xref: /netbsd-src/sys/compat/linux/common/linux_misc.c (revision 01d824d7cbb02f34569880163e54d2663275b865)
1 /*	$NetBSD: linux_misc.c,v 1.107 2002/04/10 18:19:34 christos Exp $	*/
2 
3 /*-
4  * Copyright (c) 1995, 1998, 1999 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Frank van der Linden and Eric Haszlakiewicz; by Jason R. Thorpe
9  * of the Numerical Aerospace Simulation Facility, NASA Ames Research Center.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 3. All advertising materials mentioning features or use of this software
20  *    must display the following acknowledgement:
21  *	This product includes software developed by the NetBSD
22  *	Foundation, Inc. and its contributors.
23  * 4. Neither the name of The NetBSD Foundation nor the names of its
24  *    contributors may be used to endorse or promote products derived
25  *    from this software without specific prior written permission.
26  *
27  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
28  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
29  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
30  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
31  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
32  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
33  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
34  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
35  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
36  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
37  * POSSIBILITY OF SUCH DAMAGE.
38  */
39 
40 /*
41  * Linux compatibility module. Try to deal with various Linux system calls.
42  */
43 
44 /*
45  * These functions have been moved to multiarch to allow
46  * selection of which machines include them to be
47  * determined by the individual files.linux_<arch> files.
48  *
49  * Function in multiarch:
50  *	linux_sys_break			: linux_break.c
51  *	linux_sys_alarm			: linux_misc_notalpha.c
52  *	linux_sys_getresgid		: linux_misc_notalpha.c
53  *	linux_sys_nice			: linux_misc_notalpha.c
54  *	linux_sys_readdir		: linux_misc_notalpha.c
55  *	linux_sys_setresgid		: linux_misc_notalpha.c
56  *	linux_sys_time			: linux_misc_notalpha.c
57  *	linux_sys_utime			: linux_misc_notalpha.c
58  *	linux_sys_waitpid		: linux_misc_notalpha.c
59  *	linux_sys_old_mmap		: linux_oldmmap.c
60  *	linux_sys_oldolduname		: linux_oldolduname.c
61  *	linux_sys_oldselect		: linux_oldselect.c
62  *	linux_sys_olduname		: linux_olduname.c
63  *	linux_sys_pipe			: linux_pipe.c
64  */
65 
66 #include <sys/cdefs.h>
67 __KERNEL_RCSID(0, "$NetBSD: linux_misc.c,v 1.107 2002/04/10 18:19:34 christos 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/syscallargs.h>
103 
104 #include <compat/linux/common/linux_types.h>
105 #include <compat/linux/common/linux_signal.h>
106 
107 #include <compat/linux/linux_syscallargs.h>
108 
109 #include <compat/linux/common/linux_fcntl.h>
110 #include <compat/linux/common/linux_mmap.h>
111 #include <compat/linux/common/linux_dirent.h>
112 #include <compat/linux/common/linux_util.h>
113 #include <compat/linux/common/linux_misc.h>
114 #include <compat/linux/common/linux_ptrace.h>
115 #include <compat/linux/common/linux_reboot.h>
116 #include <compat/linux/common/linux_emuldata.h>
117 
118 const int linux_ptrace_request_map[] = {
119 	LINUX_PTRACE_TRACEME,	PT_TRACE_ME,
120 	LINUX_PTRACE_PEEKTEXT,	PT_READ_I,
121 	LINUX_PTRACE_PEEKDATA,	PT_READ_D,
122 	LINUX_PTRACE_POKETEXT,	PT_WRITE_I,
123 	LINUX_PTRACE_POKEDATA,	PT_WRITE_D,
124 	LINUX_PTRACE_CONT,	PT_CONTINUE,
125 	LINUX_PTRACE_KILL,	PT_KILL,
126 	LINUX_PTRACE_ATTACH,	PT_ATTACH,
127 	LINUX_PTRACE_DETACH,	PT_DETACH,
128 #ifdef PT_STEP
129 	LINUX_PTRACE_SINGLESTEP,	PT_STEP,
130 #endif
131 	-1
132 };
133 
134 const static struct mnttypes {
135 	char *bsd;
136 	int linux;
137 } fstypes[] = {
138 	{ MOUNT_FFS,		LINUX_DEFAULT_SUPER_MAGIC	},
139 	{ MOUNT_NFS,		LINUX_NFS_SUPER_MAGIC 		},
140 	{ MOUNT_MFS,		LINUX_DEFAULT_SUPER_MAGIC	},
141 	{ MOUNT_MSDOS,		LINUX_MSDOS_SUPER_MAGIC		},
142 	{ MOUNT_LFS,		LINUX_DEFAULT_SUPER_MAGIC	},
143 	{ MOUNT_FDESC,		LINUX_DEFAULT_SUPER_MAGIC	},
144 	{ MOUNT_PORTAL,		LINUX_DEFAULT_SUPER_MAGIC	},
145 	{ MOUNT_NULL,		LINUX_DEFAULT_SUPER_MAGIC	},
146 	{ MOUNT_OVERLAY,	LINUX_DEFAULT_SUPER_MAGIC	},
147 	{ MOUNT_UMAP,		LINUX_DEFAULT_SUPER_MAGIC	},
148 	{ MOUNT_KERNFS,		LINUX_DEFAULT_SUPER_MAGIC	},
149 	{ MOUNT_PROCFS,		LINUX_PROC_SUPER_MAGIC		},
150 	{ MOUNT_AFS,		LINUX_DEFAULT_SUPER_MAGIC	},
151 	{ MOUNT_CD9660,		LINUX_ISOFS_SUPER_MAGIC		},
152 	{ MOUNT_UNION,		LINUX_DEFAULT_SUPER_MAGIC	},
153 	{ MOUNT_ADOSFS,		LINUX_ADFS_SUPER_MAGIC		},
154 	{ MOUNT_EXT2FS,		LINUX_EXT2_SUPER_MAGIC		},
155 	{ MOUNT_CFS,		LINUX_DEFAULT_SUPER_MAGIC	},
156 	{ MOUNT_CODA,		LINUX_CODA_SUPER_MAGIC		},
157 	{ MOUNT_FILECORE,	LINUX_DEFAULT_SUPER_MAGIC	},
158 	{ MOUNT_NTFS,		LINUX_DEFAULT_SUPER_MAGIC	},
159 	{ MOUNT_SMBFS,		LINUX_SMB_SUPER_MAGIC		}
160 };
161 #define FSTYPESSIZE (sizeof(fstypes) / sizeof(fstypes[0]))
162 
163 #ifdef DEBUG_LINUX
164 #define DPRINTF(a)	uprintf a
165 #else
166 #define DPRINTF(a)
167 #endif
168 
169 /* Local linux_misc.c functions: */
170 static void bsd_to_linux_statfs __P((struct statfs *, struct linux_statfs *));
171 static int linux_to_bsd_limit __P((int));
172 
173 /*
174  * The information on a terminated (or stopped) process needs
175  * to be converted in order for Linux binaries to get a valid signal
176  * number out of it.
177  */
178 void
179 bsd_to_linux_wstat(st)
180 	int *st;
181 {
182 
183 	int sig;
184 
185 	if (WIFSIGNALED(*st)) {
186 		sig = WTERMSIG(*st);
187 		if (sig >= 0 && sig < NSIG)
188 			*st= (*st& ~0177) | native_to_linux_signo[sig];
189 	} else if (WIFSTOPPED(*st)) {
190 		sig = WSTOPSIG(*st);
191 		if (sig >= 0 && sig < NSIG)
192 			*st = (*st & ~0xff00) |
193 			    (native_to_linux_signo[sig] << 8);
194 	}
195 }
196 
197 /*
198  * This is very much the same as waitpid()
199  */
200 int
201 linux_sys_wait4(p, v, retval)
202 	struct proc *p;
203 	void *v;
204 	register_t *retval;
205 {
206 	struct linux_sys_wait4_args /* {
207 		syscallarg(int) pid;
208 		syscallarg(int *) status;
209 		syscallarg(int) options;
210 		syscallarg(struct rusage *) rusage;
211 	} */ *uap = v;
212 	struct sys_wait4_args w4a;
213 	int error, *status, tstat, options, linux_options;
214 	caddr_t sg;
215 
216 	if (SCARG(uap, status) != NULL) {
217 		sg = stackgap_init(p, 0);
218 		status = (int *) stackgap_alloc(p, &sg, sizeof *status);
219 	} else
220 		status = NULL;
221 
222 	linux_options = SCARG(uap, options);
223 	options = 0;
224 	if (linux_options &
225 	    ~(LINUX_WAIT4_WNOHANG|LINUX_WAIT4_WUNTRACED|LINUX_WAIT4_WALL|
226 	      LINUX_WAIT4_WCLONE))
227 		return (EINVAL);
228 
229 	if (linux_options & LINUX_WAIT4_WNOHANG)
230 		options |= WNOHANG;
231 	if (linux_options & LINUX_WAIT4_WUNTRACED)
232 		options |= WUNTRACED;
233 	if (linux_options & LINUX_WAIT4_WALL)
234 		options |= WALLSIG;
235 	if (linux_options & LINUX_WAIT4_WCLONE)
236 		options |= WALTSIG;
237 
238 	SCARG(&w4a, pid) = SCARG(uap, pid);
239 	SCARG(&w4a, status) = status;
240 	SCARG(&w4a, options) = options;
241 	SCARG(&w4a, rusage) = SCARG(uap, rusage);
242 
243 	if ((error = sys_wait4(p, &w4a, retval)))
244 		return error;
245 
246 	sigdelset(&p->p_sigctx.ps_siglist, SIGCHLD);
247 
248 	if (status != NULL) {
249 		if ((error = copyin(status, &tstat, sizeof tstat)))
250 			return error;
251 
252 		bsd_to_linux_wstat(&tstat);
253 		return copyout(&tstat, SCARG(uap, status), sizeof tstat);
254 	}
255 
256 	return 0;
257 }
258 
259 /*
260  * Linux brk(2). The check if the new address is >= the old one is
261  * done in the kernel in Linux. NetBSD does it in the library.
262  */
263 int
264 linux_sys_brk(p, v, retval)
265 	struct proc *p;
266 	void *v;
267 	register_t *retval;
268 {
269 	struct linux_sys_brk_args /* {
270 		syscallarg(char *) nsize;
271 	} */ *uap = v;
272 	char *nbrk = SCARG(uap, nsize);
273 	struct sys_obreak_args oba;
274 	struct vmspace *vm = p->p_vmspace;
275 	struct linux_emuldata *ed = (struct linux_emuldata*)p->p_emuldata;
276 
277 	SCARG(&oba, nsize) = nbrk;
278 
279 	if ((caddr_t) nbrk > vm->vm_daddr && sys_obreak(p, &oba, retval) == 0)
280 		ed->p_break = (char*)nbrk;
281 	else
282 		nbrk = ed->p_break;
283 
284 	retval[0] = (register_t)nbrk;
285 
286 	return 0;
287 }
288 
289 /*
290  * Convert BSD statfs structure to Linux statfs structure.
291  * The Linux structure has less fields, and it also wants
292  * the length of a name in a dir entry in a field, which
293  * we fake (probably the wrong way).
294  */
295 static void
296 bsd_to_linux_statfs(bsp, lsp)
297 	struct statfs *bsp;
298 	struct linux_statfs *lsp;
299 {
300 	int i;
301 
302 	for (i = 0; i < FSTYPESSIZE; i++)
303 		if (strcmp(bsp->f_fstypename, fstypes[i].bsd) == 0)
304 			break;
305 
306 	if (i == FSTYPESSIZE) {
307 		DPRINTF(("unhandled fstype in linux emulation: %s\n",
308 		    bsp->f_fstypename));
309 		lsp->l_ftype = LINUX_DEFAULT_SUPER_MAGIC;
310 	} else {
311 		lsp->l_ftype = fstypes[i].linux;
312 	}
313 
314 	lsp->l_fbsize = bsp->f_bsize;
315 	lsp->l_fblocks = bsp->f_blocks;
316 	lsp->l_fbfree = bsp->f_bfree;
317 	lsp->l_fbavail = bsp->f_bavail;
318 	lsp->l_ffiles = bsp->f_files;
319 	lsp->l_fffree = bsp->f_ffree;
320 	/* Linux sets the fsid to 0..., we don't */
321 	lsp->l_ffsid.val[0] = bsp->f_fsid.val[0];
322 	lsp->l_ffsid.val[1] = bsp->f_fsid.val[1];
323 	lsp->l_fnamelen = MAXNAMLEN;	/* XXX */
324 	(void)memset(lsp->l_fspare, 0, sizeof(lsp->l_fspare));
325 }
326 
327 /*
328  * Implement the fs stat functions. Straightforward.
329  */
330 int
331 linux_sys_statfs(p, v, retval)
332 	struct proc *p;
333 	void *v;
334 	register_t *retval;
335 {
336 	struct linux_sys_statfs_args /* {
337 		syscallarg(const char *) path;
338 		syscallarg(struct linux_statfs *) sp;
339 	} */ *uap = v;
340 	struct statfs btmp, *bsp;
341 	struct linux_statfs ltmp;
342 	struct sys_statfs_args bsa;
343 	caddr_t sg;
344 	int error;
345 
346 	sg = stackgap_init(p, 0);
347 	bsp = (struct statfs *) stackgap_alloc(p, &sg, sizeof (struct statfs));
348 
349 	CHECK_ALT_EXIST(p, &sg, SCARG(uap, path));
350 
351 	SCARG(&bsa, path) = SCARG(uap, path);
352 	SCARG(&bsa, buf) = bsp;
353 
354 	if ((error = sys_statfs(p, &bsa, retval)))
355 		return error;
356 
357 	if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp)))
358 		return error;
359 
360 	bsd_to_linux_statfs(&btmp, &ltmp);
361 
362 	return copyout((caddr_t) &ltmp, (caddr_t) SCARG(uap, sp), sizeof ltmp);
363 }
364 
365 int
366 linux_sys_fstatfs(p, v, retval)
367 	struct proc *p;
368 	void *v;
369 	register_t *retval;
370 {
371 	struct linux_sys_fstatfs_args /* {
372 		syscallarg(int) fd;
373 		syscallarg(struct linux_statfs *) sp;
374 	} */ *uap = v;
375 	struct statfs btmp, *bsp;
376 	struct linux_statfs ltmp;
377 	struct sys_fstatfs_args bsa;
378 	caddr_t sg;
379 	int error;
380 
381 	sg = stackgap_init(p, 0);
382 	bsp = (struct statfs *) stackgap_alloc(p, &sg, sizeof (struct statfs));
383 
384 	SCARG(&bsa, fd) = SCARG(uap, fd);
385 	SCARG(&bsa, buf) = bsp;
386 
387 	if ((error = sys_fstatfs(p, &bsa, retval)))
388 		return error;
389 
390 	if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp)))
391 		return error;
392 
393 	bsd_to_linux_statfs(&btmp, &ltmp);
394 
395 	return copyout((caddr_t) &ltmp, (caddr_t) SCARG(uap, sp), sizeof ltmp);
396 }
397 
398 /*
399  * uname(). Just copy the info from the various strings stored in the
400  * kernel, and put it in the Linux utsname structure. That structure
401  * is almost the same as the NetBSD one, only it has fields 65 characters
402  * long, and an extra domainname field.
403  */
404 int
405 linux_sys_uname(p, v, retval)
406 	struct proc *p;
407 	void *v;
408 	register_t *retval;
409 {
410 	struct linux_sys_uname_args /* {
411 		syscallarg(struct linux_utsname *) up;
412 	} */ *uap = v;
413 	struct linux_utsname luts;
414 
415 	strncpy(luts.l_sysname, linux_sysname, sizeof(luts.l_sysname));
416 	strncpy(luts.l_nodename, hostname, sizeof(luts.l_nodename));
417 	strncpy(luts.l_release, linux_release, sizeof(luts.l_release));
418 	strncpy(luts.l_version, linux_version, sizeof(luts.l_version));
419 	strncpy(luts.l_machine, machine, sizeof(luts.l_machine));
420 	strncpy(luts.l_domainname, domainname, sizeof(luts.l_domainname));
421 
422 	return copyout(&luts, SCARG(uap, up), sizeof(luts));
423 }
424 
425 /* Used directly on: alpha, mips, ppc, sparc, sparc64 */
426 /* Used indirectly on: arm, i386, m68k */
427 
428 /*
429  * New type Linux mmap call.
430  * Only called directly on machines with >= 6 free regs.
431  */
432 int
433 linux_sys_mmap(p, v, retval)
434 	struct proc *p;
435 	void *v;
436 	register_t *retval;
437 {
438 	struct linux_sys_mmap_args /* {
439 		syscallarg(unsigned long) addr;
440 		syscallarg(size_t) len;
441 		syscallarg(int) prot;
442 		syscallarg(int) flags;
443 		syscallarg(int) fd;
444 		syscallarg(linux_off_t) offset;
445 	} */ *uap = v;
446 	struct sys_mmap_args cma;
447 	int flags, fl = SCARG(uap, flags);
448 
449 	flags = 0;
450 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_SHARED, MAP_SHARED);
451 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_PRIVATE, MAP_PRIVATE);
452 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_FIXED, MAP_FIXED);
453 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_ANON, MAP_ANON);
454 	/* XXX XAX ERH: Any other flags here?  There are more defined... */
455 
456 	SCARG(&cma, addr) = (void *)SCARG(uap, addr);
457 	SCARG(&cma, len) = SCARG(uap, len);
458 	SCARG(&cma, prot) = SCARG(uap, prot);
459 	if (SCARG(&cma, prot) & VM_PROT_WRITE) /* XXX */
460 		SCARG(&cma, prot) |= VM_PROT_READ;
461 	SCARG(&cma, flags) = flags;
462 	SCARG(&cma, fd) = flags & MAP_ANON ? -1 : SCARG(uap, fd);
463 	SCARG(&cma, pad) = 0;
464 	SCARG(&cma, pos) = (off_t)SCARG(uap, offset);
465 
466 	return sys_mmap(p, &cma, retval);
467 }
468 
469 int
470 linux_sys_mremap(p, v, retval)
471 	struct proc *p;
472 	void *v;
473 	register_t *retval;
474 {
475 	struct linux_sys_mremap_args /* {
476 		syscallarg(void *) old_address;
477 		syscallarg(size_t) old_size;
478 		syscallarg(size_t) new_size;
479 		syscallarg(u_long) flags;
480 	} */ *uap = v;
481 	struct sys_munmap_args mua;
482 	size_t old_size, new_size;
483 	int error;
484 
485 	old_size = round_page(SCARG(uap, old_size));
486 	new_size = round_page(SCARG(uap, new_size));
487 
488 	/*
489 	 * Growing mapped region.
490 	 */
491 	if (new_size > old_size) {
492 		/*
493 		 * XXX Implement me.  What we probably want to do is
494 		 * XXX dig out the guts of the old mapping, mmap that
495 		 * XXX object again with the new size, then munmap
496 		 * XXX the old mapping.
497 		 */
498 		*retval = 0;
499 		return (ENOMEM);
500 	}
501 
502 	/*
503 	 * Shrinking mapped region.
504 	 */
505 	if (new_size < old_size) {
506 		SCARG(&mua, addr) = (caddr_t)SCARG(uap, old_address) +
507 		    new_size;
508 		SCARG(&mua, len) = old_size - new_size;
509 		error = sys_munmap(p, &mua, retval);
510 		*retval = error ? 0 : (register_t)SCARG(uap, old_address);
511 		return (error);
512 	}
513 
514 	/*
515 	 * No change.
516 	 */
517 	*retval = (register_t)SCARG(uap, old_address);
518 	return (0);
519 }
520 
521 int
522 linux_sys_msync(p, v, retval)
523 	struct proc *p;
524 	void *v;
525 	register_t *retval;
526 {
527 	struct linux_sys_msync_args /* {
528 		syscallarg(caddr_t) addr;
529 		syscallarg(int) len;
530 		syscallarg(int) fl;
531 	} */ *uap = v;
532 
533 	struct sys___msync13_args bma;
534 
535 	/* flags are ignored */
536 	SCARG(&bma, addr) = SCARG(uap, addr);
537 	SCARG(&bma, len) = SCARG(uap, len);
538 	SCARG(&bma, flags) = SCARG(uap, fl);
539 
540 	return sys___msync13(p, &bma, retval);
541 }
542 
543 int
544 linux_sys_mprotect(p, v, retval)
545 	struct proc *p;
546 	void *v;
547 	register_t *retval;
548 {
549 	struct linux_sys_mprotect_args /* {
550 		syscallarg(const void *) start;
551 		syscallarg(unsigned long) len;
552 		syscallarg(int) prot;
553 	} */ *uap = v;
554 	unsigned long end, start = (unsigned long)SCARG(uap, start), len;
555 	int prot = SCARG(uap, prot);
556 	struct vm_map_entry *entry;
557 	struct vm_map *map = &p->p_vmspace->vm_map;
558 
559 	if (start & PAGE_MASK)
560 		return EINVAL;
561 
562 	len = round_page(SCARG(uap, len));
563 	end = start + len;
564 
565 	if (end < start)
566 		return EINVAL;
567 	else if (end == start)
568 		return 0;
569 
570 	if (SCARG(uap, prot) & ~(PROT_READ | PROT_WRITE | PROT_EXEC))
571 		return EINVAL;
572 
573 	vm_map_lock(map);
574 #ifdef notdef
575 	VM_MAP_RANGE_CHECK(map, start, end);
576 #endif
577 	if (!uvm_map_lookup_entry(map, start, &entry) || entry->start > start) {
578 		vm_map_unlock(map);
579 		return EFAULT;
580 	}
581 	vm_map_unlock(map);
582 	return uvm_map_protect(map, start, end, prot, FALSE);
583 }
584 
585 /*
586  * This code is partly stolen from src/lib/libc/compat-43/times.c
587  * XXX - CLK_TCK isn't declared in /sys, just in <time.h>, done here
588  */
589 
590 #define CLK_TCK 100
591 #define	CONVTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
592 
593 int
594 linux_sys_times(p, v, retval)
595 	struct proc *p;
596 	void *v;
597 	register_t *retval;
598 {
599 	struct linux_sys_times_args /* {
600 		syscallarg(struct times *) tms;
601 	} */ *uap = v;
602 	struct timeval t;
603 	struct linux_tms ltms;
604 	struct rusage ru;
605 	int error, s;
606 
607 	calcru(p, &ru.ru_utime, &ru.ru_stime, NULL);
608 	ltms.ltms_utime = CONVTCK(ru.ru_utime);
609 	ltms.ltms_stime = CONVTCK(ru.ru_stime);
610 
611 	ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime);
612 	ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime);
613 
614 	if ((error = copyout(&ltms, SCARG(uap, tms), sizeof ltms)))
615 		return error;
616 
617 	s = splclock();
618 	timersub(&time, &boottime, &t);
619 	splx(s);
620 
621 	retval[0] = ((linux_clock_t)(CONVTCK(t)));
622 	return 0;
623 }
624 
625 /*
626  * Linux 'readdir' call. This code is mostly taken from the
627  * SunOS getdents call (see compat/sunos/sunos_misc.c), though
628  * an attempt has been made to keep it a little cleaner (failing
629  * miserably, because of the cruft needed if count 1 is passed).
630  *
631  * The d_off field should contain the offset of the next valid entry,
632  * but in Linux it has the offset of the entry itself. We emulate
633  * that bug here.
634  *
635  * Read in BSD-style entries, convert them, and copy them out.
636  *
637  * Note that this doesn't handle union-mounted filesystems.
638  */
639 int
640 linux_sys_getdents(p, v, retval)
641 	struct proc *p;
642 	void *v;
643 	register_t *retval;
644 {
645 	struct linux_sys_getdents_args /* {
646 		syscallarg(int) fd;
647 		syscallarg(struct linux_dirent *) dent;
648 		syscallarg(unsigned int) count;
649 	} */ *uap = v;
650 	struct dirent *bdp;
651 	struct vnode *vp;
652 	caddr_t	inp, buf;		/* BSD-format */
653 	int len, reclen;		/* BSD-format */
654 	caddr_t outp;			/* Linux-format */
655 	int resid, linux_reclen = 0;	/* Linux-format */
656 	struct file *fp;
657 	struct uio auio;
658 	struct iovec aiov;
659 	struct linux_dirent idb;
660 	off_t off;		/* true file offset */
661 	int buflen, error, eofflag, nbytes, oldcall;
662 	struct vattr va;
663 	off_t *cookiebuf = NULL, *cookie;
664 	int ncookies;
665 
666 	/* getvnode() will use the descriptor for us */
667 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
668 		return (error);
669 
670 	if ((fp->f_flag & FREAD) == 0) {
671 		error = EBADF;
672 		goto out1;
673 	}
674 
675 	vp = (struct vnode *)fp->f_data;
676 	if (vp->v_type != VDIR) {
677 		error = EINVAL;
678 		goto out1;
679 	}
680 
681 	if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)))
682 		goto out1;
683 
684 	nbytes = SCARG(uap, count);
685 	if (nbytes == 1) {	/* emulating old, broken behaviour */
686 		nbytes = sizeof (idb);
687 		buflen = max(va.va_blocksize, nbytes);
688 		oldcall = 1;
689 	} else {
690 		buflen = min(MAXBSIZE, nbytes);
691 		if (buflen < va.va_blocksize)
692 			buflen = va.va_blocksize;
693 		oldcall = 0;
694 	}
695 	buf = malloc(buflen, M_TEMP, M_WAITOK);
696 
697 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
698 	off = fp->f_offset;
699 again:
700 	aiov.iov_base = buf;
701 	aiov.iov_len = buflen;
702 	auio.uio_iov = &aiov;
703 	auio.uio_iovcnt = 1;
704 	auio.uio_rw = UIO_READ;
705 	auio.uio_segflg = UIO_SYSSPACE;
706 	auio.uio_procp = p;
707 	auio.uio_resid = buflen;
708 	auio.uio_offset = off;
709 	/*
710          * First we read into the malloc'ed buffer, then
711          * we massage it into user space, one record at a time.
712          */
713 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf,
714 	    &ncookies);
715 	if (error)
716 		goto out;
717 
718 	inp = buf;
719 	outp = (caddr_t)SCARG(uap, dent);
720 	resid = nbytes;
721 	if ((len = buflen - auio.uio_resid) == 0)
722 		goto eof;
723 
724 	for (cookie = cookiebuf; len > 0; len -= reclen) {
725 		bdp = (struct dirent *)inp;
726 		reclen = bdp->d_reclen;
727 		if (reclen & 3)
728 			panic("linux_readdir");
729 		if (bdp->d_fileno == 0) {
730 			inp += reclen;	/* it is a hole; squish it out */
731 			off = *cookie++;
732 			continue;
733 		}
734 		linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen);
735 		if (reclen > len || resid < linux_reclen) {
736 			/* entry too big for buffer, so just stop */
737 			outp++;
738 			break;
739 		}
740 		/*
741 		 * Massage in place to make a Linux-shaped dirent (otherwise
742 		 * we have to worry about touching user memory outside of
743 		 * the copyout() call).
744 		 */
745 		idb.d_ino = bdp->d_fileno;
746 		/*
747 		 * The old readdir() call misuses the offset and reclen fields.
748 		 */
749 		if (oldcall) {
750 			idb.d_off = (linux_off_t)linux_reclen;
751 			idb.d_reclen = (u_short)bdp->d_namlen;
752 		} else {
753 			if (sizeof (idb.d_off) < 4 && (off >> 32) != 0) {
754 				compat_offseterr(vp, "linux_getdents");
755 				error = EINVAL;
756 				goto out;
757 			}
758 			idb.d_off = (linux_off_t)off;
759 			idb.d_reclen = (u_short)linux_reclen;
760 		}
761 		strcpy(idb.d_name, bdp->d_name);
762 		if ((error = copyout((caddr_t)&idb, outp, linux_reclen)))
763 			goto out;
764 		/* advance past this real entry */
765 		inp += reclen;
766 		off = *cookie++;	/* each entry points to itself */
767 		/* advance output past Linux-shaped entry */
768 		outp += linux_reclen;
769 		resid -= linux_reclen;
770 		if (oldcall)
771 			break;
772 	}
773 
774 	/* if we squished out the whole block, try again */
775 	if (outp == (caddr_t)SCARG(uap, dent))
776 		goto again;
777 	fp->f_offset = off;	/* update the vnode offset */
778 
779 	if (oldcall)
780 		nbytes = resid + linux_reclen;
781 
782 eof:
783 	*retval = nbytes - resid;
784 out:
785 	VOP_UNLOCK(vp, 0);
786 	if (cookiebuf)
787 		free(cookiebuf, M_TEMP);
788 	free(buf, M_TEMP);
789 out1:
790 	FILE_UNUSE(fp, p);
791 	return error;
792 }
793 
794 /*
795  * Linux 'readdir' call. This code is mostly taken from the
796  * SunOS getdents call (see compat/sunos/sunos_misc.c), though
797  * an attempt has been made to keep it a little cleaner (failing
798  * miserably, because of the cruft needed if count 1 is passed).
799  *
800  * The d_off field should contain the offset of the next valid entry,
801  * but in Linux it has the offset of the entry itself. We emulate
802  * that bug here.
803  *
804  * Read in BSD-style entries, convert them, and copy them out.
805  *
806  * Note that this doesn't handle union-mounted filesystems.
807  */
808 int
809 linux_sys_getdents64(p, v, retval)
810 	struct proc *p;
811 	void *v;
812 	register_t *retval;
813 {
814 	struct linux_sys_getdents_args /* {
815 		syscallarg(int) fd;
816 		syscallarg(struct linux_dirent64 *) dent;
817 		syscallarg(unsigned int) count;
818 	} */ *uap = v;
819 	struct dirent *bdp;
820 	struct vnode *vp;
821 	caddr_t	inp, buf;		/* BSD-format */
822 	int len, reclen;		/* BSD-format */
823 	caddr_t outp;			/* Linux-format */
824 	int resid, linux_reclen = 0;	/* Linux-format */
825 	struct file *fp;
826 	struct uio auio;
827 	struct iovec aiov;
828 	struct linux_dirent64 idb;
829 	off_t off;		/* true file offset */
830 	int buflen, error, eofflag, nbytes, oldcall;
831 	struct vattr va;
832 	off_t *cookiebuf = NULL, *cookie;
833 	int ncookies;
834 
835 	/* getvnode() will use the descriptor for us */
836 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
837 		return (error);
838 
839 	if ((fp->f_flag & FREAD) == 0) {
840 		error = EBADF;
841 		goto out1;
842 	}
843 
844 	vp = (struct vnode *)fp->f_data;
845 	if (vp->v_type != VDIR) {
846 		error = EINVAL;
847 		goto out1;
848 	}
849 
850 	if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)))
851 		goto out1;
852 
853 	nbytes = SCARG(uap, count);
854 	if (nbytes == 1) {	/* emulating old, broken behaviour */
855 		nbytes = sizeof (idb);
856 		buflen = max(va.va_blocksize, nbytes);
857 		oldcall = 1;
858 	} else {
859 		buflen = min(MAXBSIZE, nbytes);
860 		if (buflen < va.va_blocksize)
861 			buflen = va.va_blocksize;
862 		oldcall = 0;
863 	}
864 	buf = malloc(buflen, M_TEMP, M_WAITOK);
865 
866 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
867 	off = fp->f_offset;
868 again:
869 	aiov.iov_base = buf;
870 	aiov.iov_len = buflen;
871 	auio.uio_iov = &aiov;
872 	auio.uio_iovcnt = 1;
873 	auio.uio_rw = UIO_READ;
874 	auio.uio_segflg = UIO_SYSSPACE;
875 	auio.uio_procp = p;
876 	auio.uio_resid = buflen;
877 	auio.uio_offset = off;
878 	/*
879          * First we read into the malloc'ed buffer, then
880          * we massage it into user space, one record at a time.
881          */
882 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf,
883 	    &ncookies);
884 	if (error)
885 		goto out;
886 
887 	inp = buf;
888 	outp = (caddr_t)SCARG(uap, dent);
889 	resid = nbytes;
890 	if ((len = buflen - auio.uio_resid) == 0)
891 		goto eof;
892 
893 	for (cookie = cookiebuf; len > 0; len -= reclen) {
894 		bdp = (struct dirent *)inp;
895 		reclen = bdp->d_reclen;
896 		if (reclen & 3)
897 			panic("linux_readdir");
898 		if (bdp->d_fileno == 0) {
899 			inp += reclen;	/* it is a hole; squish it out */
900 			off = *cookie++;
901 			continue;
902 		}
903 		linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen);
904 		if (reclen > len || resid < linux_reclen) {
905 			/* entry too big for buffer, so just stop */
906 			outp++;
907 			break;
908 		}
909 		/*
910 		 * Massage in place to make a Linux-shaped dirent (otherwise
911 		 * we have to worry about touching user memory outside of
912 		 * the copyout() call).
913 		 */
914 		idb.d_ino = bdp->d_fileno;
915 		idb.d_type = bdp->d_type;
916 		/*
917 		 * The old readdir() call misuses the offset and reclen fields.
918 		 */
919 		if (oldcall) {
920 			idb.d_off = linux_reclen;
921 			idb.d_reclen = (u_short)bdp->d_namlen;
922 		} else {
923 			if (sizeof (idb.d_off) < 4 && (off >> 32) != 0) {
924 				compat_offseterr(vp, "linux_getdents");
925 				error = EINVAL;
926 				goto out;
927 			}
928 			idb.d_off = off;
929 			idb.d_reclen = (u_short)linux_reclen;
930 		}
931 		strcpy(idb.d_name, bdp->d_name);
932 		if ((error = copyout((caddr_t)&idb, outp, linux_reclen)))
933 			goto out;
934 		/* advance past this real entry */
935 		inp += reclen;
936 		off = *cookie++;	/* each entry points to itself */
937 		/* advance output past Linux-shaped entry */
938 		outp += linux_reclen;
939 		resid -= linux_reclen;
940 		if (oldcall)
941 			break;
942 	}
943 
944 	/* if we squished out the whole block, try again */
945 	if (outp == (caddr_t)SCARG(uap, dent))
946 		goto again;
947 	fp->f_offset = off;	/* update the vnode offset */
948 
949 	if (oldcall)
950 		nbytes = resid + linux_reclen;
951 
952 eof:
953 	*retval = nbytes - resid;
954 out:
955 	VOP_UNLOCK(vp, 0);
956 	if (cookiebuf)
957 		free(cookiebuf, M_TEMP);
958 	free(buf, M_TEMP);
959 out1:
960 	FILE_UNUSE(fp, p);
961 	return error;
962 }
963 
964 /*
965  * Even when just using registers to pass arguments to syscalls you can
966  * have 5 of them on the i386. So this newer version of select() does
967  * this.
968  */
969 int
970 linux_sys_select(p, v, retval)
971 	struct proc *p;
972 	void *v;
973 	register_t *retval;
974 {
975 	struct linux_sys_select_args /* {
976 		syscallarg(int) nfds;
977 		syscallarg(fd_set *) readfds;
978 		syscallarg(fd_set *) writefds;
979 		syscallarg(fd_set *) exceptfds;
980 		syscallarg(struct timeval *) timeout;
981 	} */ *uap = v;
982 
983 	return linux_select1(p, retval, SCARG(uap, nfds), SCARG(uap, readfds),
984 	    SCARG(uap, writefds), SCARG(uap, exceptfds), SCARG(uap, timeout));
985 }
986 
987 /*
988  * Common code for the old and new versions of select(). A couple of
989  * things are important:
990  * 1) return the amount of time left in the 'timeout' parameter
991  * 2) select never returns ERESTART on Linux, always return EINTR
992  */
993 int
994 linux_select1(p, retval, nfds, readfds, writefds, exceptfds, timeout)
995 	struct proc *p;
996 	register_t *retval;
997 	int nfds;
998 	fd_set *readfds, *writefds, *exceptfds;
999 	struct timeval *timeout;
1000 {
1001 	struct sys_select_args bsa;
1002 	struct timeval tv0, tv1, utv, *tvp;
1003 	caddr_t sg;
1004 	int error;
1005 
1006 	SCARG(&bsa, nd) = nfds;
1007 	SCARG(&bsa, in) = readfds;
1008 	SCARG(&bsa, ou) = writefds;
1009 	SCARG(&bsa, ex) = exceptfds;
1010 	SCARG(&bsa, tv) = timeout;
1011 
1012 	/*
1013 	 * Store current time for computation of the amount of
1014 	 * time left.
1015 	 */
1016 	if (timeout) {
1017 		if ((error = copyin(timeout, &utv, sizeof(utv))))
1018 			return error;
1019 		if (itimerfix(&utv)) {
1020 			/*
1021 			 * The timeval was invalid.  Convert it to something
1022 			 * valid that will act as it does under Linux.
1023 			 */
1024 			sg = stackgap_init(p, 0);
1025 			tvp = stackgap_alloc(p, &sg, sizeof(utv));
1026 			utv.tv_sec += utv.tv_usec / 1000000;
1027 			utv.tv_usec %= 1000000;
1028 			if (utv.tv_usec < 0) {
1029 				utv.tv_sec -= 1;
1030 				utv.tv_usec += 1000000;
1031 			}
1032 			if (utv.tv_sec < 0)
1033 				timerclear(&utv);
1034 			if ((error = copyout(&utv, tvp, sizeof(utv))))
1035 				return error;
1036 			SCARG(&bsa, tv) = tvp;
1037 		}
1038 		microtime(&tv0);
1039 	}
1040 
1041 	error = sys_select(p, &bsa, retval);
1042 	if (error) {
1043 		/*
1044 		 * See fs/select.c in the Linux kernel.  Without this,
1045 		 * Maelstrom doesn't work.
1046 		 */
1047 		if (error == ERESTART)
1048 			error = EINTR;
1049 		return error;
1050 	}
1051 
1052 	if (timeout) {
1053 		if (*retval) {
1054 			/*
1055 			 * Compute how much time was left of the timeout,
1056 			 * by subtracting the current time and the time
1057 			 * before we started the call, and subtracting
1058 			 * that result from the user-supplied value.
1059 			 */
1060 			microtime(&tv1);
1061 			timersub(&tv1, &tv0, &tv1);
1062 			timersub(&utv, &tv1, &utv);
1063 			if (utv.tv_sec < 0)
1064 				timerclear(&utv);
1065 		} else
1066 			timerclear(&utv);
1067 		if ((error = copyout(&utv, timeout, sizeof(utv))))
1068 			return error;
1069 	}
1070 
1071 	return 0;
1072 }
1073 
1074 /*
1075  * Get the process group of a certain process. Look it up
1076  * and return the value.
1077  */
1078 int
1079 linux_sys_getpgid(p, v, retval)
1080 	struct proc *p;
1081 	void *v;
1082 	register_t *retval;
1083 {
1084 	struct linux_sys_getpgid_args /* {
1085 		syscallarg(int) pid;
1086 	} */ *uap = v;
1087 	struct proc *targp;
1088 
1089 	if (SCARG(uap, pid) != 0 && SCARG(uap, pid) != p->p_pid) {
1090 		if ((targp = pfind(SCARG(uap, pid))) == 0)
1091 			return ESRCH;
1092 	}
1093 	else
1094 		targp = p;
1095 
1096 	retval[0] = targp->p_pgid;
1097 	return 0;
1098 }
1099 
1100 /*
1101  * Set the 'personality' (emulation mode) for the current process. Only
1102  * accept the Linux personality here (0). This call is needed because
1103  * the Linux ELF crt0 issues it in an ugly kludge to make sure that
1104  * ELF binaries run in Linux mode, not SVR4 mode.
1105  */
1106 int
1107 linux_sys_personality(p, v, retval)
1108 	struct proc *p;
1109 	void *v;
1110 	register_t *retval;
1111 {
1112 	struct linux_sys_personality_args /* {
1113 		syscallarg(int) per;
1114 	} */ *uap = v;
1115 
1116 	if (SCARG(uap, per) != 0)
1117 		return EINVAL;
1118 	retval[0] = 0;
1119 	return 0;
1120 }
1121 
1122 #if defined(__i386__) || defined(__m68k__)
1123 /*
1124  * The calls are here because of type conversions.
1125  */
1126 int
1127 linux_sys_setreuid16(p, v, retval)
1128 	struct proc *p;
1129 	void *v;
1130 	register_t *retval;
1131 {
1132 	struct linux_sys_setreuid16_args /* {
1133 		syscallarg(int) ruid;
1134 		syscallarg(int) euid;
1135 	} */ *uap = v;
1136 	struct sys_setreuid_args bsa;
1137 
1138 	SCARG(&bsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
1139 		(uid_t)-1 : SCARG(uap, ruid);
1140 	SCARG(&bsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
1141 		(uid_t)-1 : SCARG(uap, euid);
1142 
1143 	return sys_setreuid(p, &bsa, retval);
1144 }
1145 
1146 int
1147 linux_sys_setregid16(p, v, retval)
1148 	struct proc *p;
1149 	void *v;
1150 	register_t *retval;
1151 {
1152 	struct linux_sys_setregid16_args /* {
1153 		syscallarg(int) rgid;
1154 		syscallarg(int) egid;
1155 	} */ *uap = v;
1156 	struct sys_setregid_args bsa;
1157 
1158 	SCARG(&bsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
1159 		(uid_t)-1 : SCARG(uap, rgid);
1160 	SCARG(&bsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
1161 		(uid_t)-1 : SCARG(uap, egid);
1162 
1163 	return sys_setregid(p, &bsa, retval);
1164 }
1165 
1166 int
1167 linux_sys_setresuid16(p, v, retval)
1168 	struct proc *p;
1169 	void *v;
1170 	register_t *retval;
1171 {
1172 	struct linux_sys_setresuid16_args /* {
1173 		syscallarg(uid_t) ruid;
1174 		syscallarg(uid_t) euid;
1175 		syscallarg(uid_t) suid;
1176 	} */ *uap = v;
1177 	struct linux_sys_setresuid16_args lsa;
1178 
1179 	SCARG(&lsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
1180 		(uid_t)-1 : SCARG(uap, ruid);
1181 	SCARG(&lsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
1182 		(uid_t)-1 : SCARG(uap, euid);
1183 	SCARG(&lsa, suid) = ((linux_uid_t)SCARG(uap, suid) == (linux_uid_t)-1) ?
1184 		(uid_t)-1 : SCARG(uap, suid);
1185 
1186 	return linux_sys_setresuid(p, &lsa, retval);
1187 }
1188 
1189 int
1190 linux_sys_setresgid16(p, v, retval)
1191 	struct proc *p;
1192 	void *v;
1193 	register_t *retval;
1194 {
1195 	struct linux_sys_setresgid16_args /* {
1196 		syscallarg(gid_t) rgid;
1197 		syscallarg(gid_t) egid;
1198 		syscallarg(gid_t) sgid;
1199 	} */ *uap = v;
1200 	struct linux_sys_setresgid16_args lsa;
1201 
1202 	SCARG(&lsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
1203 		(gid_t)-1 : SCARG(uap, rgid);
1204 	SCARG(&lsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
1205 		(gid_t)-1 : SCARG(uap, egid);
1206 	SCARG(&lsa, sgid) = ((linux_gid_t)SCARG(uap, sgid) == (linux_gid_t)-1) ?
1207 		(gid_t)-1 : SCARG(uap, sgid);
1208 
1209 	return linux_sys_setresgid(p, &lsa, retval);
1210 }
1211 
1212 int
1213 linux_sys_getgroups16(p, v, retval)
1214 	struct proc *p;
1215 	void *v;
1216 	register_t *retval;
1217 {
1218 	struct linux_sys_getgroups16_args /* {
1219 		syscallarg(int) gidsetsize;
1220 		syscallarg(linux_gid_t *) gidset;
1221 	} */ *uap = v;
1222 	caddr_t sg;
1223 	int n, error, i;
1224 	struct sys_getgroups_args bsa;
1225 	gid_t *bset, *kbset;
1226 	linux_gid_t *lset;
1227 	struct pcred *pc = p->p_cred;
1228 
1229 	n = SCARG(uap, gidsetsize);
1230 	if (n < 0)
1231 		return EINVAL;
1232 	error = 0;
1233 	bset = kbset = NULL;
1234 	lset = NULL;
1235 	if (n > 0) {
1236 		n = min(pc->pc_ucred->cr_ngroups, n);
1237 		sg = stackgap_init(p, 0);
1238 		bset = stackgap_alloc(p, &sg, n * sizeof (gid_t));
1239 		kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK);
1240 		lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
1241 		if (bset == NULL || kbset == NULL || lset == NULL)
1242 			return ENOMEM;
1243 		SCARG(&bsa, gidsetsize) = n;
1244 		SCARG(&bsa, gidset) = bset;
1245 		error = sys_getgroups(p, &bsa, retval);
1246 		if (error != 0)
1247 			goto out;
1248 		error = copyin(bset, kbset, n * sizeof (gid_t));
1249 		if (error != 0)
1250 			goto out;
1251 		for (i = 0; i < n; i++)
1252 			lset[i] = (linux_gid_t)kbset[i];
1253 		error = copyout(lset, SCARG(uap, gidset),
1254 		    n * sizeof (linux_gid_t));
1255 	} else
1256 		*retval = pc->pc_ucred->cr_ngroups;
1257 out:
1258 	if (kbset != NULL)
1259 		free(kbset, M_TEMP);
1260 	if (lset != NULL)
1261 		free(lset, M_TEMP);
1262 	return error;
1263 }
1264 
1265 int
1266 linux_sys_setgroups16(p, v, retval)
1267 	struct proc *p;
1268 	void *v;
1269 	register_t *retval;
1270 {
1271 	struct linux_sys_setgroups16_args /* {
1272 		syscallarg(int) gidsetsize;
1273 		syscallarg(linux_gid_t *) gidset;
1274 	} */ *uap = v;
1275 	caddr_t sg;
1276 	int n;
1277 	int error, i;
1278 	struct sys_setgroups_args bsa;
1279 	gid_t *bset, *kbset;
1280 	linux_gid_t *lset;
1281 
1282 	n = SCARG(uap, gidsetsize);
1283 	if (n < 0 || n > NGROUPS)
1284 		return EINVAL;
1285 	sg = stackgap_init(p, 0);
1286 	bset = stackgap_alloc(p, &sg, n * sizeof (gid_t));
1287 	lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
1288 	kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK);
1289 	if (lset == NULL || bset == NULL)
1290 		return ENOMEM;
1291 	error = copyin(SCARG(uap, gidset), lset, n * sizeof (linux_gid_t));
1292 	if (error != 0)
1293 		goto out;
1294 	for (i = 0; i < n; i++)
1295 		kbset[i] = (gid_t)lset[i];
1296 	error = copyout(kbset, bset, n * sizeof (gid_t));
1297 	if (error != 0)
1298 		goto out;
1299 	SCARG(&bsa, gidsetsize) = n;
1300 	SCARG(&bsa, gidset) = bset;
1301 	error = sys_setgroups(p, &bsa, retval);
1302 
1303 out:
1304 	if (lset != NULL)
1305 		free(lset, M_TEMP);
1306 	if (kbset != NULL)
1307 		free(kbset, M_TEMP);
1308 
1309 	return error;
1310 }
1311 
1312 #endif /* __i386__ || __m68k__ */
1313 
1314 /*
1315  * We have nonexistent fsuid equal to uid.
1316  * If modification is requested, refuse.
1317  */
1318 int
1319 linux_sys_setfsuid(p, v, retval)
1320 	 struct proc *p;
1321 	 void *v;
1322 	 register_t *retval;
1323 {
1324 	 struct linux_sys_setfsuid_args /* {
1325 		 syscallarg(uid_t) uid;
1326 	 } */ *uap = v;
1327 	 uid_t uid;
1328 
1329 	 uid = SCARG(uap, uid);
1330 	 if (p->p_cred->p_ruid != uid)
1331 		 return sys_nosys(p, v, retval);
1332 	 else
1333 		 return (0);
1334 }
1335 
1336 /* XXX XXX XXX */
1337 #ifndef alpha
1338 int
1339 linux_sys_getfsuid(p, v, retval)
1340 	struct proc *p;
1341 	void *v;
1342 	register_t *retval;
1343 {
1344 	return sys_getuid(p, v, retval);
1345 }
1346 #endif
1347 
1348 int
1349 linux_sys_setresuid(p, v, retval)
1350 	struct proc *p;
1351 	void *v;
1352 	register_t *retval;
1353 {
1354 	struct linux_sys_setresuid_args /* {
1355 		syscallarg(uid_t) ruid;
1356 		syscallarg(uid_t) euid;
1357 		syscallarg(uid_t) suid;
1358 	} */ *uap = v;
1359 	struct pcred *pc = p->p_cred;
1360 	uid_t ruid, euid, suid;
1361 	int error;
1362 
1363 	ruid = SCARG(uap, ruid);
1364 	euid = SCARG(uap, euid);
1365 	suid = SCARG(uap, suid);
1366 
1367 	/*
1368 	 * Note: These checks are a little different than the NetBSD
1369 	 * setreuid(2) call performs.  This precisely follows the
1370 	 * behavior of the Linux kernel.
1371 	 */
1372 	if (ruid != (uid_t)-1 &&
1373 	    ruid != pc->p_ruid &&
1374 	    ruid != pc->pc_ucred->cr_uid &&
1375 	    ruid != pc->p_svuid &&
1376 	    (error = suser(pc->pc_ucred, &p->p_acflag)))
1377 		return (error);
1378 
1379 	if (euid != (uid_t)-1 &&
1380 	    euid != pc->p_ruid &&
1381 	    euid != pc->pc_ucred->cr_uid &&
1382 	    euid != pc->p_svuid &&
1383 	    (error = suser(pc->pc_ucred, &p->p_acflag)))
1384 		return (error);
1385 
1386 	if (suid != (uid_t)-1 &&
1387 	    suid != pc->p_ruid &&
1388 	    suid != pc->pc_ucred->cr_uid &&
1389 	    suid != pc->p_svuid &&
1390 	    (error = suser(pc->pc_ucred, &p->p_acflag)))
1391 		return (error);
1392 
1393 	/*
1394 	 * Now assign the new real, effective, and saved UIDs.
1395 	 * Note that Linux, unlike NetBSD in setreuid(2), does not
1396 	 * set the saved UID in this call unless the user specifies
1397 	 * it.
1398 	 */
1399 	if (ruid != (uid_t)-1) {
1400 		(void)chgproccnt(pc->p_ruid, -1);
1401 		(void)chgproccnt(ruid, 1);
1402 		pc->p_ruid = ruid;
1403 	}
1404 
1405 	if (euid != (uid_t)-1) {
1406 		pc->pc_ucred = crcopy(pc->pc_ucred);
1407 		pc->pc_ucred->cr_uid = euid;
1408 	}
1409 
1410 	if (suid != (uid_t)-1)
1411 		pc->p_svuid = suid;
1412 
1413 	if (ruid != (uid_t)-1 && euid != (uid_t)-1 && suid != (uid_t)-1)
1414 		p->p_flag |= P_SUGID;
1415 	return (0);
1416 }
1417 
1418 int
1419 linux_sys_getresuid(p, v, retval)
1420 	struct proc *p;
1421 	void *v;
1422 	register_t *retval;
1423 {
1424 	struct linux_sys_getresuid_args /* {
1425 		syscallarg(uid_t *) ruid;
1426 		syscallarg(uid_t *) euid;
1427 		syscallarg(uid_t *) suid;
1428 	} */ *uap = v;
1429 	struct pcred *pc = p->p_cred;
1430 	int error;
1431 
1432 	/*
1433 	 * Linux copies these values out to userspace like so:
1434 	 *
1435 	 *	1. Copy out ruid.
1436 	 *	2. If that succeeds, copy out euid.
1437 	 *	3. If both of those succeed, copy out suid.
1438 	 */
1439 	if ((error = copyout(&pc->p_ruid, SCARG(uap, ruid),
1440 			     sizeof(uid_t))) != 0)
1441 		return (error);
1442 
1443 	if ((error = copyout(&pc->pc_ucred->cr_uid, SCARG(uap, euid),
1444 			     sizeof(uid_t))) != 0)
1445 		return (error);
1446 
1447 	return (copyout(&pc->p_svuid, SCARG(uap, suid), sizeof(uid_t)));
1448 }
1449 
1450 int
1451 linux_sys_ptrace(p, v, retval)
1452 	struct proc *p;
1453 	void *v;
1454 	register_t *retval;
1455 {
1456 	struct linux_sys_ptrace_args /* {
1457 		i386, m68k, powerpc: T=int
1458 		alpha: T=long
1459 		syscallarg(T) request;
1460 		syscallarg(T) pid;
1461 		syscallarg(T) addr;
1462 		syscallarg(T) data;
1463 	} */ *uap = v;
1464 	const int *ptr;
1465 	int request;
1466 	int error;
1467 
1468 	ptr = linux_ptrace_request_map;
1469 	request = SCARG(uap, request);
1470 	while (*ptr != -1)
1471 		if (*ptr++ == request) {
1472 			struct sys_ptrace_args pta;
1473 
1474 			SCARG(&pta, req) = *ptr;
1475 			SCARG(&pta, pid) = SCARG(uap, pid);
1476 			SCARG(&pta, addr) = (caddr_t)SCARG(uap, addr);
1477 			SCARG(&pta, data) = SCARG(uap, data);
1478 
1479 			/*
1480 			 * Linux ptrace(PTRACE_CONT, pid, 0, 0) means actually
1481 			 * to continue where the process left off previously.
1482 			 * The same thing is achieved by addr == (caddr_t) 1
1483 			 * on NetBSD, so rewrite 'addr' appropriately.
1484 			 */
1485 			if (request == LINUX_PTRACE_CONT && SCARG(uap, addr)==0)
1486 				SCARG(&pta, addr) = (caddr_t) 1;
1487 
1488 			error = sys_ptrace(p, &pta, retval);
1489 			if (error)
1490 				return error;
1491 			switch (request) {
1492 			case LINUX_PTRACE_PEEKTEXT:
1493 			case LINUX_PTRACE_PEEKDATA:
1494 				error = copyout (retval,
1495 				    (caddr_t)SCARG(uap, data), sizeof *retval);
1496 				*retval = SCARG(uap, data);
1497 				break;
1498 			default:
1499 				break;
1500 			}
1501 			return error;
1502 		}
1503 		else
1504 			ptr++;
1505 
1506 	return LINUX_SYS_PTRACE_ARCH(p, uap, retval);
1507 }
1508 
1509 int
1510 linux_sys_reboot(struct proc *p, void *v, register_t *retval)
1511 {
1512 	struct linux_sys_reboot_args /* {
1513 		syscallarg(int) magic1;
1514 		syscallarg(int) magic2;
1515 		syscallarg(int) cmd;
1516 		syscallarg(void *) arg;
1517 	} */ *uap = v;
1518 	struct sys_reboot_args /* {
1519 		syscallarg(int) opt;
1520 		syscallarg(char *) bootstr;
1521 	} */ sra;
1522 	int error;
1523 
1524 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1525 		return(error);
1526 
1527 	if (SCARG(uap, magic1) != LINUX_REBOOT_MAGIC1)
1528 		return(EINVAL);
1529 	if (SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2 &&
1530 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2A &&
1531 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2B)
1532 		return(EINVAL);
1533 
1534 	switch (SCARG(uap, cmd)) {
1535 	case LINUX_REBOOT_CMD_RESTART:
1536 		SCARG(&sra, opt) = RB_AUTOBOOT;
1537 		break;
1538 	case LINUX_REBOOT_CMD_HALT:
1539 		SCARG(&sra, opt) = RB_HALT;
1540 		break;
1541 	case LINUX_REBOOT_CMD_POWER_OFF:
1542 		SCARG(&sra, opt) = RB_HALT|RB_POWERDOWN;
1543 		break;
1544 	case LINUX_REBOOT_CMD_RESTART2:
1545 		/* Reboot with an argument. */
1546 		SCARG(&sra, opt) = RB_AUTOBOOT|RB_STRING;
1547 		SCARG(&sra, bootstr) = SCARG(uap, arg);
1548 		break;
1549 	case LINUX_REBOOT_CMD_CAD_ON:
1550 		return(EINVAL);	/* We don't implement ctrl-alt-delete */
1551 	case LINUX_REBOOT_CMD_CAD_OFF:
1552 		return(0);
1553 	default:
1554 		return(EINVAL);
1555 	}
1556 
1557 	return(sys_reboot(p, &sra, retval));
1558 }
1559 
1560 /*
1561  * Copy of compat_12_sys_swapon().
1562  */
1563 int
1564 linux_sys_swapon(p, v, retval)
1565 	struct proc *p;
1566 	void *v;
1567 	register_t *retval;
1568 {
1569 	struct sys_swapctl_args ua;
1570 	struct linux_sys_swapon_args /* {
1571 		syscallarg(const char *) name;
1572 	} */ *uap = v;
1573 
1574 	SCARG(&ua, cmd) = SWAP_ON;
1575 	SCARG(&ua, arg) = (void *)SCARG(uap, name);
1576 	SCARG(&ua, misc) = 0;	/* priority */
1577 	return (sys_swapctl(p, &ua, retval));
1578 }
1579 
1580 /*
1581  * Stop swapping to the file or block device specified by path.
1582  */
1583 int
1584 linux_sys_swapoff(p, v, retval)
1585 	struct proc *p;
1586 	void *v;
1587 	register_t *retval;
1588 {
1589 	struct sys_swapctl_args ua;
1590 	struct linux_sys_swapoff_args /* {
1591 		syscallarg(const char *) path;
1592 	} */ *uap = v;
1593 
1594 	SCARG(&ua, cmd) = SWAP_OFF;
1595 	SCARG(&ua, arg) = (void *)SCARG(uap, path);
1596 	return (sys_swapctl(p, &ua, retval));
1597 }
1598 
1599 /*
1600  * Copy of compat_09_sys_setdomainname()
1601  */
1602 /* ARGSUSED */
1603 int
1604 linux_sys_setdomainname(p, v, retval)
1605 	struct proc *p;
1606 	void *v;
1607 	register_t *retval;
1608 {
1609 	struct linux_sys_setdomainname_args /* {
1610 		syscallarg(char *) domainname;
1611 		syscallarg(int) len;
1612 	} */ *uap = v;
1613 	int name;
1614 	int error;
1615 
1616 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1617 		return (error);
1618 	name = KERN_DOMAINNAME;
1619 	return (kern_sysctl(&name, 1, 0, 0, SCARG(uap, domainname),
1620 			    SCARG(uap, len), p));
1621 }
1622 
1623 /*
1624  * sysinfo()
1625  */
1626 /* ARGSUSED */
1627 int
1628 linux_sys_sysinfo(p, v, retval)
1629 	struct proc *p;
1630 	void *v;
1631 	register_t *retval;
1632 {
1633 	struct linux_sys_sysinfo_args /* {
1634 		syscallarg(struct linux_sysinfo *) arg;
1635 	} */ *uap = v;
1636 	struct linux_sysinfo si;
1637 	struct loadavg *la;
1638 
1639 	si.uptime = time.tv_sec - boottime.tv_sec;
1640 	la = &averunnable;
1641 	si.loads[0] = la->ldavg[0] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1642 	si.loads[1] = la->ldavg[1] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1643 	si.loads[2] = la->ldavg[2] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1644 	si.totalram = ctob(physmem);
1645 	si.freeram = uvmexp.free * uvmexp.pagesize;
1646 	si.sharedram = 0;	/* XXX */
1647 	si.bufferram = uvmexp.filepages * uvmexp.pagesize;
1648 	si.totalswap = uvmexp.swpages * uvmexp.pagesize;
1649 	si.freeswap = (uvmexp.swpages - uvmexp.swpginuse) * uvmexp.pagesize;
1650 	si.procs = nprocs;
1651 
1652 	/* The following are only present in newer Linux kernels. */
1653 	si.totalbig = 0;
1654 	si.freebig = 0;
1655 	si.mem_unit = 1;
1656 
1657 	return (copyout(&si, SCARG(uap, arg), sizeof si));
1658 }
1659 
1660 #define bsd_to_linux_rlimit1(l, b, f) \
1661     (l)->f = ((b)->f == RLIM_INFINITY || ((b)->f & 0xffffffff00000000) != 0) ? \
1662     LINUX_RLIM_INFINITY : (int32_t)(b)->f
1663 #define bsd_to_linux_rlimit(l, b) \
1664     bsd_to_linux_rlimit1(l, b, rlim_cur); \
1665     bsd_to_linux_rlimit1(l, b, rlim_max)
1666 
1667 #define linux_to_bsd_rlimit1(b, l, f) \
1668     (b)->f = (l)->f == LINUX_RLIM_INFINITY ? RLIM_INFINITY : (b)->f
1669 #define linux_to_bsd_rlimit(b, l) \
1670     linux_to_bsd_rlimit1(b, l, rlim_cur); \
1671     linux_to_bsd_rlimit1(b, l, rlim_max)
1672 
1673 static int
1674 linux_to_bsd_limit(lim)
1675 	int lim;
1676 {
1677 	switch (lim) {
1678 	case LINUX_RLIMIT_CPU:
1679 		return RLIMIT_CPU;
1680 	case LINUX_RLIMIT_FSIZE:
1681 		return RLIMIT_FSIZE;
1682 	case LINUX_RLIMIT_DATA:
1683 		return RLIMIT_DATA;
1684 	case LINUX_RLIMIT_STACK:
1685 		return RLIMIT_STACK;
1686 	case LINUX_RLIMIT_CORE:
1687 		return RLIMIT_CORE;
1688 	case LINUX_RLIMIT_RSS:
1689 		return RLIMIT_RSS;
1690 	case LINUX_RLIMIT_NPROC:
1691 		return RLIMIT_NPROC;
1692 	case LINUX_RLIMIT_NOFILE:
1693 		return RLIMIT_NOFILE;
1694 	case LINUX_RLIMIT_MEMLOCK:
1695 		return RLIMIT_MEMLOCK;
1696 	case LINUX_RLIMIT_AS:
1697 	case LINUX_RLIMIT_LOCKS:
1698 		return -EOPNOTSUPP;
1699 	default:
1700 		return -EINVAL;
1701 	}
1702 }
1703 
1704 
1705 int
1706 linux_sys_getrlimit(p, v, retval)
1707 	struct proc *p;
1708 	void *v;
1709 	register_t *retval;
1710 {
1711 	struct linux_sys_getrlimit_args /* {
1712 		syscallarg(int) which;
1713 		syscallarg(struct orlimit *) rlp;
1714 	} */ *uap = v;
1715 	caddr_t sg = stackgap_init(p, 0);
1716 	struct sys_getrlimit_args ap;
1717 	struct rlimit rl;
1718 	struct orlimit orl;
1719 	int error;
1720 
1721 	SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which));
1722 	if ((error = SCARG(&ap, which)) < 0)
1723 		return -error;
1724 	SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl);
1725 	if ((error = sys_getrlimit(p, &ap, retval)) != 0)
1726 		return error;
1727 	if ((error = copyin(SCARG(&ap, rlp), &rl, sizeof(rl))) != 0)
1728 		return error;
1729 	bsd_to_linux_rlimit(&orl, &rl);
1730 	return copyout(&orl, SCARG(uap, rlp), sizeof(orl));
1731 }
1732 
1733 int
1734 linux_sys_setrlimit(p, v, retval)
1735 	struct proc *p;
1736 	void *v;
1737 	register_t *retval;
1738 {
1739 	struct linux_sys_setrlimit_args /* {
1740 		syscallarg(int) which;
1741 		syscallarg(struct orlimit *) rlp;
1742 	} */ *uap = v;
1743 	caddr_t sg = stackgap_init(p, 0);
1744 	struct sys_setrlimit_args ap;
1745 	struct rlimit rl;
1746 	struct orlimit orl;
1747 	int error;
1748 
1749 	SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which));
1750 	SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl);
1751 	if ((error = SCARG(&ap, which)) < 0)
1752 		return -error;
1753 	if ((error = copyin(SCARG(uap, rlp), &orl, sizeof(orl))) != 0)
1754 		return error;
1755 	linux_to_bsd_rlimit(&rl, &orl);
1756 	/* XXX: alpha complains about this */
1757 	if ((error = copyout(&rl, (void *)SCARG(&ap, rlp), sizeof(rl))) != 0)
1758 		return error;
1759 	return sys_setrlimit(p, &ap, retval);
1760 }
1761 
1762 #ifndef __mips__
1763 /* XXX: this doesn't look 100% common, at least mips doesn't have it */
1764 int
1765 linux_sys_ugetrlimit(p, v, retval)
1766 	struct proc *p;
1767 	void *v;
1768 	register_t *retval;
1769 {
1770 	return linux_sys_getrlimit(p, v, retval);
1771 }
1772 #endif
1773 
1774 /*
1775  * This gets called for unsupported syscalls. The difference to sys_nosys()
1776  * is that process does not get SIGSYS, the call just returns with ENOSYS.
1777  * This is the way Linux does it and glibc depends on this behaviour.
1778  */
1779 int
1780 linux_sys_nosys(p, v, retval)
1781 	struct proc *p;
1782 	void *v;
1783 	register_t *retval;
1784 {
1785 	return (ENOSYS);
1786 }
1787