xref: /netbsd-src/sys/compat/linux/common/linux_misc.c (revision f81322cf185a4db50f71fcf7701f20198272620e)
1 /*	$NetBSD: linux_misc.c,v 1.152 2006/03/17 06:01:14 erh 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 1.152 2006/03/17 06:01:14 erh 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 #ifndef COMPAT_LINUX32
117 #include <compat/linux/common/linux_limit.h>
118 #endif
119 #include <compat/linux/common/linux_ptrace.h>
120 #include <compat/linux/common/linux_reboot.h>
121 #include <compat/linux/common/linux_emuldata.h>
122 
123 #ifndef COMPAT_LINUX32
124 const int linux_ptrace_request_map[] = {
125 	LINUX_PTRACE_TRACEME,	PT_TRACE_ME,
126 	LINUX_PTRACE_PEEKTEXT,	PT_READ_I,
127 	LINUX_PTRACE_PEEKDATA,	PT_READ_D,
128 	LINUX_PTRACE_POKETEXT,	PT_WRITE_I,
129 	LINUX_PTRACE_POKEDATA,	PT_WRITE_D,
130 	LINUX_PTRACE_CONT,	PT_CONTINUE,
131 	LINUX_PTRACE_KILL,	PT_KILL,
132 	LINUX_PTRACE_ATTACH,	PT_ATTACH,
133 	LINUX_PTRACE_DETACH,	PT_DETACH,
134 # ifdef PT_STEP
135 	LINUX_PTRACE_SINGLESTEP,	PT_STEP,
136 # endif
137 	-1
138 };
139 
140 const struct linux_mnttypes linux_fstypes[] = {
141 	{ MOUNT_FFS,		LINUX_DEFAULT_SUPER_MAGIC	},
142 	{ MOUNT_NFS,		LINUX_NFS_SUPER_MAGIC 		},
143 	{ MOUNT_MFS,		LINUX_DEFAULT_SUPER_MAGIC	},
144 	{ MOUNT_MSDOS,		LINUX_MSDOS_SUPER_MAGIC		},
145 	{ MOUNT_LFS,		LINUX_DEFAULT_SUPER_MAGIC	},
146 	{ MOUNT_FDESC,		LINUX_DEFAULT_SUPER_MAGIC	},
147 	{ MOUNT_PORTAL,		LINUX_DEFAULT_SUPER_MAGIC	},
148 	{ MOUNT_NULL,		LINUX_DEFAULT_SUPER_MAGIC	},
149 	{ MOUNT_OVERLAY,	LINUX_DEFAULT_SUPER_MAGIC	},
150 	{ MOUNT_UMAP,		LINUX_DEFAULT_SUPER_MAGIC	},
151 	{ MOUNT_KERNFS,		LINUX_DEFAULT_SUPER_MAGIC	},
152 	{ MOUNT_PROCFS,		LINUX_PROC_SUPER_MAGIC		},
153 	{ MOUNT_AFS,		LINUX_DEFAULT_SUPER_MAGIC	},
154 	{ MOUNT_CD9660,		LINUX_ISOFS_SUPER_MAGIC		},
155 	{ MOUNT_UNION,		LINUX_DEFAULT_SUPER_MAGIC	},
156 	{ MOUNT_ADOSFS,		LINUX_ADFS_SUPER_MAGIC		},
157 	{ MOUNT_EXT2FS,		LINUX_EXT2_SUPER_MAGIC		},
158 	{ MOUNT_CFS,		LINUX_DEFAULT_SUPER_MAGIC	},
159 	{ MOUNT_CODA,		LINUX_CODA_SUPER_MAGIC		},
160 	{ MOUNT_FILECORE,	LINUX_DEFAULT_SUPER_MAGIC	},
161 	{ MOUNT_NTFS,		LINUX_DEFAULT_SUPER_MAGIC	},
162 	{ MOUNT_SMBFS,		LINUX_SMB_SUPER_MAGIC		},
163 	{ MOUNT_PTYFS,		LINUX_DEVPTS_SUPER_MAGIC	},
164 	{ MOUNT_TMPFS,		LINUX_DEFAULT_SUPER_MAGIC	}
165 };
166 const int linux_fstypes_cnt = sizeof(linux_fstypes) / sizeof(linux_fstypes[0]);
167 
168 # ifdef DEBUG_LINUX
169 #define DPRINTF(a)	uprintf a
170 # else
171 #define DPRINTF(a)
172 # endif
173 
174 /* Local linux_misc.c functions: */
175 # ifndef __amd64__
176 static void bsd_to_linux_statfs __P((const struct statvfs *,
177     struct linux_statfs *));
178 # endif
179 static void linux_to_bsd_mmap_args __P((struct sys_mmap_args *,
180     const struct linux_sys_mmap_args *));
181 static int linux_mmap __P((struct lwp *, struct linux_sys_mmap_args *,
182     register_t *, off_t));
183 
184 
185 /*
186  * The information on a terminated (or stopped) process needs
187  * to be converted in order for Linux binaries to get a valid signal
188  * number out of it.
189  */
190 void
191 bsd_to_linux_wstat(st)
192 	int *st;
193 {
194 
195 	int sig;
196 
197 	if (WIFSIGNALED(*st)) {
198 		sig = WTERMSIG(*st);
199 		if (sig >= 0 && sig < NSIG)
200 			*st= (*st& ~0177) | native_to_linux_signo[sig];
201 	} else if (WIFSTOPPED(*st)) {
202 		sig = WSTOPSIG(*st);
203 		if (sig >= 0 && sig < NSIG)
204 			*st = (*st & ~0xff00) |
205 			    (native_to_linux_signo[sig] << 8);
206 	}
207 }
208 
209 /*
210  * wait4(2).  Passed on to the NetBSD call, surrounded by code to
211  * reserve some space for a NetBSD-style wait status, and converting
212  * it to what Linux wants.
213  */
214 int
215 linux_sys_wait4(l, v, retval)
216 	struct lwp *l;
217 	void *v;
218 	register_t *retval;
219 {
220 	struct linux_sys_wait4_args /* {
221 		syscallarg(int) pid;
222 		syscallarg(int *) status;
223 		syscallarg(int) options;
224 		syscallarg(struct rusage *) rusage;
225 	} */ *uap = v;
226 	struct proc *p = l->l_proc;
227 	struct sys_wait4_args w4a;
228 	int error, *status, tstat, options, linux_options;
229 	caddr_t sg;
230 
231 	if (SCARG(uap, status) != NULL) {
232 		sg = stackgap_init(p, 0);
233 		status = (int *) stackgap_alloc(p, &sg, sizeof *status);
234 	} else
235 		status = NULL;
236 
237 	linux_options = SCARG(uap, options);
238 	options = 0;
239 	if (linux_options & ~(LINUX_WAIT4_KNOWNFLAGS))
240 		return (EINVAL);
241 
242 	if (linux_options & LINUX_WAIT4_WNOHANG)
243 		options |= WNOHANG;
244 	if (linux_options & LINUX_WAIT4_WUNTRACED)
245 		options |= WUNTRACED;
246 	if (linux_options & LINUX_WAIT4_WALL)
247 		options |= WALLSIG;
248 	if (linux_options & LINUX_WAIT4_WCLONE)
249 		options |= WALTSIG;
250 # ifdef DIAGNOSTIC
251 	if (linux_options & LINUX_WAIT4_WNOTHREAD)
252 		printf("WARNING: %s: linux process %d.%d called "
253 		       "waitpid with __WNOTHREAD set!",
254 		       __FILE__, p->p_pid, l->l_lid);
255 
256 # endif
257 
258 	SCARG(&w4a, pid) = SCARG(uap, pid);
259 	SCARG(&w4a, status) = status;
260 	SCARG(&w4a, options) = options;
261 	SCARG(&w4a, rusage) = SCARG(uap, rusage);
262 
263 	if ((error = sys_wait4(l, &w4a, retval)))
264 		return error;
265 
266 	sigdelset(&p->p_sigctx.ps_siglist, SIGCHLD);
267 
268 	if (status != NULL) {
269 		if ((error = copyin(status, &tstat, sizeof tstat)))
270 			return error;
271 
272 		bsd_to_linux_wstat(&tstat);
273 		return copyout(&tstat, SCARG(uap, status), sizeof tstat);
274 	}
275 
276 	return 0;
277 }
278 
279 /*
280  * Linux brk(2). The check if the new address is >= the old one is
281  * done in the kernel in Linux. NetBSD does it in the library.
282  */
283 int
284 linux_sys_brk(l, v, retval)
285 	struct lwp *l;
286 	void *v;
287 	register_t *retval;
288 {
289 	struct linux_sys_brk_args /* {
290 		syscallarg(char *) nsize;
291 	} */ *uap = v;
292 	struct proc *p = l->l_proc;
293 	char *nbrk = SCARG(uap, nsize);
294 	struct sys_obreak_args oba;
295 	struct vmspace *vm = p->p_vmspace;
296 	struct linux_emuldata *ed = (struct linux_emuldata*)p->p_emuldata;
297 
298 	SCARG(&oba, nsize) = nbrk;
299 
300 	if ((caddr_t) nbrk > vm->vm_daddr && sys_obreak(l, &oba, retval) == 0)
301 		ed->s->p_break = (char*)nbrk;
302 	else
303 		nbrk = ed->s->p_break;
304 
305 	retval[0] = (register_t)nbrk;
306 
307 	return 0;
308 }
309 
310 # ifndef __amd64__
311 /*
312  * Convert NetBSD statvfs structure to Linux statfs structure.
313  * Linux doesn't have f_flag, and we can't set f_frsize due
314  * to glibc statvfs() bug (see below).
315  */
316 static void
317 bsd_to_linux_statfs(bsp, lsp)
318 	const struct statvfs *bsp;
319 	struct linux_statfs *lsp;
320 {
321 	int i;
322 
323 	for (i = 0; i < linux_fstypes_cnt; i++) {
324 		if (strcmp(bsp->f_fstypename, linux_fstypes[i].bsd) == 0) {
325 			lsp->l_ftype = linux_fstypes[i].linux;
326 			break;
327 		}
328 	}
329 
330 	if (i == linux_fstypes_cnt) {
331 		DPRINTF(("unhandled fstype in linux emulation: %s\n",
332 		    bsp->f_fstypename));
333 		lsp->l_ftype = LINUX_DEFAULT_SUPER_MAGIC;
334 	}
335 
336 	/*
337 	 * The sizes are expressed in number of blocks. The block
338 	 * size used for the size is f_frsize for POSIX-compliant
339 	 * statvfs. Linux statfs uses f_bsize as the block size
340 	 * (f_frsize used to not be available in Linux struct statfs).
341 	 * However, glibc 2.3.3 statvfs() wrapper fails to adjust the block
342 	 * counts for different f_frsize if f_frsize is provided by the kernel.
343 	 * POSIX conforming apps thus get wrong size if f_frsize
344 	 * is different to f_bsize. Thus, we just pretend we don't
345 	 * support f_frsize.
346 	 */
347 
348 	lsp->l_fbsize = bsp->f_frsize;
349 	lsp->l_ffrsize = 0;			/* compat */
350 	lsp->l_fblocks = bsp->f_blocks;
351 	lsp->l_fbfree = bsp->f_bfree;
352 	lsp->l_fbavail = bsp->f_bavail;
353 	lsp->l_ffiles = bsp->f_files;
354 	lsp->l_fffree = bsp->f_ffree;
355 	/* Linux sets the fsid to 0..., we don't */
356 	lsp->l_ffsid.val[0] = bsp->f_fsidx.__fsid_val[0];
357 	lsp->l_ffsid.val[1] = bsp->f_fsidx.__fsid_val[1];
358 	lsp->l_fnamelen = bsp->f_namemax;
359 	(void)memset(lsp->l_fspare, 0, sizeof(lsp->l_fspare));
360 }
361 
362 /*
363  * Implement the fs stat functions. Straightforward.
364  */
365 int
366 linux_sys_statfs(l, v, retval)
367 	struct lwp *l;
368 	void *v;
369 	register_t *retval;
370 {
371 	struct linux_sys_statfs_args /* {
372 		syscallarg(const char *) path;
373 		syscallarg(struct linux_statfs *) sp;
374 	} */ *uap = v;
375 	struct proc *p = l->l_proc;
376 	struct statvfs btmp, *bsp;
377 	struct linux_statfs ltmp;
378 	struct sys_statvfs1_args bsa;
379 	caddr_t sg;
380 	int error;
381 
382 	sg = stackgap_init(p, 0);
383 	bsp = (struct statvfs *) stackgap_alloc(p, &sg, sizeof (struct statvfs));
384 
385 	CHECK_ALT_EXIST(l, &sg, SCARG(uap, path));
386 
387 	SCARG(&bsa, path) = SCARG(uap, path);
388 	SCARG(&bsa, buf) = bsp;
389 	SCARG(&bsa, flags) = ST_WAIT;
390 
391 	if ((error = sys_statvfs1(l, &bsa, retval)))
392 		return error;
393 
394 	if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp)))
395 		return error;
396 
397 	bsd_to_linux_statfs(&btmp, &ltmp);
398 
399 	return copyout((caddr_t) &ltmp, (caddr_t) SCARG(uap, sp), sizeof ltmp);
400 }
401 
402 int
403 linux_sys_fstatfs(l, v, retval)
404 	struct lwp *l;
405 	void *v;
406 	register_t *retval;
407 {
408 	struct linux_sys_fstatfs_args /* {
409 		syscallarg(int) fd;
410 		syscallarg(struct linux_statfs *) sp;
411 	} */ *uap = v;
412 	struct proc *p = l->l_proc;
413 	struct statvfs btmp, *bsp;
414 	struct linux_statfs ltmp;
415 	struct sys_fstatvfs1_args bsa;
416 	caddr_t sg;
417 	int error;
418 
419 	sg = stackgap_init(p, 0);
420 	bsp = (struct statvfs *) stackgap_alloc(p, &sg, sizeof (struct statvfs));
421 
422 	SCARG(&bsa, fd) = SCARG(uap, fd);
423 	SCARG(&bsa, buf) = bsp;
424 	SCARG(&bsa, flags) = ST_WAIT;
425 
426 	if ((error = sys_fstatvfs1(l, &bsa, retval)))
427 		return error;
428 
429 	if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp)))
430 		return error;
431 
432 	bsd_to_linux_statfs(&btmp, &ltmp);
433 
434 	return copyout((caddr_t) &ltmp, (caddr_t) SCARG(uap, sp), sizeof ltmp);
435 }
436 # endif /* !__amd64__ */
437 
438 /*
439  * uname(). Just copy the info from the various strings stored in the
440  * kernel, and put it in the Linux utsname structure. That structure
441  * is almost the same as the NetBSD one, only it has fields 65 characters
442  * long, and an extra domainname field.
443  */
444 int
445 linux_sys_uname(l, v, retval)
446 	struct lwp *l;
447 	void *v;
448 	register_t *retval;
449 {
450 	struct linux_sys_uname_args /* {
451 		syscallarg(struct linux_utsname *) up;
452 	} */ *uap = v;
453 	struct linux_utsname luts;
454 
455 	strncpy(luts.l_sysname, linux_sysname, sizeof(luts.l_sysname));
456 	strncpy(luts.l_nodename, hostname, sizeof(luts.l_nodename));
457 	strncpy(luts.l_release, linux_release, sizeof(luts.l_release));
458 	strncpy(luts.l_version, linux_version, sizeof(luts.l_version));
459 # ifdef LINUX_UNAME_ARCH
460 	strncpy(luts.l_machine, LINUX_UNAME_ARCH, sizeof(luts.l_machine));
461 # else
462 	strncpy(luts.l_machine, machine, sizeof(luts.l_machine));
463 # endif
464 	strncpy(luts.l_domainname, domainname, sizeof(luts.l_domainname));
465 
466 	return copyout(&luts, SCARG(uap, up), sizeof(luts));
467 }
468 
469 /* Used directly on: alpha, mips, ppc, sparc, sparc64 */
470 /* Used indirectly on: arm, i386, m68k */
471 
472 /*
473  * New type Linux mmap call.
474  * Only called directly on machines with >= 6 free regs.
475  */
476 int
477 linux_sys_mmap(l, v, retval)
478 	struct lwp *l;
479 	void *v;
480 	register_t *retval;
481 {
482 	struct linux_sys_mmap_args /* {
483 		syscallarg(unsigned long) addr;
484 		syscallarg(size_t) len;
485 		syscallarg(int) prot;
486 		syscallarg(int) flags;
487 		syscallarg(int) fd;
488 		syscallarg(linux_off_t) offset;
489 	} */ *uap = v;
490 
491 	if (SCARG(uap, offset) & PAGE_MASK)
492 		return EINVAL;
493 
494 	return linux_mmap(l, uap, retval, SCARG(uap, offset));
495 }
496 
497 /*
498  * Guts of most architectures' mmap64() implementations.  This shares
499  * its list of arguments with linux_sys_mmap().
500  *
501  * The difference in linux_sys_mmap2() is that "offset" is actually
502  * (offset / pagesize), not an absolute byte count.  This translation
503  * to pagesize offsets is done inside glibc between the mmap64() call
504  * point, and the actual syscall.
505  */
506 int
507 linux_sys_mmap2(l, v, retval)
508 	struct lwp *l;
509 	void *v;
510 	register_t *retval;
511 {
512 	struct linux_sys_mmap2_args /* {
513 		syscallarg(unsigned long) addr;
514 		syscallarg(size_t) len;
515 		syscallarg(int) prot;
516 		syscallarg(int) flags;
517 		syscallarg(int) fd;
518 		syscallarg(linux_off_t) offset;
519 	} */ *uap = v;
520 
521 	return linux_mmap(l, uap, retval,
522 	    ((off_t)SCARG(uap, offset)) << PAGE_SHIFT);
523 }
524 
525 /*
526  * Massage arguments and call system mmap(2).
527  */
528 static int
529 linux_mmap(l, uap, retval, offset)
530 	struct lwp *l;
531 	struct linux_sys_mmap_args *uap;
532 	register_t *retval;
533 	off_t offset;
534 {
535 	struct sys_mmap_args cma;
536 	int error;
537 	size_t mmoff=0;
538 
539 	if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN) {
540 		/*
541 		 * Request for stack-like memory segment. On linux, this
542 		 * works by mmap()ping (small) segment, which is automatically
543 		 * extended when page fault happens below the currently
544 		 * allocated area. We emulate this by allocating (typically
545 		 * bigger) segment sized at current stack size limit, and
546 		 * offsetting the requested and returned address accordingly.
547 		 * Since physical pages are only allocated on-demand, this
548 		 * is effectively identical.
549 		 */
550 		rlim_t ssl = l->l_proc->p_rlimit[RLIMIT_STACK].rlim_cur;
551 
552 		if (SCARG(uap, len) < ssl) {
553 			/* Compute the address offset */
554 			mmoff = round_page(ssl) - SCARG(uap, len);
555 
556 			if (SCARG(uap, addr))
557 				SCARG(uap, addr) -= mmoff;
558 
559 			SCARG(uap, len) = (size_t) ssl;
560 		}
561 	}
562 
563 	linux_to_bsd_mmap_args(&cma, uap);
564 	SCARG(&cma, pos) = offset;
565 
566 	error = sys_mmap(l, &cma, retval);
567 	if (error)
568 		return (error);
569 
570 	/* Shift the returned address for stack-like segment if necessary */
571 	if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN && mmoff)
572 		retval[0] += mmoff;
573 
574 	return (0);
575 }
576 
577 static void
578 linux_to_bsd_mmap_args(cma, uap)
579 	struct sys_mmap_args *cma;
580 	const struct linux_sys_mmap_args *uap;
581 {
582 	int flags = MAP_TRYFIXED, fl = SCARG(uap, flags);
583 
584 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_SHARED, MAP_SHARED);
585 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_PRIVATE, MAP_PRIVATE);
586 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_FIXED, MAP_FIXED);
587 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_ANON, MAP_ANON);
588 	/* XXX XAX ERH: Any other flags here?  There are more defined... */
589 
590 	SCARG(cma, addr) = (void *)SCARG(uap, addr);
591 	SCARG(cma, len) = SCARG(uap, len);
592 	SCARG(cma, prot) = SCARG(uap, prot);
593 	if (SCARG(cma, prot) & VM_PROT_WRITE) /* XXX */
594 		SCARG(cma, prot) |= VM_PROT_READ;
595 	SCARG(cma, flags) = flags;
596 	SCARG(cma, fd) = flags & MAP_ANON ? -1 : SCARG(uap, fd);
597 	SCARG(cma, pad) = 0;
598 }
599 
600 #define	LINUX_MREMAP_MAYMOVE	1
601 #define	LINUX_MREMAP_FIXED	2
602 
603 int
604 linux_sys_mremap(l, v, retval)
605 	struct lwp *l;
606 	void *v;
607 	register_t *retval;
608 {
609 	struct linux_sys_mremap_args /* {
610 		syscallarg(void *) old_address;
611 		syscallarg(size_t) old_size;
612 		syscallarg(size_t) new_size;
613 		syscallarg(u_long) flags;
614 	} */ *uap = v;
615 
616 	struct proc *p;
617 	struct vm_map *map;
618 	vaddr_t oldva;
619 	vaddr_t newva;
620 	size_t oldsize;
621 	size_t newsize;
622 	int flags;
623 	int uvmflags;
624 	int error;
625 
626 	flags = SCARG(uap, flags);
627 	oldva = (vaddr_t)SCARG(uap, old_address);
628 	oldsize = round_page(SCARG(uap, old_size));
629 	newsize = round_page(SCARG(uap, new_size));
630 	if ((flags & ~(LINUX_MREMAP_FIXED|LINUX_MREMAP_MAYMOVE)) != 0) {
631 		error = EINVAL;
632 		goto done;
633 	}
634 	if ((flags & LINUX_MREMAP_FIXED) != 0) {
635 		if ((flags & LINUX_MREMAP_MAYMOVE) == 0) {
636 			error = EINVAL;
637 			goto done;
638 		}
639 #if 0 /* notyet */
640 		newva = SCARG(uap, new_address);
641 		uvmflags = UVM_MREMAP_FIXED;
642 #else /* notyet */
643 		error = EOPNOTSUPP;
644 		goto done;
645 #endif /* notyet */
646 	} else if ((flags & LINUX_MREMAP_MAYMOVE) != 0) {
647 		uvmflags = 0;
648 	} else {
649 		newva = oldva;
650 		uvmflags = UVM_MREMAP_FIXED;
651 	}
652 	p = l->l_proc;
653 	map = &p->p_vmspace->vm_map;
654 	error = uvm_mremap(map, oldva, oldsize, map, &newva, newsize, p,
655 	    uvmflags);
656 
657 done:
658 	*retval = (error != 0) ? 0 : (register_t)newva;
659 	return error;
660 }
661 
662 int
663 linux_sys_msync(l, v, retval)
664 	struct lwp *l;
665 	void *v;
666 	register_t *retval;
667 {
668 	struct linux_sys_msync_args /* {
669 		syscallarg(caddr_t) addr;
670 		syscallarg(int) len;
671 		syscallarg(int) fl;
672 	} */ *uap = v;
673 
674 	struct sys___msync13_args bma;
675 
676 	/* flags are ignored */
677 	SCARG(&bma, addr) = SCARG(uap, addr);
678 	SCARG(&bma, len) = SCARG(uap, len);
679 	SCARG(&bma, flags) = SCARG(uap, fl);
680 
681 	return sys___msync13(l, &bma, retval);
682 }
683 
684 int
685 linux_sys_mprotect(l, v, retval)
686 	struct lwp *l;
687 	void *v;
688 	register_t *retval;
689 {
690 	struct linux_sys_mprotect_args /* {
691 		syscallarg(const void *) start;
692 		syscallarg(unsigned long) len;
693 		syscallarg(int) prot;
694 	} */ *uap = v;
695 	struct vm_map_entry *entry;
696 	struct vm_map *map;
697 	struct proc *p;
698 	vaddr_t end, start, len, stacklim;
699 	int prot, grows;
700 
701 	start = (vaddr_t)SCARG(uap, start);
702 	len = round_page(SCARG(uap, len));
703 	prot = SCARG(uap, prot);
704 	grows = prot & (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP);
705 	prot &= ~grows;
706 	end = start + len;
707 
708 	if (start & PAGE_MASK)
709 		return EINVAL;
710 	if (end < start)
711 		return EINVAL;
712 	if (end == start)
713 		return 0;
714 
715 	if (prot & ~(PROT_READ | PROT_WRITE | PROT_EXEC))
716 		return EINVAL;
717 	if (grows == (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP))
718 		return EINVAL;
719 
720 	p = l->l_proc;
721 	map = &p->p_vmspace->vm_map;
722 	vm_map_lock(map);
723 # ifdef notdef
724 	VM_MAP_RANGE_CHECK(map, start, end);
725 # endif
726 	if (!uvm_map_lookup_entry(map, start, &entry) || entry->start > start) {
727 		vm_map_unlock(map);
728 		return ENOMEM;
729 	}
730 
731 	/*
732 	 * Approximate the behaviour of PROT_GROWS{DOWN,UP}.
733 	 */
734 
735 	stacklim = (vaddr_t)p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur;
736 	if (grows & LINUX_PROT_GROWSDOWN) {
737 		if (USRSTACK - stacklim <= start && start < USRSTACK) {
738 			start = USRSTACK - stacklim;
739 		} else {
740 			start = entry->start;
741 		}
742 	} else if (grows & LINUX_PROT_GROWSUP) {
743 		if (USRSTACK <= end && end < USRSTACK + stacklim) {
744 			end = USRSTACK + stacklim;
745 		} else {
746 			end = entry->end;
747 		}
748 	}
749 	vm_map_unlock(map);
750 	return uvm_map_protect(map, start, end, prot, FALSE);
751 }
752 
753 /*
754  * This code is partly stolen from src/lib/libc/compat-43/times.c
755  */
756 
757 #define	CONVTCK(r)	(r.tv_sec * hz + r.tv_usec / (1000000 / hz))
758 
759 int
760 linux_sys_times(l, v, retval)
761 	struct lwp *l;
762 	void *v;
763 	register_t *retval;
764 {
765 	struct linux_sys_times_args /* {
766 		syscallarg(struct times *) tms;
767 	} */ *uap = v;
768 	struct proc *p = l->l_proc;
769 	struct timeval t;
770 	int error, s;
771 
772 	if (SCARG(uap, tms)) {
773 		struct linux_tms ltms;
774 		struct rusage ru;
775 
776 		calcru(p, &ru.ru_utime, &ru.ru_stime, NULL);
777 		ltms.ltms_utime = CONVTCK(ru.ru_utime);
778 		ltms.ltms_stime = CONVTCK(ru.ru_stime);
779 
780 		ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime);
781 		ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime);
782 
783 		if ((error = copyout(&ltms, SCARG(uap, tms), sizeof ltms)))
784 			return error;
785 	}
786 
787 	s = splclock();
788 	timersub(&time, &boottime, &t);
789 	splx(s);
790 
791 	retval[0] = ((linux_clock_t)(CONVTCK(t)));
792 	return 0;
793 }
794 
795 #undef CONVTCK
796 
797 /*
798  * Linux 'readdir' call. This code is mostly taken from the
799  * SunOS getdents call (see compat/sunos/sunos_misc.c), though
800  * an attempt has been made to keep it a little cleaner (failing
801  * miserably, because of the cruft needed if count 1 is passed).
802  *
803  * The d_off field should contain the offset of the next valid entry,
804  * but in Linux it has the offset of the entry itself. We emulate
805  * that bug here.
806  *
807  * Read in BSD-style entries, convert them, and copy them out.
808  *
809  * Note that this doesn't handle union-mounted filesystems.
810  */
811 int
812 linux_sys_getdents(l, v, retval)
813 	struct lwp *l;
814 	void *v;
815 	register_t *retval;
816 {
817 	struct linux_sys_getdents_args /* {
818 		syscallarg(int) fd;
819 		syscallarg(struct linux_dirent *) dent;
820 		syscallarg(unsigned int) count;
821 	} */ *uap = v;
822 	struct proc *p = l->l_proc;
823 	struct dirent *bdp;
824 	struct vnode *vp;
825 	caddr_t	inp, tbuf;		/* BSD-format */
826 	int len, reclen;		/* BSD-format */
827 	caddr_t outp;			/* Linux-format */
828 	int resid, linux_reclen = 0;	/* Linux-format */
829 	struct file *fp;
830 	struct uio auio;
831 	struct iovec aiov;
832 	struct linux_dirent idb;
833 	off_t off;		/* true file offset */
834 	int buflen, error, eofflag, nbytes, oldcall;
835 	struct vattr va;
836 	off_t *cookiebuf = NULL, *cookie;
837 	int ncookies;
838 
839 	/* getvnode() will use the descriptor for us */
840 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
841 		return (error);
842 
843 	if ((fp->f_flag & FREAD) == 0) {
844 		error = EBADF;
845 		goto out1;
846 	}
847 
848 	vp = (struct vnode *)fp->f_data;
849 	if (vp->v_type != VDIR) {
850 		error = EINVAL;
851 		goto out1;
852 	}
853 
854 	if ((error = VOP_GETATTR(vp, &va, p->p_ucred, l)))
855 		goto out1;
856 
857 	nbytes = SCARG(uap, count);
858 	if (nbytes == 1) {	/* emulating old, broken behaviour */
859 		nbytes = sizeof (idb);
860 		buflen = max(va.va_blocksize, nbytes);
861 		oldcall = 1;
862 	} else {
863 		buflen = min(MAXBSIZE, nbytes);
864 		if (buflen < va.va_blocksize)
865 			buflen = va.va_blocksize;
866 		oldcall = 0;
867 	}
868 	tbuf = malloc(buflen, M_TEMP, M_WAITOK);
869 
870 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
871 	off = fp->f_offset;
872 again:
873 	aiov.iov_base = tbuf;
874 	aiov.iov_len = buflen;
875 	auio.uio_iov = &aiov;
876 	auio.uio_iovcnt = 1;
877 	auio.uio_rw = UIO_READ;
878 	auio.uio_resid = buflen;
879 	auio.uio_offset = off;
880 	UIO_SETUP_SYSSPACE(&auio);
881 	/*
882          * First we read into the malloc'ed buffer, then
883          * we massage it into user space, one record at a time.
884          */
885 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf,
886 	    &ncookies);
887 	if (error)
888 		goto out;
889 
890 	inp = tbuf;
891 	outp = (caddr_t)SCARG(uap, dent);
892 	resid = nbytes;
893 	if ((len = buflen - auio.uio_resid) == 0)
894 		goto eof;
895 
896 	for (cookie = cookiebuf; len > 0; len -= reclen) {
897 		bdp = (struct dirent *)inp;
898 		reclen = bdp->d_reclen;
899 		if (reclen & 3)
900 			panic("linux_readdir");
901 		if (bdp->d_fileno == 0) {
902 			inp += reclen;	/* it is a hole; squish it out */
903 			if (cookie)
904 				off = *cookie++;
905 			else
906 				off += reclen;
907 			continue;
908 		}
909 		linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen);
910 		if (reclen > len || resid < linux_reclen) {
911 			/* entry too big for buffer, so just stop */
912 			outp++;
913 			break;
914 		}
915 		/*
916 		 * Massage in place to make a Linux-shaped dirent (otherwise
917 		 * we have to worry about touching user memory outside of
918 		 * the copyout() call).
919 		 */
920 		idb.d_ino = bdp->d_fileno;
921 		/*
922 		 * The old readdir() call misuses the offset and reclen fields.
923 		 */
924 		if (oldcall) {
925 			idb.d_off = (linux_off_t)linux_reclen;
926 			idb.d_reclen = (u_short)bdp->d_namlen;
927 		} else {
928 			if (sizeof (idb.d_off) <= 4 && (off >> 32) != 0) {
929 				compat_offseterr(vp, "linux_getdents");
930 				error = EINVAL;
931 				goto out;
932 			}
933 			idb.d_off = (linux_off_t)off;
934 			idb.d_reclen = (u_short)linux_reclen;
935 		}
936 		strcpy(idb.d_name, bdp->d_name);
937 		if ((error = copyout((caddr_t)&idb, outp, linux_reclen)))
938 			goto out;
939 		/* advance past this real entry */
940 		inp += reclen;
941 		if (cookie)
942 			off = *cookie++; /* each entry points to itself */
943 		else
944 			off += reclen;
945 		/* advance output past Linux-shaped entry */
946 		outp += linux_reclen;
947 		resid -= linux_reclen;
948 		if (oldcall)
949 			break;
950 	}
951 
952 	/* if we squished out the whole block, try again */
953 	if (outp == (caddr_t)SCARG(uap, dent))
954 		goto again;
955 	fp->f_offset = off;	/* update the vnode offset */
956 
957 	if (oldcall)
958 		nbytes = resid + linux_reclen;
959 
960 eof:
961 	*retval = nbytes - resid;
962 out:
963 	VOP_UNLOCK(vp, 0);
964 	if (cookiebuf)
965 		free(cookiebuf, M_TEMP);
966 	free(tbuf, M_TEMP);
967 out1:
968 	FILE_UNUSE(fp, l);
969 	return error;
970 }
971 
972 /*
973  * Even when just using registers to pass arguments to syscalls you can
974  * have 5 of them on the i386. So this newer version of select() does
975  * this.
976  */
977 int
978 linux_sys_select(l, v, retval)
979 	struct lwp *l;
980 	void *v;
981 	register_t *retval;
982 {
983 	struct linux_sys_select_args /* {
984 		syscallarg(int) nfds;
985 		syscallarg(fd_set *) readfds;
986 		syscallarg(fd_set *) writefds;
987 		syscallarg(fd_set *) exceptfds;
988 		syscallarg(struct timeval *) timeout;
989 	} */ *uap = v;
990 
991 	return linux_select1(l, retval, SCARG(uap, nfds), SCARG(uap, readfds),
992 	    SCARG(uap, writefds), SCARG(uap, exceptfds), SCARG(uap, timeout));
993 }
994 
995 /*
996  * Common code for the old and new versions of select(). A couple of
997  * things are important:
998  * 1) return the amount of time left in the 'timeout' parameter
999  * 2) select never returns ERESTART on Linux, always return EINTR
1000  */
1001 int
1002 linux_select1(l, retval, nfds, readfds, writefds, exceptfds, timeout)
1003 	struct lwp *l;
1004 	register_t *retval;
1005 	int nfds;
1006 	fd_set *readfds, *writefds, *exceptfds;
1007 	struct timeval *timeout;
1008 {
1009 	struct sys_select_args bsa;
1010 	struct proc *p = l->l_proc;
1011 	struct timeval tv0, tv1, utv, *tvp;
1012 	caddr_t sg;
1013 	int error;
1014 
1015 	SCARG(&bsa, nd) = nfds;
1016 	SCARG(&bsa, in) = readfds;
1017 	SCARG(&bsa, ou) = writefds;
1018 	SCARG(&bsa, ex) = exceptfds;
1019 	SCARG(&bsa, tv) = timeout;
1020 
1021 	/*
1022 	 * Store current time for computation of the amount of
1023 	 * time left.
1024 	 */
1025 	if (timeout) {
1026 		if ((error = copyin(timeout, &utv, sizeof(utv))))
1027 			return error;
1028 		if (itimerfix(&utv)) {
1029 			/*
1030 			 * The timeval was invalid.  Convert it to something
1031 			 * valid that will act as it does under Linux.
1032 			 */
1033 			sg = stackgap_init(p, 0);
1034 			tvp = stackgap_alloc(p, &sg, sizeof(utv));
1035 			utv.tv_sec += utv.tv_usec / 1000000;
1036 			utv.tv_usec %= 1000000;
1037 			if (utv.tv_usec < 0) {
1038 				utv.tv_sec -= 1;
1039 				utv.tv_usec += 1000000;
1040 			}
1041 			if (utv.tv_sec < 0)
1042 				timerclear(&utv);
1043 			if ((error = copyout(&utv, tvp, sizeof(utv))))
1044 				return error;
1045 			SCARG(&bsa, tv) = tvp;
1046 		}
1047 		microtime(&tv0);
1048 	}
1049 
1050 	error = sys_select(l, &bsa, retval);
1051 	if (error) {
1052 		/*
1053 		 * See fs/select.c in the Linux kernel.  Without this,
1054 		 * Maelstrom doesn't work.
1055 		 */
1056 		if (error == ERESTART)
1057 			error = EINTR;
1058 		return error;
1059 	}
1060 
1061 	if (timeout) {
1062 		if (*retval) {
1063 			/*
1064 			 * Compute how much time was left of the timeout,
1065 			 * by subtracting the current time and the time
1066 			 * before we started the call, and subtracting
1067 			 * that result from the user-supplied value.
1068 			 */
1069 			microtime(&tv1);
1070 			timersub(&tv1, &tv0, &tv1);
1071 			timersub(&utv, &tv1, &utv);
1072 			if (utv.tv_sec < 0)
1073 				timerclear(&utv);
1074 		} else
1075 			timerclear(&utv);
1076 		if ((error = copyout(&utv, timeout, sizeof(utv))))
1077 			return error;
1078 	}
1079 
1080 	return 0;
1081 }
1082 
1083 /*
1084  * Get the process group of a certain process. Look it up
1085  * and return the value.
1086  */
1087 int
1088 linux_sys_getpgid(l, v, retval)
1089 	struct lwp *l;
1090 	void *v;
1091 	register_t *retval;
1092 {
1093 	struct linux_sys_getpgid_args /* {
1094 		syscallarg(int) pid;
1095 	} */ *uap = v;
1096 	struct proc *p = l->l_proc;
1097 	struct proc *targp;
1098 
1099 	if (SCARG(uap, pid) != 0 && SCARG(uap, pid) != p->p_pid) {
1100 		if ((targp = pfind(SCARG(uap, pid))) == 0)
1101 			return ESRCH;
1102 	}
1103 	else
1104 		targp = p;
1105 
1106 	retval[0] = targp->p_pgid;
1107 	return 0;
1108 }
1109 
1110 /*
1111  * Set the 'personality' (emulation mode) for the current process. Only
1112  * accept the Linux personality here (0). This call is needed because
1113  * the Linux ELF crt0 issues it in an ugly kludge to make sure that
1114  * ELF binaries run in Linux mode, not SVR4 mode.
1115  */
1116 int
1117 linux_sys_personality(l, v, retval)
1118 	struct lwp *l;
1119 	void *v;
1120 	register_t *retval;
1121 {
1122 	struct linux_sys_personality_args /* {
1123 		syscallarg(int) per;
1124 	} */ *uap = v;
1125 
1126 	if (SCARG(uap, per) != 0)
1127 		return EINVAL;
1128 	retval[0] = 0;
1129 	return 0;
1130 }
1131 #endif /* !COMPAT_LINUX32 */
1132 
1133 #if defined(__i386__) || defined(__m68k__) || defined(COMPAT_LINUX32)
1134 /*
1135  * The calls are here because of type conversions.
1136  */
1137 int
1138 linux_sys_setreuid16(l, v, retval)
1139 	struct lwp *l;
1140 	void *v;
1141 	register_t *retval;
1142 {
1143 	struct linux_sys_setreuid16_args /* {
1144 		syscallarg(int) ruid;
1145 		syscallarg(int) euid;
1146 	} */ *uap = v;
1147 	struct sys_setreuid_args bsa;
1148 
1149 	SCARG(&bsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
1150 		(uid_t)-1 : SCARG(uap, ruid);
1151 	SCARG(&bsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
1152 		(uid_t)-1 : SCARG(uap, euid);
1153 
1154 	return sys_setreuid(l, &bsa, retval);
1155 }
1156 
1157 int
1158 linux_sys_setregid16(l, v, retval)
1159 	struct lwp *l;
1160 	void *v;
1161 	register_t *retval;
1162 {
1163 	struct linux_sys_setregid16_args /* {
1164 		syscallarg(int) rgid;
1165 		syscallarg(int) egid;
1166 	} */ *uap = v;
1167 	struct sys_setregid_args bsa;
1168 
1169 	SCARG(&bsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
1170 		(uid_t)-1 : SCARG(uap, rgid);
1171 	SCARG(&bsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
1172 		(uid_t)-1 : SCARG(uap, egid);
1173 
1174 	return sys_setregid(l, &bsa, retval);
1175 }
1176 
1177 int
1178 linux_sys_setresuid16(l, v, retval)
1179 	struct lwp *l;
1180 	void *v;
1181 	register_t *retval;
1182 {
1183 	struct linux_sys_setresuid16_args /* {
1184 		syscallarg(uid_t) ruid;
1185 		syscallarg(uid_t) euid;
1186 		syscallarg(uid_t) suid;
1187 	} */ *uap = v;
1188 	struct linux_sys_setresuid16_args lsa;
1189 
1190 	SCARG(&lsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
1191 		(uid_t)-1 : SCARG(uap, ruid);
1192 	SCARG(&lsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
1193 		(uid_t)-1 : SCARG(uap, euid);
1194 	SCARG(&lsa, suid) = ((linux_uid_t)SCARG(uap, suid) == (linux_uid_t)-1) ?
1195 		(uid_t)-1 : SCARG(uap, suid);
1196 
1197 	return linux_sys_setresuid(l, &lsa, retval);
1198 }
1199 
1200 int
1201 linux_sys_setresgid16(l, v, retval)
1202 	struct lwp *l;
1203 	void *v;
1204 	register_t *retval;
1205 {
1206 	struct linux_sys_setresgid16_args /* {
1207 		syscallarg(gid_t) rgid;
1208 		syscallarg(gid_t) egid;
1209 		syscallarg(gid_t) sgid;
1210 	} */ *uap = v;
1211 	struct linux_sys_setresgid16_args lsa;
1212 
1213 	SCARG(&lsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
1214 		(gid_t)-1 : SCARG(uap, rgid);
1215 	SCARG(&lsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
1216 		(gid_t)-1 : SCARG(uap, egid);
1217 	SCARG(&lsa, sgid) = ((linux_gid_t)SCARG(uap, sgid) == (linux_gid_t)-1) ?
1218 		(gid_t)-1 : SCARG(uap, sgid);
1219 
1220 	return linux_sys_setresgid(l, &lsa, retval);
1221 }
1222 
1223 int
1224 linux_sys_getgroups16(l, v, retval)
1225 	struct lwp *l;
1226 	void *v;
1227 	register_t *retval;
1228 {
1229 	struct linux_sys_getgroups16_args /* {
1230 		syscallarg(int) gidsetsize;
1231 		syscallarg(linux_gid_t *) gidset;
1232 	} */ *uap = v;
1233 	struct proc *p = l->l_proc;
1234 	caddr_t sg;
1235 	int n, error, i;
1236 	struct sys_getgroups_args bsa;
1237 	gid_t *bset, *kbset;
1238 	linux_gid_t *lset;
1239 	struct pcred *pc = p->p_cred;
1240 
1241 	n = SCARG(uap, gidsetsize);
1242 	if (n < 0)
1243 		return EINVAL;
1244 	error = 0;
1245 	bset = kbset = NULL;
1246 	lset = NULL;
1247 	if (n > 0) {
1248 		n = min(pc->pc_ucred->cr_ngroups, n);
1249 		sg = stackgap_init(p, 0);
1250 		bset = stackgap_alloc(p, &sg, n * sizeof (gid_t));
1251 		kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK);
1252 		lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
1253 		if (bset == NULL || kbset == NULL || lset == NULL)
1254 		{
1255 			error = ENOMEM;
1256 			goto out;
1257 		}
1258 		SCARG(&bsa, gidsetsize) = n;
1259 		SCARG(&bsa, gidset) = bset;
1260 		error = sys_getgroups(l, &bsa, retval);
1261 		if (error != 0)
1262 			goto out;
1263 		error = copyin(bset, kbset, n * sizeof (gid_t));
1264 		if (error != 0)
1265 			goto out;
1266 		for (i = 0; i < n; i++)
1267 			lset[i] = (linux_gid_t)kbset[i];
1268 		error = copyout(lset, SCARG(uap, gidset),
1269 		    n * sizeof (linux_gid_t));
1270 	} else
1271 		*retval = pc->pc_ucred->cr_ngroups;
1272 out:
1273 	if (kbset != NULL)
1274 		free(kbset, M_TEMP);
1275 	if (lset != NULL)
1276 		free(lset, M_TEMP);
1277 	return error;
1278 }
1279 
1280 int
1281 linux_sys_setgroups16(l, v, retval)
1282 	struct lwp *l;
1283 	void *v;
1284 	register_t *retval;
1285 {
1286 	struct linux_sys_setgroups16_args /* {
1287 		syscallarg(int) gidsetsize;
1288 		syscallarg(linux_gid_t *) gidset;
1289 	} */ *uap = v;
1290 	struct proc *p = l->l_proc;
1291 	caddr_t sg;
1292 	int n;
1293 	int error, i;
1294 	struct sys_setgroups_args bsa;
1295 	gid_t *bset, *kbset;
1296 	linux_gid_t *lset;
1297 
1298 	n = SCARG(uap, gidsetsize);
1299 	if (n < 0 || n > NGROUPS)
1300 		return EINVAL;
1301 	sg = stackgap_init(p, 0);
1302 	bset = stackgap_alloc(p, &sg, n * sizeof (gid_t));
1303 	lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
1304 	kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK);
1305 	if (bset == NULL || kbset == NULL || lset == NULL)
1306 	{
1307 		error = ENOMEM;
1308 		goto out;
1309 	}
1310 	error = copyin(SCARG(uap, gidset), lset, n * sizeof (linux_gid_t));
1311 	if (error != 0)
1312 		goto out;
1313 	for (i = 0; i < n; i++)
1314 		kbset[i] = (gid_t)lset[i];
1315 	error = copyout(kbset, bset, n * sizeof (gid_t));
1316 	if (error != 0)
1317 		goto out;
1318 	SCARG(&bsa, gidsetsize) = n;
1319 	SCARG(&bsa, gidset) = bset;
1320 	error = sys_setgroups(l, &bsa, retval);
1321 
1322 out:
1323 	if (lset != NULL)
1324 		free(lset, M_TEMP);
1325 	if (kbset != NULL)
1326 		free(kbset, M_TEMP);
1327 
1328 	return error;
1329 }
1330 
1331 #endif /* __i386__ || __m68k__ || COMPAT_LINUX32 */
1332 
1333 #ifndef COMPAT_LINUX32
1334 /*
1335  * We have nonexistent fsuid equal to uid.
1336  * If modification is requested, refuse.
1337  */
1338 int
1339 linux_sys_setfsuid(l, v, retval)
1340 	 struct lwp *l;
1341 	 void *v;
1342 	 register_t *retval;
1343 {
1344 	 struct linux_sys_setfsuid_args /* {
1345 		 syscallarg(uid_t) uid;
1346 	 } */ *uap = v;
1347 	 struct proc *p = l->l_proc;
1348 	 uid_t uid;
1349 
1350 	 uid = SCARG(uap, uid);
1351 	 if (p->p_cred->p_ruid != uid)
1352 		 return sys_nosys(l, v, retval);
1353 	 else
1354 		 return (0);
1355 }
1356 
1357 /* XXX XXX XXX */
1358 # ifndef alpha
1359 int
1360 linux_sys_getfsuid(l, v, retval)
1361 	struct lwp *l;
1362 	void *v;
1363 	register_t *retval;
1364 {
1365 	return sys_getuid(l, v, retval);
1366 }
1367 # endif
1368 
1369 int
1370 linux_sys_setresuid(l, v, retval)
1371 	struct lwp *l;
1372 	void *v;
1373 	register_t *retval;
1374 {
1375 	struct linux_sys_setresuid_args /* {
1376 		syscallarg(uid_t) ruid;
1377 		syscallarg(uid_t) euid;
1378 		syscallarg(uid_t) suid;
1379 	} */ *uap = v;
1380 
1381 	/*
1382 	 * Note: These checks are a little different than the NetBSD
1383 	 * setreuid(2) call performs.  This precisely follows the
1384 	 * behavior of the Linux kernel.
1385 	 */
1386 
1387 	return do_setresuid(l, SCARG(uap, ruid), SCARG(uap, euid),
1388 			    SCARG(uap, suid),
1389 			    ID_R_EQ_R | ID_R_EQ_E | ID_R_EQ_S |
1390 			    ID_E_EQ_R | ID_E_EQ_E | ID_E_EQ_S |
1391 			    ID_S_EQ_R | ID_S_EQ_E | ID_S_EQ_S );
1392 }
1393 
1394 int
1395 linux_sys_getresuid(l, v, retval)
1396 	struct lwp *l;
1397 	void *v;
1398 	register_t *retval;
1399 {
1400 	struct linux_sys_getresuid_args /* {
1401 		syscallarg(uid_t *) ruid;
1402 		syscallarg(uid_t *) euid;
1403 		syscallarg(uid_t *) suid;
1404 	} */ *uap = v;
1405 	struct proc *p = l->l_proc;
1406 	struct pcred *pc = p->p_cred;
1407 	int error;
1408 
1409 	/*
1410 	 * Linux copies these values out to userspace like so:
1411 	 *
1412 	 *	1. Copy out ruid.
1413 	 *	2. If that succeeds, copy out euid.
1414 	 *	3. If both of those succeed, copy out suid.
1415 	 */
1416 	if ((error = copyout(&pc->p_ruid, SCARG(uap, ruid),
1417 			     sizeof(uid_t))) != 0)
1418 		return (error);
1419 
1420 	if ((error = copyout(&pc->pc_ucred->cr_uid, SCARG(uap, euid),
1421 			     sizeof(uid_t))) != 0)
1422 		return (error);
1423 
1424 	return (copyout(&pc->p_svuid, SCARG(uap, suid), sizeof(uid_t)));
1425 }
1426 
1427 int
1428 linux_sys_ptrace(l, v, retval)
1429 	struct lwp *l;
1430 	void *v;
1431 	register_t *retval;
1432 {
1433 	struct linux_sys_ptrace_args /* {
1434 		i386, m68k, powerpc: T=int
1435 		alpha, amd64: T=long
1436 		syscallarg(T) request;
1437 		syscallarg(T) pid;
1438 		syscallarg(T) addr;
1439 		syscallarg(T) data;
1440 	} */ *uap = v;
1441 	const int *ptr;
1442 	int request;
1443 	int error;
1444 
1445 	ptr = linux_ptrace_request_map;
1446 	request = SCARG(uap, request);
1447 	while (*ptr != -1)
1448 		if (*ptr++ == request) {
1449 			struct sys_ptrace_args pta;
1450 
1451 			SCARG(&pta, req) = *ptr;
1452 			SCARG(&pta, pid) = SCARG(uap, pid);
1453 			SCARG(&pta, addr) = (caddr_t)SCARG(uap, addr);
1454 			SCARG(&pta, data) = SCARG(uap, data);
1455 
1456 			/*
1457 			 * Linux ptrace(PTRACE_CONT, pid, 0, 0) means actually
1458 			 * to continue where the process left off previously.
1459 			 * The same thing is achieved by addr == (caddr_t) 1
1460 			 * on NetBSD, so rewrite 'addr' appropriately.
1461 			 */
1462 			if (request == LINUX_PTRACE_CONT && SCARG(uap, addr)==0)
1463 				SCARG(&pta, addr) = (caddr_t) 1;
1464 
1465 			error = sys_ptrace(l, &pta, retval);
1466 			if (error)
1467 				return error;
1468 			switch (request) {
1469 			case LINUX_PTRACE_PEEKTEXT:
1470 			case LINUX_PTRACE_PEEKDATA:
1471 				error = copyout (retval,
1472 				    (caddr_t)SCARG(uap, data),
1473 				    sizeof *retval);
1474 				*retval = SCARG(uap, data);
1475 				break;
1476 			default:
1477 				break;
1478 			}
1479 			return error;
1480 		}
1481 		else
1482 			ptr++;
1483 
1484 	return LINUX_SYS_PTRACE_ARCH(l, uap, retval);
1485 }
1486 
1487 int
1488 linux_sys_reboot(struct lwp *l, void *v, register_t *retval)
1489 {
1490 	struct linux_sys_reboot_args /* {
1491 		syscallarg(int) magic1;
1492 		syscallarg(int) magic2;
1493 		syscallarg(int) cmd;
1494 		syscallarg(void *) arg;
1495 	} */ *uap = v;
1496 	struct sys_reboot_args /* {
1497 		syscallarg(int) opt;
1498 		syscallarg(char *) bootstr;
1499 	} */ sra;
1500 	struct proc *p = l->l_proc;
1501 	int error;
1502 
1503 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1504 		return(error);
1505 
1506 	if (SCARG(uap, magic1) != LINUX_REBOOT_MAGIC1)
1507 		return(EINVAL);
1508 	if (SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2 &&
1509 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2A &&
1510 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2B)
1511 		return(EINVAL);
1512 
1513 	switch (SCARG(uap, cmd)) {
1514 	case LINUX_REBOOT_CMD_RESTART:
1515 		SCARG(&sra, opt) = RB_AUTOBOOT;
1516 		break;
1517 	case LINUX_REBOOT_CMD_HALT:
1518 		SCARG(&sra, opt) = RB_HALT;
1519 		break;
1520 	case LINUX_REBOOT_CMD_POWER_OFF:
1521 		SCARG(&sra, opt) = RB_HALT|RB_POWERDOWN;
1522 		break;
1523 	case LINUX_REBOOT_CMD_RESTART2:
1524 		/* Reboot with an argument. */
1525 		SCARG(&sra, opt) = RB_AUTOBOOT|RB_STRING;
1526 		SCARG(&sra, bootstr) = SCARG(uap, arg);
1527 		break;
1528 	case LINUX_REBOOT_CMD_CAD_ON:
1529 		return(EINVAL);	/* We don't implement ctrl-alt-delete */
1530 	case LINUX_REBOOT_CMD_CAD_OFF:
1531 		return(0);
1532 	default:
1533 		return(EINVAL);
1534 	}
1535 
1536 	return(sys_reboot(l, &sra, retval));
1537 }
1538 
1539 /*
1540  * Copy of compat_12_sys_swapon().
1541  */
1542 int
1543 linux_sys_swapon(l, v, retval)
1544 	struct lwp *l;
1545 	void *v;
1546 	register_t *retval;
1547 {
1548 	struct sys_swapctl_args ua;
1549 	struct linux_sys_swapon_args /* {
1550 		syscallarg(const char *) name;
1551 	} */ *uap = v;
1552 
1553 	SCARG(&ua, cmd) = SWAP_ON;
1554 	SCARG(&ua, arg) = (void *)__UNCONST(SCARG(uap, name));
1555 	SCARG(&ua, misc) = 0;	/* priority */
1556 	return (sys_swapctl(l, &ua, retval));
1557 }
1558 
1559 /*
1560  * Stop swapping to the file or block device specified by path.
1561  */
1562 int
1563 linux_sys_swapoff(l, v, retval)
1564 	struct lwp *l;
1565 	void *v;
1566 	register_t *retval;
1567 {
1568 	struct sys_swapctl_args ua;
1569 	struct linux_sys_swapoff_args /* {
1570 		syscallarg(const char *) path;
1571 	} */ *uap = v;
1572 
1573 	SCARG(&ua, cmd) = SWAP_OFF;
1574 	SCARG(&ua, arg) = __UNCONST(SCARG(uap, path)); /*XXXUNCONST*/
1575 	return (sys_swapctl(l, &ua, retval));
1576 }
1577 
1578 /*
1579  * Copy of compat_09_sys_setdomainname()
1580  */
1581 /* ARGSUSED */
1582 int
1583 linux_sys_setdomainname(l, v, retval)
1584 	struct lwp *l;
1585 	void *v;
1586 	register_t *retval;
1587 {
1588 	struct linux_sys_setdomainname_args /* {
1589 		syscallarg(char *) domainname;
1590 		syscallarg(int) len;
1591 	} */ *uap = v;
1592 	int name[2];
1593 
1594 	name[0] = CTL_KERN;
1595 	name[1] = KERN_DOMAINNAME;
1596 	return (old_sysctl(&name[0], 2, 0, 0, SCARG(uap, domainname),
1597 			    SCARG(uap, len), l));
1598 }
1599 
1600 /*
1601  * sysinfo()
1602  */
1603 /* ARGSUSED */
1604 int
1605 linux_sys_sysinfo(l, v, retval)
1606 	struct lwp *l;
1607 	void *v;
1608 	register_t *retval;
1609 {
1610 	struct linux_sys_sysinfo_args /* {
1611 		syscallarg(struct linux_sysinfo *) arg;
1612 	} */ *uap = v;
1613 	struct linux_sysinfo si;
1614 	struct loadavg *la;
1615 
1616 	si.uptime = time.tv_sec - boottime.tv_sec;
1617 	la = &averunnable;
1618 	si.loads[0] = la->ldavg[0] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1619 	si.loads[1] = la->ldavg[1] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1620 	si.loads[2] = la->ldavg[2] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1621 	si.totalram = ctob(physmem);
1622 	si.freeram = uvmexp.free * uvmexp.pagesize;
1623 	si.sharedram = 0;	/* XXX */
1624 	si.bufferram = uvmexp.filepages * uvmexp.pagesize;
1625 	si.totalswap = uvmexp.swpages * uvmexp.pagesize;
1626 	si.freeswap = (uvmexp.swpages - uvmexp.swpginuse) * uvmexp.pagesize;
1627 	si.procs = nprocs;
1628 
1629 	/* The following are only present in newer Linux kernels. */
1630 	si.totalbig = 0;
1631 	si.freebig = 0;
1632 	si.mem_unit = 1;
1633 
1634 	return (copyout(&si, SCARG(uap, arg), sizeof si));
1635 }
1636 
1637 int
1638 linux_sys_getrlimit(l, v, retval)
1639 	struct lwp *l;
1640 	void *v;
1641 	register_t *retval;
1642 {
1643 	struct linux_sys_getrlimit_args /* {
1644 		syscallarg(int) which;
1645 # ifdef LINUX_LARGEFILE64
1646 		syscallarg(struct rlimit *) rlp;
1647 # else
1648 		syscallarg(struct orlimit *) rlp;
1649 # endif
1650 	} */ *uap = v;
1651 	struct proc *p = l->l_proc;
1652 	caddr_t sg = stackgap_init(p, 0);
1653 	struct sys_getrlimit_args ap;
1654 	struct rlimit rl;
1655 # ifdef LINUX_LARGEFILE64
1656 	struct rlimit orl;
1657 # else
1658 	struct orlimit orl;
1659 # endif
1660 	int error;
1661 
1662 	SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which));
1663 	if ((error = SCARG(&ap, which)) < 0)
1664 		return -error;
1665 	SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl);
1666 	if ((error = sys_getrlimit(l, &ap, retval)) != 0)
1667 		return error;
1668 	if ((error = copyin(SCARG(&ap, rlp), &rl, sizeof(rl))) != 0)
1669 		return error;
1670 	bsd_to_linux_rlimit(&orl, &rl);
1671 
1672 	return copyout(&orl, SCARG(uap, rlp), sizeof(orl));
1673 }
1674 
1675 int
1676 linux_sys_setrlimit(l, v, retval)
1677 	struct lwp *l;
1678 	void *v;
1679 	register_t *retval;
1680 {
1681 	struct linux_sys_setrlimit_args /* {
1682 		syscallarg(int) which;
1683 # ifdef LINUX_LARGEFILE64
1684 		syscallarg(struct rlimit *) rlp;
1685 # else
1686 		syscallarg(struct orlimit *) rlp;
1687 # endif
1688 	} */ *uap = v;
1689 	struct proc *p = l->l_proc;
1690 	caddr_t sg = stackgap_init(p, 0);
1691 	struct sys_getrlimit_args ap;
1692 	struct rlimit rl;
1693 # ifdef LINUX_LARGEFILE64
1694 	struct rlimit orl;
1695 # else
1696 	struct orlimit orl;
1697 # endif
1698 	int error;
1699 
1700 	SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which));
1701 	SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl);
1702 	if ((error = SCARG(&ap, which)) < 0)
1703 		return -error;
1704 	if ((error = copyin(SCARG(uap, rlp), &orl, sizeof(orl))) != 0)
1705 		return error;
1706 	linux_to_bsd_rlimit(&rl, &orl);
1707 	if ((error = copyout(&rl, SCARG(&ap, rlp), sizeof(rl))) != 0)
1708 		return error;
1709 	return sys_setrlimit(l, &ap, retval);
1710 }
1711 
1712 # if !defined(__mips__) && !defined(__amd64__)
1713 /* XXX: this doesn't look 100% common, at least mips doesn't have it */
1714 int
1715 linux_sys_ugetrlimit(l, v, retval)
1716 	struct lwp *l;
1717 	void *v;
1718 	register_t *retval;
1719 {
1720 	return linux_sys_getrlimit(l, v, retval);
1721 }
1722 # endif
1723 
1724 /*
1725  * This gets called for unsupported syscalls. The difference to sys_nosys()
1726  * is that process does not get SIGSYS, the call just returns with ENOSYS.
1727  * This is the way Linux does it and glibc depends on this behaviour.
1728  */
1729 int
1730 linux_sys_nosys(l, v, retval)
1731 	struct lwp *l;
1732 	void *v;
1733 	register_t *retval;
1734 {
1735 	return (ENOSYS);
1736 }
1737 
1738 #endif /* !COMPAT_LINUX32 */
1739