1 /* $NetBSD: linux_misc.c,v 1.135.2.2 2005/11/01 22:31:17 tron 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 1.135.2.2 2005/11/01 22:31:17 tron 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/sa.h>
103 #include <sys/syscallargs.h>
104
105 #include <compat/linux/common/linux_types.h>
106 #include <compat/linux/common/linux_signal.h>
107
108 #include <compat/linux/linux_syscallargs.h>
109
110 #include <compat/linux/common/linux_fcntl.h>
111 #include <compat/linux/common/linux_mmap.h>
112 #include <compat/linux/common/linux_dirent.h>
113 #include <compat/linux/common/linux_util.h>
114 #include <compat/linux/common/linux_misc.h>
115 #include <compat/linux/common/linux_ptrace.h>
116 #include <compat/linux/common/linux_reboot.h>
117 #include <compat/linux/common/linux_emuldata.h>
118
119 const int linux_ptrace_request_map[] = {
120 LINUX_PTRACE_TRACEME, PT_TRACE_ME,
121 LINUX_PTRACE_PEEKTEXT, PT_READ_I,
122 LINUX_PTRACE_PEEKDATA, PT_READ_D,
123 LINUX_PTRACE_POKETEXT, PT_WRITE_I,
124 LINUX_PTRACE_POKEDATA, PT_WRITE_D,
125 LINUX_PTRACE_CONT, PT_CONTINUE,
126 LINUX_PTRACE_KILL, PT_KILL,
127 LINUX_PTRACE_ATTACH, PT_ATTACH,
128 LINUX_PTRACE_DETACH, PT_DETACH,
129 #ifdef PT_STEP
130 LINUX_PTRACE_SINGLESTEP, PT_STEP,
131 #endif
132 -1
133 };
134
135 const struct linux_mnttypes linux_fstypes[] = {
136 { MOUNT_FFS, LINUX_DEFAULT_SUPER_MAGIC },
137 { MOUNT_NFS, LINUX_NFS_SUPER_MAGIC },
138 { MOUNT_MFS, LINUX_DEFAULT_SUPER_MAGIC },
139 { MOUNT_MSDOS, LINUX_MSDOS_SUPER_MAGIC },
140 { MOUNT_LFS, LINUX_DEFAULT_SUPER_MAGIC },
141 { MOUNT_FDESC, LINUX_DEFAULT_SUPER_MAGIC },
142 { MOUNT_PORTAL, LINUX_DEFAULT_SUPER_MAGIC },
143 { MOUNT_NULL, LINUX_DEFAULT_SUPER_MAGIC },
144 { MOUNT_OVERLAY, LINUX_DEFAULT_SUPER_MAGIC },
145 { MOUNT_UMAP, LINUX_DEFAULT_SUPER_MAGIC },
146 { MOUNT_KERNFS, LINUX_DEFAULT_SUPER_MAGIC },
147 { MOUNT_PROCFS, LINUX_PROC_SUPER_MAGIC },
148 { MOUNT_AFS, LINUX_DEFAULT_SUPER_MAGIC },
149 { MOUNT_CD9660, LINUX_ISOFS_SUPER_MAGIC },
150 { MOUNT_UNION, LINUX_DEFAULT_SUPER_MAGIC },
151 { MOUNT_ADOSFS, LINUX_ADFS_SUPER_MAGIC },
152 { MOUNT_EXT2FS, LINUX_EXT2_SUPER_MAGIC },
153 { MOUNT_CFS, LINUX_DEFAULT_SUPER_MAGIC },
154 { MOUNT_CODA, LINUX_CODA_SUPER_MAGIC },
155 { MOUNT_FILECORE, LINUX_DEFAULT_SUPER_MAGIC },
156 { MOUNT_NTFS, LINUX_DEFAULT_SUPER_MAGIC },
157 { MOUNT_SMBFS, LINUX_SMB_SUPER_MAGIC },
158 { MOUNT_PTYFS, LINUX_DEVPTS_SUPER_MAGIC }
159 };
160 const int linux_fstypes_cnt = sizeof(linux_fstypes) / sizeof(linux_fstypes[0]);
161
162 #ifdef DEBUG_LINUX
163 #define DPRINTF(a) uprintf a
164 #else
165 #define DPRINTF(a)
166 #endif
167
168 /* Local linux_misc.c functions: */
169 static void bsd_to_linux_statfs __P((const struct statvfs *,
170 struct linux_statfs *));
171 static int linux_to_bsd_limit __P((int));
172 static void linux_to_bsd_mmap_args __P((struct sys_mmap_args *,
173 const struct linux_sys_mmap_args *));
174 static int linux_mmap __P((struct lwp *, struct linux_sys_mmap_args *,
175 register_t *, off_t));
176
177
178 /*
179 * The information on a terminated (or stopped) process needs
180 * to be converted in order for Linux binaries to get a valid signal
181 * number out of it.
182 */
183 void
184 bsd_to_linux_wstat(st)
185 int *st;
186 {
187
188 int sig;
189
190 if (WIFSIGNALED(*st)) {
191 sig = WTERMSIG(*st);
192 if (sig >= 0 && sig < NSIG)
193 *st= (*st& ~0177) | native_to_linux_signo[sig];
194 } else if (WIFSTOPPED(*st)) {
195 sig = WSTOPSIG(*st);
196 if (sig >= 0 && sig < NSIG)
197 *st = (*st & ~0xff00) |
198 (native_to_linux_signo[sig] << 8);
199 }
200 }
201
202 /*
203 * wait4(2). Passed on to the NetBSD call, surrounded by code to
204 * reserve some space for a NetBSD-style wait status, and converting
205 * it to what Linux wants.
206 */
207 int
208 linux_sys_wait4(l, v, retval)
209 struct lwp *l;
210 void *v;
211 register_t *retval;
212 {
213 struct linux_sys_wait4_args /* {
214 syscallarg(int) pid;
215 syscallarg(int *) status;
216 syscallarg(int) options;
217 syscallarg(struct rusage *) rusage;
218 } */ *uap = v;
219 struct proc *p = l->l_proc;
220 struct sys_wait4_args w4a;
221 int error, *status, tstat, options, linux_options;
222 caddr_t sg;
223
224 if (SCARG(uap, status) != NULL) {
225 sg = stackgap_init(p, 0);
226 status = (int *) stackgap_alloc(p, &sg, sizeof *status);
227 } else
228 status = NULL;
229
230 linux_options = SCARG(uap, options);
231 options = 0;
232 if (linux_options & ~(LINUX_WAIT4_KNOWNFLAGS))
233 return (EINVAL);
234
235 if (linux_options & LINUX_WAIT4_WNOHANG)
236 options |= WNOHANG;
237 if (linux_options & LINUX_WAIT4_WUNTRACED)
238 options |= WUNTRACED;
239 if (linux_options & LINUX_WAIT4_WALL)
240 options |= WALLSIG;
241 if (linux_options & LINUX_WAIT4_WCLONE)
242 options |= WALTSIG;
243 #ifdef DIAGNOSTIC
244 if (linux_options & LINUX_WAIT4_WNOTHREAD)
245 printf("WARNING: %s: linux process %d.%d called "
246 "waitpid with __WNOTHREAD set!",
247 __FILE__, p->p_pid, l->l_lid);
248
249 #endif
250
251 SCARG(&w4a, pid) = SCARG(uap, pid);
252 SCARG(&w4a, status) = status;
253 SCARG(&w4a, options) = options;
254 SCARG(&w4a, rusage) = SCARG(uap, rusage);
255
256 if ((error = sys_wait4(l, &w4a, retval)))
257 return error;
258
259 sigdelset(&p->p_sigctx.ps_siglist, SIGCHLD);
260
261 if (status != NULL) {
262 if ((error = copyin(status, &tstat, sizeof tstat)))
263 return error;
264
265 bsd_to_linux_wstat(&tstat);
266 return copyout(&tstat, SCARG(uap, status), sizeof tstat);
267 }
268
269 return 0;
270 }
271
272 /*
273 * Linux brk(2). The check if the new address is >= the old one is
274 * done in the kernel in Linux. NetBSD does it in the library.
275 */
276 int
277 linux_sys_brk(l, v, retval)
278 struct lwp *l;
279 void *v;
280 register_t *retval;
281 {
282 struct linux_sys_brk_args /* {
283 syscallarg(char *) nsize;
284 } */ *uap = v;
285 struct proc *p = l->l_proc;
286 char *nbrk = SCARG(uap, nsize);
287 struct sys_obreak_args oba;
288 struct vmspace *vm = p->p_vmspace;
289 struct linux_emuldata *ed = (struct linux_emuldata*)p->p_emuldata;
290
291 SCARG(&oba, nsize) = nbrk;
292
293 if ((caddr_t) nbrk > vm->vm_daddr && sys_obreak(l, &oba, retval) == 0)
294 ed->s->p_break = (char*)nbrk;
295 else
296 nbrk = ed->s->p_break;
297
298 retval[0] = (register_t)nbrk;
299
300 return 0;
301 }
302
303 /*
304 * Convert NetBSD statvfs structure to Linux statfs structure.
305 * Linux doesn't have f_flag, and we can't set f_frsize due
306 * to glibc statvfs() bug (see below).
307 */
308 static void
309 bsd_to_linux_statfs(bsp, lsp)
310 const struct statvfs *bsp;
311 struct linux_statfs *lsp;
312 {
313 int i;
314
315 for (i = 0; i < linux_fstypes_cnt; i++) {
316 if (strcmp(bsp->f_fstypename, linux_fstypes[i].bsd) == 0) {
317 lsp->l_ftype = linux_fstypes[i].linux;
318 break;
319 }
320 }
321
322 if (i == linux_fstypes_cnt) {
323 DPRINTF(("unhandled fstype in linux emulation: %s\n",
324 bsp->f_fstypename));
325 lsp->l_ftype = LINUX_DEFAULT_SUPER_MAGIC;
326 }
327
328 /*
329 * The sizes are expressed in number of blocks. The block
330 * size used for the size is f_frsize for POSIX-compliant
331 * statvfs. Linux statfs uses f_bsize as the block size
332 * (f_frsize used to not be available in Linux struct statfs).
333 * However, glibc 2.3.3 statvfs() wrapper fails to adjust the block
334 * counts for different f_frsize if f_frsize is provided by the kernel.
335 * POSIX conforming apps thus get wrong size if f_frsize
336 * is different to f_bsize. Thus, we just pretend we don't
337 * support f_frsize.
338 */
339
340 lsp->l_fbsize = bsp->f_frsize;
341 lsp->l_ffrsize = 0; /* compat */
342 lsp->l_fblocks = bsp->f_blocks;
343 lsp->l_fbfree = bsp->f_bfree;
344 lsp->l_fbavail = bsp->f_bavail;
345 lsp->l_ffiles = bsp->f_files;
346 lsp->l_fffree = bsp->f_ffree;
347 /* Linux sets the fsid to 0..., we don't */
348 lsp->l_ffsid.val[0] = bsp->f_fsidx.__fsid_val[0];
349 lsp->l_ffsid.val[1] = bsp->f_fsidx.__fsid_val[1];
350 lsp->l_fnamelen = bsp->f_namemax;
351 (void)memset(lsp->l_fspare, 0, sizeof(lsp->l_fspare));
352 }
353
354 /*
355 * Implement the fs stat functions. Straightforward.
356 */
357 int
358 linux_sys_statfs(l, v, retval)
359 struct lwp *l;
360 void *v;
361 register_t *retval;
362 {
363 struct linux_sys_statfs_args /* {
364 syscallarg(const char *) path;
365 syscallarg(struct linux_statfs *) sp;
366 } */ *uap = v;
367 struct proc *p = l->l_proc;
368 struct statvfs btmp, *bsp;
369 struct linux_statfs ltmp;
370 struct sys_statvfs1_args bsa;
371 caddr_t sg;
372 int error;
373
374 sg = stackgap_init(p, 0);
375 bsp = (struct statvfs *) stackgap_alloc(p, &sg, sizeof (struct statvfs));
376
377 CHECK_ALT_EXIST(p, &sg, SCARG(uap, path));
378
379 SCARG(&bsa, path) = SCARG(uap, path);
380 SCARG(&bsa, buf) = bsp;
381 SCARG(&bsa, flags) = ST_WAIT;
382
383 if ((error = sys_statvfs1(l, &bsa, retval)))
384 return error;
385
386 if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp)))
387 return error;
388
389 bsd_to_linux_statfs(&btmp, <mp);
390
391 return copyout((caddr_t) <mp, (caddr_t) SCARG(uap, sp), sizeof ltmp);
392 }
393
394 int
395 linux_sys_fstatfs(l, v, retval)
396 struct lwp *l;
397 void *v;
398 register_t *retval;
399 {
400 struct linux_sys_fstatfs_args /* {
401 syscallarg(int) fd;
402 syscallarg(struct linux_statfs *) sp;
403 } */ *uap = v;
404 struct proc *p = l->l_proc;
405 struct statvfs btmp, *bsp;
406 struct linux_statfs ltmp;
407 struct sys_fstatvfs1_args bsa;
408 caddr_t sg;
409 int error;
410
411 sg = stackgap_init(p, 0);
412 bsp = (struct statvfs *) stackgap_alloc(p, &sg, sizeof (struct statvfs));
413
414 SCARG(&bsa, fd) = SCARG(uap, fd);
415 SCARG(&bsa, buf) = bsp;
416 SCARG(&bsa, flags) = ST_WAIT;
417
418 if ((error = sys_fstatvfs1(l, &bsa, retval)))
419 return error;
420
421 if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp)))
422 return error;
423
424 bsd_to_linux_statfs(&btmp, <mp);
425
426 return copyout((caddr_t) <mp, (caddr_t) SCARG(uap, sp), sizeof ltmp);
427 }
428
429 /*
430 * uname(). Just copy the info from the various strings stored in the
431 * kernel, and put it in the Linux utsname structure. That structure
432 * is almost the same as the NetBSD one, only it has fields 65 characters
433 * long, and an extra domainname field.
434 */
435 int
436 linux_sys_uname(l, v, retval)
437 struct lwp *l;
438 void *v;
439 register_t *retval;
440 {
441 struct linux_sys_uname_args /* {
442 syscallarg(struct linux_utsname *) up;
443 } */ *uap = v;
444 struct linux_utsname luts;
445
446 strncpy(luts.l_sysname, linux_sysname, sizeof(luts.l_sysname));
447 strncpy(luts.l_nodename, hostname, sizeof(luts.l_nodename));
448 strncpy(luts.l_release, linux_release, sizeof(luts.l_release));
449 strncpy(luts.l_version, linux_version, sizeof(luts.l_version));
450 strncpy(luts.l_machine, machine, sizeof(luts.l_machine));
451 strncpy(luts.l_domainname, domainname, sizeof(luts.l_domainname));
452
453 return copyout(&luts, SCARG(uap, up), sizeof(luts));
454 }
455
456 /* Used directly on: alpha, mips, ppc, sparc, sparc64 */
457 /* Used indirectly on: arm, i386, m68k */
458
459 /*
460 * New type Linux mmap call.
461 * Only called directly on machines with >= 6 free regs.
462 */
463 int
464 linux_sys_mmap(l, v, retval)
465 struct lwp *l;
466 void *v;
467 register_t *retval;
468 {
469 struct linux_sys_mmap_args /* {
470 syscallarg(unsigned long) addr;
471 syscallarg(size_t) len;
472 syscallarg(int) prot;
473 syscallarg(int) flags;
474 syscallarg(int) fd;
475 syscallarg(linux_off_t) offset;
476 } */ *uap = v;
477
478 if (SCARG(uap, offset) & PAGE_MASK)
479 return EINVAL;
480
481 return linux_mmap(l, uap, retval, SCARG(uap, offset));
482 }
483
484 /*
485 * Guts of most architectures' mmap64() implementations. This shares
486 * its list of arguments with linux_sys_mmap().
487 *
488 * The difference in linux_sys_mmap2() is that "offset" is actually
489 * (offset / pagesize), not an absolute byte count. This translation
490 * to pagesize offsets is done inside glibc between the mmap64() call
491 * point, and the actual syscall.
492 */
493 int
494 linux_sys_mmap2(l, v, retval)
495 struct lwp *l;
496 void *v;
497 register_t *retval;
498 {
499 struct linux_sys_mmap2_args /* {
500 syscallarg(unsigned long) addr;
501 syscallarg(size_t) len;
502 syscallarg(int) prot;
503 syscallarg(int) flags;
504 syscallarg(int) fd;
505 syscallarg(linux_off_t) offset;
506 } */ *uap = v;
507
508 return linux_mmap(l, uap, retval,
509 ((off_t)SCARG(uap, offset)) << PAGE_SHIFT);
510 }
511
512 /*
513 * Massage arguments and call system mmap(2).
514 */
515 static int
516 linux_mmap(l, uap, retval, offset)
517 struct lwp *l;
518 struct linux_sys_mmap_args *uap;
519 register_t *retval;
520 off_t offset;
521 {
522 struct sys_mmap_args cma;
523 int error;
524 size_t mmoff=0;
525
526 if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN) {
527 /*
528 * Request for stack-like memory segment. On linux, this
529 * works by mmap()ping (small) segment, which is automatically
530 * extended when page fault happens below the currently
531 * allocated area. We emulate this by allocating (typically
532 * bigger) segment sized at current stack size limit, and
533 * offsetting the requested and returned address accordingly.
534 * Since physical pages are only allocated on-demand, this
535 * is effectively identical.
536 */
537 rlim_t ssl = l->l_proc->p_rlimit[RLIMIT_STACK].rlim_cur;
538
539 if (SCARG(uap, len) < ssl) {
540 /* Compute the address offset */
541 mmoff = round_page(ssl) - SCARG(uap, len);
542
543 if (SCARG(uap, addr))
544 SCARG(uap, addr) -= mmoff;
545
546 SCARG(uap, len) = (size_t) ssl;
547 }
548 }
549
550 linux_to_bsd_mmap_args(&cma, uap);
551 SCARG(&cma, pos) = offset;
552
553 error = sys_mmap(l, &cma, retval);
554 if (error)
555 return (error);
556
557 /* Shift the returned address for stack-like segment if necessary */
558 if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN && mmoff)
559 retval[0] += mmoff;
560
561 return (0);
562 }
563
564 static void
565 linux_to_bsd_mmap_args(cma, uap)
566 struct sys_mmap_args *cma;
567 const struct linux_sys_mmap_args *uap;
568 {
569 int flags = MAP_TRYFIXED, fl = SCARG(uap, flags);
570
571 flags |= cvtto_bsd_mask(fl, LINUX_MAP_SHARED, MAP_SHARED);
572 flags |= cvtto_bsd_mask(fl, LINUX_MAP_PRIVATE, MAP_PRIVATE);
573 flags |= cvtto_bsd_mask(fl, LINUX_MAP_FIXED, MAP_FIXED);
574 flags |= cvtto_bsd_mask(fl, LINUX_MAP_ANON, MAP_ANON);
575 /* XXX XAX ERH: Any other flags here? There are more defined... */
576
577 SCARG(cma, addr) = (void *)SCARG(uap, addr);
578 SCARG(cma, len) = SCARG(uap, len);
579 SCARG(cma, prot) = SCARG(uap, prot);
580 if (SCARG(cma, prot) & VM_PROT_WRITE) /* XXX */
581 SCARG(cma, prot) |= VM_PROT_READ;
582 SCARG(cma, flags) = flags;
583 SCARG(cma, fd) = flags & MAP_ANON ? -1 : SCARG(uap, fd);
584 SCARG(cma, pad) = 0;
585 }
586
587 int
588 linux_sys_mremap(l, v, retval)
589 struct lwp *l;
590 void *v;
591 register_t *retval;
592 {
593 struct linux_sys_mremap_args /* {
594 syscallarg(void *) old_address;
595 syscallarg(size_t) old_size;
596 syscallarg(size_t) new_size;
597 syscallarg(u_long) flags;
598 } */ *uap = v;
599 struct sys_munmap_args mua;
600 size_t old_size, new_size;
601 int error;
602
603 old_size = round_page(SCARG(uap, old_size));
604 new_size = round_page(SCARG(uap, new_size));
605
606 /*
607 * Growing mapped region.
608 */
609 if (new_size > old_size) {
610 /*
611 * XXX Implement me. What we probably want to do is
612 * XXX dig out the guts of the old mapping, mmap that
613 * XXX object again with the new size, then munmap
614 * XXX the old mapping.
615 */
616 *retval = 0;
617 return (ENOMEM);
618 }
619
620 /*
621 * Shrinking mapped region.
622 */
623 if (new_size < old_size) {
624 SCARG(&mua, addr) = (caddr_t)SCARG(uap, old_address) +
625 new_size;
626 SCARG(&mua, len) = old_size - new_size;
627 error = sys_munmap(l, &mua, retval);
628 *retval = error ? 0 : (register_t)SCARG(uap, old_address);
629 return (error);
630 }
631
632 /*
633 * No change.
634 */
635 *retval = (register_t)SCARG(uap, old_address);
636 return (0);
637 }
638
639 int
640 linux_sys_msync(l, v, retval)
641 struct lwp *l;
642 void *v;
643 register_t *retval;
644 {
645 struct linux_sys_msync_args /* {
646 syscallarg(caddr_t) addr;
647 syscallarg(int) len;
648 syscallarg(int) fl;
649 } */ *uap = v;
650
651 struct sys___msync13_args bma;
652
653 /* flags are ignored */
654 SCARG(&bma, addr) = SCARG(uap, addr);
655 SCARG(&bma, len) = SCARG(uap, len);
656 SCARG(&bma, flags) = SCARG(uap, fl);
657
658 return sys___msync13(l, &bma, retval);
659 }
660
661 int
662 linux_sys_mprotect(l, v, retval)
663 struct lwp *l;
664 void *v;
665 register_t *retval;
666 {
667 struct linux_sys_mprotect_args /* {
668 syscallarg(const void *) start;
669 syscallarg(unsigned long) len;
670 syscallarg(int) prot;
671 } */ *uap = v;
672 struct vm_map_entry *entry;
673 struct vm_map *map;
674 struct proc *p;
675 vaddr_t end, start, len, stacklim;
676 int prot, grows;
677
678 start = (vaddr_t)SCARG(uap, start);
679 len = round_page(SCARG(uap, len));
680 prot = SCARG(uap, prot);
681 grows = prot & (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP);
682 prot &= ~grows;
683 end = start + len;
684
685 if (start & PAGE_MASK)
686 return EINVAL;
687 if (end < start)
688 return EINVAL;
689 if (end == start)
690 return 0;
691
692 if (prot & ~(PROT_READ | PROT_WRITE | PROT_EXEC))
693 return EINVAL;
694 if (grows == (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP))
695 return EINVAL;
696
697 p = l->l_proc;
698 map = &p->p_vmspace->vm_map;
699 vm_map_lock(map);
700 #ifdef notdef
701 VM_MAP_RANGE_CHECK(map, start, end);
702 #endif
703 if (!uvm_map_lookup_entry(map, start, &entry) || entry->start > start) {
704 vm_map_unlock(map);
705 return ENOMEM;
706 }
707
708 /*
709 * Approximate the behaviour of PROT_GROWS{DOWN,UP}.
710 */
711
712 stacklim = (vaddr_t)p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur;
713 if (grows & LINUX_PROT_GROWSDOWN) {
714 if (USRSTACK - stacklim <= start && start < USRSTACK) {
715 start = USRSTACK - stacklim;
716 } else {
717 start = entry->start;
718 }
719 } else if (grows & LINUX_PROT_GROWSUP) {
720 if (USRSTACK <= end && end < USRSTACK + stacklim) {
721 end = USRSTACK + stacklim;
722 } else {
723 end = entry->end;
724 }
725 }
726 vm_map_unlock(map);
727 return uvm_map_protect(map, start, end, prot, FALSE);
728 }
729
730 /*
731 * This code is partly stolen from src/lib/libc/compat-43/times.c
732 */
733
734 #define CONVTCK(r) (r.tv_sec * hz + r.tv_usec / (1000000 / hz))
735
736 int
737 linux_sys_times(l, v, retval)
738 struct lwp *l;
739 void *v;
740 register_t *retval;
741 {
742 struct linux_sys_times_args /* {
743 syscallarg(struct times *) tms;
744 } */ *uap = v;
745 struct proc *p = l->l_proc;
746 struct timeval t;
747 int error, s;
748
749 if (SCARG(uap, tms)) {
750 struct linux_tms ltms;
751 struct rusage ru;
752
753 calcru(p, &ru.ru_utime, &ru.ru_stime, NULL);
754 ltms.ltms_utime = CONVTCK(ru.ru_utime);
755 ltms.ltms_stime = CONVTCK(ru.ru_stime);
756
757 ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime);
758 ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime);
759
760 if ((error = copyout(<ms, SCARG(uap, tms), sizeof ltms)))
761 return error;
762 }
763
764 s = splclock();
765 timersub(&time, &boottime, &t);
766 splx(s);
767
768 retval[0] = ((linux_clock_t)(CONVTCK(t)));
769 return 0;
770 }
771
772 #undef CONVTCK
773
774 /*
775 * Linux 'readdir' call. This code is mostly taken from the
776 * SunOS getdents call (see compat/sunos/sunos_misc.c), though
777 * an attempt has been made to keep it a little cleaner (failing
778 * miserably, because of the cruft needed if count 1 is passed).
779 *
780 * The d_off field should contain the offset of the next valid entry,
781 * but in Linux it has the offset of the entry itself. We emulate
782 * that bug here.
783 *
784 * Read in BSD-style entries, convert them, and copy them out.
785 *
786 * Note that this doesn't handle union-mounted filesystems.
787 */
788 int
789 linux_sys_getdents(l, v, retval)
790 struct lwp *l;
791 void *v;
792 register_t *retval;
793 {
794 struct linux_sys_getdents_args /* {
795 syscallarg(int) fd;
796 syscallarg(struct linux_dirent *) dent;
797 syscallarg(unsigned int) count;
798 } */ *uap = v;
799 struct proc *p = l->l_proc;
800 struct dirent *bdp;
801 struct vnode *vp;
802 caddr_t inp, buf; /* BSD-format */
803 int len, reclen; /* BSD-format */
804 caddr_t outp; /* Linux-format */
805 int resid, linux_reclen = 0; /* Linux-format */
806 struct file *fp;
807 struct uio auio;
808 struct iovec aiov;
809 struct linux_dirent idb;
810 off_t off; /* true file offset */
811 int buflen, error, eofflag, nbytes, oldcall;
812 struct vattr va;
813 off_t *cookiebuf = NULL, *cookie;
814 int ncookies;
815
816 /* getvnode() will use the descriptor for us */
817 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
818 return (error);
819
820 if ((fp->f_flag & FREAD) == 0) {
821 error = EBADF;
822 goto out1;
823 }
824
825 vp = (struct vnode *)fp->f_data;
826 if (vp->v_type != VDIR) {
827 error = EINVAL;
828 goto out1;
829 }
830
831 if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)))
832 goto out1;
833
834 nbytes = SCARG(uap, count);
835 if (nbytes == 1) { /* emulating old, broken behaviour */
836 nbytes = sizeof (idb);
837 buflen = max(va.va_blocksize, nbytes);
838 oldcall = 1;
839 } else {
840 buflen = min(MAXBSIZE, nbytes);
841 if (buflen < va.va_blocksize)
842 buflen = va.va_blocksize;
843 oldcall = 0;
844 }
845 buf = malloc(buflen, M_TEMP, M_WAITOK);
846
847 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
848 off = fp->f_offset;
849 again:
850 aiov.iov_base = buf;
851 aiov.iov_len = buflen;
852 auio.uio_iov = &aiov;
853 auio.uio_iovcnt = 1;
854 auio.uio_rw = UIO_READ;
855 auio.uio_segflg = UIO_SYSSPACE;
856 auio.uio_procp = NULL;
857 auio.uio_resid = buflen;
858 auio.uio_offset = off;
859 /*
860 * First we read into the malloc'ed buffer, then
861 * we massage it into user space, one record at a time.
862 */
863 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf,
864 &ncookies);
865 if (error)
866 goto out;
867
868 inp = buf;
869 outp = (caddr_t)SCARG(uap, dent);
870 resid = nbytes;
871 if ((len = buflen - auio.uio_resid) == 0)
872 goto eof;
873
874 for (cookie = cookiebuf; len > 0; len -= reclen) {
875 bdp = (struct dirent *)inp;
876 reclen = bdp->d_reclen;
877 if (reclen & 3)
878 panic("linux_readdir");
879 if (bdp->d_fileno == 0) {
880 inp += reclen; /* it is a hole; squish it out */
881 if (cookie)
882 off = *cookie++;
883 else
884 off += reclen;
885 continue;
886 }
887 linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen);
888 if (reclen > len || resid < linux_reclen) {
889 /* entry too big for buffer, so just stop */
890 outp++;
891 break;
892 }
893 /*
894 * Massage in place to make a Linux-shaped dirent (otherwise
895 * we have to worry about touching user memory outside of
896 * the copyout() call).
897 */
898 idb.d_ino = bdp->d_fileno;
899 /*
900 * The old readdir() call misuses the offset and reclen fields.
901 */
902 if (oldcall) {
903 idb.d_off = (linux_off_t)linux_reclen;
904 idb.d_reclen = (u_short)bdp->d_namlen;
905 } else {
906 if (sizeof (idb.d_off) <= 4 && (off >> 32) != 0) {
907 compat_offseterr(vp, "linux_getdents");
908 error = EINVAL;
909 goto out;
910 }
911 idb.d_off = (linux_off_t)off;
912 idb.d_reclen = (u_short)linux_reclen;
913 }
914 strcpy(idb.d_name, bdp->d_name);
915 if ((error = copyout((caddr_t)&idb, outp, linux_reclen)))
916 goto out;
917 /* advance past this real entry */
918 inp += reclen;
919 if (cookie)
920 off = *cookie++; /* each entry points to itself */
921 else
922 off += reclen;
923 /* advance output past Linux-shaped entry */
924 outp += linux_reclen;
925 resid -= linux_reclen;
926 if (oldcall)
927 break;
928 }
929
930 /* if we squished out the whole block, try again */
931 if (outp == (caddr_t)SCARG(uap, dent))
932 goto again;
933 fp->f_offset = off; /* update the vnode offset */
934
935 if (oldcall)
936 nbytes = resid + linux_reclen;
937
938 eof:
939 *retval = nbytes - resid;
940 out:
941 VOP_UNLOCK(vp, 0);
942 if (cookiebuf)
943 free(cookiebuf, M_TEMP);
944 free(buf, M_TEMP);
945 out1:
946 FILE_UNUSE(fp, p);
947 return error;
948 }
949
950 /*
951 * Even when just using registers to pass arguments to syscalls you can
952 * have 5 of them on the i386. So this newer version of select() does
953 * this.
954 */
955 int
956 linux_sys_select(l, v, retval)
957 struct lwp *l;
958 void *v;
959 register_t *retval;
960 {
961 struct linux_sys_select_args /* {
962 syscallarg(int) nfds;
963 syscallarg(fd_set *) readfds;
964 syscallarg(fd_set *) writefds;
965 syscallarg(fd_set *) exceptfds;
966 syscallarg(struct timeval *) timeout;
967 } */ *uap = v;
968
969 return linux_select1(l, retval, SCARG(uap, nfds), SCARG(uap, readfds),
970 SCARG(uap, writefds), SCARG(uap, exceptfds), SCARG(uap, timeout));
971 }
972
973 /*
974 * Common code for the old and new versions of select(). A couple of
975 * things are important:
976 * 1) return the amount of time left in the 'timeout' parameter
977 * 2) select never returns ERESTART on Linux, always return EINTR
978 */
979 int
980 linux_select1(l, retval, nfds, readfds, writefds, exceptfds, timeout)
981 struct lwp *l;
982 register_t *retval;
983 int nfds;
984 fd_set *readfds, *writefds, *exceptfds;
985 struct timeval *timeout;
986 {
987 struct sys_select_args bsa;
988 struct proc *p = l->l_proc;
989 struct timeval tv0, tv1, utv, *tvp;
990 caddr_t sg;
991 int error;
992
993 SCARG(&bsa, nd) = nfds;
994 SCARG(&bsa, in) = readfds;
995 SCARG(&bsa, ou) = writefds;
996 SCARG(&bsa, ex) = exceptfds;
997 SCARG(&bsa, tv) = timeout;
998
999 /*
1000 * Store current time for computation of the amount of
1001 * time left.
1002 */
1003 if (timeout) {
1004 if ((error = copyin(timeout, &utv, sizeof(utv))))
1005 return error;
1006 if (itimerfix(&utv)) {
1007 /*
1008 * The timeval was invalid. Convert it to something
1009 * valid that will act as it does under Linux.
1010 */
1011 sg = stackgap_init(p, 0);
1012 tvp = stackgap_alloc(p, &sg, sizeof(utv));
1013 utv.tv_sec += utv.tv_usec / 1000000;
1014 utv.tv_usec %= 1000000;
1015 if (utv.tv_usec < 0) {
1016 utv.tv_sec -= 1;
1017 utv.tv_usec += 1000000;
1018 }
1019 if (utv.tv_sec < 0)
1020 timerclear(&utv);
1021 if ((error = copyout(&utv, tvp, sizeof(utv))))
1022 return error;
1023 SCARG(&bsa, tv) = tvp;
1024 }
1025 microtime(&tv0);
1026 }
1027
1028 error = sys_select(l, &bsa, retval);
1029 if (error) {
1030 /*
1031 * See fs/select.c in the Linux kernel. Without this,
1032 * Maelstrom doesn't work.
1033 */
1034 if (error == ERESTART)
1035 error = EINTR;
1036 return error;
1037 }
1038
1039 if (timeout) {
1040 if (*retval) {
1041 /*
1042 * Compute how much time was left of the timeout,
1043 * by subtracting the current time and the time
1044 * before we started the call, and subtracting
1045 * that result from the user-supplied value.
1046 */
1047 microtime(&tv1);
1048 timersub(&tv1, &tv0, &tv1);
1049 timersub(&utv, &tv1, &utv);
1050 if (utv.tv_sec < 0)
1051 timerclear(&utv);
1052 } else
1053 timerclear(&utv);
1054 if ((error = copyout(&utv, timeout, sizeof(utv))))
1055 return error;
1056 }
1057
1058 return 0;
1059 }
1060
1061 /*
1062 * Get the process group of a certain process. Look it up
1063 * and return the value.
1064 */
1065 int
1066 linux_sys_getpgid(l, v, retval)
1067 struct lwp *l;
1068 void *v;
1069 register_t *retval;
1070 {
1071 struct linux_sys_getpgid_args /* {
1072 syscallarg(int) pid;
1073 } */ *uap = v;
1074 struct proc *p = l->l_proc;
1075 struct proc *targp;
1076
1077 if (SCARG(uap, pid) != 0 && SCARG(uap, pid) != p->p_pid) {
1078 if ((targp = pfind(SCARG(uap, pid))) == 0)
1079 return ESRCH;
1080 }
1081 else
1082 targp = p;
1083
1084 retval[0] = targp->p_pgid;
1085 return 0;
1086 }
1087
1088 /*
1089 * Set the 'personality' (emulation mode) for the current process. Only
1090 * accept the Linux personality here (0). This call is needed because
1091 * the Linux ELF crt0 issues it in an ugly kludge to make sure that
1092 * ELF binaries run in Linux mode, not SVR4 mode.
1093 */
1094 int
1095 linux_sys_personality(l, v, retval)
1096 struct lwp *l;
1097 void *v;
1098 register_t *retval;
1099 {
1100 struct linux_sys_personality_args /* {
1101 syscallarg(int) per;
1102 } */ *uap = v;
1103
1104 if (SCARG(uap, per) != 0)
1105 return EINVAL;
1106 retval[0] = 0;
1107 return 0;
1108 }
1109
1110 #if defined(__i386__) || defined(__m68k__)
1111 /*
1112 * The calls are here because of type conversions.
1113 */
1114 int
1115 linux_sys_setreuid16(l, v, retval)
1116 struct lwp *l;
1117 void *v;
1118 register_t *retval;
1119 {
1120 struct linux_sys_setreuid16_args /* {
1121 syscallarg(int) ruid;
1122 syscallarg(int) euid;
1123 } */ *uap = v;
1124 struct sys_setreuid_args bsa;
1125
1126 SCARG(&bsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
1127 (uid_t)-1 : SCARG(uap, ruid);
1128 SCARG(&bsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
1129 (uid_t)-1 : SCARG(uap, euid);
1130
1131 return sys_setreuid(l, &bsa, retval);
1132 }
1133
1134 int
1135 linux_sys_setregid16(l, v, retval)
1136 struct lwp *l;
1137 void *v;
1138 register_t *retval;
1139 {
1140 struct linux_sys_setregid16_args /* {
1141 syscallarg(int) rgid;
1142 syscallarg(int) egid;
1143 } */ *uap = v;
1144 struct sys_setregid_args bsa;
1145
1146 SCARG(&bsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
1147 (uid_t)-1 : SCARG(uap, rgid);
1148 SCARG(&bsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
1149 (uid_t)-1 : SCARG(uap, egid);
1150
1151 return sys_setregid(l, &bsa, retval);
1152 }
1153
1154 int
1155 linux_sys_setresuid16(l, v, retval)
1156 struct lwp *l;
1157 void *v;
1158 register_t *retval;
1159 {
1160 struct linux_sys_setresuid16_args /* {
1161 syscallarg(uid_t) ruid;
1162 syscallarg(uid_t) euid;
1163 syscallarg(uid_t) suid;
1164 } */ *uap = v;
1165 struct linux_sys_setresuid16_args lsa;
1166
1167 SCARG(&lsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
1168 (uid_t)-1 : SCARG(uap, ruid);
1169 SCARG(&lsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
1170 (uid_t)-1 : SCARG(uap, euid);
1171 SCARG(&lsa, suid) = ((linux_uid_t)SCARG(uap, suid) == (linux_uid_t)-1) ?
1172 (uid_t)-1 : SCARG(uap, suid);
1173
1174 return linux_sys_setresuid(l, &lsa, retval);
1175 }
1176
1177 int
1178 linux_sys_setresgid16(l, v, retval)
1179 struct lwp *l;
1180 void *v;
1181 register_t *retval;
1182 {
1183 struct linux_sys_setresgid16_args /* {
1184 syscallarg(gid_t) rgid;
1185 syscallarg(gid_t) egid;
1186 syscallarg(gid_t) sgid;
1187 } */ *uap = v;
1188 struct linux_sys_setresgid16_args lsa;
1189
1190 SCARG(&lsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
1191 (gid_t)-1 : SCARG(uap, rgid);
1192 SCARG(&lsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
1193 (gid_t)-1 : SCARG(uap, egid);
1194 SCARG(&lsa, sgid) = ((linux_gid_t)SCARG(uap, sgid) == (linux_gid_t)-1) ?
1195 (gid_t)-1 : SCARG(uap, sgid);
1196
1197 return linux_sys_setresgid(l, &lsa, retval);
1198 }
1199
1200 int
1201 linux_sys_getgroups16(l, v, retval)
1202 struct lwp *l;
1203 void *v;
1204 register_t *retval;
1205 {
1206 struct linux_sys_getgroups16_args /* {
1207 syscallarg(int) gidsetsize;
1208 syscallarg(linux_gid_t *) gidset;
1209 } */ *uap = v;
1210 struct proc *p = l->l_proc;
1211 caddr_t sg;
1212 int n, error, i;
1213 struct sys_getgroups_args bsa;
1214 gid_t *bset, *kbset;
1215 linux_gid_t *lset;
1216 struct pcred *pc = p->p_cred;
1217
1218 n = SCARG(uap, gidsetsize);
1219 if (n < 0)
1220 return EINVAL;
1221 error = 0;
1222 bset = kbset = NULL;
1223 lset = NULL;
1224 if (n > 0) {
1225 n = min(pc->pc_ucred->cr_ngroups, n);
1226 sg = stackgap_init(p, 0);
1227 bset = stackgap_alloc(p, &sg, n * sizeof (gid_t));
1228 kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK);
1229 lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
1230 if (bset == NULL || kbset == NULL || lset == NULL)
1231 return ENOMEM;
1232 SCARG(&bsa, gidsetsize) = n;
1233 SCARG(&bsa, gidset) = bset;
1234 error = sys_getgroups(l, &bsa, retval);
1235 if (error != 0)
1236 goto out;
1237 error = copyin(bset, kbset, n * sizeof (gid_t));
1238 if (error != 0)
1239 goto out;
1240 for (i = 0; i < n; i++)
1241 lset[i] = (linux_gid_t)kbset[i];
1242 error = copyout(lset, SCARG(uap, gidset),
1243 n * sizeof (linux_gid_t));
1244 } else
1245 *retval = pc->pc_ucred->cr_ngroups;
1246 out:
1247 if (kbset != NULL)
1248 free(kbset, M_TEMP);
1249 if (lset != NULL)
1250 free(lset, M_TEMP);
1251 return error;
1252 }
1253
1254 int
1255 linux_sys_setgroups16(l, v, retval)
1256 struct lwp *l;
1257 void *v;
1258 register_t *retval;
1259 {
1260 struct linux_sys_setgroups16_args /* {
1261 syscallarg(int) gidsetsize;
1262 syscallarg(linux_gid_t *) gidset;
1263 } */ *uap = v;
1264 struct proc *p = l->l_proc;
1265 caddr_t sg;
1266 int n;
1267 int error, i;
1268 struct sys_setgroups_args bsa;
1269 gid_t *bset, *kbset;
1270 linux_gid_t *lset;
1271
1272 n = SCARG(uap, gidsetsize);
1273 if (n < 0 || n > NGROUPS)
1274 return EINVAL;
1275 sg = stackgap_init(p, 0);
1276 bset = stackgap_alloc(p, &sg, n * sizeof (gid_t));
1277 lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
1278 kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK);
1279 if (lset == NULL || bset == NULL)
1280 return ENOMEM;
1281 error = copyin(SCARG(uap, gidset), lset, n * sizeof (linux_gid_t));
1282 if (error != 0)
1283 goto out;
1284 for (i = 0; i < n; i++)
1285 kbset[i] = (gid_t)lset[i];
1286 error = copyout(kbset, bset, n * sizeof (gid_t));
1287 if (error != 0)
1288 goto out;
1289 SCARG(&bsa, gidsetsize) = n;
1290 SCARG(&bsa, gidset) = bset;
1291 error = sys_setgroups(l, &bsa, retval);
1292
1293 out:
1294 if (lset != NULL)
1295 free(lset, M_TEMP);
1296 if (kbset != NULL)
1297 free(kbset, M_TEMP);
1298
1299 return error;
1300 }
1301
1302 #endif /* __i386__ || __m68k__ */
1303
1304 /*
1305 * We have nonexistent fsuid equal to uid.
1306 * If modification is requested, refuse.
1307 */
1308 int
1309 linux_sys_setfsuid(l, v, retval)
1310 struct lwp *l;
1311 void *v;
1312 register_t *retval;
1313 {
1314 struct linux_sys_setfsuid_args /* {
1315 syscallarg(uid_t) uid;
1316 } */ *uap = v;
1317 struct proc *p = l->l_proc;
1318 uid_t uid;
1319
1320 uid = SCARG(uap, uid);
1321 if (p->p_cred->p_ruid != uid)
1322 return sys_nosys(l, v, retval);
1323 else
1324 return (0);
1325 }
1326
1327 /* XXX XXX XXX */
1328 #ifndef alpha
1329 int
1330 linux_sys_getfsuid(l, v, retval)
1331 struct lwp *l;
1332 void *v;
1333 register_t *retval;
1334 {
1335 return sys_getuid(l, v, retval);
1336 }
1337 #endif
1338
1339 int
1340 linux_sys_setresuid(l, v, retval)
1341 struct lwp *l;
1342 void *v;
1343 register_t *retval;
1344 {
1345 struct linux_sys_setresuid_args /* {
1346 syscallarg(uid_t) ruid;
1347 syscallarg(uid_t) euid;
1348 syscallarg(uid_t) suid;
1349 } */ *uap = v;
1350
1351 /*
1352 * Note: These checks are a little different than the NetBSD
1353 * setreuid(2) call performs. This precisely follows the
1354 * behavior of the Linux kernel.
1355 */
1356
1357 return do_setresuid(l, SCARG(uap, ruid), SCARG(uap, euid),
1358 SCARG(uap, suid),
1359 ID_R_EQ_R | ID_R_EQ_E | ID_R_EQ_S |
1360 ID_E_EQ_R | ID_E_EQ_E | ID_E_EQ_S |
1361 ID_S_EQ_R | ID_S_EQ_E | ID_S_EQ_S );
1362 }
1363
1364 int
1365 linux_sys_getresuid(l, v, retval)
1366 struct lwp *l;
1367 void *v;
1368 register_t *retval;
1369 {
1370 struct linux_sys_getresuid_args /* {
1371 syscallarg(uid_t *) ruid;
1372 syscallarg(uid_t *) euid;
1373 syscallarg(uid_t *) suid;
1374 } */ *uap = v;
1375 struct proc *p = l->l_proc;
1376 struct pcred *pc = p->p_cred;
1377 int error;
1378
1379 /*
1380 * Linux copies these values out to userspace like so:
1381 *
1382 * 1. Copy out ruid.
1383 * 2. If that succeeds, copy out euid.
1384 * 3. If both of those succeed, copy out suid.
1385 */
1386 if ((error = copyout(&pc->p_ruid, SCARG(uap, ruid),
1387 sizeof(uid_t))) != 0)
1388 return (error);
1389
1390 if ((error = copyout(&pc->pc_ucred->cr_uid, SCARG(uap, euid),
1391 sizeof(uid_t))) != 0)
1392 return (error);
1393
1394 return (copyout(&pc->p_svuid, SCARG(uap, suid), sizeof(uid_t)));
1395 }
1396
1397 int
1398 linux_sys_ptrace(l, v, retval)
1399 struct lwp *l;
1400 void *v;
1401 register_t *retval;
1402 {
1403 struct linux_sys_ptrace_args /* {
1404 i386, m68k, powerpc: T=int
1405 alpha: T=long
1406 syscallarg(T) request;
1407 syscallarg(T) pid;
1408 syscallarg(T) addr;
1409 syscallarg(T) data;
1410 } */ *uap = v;
1411 const int *ptr;
1412 int request;
1413 int error;
1414
1415 ptr = linux_ptrace_request_map;
1416 request = SCARG(uap, request);
1417 while (*ptr != -1)
1418 if (*ptr++ == request) {
1419 struct sys_ptrace_args pta;
1420
1421 SCARG(&pta, req) = *ptr;
1422 SCARG(&pta, pid) = SCARG(uap, pid);
1423 SCARG(&pta, addr) = (caddr_t)SCARG(uap, addr);
1424 SCARG(&pta, data) = SCARG(uap, data);
1425
1426 /*
1427 * Linux ptrace(PTRACE_CONT, pid, 0, 0) means actually
1428 * to continue where the process left off previously.
1429 * The same thing is achieved by addr == (caddr_t) 1
1430 * on NetBSD, so rewrite 'addr' appropriately.
1431 */
1432 if (request == LINUX_PTRACE_CONT && SCARG(uap, addr)==0)
1433 SCARG(&pta, addr) = (caddr_t) 1;
1434
1435 error = sys_ptrace(l, &pta, retval);
1436 if (error)
1437 return error;
1438 switch (request) {
1439 case LINUX_PTRACE_PEEKTEXT:
1440 case LINUX_PTRACE_PEEKDATA:
1441 error = copyout (retval,
1442 (caddr_t)SCARG(uap, data), sizeof *retval);
1443 *retval = SCARG(uap, data);
1444 break;
1445 default:
1446 break;
1447 }
1448 return error;
1449 }
1450 else
1451 ptr++;
1452
1453 return LINUX_SYS_PTRACE_ARCH(l, uap, retval);
1454 }
1455
1456 int
1457 linux_sys_reboot(struct lwp *l, void *v, register_t *retval)
1458 {
1459 struct linux_sys_reboot_args /* {
1460 syscallarg(int) magic1;
1461 syscallarg(int) magic2;
1462 syscallarg(int) cmd;
1463 syscallarg(void *) arg;
1464 } */ *uap = v;
1465 struct sys_reboot_args /* {
1466 syscallarg(int) opt;
1467 syscallarg(char *) bootstr;
1468 } */ sra;
1469 struct proc *p = l->l_proc;
1470 int error;
1471
1472 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1473 return(error);
1474
1475 if (SCARG(uap, magic1) != LINUX_REBOOT_MAGIC1)
1476 return(EINVAL);
1477 if (SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2 &&
1478 SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2A &&
1479 SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2B)
1480 return(EINVAL);
1481
1482 switch (SCARG(uap, cmd)) {
1483 case LINUX_REBOOT_CMD_RESTART:
1484 SCARG(&sra, opt) = RB_AUTOBOOT;
1485 break;
1486 case LINUX_REBOOT_CMD_HALT:
1487 SCARG(&sra, opt) = RB_HALT;
1488 break;
1489 case LINUX_REBOOT_CMD_POWER_OFF:
1490 SCARG(&sra, opt) = RB_HALT|RB_POWERDOWN;
1491 break;
1492 case LINUX_REBOOT_CMD_RESTART2:
1493 /* Reboot with an argument. */
1494 SCARG(&sra, opt) = RB_AUTOBOOT|RB_STRING;
1495 SCARG(&sra, bootstr) = SCARG(uap, arg);
1496 break;
1497 case LINUX_REBOOT_CMD_CAD_ON:
1498 return(EINVAL); /* We don't implement ctrl-alt-delete */
1499 case LINUX_REBOOT_CMD_CAD_OFF:
1500 return(0);
1501 default:
1502 return(EINVAL);
1503 }
1504
1505 return(sys_reboot(l, &sra, retval));
1506 }
1507
1508 /*
1509 * Copy of compat_12_sys_swapon().
1510 */
1511 int
1512 linux_sys_swapon(l, v, retval)
1513 struct lwp *l;
1514 void *v;
1515 register_t *retval;
1516 {
1517 struct sys_swapctl_args ua;
1518 struct linux_sys_swapon_args /* {
1519 syscallarg(const char *) name;
1520 } */ *uap = v;
1521
1522 SCARG(&ua, cmd) = SWAP_ON;
1523 SCARG(&ua, arg) = (void *)SCARG(uap, name);
1524 SCARG(&ua, misc) = 0; /* priority */
1525 return (sys_swapctl(l, &ua, retval));
1526 }
1527
1528 /*
1529 * Stop swapping to the file or block device specified by path.
1530 */
1531 int
1532 linux_sys_swapoff(l, v, retval)
1533 struct lwp *l;
1534 void *v;
1535 register_t *retval;
1536 {
1537 struct sys_swapctl_args ua;
1538 struct linux_sys_swapoff_args /* {
1539 syscallarg(const char *) path;
1540 } */ *uap = v;
1541
1542 SCARG(&ua, cmd) = SWAP_OFF;
1543 SCARG(&ua, arg) = (void *)SCARG(uap, path);
1544 return (sys_swapctl(l, &ua, retval));
1545 }
1546
1547 /*
1548 * Copy of compat_09_sys_setdomainname()
1549 */
1550 /* ARGSUSED */
1551 int
1552 linux_sys_setdomainname(l, v, retval)
1553 struct lwp *l;
1554 void *v;
1555 register_t *retval;
1556 {
1557 struct linux_sys_setdomainname_args /* {
1558 syscallarg(char *) domainname;
1559 syscallarg(int) len;
1560 } */ *uap = v;
1561 int name[2];
1562
1563 name[0] = CTL_KERN;
1564 name[1] = KERN_DOMAINNAME;
1565 return (old_sysctl(&name[0], 2, 0, 0, SCARG(uap, domainname),
1566 SCARG(uap, len), l));
1567 }
1568
1569 /*
1570 * sysinfo()
1571 */
1572 /* ARGSUSED */
1573 int
1574 linux_sys_sysinfo(l, v, retval)
1575 struct lwp *l;
1576 void *v;
1577 register_t *retval;
1578 {
1579 struct linux_sys_sysinfo_args /* {
1580 syscallarg(struct linux_sysinfo *) arg;
1581 } */ *uap = v;
1582 struct linux_sysinfo si;
1583 struct loadavg *la;
1584
1585 si.uptime = time.tv_sec - boottime.tv_sec;
1586 la = &averunnable;
1587 si.loads[0] = la->ldavg[0] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1588 si.loads[1] = la->ldavg[1] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1589 si.loads[2] = la->ldavg[2] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1590 si.totalram = ctob(physmem);
1591 si.freeram = uvmexp.free * uvmexp.pagesize;
1592 si.sharedram = 0; /* XXX */
1593 si.bufferram = uvmexp.filepages * uvmexp.pagesize;
1594 si.totalswap = uvmexp.swpages * uvmexp.pagesize;
1595 si.freeswap = (uvmexp.swpages - uvmexp.swpginuse) * uvmexp.pagesize;
1596 si.procs = nprocs;
1597
1598 /* The following are only present in newer Linux kernels. */
1599 si.totalbig = 0;
1600 si.freebig = 0;
1601 si.mem_unit = 1;
1602
1603 return (copyout(&si, SCARG(uap, arg), sizeof si));
1604 }
1605
1606 #define bsd_to_linux_rlimit1(l, b, f) \
1607 (l)->f = ((b)->f == RLIM_INFINITY || \
1608 ((b)->f & 0xffffffff00000000ULL) != 0) ? \
1609 LINUX_RLIM_INFINITY : (int32_t)(b)->f
1610 #define bsd_to_linux_rlimit(l, b) \
1611 bsd_to_linux_rlimit1(l, b, rlim_cur); \
1612 bsd_to_linux_rlimit1(l, b, rlim_max)
1613
1614 #define linux_to_bsd_rlimit1(b, l, f) \
1615 (b)->f = (l)->f == LINUX_RLIM_INFINITY ? RLIM_INFINITY : (l)->f
1616 #define linux_to_bsd_rlimit(b, l) \
1617 linux_to_bsd_rlimit1(b, l, rlim_cur); \
1618 linux_to_bsd_rlimit1(b, l, rlim_max)
1619
1620 static int
1621 linux_to_bsd_limit(lim)
1622 int lim;
1623 {
1624 switch (lim) {
1625 case LINUX_RLIMIT_CPU:
1626 return RLIMIT_CPU;
1627 case LINUX_RLIMIT_FSIZE:
1628 return RLIMIT_FSIZE;
1629 case LINUX_RLIMIT_DATA:
1630 return RLIMIT_DATA;
1631 case LINUX_RLIMIT_STACK:
1632 return RLIMIT_STACK;
1633 case LINUX_RLIMIT_CORE:
1634 return RLIMIT_CORE;
1635 case LINUX_RLIMIT_RSS:
1636 return RLIMIT_RSS;
1637 case LINUX_RLIMIT_NPROC:
1638 return RLIMIT_NPROC;
1639 case LINUX_RLIMIT_NOFILE:
1640 return RLIMIT_NOFILE;
1641 case LINUX_RLIMIT_MEMLOCK:
1642 return RLIMIT_MEMLOCK;
1643 case LINUX_RLIMIT_AS:
1644 case LINUX_RLIMIT_LOCKS:
1645 return -EOPNOTSUPP;
1646 default:
1647 return -EINVAL;
1648 }
1649 }
1650
1651
1652 int
1653 linux_sys_getrlimit(l, v, retval)
1654 struct lwp *l;
1655 void *v;
1656 register_t *retval;
1657 {
1658 struct linux_sys_getrlimit_args /* {
1659 syscallarg(int) which;
1660 syscallarg(struct orlimit *) rlp;
1661 } */ *uap = v;
1662 struct proc *p = l->l_proc;
1663 caddr_t sg = stackgap_init(p, 0);
1664 struct sys_getrlimit_args ap;
1665 struct rlimit rl;
1666 struct orlimit orl;
1667 int error;
1668
1669 SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which));
1670 if ((error = SCARG(&ap, which)) < 0)
1671 return -error;
1672 SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl);
1673 if ((error = sys_getrlimit(l, &ap, retval)) != 0)
1674 return error;
1675 if ((error = copyin(SCARG(&ap, rlp), &rl, sizeof(rl))) != 0)
1676 return error;
1677 bsd_to_linux_rlimit(&orl, &rl);
1678 return copyout(&orl, SCARG(uap, rlp), sizeof(orl));
1679 }
1680
1681 int
1682 linux_sys_setrlimit(l, v, retval)
1683 struct lwp *l;
1684 void *v;
1685 register_t *retval;
1686 {
1687 struct linux_sys_setrlimit_args /* {
1688 syscallarg(int) which;
1689 syscallarg(struct orlimit *) rlp;
1690 } */ *uap = v;
1691 struct proc *p = l->l_proc;
1692 caddr_t sg = stackgap_init(p, 0);
1693 struct sys_setrlimit_args ap;
1694 struct rlimit rl;
1695 struct orlimit orl;
1696 int error;
1697
1698 SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which));
1699 SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl);
1700 if ((error = SCARG(&ap, which)) < 0)
1701 return -error;
1702 if ((error = copyin(SCARG(uap, rlp), &orl, sizeof(orl))) != 0)
1703 return error;
1704 linux_to_bsd_rlimit(&rl, &orl);
1705 /* XXX: alpha complains about this */
1706 if ((error = copyout(&rl, (void *)SCARG(&ap, rlp), sizeof(rl))) != 0)
1707 return error;
1708 return sys_setrlimit(l, &ap, retval);
1709 }
1710
1711 #ifndef __mips__
1712 /* XXX: this doesn't look 100% common, at least mips doesn't have it */
1713 int
1714 linux_sys_ugetrlimit(l, v, retval)
1715 struct lwp *l;
1716 void *v;
1717 register_t *retval;
1718 {
1719 return linux_sys_getrlimit(l, v, retval);
1720 }
1721 #endif
1722
1723 /*
1724 * This gets called for unsupported syscalls. The difference to sys_nosys()
1725 * is that process does not get SIGSYS, the call just returns with ENOSYS.
1726 * This is the way Linux does it and glibc depends on this behaviour.
1727 */
1728 int
1729 linux_sys_nosys(l, v, retval)
1730 struct lwp *l;
1731 void *v;
1732 register_t *retval;
1733 {
1734 return (ENOSYS);
1735 }
Cache object: 274a425e06b7580e87d98ffc3f3f1bf1
|