linux_file64.c revision 1.32.2.1 1 /* $NetBSD: linux_file64.c,v 1.32.2.1 2006/05/24 10:57:28 yamt Exp $ */
2
3 /*-
4 * Copyright (c) 1995, 1998, 2000 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.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. All advertising materials mentioning features or use of this software
19 * must display the following acknowledgement:
20 * This product includes software developed by the NetBSD
21 * Foundation, Inc. and its contributors.
22 * 4. Neither the name of The NetBSD Foundation nor the names of its
23 * contributors may be used to endorse or promote products derived
24 * from this software without specific prior written permission.
25 *
26 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
27 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
28 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
29 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
30 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
31 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
32 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
33 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
34 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
35 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
36 * POSSIBILITY OF SUCH DAMAGE.
37 */
38
39 /*
40 * Linux 64bit filesystem calls. Used on 32bit archs, not used on 64bit ones.
41 */
42
43 #include <sys/cdefs.h>
44 __KERNEL_RCSID(0, "$NetBSD: linux_file64.c,v 1.32.2.1 2006/05/24 10:57:28 yamt Exp $");
45
46 #include "opt_ktrace.h"
47
48 #include <sys/param.h>
49 #include <sys/systm.h>
50 #include <sys/namei.h>
51 #include <sys/proc.h>
52 #include <sys/dirent.h>
53 #include <sys/file.h>
54 #include <sys/stat.h>
55 #include <sys/filedesc.h>
56 #include <sys/ioctl.h>
57 #include <sys/kernel.h>
58 #include <sys/mount.h>
59 #include <sys/malloc.h>
60 #include <sys/vnode.h>
61 #include <sys/tty.h>
62 #include <sys/conf.h>
63 #if defined(KTRACE)
64 #include <sys/ktrace.h>
65 #endif /* defined(KTRACE) */
66
67 #include <sys/sa.h>
68 #include <sys/syscallargs.h>
69
70 #include <compat/linux/common/linux_types.h>
71 #include <compat/linux/common/linux_signal.h>
72 #include <compat/linux/common/linux_fcntl.h>
73 #include <compat/linux/common/linux_util.h>
74 #include <compat/linux/common/linux_machdep.h>
75 #include <compat/linux/common/linux_dirent.h>
76
77 #include <compat/linux/linux_syscallargs.h>
78
79 #ifndef alpha
80
81 # ifndef COMPAT_LINUX32
82
83 static void bsd_to_linux_stat __P((struct stat *, struct linux_stat64 *));
84 static int linux_do_stat64 __P((struct lwp *, void *, register_t *, int));
85
86 /*
87 * Convert a NetBSD stat structure to a Linux stat structure.
88 * Only the order of the fields and the padding in the structure
89 * is different. linux_fakedev is a machine-dependent function
90 * which optionally converts device driver major/minor numbers
91 * (XXX horrible, but what can you do against code that compares
92 * things against constant major device numbers? sigh)
93 */
94 static void
95 bsd_to_linux_stat(bsp, lsp)
96 struct stat *bsp;
97 struct linux_stat64 *lsp;
98 {
99 lsp->lst_dev = linux_fakedev(bsp->st_dev, 0);
100 lsp->lst_ino = bsp->st_ino;
101 lsp->lst_mode = (linux_mode_t)bsp->st_mode;
102 if (bsp->st_nlink >= (1 << 15))
103 lsp->lst_nlink = (1 << 15) - 1;
104 else
105 lsp->lst_nlink = (linux_nlink_t)bsp->st_nlink;
106 lsp->lst_uid = bsp->st_uid;
107 lsp->lst_gid = bsp->st_gid;
108 lsp->lst_rdev = linux_fakedev(bsp->st_rdev, 1);
109 lsp->lst_size = bsp->st_size;
110 lsp->lst_blksize = bsp->st_blksize;
111 lsp->lst_blocks = bsp->st_blocks;
112 lsp->lst_atime = bsp->st_atime;
113 lsp->lst_mtime = bsp->st_mtime;
114 lsp->lst_ctime = bsp->st_ctime;
115 # ifdef LINUX_STAT64_HAS_NSEC
116 lsp->lst_atime_nsec = bsp->st_atimensec;
117 lsp->lst_mtime_nsec = bsp->st_mtimensec;
118 lsp->lst_ctime_nsec = bsp->st_ctimensec;
119 # endif
120 # if LINUX_STAT64_HAS_BROKEN_ST_INO
121 lsp->__lst_ino = (linux_ino_t) bsp->st_ino;
122 # endif
123 }
124
125 /*
126 * The stat functions below are plain sailing. stat and lstat are handled
127 * by one function to avoid code duplication.
128 */
129 int
130 linux_sys_fstat64(l, v, retval)
131 struct lwp *l;
132 void *v;
133 register_t *retval;
134 {
135 struct linux_sys_fstat64_args /* {
136 syscallarg(int) fd;
137 syscallarg(struct linux_stat64 *) sp;
138 } */ *uap = v;
139 struct proc *p = l->l_proc;
140 struct sys___fstat30_args fsa;
141 struct linux_stat64 tmplst;
142 struct stat *st,tmpst;
143 caddr_t sg;
144 int error;
145
146 sg = stackgap_init(p, 0);
147
148 st = stackgap_alloc(p, &sg, sizeof (struct stat));
149
150 SCARG(&fsa, fd) = SCARG(uap, fd);
151 SCARG(&fsa, sb) = st;
152
153 if ((error = sys___fstat30(l, &fsa, retval)))
154 return error;
155
156 if ((error = copyin(st, &tmpst, sizeof tmpst)))
157 return error;
158
159 bsd_to_linux_stat(&tmpst, &tmplst);
160
161 if ((error = copyout(&tmplst, SCARG(uap, sp), sizeof tmplst)))
162 return error;
163
164 return 0;
165 }
166
167 static int
168 linux_do_stat64(l, v, retval, dolstat)
169 struct lwp *l;
170 void *v;
171 register_t *retval;
172 int dolstat;
173 {
174 struct proc *p = l->l_proc;
175 struct sys___stat30_args sa;
176 struct linux_stat64 tmplst;
177 struct stat *st, tmpst;
178 caddr_t sg;
179 int error;
180 struct linux_sys_stat64_args *uap = v;
181
182 sg = stackgap_init(p, 0);
183 st = stackgap_alloc(p, &sg, sizeof (struct stat));
184 CHECK_ALT_EXIST(l, &sg, SCARG(uap, path));
185
186 SCARG(&sa, ub) = st;
187 SCARG(&sa, path) = SCARG(uap, path);
188
189 if ((error = (dolstat ? sys___lstat30(l, &sa, retval) :
190 sys___stat30(l, &sa, retval))))
191 return error;
192
193 if ((error = copyin(st, &tmpst, sizeof tmpst)))
194 return error;
195
196 bsd_to_linux_stat(&tmpst, &tmplst);
197
198 if ((error = copyout(&tmplst, SCARG(uap, sp), sizeof tmplst)))
199 return error;
200
201 return 0;
202 }
203
204 int
205 linux_sys_stat64(l, v, retval)
206 struct lwp *l;
207 void *v;
208 register_t *retval;
209 {
210 struct linux_sys_stat64_args /* {
211 syscallarg(const char *) path;
212 syscallarg(struct linux_stat64 *) sp;
213 } */ *uap = v;
214
215 return linux_do_stat64(l, uap, retval, 0);
216 }
217
218 int
219 linux_sys_lstat64(l, v, retval)
220 struct lwp *l;
221 void *v;
222 register_t *retval;
223 {
224 struct linux_sys_lstat64_args /* {
225 syscallarg(const char *) path;
226 syscallarg(struct linux_stat64 *) sp;
227 } */ *uap = v;
228
229 return linux_do_stat64(l, uap, retval, 1);
230 }
231
232 int
233 linux_sys_truncate64(l, v, retval)
234 struct lwp *l;
235 void *v;
236 register_t *retval;
237 {
238 struct linux_sys_truncate64_args /* {
239 syscallarg(const char *) path;
240 syscallarg(off_t) length;
241 } */ *uap = v;
242 struct sys_truncate_args ta;
243 struct proc *p = l->l_proc;
244 caddr_t sg = stackgap_init(p, 0);
245
246 CHECK_ALT_EXIST(l, &sg, SCARG(uap, path));
247
248 /* Linux doesn't have the 'pad' pseudo-parameter */
249 SCARG(&ta, path) = SCARG(uap, path);
250 SCARG(&ta, pad) = 0;
251 SCARG(&ta, length) = SCARG(uap, length);
252
253 return sys_truncate(l, &ta, retval);
254 }
255
256 int
257 linux_sys_ftruncate64(l, v, retval)
258 struct lwp *l;
259 void *v;
260 register_t *retval;
261 {
262 struct linux_sys_ftruncate64_args /* {
263 syscallarg(unsigned int) fd;
264 syscallarg(off_t) length;
265 } */ *uap = v;
266 struct sys_ftruncate_args ta;
267
268 /* Linux doesn't have the 'pad' pseudo-parameter */
269 SCARG(&ta, fd) = SCARG(uap, fd);
270 SCARG(&ta, pad) = 0;
271 SCARG(&ta, length) = SCARG(uap, length);
272
273 return sys_ftruncate(l, &ta, retval);
274 }
275 # endif /* !COMPAT_LINUX32 */
276
277 # if !defined(__m68k__) && (!defined(__amd64__) || defined(COMPAT_LINUX32))
278 static void bsd_to_linux_flock64 __P((struct linux_flock64 *,
279 const struct flock *));
280 static void linux_to_bsd_flock64 __P((struct flock *,
281 const struct linux_flock64 *));
282
283 static void
284 bsd_to_linux_flock64(lfp, bfp)
285 struct linux_flock64 *lfp;
286 const struct flock *bfp;
287 {
288
289 lfp->l_start = bfp->l_start;
290 lfp->l_len = bfp->l_len;
291 lfp->l_pid = bfp->l_pid;
292 lfp->l_whence = bfp->l_whence;
293 switch (bfp->l_type) {
294 case F_RDLCK:
295 lfp->l_type = LINUX_F_RDLCK;
296 break;
297 case F_UNLCK:
298 lfp->l_type = LINUX_F_UNLCK;
299 break;
300 case F_WRLCK:
301 lfp->l_type = LINUX_F_WRLCK;
302 break;
303 }
304 }
305
306 static void
307 linux_to_bsd_flock64(bfp, lfp)
308 struct flock *bfp;
309 const struct linux_flock64 *lfp;
310 {
311
312 bfp->l_start = lfp->l_start;
313 bfp->l_len = lfp->l_len;
314 bfp->l_pid = lfp->l_pid;
315 bfp->l_whence = lfp->l_whence;
316 switch (lfp->l_type) {
317 case LINUX_F_RDLCK:
318 bfp->l_type = F_RDLCK;
319 break;
320 case LINUX_F_UNLCK:
321 bfp->l_type = F_UNLCK;
322 break;
323 case LINUX_F_WRLCK:
324 bfp->l_type = F_WRLCK;
325 break;
326 }
327 }
328
329 int
330 linux_sys_fcntl64(l, v, retval)
331 struct lwp *l;
332 void *v;
333 register_t *retval;
334 {
335 struct linux_sys_fcntl64_args /* {
336 syscallarg(int) fd;
337 syscallarg(int) cmd;
338 syscallarg(void *) arg;
339 } */ *uap = v;
340 struct proc *p = l->l_proc;
341 struct sys_fcntl_args fca;
342 struct linux_flock64 lfl;
343 struct flock bfl, *bfp;
344 int error;
345 caddr_t sg;
346 void *arg = SCARG(uap, arg);
347 int cmd = SCARG(uap, cmd);
348 int fd = SCARG(uap, fd);
349
350 switch (cmd) {
351 case LINUX_F_GETLK64:
352 sg = stackgap_init(p, 0);
353 bfp = (struct flock *) stackgap_alloc(p, &sg, sizeof *bfp);
354 if ((error = copyin(arg, &lfl, sizeof lfl)) != 0)
355 return error;
356 linux_to_bsd_flock64(&bfl, &lfl);
357 if ((error = copyout(&bfl, bfp, sizeof bfl)) != 0)
358 return error;
359 SCARG(&fca, fd) = fd;
360 SCARG(&fca, cmd) = F_GETLK;
361 SCARG(&fca, arg) = bfp;
362 if ((error = sys_fcntl(l, &fca, retval)) != 0)
363 return error;
364 if ((error = copyin(bfp, &bfl, sizeof bfl)) != 0)
365 return error;
366 bsd_to_linux_flock64(&lfl, &bfl);
367 return copyout(&lfl, arg, sizeof lfl);
368 case LINUX_F_SETLK64:
369 case LINUX_F_SETLKW64:
370 cmd = (cmd == LINUX_F_SETLK64 ? F_SETLK : F_SETLKW);
371 if ((error = copyin(arg, &lfl, sizeof lfl)) != 0)
372 return error;
373 linux_to_bsd_flock64(&bfl, &lfl);
374 sg = stackgap_init(p, 0);
375 bfp = (struct flock *) stackgap_alloc(p, &sg, sizeof *bfp);
376 if ((error = copyout(&bfl, bfp, sizeof bfl)) != 0)
377 return error;
378 SCARG(&fca, fd) = fd;
379 SCARG(&fca, cmd) = cmd;
380 SCARG(&fca, arg) = bfp;
381 return sys_fcntl(l, &fca, retval);
382 default:
383 return linux_sys_fcntl(l, v, retval);
384 }
385 }
386 # endif /* !m69k && !amd64 && !COMPAT_LINUX32 */
387
388 #endif /* !alpha */
389
390 /*
391 * Linux 'readdir' call. This code is mostly taken from the
392 * SunOS getdents call (see compat/sunos/sunos_misc.c), though
393 * an attempt has been made to keep it a little cleaner.
394 *
395 * The d_off field contains the offset of the next valid entry,
396 * unless the older Linux getdents(2), which used to have it set
397 * to the offset of the entry itself. This function also doesn't
398 * need to deal with the old count == 1 glibc problem.
399 *
400 * Read in BSD-style entries, convert them, and copy them out.
401 *
402 * Note that this doesn't handle union-mounted filesystems.
403 */
404 #ifndef COMPAT_LINUX32
405 int
406 linux_sys_getdents64(l, v, retval)
407 struct lwp *l;
408 void *v;
409 register_t *retval;
410 {
411 struct linux_sys_getdents_args /* {
412 syscallarg(int) fd;
413 syscallarg(struct linux_dirent64 *) dent;
414 syscallarg(unsigned int) count;
415 } */ *uap = v;
416 struct proc *p = l->l_proc;
417 struct dirent *bdp;
418 struct vnode *vp;
419 caddr_t inp, tbuf; /* BSD-format */
420 int len, reclen; /* BSD-format */
421 caddr_t outp; /* Linux-format */
422 int resid, linux_reclen = 0; /* Linux-format */
423 struct file *fp;
424 struct uio auio;
425 struct iovec aiov;
426 struct linux_dirent64 idb;
427 off_t off; /* true file offset */
428 int buflen, error, eofflag, nbytes;
429 struct vattr va;
430 off_t *cookiebuf = NULL, *cookie;
431 int ncookies;
432
433 /* getvnode() will use the descriptor for us */
434 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
435 return (error);
436
437 if ((fp->f_flag & FREAD) == 0) {
438 error = EBADF;
439 goto out1;
440 }
441
442 vp = (struct vnode *)fp->f_data;
443 if (vp->v_type != VDIR) {
444 error = EINVAL;
445 goto out1;
446 }
447
448 if ((error = VOP_GETATTR(vp, &va, p->p_cred, l)))
449 goto out1;
450
451 nbytes = SCARG(uap, count);
452 buflen = min(MAXBSIZE, nbytes);
453 if (buflen < va.va_blocksize)
454 buflen = va.va_blocksize;
455 tbuf = malloc(buflen, M_TEMP, M_WAITOK);
456
457 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
458 off = fp->f_offset;
459 outp = (caddr_t)SCARG(uap, dent);
460 again:
461 aiov.iov_base = tbuf;
462 aiov.iov_len = buflen;
463 auio.uio_iov = &aiov;
464 auio.uio_iovcnt = 1;
465 auio.uio_rw = UIO_READ;
466 auio.uio_resid = buflen;
467 auio.uio_offset = off;
468 UIO_SETUP_SYSSPACE(&auio);
469 /*
470 * First we read into the malloc'ed buffer, then
471 * we massage it into user space, one record at a time.
472 */
473 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf,
474 &ncookies);
475 if (error)
476 goto out;
477
478 inp = tbuf;
479 resid = nbytes;
480 if ((len = buflen - auio.uio_resid) == 0)
481 goto eof;
482
483 for (cookie = cookiebuf; len > 0; len -= reclen) {
484 bdp = (struct dirent *)inp;
485 reclen = bdp->d_reclen;
486 if (reclen & 3)
487 panic("linux_readdir");
488 if (bdp->d_fileno == 0) {
489 inp += reclen; /* it is a hole; squish it out */
490 if (cookie)
491 off = *cookie++;
492 else
493 off += reclen;
494 continue;
495 }
496 linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen);
497 if (reclen > len || resid < linux_reclen) {
498 /* entry too big for buffer, so just stop */
499 outp++;
500 break;
501 }
502 if (cookie)
503 off = *cookie++; /* each entry points to next */
504 else
505 off += reclen;
506 /*
507 * Massage in place to make a Linux-shaped dirent (otherwise
508 * we have to worry about touching user memory outside of
509 * the copyout() call).
510 */
511 idb.d_ino = bdp->d_fileno;
512 idb.d_type = bdp->d_type;
513 idb.d_off = off;
514 idb.d_reclen = (u_short)linux_reclen;
515 strcpy(idb.d_name, bdp->d_name);
516 if ((error = copyout((caddr_t)&idb, outp, linux_reclen)))
517 goto out;
518 /* advance past this real entry */
519 inp += reclen;
520 /* advance output past Linux-shaped entry */
521 outp += linux_reclen;
522 resid -= linux_reclen;
523 }
524
525 /* if we squished out the whole block, try again */
526 if (outp == (caddr_t)SCARG(uap, dent))
527 goto again;
528 fp->f_offset = off; /* update the vnode offset */
529
530 eof:
531 *retval = nbytes - resid;
532 out:
533 VOP_UNLOCK(vp, 0);
534 if (cookiebuf)
535 free(cookiebuf, M_TEMP);
536 free(tbuf, M_TEMP);
537 #if defined(KTRACE)
538 if (!error && outp > (const char *)SCARG(uap, dent) &&
539 KTRPOINT(p, KTR_GENIO)) {
540 struct iovec iov;
541 size_t done = outp - (const char *)SCARG(uap, dent);
542
543 iov.iov_base = SCARG(uap, dent);
544 iov.iov_len = done;
545 ktrgenio(l, SCARG(uap, fd), UIO_READ, &iov, done, 0);
546 }
547 #endif /* defined(KTRACE) */
548 out1:
549 FILE_UNUSE(fp, l);
550 return error;
551 }
552 #endif /* !COMPAT_LINUX32 */
553