xref: /netbsd-src/sys/compat/linux/common/linux_misc.c (revision 1ca06f9c9235889e2ff6dc77279d01d151d70a9a)
1 /*	$NetBSD: linux_misc.c,v 1.213 2009/12/05 20:11:17 pooka Exp $	*/
2 
3 /*-
4  * Copyright (c) 1995, 1998, 1999, 2008 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  *
20  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
21  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
22  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
23  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
24  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
25  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
26  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
27  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
28  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
29  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
30  * POSSIBILITY OF SUCH DAMAGE.
31  */
32 
33 /*
34  * Linux compatibility module. Try to deal with various Linux system calls.
35  */
36 
37 /*
38  * These functions have been moved to multiarch to allow
39  * selection of which machines include them to be
40  * determined by the individual files.linux_<arch> files.
41  *
42  * Function in multiarch:
43  *	linux_sys_break			: linux_break.c
44  *	linux_sys_alarm			: linux_misc_notalpha.c
45  *	linux_sys_getresgid		: linux_misc_notalpha.c
46  *	linux_sys_nice			: linux_misc_notalpha.c
47  *	linux_sys_readdir		: linux_misc_notalpha.c
48  *	linux_sys_setresgid		: linux_misc_notalpha.c
49  *	linux_sys_time			: linux_misc_notalpha.c
50  *	linux_sys_utime			: linux_misc_notalpha.c
51  *	linux_sys_waitpid		: linux_misc_notalpha.c
52  *	linux_sys_old_mmap		: linux_oldmmap.c
53  *	linux_sys_oldolduname		: linux_oldolduname.c
54  *	linux_sys_oldselect		: linux_oldselect.c
55  *	linux_sys_olduname		: linux_olduname.c
56  *	linux_sys_pipe			: linux_pipe.c
57  */
58 
59 #include <sys/cdefs.h>
60 __KERNEL_RCSID(0, "$NetBSD: linux_misc.c,v 1.213 2009/12/05 20:11:17 pooka Exp $");
61 
62 #include <sys/param.h>
63 #include <sys/systm.h>
64 #include <sys/namei.h>
65 #include <sys/proc.h>
66 #include <sys/dirent.h>
67 #include <sys/file.h>
68 #include <sys/stat.h>
69 #include <sys/filedesc.h>
70 #include <sys/ioctl.h>
71 #include <sys/kernel.h>
72 #include <sys/malloc.h>
73 #include <sys/mbuf.h>
74 #include <sys/mman.h>
75 #include <sys/mount.h>
76 #include <sys/prot.h>
77 #include <sys/reboot.h>
78 #include <sys/resource.h>
79 #include <sys/resourcevar.h>
80 #include <sys/select.h>
81 #include <sys/signal.h>
82 #include <sys/signalvar.h>
83 #include <sys/socket.h>
84 #include <sys/time.h>
85 #include <sys/times.h>
86 #include <sys/vnode.h>
87 #include <sys/uio.h>
88 #include <sys/wait.h>
89 #include <sys/utsname.h>
90 #include <sys/unistd.h>
91 #include <sys/vfs_syscalls.h>
92 #include <sys/swap.h>		/* for SWAP_ON */
93 #include <sys/sysctl.h>		/* for KERN_DOMAINNAME */
94 #include <sys/kauth.h>
95 
96 #include <sys/ptrace.h>
97 #include <machine/ptrace.h>
98 
99 #include <sys/syscall.h>
100 #include <sys/syscallargs.h>
101 
102 #include <compat/sys/resource.h>
103 
104 #include <compat/linux/common/linux_machdep.h>
105 #include <compat/linux/common/linux_types.h>
106 #include <compat/linux/common/linux_signal.h>
107 #include <compat/linux/common/linux_ipc.h>
108 #include <compat/linux/common/linux_sem.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_statfs.h>
119 #include <compat/linux/common/linux_limit.h>
120 #endif
121 #include <compat/linux/common/linux_ptrace.h>
122 #include <compat/linux/common/linux_reboot.h>
123 #include <compat/linux/common/linux_emuldata.h>
124 
125 #ifndef COMPAT_LINUX32
126 const int linux_ptrace_request_map[] = {
127 	LINUX_PTRACE_TRACEME,	PT_TRACE_ME,
128 	LINUX_PTRACE_PEEKTEXT,	PT_READ_I,
129 	LINUX_PTRACE_PEEKDATA,	PT_READ_D,
130 	LINUX_PTRACE_POKETEXT,	PT_WRITE_I,
131 	LINUX_PTRACE_POKEDATA,	PT_WRITE_D,
132 	LINUX_PTRACE_CONT,	PT_CONTINUE,
133 	LINUX_PTRACE_KILL,	PT_KILL,
134 	LINUX_PTRACE_ATTACH,	PT_ATTACH,
135 	LINUX_PTRACE_DETACH,	PT_DETACH,
136 # ifdef PT_STEP
137 	LINUX_PTRACE_SINGLESTEP,	PT_STEP,
138 # endif
139 	LINUX_PTRACE_SYSCALL,	PT_SYSCALL,
140 	-1
141 };
142 
143 const struct linux_mnttypes linux_fstypes[] = {
144 	{ MOUNT_FFS,		LINUX_DEFAULT_SUPER_MAGIC	},
145 	{ MOUNT_NFS,		LINUX_NFS_SUPER_MAGIC 		},
146 	{ MOUNT_MFS,		LINUX_DEFAULT_SUPER_MAGIC	},
147 	{ MOUNT_MSDOS,		LINUX_MSDOS_SUPER_MAGIC		},
148 	{ MOUNT_LFS,		LINUX_DEFAULT_SUPER_MAGIC	},
149 	{ MOUNT_FDESC,		LINUX_DEFAULT_SUPER_MAGIC	},
150 	{ MOUNT_NULL,		LINUX_DEFAULT_SUPER_MAGIC	},
151 	{ MOUNT_OVERLAY,	LINUX_DEFAULT_SUPER_MAGIC	},
152 	{ MOUNT_UMAP,		LINUX_DEFAULT_SUPER_MAGIC	},
153 	{ MOUNT_KERNFS,		LINUX_DEFAULT_SUPER_MAGIC	},
154 	{ MOUNT_PROCFS,		LINUX_PROC_SUPER_MAGIC		},
155 	{ MOUNT_AFS,		LINUX_DEFAULT_SUPER_MAGIC	},
156 	{ MOUNT_CD9660,		LINUX_ISOFS_SUPER_MAGIC		},
157 	{ MOUNT_UNION,		LINUX_DEFAULT_SUPER_MAGIC	},
158 	{ MOUNT_ADOSFS,		LINUX_ADFS_SUPER_MAGIC		},
159 	{ MOUNT_EXT2FS,		LINUX_EXT2_SUPER_MAGIC		},
160 	{ MOUNT_CFS,		LINUX_DEFAULT_SUPER_MAGIC	},
161 	{ MOUNT_CODA,		LINUX_CODA_SUPER_MAGIC		},
162 	{ MOUNT_FILECORE,	LINUX_DEFAULT_SUPER_MAGIC	},
163 	{ MOUNT_NTFS,		LINUX_DEFAULT_SUPER_MAGIC	},
164 	{ MOUNT_SMBFS,		LINUX_SMB_SUPER_MAGIC		},
165 	{ MOUNT_PTYFS,		LINUX_DEVPTS_SUPER_MAGIC	},
166 	{ MOUNT_TMPFS,		LINUX_TMPFS_SUPER_MAGIC		}
167 };
168 const int linux_fstypes_cnt = sizeof(linux_fstypes) / sizeof(linux_fstypes[0]);
169 
170 # ifdef DEBUG_LINUX
171 #define DPRINTF(a)	uprintf a
172 # else
173 #define DPRINTF(a)
174 # endif
175 
176 /* Local linux_misc.c functions: */
177 static void linux_to_bsd_mmap_args(struct sys_mmap_args *,
178     const struct linux_sys_mmap_args *);
179 static int linux_mmap(struct lwp *, const struct linux_sys_mmap_args *,
180     register_t *, off_t);
181 
182 
183 /*
184  * The information on a terminated (or stopped) process needs
185  * to be converted in order for Linux binaries to get a valid signal
186  * number out of it.
187  */
188 int
189 bsd_to_linux_wstat(int st)
190 {
191 
192 	int sig;
193 
194 	if (WIFSIGNALED(st)) {
195 		sig = WTERMSIG(st);
196 		if (sig >= 0 && sig < NSIG)
197 			st= (st & ~0177) | native_to_linux_signo[sig];
198 	} else if (WIFSTOPPED(st)) {
199 		sig = WSTOPSIG(st);
200 		if (sig >= 0 && sig < NSIG)
201 			st = (st & ~0xff00) |
202 			    (native_to_linux_signo[sig] << 8);
203 	}
204 	return st;
205 }
206 
207 /*
208  * wait4(2).  Passed on to the NetBSD call, surrounded by code to
209  * reserve some space for a NetBSD-style wait status, and converting
210  * it to what Linux wants.
211  */
212 int
213 linux_sys_wait4(struct lwp *l, const struct linux_sys_wait4_args *uap, register_t *retval)
214 {
215 	/* {
216 		syscallarg(int) pid;
217 		syscallarg(int *) status;
218 		syscallarg(int) options;
219 		syscallarg(struct rusage50 *) rusage;
220 	} */
221 	int error, status, options, linux_options, pid = SCARG(uap, pid);
222 	struct rusage50 ru50;
223 	struct rusage ru;
224 	proc_t *p;
225 
226 	linux_options = SCARG(uap, options);
227 	options = WOPTSCHECKED;
228 	if (linux_options & ~(LINUX_WAIT4_KNOWNFLAGS))
229 		return (EINVAL);
230 
231 	if (linux_options & LINUX_WAIT4_WNOHANG)
232 		options |= WNOHANG;
233 	if (linux_options & LINUX_WAIT4_WUNTRACED)
234 		options |= WUNTRACED;
235 	if (linux_options & LINUX_WAIT4_WALL)
236 		options |= WALLSIG;
237 	if (linux_options & LINUX_WAIT4_WCLONE)
238 		options |= WALTSIG;
239 # ifdef DIAGNOSTIC
240 	if (linux_options & LINUX_WAIT4_WNOTHREAD)
241 		printf("WARNING: %s: linux process %d.%d called "
242 		       "waitpid with __WNOTHREAD set!",
243 		       __FILE__, l->l_proc->p_pid, l->l_lid);
244 
245 # endif
246 
247 	error = do_sys_wait(&pid, &status, options,
248 	    SCARG(uap, rusage) != NULL ? &ru : NULL);
249 
250 	retval[0] = pid;
251 	if (pid == 0)
252 		return error;
253 
254 	p = curproc;
255 	mutex_enter(p->p_lock);
256 	sigdelset(&p->p_sigpend.sp_set, SIGCHLD); /* XXXAD ksiginfo leak */
257 	mutex_exit(p->p_lock);
258 
259 	if (SCARG(uap, rusage) != NULL) {
260 		rusage_to_rusage50(&ru, &ru50);
261 		error = copyout(&ru, SCARG(uap, rusage), sizeof(ru));
262 	}
263 
264 	if (error == 0 && SCARG(uap, status) != NULL) {
265 		status = bsd_to_linux_wstat(status);
266 		error = copyout(&status, SCARG(uap, status), sizeof status);
267 	}
268 
269 	return error;
270 }
271 
272 /*
273  * Linux brk(2). The check if the new address is >= the old one is
274  * done in the kernel in Linux. NetBSD does it in the library.
275  */
276 int
277 linux_sys_brk(struct lwp *l, const struct linux_sys_brk_args *uap, register_t *retval)
278 {
279 	/* {
280 		syscallarg(char *) nsize;
281 	} */
282 	struct proc *p = l->l_proc;
283 	char *nbrk = SCARG(uap, nsize);
284 	struct sys_obreak_args oba;
285 	struct vmspace *vm = p->p_vmspace;
286 	struct linux_emuldata *ed = (struct linux_emuldata*)p->p_emuldata;
287 
288 	SCARG(&oba, nsize) = nbrk;
289 
290 	if ((void *) nbrk > vm->vm_daddr && sys_obreak(l, &oba, retval) == 0)
291 		ed->s->p_break = (char*)nbrk;
292 	else
293 		nbrk = ed->s->p_break;
294 
295 	retval[0] = (register_t)nbrk;
296 
297 	return 0;
298 }
299 
300 /*
301  * Implement the fs stat functions. Straightforward.
302  */
303 int
304 linux_sys_statfs(struct lwp *l, const struct linux_sys_statfs_args *uap, register_t *retval)
305 {
306 	/* {
307 		syscallarg(const char *) path;
308 		syscallarg(struct linux_statfs *) sp;
309 	} */
310 	struct statvfs *sb;
311 	struct linux_statfs ltmp;
312 	int error;
313 
314 	sb = STATVFSBUF_GET();
315 	error = do_sys_pstatvfs(l, SCARG(uap, path), ST_WAIT, sb);
316 	if (error == 0) {
317 		bsd_to_linux_statfs(sb, &ltmp);
318 		error = copyout(&ltmp, SCARG(uap, sp), sizeof ltmp);
319 	}
320 	STATVFSBUF_PUT(sb);
321 
322 	return error;
323 }
324 
325 int
326 linux_sys_fstatfs(struct lwp *l, const struct linux_sys_fstatfs_args *uap, register_t *retval)
327 {
328 	/* {
329 		syscallarg(int) fd;
330 		syscallarg(struct linux_statfs *) sp;
331 	} */
332 	struct statvfs *sb;
333 	struct linux_statfs ltmp;
334 	int error;
335 
336 	sb = STATVFSBUF_GET();
337 	error = do_sys_fstatvfs(l, SCARG(uap, fd), ST_WAIT, sb);
338 	if (error == 0) {
339 		bsd_to_linux_statfs(sb, &ltmp);
340 		error = copyout(&ltmp, SCARG(uap, sp), sizeof ltmp);
341 	}
342 	STATVFSBUF_PUT(sb);
343 
344 	return error;
345 }
346 
347 /*
348  * uname(). Just copy the info from the various strings stored in the
349  * kernel, and put it in the Linux utsname structure. That structure
350  * is almost the same as the NetBSD one, only it has fields 65 characters
351  * long, and an extra domainname field.
352  */
353 int
354 linux_sys_uname(struct lwp *l, const struct linux_sys_uname_args *uap, register_t *retval)
355 {
356 	/* {
357 		syscallarg(struct linux_utsname *) up;
358 	} */
359 	struct linux_utsname luts;
360 
361 	strlcpy(luts.l_sysname, linux_sysname, sizeof(luts.l_sysname));
362 	strlcpy(luts.l_nodename, hostname, sizeof(luts.l_nodename));
363 	strlcpy(luts.l_release, linux_release, sizeof(luts.l_release));
364 	strlcpy(luts.l_version, linux_version, sizeof(luts.l_version));
365 	strlcpy(luts.l_machine, LINUX_UNAME_ARCH, sizeof(luts.l_machine));
366 	strlcpy(luts.l_domainname, domainname, sizeof(luts.l_domainname));
367 
368 	return copyout(&luts, SCARG(uap, up), sizeof(luts));
369 }
370 
371 /* Used directly on: alpha, mips, ppc, sparc, sparc64 */
372 /* Used indirectly on: arm, i386, m68k */
373 
374 /*
375  * New type Linux mmap call.
376  * Only called directly on machines with >= 6 free regs.
377  */
378 int
379 linux_sys_mmap(struct lwp *l, const struct linux_sys_mmap_args *uap, register_t *retval)
380 {
381 	/* {
382 		syscallarg(unsigned long) addr;
383 		syscallarg(size_t) len;
384 		syscallarg(int) prot;
385 		syscallarg(int) flags;
386 		syscallarg(int) fd;
387 		syscallarg(linux_off_t) offset;
388 	} */
389 
390 	if (SCARG(uap, offset) & PAGE_MASK)
391 		return EINVAL;
392 
393 	return linux_mmap(l, uap, retval, SCARG(uap, offset));
394 }
395 
396 /*
397  * Guts of most architectures' mmap64() implementations.  This shares
398  * its list of arguments with linux_sys_mmap().
399  *
400  * The difference in linux_sys_mmap2() is that "offset" is actually
401  * (offset / pagesize), not an absolute byte count.  This translation
402  * to pagesize offsets is done inside glibc between the mmap64() call
403  * point, and the actual syscall.
404  */
405 int
406 linux_sys_mmap2(struct lwp *l, const struct linux_sys_mmap2_args *uap, register_t *retval)
407 {
408 	/* {
409 		syscallarg(unsigned long) addr;
410 		syscallarg(size_t) len;
411 		syscallarg(int) prot;
412 		syscallarg(int) flags;
413 		syscallarg(int) fd;
414 		syscallarg(linux_off_t) offset;
415 	} */
416 
417 	return linux_mmap(l, uap, retval,
418 	    ((off_t)SCARG(uap, offset)) << PAGE_SHIFT);
419 }
420 
421 /*
422  * Massage arguments and call system mmap(2).
423  */
424 static int
425 linux_mmap(struct lwp *l, const struct linux_sys_mmap_args *uap, register_t *retval, off_t offset)
426 {
427 	struct sys_mmap_args cma;
428 	int error;
429 	size_t mmoff=0;
430 
431 	linux_to_bsd_mmap_args(&cma, uap);
432 	SCARG(&cma, pos) = offset;
433 
434 	if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN) {
435 		/*
436 		 * Request for stack-like memory segment. On linux, this
437 		 * works by mmap()ping (small) segment, which is automatically
438 		 * extended when page fault happens below the currently
439 		 * allocated area. We emulate this by allocating (typically
440 		 * bigger) segment sized at current stack size limit, and
441 		 * offsetting the requested and returned address accordingly.
442 		 * Since physical pages are only allocated on-demand, this
443 		 * is effectively identical.
444 		 */
445 		rlim_t ssl = l->l_proc->p_rlimit[RLIMIT_STACK].rlim_cur;
446 
447 		if (SCARG(&cma, len) < ssl) {
448 			/* Compute the address offset */
449 			mmoff = round_page(ssl) - SCARG(uap, len);
450 
451 			if (SCARG(&cma, addr))
452 				SCARG(&cma, addr) = (char *)SCARG(&cma, addr) - mmoff;
453 
454 			SCARG(&cma, len) = (size_t) ssl;
455 		}
456 	}
457 
458 	error = sys_mmap(l, &cma, retval);
459 	if (error)
460 		return (error);
461 
462 	/* Shift the returned address for stack-like segment if necessary */
463 	retval[0] += mmoff;
464 
465 	return (0);
466 }
467 
468 static void
469 linux_to_bsd_mmap_args(struct sys_mmap_args *cma, const struct linux_sys_mmap_args *uap)
470 {
471 	int flags = MAP_TRYFIXED, fl = SCARG(uap, flags);
472 
473 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_SHARED, MAP_SHARED);
474 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_PRIVATE, MAP_PRIVATE);
475 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_FIXED, MAP_FIXED);
476 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_ANON, MAP_ANON);
477 	/* XXX XAX ERH: Any other flags here?  There are more defined... */
478 
479 	SCARG(cma, addr) = (void *)SCARG(uap, addr);
480 	SCARG(cma, len) = SCARG(uap, len);
481 	SCARG(cma, prot) = SCARG(uap, prot);
482 	if (SCARG(cma, prot) & VM_PROT_WRITE) /* XXX */
483 		SCARG(cma, prot) |= VM_PROT_READ;
484 	SCARG(cma, flags) = flags;
485 	SCARG(cma, fd) = flags & MAP_ANON ? -1 : SCARG(uap, fd);
486 	SCARG(cma, PAD) = 0;
487 }
488 
489 #define	LINUX_MREMAP_MAYMOVE	1
490 #define	LINUX_MREMAP_FIXED	2
491 
492 int
493 linux_sys_mremap(struct lwp *l, const struct linux_sys_mremap_args *uap, register_t *retval)
494 {
495 	/* {
496 		syscallarg(void *) old_address;
497 		syscallarg(size_t) old_size;
498 		syscallarg(size_t) new_size;
499 		syscallarg(u_long) flags;
500 	} */
501 
502 	struct proc *p;
503 	struct vm_map *map;
504 	vaddr_t oldva;
505 	vaddr_t newva;
506 	size_t oldsize;
507 	size_t newsize;
508 	int flags;
509 	int uvmflags;
510 	int error;
511 
512 	flags = SCARG(uap, flags);
513 	oldva = (vaddr_t)SCARG(uap, old_address);
514 	oldsize = round_page(SCARG(uap, old_size));
515 	newsize = round_page(SCARG(uap, new_size));
516 	if ((flags & ~(LINUX_MREMAP_FIXED|LINUX_MREMAP_MAYMOVE)) != 0) {
517 		error = EINVAL;
518 		goto done;
519 	}
520 	if ((flags & LINUX_MREMAP_FIXED) != 0) {
521 		if ((flags & LINUX_MREMAP_MAYMOVE) == 0) {
522 			error = EINVAL;
523 			goto done;
524 		}
525 #if 0 /* notyet */
526 		newva = SCARG(uap, new_address);
527 		uvmflags = MAP_FIXED;
528 #else /* notyet */
529 		error = EOPNOTSUPP;
530 		goto done;
531 #endif /* notyet */
532 	} else if ((flags & LINUX_MREMAP_MAYMOVE) != 0) {
533 		uvmflags = 0;
534 	} else {
535 		newva = oldva;
536 		uvmflags = MAP_FIXED;
537 	}
538 	p = l->l_proc;
539 	map = &p->p_vmspace->vm_map;
540 	error = uvm_mremap(map, oldva, oldsize, map, &newva, newsize, p,
541 	    uvmflags);
542 
543 done:
544 	*retval = (error != 0) ? 0 : (register_t)newva;
545 	return error;
546 }
547 
548 int
549 linux_sys_mprotect(struct lwp *l, const struct linux_sys_mprotect_args *uap, register_t *retval)
550 {
551 	/* {
552 		syscallarg(const void *) start;
553 		syscallarg(unsigned long) len;
554 		syscallarg(int) prot;
555 	} */
556 	struct vm_map_entry *entry;
557 	struct vm_map *map;
558 	struct proc *p;
559 	vaddr_t end, start, len, stacklim;
560 	int prot, grows;
561 
562 	start = (vaddr_t)SCARG(uap, start);
563 	len = round_page(SCARG(uap, len));
564 	prot = SCARG(uap, prot);
565 	grows = prot & (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP);
566 	prot &= ~grows;
567 	end = start + len;
568 
569 	if (start & PAGE_MASK)
570 		return EINVAL;
571 	if (end < start)
572 		return EINVAL;
573 	if (end == start)
574 		return 0;
575 
576 	if (prot & ~(PROT_READ | PROT_WRITE | PROT_EXEC))
577 		return EINVAL;
578 	if (grows == (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP))
579 		return EINVAL;
580 
581 	p = l->l_proc;
582 	map = &p->p_vmspace->vm_map;
583 	vm_map_lock(map);
584 # ifdef notdef
585 	VM_MAP_RANGE_CHECK(map, start, end);
586 # endif
587 	if (!uvm_map_lookup_entry(map, start, &entry) || entry->start > start) {
588 		vm_map_unlock(map);
589 		return ENOMEM;
590 	}
591 
592 	/*
593 	 * Approximate the behaviour of PROT_GROWS{DOWN,UP}.
594 	 */
595 
596 	stacklim = (vaddr_t)p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur;
597 	if (grows & LINUX_PROT_GROWSDOWN) {
598 		if (USRSTACK - stacklim <= start && start < USRSTACK) {
599 			start = USRSTACK - stacklim;
600 		} else {
601 			start = entry->start;
602 		}
603 	} else if (grows & LINUX_PROT_GROWSUP) {
604 		if (USRSTACK <= end && end < USRSTACK + stacklim) {
605 			end = USRSTACK + stacklim;
606 		} else {
607 			end = entry->end;
608 		}
609 	}
610 	vm_map_unlock(map);
611 	return uvm_map_protect(map, start, end, prot, FALSE);
612 }
613 
614 /*
615  * This code is partly stolen from src/lib/libc/compat-43/times.c
616  */
617 
618 #define	CONVTCK(r)	(r.tv_sec * hz + r.tv_usec / (1000000 / hz))
619 
620 int
621 linux_sys_times(struct lwp *l, const struct linux_sys_times_args *uap, register_t *retval)
622 {
623 	/* {
624 		syscallarg(struct times *) tms;
625 	} */
626 	struct proc *p = l->l_proc;
627 	struct timeval t;
628 	int error;
629 
630 	if (SCARG(uap, tms)) {
631 		struct linux_tms ltms;
632 		struct rusage ru;
633 
634 		mutex_enter(p->p_lock);
635 		calcru(p, &ru.ru_utime, &ru.ru_stime, NULL, NULL);
636 		ltms.ltms_utime = CONVTCK(ru.ru_utime);
637 		ltms.ltms_stime = CONVTCK(ru.ru_stime);
638 		ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime);
639 		ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime);
640 		mutex_exit(p->p_lock);
641 
642 		if ((error = copyout(&ltms, SCARG(uap, tms), sizeof ltms)))
643 			return error;
644 	}
645 
646 	getmicrouptime(&t);
647 
648 	retval[0] = ((linux_clock_t)(CONVTCK(t)));
649 	return 0;
650 }
651 
652 #undef CONVTCK
653 
654 /*
655  * Linux 'readdir' call. This code is mostly taken from the
656  * SunOS getdents call (see compat/sunos/sunos_misc.c), though
657  * an attempt has been made to keep it a little cleaner (failing
658  * miserably, because of the cruft needed if count 1 is passed).
659  *
660  * The d_off field should contain the offset of the next valid entry,
661  * but in Linux it has the offset of the entry itself. We emulate
662  * that bug here.
663  *
664  * Read in BSD-style entries, convert them, and copy them out.
665  *
666  * Note that this doesn't handle union-mounted filesystems.
667  */
668 int
669 linux_sys_getdents(struct lwp *l, const struct linux_sys_getdents_args *uap, register_t *retval)
670 {
671 	/* {
672 		syscallarg(int) fd;
673 		syscallarg(struct linux_dirent *) dent;
674 		syscallarg(unsigned int) count;
675 	} */
676 	struct dirent *bdp;
677 	struct vnode *vp;
678 	char *inp, *tbuf;		/* BSD-format */
679 	int len, reclen;		/* BSD-format */
680 	char *outp;			/* Linux-format */
681 	int resid, linux_reclen = 0;	/* Linux-format */
682 	struct file *fp;
683 	struct uio auio;
684 	struct iovec aiov;
685 	struct linux_dirent idb;
686 	off_t off;		/* true file offset */
687 	int buflen, error, eofflag, nbytes, oldcall;
688 	struct vattr va;
689 	off_t *cookiebuf = NULL, *cookie;
690 	int ncookies;
691 
692 	/* fd_getvnode() will use the descriptor for us */
693 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
694 		return (error);
695 
696 	if ((fp->f_flag & FREAD) == 0) {
697 		error = EBADF;
698 		goto out1;
699 	}
700 
701 	vp = (struct vnode *)fp->f_data;
702 	if (vp->v_type != VDIR) {
703 		error = ENOTDIR;
704 		goto out1;
705 	}
706 
707 	if ((error = VOP_GETATTR(vp, &va, l->l_cred)))
708 		goto out1;
709 
710 	nbytes = SCARG(uap, count);
711 	if (nbytes == 1) {	/* emulating old, broken behaviour */
712 		nbytes = sizeof (idb);
713 		buflen = max(va.va_blocksize, nbytes);
714 		oldcall = 1;
715 	} else {
716 		buflen = min(MAXBSIZE, nbytes);
717 		if (buflen < va.va_blocksize)
718 			buflen = va.va_blocksize;
719 		oldcall = 0;
720 	}
721 	tbuf = malloc(buflen, M_TEMP, M_WAITOK);
722 
723 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
724 	off = fp->f_offset;
725 again:
726 	aiov.iov_base = tbuf;
727 	aiov.iov_len = buflen;
728 	auio.uio_iov = &aiov;
729 	auio.uio_iovcnt = 1;
730 	auio.uio_rw = UIO_READ;
731 	auio.uio_resid = buflen;
732 	auio.uio_offset = off;
733 	UIO_SETUP_SYSSPACE(&auio);
734 	/*
735          * First we read into the malloc'ed buffer, then
736          * we massage it into user space, one record at a time.
737          */
738 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf,
739 	    &ncookies);
740 	if (error)
741 		goto out;
742 
743 	inp = tbuf;
744 	outp = (void *)SCARG(uap, dent);
745 	resid = nbytes;
746 	if ((len = buflen - auio.uio_resid) == 0)
747 		goto eof;
748 
749 	for (cookie = cookiebuf; len > 0; len -= reclen) {
750 		bdp = (struct dirent *)inp;
751 		reclen = bdp->d_reclen;
752 		if (reclen & 3)
753 			panic("linux_readdir");
754 		if (bdp->d_fileno == 0) {
755 			inp += reclen;	/* it is a hole; squish it out */
756 			if (cookie)
757 				off = *cookie++;
758 			else
759 				off += reclen;
760 			continue;
761 		}
762 		linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen);
763 		if (reclen > len || resid < linux_reclen) {
764 			/* entry too big for buffer, so just stop */
765 			outp++;
766 			break;
767 		}
768 		/*
769 		 * Massage in place to make a Linux-shaped dirent (otherwise
770 		 * we have to worry about touching user memory outside of
771 		 * the copyout() call).
772 		 */
773 		idb.d_ino = bdp->d_fileno;
774 		/*
775 		 * The old readdir() call misuses the offset and reclen fields.
776 		 */
777 		if (oldcall) {
778 			idb.d_off = (linux_off_t)linux_reclen;
779 			idb.d_reclen = (u_short)bdp->d_namlen;
780 		} else {
781 			if (sizeof (idb.d_off) <= 4 && (off >> 32) != 0) {
782 				compat_offseterr(vp, "linux_getdents");
783 				error = EINVAL;
784 				goto out;
785 			}
786 			idb.d_off = (linux_off_t)off;
787 			idb.d_reclen = (u_short)linux_reclen;
788 		}
789 		strcpy(idb.d_name, bdp->d_name);
790 		if ((error = copyout((void *)&idb, outp, linux_reclen)))
791 			goto out;
792 		/* advance past this real entry */
793 		inp += reclen;
794 		if (cookie)
795 			off = *cookie++; /* each entry points to itself */
796 		else
797 			off += reclen;
798 		/* advance output past Linux-shaped entry */
799 		outp += linux_reclen;
800 		resid -= linux_reclen;
801 		if (oldcall)
802 			break;
803 	}
804 
805 	/* if we squished out the whole block, try again */
806 	if (outp == (void *)SCARG(uap, dent))
807 		goto again;
808 	fp->f_offset = off;	/* update the vnode offset */
809 
810 	if (oldcall)
811 		nbytes = resid + linux_reclen;
812 
813 eof:
814 	*retval = nbytes - resid;
815 out:
816 	VOP_UNLOCK(vp, 0);
817 	if (cookiebuf)
818 		free(cookiebuf, M_TEMP);
819 	free(tbuf, M_TEMP);
820 out1:
821 	fd_putfile(SCARG(uap, fd));
822 	return error;
823 }
824 
825 /*
826  * Even when just using registers to pass arguments to syscalls you can
827  * have 5 of them on the i386. So this newer version of select() does
828  * this.
829  */
830 int
831 linux_sys_select(struct lwp *l, const struct linux_sys_select_args *uap, register_t *retval)
832 {
833 	/* {
834 		syscallarg(int) nfds;
835 		syscallarg(fd_set *) readfds;
836 		syscallarg(fd_set *) writefds;
837 		syscallarg(fd_set *) exceptfds;
838 		syscallarg(struct timeval50 *) timeout;
839 	} */
840 
841 	return linux_select1(l, retval, SCARG(uap, nfds), SCARG(uap, readfds),
842 	    SCARG(uap, writefds), SCARG(uap, exceptfds),
843 	    (struct linux_timeval *)SCARG(uap, timeout));
844 }
845 
846 /*
847  * Common code for the old and new versions of select(). A couple of
848  * things are important:
849  * 1) return the amount of time left in the 'timeout' parameter
850  * 2) select never returns ERESTART on Linux, always return EINTR
851  */
852 int
853 linux_select1(struct lwp *l, register_t *retval, int nfds, fd_set *readfds,
854     fd_set *writefds, fd_set *exceptfds, struct linux_timeval *timeout)
855 {
856 	struct timespec ts0, ts1, uts, *ts = NULL;
857 	struct linux_timeval ltv;
858 	int error;
859 
860 	/*
861 	 * Store current time for computation of the amount of
862 	 * time left.
863 	 */
864 	if (timeout) {
865 		if ((error = copyin(timeout, &ltv, sizeof(ltv))))
866 			return error;
867 		uts.tv_sec = ltv.tv_sec;
868 		uts.tv_nsec = ltv.tv_usec * 1000;
869 		if (itimespecfix(&uts)) {
870 			/*
871 			 * The timeval was invalid.  Convert it to something
872 			 * valid that will act as it does under Linux.
873 			 */
874 			uts.tv_sec += uts.tv_nsec / 1000000000;
875 			uts.tv_nsec %= 1000000000;
876 			if (uts.tv_nsec < 0) {
877 				uts.tv_sec -= 1;
878 				uts.tv_nsec += 1000000000;
879 			}
880 			if (uts.tv_sec < 0)
881 				timespecclear(&uts);
882 		}
883 		ts = &uts;
884 		nanotime(&ts0);
885 	}
886 
887 	error = selcommon(retval, nfds, readfds, writefds, exceptfds, ts, NULL);
888 
889 	if (error) {
890 		/*
891 		 * See fs/select.c in the Linux kernel.  Without this,
892 		 * Maelstrom doesn't work.
893 		 */
894 		if (error == ERESTART)
895 			error = EINTR;
896 		return error;
897 	}
898 
899 	if (timeout) {
900 		if (*retval) {
901 			/*
902 			 * Compute how much time was left of the timeout,
903 			 * by subtracting the current time and the time
904 			 * before we started the call, and subtracting
905 			 * that result from the user-supplied value.
906 			 */
907 			nanotime(&ts1);
908 			timespecsub(&ts1, &ts0, &ts1);
909 			timespecsub(&uts, &ts1, &uts);
910 			if (uts.tv_sec < 0)
911 				timespecclear(&uts);
912 		} else
913 			timespecclear(&uts);
914 		ltv.tv_sec = uts.tv_sec;
915 		ltv.tv_usec = uts.tv_nsec / 1000;
916 		if ((error = copyout(&ltv, timeout, sizeof(ltv))))
917 			return error;
918 	}
919 
920 	return 0;
921 }
922 
923 /*
924  * Set the 'personality' (emulation mode) for the current process. Only
925  * accept the Linux personality here (0). This call is needed because
926  * the Linux ELF crt0 issues it in an ugly kludge to make sure that
927  * ELF binaries run in Linux mode, not SVR4 mode.
928  */
929 int
930 linux_sys_personality(struct lwp *l, const struct linux_sys_personality_args *uap, register_t *retval)
931 {
932 	/* {
933 		syscallarg(unsigned long) per;
934 	} */
935 
936 	switch (SCARG(uap, per)) {
937 	case LINUX_PER_LINUX:
938 	case LINUX_PER_QUERY:
939 		break;
940 	default:
941 		return EINVAL;
942 	}
943 
944 	retval[0] = LINUX_PER_LINUX;
945 	return 0;
946 }
947 
948 /*
949  * We have nonexistent fsuid equal to uid.
950  * If modification is requested, refuse.
951  */
952 int
953 linux_sys_setfsuid(struct lwp *l, const struct linux_sys_setfsuid_args *uap, register_t *retval)
954 {
955 	 /* {
956 		 syscallarg(uid_t) uid;
957 	 } */
958 	 uid_t uid;
959 
960 	 uid = SCARG(uap, uid);
961 	 if (kauth_cred_getuid(l->l_cred) != uid)
962 		 return sys_nosys(l, uap, retval);
963 
964 	 *retval = uid;
965 	 return 0;
966 }
967 
968 int
969 linux_sys_setfsgid(struct lwp *l, const struct linux_sys_setfsgid_args *uap, register_t *retval)
970 {
971 	/* {
972 		syscallarg(gid_t) gid;
973 	} */
974 	gid_t gid;
975 
976 	gid = SCARG(uap, gid);
977 	if (kauth_cred_getgid(l->l_cred) != gid)
978 		return sys_nosys(l, uap, retval);
979 
980 	*retval = gid;
981 	return 0;
982 }
983 
984 int
985 linux_sys_setresuid(struct lwp *l, const struct linux_sys_setresuid_args *uap, register_t *retval)
986 {
987 	/* {
988 		syscallarg(uid_t) ruid;
989 		syscallarg(uid_t) euid;
990 		syscallarg(uid_t) suid;
991 	} */
992 
993 	/*
994 	 * Note: These checks are a little different than the NetBSD
995 	 * setreuid(2) call performs.  This precisely follows the
996 	 * behavior of the Linux kernel.
997 	 */
998 
999 	return do_setresuid(l, SCARG(uap, ruid), SCARG(uap, euid),
1000 			    SCARG(uap, suid),
1001 			    ID_R_EQ_R | ID_R_EQ_E | ID_R_EQ_S |
1002 			    ID_E_EQ_R | ID_E_EQ_E | ID_E_EQ_S |
1003 			    ID_S_EQ_R | ID_S_EQ_E | ID_S_EQ_S );
1004 }
1005 
1006 int
1007 linux_sys_getresuid(struct lwp *l, const struct linux_sys_getresuid_args *uap, register_t *retval)
1008 {
1009 	/* {
1010 		syscallarg(uid_t *) ruid;
1011 		syscallarg(uid_t *) euid;
1012 		syscallarg(uid_t *) suid;
1013 	} */
1014 	kauth_cred_t pc = l->l_cred;
1015 	int error;
1016 	uid_t uid;
1017 
1018 	/*
1019 	 * Linux copies these values out to userspace like so:
1020 	 *
1021 	 *	1. Copy out ruid.
1022 	 *	2. If that succeeds, copy out euid.
1023 	 *	3. If both of those succeed, copy out suid.
1024 	 */
1025 	uid = kauth_cred_getuid(pc);
1026 	if ((error = copyout(&uid, SCARG(uap, ruid), sizeof(uid_t))) != 0)
1027 		return (error);
1028 
1029 	uid = kauth_cred_geteuid(pc);
1030 	if ((error = copyout(&uid, SCARG(uap, euid), sizeof(uid_t))) != 0)
1031 		return (error);
1032 
1033 	uid = kauth_cred_getsvuid(pc);
1034 
1035 	return (copyout(&uid, SCARG(uap, suid), sizeof(uid_t)));
1036 }
1037 
1038 int
1039 linux_sys_ptrace(struct lwp *l, const struct linux_sys_ptrace_args *uap, register_t *retval)
1040 {
1041 	/* {
1042 		i386, m68k, powerpc: T=int
1043 		alpha, amd64: T=long
1044 		syscallarg(T) request;
1045 		syscallarg(T) pid;
1046 		syscallarg(T) addr;
1047 		syscallarg(T) data;
1048 	} */
1049 	const int *ptr;
1050 	int request;
1051 	int error;
1052 
1053 	ptr = linux_ptrace_request_map;
1054 	request = SCARG(uap, request);
1055 	while (*ptr != -1)
1056 		if (*ptr++ == request) {
1057 			struct sys_ptrace_args pta;
1058 
1059 			SCARG(&pta, req) = *ptr;
1060 			SCARG(&pta, pid) = SCARG(uap, pid);
1061 			SCARG(&pta, addr) = (void *)SCARG(uap, addr);
1062 			SCARG(&pta, data) = SCARG(uap, data);
1063 
1064 			/*
1065 			 * Linux ptrace(PTRACE_CONT, pid, 0, 0) means actually
1066 			 * to continue where the process left off previously.
1067  			 * The same thing is achieved by addr == (void *) 1
1068 			 * on NetBSD, so rewrite 'addr' appropriately.
1069 			 */
1070 			if (request == LINUX_PTRACE_CONT && SCARG(uap, addr)==0)
1071 				SCARG(&pta, addr) = (void *) 1;
1072 
1073 			error = sysent[SYS_ptrace].sy_call(l, &pta, retval);
1074 			if (error)
1075 				return error;
1076 			switch (request) {
1077 			case LINUX_PTRACE_PEEKTEXT:
1078 			case LINUX_PTRACE_PEEKDATA:
1079 				error = copyout (retval,
1080 				    (void *)SCARG(uap, data),
1081 				    sizeof *retval);
1082 				*retval = SCARG(uap, data);
1083 				break;
1084 			default:
1085 				break;
1086 			}
1087 			return error;
1088 		}
1089 		else
1090 			ptr++;
1091 
1092 	return LINUX_SYS_PTRACE_ARCH(l, uap, retval);
1093 }
1094 
1095 int
1096 linux_sys_reboot(struct lwp *l, const struct linux_sys_reboot_args *uap, register_t *retval)
1097 {
1098 	/* {
1099 		syscallarg(int) magic1;
1100 		syscallarg(int) magic2;
1101 		syscallarg(int) cmd;
1102 		syscallarg(void *) arg;
1103 	} */
1104 	struct sys_reboot_args /* {
1105 		syscallarg(int) opt;
1106 		syscallarg(char *) bootstr;
1107 	} */ sra;
1108 	int error;
1109 
1110 	if ((error = kauth_authorize_system(l->l_cred,
1111 	    KAUTH_SYSTEM_REBOOT, 0, NULL, NULL, NULL)) != 0)
1112 		return(error);
1113 
1114 	if (SCARG(uap, magic1) != LINUX_REBOOT_MAGIC1)
1115 		return(EINVAL);
1116 	if (SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2 &&
1117 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2A &&
1118 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2B)
1119 		return(EINVAL);
1120 
1121 	switch ((unsigned long)SCARG(uap, cmd)) {
1122 	case LINUX_REBOOT_CMD_RESTART:
1123 		SCARG(&sra, opt) = RB_AUTOBOOT;
1124 		break;
1125 	case LINUX_REBOOT_CMD_HALT:
1126 		SCARG(&sra, opt) = RB_HALT;
1127 		break;
1128 	case LINUX_REBOOT_CMD_POWER_OFF:
1129 		SCARG(&sra, opt) = RB_HALT|RB_POWERDOWN;
1130 		break;
1131 	case LINUX_REBOOT_CMD_RESTART2:
1132 		/* Reboot with an argument. */
1133 		SCARG(&sra, opt) = RB_AUTOBOOT|RB_STRING;
1134 		SCARG(&sra, bootstr) = SCARG(uap, arg);
1135 		break;
1136 	case LINUX_REBOOT_CMD_CAD_ON:
1137 		return(EINVAL);	/* We don't implement ctrl-alt-delete */
1138 	case LINUX_REBOOT_CMD_CAD_OFF:
1139 		return(0);
1140 	default:
1141 		return(EINVAL);
1142 	}
1143 
1144 	return(sys_reboot(l, &sra, retval));
1145 }
1146 
1147 /*
1148  * Copy of compat_12_sys_swapon().
1149  */
1150 int
1151 linux_sys_swapon(struct lwp *l, const struct linux_sys_swapon_args *uap, register_t *retval)
1152 {
1153 	/* {
1154 		syscallarg(const char *) name;
1155 	} */
1156 	struct sys_swapctl_args ua;
1157 
1158 	SCARG(&ua, cmd) = SWAP_ON;
1159 	SCARG(&ua, arg) = (void *)__UNCONST(SCARG(uap, name));
1160 	SCARG(&ua, misc) = 0;	/* priority */
1161 	return (sys_swapctl(l, &ua, retval));
1162 }
1163 
1164 /*
1165  * Stop swapping to the file or block device specified by path.
1166  */
1167 int
1168 linux_sys_swapoff(struct lwp *l, const struct linux_sys_swapoff_args *uap, register_t *retval)
1169 {
1170 	/* {
1171 		syscallarg(const char *) path;
1172 	} */
1173 	struct sys_swapctl_args ua;
1174 
1175 	SCARG(&ua, cmd) = SWAP_OFF;
1176 	SCARG(&ua, arg) = __UNCONST(SCARG(uap, path)); /*XXXUNCONST*/
1177 	return (sys_swapctl(l, &ua, retval));
1178 }
1179 
1180 /*
1181  * Copy of compat_09_sys_setdomainname()
1182  */
1183 /* ARGSUSED */
1184 int
1185 linux_sys_setdomainname(struct lwp *l, const struct linux_sys_setdomainname_args *uap, register_t *retval)
1186 {
1187 	/* {
1188 		syscallarg(char *) domainname;
1189 		syscallarg(int) len;
1190 	} */
1191 	int name[2];
1192 
1193 	name[0] = CTL_KERN;
1194 	name[1] = KERN_DOMAINNAME;
1195 	return (old_sysctl(&name[0], 2, 0, 0, SCARG(uap, domainname),
1196 			    SCARG(uap, len), l));
1197 }
1198 
1199 /*
1200  * sysinfo()
1201  */
1202 /* ARGSUSED */
1203 int
1204 linux_sys_sysinfo(struct lwp *l, const struct linux_sys_sysinfo_args *uap, register_t *retval)
1205 {
1206 	/* {
1207 		syscallarg(struct linux_sysinfo *) arg;
1208 	} */
1209 	struct linux_sysinfo si;
1210 	struct loadavg *la;
1211 
1212 	si.uptime = time_uptime;
1213 	la = &averunnable;
1214 	si.loads[0] = la->ldavg[0] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1215 	si.loads[1] = la->ldavg[1] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1216 	si.loads[2] = la->ldavg[2] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1217 	si.totalram = ctob((u_long)physmem);
1218 	si.freeram = (u_long)uvmexp.free * uvmexp.pagesize;
1219 	si.sharedram = 0;	/* XXX */
1220 	si.bufferram = (u_long)uvmexp.filepages * uvmexp.pagesize;
1221 	si.totalswap = (u_long)uvmexp.swpages * uvmexp.pagesize;
1222 	si.freeswap =
1223 	    (u_long)(uvmexp.swpages - uvmexp.swpginuse) * uvmexp.pagesize;
1224 	si.procs = nprocs;
1225 
1226 	/* The following are only present in newer Linux kernels. */
1227 	si.totalbig = 0;
1228 	si.freebig = 0;
1229 	si.mem_unit = 1;
1230 
1231 	return (copyout(&si, SCARG(uap, arg), sizeof si));
1232 }
1233 
1234 int
1235 linux_sys_getrlimit(struct lwp *l, const struct linux_sys_getrlimit_args *uap, register_t *retval)
1236 {
1237 	/* {
1238 		syscallarg(int) which;
1239 # ifdef LINUX_LARGEFILE64
1240 		syscallarg(struct rlimit *) rlp;
1241 # else
1242 		syscallarg(struct orlimit *) rlp;
1243 # endif
1244 	} */
1245 # ifdef LINUX_LARGEFILE64
1246 	struct rlimit orl;
1247 # else
1248 	struct orlimit orl;
1249 # endif
1250 	int which;
1251 
1252 	which = linux_to_bsd_limit(SCARG(uap, which));
1253 	if (which < 0)
1254 		return -which;
1255 
1256 	bsd_to_linux_rlimit(&orl, &l->l_proc->p_rlimit[which]);
1257 
1258 	return copyout(&orl, SCARG(uap, rlp), sizeof(orl));
1259 }
1260 
1261 int
1262 linux_sys_setrlimit(struct lwp *l, const struct linux_sys_setrlimit_args *uap, register_t *retval)
1263 {
1264 	/* {
1265 		syscallarg(int) which;
1266 # ifdef LINUX_LARGEFILE64
1267 		syscallarg(struct rlimit *) rlp;
1268 # else
1269 		syscallarg(struct orlimit *) rlp;
1270 # endif
1271 	} */
1272 	struct rlimit rl;
1273 # ifdef LINUX_LARGEFILE64
1274 	struct rlimit orl;
1275 # else
1276 	struct orlimit orl;
1277 # endif
1278 	int error;
1279 	int which;
1280 
1281 	if ((error = copyin(SCARG(uap, rlp), &orl, sizeof(orl))) != 0)
1282 		return error;
1283 
1284 	which = linux_to_bsd_limit(SCARG(uap, which));
1285 	if (which < 0)
1286 		return -which;
1287 
1288 	linux_to_bsd_rlimit(&rl, &orl);
1289 	return dosetrlimit(l, l->l_proc, which, &rl);
1290 }
1291 
1292 # if !defined(__mips__) && !defined(__amd64__)
1293 /* XXX: this doesn't look 100% common, at least mips doesn't have it */
1294 int
1295 linux_sys_ugetrlimit(struct lwp *l, const struct linux_sys_ugetrlimit_args *uap, register_t *retval)
1296 {
1297 	return linux_sys_getrlimit(l, (const void *)uap, retval);
1298 }
1299 # endif
1300 
1301 /*
1302  * This gets called for unsupported syscalls. The difference to sys_nosys()
1303  * is that process does not get SIGSYS, the call just returns with ENOSYS.
1304  * This is the way Linux does it and glibc depends on this behaviour.
1305  */
1306 int
1307 linux_sys_nosys(struct lwp *l, const void *v, register_t *retval)
1308 {
1309 	return (ENOSYS);
1310 }
1311 
1312 int
1313 linux_sys_getpriority(struct lwp *l, const struct linux_sys_getpriority_args *uap, register_t *retval)
1314 {
1315         /* {
1316                 syscallarg(int) which;
1317                 syscallarg(int) who;
1318         } */
1319         struct sys_getpriority_args bsa;
1320         int error;
1321 
1322         SCARG(&bsa, which) = SCARG(uap, which);
1323         SCARG(&bsa, who) = SCARG(uap, who);
1324 
1325         if ((error = sys_getpriority(l, &bsa, retval)))
1326                 return error;
1327 
1328         *retval = NZERO - *retval;
1329 
1330         return 0;
1331 }
1332 
1333 #endif /* !COMPAT_LINUX32 */
1334