xref: /netbsd-src/sys/compat/linux/common/linux_file64.c (revision 8b0f9554ff8762542c4defc4f70e1eb76fb508fa)
1 /*	$NetBSD: linux_file64.c,v 1.43 2007/12/08 18:36:07 dsl 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 1.43 2007/12/08 18:36:07 dsl 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/namei.h>
59 #include <sys/vfs_syscalls.h>
60 #include <sys/vnode.h>
61 #include <sys/tty.h>
62 #include <sys/conf.h>
63 
64 #include <sys/syscallargs.h>
65 
66 #include <compat/linux/common/linux_types.h>
67 #include <compat/linux/common/linux_signal.h>
68 #include <compat/linux/common/linux_fcntl.h>
69 #include <compat/linux/common/linux_util.h>
70 #include <compat/linux/common/linux_machdep.h>
71 #include <compat/linux/common/linux_dirent.h>
72 #include <compat/linux/common/linux_ipc.h>
73 #include <compat/linux/common/linux_sem.h>
74 
75 #include <compat/linux/linux_syscallargs.h>
76 
77 #ifndef alpha
78 
79 # ifndef COMPAT_LINUX32
80 
81 static void bsd_to_linux_stat(struct stat *, struct linux_stat64 *);
82 static int linux_do_stat64(struct lwp *, void *, register_t *, int);
83 
84 /*
85  * Convert a NetBSD stat structure to a Linux stat structure.
86  * Only the order of the fields and the padding in the structure
87  * is different. linux_fakedev is a machine-dependent function
88  * which optionally converts device driver major/minor numbers
89  * (XXX horrible, but what can you do against code that compares
90  * things against constant major device numbers? sigh)
91  */
92 static void
93 bsd_to_linux_stat(struct stat *bsp, struct linux_stat64 *lsp)
94 {
95 	lsp->lst_dev     = linux_fakedev(bsp->st_dev, 0);
96 	lsp->lst_ino     = bsp->st_ino;
97 	lsp->lst_mode    = (linux_mode_t)bsp->st_mode;
98 	if (bsp->st_nlink >= (1 << 15))
99 		lsp->lst_nlink = (1 << 15) - 1;
100 	else
101 		lsp->lst_nlink = (linux_nlink_t)bsp->st_nlink;
102 	lsp->lst_uid     = bsp->st_uid;
103 	lsp->lst_gid     = bsp->st_gid;
104 	lsp->lst_rdev    = linux_fakedev(bsp->st_rdev, 1);
105 	lsp->lst_size    = bsp->st_size;
106 	lsp->lst_blksize = bsp->st_blksize;
107 	lsp->lst_blocks  = bsp->st_blocks;
108 	lsp->lst_atime   = bsp->st_atime;
109 	lsp->lst_mtime   = bsp->st_mtime;
110 	lsp->lst_ctime   = bsp->st_ctime;
111 #  ifdef LINUX_STAT64_HAS_NSEC
112 	lsp->lst_atime_nsec   = bsp->st_atimensec;
113 	lsp->lst_mtime_nsec   = bsp->st_mtimensec;
114 	lsp->lst_ctime_nsec   = bsp->st_ctimensec;
115 #  endif
116 #  if LINUX_STAT64_HAS_BROKEN_ST_INO
117 	lsp->__lst_ino   = (linux_ino_t) bsp->st_ino;
118 #  endif
119 }
120 
121 /*
122  * The stat functions below are plain sailing. stat and lstat are handled
123  * by one function to avoid code duplication.
124  */
125 int
126 linux_sys_fstat64(struct lwp *l, void *v, register_t *retval)
127 {
128 	struct linux_sys_fstat64_args /* {
129 		syscallarg(int) fd;
130 		syscallarg(struct linux_stat64 *) sp;
131 	} */ *uap = v;
132 	struct linux_stat64 tmplst;
133 	struct stat tmpst;
134 	int error;
135 
136 	error = do_sys_fstat(l,  SCARG(uap, fd), &tmpst);
137 	if (error != 0)
138 		return error;
139 
140 	bsd_to_linux_stat(&tmpst, &tmplst);
141 
142 	return copyout(&tmplst, SCARG(uap, sp), sizeof tmplst);
143 }
144 
145 static int
146 linux_do_stat64(struct lwp *l, void *v, register_t *retval, int flags)
147 {
148 	struct linux_stat64 tmplst;
149 	struct stat tmpst;
150 	int error;
151 	struct linux_sys_stat64_args *uap = v;
152 
153 	error = do_sys_stat(l, SCARG(uap, path), flags, &tmpst);
154 	if (error != 0)
155 		return error;
156 
157 	bsd_to_linux_stat(&tmpst, &tmplst);
158 
159 	return copyout(&tmplst, SCARG(uap, sp), sizeof tmplst);
160 }
161 
162 int
163 linux_sys_stat64(struct lwp *l, void *v, register_t *retval)
164 {
165 	struct linux_sys_stat64_args /* {
166 		syscallarg(const char *) path;
167 		syscallarg(struct linux_stat64 *) sp;
168 	} */ *uap = v;
169 
170 	return linux_do_stat64(l, uap, retval, FOLLOW);
171 }
172 
173 int
174 linux_sys_lstat64(struct lwp *l, void *v, register_t *retval)
175 {
176 	struct linux_sys_lstat64_args /* {
177 		syscallarg(const char *) path;
178 		syscallarg(struct linux_stat64 *) sp;
179 	} */ *uap = v;
180 
181 	return linux_do_stat64(l, uap, retval, NOFOLLOW);
182 }
183 
184 int
185 linux_sys_truncate64(struct lwp *l, void *v, register_t *retval)
186 {
187 	struct linux_sys_truncate64_args /* {
188 		syscallarg(const char *) path;
189 		syscallarg(off_t) length;
190 	} */ *uap = v;
191 	struct sys_truncate_args ta;
192 
193 	/* Linux doesn't have the 'pad' pseudo-parameter */
194 	SCARG(&ta, path) = SCARG(uap, path);
195 	SCARG(&ta, pad) = 0;
196 	SCARG(&ta, length) = SCARG(uap, length);
197 
198 	return sys_truncate(l, &ta, retval);
199 }
200 
201 int
202 linux_sys_ftruncate64(struct lwp *l, void *v, register_t *retval)
203 {
204 	struct linux_sys_ftruncate64_args /* {
205 		syscallarg(unsigned int) fd;
206 		syscallarg(off_t) length;
207 	} */ *uap = v;
208 	struct sys_ftruncate_args ta;
209 
210 	/* Linux doesn't have the 'pad' pseudo-parameter */
211 	SCARG(&ta, fd) = SCARG(uap, fd);
212 	SCARG(&ta, pad) = 0;
213 	SCARG(&ta, length) = SCARG(uap, length);
214 
215 	return sys_ftruncate(l, &ta, retval);
216 }
217 # endif /* !COMPAT_LINUX32 */
218 
219 # if !defined(__m68k__) && (!defined(__amd64__) || defined(COMPAT_LINUX32))
220 static void bsd_to_linux_flock64(struct linux_flock64 *,
221     const struct flock *);
222 static void linux_to_bsd_flock64(struct flock *,
223     const struct linux_flock64 *);
224 
225 static void
226 bsd_to_linux_flock64(struct linux_flock64 *lfp, const struct flock *bfp)
227 {
228 
229 	lfp->l_start = bfp->l_start;
230 	lfp->l_len = bfp->l_len;
231 	lfp->l_pid = bfp->l_pid;
232 	lfp->l_whence = bfp->l_whence;
233 	switch (bfp->l_type) {
234 	case F_RDLCK:
235 		lfp->l_type = LINUX_F_RDLCK;
236 		break;
237 	case F_UNLCK:
238 		lfp->l_type = LINUX_F_UNLCK;
239 		break;
240 	case F_WRLCK:
241 		lfp->l_type = LINUX_F_WRLCK;
242 		break;
243 	}
244 }
245 
246 static void
247 linux_to_bsd_flock64(struct flock *bfp, const struct linux_flock64 *lfp)
248 {
249 
250 	bfp->l_start = lfp->l_start;
251 	bfp->l_len = lfp->l_len;
252 	bfp->l_pid = lfp->l_pid;
253 	bfp->l_whence = lfp->l_whence;
254 	switch (lfp->l_type) {
255 	case LINUX_F_RDLCK:
256 		bfp->l_type = F_RDLCK;
257 		break;
258 	case LINUX_F_UNLCK:
259 		bfp->l_type = F_UNLCK;
260 		break;
261 	case LINUX_F_WRLCK:
262 		bfp->l_type = F_WRLCK;
263 		break;
264 	}
265 }
266 
267 int
268 linux_sys_fcntl64(struct lwp *l, void *v, register_t *retval)
269 {
270 	struct linux_sys_fcntl64_args /* {
271 		syscallarg(int) fd;
272 		syscallarg(int) cmd;
273 		syscallarg(void *) arg;
274 	} */ *uap = v;
275 	struct linux_flock64 lfl;
276 	struct flock bfl;
277 	int error;
278 	void *arg = SCARG(uap, arg);
279 	int cmd = SCARG(uap, cmd);
280 	int fd = SCARG(uap, fd);
281 
282 	switch (cmd) {
283 	case LINUX_F_GETLK64:
284 		if ((error = copyin(arg, &lfl, sizeof lfl)) != 0)
285 			return error;
286 		linux_to_bsd_flock64(&bfl, &lfl);
287 		error = do_fcntl_lock(l, fd, F_GETLK, &bfl);
288 		if (error != 0)
289 			return error;
290 		bsd_to_linux_flock64(&lfl, &bfl);
291 		return copyout(&lfl, arg, sizeof lfl);
292 	case LINUX_F_SETLK64:
293 	case LINUX_F_SETLKW64:
294 		cmd = (cmd == LINUX_F_SETLK64 ? F_SETLK : F_SETLKW);
295 		if ((error = copyin(arg, &lfl, sizeof lfl)) != 0)
296 			return error;
297 		linux_to_bsd_flock64(&bfl, &lfl);
298 		return do_fcntl_lock(l, fd, cmd, &bfl);
299 	default:
300 		return linux_sys_fcntl(l, v, retval);
301 	}
302 }
303 # endif /* !m69k && !amd64  && !COMPAT_LINUX32 */
304 
305 #endif /* !alpha */
306 
307 /*
308  * Linux 'readdir' call. This code is mostly taken from the
309  * SunOS getdents call (see compat/sunos/sunos_misc.c), though
310  * an attempt has been made to keep it a little cleaner.
311  *
312  * The d_off field contains the offset of the next valid entry,
313  * unless the older Linux getdents(2), which used to have it set
314  * to the offset of the entry itself. This function also doesn't
315  * need to deal with the old count == 1 glibc problem.
316  *
317  * Read in BSD-style entries, convert them, and copy them out.
318  *
319  * Note that this doesn't handle union-mounted filesystems.
320  */
321 #ifndef COMPAT_LINUX32
322 int
323 linux_sys_getdents64(struct lwp *l, void *v, register_t *retval)
324 {
325 	struct linux_sys_getdents_args /* {
326 		syscallarg(int) fd;
327 		syscallarg(struct linux_dirent64 *) dent;
328 		syscallarg(unsigned int) count;
329 	} */ *uap = v;
330 	struct dirent *bdp;
331 	struct vnode *vp;
332 	char *inp, *tbuf;		/* BSD-format */
333 	int len, reclen;		/* BSD-format */
334 	char *outp;			/* Linux-format */
335 	int resid, linux_reclen = 0;	/* Linux-format */
336 	struct file *fp;
337 	struct uio auio;
338 	struct iovec aiov;
339 	struct linux_dirent64 idb;
340 	off_t off;		/* true file offset */
341 	int buflen, error, eofflag, nbytes;
342 	struct vattr va;
343 	off_t *cookiebuf = NULL, *cookie;
344 	int ncookies;
345 
346 	/* getvnode() will use the descriptor for us */
347 	if ((error = getvnode(l->l_proc->p_fd, SCARG(uap, fd), &fp)) != 0)
348 		return (error);
349 
350 	if ((fp->f_flag & FREAD) == 0) {
351 		error = EBADF;
352 		goto out1;
353 	}
354 
355 	vp = (struct vnode *)fp->f_data;
356 	if (vp->v_type != VDIR) {
357 		error = EINVAL;
358 		goto out1;
359 	}
360 
361 	if ((error = VOP_GETATTR(vp, &va, l->l_cred)))
362 		goto out1;
363 
364 	nbytes = SCARG(uap, count);
365 	buflen = min(MAXBSIZE, nbytes);
366 	if (buflen < va.va_blocksize)
367 		buflen = va.va_blocksize;
368 	tbuf = malloc(buflen, M_TEMP, M_WAITOK);
369 
370 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
371 	off = fp->f_offset;
372 again:
373 	aiov.iov_base = tbuf;
374 	aiov.iov_len = buflen;
375 	auio.uio_iov = &aiov;
376 	auio.uio_iovcnt = 1;
377 	auio.uio_rw = UIO_READ;
378 	auio.uio_resid = buflen;
379 	auio.uio_offset = off;
380 	UIO_SETUP_SYSSPACE(&auio);
381 	/*
382          * First we read into the malloc'ed buffer, then
383          * we massage it into user space, one record at a time.
384          */
385 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf,
386 	    &ncookies);
387 	if (error)
388 		goto out;
389 
390 	inp = tbuf;
391 	outp = (void *)SCARG(uap, dent);
392 	resid = nbytes;
393 	if ((len = buflen - auio.uio_resid) == 0)
394 		goto eof;
395 
396 	for (cookie = cookiebuf; len > 0; len -= reclen) {
397 		bdp = (struct dirent *)inp;
398 		reclen = bdp->d_reclen;
399 		if (reclen & 3)
400 			panic("linux_readdir");
401 		if (bdp->d_fileno == 0) {
402 			inp += reclen;	/* it is a hole; squish it out */
403 			if (cookie)
404 				off = *cookie++;
405 			else
406 				off += reclen;
407 			continue;
408 		}
409 		linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen);
410 		if (reclen > len || resid < linux_reclen) {
411 			/* entry too big for buffer, so just stop */
412 			outp++;
413 			break;
414 		}
415 		if (cookie)
416 			off = *cookie++;	/* each entry points to next */
417 		else
418 			off += reclen;
419 		/*
420 		 * Massage in place to make a Linux-shaped dirent (otherwise
421 		 * we have to worry about touching user memory outside of
422 		 * the copyout() call).
423 		 */
424 		idb.d_ino = bdp->d_fileno;
425 		idb.d_type = bdp->d_type;
426 		idb.d_off = off;
427 		idb.d_reclen = (u_short)linux_reclen;
428 		strcpy(idb.d_name, bdp->d_name);
429 		if ((error = copyout((void *)&idb, outp, linux_reclen)))
430 			goto out;
431 		/* advance past this real entry */
432 		inp += reclen;
433 		/* advance output past Linux-shaped entry */
434 		outp += linux_reclen;
435 		resid -= linux_reclen;
436 	}
437 
438 	/* if we squished out the whole block, try again */
439 	if (outp == (void *)SCARG(uap, dent))
440 		goto again;
441 	fp->f_offset = off;	/* update the vnode offset */
442 
443 eof:
444 	*retval = nbytes - resid;
445 out:
446 	VOP_UNLOCK(vp, 0);
447 	if (cookiebuf)
448 		free(cookiebuf, M_TEMP);
449 	free(tbuf, M_TEMP);
450 out1:
451 	FILE_UNUSE(fp, l);
452 	return error;
453 }
454 #endif /* !COMPAT_LINUX32 */
455