xref: /netbsd-src/sys/compat/linux/common/linux_file64.c (revision fb5eed702691094bd687fbf1ded189c87457cd35)
1 /*	$NetBSD: linux_file64.c,v 1.64 2021/09/23 06:56:27 ryo Exp $	*/
2 
3 /*-
4  * Copyright (c) 1995, 1998, 2000, 2008 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  *
19  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
20  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
21  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
22  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
23  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
25  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
26  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
27  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
28  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29  * POSSIBILITY OF SUCH DAMAGE.
30  */
31 
32 /*
33  * Linux 64bit filesystem calls. Used on 32bit archs, not used on 64bit ones.
34  */
35 
36 #include <sys/cdefs.h>
37 __KERNEL_RCSID(0, "$NetBSD: linux_file64.c,v 1.64 2021/09/23 06:56:27 ryo Exp $");
38 
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/namei.h>
42 #include <sys/proc.h>
43 #include <sys/dirent.h>
44 #include <sys/file.h>
45 #include <sys/stat.h>
46 #include <sys/filedesc.h>
47 #include <sys/ioctl.h>
48 #include <sys/kernel.h>
49 #include <sys/mount.h>
50 #include <sys/malloc.h>
51 #include <sys/namei.h>
52 #include <sys/vfs_syscalls.h>
53 #include <sys/vnode.h>
54 #include <sys/tty.h>
55 #include <sys/conf.h>
56 
57 #include <sys/syscallargs.h>
58 
59 #include <compat/linux/common/linux_types.h>
60 #include <compat/linux/common/linux_signal.h>
61 #include <compat/linux/common/linux_fcntl.h>
62 #include <compat/linux/common/linux_util.h>
63 #include <compat/linux/common/linux_machdep.h>
64 #include <compat/linux/common/linux_dirent.h>
65 #include <compat/linux/common/linux_ipc.h>
66 #include <compat/linux/common/linux_sem.h>
67 
68 #include <compat/linux/linux_syscallargs.h>
69 
70 static void bsd_to_linux_stat(struct stat *, struct linux_stat64 *);
71 
72 /*
73  * Convert a NetBSD stat structure to a Linux stat structure.
74  * Only the order of the fields and the padding in the structure
75  * is different. linux_fakedev is a machine-dependent function
76  * which optionally converts device driver major/minor numbers
77  * (XXX horrible, but what can you do against code that compares
78  * things against constant major device numbers? sigh)
79  */
80 static void
81 bsd_to_linux_stat(struct stat *bsp, struct linux_stat64 *lsp)
82 {
83 	memset(lsp, 0, sizeof(*lsp));
84 	lsp->lst_dev     = linux_fakedev(bsp->st_dev, 0);
85 	lsp->lst_ino     = bsp->st_ino;
86 	lsp->lst_mode    = (linux_mode_t)bsp->st_mode;
87 	if (bsp->st_nlink >= (1 << 15))
88 		lsp->lst_nlink = (1 << 15) - 1;
89 	else
90 		lsp->lst_nlink = (linux_nlink_t)bsp->st_nlink;
91 	lsp->lst_uid     = bsp->st_uid;
92 	lsp->lst_gid     = bsp->st_gid;
93 	lsp->lst_rdev    = linux_fakedev(bsp->st_rdev, 1);
94 	lsp->lst_size    = bsp->st_size;
95 	lsp->lst_blksize = bsp->st_blksize;
96 	lsp->lst_blocks  = bsp->st_blocks;
97 	lsp->lst_atime   = bsp->st_atime;
98 	lsp->lst_mtime   = bsp->st_mtime;
99 	lsp->lst_ctime   = bsp->st_ctime;
100 #  ifdef LINUX_STAT64_HAS_NSEC
101 	lsp->lst_atime_nsec   = bsp->st_atimensec;
102 	lsp->lst_mtime_nsec   = bsp->st_mtimensec;
103 	lsp->lst_ctime_nsec   = bsp->st_ctimensec;
104 #  endif
105 #  if LINUX_STAT64_HAS_BROKEN_ST_INO
106 	lsp->__lst_ino   = (linux_ino_t) bsp->st_ino;
107 #  endif
108 }
109 
110 /*
111  * The stat functions below are plain sailing. stat and lstat are handled
112  * by one function to avoid code duplication.
113  */
114 int
115 linux_sys_fstat64(struct lwp *l, const struct linux_sys_fstat64_args *uap, register_t *retval)
116 {
117 	/* {
118 		syscallarg(int) fd;
119 		syscallarg(struct linux_stat64 *) sp;
120 	} */
121 	struct linux_stat64 tmplst;
122 	struct stat tmpst;
123 	int error;
124 
125 	error = do_sys_fstat(SCARG(uap, fd), &tmpst);
126 	if (error != 0)
127 		return error;
128 
129 	bsd_to_linux_stat(&tmpst, &tmplst);
130 
131 	return copyout(&tmplst, SCARG(uap, sp), sizeof tmplst);
132 }
133 
134 #if !defined(__aarch64__)
135 static int
136 linux_do_stat64(struct lwp *l, const struct linux_sys_stat64_args *uap, register_t *retval, int flags)
137 {
138 	struct linux_stat64 tmplst;
139 	struct stat tmpst;
140 	int error;
141 
142 	error = do_sys_stat(SCARG(uap, path), flags, &tmpst);
143 	if (error != 0)
144 		return error;
145 
146 	bsd_to_linux_stat(&tmpst, &tmplst);
147 
148 	return copyout(&tmplst, SCARG(uap, sp), sizeof tmplst);
149 }
150 
151 int
152 linux_sys_stat64(struct lwp *l, const struct linux_sys_stat64_args *uap, register_t *retval)
153 {
154 	/* {
155 		syscallarg(const char *) path;
156 		syscallarg(struct linux_stat64 *) sp;
157 	} */
158 
159 	return linux_do_stat64(l, uap, retval, FOLLOW);
160 }
161 
162 int
163 linux_sys_lstat64(struct lwp *l, const struct linux_sys_lstat64_args *uap, register_t *retval)
164 {
165 	/* {
166 		syscallarg(const char *) path;
167 		syscallarg(struct linux_stat64 *) sp;
168 	} */
169 
170 	return linux_do_stat64(l, (const void *)uap, retval, NOFOLLOW);
171 }
172 #endif
173 
174 int
175 linux_sys_fstatat64(struct lwp *l, const struct linux_sys_fstatat64_args *uap, register_t *retval)
176 {
177 	/* {
178 		syscallarg(int) fd;
179 		syscallarg(const char *) path;
180 		syscallarg(struct linux_stat64 *) sp;
181 		syscallarg(int) flag;
182 	} */
183 	struct linux_stat64 tmplst;
184 	struct stat tmpst;
185 	struct vnode *vp;
186 	int error, nd_flag, fd;
187 	uint8_t c;
188 
189 	if (SCARG(uap, flag) & LINUX_AT_EMPTY_PATH) {
190 		/*
191 		 * If path is null string:
192 		 */
193 		error = ufetch_8(SCARG(uap, path), &c);
194 		if (error != 0)
195 			return error;
196 		if (c == '\0') {
197 			fd = SCARG(uap, fd);
198 			if (fd == AT_FDCWD) {
199 				/*
200 				 * operate on current directory
201 				 */
202 				vp = l->l_proc->p_cwdi->cwdi_cdir;
203 				vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
204 				error = vn_stat(vp, &tmpst);
205 				VOP_UNLOCK(vp);
206 			} else {
207 				/*
208 				 * operate on fd
209 				 */
210 				error = do_sys_fstat(fd, &tmpst);
211 			}
212 			if (error != 0)
213 				return error;
214 			goto done;
215 		}
216 	}
217 
218 	if (SCARG(uap, flag) & LINUX_AT_SYMLINK_NOFOLLOW)
219 		nd_flag = NOFOLLOW;
220 	else
221 		nd_flag = FOLLOW;
222 
223 	error = do_sys_statat(l, SCARG(uap, fd), SCARG(uap, path), nd_flag, &tmpst);
224 	if (error != 0)
225 		return error;
226 
227 done:
228 	bsd_to_linux_stat(&tmpst, &tmplst);
229 
230 	return copyout(&tmplst, SCARG(uap, sp), sizeof tmplst);
231 }
232 
233 #ifndef __alpha__
234 int
235 linux_sys_truncate64(struct lwp *l, const struct linux_sys_truncate64_args *uap, register_t *retval)
236 {
237 	/* {
238 		syscallarg(const char *) path;
239 		syscallarg(off_t) length;
240 	} */
241 	struct sys_truncate_args ta;
242 
243 	/* Linux doesn't have the 'pad' pseudo-parameter */
244 	SCARG(&ta, path) = SCARG(uap, path);
245 	SCARG(&ta, PAD) = 0;
246 	SCARG(&ta, length) = SCARG(uap, length);
247 
248 	return sys_truncate(l, &ta, retval);
249 }
250 
251 int
252 linux_sys_ftruncate64(struct lwp *l, const struct linux_sys_ftruncate64_args *uap, register_t *retval)
253 {
254 	/* {
255 		syscallarg(unsigned int) fd;
256 		syscallarg(off_t) length;
257 	} */
258 	struct sys_ftruncate_args ta;
259 
260 	/* Linux doesn't have the 'pad' pseudo-parameter */
261 	SCARG(&ta, fd) = SCARG(uap, fd);
262 	SCARG(&ta, PAD) = 0;
263 	SCARG(&ta, length) = SCARG(uap, length);
264 
265 	return sys_ftruncate(l, &ta, retval);
266 }
267 #endif /* __alpha__ */
268 
269 /*
270  * Linux 'readdir' call. This code is mostly taken from the
271  * SunOS getdents call (see compat/sunos/sunos_misc.c), though
272  * an attempt has been made to keep it a little cleaner.
273  *
274  * The d_off field contains the offset of the next valid entry,
275  * unless the older Linux getdents(2), which used to have it set
276  * to the offset of the entry itself. This function also doesn't
277  * need to deal with the old count == 1 glibc problem.
278  *
279  * Read in BSD-style entries, convert them, and copy them out.
280  *
281  * Note that this doesn't handle union-mounted filesystems.
282  */
283 int
284 linux_sys_getdents64(struct lwp *l, const struct linux_sys_getdents64_args *uap, register_t *retval)
285 {
286 	/* {
287 		syscallarg(int) fd;
288 		syscallarg(struct linux_dirent64 *) dent;
289 		syscallarg(unsigned int) count;
290 	} */
291 	struct dirent *bdp;
292 	struct vnode *vp;
293 	char *inp, *tbuf;		/* BSD-format */
294 	int len, reclen;		/* BSD-format */
295 	char *outp;			/* Linux-format */
296 	int resid, linux_reclen = 0;	/* Linux-format */
297 	file_t *fp;
298 	struct uio auio;
299 	struct iovec aiov;
300 	struct linux_dirent64 idb;
301 	off_t off;		/* true file offset */
302 	int buflen, error, eofflag, nbytes;
303 	struct vattr va;
304 	off_t *cookiebuf = NULL, *cookie;
305 	int ncookies;
306 
307 	/* fd_getvnode() will use the descriptor for us */
308 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
309 		return (error);
310 
311 	if ((fp->f_flag & FREAD) == 0) {
312 		error = EBADF;
313 		goto out1;
314 	}
315 
316 	vp = (struct vnode *)fp->f_data;
317 	if (vp->v_type != VDIR) {
318 		error = ENOTDIR;
319 		goto out1;
320 	}
321 
322 	vn_lock(vp, LK_SHARED | LK_RETRY);
323 	error = VOP_GETATTR(vp, &va, l->l_cred);
324 	VOP_UNLOCK(vp);
325 	if (error)
326 		goto out1;
327 
328 	nbytes = SCARG(uap, count);
329 	buflen = uimin(MAXBSIZE, nbytes);
330 	if (buflen < va.va_blocksize)
331 		buflen = va.va_blocksize;
332 	tbuf = malloc(buflen, M_TEMP, M_WAITOK);
333 
334 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
335 	off = fp->f_offset;
336 again:
337 	aiov.iov_base = tbuf;
338 	aiov.iov_len = buflen;
339 	auio.uio_iov = &aiov;
340 	auio.uio_iovcnt = 1;
341 	auio.uio_rw = UIO_READ;
342 	auio.uio_resid = buflen;
343 	auio.uio_offset = off;
344 	UIO_SETUP_SYSSPACE(&auio);
345 	/*
346          * First we read into the malloc'ed buffer, then
347          * we massage it into user space, one record at a time.
348          */
349 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf,
350 	    &ncookies);
351 	if (error)
352 		goto out;
353 
354 	inp = tbuf;
355 	outp = (void *)SCARG(uap, dent);
356 	resid = nbytes;
357 	if ((len = buflen - auio.uio_resid) == 0)
358 		goto eof;
359 
360 	for (cookie = cookiebuf; len > 0; len -= reclen) {
361 		bdp = (struct dirent *)inp;
362 		reclen = bdp->d_reclen;
363 		if (reclen & 3) {
364 			error = EIO;
365 			goto out;
366 		}
367 		if (bdp->d_fileno == 0) {
368 			inp += reclen;	/* it is a hole; squish it out */
369 			if (cookie)
370 				off = *cookie++;
371 			else
372 				off += reclen;
373 			continue;
374 		}
375 		linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen);
376 		if (reclen > len || resid < linux_reclen) {
377 			/* entry too big for buffer, so just stop */
378 			outp++;
379 			break;
380 		}
381 		if (cookie)
382 			off = *cookie++;	/* each entry points to next */
383 		else
384 			off += reclen;
385 		/*
386 		 * Massage in place to make a Linux-shaped dirent (otherwise
387 		 * we have to worry about touching user memory outside of
388 		 * the copyout() call).
389 		 */
390 		memset(&idb, 0, sizeof(idb));
391 		idb.d_ino = bdp->d_fileno;
392 		idb.d_type = bdp->d_type;
393 		idb.d_off = off;
394 		idb.d_reclen = (u_short)linux_reclen;
395 		memcpy(idb.d_name, bdp->d_name, MIN(sizeof(idb.d_name),
396 		   bdp->d_namlen + 1));
397 		if ((error = copyout((void *)&idb, outp, linux_reclen)))
398 			goto out;
399 		/* advance past this real entry */
400 		inp += reclen;
401 		/* advance output past Linux-shaped entry */
402 		outp += linux_reclen;
403 		resid -= linux_reclen;
404 	}
405 
406 	/* if we squished out the whole block, try again */
407 	if (outp == (void *)SCARG(uap, dent)) {
408 		if (cookiebuf)
409 			free(cookiebuf, M_TEMP);
410 		cookiebuf = NULL;
411 		goto again;
412 	}
413 	fp->f_offset = off;	/* update the vnode offset */
414 
415 eof:
416 	*retval = nbytes - resid;
417 out:
418 	VOP_UNLOCK(vp);
419 	if (cookiebuf)
420 		free(cookiebuf, M_TEMP);
421 	free(tbuf, M_TEMP);
422 out1:
423 	fd_putfile(SCARG(uap, fd));
424 	return error;
425 }
426