linux_file64.c revision 1.36 1 /* $NetBSD: linux_file64.c,v 1.36 2007/03/04 06:01:23 christos 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 1.36 2007/03/04 06:01:23 christos Exp $");
45
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/namei.h>
49 #include <sys/proc.h>
50 #include <sys/dirent.h>
51 #include <sys/file.h>
52 #include <sys/stat.h>
53 #include <sys/filedesc.h>
54 #include <sys/ioctl.h>
55 #include <sys/kernel.h>
56 #include <sys/mount.h>
57 #include <sys/malloc.h>
58 #include <sys/vnode.h>
59 #include <sys/tty.h>
60 #include <sys/conf.h>
61
62 #include <sys/syscallargs.h>
63
64 #include <compat/linux/common/linux_types.h>
65 #include <compat/linux/common/linux_signal.h>
66 #include <compat/linux/common/linux_fcntl.h>
67 #include <compat/linux/common/linux_util.h>
68 #include <compat/linux/common/linux_machdep.h>
69 #include <compat/linux/common/linux_dirent.h>
70
71 #include <compat/linux/linux_syscallargs.h>
72
73 #ifndef alpha
74
75 # ifndef COMPAT_LINUX32
76
77 static void bsd_to_linux_stat __P((struct stat *, struct linux_stat64 *));
78 static int linux_do_stat64 __P((struct lwp *, void *, register_t *, int));
79
80 /*
81 * Convert a NetBSD stat structure to a Linux stat structure.
82 * Only the order of the fields and the padding in the structure
83 * is different. linux_fakedev is a machine-dependent function
84 * which optionally converts device driver major/minor numbers
85 * (XXX horrible, but what can you do against code that compares
86 * things against constant major device numbers? sigh)
87 */
88 static void
89 bsd_to_linux_stat(bsp, lsp)
90 struct stat *bsp;
91 struct linux_stat64 *lsp;
92 {
93 lsp->lst_dev = linux_fakedev(bsp->st_dev, 0);
94 lsp->lst_ino = bsp->st_ino;
95 lsp->lst_mode = (linux_mode_t)bsp->st_mode;
96 if (bsp->st_nlink >= (1 << 15))
97 lsp->lst_nlink = (1 << 15) - 1;
98 else
99 lsp->lst_nlink = (linux_nlink_t)bsp->st_nlink;
100 lsp->lst_uid = bsp->st_uid;
101 lsp->lst_gid = bsp->st_gid;
102 lsp->lst_rdev = linux_fakedev(bsp->st_rdev, 1);
103 lsp->lst_size = bsp->st_size;
104 lsp->lst_blksize = bsp->st_blksize;
105 lsp->lst_blocks = bsp->st_blocks;
106 lsp->lst_atime = bsp->st_atime;
107 lsp->lst_mtime = bsp->st_mtime;
108 lsp->lst_ctime = bsp->st_ctime;
109 # ifdef LINUX_STAT64_HAS_NSEC
110 lsp->lst_atime_nsec = bsp->st_atimensec;
111 lsp->lst_mtime_nsec = bsp->st_mtimensec;
112 lsp->lst_ctime_nsec = bsp->st_ctimensec;
113 # endif
114 # if LINUX_STAT64_HAS_BROKEN_ST_INO
115 lsp->__lst_ino = (linux_ino_t) bsp->st_ino;
116 # endif
117 }
118
119 /*
120 * The stat functions below are plain sailing. stat and lstat are handled
121 * by one function to avoid code duplication.
122 */
123 int
124 linux_sys_fstat64(l, v, retval)
125 struct lwp *l;
126 void *v;
127 register_t *retval;
128 {
129 struct linux_sys_fstat64_args /* {
130 syscallarg(int) fd;
131 syscallarg(struct linux_stat64 *) sp;
132 } */ *uap = v;
133 struct proc *p = l->l_proc;
134 struct sys___fstat30_args fsa;
135 struct linux_stat64 tmplst;
136 struct stat *st,tmpst;
137 void *sg;
138 int error;
139
140 sg = stackgap_init(p, 0);
141
142 st = stackgap_alloc(p, &sg, sizeof (struct stat));
143
144 SCARG(&fsa, fd) = SCARG(uap, fd);
145 SCARG(&fsa, sb) = st;
146
147 if ((error = sys___fstat30(l, &fsa, retval)))
148 return error;
149
150 if ((error = copyin(st, &tmpst, sizeof tmpst)))
151 return error;
152
153 bsd_to_linux_stat(&tmpst, &tmplst);
154
155 if ((error = copyout(&tmplst, SCARG(uap, sp), sizeof tmplst)))
156 return error;
157
158 return 0;
159 }
160
161 static int
162 linux_do_stat64(l, v, retval, dolstat)
163 struct lwp *l;
164 void *v;
165 register_t *retval;
166 int dolstat;
167 {
168 struct proc *p = l->l_proc;
169 struct sys___stat30_args sa;
170 struct linux_stat64 tmplst;
171 struct stat *st, tmpst;
172 void *sg;
173 int error;
174 struct linux_sys_stat64_args *uap = v;
175
176 sg = stackgap_init(p, 0);
177 st = stackgap_alloc(p, &sg, sizeof (struct stat));
178 CHECK_ALT_EXIST(l, &sg, SCARG(uap, path));
179
180 SCARG(&sa, ub) = st;
181 SCARG(&sa, path) = SCARG(uap, path);
182
183 if ((error = (dolstat ? sys___lstat30(l, &sa, retval) :
184 sys___stat30(l, &sa, retval))))
185 return error;
186
187 if ((error = copyin(st, &tmpst, sizeof tmpst)))
188 return error;
189
190 bsd_to_linux_stat(&tmpst, &tmplst);
191
192 if ((error = copyout(&tmplst, SCARG(uap, sp), sizeof tmplst)))
193 return error;
194
195 return 0;
196 }
197
198 int
199 linux_sys_stat64(l, v, retval)
200 struct lwp *l;
201 void *v;
202 register_t *retval;
203 {
204 struct linux_sys_stat64_args /* {
205 syscallarg(const char *) path;
206 syscallarg(struct linux_stat64 *) sp;
207 } */ *uap = v;
208
209 return linux_do_stat64(l, uap, retval, 0);
210 }
211
212 int
213 linux_sys_lstat64(l, v, retval)
214 struct lwp *l;
215 void *v;
216 register_t *retval;
217 {
218 struct linux_sys_lstat64_args /* {
219 syscallarg(const char *) path;
220 syscallarg(struct linux_stat64 *) sp;
221 } */ *uap = v;
222
223 return linux_do_stat64(l, uap, retval, 1);
224 }
225
226 int
227 linux_sys_truncate64(l, v, retval)
228 struct lwp *l;
229 void *v;
230 register_t *retval;
231 {
232 struct linux_sys_truncate64_args /* {
233 syscallarg(const char *) path;
234 syscallarg(off_t) length;
235 } */ *uap = v;
236 struct sys_truncate_args ta;
237 struct proc *p = l->l_proc;
238 void *sg = stackgap_init(p, 0);
239
240 CHECK_ALT_EXIST(l, &sg, SCARG(uap, path));
241
242 /* Linux doesn't have the 'pad' pseudo-parameter */
243 SCARG(&ta, path) = SCARG(uap, path);
244 SCARG(&ta, pad) = 0;
245 SCARG(&ta, length) = SCARG(uap, length);
246
247 return sys_truncate(l, &ta, retval);
248 }
249
250 int
251 linux_sys_ftruncate64(l, v, retval)
252 struct lwp *l;
253 void *v;
254 register_t *retval;
255 {
256 struct linux_sys_ftruncate64_args /* {
257 syscallarg(unsigned int) fd;
258 syscallarg(off_t) length;
259 } */ *uap = v;
260 struct sys_ftruncate_args ta;
261
262 /* Linux doesn't have the 'pad' pseudo-parameter */
263 SCARG(&ta, fd) = SCARG(uap, fd);
264 SCARG(&ta, pad) = 0;
265 SCARG(&ta, length) = SCARG(uap, length);
266
267 return sys_ftruncate(l, &ta, retval);
268 }
269 # endif /* !COMPAT_LINUX32 */
270
271 # if !defined(__m68k__) && (!defined(__amd64__) || defined(COMPAT_LINUX32))
272 static void bsd_to_linux_flock64 __P((struct linux_flock64 *,
273 const struct flock *));
274 static void linux_to_bsd_flock64 __P((struct flock *,
275 const struct linux_flock64 *));
276
277 static void
278 bsd_to_linux_flock64(lfp, bfp)
279 struct linux_flock64 *lfp;
280 const struct flock *bfp;
281 {
282
283 lfp->l_start = bfp->l_start;
284 lfp->l_len = bfp->l_len;
285 lfp->l_pid = bfp->l_pid;
286 lfp->l_whence = bfp->l_whence;
287 switch (bfp->l_type) {
288 case F_RDLCK:
289 lfp->l_type = LINUX_F_RDLCK;
290 break;
291 case F_UNLCK:
292 lfp->l_type = LINUX_F_UNLCK;
293 break;
294 case F_WRLCK:
295 lfp->l_type = LINUX_F_WRLCK;
296 break;
297 }
298 }
299
300 static void
301 linux_to_bsd_flock64(bfp, lfp)
302 struct flock *bfp;
303 const struct linux_flock64 *lfp;
304 {
305
306 bfp->l_start = lfp->l_start;
307 bfp->l_len = lfp->l_len;
308 bfp->l_pid = lfp->l_pid;
309 bfp->l_whence = lfp->l_whence;
310 switch (lfp->l_type) {
311 case LINUX_F_RDLCK:
312 bfp->l_type = F_RDLCK;
313 break;
314 case LINUX_F_UNLCK:
315 bfp->l_type = F_UNLCK;
316 break;
317 case LINUX_F_WRLCK:
318 bfp->l_type = F_WRLCK;
319 break;
320 }
321 }
322
323 int
324 linux_sys_fcntl64(l, v, retval)
325 struct lwp *l;
326 void *v;
327 register_t *retval;
328 {
329 struct linux_sys_fcntl64_args /* {
330 syscallarg(int) fd;
331 syscallarg(int) cmd;
332 syscallarg(void *) arg;
333 } */ *uap = v;
334 struct proc *p = l->l_proc;
335 struct sys_fcntl_args fca;
336 struct linux_flock64 lfl;
337 struct flock bfl, *bfp;
338 int error;
339 void *sg;
340 void *arg = SCARG(uap, arg);
341 int cmd = SCARG(uap, cmd);
342 int fd = SCARG(uap, fd);
343
344 switch (cmd) {
345 case LINUX_F_GETLK64:
346 sg = stackgap_init(p, 0);
347 bfp = (struct flock *) stackgap_alloc(p, &sg, sizeof *bfp);
348 if ((error = copyin(arg, &lfl, sizeof lfl)) != 0)
349 return error;
350 linux_to_bsd_flock64(&bfl, &lfl);
351 if ((error = copyout(&bfl, bfp, sizeof bfl)) != 0)
352 return error;
353 SCARG(&fca, fd) = fd;
354 SCARG(&fca, cmd) = F_GETLK;
355 SCARG(&fca, arg) = bfp;
356 if ((error = sys_fcntl(l, &fca, retval)) != 0)
357 return error;
358 if ((error = copyin(bfp, &bfl, sizeof bfl)) != 0)
359 return error;
360 bsd_to_linux_flock64(&lfl, &bfl);
361 return copyout(&lfl, arg, sizeof lfl);
362 case LINUX_F_SETLK64:
363 case LINUX_F_SETLKW64:
364 cmd = (cmd == LINUX_F_SETLK64 ? F_SETLK : F_SETLKW);
365 if ((error = copyin(arg, &lfl, sizeof lfl)) != 0)
366 return error;
367 linux_to_bsd_flock64(&bfl, &lfl);
368 sg = stackgap_init(p, 0);
369 bfp = (struct flock *) stackgap_alloc(p, &sg, sizeof *bfp);
370 if ((error = copyout(&bfl, bfp, sizeof bfl)) != 0)
371 return error;
372 SCARG(&fca, fd) = fd;
373 SCARG(&fca, cmd) = cmd;
374 SCARG(&fca, arg) = bfp;
375 return sys_fcntl(l, &fca, retval);
376 default:
377 return linux_sys_fcntl(l, v, retval);
378 }
379 }
380 # endif /* !m69k && !amd64 && !COMPAT_LINUX32 */
381
382 #endif /* !alpha */
383
384 /*
385 * Linux 'readdir' call. This code is mostly taken from the
386 * SunOS getdents call (see compat/sunos/sunos_misc.c), though
387 * an attempt has been made to keep it a little cleaner.
388 *
389 * The d_off field contains the offset of the next valid entry,
390 * unless the older Linux getdents(2), which used to have it set
391 * to the offset of the entry itself. This function also doesn't
392 * need to deal with the old count == 1 glibc problem.
393 *
394 * Read in BSD-style entries, convert them, and copy them out.
395 *
396 * Note that this doesn't handle union-mounted filesystems.
397 */
398 #ifndef COMPAT_LINUX32
399 int
400 linux_sys_getdents64(l, v, retval)
401 struct lwp *l;
402 void *v;
403 register_t *retval;
404 {
405 struct linux_sys_getdents_args /* {
406 syscallarg(int) fd;
407 syscallarg(struct linux_dirent64 *) dent;
408 syscallarg(unsigned int) count;
409 } */ *uap = v;
410 struct dirent *bdp;
411 struct vnode *vp;
412 char *inp, *tbuf; /* BSD-format */
413 int len, reclen; /* BSD-format */
414 char *outp; /* Linux-format */
415 int resid, linux_reclen = 0; /* Linux-format */
416 struct file *fp;
417 struct uio auio;
418 struct iovec aiov;
419 struct linux_dirent64 idb;
420 off_t off; /* true file offset */
421 int buflen, error, eofflag, nbytes;
422 struct vattr va;
423 off_t *cookiebuf = NULL, *cookie;
424 int ncookies;
425
426 /* getvnode() will use the descriptor for us */
427 if ((error = getvnode(l->l_proc->p_fd, SCARG(uap, fd), &fp)) != 0)
428 return (error);
429
430 if ((fp->f_flag & FREAD) == 0) {
431 error = EBADF;
432 goto out1;
433 }
434
435 vp = (struct vnode *)fp->f_data;
436 if (vp->v_type != VDIR) {
437 error = EINVAL;
438 goto out1;
439 }
440
441 if ((error = VOP_GETATTR(vp, &va, l->l_cred, l)))
442 goto out1;
443
444 nbytes = SCARG(uap, count);
445 buflen = min(MAXBSIZE, nbytes);
446 if (buflen < va.va_blocksize)
447 buflen = va.va_blocksize;
448 tbuf = malloc(buflen, M_TEMP, M_WAITOK);
449
450 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
451 off = fp->f_offset;
452 again:
453 aiov.iov_base = tbuf;
454 aiov.iov_len = buflen;
455 auio.uio_iov = &aiov;
456 auio.uio_iovcnt = 1;
457 auio.uio_rw = UIO_READ;
458 auio.uio_resid = buflen;
459 auio.uio_offset = off;
460 UIO_SETUP_SYSSPACE(&auio);
461 /*
462 * First we read into the malloc'ed buffer, then
463 * we massage it into user space, one record at a time.
464 */
465 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf,
466 &ncookies);
467 if (error)
468 goto out;
469
470 inp = tbuf;
471 outp = (void *)SCARG(uap, dent);
472 resid = nbytes;
473 if ((len = buflen - auio.uio_resid) == 0)
474 goto eof;
475
476 for (cookie = cookiebuf; len > 0; len -= reclen) {
477 bdp = (struct dirent *)inp;
478 reclen = bdp->d_reclen;
479 if (reclen & 3)
480 panic("linux_readdir");
481 if (bdp->d_fileno == 0) {
482 inp += reclen; /* it is a hole; squish it out */
483 if (cookie)
484 off = *cookie++;
485 else
486 off += reclen;
487 continue;
488 }
489 linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen);
490 if (reclen > len || resid < linux_reclen) {
491 /* entry too big for buffer, so just stop */
492 outp++;
493 break;
494 }
495 if (cookie)
496 off = *cookie++; /* each entry points to next */
497 else
498 off += reclen;
499 /*
500 * Massage in place to make a Linux-shaped dirent (otherwise
501 * we have to worry about touching user memory outside of
502 * the copyout() call).
503 */
504 idb.d_ino = bdp->d_fileno;
505 idb.d_type = bdp->d_type;
506 idb.d_off = off;
507 idb.d_reclen = (u_short)linux_reclen;
508 strcpy(idb.d_name, bdp->d_name);
509 if ((error = copyout((void *)&idb, outp, linux_reclen)))
510 goto out;
511 /* advance past this real entry */
512 inp += reclen;
513 /* advance output past Linux-shaped entry */
514 outp += linux_reclen;
515 resid -= linux_reclen;
516 }
517
518 /* if we squished out the whole block, try again */
519 if (outp == (void *)SCARG(uap, dent))
520 goto again;
521 fp->f_offset = off; /* update the vnode offset */
522
523 eof:
524 *retval = nbytes - resid;
525 out:
526 VOP_UNLOCK(vp, 0);
527 if (cookiebuf)
528 free(cookiebuf, M_TEMP);
529 free(tbuf, M_TEMP);
530 out1:
531 FILE_UNUSE(fp, l);
532 return error;
533 }
534 #endif /* !COMPAT_LINUX32 */
535