xref: /netbsd-src/sys/compat/linux/common/linux_file64.c (revision 7d62b00eb9ad855ffcd7da46b41e23feb5476fac)
1 /*	$NetBSD: linux_file64.c,v 1.67 2021/11/25 02:27:08 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 1.67 2021/11/25 02:27:08 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_syscall.h>
69 #include <compat/linux/linux_syscallargs.h>
70 
71 static void bsd_to_linux_stat64(struct stat *, struct linux_stat64 *);
72 
73 /*
74  * Convert a NetBSD stat structure to a Linux stat structure.
75  * Only the order of the fields and the padding in the structure
76  * is different. linux_fakedev is a machine-dependent function
77  * which optionally converts device driver major/minor numbers
78  * (XXX horrible, but what can you do against code that compares
79  * things against constant major device numbers? sigh)
80  */
81 static void
82 bsd_to_linux_stat64(struct stat *bsp, struct linux_stat64 *lsp)
83 {
84 	memset(lsp, 0, sizeof(*lsp));
85 	lsp->lst_dev     = linux_fakedev(bsp->st_dev, 0);
86 	lsp->lst_ino     = bsp->st_ino;
87 	lsp->lst_mode    = (linux_mode_t)bsp->st_mode;
88 	if (bsp->st_nlink >= (1 << 15))
89 		lsp->lst_nlink = (1 << 15) - 1;
90 	else
91 		lsp->lst_nlink = (linux_nlink_t)bsp->st_nlink;
92 	lsp->lst_uid     = bsp->st_uid;
93 	lsp->lst_gid     = bsp->st_gid;
94 	lsp->lst_rdev    = linux_fakedev(bsp->st_rdev, 1);
95 	lsp->lst_size    = bsp->st_size;
96 	lsp->lst_blksize = bsp->st_blksize;
97 	lsp->lst_blocks  = bsp->st_blocks;
98 	lsp->lst_atime   = bsp->st_atime;
99 	lsp->lst_mtime   = bsp->st_mtime;
100 	lsp->lst_ctime   = bsp->st_ctime;
101 #  ifdef LINUX_STAT64_HAS_NSEC
102 	lsp->lst_atime_nsec   = bsp->st_atimensec;
103 	lsp->lst_mtime_nsec   = bsp->st_mtimensec;
104 	lsp->lst_ctime_nsec   = bsp->st_ctimensec;
105 #  endif
106 #  if LINUX_STAT64_HAS_BROKEN_ST_INO
107 	lsp->__lst_ino   = (linux_ino_t) bsp->st_ino;
108 #  endif
109 }
110 
111 int
112 bsd_to_linux_statx(struct stat *st, struct linux_statx *stx,
113     unsigned int mask)
114 {
115 	if (mask & STATX__RESERVED)
116 		return EINVAL;
117 
118 	/* XXX: STATX_MNT_ID is not supported */
119 	unsigned int rmask = STATX_TYPE | STATX_MODE | STATX_NLINK |
120 	    STATX_UID | STATX_GID | STATX_ATIME | STATX_MTIME | STATX_CTIME |
121 	    STATX_INO | STATX_SIZE | STATX_BLOCKS | STATX_BTIME;
122 
123 	memset(stx, 0, sizeof(*stx));
124 
125 	if ((st->st_flags & UF_NODUMP) != 0)
126 		stx->stx_attributes |= STATX_ATTR_NODUMP;
127 	if ((st->st_flags & (UF_IMMUTABLE|SF_IMMUTABLE)) != 0)
128 		stx->stx_attributes |= STATX_ATTR_IMMUTABLE;
129 	if ((st->st_flags & (UF_APPEND|SF_APPEND)) != 0)
130 		stx->stx_attributes |= STATX_ATTR_APPEND;
131 
132 	stx->stx_attributes_mask =
133 	    STATX_ATTR_NODUMP | STATX_ATTR_IMMUTABLE | STATX_ATTR_APPEND;
134 
135 	stx->stx_blksize = st->st_blksize;
136 
137 	stx->stx_nlink = st->st_nlink;
138 	stx->stx_uid = st->st_uid;
139 	stx->stx_gid = st->st_gid;
140 	stx->stx_mode |= st->st_mode & S_IFMT;
141 	stx->stx_mode |= st->st_mode & ~S_IFMT;
142 	stx->stx_ino = st->st_ino;
143 	stx->stx_size = st->st_size;
144 	stx->stx_blocks = st->st_blocks;
145 
146 	stx->stx_atime.tv_sec = st->st_atime;
147 	stx->stx_atime.tv_nsec = st->st_atimensec;
148 
149 	/* some filesystem has no birthtime returns 0 or -1 */
150 	if ((st->st_birthtime == 0 && st->st_birthtimensec == 0) ||
151 	    (st->st_birthtime == (time_t)-1 &&
152 	    st->st_birthtimensec == (long)-1)) {
153 		rmask &= ~STATX_BTIME;
154 	} else {
155 		stx->stx_btime.tv_sec = st->st_birthtime;
156 		stx->stx_btime.tv_nsec = st->st_birthtimensec;
157 	}
158 
159 	stx->stx_ctime.tv_sec = st->st_ctime;
160 	stx->stx_ctime.tv_nsec = st->st_ctimensec;
161 
162 	stx->stx_mtime.tv_sec = st->st_mtime;
163 	stx->stx_mtime.tv_nsec = st->st_mtimensec;
164 
165 	if (S_ISCHR(st->st_mode) || S_ISBLK(st->st_mode)) {
166 		stx->stx_rdev_major = major(st->st_rdev);
167 		stx->stx_rdev_minor = minor(st->st_rdev);
168 	} else {
169 		stx->stx_dev_major = major(st->st_rdev);
170 		stx->stx_dev_minor = minor(st->st_rdev);
171 	}
172 
173 	stx->stx_mask = rmask;
174 
175 	return 0;
176 }
177 
178 /*
179  * The stat functions below are plain sailing. stat and lstat are handled
180  * by one function to avoid code duplication.
181  */
182 int
183 linux_sys_fstat64(struct lwp *l, const struct linux_sys_fstat64_args *uap, register_t *retval)
184 {
185 	/* {
186 		syscallarg(int) fd;
187 		syscallarg(struct linux_stat64 *) sp;
188 	} */
189 	struct linux_stat64 tmplst;
190 	struct stat tmpst;
191 	int error;
192 
193 	error = do_sys_fstat(SCARG(uap, fd), &tmpst);
194 	if (error != 0)
195 		return error;
196 
197 	bsd_to_linux_stat64(&tmpst, &tmplst);
198 
199 	return copyout(&tmplst, SCARG(uap, sp), sizeof tmplst);
200 }
201 
202 #if !defined(__aarch64__)
203 static int
204 linux_do_stat64(struct lwp *l, const struct linux_sys_stat64_args *uap, register_t *retval, int flags)
205 {
206 	struct linux_stat64 tmplst;
207 	struct stat tmpst;
208 	int error;
209 
210 	error = do_sys_stat(SCARG(uap, path), flags, &tmpst);
211 	if (error != 0)
212 		return error;
213 
214 	bsd_to_linux_stat64(&tmpst, &tmplst);
215 
216 	return copyout(&tmplst, SCARG(uap, sp), sizeof tmplst);
217 }
218 
219 int
220 linux_sys_stat64(struct lwp *l, const struct linux_sys_stat64_args *uap, register_t *retval)
221 {
222 	/* {
223 		syscallarg(const char *) path;
224 		syscallarg(struct linux_stat64 *) sp;
225 	} */
226 
227 	return linux_do_stat64(l, uap, retval, FOLLOW);
228 }
229 
230 int
231 linux_sys_lstat64(struct lwp *l, const struct linux_sys_lstat64_args *uap, register_t *retval)
232 {
233 	/* {
234 		syscallarg(const char *) path;
235 		syscallarg(struct linux_stat64 *) sp;
236 	} */
237 
238 	return linux_do_stat64(l, (const void *)uap, retval, NOFOLLOW);
239 }
240 #endif
241 
242 /*
243  * This is an internal function for the *statat() variant of linux,
244  * which returns struct stat, but flags and other handling are
245  * the same as in linux.
246  */
247 int
248 linux_statat(struct lwp *l, int fd, const char *path, int lflag,
249     struct stat *st)
250 {
251 	struct vnode *vp;
252 	int error, nd_flag;
253 	uint8_t c;
254 
255 	if (lflag & LINUX_AT_EMPTY_PATH) {
256 		/*
257 		 * If path is null string:
258 		 */
259 		error = ufetch_8(path, &c);
260 		if (error != 0)
261 			return error;
262 		if (c == '\0') {
263 			if (fd == LINUX_AT_FDCWD) {
264 				/*
265 				 * operate on current directory
266 				 */
267 				vp = l->l_proc->p_cwdi->cwdi_cdir;
268 				vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
269 				error = vn_stat(vp, st);
270 				VOP_UNLOCK(vp);
271 			} else {
272 				/*
273 				 * operate on fd
274 				 */
275 				error = do_sys_fstat(fd, st);
276 			}
277 			return error;
278 		}
279 	}
280 
281 	if (lflag & LINUX_AT_SYMLINK_NOFOLLOW)
282 		nd_flag = NOFOLLOW;
283 	else
284 		nd_flag = FOLLOW;
285 
286 	return do_sys_statat(l, fd, path, nd_flag, st);
287 }
288 
289 int
290 linux_sys_fstatat64(struct lwp *l, const struct linux_sys_fstatat64_args *uap, register_t *retval)
291 {
292 	/* {
293 		syscallarg(int) fd;
294 		syscallarg(const char *) path;
295 		syscallarg(struct linux_stat64 *) sp;
296 		syscallarg(int) flag;
297 	} */
298 	struct linux_stat64 tmplst;
299 	struct stat tmpst;
300 	int error;
301 
302 	error = linux_statat(l, SCARG(uap, fd), SCARG(uap, path),
303 	    SCARG(uap, flag), &tmpst);
304 	if (error != 0)
305 		return error;
306 
307 	bsd_to_linux_stat64(&tmpst, &tmplst);
308 
309 	return copyout(&tmplst, SCARG(uap, sp), sizeof tmplst);
310 }
311 
312 #ifdef LINUX_SYS_statx
313 int
314 linux_sys_statx(struct lwp *l, const struct linux_sys_statx_args *uap,
315     register_t *retval)
316 {
317 	/* {
318 		syscallarg(int) fd;
319 		syscallarg(const char *) path;
320 		syscallarg(int) flag;
321 		syscallarg(unsigned int) mask;
322 		syscallarg(struct linux_statx *) sp;
323 	} */
324 	struct linux_statx stx;
325 	struct stat st;
326 	int error;
327 
328 	error = linux_statat(l, SCARG(uap, fd), SCARG(uap, path),
329 	    SCARG(uap, flag), &st);
330 	if (error != 0)
331 		return error;
332 
333 	error = bsd_to_linux_statx(&st, &stx, SCARG(uap, mask));
334 	if (error != 0)
335 		return error;
336 
337 	return copyout(&stx, SCARG(uap, sp), sizeof stx);
338 }
339 #endif /* LINUX_SYS_statx */
340 
341 #ifndef __alpha__
342 int
343 linux_sys_truncate64(struct lwp *l, const struct linux_sys_truncate64_args *uap, register_t *retval)
344 {
345 	/* {
346 		syscallarg(const char *) path;
347 		syscallarg(off_t) length;
348 	} */
349 	struct sys_truncate_args ta;
350 
351 	/* Linux doesn't have the 'pad' pseudo-parameter */
352 	SCARG(&ta, path) = SCARG(uap, path);
353 	SCARG(&ta, PAD) = 0;
354 	SCARG(&ta, length) = SCARG(uap, length);
355 
356 	return sys_truncate(l, &ta, retval);
357 }
358 
359 int
360 linux_sys_ftruncate64(struct lwp *l, const struct linux_sys_ftruncate64_args *uap, register_t *retval)
361 {
362 	/* {
363 		syscallarg(unsigned int) fd;
364 		syscallarg(off_t) length;
365 	} */
366 	struct sys_ftruncate_args ta;
367 
368 	/* Linux doesn't have the 'pad' pseudo-parameter */
369 	SCARG(&ta, fd) = SCARG(uap, fd);
370 	SCARG(&ta, PAD) = 0;
371 	SCARG(&ta, length) = SCARG(uap, length);
372 
373 	return sys_ftruncate(l, &ta, retval);
374 }
375 #endif /* __alpha__ */
376 
377 /*
378  * Linux 'readdir' call. This code is mostly taken from the
379  * SunOS getdents call (see compat/sunos/sunos_misc.c), though
380  * an attempt has been made to keep it a little cleaner.
381  *
382  * The d_off field contains the offset of the next valid entry,
383  * unless the older Linux getdents(2), which used to have it set
384  * to the offset of the entry itself. This function also doesn't
385  * need to deal with the old count == 1 glibc problem.
386  *
387  * Read in BSD-style entries, convert them, and copy them out.
388  *
389  * Note that this doesn't handle union-mounted filesystems.
390  */
391 int
392 linux_sys_getdents64(struct lwp *l, const struct linux_sys_getdents64_args *uap, register_t *retval)
393 {
394 	/* {
395 		syscallarg(int) fd;
396 		syscallarg(struct linux_dirent64 *) dent;
397 		syscallarg(unsigned int) count;
398 	} */
399 	struct dirent *bdp;
400 	struct vnode *vp;
401 	char *inp, *tbuf;		/* BSD-format */
402 	int len, reclen;		/* BSD-format */
403 	char *outp;			/* Linux-format */
404 	int resid, linux_reclen = 0;	/* Linux-format */
405 	file_t *fp;
406 	struct uio auio;
407 	struct iovec aiov;
408 	struct linux_dirent64 idb;
409 	off_t off;		/* true file offset */
410 	int buflen, error, eofflag, nbytes;
411 	struct vattr va;
412 	off_t *cookiebuf = NULL, *cookie;
413 	int ncookies;
414 
415 	/* fd_getvnode() will use the descriptor for us */
416 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
417 		return (error);
418 
419 	if ((fp->f_flag & FREAD) == 0) {
420 		error = EBADF;
421 		goto out1;
422 	}
423 
424 	vp = (struct vnode *)fp->f_data;
425 	if (vp->v_type != VDIR) {
426 		error = ENOTDIR;
427 		goto out1;
428 	}
429 
430 	vn_lock(vp, LK_SHARED | LK_RETRY);
431 	error = VOP_GETATTR(vp, &va, l->l_cred);
432 	VOP_UNLOCK(vp);
433 	if (error)
434 		goto out1;
435 
436 	nbytes = SCARG(uap, count);
437 	buflen = uimin(MAXBSIZE, nbytes);
438 	if (buflen < va.va_blocksize)
439 		buflen = va.va_blocksize;
440 	tbuf = malloc(buflen, M_TEMP, M_WAITOK);
441 
442 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
443 	off = fp->f_offset;
444 again:
445 	aiov.iov_base = tbuf;
446 	aiov.iov_len = buflen;
447 	auio.uio_iov = &aiov;
448 	auio.uio_iovcnt = 1;
449 	auio.uio_rw = UIO_READ;
450 	auio.uio_resid = buflen;
451 	auio.uio_offset = off;
452 	UIO_SETUP_SYSSPACE(&auio);
453 	/*
454          * First we read into the malloc'ed buffer, then
455          * we massage it into user space, one record at a time.
456          */
457 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf,
458 	    &ncookies);
459 	if (error)
460 		goto out;
461 
462 	inp = tbuf;
463 	outp = (void *)SCARG(uap, dent);
464 	resid = nbytes;
465 	if ((len = buflen - auio.uio_resid) == 0)
466 		goto eof;
467 
468 	for (cookie = cookiebuf; len > 0; len -= reclen) {
469 		bdp = (struct dirent *)inp;
470 		reclen = bdp->d_reclen;
471 		if (reclen & 3) {
472 			error = EIO;
473 			goto out;
474 		}
475 		if (bdp->d_fileno == 0) {
476 			inp += reclen;	/* it is a hole; squish it out */
477 			if (cookie)
478 				off = *cookie++;
479 			else
480 				off += reclen;
481 			continue;
482 		}
483 		linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen);
484 		if (reclen > len || resid < linux_reclen) {
485 			/* entry too big for buffer, so just stop */
486 			outp++;
487 			break;
488 		}
489 		if (cookie)
490 			off = *cookie++;	/* each entry points to next */
491 		else
492 			off += reclen;
493 		/*
494 		 * Massage in place to make a Linux-shaped dirent (otherwise
495 		 * we have to worry about touching user memory outside of
496 		 * the copyout() call).
497 		 */
498 		memset(&idb, 0, sizeof(idb));
499 		idb.d_ino = bdp->d_fileno;
500 		idb.d_type = bdp->d_type;
501 		idb.d_off = off;
502 		idb.d_reclen = (u_short)linux_reclen;
503 		memcpy(idb.d_name, bdp->d_name, MIN(sizeof(idb.d_name),
504 		   bdp->d_namlen + 1));
505 		if ((error = copyout((void *)&idb, outp, linux_reclen)))
506 			goto out;
507 		/* advance past this real entry */
508 		inp += reclen;
509 		/* advance output past Linux-shaped entry */
510 		outp += linux_reclen;
511 		resid -= linux_reclen;
512 	}
513 
514 	/* if we squished out the whole block, try again */
515 	if (outp == (void *)SCARG(uap, dent)) {
516 		if (cookiebuf)
517 			free(cookiebuf, M_TEMP);
518 		cookiebuf = NULL;
519 		goto again;
520 	}
521 	fp->f_offset = off;	/* update the vnode offset */
522 
523 eof:
524 	*retval = nbytes - resid;
525 out:
526 	VOP_UNLOCK(vp);
527 	if (cookiebuf)
528 		free(cookiebuf, M_TEMP);
529 	free(tbuf, M_TEMP);
530 out1:
531 	fd_putfile(SCARG(uap, fd));
532 	return error;
533 }
534