlimit(p, which, &bsd_rlim);
1275 	PROC_UNLOCK(p);
1276 
1277 	rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1278 	rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1279 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1280 }
1281 
1282 int
1283 linux_sched_setscheduler(struct thread *td,
1284     struct linux_sched_setscheduler_args *args)
1285 {
1286 	struct sched_setscheduler_args bsd;
1287 
1288 #ifdef DEBUG
1289 	if (ldebug(sched_setscheduler))
1290 		printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1291 		    args->pid, args->policy, (const void *)args->param);
1292 #endif
1293 
1294 	switch (args->policy) {
1295 	case LINUX_SCHED_OTHER:
1296 		bsd.policy = SCHED_OTHER;
1297 		break;
1298 	case LINUX_SCHED_FIFO:
1299 		bsd.policy = SCHED_FIFO;
1300 		break;
1301 	case LINUX_SCHED_RR:
1302 		bsd.policy = SCHED_RR;
1303 		break;
1304 	default:
1305 		return (EINVAL);
1306 	}
1307 
1308 	bsd.pid = args->pid;
1309 	bsd.param = (struct sched_param *)args->param;
1310 	return (sys_sched_setscheduler(td, &bsd));
1311 }
1312 
1313 int
1314 linux_sched_getscheduler(struct thread *td,
1315     struct linux_sched_getscheduler_args *args)
1316 {
1317 	struct sched_getscheduler_args bsd;
1318 	int error;
1319 
1320 #ifdef DEBUG
1321 	if (ldebug(sched_getscheduler))
1322 		printf(ARGS(sched_getscheduler, "%d"), args->pid);
1323 #endif
1324 
1325 	bsd.pid = args->pid;
1326 	error = sys_sched_getscheduler(td, &bsd);
1327 
1328 	switch (td->td_retval[0]) {
1329 	case SCHED_OTHER:
1330 		td->td_retval[0] = LINUX_SCHED_OTHER;
1331 		break;
1332 	case SCHED_FIFO:
1333 		td->td_retval[0] = LINUX_SCHED_FIFO;
1334 		break;
1335 	case SCHED_RR:
1336 		td->td_retval[0] = LINUX_SCHED_RR;
1337 		break;
1338 	}
1339 
1340 	return (error);
1341 }
1342 
1343 int
1344 linux_sched_get_priority_max(struct thread *td,
1345     struct linux_sched_get_priority_max_args *args)
1346 {
1347 	struct sched_get_priority_max_args bsd;
1348 
1349 #ifdef DEBUG
1350 	if (ldebug(sched_get_priority_max))
1351 		printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1352 #endif
1353 
1354 	switch (args->policy) {
1355 	case LINUX_SCHED_OTHER:
1356 		bsd.policy = SCHED_OTHER;
1357 		break;
1358 	case LINUX_SCHED_FIFO:
1359 		bsd.policy = SCHED_FIFO;
1360 		break;
1361 	case LINUX_SCHED_RR:
1362 		bsd.policy = SCHED_RR;
1363 		break;
1364 	default:
1365 		return (EINVAL);
1366 	}
1367 	return (sys_sched_get_priority_max(td, &bsd));
1368 }
1369 
1370 int
1371 linux_sched_get_priority_min(struct thread *td,
1372     struct linux_sched_get_priority_min_args *args)
1373 {
1374 	struct sched_get_priority_min_args bsd;
1375 
1376 #ifdef DEBUG
1377 	if (ldebug(sched_get_priority_min))
1378 		printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1379 #endif
1380 
1381 	switch (args->policy) {
1382 	case LINUX_SCHED_OTHER:
1383 		bsd.policy = SCHED_OTHER;
1384 		break;
1385 	case LINUX_SCHED_FIFO:
1386 		bsd.policy = SCHED_FIFO;
1387 		break;
1388 	case LINUX_SCHED_RR:
1389 		bsd.policy = SCHED_RR;
1390 		break;
1391 	default:
1392 		return (EINVAL);
1393 	}
1394 	return (sys_sched_get_priority_min(td, &bsd));
1395 }
1396 
1397 #define REBOOT_CAD_ON	0x89abcdef
1398 #define REBOOT_CAD_OFF	0
1399 #define REBOOT_HALT	0xcdef0123
1400 #define REBOOT_RESTART	0x01234567
1401 #define REBOOT_RESTART2	0xA1B2C3D4
1402 #define REBOOT_POWEROFF	0x4321FEDC
1403 #define REBOOT_MAGIC1	0xfee1dead
1404 #define REBOOT_MAGIC2	0x28121969
1405 #define REBOOT_MAGIC2A	0x05121996
1406 #define REBOOT_MAGIC2B	0x16041998
1407 
1408 int
1409 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1410 {
1411 	struct reboot_args bsd_args;
1412 
1413 #ifdef DEBUG
1414 	if (ldebug(reboot))
1415 		printf(ARGS(reboot, "0x%x"), args->cmd);
1416 #endif
1417 
1418 	if (args->magic1 != REBOOT_MAGIC1)
1419 		return (EINVAL);
1420 
1421 	switch (args->magic2) {
1422 	case REBOOT_MAGIC2:
1423 	case REBOOT_MAGIC2A:
1424 	case REBOOT_MAGIC2B:
1425 		break;
1426 	default:
1427 		return (EINVAL);
1428 	}
1429 
1430 	switch (args->cmd) {
1431 	case REBOOT_CAD_ON:
1432 	case REBOOT_CAD_OFF:
1433 		return (priv_check(td, PRIV_REBOOT));
1434 	case REBOOT_HALT:
1435 		bsd_args.opt = RB_HALT;
1436 		break;
1437 	case REBOOT_RESTART:
1438 	case REBOOT_RESTART2:
1439 		bsd_args.opt = 0;
1440 		break;
1441 	case REBOOT_POWEROFF:
1442 		bsd_args.opt = RB_POWEROFF;
1443 		break;
1444 	default:
1445 		return (EINVAL);
1446 	}
1447 	return (sys_reboot(td, &bsd_args));
1448 }
1449 
1450 
1451 /*
1452  * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1453  * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
1454  * are assumed to be preserved. The following lightweight syscalls fixes
1455  * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
1456  *
1457  * linux_getpid() - MP SAFE
1458  * linux_getgid() - MP SAFE
1459  * linux_getuid() - MP SAFE
1460  */
1461 
1462 int
1463 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1464 {
1465 	struct linux_emuldata *em;
1466 
1467 #ifdef DEBUG
1468 	if (ldebug(getpid))
1469 		printf(ARGS(getpid, ""));
1470 #endif
1471 
1472 	if (linux_use26(td)) {
1473 		em = em_find(td->td_proc, EMUL_DONTLOCK);
1474 		KASSERT(em != NULL, ("getpid: emuldata not found.\n"));
1475 		td->td_retval[0] = em->shared->group_pid;
1476 	} else {
1477 		td->td_retval[0] = td->td_proc->p_pid;
1478 	}
1479 
1480 	return (0);
1481 }
1482 
1483 int
1484 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1485 {
1486 
1487 #ifdef DEBUG
1488 	if (ldebug(gettid))
1489 		printf(ARGS(gettid, ""));
1490 #endif
1491 
1492 	td->td_retval[0] = td->td_proc->p_pid;
1493 	return (0);
1494 }
1495 
1496 
1497 int
1498 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1499 {
1500 	struct linux_emuldata *em;
1501 	struct proc *p, *pp;
1502 
1503 #ifdef DEBUG
1504 	if (ldebug(getppid))
1505 		printf(ARGS(getppid, ""));
1506 #endif
1507 
1508 	if (!linux_use26(td)) {
1509 		PROC_LOCK(td->td_proc);
1510 		td->td_retval[0] = td->td_proc->p_pptr->p_pid;
1511 		PROC_UNLOCK(td->td_proc);
1512 		return (0);
1513 	}
1514 
1515 	em = em_find(td->td_proc, EMUL_DONTLOCK);
1516 
1517 	KASSERT(em != NULL, ("getppid: process emuldata not found.\n"));
1518 
1519 	/* find the group leader */
1520 	p = pfind(em->shared->group_pid);
1521 
1522 	if (p == NULL) {
1523 #ifdef DEBUG
1524 	   	printf(LMSG("parent process not found.\n"));
1525 #endif
1526 		return (0);
1527 	}
1528 
1529 	pp = p->p_pptr;		/* switch to parent */
1530 	PROC_LOCK(pp);
1531 	PROC_UNLOCK(p);
1532 
1533 	/* if its also linux process */
1534 	if (pp->p_sysent == &elf_linux_sysvec) {
1535 		em = em_find(pp, EMUL_DONTLOCK);
1536 		KASSERT(em != NULL, ("getppid: parent emuldata not found.\n"));
1537 
1538 		td->td_retval[0] = em->shared->group_pid;
1539 	} else
1540 		td->td_retval[0] = pp->p_pid;
1541 
1542 	PROC_UNLOCK(pp);
1543 
1544 	return (0);
1545 }
1546 
1547 int
1548 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1549 {
1550 
1551 #ifdef DEBUG
1552 	if (ldebug(getgid))
1553 		printf(ARGS(getgid, ""));
1554 #endif
1555 
1556 	td->td_retval[0] = td->td_ucred->cr_rgid;
1557 	return (0);
1558 }
1559 
1560 int
1561 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1562 {
1563 
1564 #ifdef DEBUG
1565 	if (ldebug(getuid))
1566 		printf(ARGS(getuid, ""));
1567 #endif
1568 
1569 	td->td_retval[0] = td->td_ucred->cr_ruid;
1570 	return (0);
1571 }
1572 
1573 
1574 int
1575 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1576 {
1577 	struct getsid_args bsd;
1578 
1579 #ifdef DEBUG
1580 	if (ldebug(getsid))
1581 		printf(ARGS(getsid, "%i"), args->pid);
1582 #endif
1583 
1584 	bsd.pid = args->pid;
1585 	return (sys_getsid(td, &bsd));
1586 }
1587 
1588 int
1589 linux_nosys(struct thread *td, struct nosys_args *ignore)
1590 {
1591 
1592 	return (ENOSYS);
1593 }
1594 
1595 int
1596 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1597 {
1598 	struct getpriority_args bsd_args;
1599 	int error;
1600 
1601 #ifdef DEBUG
1602 	if (ldebug(getpriority))
1603 		printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
1604 #endif
1605 
1606 	bsd_args.which = args->which;
1607 	bsd_args.who = args->who;
1608 	error = sys_getpriority(td, &bsd_args);
1609 	td->td_retval[0] = 20 - td->td_retval[0];
1610 	return (error);
1611 }
1612 
1613 int
1614 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1615 {
1616 	int name[2];
1617 
1618 #ifdef DEBUG
1619 	if (ldebug(sethostname))
1620 		printf(ARGS(sethostname, "*, %i"), args->len);
1621 #endif
1622 
1623 	name[0] = CTL_KERN;
1624 	name[1] = KERN_HOSTNAME;
1625 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1626 	    args->len, 0, 0));
1627 }
1628 
1629 int
1630 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1631 {
1632 	int name[2];
1633 
1634 #ifdef DEBUG
1635 	if (ldebug(setdomainname))
1636 		printf(ARGS(setdomainname, "*, %i"), args->len);
1637 #endif
1638 
1639 	name[0] = CTL_KERN;
1640 	name[1] = KERN_NISDOMAINNAME;
1641 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1642 	    args->len, 0, 0));
1643 }
1644 
1645 int
1646 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1647 {
1648 	struct linux_emuldata *em;
1649 
1650 #ifdef DEBUG
1651 	if (ldebug(exit_group))
1652 		printf(ARGS(exit_group, "%i"), args->error_code);
1653 #endif
1654 
1655 	em = em_find(td->td_proc, EMUL_DONTLOCK);
1656 	if (em->shared->refs > 1) {
1657 		EMUL_SHARED_WLOCK(&emul_shared_lock);
1658 		em->shared->flags |= EMUL_SHARED_HASXSTAT;
1659 		em->shared->xstat = W_EXITCODE(args->error_code, 0);
1660 		EMUL_SHARED_WUNLOCK(&emul_shared_lock);
1661 		if (linux_use26(td))
1662 			linux_kill_threads(td, SIGKILL);
1663 	}
1664 
1665 	/*
1666 	 * XXX: we should send a signal to the parent if
1667 	 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1668 	 * as it doesnt occur often.
1669 	 */
1670 	exit1(td, W_EXITCODE(args->error_code, 0));
1671 
1672 	return (0);
1673 }
1674 
1675 #define _LINUX_CAPABILITY_VERSION  0x19980330
1676 
1677 struct l_user_cap_header {
1678 	l_int	version;
1679 	l_int	pid;
1680 };
1681 
1682 struct l_user_cap_data {
1683 	l_int	effective;
1684 	l_int	permitted;
1685 	l_int	inheritable;
1686 };
1687 
1688 int
1689 linux_capget(struct thread *td, struct linux_capget_args *args)
1690 {
1691 	struct l_user_cap_header luch;
1692 	struct l_user_cap_data lucd;
1693 	int error;
1694 
1695 	if (args->hdrp == NULL)
1696 		return (EFAULT);
1697 
1698 	error = copyin(args->hdrp, &luch, sizeof(luch));
1699 	if (error != 0)
1700 		return (error);
1701 
1702 	if (luch.version != _LINUX_CAPABILITY_VERSION) {
1703 		luch.version = _LINUX_CAPABILITY_VERSION;
1704 		error = copyout(&luch, args->hdrp, sizeof(luch));
1705 		if (error)
1706 			return (error);
1707 		return (EINVAL);
1708 	}
1709 
1710 	if (luch.pid)
1711 		return (EPERM);
1712 
1713 	if (args->datap) {
1714 		/*
1715 		 * The current implementation doesn't support setting
1716 		 * a capability (it's essentially a stub) so indicate
1717 		 * that no capabilities are currently set or available
1718 		 * to request.
1719 		 */
1720 		bzero (&lucd, sizeof(lucd));
1721 		error = copyout(&lucd, args->datap, sizeof(lucd));
1722 	}
1723 
1724 	return (error);
1725 }
1726 
1727 int
1728 linux_capset(struct thread *td, struct linux_capset_args *args)
1729 {
1730 	struct l_user_cap_header luch;
1731 	struct l_user_cap_data lucd;
1732 	int error;
1733 
1734 	if (args->hdrp == NULL || args->datap == NULL)
1735 		return (EFAULT);
1736 
1737 	error = copyin(args->hdrp, &luch, sizeof(luch));
1738 	if (error != 0)
1739 		return (error);
1740 
1741 	if (luch.version != _LINUX_CAPABILITY_VERSION) {
1742 		luch.version = _LINUX_CAPABILITY_VERSION;
1743 		error = copyout(&luch, args->hdrp, sizeof(luch));
1744 		if (error)
1745 			return (error);
1746 		return (EINVAL);
1747 	}
1748 
1749 	if (luch.pid)
1750 		return (EPERM);
1751 
1752 	error = copyin(args->datap, &lucd, sizeof(lucd));
1753 	if (error != 0)
1754 		return (error);
1755 
1756 	/* We currently don't support setting any capabilities. */
1757 	if (lucd.effective || lucd.permitted || lucd.inheritable) {
1758 		linux_msg(td,
1759 			  "capset effective=0x%x, permitted=0x%x, "
1760 			  "inheritable=0x%x is not implemented",
1761 			  (int)lucd.effective, (int)lucd.permitted,
1762 			  (int)lucd.inheritable);
1763 		return (EPERM);
1764 	}
1765 
1766 	return (0);
1767 }
1768 
1769 int
1770 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1771 {
1772 	int error = 0, max_size;
1773 	struct proc *p = td->td_proc;
1774 	char comm[LINUX_MAX_COMM_LEN];
1775 	struct linux_emuldata *em;
1776 	int pdeath_signal;
1777 
1778 #ifdef DEBUG
1779 	if (ldebug(prctl))
1780 		printf(ARGS(prctl, "%d, %d, %d, %d, %d"), args->option,
1781 		    args->arg2, args->arg3, args->arg4, args->arg5);
1782 #endif
1783 
1784 	switch (args->option) {
1785 	case LINUX_PR_SET_PDEATHSIG:
1786 		if (!LINUX_SIG_VALID(args->arg2))
1787 			return (EINVAL);
1788 		em = em_find(p, EMUL_DOLOCK);
1789 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1790 		em->pdeath_signal = args->arg2;
1791 		EMUL_UNLOCK(&emul_lock);
1792 		break;
1793 	case LINUX_PR_GET_PDEATHSIG:
1794 		em = em_find(p, EMUL_DOLOCK);
1795 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1796 		pdeath_signal = em->pdeath_signal;
1797 		EMUL_UNLOCK(&emul_lock);
1798 		error = copyout(&pdeath_signal,
1799 		    (void *)(register_t)args->arg2,
1800 		    sizeof(pdeath_signal));
1801 		break;
1802 	case LINUX_PR_GET_KEEPCAPS:
1803 		/*
1804 		 * Indicate that we always clear the effective and
1805 		 * permitted capability sets when the user id becomes
1806 		 * non-zero (actually the capability sets are simply
1807 		 * always zero in the current implementation).
1808 		 */
1809 		td->td_retval[0] = 0;
1810 		break;
1811 	case LINUX_PR_SET_KEEPCAPS:
1812 		/*
1813 		 * Ignore requests to keep the effective and permitted
1814 		 * capability sets when the user id becomes non-zero.
1815 		 */
1816 		break;
1817 	case LINUX_PR_SET_NAME:
1818 		/*
1819 		 * To be on the safe side we need to make sure to not
1820 		 * overflow the size a linux program expects. We already
1821 		 * do this here in the copyin, so that we don't need to
1822 		 * check on copyout.
1823 		 */
1824 		max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1825 		error = copyinstr((void *)(register_t)args->arg2, comm,
1826 		    max_size, NULL);
1827 
1828 		/* Linux silently truncates the name if it is too long. */
1829 		if (error == ENAMETOOLONG) {
1830 			/*
1831 			 * XXX: copyinstr() isn't documented to populate the
1832 			 * array completely, so do a copyin() to be on the
1833 			 * safe side. This should be changed in case
1834 			 * copyinstr() is changed to guarantee this.
1835 			 */
1836 			error = copyin((void *)(register_t)args->arg2, comm,
1837 			    max_size - 1);
1838 			comm[max_size - 1] = '\0';
1839 		}
1840 		if (error)
1841 			return (error);
1842 
1843 		PROC_LOCK(p);
1844 		strlcpy(p->p_comm, comm, sizeof(p->p_comm));
1845 		PROC_UNLOCK(p);
1846 		break;
1847 	case LINUX_PR_GET_NAME:
1848 		PROC_LOCK(p);
1849 		strlcpy(comm, p->p_comm, sizeof(comm));
1850 		PROC_UNLOCK(p);
1851 		error = copyout(comm, (void *)(register_t)args->arg2,
1852 		    strlen(comm) + 1);
1853 		break;
1854 	default:
1855 		error = EINVAL;
1856 		break;
1857 	}
1858 
1859 	return (error);
1860 }
1861 
1862 /*
1863  * Get affinity of a process.
1864  */
1865 int
1866 linux_sched_getaffinity(struct thread *td,
1867     struct linux_sched_getaffinity_args *args)
1868 {
1869 	int error;
1870 	struct cpuset_getaffinity_args cga;
1871 
1872 #ifdef DEBUG
1873 	if (ldebug(sched_getaffinity))
1874 		printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid,
1875 		    args->len);
1876 #endif
1877 	if (args->len < sizeof(cpuset_t))
1878 		return (EINVAL);
1879 
1880 	cga.level = CPU_LEVEL_WHICH;
1881 	cga.which = CPU_WHICH_PID;
1882 	cga.id = args->pid;
1883 	cga.cpusetsize = sizeof(cpuset_t);
1884 	cga.mask = (cpuset_t *) args->user_mask_ptr;
1885 
1886 	if ((error = sys_cpuset_getaffinity(td, &cga)) == 0)
1887 		td->td_retval[0] = sizeof(cpuset_t);
1888 
1889 	return (error);
1890 }
1891 
1892 /*
1893  *  Set affinity of a process.
1894  */
1895 int
1896 linux_sched_setaffinity(struct thread *td,
1897     struct linux_sched_setaffinity_args *args)
1898 {
1899 	struct cpuset_setaffinity_args csa;
1900 
1901 #ifdef DEBUG
1902 	if (ldebug(sched_setaffinity))
1903 		printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid,
1904 		    args->len);
1905 #endif
1906 	if (args->len < sizeof(cpuset_t))
1907 		return (EINVAL);
1908 
1909 	csa.level = CPU_LEVEL_WHICH;
1910 	csa.which = CPU_WHICH_PID;
1911 	csa.id = args->pid;
1912 	csa.cpusetsize = sizeof(cpuset_t);
1913 	csa.mask = (cpuset_t *) args->user_mask_ptr;
1914 
1915 	return (sys_cpuset_setaffinity(td, &csa));
1916 }
1917