linux_misc.c revision 1.223.4.1 1 /* $NetBSD: linux_misc.c,v 1.223.4.1 2013/08/28 23:59:24 rmind 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 1.223.4.1 2013/08/28 23:59:24 rmind 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 #ifndef COMPAT_LINUX32
117 #include <compat/linux/common/linux_statfs.h>
118 #include <compat/linux/common/linux_limit.h>
119 #endif
120 #include <compat/linux/common/linux_ptrace.h>
121 #include <compat/linux/common/linux_reboot.h>
122 #include <compat/linux/common/linux_emuldata.h>
123 #include <compat/linux/common/linux_sched.h>
124
125 #include <compat/linux/linux_syscallargs.h>
126
127 #ifndef COMPAT_LINUX32
128 const int linux_ptrace_request_map[] = {
129 LINUX_PTRACE_TRACEME, PT_TRACE_ME,
130 LINUX_PTRACE_PEEKTEXT, PT_READ_I,
131 LINUX_PTRACE_PEEKDATA, PT_READ_D,
132 LINUX_PTRACE_POKETEXT, PT_WRITE_I,
133 LINUX_PTRACE_POKEDATA, PT_WRITE_D,
134 LINUX_PTRACE_CONT, PT_CONTINUE,
135 LINUX_PTRACE_KILL, PT_KILL,
136 LINUX_PTRACE_ATTACH, PT_ATTACH,
137 LINUX_PTRACE_DETACH, PT_DETACH,
138 # ifdef PT_STEP
139 LINUX_PTRACE_SINGLESTEP, PT_STEP,
140 # endif
141 LINUX_PTRACE_SYSCALL, PT_SYSCALL,
142 -1
143 };
144
145 const struct linux_mnttypes linux_fstypes[] = {
146 { MOUNT_FFS, LINUX_DEFAULT_SUPER_MAGIC },
147 { MOUNT_NFS, LINUX_NFS_SUPER_MAGIC },
148 { MOUNT_MFS, LINUX_DEFAULT_SUPER_MAGIC },
149 { MOUNT_MSDOS, LINUX_MSDOS_SUPER_MAGIC },
150 { MOUNT_LFS, LINUX_DEFAULT_SUPER_MAGIC },
151 { MOUNT_FDESC, LINUX_DEFAULT_SUPER_MAGIC },
152 { MOUNT_NULL, LINUX_DEFAULT_SUPER_MAGIC },
153 { MOUNT_OVERLAY, LINUX_DEFAULT_SUPER_MAGIC },
154 { MOUNT_UMAP, LINUX_DEFAULT_SUPER_MAGIC },
155 { MOUNT_KERNFS, LINUX_DEFAULT_SUPER_MAGIC },
156 { MOUNT_PROCFS, LINUX_PROC_SUPER_MAGIC },
157 { MOUNT_AFS, LINUX_DEFAULT_SUPER_MAGIC },
158 { MOUNT_CD9660, LINUX_ISOFS_SUPER_MAGIC },
159 { MOUNT_UNION, LINUX_DEFAULT_SUPER_MAGIC },
160 { MOUNT_ADOSFS, LINUX_ADFS_SUPER_MAGIC },
161 { MOUNT_EXT2FS, LINUX_EXT2_SUPER_MAGIC },
162 { MOUNT_CFS, LINUX_DEFAULT_SUPER_MAGIC },
163 { MOUNT_CODA, LINUX_CODA_SUPER_MAGIC },
164 { MOUNT_FILECORE, LINUX_DEFAULT_SUPER_MAGIC },
165 { MOUNT_NTFS, LINUX_DEFAULT_SUPER_MAGIC },
166 { MOUNT_SMBFS, LINUX_SMB_SUPER_MAGIC },
167 { MOUNT_PTYFS, LINUX_DEVPTS_SUPER_MAGIC },
168 { MOUNT_TMPFS, LINUX_TMPFS_SUPER_MAGIC }
169 };
170 const int linux_fstypes_cnt = sizeof(linux_fstypes) / sizeof(linux_fstypes[0]);
171
172 # ifdef DEBUG_LINUX
173 #define DPRINTF(a) uprintf a
174 # else
175 #define DPRINTF(a)
176 # endif
177
178 /* Local linux_misc.c functions: */
179 static void linux_to_bsd_mmap_args(struct sys_mmap_args *,
180 const struct linux_sys_mmap_args *);
181 static int linux_mmap(struct lwp *, const struct linux_sys_mmap_args *,
182 register_t *, off_t);
183
184
185 /*
186 * The information on a terminated (or stopped) process needs
187 * to be converted in order for Linux binaries to get a valid signal
188 * number out of it.
189 */
190 int
191 bsd_to_linux_wstat(int st)
192 {
193
194 int sig;
195
196 if (WIFSIGNALED(st)) {
197 sig = WTERMSIG(st);
198 if (sig >= 0 && sig < NSIG)
199 st= (st & ~0177) | native_to_linux_signo[sig];
200 } else if (WIFSTOPPED(st)) {
201 sig = WSTOPSIG(st);
202 if (sig >= 0 && sig < NSIG)
203 st = (st & ~0xff00) |
204 (native_to_linux_signo[sig] << 8);
205 }
206 return st;
207 }
208
209 /*
210 * wait4(2). Passed on to the NetBSD call, surrounded by code to
211 * reserve some space for a NetBSD-style wait status, and converting
212 * it to what Linux wants.
213 */
214 int
215 linux_sys_wait4(struct lwp *l, const struct linux_sys_wait4_args *uap, register_t *retval)
216 {
217 /* {
218 syscallarg(int) pid;
219 syscallarg(int *) status;
220 syscallarg(int) options;
221 syscallarg(struct rusage50 *) rusage;
222 } */
223 int error, status, options, linux_options, pid = SCARG(uap, pid);
224 struct rusage50 ru50;
225 struct rusage ru;
226 proc_t *p;
227
228 linux_options = SCARG(uap, options);
229 options = WOPTSCHECKED;
230 if (linux_options & ~(LINUX_WAIT4_KNOWNFLAGS))
231 return (EINVAL);
232
233 if (linux_options & LINUX_WAIT4_WNOHANG)
234 options |= WNOHANG;
235 if (linux_options & LINUX_WAIT4_WUNTRACED)
236 options |= WUNTRACED;
237 if (linux_options & LINUX_WAIT4_WALL)
238 options |= WALLSIG;
239 if (linux_options & LINUX_WAIT4_WCLONE)
240 options |= WALTSIG;
241 # ifdef DIAGNOSTIC
242 if (linux_options & LINUX_WAIT4_WNOTHREAD)
243 printf("WARNING: %s: linux process %d.%d called "
244 "waitpid with __WNOTHREAD set!",
245 __FILE__, l->l_proc->p_pid, l->l_lid);
246
247 # endif
248
249 error = do_sys_wait(&pid, &status, options,
250 SCARG(uap, rusage) != NULL ? &ru : NULL);
251
252 retval[0] = pid;
253 if (pid == 0)
254 return error;
255
256 p = curproc;
257 mutex_enter(p->p_lock);
258 sigdelset(&p->p_sigpend.sp_set, SIGCHLD); /* XXXAD ksiginfo leak */
259 mutex_exit(p->p_lock);
260
261 if (SCARG(uap, rusage) != NULL) {
262 rusage_to_rusage50(&ru, &ru50);
263 error = copyout(&ru, SCARG(uap, rusage), sizeof(ru));
264 }
265
266 if (error == 0 && SCARG(uap, status) != NULL) {
267 status = bsd_to_linux_wstat(status);
268 error = copyout(&status, SCARG(uap, status), sizeof status);
269 }
270
271 return error;
272 }
273
274 /*
275 * Linux brk(2). Like native, but always return the new break value.
276 */
277 int
278 linux_sys_brk(struct lwp *l, const struct linux_sys_brk_args *uap, register_t *retval)
279 {
280 /* {
281 syscallarg(char *) nsize;
282 } */
283 struct proc *p = l->l_proc;
284 struct vmspace *vm = p->p_vmspace;
285 struct sys_obreak_args oba;
286
287 SCARG(&oba, nsize) = SCARG(uap, nsize);
288
289 (void) sys_obreak(l, &oba, retval);
290 retval[0] = (register_t)((char *)vm->vm_daddr + ptoa(vm->vm_dsize));
291 return 0;
292 }
293
294 /*
295 * Implement the fs stat functions. Straightforward.
296 */
297 int
298 linux_sys_statfs(struct lwp *l, const struct linux_sys_statfs_args *uap, register_t *retval)
299 {
300 /* {
301 syscallarg(const char *) path;
302 syscallarg(struct linux_statfs *) sp;
303 } */
304 struct statvfs *sb;
305 struct linux_statfs ltmp;
306 int error;
307
308 sb = STATVFSBUF_GET();
309 error = do_sys_pstatvfs(l, SCARG(uap, path), ST_WAIT, sb);
310 if (error == 0) {
311 bsd_to_linux_statfs(sb, <mp);
312 error = copyout(<mp, SCARG(uap, sp), sizeof ltmp);
313 }
314 STATVFSBUF_PUT(sb);
315
316 return error;
317 }
318
319 int
320 linux_sys_fstatfs(struct lwp *l, const struct linux_sys_fstatfs_args *uap, register_t *retval)
321 {
322 /* {
323 syscallarg(int) fd;
324 syscallarg(struct linux_statfs *) sp;
325 } */
326 struct statvfs *sb;
327 struct linux_statfs ltmp;
328 int error;
329
330 sb = STATVFSBUF_GET();
331 error = do_sys_fstatvfs(l, SCARG(uap, fd), ST_WAIT, sb);
332 if (error == 0) {
333 bsd_to_linux_statfs(sb, <mp);
334 error = copyout(<mp, SCARG(uap, sp), sizeof ltmp);
335 }
336 STATVFSBUF_PUT(sb);
337
338 return error;
339 }
340
341 /*
342 * uname(). Just copy the info from the various strings stored in the
343 * kernel, and put it in the Linux utsname structure. That structure
344 * is almost the same as the NetBSD one, only it has fields 65 characters
345 * long, and an extra domainname field.
346 */
347 int
348 linux_sys_uname(struct lwp *l, const struct linux_sys_uname_args *uap, register_t *retval)
349 {
350 /* {
351 syscallarg(struct linux_utsname *) up;
352 } */
353 struct linux_utsname luts;
354
355 strlcpy(luts.l_sysname, linux_sysname, sizeof(luts.l_sysname));
356 strlcpy(luts.l_nodename, hostname, sizeof(luts.l_nodename));
357 strlcpy(luts.l_release, linux_release, sizeof(luts.l_release));
358 strlcpy(luts.l_version, linux_version, sizeof(luts.l_version));
359 strlcpy(luts.l_machine, LINUX_UNAME_ARCH, sizeof(luts.l_machine));
360 strlcpy(luts.l_domainname, domainname, sizeof(luts.l_domainname));
361
362 return copyout(&luts, SCARG(uap, up), sizeof(luts));
363 }
364
365 /* Used directly on: alpha, mips, ppc, sparc, sparc64 */
366 /* Used indirectly on: arm, i386, m68k */
367
368 /*
369 * New type Linux mmap call.
370 * Only called directly on machines with >= 6 free regs.
371 */
372 int
373 linux_sys_mmap(struct lwp *l, const struct linux_sys_mmap_args *uap, register_t *retval)
374 {
375 /* {
376 syscallarg(unsigned long) addr;
377 syscallarg(size_t) len;
378 syscallarg(int) prot;
379 syscallarg(int) flags;
380 syscallarg(int) fd;
381 syscallarg(linux_off_t) offset;
382 } */
383
384 if (SCARG(uap, offset) & PAGE_MASK)
385 return EINVAL;
386
387 return linux_mmap(l, uap, retval, SCARG(uap, offset));
388 }
389
390 /*
391 * Guts of most architectures' mmap64() implementations. This shares
392 * its list of arguments with linux_sys_mmap().
393 *
394 * The difference in linux_sys_mmap2() is that "offset" is actually
395 * (offset / pagesize), not an absolute byte count. This translation
396 * to pagesize offsets is done inside glibc between the mmap64() call
397 * point, and the actual syscall.
398 */
399 int
400 linux_sys_mmap2(struct lwp *l, const struct linux_sys_mmap2_args *uap, register_t *retval)
401 {
402 /* {
403 syscallarg(unsigned long) addr;
404 syscallarg(size_t) len;
405 syscallarg(int) prot;
406 syscallarg(int) flags;
407 syscallarg(int) fd;
408 syscallarg(linux_off_t) offset;
409 } */
410
411 return linux_mmap(l, uap, retval,
412 ((off_t)SCARG(uap, offset)) << PAGE_SHIFT);
413 }
414
415 /*
416 * Massage arguments and call system mmap(2).
417 */
418 static int
419 linux_mmap(struct lwp *l, const struct linux_sys_mmap_args *uap, register_t *retval, off_t offset)
420 {
421 struct sys_mmap_args cma;
422 int error;
423 size_t mmoff=0;
424
425 linux_to_bsd_mmap_args(&cma, uap);
426 SCARG(&cma, pos) = offset;
427
428 if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN) {
429 /*
430 * Request for stack-like memory segment. On linux, this
431 * works by mmap()ping (small) segment, which is automatically
432 * extended when page fault happens below the currently
433 * allocated area. We emulate this by allocating (typically
434 * bigger) segment sized at current stack size limit, and
435 * offsetting the requested and returned address accordingly.
436 * Since physical pages are only allocated on-demand, this
437 * is effectively identical.
438 */
439 rlim_t ssl = l->l_proc->p_rlimit[RLIMIT_STACK].rlim_cur;
440
441 if (SCARG(&cma, len) < ssl) {
442 /* Compute the address offset */
443 mmoff = round_page(ssl) - SCARG(uap, len);
444
445 if (SCARG(&cma, addr))
446 SCARG(&cma, addr) = (char *)SCARG(&cma, addr) - mmoff;
447
448 SCARG(&cma, len) = (size_t) ssl;
449 }
450 }
451
452 error = sys_mmap(l, &cma, retval);
453 if (error)
454 return (error);
455
456 /* Shift the returned address for stack-like segment if necessary */
457 retval[0] += mmoff;
458
459 return (0);
460 }
461
462 static void
463 linux_to_bsd_mmap_args(struct sys_mmap_args *cma, const struct linux_sys_mmap_args *uap)
464 {
465 int flags = MAP_TRYFIXED, fl = SCARG(uap, flags);
466
467 flags |= cvtto_bsd_mask(fl, LINUX_MAP_SHARED, MAP_SHARED);
468 flags |= cvtto_bsd_mask(fl, LINUX_MAP_PRIVATE, MAP_PRIVATE);
469 flags |= cvtto_bsd_mask(fl, LINUX_MAP_FIXED, MAP_FIXED);
470 flags |= cvtto_bsd_mask(fl, LINUX_MAP_ANON, MAP_ANON);
471 /* XXX XAX ERH: Any other flags here? There are more defined... */
472
473 SCARG(cma, addr) = (void *)SCARG(uap, addr);
474 SCARG(cma, len) = SCARG(uap, len);
475 SCARG(cma, prot) = SCARG(uap, prot);
476 if (SCARG(cma, prot) & VM_PROT_WRITE) /* XXX */
477 SCARG(cma, prot) |= VM_PROT_READ;
478 SCARG(cma, flags) = flags;
479 SCARG(cma, fd) = flags & MAP_ANON ? -1 : SCARG(uap, fd);
480 SCARG(cma, PAD) = 0;
481 }
482
483 #define LINUX_MREMAP_MAYMOVE 1
484 #define LINUX_MREMAP_FIXED 2
485
486 int
487 linux_sys_mremap(struct lwp *l, const struct linux_sys_mremap_args *uap, register_t *retval)
488 {
489 /* {
490 syscallarg(void *) old_address;
491 syscallarg(size_t) old_size;
492 syscallarg(size_t) new_size;
493 syscallarg(u_long) flags;
494 } */
495
496 struct proc *p;
497 struct vm_map *map;
498 vaddr_t oldva;
499 vaddr_t newva;
500 size_t oldsize;
501 size_t newsize;
502 int flags;
503 int uvmflags;
504 int error;
505
506 flags = SCARG(uap, flags);
507 oldva = (vaddr_t)SCARG(uap, old_address);
508 oldsize = round_page(SCARG(uap, old_size));
509 newsize = round_page(SCARG(uap, new_size));
510 if ((flags & ~(LINUX_MREMAP_FIXED|LINUX_MREMAP_MAYMOVE)) != 0) {
511 error = EINVAL;
512 goto done;
513 }
514 if ((flags & LINUX_MREMAP_FIXED) != 0) {
515 if ((flags & LINUX_MREMAP_MAYMOVE) == 0) {
516 error = EINVAL;
517 goto done;
518 }
519 #if 0 /* notyet */
520 newva = SCARG(uap, new_address);
521 uvmflags = MAP_FIXED;
522 #else /* notyet */
523 error = EOPNOTSUPP;
524 goto done;
525 #endif /* notyet */
526 } else if ((flags & LINUX_MREMAP_MAYMOVE) != 0) {
527 uvmflags = 0;
528 } else {
529 newva = oldva;
530 uvmflags = MAP_FIXED;
531 }
532 p = l->l_proc;
533 map = &p->p_vmspace->vm_map;
534 error = uvm_mremap(map, oldva, oldsize, map, &newva, newsize, p,
535 uvmflags);
536
537 done:
538 *retval = (error != 0) ? 0 : (register_t)newva;
539 return error;
540 }
541
542 #ifdef USRSTACK
543 int
544 linux_sys_mprotect(struct lwp *l, const struct linux_sys_mprotect_args *uap, register_t *retval)
545 {
546 /* {
547 syscallarg(const void *) start;
548 syscallarg(unsigned long) len;
549 syscallarg(int) prot;
550 } */
551 struct vm_map_entry *entry;
552 struct vm_map *map;
553 struct proc *p;
554 vaddr_t end, start, len, stacklim;
555 int prot, grows;
556
557 start = (vaddr_t)SCARG(uap, start);
558 len = round_page(SCARG(uap, len));
559 prot = SCARG(uap, prot);
560 grows = prot & (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP);
561 prot &= ~grows;
562 end = start + len;
563
564 if (start & PAGE_MASK)
565 return EINVAL;
566 if (end < start)
567 return EINVAL;
568 if (end == start)
569 return 0;
570
571 if (prot & ~(PROT_READ | PROT_WRITE | PROT_EXEC))
572 return EINVAL;
573 if (grows == (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP))
574 return EINVAL;
575
576 p = l->l_proc;
577 map = &p->p_vmspace->vm_map;
578 vm_map_lock(map);
579 # ifdef notdef
580 VM_MAP_RANGE_CHECK(map, start, end);
581 # endif
582 if (!uvm_map_lookup_entry(map, start, &entry) || entry->start > start) {
583 vm_map_unlock(map);
584 return ENOMEM;
585 }
586
587 /*
588 * Approximate the behaviour of PROT_GROWS{DOWN,UP}.
589 */
590
591 stacklim = (vaddr_t)p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur;
592 if (grows & LINUX_PROT_GROWSDOWN) {
593 if (USRSTACK - stacklim <= start && start < USRSTACK) {
594 start = USRSTACK - stacklim;
595 } else {
596 start = entry->start;
597 }
598 } else if (grows & LINUX_PROT_GROWSUP) {
599 if (USRSTACK <= end && end < USRSTACK + stacklim) {
600 end = USRSTACK + stacklim;
601 } else {
602 end = entry->end;
603 }
604 }
605 vm_map_unlock(map);
606 return uvm_map_protect(map, start, end, prot, FALSE);
607 }
608 #endif /* USRSTACK */
609
610 /*
611 * This code is partly stolen from src/lib/libc/compat-43/times.c
612 */
613
614 #define CONVTCK(r) (r.tv_sec * hz + r.tv_usec / (1000000 / hz))
615
616 int
617 linux_sys_times(struct lwp *l, const struct linux_sys_times_args *uap, register_t *retval)
618 {
619 /* {
620 syscallarg(struct times *) tms;
621 } */
622 struct proc *p = l->l_proc;
623 struct timeval t;
624 int error;
625
626 if (SCARG(uap, tms)) {
627 struct linux_tms ltms;
628 struct rusage ru;
629
630 mutex_enter(p->p_lock);
631 calcru(p, &ru.ru_utime, &ru.ru_stime, NULL, NULL);
632 ltms.ltms_utime = CONVTCK(ru.ru_utime);
633 ltms.ltms_stime = CONVTCK(ru.ru_stime);
634 ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime);
635 ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime);
636 mutex_exit(p->p_lock);
637
638 if ((error = copyout(<ms, SCARG(uap, tms), sizeof ltms)))
639 return error;
640 }
641
642 getmicrouptime(&t);
643
644 retval[0] = ((linux_clock_t)(CONVTCK(t)));
645 return 0;
646 }
647
648 #undef CONVTCK
649
650 /*
651 * Linux 'readdir' call. This code is mostly taken from the
652 * SunOS getdents call (see compat/sunos/sunos_misc.c), though
653 * an attempt has been made to keep it a little cleaner (failing
654 * miserably, because of the cruft needed if count 1 is passed).
655 *
656 * The d_off field should contain the offset of the next valid entry,
657 * but in Linux it has the offset of the entry itself. We emulate
658 * that bug here.
659 *
660 * Read in BSD-style entries, convert them, and copy them out.
661 *
662 * Note that this doesn't handle union-mounted filesystems.
663 */
664 int
665 linux_sys_getdents(struct lwp *l, const struct linux_sys_getdents_args *uap, register_t *retval)
666 {
667 /* {
668 syscallarg(int) fd;
669 syscallarg(struct linux_dirent *) dent;
670 syscallarg(unsigned int) count;
671 } */
672 struct dirent *bdp;
673 struct vnode *vp;
674 char *inp, *tbuf; /* BSD-format */
675 int len, reclen; /* BSD-format */
676 char *outp; /* Linux-format */
677 int resid, linux_reclen = 0; /* Linux-format */
678 struct file *fp;
679 struct uio auio;
680 struct iovec aiov;
681 struct linux_dirent idb;
682 off_t off; /* true file offset */
683 int buflen, error, eofflag, nbytes, oldcall;
684 struct vattr va;
685 off_t *cookiebuf = NULL, *cookie;
686 int ncookies;
687
688 /* fd_getvnode() will use the descriptor for us */
689 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
690 return (error);
691
692 if ((fp->f_flag & FREAD) == 0) {
693 error = EBADF;
694 goto out1;
695 }
696
697 vp = (struct vnode *)fp->f_data;
698 if (vp->v_type != VDIR) {
699 error = ENOTDIR;
700 goto out1;
701 }
702
703 vn_lock(vp, LK_SHARED | LK_RETRY);
704 error = VOP_GETATTR(vp, &va, l->l_cred);
705 VOP_UNLOCK(vp);
706 if (error)
707 goto out1;
708
709 nbytes = SCARG(uap, count);
710 if (nbytes == 1) { /* emulating old, broken behaviour */
711 nbytes = sizeof (idb);
712 buflen = max(va.va_blocksize, nbytes);
713 oldcall = 1;
714 } else {
715 buflen = min(MAXBSIZE, nbytes);
716 if (buflen < va.va_blocksize)
717 buflen = va.va_blocksize;
718 oldcall = 0;
719 }
720 tbuf = malloc(buflen, M_TEMP, M_WAITOK);
721
722 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
723 off = fp->f_offset;
724 again:
725 aiov.iov_base = tbuf;
726 aiov.iov_len = buflen;
727 auio.uio_iov = &aiov;
728 auio.uio_iovcnt = 1;
729 auio.uio_rw = UIO_READ;
730 auio.uio_resid = buflen;
731 auio.uio_offset = off;
732 UIO_SETUP_SYSSPACE(&auio);
733 /*
734 * First we read into the malloc'ed buffer, then
735 * we massage it into user space, one record at a time.
736 */
737 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf,
738 &ncookies);
739 if (error)
740 goto out;
741
742 inp = tbuf;
743 outp = (void *)SCARG(uap, dent);
744 resid = nbytes;
745 if ((len = buflen - auio.uio_resid) == 0)
746 goto eof;
747
748 for (cookie = cookiebuf; len > 0; len -= reclen) {
749 bdp = (struct dirent *)inp;
750 reclen = bdp->d_reclen;
751 if (reclen & 3)
752 panic("linux_readdir");
753 if (bdp->d_fileno == 0) {
754 inp += reclen; /* it is a hole; squish it out */
755 if (cookie)
756 off = *cookie++;
757 else
758 off += reclen;
759 continue;
760 }
761 linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen);
762 if (reclen > len || resid < linux_reclen) {
763 /* entry too big for buffer, so just stop */
764 outp++;
765 break;
766 }
767 /*
768 * Massage in place to make a Linux-shaped dirent (otherwise
769 * we have to worry about touching user memory outside of
770 * the copyout() call).
771 */
772 idb.d_ino = bdp->d_fileno;
773 /*
774 * The old readdir() call misuses the offset and reclen fields.
775 */
776 if (oldcall) {
777 idb.d_off = (linux_off_t)linux_reclen;
778 idb.d_reclen = (u_short)bdp->d_namlen;
779 } else {
780 if (sizeof (idb.d_off) <= 4 && (off >> 32) != 0) {
781 compat_offseterr(vp, "linux_getdents");
782 error = EINVAL;
783 goto out;
784 }
785 idb.d_off = (linux_off_t)off;
786 idb.d_reclen = (u_short)linux_reclen;
787 }
788 strcpy(idb.d_name, bdp->d_name);
789 idb.d_name[strlen(idb.d_name) + 1] = bdp->d_type;
790 if ((error = copyout((void *)&idb, outp, linux_reclen)))
791 goto out;
792 /* advance past this real entry */
793 inp += reclen;
794 if (cookie)
795 off = *cookie++; /* each entry points to itself */
796 else
797 off += reclen;
798 /* advance output past Linux-shaped entry */
799 outp += linux_reclen;
800 resid -= linux_reclen;
801 if (oldcall)
802 break;
803 }
804
805 /* if we squished out the whole block, try again */
806 if (outp == (void *)SCARG(uap, dent)) {
807 if (cookiebuf)
808 free(cookiebuf, M_TEMP);
809 cookiebuf = NULL;
810 goto again;
811 }
812 fp->f_offset = off; /* update the vnode offset */
813
814 if (oldcall)
815 nbytes = resid + linux_reclen;
816
817 eof:
818 *retval = nbytes - resid;
819 out:
820 VOP_UNLOCK(vp);
821 if (cookiebuf)
822 free(cookiebuf, M_TEMP);
823 free(tbuf, M_TEMP);
824 out1:
825 fd_putfile(SCARG(uap, fd));
826 return error;
827 }
828
829 /*
830 * Even when just using registers to pass arguments to syscalls you can
831 * have 5 of them on the i386. So this newer version of select() does
832 * this.
833 */
834 int
835 linux_sys_select(struct lwp *l, const struct linux_sys_select_args *uap, register_t *retval)
836 {
837 /* {
838 syscallarg(int) nfds;
839 syscallarg(fd_set *) readfds;
840 syscallarg(fd_set *) writefds;
841 syscallarg(fd_set *) exceptfds;
842 syscallarg(struct timeval50 *) timeout;
843 } */
844
845 return linux_select1(l, retval, SCARG(uap, nfds), SCARG(uap, readfds),
846 SCARG(uap, writefds), SCARG(uap, exceptfds),
847 (struct linux_timeval *)SCARG(uap, timeout));
848 }
849
850 /*
851 * Common code for the old and new versions of select(). A couple of
852 * things are important:
853 * 1) return the amount of time left in the 'timeout' parameter
854 * 2) select never returns ERESTART on Linux, always return EINTR
855 */
856 int
857 linux_select1(struct lwp *l, register_t *retval, int nfds, fd_set *readfds,
858 fd_set *writefds, fd_set *exceptfds, struct linux_timeval *timeout)
859 {
860 struct timespec ts0, ts1, uts, *ts = NULL;
861 struct linux_timeval ltv;
862 int error;
863
864 /*
865 * Store current time for computation of the amount of
866 * time left.
867 */
868 if (timeout) {
869 if ((error = copyin(timeout, <v, sizeof(ltv))))
870 return error;
871 uts.tv_sec = ltv.tv_sec;
872 uts.tv_nsec = ltv.tv_usec * 1000;
873 if (itimespecfix(&uts)) {
874 /*
875 * The timeval was invalid. Convert it to something
876 * valid that will act as it does under Linux.
877 */
878 uts.tv_sec += uts.tv_nsec / 1000000000;
879 uts.tv_nsec %= 1000000000;
880 if (uts.tv_nsec < 0) {
881 uts.tv_sec -= 1;
882 uts.tv_nsec += 1000000000;
883 }
884 if (uts.tv_sec < 0)
885 timespecclear(&uts);
886 }
887 ts = &uts;
888 nanotime(&ts0);
889 }
890
891 error = selcommon(retval, nfds, readfds, writefds, exceptfds, ts, NULL);
892
893 if (error) {
894 /*
895 * See fs/select.c in the Linux kernel. Without this,
896 * Maelstrom doesn't work.
897 */
898 if (error == ERESTART)
899 error = EINTR;
900 return error;
901 }
902
903 if (timeout) {
904 if (*retval) {
905 /*
906 * Compute how much time was left of the timeout,
907 * by subtracting the current time and the time
908 * before we started the call, and subtracting
909 * that result from the user-supplied value.
910 */
911 nanotime(&ts1);
912 timespecsub(&ts1, &ts0, &ts1);
913 timespecsub(&uts, &ts1, &uts);
914 if (uts.tv_sec < 0)
915 timespecclear(&uts);
916 } else
917 timespecclear(&uts);
918 ltv.tv_sec = uts.tv_sec;
919 ltv.tv_usec = uts.tv_nsec / 1000;
920 if ((error = copyout(<v, timeout, sizeof(ltv))))
921 return error;
922 }
923
924 return 0;
925 }
926
927 int
928 linux_sys_ppoll(struct lwp *l,
929 const struct linux_sys_ppoll_args *uap, register_t *retval)
930 {
931 /* {
932 syscallarg(struct pollfd *) fds;
933 syscallarg(int) nfds;
934 syscallarg(struct linux_timespec *) timeout;
935 syscallarg(linux_sigset_t *) sigset;
936 } */
937 struct linux_timespec lts0, *lts;
938 struct timespec ts0, *ts = NULL;
939 linux_sigset_t lsigmask0, *lsigmask;
940 sigset_t sigmask0, *sigmask = NULL;
941 int error;
942
943 lts = SCARG(uap, timeout);
944 if (lts) {
945 if ((error = copyin(lts, <s0, sizeof(lts0))) != 0)
946 return error;
947 linux_to_native_timespec(&ts0, <s0);
948 ts = &ts0;
949 }
950
951 lsigmask = SCARG(uap, sigset);
952 if (lsigmask) {
953 if ((error = copyin(lsigmask, &lsigmask0, sizeof(lsigmask0))))
954 return error;
955 linux_to_native_sigset(&sigmask0, &lsigmask0);
956 sigmask = &sigmask0;
957 }
958
959 return pollcommon(retval, SCARG(uap, fds), SCARG(uap, nfds),
960 ts, sigmask);
961 }
962
963 /*
964 * Set the 'personality' (emulation mode) for the current process. Only
965 * accept the Linux personality here (0). This call is needed because
966 * the Linux ELF crt0 issues it in an ugly kludge to make sure that
967 * ELF binaries run in Linux mode, not SVR4 mode.
968 */
969 int
970 linux_sys_personality(struct lwp *l, const struct linux_sys_personality_args *uap, register_t *retval)
971 {
972 /* {
973 syscallarg(unsigned long) per;
974 } */
975 struct linux_emuldata *led;
976 int per;
977
978 per = SCARG(uap, per);
979 led = l->l_emuldata;
980 if (per == LINUX_PER_QUERY) {
981 retval[0] = led->led_personality;
982 return 0;
983 }
984
985 switch (per & LINUX_PER_MASK) {
986 case LINUX_PER_LINUX:
987 case LINUX_PER_LINUX32:
988 led->led_personality = per;
989 break;
990
991 default:
992 return EINVAL;
993 }
994
995 retval[0] = per;
996 return 0;
997 }
998
999 /*
1000 * We have nonexistent fsuid equal to uid.
1001 * If modification is requested, refuse.
1002 */
1003 int
1004 linux_sys_setfsuid(struct lwp *l, const struct linux_sys_setfsuid_args *uap, register_t *retval)
1005 {
1006 /* {
1007 syscallarg(uid_t) uid;
1008 } */
1009 uid_t uid;
1010
1011 uid = SCARG(uap, uid);
1012 if (kauth_cred_getuid(l->l_cred) != uid)
1013 return sys_nosys(l, uap, retval);
1014
1015 *retval = uid;
1016 return 0;
1017 }
1018
1019 int
1020 linux_sys_setfsgid(struct lwp *l, const struct linux_sys_setfsgid_args *uap, register_t *retval)
1021 {
1022 /* {
1023 syscallarg(gid_t) gid;
1024 } */
1025 gid_t gid;
1026
1027 gid = SCARG(uap, gid);
1028 if (kauth_cred_getgid(l->l_cred) != gid)
1029 return sys_nosys(l, uap, retval);
1030
1031 *retval = gid;
1032 return 0;
1033 }
1034
1035 int
1036 linux_sys_setresuid(struct lwp *l, const struct linux_sys_setresuid_args *uap, register_t *retval)
1037 {
1038 /* {
1039 syscallarg(uid_t) ruid;
1040 syscallarg(uid_t) euid;
1041 syscallarg(uid_t) suid;
1042 } */
1043
1044 /*
1045 * Note: These checks are a little different than the NetBSD
1046 * setreuid(2) call performs. This precisely follows the
1047 * behavior of the Linux kernel.
1048 */
1049
1050 return do_setresuid(l, SCARG(uap, ruid), SCARG(uap, euid),
1051 SCARG(uap, suid),
1052 ID_R_EQ_R | ID_R_EQ_E | ID_R_EQ_S |
1053 ID_E_EQ_R | ID_E_EQ_E | ID_E_EQ_S |
1054 ID_S_EQ_R | ID_S_EQ_E | ID_S_EQ_S );
1055 }
1056
1057 int
1058 linux_sys_getresuid(struct lwp *l, const struct linux_sys_getresuid_args *uap, register_t *retval)
1059 {
1060 /* {
1061 syscallarg(uid_t *) ruid;
1062 syscallarg(uid_t *) euid;
1063 syscallarg(uid_t *) suid;
1064 } */
1065 kauth_cred_t pc = l->l_cred;
1066 int error;
1067 uid_t uid;
1068
1069 /*
1070 * Linux copies these values out to userspace like so:
1071 *
1072 * 1. Copy out ruid.
1073 * 2. If that succeeds, copy out euid.
1074 * 3. If both of those succeed, copy out suid.
1075 */
1076 uid = kauth_cred_getuid(pc);
1077 if ((error = copyout(&uid, SCARG(uap, ruid), sizeof(uid_t))) != 0)
1078 return (error);
1079
1080 uid = kauth_cred_geteuid(pc);
1081 if ((error = copyout(&uid, SCARG(uap, euid), sizeof(uid_t))) != 0)
1082 return (error);
1083
1084 uid = kauth_cred_getsvuid(pc);
1085
1086 return (copyout(&uid, SCARG(uap, suid), sizeof(uid_t)));
1087 }
1088
1089 int
1090 linux_sys_ptrace(struct lwp *l, const struct linux_sys_ptrace_args *uap, register_t *retval)
1091 {
1092 /* {
1093 i386, m68k, powerpc: T=int
1094 alpha, amd64: T=long
1095 syscallarg(T) request;
1096 syscallarg(T) pid;
1097 syscallarg(T) addr;
1098 syscallarg(T) data;
1099 } */
1100 const int *ptr;
1101 int request;
1102 int error;
1103
1104 ptr = linux_ptrace_request_map;
1105 request = SCARG(uap, request);
1106 while (*ptr != -1)
1107 if (*ptr++ == request) {
1108 struct sys_ptrace_args pta;
1109
1110 SCARG(&pta, req) = *ptr;
1111 SCARG(&pta, pid) = SCARG(uap, pid);
1112 SCARG(&pta, addr) = (void *)SCARG(uap, addr);
1113 SCARG(&pta, data) = SCARG(uap, data);
1114
1115 /*
1116 * Linux ptrace(PTRACE_CONT, pid, 0, 0) means actually
1117 * to continue where the process left off previously.
1118 * The same thing is achieved by addr == (void *) 1
1119 * on NetBSD, so rewrite 'addr' appropriately.
1120 */
1121 if (request == LINUX_PTRACE_CONT && SCARG(uap, addr)==0)
1122 SCARG(&pta, addr) = (void *) 1;
1123
1124 error = sysent[SYS_ptrace].sy_call(l, &pta, retval);
1125 if (error)
1126 return error;
1127 switch (request) {
1128 case LINUX_PTRACE_PEEKTEXT:
1129 case LINUX_PTRACE_PEEKDATA:
1130 error = copyout (retval,
1131 (void *)SCARG(uap, data),
1132 sizeof *retval);
1133 *retval = SCARG(uap, data);
1134 break;
1135 default:
1136 break;
1137 }
1138 return error;
1139 }
1140 else
1141 ptr++;
1142
1143 return LINUX_SYS_PTRACE_ARCH(l, uap, retval);
1144 }
1145
1146 int
1147 linux_sys_reboot(struct lwp *l, const struct linux_sys_reboot_args *uap, register_t *retval)
1148 {
1149 /* {
1150 syscallarg(int) magic1;
1151 syscallarg(int) magic2;
1152 syscallarg(int) cmd;
1153 syscallarg(void *) arg;
1154 } */
1155 struct sys_reboot_args /* {
1156 syscallarg(int) opt;
1157 syscallarg(char *) bootstr;
1158 } */ sra;
1159 int error;
1160
1161 if ((error = kauth_authorize_system(l->l_cred,
1162 KAUTH_SYSTEM_REBOOT, 0, NULL, NULL, NULL)) != 0)
1163 return(error);
1164
1165 if (SCARG(uap, magic1) != LINUX_REBOOT_MAGIC1)
1166 return(EINVAL);
1167 if (SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2 &&
1168 SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2A &&
1169 SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2B)
1170 return(EINVAL);
1171
1172 switch ((unsigned long)SCARG(uap, cmd)) {
1173 case LINUX_REBOOT_CMD_RESTART:
1174 SCARG(&sra, opt) = RB_AUTOBOOT;
1175 break;
1176 case LINUX_REBOOT_CMD_HALT:
1177 SCARG(&sra, opt) = RB_HALT;
1178 break;
1179 case LINUX_REBOOT_CMD_POWER_OFF:
1180 SCARG(&sra, opt) = RB_HALT|RB_POWERDOWN;
1181 break;
1182 case LINUX_REBOOT_CMD_RESTART2:
1183 /* Reboot with an argument. */
1184 SCARG(&sra, opt) = RB_AUTOBOOT|RB_STRING;
1185 SCARG(&sra, bootstr) = SCARG(uap, arg);
1186 break;
1187 case LINUX_REBOOT_CMD_CAD_ON:
1188 return(EINVAL); /* We don't implement ctrl-alt-delete */
1189 case LINUX_REBOOT_CMD_CAD_OFF:
1190 return(0);
1191 default:
1192 return(EINVAL);
1193 }
1194
1195 return(sys_reboot(l, &sra, retval));
1196 }
1197
1198 /*
1199 * Copy of compat_12_sys_swapon().
1200 */
1201 int
1202 linux_sys_swapon(struct lwp *l, const struct linux_sys_swapon_args *uap, register_t *retval)
1203 {
1204 /* {
1205 syscallarg(const char *) name;
1206 } */
1207 struct sys_swapctl_args ua;
1208
1209 SCARG(&ua, cmd) = SWAP_ON;
1210 SCARG(&ua, arg) = (void *)__UNCONST(SCARG(uap, name));
1211 SCARG(&ua, misc) = 0; /* priority */
1212 return (sys_swapctl(l, &ua, retval));
1213 }
1214
1215 /*
1216 * Stop swapping to the file or block device specified by path.
1217 */
1218 int
1219 linux_sys_swapoff(struct lwp *l, const struct linux_sys_swapoff_args *uap, register_t *retval)
1220 {
1221 /* {
1222 syscallarg(const char *) path;
1223 } */
1224 struct sys_swapctl_args ua;
1225
1226 SCARG(&ua, cmd) = SWAP_OFF;
1227 SCARG(&ua, arg) = __UNCONST(SCARG(uap, path)); /*XXXUNCONST*/
1228 return (sys_swapctl(l, &ua, retval));
1229 }
1230
1231 /*
1232 * Copy of compat_09_sys_setdomainname()
1233 */
1234 /* ARGSUSED */
1235 int
1236 linux_sys_setdomainname(struct lwp *l, const struct linux_sys_setdomainname_args *uap, register_t *retval)
1237 {
1238 /* {
1239 syscallarg(char *) domainname;
1240 syscallarg(int) len;
1241 } */
1242 int name[2];
1243
1244 name[0] = CTL_KERN;
1245 name[1] = KERN_DOMAINNAME;
1246 return (old_sysctl(&name[0], 2, 0, 0, SCARG(uap, domainname),
1247 SCARG(uap, len), l));
1248 }
1249
1250 /*
1251 * sysinfo()
1252 */
1253 /* ARGSUSED */
1254 int
1255 linux_sys_sysinfo(struct lwp *l, const struct linux_sys_sysinfo_args *uap, register_t *retval)
1256 {
1257 /* {
1258 syscallarg(struct linux_sysinfo *) arg;
1259 } */
1260 struct linux_sysinfo si;
1261 struct loadavg *la;
1262
1263 si.uptime = time_uptime;
1264 la = &averunnable;
1265 si.loads[0] = la->ldavg[0] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1266 si.loads[1] = la->ldavg[1] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1267 si.loads[2] = la->ldavg[2] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1268 si.totalram = ctob((u_long)physmem);
1269 si.freeram = (u_long)uvmexp.free * uvmexp.pagesize;
1270 si.sharedram = 0; /* XXX */
1271 si.bufferram = (u_long)uvmexp.filepages * uvmexp.pagesize;
1272 si.totalswap = (u_long)uvmexp.swpages * uvmexp.pagesize;
1273 si.freeswap =
1274 (u_long)(uvmexp.swpages - uvmexp.swpginuse) * uvmexp.pagesize;
1275 si.procs = nprocs;
1276
1277 /* The following are only present in newer Linux kernels. */
1278 si.totalbig = 0;
1279 si.freebig = 0;
1280 si.mem_unit = 1;
1281
1282 return (copyout(&si, SCARG(uap, arg), sizeof si));
1283 }
1284
1285 int
1286 linux_sys_getrlimit(struct lwp *l, const struct linux_sys_getrlimit_args *uap, register_t *retval)
1287 {
1288 /* {
1289 syscallarg(int) which;
1290 # ifdef LINUX_LARGEFILE64
1291 syscallarg(struct rlimit *) rlp;
1292 # else
1293 syscallarg(struct orlimit *) rlp;
1294 # endif
1295 } */
1296 # ifdef LINUX_LARGEFILE64
1297 struct rlimit orl;
1298 # else
1299 struct orlimit orl;
1300 # endif
1301 int which;
1302
1303 which = linux_to_bsd_limit(SCARG(uap, which));
1304 if (which < 0)
1305 return -which;
1306
1307 bsd_to_linux_rlimit(&orl, &l->l_proc->p_rlimit[which]);
1308
1309 return copyout(&orl, SCARG(uap, rlp), sizeof(orl));
1310 }
1311
1312 int
1313 linux_sys_setrlimit(struct lwp *l, const struct linux_sys_setrlimit_args *uap, register_t *retval)
1314 {
1315 /* {
1316 syscallarg(int) which;
1317 # ifdef LINUX_LARGEFILE64
1318 syscallarg(struct rlimit *) rlp;
1319 # else
1320 syscallarg(struct orlimit *) rlp;
1321 # endif
1322 } */
1323 struct rlimit rl;
1324 # ifdef LINUX_LARGEFILE64
1325 struct rlimit orl;
1326 # else
1327 struct orlimit orl;
1328 # endif
1329 int error;
1330 int which;
1331
1332 if ((error = copyin(SCARG(uap, rlp), &orl, sizeof(orl))) != 0)
1333 return error;
1334
1335 which = linux_to_bsd_limit(SCARG(uap, which));
1336 if (which < 0)
1337 return -which;
1338
1339 linux_to_bsd_rlimit(&rl, &orl);
1340 return dosetrlimit(l, l->l_proc, which, &rl);
1341 }
1342
1343 # if !defined(__mips__) && !defined(__amd64__)
1344 /* XXX: this doesn't look 100% common, at least mips doesn't have it */
1345 int
1346 linux_sys_ugetrlimit(struct lwp *l, const struct linux_sys_ugetrlimit_args *uap, register_t *retval)
1347 {
1348 return linux_sys_getrlimit(l, (const void *)uap, retval);
1349 }
1350 # endif
1351
1352 /*
1353 * This gets called for unsupported syscalls. The difference to sys_nosys()
1354 * is that process does not get SIGSYS, the call just returns with ENOSYS.
1355 * This is the way Linux does it and glibc depends on this behaviour.
1356 */
1357 int
1358 linux_sys_nosys(struct lwp *l, const void *v, register_t *retval)
1359 {
1360 return (ENOSYS);
1361 }
1362
1363 int
1364 linux_sys_getpriority(struct lwp *l, const struct linux_sys_getpriority_args *uap, register_t *retval)
1365 {
1366 /* {
1367 syscallarg(int) which;
1368 syscallarg(int) who;
1369 } */
1370 struct sys_getpriority_args bsa;
1371 int error;
1372
1373 SCARG(&bsa, which) = SCARG(uap, which);
1374 SCARG(&bsa, who) = SCARG(uap, who);
1375
1376 if ((error = sys_getpriority(l, &bsa, retval)))
1377 return error;
1378
1379 *retval = NZERO - *retval;
1380
1381 return 0;
1382 }
1383
1384 #ifndef __alpha__
1385 int
1386 linux_sys_utimes(struct lwp *l, const struct linux_sys_utimes_args *uap, register_t *retval)
1387 {
1388 /* {
1389 syscallarg(const char *) path;
1390 syscallarg(const struct linux_timeval) *times;
1391 } */
1392 struct linux_timeval ltv[2];
1393 struct timeval tv[2];
1394 struct timeval *tptr = NULL;
1395 int error;
1396
1397 if (SCARG(uap, times)) {
1398 if ((error = copyin(SCARG(uap, times), <v, sizeof(ltv))))
1399 return error;
1400
1401 tv[0].tv_sec = ltv[0].tv_sec;
1402 tv[0].tv_usec = ltv[0].tv_usec;
1403 tv[1].tv_sec = ltv[1].tv_sec;
1404 tv[1].tv_usec = ltv[1].tv_usec;
1405
1406 tptr = tv;
1407 }
1408
1409 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW,
1410 tptr, UIO_SYSSPACE);
1411 }
1412
1413 int linux_sys_lutimes(struct lwp *, const struct linux_sys_utimes_args *, register_t *);
1414 int
1415 linux_sys_lutimes(struct lwp *l, const struct linux_sys_utimes_args *uap, register_t *retval)
1416 {
1417 /* {
1418 syscallarg(const char *) path;
1419 syscallarg(const struct linux_timeval) *times;
1420 } */
1421 struct linux_timeval ltv[2];
1422 struct timeval tv[2];
1423 struct timeval *tptr = NULL;
1424 int error;
1425
1426 if (SCARG(uap, times)) {
1427 if ((error = copyin(SCARG(uap, times), <v, sizeof(ltv))))
1428 return error;
1429
1430 tv[0].tv_sec = ltv[0].tv_sec;
1431 tv[0].tv_usec = ltv[0].tv_usec;
1432 tv[1].tv_sec = ltv[1].tv_sec;
1433 tv[1].tv_usec = ltv[1].tv_usec;
1434
1435 tptr = tv;
1436 }
1437
1438 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW,
1439 tptr, UIO_SYSSPACE);
1440 }
1441 #endif /* __alpha__ */
1442 #endif /* !COMPAT_LINUX32 */
1443