xref: /netbsd-src/sys/compat/linux/common/linux_misc.c (revision aa73cae19608873cc4d1f712c4a0f8f8435f1ffa)
1 /*	$NetBSD: linux_misc.c,v 1.135 2005/02/26 23:10:19 perry 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 1.135 2005/02/26 23:10:19 perry 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 const struct linux_mnttypes linux_fstypes[] = {
136 	{ MOUNT_FFS,		LINUX_DEFAULT_SUPER_MAGIC	},
137 	{ MOUNT_NFS,		LINUX_NFS_SUPER_MAGIC 		},
138 	{ MOUNT_MFS,		LINUX_DEFAULT_SUPER_MAGIC	},
139 	{ MOUNT_MSDOS,		LINUX_MSDOS_SUPER_MAGIC		},
140 	{ MOUNT_LFS,		LINUX_DEFAULT_SUPER_MAGIC	},
141 	{ MOUNT_FDESC,		LINUX_DEFAULT_SUPER_MAGIC	},
142 	{ MOUNT_PORTAL,		LINUX_DEFAULT_SUPER_MAGIC	},
143 	{ MOUNT_NULL,		LINUX_DEFAULT_SUPER_MAGIC	},
144 	{ MOUNT_OVERLAY,	LINUX_DEFAULT_SUPER_MAGIC	},
145 	{ MOUNT_UMAP,		LINUX_DEFAULT_SUPER_MAGIC	},
146 	{ MOUNT_KERNFS,		LINUX_DEFAULT_SUPER_MAGIC	},
147 	{ MOUNT_PROCFS,		LINUX_PROC_SUPER_MAGIC		},
148 	{ MOUNT_AFS,		LINUX_DEFAULT_SUPER_MAGIC	},
149 	{ MOUNT_CD9660,		LINUX_ISOFS_SUPER_MAGIC		},
150 	{ MOUNT_UNION,		LINUX_DEFAULT_SUPER_MAGIC	},
151 	{ MOUNT_ADOSFS,		LINUX_ADFS_SUPER_MAGIC		},
152 	{ MOUNT_EXT2FS,		LINUX_EXT2_SUPER_MAGIC		},
153 	{ MOUNT_CFS,		LINUX_DEFAULT_SUPER_MAGIC	},
154 	{ MOUNT_CODA,		LINUX_CODA_SUPER_MAGIC		},
155 	{ MOUNT_FILECORE,	LINUX_DEFAULT_SUPER_MAGIC	},
156 	{ MOUNT_NTFS,		LINUX_DEFAULT_SUPER_MAGIC	},
157 	{ MOUNT_SMBFS,		LINUX_SMB_SUPER_MAGIC		},
158 	{ MOUNT_PTYFS,		LINUX_DEVPTS_SUPER_MAGIC	}
159 };
160 const int linux_fstypes_cnt = sizeof(linux_fstypes) / sizeof(linux_fstypes[0]);
161 
162 #ifdef DEBUG_LINUX
163 #define DPRINTF(a)	uprintf a
164 #else
165 #define DPRINTF(a)
166 #endif
167 
168 /* Local linux_misc.c functions: */
169 static void bsd_to_linux_statfs __P((const struct statvfs *,
170     struct linux_statfs *));
171 static int linux_to_bsd_limit __P((int));
172 static void linux_to_bsd_mmap_args __P((struct sys_mmap_args *,
173     const struct linux_sys_mmap_args *));
174 static int linux_mmap __P((struct lwp *, struct linux_sys_mmap_args *,
175     register_t *, off_t));
176 
177 
178 /*
179  * The information on a terminated (or stopped) process needs
180  * to be converted in order for Linux binaries to get a valid signal
181  * number out of it.
182  */
183 void
184 bsd_to_linux_wstat(st)
185 	int *st;
186 {
187 
188 	int sig;
189 
190 	if (WIFSIGNALED(*st)) {
191 		sig = WTERMSIG(*st);
192 		if (sig >= 0 && sig < NSIG)
193 			*st= (*st& ~0177) | native_to_linux_signo[sig];
194 	} else if (WIFSTOPPED(*st)) {
195 		sig = WSTOPSIG(*st);
196 		if (sig >= 0 && sig < NSIG)
197 			*st = (*st & ~0xff00) |
198 			    (native_to_linux_signo[sig] << 8);
199 	}
200 }
201 
202 /*
203  * wait4(2).  Passed on to the NetBSD call, surrounded by code to
204  * reserve some space for a NetBSD-style wait status, and converting
205  * it to what Linux wants.
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 & ~(LINUX_WAIT4_KNOWNFLAGS))
233 		return (EINVAL);
234 
235 	if (linux_options & LINUX_WAIT4_WNOHANG)
236 		options |= WNOHANG;
237 	if (linux_options & LINUX_WAIT4_WUNTRACED)
238 		options |= WUNTRACED;
239 	if (linux_options & LINUX_WAIT4_WALL)
240 		options |= WALLSIG;
241 	if (linux_options & LINUX_WAIT4_WCLONE)
242 		options |= WALTSIG;
243 #ifdef DIAGNOSTIC
244 	if (linux_options & LINUX_WAIT4_WNOTHREAD)
245 		printf("WARNING: %s: linux process %d.%d called "
246 		       "waitpid with __WNOTHREAD set!",
247 		       __FILE__, p->p_pid, l->l_lid);
248 
249 #endif
250 
251 	SCARG(&w4a, pid) = SCARG(uap, pid);
252 	SCARG(&w4a, status) = status;
253 	SCARG(&w4a, options) = options;
254 	SCARG(&w4a, rusage) = SCARG(uap, rusage);
255 
256 	if ((error = sys_wait4(l, &w4a, retval)))
257 		return error;
258 
259 	sigdelset(&p->p_sigctx.ps_siglist, SIGCHLD);
260 
261 	if (status != NULL) {
262 		if ((error = copyin(status, &tstat, sizeof tstat)))
263 			return error;
264 
265 		bsd_to_linux_wstat(&tstat);
266 		return copyout(&tstat, SCARG(uap, status), sizeof tstat);
267 	}
268 
269 	return 0;
270 }
271 
272 /*
273  * Linux brk(2). The check if the new address is >= the old one is
274  * done in the kernel in Linux. NetBSD does it in the library.
275  */
276 int
277 linux_sys_brk(l, v, retval)
278 	struct lwp *l;
279 	void *v;
280 	register_t *retval;
281 {
282 	struct linux_sys_brk_args /* {
283 		syscallarg(char *) nsize;
284 	} */ *uap = v;
285 	struct proc *p = l->l_proc;
286 	char *nbrk = SCARG(uap, nsize);
287 	struct sys_obreak_args oba;
288 	struct vmspace *vm = p->p_vmspace;
289 	struct linux_emuldata *ed = (struct linux_emuldata*)p->p_emuldata;
290 
291 	SCARG(&oba, nsize) = nbrk;
292 
293 	if ((caddr_t) nbrk > vm->vm_daddr && sys_obreak(l, &oba, retval) == 0)
294 		ed->s->p_break = (char*)nbrk;
295 	else
296 		nbrk = ed->s->p_break;
297 
298 	retval[0] = (register_t)nbrk;
299 
300 	return 0;
301 }
302 
303 /*
304  * Convert NetBSD statvfs structure to Linux statfs structure.
305  * Linux doesn't have f_flag, and we can't set f_frsize due
306  * to glibc statvfs() bug (see below).
307  */
308 static void
309 bsd_to_linux_statfs(bsp, lsp)
310 	const struct statvfs *bsp;
311 	struct linux_statfs *lsp;
312 {
313 	int i;
314 
315 	for (i = 0; i < linux_fstypes_cnt; i++) {
316 		if (strcmp(bsp->f_fstypename, linux_fstypes[i].bsd) == 0) {
317 			lsp->l_ftype = linux_fstypes[i].linux;
318 			break;
319 		}
320 	}
321 
322 	if (i == linux_fstypes_cnt) {
323 		DPRINTF(("unhandled fstype in linux emulation: %s\n",
324 		    bsp->f_fstypename));
325 		lsp->l_ftype = LINUX_DEFAULT_SUPER_MAGIC;
326 	}
327 
328 	/*
329 	 * The sizes are expressed in number of blocks. The block
330 	 * size used for the size is f_frsize for POSIX-compliant
331 	 * statvfs. Linux statfs uses f_bsize as the block size
332 	 * (f_frsize used to not be available in Linux struct statfs).
333 	 * However, glibc 2.3.3 statvfs() wrapper fails to adjust the block
334 	 * counts for different f_frsize if f_frsize is provided by the kernel.
335 	 * POSIX conforming apps thus get wrong size if f_frsize
336 	 * is different to f_bsize. Thus, we just pretend we don't
337 	 * support f_frsize.
338 	 */
339 
340 	lsp->l_fbsize = bsp->f_frsize;
341 	lsp->l_ffrsize = 0;			/* compat */
342 	lsp->l_fblocks = bsp->f_blocks;
343 	lsp->l_fbfree = bsp->f_bfree;
344 	lsp->l_fbavail = bsp->f_bavail;
345 	lsp->l_ffiles = bsp->f_files;
346 	lsp->l_fffree = bsp->f_ffree;
347 	/* Linux sets the fsid to 0..., we don't */
348 	lsp->l_ffsid.val[0] = bsp->f_fsidx.__fsid_val[0];
349 	lsp->l_ffsid.val[1] = bsp->f_fsidx.__fsid_val[1];
350 	lsp->l_fnamelen = bsp->f_namemax;
351 	(void)memset(lsp->l_fspare, 0, sizeof(lsp->l_fspare));
352 }
353 
354 /*
355  * Implement the fs stat functions. Straightforward.
356  */
357 int
358 linux_sys_statfs(l, v, retval)
359 	struct lwp *l;
360 	void *v;
361 	register_t *retval;
362 {
363 	struct linux_sys_statfs_args /* {
364 		syscallarg(const char *) path;
365 		syscallarg(struct linux_statfs *) sp;
366 	} */ *uap = v;
367 	struct proc *p = l->l_proc;
368 	struct statvfs btmp, *bsp;
369 	struct linux_statfs ltmp;
370 	struct sys_statvfs1_args bsa;
371 	caddr_t sg;
372 	int error;
373 
374 	sg = stackgap_init(p, 0);
375 	bsp = (struct statvfs *) stackgap_alloc(p, &sg, sizeof (struct statvfs));
376 
377 	CHECK_ALT_EXIST(p, &sg, SCARG(uap, path));
378 
379 	SCARG(&bsa, path) = SCARG(uap, path);
380 	SCARG(&bsa, buf) = bsp;
381 	SCARG(&bsa, flags) = ST_WAIT;
382 
383 	if ((error = sys_statvfs1(l, &bsa, retval)))
384 		return error;
385 
386 	if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp)))
387 		return error;
388 
389 	bsd_to_linux_statfs(&btmp, &ltmp);
390 
391 	return copyout((caddr_t) &ltmp, (caddr_t) SCARG(uap, sp), sizeof ltmp);
392 }
393 
394 int
395 linux_sys_fstatfs(l, v, retval)
396 	struct lwp *l;
397 	void *v;
398 	register_t *retval;
399 {
400 	struct linux_sys_fstatfs_args /* {
401 		syscallarg(int) fd;
402 		syscallarg(struct linux_statfs *) sp;
403 	} */ *uap = v;
404 	struct proc *p = l->l_proc;
405 	struct statvfs btmp, *bsp;
406 	struct linux_statfs ltmp;
407 	struct sys_fstatvfs1_args bsa;
408 	caddr_t sg;
409 	int error;
410 
411 	sg = stackgap_init(p, 0);
412 	bsp = (struct statvfs *) stackgap_alloc(p, &sg, sizeof (struct statvfs));
413 
414 	SCARG(&bsa, fd) = SCARG(uap, fd);
415 	SCARG(&bsa, buf) = bsp;
416 	SCARG(&bsa, flags) = ST_WAIT;
417 
418 	if ((error = sys_fstatvfs1(l, &bsa, retval)))
419 		return error;
420 
421 	if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp)))
422 		return error;
423 
424 	bsd_to_linux_statfs(&btmp, &ltmp);
425 
426 	return copyout((caddr_t) &ltmp, (caddr_t) SCARG(uap, sp), sizeof ltmp);
427 }
428 
429 /*
430  * uname(). Just copy the info from the various strings stored in the
431  * kernel, and put it in the Linux utsname structure. That structure
432  * is almost the same as the NetBSD one, only it has fields 65 characters
433  * long, and an extra domainname field.
434  */
435 int
436 linux_sys_uname(l, v, retval)
437 	struct lwp *l;
438 	void *v;
439 	register_t *retval;
440 {
441 	struct linux_sys_uname_args /* {
442 		syscallarg(struct linux_utsname *) up;
443 	} */ *uap = v;
444 	struct linux_utsname luts;
445 
446 	strncpy(luts.l_sysname, linux_sysname, sizeof(luts.l_sysname));
447 	strncpy(luts.l_nodename, hostname, sizeof(luts.l_nodename));
448 	strncpy(luts.l_release, linux_release, sizeof(luts.l_release));
449 	strncpy(luts.l_version, linux_version, sizeof(luts.l_version));
450 	strncpy(luts.l_machine, machine, sizeof(luts.l_machine));
451 	strncpy(luts.l_domainname, domainname, sizeof(luts.l_domainname));
452 
453 	return copyout(&luts, SCARG(uap, up), sizeof(luts));
454 }
455 
456 /* Used directly on: alpha, mips, ppc, sparc, sparc64 */
457 /* Used indirectly on: arm, i386, m68k */
458 
459 /*
460  * New type Linux mmap call.
461  * Only called directly on machines with >= 6 free regs.
462  */
463 int
464 linux_sys_mmap(l, v, retval)
465 	struct lwp *l;
466 	void *v;
467 	register_t *retval;
468 {
469 	struct linux_sys_mmap_args /* {
470 		syscallarg(unsigned long) addr;
471 		syscallarg(size_t) len;
472 		syscallarg(int) prot;
473 		syscallarg(int) flags;
474 		syscallarg(int) fd;
475 		syscallarg(linux_off_t) offset;
476 	} */ *uap = v;
477 
478 	if (SCARG(uap, offset) & PAGE_MASK)
479 		return EINVAL;
480 
481 	return linux_mmap(l, uap, retval, SCARG(uap, offset));
482 }
483 
484 /*
485  * Guts of most architectures' mmap64() implementations.  This shares
486  * its list of arguments with linux_sys_mmap().
487  *
488  * The difference in linux_sys_mmap2() is that "offset" is actually
489  * (offset / pagesize), not an absolute byte count.  This translation
490  * to pagesize offsets is done inside glibc between the mmap64() call
491  * point, and the actual syscall.
492  */
493 int
494 linux_sys_mmap2(l, v, retval)
495 	struct lwp *l;
496 	void *v;
497 	register_t *retval;
498 {
499 	struct linux_sys_mmap2_args /* {
500 		syscallarg(unsigned long) addr;
501 		syscallarg(size_t) len;
502 		syscallarg(int) prot;
503 		syscallarg(int) flags;
504 		syscallarg(int) fd;
505 		syscallarg(linux_off_t) offset;
506 	} */ *uap = v;
507 
508 	return linux_mmap(l, uap, retval,
509 	    ((off_t)SCARG(uap, offset)) << PAGE_SHIFT);
510 }
511 
512 /*
513  * Massage arguments and call system mmap(2).
514  */
515 static int
516 linux_mmap(l, uap, retval, offset)
517 	struct lwp *l;
518 	struct linux_sys_mmap_args *uap;
519 	register_t *retval;
520 	off_t offset;
521 {
522 	struct sys_mmap_args cma;
523 	int error;
524 	size_t mmoff=0;
525 
526 	if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN) {
527 		/*
528 		 * Request for stack-like memory segment. On linux, this
529 		 * works by mmap()ping (small) segment, which is automatically
530 		 * extended when page fault happens below the currently
531 		 * allocated area. We emulate this by allocating (typically
532 		 * bigger) segment sized at current stack size limit, and
533 		 * offsetting the requested and returned address accordingly.
534 		 * Since physical pages are only allocated on-demand, this
535 		 * is effectively identical.
536 		 */
537 		rlim_t ssl = l->l_proc->p_rlimit[RLIMIT_STACK].rlim_cur;
538 
539 		if (SCARG(uap, len) < ssl) {
540 			/* Compute the address offset */
541 			mmoff = round_page(ssl) - SCARG(uap, len);
542 
543 			if (SCARG(uap, addr))
544 				SCARG(uap, addr) -= mmoff;
545 
546 			SCARG(uap, len) = (size_t) ssl;
547 		}
548 	}
549 
550 	linux_to_bsd_mmap_args(&cma, uap);
551 	SCARG(&cma, pos) = offset;
552 
553 	error = sys_mmap(l, &cma, retval);
554 	if (error)
555 		return (error);
556 
557 	/* Shift the returned address for stack-like segment if necessary */
558 	if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN && mmoff)
559 		retval[0] += mmoff;
560 
561 	return (0);
562 }
563 
564 static void
565 linux_to_bsd_mmap_args(cma, uap)
566 	struct sys_mmap_args *cma;
567 	const struct linux_sys_mmap_args *uap;
568 {
569 	int flags = MAP_TRYFIXED, fl = SCARG(uap, flags);
570 
571 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_SHARED, MAP_SHARED);
572 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_PRIVATE, MAP_PRIVATE);
573 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_FIXED, MAP_FIXED);
574 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_ANON, MAP_ANON);
575 	/* XXX XAX ERH: Any other flags here?  There are more defined... */
576 
577 	SCARG(cma, addr) = (void *)SCARG(uap, addr);
578 	SCARG(cma, len) = SCARG(uap, len);
579 	SCARG(cma, prot) = SCARG(uap, prot);
580 	if (SCARG(cma, prot) & VM_PROT_WRITE) /* XXX */
581 		SCARG(cma, prot) |= VM_PROT_READ;
582 	SCARG(cma, flags) = flags;
583 	SCARG(cma, fd) = flags & MAP_ANON ? -1 : SCARG(uap, fd);
584 	SCARG(cma, pad) = 0;
585 }
586 
587 int
588 linux_sys_mremap(l, v, retval)
589 	struct lwp *l;
590 	void *v;
591 	register_t *retval;
592 {
593 	struct linux_sys_mremap_args /* {
594 		syscallarg(void *) old_address;
595 		syscallarg(size_t) old_size;
596 		syscallarg(size_t) new_size;
597 		syscallarg(u_long) flags;
598 	} */ *uap = v;
599 	struct sys_munmap_args mua;
600 	size_t old_size, new_size;
601 	int error;
602 
603 	old_size = round_page(SCARG(uap, old_size));
604 	new_size = round_page(SCARG(uap, new_size));
605 
606 	/*
607 	 * Growing mapped region.
608 	 */
609 	if (new_size > old_size) {
610 		/*
611 		 * XXX Implement me.  What we probably want to do is
612 		 * XXX dig out the guts of the old mapping, mmap that
613 		 * XXX object again with the new size, then munmap
614 		 * XXX the old mapping.
615 		 */
616 		*retval = 0;
617 		return (ENOMEM);
618 	}
619 
620 	/*
621 	 * Shrinking mapped region.
622 	 */
623 	if (new_size < old_size) {
624 		SCARG(&mua, addr) = (caddr_t)SCARG(uap, old_address) +
625 		    new_size;
626 		SCARG(&mua, len) = old_size - new_size;
627 		error = sys_munmap(l, &mua, retval);
628 		*retval = error ? 0 : (register_t)SCARG(uap, old_address);
629 		return (error);
630 	}
631 
632 	/*
633 	 * No change.
634 	 */
635 	*retval = (register_t)SCARG(uap, old_address);
636 	return (0);
637 }
638 
639 int
640 linux_sys_msync(l, v, retval)
641 	struct lwp *l;
642 	void *v;
643 	register_t *retval;
644 {
645 	struct linux_sys_msync_args /* {
646 		syscallarg(caddr_t) addr;
647 		syscallarg(int) len;
648 		syscallarg(int) fl;
649 	} */ *uap = v;
650 
651 	struct sys___msync13_args bma;
652 
653 	/* flags are ignored */
654 	SCARG(&bma, addr) = SCARG(uap, addr);
655 	SCARG(&bma, len) = SCARG(uap, len);
656 	SCARG(&bma, flags) = SCARG(uap, fl);
657 
658 	return sys___msync13(l, &bma, retval);
659 }
660 
661 int
662 linux_sys_mprotect(l, v, retval)
663 	struct lwp *l;
664 	void *v;
665 	register_t *retval;
666 {
667 	struct linux_sys_mprotect_args /* {
668 		syscallarg(const void *) start;
669 		syscallarg(unsigned long) len;
670 		syscallarg(int) prot;
671 	} */ *uap = v;
672 	unsigned long end, start = (unsigned long)SCARG(uap, start), len;
673 	int prot = SCARG(uap, prot);
674 	struct vm_map_entry *entry;
675 	struct vm_map *map = &l->l_proc->p_vmspace->vm_map;
676 
677 	if (start & PAGE_MASK)
678 		return EINVAL;
679 
680 	len = round_page(SCARG(uap, len));
681 	end = start + len;
682 
683 	if (end < start)
684 		return EINVAL;
685 	else if (end == start)
686 		return 0;
687 
688 	if (SCARG(uap, prot) & ~(PROT_READ | PROT_WRITE | PROT_EXEC))
689 		return EINVAL;
690 
691 	vm_map_lock(map);
692 #ifdef notdef
693 	VM_MAP_RANGE_CHECK(map, start, end);
694 #endif
695 	if (!uvm_map_lookup_entry(map, start, &entry) || entry->start > start) {
696 		vm_map_unlock(map);
697 		return ENOMEM;
698 	}
699 	vm_map_unlock(map);
700 	return uvm_map_protect(map, start, end, prot, FALSE);
701 }
702 
703 /*
704  * This code is partly stolen from src/lib/libc/compat-43/times.c
705  */
706 
707 #define	CONVTCK(r)	(r.tv_sec * hz + r.tv_usec / (1000000 / hz))
708 
709 int
710 linux_sys_times(l, v, retval)
711 	struct lwp *l;
712 	void *v;
713 	register_t *retval;
714 {
715 	struct linux_sys_times_args /* {
716 		syscallarg(struct times *) tms;
717 	} */ *uap = v;
718 	struct proc *p = l->l_proc;
719 	struct timeval t;
720 	int error, s;
721 
722 	if (SCARG(uap, tms)) {
723 		struct linux_tms ltms;
724 		struct rusage ru;
725 
726 		calcru(p, &ru.ru_utime, &ru.ru_stime, NULL);
727 		ltms.ltms_utime = CONVTCK(ru.ru_utime);
728 		ltms.ltms_stime = CONVTCK(ru.ru_stime);
729 
730 		ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime);
731 		ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime);
732 
733 		if ((error = copyout(&ltms, SCARG(uap, tms), sizeof ltms)))
734 			return error;
735 	}
736 
737 	s = splclock();
738 	timersub(&time, &boottime, &t);
739 	splx(s);
740 
741 	retval[0] = ((linux_clock_t)(CONVTCK(t)));
742 	return 0;
743 }
744 
745 #undef CONVTCK
746 
747 /*
748  * Linux 'readdir' call. This code is mostly taken from the
749  * SunOS getdents call (see compat/sunos/sunos_misc.c), though
750  * an attempt has been made to keep it a little cleaner (failing
751  * miserably, because of the cruft needed if count 1 is passed).
752  *
753  * The d_off field should contain the offset of the next valid entry,
754  * but in Linux it has the offset of the entry itself. We emulate
755  * that bug here.
756  *
757  * Read in BSD-style entries, convert them, and copy them out.
758  *
759  * Note that this doesn't handle union-mounted filesystems.
760  */
761 int
762 linux_sys_getdents(l, v, retval)
763 	struct lwp *l;
764 	void *v;
765 	register_t *retval;
766 {
767 	struct linux_sys_getdents_args /* {
768 		syscallarg(int) fd;
769 		syscallarg(struct linux_dirent *) dent;
770 		syscallarg(unsigned int) count;
771 	} */ *uap = v;
772 	struct proc *p = l->l_proc;
773 	struct dirent *bdp;
774 	struct vnode *vp;
775 	caddr_t	inp, buf;		/* BSD-format */
776 	int len, reclen;		/* BSD-format */
777 	caddr_t outp;			/* Linux-format */
778 	int resid, linux_reclen = 0;	/* Linux-format */
779 	struct file *fp;
780 	struct uio auio;
781 	struct iovec aiov;
782 	struct linux_dirent idb;
783 	off_t off;		/* true file offset */
784 	int buflen, error, eofflag, nbytes, oldcall;
785 	struct vattr va;
786 	off_t *cookiebuf = NULL, *cookie;
787 	int ncookies;
788 
789 	/* getvnode() will use the descriptor for us */
790 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
791 		return (error);
792 
793 	if ((fp->f_flag & FREAD) == 0) {
794 		error = EBADF;
795 		goto out1;
796 	}
797 
798 	vp = (struct vnode *)fp->f_data;
799 	if (vp->v_type != VDIR) {
800 		error = EINVAL;
801 		goto out1;
802 	}
803 
804 	if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)))
805 		goto out1;
806 
807 	nbytes = SCARG(uap, count);
808 	if (nbytes == 1) {	/* emulating old, broken behaviour */
809 		nbytes = sizeof (idb);
810 		buflen = max(va.va_blocksize, nbytes);
811 		oldcall = 1;
812 	} else {
813 		buflen = min(MAXBSIZE, nbytes);
814 		if (buflen < va.va_blocksize)
815 			buflen = va.va_blocksize;
816 		oldcall = 0;
817 	}
818 	buf = malloc(buflen, M_TEMP, M_WAITOK);
819 
820 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
821 	off = fp->f_offset;
822 again:
823 	aiov.iov_base = buf;
824 	aiov.iov_len = buflen;
825 	auio.uio_iov = &aiov;
826 	auio.uio_iovcnt = 1;
827 	auio.uio_rw = UIO_READ;
828 	auio.uio_segflg = UIO_SYSSPACE;
829 	auio.uio_procp = NULL;
830 	auio.uio_resid = buflen;
831 	auio.uio_offset = off;
832 	/*
833          * First we read into the malloc'ed buffer, then
834          * we massage it into user space, one record at a time.
835          */
836 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf,
837 	    &ncookies);
838 	if (error)
839 		goto out;
840 
841 	inp = buf;
842 	outp = (caddr_t)SCARG(uap, dent);
843 	resid = nbytes;
844 	if ((len = buflen - auio.uio_resid) == 0)
845 		goto eof;
846 
847 	for (cookie = cookiebuf; len > 0; len -= reclen) {
848 		bdp = (struct dirent *)inp;
849 		reclen = bdp->d_reclen;
850 		if (reclen & 3)
851 			panic("linux_readdir");
852 		if (bdp->d_fileno == 0) {
853 			inp += reclen;	/* it is a hole; squish it out */
854 			off = *cookie++;
855 			continue;
856 		}
857 		linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen);
858 		if (reclen > len || resid < linux_reclen) {
859 			/* entry too big for buffer, so just stop */
860 			outp++;
861 			break;
862 		}
863 		/*
864 		 * Massage in place to make a Linux-shaped dirent (otherwise
865 		 * we have to worry about touching user memory outside of
866 		 * the copyout() call).
867 		 */
868 		idb.d_ino = bdp->d_fileno;
869 		/*
870 		 * The old readdir() call misuses the offset and reclen fields.
871 		 */
872 		if (oldcall) {
873 			idb.d_off = (linux_off_t)linux_reclen;
874 			idb.d_reclen = (u_short)bdp->d_namlen;
875 		} else {
876 			if (sizeof (idb.d_off) <= 4 && (off >> 32) != 0) {
877 				compat_offseterr(vp, "linux_getdents");
878 				error = EINVAL;
879 				goto out;
880 			}
881 			idb.d_off = (linux_off_t)off;
882 			idb.d_reclen = (u_short)linux_reclen;
883 		}
884 		strcpy(idb.d_name, bdp->d_name);
885 		if ((error = copyout((caddr_t)&idb, outp, linux_reclen)))
886 			goto out;
887 		/* advance past this real entry */
888 		inp += reclen;
889 		off = *cookie++;	/* each entry points to itself */
890 		/* advance output past Linux-shaped entry */
891 		outp += linux_reclen;
892 		resid -= linux_reclen;
893 		if (oldcall)
894 			break;
895 	}
896 
897 	/* if we squished out the whole block, try again */
898 	if (outp == (caddr_t)SCARG(uap, dent))
899 		goto again;
900 	fp->f_offset = off;	/* update the vnode offset */
901 
902 	if (oldcall)
903 		nbytes = resid + linux_reclen;
904 
905 eof:
906 	*retval = nbytes - resid;
907 out:
908 	VOP_UNLOCK(vp, 0);
909 	if (cookiebuf)
910 		free(cookiebuf, M_TEMP);
911 	free(buf, M_TEMP);
912 out1:
913 	FILE_UNUSE(fp, p);
914 	return error;
915 }
916 
917 /*
918  * Even when just using registers to pass arguments to syscalls you can
919  * have 5 of them on the i386. So this newer version of select() does
920  * this.
921  */
922 int
923 linux_sys_select(l, v, retval)
924 	struct lwp *l;
925 	void *v;
926 	register_t *retval;
927 {
928 	struct linux_sys_select_args /* {
929 		syscallarg(int) nfds;
930 		syscallarg(fd_set *) readfds;
931 		syscallarg(fd_set *) writefds;
932 		syscallarg(fd_set *) exceptfds;
933 		syscallarg(struct timeval *) timeout;
934 	} */ *uap = v;
935 
936 	return linux_select1(l, retval, SCARG(uap, nfds), SCARG(uap, readfds),
937 	    SCARG(uap, writefds), SCARG(uap, exceptfds), SCARG(uap, timeout));
938 }
939 
940 /*
941  * Common code for the old and new versions of select(). A couple of
942  * things are important:
943  * 1) return the amount of time left in the 'timeout' parameter
944  * 2) select never returns ERESTART on Linux, always return EINTR
945  */
946 int
947 linux_select1(l, retval, nfds, readfds, writefds, exceptfds, timeout)
948 	struct lwp *l;
949 	register_t *retval;
950 	int nfds;
951 	fd_set *readfds, *writefds, *exceptfds;
952 	struct timeval *timeout;
953 {
954 	struct sys_select_args bsa;
955 	struct proc *p = l->l_proc;
956 	struct timeval tv0, tv1, utv, *tvp;
957 	caddr_t sg;
958 	int error;
959 
960 	SCARG(&bsa, nd) = nfds;
961 	SCARG(&bsa, in) = readfds;
962 	SCARG(&bsa, ou) = writefds;
963 	SCARG(&bsa, ex) = exceptfds;
964 	SCARG(&bsa, tv) = timeout;
965 
966 	/*
967 	 * Store current time for computation of the amount of
968 	 * time left.
969 	 */
970 	if (timeout) {
971 		if ((error = copyin(timeout, &utv, sizeof(utv))))
972 			return error;
973 		if (itimerfix(&utv)) {
974 			/*
975 			 * The timeval was invalid.  Convert it to something
976 			 * valid that will act as it does under Linux.
977 			 */
978 			sg = stackgap_init(p, 0);
979 			tvp = stackgap_alloc(p, &sg, sizeof(utv));
980 			utv.tv_sec += utv.tv_usec / 1000000;
981 			utv.tv_usec %= 1000000;
982 			if (utv.tv_usec < 0) {
983 				utv.tv_sec -= 1;
984 				utv.tv_usec += 1000000;
985 			}
986 			if (utv.tv_sec < 0)
987 				timerclear(&utv);
988 			if ((error = copyout(&utv, tvp, sizeof(utv))))
989 				return error;
990 			SCARG(&bsa, tv) = tvp;
991 		}
992 		microtime(&tv0);
993 	}
994 
995 	error = sys_select(l, &bsa, retval);
996 	if (error) {
997 		/*
998 		 * See fs/select.c in the Linux kernel.  Without this,
999 		 * Maelstrom doesn't work.
1000 		 */
1001 		if (error == ERESTART)
1002 			error = EINTR;
1003 		return error;
1004 	}
1005 
1006 	if (timeout) {
1007 		if (*retval) {
1008 			/*
1009 			 * Compute how much time was left of the timeout,
1010 			 * by subtracting the current time and the time
1011 			 * before we started the call, and subtracting
1012 			 * that result from the user-supplied value.
1013 			 */
1014 			microtime(&tv1);
1015 			timersub(&tv1, &tv0, &tv1);
1016 			timersub(&utv, &tv1, &utv);
1017 			if (utv.tv_sec < 0)
1018 				timerclear(&utv);
1019 		} else
1020 			timerclear(&utv);
1021 		if ((error = copyout(&utv, timeout, sizeof(utv))))
1022 			return error;
1023 	}
1024 
1025 	return 0;
1026 }
1027 
1028 /*
1029  * Get the process group of a certain process. Look it up
1030  * and return the value.
1031  */
1032 int
1033 linux_sys_getpgid(l, v, retval)
1034 	struct lwp *l;
1035 	void *v;
1036 	register_t *retval;
1037 {
1038 	struct linux_sys_getpgid_args /* {
1039 		syscallarg(int) pid;
1040 	} */ *uap = v;
1041 	struct proc *p = l->l_proc;
1042 	struct proc *targp;
1043 
1044 	if (SCARG(uap, pid) != 0 && SCARG(uap, pid) != p->p_pid) {
1045 		if ((targp = pfind(SCARG(uap, pid))) == 0)
1046 			return ESRCH;
1047 	}
1048 	else
1049 		targp = p;
1050 
1051 	retval[0] = targp->p_pgid;
1052 	return 0;
1053 }
1054 
1055 /*
1056  * Set the 'personality' (emulation mode) for the current process. Only
1057  * accept the Linux personality here (0). This call is needed because
1058  * the Linux ELF crt0 issues it in an ugly kludge to make sure that
1059  * ELF binaries run in Linux mode, not SVR4 mode.
1060  */
1061 int
1062 linux_sys_personality(l, v, retval)
1063 	struct lwp *l;
1064 	void *v;
1065 	register_t *retval;
1066 {
1067 	struct linux_sys_personality_args /* {
1068 		syscallarg(int) per;
1069 	} */ *uap = v;
1070 
1071 	if (SCARG(uap, per) != 0)
1072 		return EINVAL;
1073 	retval[0] = 0;
1074 	return 0;
1075 }
1076 
1077 #if defined(__i386__) || defined(__m68k__)
1078 /*
1079  * The calls are here because of type conversions.
1080  */
1081 int
1082 linux_sys_setreuid16(l, v, retval)
1083 	struct lwp *l;
1084 	void *v;
1085 	register_t *retval;
1086 {
1087 	struct linux_sys_setreuid16_args /* {
1088 		syscallarg(int) ruid;
1089 		syscallarg(int) euid;
1090 	} */ *uap = v;
1091 	struct sys_setreuid_args bsa;
1092 
1093 	SCARG(&bsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
1094 		(uid_t)-1 : SCARG(uap, ruid);
1095 	SCARG(&bsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
1096 		(uid_t)-1 : SCARG(uap, euid);
1097 
1098 	return sys_setreuid(l, &bsa, retval);
1099 }
1100 
1101 int
1102 linux_sys_setregid16(l, v, retval)
1103 	struct lwp *l;
1104 	void *v;
1105 	register_t *retval;
1106 {
1107 	struct linux_sys_setregid16_args /* {
1108 		syscallarg(int) rgid;
1109 		syscallarg(int) egid;
1110 	} */ *uap = v;
1111 	struct sys_setregid_args bsa;
1112 
1113 	SCARG(&bsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
1114 		(uid_t)-1 : SCARG(uap, rgid);
1115 	SCARG(&bsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
1116 		(uid_t)-1 : SCARG(uap, egid);
1117 
1118 	return sys_setregid(l, &bsa, retval);
1119 }
1120 
1121 int
1122 linux_sys_setresuid16(l, v, retval)
1123 	struct lwp *l;
1124 	void *v;
1125 	register_t *retval;
1126 {
1127 	struct linux_sys_setresuid16_args /* {
1128 		syscallarg(uid_t) ruid;
1129 		syscallarg(uid_t) euid;
1130 		syscallarg(uid_t) suid;
1131 	} */ *uap = v;
1132 	struct linux_sys_setresuid16_args lsa;
1133 
1134 	SCARG(&lsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
1135 		(uid_t)-1 : SCARG(uap, ruid);
1136 	SCARG(&lsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
1137 		(uid_t)-1 : SCARG(uap, euid);
1138 	SCARG(&lsa, suid) = ((linux_uid_t)SCARG(uap, suid) == (linux_uid_t)-1) ?
1139 		(uid_t)-1 : SCARG(uap, suid);
1140 
1141 	return linux_sys_setresuid(l, &lsa, retval);
1142 }
1143 
1144 int
1145 linux_sys_setresgid16(l, v, retval)
1146 	struct lwp *l;
1147 	void *v;
1148 	register_t *retval;
1149 {
1150 	struct linux_sys_setresgid16_args /* {
1151 		syscallarg(gid_t) rgid;
1152 		syscallarg(gid_t) egid;
1153 		syscallarg(gid_t) sgid;
1154 	} */ *uap = v;
1155 	struct linux_sys_setresgid16_args lsa;
1156 
1157 	SCARG(&lsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
1158 		(gid_t)-1 : SCARG(uap, rgid);
1159 	SCARG(&lsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
1160 		(gid_t)-1 : SCARG(uap, egid);
1161 	SCARG(&lsa, sgid) = ((linux_gid_t)SCARG(uap, sgid) == (linux_gid_t)-1) ?
1162 		(gid_t)-1 : SCARG(uap, sgid);
1163 
1164 	return linux_sys_setresgid(l, &lsa, retval);
1165 }
1166 
1167 int
1168 linux_sys_getgroups16(l, v, retval)
1169 	struct lwp *l;
1170 	void *v;
1171 	register_t *retval;
1172 {
1173 	struct linux_sys_getgroups16_args /* {
1174 		syscallarg(int) gidsetsize;
1175 		syscallarg(linux_gid_t *) gidset;
1176 	} */ *uap = v;
1177 	struct proc *p = l->l_proc;
1178 	caddr_t sg;
1179 	int n, error, i;
1180 	struct sys_getgroups_args bsa;
1181 	gid_t *bset, *kbset;
1182 	linux_gid_t *lset;
1183 	struct pcred *pc = p->p_cred;
1184 
1185 	n = SCARG(uap, gidsetsize);
1186 	if (n < 0)
1187 		return EINVAL;
1188 	error = 0;
1189 	bset = kbset = NULL;
1190 	lset = NULL;
1191 	if (n > 0) {
1192 		n = min(pc->pc_ucred->cr_ngroups, n);
1193 		sg = stackgap_init(p, 0);
1194 		bset = stackgap_alloc(p, &sg, n * sizeof (gid_t));
1195 		kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK);
1196 		lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
1197 		if (bset == NULL || kbset == NULL || lset == NULL)
1198 			return ENOMEM;
1199 		SCARG(&bsa, gidsetsize) = n;
1200 		SCARG(&bsa, gidset) = bset;
1201 		error = sys_getgroups(l, &bsa, retval);
1202 		if (error != 0)
1203 			goto out;
1204 		error = copyin(bset, kbset, n * sizeof (gid_t));
1205 		if (error != 0)
1206 			goto out;
1207 		for (i = 0; i < n; i++)
1208 			lset[i] = (linux_gid_t)kbset[i];
1209 		error = copyout(lset, SCARG(uap, gidset),
1210 		    n * sizeof (linux_gid_t));
1211 	} else
1212 		*retval = pc->pc_ucred->cr_ngroups;
1213 out:
1214 	if (kbset != NULL)
1215 		free(kbset, M_TEMP);
1216 	if (lset != NULL)
1217 		free(lset, M_TEMP);
1218 	return error;
1219 }
1220 
1221 int
1222 linux_sys_setgroups16(l, v, retval)
1223 	struct lwp *l;
1224 	void *v;
1225 	register_t *retval;
1226 {
1227 	struct linux_sys_setgroups16_args /* {
1228 		syscallarg(int) gidsetsize;
1229 		syscallarg(linux_gid_t *) gidset;
1230 	} */ *uap = v;
1231 	struct proc *p = l->l_proc;
1232 	caddr_t sg;
1233 	int n;
1234 	int error, i;
1235 	struct sys_setgroups_args bsa;
1236 	gid_t *bset, *kbset;
1237 	linux_gid_t *lset;
1238 
1239 	n = SCARG(uap, gidsetsize);
1240 	if (n < 0 || n > NGROUPS)
1241 		return EINVAL;
1242 	sg = stackgap_init(p, 0);
1243 	bset = stackgap_alloc(p, &sg, n * sizeof (gid_t));
1244 	lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
1245 	kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK);
1246 	if (lset == NULL || bset == NULL)
1247 		return ENOMEM;
1248 	error = copyin(SCARG(uap, gidset), lset, n * sizeof (linux_gid_t));
1249 	if (error != 0)
1250 		goto out;
1251 	for (i = 0; i < n; i++)
1252 		kbset[i] = (gid_t)lset[i];
1253 	error = copyout(kbset, bset, n * sizeof (gid_t));
1254 	if (error != 0)
1255 		goto out;
1256 	SCARG(&bsa, gidsetsize) = n;
1257 	SCARG(&bsa, gidset) = bset;
1258 	error = sys_setgroups(l, &bsa, retval);
1259 
1260 out:
1261 	if (lset != NULL)
1262 		free(lset, M_TEMP);
1263 	if (kbset != NULL)
1264 		free(kbset, M_TEMP);
1265 
1266 	return error;
1267 }
1268 
1269 #endif /* __i386__ || __m68k__ */
1270 
1271 /*
1272  * We have nonexistent fsuid equal to uid.
1273  * If modification is requested, refuse.
1274  */
1275 int
1276 linux_sys_setfsuid(l, v, retval)
1277 	 struct lwp *l;
1278 	 void *v;
1279 	 register_t *retval;
1280 {
1281 	 struct linux_sys_setfsuid_args /* {
1282 		 syscallarg(uid_t) uid;
1283 	 } */ *uap = v;
1284 	 struct proc *p = l->l_proc;
1285 	 uid_t uid;
1286 
1287 	 uid = SCARG(uap, uid);
1288 	 if (p->p_cred->p_ruid != uid)
1289 		 return sys_nosys(l, v, retval);
1290 	 else
1291 		 return (0);
1292 }
1293 
1294 /* XXX XXX XXX */
1295 #ifndef alpha
1296 int
1297 linux_sys_getfsuid(l, v, retval)
1298 	struct lwp *l;
1299 	void *v;
1300 	register_t *retval;
1301 {
1302 	return sys_getuid(l, v, retval);
1303 }
1304 #endif
1305 
1306 int
1307 linux_sys_setresuid(l, v, retval)
1308 	struct lwp *l;
1309 	void *v;
1310 	register_t *retval;
1311 {
1312 	struct linux_sys_setresuid_args /* {
1313 		syscallarg(uid_t) ruid;
1314 		syscallarg(uid_t) euid;
1315 		syscallarg(uid_t) suid;
1316 	} */ *uap = v;
1317 
1318 	/*
1319 	 * Note: These checks are a little different than the NetBSD
1320 	 * setreuid(2) call performs.  This precisely follows the
1321 	 * behavior of the Linux kernel.
1322 	 */
1323 
1324 	return do_setresuid(l, SCARG(uap, ruid), SCARG(uap, euid),
1325 			    SCARG(uap, suid),
1326 			    ID_R_EQ_R | ID_R_EQ_E | ID_R_EQ_S |
1327 			    ID_E_EQ_R | ID_E_EQ_E | ID_E_EQ_S |
1328 			    ID_S_EQ_R | ID_S_EQ_E | ID_S_EQ_S );
1329 }
1330 
1331 int
1332 linux_sys_getresuid(l, v, retval)
1333 	struct lwp *l;
1334 	void *v;
1335 	register_t *retval;
1336 {
1337 	struct linux_sys_getresuid_args /* {
1338 		syscallarg(uid_t *) ruid;
1339 		syscallarg(uid_t *) euid;
1340 		syscallarg(uid_t *) suid;
1341 	} */ *uap = v;
1342 	struct proc *p = l->l_proc;
1343 	struct pcred *pc = p->p_cred;
1344 	int error;
1345 
1346 	/*
1347 	 * Linux copies these values out to userspace like so:
1348 	 *
1349 	 *	1. Copy out ruid.
1350 	 *	2. If that succeeds, copy out euid.
1351 	 *	3. If both of those succeed, copy out suid.
1352 	 */
1353 	if ((error = copyout(&pc->p_ruid, SCARG(uap, ruid),
1354 			     sizeof(uid_t))) != 0)
1355 		return (error);
1356 
1357 	if ((error = copyout(&pc->pc_ucred->cr_uid, SCARG(uap, euid),
1358 			     sizeof(uid_t))) != 0)
1359 		return (error);
1360 
1361 	return (copyout(&pc->p_svuid, SCARG(uap, suid), sizeof(uid_t)));
1362 }
1363 
1364 int
1365 linux_sys_ptrace(l, v, retval)
1366 	struct lwp *l;
1367 	void *v;
1368 	register_t *retval;
1369 {
1370 	struct linux_sys_ptrace_args /* {
1371 		i386, m68k, powerpc: T=int
1372 		alpha: T=long
1373 		syscallarg(T) request;
1374 		syscallarg(T) pid;
1375 		syscallarg(T) addr;
1376 		syscallarg(T) data;
1377 	} */ *uap = v;
1378 	const int *ptr;
1379 	int request;
1380 	int error;
1381 
1382 	ptr = linux_ptrace_request_map;
1383 	request = SCARG(uap, request);
1384 	while (*ptr != -1)
1385 		if (*ptr++ == request) {
1386 			struct sys_ptrace_args pta;
1387 
1388 			SCARG(&pta, req) = *ptr;
1389 			SCARG(&pta, pid) = SCARG(uap, pid);
1390 			SCARG(&pta, addr) = (caddr_t)SCARG(uap, addr);
1391 			SCARG(&pta, data) = SCARG(uap, data);
1392 
1393 			/*
1394 			 * Linux ptrace(PTRACE_CONT, pid, 0, 0) means actually
1395 			 * to continue where the process left off previously.
1396 			 * The same thing is achieved by addr == (caddr_t) 1
1397 			 * on NetBSD, so rewrite 'addr' appropriately.
1398 			 */
1399 			if (request == LINUX_PTRACE_CONT && SCARG(uap, addr)==0)
1400 				SCARG(&pta, addr) = (caddr_t) 1;
1401 
1402 			error = sys_ptrace(l, &pta, retval);
1403 			if (error)
1404 				return error;
1405 			switch (request) {
1406 			case LINUX_PTRACE_PEEKTEXT:
1407 			case LINUX_PTRACE_PEEKDATA:
1408 				error = copyout (retval,
1409 				    (caddr_t)SCARG(uap, data), sizeof *retval);
1410 				*retval = SCARG(uap, data);
1411 				break;
1412 			default:
1413 				break;
1414 			}
1415 			return error;
1416 		}
1417 		else
1418 			ptr++;
1419 
1420 	return LINUX_SYS_PTRACE_ARCH(l, uap, retval);
1421 }
1422 
1423 int
1424 linux_sys_reboot(struct lwp *l, void *v, register_t *retval)
1425 {
1426 	struct linux_sys_reboot_args /* {
1427 		syscallarg(int) magic1;
1428 		syscallarg(int) magic2;
1429 		syscallarg(int) cmd;
1430 		syscallarg(void *) arg;
1431 	} */ *uap = v;
1432 	struct sys_reboot_args /* {
1433 		syscallarg(int) opt;
1434 		syscallarg(char *) bootstr;
1435 	} */ sra;
1436 	struct proc *p = l->l_proc;
1437 	int error;
1438 
1439 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1440 		return(error);
1441 
1442 	if (SCARG(uap, magic1) != LINUX_REBOOT_MAGIC1)
1443 		return(EINVAL);
1444 	if (SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2 &&
1445 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2A &&
1446 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2B)
1447 		return(EINVAL);
1448 
1449 	switch (SCARG(uap, cmd)) {
1450 	case LINUX_REBOOT_CMD_RESTART:
1451 		SCARG(&sra, opt) = RB_AUTOBOOT;
1452 		break;
1453 	case LINUX_REBOOT_CMD_HALT:
1454 		SCARG(&sra, opt) = RB_HALT;
1455 		break;
1456 	case LINUX_REBOOT_CMD_POWER_OFF:
1457 		SCARG(&sra, opt) = RB_HALT|RB_POWERDOWN;
1458 		break;
1459 	case LINUX_REBOOT_CMD_RESTART2:
1460 		/* Reboot with an argument. */
1461 		SCARG(&sra, opt) = RB_AUTOBOOT|RB_STRING;
1462 		SCARG(&sra, bootstr) = SCARG(uap, arg);
1463 		break;
1464 	case LINUX_REBOOT_CMD_CAD_ON:
1465 		return(EINVAL);	/* We don't implement ctrl-alt-delete */
1466 	case LINUX_REBOOT_CMD_CAD_OFF:
1467 		return(0);
1468 	default:
1469 		return(EINVAL);
1470 	}
1471 
1472 	return(sys_reboot(l, &sra, retval));
1473 }
1474 
1475 /*
1476  * Copy of compat_12_sys_swapon().
1477  */
1478 int
1479 linux_sys_swapon(l, v, retval)
1480 	struct lwp *l;
1481 	void *v;
1482 	register_t *retval;
1483 {
1484 	struct sys_swapctl_args ua;
1485 	struct linux_sys_swapon_args /* {
1486 		syscallarg(const char *) name;
1487 	} */ *uap = v;
1488 
1489 	SCARG(&ua, cmd) = SWAP_ON;
1490 	SCARG(&ua, arg) = (void *)SCARG(uap, name);
1491 	SCARG(&ua, misc) = 0;	/* priority */
1492 	return (sys_swapctl(l, &ua, retval));
1493 }
1494 
1495 /*
1496  * Stop swapping to the file or block device specified by path.
1497  */
1498 int
1499 linux_sys_swapoff(l, v, retval)
1500 	struct lwp *l;
1501 	void *v;
1502 	register_t *retval;
1503 {
1504 	struct sys_swapctl_args ua;
1505 	struct linux_sys_swapoff_args /* {
1506 		syscallarg(const char *) path;
1507 	} */ *uap = v;
1508 
1509 	SCARG(&ua, cmd) = SWAP_OFF;
1510 	SCARG(&ua, arg) = (void *)SCARG(uap, path);
1511 	return (sys_swapctl(l, &ua, retval));
1512 }
1513 
1514 /*
1515  * Copy of compat_09_sys_setdomainname()
1516  */
1517 /* ARGSUSED */
1518 int
1519 linux_sys_setdomainname(l, v, retval)
1520 	struct lwp *l;
1521 	void *v;
1522 	register_t *retval;
1523 {
1524 	struct linux_sys_setdomainname_args /* {
1525 		syscallarg(char *) domainname;
1526 		syscallarg(int) len;
1527 	} */ *uap = v;
1528 	int name[2];
1529 
1530 	name[0] = CTL_KERN;
1531 	name[1] = KERN_DOMAINNAME;
1532 	return (old_sysctl(&name[0], 2, 0, 0, SCARG(uap, domainname),
1533 			    SCARG(uap, len), l));
1534 }
1535 
1536 /*
1537  * sysinfo()
1538  */
1539 /* ARGSUSED */
1540 int
1541 linux_sys_sysinfo(l, v, retval)
1542 	struct lwp *l;
1543 	void *v;
1544 	register_t *retval;
1545 {
1546 	struct linux_sys_sysinfo_args /* {
1547 		syscallarg(struct linux_sysinfo *) arg;
1548 	} */ *uap = v;
1549 	struct linux_sysinfo si;
1550 	struct loadavg *la;
1551 
1552 	si.uptime = time.tv_sec - boottime.tv_sec;
1553 	la = &averunnable;
1554 	si.loads[0] = la->ldavg[0] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1555 	si.loads[1] = la->ldavg[1] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1556 	si.loads[2] = la->ldavg[2] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1557 	si.totalram = ctob(physmem);
1558 	si.freeram = uvmexp.free * uvmexp.pagesize;
1559 	si.sharedram = 0;	/* XXX */
1560 	si.bufferram = uvmexp.filepages * uvmexp.pagesize;
1561 	si.totalswap = uvmexp.swpages * uvmexp.pagesize;
1562 	si.freeswap = (uvmexp.swpages - uvmexp.swpginuse) * uvmexp.pagesize;
1563 	si.procs = nprocs;
1564 
1565 	/* The following are only present in newer Linux kernels. */
1566 	si.totalbig = 0;
1567 	si.freebig = 0;
1568 	si.mem_unit = 1;
1569 
1570 	return (copyout(&si, SCARG(uap, arg), sizeof si));
1571 }
1572 
1573 #define bsd_to_linux_rlimit1(l, b, f) \
1574     (l)->f = ((b)->f == RLIM_INFINITY || \
1575 	     ((b)->f & 0xffffffff00000000ULL) != 0) ? \
1576     LINUX_RLIM_INFINITY : (int32_t)(b)->f
1577 #define bsd_to_linux_rlimit(l, b) \
1578     bsd_to_linux_rlimit1(l, b, rlim_cur); \
1579     bsd_to_linux_rlimit1(l, b, rlim_max)
1580 
1581 #define linux_to_bsd_rlimit1(b, l, f) \
1582     (b)->f = (l)->f == LINUX_RLIM_INFINITY ? RLIM_INFINITY : (l)->f
1583 #define linux_to_bsd_rlimit(b, l) \
1584     linux_to_bsd_rlimit1(b, l, rlim_cur); \
1585     linux_to_bsd_rlimit1(b, l, rlim_max)
1586 
1587 static int
1588 linux_to_bsd_limit(lim)
1589 	int lim;
1590 {
1591 	switch (lim) {
1592 	case LINUX_RLIMIT_CPU:
1593 		return RLIMIT_CPU;
1594 	case LINUX_RLIMIT_FSIZE:
1595 		return RLIMIT_FSIZE;
1596 	case LINUX_RLIMIT_DATA:
1597 		return RLIMIT_DATA;
1598 	case LINUX_RLIMIT_STACK:
1599 		return RLIMIT_STACK;
1600 	case LINUX_RLIMIT_CORE:
1601 		return RLIMIT_CORE;
1602 	case LINUX_RLIMIT_RSS:
1603 		return RLIMIT_RSS;
1604 	case LINUX_RLIMIT_NPROC:
1605 		return RLIMIT_NPROC;
1606 	case LINUX_RLIMIT_NOFILE:
1607 		return RLIMIT_NOFILE;
1608 	case LINUX_RLIMIT_MEMLOCK:
1609 		return RLIMIT_MEMLOCK;
1610 	case LINUX_RLIMIT_AS:
1611 	case LINUX_RLIMIT_LOCKS:
1612 		return -EOPNOTSUPP;
1613 	default:
1614 		return -EINVAL;
1615 	}
1616 }
1617 
1618 
1619 int
1620 linux_sys_getrlimit(l, v, retval)
1621 	struct lwp *l;
1622 	void *v;
1623 	register_t *retval;
1624 {
1625 	struct linux_sys_getrlimit_args /* {
1626 		syscallarg(int) which;
1627 		syscallarg(struct orlimit *) rlp;
1628 	} */ *uap = v;
1629 	struct proc *p = l->l_proc;
1630 	caddr_t sg = stackgap_init(p, 0);
1631 	struct sys_getrlimit_args ap;
1632 	struct rlimit rl;
1633 	struct orlimit orl;
1634 	int error;
1635 
1636 	SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which));
1637 	if ((error = SCARG(&ap, which)) < 0)
1638 		return -error;
1639 	SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl);
1640 	if ((error = sys_getrlimit(l, &ap, retval)) != 0)
1641 		return error;
1642 	if ((error = copyin(SCARG(&ap, rlp), &rl, sizeof(rl))) != 0)
1643 		return error;
1644 	bsd_to_linux_rlimit(&orl, &rl);
1645 	return copyout(&orl, SCARG(uap, rlp), sizeof(orl));
1646 }
1647 
1648 int
1649 linux_sys_setrlimit(l, v, retval)
1650 	struct lwp *l;
1651 	void *v;
1652 	register_t *retval;
1653 {
1654 	struct linux_sys_setrlimit_args /* {
1655 		syscallarg(int) which;
1656 		syscallarg(struct orlimit *) rlp;
1657 	} */ *uap = v;
1658 	struct proc *p = l->l_proc;
1659 	caddr_t sg = stackgap_init(p, 0);
1660 	struct sys_setrlimit_args ap;
1661 	struct rlimit rl;
1662 	struct orlimit orl;
1663 	int error;
1664 
1665 	SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which));
1666 	SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl);
1667 	if ((error = SCARG(&ap, which)) < 0)
1668 		return -error;
1669 	if ((error = copyin(SCARG(uap, rlp), &orl, sizeof(orl))) != 0)
1670 		return error;
1671 	linux_to_bsd_rlimit(&rl, &orl);
1672 	/* XXX: alpha complains about this */
1673 	if ((error = copyout(&rl, (void *)SCARG(&ap, rlp), sizeof(rl))) != 0)
1674 		return error;
1675 	return sys_setrlimit(l, &ap, retval);
1676 }
1677 
1678 #ifndef __mips__
1679 /* XXX: this doesn't look 100% common, at least mips doesn't have it */
1680 int
1681 linux_sys_ugetrlimit(l, v, retval)
1682 	struct lwp *l;
1683 	void *v;
1684 	register_t *retval;
1685 {
1686 	return linux_sys_getrlimit(l, v, retval);
1687 }
1688 #endif
1689 
1690 /*
1691  * This gets called for unsupported syscalls. The difference to sys_nosys()
1692  * is that process does not get SIGSYS, the call just returns with ENOSYS.
1693  * This is the way Linux does it and glibc depends on this behaviour.
1694  */
1695 int
1696 linux_sys_nosys(l, v, retval)
1697 	struct lwp *l;
1698 	void *v;
1699 	register_t *retval;
1700 {
1701 	return (ENOSYS);
1702 }
1703