1 /*	$NetBSD: linux_misc.c,v 1.231 2015/03/14 08:32:08 njoly 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 1.231 2015/03/14 08:32:08 njoly 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/poll.h>
77 #include <sys/prot.h>
78 #include <sys/reboot.h>
79 #include <sys/resource.h>
80 #include <sys/resourcevar.h>
81 #include <sys/select.h>
82 #include <sys/signal.h>
83 #include <sys/signalvar.h>
84 #include <sys/socket.h>
85 #include <sys/time.h>
86 #include <sys/times.h>
87 #include <sys/vnode.h>
88 #include <sys/uio.h>
89 #include <sys/wait.h>
90 #include <sys/utsname.h>
91 #include <sys/unistd.h>
92 #include <sys/vfs_syscalls.h>
93 #include <sys/swap.h>		/* for SWAP_ON */
94 #include <sys/sysctl.h>		/* for KERN_DOMAINNAME */
95 #include <sys/kauth.h>
96 
97 #include <sys/ptrace.h>
98 #include <machine/ptrace.h>
99 
100 #include <sys/syscall.h>
101 #include <sys/syscallargs.h>
102 
103 #include <compat/sys/resource.h>
104 
105 #include <compat/linux/common/linux_machdep.h>
106 #include <compat/linux/common/linux_types.h>
107 #include <compat/linux/common/linux_signal.h>
108 #include <compat/linux/common/linux_ipc.h>
109 #include <compat/linux/common/linux_sem.h>
110 
111 #include <compat/linux/common/linux_fcntl.h>
112 #include <compat/linux/common/linux_mmap.h>
113 #include <compat/linux/common/linux_dirent.h>
114 #include <compat/linux/common/linux_util.h>
115 #include <compat/linux/common/linux_misc.h>
116 #include <compat/linux/common/linux_statfs.h>
117 #include <compat/linux/common/linux_limit.h>
118 #include <compat/linux/common/linux_ptrace.h>
119 #include <compat/linux/common/linux_reboot.h>
120 #include <compat/linux/common/linux_emuldata.h>
121 #include <compat/linux/common/linux_sched.h>
122 
123 #include <compat/linux/linux_syscallargs.h>
124 
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 	LINUX_PTRACE_SYSCALL,	PT_SYSCALL,
139 	-1
140 };
141 
142 const struct linux_mnttypes linux_fstypes[] = {
143 	{ MOUNT_FFS,		LINUX_DEFAULT_SUPER_MAGIC	},
144 	{ MOUNT_NFS,		LINUX_NFS_SUPER_MAGIC 		},
145 	{ MOUNT_MFS,		LINUX_DEFAULT_SUPER_MAGIC	},
146 	{ MOUNT_MSDOS,		LINUX_MSDOS_SUPER_MAGIC		},
147 	{ MOUNT_LFS,		LINUX_DEFAULT_SUPER_MAGIC	},
148 	{ MOUNT_FDESC,		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_TMPFS_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 static void linux_to_bsd_mmap_args(struct sys_mmap_args *,
177     const struct linux_sys_mmap_args *);
178 static int linux_mmap(struct lwp *, const struct linux_sys_mmap_args *,
179     register_t *, off_t);
180 
181 
182 /*
183  * The information on a terminated (or stopped) process needs
184  * to be converted in order for Linux binaries to get a valid signal
185  * number out of it.
186  */
187 int
bsd_to_linux_wstat(int st)188 bsd_to_linux_wstat(int st)
189 {
190 
191 	int sig;
192 
193 	if (WIFSIGNALED(st)) {
194 		sig = WTERMSIG(st);
195 		if (sig >= 0 && sig < NSIG)
196 			st= (st & ~0177) | native_to_linux_signo[sig];
197 	} else if (WIFSTOPPED(st)) {
198 		sig = WSTOPSIG(st);
199 		if (sig >= 0 && sig < NSIG)
200 			st = (st & ~0xff00) |
201 			    (native_to_linux_signo[sig] << 8);
202 	}
203 	return st;
204 }
205 
206 /*
207  * wait4(2).  Passed on to the NetBSD call, surrounded by code to
208  * reserve some space for a NetBSD-style wait status, and converting
209  * it to what Linux wants.
210  */
211 int
linux_sys_wait4(struct lwp * l,const struct linux_sys_wait4_args * uap,register_t * retval)212 linux_sys_wait4(struct lwp *l, const struct linux_sys_wait4_args *uap, register_t *retval)
213 {
214 	/* {
215 		syscallarg(int) pid;
216 		syscallarg(int *) status;
217 		syscallarg(int) options;
218 		syscallarg(struct rusage50 *) rusage;
219 	} */
220 	int error, status, options, linux_options, pid = SCARG(uap, pid);
221 	struct rusage50 ru50;
222 	struct rusage ru;
223 	proc_t *p;
224 
225 	linux_options = SCARG(uap, options);
226 	options = WOPTSCHECKED;
227 	if (linux_options & ~(LINUX_WAIT4_KNOWNFLAGS))
228 		return (EINVAL);
229 
230 	if (linux_options & LINUX_WAIT4_WNOHANG)
231 		options |= WNOHANG;
232 	if (linux_options & LINUX_WAIT4_WUNTRACED)
233 		options |= WUNTRACED;
234 	if (linux_options & LINUX_WAIT4_WALL)
235 		options |= WALLSIG;
236 	if (linux_options & LINUX_WAIT4_WCLONE)
237 		options |= WALTSIG;
238 # ifdef DIAGNOSTIC
239 	if (linux_options & LINUX_WAIT4_WNOTHREAD)
240 		printf("WARNING: %s: linux process %d.%d called "
241 		       "waitpid with __WNOTHREAD set!",
242 		       __FILE__, l->l_proc->p_pid, l->l_lid);
243 
244 # endif
245 
246 	error = do_sys_wait(&pid, &status, options,
247 	    SCARG(uap, rusage) != NULL ? &ru : NULL);
248 
249 	retval[0] = pid;
250 	if (pid == 0)
251 		return error;
252 
253 	p = curproc;
254 	mutex_enter(p->p_lock);
255 	sigdelset(&p->p_sigpend.sp_set, SIGCHLD); /* XXXAD ksiginfo leak */
256 	mutex_exit(p->p_lock);
257 
258 	if (SCARG(uap, rusage) != NULL) {
259 		rusage_to_rusage50(&ru, &ru50);
260 		error = copyout(&ru, SCARG(uap, rusage), sizeof(ru));
261 	}
262 
263 	if (error == 0 && SCARG(uap, status) != NULL) {
264 		status = bsd_to_linux_wstat(status);
265 		error = copyout(&status, SCARG(uap, status), sizeof status);
266 	}
267 
268 	return error;
269 }
270 
271 /*
272  * Linux brk(2).  Like native, but always return the new break value.
273  */
274 int
linux_sys_brk(struct lwp * l,const struct linux_sys_brk_args * uap,register_t * retval)275 linux_sys_brk(struct lwp *l, const struct linux_sys_brk_args *uap, register_t *retval)
276 {
277 	/* {
278 		syscallarg(char *) nsize;
279 	} */
280 	struct proc *p = l->l_proc;
281 	struct vmspace *vm = p->p_vmspace;
282 	struct sys_obreak_args oba;
283 
284 	SCARG(&oba, nsize) = SCARG(uap, nsize);
285 
286 	(void) sys_obreak(l, &oba, retval);
287 	retval[0] = (register_t)((char *)vm->vm_daddr + ptoa(vm->vm_dsize));
288 	return 0;
289 }
290 
291 /*
292  * Implement the fs stat functions. Straightforward.
293  */
294 int
linux_sys_statfs(struct lwp * l,const struct linux_sys_statfs_args * uap,register_t * retval)295 linux_sys_statfs(struct lwp *l, const struct linux_sys_statfs_args *uap, register_t *retval)
296 {
297 	/* {
298 		syscallarg(const char *) path;
299 		syscallarg(struct linux_statfs *) sp;
300 	} */
301 	struct statvfs *sb;
302 	struct linux_statfs ltmp;
303 	int error;
304 
305 	sb = STATVFSBUF_GET();
306 	error = do_sys_pstatvfs(l, SCARG(uap, path), ST_WAIT, sb);
307 	if (error == 0) {
308 		bsd_to_linux_statfs(sb, &ltmp);
309 		error = copyout(&ltmp, SCARG(uap, sp), sizeof ltmp);
310 	}
311 	STATVFSBUF_PUT(sb);
312 
313 	return error;
314 }
315 
316 int
linux_sys_fstatfs(struct lwp * l,const struct linux_sys_fstatfs_args * uap,register_t * retval)317 linux_sys_fstatfs(struct lwp *l, const struct linux_sys_fstatfs_args *uap, register_t *retval)
318 {
319 	/* {
320 		syscallarg(int) fd;
321 		syscallarg(struct linux_statfs *) sp;
322 	} */
323 	struct statvfs *sb;
324 	struct linux_statfs ltmp;
325 	int error;
326 
327 	sb = STATVFSBUF_GET();
328 	error = do_sys_fstatvfs(l, SCARG(uap, fd), ST_WAIT, sb);
329 	if (error == 0) {
330 		bsd_to_linux_statfs(sb, &ltmp);
331 		error = copyout(&ltmp, SCARG(uap, sp), sizeof ltmp);
332 	}
333 	STATVFSBUF_PUT(sb);
334 
335 	return error;
336 }
337 
338 /*
339  * uname(). Just copy the info from the various strings stored in the
340  * kernel, and put it in the Linux utsname structure. That structure
341  * is almost the same as the NetBSD one, only it has fields 65 characters
342  * long, and an extra domainname field.
343  */
344 int
linux_sys_uname(struct lwp * l,const struct linux_sys_uname_args * uap,register_t * retval)345 linux_sys_uname(struct lwp *l, const struct linux_sys_uname_args *uap, register_t *retval)
346 {
347 	/* {
348 		syscallarg(struct linux_utsname *) up;
349 	} */
350 	struct linux_utsname luts;
351 
352 	strlcpy(luts.l_sysname, linux_sysname, sizeof(luts.l_sysname));
353 	strlcpy(luts.l_nodename, hostname, sizeof(luts.l_nodename));
354 	strlcpy(luts.l_release, linux_release, sizeof(luts.l_release));
355 	strlcpy(luts.l_version, linux_version, sizeof(luts.l_version));
356 	strlcpy(luts.l_machine, LINUX_UNAME_ARCH, sizeof(luts.l_machine));
357 	strlcpy(luts.l_domainname, domainname, sizeof(luts.l_domainname));
358 
359 	return copyout(&luts, SCARG(uap, up), sizeof(luts));
360 }
361 
362 /* Used directly on: alpha, mips, ppc, sparc, sparc64 */
363 /* Used indirectly on: arm, i386, m68k */
364 
365 /*
366  * New type Linux mmap call.
367  * Only called directly on machines with >= 6 free regs.
368  */
369 int
linux_sys_mmap(struct lwp * l,const struct linux_sys_mmap_args * uap,register_t * retval)370 linux_sys_mmap(struct lwp *l, const struct linux_sys_mmap_args *uap, register_t *retval)
371 {
372 	/* {
373 		syscallarg(unsigned long) addr;
374 		syscallarg(size_t) len;
375 		syscallarg(int) prot;
376 		syscallarg(int) flags;
377 		syscallarg(int) fd;
378 		syscallarg(linux_off_t) offset;
379 	} */
380 
381 	if (SCARG(uap, offset) & PAGE_MASK)
382 		return EINVAL;
383 
384 	return linux_mmap(l, uap, retval, SCARG(uap, offset));
385 }
386 
387 /*
388  * Guts of most architectures' mmap64() implementations.  This shares
389  * its list of arguments with linux_sys_mmap().
390  *
391  * The difference in linux_sys_mmap2() is that "offset" is actually
392  * (offset / pagesize), not an absolute byte count.  This translation
393  * to pagesize offsets is done inside glibc between the mmap64() call
394  * point, and the actual syscall.
395  */
396 int
linux_sys_mmap2(struct lwp * l,const struct linux_sys_mmap2_args * uap,register_t * retval)397 linux_sys_mmap2(struct lwp *l, const struct linux_sys_mmap2_args *uap, register_t *retval)
398 {
399 	/* {
400 		syscallarg(unsigned long) addr;
401 		syscallarg(size_t) len;
402 		syscallarg(int) prot;
403 		syscallarg(int) flags;
404 		syscallarg(int) fd;
405 		syscallarg(linux_off_t) offset;
406 	} */
407 
408 	return linux_mmap(l, uap, retval,
409 	    ((off_t)SCARG(uap, offset)) << PAGE_SHIFT);
410 }
411 
412 /*
413  * Massage arguments and call system mmap(2).
414  */
415 static int
linux_mmap(struct lwp * l,const struct linux_sys_mmap_args * uap,register_t * retval,off_t offset)416 linux_mmap(struct lwp *l, const struct linux_sys_mmap_args *uap, register_t *retval, off_t offset)
417 {
418 	struct sys_mmap_args cma;
419 	int error;
420 	size_t mmoff=0;
421 
422 	linux_to_bsd_mmap_args(&cma, uap);
423 	SCARG(&cma, pos) = offset;
424 
425 	if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN) {
426 		/*
427 		 * Request for stack-like memory segment. On linux, this
428 		 * works by mmap()ping (small) segment, which is automatically
429 		 * extended when page fault happens below the currently
430 		 * allocated area. We emulate this by allocating (typically
431 		 * bigger) segment sized at current stack size limit, and
432 		 * offsetting the requested and returned address accordingly.
433 		 * Since physical pages are only allocated on-demand, this
434 		 * is effectively identical.
435 		 */
436 		rlim_t ssl = l->l_proc->p_rlimit[RLIMIT_STACK].rlim_cur;
437 
438 		if (SCARG(&cma, len) < ssl) {
439 			/* Compute the address offset */
440 			mmoff = round_page(ssl) - SCARG(uap, len);
441 
442 			if (SCARG(&cma, addr))
443 				SCARG(&cma, addr) = (char *)SCARG(&cma, addr) - mmoff;
444 
445 			SCARG(&cma, len) = (size_t) ssl;
446 		}
447 	}
448 
449 	error = sys_mmap(l, &cma, retval);
450 	if (error)
451 		return (error);
452 
453 	/* Shift the returned address for stack-like segment if necessary */
454 	retval[0] += mmoff;
455 
456 	return (0);
457 }
458 
459 static void
linux_to_bsd_mmap_args(struct sys_mmap_args * cma,const struct linux_sys_mmap_args * uap)460 linux_to_bsd_mmap_args(struct sys_mmap_args *cma, const struct linux_sys_mmap_args *uap)
461 {
462 	int flags = MAP_TRYFIXED, fl = SCARG(uap, flags);
463 
464 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_SHARED, MAP_SHARED);
465 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_PRIVATE, MAP_PRIVATE);
466 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_FIXED, MAP_FIXED);
467 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_ANON, MAP_ANON);
468 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_LOCKED, MAP_WIRED);
469 	/* XXX XAX ERH: Any other flags here?  There are more defined... */
470 
471 	SCARG(cma, addr) = (void *)SCARG(uap, addr);
472 	SCARG(cma, len) = SCARG(uap, len);
473 	SCARG(cma, prot) = SCARG(uap, prot);
474 	if (SCARG(cma, prot) & VM_PROT_WRITE) /* XXX */
475 		SCARG(cma, prot) |= VM_PROT_READ;
476 	SCARG(cma, flags) = flags;
477 	SCARG(cma, fd) = flags & MAP_ANON ? -1 : SCARG(uap, fd);
478 	SCARG(cma, PAD) = 0;
479 }
480 
481 #define	LINUX_MREMAP_MAYMOVE	1
482 #define	LINUX_MREMAP_FIXED	2
483 
484 int
linux_sys_mremap(struct lwp * l,const struct linux_sys_mremap_args * uap,register_t * retval)485 linux_sys_mremap(struct lwp *l, const struct linux_sys_mremap_args *uap, register_t *retval)
486 {
487 	/* {
488 		syscallarg(void *) old_address;
489 		syscallarg(size_t) old_size;
490 		syscallarg(size_t) new_size;
491 		syscallarg(u_long) flags;
492 	} */
493 
494 	struct proc *p;
495 	struct vm_map *map;
496 	vaddr_t oldva;
497 	vaddr_t newva;
498 	size_t oldsize;
499 	size_t newsize;
500 	int flags;
501 	int uvmflags;
502 	int error;
503 
504 	flags = SCARG(uap, flags);
505 	oldva = (vaddr_t)SCARG(uap, old_address);
506 	oldsize = round_page(SCARG(uap, old_size));
507 	newsize = round_page(SCARG(uap, new_size));
508 	if ((flags & ~(LINUX_MREMAP_FIXED|LINUX_MREMAP_MAYMOVE)) != 0) {
509 		error = EINVAL;
510 		goto done;
511 	}
512 	if ((flags & LINUX_MREMAP_FIXED) != 0) {
513 		if ((flags & LINUX_MREMAP_MAYMOVE) == 0) {
514 			error = EINVAL;
515 			goto done;
516 		}
517 #if 0 /* notyet */
518 		newva = SCARG(uap, new_address);
519 		uvmflags = MAP_FIXED;
520 #else /* notyet */
521 		error = EOPNOTSUPP;
522 		goto done;
523 #endif /* notyet */
524 	} else if ((flags & LINUX_MREMAP_MAYMOVE) != 0) {
525 		uvmflags = 0;
526 	} else {
527 		newva = oldva;
528 		uvmflags = MAP_FIXED;
529 	}
530 	p = l->l_proc;
531 	map = &p->p_vmspace->vm_map;
532 	error = uvm_mremap(map, oldva, oldsize, map, &newva, newsize, p,
533 	    uvmflags);
534 
535 done:
536 	*retval = (error != 0) ? 0 : (register_t)newva;
537 	return error;
538 }
539 
540 #ifdef USRSTACK
541 int
linux_sys_mprotect(struct lwp * l,const struct linux_sys_mprotect_args * uap,register_t * retval)542 linux_sys_mprotect(struct lwp *l, const struct linux_sys_mprotect_args *uap, register_t *retval)
543 {
544 	/* {
545 		syscallarg(const void *) start;
546 		syscallarg(unsigned long) len;
547 		syscallarg(int) prot;
548 	} */
549 	struct vm_map_entry *entry;
550 	struct vm_map *map;
551 	struct proc *p;
552 	vaddr_t end, start, len, stacklim;
553 	int prot, grows;
554 
555 	start = (vaddr_t)SCARG(uap, start);
556 	len = round_page(SCARG(uap, len));
557 	prot = SCARG(uap, prot);
558 	grows = prot & (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP);
559 	prot &= ~grows;
560 	end = start + len;
561 
562 	if (start & PAGE_MASK)
563 		return EINVAL;
564 	if (end < start)
565 		return EINVAL;
566 	if (end == start)
567 		return 0;
568 
569 	if (prot & ~(PROT_READ | PROT_WRITE | PROT_EXEC))
570 		return EINVAL;
571 	if (grows == (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP))
572 		return EINVAL;
573 
574 	p = l->l_proc;
575 	map = &p->p_vmspace->vm_map;
576 	vm_map_lock(map);
577 # ifdef notdef
578 	VM_MAP_RANGE_CHECK(map, start, end);
579 # endif
580 	if (!uvm_map_lookup_entry(map, start, &entry) || entry->start > start) {
581 		vm_map_unlock(map);
582 		return ENOMEM;
583 	}
584 
585 	/*
586 	 * Approximate the behaviour of PROT_GROWS{DOWN,UP}.
587 	 */
588 
589 	stacklim = (vaddr_t)p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur;
590 	if (grows & LINUX_PROT_GROWSDOWN) {
591 		if (USRSTACK - stacklim <= start && start < USRSTACK) {
592 			start = USRSTACK - stacklim;
593 		} else {
594 			start = entry->start;
595 		}
596 	} else if (grows & LINUX_PROT_GROWSUP) {
597 		if (USRSTACK <= end && end < USRSTACK + stacklim) {
598 			end = USRSTACK + stacklim;
599 		} else {
600 			end = entry->end;
601 		}
602 	}
603 	vm_map_unlock(map);
604 	return uvm_map_protect(map, start, end, prot, FALSE);
605 }
606 #endif /* USRSTACK */
607 
608 /*
609  * This code is partly stolen from src/lib/libc/compat-43/times.c
610  */
611 
612 #define	CONVTCK(r)	(r.tv_sec * hz + r.tv_usec / (1000000 / hz))
613 
614 int
linux_sys_times(struct lwp * l,const struct linux_sys_times_args * uap,register_t * retval)615 linux_sys_times(struct lwp *l, const struct linux_sys_times_args *uap, register_t *retval)
616 {
617 	/* {
618 		syscallarg(struct times *) tms;
619 	} */
620 	struct proc *p = l->l_proc;
621 	struct timeval t;
622 	int error;
623 
624 	if (SCARG(uap, tms)) {
625 		struct linux_tms ltms;
626 		struct rusage ru;
627 
628 		mutex_enter(p->p_lock);
629 		calcru(p, &ru.ru_utime, &ru.ru_stime, NULL, NULL);
630 		ltms.ltms_utime = CONVTCK(ru.ru_utime);
631 		ltms.ltms_stime = CONVTCK(ru.ru_stime);
632 		ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime);
633 		ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime);
634 		mutex_exit(p->p_lock);
635 
636 		if ((error = copyout(&ltms, SCARG(uap, tms), sizeof ltms)))
637 			return error;
638 	}
639 
640 	getmicrouptime(&t);
641 
642 	retval[0] = ((linux_clock_t)(CONVTCK(t)));
643 	return 0;
644 }
645 
646 #undef CONVTCK
647 
648 /*
649  * Linux 'readdir' call. This code is mostly taken from the
650  * SunOS getdents call (see compat/sunos/sunos_misc.c), though
651  * an attempt has been made to keep it a little cleaner (failing
652  * miserably, because of the cruft needed if count 1 is passed).
653  *
654  * The d_off field should contain the offset of the next valid entry,
655  * but in Linux it has the offset of the entry itself. We emulate
656  * that bug here.
657  *
658  * Read in BSD-style entries, convert them, and copy them out.
659  *
660  * Note that this doesn't handle union-mounted filesystems.
661  */
662 int
linux_sys_getdents(struct lwp * l,const struct linux_sys_getdents_args * uap,register_t * retval)663 linux_sys_getdents(struct lwp *l, const struct linux_sys_getdents_args *uap, register_t *retval)
664 {
665 	/* {
666 		syscallarg(int) fd;
667 		syscallarg(struct linux_dirent *) dent;
668 		syscallarg(unsigned int) count;
669 	} */
670 	struct dirent *bdp;
671 	struct vnode *vp;
672 	char *inp, *tbuf;		/* BSD-format */
673 	int len, reclen;		/* BSD-format */
674 	char *outp;			/* Linux-format */
675 	int resid, linux_reclen = 0;	/* Linux-format */
676 	struct file *fp;
677 	struct uio auio;
678 	struct iovec aiov;
679 	struct linux_dirent idb;
680 	off_t off;		/* true file offset */
681 	int buflen, error, eofflag, nbytes, oldcall;
682 	struct vattr va;
683 	off_t *cookiebuf = NULL, *cookie;
684 	int ncookies;
685 
686 	/* fd_getvnode() will use the descriptor for us */
687 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
688 		return (error);
689 
690 	if ((fp->f_flag & FREAD) == 0) {
691 		error = EBADF;
692 		goto out1;
693 	}
694 
695 	vp = (struct vnode *)fp->f_data;
696 	if (vp->v_type != VDIR) {
697 		error = ENOTDIR;
698 		goto out1;
699 	}
700 
701 	vn_lock(vp, LK_SHARED | LK_RETRY);
702 	error = VOP_GETATTR(vp, &va, l->l_cred);
703 	VOP_UNLOCK(vp);
704 	if (error)
705 		goto out1;
706 
707 	nbytes = SCARG(uap, count);
708 	if (nbytes == 1) {	/* emulating old, broken behaviour */
709 		nbytes = sizeof (idb);
710 		buflen = max(va.va_blocksize, nbytes);
711 		oldcall = 1;
712 	} else {
713 		buflen = min(MAXBSIZE, nbytes);
714 		if (buflen < va.va_blocksize)
715 			buflen = va.va_blocksize;
716 		oldcall = 0;
717 	}
718 	tbuf = malloc(buflen, M_TEMP, M_WAITOK);
719 
720 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
721 	off = fp->f_offset;
722 again:
723 	aiov.iov_base = tbuf;
724 	aiov.iov_len = buflen;
725 	auio.uio_iov = &aiov;
726 	auio.uio_iovcnt = 1;
727 	auio.uio_rw = UIO_READ;
728 	auio.uio_resid = buflen;
729 	auio.uio_offset = off;
730 	UIO_SETUP_SYSSPACE(&auio);
731 	/*
732          * First we read into the malloc'ed buffer, then
733          * we massage it into user space, one record at a time.
734          */
735 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf,
736 	    &ncookies);
737 	if (error)
738 		goto out;
739 
740 	inp = tbuf;
741 	outp = (void *)SCARG(uap, dent);
742 	resid = nbytes;
743 	if ((len = buflen - auio.uio_resid) == 0)
744 		goto eof;
745 
746 	for (cookie = cookiebuf; len > 0; len -= reclen) {
747 		bdp = (struct dirent *)inp;
748 		reclen = bdp->d_reclen;
749 		if (reclen & 3)
750 			panic("linux_readdir");
751 		if (bdp->d_fileno == 0) {
752 			inp += reclen;	/* it is a hole; squish it out */
753 			if (cookie)
754 				off = *cookie++;
755 			else
756 				off += reclen;
757 			continue;
758 		}
759 		linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen);
760 		if (reclen > len || resid < linux_reclen) {
761 			/* entry too big for buffer, so just stop */
762 			outp++;
763 			break;
764 		}
765 		/*
766 		 * Massage in place to make a Linux-shaped dirent (otherwise
767 		 * we have to worry about touching user memory outside of
768 		 * the copyout() call).
769 		 */
770 		idb.d_ino = bdp->d_fileno;
771 		/*
772 		 * The old readdir() call misuses the offset and reclen fields.
773 		 */
774 		if (oldcall) {
775 			idb.d_off = (linux_off_t)linux_reclen;
776 			idb.d_reclen = (u_short)bdp->d_namlen;
777 		} else {
778 			if (sizeof (idb.d_off) <= 4 && (off >> 32) != 0) {
779 				compat_offseterr(vp, "linux_getdents");
780 				error = EINVAL;
781 				goto out;
782 			}
783 			idb.d_off = (linux_off_t)off;
784 			idb.d_reclen = (u_short)linux_reclen;
785 			/* Linux puts d_type at the end of each record */
786 			*((char *)&idb + idb.d_reclen - 1) = bdp->d_type;
787 		}
788 		strcpy(idb.d_name, bdp->d_name);
789 		if ((error = copyout((void *)&idb, outp, linux_reclen)))
790 			goto out;
791 		/* advance past this real entry */
792 		inp += reclen;
793 		if (cookie)
794 			off = *cookie++; /* each entry points to itself */
795 		else
796 			off += reclen;
797 		/* advance output past Linux-shaped entry */
798 		outp += linux_reclen;
799 		resid -= linux_reclen;
800 		if (oldcall)
801 			break;
802 	}
803 
804 	/* if we squished out the whole block, try again */
805 	if (outp == (void *)SCARG(uap, dent)) {
806 		if (cookiebuf)
807 			free(cookiebuf, M_TEMP);
808 		cookiebuf = NULL;
809 		goto again;
810 	}
811 	fp->f_offset = off;	/* update the vnode offset */
812 
813 	if (oldcall)
814 		nbytes = resid + linux_reclen;
815 
816 eof:
817 	*retval = nbytes - resid;
818 out:
819 	VOP_UNLOCK(vp);
820 	if (cookiebuf)
821 		free(cookiebuf, M_TEMP);
822 	free(tbuf, M_TEMP);
823 out1:
824 	fd_putfile(SCARG(uap, fd));
825 	return error;
826 }
827 
828 /*
829  * Even when just using registers to pass arguments to syscalls you can
830  * have 5 of them on the i386. So this newer version of select() does
831  * this.
832  */
833 int
linux_sys_select(struct lwp * l,const struct linux_sys_select_args * uap,register_t * retval)834 linux_sys_select(struct lwp *l, const struct linux_sys_select_args *uap, register_t *retval)
835 {
836 	/* {
837 		syscallarg(int) nfds;
838 		syscallarg(fd_set *) readfds;
839 		syscallarg(fd_set *) writefds;
840 		syscallarg(fd_set *) exceptfds;
841 		syscallarg(struct timeval50 *) timeout;
842 	} */
843 
844 	return linux_select1(l, retval, SCARG(uap, nfds), SCARG(uap, readfds),
845 	    SCARG(uap, writefds), SCARG(uap, exceptfds),
846 	    (struct linux_timeval *)SCARG(uap, timeout));
847 }
848 
849 /*
850  * Common code for the old and new versions of select(). A couple of
851  * things are important:
852  * 1) return the amount of time left in the 'timeout' parameter
853  * 2) select never returns ERESTART on Linux, always return EINTR
854  */
855 int
linux_select1(struct lwp * l,register_t * retval,int nfds,fd_set * readfds,fd_set * writefds,fd_set * exceptfds,struct linux_timeval * timeout)856 linux_select1(struct lwp *l, register_t *retval, int nfds, fd_set *readfds,
857     fd_set *writefds, fd_set *exceptfds, struct linux_timeval *timeout)
858 {
859 	struct timespec ts0, ts1, uts, *ts = NULL;
860 	struct linux_timeval ltv;
861 	int error;
862 
863 	/*
864 	 * Store current time for computation of the amount of
865 	 * time left.
866 	 */
867 	if (timeout) {
868 		if ((error = copyin(timeout, &ltv, sizeof(ltv))))
869 			return error;
870 		uts.tv_sec = ltv.tv_sec;
871 		uts.tv_nsec = ltv.tv_usec * 1000;
872 		if (itimespecfix(&uts)) {
873 			/*
874 			 * The timeval was invalid.  Convert it to something
875 			 * valid that will act as it does under Linux.
876 			 */
877 			uts.tv_sec += uts.tv_nsec / 1000000000;
878 			uts.tv_nsec %= 1000000000;
879 			if (uts.tv_nsec < 0) {
880 				uts.tv_sec -= 1;
881 				uts.tv_nsec += 1000000000;
882 			}
883 			if (uts.tv_sec < 0)
884 				timespecclear(&uts);
885 		}
886 		ts = &uts;
887 		nanotime(&ts0);
888 	}
889 
890 	error = selcommon(retval, nfds, readfds, writefds, exceptfds, ts, NULL);
891 
892 	if (error) {
893 		/*
894 		 * See fs/select.c in the Linux kernel.  Without this,
895 		 * Maelstrom doesn't work.
896 		 */
897 		if (error == ERESTART)
898 			error = EINTR;
899 		return error;
900 	}
901 
902 	if (timeout) {
903 		if (*retval) {
904 			/*
905 			 * Compute how much time was left of the timeout,
906 			 * by subtracting the current time and the time
907 			 * before we started the call, and subtracting
908 			 * that result from the user-supplied value.
909 			 */
910 			nanotime(&ts1);
911 			timespecsub(&ts1, &ts0, &ts1);
912 			timespecsub(&uts, &ts1, &uts);
913 			if (uts.tv_sec < 0)
914 				timespecclear(&uts);
915 		} else
916 			timespecclear(&uts);
917 		ltv.tv_sec = uts.tv_sec;
918 		ltv.tv_usec = uts.tv_nsec / 1000;
919 		if ((error = copyout(&ltv, timeout, sizeof(ltv))))
920 			return error;
921 	}
922 
923 	return 0;
924 }
925 
926 int
linux_sys_ppoll(struct lwp * l,const struct linux_sys_ppoll_args * uap,register_t * retval)927 linux_sys_ppoll(struct lwp *l,
928 	const struct linux_sys_ppoll_args *uap, register_t *retval)
929 {
930 	/* {
931 		syscallarg(struct pollfd *) fds;
932 		syscallarg(u_int) nfds;
933 		syscallarg(struct linux_timespec *) timeout;
934 		syscallarg(linux_sigset_t *) sigset;
935 	} */
936 	struct linux_timespec lts0, *lts;
937 	struct timespec ts0, *ts = NULL;
938 	linux_sigset_t lsigmask0, *lsigmask;
939 	sigset_t sigmask0, *sigmask = NULL;
940 	int error;
941 
942 	lts = SCARG(uap, timeout);
943 	if (lts) {
944 		if ((error = copyin(lts, &lts0, sizeof(lts0))) != 0)
945 			return error;
946 		linux_to_native_timespec(&ts0, &lts0);
947 		ts = &ts0;
948 	}
949 
950 	lsigmask = SCARG(uap, sigset);
951 	if (lsigmask) {
952 		if ((error = copyin(lsigmask, &lsigmask0, sizeof(lsigmask0))))
953 			return error;
954 		linux_to_native_sigset(&sigmask0, &lsigmask0);
955 		sigmask = &sigmask0;
956 	}
957 
958 	return pollcommon(retval, SCARG(uap, fds), SCARG(uap, nfds),
959 	    ts, sigmask);
960 }
961 
962 /*
963  * Set the 'personality' (emulation mode) for the current process. Only
964  * accept the Linux personality here (0). This call is needed because
965  * the Linux ELF crt0 issues it in an ugly kludge to make sure that
966  * ELF binaries run in Linux mode, not SVR4 mode.
967  */
968 int
linux_sys_personality(struct lwp * l,const struct linux_sys_personality_args * uap,register_t * retval)969 linux_sys_personality(struct lwp *l, const struct linux_sys_personality_args *uap, register_t *retval)
970 {
971 	/* {
972 		syscallarg(unsigned long) per;
973 	} */
974 	struct linux_emuldata *led;
975 	int per;
976 
977 	per = SCARG(uap, per);
978 	led = l->l_emuldata;
979 	if (per == LINUX_PER_QUERY) {
980 		retval[0] = led->led_personality;
981 		return 0;
982 	}
983 
984 	switch (per & LINUX_PER_MASK) {
985 	case LINUX_PER_LINUX:
986 	case LINUX_PER_LINUX32:
987 		led->led_personality = per;
988 		break;
989 
990 	default:
991 		return EINVAL;
992 	}
993 
994 	retval[0] = per;
995 	return 0;
996 }
997 
998 /*
999  * We have nonexistent fsuid equal to uid.
1000  * If modification is requested, refuse.
1001  */
1002 int
linux_sys_setfsuid(struct lwp * l,const struct linux_sys_setfsuid_args * uap,register_t * retval)1003 linux_sys_setfsuid(struct lwp *l, const struct linux_sys_setfsuid_args *uap, register_t *retval)
1004 {
1005 	 /* {
1006 		 syscallarg(uid_t) uid;
1007 	 } */
1008 	 uid_t uid;
1009 
1010 	 uid = SCARG(uap, uid);
1011 	 if (kauth_cred_getuid(l->l_cred) != uid)
1012 		 return sys_nosys(l, uap, retval);
1013 
1014 	 *retval = uid;
1015 	 return 0;
1016 }
1017 
1018 int
linux_sys_setfsgid(struct lwp * l,const struct linux_sys_setfsgid_args * uap,register_t * retval)1019 linux_sys_setfsgid(struct lwp *l, const struct linux_sys_setfsgid_args *uap, register_t *retval)
1020 {
1021 	/* {
1022 		syscallarg(gid_t) gid;
1023 	} */
1024 	gid_t gid;
1025 
1026 	gid = SCARG(uap, gid);
1027 	if (kauth_cred_getgid(l->l_cred) != gid)
1028 		return sys_nosys(l, uap, retval);
1029 
1030 	*retval = gid;
1031 	return 0;
1032 }
1033 
1034 int
linux_sys_setresuid(struct lwp * l,const struct linux_sys_setresuid_args * uap,register_t * retval)1035 linux_sys_setresuid(struct lwp *l, const struct linux_sys_setresuid_args *uap, register_t *retval)
1036 {
1037 	/* {
1038 		syscallarg(uid_t) ruid;
1039 		syscallarg(uid_t) euid;
1040 		syscallarg(uid_t) suid;
1041 	} */
1042 
1043 	/*
1044 	 * Note: These checks are a little different than the NetBSD
1045 	 * setreuid(2) call performs.  This precisely follows the
1046 	 * behavior of the Linux kernel.
1047 	 */
1048 
1049 	return do_setresuid(l, SCARG(uap, ruid), SCARG(uap, euid),
1050 			    SCARG(uap, suid),
1051 			    ID_R_EQ_R | ID_R_EQ_E | ID_R_EQ_S |
1052 			    ID_E_EQ_R | ID_E_EQ_E | ID_E_EQ_S |
1053 			    ID_S_EQ_R | ID_S_EQ_E | ID_S_EQ_S );
1054 }
1055 
1056 int
linux_sys_getresuid(struct lwp * l,const struct linux_sys_getresuid_args * uap,register_t * retval)1057 linux_sys_getresuid(struct lwp *l, const struct linux_sys_getresuid_args *uap, register_t *retval)
1058 {
1059 	/* {
1060 		syscallarg(uid_t *) ruid;
1061 		syscallarg(uid_t *) euid;
1062 		syscallarg(uid_t *) suid;
1063 	} */
1064 	kauth_cred_t pc = l->l_cred;
1065 	int error;
1066 	uid_t uid;
1067 
1068 	/*
1069 	 * Linux copies these values out to userspace like so:
1070 	 *
1071 	 *	1. Copy out ruid.
1072 	 *	2. If that succeeds, copy out euid.
1073 	 *	3. If both of those succeed, copy out suid.
1074 	 */
1075 	uid = kauth_cred_getuid(pc);
1076 	if ((error = copyout(&uid, SCARG(uap, ruid), sizeof(uid_t))) != 0)
1077 		return (error);
1078 
1079 	uid = kauth_cred_geteuid(pc);
1080 	if ((error = copyout(&uid, SCARG(uap, euid), sizeof(uid_t))) != 0)
1081 		return (error);
1082 
1083 	uid = kauth_cred_getsvuid(pc);
1084 
1085 	return (copyout(&uid, SCARG(uap, suid), sizeof(uid_t)));
1086 }
1087 
1088 int
linux_sys_ptrace(struct lwp * l,const struct linux_sys_ptrace_args * uap,register_t * retval)1089 linux_sys_ptrace(struct lwp *l, const struct linux_sys_ptrace_args *uap, register_t *retval)
1090 {
1091 	/* {
1092 		i386, m68k, powerpc: T=int
1093 		alpha, amd64: T=long
1094 		syscallarg(T) request;
1095 		syscallarg(T) pid;
1096 		syscallarg(T) addr;
1097 		syscallarg(T) data;
1098 	} */
1099 	const int *ptr;
1100 	int request;
1101 	int error;
1102 
1103 	ptr = linux_ptrace_request_map;
1104 	request = SCARG(uap, request);
1105 	while (*ptr != -1)
1106 		if (*ptr++ == request) {
1107 			struct sys_ptrace_args pta;
1108 
1109 			SCARG(&pta, req) = *ptr;
1110 			SCARG(&pta, pid) = SCARG(uap, pid);
1111 			SCARG(&pta, addr) = (void *)SCARG(uap, addr);
1112 			SCARG(&pta, data) = SCARG(uap, data);
1113 
1114 			/*
1115 			 * Linux ptrace(PTRACE_CONT, pid, 0, 0) means actually
1116 			 * to continue where the process left off previously.
1117  			 * The same thing is achieved by addr == (void *) 1
1118 			 * on NetBSD, so rewrite 'addr' appropriately.
1119 			 */
1120 			if (request == LINUX_PTRACE_CONT && SCARG(uap, addr)==0)
1121 				SCARG(&pta, addr) = (void *) 1;
1122 
1123 			error = sysent[SYS_ptrace].sy_call(l, &pta, retval);
1124 			if (error)
1125 				return error;
1126 			switch (request) {
1127 			case LINUX_PTRACE_PEEKTEXT:
1128 			case LINUX_PTRACE_PEEKDATA:
1129 				error = copyout (retval,
1130 				    (void *)SCARG(uap, data),
1131 				    sizeof *retval);
1132 				*retval = SCARG(uap, data);
1133 				break;
1134 			default:
1135 				break;
1136 			}
1137 			return error;
1138 		}
1139 		else
1140 			ptr++;
1141 
1142 	return LINUX_SYS_PTRACE_ARCH(l, uap, retval);
1143 }
1144 
1145 int
linux_sys_reboot(struct lwp * l,const struct linux_sys_reboot_args * uap,register_t * retval)1146 linux_sys_reboot(struct lwp *l, const struct linux_sys_reboot_args *uap, register_t *retval)
1147 {
1148 	/* {
1149 		syscallarg(int) magic1;
1150 		syscallarg(int) magic2;
1151 		syscallarg(int) cmd;
1152 		syscallarg(void *) arg;
1153 	} */
1154 	struct sys_reboot_args /* {
1155 		syscallarg(int) opt;
1156 		syscallarg(char *) bootstr;
1157 	} */ sra;
1158 	int error;
1159 
1160 	if ((error = kauth_authorize_system(l->l_cred,
1161 	    KAUTH_SYSTEM_REBOOT, 0, NULL, NULL, NULL)) != 0)
1162 		return(error);
1163 
1164 	if (SCARG(uap, magic1) != LINUX_REBOOT_MAGIC1)
1165 		return(EINVAL);
1166 	if (SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2 &&
1167 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2A &&
1168 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2B)
1169 		return(EINVAL);
1170 
1171 	switch ((unsigned long)SCARG(uap, cmd)) {
1172 	case LINUX_REBOOT_CMD_RESTART:
1173 		SCARG(&sra, opt) = RB_AUTOBOOT;
1174 		break;
1175 	case LINUX_REBOOT_CMD_HALT:
1176 		SCARG(&sra, opt) = RB_HALT;
1177 		break;
1178 	case LINUX_REBOOT_CMD_POWER_OFF:
1179 		SCARG(&sra, opt) = RB_HALT|RB_POWERDOWN;
1180 		break;
1181 	case LINUX_REBOOT_CMD_RESTART2:
1182 		/* Reboot with an argument. */
1183 		SCARG(&sra, opt) = RB_AUTOBOOT|RB_STRING;
1184 		SCARG(&sra, bootstr) = SCARG(uap, arg);
1185 		break;
1186 	case LINUX_REBOOT_CMD_CAD_ON:
1187 		return(EINVAL);	/* We don't implement ctrl-alt-delete */
1188 	case LINUX_REBOOT_CMD_CAD_OFF:
1189 		return(0);
1190 	default:
1191 		return(EINVAL);
1192 	}
1193 
1194 	return(sys_reboot(l, &sra, retval));
1195 }
1196 
1197 /*
1198  * Copy of compat_12_sys_swapon().
1199  */
1200 int
linux_sys_swapon(struct lwp * l,const struct linux_sys_swapon_args * uap,register_t * retval)1201 linux_sys_swapon(struct lwp *l, const struct linux_sys_swapon_args *uap, register_t *retval)
1202 {
1203 	/* {
1204 		syscallarg(const char *) name;
1205 	} */
1206 	struct sys_swapctl_args ua;
1207 
1208 	SCARG(&ua, cmd) = SWAP_ON;
1209 	SCARG(&ua, arg) = (void *)__UNCONST(SCARG(uap, name));
1210 	SCARG(&ua, misc) = 0;	/* priority */
1211 	return (sys_swapctl(l, &ua, retval));
1212 }
1213 
1214 /*
1215  * Stop swapping to the file or block device specified by path.
1216  */
1217 int
linux_sys_swapoff(struct lwp * l,const struct linux_sys_swapoff_args * uap,register_t * retval)1218 linux_sys_swapoff(struct lwp *l, const struct linux_sys_swapoff_args *uap, register_t *retval)
1219 {
1220 	/* {
1221 		syscallarg(const char *) path;
1222 	} */
1223 	struct sys_swapctl_args ua;
1224 
1225 	SCARG(&ua, cmd) = SWAP_OFF;
1226 	SCARG(&ua, arg) = __UNCONST(SCARG(uap, path)); /*XXXUNCONST*/
1227 	return (sys_swapctl(l, &ua, retval));
1228 }
1229 
1230 /*
1231  * Copy of compat_09_sys_setdomainname()
1232  */
1233 /* ARGSUSED */
1234 int
linux_sys_setdomainname(struct lwp * l,const struct linux_sys_setdomainname_args * uap,register_t * retval)1235 linux_sys_setdomainname(struct lwp *l, const struct linux_sys_setdomainname_args *uap, register_t *retval)
1236 {
1237 	/* {
1238 		syscallarg(char *) domainname;
1239 		syscallarg(int) len;
1240 	} */
1241 	int name[2];
1242 
1243 	name[0] = CTL_KERN;
1244 	name[1] = KERN_DOMAINNAME;
1245 	return (old_sysctl(&name[0], 2, 0, 0, SCARG(uap, domainname),
1246 			    SCARG(uap, len), l));
1247 }
1248 
1249 /*
1250  * sysinfo()
1251  */
1252 /* ARGSUSED */
1253 int
linux_sys_sysinfo(struct lwp * l,const struct linux_sys_sysinfo_args * uap,register_t * retval)1254 linux_sys_sysinfo(struct lwp *l, const struct linux_sys_sysinfo_args *uap, register_t *retval)
1255 {
1256 	/* {
1257 		syscallarg(struct linux_sysinfo *) arg;
1258 	} */
1259 	struct linux_sysinfo si;
1260 	struct loadavg *la;
1261 
1262 	si.uptime = time_uptime;
1263 	la = &averunnable;
1264 	si.loads[0] = la->ldavg[0] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1265 	si.loads[1] = la->ldavg[1] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1266 	si.loads[2] = la->ldavg[2] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1267 	si.totalram = ctob((u_long)physmem);
1268 	si.freeram = (u_long)uvmexp.free * uvmexp.pagesize;
1269 	si.sharedram = 0;	/* XXX */
1270 	si.bufferram = (u_long)uvmexp.filepages * uvmexp.pagesize;
1271 	si.totalswap = (u_long)uvmexp.swpages * uvmexp.pagesize;
1272 	si.freeswap =
1273 	    (u_long)(uvmexp.swpages - uvmexp.swpginuse) * uvmexp.pagesize;
1274 	si.procs = nprocs;
1275 
1276 	/* The following are only present in newer Linux kernels. */
1277 	si.totalbig = 0;
1278 	si.freebig = 0;
1279 	si.mem_unit = 1;
1280 
1281 	return (copyout(&si, SCARG(uap, arg), sizeof si));
1282 }
1283 
1284 int
linux_sys_getrlimit(struct lwp * l,const struct linux_sys_getrlimit_args * uap,register_t * retval)1285 linux_sys_getrlimit(struct lwp *l, const struct linux_sys_getrlimit_args *uap, register_t *retval)
1286 {
1287 	/* {
1288 		syscallarg(int) which;
1289 # ifdef LINUX_LARGEFILE64
1290 		syscallarg(struct rlimit *) rlp;
1291 # else
1292 		syscallarg(struct orlimit *) rlp;
1293 # endif
1294 	} */
1295 # ifdef LINUX_LARGEFILE64
1296 	struct rlimit orl;
1297 # else
1298 	struct orlimit orl;
1299 # endif
1300 	int which;
1301 
1302 	which = linux_to_bsd_limit(SCARG(uap, which));
1303 	if (which < 0)
1304 		return -which;
1305 
1306 	bsd_to_linux_rlimit(&orl, &l->l_proc->p_rlimit[which]);
1307 
1308 	return copyout(&orl, SCARG(uap, rlp), sizeof(orl));
1309 }
1310 
1311 int
linux_sys_setrlimit(struct lwp * l,const struct linux_sys_setrlimit_args * uap,register_t * retval)1312 linux_sys_setrlimit(struct lwp *l, const struct linux_sys_setrlimit_args *uap, register_t *retval)
1313 {
1314 	/* {
1315 		syscallarg(int) which;
1316 # ifdef LINUX_LARGEFILE64
1317 		syscallarg(struct rlimit *) rlp;
1318 # else
1319 		syscallarg(struct orlimit *) rlp;
1320 # endif
1321 	} */
1322 	struct rlimit rl;
1323 # ifdef LINUX_LARGEFILE64
1324 	struct rlimit orl;
1325 # else
1326 	struct orlimit orl;
1327 # endif
1328 	int error;
1329 	int which;
1330 
1331 	if ((error = copyin(SCARG(uap, rlp), &orl, sizeof(orl))) != 0)
1332 		return error;
1333 
1334 	which = linux_to_bsd_limit(SCARG(uap, which));
1335 	if (which < 0)
1336 		return -which;
1337 
1338 	linux_to_bsd_rlimit(&rl, &orl);
1339 	return dosetrlimit(l, l->l_proc, which, &rl);
1340 }
1341 
1342 # if !defined(__mips__) && !defined(__amd64__)
1343 /* XXX: this doesn't look 100% common, at least mips doesn't have it */
1344 int
linux_sys_ugetrlimit(struct lwp * l,const struct linux_sys_ugetrlimit_args * uap,register_t * retval)1345 linux_sys_ugetrlimit(struct lwp *l, const struct linux_sys_ugetrlimit_args *uap, register_t *retval)
1346 {
1347 	return linux_sys_getrlimit(l, (const void *)uap, retval);
1348 }
1349 # endif
1350 
1351 /*
1352  * This gets called for unsupported syscalls. The difference to sys_nosys()
1353  * is that process does not get SIGSYS, the call just returns with ENOSYS.
1354  * This is the way Linux does it and glibc depends on this behaviour.
1355  */
1356 int
linux_sys_nosys(struct lwp * l,const void * v,register_t * retval)1357 linux_sys_nosys(struct lwp *l, const void *v, register_t *retval)
1358 {
1359 	return (ENOSYS);
1360 }
1361 
1362 int
linux_sys_getpriority(struct lwp * l,const struct linux_sys_getpriority_args * uap,register_t * retval)1363 linux_sys_getpriority(struct lwp *l, const struct linux_sys_getpriority_args *uap, register_t *retval)
1364 {
1365         /* {
1366                 syscallarg(int) which;
1367                 syscallarg(int) who;
1368         } */
1369         struct sys_getpriority_args bsa;
1370         int error;
1371 
1372         SCARG(&bsa, which) = SCARG(uap, which);
1373         SCARG(&bsa, who) = SCARG(uap, who);
1374 
1375         if ((error = sys_getpriority(l, &bsa, retval)))
1376                 return error;
1377 
1378         *retval = NZERO - *retval;
1379 
1380         return 0;
1381 }
1382 
1383 int
linux_do_sys_utimensat(struct lwp * l,int fd,const char * path,struct timespec * tsp,int flags,register_t * retval)1384 linux_do_sys_utimensat(struct lwp *l, int fd, const char *path, struct timespec *tsp, int flags, register_t *retval)
1385 {
1386 	int follow, error;
1387 
1388 	follow = (flags & LINUX_AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
1389 
1390 	if (path == NULL && fd != AT_FDCWD) {
1391 		file_t *fp;
1392 
1393 		/* fd_getvnode() will use the descriptor for us */
1394 		if ((error = fd_getvnode(fd, &fp)) != 0)
1395 			return error;
1396 		error = do_sys_utimensat(l, AT_FDCWD, fp->f_data, NULL, 0,
1397 		    tsp, UIO_SYSSPACE);
1398 		fd_putfile(fd);
1399 		return error;
1400 	}
1401 
1402 	return do_sys_utimensat(l, fd, NULL, path, follow, tsp, UIO_SYSSPACE);
1403 }
1404 
1405 int
linux_sys_utimensat(struct lwp * l,const struct linux_sys_utimensat_args * uap,register_t * retval)1406 linux_sys_utimensat(struct lwp *l, const struct linux_sys_utimensat_args *uap,
1407 	register_t *retval)
1408 {
1409 	/* {
1410 		syscallarg(int) fd;
1411 		syscallarg(const char *) path;
1412 		syscallarg(const struct linux_timespec *) times;
1413 		syscallarg(int) flag;
1414 	} */
1415 	int error;
1416 	struct linux_timespec lts[2];
1417 	struct timespec *tsp = NULL, ts[2];
1418 
1419 	if (SCARG(uap, times)) {
1420 		error = copyin(SCARG(uap, times), &lts, sizeof(lts));
1421 		if (error != 0)
1422 			return error;
1423 		linux_to_native_timespec(&ts[0], &lts[0]);
1424 		linux_to_native_timespec(&ts[1], &lts[1]);
1425 		tsp = ts;
1426 	}
1427 
1428 	return linux_do_sys_utimensat(l, SCARG(uap, fd), SCARG(uap, path),
1429 	    tsp, SCARG(uap, flag), retval);
1430 }
1431