xref: /netbsd-src/sys/compat/linux/common/linux_misc.c (revision de1dfb1250df962f1ff3a011772cf58e605aed11)
1 /*	$NetBSD: linux_misc.c,v 1.129 2004/09/13 20:09:44 jdolecek 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 1.129 2004/09/13 20:09:44 jdolecek 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/sa.h>
103 #include <sys/syscallargs.h>
104 
105 #include <compat/linux/common/linux_types.h>
106 #include <compat/linux/common/linux_signal.h>
107 
108 #include <compat/linux/linux_syscallargs.h>
109 
110 #include <compat/linux/common/linux_fcntl.h>
111 #include <compat/linux/common/linux_mmap.h>
112 #include <compat/linux/common/linux_dirent.h>
113 #include <compat/linux/common/linux_util.h>
114 #include <compat/linux/common/linux_misc.h>
115 #include <compat/linux/common/linux_ptrace.h>
116 #include <compat/linux/common/linux_reboot.h>
117 #include <compat/linux/common/linux_emuldata.h>
118 
119 const int linux_ptrace_request_map[] = {
120 	LINUX_PTRACE_TRACEME,	PT_TRACE_ME,
121 	LINUX_PTRACE_PEEKTEXT,	PT_READ_I,
122 	LINUX_PTRACE_PEEKDATA,	PT_READ_D,
123 	LINUX_PTRACE_POKETEXT,	PT_WRITE_I,
124 	LINUX_PTRACE_POKEDATA,	PT_WRITE_D,
125 	LINUX_PTRACE_CONT,	PT_CONTINUE,
126 	LINUX_PTRACE_KILL,	PT_KILL,
127 	LINUX_PTRACE_ATTACH,	PT_ATTACH,
128 	LINUX_PTRACE_DETACH,	PT_DETACH,
129 #ifdef PT_STEP
130 	LINUX_PTRACE_SINGLESTEP,	PT_STEP,
131 #endif
132 	-1
133 };
134 
135 static const struct mnttypes {
136 	char *bsd;
137 	int linux;
138 } fstypes[] = {
139 	{ MOUNT_FFS,		LINUX_DEFAULT_SUPER_MAGIC	},
140 	{ MOUNT_NFS,		LINUX_NFS_SUPER_MAGIC 		},
141 	{ MOUNT_MFS,		LINUX_DEFAULT_SUPER_MAGIC	},
142 	{ MOUNT_MSDOS,		LINUX_MSDOS_SUPER_MAGIC		},
143 	{ MOUNT_LFS,		LINUX_DEFAULT_SUPER_MAGIC	},
144 	{ MOUNT_FDESC,		LINUX_DEFAULT_SUPER_MAGIC	},
145 	{ MOUNT_PORTAL,		LINUX_DEFAULT_SUPER_MAGIC	},
146 	{ MOUNT_NULL,		LINUX_DEFAULT_SUPER_MAGIC	},
147 	{ MOUNT_OVERLAY,	LINUX_DEFAULT_SUPER_MAGIC	},
148 	{ MOUNT_UMAP,		LINUX_DEFAULT_SUPER_MAGIC	},
149 	{ MOUNT_KERNFS,		LINUX_DEFAULT_SUPER_MAGIC	},
150 	{ MOUNT_PROCFS,		LINUX_PROC_SUPER_MAGIC		},
151 	{ MOUNT_AFS,		LINUX_DEFAULT_SUPER_MAGIC	},
152 	{ MOUNT_CD9660,		LINUX_ISOFS_SUPER_MAGIC		},
153 	{ MOUNT_UNION,		LINUX_DEFAULT_SUPER_MAGIC	},
154 	{ MOUNT_ADOSFS,		LINUX_ADFS_SUPER_MAGIC		},
155 	{ MOUNT_EXT2FS,		LINUX_EXT2_SUPER_MAGIC		},
156 	{ MOUNT_CFS,		LINUX_DEFAULT_SUPER_MAGIC	},
157 	{ MOUNT_CODA,		LINUX_CODA_SUPER_MAGIC		},
158 	{ MOUNT_FILECORE,	LINUX_DEFAULT_SUPER_MAGIC	},
159 	{ MOUNT_NTFS,		LINUX_DEFAULT_SUPER_MAGIC	},
160 	{ MOUNT_SMBFS,		LINUX_SMB_SUPER_MAGIC		}
161 };
162 #define FSTYPESSIZE (sizeof(fstypes) / sizeof(fstypes[0]))
163 
164 #ifdef DEBUG_LINUX
165 #define DPRINTF(a)	uprintf a
166 #else
167 #define DPRINTF(a)
168 #endif
169 
170 /* Local linux_misc.c functions: */
171 static void bsd_to_linux_statfs __P((const struct statvfs *,
172     struct linux_statfs *));
173 static int linux_to_bsd_limit __P((int));
174 static void linux_to_bsd_mmap_args __P((struct sys_mmap_args *,
175     const struct linux_sys_mmap_args *));
176 static int linux_mmap __P((struct lwp *, struct linux_sys_mmap_args *,
177     register_t *, off_t));
178 
179 
180 /*
181  * The information on a terminated (or stopped) process needs
182  * to be converted in order for Linux binaries to get a valid signal
183  * number out of it.
184  */
185 void
186 bsd_to_linux_wstat(st)
187 	int *st;
188 {
189 
190 	int sig;
191 
192 	if (WIFSIGNALED(*st)) {
193 		sig = WTERMSIG(*st);
194 		if (sig >= 0 && sig < NSIG)
195 			*st= (*st& ~0177) | native_to_linux_signo[sig];
196 	} else if (WIFSTOPPED(*st)) {
197 		sig = WSTOPSIG(*st);
198 		if (sig >= 0 && sig < NSIG)
199 			*st = (*st & ~0xff00) |
200 			    (native_to_linux_signo[sig] << 8);
201 	}
202 }
203 
204 /*
205  * This is very much the same as waitpid()
206  */
207 int
208 linux_sys_wait4(l, v, retval)
209 	struct lwp *l;
210 	void *v;
211 	register_t *retval;
212 {
213 	struct linux_sys_wait4_args /* {
214 		syscallarg(int) pid;
215 		syscallarg(int *) status;
216 		syscallarg(int) options;
217 		syscallarg(struct rusage *) rusage;
218 	} */ *uap = v;
219 	struct proc *p = l->l_proc;
220 	struct sys_wait4_args w4a;
221 	int error, *status, tstat, options, linux_options;
222 	caddr_t sg;
223 
224 	if (SCARG(uap, status) != NULL) {
225 		sg = stackgap_init(p, 0);
226 		status = (int *) stackgap_alloc(p, &sg, sizeof *status);
227 	} else
228 		status = NULL;
229 
230 	linux_options = SCARG(uap, options);
231 	options = 0;
232 	if (linux_options &
233 	    ~(LINUX_WAIT4_WNOHANG|LINUX_WAIT4_WUNTRACED|LINUX_WAIT4_WALL|
234 	      LINUX_WAIT4_WCLONE))
235 		return (EINVAL);
236 
237 	if (linux_options & LINUX_WAIT4_WNOHANG)
238 		options |= WNOHANG;
239 	if (linux_options & LINUX_WAIT4_WUNTRACED)
240 		options |= WUNTRACED;
241 	if (linux_options & LINUX_WAIT4_WALL)
242 		options |= WALLSIG;
243 	if (linux_options & LINUX_WAIT4_WCLONE)
244 		options |= WALTSIG;
245 
246 	SCARG(&w4a, pid) = SCARG(uap, pid);
247 	SCARG(&w4a, status) = status;
248 	SCARG(&w4a, options) = options;
249 	SCARG(&w4a, rusage) = SCARG(uap, rusage);
250 
251 	if ((error = sys_wait4(l, &w4a, retval)))
252 		return error;
253 
254 	sigdelset(&p->p_sigctx.ps_siglist, SIGCHLD);
255 
256 	if (status != NULL) {
257 		if ((error = copyin(status, &tstat, sizeof tstat)))
258 			return error;
259 
260 		bsd_to_linux_wstat(&tstat);
261 		return copyout(&tstat, SCARG(uap, status), sizeof tstat);
262 	}
263 
264 	return 0;
265 }
266 
267 /*
268  * Linux brk(2). The check if the new address is >= the old one is
269  * done in the kernel in Linux. NetBSD does it in the library.
270  */
271 int
272 linux_sys_brk(l, v, retval)
273 	struct lwp *l;
274 	void *v;
275 	register_t *retval;
276 {
277 	struct linux_sys_brk_args /* {
278 		syscallarg(char *) nsize;
279 	} */ *uap = v;
280 	struct proc *p = l->l_proc;
281 	char *nbrk = SCARG(uap, nsize);
282 	struct sys_obreak_args oba;
283 	struct vmspace *vm = p->p_vmspace;
284 	struct linux_emuldata *ed = (struct linux_emuldata*)p->p_emuldata;
285 
286 	SCARG(&oba, nsize) = nbrk;
287 
288 	if ((caddr_t) nbrk > vm->vm_daddr && sys_obreak(l, &oba, retval) == 0)
289 		ed->s->p_break = (char*)nbrk;
290 	else
291 		nbrk = ed->s->p_break;
292 
293 	retval[0] = (register_t)nbrk;
294 
295 	return 0;
296 }
297 
298 /*
299  * Convert BSD statfs structure to Linux statfs structure.
300  * The Linux structure has less fields, and it also wants
301  * the length of a name in a dir entry in a field, which
302  * we fake (probably the wrong way).
303  */
304 static void
305 bsd_to_linux_statfs(bsp, lsp)
306 	const struct statvfs *bsp;
307 	struct linux_statfs *lsp;
308 {
309 	int i;
310 
311 	for (i = 0; i < FSTYPESSIZE; i++)
312 		if (strcmp(bsp->f_fstypename, fstypes[i].bsd) == 0)
313 			break;
314 
315 	if (i == FSTYPESSIZE) {
316 		DPRINTF(("unhandled fstype in linux emulation: %s\n",
317 		    bsp->f_fstypename));
318 		lsp->l_ftype = LINUX_DEFAULT_SUPER_MAGIC;
319 	} else {
320 		lsp->l_ftype = fstypes[i].linux;
321 	}
322 
323 	lsp->l_fbsize = bsp->f_frsize;
324 	lsp->l_fblocks = bsp->f_blocks;
325 	lsp->l_fbfree = bsp->f_bfree;
326 	lsp->l_fbavail = bsp->f_bavail;
327 	lsp->l_ffiles = bsp->f_files;
328 	lsp->l_fffree = bsp->f_ffree;
329 	/* Linux sets the fsid to 0..., we don't */
330 	lsp->l_ffsid.val[0] = bsp->f_fsidx.__fsid_val[0];
331 	lsp->l_ffsid.val[1] = bsp->f_fsidx.__fsid_val[1];
332 	lsp->l_fnamelen = bsp->f_namemax;
333 	(void)memset(lsp->l_fspare, 0, sizeof(lsp->l_fspare));
334 }
335 
336 /*
337  * Implement the fs stat functions. Straightforward.
338  */
339 int
340 linux_sys_statfs(l, v, retval)
341 	struct lwp *l;
342 	void *v;
343 	register_t *retval;
344 {
345 	struct linux_sys_statfs_args /* {
346 		syscallarg(const char *) path;
347 		syscallarg(struct linux_statfs *) sp;
348 	} */ *uap = v;
349 	struct proc *p = l->l_proc;
350 	struct statvfs btmp, *bsp;
351 	struct linux_statfs ltmp;
352 	struct sys_statvfs1_args bsa;
353 	caddr_t sg;
354 	int error;
355 
356 	sg = stackgap_init(p, 0);
357 	bsp = (struct statvfs *) stackgap_alloc(p, &sg, sizeof (struct statvfs));
358 
359 	CHECK_ALT_EXIST(p, &sg, SCARG(uap, path));
360 
361 	SCARG(&bsa, path) = SCARG(uap, path);
362 	SCARG(&bsa, buf) = bsp;
363 	SCARG(&bsa, flags) = ST_WAIT;
364 
365 	if ((error = sys_statvfs1(l, &bsa, retval)))
366 		return error;
367 
368 	if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp)))
369 		return error;
370 
371 	bsd_to_linux_statfs(&btmp, &ltmp);
372 
373 	return copyout((caddr_t) &ltmp, (caddr_t) SCARG(uap, sp), sizeof ltmp);
374 }
375 
376 int
377 linux_sys_fstatfs(l, v, retval)
378 	struct lwp *l;
379 	void *v;
380 	register_t *retval;
381 {
382 	struct linux_sys_fstatfs_args /* {
383 		syscallarg(int) fd;
384 		syscallarg(struct linux_statfs *) sp;
385 	} */ *uap = v;
386 	struct proc *p = l->l_proc;
387 	struct statvfs btmp, *bsp;
388 	struct linux_statfs ltmp;
389 	struct sys_fstatvfs1_args bsa;
390 	caddr_t sg;
391 	int error;
392 
393 	sg = stackgap_init(p, 0);
394 	bsp = (struct statvfs *) stackgap_alloc(p, &sg, sizeof (struct statvfs));
395 
396 	SCARG(&bsa, fd) = SCARG(uap, fd);
397 	SCARG(&bsa, buf) = bsp;
398 	SCARG(&bsa, flags) = ST_WAIT;
399 
400 	if ((error = sys_fstatvfs1(l, &bsa, retval)))
401 		return error;
402 
403 	if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp)))
404 		return error;
405 
406 	bsd_to_linux_statfs(&btmp, &ltmp);
407 
408 	return copyout((caddr_t) &ltmp, (caddr_t) SCARG(uap, sp), sizeof ltmp);
409 }
410 
411 /*
412  * uname(). Just copy the info from the various strings stored in the
413  * kernel, and put it in the Linux utsname structure. That structure
414  * is almost the same as the NetBSD one, only it has fields 65 characters
415  * long, and an extra domainname field.
416  */
417 int
418 linux_sys_uname(l, v, retval)
419 	struct lwp *l;
420 	void *v;
421 	register_t *retval;
422 {
423 	struct linux_sys_uname_args /* {
424 		syscallarg(struct linux_utsname *) up;
425 	} */ *uap = v;
426 	struct linux_utsname luts;
427 
428 	strncpy(luts.l_sysname, linux_sysname, sizeof(luts.l_sysname));
429 	strncpy(luts.l_nodename, hostname, sizeof(luts.l_nodename));
430 	strncpy(luts.l_release, linux_release, sizeof(luts.l_release));
431 	strncpy(luts.l_version, linux_version, sizeof(luts.l_version));
432 	strncpy(luts.l_machine, machine, sizeof(luts.l_machine));
433 	strncpy(luts.l_domainname, domainname, sizeof(luts.l_domainname));
434 
435 	return copyout(&luts, SCARG(uap, up), sizeof(luts));
436 }
437 
438 /* Used directly on: alpha, mips, ppc, sparc, sparc64 */
439 /* Used indirectly on: arm, i386, m68k */
440 
441 /*
442  * New type Linux mmap call.
443  * Only called directly on machines with >= 6 free regs.
444  */
445 int
446 linux_sys_mmap(l, v, retval)
447 	struct lwp *l;
448 	void *v;
449 	register_t *retval;
450 {
451 	struct linux_sys_mmap_args /* {
452 		syscallarg(unsigned long) addr;
453 		syscallarg(size_t) len;
454 		syscallarg(int) prot;
455 		syscallarg(int) flags;
456 		syscallarg(int) fd;
457 		syscallarg(linux_off_t) offset;
458 	} */ *uap = v;
459 
460 	if (SCARG(uap, offset) & PAGE_MASK)
461 		return EINVAL;
462 
463 	return linux_mmap(l, uap, retval, SCARG(uap, offset));
464 }
465 
466 /*
467  * Guts of most architectures' mmap64() implementations.  This shares
468  * its list of arguments with linux_sys_mmap().
469  *
470  * The difference in linux_sys_mmap2() is that "offset" is actually
471  * (offset / pagesize), not an absolute byte count.  This translation
472  * to pagesize offsets is done inside glibc between the mmap64() call
473  * point, and the actual syscall.
474  */
475 int
476 linux_sys_mmap2(l, v, retval)
477 	struct lwp *l;
478 	void *v;
479 	register_t *retval;
480 {
481 	struct linux_sys_mmap2_args /* {
482 		syscallarg(unsigned long) addr;
483 		syscallarg(size_t) len;
484 		syscallarg(int) prot;
485 		syscallarg(int) flags;
486 		syscallarg(int) fd;
487 		syscallarg(linux_off_t) offset;
488 	} */ *uap = v;
489 
490 	return linux_mmap(l, uap, retval,
491 	    ((off_t)SCARG(uap, offset)) << PAGE_SHIFT);
492 }
493 
494 /*
495  * Massage arguments and call system mmap(2).
496  */
497 static int
498 linux_mmap(l, uap, retval, offset)
499 	struct lwp *l;
500 	struct linux_sys_mmap_args *uap;
501 	register_t *retval;
502 	off_t offset;
503 {
504 	struct sys_mmap_args cma;
505 	int error;
506 	size_t mmoff=0;
507 
508 	if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN) {
509 		/*
510 		 * Request for stack-like memory segment. On linux, this
511 		 * works by mmap()ping (small) segment, which is automatically
512 		 * extended when page fault happens below the currently
513 		 * allocated area. We emulate this by allocating (typically
514 		 * bigger) segment sized at current stack size limit, and
515 		 * offsetting the requested and returned address accordingly.
516 		 * Since physical pages are only allocated on-demand, this
517 		 * is effectively identical.
518 		 */
519 		rlim_t ssl = l->l_proc->p_rlimit[RLIMIT_STACK].rlim_cur;
520 
521 		if (SCARG(uap, len) < ssl) {
522 			/* Compute the address offset */
523 			mmoff = round_page(ssl) - SCARG(uap, len);
524 
525 			if (SCARG(uap, addr))
526 				SCARG(uap, addr) -= mmoff;
527 
528 			SCARG(uap, len) = (size_t) ssl;
529 		}
530 	}
531 
532 	linux_to_bsd_mmap_args(&cma, uap);
533 	SCARG(&cma, pos) = offset;
534 
535 	error = sys_mmap(l, &cma, retval);
536 	if (error)
537 		return (error);
538 
539 	/* Shift the returned address for stack-like segment if necessary */
540 	if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN && mmoff)
541 		retval[0] += mmoff;
542 
543 	return (0);
544 }
545 
546 static void
547 linux_to_bsd_mmap_args(cma, uap)
548 	struct sys_mmap_args *cma;
549 	const struct linux_sys_mmap_args *uap;
550 {
551 	int flags = MAP_TRYFIXED, fl = SCARG(uap, flags);
552 
553 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_SHARED, MAP_SHARED);
554 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_PRIVATE, MAP_PRIVATE);
555 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_FIXED, MAP_FIXED);
556 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_ANON, MAP_ANON);
557 	/* XXX XAX ERH: Any other flags here?  There are more defined... */
558 
559 	SCARG(cma, addr) = (void *)SCARG(uap, addr);
560 	SCARG(cma, len) = SCARG(uap, len);
561 	SCARG(cma, prot) = SCARG(uap, prot);
562 	if (SCARG(cma, prot) & VM_PROT_WRITE) /* XXX */
563 		SCARG(cma, prot) |= VM_PROT_READ;
564 	SCARG(cma, flags) = flags;
565 	SCARG(cma, fd) = flags & MAP_ANON ? -1 : SCARG(uap, fd);
566 	SCARG(cma, pad) = 0;
567 }
568 
569 int
570 linux_sys_mremap(l, v, retval)
571 	struct lwp *l;
572 	void *v;
573 	register_t *retval;
574 {
575 	struct linux_sys_mremap_args /* {
576 		syscallarg(void *) old_address;
577 		syscallarg(size_t) old_size;
578 		syscallarg(size_t) new_size;
579 		syscallarg(u_long) flags;
580 	} */ *uap = v;
581 	struct sys_munmap_args mua;
582 	size_t old_size, new_size;
583 	int error;
584 
585 	old_size = round_page(SCARG(uap, old_size));
586 	new_size = round_page(SCARG(uap, new_size));
587 
588 	/*
589 	 * Growing mapped region.
590 	 */
591 	if (new_size > old_size) {
592 		/*
593 		 * XXX Implement me.  What we probably want to do is
594 		 * XXX dig out the guts of the old mapping, mmap that
595 		 * XXX object again with the new size, then munmap
596 		 * XXX the old mapping.
597 		 */
598 		*retval = 0;
599 		return (ENOMEM);
600 	}
601 
602 	/*
603 	 * Shrinking mapped region.
604 	 */
605 	if (new_size < old_size) {
606 		SCARG(&mua, addr) = (caddr_t)SCARG(uap, old_address) +
607 		    new_size;
608 		SCARG(&mua, len) = old_size - new_size;
609 		error = sys_munmap(l, &mua, retval);
610 		*retval = error ? 0 : (register_t)SCARG(uap, old_address);
611 		return (error);
612 	}
613 
614 	/*
615 	 * No change.
616 	 */
617 	*retval = (register_t)SCARG(uap, old_address);
618 	return (0);
619 }
620 
621 int
622 linux_sys_msync(l, v, retval)
623 	struct lwp *l;
624 	void *v;
625 	register_t *retval;
626 {
627 	struct linux_sys_msync_args /* {
628 		syscallarg(caddr_t) addr;
629 		syscallarg(int) len;
630 		syscallarg(int) fl;
631 	} */ *uap = v;
632 
633 	struct sys___msync13_args bma;
634 
635 	/* flags are ignored */
636 	SCARG(&bma, addr) = SCARG(uap, addr);
637 	SCARG(&bma, len) = SCARG(uap, len);
638 	SCARG(&bma, flags) = SCARG(uap, fl);
639 
640 	return sys___msync13(l, &bma, retval);
641 }
642 
643 int
644 linux_sys_mprotect(l, v, retval)
645 	struct lwp *l;
646 	void *v;
647 	register_t *retval;
648 {
649 	struct linux_sys_mprotect_args /* {
650 		syscallarg(const void *) start;
651 		syscallarg(unsigned long) len;
652 		syscallarg(int) prot;
653 	} */ *uap = v;
654 	unsigned long end, start = (unsigned long)SCARG(uap, start), len;
655 	int prot = SCARG(uap, prot);
656 	struct vm_map_entry *entry;
657 	struct vm_map *map = &l->l_proc->p_vmspace->vm_map;
658 
659 	if (start & PAGE_MASK)
660 		return EINVAL;
661 
662 	len = round_page(SCARG(uap, len));
663 	end = start + len;
664 
665 	if (end < start)
666 		return EINVAL;
667 	else if (end == start)
668 		return 0;
669 
670 	if (SCARG(uap, prot) & ~(PROT_READ | PROT_WRITE | PROT_EXEC))
671 		return EINVAL;
672 
673 	vm_map_lock(map);
674 #ifdef notdef
675 	VM_MAP_RANGE_CHECK(map, start, end);
676 #endif
677 	if (!uvm_map_lookup_entry(map, start, &entry) || entry->start > start) {
678 		vm_map_unlock(map);
679 		return ENOMEM;
680 	}
681 	vm_map_unlock(map);
682 	return uvm_map_protect(map, start, end, prot, FALSE);
683 }
684 
685 /*
686  * This code is partly stolen from src/lib/libc/compat-43/times.c
687  */
688 
689 #define	CONVTCK(r)	(r.tv_sec * hz + r.tv_usec / (1000000 / hz))
690 
691 int
692 linux_sys_times(l, v, retval)
693 	struct lwp *l;
694 	void *v;
695 	register_t *retval;
696 {
697 	struct linux_sys_times_args /* {
698 		syscallarg(struct times *) tms;
699 	} */ *uap = v;
700 	struct proc *p = l->l_proc;
701 	struct timeval t;
702 	int error, s;
703 
704 	if (SCARG(uap, tms)) {
705 		struct linux_tms ltms;
706 		struct rusage ru;
707 
708 		calcru(p, &ru.ru_utime, &ru.ru_stime, NULL);
709 		ltms.ltms_utime = CONVTCK(ru.ru_utime);
710 		ltms.ltms_stime = CONVTCK(ru.ru_stime);
711 
712 		ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime);
713 		ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime);
714 
715 		if ((error = copyout(&ltms, SCARG(uap, tms), sizeof ltms)))
716 			return error;
717 	}
718 
719 	s = splclock();
720 	timersub(&time, &boottime, &t);
721 	splx(s);
722 
723 	retval[0] = ((linux_clock_t)(CONVTCK(t)));
724 	return 0;
725 }
726 
727 #undef CONVTCK
728 
729 /*
730  * Linux 'readdir' call. This code is mostly taken from the
731  * SunOS getdents call (see compat/sunos/sunos_misc.c), though
732  * an attempt has been made to keep it a little cleaner (failing
733  * miserably, because of the cruft needed if count 1 is passed).
734  *
735  * The d_off field should contain the offset of the next valid entry,
736  * but in Linux it has the offset of the entry itself. We emulate
737  * that bug here.
738  *
739  * Read in BSD-style entries, convert them, and copy them out.
740  *
741  * Note that this doesn't handle union-mounted filesystems.
742  */
743 int
744 linux_sys_getdents(l, v, retval)
745 	struct lwp *l;
746 	void *v;
747 	register_t *retval;
748 {
749 	struct linux_sys_getdents_args /* {
750 		syscallarg(int) fd;
751 		syscallarg(struct linux_dirent *) dent;
752 		syscallarg(unsigned int) count;
753 	} */ *uap = v;
754 	struct proc *p = l->l_proc;
755 	struct dirent *bdp;
756 	struct vnode *vp;
757 	caddr_t	inp, buf;		/* BSD-format */
758 	int len, reclen;		/* BSD-format */
759 	caddr_t outp;			/* Linux-format */
760 	int resid, linux_reclen = 0;	/* Linux-format */
761 	struct file *fp;
762 	struct uio auio;
763 	struct iovec aiov;
764 	struct linux_dirent idb;
765 	off_t off;		/* true file offset */
766 	int buflen, error, eofflag, nbytes, oldcall;
767 	struct vattr va;
768 	off_t *cookiebuf = NULL, *cookie;
769 	int ncookies;
770 
771 	/* getvnode() will use the descriptor for us */
772 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
773 		return (error);
774 
775 	if ((fp->f_flag & FREAD) == 0) {
776 		error = EBADF;
777 		goto out1;
778 	}
779 
780 	vp = (struct vnode *)fp->f_data;
781 	if (vp->v_type != VDIR) {
782 		error = EINVAL;
783 		goto out1;
784 	}
785 
786 	if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)))
787 		goto out1;
788 
789 	nbytes = SCARG(uap, count);
790 	if (nbytes == 1) {	/* emulating old, broken behaviour */
791 		nbytes = sizeof (idb);
792 		buflen = max(va.va_blocksize, nbytes);
793 		oldcall = 1;
794 	} else {
795 		buflen = min(MAXBSIZE, nbytes);
796 		if (buflen < va.va_blocksize)
797 			buflen = va.va_blocksize;
798 		oldcall = 0;
799 	}
800 	buf = malloc(buflen, M_TEMP, M_WAITOK);
801 
802 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
803 	off = fp->f_offset;
804 again:
805 	aiov.iov_base = buf;
806 	aiov.iov_len = buflen;
807 	auio.uio_iov = &aiov;
808 	auio.uio_iovcnt = 1;
809 	auio.uio_rw = UIO_READ;
810 	auio.uio_segflg = UIO_SYSSPACE;
811 	auio.uio_procp = p;
812 	auio.uio_resid = buflen;
813 	auio.uio_offset = off;
814 	/*
815          * First we read into the malloc'ed buffer, then
816          * we massage it into user space, one record at a time.
817          */
818 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf,
819 	    &ncookies);
820 	if (error)
821 		goto out;
822 
823 	inp = buf;
824 	outp = (caddr_t)SCARG(uap, dent);
825 	resid = nbytes;
826 	if ((len = buflen - auio.uio_resid) == 0)
827 		goto eof;
828 
829 	for (cookie = cookiebuf; len > 0; len -= reclen) {
830 		bdp = (struct dirent *)inp;
831 		reclen = bdp->d_reclen;
832 		if (reclen & 3)
833 			panic("linux_readdir");
834 		if (bdp->d_fileno == 0) {
835 			inp += reclen;	/* it is a hole; squish it out */
836 			off = *cookie++;
837 			continue;
838 		}
839 		linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen);
840 		if (reclen > len || resid < linux_reclen) {
841 			/* entry too big for buffer, so just stop */
842 			outp++;
843 			break;
844 		}
845 		/*
846 		 * Massage in place to make a Linux-shaped dirent (otherwise
847 		 * we have to worry about touching user memory outside of
848 		 * the copyout() call).
849 		 */
850 		idb.d_ino = bdp->d_fileno;
851 		/*
852 		 * The old readdir() call misuses the offset and reclen fields.
853 		 */
854 		if (oldcall) {
855 			idb.d_off = (linux_off_t)linux_reclen;
856 			idb.d_reclen = (u_short)bdp->d_namlen;
857 		} else {
858 			if (sizeof (idb.d_off) <= 4 && (off >> 32) != 0) {
859 				compat_offseterr(vp, "linux_getdents");
860 				error = EINVAL;
861 				goto out;
862 			}
863 			idb.d_off = (linux_off_t)off;
864 			idb.d_reclen = (u_short)linux_reclen;
865 		}
866 		strcpy(idb.d_name, bdp->d_name);
867 		if ((error = copyout((caddr_t)&idb, outp, linux_reclen)))
868 			goto out;
869 		/* advance past this real entry */
870 		inp += reclen;
871 		off = *cookie++;	/* each entry points to itself */
872 		/* advance output past Linux-shaped entry */
873 		outp += linux_reclen;
874 		resid -= linux_reclen;
875 		if (oldcall)
876 			break;
877 	}
878 
879 	/* if we squished out the whole block, try again */
880 	if (outp == (caddr_t)SCARG(uap, dent))
881 		goto again;
882 	fp->f_offset = off;	/* update the vnode offset */
883 
884 	if (oldcall)
885 		nbytes = resid + linux_reclen;
886 
887 eof:
888 	*retval = nbytes - resid;
889 out:
890 	VOP_UNLOCK(vp, 0);
891 	if (cookiebuf)
892 		free(cookiebuf, M_TEMP);
893 	free(buf, M_TEMP);
894 out1:
895 	FILE_UNUSE(fp, p);
896 	return error;
897 }
898 
899 /*
900  * Even when just using registers to pass arguments to syscalls you can
901  * have 5 of them on the i386. So this newer version of select() does
902  * this.
903  */
904 int
905 linux_sys_select(l, v, retval)
906 	struct lwp *l;
907 	void *v;
908 	register_t *retval;
909 {
910 	struct linux_sys_select_args /* {
911 		syscallarg(int) nfds;
912 		syscallarg(fd_set *) readfds;
913 		syscallarg(fd_set *) writefds;
914 		syscallarg(fd_set *) exceptfds;
915 		syscallarg(struct timeval *) timeout;
916 	} */ *uap = v;
917 
918 	return linux_select1(l, retval, SCARG(uap, nfds), SCARG(uap, readfds),
919 	    SCARG(uap, writefds), SCARG(uap, exceptfds), SCARG(uap, timeout));
920 }
921 
922 /*
923  * Common code for the old and new versions of select(). A couple of
924  * things are important:
925  * 1) return the amount of time left in the 'timeout' parameter
926  * 2) select never returns ERESTART on Linux, always return EINTR
927  */
928 int
929 linux_select1(l, retval, nfds, readfds, writefds, exceptfds, timeout)
930 	struct lwp *l;
931 	register_t *retval;
932 	int nfds;
933 	fd_set *readfds, *writefds, *exceptfds;
934 	struct timeval *timeout;
935 {
936 	struct sys_select_args bsa;
937 	struct proc *p = l->l_proc;
938 	struct timeval tv0, tv1, utv, *tvp;
939 	caddr_t sg;
940 	int error;
941 
942 	SCARG(&bsa, nd) = nfds;
943 	SCARG(&bsa, in) = readfds;
944 	SCARG(&bsa, ou) = writefds;
945 	SCARG(&bsa, ex) = exceptfds;
946 	SCARG(&bsa, tv) = timeout;
947 
948 	/*
949 	 * Store current time for computation of the amount of
950 	 * time left.
951 	 */
952 	if (timeout) {
953 		if ((error = copyin(timeout, &utv, sizeof(utv))))
954 			return error;
955 		if (itimerfix(&utv)) {
956 			/*
957 			 * The timeval was invalid.  Convert it to something
958 			 * valid that will act as it does under Linux.
959 			 */
960 			sg = stackgap_init(p, 0);
961 			tvp = stackgap_alloc(p, &sg, sizeof(utv));
962 			utv.tv_sec += utv.tv_usec / 1000000;
963 			utv.tv_usec %= 1000000;
964 			if (utv.tv_usec < 0) {
965 				utv.tv_sec -= 1;
966 				utv.tv_usec += 1000000;
967 			}
968 			if (utv.tv_sec < 0)
969 				timerclear(&utv);
970 			if ((error = copyout(&utv, tvp, sizeof(utv))))
971 				return error;
972 			SCARG(&bsa, tv) = tvp;
973 		}
974 		microtime(&tv0);
975 	}
976 
977 	error = sys_select(l, &bsa, retval);
978 	if (error) {
979 		/*
980 		 * See fs/select.c in the Linux kernel.  Without this,
981 		 * Maelstrom doesn't work.
982 		 */
983 		if (error == ERESTART)
984 			error = EINTR;
985 		return error;
986 	}
987 
988 	if (timeout) {
989 		if (*retval) {
990 			/*
991 			 * Compute how much time was left of the timeout,
992 			 * by subtracting the current time and the time
993 			 * before we started the call, and subtracting
994 			 * that result from the user-supplied value.
995 			 */
996 			microtime(&tv1);
997 			timersub(&tv1, &tv0, &tv1);
998 			timersub(&utv, &tv1, &utv);
999 			if (utv.tv_sec < 0)
1000 				timerclear(&utv);
1001 		} else
1002 			timerclear(&utv);
1003 		if ((error = copyout(&utv, timeout, sizeof(utv))))
1004 			return error;
1005 	}
1006 
1007 	return 0;
1008 }
1009 
1010 /*
1011  * Get the process group of a certain process. Look it up
1012  * and return the value.
1013  */
1014 int
1015 linux_sys_getpgid(l, v, retval)
1016 	struct lwp *l;
1017 	void *v;
1018 	register_t *retval;
1019 {
1020 	struct linux_sys_getpgid_args /* {
1021 		syscallarg(int) pid;
1022 	} */ *uap = v;
1023 	struct proc *p = l->l_proc;
1024 	struct proc *targp;
1025 
1026 	if (SCARG(uap, pid) != 0 && SCARG(uap, pid) != p->p_pid) {
1027 		if ((targp = pfind(SCARG(uap, pid))) == 0)
1028 			return ESRCH;
1029 	}
1030 	else
1031 		targp = p;
1032 
1033 	retval[0] = targp->p_pgid;
1034 	return 0;
1035 }
1036 
1037 /*
1038  * Set the 'personality' (emulation mode) for the current process. Only
1039  * accept the Linux personality here (0). This call is needed because
1040  * the Linux ELF crt0 issues it in an ugly kludge to make sure that
1041  * ELF binaries run in Linux mode, not SVR4 mode.
1042  */
1043 int
1044 linux_sys_personality(l, v, retval)
1045 	struct lwp *l;
1046 	void *v;
1047 	register_t *retval;
1048 {
1049 	struct linux_sys_personality_args /* {
1050 		syscallarg(int) per;
1051 	} */ *uap = v;
1052 
1053 	if (SCARG(uap, per) != 0)
1054 		return EINVAL;
1055 	retval[0] = 0;
1056 	return 0;
1057 }
1058 
1059 #if defined(__i386__) || defined(__m68k__)
1060 /*
1061  * The calls are here because of type conversions.
1062  */
1063 int
1064 linux_sys_setreuid16(l, v, retval)
1065 	struct lwp *l;
1066 	void *v;
1067 	register_t *retval;
1068 {
1069 	struct linux_sys_setreuid16_args /* {
1070 		syscallarg(int) ruid;
1071 		syscallarg(int) euid;
1072 	} */ *uap = v;
1073 	struct sys_setreuid_args bsa;
1074 
1075 	SCARG(&bsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
1076 		(uid_t)-1 : SCARG(uap, ruid);
1077 	SCARG(&bsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
1078 		(uid_t)-1 : SCARG(uap, euid);
1079 
1080 	return sys_setreuid(l, &bsa, retval);
1081 }
1082 
1083 int
1084 linux_sys_setregid16(l, v, retval)
1085 	struct lwp *l;
1086 	void *v;
1087 	register_t *retval;
1088 {
1089 	struct linux_sys_setregid16_args /* {
1090 		syscallarg(int) rgid;
1091 		syscallarg(int) egid;
1092 	} */ *uap = v;
1093 	struct sys_setregid_args bsa;
1094 
1095 	SCARG(&bsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
1096 		(uid_t)-1 : SCARG(uap, rgid);
1097 	SCARG(&bsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
1098 		(uid_t)-1 : SCARG(uap, egid);
1099 
1100 	return sys_setregid(l, &bsa, retval);
1101 }
1102 
1103 int
1104 linux_sys_setresuid16(l, v, retval)
1105 	struct lwp *l;
1106 	void *v;
1107 	register_t *retval;
1108 {
1109 	struct linux_sys_setresuid16_args /* {
1110 		syscallarg(uid_t) ruid;
1111 		syscallarg(uid_t) euid;
1112 		syscallarg(uid_t) suid;
1113 	} */ *uap = v;
1114 	struct linux_sys_setresuid16_args lsa;
1115 
1116 	SCARG(&lsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
1117 		(uid_t)-1 : SCARG(uap, ruid);
1118 	SCARG(&lsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
1119 		(uid_t)-1 : SCARG(uap, euid);
1120 	SCARG(&lsa, suid) = ((linux_uid_t)SCARG(uap, suid) == (linux_uid_t)-1) ?
1121 		(uid_t)-1 : SCARG(uap, suid);
1122 
1123 	return linux_sys_setresuid(l, &lsa, retval);
1124 }
1125 
1126 int
1127 linux_sys_setresgid16(l, v, retval)
1128 	struct lwp *l;
1129 	void *v;
1130 	register_t *retval;
1131 {
1132 	struct linux_sys_setresgid16_args /* {
1133 		syscallarg(gid_t) rgid;
1134 		syscallarg(gid_t) egid;
1135 		syscallarg(gid_t) sgid;
1136 	} */ *uap = v;
1137 	struct linux_sys_setresgid16_args lsa;
1138 
1139 	SCARG(&lsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
1140 		(gid_t)-1 : SCARG(uap, rgid);
1141 	SCARG(&lsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
1142 		(gid_t)-1 : SCARG(uap, egid);
1143 	SCARG(&lsa, sgid) = ((linux_gid_t)SCARG(uap, sgid) == (linux_gid_t)-1) ?
1144 		(gid_t)-1 : SCARG(uap, sgid);
1145 
1146 	return linux_sys_setresgid(l, &lsa, retval);
1147 }
1148 
1149 int
1150 linux_sys_getgroups16(l, v, retval)
1151 	struct lwp *l;
1152 	void *v;
1153 	register_t *retval;
1154 {
1155 	struct linux_sys_getgroups16_args /* {
1156 		syscallarg(int) gidsetsize;
1157 		syscallarg(linux_gid_t *) gidset;
1158 	} */ *uap = v;
1159 	struct proc *p = l->l_proc;
1160 	caddr_t sg;
1161 	int n, error, i;
1162 	struct sys_getgroups_args bsa;
1163 	gid_t *bset, *kbset;
1164 	linux_gid_t *lset;
1165 	struct pcred *pc = p->p_cred;
1166 
1167 	n = SCARG(uap, gidsetsize);
1168 	if (n < 0)
1169 		return EINVAL;
1170 	error = 0;
1171 	bset = kbset = NULL;
1172 	lset = NULL;
1173 	if (n > 0) {
1174 		n = min(pc->pc_ucred->cr_ngroups, n);
1175 		sg = stackgap_init(p, 0);
1176 		bset = stackgap_alloc(p, &sg, n * sizeof (gid_t));
1177 		kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK);
1178 		lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
1179 		if (bset == NULL || kbset == NULL || lset == NULL)
1180 			return ENOMEM;
1181 		SCARG(&bsa, gidsetsize) = n;
1182 		SCARG(&bsa, gidset) = bset;
1183 		error = sys_getgroups(l, &bsa, retval);
1184 		if (error != 0)
1185 			goto out;
1186 		error = copyin(bset, kbset, n * sizeof (gid_t));
1187 		if (error != 0)
1188 			goto out;
1189 		for (i = 0; i < n; i++)
1190 			lset[i] = (linux_gid_t)kbset[i];
1191 		error = copyout(lset, SCARG(uap, gidset),
1192 		    n * sizeof (linux_gid_t));
1193 	} else
1194 		*retval = pc->pc_ucred->cr_ngroups;
1195 out:
1196 	if (kbset != NULL)
1197 		free(kbset, M_TEMP);
1198 	if (lset != NULL)
1199 		free(lset, M_TEMP);
1200 	return error;
1201 }
1202 
1203 int
1204 linux_sys_setgroups16(l, v, retval)
1205 	struct lwp *l;
1206 	void *v;
1207 	register_t *retval;
1208 {
1209 	struct linux_sys_setgroups16_args /* {
1210 		syscallarg(int) gidsetsize;
1211 		syscallarg(linux_gid_t *) gidset;
1212 	} */ *uap = v;
1213 	struct proc *p = l->l_proc;
1214 	caddr_t sg;
1215 	int n;
1216 	int error, i;
1217 	struct sys_setgroups_args bsa;
1218 	gid_t *bset, *kbset;
1219 	linux_gid_t *lset;
1220 
1221 	n = SCARG(uap, gidsetsize);
1222 	if (n < 0 || n > NGROUPS)
1223 		return EINVAL;
1224 	sg = stackgap_init(p, 0);
1225 	bset = stackgap_alloc(p, &sg, n * sizeof (gid_t));
1226 	lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
1227 	kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK);
1228 	if (lset == NULL || bset == NULL)
1229 		return ENOMEM;
1230 	error = copyin(SCARG(uap, gidset), lset, n * sizeof (linux_gid_t));
1231 	if (error != 0)
1232 		goto out;
1233 	for (i = 0; i < n; i++)
1234 		kbset[i] = (gid_t)lset[i];
1235 	error = copyout(kbset, bset, n * sizeof (gid_t));
1236 	if (error != 0)
1237 		goto out;
1238 	SCARG(&bsa, gidsetsize) = n;
1239 	SCARG(&bsa, gidset) = bset;
1240 	error = sys_setgroups(l, &bsa, retval);
1241 
1242 out:
1243 	if (lset != NULL)
1244 		free(lset, M_TEMP);
1245 	if (kbset != NULL)
1246 		free(kbset, M_TEMP);
1247 
1248 	return error;
1249 }
1250 
1251 #endif /* __i386__ || __m68k__ */
1252 
1253 /*
1254  * We have nonexistent fsuid equal to uid.
1255  * If modification is requested, refuse.
1256  */
1257 int
1258 linux_sys_setfsuid(l, v, retval)
1259 	 struct lwp *l;
1260 	 void *v;
1261 	 register_t *retval;
1262 {
1263 	 struct linux_sys_setfsuid_args /* {
1264 		 syscallarg(uid_t) uid;
1265 	 } */ *uap = v;
1266 	 struct proc *p = l->l_proc;
1267 	 uid_t uid;
1268 
1269 	 uid = SCARG(uap, uid);
1270 	 if (p->p_cred->p_ruid != uid)
1271 		 return sys_nosys(l, v, retval);
1272 	 else
1273 		 return (0);
1274 }
1275 
1276 /* XXX XXX XXX */
1277 #ifndef alpha
1278 int
1279 linux_sys_getfsuid(l, v, retval)
1280 	struct lwp *l;
1281 	void *v;
1282 	register_t *retval;
1283 {
1284 	return sys_getuid(l, v, retval);
1285 }
1286 #endif
1287 
1288 int
1289 linux_sys_setresuid(l, v, retval)
1290 	struct lwp *l;
1291 	void *v;
1292 	register_t *retval;
1293 {
1294 	struct linux_sys_setresuid_args /* {
1295 		syscallarg(uid_t) ruid;
1296 		syscallarg(uid_t) euid;
1297 		syscallarg(uid_t) suid;
1298 	} */ *uap = v;
1299 
1300 	/*
1301 	 * Note: These checks are a little different than the NetBSD
1302 	 * setreuid(2) call performs.  This precisely follows the
1303 	 * behavior of the Linux kernel.
1304 	 */
1305 
1306 	return do_setresuid(l, SCARG(uap, ruid), SCARG(uap, euid),
1307 			    SCARG(uap, suid),
1308 			    ID_R_EQ_R | ID_R_EQ_E | ID_R_EQ_S |
1309 			    ID_E_EQ_R | ID_E_EQ_E | ID_E_EQ_S |
1310 			    ID_S_EQ_R | ID_S_EQ_E | ID_S_EQ_S );
1311 }
1312 
1313 int
1314 linux_sys_getresuid(l, v, retval)
1315 	struct lwp *l;
1316 	void *v;
1317 	register_t *retval;
1318 {
1319 	struct linux_sys_getresuid_args /* {
1320 		syscallarg(uid_t *) ruid;
1321 		syscallarg(uid_t *) euid;
1322 		syscallarg(uid_t *) suid;
1323 	} */ *uap = v;
1324 	struct proc *p = l->l_proc;
1325 	struct pcred *pc = p->p_cred;
1326 	int error;
1327 
1328 	/*
1329 	 * Linux copies these values out to userspace like so:
1330 	 *
1331 	 *	1. Copy out ruid.
1332 	 *	2. If that succeeds, copy out euid.
1333 	 *	3. If both of those succeed, copy out suid.
1334 	 */
1335 	if ((error = copyout(&pc->p_ruid, SCARG(uap, ruid),
1336 			     sizeof(uid_t))) != 0)
1337 		return (error);
1338 
1339 	if ((error = copyout(&pc->pc_ucred->cr_uid, SCARG(uap, euid),
1340 			     sizeof(uid_t))) != 0)
1341 		return (error);
1342 
1343 	return (copyout(&pc->p_svuid, SCARG(uap, suid), sizeof(uid_t)));
1344 }
1345 
1346 int
1347 linux_sys_ptrace(l, v, retval)
1348 	struct lwp *l;
1349 	void *v;
1350 	register_t *retval;
1351 {
1352 	struct linux_sys_ptrace_args /* {
1353 		i386, m68k, powerpc: T=int
1354 		alpha: T=long
1355 		syscallarg(T) request;
1356 		syscallarg(T) pid;
1357 		syscallarg(T) addr;
1358 		syscallarg(T) data;
1359 	} */ *uap = v;
1360 	const int *ptr;
1361 	int request;
1362 	int error;
1363 
1364 	ptr = linux_ptrace_request_map;
1365 	request = SCARG(uap, request);
1366 	while (*ptr != -1)
1367 		if (*ptr++ == request) {
1368 			struct sys_ptrace_args pta;
1369 
1370 			SCARG(&pta, req) = *ptr;
1371 			SCARG(&pta, pid) = SCARG(uap, pid);
1372 			SCARG(&pta, addr) = (caddr_t)SCARG(uap, addr);
1373 			SCARG(&pta, data) = SCARG(uap, data);
1374 
1375 			/*
1376 			 * Linux ptrace(PTRACE_CONT, pid, 0, 0) means actually
1377 			 * to continue where the process left off previously.
1378 			 * The same thing is achieved by addr == (caddr_t) 1
1379 			 * on NetBSD, so rewrite 'addr' appropriately.
1380 			 */
1381 			if (request == LINUX_PTRACE_CONT && SCARG(uap, addr)==0)
1382 				SCARG(&pta, addr) = (caddr_t) 1;
1383 
1384 			error = sys_ptrace(l, &pta, retval);
1385 			if (error)
1386 				return error;
1387 			switch (request) {
1388 			case LINUX_PTRACE_PEEKTEXT:
1389 			case LINUX_PTRACE_PEEKDATA:
1390 				error = copyout (retval,
1391 				    (caddr_t)SCARG(uap, data), sizeof *retval);
1392 				*retval = SCARG(uap, data);
1393 				break;
1394 			default:
1395 				break;
1396 			}
1397 			return error;
1398 		}
1399 		else
1400 			ptr++;
1401 
1402 	return LINUX_SYS_PTRACE_ARCH(l, uap, retval);
1403 }
1404 
1405 int
1406 linux_sys_reboot(struct lwp *l, void *v, register_t *retval)
1407 {
1408 	struct linux_sys_reboot_args /* {
1409 		syscallarg(int) magic1;
1410 		syscallarg(int) magic2;
1411 		syscallarg(int) cmd;
1412 		syscallarg(void *) arg;
1413 	} */ *uap = v;
1414 	struct sys_reboot_args /* {
1415 		syscallarg(int) opt;
1416 		syscallarg(char *) bootstr;
1417 	} */ sra;
1418 	struct proc *p = l->l_proc;
1419 	int error;
1420 
1421 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1422 		return(error);
1423 
1424 	if (SCARG(uap, magic1) != LINUX_REBOOT_MAGIC1)
1425 		return(EINVAL);
1426 	if (SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2 &&
1427 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2A &&
1428 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2B)
1429 		return(EINVAL);
1430 
1431 	switch (SCARG(uap, cmd)) {
1432 	case LINUX_REBOOT_CMD_RESTART:
1433 		SCARG(&sra, opt) = RB_AUTOBOOT;
1434 		break;
1435 	case LINUX_REBOOT_CMD_HALT:
1436 		SCARG(&sra, opt) = RB_HALT;
1437 		break;
1438 	case LINUX_REBOOT_CMD_POWER_OFF:
1439 		SCARG(&sra, opt) = RB_HALT|RB_POWERDOWN;
1440 		break;
1441 	case LINUX_REBOOT_CMD_RESTART2:
1442 		/* Reboot with an argument. */
1443 		SCARG(&sra, opt) = RB_AUTOBOOT|RB_STRING;
1444 		SCARG(&sra, bootstr) = SCARG(uap, arg);
1445 		break;
1446 	case LINUX_REBOOT_CMD_CAD_ON:
1447 		return(EINVAL);	/* We don't implement ctrl-alt-delete */
1448 	case LINUX_REBOOT_CMD_CAD_OFF:
1449 		return(0);
1450 	default:
1451 		return(EINVAL);
1452 	}
1453 
1454 	return(sys_reboot(l, &sra, retval));
1455 }
1456 
1457 /*
1458  * Copy of compat_12_sys_swapon().
1459  */
1460 int
1461 linux_sys_swapon(l, v, retval)
1462 	struct lwp *l;
1463 	void *v;
1464 	register_t *retval;
1465 {
1466 	struct sys_swapctl_args ua;
1467 	struct linux_sys_swapon_args /* {
1468 		syscallarg(const char *) name;
1469 	} */ *uap = v;
1470 
1471 	SCARG(&ua, cmd) = SWAP_ON;
1472 	SCARG(&ua, arg) = (void *)SCARG(uap, name);
1473 	SCARG(&ua, misc) = 0;	/* priority */
1474 	return (sys_swapctl(l, &ua, retval));
1475 }
1476 
1477 /*
1478  * Stop swapping to the file or block device specified by path.
1479  */
1480 int
1481 linux_sys_swapoff(l, v, retval)
1482 	struct lwp *l;
1483 	void *v;
1484 	register_t *retval;
1485 {
1486 	struct sys_swapctl_args ua;
1487 	struct linux_sys_swapoff_args /* {
1488 		syscallarg(const char *) path;
1489 	} */ *uap = v;
1490 
1491 	SCARG(&ua, cmd) = SWAP_OFF;
1492 	SCARG(&ua, arg) = (void *)SCARG(uap, path);
1493 	return (sys_swapctl(l, &ua, retval));
1494 }
1495 
1496 /*
1497  * Copy of compat_09_sys_setdomainname()
1498  */
1499 /* ARGSUSED */
1500 int
1501 linux_sys_setdomainname(l, v, retval)
1502 	struct lwp *l;
1503 	void *v;
1504 	register_t *retval;
1505 {
1506 	struct linux_sys_setdomainname_args /* {
1507 		syscallarg(char *) domainname;
1508 		syscallarg(int) len;
1509 	} */ *uap = v;
1510 	int name[2];
1511 
1512 	name[0] = CTL_KERN;
1513 	name[1] = KERN_DOMAINNAME;
1514 	return (old_sysctl(&name[0], 2, 0, 0, SCARG(uap, domainname),
1515 			    SCARG(uap, len), l));
1516 }
1517 
1518 /*
1519  * sysinfo()
1520  */
1521 /* ARGSUSED */
1522 int
1523 linux_sys_sysinfo(l, v, retval)
1524 	struct lwp *l;
1525 	void *v;
1526 	register_t *retval;
1527 {
1528 	struct linux_sys_sysinfo_args /* {
1529 		syscallarg(struct linux_sysinfo *) arg;
1530 	} */ *uap = v;
1531 	struct linux_sysinfo si;
1532 	struct loadavg *la;
1533 
1534 	si.uptime = time.tv_sec - boottime.tv_sec;
1535 	la = &averunnable;
1536 	si.loads[0] = la->ldavg[0] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1537 	si.loads[1] = la->ldavg[1] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1538 	si.loads[2] = la->ldavg[2] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1539 	si.totalram = ctob(physmem);
1540 	si.freeram = uvmexp.free * uvmexp.pagesize;
1541 	si.sharedram = 0;	/* XXX */
1542 	si.bufferram = uvmexp.filepages * uvmexp.pagesize;
1543 	si.totalswap = uvmexp.swpages * uvmexp.pagesize;
1544 	si.freeswap = (uvmexp.swpages - uvmexp.swpginuse) * uvmexp.pagesize;
1545 	si.procs = nprocs;
1546 
1547 	/* The following are only present in newer Linux kernels. */
1548 	si.totalbig = 0;
1549 	si.freebig = 0;
1550 	si.mem_unit = 1;
1551 
1552 	return (copyout(&si, SCARG(uap, arg), sizeof si));
1553 }
1554 
1555 #define bsd_to_linux_rlimit1(l, b, f) \
1556     (l)->f = ((b)->f == RLIM_INFINITY || \
1557 	     ((b)->f & 0xffffffff00000000ULL) != 0) ? \
1558     LINUX_RLIM_INFINITY : (int32_t)(b)->f
1559 #define bsd_to_linux_rlimit(l, b) \
1560     bsd_to_linux_rlimit1(l, b, rlim_cur); \
1561     bsd_to_linux_rlimit1(l, b, rlim_max)
1562 
1563 #define linux_to_bsd_rlimit1(b, l, f) \
1564     (b)->f = (l)->f == LINUX_RLIM_INFINITY ? RLIM_INFINITY : (l)->f
1565 #define linux_to_bsd_rlimit(b, l) \
1566     linux_to_bsd_rlimit1(b, l, rlim_cur); \
1567     linux_to_bsd_rlimit1(b, l, rlim_max)
1568 
1569 static int
1570 linux_to_bsd_limit(lim)
1571 	int lim;
1572 {
1573 	switch (lim) {
1574 	case LINUX_RLIMIT_CPU:
1575 		return RLIMIT_CPU;
1576 	case LINUX_RLIMIT_FSIZE:
1577 		return RLIMIT_FSIZE;
1578 	case LINUX_RLIMIT_DATA:
1579 		return RLIMIT_DATA;
1580 	case LINUX_RLIMIT_STACK:
1581 		return RLIMIT_STACK;
1582 	case LINUX_RLIMIT_CORE:
1583 		return RLIMIT_CORE;
1584 	case LINUX_RLIMIT_RSS:
1585 		return RLIMIT_RSS;
1586 	case LINUX_RLIMIT_NPROC:
1587 		return RLIMIT_NPROC;
1588 	case LINUX_RLIMIT_NOFILE:
1589 		return RLIMIT_NOFILE;
1590 	case LINUX_RLIMIT_MEMLOCK:
1591 		return RLIMIT_MEMLOCK;
1592 	case LINUX_RLIMIT_AS:
1593 	case LINUX_RLIMIT_LOCKS:
1594 		return -EOPNOTSUPP;
1595 	default:
1596 		return -EINVAL;
1597 	}
1598 }
1599 
1600 
1601 int
1602 linux_sys_getrlimit(l, v, retval)
1603 	struct lwp *l;
1604 	void *v;
1605 	register_t *retval;
1606 {
1607 	struct linux_sys_getrlimit_args /* {
1608 		syscallarg(int) which;
1609 		syscallarg(struct orlimit *) rlp;
1610 	} */ *uap = v;
1611 	struct proc *p = l->l_proc;
1612 	caddr_t sg = stackgap_init(p, 0);
1613 	struct sys_getrlimit_args ap;
1614 	struct rlimit rl;
1615 	struct orlimit orl;
1616 	int error;
1617 
1618 	SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which));
1619 	if ((error = SCARG(&ap, which)) < 0)
1620 		return -error;
1621 	SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl);
1622 	if ((error = sys_getrlimit(l, &ap, retval)) != 0)
1623 		return error;
1624 	if ((error = copyin(SCARG(&ap, rlp), &rl, sizeof(rl))) != 0)
1625 		return error;
1626 	bsd_to_linux_rlimit(&orl, &rl);
1627 	return copyout(&orl, SCARG(uap, rlp), sizeof(orl));
1628 }
1629 
1630 int
1631 linux_sys_setrlimit(l, v, retval)
1632 	struct lwp *l;
1633 	void *v;
1634 	register_t *retval;
1635 {
1636 	struct linux_sys_setrlimit_args /* {
1637 		syscallarg(int) which;
1638 		syscallarg(struct orlimit *) rlp;
1639 	} */ *uap = v;
1640 	struct proc *p = l->l_proc;
1641 	caddr_t sg = stackgap_init(p, 0);
1642 	struct sys_setrlimit_args ap;
1643 	struct rlimit rl;
1644 	struct orlimit orl;
1645 	int error;
1646 
1647 	SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which));
1648 	SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl);
1649 	if ((error = SCARG(&ap, which)) < 0)
1650 		return -error;
1651 	if ((error = copyin(SCARG(uap, rlp), &orl, sizeof(orl))) != 0)
1652 		return error;
1653 	linux_to_bsd_rlimit(&rl, &orl);
1654 	/* XXX: alpha complains about this */
1655 	if ((error = copyout(&rl, (void *)SCARG(&ap, rlp), sizeof(rl))) != 0)
1656 		return error;
1657 	return sys_setrlimit(l, &ap, retval);
1658 }
1659 
1660 #ifndef __mips__
1661 /* XXX: this doesn't look 100% common, at least mips doesn't have it */
1662 int
1663 linux_sys_ugetrlimit(l, v, retval)
1664 	struct lwp *l;
1665 	void *v;
1666 	register_t *retval;
1667 {
1668 	return linux_sys_getrlimit(l, v, retval);
1669 }
1670 #endif
1671 
1672 /*
1673  * This gets called for unsupported syscalls. The difference to sys_nosys()
1674  * is that process does not get SIGSYS, the call just returns with ENOSYS.
1675  * This is the way Linux does it and glibc depends on this behaviour.
1676  */
1677 int
1678 linux_sys_nosys(l, v, retval)
1679 	struct lwp *l;
1680 	void *v;
1681 	register_t *retval;
1682 {
1683 	return (ENOSYS);
1684 }
1685