xref: /freebsd/sys/compat/linux/linux_misc.c (revision aad12c28e09753c27e92e772d42978a0c245e14c)
1 /*-
2  * Copyright (c) 2002 Doug Rabson
3  * Copyright (c) 1994-1995 Søren Schmidt
4  * All rights reserved.
5  *
6  * Redistribution and use in source and binary forms, with or without
7  * modification, are permitted provided that the following conditions
8  * are met:
9  * 1. Redistributions of source code must retain the above copyright
10  *    notice, this list of conditions and the following disclaimer
11  *    in this position and unchanged.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. The name of the author may not be used to endorse or promote products
16  *    derived from this software without specific prior written permission
17  *
18  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
19  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
20  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
21  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
22  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
23  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
24  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
25  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
26  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
27  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
28  */
29 
30 #include <sys/cdefs.h>
31 __FBSDID("$FreeBSD$");
32 
33 #include "opt_compat.h"
34 #include "opt_kdtrace.h"
35 
36 #include <sys/param.h>
37 #include <sys/blist.h>
38 #include <sys/fcntl.h>
39 #if defined(__i386__)
40 #include <sys/imgact_aout.h>
41 #endif
42 #include <sys/jail.h>
43 #include <sys/kernel.h>
44 #include <sys/limits.h>
45 #include <sys/lock.h>
46 #include <sys/malloc.h>
47 #include <sys/mman.h>
48 #include <sys/mount.h>
49 #include <sys/mutex.h>
50 #include <sys/namei.h>
51 #include <sys/priv.h>
52 #include <sys/proc.h>
53 #include <sys/reboot.h>
54 #include <sys/racct.h>
55 #include <sys/resourcevar.h>
56 #include <sys/sched.h>
57 #include <sys/sdt.h>
58 #include <sys/signalvar.h>
59 #include <sys/stat.h>
60 #include <sys/syscallsubr.h>
61 #include <sys/sysctl.h>
62 #include <sys/sysproto.h>
63 #include <sys/systm.h>
64 #include <sys/time.h>
65 #include <sys/vmmeter.h>
66 #include <sys/vnode.h>
67 #include <sys/wait.h>
68 #include <sys/cpuset.h>
69 
70 #include <security/mac/mac_framework.h>
71 
72 #include <vm/vm.h>
73 #include <vm/pmap.h>
74 #include <vm/vm_kern.h>
75 #include <vm/vm_map.h>
76 #include <vm/vm_extern.h>
77 #include <vm/vm_object.h>
78 #include <vm/swap_pager.h>
79 
80 #ifdef COMPAT_LINUX32
81 #include <machine/../linux32/linux.h>
82 #include <machine/../linux32/linux32_proto.h>
83 #else
84 #include <machine/../linux/linux.h>
85 #include <machine/../linux/linux_proto.h>
86 #endif
87 
88 #include <compat/linux/linux_dtrace.h>
89 #include <compat/linux/linux_file.h>
90 #include <compat/linux/linux_mib.h>
91 #include <compat/linux/linux_signal.h>
92 #include <compat/linux/linux_util.h>
93 #include <compat/linux/linux_sysproto.h>
94 #include <compat/linux/linux_emul.h>
95 #include <compat/linux/linux_misc.h>
96 
97 /* DTrace init */
98 LIN_SDT_PROVIDER_DECLARE(LINUX_DTRACE);
99 
100 /* Linuxulator-global DTrace probes */
101 LIN_SDT_PROBE_DECLARE(locks, emul_lock, locked);
102 LIN_SDT_PROBE_DECLARE(locks, emul_lock, unlock);
103 LIN_SDT_PROBE_DECLARE(locks, emul_shared_rlock, locked);
104 LIN_SDT_PROBE_DECLARE(locks, emul_shared_rlock, unlock);
105 LIN_SDT_PROBE_DECLARE(locks, emul_shared_wlock, locked);
106 LIN_SDT_PROBE_DECLARE(locks, emul_shared_wlock, unlock);
107 
108 int stclohz;				/* Statistics clock frequency */
109 
110 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
111 	RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
112 	RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
113 	RLIMIT_MEMLOCK, RLIMIT_AS
114 };
115 
116 struct l_sysinfo {
117 	l_long		uptime;		/* Seconds since boot */
118 	l_ulong		loads[3];	/* 1, 5, and 15 minute load averages */
119 #define LINUX_SYSINFO_LOADS_SCALE 65536
120 	l_ulong		totalram;	/* Total usable main memory size */
121 	l_ulong		freeram;	/* Available memory size */
122 	l_ulong		sharedram;	/* Amount of shared memory */
123 	l_ulong		bufferram;	/* Memory used by buffers */
124 	l_ulong		totalswap;	/* Total swap space size */
125 	l_ulong		freeswap;	/* swap space still available */
126 	l_ushort	procs;		/* Number of current processes */
127 	l_ushort	pads;
128 	l_ulong		totalbig;
129 	l_ulong		freebig;
130 	l_uint		mem_unit;
131 	char		_f[20-2*sizeof(l_long)-sizeof(l_int)];	/* padding */
132 };
133 int
134 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
135 {
136 	struct l_sysinfo sysinfo;
137 	vm_object_t object;
138 	int i, j;
139 	struct timespec ts;
140 
141 	getnanouptime(&ts);
142 	if (ts.tv_nsec != 0)
143 		ts.tv_sec++;
144 	sysinfo.uptime = ts.tv_sec;
145 
146 	/* Use the information from the mib to get our load averages */
147 	for (i = 0; i < 3; i++)
148 		sysinfo.loads[i] = averunnable.ldavg[i] *
149 		    LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
150 
151 	sysinfo.totalram = physmem * PAGE_SIZE;
152 	sysinfo.freeram = sysinfo.totalram - cnt.v_wire_count * PAGE_SIZE;
153 
154 	sysinfo.sharedram = 0;
155 	mtx_lock(&vm_object_list_mtx);
156 	TAILQ_FOREACH(object, &vm_object_list, object_list)
157 		if (object->shadow_count > 1)
158 			sysinfo.sharedram += object->resident_page_count;
159 	mtx_unlock(&vm_object_list_mtx);
160 
161 	sysinfo.sharedram *= PAGE_SIZE;
162 	sysinfo.bufferram = 0;
163 
164 	swap_pager_status(&i, &j);
165 	sysinfo.totalswap = i * PAGE_SIZE;
166 	sysinfo.freeswap = (i - j) * PAGE_SIZE;
167 
168 	sysinfo.procs = nprocs;
169 
170 	/* The following are only present in newer Linux kernels. */
171 	sysinfo.totalbig = 0;
172 	sysinfo.freebig = 0;
173 	sysinfo.mem_unit = 1;
174 
175 	return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
176 }
177 
178 int
179 linux_alarm(struct thread *td, struct linux_alarm_args *args)
180 {
181 	struct itimerval it, old_it;
182 	u_int secs;
183 	int error;
184 
185 #ifdef DEBUG
186 	if (ldebug(alarm))
187 		printf(ARGS(alarm, "%u"), args->secs);
188 #endif
189 
190 	secs = args->secs;
191 
192 	if (secs > INT_MAX)
193 		secs = INT_MAX;
194 
195 	it.it_value.tv_sec = (long) secs;
196 	it.it_value.tv_usec = 0;
197 	it.it_interval.tv_sec = 0;
198 	it.it_interval.tv_usec = 0;
199 	error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
200 	if (error)
201 		return (error);
202 	if (timevalisset(&old_it.it_value)) {
203 		if (old_it.it_value.tv_usec != 0)
204 			old_it.it_value.tv_sec++;
205 		td->td_retval[0] = old_it.it_value.tv_sec;
206 	}
207 	return (0);
208 }
209 
210 int
211 linux_brk(struct thread *td, struct linux_brk_args *args)
212 {
213 	struct vmspace *vm = td->td_proc->p_vmspace;
214 	vm_offset_t new, old;
215 	struct obreak_args /* {
216 		char * nsize;
217 	} */ tmp;
218 
219 #ifdef DEBUG
220 	if (ldebug(brk))
221 		printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
222 #endif
223 	old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize);
224 	new = (vm_offset_t)args->dsend;
225 	tmp.nsize = (char *)new;
226 	if (((caddr_t)new > vm->vm_daddr) && !sys_obreak(td, &tmp))
227 		td->td_retval[0] = (long)new;
228 	else
229 		td->td_retval[0] = (long)old;
230 
231 	return (0);
232 }
233 
234 #if defined(__i386__)
235 /* XXX: what about amd64/linux32? */
236 
237 int
238 linux_uselib(struct thread *td, struct linux_uselib_args *args)
239 {
240 	struct nameidata ni;
241 	struct vnode *vp;
242 	struct exec *a_out;
243 	struct vattr attr;
244 	vm_offset_t vmaddr;
245 	unsigned long file_offset;
246 	unsigned long bss_size;
247 	char *library;
248 	ssize_t aresid;
249 	int error;
250 	int locked;
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 	if (vp->v_writecount) {
281 		error = ETXTBSY;
282 		goto cleanup;
283 	}
284 
285 	/* Executable? */
286 	error = VOP_GETATTR(vp, &attr, td->td_ucred);
287 	if (error)
288 		goto cleanup;
289 
290 	if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
291 	    ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
292 		/* EACCESS is what exec(2) returns. */
293 		error = ENOEXEC;
294 		goto cleanup;
295 	}
296 
297 	/* Sensible size? */
298 	if (attr.va_size == 0) {
299 		error = ENOEXEC;
300 		goto cleanup;
301 	}
302 
303 	/* Can we access it? */
304 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
305 	if (error)
306 		goto cleanup;
307 
308 	/*
309 	 * XXX: This should use vn_open() so that it is properly authorized,
310 	 * and to reduce code redundancy all over the place here.
311 	 * XXX: Not really, it duplicates far more of exec_check_permissions()
312 	 * than vn_open().
313 	 */
314 #ifdef MAC
315 	error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
316 	if (error)
317 		goto cleanup;
318 #endif
319 	error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
320 	if (error)
321 		goto cleanup;
322 
323 	/* Pull in executable header into exec_map */
324 	error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
325 	    VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
326 	if (error)
327 		goto cleanup;
328 
329 	/* Is it a Linux binary ? */
330 	if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
331 		error = ENOEXEC;
332 		goto cleanup;
333 	}
334 
335 	/*
336 	 * While we are here, we should REALLY do some more checks
337 	 */
338 
339 	/* Set file/virtual offset based on a.out variant. */
340 	switch ((int)(a_out->a_magic & 0xffff)) {
341 	case 0413:			/* ZMAGIC */
342 		file_offset = 1024;
343 		break;
344 	case 0314:			/* QMAGIC */
345 		file_offset = 0;
346 		break;
347 	default:
348 		error = ENOEXEC;
349 		goto cleanup;
350 	}
351 
352 	bss_size = round_page(a_out->a_bss);
353 
354 	/* Check various fields in header for validity/bounds. */
355 	if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
356 		error = ENOEXEC;
357 		goto cleanup;
358 	}
359 
360 	/* text + data can't exceed file size */
361 	if (a_out->a_data + a_out->a_text > attr.va_size) {
362 		error = EFAULT;
363 		goto cleanup;
364 	}
365 
366 	/*
367 	 * text/data/bss must not exceed limits
368 	 * XXX - this is not complete. it should check current usage PLUS
369 	 * the resources needed by this library.
370 	 */
371 	PROC_LOCK(td->td_proc);
372 	if (a_out->a_text > maxtsiz ||
373 	    a_out->a_data + bss_size > lim_cur(td->td_proc, RLIMIT_DATA) ||
374 	    racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
375 	    bss_size) != 0) {
376 		PROC_UNLOCK(td->td_proc);
377 		error = ENOMEM;
378 		goto cleanup;
379 	}
380 	PROC_UNLOCK(td->td_proc);
381 
382 	/*
383 	 * Prevent more writers.
384 	 * XXX: Note that if any of the VM operations fail below we don't
385 	 * clear this flag.
386 	 */
387 	VOP_SET_TEXT(vp);
388 
389 	/*
390 	 * Lock no longer needed
391 	 */
392 	locked = 0;
393 	VOP_UNLOCK(vp, 0);
394 
395 	/*
396 	 * Check if file_offset page aligned. Currently we cannot handle
397 	 * misalinged file offsets, and so we read in the entire image
398 	 * (what a waste).
399 	 */
400 	if (file_offset & PAGE_MASK) {
401 #ifdef DEBUG
402 		printf("uselib: Non page aligned binary %lu\n", file_offset);
403 #endif
404 		/* Map text+data read/write/execute */
405 
406 		/* a_entry is the load address and is page aligned */
407 		vmaddr = trunc_page(a_out->a_entry);
408 
409 		/* get anon user mapping, read+write+execute */
410 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
411 		    &vmaddr, a_out->a_text + a_out->a_data, FALSE, VM_PROT_ALL,
412 		    VM_PROT_ALL, 0);
413 		if (error)
414 			goto cleanup;
415 
416 		error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
417 		    a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
418 		    td->td_ucred, NOCRED, &aresid, td);
419 		if (error != 0)
420 			goto cleanup;
421 		if (aresid != 0) {
422 			error = ENOEXEC;
423 			goto cleanup;
424 		}
425 	} else {
426 #ifdef DEBUG
427 		printf("uselib: Page aligned binary %lu\n", file_offset);
428 #endif
429 		/*
430 		 * for QMAGIC, a_entry is 20 bytes beyond the load address
431 		 * to skip the executable header
432 		 */
433 		vmaddr = trunc_page(a_out->a_entry);
434 
435 		/*
436 		 * Map it all into the process's space as a single
437 		 * copy-on-write "data" segment.
438 		 */
439 		error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr,
440 		    a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
441 		    MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
442 		if (error)
443 			goto cleanup;
444 	}
445 #ifdef DEBUG
446 	printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
447 	    ((long *)vmaddr)[1]);
448 #endif
449 	if (bss_size != 0) {
450 		/* Calculate BSS start address */
451 		vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
452 		    a_out->a_data;
453 
454 		/* allocate some 'anon' space */
455 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
456 		    &vmaddr, bss_size, FALSE, VM_PROT_ALL, VM_PROT_ALL, 0);
457 		if (error)
458 			goto cleanup;
459 	}
460 
461 cleanup:
462 	/* Unlock vnode if needed */
463 	if (locked)
464 		VOP_UNLOCK(vp, 0);
465 
466 	/* Release the temporary mapping. */
467 	if (a_out)
468 		kmem_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
469 
470 	return (error);
471 }
472 
473 #endif	/* __i386__ */
474 
475 int
476 linux_select(struct thread *td, struct linux_select_args *args)
477 {
478 	l_timeval ltv;
479 	struct timeval tv0, tv1, utv, *tvp;
480 	int error;
481 
482 #ifdef DEBUG
483 	if (ldebug(select))
484 		printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
485 		    (void *)args->readfds, (void *)args->writefds,
486 		    (void *)args->exceptfds, (void *)args->timeout);
487 #endif
488 
489 	/*
490 	 * Store current time for computation of the amount of
491 	 * time left.
492 	 */
493 	if (args->timeout) {
494 		if ((error = copyin(args->timeout, &ltv, sizeof(ltv))))
495 			goto select_out;
496 		utv.tv_sec = ltv.tv_sec;
497 		utv.tv_usec = ltv.tv_usec;
498 #ifdef DEBUG
499 		if (ldebug(select))
500 			printf(LMSG("incoming timeout (%jd/%ld)"),
501 			    (intmax_t)utv.tv_sec, utv.tv_usec);
502 #endif
503 
504 		if (itimerfix(&utv)) {
505 			/*
506 			 * The timeval was invalid.  Convert it to something
507 			 * valid that will act as it does under Linux.
508 			 */
509 			utv.tv_sec += utv.tv_usec / 1000000;
510 			utv.tv_usec %= 1000000;
511 			if (utv.tv_usec < 0) {
512 				utv.tv_sec -= 1;
513 				utv.tv_usec += 1000000;
514 			}
515 			if (utv.tv_sec < 0)
516 				timevalclear(&utv);
517 		}
518 		microtime(&tv0);
519 		tvp = &utv;
520 	} else
521 		tvp = NULL;
522 
523 	error = kern_select(td, args->nfds, args->readfds, args->writefds,
524 	    args->exceptfds, tvp, sizeof(l_int) * 8);
525 
526 #ifdef DEBUG
527 	if (ldebug(select))
528 		printf(LMSG("real select returns %d"), error);
529 #endif
530 	if (error)
531 		goto select_out;
532 
533 	if (args->timeout) {
534 		if (td->td_retval[0]) {
535 			/*
536 			 * Compute how much time was left of the timeout,
537 			 * by subtracting the current time and the time
538 			 * before we started the call, and subtracting
539 			 * that result from the user-supplied value.
540 			 */
541 			microtime(&tv1);
542 			timevalsub(&tv1, &tv0);
543 			timevalsub(&utv, &tv1);
544 			if (utv.tv_sec < 0)
545 				timevalclear(&utv);
546 		} else
547 			timevalclear(&utv);
548 #ifdef DEBUG
549 		if (ldebug(select))
550 			printf(LMSG("outgoing timeout (%jd/%ld)"),
551 			    (intmax_t)utv.tv_sec, utv.tv_usec);
552 #endif
553 		ltv.tv_sec = utv.tv_sec;
554 		ltv.tv_usec = utv.tv_usec;
555 		if ((error = copyout(&ltv, args->timeout, sizeof(ltv))))
556 			goto select_out;
557 	}
558 
559 select_out:
560 #ifdef DEBUG
561 	if (ldebug(select))
562 		printf(LMSG("select_out -> %d"), error);
563 #endif
564 	return (error);
565 }
566 
567 int
568 linux_mremap(struct thread *td, struct linux_mremap_args *args)
569 {
570 	struct munmap_args /* {
571 		void *addr;
572 		size_t len;
573 	} */ bsd_args;
574 	int error = 0;
575 
576 #ifdef DEBUG
577 	if (ldebug(mremap))
578 		printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
579 		    (void *)(uintptr_t)args->addr,
580 		    (unsigned long)args->old_len,
581 		    (unsigned long)args->new_len,
582 		    (unsigned long)args->flags);
583 #endif
584 
585 	if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
586 		td->td_retval[0] = 0;
587 		return (EINVAL);
588 	}
589 
590 	/*
591 	 * Check for the page alignment.
592 	 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
593 	 */
594 	if (args->addr & PAGE_MASK) {
595 		td->td_retval[0] = 0;
596 		return (EINVAL);
597 	}
598 
599 	args->new_len = round_page(args->new_len);
600 	args->old_len = round_page(args->old_len);
601 
602 	if (args->new_len > args->old_len) {
603 		td->td_retval[0] = 0;
604 		return (ENOMEM);
605 	}
606 
607 	if (args->new_len < args->old_len) {
608 		bsd_args.addr =
609 		    (caddr_t)((uintptr_t)args->addr + args->new_len);
610 		bsd_args.len = args->old_len - args->new_len;
611 		error = sys_munmap(td, &bsd_args);
612 	}
613 
614 	td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
615 	return (error);
616 }
617 
618 #define LINUX_MS_ASYNC       0x0001
619 #define LINUX_MS_INVALIDATE  0x0002
620 #define LINUX_MS_SYNC        0x0004
621 
622 int
623 linux_msync(struct thread *td, struct linux_msync_args *args)
624 {
625 	struct msync_args bsd_args;
626 
627 	bsd_args.addr = (caddr_t)(uintptr_t)args->addr;
628 	bsd_args.len = (uintptr_t)args->len;
629 	bsd_args.flags = args->fl & ~LINUX_MS_SYNC;
630 
631 	return (sys_msync(td, &bsd_args));
632 }
633 
634 int
635 linux_time(struct thread *td, struct linux_time_args *args)
636 {
637 	struct timeval tv;
638 	l_time_t tm;
639 	int error;
640 
641 #ifdef DEBUG
642 	if (ldebug(time))
643 		printf(ARGS(time, "*"));
644 #endif
645 
646 	microtime(&tv);
647 	tm = tv.tv_sec;
648 	if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
649 		return (error);
650 	td->td_retval[0] = tm;
651 	return (0);
652 }
653 
654 struct l_times_argv {
655 	l_clock_t	tms_utime;
656 	l_clock_t	tms_stime;
657 	l_clock_t	tms_cutime;
658 	l_clock_t	tms_cstime;
659 };
660 
661 
662 /*
663  * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
664  * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
665  * auxiliary vector entry.
666  */
667 #define	CLK_TCK		100
668 
669 #define	CONVOTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
670 #define	CONVNTCK(r)	(r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
671 
672 #define	CONVTCK(r)	(linux_kernver(td) >= LINUX_KERNVER_2004000 ?		\
673 			    CONVNTCK(r) : CONVOTCK(r))
674 
675 int
676 linux_times(struct thread *td, struct linux_times_args *args)
677 {
678 	struct timeval tv, utime, stime, cutime, cstime;
679 	struct l_times_argv tms;
680 	struct proc *p;
681 	int error;
682 
683 #ifdef DEBUG
684 	if (ldebug(times))
685 		printf(ARGS(times, "*"));
686 #endif
687 
688 	if (args->buf != NULL) {
689 		p = td->td_proc;
690 		PROC_LOCK(p);
691 		PROC_SLOCK(p);
692 		calcru(p, &utime, &stime);
693 		PROC_SUNLOCK(p);
694 		calccru(p, &cutime, &cstime);
695 		PROC_UNLOCK(p);
696 
697 		tms.tms_utime = CONVTCK(utime);
698 		tms.tms_stime = CONVTCK(stime);
699 
700 		tms.tms_cutime = CONVTCK(cutime);
701 		tms.tms_cstime = CONVTCK(cstime);
702 
703 		if ((error = copyout(&tms, args->buf, sizeof(tms))))
704 			return (error);
705 	}
706 
707 	microuptime(&tv);
708 	td->td_retval[0] = (int)CONVTCK(tv);
709 	return (0);
710 }
711 
712 int
713 linux_newuname(struct thread *td, struct linux_newuname_args *args)
714 {
715 	struct l_new_utsname utsname;
716 	char osname[LINUX_MAX_UTSNAME];
717 	char osrelease[LINUX_MAX_UTSNAME];
718 	char *p;
719 
720 #ifdef DEBUG
721 	if (ldebug(newuname))
722 		printf(ARGS(newuname, "*"));
723 #endif
724 
725 	linux_get_osname(td, osname);
726 	linux_get_osrelease(td, osrelease);
727 
728 	bzero(&utsname, sizeof(utsname));
729 	strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
730 	getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
731 	getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
732 	strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
733 	strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
734 	for (p = utsname.version; *p != '\0'; ++p)
735 		if (*p == '\n') {
736 			*p = '\0';
737 			break;
738 		}
739 	strlcpy(utsname.machine, linux_platform, LINUX_MAX_UTSNAME);
740 
741 	return (copyout(&utsname, args->buf, sizeof(utsname)));
742 }
743 
744 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
745 struct l_utimbuf {
746 	l_time_t l_actime;
747 	l_time_t l_modtime;
748 };
749 
750 int
751 linux_utime(struct thread *td, struct linux_utime_args *args)
752 {
753 	struct timeval tv[2], *tvp;
754 	struct l_utimbuf lut;
755 	char *fname;
756 	int error;
757 
758 	LCONVPATHEXIST(td, args->fname, &fname);
759 
760 #ifdef DEBUG
761 	if (ldebug(utime))
762 		printf(ARGS(utime, "%s, *"), fname);
763 #endif
764 
765 	if (args->times) {
766 		if ((error = copyin(args->times, &lut, sizeof lut))) {
767 			LFREEPATH(fname);
768 			return (error);
769 		}
770 		tv[0].tv_sec = lut.l_actime;
771 		tv[0].tv_usec = 0;
772 		tv[1].tv_sec = lut.l_modtime;
773 		tv[1].tv_usec = 0;
774 		tvp = tv;
775 	} else
776 		tvp = NULL;
777 
778 	error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
779 	LFREEPATH(fname);
780 	return (error);
781 }
782 
783 int
784 linux_utimes(struct thread *td, struct linux_utimes_args *args)
785 {
786 	l_timeval ltv[2];
787 	struct timeval tv[2], *tvp = NULL;
788 	char *fname;
789 	int error;
790 
791 	LCONVPATHEXIST(td, args->fname, &fname);
792 
793 #ifdef DEBUG
794 	if (ldebug(utimes))
795 		printf(ARGS(utimes, "%s, *"), fname);
796 #endif
797 
798 	if (args->tptr != NULL) {
799 		if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
800 			LFREEPATH(fname);
801 			return (error);
802 		}
803 		tv[0].tv_sec = ltv[0].tv_sec;
804 		tv[0].tv_usec = ltv[0].tv_usec;
805 		tv[1].tv_sec = ltv[1].tv_sec;
806 		tv[1].tv_usec = ltv[1].tv_usec;
807 		tvp = tv;
808 	}
809 
810 	error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
811 	LFREEPATH(fname);
812 	return (error);
813 }
814 
815 int
816 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
817 {
818 	l_timeval ltv[2];
819 	struct timeval tv[2], *tvp = NULL;
820 	char *fname;
821 	int error, dfd;
822 
823 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
824 	LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
825 
826 #ifdef DEBUG
827 	if (ldebug(futimesat))
828 		printf(ARGS(futimesat, "%s, *"), fname);
829 #endif
830 
831 	if (args->utimes != NULL) {
832 		if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
833 			LFREEPATH(fname);
834 			return (error);
835 		}
836 		tv[0].tv_sec = ltv[0].tv_sec;
837 		tv[0].tv_usec = ltv[0].tv_usec;
838 		tv[1].tv_sec = ltv[1].tv_sec;
839 		tv[1].tv_usec = ltv[1].tv_usec;
840 		tvp = tv;
841 	}
842 
843 	error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
844 	LFREEPATH(fname);
845 	return (error);
846 }
847 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
848 
849 int
850 linux_common_wait(struct thread *td, int pid, int *status,
851     int options, struct rusage *ru)
852 {
853 	int error, tmpstat;
854 
855 	error = kern_wait(td, pid, &tmpstat, options, ru);
856 	if (error)
857 		return (error);
858 
859 	if (status) {
860 		tmpstat &= 0xffff;
861 		if (WIFSIGNALED(tmpstat))
862 			tmpstat = (tmpstat & 0xffffff80) |
863 			    BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat));
864 		else if (WIFSTOPPED(tmpstat))
865 			tmpstat = (tmpstat & 0xffff00ff) |
866 			    (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8);
867 		error = copyout(&tmpstat, status, sizeof(int));
868 	}
869 
870 	return (error);
871 }
872 
873 int
874 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
875 {
876 	int options;
877 
878 #ifdef DEBUG
879 	if (ldebug(waitpid))
880 		printf(ARGS(waitpid, "%d, %p, %d"),
881 		    args->pid, (void *)args->status, args->options);
882 #endif
883 	/*
884 	 * this is necessary because the test in kern_wait doesn't work
885 	 * because we mess with the options here
886 	 */
887 	if (args->options & ~(WUNTRACED | WNOHANG | WCONTINUED | __WCLONE))
888 		return (EINVAL);
889 
890 	options = (args->options & (WNOHANG | WUNTRACED));
891 	/* WLINUXCLONE should be equal to __WCLONE, but we make sure */
892 	if (args->options & __WCLONE)
893 		options |= WLINUXCLONE;
894 
895 	return (linux_common_wait(td, args->pid, args->status, options, NULL));
896 }
897 
898 
899 int
900 linux_mknod(struct thread *td, struct linux_mknod_args *args)
901 {
902 	char *path;
903 	int error;
904 
905 	LCONVPATHCREAT(td, args->path, &path);
906 
907 #ifdef DEBUG
908 	if (ldebug(mknod))
909 		printf(ARGS(mknod, "%s, %d, %d"), path, args->mode, args->dev);
910 #endif
911 
912 	switch (args->mode & S_IFMT) {
913 	case S_IFIFO:
914 	case S_IFSOCK:
915 		error = kern_mkfifo(td, path, UIO_SYSSPACE, args->mode);
916 		break;
917 
918 	case S_IFCHR:
919 	case S_IFBLK:
920 		error = kern_mknod(td, path, UIO_SYSSPACE, args->mode,
921 		    args->dev);
922 		break;
923 
924 	case S_IFDIR:
925 		error = EPERM;
926 		break;
927 
928 	case 0:
929 		args->mode |= S_IFREG;
930 		/* FALLTHROUGH */
931 	case S_IFREG:
932 		error = kern_open(td, path, UIO_SYSSPACE,
933 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
934 		if (error == 0)
935 			kern_close(td, td->td_retval[0]);
936 		break;
937 
938 	default:
939 		error = EINVAL;
940 		break;
941 	}
942 	LFREEPATH(path);
943 	return (error);
944 }
945 
946 int
947 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
948 {
949 	char *path;
950 	int error, dfd;
951 
952 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
953 	LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
954 
955 #ifdef DEBUG
956 	if (ldebug(mknodat))
957 		printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev);
958 #endif
959 
960 	switch (args->mode & S_IFMT) {
961 	case S_IFIFO:
962 	case S_IFSOCK:
963 		error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
964 		break;
965 
966 	case S_IFCHR:
967 	case S_IFBLK:
968 		error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
969 		    args->dev);
970 		break;
971 
972 	case S_IFDIR:
973 		error = EPERM;
974 		break;
975 
976 	case 0:
977 		args->mode |= S_IFREG;
978 		/* FALLTHROUGH */
979 	case S_IFREG:
980 		error = kern_openat(td, dfd, path, UIO_SYSSPACE,
981 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
982 		if (error == 0)
983 			kern_close(td, td->td_retval[0]);
984 		break;
985 
986 	default:
987 		error = EINVAL;
988 		break;
989 	}
990 	LFREEPATH(path);
991 	return (error);
992 }
993 
994 /*
995  * UGH! This is just about the dumbest idea I've ever heard!!
996  */
997 int
998 linux_personality(struct thread *td, struct linux_personality_args *args)
999 {
1000 #ifdef DEBUG
1001 	if (ldebug(personality))
1002 		printf(ARGS(personality, "%lu"), (unsigned long)args->per);
1003 #endif
1004 	if (args->per != 0)
1005 		return (EINVAL);
1006 
1007 	/* Yes Jim, it's still a Linux... */
1008 	td->td_retval[0] = 0;
1009 	return (0);
1010 }
1011 
1012 struct l_itimerval {
1013 	l_timeval it_interval;
1014 	l_timeval it_value;
1015 };
1016 
1017 #define	B2L_ITIMERVAL(bip, lip) 					\
1018 	(bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec;		\
1019 	(bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec;	\
1020 	(bip)->it_value.tv_sec = (lip)->it_value.tv_sec;		\
1021 	(bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1022 
1023 int
1024 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1025 {
1026 	int error;
1027 	struct l_itimerval ls;
1028 	struct itimerval aitv, oitv;
1029 
1030 #ifdef DEBUG
1031 	if (ldebug(setitimer))
1032 		printf(ARGS(setitimer, "%p, %p"),
1033 		    (void *)uap->itv, (void *)uap->oitv);
1034 #endif
1035 
1036 	if (uap->itv == NULL) {
1037 		uap->itv = uap->oitv;
1038 		return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1039 	}
1040 
1041 	error = copyin(uap->itv, &ls, sizeof(ls));
1042 	if (error != 0)
1043 		return (error);
1044 	B2L_ITIMERVAL(&aitv, &ls);
1045 #ifdef DEBUG
1046 	if (ldebug(setitimer)) {
1047 		printf("setitimer: value: sec: %jd, usec: %ld\n",
1048 		    (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1049 		printf("setitimer: interval: sec: %jd, usec: %ld\n",
1050 		    (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1051 	}
1052 #endif
1053 	error = kern_setitimer(td, uap->which, &aitv, &oitv);
1054 	if (error != 0 || uap->oitv == NULL)
1055 		return (error);
1056 	B2L_ITIMERVAL(&ls, &oitv);
1057 
1058 	return (copyout(&ls, uap->oitv, sizeof(ls)));
1059 }
1060 
1061 int
1062 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1063 {
1064 	int error;
1065 	struct l_itimerval ls;
1066 	struct itimerval aitv;
1067 
1068 #ifdef DEBUG
1069 	if (ldebug(getitimer))
1070 		printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1071 #endif
1072 	error = kern_getitimer(td, uap->which, &aitv);
1073 	if (error != 0)
1074 		return (error);
1075 	B2L_ITIMERVAL(&ls, &aitv);
1076 	return (copyout(&ls, uap->itv, sizeof(ls)));
1077 }
1078 
1079 int
1080 linux_nice(struct thread *td, struct linux_nice_args *args)
1081 {
1082 	struct setpriority_args bsd_args;
1083 
1084 	bsd_args.which = PRIO_PROCESS;
1085 	bsd_args.who = 0;		/* current process */
1086 	bsd_args.prio = args->inc;
1087 	return (sys_setpriority(td, &bsd_args));
1088 }
1089 
1090 int
1091 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1092 {
1093 	struct ucred *newcred, *oldcred;
1094 	l_gid_t *linux_gidset;
1095 	gid_t *bsd_gidset;
1096 	int ngrp, error;
1097 	struct proc *p;
1098 
1099 	ngrp = args->gidsetsize;
1100 	if (ngrp < 0 || ngrp >= ngroups_max + 1)
1101 		return (EINVAL);
1102 	linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_TEMP, M_WAITOK);
1103 	error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1104 	if (error)
1105 		goto out;
1106 	newcred = crget();
1107 	p = td->td_proc;
1108 	PROC_LOCK(p);
1109 	oldcred = crcopysafe(p, newcred);
1110 
1111 	/*
1112 	 * cr_groups[0] holds egid. Setting the whole set from
1113 	 * the supplied set will cause egid to be changed too.
1114 	 * Keep cr_groups[0] unchanged to prevent that.
1115 	 */
1116 
1117 	if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) {
1118 		PROC_UNLOCK(p);
1119 		crfree(newcred);
1120 		goto out;
1121 	}
1122 
1123 	if (ngrp > 0) {
1124 		newcred->cr_ngroups = ngrp + 1;
1125 
1126 		bsd_gidset = newcred->cr_groups;
1127 		ngrp--;
1128 		while (ngrp >= 0) {
1129 			bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1130 			ngrp--;
1131 		}
1132 	} else
1133 		newcred->cr_ngroups = 1;
1134 
1135 	setsugid(p);
1136 	p->p_ucred = newcred;
1137 	PROC_UNLOCK(p);
1138 	crfree(oldcred);
1139 	error = 0;
1140 out:
1141 	free(linux_gidset, M_TEMP);
1142 	return (error);
1143 }
1144 
1145 int
1146 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1147 {
1148 	struct ucred *cred;
1149 	l_gid_t *linux_gidset;
1150 	gid_t *bsd_gidset;
1151 	int bsd_gidsetsz, ngrp, error;
1152 
1153 	cred = td->td_ucred;
1154 	bsd_gidset = cred->cr_groups;
1155 	bsd_gidsetsz = cred->cr_ngroups - 1;
1156 
1157 	/*
1158 	 * cr_groups[0] holds egid. Returning the whole set
1159 	 * here will cause a duplicate. Exclude cr_groups[0]
1160 	 * to prevent that.
1161 	 */
1162 
1163 	if ((ngrp = args->gidsetsize) == 0) {
1164 		td->td_retval[0] = bsd_gidsetsz;
1165 		return (0);
1166 	}
1167 
1168 	if (ngrp < bsd_gidsetsz)
1169 		return (EINVAL);
1170 
1171 	ngrp = 0;
1172 	linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1173 	    M_TEMP, M_WAITOK);
1174 	while (ngrp < bsd_gidsetsz) {
1175 		linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1176 		ngrp++;
1177 	}
1178 
1179 	error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1180 	free(linux_gidset, M_TEMP);
1181 	if (error)
1182 		return (error);
1183 
1184 	td->td_retval[0] = ngrp;
1185 	return (0);
1186 }
1187 
1188 int
1189 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1190 {
1191 	struct rlimit bsd_rlim;
1192 	struct l_rlimit rlim;
1193 	u_int which;
1194 	int error;
1195 
1196 #ifdef DEBUG
1197 	if (ldebug(setrlimit))
1198 		printf(ARGS(setrlimit, "%d, %p"),
1199 		    args->resource, (void *)args->rlim);
1200 #endif
1201 
1202 	if (args->resource >= LINUX_RLIM_NLIMITS)
1203 		return (EINVAL);
1204 
1205 	which = linux_to_bsd_resource[args->resource];
1206 	if (which == -1)
1207 		return (EINVAL);
1208 
1209 	error = copyin(args->rlim, &rlim, sizeof(rlim));
1210 	if (error)
1211 		return (error);
1212 
1213 	bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1214 	bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1215 	return (kern_setrlimit(td, which, &bsd_rlim));
1216 }
1217 
1218 int
1219 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1220 {
1221 	struct l_rlimit rlim;
1222 	struct proc *p = td->td_proc;
1223 	struct rlimit bsd_rlim;
1224 	u_int which;
1225 
1226 #ifdef DEBUG
1227 	if (ldebug(old_getrlimit))
1228 		printf(ARGS(old_getrlimit, "%d, %p"),
1229 		    args->resource, (void *)args->rlim);
1230 #endif
1231 
1232 	if (args->resource >= LINUX_RLIM_NLIMITS)
1233 		return (EINVAL);
1234 
1235 	which = linux_to_bsd_resource[args->resource];
1236 	if (which == -1)
1237 		return (EINVAL);
1238 
1239 	PROC_LOCK(p);
1240 	lim_rlimit(p, which, &bsd_rlim);
1241 	PROC_UNLOCK(p);
1242 
1243 #ifdef COMPAT_LINUX32
1244 	rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1245 	if (rlim.rlim_cur == UINT_MAX)
1246 		rlim.rlim_cur = INT_MAX;
1247 	rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1248 	if (rlim.rlim_max == UINT_MAX)
1249 		rlim.rlim_max = INT_MAX;
1250 #else
1251 	rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1252 	if (rlim.rlim_cur == ULONG_MAX)
1253 		rlim.rlim_cur = LONG_MAX;
1254 	rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1255 	if (rlim.rlim_max == ULONG_MAX)
1256 		rlim.rlim_max = LONG_MAX;
1257 #endif
1258 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1259 }
1260 
1261 int
1262 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1263 {
1264 	struct l_rlimit rlim;
1265 	struct proc *p = td->td_proc;
1266 	struct rlimit bsd_rlim;
1267 	u_int which;
1268 
1269 #ifdef DEBUG
1270 	if (ldebug(getrlimit))
1271 		printf(ARGS(getrlimit, "%d, %p"),
1272 		    args->resource, (void *)args->rlim);
1273 #endif
1274 
1275 	if (args->resource >= LINUX_RLIM_NLIMITS)
1276 		return (EINVAL);
1277 
1278 	which = linux_to_bsd_resource[args->resource];
1279 	if (which == -1)
1280 		return (EINVAL);
1281 
1282 	PROC_LOCK(p);
1283 	lim_rlimit(p, which, &bsd_rlim);
1284 	PROC_UNLOCK(p);
1285 
1286 	rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1287 	rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1288 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1289 }
1290 
1291 int
1292 linux_sched_setscheduler(struct thread *td,
1293     struct linux_sched_setscheduler_args *args)
1294 {
1295 	struct sched_setscheduler_args bsd;
1296 
1297 #ifdef DEBUG
1298 	if (ldebug(sched_setscheduler))
1299 		printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1300 		    args->pid, args->policy, (const void *)args->param);
1301 #endif
1302 
1303 	switch (args->policy) {
1304 	case LINUX_SCHED_OTHER:
1305 		bsd.policy = SCHED_OTHER;
1306 		break;
1307 	case LINUX_SCHED_FIFO:
1308 		bsd.policy = SCHED_FIFO;
1309 		break;
1310 	case LINUX_SCHED_RR:
1311 		bsd.policy = SCHED_RR;
1312 		break;
1313 	default:
1314 		return (EINVAL);
1315 	}
1316 
1317 	bsd.pid = args->pid;
1318 	bsd.param = (struct sched_param *)args->param;
1319 	return (sys_sched_setscheduler(td, &bsd));
1320 }
1321 
1322 int
1323 linux_sched_getscheduler(struct thread *td,
1324     struct linux_sched_getscheduler_args *args)
1325 {
1326 	struct sched_getscheduler_args bsd;
1327 	int error;
1328 
1329 #ifdef DEBUG
1330 	if (ldebug(sched_getscheduler))
1331 		printf(ARGS(sched_getscheduler, "%d"), args->pid);
1332 #endif
1333 
1334 	bsd.pid = args->pid;
1335 	error = sys_sched_getscheduler(td, &bsd);
1336 
1337 	switch (td->td_retval[0]) {
1338 	case SCHED_OTHER:
1339 		td->td_retval[0] = LINUX_SCHED_OTHER;
1340 		break;
1341 	case SCHED_FIFO:
1342 		td->td_retval[0] = LINUX_SCHED_FIFO;
1343 		break;
1344 	case SCHED_RR:
1345 		td->td_retval[0] = LINUX_SCHED_RR;
1346 		break;
1347 	}
1348 
1349 	return (error);
1350 }
1351 
1352 int
1353 linux_sched_get_priority_max(struct thread *td,
1354     struct linux_sched_get_priority_max_args *args)
1355 {
1356 	struct sched_get_priority_max_args bsd;
1357 
1358 #ifdef DEBUG
1359 	if (ldebug(sched_get_priority_max))
1360 		printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1361 #endif
1362 
1363 	switch (args->policy) {
1364 	case LINUX_SCHED_OTHER:
1365 		bsd.policy = SCHED_OTHER;
1366 		break;
1367 	case LINUX_SCHED_FIFO:
1368 		bsd.policy = SCHED_FIFO;
1369 		break;
1370 	case LINUX_SCHED_RR:
1371 		bsd.policy = SCHED_RR;
1372 		break;
1373 	default:
1374 		return (EINVAL);
1375 	}
1376 	return (sys_sched_get_priority_max(td, &bsd));
1377 }
1378 
1379 int
1380 linux_sched_get_priority_min(struct thread *td,
1381     struct linux_sched_get_priority_min_args *args)
1382 {
1383 	struct sched_get_priority_min_args bsd;
1384 
1385 #ifdef DEBUG
1386 	if (ldebug(sched_get_priority_min))
1387 		printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1388 #endif
1389 
1390 	switch (args->policy) {
1391 	case LINUX_SCHED_OTHER:
1392 		bsd.policy = SCHED_OTHER;
1393 		break;
1394 	case LINUX_SCHED_FIFO:
1395 		bsd.policy = SCHED_FIFO;
1396 		break;
1397 	case LINUX_SCHED_RR:
1398 		bsd.policy = SCHED_RR;
1399 		break;
1400 	default:
1401 		return (EINVAL);
1402 	}
1403 	return (sys_sched_get_priority_min(td, &bsd));
1404 }
1405 
1406 #define REBOOT_CAD_ON	0x89abcdef
1407 #define REBOOT_CAD_OFF	0
1408 #define REBOOT_HALT	0xcdef0123
1409 #define REBOOT_RESTART	0x01234567
1410 #define REBOOT_RESTART2	0xA1B2C3D4
1411 #define REBOOT_POWEROFF	0x4321FEDC
1412 #define REBOOT_MAGIC1	0xfee1dead
1413 #define REBOOT_MAGIC2	0x28121969
1414 #define REBOOT_MAGIC2A	0x05121996
1415 #define REBOOT_MAGIC2B	0x16041998
1416 
1417 int
1418 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1419 {
1420 	struct reboot_args bsd_args;
1421 
1422 #ifdef DEBUG
1423 	if (ldebug(reboot))
1424 		printf(ARGS(reboot, "0x%x"), args->cmd);
1425 #endif
1426 
1427 	if (args->magic1 != REBOOT_MAGIC1)
1428 		return (EINVAL);
1429 
1430 	switch (args->magic2) {
1431 	case REBOOT_MAGIC2:
1432 	case REBOOT_MAGIC2A:
1433 	case REBOOT_MAGIC2B:
1434 		break;
1435 	default:
1436 		return (EINVAL);
1437 	}
1438 
1439 	switch (args->cmd) {
1440 	case REBOOT_CAD_ON:
1441 	case REBOOT_CAD_OFF:
1442 		return (priv_check(td, PRIV_REBOOT));
1443 	case REBOOT_HALT:
1444 		bsd_args.opt = RB_HALT;
1445 		break;
1446 	case REBOOT_RESTART:
1447 	case REBOOT_RESTART2:
1448 		bsd_args.opt = 0;
1449 		break;
1450 	case REBOOT_POWEROFF:
1451 		bsd_args.opt = RB_POWEROFF;
1452 		break;
1453 	default:
1454 		return (EINVAL);
1455 	}
1456 	return (sys_reboot(td, &bsd_args));
1457 }
1458 
1459 
1460 /*
1461  * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1462  * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
1463  * are assumed to be preserved. The following lightweight syscalls fixes
1464  * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
1465  *
1466  * linux_getpid() - MP SAFE
1467  * linux_getgid() - MP SAFE
1468  * linux_getuid() - MP SAFE
1469  */
1470 
1471 int
1472 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1473 {
1474 	struct linux_emuldata *em;
1475 
1476 #ifdef DEBUG
1477 	if (ldebug(getpid))
1478 		printf(ARGS(getpid, ""));
1479 #endif
1480 
1481 	if (linux_use26(td)) {
1482 		em = em_find(td->td_proc, EMUL_DONTLOCK);
1483 		KASSERT(em != NULL, ("getpid: emuldata not found.\n"));
1484 		td->td_retval[0] = em->shared->group_pid;
1485 	} else {
1486 		td->td_retval[0] = td->td_proc->p_pid;
1487 	}
1488 
1489 	return (0);
1490 }
1491 
1492 int
1493 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1494 {
1495 
1496 #ifdef DEBUG
1497 	if (ldebug(gettid))
1498 		printf(ARGS(gettid, ""));
1499 #endif
1500 
1501 	td->td_retval[0] = td->td_proc->p_pid;
1502 	return (0);
1503 }
1504 
1505 
1506 int
1507 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1508 {
1509 	struct linux_emuldata *em;
1510 	struct proc *p, *pp;
1511 
1512 #ifdef DEBUG
1513 	if (ldebug(getppid))
1514 		printf(ARGS(getppid, ""));
1515 #endif
1516 
1517 	if (!linux_use26(td)) {
1518 		PROC_LOCK(td->td_proc);
1519 		td->td_retval[0] = td->td_proc->p_pptr->p_pid;
1520 		PROC_UNLOCK(td->td_proc);
1521 		return (0);
1522 	}
1523 
1524 	em = em_find(td->td_proc, EMUL_DONTLOCK);
1525 
1526 	KASSERT(em != NULL, ("getppid: process emuldata not found.\n"));
1527 
1528 	/* find the group leader */
1529 	p = pfind(em->shared->group_pid);
1530 
1531 	if (p == NULL) {
1532 #ifdef DEBUG
1533 	   	printf(LMSG("parent process not found.\n"));
1534 #endif
1535 		return (0);
1536 	}
1537 
1538 	pp = p->p_pptr;		/* switch to parent */
1539 	PROC_LOCK(pp);
1540 	PROC_UNLOCK(p);
1541 
1542 	/* if its also linux process */
1543 	if (pp->p_sysent == &elf_linux_sysvec) {
1544 		em = em_find(pp, EMUL_DONTLOCK);
1545 		KASSERT(em != NULL, ("getppid: parent emuldata not found.\n"));
1546 
1547 		td->td_retval[0] = em->shared->group_pid;
1548 	} else
1549 		td->td_retval[0] = pp->p_pid;
1550 
1551 	PROC_UNLOCK(pp);
1552 
1553 	return (0);
1554 }
1555 
1556 int
1557 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1558 {
1559 
1560 #ifdef DEBUG
1561 	if (ldebug(getgid))
1562 		printf(ARGS(getgid, ""));
1563 #endif
1564 
1565 	td->td_retval[0] = td->td_ucred->cr_rgid;
1566 	return (0);
1567 }
1568 
1569 int
1570 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1571 {
1572 
1573 #ifdef DEBUG
1574 	if (ldebug(getuid))
1575 		printf(ARGS(getuid, ""));
1576 #endif
1577 
1578 	td->td_retval[0] = td->td_ucred->cr_ruid;
1579 	return (0);
1580 }
1581 
1582 
1583 int
1584 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1585 {
1586 	struct getsid_args bsd;
1587 
1588 #ifdef DEBUG
1589 	if (ldebug(getsid))
1590 		printf(ARGS(getsid, "%i"), args->pid);
1591 #endif
1592 
1593 	bsd.pid = args->pid;
1594 	return (sys_getsid(td, &bsd));
1595 }
1596 
1597 int
1598 linux_nosys(struct thread *td, struct nosys_args *ignore)
1599 {
1600 
1601 	return (ENOSYS);
1602 }
1603 
1604 int
1605 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1606 {
1607 	struct getpriority_args bsd_args;
1608 	int error;
1609 
1610 #ifdef DEBUG
1611 	if (ldebug(getpriority))
1612 		printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
1613 #endif
1614 
1615 	bsd_args.which = args->which;
1616 	bsd_args.who = args->who;
1617 	error = sys_getpriority(td, &bsd_args);
1618 	td->td_retval[0] = 20 - td->td_retval[0];
1619 	return (error);
1620 }
1621 
1622 int
1623 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1624 {
1625 	int name[2];
1626 
1627 #ifdef DEBUG
1628 	if (ldebug(sethostname))
1629 		printf(ARGS(sethostname, "*, %i"), args->len);
1630 #endif
1631 
1632 	name[0] = CTL_KERN;
1633 	name[1] = KERN_HOSTNAME;
1634 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1635 	    args->len, 0, 0));
1636 }
1637 
1638 int
1639 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1640 {
1641 	int name[2];
1642 
1643 #ifdef DEBUG
1644 	if (ldebug(setdomainname))
1645 		printf(ARGS(setdomainname, "*, %i"), args->len);
1646 #endif
1647 
1648 	name[0] = CTL_KERN;
1649 	name[1] = KERN_NISDOMAINNAME;
1650 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1651 	    args->len, 0, 0));
1652 }
1653 
1654 int
1655 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1656 {
1657 	struct linux_emuldata *em;
1658 
1659 #ifdef DEBUG
1660 	if (ldebug(exit_group))
1661 		printf(ARGS(exit_group, "%i"), args->error_code);
1662 #endif
1663 
1664 	em = em_find(td->td_proc, EMUL_DONTLOCK);
1665 	if (em->shared->refs > 1) {
1666 		EMUL_SHARED_WLOCK(&emul_shared_lock);
1667 		em->shared->flags |= EMUL_SHARED_HASXSTAT;
1668 		em->shared->xstat = W_EXITCODE(args->error_code, 0);
1669 		EMUL_SHARED_WUNLOCK(&emul_shared_lock);
1670 		if (linux_use26(td))
1671 			linux_kill_threads(td, SIGKILL);
1672 	}
1673 
1674 	/*
1675 	 * XXX: we should send a signal to the parent if
1676 	 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1677 	 * as it doesnt occur often.
1678 	 */
1679 	exit1(td, W_EXITCODE(args->error_code, 0));
1680 
1681 	return (0);
1682 }
1683 
1684 #define _LINUX_CAPABILITY_VERSION  0x19980330
1685 
1686 struct l_user_cap_header {
1687 	l_int	version;
1688 	l_int	pid;
1689 };
1690 
1691 struct l_user_cap_data {
1692 	l_int	effective;
1693 	l_int	permitted;
1694 	l_int	inheritable;
1695 };
1696 
1697 int
1698 linux_capget(struct thread *td, struct linux_capget_args *args)
1699 {
1700 	struct l_user_cap_header luch;
1701 	struct l_user_cap_data lucd;
1702 	int error;
1703 
1704 	if (args->hdrp == NULL)
1705 		return (EFAULT);
1706 
1707 	error = copyin(args->hdrp, &luch, sizeof(luch));
1708 	if (error != 0)
1709 		return (error);
1710 
1711 	if (luch.version != _LINUX_CAPABILITY_VERSION) {
1712 		luch.version = _LINUX_CAPABILITY_VERSION;
1713 		error = copyout(&luch, args->hdrp, sizeof(luch));
1714 		if (error)
1715 			return (error);
1716 		return (EINVAL);
1717 	}
1718 
1719 	if (luch.pid)
1720 		return (EPERM);
1721 
1722 	if (args->datap) {
1723 		/*
1724 		 * The current implementation doesn't support setting
1725 		 * a capability (it's essentially a stub) so indicate
1726 		 * that no capabilities are currently set or available
1727 		 * to request.
1728 		 */
1729 		bzero (&lucd, sizeof(lucd));
1730 		error = copyout(&lucd, args->datap, sizeof(lucd));
1731 	}
1732 
1733 	return (error);
1734 }
1735 
1736 int
1737 linux_capset(struct thread *td, struct linux_capset_args *args)
1738 {
1739 	struct l_user_cap_header luch;
1740 	struct l_user_cap_data lucd;
1741 	int error;
1742 
1743 	if (args->hdrp == NULL || args->datap == NULL)
1744 		return (EFAULT);
1745 
1746 	error = copyin(args->hdrp, &luch, sizeof(luch));
1747 	if (error != 0)
1748 		return (error);
1749 
1750 	if (luch.version != _LINUX_CAPABILITY_VERSION) {
1751 		luch.version = _LINUX_CAPABILITY_VERSION;
1752 		error = copyout(&luch, args->hdrp, sizeof(luch));
1753 		if (error)
1754 			return (error);
1755 		return (EINVAL);
1756 	}
1757 
1758 	if (luch.pid)
1759 		return (EPERM);
1760 
1761 	error = copyin(args->datap, &lucd, sizeof(lucd));
1762 	if (error != 0)
1763 		return (error);
1764 
1765 	/* We currently don't support setting any capabilities. */
1766 	if (lucd.effective || lucd.permitted || lucd.inheritable) {
1767 		linux_msg(td,
1768 			  "capset effective=0x%x, permitted=0x%x, "
1769 			  "inheritable=0x%x is not implemented",
1770 			  (int)lucd.effective, (int)lucd.permitted,
1771 			  (int)lucd.inheritable);
1772 		return (EPERM);
1773 	}
1774 
1775 	return (0);
1776 }
1777 
1778 int
1779 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1780 {
1781 	int error = 0, max_size;
1782 	struct proc *p = td->td_proc;
1783 	char comm[LINUX_MAX_COMM_LEN];
1784 	struct linux_emuldata *em;
1785 	int pdeath_signal;
1786 
1787 #ifdef DEBUG
1788 	if (ldebug(prctl))
1789 		printf(ARGS(prctl, "%d, %d, %d, %d, %d"), args->option,
1790 		    args->arg2, args->arg3, args->arg4, args->arg5);
1791 #endif
1792 
1793 	switch (args->option) {
1794 	case LINUX_PR_SET_PDEATHSIG:
1795 		if (!LINUX_SIG_VALID(args->arg2))
1796 			return (EINVAL);
1797 		em = em_find(p, EMUL_DOLOCK);
1798 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1799 		em->pdeath_signal = args->arg2;
1800 		EMUL_UNLOCK(&emul_lock);
1801 		break;
1802 	case LINUX_PR_GET_PDEATHSIG:
1803 		em = em_find(p, EMUL_DOLOCK);
1804 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1805 		pdeath_signal = em->pdeath_signal;
1806 		EMUL_UNLOCK(&emul_lock);
1807 		error = copyout(&pdeath_signal,
1808 		    (void *)(register_t)args->arg2,
1809 		    sizeof(pdeath_signal));
1810 		break;
1811 	case LINUX_PR_GET_KEEPCAPS:
1812 		/*
1813 		 * Indicate that we always clear the effective and
1814 		 * permitted capability sets when the user id becomes
1815 		 * non-zero (actually the capability sets are simply
1816 		 * always zero in the current implementation).
1817 		 */
1818 		td->td_retval[0] = 0;
1819 		break;
1820 	case LINUX_PR_SET_KEEPCAPS:
1821 		/*
1822 		 * Ignore requests to keep the effective and permitted
1823 		 * capability sets when the user id becomes non-zero.
1824 		 */
1825 		break;
1826 	case LINUX_PR_SET_NAME:
1827 		/*
1828 		 * To be on the safe side we need to make sure to not
1829 		 * overflow the size a linux program expects. We already
1830 		 * do this here in the copyin, so that we don't need to
1831 		 * check on copyout.
1832 		 */
1833 		max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1834 		error = copyinstr((void *)(register_t)args->arg2, comm,
1835 		    max_size, NULL);
1836 
1837 		/* Linux silently truncates the name if it is too long. */
1838 		if (error == ENAMETOOLONG) {
1839 			/*
1840 			 * XXX: copyinstr() isn't documented to populate the
1841 			 * array completely, so do a copyin() to be on the
1842 			 * safe side. This should be changed in case
1843 			 * copyinstr() is changed to guarantee this.
1844 			 */
1845 			error = copyin((void *)(register_t)args->arg2, comm,
1846 			    max_size - 1);
1847 			comm[max_size - 1] = '\0';
1848 		}
1849 		if (error)
1850 			return (error);
1851 
1852 		PROC_LOCK(p);
1853 		strlcpy(p->p_comm, comm, sizeof(p->p_comm));
1854 		PROC_UNLOCK(p);
1855 		break;
1856 	case LINUX_PR_GET_NAME:
1857 		PROC_LOCK(p);
1858 		strlcpy(comm, p->p_comm, sizeof(comm));
1859 		PROC_UNLOCK(p);
1860 		error = copyout(comm, (void *)(register_t)args->arg2,
1861 		    strlen(comm) + 1);
1862 		break;
1863 	default:
1864 		error = EINVAL;
1865 		break;
1866 	}
1867 
1868 	return (error);
1869 }
1870 
1871 /*
1872  * Get affinity of a process.
1873  */
1874 int
1875 linux_sched_getaffinity(struct thread *td,
1876     struct linux_sched_getaffinity_args *args)
1877 {
1878 	int error;
1879 	struct cpuset_getaffinity_args cga;
1880 
1881 #ifdef DEBUG
1882 	if (ldebug(sched_getaffinity))
1883 		printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid,
1884 		    args->len);
1885 #endif
1886 	if (args->len < sizeof(cpuset_t))
1887 		return (EINVAL);
1888 
1889 	cga.level = CPU_LEVEL_WHICH;
1890 	cga.which = CPU_WHICH_PID;
1891 	cga.id = args->pid;
1892 	cga.cpusetsize = sizeof(cpuset_t);
1893 	cga.mask = (cpuset_t *) args->user_mask_ptr;
1894 
1895 	if ((error = sys_cpuset_getaffinity(td, &cga)) == 0)
1896 		td->td_retval[0] = sizeof(cpuset_t);
1897 
1898 	return (error);
1899 }
1900 
1901 /*
1902  *  Set affinity of a process.
1903  */
1904 int
1905 linux_sched_setaffinity(struct thread *td,
1906     struct linux_sched_setaffinity_args *args)
1907 {
1908 	struct cpuset_setaffinity_args csa;
1909 
1910 #ifdef DEBUG
1911 	if (ldebug(sched_setaffinity))
1912 		printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid,
1913 		    args->len);
1914 #endif
1915 	if (args->len < sizeof(cpuset_t))
1916 		return (EINVAL);
1917 
1918 	csa.level = CPU_LEVEL_WHICH;
1919 	csa.which = CPU_WHICH_PID;
1920 	csa.id = args->pid;
1921 	csa.cpusetsize = sizeof(cpuset_t);
1922 	csa.mask = (cpuset_t *) args->user_mask_ptr;
1923 
1924 	return (sys_cpuset_setaffinity(td, &csa));
1925 }
1926