xref: /netbsd-src/sys/compat/linux/common/linux_misc.c (revision d48f14661dda8638fee055ba15d35bdfb29b9fa8)
1 /*	$NetBSD: linux_misc.c,v 1.157 2006/06/13 16:54:56 christos Exp $	*/
2 
3 /*-
4  * Copyright (c) 1995, 1998, 1999 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Frank van der Linden and Eric Haszlakiewicz; by Jason R. Thorpe
9  * of the Numerical Aerospace Simulation Facility, NASA Ames Research Center.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 3. All advertising materials mentioning features or use of this software
20  *    must display the following acknowledgement:
21  *	This product includes software developed by the NetBSD
22  *	Foundation, Inc. and its contributors.
23  * 4. Neither the name of The NetBSD Foundation nor the names of its
24  *    contributors may be used to endorse or promote products derived
25  *    from this software without specific prior written permission.
26  *
27  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
28  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
29  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
30  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
31  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
32  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
33  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
34  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
35  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
36  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
37  * POSSIBILITY OF SUCH DAMAGE.
38  */
39 
40 /*
41  * Linux compatibility module. Try to deal with various Linux system calls.
42  */
43 
44 /*
45  * These functions have been moved to multiarch to allow
46  * selection of which machines include them to be
47  * determined by the individual files.linux_<arch> files.
48  *
49  * Function in multiarch:
50  *	linux_sys_break			: linux_break.c
51  *	linux_sys_alarm			: linux_misc_notalpha.c
52  *	linux_sys_getresgid		: linux_misc_notalpha.c
53  *	linux_sys_nice			: linux_misc_notalpha.c
54  *	linux_sys_readdir		: linux_misc_notalpha.c
55  *	linux_sys_setresgid		: linux_misc_notalpha.c
56  *	linux_sys_time			: linux_misc_notalpha.c
57  *	linux_sys_utime			: linux_misc_notalpha.c
58  *	linux_sys_waitpid		: linux_misc_notalpha.c
59  *	linux_sys_old_mmap		: linux_oldmmap.c
60  *	linux_sys_oldolduname		: linux_oldolduname.c
61  *	linux_sys_oldselect		: linux_oldselect.c
62  *	linux_sys_olduname		: linux_olduname.c
63  *	linux_sys_pipe			: linux_pipe.c
64  */
65 
66 #include <sys/cdefs.h>
67 __KERNEL_RCSID(0, "$NetBSD: linux_misc.c,v 1.157 2006/06/13 16:54:56 christos Exp $");
68 
69 #include <sys/param.h>
70 #include <sys/systm.h>
71 #include <sys/namei.h>
72 #include <sys/proc.h>
73 #include <sys/dirent.h>
74 #include <sys/file.h>
75 #include <sys/stat.h>
76 #include <sys/filedesc.h>
77 #include <sys/ioctl.h>
78 #include <sys/kernel.h>
79 #include <sys/malloc.h>
80 #include <sys/mbuf.h>
81 #include <sys/mman.h>
82 #include <sys/mount.h>
83 #include <sys/reboot.h>
84 #include <sys/resource.h>
85 #include <sys/resourcevar.h>
86 #include <sys/signal.h>
87 #include <sys/signalvar.h>
88 #include <sys/socket.h>
89 #include <sys/time.h>
90 #include <sys/times.h>
91 #include <sys/vnode.h>
92 #include <sys/uio.h>
93 #include <sys/wait.h>
94 #include <sys/utsname.h>
95 #include <sys/unistd.h>
96 #include <sys/swap.h>		/* for SWAP_ON */
97 #include <sys/sysctl.h>		/* for KERN_DOMAINNAME */
98 #include <sys/kauth.h>
99 
100 #include <sys/ptrace.h>
101 #include <machine/ptrace.h>
102 
103 #include <sys/sa.h>
104 #include <sys/syscallargs.h>
105 
106 #include <compat/linux/common/linux_machdep.h>
107 #include <compat/linux/common/linux_types.h>
108 #include <compat/linux/common/linux_signal.h>
109 
110 #include <compat/linux/linux_syscallargs.h>
111 
112 #include <compat/linux/common/linux_fcntl.h>
113 #include <compat/linux/common/linux_mmap.h>
114 #include <compat/linux/common/linux_dirent.h>
115 #include <compat/linux/common/linux_util.h>
116 #include <compat/linux/common/linux_misc.h>
117 #ifndef COMPAT_LINUX32
118 #include <compat/linux/common/linux_limit.h>
119 #endif
120 #include <compat/linux/common/linux_ptrace.h>
121 #include <compat/linux/common/linux_reboot.h>
122 #include <compat/linux/common/linux_emuldata.h>
123 
124 #ifndef COMPAT_LINUX32
125 const int linux_ptrace_request_map[] = {
126 	LINUX_PTRACE_TRACEME,	PT_TRACE_ME,
127 	LINUX_PTRACE_PEEKTEXT,	PT_READ_I,
128 	LINUX_PTRACE_PEEKDATA,	PT_READ_D,
129 	LINUX_PTRACE_POKETEXT,	PT_WRITE_I,
130 	LINUX_PTRACE_POKEDATA,	PT_WRITE_D,
131 	LINUX_PTRACE_CONT,	PT_CONTINUE,
132 	LINUX_PTRACE_KILL,	PT_KILL,
133 	LINUX_PTRACE_ATTACH,	PT_ATTACH,
134 	LINUX_PTRACE_DETACH,	PT_DETACH,
135 # ifdef PT_STEP
136 	LINUX_PTRACE_SINGLESTEP,	PT_STEP,
137 # endif
138 	-1
139 };
140 
141 const struct linux_mnttypes linux_fstypes[] = {
142 	{ MOUNT_FFS,		LINUX_DEFAULT_SUPER_MAGIC	},
143 	{ MOUNT_NFS,		LINUX_NFS_SUPER_MAGIC 		},
144 	{ MOUNT_MFS,		LINUX_DEFAULT_SUPER_MAGIC	},
145 	{ MOUNT_MSDOS,		LINUX_MSDOS_SUPER_MAGIC		},
146 	{ MOUNT_LFS,		LINUX_DEFAULT_SUPER_MAGIC	},
147 	{ MOUNT_FDESC,		LINUX_DEFAULT_SUPER_MAGIC	},
148 	{ MOUNT_PORTAL,		LINUX_DEFAULT_SUPER_MAGIC	},
149 	{ MOUNT_NULL,		LINUX_DEFAULT_SUPER_MAGIC	},
150 	{ MOUNT_OVERLAY,	LINUX_DEFAULT_SUPER_MAGIC	},
151 	{ MOUNT_UMAP,		LINUX_DEFAULT_SUPER_MAGIC	},
152 	{ MOUNT_KERNFS,		LINUX_DEFAULT_SUPER_MAGIC	},
153 	{ MOUNT_PROCFS,		LINUX_PROC_SUPER_MAGIC		},
154 	{ MOUNT_AFS,		LINUX_DEFAULT_SUPER_MAGIC	},
155 	{ MOUNT_CD9660,		LINUX_ISOFS_SUPER_MAGIC		},
156 	{ MOUNT_UNION,		LINUX_DEFAULT_SUPER_MAGIC	},
157 	{ MOUNT_ADOSFS,		LINUX_ADFS_SUPER_MAGIC		},
158 	{ MOUNT_EXT2FS,		LINUX_EXT2_SUPER_MAGIC		},
159 	{ MOUNT_CFS,		LINUX_DEFAULT_SUPER_MAGIC	},
160 	{ MOUNT_CODA,		LINUX_CODA_SUPER_MAGIC		},
161 	{ MOUNT_FILECORE,	LINUX_DEFAULT_SUPER_MAGIC	},
162 	{ MOUNT_NTFS,		LINUX_DEFAULT_SUPER_MAGIC	},
163 	{ MOUNT_SMBFS,		LINUX_SMB_SUPER_MAGIC		},
164 	{ MOUNT_PTYFS,		LINUX_DEVPTS_SUPER_MAGIC	},
165 	{ MOUNT_TMPFS,		LINUX_DEFAULT_SUPER_MAGIC	}
166 };
167 const int linux_fstypes_cnt = sizeof(linux_fstypes) / sizeof(linux_fstypes[0]);
168 
169 # ifdef DEBUG_LINUX
170 #define DPRINTF(a)	uprintf a
171 # else
172 #define DPRINTF(a)
173 # endif
174 
175 /* Local linux_misc.c functions: */
176 # ifndef __amd64__
177 static void bsd_to_linux_statfs __P((const struct statvfs *,
178     struct linux_statfs *));
179 # endif
180 static void linux_to_bsd_mmap_args __P((struct sys_mmap_args *,
181     const struct linux_sys_mmap_args *));
182 static int linux_mmap __P((struct lwp *, struct linux_sys_mmap_args *,
183     register_t *, off_t));
184 
185 
186 /*
187  * The information on a terminated (or stopped) process needs
188  * to be converted in order for Linux binaries to get a valid signal
189  * number out of it.
190  */
191 void
192 bsd_to_linux_wstat(st)
193 	int *st;
194 {
195 
196 	int sig;
197 
198 	if (WIFSIGNALED(*st)) {
199 		sig = WTERMSIG(*st);
200 		if (sig >= 0 && sig < NSIG)
201 			*st= (*st& ~0177) | native_to_linux_signo[sig];
202 	} else if (WIFSTOPPED(*st)) {
203 		sig = WSTOPSIG(*st);
204 		if (sig >= 0 && sig < NSIG)
205 			*st = (*st & ~0xff00) |
206 			    (native_to_linux_signo[sig] << 8);
207 	}
208 }
209 
210 /*
211  * wait4(2).  Passed on to the NetBSD call, surrounded by code to
212  * reserve some space for a NetBSD-style wait status, and converting
213  * it to what Linux wants.
214  */
215 int
216 linux_sys_wait4(l, v, retval)
217 	struct lwp *l;
218 	void *v;
219 	register_t *retval;
220 {
221 	struct linux_sys_wait4_args /* {
222 		syscallarg(int) pid;
223 		syscallarg(int *) status;
224 		syscallarg(int) options;
225 		syscallarg(struct rusage *) rusage;
226 	} */ *uap = v;
227 	struct proc *p = l->l_proc;
228 	struct sys_wait4_args w4a;
229 	int error, *status, tstat, options, linux_options;
230 	caddr_t sg;
231 
232 	if (SCARG(uap, status) != NULL) {
233 		sg = stackgap_init(p, 0);
234 		status = (int *) stackgap_alloc(p, &sg, sizeof *status);
235 	} else
236 		status = NULL;
237 
238 	linux_options = SCARG(uap, options);
239 	options = 0;
240 	if (linux_options & ~(LINUX_WAIT4_KNOWNFLAGS))
241 		return (EINVAL);
242 
243 	if (linux_options & LINUX_WAIT4_WNOHANG)
244 		options |= WNOHANG;
245 	if (linux_options & LINUX_WAIT4_WUNTRACED)
246 		options |= WUNTRACED;
247 	if (linux_options & LINUX_WAIT4_WALL)
248 		options |= WALLSIG;
249 	if (linux_options & LINUX_WAIT4_WCLONE)
250 		options |= WALTSIG;
251 # ifdef DIAGNOSTIC
252 	if (linux_options & LINUX_WAIT4_WNOTHREAD)
253 		printf("WARNING: %s: linux process %d.%d called "
254 		       "waitpid with __WNOTHREAD set!",
255 		       __FILE__, p->p_pid, l->l_lid);
256 
257 # endif
258 
259 	SCARG(&w4a, pid) = SCARG(uap, pid);
260 	SCARG(&w4a, status) = status;
261 	SCARG(&w4a, options) = options;
262 	SCARG(&w4a, rusage) = SCARG(uap, rusage);
263 
264 	if ((error = sys_wait4(l, &w4a, retval)))
265 		return error;
266 
267 	sigdelset(&p->p_sigctx.ps_siglist, SIGCHLD);
268 
269 	if (status != NULL) {
270 		if ((error = copyin(status, &tstat, sizeof tstat)))
271 			return error;
272 
273 		bsd_to_linux_wstat(&tstat);
274 		return copyout(&tstat, SCARG(uap, status), sizeof tstat);
275 	}
276 
277 	return 0;
278 }
279 
280 /*
281  * Linux brk(2). The check if the new address is >= the old one is
282  * done in the kernel in Linux. NetBSD does it in the library.
283  */
284 int
285 linux_sys_brk(l, v, retval)
286 	struct lwp *l;
287 	void *v;
288 	register_t *retval;
289 {
290 	struct linux_sys_brk_args /* {
291 		syscallarg(char *) nsize;
292 	} */ *uap = v;
293 	struct proc *p = l->l_proc;
294 	char *nbrk = SCARG(uap, nsize);
295 	struct sys_obreak_args oba;
296 	struct vmspace *vm = p->p_vmspace;
297 	struct linux_emuldata *ed = (struct linux_emuldata*)p->p_emuldata;
298 
299 	SCARG(&oba, nsize) = nbrk;
300 
301 	if ((caddr_t) nbrk > vm->vm_daddr && sys_obreak(l, &oba, retval) == 0)
302 		ed->s->p_break = (char*)nbrk;
303 	else
304 		nbrk = ed->s->p_break;
305 
306 	retval[0] = (register_t)nbrk;
307 
308 	return 0;
309 }
310 
311 # ifndef __amd64__
312 /*
313  * Convert NetBSD statvfs structure to Linux statfs structure.
314  * Linux doesn't have f_flag, and we can't set f_frsize due
315  * to glibc statvfs() bug (see below).
316  */
317 static void
318 bsd_to_linux_statfs(bsp, lsp)
319 	const struct statvfs *bsp;
320 	struct linux_statfs *lsp;
321 {
322 	int i;
323 
324 	for (i = 0; i < linux_fstypes_cnt; i++) {
325 		if (strcmp(bsp->f_fstypename, linux_fstypes[i].bsd) == 0) {
326 			lsp->l_ftype = linux_fstypes[i].linux;
327 			break;
328 		}
329 	}
330 
331 	if (i == linux_fstypes_cnt) {
332 		DPRINTF(("unhandled fstype in linux emulation: %s\n",
333 		    bsp->f_fstypename));
334 		lsp->l_ftype = LINUX_DEFAULT_SUPER_MAGIC;
335 	}
336 
337 	/*
338 	 * The sizes are expressed in number of blocks. The block
339 	 * size used for the size is f_frsize for POSIX-compliant
340 	 * statvfs. Linux statfs uses f_bsize as the block size
341 	 * (f_frsize used to not be available in Linux struct statfs).
342 	 * However, glibc 2.3.3 statvfs() wrapper fails to adjust the block
343 	 * counts for different f_frsize if f_frsize is provided by the kernel.
344 	 * POSIX conforming apps thus get wrong size if f_frsize
345 	 * is different to f_bsize. Thus, we just pretend we don't
346 	 * support f_frsize.
347 	 */
348 
349 	lsp->l_fbsize = bsp->f_frsize;
350 	lsp->l_ffrsize = 0;			/* compat */
351 	lsp->l_fblocks = bsp->f_blocks;
352 	lsp->l_fbfree = bsp->f_bfree;
353 	lsp->l_fbavail = bsp->f_bavail;
354 	lsp->l_ffiles = bsp->f_files;
355 	lsp->l_fffree = bsp->f_ffree;
356 	/* Linux sets the fsid to 0..., we don't */
357 	lsp->l_ffsid.val[0] = bsp->f_fsidx.__fsid_val[0];
358 	lsp->l_ffsid.val[1] = bsp->f_fsidx.__fsid_val[1];
359 	lsp->l_fnamelen = bsp->f_namemax;
360 	(void)memset(lsp->l_fspare, 0, sizeof(lsp->l_fspare));
361 }
362 
363 /*
364  * Implement the fs stat functions. Straightforward.
365  */
366 int
367 linux_sys_statfs(l, v, retval)
368 	struct lwp *l;
369 	void *v;
370 	register_t *retval;
371 {
372 	struct linux_sys_statfs_args /* {
373 		syscallarg(const char *) path;
374 		syscallarg(struct linux_statfs *) sp;
375 	} */ *uap = v;
376 	struct proc *p = l->l_proc;
377 	struct statvfs *btmp, *bsp;
378 	struct linux_statfs ltmp;
379 	struct sys_statvfs1_args bsa;
380 	caddr_t sg;
381 	int error;
382 
383 	sg = stackgap_init(p, 0);
384 	bsp = stackgap_alloc(p, &sg, sizeof (struct statvfs));
385 
386 	CHECK_ALT_EXIST(l, &sg, SCARG(uap, path));
387 
388 	SCARG(&bsa, path) = SCARG(uap, path);
389 	SCARG(&bsa, buf) = bsp;
390 	SCARG(&bsa, flags) = ST_WAIT;
391 
392 	if ((error = sys_statvfs1(l, &bsa, retval)))
393 		return error;
394 
395 	btmp = STATVFSBUF_GET();
396 	error = copyin(bsp, btmp, sizeof(*btmp));
397 	if (error) {
398 		goto out;
399 	}
400 	bsd_to_linux_statfs(btmp, &ltmp);
401 	error = copyout(&ltmp, SCARG(uap, sp), sizeof ltmp);
402 out:
403 	STATVFSBUF_PUT(btmp);
404 	return error;
405 }
406 
407 int
408 linux_sys_fstatfs(l, v, retval)
409 	struct lwp *l;
410 	void *v;
411 	register_t *retval;
412 {
413 	struct linux_sys_fstatfs_args /* {
414 		syscallarg(int) fd;
415 		syscallarg(struct linux_statfs *) sp;
416 	} */ *uap = v;
417 	struct proc *p = l->l_proc;
418 	struct statvfs *btmp, *bsp;
419 	struct linux_statfs ltmp;
420 	struct sys_fstatvfs1_args bsa;
421 	caddr_t sg;
422 	int error;
423 
424 	sg = stackgap_init(p, 0);
425 	bsp = stackgap_alloc(p, &sg, sizeof (struct statvfs));
426 
427 	SCARG(&bsa, fd) = SCARG(uap, fd);
428 	SCARG(&bsa, buf) = bsp;
429 	SCARG(&bsa, flags) = ST_WAIT;
430 
431 	if ((error = sys_fstatvfs1(l, &bsa, retval)))
432 		return error;
433 
434 	btmp = STATVFSBUF_GET();
435 	error = copyin(bsp, btmp, sizeof(*btmp));
436 	if (error) {
437 		goto out;
438 	}
439 	bsd_to_linux_statfs(btmp, &ltmp);
440 	error = copyout(&ltmp, SCARG(uap, sp), sizeof ltmp);
441 out:
442 	STATVFSBUF_PUT(btmp);
443 	return error;
444 }
445 # endif /* !__amd64__ */
446 
447 /*
448  * uname(). Just copy the info from the various strings stored in the
449  * kernel, and put it in the Linux utsname structure. That structure
450  * is almost the same as the NetBSD one, only it has fields 65 characters
451  * long, and an extra domainname field.
452  */
453 int
454 linux_sys_uname(l, v, retval)
455 	struct lwp *l;
456 	void *v;
457 	register_t *retval;
458 {
459 	struct linux_sys_uname_args /* {
460 		syscallarg(struct linux_utsname *) up;
461 	} */ *uap = v;
462 	struct linux_utsname luts;
463 
464 	strncpy(luts.l_sysname, linux_sysname, sizeof(luts.l_sysname));
465 	strncpy(luts.l_nodename, hostname, sizeof(luts.l_nodename));
466 	strncpy(luts.l_release, linux_release, sizeof(luts.l_release));
467 	strncpy(luts.l_version, linux_version, sizeof(luts.l_version));
468 # ifdef LINUX_UNAME_ARCH
469 	strncpy(luts.l_machine, LINUX_UNAME_ARCH, sizeof(luts.l_machine));
470 # else
471 	strncpy(luts.l_machine, machine, sizeof(luts.l_machine));
472 # endif
473 	strncpy(luts.l_domainname, domainname, sizeof(luts.l_domainname));
474 
475 	return copyout(&luts, SCARG(uap, up), sizeof(luts));
476 }
477 
478 /* Used directly on: alpha, mips, ppc, sparc, sparc64 */
479 /* Used indirectly on: arm, i386, m68k */
480 
481 /*
482  * New type Linux mmap call.
483  * Only called directly on machines with >= 6 free regs.
484  */
485 int
486 linux_sys_mmap(l, v, retval)
487 	struct lwp *l;
488 	void *v;
489 	register_t *retval;
490 {
491 	struct linux_sys_mmap_args /* {
492 		syscallarg(unsigned long) addr;
493 		syscallarg(size_t) len;
494 		syscallarg(int) prot;
495 		syscallarg(int) flags;
496 		syscallarg(int) fd;
497 		syscallarg(linux_off_t) offset;
498 	} */ *uap = v;
499 
500 	if (SCARG(uap, offset) & PAGE_MASK)
501 		return EINVAL;
502 
503 	return linux_mmap(l, uap, retval, SCARG(uap, offset));
504 }
505 
506 /*
507  * Guts of most architectures' mmap64() implementations.  This shares
508  * its list of arguments with linux_sys_mmap().
509  *
510  * The difference in linux_sys_mmap2() is that "offset" is actually
511  * (offset / pagesize), not an absolute byte count.  This translation
512  * to pagesize offsets is done inside glibc between the mmap64() call
513  * point, and the actual syscall.
514  */
515 int
516 linux_sys_mmap2(l, v, retval)
517 	struct lwp *l;
518 	void *v;
519 	register_t *retval;
520 {
521 	struct linux_sys_mmap2_args /* {
522 		syscallarg(unsigned long) addr;
523 		syscallarg(size_t) len;
524 		syscallarg(int) prot;
525 		syscallarg(int) flags;
526 		syscallarg(int) fd;
527 		syscallarg(linux_off_t) offset;
528 	} */ *uap = v;
529 
530 	return linux_mmap(l, uap, retval,
531 	    ((off_t)SCARG(uap, offset)) << PAGE_SHIFT);
532 }
533 
534 /*
535  * Massage arguments and call system mmap(2).
536  */
537 static int
538 linux_mmap(l, uap, retval, offset)
539 	struct lwp *l;
540 	struct linux_sys_mmap_args *uap;
541 	register_t *retval;
542 	off_t offset;
543 {
544 	struct sys_mmap_args cma;
545 	int error;
546 	size_t mmoff=0;
547 
548 	if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN) {
549 		/*
550 		 * Request for stack-like memory segment. On linux, this
551 		 * works by mmap()ping (small) segment, which is automatically
552 		 * extended when page fault happens below the currently
553 		 * allocated area. We emulate this by allocating (typically
554 		 * bigger) segment sized at current stack size limit, and
555 		 * offsetting the requested and returned address accordingly.
556 		 * Since physical pages are only allocated on-demand, this
557 		 * is effectively identical.
558 		 */
559 		rlim_t ssl = l->l_proc->p_rlimit[RLIMIT_STACK].rlim_cur;
560 
561 		if (SCARG(uap, len) < ssl) {
562 			/* Compute the address offset */
563 			mmoff = round_page(ssl) - SCARG(uap, len);
564 
565 			if (SCARG(uap, addr))
566 				SCARG(uap, addr) -= mmoff;
567 
568 			SCARG(uap, len) = (size_t) ssl;
569 		}
570 	}
571 
572 	linux_to_bsd_mmap_args(&cma, uap);
573 	SCARG(&cma, pos) = offset;
574 
575 	error = sys_mmap(l, &cma, retval);
576 	if (error)
577 		return (error);
578 
579 	/* Shift the returned address for stack-like segment if necessary */
580 	if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN && mmoff)
581 		retval[0] += mmoff;
582 
583 	return (0);
584 }
585 
586 static void
587 linux_to_bsd_mmap_args(cma, uap)
588 	struct sys_mmap_args *cma;
589 	const struct linux_sys_mmap_args *uap;
590 {
591 	int flags = MAP_TRYFIXED, fl = SCARG(uap, flags);
592 
593 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_SHARED, MAP_SHARED);
594 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_PRIVATE, MAP_PRIVATE);
595 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_FIXED, MAP_FIXED);
596 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_ANON, MAP_ANON);
597 	/* XXX XAX ERH: Any other flags here?  There are more defined... */
598 
599 	SCARG(cma, addr) = (void *)SCARG(uap, addr);
600 	SCARG(cma, len) = SCARG(uap, len);
601 	SCARG(cma, prot) = SCARG(uap, prot);
602 	if (SCARG(cma, prot) & VM_PROT_WRITE) /* XXX */
603 		SCARG(cma, prot) |= VM_PROT_READ;
604 	SCARG(cma, flags) = flags;
605 	SCARG(cma, fd) = flags & MAP_ANON ? -1 : SCARG(uap, fd);
606 	SCARG(cma, pad) = 0;
607 }
608 
609 #define	LINUX_MREMAP_MAYMOVE	1
610 #define	LINUX_MREMAP_FIXED	2
611 
612 int
613 linux_sys_mremap(l, v, retval)
614 	struct lwp *l;
615 	void *v;
616 	register_t *retval;
617 {
618 	struct linux_sys_mremap_args /* {
619 		syscallarg(void *) old_address;
620 		syscallarg(size_t) old_size;
621 		syscallarg(size_t) new_size;
622 		syscallarg(u_long) flags;
623 	} */ *uap = v;
624 
625 	struct proc *p;
626 	struct vm_map *map;
627 	vaddr_t oldva;
628 	vaddr_t newva;
629 	size_t oldsize;
630 	size_t newsize;
631 	int flags;
632 	int uvmflags;
633 	int error;
634 
635 	flags = SCARG(uap, flags);
636 	oldva = (vaddr_t)SCARG(uap, old_address);
637 	oldsize = round_page(SCARG(uap, old_size));
638 	newsize = round_page(SCARG(uap, new_size));
639 	if ((flags & ~(LINUX_MREMAP_FIXED|LINUX_MREMAP_MAYMOVE)) != 0) {
640 		error = EINVAL;
641 		goto done;
642 	}
643 	if ((flags & LINUX_MREMAP_FIXED) != 0) {
644 		if ((flags & LINUX_MREMAP_MAYMOVE) == 0) {
645 			error = EINVAL;
646 			goto done;
647 		}
648 #if 0 /* notyet */
649 		newva = SCARG(uap, new_address);
650 		uvmflags = UVM_MREMAP_FIXED;
651 #else /* notyet */
652 		error = EOPNOTSUPP;
653 		goto done;
654 #endif /* notyet */
655 	} else if ((flags & LINUX_MREMAP_MAYMOVE) != 0) {
656 		uvmflags = 0;
657 	} else {
658 		newva = oldva;
659 		uvmflags = UVM_MREMAP_FIXED;
660 	}
661 	p = l->l_proc;
662 	map = &p->p_vmspace->vm_map;
663 	error = uvm_mremap(map, oldva, oldsize, map, &newva, newsize, p,
664 	    uvmflags);
665 
666 done:
667 	*retval = (error != 0) ? 0 : (register_t)newva;
668 	return error;
669 }
670 
671 int
672 linux_sys_msync(l, v, retval)
673 	struct lwp *l;
674 	void *v;
675 	register_t *retval;
676 {
677 	struct linux_sys_msync_args /* {
678 		syscallarg(caddr_t) addr;
679 		syscallarg(int) len;
680 		syscallarg(int) fl;
681 	} */ *uap = v;
682 
683 	struct sys___msync13_args bma;
684 
685 	/* flags are ignored */
686 	SCARG(&bma, addr) = SCARG(uap, addr);
687 	SCARG(&bma, len) = SCARG(uap, len);
688 	SCARG(&bma, flags) = SCARG(uap, fl);
689 
690 	return sys___msync13(l, &bma, retval);
691 }
692 
693 int
694 linux_sys_mprotect(l, v, retval)
695 	struct lwp *l;
696 	void *v;
697 	register_t *retval;
698 {
699 	struct linux_sys_mprotect_args /* {
700 		syscallarg(const void *) start;
701 		syscallarg(unsigned long) len;
702 		syscallarg(int) prot;
703 	} */ *uap = v;
704 	struct vm_map_entry *entry;
705 	struct vm_map *map;
706 	struct proc *p;
707 	vaddr_t end, start, len, stacklim;
708 	int prot, grows;
709 
710 	start = (vaddr_t)SCARG(uap, start);
711 	len = round_page(SCARG(uap, len));
712 	prot = SCARG(uap, prot);
713 	grows = prot & (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP);
714 	prot &= ~grows;
715 	end = start + len;
716 
717 	if (start & PAGE_MASK)
718 		return EINVAL;
719 	if (end < start)
720 		return EINVAL;
721 	if (end == start)
722 		return 0;
723 
724 	if (prot & ~(PROT_READ | PROT_WRITE | PROT_EXEC))
725 		return EINVAL;
726 	if (grows == (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP))
727 		return EINVAL;
728 
729 	p = l->l_proc;
730 	map = &p->p_vmspace->vm_map;
731 	vm_map_lock(map);
732 # ifdef notdef
733 	VM_MAP_RANGE_CHECK(map, start, end);
734 # endif
735 	if (!uvm_map_lookup_entry(map, start, &entry) || entry->start > start) {
736 		vm_map_unlock(map);
737 		return ENOMEM;
738 	}
739 
740 	/*
741 	 * Approximate the behaviour of PROT_GROWS{DOWN,UP}.
742 	 */
743 
744 	stacklim = (vaddr_t)p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur;
745 	if (grows & LINUX_PROT_GROWSDOWN) {
746 		if (USRSTACK - stacklim <= start && start < USRSTACK) {
747 			start = USRSTACK - stacklim;
748 		} else {
749 			start = entry->start;
750 		}
751 	} else if (grows & LINUX_PROT_GROWSUP) {
752 		if (USRSTACK <= end && end < USRSTACK + stacklim) {
753 			end = USRSTACK + stacklim;
754 		} else {
755 			end = entry->end;
756 		}
757 	}
758 	vm_map_unlock(map);
759 	return uvm_map_protect(map, start, end, prot, FALSE);
760 }
761 
762 /*
763  * This code is partly stolen from src/lib/libc/compat-43/times.c
764  */
765 
766 #define	CONVTCK(r)	(r.tv_sec * hz + r.tv_usec / (1000000 / hz))
767 
768 int
769 linux_sys_times(l, v, retval)
770 	struct lwp *l;
771 	void *v;
772 	register_t *retval;
773 {
774 	struct linux_sys_times_args /* {
775 		syscallarg(struct times *) tms;
776 	} */ *uap = v;
777 	struct proc *p = l->l_proc;
778 	struct timeval t;
779 	int error;
780 
781 	if (SCARG(uap, tms)) {
782 		struct linux_tms ltms;
783 		struct rusage ru;
784 
785 		calcru(p, &ru.ru_utime, &ru.ru_stime, NULL);
786 		ltms.ltms_utime = CONVTCK(ru.ru_utime);
787 		ltms.ltms_stime = CONVTCK(ru.ru_stime);
788 
789 		ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime);
790 		ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime);
791 
792 		if ((error = copyout(&ltms, SCARG(uap, tms), sizeof ltms)))
793 			return error;
794 	}
795 
796 	getmicrouptime(&t);
797 
798 	retval[0] = ((linux_clock_t)(CONVTCK(t)));
799 	return 0;
800 }
801 
802 #undef CONVTCK
803 
804 /*
805  * Linux 'readdir' call. This code is mostly taken from the
806  * SunOS getdents call (see compat/sunos/sunos_misc.c), though
807  * an attempt has been made to keep it a little cleaner (failing
808  * miserably, because of the cruft needed if count 1 is passed).
809  *
810  * The d_off field should contain the offset of the next valid entry,
811  * but in Linux it has the offset of the entry itself. We emulate
812  * that bug here.
813  *
814  * Read in BSD-style entries, convert them, and copy them out.
815  *
816  * Note that this doesn't handle union-mounted filesystems.
817  */
818 int
819 linux_sys_getdents(l, v, retval)
820 	struct lwp *l;
821 	void *v;
822 	register_t *retval;
823 {
824 	struct linux_sys_getdents_args /* {
825 		syscallarg(int) fd;
826 		syscallarg(struct linux_dirent *) dent;
827 		syscallarg(unsigned int) count;
828 	} */ *uap = v;
829 	struct proc *p = l->l_proc;
830 	struct dirent *bdp;
831 	struct vnode *vp;
832 	caddr_t	inp, tbuf;		/* BSD-format */
833 	int len, reclen;		/* BSD-format */
834 	caddr_t outp;			/* Linux-format */
835 	int resid, linux_reclen = 0;	/* Linux-format */
836 	struct file *fp;
837 	struct uio auio;
838 	struct iovec aiov;
839 	struct linux_dirent idb;
840 	off_t off;		/* true file offset */
841 	int buflen, error, eofflag, nbytes, oldcall;
842 	struct vattr va;
843 	off_t *cookiebuf = NULL, *cookie;
844 	int ncookies;
845 
846 	/* getvnode() will use the descriptor for us */
847 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
848 		return (error);
849 
850 	if ((fp->f_flag & FREAD) == 0) {
851 		error = EBADF;
852 		goto out1;
853 	}
854 
855 	vp = (struct vnode *)fp->f_data;
856 	if (vp->v_type != VDIR) {
857 		error = EINVAL;
858 		goto out1;
859 	}
860 
861 	if ((error = VOP_GETATTR(vp, &va, p->p_cred, l)))
862 		goto out1;
863 
864 	nbytes = SCARG(uap, count);
865 	if (nbytes == 1) {	/* emulating old, broken behaviour */
866 		nbytes = sizeof (idb);
867 		buflen = max(va.va_blocksize, nbytes);
868 		oldcall = 1;
869 	} else {
870 		buflen = min(MAXBSIZE, nbytes);
871 		if (buflen < va.va_blocksize)
872 			buflen = va.va_blocksize;
873 		oldcall = 0;
874 	}
875 	tbuf = malloc(buflen, M_TEMP, M_WAITOK);
876 
877 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
878 	off = fp->f_offset;
879 again:
880 	aiov.iov_base = tbuf;
881 	aiov.iov_len = buflen;
882 	auio.uio_iov = &aiov;
883 	auio.uio_iovcnt = 1;
884 	auio.uio_rw = UIO_READ;
885 	auio.uio_resid = buflen;
886 	auio.uio_offset = off;
887 	UIO_SETUP_SYSSPACE(&auio);
888 	/*
889          * First we read into the malloc'ed buffer, then
890          * we massage it into user space, one record at a time.
891          */
892 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf,
893 	    &ncookies);
894 	if (error)
895 		goto out;
896 
897 	inp = tbuf;
898 	outp = (caddr_t)SCARG(uap, dent);
899 	resid = nbytes;
900 	if ((len = buflen - auio.uio_resid) == 0)
901 		goto eof;
902 
903 	for (cookie = cookiebuf; len > 0; len -= reclen) {
904 		bdp = (struct dirent *)inp;
905 		reclen = bdp->d_reclen;
906 		if (reclen & 3)
907 			panic("linux_readdir");
908 		if (bdp->d_fileno == 0) {
909 			inp += reclen;	/* it is a hole; squish it out */
910 			if (cookie)
911 				off = *cookie++;
912 			else
913 				off += reclen;
914 			continue;
915 		}
916 		linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen);
917 		if (reclen > len || resid < linux_reclen) {
918 			/* entry too big for buffer, so just stop */
919 			outp++;
920 			break;
921 		}
922 		/*
923 		 * Massage in place to make a Linux-shaped dirent (otherwise
924 		 * we have to worry about touching user memory outside of
925 		 * the copyout() call).
926 		 */
927 		idb.d_ino = bdp->d_fileno;
928 		/*
929 		 * The old readdir() call misuses the offset and reclen fields.
930 		 */
931 		if (oldcall) {
932 			idb.d_off = (linux_off_t)linux_reclen;
933 			idb.d_reclen = (u_short)bdp->d_namlen;
934 		} else {
935 			if (sizeof (idb.d_off) <= 4 && (off >> 32) != 0) {
936 				compat_offseterr(vp, "linux_getdents");
937 				error = EINVAL;
938 				goto out;
939 			}
940 			idb.d_off = (linux_off_t)off;
941 			idb.d_reclen = (u_short)linux_reclen;
942 		}
943 		strcpy(idb.d_name, bdp->d_name);
944 		if ((error = copyout((caddr_t)&idb, outp, linux_reclen)))
945 			goto out;
946 		/* advance past this real entry */
947 		inp += reclen;
948 		if (cookie)
949 			off = *cookie++; /* each entry points to itself */
950 		else
951 			off += reclen;
952 		/* advance output past Linux-shaped entry */
953 		outp += linux_reclen;
954 		resid -= linux_reclen;
955 		if (oldcall)
956 			break;
957 	}
958 
959 	/* if we squished out the whole block, try again */
960 	if (outp == (caddr_t)SCARG(uap, dent))
961 		goto again;
962 	fp->f_offset = off;	/* update the vnode offset */
963 
964 	if (oldcall)
965 		nbytes = resid + linux_reclen;
966 
967 eof:
968 	*retval = nbytes - resid;
969 out:
970 	VOP_UNLOCK(vp, 0);
971 	if (cookiebuf)
972 		free(cookiebuf, M_TEMP);
973 	free(tbuf, M_TEMP);
974 out1:
975 	FILE_UNUSE(fp, l);
976 	return error;
977 }
978 
979 /*
980  * Even when just using registers to pass arguments to syscalls you can
981  * have 5 of them on the i386. So this newer version of select() does
982  * this.
983  */
984 int
985 linux_sys_select(l, v, retval)
986 	struct lwp *l;
987 	void *v;
988 	register_t *retval;
989 {
990 	struct linux_sys_select_args /* {
991 		syscallarg(int) nfds;
992 		syscallarg(fd_set *) readfds;
993 		syscallarg(fd_set *) writefds;
994 		syscallarg(fd_set *) exceptfds;
995 		syscallarg(struct timeval *) timeout;
996 	} */ *uap = v;
997 
998 	return linux_select1(l, retval, SCARG(uap, nfds), SCARG(uap, readfds),
999 	    SCARG(uap, writefds), SCARG(uap, exceptfds), SCARG(uap, timeout));
1000 }
1001 
1002 /*
1003  * Common code for the old and new versions of select(). A couple of
1004  * things are important:
1005  * 1) return the amount of time left in the 'timeout' parameter
1006  * 2) select never returns ERESTART on Linux, always return EINTR
1007  */
1008 int
1009 linux_select1(l, retval, nfds, readfds, writefds, exceptfds, timeout)
1010 	struct lwp *l;
1011 	register_t *retval;
1012 	int nfds;
1013 	fd_set *readfds, *writefds, *exceptfds;
1014 	struct timeval *timeout;
1015 {
1016 	struct sys_select_args bsa;
1017 	struct proc *p = l->l_proc;
1018 	struct timeval tv0, tv1, utv, *tvp;
1019 	caddr_t sg;
1020 	int error;
1021 
1022 	SCARG(&bsa, nd) = nfds;
1023 	SCARG(&bsa, in) = readfds;
1024 	SCARG(&bsa, ou) = writefds;
1025 	SCARG(&bsa, ex) = exceptfds;
1026 	SCARG(&bsa, tv) = timeout;
1027 
1028 	/*
1029 	 * Store current time for computation of the amount of
1030 	 * time left.
1031 	 */
1032 	if (timeout) {
1033 		if ((error = copyin(timeout, &utv, sizeof(utv))))
1034 			return error;
1035 		if (itimerfix(&utv)) {
1036 			/*
1037 			 * The timeval was invalid.  Convert it to something
1038 			 * valid that will act as it does under Linux.
1039 			 */
1040 			sg = stackgap_init(p, 0);
1041 			tvp = stackgap_alloc(p, &sg, sizeof(utv));
1042 			utv.tv_sec += utv.tv_usec / 1000000;
1043 			utv.tv_usec %= 1000000;
1044 			if (utv.tv_usec < 0) {
1045 				utv.tv_sec -= 1;
1046 				utv.tv_usec += 1000000;
1047 			}
1048 			if (utv.tv_sec < 0)
1049 				timerclear(&utv);
1050 			if ((error = copyout(&utv, tvp, sizeof(utv))))
1051 				return error;
1052 			SCARG(&bsa, tv) = tvp;
1053 		}
1054 		microtime(&tv0);
1055 	}
1056 
1057 	error = sys_select(l, &bsa, retval);
1058 	if (error) {
1059 		/*
1060 		 * See fs/select.c in the Linux kernel.  Without this,
1061 		 * Maelstrom doesn't work.
1062 		 */
1063 		if (error == ERESTART)
1064 			error = EINTR;
1065 		return error;
1066 	}
1067 
1068 	if (timeout) {
1069 		if (*retval) {
1070 			/*
1071 			 * Compute how much time was left of the timeout,
1072 			 * by subtracting the current time and the time
1073 			 * before we started the call, and subtracting
1074 			 * that result from the user-supplied value.
1075 			 */
1076 			microtime(&tv1);
1077 			timersub(&tv1, &tv0, &tv1);
1078 			timersub(&utv, &tv1, &utv);
1079 			if (utv.tv_sec < 0)
1080 				timerclear(&utv);
1081 		} else
1082 			timerclear(&utv);
1083 		if ((error = copyout(&utv, timeout, sizeof(utv))))
1084 			return error;
1085 	}
1086 
1087 	return 0;
1088 }
1089 
1090 /*
1091  * Get the process group of a certain process. Look it up
1092  * and return the value.
1093  */
1094 int
1095 linux_sys_getpgid(l, v, retval)
1096 	struct lwp *l;
1097 	void *v;
1098 	register_t *retval;
1099 {
1100 	struct linux_sys_getpgid_args /* {
1101 		syscallarg(int) pid;
1102 	} */ *uap = v;
1103 	struct proc *p = l->l_proc;
1104 	struct proc *targp;
1105 
1106 	if (SCARG(uap, pid) != 0 && SCARG(uap, pid) != p->p_pid) {
1107 		if ((targp = pfind(SCARG(uap, pid))) == 0)
1108 			return ESRCH;
1109 	}
1110 	else
1111 		targp = p;
1112 
1113 	retval[0] = targp->p_pgid;
1114 	return 0;
1115 }
1116 
1117 /*
1118  * Set the 'personality' (emulation mode) for the current process. Only
1119  * accept the Linux personality here (0). This call is needed because
1120  * the Linux ELF crt0 issues it in an ugly kludge to make sure that
1121  * ELF binaries run in Linux mode, not SVR4 mode.
1122  */
1123 int
1124 linux_sys_personality(l, v, retval)
1125 	struct lwp *l;
1126 	void *v;
1127 	register_t *retval;
1128 {
1129 	struct linux_sys_personality_args /* {
1130 		syscallarg(int) per;
1131 	} */ *uap = v;
1132 
1133 	if (SCARG(uap, per) != 0)
1134 		return EINVAL;
1135 	retval[0] = 0;
1136 	return 0;
1137 }
1138 #endif /* !COMPAT_LINUX32 */
1139 
1140 #if defined(__i386__) || defined(__m68k__) || defined(COMPAT_LINUX32)
1141 /*
1142  * The calls are here because of type conversions.
1143  */
1144 int
1145 linux_sys_setreuid16(l, v, retval)
1146 	struct lwp *l;
1147 	void *v;
1148 	register_t *retval;
1149 {
1150 	struct linux_sys_setreuid16_args /* {
1151 		syscallarg(int) ruid;
1152 		syscallarg(int) euid;
1153 	} */ *uap = v;
1154 	struct sys_setreuid_args bsa;
1155 
1156 	SCARG(&bsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
1157 		(uid_t)-1 : SCARG(uap, ruid);
1158 	SCARG(&bsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
1159 		(uid_t)-1 : SCARG(uap, euid);
1160 
1161 	return sys_setreuid(l, &bsa, retval);
1162 }
1163 
1164 int
1165 linux_sys_setregid16(l, v, retval)
1166 	struct lwp *l;
1167 	void *v;
1168 	register_t *retval;
1169 {
1170 	struct linux_sys_setregid16_args /* {
1171 		syscallarg(int) rgid;
1172 		syscallarg(int) egid;
1173 	} */ *uap = v;
1174 	struct sys_setregid_args bsa;
1175 
1176 	SCARG(&bsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
1177 		(uid_t)-1 : SCARG(uap, rgid);
1178 	SCARG(&bsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
1179 		(uid_t)-1 : SCARG(uap, egid);
1180 
1181 	return sys_setregid(l, &bsa, retval);
1182 }
1183 
1184 int
1185 linux_sys_setresuid16(l, v, retval)
1186 	struct lwp *l;
1187 	void *v;
1188 	register_t *retval;
1189 {
1190 	struct linux_sys_setresuid16_args /* {
1191 		syscallarg(uid_t) ruid;
1192 		syscallarg(uid_t) euid;
1193 		syscallarg(uid_t) suid;
1194 	} */ *uap = v;
1195 	struct linux_sys_setresuid16_args lsa;
1196 
1197 	SCARG(&lsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
1198 		(uid_t)-1 : SCARG(uap, ruid);
1199 	SCARG(&lsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
1200 		(uid_t)-1 : SCARG(uap, euid);
1201 	SCARG(&lsa, suid) = ((linux_uid_t)SCARG(uap, suid) == (linux_uid_t)-1) ?
1202 		(uid_t)-1 : SCARG(uap, suid);
1203 
1204 	return linux_sys_setresuid(l, &lsa, retval);
1205 }
1206 
1207 int
1208 linux_sys_setresgid16(l, v, retval)
1209 	struct lwp *l;
1210 	void *v;
1211 	register_t *retval;
1212 {
1213 	struct linux_sys_setresgid16_args /* {
1214 		syscallarg(gid_t) rgid;
1215 		syscallarg(gid_t) egid;
1216 		syscallarg(gid_t) sgid;
1217 	} */ *uap = v;
1218 	struct linux_sys_setresgid16_args lsa;
1219 
1220 	SCARG(&lsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
1221 		(gid_t)-1 : SCARG(uap, rgid);
1222 	SCARG(&lsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
1223 		(gid_t)-1 : SCARG(uap, egid);
1224 	SCARG(&lsa, sgid) = ((linux_gid_t)SCARG(uap, sgid) == (linux_gid_t)-1) ?
1225 		(gid_t)-1 : SCARG(uap, sgid);
1226 
1227 	return linux_sys_setresgid(l, &lsa, retval);
1228 }
1229 
1230 int
1231 linux_sys_getgroups16(l, v, retval)
1232 	struct lwp *l;
1233 	void *v;
1234 	register_t *retval;
1235 {
1236 	struct linux_sys_getgroups16_args /* {
1237 		syscallarg(int) gidsetsize;
1238 		syscallarg(linux_gid_t *) gidset;
1239 	} */ *uap = v;
1240 	struct proc *p = l->l_proc;
1241 	caddr_t sg;
1242 	int n, error, i;
1243 	struct sys_getgroups_args bsa;
1244 	gid_t *bset, *kbset;
1245 	linux_gid_t *lset;
1246 	kauth_cred_t pc = p->p_cred;
1247 
1248 	n = SCARG(uap, gidsetsize);
1249 	if (n < 0)
1250 		return EINVAL;
1251 	error = 0;
1252 	bset = kbset = NULL;
1253 	lset = NULL;
1254 	if (n > 0) {
1255 		n = min(kauth_cred_ngroups(pc), n);
1256 		sg = stackgap_init(p, 0);
1257 		bset = stackgap_alloc(p, &sg, n * sizeof (gid_t));
1258 		kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK);
1259 		lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
1260 		if (bset == NULL || kbset == NULL || lset == NULL)
1261 		{
1262 			error = ENOMEM;
1263 			goto out;
1264 		}
1265 		SCARG(&bsa, gidsetsize) = n;
1266 		SCARG(&bsa, gidset) = bset;
1267 		error = sys_getgroups(l, &bsa, retval);
1268 		if (error != 0)
1269 			goto out;
1270 		error = copyin(bset, kbset, n * sizeof (gid_t));
1271 		if (error != 0)
1272 			goto out;
1273 		for (i = 0; i < n; i++)
1274 			lset[i] = (linux_gid_t)kbset[i];
1275 		error = copyout(lset, SCARG(uap, gidset),
1276 		    n * sizeof (linux_gid_t));
1277 	} else
1278 		*retval = kauth_cred_ngroups(pc);
1279 out:
1280 	if (kbset != NULL)
1281 		free(kbset, M_TEMP);
1282 	if (lset != NULL)
1283 		free(lset, M_TEMP);
1284 	return error;
1285 }
1286 
1287 int
1288 linux_sys_setgroups16(l, v, retval)
1289 	struct lwp *l;
1290 	void *v;
1291 	register_t *retval;
1292 {
1293 	struct linux_sys_setgroups16_args /* {
1294 		syscallarg(int) gidsetsize;
1295 		syscallarg(linux_gid_t *) gidset;
1296 	} */ *uap = v;
1297 	struct proc *p = l->l_proc;
1298 	caddr_t sg;
1299 	int n;
1300 	int error, i;
1301 	struct sys_setgroups_args bsa;
1302 	gid_t *bset, *kbset;
1303 	linux_gid_t *lset;
1304 
1305 	n = SCARG(uap, gidsetsize);
1306 	if (n < 0 || n > NGROUPS)
1307 		return EINVAL;
1308 	sg = stackgap_init(p, 0);
1309 	bset = stackgap_alloc(p, &sg, n * sizeof (gid_t));
1310 	lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
1311 	kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK);
1312 	if (bset == NULL || kbset == NULL || lset == NULL)
1313 	{
1314 		error = ENOMEM;
1315 		goto out;
1316 	}
1317 	error = copyin(SCARG(uap, gidset), lset, n * sizeof (linux_gid_t));
1318 	if (error != 0)
1319 		goto out;
1320 	for (i = 0; i < n; i++)
1321 		kbset[i] = (gid_t)lset[i];
1322 	error = copyout(kbset, bset, n * sizeof (gid_t));
1323 	if (error != 0)
1324 		goto out;
1325 	SCARG(&bsa, gidsetsize) = n;
1326 	SCARG(&bsa, gidset) = bset;
1327 	error = sys_setgroups(l, &bsa, retval);
1328 
1329 out:
1330 	if (lset != NULL)
1331 		free(lset, M_TEMP);
1332 	if (kbset != NULL)
1333 		free(kbset, M_TEMP);
1334 
1335 	return error;
1336 }
1337 
1338 #endif /* __i386__ || __m68k__ || COMPAT_LINUX32 */
1339 
1340 #ifndef COMPAT_LINUX32
1341 /*
1342  * We have nonexistent fsuid equal to uid.
1343  * If modification is requested, refuse.
1344  */
1345 int
1346 linux_sys_setfsuid(l, v, retval)
1347 	 struct lwp *l;
1348 	 void *v;
1349 	 register_t *retval;
1350 {
1351 	 struct linux_sys_setfsuid_args /* {
1352 		 syscallarg(uid_t) uid;
1353 	 } */ *uap = v;
1354 	 struct proc *p = l->l_proc;
1355 	 uid_t uid;
1356 
1357 	 uid = SCARG(uap, uid);
1358 	 if (kauth_cred_getuid(p->p_cred) != uid)
1359 		 return sys_nosys(l, v, retval);
1360 	 else
1361 		 return (0);
1362 }
1363 
1364 /* XXX XXX XXX */
1365 # ifndef alpha
1366 int
1367 linux_sys_getfsuid(l, v, retval)
1368 	struct lwp *l;
1369 	void *v;
1370 	register_t *retval;
1371 {
1372 	return sys_getuid(l, v, retval);
1373 }
1374 # endif
1375 
1376 int
1377 linux_sys_setresuid(l, v, retval)
1378 	struct lwp *l;
1379 	void *v;
1380 	register_t *retval;
1381 {
1382 	struct linux_sys_setresuid_args /* {
1383 		syscallarg(uid_t) ruid;
1384 		syscallarg(uid_t) euid;
1385 		syscallarg(uid_t) suid;
1386 	} */ *uap = v;
1387 
1388 	/*
1389 	 * Note: These checks are a little different than the NetBSD
1390 	 * setreuid(2) call performs.  This precisely follows the
1391 	 * behavior of the Linux kernel.
1392 	 */
1393 
1394 	return do_setresuid(l, SCARG(uap, ruid), SCARG(uap, euid),
1395 			    SCARG(uap, suid),
1396 			    ID_R_EQ_R | ID_R_EQ_E | ID_R_EQ_S |
1397 			    ID_E_EQ_R | ID_E_EQ_E | ID_E_EQ_S |
1398 			    ID_S_EQ_R | ID_S_EQ_E | ID_S_EQ_S );
1399 }
1400 
1401 int
1402 linux_sys_getresuid(l, v, retval)
1403 	struct lwp *l;
1404 	void *v;
1405 	register_t *retval;
1406 {
1407 	struct linux_sys_getresuid_args /* {
1408 		syscallarg(uid_t *) ruid;
1409 		syscallarg(uid_t *) euid;
1410 		syscallarg(uid_t *) suid;
1411 	} */ *uap = v;
1412 	struct proc *p = l->l_proc;
1413 	kauth_cred_t pc = p->p_cred;
1414 	int error;
1415 	uid_t uid;
1416 
1417 	/*
1418 	 * Linux copies these values out to userspace like so:
1419 	 *
1420 	 *	1. Copy out ruid.
1421 	 *	2. If that succeeds, copy out euid.
1422 	 *	3. If both of those succeed, copy out suid.
1423 	 */
1424 	uid = kauth_cred_getuid(pc);
1425 	if ((error = copyout(&uid, SCARG(uap, ruid), sizeof(uid_t))) != 0)
1426 		return (error);
1427 
1428 	uid = kauth_cred_geteuid(pc);
1429 	if ((error = copyout(&uid, SCARG(uap, euid), sizeof(uid_t))) != 0)
1430 		return (error);
1431 
1432 	uid = kauth_cred_getsvuid(pc);
1433 
1434 	return (copyout(&uid, SCARG(uap, suid), sizeof(uid_t)));
1435 }
1436 
1437 int
1438 linux_sys_ptrace(l, v, retval)
1439 	struct lwp *l;
1440 	void *v;
1441 	register_t *retval;
1442 {
1443 	struct linux_sys_ptrace_args /* {
1444 		i386, m68k, powerpc: T=int
1445 		alpha, amd64: T=long
1446 		syscallarg(T) request;
1447 		syscallarg(T) pid;
1448 		syscallarg(T) addr;
1449 		syscallarg(T) data;
1450 	} */ *uap = v;
1451 	const int *ptr;
1452 	int request;
1453 	int error;
1454 
1455 	ptr = linux_ptrace_request_map;
1456 	request = SCARG(uap, request);
1457 	while (*ptr != -1)
1458 		if (*ptr++ == request) {
1459 			struct sys_ptrace_args pta;
1460 
1461 			SCARG(&pta, req) = *ptr;
1462 			SCARG(&pta, pid) = SCARG(uap, pid);
1463 			SCARG(&pta, addr) = (caddr_t)SCARG(uap, addr);
1464 			SCARG(&pta, data) = SCARG(uap, data);
1465 
1466 			/*
1467 			 * Linux ptrace(PTRACE_CONT, pid, 0, 0) means actually
1468 			 * to continue where the process left off previously.
1469 			 * The same thing is achieved by addr == (caddr_t) 1
1470 			 * on NetBSD, so rewrite 'addr' appropriately.
1471 			 */
1472 			if (request == LINUX_PTRACE_CONT && SCARG(uap, addr)==0)
1473 				SCARG(&pta, addr) = (caddr_t) 1;
1474 
1475 			error = sys_ptrace(l, &pta, retval);
1476 			if (error)
1477 				return error;
1478 			switch (request) {
1479 			case LINUX_PTRACE_PEEKTEXT:
1480 			case LINUX_PTRACE_PEEKDATA:
1481 				error = copyout (retval,
1482 				    (caddr_t)SCARG(uap, data),
1483 				    sizeof *retval);
1484 				*retval = SCARG(uap, data);
1485 				break;
1486 			default:
1487 				break;
1488 			}
1489 			return error;
1490 		}
1491 		else
1492 			ptr++;
1493 
1494 	return LINUX_SYS_PTRACE_ARCH(l, uap, retval);
1495 }
1496 
1497 int
1498 linux_sys_reboot(struct lwp *l, void *v, register_t *retval)
1499 {
1500 	struct linux_sys_reboot_args /* {
1501 		syscallarg(int) magic1;
1502 		syscallarg(int) magic2;
1503 		syscallarg(int) cmd;
1504 		syscallarg(void *) arg;
1505 	} */ *uap = v;
1506 	struct sys_reboot_args /* {
1507 		syscallarg(int) opt;
1508 		syscallarg(char *) bootstr;
1509 	} */ sra;
1510 	struct proc *p = l->l_proc;
1511 	int error;
1512 
1513 	if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, &p->p_acflag)) != 0)
1514 		return(error);
1515 
1516 	if (SCARG(uap, magic1) != LINUX_REBOOT_MAGIC1)
1517 		return(EINVAL);
1518 	if (SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2 &&
1519 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2A &&
1520 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2B)
1521 		return(EINVAL);
1522 
1523 	switch (SCARG(uap, cmd)) {
1524 	case LINUX_REBOOT_CMD_RESTART:
1525 		SCARG(&sra, opt) = RB_AUTOBOOT;
1526 		break;
1527 	case LINUX_REBOOT_CMD_HALT:
1528 		SCARG(&sra, opt) = RB_HALT;
1529 		break;
1530 	case LINUX_REBOOT_CMD_POWER_OFF:
1531 		SCARG(&sra, opt) = RB_HALT|RB_POWERDOWN;
1532 		break;
1533 	case LINUX_REBOOT_CMD_RESTART2:
1534 		/* Reboot with an argument. */
1535 		SCARG(&sra, opt) = RB_AUTOBOOT|RB_STRING;
1536 		SCARG(&sra, bootstr) = SCARG(uap, arg);
1537 		break;
1538 	case LINUX_REBOOT_CMD_CAD_ON:
1539 		return(EINVAL);	/* We don't implement ctrl-alt-delete */
1540 	case LINUX_REBOOT_CMD_CAD_OFF:
1541 		return(0);
1542 	default:
1543 		return(EINVAL);
1544 	}
1545 
1546 	return(sys_reboot(l, &sra, retval));
1547 }
1548 
1549 /*
1550  * Copy of compat_12_sys_swapon().
1551  */
1552 int
1553 linux_sys_swapon(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_swapon_args /* {
1560 		syscallarg(const char *) name;
1561 	} */ *uap = v;
1562 
1563 	SCARG(&ua, cmd) = SWAP_ON;
1564 	SCARG(&ua, arg) = (void *)__UNCONST(SCARG(uap, name));
1565 	SCARG(&ua, misc) = 0;	/* priority */
1566 	return (sys_swapctl(l, &ua, retval));
1567 }
1568 
1569 /*
1570  * Stop swapping to the file or block device specified by path.
1571  */
1572 int
1573 linux_sys_swapoff(l, v, retval)
1574 	struct lwp *l;
1575 	void *v;
1576 	register_t *retval;
1577 {
1578 	struct sys_swapctl_args ua;
1579 	struct linux_sys_swapoff_args /* {
1580 		syscallarg(const char *) path;
1581 	} */ *uap = v;
1582 
1583 	SCARG(&ua, cmd) = SWAP_OFF;
1584 	SCARG(&ua, arg) = __UNCONST(SCARG(uap, path)); /*XXXUNCONST*/
1585 	return (sys_swapctl(l, &ua, retval));
1586 }
1587 
1588 /*
1589  * Copy of compat_09_sys_setdomainname()
1590  */
1591 /* ARGSUSED */
1592 int
1593 linux_sys_setdomainname(l, v, retval)
1594 	struct lwp *l;
1595 	void *v;
1596 	register_t *retval;
1597 {
1598 	struct linux_sys_setdomainname_args /* {
1599 		syscallarg(char *) domainname;
1600 		syscallarg(int) len;
1601 	} */ *uap = v;
1602 	int name[2];
1603 
1604 	name[0] = CTL_KERN;
1605 	name[1] = KERN_DOMAINNAME;
1606 	return (old_sysctl(&name[0], 2, 0, 0, SCARG(uap, domainname),
1607 			    SCARG(uap, len), l));
1608 }
1609 
1610 /*
1611  * sysinfo()
1612  */
1613 /* ARGSUSED */
1614 int
1615 linux_sys_sysinfo(l, v, retval)
1616 	struct lwp *l;
1617 	void *v;
1618 	register_t *retval;
1619 {
1620 	struct linux_sys_sysinfo_args /* {
1621 		syscallarg(struct linux_sysinfo *) arg;
1622 	} */ *uap = v;
1623 	struct linux_sysinfo si;
1624 	struct loadavg *la;
1625 
1626 	si.uptime = time_uptime;
1627 	la = &averunnable;
1628 	si.loads[0] = la->ldavg[0] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1629 	si.loads[1] = la->ldavg[1] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1630 	si.loads[2] = la->ldavg[2] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1631 	si.totalram = ctob(physmem);
1632 	si.freeram = uvmexp.free * uvmexp.pagesize;
1633 	si.sharedram = 0;	/* XXX */
1634 	si.bufferram = uvmexp.filepages * uvmexp.pagesize;
1635 	si.totalswap = uvmexp.swpages * uvmexp.pagesize;
1636 	si.freeswap = (uvmexp.swpages - uvmexp.swpginuse) * uvmexp.pagesize;
1637 	si.procs = nprocs;
1638 
1639 	/* The following are only present in newer Linux kernels. */
1640 	si.totalbig = 0;
1641 	si.freebig = 0;
1642 	si.mem_unit = 1;
1643 
1644 	return (copyout(&si, SCARG(uap, arg), sizeof si));
1645 }
1646 
1647 int
1648 linux_sys_getrlimit(l, v, retval)
1649 	struct lwp *l;
1650 	void *v;
1651 	register_t *retval;
1652 {
1653 	struct linux_sys_getrlimit_args /* {
1654 		syscallarg(int) which;
1655 # ifdef LINUX_LARGEFILE64
1656 		syscallarg(struct rlimit *) rlp;
1657 # else
1658 		syscallarg(struct orlimit *) rlp;
1659 # endif
1660 	} */ *uap = v;
1661 	struct proc *p = l->l_proc;
1662 	caddr_t sg = stackgap_init(p, 0);
1663 	struct sys_getrlimit_args ap;
1664 	struct rlimit rl;
1665 # ifdef LINUX_LARGEFILE64
1666 	struct rlimit orl;
1667 # else
1668 	struct orlimit orl;
1669 # endif
1670 	int error;
1671 
1672 	SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which));
1673 	if ((error = SCARG(&ap, which)) < 0)
1674 		return -error;
1675 	SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl);
1676 	if ((error = sys_getrlimit(l, &ap, retval)) != 0)
1677 		return error;
1678 	if ((error = copyin(SCARG(&ap, rlp), &rl, sizeof(rl))) != 0)
1679 		return error;
1680 	bsd_to_linux_rlimit(&orl, &rl);
1681 
1682 	return copyout(&orl, SCARG(uap, rlp), sizeof(orl));
1683 }
1684 
1685 int
1686 linux_sys_setrlimit(l, v, retval)
1687 	struct lwp *l;
1688 	void *v;
1689 	register_t *retval;
1690 {
1691 	struct linux_sys_setrlimit_args /* {
1692 		syscallarg(int) which;
1693 # ifdef LINUX_LARGEFILE64
1694 		syscallarg(struct rlimit *) rlp;
1695 # else
1696 		syscallarg(struct orlimit *) rlp;
1697 # endif
1698 	} */ *uap = v;
1699 	struct proc *p = l->l_proc;
1700 	caddr_t sg = stackgap_init(p, 0);
1701 	struct sys_getrlimit_args ap;
1702 	struct rlimit rl;
1703 # ifdef LINUX_LARGEFILE64
1704 	struct rlimit orl;
1705 # else
1706 	struct orlimit orl;
1707 # endif
1708 	int error;
1709 
1710 	SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which));
1711 	SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl);
1712 	if ((error = SCARG(&ap, which)) < 0)
1713 		return -error;
1714 	if ((error = copyin(SCARG(uap, rlp), &orl, sizeof(orl))) != 0)
1715 		return error;
1716 	linux_to_bsd_rlimit(&rl, &orl);
1717 	if ((error = copyout(&rl, SCARG(&ap, rlp), sizeof(rl))) != 0)
1718 		return error;
1719 	return sys_setrlimit(l, &ap, retval);
1720 }
1721 
1722 # if !defined(__mips__) && !defined(__amd64__)
1723 /* XXX: this doesn't look 100% common, at least mips doesn't have it */
1724 int
1725 linux_sys_ugetrlimit(l, v, retval)
1726 	struct lwp *l;
1727 	void *v;
1728 	register_t *retval;
1729 {
1730 	return linux_sys_getrlimit(l, v, retval);
1731 }
1732 # endif
1733 
1734 /*
1735  * This gets called for unsupported syscalls. The difference to sys_nosys()
1736  * is that process does not get SIGSYS, the call just returns with ENOSYS.
1737  * This is the way Linux does it and glibc depends on this behaviour.
1738  */
1739 int
1740 linux_sys_nosys(l, v, retval)
1741 	struct lwp *l;
1742 	void *v;
1743 	register_t *retval;
1744 {
1745 	return (ENOSYS);
1746 }
1747 
1748 int
1749 linux_sys_getpriority(l, v, retval)
1750         struct lwp *l;
1751         void *v;
1752         register_t *retval;
1753 {
1754         struct linux_sys_getpriority_args /* {
1755                 syscallarg(int) which;
1756                 syscallarg(int) who;
1757         } */ *uap = v;
1758         struct sys_getpriority_args bsa;
1759         int error;
1760 
1761         SCARG(&bsa, which) = SCARG(uap, which);
1762         SCARG(&bsa, who) = SCARG(uap, who);
1763 
1764         if ((error = sys_getpriority(l, &bsa, retval)))
1765                 return error;
1766 
1767         *retval = NZERO - *retval;
1768 
1769         return 0;
1770 }
1771 
1772 #endif /* !COMPAT_LINUX32 */
1773