xref: /netbsd-src/sys/compat/linux/common/linux_misc.c (revision 1ca5c1b28139779176bd5c13ad7c5f25c0bcd5f8)
1 /*	$NetBSD: linux_misc.c,v 1.96 2001/12/09 03:07:43 chs Exp $	*/
2 
3 /*-
4  * Copyright (c) 1995, 1998, 1999 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Frank van der Linden and Eric Haszlakiewicz; by Jason R. Thorpe
9  * of the Numerical Aerospace Simulation Facility, NASA Ames Research Center.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 3. All advertising materials mentioning features or use of this software
20  *    must display the following acknowledgement:
21  *	This product includes software developed by the NetBSD
22  *	Foundation, Inc. and its contributors.
23  * 4. Neither the name of The NetBSD Foundation nor the names of its
24  *    contributors may be used to endorse or promote products derived
25  *    from this software without specific prior written permission.
26  *
27  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
28  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
29  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
30  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
31  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
32  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
33  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
34  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
35  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
36  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
37  * POSSIBILITY OF SUCH DAMAGE.
38  */
39 
40 /*
41  * Linux compatibility module. Try to deal with various Linux system calls.
42  */
43 
44 /*
45  * These functions have been moved to multiarch to allow
46  * selection of which machines include them to be
47  * determined by the individual files.linux_<arch> files.
48  *
49  * Function in multiarch:
50  *	linux_sys_break			: linux_break.c
51  *	linux_sys_alarm			: linux_misc_notalpha.c
52  *	linux_sys_getresgid		: linux_misc_notalpha.c
53  *	linux_sys_nice			: linux_misc_notalpha.c
54  *	linux_sys_readdir		: linux_misc_notalpha.c
55  *	linux_sys_setresgid		: linux_misc_notalpha.c
56  *	linux_sys_time			: linux_misc_notalpha.c
57  *	linux_sys_utime			: linux_misc_notalpha.c
58  *	linux_sys_waitpid		: linux_misc_notalpha.c
59  *	linux_sys_old_mmap		: linux_oldmmap.c
60  *	linux_sys_oldolduname		: linux_oldolduname.c
61  *	linux_sys_oldselect		: linux_oldselect.c
62  *	linux_sys_olduname		: linux_olduname.c
63  *	linux_sys_pipe			: linux_pipe.c
64  */
65 
66 #include <sys/cdefs.h>
67 __KERNEL_RCSID(0, "$NetBSD: linux_misc.c,v 1.96 2001/12/09 03:07:43 chs Exp $");
68 
69 #include <sys/param.h>
70 #include <sys/systm.h>
71 #include <sys/namei.h>
72 #include <sys/proc.h>
73 #include <sys/dirent.h>
74 #include <sys/file.h>
75 #include <sys/stat.h>
76 #include <sys/filedesc.h>
77 #include <sys/ioctl.h>
78 #include <sys/kernel.h>
79 #include <sys/malloc.h>
80 #include <sys/mbuf.h>
81 #include <sys/mman.h>
82 #include <sys/mount.h>
83 #include <sys/reboot.h>
84 #include <sys/resource.h>
85 #include <sys/resourcevar.h>
86 #include <sys/signal.h>
87 #include <sys/signalvar.h>
88 #include <sys/socket.h>
89 #include <sys/time.h>
90 #include <sys/times.h>
91 #include <sys/vnode.h>
92 #include <sys/uio.h>
93 #include <sys/wait.h>
94 #include <sys/utsname.h>
95 #include <sys/unistd.h>
96 #include <sys/swap.h>		/* for SWAP_ON */
97 #include <sys/sysctl.h>		/* for KERN_DOMAINNAME */
98 
99 #include <sys/ptrace.h>
100 #include <machine/ptrace.h>
101 
102 #include <sys/syscallargs.h>
103 
104 #include <compat/linux/common/linux_types.h>
105 #include <compat/linux/common/linux_signal.h>
106 
107 #include <compat/linux/linux_syscallargs.h>
108 
109 #include <compat/linux/common/linux_fcntl.h>
110 #include <compat/linux/common/linux_mmap.h>
111 #include <compat/linux/common/linux_dirent.h>
112 #include <compat/linux/common/linux_util.h>
113 #include <compat/linux/common/linux_misc.h>
114 #include <compat/linux/common/linux_ptrace.h>
115 #include <compat/linux/common/linux_reboot.h>
116 #include <compat/linux/common/linux_emuldata.h>
117 
118 const int linux_ptrace_request_map[] = {
119 	LINUX_PTRACE_TRACEME,	PT_TRACE_ME,
120 	LINUX_PTRACE_PEEKTEXT,	PT_READ_I,
121 	LINUX_PTRACE_PEEKDATA,	PT_READ_D,
122 	LINUX_PTRACE_POKETEXT,	PT_WRITE_I,
123 	LINUX_PTRACE_POKEDATA,	PT_WRITE_D,
124 	LINUX_PTRACE_CONT,	PT_CONTINUE,
125 	LINUX_PTRACE_KILL,	PT_KILL,
126 	LINUX_PTRACE_ATTACH,	PT_ATTACH,
127 	LINUX_PTRACE_DETACH,	PT_DETACH,
128 #ifdef PT_STEP
129 	LINUX_PTRACE_SINGLESTEP,	PT_STEP,
130 #endif
131 	-1
132 };
133 
134 /* Local linux_misc.c functions: */
135 static void bsd_to_linux_statfs __P((struct statfs *, struct linux_statfs *));
136 
137 /*
138  * The information on a terminated (or stopped) process needs
139  * to be converted in order for Linux binaries to get a valid signal
140  * number out of it.
141  */
142 void
143 bsd_to_linux_wstat(st)
144 	int *st;
145 {
146 
147 	int sig;
148 
149 	if (WIFSIGNALED(*st)) {
150 		sig = WTERMSIG(*st);
151 		if (sig >= 0 && sig < NSIG)
152 			*st= (*st& ~0177) | native_to_linux_sig[sig];
153 	} else if (WIFSTOPPED(*st)) {
154 		sig = WSTOPSIG(*st);
155 		if (sig >= 0 && sig < NSIG)
156 			*st = (*st & ~0xff00) | (native_to_linux_sig[sig] << 8);
157 	}
158 }
159 
160 /*
161  * This is very much the same as waitpid()
162  */
163 int
164 linux_sys_wait4(p, v, retval)
165 	struct proc *p;
166 	void *v;
167 	register_t *retval;
168 {
169 	struct linux_sys_wait4_args /* {
170 		syscallarg(int) pid;
171 		syscallarg(int *) status;
172 		syscallarg(int) options;
173 		syscallarg(struct rusage *) rusage;
174 	} */ *uap = v;
175 	struct sys_wait4_args w4a;
176 	int error, *status, tstat, options, linux_options;
177 	caddr_t sg;
178 
179 	if (SCARG(uap, status) != NULL) {
180 		sg = stackgap_init(p->p_emul);
181 		status = (int *) stackgap_alloc(&sg, sizeof *status);
182 	} else
183 		status = NULL;
184 
185 	linux_options = SCARG(uap, options);
186 	options = 0;
187 	if (linux_options &
188 	    ~(LINUX_WAIT4_WNOHANG|LINUX_WAIT4_WUNTRACED|LINUX_WAIT4_WALL|
189 	      LINUX_WAIT4_WCLONE))
190 		return (EINVAL);
191 
192 	if (linux_options & LINUX_WAIT4_WNOHANG)
193 		options |= WNOHANG;
194 	if (linux_options & LINUX_WAIT4_WUNTRACED)
195 		options |= WUNTRACED;
196 	if (linux_options & LINUX_WAIT4_WALL)
197 		options |= WALLSIG;
198 	if (linux_options & LINUX_WAIT4_WCLONE)
199 		options |= WALTSIG;
200 
201 	SCARG(&w4a, pid) = SCARG(uap, pid);
202 	SCARG(&w4a, status) = status;
203 	SCARG(&w4a, options) = options;
204 	SCARG(&w4a, rusage) = SCARG(uap, rusage);
205 
206 	if ((error = sys_wait4(p, &w4a, retval)))
207 		return error;
208 
209 	sigdelset(&p->p_sigctx.ps_siglist, SIGCHLD);
210 
211 	if (status != NULL) {
212 		if ((error = copyin(status, &tstat, sizeof tstat)))
213 			return error;
214 
215 		bsd_to_linux_wstat(&tstat);
216 		return copyout(&tstat, SCARG(uap, status), sizeof tstat);
217 	}
218 
219 	return 0;
220 }
221 
222 /*
223  * Linux brk(2). The check if the new address is >= the old one is
224  * done in the kernel in Linux. NetBSD does it in the library.
225  */
226 int
227 linux_sys_brk(p, v, retval)
228 	struct proc *p;
229 	void *v;
230 	register_t *retval;
231 {
232 	struct linux_sys_brk_args /* {
233 		syscallarg(char *) nsize;
234 	} */ *uap = v;
235 	char *nbrk = SCARG(uap, nsize);
236 	struct sys_obreak_args oba;
237 	struct vmspace *vm = p->p_vmspace;
238 	struct linux_emuldata *ed = (struct linux_emuldata*)p->p_emuldata;
239 
240 	SCARG(&oba, nsize) = nbrk;
241 
242 	if ((caddr_t) nbrk > vm->vm_daddr && sys_obreak(p, &oba, retval) == 0)
243 		ed->p_break = (char*)nbrk;
244 	else
245 		nbrk = ed->p_break;
246 
247 	retval[0] = (register_t)nbrk;
248 
249 	return 0;
250 }
251 
252 /*
253  * Convert BSD statfs structure to Linux statfs structure.
254  * The Linux structure has less fields, and it also wants
255  * the length of a name in a dir entry in a field, which
256  * we fake (probably the wrong way).
257  */
258 static void
259 bsd_to_linux_statfs(bsp, lsp)
260 	struct statfs *bsp;
261 	struct linux_statfs *lsp;
262 {
263 
264 	lsp->l_ftype = bsp->f_type;
265 	lsp->l_fbsize = bsp->f_bsize;
266 	lsp->l_fblocks = bsp->f_blocks;
267 	lsp->l_fbfree = bsp->f_bfree;
268 	lsp->l_fbavail = bsp->f_bavail;
269 	lsp->l_ffiles = bsp->f_files;
270 	lsp->l_fffree = bsp->f_ffree;
271 	lsp->l_ffsid.val[0] = bsp->f_fsid.val[0];
272 	lsp->l_ffsid.val[1] = bsp->f_fsid.val[1];
273 	lsp->l_fnamelen = MAXNAMLEN;	/* XXX */
274 }
275 
276 /*
277  * Implement the fs stat functions. Straightforward.
278  */
279 int
280 linux_sys_statfs(p, v, retval)
281 	struct proc *p;
282 	void *v;
283 	register_t *retval;
284 {
285 	struct linux_sys_statfs_args /* {
286 		syscallarg(const char *) path;
287 		syscallarg(struct linux_statfs *) sp;
288 	} */ *uap = v;
289 	struct statfs btmp, *bsp;
290 	struct linux_statfs ltmp;
291 	struct sys_statfs_args bsa;
292 	caddr_t sg;
293 	int error;
294 
295 	sg = stackgap_init(p->p_emul);
296 	bsp = (struct statfs *) stackgap_alloc(&sg, sizeof (struct statfs));
297 
298 	CHECK_ALT_EXIST(p, &sg, SCARG(uap, path));
299 
300 	SCARG(&bsa, path) = SCARG(uap, path);
301 	SCARG(&bsa, buf) = bsp;
302 
303 	if ((error = sys_statfs(p, &bsa, retval)))
304 		return error;
305 
306 	if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp)))
307 		return error;
308 
309 	bsd_to_linux_statfs(&btmp, &ltmp);
310 
311 	return copyout((caddr_t) &ltmp, (caddr_t) SCARG(uap, sp), sizeof ltmp);
312 }
313 
314 int
315 linux_sys_fstatfs(p, v, retval)
316 	struct proc *p;
317 	void *v;
318 	register_t *retval;
319 {
320 	struct linux_sys_fstatfs_args /* {
321 		syscallarg(int) fd;
322 		syscallarg(struct linux_statfs *) sp;
323 	} */ *uap = v;
324 	struct statfs btmp, *bsp;
325 	struct linux_statfs ltmp;
326 	struct sys_fstatfs_args bsa;
327 	caddr_t sg;
328 	int error;
329 
330 	sg = stackgap_init(p->p_emul);
331 	bsp = (struct statfs *) stackgap_alloc(&sg, sizeof (struct statfs));
332 
333 	SCARG(&bsa, fd) = SCARG(uap, fd);
334 	SCARG(&bsa, buf) = bsp;
335 
336 	if ((error = sys_fstatfs(p, &bsa, retval)))
337 		return error;
338 
339 	if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp)))
340 		return error;
341 
342 	bsd_to_linux_statfs(&btmp, &ltmp);
343 
344 	return copyout((caddr_t) &ltmp, (caddr_t) SCARG(uap, sp), sizeof ltmp);
345 }
346 
347 char linux_sysname[] = "Linux";
348 char linux_release[] = "2.0.38";
349 char linux_version[] = "#0 Sun Apr 1 11:11:11 MET 2000";
350 
351 /*
352  * uname(). Just copy the info from the various strings stored in the
353  * kernel, and put it in the Linux utsname structure. That structure
354  * is almost the same as the NetBSD one, only it has fields 65 characters
355  * long, and an extra domainname field.
356  */
357 int
358 linux_sys_uname(p, v, retval)
359 	struct proc *p;
360 	void *v;
361 	register_t *retval;
362 {
363 	struct linux_sys_uname_args /* {
364 		syscallarg(struct linux_utsname *) up;
365 	} */ *uap = v;
366 	struct linux_utsname luts;
367 
368 	strncpy(luts.l_sysname, linux_sysname, sizeof(luts.l_sysname));
369 	strncpy(luts.l_nodename, hostname, sizeof(luts.l_nodename));
370 	strncpy(luts.l_release, linux_release, sizeof(luts.l_release));
371 	strncpy(luts.l_version, linux_version, sizeof(luts.l_version));
372 	strncpy(luts.l_machine, machine, sizeof(luts.l_machine));
373 	strncpy(luts.l_domainname, domainname, sizeof(luts.l_domainname));
374 
375 	return copyout(&luts, SCARG(uap, up), sizeof(luts));
376 }
377 
378 /* Used directly on: alpha, mips, ppc, sparc, sparc64 */
379 /* Used indirectly on: arm, i386, m68k */
380 
381 /*
382  * New type Linux mmap call.
383  * Only called directly on machines with >= 6 free regs.
384  */
385 int
386 linux_sys_mmap(p, v, retval)
387 	struct proc *p;
388 	void *v;
389 	register_t *retval;
390 {
391 	struct linux_sys_mmap_args /* {
392 		syscallarg(unsigned long) addr;
393 		syscallarg(size_t) len;
394 		syscallarg(int) prot;
395 		syscallarg(int) flags;
396 		syscallarg(int) fd;
397 		syscallarg(linux_off_t) offset;
398 	} */ *uap = v;
399 	struct sys_mmap_args cma;
400 	int flags;
401 
402 	flags = 0;
403 	flags |= cvtto_bsd_mask(SCARG(uap,flags), LINUX_MAP_SHARED, MAP_SHARED);
404 	flags |= cvtto_bsd_mask(SCARG(uap,flags), LINUX_MAP_PRIVATE, MAP_PRIVATE);
405 	flags |= cvtto_bsd_mask(SCARG(uap,flags), LINUX_MAP_FIXED, MAP_FIXED);
406 	flags |= cvtto_bsd_mask(SCARG(uap,flags), LINUX_MAP_ANON, MAP_ANON);
407 	/* XXX XAX ERH: Any other flags here?  There are more defined... */
408 
409 	SCARG(&cma,addr) = (void *)SCARG(uap, addr);
410 	SCARG(&cma,len) = SCARG(uap, len);
411 	SCARG(&cma,prot) = SCARG(uap, prot);
412 	if (SCARG(&cma,prot) & VM_PROT_WRITE) /* XXX */
413 		SCARG(&cma,prot) |= VM_PROT_READ;
414 	SCARG(&cma,flags) = flags;
415 	SCARG(&cma,fd) = flags & MAP_ANON ? -1 : SCARG(uap, fd);
416 	SCARG(&cma,pad) = 0;
417 	SCARG(&cma,pos) = (off_t)SCARG(uap, offset);
418 
419 	return sys_mmap(p, &cma, retval);
420 }
421 
422 int
423 linux_sys_mremap(p, v, retval)
424 	struct proc *p;
425 	void *v;
426 	register_t *retval;
427 {
428 	struct linux_sys_mremap_args /* {
429 		syscallarg(void *) old_address;
430 		syscallarg(size_t) old_size;
431 		syscallarg(size_t) new_size;
432 		syscallarg(u_long) flags;
433 	} */ *uap = v;
434 	struct sys_munmap_args mua;
435 	size_t old_size, new_size;
436 	int error;
437 
438 	old_size = round_page(SCARG(uap, old_size));
439 	new_size = round_page(SCARG(uap, new_size));
440 
441 	/*
442 	 * Growing mapped region.
443 	 */
444 	if (new_size > old_size) {
445 		/*
446 		 * XXX Implement me.  What we probably want to do is
447 		 * XXX dig out the guts of the old mapping, mmap that
448 		 * XXX object again with the new size, then munmap
449 		 * XXX the old mapping.
450 		 */
451 		*retval = 0;
452 		return (ENOMEM);
453 	}
454 
455 	/*
456 	 * Shrinking mapped region.
457 	 */
458 	if (new_size < old_size) {
459 		SCARG(&mua, addr) = (caddr_t)SCARG(uap, old_address) +
460 		    new_size;
461 		SCARG(&mua, len) = old_size - new_size;
462 		error = sys_munmap(p, &mua, retval);
463 		*retval = error ? 0 : (register_t)SCARG(uap, old_address);
464 		return (error);
465 	}
466 
467 	/*
468 	 * No change.
469 	 */
470 	*retval = (register_t)SCARG(uap, old_address);
471 	return (0);
472 }
473 
474 int
475 linux_sys_msync(p, v, retval)
476 	struct proc *p;
477 	void *v;
478 	register_t *retval;
479 {
480 	struct linux_sys_msync_args /* {
481 		syscallarg(caddr_t) addr;
482 		syscallarg(int) len;
483 		syscallarg(int) fl;
484 	} */ *uap = v;
485 
486 	struct sys___msync13_args bma;
487 
488 	/* flags are ignored */
489 	SCARG(&bma, addr) = SCARG(uap, addr);
490 	SCARG(&bma, len) = SCARG(uap, len);
491 	SCARG(&bma, flags) = SCARG(uap, fl);
492 
493 	return sys___msync13(p, &bma, retval);
494 }
495 
496 /*
497  * This code is partly stolen from src/lib/libc/compat-43/times.c
498  * XXX - CLK_TCK isn't declared in /sys, just in <time.h>, done here
499  */
500 
501 #define CLK_TCK 100
502 #define	CONVTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
503 
504 int
505 linux_sys_times(p, v, retval)
506 	struct proc *p;
507 	void *v;
508 	register_t *retval;
509 {
510 	struct linux_sys_times_args /* {
511 		syscallarg(struct times *) tms;
512 	} */ *uap = v;
513 	struct timeval t;
514 	struct linux_tms ltms;
515 	struct rusage ru;
516 	int error, s;
517 
518 	calcru(p, &ru.ru_utime, &ru.ru_stime, NULL);
519 	ltms.ltms_utime = CONVTCK(ru.ru_utime);
520 	ltms.ltms_stime = CONVTCK(ru.ru_stime);
521 
522 	ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime);
523 	ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime);
524 
525 	if ((error = copyout(&ltms, SCARG(uap, tms), sizeof ltms)))
526 		return error;
527 
528 	s = splclock();
529 	timersub(&time, &boottime, &t);
530 	splx(s);
531 
532 	retval[0] = ((linux_clock_t)(CONVTCK(t)));
533 	return 0;
534 }
535 
536 /*
537  * Linux 'readdir' call. This code is mostly taken from the
538  * SunOS getdents call (see compat/sunos/sunos_misc.c), though
539  * an attempt has been made to keep it a little cleaner (failing
540  * miserably, because of the cruft needed if count 1 is passed).
541  *
542  * The d_off field should contain the offset of the next valid entry,
543  * but in Linux it has the offset of the entry itself. We emulate
544  * that bug here.
545  *
546  * Read in BSD-style entries, convert them, and copy them out.
547  *
548  * Note that this doesn't handle union-mounted filesystems.
549  */
550 int
551 linux_sys_getdents(p, v, retval)
552 	struct proc *p;
553 	void *v;
554 	register_t *retval;
555 {
556 	struct linux_sys_getdents_args /* {
557 		syscallarg(int) fd;
558 		syscallarg(struct linux_dirent *) dent;
559 		syscallarg(unsigned int) count;
560 	} */ *uap = v;
561 	struct dirent *bdp;
562 	struct vnode *vp;
563 	caddr_t	inp, buf;		/* BSD-format */
564 	int len, reclen;		/* BSD-format */
565 	caddr_t outp;			/* Linux-format */
566 	int resid, linux_reclen = 0;	/* Linux-format */
567 	struct file *fp;
568 	struct uio auio;
569 	struct iovec aiov;
570 	struct linux_dirent idb;
571 	off_t off;		/* true file offset */
572 	int buflen, error, eofflag, nbytes, oldcall;
573 	struct vattr va;
574 	off_t *cookiebuf = NULL, *cookie;
575 	int ncookies;
576 
577 	/* getvnode() will use the descriptor for us */
578 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
579 		return (error);
580 
581 	if ((fp->f_flag & FREAD) == 0) {
582 		error = EBADF;
583 		goto out1;
584 	}
585 
586 	vp = (struct vnode *)fp->f_data;
587 	if (vp->v_type != VDIR) {
588 		error = EINVAL;
589 		goto out1;
590 	}
591 
592 	if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)))
593 		goto out1;
594 
595 	nbytes = SCARG(uap, count);
596 	if (nbytes == 1) {	/* emulating old, broken behaviour */
597 		nbytes = sizeof (struct linux_dirent);
598 		buflen = max(va.va_blocksize, nbytes);
599 		oldcall = 1;
600 	} else {
601 		buflen = min(MAXBSIZE, nbytes);
602 		if (buflen < va.va_blocksize)
603 			buflen = va.va_blocksize;
604 		oldcall = 0;
605 	}
606 	buf = malloc(buflen, M_TEMP, M_WAITOK);
607 
608 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
609 	off = fp->f_offset;
610 again:
611 	aiov.iov_base = buf;
612 	aiov.iov_len = buflen;
613 	auio.uio_iov = &aiov;
614 	auio.uio_iovcnt = 1;
615 	auio.uio_rw = UIO_READ;
616 	auio.uio_segflg = UIO_SYSSPACE;
617 	auio.uio_procp = p;
618 	auio.uio_resid = buflen;
619 	auio.uio_offset = off;
620 	/*
621          * First we read into the malloc'ed buffer, then
622          * we massage it into user space, one record at a time.
623          */
624 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf,
625 	    &ncookies);
626 	if (error)
627 		goto out;
628 
629 	inp = buf;
630 	outp = (caddr_t)SCARG(uap, dent);
631 	resid = nbytes;
632 	if ((len = buflen - auio.uio_resid) == 0)
633 		goto eof;
634 
635 	for (cookie = cookiebuf; len > 0; len -= reclen) {
636 		bdp = (struct dirent *)inp;
637 		reclen = bdp->d_reclen;
638 		if (reclen & 3)
639 			panic("linux_readdir");
640 		if (bdp->d_fileno == 0) {
641 			inp += reclen;	/* it is a hole; squish it out */
642 			off = *cookie++;
643 			continue;
644 		}
645 		linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen);
646 		if (reclen > len || resid < linux_reclen) {
647 			/* entry too big for buffer, so just stop */
648 			outp++;
649 			break;
650 		}
651 		/*
652 		 * Massage in place to make a Linux-shaped dirent (otherwise
653 		 * we have to worry about touching user memory outside of
654 		 * the copyout() call).
655 		 */
656 		idb.d_ino = (linux_ino_t)bdp->d_fileno;
657 		/*
658 		 * The old readdir() call misuses the offset and reclen fields.
659 		 */
660 		if (oldcall) {
661 			idb.d_off = (linux_off_t)linux_reclen;
662 			idb.d_reclen = (u_short)bdp->d_namlen;
663 		} else {
664 			if (sizeof (linux_off_t) < 4 && (off >> 32) != 0) {
665 				compat_offseterr(vp, "linux_getdents");
666 				error = EINVAL;
667 				goto out;
668 			}
669 			idb.d_off = (linux_off_t)off;
670 			idb.d_reclen = (u_short)linux_reclen;
671 		}
672 		strcpy(idb.d_name, bdp->d_name);
673 		if ((error = copyout((caddr_t)&idb, outp, linux_reclen)))
674 			goto out;
675 		/* advance past this real entry */
676 		inp += reclen;
677 		off = *cookie++;	/* each entry points to itself */
678 		/* advance output past Linux-shaped entry */
679 		outp += linux_reclen;
680 		resid -= linux_reclen;
681 		if (oldcall)
682 			break;
683 	}
684 
685 	/* if we squished out the whole block, try again */
686 	if (outp == (caddr_t)SCARG(uap, dent))
687 		goto again;
688 	fp->f_offset = off;	/* update the vnode offset */
689 
690 	if (oldcall)
691 		nbytes = resid + linux_reclen;
692 
693 eof:
694 	*retval = nbytes - resid;
695 out:
696 	VOP_UNLOCK(vp, 0);
697 	if (cookiebuf)
698 		free(cookiebuf, M_TEMP);
699 	free(buf, M_TEMP);
700  out1:
701 	FILE_UNUSE(fp, p);
702 	return error;
703 }
704 
705 /*
706  * Even when just using registers to pass arguments to syscalls you can
707  * have 5 of them on the i386. So this newer version of select() does
708  * this.
709  */
710 int
711 linux_sys_select(p, v, retval)
712 	struct proc *p;
713 	void *v;
714 	register_t *retval;
715 {
716 	struct linux_sys_select_args /* {
717 		syscallarg(int) nfds;
718 		syscallarg(fd_set *) readfds;
719 		syscallarg(fd_set *) writefds;
720 		syscallarg(fd_set *) exceptfds;
721 		syscallarg(struct timeval *) timeout;
722 	} */ *uap = v;
723 
724 	return linux_select1(p, retval, SCARG(uap, nfds), SCARG(uap, readfds),
725 	    SCARG(uap, writefds), SCARG(uap, exceptfds), SCARG(uap, timeout));
726 }
727 
728 /*
729  * Common code for the old and new versions of select(). A couple of
730  * things are important:
731  * 1) return the amount of time left in the 'timeout' parameter
732  * 2) select never returns ERESTART on Linux, always return EINTR
733  */
734 int
735 linux_select1(p, retval, nfds, readfds, writefds, exceptfds, timeout)
736 	struct proc *p;
737 	register_t *retval;
738 	int nfds;
739 	fd_set *readfds, *writefds, *exceptfds;
740 	struct timeval *timeout;
741 {
742 	struct sys_select_args bsa;
743 	struct timeval tv0, tv1, utv, *tvp;
744 	caddr_t sg;
745 	int error;
746 
747 	SCARG(&bsa, nd) = nfds;
748 	SCARG(&bsa, in) = readfds;
749 	SCARG(&bsa, ou) = writefds;
750 	SCARG(&bsa, ex) = exceptfds;
751 	SCARG(&bsa, tv) = timeout;
752 
753 	/*
754 	 * Store current time for computation of the amount of
755 	 * time left.
756 	 */
757 	if (timeout) {
758 		if ((error = copyin(timeout, &utv, sizeof(utv))))
759 			return error;
760 		if (itimerfix(&utv)) {
761 			/*
762 			 * The timeval was invalid.  Convert it to something
763 			 * valid that will act as it does under Linux.
764 			 */
765 			sg = stackgap_init(p->p_emul);
766 			tvp = stackgap_alloc(&sg, sizeof(utv));
767 			utv.tv_sec += utv.tv_usec / 1000000;
768 			utv.tv_usec %= 1000000;
769 			if (utv.tv_usec < 0) {
770 				utv.tv_sec -= 1;
771 				utv.tv_usec += 1000000;
772 			}
773 			if (utv.tv_sec < 0)
774 				timerclear(&utv);
775 			if ((error = copyout(&utv, tvp, sizeof(utv))))
776 				return error;
777 			SCARG(&bsa, tv) = tvp;
778 		}
779 		microtime(&tv0);
780 	}
781 
782 	error = sys_select(p, &bsa, retval);
783 	if (error) {
784 		/*
785 		 * See fs/select.c in the Linux kernel.  Without this,
786 		 * Maelstrom doesn't work.
787 		 */
788 		if (error == ERESTART)
789 			error = EINTR;
790 		return error;
791 	}
792 
793 	if (timeout) {
794 		if (*retval) {
795 			/*
796 			 * Compute how much time was left of the timeout,
797 			 * by subtracting the current time and the time
798 			 * before we started the call, and subtracting
799 			 * that result from the user-supplied value.
800 			 */
801 			microtime(&tv1);
802 			timersub(&tv1, &tv0, &tv1);
803 			timersub(&utv, &tv1, &utv);
804 			if (utv.tv_sec < 0)
805 				timerclear(&utv);
806 		} else
807 			timerclear(&utv);
808 		if ((error = copyout(&utv, timeout, sizeof(utv))))
809 			return error;
810 	}
811 
812 	return 0;
813 }
814 
815 /*
816  * Get the process group of a certain process. Look it up
817  * and return the value.
818  */
819 int
820 linux_sys_getpgid(p, v, retval)
821 	struct proc *p;
822 	void *v;
823 	register_t *retval;
824 {
825 	struct linux_sys_getpgid_args /* {
826 		syscallarg(int) pid;
827 	} */ *uap = v;
828 	struct proc *targp;
829 
830 	if (SCARG(uap, pid) != 0 && SCARG(uap, pid) != p->p_pid) {
831 		if ((targp = pfind(SCARG(uap, pid))) == 0)
832 			return ESRCH;
833 	}
834 	else
835 		targp = p;
836 
837 	retval[0] = targp->p_pgid;
838 	return 0;
839 }
840 
841 /*
842  * Set the 'personality' (emulation mode) for the current process. Only
843  * accept the Linux personality here (0). This call is needed because
844  * the Linux ELF crt0 issues it in an ugly kludge to make sure that
845  * ELF binaries run in Linux mode, not SVR4 mode.
846  */
847 int
848 linux_sys_personality(p, v, retval)
849 	struct proc *p;
850 	void *v;
851 	register_t *retval;
852 {
853 	struct linux_sys_personality_args /* {
854 		syscallarg(int) per;
855 	} */ *uap = v;
856 
857 	if (SCARG(uap, per) != 0)
858 		return EINVAL;
859 	retval[0] = 0;
860 	return 0;
861 }
862 
863 #if defined(__i386__) || defined(__m68k__)
864 /*
865  * The calls are here because of type conversions.
866  */
867 int
868 linux_sys_setreuid16(p, v, retval)
869 	struct proc *p;
870 	void *v;
871 	register_t *retval;
872 {
873 	struct linux_sys_setreuid16_args /* {
874 		syscallarg(int) ruid;
875 		syscallarg(int) euid;
876 	} */ *uap = v;
877 	struct sys_setreuid_args bsa;
878 
879 	SCARG(&bsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
880 		(uid_t)-1 : SCARG(uap, ruid);
881 	SCARG(&bsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
882 		(uid_t)-1 : SCARG(uap, euid);
883 
884 	return sys_setreuid(p, &bsa, retval);
885 }
886 
887 int
888 linux_sys_setregid16(p, v, retval)
889 	struct proc *p;
890 	void *v;
891 	register_t *retval;
892 {
893 	struct linux_sys_setregid16_args /* {
894 		syscallarg(int) rgid;
895 		syscallarg(int) egid;
896 	} */ *uap = v;
897 	struct sys_setregid_args bsa;
898 
899 	SCARG(&bsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
900 		(uid_t)-1 : SCARG(uap, rgid);
901 	SCARG(&bsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
902 		(uid_t)-1 : SCARG(uap, egid);
903 
904 	return sys_setregid(p, &bsa, retval);
905 }
906 
907 int
908 linux_sys_setresuid16(p, v, retval)
909 	struct proc *p;
910 	void *v;
911 	register_t *retval;
912 {
913 	struct linux_sys_setresuid16_args /* {
914 		syscallarg(uid_t) ruid;
915 		syscallarg(uid_t) euid;
916 		syscallarg(uid_t) suid;
917 	} */ *uap = v;
918 	struct linux_sys_setresuid16_args lsa;
919 
920 	SCARG(&lsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
921 		(uid_t)-1 : SCARG(uap, ruid);
922 	SCARG(&lsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
923 		(uid_t)-1 : SCARG(uap, euid);
924 	SCARG(&lsa, suid) = ((linux_uid_t)SCARG(uap, suid) == (linux_uid_t)-1) ?
925 		(uid_t)-1 : SCARG(uap, suid);
926 
927 	return linux_sys_setresuid(p, &lsa, retval);
928 }
929 
930 int
931 linux_sys_setresgid16(p, v, retval)
932 	struct proc *p;
933 	void *v;
934 	register_t *retval;
935 {
936 	struct linux_sys_setresgid16_args /* {
937 		syscallarg(gid_t) rgid;
938 		syscallarg(gid_t) egid;
939 		syscallarg(gid_t) sgid;
940 	} */ *uap = v;
941 	struct linux_sys_setresgid16_args lsa;
942 
943 	SCARG(&lsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
944 		(gid_t)-1 : SCARG(uap, rgid);
945 	SCARG(&lsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
946 		(gid_t)-1 : SCARG(uap, egid);
947 	SCARG(&lsa, sgid) = ((linux_gid_t)SCARG(uap, sgid) == (linux_gid_t)-1) ?
948 		(gid_t)-1 : SCARG(uap, sgid);
949 
950 	return linux_sys_setresgid(p, &lsa, retval);
951 }
952 
953 int
954 linux_sys_getgroups16(p, v, retval)
955 	struct proc *p;
956 	void *v;
957 	register_t *retval;
958 {
959 	struct linux_sys_getgroups16_args /* {
960 		syscallarg(int) gidsetsize;
961 		syscallarg(linux_gid_t *) gidset;
962 	} */ *uap = v;
963 	caddr_t sg;
964 	int n, error, i;
965 	struct sys_getgroups_args bsa;
966 	gid_t *bset, *kbset;
967 	linux_gid_t *lset;
968 	struct pcred *pc = p->p_cred;
969 
970 	n = SCARG(uap, gidsetsize);
971 	if (n < 0)
972 		return EINVAL;
973 	error = 0;
974 	bset = kbset = NULL;
975 	lset = NULL;
976 	if (n > 0) {
977 		n = min(pc->pc_ucred->cr_ngroups, n);
978 		sg = stackgap_init(p->p_emul);
979 		bset = stackgap_alloc(&sg, n * sizeof (gid_t));
980 		kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK);
981 		lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
982 		if (bset == NULL || kbset == NULL || lset == NULL)
983 			return ENOMEM;
984 		SCARG(&bsa, gidsetsize) = n;
985 		SCARG(&bsa, gidset) = bset;
986 		error = sys_getgroups(p, &bsa, retval);
987 		if (error != 0)
988 			goto out;
989 		error = copyin(bset, kbset, n * sizeof (gid_t));
990 		if (error != 0)
991 			goto out;
992 		for (i = 0; i < n; i++)
993 			lset[i] = (linux_gid_t)kbset[i];
994 		error = copyout(lset, SCARG(uap, gidset),
995 		    n * sizeof (linux_gid_t));
996 	} else
997 		*retval = pc->pc_ucred->cr_ngroups;
998 out:
999 	if (kbset != NULL)
1000 		free(kbset, M_TEMP);
1001 	if (lset != NULL)
1002 		free(lset, M_TEMP);
1003 	return error;
1004 }
1005 
1006 int
1007 linux_sys_setgroups16(p, v, retval)
1008 	struct proc *p;
1009 	void *v;
1010 	register_t *retval;
1011 {
1012 	struct linux_sys_setgroups16_args /* {
1013 		syscallarg(int) gidsetsize;
1014 		syscallarg(linux_gid_t *) gidset;
1015 	} */ *uap = v;
1016 	caddr_t sg;
1017 	int n;
1018 	int error, i;
1019 	struct sys_setgroups_args bsa;
1020 	gid_t *bset, *kbset;
1021 	linux_gid_t *lset;
1022 
1023 	n = SCARG(uap, gidsetsize);
1024 	if (n < 0 || n > NGROUPS)
1025 		return EINVAL;
1026 	sg = stackgap_init(p->p_emul);
1027 	bset = stackgap_alloc(&sg, n * sizeof (gid_t));
1028 	lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
1029 	kbset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
1030 	if (lset == NULL || bset == NULL)
1031 		return ENOMEM;
1032 	error = copyin(SCARG(uap, gidset), lset, n * sizeof (linux_gid_t));
1033 	if (error != 0)
1034 		goto out;
1035 	for (i = 0; i < n; i++)
1036 		kbset[i] = (gid_t)lset[i];
1037 	error = copyout(kbset, bset, n * sizeof (gid_t));
1038 	if (error != 0)
1039 		goto out;
1040 	SCARG(&bsa, gidsetsize) = n;
1041 	SCARG(&bsa, gidset) = bset;
1042 	error = sys_setgroups(p, &bsa, retval);
1043 
1044 out:
1045 	if (lset != NULL)
1046 		free(lset, M_TEMP);
1047 	if (kbset != NULL)
1048 		free(kbset, M_TEMP);
1049 
1050 	return error;
1051 }
1052 
1053 #endif /* __i386__ || __m68k__ */
1054 
1055 /*
1056  * We have nonexistent fsuid equal to uid.
1057  * If modification is requested, refuse.
1058  */
1059 int
1060 linux_sys_setfsuid(p, v, retval)
1061 	 struct proc *p;
1062 	 void *v;
1063 	 register_t *retval;
1064 {
1065 	 struct linux_sys_setfsuid_args /* {
1066 		 syscallarg(uid_t) uid;
1067 	 } */ *uap = v;
1068 	 uid_t uid;
1069 
1070 	 uid = SCARG(uap, uid);
1071 	 if (p->p_cred->p_ruid != uid)
1072 		 return sys_nosys(p, v, retval);
1073 	 else
1074 		 return (0);
1075 }
1076 
1077 /* XXX XXX XXX */
1078 #ifndef alpha
1079 int
1080 linux_sys_getfsuid(p, v, retval)
1081 	struct proc *p;
1082 	void *v;
1083 	register_t *retval;
1084 {
1085 	return sys_getuid(p, v, retval);
1086 }
1087 #endif
1088 
1089 int
1090 linux_sys___sysctl(p, v, retval)
1091 	struct proc *p;
1092 	void *v;
1093 	register_t *retval;
1094 {
1095 	struct linux_sys___sysctl_args /* {
1096 		syscallarg(struct linux___sysctl *) lsp;
1097 	} */ *uap = v;
1098 	struct linux___sysctl ls;
1099 	struct sys___sysctl_args bsa;
1100 	int error;
1101 
1102 	if ((error = copyin(SCARG(uap, lsp), &ls, sizeof ls)))
1103 		return error;
1104 	SCARG(&bsa, name) = ls.name;
1105 	SCARG(&bsa, namelen) = ls.namelen;
1106 	SCARG(&bsa, old) = ls.old;
1107 	SCARG(&bsa, oldlenp) = ls.oldlenp;
1108 	SCARG(&bsa, new) = ls.new;
1109 	SCARG(&bsa, newlen) = ls.newlen;
1110 
1111 	return sys___sysctl(p, &bsa, retval);
1112 }
1113 
1114 int
1115 linux_sys_setresuid(p, v, retval)
1116 	struct proc *p;
1117 	void *v;
1118 	register_t *retval;
1119 {
1120 	struct linux_sys_setresuid_args /* {
1121 		syscallarg(uid_t) ruid;
1122 		syscallarg(uid_t) euid;
1123 		syscallarg(uid_t) suid;
1124 	} */ *uap = v;
1125 	struct pcred *pc = p->p_cred;
1126 	uid_t ruid, euid, suid;
1127 	int error;
1128 
1129 	ruid = SCARG(uap, ruid);
1130 	euid = SCARG(uap, euid);
1131 	suid = SCARG(uap, suid);
1132 
1133 	/*
1134 	 * Note: These checks are a little different than the NetBSD
1135 	 * setreuid(2) call performs.  This precisely follows the
1136 	 * behavior of the Linux kernel.
1137 	 */
1138 	if (ruid != (uid_t)-1 &&
1139 	    ruid != pc->p_ruid &&
1140 	    ruid != pc->pc_ucred->cr_uid &&
1141 	    ruid != pc->p_svuid &&
1142 	    (error = suser(pc->pc_ucred, &p->p_acflag)))
1143 		return (error);
1144 
1145 	if (euid != (uid_t)-1 &&
1146 	    euid != pc->p_ruid &&
1147 	    euid != pc->pc_ucred->cr_uid &&
1148 	    euid != pc->p_svuid &&
1149 	    (error = suser(pc->pc_ucred, &p->p_acflag)))
1150 		return (error);
1151 
1152 	if (suid != (uid_t)-1 &&
1153 	    suid != pc->p_ruid &&
1154 	    suid != pc->pc_ucred->cr_uid &&
1155 	    suid != pc->p_svuid &&
1156 	    (error = suser(pc->pc_ucred, &p->p_acflag)))
1157 		return (error);
1158 
1159 	/*
1160 	 * Now assign the new real, effective, and saved UIDs.
1161 	 * Note that Linux, unlike NetBSD in setreuid(2), does not
1162 	 * set the saved UID in this call unless the user specifies
1163 	 * it.
1164 	 */
1165 	if (ruid != (uid_t)-1) {
1166 		(void)chgproccnt(pc->p_ruid, -1);
1167 		(void)chgproccnt(ruid, 1);
1168 		pc->p_ruid = ruid;
1169 	}
1170 
1171 	if (euid != (uid_t)-1) {
1172 		pc->pc_ucred = crcopy(pc->pc_ucred);
1173 		pc->pc_ucred->cr_uid = euid;
1174 	}
1175 
1176 	if (suid != (uid_t)-1)
1177 		pc->p_svuid = suid;
1178 
1179 	if (ruid != (uid_t)-1 && euid != (uid_t)-1 && suid != (uid_t)-1)
1180 		p->p_flag |= P_SUGID;
1181 	return (0);
1182 }
1183 
1184 int
1185 linux_sys_getresuid(p, v, retval)
1186 	struct proc *p;
1187 	void *v;
1188 	register_t *retval;
1189 {
1190 	struct linux_sys_getresuid_args /* {
1191 		syscallarg(uid_t *) ruid;
1192 		syscallarg(uid_t *) euid;
1193 		syscallarg(uid_t *) suid;
1194 	} */ *uap = v;
1195 	struct pcred *pc = p->p_cred;
1196 	int error;
1197 
1198 	/*
1199 	 * Linux copies these values out to userspace like so:
1200 	 *
1201 	 *	1. Copy out ruid.
1202 	 *	2. If that succeeds, copy out euid.
1203 	 *	3. If both of those succeed, copy out suid.
1204 	 */
1205 	if ((error = copyout(&pc->p_ruid, SCARG(uap, ruid),
1206 			     sizeof(uid_t))) != 0)
1207 		return (error);
1208 
1209 	if ((error = copyout(&pc->pc_ucred->cr_uid, SCARG(uap, euid),
1210 			     sizeof(uid_t))) != 0)
1211 		return (error);
1212 
1213 	return (copyout(&pc->p_svuid, SCARG(uap, suid), sizeof(uid_t)));
1214 }
1215 
1216 int
1217 linux_sys_ptrace(p, v, retval)
1218 	struct proc *p;
1219 	void *v;
1220 	register_t *retval;
1221 {
1222 	struct linux_sys_ptrace_args /* {
1223 		i386, m68k, powerpc: T=int
1224 		alpha: T=long
1225 		syscallarg(T) request;
1226 		syscallarg(T) pid;
1227 		syscallarg(T) addr;
1228 		syscallarg(T) data;
1229 	} */ *uap = v;
1230 	const int *ptr;
1231 	int request;
1232 	int error;
1233 
1234 	ptr = linux_ptrace_request_map;
1235 	request = SCARG(uap, request);
1236 	while (*ptr != -1)
1237 		if (*ptr++ == request) {
1238 			struct sys_ptrace_args pta;
1239 
1240 			SCARG(&pta, req) = *ptr;
1241 			SCARG(&pta, pid) = SCARG(uap, pid);
1242 			SCARG(&pta, addr) = (caddr_t)SCARG(uap, addr);
1243 			SCARG(&pta, data) = SCARG(uap, data);
1244 
1245 			/*
1246 			 * Linux ptrace(PTRACE_CONT, pid, 0, 0) means actually
1247 			 * to continue where the process left off previously.
1248 			 * The same thing is achieved by addr == (caddr_t) 1
1249 			 * on NetBSD, so rewrite 'addr' appropriately.
1250 			 */
1251 			if (request == LINUX_PTRACE_CONT && SCARG(uap, addr)==0)
1252 				SCARG(&pta, addr) = (caddr_t) 1;
1253 
1254 			error = sys_ptrace(p, &pta, retval);
1255 			if (error)
1256 				return error;
1257 			switch (request) {
1258 			case LINUX_PTRACE_PEEKTEXT:
1259 			case LINUX_PTRACE_PEEKDATA:
1260 				error = copyout (retval,
1261 				    (caddr_t)SCARG(uap, data), sizeof *retval);
1262 				*retval = SCARG(uap, data);
1263 				break;
1264 			default:
1265 				break;
1266 			}
1267 			return error;
1268 		}
1269 		else
1270 			ptr++;
1271 
1272 	return LINUX_SYS_PTRACE_ARCH(p, uap, retval);
1273 }
1274 
1275 int
1276 linux_sys_reboot(struct proc *p, void *v, register_t *retval)
1277 {
1278 	struct linux_sys_reboot_args /* {
1279 		syscallarg(int) magic1;
1280 		syscallarg(int) magic2;
1281 		syscallarg(int) cmd;
1282 		syscallarg(void *) arg;
1283 	} */ *uap = v;
1284 	struct sys_reboot_args /* {
1285 		syscallarg(int) opt;
1286 		syscallarg(char *) bootstr;
1287 	} */ sra;
1288 	int error;
1289 
1290 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1291 		return(error);
1292 
1293 	if (SCARG(uap, magic1) != LINUX_REBOOT_MAGIC1)
1294 		return(EINVAL);
1295 	if (SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2 &&
1296 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2A &&
1297 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2B)
1298 		return(EINVAL);
1299 
1300 	switch (SCARG(uap, cmd)) {
1301 	case LINUX_REBOOT_CMD_RESTART:
1302 		SCARG(&sra, opt) = RB_AUTOBOOT;
1303 		break;
1304 	case LINUX_REBOOT_CMD_HALT:
1305 		SCARG(&sra, opt) = RB_HALT;
1306 		break;
1307 	case LINUX_REBOOT_CMD_POWER_OFF:
1308 		SCARG(&sra, opt) = RB_HALT|RB_POWERDOWN;
1309 		break;
1310 	case LINUX_REBOOT_CMD_RESTART2:
1311 		/* Reboot with an argument. */
1312 		SCARG(&sra, opt) = RB_AUTOBOOT|RB_STRING;
1313 		SCARG(&sra, bootstr) = SCARG(uap, arg);
1314 		break;
1315 	case LINUX_REBOOT_CMD_CAD_ON:
1316 		return(EINVAL);	/* We don't implement ctrl-alt-delete */
1317 	case LINUX_REBOOT_CMD_CAD_OFF:
1318 		return(0);
1319 	default:
1320 		return(EINVAL);
1321 	}
1322 
1323 	return(sys_reboot(p, &sra, retval));
1324 }
1325 
1326 /*
1327  * Copy of compat_12_sys_swapon().
1328  */
1329 int
1330 linux_sys_swapon(p, v, retval)
1331 	struct proc *p;
1332 	void *v;
1333 	register_t *retval;
1334 {
1335 	struct sys_swapctl_args ua;
1336 	struct linux_sys_swapon_args /* {
1337 		syscallarg(const char *) name;
1338 	} */ *uap = v;
1339 
1340 	SCARG(&ua, cmd) = SWAP_ON;
1341 	SCARG(&ua, arg) = (void *)SCARG(uap, name);
1342 	SCARG(&ua, misc) = 0;	/* priority */
1343 	return (sys_swapctl(p, &ua, retval));
1344 }
1345 
1346 /*
1347  * Stop swapping to the file or block device specified by path.
1348  */
1349 int
1350 linux_sys_swapoff(p, v, retval)
1351 	struct proc *p;
1352 	void *v;
1353 	register_t *retval;
1354 {
1355 	struct sys_swapctl_args ua;
1356 	struct linux_sys_swapoff_args /* {
1357 		syscallarg(const char *) path;
1358 	} */ *uap = v;
1359 
1360 	SCARG(&ua, cmd) = SWAP_OFF;
1361 	SCARG(&ua, arg) = (void *)SCARG(uap, path);
1362 	return (sys_swapctl(p, &ua, retval));
1363 }
1364 
1365 /*
1366  * Copy of compat_09_sys_setdomainname()
1367  */
1368 /* ARGSUSED */
1369 int
1370 linux_sys_setdomainname(p, v, retval)
1371 	struct proc *p;
1372 	void *v;
1373 	register_t *retval;
1374 {
1375 	struct linux_sys_setdomainname_args /* {
1376 		syscallarg(char *) domainname;
1377 		syscallarg(int) len;
1378 	} */ *uap = v;
1379 	int name;
1380 	int error;
1381 
1382 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1383 		return (error);
1384 	name = KERN_DOMAINNAME;
1385 	return (kern_sysctl(&name, 1, 0, 0, SCARG(uap, domainname),
1386 			    SCARG(uap, len), p));
1387 }
1388 
1389 /*
1390  * sysinfo()
1391  */
1392 /* ARGSUSED */
1393 int
1394 linux_sys_sysinfo(p, v, retval)
1395 	struct proc *p;
1396 	void *v;
1397 	register_t *retval;
1398 {
1399 	struct linux_sys_sysinfo_args /* {
1400 		syscallarg(struct linux_sysinfo *) arg;
1401 	} */ *uap = v;
1402 	struct linux_sysinfo si;
1403 	struct loadavg *la;
1404 
1405 	si.uptime = time.tv_sec - boottime.tv_sec;
1406 	la = &averunnable;
1407 	si.loads[0] = la->ldavg[0] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1408 	si.loads[1] = la->ldavg[1] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1409 	si.loads[2] = la->ldavg[2] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1410 	si.totalram = ctob(physmem);
1411 	si.freeram = uvmexp.free * uvmexp.pagesize;
1412 	si.sharedram = 0;	/* XXX */
1413 	si.bufferram = uvmexp.filepages * uvmexp.pagesize;
1414 	si.totalswap = uvmexp.swpages * uvmexp.pagesize;
1415 	si.freeswap = (uvmexp.swpages - uvmexp.swpginuse) * uvmexp.pagesize;
1416 	si.procs = nprocs;
1417 
1418 	/* The following are only present in newer Linux kernels. */
1419 	si.totalbig = 0;
1420 	si.freebig = 0;
1421 	si.mem_unit = 1;
1422 
1423 	return (copyout(&si, SCARG(uap, arg), sizeof si));
1424 }
1425 
1426 /*
1427  * This gets called for unsupported syscalls. The difference to sys_nosys()
1428  * is that process does not get SIGSYS, the call just returns with ENOSYS.
1429  * This is the way Linux does it and glibc depends on this behaviour.
1430  */
1431 int
1432 linux_sys_nosys(p, v, retval)
1433 	struct proc *p;
1434 	void *v;
1435 	register_t *retval;
1436 {
1437 	return (ENOSYS);
1438 }
1439