xref: /netbsd-src/sys/compat/linux/common/linux_misc.c (revision 0dd5877adce57db949b16ae963e5a6831cccdfb6)
1 /*	$NetBSD: linux_misc.c,v 1.101 2002/02/20 17:03:03 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 1.101 2002/02/20 17:03:03 christos Exp $");
68 
69 #include <sys/param.h>
70 #include <sys/systm.h>
71 #include <sys/namei.h>
72 #include <sys/proc.h>
73 #include <sys/dirent.h>
74 #include <sys/file.h>
75 #include <sys/stat.h>
76 #include <sys/filedesc.h>
77 #include <sys/ioctl.h>
78 #include <sys/kernel.h>
79 #include <sys/malloc.h>
80 #include <sys/mbuf.h>
81 #include <sys/mman.h>
82 #include <sys/mount.h>
83 #include <sys/reboot.h>
84 #include <sys/resource.h>
85 #include <sys/resourcevar.h>
86 #include <sys/signal.h>
87 #include <sys/signalvar.h>
88 #include <sys/socket.h>
89 #include <sys/time.h>
90 #include <sys/times.h>
91 #include <sys/vnode.h>
92 #include <sys/uio.h>
93 #include <sys/wait.h>
94 #include <sys/utsname.h>
95 #include <sys/unistd.h>
96 #include <sys/swap.h>		/* for SWAP_ON */
97 #include <sys/sysctl.h>		/* for KERN_DOMAINNAME */
98 
99 #include <sys/ptrace.h>
100 #include <machine/ptrace.h>
101 
102 #include <sys/syscallargs.h>
103 
104 #include <compat/linux/common/linux_types.h>
105 #include <compat/linux/common/linux_signal.h>
106 
107 #include <compat/linux/linux_syscallargs.h>
108 
109 #include <compat/linux/common/linux_fcntl.h>
110 #include <compat/linux/common/linux_mmap.h>
111 #include <compat/linux/common/linux_dirent.h>
112 #include <compat/linux/common/linux_util.h>
113 #include <compat/linux/common/linux_misc.h>
114 #include <compat/linux/common/linux_ptrace.h>
115 #include <compat/linux/common/linux_reboot.h>
116 #include <compat/linux/common/linux_emuldata.h>
117 
118 const int linux_ptrace_request_map[] = {
119 	LINUX_PTRACE_TRACEME,	PT_TRACE_ME,
120 	LINUX_PTRACE_PEEKTEXT,	PT_READ_I,
121 	LINUX_PTRACE_PEEKDATA,	PT_READ_D,
122 	LINUX_PTRACE_POKETEXT,	PT_WRITE_I,
123 	LINUX_PTRACE_POKEDATA,	PT_WRITE_D,
124 	LINUX_PTRACE_CONT,	PT_CONTINUE,
125 	LINUX_PTRACE_KILL,	PT_KILL,
126 	LINUX_PTRACE_ATTACH,	PT_ATTACH,
127 	LINUX_PTRACE_DETACH,	PT_DETACH,
128 #ifdef PT_STEP
129 	LINUX_PTRACE_SINGLESTEP,	PT_STEP,
130 #endif
131 	-1
132 };
133 
134 const static struct mnttypes {
135 	char *bsd;
136 	int linux;
137 } fstypes[] = {
138 	{ MOUNT_FFS,		LINUX_DEFAULT_SUPER_MAGIC	},
139 	{ MOUNT_NFS,		LINUX_NFS_SUPER_MAGIC 		},
140 	{ MOUNT_MFS,		LINUX_DEFAULT_SUPER_MAGIC	},
141 	{ MOUNT_MSDOS,		LINUX_MSDOS_SUPER_MAGIC		},
142 	{ MOUNT_LFS,		LINUX_DEFAULT_SUPER_MAGIC	},
143 	{ MOUNT_FDESC,		LINUX_DEFAULT_SUPER_MAGIC	},
144 	{ MOUNT_PORTAL,		LINUX_DEFAULT_SUPER_MAGIC	},
145 	{ MOUNT_NULL,		LINUX_DEFAULT_SUPER_MAGIC	},
146 	{ MOUNT_OVERLAY,	LINUX_DEFAULT_SUPER_MAGIC	},
147 	{ MOUNT_UMAP,		LINUX_DEFAULT_SUPER_MAGIC	},
148 	{ MOUNT_KERNFS,		LINUX_DEFAULT_SUPER_MAGIC	},
149 	{ MOUNT_PROCFS,		LINUX_PROC_SUPER_MAGIC		},
150 	{ MOUNT_AFS,		LINUX_DEFAULT_SUPER_MAGIC	},
151 	{ MOUNT_CD9660,		LINUX_ISOFS_SUPER_MAGIC		},
152 	{ MOUNT_UNION,		LINUX_DEFAULT_SUPER_MAGIC	},
153 	{ MOUNT_ADOSFS,		LINUX_ADFS_SUPER_MAGIC		},
154 	{ MOUNT_EXT2FS,		LINUX_EXT2_SUPER_MAGIC		},
155 	{ MOUNT_CFS,		LINUX_DEFAULT_SUPER_MAGIC	},
156 	{ MOUNT_CODA,		LINUX_CODA_SUPER_MAGIC		},
157 	{ MOUNT_FILECORE,	LINUX_DEFAULT_SUPER_MAGIC	},
158 	{ MOUNT_NTFS,		LINUX_DEFAULT_SUPER_MAGIC	},
159 	{ MOUNT_SMBFS,		LINUX_SMB_SUPER_MAGIC		}
160 };
161 #define FSTYPESSIZE (sizeof(fstypes) / sizeof(fstypes[0]))
162 
163 /* Local linux_misc.c functions: */
164 static void bsd_to_linux_statfs __P((struct statfs *, struct linux_statfs *));
165 static int linux_to_bsd_limit __P((int));
166 
167 /*
168  * The information on a terminated (or stopped) process needs
169  * to be converted in order for Linux binaries to get a valid signal
170  * number out of it.
171  */
172 void
173 bsd_to_linux_wstat(st)
174 	int *st;
175 {
176 
177 	int sig;
178 
179 	if (WIFSIGNALED(*st)) {
180 		sig = WTERMSIG(*st);
181 		if (sig >= 0 && sig < NSIG)
182 			*st= (*st& ~0177) | native_to_linux_sig[sig];
183 	} else if (WIFSTOPPED(*st)) {
184 		sig = WSTOPSIG(*st);
185 		if (sig >= 0 && sig < NSIG)
186 			*st = (*st & ~0xff00) | (native_to_linux_sig[sig] << 8);
187 	}
188 }
189 
190 /*
191  * This is very much the same as waitpid()
192  */
193 int
194 linux_sys_wait4(p, v, retval)
195 	struct proc *p;
196 	void *v;
197 	register_t *retval;
198 {
199 	struct linux_sys_wait4_args /* {
200 		syscallarg(int) pid;
201 		syscallarg(int *) status;
202 		syscallarg(int) options;
203 		syscallarg(struct rusage *) rusage;
204 	} */ *uap = v;
205 	struct sys_wait4_args w4a;
206 	int error, *status, tstat, options, linux_options;
207 	caddr_t sg;
208 
209 	if (SCARG(uap, status) != NULL) {
210 		sg = stackgap_init(p->p_emul);
211 		status = (int *) stackgap_alloc(&sg, sizeof *status);
212 	} else
213 		status = NULL;
214 
215 	linux_options = SCARG(uap, options);
216 	options = 0;
217 	if (linux_options &
218 	    ~(LINUX_WAIT4_WNOHANG|LINUX_WAIT4_WUNTRACED|LINUX_WAIT4_WALL|
219 	      LINUX_WAIT4_WCLONE))
220 		return (EINVAL);
221 
222 	if (linux_options & LINUX_WAIT4_WNOHANG)
223 		options |= WNOHANG;
224 	if (linux_options & LINUX_WAIT4_WUNTRACED)
225 		options |= WUNTRACED;
226 	if (linux_options & LINUX_WAIT4_WALL)
227 		options |= WALLSIG;
228 	if (linux_options & LINUX_WAIT4_WCLONE)
229 		options |= WALTSIG;
230 
231 	SCARG(&w4a, pid) = SCARG(uap, pid);
232 	SCARG(&w4a, status) = status;
233 	SCARG(&w4a, options) = options;
234 	SCARG(&w4a, rusage) = SCARG(uap, rusage);
235 
236 	if ((error = sys_wait4(p, &w4a, retval)))
237 		return error;
238 
239 	sigdelset(&p->p_sigctx.ps_siglist, SIGCHLD);
240 
241 	if (status != NULL) {
242 		if ((error = copyin(status, &tstat, sizeof tstat)))
243 			return error;
244 
245 		bsd_to_linux_wstat(&tstat);
246 		return copyout(&tstat, SCARG(uap, status), sizeof tstat);
247 	}
248 
249 	return 0;
250 }
251 
252 /*
253  * Linux brk(2). The check if the new address is >= the old one is
254  * done in the kernel in Linux. NetBSD does it in the library.
255  */
256 int
257 linux_sys_brk(p, v, retval)
258 	struct proc *p;
259 	void *v;
260 	register_t *retval;
261 {
262 	struct linux_sys_brk_args /* {
263 		syscallarg(char *) nsize;
264 	} */ *uap = v;
265 	char *nbrk = SCARG(uap, nsize);
266 	struct sys_obreak_args oba;
267 	struct vmspace *vm = p->p_vmspace;
268 	struct linux_emuldata *ed = (struct linux_emuldata*)p->p_emuldata;
269 
270 	SCARG(&oba, nsize) = nbrk;
271 
272 	if ((caddr_t) nbrk > vm->vm_daddr && sys_obreak(p, &oba, retval) == 0)
273 		ed->p_break = (char*)nbrk;
274 	else
275 		nbrk = ed->p_break;
276 
277 	retval[0] = (register_t)nbrk;
278 
279 	return 0;
280 }
281 
282 /*
283  * Convert BSD statfs structure to Linux statfs structure.
284  * The Linux structure has less fields, and it also wants
285  * the length of a name in a dir entry in a field, which
286  * we fake (probably the wrong way).
287  */
288 static void
289 bsd_to_linux_statfs(bsp, lsp)
290 	struct statfs *bsp;
291 	struct linux_statfs *lsp;
292 {
293 	int i;
294 
295 	for (i = 0; i < FSTYPESSIZE; i++)
296 		if (strcmp(bsp->f_fstypename, fstypes[i].bsd) == 0)
297 			break;
298 
299 	if (i == FSTYPESSIZE) {
300 #ifdef DIAGNOSTIC
301 		printf("unhandled fstype in linux emulation: %s\n",
302 		    bsp->f_fstypename);
303 #endif
304 		lsp->l_ftype = LINUX_DEFAULT_SUPER_MAGIC;
305 	} else {
306 		lsp->l_ftype = fstypes[i].linux;
307 	}
308 
309 	lsp->l_fbsize = bsp->f_bsize;
310 	lsp->l_fblocks = bsp->f_blocks;
311 	lsp->l_fbfree = bsp->f_bfree;
312 	lsp->l_fbavail = bsp->f_bavail;
313 	lsp->l_ffiles = bsp->f_files;
314 	lsp->l_fffree = bsp->f_ffree;
315 	/* Linux sets the fsid to 0..., we don't */
316 	lsp->l_ffsid.val[0] = bsp->f_fsid.val[0];
317 	lsp->l_ffsid.val[1] = bsp->f_fsid.val[1];
318 	lsp->l_fnamelen = MAXNAMLEN;	/* XXX */
319 	(void)memset(lsp->l_fspare, 0, sizeof(lsp->l_fspare));
320 }
321 
322 /*
323  * Implement the fs stat functions. Straightforward.
324  */
325 int
326 linux_sys_statfs(p, v, retval)
327 	struct proc *p;
328 	void *v;
329 	register_t *retval;
330 {
331 	struct linux_sys_statfs_args /* {
332 		syscallarg(const char *) path;
333 		syscallarg(struct linux_statfs *) sp;
334 	} */ *uap = v;
335 	struct statfs btmp, *bsp;
336 	struct linux_statfs ltmp;
337 	struct sys_statfs_args bsa;
338 	caddr_t sg;
339 	int error;
340 
341 	sg = stackgap_init(p->p_emul);
342 	bsp = (struct statfs *) stackgap_alloc(&sg, sizeof (struct statfs));
343 
344 	CHECK_ALT_EXIST(p, &sg, SCARG(uap, path));
345 
346 	SCARG(&bsa, path) = SCARG(uap, path);
347 	SCARG(&bsa, buf) = bsp;
348 
349 	if ((error = sys_statfs(p, &bsa, retval)))
350 		return error;
351 
352 	if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp)))
353 		return error;
354 
355 	bsd_to_linux_statfs(&btmp, &ltmp);
356 
357 	return copyout((caddr_t) &ltmp, (caddr_t) SCARG(uap, sp), sizeof ltmp);
358 }
359 
360 int
361 linux_sys_fstatfs(p, v, retval)
362 	struct proc *p;
363 	void *v;
364 	register_t *retval;
365 {
366 	struct linux_sys_fstatfs_args /* {
367 		syscallarg(int) fd;
368 		syscallarg(struct linux_statfs *) sp;
369 	} */ *uap = v;
370 	struct statfs btmp, *bsp;
371 	struct linux_statfs ltmp;
372 	struct sys_fstatfs_args bsa;
373 	caddr_t sg;
374 	int error;
375 
376 	sg = stackgap_init(p->p_emul);
377 	bsp = (struct statfs *) stackgap_alloc(&sg, sizeof (struct statfs));
378 
379 	SCARG(&bsa, fd) = SCARG(uap, fd);
380 	SCARG(&bsa, buf) = bsp;
381 
382 	if ((error = sys_fstatfs(p, &bsa, retval)))
383 		return error;
384 
385 	if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp)))
386 		return error;
387 
388 	bsd_to_linux_statfs(&btmp, &ltmp);
389 
390 	return copyout((caddr_t) &ltmp, (caddr_t) SCARG(uap, sp), sizeof ltmp);
391 }
392 
393 /*
394  * uname(). Just copy the info from the various strings stored in the
395  * kernel, and put it in the Linux utsname structure. That structure
396  * is almost the same as the NetBSD one, only it has fields 65 characters
397  * long, and an extra domainname field.
398  */
399 int
400 linux_sys_uname(p, v, retval)
401 	struct proc *p;
402 	void *v;
403 	register_t *retval;
404 {
405 	struct linux_sys_uname_args /* {
406 		syscallarg(struct linux_utsname *) up;
407 	} */ *uap = v;
408 	struct linux_utsname luts;
409 
410 	strncpy(luts.l_sysname, linux_sysname, sizeof(luts.l_sysname));
411 	strncpy(luts.l_nodename, hostname, sizeof(luts.l_nodename));
412 	strncpy(luts.l_release, linux_release, sizeof(luts.l_release));
413 	strncpy(luts.l_version, linux_version, sizeof(luts.l_version));
414 	strncpy(luts.l_machine, machine, sizeof(luts.l_machine));
415 	strncpy(luts.l_domainname, domainname, sizeof(luts.l_domainname));
416 
417 	return copyout(&luts, SCARG(uap, up), sizeof(luts));
418 }
419 
420 /* Used directly on: alpha, mips, ppc, sparc, sparc64 */
421 /* Used indirectly on: arm, i386, m68k */
422 
423 /*
424  * New type Linux mmap call.
425  * Only called directly on machines with >= 6 free regs.
426  */
427 int
428 linux_sys_mmap(p, v, retval)
429 	struct proc *p;
430 	void *v;
431 	register_t *retval;
432 {
433 	struct linux_sys_mmap_args /* {
434 		syscallarg(unsigned long) addr;
435 		syscallarg(size_t) len;
436 		syscallarg(int) prot;
437 		syscallarg(int) flags;
438 		syscallarg(int) fd;
439 		syscallarg(linux_off_t) offset;
440 	} */ *uap = v;
441 	struct sys_mmap_args cma;
442 	int flags;
443 
444 	flags = 0;
445 	flags |= cvtto_bsd_mask(SCARG(uap,flags), LINUX_MAP_SHARED, MAP_SHARED);
446 	flags |= cvtto_bsd_mask(SCARG(uap,flags), LINUX_MAP_PRIVATE, MAP_PRIVATE);
447 	flags |= cvtto_bsd_mask(SCARG(uap,flags), LINUX_MAP_FIXED, MAP_FIXED);
448 	flags |= cvtto_bsd_mask(SCARG(uap,flags), LINUX_MAP_ANON, MAP_ANON);
449 	/* XXX XAX ERH: Any other flags here?  There are more defined... */
450 
451 	SCARG(&cma,addr) = (void *)SCARG(uap, addr);
452 	SCARG(&cma,len) = SCARG(uap, len);
453 	SCARG(&cma,prot) = SCARG(uap, prot);
454 	if (SCARG(&cma,prot) & VM_PROT_WRITE) /* XXX */
455 		SCARG(&cma,prot) |= VM_PROT_READ;
456 	SCARG(&cma,flags) = flags;
457 	SCARG(&cma,fd) = flags & MAP_ANON ? -1 : SCARG(uap, fd);
458 	SCARG(&cma,pad) = 0;
459 	SCARG(&cma,pos) = (off_t)SCARG(uap, offset);
460 
461 	return sys_mmap(p, &cma, retval);
462 }
463 
464 /*
465  * Newer type Linux mmap call.
466  */
467 int
468 linux_sys_mmap2(p, v, retval)
469 	struct proc *p;
470 	void *v;
471 	register_t *retval;
472 {
473 	struct linux_sys_mmap2_args /* {
474 		syscallarg(void *) addr;
475 		syscallarg(size_t) len;
476 		syscallarg(int) prot;
477 		syscallarg(int) flags;
478 		syscallarg(int) fd;
479 		syscallarg(off_t) offset;
480 	} */ *uap = v;
481 	struct sys_mmap_args cma;
482 	int flags;
483 
484 	flags = 0;
485 	flags |= cvtto_bsd_mask(SCARG(uap,flags), LINUX_MAP_SHARED, MAP_SHARED);
486 	flags |= cvtto_bsd_mask(SCARG(uap,flags), LINUX_MAP_PRIVATE, MAP_PRIVATE);
487 	flags |= cvtto_bsd_mask(SCARG(uap,flags), LINUX_MAP_FIXED, MAP_FIXED);
488 	flags |= cvtto_bsd_mask(SCARG(uap,flags), LINUX_MAP_ANON, MAP_ANON);
489 	/* XXX XAX ERH: Any other flags here?  There are more defined... */
490 
491 	SCARG(&cma,addr) = (void *)SCARG(uap, addr);
492 	SCARG(&cma,len) = SCARG(uap, len);
493 	SCARG(&cma,prot) = SCARG(uap, prot);
494 	if (SCARG(&cma,prot) & VM_PROT_WRITE) /* XXX */
495 		SCARG(&cma,prot) |= VM_PROT_READ;
496 	SCARG(&cma,flags) = flags;
497 	SCARG(&cma,fd) = flags & MAP_ANON ? -1 : SCARG(uap, fd);
498 	SCARG(&cma,pos) = (off_t)SCARG(uap, offset);
499 
500 	return sys_mmap(p, &cma, retval);
501 }
502 
503 int
504 linux_sys_mremap(p, v, retval)
505 	struct proc *p;
506 	void *v;
507 	register_t *retval;
508 {
509 	struct linux_sys_mremap_args /* {
510 		syscallarg(void *) old_address;
511 		syscallarg(size_t) old_size;
512 		syscallarg(size_t) new_size;
513 		syscallarg(u_long) flags;
514 	} */ *uap = v;
515 	struct sys_munmap_args mua;
516 	size_t old_size, new_size;
517 	int error;
518 
519 	old_size = round_page(SCARG(uap, old_size));
520 	new_size = round_page(SCARG(uap, new_size));
521 
522 	/*
523 	 * Growing mapped region.
524 	 */
525 	if (new_size > old_size) {
526 		/*
527 		 * XXX Implement me.  What we probably want to do is
528 		 * XXX dig out the guts of the old mapping, mmap that
529 		 * XXX object again with the new size, then munmap
530 		 * XXX the old mapping.
531 		 */
532 		*retval = 0;
533 		return (ENOMEM);
534 	}
535 
536 	/*
537 	 * Shrinking mapped region.
538 	 */
539 	if (new_size < old_size) {
540 		SCARG(&mua, addr) = (caddr_t)SCARG(uap, old_address) +
541 		    new_size;
542 		SCARG(&mua, len) = old_size - new_size;
543 		error = sys_munmap(p, &mua, retval);
544 		*retval = error ? 0 : (register_t)SCARG(uap, old_address);
545 		return (error);
546 	}
547 
548 	/*
549 	 * No change.
550 	 */
551 	*retval = (register_t)SCARG(uap, old_address);
552 	return (0);
553 }
554 
555 int
556 linux_sys_msync(p, v, retval)
557 	struct proc *p;
558 	void *v;
559 	register_t *retval;
560 {
561 	struct linux_sys_msync_args /* {
562 		syscallarg(caddr_t) addr;
563 		syscallarg(int) len;
564 		syscallarg(int) fl;
565 	} */ *uap = v;
566 
567 	struct sys___msync13_args bma;
568 
569 	/* flags are ignored */
570 	SCARG(&bma, addr) = SCARG(uap, addr);
571 	SCARG(&bma, len) = SCARG(uap, len);
572 	SCARG(&bma, flags) = SCARG(uap, fl);
573 
574 	return sys___msync13(p, &bma, retval);
575 }
576 
577 /*
578  * This code is partly stolen from src/lib/libc/compat-43/times.c
579  * XXX - CLK_TCK isn't declared in /sys, just in <time.h>, done here
580  */
581 
582 #define CLK_TCK 100
583 #define	CONVTCK(r)	(r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
584 
585 int
586 linux_sys_times(p, v, retval)
587 	struct proc *p;
588 	void *v;
589 	register_t *retval;
590 {
591 	struct linux_sys_times_args /* {
592 		syscallarg(struct times *) tms;
593 	} */ *uap = v;
594 	struct timeval t;
595 	struct linux_tms ltms;
596 	struct rusage ru;
597 	int error, s;
598 
599 	calcru(p, &ru.ru_utime, &ru.ru_stime, NULL);
600 	ltms.ltms_utime = CONVTCK(ru.ru_utime);
601 	ltms.ltms_stime = CONVTCK(ru.ru_stime);
602 
603 	ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime);
604 	ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime);
605 
606 	if ((error = copyout(&ltms, SCARG(uap, tms), sizeof ltms)))
607 		return error;
608 
609 	s = splclock();
610 	timersub(&time, &boottime, &t);
611 	splx(s);
612 
613 	retval[0] = ((linux_clock_t)(CONVTCK(t)));
614 	return 0;
615 }
616 
617 /*
618  * Linux 'readdir' call. This code is mostly taken from the
619  * SunOS getdents call (see compat/sunos/sunos_misc.c), though
620  * an attempt has been made to keep it a little cleaner (failing
621  * miserably, because of the cruft needed if count 1 is passed).
622  *
623  * The d_off field should contain the offset of the next valid entry,
624  * but in Linux it has the offset of the entry itself. We emulate
625  * that bug here.
626  *
627  * Read in BSD-style entries, convert them, and copy them out.
628  *
629  * Note that this doesn't handle union-mounted filesystems.
630  */
631 int
632 linux_sys_getdents(p, v, retval)
633 	struct proc *p;
634 	void *v;
635 	register_t *retval;
636 {
637 	struct linux_sys_getdents_args /* {
638 		syscallarg(int) fd;
639 		syscallarg(struct linux_dirent *) dent;
640 		syscallarg(unsigned int) count;
641 	} */ *uap = v;
642 	struct dirent *bdp;
643 	struct vnode *vp;
644 	caddr_t	inp, buf;		/* BSD-format */
645 	int len, reclen;		/* BSD-format */
646 	caddr_t outp;			/* Linux-format */
647 	int resid, linux_reclen = 0;	/* Linux-format */
648 	struct file *fp;
649 	struct uio auio;
650 	struct iovec aiov;
651 	struct linux_dirent idb;
652 	off_t off;		/* true file offset */
653 	int buflen, error, eofflag, nbytes, oldcall;
654 	struct vattr va;
655 	off_t *cookiebuf = NULL, *cookie;
656 	int ncookies;
657 
658 	/* getvnode() will use the descriptor for us */
659 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
660 		return (error);
661 
662 	if ((fp->f_flag & FREAD) == 0) {
663 		error = EBADF;
664 		goto out1;
665 	}
666 
667 	vp = (struct vnode *)fp->f_data;
668 	if (vp->v_type != VDIR) {
669 		error = EINVAL;
670 		goto out1;
671 	}
672 
673 	if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)))
674 		goto out1;
675 
676 	nbytes = SCARG(uap, count);
677 	if (nbytes == 1) {	/* emulating old, broken behaviour */
678 		nbytes = sizeof (struct linux_dirent);
679 		buflen = max(va.va_blocksize, nbytes);
680 		oldcall = 1;
681 	} else {
682 		buflen = min(MAXBSIZE, nbytes);
683 		if (buflen < va.va_blocksize)
684 			buflen = va.va_blocksize;
685 		oldcall = 0;
686 	}
687 	buf = malloc(buflen, M_TEMP, M_WAITOK);
688 
689 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
690 	off = fp->f_offset;
691 again:
692 	aiov.iov_base = buf;
693 	aiov.iov_len = buflen;
694 	auio.uio_iov = &aiov;
695 	auio.uio_iovcnt = 1;
696 	auio.uio_rw = UIO_READ;
697 	auio.uio_segflg = UIO_SYSSPACE;
698 	auio.uio_procp = p;
699 	auio.uio_resid = buflen;
700 	auio.uio_offset = off;
701 	/*
702          * First we read into the malloc'ed buffer, then
703          * we massage it into user space, one record at a time.
704          */
705 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf,
706 	    &ncookies);
707 	if (error)
708 		goto out;
709 
710 	inp = buf;
711 	outp = (caddr_t)SCARG(uap, dent);
712 	resid = nbytes;
713 	if ((len = buflen - auio.uio_resid) == 0)
714 		goto eof;
715 
716 	for (cookie = cookiebuf; len > 0; len -= reclen) {
717 		bdp = (struct dirent *)inp;
718 		reclen = bdp->d_reclen;
719 		if (reclen & 3)
720 			panic("linux_readdir");
721 		if (bdp->d_fileno == 0) {
722 			inp += reclen;	/* it is a hole; squish it out */
723 			off = *cookie++;
724 			continue;
725 		}
726 		linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen);
727 		if (reclen > len || resid < linux_reclen) {
728 			/* entry too big for buffer, so just stop */
729 			outp++;
730 			break;
731 		}
732 		/*
733 		 * Massage in place to make a Linux-shaped dirent (otherwise
734 		 * we have to worry about touching user memory outside of
735 		 * the copyout() call).
736 		 */
737 		idb.d_ino = (linux_ino_t)bdp->d_fileno;
738 		/*
739 		 * The old readdir() call misuses the offset and reclen fields.
740 		 */
741 		if (oldcall) {
742 			idb.d_off = (linux_off_t)linux_reclen;
743 			idb.d_reclen = (u_short)bdp->d_namlen;
744 		} else {
745 			if (sizeof (linux_off_t) < 4 && (off >> 32) != 0) {
746 				compat_offseterr(vp, "linux_getdents");
747 				error = EINVAL;
748 				goto out;
749 			}
750 			idb.d_off = (linux_off_t)off;
751 			idb.d_reclen = (u_short)linux_reclen;
752 		}
753 		strcpy(idb.d_name, bdp->d_name);
754 		if ((error = copyout((caddr_t)&idb, outp, linux_reclen)))
755 			goto out;
756 		/* advance past this real entry */
757 		inp += reclen;
758 		off = *cookie++;	/* each entry points to itself */
759 		/* advance output past Linux-shaped entry */
760 		outp += linux_reclen;
761 		resid -= linux_reclen;
762 		if (oldcall)
763 			break;
764 	}
765 
766 	/* if we squished out the whole block, try again */
767 	if (outp == (caddr_t)SCARG(uap, dent))
768 		goto again;
769 	fp->f_offset = off;	/* update the vnode offset */
770 
771 	if (oldcall)
772 		nbytes = resid + linux_reclen;
773 
774 eof:
775 	*retval = nbytes - resid;
776 out:
777 	VOP_UNLOCK(vp, 0);
778 	if (cookiebuf)
779 		free(cookiebuf, M_TEMP);
780 	free(buf, M_TEMP);
781  out1:
782 	FILE_UNUSE(fp, p);
783 	return error;
784 }
785 
786 /*
787  * Even when just using registers to pass arguments to syscalls you can
788  * have 5 of them on the i386. So this newer version of select() does
789  * this.
790  */
791 int
792 linux_sys_select(p, v, retval)
793 	struct proc *p;
794 	void *v;
795 	register_t *retval;
796 {
797 	struct linux_sys_select_args /* {
798 		syscallarg(int) nfds;
799 		syscallarg(fd_set *) readfds;
800 		syscallarg(fd_set *) writefds;
801 		syscallarg(fd_set *) exceptfds;
802 		syscallarg(struct timeval *) timeout;
803 	} */ *uap = v;
804 
805 	return linux_select1(p, retval, SCARG(uap, nfds), SCARG(uap, readfds),
806 	    SCARG(uap, writefds), SCARG(uap, exceptfds), SCARG(uap, timeout));
807 }
808 
809 /*
810  * Common code for the old and new versions of select(). A couple of
811  * things are important:
812  * 1) return the amount of time left in the 'timeout' parameter
813  * 2) select never returns ERESTART on Linux, always return EINTR
814  */
815 int
816 linux_select1(p, retval, nfds, readfds, writefds, exceptfds, timeout)
817 	struct proc *p;
818 	register_t *retval;
819 	int nfds;
820 	fd_set *readfds, *writefds, *exceptfds;
821 	struct timeval *timeout;
822 {
823 	struct sys_select_args bsa;
824 	struct timeval tv0, tv1, utv, *tvp;
825 	caddr_t sg;
826 	int error;
827 
828 	SCARG(&bsa, nd) = nfds;
829 	SCARG(&bsa, in) = readfds;
830 	SCARG(&bsa, ou) = writefds;
831 	SCARG(&bsa, ex) = exceptfds;
832 	SCARG(&bsa, tv) = timeout;
833 
834 	/*
835 	 * Store current time for computation of the amount of
836 	 * time left.
837 	 */
838 	if (timeout) {
839 		if ((error = copyin(timeout, &utv, sizeof(utv))))
840 			return error;
841 		if (itimerfix(&utv)) {
842 			/*
843 			 * The timeval was invalid.  Convert it to something
844 			 * valid that will act as it does under Linux.
845 			 */
846 			sg = stackgap_init(p->p_emul);
847 			tvp = stackgap_alloc(&sg, sizeof(utv));
848 			utv.tv_sec += utv.tv_usec / 1000000;
849 			utv.tv_usec %= 1000000;
850 			if (utv.tv_usec < 0) {
851 				utv.tv_sec -= 1;
852 				utv.tv_usec += 1000000;
853 			}
854 			if (utv.tv_sec < 0)
855 				timerclear(&utv);
856 			if ((error = copyout(&utv, tvp, sizeof(utv))))
857 				return error;
858 			SCARG(&bsa, tv) = tvp;
859 		}
860 		microtime(&tv0);
861 	}
862 
863 	error = sys_select(p, &bsa, retval);
864 	if (error) {
865 		/*
866 		 * See fs/select.c in the Linux kernel.  Without this,
867 		 * Maelstrom doesn't work.
868 		 */
869 		if (error == ERESTART)
870 			error = EINTR;
871 		return error;
872 	}
873 
874 	if (timeout) {
875 		if (*retval) {
876 			/*
877 			 * Compute how much time was left of the timeout,
878 			 * by subtracting the current time and the time
879 			 * before we started the call, and subtracting
880 			 * that result from the user-supplied value.
881 			 */
882 			microtime(&tv1);
883 			timersub(&tv1, &tv0, &tv1);
884 			timersub(&utv, &tv1, &utv);
885 			if (utv.tv_sec < 0)
886 				timerclear(&utv);
887 		} else
888 			timerclear(&utv);
889 		if ((error = copyout(&utv, timeout, sizeof(utv))))
890 			return error;
891 	}
892 
893 	return 0;
894 }
895 
896 /*
897  * Get the process group of a certain process. Look it up
898  * and return the value.
899  */
900 int
901 linux_sys_getpgid(p, v, retval)
902 	struct proc *p;
903 	void *v;
904 	register_t *retval;
905 {
906 	struct linux_sys_getpgid_args /* {
907 		syscallarg(int) pid;
908 	} */ *uap = v;
909 	struct proc *targp;
910 
911 	if (SCARG(uap, pid) != 0 && SCARG(uap, pid) != p->p_pid) {
912 		if ((targp = pfind(SCARG(uap, pid))) == 0)
913 			return ESRCH;
914 	}
915 	else
916 		targp = p;
917 
918 	retval[0] = targp->p_pgid;
919 	return 0;
920 }
921 
922 /*
923  * Set the 'personality' (emulation mode) for the current process. Only
924  * accept the Linux personality here (0). This call is needed because
925  * the Linux ELF crt0 issues it in an ugly kludge to make sure that
926  * ELF binaries run in Linux mode, not SVR4 mode.
927  */
928 int
929 linux_sys_personality(p, v, retval)
930 	struct proc *p;
931 	void *v;
932 	register_t *retval;
933 {
934 	struct linux_sys_personality_args /* {
935 		syscallarg(int) per;
936 	} */ *uap = v;
937 
938 	if (SCARG(uap, per) != 0)
939 		return EINVAL;
940 	retval[0] = 0;
941 	return 0;
942 }
943 
944 #if defined(__i386__) || defined(__m68k__)
945 /*
946  * The calls are here because of type conversions.
947  */
948 int
949 linux_sys_setreuid16(p, v, retval)
950 	struct proc *p;
951 	void *v;
952 	register_t *retval;
953 {
954 	struct linux_sys_setreuid16_args /* {
955 		syscallarg(int) ruid;
956 		syscallarg(int) euid;
957 	} */ *uap = v;
958 	struct sys_setreuid_args bsa;
959 
960 	SCARG(&bsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
961 		(uid_t)-1 : SCARG(uap, ruid);
962 	SCARG(&bsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
963 		(uid_t)-1 : SCARG(uap, euid);
964 
965 	return sys_setreuid(p, &bsa, retval);
966 }
967 
968 int
969 linux_sys_setregid16(p, v, retval)
970 	struct proc *p;
971 	void *v;
972 	register_t *retval;
973 {
974 	struct linux_sys_setregid16_args /* {
975 		syscallarg(int) rgid;
976 		syscallarg(int) egid;
977 	} */ *uap = v;
978 	struct sys_setregid_args bsa;
979 
980 	SCARG(&bsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
981 		(uid_t)-1 : SCARG(uap, rgid);
982 	SCARG(&bsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
983 		(uid_t)-1 : SCARG(uap, egid);
984 
985 	return sys_setregid(p, &bsa, retval);
986 }
987 
988 int
989 linux_sys_setresuid16(p, v, retval)
990 	struct proc *p;
991 	void *v;
992 	register_t *retval;
993 {
994 	struct linux_sys_setresuid16_args /* {
995 		syscallarg(uid_t) ruid;
996 		syscallarg(uid_t) euid;
997 		syscallarg(uid_t) suid;
998 	} */ *uap = v;
999 	struct linux_sys_setresuid16_args lsa;
1000 
1001 	SCARG(&lsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
1002 		(uid_t)-1 : SCARG(uap, ruid);
1003 	SCARG(&lsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
1004 		(uid_t)-1 : SCARG(uap, euid);
1005 	SCARG(&lsa, suid) = ((linux_uid_t)SCARG(uap, suid) == (linux_uid_t)-1) ?
1006 		(uid_t)-1 : SCARG(uap, suid);
1007 
1008 	return linux_sys_setresuid(p, &lsa, retval);
1009 }
1010 
1011 int
1012 linux_sys_setresgid16(p, v, retval)
1013 	struct proc *p;
1014 	void *v;
1015 	register_t *retval;
1016 {
1017 	struct linux_sys_setresgid16_args /* {
1018 		syscallarg(gid_t) rgid;
1019 		syscallarg(gid_t) egid;
1020 		syscallarg(gid_t) sgid;
1021 	} */ *uap = v;
1022 	struct linux_sys_setresgid16_args lsa;
1023 
1024 	SCARG(&lsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
1025 		(gid_t)-1 : SCARG(uap, rgid);
1026 	SCARG(&lsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
1027 		(gid_t)-1 : SCARG(uap, egid);
1028 	SCARG(&lsa, sgid) = ((linux_gid_t)SCARG(uap, sgid) == (linux_gid_t)-1) ?
1029 		(gid_t)-1 : SCARG(uap, sgid);
1030 
1031 	return linux_sys_setresgid(p, &lsa, retval);
1032 }
1033 
1034 int
1035 linux_sys_getgroups16(p, v, retval)
1036 	struct proc *p;
1037 	void *v;
1038 	register_t *retval;
1039 {
1040 	struct linux_sys_getgroups16_args /* {
1041 		syscallarg(int) gidsetsize;
1042 		syscallarg(linux_gid_t *) gidset;
1043 	} */ *uap = v;
1044 	caddr_t sg;
1045 	int n, error, i;
1046 	struct sys_getgroups_args bsa;
1047 	gid_t *bset, *kbset;
1048 	linux_gid_t *lset;
1049 	struct pcred *pc = p->p_cred;
1050 
1051 	n = SCARG(uap, gidsetsize);
1052 	if (n < 0)
1053 		return EINVAL;
1054 	error = 0;
1055 	bset = kbset = NULL;
1056 	lset = NULL;
1057 	if (n > 0) {
1058 		n = min(pc->pc_ucred->cr_ngroups, n);
1059 		sg = stackgap_init(p->p_emul);
1060 		bset = stackgap_alloc(&sg, n * sizeof (gid_t));
1061 		kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK);
1062 		lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
1063 		if (bset == NULL || kbset == NULL || lset == NULL)
1064 			return ENOMEM;
1065 		SCARG(&bsa, gidsetsize) = n;
1066 		SCARG(&bsa, gidset) = bset;
1067 		error = sys_getgroups(p, &bsa, retval);
1068 		if (error != 0)
1069 			goto out;
1070 		error = copyin(bset, kbset, n * sizeof (gid_t));
1071 		if (error != 0)
1072 			goto out;
1073 		for (i = 0; i < n; i++)
1074 			lset[i] = (linux_gid_t)kbset[i];
1075 		error = copyout(lset, SCARG(uap, gidset),
1076 		    n * sizeof (linux_gid_t));
1077 	} else
1078 		*retval = pc->pc_ucred->cr_ngroups;
1079 out:
1080 	if (kbset != NULL)
1081 		free(kbset, M_TEMP);
1082 	if (lset != NULL)
1083 		free(lset, M_TEMP);
1084 	return error;
1085 }
1086 
1087 int
1088 linux_sys_setgroups16(p, v, retval)
1089 	struct proc *p;
1090 	void *v;
1091 	register_t *retval;
1092 {
1093 	struct linux_sys_setgroups16_args /* {
1094 		syscallarg(int) gidsetsize;
1095 		syscallarg(linux_gid_t *) gidset;
1096 	} */ *uap = v;
1097 	caddr_t sg;
1098 	int n;
1099 	int error, i;
1100 	struct sys_setgroups_args bsa;
1101 	gid_t *bset, *kbset;
1102 	linux_gid_t *lset;
1103 
1104 	n = SCARG(uap, gidsetsize);
1105 	if (n < 0 || n > NGROUPS)
1106 		return EINVAL;
1107 	sg = stackgap_init(p->p_emul);
1108 	bset = stackgap_alloc(&sg, n * sizeof (gid_t));
1109 	lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
1110 	kbset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
1111 	if (lset == NULL || bset == NULL)
1112 		return ENOMEM;
1113 	error = copyin(SCARG(uap, gidset), lset, n * sizeof (linux_gid_t));
1114 	if (error != 0)
1115 		goto out;
1116 	for (i = 0; i < n; i++)
1117 		kbset[i] = (gid_t)lset[i];
1118 	error = copyout(kbset, bset, n * sizeof (gid_t));
1119 	if (error != 0)
1120 		goto out;
1121 	SCARG(&bsa, gidsetsize) = n;
1122 	SCARG(&bsa, gidset) = bset;
1123 	error = sys_setgroups(p, &bsa, retval);
1124 
1125 out:
1126 	if (lset != NULL)
1127 		free(lset, M_TEMP);
1128 	if (kbset != NULL)
1129 		free(kbset, M_TEMP);
1130 
1131 	return error;
1132 }
1133 
1134 #endif /* __i386__ || __m68k__ */
1135 
1136 /*
1137  * We have nonexistent fsuid equal to uid.
1138  * If modification is requested, refuse.
1139  */
1140 int
1141 linux_sys_setfsuid(p, v, retval)
1142 	 struct proc *p;
1143 	 void *v;
1144 	 register_t *retval;
1145 {
1146 	 struct linux_sys_setfsuid_args /* {
1147 		 syscallarg(uid_t) uid;
1148 	 } */ *uap = v;
1149 	 uid_t uid;
1150 
1151 	 uid = SCARG(uap, uid);
1152 	 if (p->p_cred->p_ruid != uid)
1153 		 return sys_nosys(p, v, retval);
1154 	 else
1155 		 return (0);
1156 }
1157 
1158 /* XXX XXX XXX */
1159 #ifndef alpha
1160 int
1161 linux_sys_getfsuid(p, v, retval)
1162 	struct proc *p;
1163 	void *v;
1164 	register_t *retval;
1165 {
1166 	return sys_getuid(p, v, retval);
1167 }
1168 #endif
1169 
1170 int
1171 linux_sys_setresuid(p, v, retval)
1172 	struct proc *p;
1173 	void *v;
1174 	register_t *retval;
1175 {
1176 	struct linux_sys_setresuid_args /* {
1177 		syscallarg(uid_t) ruid;
1178 		syscallarg(uid_t) euid;
1179 		syscallarg(uid_t) suid;
1180 	} */ *uap = v;
1181 	struct pcred *pc = p->p_cred;
1182 	uid_t ruid, euid, suid;
1183 	int error;
1184 
1185 	ruid = SCARG(uap, ruid);
1186 	euid = SCARG(uap, euid);
1187 	suid = SCARG(uap, suid);
1188 
1189 	/*
1190 	 * Note: These checks are a little different than the NetBSD
1191 	 * setreuid(2) call performs.  This precisely follows the
1192 	 * behavior of the Linux kernel.
1193 	 */
1194 	if (ruid != (uid_t)-1 &&
1195 	    ruid != pc->p_ruid &&
1196 	    ruid != pc->pc_ucred->cr_uid &&
1197 	    ruid != pc->p_svuid &&
1198 	    (error = suser(pc->pc_ucred, &p->p_acflag)))
1199 		return (error);
1200 
1201 	if (euid != (uid_t)-1 &&
1202 	    euid != pc->p_ruid &&
1203 	    euid != pc->pc_ucred->cr_uid &&
1204 	    euid != pc->p_svuid &&
1205 	    (error = suser(pc->pc_ucred, &p->p_acflag)))
1206 		return (error);
1207 
1208 	if (suid != (uid_t)-1 &&
1209 	    suid != pc->p_ruid &&
1210 	    suid != pc->pc_ucred->cr_uid &&
1211 	    suid != pc->p_svuid &&
1212 	    (error = suser(pc->pc_ucred, &p->p_acflag)))
1213 		return (error);
1214 
1215 	/*
1216 	 * Now assign the new real, effective, and saved UIDs.
1217 	 * Note that Linux, unlike NetBSD in setreuid(2), does not
1218 	 * set the saved UID in this call unless the user specifies
1219 	 * it.
1220 	 */
1221 	if (ruid != (uid_t)-1) {
1222 		(void)chgproccnt(pc->p_ruid, -1);
1223 		(void)chgproccnt(ruid, 1);
1224 		pc->p_ruid = ruid;
1225 	}
1226 
1227 	if (euid != (uid_t)-1) {
1228 		pc->pc_ucred = crcopy(pc->pc_ucred);
1229 		pc->pc_ucred->cr_uid = euid;
1230 	}
1231 
1232 	if (suid != (uid_t)-1)
1233 		pc->p_svuid = suid;
1234 
1235 	if (ruid != (uid_t)-1 && euid != (uid_t)-1 && suid != (uid_t)-1)
1236 		p->p_flag |= P_SUGID;
1237 	return (0);
1238 }
1239 
1240 int
1241 linux_sys_getresuid(p, v, retval)
1242 	struct proc *p;
1243 	void *v;
1244 	register_t *retval;
1245 {
1246 	struct linux_sys_getresuid_args /* {
1247 		syscallarg(uid_t *) ruid;
1248 		syscallarg(uid_t *) euid;
1249 		syscallarg(uid_t *) suid;
1250 	} */ *uap = v;
1251 	struct pcred *pc = p->p_cred;
1252 	int error;
1253 
1254 	/*
1255 	 * Linux copies these values out to userspace like so:
1256 	 *
1257 	 *	1. Copy out ruid.
1258 	 *	2. If that succeeds, copy out euid.
1259 	 *	3. If both of those succeed, copy out suid.
1260 	 */
1261 	if ((error = copyout(&pc->p_ruid, SCARG(uap, ruid),
1262 			     sizeof(uid_t))) != 0)
1263 		return (error);
1264 
1265 	if ((error = copyout(&pc->pc_ucred->cr_uid, SCARG(uap, euid),
1266 			     sizeof(uid_t))) != 0)
1267 		return (error);
1268 
1269 	return (copyout(&pc->p_svuid, SCARG(uap, suid), sizeof(uid_t)));
1270 }
1271 
1272 int
1273 linux_sys_ptrace(p, v, retval)
1274 	struct proc *p;
1275 	void *v;
1276 	register_t *retval;
1277 {
1278 	struct linux_sys_ptrace_args /* {
1279 		i386, m68k, powerpc: T=int
1280 		alpha: T=long
1281 		syscallarg(T) request;
1282 		syscallarg(T) pid;
1283 		syscallarg(T) addr;
1284 		syscallarg(T) data;
1285 	} */ *uap = v;
1286 	const int *ptr;
1287 	int request;
1288 	int error;
1289 
1290 	ptr = linux_ptrace_request_map;
1291 	request = SCARG(uap, request);
1292 	while (*ptr != -1)
1293 		if (*ptr++ == request) {
1294 			struct sys_ptrace_args pta;
1295 
1296 			SCARG(&pta, req) = *ptr;
1297 			SCARG(&pta, pid) = SCARG(uap, pid);
1298 			SCARG(&pta, addr) = (caddr_t)SCARG(uap, addr);
1299 			SCARG(&pta, data) = SCARG(uap, data);
1300 
1301 			/*
1302 			 * Linux ptrace(PTRACE_CONT, pid, 0, 0) means actually
1303 			 * to continue where the process left off previously.
1304 			 * The same thing is achieved by addr == (caddr_t) 1
1305 			 * on NetBSD, so rewrite 'addr' appropriately.
1306 			 */
1307 			if (request == LINUX_PTRACE_CONT && SCARG(uap, addr)==0)
1308 				SCARG(&pta, addr) = (caddr_t) 1;
1309 
1310 			error = sys_ptrace(p, &pta, retval);
1311 			if (error)
1312 				return error;
1313 			switch (request) {
1314 			case LINUX_PTRACE_PEEKTEXT:
1315 			case LINUX_PTRACE_PEEKDATA:
1316 				error = copyout (retval,
1317 				    (caddr_t)SCARG(uap, data), sizeof *retval);
1318 				*retval = SCARG(uap, data);
1319 				break;
1320 			default:
1321 				break;
1322 			}
1323 			return error;
1324 		}
1325 		else
1326 			ptr++;
1327 
1328 	return LINUX_SYS_PTRACE_ARCH(p, uap, retval);
1329 }
1330 
1331 int
1332 linux_sys_reboot(struct proc *p, void *v, register_t *retval)
1333 {
1334 	struct linux_sys_reboot_args /* {
1335 		syscallarg(int) magic1;
1336 		syscallarg(int) magic2;
1337 		syscallarg(int) cmd;
1338 		syscallarg(void *) arg;
1339 	} */ *uap = v;
1340 	struct sys_reboot_args /* {
1341 		syscallarg(int) opt;
1342 		syscallarg(char *) bootstr;
1343 	} */ sra;
1344 	int error;
1345 
1346 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1347 		return(error);
1348 
1349 	if (SCARG(uap, magic1) != LINUX_REBOOT_MAGIC1)
1350 		return(EINVAL);
1351 	if (SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2 &&
1352 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2A &&
1353 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2B)
1354 		return(EINVAL);
1355 
1356 	switch (SCARG(uap, cmd)) {
1357 	case LINUX_REBOOT_CMD_RESTART:
1358 		SCARG(&sra, opt) = RB_AUTOBOOT;
1359 		break;
1360 	case LINUX_REBOOT_CMD_HALT:
1361 		SCARG(&sra, opt) = RB_HALT;
1362 		break;
1363 	case LINUX_REBOOT_CMD_POWER_OFF:
1364 		SCARG(&sra, opt) = RB_HALT|RB_POWERDOWN;
1365 		break;
1366 	case LINUX_REBOOT_CMD_RESTART2:
1367 		/* Reboot with an argument. */
1368 		SCARG(&sra, opt) = RB_AUTOBOOT|RB_STRING;
1369 		SCARG(&sra, bootstr) = SCARG(uap, arg);
1370 		break;
1371 	case LINUX_REBOOT_CMD_CAD_ON:
1372 		return(EINVAL);	/* We don't implement ctrl-alt-delete */
1373 	case LINUX_REBOOT_CMD_CAD_OFF:
1374 		return(0);
1375 	default:
1376 		return(EINVAL);
1377 	}
1378 
1379 	return(sys_reboot(p, &sra, retval));
1380 }
1381 
1382 /*
1383  * Copy of compat_12_sys_swapon().
1384  */
1385 int
1386 linux_sys_swapon(p, v, retval)
1387 	struct proc *p;
1388 	void *v;
1389 	register_t *retval;
1390 {
1391 	struct sys_swapctl_args ua;
1392 	struct linux_sys_swapon_args /* {
1393 		syscallarg(const char *) name;
1394 	} */ *uap = v;
1395 
1396 	SCARG(&ua, cmd) = SWAP_ON;
1397 	SCARG(&ua, arg) = (void *)SCARG(uap, name);
1398 	SCARG(&ua, misc) = 0;	/* priority */
1399 	return (sys_swapctl(p, &ua, retval));
1400 }
1401 
1402 /*
1403  * Stop swapping to the file or block device specified by path.
1404  */
1405 int
1406 linux_sys_swapoff(p, v, retval)
1407 	struct proc *p;
1408 	void *v;
1409 	register_t *retval;
1410 {
1411 	struct sys_swapctl_args ua;
1412 	struct linux_sys_swapoff_args /* {
1413 		syscallarg(const char *) path;
1414 	} */ *uap = v;
1415 
1416 	SCARG(&ua, cmd) = SWAP_OFF;
1417 	SCARG(&ua, arg) = (void *)SCARG(uap, path);
1418 	return (sys_swapctl(p, &ua, retval));
1419 }
1420 
1421 /*
1422  * Copy of compat_09_sys_setdomainname()
1423  */
1424 /* ARGSUSED */
1425 int
1426 linux_sys_setdomainname(p, v, retval)
1427 	struct proc *p;
1428 	void *v;
1429 	register_t *retval;
1430 {
1431 	struct linux_sys_setdomainname_args /* {
1432 		syscallarg(char *) domainname;
1433 		syscallarg(int) len;
1434 	} */ *uap = v;
1435 	int name;
1436 	int error;
1437 
1438 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1439 		return (error);
1440 	name = KERN_DOMAINNAME;
1441 	return (kern_sysctl(&name, 1, 0, 0, SCARG(uap, domainname),
1442 			    SCARG(uap, len), p));
1443 }
1444 
1445 /*
1446  * sysinfo()
1447  */
1448 /* ARGSUSED */
1449 int
1450 linux_sys_sysinfo(p, v, retval)
1451 	struct proc *p;
1452 	void *v;
1453 	register_t *retval;
1454 {
1455 	struct linux_sys_sysinfo_args /* {
1456 		syscallarg(struct linux_sysinfo *) arg;
1457 	} */ *uap = v;
1458 	struct linux_sysinfo si;
1459 	struct loadavg *la;
1460 
1461 	si.uptime = time.tv_sec - boottime.tv_sec;
1462 	la = &averunnable;
1463 	si.loads[0] = la->ldavg[0] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1464 	si.loads[1] = la->ldavg[1] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1465 	si.loads[2] = la->ldavg[2] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1466 	si.totalram = ctob(physmem);
1467 	si.freeram = uvmexp.free * uvmexp.pagesize;
1468 	si.sharedram = 0;	/* XXX */
1469 	si.bufferram = uvmexp.filepages * uvmexp.pagesize;
1470 	si.totalswap = uvmexp.swpages * uvmexp.pagesize;
1471 	si.freeswap = (uvmexp.swpages - uvmexp.swpginuse) * uvmexp.pagesize;
1472 	si.procs = nprocs;
1473 
1474 	/* The following are only present in newer Linux kernels. */
1475 	si.totalbig = 0;
1476 	si.freebig = 0;
1477 	si.mem_unit = 1;
1478 
1479 	return (copyout(&si, SCARG(uap, arg), sizeof si));
1480 }
1481 
1482 #define bsd_to_linux_rlimit1(l, b, f) \
1483     (l)->f = ((b)->f == RLIM_INFINITY || ((b)->f & 0xffffffff00000000) != 0) ? \
1484     LINUX_RLIM_INFINITY : (int32_t)(b)->f
1485 #define bsd_to_linux_rlimit(l, b) \
1486     bsd_to_linux_rlimit1(l, b, rlim_cur); \
1487     bsd_to_linux_rlimit1(l, b, rlim_max)
1488 
1489 #define linux_to_bsd_rlimit1(b, l, f) \
1490     (b)->f = (l)->f == LINUX_RLIM_INFINITY ? RLIM_INFINITY : (b)->f
1491 #define linux_to_bsd_rlimit(b, l) \
1492     linux_to_bsd_rlimit1(b, l, rlim_cur); \
1493     linux_to_bsd_rlimit1(b, l, rlim_max)
1494 
1495 static int
1496 linux_to_bsd_limit(lim)
1497 	int lim;
1498 {
1499 	switch (lim) {
1500 	case LINUX_RLIMIT_CPU:
1501 		return RLIMIT_CPU;
1502 	case LINUX_RLIMIT_FSIZE:
1503 		return RLIMIT_FSIZE;
1504 	case LINUX_RLIMIT_DATA:
1505 		return RLIMIT_DATA;
1506 	case LINUX_RLIMIT_STACK:
1507 		return RLIMIT_STACK;
1508 	case LINUX_RLIMIT_CORE:
1509 		return RLIMIT_CORE;
1510 	case LINUX_RLIMIT_RSS:
1511 		return RLIMIT_RSS;
1512 	case LINUX_RLIMIT_NPROC:
1513 		return RLIMIT_NPROC;
1514 	case LINUX_RLIMIT_NOFILE:
1515 		return RLIMIT_NOFILE;
1516 	case LINUX_RLIMIT_MEMLOCK:
1517 		return RLIMIT_MEMLOCK;
1518 	case LINUX_RLIMIT_AS:
1519 	case LINUX_RLIMIT_LOCKS:
1520 		return -EOPNOTSUPP;
1521 	default:
1522 		return -EINVAL;
1523 	}
1524 }
1525 
1526 
1527 int
1528 linux_sys_getrlimit(p, v, retval)
1529 	struct proc *p;
1530 	void *v;
1531 	register_t *retval;
1532 {
1533 	struct linux_sys_getrlimit_args /* {
1534 		syscallarg(int) which;
1535 		syscallarg(struct orlimit *) rlp;
1536 	} */ *uap = v;
1537 	caddr_t sg = stackgap_init(p->p_emul);
1538 	struct sys_getrlimit_args ap;
1539 	struct rlimit rl;
1540 	struct orlimit orl;
1541 	int error;
1542 
1543 	SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which));
1544 	if ((error = SCARG(&ap, which)) < 0)
1545 		return -error;
1546 	SCARG(&ap, rlp) = stackgap_alloc(&sg, sizeof rl);
1547 	if ((error = sys_getrlimit(p, &ap, retval)) != 0)
1548 		return error;
1549 	if ((error = copyin(SCARG(&ap, rlp), &rl, sizeof(rl))) != 0)
1550 		return error;
1551 	bsd_to_linux_rlimit(&orl, &rl);
1552 	return copyout(&orl, SCARG(uap, rlp), sizeof(orl));
1553 }
1554 
1555 int
1556 linux_sys_setrlimit(p, v, retval)
1557 	struct proc *p;
1558 	void *v;
1559 	register_t *retval;
1560 {
1561 	struct linux_sys_setrlimit_args /* {
1562 		syscallarg(int) which;
1563 		syscallarg(struct orlimit *) rlp;
1564 	} */ *uap = v;
1565 	caddr_t sg = stackgap_init(p->p_emul);
1566 	struct sys_setrlimit_args ap;
1567 	struct rlimit rl;
1568 	struct orlimit orl;
1569 	int error;
1570 
1571 	SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which));
1572 	SCARG(&ap, rlp) = stackgap_alloc(&sg, sizeof rl);
1573 	if ((error = SCARG(&ap, which)) < 0)
1574 		return -error;
1575 	if ((error = copyin(SCARG(uap, rlp), &orl, sizeof(orl))) != 0)
1576 		return error;
1577 	linux_to_bsd_rlimit(&rl, &orl);
1578 	/* XXX: alpha complains about this */
1579 	if ((error = copyout(&rl, (void *)SCARG(&ap, rlp), sizeof(rl))) != 0)
1580 		return error;
1581 	return sys_setrlimit(p, &ap, retval);
1582 }
1583 
1584 #ifndef __mips__
1585 /* XXX: this doesn't look 100% common, at least mips doesn't have it */
1586 int
1587 linux_sys_ugetrlimit(p, v, retval)
1588 	struct proc *p;
1589 	void *v;
1590 	register_t *retval;
1591 {
1592 	return linux_sys_getrlimit(p, v, retval);
1593 }
1594 #endif
1595 
1596 /*
1597  * This gets called for unsupported syscalls. The difference to sys_nosys()
1598  * is that process does not get SIGSYS, the call just returns with ENOSYS.
1599  * This is the way Linux does it and glibc depends on this behaviour.
1600  */
1601 int
1602 linux_sys_nosys(p, v, retval)
1603 	struct proc *p;
1604 	void *v;
1605 	register_t *retval;
1606 {
1607 	return (ENOSYS);
1608 }
1609