xref: /netbsd-src/sys/compat/linux/common/linux_misc.c (revision ce2c90c7c172d95d2402a5b3d96d8f8e6d138a21)
1 /*	$NetBSD: linux_misc.c,v 1.163 2006/10/12 01:30:48 christos 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 1.163 2006/10/12 01:30:48 christos Exp $");
68 
69 #if defined(_KERNEL_OPT)
70 #include "opt_ptrace.h"
71 #endif
72 
73 #include <sys/param.h>
74 #include <sys/systm.h>
75 #include <sys/namei.h>
76 #include <sys/proc.h>
77 #include <sys/dirent.h>
78 #include <sys/file.h>
79 #include <sys/stat.h>
80 #include <sys/filedesc.h>
81 #include <sys/ioctl.h>
82 #include <sys/kernel.h>
83 #include <sys/malloc.h>
84 #include <sys/mbuf.h>
85 #include <sys/mman.h>
86 #include <sys/mount.h>
87 #include <sys/reboot.h>
88 #include <sys/resource.h>
89 #include <sys/resourcevar.h>
90 #include <sys/signal.h>
91 #include <sys/signalvar.h>
92 #include <sys/socket.h>
93 #include <sys/time.h>
94 #include <sys/times.h>
95 #include <sys/vnode.h>
96 #include <sys/uio.h>
97 #include <sys/wait.h>
98 #include <sys/utsname.h>
99 #include <sys/unistd.h>
100 #include <sys/swap.h>		/* for SWAP_ON */
101 #include <sys/sysctl.h>		/* for KERN_DOMAINNAME */
102 #include <sys/kauth.h>
103 
104 #include <sys/ptrace.h>
105 #include <machine/ptrace.h>
106 
107 #include <sys/sa.h>
108 #include <sys/syscall.h>
109 #include <sys/syscallargs.h>
110 
111 #include <compat/linux/common/linux_machdep.h>
112 #include <compat/linux/common/linux_types.h>
113 #include <compat/linux/common/linux_signal.h>
114 
115 #include <compat/linux/linux_syscallargs.h>
116 
117 #include <compat/linux/common/linux_fcntl.h>
118 #include <compat/linux/common/linux_mmap.h>
119 #include <compat/linux/common/linux_dirent.h>
120 #include <compat/linux/common/linux_util.h>
121 #include <compat/linux/common/linux_misc.h>
122 #ifndef COMPAT_LINUX32
123 #include <compat/linux/common/linux_limit.h>
124 #endif
125 #include <compat/linux/common/linux_ptrace.h>
126 #include <compat/linux/common/linux_reboot.h>
127 #include <compat/linux/common/linux_emuldata.h>
128 
129 #ifndef COMPAT_LINUX32
130 const int linux_ptrace_request_map[] = {
131 	LINUX_PTRACE_TRACEME,	PT_TRACE_ME,
132 	LINUX_PTRACE_PEEKTEXT,	PT_READ_I,
133 	LINUX_PTRACE_PEEKDATA,	PT_READ_D,
134 	LINUX_PTRACE_POKETEXT,	PT_WRITE_I,
135 	LINUX_PTRACE_POKEDATA,	PT_WRITE_D,
136 	LINUX_PTRACE_CONT,	PT_CONTINUE,
137 	LINUX_PTRACE_KILL,	PT_KILL,
138 	LINUX_PTRACE_ATTACH,	PT_ATTACH,
139 	LINUX_PTRACE_DETACH,	PT_DETACH,
140 # ifdef PT_STEP
141 	LINUX_PTRACE_SINGLESTEP,	PT_STEP,
142 # endif
143 	-1
144 };
145 
146 const struct linux_mnttypes linux_fstypes[] = {
147 	{ MOUNT_FFS,		LINUX_DEFAULT_SUPER_MAGIC	},
148 	{ MOUNT_NFS,		LINUX_NFS_SUPER_MAGIC 		},
149 	{ MOUNT_MFS,		LINUX_DEFAULT_SUPER_MAGIC	},
150 	{ MOUNT_MSDOS,		LINUX_MSDOS_SUPER_MAGIC		},
151 	{ MOUNT_LFS,		LINUX_DEFAULT_SUPER_MAGIC	},
152 	{ MOUNT_FDESC,		LINUX_DEFAULT_SUPER_MAGIC	},
153 	{ MOUNT_PORTAL,		LINUX_DEFAULT_SUPER_MAGIC	},
154 	{ MOUNT_NULL,		LINUX_DEFAULT_SUPER_MAGIC	},
155 	{ MOUNT_OVERLAY,	LINUX_DEFAULT_SUPER_MAGIC	},
156 	{ MOUNT_UMAP,		LINUX_DEFAULT_SUPER_MAGIC	},
157 	{ MOUNT_KERNFS,		LINUX_DEFAULT_SUPER_MAGIC	},
158 	{ MOUNT_PROCFS,		LINUX_PROC_SUPER_MAGIC		},
159 	{ MOUNT_AFS,		LINUX_DEFAULT_SUPER_MAGIC	},
160 	{ MOUNT_CD9660,		LINUX_ISOFS_SUPER_MAGIC		},
161 	{ MOUNT_UNION,		LINUX_DEFAULT_SUPER_MAGIC	},
162 	{ MOUNT_ADOSFS,		LINUX_ADFS_SUPER_MAGIC		},
163 	{ MOUNT_EXT2FS,		LINUX_EXT2_SUPER_MAGIC		},
164 	{ MOUNT_CFS,		LINUX_DEFAULT_SUPER_MAGIC	},
165 	{ MOUNT_CODA,		LINUX_CODA_SUPER_MAGIC		},
166 	{ MOUNT_FILECORE,	LINUX_DEFAULT_SUPER_MAGIC	},
167 	{ MOUNT_NTFS,		LINUX_DEFAULT_SUPER_MAGIC	},
168 	{ MOUNT_SMBFS,		LINUX_SMB_SUPER_MAGIC		},
169 	{ MOUNT_PTYFS,		LINUX_DEVPTS_SUPER_MAGIC	},
170 	{ MOUNT_TMPFS,		LINUX_DEFAULT_SUPER_MAGIC	}
171 };
172 const int linux_fstypes_cnt = sizeof(linux_fstypes) / sizeof(linux_fstypes[0]);
173 
174 # ifdef DEBUG_LINUX
175 #define DPRINTF(a)	uprintf a
176 # else
177 #define DPRINTF(a)
178 # endif
179 
180 /* Local linux_misc.c functions: */
181 # ifndef __amd64__
182 static void bsd_to_linux_statfs __P((const struct statvfs *,
183     struct linux_statfs *));
184 # endif
185 static void linux_to_bsd_mmap_args __P((struct sys_mmap_args *,
186     const struct linux_sys_mmap_args *));
187 static int linux_mmap __P((struct lwp *, struct linux_sys_mmap_args *,
188     register_t *, off_t));
189 
190 
191 /*
192  * The information on a terminated (or stopped) process needs
193  * to be converted in order for Linux binaries to get a valid signal
194  * number out of it.
195  */
196 void
197 bsd_to_linux_wstat(st)
198 	int *st;
199 {
200 
201 	int sig;
202 
203 	if (WIFSIGNALED(*st)) {
204 		sig = WTERMSIG(*st);
205 		if (sig >= 0 && sig < NSIG)
206 			*st= (*st& ~0177) | native_to_linux_signo[sig];
207 	} else if (WIFSTOPPED(*st)) {
208 		sig = WSTOPSIG(*st);
209 		if (sig >= 0 && sig < NSIG)
210 			*st = (*st & ~0xff00) |
211 			    (native_to_linux_signo[sig] << 8);
212 	}
213 }
214 
215 /*
216  * wait4(2).  Passed on to the NetBSD call, surrounded by code to
217  * reserve some space for a NetBSD-style wait status, and converting
218  * it to what Linux wants.
219  */
220 int
221 linux_sys_wait4(l, v, retval)
222 	struct lwp *l;
223 	void *v;
224 	register_t *retval;
225 {
226 	struct linux_sys_wait4_args /* {
227 		syscallarg(int) pid;
228 		syscallarg(int *) status;
229 		syscallarg(int) options;
230 		syscallarg(struct rusage *) rusage;
231 	} */ *uap = v;
232 	struct proc *p = l->l_proc;
233 	struct sys_wait4_args w4a;
234 	int error, *status, tstat, options, linux_options;
235 	caddr_t sg;
236 
237 	if (SCARG(uap, status) != NULL) {
238 		sg = stackgap_init(p, 0);
239 		status = (int *) stackgap_alloc(p, &sg, sizeof *status);
240 	} else
241 		status = NULL;
242 
243 	linux_options = SCARG(uap, options);
244 	options = 0;
245 	if (linux_options & ~(LINUX_WAIT4_KNOWNFLAGS))
246 		return (EINVAL);
247 
248 	if (linux_options & LINUX_WAIT4_WNOHANG)
249 		options |= WNOHANG;
250 	if (linux_options & LINUX_WAIT4_WUNTRACED)
251 		options |= WUNTRACED;
252 	if (linux_options & LINUX_WAIT4_WALL)
253 		options |= WALLSIG;
254 	if (linux_options & LINUX_WAIT4_WCLONE)
255 		options |= WALTSIG;
256 # ifdef DIAGNOSTIC
257 	if (linux_options & LINUX_WAIT4_WNOTHREAD)
258 		printf("WARNING: %s: linux process %d.%d called "
259 		       "waitpid with __WNOTHREAD set!",
260 		       __FILE__, p->p_pid, l->l_lid);
261 
262 # endif
263 
264 	SCARG(&w4a, pid) = SCARG(uap, pid);
265 	SCARG(&w4a, status) = status;
266 	SCARG(&w4a, options) = options;
267 	SCARG(&w4a, rusage) = SCARG(uap, rusage);
268 
269 	if ((error = sys_wait4(l, &w4a, retval)))
270 		return error;
271 
272 	sigdelset(&p->p_sigctx.ps_siglist, SIGCHLD);
273 
274 	if (status != NULL) {
275 		if ((error = copyin(status, &tstat, sizeof tstat)))
276 			return error;
277 
278 		bsd_to_linux_wstat(&tstat);
279 		return copyout(&tstat, SCARG(uap, status), sizeof tstat);
280 	}
281 
282 	return 0;
283 }
284 
285 /*
286  * Linux brk(2). The check if the new address is >= the old one is
287  * done in the kernel in Linux. NetBSD does it in the library.
288  */
289 int
290 linux_sys_brk(l, v, retval)
291 	struct lwp *l;
292 	void *v;
293 	register_t *retval;
294 {
295 	struct linux_sys_brk_args /* {
296 		syscallarg(char *) nsize;
297 	} */ *uap = v;
298 	struct proc *p = l->l_proc;
299 	char *nbrk = SCARG(uap, nsize);
300 	struct sys_obreak_args oba;
301 	struct vmspace *vm = p->p_vmspace;
302 	struct linux_emuldata *ed = (struct linux_emuldata*)p->p_emuldata;
303 
304 	SCARG(&oba, nsize) = nbrk;
305 
306 	if ((caddr_t) nbrk > vm->vm_daddr && sys_obreak(l, &oba, retval) == 0)
307 		ed->s->p_break = (char*)nbrk;
308 	else
309 		nbrk = ed->s->p_break;
310 
311 	retval[0] = (register_t)nbrk;
312 
313 	return 0;
314 }
315 
316 # ifndef __amd64__
317 /*
318  * Convert NetBSD statvfs structure to Linux statfs structure.
319  * Linux doesn't have f_flag, and we can't set f_frsize due
320  * to glibc statvfs() bug (see below).
321  */
322 static void
323 bsd_to_linux_statfs(bsp, lsp)
324 	const struct statvfs *bsp;
325 	struct linux_statfs *lsp;
326 {
327 	int i;
328 
329 	for (i = 0; i < linux_fstypes_cnt; i++) {
330 		if (strcmp(bsp->f_fstypename, linux_fstypes[i].bsd) == 0) {
331 			lsp->l_ftype = linux_fstypes[i].linux;
332 			break;
333 		}
334 	}
335 
336 	if (i == linux_fstypes_cnt) {
337 		DPRINTF(("unhandled fstype in linux emulation: %s\n",
338 		    bsp->f_fstypename));
339 		lsp->l_ftype = LINUX_DEFAULT_SUPER_MAGIC;
340 	}
341 
342 	/*
343 	 * The sizes are expressed in number of blocks. The block
344 	 * size used for the size is f_frsize for POSIX-compliant
345 	 * statvfs. Linux statfs uses f_bsize as the block size
346 	 * (f_frsize used to not be available in Linux struct statfs).
347 	 * However, glibc 2.3.3 statvfs() wrapper fails to adjust the block
348 	 * counts for different f_frsize if f_frsize is provided by the kernel.
349 	 * POSIX conforming apps thus get wrong size if f_frsize
350 	 * is different to f_bsize. Thus, we just pretend we don't
351 	 * support f_frsize.
352 	 */
353 
354 	lsp->l_fbsize = bsp->f_frsize;
355 	lsp->l_ffrsize = 0;			/* compat */
356 	lsp->l_fblocks = bsp->f_blocks;
357 	lsp->l_fbfree = bsp->f_bfree;
358 	lsp->l_fbavail = bsp->f_bavail;
359 	lsp->l_ffiles = bsp->f_files;
360 	lsp->l_fffree = bsp->f_ffree;
361 	/* Linux sets the fsid to 0..., we don't */
362 	lsp->l_ffsid.val[0] = bsp->f_fsidx.__fsid_val[0];
363 	lsp->l_ffsid.val[1] = bsp->f_fsidx.__fsid_val[1];
364 	lsp->l_fnamelen = bsp->f_namemax;
365 	(void)memset(lsp->l_fspare, 0, sizeof(lsp->l_fspare));
366 }
367 
368 /*
369  * Implement the fs stat functions. Straightforward.
370  */
371 int
372 linux_sys_statfs(l, v, retval)
373 	struct lwp *l;
374 	void *v;
375 	register_t *retval;
376 {
377 	struct linux_sys_statfs_args /* {
378 		syscallarg(const char *) path;
379 		syscallarg(struct linux_statfs *) sp;
380 	} */ *uap = v;
381 	struct proc *p = l->l_proc;
382 	struct statvfs *btmp, *bsp;
383 	struct linux_statfs ltmp;
384 	struct sys_statvfs1_args bsa;
385 	caddr_t sg;
386 	int error;
387 
388 	sg = stackgap_init(p, 0);
389 	bsp = stackgap_alloc(p, &sg, sizeof (struct statvfs));
390 
391 	CHECK_ALT_EXIST(l, &sg, SCARG(uap, path));
392 
393 	SCARG(&bsa, path) = SCARG(uap, path);
394 	SCARG(&bsa, buf) = bsp;
395 	SCARG(&bsa, flags) = ST_WAIT;
396 
397 	if ((error = sys_statvfs1(l, &bsa, retval)))
398 		return error;
399 
400 	btmp = STATVFSBUF_GET();
401 	error = copyin(bsp, btmp, sizeof(*btmp));
402 	if (error) {
403 		goto out;
404 	}
405 	bsd_to_linux_statfs(btmp, &ltmp);
406 	error = copyout(&ltmp, SCARG(uap, sp), sizeof ltmp);
407 out:
408 	STATVFSBUF_PUT(btmp);
409 	return error;
410 }
411 
412 int
413 linux_sys_fstatfs(l, v, retval)
414 	struct lwp *l;
415 	void *v;
416 	register_t *retval;
417 {
418 	struct linux_sys_fstatfs_args /* {
419 		syscallarg(int) fd;
420 		syscallarg(struct linux_statfs *) sp;
421 	} */ *uap = v;
422 	struct proc *p = l->l_proc;
423 	struct statvfs *btmp, *bsp;
424 	struct linux_statfs ltmp;
425 	struct sys_fstatvfs1_args bsa;
426 	caddr_t sg;
427 	int error;
428 
429 	sg = stackgap_init(p, 0);
430 	bsp = stackgap_alloc(p, &sg, sizeof (struct statvfs));
431 
432 	SCARG(&bsa, fd) = SCARG(uap, fd);
433 	SCARG(&bsa, buf) = bsp;
434 	SCARG(&bsa, flags) = ST_WAIT;
435 
436 	if ((error = sys_fstatvfs1(l, &bsa, retval)))
437 		return error;
438 
439 	btmp = STATVFSBUF_GET();
440 	error = copyin(bsp, btmp, sizeof(*btmp));
441 	if (error) {
442 		goto out;
443 	}
444 	bsd_to_linux_statfs(btmp, &ltmp);
445 	error = copyout(&ltmp, SCARG(uap, sp), sizeof ltmp);
446 out:
447 	STATVFSBUF_PUT(btmp);
448 	return error;
449 }
450 # endif /* !__amd64__ */
451 
452 /*
453  * uname(). Just copy the info from the various strings stored in the
454  * kernel, and put it in the Linux utsname structure. That structure
455  * is almost the same as the NetBSD one, only it has fields 65 characters
456  * long, and an extra domainname field.
457  */
458 int
459 linux_sys_uname(struct lwp *l __unused, void *v, register_t *retval __unused)
460 {
461 	struct linux_sys_uname_args /* {
462 		syscallarg(struct linux_utsname *) up;
463 	} */ *uap = v;
464 	struct linux_utsname luts;
465 
466 	strncpy(luts.l_sysname, linux_sysname, sizeof(luts.l_sysname));
467 	strncpy(luts.l_nodename, hostname, sizeof(luts.l_nodename));
468 	strncpy(luts.l_release, linux_release, sizeof(luts.l_release));
469 	strncpy(luts.l_version, linux_version, sizeof(luts.l_version));
470 # ifdef LINUX_UNAME_ARCH
471 	strncpy(luts.l_machine, LINUX_UNAME_ARCH, sizeof(luts.l_machine));
472 # else
473 	strncpy(luts.l_machine, machine, sizeof(luts.l_machine));
474 # endif
475 	strncpy(luts.l_domainname, domainname, sizeof(luts.l_domainname));
476 
477 	return copyout(&luts, SCARG(uap, up), sizeof(luts));
478 }
479 
480 /* Used directly on: alpha, mips, ppc, sparc, sparc64 */
481 /* Used indirectly on: arm, i386, m68k */
482 
483 /*
484  * New type Linux mmap call.
485  * Only called directly on machines with >= 6 free regs.
486  */
487 int
488 linux_sys_mmap(l, v, retval)
489 	struct lwp *l;
490 	void *v;
491 	register_t *retval;
492 {
493 	struct linux_sys_mmap_args /* {
494 		syscallarg(unsigned long) addr;
495 		syscallarg(size_t) len;
496 		syscallarg(int) prot;
497 		syscallarg(int) flags;
498 		syscallarg(int) fd;
499 		syscallarg(linux_off_t) offset;
500 	} */ *uap = v;
501 
502 	if (SCARG(uap, offset) & PAGE_MASK)
503 		return EINVAL;
504 
505 	return linux_mmap(l, uap, retval, SCARG(uap, offset));
506 }
507 
508 /*
509  * Guts of most architectures' mmap64() implementations.  This shares
510  * its list of arguments with linux_sys_mmap().
511  *
512  * The difference in linux_sys_mmap2() is that "offset" is actually
513  * (offset / pagesize), not an absolute byte count.  This translation
514  * to pagesize offsets is done inside glibc between the mmap64() call
515  * point, and the actual syscall.
516  */
517 int
518 linux_sys_mmap2(l, v, retval)
519 	struct lwp *l;
520 	void *v;
521 	register_t *retval;
522 {
523 	struct linux_sys_mmap2_args /* {
524 		syscallarg(unsigned long) addr;
525 		syscallarg(size_t) len;
526 		syscallarg(int) prot;
527 		syscallarg(int) flags;
528 		syscallarg(int) fd;
529 		syscallarg(linux_off_t) offset;
530 	} */ *uap = v;
531 
532 	return linux_mmap(l, uap, retval,
533 	    ((off_t)SCARG(uap, offset)) << PAGE_SHIFT);
534 }
535 
536 /*
537  * Massage arguments and call system mmap(2).
538  */
539 static int
540 linux_mmap(l, uap, retval, offset)
541 	struct lwp *l;
542 	struct linux_sys_mmap_args *uap;
543 	register_t *retval;
544 	off_t offset;
545 {
546 	struct sys_mmap_args cma;
547 	int error;
548 	size_t mmoff=0;
549 
550 	if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN) {
551 		/*
552 		 * Request for stack-like memory segment. On linux, this
553 		 * works by mmap()ping (small) segment, which is automatically
554 		 * extended when page fault happens below the currently
555 		 * allocated area. We emulate this by allocating (typically
556 		 * bigger) segment sized at current stack size limit, and
557 		 * offsetting the requested and returned address accordingly.
558 		 * Since physical pages are only allocated on-demand, this
559 		 * is effectively identical.
560 		 */
561 		rlim_t ssl = l->l_proc->p_rlimit[RLIMIT_STACK].rlim_cur;
562 
563 		if (SCARG(uap, len) < ssl) {
564 			/* Compute the address offset */
565 			mmoff = round_page(ssl) - SCARG(uap, len);
566 
567 			if (SCARG(uap, addr))
568 				SCARG(uap, addr) -= mmoff;
569 
570 			SCARG(uap, len) = (size_t) ssl;
571 		}
572 	}
573 
574 	linux_to_bsd_mmap_args(&cma, uap);
575 	SCARG(&cma, pos) = offset;
576 
577 	error = sys_mmap(l, &cma, retval);
578 	if (error)
579 		return (error);
580 
581 	/* Shift the returned address for stack-like segment if necessary */
582 	if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN && mmoff)
583 		retval[0] += mmoff;
584 
585 	return (0);
586 }
587 
588 static void
589 linux_to_bsd_mmap_args(cma, uap)
590 	struct sys_mmap_args *cma;
591 	const struct linux_sys_mmap_args *uap;
592 {
593 	int flags = MAP_TRYFIXED, fl = SCARG(uap, flags);
594 
595 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_SHARED, MAP_SHARED);
596 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_PRIVATE, MAP_PRIVATE);
597 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_FIXED, MAP_FIXED);
598 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_ANON, MAP_ANON);
599 	/* XXX XAX ERH: Any other flags here?  There are more defined... */
600 
601 	SCARG(cma, addr) = (void *)SCARG(uap, addr);
602 	SCARG(cma, len) = SCARG(uap, len);
603 	SCARG(cma, prot) = SCARG(uap, prot);
604 	if (SCARG(cma, prot) & VM_PROT_WRITE) /* XXX */
605 		SCARG(cma, prot) |= VM_PROT_READ;
606 	SCARG(cma, flags) = flags;
607 	SCARG(cma, fd) = flags & MAP_ANON ? -1 : SCARG(uap, fd);
608 	SCARG(cma, pad) = 0;
609 }
610 
611 #define	LINUX_MREMAP_MAYMOVE	1
612 #define	LINUX_MREMAP_FIXED	2
613 
614 int
615 linux_sys_mremap(l, v, retval)
616 	struct lwp *l;
617 	void *v;
618 	register_t *retval;
619 {
620 	struct linux_sys_mremap_args /* {
621 		syscallarg(void *) old_address;
622 		syscallarg(size_t) old_size;
623 		syscallarg(size_t) new_size;
624 		syscallarg(u_long) flags;
625 	} */ *uap = v;
626 
627 	struct proc *p;
628 	struct vm_map *map;
629 	vaddr_t oldva;
630 	vaddr_t newva;
631 	size_t oldsize;
632 	size_t newsize;
633 	int flags;
634 	int uvmflags;
635 	int error;
636 
637 	flags = SCARG(uap, flags);
638 	oldva = (vaddr_t)SCARG(uap, old_address);
639 	oldsize = round_page(SCARG(uap, old_size));
640 	newsize = round_page(SCARG(uap, new_size));
641 	if ((flags & ~(LINUX_MREMAP_FIXED|LINUX_MREMAP_MAYMOVE)) != 0) {
642 		error = EINVAL;
643 		goto done;
644 	}
645 	if ((flags & LINUX_MREMAP_FIXED) != 0) {
646 		if ((flags & LINUX_MREMAP_MAYMOVE) == 0) {
647 			error = EINVAL;
648 			goto done;
649 		}
650 #if 0 /* notyet */
651 		newva = SCARG(uap, new_address);
652 		uvmflags = UVM_MREMAP_FIXED;
653 #else /* notyet */
654 		error = EOPNOTSUPP;
655 		goto done;
656 #endif /* notyet */
657 	} else if ((flags & LINUX_MREMAP_MAYMOVE) != 0) {
658 		uvmflags = 0;
659 	} else {
660 		newva = oldva;
661 		uvmflags = UVM_MREMAP_FIXED;
662 	}
663 	p = l->l_proc;
664 	map = &p->p_vmspace->vm_map;
665 	error = uvm_mremap(map, oldva, oldsize, map, &newva, newsize, p,
666 	    uvmflags);
667 
668 done:
669 	*retval = (error != 0) ? 0 : (register_t)newva;
670 	return error;
671 }
672 
673 int
674 linux_sys_msync(l, v, retval)
675 	struct lwp *l;
676 	void *v;
677 	register_t *retval;
678 {
679 	struct linux_sys_msync_args /* {
680 		syscallarg(caddr_t) addr;
681 		syscallarg(int) len;
682 		syscallarg(int) fl;
683 	} */ *uap = v;
684 
685 	struct sys___msync13_args bma;
686 
687 	/* flags are ignored */
688 	SCARG(&bma, addr) = SCARG(uap, addr);
689 	SCARG(&bma, len) = SCARG(uap, len);
690 	SCARG(&bma, flags) = SCARG(uap, fl);
691 
692 	return sys___msync13(l, &bma, retval);
693 }
694 
695 int
696 linux_sys_mprotect(struct lwp *l, void *v, register_t *retval __unused)
697 {
698 	struct linux_sys_mprotect_args /* {
699 		syscallarg(const void *) start;
700 		syscallarg(unsigned long) len;
701 		syscallarg(int) prot;
702 	} */ *uap = v;
703 	struct vm_map_entry *entry;
704 	struct vm_map *map;
705 	struct proc *p;
706 	vaddr_t end, start, len, stacklim;
707 	int prot, grows;
708 
709 	start = (vaddr_t)SCARG(uap, start);
710 	len = round_page(SCARG(uap, len));
711 	prot = SCARG(uap, prot);
712 	grows = prot & (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP);
713 	prot &= ~grows;
714 	end = start + len;
715 
716 	if (start & PAGE_MASK)
717 		return EINVAL;
718 	if (end < start)
719 		return EINVAL;
720 	if (end == start)
721 		return 0;
722 
723 	if (prot & ~(PROT_READ | PROT_WRITE | PROT_EXEC))
724 		return EINVAL;
725 	if (grows == (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP))
726 		return EINVAL;
727 
728 	p = l->l_proc;
729 	map = &p->p_vmspace->vm_map;
730 	vm_map_lock(map);
731 # ifdef notdef
732 	VM_MAP_RANGE_CHECK(map, start, end);
733 # endif
734 	if (!uvm_map_lookup_entry(map, start, &entry) || entry->start > start) {
735 		vm_map_unlock(map);
736 		return ENOMEM;
737 	}
738 
739 	/*
740 	 * Approximate the behaviour of PROT_GROWS{DOWN,UP}.
741 	 */
742 
743 	stacklim = (vaddr_t)p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur;
744 	if (grows & LINUX_PROT_GROWSDOWN) {
745 		if (USRSTACK - stacklim <= start && start < USRSTACK) {
746 			start = USRSTACK - stacklim;
747 		} else {
748 			start = entry->start;
749 		}
750 	} else if (grows & LINUX_PROT_GROWSUP) {
751 		if (USRSTACK <= end && end < USRSTACK + stacklim) {
752 			end = USRSTACK + stacklim;
753 		} else {
754 			end = entry->end;
755 		}
756 	}
757 	vm_map_unlock(map);
758 	return uvm_map_protect(map, start, end, prot, FALSE);
759 }
760 
761 /*
762  * This code is partly stolen from src/lib/libc/compat-43/times.c
763  */
764 
765 #define	CONVTCK(r)	(r.tv_sec * hz + r.tv_usec / (1000000 / hz))
766 
767 int
768 linux_sys_times(l, v, retval)
769 	struct lwp *l;
770 	void *v;
771 	register_t *retval;
772 {
773 	struct linux_sys_times_args /* {
774 		syscallarg(struct times *) tms;
775 	} */ *uap = v;
776 	struct proc *p = l->l_proc;
777 	struct timeval t;
778 	int error;
779 
780 	if (SCARG(uap, tms)) {
781 		struct linux_tms ltms;
782 		struct rusage ru;
783 
784 		calcru(p, &ru.ru_utime, &ru.ru_stime, NULL);
785 		ltms.ltms_utime = CONVTCK(ru.ru_utime);
786 		ltms.ltms_stime = CONVTCK(ru.ru_stime);
787 
788 		ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime);
789 		ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime);
790 
791 		if ((error = copyout(&ltms, SCARG(uap, tms), sizeof ltms)))
792 			return error;
793 	}
794 
795 	getmicrouptime(&t);
796 
797 	retval[0] = ((linux_clock_t)(CONVTCK(t)));
798 	return 0;
799 }
800 
801 #undef CONVTCK
802 
803 /*
804  * Linux 'readdir' call. This code is mostly taken from the
805  * SunOS getdents call (see compat/sunos/sunos_misc.c), though
806  * an attempt has been made to keep it a little cleaner (failing
807  * miserably, because of the cruft needed if count 1 is passed).
808  *
809  * The d_off field should contain the offset of the next valid entry,
810  * but in Linux it has the offset of the entry itself. We emulate
811  * that bug here.
812  *
813  * Read in BSD-style entries, convert them, and copy them out.
814  *
815  * Note that this doesn't handle union-mounted filesystems.
816  */
817 int
818 linux_sys_getdents(l, v, retval)
819 	struct lwp *l;
820 	void *v;
821 	register_t *retval;
822 {
823 	struct linux_sys_getdents_args /* {
824 		syscallarg(int) fd;
825 		syscallarg(struct linux_dirent *) dent;
826 		syscallarg(unsigned int) count;
827 	} */ *uap = v;
828 	struct dirent *bdp;
829 	struct vnode *vp;
830 	caddr_t	inp, tbuf;		/* BSD-format */
831 	int len, reclen;		/* BSD-format */
832 	caddr_t outp;			/* Linux-format */
833 	int resid, linux_reclen = 0;	/* Linux-format */
834 	struct file *fp;
835 	struct uio auio;
836 	struct iovec aiov;
837 	struct linux_dirent idb;
838 	off_t off;		/* true file offset */
839 	int buflen, error, eofflag, nbytes, oldcall;
840 	struct vattr va;
841 	off_t *cookiebuf = NULL, *cookie;
842 	int ncookies;
843 
844 	/* getvnode() will use the descriptor for us */
845 	if ((error = getvnode(l->l_proc->p_fd, SCARG(uap, fd), &fp)) != 0)
846 		return (error);
847 
848 	if ((fp->f_flag & FREAD) == 0) {
849 		error = EBADF;
850 		goto out1;
851 	}
852 
853 	vp = (struct vnode *)fp->f_data;
854 	if (vp->v_type != VDIR) {
855 		error = EINVAL;
856 		goto out1;
857 	}
858 
859 	if ((error = VOP_GETATTR(vp, &va, l->l_cred, l)))
860 		goto out1;
861 
862 	nbytes = SCARG(uap, count);
863 	if (nbytes == 1) {	/* emulating old, broken behaviour */
864 		nbytes = sizeof (idb);
865 		buflen = max(va.va_blocksize, nbytes);
866 		oldcall = 1;
867 	} else {
868 		buflen = min(MAXBSIZE, nbytes);
869 		if (buflen < va.va_blocksize)
870 			buflen = va.va_blocksize;
871 		oldcall = 0;
872 	}
873 	tbuf = malloc(buflen, M_TEMP, M_WAITOK);
874 
875 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
876 	off = fp->f_offset;
877 again:
878 	aiov.iov_base = tbuf;
879 	aiov.iov_len = buflen;
880 	auio.uio_iov = &aiov;
881 	auio.uio_iovcnt = 1;
882 	auio.uio_rw = UIO_READ;
883 	auio.uio_resid = buflen;
884 	auio.uio_offset = off;
885 	UIO_SETUP_SYSSPACE(&auio);
886 	/*
887          * First we read into the malloc'ed buffer, then
888          * we massage it into user space, one record at a time.
889          */
890 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf,
891 	    &ncookies);
892 	if (error)
893 		goto out;
894 
895 	inp = tbuf;
896 	outp = (caddr_t)SCARG(uap, dent);
897 	resid = nbytes;
898 	if ((len = buflen - auio.uio_resid) == 0)
899 		goto eof;
900 
901 	for (cookie = cookiebuf; len > 0; len -= reclen) {
902 		bdp = (struct dirent *)inp;
903 		reclen = bdp->d_reclen;
904 		if (reclen & 3)
905 			panic("linux_readdir");
906 		if (bdp->d_fileno == 0) {
907 			inp += reclen;	/* it is a hole; squish it out */
908 			if (cookie)
909 				off = *cookie++;
910 			else
911 				off += reclen;
912 			continue;
913 		}
914 		linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen);
915 		if (reclen > len || resid < linux_reclen) {
916 			/* entry too big for buffer, so just stop */
917 			outp++;
918 			break;
919 		}
920 		/*
921 		 * Massage in place to make a Linux-shaped dirent (otherwise
922 		 * we have to worry about touching user memory outside of
923 		 * the copyout() call).
924 		 */
925 		idb.d_ino = bdp->d_fileno;
926 		/*
927 		 * The old readdir() call misuses the offset and reclen fields.
928 		 */
929 		if (oldcall) {
930 			idb.d_off = (linux_off_t)linux_reclen;
931 			idb.d_reclen = (u_short)bdp->d_namlen;
932 		} else {
933 			if (sizeof (idb.d_off) <= 4 && (off >> 32) != 0) {
934 				compat_offseterr(vp, "linux_getdents");
935 				error = EINVAL;
936 				goto out;
937 			}
938 			idb.d_off = (linux_off_t)off;
939 			idb.d_reclen = (u_short)linux_reclen;
940 		}
941 		strcpy(idb.d_name, bdp->d_name);
942 		if ((error = copyout((caddr_t)&idb, outp, linux_reclen)))
943 			goto out;
944 		/* advance past this real entry */
945 		inp += reclen;
946 		if (cookie)
947 			off = *cookie++; /* each entry points to itself */
948 		else
949 			off += reclen;
950 		/* advance output past Linux-shaped entry */
951 		outp += linux_reclen;
952 		resid -= linux_reclen;
953 		if (oldcall)
954 			break;
955 	}
956 
957 	/* if we squished out the whole block, try again */
958 	if (outp == (caddr_t)SCARG(uap, dent))
959 		goto again;
960 	fp->f_offset = off;	/* update the vnode offset */
961 
962 	if (oldcall)
963 		nbytes = resid + linux_reclen;
964 
965 eof:
966 	*retval = nbytes - resid;
967 out:
968 	VOP_UNLOCK(vp, 0);
969 	if (cookiebuf)
970 		free(cookiebuf, M_TEMP);
971 	free(tbuf, M_TEMP);
972 out1:
973 	FILE_UNUSE(fp, l);
974 	return error;
975 }
976 
977 /*
978  * Even when just using registers to pass arguments to syscalls you can
979  * have 5 of them on the i386. So this newer version of select() does
980  * this.
981  */
982 int
983 linux_sys_select(l, v, retval)
984 	struct lwp *l;
985 	void *v;
986 	register_t *retval;
987 {
988 	struct linux_sys_select_args /* {
989 		syscallarg(int) nfds;
990 		syscallarg(fd_set *) readfds;
991 		syscallarg(fd_set *) writefds;
992 		syscallarg(fd_set *) exceptfds;
993 		syscallarg(struct timeval *) timeout;
994 	} */ *uap = v;
995 
996 	return linux_select1(l, retval, SCARG(uap, nfds), SCARG(uap, readfds),
997 	    SCARG(uap, writefds), SCARG(uap, exceptfds), SCARG(uap, timeout));
998 }
999 
1000 /*
1001  * Common code for the old and new versions of select(). A couple of
1002  * things are important:
1003  * 1) return the amount of time left in the 'timeout' parameter
1004  * 2) select never returns ERESTART on Linux, always return EINTR
1005  */
1006 int
1007 linux_select1(l, retval, nfds, readfds, writefds, exceptfds, timeout)
1008 	struct lwp *l;
1009 	register_t *retval;
1010 	int nfds;
1011 	fd_set *readfds, *writefds, *exceptfds;
1012 	struct timeval *timeout;
1013 {
1014 	struct sys_select_args bsa;
1015 	struct proc *p = l->l_proc;
1016 	struct timeval tv0, tv1, utv, *tvp;
1017 	caddr_t sg;
1018 	int error;
1019 
1020 	SCARG(&bsa, nd) = nfds;
1021 	SCARG(&bsa, in) = readfds;
1022 	SCARG(&bsa, ou) = writefds;
1023 	SCARG(&bsa, ex) = exceptfds;
1024 	SCARG(&bsa, tv) = timeout;
1025 
1026 	/*
1027 	 * Store current time for computation of the amount of
1028 	 * time left.
1029 	 */
1030 	if (timeout) {
1031 		if ((error = copyin(timeout, &utv, sizeof(utv))))
1032 			return error;
1033 		if (itimerfix(&utv)) {
1034 			/*
1035 			 * The timeval was invalid.  Convert it to something
1036 			 * valid that will act as it does under Linux.
1037 			 */
1038 			sg = stackgap_init(p, 0);
1039 			tvp = stackgap_alloc(p, &sg, sizeof(utv));
1040 			utv.tv_sec += utv.tv_usec / 1000000;
1041 			utv.tv_usec %= 1000000;
1042 			if (utv.tv_usec < 0) {
1043 				utv.tv_sec -= 1;
1044 				utv.tv_usec += 1000000;
1045 			}
1046 			if (utv.tv_sec < 0)
1047 				timerclear(&utv);
1048 			if ((error = copyout(&utv, tvp, sizeof(utv))))
1049 				return error;
1050 			SCARG(&bsa, tv) = tvp;
1051 		}
1052 		microtime(&tv0);
1053 	}
1054 
1055 	error = sys_select(l, &bsa, retval);
1056 	if (error) {
1057 		/*
1058 		 * See fs/select.c in the Linux kernel.  Without this,
1059 		 * Maelstrom doesn't work.
1060 		 */
1061 		if (error == ERESTART)
1062 			error = EINTR;
1063 		return error;
1064 	}
1065 
1066 	if (timeout) {
1067 		if (*retval) {
1068 			/*
1069 			 * Compute how much time was left of the timeout,
1070 			 * by subtracting the current time and the time
1071 			 * before we started the call, and subtracting
1072 			 * that result from the user-supplied value.
1073 			 */
1074 			microtime(&tv1);
1075 			timersub(&tv1, &tv0, &tv1);
1076 			timersub(&utv, &tv1, &utv);
1077 			if (utv.tv_sec < 0)
1078 				timerclear(&utv);
1079 		} else
1080 			timerclear(&utv);
1081 		if ((error = copyout(&utv, timeout, sizeof(utv))))
1082 			return error;
1083 	}
1084 
1085 	return 0;
1086 }
1087 
1088 /*
1089  * Get the process group of a certain process. Look it up
1090  * and return the value.
1091  */
1092 int
1093 linux_sys_getpgid(l, v, retval)
1094 	struct lwp *l;
1095 	void *v;
1096 	register_t *retval;
1097 {
1098 	struct linux_sys_getpgid_args /* {
1099 		syscallarg(int) pid;
1100 	} */ *uap = v;
1101 	struct proc *p = l->l_proc;
1102 	struct proc *targp;
1103 
1104 	if (SCARG(uap, pid) != 0 && SCARG(uap, pid) != p->p_pid) {
1105 		if ((targp = pfind(SCARG(uap, pid))) == 0)
1106 			return ESRCH;
1107 	}
1108 	else
1109 		targp = p;
1110 
1111 	retval[0] = targp->p_pgid;
1112 	return 0;
1113 }
1114 
1115 /*
1116  * Set the 'personality' (emulation mode) for the current process. Only
1117  * accept the Linux personality here (0). This call is needed because
1118  * the Linux ELF crt0 issues it in an ugly kludge to make sure that
1119  * ELF binaries run in Linux mode, not SVR4 mode.
1120  */
1121 int
1122 linux_sys_personality(struct lwp *l __unused, void *v, register_t *retval)
1123 {
1124 	struct linux_sys_personality_args /* {
1125 		syscallarg(int) per;
1126 	} */ *uap = v;
1127 
1128 	if (SCARG(uap, per) != 0)
1129 		return EINVAL;
1130 	retval[0] = 0;
1131 	return 0;
1132 }
1133 #endif /* !COMPAT_LINUX32 */
1134 
1135 #if defined(__i386__) || defined(__m68k__) || defined(COMPAT_LINUX32)
1136 /*
1137  * The calls are here because of type conversions.
1138  */
1139 int
1140 linux_sys_setreuid16(l, v, retval)
1141 	struct lwp *l;
1142 	void *v;
1143 	register_t *retval;
1144 {
1145 	struct linux_sys_setreuid16_args /* {
1146 		syscallarg(int) ruid;
1147 		syscallarg(int) euid;
1148 	} */ *uap = v;
1149 	struct sys_setreuid_args bsa;
1150 
1151 	SCARG(&bsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
1152 		(uid_t)-1 : SCARG(uap, ruid);
1153 	SCARG(&bsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
1154 		(uid_t)-1 : SCARG(uap, euid);
1155 
1156 	return sys_setreuid(l, &bsa, retval);
1157 }
1158 
1159 int
1160 linux_sys_setregid16(l, v, retval)
1161 	struct lwp *l;
1162 	void *v;
1163 	register_t *retval;
1164 {
1165 	struct linux_sys_setregid16_args /* {
1166 		syscallarg(int) rgid;
1167 		syscallarg(int) egid;
1168 	} */ *uap = v;
1169 	struct sys_setregid_args bsa;
1170 
1171 	SCARG(&bsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
1172 		(uid_t)-1 : SCARG(uap, rgid);
1173 	SCARG(&bsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
1174 		(uid_t)-1 : SCARG(uap, egid);
1175 
1176 	return sys_setregid(l, &bsa, retval);
1177 }
1178 
1179 int
1180 linux_sys_setresuid16(l, v, retval)
1181 	struct lwp *l;
1182 	void *v;
1183 	register_t *retval;
1184 {
1185 	struct linux_sys_setresuid16_args /* {
1186 		syscallarg(uid_t) ruid;
1187 		syscallarg(uid_t) euid;
1188 		syscallarg(uid_t) suid;
1189 	} */ *uap = v;
1190 	struct linux_sys_setresuid16_args lsa;
1191 
1192 	SCARG(&lsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
1193 		(uid_t)-1 : SCARG(uap, ruid);
1194 	SCARG(&lsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
1195 		(uid_t)-1 : SCARG(uap, euid);
1196 	SCARG(&lsa, suid) = ((linux_uid_t)SCARG(uap, suid) == (linux_uid_t)-1) ?
1197 		(uid_t)-1 : SCARG(uap, suid);
1198 
1199 	return linux_sys_setresuid(l, &lsa, retval);
1200 }
1201 
1202 int
1203 linux_sys_setresgid16(l, v, retval)
1204 	struct lwp *l;
1205 	void *v;
1206 	register_t *retval;
1207 {
1208 	struct linux_sys_setresgid16_args /* {
1209 		syscallarg(gid_t) rgid;
1210 		syscallarg(gid_t) egid;
1211 		syscallarg(gid_t) sgid;
1212 	} */ *uap = v;
1213 	struct linux_sys_setresgid16_args lsa;
1214 
1215 	SCARG(&lsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
1216 		(gid_t)-1 : SCARG(uap, rgid);
1217 	SCARG(&lsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
1218 		(gid_t)-1 : SCARG(uap, egid);
1219 	SCARG(&lsa, sgid) = ((linux_gid_t)SCARG(uap, sgid) == (linux_gid_t)-1) ?
1220 		(gid_t)-1 : SCARG(uap, sgid);
1221 
1222 	return linux_sys_setresgid(l, &lsa, retval);
1223 }
1224 
1225 int
1226 linux_sys_getgroups16(l, v, retval)
1227 	struct lwp *l;
1228 	void *v;
1229 	register_t *retval;
1230 {
1231 	struct linux_sys_getgroups16_args /* {
1232 		syscallarg(int) gidsetsize;
1233 		syscallarg(linux_gid_t *) gidset;
1234 	} */ *uap = v;
1235 	struct proc *p = l->l_proc;
1236 	caddr_t sg;
1237 	int n, error, i;
1238 	struct sys_getgroups_args bsa;
1239 	gid_t *bset, *kbset;
1240 	linux_gid_t *lset;
1241 	kauth_cred_t pc = l->l_cred;
1242 
1243 	n = SCARG(uap, gidsetsize);
1244 	if (n < 0)
1245 		return EINVAL;
1246 	error = 0;
1247 	bset = kbset = NULL;
1248 	lset = NULL;
1249 	if (n > 0) {
1250 		n = min(kauth_cred_ngroups(pc), n);
1251 		sg = stackgap_init(p, 0);
1252 		bset = stackgap_alloc(p, &sg, n * sizeof (gid_t));
1253 		kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK);
1254 		lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
1255 		if (bset == NULL || kbset == NULL || lset == NULL)
1256 		{
1257 			error = ENOMEM;
1258 			goto out;
1259 		}
1260 		SCARG(&bsa, gidsetsize) = n;
1261 		SCARG(&bsa, gidset) = bset;
1262 		error = sys_getgroups(l, &bsa, retval);
1263 		if (error != 0)
1264 			goto out;
1265 		error = copyin(bset, kbset, n * sizeof (gid_t));
1266 		if (error != 0)
1267 			goto out;
1268 		for (i = 0; i < n; i++)
1269 			lset[i] = (linux_gid_t)kbset[i];
1270 		error = copyout(lset, SCARG(uap, gidset),
1271 		    n * sizeof (linux_gid_t));
1272 	} else
1273 		*retval = kauth_cred_ngroups(pc);
1274 out:
1275 	if (kbset != NULL)
1276 		free(kbset, M_TEMP);
1277 	if (lset != NULL)
1278 		free(lset, M_TEMP);
1279 	return error;
1280 }
1281 
1282 int
1283 linux_sys_setgroups16(l, v, retval)
1284 	struct lwp *l;
1285 	void *v;
1286 	register_t *retval;
1287 {
1288 	struct linux_sys_setgroups16_args /* {
1289 		syscallarg(int) gidsetsize;
1290 		syscallarg(linux_gid_t *) gidset;
1291 	} */ *uap = v;
1292 	struct proc *p = l->l_proc;
1293 	caddr_t sg;
1294 	int n;
1295 	int error, i;
1296 	struct sys_setgroups_args bsa;
1297 	gid_t *bset, *kbset;
1298 	linux_gid_t *lset;
1299 
1300 	n = SCARG(uap, gidsetsize);
1301 	if (n < 0 || n > NGROUPS)
1302 		return EINVAL;
1303 	sg = stackgap_init(p, 0);
1304 	bset = stackgap_alloc(p, &sg, n * sizeof (gid_t));
1305 	lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
1306 	kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK);
1307 	if (bset == NULL || kbset == NULL || lset == NULL)
1308 	{
1309 		error = ENOMEM;
1310 		goto out;
1311 	}
1312 	error = copyin(SCARG(uap, gidset), lset, n * sizeof (linux_gid_t));
1313 	if (error != 0)
1314 		goto out;
1315 	for (i = 0; i < n; i++)
1316 		kbset[i] = (gid_t)lset[i];
1317 	error = copyout(kbset, bset, n * sizeof (gid_t));
1318 	if (error != 0)
1319 		goto out;
1320 	SCARG(&bsa, gidsetsize) = n;
1321 	SCARG(&bsa, gidset) = bset;
1322 	error = sys_setgroups(l, &bsa, retval);
1323 
1324 out:
1325 	if (lset != NULL)
1326 		free(lset, M_TEMP);
1327 	if (kbset != NULL)
1328 		free(kbset, M_TEMP);
1329 
1330 	return error;
1331 }
1332 
1333 #endif /* __i386__ || __m68k__ || COMPAT_LINUX32 */
1334 
1335 #ifndef COMPAT_LINUX32
1336 /*
1337  * We have nonexistent fsuid equal to uid.
1338  * If modification is requested, refuse.
1339  */
1340 int
1341 linux_sys_setfsuid(l, v, retval)
1342 	 struct lwp *l;
1343 	 void *v;
1344 	 register_t *retval;
1345 {
1346 	 struct linux_sys_setfsuid_args /* {
1347 		 syscallarg(uid_t) uid;
1348 	 } */ *uap = v;
1349 	 uid_t uid;
1350 
1351 	 uid = SCARG(uap, uid);
1352 	 if (kauth_cred_getuid(l->l_cred) != uid)
1353 		 return sys_nosys(l, v, retval);
1354 	 else
1355 		 return (0);
1356 }
1357 
1358 /* XXX XXX XXX */
1359 # ifndef alpha
1360 int
1361 linux_sys_getfsuid(l, v, retval)
1362 	struct lwp *l;
1363 	void *v;
1364 	register_t *retval;
1365 {
1366 	return sys_getuid(l, v, retval);
1367 }
1368 # endif
1369 
1370 int
1371 linux_sys_setresuid(struct lwp *l, void *v, register_t *retval __unused)
1372 {
1373 	struct linux_sys_setresuid_args /* {
1374 		syscallarg(uid_t) ruid;
1375 		syscallarg(uid_t) euid;
1376 		syscallarg(uid_t) suid;
1377 	} */ *uap = v;
1378 
1379 	/*
1380 	 * Note: These checks are a little different than the NetBSD
1381 	 * setreuid(2) call performs.  This precisely follows the
1382 	 * behavior of the Linux kernel.
1383 	 */
1384 
1385 	return do_setresuid(l, SCARG(uap, ruid), SCARG(uap, euid),
1386 			    SCARG(uap, suid),
1387 			    ID_R_EQ_R | ID_R_EQ_E | ID_R_EQ_S |
1388 			    ID_E_EQ_R | ID_E_EQ_E | ID_E_EQ_S |
1389 			    ID_S_EQ_R | ID_S_EQ_E | ID_S_EQ_S );
1390 }
1391 
1392 int
1393 linux_sys_getresuid(struct lwp *l, void *v, register_t *retval __unused)
1394 {
1395 	struct linux_sys_getresuid_args /* {
1396 		syscallarg(uid_t *) ruid;
1397 		syscallarg(uid_t *) euid;
1398 		syscallarg(uid_t *) suid;
1399 	} */ *uap = v;
1400 	kauth_cred_t pc = l->l_cred;
1401 	int error;
1402 	uid_t uid;
1403 
1404 	/*
1405 	 * Linux copies these values out to userspace like so:
1406 	 *
1407 	 *	1. Copy out ruid.
1408 	 *	2. If that succeeds, copy out euid.
1409 	 *	3. If both of those succeed, copy out suid.
1410 	 */
1411 	uid = kauth_cred_getuid(pc);
1412 	if ((error = copyout(&uid, SCARG(uap, ruid), sizeof(uid_t))) != 0)
1413 		return (error);
1414 
1415 	uid = kauth_cred_geteuid(pc);
1416 	if ((error = copyout(&uid, SCARG(uap, euid), sizeof(uid_t))) != 0)
1417 		return (error);
1418 
1419 	uid = kauth_cred_getsvuid(pc);
1420 
1421 	return (copyout(&uid, SCARG(uap, suid), sizeof(uid_t)));
1422 }
1423 
1424 int
1425 linux_sys_ptrace(l, v, retval)
1426 	struct lwp *l;
1427 	void *v;
1428 	register_t *retval;
1429 {
1430 #if defined(PTRACE) || defined(_LKM)
1431 	struct linux_sys_ptrace_args /* {
1432 		i386, m68k, powerpc: T=int
1433 		alpha, amd64: T=long
1434 		syscallarg(T) request;
1435 		syscallarg(T) pid;
1436 		syscallarg(T) addr;
1437 		syscallarg(T) data;
1438 	} */ *uap = v;
1439 	const int *ptr;
1440 	int request;
1441 	int error;
1442 #ifdef _LKM
1443 #define sys_ptrace (*sysent[SYS_ptrace].sy_call)
1444 #endif
1445 
1446 	ptr = linux_ptrace_request_map;
1447 	request = SCARG(uap, request);
1448 	while (*ptr != -1)
1449 		if (*ptr++ == request) {
1450 			struct sys_ptrace_args pta;
1451 
1452 			SCARG(&pta, req) = *ptr;
1453 			SCARG(&pta, pid) = SCARG(uap, pid);
1454 			SCARG(&pta, addr) = (caddr_t)SCARG(uap, addr);
1455 			SCARG(&pta, data) = SCARG(uap, data);
1456 
1457 			/*
1458 			 * Linux ptrace(PTRACE_CONT, pid, 0, 0) means actually
1459 			 * to continue where the process left off previously.
1460 			 * The same thing is achieved by addr == (caddr_t) 1
1461 			 * on NetBSD, so rewrite 'addr' appropriately.
1462 			 */
1463 			if (request == LINUX_PTRACE_CONT && SCARG(uap, addr)==0)
1464 				SCARG(&pta, addr) = (caddr_t) 1;
1465 
1466 			error = sys_ptrace(l, &pta, retval);
1467 			if (error)
1468 				return error;
1469 			switch (request) {
1470 			case LINUX_PTRACE_PEEKTEXT:
1471 			case LINUX_PTRACE_PEEKDATA:
1472 				error = copyout (retval,
1473 				    (caddr_t)SCARG(uap, data),
1474 				    sizeof *retval);
1475 				*retval = SCARG(uap, data);
1476 				break;
1477 			default:
1478 				break;
1479 			}
1480 			return error;
1481 		}
1482 		else
1483 			ptr++;
1484 
1485 	return LINUX_SYS_PTRACE_ARCH(l, uap, retval);
1486 #else
1487 	return ENOSYS;
1488 #endif /* PTRACE || _LKM */
1489 }
1490 
1491 int
1492 linux_sys_reboot(struct lwp *l, void *v, register_t *retval)
1493 {
1494 	struct linux_sys_reboot_args /* {
1495 		syscallarg(int) magic1;
1496 		syscallarg(int) magic2;
1497 		syscallarg(int) cmd;
1498 		syscallarg(void *) arg;
1499 	} */ *uap = v;
1500 	struct sys_reboot_args /* {
1501 		syscallarg(int) opt;
1502 		syscallarg(char *) bootstr;
1503 	} */ sra;
1504 	int error;
1505 
1506 	if ((error = kauth_authorize_generic(l->l_cred,
1507 	    KAUTH_GENERIC_ISSUSER, &l->l_acflag)) != 0)
1508 		return(error);
1509 
1510 	if (SCARG(uap, magic1) != LINUX_REBOOT_MAGIC1)
1511 		return(EINVAL);
1512 	if (SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2 &&
1513 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2A &&
1514 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2B)
1515 		return(EINVAL);
1516 
1517 	switch (SCARG(uap, cmd)) {
1518 	case LINUX_REBOOT_CMD_RESTART:
1519 		SCARG(&sra, opt) = RB_AUTOBOOT;
1520 		break;
1521 	case LINUX_REBOOT_CMD_HALT:
1522 		SCARG(&sra, opt) = RB_HALT;
1523 		break;
1524 	case LINUX_REBOOT_CMD_POWER_OFF:
1525 		SCARG(&sra, opt) = RB_HALT|RB_POWERDOWN;
1526 		break;
1527 	case LINUX_REBOOT_CMD_RESTART2:
1528 		/* Reboot with an argument. */
1529 		SCARG(&sra, opt) = RB_AUTOBOOT|RB_STRING;
1530 		SCARG(&sra, bootstr) = SCARG(uap, arg);
1531 		break;
1532 	case LINUX_REBOOT_CMD_CAD_ON:
1533 		return(EINVAL);	/* We don't implement ctrl-alt-delete */
1534 	case LINUX_REBOOT_CMD_CAD_OFF:
1535 		return(0);
1536 	default:
1537 		return(EINVAL);
1538 	}
1539 
1540 	return(sys_reboot(l, &sra, retval));
1541 }
1542 
1543 /*
1544  * Copy of compat_12_sys_swapon().
1545  */
1546 int
1547 linux_sys_swapon(l, v, retval)
1548 	struct lwp *l;
1549 	void *v;
1550 	register_t *retval;
1551 {
1552 	struct sys_swapctl_args ua;
1553 	struct linux_sys_swapon_args /* {
1554 		syscallarg(const char *) name;
1555 	} */ *uap = v;
1556 
1557 	SCARG(&ua, cmd) = SWAP_ON;
1558 	SCARG(&ua, arg) = (void *)__UNCONST(SCARG(uap, name));
1559 	SCARG(&ua, misc) = 0;	/* priority */
1560 	return (sys_swapctl(l, &ua, retval));
1561 }
1562 
1563 /*
1564  * Stop swapping to the file or block device specified by path.
1565  */
1566 int
1567 linux_sys_swapoff(l, v, retval)
1568 	struct lwp *l;
1569 	void *v;
1570 	register_t *retval;
1571 {
1572 	struct sys_swapctl_args ua;
1573 	struct linux_sys_swapoff_args /* {
1574 		syscallarg(const char *) path;
1575 	} */ *uap = v;
1576 
1577 	SCARG(&ua, cmd) = SWAP_OFF;
1578 	SCARG(&ua, arg) = __UNCONST(SCARG(uap, path)); /*XXXUNCONST*/
1579 	return (sys_swapctl(l, &ua, retval));
1580 }
1581 
1582 /*
1583  * Copy of compat_09_sys_setdomainname()
1584  */
1585 /* ARGSUSED */
1586 int
1587 linux_sys_setdomainname(struct lwp *l, void *v, register_t *retval __unused)
1588 {
1589 	struct linux_sys_setdomainname_args /* {
1590 		syscallarg(char *) domainname;
1591 		syscallarg(int) len;
1592 	} */ *uap = v;
1593 	int name[2];
1594 
1595 	name[0] = CTL_KERN;
1596 	name[1] = KERN_DOMAINNAME;
1597 	return (old_sysctl(&name[0], 2, 0, 0, SCARG(uap, domainname),
1598 			    SCARG(uap, len), l));
1599 }
1600 
1601 /*
1602  * sysinfo()
1603  */
1604 /* ARGSUSED */
1605 int
1606 linux_sys_sysinfo(struct lwp *l __unused, void *v, register_t *retval __unused)
1607 {
1608 	struct linux_sys_sysinfo_args /* {
1609 		syscallarg(struct linux_sysinfo *) arg;
1610 	} */ *uap = v;
1611 	struct linux_sysinfo si;
1612 	struct loadavg *la;
1613 
1614 	si.uptime = time_uptime;
1615 	la = &averunnable;
1616 	si.loads[0] = la->ldavg[0] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1617 	si.loads[1] = la->ldavg[1] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1618 	si.loads[2] = la->ldavg[2] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1619 	si.totalram = ctob((u_long)physmem);
1620 	si.freeram = (u_long)uvmexp.free * uvmexp.pagesize;
1621 	si.sharedram = 0;	/* XXX */
1622 	si.bufferram = (u_long)uvmexp.filepages * uvmexp.pagesize;
1623 	si.totalswap = (u_long)uvmexp.swpages * uvmexp.pagesize;
1624 	si.freeswap =
1625 	    (u_long)(uvmexp.swpages - uvmexp.swpginuse) * uvmexp.pagesize;
1626 	si.procs = nprocs;
1627 
1628 	/* The following are only present in newer Linux kernels. */
1629 	si.totalbig = 0;
1630 	si.freebig = 0;
1631 	si.mem_unit = 1;
1632 
1633 	return (copyout(&si, SCARG(uap, arg), sizeof si));
1634 }
1635 
1636 int
1637 linux_sys_getrlimit(l, v, retval)
1638 	struct lwp *l;
1639 	void *v;
1640 	register_t *retval;
1641 {
1642 	struct linux_sys_getrlimit_args /* {
1643 		syscallarg(int) which;
1644 # ifdef LINUX_LARGEFILE64
1645 		syscallarg(struct rlimit *) rlp;
1646 # else
1647 		syscallarg(struct orlimit *) rlp;
1648 # endif
1649 	} */ *uap = v;
1650 	struct proc *p = l->l_proc;
1651 	caddr_t sg = stackgap_init(p, 0);
1652 	struct sys_getrlimit_args ap;
1653 	struct rlimit rl;
1654 # ifdef LINUX_LARGEFILE64
1655 	struct rlimit orl;
1656 # else
1657 	struct orlimit orl;
1658 # endif
1659 	int error;
1660 
1661 	SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which));
1662 	if ((error = SCARG(&ap, which)) < 0)
1663 		return -error;
1664 	SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl);
1665 	if ((error = sys_getrlimit(l, &ap, retval)) != 0)
1666 		return error;
1667 	if ((error = copyin(SCARG(&ap, rlp), &rl, sizeof(rl))) != 0)
1668 		return error;
1669 	bsd_to_linux_rlimit(&orl, &rl);
1670 
1671 	return copyout(&orl, SCARG(uap, rlp), sizeof(orl));
1672 }
1673 
1674 int
1675 linux_sys_setrlimit(l, v, retval)
1676 	struct lwp *l;
1677 	void *v;
1678 	register_t *retval;
1679 {
1680 	struct linux_sys_setrlimit_args /* {
1681 		syscallarg(int) which;
1682 # ifdef LINUX_LARGEFILE64
1683 		syscallarg(struct rlimit *) rlp;
1684 # else
1685 		syscallarg(struct orlimit *) rlp;
1686 # endif
1687 	} */ *uap = v;
1688 	struct proc *p = l->l_proc;
1689 	caddr_t sg = stackgap_init(p, 0);
1690 	struct sys_getrlimit_args ap;
1691 	struct rlimit rl;
1692 # ifdef LINUX_LARGEFILE64
1693 	struct rlimit orl;
1694 # else
1695 	struct orlimit orl;
1696 # endif
1697 	int error;
1698 
1699 	SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which));
1700 	SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl);
1701 	if ((error = SCARG(&ap, which)) < 0)
1702 		return -error;
1703 	if ((error = copyin(SCARG(uap, rlp), &orl, sizeof(orl))) != 0)
1704 		return error;
1705 	linux_to_bsd_rlimit(&rl, &orl);
1706 	if ((error = copyout(&rl, SCARG(&ap, rlp), sizeof(rl))) != 0)
1707 		return error;
1708 	return sys_setrlimit(l, &ap, retval);
1709 }
1710 
1711 # if !defined(__mips__) && !defined(__amd64__)
1712 /* XXX: this doesn't look 100% common, at least mips doesn't have it */
1713 int
1714 linux_sys_ugetrlimit(l, v, retval)
1715 	struct lwp *l;
1716 	void *v;
1717 	register_t *retval;
1718 {
1719 	return linux_sys_getrlimit(l, v, retval);
1720 }
1721 # endif
1722 
1723 /*
1724  * This gets called for unsupported syscalls. The difference to sys_nosys()
1725  * is that process does not get SIGSYS, the call just returns with ENOSYS.
1726  * This is the way Linux does it and glibc depends on this behaviour.
1727  */
1728 int
1729 linux_sys_nosys(struct lwp *l __unused, void *v __unused,
1730     register_t *retval __unused)
1731 {
1732 	return (ENOSYS);
1733 }
1734 
1735 int
1736 linux_sys_getpriority(l, v, retval)
1737         struct lwp *l;
1738         void *v;
1739         register_t *retval;
1740 {
1741         struct linux_sys_getpriority_args /* {
1742                 syscallarg(int) which;
1743                 syscallarg(int) who;
1744         } */ *uap = v;
1745         struct sys_getpriority_args bsa;
1746         int error;
1747 
1748         SCARG(&bsa, which) = SCARG(uap, which);
1749         SCARG(&bsa, who) = SCARG(uap, who);
1750 
1751         if ((error = sys_getpriority(l, &bsa, retval)))
1752                 return error;
1753 
1754         *retval = NZERO - *retval;
1755 
1756         return 0;
1757 }
1758 
1759 #endif /* !COMPAT_LINUX32 */
1760