xref: /linux/fs/ntfs3/dir.c (revision 42422993cf28d456778ee9168d73758ec037cd51)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  *
4  * Copyright (C) 2019-2021 Paragon Software GmbH, All rights reserved.
5  *
6  *  Directory handling functions for NTFS-based filesystems.
7  *
8  */
9 
10 #include <linux/fs.h>
11 #include <linux/nls.h>
12 
13 #include "debug.h"
14 #include "ntfs.h"
15 #include "ntfs_fs.h"
16 
17 /* Convert little endian UTF-16 to NLS string. */
18 int ntfs_utf16_to_nls(struct ntfs_sb_info *sbi, const __le16 *name, u32 len,
19 		      u8 *buf, int buf_len)
20 {
21 	int ret, warn;
22 	u8 *op;
23 	struct nls_table *nls = sbi->options->nls;
24 
25 	static_assert(sizeof(wchar_t) == sizeof(__le16));
26 
27 	if (!nls) {
28 		/* UTF-16 -> UTF-8 */
29 		ret = utf16s_to_utf8s((wchar_t *)name, len, UTF16_LITTLE_ENDIAN,
30 				      buf, buf_len);
31 		buf[ret] = '\0';
32 		return ret;
33 	}
34 
35 	op = buf;
36 	warn = 0;
37 
38 	while (len--) {
39 		u16 ec;
40 		int charlen;
41 		char dump[5];
42 
43 		if (buf_len < NLS_MAX_CHARSET_SIZE) {
44 			ntfs_warn(sbi->sb,
45 				  "filename was truncated while converting.");
46 			break;
47 		}
48 
49 		ec = le16_to_cpu(*name++);
50 		charlen = nls->uni2char(ec, op, buf_len);
51 
52 		if (charlen > 0) {
53 			op += charlen;
54 			buf_len -= charlen;
55 			continue;
56 		}
57 
58 		*op++ = '_';
59 		buf_len -= 1;
60 		if (warn)
61 			continue;
62 
63 		warn = 1;
64 		hex_byte_pack(&dump[0], ec >> 8);
65 		hex_byte_pack(&dump[2], ec);
66 		dump[4] = 0;
67 
68 		ntfs_err(sbi->sb, "failed to convert \"%s\" to %s", dump,
69 			 nls->charset);
70 	}
71 
72 	*op = '\0';
73 	return op - buf;
74 }
75 
76 // clang-format off
77 #define PLANE_SIZE	0x00010000
78 
79 #define SURROGATE_PAIR	0x0000d800
80 #define SURROGATE_LOW	0x00000400
81 #define SURROGATE_BITS	0x000003ff
82 // clang-format on
83 
84 /*
85  * put_utf16 - Modified version of put_utf16 from fs/nls/nls_base.c
86  *
87  * Function is sparse warnings free.
88  */
89 static inline void put_utf16(wchar_t *s, unsigned int c,
90 			     enum utf16_endian endian)
91 {
92 	static_assert(sizeof(wchar_t) == sizeof(__le16));
93 	static_assert(sizeof(wchar_t) == sizeof(__be16));
94 
95 	switch (endian) {
96 	default:
97 		*s = (wchar_t)c;
98 		break;
99 	case UTF16_LITTLE_ENDIAN:
100 		*(__le16 *)s = __cpu_to_le16(c);
101 		break;
102 	case UTF16_BIG_ENDIAN:
103 		*(__be16 *)s = __cpu_to_be16(c);
104 		break;
105 	}
106 }
107 
108 /*
109  * _utf8s_to_utf16s
110  *
111  * Modified version of 'utf8s_to_utf16s' allows to
112  * detect -ENAMETOOLONG without writing out of expected maximum.
113  */
114 static int _utf8s_to_utf16s(const u8 *s, int inlen, enum utf16_endian endian,
115 			    wchar_t *pwcs, int maxout)
116 {
117 	u16 *op;
118 	int size;
119 	unicode_t u;
120 
121 	op = pwcs;
122 	while (inlen > 0 && *s) {
123 		if (*s & 0x80) {
124 			size = utf8_to_utf32(s, inlen, &u);
125 			if (size < 0)
126 				return -EINVAL;
127 			s += size;
128 			inlen -= size;
129 
130 			if (u >= PLANE_SIZE) {
131 				if (maxout < 2)
132 					return -ENAMETOOLONG;
133 
134 				u -= PLANE_SIZE;
135 				put_utf16(op++,
136 					  SURROGATE_PAIR |
137 						  ((u >> 10) & SURROGATE_BITS),
138 					  endian);
139 				put_utf16(op++,
140 					  SURROGATE_PAIR | SURROGATE_LOW |
141 						  (u & SURROGATE_BITS),
142 					  endian);
143 				maxout -= 2;
144 			} else {
145 				if (maxout < 1)
146 					return -ENAMETOOLONG;
147 
148 				put_utf16(op++, u, endian);
149 				maxout--;
150 			}
151 		} else {
152 			if (maxout < 1)
153 				return -ENAMETOOLONG;
154 
155 			put_utf16(op++, *s++, endian);
156 			inlen--;
157 			maxout--;
158 		}
159 	}
160 	return op - pwcs;
161 }
162 
163 /*
164  * ntfs_nls_to_utf16 - Convert input string to UTF-16.
165  * @name:	Input name.
166  * @name_len:	Input name length.
167  * @uni:	Destination memory.
168  * @max_ulen:	Destination memory.
169  * @endian:	Endian of target UTF-16 string.
170  *
171  * This function is called:
172  * - to create NTFS name
173  * - to create symlink
174  *
175  * Return: UTF-16 string length or error (if negative).
176  */
177 int ntfs_nls_to_utf16(struct ntfs_sb_info *sbi, const u8 *name, u32 name_len,
178 		      struct cpu_str *uni, u32 max_ulen,
179 		      enum utf16_endian endian)
180 {
181 	int ret, slen;
182 	const u8 *end;
183 	struct nls_table *nls = sbi->options->nls;
184 	u16 *uname = uni->name;
185 
186 	static_assert(sizeof(wchar_t) == sizeof(u16));
187 
188 	if (!nls) {
189 		/* utf8 -> utf16 */
190 		ret = _utf8s_to_utf16s(name, name_len, endian, uname, max_ulen);
191 		uni->len = ret;
192 		return ret;
193 	}
194 
195 	for (ret = 0, end = name + name_len; name < end; ret++, name += slen) {
196 		if (ret >= max_ulen)
197 			return -ENAMETOOLONG;
198 
199 		slen = nls->char2uni(name, end - name, uname + ret);
200 		if (!slen)
201 			return -EINVAL;
202 		if (slen < 0)
203 			return slen;
204 	}
205 
206 #ifdef __BIG_ENDIAN
207 	if (endian == UTF16_LITTLE_ENDIAN) {
208 		int i = ret;
209 
210 		while (i--) {
211 			__cpu_to_le16s(uname);
212 			uname++;
213 		}
214 	}
215 #else
216 	if (endian == UTF16_BIG_ENDIAN) {
217 		int i = ret;
218 
219 		while (i--) {
220 			__cpu_to_be16s(uname);
221 			uname++;
222 		}
223 	}
224 #endif
225 
226 	uni->len = ret;
227 	return ret;
228 }
229 
230 /*
231  * dir_search_u - Helper function.
232  */
233 struct inode *dir_search_u(struct inode *dir, const struct cpu_str *uni,
234 			   struct ntfs_fnd *fnd)
235 {
236 	int err = 0;
237 	struct super_block *sb = dir->i_sb;
238 	struct ntfs_sb_info *sbi = sb->s_fs_info;
239 	struct ntfs_inode *ni = ntfs_i(dir);
240 	struct NTFS_DE *e;
241 	int diff;
242 	struct inode *inode = NULL;
243 	struct ntfs_fnd *fnd_a = NULL;
244 
245 	if (!fnd) {
246 		fnd_a = fnd_get();
247 		if (!fnd_a) {
248 			err = -ENOMEM;
249 			goto out;
250 		}
251 		fnd = fnd_a;
252 	}
253 
254 	err = indx_find(&ni->dir, ni, NULL, uni, 0, sbi, &diff, &e, fnd);
255 
256 	if (err)
257 		goto out;
258 
259 	if (diff) {
260 		err = -ENOENT;
261 		goto out;
262 	}
263 
264 	inode = ntfs_iget5(sb, &e->ref, uni);
265 	if (!IS_ERR(inode) && is_bad_inode(inode)) {
266 		iput(inode);
267 		err = -EINVAL;
268 	}
269 out:
270 	fnd_put(fnd_a);
271 
272 	return err == -ENOENT ? NULL : err ? ERR_PTR(err) : inode;
273 }
274 
275 static inline int ntfs_filldir(struct ntfs_sb_info *sbi, struct ntfs_inode *ni,
276 			       const struct NTFS_DE *e, u8 *name,
277 			       struct dir_context *ctx)
278 {
279 	const struct ATTR_FILE_NAME *fname;
280 	unsigned long ino;
281 	int name_len;
282 	u32 dt_type;
283 
284 	fname = Add2Ptr(e, sizeof(struct NTFS_DE));
285 
286 	if (fname->type == FILE_NAME_DOS)
287 		return 0;
288 
289 	if (!mi_is_ref(&ni->mi, &fname->home))
290 		return 0;
291 
292 	ino = ino_get(&e->ref);
293 
294 	if (ino == MFT_REC_ROOT)
295 		return 0;
296 
297 	/* Skip meta files. Unless option to show metafiles is set. */
298 	if (!sbi->options->showmeta && ntfs_is_meta_file(sbi, ino))
299 		return 0;
300 
301 	if (sbi->options->nohidden && (fname->dup.fa & FILE_ATTRIBUTE_HIDDEN))
302 		return 0;
303 
304 	name_len = ntfs_utf16_to_nls(sbi, fname->name, fname->name_len, name,
305 				     PATH_MAX);
306 	if (name_len <= 0) {
307 		ntfs_warn(sbi->sb, "failed to convert name for inode %lx.",
308 			  ino);
309 		return 0;
310 	}
311 
312 	/* NTFS: symlinks are "dir + reparse" or "file + reparse" */
313 	if (fname->dup.fa & FILE_ATTRIBUTE_REPARSE_POINT)
314 		dt_type = DT_LNK;
315 	else
316 		dt_type = (fname->dup.fa & FILE_ATTRIBUTE_DIRECTORY) ? DT_DIR : DT_REG;
317 
318 	return !dir_emit(ctx, (s8 *)name, name_len, ino, dt_type);
319 }
320 
321 /*
322  * ntfs_read_hdr - Helper function for ntfs_readdir().
323  */
324 static int ntfs_read_hdr(struct ntfs_sb_info *sbi, struct ntfs_inode *ni,
325 			 const struct INDEX_HDR *hdr, u64 vbo, u64 pos,
326 			 u8 *name, struct dir_context *ctx)
327 {
328 	int err;
329 	const struct NTFS_DE *e;
330 	u32 e_size;
331 	u32 end = le32_to_cpu(hdr->used);
332 	u32 off = le32_to_cpu(hdr->de_off);
333 
334 	for (;; off += e_size) {
335 		if (off + sizeof(struct NTFS_DE) > end)
336 			return -1;
337 
338 		e = Add2Ptr(hdr, off);
339 		e_size = le16_to_cpu(e->size);
340 		if (e_size < sizeof(struct NTFS_DE) || off + e_size > end)
341 			return -1;
342 
343 		if (de_is_last(e))
344 			return 0;
345 
346 		/* Skip already enumerated. */
347 		if (vbo + off < pos)
348 			continue;
349 
350 		if (le16_to_cpu(e->key_size) < SIZEOF_ATTRIBUTE_FILENAME)
351 			return -1;
352 
353 		ctx->pos = vbo + off;
354 
355 		/* Submit the name to the filldir callback. */
356 		err = ntfs_filldir(sbi, ni, e, name, ctx);
357 		if (err)
358 			return err;
359 	}
360 }
361 
362 /*
363  * ntfs_readdir - file_operations::iterate_shared
364  *
365  * Use non sorted enumeration.
366  * We have an example of broken volume where sorted enumeration
367  * counts each name twice.
368  */
369 static int ntfs_readdir(struct file *file, struct dir_context *ctx)
370 {
371 	const struct INDEX_ROOT *root;
372 	u64 vbo;
373 	size_t bit;
374 	loff_t eod;
375 	int err = 0;
376 	struct inode *dir = file_inode(file);
377 	struct ntfs_inode *ni = ntfs_i(dir);
378 	struct super_block *sb = dir->i_sb;
379 	struct ntfs_sb_info *sbi = sb->s_fs_info;
380 	loff_t i_size = i_size_read(dir);
381 	u32 pos = ctx->pos;
382 	u8 *name = NULL;
383 	struct indx_node *node = NULL;
384 	u8 index_bits = ni->dir.index_bits;
385 
386 	/* Name is a buffer of PATH_MAX length. */
387 	static_assert(NTFS_NAME_LEN * 4 < PATH_MAX);
388 
389 	eod = i_size + sbi->record_size;
390 
391 	if (pos >= eod)
392 		return 0;
393 
394 	if (!dir_emit_dots(file, ctx))
395 		return 0;
396 
397 	/* Allocate PATH_MAX bytes. */
398 	name = __getname();
399 	if (!name)
400 		return -ENOMEM;
401 
402 	if (!ni->mi_loaded && ni->attr_list.size) {
403 		/*
404 		 * Directory inode is locked for read.
405 		 * Load all subrecords to avoid 'write' access to 'ni' during
406 		 * directory reading.
407 		 */
408 		ni_lock(ni);
409 		if (!ni->mi_loaded && ni->attr_list.size) {
410 			err = ni_load_all_mi(ni);
411 			if (!err)
412 				ni->mi_loaded = true;
413 		}
414 		ni_unlock(ni);
415 		if (err)
416 			goto out;
417 	}
418 
419 	root = indx_get_root(&ni->dir, ni, NULL, NULL);
420 	if (!root) {
421 		err = -EINVAL;
422 		goto out;
423 	}
424 
425 	if (pos >= sbi->record_size) {
426 		bit = (pos - sbi->record_size) >> index_bits;
427 	} else {
428 		err = ntfs_read_hdr(sbi, ni, &root->ihdr, 0, pos, name, ctx);
429 		if (err)
430 			goto out;
431 		bit = 0;
432 	}
433 
434 	if (!i_size) {
435 		ctx->pos = eod;
436 		goto out;
437 	}
438 
439 	for (;;) {
440 		vbo = (u64)bit << index_bits;
441 		if (vbo >= i_size) {
442 			ctx->pos = eod;
443 			goto out;
444 		}
445 
446 		err = indx_used_bit(&ni->dir, ni, &bit);
447 		if (err)
448 			goto out;
449 
450 		if (bit == MINUS_ONE_T) {
451 			ctx->pos = eod;
452 			goto out;
453 		}
454 
455 		vbo = (u64)bit << index_bits;
456 		if (vbo >= i_size) {
457 			ntfs_inode_err(dir, "Looks like your dir is corrupt");
458 			err = -EINVAL;
459 			goto out;
460 		}
461 
462 		err = indx_read(&ni->dir, ni, bit << ni->dir.idx2vbn_bits,
463 				&node);
464 		if (err)
465 			goto out;
466 
467 		err = ntfs_read_hdr(sbi, ni, &node->index->ihdr,
468 				    vbo + sbi->record_size, pos, name, ctx);
469 		if (err)
470 			goto out;
471 
472 		bit += 1;
473 	}
474 
475 out:
476 
477 	__putname(name);
478 	put_indx_node(node);
479 
480 	if (err == -ENOENT) {
481 		err = 0;
482 		ctx->pos = pos;
483 	}
484 
485 	return err;
486 }
487 
488 static int ntfs_dir_count(struct inode *dir, bool *is_empty, size_t *dirs,
489 			  size_t *files)
490 {
491 	int err = 0;
492 	struct ntfs_inode *ni = ntfs_i(dir);
493 	struct NTFS_DE *e = NULL;
494 	struct INDEX_ROOT *root;
495 	struct INDEX_HDR *hdr;
496 	const struct ATTR_FILE_NAME *fname;
497 	u32 e_size, off, end;
498 	u64 vbo = 0;
499 	size_t drs = 0, fles = 0, bit = 0;
500 	loff_t i_size = ni->vfs_inode.i_size;
501 	struct indx_node *node = NULL;
502 	u8 index_bits = ni->dir.index_bits;
503 
504 	if (is_empty)
505 		*is_empty = true;
506 
507 	root = indx_get_root(&ni->dir, ni, NULL, NULL);
508 	if (!root)
509 		return -EINVAL;
510 
511 	hdr = &root->ihdr;
512 
513 	for (;;) {
514 		end = le32_to_cpu(hdr->used);
515 		off = le32_to_cpu(hdr->de_off);
516 
517 		for (; off + sizeof(struct NTFS_DE) <= end; off += e_size) {
518 			e = Add2Ptr(hdr, off);
519 			e_size = le16_to_cpu(e->size);
520 			if (e_size < sizeof(struct NTFS_DE) ||
521 			    off + e_size > end)
522 				break;
523 
524 			if (de_is_last(e))
525 				break;
526 
527 			fname = de_get_fname(e);
528 			if (!fname)
529 				continue;
530 
531 			if (fname->type == FILE_NAME_DOS)
532 				continue;
533 
534 			if (is_empty) {
535 				*is_empty = false;
536 				if (!dirs && !files)
537 					goto out;
538 			}
539 
540 			if (fname->dup.fa & FILE_ATTRIBUTE_DIRECTORY)
541 				drs += 1;
542 			else
543 				fles += 1;
544 		}
545 
546 		if (vbo >= i_size)
547 			goto out;
548 
549 		err = indx_used_bit(&ni->dir, ni, &bit);
550 		if (err)
551 			goto out;
552 
553 		if (bit == MINUS_ONE_T)
554 			goto out;
555 
556 		vbo = (u64)bit << index_bits;
557 		if (vbo >= i_size)
558 			goto out;
559 
560 		err = indx_read(&ni->dir, ni, bit << ni->dir.idx2vbn_bits,
561 				&node);
562 		if (err)
563 			goto out;
564 
565 		hdr = &node->index->ihdr;
566 		bit += 1;
567 		vbo = (u64)bit << ni->dir.idx2vbn_bits;
568 	}
569 
570 out:
571 	put_indx_node(node);
572 	if (dirs)
573 		*dirs = drs;
574 	if (files)
575 		*files = fles;
576 
577 	return err;
578 }
579 
580 bool dir_is_empty(struct inode *dir)
581 {
582 	bool is_empty = false;
583 
584 	ntfs_dir_count(dir, &is_empty, NULL, NULL);
585 
586 	return is_empty;
587 }
588 
589 // clang-format off
590 const struct file_operations ntfs_dir_operations = {
591 	.llseek		= generic_file_llseek,
592 	.read		= generic_read_dir,
593 	.iterate_shared	= ntfs_readdir,
594 	.fsync		= generic_file_fsync,
595 	.open		= ntfs_file_open,
596 };
597 // clang-format on
598