xref: /netbsd-src/sys/compat/linux/common/linux_misc.c (revision b7ae68fde0d8ef1c03714e8bbb1ee7c6118ea93b)
1 /*	$NetBSD: linux_misc.c,v 1.162 2006/09/13 19:55:49 manu 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 1.162 2006/09/13 19:55:49 manu 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(l, v, retval)
460 	struct lwp *l;
461 	void *v;
462 	register_t *retval;
463 {
464 	struct linux_sys_uname_args /* {
465 		syscallarg(struct linux_utsname *) up;
466 	} */ *uap = v;
467 	struct linux_utsname luts;
468 
469 	strncpy(luts.l_sysname, linux_sysname, sizeof(luts.l_sysname));
470 	strncpy(luts.l_nodename, hostname, sizeof(luts.l_nodename));
471 	strncpy(luts.l_release, linux_release, sizeof(luts.l_release));
472 	strncpy(luts.l_version, linux_version, sizeof(luts.l_version));
473 # ifdef LINUX_UNAME_ARCH
474 	strncpy(luts.l_machine, LINUX_UNAME_ARCH, sizeof(luts.l_machine));
475 # else
476 	strncpy(luts.l_machine, machine, sizeof(luts.l_machine));
477 # endif
478 	strncpy(luts.l_domainname, domainname, sizeof(luts.l_domainname));
479 
480 	return copyout(&luts, SCARG(uap, up), sizeof(luts));
481 }
482 
483 /* Used directly on: alpha, mips, ppc, sparc, sparc64 */
484 /* Used indirectly on: arm, i386, m68k */
485 
486 /*
487  * New type Linux mmap call.
488  * Only called directly on machines with >= 6 free regs.
489  */
490 int
491 linux_sys_mmap(l, v, retval)
492 	struct lwp *l;
493 	void *v;
494 	register_t *retval;
495 {
496 	struct linux_sys_mmap_args /* {
497 		syscallarg(unsigned long) addr;
498 		syscallarg(size_t) len;
499 		syscallarg(int) prot;
500 		syscallarg(int) flags;
501 		syscallarg(int) fd;
502 		syscallarg(linux_off_t) offset;
503 	} */ *uap = v;
504 
505 	if (SCARG(uap, offset) & PAGE_MASK)
506 		return EINVAL;
507 
508 	return linux_mmap(l, uap, retval, SCARG(uap, offset));
509 }
510 
511 /*
512  * Guts of most architectures' mmap64() implementations.  This shares
513  * its list of arguments with linux_sys_mmap().
514  *
515  * The difference in linux_sys_mmap2() is that "offset" is actually
516  * (offset / pagesize), not an absolute byte count.  This translation
517  * to pagesize offsets is done inside glibc between the mmap64() call
518  * point, and the actual syscall.
519  */
520 int
521 linux_sys_mmap2(l, v, retval)
522 	struct lwp *l;
523 	void *v;
524 	register_t *retval;
525 {
526 	struct linux_sys_mmap2_args /* {
527 		syscallarg(unsigned long) addr;
528 		syscallarg(size_t) len;
529 		syscallarg(int) prot;
530 		syscallarg(int) flags;
531 		syscallarg(int) fd;
532 		syscallarg(linux_off_t) offset;
533 	} */ *uap = v;
534 
535 	return linux_mmap(l, uap, retval,
536 	    ((off_t)SCARG(uap, offset)) << PAGE_SHIFT);
537 }
538 
539 /*
540  * Massage arguments and call system mmap(2).
541  */
542 static int
543 linux_mmap(l, uap, retval, offset)
544 	struct lwp *l;
545 	struct linux_sys_mmap_args *uap;
546 	register_t *retval;
547 	off_t offset;
548 {
549 	struct sys_mmap_args cma;
550 	int error;
551 	size_t mmoff=0;
552 
553 	if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN) {
554 		/*
555 		 * Request for stack-like memory segment. On linux, this
556 		 * works by mmap()ping (small) segment, which is automatically
557 		 * extended when page fault happens below the currently
558 		 * allocated area. We emulate this by allocating (typically
559 		 * bigger) segment sized at current stack size limit, and
560 		 * offsetting the requested and returned address accordingly.
561 		 * Since physical pages are only allocated on-demand, this
562 		 * is effectively identical.
563 		 */
564 		rlim_t ssl = l->l_proc->p_rlimit[RLIMIT_STACK].rlim_cur;
565 
566 		if (SCARG(uap, len) < ssl) {
567 			/* Compute the address offset */
568 			mmoff = round_page(ssl) - SCARG(uap, len);
569 
570 			if (SCARG(uap, addr))
571 				SCARG(uap, addr) -= mmoff;
572 
573 			SCARG(uap, len) = (size_t) ssl;
574 		}
575 	}
576 
577 	linux_to_bsd_mmap_args(&cma, uap);
578 	SCARG(&cma, pos) = offset;
579 
580 	error = sys_mmap(l, &cma, retval);
581 	if (error)
582 		return (error);
583 
584 	/* Shift the returned address for stack-like segment if necessary */
585 	if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN && mmoff)
586 		retval[0] += mmoff;
587 
588 	return (0);
589 }
590 
591 static void
592 linux_to_bsd_mmap_args(cma, uap)
593 	struct sys_mmap_args *cma;
594 	const struct linux_sys_mmap_args *uap;
595 {
596 	int flags = MAP_TRYFIXED, fl = SCARG(uap, flags);
597 
598 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_SHARED, MAP_SHARED);
599 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_PRIVATE, MAP_PRIVATE);
600 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_FIXED, MAP_FIXED);
601 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_ANON, MAP_ANON);
602 	/* XXX XAX ERH: Any other flags here?  There are more defined... */
603 
604 	SCARG(cma, addr) = (void *)SCARG(uap, addr);
605 	SCARG(cma, len) = SCARG(uap, len);
606 	SCARG(cma, prot) = SCARG(uap, prot);
607 	if (SCARG(cma, prot) & VM_PROT_WRITE) /* XXX */
608 		SCARG(cma, prot) |= VM_PROT_READ;
609 	SCARG(cma, flags) = flags;
610 	SCARG(cma, fd) = flags & MAP_ANON ? -1 : SCARG(uap, fd);
611 	SCARG(cma, pad) = 0;
612 }
613 
614 #define	LINUX_MREMAP_MAYMOVE	1
615 #define	LINUX_MREMAP_FIXED	2
616 
617 int
618 linux_sys_mremap(l, v, retval)
619 	struct lwp *l;
620 	void *v;
621 	register_t *retval;
622 {
623 	struct linux_sys_mremap_args /* {
624 		syscallarg(void *) old_address;
625 		syscallarg(size_t) old_size;
626 		syscallarg(size_t) new_size;
627 		syscallarg(u_long) flags;
628 	} */ *uap = v;
629 
630 	struct proc *p;
631 	struct vm_map *map;
632 	vaddr_t oldva;
633 	vaddr_t newva;
634 	size_t oldsize;
635 	size_t newsize;
636 	int flags;
637 	int uvmflags;
638 	int error;
639 
640 	flags = SCARG(uap, flags);
641 	oldva = (vaddr_t)SCARG(uap, old_address);
642 	oldsize = round_page(SCARG(uap, old_size));
643 	newsize = round_page(SCARG(uap, new_size));
644 	if ((flags & ~(LINUX_MREMAP_FIXED|LINUX_MREMAP_MAYMOVE)) != 0) {
645 		error = EINVAL;
646 		goto done;
647 	}
648 	if ((flags & LINUX_MREMAP_FIXED) != 0) {
649 		if ((flags & LINUX_MREMAP_MAYMOVE) == 0) {
650 			error = EINVAL;
651 			goto done;
652 		}
653 #if 0 /* notyet */
654 		newva = SCARG(uap, new_address);
655 		uvmflags = UVM_MREMAP_FIXED;
656 #else /* notyet */
657 		error = EOPNOTSUPP;
658 		goto done;
659 #endif /* notyet */
660 	} else if ((flags & LINUX_MREMAP_MAYMOVE) != 0) {
661 		uvmflags = 0;
662 	} else {
663 		newva = oldva;
664 		uvmflags = UVM_MREMAP_FIXED;
665 	}
666 	p = l->l_proc;
667 	map = &p->p_vmspace->vm_map;
668 	error = uvm_mremap(map, oldva, oldsize, map, &newva, newsize, p,
669 	    uvmflags);
670 
671 done:
672 	*retval = (error != 0) ? 0 : (register_t)newva;
673 	return error;
674 }
675 
676 int
677 linux_sys_msync(l, v, retval)
678 	struct lwp *l;
679 	void *v;
680 	register_t *retval;
681 {
682 	struct linux_sys_msync_args /* {
683 		syscallarg(caddr_t) addr;
684 		syscallarg(int) len;
685 		syscallarg(int) fl;
686 	} */ *uap = v;
687 
688 	struct sys___msync13_args bma;
689 
690 	/* flags are ignored */
691 	SCARG(&bma, addr) = SCARG(uap, addr);
692 	SCARG(&bma, len) = SCARG(uap, len);
693 	SCARG(&bma, flags) = SCARG(uap, fl);
694 
695 	return sys___msync13(l, &bma, retval);
696 }
697 
698 int
699 linux_sys_mprotect(l, v, retval)
700 	struct lwp *l;
701 	void *v;
702 	register_t *retval;
703 {
704 	struct linux_sys_mprotect_args /* {
705 		syscallarg(const void *) start;
706 		syscallarg(unsigned long) len;
707 		syscallarg(int) prot;
708 	} */ *uap = v;
709 	struct vm_map_entry *entry;
710 	struct vm_map *map;
711 	struct proc *p;
712 	vaddr_t end, start, len, stacklim;
713 	int prot, grows;
714 
715 	start = (vaddr_t)SCARG(uap, start);
716 	len = round_page(SCARG(uap, len));
717 	prot = SCARG(uap, prot);
718 	grows = prot & (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP);
719 	prot &= ~grows;
720 	end = start + len;
721 
722 	if (start & PAGE_MASK)
723 		return EINVAL;
724 	if (end < start)
725 		return EINVAL;
726 	if (end == start)
727 		return 0;
728 
729 	if (prot & ~(PROT_READ | PROT_WRITE | PROT_EXEC))
730 		return EINVAL;
731 	if (grows == (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP))
732 		return EINVAL;
733 
734 	p = l->l_proc;
735 	map = &p->p_vmspace->vm_map;
736 	vm_map_lock(map);
737 # ifdef notdef
738 	VM_MAP_RANGE_CHECK(map, start, end);
739 # endif
740 	if (!uvm_map_lookup_entry(map, start, &entry) || entry->start > start) {
741 		vm_map_unlock(map);
742 		return ENOMEM;
743 	}
744 
745 	/*
746 	 * Approximate the behaviour of PROT_GROWS{DOWN,UP}.
747 	 */
748 
749 	stacklim = (vaddr_t)p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur;
750 	if (grows & LINUX_PROT_GROWSDOWN) {
751 		if (USRSTACK - stacklim <= start && start < USRSTACK) {
752 			start = USRSTACK - stacklim;
753 		} else {
754 			start = entry->start;
755 		}
756 	} else if (grows & LINUX_PROT_GROWSUP) {
757 		if (USRSTACK <= end && end < USRSTACK + stacklim) {
758 			end = USRSTACK + stacklim;
759 		} else {
760 			end = entry->end;
761 		}
762 	}
763 	vm_map_unlock(map);
764 	return uvm_map_protect(map, start, end, prot, FALSE);
765 }
766 
767 /*
768  * This code is partly stolen from src/lib/libc/compat-43/times.c
769  */
770 
771 #define	CONVTCK(r)	(r.tv_sec * hz + r.tv_usec / (1000000 / hz))
772 
773 int
774 linux_sys_times(l, v, retval)
775 	struct lwp *l;
776 	void *v;
777 	register_t *retval;
778 {
779 	struct linux_sys_times_args /* {
780 		syscallarg(struct times *) tms;
781 	} */ *uap = v;
782 	struct proc *p = l->l_proc;
783 	struct timeval t;
784 	int error;
785 
786 	if (SCARG(uap, tms)) {
787 		struct linux_tms ltms;
788 		struct rusage ru;
789 
790 		calcru(p, &ru.ru_utime, &ru.ru_stime, NULL);
791 		ltms.ltms_utime = CONVTCK(ru.ru_utime);
792 		ltms.ltms_stime = CONVTCK(ru.ru_stime);
793 
794 		ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime);
795 		ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime);
796 
797 		if ((error = copyout(&ltms, SCARG(uap, tms), sizeof ltms)))
798 			return error;
799 	}
800 
801 	getmicrouptime(&t);
802 
803 	retval[0] = ((linux_clock_t)(CONVTCK(t)));
804 	return 0;
805 }
806 
807 #undef CONVTCK
808 
809 /*
810  * Linux 'readdir' call. This code is mostly taken from the
811  * SunOS getdents call (see compat/sunos/sunos_misc.c), though
812  * an attempt has been made to keep it a little cleaner (failing
813  * miserably, because of the cruft needed if count 1 is passed).
814  *
815  * The d_off field should contain the offset of the next valid entry,
816  * but in Linux it has the offset of the entry itself. We emulate
817  * that bug here.
818  *
819  * Read in BSD-style entries, convert them, and copy them out.
820  *
821  * Note that this doesn't handle union-mounted filesystems.
822  */
823 int
824 linux_sys_getdents(l, v, retval)
825 	struct lwp *l;
826 	void *v;
827 	register_t *retval;
828 {
829 	struct linux_sys_getdents_args /* {
830 		syscallarg(int) fd;
831 		syscallarg(struct linux_dirent *) dent;
832 		syscallarg(unsigned int) count;
833 	} */ *uap = v;
834 	struct dirent *bdp;
835 	struct vnode *vp;
836 	caddr_t	inp, tbuf;		/* BSD-format */
837 	int len, reclen;		/* BSD-format */
838 	caddr_t outp;			/* Linux-format */
839 	int resid, linux_reclen = 0;	/* Linux-format */
840 	struct file *fp;
841 	struct uio auio;
842 	struct iovec aiov;
843 	struct linux_dirent idb;
844 	off_t off;		/* true file offset */
845 	int buflen, error, eofflag, nbytes, oldcall;
846 	struct vattr va;
847 	off_t *cookiebuf = NULL, *cookie;
848 	int ncookies;
849 
850 	/* getvnode() will use the descriptor for us */
851 	if ((error = getvnode(l->l_proc->p_fd, SCARG(uap, fd), &fp)) != 0)
852 		return (error);
853 
854 	if ((fp->f_flag & FREAD) == 0) {
855 		error = EBADF;
856 		goto out1;
857 	}
858 
859 	vp = (struct vnode *)fp->f_data;
860 	if (vp->v_type != VDIR) {
861 		error = EINVAL;
862 		goto out1;
863 	}
864 
865 	if ((error = VOP_GETATTR(vp, &va, l->l_cred, l)))
866 		goto out1;
867 
868 	nbytes = SCARG(uap, count);
869 	if (nbytes == 1) {	/* emulating old, broken behaviour */
870 		nbytes = sizeof (idb);
871 		buflen = max(va.va_blocksize, nbytes);
872 		oldcall = 1;
873 	} else {
874 		buflen = min(MAXBSIZE, nbytes);
875 		if (buflen < va.va_blocksize)
876 			buflen = va.va_blocksize;
877 		oldcall = 0;
878 	}
879 	tbuf = malloc(buflen, M_TEMP, M_WAITOK);
880 
881 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
882 	off = fp->f_offset;
883 again:
884 	aiov.iov_base = tbuf;
885 	aiov.iov_len = buflen;
886 	auio.uio_iov = &aiov;
887 	auio.uio_iovcnt = 1;
888 	auio.uio_rw = UIO_READ;
889 	auio.uio_resid = buflen;
890 	auio.uio_offset = off;
891 	UIO_SETUP_SYSSPACE(&auio);
892 	/*
893          * First we read into the malloc'ed buffer, then
894          * we massage it into user space, one record at a time.
895          */
896 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf,
897 	    &ncookies);
898 	if (error)
899 		goto out;
900 
901 	inp = tbuf;
902 	outp = (caddr_t)SCARG(uap, dent);
903 	resid = nbytes;
904 	if ((len = buflen - auio.uio_resid) == 0)
905 		goto eof;
906 
907 	for (cookie = cookiebuf; len > 0; len -= reclen) {
908 		bdp = (struct dirent *)inp;
909 		reclen = bdp->d_reclen;
910 		if (reclen & 3)
911 			panic("linux_readdir");
912 		if (bdp->d_fileno == 0) {
913 			inp += reclen;	/* it is a hole; squish it out */
914 			if (cookie)
915 				off = *cookie++;
916 			else
917 				off += reclen;
918 			continue;
919 		}
920 		linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen);
921 		if (reclen > len || resid < linux_reclen) {
922 			/* entry too big for buffer, so just stop */
923 			outp++;
924 			break;
925 		}
926 		/*
927 		 * Massage in place to make a Linux-shaped dirent (otherwise
928 		 * we have to worry about touching user memory outside of
929 		 * the copyout() call).
930 		 */
931 		idb.d_ino = bdp->d_fileno;
932 		/*
933 		 * The old readdir() call misuses the offset and reclen fields.
934 		 */
935 		if (oldcall) {
936 			idb.d_off = (linux_off_t)linux_reclen;
937 			idb.d_reclen = (u_short)bdp->d_namlen;
938 		} else {
939 			if (sizeof (idb.d_off) <= 4 && (off >> 32) != 0) {
940 				compat_offseterr(vp, "linux_getdents");
941 				error = EINVAL;
942 				goto out;
943 			}
944 			idb.d_off = (linux_off_t)off;
945 			idb.d_reclen = (u_short)linux_reclen;
946 		}
947 		strcpy(idb.d_name, bdp->d_name);
948 		if ((error = copyout((caddr_t)&idb, outp, linux_reclen)))
949 			goto out;
950 		/* advance past this real entry */
951 		inp += reclen;
952 		if (cookie)
953 			off = *cookie++; /* each entry points to itself */
954 		else
955 			off += reclen;
956 		/* advance output past Linux-shaped entry */
957 		outp += linux_reclen;
958 		resid -= linux_reclen;
959 		if (oldcall)
960 			break;
961 	}
962 
963 	/* if we squished out the whole block, try again */
964 	if (outp == (caddr_t)SCARG(uap, dent))
965 		goto again;
966 	fp->f_offset = off;	/* update the vnode offset */
967 
968 	if (oldcall)
969 		nbytes = resid + linux_reclen;
970 
971 eof:
972 	*retval = nbytes - resid;
973 out:
974 	VOP_UNLOCK(vp, 0);
975 	if (cookiebuf)
976 		free(cookiebuf, M_TEMP);
977 	free(tbuf, M_TEMP);
978 out1:
979 	FILE_UNUSE(fp, l);
980 	return error;
981 }
982 
983 /*
984  * Even when just using registers to pass arguments to syscalls you can
985  * have 5 of them on the i386. So this newer version of select() does
986  * this.
987  */
988 int
989 linux_sys_select(l, v, retval)
990 	struct lwp *l;
991 	void *v;
992 	register_t *retval;
993 {
994 	struct linux_sys_select_args /* {
995 		syscallarg(int) nfds;
996 		syscallarg(fd_set *) readfds;
997 		syscallarg(fd_set *) writefds;
998 		syscallarg(fd_set *) exceptfds;
999 		syscallarg(struct timeval *) timeout;
1000 	} */ *uap = v;
1001 
1002 	return linux_select1(l, retval, SCARG(uap, nfds), SCARG(uap, readfds),
1003 	    SCARG(uap, writefds), SCARG(uap, exceptfds), SCARG(uap, timeout));
1004 }
1005 
1006 /*
1007  * Common code for the old and new versions of select(). A couple of
1008  * things are important:
1009  * 1) return the amount of time left in the 'timeout' parameter
1010  * 2) select never returns ERESTART on Linux, always return EINTR
1011  */
1012 int
1013 linux_select1(l, retval, nfds, readfds, writefds, exceptfds, timeout)
1014 	struct lwp *l;
1015 	register_t *retval;
1016 	int nfds;
1017 	fd_set *readfds, *writefds, *exceptfds;
1018 	struct timeval *timeout;
1019 {
1020 	struct sys_select_args bsa;
1021 	struct proc *p = l->l_proc;
1022 	struct timeval tv0, tv1, utv, *tvp;
1023 	caddr_t sg;
1024 	int error;
1025 
1026 	SCARG(&bsa, nd) = nfds;
1027 	SCARG(&bsa, in) = readfds;
1028 	SCARG(&bsa, ou) = writefds;
1029 	SCARG(&bsa, ex) = exceptfds;
1030 	SCARG(&bsa, tv) = timeout;
1031 
1032 	/*
1033 	 * Store current time for computation of the amount of
1034 	 * time left.
1035 	 */
1036 	if (timeout) {
1037 		if ((error = copyin(timeout, &utv, sizeof(utv))))
1038 			return error;
1039 		if (itimerfix(&utv)) {
1040 			/*
1041 			 * The timeval was invalid.  Convert it to something
1042 			 * valid that will act as it does under Linux.
1043 			 */
1044 			sg = stackgap_init(p, 0);
1045 			tvp = stackgap_alloc(p, &sg, sizeof(utv));
1046 			utv.tv_sec += utv.tv_usec / 1000000;
1047 			utv.tv_usec %= 1000000;
1048 			if (utv.tv_usec < 0) {
1049 				utv.tv_sec -= 1;
1050 				utv.tv_usec += 1000000;
1051 			}
1052 			if (utv.tv_sec < 0)
1053 				timerclear(&utv);
1054 			if ((error = copyout(&utv, tvp, sizeof(utv))))
1055 				return error;
1056 			SCARG(&bsa, tv) = tvp;
1057 		}
1058 		microtime(&tv0);
1059 	}
1060 
1061 	error = sys_select(l, &bsa, retval);
1062 	if (error) {
1063 		/*
1064 		 * See fs/select.c in the Linux kernel.  Without this,
1065 		 * Maelstrom doesn't work.
1066 		 */
1067 		if (error == ERESTART)
1068 			error = EINTR;
1069 		return error;
1070 	}
1071 
1072 	if (timeout) {
1073 		if (*retval) {
1074 			/*
1075 			 * Compute how much time was left of the timeout,
1076 			 * by subtracting the current time and the time
1077 			 * before we started the call, and subtracting
1078 			 * that result from the user-supplied value.
1079 			 */
1080 			microtime(&tv1);
1081 			timersub(&tv1, &tv0, &tv1);
1082 			timersub(&utv, &tv1, &utv);
1083 			if (utv.tv_sec < 0)
1084 				timerclear(&utv);
1085 		} else
1086 			timerclear(&utv);
1087 		if ((error = copyout(&utv, timeout, sizeof(utv))))
1088 			return error;
1089 	}
1090 
1091 	return 0;
1092 }
1093 
1094 /*
1095  * Get the process group of a certain process. Look it up
1096  * and return the value.
1097  */
1098 int
1099 linux_sys_getpgid(l, v, retval)
1100 	struct lwp *l;
1101 	void *v;
1102 	register_t *retval;
1103 {
1104 	struct linux_sys_getpgid_args /* {
1105 		syscallarg(int) pid;
1106 	} */ *uap = v;
1107 	struct proc *p = l->l_proc;
1108 	struct proc *targp;
1109 
1110 	if (SCARG(uap, pid) != 0 && SCARG(uap, pid) != p->p_pid) {
1111 		if ((targp = pfind(SCARG(uap, pid))) == 0)
1112 			return ESRCH;
1113 	}
1114 	else
1115 		targp = p;
1116 
1117 	retval[0] = targp->p_pgid;
1118 	return 0;
1119 }
1120 
1121 /*
1122  * Set the 'personality' (emulation mode) for the current process. Only
1123  * accept the Linux personality here (0). This call is needed because
1124  * the Linux ELF crt0 issues it in an ugly kludge to make sure that
1125  * ELF binaries run in Linux mode, not SVR4 mode.
1126  */
1127 int
1128 linux_sys_personality(l, v, retval)
1129 	struct lwp *l;
1130 	void *v;
1131 	register_t *retval;
1132 {
1133 	struct linux_sys_personality_args /* {
1134 		syscallarg(int) per;
1135 	} */ *uap = v;
1136 
1137 	if (SCARG(uap, per) != 0)
1138 		return EINVAL;
1139 	retval[0] = 0;
1140 	return 0;
1141 }
1142 #endif /* !COMPAT_LINUX32 */
1143 
1144 #if defined(__i386__) || defined(__m68k__) || defined(COMPAT_LINUX32)
1145 /*
1146  * The calls are here because of type conversions.
1147  */
1148 int
1149 linux_sys_setreuid16(l, v, retval)
1150 	struct lwp *l;
1151 	void *v;
1152 	register_t *retval;
1153 {
1154 	struct linux_sys_setreuid16_args /* {
1155 		syscallarg(int) ruid;
1156 		syscallarg(int) euid;
1157 	} */ *uap = v;
1158 	struct sys_setreuid_args bsa;
1159 
1160 	SCARG(&bsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
1161 		(uid_t)-1 : SCARG(uap, ruid);
1162 	SCARG(&bsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
1163 		(uid_t)-1 : SCARG(uap, euid);
1164 
1165 	return sys_setreuid(l, &bsa, retval);
1166 }
1167 
1168 int
1169 linux_sys_setregid16(l, v, retval)
1170 	struct lwp *l;
1171 	void *v;
1172 	register_t *retval;
1173 {
1174 	struct linux_sys_setregid16_args /* {
1175 		syscallarg(int) rgid;
1176 		syscallarg(int) egid;
1177 	} */ *uap = v;
1178 	struct sys_setregid_args bsa;
1179 
1180 	SCARG(&bsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
1181 		(uid_t)-1 : SCARG(uap, rgid);
1182 	SCARG(&bsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
1183 		(uid_t)-1 : SCARG(uap, egid);
1184 
1185 	return sys_setregid(l, &bsa, retval);
1186 }
1187 
1188 int
1189 linux_sys_setresuid16(l, v, retval)
1190 	struct lwp *l;
1191 	void *v;
1192 	register_t *retval;
1193 {
1194 	struct linux_sys_setresuid16_args /* {
1195 		syscallarg(uid_t) ruid;
1196 		syscallarg(uid_t) euid;
1197 		syscallarg(uid_t) suid;
1198 	} */ *uap = v;
1199 	struct linux_sys_setresuid16_args lsa;
1200 
1201 	SCARG(&lsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
1202 		(uid_t)-1 : SCARG(uap, ruid);
1203 	SCARG(&lsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
1204 		(uid_t)-1 : SCARG(uap, euid);
1205 	SCARG(&lsa, suid) = ((linux_uid_t)SCARG(uap, suid) == (linux_uid_t)-1) ?
1206 		(uid_t)-1 : SCARG(uap, suid);
1207 
1208 	return linux_sys_setresuid(l, &lsa, retval);
1209 }
1210 
1211 int
1212 linux_sys_setresgid16(l, v, retval)
1213 	struct lwp *l;
1214 	void *v;
1215 	register_t *retval;
1216 {
1217 	struct linux_sys_setresgid16_args /* {
1218 		syscallarg(gid_t) rgid;
1219 		syscallarg(gid_t) egid;
1220 		syscallarg(gid_t) sgid;
1221 	} */ *uap = v;
1222 	struct linux_sys_setresgid16_args lsa;
1223 
1224 	SCARG(&lsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
1225 		(gid_t)-1 : SCARG(uap, rgid);
1226 	SCARG(&lsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
1227 		(gid_t)-1 : SCARG(uap, egid);
1228 	SCARG(&lsa, sgid) = ((linux_gid_t)SCARG(uap, sgid) == (linux_gid_t)-1) ?
1229 		(gid_t)-1 : SCARG(uap, sgid);
1230 
1231 	return linux_sys_setresgid(l, &lsa, retval);
1232 }
1233 
1234 int
1235 linux_sys_getgroups16(l, v, retval)
1236 	struct lwp *l;
1237 	void *v;
1238 	register_t *retval;
1239 {
1240 	struct linux_sys_getgroups16_args /* {
1241 		syscallarg(int) gidsetsize;
1242 		syscallarg(linux_gid_t *) gidset;
1243 	} */ *uap = v;
1244 	struct proc *p = l->l_proc;
1245 	caddr_t sg;
1246 	int n, error, i;
1247 	struct sys_getgroups_args bsa;
1248 	gid_t *bset, *kbset;
1249 	linux_gid_t *lset;
1250 	kauth_cred_t pc = l->l_cred;
1251 
1252 	n = SCARG(uap, gidsetsize);
1253 	if (n < 0)
1254 		return EINVAL;
1255 	error = 0;
1256 	bset = kbset = NULL;
1257 	lset = NULL;
1258 	if (n > 0) {
1259 		n = min(kauth_cred_ngroups(pc), n);
1260 		sg = stackgap_init(p, 0);
1261 		bset = stackgap_alloc(p, &sg, n * sizeof (gid_t));
1262 		kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK);
1263 		lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
1264 		if (bset == NULL || kbset == NULL || lset == NULL)
1265 		{
1266 			error = ENOMEM;
1267 			goto out;
1268 		}
1269 		SCARG(&bsa, gidsetsize) = n;
1270 		SCARG(&bsa, gidset) = bset;
1271 		error = sys_getgroups(l, &bsa, retval);
1272 		if (error != 0)
1273 			goto out;
1274 		error = copyin(bset, kbset, n * sizeof (gid_t));
1275 		if (error != 0)
1276 			goto out;
1277 		for (i = 0; i < n; i++)
1278 			lset[i] = (linux_gid_t)kbset[i];
1279 		error = copyout(lset, SCARG(uap, gidset),
1280 		    n * sizeof (linux_gid_t));
1281 	} else
1282 		*retval = kauth_cred_ngroups(pc);
1283 out:
1284 	if (kbset != NULL)
1285 		free(kbset, M_TEMP);
1286 	if (lset != NULL)
1287 		free(lset, M_TEMP);
1288 	return error;
1289 }
1290 
1291 int
1292 linux_sys_setgroups16(l, v, retval)
1293 	struct lwp *l;
1294 	void *v;
1295 	register_t *retval;
1296 {
1297 	struct linux_sys_setgroups16_args /* {
1298 		syscallarg(int) gidsetsize;
1299 		syscallarg(linux_gid_t *) gidset;
1300 	} */ *uap = v;
1301 	struct proc *p = l->l_proc;
1302 	caddr_t sg;
1303 	int n;
1304 	int error, i;
1305 	struct sys_setgroups_args bsa;
1306 	gid_t *bset, *kbset;
1307 	linux_gid_t *lset;
1308 
1309 	n = SCARG(uap, gidsetsize);
1310 	if (n < 0 || n > NGROUPS)
1311 		return EINVAL;
1312 	sg = stackgap_init(p, 0);
1313 	bset = stackgap_alloc(p, &sg, n * sizeof (gid_t));
1314 	lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
1315 	kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK);
1316 	if (bset == NULL || kbset == NULL || lset == NULL)
1317 	{
1318 		error = ENOMEM;
1319 		goto out;
1320 	}
1321 	error = copyin(SCARG(uap, gidset), lset, n * sizeof (linux_gid_t));
1322 	if (error != 0)
1323 		goto out;
1324 	for (i = 0; i < n; i++)
1325 		kbset[i] = (gid_t)lset[i];
1326 	error = copyout(kbset, bset, n * sizeof (gid_t));
1327 	if (error != 0)
1328 		goto out;
1329 	SCARG(&bsa, gidsetsize) = n;
1330 	SCARG(&bsa, gidset) = bset;
1331 	error = sys_setgroups(l, &bsa, retval);
1332 
1333 out:
1334 	if (lset != NULL)
1335 		free(lset, M_TEMP);
1336 	if (kbset != NULL)
1337 		free(kbset, M_TEMP);
1338 
1339 	return error;
1340 }
1341 
1342 #endif /* __i386__ || __m68k__ || COMPAT_LINUX32 */
1343 
1344 #ifndef COMPAT_LINUX32
1345 /*
1346  * We have nonexistent fsuid equal to uid.
1347  * If modification is requested, refuse.
1348  */
1349 int
1350 linux_sys_setfsuid(l, v, retval)
1351 	 struct lwp *l;
1352 	 void *v;
1353 	 register_t *retval;
1354 {
1355 	 struct linux_sys_setfsuid_args /* {
1356 		 syscallarg(uid_t) uid;
1357 	 } */ *uap = v;
1358 	 uid_t uid;
1359 
1360 	 uid = SCARG(uap, uid);
1361 	 if (kauth_cred_getuid(l->l_cred) != uid)
1362 		 return sys_nosys(l, v, retval);
1363 	 else
1364 		 return (0);
1365 }
1366 
1367 /* XXX XXX XXX */
1368 # ifndef alpha
1369 int
1370 linux_sys_getfsuid(l, v, retval)
1371 	struct lwp *l;
1372 	void *v;
1373 	register_t *retval;
1374 {
1375 	return sys_getuid(l, v, retval);
1376 }
1377 # endif
1378 
1379 int
1380 linux_sys_setresuid(l, v, retval)
1381 	struct lwp *l;
1382 	void *v;
1383 	register_t *retval;
1384 {
1385 	struct linux_sys_setresuid_args /* {
1386 		syscallarg(uid_t) ruid;
1387 		syscallarg(uid_t) euid;
1388 		syscallarg(uid_t) suid;
1389 	} */ *uap = v;
1390 
1391 	/*
1392 	 * Note: These checks are a little different than the NetBSD
1393 	 * setreuid(2) call performs.  This precisely follows the
1394 	 * behavior of the Linux kernel.
1395 	 */
1396 
1397 	return do_setresuid(l, SCARG(uap, ruid), SCARG(uap, euid),
1398 			    SCARG(uap, suid),
1399 			    ID_R_EQ_R | ID_R_EQ_E | ID_R_EQ_S |
1400 			    ID_E_EQ_R | ID_E_EQ_E | ID_E_EQ_S |
1401 			    ID_S_EQ_R | ID_S_EQ_E | ID_S_EQ_S );
1402 }
1403 
1404 int
1405 linux_sys_getresuid(l, v, retval)
1406 	struct lwp *l;
1407 	void *v;
1408 	register_t *retval;
1409 {
1410 	struct linux_sys_getresuid_args /* {
1411 		syscallarg(uid_t *) ruid;
1412 		syscallarg(uid_t *) euid;
1413 		syscallarg(uid_t *) suid;
1414 	} */ *uap = v;
1415 	kauth_cred_t pc = l->l_cred;
1416 	int error;
1417 	uid_t uid;
1418 
1419 	/*
1420 	 * Linux copies these values out to userspace like so:
1421 	 *
1422 	 *	1. Copy out ruid.
1423 	 *	2. If that succeeds, copy out euid.
1424 	 *	3. If both of those succeed, copy out suid.
1425 	 */
1426 	uid = kauth_cred_getuid(pc);
1427 	if ((error = copyout(&uid, SCARG(uap, ruid), sizeof(uid_t))) != 0)
1428 		return (error);
1429 
1430 	uid = kauth_cred_geteuid(pc);
1431 	if ((error = copyout(&uid, SCARG(uap, euid), sizeof(uid_t))) != 0)
1432 		return (error);
1433 
1434 	uid = kauth_cred_getsvuid(pc);
1435 
1436 	return (copyout(&uid, SCARG(uap, suid), sizeof(uid_t)));
1437 }
1438 
1439 int
1440 linux_sys_ptrace(l, v, retval)
1441 	struct lwp *l;
1442 	void *v;
1443 	register_t *retval;
1444 {
1445 #if defined(PTRACE) || defined(_LKM)
1446 	struct linux_sys_ptrace_args /* {
1447 		i386, m68k, powerpc: T=int
1448 		alpha, amd64: T=long
1449 		syscallarg(T) request;
1450 		syscallarg(T) pid;
1451 		syscallarg(T) addr;
1452 		syscallarg(T) data;
1453 	} */ *uap = v;
1454 	const int *ptr;
1455 	int request;
1456 	int error;
1457 #ifdef _LKM
1458 #define sys_ptrace (*sysent[SYS_ptrace].sy_call)
1459 #endif
1460 
1461 	ptr = linux_ptrace_request_map;
1462 	request = SCARG(uap, request);
1463 	while (*ptr != -1)
1464 		if (*ptr++ == request) {
1465 			struct sys_ptrace_args pta;
1466 
1467 			SCARG(&pta, req) = *ptr;
1468 			SCARG(&pta, pid) = SCARG(uap, pid);
1469 			SCARG(&pta, addr) = (caddr_t)SCARG(uap, addr);
1470 			SCARG(&pta, data) = SCARG(uap, data);
1471 
1472 			/*
1473 			 * Linux ptrace(PTRACE_CONT, pid, 0, 0) means actually
1474 			 * to continue where the process left off previously.
1475 			 * The same thing is achieved by addr == (caddr_t) 1
1476 			 * on NetBSD, so rewrite 'addr' appropriately.
1477 			 */
1478 			if (request == LINUX_PTRACE_CONT && SCARG(uap, addr)==0)
1479 				SCARG(&pta, addr) = (caddr_t) 1;
1480 
1481 			error = sys_ptrace(l, &pta, retval);
1482 			if (error)
1483 				return error;
1484 			switch (request) {
1485 			case LINUX_PTRACE_PEEKTEXT:
1486 			case LINUX_PTRACE_PEEKDATA:
1487 				error = copyout (retval,
1488 				    (caddr_t)SCARG(uap, data),
1489 				    sizeof *retval);
1490 				*retval = SCARG(uap, data);
1491 				break;
1492 			default:
1493 				break;
1494 			}
1495 			return error;
1496 		}
1497 		else
1498 			ptr++;
1499 
1500 	return LINUX_SYS_PTRACE_ARCH(l, uap, retval);
1501 #else
1502 	return ENOSYS;
1503 #endif /* PTRACE || _LKM */
1504 }
1505 
1506 int
1507 linux_sys_reboot(struct lwp *l, void *v, register_t *retval)
1508 {
1509 	struct linux_sys_reboot_args /* {
1510 		syscallarg(int) magic1;
1511 		syscallarg(int) magic2;
1512 		syscallarg(int) cmd;
1513 		syscallarg(void *) arg;
1514 	} */ *uap = v;
1515 	struct sys_reboot_args /* {
1516 		syscallarg(int) opt;
1517 		syscallarg(char *) bootstr;
1518 	} */ sra;
1519 	int error;
1520 
1521 	if ((error = kauth_authorize_generic(l->l_cred,
1522 	    KAUTH_GENERIC_ISSUSER, &l->l_acflag)) != 0)
1523 		return(error);
1524 
1525 	if (SCARG(uap, magic1) != LINUX_REBOOT_MAGIC1)
1526 		return(EINVAL);
1527 	if (SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2 &&
1528 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2A &&
1529 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2B)
1530 		return(EINVAL);
1531 
1532 	switch (SCARG(uap, cmd)) {
1533 	case LINUX_REBOOT_CMD_RESTART:
1534 		SCARG(&sra, opt) = RB_AUTOBOOT;
1535 		break;
1536 	case LINUX_REBOOT_CMD_HALT:
1537 		SCARG(&sra, opt) = RB_HALT;
1538 		break;
1539 	case LINUX_REBOOT_CMD_POWER_OFF:
1540 		SCARG(&sra, opt) = RB_HALT|RB_POWERDOWN;
1541 		break;
1542 	case LINUX_REBOOT_CMD_RESTART2:
1543 		/* Reboot with an argument. */
1544 		SCARG(&sra, opt) = RB_AUTOBOOT|RB_STRING;
1545 		SCARG(&sra, bootstr) = SCARG(uap, arg);
1546 		break;
1547 	case LINUX_REBOOT_CMD_CAD_ON:
1548 		return(EINVAL);	/* We don't implement ctrl-alt-delete */
1549 	case LINUX_REBOOT_CMD_CAD_OFF:
1550 		return(0);
1551 	default:
1552 		return(EINVAL);
1553 	}
1554 
1555 	return(sys_reboot(l, &sra, retval));
1556 }
1557 
1558 /*
1559  * Copy of compat_12_sys_swapon().
1560  */
1561 int
1562 linux_sys_swapon(l, v, retval)
1563 	struct lwp *l;
1564 	void *v;
1565 	register_t *retval;
1566 {
1567 	struct sys_swapctl_args ua;
1568 	struct linux_sys_swapon_args /* {
1569 		syscallarg(const char *) name;
1570 	} */ *uap = v;
1571 
1572 	SCARG(&ua, cmd) = SWAP_ON;
1573 	SCARG(&ua, arg) = (void *)__UNCONST(SCARG(uap, name));
1574 	SCARG(&ua, misc) = 0;	/* priority */
1575 	return (sys_swapctl(l, &ua, retval));
1576 }
1577 
1578 /*
1579  * Stop swapping to the file or block device specified by path.
1580  */
1581 int
1582 linux_sys_swapoff(l, v, retval)
1583 	struct lwp *l;
1584 	void *v;
1585 	register_t *retval;
1586 {
1587 	struct sys_swapctl_args ua;
1588 	struct linux_sys_swapoff_args /* {
1589 		syscallarg(const char *) path;
1590 	} */ *uap = v;
1591 
1592 	SCARG(&ua, cmd) = SWAP_OFF;
1593 	SCARG(&ua, arg) = __UNCONST(SCARG(uap, path)); /*XXXUNCONST*/
1594 	return (sys_swapctl(l, &ua, retval));
1595 }
1596 
1597 /*
1598  * Copy of compat_09_sys_setdomainname()
1599  */
1600 /* ARGSUSED */
1601 int
1602 linux_sys_setdomainname(l, v, retval)
1603 	struct lwp *l;
1604 	void *v;
1605 	register_t *retval;
1606 {
1607 	struct linux_sys_setdomainname_args /* {
1608 		syscallarg(char *) domainname;
1609 		syscallarg(int) len;
1610 	} */ *uap = v;
1611 	int name[2];
1612 
1613 	name[0] = CTL_KERN;
1614 	name[1] = KERN_DOMAINNAME;
1615 	return (old_sysctl(&name[0], 2, 0, 0, SCARG(uap, domainname),
1616 			    SCARG(uap, len), l));
1617 }
1618 
1619 /*
1620  * sysinfo()
1621  */
1622 /* ARGSUSED */
1623 int
1624 linux_sys_sysinfo(l, v, retval)
1625 	struct lwp *l;
1626 	void *v;
1627 	register_t *retval;
1628 {
1629 	struct linux_sys_sysinfo_args /* {
1630 		syscallarg(struct linux_sysinfo *) arg;
1631 	} */ *uap = v;
1632 	struct linux_sysinfo si;
1633 	struct loadavg *la;
1634 
1635 	si.uptime = time_uptime;
1636 	la = &averunnable;
1637 	si.loads[0] = la->ldavg[0] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1638 	si.loads[1] = la->ldavg[1] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1639 	si.loads[2] = la->ldavg[2] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1640 	si.totalram = ctob((u_long)physmem);
1641 	si.freeram = (u_long)uvmexp.free * uvmexp.pagesize;
1642 	si.sharedram = 0;	/* XXX */
1643 	si.bufferram = (u_long)uvmexp.filepages * uvmexp.pagesize;
1644 	si.totalswap = (u_long)uvmexp.swpages * uvmexp.pagesize;
1645 	si.freeswap =
1646 	    (u_long)(uvmexp.swpages - uvmexp.swpginuse) * uvmexp.pagesize;
1647 	si.procs = nprocs;
1648 
1649 	/* The following are only present in newer Linux kernels. */
1650 	si.totalbig = 0;
1651 	si.freebig = 0;
1652 	si.mem_unit = 1;
1653 
1654 	return (copyout(&si, SCARG(uap, arg), sizeof si));
1655 }
1656 
1657 int
1658 linux_sys_getrlimit(l, v, retval)
1659 	struct lwp *l;
1660 	void *v;
1661 	register_t *retval;
1662 {
1663 	struct linux_sys_getrlimit_args /* {
1664 		syscallarg(int) which;
1665 # ifdef LINUX_LARGEFILE64
1666 		syscallarg(struct rlimit *) rlp;
1667 # else
1668 		syscallarg(struct orlimit *) rlp;
1669 # endif
1670 	} */ *uap = v;
1671 	struct proc *p = l->l_proc;
1672 	caddr_t sg = stackgap_init(p, 0);
1673 	struct sys_getrlimit_args ap;
1674 	struct rlimit rl;
1675 # ifdef LINUX_LARGEFILE64
1676 	struct rlimit orl;
1677 # else
1678 	struct orlimit orl;
1679 # endif
1680 	int error;
1681 
1682 	SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which));
1683 	if ((error = SCARG(&ap, which)) < 0)
1684 		return -error;
1685 	SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl);
1686 	if ((error = sys_getrlimit(l, &ap, retval)) != 0)
1687 		return error;
1688 	if ((error = copyin(SCARG(&ap, rlp), &rl, sizeof(rl))) != 0)
1689 		return error;
1690 	bsd_to_linux_rlimit(&orl, &rl);
1691 
1692 	return copyout(&orl, SCARG(uap, rlp), sizeof(orl));
1693 }
1694 
1695 int
1696 linux_sys_setrlimit(l, v, retval)
1697 	struct lwp *l;
1698 	void *v;
1699 	register_t *retval;
1700 {
1701 	struct linux_sys_setrlimit_args /* {
1702 		syscallarg(int) which;
1703 # ifdef LINUX_LARGEFILE64
1704 		syscallarg(struct rlimit *) rlp;
1705 # else
1706 		syscallarg(struct orlimit *) rlp;
1707 # endif
1708 	} */ *uap = v;
1709 	struct proc *p = l->l_proc;
1710 	caddr_t sg = stackgap_init(p, 0);
1711 	struct sys_getrlimit_args ap;
1712 	struct rlimit rl;
1713 # ifdef LINUX_LARGEFILE64
1714 	struct rlimit orl;
1715 # else
1716 	struct orlimit orl;
1717 # endif
1718 	int error;
1719 
1720 	SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which));
1721 	SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl);
1722 	if ((error = SCARG(&ap, which)) < 0)
1723 		return -error;
1724 	if ((error = copyin(SCARG(uap, rlp), &orl, sizeof(orl))) != 0)
1725 		return error;
1726 	linux_to_bsd_rlimit(&rl, &orl);
1727 	if ((error = copyout(&rl, SCARG(&ap, rlp), sizeof(rl))) != 0)
1728 		return error;
1729 	return sys_setrlimit(l, &ap, retval);
1730 }
1731 
1732 # if !defined(__mips__) && !defined(__amd64__)
1733 /* XXX: this doesn't look 100% common, at least mips doesn't have it */
1734 int
1735 linux_sys_ugetrlimit(l, v, retval)
1736 	struct lwp *l;
1737 	void *v;
1738 	register_t *retval;
1739 {
1740 	return linux_sys_getrlimit(l, v, retval);
1741 }
1742 # endif
1743 
1744 /*
1745  * This gets called for unsupported syscalls. The difference to sys_nosys()
1746  * is that process does not get SIGSYS, the call just returns with ENOSYS.
1747  * This is the way Linux does it and glibc depends on this behaviour.
1748  */
1749 int
1750 linux_sys_nosys(l, v, retval)
1751 	struct lwp *l;
1752 	void *v;
1753 	register_t *retval;
1754 {
1755 	return (ENOSYS);
1756 }
1757 
1758 int
1759 linux_sys_getpriority(l, v, retval)
1760         struct lwp *l;
1761         void *v;
1762         register_t *retval;
1763 {
1764         struct linux_sys_getpriority_args /* {
1765                 syscallarg(int) which;
1766                 syscallarg(int) who;
1767         } */ *uap = v;
1768         struct sys_getpriority_args bsa;
1769         int error;
1770 
1771         SCARG(&bsa, which) = SCARG(uap, which);
1772         SCARG(&bsa, who) = SCARG(uap, who);
1773 
1774         if ((error = sys_getpriority(l, &bsa, retval)))
1775                 return error;
1776 
1777         *retval = NZERO - *retval;
1778 
1779         return 0;
1780 }
1781 
1782 #endif /* !COMPAT_LINUX32 */
1783