xref: /freebsd/sys/compat/linux/linux_misc.c (revision e45b2259ec3d7d128e1da66cf5b1a76f6618d9e9)
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("$FreeBSD$");
32 
33 #include "opt_compat.h"
34 #include "opt_mac.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/resourcevar.h>
55 #include <sys/sched.h>
56 #include <sys/signalvar.h>
57 #include <sys/stat.h>
58 #include <sys/syscallsubr.h>
59 #include <sys/sysctl.h>
60 #include <sys/sysproto.h>
61 #include <sys/systm.h>
62 #include <sys/time.h>
63 #include <sys/vmmeter.h>
64 #include <sys/vnode.h>
65 #include <sys/wait.h>
66 #include <sys/cpuset.h>
67 #include <sys/vimage.h>
68 
69 #include <security/mac/mac_framework.h>
70 
71 #include <vm/vm.h>
72 #include <vm/pmap.h>
73 #include <vm/vm_kern.h>
74 #include <vm/vm_map.h>
75 #include <vm/vm_extern.h>
76 #include <vm/vm_object.h>
77 #include <vm/swap_pager.h>
78 
79 #ifdef COMPAT_LINUX32
80 #include <machine/../linux32/linux.h>
81 #include <machine/../linux32/linux32_proto.h>
82 #else
83 #include <machine/../linux/linux.h>
84 #include <machine/../linux/linux_proto.h>
85 #endif
86 
87 #include <compat/linux/linux_file.h>
88 #include <compat/linux/linux_mib.h>
89 #include <compat/linux/linux_signal.h>
90 #include <compat/linux/linux_util.h>
91 #include <compat/linux/linux_sysproto.h>
92 #include <compat/linux/linux_emul.h>
93 #include <compat/linux/linux_misc.h>
94 
95 #ifdef __i386__
96 #include <machine/cputypes.h>
97 #endif
98 
99 #define BSD_TO_LINUX_SIGNAL(sig)	\
100 	(((sig) <= LINUX_SIGTBLSZ) ? bsd_to_linux_signal[_SIG_IDX(sig)] : sig)
101 
102 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
103 	RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
104 	RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
105 	RLIMIT_MEMLOCK, RLIMIT_AS
106 };
107 
108 struct l_sysinfo {
109 	l_long		uptime;		/* Seconds since boot */
110 	l_ulong		loads[3];	/* 1, 5, and 15 minute load averages */
111 #define LINUX_SYSINFO_LOADS_SCALE 65536
112 	l_ulong		totalram;	/* Total usable main memory size */
113 	l_ulong		freeram;	/* Available memory size */
114 	l_ulong		sharedram;	/* Amount of shared memory */
115 	l_ulong		bufferram;	/* Memory used by buffers */
116 	l_ulong		totalswap;	/* Total swap space size */
117 	l_ulong		freeswap;	/* swap space still available */
118 	l_ushort	procs;		/* Number of current processes */
119 	l_ushort	pads;
120 	l_ulong		totalbig;
121 	l_ulong		freebig;
122 	l_uint		mem_unit;
123 	char		_f[20-2*sizeof(l_long)-sizeof(l_int)];	/* padding */
124 };
125 int
126 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
127 {
128 	struct l_sysinfo sysinfo;
129 	vm_object_t object;
130 	int i, j;
131 	struct timespec ts;
132 
133 	getnanouptime(&ts);
134 	if (ts.tv_nsec != 0)
135 		ts.tv_sec++;
136 	sysinfo.uptime = ts.tv_sec;
137 
138 	/* Use the information from the mib to get our load averages */
139 	for (i = 0; i < 3; i++)
140 		sysinfo.loads[i] = averunnable.ldavg[i] *
141 		    LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
142 
143 	sysinfo.totalram = physmem * PAGE_SIZE;
144 	sysinfo.freeram = sysinfo.totalram - cnt.v_wire_count * PAGE_SIZE;
145 
146 	sysinfo.sharedram = 0;
147 	mtx_lock(&vm_object_list_mtx);
148 	TAILQ_FOREACH(object, &vm_object_list, object_list)
149 		if (object->shadow_count > 1)
150 			sysinfo.sharedram += object->resident_page_count;
151 	mtx_unlock(&vm_object_list_mtx);
152 
153 	sysinfo.sharedram *= PAGE_SIZE;
154 	sysinfo.bufferram = 0;
155 
156 	swap_pager_status(&i, &j);
157 	sysinfo.totalswap = i * PAGE_SIZE;
158 	sysinfo.freeswap = (i - j) * PAGE_SIZE;
159 
160 	sysinfo.procs = nprocs;
161 
162 	/* The following are only present in newer Linux kernels. */
163 	sysinfo.totalbig = 0;
164 	sysinfo.freebig = 0;
165 	sysinfo.mem_unit = 1;
166 
167 	return copyout(&sysinfo, args->info, sizeof(sysinfo));
168 }
169 
170 int
171 linux_alarm(struct thread *td, struct linux_alarm_args *args)
172 {
173 	struct itimerval it, old_it;
174 	u_int secs;
175 	int error;
176 
177 #ifdef DEBUG
178 	if (ldebug(alarm))
179 		printf(ARGS(alarm, "%u"), args->secs);
180 #endif
181 
182 	secs = args->secs;
183 
184 	if (secs > INT_MAX)
185 		secs = INT_MAX;
186 
187 	it.it_value.tv_sec = (long) secs;
188 	it.it_value.tv_usec = 0;
189 	it.it_interval.tv_sec = 0;
190 	it.it_interval.tv_usec = 0;
191 	error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
192 	if (error)
193 		return (error);
194 	if (timevalisset(&old_it.it_value)) {
195 		if (old_it.it_value.tv_usec != 0)
196 			old_it.it_value.tv_sec++;
197 		td->td_retval[0] = old_it.it_value.tv_sec;
198 	}
199 	return (0);
200 }
201 
202 int
203 linux_brk(struct thread *td, struct linux_brk_args *args)
204 {
205 	struct vmspace *vm = td->td_proc->p_vmspace;
206 	vm_offset_t new, old;
207 	struct obreak_args /* {
208 		char * nsize;
209 	} */ tmp;
210 
211 #ifdef DEBUG
212 	if (ldebug(brk))
213 		printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
214 #endif
215 	old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize);
216 	new = (vm_offset_t)args->dsend;
217 	tmp.nsize = (char *)new;
218 	if (((caddr_t)new > vm->vm_daddr) && !obreak(td, &tmp))
219 		td->td_retval[0] = (long)new;
220 	else
221 		td->td_retval[0] = (long)old;
222 
223 	return 0;
224 }
225 
226 #if defined(__i386__)
227 /* XXX: what about amd64/linux32? */
228 
229 int
230 linux_uselib(struct thread *td, struct linux_uselib_args *args)
231 {
232 	struct nameidata ni;
233 	struct vnode *vp;
234 	struct exec *a_out;
235 	struct vattr attr;
236 	vm_offset_t vmaddr;
237 	unsigned long file_offset;
238 	vm_offset_t buffer;
239 	unsigned long bss_size;
240 	char *library;
241 	int error;
242 	int locked, vfslocked;
243 
244 	LCONVPATHEXIST(td, args->library, &library);
245 
246 #ifdef DEBUG
247 	if (ldebug(uselib))
248 		printf(ARGS(uselib, "%s"), library);
249 #endif
250 
251 	a_out = NULL;
252 	vfslocked = 0;
253 	locked = 0;
254 	vp = NULL;
255 
256 	NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
257 	    UIO_SYSSPACE, library, td);
258 	error = namei(&ni);
259 	LFREEPATH(library);
260 	if (error)
261 		goto cleanup;
262 
263 	vp = ni.ni_vp;
264 	vfslocked = NDHASGIANT(&ni);
265 	NDFREE(&ni, NDF_ONLY_PNBUF);
266 
267 	/*
268 	 * From here on down, we have a locked vnode that must be unlocked.
269 	 * XXX: The code below largely duplicates exec_check_permissions().
270 	 */
271 	locked = 1;
272 
273 	/* Writable? */
274 	if (vp->v_writecount) {
275 		error = ETXTBSY;
276 		goto cleanup;
277 	}
278 
279 	/* Executable? */
280 	error = VOP_GETATTR(vp, &attr, td->td_ucred);
281 	if (error)
282 		goto cleanup;
283 
284 	if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
285 	    ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
286 		/* EACCESS is what exec(2) returns. */
287 		error = ENOEXEC;
288 		goto cleanup;
289 	}
290 
291 	/* Sensible size? */
292 	if (attr.va_size == 0) {
293 		error = ENOEXEC;
294 		goto cleanup;
295 	}
296 
297 	/* Can we access it? */
298 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
299 	if (error)
300 		goto cleanup;
301 
302 	/*
303 	 * XXX: This should use vn_open() so that it is properly authorized,
304 	 * and to reduce code redundancy all over the place here.
305 	 * XXX: Not really, it duplicates far more of exec_check_permissions()
306 	 * than vn_open().
307 	 */
308 #ifdef MAC
309 	error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
310 	if (error)
311 		goto cleanup;
312 #endif
313 	error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
314 	if (error)
315 		goto cleanup;
316 
317 	/* Pull in executable header into kernel_map */
318 	error = vm_mmap(kernel_map, (vm_offset_t *)&a_out, PAGE_SIZE,
319 	    VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
320 	if (error)
321 		goto cleanup;
322 
323 	/* Is it a Linux binary ? */
324 	if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
325 		error = ENOEXEC;
326 		goto cleanup;
327 	}
328 
329 	/*
330 	 * While we are here, we should REALLY do some more checks
331 	 */
332 
333 	/* Set file/virtual offset based on a.out variant. */
334 	switch ((int)(a_out->a_magic & 0xffff)) {
335 	case 0413:			/* ZMAGIC */
336 		file_offset = 1024;
337 		break;
338 	case 0314:			/* QMAGIC */
339 		file_offset = 0;
340 		break;
341 	default:
342 		error = ENOEXEC;
343 		goto cleanup;
344 	}
345 
346 	bss_size = round_page(a_out->a_bss);
347 
348 	/* Check various fields in header for validity/bounds. */
349 	if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
350 		error = ENOEXEC;
351 		goto cleanup;
352 	}
353 
354 	/* text + data can't exceed file size */
355 	if (a_out->a_data + a_out->a_text > attr.va_size) {
356 		error = EFAULT;
357 		goto cleanup;
358 	}
359 
360 	/*
361 	 * text/data/bss must not exceed limits
362 	 * XXX - this is not complete. it should check current usage PLUS
363 	 * the resources needed by this library.
364 	 */
365 	PROC_LOCK(td->td_proc);
366 	if (a_out->a_text > maxtsiz ||
367 	    a_out->a_data + bss_size > lim_cur(td->td_proc, RLIMIT_DATA)) {
368 		PROC_UNLOCK(td->td_proc);
369 		error = ENOMEM;
370 		goto cleanup;
371 	}
372 	PROC_UNLOCK(td->td_proc);
373 
374 	/*
375 	 * Prevent more writers.
376 	 * XXX: Note that if any of the VM operations fail below we don't
377 	 * clear this flag.
378 	 */
379 	vp->v_vflag |= VV_TEXT;
380 
381 	/*
382 	 * Lock no longer needed
383 	 */
384 	locked = 0;
385 	VOP_UNLOCK(vp, 0);
386 	VFS_UNLOCK_GIANT(vfslocked);
387 
388 	/*
389 	 * Check if file_offset page aligned. Currently we cannot handle
390 	 * misalinged file offsets, and so we read in the entire image
391 	 * (what a waste).
392 	 */
393 	if (file_offset & PAGE_MASK) {
394 #ifdef DEBUG
395 		printf("uselib: Non page aligned binary %lu\n", file_offset);
396 #endif
397 		/* Map text+data read/write/execute */
398 
399 		/* a_entry is the load address and is page aligned */
400 		vmaddr = trunc_page(a_out->a_entry);
401 
402 		/* get anon user mapping, read+write+execute */
403 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
404 		    &vmaddr, a_out->a_text + a_out->a_data, FALSE, VM_PROT_ALL,
405 		    VM_PROT_ALL, 0);
406 		if (error)
407 			goto cleanup;
408 
409 		/* map file into kernel_map */
410 		error = vm_mmap(kernel_map, &buffer,
411 		    round_page(a_out->a_text + a_out->a_data + file_offset),
412 		    VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp,
413 		    trunc_page(file_offset));
414 		if (error)
415 			goto cleanup;
416 
417 		/* copy from kernel VM space to user space */
418 		error = copyout(PTRIN(buffer + file_offset),
419 		    (void *)vmaddr, a_out->a_text + a_out->a_data);
420 
421 		/* release temporary kernel space */
422 		vm_map_remove(kernel_map, buffer, buffer +
423 		    round_page(a_out->a_text + a_out->a_data + file_offset));
424 
425 		if (error)
426 			goto cleanup;
427 	} else {
428 #ifdef DEBUG
429 		printf("uselib: Page aligned binary %lu\n", file_offset);
430 #endif
431 		/*
432 		 * for QMAGIC, a_entry is 20 bytes beyond the load address
433 		 * to skip the executable header
434 		 */
435 		vmaddr = trunc_page(a_out->a_entry);
436 
437 		/*
438 		 * Map it all into the process's space as a single
439 		 * copy-on-write "data" segment.
440 		 */
441 		error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr,
442 		    a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
443 		    MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
444 		if (error)
445 			goto cleanup;
446 	}
447 #ifdef DEBUG
448 	printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
449 	    ((long *)vmaddr)[1]);
450 #endif
451 	if (bss_size != 0) {
452 		/* Calculate BSS start address */
453 		vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
454 		    a_out->a_data;
455 
456 		/* allocate some 'anon' space */
457 		error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
458 		    &vmaddr, bss_size, FALSE, VM_PROT_ALL, VM_PROT_ALL, 0);
459 		if (error)
460 			goto cleanup;
461 	}
462 
463 cleanup:
464 	/* Unlock vnode if needed */
465 	if (locked) {
466 		VOP_UNLOCK(vp, 0);
467 		VFS_UNLOCK_GIANT(vfslocked);
468 	}
469 
470 	/* Release the kernel mapping. */
471 	if (a_out)
472 		vm_map_remove(kernel_map, (vm_offset_t)a_out,
473 		    (vm_offset_t)a_out + PAGE_SIZE);
474 
475 	return error;
476 }
477 
478 #endif	/* __i386__ */
479 
480 int
481 linux_select(struct thread *td, struct linux_select_args *args)
482 {
483 	l_timeval ltv;
484 	struct timeval tv0, tv1, utv, *tvp;
485 	int error;
486 
487 #ifdef DEBUG
488 	if (ldebug(select))
489 		printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
490 		    (void *)args->readfds, (void *)args->writefds,
491 		    (void *)args->exceptfds, (void *)args->timeout);
492 #endif
493 
494 	/*
495 	 * Store current time for computation of the amount of
496 	 * time left.
497 	 */
498 	if (args->timeout) {
499 		if ((error = copyin(args->timeout, &ltv, sizeof(ltv))))
500 			goto select_out;
501 		utv.tv_sec = ltv.tv_sec;
502 		utv.tv_usec = ltv.tv_usec;
503 #ifdef DEBUG
504 		if (ldebug(select))
505 			printf(LMSG("incoming timeout (%jd/%ld)"),
506 			    (intmax_t)utv.tv_sec, utv.tv_usec);
507 #endif
508 
509 		if (itimerfix(&utv)) {
510 			/*
511 			 * The timeval was invalid.  Convert it to something
512 			 * valid that will act as it does under Linux.
513 			 */
514 			utv.tv_sec += utv.tv_usec / 1000000;
515 			utv.tv_usec %= 1000000;
516 			if (utv.tv_usec < 0) {
517 				utv.tv_sec -= 1;
518 				utv.tv_usec += 1000000;
519 			}
520 			if (utv.tv_sec < 0)
521 				timevalclear(&utv);
522 		}
523 		microtime(&tv0);
524 		tvp = &utv;
525 	} else
526 		tvp = NULL;
527 
528 	error = kern_select(td, args->nfds, args->readfds, args->writefds,
529 	    args->exceptfds, tvp);
530 
531 #ifdef DEBUG
532 	if (ldebug(select))
533 		printf(LMSG("real select returns %d"), error);
534 #endif
535 	if (error)
536 		goto select_out;
537 
538 	if (args->timeout) {
539 		if (td->td_retval[0]) {
540 			/*
541 			 * Compute how much time was left of the timeout,
542 			 * by subtracting the current time and the time
543 			 * before we started the call, and subtracting
544 			 * that result from the user-supplied value.
545 			 */
546 			microtime(&tv1);
547 			timevalsub(&tv1, &tv0);
548 			timevalsub(&utv, &tv1);
549 			if (utv.tv_sec < 0)
550 				timevalclear(&utv);
551 		} else
552 			timevalclear(&utv);
553 #ifdef DEBUG
554 		if (ldebug(select))
555 			printf(LMSG("outgoing timeout (%jd/%ld)"),
556 			    (intmax_t)utv.tv_sec, utv.tv_usec);
557 #endif
558 		ltv.tv_sec = utv.tv_sec;
559 		ltv.tv_usec = utv.tv_usec;
560 		if ((error = copyout(&ltv, args->timeout, sizeof(ltv))))
561 			goto select_out;
562 	}
563 
564 select_out:
565 #ifdef DEBUG
566 	if (ldebug(select))
567 		printf(LMSG("select_out -> %d"), error);
568 #endif
569 	return error;
570 }
571 
572 int
573 linux_mremap(struct thread *td, struct linux_mremap_args *args)
574 {
575 	struct munmap_args /* {
576 		void *addr;
577 		size_t len;
578 	} */ bsd_args;
579 	int error = 0;
580 
581 #ifdef DEBUG
582 	if (ldebug(mremap))
583 		printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
584 		    (void *)(uintptr_t)args->addr,
585 		    (unsigned long)args->old_len,
586 		    (unsigned long)args->new_len,
587 		    (unsigned long)args->flags);
588 #endif
589 
590 	if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
591 		td->td_retval[0] = 0;
592 		return (EINVAL);
593 	}
594 
595 	/*
596 	 * Check for the page alignment.
597 	 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
598 	 */
599 	if (args->addr & PAGE_MASK) {
600 		td->td_retval[0] = 0;
601 		return (EINVAL);
602 	}
603 
604 	args->new_len = round_page(args->new_len);
605 	args->old_len = round_page(args->old_len);
606 
607 	if (args->new_len > args->old_len) {
608 		td->td_retval[0] = 0;
609 		return ENOMEM;
610 	}
611 
612 	if (args->new_len < args->old_len) {
613 		bsd_args.addr =
614 		    (caddr_t)((uintptr_t)args->addr + args->new_len);
615 		bsd_args.len = args->old_len - args->new_len;
616 		error = munmap(td, &bsd_args);
617 	}
618 
619 	td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
620 	return error;
621 }
622 
623 #define LINUX_MS_ASYNC       0x0001
624 #define LINUX_MS_INVALIDATE  0x0002
625 #define LINUX_MS_SYNC        0x0004
626 
627 int
628 linux_msync(struct thread *td, struct linux_msync_args *args)
629 {
630 	struct msync_args bsd_args;
631 
632 	bsd_args.addr = (caddr_t)(uintptr_t)args->addr;
633 	bsd_args.len = (uintptr_t)args->len;
634 	bsd_args.flags = args->fl & ~LINUX_MS_SYNC;
635 
636 	return msync(td, &bsd_args);
637 }
638 
639 int
640 linux_time(struct thread *td, struct linux_time_args *args)
641 {
642 	struct timeval tv;
643 	l_time_t tm;
644 	int error;
645 
646 #ifdef DEBUG
647 	if (ldebug(time))
648 		printf(ARGS(time, "*"));
649 #endif
650 
651 	microtime(&tv);
652 	tm = tv.tv_sec;
653 	if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
654 		return error;
655 	td->td_retval[0] = tm;
656 	return 0;
657 }
658 
659 struct l_times_argv {
660 	l_long	tms_utime;
661 	l_long	tms_stime;
662 	l_long	tms_cutime;
663 	l_long	tms_cstime;
664 };
665 
666 #define CLK_TCK 100			/* Linux uses 100 */
667 
668 #define CONVTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
669 
670 int
671 linux_times(struct thread *td, struct linux_times_args *args)
672 {
673 	struct timeval tv, utime, stime, cutime, cstime;
674 	struct l_times_argv tms;
675 	struct proc *p;
676 	int error;
677 
678 #ifdef DEBUG
679 	if (ldebug(times))
680 		printf(ARGS(times, "*"));
681 #endif
682 
683 	if (args->buf != NULL) {
684 		p = td->td_proc;
685 		PROC_LOCK(p);
686 		PROC_SLOCK(p);
687 		calcru(p, &utime, &stime);
688 		PROC_SUNLOCK(p);
689 		calccru(p, &cutime, &cstime);
690 		PROC_UNLOCK(p);
691 
692 		tms.tms_utime = CONVTCK(utime);
693 		tms.tms_stime = CONVTCK(stime);
694 
695 		tms.tms_cutime = CONVTCK(cutime);
696 		tms.tms_cstime = CONVTCK(cstime);
697 
698 		if ((error = copyout(&tms, args->buf, sizeof(tms))))
699 			return error;
700 	}
701 
702 	microuptime(&tv);
703 	td->td_retval[0] = (int)CONVTCK(tv);
704 	return 0;
705 }
706 
707 int
708 linux_newuname(struct thread *td, struct linux_newuname_args *args)
709 {
710 	struct l_new_utsname utsname;
711 	char osname[LINUX_MAX_UTSNAME];
712 	char osrelease[LINUX_MAX_UTSNAME];
713 	char *p;
714 
715 #ifdef DEBUG
716 	if (ldebug(newuname))
717 		printf(ARGS(newuname, "*"));
718 #endif
719 
720 	linux_get_osname(td, osname);
721 	linux_get_osrelease(td, osrelease);
722 
723 	bzero(&utsname, sizeof(utsname));
724 	strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
725 	getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
726 	strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
727 	strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
728 	for (p = utsname.version; *p != '\0'; ++p)
729 		if (*p == '\n') {
730 			*p = '\0';
731 			break;
732 		}
733 #ifdef __i386__
734 	{
735 		const char *class;
736 
737 		switch (cpu_class) {
738 		case CPUCLASS_686:
739 			class = "i686";
740 			break;
741 		case CPUCLASS_586:
742 			class = "i586";
743 			break;
744 		case CPUCLASS_486:
745 			class = "i486";
746 			break;
747 		default:
748 			class = "i386";
749 		}
750 		strlcpy(utsname.machine, class, LINUX_MAX_UTSNAME);
751 	}
752 #elif defined(__amd64__)	/* XXX: Linux can change 'personality'. */
753 #ifdef COMPAT_LINUX32
754 	strlcpy(utsname.machine, "i686", LINUX_MAX_UTSNAME);
755 #else
756 	strlcpy(utsname.machine, "x86_64", LINUX_MAX_UTSNAME);
757 #endif /* COMPAT_LINUX32 */
758 #else /* something other than i386 or amd64 - assume we and Linux agree */
759 	strlcpy(utsname.machine, machine, LINUX_MAX_UTSNAME);
760 #endif /* __i386__ */
761 	mtx_lock(&hostname_mtx);
762 	strlcpy(utsname.domainname, V_domainname, LINUX_MAX_UTSNAME);
763 	mtx_unlock(&hostname_mtx);
764 
765 	return (copyout(&utsname, args->buf, sizeof(utsname)));
766 }
767 
768 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
769 struct l_utimbuf {
770 	l_time_t l_actime;
771 	l_time_t l_modtime;
772 };
773 
774 int
775 linux_utime(struct thread *td, struct linux_utime_args *args)
776 {
777 	struct timeval tv[2], *tvp;
778 	struct l_utimbuf lut;
779 	char *fname;
780 	int error;
781 
782 	LCONVPATHEXIST(td, args->fname, &fname);
783 
784 #ifdef DEBUG
785 	if (ldebug(utime))
786 		printf(ARGS(utime, "%s, *"), fname);
787 #endif
788 
789 	if (args->times) {
790 		if ((error = copyin(args->times, &lut, sizeof lut))) {
791 			LFREEPATH(fname);
792 			return error;
793 		}
794 		tv[0].tv_sec = lut.l_actime;
795 		tv[0].tv_usec = 0;
796 		tv[1].tv_sec = lut.l_modtime;
797 		tv[1].tv_usec = 0;
798 		tvp = tv;
799 	} else
800 		tvp = NULL;
801 
802 	error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
803 	LFREEPATH(fname);
804 	return (error);
805 }
806 
807 int
808 linux_utimes(struct thread *td, struct linux_utimes_args *args)
809 {
810 	l_timeval ltv[2];
811 	struct timeval tv[2], *tvp = NULL;
812 	char *fname;
813 	int error;
814 
815 	LCONVPATHEXIST(td, args->fname, &fname);
816 
817 #ifdef DEBUG
818 	if (ldebug(utimes))
819 		printf(ARGS(utimes, "%s, *"), fname);
820 #endif
821 
822 	if (args->tptr != NULL) {
823 		if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
824 			LFREEPATH(fname);
825 			return (error);
826 		}
827 		tv[0].tv_sec = ltv[0].tv_sec;
828 		tv[0].tv_usec = ltv[0].tv_usec;
829 		tv[1].tv_sec = ltv[1].tv_sec;
830 		tv[1].tv_usec = ltv[1].tv_usec;
831 		tvp = tv;
832 	}
833 
834 	error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
835 	LFREEPATH(fname);
836 	return (error);
837 }
838 
839 int
840 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
841 {
842 	l_timeval ltv[2];
843 	struct timeval tv[2], *tvp = NULL;
844 	char *fname;
845 	int error, dfd;
846 
847 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
848 	LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
849 
850 #ifdef DEBUG
851 	if (ldebug(futimesat))
852 		printf(ARGS(futimesat, "%s, *"), fname);
853 #endif
854 
855 	if (args->utimes != NULL) {
856 		if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
857 			LFREEPATH(fname);
858 			return (error);
859 		}
860 		tv[0].tv_sec = ltv[0].tv_sec;
861 		tv[0].tv_usec = ltv[0].tv_usec;
862 		tv[1].tv_sec = ltv[1].tv_sec;
863 		tv[1].tv_usec = ltv[1].tv_usec;
864 		tvp = tv;
865 	}
866 
867 	error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
868 	LFREEPATH(fname);
869 	return (error);
870 }
871 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
872 
873 #define __WCLONE 0x80000000
874 
875 int
876 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
877 {
878 	int error, options, tmpstat;
879 
880 #ifdef DEBUG
881 	if (ldebug(waitpid))
882 		printf(ARGS(waitpid, "%d, %p, %d"),
883 		    args->pid, (void *)args->status, args->options);
884 #endif
885 	/*
886 	 * this is necessary because the test in kern_wait doesn't work
887 	 * because we mess with the options here
888 	 */
889 	if (args->options & ~(WUNTRACED | WNOHANG | WCONTINUED | __WCLONE))
890 		return (EINVAL);
891 
892 	options = (args->options & (WNOHANG | WUNTRACED));
893 	/* WLINUXCLONE should be equal to __WCLONE, but we make sure */
894 	if (args->options & __WCLONE)
895 		options |= WLINUXCLONE;
896 
897 	error = kern_wait(td, args->pid, &tmpstat, options, NULL);
898 	if (error)
899 		return error;
900 
901 	if (args->status) {
902 		tmpstat &= 0xffff;
903 		if (WIFSIGNALED(tmpstat))
904 			tmpstat = (tmpstat & 0xffffff80) |
905 			    BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat));
906 		else if (WIFSTOPPED(tmpstat))
907 			tmpstat = (tmpstat & 0xffff00ff) |
908 			    (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8);
909 		return copyout(&tmpstat, args->status, sizeof(int));
910 	}
911 
912 	return 0;
913 }
914 
915 int
916 linux_wait4(struct thread *td, struct linux_wait4_args *args)
917 {
918 	int error, options, tmpstat;
919 	struct rusage ru, *rup;
920 	struct proc *p;
921 
922 #ifdef DEBUG
923 	if (ldebug(wait4))
924 		printf(ARGS(wait4, "%d, %p, %d, %p"),
925 		    args->pid, (void *)args->status, args->options,
926 		    (void *)args->rusage);
927 #endif
928 
929 	options = (args->options & (WNOHANG | WUNTRACED));
930 	/* WLINUXCLONE should be equal to __WCLONE, but we make sure */
931 	if (args->options & __WCLONE)
932 		options |= WLINUXCLONE;
933 
934 	if (args->rusage != NULL)
935 		rup = &ru;
936 	else
937 		rup = NULL;
938 	error = kern_wait(td, args->pid, &tmpstat, options, rup);
939 	if (error)
940 		return error;
941 
942 	p = td->td_proc;
943 	PROC_LOCK(p);
944 	sigqueue_delete(&p->p_sigqueue, SIGCHLD);
945 	PROC_UNLOCK(p);
946 
947 	if (args->status) {
948 		tmpstat &= 0xffff;
949 		if (WIFSIGNALED(tmpstat))
950 			tmpstat = (tmpstat & 0xffffff80) |
951 			    BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat));
952 		else if (WIFSTOPPED(tmpstat))
953 			tmpstat = (tmpstat & 0xffff00ff) |
954 			    (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8);
955 		error = copyout(&tmpstat, args->status, sizeof(int));
956 	}
957 	if (args->rusage != NULL && error == 0)
958 		error = copyout(&ru, args->rusage, sizeof(ru));
959 
960 	return (error);
961 }
962 
963 int
964 linux_mknod(struct thread *td, struct linux_mknod_args *args)
965 {
966 	char *path;
967 	int error;
968 
969 	LCONVPATHCREAT(td, args->path, &path);
970 
971 #ifdef DEBUG
972 	if (ldebug(mknod))
973 		printf(ARGS(mknod, "%s, %d, %d"), path, args->mode, args->dev);
974 #endif
975 
976 	switch (args->mode & S_IFMT) {
977 	case S_IFIFO:
978 	case S_IFSOCK:
979 		error = kern_mkfifo(td, path, UIO_SYSSPACE, args->mode);
980 		break;
981 
982 	case S_IFCHR:
983 	case S_IFBLK:
984 		error = kern_mknod(td, path, UIO_SYSSPACE, args->mode,
985 		    args->dev);
986 		break;
987 
988 	case S_IFDIR:
989 		error = EPERM;
990 		break;
991 
992 	case 0:
993 		args->mode |= S_IFREG;
994 		/* FALLTHROUGH */
995 	case S_IFREG:
996 		error = kern_open(td, path, UIO_SYSSPACE,
997 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
998 		if (error == 0)
999 			kern_close(td, td->td_retval[0]);
1000 		break;
1001 
1002 	default:
1003 		error = EINVAL;
1004 		break;
1005 	}
1006 	LFREEPATH(path);
1007 	return (error);
1008 }
1009 
1010 int
1011 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
1012 {
1013 	char *path;
1014 	int error, dfd;
1015 
1016 	dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
1017 	LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
1018 
1019 #ifdef DEBUG
1020 	if (ldebug(mknodat))
1021 		printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev);
1022 #endif
1023 
1024 	switch (args->mode & S_IFMT) {
1025 	case S_IFIFO:
1026 	case S_IFSOCK:
1027 		error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
1028 		break;
1029 
1030 	case S_IFCHR:
1031 	case S_IFBLK:
1032 		error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
1033 		    args->dev);
1034 		break;
1035 
1036 	case S_IFDIR:
1037 		error = EPERM;
1038 		break;
1039 
1040 	case 0:
1041 		args->mode |= S_IFREG;
1042 		/* FALLTHROUGH */
1043 	case S_IFREG:
1044 		error = kern_openat(td, dfd, path, UIO_SYSSPACE,
1045 		    O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1046 		if (error == 0)
1047 			kern_close(td, td->td_retval[0]);
1048 		break;
1049 
1050 	default:
1051 		error = EINVAL;
1052 		break;
1053 	}
1054 	LFREEPATH(path);
1055 	return (error);
1056 }
1057 
1058 /*
1059  * UGH! This is just about the dumbest idea I've ever heard!!
1060  */
1061 int
1062 linux_personality(struct thread *td, struct linux_personality_args *args)
1063 {
1064 #ifdef DEBUG
1065 	if (ldebug(personality))
1066 		printf(ARGS(personality, "%lu"), (unsigned long)args->per);
1067 #endif
1068 	if (args->per != 0)
1069 		return EINVAL;
1070 
1071 	/* Yes Jim, it's still a Linux... */
1072 	td->td_retval[0] = 0;
1073 	return 0;
1074 }
1075 
1076 struct l_itimerval {
1077 	l_timeval it_interval;
1078 	l_timeval it_value;
1079 };
1080 
1081 #define	B2L_ITIMERVAL(bip, lip) 					\
1082 	(bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec;		\
1083 	(bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec;	\
1084 	(bip)->it_value.tv_sec = (lip)->it_value.tv_sec;		\
1085 	(bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1086 
1087 int
1088 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1089 {
1090 	int error;
1091 	struct l_itimerval ls;
1092 	struct itimerval aitv, oitv;
1093 
1094 #ifdef DEBUG
1095 	if (ldebug(setitimer))
1096 		printf(ARGS(setitimer, "%p, %p"),
1097 		    (void *)uap->itv, (void *)uap->oitv);
1098 #endif
1099 
1100 	if (uap->itv == NULL) {
1101 		uap->itv = uap->oitv;
1102 		return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1103 	}
1104 
1105 	error = copyin(uap->itv, &ls, sizeof(ls));
1106 	if (error != 0)
1107 		return (error);
1108 	B2L_ITIMERVAL(&aitv, &ls);
1109 #ifdef DEBUG
1110 	if (ldebug(setitimer)) {
1111 		printf("setitimer: value: sec: %jd, usec: %ld\n",
1112 		    (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1113 		printf("setitimer: interval: sec: %jd, usec: %ld\n",
1114 		    (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1115 	}
1116 #endif
1117 	error = kern_setitimer(td, uap->which, &aitv, &oitv);
1118 	if (error != 0 || uap->oitv == NULL)
1119 		return (error);
1120 	B2L_ITIMERVAL(&ls, &oitv);
1121 
1122 	return (copyout(&ls, uap->oitv, sizeof(ls)));
1123 }
1124 
1125 int
1126 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1127 {
1128 	int error;
1129 	struct l_itimerval ls;
1130 	struct itimerval aitv;
1131 
1132 #ifdef DEBUG
1133 	if (ldebug(getitimer))
1134 		printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1135 #endif
1136 	error = kern_getitimer(td, uap->which, &aitv);
1137 	if (error != 0)
1138 		return (error);
1139 	B2L_ITIMERVAL(&ls, &aitv);
1140 	return (copyout(&ls, uap->itv, sizeof(ls)));
1141 }
1142 
1143 int
1144 linux_nice(struct thread *td, struct linux_nice_args *args)
1145 {
1146 	struct setpriority_args bsd_args;
1147 
1148 	bsd_args.which = PRIO_PROCESS;
1149 	bsd_args.who = 0;		/* current process */
1150 	bsd_args.prio = args->inc;
1151 	return setpriority(td, &bsd_args);
1152 }
1153 
1154 int
1155 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1156 {
1157 	struct ucred *newcred, *oldcred;
1158 	l_gid_t linux_gidset[NGROUPS];
1159 	gid_t *bsd_gidset;
1160 	int ngrp, error;
1161 	struct proc *p;
1162 
1163 	ngrp = args->gidsetsize;
1164 	if (ngrp < 0 || ngrp >= NGROUPS)
1165 		return (EINVAL);
1166 	error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1167 	if (error)
1168 		return (error);
1169 	newcred = crget();
1170 	p = td->td_proc;
1171 	PROC_LOCK(p);
1172 	oldcred = p->p_ucred;
1173 
1174 	/*
1175 	 * cr_groups[0] holds egid. Setting the whole set from
1176 	 * the supplied set will cause egid to be changed too.
1177 	 * Keep cr_groups[0] unchanged to prevent that.
1178 	 */
1179 
1180 	if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) {
1181 		PROC_UNLOCK(p);
1182 		crfree(newcred);
1183 		return (error);
1184 	}
1185 
1186 	crcopy(newcred, oldcred);
1187 	if (ngrp > 0) {
1188 		newcred->cr_ngroups = ngrp + 1;
1189 
1190 		bsd_gidset = newcred->cr_groups;
1191 		ngrp--;
1192 		while (ngrp >= 0) {
1193 			bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1194 			ngrp--;
1195 		}
1196 	} else
1197 		newcred->cr_ngroups = 1;
1198 
1199 	setsugid(p);
1200 	p->p_ucred = newcred;
1201 	PROC_UNLOCK(p);
1202 	crfree(oldcred);
1203 	return (0);
1204 }
1205 
1206 int
1207 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1208 {
1209 	struct ucred *cred;
1210 	l_gid_t linux_gidset[NGROUPS];
1211 	gid_t *bsd_gidset;
1212 	int bsd_gidsetsz, ngrp, error;
1213 
1214 	cred = td->td_ucred;
1215 	bsd_gidset = cred->cr_groups;
1216 	bsd_gidsetsz = cred->cr_ngroups - 1;
1217 
1218 	/*
1219 	 * cr_groups[0] holds egid. Returning the whole set
1220 	 * here will cause a duplicate. Exclude cr_groups[0]
1221 	 * to prevent that.
1222 	 */
1223 
1224 	if ((ngrp = args->gidsetsize) == 0) {
1225 		td->td_retval[0] = bsd_gidsetsz;
1226 		return (0);
1227 	}
1228 
1229 	if (ngrp < bsd_gidsetsz)
1230 		return (EINVAL);
1231 
1232 	ngrp = 0;
1233 	while (ngrp < bsd_gidsetsz) {
1234 		linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1235 		ngrp++;
1236 	}
1237 
1238 	if ((error = copyout(linux_gidset, args->grouplist,
1239 	    ngrp * sizeof(l_gid_t))))
1240 		return (error);
1241 
1242 	td->td_retval[0] = ngrp;
1243 	return (0);
1244 }
1245 
1246 int
1247 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1248 {
1249 	struct rlimit bsd_rlim;
1250 	struct l_rlimit rlim;
1251 	u_int which;
1252 	int error;
1253 
1254 #ifdef DEBUG
1255 	if (ldebug(setrlimit))
1256 		printf(ARGS(setrlimit, "%d, %p"),
1257 		    args->resource, (void *)args->rlim);
1258 #endif
1259 
1260 	if (args->resource >= LINUX_RLIM_NLIMITS)
1261 		return (EINVAL);
1262 
1263 	which = linux_to_bsd_resource[args->resource];
1264 	if (which == -1)
1265 		return (EINVAL);
1266 
1267 	error = copyin(args->rlim, &rlim, sizeof(rlim));
1268 	if (error)
1269 		return (error);
1270 
1271 	bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1272 	bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1273 	return (kern_setrlimit(td, which, &bsd_rlim));
1274 }
1275 
1276 int
1277 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1278 {
1279 	struct l_rlimit rlim;
1280 	struct proc *p = td->td_proc;
1281 	struct rlimit bsd_rlim;
1282 	u_int which;
1283 
1284 #ifdef DEBUG
1285 	if (ldebug(old_getrlimit))
1286 		printf(ARGS(old_getrlimit, "%d, %p"),
1287 		    args->resource, (void *)args->rlim);
1288 #endif
1289 
1290 	if (args->resource >= LINUX_RLIM_NLIMITS)
1291 		return (EINVAL);
1292 
1293 	which = linux_to_bsd_resource[args->resource];
1294 	if (which == -1)
1295 		return (EINVAL);
1296 
1297 	PROC_LOCK(p);
1298 	lim_rlimit(p, which, &bsd_rlim);
1299 	PROC_UNLOCK(p);
1300 
1301 #ifdef COMPAT_LINUX32
1302 	rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1303 	if (rlim.rlim_cur == UINT_MAX)
1304 		rlim.rlim_cur = INT_MAX;
1305 	rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1306 	if (rlim.rlim_max == UINT_MAX)
1307 		rlim.rlim_max = INT_MAX;
1308 #else
1309 	rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1310 	if (rlim.rlim_cur == ULONG_MAX)
1311 		rlim.rlim_cur = LONG_MAX;
1312 	rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1313 	if (rlim.rlim_max == ULONG_MAX)
1314 		rlim.rlim_max = LONG_MAX;
1315 #endif
1316 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1317 }
1318 
1319 int
1320 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1321 {
1322 	struct l_rlimit rlim;
1323 	struct proc *p = td->td_proc;
1324 	struct rlimit bsd_rlim;
1325 	u_int which;
1326 
1327 #ifdef DEBUG
1328 	if (ldebug(getrlimit))
1329 		printf(ARGS(getrlimit, "%d, %p"),
1330 		    args->resource, (void *)args->rlim);
1331 #endif
1332 
1333 	if (args->resource >= LINUX_RLIM_NLIMITS)
1334 		return (EINVAL);
1335 
1336 	which = linux_to_bsd_resource[args->resource];
1337 	if (which == -1)
1338 		return (EINVAL);
1339 
1340 	PROC_LOCK(p);
1341 	lim_rlimit(p, which, &bsd_rlim);
1342 	PROC_UNLOCK(p);
1343 
1344 	rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1345 	rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1346 	return (copyout(&rlim, args->rlim, sizeof(rlim)));
1347 }
1348 
1349 int
1350 linux_sched_setscheduler(struct thread *td,
1351     struct linux_sched_setscheduler_args *args)
1352 {
1353 	struct sched_setscheduler_args bsd;
1354 
1355 #ifdef DEBUG
1356 	if (ldebug(sched_setscheduler))
1357 		printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1358 		    args->pid, args->policy, (const void *)args->param);
1359 #endif
1360 
1361 	switch (args->policy) {
1362 	case LINUX_SCHED_OTHER:
1363 		bsd.policy = SCHED_OTHER;
1364 		break;
1365 	case LINUX_SCHED_FIFO:
1366 		bsd.policy = SCHED_FIFO;
1367 		break;
1368 	case LINUX_SCHED_RR:
1369 		bsd.policy = SCHED_RR;
1370 		break;
1371 	default:
1372 		return EINVAL;
1373 	}
1374 
1375 	bsd.pid = args->pid;
1376 	bsd.param = (struct sched_param *)args->param;
1377 	return sched_setscheduler(td, &bsd);
1378 }
1379 
1380 int
1381 linux_sched_getscheduler(struct thread *td,
1382     struct linux_sched_getscheduler_args *args)
1383 {
1384 	struct sched_getscheduler_args bsd;
1385 	int error;
1386 
1387 #ifdef DEBUG
1388 	if (ldebug(sched_getscheduler))
1389 		printf(ARGS(sched_getscheduler, "%d"), args->pid);
1390 #endif
1391 
1392 	bsd.pid = args->pid;
1393 	error = sched_getscheduler(td, &bsd);
1394 
1395 	switch (td->td_retval[0]) {
1396 	case SCHED_OTHER:
1397 		td->td_retval[0] = LINUX_SCHED_OTHER;
1398 		break;
1399 	case SCHED_FIFO:
1400 		td->td_retval[0] = LINUX_SCHED_FIFO;
1401 		break;
1402 	case SCHED_RR:
1403 		td->td_retval[0] = LINUX_SCHED_RR;
1404 		break;
1405 	}
1406 
1407 	return error;
1408 }
1409 
1410 int
1411 linux_sched_get_priority_max(struct thread *td,
1412     struct linux_sched_get_priority_max_args *args)
1413 {
1414 	struct sched_get_priority_max_args bsd;
1415 
1416 #ifdef DEBUG
1417 	if (ldebug(sched_get_priority_max))
1418 		printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1419 #endif
1420 
1421 	switch (args->policy) {
1422 	case LINUX_SCHED_OTHER:
1423 		bsd.policy = SCHED_OTHER;
1424 		break;
1425 	case LINUX_SCHED_FIFO:
1426 		bsd.policy = SCHED_FIFO;
1427 		break;
1428 	case LINUX_SCHED_RR:
1429 		bsd.policy = SCHED_RR;
1430 		break;
1431 	default:
1432 		return EINVAL;
1433 	}
1434 	return sched_get_priority_max(td, &bsd);
1435 }
1436 
1437 int
1438 linux_sched_get_priority_min(struct thread *td,
1439     struct linux_sched_get_priority_min_args *args)
1440 {
1441 	struct sched_get_priority_min_args bsd;
1442 
1443 #ifdef DEBUG
1444 	if (ldebug(sched_get_priority_min))
1445 		printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1446 #endif
1447 
1448 	switch (args->policy) {
1449 	case LINUX_SCHED_OTHER:
1450 		bsd.policy = SCHED_OTHER;
1451 		break;
1452 	case LINUX_SCHED_FIFO:
1453 		bsd.policy = SCHED_FIFO;
1454 		break;
1455 	case LINUX_SCHED_RR:
1456 		bsd.policy = SCHED_RR;
1457 		break;
1458 	default:
1459 		return EINVAL;
1460 	}
1461 	return sched_get_priority_min(td, &bsd);
1462 }
1463 
1464 #define REBOOT_CAD_ON	0x89abcdef
1465 #define REBOOT_CAD_OFF	0
1466 #define REBOOT_HALT	0xcdef0123
1467 #define REBOOT_RESTART	0x01234567
1468 #define REBOOT_RESTART2	0xA1B2C3D4
1469 #define REBOOT_POWEROFF	0x4321FEDC
1470 #define REBOOT_MAGIC1	0xfee1dead
1471 #define REBOOT_MAGIC2	0x28121969
1472 #define REBOOT_MAGIC2A	0x05121996
1473 #define REBOOT_MAGIC2B	0x16041998
1474 
1475 int
1476 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1477 {
1478 	struct reboot_args bsd_args;
1479 
1480 #ifdef DEBUG
1481 	if (ldebug(reboot))
1482 		printf(ARGS(reboot, "0x%x"), args->cmd);
1483 #endif
1484 
1485 	if (args->magic1 != REBOOT_MAGIC1)
1486 		return EINVAL;
1487 
1488 	switch (args->magic2) {
1489 	case REBOOT_MAGIC2:
1490 	case REBOOT_MAGIC2A:
1491 	case REBOOT_MAGIC2B:
1492 		break;
1493 	default:
1494 		return EINVAL;
1495 	}
1496 
1497 	switch (args->cmd) {
1498 	case REBOOT_CAD_ON:
1499 	case REBOOT_CAD_OFF:
1500 		return (priv_check(td, PRIV_REBOOT));
1501 	case REBOOT_HALT:
1502 		bsd_args.opt = RB_HALT;
1503 		break;
1504 	case REBOOT_RESTART:
1505 	case REBOOT_RESTART2:
1506 		bsd_args.opt = 0;
1507 		break;
1508 	case REBOOT_POWEROFF:
1509 		bsd_args.opt = RB_POWEROFF;
1510 		break;
1511 	default:
1512 		return EINVAL;
1513 	}
1514 	return reboot(td, &bsd_args);
1515 }
1516 
1517 
1518 /*
1519  * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1520  * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
1521  * are assumed to be preserved. The following lightweight syscalls fixes
1522  * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
1523  *
1524  * linux_getpid() - MP SAFE
1525  * linux_getgid() - MP SAFE
1526  * linux_getuid() - MP SAFE
1527  */
1528 
1529 int
1530 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1531 {
1532 	struct linux_emuldata *em;
1533 
1534 #ifdef DEBUG
1535 	if (ldebug(getpid))
1536 		printf(ARGS(getpid, ""));
1537 #endif
1538 
1539 	if (linux_use26(td)) {
1540 		em = em_find(td->td_proc, EMUL_DONTLOCK);
1541 		KASSERT(em != NULL, ("getpid: emuldata not found.\n"));
1542 		td->td_retval[0] = em->shared->group_pid;
1543 	} else {
1544 		td->td_retval[0] = td->td_proc->p_pid;
1545 	}
1546 
1547 	return (0);
1548 }
1549 
1550 int
1551 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1552 {
1553 
1554 #ifdef DEBUG
1555 	if (ldebug(gettid))
1556 		printf(ARGS(gettid, ""));
1557 #endif
1558 
1559 	td->td_retval[0] = td->td_proc->p_pid;
1560 	return (0);
1561 }
1562 
1563 
1564 int
1565 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1566 {
1567 	struct linux_emuldata *em;
1568 	struct proc *p, *pp;
1569 
1570 #ifdef DEBUG
1571 	if (ldebug(getppid))
1572 		printf(ARGS(getppid, ""));
1573 #endif
1574 
1575 	if (!linux_use26(td)) {
1576 		PROC_LOCK(td->td_proc);
1577 		td->td_retval[0] = td->td_proc->p_pptr->p_pid;
1578 		PROC_UNLOCK(td->td_proc);
1579 		return (0);
1580 	}
1581 
1582 	em = em_find(td->td_proc, EMUL_DONTLOCK);
1583 
1584 	KASSERT(em != NULL, ("getppid: process emuldata not found.\n"));
1585 
1586 	/* find the group leader */
1587 	p = pfind(em->shared->group_pid);
1588 
1589 	if (p == NULL) {
1590 #ifdef DEBUG
1591 	   	printf(LMSG("parent process not found.\n"));
1592 #endif
1593 		return (0);
1594 	}
1595 
1596 	pp = p->p_pptr;		/* switch to parent */
1597 	PROC_LOCK(pp);
1598 	PROC_UNLOCK(p);
1599 
1600 	/* if its also linux process */
1601 	if (pp->p_sysent == &elf_linux_sysvec) {
1602 		em = em_find(pp, EMUL_DONTLOCK);
1603 		KASSERT(em != NULL, ("getppid: parent emuldata not found.\n"));
1604 
1605 		td->td_retval[0] = em->shared->group_pid;
1606 	} else
1607 		td->td_retval[0] = pp->p_pid;
1608 
1609 	PROC_UNLOCK(pp);
1610 
1611 	return (0);
1612 }
1613 
1614 int
1615 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1616 {
1617 
1618 #ifdef DEBUG
1619 	if (ldebug(getgid))
1620 		printf(ARGS(getgid, ""));
1621 #endif
1622 
1623 	td->td_retval[0] = td->td_ucred->cr_rgid;
1624 	return (0);
1625 }
1626 
1627 int
1628 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1629 {
1630 
1631 #ifdef DEBUG
1632 	if (ldebug(getuid))
1633 		printf(ARGS(getuid, ""));
1634 #endif
1635 
1636 	td->td_retval[0] = td->td_ucred->cr_ruid;
1637 	return (0);
1638 }
1639 
1640 
1641 int
1642 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1643 {
1644 	struct getsid_args bsd;
1645 
1646 #ifdef DEBUG
1647 	if (ldebug(getsid))
1648 		printf(ARGS(getsid, "%i"), args->pid);
1649 #endif
1650 
1651 	bsd.pid = args->pid;
1652 	return getsid(td, &bsd);
1653 }
1654 
1655 int
1656 linux_nosys(struct thread *td, struct nosys_args *ignore)
1657 {
1658 
1659 	return (ENOSYS);
1660 }
1661 
1662 int
1663 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1664 {
1665 	struct getpriority_args bsd_args;
1666 	int error;
1667 
1668 #ifdef DEBUG
1669 	if (ldebug(getpriority))
1670 		printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
1671 #endif
1672 
1673 	bsd_args.which = args->which;
1674 	bsd_args.who = args->who;
1675 	error = getpriority(td, &bsd_args);
1676 	td->td_retval[0] = 20 - td->td_retval[0];
1677 	return error;
1678 }
1679 
1680 int
1681 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1682 {
1683 	int name[2];
1684 
1685 #ifdef DEBUG
1686 	if (ldebug(sethostname))
1687 		printf(ARGS(sethostname, "*, %i"), args->len);
1688 #endif
1689 
1690 	name[0] = CTL_KERN;
1691 	name[1] = KERN_HOSTNAME;
1692 	return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1693 	    args->len, 0, 0));
1694 }
1695 
1696 int
1697 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1698 {
1699 	struct linux_emuldata *em, *td_em, *tmp_em;
1700 	struct proc *sp;
1701 
1702 #ifdef DEBUG
1703 	if (ldebug(exit_group))
1704 		printf(ARGS(exit_group, "%i"), args->error_code);
1705 #endif
1706 
1707 	if (linux_use26(td)) {
1708 		td_em = em_find(td->td_proc, EMUL_DONTLOCK);
1709 
1710 		KASSERT(td_em != NULL, ("exit_group: emuldata not found.\n"));
1711 
1712 		EMUL_SHARED_RLOCK(&emul_shared_lock);
1713 		LIST_FOREACH_SAFE(em, &td_em->shared->threads, threads, tmp_em) {
1714 			if (em->pid == td_em->pid)
1715 				continue;
1716 
1717 			sp = pfind(em->pid);
1718 			psignal(sp, SIGKILL);
1719 			PROC_UNLOCK(sp);
1720 #ifdef DEBUG
1721 			printf(LMSG("linux_sys_exit_group: kill PID %d\n"), em->pid);
1722 #endif
1723 		}
1724 
1725 		EMUL_SHARED_RUNLOCK(&emul_shared_lock);
1726 	}
1727 	/*
1728 	 * XXX: we should send a signal to the parent if
1729 	 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1730 	 * as it doesnt occur often.
1731 	 */
1732 	exit1(td, W_EXITCODE(args->error_code, 0));
1733 
1734 	return (0);
1735 }
1736 
1737 int
1738 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1739 {
1740 	int error = 0, max_size;
1741 	struct proc *p = td->td_proc;
1742 	char comm[LINUX_MAX_COMM_LEN];
1743 	struct linux_emuldata *em;
1744 	int pdeath_signal;
1745 
1746 #ifdef DEBUG
1747 	if (ldebug(prctl))
1748 		printf(ARGS(prctl, "%d, %d, %d, %d, %d"), args->option,
1749 		    args->arg2, args->arg3, args->arg4, args->arg5);
1750 #endif
1751 
1752 	switch (args->option) {
1753 	case LINUX_PR_SET_PDEATHSIG:
1754 		if (!LINUX_SIG_VALID(args->arg2))
1755 			return (EINVAL);
1756 		em = em_find(p, EMUL_DOLOCK);
1757 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1758 		em->pdeath_signal = args->arg2;
1759 		EMUL_UNLOCK(&emul_lock);
1760 		break;
1761 	case LINUX_PR_GET_PDEATHSIG:
1762 		em = em_find(p, EMUL_DOLOCK);
1763 		KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1764 		pdeath_signal = em->pdeath_signal;
1765 		EMUL_UNLOCK(&emul_lock);
1766 		error = copyout(&pdeath_signal,
1767 		    (void *)(register_t)args->arg2,
1768 		    sizeof(pdeath_signal));
1769 		break;
1770 	case LINUX_PR_SET_NAME:
1771 		/*
1772 		 * To be on the safe side we need to make sure to not
1773 		 * overflow the size a linux program expects. We already
1774 		 * do this here in the copyin, so that we don't need to
1775 		 * check on copyout.
1776 		 */
1777 		max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1778 		error = copyinstr((void *)(register_t)args->arg2, comm,
1779 		    max_size, NULL);
1780 
1781 		/* Linux silently truncates the name if it is too long. */
1782 		if (error == ENAMETOOLONG) {
1783 			/*
1784 			 * XXX: copyinstr() isn't documented to populate the
1785 			 * array completely, so do a copyin() to be on the
1786 			 * safe side. This should be changed in case
1787 			 * copyinstr() is changed to guarantee this.
1788 			 */
1789 			error = copyin((void *)(register_t)args->arg2, comm,
1790 			    max_size - 1);
1791 			comm[max_size - 1] = '\0';
1792 		}
1793 		if (error)
1794 			return (error);
1795 
1796 		PROC_LOCK(p);
1797 		strlcpy(p->p_comm, comm, sizeof(p->p_comm));
1798 		PROC_UNLOCK(p);
1799 		break;
1800 	case LINUX_PR_GET_NAME:
1801 		PROC_LOCK(p);
1802 		strlcpy(comm, p->p_comm, sizeof(comm));
1803 		PROC_UNLOCK(p);
1804 		error = copyout(comm, (void *)(register_t)args->arg2,
1805 		    strlen(comm) + 1);
1806 		break;
1807 	default:
1808 		error = EINVAL;
1809 		break;
1810 	}
1811 
1812 	return (error);
1813 }
1814 
1815 /*
1816  * Get affinity of a process.
1817  */
1818 int
1819 linux_sched_getaffinity(struct thread *td,
1820     struct linux_sched_getaffinity_args *args)
1821 {
1822 	int error;
1823 	struct cpuset_getaffinity_args cga;
1824 
1825 #ifdef DEBUG
1826 	if (ldebug(sched_getaffinity))
1827 		printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid,
1828 		    args->len);
1829 #endif
1830 
1831 	cga.level = CPU_LEVEL_WHICH;
1832 	cga.which = CPU_WHICH_PID;
1833 	cga.id = args->pid;
1834 	cga.cpusetsize = sizeof(cpumask_t);
1835 	cga.mask = (cpuset_t *) args->user_mask_ptr;
1836 
1837 	if ((error = cpuset_getaffinity(td, &cga)) == 0)
1838 		td->td_retval[0] = sizeof(cpumask_t);
1839 
1840 	return (error);
1841 }
1842 
1843 /*
1844  *  Set affinity of a process.
1845  */
1846 int
1847 linux_sched_setaffinity(struct thread *td,
1848     struct linux_sched_setaffinity_args *args)
1849 {
1850 	struct cpuset_setaffinity_args csa;
1851 
1852 #ifdef DEBUG
1853 	if (ldebug(sched_setaffinity))
1854 		printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid,
1855 		    args->len);
1856 #endif
1857 	csa.level = CPU_LEVEL_WHICH;
1858 	csa.which = CPU_WHICH_PID;
1859 	csa.id = args->pid;
1860 	csa.cpusetsize = args->len;
1861 	csa.mask = (cpuset_t *) args->user_mask_ptr;
1862 
1863 	return (cpuset_setaffinity(td, &csa));
1864 }
1865