xref: /netbsd-src/sys/compat/linux/common/linux_misc.c (revision 27527e67bbdf8d9ec84fd58803048ed6d181ece2)
1 /*	$NetBSD: linux_misc.c,v 1.149 2006/01/31 14:02:55 yamt 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 1.149 2006/01/31 14:02:55 yamt 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_machdep.h>
106 #include <compat/linux/common/linux_types.h>
107 #include <compat/linux/common/linux_signal.h>
108 
109 #include <compat/linux/linux_syscallargs.h>
110 
111 #include <compat/linux/common/linux_fcntl.h>
112 #include <compat/linux/common/linux_mmap.h>
113 #include <compat/linux/common/linux_dirent.h>
114 #include <compat/linux/common/linux_util.h>
115 #include <compat/linux/common/linux_misc.h>
116 #include <compat/linux/common/linux_ptrace.h>
117 #include <compat/linux/common/linux_reboot.h>
118 #include <compat/linux/common/linux_emuldata.h>
119 
120 const int linux_ptrace_request_map[] = {
121 	LINUX_PTRACE_TRACEME,	PT_TRACE_ME,
122 	LINUX_PTRACE_PEEKTEXT,	PT_READ_I,
123 	LINUX_PTRACE_PEEKDATA,	PT_READ_D,
124 	LINUX_PTRACE_POKETEXT,	PT_WRITE_I,
125 	LINUX_PTRACE_POKEDATA,	PT_WRITE_D,
126 	LINUX_PTRACE_CONT,	PT_CONTINUE,
127 	LINUX_PTRACE_KILL,	PT_KILL,
128 	LINUX_PTRACE_ATTACH,	PT_ATTACH,
129 	LINUX_PTRACE_DETACH,	PT_DETACH,
130 #ifdef PT_STEP
131 	LINUX_PTRACE_SINGLESTEP,	PT_STEP,
132 #endif
133 	-1
134 };
135 
136 const struct linux_mnttypes linux_fstypes[] = {
137 	{ MOUNT_FFS,		LINUX_DEFAULT_SUPER_MAGIC	},
138 	{ MOUNT_NFS,		LINUX_NFS_SUPER_MAGIC 		},
139 	{ MOUNT_MFS,		LINUX_DEFAULT_SUPER_MAGIC	},
140 	{ MOUNT_MSDOS,		LINUX_MSDOS_SUPER_MAGIC		},
141 	{ MOUNT_LFS,		LINUX_DEFAULT_SUPER_MAGIC	},
142 	{ MOUNT_FDESC,		LINUX_DEFAULT_SUPER_MAGIC	},
143 	{ MOUNT_PORTAL,		LINUX_DEFAULT_SUPER_MAGIC	},
144 	{ MOUNT_NULL,		LINUX_DEFAULT_SUPER_MAGIC	},
145 	{ MOUNT_OVERLAY,	LINUX_DEFAULT_SUPER_MAGIC	},
146 	{ MOUNT_UMAP,		LINUX_DEFAULT_SUPER_MAGIC	},
147 	{ MOUNT_KERNFS,		LINUX_DEFAULT_SUPER_MAGIC	},
148 	{ MOUNT_PROCFS,		LINUX_PROC_SUPER_MAGIC		},
149 	{ MOUNT_AFS,		LINUX_DEFAULT_SUPER_MAGIC	},
150 	{ MOUNT_CD9660,		LINUX_ISOFS_SUPER_MAGIC		},
151 	{ MOUNT_UNION,		LINUX_DEFAULT_SUPER_MAGIC	},
152 	{ MOUNT_ADOSFS,		LINUX_ADFS_SUPER_MAGIC		},
153 	{ MOUNT_EXT2FS,		LINUX_EXT2_SUPER_MAGIC		},
154 	{ MOUNT_CFS,		LINUX_DEFAULT_SUPER_MAGIC	},
155 	{ MOUNT_CODA,		LINUX_CODA_SUPER_MAGIC		},
156 	{ MOUNT_FILECORE,	LINUX_DEFAULT_SUPER_MAGIC	},
157 	{ MOUNT_NTFS,		LINUX_DEFAULT_SUPER_MAGIC	},
158 	{ MOUNT_SMBFS,		LINUX_SMB_SUPER_MAGIC		},
159 	{ MOUNT_PTYFS,		LINUX_DEVPTS_SUPER_MAGIC	},
160 	{ MOUNT_TMPFS,		LINUX_DEFAULT_SUPER_MAGIC	}
161 };
162 const int linux_fstypes_cnt = sizeof(linux_fstypes) / sizeof(linux_fstypes[0]);
163 
164 #ifdef DEBUG_LINUX
165 #define DPRINTF(a)	uprintf a
166 #else
167 #define DPRINTF(a)
168 #endif
169 
170 /* Local linux_misc.c functions: */
171 #ifndef __amd64__
172 static void bsd_to_linux_statfs __P((const struct statvfs *,
173     struct linux_statfs *));
174 #endif
175 static int linux_to_bsd_limit __P((int));
176 static void linux_to_bsd_mmap_args __P((struct sys_mmap_args *,
177     const struct linux_sys_mmap_args *));
178 static int linux_mmap __P((struct lwp *, struct linux_sys_mmap_args *,
179     register_t *, off_t));
180 
181 
182 /*
183  * The information on a terminated (or stopped) process needs
184  * to be converted in order for Linux binaries to get a valid signal
185  * number out of it.
186  */
187 void
188 bsd_to_linux_wstat(st)
189 	int *st;
190 {
191 
192 	int sig;
193 
194 	if (WIFSIGNALED(*st)) {
195 		sig = WTERMSIG(*st);
196 		if (sig >= 0 && sig < NSIG)
197 			*st= (*st& ~0177) | native_to_linux_signo[sig];
198 	} else if (WIFSTOPPED(*st)) {
199 		sig = WSTOPSIG(*st);
200 		if (sig >= 0 && sig < NSIG)
201 			*st = (*st & ~0xff00) |
202 			    (native_to_linux_signo[sig] << 8);
203 	}
204 }
205 
206 /*
207  * wait4(2).  Passed on to the NetBSD call, surrounded by code to
208  * reserve some space for a NetBSD-style wait status, and converting
209  * it to what Linux wants.
210  */
211 int
212 linux_sys_wait4(l, v, retval)
213 	struct lwp *l;
214 	void *v;
215 	register_t *retval;
216 {
217 	struct linux_sys_wait4_args /* {
218 		syscallarg(int) pid;
219 		syscallarg(int *) status;
220 		syscallarg(int) options;
221 		syscallarg(struct rusage *) rusage;
222 	} */ *uap = v;
223 	struct proc *p = l->l_proc;
224 	struct sys_wait4_args w4a;
225 	int error, *status, tstat, options, linux_options;
226 	caddr_t sg;
227 
228 	if (SCARG(uap, status) != NULL) {
229 		sg = stackgap_init(p, 0);
230 		status = (int *) stackgap_alloc(p, &sg, sizeof *status);
231 	} else
232 		status = NULL;
233 
234 	linux_options = SCARG(uap, options);
235 	options = 0;
236 	if (linux_options & ~(LINUX_WAIT4_KNOWNFLAGS))
237 		return (EINVAL);
238 
239 	if (linux_options & LINUX_WAIT4_WNOHANG)
240 		options |= WNOHANG;
241 	if (linux_options & LINUX_WAIT4_WUNTRACED)
242 		options |= WUNTRACED;
243 	if (linux_options & LINUX_WAIT4_WALL)
244 		options |= WALLSIG;
245 	if (linux_options & LINUX_WAIT4_WCLONE)
246 		options |= WALTSIG;
247 #ifdef DIAGNOSTIC
248 	if (linux_options & LINUX_WAIT4_WNOTHREAD)
249 		printf("WARNING: %s: linux process %d.%d called "
250 		       "waitpid with __WNOTHREAD set!",
251 		       __FILE__, p->p_pid, l->l_lid);
252 
253 #endif
254 
255 	SCARG(&w4a, pid) = SCARG(uap, pid);
256 	SCARG(&w4a, status) = status;
257 	SCARG(&w4a, options) = options;
258 	SCARG(&w4a, rusage) = SCARG(uap, rusage);
259 
260 	if ((error = sys_wait4(l, &w4a, retval)))
261 		return error;
262 
263 	sigdelset(&p->p_sigctx.ps_siglist, SIGCHLD);
264 
265 	if (status != NULL) {
266 		if ((error = copyin(status, &tstat, sizeof tstat)))
267 			return error;
268 
269 		bsd_to_linux_wstat(&tstat);
270 		return copyout(&tstat, SCARG(uap, status), sizeof tstat);
271 	}
272 
273 	return 0;
274 }
275 
276 /*
277  * Linux brk(2). The check if the new address is >= the old one is
278  * done in the kernel in Linux. NetBSD does it in the library.
279  */
280 int
281 linux_sys_brk(l, v, retval)
282 	struct lwp *l;
283 	void *v;
284 	register_t *retval;
285 {
286 	struct linux_sys_brk_args /* {
287 		syscallarg(char *) nsize;
288 	} */ *uap = v;
289 	struct proc *p = l->l_proc;
290 	char *nbrk = SCARG(uap, nsize);
291 	struct sys_obreak_args oba;
292 	struct vmspace *vm = p->p_vmspace;
293 	struct linux_emuldata *ed = (struct linux_emuldata*)p->p_emuldata;
294 
295 	SCARG(&oba, nsize) = nbrk;
296 
297 	if ((caddr_t) nbrk > vm->vm_daddr && sys_obreak(l, &oba, retval) == 0)
298 		ed->s->p_break = (char*)nbrk;
299 	else
300 		nbrk = ed->s->p_break;
301 
302 	retval[0] = (register_t)nbrk;
303 
304 	return 0;
305 }
306 
307 #ifndef __amd64__
308 /*
309  * Convert NetBSD statvfs structure to Linux statfs structure.
310  * Linux doesn't have f_flag, and we can't set f_frsize due
311  * to glibc statvfs() bug (see below).
312  */
313 static void
314 bsd_to_linux_statfs(bsp, lsp)
315 	const struct statvfs *bsp;
316 	struct linux_statfs *lsp;
317 {
318 	int i;
319 
320 	for (i = 0; i < linux_fstypes_cnt; i++) {
321 		if (strcmp(bsp->f_fstypename, linux_fstypes[i].bsd) == 0) {
322 			lsp->l_ftype = linux_fstypes[i].linux;
323 			break;
324 		}
325 	}
326 
327 	if (i == linux_fstypes_cnt) {
328 		DPRINTF(("unhandled fstype in linux emulation: %s\n",
329 		    bsp->f_fstypename));
330 		lsp->l_ftype = LINUX_DEFAULT_SUPER_MAGIC;
331 	}
332 
333 	/*
334 	 * The sizes are expressed in number of blocks. The block
335 	 * size used for the size is f_frsize for POSIX-compliant
336 	 * statvfs. Linux statfs uses f_bsize as the block size
337 	 * (f_frsize used to not be available in Linux struct statfs).
338 	 * However, glibc 2.3.3 statvfs() wrapper fails to adjust the block
339 	 * counts for different f_frsize if f_frsize is provided by the kernel.
340 	 * POSIX conforming apps thus get wrong size if f_frsize
341 	 * is different to f_bsize. Thus, we just pretend we don't
342 	 * support f_frsize.
343 	 */
344 
345 	lsp->l_fbsize = bsp->f_frsize;
346 	lsp->l_ffrsize = 0;			/* compat */
347 	lsp->l_fblocks = bsp->f_blocks;
348 	lsp->l_fbfree = bsp->f_bfree;
349 	lsp->l_fbavail = bsp->f_bavail;
350 	lsp->l_ffiles = bsp->f_files;
351 	lsp->l_fffree = bsp->f_ffree;
352 	/* Linux sets the fsid to 0..., we don't */
353 	lsp->l_ffsid.val[0] = bsp->f_fsidx.__fsid_val[0];
354 	lsp->l_ffsid.val[1] = bsp->f_fsidx.__fsid_val[1];
355 	lsp->l_fnamelen = bsp->f_namemax;
356 	(void)memset(lsp->l_fspare, 0, sizeof(lsp->l_fspare));
357 }
358 
359 /*
360  * Implement the fs stat functions. Straightforward.
361  */
362 int
363 linux_sys_statfs(l, v, retval)
364 	struct lwp *l;
365 	void *v;
366 	register_t *retval;
367 {
368 	struct linux_sys_statfs_args /* {
369 		syscallarg(const char *) path;
370 		syscallarg(struct linux_statfs *) sp;
371 	} */ *uap = v;
372 	struct proc *p = l->l_proc;
373 	struct statvfs btmp, *bsp;
374 	struct linux_statfs ltmp;
375 	struct sys_statvfs1_args bsa;
376 	caddr_t sg;
377 	int error;
378 
379 	sg = stackgap_init(p, 0);
380 	bsp = (struct statvfs *) stackgap_alloc(p, &sg, sizeof (struct statvfs));
381 
382 	CHECK_ALT_EXIST(l, &sg, SCARG(uap, path));
383 
384 	SCARG(&bsa, path) = SCARG(uap, path);
385 	SCARG(&bsa, buf) = bsp;
386 	SCARG(&bsa, flags) = ST_WAIT;
387 
388 	if ((error = sys_statvfs1(l, &bsa, retval)))
389 		return error;
390 
391 	if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp)))
392 		return error;
393 
394 	bsd_to_linux_statfs(&btmp, &ltmp);
395 
396 	return copyout((caddr_t) &ltmp, (caddr_t) SCARG(uap, sp), sizeof ltmp);
397 }
398 
399 int
400 linux_sys_fstatfs(l, v, retval)
401 	struct lwp *l;
402 	void *v;
403 	register_t *retval;
404 {
405 	struct linux_sys_fstatfs_args /* {
406 		syscallarg(int) fd;
407 		syscallarg(struct linux_statfs *) sp;
408 	} */ *uap = v;
409 	struct proc *p = l->l_proc;
410 	struct statvfs btmp, *bsp;
411 	struct linux_statfs ltmp;
412 	struct sys_fstatvfs1_args bsa;
413 	caddr_t sg;
414 	int error;
415 
416 	sg = stackgap_init(p, 0);
417 	bsp = (struct statvfs *) stackgap_alloc(p, &sg, sizeof (struct statvfs));
418 
419 	SCARG(&bsa, fd) = SCARG(uap, fd);
420 	SCARG(&bsa, buf) = bsp;
421 	SCARG(&bsa, flags) = ST_WAIT;
422 
423 	if ((error = sys_fstatvfs1(l, &bsa, retval)))
424 		return error;
425 
426 	if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp)))
427 		return error;
428 
429 	bsd_to_linux_statfs(&btmp, &ltmp);
430 
431 	return copyout((caddr_t) &ltmp, (caddr_t) SCARG(uap, sp), sizeof ltmp);
432 }
433 #endif /* __amd64__ */
434 
435 /*
436  * uname(). Just copy the info from the various strings stored in the
437  * kernel, and put it in the Linux utsname structure. That structure
438  * is almost the same as the NetBSD one, only it has fields 65 characters
439  * long, and an extra domainname field.
440  */
441 int
442 linux_sys_uname(l, v, retval)
443 	struct lwp *l;
444 	void *v;
445 	register_t *retval;
446 {
447 	struct linux_sys_uname_args /* {
448 		syscallarg(struct linux_utsname *) up;
449 	} */ *uap = v;
450 	struct linux_utsname luts;
451 
452 	strncpy(luts.l_sysname, linux_sysname, sizeof(luts.l_sysname));
453 	strncpy(luts.l_nodename, hostname, sizeof(luts.l_nodename));
454 	strncpy(luts.l_release, linux_release, sizeof(luts.l_release));
455 	strncpy(luts.l_version, linux_version, sizeof(luts.l_version));
456 #ifdef LINUX_UNAME_ARCH
457 	strncpy(luts.l_machine, LINUX_UNAME_ARCH, sizeof(luts.l_machine));
458 #else
459 	strncpy(luts.l_machine, machine, sizeof(luts.l_machine));
460 #endif
461 	strncpy(luts.l_domainname, domainname, sizeof(luts.l_domainname));
462 
463 	return copyout(&luts, SCARG(uap, up), sizeof(luts));
464 }
465 
466 /* Used directly on: alpha, mips, ppc, sparc, sparc64 */
467 /* Used indirectly on: arm, i386, m68k */
468 
469 /*
470  * New type Linux mmap call.
471  * Only called directly on machines with >= 6 free regs.
472  */
473 int
474 linux_sys_mmap(l, v, retval)
475 	struct lwp *l;
476 	void *v;
477 	register_t *retval;
478 {
479 	struct linux_sys_mmap_args /* {
480 		syscallarg(unsigned long) addr;
481 		syscallarg(size_t) len;
482 		syscallarg(int) prot;
483 		syscallarg(int) flags;
484 		syscallarg(int) fd;
485 		syscallarg(linux_off_t) offset;
486 	} */ *uap = v;
487 
488 	if (SCARG(uap, offset) & PAGE_MASK)
489 		return EINVAL;
490 
491 	return linux_mmap(l, uap, retval, SCARG(uap, offset));
492 }
493 
494 /*
495  * Guts of most architectures' mmap64() implementations.  This shares
496  * its list of arguments with linux_sys_mmap().
497  *
498  * The difference in linux_sys_mmap2() is that "offset" is actually
499  * (offset / pagesize), not an absolute byte count.  This translation
500  * to pagesize offsets is done inside glibc between the mmap64() call
501  * point, and the actual syscall.
502  */
503 int
504 linux_sys_mmap2(l, v, retval)
505 	struct lwp *l;
506 	void *v;
507 	register_t *retval;
508 {
509 	struct linux_sys_mmap2_args /* {
510 		syscallarg(unsigned long) addr;
511 		syscallarg(size_t) len;
512 		syscallarg(int) prot;
513 		syscallarg(int) flags;
514 		syscallarg(int) fd;
515 		syscallarg(linux_off_t) offset;
516 	} */ *uap = v;
517 
518 	return linux_mmap(l, uap, retval,
519 	    ((off_t)SCARG(uap, offset)) << PAGE_SHIFT);
520 }
521 
522 /*
523  * Massage arguments and call system mmap(2).
524  */
525 static int
526 linux_mmap(l, uap, retval, offset)
527 	struct lwp *l;
528 	struct linux_sys_mmap_args *uap;
529 	register_t *retval;
530 	off_t offset;
531 {
532 	struct sys_mmap_args cma;
533 	int error;
534 	size_t mmoff=0;
535 
536 	if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN) {
537 		/*
538 		 * Request for stack-like memory segment. On linux, this
539 		 * works by mmap()ping (small) segment, which is automatically
540 		 * extended when page fault happens below the currently
541 		 * allocated area. We emulate this by allocating (typically
542 		 * bigger) segment sized at current stack size limit, and
543 		 * offsetting the requested and returned address accordingly.
544 		 * Since physical pages are only allocated on-demand, this
545 		 * is effectively identical.
546 		 */
547 		rlim_t ssl = l->l_proc->p_rlimit[RLIMIT_STACK].rlim_cur;
548 
549 		if (SCARG(uap, len) < ssl) {
550 			/* Compute the address offset */
551 			mmoff = round_page(ssl) - SCARG(uap, len);
552 
553 			if (SCARG(uap, addr))
554 				SCARG(uap, addr) -= mmoff;
555 
556 			SCARG(uap, len) = (size_t) ssl;
557 		}
558 	}
559 
560 	linux_to_bsd_mmap_args(&cma, uap);
561 	SCARG(&cma, pos) = offset;
562 
563 	error = sys_mmap(l, &cma, retval);
564 	if (error)
565 		return (error);
566 
567 	/* Shift the returned address for stack-like segment if necessary */
568 	if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN && mmoff)
569 		retval[0] += mmoff;
570 
571 	return (0);
572 }
573 
574 static void
575 linux_to_bsd_mmap_args(cma, uap)
576 	struct sys_mmap_args *cma;
577 	const struct linux_sys_mmap_args *uap;
578 {
579 	int flags = MAP_TRYFIXED, fl = SCARG(uap, flags);
580 
581 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_SHARED, MAP_SHARED);
582 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_PRIVATE, MAP_PRIVATE);
583 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_FIXED, MAP_FIXED);
584 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_ANON, MAP_ANON);
585 	/* XXX XAX ERH: Any other flags here?  There are more defined... */
586 
587 	SCARG(cma, addr) = (void *)SCARG(uap, addr);
588 	SCARG(cma, len) = SCARG(uap, len);
589 	SCARG(cma, prot) = SCARG(uap, prot);
590 	if (SCARG(cma, prot) & VM_PROT_WRITE) /* XXX */
591 		SCARG(cma, prot) |= VM_PROT_READ;
592 	SCARG(cma, flags) = flags;
593 	SCARG(cma, fd) = flags & MAP_ANON ? -1 : SCARG(uap, fd);
594 	SCARG(cma, pad) = 0;
595 }
596 
597 #define	LINUX_MREMAP_MAYMOVE	1
598 #define	LINUX_MREMAP_FIXED	2
599 
600 int
601 linux_sys_mremap(l, v, retval)
602 	struct lwp *l;
603 	void *v;
604 	register_t *retval;
605 {
606 	struct linux_sys_mremap_args /* {
607 		syscallarg(void *) old_address;
608 		syscallarg(size_t) old_size;
609 		syscallarg(size_t) new_size;
610 		syscallarg(u_long) flags;
611 	} */ *uap = v;
612 
613 	struct proc *p;
614 	struct vm_map *map;
615 	vaddr_t oldva;
616 	vaddr_t newva;
617 	size_t oldsize;
618 	size_t newsize;
619 	int flags;
620 	int uvmflags;
621 	int error;
622 
623 	flags = SCARG(uap, flags);
624 	oldva = (vaddr_t)SCARG(uap, old_address);
625 	oldsize = round_page(SCARG(uap, old_size));
626 	newsize = round_page(SCARG(uap, new_size));
627 	if ((flags & ~(LINUX_MREMAP_FIXED|LINUX_MREMAP_MAYMOVE)) != 0) {
628 		error = EINVAL;
629 		goto done;
630 	}
631 	if ((flags & LINUX_MREMAP_FIXED) != 0) {
632 		if ((flags & LINUX_MREMAP_MAYMOVE) == 0) {
633 			error = EINVAL;
634 			goto done;
635 		}
636 #if 0 /* notyet */
637 		newva = SCARG(uap, new_address);
638 		uvmflags = UVM_MREMAP_FIXED;
639 #else /* notyet */
640 		error = EOPNOTSUPP;
641 		goto done;
642 #endif /* notyet */
643 	} else if ((flags & LINUX_MREMAP_MAYMOVE) != 0) {
644 		uvmflags = 0;
645 	} else {
646 		newva = oldva;
647 		uvmflags = UVM_MREMAP_FIXED;
648 	}
649 	p = l->l_proc;
650 	map = &p->p_vmspace->vm_map;
651 	error = uvm_mremap(map, oldva, oldsize, map, &newva, newsize, p,
652 	    uvmflags);
653 
654 done:
655 	*retval = (error != 0) ? 0 : (register_t)newva;
656 	return error;
657 }
658 
659 int
660 linux_sys_msync(l, v, retval)
661 	struct lwp *l;
662 	void *v;
663 	register_t *retval;
664 {
665 	struct linux_sys_msync_args /* {
666 		syscallarg(caddr_t) addr;
667 		syscallarg(int) len;
668 		syscallarg(int) fl;
669 	} */ *uap = v;
670 
671 	struct sys___msync13_args bma;
672 
673 	/* flags are ignored */
674 	SCARG(&bma, addr) = SCARG(uap, addr);
675 	SCARG(&bma, len) = SCARG(uap, len);
676 	SCARG(&bma, flags) = SCARG(uap, fl);
677 
678 	return sys___msync13(l, &bma, retval);
679 }
680 
681 int
682 linux_sys_mprotect(l, v, retval)
683 	struct lwp *l;
684 	void *v;
685 	register_t *retval;
686 {
687 	struct linux_sys_mprotect_args /* {
688 		syscallarg(const void *) start;
689 		syscallarg(unsigned long) len;
690 		syscallarg(int) prot;
691 	} */ *uap = v;
692 	struct vm_map_entry *entry;
693 	struct vm_map *map;
694 	struct proc *p;
695 	vaddr_t end, start, len, stacklim;
696 	int prot, grows;
697 
698 	start = (vaddr_t)SCARG(uap, start);
699 	len = round_page(SCARG(uap, len));
700 	prot = SCARG(uap, prot);
701 	grows = prot & (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP);
702 	prot &= ~grows;
703 	end = start + len;
704 
705 	if (start & PAGE_MASK)
706 		return EINVAL;
707 	if (end < start)
708 		return EINVAL;
709 	if (end == start)
710 		return 0;
711 
712 	if (prot & ~(PROT_READ | PROT_WRITE | PROT_EXEC))
713 		return EINVAL;
714 	if (grows == (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP))
715 		return EINVAL;
716 
717 	p = l->l_proc;
718 	map = &p->p_vmspace->vm_map;
719 	vm_map_lock(map);
720 #ifdef notdef
721 	VM_MAP_RANGE_CHECK(map, start, end);
722 #endif
723 	if (!uvm_map_lookup_entry(map, start, &entry) || entry->start > start) {
724 		vm_map_unlock(map);
725 		return ENOMEM;
726 	}
727 
728 	/*
729 	 * Approximate the behaviour of PROT_GROWS{DOWN,UP}.
730 	 */
731 
732 	stacklim = (vaddr_t)p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur;
733 	if (grows & LINUX_PROT_GROWSDOWN) {
734 		if (USRSTACK - stacklim <= start && start < USRSTACK) {
735 			start = USRSTACK - stacklim;
736 		} else {
737 			start = entry->start;
738 		}
739 	} else if (grows & LINUX_PROT_GROWSUP) {
740 		if (USRSTACK <= end && end < USRSTACK + stacklim) {
741 			end = USRSTACK + stacklim;
742 		} else {
743 			end = entry->end;
744 		}
745 	}
746 	vm_map_unlock(map);
747 	return uvm_map_protect(map, start, end, prot, FALSE);
748 }
749 
750 /*
751  * This code is partly stolen from src/lib/libc/compat-43/times.c
752  */
753 
754 #define	CONVTCK(r)	(r.tv_sec * hz + r.tv_usec / (1000000 / hz))
755 
756 int
757 linux_sys_times(l, v, retval)
758 	struct lwp *l;
759 	void *v;
760 	register_t *retval;
761 {
762 	struct linux_sys_times_args /* {
763 		syscallarg(struct times *) tms;
764 	} */ *uap = v;
765 	struct proc *p = l->l_proc;
766 	struct timeval t;
767 	int error, s;
768 
769 	if (SCARG(uap, tms)) {
770 		struct linux_tms ltms;
771 		struct rusage ru;
772 
773 		calcru(p, &ru.ru_utime, &ru.ru_stime, NULL);
774 		ltms.ltms_utime = CONVTCK(ru.ru_utime);
775 		ltms.ltms_stime = CONVTCK(ru.ru_stime);
776 
777 		ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime);
778 		ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime);
779 
780 		if ((error = copyout(&ltms, SCARG(uap, tms), sizeof ltms)))
781 			return error;
782 	}
783 
784 	s = splclock();
785 	timersub(&time, &boottime, &t);
786 	splx(s);
787 
788 	retval[0] = ((linux_clock_t)(CONVTCK(t)));
789 	return 0;
790 }
791 
792 #undef CONVTCK
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_getdents(l, v, retval)
810 	struct lwp *l;
811 	void *v;
812 	register_t *retval;
813 {
814 	struct linux_sys_getdents_args /* {
815 		syscallarg(int) fd;
816 		syscallarg(struct linux_dirent *) dent;
817 		syscallarg(unsigned int) count;
818 	} */ *uap = v;
819 	struct proc *p = l->l_proc;
820 	struct dirent *bdp;
821 	struct vnode *vp;
822 	caddr_t	inp, tbuf;		/* BSD-format */
823 	int len, reclen;		/* BSD-format */
824 	caddr_t outp;			/* Linux-format */
825 	int resid, linux_reclen = 0;	/* Linux-format */
826 	struct file *fp;
827 	struct uio auio;
828 	struct iovec aiov;
829 	struct linux_dirent idb;
830 	off_t off;		/* true file offset */
831 	int buflen, error, eofflag, nbytes, oldcall;
832 	struct vattr va;
833 	off_t *cookiebuf = NULL, *cookie;
834 	int ncookies;
835 
836 	/* getvnode() will use the descriptor for us */
837 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
838 		return (error);
839 
840 	if ((fp->f_flag & FREAD) == 0) {
841 		error = EBADF;
842 		goto out1;
843 	}
844 
845 	vp = (struct vnode *)fp->f_data;
846 	if (vp->v_type != VDIR) {
847 		error = EINVAL;
848 		goto out1;
849 	}
850 
851 	if ((error = VOP_GETATTR(vp, &va, p->p_ucred, l)))
852 		goto out1;
853 
854 	nbytes = SCARG(uap, count);
855 	if (nbytes == 1) {	/* emulating old, broken behaviour */
856 		nbytes = sizeof (idb);
857 		buflen = max(va.va_blocksize, nbytes);
858 		oldcall = 1;
859 	} else {
860 		buflen = min(MAXBSIZE, nbytes);
861 		if (buflen < va.va_blocksize)
862 			buflen = va.va_blocksize;
863 		oldcall = 0;
864 	}
865 	tbuf = malloc(buflen, M_TEMP, M_WAITOK);
866 
867 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
868 	off = fp->f_offset;
869 again:
870 	aiov.iov_base = tbuf;
871 	aiov.iov_len = buflen;
872 	auio.uio_iov = &aiov;
873 	auio.uio_iovcnt = 1;
874 	auio.uio_rw = UIO_READ;
875 	auio.uio_segflg = UIO_SYSSPACE;
876 	auio.uio_lwp = NULL;
877 	auio.uio_resid = buflen;
878 	auio.uio_offset = off;
879 	/*
880          * First we read into the malloc'ed buffer, then
881          * we massage it into user space, one record at a time.
882          */
883 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf,
884 	    &ncookies);
885 	if (error)
886 		goto out;
887 
888 	inp = tbuf;
889 	outp = (caddr_t)SCARG(uap, dent);
890 	resid = nbytes;
891 	if ((len = buflen - auio.uio_resid) == 0)
892 		goto eof;
893 
894 	for (cookie = cookiebuf; len > 0; len -= reclen) {
895 		bdp = (struct dirent *)inp;
896 		reclen = bdp->d_reclen;
897 		if (reclen & 3)
898 			panic("linux_readdir");
899 		if (bdp->d_fileno == 0) {
900 			inp += reclen;	/* it is a hole; squish it out */
901 			if (cookie)
902 				off = *cookie++;
903 			else
904 				off += reclen;
905 			continue;
906 		}
907 		linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen);
908 		if (reclen > len || resid < linux_reclen) {
909 			/* entry too big for buffer, so just stop */
910 			outp++;
911 			break;
912 		}
913 		/*
914 		 * Massage in place to make a Linux-shaped dirent (otherwise
915 		 * we have to worry about touching user memory outside of
916 		 * the copyout() call).
917 		 */
918 		idb.d_ino = bdp->d_fileno;
919 		/*
920 		 * The old readdir() call misuses the offset and reclen fields.
921 		 */
922 		if (oldcall) {
923 			idb.d_off = (linux_off_t)linux_reclen;
924 			idb.d_reclen = (u_short)bdp->d_namlen;
925 		} else {
926 			if (sizeof (idb.d_off) <= 4 && (off >> 32) != 0) {
927 				compat_offseterr(vp, "linux_getdents");
928 				error = EINVAL;
929 				goto out;
930 			}
931 			idb.d_off = (linux_off_t)off;
932 			idb.d_reclen = (u_short)linux_reclen;
933 		}
934 		strcpy(idb.d_name, bdp->d_name);
935 		if ((error = copyout((caddr_t)&idb, outp, linux_reclen)))
936 			goto out;
937 		/* advance past this real entry */
938 		inp += reclen;
939 		if (cookie)
940 			off = *cookie++; /* each entry points to itself */
941 		else
942 			off += reclen;
943 		/* advance output past Linux-shaped entry */
944 		outp += linux_reclen;
945 		resid -= linux_reclen;
946 		if (oldcall)
947 			break;
948 	}
949 
950 	/* if we squished out the whole block, try again */
951 	if (outp == (caddr_t)SCARG(uap, dent))
952 		goto again;
953 	fp->f_offset = off;	/* update the vnode offset */
954 
955 	if (oldcall)
956 		nbytes = resid + linux_reclen;
957 
958 eof:
959 	*retval = nbytes - resid;
960 out:
961 	VOP_UNLOCK(vp, 0);
962 	if (cookiebuf)
963 		free(cookiebuf, M_TEMP);
964 	free(tbuf, M_TEMP);
965 out1:
966 	FILE_UNUSE(fp, l);
967 	return error;
968 }
969 
970 /*
971  * Even when just using registers to pass arguments to syscalls you can
972  * have 5 of them on the i386. So this newer version of select() does
973  * this.
974  */
975 int
976 linux_sys_select(l, v, retval)
977 	struct lwp *l;
978 	void *v;
979 	register_t *retval;
980 {
981 	struct linux_sys_select_args /* {
982 		syscallarg(int) nfds;
983 		syscallarg(fd_set *) readfds;
984 		syscallarg(fd_set *) writefds;
985 		syscallarg(fd_set *) exceptfds;
986 		syscallarg(struct timeval *) timeout;
987 	} */ *uap = v;
988 
989 	return linux_select1(l, retval, SCARG(uap, nfds), SCARG(uap, readfds),
990 	    SCARG(uap, writefds), SCARG(uap, exceptfds), SCARG(uap, timeout));
991 }
992 
993 /*
994  * Common code for the old and new versions of select(). A couple of
995  * things are important:
996  * 1) return the amount of time left in the 'timeout' parameter
997  * 2) select never returns ERESTART on Linux, always return EINTR
998  */
999 int
1000 linux_select1(l, retval, nfds, readfds, writefds, exceptfds, timeout)
1001 	struct lwp *l;
1002 	register_t *retval;
1003 	int nfds;
1004 	fd_set *readfds, *writefds, *exceptfds;
1005 	struct timeval *timeout;
1006 {
1007 	struct sys_select_args bsa;
1008 	struct proc *p = l->l_proc;
1009 	struct timeval tv0, tv1, utv, *tvp;
1010 	caddr_t sg;
1011 	int error;
1012 
1013 	SCARG(&bsa, nd) = nfds;
1014 	SCARG(&bsa, in) = readfds;
1015 	SCARG(&bsa, ou) = writefds;
1016 	SCARG(&bsa, ex) = exceptfds;
1017 	SCARG(&bsa, tv) = timeout;
1018 
1019 	/*
1020 	 * Store current time for computation of the amount of
1021 	 * time left.
1022 	 */
1023 	if (timeout) {
1024 		if ((error = copyin(timeout, &utv, sizeof(utv))))
1025 			return error;
1026 		if (itimerfix(&utv)) {
1027 			/*
1028 			 * The timeval was invalid.  Convert it to something
1029 			 * valid that will act as it does under Linux.
1030 			 */
1031 			sg = stackgap_init(p, 0);
1032 			tvp = stackgap_alloc(p, &sg, sizeof(utv));
1033 			utv.tv_sec += utv.tv_usec / 1000000;
1034 			utv.tv_usec %= 1000000;
1035 			if (utv.tv_usec < 0) {
1036 				utv.tv_sec -= 1;
1037 				utv.tv_usec += 1000000;
1038 			}
1039 			if (utv.tv_sec < 0)
1040 				timerclear(&utv);
1041 			if ((error = copyout(&utv, tvp, sizeof(utv))))
1042 				return error;
1043 			SCARG(&bsa, tv) = tvp;
1044 		}
1045 		microtime(&tv0);
1046 	}
1047 
1048 	error = sys_select(l, &bsa, retval);
1049 	if (error) {
1050 		/*
1051 		 * See fs/select.c in the Linux kernel.  Without this,
1052 		 * Maelstrom doesn't work.
1053 		 */
1054 		if (error == ERESTART)
1055 			error = EINTR;
1056 		return error;
1057 	}
1058 
1059 	if (timeout) {
1060 		if (*retval) {
1061 			/*
1062 			 * Compute how much time was left of the timeout,
1063 			 * by subtracting the current time and the time
1064 			 * before we started the call, and subtracting
1065 			 * that result from the user-supplied value.
1066 			 */
1067 			microtime(&tv1);
1068 			timersub(&tv1, &tv0, &tv1);
1069 			timersub(&utv, &tv1, &utv);
1070 			if (utv.tv_sec < 0)
1071 				timerclear(&utv);
1072 		} else
1073 			timerclear(&utv);
1074 		if ((error = copyout(&utv, timeout, sizeof(utv))))
1075 			return error;
1076 	}
1077 
1078 	return 0;
1079 }
1080 
1081 /*
1082  * Get the process group of a certain process. Look it up
1083  * and return the value.
1084  */
1085 int
1086 linux_sys_getpgid(l, v, retval)
1087 	struct lwp *l;
1088 	void *v;
1089 	register_t *retval;
1090 {
1091 	struct linux_sys_getpgid_args /* {
1092 		syscallarg(int) pid;
1093 	} */ *uap = v;
1094 	struct proc *p = l->l_proc;
1095 	struct proc *targp;
1096 
1097 	if (SCARG(uap, pid) != 0 && SCARG(uap, pid) != p->p_pid) {
1098 		if ((targp = pfind(SCARG(uap, pid))) == 0)
1099 			return ESRCH;
1100 	}
1101 	else
1102 		targp = p;
1103 
1104 	retval[0] = targp->p_pgid;
1105 	return 0;
1106 }
1107 
1108 /*
1109  * Set the 'personality' (emulation mode) for the current process. Only
1110  * accept the Linux personality here (0). This call is needed because
1111  * the Linux ELF crt0 issues it in an ugly kludge to make sure that
1112  * ELF binaries run in Linux mode, not SVR4 mode.
1113  */
1114 int
1115 linux_sys_personality(l, v, retval)
1116 	struct lwp *l;
1117 	void *v;
1118 	register_t *retval;
1119 {
1120 	struct linux_sys_personality_args /* {
1121 		syscallarg(int) per;
1122 	} */ *uap = v;
1123 
1124 	if (SCARG(uap, per) != 0)
1125 		return EINVAL;
1126 	retval[0] = 0;
1127 	return 0;
1128 }
1129 
1130 #if defined(__i386__) || defined(__m68k__)
1131 /*
1132  * The calls are here because of type conversions.
1133  */
1134 int
1135 linux_sys_setreuid16(l, v, retval)
1136 	struct lwp *l;
1137 	void *v;
1138 	register_t *retval;
1139 {
1140 	struct linux_sys_setreuid16_args /* {
1141 		syscallarg(int) ruid;
1142 		syscallarg(int) euid;
1143 	} */ *uap = v;
1144 	struct sys_setreuid_args bsa;
1145 
1146 	SCARG(&bsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
1147 		(uid_t)-1 : SCARG(uap, ruid);
1148 	SCARG(&bsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
1149 		(uid_t)-1 : SCARG(uap, euid);
1150 
1151 	return sys_setreuid(l, &bsa, retval);
1152 }
1153 
1154 int
1155 linux_sys_setregid16(l, v, retval)
1156 	struct lwp *l;
1157 	void *v;
1158 	register_t *retval;
1159 {
1160 	struct linux_sys_setregid16_args /* {
1161 		syscallarg(int) rgid;
1162 		syscallarg(int) egid;
1163 	} */ *uap = v;
1164 	struct sys_setregid_args bsa;
1165 
1166 	SCARG(&bsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
1167 		(uid_t)-1 : SCARG(uap, rgid);
1168 	SCARG(&bsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
1169 		(uid_t)-1 : SCARG(uap, egid);
1170 
1171 	return sys_setregid(l, &bsa, retval);
1172 }
1173 
1174 int
1175 linux_sys_setresuid16(l, v, retval)
1176 	struct lwp *l;
1177 	void *v;
1178 	register_t *retval;
1179 {
1180 	struct linux_sys_setresuid16_args /* {
1181 		syscallarg(uid_t) ruid;
1182 		syscallarg(uid_t) euid;
1183 		syscallarg(uid_t) suid;
1184 	} */ *uap = v;
1185 	struct linux_sys_setresuid16_args lsa;
1186 
1187 	SCARG(&lsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
1188 		(uid_t)-1 : SCARG(uap, ruid);
1189 	SCARG(&lsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
1190 		(uid_t)-1 : SCARG(uap, euid);
1191 	SCARG(&lsa, suid) = ((linux_uid_t)SCARG(uap, suid) == (linux_uid_t)-1) ?
1192 		(uid_t)-1 : SCARG(uap, suid);
1193 
1194 	return linux_sys_setresuid(l, &lsa, retval);
1195 }
1196 
1197 int
1198 linux_sys_setresgid16(l, v, retval)
1199 	struct lwp *l;
1200 	void *v;
1201 	register_t *retval;
1202 {
1203 	struct linux_sys_setresgid16_args /* {
1204 		syscallarg(gid_t) rgid;
1205 		syscallarg(gid_t) egid;
1206 		syscallarg(gid_t) sgid;
1207 	} */ *uap = v;
1208 	struct linux_sys_setresgid16_args lsa;
1209 
1210 	SCARG(&lsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
1211 		(gid_t)-1 : SCARG(uap, rgid);
1212 	SCARG(&lsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
1213 		(gid_t)-1 : SCARG(uap, egid);
1214 	SCARG(&lsa, sgid) = ((linux_gid_t)SCARG(uap, sgid) == (linux_gid_t)-1) ?
1215 		(gid_t)-1 : SCARG(uap, sgid);
1216 
1217 	return linux_sys_setresgid(l, &lsa, retval);
1218 }
1219 
1220 int
1221 linux_sys_getgroups16(l, v, retval)
1222 	struct lwp *l;
1223 	void *v;
1224 	register_t *retval;
1225 {
1226 	struct linux_sys_getgroups16_args /* {
1227 		syscallarg(int) gidsetsize;
1228 		syscallarg(linux_gid_t *) gidset;
1229 	} */ *uap = v;
1230 	struct proc *p = l->l_proc;
1231 	caddr_t sg;
1232 	int n, error, i;
1233 	struct sys_getgroups_args bsa;
1234 	gid_t *bset, *kbset;
1235 	linux_gid_t *lset;
1236 	struct pcred *pc = p->p_cred;
1237 
1238 	n = SCARG(uap, gidsetsize);
1239 	if (n < 0)
1240 		return EINVAL;
1241 	error = 0;
1242 	bset = kbset = NULL;
1243 	lset = NULL;
1244 	if (n > 0) {
1245 		n = min(pc->pc_ucred->cr_ngroups, n);
1246 		sg = stackgap_init(p, 0);
1247 		bset = stackgap_alloc(p, &sg, n * sizeof (gid_t));
1248 		kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK);
1249 		lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
1250 		if (bset == NULL || kbset == NULL || lset == NULL)
1251 			return ENOMEM;
1252 		SCARG(&bsa, gidsetsize) = n;
1253 		SCARG(&bsa, gidset) = bset;
1254 		error = sys_getgroups(l, &bsa, retval);
1255 		if (error != 0)
1256 			goto out;
1257 		error = copyin(bset, kbset, n * sizeof (gid_t));
1258 		if (error != 0)
1259 			goto out;
1260 		for (i = 0; i < n; i++)
1261 			lset[i] = (linux_gid_t)kbset[i];
1262 		error = copyout(lset, SCARG(uap, gidset),
1263 		    n * sizeof (linux_gid_t));
1264 	} else
1265 		*retval = pc->pc_ucred->cr_ngroups;
1266 out:
1267 	if (kbset != NULL)
1268 		free(kbset, M_TEMP);
1269 	if (lset != NULL)
1270 		free(lset, M_TEMP);
1271 	return error;
1272 }
1273 
1274 int
1275 linux_sys_setgroups16(l, v, retval)
1276 	struct lwp *l;
1277 	void *v;
1278 	register_t *retval;
1279 {
1280 	struct linux_sys_setgroups16_args /* {
1281 		syscallarg(int) gidsetsize;
1282 		syscallarg(linux_gid_t *) gidset;
1283 	} */ *uap = v;
1284 	struct proc *p = l->l_proc;
1285 	caddr_t sg;
1286 	int n;
1287 	int error, i;
1288 	struct sys_setgroups_args bsa;
1289 	gid_t *bset, *kbset;
1290 	linux_gid_t *lset;
1291 
1292 	n = SCARG(uap, gidsetsize);
1293 	if (n < 0 || n > NGROUPS)
1294 		return EINVAL;
1295 	sg = stackgap_init(p, 0);
1296 	bset = stackgap_alloc(p, &sg, n * sizeof (gid_t));
1297 	lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
1298 	kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK);
1299 	if (lset == NULL || bset == NULL)
1300 		return ENOMEM;
1301 	error = copyin(SCARG(uap, gidset), lset, n * sizeof (linux_gid_t));
1302 	if (error != 0)
1303 		goto out;
1304 	for (i = 0; i < n; i++)
1305 		kbset[i] = (gid_t)lset[i];
1306 	error = copyout(kbset, bset, n * sizeof (gid_t));
1307 	if (error != 0)
1308 		goto out;
1309 	SCARG(&bsa, gidsetsize) = n;
1310 	SCARG(&bsa, gidset) = bset;
1311 	error = sys_setgroups(l, &bsa, retval);
1312 
1313 out:
1314 	if (lset != NULL)
1315 		free(lset, M_TEMP);
1316 	if (kbset != NULL)
1317 		free(kbset, M_TEMP);
1318 
1319 	return error;
1320 }
1321 
1322 #endif /* __i386__ || __m68k__ || __amd64__ */
1323 
1324 /*
1325  * We have nonexistent fsuid equal to uid.
1326  * If modification is requested, refuse.
1327  */
1328 int
1329 linux_sys_setfsuid(l, v, retval)
1330 	 struct lwp *l;
1331 	 void *v;
1332 	 register_t *retval;
1333 {
1334 	 struct linux_sys_setfsuid_args /* {
1335 		 syscallarg(uid_t) uid;
1336 	 } */ *uap = v;
1337 	 struct proc *p = l->l_proc;
1338 	 uid_t uid;
1339 
1340 	 uid = SCARG(uap, uid);
1341 	 if (p->p_cred->p_ruid != uid)
1342 		 return sys_nosys(l, v, retval);
1343 	 else
1344 		 return (0);
1345 }
1346 
1347 /* XXX XXX XXX */
1348 #ifndef alpha
1349 int
1350 linux_sys_getfsuid(l, v, retval)
1351 	struct lwp *l;
1352 	void *v;
1353 	register_t *retval;
1354 {
1355 	return sys_getuid(l, v, retval);
1356 }
1357 #endif
1358 
1359 int
1360 linux_sys_setresuid(l, v, retval)
1361 	struct lwp *l;
1362 	void *v;
1363 	register_t *retval;
1364 {
1365 	struct linux_sys_setresuid_args /* {
1366 		syscallarg(uid_t) ruid;
1367 		syscallarg(uid_t) euid;
1368 		syscallarg(uid_t) suid;
1369 	} */ *uap = v;
1370 
1371 	/*
1372 	 * Note: These checks are a little different than the NetBSD
1373 	 * setreuid(2) call performs.  This precisely follows the
1374 	 * behavior of the Linux kernel.
1375 	 */
1376 
1377 	return do_setresuid(l, SCARG(uap, ruid), SCARG(uap, euid),
1378 			    SCARG(uap, suid),
1379 			    ID_R_EQ_R | ID_R_EQ_E | ID_R_EQ_S |
1380 			    ID_E_EQ_R | ID_E_EQ_E | ID_E_EQ_S |
1381 			    ID_S_EQ_R | ID_S_EQ_E | ID_S_EQ_S );
1382 }
1383 
1384 int
1385 linux_sys_getresuid(l, v, retval)
1386 	struct lwp *l;
1387 	void *v;
1388 	register_t *retval;
1389 {
1390 	struct linux_sys_getresuid_args /* {
1391 		syscallarg(uid_t *) ruid;
1392 		syscallarg(uid_t *) euid;
1393 		syscallarg(uid_t *) suid;
1394 	} */ *uap = v;
1395 	struct proc *p = l->l_proc;
1396 	struct pcred *pc = p->p_cred;
1397 	int error;
1398 
1399 	/*
1400 	 * Linux copies these values out to userspace like so:
1401 	 *
1402 	 *	1. Copy out ruid.
1403 	 *	2. If that succeeds, copy out euid.
1404 	 *	3. If both of those succeed, copy out suid.
1405 	 */
1406 	if ((error = copyout(&pc->p_ruid, SCARG(uap, ruid),
1407 			     sizeof(uid_t))) != 0)
1408 		return (error);
1409 
1410 	if ((error = copyout(&pc->pc_ucred->cr_uid, SCARG(uap, euid),
1411 			     sizeof(uid_t))) != 0)
1412 		return (error);
1413 
1414 	return (copyout(&pc->p_svuid, SCARG(uap, suid), sizeof(uid_t)));
1415 }
1416 
1417 int
1418 linux_sys_ptrace(l, v, retval)
1419 	struct lwp *l;
1420 	void *v;
1421 	register_t *retval;
1422 {
1423 	struct linux_sys_ptrace_args /* {
1424 		i386, m68k, powerpc: T=int
1425 		alpha, amd64: T=long
1426 		syscallarg(T) request;
1427 		syscallarg(T) pid;
1428 		syscallarg(T) addr;
1429 		syscallarg(T) data;
1430 	} */ *uap = v;
1431 	const int *ptr;
1432 	int request;
1433 	int error;
1434 
1435 	ptr = linux_ptrace_request_map;
1436 	request = SCARG(uap, request);
1437 	while (*ptr != -1)
1438 		if (*ptr++ == request) {
1439 			struct sys_ptrace_args pta;
1440 
1441 			SCARG(&pta, req) = *ptr;
1442 			SCARG(&pta, pid) = SCARG(uap, pid);
1443 			SCARG(&pta, addr) = (caddr_t)SCARG(uap, addr);
1444 			SCARG(&pta, data) = SCARG(uap, data);
1445 
1446 			/*
1447 			 * Linux ptrace(PTRACE_CONT, pid, 0, 0) means actually
1448 			 * to continue where the process left off previously.
1449 			 * The same thing is achieved by addr == (caddr_t) 1
1450 			 * on NetBSD, so rewrite 'addr' appropriately.
1451 			 */
1452 			if (request == LINUX_PTRACE_CONT && SCARG(uap, addr)==0)
1453 				SCARG(&pta, addr) = (caddr_t) 1;
1454 
1455 			error = sys_ptrace(l, &pta, retval);
1456 			if (error)
1457 				return error;
1458 			switch (request) {
1459 			case LINUX_PTRACE_PEEKTEXT:
1460 			case LINUX_PTRACE_PEEKDATA:
1461 				error = copyout (retval,
1462 				    (caddr_t)SCARG(uap, data),
1463 				    sizeof *retval);
1464 				*retval = SCARG(uap, data);
1465 				break;
1466 			default:
1467 				break;
1468 			}
1469 			return error;
1470 		}
1471 		else
1472 			ptr++;
1473 
1474 	return LINUX_SYS_PTRACE_ARCH(l, uap, retval);
1475 }
1476 
1477 int
1478 linux_sys_reboot(struct lwp *l, void *v, register_t *retval)
1479 {
1480 	struct linux_sys_reboot_args /* {
1481 		syscallarg(int) magic1;
1482 		syscallarg(int) magic2;
1483 		syscallarg(int) cmd;
1484 		syscallarg(void *) arg;
1485 	} */ *uap = v;
1486 	struct sys_reboot_args /* {
1487 		syscallarg(int) opt;
1488 		syscallarg(char *) bootstr;
1489 	} */ sra;
1490 	struct proc *p = l->l_proc;
1491 	int error;
1492 
1493 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1494 		return(error);
1495 
1496 	if (SCARG(uap, magic1) != LINUX_REBOOT_MAGIC1)
1497 		return(EINVAL);
1498 	if (SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2 &&
1499 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2A &&
1500 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2B)
1501 		return(EINVAL);
1502 
1503 	switch (SCARG(uap, cmd)) {
1504 	case LINUX_REBOOT_CMD_RESTART:
1505 		SCARG(&sra, opt) = RB_AUTOBOOT;
1506 		break;
1507 	case LINUX_REBOOT_CMD_HALT:
1508 		SCARG(&sra, opt) = RB_HALT;
1509 		break;
1510 	case LINUX_REBOOT_CMD_POWER_OFF:
1511 		SCARG(&sra, opt) = RB_HALT|RB_POWERDOWN;
1512 		break;
1513 	case LINUX_REBOOT_CMD_RESTART2:
1514 		/* Reboot with an argument. */
1515 		SCARG(&sra, opt) = RB_AUTOBOOT|RB_STRING;
1516 		SCARG(&sra, bootstr) = SCARG(uap, arg);
1517 		break;
1518 	case LINUX_REBOOT_CMD_CAD_ON:
1519 		return(EINVAL);	/* We don't implement ctrl-alt-delete */
1520 	case LINUX_REBOOT_CMD_CAD_OFF:
1521 		return(0);
1522 	default:
1523 		return(EINVAL);
1524 	}
1525 
1526 	return(sys_reboot(l, &sra, retval));
1527 }
1528 
1529 /*
1530  * Copy of compat_12_sys_swapon().
1531  */
1532 int
1533 linux_sys_swapon(l, v, retval)
1534 	struct lwp *l;
1535 	void *v;
1536 	register_t *retval;
1537 {
1538 	struct sys_swapctl_args ua;
1539 	struct linux_sys_swapon_args /* {
1540 		syscallarg(const char *) name;
1541 	} */ *uap = v;
1542 
1543 	SCARG(&ua, cmd) = SWAP_ON;
1544 	SCARG(&ua, arg) = (void *)__UNCONST(SCARG(uap, name));
1545 	SCARG(&ua, misc) = 0;	/* priority */
1546 	return (sys_swapctl(l, &ua, retval));
1547 }
1548 
1549 /*
1550  * Stop swapping to the file or block device specified by path.
1551  */
1552 int
1553 linux_sys_swapoff(l, v, retval)
1554 	struct lwp *l;
1555 	void *v;
1556 	register_t *retval;
1557 {
1558 	struct sys_swapctl_args ua;
1559 	struct linux_sys_swapoff_args /* {
1560 		syscallarg(const char *) path;
1561 	} */ *uap = v;
1562 
1563 	SCARG(&ua, cmd) = SWAP_OFF;
1564 	SCARG(&ua, arg) = __UNCONST(SCARG(uap, path)); /*XXXUNCONST*/
1565 	return (sys_swapctl(l, &ua, retval));
1566 }
1567 
1568 /*
1569  * Copy of compat_09_sys_setdomainname()
1570  */
1571 /* ARGSUSED */
1572 int
1573 linux_sys_setdomainname(l, v, retval)
1574 	struct lwp *l;
1575 	void *v;
1576 	register_t *retval;
1577 {
1578 	struct linux_sys_setdomainname_args /* {
1579 		syscallarg(char *) domainname;
1580 		syscallarg(int) len;
1581 	} */ *uap = v;
1582 	int name[2];
1583 
1584 	name[0] = CTL_KERN;
1585 	name[1] = KERN_DOMAINNAME;
1586 	return (old_sysctl(&name[0], 2, 0, 0, SCARG(uap, domainname),
1587 			    SCARG(uap, len), l));
1588 }
1589 
1590 /*
1591  * sysinfo()
1592  */
1593 /* ARGSUSED */
1594 int
1595 linux_sys_sysinfo(l, v, retval)
1596 	struct lwp *l;
1597 	void *v;
1598 	register_t *retval;
1599 {
1600 	struct linux_sys_sysinfo_args /* {
1601 		syscallarg(struct linux_sysinfo *) arg;
1602 	} */ *uap = v;
1603 	struct linux_sysinfo si;
1604 	struct loadavg *la;
1605 
1606 	si.uptime = time.tv_sec - boottime.tv_sec;
1607 	la = &averunnable;
1608 	si.loads[0] = la->ldavg[0] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1609 	si.loads[1] = la->ldavg[1] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1610 	si.loads[2] = la->ldavg[2] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1611 	si.totalram = ctob(physmem);
1612 	si.freeram = uvmexp.free * uvmexp.pagesize;
1613 	si.sharedram = 0;	/* XXX */
1614 	si.bufferram = uvmexp.filepages * uvmexp.pagesize;
1615 	si.totalswap = uvmexp.swpages * uvmexp.pagesize;
1616 	si.freeswap = (uvmexp.swpages - uvmexp.swpginuse) * uvmexp.pagesize;
1617 	si.procs = nprocs;
1618 
1619 	/* The following are only present in newer Linux kernels. */
1620 	si.totalbig = 0;
1621 	si.freebig = 0;
1622 	si.mem_unit = 1;
1623 
1624 	return (copyout(&si, SCARG(uap, arg), sizeof si));
1625 }
1626 
1627 #ifdef LINUX_LARGEFILE64
1628 #define bsd_to_linux_rlimit1(l, b, f) \
1629     (l)->f = ((b)->f == RLIM_INFINITY || \
1630 	     ((b)->f & 0x8000000000000000UL) != 0) ? \
1631     LINUX_RLIM_INFINITY : (b)->f
1632 #else
1633 #define bsd_to_linux_rlimit1(l, b, f) \
1634     (l)->f = ((b)->f == RLIM_INFINITY || \
1635 	     ((b)->f & 0xffffffff00000000ULL) != 0) ? \
1636     LINUX_RLIM_INFINITY : (int32_t)(b)->f
1637 #endif
1638 #define bsd_to_linux_rlimit(l, b) \
1639     bsd_to_linux_rlimit1(l, b, rlim_cur); \
1640     bsd_to_linux_rlimit1(l, b, rlim_max)
1641 
1642 #define linux_to_bsd_rlimit1(b, l, f) \
1643     (b)->f = (l)->f == LINUX_RLIM_INFINITY ? RLIM_INFINITY : (l)->f
1644 #define linux_to_bsd_rlimit(b, l) \
1645     linux_to_bsd_rlimit1(b, l, rlim_cur); \
1646     linux_to_bsd_rlimit1(b, l, rlim_max)
1647 
1648 static int
1649 linux_to_bsd_limit(lim)
1650 	int lim;
1651 {
1652 	switch (lim) {
1653 	case LINUX_RLIMIT_CPU:
1654 		return RLIMIT_CPU;
1655 	case LINUX_RLIMIT_FSIZE:
1656 		return RLIMIT_FSIZE;
1657 	case LINUX_RLIMIT_DATA:
1658 		return RLIMIT_DATA;
1659 	case LINUX_RLIMIT_STACK:
1660 		return RLIMIT_STACK;
1661 	case LINUX_RLIMIT_CORE:
1662 		return RLIMIT_CORE;
1663 	case LINUX_RLIMIT_RSS:
1664 		return RLIMIT_RSS;
1665 	case LINUX_RLIMIT_NPROC:
1666 		return RLIMIT_NPROC;
1667 	case LINUX_RLIMIT_NOFILE:
1668 		return RLIMIT_NOFILE;
1669 	case LINUX_RLIMIT_MEMLOCK:
1670 		return RLIMIT_MEMLOCK;
1671 	case LINUX_RLIMIT_AS:
1672 	case LINUX_RLIMIT_LOCKS:
1673 		return -EOPNOTSUPP;
1674 	default:
1675 		return -EINVAL;
1676 	}
1677 }
1678 
1679 
1680 int
1681 linux_sys_getrlimit(l, v, retval)
1682 	struct lwp *l;
1683 	void *v;
1684 	register_t *retval;
1685 {
1686 	struct linux_sys_getrlimit_args /* {
1687 		syscallarg(int) which;
1688 #ifdef LINUX_LARGEFILE64
1689 		syscallarg(struct rlimit *) rlp;
1690 #else
1691 		syscallarg(struct orlimit *) rlp;
1692 #endif
1693 	} */ *uap = v;
1694 	struct proc *p = l->l_proc;
1695 	caddr_t sg = stackgap_init(p, 0);
1696 	struct sys_getrlimit_args ap;
1697 	struct rlimit rl;
1698 #ifdef LINUX_LARGEFILE64
1699 	struct rlimit orl;
1700 #else
1701 	struct orlimit orl;
1702 #endif
1703 	int error;
1704 
1705 	SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which));
1706 	if ((error = SCARG(&ap, which)) < 0)
1707 		return -error;
1708 	SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl);
1709 	if ((error = sys_getrlimit(l, &ap, retval)) != 0)
1710 		return error;
1711 	if ((error = copyin(SCARG(&ap, rlp), &rl, sizeof(rl))) != 0)
1712 		return error;
1713 	bsd_to_linux_rlimit(&orl, &rl);
1714 
1715 	return copyout(&orl, SCARG(uap, rlp), sizeof(orl));
1716 }
1717 
1718 int
1719 linux_sys_setrlimit(l, v, retval)
1720 	struct lwp *l;
1721 	void *v;
1722 	register_t *retval;
1723 {
1724 	struct linux_sys_setrlimit_args /* {
1725 		syscallarg(int) which;
1726 #ifdef LINUX_LARGEFILE64
1727 		syscallarg(struct rlimit *) rlp;
1728 #else
1729 		syscallarg(struct orlimit *) rlp;
1730 #endif
1731 	} */ *uap = v;
1732 	struct proc *p = l->l_proc;
1733 	caddr_t sg = stackgap_init(p, 0);
1734 	struct sys_getrlimit_args ap;
1735 	struct rlimit rl;
1736 #ifdef LINUX_LARGEFILE64
1737 	struct rlimit orl;
1738 #else
1739 	struct orlimit orl;
1740 #endif
1741 	int error;
1742 
1743 	SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which));
1744 	SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl);
1745 	if ((error = SCARG(&ap, which)) < 0)
1746 		return -error;
1747 	if ((error = copyin(SCARG(uap, rlp), &orl, sizeof(orl))) != 0)
1748 		return error;
1749 	linux_to_bsd_rlimit(&rl, &orl);
1750 	if ((error = copyout(&rl, SCARG(&ap, rlp), sizeof(rl))) != 0)
1751 		return error;
1752 	return sys_setrlimit(l, &ap, retval);
1753 }
1754 
1755 #if !defined(__mips__) && !defined(__amd64__)
1756 /* XXX: this doesn't look 100% common, at least mips doesn't have it */
1757 int
1758 linux_sys_ugetrlimit(l, v, retval)
1759 	struct lwp *l;
1760 	void *v;
1761 	register_t *retval;
1762 {
1763 	return linux_sys_getrlimit(l, v, retval);
1764 }
1765 #endif
1766 
1767 /*
1768  * This gets called for unsupported syscalls. The difference to sys_nosys()
1769  * is that process does not get SIGSYS, the call just returns with ENOSYS.
1770  * This is the way Linux does it and glibc depends on this behaviour.
1771  */
1772 int
1773 linux_sys_nosys(l, v, retval)
1774 	struct lwp *l;
1775 	void *v;
1776 	register_t *retval;
1777 {
1778 	return (ENOSYS);
1779 }
1780