xref: /netbsd-src/sys/compat/linux/common/linux_misc.c (revision 93086a3d8d0057acb6d878df423458ce6cbc1fc5)
1 /*	$NetBSD: linux_misc.c,v 1.133 2004/10/07 19:30:28 erh 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 1.133 2004/10/07 19:30:28 erh 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 };
159 const int linux_fstypes_cnt = sizeof(linux_fstypes) / sizeof(linux_fstypes[0]);
160 
161 #ifdef DEBUG_LINUX
162 #define DPRINTF(a)	uprintf a
163 #else
164 #define DPRINTF(a)
165 #endif
166 
167 /* Local linux_misc.c functions: */
168 static void bsd_to_linux_statfs __P((const struct statvfs *,
169     struct linux_statfs *));
170 static int linux_to_bsd_limit __P((int));
171 static void linux_to_bsd_mmap_args __P((struct sys_mmap_args *,
172     const struct linux_sys_mmap_args *));
173 static int linux_mmap __P((struct lwp *, struct linux_sys_mmap_args *,
174     register_t *, off_t));
175 
176 
177 /*
178  * The information on a terminated (or stopped) process needs
179  * to be converted in order for Linux binaries to get a valid signal
180  * number out of it.
181  */
182 void
183 bsd_to_linux_wstat(st)
184 	int *st;
185 {
186 
187 	int sig;
188 
189 	if (WIFSIGNALED(*st)) {
190 		sig = WTERMSIG(*st);
191 		if (sig >= 0 && sig < NSIG)
192 			*st= (*st& ~0177) | native_to_linux_signo[sig];
193 	} else if (WIFSTOPPED(*st)) {
194 		sig = WSTOPSIG(*st);
195 		if (sig >= 0 && sig < NSIG)
196 			*st = (*st & ~0xff00) |
197 			    (native_to_linux_signo[sig] << 8);
198 	}
199 }
200 
201 /*
202  * wait4(2).  Passed on to the NetBSD call, surrounded by code to
203  * reserve some space for a NetBSD-style wait status, and converting
204  * it to what Linux wants.
205  */
206 int
207 linux_sys_wait4(l, v, retval)
208 	struct lwp *l;
209 	void *v;
210 	register_t *retval;
211 {
212 	struct linux_sys_wait4_args /* {
213 		syscallarg(int) pid;
214 		syscallarg(int *) status;
215 		syscallarg(int) options;
216 		syscallarg(struct rusage *) rusage;
217 	} */ *uap = v;
218 	struct proc *p = l->l_proc;
219 	struct sys_wait4_args w4a;
220 	int error, *status, tstat, options, linux_options;
221 	caddr_t sg;
222 
223 	if (SCARG(uap, status) != NULL) {
224 		sg = stackgap_init(p, 0);
225 		status = (int *) stackgap_alloc(p, &sg, sizeof *status);
226 	} else
227 		status = NULL;
228 
229 	linux_options = SCARG(uap, options);
230 	options = 0;
231 	if (linux_options & ~(LINUX_WAIT4_KNOWNFLAGS))
232 		return (EINVAL);
233 
234 	if (linux_options & LINUX_WAIT4_WNOHANG)
235 		options |= WNOHANG;
236 	if (linux_options & LINUX_WAIT4_WUNTRACED)
237 		options |= WUNTRACED;
238 	if (linux_options & LINUX_WAIT4_WALL)
239 		options |= WALLSIG;
240 	if (linux_options & LINUX_WAIT4_WCLONE)
241 		options |= WALTSIG;
242 #ifdef DIAGNOSTIC
243 	if (linux_options & LINUX_WAIT4_WNOTHREAD)
244 		printf("WARNING: %s: linux process %d.%d called "
245 		       "waitpid with __WNOTHREAD set!",
246 		       __FILE__, p->p_pid, l->l_lid);
247 
248 #endif
249 
250 	SCARG(&w4a, pid) = SCARG(uap, pid);
251 	SCARG(&w4a, status) = status;
252 	SCARG(&w4a, options) = options;
253 	SCARG(&w4a, rusage) = SCARG(uap, rusage);
254 
255 	if ((error = sys_wait4(l, &w4a, retval)))
256 		return error;
257 
258 	sigdelset(&p->p_sigctx.ps_siglist, SIGCHLD);
259 
260 	if (status != NULL) {
261 		if ((error = copyin(status, &tstat, sizeof tstat)))
262 			return error;
263 
264 		bsd_to_linux_wstat(&tstat);
265 		return copyout(&tstat, SCARG(uap, status), sizeof tstat);
266 	}
267 
268 	return 0;
269 }
270 
271 /*
272  * Linux brk(2). The check if the new address is >= the old one is
273  * done in the kernel in Linux. NetBSD does it in the library.
274  */
275 int
276 linux_sys_brk(l, v, retval)
277 	struct lwp *l;
278 	void *v;
279 	register_t *retval;
280 {
281 	struct linux_sys_brk_args /* {
282 		syscallarg(char *) nsize;
283 	} */ *uap = v;
284 	struct proc *p = l->l_proc;
285 	char *nbrk = SCARG(uap, nsize);
286 	struct sys_obreak_args oba;
287 	struct vmspace *vm = p->p_vmspace;
288 	struct linux_emuldata *ed = (struct linux_emuldata*)p->p_emuldata;
289 
290 	SCARG(&oba, nsize) = nbrk;
291 
292 	if ((caddr_t) nbrk > vm->vm_daddr && sys_obreak(l, &oba, retval) == 0)
293 		ed->s->p_break = (char*)nbrk;
294 	else
295 		nbrk = ed->s->p_break;
296 
297 	retval[0] = (register_t)nbrk;
298 
299 	return 0;
300 }
301 
302 /*
303  * Convert NetBSD statvfs structure to Linux statfs structure.
304  * Linux doesn't have f_flag, and we can't set f_frsize due
305  * to glibc statvfs() bug (see below).
306  */
307 static void
308 bsd_to_linux_statfs(bsp, lsp)
309 	const struct statvfs *bsp;
310 	struct linux_statfs *lsp;
311 {
312 	int i;
313 
314 	for (i = 0; i < linux_fstypes_cnt; i++) {
315 		if (strcmp(bsp->f_fstypename, linux_fstypes[i].bsd) == 0) {
316 			lsp->l_ftype = linux_fstypes[i].linux;
317 			break;
318 		}
319 	}
320 
321 	if (i == linux_fstypes_cnt) {
322 		DPRINTF(("unhandled fstype in linux emulation: %s\n",
323 		    bsp->f_fstypename));
324 		lsp->l_ftype = LINUX_DEFAULT_SUPER_MAGIC;
325 	}
326 
327 	/*
328 	 * The sizes are expressed in number of blocks. The block
329 	 * size used for the size is f_frsize for POSIX-compliant
330 	 * statvfs. Linux statfs uses f_bsize as the block size
331 	 * (f_frsize used to not be available in Linux struct statfs).
332 	 * However, glibc 2.3.3 statvfs() wrapper fails to adjust the block
333 	 * counts for different f_frsize if f_frsize is provided by the kernel.
334 	 * POSIX conforming apps thus get wrong size if f_frsize
335 	 * is different to f_bsize. Thus, we just pretend we don't
336 	 * support f_frsize.
337 	 */
338 
339 	lsp->l_fbsize = bsp->f_frsize;
340 	lsp->l_ffrsize = 0;			/* compat */
341 	lsp->l_fblocks = bsp->f_blocks;
342 	lsp->l_fbfree = bsp->f_bfree;
343 	lsp->l_fbavail = bsp->f_bavail;
344 	lsp->l_ffiles = bsp->f_files;
345 	lsp->l_fffree = bsp->f_ffree;
346 	/* Linux sets the fsid to 0..., we don't */
347 	lsp->l_ffsid.val[0] = bsp->f_fsidx.__fsid_val[0];
348 	lsp->l_ffsid.val[1] = bsp->f_fsidx.__fsid_val[1];
349 	lsp->l_fnamelen = bsp->f_namemax;
350 	(void)memset(lsp->l_fspare, 0, sizeof(lsp->l_fspare));
351 }
352 
353 /*
354  * Implement the fs stat functions. Straightforward.
355  */
356 int
357 linux_sys_statfs(l, v, retval)
358 	struct lwp *l;
359 	void *v;
360 	register_t *retval;
361 {
362 	struct linux_sys_statfs_args /* {
363 		syscallarg(const char *) path;
364 		syscallarg(struct linux_statfs *) sp;
365 	} */ *uap = v;
366 	struct proc *p = l->l_proc;
367 	struct statvfs btmp, *bsp;
368 	struct linux_statfs ltmp;
369 	struct sys_statvfs1_args bsa;
370 	caddr_t sg;
371 	int error;
372 
373 	sg = stackgap_init(p, 0);
374 	bsp = (struct statvfs *) stackgap_alloc(p, &sg, sizeof (struct statvfs));
375 
376 	CHECK_ALT_EXIST(p, &sg, SCARG(uap, path));
377 
378 	SCARG(&bsa, path) = SCARG(uap, path);
379 	SCARG(&bsa, buf) = bsp;
380 	SCARG(&bsa, flags) = ST_WAIT;
381 
382 	if ((error = sys_statvfs1(l, &bsa, retval)))
383 		return error;
384 
385 	if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp)))
386 		return error;
387 
388 	bsd_to_linux_statfs(&btmp, &ltmp);
389 
390 	return copyout((caddr_t) &ltmp, (caddr_t) SCARG(uap, sp), sizeof ltmp);
391 }
392 
393 int
394 linux_sys_fstatfs(l, v, retval)
395 	struct lwp *l;
396 	void *v;
397 	register_t *retval;
398 {
399 	struct linux_sys_fstatfs_args /* {
400 		syscallarg(int) fd;
401 		syscallarg(struct linux_statfs *) sp;
402 	} */ *uap = v;
403 	struct proc *p = l->l_proc;
404 	struct statvfs btmp, *bsp;
405 	struct linux_statfs ltmp;
406 	struct sys_fstatvfs1_args bsa;
407 	caddr_t sg;
408 	int error;
409 
410 	sg = stackgap_init(p, 0);
411 	bsp = (struct statvfs *) stackgap_alloc(p, &sg, sizeof (struct statvfs));
412 
413 	SCARG(&bsa, fd) = SCARG(uap, fd);
414 	SCARG(&bsa, buf) = bsp;
415 	SCARG(&bsa, flags) = ST_WAIT;
416 
417 	if ((error = sys_fstatvfs1(l, &bsa, retval)))
418 		return error;
419 
420 	if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp)))
421 		return error;
422 
423 	bsd_to_linux_statfs(&btmp, &ltmp);
424 
425 	return copyout((caddr_t) &ltmp, (caddr_t) SCARG(uap, sp), sizeof ltmp);
426 }
427 
428 /*
429  * uname(). Just copy the info from the various strings stored in the
430  * kernel, and put it in the Linux utsname structure. That structure
431  * is almost the same as the NetBSD one, only it has fields 65 characters
432  * long, and an extra domainname field.
433  */
434 int
435 linux_sys_uname(l, v, retval)
436 	struct lwp *l;
437 	void *v;
438 	register_t *retval;
439 {
440 	struct linux_sys_uname_args /* {
441 		syscallarg(struct linux_utsname *) up;
442 	} */ *uap = v;
443 	struct linux_utsname luts;
444 
445 	strncpy(luts.l_sysname, linux_sysname, sizeof(luts.l_sysname));
446 	strncpy(luts.l_nodename, hostname, sizeof(luts.l_nodename));
447 	strncpy(luts.l_release, linux_release, sizeof(luts.l_release));
448 	strncpy(luts.l_version, linux_version, sizeof(luts.l_version));
449 	strncpy(luts.l_machine, machine, sizeof(luts.l_machine));
450 	strncpy(luts.l_domainname, domainname, sizeof(luts.l_domainname));
451 
452 	return copyout(&luts, SCARG(uap, up), sizeof(luts));
453 }
454 
455 /* Used directly on: alpha, mips, ppc, sparc, sparc64 */
456 /* Used indirectly on: arm, i386, m68k */
457 
458 /*
459  * New type Linux mmap call.
460  * Only called directly on machines with >= 6 free regs.
461  */
462 int
463 linux_sys_mmap(l, v, retval)
464 	struct lwp *l;
465 	void *v;
466 	register_t *retval;
467 {
468 	struct linux_sys_mmap_args /* {
469 		syscallarg(unsigned long) addr;
470 		syscallarg(size_t) len;
471 		syscallarg(int) prot;
472 		syscallarg(int) flags;
473 		syscallarg(int) fd;
474 		syscallarg(linux_off_t) offset;
475 	} */ *uap = v;
476 
477 	if (SCARG(uap, offset) & PAGE_MASK)
478 		return EINVAL;
479 
480 	return linux_mmap(l, uap, retval, SCARG(uap, offset));
481 }
482 
483 /*
484  * Guts of most architectures' mmap64() implementations.  This shares
485  * its list of arguments with linux_sys_mmap().
486  *
487  * The difference in linux_sys_mmap2() is that "offset" is actually
488  * (offset / pagesize), not an absolute byte count.  This translation
489  * to pagesize offsets is done inside glibc between the mmap64() call
490  * point, and the actual syscall.
491  */
492 int
493 linux_sys_mmap2(l, v, retval)
494 	struct lwp *l;
495 	void *v;
496 	register_t *retval;
497 {
498 	struct linux_sys_mmap2_args /* {
499 		syscallarg(unsigned long) addr;
500 		syscallarg(size_t) len;
501 		syscallarg(int) prot;
502 		syscallarg(int) flags;
503 		syscallarg(int) fd;
504 		syscallarg(linux_off_t) offset;
505 	} */ *uap = v;
506 
507 	return linux_mmap(l, uap, retval,
508 	    ((off_t)SCARG(uap, offset)) << PAGE_SHIFT);
509 }
510 
511 /*
512  * Massage arguments and call system mmap(2).
513  */
514 static int
515 linux_mmap(l, uap, retval, offset)
516 	struct lwp *l;
517 	struct linux_sys_mmap_args *uap;
518 	register_t *retval;
519 	off_t offset;
520 {
521 	struct sys_mmap_args cma;
522 	int error;
523 	size_t mmoff=0;
524 
525 	if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN) {
526 		/*
527 		 * Request for stack-like memory segment. On linux, this
528 		 * works by mmap()ping (small) segment, which is automatically
529 		 * extended when page fault happens below the currently
530 		 * allocated area. We emulate this by allocating (typically
531 		 * bigger) segment sized at current stack size limit, and
532 		 * offsetting the requested and returned address accordingly.
533 		 * Since physical pages are only allocated on-demand, this
534 		 * is effectively identical.
535 		 */
536 		rlim_t ssl = l->l_proc->p_rlimit[RLIMIT_STACK].rlim_cur;
537 
538 		if (SCARG(uap, len) < ssl) {
539 			/* Compute the address offset */
540 			mmoff = round_page(ssl) - SCARG(uap, len);
541 
542 			if (SCARG(uap, addr))
543 				SCARG(uap, addr) -= mmoff;
544 
545 			SCARG(uap, len) = (size_t) ssl;
546 		}
547 	}
548 
549 	linux_to_bsd_mmap_args(&cma, uap);
550 	SCARG(&cma, pos) = offset;
551 
552 	error = sys_mmap(l, &cma, retval);
553 	if (error)
554 		return (error);
555 
556 	/* Shift the returned address for stack-like segment if necessary */
557 	if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN && mmoff)
558 		retval[0] += mmoff;
559 
560 	return (0);
561 }
562 
563 static void
564 linux_to_bsd_mmap_args(cma, uap)
565 	struct sys_mmap_args *cma;
566 	const struct linux_sys_mmap_args *uap;
567 {
568 	int flags = MAP_TRYFIXED, fl = SCARG(uap, flags);
569 
570 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_SHARED, MAP_SHARED);
571 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_PRIVATE, MAP_PRIVATE);
572 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_FIXED, MAP_FIXED);
573 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_ANON, MAP_ANON);
574 	/* XXX XAX ERH: Any other flags here?  There are more defined... */
575 
576 	SCARG(cma, addr) = (void *)SCARG(uap, addr);
577 	SCARG(cma, len) = SCARG(uap, len);
578 	SCARG(cma, prot) = SCARG(uap, prot);
579 	if (SCARG(cma, prot) & VM_PROT_WRITE) /* XXX */
580 		SCARG(cma, prot) |= VM_PROT_READ;
581 	SCARG(cma, flags) = flags;
582 	SCARG(cma, fd) = flags & MAP_ANON ? -1 : SCARG(uap, fd);
583 	SCARG(cma, pad) = 0;
584 }
585 
586 int
587 linux_sys_mremap(l, v, retval)
588 	struct lwp *l;
589 	void *v;
590 	register_t *retval;
591 {
592 	struct linux_sys_mremap_args /* {
593 		syscallarg(void *) old_address;
594 		syscallarg(size_t) old_size;
595 		syscallarg(size_t) new_size;
596 		syscallarg(u_long) flags;
597 	} */ *uap = v;
598 	struct sys_munmap_args mua;
599 	size_t old_size, new_size;
600 	int error;
601 
602 	old_size = round_page(SCARG(uap, old_size));
603 	new_size = round_page(SCARG(uap, new_size));
604 
605 	/*
606 	 * Growing mapped region.
607 	 */
608 	if (new_size > old_size) {
609 		/*
610 		 * XXX Implement me.  What we probably want to do is
611 		 * XXX dig out the guts of the old mapping, mmap that
612 		 * XXX object again with the new size, then munmap
613 		 * XXX the old mapping.
614 		 */
615 		*retval = 0;
616 		return (ENOMEM);
617 	}
618 
619 	/*
620 	 * Shrinking mapped region.
621 	 */
622 	if (new_size < old_size) {
623 		SCARG(&mua, addr) = (caddr_t)SCARG(uap, old_address) +
624 		    new_size;
625 		SCARG(&mua, len) = old_size - new_size;
626 		error = sys_munmap(l, &mua, retval);
627 		*retval = error ? 0 : (register_t)SCARG(uap, old_address);
628 		return (error);
629 	}
630 
631 	/*
632 	 * No change.
633 	 */
634 	*retval = (register_t)SCARG(uap, old_address);
635 	return (0);
636 }
637 
638 int
639 linux_sys_msync(l, v, retval)
640 	struct lwp *l;
641 	void *v;
642 	register_t *retval;
643 {
644 	struct linux_sys_msync_args /* {
645 		syscallarg(caddr_t) addr;
646 		syscallarg(int) len;
647 		syscallarg(int) fl;
648 	} */ *uap = v;
649 
650 	struct sys___msync13_args bma;
651 
652 	/* flags are ignored */
653 	SCARG(&bma, addr) = SCARG(uap, addr);
654 	SCARG(&bma, len) = SCARG(uap, len);
655 	SCARG(&bma, flags) = SCARG(uap, fl);
656 
657 	return sys___msync13(l, &bma, retval);
658 }
659 
660 int
661 linux_sys_mprotect(l, v, retval)
662 	struct lwp *l;
663 	void *v;
664 	register_t *retval;
665 {
666 	struct linux_sys_mprotect_args /* {
667 		syscallarg(const void *) start;
668 		syscallarg(unsigned long) len;
669 		syscallarg(int) prot;
670 	} */ *uap = v;
671 	unsigned long end, start = (unsigned long)SCARG(uap, start), len;
672 	int prot = SCARG(uap, prot);
673 	struct vm_map_entry *entry;
674 	struct vm_map *map = &l->l_proc->p_vmspace->vm_map;
675 
676 	if (start & PAGE_MASK)
677 		return EINVAL;
678 
679 	len = round_page(SCARG(uap, len));
680 	end = start + len;
681 
682 	if (end < start)
683 		return EINVAL;
684 	else if (end == start)
685 		return 0;
686 
687 	if (SCARG(uap, prot) & ~(PROT_READ | PROT_WRITE | PROT_EXEC))
688 		return EINVAL;
689 
690 	vm_map_lock(map);
691 #ifdef notdef
692 	VM_MAP_RANGE_CHECK(map, start, end);
693 #endif
694 	if (!uvm_map_lookup_entry(map, start, &entry) || entry->start > start) {
695 		vm_map_unlock(map);
696 		return ENOMEM;
697 	}
698 	vm_map_unlock(map);
699 	return uvm_map_protect(map, start, end, prot, FALSE);
700 }
701 
702 /*
703  * This code is partly stolen from src/lib/libc/compat-43/times.c
704  */
705 
706 #define	CONVTCK(r)	(r.tv_sec * hz + r.tv_usec / (1000000 / hz))
707 
708 int
709 linux_sys_times(l, v, retval)
710 	struct lwp *l;
711 	void *v;
712 	register_t *retval;
713 {
714 	struct linux_sys_times_args /* {
715 		syscallarg(struct times *) tms;
716 	} */ *uap = v;
717 	struct proc *p = l->l_proc;
718 	struct timeval t;
719 	int error, s;
720 
721 	if (SCARG(uap, tms)) {
722 		struct linux_tms ltms;
723 		struct rusage ru;
724 
725 		calcru(p, &ru.ru_utime, &ru.ru_stime, NULL);
726 		ltms.ltms_utime = CONVTCK(ru.ru_utime);
727 		ltms.ltms_stime = CONVTCK(ru.ru_stime);
728 
729 		ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime);
730 		ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime);
731 
732 		if ((error = copyout(&ltms, SCARG(uap, tms), sizeof ltms)))
733 			return error;
734 	}
735 
736 	s = splclock();
737 	timersub(&time, &boottime, &t);
738 	splx(s);
739 
740 	retval[0] = ((linux_clock_t)(CONVTCK(t)));
741 	return 0;
742 }
743 
744 #undef CONVTCK
745 
746 /*
747  * Linux 'readdir' call. This code is mostly taken from the
748  * SunOS getdents call (see compat/sunos/sunos_misc.c), though
749  * an attempt has been made to keep it a little cleaner (failing
750  * miserably, because of the cruft needed if count 1 is passed).
751  *
752  * The d_off field should contain the offset of the next valid entry,
753  * but in Linux it has the offset of the entry itself. We emulate
754  * that bug here.
755  *
756  * Read in BSD-style entries, convert them, and copy them out.
757  *
758  * Note that this doesn't handle union-mounted filesystems.
759  */
760 int
761 linux_sys_getdents(l, v, retval)
762 	struct lwp *l;
763 	void *v;
764 	register_t *retval;
765 {
766 	struct linux_sys_getdents_args /* {
767 		syscallarg(int) fd;
768 		syscallarg(struct linux_dirent *) dent;
769 		syscallarg(unsigned int) count;
770 	} */ *uap = v;
771 	struct proc *p = l->l_proc;
772 	struct dirent *bdp;
773 	struct vnode *vp;
774 	caddr_t	inp, buf;		/* BSD-format */
775 	int len, reclen;		/* BSD-format */
776 	caddr_t outp;			/* Linux-format */
777 	int resid, linux_reclen = 0;	/* Linux-format */
778 	struct file *fp;
779 	struct uio auio;
780 	struct iovec aiov;
781 	struct linux_dirent idb;
782 	off_t off;		/* true file offset */
783 	int buflen, error, eofflag, nbytes, oldcall;
784 	struct vattr va;
785 	off_t *cookiebuf = NULL, *cookie;
786 	int ncookies;
787 
788 	/* getvnode() will use the descriptor for us */
789 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
790 		return (error);
791 
792 	if ((fp->f_flag & FREAD) == 0) {
793 		error = EBADF;
794 		goto out1;
795 	}
796 
797 	vp = (struct vnode *)fp->f_data;
798 	if (vp->v_type != VDIR) {
799 		error = EINVAL;
800 		goto out1;
801 	}
802 
803 	if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)))
804 		goto out1;
805 
806 	nbytes = SCARG(uap, count);
807 	if (nbytes == 1) {	/* emulating old, broken behaviour */
808 		nbytes = sizeof (idb);
809 		buflen = max(va.va_blocksize, nbytes);
810 		oldcall = 1;
811 	} else {
812 		buflen = min(MAXBSIZE, nbytes);
813 		if (buflen < va.va_blocksize)
814 			buflen = va.va_blocksize;
815 		oldcall = 0;
816 	}
817 	buf = malloc(buflen, M_TEMP, M_WAITOK);
818 
819 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
820 	off = fp->f_offset;
821 again:
822 	aiov.iov_base = buf;
823 	aiov.iov_len = buflen;
824 	auio.uio_iov = &aiov;
825 	auio.uio_iovcnt = 1;
826 	auio.uio_rw = UIO_READ;
827 	auio.uio_segflg = UIO_SYSSPACE;
828 	auio.uio_procp = NULL;
829 	auio.uio_resid = buflen;
830 	auio.uio_offset = off;
831 	/*
832          * First we read into the malloc'ed buffer, then
833          * we massage it into user space, one record at a time.
834          */
835 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf,
836 	    &ncookies);
837 	if (error)
838 		goto out;
839 
840 	inp = buf;
841 	outp = (caddr_t)SCARG(uap, dent);
842 	resid = nbytes;
843 	if ((len = buflen - auio.uio_resid) == 0)
844 		goto eof;
845 
846 	for (cookie = cookiebuf; len > 0; len -= reclen) {
847 		bdp = (struct dirent *)inp;
848 		reclen = bdp->d_reclen;
849 		if (reclen & 3)
850 			panic("linux_readdir");
851 		if (bdp->d_fileno == 0) {
852 			inp += reclen;	/* it is a hole; squish it out */
853 			off = *cookie++;
854 			continue;
855 		}
856 		linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen);
857 		if (reclen > len || resid < linux_reclen) {
858 			/* entry too big for buffer, so just stop */
859 			outp++;
860 			break;
861 		}
862 		/*
863 		 * Massage in place to make a Linux-shaped dirent (otherwise
864 		 * we have to worry about touching user memory outside of
865 		 * the copyout() call).
866 		 */
867 		idb.d_ino = bdp->d_fileno;
868 		/*
869 		 * The old readdir() call misuses the offset and reclen fields.
870 		 */
871 		if (oldcall) {
872 			idb.d_off = (linux_off_t)linux_reclen;
873 			idb.d_reclen = (u_short)bdp->d_namlen;
874 		} else {
875 			if (sizeof (idb.d_off) <= 4 && (off >> 32) != 0) {
876 				compat_offseterr(vp, "linux_getdents");
877 				error = EINVAL;
878 				goto out;
879 			}
880 			idb.d_off = (linux_off_t)off;
881 			idb.d_reclen = (u_short)linux_reclen;
882 		}
883 		strcpy(idb.d_name, bdp->d_name);
884 		if ((error = copyout((caddr_t)&idb, outp, linux_reclen)))
885 			goto out;
886 		/* advance past this real entry */
887 		inp += reclen;
888 		off = *cookie++;	/* each entry points to itself */
889 		/* advance output past Linux-shaped entry */
890 		outp += linux_reclen;
891 		resid -= linux_reclen;
892 		if (oldcall)
893 			break;
894 	}
895 
896 	/* if we squished out the whole block, try again */
897 	if (outp == (caddr_t)SCARG(uap, dent))
898 		goto again;
899 	fp->f_offset = off;	/* update the vnode offset */
900 
901 	if (oldcall)
902 		nbytes = resid + linux_reclen;
903 
904 eof:
905 	*retval = nbytes - resid;
906 out:
907 	VOP_UNLOCK(vp, 0);
908 	if (cookiebuf)
909 		free(cookiebuf, M_TEMP);
910 	free(buf, M_TEMP);
911 out1:
912 	FILE_UNUSE(fp, p);
913 	return error;
914 }
915 
916 /*
917  * Even when just using registers to pass arguments to syscalls you can
918  * have 5 of them on the i386. So this newer version of select() does
919  * this.
920  */
921 int
922 linux_sys_select(l, v, retval)
923 	struct lwp *l;
924 	void *v;
925 	register_t *retval;
926 {
927 	struct linux_sys_select_args /* {
928 		syscallarg(int) nfds;
929 		syscallarg(fd_set *) readfds;
930 		syscallarg(fd_set *) writefds;
931 		syscallarg(fd_set *) exceptfds;
932 		syscallarg(struct timeval *) timeout;
933 	} */ *uap = v;
934 
935 	return linux_select1(l, retval, SCARG(uap, nfds), SCARG(uap, readfds),
936 	    SCARG(uap, writefds), SCARG(uap, exceptfds), SCARG(uap, timeout));
937 }
938 
939 /*
940  * Common code for the old and new versions of select(). A couple of
941  * things are important:
942  * 1) return the amount of time left in the 'timeout' parameter
943  * 2) select never returns ERESTART on Linux, always return EINTR
944  */
945 int
946 linux_select1(l, retval, nfds, readfds, writefds, exceptfds, timeout)
947 	struct lwp *l;
948 	register_t *retval;
949 	int nfds;
950 	fd_set *readfds, *writefds, *exceptfds;
951 	struct timeval *timeout;
952 {
953 	struct sys_select_args bsa;
954 	struct proc *p = l->l_proc;
955 	struct timeval tv0, tv1, utv, *tvp;
956 	caddr_t sg;
957 	int error;
958 
959 	SCARG(&bsa, nd) = nfds;
960 	SCARG(&bsa, in) = readfds;
961 	SCARG(&bsa, ou) = writefds;
962 	SCARG(&bsa, ex) = exceptfds;
963 	SCARG(&bsa, tv) = timeout;
964 
965 	/*
966 	 * Store current time for computation of the amount of
967 	 * time left.
968 	 */
969 	if (timeout) {
970 		if ((error = copyin(timeout, &utv, sizeof(utv))))
971 			return error;
972 		if (itimerfix(&utv)) {
973 			/*
974 			 * The timeval was invalid.  Convert it to something
975 			 * valid that will act as it does under Linux.
976 			 */
977 			sg = stackgap_init(p, 0);
978 			tvp = stackgap_alloc(p, &sg, sizeof(utv));
979 			utv.tv_sec += utv.tv_usec / 1000000;
980 			utv.tv_usec %= 1000000;
981 			if (utv.tv_usec < 0) {
982 				utv.tv_sec -= 1;
983 				utv.tv_usec += 1000000;
984 			}
985 			if (utv.tv_sec < 0)
986 				timerclear(&utv);
987 			if ((error = copyout(&utv, tvp, sizeof(utv))))
988 				return error;
989 			SCARG(&bsa, tv) = tvp;
990 		}
991 		microtime(&tv0);
992 	}
993 
994 	error = sys_select(l, &bsa, retval);
995 	if (error) {
996 		/*
997 		 * See fs/select.c in the Linux kernel.  Without this,
998 		 * Maelstrom doesn't work.
999 		 */
1000 		if (error == ERESTART)
1001 			error = EINTR;
1002 		return error;
1003 	}
1004 
1005 	if (timeout) {
1006 		if (*retval) {
1007 			/*
1008 			 * Compute how much time was left of the timeout,
1009 			 * by subtracting the current time and the time
1010 			 * before we started the call, and subtracting
1011 			 * that result from the user-supplied value.
1012 			 */
1013 			microtime(&tv1);
1014 			timersub(&tv1, &tv0, &tv1);
1015 			timersub(&utv, &tv1, &utv);
1016 			if (utv.tv_sec < 0)
1017 				timerclear(&utv);
1018 		} else
1019 			timerclear(&utv);
1020 		if ((error = copyout(&utv, timeout, sizeof(utv))))
1021 			return error;
1022 	}
1023 
1024 	return 0;
1025 }
1026 
1027 /*
1028  * Get the process group of a certain process. Look it up
1029  * and return the value.
1030  */
1031 int
1032 linux_sys_getpgid(l, v, retval)
1033 	struct lwp *l;
1034 	void *v;
1035 	register_t *retval;
1036 {
1037 	struct linux_sys_getpgid_args /* {
1038 		syscallarg(int) pid;
1039 	} */ *uap = v;
1040 	struct proc *p = l->l_proc;
1041 	struct proc *targp;
1042 
1043 	if (SCARG(uap, pid) != 0 && SCARG(uap, pid) != p->p_pid) {
1044 		if ((targp = pfind(SCARG(uap, pid))) == 0)
1045 			return ESRCH;
1046 	}
1047 	else
1048 		targp = p;
1049 
1050 	retval[0] = targp->p_pgid;
1051 	return 0;
1052 }
1053 
1054 /*
1055  * Set the 'personality' (emulation mode) for the current process. Only
1056  * accept the Linux personality here (0). This call is needed because
1057  * the Linux ELF crt0 issues it in an ugly kludge to make sure that
1058  * ELF binaries run in Linux mode, not SVR4 mode.
1059  */
1060 int
1061 linux_sys_personality(l, v, retval)
1062 	struct lwp *l;
1063 	void *v;
1064 	register_t *retval;
1065 {
1066 	struct linux_sys_personality_args /* {
1067 		syscallarg(int) per;
1068 	} */ *uap = v;
1069 
1070 	if (SCARG(uap, per) != 0)
1071 		return EINVAL;
1072 	retval[0] = 0;
1073 	return 0;
1074 }
1075 
1076 #if defined(__i386__) || defined(__m68k__)
1077 /*
1078  * The calls are here because of type conversions.
1079  */
1080 int
1081 linux_sys_setreuid16(l, v, retval)
1082 	struct lwp *l;
1083 	void *v;
1084 	register_t *retval;
1085 {
1086 	struct linux_sys_setreuid16_args /* {
1087 		syscallarg(int) ruid;
1088 		syscallarg(int) euid;
1089 	} */ *uap = v;
1090 	struct sys_setreuid_args bsa;
1091 
1092 	SCARG(&bsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
1093 		(uid_t)-1 : SCARG(uap, ruid);
1094 	SCARG(&bsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
1095 		(uid_t)-1 : SCARG(uap, euid);
1096 
1097 	return sys_setreuid(l, &bsa, retval);
1098 }
1099 
1100 int
1101 linux_sys_setregid16(l, v, retval)
1102 	struct lwp *l;
1103 	void *v;
1104 	register_t *retval;
1105 {
1106 	struct linux_sys_setregid16_args /* {
1107 		syscallarg(int) rgid;
1108 		syscallarg(int) egid;
1109 	} */ *uap = v;
1110 	struct sys_setregid_args bsa;
1111 
1112 	SCARG(&bsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
1113 		(uid_t)-1 : SCARG(uap, rgid);
1114 	SCARG(&bsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
1115 		(uid_t)-1 : SCARG(uap, egid);
1116 
1117 	return sys_setregid(l, &bsa, retval);
1118 }
1119 
1120 int
1121 linux_sys_setresuid16(l, v, retval)
1122 	struct lwp *l;
1123 	void *v;
1124 	register_t *retval;
1125 {
1126 	struct linux_sys_setresuid16_args /* {
1127 		syscallarg(uid_t) ruid;
1128 		syscallarg(uid_t) euid;
1129 		syscallarg(uid_t) suid;
1130 	} */ *uap = v;
1131 	struct linux_sys_setresuid16_args lsa;
1132 
1133 	SCARG(&lsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
1134 		(uid_t)-1 : SCARG(uap, ruid);
1135 	SCARG(&lsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
1136 		(uid_t)-1 : SCARG(uap, euid);
1137 	SCARG(&lsa, suid) = ((linux_uid_t)SCARG(uap, suid) == (linux_uid_t)-1) ?
1138 		(uid_t)-1 : SCARG(uap, suid);
1139 
1140 	return linux_sys_setresuid(l, &lsa, retval);
1141 }
1142 
1143 int
1144 linux_sys_setresgid16(l, v, retval)
1145 	struct lwp *l;
1146 	void *v;
1147 	register_t *retval;
1148 {
1149 	struct linux_sys_setresgid16_args /* {
1150 		syscallarg(gid_t) rgid;
1151 		syscallarg(gid_t) egid;
1152 		syscallarg(gid_t) sgid;
1153 	} */ *uap = v;
1154 	struct linux_sys_setresgid16_args lsa;
1155 
1156 	SCARG(&lsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
1157 		(gid_t)-1 : SCARG(uap, rgid);
1158 	SCARG(&lsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
1159 		(gid_t)-1 : SCARG(uap, egid);
1160 	SCARG(&lsa, sgid) = ((linux_gid_t)SCARG(uap, sgid) == (linux_gid_t)-1) ?
1161 		(gid_t)-1 : SCARG(uap, sgid);
1162 
1163 	return linux_sys_setresgid(l, &lsa, retval);
1164 }
1165 
1166 int
1167 linux_sys_getgroups16(l, v, retval)
1168 	struct lwp *l;
1169 	void *v;
1170 	register_t *retval;
1171 {
1172 	struct linux_sys_getgroups16_args /* {
1173 		syscallarg(int) gidsetsize;
1174 		syscallarg(linux_gid_t *) gidset;
1175 	} */ *uap = v;
1176 	struct proc *p = l->l_proc;
1177 	caddr_t sg;
1178 	int n, error, i;
1179 	struct sys_getgroups_args bsa;
1180 	gid_t *bset, *kbset;
1181 	linux_gid_t *lset;
1182 	struct pcred *pc = p->p_cred;
1183 
1184 	n = SCARG(uap, gidsetsize);
1185 	if (n < 0)
1186 		return EINVAL;
1187 	error = 0;
1188 	bset = kbset = NULL;
1189 	lset = NULL;
1190 	if (n > 0) {
1191 		n = min(pc->pc_ucred->cr_ngroups, n);
1192 		sg = stackgap_init(p, 0);
1193 		bset = stackgap_alloc(p, &sg, n * sizeof (gid_t));
1194 		kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK);
1195 		lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
1196 		if (bset == NULL || kbset == NULL || lset == NULL)
1197 			return ENOMEM;
1198 		SCARG(&bsa, gidsetsize) = n;
1199 		SCARG(&bsa, gidset) = bset;
1200 		error = sys_getgroups(l, &bsa, retval);
1201 		if (error != 0)
1202 			goto out;
1203 		error = copyin(bset, kbset, n * sizeof (gid_t));
1204 		if (error != 0)
1205 			goto out;
1206 		for (i = 0; i < n; i++)
1207 			lset[i] = (linux_gid_t)kbset[i];
1208 		error = copyout(lset, SCARG(uap, gidset),
1209 		    n * sizeof (linux_gid_t));
1210 	} else
1211 		*retval = pc->pc_ucred->cr_ngroups;
1212 out:
1213 	if (kbset != NULL)
1214 		free(kbset, M_TEMP);
1215 	if (lset != NULL)
1216 		free(lset, M_TEMP);
1217 	return error;
1218 }
1219 
1220 int
1221 linux_sys_setgroups16(l, v, retval)
1222 	struct lwp *l;
1223 	void *v;
1224 	register_t *retval;
1225 {
1226 	struct linux_sys_setgroups16_args /* {
1227 		syscallarg(int) gidsetsize;
1228 		syscallarg(linux_gid_t *) gidset;
1229 	} */ *uap = v;
1230 	struct proc *p = l->l_proc;
1231 	caddr_t sg;
1232 	int n;
1233 	int error, i;
1234 	struct sys_setgroups_args bsa;
1235 	gid_t *bset, *kbset;
1236 	linux_gid_t *lset;
1237 
1238 	n = SCARG(uap, gidsetsize);
1239 	if (n < 0 || n > NGROUPS)
1240 		return EINVAL;
1241 	sg = stackgap_init(p, 0);
1242 	bset = stackgap_alloc(p, &sg, n * sizeof (gid_t));
1243 	lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
1244 	kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK);
1245 	if (lset == NULL || bset == NULL)
1246 		return ENOMEM;
1247 	error = copyin(SCARG(uap, gidset), lset, n * sizeof (linux_gid_t));
1248 	if (error != 0)
1249 		goto out;
1250 	for (i = 0; i < n; i++)
1251 		kbset[i] = (gid_t)lset[i];
1252 	error = copyout(kbset, bset, n * sizeof (gid_t));
1253 	if (error != 0)
1254 		goto out;
1255 	SCARG(&bsa, gidsetsize) = n;
1256 	SCARG(&bsa, gidset) = bset;
1257 	error = sys_setgroups(l, &bsa, retval);
1258 
1259 out:
1260 	if (lset != NULL)
1261 		free(lset, M_TEMP);
1262 	if (kbset != NULL)
1263 		free(kbset, M_TEMP);
1264 
1265 	return error;
1266 }
1267 
1268 #endif /* __i386__ || __m68k__ */
1269 
1270 /*
1271  * We have nonexistent fsuid equal to uid.
1272  * If modification is requested, refuse.
1273  */
1274 int
1275 linux_sys_setfsuid(l, v, retval)
1276 	 struct lwp *l;
1277 	 void *v;
1278 	 register_t *retval;
1279 {
1280 	 struct linux_sys_setfsuid_args /* {
1281 		 syscallarg(uid_t) uid;
1282 	 } */ *uap = v;
1283 	 struct proc *p = l->l_proc;
1284 	 uid_t uid;
1285 
1286 	 uid = SCARG(uap, uid);
1287 	 if (p->p_cred->p_ruid != uid)
1288 		 return sys_nosys(l, v, retval);
1289 	 else
1290 		 return (0);
1291 }
1292 
1293 /* XXX XXX XXX */
1294 #ifndef alpha
1295 int
1296 linux_sys_getfsuid(l, v, retval)
1297 	struct lwp *l;
1298 	void *v;
1299 	register_t *retval;
1300 {
1301 	return sys_getuid(l, v, retval);
1302 }
1303 #endif
1304 
1305 int
1306 linux_sys_setresuid(l, v, retval)
1307 	struct lwp *l;
1308 	void *v;
1309 	register_t *retval;
1310 {
1311 	struct linux_sys_setresuid_args /* {
1312 		syscallarg(uid_t) ruid;
1313 		syscallarg(uid_t) euid;
1314 		syscallarg(uid_t) suid;
1315 	} */ *uap = v;
1316 
1317 	/*
1318 	 * Note: These checks are a little different than the NetBSD
1319 	 * setreuid(2) call performs.  This precisely follows the
1320 	 * behavior of the Linux kernel.
1321 	 */
1322 
1323 	return do_setresuid(l, SCARG(uap, ruid), SCARG(uap, euid),
1324 			    SCARG(uap, suid),
1325 			    ID_R_EQ_R | ID_R_EQ_E | ID_R_EQ_S |
1326 			    ID_E_EQ_R | ID_E_EQ_E | ID_E_EQ_S |
1327 			    ID_S_EQ_R | ID_S_EQ_E | ID_S_EQ_S );
1328 }
1329 
1330 int
1331 linux_sys_getresuid(l, v, retval)
1332 	struct lwp *l;
1333 	void *v;
1334 	register_t *retval;
1335 {
1336 	struct linux_sys_getresuid_args /* {
1337 		syscallarg(uid_t *) ruid;
1338 		syscallarg(uid_t *) euid;
1339 		syscallarg(uid_t *) suid;
1340 	} */ *uap = v;
1341 	struct proc *p = l->l_proc;
1342 	struct pcred *pc = p->p_cred;
1343 	int error;
1344 
1345 	/*
1346 	 * Linux copies these values out to userspace like so:
1347 	 *
1348 	 *	1. Copy out ruid.
1349 	 *	2. If that succeeds, copy out euid.
1350 	 *	3. If both of those succeed, copy out suid.
1351 	 */
1352 	if ((error = copyout(&pc->p_ruid, SCARG(uap, ruid),
1353 			     sizeof(uid_t))) != 0)
1354 		return (error);
1355 
1356 	if ((error = copyout(&pc->pc_ucred->cr_uid, SCARG(uap, euid),
1357 			     sizeof(uid_t))) != 0)
1358 		return (error);
1359 
1360 	return (copyout(&pc->p_svuid, SCARG(uap, suid), sizeof(uid_t)));
1361 }
1362 
1363 int
1364 linux_sys_ptrace(l, v, retval)
1365 	struct lwp *l;
1366 	void *v;
1367 	register_t *retval;
1368 {
1369 	struct linux_sys_ptrace_args /* {
1370 		i386, m68k, powerpc: T=int
1371 		alpha: T=long
1372 		syscallarg(T) request;
1373 		syscallarg(T) pid;
1374 		syscallarg(T) addr;
1375 		syscallarg(T) data;
1376 	} */ *uap = v;
1377 	const int *ptr;
1378 	int request;
1379 	int error;
1380 
1381 	ptr = linux_ptrace_request_map;
1382 	request = SCARG(uap, request);
1383 	while (*ptr != -1)
1384 		if (*ptr++ == request) {
1385 			struct sys_ptrace_args pta;
1386 
1387 			SCARG(&pta, req) = *ptr;
1388 			SCARG(&pta, pid) = SCARG(uap, pid);
1389 			SCARG(&pta, addr) = (caddr_t)SCARG(uap, addr);
1390 			SCARG(&pta, data) = SCARG(uap, data);
1391 
1392 			/*
1393 			 * Linux ptrace(PTRACE_CONT, pid, 0, 0) means actually
1394 			 * to continue where the process left off previously.
1395 			 * The same thing is achieved by addr == (caddr_t) 1
1396 			 * on NetBSD, so rewrite 'addr' appropriately.
1397 			 */
1398 			if (request == LINUX_PTRACE_CONT && SCARG(uap, addr)==0)
1399 				SCARG(&pta, addr) = (caddr_t) 1;
1400 
1401 			error = sys_ptrace(l, &pta, retval);
1402 			if (error)
1403 				return error;
1404 			switch (request) {
1405 			case LINUX_PTRACE_PEEKTEXT:
1406 			case LINUX_PTRACE_PEEKDATA:
1407 				error = copyout (retval,
1408 				    (caddr_t)SCARG(uap, data), sizeof *retval);
1409 				*retval = SCARG(uap, data);
1410 				break;
1411 			default:
1412 				break;
1413 			}
1414 			return error;
1415 		}
1416 		else
1417 			ptr++;
1418 
1419 	return LINUX_SYS_PTRACE_ARCH(l, uap, retval);
1420 }
1421 
1422 int
1423 linux_sys_reboot(struct lwp *l, void *v, register_t *retval)
1424 {
1425 	struct linux_sys_reboot_args /* {
1426 		syscallarg(int) magic1;
1427 		syscallarg(int) magic2;
1428 		syscallarg(int) cmd;
1429 		syscallarg(void *) arg;
1430 	} */ *uap = v;
1431 	struct sys_reboot_args /* {
1432 		syscallarg(int) opt;
1433 		syscallarg(char *) bootstr;
1434 	} */ sra;
1435 	struct proc *p = l->l_proc;
1436 	int error;
1437 
1438 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1439 		return(error);
1440 
1441 	if (SCARG(uap, magic1) != LINUX_REBOOT_MAGIC1)
1442 		return(EINVAL);
1443 	if (SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2 &&
1444 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2A &&
1445 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2B)
1446 		return(EINVAL);
1447 
1448 	switch (SCARG(uap, cmd)) {
1449 	case LINUX_REBOOT_CMD_RESTART:
1450 		SCARG(&sra, opt) = RB_AUTOBOOT;
1451 		break;
1452 	case LINUX_REBOOT_CMD_HALT:
1453 		SCARG(&sra, opt) = RB_HALT;
1454 		break;
1455 	case LINUX_REBOOT_CMD_POWER_OFF:
1456 		SCARG(&sra, opt) = RB_HALT|RB_POWERDOWN;
1457 		break;
1458 	case LINUX_REBOOT_CMD_RESTART2:
1459 		/* Reboot with an argument. */
1460 		SCARG(&sra, opt) = RB_AUTOBOOT|RB_STRING;
1461 		SCARG(&sra, bootstr) = SCARG(uap, arg);
1462 		break;
1463 	case LINUX_REBOOT_CMD_CAD_ON:
1464 		return(EINVAL);	/* We don't implement ctrl-alt-delete */
1465 	case LINUX_REBOOT_CMD_CAD_OFF:
1466 		return(0);
1467 	default:
1468 		return(EINVAL);
1469 	}
1470 
1471 	return(sys_reboot(l, &sra, retval));
1472 }
1473 
1474 /*
1475  * Copy of compat_12_sys_swapon().
1476  */
1477 int
1478 linux_sys_swapon(l, v, retval)
1479 	struct lwp *l;
1480 	void *v;
1481 	register_t *retval;
1482 {
1483 	struct sys_swapctl_args ua;
1484 	struct linux_sys_swapon_args /* {
1485 		syscallarg(const char *) name;
1486 	} */ *uap = v;
1487 
1488 	SCARG(&ua, cmd) = SWAP_ON;
1489 	SCARG(&ua, arg) = (void *)SCARG(uap, name);
1490 	SCARG(&ua, misc) = 0;	/* priority */
1491 	return (sys_swapctl(l, &ua, retval));
1492 }
1493 
1494 /*
1495  * Stop swapping to the file or block device specified by path.
1496  */
1497 int
1498 linux_sys_swapoff(l, v, retval)
1499 	struct lwp *l;
1500 	void *v;
1501 	register_t *retval;
1502 {
1503 	struct sys_swapctl_args ua;
1504 	struct linux_sys_swapoff_args /* {
1505 		syscallarg(const char *) path;
1506 	} */ *uap = v;
1507 
1508 	SCARG(&ua, cmd) = SWAP_OFF;
1509 	SCARG(&ua, arg) = (void *)SCARG(uap, path);
1510 	return (sys_swapctl(l, &ua, retval));
1511 }
1512 
1513 /*
1514  * Copy of compat_09_sys_setdomainname()
1515  */
1516 /* ARGSUSED */
1517 int
1518 linux_sys_setdomainname(l, v, retval)
1519 	struct lwp *l;
1520 	void *v;
1521 	register_t *retval;
1522 {
1523 	struct linux_sys_setdomainname_args /* {
1524 		syscallarg(char *) domainname;
1525 		syscallarg(int) len;
1526 	} */ *uap = v;
1527 	int name[2];
1528 
1529 	name[0] = CTL_KERN;
1530 	name[1] = KERN_DOMAINNAME;
1531 	return (old_sysctl(&name[0], 2, 0, 0, SCARG(uap, domainname),
1532 			    SCARG(uap, len), l));
1533 }
1534 
1535 /*
1536  * sysinfo()
1537  */
1538 /* ARGSUSED */
1539 int
1540 linux_sys_sysinfo(l, v, retval)
1541 	struct lwp *l;
1542 	void *v;
1543 	register_t *retval;
1544 {
1545 	struct linux_sys_sysinfo_args /* {
1546 		syscallarg(struct linux_sysinfo *) arg;
1547 	} */ *uap = v;
1548 	struct linux_sysinfo si;
1549 	struct loadavg *la;
1550 
1551 	si.uptime = time.tv_sec - boottime.tv_sec;
1552 	la = &averunnable;
1553 	si.loads[0] = la->ldavg[0] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1554 	si.loads[1] = la->ldavg[1] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1555 	si.loads[2] = la->ldavg[2] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1556 	si.totalram = ctob(physmem);
1557 	si.freeram = uvmexp.free * uvmexp.pagesize;
1558 	si.sharedram = 0;	/* XXX */
1559 	si.bufferram = uvmexp.filepages * uvmexp.pagesize;
1560 	si.totalswap = uvmexp.swpages * uvmexp.pagesize;
1561 	si.freeswap = (uvmexp.swpages - uvmexp.swpginuse) * uvmexp.pagesize;
1562 	si.procs = nprocs;
1563 
1564 	/* The following are only present in newer Linux kernels. */
1565 	si.totalbig = 0;
1566 	si.freebig = 0;
1567 	si.mem_unit = 1;
1568 
1569 	return (copyout(&si, SCARG(uap, arg), sizeof si));
1570 }
1571 
1572 #define bsd_to_linux_rlimit1(l, b, f) \
1573     (l)->f = ((b)->f == RLIM_INFINITY || \
1574 	     ((b)->f & 0xffffffff00000000ULL) != 0) ? \
1575     LINUX_RLIM_INFINITY : (int32_t)(b)->f
1576 #define bsd_to_linux_rlimit(l, b) \
1577     bsd_to_linux_rlimit1(l, b, rlim_cur); \
1578     bsd_to_linux_rlimit1(l, b, rlim_max)
1579 
1580 #define linux_to_bsd_rlimit1(b, l, f) \
1581     (b)->f = (l)->f == LINUX_RLIM_INFINITY ? RLIM_INFINITY : (l)->f
1582 #define linux_to_bsd_rlimit(b, l) \
1583     linux_to_bsd_rlimit1(b, l, rlim_cur); \
1584     linux_to_bsd_rlimit1(b, l, rlim_max)
1585 
1586 static int
1587 linux_to_bsd_limit(lim)
1588 	int lim;
1589 {
1590 	switch (lim) {
1591 	case LINUX_RLIMIT_CPU:
1592 		return RLIMIT_CPU;
1593 	case LINUX_RLIMIT_FSIZE:
1594 		return RLIMIT_FSIZE;
1595 	case LINUX_RLIMIT_DATA:
1596 		return RLIMIT_DATA;
1597 	case LINUX_RLIMIT_STACK:
1598 		return RLIMIT_STACK;
1599 	case LINUX_RLIMIT_CORE:
1600 		return RLIMIT_CORE;
1601 	case LINUX_RLIMIT_RSS:
1602 		return RLIMIT_RSS;
1603 	case LINUX_RLIMIT_NPROC:
1604 		return RLIMIT_NPROC;
1605 	case LINUX_RLIMIT_NOFILE:
1606 		return RLIMIT_NOFILE;
1607 	case LINUX_RLIMIT_MEMLOCK:
1608 		return RLIMIT_MEMLOCK;
1609 	case LINUX_RLIMIT_AS:
1610 	case LINUX_RLIMIT_LOCKS:
1611 		return -EOPNOTSUPP;
1612 	default:
1613 		return -EINVAL;
1614 	}
1615 }
1616 
1617 
1618 int
1619 linux_sys_getrlimit(l, v, retval)
1620 	struct lwp *l;
1621 	void *v;
1622 	register_t *retval;
1623 {
1624 	struct linux_sys_getrlimit_args /* {
1625 		syscallarg(int) which;
1626 		syscallarg(struct orlimit *) rlp;
1627 	} */ *uap = v;
1628 	struct proc *p = l->l_proc;
1629 	caddr_t sg = stackgap_init(p, 0);
1630 	struct sys_getrlimit_args ap;
1631 	struct rlimit rl;
1632 	struct orlimit orl;
1633 	int error;
1634 
1635 	SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which));
1636 	if ((error = SCARG(&ap, which)) < 0)
1637 		return -error;
1638 	SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl);
1639 	if ((error = sys_getrlimit(l, &ap, retval)) != 0)
1640 		return error;
1641 	if ((error = copyin(SCARG(&ap, rlp), &rl, sizeof(rl))) != 0)
1642 		return error;
1643 	bsd_to_linux_rlimit(&orl, &rl);
1644 	return copyout(&orl, SCARG(uap, rlp), sizeof(orl));
1645 }
1646 
1647 int
1648 linux_sys_setrlimit(l, v, retval)
1649 	struct lwp *l;
1650 	void *v;
1651 	register_t *retval;
1652 {
1653 	struct linux_sys_setrlimit_args /* {
1654 		syscallarg(int) which;
1655 		syscallarg(struct orlimit *) rlp;
1656 	} */ *uap = v;
1657 	struct proc *p = l->l_proc;
1658 	caddr_t sg = stackgap_init(p, 0);
1659 	struct sys_setrlimit_args ap;
1660 	struct rlimit rl;
1661 	struct orlimit orl;
1662 	int error;
1663 
1664 	SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which));
1665 	SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl);
1666 	if ((error = SCARG(&ap, which)) < 0)
1667 		return -error;
1668 	if ((error = copyin(SCARG(uap, rlp), &orl, sizeof(orl))) != 0)
1669 		return error;
1670 	linux_to_bsd_rlimit(&rl, &orl);
1671 	/* XXX: alpha complains about this */
1672 	if ((error = copyout(&rl, (void *)SCARG(&ap, rlp), sizeof(rl))) != 0)
1673 		return error;
1674 	return sys_setrlimit(l, &ap, retval);
1675 }
1676 
1677 #ifndef __mips__
1678 /* XXX: this doesn't look 100% common, at least mips doesn't have it */
1679 int
1680 linux_sys_ugetrlimit(l, v, retval)
1681 	struct lwp *l;
1682 	void *v;
1683 	register_t *retval;
1684 {
1685 	return linux_sys_getrlimit(l, v, retval);
1686 }
1687 #endif
1688 
1689 /*
1690  * This gets called for unsupported syscalls. The difference to sys_nosys()
1691  * is that process does not get SIGSYS, the call just returns with ENOSYS.
1692  * This is the way Linux does it and glibc depends on this behaviour.
1693  */
1694 int
1695 linux_sys_nosys(l, v, retval)
1696 	struct lwp *l;
1697 	void *v;
1698 	register_t *retval;
1699 {
1700 	return (ENOSYS);
1701 }
1702