linux_file64.c revision 1.18 1 /* $NetBSD: linux_file64.c,v 1.18 2002/05/20 06:45:11 jdolecek 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 1.18 2002/05/20 06:45:11 jdolecek 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 static void bsd_to_linux_stat __P((struct stat *, struct linux_stat64 *));
76 static int linux_do_stat64 __P((struct proc *, void *, register_t *, int));
77
78 /*
79 * Convert a NetBSD stat structure to a Linux stat structure.
80 * Only the order of the fields and the padding in the structure
81 * is different. linux_fakedev is a machine-dependent function
82 * which optionally converts device driver major/minor numbers
83 * (XXX horrible, but what can you do against code that compares
84 * things against constant major device numbers? sigh)
85 */
86 static void
87 bsd_to_linux_stat(bsp, lsp)
88 struct stat *bsp;
89 struct linux_stat64 *lsp;
90 {
91 lsp->lst_dev = linux_fakedev(bsp->st_dev, 0);
92 lsp->lst_ino = bsp->st_ino;
93 lsp->lst_mode = (linux_mode_t)bsp->st_mode;
94 if (bsp->st_nlink >= (1 << 15))
95 lsp->lst_nlink = (1 << 15) - 1;
96 else
97 lsp->lst_nlink = (linux_nlink_t)bsp->st_nlink;
98 lsp->lst_uid = bsp->st_uid;
99 lsp->lst_gid = bsp->st_gid;
100 lsp->lst_rdev = linux_fakedev(bsp->st_rdev, 1);
101 lsp->lst_size = bsp->st_size;
102 lsp->lst_blksize = bsp->st_blksize;
103 lsp->lst_blocks = bsp->st_blocks;
104 lsp->lst_atime = bsp->st_atime;
105 lsp->lst_mtime = bsp->st_mtime;
106 lsp->lst_ctime = bsp->st_ctime;
107 #if LINUX_STAT64_HAS_BROKEN_ST_INO
108 lsp->__lst_ino = (linux_ino_t) bsp->st_ino;
109 #endif
110 }
111
112 /*
113 * The stat functions below are plain sailing. stat and lstat are handled
114 * by one function to avoid code duplication.
115 */
116 int
117 linux_sys_fstat64(p, v, retval)
118 struct proc *p;
119 void *v;
120 register_t *retval;
121 {
122 struct linux_sys_fstat64_args /* {
123 syscallarg(int) fd;
124 syscallarg(struct linux_stat64 *) sp;
125 } */ *uap = v;
126 struct sys___fstat13_args fsa;
127 struct linux_stat64 tmplst;
128 struct stat *st,tmpst;
129 caddr_t sg;
130 int error;
131
132 sg = stackgap_init(p, 0);
133
134 st = stackgap_alloc(p, &sg, sizeof (struct stat));
135
136 SCARG(&fsa, fd) = SCARG(uap, fd);
137 SCARG(&fsa, sb) = st;
138
139 if ((error = sys___fstat13(p, &fsa, retval)))
140 return error;
141
142 if ((error = copyin(st, &tmpst, sizeof tmpst)))
143 return error;
144
145 bsd_to_linux_stat(&tmpst, &tmplst);
146
147 if ((error = copyout(&tmplst, SCARG(uap, sp), sizeof tmplst)))
148 return error;
149
150 return 0;
151 }
152
153 static int
154 linux_do_stat64(p, v, retval, dolstat)
155 struct proc *p;
156 void *v;
157 register_t *retval;
158 int dolstat;
159 {
160 struct sys___stat13_args sa;
161 struct linux_stat64 tmplst;
162 struct stat *st, tmpst;
163 caddr_t sg;
164 int error;
165 struct linux_sys_stat64_args *uap = v;
166
167 sg = stackgap_init(p, 0);
168 st = stackgap_alloc(p, &sg, sizeof (struct stat));
169 CHECK_ALT_EXIST(p, &sg, SCARG(uap, path));
170
171 SCARG(&sa, ub) = st;
172 SCARG(&sa, path) = SCARG(uap, path);
173
174 if ((error = (dolstat ? sys___lstat13(p, &sa, retval) :
175 sys___stat13(p, &sa, retval))))
176 return error;
177
178 if ((error = copyin(st, &tmpst, sizeof tmpst)))
179 return error;
180
181 bsd_to_linux_stat(&tmpst, &tmplst);
182
183 if ((error = copyout(&tmplst, SCARG(uap, sp), sizeof tmplst)))
184 return error;
185
186 return 0;
187 }
188
189 int
190 linux_sys_stat64(p, v, retval)
191 struct proc *p;
192 void *v;
193 register_t *retval;
194 {
195 struct linux_sys_stat64_args /* {
196 syscallarg(const char *) path;
197 syscallarg(struct linux_stat64 *) sp;
198 } */ *uap = v;
199
200 return linux_do_stat64(p, uap, retval, 0);
201 }
202
203 int
204 linux_sys_lstat64(p, v, retval)
205 struct proc *p;
206 void *v;
207 register_t *retval;
208 {
209 struct linux_sys_lstat64_args /* {
210 syscallarg(const char *) path;
211 syscallarg(struct linux_stat64 *) sp;
212 } */ *uap = v;
213
214 return linux_do_stat64(p, uap, retval, 1);
215 }
216
217 int
218 linux_sys_truncate64(p, v, retval)
219 struct proc *p;
220 void *v;
221 register_t *retval;
222 {
223 struct linux_sys_truncate64_args /* {
224 syscallarg(const char *) path;
225 syscallarg(off_t) length;
226 } */ *uap = v;
227 caddr_t sg = stackgap_init(p, 0);
228
229 CHECK_ALT_EXIST(p, &sg, SCARG(uap, path));
230
231 return sys_truncate(p, uap, retval);
232 }
233
234 #if !defined(__m68k__)
235 static void bsd_to_linux_flock64 __P((struct linux_flock64 *,
236 const struct flock *));
237 static void linux_to_bsd_flock64 __P((struct flock *,
238 const struct linux_flock64 *));
239
240 static void
241 bsd_to_linux_flock64(lfp, bfp)
242 struct linux_flock64 *lfp;
243 const struct flock *bfp;
244 {
245
246 lfp->l_start = bfp->l_start;
247 lfp->l_len = bfp->l_len;
248 lfp->l_pid = bfp->l_pid;
249 lfp->l_whence = bfp->l_whence;
250 switch (bfp->l_type) {
251 case F_RDLCK:
252 lfp->l_type = LINUX_F_RDLCK;
253 break;
254 case F_UNLCK:
255 lfp->l_type = LINUX_F_UNLCK;
256 break;
257 case F_WRLCK:
258 lfp->l_type = LINUX_F_WRLCK;
259 break;
260 }
261 }
262
263 static void
264 linux_to_bsd_flock64(bfp, lfp)
265 struct flock *bfp;
266 const struct linux_flock64 *lfp;
267 {
268
269 bfp->l_start = lfp->l_start;
270 bfp->l_len = lfp->l_len;
271 bfp->l_pid = lfp->l_pid;
272 bfp->l_whence = lfp->l_whence;
273 switch (lfp->l_type) {
274 case LINUX_F_RDLCK:
275 bfp->l_type = F_RDLCK;
276 break;
277 case LINUX_F_UNLCK:
278 bfp->l_type = F_UNLCK;
279 break;
280 case LINUX_F_WRLCK:
281 bfp->l_type = F_WRLCK;
282 break;
283 }
284 }
285
286 int
287 linux_sys_fcntl64(p, v, retval)
288 struct proc *p;
289 void *v;
290 register_t *retval;
291 {
292 struct linux_sys_fcntl64_args /* {
293 syscallarg(int) fd;
294 syscallarg(int) cmd;
295 syscallarg(void *) arg;
296 } */ *uap = v;
297 struct sys_fcntl_args fca;
298 struct linux_flock64 lfl;
299 struct flock bfl, *bfp;
300 int error;
301 caddr_t sg;
302 void *arg = SCARG(uap, arg);
303 int cmd = SCARG(uap, cmd);
304 int fd = SCARG(uap, fd);
305
306 switch (cmd) {
307 case LINUX_F_GETLK64:
308 sg = stackgap_init(p, 0);
309 bfp = (struct flock *) stackgap_alloc(p, &sg, sizeof *bfp);
310 if ((error = copyin(arg, &lfl, sizeof lfl)) != 0)
311 return error;
312 linux_to_bsd_flock64(&bfl, &lfl);
313 if ((error = copyout(&bfl, bfp, sizeof bfl)) != 0)
314 return error;
315 SCARG(&fca, fd) = fd;
316 SCARG(&fca, cmd) = F_GETLK;
317 SCARG(&fca, arg) = bfp;
318 if ((error = sys_fcntl(p, &fca, retval)) != 0)
319 return error;
320 if ((error = copyin(bfp, &bfl, sizeof bfl)) != 0)
321 return error;
322 bsd_to_linux_flock64(&lfl, &bfl);
323 return copyout(&lfl, arg, sizeof lfl);
324 case LINUX_F_SETLK64:
325 case LINUX_F_SETLKW64:
326 cmd = (cmd == LINUX_F_SETLK64 ? F_SETLK : F_SETLKW);
327 if ((error = copyin(arg, &lfl, sizeof lfl)) != 0)
328 return error;
329 linux_to_bsd_flock64(&bfl, &lfl);
330 sg = stackgap_init(p, 0);
331 bfp = (struct flock *) stackgap_alloc(p, &sg, sizeof *bfp);
332 if ((error = copyout(&bfl, bfp, sizeof bfl)) != 0)
333 return error;
334 SCARG(&fca, fd) = fd;
335 SCARG(&fca, cmd) = cmd;
336 SCARG(&fca, arg) = bfp;
337 return sys_fcntl(p, &fca, retval);
338 default:
339 return linux_sys_fcntl(p, v, retval);
340 }
341 }
342 #endif /* !m68k */
343
344 #endif /* !alpha */
345
346 /*
347 * Linux 'readdir' call. This code is mostly taken from the
348 * SunOS getdents call (see compat/sunos/sunos_misc.c), though
349 * an attempt has been made to keep it a little cleaner.
350 *
351 * The d_off field contains the offset of the next valid entry,
352 * unless the older Linux getdents(2), which used to have it set
353 * to the offset of the entry itself. This function also doesn't
354 * need to deal with the old count == 1 glibc problem.
355 *
356 * Read in BSD-style entries, convert them, and copy them out.
357 *
358 * Note that this doesn't handle union-mounted filesystems.
359 */
360 int
361 linux_sys_getdents64(p, v, retval)
362 struct proc *p;
363 void *v;
364 register_t *retval;
365 {
366 struct linux_sys_getdents_args /* {
367 syscallarg(int) fd;
368 syscallarg(struct linux_dirent64 *) dent;
369 syscallarg(unsigned int) count;
370 } */ *uap = v;
371 struct dirent *bdp;
372 struct vnode *vp;
373 caddr_t inp, buf; /* BSD-format */
374 int len, reclen; /* BSD-format */
375 caddr_t outp; /* Linux-format */
376 int resid, linux_reclen = 0; /* Linux-format */
377 struct file *fp;
378 struct uio auio;
379 struct iovec aiov;
380 struct linux_dirent64 idb;
381 off_t off; /* true file offset */
382 int buflen, error, eofflag, nbytes;
383 struct vattr va;
384 off_t *cookiebuf = NULL, *cookie;
385 int ncookies;
386
387 /* getvnode() will use the descriptor for us */
388 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
389 return (error);
390
391 if ((fp->f_flag & FREAD) == 0) {
392 error = EBADF;
393 goto out1;
394 }
395
396 vp = (struct vnode *)fp->f_data;
397 if (vp->v_type != VDIR) {
398 error = EINVAL;
399 goto out1;
400 }
401
402 if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)))
403 goto out1;
404
405 nbytes = SCARG(uap, count);
406 buflen = min(MAXBSIZE, nbytes);
407 if (buflen < va.va_blocksize)
408 buflen = va.va_blocksize;
409 buf = malloc(buflen, M_TEMP, M_WAITOK);
410
411 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
412 off = fp->f_offset;
413 again:
414 aiov.iov_base = buf;
415 aiov.iov_len = buflen;
416 auio.uio_iov = &aiov;
417 auio.uio_iovcnt = 1;
418 auio.uio_rw = UIO_READ;
419 auio.uio_segflg = UIO_SYSSPACE;
420 auio.uio_procp = p;
421 auio.uio_resid = buflen;
422 auio.uio_offset = off;
423 /*
424 * First we read into the malloc'ed buffer, then
425 * we massage it into user space, one record at a time.
426 */
427 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf,
428 &ncookies);
429 if (error)
430 goto out;
431
432 inp = buf;
433 outp = (caddr_t)SCARG(uap, dent);
434 resid = nbytes;
435 if ((len = buflen - auio.uio_resid) == 0)
436 goto eof;
437
438 for (cookie = cookiebuf; len > 0; len -= reclen) {
439 bdp = (struct dirent *)inp;
440 reclen = bdp->d_reclen;
441 if (reclen & 3)
442 panic("linux_readdir");
443 if (bdp->d_fileno == 0) {
444 inp += reclen; /* it is a hole; squish it out */
445 off = *cookie++;
446 continue;
447 }
448 linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen);
449 if (reclen > len || resid < linux_reclen) {
450 /* entry too big for buffer, so just stop */
451 outp++;
452 break;
453 }
454 off = *cookie++; /* each entry points to next */
455 /*
456 * Massage in place to make a Linux-shaped dirent (otherwise
457 * we have to worry about touching user memory outside of
458 * the copyout() call).
459 */
460 idb.d_ino = bdp->d_fileno;
461 idb.d_type = bdp->d_type;
462 idb.d_off = off;
463 idb.d_reclen = (u_short)linux_reclen;
464 strcpy(idb.d_name, bdp->d_name);
465 if ((error = copyout((caddr_t)&idb, outp, linux_reclen)))
466 goto out;
467 /* advance past this real entry */
468 inp += reclen;
469 /* advance output past Linux-shaped entry */
470 outp += linux_reclen;
471 resid -= linux_reclen;
472 }
473
474 /* if we squished out the whole block, try again */
475 if (outp == (caddr_t)SCARG(uap, dent))
476 goto again;
477 fp->f_offset = off; /* update the vnode offset */
478
479 eof:
480 *retval = nbytes - resid;
481 out:
482 VOP_UNLOCK(vp, 0);
483 if (cookiebuf)
484 free(cookiebuf, M_TEMP);
485 free(buf, M_TEMP);
486 out1:
487 FILE_UNUSE(fp, p);
488 return error;
489 }
490