linux_misc.c revision 301051
1/*-
2 * Copyright (c) 2002 Doug Rabson
3 * Copyright (c) 1994-1995 S��ren 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: releng/10.2/sys/compat/linux/linux_misc.c 301051 2016-05-31 16:55:45Z glebius $");
32
33#include "opt_compat.h"
34#include "opt_kdtrace.h"
35
36#include <sys/param.h>
37#include <sys/blist.h>
38#include <sys/fcntl.h>
39#if defined(__i386__)
40#include <sys/imgact_aout.h>
41#endif
42#include <sys/jail.h>
43#include <sys/kernel.h>
44#include <sys/limits.h>
45#include <sys/lock.h>
46#include <sys/malloc.h>
47#include <sys/mman.h>
48#include <sys/mount.h>
49#include <sys/mutex.h>
50#include <sys/namei.h>
51#include <sys/priv.h>
52#include <sys/proc.h>
53#include <sys/reboot.h>
54#include <sys/racct.h>
55#include <sys/resourcevar.h>
56#include <sys/sched.h>
57#include <sys/sdt.h>
58#include <sys/signalvar.h>
59#include <sys/stat.h>
60#include <sys/syscallsubr.h>
61#include <sys/sysctl.h>
62#include <sys/sysproto.h>
63#include <sys/systm.h>
64#include <sys/time.h>
65#include <sys/vmmeter.h>
66#include <sys/vnode.h>
67#include <sys/wait.h>
68#include <sys/cpuset.h>
69
70#include <security/mac/mac_framework.h>
71
72#include <vm/vm.h>
73#include <vm/pmap.h>
74#include <vm/vm_kern.h>
75#include <vm/vm_map.h>
76#include <vm/vm_extern.h>
77#include <vm/vm_object.h>
78#include <vm/swap_pager.h>
79
80#ifdef COMPAT_LINUX32
81#include <machine/../linux32/linux.h>
82#include <machine/../linux32/linux32_proto.h>
83#else
84#include <machine/../linux/linux.h>
85#include <machine/../linux/linux_proto.h>
86#endif
87
88#include <compat/linux/linux_dtrace.h>
89#include <compat/linux/linux_file.h>
90#include <compat/linux/linux_mib.h>
91#include <compat/linux/linux_signal.h>
92#include <compat/linux/linux_util.h>
93#include <compat/linux/linux_sysproto.h>
94#include <compat/linux/linux_emul.h>
95#include <compat/linux/linux_misc.h>
96
97/* DTrace init */
98LIN_SDT_PROVIDER_DECLARE(LINUX_DTRACE);
99
100/* Linuxulator-global DTrace probes */
101LIN_SDT_PROBE_DECLARE(locks, emul_lock, locked);
102LIN_SDT_PROBE_DECLARE(locks, emul_lock, unlock);
103LIN_SDT_PROBE_DECLARE(locks, emul_shared_rlock, locked);
104LIN_SDT_PROBE_DECLARE(locks, emul_shared_rlock, unlock);
105LIN_SDT_PROBE_DECLARE(locks, emul_shared_wlock, locked);
106LIN_SDT_PROBE_DECLARE(locks, emul_shared_wlock, unlock);
107
108int stclohz;				/* Statistics clock frequency */
109
110static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
111	RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
112	RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
113	RLIMIT_MEMLOCK, RLIMIT_AS
114};
115
116struct l_sysinfo {
117	l_long		uptime;		/* Seconds since boot */
118	l_ulong		loads[3];	/* 1, 5, and 15 minute load averages */
119#define LINUX_SYSINFO_LOADS_SCALE 65536
120	l_ulong		totalram;	/* Total usable main memory size */
121	l_ulong		freeram;	/* Available memory size */
122	l_ulong		sharedram;	/* Amount of shared memory */
123	l_ulong		bufferram;	/* Memory used by buffers */
124	l_ulong		totalswap;	/* Total swap space size */
125	l_ulong		freeswap;	/* swap space still available */
126	l_ushort	procs;		/* Number of current processes */
127	l_ushort	pads;
128	l_ulong		totalbig;
129	l_ulong		freebig;
130	l_uint		mem_unit;
131	char		_f[20-2*sizeof(l_long)-sizeof(l_int)];	/* padding */
132};
133int
134linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
135{
136	struct l_sysinfo sysinfo;
137	vm_object_t object;
138	int i, j;
139	struct timespec ts;
140
141	bzero(&sysinfo, sizeof(sysinfo));
142	getnanouptime(&ts);
143	if (ts.tv_nsec != 0)
144		ts.tv_sec++;
145	sysinfo.uptime = ts.tv_sec;
146
147	/* Use the information from the mib to get our load averages */
148	for (i = 0; i < 3; i++)
149		sysinfo.loads[i] = averunnable.ldavg[i] *
150		    LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
151
152	sysinfo.totalram = physmem * PAGE_SIZE;
153	sysinfo.freeram = sysinfo.totalram - cnt.v_wire_count * PAGE_SIZE;
154
155	sysinfo.sharedram = 0;
156	mtx_lock(&vm_object_list_mtx);
157	TAILQ_FOREACH(object, &vm_object_list, object_list)
158		if (object->shadow_count > 1)
159			sysinfo.sharedram += object->resident_page_count;
160	mtx_unlock(&vm_object_list_mtx);
161
162	sysinfo.sharedram *= PAGE_SIZE;
163	sysinfo.bufferram = 0;
164
165	swap_pager_status(&i, &j);
166	sysinfo.totalswap = i * PAGE_SIZE;
167	sysinfo.freeswap = (i - j) * PAGE_SIZE;
168
169	sysinfo.procs = nprocs;
170
171	/* The following are only present in newer Linux kernels. */
172	sysinfo.totalbig = 0;
173	sysinfo.freebig = 0;
174	sysinfo.mem_unit = 1;
175
176	return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
177}
178
179int
180linux_alarm(struct thread *td, struct linux_alarm_args *args)
181{
182	struct itimerval it, old_it;
183	u_int secs;
184	int error;
185
186#ifdef DEBUG
187	if (ldebug(alarm))
188		printf(ARGS(alarm, "%u"), args->secs);
189#endif
190
191	secs = args->secs;
192
193	if (secs > INT_MAX)
194		secs = INT_MAX;
195
196	it.it_value.tv_sec = (long) secs;
197	it.it_value.tv_usec = 0;
198	it.it_interval.tv_sec = 0;
199	it.it_interval.tv_usec = 0;
200	error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
201	if (error)
202		return (error);
203	if (timevalisset(&old_it.it_value)) {
204		if (old_it.it_value.tv_usec != 0)
205			old_it.it_value.tv_sec++;
206		td->td_retval[0] = old_it.it_value.tv_sec;
207	}
208	return (0);
209}
210
211int
212linux_brk(struct thread *td, struct linux_brk_args *args)
213{
214	struct vmspace *vm = td->td_proc->p_vmspace;
215	vm_offset_t new, old;
216	struct obreak_args /* {
217		char * nsize;
218	} */ tmp;
219
220#ifdef DEBUG
221	if (ldebug(brk))
222		printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
223#endif
224	old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize);
225	new = (vm_offset_t)args->dsend;
226	tmp.nsize = (char *)new;
227	if (((caddr_t)new > vm->vm_daddr) && !sys_obreak(td, &tmp))
228		td->td_retval[0] = (long)new;
229	else
230		td->td_retval[0] = (long)old;
231
232	return (0);
233}
234
235#if defined(__i386__)
236/* XXX: what about amd64/linux32? */
237
238int
239linux_uselib(struct thread *td, struct linux_uselib_args *args)
240{
241	struct nameidata ni;
242	struct vnode *vp;
243	struct exec *a_out;
244	struct vattr attr;
245	vm_offset_t vmaddr;
246	unsigned long file_offset;
247	unsigned long bss_size;
248	char *library;
249	ssize_t aresid;
250	int error, locked, writecount;
251
252	LCONVPATHEXIST(td, args->library, &library);
253
254#ifdef DEBUG
255	if (ldebug(uselib))
256		printf(ARGS(uselib, "%s"), library);
257#endif
258
259	a_out = NULL;
260	locked = 0;
261	vp = NULL;
262
263	NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
264	    UIO_SYSSPACE, library, td);
265	error = namei(&ni);
266	LFREEPATH(library);
267	if (error)
268		goto cleanup;
269
270	vp = ni.ni_vp;
271	NDFREE(&ni, NDF_ONLY_PNBUF);
272
273	/*
274	 * From here on down, we have a locked vnode that must be unlocked.
275	 * XXX: The code below largely duplicates exec_check_permissions().
276	 */
277	locked = 1;
278
279	/* Writable? */
280	error = VOP_GET_WRITECOUNT(vp, &writecount);
281	if (error != 0)
282		goto cleanup;
283	if (writecount != 0) {
284		error = ETXTBSY;
285		goto cleanup;
286	}
287
288	/* Executable? */
289	error = VOP_GETATTR(vp, &attr, td->td_ucred);
290	if (error)
291		goto cleanup;
292
293	if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
294	    ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
295		/* EACCESS is what exec(2) returns. */
296		error = ENOEXEC;
297		goto cleanup;
298	}
299
300	/* Sensible size? */
301	if (attr.va_size == 0) {
302		error = ENOEXEC;
303		goto cleanup;
304	}
305
306	/* Can we access it? */
307	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
308	if (error)
309		goto cleanup;
310
311	/*
312	 * XXX: This should use vn_open() so that it is properly authorized,
313	 * and to reduce code redundancy all over the place here.
314	 * XXX: Not really, it duplicates far more of exec_check_permissions()
315	 * than vn_open().
316	 */
317#ifdef MAC
318	error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
319	if (error)
320		goto cleanup;
321#endif
322	error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
323	if (error)
324		goto cleanup;
325
326	/* Pull in executable header into exec_map */
327	error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
328	    VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
329	if (error)
330		goto cleanup;
331
332	/* Is it a Linux binary ? */
333	if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
334		error = ENOEXEC;
335		goto cleanup;
336	}
337
338	/*
339	 * While we are here, we should REALLY do some more checks
340	 */
341
342	/* Set file/virtual offset based on a.out variant. */
343	switch ((int)(a_out->a_magic & 0xffff)) {
344	case 0413:			/* ZMAGIC */
345		file_offset = 1024;
346		break;
347	case 0314:			/* QMAGIC */
348		file_offset = 0;
349		break;
350	default:
351		error = ENOEXEC;
352		goto cleanup;
353	}
354
355	bss_size = round_page(a_out->a_bss);
356
357	/* Check various fields in header for validity/bounds. */
358	if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
359		error = ENOEXEC;
360		goto cleanup;
361	}
362
363	/* text + data can't exceed file size */
364	if (a_out->a_data + a_out->a_text > attr.va_size) {
365		error = EFAULT;
366		goto cleanup;
367	}
368
369	/*
370	 * text/data/bss must not exceed limits
371	 * XXX - this is not complete. it should check current usage PLUS
372	 * the resources needed by this library.
373	 */
374	PROC_LOCK(td->td_proc);
375	if (a_out->a_text > maxtsiz ||
376	    a_out->a_data + bss_size > lim_cur(td->td_proc, RLIMIT_DATA) ||
377	    racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
378	    bss_size) != 0) {
379		PROC_UNLOCK(td->td_proc);
380		error = ENOMEM;
381		goto cleanup;
382	}
383	PROC_UNLOCK(td->td_proc);
384
385	/*
386	 * Prevent more writers.
387	 * XXX: Note that if any of the VM operations fail below we don't
388	 * clear this flag.
389	 */
390	VOP_SET_TEXT(vp);
391
392	/*
393	 * Lock no longer needed
394	 */
395	locked = 0;
396	VOP_UNLOCK(vp, 0);
397
398	/*
399	 * Check if file_offset page aligned. Currently we cannot handle
400	 * misalinged file offsets, and so we read in the entire image
401	 * (what a waste).
402	 */
403	if (file_offset & PAGE_MASK) {
404#ifdef DEBUG
405		printf("uselib: Non page aligned binary %lu\n", file_offset);
406#endif
407		/* Map text+data read/write/execute */
408
409		/* a_entry is the load address and is page aligned */
410		vmaddr = trunc_page(a_out->a_entry);
411
412		/* get anon user mapping, read+write+execute */
413		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
414		    &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
415		    VM_PROT_ALL, VM_PROT_ALL, 0);
416		if (error)
417			goto cleanup;
418
419		error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
420		    a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
421		    td->td_ucred, NOCRED, &aresid, td);
422		if (error != 0)
423			goto cleanup;
424		if (aresid != 0) {
425			error = ENOEXEC;
426			goto cleanup;
427		}
428	} else {
429#ifdef DEBUG
430		printf("uselib: Page aligned binary %lu\n", file_offset);
431#endif
432		/*
433		 * for QMAGIC, a_entry is 20 bytes beyond the load address
434		 * to skip the executable header
435		 */
436		vmaddr = trunc_page(a_out->a_entry);
437
438		/*
439		 * Map it all into the process's space as a single
440		 * copy-on-write "data" segment.
441		 */
442		error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr,
443		    a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
444		    MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
445		if (error)
446			goto cleanup;
447	}
448#ifdef DEBUG
449	printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
450	    ((long *)vmaddr)[1]);
451#endif
452	if (bss_size != 0) {
453		/* Calculate BSS start address */
454		vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
455		    a_out->a_data;
456
457		/* allocate some 'anon' space */
458		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
459		    &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
460		    VM_PROT_ALL, 0);
461		if (error)
462			goto cleanup;
463	}
464
465cleanup:
466	/* Unlock vnode if needed */
467	if (locked)
468		VOP_UNLOCK(vp, 0);
469
470	/* Release the temporary mapping. */
471	if (a_out)
472		kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
473
474	return (error);
475}
476
477#endif	/* __i386__ */
478
479int
480linux_select(struct thread *td, struct linux_select_args *args)
481{
482	l_timeval ltv;
483	struct timeval tv0, tv1, utv, *tvp;
484	int error;
485
486#ifdef DEBUG
487	if (ldebug(select))
488		printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
489		    (void *)args->readfds, (void *)args->writefds,
490		    (void *)args->exceptfds, (void *)args->timeout);
491#endif
492
493	/*
494	 * Store current time for computation of the amount of
495	 * time left.
496	 */
497	if (args->timeout) {
498		if ((error = copyin(args->timeout, &ltv, sizeof(ltv))))
499			goto select_out;
500		utv.tv_sec = ltv.tv_sec;
501		utv.tv_usec = ltv.tv_usec;
502#ifdef DEBUG
503		if (ldebug(select))
504			printf(LMSG("incoming timeout (%jd/%ld)"),
505			    (intmax_t)utv.tv_sec, utv.tv_usec);
506#endif
507
508		if (itimerfix(&utv)) {
509			/*
510			 * The timeval was invalid.  Convert it to something
511			 * valid that will act as it does under Linux.
512			 */
513			utv.tv_sec += utv.tv_usec / 1000000;
514			utv.tv_usec %= 1000000;
515			if (utv.tv_usec < 0) {
516				utv.tv_sec -= 1;
517				utv.tv_usec += 1000000;
518			}
519			if (utv.tv_sec < 0)
520				timevalclear(&utv);
521		}
522		microtime(&tv0);
523		tvp = &utv;
524	} else
525		tvp = NULL;
526
527	error = kern_select(td, args->nfds, args->readfds, args->writefds,
528	    args->exceptfds, tvp, sizeof(l_int) * 8);
529
530#ifdef DEBUG
531	if (ldebug(select))
532		printf(LMSG("real select returns %d"), error);
533#endif
534	if (error)
535		goto select_out;
536
537	if (args->timeout) {
538		if (td->td_retval[0]) {
539			/*
540			 * Compute how much time was left of the timeout,
541			 * by subtracting the current time and the time
542			 * before we started the call, and subtracting
543			 * that result from the user-supplied value.
544			 */
545			microtime(&tv1);
546			timevalsub(&tv1, &tv0);
547			timevalsub(&utv, &tv1);
548			if (utv.tv_sec < 0)
549				timevalclear(&utv);
550		} else
551			timevalclear(&utv);
552#ifdef DEBUG
553		if (ldebug(select))
554			printf(LMSG("outgoing timeout (%jd/%ld)"),
555			    (intmax_t)utv.tv_sec, utv.tv_usec);
556#endif
557		ltv.tv_sec = utv.tv_sec;
558		ltv.tv_usec = utv.tv_usec;
559		if ((error = copyout(&ltv, args->timeout, sizeof(ltv))))
560			goto select_out;
561	}
562
563select_out:
564#ifdef DEBUG
565	if (ldebug(select))
566		printf(LMSG("select_out -> %d"), error);
567#endif
568	return (error);
569}
570
571int
572linux_mremap(struct thread *td, struct linux_mremap_args *args)
573{
574	struct munmap_args /* {
575		void *addr;
576		size_t len;
577	} */ bsd_args;
578	int error = 0;
579
580#ifdef DEBUG
581	if (ldebug(mremap))
582		printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
583		    (void *)(uintptr_t)args->addr,
584		    (unsigned long)args->old_len,
585		    (unsigned long)args->new_len,
586		    (unsigned long)args->flags);
587#endif
588
589	if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
590		td->td_retval[0] = 0;
591		return (EINVAL);
592	}
593
594	/*
595	 * Check for the page alignment.
596	 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
597	 */
598	if (args->addr & PAGE_MASK) {
599		td->td_retval[0] = 0;
600		return (EINVAL);
601	}
602
603	args->new_len = round_page(args->new_len);
604	args->old_len = round_page(args->old_len);
605
606	if (args->new_len > args->old_len) {
607		td->td_retval[0] = 0;
608		return (ENOMEM);
609	}
610
611	if (args->new_len < args->old_len) {
612		bsd_args.addr =
613		    (caddr_t)((uintptr_t)args->addr + args->new_len);
614		bsd_args.len = args->old_len - args->new_len;
615		error = sys_munmap(td, &bsd_args);
616	}
617
618	td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
619	return (error);
620}
621
622#define LINUX_MS_ASYNC       0x0001
623#define LINUX_MS_INVALIDATE  0x0002
624#define LINUX_MS_SYNC        0x0004
625
626int
627linux_msync(struct thread *td, struct linux_msync_args *args)
628{
629	struct msync_args bsd_args;
630
631	bsd_args.addr = (caddr_t)(uintptr_t)args->addr;
632	bsd_args.len = (uintptr_t)args->len;
633	bsd_args.flags = args->fl & ~LINUX_MS_SYNC;
634
635	return (sys_msync(td, &bsd_args));
636}
637
638int
639linux_time(struct thread *td, struct linux_time_args *args)
640{
641	struct timeval tv;
642	l_time_t tm;
643	int error;
644
645#ifdef DEBUG
646	if (ldebug(time))
647		printf(ARGS(time, "*"));
648#endif
649
650	microtime(&tv);
651	tm = tv.tv_sec;
652	if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
653		return (error);
654	td->td_retval[0] = tm;
655	return (0);
656}
657
658struct l_times_argv {
659	l_clock_t	tms_utime;
660	l_clock_t	tms_stime;
661	l_clock_t	tms_cutime;
662	l_clock_t	tms_cstime;
663};
664
665
666/*
667 * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
668 * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
669 * auxiliary vector entry.
670 */
671#define	CLK_TCK		100
672
673#define	CONVOTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
674#define	CONVNTCK(r)	(r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
675
676#define	CONVTCK(r)	(linux_kernver(td) >= LINUX_KERNVER_2004000 ?		\
677			    CONVNTCK(r) : CONVOTCK(r))
678
679int
680linux_times(struct thread *td, struct linux_times_args *args)
681{
682	struct timeval tv, utime, stime, cutime, cstime;
683	struct l_times_argv tms;
684	struct proc *p;
685	int error;
686
687#ifdef DEBUG
688	if (ldebug(times))
689		printf(ARGS(times, "*"));
690#endif
691
692	if (args->buf != NULL) {
693		p = td->td_proc;
694		PROC_LOCK(p);
695		PROC_SLOCK(p);
696		calcru(p, &utime, &stime);
697		PROC_SUNLOCK(p);
698		calccru(p, &cutime, &cstime);
699		PROC_UNLOCK(p);
700
701		tms.tms_utime = CONVTCK(utime);
702		tms.tms_stime = CONVTCK(stime);
703
704		tms.tms_cutime = CONVTCK(cutime);
705		tms.tms_cstime = CONVTCK(cstime);
706
707		if ((error = copyout(&tms, args->buf, sizeof(tms))))
708			return (error);
709	}
710
711	microuptime(&tv);
712	td->td_retval[0] = (int)CONVTCK(tv);
713	return (0);
714}
715
716int
717linux_newuname(struct thread *td, struct linux_newuname_args *args)
718{
719	struct l_new_utsname utsname;
720	char osname[LINUX_MAX_UTSNAME];
721	char osrelease[LINUX_MAX_UTSNAME];
722	char *p;
723
724#ifdef DEBUG
725	if (ldebug(newuname))
726		printf(ARGS(newuname, "*"));
727#endif
728
729	linux_get_osname(td, osname);
730	linux_get_osrelease(td, osrelease);
731
732	bzero(&utsname, sizeof(utsname));
733	strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
734	getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
735	getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
736	strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
737	strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
738	for (p = utsname.version; *p != '\0'; ++p)
739		if (*p == '\n') {
740			*p = '\0';
741			break;
742		}
743	strlcpy(utsname.machine, linux_platform, LINUX_MAX_UTSNAME);
744
745	return (copyout(&utsname, args->buf, sizeof(utsname)));
746}
747
748#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
749struct l_utimbuf {
750	l_time_t l_actime;
751	l_time_t l_modtime;
752};
753
754int
755linux_utime(struct thread *td, struct linux_utime_args *args)
756{
757	struct timeval tv[2], *tvp;
758	struct l_utimbuf lut;
759	char *fname;
760	int error;
761
762	LCONVPATHEXIST(td, args->fname, &fname);
763
764#ifdef DEBUG
765	if (ldebug(utime))
766		printf(ARGS(utime, "%s, *"), fname);
767#endif
768
769	if (args->times) {
770		if ((error = copyin(args->times, &lut, sizeof lut))) {
771			LFREEPATH(fname);
772			return (error);
773		}
774		tv[0].tv_sec = lut.l_actime;
775		tv[0].tv_usec = 0;
776		tv[1].tv_sec = lut.l_modtime;
777		tv[1].tv_usec = 0;
778		tvp = tv;
779	} else
780		tvp = NULL;
781
782	error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
783	LFREEPATH(fname);
784	return (error);
785}
786
787int
788linux_utimes(struct thread *td, struct linux_utimes_args *args)
789{
790	l_timeval ltv[2];
791	struct timeval tv[2], *tvp = NULL;
792	char *fname;
793	int error;
794
795	LCONVPATHEXIST(td, args->fname, &fname);
796
797#ifdef DEBUG
798	if (ldebug(utimes))
799		printf(ARGS(utimes, "%s, *"), fname);
800#endif
801
802	if (args->tptr != NULL) {
803		if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
804			LFREEPATH(fname);
805			return (error);
806		}
807		tv[0].tv_sec = ltv[0].tv_sec;
808		tv[0].tv_usec = ltv[0].tv_usec;
809		tv[1].tv_sec = ltv[1].tv_sec;
810		tv[1].tv_usec = ltv[1].tv_usec;
811		tvp = tv;
812	}
813
814	error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
815	LFREEPATH(fname);
816	return (error);
817}
818
819int
820linux_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
853int
854linux_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
877int
878linux_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
903int
904linux_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_mkfifo(td, path, UIO_SYSSPACE, args->mode);
920		break;
921
922	case S_IFCHR:
923	case S_IFBLK:
924		error = kern_mknod(td, path, UIO_SYSSPACE, args->mode,
925		    args->dev);
926		break;
927
928	case S_IFDIR:
929		error = EPERM;
930		break;
931
932	case 0:
933		args->mode |= S_IFREG;
934		/* FALLTHROUGH */
935	case S_IFREG:
936		error = kern_open(td, path, UIO_SYSSPACE,
937		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
938		if (error == 0)
939			kern_close(td, td->td_retval[0]);
940		break;
941
942	default:
943		error = EINVAL;
944		break;
945	}
946	LFREEPATH(path);
947	return (error);
948}
949
950int
951linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
952{
953	char *path;
954	int error, dfd;
955
956	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
957	LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
958
959#ifdef DEBUG
960	if (ldebug(mknodat))
961		printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev);
962#endif
963
964	switch (args->mode & S_IFMT) {
965	case S_IFIFO:
966	case S_IFSOCK:
967		error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
968		break;
969
970	case S_IFCHR:
971	case S_IFBLK:
972		error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
973		    args->dev);
974		break;
975
976	case S_IFDIR:
977		error = EPERM;
978		break;
979
980	case 0:
981		args->mode |= S_IFREG;
982		/* FALLTHROUGH */
983	case S_IFREG:
984		error = kern_openat(td, dfd, path, UIO_SYSSPACE,
985		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
986		if (error == 0)
987			kern_close(td, td->td_retval[0]);
988		break;
989
990	default:
991		error = EINVAL;
992		break;
993	}
994	LFREEPATH(path);
995	return (error);
996}
997
998/*
999 * UGH! This is just about the dumbest idea I've ever heard!!
1000 */
1001int
1002linux_personality(struct thread *td, struct linux_personality_args *args)
1003{
1004#ifdef DEBUG
1005	if (ldebug(personality))
1006		printf(ARGS(personality, "%lu"), (unsigned long)args->per);
1007#endif
1008	if (args->per != 0)
1009		return (EINVAL);
1010
1011	/* Yes Jim, it's still a Linux... */
1012	td->td_retval[0] = 0;
1013	return (0);
1014}
1015
1016struct l_itimerval {
1017	l_timeval it_interval;
1018	l_timeval it_value;
1019};
1020
1021#define	B2L_ITIMERVAL(bip, lip) 					\
1022	(bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec;		\
1023	(bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec;	\
1024	(bip)->it_value.tv_sec = (lip)->it_value.tv_sec;		\
1025	(bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1026
1027int
1028linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1029{
1030	int error;
1031	struct l_itimerval ls;
1032	struct itimerval aitv, oitv;
1033
1034#ifdef DEBUG
1035	if (ldebug(setitimer))
1036		printf(ARGS(setitimer, "%p, %p"),
1037		    (void *)uap->itv, (void *)uap->oitv);
1038#endif
1039
1040	if (uap->itv == NULL) {
1041		uap->itv = uap->oitv;
1042		return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1043	}
1044
1045	error = copyin(uap->itv, &ls, sizeof(ls));
1046	if (error != 0)
1047		return (error);
1048	B2L_ITIMERVAL(&aitv, &ls);
1049#ifdef DEBUG
1050	if (ldebug(setitimer)) {
1051		printf("setitimer: value: sec: %jd, usec: %ld\n",
1052		    (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1053		printf("setitimer: interval: sec: %jd, usec: %ld\n",
1054		    (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1055	}
1056#endif
1057	error = kern_setitimer(td, uap->which, &aitv, &oitv);
1058	if (error != 0 || uap->oitv == NULL)
1059		return (error);
1060	B2L_ITIMERVAL(&ls, &oitv);
1061
1062	return (copyout(&ls, uap->oitv, sizeof(ls)));
1063}
1064
1065int
1066linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1067{
1068	int error;
1069	struct l_itimerval ls;
1070	struct itimerval aitv;
1071
1072#ifdef DEBUG
1073	if (ldebug(getitimer))
1074		printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1075#endif
1076	error = kern_getitimer(td, uap->which, &aitv);
1077	if (error != 0)
1078		return (error);
1079	B2L_ITIMERVAL(&ls, &aitv);
1080	return (copyout(&ls, uap->itv, sizeof(ls)));
1081}
1082
1083int
1084linux_nice(struct thread *td, struct linux_nice_args *args)
1085{
1086	struct setpriority_args bsd_args;
1087
1088	bsd_args.which = PRIO_PROCESS;
1089	bsd_args.who = 0;		/* current process */
1090	bsd_args.prio = args->inc;
1091	return (sys_setpriority(td, &bsd_args));
1092}
1093
1094int
1095linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1096{
1097	struct ucred *newcred, *oldcred;
1098	l_gid_t *linux_gidset;
1099	gid_t *bsd_gidset;
1100	int ngrp, error;
1101	struct proc *p;
1102
1103	ngrp = args->gidsetsize;
1104	if (ngrp < 0 || ngrp >= ngroups_max + 1)
1105		return (EINVAL);
1106	linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_TEMP, M_WAITOK);
1107	error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1108	if (error)
1109		goto out;
1110	newcred = crget();
1111	crextend(newcred, ngrp + 1);
1112	p = td->td_proc;
1113	PROC_LOCK(p);
1114	oldcred = p->p_ucred;
1115	crcopy(newcred, oldcred);
1116
1117	/*
1118	 * cr_groups[0] holds egid. Setting the whole set from
1119	 * the supplied set will cause egid to be changed too.
1120	 * Keep cr_groups[0] unchanged to prevent that.
1121	 */
1122
1123	if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) {
1124		PROC_UNLOCK(p);
1125		crfree(newcred);
1126		goto out;
1127	}
1128
1129	if (ngrp > 0) {
1130		newcred->cr_ngroups = ngrp + 1;
1131
1132		bsd_gidset = newcred->cr_groups;
1133		ngrp--;
1134		while (ngrp >= 0) {
1135			bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1136			ngrp--;
1137		}
1138	} else
1139		newcred->cr_ngroups = 1;
1140
1141	setsugid(p);
1142	p->p_ucred = newcred;
1143	PROC_UNLOCK(p);
1144	crfree(oldcred);
1145	error = 0;
1146out:
1147	free(linux_gidset, M_TEMP);
1148	return (error);
1149}
1150
1151int
1152linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1153{
1154	struct ucred *cred;
1155	l_gid_t *linux_gidset;
1156	gid_t *bsd_gidset;
1157	int bsd_gidsetsz, ngrp, error;
1158
1159	cred = td->td_ucred;
1160	bsd_gidset = cred->cr_groups;
1161	bsd_gidsetsz = cred->cr_ngroups - 1;
1162
1163	/*
1164	 * cr_groups[0] holds egid. Returning the whole set
1165	 * here will cause a duplicate. Exclude cr_groups[0]
1166	 * to prevent that.
1167	 */
1168
1169	if ((ngrp = args->gidsetsize) == 0) {
1170		td->td_retval[0] = bsd_gidsetsz;
1171		return (0);
1172	}
1173
1174	if (ngrp < bsd_gidsetsz)
1175		return (EINVAL);
1176
1177	ngrp = 0;
1178	linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1179	    M_TEMP, M_WAITOK);
1180	while (ngrp < bsd_gidsetsz) {
1181		linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1182		ngrp++;
1183	}
1184
1185	error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1186	free(linux_gidset, M_TEMP);
1187	if (error)
1188		return (error);
1189
1190	td->td_retval[0] = ngrp;
1191	return (0);
1192}
1193
1194int
1195linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1196{
1197	struct rlimit bsd_rlim;
1198	struct l_rlimit rlim;
1199	u_int which;
1200	int error;
1201
1202#ifdef DEBUG
1203	if (ldebug(setrlimit))
1204		printf(ARGS(setrlimit, "%d, %p"),
1205		    args->resource, (void *)args->rlim);
1206#endif
1207
1208	if (args->resource >= LINUX_RLIM_NLIMITS)
1209		return (EINVAL);
1210
1211	which = linux_to_bsd_resource[args->resource];
1212	if (which == -1)
1213		return (EINVAL);
1214
1215	error = copyin(args->rlim, &rlim, sizeof(rlim));
1216	if (error)
1217		return (error);
1218
1219	bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1220	bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1221	return (kern_setrlimit(td, which, &bsd_rlim));
1222}
1223
1224int
1225linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1226{
1227	struct l_rlimit rlim;
1228	struct proc *p = td->td_proc;
1229	struct rlimit bsd_rlim;
1230	u_int which;
1231
1232#ifdef DEBUG
1233	if (ldebug(old_getrlimit))
1234		printf(ARGS(old_getrlimit, "%d, %p"),
1235		    args->resource, (void *)args->rlim);
1236#endif
1237
1238	if (args->resource >= LINUX_RLIM_NLIMITS)
1239		return (EINVAL);
1240
1241	which = linux_to_bsd_resource[args->resource];
1242	if (which == -1)
1243		return (EINVAL);
1244
1245	PROC_LOCK(p);
1246	lim_rlimit(p, which, &bsd_rlim);
1247	PROC_UNLOCK(p);
1248
1249#ifdef COMPAT_LINUX32
1250	rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1251	if (rlim.rlim_cur == UINT_MAX)
1252		rlim.rlim_cur = INT_MAX;
1253	rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1254	if (rlim.rlim_max == UINT_MAX)
1255		rlim.rlim_max = INT_MAX;
1256#else
1257	rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1258	if (rlim.rlim_cur == ULONG_MAX)
1259		rlim.rlim_cur = LONG_MAX;
1260	rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1261	if (rlim.rlim_max == ULONG_MAX)
1262		rlim.rlim_max = LONG_MAX;
1263#endif
1264	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1265}
1266
1267int
1268linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1269{
1270	struct l_rlimit rlim;
1271	struct proc *p = td->td_proc;
1272	struct rlimit bsd_rlim;
1273	u_int which;
1274
1275#ifdef DEBUG
1276	if (ldebug(getrlimit))
1277		printf(ARGS(getrlimit, "%d, %p"),
1278		    args->resource, (void *)args->rlim);
1279#endif
1280
1281	if (args->resource >= LINUX_RLIM_NLIMITS)
1282		return (EINVAL);
1283
1284	which = linux_to_bsd_resource[args->resource];
1285	if (which == -1)
1286		return (EINVAL);
1287
1288	PROC_LOCK(p);
1289	lim_rlimit(p, which, &bsd_rlim);
1290	PROC_UNLOCK(p);
1291
1292	rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1293	rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1294	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1295}
1296
1297int
1298linux_sched_setscheduler(struct thread *td,
1299    struct linux_sched_setscheduler_args *args)
1300{
1301	struct sched_setscheduler_args bsd;
1302
1303#ifdef DEBUG
1304	if (ldebug(sched_setscheduler))
1305		printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1306		    args->pid, args->policy, (const void *)args->param);
1307#endif
1308
1309	switch (args->policy) {
1310	case LINUX_SCHED_OTHER:
1311		bsd.policy = SCHED_OTHER;
1312		break;
1313	case LINUX_SCHED_FIFO:
1314		bsd.policy = SCHED_FIFO;
1315		break;
1316	case LINUX_SCHED_RR:
1317		bsd.policy = SCHED_RR;
1318		break;
1319	default:
1320		return (EINVAL);
1321	}
1322
1323	bsd.pid = args->pid;
1324	bsd.param = (struct sched_param *)args->param;
1325	return (sys_sched_setscheduler(td, &bsd));
1326}
1327
1328int
1329linux_sched_getscheduler(struct thread *td,
1330    struct linux_sched_getscheduler_args *args)
1331{
1332	struct sched_getscheduler_args bsd;
1333	int error;
1334
1335#ifdef DEBUG
1336	if (ldebug(sched_getscheduler))
1337		printf(ARGS(sched_getscheduler, "%d"), args->pid);
1338#endif
1339
1340	bsd.pid = args->pid;
1341	error = sys_sched_getscheduler(td, &bsd);
1342
1343	switch (td->td_retval[0]) {
1344	case SCHED_OTHER:
1345		td->td_retval[0] = LINUX_SCHED_OTHER;
1346		break;
1347	case SCHED_FIFO:
1348		td->td_retval[0] = LINUX_SCHED_FIFO;
1349		break;
1350	case SCHED_RR:
1351		td->td_retval[0] = LINUX_SCHED_RR;
1352		break;
1353	}
1354
1355	return (error);
1356}
1357
1358int
1359linux_sched_get_priority_max(struct thread *td,
1360    struct linux_sched_get_priority_max_args *args)
1361{
1362	struct sched_get_priority_max_args bsd;
1363
1364#ifdef DEBUG
1365	if (ldebug(sched_get_priority_max))
1366		printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1367#endif
1368
1369	switch (args->policy) {
1370	case LINUX_SCHED_OTHER:
1371		bsd.policy = SCHED_OTHER;
1372		break;
1373	case LINUX_SCHED_FIFO:
1374		bsd.policy = SCHED_FIFO;
1375		break;
1376	case LINUX_SCHED_RR:
1377		bsd.policy = SCHED_RR;
1378		break;
1379	default:
1380		return (EINVAL);
1381	}
1382	return (sys_sched_get_priority_max(td, &bsd));
1383}
1384
1385int
1386linux_sched_get_priority_min(struct thread *td,
1387    struct linux_sched_get_priority_min_args *args)
1388{
1389	struct sched_get_priority_min_args bsd;
1390
1391#ifdef DEBUG
1392	if (ldebug(sched_get_priority_min))
1393		printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1394#endif
1395
1396	switch (args->policy) {
1397	case LINUX_SCHED_OTHER:
1398		bsd.policy = SCHED_OTHER;
1399		break;
1400	case LINUX_SCHED_FIFO:
1401		bsd.policy = SCHED_FIFO;
1402		break;
1403	case LINUX_SCHED_RR:
1404		bsd.policy = SCHED_RR;
1405		break;
1406	default:
1407		return (EINVAL);
1408	}
1409	return (sys_sched_get_priority_min(td, &bsd));
1410}
1411
1412#define REBOOT_CAD_ON	0x89abcdef
1413#define REBOOT_CAD_OFF	0
1414#define REBOOT_HALT	0xcdef0123
1415#define REBOOT_RESTART	0x01234567
1416#define REBOOT_RESTART2	0xA1B2C3D4
1417#define REBOOT_POWEROFF	0x4321FEDC
1418#define REBOOT_MAGIC1	0xfee1dead
1419#define REBOOT_MAGIC2	0x28121969
1420#define REBOOT_MAGIC2A	0x05121996
1421#define REBOOT_MAGIC2B	0x16041998
1422
1423int
1424linux_reboot(struct thread *td, struct linux_reboot_args *args)
1425{
1426	struct reboot_args bsd_args;
1427
1428#ifdef DEBUG
1429	if (ldebug(reboot))
1430		printf(ARGS(reboot, "0x%x"), args->cmd);
1431#endif
1432
1433	if (args->magic1 != REBOOT_MAGIC1)
1434		return (EINVAL);
1435
1436	switch (args->magic2) {
1437	case REBOOT_MAGIC2:
1438	case REBOOT_MAGIC2A:
1439	case REBOOT_MAGIC2B:
1440		break;
1441	default:
1442		return (EINVAL);
1443	}
1444
1445	switch (args->cmd) {
1446	case REBOOT_CAD_ON:
1447	case REBOOT_CAD_OFF:
1448		return (priv_check(td, PRIV_REBOOT));
1449	case REBOOT_HALT:
1450		bsd_args.opt = RB_HALT;
1451		break;
1452	case REBOOT_RESTART:
1453	case REBOOT_RESTART2:
1454		bsd_args.opt = 0;
1455		break;
1456	case REBOOT_POWEROFF:
1457		bsd_args.opt = RB_POWEROFF;
1458		break;
1459	default:
1460		return (EINVAL);
1461	}
1462	return (sys_reboot(td, &bsd_args));
1463}
1464
1465
1466/*
1467 * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1468 * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
1469 * are assumed to be preserved. The following lightweight syscalls fixes
1470 * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
1471 *
1472 * linux_getpid() - MP SAFE
1473 * linux_getgid() - MP SAFE
1474 * linux_getuid() - MP SAFE
1475 */
1476
1477int
1478linux_getpid(struct thread *td, struct linux_getpid_args *args)
1479{
1480	struct linux_emuldata *em;
1481
1482#ifdef DEBUG
1483	if (ldebug(getpid))
1484		printf(ARGS(getpid, ""));
1485#endif
1486
1487	if (linux_use26(td)) {
1488		em = em_find(td->td_proc, EMUL_DONTLOCK);
1489		KASSERT(em != NULL, ("getpid: emuldata not found.\n"));
1490		td->td_retval[0] = em->shared->group_pid;
1491	} else {
1492		td->td_retval[0] = td->td_proc->p_pid;
1493	}
1494
1495	return (0);
1496}
1497
1498int
1499linux_gettid(struct thread *td, struct linux_gettid_args *args)
1500{
1501
1502#ifdef DEBUG
1503	if (ldebug(gettid))
1504		printf(ARGS(gettid, ""));
1505#endif
1506
1507	td->td_retval[0] = td->td_proc->p_pid;
1508	return (0);
1509}
1510
1511
1512int
1513linux_getppid(struct thread *td, struct linux_getppid_args *args)
1514{
1515	struct linux_emuldata *em;
1516	struct proc *p, *pp;
1517
1518#ifdef DEBUG
1519	if (ldebug(getppid))
1520		printf(ARGS(getppid, ""));
1521#endif
1522
1523	if (!linux_use26(td)) {
1524		PROC_LOCK(td->td_proc);
1525		td->td_retval[0] = td->td_proc->p_pptr->p_pid;
1526		PROC_UNLOCK(td->td_proc);
1527		return (0);
1528	}
1529
1530	em = em_find(td->td_proc, EMUL_DONTLOCK);
1531
1532	KASSERT(em != NULL, ("getppid: process emuldata not found.\n"));
1533
1534	/* find the group leader */
1535	p = pfind(em->shared->group_pid);
1536
1537	if (p == NULL) {
1538#ifdef DEBUG
1539	   	printf(LMSG("parent process not found.\n"));
1540#endif
1541		return (0);
1542	}
1543
1544	pp = p->p_pptr;		/* switch to parent */
1545	PROC_LOCK(pp);
1546	PROC_UNLOCK(p);
1547
1548	/* if its also linux process */
1549	if (pp->p_sysent == &elf_linux_sysvec) {
1550		em = em_find(pp, EMUL_DONTLOCK);
1551		KASSERT(em != NULL, ("getppid: parent emuldata not found.\n"));
1552
1553		td->td_retval[0] = em->shared->group_pid;
1554	} else
1555		td->td_retval[0] = pp->p_pid;
1556
1557	PROC_UNLOCK(pp);
1558
1559	return (0);
1560}
1561
1562int
1563linux_getgid(struct thread *td, struct linux_getgid_args *args)
1564{
1565
1566#ifdef DEBUG
1567	if (ldebug(getgid))
1568		printf(ARGS(getgid, ""));
1569#endif
1570
1571	td->td_retval[0] = td->td_ucred->cr_rgid;
1572	return (0);
1573}
1574
1575int
1576linux_getuid(struct thread *td, struct linux_getuid_args *args)
1577{
1578
1579#ifdef DEBUG
1580	if (ldebug(getuid))
1581		printf(ARGS(getuid, ""));
1582#endif
1583
1584	td->td_retval[0] = td->td_ucred->cr_ruid;
1585	return (0);
1586}
1587
1588
1589int
1590linux_getsid(struct thread *td, struct linux_getsid_args *args)
1591{
1592	struct getsid_args bsd;
1593
1594#ifdef DEBUG
1595	if (ldebug(getsid))
1596		printf(ARGS(getsid, "%i"), args->pid);
1597#endif
1598
1599	bsd.pid = args->pid;
1600	return (sys_getsid(td, &bsd));
1601}
1602
1603int
1604linux_nosys(struct thread *td, struct nosys_args *ignore)
1605{
1606
1607	return (ENOSYS);
1608}
1609
1610int
1611linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1612{
1613	struct getpriority_args bsd_args;
1614	int error;
1615
1616#ifdef DEBUG
1617	if (ldebug(getpriority))
1618		printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
1619#endif
1620
1621	bsd_args.which = args->which;
1622	bsd_args.who = args->who;
1623	error = sys_getpriority(td, &bsd_args);
1624	td->td_retval[0] = 20 - td->td_retval[0];
1625	return (error);
1626}
1627
1628int
1629linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1630{
1631	int name[2];
1632
1633#ifdef DEBUG
1634	if (ldebug(sethostname))
1635		printf(ARGS(sethostname, "*, %i"), args->len);
1636#endif
1637
1638	name[0] = CTL_KERN;
1639	name[1] = KERN_HOSTNAME;
1640	return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1641	    args->len, 0, 0));
1642}
1643
1644int
1645linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1646{
1647	int name[2];
1648
1649#ifdef DEBUG
1650	if (ldebug(setdomainname))
1651		printf(ARGS(setdomainname, "*, %i"), args->len);
1652#endif
1653
1654	name[0] = CTL_KERN;
1655	name[1] = KERN_NISDOMAINNAME;
1656	return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1657	    args->len, 0, 0));
1658}
1659
1660int
1661linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1662{
1663	struct linux_emuldata *em;
1664
1665#ifdef DEBUG
1666	if (ldebug(exit_group))
1667		printf(ARGS(exit_group, "%i"), args->error_code);
1668#endif
1669
1670	em = em_find(td->td_proc, EMUL_DONTLOCK);
1671	if (em->shared->refs > 1) {
1672		EMUL_SHARED_WLOCK(&emul_shared_lock);
1673		em->shared->flags |= EMUL_SHARED_HASXSTAT;
1674		em->shared->xstat = W_EXITCODE(args->error_code, 0);
1675		EMUL_SHARED_WUNLOCK(&emul_shared_lock);
1676		if (linux_use26(td))
1677			linux_kill_threads(td, SIGKILL);
1678	}
1679
1680	/*
1681	 * XXX: we should send a signal to the parent if
1682	 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1683	 * as it doesnt occur often.
1684	 */
1685	exit1(td, W_EXITCODE(args->error_code, 0));
1686
1687	return (0);
1688}
1689
1690#define _LINUX_CAPABILITY_VERSION  0x19980330
1691
1692struct l_user_cap_header {
1693	l_int	version;
1694	l_int	pid;
1695};
1696
1697struct l_user_cap_data {
1698	l_int	effective;
1699	l_int	permitted;
1700	l_int	inheritable;
1701};
1702
1703int
1704linux_capget(struct thread *td, struct linux_capget_args *args)
1705{
1706	struct l_user_cap_header luch;
1707	struct l_user_cap_data lucd;
1708	int error;
1709
1710	if (args->hdrp == NULL)
1711		return (EFAULT);
1712
1713	error = copyin(args->hdrp, &luch, sizeof(luch));
1714	if (error != 0)
1715		return (error);
1716
1717	if (luch.version != _LINUX_CAPABILITY_VERSION) {
1718		luch.version = _LINUX_CAPABILITY_VERSION;
1719		error = copyout(&luch, args->hdrp, sizeof(luch));
1720		if (error)
1721			return (error);
1722		return (EINVAL);
1723	}
1724
1725	if (luch.pid)
1726		return (EPERM);
1727
1728	if (args->datap) {
1729		/*
1730		 * The current implementation doesn't support setting
1731		 * a capability (it's essentially a stub) so indicate
1732		 * that no capabilities are currently set or available
1733		 * to request.
1734		 */
1735		bzero (&lucd, sizeof(lucd));
1736		error = copyout(&lucd, args->datap, sizeof(lucd));
1737	}
1738
1739	return (error);
1740}
1741
1742int
1743linux_capset(struct thread *td, struct linux_capset_args *args)
1744{
1745	struct l_user_cap_header luch;
1746	struct l_user_cap_data lucd;
1747	int error;
1748
1749	if (args->hdrp == NULL || args->datap == NULL)
1750		return (EFAULT);
1751
1752	error = copyin(args->hdrp, &luch, sizeof(luch));
1753	if (error != 0)
1754		return (error);
1755
1756	if (luch.version != _LINUX_CAPABILITY_VERSION) {
1757		luch.version = _LINUX_CAPABILITY_VERSION;
1758		error = copyout(&luch, args->hdrp, sizeof(luch));
1759		if (error)
1760			return (error);
1761		return (EINVAL);
1762	}
1763
1764	if (luch.pid)
1765		return (EPERM);
1766
1767	error = copyin(args->datap, &lucd, sizeof(lucd));
1768	if (error != 0)
1769		return (error);
1770
1771	/* We currently don't support setting any capabilities. */
1772	if (lucd.effective || lucd.permitted || lucd.inheritable) {
1773		linux_msg(td,
1774			  "capset effective=0x%x, permitted=0x%x, "
1775			  "inheritable=0x%x is not implemented",
1776			  (int)lucd.effective, (int)lucd.permitted,
1777			  (int)lucd.inheritable);
1778		return (EPERM);
1779	}
1780
1781	return (0);
1782}
1783
1784int
1785linux_prctl(struct thread *td, struct linux_prctl_args *args)
1786{
1787	int error = 0, max_size;
1788	struct proc *p = td->td_proc;
1789	char comm[LINUX_MAX_COMM_LEN];
1790	struct linux_emuldata *em;
1791	int pdeath_signal;
1792
1793#ifdef DEBUG
1794	if (ldebug(prctl))
1795		printf(ARGS(prctl, "%d, %d, %d, %d, %d"), args->option,
1796		    args->arg2, args->arg3, args->arg4, args->arg5);
1797#endif
1798
1799	switch (args->option) {
1800	case LINUX_PR_SET_PDEATHSIG:
1801		if (!LINUX_SIG_VALID(args->arg2))
1802			return (EINVAL);
1803		em = em_find(p, EMUL_DOLOCK);
1804		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1805		em->pdeath_signal = args->arg2;
1806		EMUL_UNLOCK(&emul_lock);
1807		break;
1808	case LINUX_PR_GET_PDEATHSIG:
1809		em = em_find(p, EMUL_DOLOCK);
1810		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1811		pdeath_signal = em->pdeath_signal;
1812		EMUL_UNLOCK(&emul_lock);
1813		error = copyout(&pdeath_signal,
1814		    (void *)(register_t)args->arg2,
1815		    sizeof(pdeath_signal));
1816		break;
1817	case LINUX_PR_GET_KEEPCAPS:
1818		/*
1819		 * Indicate that we always clear the effective and
1820		 * permitted capability sets when the user id becomes
1821		 * non-zero (actually the capability sets are simply
1822		 * always zero in the current implementation).
1823		 */
1824		td->td_retval[0] = 0;
1825		break;
1826	case LINUX_PR_SET_KEEPCAPS:
1827		/*
1828		 * Ignore requests to keep the effective and permitted
1829		 * capability sets when the user id becomes non-zero.
1830		 */
1831		break;
1832	case LINUX_PR_SET_NAME:
1833		/*
1834		 * To be on the safe side we need to make sure to not
1835		 * overflow the size a linux program expects. We already
1836		 * do this here in the copyin, so that we don't need to
1837		 * check on copyout.
1838		 */
1839		max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1840		error = copyinstr((void *)(register_t)args->arg2, comm,
1841		    max_size, NULL);
1842
1843		/* Linux silently truncates the name if it is too long. */
1844		if (error == ENAMETOOLONG) {
1845			/*
1846			 * XXX: copyinstr() isn't documented to populate the
1847			 * array completely, so do a copyin() to be on the
1848			 * safe side. This should be changed in case
1849			 * copyinstr() is changed to guarantee this.
1850			 */
1851			error = copyin((void *)(register_t)args->arg2, comm,
1852			    max_size - 1);
1853			comm[max_size - 1] = '\0';
1854		}
1855		if (error)
1856			return (error);
1857
1858		PROC_LOCK(p);
1859		strlcpy(p->p_comm, comm, sizeof(p->p_comm));
1860		PROC_UNLOCK(p);
1861		break;
1862	case LINUX_PR_GET_NAME:
1863		PROC_LOCK(p);
1864		strlcpy(comm, p->p_comm, sizeof(comm));
1865		PROC_UNLOCK(p);
1866		error = copyout(comm, (void *)(register_t)args->arg2,
1867		    strlen(comm) + 1);
1868		break;
1869	default:
1870		error = EINVAL;
1871		break;
1872	}
1873
1874	return (error);
1875}
1876
1877/*
1878 * Get affinity of a process.
1879 */
1880int
1881linux_sched_getaffinity(struct thread *td,
1882    struct linux_sched_getaffinity_args *args)
1883{
1884	int error;
1885	struct cpuset_getaffinity_args cga;
1886
1887#ifdef DEBUG
1888	if (ldebug(sched_getaffinity))
1889		printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid,
1890		    args->len);
1891#endif
1892	if (args->len < sizeof(cpuset_t))
1893		return (EINVAL);
1894
1895	cga.level = CPU_LEVEL_WHICH;
1896	cga.which = CPU_WHICH_PID;
1897	cga.id = args->pid;
1898	cga.cpusetsize = sizeof(cpuset_t);
1899	cga.mask = (cpuset_t *) args->user_mask_ptr;
1900
1901	if ((error = sys_cpuset_getaffinity(td, &cga)) == 0)
1902		td->td_retval[0] = sizeof(cpuset_t);
1903
1904	return (error);
1905}
1906
1907/*
1908 *  Set affinity of a process.
1909 */
1910int
1911linux_sched_setaffinity(struct thread *td,
1912    struct linux_sched_setaffinity_args *args)
1913{
1914	struct cpuset_setaffinity_args csa;
1915
1916#ifdef DEBUG
1917	if (ldebug(sched_setaffinity))
1918		printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid,
1919		    args->len);
1920#endif
1921	if (args->len < sizeof(cpuset_t))
1922		return (EINVAL);
1923
1924	csa.level = CPU_LEVEL_WHICH;
1925	csa.which = CPU_WHICH_PID;
1926	csa.id = args->pid;
1927	csa.cpusetsize = sizeof(cpuset_t);
1928	csa.mask = (cpuset_t *) args->user_mask_ptr;
1929
1930	return (sys_cpuset_setaffinity(td, &csa));
1931}
1932