xref: /freebsd/sys/compat/linux/linux_misc.c (revision 4b2eaea43fec8e8792be611dea204071a10b655a)
1 /*-
2  * Copyright (c) 1994-1995 S�ren Schmidt
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer
10  *    in this position and unchanged.
11  * 2. Redistributions in binary form must reproduce the above copyright
12  *    notice, this list of conditions and the following disclaimer in the
13  *    documentation and/or other materials provided with the distribution.
14  * 3. The name of the author may not be used to endorse or promote products
15  *    derived from this software without specific prior written permission
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
18  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
19  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
20  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
21  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
22  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
26  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
27  *
28  * $FreeBSD$
29  */
30 
31 #include "opt_mac.h"
32 
33 #include <sys/param.h>
34 #include <sys/blist.h>
35 #include <sys/fcntl.h>
36 #include <sys/imgact_aout.h>
37 #include <sys/jail.h>
38 #include <sys/kernel.h>
39 #include <sys/lock.h>
40 #include <sys/mac.h>
41 #include <sys/malloc.h>
42 #include <sys/mman.h>
43 #include <sys/mount.h>
44 #include <sys/mutex.h>
45 #include <sys/namei.h>
46 #include <sys/proc.h>
47 #include <sys/reboot.h>
48 #include <sys/resourcevar.h>
49 #include <sys/signalvar.h>
50 #include <sys/stat.h>
51 #include <sys/syscallsubr.h>
52 #include <sys/sysctl.h>
53 #include <sys/sysproto.h>
54 #include <sys/systm.h>
55 #include <sys/time.h>
56 #include <sys/vmmeter.h>
57 #include <sys/vnode.h>
58 #include <sys/wait.h>
59 
60 #include <vm/vm.h>
61 #include <vm/pmap.h>
62 #include <vm/vm_kern.h>
63 #include <vm/vm_map.h>
64 #include <vm/vm_extern.h>
65 #include <vm/vm_object.h>
66 #include <vm/swap_pager.h>
67 
68 #include <machine/limits.h>
69 
70 #include <posix4/sched.h>
71 
72 #include <machine/../linux/linux.h>
73 #include <machine/../linux/linux_proto.h>
74 
75 #include <compat/linux/linux_mib.h>
76 #include <compat/linux/linux_util.h>
77 
78 #ifdef __alpha__
79 #define BSD_TO_LINUX_SIGNAL(sig)       (sig)
80 #else
81 #define BSD_TO_LINUX_SIGNAL(sig)	\
82 	(((sig) <= LINUX_SIGTBLSZ) ? bsd_to_linux_signal[_SIG_IDX(sig)] : sig)
83 #endif
84 
85 #ifndef __alpha__
86 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
87 	RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
88 	RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
89 	RLIMIT_MEMLOCK, -1
90 };
91 #endif /*!__alpha__*/
92 
93 struct l_sysinfo {
94 	l_long		uptime;		/* Seconds since boot */
95 	l_ulong		loads[3];	/* 1, 5, and 15 minute load averages */
96 	l_ulong		totalram;	/* Total usable main memory size */
97 	l_ulong		freeram;	/* Available memory size */
98 	l_ulong		sharedram;	/* Amount of shared memory */
99 	l_ulong		bufferram;	/* Memory used by buffers */
100 	l_ulong		totalswap;	/* Total swap space size */
101 	l_ulong		freeswap;	/* swap space still available */
102 	l_ushort	procs;		/* Number of current processes */
103 	char		_f[22];		/* Pads structure to 64 bytes */
104 };
105 #ifndef __alpha__
106 int
107 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
108 {
109 	struct l_sysinfo sysinfo;
110 	vm_object_t object;
111 	int i;
112 	struct timespec ts;
113 
114 	/* Uptime is copied out of print_uptime() in kern_shutdown.c */
115 	getnanouptime(&ts);
116 	i = 0;
117 	if (ts.tv_sec >= 86400) {
118 		ts.tv_sec %= 86400;
119 		i = 1;
120 	}
121 	if (i || ts.tv_sec >= 3600) {
122 		ts.tv_sec %= 3600;
123 		i = 1;
124 	}
125 	if (i || ts.tv_sec >= 60) {
126 		ts.tv_sec %= 60;
127 		i = 1;
128 	}
129 	sysinfo.uptime=ts.tv_sec;
130 
131 	/* Use the information from the mib to get our load averages */
132 	for (i = 0; i < 3; i++)
133 		sysinfo.loads[i] = averunnable.ldavg[i];
134 
135 	sysinfo.totalram = physmem * PAGE_SIZE;
136 	sysinfo.freeram = sysinfo.totalram - cnt.v_wire_count * PAGE_SIZE;
137 
138 	sysinfo.sharedram = 0;
139 	for (object = TAILQ_FIRST(&vm_object_list); object != NULL;
140 	     object = TAILQ_NEXT(object, object_list))
141 		if (object->shadow_count > 1)
142 			sysinfo.sharedram += object->resident_page_count;
143 
144 	sysinfo.sharedram *= PAGE_SIZE;
145 	sysinfo.bufferram = 0;
146 
147 	if (swapblist == NULL) {
148 		sysinfo.totalswap= 0;
149 		sysinfo.freeswap = 0;
150 	} else {
151 		sysinfo.totalswap = swapblist->bl_blocks * 1024;
152 		sysinfo.freeswap = swapblist->bl_root->u.bmu_avail * PAGE_SIZE;
153 	}
154 
155 	sysinfo.procs = 20; /* Hack */
156 
157 	return copyout(&sysinfo, (caddr_t)args->info, sizeof(sysinfo));
158 }
159 #endif /*!__alpha__*/
160 
161 #ifndef __alpha__
162 int
163 linux_alarm(struct thread *td, struct linux_alarm_args *args)
164 {
165 	struct itimerval it, old_it;
166 	struct timeval tv;
167 	int s;
168 
169 #ifdef DEBUG
170 	if (ldebug(alarm))
171 		printf(ARGS(alarm, "%u"), args->secs);
172 #endif
173 
174 	if (args->secs > 100000000)
175 		return EINVAL;
176 
177 	it.it_value.tv_sec = (long)args->secs;
178 	it.it_value.tv_usec = 0;
179 	it.it_interval.tv_sec = 0;
180 	it.it_interval.tv_usec = 0;
181 	s = splsoftclock();
182 	old_it = td->td_proc->p_realtimer;
183 	getmicrouptime(&tv);
184 	if (timevalisset(&old_it.it_value))
185 		callout_stop(&td->td_proc->p_itcallout);
186 	if (it.it_value.tv_sec != 0) {
187 		callout_reset(&td->td_proc->p_itcallout, tvtohz(&it.it_value),
188 		    realitexpire, td->td_proc);
189 		timevaladd(&it.it_value, &tv);
190 	}
191 	td->td_proc->p_realtimer = it;
192 	splx(s);
193 	if (timevalcmp(&old_it.it_value, &tv, >)) {
194 		timevalsub(&old_it.it_value, &tv);
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 #endif /*!__alpha__*/
202 
203 int
204 linux_brk(struct thread *td, struct linux_brk_args *args)
205 {
206 	struct vmspace *vm = td->td_proc->p_vmspace;
207 	vm_offset_t new, old;
208 	struct obreak_args /* {
209 		char * nsize;
210 	} */ tmp;
211 
212 #ifdef DEBUG
213 	if (ldebug(brk))
214 		printf(ARGS(brk, "%p"), (void *)args->dsend);
215 #endif
216 	old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize);
217 	new = (vm_offset_t)args->dsend;
218 	tmp.nsize = (char *) new;
219 	if (((caddr_t)new > vm->vm_daddr) && !obreak(td, &tmp))
220 		td->td_retval[0] = (long)new;
221 	else
222 		td->td_retval[0] = (long)old;
223 
224 	return 0;
225 }
226 
227 int
228 linux_uselib(struct thread *td, struct linux_uselib_args *args)
229 {
230 	struct nameidata ni;
231 	struct vnode *vp;
232 	struct exec *a_out;
233 	struct vattr attr;
234 	vm_offset_t vmaddr;
235 	unsigned long file_offset;
236 	vm_offset_t buffer;
237 	unsigned long bss_size;
238 	char *library;
239 	int error;
240 	int locked;
241 
242 	LCONVPATHEXIST(td, args->library, &library);
243 
244 #ifdef DEBUG
245 	if (ldebug(uselib))
246 		printf(ARGS(uselib, "%s"), library);
247 #endif
248 
249 	a_out = NULL;
250 	locked = 0;
251 	vp = NULL;
252 
253 	/*
254 	 * XXX: This code should make use of vn_open(), rather than doing
255 	 * all this stuff itself.
256 	 */
257 	NDINIT(&ni, LOOKUP, FOLLOW|LOCKLEAF, UIO_SYSSPACE, library, td);
258 	error = namei(&ni);
259 	LFREEPATH(library);
260 	if (error)
261 		goto cleanup;
262 
263 	vp = ni.ni_vp;
264 	/*
265 	 * XXX - This looks like a bogus check. A LOCKLEAF namei should not
266 	 * succeed without returning a vnode.
267 	 */
268 	if (vp == NULL) {
269 		error = ENOEXEC;	/* ?? */
270 		goto cleanup;
271 	}
272 	NDFREE(&ni, NDF_ONLY_PNBUF);
273 
274 	/*
275 	 * From here on down, we have a locked vnode that must be unlocked.
276 	 */
277 	locked++;
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, td);
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 		error = ENOEXEC;
293 		goto cleanup;
294 	}
295 
296 	/* Sensible size? */
297 	if (attr.va_size == 0) {
298 		error = ENOEXEC;
299 		goto cleanup;
300 	}
301 
302 	/* Can we access it? */
303 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
304 	if (error)
305 		goto cleanup;
306 
307 	/*
308 	 * XXX: This should use vn_open() so that it is properly authorized,
309 	 * and to reduce code redundancy all over the place here.
310 	 */
311 #ifdef MAC
312 	error = mac_check_vnode_open(td->td_ucred, vp, FREAD);
313 	if (error)
314 		goto cleanup;
315 #endif
316 	error = VOP_OPEN(vp, FREAD, td->td_ucred, td);
317 	if (error)
318 		goto cleanup;
319 
320 	/* Pull in executable header into kernel_map */
321 	error = vm_mmap(kernel_map, (vm_offset_t *)&a_out, PAGE_SIZE,
322 	    VM_PROT_READ, VM_PROT_READ, 0, (caddr_t)vp, 0);
323 	/*
324 	 * Lock no longer needed
325 	 */
326 	locked = 0;
327 	VOP_UNLOCK(vp, 0, td);
328 
329 	if (error)
330 		goto cleanup;
331 
332 	/* Is it a Linux binary ? */
333 	if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
334 		error = ENOEXEC;
335 		goto cleanup;
336 	}
337 
338 	/*
339 	 * While we are here, we should REALLY do some more checks
340 	 */
341 
342 	/* Set file/virtual offset based on a.out variant. */
343 	switch ((int)(a_out->a_magic & 0xffff)) {
344 	case 0413:	/* ZMAGIC */
345 		file_offset = 1024;
346 		break;
347 	case 0314:	/* QMAGIC */
348 		file_offset = 0;
349 		break;
350 	default:
351 		error = ENOEXEC;
352 		goto cleanup;
353 	}
354 
355 	bss_size = round_page(a_out->a_bss);
356 
357 	/* Check various fields in header for validity/bounds. */
358 	if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
359 		error = ENOEXEC;
360 		goto cleanup;
361 	}
362 
363 	/* text + data can't exceed file size */
364 	if (a_out->a_data + a_out->a_text > attr.va_size) {
365 		error = EFAULT;
366 		goto cleanup;
367 	}
368 
369 	/* To protect td->td_proc->p_rlimit in the if condition. */
370 	mtx_assert(&Giant, MA_OWNED);
371 
372 	/*
373 	 * text/data/bss must not exceed limits
374 	 * XXX - this is not complete. it should check current usage PLUS
375 	 * the resources needed by this library.
376 	 */
377 	if (a_out->a_text > maxtsiz ||
378 	    a_out->a_data + bss_size >
379 	    td->td_proc->p_rlimit[RLIMIT_DATA].rlim_cur) {
380 		error = ENOMEM;
381 		goto cleanup;
382 	}
383 
384 	mp_fixme("Unlocked vflags access.");
385 	/* prevent more writers */
386 	vp->v_vflag |= VV_TEXT;
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, (caddr_t)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((caddr_t)(uintptr_t)(buffer + file_offset),
419 		    (caddr_t)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, (caddr_t)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, td);
467 
468 	/* Release the kernel mapping. */
469 	if (a_out)
470 		vm_map_remove(kernel_map, (vm_offset_t)a_out,
471 		    (vm_offset_t)a_out + PAGE_SIZE);
472 
473 	return error;
474 }
475 
476 int
477 linux_select(struct thread *td, struct linux_select_args *args)
478 {
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((caddr_t)args->timeout, &utv,
495 		    sizeof(utv))))
496 			goto select_out;
497 #ifdef DEBUG
498 		if (ldebug(select))
499 			printf(LMSG("incoming timeout (%ld/%ld)"),
500 			    utv.tv_sec, utv.tv_usec);
501 #endif
502 
503 		if (itimerfix(&utv)) {
504 			/*
505 			 * The timeval was invalid.  Convert it to something
506 			 * valid that will act as it does under Linux.
507 			 */
508 			utv.tv_sec += utv.tv_usec / 1000000;
509 			utv.tv_usec %= 1000000;
510 			if (utv.tv_usec < 0) {
511 				utv.tv_sec -= 1;
512 				utv.tv_usec += 1000000;
513 			}
514 			if (utv.tv_sec < 0)
515 				timevalclear(&utv);
516 		}
517 		microtime(&tv0);
518 		tvp = &utv;
519 	} else
520 		tvp = NULL;
521 
522 	error = kern_select(td, args->nfds, args->readfds, args->writefds,
523 	    args->exceptfds, tvp);
524 
525 #ifdef DEBUG
526 	if (ldebug(select))
527 		printf(LMSG("real select returns %d"), error);
528 #endif
529 	if (error) {
530 		/*
531 		 * See fs/select.c in the Linux kernel.  Without this,
532 		 * Maelstrom doesn't work.
533 		 */
534 		if (error == ERESTART)
535 			error = EINTR;
536 		goto select_out;
537 	}
538 
539 	if (args->timeout) {
540 		if (td->td_retval[0]) {
541 			/*
542 			 * Compute how much time was left of the timeout,
543 			 * by subtracting the current time and the time
544 			 * before we started the call, and subtracting
545 			 * that result from the user-supplied value.
546 			 */
547 			microtime(&tv1);
548 			timevalsub(&tv1, &tv0);
549 			timevalsub(&utv, &tv1);
550 			if (utv.tv_sec < 0)
551 				timevalclear(&utv);
552 		} else
553 			timevalclear(&utv);
554 #ifdef DEBUG
555 		if (ldebug(select))
556 			printf(LMSG("outgoing timeout (%ld/%ld)"),
557 			    utv.tv_sec, utv.tv_usec);
558 #endif
559 		if ((error = copyout(&utv, (caddr_t)args->timeout,
560 		    sizeof(utv))))
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 *)args->addr,
585 		    (unsigned long)args->old_len,
586 		    (unsigned long)args->new_len,
587 		    (unsigned long)args->flags);
588 #endif
589 	args->new_len = round_page(args->new_len);
590 	args->old_len = round_page(args->old_len);
591 
592 	if (args->new_len > args->old_len) {
593 		td->td_retval[0] = 0;
594 		return ENOMEM;
595 	}
596 
597 	if (args->new_len < args->old_len) {
598 		bsd_args.addr = (caddr_t)(args->addr + args->new_len);
599 		bsd_args.len = args->old_len - args->new_len;
600 		error = munmap(td, &bsd_args);
601 	}
602 
603 	td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
604 	return error;
605 }
606 
607 #define LINUX_MS_ASYNC       0x0001
608 #define LINUX_MS_INVALIDATE  0x0002
609 #define LINUX_MS_SYNC        0x0004
610 
611 int
612 linux_msync(struct thread *td, struct linux_msync_args *args)
613 {
614 	struct msync_args bsd_args;
615 
616 	bsd_args.addr = (caddr_t)args->addr;
617 	bsd_args.len = args->len;
618 	bsd_args.flags = args->fl & ~LINUX_MS_SYNC;
619 
620 	return msync(td, &bsd_args);
621 }
622 
623 #ifndef __alpha__
624 int
625 linux_time(struct thread *td, struct linux_time_args *args)
626 {
627 	struct timeval tv;
628 	l_time_t tm;
629 	int error;
630 
631 #ifdef DEBUG
632 	if (ldebug(time))
633 		printf(ARGS(time, "*"));
634 #endif
635 
636 	microtime(&tv);
637 	tm = tv.tv_sec;
638 	if (args->tm && (error = copyout(&tm, (caddr_t)args->tm, sizeof(tm))))
639 		return error;
640 	td->td_retval[0] = tm;
641 	return 0;
642 }
643 #endif	/*!__alpha__*/
644 
645 struct l_times_argv {
646 	l_long		tms_utime;
647 	l_long		tms_stime;
648 	l_long		tms_cutime;
649 	l_long		tms_cstime;
650 };
651 
652 #ifdef __alpha__
653 #define CLK_TCK 1024	/* Linux uses 1024 on alpha */
654 #else
655 #define CLK_TCK 100	/* Linux uses 100 */
656 #endif
657 
658 #define CONVTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
659 
660 int
661 linux_times(struct thread *td, struct linux_times_args *args)
662 {
663 	struct timeval tv;
664 	struct l_times_argv tms;
665 	struct rusage ru;
666 	int error;
667 
668 #ifdef DEBUG
669 	if (ldebug(times))
670 		printf(ARGS(times, "*"));
671 #endif
672 
673 	mtx_lock_spin(&sched_lock);
674 	calcru(td->td_proc, &ru.ru_utime, &ru.ru_stime, NULL);
675 	mtx_unlock_spin(&sched_lock);
676 
677 	tms.tms_utime = CONVTCK(ru.ru_utime);
678 	tms.tms_stime = CONVTCK(ru.ru_stime);
679 
680 	tms.tms_cutime = CONVTCK(td->td_proc->p_stats->p_cru.ru_utime);
681 	tms.tms_cstime = CONVTCK(td->td_proc->p_stats->p_cru.ru_stime);
682 
683 	if ((error = copyout(&tms, (caddr_t)args->buf, sizeof(tms))))
684 		return error;
685 
686 	microuptime(&tv);
687 	td->td_retval[0] = (int)CONVTCK(tv);
688 	return 0;
689 }
690 
691 int
692 linux_newuname(struct thread *td, struct linux_newuname_args *args)
693 {
694 	struct l_new_utsname utsname;
695 	char osname[LINUX_MAX_UTSNAME];
696 	char osrelease[LINUX_MAX_UTSNAME];
697 
698 #ifdef DEBUG
699 	if (ldebug(newuname))
700 		printf(ARGS(newuname, "*"));
701 #endif
702 
703 	linux_get_osname(td->td_proc, osname);
704 	linux_get_osrelease(td->td_proc, osrelease);
705 
706 	bzero(&utsname, sizeof(utsname));
707 	strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
708 	getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
709 	strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
710 	strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
711 	strlcpy(utsname.machine, machine, LINUX_MAX_UTSNAME);
712 	strlcpy(utsname.domainname, domainname, LINUX_MAX_UTSNAME);
713 
714 	return (copyout(&utsname, (caddr_t)args->buf, sizeof(utsname)));
715 }
716 
717 #if defined(__i386__)
718 struct l_utimbuf {
719 	l_time_t l_actime;
720 	l_time_t l_modtime;
721 };
722 
723 int
724 linux_utime(struct thread *td, struct linux_utime_args *args)
725 {
726 	struct timeval tv[2], *tvp;
727 	struct l_utimbuf lut;
728 	char *fname;
729 	int error;
730 
731 	LCONVPATHEXIST(td, args->fname, &fname);
732 
733 #ifdef DEBUG
734 	if (ldebug(utime))
735 		printf(ARGS(utime, "%s, *"), fname);
736 #endif
737 
738 	if (args->times) {
739 		if ((error = copyin((caddr_t)args->times, &lut, sizeof lut))) {
740 			LFREEPATH(fname);
741 			return error;
742 		}
743 		tv[0].tv_sec = lut.l_actime;
744 		tv[0].tv_usec = 0;
745 		tv[1].tv_sec = lut.l_modtime;
746 		tv[1].tv_usec = 0;
747 		tvp = tv;
748 	} else
749 		tvp = NULL;
750 
751 	error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
752 	LFREEPATH(fname);
753 	return (error);
754 }
755 #endif /* __i386__ */
756 
757 #define __WCLONE 0x80000000
758 
759 #ifndef __alpha__
760 int
761 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
762 {
763 	struct wait_args /* {
764 		int pid;
765 		int *status;
766 		int options;
767 		struct	rusage *rusage;
768 	} */ tmp;
769 	int error, tmpstat;
770 
771 #ifdef DEBUG
772 	if (ldebug(waitpid))
773 		printf(ARGS(waitpid, "%d, %p, %d"),
774 		    args->pid, (void *)args->status, args->options);
775 #endif
776 
777 	tmp.pid = args->pid;
778 	tmp.status = args->status;
779 	tmp.options = (args->options & (WNOHANG | WUNTRACED));
780 	/* WLINUXCLONE should be equal to __WCLONE, but we make sure */
781 	if (args->options & __WCLONE)
782 		tmp.options |= WLINUXCLONE;
783 	tmp.rusage = NULL;
784 
785 	if ((error = wait4(td, &tmp)) != 0)
786 		return error;
787 
788 	if (args->status) {
789 		if ((error = copyin((caddr_t)args->status, &tmpstat,
790 		    sizeof(int))) != 0)
791 			return error;
792 		tmpstat &= 0xffff;
793 		if (WIFSIGNALED(tmpstat))
794 			tmpstat = (tmpstat & 0xffffff80) |
795 			    BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat));
796 		else if (WIFSTOPPED(tmpstat))
797 			tmpstat = (tmpstat & 0xffff00ff) |
798 			    (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8);
799 		return copyout(&tmpstat, (caddr_t)args->status, sizeof(int));
800 	}
801 
802 	return 0;
803 }
804 #endif	/*!__alpha__*/
805 
806 int
807 linux_wait4(struct thread *td, struct linux_wait4_args *args)
808 {
809 	struct wait_args /* {
810 		int pid;
811 		int *status;
812 		int options;
813 		struct	rusage *rusage;
814 	} */ tmp;
815 	int error, tmpstat;
816 
817 #ifdef DEBUG
818 	if (ldebug(wait4))
819 		printf(ARGS(wait4, "%d, %p, %d, %p"),
820 		    args->pid, (void *)args->status, args->options,
821 		    (void *)args->rusage);
822 #endif
823 
824 	tmp.pid = args->pid;
825 	tmp.status = args->status;
826 	tmp.options = (args->options & (WNOHANG | WUNTRACED));
827 	/* WLINUXCLONE should be equal to __WCLONE, but we make sure */
828 	if (args->options & __WCLONE)
829 		tmp.options |= WLINUXCLONE;
830 	tmp.rusage = (struct rusage *)args->rusage;
831 
832 	if ((error = wait4(td, &tmp)) != 0)
833 		return error;
834 
835 	SIGDELSET(td->td_proc->p_siglist, SIGCHLD);
836 
837 	if (args->status) {
838 		if ((error = copyin((caddr_t)args->status, &tmpstat,
839 		    sizeof(int))) != 0)
840 			return error;
841 		tmpstat &= 0xffff;
842 		if (WIFSIGNALED(tmpstat))
843 			tmpstat = (tmpstat & 0xffffff80) |
844 			    BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat));
845 		else if (WIFSTOPPED(tmpstat))
846 			tmpstat = (tmpstat & 0xffff00ff) |
847 			    (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8);
848 		return copyout(&tmpstat, (caddr_t)args->status, sizeof(int));
849 	}
850 
851 	return 0;
852 }
853 
854 int
855 linux_mknod(struct thread *td, struct linux_mknod_args *args)
856 {
857 	char *path;
858 	int error;
859 
860 	LCONVPATHCREAT(td, args->path, &path);
861 
862 #ifdef DEBUG
863 	if (ldebug(mknod))
864 		printf(ARGS(mknod, "%s, %d, %d"), path, args->mode, args->dev);
865 #endif
866 
867 	if (args->mode & S_IFIFO)
868 		error = kern_mkfifo(td, path, UIO_SYSSPACE, args->mode);
869 	else
870 		error = kern_mknod(td, path, UIO_SYSSPACE, args->mode,
871 		    args->dev);
872 	LFREEPATH(path);
873 	return (error);
874 }
875 
876 /*
877  * UGH! This is just about the dumbest idea I've ever heard!!
878  */
879 int
880 linux_personality(struct thread *td, struct linux_personality_args *args)
881 {
882 #ifdef DEBUG
883 	if (ldebug(personality))
884 		printf(ARGS(personality, "%d"), args->per);
885 #endif
886 #ifndef __alpha__
887 	if (args->per != 0)
888 		return EINVAL;
889 #endif
890 
891 	/* Yes Jim, it's still a Linux... */
892 	td->td_retval[0] = 0;
893 	return 0;
894 }
895 
896 /*
897  * Wrappers for get/setitimer for debugging..
898  */
899 int
900 linux_setitimer(struct thread *td, struct linux_setitimer_args *args)
901 {
902 	struct setitimer_args bsa;
903 	struct itimerval foo;
904 	int error;
905 
906 #ifdef DEBUG
907 	if (ldebug(setitimer))
908 		printf(ARGS(setitimer, "%p, %p"),
909 		    (void *)args->itv, (void *)args->oitv);
910 #endif
911 	bsa.which = args->which;
912 	bsa.itv = (struct itimerval *)args->itv;
913 	bsa.oitv = (struct itimerval *)args->oitv;
914 	if (args->itv) {
915 	    if ((error = copyin((caddr_t)args->itv, &foo, sizeof(foo))))
916 		return error;
917 #ifdef DEBUG
918 	    if (ldebug(setitimer)) {
919 	        printf("setitimer: value: sec: %ld, usec: %ld\n",
920 		    foo.it_value.tv_sec, foo.it_value.tv_usec);
921 	        printf("setitimer: interval: sec: %ld, usec: %ld\n",
922 		    foo.it_interval.tv_sec, foo.it_interval.tv_usec);
923 	    }
924 #endif
925 	}
926 	return setitimer(td, &bsa);
927 }
928 
929 int
930 linux_getitimer(struct thread *td, struct linux_getitimer_args *args)
931 {
932 	struct getitimer_args bsa;
933 #ifdef DEBUG
934 	if (ldebug(getitimer))
935 		printf(ARGS(getitimer, "%p"), (void *)args->itv);
936 #endif
937 	bsa.which = args->which;
938 	bsa.itv = (struct itimerval *)args->itv;
939 	return getitimer(td, &bsa);
940 }
941 
942 #ifndef __alpha__
943 int
944 linux_nice(struct thread *td, struct linux_nice_args *args)
945 {
946 	struct setpriority_args	bsd_args;
947 
948 	bsd_args.which = PRIO_PROCESS;
949 	bsd_args.who = 0;	/* current process */
950 	bsd_args.prio = args->inc;
951 	return setpriority(td, &bsd_args);
952 }
953 #endif	/*!__alpha__*/
954 
955 int
956 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
957 {
958 	struct ucred *newcred, *oldcred;
959 	l_gid_t linux_gidset[NGROUPS];
960 	gid_t *bsd_gidset;
961 	int ngrp, error;
962 	struct proc *p;
963 
964 	ngrp = args->gidsetsize;
965 	if (ngrp >= NGROUPS)
966 		return (EINVAL);
967 	error = copyin((caddr_t)args->grouplist, linux_gidset,
968 	    ngrp * sizeof(l_gid_t));
969 	if (error)
970 		return (error);
971 	newcred = crget();
972 	p = td->td_proc;
973 	PROC_LOCK(p);
974 	oldcred = p->p_ucred;
975 
976 	/*
977 	 * cr_groups[0] holds egid. Setting the whole set from
978 	 * the supplied set will cause egid to be changed too.
979 	 * Keep cr_groups[0] unchanged to prevent that.
980 	 */
981 
982 	if ((error = suser_cred(oldcred, PRISON_ROOT)) != 0) {
983 		PROC_UNLOCK(p);
984 		crfree(newcred);
985 		return (error);
986 	}
987 
988 	crcopy(newcred, oldcred);
989 	if (ngrp > 0) {
990 		newcred->cr_ngroups = ngrp + 1;
991 
992 		bsd_gidset = newcred->cr_groups;
993 		ngrp--;
994 		while (ngrp >= 0) {
995 			bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
996 			ngrp--;
997 		}
998 	}
999 	else
1000 		newcred->cr_ngroups = 1;
1001 
1002 	setsugid(p);
1003 	p->p_ucred = newcred;
1004 	PROC_UNLOCK(p);
1005 	crfree(oldcred);
1006 	return (0);
1007 }
1008 
1009 int
1010 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1011 {
1012 	struct ucred *cred;
1013 	l_gid_t linux_gidset[NGROUPS];
1014 	gid_t *bsd_gidset;
1015 	int bsd_gidsetsz, ngrp, error;
1016 
1017 	cred = td->td_ucred;
1018 	bsd_gidset = cred->cr_groups;
1019 	bsd_gidsetsz = cred->cr_ngroups - 1;
1020 
1021 	/*
1022 	 * cr_groups[0] holds egid. Returning the whole set
1023 	 * here will cause a duplicate. Exclude cr_groups[0]
1024 	 * to prevent that.
1025 	 */
1026 
1027 	if ((ngrp = args->gidsetsize) == 0) {
1028 		td->td_retval[0] = bsd_gidsetsz;
1029 		return (0);
1030 	}
1031 
1032 	if (ngrp < bsd_gidsetsz)
1033 		return (EINVAL);
1034 
1035 	ngrp = 0;
1036 	while (ngrp < bsd_gidsetsz) {
1037 		linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1038 		ngrp++;
1039 	}
1040 
1041 	if ((error = copyout(linux_gidset, (caddr_t)args->grouplist,
1042 	    ngrp * sizeof(l_gid_t))))
1043 		return (error);
1044 
1045 	td->td_retval[0] = ngrp;
1046 	return (0);
1047 }
1048 
1049 #ifndef __alpha__
1050 int
1051 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1052 {
1053 	struct rlimit bsd_rlim;
1054 	struct l_rlimit rlim;
1055 	u_int which;
1056 	int error;
1057 
1058 #ifdef DEBUG
1059 	if (ldebug(setrlimit))
1060 		printf(ARGS(setrlimit, "%d, %p"),
1061 		    args->resource, (void *)args->rlim);
1062 #endif
1063 
1064 	if (args->resource >= LINUX_RLIM_NLIMITS)
1065 		return (EINVAL);
1066 
1067 	which = linux_to_bsd_resource[args->resource];
1068 	if (which == -1)
1069 		return (EINVAL);
1070 
1071 	error = copyin((caddr_t)args->rlim, &rlim, sizeof(rlim));
1072 	if (error)
1073 		return (error);
1074 
1075 	bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1076 	bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1077 	return (dosetrlimit(td, which, &bsd_rlim));
1078 }
1079 
1080 int
1081 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1082 {
1083 	struct l_rlimit rlim;
1084 	struct proc *p = td->td_proc;
1085 	struct rlimit *bsd_rlp;
1086 	u_int which;
1087 
1088 #ifdef DEBUG
1089 	if (ldebug(old_getrlimit))
1090 		printf(ARGS(old_getrlimit, "%d, %p"),
1091 		    args->resource, (void *)args->rlim);
1092 #endif
1093 
1094 	if (args->resource >= LINUX_RLIM_NLIMITS)
1095 		return (EINVAL);
1096 
1097 	which = linux_to_bsd_resource[args->resource];
1098 	if (which == -1)
1099 		return (EINVAL);
1100 	bsd_rlp = &p->p_rlimit[which];
1101 
1102 	rlim.rlim_cur = (unsigned long)bsd_rlp->rlim_cur;
1103 	if (rlim.rlim_cur == ULONG_MAX)
1104 		rlim.rlim_cur = LONG_MAX;
1105 	rlim.rlim_max = (unsigned long)bsd_rlp->rlim_max;
1106 	if (rlim.rlim_max == ULONG_MAX)
1107 		rlim.rlim_max = LONG_MAX;
1108 	return (copyout(&rlim, (caddr_t)args->rlim, sizeof(rlim)));
1109 }
1110 
1111 int
1112 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1113 {
1114 	struct l_rlimit rlim;
1115 	struct proc *p = td->td_proc;
1116 	struct rlimit *bsd_rlp;
1117 	u_int which;
1118 
1119 #ifdef DEBUG
1120 	if (ldebug(getrlimit))
1121 		printf(ARGS(getrlimit, "%d, %p"),
1122 		    args->resource, (void *)args->rlim);
1123 #endif
1124 
1125 	if (args->resource >= LINUX_RLIM_NLIMITS)
1126 		return (EINVAL);
1127 
1128 	which = linux_to_bsd_resource[args->resource];
1129 	if (which == -1)
1130 		return (EINVAL);
1131 	bsd_rlp = &p->p_rlimit[which];
1132 
1133 	rlim.rlim_cur = (l_ulong)bsd_rlp->rlim_cur;
1134 	rlim.rlim_max = (l_ulong)bsd_rlp->rlim_max;
1135 	return (copyout(&rlim, (caddr_t)args->rlim, sizeof(rlim)));
1136 }
1137 #endif /*!__alpha__*/
1138 
1139 int
1140 linux_sched_setscheduler(struct thread *td,
1141     struct linux_sched_setscheduler_args *args)
1142 {
1143 	struct sched_setscheduler_args bsd;
1144 
1145 #ifdef DEBUG
1146 	if (ldebug(sched_setscheduler))
1147 		printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1148 		    args->pid, args->policy, (const void *)args->param);
1149 #endif
1150 
1151 	switch (args->policy) {
1152 	case LINUX_SCHED_OTHER:
1153 		bsd.policy = SCHED_OTHER;
1154 		break;
1155 	case LINUX_SCHED_FIFO:
1156 		bsd.policy = SCHED_FIFO;
1157 		break;
1158 	case LINUX_SCHED_RR:
1159 		bsd.policy = SCHED_RR;
1160 		break;
1161 	default:
1162 		return EINVAL;
1163 	}
1164 
1165 	bsd.pid = args->pid;
1166 	bsd.param = (struct sched_param *)args->param;
1167 	return sched_setscheduler(td, &bsd);
1168 }
1169 
1170 int
1171 linux_sched_getscheduler(struct thread *td,
1172     struct linux_sched_getscheduler_args *args)
1173 {
1174 	struct sched_getscheduler_args bsd;
1175 	int error;
1176 
1177 #ifdef DEBUG
1178 	if (ldebug(sched_getscheduler))
1179 		printf(ARGS(sched_getscheduler, "%d"), args->pid);
1180 #endif
1181 
1182 	bsd.pid = args->pid;
1183 	error = sched_getscheduler(td, &bsd);
1184 
1185 	switch (td->td_retval[0]) {
1186 	case SCHED_OTHER:
1187 		td->td_retval[0] = LINUX_SCHED_OTHER;
1188 		break;
1189 	case SCHED_FIFO:
1190 		td->td_retval[0] = LINUX_SCHED_FIFO;
1191 		break;
1192 	case SCHED_RR:
1193 		td->td_retval[0] = LINUX_SCHED_RR;
1194 		break;
1195 	}
1196 
1197 	return error;
1198 }
1199 
1200 int
1201 linux_sched_get_priority_max(struct thread *td,
1202     struct linux_sched_get_priority_max_args *args)
1203 {
1204 	struct sched_get_priority_max_args bsd;
1205 
1206 #ifdef DEBUG
1207 	if (ldebug(sched_get_priority_max))
1208 		printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1209 #endif
1210 
1211 	switch (args->policy) {
1212 	case LINUX_SCHED_OTHER:
1213 		bsd.policy = SCHED_OTHER;
1214 		break;
1215 	case LINUX_SCHED_FIFO:
1216 		bsd.policy = SCHED_FIFO;
1217 		break;
1218 	case LINUX_SCHED_RR:
1219 		bsd.policy = SCHED_RR;
1220 		break;
1221 	default:
1222 		return EINVAL;
1223 	}
1224 	return sched_get_priority_max(td, &bsd);
1225 }
1226 
1227 int
1228 linux_sched_get_priority_min(struct thread *td,
1229     struct linux_sched_get_priority_min_args *args)
1230 {
1231 	struct sched_get_priority_min_args bsd;
1232 
1233 #ifdef DEBUG
1234 	if (ldebug(sched_get_priority_min))
1235 		printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1236 #endif
1237 
1238 	switch (args->policy) {
1239 	case LINUX_SCHED_OTHER:
1240 		bsd.policy = SCHED_OTHER;
1241 		break;
1242 	case LINUX_SCHED_FIFO:
1243 		bsd.policy = SCHED_FIFO;
1244 		break;
1245 	case LINUX_SCHED_RR:
1246 		bsd.policy = SCHED_RR;
1247 		break;
1248 	default:
1249 		return EINVAL;
1250 	}
1251 	return sched_get_priority_min(td, &bsd);
1252 }
1253 
1254 #define REBOOT_CAD_ON	0x89abcdef
1255 #define REBOOT_CAD_OFF	0
1256 #define REBOOT_HALT	0xcdef0123
1257 
1258 int
1259 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1260 {
1261 	struct reboot_args bsd_args;
1262 
1263 #ifdef DEBUG
1264 	if (ldebug(reboot))
1265 		printf(ARGS(reboot, "0x%x"), args->cmd);
1266 #endif
1267 	if (args->cmd == REBOOT_CAD_ON || args->cmd == REBOOT_CAD_OFF)
1268 		return (0);
1269 	bsd_args.opt = (args->cmd == REBOOT_HALT) ? RB_HALT : 0;
1270 	return (reboot(td, &bsd_args));
1271 }
1272 
1273 #ifndef __alpha__
1274 
1275 /*
1276  * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1277  * td->td_retval[1] when COMPAT_43 or COMPAT_SUNOS is defined. This
1278  * globbers registers that are assumed to be preserved. The following
1279  * lightweight syscalls fixes this. See also linux_getgid16() and
1280  * linux_getuid16() in linux_uid16.c.
1281  *
1282  * linux_getpid() - MP SAFE
1283  * linux_getgid() - MP SAFE
1284  * linux_getuid() - MP SAFE
1285  */
1286 
1287 int
1288 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1289 {
1290 
1291 	td->td_retval[0] = td->td_proc->p_pid;
1292 	return (0);
1293 }
1294 
1295 int
1296 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1297 {
1298 
1299 	td->td_retval[0] = td->td_ucred->cr_rgid;
1300 	return (0);
1301 }
1302 
1303 int
1304 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1305 {
1306 
1307 	td->td_retval[0] = td->td_ucred->cr_ruid;
1308 	return (0);
1309 }
1310 
1311 #endif /*!__alpha__*/
1312 
1313 int
1314 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1315 {
1316 	struct getsid_args bsd;
1317 	bsd.pid = args->pid;
1318 	return getsid(td, &bsd);
1319 }
1320