xref: /netbsd-src/sys/compat/linux/common/linux_misc.c (revision df0caa2637da0538ecdf6b878c4d08e684b43d8f)
1 /*	$NetBSD: linux_misc.c,v 1.139 2005/06/02 13:03:27 drochner 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 1.139 2005/06/02 13:03:27 drochner Exp $");
68 
69 #include <sys/param.h>
70 #include <sys/systm.h>
71 #include <sys/namei.h>
72 #include <sys/proc.h>
73 #include <sys/dirent.h>
74 #include <sys/file.h>
75 #include <sys/stat.h>
76 #include <sys/filedesc.h>
77 #include <sys/ioctl.h>
78 #include <sys/kernel.h>
79 #include <sys/malloc.h>
80 #include <sys/mbuf.h>
81 #include <sys/mman.h>
82 #include <sys/mount.h>
83 #include <sys/reboot.h>
84 #include <sys/resource.h>
85 #include <sys/resourcevar.h>
86 #include <sys/signal.h>
87 #include <sys/signalvar.h>
88 #include <sys/socket.h>
89 #include <sys/time.h>
90 #include <sys/times.h>
91 #include <sys/vnode.h>
92 #include <sys/uio.h>
93 #include <sys/wait.h>
94 #include <sys/utsname.h>
95 #include <sys/unistd.h>
96 #include <sys/swap.h>		/* for SWAP_ON */
97 #include <sys/sysctl.h>		/* for KERN_DOMAINNAME */
98 
99 #include <sys/ptrace.h>
100 #include <machine/ptrace.h>
101 
102 #include <sys/sa.h>
103 #include <sys/syscallargs.h>
104 
105 #include <compat/linux/common/linux_types.h>
106 #include <compat/linux/common/linux_signal.h>
107 
108 #include <compat/linux/linux_syscallargs.h>
109 
110 #include <compat/linux/common/linux_fcntl.h>
111 #include <compat/linux/common/linux_mmap.h>
112 #include <compat/linux/common/linux_dirent.h>
113 #include <compat/linux/common/linux_util.h>
114 #include <compat/linux/common/linux_misc.h>
115 #include <compat/linux/common/linux_ptrace.h>
116 #include <compat/linux/common/linux_reboot.h>
117 #include <compat/linux/common/linux_emuldata.h>
118 
119 const int linux_ptrace_request_map[] = {
120 	LINUX_PTRACE_TRACEME,	PT_TRACE_ME,
121 	LINUX_PTRACE_PEEKTEXT,	PT_READ_I,
122 	LINUX_PTRACE_PEEKDATA,	PT_READ_D,
123 	LINUX_PTRACE_POKETEXT,	PT_WRITE_I,
124 	LINUX_PTRACE_POKEDATA,	PT_WRITE_D,
125 	LINUX_PTRACE_CONT,	PT_CONTINUE,
126 	LINUX_PTRACE_KILL,	PT_KILL,
127 	LINUX_PTRACE_ATTACH,	PT_ATTACH,
128 	LINUX_PTRACE_DETACH,	PT_DETACH,
129 #ifdef PT_STEP
130 	LINUX_PTRACE_SINGLESTEP,	PT_STEP,
131 #endif
132 	-1
133 };
134 
135 const struct linux_mnttypes linux_fstypes[] = {
136 	{ MOUNT_FFS,		LINUX_DEFAULT_SUPER_MAGIC	},
137 	{ MOUNT_NFS,		LINUX_NFS_SUPER_MAGIC 		},
138 	{ MOUNT_MFS,		LINUX_DEFAULT_SUPER_MAGIC	},
139 	{ MOUNT_MSDOS,		LINUX_MSDOS_SUPER_MAGIC		},
140 	{ MOUNT_LFS,		LINUX_DEFAULT_SUPER_MAGIC	},
141 	{ MOUNT_FDESC,		LINUX_DEFAULT_SUPER_MAGIC	},
142 	{ MOUNT_PORTAL,		LINUX_DEFAULT_SUPER_MAGIC	},
143 	{ MOUNT_NULL,		LINUX_DEFAULT_SUPER_MAGIC	},
144 	{ MOUNT_OVERLAY,	LINUX_DEFAULT_SUPER_MAGIC	},
145 	{ MOUNT_UMAP,		LINUX_DEFAULT_SUPER_MAGIC	},
146 	{ MOUNT_KERNFS,		LINUX_DEFAULT_SUPER_MAGIC	},
147 	{ MOUNT_PROCFS,		LINUX_PROC_SUPER_MAGIC		},
148 	{ MOUNT_AFS,		LINUX_DEFAULT_SUPER_MAGIC	},
149 	{ MOUNT_CD9660,		LINUX_ISOFS_SUPER_MAGIC		},
150 	{ MOUNT_UNION,		LINUX_DEFAULT_SUPER_MAGIC	},
151 	{ MOUNT_ADOSFS,		LINUX_ADFS_SUPER_MAGIC		},
152 	{ MOUNT_EXT2FS,		LINUX_EXT2_SUPER_MAGIC		},
153 	{ MOUNT_CFS,		LINUX_DEFAULT_SUPER_MAGIC	},
154 	{ MOUNT_CODA,		LINUX_CODA_SUPER_MAGIC		},
155 	{ MOUNT_FILECORE,	LINUX_DEFAULT_SUPER_MAGIC	},
156 	{ MOUNT_NTFS,		LINUX_DEFAULT_SUPER_MAGIC	},
157 	{ MOUNT_SMBFS,		LINUX_SMB_SUPER_MAGIC		},
158 	{ MOUNT_PTYFS,		LINUX_DEVPTS_SUPER_MAGIC	}
159 };
160 const int linux_fstypes_cnt = sizeof(linux_fstypes) / sizeof(linux_fstypes[0]);
161 
162 #ifdef DEBUG_LINUX
163 #define DPRINTF(a)	uprintf a
164 #else
165 #define DPRINTF(a)
166 #endif
167 
168 /* Local linux_misc.c functions: */
169 #ifndef __amd64__
170 static void bsd_to_linux_statfs __P((const struct statvfs *,
171     struct linux_statfs *));
172 #endif
173 static int linux_to_bsd_limit __P((int));
174 static void linux_to_bsd_mmap_args __P((struct sys_mmap_args *,
175     const struct linux_sys_mmap_args *));
176 static int linux_mmap __P((struct lwp *, struct linux_sys_mmap_args *,
177     register_t *, off_t));
178 
179 
180 /*
181  * The information on a terminated (or stopped) process needs
182  * to be converted in order for Linux binaries to get a valid signal
183  * number out of it.
184  */
185 void
186 bsd_to_linux_wstat(st)
187 	int *st;
188 {
189 
190 	int sig;
191 
192 	if (WIFSIGNALED(*st)) {
193 		sig = WTERMSIG(*st);
194 		if (sig >= 0 && sig < NSIG)
195 			*st= (*st& ~0177) | native_to_linux_signo[sig];
196 	} else if (WIFSTOPPED(*st)) {
197 		sig = WSTOPSIG(*st);
198 		if (sig >= 0 && sig < NSIG)
199 			*st = (*st & ~0xff00) |
200 			    (native_to_linux_signo[sig] << 8);
201 	}
202 }
203 
204 /*
205  * wait4(2).  Passed on to the NetBSD call, surrounded by code to
206  * reserve some space for a NetBSD-style wait status, and converting
207  * it to what Linux wants.
208  */
209 int
210 linux_sys_wait4(l, v, retval)
211 	struct lwp *l;
212 	void *v;
213 	register_t *retval;
214 {
215 	struct linux_sys_wait4_args /* {
216 		syscallarg(int) pid;
217 		syscallarg(int *) status;
218 		syscallarg(int) options;
219 		syscallarg(struct rusage *) rusage;
220 	} */ *uap = v;
221 	struct proc *p = l->l_proc;
222 	struct sys_wait4_args w4a;
223 	int error, *status, tstat, options, linux_options;
224 	caddr_t sg;
225 
226 	if (SCARG(uap, status) != NULL) {
227 		sg = stackgap_init(p, 0);
228 		status = (int *) stackgap_alloc(p, &sg, sizeof *status);
229 	} else
230 		status = NULL;
231 
232 	linux_options = SCARG(uap, options);
233 	options = 0;
234 	if (linux_options & ~(LINUX_WAIT4_KNOWNFLAGS))
235 		return (EINVAL);
236 
237 	if (linux_options & LINUX_WAIT4_WNOHANG)
238 		options |= WNOHANG;
239 	if (linux_options & LINUX_WAIT4_WUNTRACED)
240 		options |= WUNTRACED;
241 	if (linux_options & LINUX_WAIT4_WALL)
242 		options |= WALLSIG;
243 	if (linux_options & LINUX_WAIT4_WCLONE)
244 		options |= WALTSIG;
245 #ifdef DIAGNOSTIC
246 	if (linux_options & LINUX_WAIT4_WNOTHREAD)
247 		printf("WARNING: %s: linux process %d.%d called "
248 		       "waitpid with __WNOTHREAD set!",
249 		       __FILE__, p->p_pid, l->l_lid);
250 
251 #endif
252 
253 	SCARG(&w4a, pid) = SCARG(uap, pid);
254 	SCARG(&w4a, status) = status;
255 	SCARG(&w4a, options) = options;
256 	SCARG(&w4a, rusage) = SCARG(uap, rusage);
257 
258 	if ((error = sys_wait4(l, &w4a, retval)))
259 		return error;
260 
261 	sigdelset(&p->p_sigctx.ps_siglist, SIGCHLD);
262 
263 	if (status != NULL) {
264 		if ((error = copyin(status, &tstat, sizeof tstat)))
265 			return error;
266 
267 		bsd_to_linux_wstat(&tstat);
268 		return copyout(&tstat, SCARG(uap, status), sizeof tstat);
269 	}
270 
271 	return 0;
272 }
273 
274 /*
275  * Linux brk(2). The check if the new address is >= the old one is
276  * done in the kernel in Linux. NetBSD does it in the library.
277  */
278 int
279 linux_sys_brk(l, v, retval)
280 	struct lwp *l;
281 	void *v;
282 	register_t *retval;
283 {
284 	struct linux_sys_brk_args /* {
285 		syscallarg(char *) nsize;
286 	} */ *uap = v;
287 	struct proc *p = l->l_proc;
288 	char *nbrk = SCARG(uap, nsize);
289 	struct sys_obreak_args oba;
290 	struct vmspace *vm = p->p_vmspace;
291 	struct linux_emuldata *ed = (struct linux_emuldata*)p->p_emuldata;
292 
293 	SCARG(&oba, nsize) = nbrk;
294 
295 	if ((caddr_t) nbrk > vm->vm_daddr && sys_obreak(l, &oba, retval) == 0)
296 		ed->s->p_break = (char*)nbrk;
297 	else
298 		nbrk = ed->s->p_break;
299 
300 	retval[0] = (register_t)nbrk;
301 
302 	return 0;
303 }
304 
305 #ifndef __amd64__
306 /*
307  * Convert NetBSD statvfs structure to Linux statfs structure.
308  * Linux doesn't have f_flag, and we can't set f_frsize due
309  * to glibc statvfs() bug (see below).
310  */
311 static void
312 bsd_to_linux_statfs(bsp, lsp)
313 	const struct statvfs *bsp;
314 	struct linux_statfs *lsp;
315 {
316 	int i;
317 
318 	for (i = 0; i < linux_fstypes_cnt; i++) {
319 		if (strcmp(bsp->f_fstypename, linux_fstypes[i].bsd) == 0) {
320 			lsp->l_ftype = linux_fstypes[i].linux;
321 			break;
322 		}
323 	}
324 
325 	if (i == linux_fstypes_cnt) {
326 		DPRINTF(("unhandled fstype in linux emulation: %s\n",
327 		    bsp->f_fstypename));
328 		lsp->l_ftype = LINUX_DEFAULT_SUPER_MAGIC;
329 	}
330 
331 	/*
332 	 * The sizes are expressed in number of blocks. The block
333 	 * size used for the size is f_frsize for POSIX-compliant
334 	 * statvfs. Linux statfs uses f_bsize as the block size
335 	 * (f_frsize used to not be available in Linux struct statfs).
336 	 * However, glibc 2.3.3 statvfs() wrapper fails to adjust the block
337 	 * counts for different f_frsize if f_frsize is provided by the kernel.
338 	 * POSIX conforming apps thus get wrong size if f_frsize
339 	 * is different to f_bsize. Thus, we just pretend we don't
340 	 * support f_frsize.
341 	 */
342 
343 	lsp->l_fbsize = bsp->f_frsize;
344 	lsp->l_ffrsize = 0;			/* compat */
345 	lsp->l_fblocks = bsp->f_blocks;
346 	lsp->l_fbfree = bsp->f_bfree;
347 	lsp->l_fbavail = bsp->f_bavail;
348 	lsp->l_ffiles = bsp->f_files;
349 	lsp->l_fffree = bsp->f_ffree;
350 	/* Linux sets the fsid to 0..., we don't */
351 	lsp->l_ffsid.val[0] = bsp->f_fsidx.__fsid_val[0];
352 	lsp->l_ffsid.val[1] = bsp->f_fsidx.__fsid_val[1];
353 	lsp->l_fnamelen = bsp->f_namemax;
354 	(void)memset(lsp->l_fspare, 0, sizeof(lsp->l_fspare));
355 }
356 
357 /*
358  * Implement the fs stat functions. Straightforward.
359  */
360 int
361 linux_sys_statfs(l, v, retval)
362 	struct lwp *l;
363 	void *v;
364 	register_t *retval;
365 {
366 	struct linux_sys_statfs_args /* {
367 		syscallarg(const char *) path;
368 		syscallarg(struct linux_statfs *) sp;
369 	} */ *uap = v;
370 	struct proc *p = l->l_proc;
371 	struct statvfs btmp, *bsp;
372 	struct linux_statfs ltmp;
373 	struct sys_statvfs1_args bsa;
374 	caddr_t sg;
375 	int error;
376 
377 	sg = stackgap_init(p, 0);
378 	bsp = (struct statvfs *) stackgap_alloc(p, &sg, sizeof (struct statvfs));
379 
380 	CHECK_ALT_EXIST(p, &sg, SCARG(uap, path));
381 
382 	SCARG(&bsa, path) = SCARG(uap, path);
383 	SCARG(&bsa, buf) = bsp;
384 	SCARG(&bsa, flags) = ST_WAIT;
385 
386 	if ((error = sys_statvfs1(l, &bsa, retval)))
387 		return error;
388 
389 	if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp)))
390 		return error;
391 
392 	bsd_to_linux_statfs(&btmp, &ltmp);
393 
394 	return copyout((caddr_t) &ltmp, (caddr_t) SCARG(uap, sp), sizeof ltmp);
395 }
396 
397 int
398 linux_sys_fstatfs(l, v, retval)
399 	struct lwp *l;
400 	void *v;
401 	register_t *retval;
402 {
403 	struct linux_sys_fstatfs_args /* {
404 		syscallarg(int) fd;
405 		syscallarg(struct linux_statfs *) sp;
406 	} */ *uap = v;
407 	struct proc *p = l->l_proc;
408 	struct statvfs btmp, *bsp;
409 	struct linux_statfs ltmp;
410 	struct sys_fstatvfs1_args bsa;
411 	caddr_t sg;
412 	int error;
413 
414 	sg = stackgap_init(p, 0);
415 	bsp = (struct statvfs *) stackgap_alloc(p, &sg, sizeof (struct statvfs));
416 
417 	SCARG(&bsa, fd) = SCARG(uap, fd);
418 	SCARG(&bsa, buf) = bsp;
419 	SCARG(&bsa, flags) = ST_WAIT;
420 
421 	if ((error = sys_fstatvfs1(l, &bsa, retval)))
422 		return error;
423 
424 	if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp)))
425 		return error;
426 
427 	bsd_to_linux_statfs(&btmp, &ltmp);
428 
429 	return copyout((caddr_t) &ltmp, (caddr_t) SCARG(uap, sp), sizeof ltmp);
430 }
431 #endif /* __amd64__ */
432 
433 /*
434  * uname(). Just copy the info from the various strings stored in the
435  * kernel, and put it in the Linux utsname structure. That structure
436  * is almost the same as the NetBSD one, only it has fields 65 characters
437  * long, and an extra domainname field.
438  */
439 int
440 linux_sys_uname(l, v, retval)
441 	struct lwp *l;
442 	void *v;
443 	register_t *retval;
444 {
445 	struct linux_sys_uname_args /* {
446 		syscallarg(struct linux_utsname *) up;
447 	} */ *uap = v;
448 	struct linux_utsname luts;
449 
450 	strncpy(luts.l_sysname, linux_sysname, sizeof(luts.l_sysname));
451 	strncpy(luts.l_nodename, hostname, sizeof(luts.l_nodename));
452 	strncpy(luts.l_release, linux_release, sizeof(luts.l_release));
453 	strncpy(luts.l_version, linux_version, sizeof(luts.l_version));
454 	strncpy(luts.l_machine, machine, sizeof(luts.l_machine));
455 	strncpy(luts.l_domainname, domainname, sizeof(luts.l_domainname));
456 
457 	return copyout(&luts, SCARG(uap, up), sizeof(luts));
458 }
459 
460 /* Used directly on: alpha, mips, ppc, sparc, sparc64 */
461 /* Used indirectly on: arm, i386, m68k */
462 
463 /*
464  * New type Linux mmap call.
465  * Only called directly on machines with >= 6 free regs.
466  */
467 int
468 linux_sys_mmap(l, v, retval)
469 	struct lwp *l;
470 	void *v;
471 	register_t *retval;
472 {
473 	struct linux_sys_mmap_args /* {
474 		syscallarg(unsigned long) addr;
475 		syscallarg(size_t) len;
476 		syscallarg(int) prot;
477 		syscallarg(int) flags;
478 		syscallarg(int) fd;
479 		syscallarg(linux_off_t) offset;
480 	} */ *uap = v;
481 
482 	if (SCARG(uap, offset) & PAGE_MASK)
483 		return EINVAL;
484 
485 	return linux_mmap(l, uap, retval, SCARG(uap, offset));
486 }
487 
488 /*
489  * Guts of most architectures' mmap64() implementations.  This shares
490  * its list of arguments with linux_sys_mmap().
491  *
492  * The difference in linux_sys_mmap2() is that "offset" is actually
493  * (offset / pagesize), not an absolute byte count.  This translation
494  * to pagesize offsets is done inside glibc between the mmap64() call
495  * point, and the actual syscall.
496  */
497 int
498 linux_sys_mmap2(l, v, retval)
499 	struct lwp *l;
500 	void *v;
501 	register_t *retval;
502 {
503 	struct linux_sys_mmap2_args /* {
504 		syscallarg(unsigned long) addr;
505 		syscallarg(size_t) len;
506 		syscallarg(int) prot;
507 		syscallarg(int) flags;
508 		syscallarg(int) fd;
509 		syscallarg(linux_off_t) offset;
510 	} */ *uap = v;
511 
512 	return linux_mmap(l, uap, retval,
513 	    ((off_t)SCARG(uap, offset)) << PAGE_SHIFT);
514 }
515 
516 /*
517  * Massage arguments and call system mmap(2).
518  */
519 static int
520 linux_mmap(l, uap, retval, offset)
521 	struct lwp *l;
522 	struct linux_sys_mmap_args *uap;
523 	register_t *retval;
524 	off_t offset;
525 {
526 	struct sys_mmap_args cma;
527 	int error;
528 	size_t mmoff=0;
529 
530 	if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN) {
531 		/*
532 		 * Request for stack-like memory segment. On linux, this
533 		 * works by mmap()ping (small) segment, which is automatically
534 		 * extended when page fault happens below the currently
535 		 * allocated area. We emulate this by allocating (typically
536 		 * bigger) segment sized at current stack size limit, and
537 		 * offsetting the requested and returned address accordingly.
538 		 * Since physical pages are only allocated on-demand, this
539 		 * is effectively identical.
540 		 */
541 		rlim_t ssl = l->l_proc->p_rlimit[RLIMIT_STACK].rlim_cur;
542 
543 		if (SCARG(uap, len) < ssl) {
544 			/* Compute the address offset */
545 			mmoff = round_page(ssl) - SCARG(uap, len);
546 
547 			if (SCARG(uap, addr))
548 				SCARG(uap, addr) -= mmoff;
549 
550 			SCARG(uap, len) = (size_t) ssl;
551 		}
552 	}
553 
554 	linux_to_bsd_mmap_args(&cma, uap);
555 	SCARG(&cma, pos) = offset;
556 
557 	error = sys_mmap(l, &cma, retval);
558 	if (error)
559 		return (error);
560 
561 	/* Shift the returned address for stack-like segment if necessary */
562 	if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN && mmoff)
563 		retval[0] += mmoff;
564 
565 	return (0);
566 }
567 
568 static void
569 linux_to_bsd_mmap_args(cma, uap)
570 	struct sys_mmap_args *cma;
571 	const struct linux_sys_mmap_args *uap;
572 {
573 	int flags = MAP_TRYFIXED, fl = SCARG(uap, flags);
574 
575 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_SHARED, MAP_SHARED);
576 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_PRIVATE, MAP_PRIVATE);
577 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_FIXED, MAP_FIXED);
578 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_ANON, MAP_ANON);
579 	/* XXX XAX ERH: Any other flags here?  There are more defined... */
580 
581 	SCARG(cma, addr) = (void *)SCARG(uap, addr);
582 	SCARG(cma, len) = SCARG(uap, len);
583 	SCARG(cma, prot) = SCARG(uap, prot);
584 	if (SCARG(cma, prot) & VM_PROT_WRITE) /* XXX */
585 		SCARG(cma, prot) |= VM_PROT_READ;
586 	SCARG(cma, flags) = flags;
587 	SCARG(cma, fd) = flags & MAP_ANON ? -1 : SCARG(uap, fd);
588 	SCARG(cma, pad) = 0;
589 }
590 
591 int
592 linux_sys_mremap(l, v, retval)
593 	struct lwp *l;
594 	void *v;
595 	register_t *retval;
596 {
597 	struct linux_sys_mremap_args /* {
598 		syscallarg(void *) old_address;
599 		syscallarg(size_t) old_size;
600 		syscallarg(size_t) new_size;
601 		syscallarg(u_long) flags;
602 	} */ *uap = v;
603 	struct sys_munmap_args mua;
604 	size_t old_size, new_size;
605 	int error;
606 
607 	old_size = round_page(SCARG(uap, old_size));
608 	new_size = round_page(SCARG(uap, new_size));
609 
610 	/*
611 	 * Growing mapped region.
612 	 */
613 	if (new_size > old_size) {
614 		/*
615 		 * XXX Implement me.  What we probably want to do is
616 		 * XXX dig out the guts of the old mapping, mmap that
617 		 * XXX object again with the new size, then munmap
618 		 * XXX the old mapping.
619 		 */
620 		*retval = 0;
621 		return (ENOMEM);
622 	}
623 
624 	/*
625 	 * Shrinking mapped region.
626 	 */
627 	if (new_size < old_size) {
628 		SCARG(&mua, addr) = (caddr_t)SCARG(uap, old_address) +
629 		    new_size;
630 		SCARG(&mua, len) = old_size - new_size;
631 		error = sys_munmap(l, &mua, retval);
632 		*retval = error ? 0 : (register_t)SCARG(uap, old_address);
633 		return (error);
634 	}
635 
636 	/*
637 	 * No change.
638 	 */
639 	*retval = (register_t)SCARG(uap, old_address);
640 	return (0);
641 }
642 
643 int
644 linux_sys_msync(l, v, retval)
645 	struct lwp *l;
646 	void *v;
647 	register_t *retval;
648 {
649 	struct linux_sys_msync_args /* {
650 		syscallarg(caddr_t) addr;
651 		syscallarg(int) len;
652 		syscallarg(int) fl;
653 	} */ *uap = v;
654 
655 	struct sys___msync13_args bma;
656 
657 	/* flags are ignored */
658 	SCARG(&bma, addr) = SCARG(uap, addr);
659 	SCARG(&bma, len) = SCARG(uap, len);
660 	SCARG(&bma, flags) = SCARG(uap, fl);
661 
662 	return sys___msync13(l, &bma, retval);
663 }
664 
665 int
666 linux_sys_mprotect(l, v, retval)
667 	struct lwp *l;
668 	void *v;
669 	register_t *retval;
670 {
671 	struct linux_sys_mprotect_args /* {
672 		syscallarg(const void *) start;
673 		syscallarg(unsigned long) len;
674 		syscallarg(int) prot;
675 	} */ *uap = v;
676 	unsigned long end, start = (unsigned long)SCARG(uap, start), len;
677 	int prot = SCARG(uap, prot);
678 	struct vm_map_entry *entry;
679 	struct vm_map *map = &l->l_proc->p_vmspace->vm_map;
680 
681 	if (start & PAGE_MASK)
682 		return EINVAL;
683 
684 	len = round_page(SCARG(uap, len));
685 	end = start + len;
686 
687 	if (end < start)
688 		return EINVAL;
689 	else if (end == start)
690 		return 0;
691 
692 	if (SCARG(uap, prot) & ~(PROT_READ | PROT_WRITE | PROT_EXEC))
693 		return EINVAL;
694 
695 	vm_map_lock(map);
696 #ifdef notdef
697 	VM_MAP_RANGE_CHECK(map, start, end);
698 #endif
699 	if (!uvm_map_lookup_entry(map, start, &entry) || entry->start > start) {
700 		vm_map_unlock(map);
701 		return ENOMEM;
702 	}
703 	vm_map_unlock(map);
704 	return uvm_map_protect(map, start, end, prot, FALSE);
705 }
706 
707 /*
708  * This code is partly stolen from src/lib/libc/compat-43/times.c
709  */
710 
711 #define	CONVTCK(r)	(r.tv_sec * hz + r.tv_usec / (1000000 / hz))
712 
713 int
714 linux_sys_times(l, v, retval)
715 	struct lwp *l;
716 	void *v;
717 	register_t *retval;
718 {
719 	struct linux_sys_times_args /* {
720 		syscallarg(struct times *) tms;
721 	} */ *uap = v;
722 	struct proc *p = l->l_proc;
723 	struct timeval t;
724 	int error, s;
725 
726 	if (SCARG(uap, tms)) {
727 		struct linux_tms ltms;
728 		struct rusage ru;
729 
730 		calcru(p, &ru.ru_utime, &ru.ru_stime, NULL);
731 		ltms.ltms_utime = CONVTCK(ru.ru_utime);
732 		ltms.ltms_stime = CONVTCK(ru.ru_stime);
733 
734 		ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime);
735 		ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime);
736 
737 		if ((error = copyout(&ltms, SCARG(uap, tms), sizeof ltms)))
738 			return error;
739 	}
740 
741 	s = splclock();
742 	timersub(&time, &boottime, &t);
743 	splx(s);
744 
745 	retval[0] = ((linux_clock_t)(CONVTCK(t)));
746 	return 0;
747 }
748 
749 #undef CONVTCK
750 
751 /*
752  * Linux 'readdir' call. This code is mostly taken from the
753  * SunOS getdents call (see compat/sunos/sunos_misc.c), though
754  * an attempt has been made to keep it a little cleaner (failing
755  * miserably, because of the cruft needed if count 1 is passed).
756  *
757  * The d_off field should contain the offset of the next valid entry,
758  * but in Linux it has the offset of the entry itself. We emulate
759  * that bug here.
760  *
761  * Read in BSD-style entries, convert them, and copy them out.
762  *
763  * Note that this doesn't handle union-mounted filesystems.
764  */
765 int
766 linux_sys_getdents(l, v, retval)
767 	struct lwp *l;
768 	void *v;
769 	register_t *retval;
770 {
771 	struct linux_sys_getdents_args /* {
772 		syscallarg(int) fd;
773 		syscallarg(struct linux_dirent *) dent;
774 		syscallarg(unsigned int) count;
775 	} */ *uap = v;
776 	struct proc *p = l->l_proc;
777 	struct dirent *bdp;
778 	struct vnode *vp;
779 	caddr_t	inp, tbuf;		/* BSD-format */
780 	int len, reclen;		/* BSD-format */
781 	caddr_t outp;			/* Linux-format */
782 	int resid, linux_reclen = 0;	/* Linux-format */
783 	struct file *fp;
784 	struct uio auio;
785 	struct iovec aiov;
786 	struct linux_dirent idb;
787 	off_t off;		/* true file offset */
788 	int buflen, error, eofflag, nbytes, oldcall;
789 	struct vattr va;
790 	off_t *cookiebuf = NULL, *cookie;
791 	int ncookies;
792 
793 	/* getvnode() will use the descriptor for us */
794 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
795 		return (error);
796 
797 	if ((fp->f_flag & FREAD) == 0) {
798 		error = EBADF;
799 		goto out1;
800 	}
801 
802 	vp = (struct vnode *)fp->f_data;
803 	if (vp->v_type != VDIR) {
804 		error = EINVAL;
805 		goto out1;
806 	}
807 
808 	if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)))
809 		goto out1;
810 
811 	nbytes = SCARG(uap, count);
812 	if (nbytes == 1) {	/* emulating old, broken behaviour */
813 		nbytes = sizeof (idb);
814 		buflen = max(va.va_blocksize, nbytes);
815 		oldcall = 1;
816 	} else {
817 		buflen = min(MAXBSIZE, nbytes);
818 		if (buflen < va.va_blocksize)
819 			buflen = va.va_blocksize;
820 		oldcall = 0;
821 	}
822 	tbuf = malloc(buflen, M_TEMP, M_WAITOK);
823 
824 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
825 	off = fp->f_offset;
826 again:
827 	aiov.iov_base = tbuf;
828 	aiov.iov_len = buflen;
829 	auio.uio_iov = &aiov;
830 	auio.uio_iovcnt = 1;
831 	auio.uio_rw = UIO_READ;
832 	auio.uio_segflg = UIO_SYSSPACE;
833 	auio.uio_procp = NULL;
834 	auio.uio_resid = buflen;
835 	auio.uio_offset = off;
836 	/*
837          * First we read into the malloc'ed buffer, then
838          * we massage it into user space, one record at a time.
839          */
840 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf,
841 	    &ncookies);
842 	if (error)
843 		goto out;
844 
845 	inp = tbuf;
846 	outp = (caddr_t)SCARG(uap, dent);
847 	resid = nbytes;
848 	if ((len = buflen - auio.uio_resid) == 0)
849 		goto eof;
850 
851 	for (cookie = cookiebuf; len > 0; len -= reclen) {
852 		bdp = (struct dirent *)inp;
853 		reclen = bdp->d_reclen;
854 		if (reclen & 3)
855 			panic("linux_readdir");
856 		if (bdp->d_fileno == 0) {
857 			inp += reclen;	/* it is a hole; squish it out */
858 			if (cookie)
859 				off = *cookie++;
860 			else
861 				off += reclen;
862 			continue;
863 		}
864 		linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen);
865 		if (reclen > len || resid < linux_reclen) {
866 			/* entry too big for buffer, so just stop */
867 			outp++;
868 			break;
869 		}
870 		/*
871 		 * Massage in place to make a Linux-shaped dirent (otherwise
872 		 * we have to worry about touching user memory outside of
873 		 * the copyout() call).
874 		 */
875 		idb.d_ino = bdp->d_fileno;
876 		/*
877 		 * The old readdir() call misuses the offset and reclen fields.
878 		 */
879 		if (oldcall) {
880 			idb.d_off = (linux_off_t)linux_reclen;
881 			idb.d_reclen = (u_short)bdp->d_namlen;
882 		} else {
883 			if (sizeof (idb.d_off) <= 4 && (off >> 32) != 0) {
884 				compat_offseterr(vp, "linux_getdents");
885 				error = EINVAL;
886 				goto out;
887 			}
888 			idb.d_off = (linux_off_t)off;
889 			idb.d_reclen = (u_short)linux_reclen;
890 		}
891 		strcpy(idb.d_name, bdp->d_name);
892 		if ((error = copyout((caddr_t)&idb, outp, linux_reclen)))
893 			goto out;
894 		/* advance past this real entry */
895 		inp += reclen;
896 		if (cookie)
897 			off = *cookie++; /* each entry points to itself */
898 		else
899 			off += reclen;
900 		/* advance output past Linux-shaped entry */
901 		outp += linux_reclen;
902 		resid -= linux_reclen;
903 		if (oldcall)
904 			break;
905 	}
906 
907 	/* if we squished out the whole block, try again */
908 	if (outp == (caddr_t)SCARG(uap, dent))
909 		goto again;
910 	fp->f_offset = off;	/* update the vnode offset */
911 
912 	if (oldcall)
913 		nbytes = resid + linux_reclen;
914 
915 eof:
916 	*retval = nbytes - resid;
917 out:
918 	VOP_UNLOCK(vp, 0);
919 	if (cookiebuf)
920 		free(cookiebuf, M_TEMP);
921 	free(tbuf, M_TEMP);
922 out1:
923 	FILE_UNUSE(fp, p);
924 	return error;
925 }
926 
927 /*
928  * Even when just using registers to pass arguments to syscalls you can
929  * have 5 of them on the i386. So this newer version of select() does
930  * this.
931  */
932 int
933 linux_sys_select(l, v, retval)
934 	struct lwp *l;
935 	void *v;
936 	register_t *retval;
937 {
938 	struct linux_sys_select_args /* {
939 		syscallarg(int) nfds;
940 		syscallarg(fd_set *) readfds;
941 		syscallarg(fd_set *) writefds;
942 		syscallarg(fd_set *) exceptfds;
943 		syscallarg(struct timeval *) timeout;
944 	} */ *uap = v;
945 
946 	return linux_select1(l, retval, SCARG(uap, nfds), SCARG(uap, readfds),
947 	    SCARG(uap, writefds), SCARG(uap, exceptfds), SCARG(uap, timeout));
948 }
949 
950 /*
951  * Common code for the old and new versions of select(). A couple of
952  * things are important:
953  * 1) return the amount of time left in the 'timeout' parameter
954  * 2) select never returns ERESTART on Linux, always return EINTR
955  */
956 int
957 linux_select1(l, retval, nfds, readfds, writefds, exceptfds, timeout)
958 	struct lwp *l;
959 	register_t *retval;
960 	int nfds;
961 	fd_set *readfds, *writefds, *exceptfds;
962 	struct timeval *timeout;
963 {
964 	struct sys_select_args bsa;
965 	struct proc *p = l->l_proc;
966 	struct timeval tv0, tv1, utv, *tvp;
967 	caddr_t sg;
968 	int error;
969 
970 	SCARG(&bsa, nd) = nfds;
971 	SCARG(&bsa, in) = readfds;
972 	SCARG(&bsa, ou) = writefds;
973 	SCARG(&bsa, ex) = exceptfds;
974 	SCARG(&bsa, tv) = timeout;
975 
976 	/*
977 	 * Store current time for computation of the amount of
978 	 * time left.
979 	 */
980 	if (timeout) {
981 		if ((error = copyin(timeout, &utv, sizeof(utv))))
982 			return error;
983 		if (itimerfix(&utv)) {
984 			/*
985 			 * The timeval was invalid.  Convert it to something
986 			 * valid that will act as it does under Linux.
987 			 */
988 			sg = stackgap_init(p, 0);
989 			tvp = stackgap_alloc(p, &sg, sizeof(utv));
990 			utv.tv_sec += utv.tv_usec / 1000000;
991 			utv.tv_usec %= 1000000;
992 			if (utv.tv_usec < 0) {
993 				utv.tv_sec -= 1;
994 				utv.tv_usec += 1000000;
995 			}
996 			if (utv.tv_sec < 0)
997 				timerclear(&utv);
998 			if ((error = copyout(&utv, tvp, sizeof(utv))))
999 				return error;
1000 			SCARG(&bsa, tv) = tvp;
1001 		}
1002 		microtime(&tv0);
1003 	}
1004 
1005 	error = sys_select(l, &bsa, retval);
1006 	if (error) {
1007 		/*
1008 		 * See fs/select.c in the Linux kernel.  Without this,
1009 		 * Maelstrom doesn't work.
1010 		 */
1011 		if (error == ERESTART)
1012 			error = EINTR;
1013 		return error;
1014 	}
1015 
1016 	if (timeout) {
1017 		if (*retval) {
1018 			/*
1019 			 * Compute how much time was left of the timeout,
1020 			 * by subtracting the current time and the time
1021 			 * before we started the call, and subtracting
1022 			 * that result from the user-supplied value.
1023 			 */
1024 			microtime(&tv1);
1025 			timersub(&tv1, &tv0, &tv1);
1026 			timersub(&utv, &tv1, &utv);
1027 			if (utv.tv_sec < 0)
1028 				timerclear(&utv);
1029 		} else
1030 			timerclear(&utv);
1031 		if ((error = copyout(&utv, timeout, sizeof(utv))))
1032 			return error;
1033 	}
1034 
1035 	return 0;
1036 }
1037 
1038 /*
1039  * Get the process group of a certain process. Look it up
1040  * and return the value.
1041  */
1042 int
1043 linux_sys_getpgid(l, v, retval)
1044 	struct lwp *l;
1045 	void *v;
1046 	register_t *retval;
1047 {
1048 	struct linux_sys_getpgid_args /* {
1049 		syscallarg(int) pid;
1050 	} */ *uap = v;
1051 	struct proc *p = l->l_proc;
1052 	struct proc *targp;
1053 
1054 	if (SCARG(uap, pid) != 0 && SCARG(uap, pid) != p->p_pid) {
1055 		if ((targp = pfind(SCARG(uap, pid))) == 0)
1056 			return ESRCH;
1057 	}
1058 	else
1059 		targp = p;
1060 
1061 	retval[0] = targp->p_pgid;
1062 	return 0;
1063 }
1064 
1065 /*
1066  * Set the 'personality' (emulation mode) for the current process. Only
1067  * accept the Linux personality here (0). This call is needed because
1068  * the Linux ELF crt0 issues it in an ugly kludge to make sure that
1069  * ELF binaries run in Linux mode, not SVR4 mode.
1070  */
1071 int
1072 linux_sys_personality(l, v, retval)
1073 	struct lwp *l;
1074 	void *v;
1075 	register_t *retval;
1076 {
1077 	struct linux_sys_personality_args /* {
1078 		syscallarg(int) per;
1079 	} */ *uap = v;
1080 
1081 	if (SCARG(uap, per) != 0)
1082 		return EINVAL;
1083 	retval[0] = 0;
1084 	return 0;
1085 }
1086 
1087 #if defined(__i386__) || defined(__m68k__)
1088 /*
1089  * The calls are here because of type conversions.
1090  */
1091 int
1092 linux_sys_setreuid16(l, v, retval)
1093 	struct lwp *l;
1094 	void *v;
1095 	register_t *retval;
1096 {
1097 	struct linux_sys_setreuid16_args /* {
1098 		syscallarg(int) ruid;
1099 		syscallarg(int) euid;
1100 	} */ *uap = v;
1101 	struct sys_setreuid_args bsa;
1102 
1103 	SCARG(&bsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
1104 		(uid_t)-1 : SCARG(uap, ruid);
1105 	SCARG(&bsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
1106 		(uid_t)-1 : SCARG(uap, euid);
1107 
1108 	return sys_setreuid(l, &bsa, retval);
1109 }
1110 
1111 int
1112 linux_sys_setregid16(l, v, retval)
1113 	struct lwp *l;
1114 	void *v;
1115 	register_t *retval;
1116 {
1117 	struct linux_sys_setregid16_args /* {
1118 		syscallarg(int) rgid;
1119 		syscallarg(int) egid;
1120 	} */ *uap = v;
1121 	struct sys_setregid_args bsa;
1122 
1123 	SCARG(&bsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
1124 		(uid_t)-1 : SCARG(uap, rgid);
1125 	SCARG(&bsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
1126 		(uid_t)-1 : SCARG(uap, egid);
1127 
1128 	return sys_setregid(l, &bsa, retval);
1129 }
1130 
1131 int
1132 linux_sys_setresuid16(l, v, retval)
1133 	struct lwp *l;
1134 	void *v;
1135 	register_t *retval;
1136 {
1137 	struct linux_sys_setresuid16_args /* {
1138 		syscallarg(uid_t) ruid;
1139 		syscallarg(uid_t) euid;
1140 		syscallarg(uid_t) suid;
1141 	} */ *uap = v;
1142 	struct linux_sys_setresuid16_args lsa;
1143 
1144 	SCARG(&lsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
1145 		(uid_t)-1 : SCARG(uap, ruid);
1146 	SCARG(&lsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
1147 		(uid_t)-1 : SCARG(uap, euid);
1148 	SCARG(&lsa, suid) = ((linux_uid_t)SCARG(uap, suid) == (linux_uid_t)-1) ?
1149 		(uid_t)-1 : SCARG(uap, suid);
1150 
1151 	return linux_sys_setresuid(l, &lsa, retval);
1152 }
1153 
1154 int
1155 linux_sys_setresgid16(l, v, retval)
1156 	struct lwp *l;
1157 	void *v;
1158 	register_t *retval;
1159 {
1160 	struct linux_sys_setresgid16_args /* {
1161 		syscallarg(gid_t) rgid;
1162 		syscallarg(gid_t) egid;
1163 		syscallarg(gid_t) sgid;
1164 	} */ *uap = v;
1165 	struct linux_sys_setresgid16_args lsa;
1166 
1167 	SCARG(&lsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
1168 		(gid_t)-1 : SCARG(uap, rgid);
1169 	SCARG(&lsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
1170 		(gid_t)-1 : SCARG(uap, egid);
1171 	SCARG(&lsa, sgid) = ((linux_gid_t)SCARG(uap, sgid) == (linux_gid_t)-1) ?
1172 		(gid_t)-1 : SCARG(uap, sgid);
1173 
1174 	return linux_sys_setresgid(l, &lsa, retval);
1175 }
1176 
1177 int
1178 linux_sys_getgroups16(l, v, retval)
1179 	struct lwp *l;
1180 	void *v;
1181 	register_t *retval;
1182 {
1183 	struct linux_sys_getgroups16_args /* {
1184 		syscallarg(int) gidsetsize;
1185 		syscallarg(linux_gid_t *) gidset;
1186 	} */ *uap = v;
1187 	struct proc *p = l->l_proc;
1188 	caddr_t sg;
1189 	int n, error, i;
1190 	struct sys_getgroups_args bsa;
1191 	gid_t *bset, *kbset;
1192 	linux_gid_t *lset;
1193 	struct pcred *pc = p->p_cred;
1194 
1195 	n = SCARG(uap, gidsetsize);
1196 	if (n < 0)
1197 		return EINVAL;
1198 	error = 0;
1199 	bset = kbset = NULL;
1200 	lset = NULL;
1201 	if (n > 0) {
1202 		n = min(pc->pc_ucred->cr_ngroups, n);
1203 		sg = stackgap_init(p, 0);
1204 		bset = stackgap_alloc(p, &sg, n * sizeof (gid_t));
1205 		kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK);
1206 		lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
1207 		if (bset == NULL || kbset == NULL || lset == NULL)
1208 			return ENOMEM;
1209 		SCARG(&bsa, gidsetsize) = n;
1210 		SCARG(&bsa, gidset) = bset;
1211 		error = sys_getgroups(l, &bsa, retval);
1212 		if (error != 0)
1213 			goto out;
1214 		error = copyin(bset, kbset, n * sizeof (gid_t));
1215 		if (error != 0)
1216 			goto out;
1217 		for (i = 0; i < n; i++)
1218 			lset[i] = (linux_gid_t)kbset[i];
1219 		error = copyout(lset, SCARG(uap, gidset),
1220 		    n * sizeof (linux_gid_t));
1221 	} else
1222 		*retval = pc->pc_ucred->cr_ngroups;
1223 out:
1224 	if (kbset != NULL)
1225 		free(kbset, M_TEMP);
1226 	if (lset != NULL)
1227 		free(lset, M_TEMP);
1228 	return error;
1229 }
1230 
1231 int
1232 linux_sys_setgroups16(l, v, retval)
1233 	struct lwp *l;
1234 	void *v;
1235 	register_t *retval;
1236 {
1237 	struct linux_sys_setgroups16_args /* {
1238 		syscallarg(int) gidsetsize;
1239 		syscallarg(linux_gid_t *) gidset;
1240 	} */ *uap = v;
1241 	struct proc *p = l->l_proc;
1242 	caddr_t sg;
1243 	int n;
1244 	int error, i;
1245 	struct sys_setgroups_args bsa;
1246 	gid_t *bset, *kbset;
1247 	linux_gid_t *lset;
1248 
1249 	n = SCARG(uap, gidsetsize);
1250 	if (n < 0 || n > NGROUPS)
1251 		return EINVAL;
1252 	sg = stackgap_init(p, 0);
1253 	bset = stackgap_alloc(p, &sg, n * sizeof (gid_t));
1254 	lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
1255 	kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK);
1256 	if (lset == NULL || bset == NULL)
1257 		return ENOMEM;
1258 	error = copyin(SCARG(uap, gidset), lset, n * sizeof (linux_gid_t));
1259 	if (error != 0)
1260 		goto out;
1261 	for (i = 0; i < n; i++)
1262 		kbset[i] = (gid_t)lset[i];
1263 	error = copyout(kbset, bset, n * sizeof (gid_t));
1264 	if (error != 0)
1265 		goto out;
1266 	SCARG(&bsa, gidsetsize) = n;
1267 	SCARG(&bsa, gidset) = bset;
1268 	error = sys_setgroups(l, &bsa, retval);
1269 
1270 out:
1271 	if (lset != NULL)
1272 		free(lset, M_TEMP);
1273 	if (kbset != NULL)
1274 		free(kbset, M_TEMP);
1275 
1276 	return error;
1277 }
1278 
1279 #endif /* __i386__ || __m68k__ || __amd64__ */
1280 
1281 /*
1282  * We have nonexistent fsuid equal to uid.
1283  * If modification is requested, refuse.
1284  */
1285 int
1286 linux_sys_setfsuid(l, v, retval)
1287 	 struct lwp *l;
1288 	 void *v;
1289 	 register_t *retval;
1290 {
1291 	 struct linux_sys_setfsuid_args /* {
1292 		 syscallarg(uid_t) uid;
1293 	 } */ *uap = v;
1294 	 struct proc *p = l->l_proc;
1295 	 uid_t uid;
1296 
1297 	 uid = SCARG(uap, uid);
1298 	 if (p->p_cred->p_ruid != uid)
1299 		 return sys_nosys(l, v, retval);
1300 	 else
1301 		 return (0);
1302 }
1303 
1304 /* XXX XXX XXX */
1305 #ifndef alpha
1306 int
1307 linux_sys_getfsuid(l, v, retval)
1308 	struct lwp *l;
1309 	void *v;
1310 	register_t *retval;
1311 {
1312 	return sys_getuid(l, v, retval);
1313 }
1314 #endif
1315 
1316 int
1317 linux_sys_setresuid(l, v, retval)
1318 	struct lwp *l;
1319 	void *v;
1320 	register_t *retval;
1321 {
1322 	struct linux_sys_setresuid_args /* {
1323 		syscallarg(uid_t) ruid;
1324 		syscallarg(uid_t) euid;
1325 		syscallarg(uid_t) suid;
1326 	} */ *uap = v;
1327 
1328 	/*
1329 	 * Note: These checks are a little different than the NetBSD
1330 	 * setreuid(2) call performs.  This precisely follows the
1331 	 * behavior of the Linux kernel.
1332 	 */
1333 
1334 	return do_setresuid(l, SCARG(uap, ruid), SCARG(uap, euid),
1335 			    SCARG(uap, suid),
1336 			    ID_R_EQ_R | ID_R_EQ_E | ID_R_EQ_S |
1337 			    ID_E_EQ_R | ID_E_EQ_E | ID_E_EQ_S |
1338 			    ID_S_EQ_R | ID_S_EQ_E | ID_S_EQ_S );
1339 }
1340 
1341 int
1342 linux_sys_getresuid(l, v, retval)
1343 	struct lwp *l;
1344 	void *v;
1345 	register_t *retval;
1346 {
1347 	struct linux_sys_getresuid_args /* {
1348 		syscallarg(uid_t *) ruid;
1349 		syscallarg(uid_t *) euid;
1350 		syscallarg(uid_t *) suid;
1351 	} */ *uap = v;
1352 	struct proc *p = l->l_proc;
1353 	struct pcred *pc = p->p_cred;
1354 	int error;
1355 
1356 	/*
1357 	 * Linux copies these values out to userspace like so:
1358 	 *
1359 	 *	1. Copy out ruid.
1360 	 *	2. If that succeeds, copy out euid.
1361 	 *	3. If both of those succeed, copy out suid.
1362 	 */
1363 	if ((error = copyout(&pc->p_ruid, SCARG(uap, ruid),
1364 			     sizeof(uid_t))) != 0)
1365 		return (error);
1366 
1367 	if ((error = copyout(&pc->pc_ucred->cr_uid, SCARG(uap, euid),
1368 			     sizeof(uid_t))) != 0)
1369 		return (error);
1370 
1371 	return (copyout(&pc->p_svuid, SCARG(uap, suid), sizeof(uid_t)));
1372 }
1373 
1374 int
1375 linux_sys_ptrace(l, v, retval)
1376 	struct lwp *l;
1377 	void *v;
1378 	register_t *retval;
1379 {
1380 	struct linux_sys_ptrace_args /* {
1381 		i386, m68k, powerpc: T=int
1382 		alpha, amd64: T=long
1383 		syscallarg(T) request;
1384 		syscallarg(T) pid;
1385 		syscallarg(T) addr;
1386 		syscallarg(T) data;
1387 	} */ *uap = v;
1388 	const int *ptr;
1389 	int request;
1390 	int error;
1391 
1392 	ptr = linux_ptrace_request_map;
1393 	request = SCARG(uap, request);
1394 	while (*ptr != -1)
1395 		if (*ptr++ == request) {
1396 			struct sys_ptrace_args pta;
1397 
1398 			SCARG(&pta, req) = *ptr;
1399 			SCARG(&pta, pid) = SCARG(uap, pid);
1400 			SCARG(&pta, addr) = (caddr_t)SCARG(uap, addr);
1401 			SCARG(&pta, data) = SCARG(uap, data);
1402 
1403 			/*
1404 			 * Linux ptrace(PTRACE_CONT, pid, 0, 0) means actually
1405 			 * to continue where the process left off previously.
1406 			 * The same thing is achieved by addr == (caddr_t) 1
1407 			 * on NetBSD, so rewrite 'addr' appropriately.
1408 			 */
1409 			if (request == LINUX_PTRACE_CONT && SCARG(uap, addr)==0)
1410 				SCARG(&pta, addr) = (caddr_t) 1;
1411 
1412 			error = sys_ptrace(l, &pta, retval);
1413 			if (error)
1414 				return error;
1415 			switch (request) {
1416 			case LINUX_PTRACE_PEEKTEXT:
1417 			case LINUX_PTRACE_PEEKDATA:
1418 				error = copyout (retval,
1419 				    (caddr_t)SCARG(uap, data),
1420 				    sizeof *retval);
1421 				*retval = SCARG(uap, data);
1422 				break;
1423 			default:
1424 				break;
1425 			}
1426 			return error;
1427 		}
1428 		else
1429 			ptr++;
1430 
1431 	return LINUX_SYS_PTRACE_ARCH(l, uap, retval);
1432 }
1433 
1434 int
1435 linux_sys_reboot(struct lwp *l, void *v, register_t *retval)
1436 {
1437 	struct linux_sys_reboot_args /* {
1438 		syscallarg(int) magic1;
1439 		syscallarg(int) magic2;
1440 		syscallarg(int) cmd;
1441 		syscallarg(void *) arg;
1442 	} */ *uap = v;
1443 	struct sys_reboot_args /* {
1444 		syscallarg(int) opt;
1445 		syscallarg(char *) bootstr;
1446 	} */ sra;
1447 	struct proc *p = l->l_proc;
1448 	int error;
1449 
1450 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1451 		return(error);
1452 
1453 	if (SCARG(uap, magic1) != LINUX_REBOOT_MAGIC1)
1454 		return(EINVAL);
1455 	if (SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2 &&
1456 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2A &&
1457 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2B)
1458 		return(EINVAL);
1459 
1460 	switch (SCARG(uap, cmd)) {
1461 	case LINUX_REBOOT_CMD_RESTART:
1462 		SCARG(&sra, opt) = RB_AUTOBOOT;
1463 		break;
1464 	case LINUX_REBOOT_CMD_HALT:
1465 		SCARG(&sra, opt) = RB_HALT;
1466 		break;
1467 	case LINUX_REBOOT_CMD_POWER_OFF:
1468 		SCARG(&sra, opt) = RB_HALT|RB_POWERDOWN;
1469 		break;
1470 	case LINUX_REBOOT_CMD_RESTART2:
1471 		/* Reboot with an argument. */
1472 		SCARG(&sra, opt) = RB_AUTOBOOT|RB_STRING;
1473 		SCARG(&sra, bootstr) = SCARG(uap, arg);
1474 		break;
1475 	case LINUX_REBOOT_CMD_CAD_ON:
1476 		return(EINVAL);	/* We don't implement ctrl-alt-delete */
1477 	case LINUX_REBOOT_CMD_CAD_OFF:
1478 		return(0);
1479 	default:
1480 		return(EINVAL);
1481 	}
1482 
1483 	return(sys_reboot(l, &sra, retval));
1484 }
1485 
1486 /*
1487  * Copy of compat_12_sys_swapon().
1488  */
1489 int
1490 linux_sys_swapon(l, v, retval)
1491 	struct lwp *l;
1492 	void *v;
1493 	register_t *retval;
1494 {
1495 	struct sys_swapctl_args ua;
1496 	struct linux_sys_swapon_args /* {
1497 		syscallarg(const char *) name;
1498 	} */ *uap = v;
1499 
1500 	SCARG(&ua, cmd) = SWAP_ON;
1501 	SCARG(&ua, arg) = (void *)__UNCONST(SCARG(uap, name));
1502 	SCARG(&ua, misc) = 0;	/* priority */
1503 	return (sys_swapctl(l, &ua, retval));
1504 }
1505 
1506 /*
1507  * Stop swapping to the file or block device specified by path.
1508  */
1509 int
1510 linux_sys_swapoff(l, v, retval)
1511 	struct lwp *l;
1512 	void *v;
1513 	register_t *retval;
1514 {
1515 	struct sys_swapctl_args ua;
1516 	struct linux_sys_swapoff_args /* {
1517 		syscallarg(const char *) path;
1518 	} */ *uap = v;
1519 
1520 	SCARG(&ua, cmd) = SWAP_OFF;
1521 	SCARG(&ua, arg) = __UNCONST(SCARG(uap, path)); /*XXXUNCONST*/
1522 	return (sys_swapctl(l, &ua, retval));
1523 }
1524 
1525 /*
1526  * Copy of compat_09_sys_setdomainname()
1527  */
1528 /* ARGSUSED */
1529 int
1530 linux_sys_setdomainname(l, v, retval)
1531 	struct lwp *l;
1532 	void *v;
1533 	register_t *retval;
1534 {
1535 	struct linux_sys_setdomainname_args /* {
1536 		syscallarg(char *) domainname;
1537 		syscallarg(int) len;
1538 	} */ *uap = v;
1539 	int name[2];
1540 
1541 	name[0] = CTL_KERN;
1542 	name[1] = KERN_DOMAINNAME;
1543 	return (old_sysctl(&name[0], 2, 0, 0, SCARG(uap, domainname),
1544 			    SCARG(uap, len), l));
1545 }
1546 
1547 /*
1548  * sysinfo()
1549  */
1550 /* ARGSUSED */
1551 int
1552 linux_sys_sysinfo(l, v, retval)
1553 	struct lwp *l;
1554 	void *v;
1555 	register_t *retval;
1556 {
1557 	struct linux_sys_sysinfo_args /* {
1558 		syscallarg(struct linux_sysinfo *) arg;
1559 	} */ *uap = v;
1560 	struct linux_sysinfo si;
1561 	struct loadavg *la;
1562 
1563 	si.uptime = time.tv_sec - boottime.tv_sec;
1564 	la = &averunnable;
1565 	si.loads[0] = la->ldavg[0] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1566 	si.loads[1] = la->ldavg[1] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1567 	si.loads[2] = la->ldavg[2] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1568 	si.totalram = ctob(physmem);
1569 	si.freeram = uvmexp.free * uvmexp.pagesize;
1570 	si.sharedram = 0;	/* XXX */
1571 	si.bufferram = uvmexp.filepages * uvmexp.pagesize;
1572 	si.totalswap = uvmexp.swpages * uvmexp.pagesize;
1573 	si.freeswap = (uvmexp.swpages - uvmexp.swpginuse) * uvmexp.pagesize;
1574 	si.procs = nprocs;
1575 
1576 	/* The following are only present in newer Linux kernels. */
1577 	si.totalbig = 0;
1578 	si.freebig = 0;
1579 	si.mem_unit = 1;
1580 
1581 	return (copyout(&si, SCARG(uap, arg), sizeof si));
1582 }
1583 
1584 #define bsd_to_linux_rlimit1(l, b, f) \
1585     (l)->f = ((b)->f == RLIM_INFINITY || \
1586 	     ((b)->f & 0xffffffff00000000ULL) != 0) ? \
1587     LINUX_RLIM_INFINITY : (int32_t)(b)->f
1588 #define bsd_to_linux_rlimit(l, b) \
1589     bsd_to_linux_rlimit1(l, b, rlim_cur); \
1590     bsd_to_linux_rlimit1(l, b, rlim_max)
1591 
1592 #define linux_to_bsd_rlimit1(b, l, f) \
1593     (b)->f = (l)->f == LINUX_RLIM_INFINITY ? RLIM_INFINITY : (l)->f
1594 #define linux_to_bsd_rlimit(b, l) \
1595     linux_to_bsd_rlimit1(b, l, rlim_cur); \
1596     linux_to_bsd_rlimit1(b, l, rlim_max)
1597 
1598 static int
1599 linux_to_bsd_limit(lim)
1600 	int lim;
1601 {
1602 	switch (lim) {
1603 	case LINUX_RLIMIT_CPU:
1604 		return RLIMIT_CPU;
1605 	case LINUX_RLIMIT_FSIZE:
1606 		return RLIMIT_FSIZE;
1607 	case LINUX_RLIMIT_DATA:
1608 		return RLIMIT_DATA;
1609 	case LINUX_RLIMIT_STACK:
1610 		return RLIMIT_STACK;
1611 	case LINUX_RLIMIT_CORE:
1612 		return RLIMIT_CORE;
1613 	case LINUX_RLIMIT_RSS:
1614 		return RLIMIT_RSS;
1615 	case LINUX_RLIMIT_NPROC:
1616 		return RLIMIT_NPROC;
1617 	case LINUX_RLIMIT_NOFILE:
1618 		return RLIMIT_NOFILE;
1619 	case LINUX_RLIMIT_MEMLOCK:
1620 		return RLIMIT_MEMLOCK;
1621 	case LINUX_RLIMIT_AS:
1622 	case LINUX_RLIMIT_LOCKS:
1623 		return -EOPNOTSUPP;
1624 	default:
1625 		return -EINVAL;
1626 	}
1627 }
1628 
1629 
1630 int
1631 linux_sys_getrlimit(l, v, retval)
1632 	struct lwp *l;
1633 	void *v;
1634 	register_t *retval;
1635 {
1636 	struct linux_sys_getrlimit_args /* {
1637 		syscallarg(int) which;
1638 		syscallarg(struct orlimit *) rlp;
1639 	} */ *uap = v;
1640 	struct proc *p = l->l_proc;
1641 	caddr_t sg = stackgap_init(p, 0);
1642 	struct sys_getrlimit_args ap;
1643 	struct rlimit rl;
1644 	struct orlimit orl;
1645 	int error;
1646 
1647 	SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which));
1648 	if ((error = SCARG(&ap, which)) < 0)
1649 		return -error;
1650 	SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl);
1651 	if ((error = sys_getrlimit(l, &ap, retval)) != 0)
1652 		return error;
1653 	if ((error = copyin(SCARG(&ap, rlp), &rl, sizeof(rl))) != 0)
1654 		return error;
1655 	bsd_to_linux_rlimit(&orl, &rl);
1656 	return copyout(&orl, SCARG(uap, rlp), sizeof(orl));
1657 }
1658 
1659 int
1660 linux_sys_setrlimit(l, v, retval)
1661 	struct lwp *l;
1662 	void *v;
1663 	register_t *retval;
1664 {
1665 	struct linux_sys_setrlimit_args /* {
1666 		syscallarg(int) which;
1667 		syscallarg(struct orlimit *) rlp;
1668 	} */ *uap = v;
1669 	struct proc *p = l->l_proc;
1670 	caddr_t sg = stackgap_init(p, 0);
1671 	struct sys_getrlimit_args ap;
1672 	struct rlimit rl;
1673 	struct orlimit orl;
1674 	int error;
1675 
1676 	SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which));
1677 	SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl);
1678 	if ((error = SCARG(&ap, which)) < 0)
1679 		return -error;
1680 	if ((error = copyin(SCARG(uap, rlp), &orl, sizeof(orl))) != 0)
1681 		return error;
1682 	linux_to_bsd_rlimit(&rl, &orl);
1683 	if ((error = copyout(&rl, SCARG(&ap, rlp), sizeof(rl))) != 0)
1684 		return error;
1685 	return sys_setrlimit(l, &ap, retval);
1686 }
1687 
1688 #if !defined(__mips__) && !defined(__amd64__)
1689 /* XXX: this doesn't look 100% common, at least mips doesn't have it */
1690 int
1691 linux_sys_ugetrlimit(l, v, retval)
1692 	struct lwp *l;
1693 	void *v;
1694 	register_t *retval;
1695 {
1696 	return linux_sys_getrlimit(l, v, retval);
1697 }
1698 #endif
1699 
1700 /*
1701  * This gets called for unsupported syscalls. The difference to sys_nosys()
1702  * is that process does not get SIGSYS, the call just returns with ENOSYS.
1703  * This is the way Linux does it and glibc depends on this behaviour.
1704  */
1705 int
1706 linux_sys_nosys(l, v, retval)
1707 	struct lwp *l;
1708 	void *v;
1709 	register_t *retval;
1710 {
1711 	return (ENOSYS);
1712 }
1713