linux_file64.c revision 1.11 1 /* $NetBSD: linux_file64.c,v 1.11 2002/05/10 14:38:08 tron 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 1.11 2002/05/10 14:38:08 tron 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 static void bsd_to_linux_stat __P((struct stat *, struct linux_stat64 *));
74 static int linux_do_stat64 __P((struct proc *, void *, register_t *, int));
75
76 /*
77 * Convert a NetBSD stat structure to a Linux stat structure.
78 * Only the order of the fields and the padding in the structure
79 * is different. linux_fakedev is a machine-dependent function
80 * which optionally converts device driver major/minor numbers
81 * (XXX horrible, but what can you do against code that compares
82 * things against constant major device numbers? sigh)
83 */
84 static void
85 bsd_to_linux_stat(bsp, lsp)
86 struct stat *bsp;
87 struct linux_stat64 *lsp;
88 {
89 lsp->lst_dev = linux_fakedev(bsp->st_dev, 0);
90 lsp->lst_ino = bsp->st_ino;
91 lsp->lst_mode = (linux_mode_t)bsp->st_mode;
92 if (bsp->st_nlink >= (1 << 15))
93 lsp->lst_nlink = (1 << 15) - 1;
94 else
95 lsp->lst_nlink = (linux_nlink_t)bsp->st_nlink;
96 lsp->lst_uid = bsp->st_uid;
97 lsp->lst_gid = bsp->st_gid;
98 lsp->lst_rdev = linux_fakedev(bsp->st_rdev, 1);
99 lsp->lst_size = bsp->st_size;
100 lsp->lst_blksize = bsp->st_blksize;
101 lsp->lst_blocks = bsp->st_blocks;
102 lsp->lst_atime = bsp->st_atime;
103 lsp->lst_mtime = bsp->st_mtime;
104 lsp->lst_ctime = bsp->st_ctime;
105 lsp->lst_ino64 = bsp->st_ino;
106 }
107
108 /*
109 * The stat functions below are plain sailing. stat and lstat are handled
110 * by one function to avoid code duplication.
111 */
112 int
113 linux_sys_fstat64(p, v, retval)
114 struct proc *p;
115 void *v;
116 register_t *retval;
117 {
118 struct linux_sys_fstat64_args /* {
119 syscallarg(int) fd;
120 syscallarg(struct linux_stat64 *) sp;
121 } */ *uap = v;
122 struct sys___fstat13_args fsa;
123 struct linux_stat64 tmplst;
124 struct stat *st,tmpst;
125 caddr_t sg;
126 int error;
127
128 sg = stackgap_init(p, 0);
129
130 st = stackgap_alloc(p, &sg, sizeof (struct stat));
131
132 SCARG(&fsa, fd) = SCARG(uap, fd);
133 SCARG(&fsa, sb) = st;
134
135 if ((error = sys___fstat13(p, &fsa, retval)))
136 return error;
137
138 if ((error = copyin(st, &tmpst, sizeof tmpst)))
139 return error;
140
141 bsd_to_linux_stat(&tmpst, &tmplst);
142
143 if ((error = copyout(&tmplst, SCARG(uap, sp), sizeof tmplst)))
144 return error;
145
146 return 0;
147 }
148
149 static int
150 linux_do_stat64(p, v, retval, dolstat)
151 struct proc *p;
152 void *v;
153 register_t *retval;
154 int dolstat;
155 {
156 struct sys___stat13_args sa;
157 struct linux_stat64 tmplst;
158 struct stat *st, tmpst;
159 caddr_t sg;
160 int error;
161 struct linux_sys_stat64_args *uap = v;
162
163 sg = stackgap_init(p, 0);
164 st = stackgap_alloc(p, &sg, sizeof (struct stat));
165 CHECK_ALT_EXIST(p, &sg, SCARG(uap, path));
166
167 SCARG(&sa, ub) = st;
168 SCARG(&sa, path) = SCARG(uap, path);
169
170 if ((error = (dolstat ? sys___lstat13(p, &sa, retval) :
171 sys___stat13(p, &sa, retval))))
172 return error;
173
174 if ((error = copyin(st, &tmpst, sizeof tmpst)))
175 return error;
176
177 bsd_to_linux_stat(&tmpst, &tmplst);
178
179 if ((error = copyout(&tmplst, SCARG(uap, sp), sizeof tmplst)))
180 return error;
181
182 return 0;
183 }
184
185 int
186 linux_sys_stat64(p, v, retval)
187 struct proc *p;
188 void *v;
189 register_t *retval;
190 {
191 struct linux_sys_stat64_args /* {
192 syscallarg(const char *) path;
193 syscallarg(struct linux_stat64 *) sp;
194 } */ *uap = v;
195
196 return linux_do_stat64(p, uap, retval, 0);
197 }
198
199 int
200 linux_sys_lstat64(p, v, retval)
201 struct proc *p;
202 void *v;
203 register_t *retval;
204 {
205 struct linux_sys_lstat64_args /* {
206 syscallarg(const char *) path;
207 syscallarg(struct linux_stat64 *) sp;
208 } */ *uap = v;
209
210 return linux_do_stat64(p, uap, retval, 1);
211 }
212
213 int
214 linux_sys_truncate64(p, v, retval)
215 struct proc *p;
216 void *v;
217 register_t *retval;
218 {
219 struct linux_sys_truncate64_args /* {
220 syscallarg(const char *) path;
221 syscallarg(off_t) length;
222 } */ *uap = v;
223 caddr_t sg = stackgap_init(p, 0);
224
225 CHECK_ALT_EXIST(p, &sg, SCARG(uap, path));
226
227 return sys_truncate(p, uap, retval);
228 }
229
230 #if defined(__mips__) || defined(__i386__) /* powerpc could use it too */
231 static void bsd_to_linux_flock64 __P((struct linux_flock64 *,
232 const struct flock *));
233 static void linux_to_bsd_flock64 __P((struct flock *,
234 const struct linux_flock64 *));
235
236 static void
237 bsd_to_linux_flock64(lfp, bfp)
238 struct linux_flock64 *lfp;
239 const struct flock *bfp;
240 {
241
242 lfp->l_start = bfp->l_start;
243 lfp->l_len = bfp->l_len;
244 lfp->l_pid = bfp->l_pid;
245 lfp->l_whence = bfp->l_whence;
246 switch (bfp->l_type) {
247 case F_RDLCK:
248 lfp->l_type = LINUX_F_RDLCK;
249 break;
250 case F_UNLCK:
251 lfp->l_type = LINUX_F_UNLCK;
252 break;
253 case F_WRLCK:
254 lfp->l_type = LINUX_F_WRLCK;
255 break;
256 }
257 }
258
259 static void
260 linux_to_bsd_flock64(bfp, lfp)
261 struct flock *bfp;
262 const struct linux_flock64 *lfp;
263 {
264
265 bfp->l_start = lfp->l_start;
266 bfp->l_len = lfp->l_len;
267 bfp->l_pid = lfp->l_pid;
268 bfp->l_whence = lfp->l_whence;
269 switch (lfp->l_type) {
270 case LINUX_F_RDLCK:
271 bfp->l_type = F_RDLCK;
272 break;
273 case LINUX_F_UNLCK:
274 bfp->l_type = F_UNLCK;
275 break;
276 case LINUX_F_WRLCK:
277 bfp->l_type = F_WRLCK;
278 break;
279 }
280 }
281 int
282 linux_sys_fcntl64(p, v, retval)
283 struct proc *p;
284 void *v;
285 register_t *retval;
286 {
287 struct linux_sys_fcntl64_args /* {
288 syscallarg(int) fd;
289 syscallarg(int) cmd;
290 syscallarg(void *) arg;
291 } */ *uap = v;
292 struct sys_fcntl_args fca;
293 struct linux_flock64 lfl;
294 struct flock bfl, *bfp;
295 int error;
296 caddr_t sg;
297 void *arg = SCARG(uap, arg);
298 int cmd = SCARG(uap, cmd);
299 int fd = SCARG(uap, fd);
300
301 switch (cmd) {
302 case LINUX_F_GETLK64:
303 sg = stackgap_init(p, 0);
304 bfp = (struct flock *) stackgap_alloc(p, &sg, sizeof *bfp);
305 if ((error = copyin(arg, &lfl, sizeof lfl)) != 0)
306 return error;
307 linux_to_bsd_flock64(&bfl, &lfl);
308 if ((error = copyout(&bfl, bfp, sizeof bfl)) != 0)
309 return error;
310 SCARG(&fca, fd) = fd;
311 SCARG(&fca, cmd) = F_GETLK;
312 SCARG(&fca, arg) = bfp;
313 if ((error = sys_fcntl(p, &fca, retval)) != 0)
314 return error;
315 if ((error = copyin(bfp, &bfl, sizeof bfl)) != 0)
316 return error;
317 bsd_to_linux_flock64(&lfl, &bfl);
318 return copyout(&lfl, arg, sizeof lfl);
319 case LINUX_F_SETLK64:
320 case LINUX_F_SETLKW64:
321 cmd = (cmd == LINUX_F_SETLK64 ? F_SETLK : F_SETLKW);
322 if ((error = copyin(arg, &lfl, sizeof lfl)) != 0)
323 return error;
324 linux_to_bsd_flock64(&bfl, &lfl);
325 sg = stackgap_init(p, 0);
326 bfp = (struct flock *) stackgap_alloc(p, &sg, sizeof *bfp);
327 if ((error = copyout(&bfl, bfp, sizeof bfl)) != 0)
328 return error;
329 SCARG(&fca, fd) = fd;
330 SCARG(&fca, cmd) = cmd;
331 SCARG(&fca, arg) = bfp;
332 return sys_fcntl(p, &fca, retval);
333 default:
334 return linux_sys_fcntl(p, v, retval);
335 }
336
337 return error;
338 }
339
340 /*
341 * Linux 'readdir' call. This code is mostly taken from the
342 * SunOS getdents call (see compat/sunos/sunos_misc.c), though
343 * an attempt has been made to keep it a little cleaner (failing
344 * miserably, because of the cruft needed if count 1 is passed).
345 *
346 * The d_off field should contain the offset of the next valid entry,
347 * but in Linux it has the offset of the entry itself. We emulate
348 * that bug here.
349 *
350 * Read in BSD-style entries, convert them, and copy them out.
351 *
352 * Note that this doesn't handle union-mounted filesystems.
353 */
354 int
355 linux_sys_getdents64(p, v, retval)
356 struct proc *p;
357 void *v;
358 register_t *retval;
359 {
360 struct linux_sys_getdents_args /* {
361 syscallarg(int) fd;
362 syscallarg(struct linux_dirent64 *) dent;
363 syscallarg(unsigned int) count;
364 } */ *uap = v;
365 struct dirent *bdp;
366 struct vnode *vp;
367 caddr_t inp, buf; /* BSD-format */
368 int len, reclen; /* BSD-format */
369 caddr_t outp; /* Linux-format */
370 int resid, linux_reclen = 0; /* Linux-format */
371 struct file *fp;
372 struct uio auio;
373 struct iovec aiov;
374 struct linux_dirent64 idb;
375 off_t off; /* true file offset */
376 int buflen, error, eofflag, nbytes, oldcall;
377 struct vattr va;
378 off_t *cookiebuf = NULL, *cookie;
379 int ncookies;
380
381 /* getvnode() will use the descriptor for us */
382 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
383 return (error);
384
385 if ((fp->f_flag & FREAD) == 0) {
386 error = EBADF;
387 goto out1;
388 }
389
390 vp = (struct vnode *)fp->f_data;
391 if (vp->v_type != VDIR) {
392 error = EINVAL;
393 goto out1;
394 }
395
396 if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)))
397 goto out1;
398
399 nbytes = SCARG(uap, count);
400 if (nbytes == 1) { /* emulating old, broken behaviour */
401 nbytes = sizeof (idb);
402 buflen = max(va.va_blocksize, nbytes);
403 oldcall = 1;
404 } else {
405 buflen = min(MAXBSIZE, nbytes);
406 if (buflen < va.va_blocksize)
407 buflen = va.va_blocksize;
408 oldcall = 0;
409 }
410 buf = malloc(buflen, M_TEMP, M_WAITOK);
411
412 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
413 off = fp->f_offset;
414 again:
415 aiov.iov_base = buf;
416 aiov.iov_len = buflen;
417 auio.uio_iov = &aiov;
418 auio.uio_iovcnt = 1;
419 auio.uio_rw = UIO_READ;
420 auio.uio_segflg = UIO_SYSSPACE;
421 auio.uio_procp = p;
422 auio.uio_resid = buflen;
423 auio.uio_offset = off;
424 /*
425 * First we read into the malloc'ed buffer, then
426 * we massage it into user space, one record at a time.
427 */
428 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf,
429 &ncookies);
430 if (error)
431 goto out;
432
433 inp = buf;
434 outp = (caddr_t)SCARG(uap, dent);
435 resid = nbytes;
436 if ((len = buflen - auio.uio_resid) == 0)
437 goto eof;
438
439 for (cookie = cookiebuf; len > 0; len -= reclen) {
440 bdp = (struct dirent *)inp;
441 reclen = bdp->d_reclen;
442 if (reclen & 3)
443 panic("linux_readdir");
444 if (bdp->d_fileno == 0) {
445 inp += reclen; /* it is a hole; squish it out */
446 off = *cookie++;
447 continue;
448 }
449 linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen);
450 if (reclen > len || resid < linux_reclen) {
451 /* entry too big for buffer, so just stop */
452 outp++;
453 break;
454 }
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 /*
463 * The old readdir() call misuses the offset and reclen fields.
464 */
465 if (oldcall) {
466 idb.d_off = linux_reclen;
467 idb.d_reclen = (u_short)bdp->d_namlen;
468 } else {
469 if (sizeof (idb.d_off) < 4 && (off >> 32) != 0) {
470 compat_offseterr(vp, "linux_getdents");
471 error = EINVAL;
472 goto out;
473 }
474 idb.d_off = off;
475 idb.d_reclen = (u_short)linux_reclen;
476 }
477 strcpy(idb.d_name, bdp->d_name);
478 if ((error = copyout((caddr_t)&idb, outp, linux_reclen)))
479 goto out;
480 /* advance past this real entry */
481 inp += reclen;
482 off = *cookie++; /* each entry points to itself */
483 /* advance output past Linux-shaped entry */
484 outp += linux_reclen;
485 resid -= linux_reclen;
486 if (oldcall)
487 break;
488 }
489
490 /* if we squished out the whole block, try again */
491 if (outp == (caddr_t)SCARG(uap, dent))
492 goto again;
493 fp->f_offset = off; /* update the vnode offset */
494
495 if (oldcall)
496 nbytes = resid + linux_reclen;
497
498 eof:
499 *retval = nbytes - resid;
500 out:
501 VOP_UNLOCK(vp, 0);
502 if (cookiebuf)
503 free(cookiebuf, M_TEMP);
504 free(buf, M_TEMP);
505 out1:
506 FILE_UNUSE(fp, p);
507 return error;
508 }
509 #endif /* __mips__ || __i386__ */
510